mirror of https://github.com/vladmandic/human
8057 lines
1.6 MiB
8057 lines
1.6 MiB
/*
|
|
Human
|
|
homepage: <https://github.com/vladmandic/human>
|
|
author: <https://github.com/vladmandic>'
|
|
*/
|
|
|
|
var Human=(()=>{var jc=Object.defineProperty;var jE=Object.getOwnPropertyDescriptor;var HE=Object.getOwnPropertyNames;var qE=Object.prototype.hasOwnProperty;var XE=(e,t,r)=>t in e?jc(e,t,{enumerable:!0,configurable:!0,writable:!0,value:r}):e[t]=r;var vs=(e,t)=>{for(var r in t)jc(e,r,{get:t[r],enumerable:!0})},KE=(e,t,r,n)=>{if(t&&typeof t=="object"||typeof t=="function")for(let a of HE(t))!qE.call(e,a)&&a!==r&&jc(e,a,{get:()=>t[a],enumerable:!(n=jE(t,a))||n.enumerable});return e};var ZE=e=>KE(jc({},"__esModule",{value:!0}),e);var fe=(e,t,r)=>(XE(e,typeof t!="symbol"?t+"":t,r),r),_3=(e,t,r)=>{if(!t.has(e))throw TypeError("Cannot "+r)};var fp=(e,t,r)=>(_3(e,t,"read from private field"),r?r.call(e):t.get(e)),mp=(e,t,r)=>{if(t.has(e))throw TypeError("Cannot add the same private member more than once");t instanceof WeakSet?t.add(e):t.set(e,r)},gp=(e,t,r,n)=>(_3(e,t,"write to private field"),n?n.call(e,r):t.set(e,r),r);var Exe={};vs(Exe,{Human:()=>c3,default:()=>c3,defaults:()=>ws,draw:()=>i3,env:()=>he,match:()=>h3,models:()=>y1});function se(...e){let t=new Date,r=`${t.getHours().toString().padStart(2,"0")}:${t.getMinutes().toString().padStart(2,"0")}:${t.getSeconds().toString().padStart(2,"0")}.${t.getMilliseconds().toString().padStart(3,"0")}`;e&&console.log(r,"Human:",...e)}function z3(e,t){let r=e.endsWith("/")?"":"/",a=t.startsWith(".")||t.startsWith("/")||t.startsWith("http:")||t.startsWith("https:")||t.startsWith("file:")?`${t}`:`${e}${r}${t}`;if(!a.toLocaleLowerCase().includes(".json"))throw new Error(`modelpath error: expecting json file: ${a}`);return a}var oe=()=>typeof performance!="undefined"?performance.now():parseInt((Number(process.hrtime.bigint())/1e3/1e3).toString());function Q2(e,t,r="config",n=[]){for(let a of Object.keys(t))if(typeof t[a]=="object")Q2(e[a],t[a],a,n);else{let s=e&&typeof e[a]!="undefined";s||n.push({reason:"unknown property",where:`${r}.${a} = ${t[a]}`});let i=e&&typeof e[a]==typeof t[a];s&&!i&&n.push({reason:"property type mismatch",where:`${r}.${a} = ${t[a]}`,expected:typeof e[a]})}return t.debug&&r==="config"&&n.length>0&&se("invalid configuration",n),n}function Gt(...e){let t=r=>r&&typeof r=="object";return e.reduce((r,n)=>(Object.keys(n||{}).forEach(a=>{let s=r[a],i=n[a];Array.isArray(s)&&Array.isArray(i)?r[a]=s.concat(...i):t(s)&&t(i)?r[a]=Gt(s,i):r[a]=i}),r),{})}var ws={backend:"",modelBasePath:"",cacheModels:!0,wasmPath:"",wasmPlatformFetch:!1,debug:!0,async:!0,warmup:"full",cacheSensitivity:.7,skipAllowed:!1,deallocate:!1,filter:{enabled:!0,equalization:!1,width:0,height:0,flip:!1,return:!0,brightness:0,contrast:0,sharpness:0,blur:0,saturation:0,hue:0,negative:!1,sepia:!1,vintage:!1,kodachrome:!1,technicolor:!1,polaroid:!1,pixelate:0},gesture:{enabled:!0},face:{enabled:!0,detector:{modelPath:"blazeface.json",rotation:!0,maxDetected:1,skipFrames:99,skipTime:2500,minConfidence:.2,iouThreshold:.1,mask:!1,return:!1},mesh:{enabled:!0,modelPath:"facemesh.json"},attention:{enabled:!1,modelPath:"facemesh-attention.json"},iris:{enabled:!0,modelPath:"iris.json"},emotion:{enabled:!0,minConfidence:.1,skipFrames:99,skipTime:1500,modelPath:"emotion.json"},description:{enabled:!0,modelPath:"faceres.json",skipFrames:99,skipTime:3e3,minConfidence:.1},antispoof:{enabled:!1,skipFrames:99,skipTime:4e3,modelPath:"antispoof.json"},liveness:{enabled:!1,skipFrames:99,skipTime:4e3,modelPath:"liveness.json"}},body:{enabled:!0,modelPath:"movenet-lightning.json",maxDetected:-1,minConfidence:.3,skipFrames:1,skipTime:200},hand:{enabled:!0,rotation:!0,skipFrames:99,skipTime:1e3,minConfidence:.5,iouThreshold:.2,maxDetected:-1,landmarks:!0,detector:{modelPath:"handtrack.json"},skeleton:{modelPath:"handlandmark-full.json"}},object:{enabled:!1,modelPath:"mb3-centernet.json",minConfidence:.2,iouThreshold:.4,maxDetected:10,skipFrames:99,skipTime:2e3},segmentation:{enabled:!1,modelPath:"selfie.json",blur:8}};var Ue={};vs(Ue,{Abs:()=>Ho,Acos:()=>Du,Acosh:()=>Lu,AdadeltaOptimizer:()=>qf,AdagradOptimizer:()=>Xf,AdamOptimizer:()=>Kf,AdamaxOptimizer:()=>Zf,Add:()=>Ja,AddN:()=>Zs,All:()=>Bu,Any:()=>Wu,ArgMax:()=>Ys,ArgMin:()=>Vu,Asin:()=>Uu,Asinh:()=>Gu,Atan:()=>ju,Atan2:()=>qu,Atanh:()=>Hu,AvgPool:()=>Js,AvgPool3D:()=>Yp,AvgPool3DGrad:()=>Q0,AvgPoolGrad:()=>J0,BackendWasm:()=>mC,BatchMatMul:()=>Qs,BatchToSpaceND:()=>qo,Bincount:()=>ef,BroadcastArgs:()=>tf,BroadcastTo:()=>Dv,Callback:()=>fk,CallbackList:()=>f6,Cast:()=>ei,Ceil:()=>ti,ClipByValue:()=>Qa,Complex:()=>Jp,ComplexAbs:()=>Qp,Concat:()=>Xo,Conv2D:()=>ri,Conv2DBackpropFilter:()=>rf,Conv2DBackpropInput:()=>ni,Conv3D:()=>eh,Conv3DBackpropFilterV2:()=>nf,Conv3DBackpropInputV2:()=>af,Cos:()=>ai,Cosh:()=>si,CropAndResize:()=>Zo,Cumprod:()=>Ko,Cumsum:()=>ii,CustomCallback:()=>g6,DataStorage:()=>Zp,DenseBincount:()=>sf,DepthToSpace:()=>Yo,DepthwiseConv2dNative:()=>oi,DepthwiseConv2dNativeBackpropFilter:()=>of,DepthwiseConv2dNativeBackpropInput:()=>lf,Diag:()=>uf,Dilation2D:()=>th,Dilation2DBackpropFilter:()=>b0,Dilation2DBackpropInput:()=>x0,ENV:()=>ba,EarlyStopping:()=>mk,Einsum:()=>rh,Elu:()=>ui,EluGrad:()=>df,Environment:()=>zv,Equal:()=>Jo,Erf:()=>Xu,Exp:()=>di,ExpandDims:()=>Qo,Expm1:()=>el,FFT:()=>pf,Fill:()=>Ku,FlipLeftRight:()=>tl,Floor:()=>pi,FloorDiv:()=>hi,FromPixels:()=>Op,FusedBatchNorm:()=>ci,FusedConv2D:()=>_s,FusedDepthwiseConv2D:()=>zs,GPGPUContext:()=>vu,GatherNd:()=>nl,GatherV2:()=>rl,GraphModel:()=>vm,Greater:()=>al,GreaterEqual:()=>fi,History:()=>m6,IFFT:()=>hf,Identity:()=>mi,Imag:()=>nh,InputSpec:()=>Zt,IsFinite:()=>Zu,IsInf:()=>Yu,IsNan:()=>Ju,KernelBackend:()=>zu,LRN:()=>sh,LRNGrad:()=>ff,LayerVariable:()=>s6,LayersModel:()=>Ka,LeakyRelu:()=>gi,Less:()=>sl,LessEqual:()=>il,LinSpace:()=>cf,Log:()=>yi,Log1p:()=>Qu,LogSoftmax:()=>Lv,LogicalAnd:()=>ol,LogicalNot:()=>ed,LogicalOr:()=>ah,LowerBound:()=>WR,MathBackendCPU:()=>Qx,MathBackendWebGL:()=>Bh,Max:()=>Ai,MaxPool:()=>bi,MaxPool3D:()=>ih,MaxPool3DGrad:()=>gf,MaxPoolGrad:()=>mf,MaxPoolWithArgmax:()=>yf,Maximum:()=>xi,Mean:()=>vi,Min:()=>wi,Minimum:()=>ki,MirrorPad:()=>Ii,Mod:()=>td,MomentumOptimizer:()=>Yf,Multinomial:()=>Af,Multiply:()=>Si,Neg:()=>ll,NonMaxSuppressionV3:()=>dl,NonMaxSuppressionV4:()=>rd,NonMaxSuppressionV5:()=>pl,NotEqual:()=>ul,OP_SCOPE_SUFFIX:()=>e7,OneHot:()=>cl,OnesLike:()=>hl,Optimizer:()=>ns,OptimizerConstructors:()=>Is,Pack:()=>fl,PadV2:()=>Ci,Pool:()=>VR,Pow:()=>Ti,Prelu:()=>Ni,Prod:()=>Ei,RMSPropOptimizer:()=>Jf,RNN:()=>as,Range:()=>nd,Rank:()=>Gv,Real:()=>oh,RealDiv:()=>li,Reciprocal:()=>ad,Reduction:()=>Uw,Relu:()=>Ri,Relu6:()=>Mi,Reshape:()=>ml,ResizeBilinear:()=>$i,ResizeBilinearGrad:()=>bf,ResizeNearestNeighbor:()=>sd,ResizeNearestNeighborGrad:()=>xf,Reverse:()=>gl,RotateWithOffset:()=>Rl,Round:()=>yl,Rsqrt:()=>Fi,SGDOptimizer:()=>Ch,ScatterNd:()=>Al,SearchSorted:()=>vf,Select:()=>xl,Selu:()=>id,Sequential:()=>hm,Sigmoid:()=>_i,Sign:()=>od,Sin:()=>Pi,Sinh:()=>vl,Slice:()=>bl,Softmax:()=>Di,Softplus:()=>ld,SpaceToBatchND:()=>wl,SparseFillEmptyRows:()=>lh,SparseReshape:()=>ud,SparseSegmentMean:()=>uh,SparseSegmentSum:()=>dh,SparseToDense:()=>ph,SplitV:()=>kl,Sqrt:()=>zi,Square:()=>dd,SquaredDifference:()=>Li,Step:()=>Ui,StridedSlice:()=>Il,StringNGrams:()=>hh,StringSplit:()=>wf,StringToHashBucketFast:()=>kf,Sub:()=>Bi,Sum:()=>Oi,SymbolicTensor:()=>pa,Tan:()=>Sl,Tanh:()=>Wi,Tensor:()=>nt,TensorBuffer:()=>ir,Tile:()=>es,TopK:()=>Cl,Transform:()=>Tl,Transpose:()=>Vi,Unique:()=>If,Unpack:()=>Nl,UnsortedSegmentSum:()=>ch,UpperBound:()=>UR,Variable:()=>Bp,ZerosLike:()=>El,_FusedMatMul:()=>Ps,abs:()=>ar,acos:()=>F7,acosh:()=>P7,add:()=>le,addN:()=>Cf,all:()=>Dy,any:()=>C0,argMax:()=>Rn,argMin:()=>_7,asin:()=>z7,asinh:()=>O7,atan:()=>D7,atan2:()=>L7,atanh:()=>B7,avgPool:()=>Tf,avgPool3d:()=>By,backend:()=>Dn,backend_util:()=>C,basicLSTMCell:()=>RF,batchNorm:()=>Su,batchNorm2d:()=>G7,batchNorm3d:()=>j7,batchNorm4d:()=>H7,batchToSpaceND:()=>Nf,bincount:()=>Wy,booleanMaskAsync:()=>Gz,broadcastArgs:()=>q7,broadcastTo:()=>Mp,broadcast_util:()=>ji,browser:()=>On,buffer:()=>We,callbacks:()=>wj,cast:()=>me,ceil:()=>X7,clipByValue:()=>cn,clone:()=>Wr,complex:()=>Os,concat:()=>St,concat1d:()=>K7,concat2d:()=>hd,concat3d:()=>Z7,concat4d:()=>Y7,constraints:()=>d6,conv1d:()=>Vy,conv2d:()=>Ls,conv2dTranspose:()=>Gy,conv3d:()=>jy,conv3dTranspose:()=>Q7,copyRegisteredKernels:()=>qR,cos:()=>Ef,cosh:()=>Hy,cosineWindow:()=>yA,cumprod:()=>N0,cumsum:()=>qy,customGrad:()=>Pa,data:()=>Bk,denseBincount:()=>ew,deprecationWarn:()=>_y,depthToSpace:()=>tw,depthwiseConv2d:()=>vh,deregisterOp:()=>Sj,device_util:()=>yh,diag:()=>oP,dilation2d:()=>rw,disableDeprecationWarnings:()=>HM,dispose:()=>te,disposeVariables:()=>qM,div:()=>pe,divNoNan:()=>nw,dot:()=>fP,dropout:()=>Pw,einsum:()=>aw,elu:()=>wh,enableDebugMode:()=>jM,enableProdMode:()=>Py,enclosingPowerOfTwo:()=>_w,engine:()=>Xt,env:()=>Z,equal:()=>$n,erf:()=>sw,exp:()=>Mn,expandDims:()=>Kt,expm1:()=>iw,eye:()=>Xy,fft:()=>Bf,fill:()=>cd,findBackend:()=>Oy,findBackendFactory:()=>YM,floor:()=>kh,floorDiv:()=>xh,forceHalfFloat:()=>l9,fused:()=>Vs,gather:()=>Cu,gatherND:()=>Fw,gather_util:()=>Ty,getBackend:()=>Gr,getGradient:()=>mg,getKernel:()=>v0,getKernelsForBackend:()=>Ma,getThreadsCount:()=>Pye,gpgpu_util:()=>LI,grad:()=>LP,grads:()=>BP,greater:()=>mn,greaterEqual:()=>Ml,ifft:()=>Gp,imag:()=>Rf,image:()=>Ie,inTopKAsync:()=>tO,initializers:()=>p6,input:()=>R6,io:()=>Cr,irfft:()=>hA,isFinite:()=>EP,isInf:()=>$P,isNaN:()=>ow,keep:()=>gr,kernel_impls:()=>Kn,layers:()=>h6,leakyRelu:()=>$f,less:()=>Ky,lessEqual:()=>Fl,linalg:()=>Gw,linspace:()=>lw,loadGraphModel:()=>NH,loadLayersModel:()=>FU,localResponseNormalization:()=>uw,log:()=>Fn,log1p:()=>Mf,logSigmoid:()=>HP,logSoftmax:()=>Zy,logSumExp:()=>fw,logicalAnd:()=>ga,logicalNot:()=>Pf,logicalOr:()=>Qy,logicalXor:()=>a_,losses:()=>LD,lowerBound:()=>mw,matMul:()=>Ze,math:()=>f7,max:()=>Ar,maxPool:()=>_f,maxPool3d:()=>tA,maxPoolWithArgmax:()=>gw,maximum:()=>ts,mean:()=>Vt,memory:()=>S0,meshgrid:()=>p_,metrics:()=>pk,min:()=>Bs,minimum:()=>Ih,mirrorPad:()=>yw,mod:()=>md,model:()=>$U,models:()=>hk,moments:()=>zf,movingAverage:()=>qz,mul:()=>L,multiRNNCell:()=>x_,multinomial:()=>Aw,neg:()=>Dt,nextFrame:()=>bA,norm:()=>mA,notEqual:()=>Tu,oneHot:()=>Vp,ones:()=>hn,onesLike:()=>Pn,op:()=>W,outerProduct:()=>I_,pad:()=>Xn,pad1d:()=>T_,pad2d:()=>E_,pad3d:()=>$_,pad4d:()=>F_,pool:()=>D_,pow:()=>Ws,prelu:()=>Df,print:()=>d7,prod:()=>rA,profile:()=>XM,rand:()=>U_,randomGamma:()=>q_,randomNormal:()=>xw,randomUniform:()=>gd,range:()=>Nu,ready:()=>pd,real:()=>Up,reciprocal:()=>bw,registerBackend:()=>$l,registerCallbackConstructor:()=>PU,registerGradient:()=>Bv,registerKernel:()=>qn,registerOp:()=>Ij,regularizers:()=>ck,relu:()=>Oa,relu6:()=>sA,removeBackend:()=>ZM,reshape:()=>U,reverse:()=>_n,reverse1d:()=>rz,reverse2d:()=>az,reverse3d:()=>iz,reverse4d:()=>lz,rfft:()=>Wf,round:()=>iA,rsqrt:()=>oA,scalar:()=>Se,scatterND:()=>Mw,scatter_util:()=>Ny,searchSorted:()=>eA,selu:()=>lA,separableConv2d:()=>vw,sequential:()=>MU,serialization:()=>ue,setBackend:()=>zy,setPlatform:()=>JM,setThreadsCount:()=>Fye,setWasmPath:()=>Mye,setWasmPaths:()=>_5,setWebGLContext:()=>Im,setdiff1dAsync:()=>ww,shared:()=>wm,sigmoid:()=>Tr,sign:()=>kw,signal:()=>DD,sin:()=>uA,sinh:()=>dA,slice:()=>Pe,slice1d:()=>Lf,slice2d:()=>pA,slice3d:()=>Pl,slice4d:()=>Oo,slice_util:()=>Ot,softmax:()=>yd,softplus:()=>fd,spaceToBatchND:()=>Of,sparse:()=>kp,sparseToDense:()=>gA,spectral:()=>OD,split:()=>Yt,sqrt:()=>Er,square:()=>At,squaredDifference:()=>cA,squeeze:()=>tt,stack:()=>ur,step:()=>Sh,stridedSlice:()=>Iw,string:()=>i0,sub:()=>ce,sum:()=>ke,sumOutType:()=>gh,tan:()=>Sw,tanh:()=>Iu,tensor:()=>ft,tensor1d:()=>Tt,tensor2d:()=>ca,tensor3d:()=>g7,tensor4d:()=>Pz,tensor5d:()=>_z,tensor6d:()=>zz,tensor_util:()=>ha,test_util:()=>R7,tidy:()=>X,tile:()=>Gn,time:()=>KM,topk:()=>Cw,train:()=>bo,transpose:()=>et,truncatedNormal:()=>Vf,unique:()=>Ng,unregisterGradient:()=>HR,unregisterKernel:()=>jR,unsortedSegmentSum:()=>Tw,unstack:()=>nn,upcastType:()=>Nr,upperBound:()=>Nw,util:()=>v,valueAndGrad:()=>WP,valueAndGrads:()=>VP,variable:()=>Ew,variableGrads:()=>dw,version:()=>Kh,version_converter:()=>EH,version_core:()=>Fy,version_cpu:()=>mX,version_layers:()=>UA,version_wasm:()=>_ye,version_webgl:()=>Gte,webgl:()=>jte,webgl_util:()=>uI,webgpu:()=>dS,where:()=>Vr,whereAsync:()=>fA,zeros:()=>_t,zerosLike:()=>at});var YE=Object.create,X0=Object.defineProperty,JE=Object.getOwnPropertyDescriptor,Iv=Object.getOwnPropertyNames,QE=Object.getPrototypeOf,eR=Object.prototype.hasOwnProperty,tR=e=>X0(e,"__esModule",{value:!0}),pr=(e,t)=>function(){return t||(0,e[Iv(e)[0]])((t={exports:{}}).exports,t),t.exports},Le=(e,t)=>{for(var r in t)X0(e,r,{get:t[r],enumerable:!0})},rR=(e,t,r,n)=>{if(t&&typeof t=="object"||typeof t=="function")for(let a of Iv(t))!eR.call(e,a)&&(r||a!=="default")&&X0(e,a,{get:()=>t[a],enumerable:!(n=JE(t,a))||n.enumerable});return e},Go=(e,t)=>rR(tR(X0(e!=null?YE(QE(e)):{},"default",!t&&e&&e.__esModule?{get:()=>e.default,enumerable:!0}:{value:e,enumerable:!0})),e),nR=pr({"src/node_modules/long/src/long.js"(e,t){t.exports=n;var r=null;try{r=new WebAssembly.Instance(new WebAssembly.Module(new Uint8Array([0,97,115,109,1,0,0,0,1,13,2,96,0,1,127,96,4,127,127,127,127,1,127,3,7,6,0,1,1,1,1,1,6,6,1,127,1,65,0,11,7,50,6,3,109,117,108,0,1,5,100,105,118,95,115,0,2,5,100,105,118,95,117,0,3,5,114,101,109,95,115,0,4,5,114,101,109,95,117,0,5,8,103,101,116,95,104,105,103,104,0,0,10,191,1,6,4,0,35,0,11,36,1,1,126,32,0,173,32,1,173,66,32,134,132,32,2,173,32,3,173,66,32,134,132,126,34,4,66,32,135,167,36,0,32,4,167,11,36,1,1,126,32,0,173,32,1,173,66,32,134,132,32,2,173,32,3,173,66,32,134,132,127,34,4,66,32,135,167,36,0,32,4,167,11,36,1,1,126,32,0,173,32,1,173,66,32,134,132,32,2,173,32,3,173,66,32,134,132,128,34,4,66,32,135,167,36,0,32,4,167,11,36,1,1,126,32,0,173,32,1,173,66,32,134,132,32,2,173,32,3,173,66,32,134,132,129,34,4,66,32,135,167,36,0,32,4,167,11,36,1,1,126,32,0,173,32,1,173,66,32,134,132,32,2,173,32,3,173,66,32,134,132,130,34,4,66,32,135,167,36,0,32,4,167,11])),{}).exports}catch(I){}function n(I,_,O){this.low=I|0,this.high=_|0,this.unsigned=!!O}n.prototype.__isLong__,Object.defineProperty(n.prototype,"__isLong__",{value:!0});function a(I){return(I&&I.__isLong__)===!0}n.isLong=a;var s={},i={};function o(I,_){var O,q,K;return _?(I>>>=0,(K=0<=I&&I<256)&&(q=i[I],q)?q:(O=u(I,(I|0)<0?-1:0,!0),K&&(i[I]=O),O)):(I|=0,(K=-128<=I&&I<128)&&(q=s[I],q)?q:(O=u(I,I<0?-1:0,!1),K&&(s[I]=O),O))}n.fromInt=o;function l(I,_){if(isNaN(I))return _?b:x;if(_){if(I<0)return b;if(I>=g)return R}else{if(I<=-y)return z;if(I+1>=y)return E}return I<0?l(-I,_).neg():u(I%m|0,I/m|0,_)}n.fromNumber=l;function u(I,_,O){return new n(I,_,O)}n.fromBits=u;var d=Math.pow;function h(I,_,O){if(I.length===0)throw Error("empty string");if(I==="NaN"||I==="Infinity"||I==="+Infinity"||I==="-Infinity")return x;if(typeof _=="number"?(O=_,_=!1):_=!!_,O=O||10,O<2||36<O)throw RangeError("radix");var q;if((q=I.indexOf("-"))>0)throw Error("interior hyphen");if(q===0)return h(I.substring(1),_,O).neg();for(var K=l(d(O,8)),D=x,Q=0;Q<I.length;Q+=8){var V=Math.min(8,I.length-Q),re=parseInt(I.substring(Q,Q+V),O);if(V<8){var Y=l(d(O,V));D=D.mul(Y).add(l(re))}else D=D.mul(K),D=D.add(l(re))}return D.unsigned=_,D}n.fromString=h;function p(I,_){return typeof I=="number"?l(I,_):typeof I=="string"?h(I,_):u(I.low,I.high,typeof _=="boolean"?_:I.unsigned)}n.fromValue=p;var c=1<<16,f=1<<24,m=c*c,g=m*m,y=g/2,A=o(f),x=o(0);n.ZERO=x;var b=o(0,!0);n.UZERO=b;var w=o(1);n.ONE=w;var S=o(1,!0);n.UONE=S;var N=o(-1);n.NEG_ONE=N;var E=u(-1,2147483647,!1);n.MAX_VALUE=E;var R=u(-1,-1,!0);n.MAX_UNSIGNED_VALUE=R;var z=u(0,-2147483648,!1);n.MIN_VALUE=z;var M=n.prototype;M.toInt=function(){return this.unsigned?this.low>>>0:this.low},M.toNumber=function(){return this.unsigned?(this.high>>>0)*m+(this.low>>>0):this.high*m+(this.low>>>0)},M.toString=function(I){if(I=I||10,I<2||36<I)throw RangeError("radix");if(this.isZero())return"0";if(this.isNegative())if(this.eq(z)){var _=l(I),O=this.div(_),q=O.mul(_).sub(this);return O.toString(I)+q.toInt().toString(I)}else return"-"+this.neg().toString(I);for(var K=l(d(I,6),this.unsigned),D=this,Q="";;){var V=D.div(K),re=D.sub(V.mul(K)).toInt()>>>0,Y=re.toString(I);if(D=V,D.isZero())return Y+Q;for(;Y.length<6;)Y="0"+Y;Q=""+Y+Q}},M.getHighBits=function(){return this.high},M.getHighBitsUnsigned=function(){return this.high>>>0},M.getLowBits=function(){return this.low},M.getLowBitsUnsigned=function(){return this.low>>>0},M.getNumBitsAbs=function(){if(this.isNegative())return this.eq(z)?64:this.neg().getNumBitsAbs();for(var I=this.high!=0?this.high:this.low,_=31;_>0&&(I&1<<_)==0;_--);return this.high!=0?_+33:_+1},M.isZero=function(){return this.high===0&&this.low===0},M.eqz=M.isZero,M.isNegative=function(){return!this.unsigned&&this.high<0},M.isPositive=function(){return this.unsigned||this.high>=0},M.isOdd=function(){return(this.low&1)===1},M.isEven=function(){return(this.low&1)===0},M.equals=function(I){return a(I)||(I=p(I)),this.unsigned!==I.unsigned&&this.high>>>31===1&&I.high>>>31===1?!1:this.high===I.high&&this.low===I.low},M.eq=M.equals,M.notEquals=function(I){return!this.eq(I)},M.neq=M.notEquals,M.ne=M.notEquals,M.lessThan=function(I){return this.comp(I)<0},M.lt=M.lessThan,M.lessThanOrEqual=function(I){return this.comp(I)<=0},M.lte=M.lessThanOrEqual,M.le=M.lessThanOrEqual,M.greaterThan=function(I){return this.comp(I)>0},M.gt=M.greaterThan,M.greaterThanOrEqual=function(I){return this.comp(I)>=0},M.gte=M.greaterThanOrEqual,M.ge=M.greaterThanOrEqual,M.compare=function(I){if(a(I)||(I=p(I)),this.eq(I))return 0;var _=this.isNegative(),O=I.isNegative();return _&&!O?-1:!_&&O?1:this.unsigned?I.high>>>0>this.high>>>0||I.high===this.high&&I.low>>>0>this.low>>>0?-1:1:this.sub(I).isNegative()?-1:1},M.comp=M.compare,M.negate=function(){return!this.unsigned&&this.eq(z)?z:this.not().add(w)},M.neg=M.negate,M.add=function(I){a(I)||(I=p(I));var _=this.high>>>16,O=this.high&65535,q=this.low>>>16,K=this.low&65535,D=I.high>>>16,Q=I.high&65535,V=I.low>>>16,re=I.low&65535,Y=0,ie=0,J=0,ae=0;return ae+=K+re,J+=ae>>>16,ae&=65535,J+=q+V,ie+=J>>>16,J&=65535,ie+=O+Q,Y+=ie>>>16,ie&=65535,Y+=_+D,Y&=65535,u(J<<16|ae,Y<<16|ie,this.unsigned)},M.subtract=function(I){return a(I)||(I=p(I)),this.add(I.neg())},M.sub=M.subtract,M.multiply=function(I){if(this.isZero())return x;if(a(I)||(I=p(I)),r){var _=r.mul(this.low,this.high,I.low,I.high);return u(_,r.get_high(),this.unsigned)}if(I.isZero())return x;if(this.eq(z))return I.isOdd()?z:x;if(I.eq(z))return this.isOdd()?z:x;if(this.isNegative())return I.isNegative()?this.neg().mul(I.neg()):this.neg().mul(I).neg();if(I.isNegative())return this.mul(I.neg()).neg();if(this.lt(A)&&I.lt(A))return l(this.toNumber()*I.toNumber(),this.unsigned);var O=this.high>>>16,q=this.high&65535,K=this.low>>>16,D=this.low&65535,Q=I.high>>>16,V=I.high&65535,re=I.low>>>16,Y=I.low&65535,ie=0,J=0,ae=0,de=0;return de+=D*Y,ae+=de>>>16,de&=65535,ae+=K*Y,J+=ae>>>16,ae&=65535,ae+=D*re,J+=ae>>>16,ae&=65535,J+=q*Y,ie+=J>>>16,J&=65535,J+=K*re,ie+=J>>>16,J&=65535,J+=D*V,ie+=J>>>16,J&=65535,ie+=O*Y+q*re+K*V+D*Q,ie&=65535,u(ae<<16|de,ie<<16|J,this.unsigned)},M.mul=M.multiply,M.divide=function(I){if(a(I)||(I=p(I)),I.isZero())throw Error("division by zero");if(r){if(!this.unsigned&&this.high===-2147483648&&I.low===-1&&I.high===-1)return this;var _=(this.unsigned?r.div_u:r.div_s)(this.low,this.high,I.low,I.high);return u(_,r.get_high(),this.unsigned)}if(this.isZero())return this.unsigned?b:x;var O,q,K;if(this.unsigned){if(I.unsigned||(I=I.toUnsigned()),I.gt(this))return b;if(I.gt(this.shru(1)))return S;K=b}else{if(this.eq(z)){if(I.eq(w)||I.eq(N))return z;if(I.eq(z))return w;var D=this.shr(1);return O=D.div(I).shl(1),O.eq(x)?I.isNegative()?w:N:(q=this.sub(I.mul(O)),K=O.add(q.div(I)),K)}else if(I.eq(z))return this.unsigned?b:x;if(this.isNegative())return I.isNegative()?this.neg().div(I.neg()):this.neg().div(I).neg();if(I.isNegative())return this.div(I.neg()).neg();K=x}for(q=this;q.gte(I);){O=Math.max(1,Math.floor(q.toNumber()/I.toNumber()));for(var Q=Math.ceil(Math.log(O)/Math.LN2),V=Q<=48?1:d(2,Q-48),re=l(O),Y=re.mul(I);Y.isNegative()||Y.gt(q);)O-=V,re=l(O,this.unsigned),Y=re.mul(I);re.isZero()&&(re=w),K=K.add(re),q=q.sub(Y)}return K},M.div=M.divide,M.modulo=function(I){if(a(I)||(I=p(I)),r){var _=(this.unsigned?r.rem_u:r.rem_s)(this.low,this.high,I.low,I.high);return u(_,r.get_high(),this.unsigned)}return this.sub(this.div(I).mul(I))},M.mod=M.modulo,M.rem=M.modulo,M.not=function(){return u(~this.low,~this.high,this.unsigned)},M.and=function(I){return a(I)||(I=p(I)),u(this.low&I.low,this.high&I.high,this.unsigned)},M.or=function(I){return a(I)||(I=p(I)),u(this.low|I.low,this.high|I.high,this.unsigned)},M.xor=function(I){return a(I)||(I=p(I)),u(this.low^I.low,this.high^I.high,this.unsigned)},M.shiftLeft=function(I){return a(I)&&(I=I.toInt()),(I&=63)===0?this:I<32?u(this.low<<I,this.high<<I|this.low>>>32-I,this.unsigned):u(0,this.low<<I-32,this.unsigned)},M.shl=M.shiftLeft,M.shiftRight=function(I){return a(I)&&(I=I.toInt()),(I&=63)===0?this:I<32?u(this.low>>>I|this.high<<32-I,this.high>>I,this.unsigned):u(this.high>>I-32,this.high>=0?0:-1,this.unsigned)},M.shr=M.shiftRight,M.shiftRightUnsigned=function(I){if(a(I)&&(I=I.toInt()),I&=63,I===0)return this;var _=this.high;if(I<32){var O=this.low;return u(O>>>I|_<<32-I,_>>>I,this.unsigned)}else return I===32?u(_,0,this.unsigned):u(_>>>I-32,0,this.unsigned)},M.shru=M.shiftRightUnsigned,M.shr_u=M.shiftRightUnsigned,M.toSigned=function(){return this.unsigned?u(this.low,this.high,!1):this},M.toUnsigned=function(){return this.unsigned?this:u(this.low,this.high,!0)},M.toBytes=function(I){return I?this.toBytesLE():this.toBytesBE()},M.toBytesLE=function(){var I=this.high,_=this.low;return[_&255,_>>>8&255,_>>>16&255,_>>>24,I&255,I>>>8&255,I>>>16&255,I>>>24]},M.toBytesBE=function(){var I=this.high,_=this.low;return[I>>>24,I>>>16&255,I>>>8&255,I&255,_>>>24,_>>>16&255,_>>>8&255,_&255]},n.fromBytes=function(I,_,O){return O?n.fromBytesLE(I,_):n.fromBytesBE(I,_)},n.fromBytesLE=function(I,_){return new n(I[0]|I[1]<<8|I[2]<<16|I[3]<<24,I[4]|I[5]<<8|I[6]<<16|I[7]<<24,_)},n.fromBytesBE=function(I,_){return new n(I[4]<<24|I[5]<<16|I[6]<<8|I[7],I[0]<<24|I[1]<<16|I[2]<<8|I[3],_)}}}),aR=pr({"(disabled):src/node_modules/node-fetch/browser.js"(){}}),sR=pr({"(disabled):util"(){}}),iR=pr({"src/node_modules/seedrandom/lib/alea.js"(e,t){(function(r,n,a){function s(u){var d=this,h=l();d.next=function(){var p=2091639*d.s0+d.c*23283064365386963e-26;return d.s0=d.s1,d.s1=d.s2,d.s2=p-(d.c=p|0)},d.c=1,d.s0=h(" "),d.s1=h(" "),d.s2=h(" "),d.s0-=h(u),d.s0<0&&(d.s0+=1),d.s1-=h(u),d.s1<0&&(d.s1+=1),d.s2-=h(u),d.s2<0&&(d.s2+=1),h=null}function i(u,d){return d.c=u.c,d.s0=u.s0,d.s1=u.s1,d.s2=u.s2,d}function o(u,d){var h=new s(u),p=d&&d.state,c=h.next;return c.int32=function(){return h.next()*4294967296|0},c.double=function(){return c()+(c()*2097152|0)*11102230246251565e-32},c.quick=c,p&&(typeof p=="object"&&i(p,h),c.state=function(){return i(h,{})}),c}function l(){var u=4022871197,d=function(h){h=String(h);for(var p=0;p<h.length;p++){u+=h.charCodeAt(p);var c=.02519603282416938*u;u=c>>>0,c-=u,c*=u,u=c>>>0,c-=u,u+=c*4294967296}return(u>>>0)*23283064365386963e-26};return d}n&&n.exports?n.exports=o:a&&a.amd?a(function(){return o}):this.alea=o})(e,typeof t=="object"&&t,typeof define=="function"&&define)}}),oR=pr({"src/node_modules/seedrandom/lib/xor128.js"(e,t){(function(r,n,a){function s(l){var u=this,d="";u.x=0,u.y=0,u.z=0,u.w=0,u.next=function(){var p=u.x^u.x<<11;return u.x=u.y,u.y=u.z,u.z=u.w,u.w^=u.w>>>19^p^p>>>8},l===(l|0)?u.x=l:d+=l;for(var h=0;h<d.length+64;h++)u.x^=d.charCodeAt(h)|0,u.next()}function i(l,u){return u.x=l.x,u.y=l.y,u.z=l.z,u.w=l.w,u}function o(l,u){var d=new s(l),h=u&&u.state,p=function(){return(d.next()>>>0)/4294967296};return p.double=function(){do var c=d.next()>>>11,f=(d.next()>>>0)/4294967296,m=(c+f)/(1<<21);while(m===0);return m},p.int32=d.next,p.quick=p,h&&(typeof h=="object"&&i(h,d),p.state=function(){return i(d,{})}),p}n&&n.exports?n.exports=o:a&&a.amd?a(function(){return o}):this.xor128=o})(e,typeof t=="object"&&t,typeof define=="function"&&define)}}),lR=pr({"src/node_modules/seedrandom/lib/xorwow.js"(e,t){(function(r,n,a){function s(l){var u=this,d="";u.next=function(){var p=u.x^u.x>>>2;return u.x=u.y,u.y=u.z,u.z=u.w,u.w=u.v,(u.d=u.d+362437|0)+(u.v=u.v^u.v<<4^(p^p<<1))|0},u.x=0,u.y=0,u.z=0,u.w=0,u.v=0,l===(l|0)?u.x=l:d+=l;for(var h=0;h<d.length+64;h++)u.x^=d.charCodeAt(h)|0,h==d.length&&(u.d=u.x<<10^u.x>>>4),u.next()}function i(l,u){return u.x=l.x,u.y=l.y,u.z=l.z,u.w=l.w,u.v=l.v,u.d=l.d,u}function o(l,u){var d=new s(l),h=u&&u.state,p=function(){return(d.next()>>>0)/4294967296};return p.double=function(){do var c=d.next()>>>11,f=(d.next()>>>0)/4294967296,m=(c+f)/(1<<21);while(m===0);return m},p.int32=d.next,p.quick=p,h&&(typeof h=="object"&&i(h,d),p.state=function(){return i(d,{})}),p}n&&n.exports?n.exports=o:a&&a.amd?a(function(){return o}):this.xorwow=o})(e,typeof t=="object"&&t,typeof define=="function"&&define)}}),uR=pr({"src/node_modules/seedrandom/lib/xorshift7.js"(e,t){(function(r,n,a){function s(l){var u=this;u.next=function(){var h=u.x,p=u.i,c,f,m;return c=h[p],c^=c>>>7,f=c^c<<24,c=h[p+1&7],f^=c^c>>>10,c=h[p+3&7],f^=c^c>>>3,c=h[p+4&7],f^=c^c<<7,c=h[p+7&7],c=c^c<<13,f^=c^c<<9,h[p]=f,u.i=p+1&7,f};function d(h,p){var c,f,m=[];if(p===(p|0))f=m[0]=p;else for(p=""+p,c=0;c<p.length;++c)m[c&7]=m[c&7]<<15^p.charCodeAt(c)+m[c+1&7]<<13;for(;m.length<8;)m.push(0);for(c=0;c<8&&m[c]===0;++c);for(c==8?f=m[7]=-1:f=m[c],h.x=m,h.i=0,c=256;c>0;--c)h.next()}d(u,l)}function i(l,u){return u.x=l.x.slice(),u.i=l.i,u}function o(l,u){l==null&&(l=+new Date);var d=new s(l),h=u&&u.state,p=function(){return(d.next()>>>0)/4294967296};return p.double=function(){do var c=d.next()>>>11,f=(d.next()>>>0)/4294967296,m=(c+f)/(1<<21);while(m===0);return m},p.int32=d.next,p.quick=p,h&&(h.x&&i(h,d),p.state=function(){return i(d,{})}),p}n&&n.exports?n.exports=o:a&&a.amd?a(function(){return o}):this.xorshift7=o})(e,typeof t=="object"&&t,typeof define=="function"&&define)}}),dR=pr({"src/node_modules/seedrandom/lib/xor4096.js"(e,t){(function(r,n,a){function s(l){var u=this;u.next=function(){var h=u.w,p=u.X,c=u.i,f,m;return u.w=h=h+1640531527|0,m=p[c+34&127],f=p[c=c+1&127],m^=m<<13,f^=f<<17,m^=m>>>15,f^=f>>>12,m=p[c]=m^f,u.i=c,m+(h^h>>>16)|0};function d(h,p){var c,f,m,g,y,A=[],x=128;for(p===(p|0)?(f=p,p=null):(p=p+"\0",f=0,x=Math.max(x,p.length)),m=0,g=-32;g<x;++g)p&&(f^=p.charCodeAt((g+32)%p.length)),g===0&&(y=f),f^=f<<10,f^=f>>>15,f^=f<<4,f^=f>>>13,g>=0&&(y=y+1640531527|0,c=A[g&127]^=f+y,m=c==0?m+1:0);for(m>=128&&(A[(p&&p.length||0)&127]=-1),m=127,g=4*128;g>0;--g)f=A[m+34&127],c=A[m=m+1&127],f^=f<<13,c^=c<<17,f^=f>>>15,c^=c>>>12,A[m]=f^c;h.w=y,h.X=A,h.i=m}d(u,l)}function i(l,u){return u.i=l.i,u.w=l.w,u.X=l.X.slice(),u}function o(l,u){l==null&&(l=+new Date);var d=new s(l),h=u&&u.state,p=function(){return(d.next()>>>0)/4294967296};return p.double=function(){do var c=d.next()>>>11,f=(d.next()>>>0)/4294967296,m=(c+f)/(1<<21);while(m===0);return m},p.int32=d.next,p.quick=p,h&&(h.X&&i(h,d),p.state=function(){return i(d,{})}),p}n&&n.exports?n.exports=o:a&&a.amd?a(function(){return o}):this.xor4096=o})(e,typeof t=="object"&&t,typeof define=="function"&&define)}}),pR=pr({"src/node_modules/seedrandom/lib/tychei.js"(e,t){(function(r,n,a){function s(l){var u=this,d="";u.next=function(){var p=u.b,c=u.c,f=u.d,m=u.a;return p=p<<25^p>>>7^c,c=c-f|0,f=f<<24^f>>>8^m,m=m-p|0,u.b=p=p<<20^p>>>12^c,u.c=c=c-f|0,u.d=f<<16^c>>>16^m,u.a=m-p|0},u.a=0,u.b=0,u.c=-1640531527,u.d=1367130551,l===Math.floor(l)?(u.a=l/4294967296|0,u.b=l|0):d+=l;for(var h=0;h<d.length+20;h++)u.b^=d.charCodeAt(h)|0,u.next()}function i(l,u){return u.a=l.a,u.b=l.b,u.c=l.c,u.d=l.d,u}function o(l,u){var d=new s(l),h=u&&u.state,p=function(){return(d.next()>>>0)/4294967296};return p.double=function(){do var c=d.next()>>>11,f=(d.next()>>>0)/4294967296,m=(c+f)/(1<<21);while(m===0);return m},p.int32=d.next,p.quick=p,h&&(typeof h=="object"&&i(h,d),p.state=function(){return i(d,{})}),p}n&&n.exports?n.exports=o:a&&a.amd?a(function(){return o}):this.tychei=o})(e,typeof t=="object"&&t,typeof define=="function"&&define)}}),hR=pr({"(disabled):crypto"(){}}),cR=pr({"src/node_modules/seedrandom/seedrandom.js"(e,t){(function(r,n,a){var s=256,i=6,o=52,l="random",u=a.pow(s,i),d=a.pow(2,o),h=d*2,p=s-1,c;function f(w,S,N){var E=[];S=S==!0?{entropy:!0}:S||{};var R=A(y(S.entropy?[w,b(n)]:w==null?x():w,3),E),z=new m(E),M=function(){for(var I=z.g(i),_=u,O=0;I<d;)I=(I+O)*s,_*=s,O=z.g(1);for(;I>=h;)I/=2,_/=2,O>>>=1;return(I+O)/_};return M.int32=function(){return z.g(4)|0},M.quick=function(){return z.g(4)/4294967296},M.double=M,A(b(z.S),n),(S.pass||N||function(I,_,O,q){return q&&(q.S&&g(q,z),I.state=function(){return g(z,{})}),O?(a[l]=I,_):I})(M,R,"global"in S?S.global:this==a,S.state)}function m(w){var S,N=w.length,E=this,R=0,z=E.i=E.j=0,M=E.S=[];for(N||(w=[N++]);R<s;)M[R]=R++;for(R=0;R<s;R++)M[R]=M[z=p&z+w[R%N]+(S=M[R])],M[z]=S;(E.g=function(I){for(var _,O=0,q=E.i,K=E.j,D=E.S;I--;)_=D[q=p&q+1],O=O*s+D[p&(D[q]=D[K=p&K+_])+(D[K]=_)];return E.i=q,E.j=K,O})(s)}function g(w,S){return S.i=w.i,S.j=w.j,S.S=w.S.slice(),S}function y(w,S){var N=[],E=typeof w,R;if(S&&E=="object")for(R in w)try{N.push(y(w[R],S-1))}catch(z){}return N.length?N:E=="string"?w:w+"\0"}function A(w,S){for(var N=w+"",E,R=0;R<N.length;)S[p&R]=p&(E^=S[p&R]*19)+N.charCodeAt(R++);return b(S)}function x(){try{var w;return c&&(w=c.randomBytes)?w=w(s):(w=new Uint8Array(s),(r.crypto||r.msCrypto).getRandomValues(w)),b(w)}catch(E){var S=r.navigator,N=S&&S.plugins;return[+new Date,r,N,r.screen,b(n)]}}function b(w){return String.fromCharCode.apply(0,w)}if(A(a.random(),n),typeof t=="object"&&t.exports){t.exports=f;try{c=hR()}catch(w){}}else typeof define=="function"&&define.amd?define(function(){return f}):a["seed"+l]=f})(typeof self!="undefined"?self:e,[],Math)}}),K0=pr({"src/node_modules/seedrandom/index.js"(e,t){var r=iR(),n=oR(),a=lR(),s=uR(),i=dR(),o=pR(),l=cR();l.alea=r,l.xor128=n,l.xorwow=a,l.xorshift7=s,l.xor4096=i,l.tychei=o,t.exports=l}}),Sv=pr({"(disabled):src/node_modules/string_decoder/index.js"(){}}),fy=pr({"(disabled):fs"(){}}),g0=pr({"(disabled):path"(){}}),fR=pr({"(disabled):worker_threads"(){}}),mR=pr({"(disabled):perf_hooks"(){}}),gR=pr({"(disabled):os"(){}}),yR=pr({"src/tfjs-backend-wasm/wasm-out/tfjs-backend-wasm-threaded-simd.js"(e,t){var r=(()=>{var n=typeof document!="undefined"&&document.currentScript?document.currentScript.src:void 0;return typeof __filename!="undefined"&&(n=n||__filename),function(a){a=a||{};function s(){return Fe.buffer!=_r&&ra(Fe.buffer),dc}function i(){return Fe.buffer!=_r&&ra(Fe.buffer),pc}function o(){return Fe.buffer!=_r&&ra(Fe.buffer),np}function l(){return Fe.buffer!=_r&&ra(Fe.buffer),hc}function u(){return Fe.buffer!=_r&&ra(Fe.buffer),cc}function d(){return Fe.buffer!=_r&&ra(Fe.buffer),fc}function h(){return Fe.buffer!=_r&&ra(Fe.buffer),mc}var p=typeof a!="undefined"?a:{},c,f;p.ready=new Promise(function(T,F){c=T,f=F});var m;typeof process!="undefined"&&process.listeners&&(m={uncaughtException:process.listeners("uncaughtException"),unhandledRejection:process.listeners("unhandledRejection")});var g=Object.assign({},p),y=[],A="./this.program",x=(T,F)=>{throw F},b=typeof window=="object",w=typeof importScripts=="function",S=typeof process=="object"&&typeof process.versions=="object"&&typeof process.versions.node=="string",N=p.ENVIRONMENT_IS_PTHREAD||!1,E="";function R(T){return p.locateFile?p.locateFile(T,E):E+T}var z,M,I,_;function O(T){T instanceof hp||Y("exiting due to exception: "+T)}var q,K,D;if(S){w?E=g0().dirname(E)+"/":E=__dirname+"/",D=()=>{K||(q=fy(),K=g0())},z=function(F,G){return D(),F=K.normalize(F),q.readFileSync(F,G?void 0:"utf8")},I=F=>{var G=z(F,!0);return G.buffer||(G=new Uint8Array(G)),G},M=(F,G,ee)=>{D(),F=K.normalize(F),q.readFile(F,function(ge,Ae){ge?ee(ge):G(Ae.buffer)})},process.argv.length>1&&(A=process.argv[1].replace(/\\/g,"/")),y=process.argv.slice(2),process.on("uncaughtException",function(F){if(!(F instanceof hp))throw F}),process.on("unhandledRejection",function(F){throw F}),x=(F,G)=>{if(ho())throw process.exitCode=F,G;O(G),process.exit(F)},p.inspect=function(){return"[Emscripten Module object]"};let T;try{T=fR()}catch(F){throw console.error('The "worker_threads" module is not supported in this node.js build - perhaps a newer version is needed?'),F}global.Worker=T.Worker}else(b||w)&&(w?E=self.location.href:typeof document!="undefined"&&document.currentScript&&(E=document.currentScript.src),typeof n!="undefined"&&n&&(E=n),E.indexOf("blob:")!==0?E=E.substr(0,E.replace(/[?#].*/,"").lastIndexOf("/")+1):E="",S||(z=T=>{var F=new XMLHttpRequest;return F.open("GET",T,!1),F.send(null),F.responseText},w&&(I=T=>{var F=new XMLHttpRequest;return F.open("GET",T,!1),F.responseType="arraybuffer",F.send(null),new Uint8Array(F.response)}),M=(T,F,G)=>{var ee=new XMLHttpRequest;ee.open("GET",T,!0),ee.responseType="arraybuffer",ee.onload=()=>{if(ee.status==200||ee.status==0&&ee.response){F(ee.response);return}G()},ee.onerror=G,ee.send(null)}),_=T=>document.title=T);S&&typeof performance=="undefined"&&(global.performance=mR().performance);var Q=console.log.bind(console),V=console.warn.bind(console);S&&(D(),Q=T=>q.writeSync(1,T+`
|
|
`),V=T=>q.writeSync(2,T+`
|
|
`));var re=p.print||Q,Y=p.printErr||V;Object.assign(p,g),g=null,p.arguments&&(y=p.arguments),p.thisProgram&&(A=p.thisProgram),p.quit&&(x=p.quit);var ie=4;function J(T){J.shown||(J.shown={}),J.shown[T]||(J.shown[T]=1,Y(T))}function ae(T,F){if(typeof WebAssembly.Function=="function"){for(var G={i:"i32",j:"i64",f:"f32",d:"f64"},ee={parameters:[],results:F[0]=="v"?[]:[G[F[0]]]},ge=1;ge<F.length;++ge)ee.parameters.push(G[F[ge]]);return new WebAssembly.Function(ee,T)}var Ae=[1,0,1,96],Te=F.slice(0,1),_e=F.slice(1),Pt={i:127,j:126,f:125,d:124};Ae.push(_e.length);for(var ge=0;ge<_e.length;++ge)Ae.push(Pt[_e[ge]]);Te=="v"?Ae.push(0):Ae=Ae.concat([1,Pt[Te]]),Ae[1]=Ae.length-2;var ia=new Uint8Array([0,97,115,109,1,0,0,0].concat(Ae,[2,7,1,1,101,1,102,0,0,7,5,1,1,102,0,0])),oa=new WebAssembly.Module(ia),Gc=new WebAssembly.Instance(oa,{e:{f:T}}),cp=Gc.exports.f;return cp}var de=[],xe;function be(){if(de.length)return de.pop();try{Sn.grow(1)}catch(T){throw T instanceof RangeError?"Unable to grow wasm table. Set ALLOW_TABLE_GROWTH.":T}return Sn.length-1}function Ee(T,F){for(var G=T;G<T+F;G++){var ee=tu(G);ee&&xe.set(ee,G)}}var $e=0,De=T=>{$e=T},Be=Atomics.load,Xe=Atomics.store,ot=Atomics.compareExchange,pt;p.wasmBinary&&(pt=p.wasmBinary);var ht=p.noExitRuntime||!0;typeof WebAssembly!="object"&&Jl("no native wasm support detected");var Fe,wt,yt=!1,Pr;function cr(T,F){T||Jl(F)}function Jr(T){var F=p["_"+T];return F}function rr(T,F,G,ee,ge){var Ae={string:function(Cn){var lu=0;if(Cn!=null&&Cn!==0){var P3=(Cn.length<<2)+1;lu=ou(P3),uo(Cn,lu,P3)}return lu},array:function(Cn){var lu=ou(Cn.length);return Ua(Cn,lu),lu}};function Te(Cn){return F==="string"?In(Cn):F==="boolean"?Boolean(Cn):Cn}var _e=Jr(T),Pt=[],ia=0;if(ee)for(var oa=0;oa<ee.length;oa++){var Gc=Ae[G[oa]];Gc?(ia===0&&(ia=Y2()),Pt[oa]=Gc(ee[oa])):Pt[oa]=ee[oa]}var cp=_e.apply(null,Pt);function GE(Cn){return ia!==0&&Bc(ia),Te(Cn)}return cp=GE(cp),cp}function fr(T,F,G,ee){G=G||[];var ge=G.every(function(Te){return Te==="number"}),Ae=F!=="string";return Ae&&ge&&!ee?Jr(T):function(){return rr(T,F,G,arguments,ee)}}var ta=1;function Qr(T){var F=new TextDecoder(T);this.decode=G=>(G.buffer instanceof SharedArrayBuffer&&(G=new Uint8Array(G)),F.decode.call(F,G))}var nr=typeof TextDecoder!="undefined"?new Qr("utf8"):void 0;function kn(T,F,G){for(var ee=F+G,ge=F;T[ge]&&!(ge>=ee);)++ge;if(ge-F>16&&T.subarray&&nr)return nr.decode(T.subarray(F,ge));for(var Ae="";F<ge;){var Te=T[F++];if(!(Te&128)){Ae+=String.fromCharCode(Te);continue}var _e=T[F++]&63;if((Te&224)==192){Ae+=String.fromCharCode((Te&31)<<6|_e);continue}var Pt=T[F++]&63;if((Te&240)==224?Te=(Te&15)<<12|_e<<6|Pt:Te=(Te&7)<<18|_e<<12|Pt<<6|T[F++]&63,Te<65536)Ae+=String.fromCharCode(Te);else{var ia=Te-65536;Ae+=String.fromCharCode(55296|ia>>10,56320|ia&1023)}}return Ae}function In(T,F){return T?kn(i(),T,F):""}function gs(T,F,G,ee){if(!(ee>0))return 0;for(var ge=G,Ae=G+ee-1,Te=0;Te<T.length;++Te){var _e=T.charCodeAt(Te);if(_e>=55296&&_e<=57343){var Pt=T.charCodeAt(++Te);_e=65536+((_e&1023)<<10)|Pt&1023}if(_e<=127){if(G>=Ae)break;F[G++]=_e}else if(_e<=2047){if(G+1>=Ae)break;F[G++]=192|_e>>6,F[G++]=128|_e&63}else if(_e<=65535){if(G+2>=Ae)break;F[G++]=224|_e>>12,F[G++]=128|_e>>6&63,F[G++]=128|_e&63}else{if(G+3>=Ae)break;F[G++]=240|_e>>18,F[G++]=128|_e>>12&63,F[G++]=128|_e>>6&63,F[G++]=128|_e&63}}return F[G]=0,G-ge}function uo(T,F,G){return gs(T,i(),F,G)}function uc(T){for(var F=0,G=0;G<T.length;++G){var ee=T.charCodeAt(G);ee>=55296&&ee<=57343&&(ee=65536+((ee&1023)<<10)|T.charCodeAt(++G)&1023),ee<=127?++F:ee<=2047?F+=2:ee<=65535?F+=3:F+=4}return F}var ys=typeof TextDecoder!="undefined"?new Qr("utf-16le"):void 0;function Ua(T,F){s().set(T,F)}function rp(T,F,G){for(var ee=0;ee<T.length;++ee)s()[F++>>0]=T.charCodeAt(ee);G||(s()[F>>0]=0)}function Zl(T,F){return T%F>0&&(T+=F-T%F),T}var _r,dc,pc,np,hc,cc,f3,fc,mc;N&&(_r=p.buffer);function ra(T){_r=T,p.HEAP8=dc=new Int8Array(T),p.HEAP16=np=new Int16Array(T),p.HEAP32=cc=new Int32Array(T),p.HEAPU8=pc=new Uint8Array(T),p.HEAPU16=hc=new Uint16Array(T),p.HEAPU32=f3=new Uint32Array(T),p.HEAPF32=fc=new Float32Array(T),p.HEAPF64=mc=new Float64Array(T)}var gc=p.INITIAL_MEMORY||16777216;if(N)Fe=p.wasmMemory,_r=p.buffer;else if(p.wasmMemory)Fe=p.wasmMemory;else if(Fe=new WebAssembly.Memory({initial:gc/65536,maximum:32768,shared:!0}),!(Fe.buffer instanceof SharedArrayBuffer))throw Y("requested a shared WebAssembly.Memory but the returned buffer is not a SharedArrayBuffer, indicating that while the browser has SharedArrayBuffer it does not have WebAssembly threads support - you may need to set a flag"),S&&console.log("(on node you may need: --experimental-wasm-threads --experimental-wasm-bulk-memory and also use a recent version)"),Error("bad memory");Fe&&(_r=Fe.buffer),gc=_r.byteLength,ra(_r);var Sn,Yl=[],As=[],w1=[],yc=[],po=!1,k1=!1,Ac=0;function ho(){return ht||Ac>0}function zr(){if(p.preRun)for(typeof p.preRun=="function"&&(p.preRun=[p.preRun]);p.preRun.length;)m3(p.preRun.shift());kc(Yl)}function ap(){po=!0,!N&&kc(As)}function I1(){N||(ze.terminateAllThreads(),k1=!0)}function S1(){if(!N){if(p.postRun)for(typeof p.postRun=="function"&&(p.postRun=[p.postRun]);p.postRun.length;)sp(p.postRun.shift());kc(yc)}}function m3(T){Yl.unshift(T)}function g3(T){As.unshift(T)}function sp(T){yc.unshift(T)}var xs=0,xc=null,na=null;function ip(T){xs++,p.monitorRunDependencies&&p.monitorRunDependencies(xs)}function y3(T){if(xs--,p.monitorRunDependencies&&p.monitorRunDependencies(xs),xs==0&&(xc!==null&&(clearInterval(xc),xc=null),na)){var F=na;na=null,F()}}p.preloadedImages={},p.preloadedAudios={};function Jl(T){N?postMessage({cmd:"onAbort",arg:T}):p.onAbort&&p.onAbort(T),T="Aborted("+T+")",Y(T),yt=!0,Pr=1,T+=". Build with -s ASSERTIONS=1 for more info.";var F=new WebAssembly.RuntimeError(T);throw f(F),F}var C1="data:application/octet-stream;base64,";function bc(T){return T.startsWith(C1)}function vc(T){return T.startsWith("file://")}var Or;Or="tfjs-backend-wasm-threaded-simd.wasm",bc(Or)||(Or=R(Or));function wc(T){try{if(T==Or&&pt)return new Uint8Array(pt);if(I)return I(T);throw"both async and sync fetching of the wasm failed"}catch(F){Jl(F)}}function Ql(){if(!pt&&(b||w)){if(typeof fetch=="function"&&!vc(Or))return fetch(Or,{credentials:"same-origin"}).then(function(T){if(!T.ok)throw"failed to load wasm binary file at '"+Or+"'";return T.arrayBuffer()}).catch(function(){return wc(Or)});if(M)return new Promise(function(T,F){M(Or,function(G){T(new Uint8Array(G))},F)})}return Promise.resolve().then(function(){return wc(Or)})}function T1(){var T={env:_c,wasi_snapshot_preview1:_c};function F(Te,_e){var Pt=Te.exports;if(p.asm=Pt,P1(p.asm.emscripten_tls_init),Sn=p.asm.__indirect_function_table,g3(p.asm.__wasm_call_ctors),wt=_e,!N){var ia=ze.unusedWorkers.length;ze.unusedWorkers.forEach(function(oa){ze.loadWasmModuleToWorker(oa,function(){--ia||y3("wasm-instantiate")})})}}N||ip("wasm-instantiate");function G(Te){F(Te.instance,Te.module)}function ee(Te){return Ql().then(function(_e){return WebAssembly.instantiate(_e,T)}).then(function(_e){return _e}).then(Te,function(_e){Y("failed to asynchronously prepare wasm: "+_e),Jl(_e)})}function ge(){return!pt&&typeof WebAssembly.instantiateStreaming=="function"&&!bc(Or)&&!vc(Or)&&typeof fetch=="function"?fetch(Or,{credentials:"same-origin"}).then(function(Te){var _e=WebAssembly.instantiateStreaming(Te,T);return _e.then(G,function(Pt){return Y("wasm streaming compile failed: "+Pt),Y("falling back to ArrayBuffer instantiation"),ee(G)})}):ee(G)}if(p.instantiateWasm)try{var Ae=p.instantiateWasm(T,F);return Ae}catch(Te){return Y("Module.instantiateWasm callback failed with error: "+Te),!1}return ge().catch(f),{}}var A3,x3,N1={};function kc(T){for(;T.length>0;){var F=T.shift();if(typeof F=="function"){F(p);continue}var G=F.func;typeof G=="number"?F.arg===void 0?tu(G)():tu(G)(F.arg):G(F.arg===void 0?null:F.arg)}}function eu(T){var F=Y2(),G=T();return Bc(F),G}function JN(T){return T}function b3(T){var F=/\b_Z[\w\d_]+/g;return T.replace(F,function(G){var ee=G;return G===ee?G:ee+" ["+G+"]"})}function E1(T){u()[T>>2]=0;var F=ze.pthreads[T];delete ze.pthreads[T],F.worker.terminate(),Z2(T),ze.runningWorkers.splice(ze.runningWorkers.indexOf(F.worker),1),F.worker.pthread=void 0}function R1(T){var F=ze.pthreads[T];F.worker.postMessage({cmd:"cancel"})}function Ic(T){var F=ze.pthreads[T];if(F){u()[T>>2]=0;var G=F.worker;ze.returnWorkerToPool(G)}}function Sc(T){WE(T)}function $1(T){if(T instanceof hp||T=="unwind")return Pr;x(1,T)}var ze={unusedWorkers:[],runningWorkers:[],tlsInitFunctions:[],init:function(){N?ze.initWorker():ze.initMainThread()},initMainThread:function(){for(var T=8,F=0;F<T;++F)ze.allocateUnusedWorker()},initWorker:function(){ht=!1},pthreads:{},setExitStatus:function(T){Pr=T},terminateAllThreads:function(){for(var T in ze.pthreads){var F=ze.pthreads[T];F&&F.worker&&ze.returnWorkerToPool(F.worker)}for(var G=0;G<ze.unusedWorkers.length;++G){var ee=ze.unusedWorkers[G];ee.terminate()}ze.unusedWorkers=[]},returnWorkerToPool:function(T){ze.runWithoutMainThreadQueuedCalls(function(){delete ze.pthreads[T.pthread.threadInfoStruct],ze.unusedWorkers.push(T),ze.runningWorkers.splice(ze.runningWorkers.indexOf(T),1),Z2(T.pthread.threadInfoStruct),T.pthread=void 0})},runWithoutMainThreadQueuedCalls:function(T){u()[F3>>2]=0;try{T()}finally{u()[F3>>2]=1}},receiveObjectTransfer:function(T){},threadInit:function(){for(var T in ze.tlsInitFunctions)ze.tlsInitFunctions[T]()},loadWasmModuleToWorker:function(T,F){T.onmessage=G=>{var ee=G.data,ge=ee.cmd;if(T.pthread&&(ze.currentProxiedOperationCallerThread=T.pthread.threadInfoStruct),ee.targetThread&&ee.targetThread!=Lc()){var Ae=ze.pthreads[ee.targetThread];Ae?Ae.worker.postMessage(ee,ee.transferList):Y('Internal error! Worker sent a message "'+ge+'" to target pthread '+ee.targetThread+", but that thread no longer exists!"),ze.currentProxiedOperationCallerThread=void 0;return}ge==="processQueuedMainThreadWork"?N3():ge==="spawnThread"?Tc(ee):ge==="cleanupThread"?Ic(ee.thread):ge==="killThread"?E1(ee.thread):ge==="cancelThread"?R1(ee.thread):ge==="loaded"?(T.loaded=!0,F&&F(T),T.runPthread&&(T.runPthread(),delete T.runPthread)):ge==="print"?re("Thread "+ee.threadId+": "+ee.text):ge==="printErr"?Y("Thread "+ee.threadId+": "+ee.text):ge==="alert"?alert("Thread "+ee.threadId+": "+ee.text):ee.target==="setimmediate"?T.postMessage(ee):ge==="onAbort"?p.onAbort&&p.onAbort(ee.arg):Y("worker sent an unknown command "+ge),ze.currentProxiedOperationCallerThread=void 0},T.onerror=G=>{var ee="worker sent an error!";throw Y(ee+" "+G.filename+":"+G.lineno+": "+G.message),G},S&&(T.on("message",function(G){T.onmessage({data:G})}),T.on("error",function(G){T.onerror(G)}),T.on("detachedExit",function(){})),T.postMessage({cmd:"load",urlOrBlob:p.mainScriptUrlOrBlob||n,wasmMemory:Fe,wasmModule:wt})},allocateUnusedWorker:function(){var T=R("tfjs-backend-wasm-threaded-simd.worker.js");ze.unusedWorkers.push(new Worker(T))},getNewWorker:function(){return ze.unusedWorkers.length==0&&(ze.allocateUnusedWorker(),ze.loadWasmModuleToWorker(ze.unusedWorkers[0])),ze.unusedWorkers.pop()}};function M1(){var T=Lc(),F=u()[T+44>>2],G=u()[T+48>>2],ee=F-G;M3(F,ee),Bc(F)}p.establishStackSpace=M1;function Cc(T){if(N)return mo(1,0,T);try{Sc(T)}catch(F){$1(F)}}var co=[];function tu(T){var F=co[T];return F||(T>=co.length&&(co.length=T+1),co[T]=F=Sn.get(T)),F}function F1(T,F){return tu(T)(F)}p.invokeEntryPoint=F1;function v3(){var T=new Error;if(!T.stack){try{throw new Error}catch(F){T=F}if(!T.stack)return"(no stack trace available)"}return T.stack.toString()}function P1(T,F,G){ze.tlsInitFunctions.push(T)}function w3(T,F){Sn.set(T,F),co[T]=F}var fo;S?fo=()=>{var T=process.hrtime();return T[0]*1e3+T[1]/1e6}:N?fo=()=>performance.now()-p.__performance_now_clock_drift:fo=()=>performance.now();var _1=!0;function z1(T){return u()[T3()>>2]=T,T}function O1(T,F){var G;if(T===0)G=Date.now();else if((T===1||T===4)&&_1)G=fo();else return z1(28),-1;return u()[F>>2]=G/1e3|0,u()[F+4>>2]=G%1e3*1e3*1e3|0,0}function D1(T,F){return O1(T,F)}function L1(T){E3(T,!w,1,!b),ze.threadInit()}function B1(T){N?postMessage({cmd:"cleanupThread",thread:T}):Ic(T)}function Tc(T){var F=ze.getNewWorker();if(!F)return 6;ze.runningWorkers.push(F);var G=ze.pthreads[T.pthread_ptr]={worker:F,threadInfoStruct:T.pthread_ptr};F.pthread=G;var ee={cmd:"run",start_routine:T.startRoutine,arg:T.arg,threadInfoStruct:T.pthread_ptr};return F.runPthread=()=>{ee.time=performance.now(),F.postMessage(ee,T.transferList)},F.loaded&&(F.runPthread(),delete F.runPthread),0}function W1(T,F,G,ee){if(typeof SharedArrayBuffer=="undefined")return Y("Current environment does not support SharedArrayBuffer, pthreads are not available!"),6;var ge=[],Ae=0;if(N&&(ge.length===0||Ae))return R3(687865856,T,F,G,ee);if(Ae)return Ae;var Te={startRoutine:G,pthread_ptr:T,arg:ee,transferList:ge};return N?(Te.cmd="spawnThread",postMessage(Te,ge),0):Tc(Te)}function V1(){return 2097152}function U1(T,F){if(T==F)postMessage({cmd:"processQueuedMainThreadWork"});else if(N)postMessage({targetThread:T,cmd:"processThreadQueue"});else{var G=ze.pthreads[T],ee=G&&G.worker;if(!ee)return;ee.postMessage({cmd:"processThreadQueue"})}return 1}function G1(){Jl("")}function j1(){S||w||J("Blocking on the main thread is very dangerous, see https://emscripten.org/docs/porting/pthreads.html#blocking-on-the-main-browser-thread")}function Nc(){return 2147483648}function H1(T,F,G){i().copyWithin(T,F,F+G)}function q1(){return S?gR().cpus().length:navigator.hardwareConcurrency}function mo(T,F){var G=arguments.length-2,ee=arguments;return eu(function(){for(var ge=G,Ae=ou(ge*8),Te=Ae>>3,_e=0;_e<G;_e++){var Pt=ee[2+_e];h()[Te+_e]=Pt}return $3(T,ge,Ae,F)})}var op=[];function X1(T,F,G){op.length=F;for(var ee=G>>3,ge=0;ge<F;ge++)op[ge]=h()[ee+ge];var Ae=T<0,Te=Ae?N1[-T-1]:c2[T];return Te.apply(null,op)}function K1(T){try{return Fe.grow(T-_r.byteLength+65535>>>16),ra(Fe.buffer),1}catch(F){}}function Z1(T){var F=i().length;if(T=T>>>0,T<=F)return!1;var G=Nc();if(T>G)return!1;for(var ee=1;ee<=4;ee*=2){var ge=F*(1+.2/ee);ge=Math.min(ge,T+100663296);var Ae=Math.min(G,Zl(Math.max(T,ge),65536)),Te=K1(Ae);if(Te)return!0}return!1}var qe={inEventHandler:0,removeAllEventListeners:function(){for(var T=qe.eventHandlers.length-1;T>=0;--T)qe._removeHandler(T);qe.eventHandlers=[],qe.deferredCalls=[]},registerRemoveEventListeners:function(){qe.removeEventListenersRegistered||(w1.push(qe.removeAllEventListeners),qe.removeEventListenersRegistered=!0)},deferredCalls:[],deferCall:function(T,F,G){function ee(Te,_e){if(Te.length!=_e.length)return!1;for(var Pt in Te)if(Te[Pt]!=_e[Pt])return!1;return!0}for(var ge in qe.deferredCalls){var Ae=qe.deferredCalls[ge];if(Ae.targetFunction==T&&ee(Ae.argsList,G))return}qe.deferredCalls.push({targetFunction:T,precedence:F,argsList:G}),qe.deferredCalls.sort(function(Te,_e){return Te.precedence<_e.precedence})},removeDeferredCalls:function(T){for(var F=0;F<qe.deferredCalls.length;++F)qe.deferredCalls[F].targetFunction==T&&(qe.deferredCalls.splice(F,1),--F)},canPerformEventHandlerRequests:function(){return qe.inEventHandler&&qe.currentEventHandler.allowsDeferredCalls},runDeferredCalls:function(){if(qe.canPerformEventHandlerRequests())for(var T=0;T<qe.deferredCalls.length;++T){var F=qe.deferredCalls[T];qe.deferredCalls.splice(T,1),--T,F.targetFunction.apply(null,F.argsList)}},eventHandlers:[],removeAllHandlersOnTarget:function(T,F){for(var G=0;G<qe.eventHandlers.length;++G)qe.eventHandlers[G].target==T&&(!F||F==qe.eventHandlers[G].eventTypeString)&&qe._removeHandler(G--)},_removeHandler:function(T){var F=qe.eventHandlers[T];F.target.removeEventListener(F.eventTypeString,F.eventListenerFunc,F.useCapture),qe.eventHandlers.splice(T,1)},registerOrRemoveHandler:function(T){var F=function(ee){++qe.inEventHandler,qe.currentEventHandler=T,qe.runDeferredCalls(),T.handlerFunc(ee),qe.runDeferredCalls(),--qe.inEventHandler};if(T.callbackfunc)T.eventListenerFunc=F,T.target.addEventListener(T.eventTypeString,F,T.useCapture),qe.eventHandlers.push(T),qe.registerRemoveEventListeners();else for(var G=0;G<qe.eventHandlers.length;++G)qe.eventHandlers[G].target==T.target&&qe.eventHandlers[G].eventTypeString==T.eventTypeString&&qe._removeHandler(G--)},queueEventHandlerOnThread_iiii:function(T,F,G,ee,ge){eu(function(){var Ae=ou(12);u()[Ae>>2]=G,u()[Ae+4>>2]=ee,u()[Ae+8>>2]=ge,K2(T,637534208,F,ee,Ae)})},getTargetThreadForEventCallback:function(T){switch(T){case 1:return 0;case 2:return ze.currentProxiedOperationCallerThread;default:return T}},getNodeNameForTarget:function(T){return T?T==window?"#window":T==screen?"#screen":T&&T.nodeName?T.nodeName:"":""},fullscreenEnabled:function(){return document.fullscreenEnabled||document.webkitFullscreenEnabled}};function Y1(T){var F=uc(T)+1,G=X2(F);return uo(T,G,F),G}function J1(T,F,G,ee){eu(function(){var ge=ou(12),Ae=0;F&&(Ae=Y1(F)),u()[ge>>2]=Ae,u()[ge+4>>2]=G,u()[ge+8>>2]=ee,K2(T,657457152,0,Ae,ge)})}function Q1(T,F,G,ee){F=F?In(F):"",J1(T,F,G,ee)}function e2(T){return T>2?In(T):T}var t2=[0,typeof document!="undefined"?document:0,typeof window!="undefined"?window:0];function r2(T){T=e2(T);var F=t2[T]||(typeof document!="undefined"?document.querySelector(T):void 0);return F}function lp(T){return r2(T)}function Ec(T,F,G){var ee=lp(T);if(!ee)return-4;if(ee.canvasSharedPtr&&(u()[ee.canvasSharedPtr>>2]=F,u()[ee.canvasSharedPtr+4>>2]=G),ee.offscreenCanvas||!ee.controlTransferredOffscreen){ee.offscreenCanvas&&(ee=ee.offscreenCanvas);var ge=!1;if(ee.GLctxObject&&ee.GLctxObject.GLctx){var Ae=ee.GLctxObject.GLctx.getParameter(2978);ge=Ae[0]===0&&Ae[1]===0&&Ae[2]===ee.width&&Ae[3]===ee.height}ee.width=F,ee.height=G,ge&&ee.GLctxObject.GLctx.viewport(0,0,F,G)}else if(ee.canvasSharedPtr){var Te=u()[ee.canvasSharedPtr+8>>2];return Q1(Te,T,F,G),1}else return-4;return 0}function Rc(T,F,G){return N?mo(2,1,T,F,G):Ec(T,F,G)}function n2(T,F,G){var ee=lp(T);return ee?Ec(T,F,G):Rc(T,F,G)}function a2(){throw"unwind"}function s2(T){var F=T.getExtension("ANGLE_instanced_arrays");if(F)return T.vertexAttribDivisor=function(G,ee){F.vertexAttribDivisorANGLE(G,ee)},T.drawArraysInstanced=function(G,ee,ge,Ae){F.drawArraysInstancedANGLE(G,ee,ge,Ae)},T.drawElementsInstanced=function(G,ee,ge,Ae,Te){F.drawElementsInstancedANGLE(G,ee,ge,Ae,Te)},1}function i2(T){var F=T.getExtension("OES_vertex_array_object");if(F)return T.createVertexArray=function(){return F.createVertexArrayOES()},T.deleteVertexArray=function(G){F.deleteVertexArrayOES(G)},T.bindVertexArray=function(G){F.bindVertexArrayOES(G)},T.isVertexArray=function(G){return F.isVertexArrayOES(G)},1}function o2(T){var F=T.getExtension("WEBGL_draw_buffers");if(F)return T.drawBuffers=function(G,ee){F.drawBuffersWEBGL(G,ee)},1}function l2(T){return!!(T.multiDrawWebgl=T.getExtension("WEBGL_multi_draw"))}var Ft={counter:1,buffers:[],programs:[],framebuffers:[],renderbuffers:[],textures:[],shaders:[],vaos:[],contexts:{},offscreenCanvases:{},queries:[],stringCache:{},unpackAlignment:4,recordError:function(T){Ft.lastError||(Ft.lastError=T)},getNewId:function(T){for(var F=Ft.counter++,G=T.length;G<F;G++)T[G]=null;return F},getSource:function(T,F,G,ee){for(var ge="",Ae=0;Ae<F;++Ae){var Te=ee?u()[ee+Ae*4>>2]:-1;ge+=In(u()[G+Ae*4>>2],Te<0?void 0:Te)}return ge},createContext:function(T,F){T.getContextSafariWebGL2Fixed||(T.getContextSafariWebGL2Fixed=T.getContext,T.getContext=function(ge,Ae){var Te=T.getContextSafariWebGL2Fixed(ge,Ae);return ge=="webgl"==Te instanceof WebGLRenderingContext?Te:null});var G=T.getContext("webgl",F);if(!G)return 0;var ee=Ft.registerContext(G,F);return ee},registerContext:function(T,F){var G=X2(8);u()[G+4>>2]=Lc();var ee={handle:G,attributes:F,version:F.majorVersion,GLctx:T};return T.canvas&&(T.canvas.GLctxObject=ee),Ft.contexts[G]=ee,(typeof F.enableExtensionsByDefault=="undefined"||F.enableExtensionsByDefault)&&Ft.initExtensions(ee),G},makeContextCurrent:function(T){return Ft.currentContext=Ft.contexts[T],p.ctx=Pc=Ft.currentContext&&Ft.currentContext.GLctx,!(T&&!Pc)},getContext:function(T){return Ft.contexts[T]},deleteContext:function(T){Ft.currentContext===Ft.contexts[T]&&(Ft.currentContext=null),typeof qe=="object"&&qe.removeAllHandlersOnTarget(Ft.contexts[T].GLctx.canvas),Ft.contexts[T]&&Ft.contexts[T].GLctx.canvas&&(Ft.contexts[T].GLctx.canvas.GLctxObject=void 0),C3(Ft.contexts[T].handle),Ft.contexts[T]=null},initExtensions:function(T){if(T||(T=Ft.currentContext),!T.initExtensionsDone){T.initExtensionsDone=!0;var F=T.GLctx;s2(F),i2(F),o2(F),F.disjointTimerQueryExt=F.getExtension("EXT_disjoint_timer_query"),l2(F);var G=F.getSupportedExtensions()||[];G.forEach(function(ee){!ee.includes("lose_context")&&!ee.includes("debug")&&F.getExtension(ee)})}}},u2=["default","low-power","high-performance"];function d2(T,F){var G=F>>2,ee=u()[G+6],ge={alpha:!!u()[G+0],depth:!!u()[G+1],stencil:!!u()[G+2],antialias:!!u()[G+3],premultipliedAlpha:!!u()[G+4],preserveDrawingBuffer:!!u()[G+5],powerPreference:u2[ee],failIfMajorPerformanceCaveat:!!u()[G+7],majorVersion:u()[G+8],minorVersion:u()[G+9],enableExtensionsByDefault:u()[G+10],explicitSwapControl:u()[G+11],proxyContextToMainThread:u()[G+12],renderViaOffscreenBackBuffer:u()[G+13]},Ae=lp(T);if(!Ae||ge.explicitSwapControl)return 0;var Te=Ft.createContext(Ae,ge);return Te}function p2(T,F){return d2(T,F)}var ru={mappings:{},buffers:[null,[],[]],printChar:function(T,F){var G=ru.buffers[T];F===0||F===10?((T===1?re:Y)(kn(G,0)),G.length=0):G.push(F)},varargs:void 0,get:function(){ru.varargs+=4;var T=u()[ru.varargs-4>>2];return T},getStr:function(T){var F=In(T);return F},get64:function(T,F){return T}};function $c(T){return N?mo(3,1,T):0}function Mc(T,F,G,ee,ge){if(N)return mo(4,1,T,F,G,ee,ge)}function Fc(T,F,G,ee){if(N)return mo(5,1,T,F,G,ee);for(var ge=0,Ae=0;Ae<G;Ae++){var Te=u()[F>>2],_e=u()[F+4>>2];F+=8;for(var Pt=0;Pt<_e;Pt++)ru.printChar(T,i()[Te+Pt]);ge+=_e}return u()[ee>>2]=ge,0}function h2(T){De(T)}ze.init();var Pc,c2=[null,Cc,Rc,$c,Mc,Fc],k3=!1,_c={__clock_gettime:D1,__emscripten_init_main_thread_js:L1,__emscripten_thread_cleanup:B1,__pthread_create_js:W1,_emscripten_default_pthread_stack_size:V1,_emscripten_notify_thread_queue:U1,abort:G1,emscripten_check_blocking_allowed:j1,emscripten_get_heap_max:Nc,emscripten_get_now:fo,emscripten_memcpy_big:H1,emscripten_num_logical_cores:q1,emscripten_receive_on_main_thread_js:X1,emscripten_resize_heap:Z1,emscripten_set_canvas_element_size:n2,emscripten_unwind_to_js_event_loop:a2,emscripten_webgl_create_context:p2,exit:Sc,fd_close:$c,fd_seek:Mc,fd_write:Fc,memory:Fe||p.wasmMemory,setTempRet0:h2},I3=T1(),f2=p.___wasm_call_ctors=function(){return(f2=p.___wasm_call_ctors=p.asm.__wasm_call_ctors).apply(null,arguments)},m2=p._init=function(){return(m2=p._init=p.asm.init).apply(null,arguments)},g2=p._init_with_threads_count=function(){return(g2=p._init_with_threads_count=p.asm.init_with_threads_count).apply(null,arguments)},y2=p._get_threads_count=function(){return(y2=p._get_threads_count=p.asm.get_threads_count).apply(null,arguments)},A2=p._register_tensor=function(){return(A2=p._register_tensor=p.asm.register_tensor).apply(null,arguments)},x2=p._dispose_data=function(){return(x2=p._dispose_data=p.asm.dispose_data).apply(null,arguments)},b2=p._dispose=function(){return(b2=p._dispose=p.asm.dispose).apply(null,arguments)},v2=p._Abs=function(){return(v2=p._Abs=p.asm.Abs).apply(null,arguments)},w2=p._Add=function(){return(w2=p._Add=p.asm.Add).apply(null,arguments)},k2=p._AddN=function(){return(k2=p._AddN=p.asm.AddN).apply(null,arguments)},I2=p._All=function(){return(I2=p._All=p.asm.All).apply(null,arguments)},S2=p._Any=function(){return(S2=p._Any=p.asm.Any).apply(null,arguments)},C2=p._ArgMax=function(){return(C2=p._ArgMax=p.asm.ArgMax).apply(null,arguments)},T2=p._AvgPool=function(){return(T2=p._AvgPool=p.asm.AvgPool).apply(null,arguments)},N2=p._BatchMatMul=function(){return(N2=p._BatchMatMul=p.asm.BatchMatMul).apply(null,arguments)},E2=p._Ceil=function(){return(E2=p._Ceil=p.asm.Ceil).apply(null,arguments)},R2=p._ClipByValue=function(){return(R2=p._ClipByValue=p.asm.ClipByValue).apply(null,arguments)},$2=p._Conv2D=function(){return($2=p._Conv2D=p.asm.Conv2D).apply(null,arguments)},M2=p._Conv2DBackpropInput=function(){return(M2=p._Conv2DBackpropInput=p.asm.Conv2DBackpropInput).apply(null,arguments)},F2=p._Cos=function(){return(F2=p._Cos=p.asm.Cos).apply(null,arguments)},P2=p._Cosh=function(){return(P2=p._Cosh=p.asm.Cosh).apply(null,arguments)},_2=p._CropAndResize=function(){return(_2=p._CropAndResize=p.asm.CropAndResize).apply(null,arguments)},z2=p._Cumprod=function(){return(z2=p._Cumprod=p.asm.Cumprod).apply(null,arguments)},O2=p._Cumsum=function(){return(O2=p._Cumsum=p.asm.Cumsum).apply(null,arguments)},D2=p._DepthToSpace=function(){return(D2=p._DepthToSpace=p.asm.DepthToSpace).apply(null,arguments)},L2=p._DepthwiseConv2dNative=function(){return(L2=p._DepthwiseConv2dNative=p.asm.DepthwiseConv2dNative).apply(null,arguments)},B2=p._Elu=function(){return(B2=p._Elu=p.asm.Elu).apply(null,arguments)},W2=p._Equal=function(){return(W2=p._Equal=p.asm.Equal).apply(null,arguments)},V2=p._Exp=function(){return(V2=p._Exp=p.asm.Exp).apply(null,arguments)},U2=p._FlipLeftRight=function(){return(U2=p._FlipLeftRight=p.asm.FlipLeftRight).apply(null,arguments)},zc=p._Floor=function(){return(zc=p._Floor=p.asm.Floor).apply(null,arguments)},Oc=p._FloorDiv=function(){return(Oc=p._FloorDiv=p.asm.FloorDiv).apply(null,arguments)},up=p._FusedBatchNorm=function(){return(up=p._FusedBatchNorm=p.asm.FusedBatchNorm).apply(null,arguments)},G2=p._FusedConv2D=function(){return(G2=p._FusedConv2D=p.asm.FusedConv2D).apply(null,arguments)},j2=p._FusedDepthwiseConv2D=function(){return(j2=p._FusedDepthwiseConv2D=p.asm.FusedDepthwiseConv2D).apply(null,arguments)},nu=p._Gather=function(){return(nu=p._Gather=p.asm.Gather).apply(null,arguments)},dp=p._GatherNd=function(){return(dp=p._GatherNd=p.asm.GatherNd).apply(null,arguments)},pp=p._Greater=function(){return(pp=p._Greater=p.asm.Greater).apply(null,arguments)},S3=p._GreaterEqual=function(){return(S3=p._GreaterEqual=p.asm.GreaterEqual).apply(null,arguments)},au=p._LeakyRelu=function(){return(au=p._LeakyRelu=p.asm.LeakyRelu).apply(null,arguments)},su=p._Less=function(){return(su=p._Less=p.asm.Less).apply(null,arguments)},H2=p._LessEqual=function(){return(H2=p._LessEqual=p.asm.LessEqual).apply(null,arguments)},j=p._Log=function(){return(j=p._Log=p.asm.Log).apply(null,arguments)},ne=p._LogicalAnd=function(){return(ne=p._LogicalAnd=p.asm.LogicalAnd).apply(null,arguments)},ye=p._Max=function(){return(ye=p._Max=p.asm.Max).apply(null,arguments)},Re=p._MaxPool=function(){return(Re=p._MaxPool=p.asm.MaxPool).apply(null,arguments)},lt=p._Maximum=function(){return(lt=p._Maximum=p.asm.Maximum).apply(null,arguments)},ct=p._Mean=function(){return(ct=p._Mean=p.asm.Mean).apply(null,arguments)},Ke=p._Min=function(){return(Ke=p._Min=p.asm.Min).apply(null,arguments)},He=p._Minimum=function(){return(He=p._Minimum=p.asm.Minimum).apply(null,arguments)},qt=p._MirrorPad=function(){return(qt=p._MirrorPad=p.asm.MirrorPad).apply(null,arguments)},aa=p._Multiply=function(){return(aa=p._Multiply=p.asm.Multiply).apply(null,arguments)},sa=p._Neg=function(){return(sa=p._Neg=p.asm.Neg).apply(null,arguments)},iu=p._NonMaxSuppressionV3=function(){return(iu=p._NonMaxSuppressionV3=p.asm.NonMaxSuppressionV3).apply(null,arguments)},go=p._NonMaxSuppressionV4=function(){return(go=p._NonMaxSuppressionV4=p.asm.NonMaxSuppressionV4).apply(null,arguments)},q2=p._NonMaxSuppressionV5=function(){return(q2=p._NonMaxSuppressionV5=p.asm.NonMaxSuppressionV5).apply(null,arguments)},en=p._NotEqual=function(){return(en=p._NotEqual=p.asm.NotEqual).apply(null,arguments)},bs=p._OneHot=function(){return(bs=p._OneHot=p.asm.OneHot).apply(null,arguments)},Dc=p._PadV2=function(){return(Dc=p._PadV2=p.asm.PadV2).apply(null,arguments)},QN=p._Pow=function(){return(QN=p._Pow=p.asm.Pow).apply(null,arguments)},eE=p._Prelu=function(){return(eE=p._Prelu=p.asm.Prelu).apply(null,arguments)},tE=p._Prod=function(){return(tE=p._Prod=p.asm.Prod).apply(null,arguments)},rE=p._RealDiv=function(){return(rE=p._RealDiv=p.asm.RealDiv).apply(null,arguments)},nE=p._Relu=function(){return(nE=p._Relu=p.asm.Relu).apply(null,arguments)},aE=p._Relu6=function(){return(aE=p._Relu6=p.asm.Relu6).apply(null,arguments)},sE=p._ResizeBilinear=function(){return(sE=p._ResizeBilinear=p.asm.ResizeBilinear).apply(null,arguments)},iE=p._Reverse=function(){return(iE=p._Reverse=p.asm.Reverse).apply(null,arguments)},oE=p._RotateWithOffset=function(){return(oE=p._RotateWithOffset=p.asm.RotateWithOffset).apply(null,arguments)},lE=p._Round=function(){return(lE=p._Round=p.asm.Round).apply(null,arguments)},uE=p._Rsqrt=function(){return(uE=p._Rsqrt=p.asm.Rsqrt).apply(null,arguments)},dE=p._ScatterNd=function(){return(dE=p._ScatterNd=p.asm.ScatterNd).apply(null,arguments)},pE=p._SelectV2=function(){return(pE=p._SelectV2=p.asm.SelectV2).apply(null,arguments)},hE=p._Sigmoid=function(){return(hE=p._Sigmoid=p.asm.Sigmoid).apply(null,arguments)},cE=p._Sin=function(){return(cE=p._Sin=p.asm.Sin).apply(null,arguments)},fE=p._Softmax=function(){return(fE=p._Softmax=p.asm.Softmax).apply(null,arguments)},mE=p._SparseFillEmptyRows=function(){return(mE=p._SparseFillEmptyRows=p.asm.SparseFillEmptyRows).apply(null,arguments)},gE=p._SparseReshape=function(){return(gE=p._SparseReshape=p.asm.SparseReshape).apply(null,arguments)},yE=p._SparseSegmentReduction=function(){return(yE=p._SparseSegmentReduction=p.asm.SparseSegmentReduction).apply(null,arguments)},AE=p._Sqrt=function(){return(AE=p._Sqrt=p.asm.Sqrt).apply(null,arguments)},xE=p._Square=function(){return(xE=p._Square=p.asm.Square).apply(null,arguments)},bE=p._SquaredDifference=function(){return(bE=p._SquaredDifference=p.asm.SquaredDifference).apply(null,arguments)},vE=p._Step=function(){return(vE=p._Step=p.asm.Step).apply(null,arguments)},wE=p._StridedSlice=function(){return(wE=p._StridedSlice=p.asm.StridedSlice).apply(null,arguments)},kE=p._Sub=function(){return(kE=p._Sub=p.asm.Sub).apply(null,arguments)},IE=p._Sum=function(){return(IE=p._Sum=p.asm.Sum).apply(null,arguments)},SE=p._Tan=function(){return(SE=p._Tan=p.asm.Tan).apply(null,arguments)},CE=p._Tanh=function(){return(CE=p._Tanh=p.asm.Tanh).apply(null,arguments)},TE=p._Tile=function(){return(TE=p._Tile=p.asm.Tile).apply(null,arguments)},NE=p._TopK=function(){return(NE=p._TopK=p.asm.TopK).apply(null,arguments)},EE=p._Transform=function(){return(EE=p._Transform=p.asm.Transform).apply(null,arguments)},RE=p._Transpose=function(){return(RE=p._Transpose=p.asm.Transpose).apply(null,arguments)},$E=p.__FusedMatMul=function(){return($E=p.__FusedMatMul=p.asm._FusedMatMul).apply(null,arguments)},X2=p._malloc=function(){return(X2=p._malloc=p.asm.malloc).apply(null,arguments)},C3=p._free=function(){return(C3=p._free=p.asm.free).apply(null,arguments)},ME=p._emscripten_tls_init=function(){return(ME=p._emscripten_tls_init=p.asm.emscripten_tls_init).apply(null,arguments)},T3=p.___errno_location=function(){return(T3=p.___errno_location=p.asm.__errno_location).apply(null,arguments)},Lc=p._pthread_self=function(){return(Lc=p._pthread_self=p.asm.pthread_self).apply(null,arguments)},N3=p._emscripten_main_thread_process_queued_calls=function(){return(N3=p._emscripten_main_thread_process_queued_calls=p.asm.emscripten_main_thread_process_queued_calls).apply(null,arguments)},FE=p.__emscripten_thread_crashed=function(){return(FE=p.__emscripten_thread_crashed=p.asm._emscripten_thread_crashed).apply(null,arguments)},E3=p.__emscripten_thread_init=function(){return(E3=p.__emscripten_thread_init=p.asm._emscripten_thread_init).apply(null,arguments)},PE=p._emscripten_current_thread_process_queued_calls=function(){return(PE=p._emscripten_current_thread_process_queued_calls=p.asm.emscripten_current_thread_process_queued_calls).apply(null,arguments)},_E=p._emscripten_main_browser_thread_id=function(){return(_E=p._emscripten_main_browser_thread_id=p.asm.emscripten_main_browser_thread_id).apply(null,arguments)},zE=p._emscripten_sync_run_in_main_thread_2=function(){return(zE=p._emscripten_sync_run_in_main_thread_2=p.asm.emscripten_sync_run_in_main_thread_2).apply(null,arguments)},R3=p._emscripten_sync_run_in_main_thread_4=function(){return(R3=p._emscripten_sync_run_in_main_thread_4=p.asm.emscripten_sync_run_in_main_thread_4).apply(null,arguments)},$3=p._emscripten_run_in_main_runtime_thread_js=function(){return($3=p._emscripten_run_in_main_runtime_thread_js=p.asm.emscripten_run_in_main_runtime_thread_js).apply(null,arguments)},K2=p._emscripten_dispatch_to_thread_=function(){return(K2=p._emscripten_dispatch_to_thread_=p.asm.emscripten_dispatch_to_thread_).apply(null,arguments)},Z2=p.__emscripten_thread_free_data=function(){return(Z2=p.__emscripten_thread_free_data=p.asm._emscripten_thread_free_data).apply(null,arguments)},OE=p.__emscripten_thread_exit=function(){return(OE=p.__emscripten_thread_exit=p.asm._emscripten_thread_exit).apply(null,arguments)},DE=p._memalign=function(){return(DE=p._memalign=p.asm.memalign).apply(null,arguments)},M3=p._emscripten_stack_set_limits=function(){return(M3=p._emscripten_stack_set_limits=p.asm.emscripten_stack_set_limits).apply(null,arguments)},Y2=p.stackSave=function(){return(Y2=p.stackSave=p.asm.stackSave).apply(null,arguments)},Bc=p.stackRestore=function(){return(Bc=p.stackRestore=p.asm.stackRestore).apply(null,arguments)},ou=p.stackAlloc=function(){return(ou=p.stackAlloc=p.asm.stackAlloc).apply(null,arguments)},LE=p.dynCall_iijjiiii=function(){return(LE=p.dynCall_iijjiiii=p.asm.dynCall_iijjiiii).apply(null,arguments)},BE=p.dynCall_jiji=function(){return(BE=p.dynCall_jiji=p.asm.dynCall_jiji).apply(null,arguments)},F3=p.__emscripten_allow_main_runtime_queued_calls=21456;p.cwrap=fr,p.keepRuntimeAlive=ho,p.PThread=ze,p.PThread=ze,p.wasmMemory=Fe,p.ExitStatus=hp;var Wc;function hp(T){this.name="ExitStatus",this.message="Program terminated with exit("+T+")",this.status=T}na=function T(){Wc||J2(),Wc||(na=T)};function J2(T){if(T=T||y,xs>0)return;if(N){c(p),ap(),postMessage({cmd:"loaded"});return}if(zr(),xs>0)return;function F(){Wc||(Wc=!0,p.calledRun=!0,!yt&&(ap(),c(p),p.onRuntimeInitialized&&p.onRuntimeInitialized(),S1()))}p.setStatus?(p.setStatus("Running..."),setTimeout(function(){setTimeout(function(){p.setStatus("")},1),F()},1)):F()}p.run=J2;function WE(T,F){if(Pr=T,!F&&N)throw Cc(T),"unwind";ho()||I1(),VE(T)}function VE(T){Pr=T,ho()||(ze.terminateAllThreads(),p.onExit&&p.onExit(T),yt=!0),x(T,new hp(T))}if(p.preInit)for(typeof p.preInit=="function"&&(p.preInit=[p.preInit]);p.preInit.length>0;)p.preInit.pop()();J2();var Vc;m&&(Vc={uncaughtException:process.listeners("uncaughtException").filter(function(T){return!m.uncaughtException.indexOf(T)>-1}),unhandledRejection:process.listeners("unhandledRejection").filter(function(T){return!m.unhandledRejection.indexOf(T)>-1})});var Uc;if(typeof WasmBackendModule!="undefined")Uc=WasmBackendModule;else if(typeof a!="undefined")Uc=a;else throw new Error("Could not find wasm module in post.js");if(Vc){var UE=Uc._dispose;Uc._dispose=function(){UE(),Vc.uncaughtException.forEach(function(T){process.removeListener("uncaughtException",T)}),Vc.unhandledRejection.forEach(function(T){process.removeListener("unhandledRejection",T)})}}return a.ready}})();typeof e=="object"&&typeof t=="object"?t.exports=r:typeof define=="function"&&define.amd?define([],function(){return r}):typeof e=="object"&&(e.WasmBackendModuleThreadedSimd=r)}}),AR=pr({"src/tfjs-backend-wasm/wasm-out/tfjs-backend-wasm.js"(e,t){var r=(()=>{var n=typeof document!="undefined"&&document.currentScript?document.currentScript.src:void 0;return typeof __filename!="undefined"&&(n=n||__filename),function(a){a=a||{};var s=typeof a!="undefined"?a:{},i,o;s.ready=new Promise(function(j,ne){i=j,o=ne});var l;typeof process!="undefined"&&process.listeners&&(l={uncaughtException:process.listeners("uncaughtException"),unhandledRejection:process.listeners("unhandledRejection")});var u=Object.assign({},s),d=[],h="./this.program",p=(j,ne)=>{throw ne},c=typeof window=="object",f=typeof importScripts=="function",m=typeof process=="object"&&typeof process.versions=="object"&&typeof process.versions.node=="string",g="";function y(j){return s.locateFile?s.locateFile(j,g):g+j}var A,x,b,w;function S(j){j instanceof dp||M("exiting due to exception: "+j)}var N,E,R;m?(f?g=g0().dirname(g)+"/":g=__dirname+"/",R=()=>{E||(N=fy(),E=g0())},A=function(j,ne){return R(),j=E.normalize(j),N.readFileSync(j,ne?void 0:"utf8")},b=j=>{var ne=A(j,!0);return ne.buffer||(ne=new Uint8Array(ne)),ne},x=(j,ne,ye)=>{R(),j=E.normalize(j),N.readFile(j,function(Re,lt){Re?ye(Re):ne(lt.buffer)})},process.argv.length>1&&(h=process.argv[1].replace(/\\/g,"/")),d=process.argv.slice(2),process.on("uncaughtException",function(j){if(!(j instanceof dp))throw j}),process.on("unhandledRejection",function(j){throw j}),p=(j,ne)=>{if(np())throw process.exitCode=j,ne;S(ne),process.exit(j)},s.inspect=function(){return"[Emscripten Module object]"}):(c||f)&&(f?g=self.location.href:typeof document!="undefined"&&document.currentScript&&(g=document.currentScript.src),n&&(g=n),g.indexOf("blob:")!==0?g=g.substr(0,g.replace(/[?#].*/,"").lastIndexOf("/")+1):g="",A=j=>{var ne=new XMLHttpRequest;return ne.open("GET",j,!1),ne.send(null),ne.responseText},f&&(b=j=>{var ne=new XMLHttpRequest;return ne.open("GET",j,!1),ne.responseType="arraybuffer",ne.send(null),new Uint8Array(ne.response)}),x=(j,ne,ye)=>{var Re=new XMLHttpRequest;Re.open("GET",j,!0),Re.responseType="arraybuffer",Re.onload=()=>{if(Re.status==200||Re.status==0&&Re.response){ne(Re.response);return}ye()},Re.onerror=ye,Re.send(null)},w=j=>document.title=j);var z=s.print||console.log.bind(console),M=s.printErr||console.warn.bind(console);Object.assign(s,u),u=null,s.arguments&&(d=s.arguments),s.thisProgram&&(h=s.thisProgram),s.quit&&(p=s.quit);var I=4;function _(j){_.shown||(_.shown={}),_.shown[j]||(_.shown[j]=1,M(j))}function O(j,ne){if(typeof WebAssembly.Function=="function"){for(var ye={i:"i32",j:"i64",f:"f32",d:"f64"},Re={parameters:[],results:ne[0]=="v"?[]:[ye[ne[0]]]},lt=1;lt<ne.length;++lt)Re.parameters.push(ye[ne[lt]]);return new WebAssembly.Function(Re,j)}var ct=[1,0,1,96],Ke=ne.slice(0,1),He=ne.slice(1),qt={i:127,j:126,f:125,d:124};ct.push(He.length);for(var lt=0;lt<He.length;++lt)ct.push(qt[He[lt]]);Ke=="v"?ct.push(0):ct=ct.concat([1,qt[Ke]]),ct[1]=ct.length-2;var aa=new Uint8Array([0,97,115,109,1,0,0,0].concat(ct,[2,7,1,1,101,1,102,0,0,7,5,1,1,102,0,0])),sa=new WebAssembly.Module(aa),iu=new WebAssembly.Instance(sa,{e:{f:j}}),go=iu.exports.f;return go}var q=[],K;function D(){if(q.length)return q.pop();try{ys.grow(1)}catch(j){throw j instanceof RangeError?"Unable to grow wasm table. Set ALLOW_TABLE_GROWTH.":j}return ys.length-1}function Q(j,ne){for(var ye=j;ye<j+ne;ye++){var Re=ip(ye);Re&&K.set(Re,ye)}}var V=0,re=j=>{V=j},Y;s.wasmBinary&&(Y=s.wasmBinary);var ie=s.noExitRuntime||!0;typeof WebAssembly!="object"&&po("no native wasm support detected");var J,ae=!1,de;function xe(j,ne){j||po(ne)}function be(j){var ne=s["_"+j];return ne}function Ee(j,ne,ye,Re,lt){var ct={string:function(en){var bs=0;if(en!=null&&en!==0){var Dc=(en.length<<2)+1;bs=up(Dc),ht(en,bs,Dc)}return bs},array:function(en){var bs=up(en.length);return yt(en,bs),bs}};function Ke(en){return ne==="string"?ot(en):ne==="boolean"?Boolean(en):en}var He=be(j),qt=[],aa=0;if(Re)for(var sa=0;sa<Re.length;sa++){var iu=ct[ye[sa]];iu?(aa===0&&(aa=zc()),qt[sa]=iu(Re[sa])):qt[sa]=Re[sa]}var go=He.apply(null,qt);function q2(en){return aa!==0&&Oc(aa),Ke(en)}return go=q2(go),go}function $e(j,ne,ye,Re){ye=ye||[];var lt=ye.every(function(Ke){return Ke==="number"}),ct=ne!=="string";return ct&<&&!Re?be(j):function(){return Ee(j,ne,ye,arguments,Re)}}var De=1,Be=typeof TextDecoder!="undefined"?new TextDecoder("utf8"):void 0;function Xe(j,ne,ye){for(var Re=ne+ye,lt=ne;j[lt]&&!(lt>=Re);)++lt;if(lt-ne>16&&j.subarray&&Be)return Be.decode(j.subarray(ne,lt));for(var ct="";ne<lt;){var Ke=j[ne++];if(!(Ke&128)){ct+=String.fromCharCode(Ke);continue}var He=j[ne++]&63;if((Ke&224)==192){ct+=String.fromCharCode((Ke&31)<<6|He);continue}var qt=j[ne++]&63;if((Ke&240)==224?Ke=(Ke&15)<<12|He<<6|qt:Ke=(Ke&7)<<18|He<<12|qt<<6|j[ne++]&63,Ke<65536)ct+=String.fromCharCode(Ke);else{var aa=Ke-65536;ct+=String.fromCharCode(55296|aa>>10,56320|aa&1023)}}return ct}function ot(j,ne){return j?Xe(fr,j,ne):""}function pt(j,ne,ye,Re){if(!(Re>0))return 0;for(var lt=ye,ct=ye+Re-1,Ke=0;Ke<j.length;++Ke){var He=j.charCodeAt(Ke);if(He>=55296&&He<=57343){var qt=j.charCodeAt(++Ke);He=65536+((He&1023)<<10)|qt&1023}if(He<=127){if(ye>=ct)break;ne[ye++]=He}else if(He<=2047){if(ye+1>=ct)break;ne[ye++]=192|He>>6,ne[ye++]=128|He&63}else if(He<=65535){if(ye+2>=ct)break;ne[ye++]=224|He>>12,ne[ye++]=128|He>>6&63,ne[ye++]=128|He&63}else{if(ye+3>=ct)break;ne[ye++]=240|He>>18,ne[ye++]=128|He>>12&63,ne[ye++]=128|He>>6&63,ne[ye++]=128|He&63}}return ne[ye]=0,ye-lt}function ht(j,ne,ye){return pt(j,fr,ne,ye)}function Fe(j){for(var ne=0,ye=0;ye<j.length;++ye){var Re=j.charCodeAt(ye);Re>=55296&&Re<=57343&&(Re=65536+((Re&1023)<<10)|j.charCodeAt(++ye)&1023),Re<=127?++ne:Re<=2047?ne+=2:Re<=65535?ne+=3:ne+=4}return ne}var wt=typeof TextDecoder!="undefined"?new TextDecoder("utf-16le"):void 0;function yt(j,ne){rr.set(j,ne)}function Pr(j,ne,ye){for(var Re=0;Re<j.length;++Re)rr[ne++>>0]=j.charCodeAt(Re);ye||(rr[ne>>0]=0)}function cr(j,ne){return j%ne>0&&(j+=ne-j%ne),j}var Jr,rr,fr,ta,Qr,nr,kn,In,gs;function uo(j){Jr=j,s.HEAP8=rr=new Int8Array(j),s.HEAP16=ta=new Int16Array(j),s.HEAP32=nr=new Int32Array(j),s.HEAPU8=fr=new Uint8Array(j),s.HEAPU16=Qr=new Uint16Array(j),s.HEAPU32=kn=new Uint32Array(j),s.HEAPF32=In=new Float32Array(j),s.HEAPF64=gs=new Float64Array(j)}var uc=s.INITIAL_MEMORY||16777216,ys,Ua=[],rp=[],Zl=[],_r=!1,dc=!1,pc=0;function np(){return ie||pc>0}function hc(){if(s.preRun)for(typeof s.preRun=="function"&&(s.preRun=[s.preRun]);s.preRun.length;)mc(s.preRun.shift());sp(Ua)}function cc(){_r=!0,sp(rp)}function f3(){dc=!0}function fc(){if(s.postRun)for(typeof s.postRun=="function"&&(s.postRun=[s.postRun]);s.postRun.length;)gc(s.postRun.shift());sp(Zl)}function mc(j){Ua.unshift(j)}function ra(j){rp.unshift(j)}function gc(j){Zl.unshift(j)}var Sn=0,Yl=null,As=null;function w1(j){Sn++,s.monitorRunDependencies&&s.monitorRunDependencies(Sn)}function yc(j){if(Sn--,s.monitorRunDependencies&&s.monitorRunDependencies(Sn),Sn==0&&(Yl!==null&&(clearInterval(Yl),Yl=null),As)){var ne=As;As=null,ne()}}s.preloadedImages={},s.preloadedAudios={};function po(j){s.onAbort&&s.onAbort(j),j="Aborted("+j+")",M(j),ae=!0,de=1,j+=". Build with -s ASSERTIONS=1 for more info.";var ne=new WebAssembly.RuntimeError(j);throw o(ne),ne}var k1="data:application/octet-stream;base64,";function Ac(j){return j.startsWith(k1)}function ho(j){return j.startsWith("file://")}var zr;zr="tfjs-backend-wasm.wasm",Ac(zr)||(zr=y(zr));function ap(j){try{if(j==zr&&Y)return new Uint8Array(Y);if(b)return b(j);throw"both async and sync fetching of the wasm failed"}catch(ne){po(ne)}}function I1(){if(!Y&&(c||f)){if(typeof fetch=="function"&&!ho(zr))return fetch(zr,{credentials:"same-origin"}).then(function(j){if(!j.ok)throw"failed to load wasm binary file at '"+zr+"'";return j.arrayBuffer()}).catch(function(){return ap(zr)});if(x)return new Promise(function(j,ne){x(zr,function(ye){j(new Uint8Array(ye))},ne)})}return Promise.resolve().then(function(){return ap(zr)})}function S1(){var j={env:eu,wasi_snapshot_preview1:eu};function ne(Ke,He){var qt=Ke.exports;s.asm=qt,J=s.asm.memory,uo(J.buffer),ys=s.asm.__indirect_function_table,ra(s.asm.__wasm_call_ctors),yc("wasm-instantiate")}w1("wasm-instantiate");function ye(Ke){ne(Ke.instance)}function Re(Ke){return I1().then(function(He){return WebAssembly.instantiate(He,j)}).then(function(He){return He}).then(Ke,function(He){M("failed to asynchronously prepare wasm: "+He),po(He)})}function lt(){return!Y&&typeof WebAssembly.instantiateStreaming=="function"&&!Ac(zr)&&!ho(zr)&&typeof fetch=="function"?fetch(zr,{credentials:"same-origin"}).then(function(Ke){var He=WebAssembly.instantiateStreaming(Ke,j);return He.then(ye,function(qt){return M("wasm streaming compile failed: "+qt),M("falling back to ArrayBuffer instantiation"),Re(ye)})}):Re(ye)}if(s.instantiateWasm)try{var ct=s.instantiateWasm(j,ne);return ct}catch(Ke){return M("Module.instantiateWasm callback failed with error: "+Ke),!1}return lt().catch(o),{}}var m3,g3;function sp(j){for(;j.length>0;){var ne=j.shift();if(typeof ne=="function"){ne(s);continue}var ye=ne.func;typeof ye=="number"?ne.arg===void 0?ip(ye)():ip(ye)(ne.arg):ye(ne.arg===void 0?null:ne.arg)}}function xs(j){return j}function xc(j){var ne=/\b_Z[\w\d_]+/g;return j.replace(ne,function(ye){var Re=ye;return ye===Re?ye:Re+" ["+ye+"]"})}var na=[];function ip(j){var ne=na[j];return ne||(j>=na.length&&(na.length=j+1),na[j]=ne=ys.get(j)),ne}function y3(){var j=new Error;if(!j.stack){try{throw new Error}catch(ne){j=ne}if(!j.stack)return"(no stack trace available)"}return j.stack.toString()}function Jl(j,ne){ys.set(j,ne),na[j]=ne}function C1(){po("")}function bc(j,ne,ye){fr.copyWithin(j,ne,ne+ye)}function vc(){return 2147483648}function Or(j){try{return J.grow(j-Jr.byteLength+65535>>>16),uo(J.buffer),1}catch(ne){}}function wc(j){var ne=fr.length;j=j>>>0;var ye=vc();if(j>ye)return!1;for(var Re=1;Re<=4;Re*=2){var lt=ne*(1+.2/Re);lt=Math.min(lt,j+100663296);var ct=Math.min(ye,cr(Math.max(j,lt),65536)),Ke=Or(ct);if(Ke)return!0}return!1}var Ql={mappings:{},buffers:[null,[],[]],printChar:function(j,ne){var ye=Ql.buffers[j];ne===0||ne===10?((j===1?z:M)(Xe(ye,0)),ye.length=0):ye.push(ne)},varargs:void 0,get:function(){Ql.varargs+=4;var j=nr[Ql.varargs-4>>2];return j},getStr:function(j){var ne=ot(j);return ne},get64:function(j,ne){return j}};function T1(j){return 0}function A3(j,ne,ye,Re,lt){}function x3(j,ne,ye,Re){for(var lt=0,ct=0;ct<ye;ct++){var Ke=nr[ne>>2],He=nr[ne+4>>2];ne+=8;for(var qt=0;qt<He;qt++)Ql.printChar(j,fr[Ke+qt]);lt+=He}return nr[Re>>2]=lt,0}function N1(j){re(j)}var kc=!1,eu={abort:C1,emscripten_memcpy_big:bc,emscripten_resize_heap:wc,fd_close:T1,fd_seek:A3,fd_write:x3,setTempRet0:N1},JN=S1(),b3=s.___wasm_call_ctors=function(){return(b3=s.___wasm_call_ctors=s.asm.__wasm_call_ctors).apply(null,arguments)},E1=s._init=function(){return(E1=s._init=s.asm.init).apply(null,arguments)},R1=s._init_with_threads_count=function(){return(R1=s._init_with_threads_count=s.asm.init_with_threads_count).apply(null,arguments)},Ic=s._get_threads_count=function(){return(Ic=s._get_threads_count=s.asm.get_threads_count).apply(null,arguments)},Sc=s._register_tensor=function(){return(Sc=s._register_tensor=s.asm.register_tensor).apply(null,arguments)},$1=s._dispose_data=function(){return($1=s._dispose_data=s.asm.dispose_data).apply(null,arguments)},ze=s._dispose=function(){return(ze=s._dispose=s.asm.dispose).apply(null,arguments)},M1=s._Abs=function(){return(M1=s._Abs=s.asm.Abs).apply(null,arguments)},Cc=s._Add=function(){return(Cc=s._Add=s.asm.Add).apply(null,arguments)},co=s._AddN=function(){return(co=s._AddN=s.asm.AddN).apply(null,arguments)},tu=s._All=function(){return(tu=s._All=s.asm.All).apply(null,arguments)},F1=s._Any=function(){return(F1=s._Any=s.asm.Any).apply(null,arguments)},v3=s._ArgMax=function(){return(v3=s._ArgMax=s.asm.ArgMax).apply(null,arguments)},P1=s._AvgPool=function(){return(P1=s._AvgPool=s.asm.AvgPool).apply(null,arguments)},w3=s._BatchMatMul=function(){return(w3=s._BatchMatMul=s.asm.BatchMatMul).apply(null,arguments)},fo=s._Ceil=function(){return(fo=s._Ceil=s.asm.Ceil).apply(null,arguments)},_1=s._ClipByValue=function(){return(_1=s._ClipByValue=s.asm.ClipByValue).apply(null,arguments)},z1=s._Conv2D=function(){return(z1=s._Conv2D=s.asm.Conv2D).apply(null,arguments)},O1=s._Conv2DBackpropInput=function(){return(O1=s._Conv2DBackpropInput=s.asm.Conv2DBackpropInput).apply(null,arguments)},D1=s._Cos=function(){return(D1=s._Cos=s.asm.Cos).apply(null,arguments)},L1=s._Cosh=function(){return(L1=s._Cosh=s.asm.Cosh).apply(null,arguments)},B1=s._CropAndResize=function(){return(B1=s._CropAndResize=s.asm.CropAndResize).apply(null,arguments)},Tc=s._Cumprod=function(){return(Tc=s._Cumprod=s.asm.Cumprod).apply(null,arguments)},W1=s._Cumsum=function(){return(W1=s._Cumsum=s.asm.Cumsum).apply(null,arguments)},V1=s._DepthToSpace=function(){return(V1=s._DepthToSpace=s.asm.DepthToSpace).apply(null,arguments)},U1=s._DepthwiseConv2dNative=function(){return(U1=s._DepthwiseConv2dNative=s.asm.DepthwiseConv2dNative).apply(null,arguments)},G1=s._Elu=function(){return(G1=s._Elu=s.asm.Elu).apply(null,arguments)},j1=s._Equal=function(){return(j1=s._Equal=s.asm.Equal).apply(null,arguments)},Nc=s._Exp=function(){return(Nc=s._Exp=s.asm.Exp).apply(null,arguments)},H1=s._FlipLeftRight=function(){return(H1=s._FlipLeftRight=s.asm.FlipLeftRight).apply(null,arguments)},q1=s._Floor=function(){return(q1=s._Floor=s.asm.Floor).apply(null,arguments)},mo=s._FloorDiv=function(){return(mo=s._FloorDiv=s.asm.FloorDiv).apply(null,arguments)},op=s._FusedBatchNorm=function(){return(op=s._FusedBatchNorm=s.asm.FusedBatchNorm).apply(null,arguments)},X1=s._FusedConv2D=function(){return(X1=s._FusedConv2D=s.asm.FusedConv2D).apply(null,arguments)},K1=s._FusedDepthwiseConv2D=function(){return(K1=s._FusedDepthwiseConv2D=s.asm.FusedDepthwiseConv2D).apply(null,arguments)},Z1=s._Gather=function(){return(Z1=s._Gather=s.asm.Gather).apply(null,arguments)},qe=s._GatherNd=function(){return(qe=s._GatherNd=s.asm.GatherNd).apply(null,arguments)},Y1=s._Greater=function(){return(Y1=s._Greater=s.asm.Greater).apply(null,arguments)},J1=s._GreaterEqual=function(){return(J1=s._GreaterEqual=s.asm.GreaterEqual).apply(null,arguments)},Q1=s._LeakyRelu=function(){return(Q1=s._LeakyRelu=s.asm.LeakyRelu).apply(null,arguments)},e2=s._Less=function(){return(e2=s._Less=s.asm.Less).apply(null,arguments)},t2=s._LessEqual=function(){return(t2=s._LessEqual=s.asm.LessEqual).apply(null,arguments)},r2=s._Log=function(){return(r2=s._Log=s.asm.Log).apply(null,arguments)},lp=s._LogicalAnd=function(){return(lp=s._LogicalAnd=s.asm.LogicalAnd).apply(null,arguments)},Ec=s._Max=function(){return(Ec=s._Max=s.asm.Max).apply(null,arguments)},Rc=s._MaxPool=function(){return(Rc=s._MaxPool=s.asm.MaxPool).apply(null,arguments)},n2=s._Maximum=function(){return(n2=s._Maximum=s.asm.Maximum).apply(null,arguments)},a2=s._Mean=function(){return(a2=s._Mean=s.asm.Mean).apply(null,arguments)},s2=s._Min=function(){return(s2=s._Min=s.asm.Min).apply(null,arguments)},i2=s._Minimum=function(){return(i2=s._Minimum=s.asm.Minimum).apply(null,arguments)},o2=s._MirrorPad=function(){return(o2=s._MirrorPad=s.asm.MirrorPad).apply(null,arguments)},l2=s._Multiply=function(){return(l2=s._Multiply=s.asm.Multiply).apply(null,arguments)},Ft=s._Neg=function(){return(Ft=s._Neg=s.asm.Neg).apply(null,arguments)},u2=s._NonMaxSuppressionV3=function(){return(u2=s._NonMaxSuppressionV3=s.asm.NonMaxSuppressionV3).apply(null,arguments)},d2=s._NonMaxSuppressionV4=function(){return(d2=s._NonMaxSuppressionV4=s.asm.NonMaxSuppressionV4).apply(null,arguments)},p2=s._NonMaxSuppressionV5=function(){return(p2=s._NonMaxSuppressionV5=s.asm.NonMaxSuppressionV5).apply(null,arguments)},ru=s._NotEqual=function(){return(ru=s._NotEqual=s.asm.NotEqual).apply(null,arguments)},$c=s._OneHot=function(){return($c=s._OneHot=s.asm.OneHot).apply(null,arguments)},Mc=s._PadV2=function(){return(Mc=s._PadV2=s.asm.PadV2).apply(null,arguments)},Fc=s._Pow=function(){return(Fc=s._Pow=s.asm.Pow).apply(null,arguments)},h2=s._Prelu=function(){return(h2=s._Prelu=s.asm.Prelu).apply(null,arguments)},Pc=s._Prod=function(){return(Pc=s._Prod=s.asm.Prod).apply(null,arguments)},c2=s._RealDiv=function(){return(c2=s._RealDiv=s.asm.RealDiv).apply(null,arguments)},k3=s._Relu=function(){return(k3=s._Relu=s.asm.Relu).apply(null,arguments)},_c=s._Relu6=function(){return(_c=s._Relu6=s.asm.Relu6).apply(null,arguments)},I3=s._ResizeBilinear=function(){return(I3=s._ResizeBilinear=s.asm.ResizeBilinear).apply(null,arguments)},f2=s._Reverse=function(){return(f2=s._Reverse=s.asm.Reverse).apply(null,arguments)},m2=s._RotateWithOffset=function(){return(m2=s._RotateWithOffset=s.asm.RotateWithOffset).apply(null,arguments)},g2=s._Round=function(){return(g2=s._Round=s.asm.Round).apply(null,arguments)},y2=s._Rsqrt=function(){return(y2=s._Rsqrt=s.asm.Rsqrt).apply(null,arguments)},A2=s._ScatterNd=function(){return(A2=s._ScatterNd=s.asm.ScatterNd).apply(null,arguments)},x2=s._SelectV2=function(){return(x2=s._SelectV2=s.asm.SelectV2).apply(null,arguments)},b2=s._Sigmoid=function(){return(b2=s._Sigmoid=s.asm.Sigmoid).apply(null,arguments)},v2=s._Sin=function(){return(v2=s._Sin=s.asm.Sin).apply(null,arguments)},w2=s._Softmax=function(){return(w2=s._Softmax=s.asm.Softmax).apply(null,arguments)},k2=s._SparseFillEmptyRows=function(){return(k2=s._SparseFillEmptyRows=s.asm.SparseFillEmptyRows).apply(null,arguments)},I2=s._SparseReshape=function(){return(I2=s._SparseReshape=s.asm.SparseReshape).apply(null,arguments)},S2=s._SparseSegmentReduction=function(){return(S2=s._SparseSegmentReduction=s.asm.SparseSegmentReduction).apply(null,arguments)},C2=s._Sqrt=function(){return(C2=s._Sqrt=s.asm.Sqrt).apply(null,arguments)},T2=s._Square=function(){return(T2=s._Square=s.asm.Square).apply(null,arguments)},N2=s._SquaredDifference=function(){return(N2=s._SquaredDifference=s.asm.SquaredDifference).apply(null,arguments)},E2=s._Step=function(){return(E2=s._Step=s.asm.Step).apply(null,arguments)},R2=s._StridedSlice=function(){return(R2=s._StridedSlice=s.asm.StridedSlice).apply(null,arguments)},$2=s._Sub=function(){return($2=s._Sub=s.asm.Sub).apply(null,arguments)},M2=s._Sum=function(){return(M2=s._Sum=s.asm.Sum).apply(null,arguments)},F2=s._Tan=function(){return(F2=s._Tan=s.asm.Tan).apply(null,arguments)},P2=s._Tanh=function(){return(P2=s._Tanh=s.asm.Tanh).apply(null,arguments)},_2=s._Tile=function(){return(_2=s._Tile=s.asm.Tile).apply(null,arguments)},z2=s._TopK=function(){return(z2=s._TopK=s.asm.TopK).apply(null,arguments)},O2=s._Transform=function(){return(O2=s._Transform=s.asm.Transform).apply(null,arguments)},D2=s._Transpose=function(){return(D2=s._Transpose=s.asm.Transpose).apply(null,arguments)},L2=s.__FusedMatMul=function(){return(L2=s.__FusedMatMul=s.asm._FusedMatMul).apply(null,arguments)},B2=s._malloc=function(){return(B2=s._malloc=s.asm.malloc).apply(null,arguments)},W2=s._free=function(){return(W2=s._free=s.asm.free).apply(null,arguments)},V2=s.___errno_location=function(){return(V2=s.___errno_location=s.asm.__errno_location).apply(null,arguments)},U2=s._emscripten_main_thread_process_queued_calls=function(){return(U2=s._emscripten_main_thread_process_queued_calls=s.asm.emscripten_main_thread_process_queued_calls).apply(null,arguments)},zc=s.stackSave=function(){return(zc=s.stackSave=s.asm.stackSave).apply(null,arguments)},Oc=s.stackRestore=function(){return(Oc=s.stackRestore=s.asm.stackRestore).apply(null,arguments)},up=s.stackAlloc=function(){return(up=s.stackAlloc=s.asm.stackAlloc).apply(null,arguments)},G2=s.dynCall_iijjiiii=function(){return(G2=s.dynCall_iijjiiii=s.asm.dynCall_iijjiiii).apply(null,arguments)},j2=s.dynCall_jiji=function(){return(j2=s.dynCall_jiji=s.asm.dynCall_jiji).apply(null,arguments)};s.cwrap=$e;var nu;function dp(j){this.name="ExitStatus",this.message="Program terminated with exit("+j+")",this.status=j}As=function j(){nu||pp(),nu||(As=j)};function pp(j){if(j=j||d,Sn>0||(hc(),Sn>0))return;function ne(){nu||(nu=!0,s.calledRun=!0,!ae&&(cc(),i(s),s.onRuntimeInitialized&&s.onRuntimeInitialized(),fc()))}s.setStatus?(s.setStatus("Running..."),setTimeout(function(){setTimeout(function(){s.setStatus("")},1),ne()},1)):ne()}s.run=pp;function S3(j){de=j,np()||(s.onExit&&s.onExit(j),ae=!0),p(j,new dp(j))}if(s.preInit)for(typeof s.preInit=="function"&&(s.preInit=[s.preInit]);s.preInit.length>0;)s.preInit.pop()();pp();var au;l&&(au={uncaughtException:process.listeners("uncaughtException").filter(function(j){return!l.uncaughtException.indexOf(j)>-1}),unhandledRejection:process.listeners("unhandledRejection").filter(function(j){return!l.unhandledRejection.indexOf(j)>-1})});var su;if(typeof a!="undefined")su=a;else if(typeof WasmBackendModuleThreadedSimd!="undefined")su=WasmBackendModuleThreadedSimd;else throw new Error("Could not find wasm module in post.js");if(au){var H2=su._dispose;su._dispose=function(){H2(),au.uncaughtException.forEach(function(j){process.removeListener("uncaughtException",j)}),au.unhandledRejection.forEach(function(j){process.removeListener("unhandledRejection",j)})}}return a.ready}})();typeof e=="object"&&typeof t=="object"?t.exports=r:typeof define=="function"&&define.amd?define([],function(){return r}):typeof e=="object"&&(e.WasmBackendModule=r)}}),xR=1e-7,bR=1e-4,Zp=class{constructor(e,t){this.backend=e,this.dataMover=t,this.data=new WeakMap,this.dataIdsCount=0}get(e){return this.data.has(e)||this.dataMover.moveData(this.backend,e),this.data.get(e)}set(e,t){this.dataIdsCount++,this.data.set(e,t)}has(e){return this.data.has(e)}delete(e){return this.dataIdsCount--,this.data.delete(e)}numDataIds(){return this.dataIdsCount}},zu=class{refCount(e){return Tn("refCount")}incRef(e){return Tn("incRef")}timerAvailable(){return!0}time(e){return Tn("time")}read(e){return Tn("read")}readSync(e){return Tn("readSync")}readToGPU(e,t){return Tn("readToGPU")}numDataIds(){return Tn("numDataIds")}disposeData(e,t){return Tn("disposeData")}write(e,t,r){return Tn("write")}move(e,t,r,n,a){return Tn("move")}memory(){return Tn("memory")}floatPrecision(){return Tn("floatPrecision")}epsilon(){return this.floatPrecision()===32?xR:bR}dispose(){return Tn("dispose")}};function Tn(e){throw new Error(`'${e}' not yet implemented or not found in the registry. This kernel may not be supported by the tfjs backend you have chosen`)}function Cv(e){let t=e.length,r=0;for(;t>0;)r=Math.random()*t|0,t--,y0(e,t,r)}function vR(e,t){if(e.length!==t.length)throw new Error(`Array sizes must match to be shuffled together First array length was ${e.length}Second array length was ${t.length}`);let r=e.length,n=0;for(;r>0;)n=Math.random()*r|0,r--,y0(e,r,n),y0(t,r,n)}function zp(e,t,r){return Math.max(e,Math.min(t,r))}function wR(e){return e%2===0?e:e+1}function y0(e,t,r){let n=e[t];e[t]=e[r],e[r]=n}function kR(e){let t=0;for(let r=0;r<e.length;r++)t+=e[r];return t}function IR(e,t){let r=Math.random();return t*r+(1-r)*e}function SR(e,t){let r=0;for(let n=0;n<e.length;n++){let a=Number(e[n])-Number(t[n]);r+=a*a}return r}function P(e,t){if(!e)throw new Error(typeof t=="string"?t:t())}function Ur(e,t,r=""){P(Ks(e,t),()=>r+` Shapes ${e} and ${t} must match`)}function jo(e){P(e!=null,()=>"The input to the tensor constructor must be a non-null value.")}function Fo(e,t=[],r=!1){if(t==null&&(t=[]),Array.isArray(e)||Sr(e)&&!r)for(let n=0;n<e.length;++n)Fo(e[n],t,r);else t.push(e);return t}function It(e){if(e.length===0)return 1;let t=e[0];for(let r=1;r<e.length;r++)t*=e[r];return t}function CR(e){return e.length===0}function Ks(e,t){if(e===t)return!0;if(e==null||t==null||e.length!==t.length)return!1;for(let r=0;r<e.length;r++)if(e[r]!==t[r])return!1;return!0}function wu(e){return e%1===0}function TR(e){if(Math.tanh!=null)return Math.tanh(e);if(e===1/0)return 1;if(e===-1/0)return-1;{let t=Math.exp(2*e);return(t-1)/(t+1)}}function NR(e){let t=Math.ceil(Math.sqrt(e));return[t,Math.ceil(e/t)]}function ER(e){let t=new Uint32Array(e);for(let r=0;r<e;++r)t[r]=r;return Cv(t),t}function $p(e,t){return t<=e.length?e:e+" ".repeat(t-e.length)}function RR(e,t=n=>0,r){return new Promise((n,a)=>{let s=0,i=()=>{if(e()){n();return}s++;let o=t(s);if(r!=null&&s>=r){a();return}setTimeout(i,o)};i()})}function $R(e,t){let r=1,n=-1;for(let s=0;s<e.length;++s)if(e[s]>=0)r*=e[s];else if(e[s]===-1){if(n!==-1)throw Error(`Shapes can only have 1 implicit size. Found -1 at dim ${n} and dim ${s}`);n=s}else if(e[s]<0)throw Error(`Shapes can not be < 0. Found ${e[s]} at dim ${s}`);if(n===-1){if(t>0&&t!==r)throw Error(`Size(${t}) must match the product of shape ${e}`);return e}if(r===0)throw Error(`Cannot infer the missing size in [${e}] when there are 0 elements`);if(t%r!==0)throw Error(`The implicit shape can't be a fractional number. Got ${t} / ${r}`);let a=e.slice();return a[n]=t/r,a}function Hn(e,t){let r=t.length;return e=e==null?t.map((n,a)=>a):[].concat(e),P(e.every(n=>n>=-r&&n<r),()=>`All values in axis param must be in range [-${r}, ${r}) but got axis ${e}`),P(e.every(n=>wu(n)),()=>`All values in axis param must be integers but got axis ${e}`),e.map(n=>n<0?r+n:n)}function Tv(e,t){let r=[],n=[],a=t!=null&&Array.isArray(t)&&t.length===0,s=t==null||a?null:Hn(t,e).sort(),i=0;for(let o=0;o<e.length;++o){if(s!=null){if(s[i]===o&&e[o]!==1)throw new Error(`Can't squeeze axis ${o} since its dim '${e[o]}' is not 1`);(s[i]==null||s[i]>o)&&e[o]===1&&(r.push(e[o]),n.push(o)),s[i]<=o&&i++}e[o]!==1&&(r.push(e[o]),n.push(o))}return{newShape:r,keptDims:n}}function Nv(e,t){let r=null;if(e==null||e==="float32")r=new Float32Array(t);else if(e==="int32")r=new Int32Array(t);else if(e==="bool")r=new Uint8Array(t);else throw new Error(`Unknown data type ${e}`);return r}function Ev(e,t){let r=null;if(e==null||e==="float32")r=new Float32Array(t);else if(e==="int32")r=new Int32Array(t);else if(e==="bool")r=new Uint8Array(t);else if(e==="string")r=new Array(t);else throw new Error(`Unknown data type ${e}`);return r}function Rv(e,t){for(let r=0;r<e.length;r++){let n=e[r];if(isNaN(n)||!isFinite(n))throw Error(`A tensor of type ${t} being uploaded contains ${n}.`)}}function $v(e){return e==="bool"||e==="complex64"||e==="float32"||e==="int32"||e==="string"}function MR(e,t){return!(t==="complex64"||t==="float32"&&e!=="complex64"||t==="int32"&&e!=="float32"&&e!=="complex64"||t==="bool"&&e==="bool")}function Sr(e){return e instanceof Float32Array||e instanceof Int32Array||e instanceof Uint8Array||e instanceof Uint8ClampedArray}function fg(e){if(e==="float32"||e==="int32")return 4;if(e==="complex64")return 8;if(e==="bool")return 1;throw new Error(`Unknown dtype ${e}`)}function Mv(e){if(e==null)return 0;let t=0;return e.forEach(r=>t+=r.length),t}function Ts(e){return typeof e=="string"||e instanceof String}function Fv(e){return typeof e=="boolean"}function Pv(e){return typeof e=="number"}function Z0(e){return Array.isArray(e)?Z0(e[0]):e instanceof Float32Array?"float32":e instanceof Int32Array||e instanceof Uint8Array||e instanceof Uint8ClampedArray?"int32":Pv(e)?"float32":Ts(e)?"string":Fv(e)?"bool":"float32"}function Fs(e){return!!(e&&e.constructor&&e.call&&e.apply)}function A0(e,t){for(let r=t;r<e;++r)if(e%r===0)return r;return e}function Ou(e){let t=e.length;if(t<2)return[];let r=new Array(t-1);r[t-2]=e[t-1];for(let n=t-3;n>=0;--n)r[n]=r[n+1]*e[n+1];return r}function _v(e,t,r,n=!1){let a=new Array;if(t.length===1){let s=t[0]*(n?2:1);for(let i=0;i<s;i++)a[i]=r[e+i]}else{let s=t[0],i=t.slice(1),o=i.reduce((l,u)=>l*u)*(n?2:1);for(let l=0;l<s;l++)a[l]=_v(e+l*o,i,r,n)}return a}function yu(e,t,r=!1){if(e.length===0)return t[0];let n=e.reduce((a,s)=>a*s)*(r?2:1);if(n===0)return[];if(n!==t.length)throw new Error(`[${e}] does not match the input size ${t.length}${r?" for a complex tensor":""}.`);return _v(0,e,t,r)}function my(e,t){let r=Y0(e,t);for(let n=0;n<r.length;n++)r[n]=1;return r}function Y0(e,t){if(t==null||t==="float32"||t==="complex64")return new Float32Array(e);if(t==="int32")return new Int32Array(e);if(t==="bool")return new Uint8Array(e);throw new Error(`Unknown data type ${t}`)}function FR(e,t){let r=e.reduce((n,a)=>n*a,1);if(t==null||t==="float32")return yu(e,new Float32Array(r));if(t==="int32")return yu(e,new Int32Array(r));if(t==="bool")return yu(e,new Uint8Array(r));throw new Error(`Unknown data type ${t}`)}function gy(e){e.forEach(t=>{P(Number.isInteger(t)&&t>=0,()=>`Tensor must have a shape comprised of positive integers but got shape [${e}].`)})}function PR(e,t,r){if(t===0)return 0;if(t===1)return e[0];let n=e[e.length-1];for(let a=0;a<e.length-1;++a)n+=r[a]*e[a];return n}function _R(e,t,r){if(t===0)return[];if(t===1)return[e];let n=new Array(t);for(let a=0;a<n.length-1;++a)n[a]=Math.floor(e/r[a]),e-=n[a]*r[a];return n[n.length-1]=e,n}function yy(e){return e&&e.then&&typeof e.then=="function"}var O3="tfjsflags",zv=class{constructor(e){this.global=e,this.flags={},this.flagRegistry={},this.urlFlags={},this.getQueryParams=zR,this.populateURLFlags()}setPlatform(e,t){this.platform!=null&&(Z().getBool("IS_TEST")||Z().getBool("PROD")||console.warn(`Platform ${this.platformName} has already been set. Overwriting the platform with ${e}.`)),this.platformName=e,this.platform=t}registerFlag(e,t,r){if(this.flagRegistry[e]={evaluationFn:t,setHook:r},this.urlFlags[e]!=null){let n=this.urlFlags[e];Z().getBool("IS_TEST")||Z().getBool("PROD")||console.warn(`Setting feature override from URL ${e}: ${n}.`),this.set(e,n)}}async getAsync(e){return e in this.flags?this.flags[e]:(this.flags[e]=await this.evaluateFlag(e),this.flags[e])}get(e){if(e in this.flags)return this.flags[e];let t=this.evaluateFlag(e);if(yy(t))throw new Error(`Flag ${e} cannot be synchronously evaluated. Please use getAsync() instead.`);return this.flags[e]=t,this.flags[e]}getNumber(e){return this.get(e)}getBool(e){return this.get(e)}getFlags(){return this.flags}get features(){return this.flags}set(e,t){if(this.flagRegistry[e]==null)throw new Error(`Cannot set flag ${e} as it has not been registered.`);this.flags[e]=t,this.flagRegistry[e].setHook!=null&&this.flagRegistry[e].setHook(t)}evaluateFlag(e){if(this.flagRegistry[e]==null)throw new Error(`Cannot evaluate flag '${e}': no evaluation function found.`);return this.flagRegistry[e].evaluationFn()}setFlags(e){this.flags=Object.assign({},e)}reset(){this.flags={},this.urlFlags={},this.populateURLFlags()}populateURLFlags(){if(typeof this.global=="undefined"||typeof this.global.location=="undefined"||typeof this.global.location.search=="undefined")return;let e=this.getQueryParams(this.global.location.search);O3 in e&&e[O3].split(",").forEach(t=>{let[r,n]=t.split(":");this.urlFlags[r]=DR(r,n)})}};function zR(e){let t={};return e.replace(/[?&]([^=?&]+)(?:=([^&]*))?/g,(r,...n)=>(OR(t,n[0],n[1]),n.join("="))),t}function OR(e,t,r){e[decodeURIComponent(t)]=decodeURIComponent(r||"")}function DR(e,t){if(t=t.toLowerCase(),t==="true"||t==="false")return t==="true";if(`${+t}`===t)return+t;throw new Error(`Could not parse value flag value ${t} for flag ${e}.`)}function Z(){return ba}var ba=null;function LR(e){ba=e}var eg;function Ov(){if(eg==null){let e;if(typeof window!="undefined")e=window;else if(typeof global!="undefined")e=global;else if(typeof process!="undefined")e=process;else if(typeof self!="undefined")e=self;else throw new Error("Could not find a global object");eg=e}return eg}function BR(){let e=Ov();return e._tfGlobals==null&&(e._tfGlobals=new Map),e._tfGlobals}function Ay(e,t){let r=BR();if(r.has(e))return r.get(e);{let n=t();return r.set(e,n),r.get(e)}}var Ho="Abs",Du="Acos",Lu="Acosh",Ja="Add",Zs="AddN",Bu="All",Wu="Any",Ys="ArgMax",Vu="ArgMin",Uu="Asin",Gu="Asinh",ju="Atan",Hu="Atanh",qu="Atan2",Js="AvgPool",J0="AvgPoolGrad",Yp="AvgPool3D",Q0="AvgPool3DGrad",Qs="BatchMatMul",qo="BatchToSpaceND",ef="Bincount",Dv="BroadcastTo",tf="BroadcastArgs",ei="Cast",ti="Ceil",Qa="ClipByValue",Jp="Complex",Qp="ComplexAbs",Xo="Concat",ri="Conv2D",rf="Conv2DBackpropFilter",ni="Conv2DBackpropInput",eh="Conv3D",nf="Conv3DBackpropFilterV2",af="Conv3DBackpropInputV2",ai="Cos",si="Cosh",Ko="Cumprod",ii="Cumsum",Zo="CropAndResize",sf="DenseBincount",Yo="DepthToSpace",oi="DepthwiseConv2dNative",of="DepthwiseConv2dNativeBackpropFilter",lf="DepthwiseConv2dNativeBackpropInput",uf="Diag",th="Dilation2D",x0="Dilation2DBackpropInput",b0="Dilation2DBackpropFilter",li="RealDiv",rh="Einsum",ui="Elu",df="EluGrad",Xu="Erf",Jo="Equal",di="Exp",Qo="ExpandDims",el="Expm1",pf="FFT",Ku="Fill",tl="FlipLeftRight",pi="Floor",hi="FloorDiv",ci="FusedBatchNorm",rl="GatherV2",nl="GatherNd",al="Greater",fi="GreaterEqual",mi="Identity",hf="IFFT",nh="Imag",Zu="IsFinite",Yu="IsInf",Ju="IsNan",gi="LeakyRelu",sl="Less",il="LessEqual",cf="LinSpace",yi="Log",Qu="Log1p",ol="LogicalAnd",ed="LogicalNot",ah="LogicalOr",Lv="LogSoftmax",WR="LowerBound",sh="LRN",ff="LRNGrad",Ai="Max",xi="Maximum",bi="MaxPool",mf="MaxPoolGrad",ih="MaxPool3D",gf="MaxPool3DGrad",yf="MaxPoolWithArgmax",vi="Mean",wi="Min",ki="Minimum",Ii="MirrorPad",td="Mod",Af="Multinomial",Si="Multiply",ll="Neg",ul="NotEqual",dl="NonMaxSuppressionV3",rd="NonMaxSuppressionV4",pl="NonMaxSuppressionV5",hl="OnesLike",cl="OneHot",fl="Pack",Ci="PadV2",VR="Pool",Ti="Pow",Ni="Prelu",Ei="Prod",nd="Range",oh="Real",ad="Reciprocal",Ri="Relu",ml="Reshape",sd="ResizeNearestNeighbor",xf="ResizeNearestNeighborGrad",$i="ResizeBilinear",bf="ResizeBilinearGrad",Mi="Relu6",gl="Reverse",yl="Round",Fi="Rsqrt",Al="ScatterNd",vf="SearchSorted",xl="Select",id="Selu",bl="Slice",Pi="Sin",vl="Sinh",od="Sign",_i="Sigmoid",ld="Softplus",zi="Sqrt",Oi="Sum",wl="SpaceToBatchND",kl="SplitV",Di="Softmax",lh="SparseFillEmptyRows",ud="SparseReshape",uh="SparseSegmentMean",dh="SparseSegmentSum",ph="SparseToDense",Li="SquaredDifference",dd="Square",Il="StridedSlice",hh="StringNGrams",wf="StringSplit",kf="StringToHashBucketFast",Bi="Sub",Sl="Tan",Wi="Tanh",es="Tile",Cl="TopK",Tl="Transform",Vi="Transpose",If="Unique",Nl="Unpack",ch="UnsortedSegmentSum",UR="UpperBound",El="ZerosLike",Ui="Step",Op="FromPixels",Rl="RotateWithOffset",Ps="_FusedMatMul",_s="FusedConv2D",zs="FusedDepthwiseConv2D";function Cs(...e){Z().getBool("IS_TEST")||Z().getBool("PROD")||console.warn(...e)}function GR(...e){Z().getBool("IS_TEST")||Z().getBool("PROD")||console.log(...e)}var ku=Ay("kernelRegistry",()=>new Map),Dp=Ay("gradRegistry",()=>new Map);function v0(e,t){let r=xy(e,t);return ku.get(r)}function mg(e){return Dp.get(e)}function Ma(e){let t=ku.entries(),r=[];for(;;){let{done:n,value:a}=t.next();if(n)break;let[s,i]=a,[o]=s.split("_");o===e&&r.push(i)}return r}function qn(e){let{kernelName:t,backendName:r}=e,n=xy(t,r);ku.has(n)&&Cs(`The kernel '${t}' for backend '${r}' is already registered`),ku.set(n,e)}function Bv(e){let{kernelName:t}=e;Dp.has(t)&&Z().getBool("DEBUG")&&Cs(`Overriding the gradient for '${t}'`),Dp.set(t,e)}function jR(e,t){let r=xy(e,t);if(!ku.has(r))throw new Error(`The kernel '${e}' for backend '${t}' is not registered`);ku.delete(r)}function HR(e){if(!Dp.has(e))throw new Error(`The gradient '${e}' for backend is not registered`);Dp.delete(e)}function qR(e,t){Ma(e).forEach(r=>{let n=Object.assign({},r,{backendName:t});qn(n)})}function xy(e,t){return`${t}_${e}`}var v={};Le(v,{arraysEqual:()=>Ks,assert:()=>P,assertNonNegativeIntegerDimensions:()=>gy,assertNonNull:()=>jo,assertShapesMatch:()=>Ur,bytesFromStringArray:()=>Mv,bytesPerElement:()=>fg,checkConversionForErrors:()=>Rv,clamp:()=>zp,computeStrides:()=>Ou,createScalarValue:()=>QR,createShuffledIndices:()=>ER,decodeString:()=>w0,distSquared:()=>SR,encodeString:()=>mh,fetch:()=>t$,fingerPrint64:()=>JR,flatten:()=>Fo,getArrayFromDType:()=>Ev,getTypedArrayFromDType:()=>Nv,hasEncodingLoss:()=>MR,hexToLong:()=>fh,indexToLoc:()=>_R,inferDtype:()=>Z0,inferFromImplicitShape:()=>$R,isBoolean:()=>Fv,isFunction:()=>Fs,isInt:()=>wu,isNumber:()=>Pv,isPromise:()=>yy,isScalarShape:()=>CR,isString:()=>Ts,isTypedArray:()=>Sr,isValidDtype:()=>$v,locToIndex:()=>PR,makeOnesTypedArray:()=>my,makeZerosNestedTypedArray:()=>FR,makeZerosTypedArray:()=>Y0,nearestDivisor:()=>A0,nearestLargerEven:()=>wR,now:()=>Lp,parseAxisParam:()=>Hn,randUniform:()=>IR,repeatedTry:()=>RR,rightPad:()=>$p,shuffle:()=>Cv,shuffleCombo:()=>vR,sizeFromShape:()=>It,sizeToSquarishShape:()=>NR,squeezeShape:()=>Tv,sum:()=>kR,swap:()=>y0,tanh:()=>TR,toNestedArray:()=>yu,toTypedArray:()=>Sf});var D3=Go(nR()),wo=D3.default||D3;function fh(e){return wo.fromString(e,!0,16)}var Wv=fh("c3a5c85c97cb3127"),xo=fh("b492b66fbe98f273"),Dr=fh("9ae16a3b2f90404f");function gg(e){return e.xor(e.shru(47))}function Vv(e,t,r){let n=e.slice(t,t+r);return wo.fromBytes(Array.from(n),!0,!0)}function kt(e,t){return Vv(e,t,8)}function L3(e,t){return Vv(e,t,4)}function mr(e,t){return t===0?e:e.shru(t).or(e.shl(64-t))}function Es(e,t,r=fh("9ddfea08eb382d69")){let n=e.xor(t).mul(r);n=n.xor(n.shru(47));let a=t.xor(n).mul(r);return a=a.xor(a.shru(47)),a=a.mul(r),a}function XR(e,t,r,n,a,s){a=a.add(e),s=mr(s.add(a).add(n),21);let i=a;return a=a.add(t),a=a.add(r),s=s.add(mr(a,44)),[a.add(n),s.add(i)]}function Hc(e,t,r,n){return XR(kt(e,t),kt(e,t+8),kt(e,t+16),kt(e,t+24),r,n)}function KR(e,t=e.length){if(t>=8){let r=Dr.add(t*2),n=kt(e,0).add(Dr),a=kt(e,t-8),s=mr(a,37).mul(r).add(n),i=mr(n,25).add(a).mul(r);return Es(s,i,r)}if(t>=4){let r=Dr.add(t*2),n=L3(e,0);return Es(n.shl(3).add(t),L3(e,t-4),r)}if(t>0){let r=e[0],n=e[t>>1],a=e[t-1],s=r+(n<<8),i=t+(a<<2);return gg(Dr.mul(s).xor(Wv.mul(i))).mul(Dr)}return Dr}function ZR(e,t=e.length){let r=Dr.add(t*2),n=kt(e,0).mul(xo),a=kt(e,8),s=kt(e,t-8).mul(r),i=kt(e,t-16).mul(Dr);return Es(mr(n.add(a),43).add(mr(s,30)).add(i),n.add(mr(a.add(Dr),18)).add(s),r)}function YR(e,t=e.length){let r=Dr.add(t*2),n=kt(e,0).mul(Dr),a=kt(e,8),s=kt(e,t-8).mul(r),i=kt(e,t-16).mul(Dr),o=mr(n.add(a),43).add(mr(s,30)).add(i),l=Es(o,n.add(mr(a.add(Dr),18)).add(s),r),u=kt(e,16).mul(r),d=kt(e,24),h=o.add(kt(e,t-32)).mul(r),p=l.add(kt(e,t-24)).mul(r);return Es(mr(u.add(d),43).add(mr(h,30)).add(p),u.add(mr(d.add(n),18)).add(h),r)}function JR(e,t=e.length){let r=wo.fromNumber(81,!0);if(t<=32)return t<=16?KR(e,t):ZR(e,t);if(t<=64)return YR(e,t);let n=r,a=r.mul(xo).add(113),s=gg(a.mul(Dr).add(113)).mul(Dr),i=[wo.UZERO,wo.UZERO],o=[wo.UZERO,wo.UZERO];n=n.mul(Dr).add(kt(e,0));let l=0,u=(t-1>>6)*64,d=u+(t-1&63)-63;do n=mr(n.add(a).add(i[0]).add(kt(e,l+8)),37).mul(xo),a=mr(a.add(i[1]).add(kt(e,l+48)),42).mul(xo),n=n.xor(o[1]),a=a.add(i[0]).add(kt(e,l+40)),s=mr(s.add(o[0]),33).mul(xo),i=Hc(e,l,i[1].mul(xo),n.add(o[0])),o=Hc(e,l+32,s.add(o[1]),a.add(kt(e,l+16))),[s,n]=[n,s],l+=64;while(l!==u);let h=xo.add(s.and(255).shl(1));return l=d,o[0]=o[0].add(t-1&63),i[0]=i[0].add(o[0]),o[0]=o[0].add(i[0]),n=mr(n.add(a).add(i[0]).add(kt(e,l+8)),37).mul(h),a=mr(a.add(i[1]).add(kt(e,l+48)),42).mul(h),n=n.xor(o[1].mul(9)),a=a.add(i[0].mul(9).add(kt(e,l+40))),s=mr(s.add(o[0]),33).mul(h),i=Hc(e,l,i[1].mul(h),n.add(o[0])),o=Hc(e,l+32,s.add(o[1]),a.add(kt(e,l+16))),[s,n]=[n,s],Es(Es(i[0],o[0],h).add(gg(a).mul(Wv)).add(s),Es(i[1],o[1],h).add(n),h)}function QR(e,t){return t==="string"?mh(e):Sf([e],t)}function e$(e,t){return e instanceof Float32Array&&t==="float32"||e instanceof Int32Array&&t==="int32"||e instanceof Uint8Array&&t==="bool"}function Sf(e,t){if(t==="string")throw new Error("Cannot convert a string[] to a TypedArray");if(Array.isArray(e)&&(e=Fo(e)),Z().getBool("DEBUG")&&Rv(e,t),e$(e,t))return e;if(t==null||t==="float32"||t==="complex64")return new Float32Array(e);if(t==="int32")return new Int32Array(e);if(t==="bool"){let r=new Uint8Array(e.length);for(let n=0;n<r.length;++n)Math.round(e[n])!==0&&(r[n]=1);return r}else throw new Error(`Unknown data type ${t}`)}function Lp(){return Z().platform.now()}function t$(e,t){return Z().platform.fetch(e,t)}function mh(e,t="utf-8"){return t=t||"utf-8",Z().platform.encode(e,t)}function w0(e,t="utf-8"){return t=t||"utf-8",Z().platform.decode(e,t)}var r$=class{constructor(e,t){this.backendTimer=e,this.logger=t,t==null&&(this.logger=new a$)}profileKernel(e,t,r){let n,a=()=>{n=r()},s,i=Lp();if(this.backendTimer.timerAvailable())s=this.backendTimer.time(a);else{a();for(let o of n)o.dataSync();s=Promise.resolve({kernelMs:Lp()-i})}if(Z().getBool("CHECK_COMPUTATION_FOR_ERRORS"))for(let o=0;o<n.length;o++){let l=n[o];l.data().then(u=>{n$(u,l.dtype,e)})}return{kernelName:e,outputs:n,inputs:t,timeMs:s.then(o=>o.kernelMs),extraInfo:s.then(o=>o.getExtraProfileInfo!=null?o.getExtraProfileInfo():"")}}logKernelProfile(e){let{kernelName:t,outputs:r,timeMs:n,inputs:a,extraInfo:s}=e;r.forEach(i=>{Promise.all([i.data(),n,s]).then(o=>{this.logger.logKernelProfile(t,i,o[0],o[1],a,o[2])})})}};function n$(e,t,r){if(t!=="float32")return!1;for(let n=0;n<e.length;n++){let a=e[n];if(isNaN(a)||!isFinite(a))return console.warn(`Found ${a} in the result of '${r}'`),!0}return!1}var a$=class{logKernelProfile(e,t,r,n,a,s){let i=typeof n=="number"?$p(`${n}ms`,9):n.error,o=$p(e,25),l=t.rank,u=t.size,d=$p(t.shape.toString(),14),h="";for(let p in a){let c=a[p];if(c!=null){let f=c.shape||t.shape,m=f.length;h+=`${p}: ${m}D ${m>0?f:""} `}}console.log(`%c${o} %c${i} %c${l}D ${d} %c${u} %c${h} %c${s}`,"font-weight:bold","color:red","color:blue","color: orange","color: green","color: steelblue")}};function s$(e,t,r){let n={},a={};for(let l=0;l<t.length;l++)n[t[l].id]=!0;for(let l=0;l<e.length;l++){let u=e[l],d=u.inputs;for(let h in d){let p=d[h],c=!1;for(let f=0;f<t.length;f++)if(n[p.id]){u.outputs.forEach(m=>n[m.id]=!0),c=!0,a[u.id]=!0;break}if(c)break}}let s={};s[r.id]=!0;let i={};for(let l=e.length-1;l>=0;l--){let u=e[l],d=u.inputs;for(let h=0;h<u.outputs.length;h++)if(s[u.outputs[h].id]){for(let p in d)s[d[p].id]=!0,i[u.id]=!0;break}}let o=[];for(let l=0;l<e.length;l++){let u=e[l];if(a[u.id]&&i[u.id]){let d={};for(let p in u.inputs){let c=u.inputs[p];n[c.id]&&(d[p]=c)}let h=Object.assign({},u);h.inputs=d,h.outputs=u.outputs,o.push(h)}}return o}function i$(e,t,r,n){for(let a=t.length-1;a>=0;a--){let s=t[a],i=[];if(s.outputs.forEach(l=>{let u=e[l.id];u!=null?i.push(u):i.push(null)}),s.gradient==null)throw new Error(`Cannot compute gradient: gradient function not found for ${s.kernelName}.`);let o=s.gradient(i);for(let l in s.inputs){if(!(l in o))throw new Error(`Cannot backprop through input ${l}. Available gradients found: ${Object.keys(o)}.`);let u=r(()=>o[l]());if(u.dtype!=="float32")throw new Error(`Error in gradient for op ${s.kernelName}. The gradient of input ${l} must have 'float32' dtype, but has '${u.dtype}'`);let d=s.inputs[l];if(!Ks(u.shape,d.shape))throw new Error(`Error in gradient for op ${s.kernelName}. The gradient of input '${l}' has shape '${u.shape}', which does not match the shape of the input '${d.shape}'`);if(e[d.id]==null)e[d.id]=u;else{let h=e[d.id];e[d.id]=n(h,u),h.dispose()}}}}var B3=20,yp=3,tg=7;function o$(e,t,r,n){let a=Ou(t),s=l$(e,t,r,a),i=t.length,o=a0(e,t,r,a,s),l=["Tensor"];return n&&(l.push(` dtype: ${r}`),l.push(` rank: ${i}`),l.push(` shape: [${t}]`),l.push(" values:")),l.push(o.map(u=>" "+u).join(`
|
|
`)),l.join(`
|
|
`)}function l$(e,t,r,n){let a=It(t),s=n[n.length-1],i=new Array(s).fill(0),o=t.length,l=r==="complex64"?wp(e):e;if(o>1)for(let u=0;u<a/s;u++){let d=u*s;for(let h=0;h<s;h++)i[h]=Math.max(i[h],vp(l[d+h],0,r).length)}return i}function vp(e,t,r){let n;return Array.isArray(e)?n=`${parseFloat(e[0].toFixed(tg))} + ${parseFloat(e[1].toFixed(tg))}j`:Ts(e)?n=`'${e}'`:r==="bool"?n=Uv(e):n=parseFloat(e.toFixed(tg)).toString(),$p(n,t)}function Uv(e){return e===0?"false":"true"}function a0(e,t,r,n,a,s=!0){let i=r==="complex64"?2:1,o=t[0],l=t.length;if(l===0){if(r==="complex64"){let m=wp(e);return[vp(m[0],0,r)]}return r==="bool"?[Uv(e[0])]:[e[0].toString()]}if(l===1){if(o>B3){let g=yp*i,y=Array.from(e.slice(0,g)),A=Array.from(e.slice((o-yp)*i,o*i));return r==="complex64"&&(y=wp(y),A=wp(A)),["["+y.map((x,b)=>vp(x,a[b],r)).join(", ")+", ..., "+A.map((x,b)=>vp(x,a[o-yp+b],r)).join(", ")+"]"]}let m=r==="complex64"?wp(e):Array.from(e);return["["+m.map((g,y)=>vp(g,a[y],r)).join(", ")+"]"]}let u=t.slice(1),d=n.slice(1),h=n[0]*i,p=[];if(o>B3){for(let m=0;m<yp;m++){let g=m*h,y=g+h;p.push(...a0(e.slice(g,y),u,r,d,a,!1))}p.push("...");for(let m=o-yp;m<o;m++){let g=m*h,y=g+h;p.push(...a0(e.slice(g,y),u,r,d,a,m===o-1))}}else for(let m=0;m<o;m++){let g=m*h,y=g+h;p.push(...a0(e.slice(g,y),u,r,d,a,m===o-1))}let c=l===2?",":"";p[0]="["+p[0]+c;for(let m=1;m<p.length-1;m++)p[m]=" "+p[m]+c;let f=`,
|
|
`;for(let m=2;m<l;m++)f+=`
|
|
`;return p[p.length-1]=" "+p[p.length-1]+"]"+(s?"":f),p}function wp(e){let t=[];for(let r=0;r<e.length;r+=2)t.push([e[r],e[r+1]]);return t}var ir=class{constructor(e,t,r){if(this.dtype=t,this.shape=e.slice(),this.size=It(e),r!=null){let n=r.length;P(n===this.size,()=>`Length of values '${n}' does not match the size inferred by the shape '${this.size}'.`)}if(t==="complex64")throw new Error("complex64 dtype TensorBuffers are not supported. Please create a TensorBuffer for the real and imaginary parts separately and call tf.complex(real, imag).");this.values=r||Ev(t,this.size),this.strides=Ou(e)}set(e,...t){t.length===0&&(t=[0]),P(t.length===this.rank,()=>`The number of provided coordinates (${t.length}) must match the rank (${this.rank})`);let r=this.locToIndex(t);this.values[r]=e}get(...e){e.length===0&&(e=[0]);let t=0;for(let n of e){if(n<0||n>=this.shape[t]){let a=`Requested out of range element at ${e}. Buffer shape=${this.shape}`;throw new Error(a)}t++}let r=e[e.length-1];for(let n=0;n<e.length-1;++n)r+=this.strides[n]*e[n];return this.values[r]}locToIndex(e){if(this.rank===0)return 0;if(this.rank===1)return e[0];let t=e[e.length-1];for(let r=0;r<e.length-1;++r)t+=this.strides[r]*e[r];return t}indexToLoc(e){if(this.rank===0)return[];if(this.rank===1)return[e];let t=new Array(this.shape.length);for(let r=0;r<t.length-1;++r)t[r]=Math.floor(e/this.strides[r]),e-=t[r]*this.strides[r];return t[t.length-1]=e,t}get rank(){return this.shape.length}toTensor(){return la().makeTensor(this.values,this.shape,this.dtype)}},la=null,fu=null,u$=null;function d$(e){la=e}function p$(e){fu=e}function h$(e){u$=e}var nt=class{constructor(e,t,r,n){this.kept=!1,this.isDisposedInternal=!1,this.shape=e.slice(),this.dtype=t||"float32",this.size=It(e),this.strides=Ou(e),this.dataId=r,this.id=n,this.rankType=this.rank<5?this.rank.toString():"higher"}get rank(){return this.shape.length}async buffer(){let e=await this.data();return fu.buffer(this.shape,this.dtype,e)}bufferSync(){return fu.buffer(this.shape,this.dtype,this.dataSync())}async array(){let e=await this.data();return yu(this.shape,e,this.dtype==="complex64")}arraySync(){return yu(this.shape,this.dataSync(),this.dtype==="complex64")}async data(){this.throwIfDisposed();let e=la().read(this.dataId);if(this.dtype==="string"){let t=await e;try{return t.map(r=>w0(r))}catch(r){throw new Error("Failed to decode the string bytes into utf-8. To get the original bytes, call tensor.bytes().")}}return e}dataToGPU(e){return this.throwIfDisposed(),la().readToGPU(this.dataId,e)}dataSync(){this.throwIfDisposed();let e=la().readSync(this.dataId);if(this.dtype==="string")try{return e.map(t=>w0(t))}catch(t){throw new Error("Failed to decode the string bytes into utf-8. To get the original bytes, call tensor.bytes().")}return e}async bytes(){this.throwIfDisposed();let e=await la().read(this.dataId);return this.dtype==="string"?e:new Uint8Array(e.buffer)}dispose(){this.isDisposed||(la().disposeTensor(this),this.isDisposedInternal=!0)}get isDisposed(){return this.isDisposedInternal}throwIfDisposed(){if(this.isDisposed)throw new Error("Tensor is disposed.")}print(e=!1){return fu.print(this,e)}clone(){return this.throwIfDisposed(),fu.clone(this)}toString(e=!1){let t=this.dataSync();return o$(t,this.shape,this.dtype,e)}cast(e){return this.throwIfDisposed(),fu.cast(this,e)}variable(e=!0,t,r){return this.throwIfDisposed(),la().makeVariable(this,e,t,r)}};Object.defineProperty(nt,Symbol.hasInstance,{value:e=>!!e&&e.data!=null&&e.dataSync!=null&&e.throwIfDisposed!=null});function c$(){return Ay("Tensor",()=>nt)}c$();var Bp=class extends nt{constructor(e,t,r,n){super(e.shape,e.dtype,e.dataId,n),this.trainable=t,this.name=r}assign(e){if(e.dtype!==this.dtype)throw new Error(`dtype of the new value (${e.dtype}) and previous value (${this.dtype}) must match`);if(!Ks(e.shape,this.shape))throw new Error(`shape of the new value (${e.shape}) and previous value (${this.shape}) must match`);la().disposeTensor(this),this.dataId=e.dataId,la().incRef(this,null)}dispose(){la().disposeVariable(this),this.isDisposedInternal=!0}};Object.defineProperty(Bp,Symbol.hasInstance,{value:e=>e instanceof nt&&e.assign!=null&&e.assign instanceof Function});var ha={};Le(ha,{assertTypesMatch:()=>Kv,getTensorsInContainer:()=>by,isTensorInList:()=>m$,makeTypesMatch:()=>Lt});var Gv=(e=>(e.R0="R0",e.R1="R1",e.R2="R2",e.R3="R3",e.R4="R4",e.R5="R5",e.R6="R6",e))(Gv||{}),jv=(e=>(e.float32="float32",e.int32="int32",e.bool="int32",e.complex64="complex64",e))(jv||{}),Hv=(e=>(e.float32="float32",e.int32="int32",e.bool="bool",e.complex64="complex64",e))(Hv||{}),qv=(e=>(e.float32="float32",e.int32="float32",e.bool="float32",e.complex64="complex64",e))(qv||{}),Xv=(e=>(e.float32="complex64",e.int32="complex64",e.bool="complex64",e.complex64="complex64",e))(Xv||{}),f$={float32:qv,int32:jv,bool:Hv,complex64:Xv};function Nr(e,t){if(e==="string"||t==="string"){if(e==="string"&&t==="string")return"string";throw new Error(`Can not upcast ${e} with ${t}`)}return f$[e][t]}function gh(e){return Nr(e,"int32")}function Lt(e,t){if(e.dtype===t.dtype)return[e,t];let r=Nr(e.dtype,t.dtype);return[e.cast(r),t.cast(r)]}function Kv(e,t){P(e.dtype===t.dtype,()=>`The dtypes of the first(${e.dtype}) and second(${t.dtype}) input must match`)}function m$(e,t){return t.some(r=>r.id===e.id)}function by(e){let t=[];return Zv(e,t,new Set),t}function Zv(e,t,r){if(e==null)return;if(e instanceof nt){t.push(e);return}if(!g$(e))return;let n=e;for(let a in n){let s=n[a];r.has(s)||(r.add(s),Zv(s,t,r))}}function g$(e){return Array.isArray(e)||typeof e=="object"}function rg(e){return e.kernelName!=null}var W3=class{constructor(){this.registeredVariables={},this.nextTapeNodeId=0,this.numBytes=0,this.numTensors=0,this.numStringTensors=0,this.numDataBuffers=0,this.gradientDepth=0,this.kernelDepth=0,this.scopeStack=[],this.numDataMovesStack=[],this.nextScopeId=0,this.tensorInfo=new WeakMap,this.profiling=!1,this.activeProfile={newBytes:0,newTensors:0,peakBytes:0,kernels:[],result:null,get kernelNames(){return Array.from(new Set(this.kernels.map(e=>e.name)))}}}dispose(){for(let e in this.registeredVariables)this.registeredVariables[e].dispose()}},yg=class{constructor(e){this.ENV=e,this.registry={},this.registryFactory={},this.pendingBackendInitId=0,this.state=new W3}async ready(){if(this.pendingBackendInit!=null)return this.pendingBackendInit.then(()=>{});if(this.backendInstance!=null)return;let e=this.getSortedBackends();for(let t=0;t<e.length;t++){let r=e[t];if(await this.initializeBackend(r).success){await this.setBackend(r);return}}throw new Error("Could not initialize any backends, all backend initializations failed.")}get backend(){if(this.pendingBackendInit!=null)throw new Error(`Backend '${this.backendName}' has not yet been initialized. Make sure to await tf.ready() or await tf.setBackend() before calling other methods`);if(this.backendInstance==null){let{name:e,asyncInit:t}=this.initializeBackendsAndReturnBest();if(t)throw new Error(`The highest priority backend '${e}' has not yet been initialized. Make sure to await tf.ready() or await tf.setBackend() before calling other methods`);this.setBackend(e)}return this.backendInstance}backendNames(){return Object.keys(this.registryFactory)}findBackend(e){if(!(e in this.registry))if(e in this.registryFactory){let{asyncInit:t}=this.initializeBackend(e);if(t)return null}else return null;return this.registry[e]}findBackendFactory(e){return e in this.registryFactory?this.registryFactory[e].factory:null}registerBackend(e,t,r=1){return e in this.registryFactory?(Cs(`${e} backend was already registered. Reusing existing backend factory.`),!1):(this.registryFactory[e]={factory:t,priority:r},!0)}async setBackend(e){if(this.registryFactory[e]==null)throw new Error(`Backend name '${e}' not found in registry`);if(this.backendName=e,this.registry[e]==null){this.backendInstance=null;let{success:t,asyncInit:r}=this.initializeBackend(e);if(!(r?await t:t))return!1}return this.backendInstance=this.registry[e],this.setupRegisteredKernels(),this.profiler=new r$(this.backendInstance),!0}setupRegisteredKernels(){Ma(this.backendName).forEach(e=>{e.setupFunc!=null&&e.setupFunc(this.backendInstance)})}disposeRegisteredKernels(e){Ma(e).forEach(t=>{t.disposeFunc!=null&&t.disposeFunc(this.registry[e])})}initializeBackend(e){let t=this.registryFactory[e];if(t==null)throw new Error(`Cannot initialize backend ${e}, no registration found.`);try{let r=t.factory();if(r&&!(r instanceof zu)&&typeof r.then=="function"){let n=++this.pendingBackendInitId,a=r.then(s=>n<this.pendingBackendInitId?!1:(this.registry[e]=s,this.pendingBackendInit=null,!0)).catch(s=>(n<this.pendingBackendInitId||(this.pendingBackendInit=null,Cs(`Initialization of backend ${e} failed`),Cs(s.stack||s.message)),!1));return this.pendingBackendInit=a,{success:a,asyncInit:!0}}else return this.registry[e]=r,{success:!0,asyncInit:!1}}catch(r){return Cs(`Initialization of backend ${e} failed`),Cs(r.stack||r.message),{success:!1,asyncInit:!1}}}removeBackend(e){if(!(e in this.registryFactory))throw new Error(`${e} backend not found in registry`);this.backendName===e&&this.pendingBackendInit!=null&&this.pendingBackendInitId++,e in this.registry&&(this.disposeRegisteredKernels(e),this.registry[e].dispose(),delete this.registry[e]),delete this.registryFactory[e],this.backendName===e&&(this.pendingBackendInit=null,this.backendName=null,this.backendInstance=null)}getSortedBackends(){if(Object.keys(this.registryFactory).length===0)throw new Error("No backend found in registry.");return Object.keys(this.registryFactory).sort((e,t)=>this.registryFactory[t].priority-this.registryFactory[e].priority)}initializeBackendsAndReturnBest(){let e=this.getSortedBackends();for(let t=0;t<e.length;t++){let r=e[t],{success:n,asyncInit:a}=this.initializeBackend(r);if(a||n)return{name:r,asyncInit:a}}throw new Error("Could not initialize any backends, all backend initializations failed.")}moveData(e,t){let r=this.state.tensorInfo.get(t),n=r.backend,a=this.readSync(t),s=n.refCount(t);n.disposeData(t,!0),r.backend=e,e.move(t,a,r.shape,r.dtype,s),this.shouldCheckForMemLeaks()&&this.state.numDataMovesStack[this.state.numDataMovesStack.length-1]++}tidy(e,t){let r=null;if(t==null){if(typeof e!="function")throw new Error("Please provide a function to tidy()");t=e}else{if(typeof e!="string"&&!(e instanceof String))throw new Error("When calling with two arguments, the first argument to tidy() must be a string");if(typeof t!="function")throw new Error("When calling with two arguments, the 2nd argument to tidy() must be a function");r=e}let n;return this.scopedRun(()=>this.startScope(r),()=>this.endScope(n),()=>(n=t(),n instanceof Promise&&console.error("Cannot return a Promise inside of tidy."),n))}scopedRun(e,t,r){e();try{let n=r();return t(),n}catch(n){throw t(),n}}nextTensorId(){return yg.nextTensorId++}nextVariableId(){return yg.nextVariableId++}clone(e){let t=B.runKernel(mi,{x:e}),r={x:e},n=s=>({x:()=>{let i="float32",o={x:s},l={dtype:i};return B.runKernel(ei,o,l)}}),a=[];return this.addTapeNode(this.state.activeScope.name,r,[t],n,a,{}),t}runKernel(e,t,r){if(this.backendName==null&&this.backend,v0(e,this.backendName)==null)throw new Error(`Kernel '${e}' not registered for backend '${this.backendName}'`);return this.runKernelFunc({kernelName:e,inputs:t,attrs:r})}shouldCheckForMemLeaks(){return this.ENV.getBool("IS_TEST")}checkKernelForMemLeak(e,t,r){let n=this.backend.numDataIds(),a=0;r.forEach(o=>{a+=o.dtype==="complex64"?3:1});let s=this.state.numDataMovesStack[this.state.numDataMovesStack.length-1],i=n-t-a-s;if(i>0)throw new Error(`Backend '${this.backendName}' has an internal memory leak (${i} data ids) after running '${e}'`)}runKernelFunc(e){let t,r=[],n=this.isTapeOn(),a=this.state.numBytes,s=this.state.numTensors;this.shouldCheckForMemLeaks()&&this.state.numDataMovesStack.push(0);let i;this.backendName==null&&this.backend;let o,l=rg(e)?e.kernelName:this.state.activeScope!=null?this.state.activeScope.name:"";if(rg(e)){let{kernelName:c,inputs:f,attrs:m}=e;this.backendName==null&&this.backend;let g=v0(c,this.backendName);P(g!=null,()=>`Cannot find registered kernel '${c}' for backend '${this.backendName}'`),i=()=>{let y=this.backend.numDataIds();o=g.kernelFunc({inputs:f,attrs:m,backend:this.backend});let A=Array.isArray(o)?o:[o];this.shouldCheckForMemLeaks()&&this.checkKernelForMemLeak(c,y,A);let x=A.map(b=>b.rank!=null?b:this.makeTensorFromTensorInfo(b));if(n){let b=this.getTensorsForGradient(c,f,x);r=this.saveTensorsForBackwardMode(b)}return x}}else{let{forwardFunc:c}=e,f=m=>{!n||(r=m.map(g=>this.keep(this.clone(g))))};i=()=>{let m=this.backend.numDataIds();o=this.tidy(()=>c(this.backend,f));let g=Array.isArray(o)?o:[o];return this.shouldCheckForMemLeaks()&&this.checkKernelForMemLeak(l,m,g),g}}let{inputs:u,attrs:d}=e,h=rg(e)?null:e.backwardsFunc,p;return this.scopedRun(()=>this.state.kernelDepth++,()=>this.state.kernelDepth--,()=>{!this.ENV.getBool("DEBUG")&&!this.state.profiling?t=i():(p=this.profiler.profileKernel(l,u,()=>i()),this.ENV.getBool("DEBUG")&&this.profiler.logKernelProfile(p),t=p.outputs)}),n&&this.addTapeNode(l,u,t,h,r,d),this.state.profiling&&this.state.activeProfile.kernels.push({name:l,bytesAdded:this.state.numBytes-a,totalBytesSnapshot:this.state.numBytes,tensorsAdded:this.state.numTensors-s,totalTensorsSnapshot:this.state.numTensors,inputShapes:Object.keys(u).map(c=>u[c]!=null?u[c].shape:null),outputShapes:t.map(c=>c.shape),kernelTimeMs:p.timeMs,extraInfo:p.extraInfo}),Array.isArray(o)?t:t[0]}saveTensorsForBackwardMode(e){return e.map(t=>this.keep(this.clone(t)))}getTensorsForGradient(e,t,r){let n=mg(e);if(n!=null){let a=n.inputsToSave||[],s=n.outputsToSave||[],i;n.saveAllInputs?(P(Array.isArray(t),()=>"saveAllInputs is true, expected inputs to be an array."),i=Object.keys(t).map(l=>t[l])):i=a.map(l=>t[l]);let o=r.filter((l,u)=>s[u]);return i.concat(o)}return[]}makeTensor(e,t,r,n){if(e==null)throw new Error("Values passed to engine.makeTensor() are null");r=r||"float32",n=n||this.backend;let a=e;r==="string"&&Ts(e[0])&&(a=e.map(o=>mh(o)));let s=n.write(a,t,r),i=new nt(t,r,s,this.nextTensorId());if(this.trackTensor(i,n),r==="string"){let o=this.state.tensorInfo.get(s),l=Mv(a);this.state.numBytes+=l-o.bytes,o.bytes=l}return i}makeTensorFromTensorInfo(e,t){let{dataId:r,shape:n,dtype:a}=e,s=new nt(n,a,r,this.nextTensorId());return this.trackTensor(s,t),s}makeVariable(e,t=!0,r,n){r=r||this.nextVariableId().toString(),n!=null&&n!==e.dtype&&(e=e.cast(n));let a=new Bp(e,t,r,this.nextTensorId());if(this.state.registeredVariables[a.name]!=null)throw new Error(`Variable with name ${a.name} was already registered`);return this.state.registeredVariables[a.name]=a,this.incRef(a,this.backend),a}trackTensor(e,t){this.state.numTensors++,e.dtype==="string"&&this.state.numStringTensors++;let r=0;e.dtype!=="complex64"&&e.dtype!=="string"&&(r=e.size*fg(e.dtype)),this.state.numBytes+=r,this.state.tensorInfo.has(e.dataId)||(this.state.numDataBuffers++,this.state.tensorInfo.set(e.dataId,{backend:t||this.backend,dtype:e.dtype,shape:e.shape,bytes:r})),e instanceof Bp||this.track(e)}incRef(e,t){this.trackTensor(e,t),this.backend.incRef(e.dataId)}removeDataId(e,t){this.state.tensorInfo.has(e)&&this.state.tensorInfo.get(e).backend===t&&(this.state.tensorInfo.delete(e),this.state.numDataBuffers--)}disposeTensor(e){if(!this.state.tensorInfo.has(e.dataId))return;let t=this.state.tensorInfo.get(e.dataId);if(this.state.numTensors--,e.dtype==="string"&&(this.state.numStringTensors--,this.state.numBytes-=t.bytes),e.dtype!=="complex64"&&e.dtype!=="string"){let r=e.size*fg(e.dtype);this.state.numBytes-=r}t.backend.disposeData(e.dataId)&&this.removeDataId(e.dataId,t.backend)}disposeVariables(){for(let e in this.state.registeredVariables){let t=this.state.registeredVariables[e];this.disposeVariable(t)}}disposeVariable(e){this.disposeTensor(e),this.state.registeredVariables[e.name]!=null&&delete this.state.registeredVariables[e.name]}memory(){let e=this.backend.memory();return e.numTensors=this.state.numTensors,e.numDataBuffers=this.state.numDataBuffers,e.numBytes=this.state.numBytes,this.state.numStringTensors>0&&(e.unreliable=!0,e.reasons==null&&(e.reasons=[]),e.reasons.push("Memory usage by string tensors is approximate (2 bytes per character)")),e}async profile(e){this.state.profiling=!0;let t=this.state.numBytes,r=this.state.numTensors;this.state.activeProfile.kernels=[],this.state.activeProfile.result=await e(),this.state.profiling=!1,this.state.activeProfile.peakBytes=Math.max(...this.state.activeProfile.kernels.map(n=>n.totalBytesSnapshot)),this.state.activeProfile.newBytes=this.state.numBytes-t,this.state.activeProfile.newTensors=this.state.numTensors-r;for(let n of this.state.activeProfile.kernels)n.kernelTimeMs=await n.kernelTimeMs,n.extraInfo=await n.extraInfo;return this.state.activeProfile}isTapeOn(){return this.state.gradientDepth>0&&this.state.kernelDepth===0}addTapeNode(e,t,r,n,a,s){let i={id:this.state.nextTapeNodeId++,kernelName:e,inputs:t,outputs:r,saved:a},o=mg(e);o!=null&&(n=o.gradFunc),n!=null&&(i.gradient=l=>(l=l.map((u,d)=>{if(u==null){let h=r[d],p=Y0(h.size,h.dtype);return this.makeTensor(p,h.shape,h.dtype)}return u}),n(l.length>1?l:l[0],a,s))),this.state.activeTape.push(i)}keep(e){return e.kept=!0,e}startTape(){this.state.gradientDepth===0&&(this.state.activeTape=[]),this.state.gradientDepth++}endTape(){this.state.gradientDepth--}startScope(e){let t={track:[],name:"unnamed scope",id:this.state.nextScopeId++};e&&(t.name=e),this.state.scopeStack.push(t),this.state.activeScope=t}endScope(e){let t=by(e),r=new Set(t.map(a=>a.id));for(let a=0;a<this.state.activeScope.track.length;a++){let s=this.state.activeScope.track[a];!s.kept&&!r.has(s.id)&&s.dispose()}let n=this.state.scopeStack.pop();this.state.activeScope=this.state.scopeStack.length===0?null:this.state.scopeStack[this.state.scopeStack.length-1],t.forEach(a=>{!a.kept&&a.scopeId===n.id&&this.track(a)})}gradients(e,t,r,n=!1){if(P(t.length>0,()=>"gradients() received an empty list of xs."),r!=null&&r.dtype!=="float32")throw new Error(`dy must have 'float32' dtype, but has '${r.dtype}'`);let a=this.scopedRun(()=>this.startTape(),()=>this.endTape(),()=>this.tidy("forward",e));P(a instanceof nt,()=>"The result y returned by f() must be a tensor.");let s=s$(this.state.activeTape,t,a);if(!n&&s.length===0&&t.length>0)throw new Error("Cannot compute gradient of y=f(x) with respect to x. Make sure that the f you passed encloses all operations that lead from x to y.");return this.tidy("backward",()=>{let i={};i[a.id]=r==null?y$(a.shape):r,i$(i,s,l=>this.tidy(l),A$);let o=t.map(l=>i[l.id]);return this.state.gradientDepth===0&&(this.state.activeTape.forEach(l=>{for(let u of l.saved)u.dispose()}),this.state.activeTape=null),{value:a,grads:o}})}customGrad(e){return P(Fs(e),()=>"The f passed in customGrad(f) must be a function."),(...t)=>{P(t.every(i=>i instanceof nt),()=>"The args passed in customGrad(f)(x1, x2,...) must all be tensors");let r,n={};t.forEach((i,o)=>{n[o]=i});let a=(i,o)=>(r=e(...t,o),P(r.value instanceof nt,()=>"The function f passed in customGrad(f) must return an object where `obj.value` is a tensor"),P(Fs(r.gradFunc),()=>"The function f passed in customGrad(f) must return an object where `obj.gradFunc` is a function."),r.value),s=(i,o)=>{let l=r.gradFunc(i,o),u=Array.isArray(l)?l:[l];P(u.length===t.length,()=>"The function f passed in customGrad(f) must return an object where `obj.gradFunc` is a function that returns the same number of tensors as inputs passed to f(...)."),P(u.every(h=>h instanceof nt),()=>"The function f passed in customGrad(f) must return an object where `obj.gradFunc` is a function that returns a list of only tensors.");let d={};return u.forEach((h,p)=>{d[p]=()=>h}),d};return this.runKernelFunc({forwardFunc:a,backwardsFunc:s,inputs:n})}}readSync(e){return this.state.tensorInfo.get(e).backend.readSync(e)}read(e){return this.state.tensorInfo.get(e).backend.read(e)}readToGPU(e,t){return this.state.tensorInfo.get(e).backend.readToGPU(e,t)}async time(e){let t=Lp(),r=await this.backend.time(e);return r.wallMs=Lp()-t,r}track(e){return this.state.activeScope!=null&&(e.scopeId=this.state.activeScope.id,this.state.activeScope.track.push(e)),e}get registeredVariables(){return this.state.registeredVariables}reset(){this.pendingBackendInitId++,this.state.dispose(),this.ENV.reset(),this.state=new W3;for(let e in this.registry)this.disposeRegisteredKernels(e),this.registry[e].dispose(),delete this.registry[e];this.backendName=null,this.backendInstance=null,this.pendingBackendInit=null}},vy=yg;vy.nextTensorId=0;vy.nextVariableId=0;function y$(e){let t=my(It(e),"float32");return B.makeTensor(t,e,"float32")}function Yv(){let e=Ov();if(e._tfengine==null){let t=new zv(e);e._tfengine=new vy(t)}return LR(e._tfengine.ENV),d$(()=>e._tfengine),e._tfengine}var B=Yv();function A$(e,t){let r={a:e,b:t};return B.runKernel(Ja,r)}var yh={};Le(yh,{isBrowser:()=>Jv,isMobile:()=>v$,mockIsMobile:()=>b$});function x$(){return typeof navigator!="undefined"&&navigator!=null}var Ag;function b$(e){Ag=e}function v$(e){if(Ag!==void 0)return Ag;if(e||x$()){if(e||(e=navigator),e.product==="ReactNative")return!0;let t=e.userAgent||e.vendor||(typeof window!="undefined"?window.opera:"");if(!t){let r=e;return r.userAgentData&&r.userAgentData.mobile}return/(android|bb\d+|meego).+mobile|avantgo|bada\/|blackberry|blazer|compal|elaine|fennec|hiptop|iemobile|ip(hone|od)|iris|kindle|lge |maemo|midp|mmp|mobile.+firefox|netfront|opera m(ob|in)i|palm( os)?|phone|p(ixi|re)\/|plucker|pocket|psp|series(4|6)0|symbian|treo|up\.(browser|link)|vodafone|wap|windows ce|xda|xiino/i.test(t)||/1207|6310|6590|3gso|4thp|50[1-6]i|770s|802s|a wa|abac|ac(er|oo|s\-)|ai(ko|rn)|al(av|ca|co)|amoi|an(ex|ny|yw)|aptu|ar(ch|go)|as(te|us)|attw|au(di|\-m|r |s )|avan|be(ck|ll|nq)|bi(lb|rd)|bl(ac|az)|br(e|v)w|bumb|bw\-(n|u)|c55\/|capi|ccwa|cdm\-|cell|chtm|cldc|cmd\-|co(mp|nd)|craw|da(it|ll|ng)|dbte|dc\-s|devi|dica|dmob|do(c|p)o|ds(12|\-d)|el(49|ai)|em(l2|ul)|er(ic|k0)|esl8|ez([4-7]0|os|wa|ze)|fetc|fly(\-|_)|g1 u|g560|gene|gf\-5|g\-mo|go(\.w|od)|gr(ad|un)|haie|hcit|hd\-(m|p|t)|hei\-|hi(pt|ta)|hp( i|ip)|hs\-c|ht(c(\-| |_|a|g|p|s|t)|tp)|hu(aw|tc)|i\-(20|go|ma)|i230|iac( |\-|\/)|ibro|idea|ig01|ikom|im1k|inno|ipaq|iris|ja(t|v)a|jbro|jemu|jigs|kddi|keji|kgt( |\/)|klon|kpt |kwc\-|kyo(c|k)|le(no|xi)|lg( g|\/(k|l|u)|50|54|\-[a-w])|libw|lynx|m1\-w|m3ga|m50\/|ma(te|ui|xo)|mc(01|21|ca)|m\-cr|me(rc|ri)|mi(o8|oa|ts)|mmef|mo(01|02|bi|de|do|t(\-| |o|v)|zz)|mt(50|p1|v )|mwbp|mywa|n10[0-2]|n20[2-3]|n30(0|2)|n50(0|2|5)|n7(0(0|1)|10)|ne((c|m)\-|on|tf|wf|wg|wt)|nok(6|i)|nzph|o2im|op(ti|wv)|oran|owg1|p800|pan(a|d|t)|pdxg|pg(13|\-([1-8]|c))|phil|pire|pl(ay|uc)|pn\-2|po(ck|rt|se)|prox|psio|pt\-g|qa\-a|qc(07|12|21|32|60|\-[2-7]|i\-)|qtek|r380|r600|raks|rim9|ro(ve|zo)|s55\/|sa(ge|ma|mm|ms|ny|va)|sc(01|h\-|oo|p\-)|sdk\/|se(c(\-|0|1)|47|mc|nd|ri)|sgh\-|shar|sie(\-|m)|sk\-0|sl(45|id)|sm(al|ar|b3|it|t5)|so(ft|ny)|sp(01|h\-|v\-|v )|sy(01|mb)|t2(18|50)|t6(00|10|18)|ta(gt|lk)|tcl\-|tdg\-|tel(i|m)|tim\-|t\-mo|to(pl|sh)|ts(70|m\-|m3|m5)|tx\-9|up(\.b|g1|si)|utst|v400|v750|veri|vi(rg|te)|vk(40|5[0-3]|\-v)|vm40|voda|vulc|vx(52|53|60|61|70|80|81|83|85|98)|w3c(\-| )|webc|whit|wi(g |nc|nw)|wmlb|wonu|x700|yas\-|your|zeto|zte\-/i.test(t.substr(0,4))}return!1}function Jv(){return typeof window!="undefined"&&window.document!=null||typeof WorkerGlobalScope!="undefined"}var jn=Z();jn.registerFlag("DEBUG",()=>!1,e=>{e&&console.warn("Debugging mode is ON. The output of every math call will be downloaded to CPU and checked for NaNs. This significantly impacts performance.")});jn.registerFlag("IS_BROWSER",()=>Jv());jn.registerFlag("IS_NODE",()=>typeof process!="undefined"&&typeof process.versions!="undefined"&&typeof process.versions.node!="undefined");jn.registerFlag("IS_CHROME",()=>typeof navigator!="undefined"&&navigator!=null&&navigator.userAgent!=null&&/Chrome/.test(navigator.userAgent)&&/Google Inc/.test(navigator.vendor));jn.registerFlag("PROD",()=>!1);jn.registerFlag("TENSORLIKE_CHECK_SHAPE_CONSISTENCY",()=>jn.getBool("DEBUG"));jn.registerFlag("DEPRECATION_WARNINGS_ENABLED",()=>!0);jn.registerFlag("IS_TEST",()=>!1);jn.registerFlag("CHECK_COMPUTATION_FOR_ERRORS",()=>!0);jn.registerFlag("WRAP_TO_IMAGEBITMAP",()=>!1);jn.registerFlag("ENGINE_COMPILE_ONLY",()=>!1);function Fa(e,t){let r=e;if(Sr(e))return t==="string"?[]:[e.length];if(!Array.isArray(e))return[];let n=[];for(;Array.isArray(r)||Sr(r)&&t!=="string";)n.push(r.length),r=r[0];return Array.isArray(e)&&Z().getBool("TENSORLIKE_CHECK_SHAPE_CONSISTENCY")&&Qv(e,n,[]),n}function Qv(e,t,r){if(r=r||[],!Array.isArray(e)&&!Sr(e)){P(t.length===0,()=>`Element arr[${r.join("][")}] is a primitive, but should be an array/TypedArray of ${t[0]} elements`);return}P(t.length>0,()=>`Element arr[${r.join("][")}] should be a primitive, but is an array of ${e.length} elements`),P(e.length===t[0],()=>`Element arr[${r.join("][")}] should have ${t[0]} elements, but has ${e.length} elements`);let n=t.slice(1);for(let a=0;a<e.length;++a)Qv(e[a],n,r.concat(a))}function V3(e,t,r,n){if(e!=="string_or_numeric"){if(e==null)throw new Error("Expected dtype cannot be null.");if(e!=="numeric"&&e!==t||e==="numeric"&&t==="string")throw new Error(`Argument '${r}' passed to '${n}' must be ${e} tensor, but got ${t} tensor`)}}function $(e,t,r,n="numeric"){if(e instanceof nt)return V3(n,e.dtype,t,r),e;let a=Z0(e);if(a!=="string"&&["bool","int32","float32"].indexOf(n)>=0&&(a=n),V3(n,a,t,r),e==null||!Sr(e)&&!Array.isArray(e)&&typeof e!="number"&&typeof e!="boolean"&&typeof e!="string"){let o=e==null?"null":e.constructor.name;throw new Error(`Argument '${t}' passed to '${r}' must be a Tensor or TensorLike, but got '${o}'`)}let s=Fa(e,a);!Sr(e)&&!Array.isArray(e)&&(e=[e]);let i=a!=="string"?Sf(e,a):Fo(e,[],!0);return B.makeTensor(i,s,a)}function Wp(e,t,r,n="numeric"){if(!Array.isArray(e))throw new Error(`Argument ${t} passed to ${r} must be a \`Tensor[]\` or \`TensorLike[]\``);return e.map((a,s)=>$(a,`${t}[${s}]`,r,n))}var e7="__op";function W(e){let t=Object.keys(e);if(t.length!==1)throw new Error(`Please provide an object with a single key (operation name) mapping to a function. Got an object with ${t.length} keys.`);let r=t[0],n=e[r];r.endsWith("_")&&(r=r.substring(0,r.length-1)),r=r+e7;let a=(...s)=>{B.startScope(r);try{let i=n(...s);return yy(i)&&console.error("Cannot return a Promise inside of tidy."),B.endScope(i),i}catch(i){throw B.endScope(null),i}};return Object.defineProperty(a,"name",{value:r,configurable:!0}),a}function w$(e,t){let r=$(e,"real","complex"),n=$(t,"imag","complex");Ur(r.shape,n.shape,`real and imag shapes, ${r.shape} and ${n.shape}, must match in call to tf.complex().`);let a={real:r,imag:n};return B.runKernel(Jp,a)}var Os=W({complex_:w$});function Gi(e,t,r,n){if(n==null&&(n=Z0(e)),n==="complex64")throw new Error("Cannot construct a complex64 tensor directly. Please use tf.complex(real, imag).");if(!Sr(e)&&!Array.isArray(e)&&typeof e!="number"&&typeof e!="boolean"&&typeof e!="string")throw new Error("values passed to tensor(values) must be a number/boolean/string or an array of numbers/booleans/strings, or a TypedArray");if(t!=null){gy(t);let a=It(t),s=It(r);P(a===s,()=>`Based on the provided shape, [${t}], the tensor should have ${a} values but has ${s}`);for(let i=0;i<r.length;++i){let o=r[i],l=i===r.length-1?o!==It(t.slice(i)):!0;P(r[i]===t[i]||!l,()=>`Error creating a new Tensor. Inferred shape (${r}) does not match the provided shape (${t}). `)}}return!Sr(e)&&!Array.isArray(e)&&(e=[e]),t=t||r,e=n!=="string"?Sf(e,n):Fo(e,[],!0),B.makeTensor(e,t,n)}function ft(e,t,r){let n=Fa(e,r);return Gi(e,t,n,r)}var xg={float32:4,float16:2,int32:4,uint16:2,uint8:1,bool:1,complex64:8},k0=4;async function k$(e,t){let r=[],n=[],a=Array.isArray(e)?e.map(i=>i.name):Object.keys(e);for(let i=0;i<a.length;++i){let o=a[i],l=Array.isArray(e)?e[i].tensor:e[o];if(l.dtype!=="float32"&&l.dtype!=="int32"&&l.dtype!=="bool"&&l.dtype!=="string"&&l.dtype!=="complex64")throw new Error(`Unsupported dtype in weight '${o}': ${l.dtype}`);let u={name:o,shape:l.shape,dtype:l.dtype};if(l.dtype==="string"){let d=new Promise(async h=>{let p=await l.bytes(),c=p.reduce((g,y)=>g+y.length,0)+k0*p.length,f=new Uint8Array(c),m=0;for(let g=0;g<p.length;g++){let y=p[g],A=new Uint8Array(new Uint32Array([y.length]).buffer);f.set(A,m),m+=k0,f.set(y,m),m+=y.length}h(f)});n.push(d)}else n.push(l.data());t!=null&&(u.group=t),r.push(u)}let s=await Promise.all(n);return{data:I$(s),specs:r}}function t7(e,t){let r={},n,a=0;for(let s of t){let i=s.name,o=s.dtype,l=s.shape,u=It(l),d;if("quantization"in s){let h=s.quantization;if(h.dtype==="uint8"||h.dtype==="uint16"){if(!("min"in h&&"scale"in h))throw new Error(`Weight ${s.name} with quantization ${h.dtype} doesn't have corresponding metadata min and scale.`)}else if(h.dtype==="float16"){if(o!=="float32")throw new Error(`Weight ${s.name} is quantized with ${h.dtype} which only supports weights of type float32 not ${o}.`)}else throw new Error(`Weight ${s.name} has unknown quantization dtype ${h.dtype}. Supported quantization dtypes are: 'uint8', 'uint16', and 'float16'.`);let p=xg[h.dtype],c=e.slice(a,a+u*p),f=h.dtype==="uint8"?new Uint8Array(c):new Uint16Array(c);if(o==="float32")if(h.dtype==="uint8"||h.dtype==="uint16"){d=new Float32Array(f.length);for(let m=0;m<f.length;m++){let g=f[m];d[m]=g*h.scale+h.min}}else if(h.dtype==="float16")n===void 0&&(n=R$()),d=n(f);else throw new Error(`Unsupported quantization type ${h.dtype} for weight type float32.`);else if(o==="int32"){if(h.dtype!=="uint8"&&h.dtype!=="uint16")throw new Error(`Unsupported quantization type ${h.dtype} for weight type int32.`);d=new Int32Array(f.length);for(let m=0;m<f.length;m++){let g=f[m];d[m]=Math.round(g*h.scale+h.min)}}else throw new Error(`Unsupported dtype in weight '${i}': ${o}`);a+=u*p}else if(o==="string"){let h=It(s.shape);d=[];for(let p=0;p<h;p++){let c=new Uint32Array(e.slice(a,a+k0))[0];a+=k0;let f=new Uint8Array(e.slice(a,a+c));d.push(f),a+=c}}else{let h=xg[o],p=e.slice(a,a+u*h);if(o==="float32")d=new Float32Array(p);else if(o==="int32")d=new Int32Array(p);else if(o==="bool")d=new Uint8Array(p);else if(o==="complex64"){d=new Float32Array(p);let c=new Float32Array(d.length/2),f=new Float32Array(d.length/2);for(let y=0;y<c.length;y++)c[y]=d[y*2],f[y]=d[y*2+1];let m=ft(c,l,"float32"),g=ft(f,l,"float32");r[i]=Os(m,g),m.dispose(),g.dispose()}else throw new Error(`Unsupported dtype in weight '${i}': ${o}`);a+=u*h}o!=="complex64"&&(r[i]=ft(d,l,o))}return r}function I$(e){if(e===null)throw new Error(`Invalid input value: ${JSON.stringify(e)}`);let t=0,r=[];e.forEach(s=>{if(t+=s.byteLength,r.push(s.byteLength===s.buffer.byteLength?s:new s.constructor(s)),!(s instanceof Float32Array||s instanceof Int32Array||s instanceof Uint8Array))throw new Error(`Unsupported TypedArray subtype: ${s.constructor.name}`)});let n=new Uint8Array(t),a=0;return r.forEach(s=>{n.set(new Uint8Array(s.buffer),a),a+=s.byteLength}),n.buffer}var wy=typeof Buffer!="undefined"&&(typeof Blob=="undefined"||typeof atob=="undefined"||typeof btoa=="undefined");function U3(e){return wy?Buffer.byteLength(e):new Blob([e]).size}function S$(e){if(wy)return Buffer.from(e).toString("base64");let t=new Uint8Array(e),r="";for(let n=0,a=t.length;n<a;n++)r+=String.fromCharCode(t[n]);return btoa(r)}function C$(e){if(wy){let n=Buffer.from(e,"base64");return n.buffer.slice(n.byteOffset,n.byteOffset+n.byteLength)}let t=atob(e),r=new Uint8Array(t.length);for(let n=0;n<t.length;++n)r.set([t.charCodeAt(n)],n);return r.buffer}function ky(e){if(e.length===1)return e[0];let t=0;e.forEach(a=>{t+=a.byteLength});let r=new Uint8Array(t),n=0;return e.forEach(a=>{r.set(new Uint8Array(a),n),n+=a.byteLength}),r.buffer}function G3(e){let t="/";for(e=e.trim();e.endsWith(t);)e=e.slice(0,e.length-1);let r=e.split(t);return r[r.length-1]}function r7(e,t){let r={modelTopology:e.modelTopology,format:e.format,generatedBy:e.generatedBy,convertedBy:e.convertedBy,weightsManifest:t};return e.signature!=null&&(r.signature=e.signature),e.userDefinedMetadata!=null&&(r.userDefinedMetadata=e.userDefinedMetadata),e.modelInitializer!=null&&(r.modelInitializer=e.modelInitializer),e.trainingConfig!=null&&(r.trainingConfig=e.trainingConfig),r}async function Iy(e,t){let r={modelTopology:e.modelTopology,format:e.format,generatedBy:e.generatedBy,convertedBy:e.convertedBy};if(e.trainingConfig!=null&&(r.trainingConfig=e.trainingConfig),e.weightsManifest!=null){let[n,a]=await t(e.weightsManifest);r.weightSpecs=n,r.weightData=a}return e.signature!=null&&(r.signature=e.signature),e.userDefinedMetadata!=null&&(r.userDefinedMetadata=e.userDefinedMetadata),e.modelInitializer!=null&&(r.modelInitializer=e.modelInitializer),r}function Ah(e){if(e.modelTopology instanceof ArrayBuffer)throw new Error("Expected JSON model topology, received ArrayBuffer.");return{dateSaved:new Date,modelTopologyType:"JSON",modelTopologyBytes:e.modelTopology==null?0:U3(JSON.stringify(e.modelTopology)),weightSpecsBytes:e.weightSpecs==null?0:U3(JSON.stringify(e.weightSpecs)),weightDataBytes:e.weightData==null?0:e.weightData.byteLength}}function T$(){let e=r=>{let n=r<<13,a=0;for(;(n&8388608)===0;)a-=8388608,n<<=1;return n&=-8388609,a+=947912704,n|a},t=new Uint32Array(2048);t[0]=0;for(let r=1;r<1024;r++)t[r]=e(r);for(let r=1024;r<2048;r++)t[r]=939524096+(r-1024<<13);return t}function N$(){let e=new Uint32Array(64);e[0]=0,e[31]=1199570944,e[32]=2147483648,e[63]=3347054592;for(let t=1;t<31;t++)e[t]=t<<23;for(let t=33;t<63;t++)e[t]=2147483648+(t-32<<23);return e}function E$(){let e=new Uint32Array(64);for(let t=0;t<64;t++)e[t]=1024;return e[0]=e[32]=0,e}function R$(){let e=T$(),t=N$(),r=E$();return n=>{let a=new ArrayBuffer(4*n.length),s=new Uint32Array(a);for(let i=0;i<n.length;i++){let o=n[i],l=e[r[o>>10]+(o&1023)]+t[o>>10];s[i]=l}return new Float32Array(a)}}var Wt=class{constructor(){this.saveRouters=[],this.loadRouters=[]}static getInstance(){return Wt.instance==null&&(Wt.instance=new Wt),Wt.instance}static registerSaveRouter(e){Wt.getInstance().saveRouters.push(e)}static registerLoadRouter(e){Wt.getInstance().loadRouters.push(e)}static getSaveHandlers(e){return Wt.getHandlers(e,"save")}static getLoadHandlers(e,t){return Wt.getHandlers(e,"load",t)}static getHandlers(e,t,r){let n=[];return(t==="load"?Wt.getInstance().loadRouters:Wt.getInstance().saveRouters).forEach(a=>{let s=a(e,r);s!==null&&n.push(s)}),n}},$$=e=>Wt.registerSaveRouter(e),M$=e=>Wt.registerLoadRouter(e),F$=e=>Wt.getSaveHandlers(e),P$=(e,t)=>Wt.getLoadHandlers(e,t),bg="tensorflowjs",vg=1,Co="models_store",Ns="model_info_store";function n7(){if(!Z().getBool("IS_BROWSER"))throw new Error("Failed to obtain IndexedDB factory because the current environmentis not a web browser.");let e=typeof window=="undefined"?self:window,t=e.indexedDB||e.mozIndexedDB||e.webkitIndexedDB||e.msIndexedDB||e.shimIndexedDB;if(t==null)throw new Error("The current browser does not appear to support IndexedDB.");return t}function wg(e){let t=e.result;t.createObjectStore(Co,{keyPath:"modelPath"}),t.createObjectStore(Ns,{keyPath:"modelPath"})}var Po=class{constructor(e){if(this.indexedDB=n7(),e==null||!e)throw new Error("For IndexedDB, modelPath must not be null, undefined or empty.");this.modelPath=e}async save(e){if(e.modelTopology instanceof ArrayBuffer)throw new Error("BrowserLocalStorage.save() does not support saving model topology in binary formats yet.");return this.databaseAction(this.modelPath,e)}async load(){return this.databaseAction(this.modelPath)}databaseAction(e,t){return new Promise((r,n)=>{let a=this.indexedDB.open(bg,vg);a.onupgradeneeded=()=>wg(a),a.onsuccess=()=>{let s=a.result;if(t==null){let i=s.transaction(Co,"readonly"),o=i.objectStore(Co).get(this.modelPath);o.onsuccess=()=>{if(o.result==null)return s.close(),n(new Error(`Cannot find model with path '${this.modelPath}' in IndexedDB.`));r(o.result.modelArtifacts)},o.onerror=l=>(s.close(),n(o.error)),i.oncomplete=()=>s.close()}else{let i=Ah(t),o=s.transaction(Ns,"readwrite"),l=o.objectStore(Ns),u=l.put({modelPath:this.modelPath,modelArtifactsInfo:i}),d;u.onsuccess=()=>{d=s.transaction(Co,"readwrite");let h=d.objectStore(Co).put({modelPath:this.modelPath,modelArtifacts:t,modelArtifactsInfo:i});h.onsuccess=()=>r({modelArtifactsInfo:i}),h.onerror=p=>{l=o.objectStore(Ns);let c=l.delete(this.modelPath);c.onsuccess=()=>(s.close(),n(h.error)),c.onerror=f=>(s.close(),n(h.error))}},u.onerror=h=>(s.close(),n(u.error)),o.oncomplete=()=>{d==null?s.close():d.oncomplete=()=>s.close()}}},a.onerror=s=>n(a.error)})}};Po.URL_SCHEME="indexeddb://";var a7=e=>Z().getBool("IS_BROWSER")&&!Array.isArray(e)&&e.startsWith(Po.URL_SCHEME)?_$(e.slice(Po.URL_SCHEME.length)):null;Wt.registerSaveRouter(a7);Wt.registerLoadRouter(a7);function _$(e){return new Po(e)}function z$(e){return e.startsWith(Po.URL_SCHEME)?e.slice(Po.URL_SCHEME.length):e}var O$=class{constructor(){this.indexedDB=n7()}async listModels(){return new Promise((e,t)=>{let r=this.indexedDB.open(bg,vg);r.onupgradeneeded=()=>wg(r),r.onsuccess=()=>{let n=r.result,a=n.transaction(Ns,"readonly"),s=a.objectStore(Ns).getAll();s.onsuccess=()=>{let i={};for(let o of s.result)i[o.modelPath]=o.modelArtifactsInfo;e(i)},s.onerror=i=>(n.close(),t(s.error)),a.oncomplete=()=>n.close()},r.onerror=n=>t(r.error)})}async removeModel(e){return e=z$(e),new Promise((t,r)=>{let n=this.indexedDB.open(bg,vg);n.onupgradeneeded=()=>wg(n),n.onsuccess=()=>{let a=n.result,s=a.transaction(Ns,"readwrite"),i=s.objectStore(Ns),o=i.get(e),l;o.onsuccess=()=>{if(o.result==null)return a.close(),r(new Error(`Cannot find model with path '${e}' in IndexedDB.`));{let u=i.delete(e),d=()=>{l=a.transaction(Co,"readwrite");let h=l.objectStore(Co).delete(e);h.onsuccess=()=>t(o.result.modelArtifactsInfo),h.onerror=p=>r(o.error)};u.onsuccess=d,u.onerror=h=>(d(),a.close(),r(o.error))}},o.onerror=u=>(a.close(),r(o.error)),s.oncomplete=()=>{l==null?a.close():l.oncomplete=()=>a.close()}},n.onerror=a=>r(n.error)})}},qa="/",mu="tensorflowjs_models",s7="info",D$="model_topology",L$="weight_specs",B$="weight_data",W$="model_metadata";function i7(e){return{info:[mu,e,s7].join(qa),topology:[mu,e,D$].join(qa),weightSpecs:[mu,e,L$].join(qa),weightData:[mu,e,B$].join(qa),modelMetadata:[mu,e,W$].join(qa)}}function o7(e){for(let t of Object.values(e))window.localStorage.removeItem(t)}function V$(e){let t=e.split(qa);if(t.length<3)throw new Error(`Invalid key format: ${e}`);return t.slice(1,t.length-1).join(qa)}function U$(e){return e.startsWith(_o.URL_SCHEME)?e.slice(_o.URL_SCHEME.length):e}var _o=class{constructor(e){if(!Z().getBool("IS_BROWSER")||typeof window=="undefined"||typeof window.localStorage=="undefined")throw new Error("The current environment does not support local storage.");if(this.LS=window.localStorage,e==null||!e)throw new Error("For local storage, modelPath must not be null, undefined or empty.");this.modelPath=e,this.keys=i7(this.modelPath)}async save(e){if(e.modelTopology instanceof ArrayBuffer)throw new Error("BrowserLocalStorage.save() does not support saving model topology in binary formats yet.");{let t=JSON.stringify(e.modelTopology),r=JSON.stringify(e.weightSpecs),n=Ah(e);try{this.LS.setItem(this.keys.info,JSON.stringify(n)),this.LS.setItem(this.keys.topology,t),this.LS.setItem(this.keys.weightSpecs,r),this.LS.setItem(this.keys.weightData,S$(e.weightData));let a={format:e.format,generatedBy:e.generatedBy,convertedBy:e.convertedBy,signature:e.signature!=null?e.signature:void 0,userDefinedMetadata:e.userDefinedMetadata!=null?e.userDefinedMetadata:void 0,modelInitializer:e.modelInitializer!=null?e.modelInitializer:void 0,trainingConfig:e.trainingConfig!=null?e.trainingConfig:void 0};return this.LS.setItem(this.keys.modelMetadata,JSON.stringify(a)),{modelArtifactsInfo:n}}catch(a){throw o7(this.keys),new Error(`Failed to save model '${this.modelPath}' to local storage: size quota being exceeded is a possible cause of this failure: modelTopologyBytes=${n.modelTopologyBytes}, weightSpecsBytes=${n.weightSpecsBytes}, weightDataBytes=${n.weightDataBytes}.`)}}}async load(){let e=JSON.parse(this.LS.getItem(this.keys.info));if(e==null)throw new Error(`In local storage, there is no model with name '${this.modelPath}'`);if(e.modelTopologyType!=="JSON")throw new Error("BrowserLocalStorage does not support loading non-JSON model topology yet.");let t={},r=JSON.parse(this.LS.getItem(this.keys.topology));if(r==null)throw new Error(`In local storage, the topology of model '${this.modelPath}' is missing.`);t.modelTopology=r;let n=JSON.parse(this.LS.getItem(this.keys.weightSpecs));if(n==null)throw new Error(`In local storage, the weight specs of model '${this.modelPath}' are missing.`);t.weightSpecs=n;let a=this.LS.getItem(this.keys.modelMetadata);if(a!=null){let i=JSON.parse(a);t.format=i.format,t.generatedBy=i.generatedBy,t.convertedBy=i.convertedBy,i.signature!=null&&(t.signature=i.signature),i.userDefinedMetadata!=null&&(t.userDefinedMetadata=i.userDefinedMetadata),i.modelInitializer!=null&&(t.modelInitializer=i.modelInitializer),i.trainingConfig!=null&&(t.trainingConfig=i.trainingConfig)}let s=this.LS.getItem(this.keys.weightData);if(s==null)throw new Error(`In local storage, the binary weight values of model '${this.modelPath}' are missing.`);return t.weightData=C$(s),t}};_o.URL_SCHEME="localstorage://";var l7=e=>Z().getBool("IS_BROWSER")&&!Array.isArray(e)&&e.startsWith(_o.URL_SCHEME)?G$(e.slice(_o.URL_SCHEME.length)):null;Wt.registerSaveRouter(l7);Wt.registerLoadRouter(l7);function G$(e){return new _o(e)}var j$=class{constructor(){P(Z().getBool("IS_BROWSER"),()=>"Current environment is not a web browser"),P(typeof window=="undefined"||typeof window.localStorage!="undefined",()=>"Current browser does not appear to support localStorage"),this.LS=window.localStorage}async listModels(){let e={},t=mu+qa,r=qa+s7;for(let n=0;n<this.LS.length;++n){let a=this.LS.key(n);if(a.startsWith(t)&&a.endsWith(r)){let s=V$(a);e[s]=JSON.parse(this.LS.getItem(a))}}return e}async removeModel(e){e=U$(e);let t=i7(e);if(this.LS.getItem(t.info)==null)throw new Error(`Cannot find model at path '${e}'`);let r=JSON.parse(this.LS.getItem(t.info));return o7(t),r}},Au="://",Nn=class{constructor(){this.managers={}}static getInstance(){return Nn.instance==null&&(Nn.instance=new Nn),Nn.instance}static registerManager(e,t){P(e!=null,()=>"scheme must not be undefined or null."),e.endsWith(Au)&&(e=e.slice(0,e.indexOf(Au))),P(e.length>0,()=>"scheme must not be an empty string.");let r=Nn.getInstance();P(r.managers[e]==null,()=>`A model store manager is already registered for scheme '${e}'.`),r.managers[e]=t}static getManager(e){let t=this.getInstance().managers[e];if(t==null)throw new Error(`Cannot find model manager for scheme '${e}'`);return t}static getSchemes(){return Object.keys(this.getInstance().managers)}};function s0(e){if(e.indexOf(Au)===-1)throw new Error(`The url string provided does not contain a scheme. Supported schemes are: ${Nn.getSchemes().join(",")}`);return{scheme:e.split(Au)[0],path:e.split(Au)[1]}}async function u7(e,t,r=!1){P(e!==t,()=>`Old path and new path are the same: '${e}'`);let n=Wt.getLoadHandlers(e);P(n.length>0,()=>`Copying failed because no load handler is found for source URL ${e}.`),P(n.length<2,()=>`Copying failed because more than one (${n.length}) load handlers for source URL ${e}.`);let a=n[0],s=Wt.getSaveHandlers(t);P(s.length>0,()=>`Copying failed because no save handler is found for destination URL ${t}.`),P(s.length<2,()=>`Copying failed because more than one (${n.length}) save handlers for destination URL ${t}.`);let i=s[0],o=s0(e).scheme,l=s0(e).path,u=o===s0(e).scheme,d=await a.load();r&&u&&await Nn.getManager(o).removeModel(l);let h=await i.save(d);return r&&!u&&await Nn.getManager(o).removeModel(l),h.modelArtifactsInfo}async function H$(){let e=Nn.getSchemes(),t={};for(let r of e){let n=await Nn.getManager(r).listModels();for(let a in n){let s=r+Au+a;t[s]=n[a]}}return t}async function q$(e){let t=s0(e);return Nn.getManager(t.scheme).removeModel(t.path)}async function X$(e,t){return u7(e,t,!1)}async function K$(e,t){return u7(e,t,!0)}var Z$=class{fetch(e,t){return fetch(e,t)}now(){return performance.now()}encode(e,t){if(t!=="utf-8"&&t!=="utf8")throw new Error(`Browser's encoder only supports utf-8, but got ${t}`);return this.textEncoder==null&&(this.textEncoder=new TextEncoder),this.textEncoder.encode(e)}decode(e,t){return new TextDecoder(t).decode(e)}};if(Z().get("IS_BROWSER")){Z().setPlatform("browser",new Z$);try{Nn.registerManager(_o.URL_SCHEME,new j$)}catch(e){}try{Nn.registerManager(Po.URL_SCHEME,new O$)}catch(e){}}var Y$={importFetch:()=>aR()},ng,J$=class{constructor(){this.util=sR(),this.textEncoder=new this.util.TextEncoder}fetch(e,t){return Z().global.fetch!=null?Z().global.fetch(e,t):(ng==null&&(ng=Y$.importFetch()),ng(e,t))}now(){let e=process.hrtime();return e[0]*1e3+e[1]/1e6}encode(e,t){if(t!=="utf-8"&&t!=="utf8")throw new Error(`Node built-in encoder only supports utf-8, but got ${t}`);return this.textEncoder.encode(e)}decode(e,t){return e.length===0?"":new this.util.TextDecoder(t).decode(e)}};Z().get("IS_NODE")&&!Z().get("IS_BROWSER")&&Z().setPlatform("node",new J$);function We(e,t="float32",r){return t=t||"float32",gy(e),new ir(e,t,r)}function Q$(e,t){let r=$(e,"x","cast");if(!$v(t))throw new Error(`Failed to cast to unknown dtype ${t}`);if(t==="string"&&r.dtype!=="string"||t!=="string"&&r.dtype==="string")throw new Error("Only strings can be casted to strings");let n={x:r},a={dtype:t};return B.runKernel(ei,n,a)}var me=W({cast_:Q$});function eM(e){let t={x:$(e,"x","clone","string_or_numeric")};return B.runKernel(mi,t)}var Wr=W({clone_:eM});function d7(e,t=!1){console.log(e.toString(t))}Yv();var tM={buffer:We,cast:me,clone:Wr,print:d7};p$(tM);var Cr={};Le(Cr,{browserFiles:()=>lM,browserHTTPRequest:()=>cM,concatenateArrayBuffers:()=>ky,copyModel:()=>X$,decodeWeights:()=>t7,encodeWeights:()=>k$,fromMemory:()=>mM,getLoadHandlers:()=>P$,getModelArtifactsForJSON:()=>Iy,getModelArtifactsInfoForJSON:()=>Ah,getSaveHandlers:()=>F$,http:()=>Cy,isHTTPScheme:()=>Ig,listModels:()=>H$,loadWeights:()=>uM,moveModel:()=>K$,registerLoadRouter:()=>M$,registerSaveRouter:()=>$$,removeModel:()=>q$,weightsLoaderFactory:()=>h7,withSaveHandler:()=>gM});var rM="model",nM=".json",aM=".weights.bin";function j3(e){return new Promise(t=>setTimeout(t)).then(e)}var kg=class{constructor(e){if(!Z().getBool("IS_BROWSER"))throw new Error("browserDownloads() cannot proceed because the current environment is not a browser.");e.startsWith(kg.URL_SCHEME)&&(e=e.slice(kg.URL_SCHEME.length)),(e==null||e.length===0)&&(e=rM),this.modelJsonFileName=e+nM,this.weightDataFileName=e+aM}async save(e){if(typeof document=="undefined")throw new Error("Browser downloads are not supported in this environment since `document` is not present");let t=window.URL.createObjectURL(new Blob([e.weightData],{type:"application/octet-stream"}));if(e.modelTopology instanceof ArrayBuffer)throw new Error("BrowserDownloads.save() does not support saving model topology in binary formats yet.");{let r=[{paths:["./"+this.weightDataFileName],weights:e.weightSpecs}],n=r7(e,r),a=window.URL.createObjectURL(new Blob([JSON.stringify(n)],{type:"application/json"})),s=this.modelJsonAnchor==null?document.createElement("a"):this.modelJsonAnchor;if(s.download=this.modelJsonFileName,s.href=a,await j3(()=>s.dispatchEvent(new MouseEvent("click"))),e.weightData!=null){let i=this.weightDataAnchor==null?document.createElement("a"):this.weightDataAnchor;i.download=this.weightDataFileName,i.href=t,await j3(()=>i.dispatchEvent(new MouseEvent("click")))}return{modelArtifactsInfo:Ah(e)}}}},I0=kg;I0.URL_SCHEME="downloads://";var sM=class{constructor(e){if(e==null||e.length<1)throw new Error(`When calling browserFiles, at least 1 file is required, but received ${e}`);this.jsonFile=e[0],this.weightsFiles=e.slice(1)}async load(){return new Promise((e,t)=>{let r=new FileReader;r.onload=n=>{let a=JSON.parse(n.target.result),s=a.modelTopology;if(s==null){t(new Error(`modelTopology field is missing from file ${this.jsonFile.name}`));return}if(a.weightsManifest==null){t(new Error(`weightManifest field is missing from file ${this.jsonFile.name}`));return}if(this.weightsFiles.length===0){e({modelTopology:s});return}let i=Iy(a,o=>this.loadWeights(o));e(i)},r.onerror=n=>t(`Failed to read model topology and weights manifest JSON from file '${this.jsonFile.name}'. BrowserFiles supports loading Keras-style tf.Model artifacts only.`),r.readAsText(this.jsonFile)})}loadWeights(e){let t=[],r=[];for(let s of e)t.push(...s.weights),r.push(...s.paths);let n=this.checkManifestAndWeightFiles(e),a=r.map(s=>this.loadWeightsFile(s,n[s]));return Promise.all(a).then(s=>[t,ky(s)])}loadWeightsFile(e,t){return new Promise((r,n)=>{let a=new FileReader;a.onload=s=>{let i=s.target.result;r(i)},a.onerror=s=>n(`Failed to weights data from file of path '${e}'.`),a.readAsArrayBuffer(t)})}checkManifestAndWeightFiles(e){let t=[],r=this.weightsFiles.map(a=>G3(a.name)),n={};for(let a of e)a.paths.forEach(s=>{let i=G3(s);if(t.indexOf(i)!==-1)throw new Error(`Duplicate file basename found in weights manifest: '${i}'`);if(t.push(i),r.indexOf(i)===-1)throw new Error(`Weight file with basename '${i}' is not provided.`);n[s]=this.weightsFiles[r.indexOf(i)]});if(t.length!==this.weightsFiles.length)throw new Error(`Mismatch in the number of files in weights manifest (${t.length}) and the number of weight files provided (${this.weightsFiles.length}).`);return n}},iM=e=>Z().getBool("IS_BROWSER")&&!Array.isArray(e)&&e.startsWith(I0.URL_SCHEME)?oM(e.slice(I0.URL_SCHEME.length)):null;Wt.registerSaveRouter(iM);function oM(e="model"){return new I0(e)}function lM(e){return new sM(e)}function H3(e,t,r,n){i(e),r=r==null?0:r,n=n==null?1:n,o(r,n);let a=0,s=l=>(l.then(u=>{let d=r+ ++a/e.length*(n-r);return t(d),u}),l);function i(l){P(l!=null&&Array.isArray(l)&&l.length>0,()=>"promises must be a none empty array")}function o(l,u){P(l>=0&&l<=1,()=>`Progress fraction must be in range [0, 1], but got startFraction ${l}`),P(u>=0&&u<=1,()=>`Progress fraction must be in range [0, 1], but got endFraction ${u}`),P(u>=l,()=>`startFraction must be no more than endFraction, but got startFraction ${l} and endFraction ${u}`)}return Promise.all(e.map(s))}async function p7(e,t){t==null&&(t={});let r=t.fetchFunc==null?Z().platform.fetch:t.fetchFunc,n=e.map(u=>r(u,t.requestInit,{isBinary:!0})),a=0,s=.5,i=(t.onProgress==null?await Promise.all(n):await H3(n,t.onProgress,a,s)).map(u=>u.arrayBuffer()),o=.5,l=1;return t.onProgress==null?await Promise.all(i):await H3(i,t.onProgress,o,l)}async function uM(e,t="",r,n){return h7(a=>p7(a,{requestInit:n}))(e,t,r)}function h7(e){return async(t,r="",n)=>{let a=t.map(()=>!1),s={},i=n!=null?n.map(()=>!1):[],o=[];if(t.forEach((c,f)=>{let m=0;c.weights.forEach(g=>{let y="quantization"in g?g.quantization.dtype:g.dtype,A=xg[y]*It(g.shape),x=()=>{a[f]=!0,s[f]==null&&(s[f]=[]),s[f].push({manifestEntry:g,groupOffset:m,sizeBytes:A})};n!=null?n.forEach((b,w)=>{b===g.name&&(x(),i[w]=!0)}):x(),o.push(g.name),m+=A})}),!i.every(c=>c)){let c=n.filter((f,m)=>!i[m]);throw new Error(`Could not find weights in manifest with names: ${c.join(", ")}.
|
|
Manifest JSON has weights with names: ${o.join(", ")}.`)}let l=a.reduce((c,f,m)=>(f&&c.push(m),c),[]),u=[];l.forEach(c=>{t[c].paths.forEach(f=>{let m=r+(r.endsWith("/")?"":"/")+f;u.push(m)})});let d=await e(u),h={},p=0;return l.forEach(c=>{let f=t[c].paths.length,m=0;for(let x=0;x<f;x++)m+=d[p+x].byteLength;let g=new ArrayBuffer(m),y=new Uint8Array(g),A=0;for(let x=0;x<f;x++){let b=new Uint8Array(d[p+x]);y.set(b,A),A+=b.byteLength}s[c].forEach(x=>{let b=g.slice(x.groupOffset,x.groupOffset+x.sizeBytes),w=t7(b,[x.manifestEntry]);for(let S in w)h[S]=w[S]}),p+=f}),h}}var dM="application/octet-stream",pM="application/json",Sy=class{constructor(e,t){if(this.DEFAULT_METHOD="POST",t==null&&(t={}),this.weightPathPrefix=t.weightPathPrefix,this.onProgress=t.onProgress,this.weightUrlConverter=t.weightUrlConverter,t.fetchFunc!=null?(P(typeof t.fetchFunc=="function",()=>"Must pass a function that matches the signature of `fetch` (see https://developer.mozilla.org/en-US/docs/Web/API/Fetch_API)"),this.fetch=t.fetchFunc):this.fetch=Z().platform.fetch,P(e!=null&&e.length>0,()=>"URL path for http must not be null, undefined or empty."),Array.isArray(e)&&P(e.length===2,()=>`URL paths for http must have a length of 2, (actual length is ${e.length}).`),this.path=e,t.requestInit!=null&&t.requestInit.body!=null)throw new Error("requestInit is expected to have no pre-existing body, but has one.");this.requestInit=t.requestInit||{}}async save(e){if(e.modelTopology instanceof ArrayBuffer)throw new Error("BrowserHTTPRequest.save() does not support saving model topology in binary formats yet.");let t=Object.assign({method:this.DEFAULT_METHOD},this.requestInit);t.body=new FormData;let r=[{paths:["./model.weights.bin"],weights:e.weightSpecs}],n=r7(e,r);t.body.append("model.json",new Blob([JSON.stringify(n)],{type:pM}),"model.json"),e.weightData!=null&&t.body.append("model.weights.bin",new Blob([e.weightData],{type:dM}),"model.weights.bin");let a=await this.fetch(this.path,t);if(a.ok)return{modelArtifactsInfo:Ah(e),responses:[a]};throw new Error(`BrowserHTTPRequest.save() failed due to HTTP response status ${a.status}.`)}async load(){let e=await this.fetch(this.path,this.requestInit);if(!e.ok)throw new Error(`Request to ${this.path} failed with status code ${e.status}. Please verify this URL points to the model JSON of the model to load.`);let t;try{t=await e.json()}catch(a){let s=`Failed to parse model JSON of response from ${this.path}.`;throw this.path.endsWith(".pb")?s+=" Your path contains a .pb file extension. Support for .pb models have been removed in TensorFlow.js 1.0 in favor of .json models. You can re-convert your Python TensorFlow model using the TensorFlow.js 1.0 conversion scripts or you can convert your.pb models with the 'pb2json'NPM script in the tensorflow/tfjs-converter repository.":s+=" Please make sure the server is serving valid JSON for this request.",new Error(s)}let r=t.modelTopology,n=t.weightsManifest;if(r==null&&n==null)throw new Error(`The JSON from HTTP path ${this.path} contains neither model topology or manifest for weights.`);return Iy(t,a=>this.loadWeights(a))}async loadWeights(e){let t=Array.isArray(this.path)?this.path[1]:this.path,[r,n]=hM(t),a=this.weightPathPrefix||r,s=[];for(let u of e)s.push(...u.weights);let i=[],o=[];for(let u of e)for(let d of u.paths)this.weightUrlConverter!=null?o.push(this.weightUrlConverter(d)):i.push(a+d+n);this.weightUrlConverter&&i.push(...await Promise.all(o));let l=await p7(i,{requestInit:this.requestInit,fetchFunc:this.fetch,onProgress:this.onProgress});return[s,ky(l)]}};Sy.URL_SCHEME_REGEX=/^https?:\/\//;function hM(e){let t=e.lastIndexOf("/"),r=e.lastIndexOf("?"),n=e.substring(0,t),a=r>t?e.substring(r):"";return[n+"/",a]}function Ig(e){return e.match(Sy.URL_SCHEME_REGEX)!=null}var c7=(e,t)=>{if(typeof fetch=="undefined"&&(t==null||t.fetchFunc==null))return null;{let r=!0;if(Array.isArray(e)?r=e.every(n=>Ig(n)):r=Ig(e),r)return Cy(e,t)}return null};Wt.registerSaveRouter(c7);Wt.registerLoadRouter(c7);function Cy(e,t){return new Sy(e,t)}function cM(e,t){return Cy(e,t)}var ag=class{constructor(e){this.modelArtifacts=e}async load(){return this.modelArtifacts}},fM=class{constructor(e){this.saveHandler=e}async save(e){return this.saveHandler(e)}};function mM(e,t,r,n){return arguments.length===1?e.modelTopology!=null||e.weightSpecs!=null?new ag(e):(console.warn("Please call tf.io.fromMemory() with only one argument. The argument should be of type ModelArtifacts. The multi-argument signature of tf.io.fromMemory() has been deprecated and will be removed in a future release."),new ag({modelTopology:e})):(console.warn("Please call tf.io.fromMemory() with only one argument. The argument should be of type ModelArtifacts. The multi-argument signature of tf.io.fromMemory() has been deprecated and will be removed in a future release."),new ag({modelTopology:e,weightSpecs:t,weightData:r,trainingConfig:n}))}function gM(e){return new fM(e)}var f7={};Le(f7,{confusionMatrix:()=>vM});function yM(e,t,r=!1,n=!1){let a=$(e,"a","matMul"),s=$(t,"b","matMul");[a,s]=Lt(a,s);let i={a,b:s},o={transposeA:r,transposeB:n};return B.runKernel(Qs,i,o)}var Ze=W({matMul_:yM});function AM(e,t,r=1,n=0){if(t<2)throw new Error(`Error in oneHot: depth must be >=2, but it is ${t}`);let a={indices:$(e,"indices","oneHot","int32")},s={depth:t,onValue:r,offValue:n};return B.runKernel(cl,a,s)}var Vp=W({oneHot_:AM});function xM(e,t){let r=$(e,"x","transpose");if(t==null&&(t=r.shape.map((s,i)=>i).reverse()),P(r.rank===t.length,()=>`Error in transpose: rank of input ${r.rank} must match length of perm ${t}.`),t.forEach(s=>{P(s>=0&&s<r.rank,()=>`All entries in 'perm' must be between 0 and ${r.rank-1} but got ${t}`)}),r.rank<=1)return r.clone();let n={x:r},a={perm:t};return B.runKernel(Vi,n,a)}var et=W({transpose_:xM});function bM(e,t,r){let n=$(e,"labels","confusionMatrix"),a=$(t,"predictions","confusionMatrix");P(r==null||r>0&&Number.isInteger(r),()=>`If provided, numClasses must be a positive integer, but got ${r}`),P(n.rank===1,()=>`Expected the rank of labels to be 1, but got ${n.rank}`),P(a.rank===1,()=>`Expected the rank of predictions to be 1, but got ${a.rank}`),P(n.shape[0]===a.shape[0],()=>`Mismatch in the number of examples: ${n.shape[0]} vs. ${a.shape[0]}. Labels and predictions should have the same number of elements.`),P(r>0&&Number.isInteger(r),()=>`numClasses is required to be a positive integer, but got ${r}`);let s=Vp(me(n,"int32"),r),i=Vp(me(a,"int32"),r),o=et(s),l=Ze(o,i);return me(l,"int32")}var vM=W({confusionMatrix_:bM}),ji={};Le(ji,{assertAndGetBroadcastShape:()=>bt,getBroadcastDims:()=>m7,getReductionAxes:()=>Jt});function m7(e,t){let r=e.length,n=[];for(let a=0;a<r;a++){let s=r-1-a,i=e[s]||1;(t[t.length-1-a]||1)>1&&i===1&&n.unshift(s)}return n}function Jt(e,t){let r=[];for(let n=0;n<t.length;n++){let a=e[e.length-n-1],s=t.length-n-1,i=t[s];(a==null||a===1&&i>1)&&r.unshift(s)}return r}function bt(e,t){let r=[],n=Math.max(e.length,t.length);for(let a=0;a<n;a++){let s=e[e.length-a-1];s==null&&(s=1);let i=t[t.length-a-1];if(i==null&&(i=1),s===1)r.unshift(i);else if(i===1)r.unshift(s);else if(s!==i){let o=`Operands could not be broadcast together with shapes ${e} and ${t}.`;throw Error(o)}else r.unshift(s)}return r}var On={};Le(On,{fromPixels:()=>NM,fromPixelsAsync:()=>CM,toPixels:()=>TM});function g7(e,t,r){if(jo(e),t!=null&&t.length!==3)throw new Error("tensor3d() requires shape to have three numbers");let n=Fa(e,r);if(n.length!==3&&n.length!==1)throw new Error("tensor3d() requires values to be number[][][] or flat/TypedArray");if(n.length===1&&t==null)throw new Error("tensor3d() requires shape to be provided when `values` are a flat array");return Gi(e,t,n,r)}var yo;function y7(e,t=3){if(t>4)throw new Error("Cannot construct Tensor with more than 4 channels from pixels.");if(e==null)throw new Error("pixels passed to tf.browser.fromPixels() can not be null");let r=!1,n=!1,a=!1,s=!1,i=!1,o=!1;if(e.data instanceof Uint8Array)r=!0;else if(typeof ImageData!="undefined"&&e instanceof ImageData)n=!0;else if(typeof HTMLVideoElement!="undefined"&&e instanceof HTMLVideoElement)a=!0;else if(typeof HTMLImageElement!="undefined"&&e instanceof HTMLImageElement)s=!0;else if(e.getContext!=null)i=!0;else if(typeof ImageBitmap!="undefined"&&e instanceof ImageBitmap)o=!0;else throw new Error(`pixels passed to tf.browser.fromPixels() must be either an HTMLVideoElement, HTMLImageElement, HTMLCanvasElement, ImageData in browser, or OffscreenCanvas, ImageData in webworker or {data: Uint32Array, width: number, height: number}, but was ${e.constructor.name}`);if(a&&a&&e.readyState<2)throw new Error("The video element has not loaded data yet. Please wait for `loadeddata` event on the <video> element.");if(v0(Op,B.backendName)!=null){let p={pixels:e},c={numChannels:t};return B.runKernel(Op,p,c)}let[l,u]=a?[e.videoWidth,e.videoHeight]:[e.width,e.height],d;if(i)d=e.getContext("2d").getImageData(0,0,l,u).data;else if(n||r)d=e.data;else if(s||a||o){if(yo==null)if(typeof document=="undefined")if(typeof OffscreenCanvas!="undefined"&&typeof OffscreenCanvasRenderingContext2D!="undefined")yo=new OffscreenCanvas(1,1).getContext("2d");else throw new Error("Cannot parse input in current context. Reason: OffscreenCanvas Context2D rendering is not supported.");else yo=document.createElement("canvas").getContext("2d");yo.canvas.width=l,yo.canvas.height=u,yo.drawImage(e,0,0,l,u),d=yo.getImageData(0,0,l,u).data}let h;if(t===4)h=new Int32Array(d);else{let p=l*u;h=new Int32Array(p*t);for(let c=0;c<p;c++)for(let f=0;f<t;++f)h[c*t+f]=d[c*4+f]}return g7(h,[u,l,t],"int32")}function wM(e){return e!=null&&e.data instanceof Uint8Array}function kM(){return typeof window!="undefined"&&typeof ImageBitmap!="undefined"&&window.hasOwnProperty("createImageBitmap")}function IM(e){return e!=null&&e.width!==0&&e.height!==0}function SM(e){return kM()&&!(e instanceof ImageBitmap)&&IM(e)&&!wM(e)}async function CM(e,t=3){let r=null;if(Z().getBool("WRAP_TO_IMAGEBITMAP")&&SM(e)){let n;try{n=await createImageBitmap(e,{premultiplyAlpha:"none"})}catch(a){n=null}n!=null&&n.width===e.width&&n.height===e.height?r=n:r=e}else r=e;return y7(r,t)}async function TM(e,t){let r=$(e,"img","toPixels");if(!(e instanceof nt)){let u=r;r=me(u,"int32"),u.dispose()}if(r.rank!==2&&r.rank!==3)throw new Error(`toPixels only supports rank 2 or 3 tensors, got rank ${r.rank}.`);let[n,a]=r.shape.slice(0,2),s=r.rank===2?1:r.shape[2];if(s>4||s===2)throw new Error(`toPixels only supports depth of size 1, 3 or 4 but got ${s}`);if(r.dtype!=="float32"&&r.dtype!=="int32")throw new Error(`Unsupported type for toPixels: ${r.dtype}. Please use float32 or int32 tensors.`);let i=await r.data(),o=r.dtype==="float32"?255:1,l=new Uint8ClampedArray(a*n*4);for(let u=0;u<n*a;++u){let d=[0,0,0,255];for(let p=0;p<s;p++){let c=i[u*s+p];if(r.dtype==="float32"){if(c<0||c>1)throw new Error(`Tensor values for a float32 Tensor must be in the range [0 - 1] but encountered ${c}.`)}else if(r.dtype==="int32"&&(c<0||c>255))throw new Error(`Tensor values for a int32 Tensor must be in the range [0 - 255] but encountered ${c}.`);s===1?(d[0]=c*o,d[1]=c*o,d[2]=c*o):d[p]=c*o}let h=u*4;l[h+0]=Math.round(d[0]),l[h+1]=Math.round(d[1]),l[h+2]=Math.round(d[2]),l[h+3]=Math.round(d[3])}if(t!=null){t.width=a,t.height=n;let u=t.getContext("2d"),d=new ImageData(l,a,n);u.putImageData(d,0,0)}return r!==e&&r.dispose(),l}var NM=W({fromPixels_:y7}),Ty={};Le(Ty,{prepareAndValidate:()=>A7});function A7(e,t){let r=e.shape.length,n=t.shape.length;if(r<1)throw new Error(`tf.gatherND() expects the input to be rank 1 or higher, but the rank was ${r}.`);if(n<1)throw new Error(`tf.gatherND() expects the indices to be rank 1 or higher, but the rank was ${n}.`);if(t.dtype!=="int32")throw new Error(`tf.gatherND() expects the indices to be int32 type, but the dtype was ${t.dtype}.`);if(t.shape[n-1]>r)throw new Error(`index innermost dimension length must be <= tensor rank; saw: ${t.shape[n-1]} vs. ${r}`);if(It(e.shape)===0)throw new Error(`Requested more than 0 entries, but input is empty. Input shape: ${e.shape}.`);let a=t.shape,s=a[a.length-1],i=1;for(let h=0;h<a.length-1;++h)i*=a[h];let o=e.shape,l=a.slice();l.pop();let u=1;for(let h=s;h<r;++h)u*=o[h],l.push(o[h]);let d=[...Ou(e.shape).map(h=>h/u),1].slice(0,s);return[l,i,u,d]}var Ny={};Le(Ny,{calculateShapes:()=>x7,validateInput:()=>Ry,validateUpdateShape:()=>Ey});function Ey(e,t,r){let n=t.rank>1?t.shape[t.rank-1]:1,a=t.rank>1?t.rank-1:1,s=`Must have updates.shape = indices.shape[:batchDim] + shape[sliceDim:], got updates.shape: ${r.shape}, indices.shape: ${t.shape}, shape: ${e}, sliceDim: ${n}, and batchDim: ${a}.`;if(r.rank<a)throw new Error(s+` update.rank < ${a}. `);if(e.length<n+(r.rank-a))throw new Error(s+` Output shape length < ${n+(r.rank-a)}`);if(r.rank!==a+e.length-n)throw new Error(s+` update.rank != ${a+e.length-n}`);for(let i=0;i<a;++i)if(r.shape[i]!==t.shape[i])throw new Error(s+` updates.shape[${i}] (${r.shape[i]}) != indices.shape[${i}] (${t.shape[i]}).`);for(let i=0;i<r.rank-a;++i)if(r.shape[i+a]!==e[i+n])throw new Error(s+` updates.shape[${i+a}] (${r.shape[i+a]}) != shape[${i+a}] (${e[i+a]})`)}function Ry(e,t,r){if(t.rank<1)throw new Error(`tf.scatterND() expects the indices to be rank 1 or higher, but the rank was ${t.rank}.`);if(e.rank<1)throw new Error(`tf.scatterND() expects the updates to be rank 1 or higher, but the rank was ${e.rank}.`);if(t.dtype!=="int32")throw new Error(`The dtype of 'indices' should be int32, but got dtype: ${t.dtype}`);if(r.length<1)throw new Error(`Output rank must be greater or equal to 1, but got shape: ${r}`);if(r.length===0){if(t.size===0)throw new Error(`Indices specified for empty output. indices shape: ${t.shape}`);if(e.size===0)throw new Error(`Updates specified for empty output. updates shape: ${e.shape}`)}Ey(r,t,e)}function x7(e,t,r){let n=t.shape.length,a=n>1?t.shape[n-1]:1,s=r.length,i=1;for(let h=a;h<s;++h)i*=r[h];let o=a<1?1:a,l=It(t.shape)/o,u=[...Ou(r.slice(0,a)),1],d=It(r);return{sliceRank:a,numUpdates:l,sliceSize:i,strides:u,outputSize:d}}var Ot={};Le(Ot,{assertParamsValid:()=>RM,computeFlatOffset:()=>_M,computeOutShape:()=>MM,getNormalizedAxes:()=>FM,isSliceContinous:()=>PM,maskToAxes:()=>$M,parseSliceParams:()=>N7,sliceInfo:()=>zM,startForAxis:()=>C7,startIndicesWithElidedDims:()=>k7,stopForAxis:()=>T7,stopIndicesWithElidedDims:()=>I7,stridesForAxis:()=>S7,stridesWithElidedDims:()=>b7});var Sg=-2,EM=-1;function RM(e,t,r){let n=e.shape.length;P(n===t.length,()=>`Error in slice${n}D: Length of begin ${t} must match the rank of the array (${n}).`),P(n===r.length,()=>`Error in slice${n}D: Length of size ${r} must match the rank of the array (${n}).`);for(let a=0;a<n;++a)P(t[a]+r[a]<=e.shape[a],()=>`Error in slice${n}D: begin[${a}] + size[${a}] (${t[a]+r[a]}) would overflow input.shape[${a}] (${e.shape[a]})`)}function $M(e){let t=[],r=0;for(;e>0;)e&1&&t.push(r),e/=2,r++;return t}function MM(e,t,r){let n=[];for(let a=0;a<e.length;a++)n[a]=Math.ceil((t[a]-e[a])/r[a]);return n}function b7(e,t,r,n){let a=[...e];for(let s=a.length;s<n.length;s++)a.push(1);for(let s=0;s<r;s++)s===0?a[t]=1:(a.splice(t,0,1),a.pop());return a}function v7(e,t,r){return r<=e?r:r-(t-1)}function w7(e,t){let r=[];for(let n=0;n<e;n++)r.push(t+n);return r}function FM(e,t,r,n,a,s,i,o,l){let u=e.length,d=new Array(u),h=new Array(u),p=new Array(u);if(t.length&&r>0){let c=t[0],f=r+1;d=k7(i,c,f,n,e),h=I7(o,c,f,a,e),p=b7(s,c,f,e)}else for(let c=0;c<u;c++)d[c]=C7(i,n,s,e,c,l),h[c]=T7(o,a,s,e,c,l),p[c]=S7(s,c,l);return{begin:d,end:h,strides:p}}function k7(e,t,r,n,a){let s=[...a],i=w7(r,t);for(let o=0;o<s.length;o++)if(i.indexOf(o)>-1)s[o]=0;else{let l=v7(t,r,o),u=n[l];e&1<<l&&(u=0),s[o]=u}return s}function I7(e,t,r,n,a){let s=[...a],i=w7(r,t);for(let o=0;o<s.length;o++)if(i.indexOf(o)>-1)s[o]=Number.MAX_SAFE_INTEGER;else{let l=v7(t,r,o),u=n[l];e&1<<l&&(u=Number.MAX_SAFE_INTEGER),s[o]=u}for(let o=0;o<s.length;o++){let l=a[o];s[o]<0&&(s[o]+=l),s[o]=zp(0,s[o],a[o])}return s}function S7(e,t,r){let n=e[t];return(r&1<<t||n==null)&&(n=1),n}function C7(e,t,r,n,a,s){let i=t[a],o=r[a]||1;(e&1<<a||s&1<<a||i==null)&&(o>0?i=Number.MIN_SAFE_INTEGER:i=Number.MAX_SAFE_INTEGER);let l=n[a];return i<0&&(i+=l),i=zp(0,i,l-1),i}function T7(e,t,r,n,a,s){let i=t[a],o=r[a]||1;(e&1<<a||s&1<<a||i==null)&&(o>0?i=Number.MAX_SAFE_INTEGER:i=Number.MIN_SAFE_INTEGER);let l=n[a];return i<0&&(i+=l),o>0?i=zp(0,i,l):i=zp(-1,i,l-1),i}function PM(e,t,r){let n=r.length;for(let a=0;a<r.length;a++)if(r[a]>1){n=a;break}for(let a=n+1;a<r.length;a++)if(t[a]>0||r[a]!==e[a])return!1;return!0}function _M(e,t){let r=e.length>0?e[e.length-1]:1;for(let n=0;n<e.length-1;n++)r+=e[n]*t[n];return r}function N7(e,t,r){let n,a=e.shape.length;typeof t=="number"?n=[t,...new Array(a-1).fill(0)]:t.length<a?n=t.concat(new Array(a-t.length).fill(0)):n=t.slice(),n.forEach(i=>{P(i!==-1,()=>"slice() does not support negative begin indexing.")});let s;return r==null?s=new Array(a).fill(-1):typeof r=="number"?s=[r,...new Array(a-1).fill(-1)]:r.length<a?s=r.concat(new Array(a-r.length).fill(-1)):s=r,s=s.map((i,o)=>i>=0?i:(P(i===-1,()=>`Negative size values should be exactly -1 but got ${i} for the slice() size at index ${o}.`),e.shape[o]-n[o])),[n,s]}function zM(e,t,r,n,a,s,i,o,l){let u;if(n==null?(u=new Array(t.length),u.fill(1)):u=n,i!=null&&(i&i-1)!==0)throw new Error("Multiple ellipses in slice is not allowed.");let d=!1,h={dims:u.length,numAddAxisAfterEllipsis:0,begin:t.slice(),end:r.slice(),strides:u.slice(),beginMask:a,endMask:s,ellipsisMask:i,newAxisMask:o,shrinkAxisMask:l};for(let A=0;A<h.dims;A++)d&&(1<<A&o)!==0&&h.numAddAxisAfterEllipsis++,1<<A&i&&(d=!0);d||(h.ellipsisMask|=1<<h.dims,h.dims++);let p={dims:e.length,beginMask:0,endMask:0,beginValid:!1,endValid:!1};OM(h,p);let c=!0,f=!0,m=!0,g=[],y=[];for(let A=0;A<e.length;++A){if(p.strides[A]===0)throw Error(`strides[${A}] must be non-zero`);let x=!!(p.shrinkAxisMask&1<<A),b=e[A];if(b===-1){g.push(x?1:-1);continue}let w=[p.beginMask&1<<A,p.endMask&1<<A],S=[p.strides[A]>0?0:-1,p.strides[A]>0?b:b-1];if(x&&p.strides[A]<=0)throw Error("only stride 1 allowed on non-range indexing.");m=m&&p.strides[A]===1;let N=!!(p.beginMask&1<<A&&p.endMask&1<<A);if(p.beginValid&&p.endValid){if(x){let M=p.begin[A]<0?b+p.begin[A]:p.begin[A];if(p.begin[A]=M,p.end[A]=p.begin[A]+1,M<0||M>=b)throw Error(`slice index ${p.begin[A]} of dimension ${A} out of bounds.`)}else p.begin[A]=q3(p.begin[A],0,p.strides[A],b,w,S),p.end[A]=q3(p.end[A],1,p.strides[A],b,w,S);let z=p.strides[A]===1&&p.begin[A]===0&&p.end[A]===b;c=c&&z,f=f&&(A===0&&p.strides[A]===1||z)}else c=c&&p.strides[A]===1&&N,f=f&&(A===0&&p.strides[A]===1||N);let E,R=!1;if(p.beginValid&&p.endValid?(E=p.end[A]-p.begin[A],R=!0):x?(E=1,R=!0):N&&b>=0&&(p.strides[A]<0?E=-b:E=b,R=!0),R){let z;E===0||E<0!=p.strides[A]<0?z=0:z=Math.trunc(E/p.strides[A])+(E%p.strides[A]!==0?1:0),g.push(z)}else g.push(-1)}for(let A=0;A<p.finalShapeGatherIndices.length;++A){let x=p.finalShapeGatherIndices[A];x>=0?y.push(g[x]):x===Sg&&y.push(1)}return{finalShapeSparse:y.filter((A,x)=>p.finalShapeGatherIndices[x]!==Sg),finalShape:y,isIdentity:c,sliceDim0:f,isSimpleSlice:m,begin:p.begin,end:p.end,strides:p.strides}}function OM(e,t){t.beginMask=0,t.endMask=0,t.shrinkAxisMask=0;let r=0;t.beginValid=e.begin!=null,t.endValid=e.end!=null,t.begin=new Array(t.dims),t.end=new Array(t.dims),t.strides=new Array(t.dims),t.finalShapeGatherIndices=[],t.finalShapeGatherIndicesSparse=[],t.inputShapeGatherIndicesSparse=new Array(t.dims);for(let n=0;n<e.dims;n++)if(1<<n&e.ellipsisMask){let a=Math.min(t.dims-(e.dims-n)+1+e.numAddAxisAfterEllipsis,t.dims);for(;r<a;r++)t.begin[r]=0,t.end[r]=0,t.strides[r]=1,t.beginMask|=1<<r,t.endMask|=1<<r,t.finalShapeGatherIndices.push(r),t.finalShapeGatherIndicesSparse.push(-1),t.inputShapeGatherIndicesSparse[r]=n}else if(1<<n&e.newAxisMask)t.finalShapeGatherIndices.push(Sg),t.finalShapeGatherIndicesSparse.push(-1);else{if(r===t.begin.length)throw Error(`Index out of range using input dim ${r}; input has only ${t.dims} dims, ${t.begin.length}.`);e.begin!=null&&(t.begin[r]=e.begin[n]),e.end!=null&&(t.end[r]=e.end[n]),t.strides[r]=e.strides[n],e.beginMask&1<<n&&(t.beginMask|=1<<r),e.endMask&1<<n&&(t.endMask|=1<<r),e.shrinkAxisMask&1<<n?(t.finalShapeGatherIndices.push(EM),t.finalShapeGatherIndicesSparse.push(-1),t.shrinkAxisMask|=1<<r):(t.finalShapeGatherIndices.push(r),t.finalShapeGatherIndicesSparse.push(n)),t.inputShapeGatherIndicesSparse[r]=n,r++}}function q3(e,t,r,n,a,s){if(a[t])return r>0?s[t]:s[t+1&1];{let i=e<0?n+e:e;return i<s[0]?s[0]:i>s[1]?s[1]:i}}var ue={};Le(ue,{Serializable:()=>E7,SerializationMap:()=>ko,registerClass:()=>Hi});var E7=class{getClassName(){return this.constructor.className}static fromConfig(e,t){return new e(t)}},ko=class{constructor(){this.classNameMap={}}static getMap(){return ko.instance==null&&(ko.instance=new ko),ko.instance}static register(e){ko.getMap().classNameMap[e.className]=[e,e.fromConfig]}};function Hi(e){P(e.className!=null,()=>"Class being registered does not have the static className property defined."),P(typeof e.className=="string",()=>"className is required to be a string, but got type "+typeof e.className),P(e.className.length>0,()=>"Class being registered has an empty-string as its className, which is disallowed."),ko.register(e)}var R7={};Le(R7,{TEST_EPSILON_FLOAT16:()=>$7,encodeStrings:()=>M7,expectArrayBuffersEqual:()=>GM,expectArraysClose:()=>LM,expectArraysEqual:()=>WM,expectNumbersClose:()=>VM,expectPromiseToFail:()=>BM,expectValuesInRange:()=>UM,testEpsilon:()=>$y});var DM=.001,$7=.1;function LM(e,t,r){return r==null&&(r=$y()),Cg(e,t,(n,a)=>My(n,a,r))}function $y(){return B.backend.floatPrecision()===32?DM:$7}function Cg(e,t,r){let n=!0;if((Sr(e)||Sr(t))&&(n=!1),Sr(e)&&Sr(t)&&(n=!0),n){let i=e.constructor.name,o=t.constructor.name;if(i!==o)throw new Error(`Arrays are of different type. Actual: ${i}. Expected: ${o}`)}if(Array.isArray(e)&&Array.isArray(t)){let i=Fa(e),o=Fa(t);if(!Ks(i,o))throw new Error(`Arrays have different shapes. Actual: [${i}]. Expected: [${o}]`)}let a=Sr(e)?e:Fo(e),s=Sr(t)?t:Fo(t);if(a.length!==s.length)throw new Error(`Arrays have different lengths actual: ${a.length} vs expected: ${s.length}.
|
|
Actual: ${a}.
|
|
Expected: ${s}.`);for(let i=0;i<s.length;++i){let o=a[i],l=s[i];if(!r(o,l))throw new Error(`Arrays differ: actual[${i}] = ${o}, expected[${i}] = ${l}.
|
|
Actual: ${a}.
|
|
Expected: ${s}.`)}}function BM(e,t){e().then(()=>t.fail(),()=>t())}function WM(e,t){let r=typeof t=="string"||typeof t=="number"||typeof t=="boolean"?[t]:t;return Ts(e)||Ts(e[0])||Ts(t)||Ts(t[0])?Cg(e,r,(n,a)=>n==a):Cg(e,t,(n,a)=>My(n,a,0))}function VM(e,t,r){if(r==null&&(r=$y()),!My(e,t,r))throw new Error(`Numbers differ: actual === ${e}, expected === ${t}`)}function My(e,t,r){return!isFinite(e)&&!isFinite(t)?!0:!(isNaN(e)||isNaN(t)||Math.abs(e-t)>r)}function UM(e,t,r){for(let n=0;n<e.length;n++)if(e[n]<t||e[n]>r)throw new Error(`Value out of range:${e[n]} low: ${t}, high: ${r}`)}function GM(e,t){let r=new Float32Array(e),n=new Float32Array(t);if(r.length!==n.length)throw new Error(`Expected ArrayBuffer to be of length ${n.length}, but it was ${r.length}`);for(let a=0;a<n.length;a++)if(r[a]!==n[a])throw new Error(`Expected ArrayBuffer value at ${a} to be ${n[a]} but got ${r[a]} instead`)}function M7(e){for(let t=0;t<e.length;t++){let r=e[t];Array.isArray(r)?M7(r):e[t]=mh(r)}return e}var Fy="0.0.0";function Py(){Z().set("PROD",!0)}function jM(){Z().set("DEBUG",!0)}function HM(){Z().set("DEPRECATION_WARNINGS_ENABLED",!1),console.warn("TensorFlow.js deprecation warnings have been disabled.")}function _y(e){Z().getBool("DEPRECATION_WARNINGS_ENABLED")&&console.warn(e+" You can disable deprecation warnings with tf.disableDeprecationWarnings().")}h$(_y);function qM(){B.disposeVariables()}function Xt(){return B}function S0(){return B.memory()}function XM(e){return B.profile(e)}function X(e,t){return B.tidy(e,t)}function te(e){by(e).forEach(t=>t.dispose())}function gr(e){return B.keep(e)}function KM(e){return B.time(e)}function zy(e){return B.setBackend(e)}function pd(){return B.ready()}function Gr(){return B.backendName}function ZM(e){B.removeBackend(e)}function Oy(e){return B.findBackend(e)}function YM(e){return B.findBackendFactory(e)}function $l(e,t,r=1){return B.registerBackend(e,t,r)}function Dn(){return B.backend}function JM(e,t){Z().setPlatform(e,t)}function QM(e,t){let r=$(e,"a","add"),n=$(t,"b","add");[r,n]=Lt(r,n);let a={a:r,b:n};return B.runKernel(Ja,a)}var le=W({add_:QM});function eF(e,t){let r=$(e,"a","floorDiv"),n=$(t,"b","floorDiv");[r,n]=Lt(r,n);let a={a:r,b:n};return B.runKernel(hi,a)}var xh=W({floorDiv_:eF});function tF(e,t){let r=$(e,"a","div"),n=$(t,"b","div");if([r,n]=Lt(r,n),r.dtype==="int32"&&n.dtype==="int32")return xh(r,n);let a={a:r,b:n},s={};return B.runKernel(li,a,s)}var pe=W({div_:tF});function rF(e,t){let r=$(e,"a","mul"),n=$(t,"b","mul");[r,n]=Lt(r,n);let a={a:r,b:n};return B.runKernel(Si,a)}var L=W({mul_:rF});function nF(e){let t=$(e,"x","abs");if(t.dtype==="complex64"){let r={x:t};return B.runKernel(Qp,r)}else{let r={x:t};return B.runKernel(Ho,r)}}var ar=W({abs_:nF});function aF(e){let t={x:$(e,"x","acos")};return B.runKernel(Du,t)}var F7=W({acos_:aF});function sF(e){let t={x:$(e,"x","acosh")};return B.runKernel(Lu,t)}var P7=W({acosh_:sF});function iF(e){P(Array.isArray(e),()=>"The argument passed to tf.addN() must be a list of tensors"),P(e.length>=1,()=>`Must pass at least one tensor to tf.addN(), but got ${e.length}`);let t=e.map((a,s)=>$(a,`tensors${s}`,"addN")),r=t[0];t.forEach(a=>{if(a.dtype!==r.dtype)throw new Error("All tensors passed to tf.addN() must have the same dtype")}),t.forEach(a=>{if(!Ks(a.shape,r.shape))throw new Error("All tensors passed to tf.addN() must have the same shape")});let n=t;return B.runKernel(Zs,n)}var Cf=W({addN_:iF});function oF(e,t=null,r=!1){let n={x:$(e,"x","all","bool")},a={axis:t,keepDims:r};return B.runKernel(Bu,n,a)}var Dy=W({all_:oF});function lF(e,t=null,r=!1){let n={x:$(e,"x","any","bool")},a={axis:t,keepDims:r};return B.runKernel(Wu,n,a)}var C0=W({any_:lF});function uF(e,t=0){let r={x:$(e,"x","argMax")},n={axis:t};return B.runKernel(Ys,r,n)}var Rn=W({argMax_:uF});function dF(e,t=0){let r={x:$(e,"x","argMin")},n={axis:t};return B.runKernel(Vu,r,n)}var _7=W({argMin_:dF});function pF(e){let t={x:$(e,"x","asin")};return B.runKernel(Uu,t)}var z7=W({asin_:pF});function hF(e){let t={x:$(e,"x","asinh")};return B.runKernel(Gu,t)}var O7=W({asinh_:hF});function cF(e){let t={x:$(e,"x","atan")};return B.runKernel(ju,t)}var D7=W({atan_:cF});function fF(e,t){let r=$(e,"a","atan2"),n=$(t,"b","atan2");[r,n]=Lt(r,n);let a={a:r,b:n};return B.runKernel(qu,a)}var L7=W({atan2_:fF});function mF(e){let t={x:$(e,"x","atanh")};return B.runKernel(Hu,t)}var B7=W({atanh_:mF});function gF(e,t,r,n,a="NHWC",s){let i=e[3],o=[...t,i],l=U7(a);return bh(e,o,r,s,n,null,null,l)}function W7(e,t,r,n,a,s,i="channelsLast"){let[o,l]=T0(t),u;if(i==="channelsLast")u=[o,l,e[3],e[3]];else if(i==="channelsFirst")u=[o,l,e[1],e[1]];else throw new Error(`Unknown dataFormat ${i}`);return bh(e,u,r,n,a,s,!1,i)}function yF(e,t,r,n,a,s,i="NDHWC"){let[o,l,u]=Tg(t),d,h;if(i==="NDHWC")h="channelsLast",d=[o,l,u,e[4],e[4]];else if(i==="NCDHW")h="channelsFirst",d=[o,l,u,e[1],e[1]];else throw new Error(`Unknown dataFormat ${i}`);return V7(e,d,r,n,a,!1,h,s)}function bh(e,t,r,n,a,s,i=!1,o="channelsLast"){let[l,u,d,h]=[-1,-1,-1,-1];if(o==="channelsLast")[l,u,d,h]=e;else if(o==="channelsFirst")[l,h,u,d]=e;else throw new Error(`Unknown dataFormat ${o}`);let[p,c,,f]=t,[m,g]=T0(r),[y,A]=T0(n),x=xu(p,y),b=xu(c,A),{padInfo:w,outHeight:S,outWidth:N}=bF(a,u,d,m,g,x,b,s,o),E=i?f*h:f,R;return o==="channelsFirst"?R=[l,E,S,N]:o==="channelsLast"&&(R=[l,S,N,E]),{batchSize:l,dataFormat:o,inHeight:u,inWidth:d,inChannels:h,outHeight:S,outWidth:N,outChannels:E,padInfo:w,strideHeight:m,strideWidth:g,filterHeight:p,filterWidth:c,effectiveFilterHeight:x,effectiveFilterWidth:b,dilationHeight:y,dilationWidth:A,inShape:e,outShape:R,filterShape:t}}function V7(e,t,r,n,a,s=!1,i="channelsLast",o){let[l,u,d,h,p]=[-1,-1,-1,-1,-1];if(i==="channelsLast")[l,u,d,h,p]=e;else if(i==="channelsFirst")[l,p,u,d,h]=e;else throw new Error(`Unknown dataFormat ${i}`);let[c,f,m,,g]=t,[y,A,x]=Tg(r),[b,w,S]=Tg(n),N=xu(c,b),E=xu(f,w),R=xu(m,S),{padInfo:z,outDepth:M,outHeight:I,outWidth:_}=vF(a,u,d,h,y,A,x,N,E,R,o),O=s?g*p:g,q;return i==="channelsFirst"?q=[l,O,M,I,_]:i==="channelsLast"&&(q=[l,M,I,_,O]),{batchSize:l,dataFormat:i,inDepth:u,inHeight:d,inWidth:h,inChannels:p,outDepth:M,outHeight:I,outWidth:_,outChannels:O,padInfo:z,strideDepth:y,strideHeight:A,strideWidth:x,filterDepth:c,filterHeight:f,filterWidth:m,effectiveFilterDepth:N,effectiveFilterHeight:E,effectiveFilterWidth:R,dilationDepth:b,dilationHeight:w,dilationWidth:S,inShape:e,outShape:q,filterShape:t}}function AF(e,t,r,n,a){n==null&&(n=Ly(e,t,r));let s=e[0],i=e[1],o=Eo((s-t+2*n)/r+1,a),l=Eo((i-t+2*n)/r+1,a);return[o,l]}function xF(e,t,r,n,a,s){a==null&&(a=Ly(e,t,n));let i=e[0],o=e[1],l=e[2],u=Eo((i-t+2*a)/n+1,s),d=Eo((o-t+2*a)/n+1,s),h=Eo((l-t+2*a)/n+1,s);return[u,d,h,r]}function Ly(e,t,r,n=1){let a=xu(t,n);return Math.floor((e[0]*(r-1)-r+a)/2)}function T0(e){return typeof e=="number"?[e,e,e]:e.length===2?[e[0],e[1],1]:e}function Tg(e){return typeof e=="number"?[e,e,e]:e}function xu(e,t){return t<=1?e:e+(e-1)*(t-1)}function bF(e,t,r,n,a,s,i,o,l){let u,d,h;if(typeof e=="number"){u={top:e,bottom:e,left:e,right:e,type:e===0?"VALID":"NUMBER"};let p=AF([t,r],s,n,e,o);d=p[0],h=p[1]}else if(e==="same"){d=Math.ceil(t/n),h=Math.ceil(r/a);let p=Math.max(0,(d-1)*n+s-t),c=Math.max(0,(h-1)*a+i-r),f=Math.floor(p/2),m=p-f,g=Math.floor(c/2),y=c-g;u={top:f,bottom:m,left:g,right:y,type:"SAME"}}else if(e==="valid")u={top:0,bottom:0,left:0,right:0,type:"VALID"},d=Math.ceil((t-s+1)/n),h=Math.ceil((r-i+1)/a);else if(typeof e=="object"){let p=l==="channelsLast"?e[1][0]:e[2][0],c=l==="channelsLast"?e[1][1]:e[2][1],f=l==="channelsLast"?e[2][0]:e[3][0],m=l==="channelsLast"?e[2][1]:e[3][1];u={top:p,bottom:c,left:f,right:m,type:p===0&&c===0&&f===0&&m===0?"VALID":"EXPLICIT"},d=Eo((t-s+p+c)/n+1,o),h=Eo((r-i+f+m)/a+1,o)}else throw Error(`Unknown padding parameter: ${e}`);return{padInfo:u,outHeight:d,outWidth:h}}function vF(e,t,r,n,a,s,i,o,l,u,d){let h,p,c,f;if(typeof e=="number"){h={top:e,bottom:e,left:e,right:e,front:e,back:e,type:e===0?"VALID":"NUMBER"};let m=xF([t,r,n,1],o,1,a,e,d);p=m[0],c=m[1],f=m[2]}else if(e==="same"){p=Math.ceil(t/a),c=Math.ceil(r/s),f=Math.ceil(n/i);let m=(p-1)*a+o-t,g=(c-1)*s+l-r,y=(f-1)*i+u-n,A=Math.floor(m/2),x=m-A,b=Math.floor(g/2),w=g-b,S=Math.floor(y/2),N=y-S;h={top:b,bottom:w,left:S,right:N,front:A,back:x,type:"SAME"}}else if(e==="valid")h={top:0,bottom:0,left:0,right:0,front:0,back:0,type:"VALID"},p=Math.ceil((t-o+1)/a),c=Math.ceil((r-l+1)/s),f=Math.ceil((n-u+1)/i);else throw Error(`Unknown padding parameter: ${e}`);return{padInfo:h,outDepth:p,outHeight:c,outWidth:f}}function Eo(e,t){if(!t)return Math.trunc(e);switch(t){case"round":return Math.round(e);case"ceil":return Math.ceil(e);case"floor":return Math.floor(e);default:throw new Error(`Unknown roundingMode ${t}`)}}function Ds(e){let[t,r,n]=T0(e);return t===1&&r===1&&n===1}function za(e,t){return Ds(e)||Ds(t)}function U7(e){if(e==="NHWC")return"channelsLast";if(e==="NCHW")return"channelsFirst";throw new Error(`Unknown dataFormat ${e}`)}function jr(e,t,r){if(r!=null){if(typeof t=="string")throw Error(`Error in ${e}: pad must be an integer when using dimRoundingMode ${r} but got pad ${t}.`);if(typeof t=="number")P(wu(t),()=>`Error in ${e}: pad must be an integer when using dimRoundingMode ${r} but got pad ${t}.`);else if(typeof t=="object")t.forEach(n=>{n.forEach(a=>{P(wu(a),()=>`Error in ${e}: pad must be an integer when using dimRoundingMode ${r} but got pad ${a}.`)})});else throw Error(`Error in ${e}: Unknown padding parameter: ${t}`)}}function wF(e,t){let r={x:$(e,"x","reshape","string_or_numeric")},n={shape:t};return B.runKernel(ml,r,n)}var U=W({reshape_:wF});function kF(e,t,r,n,a){let s=$(e,"x","avgPool","float32"),i=1;P(za(r,i),()=>`Error in avgPool: Either strides or dilations must be 1. Got strides ${r} and dilations '${i}'`);let o=s,l=!1;s.rank===3&&(l=!0,o=U(s,[1,s.shape[0],s.shape[1],s.shape[2]])),P(o.rank===4,()=>`Error in avgPool: x must be rank 4 but got rank ${o.rank}.`),jr("avgPool",n,a);let u={x:o},d={filterSize:t,strides:r,pad:n,dimRoundingMode:a},h=B.runKernel(Js,u,d);return h=me(h,s.dtype),l?U(h,[h.shape[1],h.shape[2],h.shape[3]]):h}var Tf=W({avgPool_:kF});function IF(e,t,r,n,a,s="NDHWC"){let i=$(e,"x","avgPool3d","float32"),o=i,l=!1;i.rank===4&&(l=!0,o=U(i,[1,i.shape[0],i.shape[1],i.shape[2],i.shape[3]])),P(o.rank===5,()=>`Error in avgPool3d: x must be rank 5 but got rank ${o.rank}.`),P(s==="NDHWC",()=>`Error in avgPool3d: Only NDHWC is currently supported, but got dataFormat of ${s}`),jr("avgPool3d",n,a);let u={x:o},d={filterSize:t,strides:r,pad:n,dimRoundingMode:a,dataFormat:s},h=B.runKernel(Yp,u,d);return h=me(h,o.dtype),l?U(h,[h.shape[1],h.shape[2],h.shape[3],h.shape[4]]):h}var By=W({avgPool3d_:IF});function SF(e,t=0){P(e.length>=1,()=>"Pass at least one tensor to concat");let r=Wp(e,"tensors","concat","string_or_numeric");if(r[0].dtype==="complex64"&&r.forEach(s=>{if(s.dtype!=="complex64")throw new Error(`Cannot concatenate complex64 tensors with a tensor
|
|
with dtype ${s.dtype}. `)}),r.length===1)return Wr(r[0]);let n=r,a={axis:t};return B.runKernel(Xo,n,a)}var St=W({concat_:SF});function CF(e){let t={x:$(e,"x","sigmoid","float32")};return B.runKernel(_i,t)}var Tr=W({sigmoid_:CF});function TF(e,t,r){let n=$(e,"x","slice","string_or_numeric");if(n.rank===0)throw new Error("Slicing scalar is not possible");let a={x:n},s={begin:t,size:r};return B.runKernel(bl,a,s)}var Pe=W({slice_:TF});function NF(e){let t={x:$(e,"x","tanh","float32")};return B.runKernel(Wi,t)}var Iu=W({tanh_:NF});function EF(e,t,r,n,a,s){let i=$(e,"forgetBias","basicLSTMCell"),o=$(t,"lstmKernel","basicLSTMCell"),l=$(r,"lstmBias","basicLSTMCell"),u=$(n,"data","basicLSTMCell"),d=$(a,"c","basicLSTMCell"),h=$(s,"h","basicLSTMCell"),p=St([u,h],1),c=Ze(p,o),f=le(c,l),m=f.shape[0],g=f.shape[1]/4,y=[m,g],A=Pe(f,[0,0],y),x=Pe(f,[0,g],y),b=Pe(f,[0,g*2],y),w=Pe(f,[0,g*3],y),S=le(L(Tr(A),Iu(x)),L(d,Tr(le(i,b)))),N=L(Iu(S),Tr(w));return[S,N]}var RF=W({basicLSTMCell_:EF});function $F(e,t,r){let n=$(e,"x","batchToSpaceND"),a=t.reduce((o,l)=>o*l);P(n.rank>=1+t.length,()=>`input rank is ${n.rank} but should be > than blockShape.length ${t.length}`),P(r.length===t.length,()=>`crops.length is ${r.length} but should be equal to blockShape.length ${t.length}`),P(n.shape[0]%a===0,()=>`input tensor batch is ${n.shape[0]} but is not divisible by the product of the elements of blockShape ${t.join(" * ")} === ${a}`);let s={x:n},i={blockShape:t,crops:r};return B.runKernel(qo,s,i)}var Nf=W({batchToSpaceND_:$F});function MF(e){let t;return e.rank===0||e.rank===1?t=U(e,[1,1,1,e.size]):e.rank===2?t=U(e,[1,1,e.shape[0],e.shape[1]]):e.rank===3?t=U(e,[1,e.shape[0],e.shape[1],e.shape[2]]):t=e,t}function FF(e,t,r,n,a,s){s==null&&(s=.001);let i=$(e,"x","batchNorm"),o=$(t,"mean","batchNorm"),l=$(r,"variance","batchNorm"),u;a!=null&&(u=$(a,"scale","batchNorm"));let d;n!=null&&(d=$(n,"offset","batchNorm")),P(o.rank===l.rank,()=>"Batch normalization gradient requires mean and variance to have equal ranks."),P(d==null||o.rank===d.rank,()=>"Batch normalization gradient requires mean and offset to have equal ranks."),P(u==null||o.rank===u.rank,()=>"Batch normalization gradient requires mean and scale to have equal ranks.");let h={x:MF(i),scale:u,offset:d,mean:o,variance:l},p={varianceEpsilon:s},c=B.runKernel(ci,h,p);return U(c,i.shape)}var Su=W({batchNorm_:FF});function PF(e,t,r,n,a,s){let i=$(e,"x","batchNorm"),o=$(t,"mean","batchNorm"),l=$(r,"variance","batchNorm"),u;a!=null&&(u=$(a,"scale","batchNorm"));let d;return n!=null&&(d=$(n,"offset","batchNorm")),P(i.rank===2,()=>`Error in batchNorm2D: x must be rank 2 but got rank ${i.rank}.`),P(o.rank===2||o.rank===1,()=>`Error in batchNorm2D: mean must be rank 2 or rank 1 but got rank ${o.rank}.`),P(l.rank===2||l.rank===1,()=>`Error in batchNorm2D: variance must be rank 2 or rank 1 but got rank ${l.rank}.`),u!=null&&P(u.rank===2||u.rank===1,()=>`Error in batchNorm2D: scale must be rank 2 or rank 1 but got rank ${u.rank}.`),d!=null&&P(d.rank===2||d.rank===1,()=>`Error in batchNorm2D: offset must be rank 2 or rank 1 but got rank ${d.rank}.`),Su(i,o,l,d,u,s)}var G7=W({batchNorm2d_:PF});function _F(e,t,r,n,a,s){let i=$(e,"x","batchNorm"),o=$(t,"mean","batchNorm"),l=$(r,"variance","batchNorm"),u;a!=null&&(u=$(a,"scale","batchNorm"));let d;return n!=null&&(d=$(n,"offset","batchNorm")),P(i.rank===3,()=>`Error in batchNorm3D: x must be rank 3 but got rank ${i.rank}.`),P(o.rank===3||o.rank===1,()=>`Error in batchNorm3D: mean must be rank 3 or rank 1 but got rank ${o.rank}.`),P(l.rank===3||l.rank===1,()=>`Error in batchNorm3D: variance must be rank 3 or rank 1 but got rank ${l.rank}.`),u!=null&&P(u.rank===3||u.rank===1,()=>`Error in batchNorm3D: scale must be rank 3 or rank 1 but got rank ${u.rank}.`),d!=null&&P(d.rank===3||d.rank===1,()=>`Error in batchNorm3D: offset must be rank 3 or rank 1 but got rank ${d.rank}.`),Su(i,o,l,d,u,s)}var j7=W({batchNorm3d_:_F});function zF(e,t,r,n,a,s){let i=$(e,"x","batchNorm"),o=$(t,"mean","batchNorm"),l=$(r,"variance","batchNorm"),u;a!=null&&(u=$(a,"scale","batchNorm"));let d;return n!=null&&(d=$(n,"offset","batchNorm")),P(i.rank===4,()=>`Error in batchNorm4D: x must be rank 4 but got rank ${i.rank}.`),P(o.rank===4||o.rank===1,()=>`Error in batchNorm4D: mean must be rank 4 or rank 1 but got rank ${o.rank}.`),P(l.rank===4||l.rank===1,()=>`Error in batchNorm4D: variance must be rank 4 or rank 1 but got rank ${l.rank}.`),u!=null&&P(u.rank===4||u.rank===1,()=>`Error in batchNorm4D: scale must be rank 4 or rank 1 but got rank ${u.rank}.`),d!=null&&P(d.rank===4||d.rank===1,()=>`Error in batchNorm4D: offset must be rank 4 or rank 1 but got rank ${d.rank}.`),Su(i,o,l,d,u,s)}var H7=W({batchNorm4d_:zF});function OF(e,t,r){let n=$(e,"x","bincount"),a=$(t,"weights","bincount");P(n.dtype==="int32",()=>`Error in bincount: input dtype must be int32, but got ${n.dtype}`),P(r>=0,()=>`size must be non-negative, but got ${r}.`),P(a.size===n.size||a.size===0,()=>`Error in bincount: weights must have the same size as input or0-length, but got input shape: ${n.shape}, weights shape: ${a.shape}.`);let s={x:n,weights:a},i={size:r};return B.runKernel(ef,s,i)}var Wy=W({bincount_:OF});function DF(e,t){let r=$(e,"s0","broadcastArgs","int32"),n=$(t,"s1","broadcastArgs","int32");if(r.rank!==1)throw new Error(`broadcastArgs(): first input must be a vector (rank=1). Has rank ${r.rank}`);if(n.rank!==1)throw new Error(`broadcastArgs(): second input must be a vector (rank=1). Has rank ${n.rank}`);let a={s0:r,s1:n};return B.runKernel(tf,a)}var q7=W({broadcastArgs_:DF});function LF(e,t){let r=$(e,"broadcastTo","x"),n=r.shape;if(t.some(l=>!(l>0)||l%1!==0))throw new Error(`broadcastTo(): Invalid broadcast shape [${t}].`);if(t.length<r.rank)throw new Error(`broadcastTo(): shape.length=${t.length} < input.rank=${r.rank}.`);if(t.length>r.rank){let l=r.shape.slice();for(;l.length<t.length;)l.unshift(1);r=U(r,l)}let a=r.shape,s=Array.from(t);for(let l=t.length-1;l>=0;l--)if(a[l]===t[l])s[l]=1;else if(r.shape[l]!==1)throw new Error(`broadcastTo(): [${n}] cannot be broadcast to [${t}].`);if(s.map((l,u)=>l>1?u:-1).filter(l=>l>=0).length===0)return Wr(r);let i={x:r},o={reps:s};return B.runKernel(es,i,o)}var Mp=W({broadcastTo_:LF});function BF(e){let t={x:$(e,"x","ceil","float32")};return B.runKernel(ti,t)}var X7=W({ceil_:BF});function WF(e,t,r){let n=$(e,"x","clipByValue");P(t<=r,()=>`Error in clip: min (${t}) must be less than or equal to max (${r}).`);let a={x:n},s={clipValueMin:t,clipValueMax:r};return B.runKernel(Qa,a,s)}var cn=W({clipByValue_:WF});function VF(e){return St(e,0)}var K7=W({concat1d_:VF});function UF(e,t){return St(e,t)}var hd=W({concat2d_:UF});function GF(e,t){return St(e,t)}var Z7=W({concat3d_:GF});function jF(e,t){return St(e,t)}var Y7=W({concat4d_:jF});function HF(e,t,r,n,a="NHWC",s=[1,1],i){let o=$(e,"x","conv2d","float32"),l=$(t,"filter","conv2d","float32"),u=o,d=!1;o.rank===3&&(d=!0,u=U(o,[1,o.shape[0],o.shape[1],o.shape[2]])),P(u.rank===4,()=>`Error in conv2d: input must be rank 4, but got rank ${u.rank}.`),P(l.rank===4,()=>`Error in conv2d: filter must be rank 4, but got rank ${l.rank}.`),jr("conv2d",n,i);let h=a==="NHWC"?u.shape[3]:u.shape[1];P(h===l.shape[2],()=>`Error in conv2d: depth of input (${h}) must match input depth for filter ${l.shape[2]}.`),P(za(r,s),()=>`Error in conv2D: Either strides or dilations must be 1. Got strides ${r} and dilations '${s}'`);let p={x:u,filter:l},c={strides:r,pad:n,dataFormat:a,dilations:s,dimRoundingMode:i},f=B.runKernel(ri,p,c);return d?U(f,[f.shape[1],f.shape[2],f.shape[3]]):f}var Ls=W({conv2d_:HF});function qF(e,t,r,n,a="NWC",s=1,i){let o=$(e,"x","conv1d"),l=$(t,"filter","conv1d"),u=o,d=!1;o.rank===2&&(d=!0,u=U(o,[1,o.shape[0],o.shape[1]])),P(u.rank===3,()=>`Error in conv1d: input must be rank 3, but got rank ${u.rank}.`),P(l.rank===3,()=>`Error in conv1d: filter must be rank 3, but got rank ${l.rank}.`),jr("conv1d",n,i),P(u.shape[2]===l.shape[1],()=>`Error in conv1d: depth of input (${u.shape[2]}) must match input depth for filter ${l.shape[1]}.`),P(za(r,s),()=>`Error in conv1D: Either stride or dilation must be 1. Got stride ${r} and dilation '${s}'`),P(a==="NWC",()=>`Error in conv1d: got dataFormat of ${a} but only NWC is currently supported.`);let h=U(l,[1,l.shape[0],l.shape[1],l.shape[2]]),p=U(u,[u.shape[0],1,u.shape[1],u.shape[2]]),c=Ls(p,h,[1,r],n,"NHWC",[1,s],i);return d?U(c,[c.shape[2],c.shape[3]]):U(c,[c.shape[0],c.shape[2],c.shape[3]])}var Vy=W({conv1d_:qF});function XF(e,t,r,n,a,s="NHWC",i){P(e.length===t.rank,()=>`Length of inShape (${e.length}) and rank of dy (${t.rank}) must match`);let o=e,l=t,u=!1;t.rank===3&&(u=!0,l=U(t,[1,t.shape[0],t.shape[1],t.shape[2]]),o=[1,e[0],e[1],e[2]]),P(o.length===4,()=>`Error in conv2dDerInput: inShape must be length 4, but got length ${o.length}.`),P(l.rank===4,()=>`Error in conv2dDerInput: dy must be rank 4, but got rank ${l.rank}`),P(r.rank===4,()=>`Error in conv2dDerInput: filter must be rank 4, but got rank ${r.rank}`);let d=s==="NHWC"?o[3]:o[1],h=s==="NHWC"?l.shape[3]:l.shape[1];P(d===r.shape[2],()=>`Error in conv2dDerInput: depth of input (${d}) must match input depth for filter ${r.shape[2]}.`),P(h===r.shape[3],()=>`Error in conv2dDerInput: depth of output (${h}) must match output depth for filter ${r.shape[3]}.`),jr("conv2dDerInput",a,i);let p={dy:l,filter:r},c={strides:n,pad:a,dataFormat:s,dimRoundingMode:i,inputShape:o},f=B.runKernel(ni,p,c);return u?U(f,[f.shape[1],f.shape[2],f.shape[3]]):f}var Uy=W({conv2DBackpropInput_:XF});function KF(e,t,r,n,a,s){let i=$(e,"x","conv2dTranspose"),o=$(t,"filter","conv2dTranspose");return Uy(r,i,o,n,a,"NHWC",s)}var Gy=W({conv2dTranspose_:KF});function ZF(e,t,r,n,a="NDHWC",s=[1,1,1]){let i=$(e,"x","conv3d"),o=$(t,"filter","conv3d"),l=i,u=!1;i.rank===4&&(u=!0,l=U(i,[1,i.shape[0],i.shape[1],i.shape[2],i.shape[3]])),P(l.rank===5,()=>`Error in conv3d: input must be rank 5, but got rank ${l.rank}.`),P(o.rank===5,()=>`Error in conv3d: filter must be rank 5, but got rank ${o.rank}.`),P(l.shape[4]===o.shape[3],()=>`Error in conv3d: depth of input (${l.shape[4]}) must match input depth for filter ${o.shape[3]}.`),P(za(r,s),()=>`Error in conv3D: Either strides or dilations must be 1. Got strides ${r} and dilations '${s}'`),P(a==="NDHWC",()=>`Error in conv3d: got dataFormat of ${a} but only NDHWC is currently supported.`);let d={x:l,filter:o},h={strides:r,pad:n,dataFormat:a,dilations:s},p=B.runKernel(eh,d,h);return u?U(p,[p.shape[1],p.shape[2],p.shape[3],p.shape[4]]):p}var jy=W({conv3d_:ZF});function YF(e,t,r,n,a){P(e.length===t.rank,()=>`Length of inShape (${e.length}) and rank of dy (${t.rank}) must match`);let s=e,i=t,o=!1;t.rank===4&&(o=!0,i=U(t,[1,t.shape[0],t.shape[1],t.shape[2],t.shape[3]]),s=[1,e[0],e[1],e[2],e[3]]);let l=s[4],u=i.shape[4];P(s.length===5,()=>`Error in conv3dDerInput: inShape must be length 5, but got length ${s.length}.`),P(i.rank===5,()=>`Error in conv3dDerInput: dy must be rank 5, but got rank ${i.rank}`),P(r.rank===5,()=>`Error in conv3dDerInput: filter must be rank 5, but got rank ${r.rank}`),P(l===r.shape[3],()=>`Error in conv3dDerInput: depth of input (${l}) must match input depth for filter ${r.shape[3]}.`),P(u===r.shape[4],()=>`Error in conv3dDerInput: depth of output (${u}) must match output depth for filter ${r.shape[4]}.`);let d={dy:i,filter:r},h={pad:a,strides:n,inputShape:s},p=B.runKernel(af,d,h);return o?U(p,[p.shape[1],p.shape[2],p.shape[3],p.shape[4]]):p}var J7=W({conv3DBackpropInput_:YF});function JF(e,t,r,n,a){let s=$(e,"x","conv3dTranspose"),i=$(t,"filter","conv3dTranspose");return J7(r,s,i,n,a)}var Q7=W({conv3dTranspose_:JF});function QF(e){let t={x:$(e,"x","cos","float32")};return B.runKernel(ai,t)}var Ef=W({cos_:QF});function eP(e){let t={x:$(e,"x","cosh","float32")};return B.runKernel(si,t)}var Hy=W({cosh_:eP});function tP(e,t=0,r=!1,n=!1){let a={x:$(e,"x","cumprod")},s={axis:t,exclusive:r,reverse:n};return B.runKernel(Ko,a,s)}var N0=W({cumprod_:tP});function rP(e,t=0,r=!1,n=!1){let a={x:$(e,"x","cumsum")},s={axis:t,exclusive:r,reverse:n};return B.runKernel(ii,a,s)}var qy=W({cumsum_:rP});function nP(e,t,r,n=!1){let a=$(e,"x","denseBincount"),s=$(t,"weights","denseBincount");P(a.dtype==="int32",()=>`Error in denseBincount: input dtype must be int32, but got ${a.dtype}`),P(a.rank<=2,()=>`Error in denseBincount: input must be at most rank 2, but got rank ${a.rank}.`),P(r>=0,()=>`size must be non-negative, but got ${r}.`),P(s.size===a.size||s.size===0,()=>`Error in denseBincount: weights must have the same shape as x or 0-length, but got x shape: ${a.shape}, weights shape: ${s.shape}.`);let i={x:a,weights:s},o={size:r,binaryOutput:n};return B.runKernel(sf,i,o)}var ew=W({denseBincount_:nP});function aP(e,t,r="NHWC"){let n=$(e,"x","depthToSpace","float32"),a=r==="NHWC"?n.shape[1]:n.shape[2],s=r==="NHWC"?n.shape[2]:n.shape[3],i=r==="NHWC"?n.shape[3]:n.shape[1];P(t>1,()=>`blockSize should be > 1 for depthToSpace, but was: ${t}`),P(a*t>=0,()=>`Negative dimension size caused by overflow when multiplying
|
|
${a} and ${t} for depthToSpace with input shape
|
|
${n.shape}`),P(s*t>=0,()=>`Negative dimension size caused by overflow when multiplying
|
|
${s} and ${t} for depthToSpace with input shape
|
|
${n.shape}`),P(i%(t*t)===0,()=>`Dimension size must be evenly divisible by ${t*t} but is ${i} for depthToSpace with input shape ${n.shape}`);let o={x:n},l={blockSize:t,dataFormat:r};return B.runKernel(Yo,o,l)}var tw=W({depthToSpace_:aP});function sP(e,t,r,n,a="NHWC",s=[1,1],i){let o=$(e,"x","depthwiseConv2d","float32"),l=$(t,"filter","depthwiseConv2d","float32"),u=o,d=!1;o.rank===3&&(d=!0,u=U(o,[1,o.shape[0],o.shape[1],o.shape[2]])),P(u.rank===4,()=>`Error in depthwiseConv2d: input must be rank 4, but got rank ${u.rank}.`),P(l.rank===4,()=>`Error in depthwiseConv2d: filter must be rank 4, but got rank ${l.rank}.`),P(u.shape[3]===l.shape[2],()=>`Error in depthwiseConv2d: number of input channels (${u.shape[3]}) must match the inChannels dimension in filter ${l.shape[2]}.`),jr("depthwiseConv2d",n,i);let h={x:u,filter:l},p={strides:r,pad:n,dataFormat:a,dilations:s,dimRoundingMode:i},c=B.runKernel(oi,h,p);return d?U(c,[c.shape[1],c.shape[2],c.shape[3]]):c}var vh=W({depthwiseConv2d_:sP});function iP(e){let t={x:$(e,"x","diag")};return B.runKernel(uf,t)}var oP=W({diag_:iP});function lP(e,t,r,n,a=[1,1],s="NHWC"){let i=$(e,"x","dilation2d"),o=$(t,"filter","dilation2d");P(i.rank===3||i.rank===4,()=>`Error in dilation2d: input must be rank 3 or 4, but got rank ${i.rank}.`),P(o.rank===3,()=>`Error in dilation2d: filter must be rank 3, but got rank ${o.rank}.`),P(s==="NHWC",()=>`Error in dilation2d: Only NHWC is currently supported, but got dataFormat of ${s}`);let l=i,u=!1;i.rank===3&&(l=U(i,[1,i.shape[0],i.shape[1],i.shape[2]]),u=!0);let d={x:l,filter:o},h={strides:r,pad:n,dilations:a},p=B.runKernel(th,d,h);return u?U(p,[p.shape[1],p.shape[2],p.shape[3]]):p}var rw=W({dilation2d_:lP});function uP(e,t){let r=$(e,"a","equal","string_or_numeric"),n=$(t,"b","equal","string_or_numeric");[r,n]=Lt(r,n),bt(r.shape,n.shape);let a={a:r,b:n};return B.runKernel(Jo,a)}var $n=W({equal_:uP});function dP(e,t,r){let n=$(t,"a","where"),a=$(r,"b","where"),s=$(e,"condition","where","bool"),i=bt(bt(s.shape,n.shape),a.shape),o=Mp(s,i),l=Mp(n,i),u=Mp(a,i),d={condition:o,t:l,e:u};return B.runKernel(xl,d)}var Vr=W({where_:dP});function pP(e){let t={x:$(e,"x","zerosLike")};return B.runKernel(El,t)}var at=W({zerosLike_:pP});function hP(e,t){let r=$(e,"a","div"),n=$(t,"b","div");[r,n]=Lt(r,n);let a=pe(r,n),s=at(a),i=$n(n,s);return Vr(i,s,a)}var nw=W({divNoNan_:hP});function cP(e,t){let r=$(e,"t1","dot"),n=$(t,"t2","dot");P((r.rank===1||r.rank===2)&&(n.rank===1||n.rank===2),()=>`Error in dot: inputs must all be rank 1 or 2, but got ranks ${r.rank} and ${n.rank}.`);let a=r.rank===1?r.size:r.shape[1],s=n.rank===1?n.size:n.shape[0];if(P(a===s,()=>`Error in dot: inner dimensions of inputs must match, but got ${a} and ${s}.`),r.rank===1&&n.rank===1){let i=U(r,[1,-1]),o=U(n,[-1,1]),l=Ze(i,o);return U(l,[])}else if(r.rank===1&&n.rank===2){let i=U(r,[1,-1]),o=U(n,[n.shape[0],n.shape[1]]),l=Ze(i,o);return U(l,[l.size])}else if(r.rank===2&&n.rank===1){let i=U(n,[-1,1]),o=Ze(r,i);return U(o,[o.size])}else{let i=U(n,[n.shape[0],n.shape[1]]);return Ze(r,i)}}var fP=W({dot_:cP});function mP(e,...t){let r=t.map((a,s)=>$(a,`tensors${s}`,"einsum")),n={equation:e};return B.runKernel(rh,r,n)}var aw=W({einsum_:mP});function gP(e){let t={x:$(e,"x","elu","float32")};return B.runKernel(ui,t)}var wh=W({elu_:gP});function yP(e){let t=$(e,"x","erf");P(t.dtype==="int32"||t.dtype==="float32",()=>"Input dtype must be `int32` or `float32`."),t.dtype==="int32"&&(t=me(t,"float32"));let r={x:t};return B.runKernel(Xu,r)}var sw=W({erf_:yP});function AP(e){let t={x:$(e,"x","exp")};return B.runKernel(di,t)}var Mn=W({exp_:AP});function xP(e,t=0){let r=$(e,"x","expandDims","string_or_numeric");P(t<=r.rank,()=>"Axis must be <= rank of the tensor");let n={input:r},a={dim:t};return B.runKernel(Qo,n,a)}var Kt=W({expandDims_:xP});function bP(e){let t={x:$(e,"x","expm1")};return B.runKernel(el,t)}var iw=W({expm1_:bP});function vP(e,t){let r=$(e,"x","tile","string_or_numeric");P(r.rank===t.length,()=>`Error in transpose: rank of input ${r.rank} must match length of reps ${t}.`);let n={x:r},a={reps:t};return B.runKernel(es,n,a)}var Gn=W({tile_:vP});function wP(e,t,r,n="float32"){t==null&&(t=e);let a=We([e,t],n),s=e<=t?e:t;for(let o=0;o<s;++o)a.set(1,o,o);let i=U(a.toTensor(),[e,t]);if(r==null)return i;if(r.length===1)return Gn(Kt(i,0),[r[0],1,1]);if(r.length===2)return Gn(Kt(Kt(i,0),0),[r[0],r[1],1,1]);if(r.length===3)return Gn(Kt(Kt(Kt(i,0),0),0),[r[0],r[1],r[2],1,1]);throw new Error(`eye() currently supports only 1D and 2D batchShapes, but received ${r.length}D.`)}var Xy=W({eye_:wP});function cd(e,t,r){let n={shape:e,value:t,dtype:r};return B.runKernel(Ku,{},n)}function kP(e){let t={x:$(e,"x","floor","float32")};return B.runKernel(pi,t)}var kh=W({floor_:kP});function IP(e,t,r=0,n=0){let a=$(e,"x","gather"),s=$(t,"indices","gather","int32"),i={x:a,indices:s},o={axis:r,batchDims:n};return B.runKernel(rl,i,o)}var Cu=W({gather_:IP});function SP(e,t){let r=$(e,"a","greater","string_or_numeric"),n=$(t,"b","greater","string_or_numeric");[r,n]=Lt(r,n),bt(r.shape,n.shape);let a={a:r,b:n};return B.runKernel(al,a)}var mn=W({greater_:SP});function CP(e,t){let r=$(e,"a","greaterEqual","string_or_numeric"),n=$(t,"b","greaterEqual","string_or_numeric");[r,n]=Lt(r,n),bt(r.shape,n.shape);let a={a:r,b:n};return B.runKernel(fi,a)}var Ml=W({greaterEqual_:CP});function TP(e){let t={input:$(e,"input","imag")};return B.runKernel(nh,t)}var Rf=W({imag_:TP});function NP(e){let t={x:$(e,"x","isFinite")};return B.runKernel(Zu,t)}var EP=W({isFinite_:NP});function RP(e){let t={x:$(e,"x","isInf")};return B.runKernel(Yu,t)}var $P=W({isInf_:RP});function MP(e){let t={x:$(e,"x","isNaN")};return B.runKernel(Ju,t)}var ow=W({isNaN_:MP});function FP(e,t=.2){let r={x:$(e,"x","leakyRelu")},n={alpha:t};return B.runKernel(gi,r,n)}var $f=W({leakyRelu_:FP});function PP(e,t){let r=$(e,"a","less","string_or_numeric"),n=$(t,"b","less","string_or_numeric");[r,n]=Lt(r,n),bt(r.shape,n.shape);let a={a:r,b:n};return B.runKernel(sl,a)}var Ky=W({less_:PP});function _P(e,t){let r=$(e,"a","lessEqual","string_or_numeric"),n=$(t,"b","lessEqual","string_or_numeric");[r,n]=Lt(r,n),bt(r.shape,n.shape);let a={a:r,b:n};return B.runKernel(il,a)}var Fl=W({lessEqual_:_P});function lw(e,t,r){if(r<=0)throw new Error("The number of values should be positive.");let n={start:e,stop:t,num:r};return B.runKernel(cf,{},n)}function zP(e,t=5,r=1,n=1,a=.5){let s=$(e,"x","localResponseNormalization");P(s.rank===4||s.rank===3,()=>`Error in localResponseNormalization: x must be rank 3 or 4 but got
|
|
rank ${s.rank}.`),P(wu(t),()=>`Error in localResponseNormalization: depthRadius must be an integer but got depthRadius ${t}.`);let i=s,o=!1;s.rank===3&&(o=!0,i=U(s,[1,s.shape[0],s.shape[1],s.shape[2]]));let l={x:i},u={depthRadius:t,bias:r,alpha:n,beta:a},d=B.runKernel(sh,l,u);return o?U(d,[d.shape[1],d.shape[2],d.shape[3]]):d}var uw=W({localResponseNormalization_:zP});function OP(e){let t={x:$(e,"x","log","float32")};return B.runKernel(yi,t)}var Fn=W({log_:OP});function DP(e){let t={x:$(e,"x","log1p")};return B.runKernel(Qu,t)}var Mf=W({log1p_:DP});function LP(e){return P(Fs(e),()=>"The f passed in grad(f) must be a function"),(t,r)=>{let n=$(t,"x","tf.grad","string_or_numeric"),a=r!=null?$(r,"dy","tf.grad"):null;return B.tidy(()=>{let{value:s,grads:i}=B.gradients(()=>e(n),[n],a);return a!=null&&Ur(s.shape,a.shape,"The shape of dy passed in grad(f)(x, dy) must match the shape returned by f(x)"),Ff(i),i[0]})}}function BP(e){return P(Fs(e),()=>"The f passed in grads(f) must be a function"),(t,r)=>{P(Array.isArray(t),()=>"The args passed in grads(f)(args) must be an array of `Tensor`s or `TensorLike`s");let n=Wp(t,"args","tf.grads","string_or_numeric"),a=r!=null?$(r,"dy","tf.grads"):null;return B.tidy(()=>{let{value:s,grads:i}=B.gradients(()=>e(...n),n,a);return a!=null&&Ur(s.shape,a.shape,"The shape of dy passed in grads(f)([x1,...], dy) must match the shape returned by f([x1,...])"),Ff(i),i})}}function WP(e){return P(Fs(e),()=>"The f passed in valueAndGrad(f) must be a function"),(t,r)=>{P(t instanceof nt,()=>"The x passed in valueAndGrad(f)(x) must be a tensor"),P(r==null||r instanceof nt,()=>"The dy passed in valueAndGrad(f)(x, dy) must be a tensor");let{grads:n,value:a}=B.gradients(()=>e(t),[t],r);return Ff(n),{grad:n[0],value:a}}}function VP(e){return P(Fs(e),()=>"The f passed in valueAndGrads(f) must be a function"),(t,r)=>{P(Array.isArray(t)&&t.every(a=>a instanceof nt),()=>"The args passed in valueAndGrads(f)(args) must be array of tensors"),P(r==null||r instanceof nt,()=>"The dy passed in valueAndGrads(f)(args, dy) must be a tensor");let n=B.gradients(()=>e(...t),t,r);return r!=null&&Ur(n.value.shape,r.shape,"The shape of dy passed in valueAndGrads(f)([x1,...], dy) must match the shape returned by f([x1,...])"),Ff(n.grads),n}}function dw(e,t){P(Fs(e),()=>"The f passed in variableGrads(f) must be a function"),P(t==null||Array.isArray(t)&&t.every(u=>u instanceof Bp),()=>"The varList passed in variableGrads(f, varList) must be an array of variables");let r=t!=null;if(!r){t=[];for(let u in B.registeredVariables)t.push(B.registeredVariables[u])}let n=r?t.filter(u=>!u.trainable):null,a=t.length;t=t.filter(u=>u.trainable),P(t.length>0,()=>`variableGrads() expects at least one of the input variables to be trainable, but none of the ${a} variables is trainable.`);let s=!0,{value:i,grads:o}=B.gradients(e,t,null,s);P(o.some(u=>u!=null),()=>"Cannot find a connection between any variable and the result of the loss function y=f(x). Please make sure the operations that use variables are inside the function f passed to minimize()."),P(i.rank===0,()=>`The f passed in variableGrads(f) must return a scalar, but it returned a rank-${i.rank} tensor`);let l={};return t.forEach((u,d)=>{o[d]!=null&&(l[u.name]=o[d])}),n!=null&&n.forEach(u=>l[u.name]=null),{value:i,grads:l}}function Pa(e){return B.customGrad(e)}function Ff(e){if(e.filter(t=>t==null).length>0)throw new Error(`Cannot compute gradient of y=f(x) with respect to x. Make sure that
|
|
the f you passed encloses all operations that lead from x to y.`)}function UP(e){let t={x:$(e,"x","neg")};return B.runKernel(ll,t)}var Dt=W({neg_:UP});function GP(e){let t={x:$(e,"x","softplus")};return B.runKernel(ld,t)}var fd=W({softplus_:GP});function jP(e){let t=$(e,"x","logSigmoid");return Pa(r=>({value:Dt(fd(Dt(r))),gradFunc:n=>L(n,Tr(Dt(r)))}))(t)}var HP=W({logSigmoid_:jP});function qP(e,t=null,r=!1){let n={x:$(e,"x","max")},a={reductionIndices:t,keepDims:r};return B.runKernel(Ai,n,a)}var Ar=W({max_:qP});function XP(e,t){let r=$(e,"a","sub"),n=$(t,"b","sub");[r,n]=Lt(r,n);let a={a:r,b:n};return B.runKernel(Bi,a)}var ce=W({sub_:XP});function KP(e,t=null,r=!1){let n=$(e,"x","sum");n.dtype==="bool"&&(n=me(n,"int32"));let a={x:n},s={axis:t,keepDims:r};return B.runKernel(Oi,a,s)}var ke=W({sum_:KP});function ZP(e,t=-1){let r=$(e,"logits","logSoftmax");if(t===-1&&(t=r.rank-1),t!==r.rank-1)throw Error(`Log Softmax along a non-last dimension is not yet supported. Logits was rank ${r.rank} and axis was ${t}`);return Pa((n,a)=>{let s=Ar(n,t,!0),i=ce(n,s),o=ce(me(i,"float32"),Fn(ke(Mn(i),t,!0)));return a([o]),{value:o,gradFunc:(l,u)=>{let[d]=u,h=!0,p=Mn(d);return ce(l,L(ke(l,t,h),p))}}})(r)}var Zy=W({logSoftmax_:ZP});function Yy(e,t){for(let r=0;r<e.length;++r)if(e[e.length-r-1]!==t-1-r)return!1;return!0}function pw(e,t,r){let n=e.length+t.length,a=[],s=0,i=0;for(let o=0;o<n;o++)r.indexOf(o)===-1?a.push(e[s++]):a.push(t[i++]);return a}function hw(e,t){let r=[],n=e.length;for(let s=0;s<n;s++)t.indexOf(s)===-1&&r.push(e[s]);let a=t.map(s=>e[s]);return[r,a]}function zo(e,t){let r=t.map(n=>1);return pw(e,r,t)}function YP(e,t,r){P(Yy(t,r),()=>`${e} supports only inner-most axes for now. Got axes ${t} and rank-${r} input.`)}function cw(e,t){if(Yy(e,t))return null;let r=[];for(let n=0;n<t;++n)e.indexOf(n)===-1&&r.push(n);return e.forEach(n=>r.push(n)),r}function Jy(e){return e.map((t,r)=>[r,t]).sort((t,r)=>t[1]-r[1]).map(t=>t[0])}function JP(e,t){let r=[];for(let n=t-e;n<t;++n)r.push(n);return r}function QP(e,t=null,r=!1){let n=$(e,"x","logSumExp"),a=Hn(t,n.shape),s=Ar(n,a,!0),i=ce(n,s),o=Mn(i),l=ke(o,a),u=Fn(l),d=le(U(s,u.shape),u);if(r){let h=zo(d.shape,a);return U(d,h)}return d}var fw=W({logSumExp_:QP});function e_(e,t){let r=$(e,"a","logicalAnd","bool"),n=$(t,"b","logicalAnd","bool");bt(r.shape,n.shape);let a={a:r,b:n};return B.runKernel(ol,a)}var ga=W({logicalAnd_:e_});function t_(e){let t={x:$(e,"x","logicalNot","bool")};return B.runKernel(ed,t)}var Pf=W({logicalNot_:t_});function r_(e,t){let r=$(e,"a","logicalOr","bool"),n=$(t,"b","logicalOr","bool");bt(r.shape,n.shape);let a={a:r,b:n};return B.runKernel(ah,a)}var Qy=W({logicalOr_:r_});function n_(e,t){let r=$(e,"a","logicalXor","bool"),n=$(t,"b","logicalXor","bool");return bt(r.shape,n.shape),ga(Qy(e,t),Pf(ga(e,t)))}var a_=W({logicalXor_:n_}),qc=2147483648;function s_(e,t,r="left"){let n=$(e,"sortedSequence","searchSorted"),a=$(t,"values","searchSorted"),s=n.shape[n.shape.length-1],i=a.shape[a.shape.length-1],o=U(n,[-1,s]),l=U(a,[-1,i]);if(o.rank<2)throw new Error("Sorted input argument must be at least 2-dimensional");if(o.shape[0]!==l.shape[0])throw new Error("Leading dimension of 'sortedSequence' and 'values' must match.");if(It(l.shape)>=qc)throw new Error(`values tensor size must less than ${qc}`);if(o.shape[1]>=qc)throw new Error(`trailing dim_size must less than ${qc} for int32 output type, was ${o.shape[1]}`);let u={sortedSequence:o,values:l},d={side:r};return B.runKernel(vf,u,d)}var eA=W({searchSorted_:s_});function mw(e,t){return eA(e,t,"left")}function i_(e,t,r,n,a){let s=$(e,"x","maxPool"),i=1,o=s,l=!1;s.rank===3&&(l=!0,o=U(s,[1,s.shape[0],s.shape[1],s.shape[2]])),P(o.rank===4,()=>`Error in maxPool: input must be rank 4 but got rank ${o.rank}.`),P(za(r,i),()=>`Error in maxPool: Either strides or dilations must be 1. Got strides ${r} and dilations '${i}'`),jr("maxPool",n,a);let u={x:o},d={filterSize:t,strides:r,pad:n,dimRoundingMode:a},h=B.runKernel(bi,u,d);return l?U(h,[h.shape[1],h.shape[2],h.shape[3]]):h}var _f=W({maxPool_:i_});function o_(e,t=[1,1,1],r,n,a,s="NDHWC"){let i=$(e,"x","maxPool3d"),o=i,l=!1;i.rank===4&&(l=!0,o=U(i,[1,i.shape[0],i.shape[1],i.shape[2],i.shape[3]])),P(o.rank===5,()=>`Error in maxPool3d: x must be rank 5 but got rank ${o.rank}.`),P(s==="NDHWC",()=>`Error in maxPool3d: Only NDHWC is currently supported, but got dataFormat of ${s}`),jr("maxPool3d",n,a);let u={x:o},d={filterSize:t,strides:r,pad:n,dimRoundingMode:a,dataFormat:s},h=B.runKernel(ih,u,d);return l?U(h,[h.shape[1],h.shape[2],h.shape[3],h.shape[4]]):h}var tA=W({maxPool3d_:o_});function l_(e,t,r,n,a=!1){let s={x:$(e,"x","maxPoolWithArgmax")},i={filterSize:t,strides:r,pad:n,includeBatchInIndex:a},o=B.runKernel(yf,s,i);return{result:o[0],indexes:o[1]}}var gw=W({maxPoolWithArgmax_:l_});function u_(e,t){let r=$(e,"a","maximum"),n=$(t,"b","maximum");[r,n]=Lt(r,n),r.dtype==="bool"&&(r=me(r,"int32"),n=me(n,"int32")),bt(r.shape,n.shape);let a={a:r,b:n};return B.runKernel(xi,a)}var ts=W({maximum_:u_});function d_(e,t=null,r=!1){let n={x:$(e,"x","mean")},a={axis:t,keepDims:r};return B.runKernel(vi,n,a)}var Vt=W({mean_:d_});function _t(e,t="float32"){if(t==="complex64"){let n=_t(e,"float32"),a=_t(e,"float32");return Os(n,a)}let r=Y0(It(e),t);return B.makeTensor(r,e,t)}function hn(e,t="float32"){if(t==="complex64"){let n=hn(e,"float32"),a=_t(e,"float32");return Os(n,a)}let r=my(It(e),t);return B.makeTensor(r,e,t)}function p_(e,t,{indexing:r="xy"}={}){if(r!=="xy"&&r!=="ij")throw new TypeError(`${r} is not a valid third argument to meshgrid`);if(e===void 0)return[];let n=$(e,"x","meshgrid",e instanceof nt?e.dtype:"float32");if(t===void 0)return[n];let a=$(t,"y","meshgrid",t instanceof nt?t.dtype:"float32"),s=It(n.shape),i=It(a.shape);return r==="xy"?(n=U(n,[1,-1]),a=U(a,[-1,1]),[Ze(hn([i,1],n.dtype),n),Ze(a,hn([1,s],a.dtype))]):(n=U(n,[-1,1]),a=U(a,[1,-1]),[Ze(n,hn([1,i],n.dtype)),Ze(hn([s,1],a.dtype),a)])}function h_(e,t=null,r=!1){let n={x:$(e,"x","min")},a={axis:t,keepDims:r};return B.runKernel(wi,n,a)}var Bs=W({min_:h_});function c_(e,t){let r=$(e,"a","minimum"),n=$(t,"b","minimum");[r,n]=Lt(r,n),r.dtype==="bool"&&(r=me(r,"int32"),n=me(n,"int32")),bt(r.shape,n.shape);let a={a:r,b:n};return B.runKernel(ki,a)}var Ih=W({minimum_:c_});function f_(e,t,r){P(r==="reflect"||r==="symmetric",()=>`Invalid mode. Mode must be either reflect or symmetric. Got ${r}.`);let n=$(e,"x","mirrorPad");if(n.rank===0)throw new Error("mirrorPad(scalar) is not defined. Pass non-scalar to mirrorPad");P(t.length===n.rank,()=>`Padding doesn't match input. Must be ${n.rank}. Got ${t.length}.`);let a=r==="reflect"?1:0;for(let o=0;o<n.rank;o++)P(t[o].length===2,()=>"Invalid number of paddings. Must be length of 2 each."),P(t[o][0]>=0&&t[o][0]<=n.shape[o]-a&&t[o][1]>=0&&t[o][1]<=n.shape[o]-a,()=>`Padding in dimension ${o} cannot be greater than or equal to ${n.shape[o]-a} or less than 0 for input of shape ${n.shape}`);let s={paddings:t,mode:r},i={x:n};return B.runKernel(Ii,i,s)}var yw=W({mirrorPad_:f_});function m_(e,t){let r=$(e,"a","mod"),n=$(t,"b","mod");[r,n]=Lt(r,n);let a={a:r,b:n};return B.runKernel(td,a)}var md=W({mod_:m_});function g_(e){let t=$(e,"x","square"),r={};return B.runKernel("Square",{x:t},r)}var At=W({square_:g_});function y_(e,t=null,r=!1){e=$(e,"x","moments");let n=Hn(t,e.shape),a=Vt(e,n,r),s=a.shape;r||(s=zo(a.shape,n));let i=At(ce(me(e,"float32"),U(a,s))),o=Vt(i,n,r);return{mean:a,variance:o}}var zf=W({moments_:y_});function A_(e,t,r,n){let a=$(t,"data","multiRNNCell"),s=Wp(r,"c","multiRNNCell"),i=Wp(n,"h","multiRNNCell"),o=a,l=[];for(let h=0;h<e.length;h++){let p=e[h](o,s[h],i[h]);l.push(p[0]),l.push(p[1]),o=p[1]}let u=[],d=[];for(let h=0;h<l.length;h+=2)u.push(l[h]),d.push(l[h+1]);return[u,d]}var x_=W({multiRNNCell_:A_});function b_(e,t,r,n=!1){let a=$(e,"logits","multinomial"),s=a.size,i=a.rank;if(s<2)throw new Error(`Error in multinomial: you need at least 2 outcomes, but got ${s}.`);if(i>2)throw new Error(`Rank of probabilities must be 1 or 2, but is ${i}`);r=r||Math.random();let o={logits:i===1?U(a,[1,-1]):a},l={numSamples:t,seed:r,normalized:n},u=B.runKernel(Af,o,l);return i===1?U(u,[u.size]):u}var Aw=W({multinomial_:b_});function v_(e,t){let r=$(e,"a","notEqual","string_or_numeric"),n=$(t,"b","notEqual","string_or_numeric");[r,n]=Lt(r,n),bt(r.shape,n.shape);let a={a:r,b:n};return B.runKernel(ul,a)}var Tu=W({notEqual_:v_});function w_(e){let t={x:$(e,"x","onesLike")};return B.runKernel(hl,t)}var Pn=W({onesLike_:w_});function k_(e,t){let r=$(e,"v1","outerProduct"),n=$(t,"v2","outerProduct");P(r.rank===1&&n.rank===1,()=>`Error in outerProduct: inputs must be rank 1, but got ranks ${r.rank} and ${n.rank}.`);let a=U(r,[-1,1]),s=U(n,[1,-1]);return Ze(a,s)}var I_=W({outerProduct_:k_});function S_(e,t,r=0){let n=$(e,"x","pad");if(n.rank===0)throw new Error("pad(scalar) is not defined. Pass non-scalar to pad");let a={paddings:t,constantValue:r},s={x:n};return B.runKernel(Ci,s,a)}var Xn=W({pad_:S_});function C_(e,t,r=0){return P(t.length===2,()=>"Invalid number of paddings. Must be length of 2."),Xn(e,[t],r)}var T_=W({pad1d_:C_});function N_(e,t,r=0){return P(t.length===2&&t[0].length===2&&t[1].length===2,()=>"Invalid number of paddings. Must be length of 2 each."),Xn(e,t,r)}var E_=W({pad2d_:N_});function R_(e,t,r=0){return P(t.length===3&&t[0].length===2&&t[1].length===2&&t[2].length===2,()=>"Invalid number of paddings. Must be length of 2 each."),Xn(e,t,r)}var $_=W({pad3d_:R_});function M_(e,t,r=0){return P(t.length===4&&t[0].length===2&&t[1].length===2&&t[2].length===2&&t[3].length===2,()=>"Invalid number of paddings. Must be length of 2 each."),Xn(e,t,r)}var F_=W({pad4d_:M_});function P_(e,t,r){let n=$(e,"x","spaceToBatchND");P(n.rank>=1+t.length,()=>`input rank ${n.rank} should be > than [blockShape] ${t.length}`),P(r.length===t.length,()=>`paddings.shape[0] ${r.length} must be equal to [blockShape] ${t.length}`),P(n.shape.reduce((i,o,l)=>l>0&&l<=t.length?i&&(o+r[l-1][0]+r[l-1][1])%t[l-1]===0:i,!0),()=>`input spatial dimensions ${n.shape.slice(1)} with paddings ${r.toString()} must be divisible by blockShapes ${t.toString()}`);let a={x:n},s={blockShape:t,paddings:r};return B.runKernel(wl,a,s)}var Of=W({spaceToBatchND_:P_});function __(e,t,r,n,a,s,i){a==null&&(a=[1,1]),s==null&&(s=1),n===0&&(n="valid");let o=$(e,"x","maxPool"),l=o,u=!1;o.rank===3&&(u=!0,l=U(o,[1,o.shape[0],o.shape[1],o.shape[2]])),P(za(s,a),()=>`Error in pool: Either strides or dilations must be 1. Got strides ${s} and dilations '${a}'`);let d=W7(l.shape,t,s,a,n),h=[d.dilationHeight,d.dilationWidth],p;n==="same"?p=O_([d.filterHeight,d.filterWidth],h):p=[[0,0],[0,0]];let c=h[0]===1&&h[1]===1,[f,m]=z_([d.inHeight,d.inWidth],h,p),g=c?n:"valid",y=c?l:Of(l,h,f),A=(r==="avg"?()=>Tf(y,t,s,g,i):()=>_f(y,t,s,g,i))(),x=c?A:Nf(A,h,m);return u?U(x,[x.shape[1],x.shape[2],x.shape[3]]):x}function z_(e,t,r){let n=r.map(d=>d[0]),a=r.map(d=>d[1]),s=e.concat(n,a),i=t.map((d,h)=>(d-s[h]%d)%d),o=a.map((d,h)=>d+i[h]),l=t.map((d,h)=>[n[h],o[h]]),u=t.map((d,h)=>[0,i[h]]);return[l,u]}function O_(e,t){let r=e.map((s,i)=>s+(s-1)*(t[i]-1)).map(s=>s-1),n=r.map(s=>Math.floor(s/2)),a=r.map((s,i)=>s-n[i]);return r.map((s,i)=>[n[i],a[i]])}var D_=W({pool_:__});function L_(e,t){let r=$(e,"base","pow"),n=$(t,"exp","pow");[r,n]=Lt(r,n);let a={a:r,b:n};return B.runKernel(Ti,a)}var Ws=W({pow_:L_});function B_(e,t){let r=$(e,"x","prelu"),n=$(t,"alpha","prelu"),a={x:r,alpha:n};return B.runKernel(Ni,a)}var Df=W({prelu_:B_});function W_(e,t=null,r=!1){let n=$(e,"x","prod");n.dtype==="bool"&&(n=me(n,"int32"));let a={x:n},s={axis:t,keepDims:r};return B.runKernel(Ei,a,s)}var rA=W({prod_:W_});function V_(e,t,r){let n=It(e),a=null;if(r==null||r==="float32")a=new Float32Array(n);else if(r==="int32")a=new Int32Array(n);else if(r==="bool")a=new Uint8Array(n);else throw new Error(`Unknown data type ${r}`);for(let s=0;s<n;s++)a[s]=t();return B.makeTensor(a,e,r)}var U_=W({rand_:V_}),nA=Go(K0()),aA=class{constructor(e,t,r,n,a){this.mean=e,this.stdDev=t,this.dtype=r,this.nextVal=NaN,this.truncated=n,this.truncated&&(this.upper=this.mean+this.stdDev*2,this.lower=this.mean-this.stdDev*2);let s=a||Math.random();this.random=nA.alea(s.toString())}nextValue(){if(!isNaN(this.nextVal)){let n=this.nextVal;return this.nextVal=NaN,n}let e,t,r=!1;for(;!r;){let n,a,s;do n=2*this.random()-1,a=2*this.random()-1,s=n*n+a*a;while(s>=1||s===0);let i=Math.sqrt(-2*Math.log(s)/s);e=this.mean+this.stdDev*n*i,t=this.mean+this.stdDev*a*i,(!this.truncated||this.isValidTruncated(e))&&(r=!0)}return(!this.truncated||this.isValidTruncated(t))&&(this.nextVal=this.convertValue(t)),this.convertValue(e)}convertValue(e){return this.dtype==null||this.dtype==="float32"?e:Math.round(e)}isValidTruncated(e){return e<=this.upper&&e>=this.lower}},G_=class{constructor(e,t,r,n){this.alpha=e,this.beta=1/t,this.dtype=r;let a=n||Math.random();this.randu=nA.alea(a.toString()),this.randn=new aA(0,1,r,!1,this.randu()),e<1?this.d=e+2/3:this.d=e-1/3,this.c=1/Math.sqrt(9*this.d)}nextValue(){let e,t,r,n,a,s;for(;;){do n=this.randn.nextValue(),s=1+this.c*n;while(s<=0);if(s*=s*s,e=n*n,t=1-.331*e*e,r=.5*e+this.d*(1-s+Math.log(s)),a=this.randu(),a<t||Math.log(a)<r)break}return s=1/this.beta*this.d*s,this.alpha<1&&(s*=Math.pow(this.randu(),1/this.alpha)),this.convertValue(s)}convertValue(e){return this.dtype==="float32"?e:Math.round(e)}},j_=class{constructor(e=0,t=1,r,n){if(this.canReturnFloat=()=>this.dtype==null||this.dtype==="float32",this.min=e,this.range=t-e,this.dtype=r,n==null&&(n=Math.random()),typeof n=="number"&&(n=n.toString()),!this.canReturnFloat()&&this.range<=1)throw new Error(`The difference between ${e} - ${t} <= 1 and dtype is not float`);this.random=nA.alea(n)}convertValue(e){return this.canReturnFloat()?e:Math.round(e)}nextValue(){return this.convertValue(this.min+this.range*this.random())}};function H_(e,t,r=1,n="float32",a){if(r==null&&(r=1),n==null&&(n="float32"),n!=="float32"&&n!=="int32")throw new Error(`Unsupported data type ${n}`);let s=new G_(t,r,n,a),i=We(e,n);for(let o=0;o<i.values.length;o++)i.values[o]=s.nextValue();return i.toTensor()}var q_=W({randomGamma_:H_});function X_(e,t=0,r=1,n,a){if(n!=null&&n==="bool")throw new Error(`Unsupported data type ${n}`);let s=new aA(t,r,n,!1,a),i=We(e,n);for(let o=0;o<i.values.length;o++)i.values[o]=s.nextValue();return i.toTensor()}var xw=W({randomNormal_:X_});function K_(e,t=0,r=1,n="float32",a){let s=We(e,n),i=new j_(t,r,null,a);for(let o=0;o<s.values.length;o++)s.values[o]=i.nextValue();return s.toTensor()}var gd=W({randomUniform_:K_});function Nu(e,t,r=1,n="float32"){if(r===0)throw new Error("Cannot have a step of zero");let a={start:e,stop:t,step:r,dtype:n};return B.runKernel(nd,{},a)}function Z_(e){let t={input:$(e,"input","real")};return B.runKernel(oh,t)}var Up=W({real_:Z_});function Y_(e){let t={x:$(e,"x","reciprocal")};return B.runKernel(ad,t)}var bw=W({reciprocal_:Y_});function J_(e){let t={x:$(e,"x","relu")};return B.runKernel(Ri,t)}var Oa=W({relu_:J_});function Q_(e){let t={x:$(e,"x","relu6")};return B.runKernel(Mi,t)}var sA=W({relu6_:Q_});function ez(e,t){let r={x:$(e,"x","reverse")},n={dims:t};return B.runKernel(gl,r,n)}var _n=W({reverse_:ez});function tz(e){let t=$(e,"x","reverse");return P(t.rank===1,()=>`Error in reverse1D: x must be rank 1 but got rank ${t.rank}.`),_n(t,0)}var rz=W({reverse1d_:tz});function nz(e,t){let r=$(e,"x","reverse");return P(r.rank===2,()=>`Error in reverse2D: x must be rank 2 but got rank ${r.rank}.`),_n(r,t)}var az=W({reverse2d_:nz});function sz(e,t){let r=$(e,"x","reverse");return P(r.rank===3,()=>`Error in reverse3D: x must be rank 3 but got rank ${r.rank}.`),_n(r,t)}var iz=W({reverse3d_:sz});function oz(e,t){let r=$(e,"x","reverse");return P(r.rank===4,()=>`Error in reverse4D: x must be rank 4 but got rank ${r.rank}.`),_n(r,t)}var lz=W({reverse4d_:oz});function uz(e){let t={x:$(e,"x","round")};return B.runKernel(yl,t)}var iA=W({round_:uz});function dz(e){let t={x:$(e,"x","rsqrt","float32")};return B.runKernel(Fi,t)}var oA=W({rsqrt_:dz});function Se(e,t){if((Sr(e)&&t!=="string"||Array.isArray(e))&&t!=="complex64")throw new Error("Error creating a new Scalar: value must be a primitive (number|boolean|string)");if(t==="string"&&Sr(e)&&!(e instanceof Uint8Array))throw new Error("When making a scalar from encoded string, the value must be `Uint8Array`.");return Gi(e,[],[],t)}function pz(e){let t={x:$(e,"x","selu")};return B.runKernel(id,t)}var lA=W({selu_:pz});function hz(e,t,r,n,a,s=[1,1],i="NHWC"){let o=$(e,"x","separableConv2d"),l=$(t,"depthwiseFilter","separableConv2d"),u=$(r,"pointwiseFilter","separableConv2d"),d=o,h=!1;if(o.rank===3&&(h=!0,d=U(o,[1,o.shape[0],o.shape[1],o.shape[2]])),i==="NCHW")throw new Error("separableConv2d currently does not support dataFormat NCHW; only NHWC is supported");P(d.rank===4,()=>`Error in separableConv2d: input must be rank 4, but got rank ${d.rank}.`),P(l.rank===4,()=>`Error in separableConv2d: depthwise filter must be rank 4, but got rank ${l.rank}.`),P(u.rank===4,()=>`Error in separableConv2d: pointwise filter must be rank 4, but got rank ${l.rank}.`),P(u.shape[0]===1,()=>`Error in separableConv2d: the first dimension of pointwise filter must be 1, but got ${u.shape[0]}.`),P(u.shape[1]===1,()=>`Error in separableConv2d: the second dimension of pointwise filter must be 1, but got ${u.shape[1]}.`);let p=l.shape[2],c=l.shape[3];P(u.shape[2]===p*c,()=>`Error in separableConv2d: the third dimension of pointwise filter must be ${p*c}, but got ${u.shape[2]}.`);let f=vh(d,l,n,a,i,s),m=Ls(f,u,1,"valid",i);return h?U(m,[m.shape[1],m.shape[2],m.shape[3]]):m}var vw=W({separableConv2d_:hz});async function cz(e,t){let r=$(e,"x","setdiff1d"),n=$(t,"y","setdiff1d");P(r.dtype===n.dtype,()=>`x and y should have the same dtype, but got x (${r.dtype}) and y (${n.dtype}).`),P(r.rank===1,()=>`x should be 1D tensor, but got x (${r.shape}).`),P(n.rank===1,()=>`y should be 1D tensor, but got y (${n.shape}).`);let a=await r.data(),s=await n.data(),i=new Set(s),o=0;for(let d=0;d<a.length;d++)i.has(a[d])||o++;let l=new ir([o],r.dtype),u=new ir([o],"int32");for(let d=0,h=0;d<a.length;d++)i.has(a[d])||(l.values[h]=a[d],u.values[h]=d,h++);return[l.toTensor(),u.toTensor()]}var ww=cz;function fz(e){let t={x:$(e,"x","sign")};return B.runKernel(od,t)}var kw=W({sign_:fz});function mz(e){let t={x:$(e,"x","sin","float32")};return B.runKernel(Pi,t)}var uA=W({sin_:mz});function gz(e){let t={x:$(e,"x","sinh")};return B.runKernel(vl,t)}var dA=W({sinh_:gz});function yz(e,t,r){let n=$(e,"x","slice1d");return P(n.rank===1,()=>`slice1d expects a rank-1 tensor, but got a rank-${n.rank} tensor`),Pe(n,[t],[r])}var Lf=W({slice1d_:yz});function Az(e,t,r){let n=$(e,"x","slice2d");return P(n.rank===2,()=>`slice2d expects a rank-2 tensor, but got a rank-${n.rank} tensor`),Pe(n,t,r)}var pA=W({slice2d_:Az});function xz(e,t,r){let n=$(e,"x","slice3d");return P(n.rank===3,()=>`slice3d expects a rank-3 tensor, but got a rank-${n.rank} tensor`),Pe(n,t,r)}var Pl=W({slice3d_:xz});function bz(e,t,r){let n=$(e,"x","slice4d");return P(n.rank===4,()=>`slice4d expects a rank-4 tensor, but got a rank-${n.rank} tensor`),Pe(n,t,r)}var Oo=W({slice4d_:bz});function vz(e,t=-1){let r=$(e,"logits","softmax","float32");if(t===-1&&(t=r.rank-1),t!==r.rank-1)throw Error(`Softmax along a non-last dimension is not yet supported. Logits was rank ${r.rank} and dim was ${t}`);let n={logits:r},a={dim:t};return B.runKernel(Di,n,a)}var yd=W({softmax_:vz});function wz(e){P(e.dtype==="complex64",()=>`The dtype for tf.spectral.fft() must be complex64 but got ${e.dtype}.`);let t={input:e};return B.runKernel(pf,t)}var Bf=W({fft_:wz});function kz(e){P(e.dtype==="complex64",()=>`The dtype for tf.spectral.ifft() must be complex64 but got ${e.dtype}.`);let t={input:e};return B.runKernel(hf,t)}var Gp=W({ifft_:kz});function Iz(e){let t=e.shape[e.shape.length-1],r=e.size/t,n;if(t<=2){let a=U(e,[r,t]);n=Gp(a)}else{let a=[r,2*(t-1)],s=U(Up(e),[r,t]),i=U(Rf(e),[r,t]),o=_n(Pe(s,[0,1],[r,t-2]),1),l=L(_n(Pe(i,[0,1],[r,t-2]),1),Se(-1)),u=St([s,o],1),d=St([i,l],1),h=U(Os(u,d),[a[0],a[1]]);n=Gp(h)}if(n=Up(n),e.rank===3&&e.shape[0]!==0){let a=n,s=e.shape[0];n=U(n,[s,n.shape[0]/s,n.shape[1]]),a.dispose()}return n}var hA=W({irfft_:Iz});function Sz(e,t,r=0){let n={x:$(e,"x","split")},a={numOrSizeSplits:t,axis:r};return B.runKernel(kl,n,a)}var Yt=W({split_:Sz});function Cz(e,t){P(e.dtype==="float32",()=>`The dtype for rfft() must be real value but got ${e.dtype}`);let r=e.shape[e.shape.length-1],n=e.size/r,a;if(t!=null&&t<r){let f=e.shape.map(g=>0),m=e.shape.map(g=>g);m[e.shape.length-1]=t,a=Pe(e,f,m),r=t}else if(t!=null&&t>r){let f=e.shape.map(m=>m);f[e.shape.length-1]=t-r,a=St([e,_t(f)],e.shape.length-1),r=t}else a=e;let s=at(a),i=U(Os(a,s),[n,r]),o=Bf(i),l=Math.floor(r/2)+1,u=Up(o),d=Rf(o),h=Yt(u,[l,r-l],u.shape.length-1),p=Yt(d,[l,r-l],d.shape.length-1),c=a.shape.slice();return c[a.shape.length-1]=l,U(Os(h[0],p[0]),c)}var Wf=W({rfft_:Cz});function Tz(e){let t={x:$(e,"x","sqrt","float32")};return B.runKernel(zi,t)}var Er=W({sqrt_:Tz});function Nz(e,t){let r=$(e,"a","squaredDifference"),n=$(t,"b","squaredDifference");[r,n]=Lt(r,n),bt(r.shape,n.shape);let a={a:r,b:n},s={};return B.runKernel(Li,a,s)}var cA=W({squaredDifference_:Nz});function Ez(e,t){let r=$(e,"x","squeeze");return U(r,Tv(r.shape,t).newShape)}var tt=W({squeeze_:Ez});function Rz(e,t=0){let r=Wp(e,"tensors","stack","string_or_numeric");P(r.length>=1,()=>"Pass at least one tensor to tf.stack"),r.length>0&&P(t<=r[0].rank,()=>"Axis must be <= rank of the tensor");let n=r,a={axis:t};return B.runKernel(fl,n,a)}var ur=W({stack_:Rz});function $z(e,t=0){let r={x:$(e,"x","step")},n={alpha:t};return B.runKernel(Ui,r,n)}var Sh=W({step_:$z});function Mz(e,t,r,n,a=0,s=0,i=0,o=0,l=0){let u={x:$(e,"x","stridedSlice","string_or_numeric")},d={begin:t,end:r,strides:n,beginMask:a,endMask:s,ellipsisMask:i,newAxisMask:o,shrinkAxisMask:l};return B.runKernel(Il,u,d)}var Iw=W({stridedSlice_:Mz});function Fz(e){let t={x:$(e,"x","tan","float32")};return B.runKernel(Sl,t)}var Sw=W({tan_:Fz});function Tt(e,t){jo(e);let r=Fa(e,t);if(r.length!==1)throw new Error("tensor1d() requires values to be a flat/TypedArray");return Gi(e,null,r,t)}function ca(e,t,r){if(jo(e),t!=null&&t.length!==2)throw new Error("tensor2d() requires shape to have two numbers");let n=Fa(e,r);if(n.length!==2&&n.length!==1)throw new Error("tensor2d() requires values to be number[][] or flat/TypedArray");if(n.length===1&&t==null)throw new Error("tensor2d() requires shape to be provided when `values` are a flat/TypedArray");return Gi(e,t,n,r)}function Pz(e,t,r){if(jo(e),t!=null&&t.length!==4)throw new Error("tensor4d() requires shape to have four numbers");let n=Fa(e,r);if(n.length!==4&&n.length!==1)throw new Error("tensor4d() requires values to be number[][][][] or flat/TypedArray");if(n.length===1&&t==null)throw new Error("tensor4d() requires shape to be provided when `values` are a flat array");return Gi(e,t,n,r)}function _z(e,t,r){if(jo(e),t!=null&&t.length!==5)throw new Error("tensor5d() requires shape to have five numbers");let n=Fa(e,r);if(n.length!==5&&n.length!==1)throw new Error("tensor5d() requires values to be number[][][][][] or flat/TypedArray");if(n.length===1&&t==null)throw new Error("tensor5d() requires shape to be provided when `values` are a flat array");return Gi(e,t,n,r)}function zz(e,t,r){if(jo(e),t!=null&&t.length!==6)throw new Error("tensor6d() requires shape to have six numbers");let n=Fa(e,r);if(n.length!==6&&n.length!==1)throw new Error("tensor6d() requires values to be number[][][][][][] or flat/TypedArray");if(n.length===1&&t==null)throw new Error("tensor6d() requires shape to be provided when `values` are a flat array");return t=t||n,Gi(e,t,n,r)}function Oz(e,t=1,r=!0){let n=$(e,"x","topk");if(n.rank===0)throw new Error("topk() expects the input to be of rank 1 or higher");let a=n.shape[n.shape.length-1];if(t<0)throw new Error(`'k' passed to topk() must be >= 0 but got ${t}`);if(t>a)throw new Error(`'k' passed to topk() must be <= the last dimension (${a}) but got ${t}`);let s={x:n},i={k:t,sorted:r},[o,l]=B.runKernel(Cl,s,i);return{values:o,indices:l}}var Cw=W({topk_:Oz});function Dz(e,t=0,r=1,n,a){if(n!=null&&n==="bool")throw new Error("Unsupported data type $ { dtype }");let s=new aA(t,r,n,!0,a),i=We(e,n);for(let o=0;o<i.values.length;o++)i.values[o]=s.nextValue();return i.toTensor()}var Vf=W({truncatedNormal_:Dz});function Lz(e,t=0){let r=$(e,"x","unique","string_or_numeric");P(r.rank>0,()=>"The input tensor must be at least 1D");let n={x:r},a={axis:t},[s,i]=B.runKernel(If,n,a);return{values:s,indices:i}}var Ng=W({unique_:Lz});function Bz(e,t,r){let n=$(e,"x","unsortedSegmentSum"),a=$(t,"segmentIds","unsortedSegmentSum","int32");P(wu(r),()=>"numSegments must be of dtype int");let s={x:n,segmentIds:a},i={numSegments:r};return B.runKernel(ch,s,i)}var Tw=W({unsortedSegmentSum_:Bz});function Wz(e,t=0){let r=$(e,"x","unstack","string_or_numeric");P(t>=-r.shape.length&&t<r.shape.length,()=>`Axis = ${t} is not in [-${r.shape.length}, ${r.shape.length})`);let n={value:r},a={axis:t};return B.runKernel(Nl,n,a)}var nn=W({unstack_:Wz});function Nw(e,t){return eA(e,t,"right")}function Ew(e,t=!0,r,n){return B.makeVariable(e,t,r,n)}function Rw(e,t){let r=[];for(let s=0;s<t.length;s++)t[s]&&r.push(s);let n=We(e,"int32"),a=We([r.length,e.length],"int32");for(let s=0;s<r.length;s++){let i=n.indexToLoc(r[s]),o=s*e.length;a.values.set(i,o)}return a.toTensor()}async function Vz(e){let t=$(e,"condition","whereAsync","bool"),r=await t.data(),n=Rw(t.shape,r);return e!==t&&t.dispose(),n}var fA=Vz;async function Uz(e,t,r){let n=$(e,"tensor","boolMask"),a=$(t,"mask","boolMask","bool"),s=r==null?0:r,i=a.rank,o=n.shape;P(i>0,()=>"mask cannot be scalar"),Ur(o.slice(s,s+i),a.shape,"mask's shape must match the first K dimensions of tensor's shape,");let l=1;for(let m=s;m<s+i;m++)l*=o[m];let u=o.slice(0,s).concat([l],o.slice(s+i)),d=U(n,u),h=U(a,[-1]),p=await fA(h),c=tt(p,[1]),f=Cu(d,c,s);return e!==n&&n.dispose(),t!==a&&a.dispose(),c.dispose(),d.dispose(),h.dispose(),p.dispose(),f}var Gz=Uz;function jz(e,t="euclidean",r=null,n=!1){e=$(e,"x","norm");let a=$w(e,t,r),s=a.shape;if(n){let i=Hn(r,e.shape);s=zo(a.shape,i)}return U(a,s)}function $w(e,t,r=null){if(e.rank===0)return ar(e);if(e.rank!==1&&r===null)return $w(U(e,[-1]),t,r);if(e.rank===1||typeof r=="number"||Array.isArray(r)&&r.length===1){if(t===1)return ke(ar(e),r);if(t===1/0)return Ar(ar(e),r);if(t===-1/0)return Bs(ar(e),r);if(t==="euclidean"||t===2)return Er(ke(Ws(ar(e),Se(2,"int32")),r));throw new Error(`Error in norm: invalid ord value: ${t}`)}if(Array.isArray(r)&&r.length===2){if(t===1)return Ar(ke(ar(e),r[0]),r[1]-1);if(t===1/0)return Ar(ke(ar(e),r[1]),r[0]);if(t===-1/0)return Bs(ke(ar(e),r[1]),r[0]);if(t==="fro"||t==="euclidean")return Er(ke(At(e),r));throw new Error(`Error in norm: invalid ord value: ${t}`)}throw new Error(`Error in norm: invalid axis: ${r}`)}var mA=W({norm_:jz});function Hz(e,t,r,n,a=!0){let s=$(e,"v","movingAverage"),i=$(t,"x","movingAverage"),o=$(r,"decay","movingAverage");Kv(s,i),P(Ks(s.shape,i.shape),()=>"Shape mismatch in v and x");let l=Se(1),u=ce(l,o),d=L(ce(i,s),u);if(a){P(n!=null,()=>"When using zeroDebias: true, step is required.");let h=$(n,"step","movingAverage");d=pe(d,ce(l,Ws(o,h)))}return le(s,d)}var qz=W({movingAverage_:Hz});function Xz(e,t,r){let n=$(e,"indices","scatterND","int32"),a=$(t,"updates","scatterND");Ry(a,n,r);let s={indices:n,updates:a},i={shape:r};return B.runKernel(Al,s,i)}var Mw=W({scatterND_:Xz});function Kz(e,t,r,n){if(e.dtype!=="int32")throw new Error(`tf.sparseToDense() expects the indices to be int32 type, but the dtype was ${e.dtype}.`);if(e.rank>2)throw new Error(`sparseIndices should be a scalar, vector, or matrix, but got shape ${e.shape}.`);let a=e.rank>0?e.shape[0]:1,s=e.rank>1?e.shape[1]:1;if(r.length!==s)throw new Error(`outputShape has incorrect number of elements:, ${r.length}, should be: ${s}.`);let i=t.size;if(!(t.rank===0||t.rank===1&&i===a))throw new Error(`sparseValues has incorrect shape ${t.shape}, should be [] or [${a}]`);if(t.dtype!==n.dtype)throw new Error("sparseValues.dtype must match defaultValues.dtype")}function Zz(e,t,r,n=0){let a=$(e,"sparseIndices","sparseToDense","int32"),s=$(t,"sparseValues","sparseToDense"),i=$(n,"defaultValue","sparseToDense",s.dtype);Kz(a,s,r,i);let o={sparseIndices:a,sparseValues:s,defaultValue:i},l={outputShape:r};return B.runKernel(ph,o,l)}var gA=W({sparseToDense_:Zz});function Yz(e,t){let r=$(t,"indices","gatherND","int32"),n={params:$(e,"x","gatherND","string_or_numeric"),indices:r};return B.runKernel(nl,n)}var Fw=W({gatherND_:Yz});function Jz(e,t){if(t==null)return e.shape.slice();if(Ks(e.shape,t))return t;if(e.shape.length===t.length){let r=[];for(let n=0;n<e.shape.length;n++)t[n]==null&&e.shape[n]!=null?r.push(e.shape[n]):r.push(t[n]);return r}return t}function Qz(e,t,r,n){let a=$(e,"x","dropout");if(P(a.dtype==="float32",()=>`x has to be a floating point tensor since it's going to be scaled, but got a ${a.dtype} tensor instead.`),P(t>=0&&t<1,()=>`rate must be a float in the range [0, 1), but got ${t}.`),t===0)return e instanceof nt?a.clone():a;let s=Jz(a,r),i=1-t,o=pe(kh(le(gd(s,0,1,"float32",n),i)),i);return L(a,o)}var Pw=W({dropout_:Qz});function _w(e){return Math.floor(Math.pow(2,Math.ceil(Math.log(e)/Math.log(2))))}function yA(e,t,r){let n=1-e%2,a=new Float32Array(e);for(let s=0;s<e;++s){let i=2*Math.PI*s/(e+n-1);a[s]=t-r*Math.cos(i)}return Tt(a,"float32")}async function eO(e,t,r=1){let n=$(e,"predictions","inTopK"),a=$(t,"targets","inTopK");P(n.rank>1,()=>`inTopK() expects the predictions to be of rank 2 or higher, but got ${n.rank}`),P(n.rank-1===a.rank,()=>`predictions rank should be 1 larger than targets rank, but got predictions rank ${n.rank} and targets rank ${a.rank}`),Ur(n.shape.slice(0,n.shape.length-1),a.shape,"predictions's shape should be align with the targets' shape, except the last dimension.");let s=n.shape[n.shape.length-1];P(r>0&&r<=s,()=>`'k' passed to inTopK() must be > 0 && <= the predictions last dimension (${s}), but got ${r}`);let i=await n.data(),o=await a.data(),[l,u]=[i.length/s,s],d=Nv("bool",l);for(let h=0;h<l;h++){let p=h*u,c=i.subarray(p,p+u),f=[];for(let m=0;m<c.length;m++)f.push({value:c[m],index:m});f.sort((m,g)=>g.value-m.value),d[h]=0;for(let m=0;m<r;m++)if(f[m].index===o[h]){d[h]=1;break}}return e!==n&&n.dispose(),t!==a&&a.dispose(),ft(d,a.shape,"bool")}var tO=eO,Vs={};Le(Vs,{conv2d:()=>aO,depthwiseConv2d:()=>lO,matMul:()=>dO});function rO(e,t,r,n,a,s="NHWC",i){let o=e;e.rank===3&&(o=U(e,[1,e.shape[0],e.shape[1],e.shape[2]]));let l=t;l.rank===3&&(l=U(t,[1,t.shape[0],t.shape[1],t.shape[2]])),P(o.rank===4,()=>`Error in conv2dDerFilter: input must be rank 4, but got shape ${o.shape}.`),P(l.rank===4,()=>`Error in conv2dDerFilter: dy must be rank 4, but got shape ${l.shape}.`),P(r.length===4,()=>`Error in conv2dDerFilter: filterShape must be length 4, but got ${r}.`);let u=s==="NHWC"?o.shape[3]:o.shape[1],d=s==="NHWC"?l.shape[3]:l.shape[1];P(u===r[2],()=>`Error in conv2dDerFilter: depth of input ${u}) must match input depth in filter (${r[2]}.`),P(d===r[3],()=>`Error in conv2dDerFilter: depth of dy (${d}) must match output depth for filter (${r[3]}).`),jr("conv2dDerFilter",a,i);let h={x:o,dy:l},p={strides:n,pad:a,dataFormat:s,dimRoundingMode:i,filterShape:r};return B.runKernel(rf,h,p)}var AA=W({conv2DBackpropFilter_:rO});function Uf(e,t,r){if(r==null||r==="linear")return e;if(r==="relu")return L(e,Sh(t));throw new Error(`Cannot compute gradient for fused activation ${r}.`)}function Gf(e,t){let r=t,n=Jt(e.shape,t.shape);return n.length>0&&(r=ke(r,n)),U(r,e.shape)}function jf(e,t,r,n){if(t==="linear")return e;if(t==="relu")return Oa(e);if(t==="elu")return wh(e);if(t==="relu6")return sA(e);if(t==="prelu")return Df(e,r);if(t==="leakyrelu")return $f(e,n);if(t==="sigmoid")return Tr(e);throw new Error(`Unknown fused activation ${t}.`)}var Hf=(e,t)=>!(e>0)||t==="linear";function nO({x:e,filter:t,strides:r,pad:n,dataFormat:a="NHWC",dilations:s=[1,1],dimRoundingMode:i,bias:o,activation:l="linear",preluActivationWeights:u,leakyreluAlpha:d}){if(l=l||"linear",Hf(B.state.gradientDepth,l)===!1){let w=Ls(e,t,r,n,a,s,i);return o!=null&&(w=le(w,o)),jf(w,l,u,d)}let h=$(e,"x","conv2d","float32"),p=$(t,"filter","conv2d","float32"),c=h,f=!1;h.rank===3&&(f=!0,c=U(h,[1,h.shape[0],h.shape[1],h.shape[2]])),P(c.rank===4,()=>`Error in fused conv2d: input must be rank 4, but got rank ${c.rank}.`),P(p.rank===4,()=>`Error in fused conv2d: filter must be rank 4, but got rank ${p.rank}.`),jr("fused conv2d",n,i),P(c.shape[3]===p.shape[2],()=>`Error in conv2d: depth of input (${c.shape[3]}) must match input depth for filter ${p.shape[2]}.`),P(za(r,s),()=>`Error in conv2D: Either strides or dilations must be 1. Got strides ${r} and dilations '${s}'`),P(a==="NHWC",()=>`Error in conv2d: got dataFormat of ${a} but only NHWC is currently supported.`);let m=bh(c.shape,p.shape,r,s,n,i),g;o!=null&&(g=$(o,"bias","fused conv2d"),[g]=Lt(g,h),bt(m.outShape,g.shape));let y;u!=null&&(y=$(u,"prelu weights","fused conv2d"));let A=(w,S)=>{let[N,E,R,z]=S,M=Uf(w,R,l);P(Ds(s),()=>`Error in gradient of fused conv2D: dilation rates greater than 1 are not yet supported in gradients. Got dilations '${s}'`);let I=Uy(E.shape,M,N,r,n),_=AA(E,M,N.shape,r,n),O=[I,_];if(z!=null){let q=Gf(z,M);O.push(q)}return O},x={x:c,filter:p,bias:g,preluActivationWeights:y},b={strides:r,pad:n,dataFormat:a,dilations:s,dimRoundingMode:i,activation:l,leakyreluAlpha:d};return o==null?Pa((w,S,N)=>{let E=B.runKernel(_s,x,b);return N([S,w,E]),f&&(E=U(E,[E.shape[1],E.shape[2],E.shape[3]])),{value:E,gradFunc:A}})(c,p):Pa((w,S,N,E)=>{let R=B.runKernel(_s,x,b);return E([S,w,R,N]),f&&(R=U(R,[R.shape[1],R.shape[2],R.shape[3]])),{value:R,gradFunc:A}})(c,p,g)}var aO=W({fusedConv2d_:nO});function sO(e,t,r,n,a,s=[1,1],i){let o=e;e.rank===3&&(o=U(e,[1,e.shape[0],e.shape[1],e.shape[2]]));let l=t;l.rank===3&&(l=U(t,[1,t.shape[0],t.shape[1],t.shape[2]]));let u={x:o,dy:l},d={strides:n,pad:a,dimRoundingMode:i,dilations:s,filterShape:r};return B.runKernel(of,u,d)}var zw=W({depthwiseConv2dNativeBackpropFilter_:sO});function iO(e,t,r,n,a,s=[1,1],i){let o=t,l=!1;t.rank===3&&(l=!0,o=U(t,[1,t.shape[0],t.shape[1],t.shape[2]]));let u={dy:o,filter:r},d={strides:n,pad:a,dimRoundingMode:i,dilations:s,inputShape:e},h=B.runKernel(lf,u,d);return l?U(h,[h.shape[1],h.shape[2],h.shape[3]]):h}var Ow=W({depthwiseConv2dNativeBackpropInput_:iO});function oO({x:e,filter:t,strides:r,pad:n,dataFormat:a="NHWC",dilations:s=[1,1],dimRoundingMode:i,bias:o,activation:l="linear",preluActivationWeights:u,leakyreluAlpha:d}){if(Hf(B.state.gradientDepth,l)===!1){let w=vh(e,t,r,n,a,s,i);return o!=null&&(w=le(w,o)),jf(w,l,u,d)}let h=$(e,"x","depthwiseConv2d","float32"),p=$(t,"filter","depthwiseConv2d","float32"),c=h,f=!1;h.rank===3&&(f=!0,c=U(h,[1,h.shape[0],h.shape[1],h.shape[2]])),P(c.rank===4,()=>`Error in fused depthwiseConv2d: input must be rank 4, but got rank ${c.rank}.`),P(p.rank===4,()=>`Error in fused depthwiseConv2d: filter must be rank 4, but got rank ${p.rank}.`),P(c.shape[3]===p.shape[2],()=>`Error in fused depthwiseConv2d: number of input channels (${c.shape[3]}) must match the inChannels dimension in filter ${p.shape[2]}.`),s==null&&(s=[1,1]),P(za(r,s),()=>`Error in fused depthwiseConv2d: Either strides or dilations must be 1. Got strides ${r} and dilations '${s}'`),jr("fused depthwiseConv2d",n,i);let m=bh(c.shape,p.shape,r,s,n,i,!0),g;o!=null&&(g=$(o,"bias","fused conv2d"),[g]=Lt(g,h),bt(m.outShape,g.shape));let y;u!=null&&(y=$(u,"prelu weights","fused depthwiseConv2d"));let A=(w,S)=>{P(Ds(s),()=>`Error in gradient of fused depthwiseConv2d: dilation rates greater than 1 are not yet supported. Got dilations '${s}'`);let[N,E,R,z]=S,M=Uf(w,R,l),I=Ow(E.shape,M,N,r,n,s,i),_=zw(E,M,N.shape,r,n,s,i);if(z!=null){let O=Gf(g,M);return[I,_,O]}return[I,_]},x={x:c,filter:p,bias:g,preluActivationWeights:y},b={strides:r,pad:n,dataFormat:a,dilations:s,dimRoundingMode:i,activation:l,leakyreluAlpha:d};return o==null?Pa((w,S,N)=>{let E=B.runKernel(zs,x,b);return N([S,w,E]),f&&(E=U(E,[E.shape[1],E.shape[2],E.shape[3]])),{value:E,gradFunc:A}})(c,p):Pa((w,S,N,E)=>{let R=B.runKernel(zs,x,b);return E([S,w,R,N]),f&&(R=U(R,[R.shape[1],R.shape[2],R.shape[3]])),{value:R,gradFunc:A}})(c,p,g)}var lO=W({fusedDepthwiseConv2d_:oO});function uO({a:e,b:t,transposeA:r=!1,transposeB:n=!1,bias:a,activation:s="linear",preluActivationWeights:i,leakyreluAlpha:o}){if(Hf(B.state.gradientDepth,s)===!1){let z=Ze(e,t,r,n);return a!=null&&(z=le(z,a)),jf(z,s,i,o)}let l=$(e,"a","fused matMul"),u=$(t,"b","fused matMul");[l,u]=Lt(l,u);let d=r?l.shape[l.rank-2]:l.shape[l.rank-1],h=n?u.shape[u.rank-1]:u.shape[u.rank-2],p=r?l.shape[l.rank-1]:l.shape[l.rank-2],c=n?u.shape[u.rank-2]:u.shape[u.rank-1],f=l.shape.slice(0,-2),m=u.shape.slice(0,-2),g=It(f),y=It(m);P(d===h,()=>`Error in fused matMul: inner shapes (${d}) and (${h}) of Tensors with shapes ${l.shape} and ${u.shape} and transposeA=${r} and transposeB=${n} must match.`);let A=bt(l.shape.slice(0,-2),u.shape.slice(0,-2)).concat([p,c]),x=r?U(l,[g,d,p]):U(l,[g,p,d]),b=n?U(u,[y,c,h]):U(u,[y,h,c]),w;a!=null&&(w=$(a,"bias","fused matMul"),[w]=Lt(w,l),bt(A,w.shape));let S;i!=null&&(S=$(i,"prelu weights","fused matMul"));let N=(z,M)=>{let[I,_,O,q]=M,K=Uf(U(z,O.shape),O,s),D,Q;if(!r&&!n?(D=Ze(K,_,!1,!0),Q=Ze(I,K,!0,!1)):!r&&n?(D=Ze(K,_,!1,!1),Q=Ze(K,I,!0,!1)):r&&!n?(D=Ze(_,K,!1,!0),Q=Ze(I,K,!1,!1)):(D=Ze(_,K,!0,!0),Q=Ze(K,I,!0,!0)),a!=null){let V=Gf(q,K);return[D,Q,V]}else return[D,Q]},E={a:x,b,bias:w,preluActivationWeights:S},R={transposeA:r,transposeB:n,activation:s,leakyreluAlpha:o};return a==null?Pa((z,M,I)=>{let _=B.runKernel(Ps,E,R);return I([z,M,_]),{value:U(_,A),gradFunc:N}})(x,b):Pa((z,M,I,_)=>{let O=B.runKernel(Ps,E,R);return _([z,M,O,I]),{value:U(O,A),gradFunc:N}})(x,b,w)}var dO=W({fusedMatMul_:uO});function pO(e){return yA(e,.54,.46)}var hO=W({hammingWindow_:pO});function cO(e){return yA(e,.5,.5)}var Dw=W({hannWindow_:cO});function fO(e,t,r,n=!1,a=0){let s=0,i=[];for(;s+t<=e.size;)i.push(Pe(e,s,t)),s+=r;if(n)for(;s<e.size;){let o=s+t-e.size,l=St([Pe(e,s,t-o),cd([o],a)]);i.push(l),s+=r}return i.length===0?ca([],[0,t]):U(St(i),[i.length,t])}var Lw=W({frame_:fO});function mO(e,t,r,n,a=Dw){n==null&&(n=_w(t));let s=Lw(e,t,r),i=L(s,a(t));return Wf(i,n)}var gO=W({stft_:mO});function yO(e,t,r,n,a="bilinear",s=0){let i=$(e,"image","cropAndResize"),o=$(t,"boxes","cropAndResize","float32"),l=$(r,"boxInd","cropAndResize","int32"),u=o.shape[0];P(i.rank===4,()=>`Error in cropAndResize: image must be rank 4,but got rank ${i.rank}.`),P(o.rank===2&&o.shape[1]===4,()=>`Error in cropAndResize: boxes must be have size [${u},4] but had shape ${o.shape}.`),P(l.rank===1&&l.shape[0]===u,()=>`Error in cropAndResize: boxInd must be have size [${u}] but had shape ${o.shape}.`),P(n.length===2,()=>`Error in cropAndResize: cropSize must be of length 2, but got length ${n.length}.`),P(n[0]>=1&&n[1]>=1,()=>`cropSize must be atleast [1,1], but was ${n}`),P(a==="bilinear"||a==="nearest",()=>`method must be bilinear or nearest, but was ${a}`);let d={image:i,boxes:o,boxInd:l},h={method:a,extrapolationValue:s,cropSize:n};return B.runKernel(Zo,d,h)}var AO=W({cropAndResize_:yO});function xO(e){let t=$(e,"image","flipLeftRight","float32");P(t.rank===4,()=>`Error in flipLeftRight: image must be rank 4,but got rank ${t.rank}.`);let r={image:t};return B.runKernel(tl,r,{})}var bO=W({flipLeftRight_:xO});function vO(e){let t=$(e,"image","grayscaleToRGB"),r=t.rank-1,n=t.shape[r];P(t.rank>=2,()=>`Error in grayscaleToRGB: images must be at least rank 2, but got rank ${t.rank}.`),P(n===1,()=>`Error in grayscaleToRGB: last dimension of a grayscale image should be size 1, but got size ${n}.`);let a=new Array(t.rank);return a.fill(1,0,r),a[r]=3,Gn(t,a)}var wO=W({grayscaleToRGB_:vO});function kO(e,t,r=0,n=.5){let a=$(e,"image","rotateWithOffset","float32");P(a.rank===4,()=>`Error in rotateWithOffset: image must be rank 4,but got rank ${a.rank}.`);let s={image:a},i={radians:t,fillValue:r,center:n};return B.runKernel(Rl,s,i)}var IO=W({rotateWithOffset_:kO});function Ad(e,t,r,n,a,s){n==null&&(n=.5),a==null&&(a=Number.NEGATIVE_INFINITY),s==null&&(s=0);let i=e.shape[0];return r=Math.min(r,i),P(0<=n&&n<=1,()=>`iouThreshold must be in [0, 1], but was '${n}'`),P(e.rank===2,()=>`boxes must be a 2D tensor, but was of rank '${e.rank}'`),P(e.shape[1]===4,()=>`boxes must have 4 columns, but 2nd dimension was ${e.shape[1]}`),P(t.rank===1,()=>"scores must be a 1D tensor"),P(t.shape[0]===i,()=>`scores has incompatible shape with boxes. Expected ${i}, but was ${t.shape[0]}`),P(0<=s&&s<=1,()=>`softNmsSigma must be in [0, 1], but was '${s}'`),{maxOutputSize:r,iouThreshold:n,scoreThreshold:a,softNmsSigma:s}}function SO(e,t,r,n=.5,a=Number.NEGATIVE_INFINITY){let s=$(e,"boxes","nonMaxSuppression","float32"),i=$(t,"scores","nonMaxSuppression","float32"),o=Ad(s,i,r,n,a);r=o.maxOutputSize,n=o.iouThreshold,a=o.scoreThreshold;let l={maxOutputSize:r,iouThreshold:n,scoreThreshold:a};return B.runKernel(dl,{boxes:s,scores:i},l)}var CO=W({nonMaxSuppression_:SO});function TO(e,t,r){let n=NO(e,t,r),a=n<0?-(n+1):n;e.splice(a,0,t)}function NO(e,t,r){return RO(e,t,r||EO)}function EO(e,t){return e>t?1:e<t?-1:0}function RO(e,t,r){let n=0,a=e.length,s=0,i=!1;for(;n<a;){s=n+(a-n>>>1);let o=r(t,e[s]);o>0?n=s+1:(a=s,i=!o)}return i?n:-n-1}function Bw(e,t,r,n,a){return xA(e,t,r,n,a,0)}function Ww(e,t,r,n,a,s){return xA(e,t,r,n,a,0,!1,s,!0)}function Vw(e,t,r,n,a,s){return xA(e,t,r,n,a,s,!0)}function xA(e,t,r,n,a,s,i=!1,o=!1,l=!1){let u=[];for(let g=0;g<t.length;g++)t[g]>a&&u.push({score:t[g],boxIndex:g,suppressBeginIndex:0});u.sort(X3);let d=s>0?-.5/s:0,h=[],p=[];for(;h.length<r&&u.length>0;){let g=u.pop(),{score:y,boxIndex:A,suppressBeginIndex:x}=g;if(y<a)break;let b=!1;for(let w=h.length-1;w>=x;--w){let S=$O(e,A,h[w]);if(S>=n){b=!0;break}if(g.score=g.score*MO(n,d,S),g.score<=a)break}g.suppressBeginIndex=h.length,b||(g.score===y?(h.push(A),p.push(g.score)):g.score>a&&TO(u,g,X3))}let c=h.length,f=r-c;o&&f>0&&(h.push(...new Array(f).fill(0)),p.push(...new Array(f).fill(0)));let m={selectedIndices:h};return i&&(m.selectedScores=p),l&&(m.validOutputs=c),m}function $O(e,t,r){let n=e.subarray(t*4,t*4+4),a=e.subarray(r*4,r*4+4),s=Math.min(n[0],n[2]),i=Math.min(n[1],n[3]),o=Math.max(n[0],n[2]),l=Math.max(n[1],n[3]),u=Math.min(a[0],a[2]),d=Math.min(a[1],a[3]),h=Math.max(a[0],a[2]),p=Math.max(a[1],a[3]),c=(o-s)*(l-i),f=(h-u)*(p-d);if(c<=0||f<=0)return 0;let m=Math.max(s,u),g=Math.max(i,d),y=Math.min(o,h),A=Math.min(l,p),x=Math.max(y-m,0)*Math.max(A-g,0);return x/(c+f-x)}function MO(e,t,r){let n=Math.exp(t*r*r);return r<=e?n:0}function X3(e,t){return e.score-t.score||e.score===t.score&&t.boxIndex-e.boxIndex}async function FO(e,t,r,n=.5,a=Number.NEGATIVE_INFINITY){let s=$(e,"boxes","nonMaxSuppressionAsync"),i=$(t,"scores","nonMaxSuppressionAsync"),o=Ad(s,i,r,n,a);r=o.maxOutputSize,n=o.iouThreshold,a=o.scoreThreshold;let l=await Promise.all([s.data(),i.data()]),u=l[0],d=l[1],{selectedIndices:h}=Bw(u,d,r,n,a);return s!==e&&s.dispose(),i!==t&&i.dispose(),Tt(h,"int32")}var PO=FO;function _O(e,t,r,n=.5,a=Number.NEGATIVE_INFINITY,s=0){let i=$(e,"boxes","nonMaxSuppression"),o=$(t,"scores","nonMaxSuppression"),l=Ad(i,o,r,n,a,s);r=l.maxOutputSize,n=l.iouThreshold,a=l.scoreThreshold,s=l.softNmsSigma;let u={boxes:i,scores:o},d={maxOutputSize:r,iouThreshold:n,scoreThreshold:a,softNmsSigma:s},h=B.runKernel(pl,u,d);return{selectedIndices:h[0],selectedScores:h[1]}}var zO=W({nonMaxSuppressionWithScore_:_O});async function OO(e,t,r,n=.5,a=Number.NEGATIVE_INFINITY,s=0){let i=$(e,"boxes","nonMaxSuppressionAsync"),o=$(t,"scores","nonMaxSuppressionAsync"),l=Ad(i,o,r,n,a,s);r=l.maxOutputSize,n=l.iouThreshold,a=l.scoreThreshold,s=l.softNmsSigma;let u=await Promise.all([i.data(),o.data()]),d=u[0],h=u[1],{selectedIndices:p,selectedScores:c}=Vw(d,h,r,n,a,s);return i!==e&&i.dispose(),o!==t&&o.dispose(),{selectedIndices:Tt(p,"int32"),selectedScores:Tt(c)}}var DO=OO;function LO(e,t,r,n=.5,a=Number.NEGATIVE_INFINITY,s=!1){let i=$(e,"boxes","nonMaxSuppression"),o=$(t,"scores","nonMaxSuppression"),l=Ad(i,o,r,n,a,null),u=l.maxOutputSize,d=l.iouThreshold,h=l.scoreThreshold,p={boxes:i,scores:o},c={maxOutputSize:u,iouThreshold:d,scoreThreshold:h,padToMaxOutputSize:s},f=B.runKernel(rd,p,c);return{selectedIndices:f[0],validOutputs:f[1]}}var BO=W({nonMaxSuppressionPadded_:LO});async function WO(e,t,r,n=.5,a=Number.NEGATIVE_INFINITY,s=!1){let i=$(e,"boxes","nonMaxSuppressionAsync"),o=$(t,"scores","nonMaxSuppressionAsync"),l=Ad(i,o,r,n,a,null),u=l.maxOutputSize,d=l.iouThreshold,h=l.scoreThreshold,[p,c]=await Promise.all([i.data(),o.data()]),{selectedIndices:f,validOutputs:m}=Ww(p,c,u,d,h,s);return i!==e&&i.dispose(),o!==t&&o.dispose(),{selectedIndices:Tt(f,"int32"),validOutputs:Se(m,"int32")}}var VO=WO;function UO(e,t,r=!1,n=!1){let a=$(e,"images","resizeBilinear");P(a.rank===3||a.rank===4,()=>`Error in resizeBilinear: x must be rank 3 or 4, but got rank ${a.rank}.`),P(t.length===2,()=>`Error in resizeBilinear: new shape must 2D, but got shape ${t}.`),P(n===!1||r===!1,()=>"Error in resizeBilinear: If halfPixelCenters is true, alignCorners must be false.");let s=a,i=!1;a.rank===3&&(i=!0,s=U(a,[1,a.shape[0],a.shape[1],a.shape[2]]));let[]=t,o={images:s},l={alignCorners:r,halfPixelCenters:n,size:t},u=B.runKernel($i,o,l);return i?U(u,[u.shape[1],u.shape[2],u.shape[3]]):u}var GO=W({resizeBilinear_:UO});function jO(e,t,r=!1,n=!1){let a=$(e,"images","resizeNearestNeighbor");P(a.rank===3||a.rank===4,()=>`Error in resizeNearestNeighbor: x must be rank 3 or 4, but got rank ${a.rank}.`),P(t.length===2,()=>`Error in resizeNearestNeighbor: new shape must 2D, but got shape ${t}.`),P(a.dtype==="float32"||a.dtype==="int32",()=>"`images` must have `int32` or `float32` as dtype"),P(n===!1||r===!1,()=>"Error in resizeNearestNeighbor: If halfPixelCenters is true, alignCorners must be false.");let s=a,i=!1;a.rank===3&&(i=!0,s=U(a,[1,a.shape[0],a.shape[1],a.shape[2]]));let[]=t,o={images:s},l={alignCorners:r,halfPixelCenters:n,size:t},u=B.runKernel(sd,o,l);return i?U(u,[u.shape[1],u.shape[2],u.shape[3]]):u}var HO=W({resizeNearestNeighbor_:jO});function qO(e,t="binary",r=!1,n=.5){let a=$(e,"image","threshold"),s=.2989,i=.587,o=.114,l=a.shape[0]*a.shape[1],u=L(Tt([n]),255),d,h,p,c;if(P(a.rank===3,()=>`Error in threshold: image must be rank 3,but got rank ${a.rank}.`),P(a.shape[2]===3||a.shape[2]===1,()=>`Error in threshold: image color channel must be equal to 3 or 1but got ${a.shape[2]}.`),P(a.dtype==="int32"||a.dtype==="float32",()=>`Error in dtype: image dtype must be int32 or float32,but got dtype ${a.dtype}.`),P(t==="otsu"||t==="binary",()=>`Method must be binary or otsu, but was ${t}`),a.shape[2]===3){[d,h,p]=Yt(a,[1,1,1],-1);let m=L(d,s),g=L(h,i),y=L(p,o);c=le(le(m,g),y)}else c=e;if(t==="otsu"){let m=Wy(me(iA(c),"int32"),ft([]),256);u=XO(m,l)}let f=r?Fl(c,u):mn(c,u);return me(L(f,255),"int32")}function XO(e,t){let r=Tt([-1]),n=Tt([0]),a=Tt([0]),s,i,o,l,u,d;for(let h=0;h<e.size-1;h++){s=Pe(e,0,h+1),i=Pe(e,h+1),u=pe(ke(s),t),d=pe(ke(i),t);let p=ke(L(s,Nu(0,s.size)));o=pe(p,ke(s));let c=cd(i.shape,s.size),f=le(Nu(0,i.size),c),m=L(i,f);l=pe(ke(m),ke(i));let g=ce(o,l),y=ce(o,l),A=L(u,d);a=L(L(A,g),y);let x=mn(a,n);n=Vr(x,a,n),r=Vr(x,Tt([h]),r)}return r}var KO=W({threshold_:qO});function ZO(e,t,r="nearest",n="constant",a=0,s){let i=$(e,"image","transform","float32"),o=$(t,"transforms","transform","float32");P(i.rank===4,()=>`Error in transform: image must be rank 4,but got rank ${i.rank}.`),P(o.rank===2&&(o.shape[0]===i.shape[0]||o.shape[0]===1)&&o.shape[1]===8,()=>"Error in transform: Input transform should be batch x 8 or 1 x 8"),P(s==null||s.length===2,()=>`Error in transform: outputShape must be [height, width] or null, but got ${s}.`);let l={image:i,transforms:o},u={interpolation:r,fillMode:n,fillValue:a,outputShape:s};return B.runKernel(Tl,l,u)}var YO=W({transform_:ZO});function JO(e,t,r){P(t%1===0,()=>`bandPart(): numLower must be an integer, got ${t}.`),P(r%1===0,()=>`bandPart(): numUpper must be an integer, got ${r}.`);let n=$(e,"a","bandPart");P(n.rank>=2,()=>`bandPart(): Rank must be at least 2, got ${n.rank}.`);let a=n.shape,[s,i]=n.shape.slice(-2);if(!(t<=s))throw new Error(`bandPart(): numLower (${t}) must not be greater than the number of rows (${s}).`);if(!(r<=i))throw new Error(`bandPart(): numUpper (${r}) must not be greater than the number of columns (${i}).`);t<0&&(t=s),r<0&&(r=i);let o=U(Nu(0,s,1,"int32"),[-1,1]),l=Nu(0,i,1,"int32"),u=ce(o,l),d=ga(Fl(u,Se(+t,"int32")),Ml(u,Se(-r,"int32"))),h=_t([s,i],n.dtype);return U(ur(nn(U(n,[-1,s,i])).map(p=>Vr(d,p,h))),a)}var QO=W({bandPart_:JO});function eD(e){let t;if(Array.isArray(e)){t=!1,P(e!=null&&e.length>0,()=>"Gram-Schmidt process: input must not be null, undefined, or empty");let a=e[0].shape[0];for(let s=1;s<e.length;++s)P(e[s].shape[0]===a,()=>`Gram-Schmidt: Non-unique lengths found in the input vectors: (${e[s].shape[0]} vs. ${a})`)}else t=!0,e=Yt(e,e.shape[0],0).map(a=>tt(a,[0]));P(e.length<=e[0].shape[0],()=>`Gram-Schmidt: Number of vectors (${e.length}) exceeds number of dimensions (${e[0].shape[0]}).`);let r=[],n=e;for(let a=0;a<e.length;++a)r.push(B.tidy(()=>{let s=n[a];if(a>0)for(let i=0;i<a;++i){let o=L(ke(L(r[i],s)),r[i]);s=ce(s,o)}return pe(s,mA(s,"euclidean"))}));return t?ur(r,0):r}var tD=W({gramSchmidt_:eD});function rD(e,t=!1){if(P(e.rank>=2,()=>`qr() requires input tensor to have a rank >= 2, but got rank ${e.rank}`),e.rank===2)return K3(e,t);{let r=e.shape.slice(0,e.shape.length-2).reduce((l,u)=>l*u),n=nn(U(e,[r,e.shape[e.shape.length-2],e.shape[e.shape.length-1]]),0),a=[],s=[];n.forEach(l=>{let[u,d]=K3(l,t);a.push(u),s.push(d)});let i=U(ur(a,0),e.shape),o=U(ur(s,0),e.shape);return[i,o]}}function K3(e,t=!1){return B.tidy(()=>{P(e.shape.length===2,()=>`qr2d() requires a 2D Tensor, but got a ${e.shape.length}D Tensor.`);let r=e.shape[0],n=e.shape[1],a=Xy(r),s=Wr(e),i=ca([[1]],[1,1]),o=Wr(i),l=r>=n?n:r;for(let u=0;u<l;++u){let d=s,h=o,p=a;[o,s,a]=B.tidy(()=>{let c=Pe(s,[u,u],[r-u,1]),f=mA(c),m=Pe(s,[u,u],[1,1]),g=Vr(mn(m,0),ca([[-1]]),ca([[1]])),y=ce(m,L(g,f)),A=pe(c,y);A.shape[0]===1?o=Wr(i):o=St([i,Pe(A,[1,0],[A.shape[0]-1,A.shape[1]])],0);let x=Dt(pe(Ze(g,y),f)),b=Pe(s,[u,0],[r-u,n]),w=L(x,o),S=et(o);if(u===0)s=ce(b,Ze(w,Ze(S,b)));else{let R=ce(b,Ze(w,Ze(S,b)));s=St([Pe(s,[0,0],[u,n]),R],0)}let N=et(w),E=Pe(a,[0,u],[r,a.shape[1]-u]);if(u===0)a=ce(E,Ze(Ze(E,o),N));else{let R=ce(E,Ze(Ze(E,o),N));a=St([Pe(a,[0,0],[r,u]),R],1)}return[o,s,a]}),te([d,h,p])}return!t&&r>n&&(a=Pe(a,[0,0],[r,n]),s=Pe(s,[0,0],[n,n])),[a,s]})}var nD=W({qr_:rD}),Uw=(e=>(e[e.NONE=0]="NONE",e[e.MEAN=1]="MEAN",e[e.SUM=2]="SUM",e[e.SUM_BY_NONZERO_WEIGHTS=3]="SUM_BY_NONZERO_WEIGHTS",e))(Uw||{});function aD(e,t,r=3){let n=$(e,"losses","computeWeightedLoss"),a=null;t!=null&&(a=$(t,"weights","computeWeightedLoss"));let s=a==null?n:L(n,a);if(r===0)return s;if(r===2)return ke(s);if(r===1){if(a==null)return Vt(s);{let i=n.size/a.size,o=pe(ke(s),ke(a));return i>1?pe(o,Se(i)):o}}if(r===3){if(a==null)return pe(ke(s),Se(n.size));{let i=L(a,hn(n.shape)),o=me(ke(Tu(i,Se(0))),"float32");return pe(ke(s),o)}}throw Error(`Unknown reduction: ${r}`)}var rs=W({computeWeightedLoss_:aD});function sD(e,t,r,n=3){let a=$(e,"labels","absoluteDifference"),s=$(t,"predictions","absoluteDifference"),i=null;r!=null&&(i=$(r,"weights","absoluteDifference")),Ur(a.shape,s.shape,"Error in absoluteDifference: ");let o=ar(ce(a,s));return rs(o,i,n)}var iD=W({absoluteDifference_:sD});function oD(e,t,r,n,a=3){let s=$(e,"labels","cosineDistance"),i=$(t,"predictions","cosineDistance"),o=null;n!=null&&(o=$(n,"weights","cosineDistance")),Ur(s.shape,i.shape,"Error in cosineDistance: ");let l=Se(1),u=ce(l,ke(L(s,i),r,!0));return rs(u,o,a)}var lD=W({cosineDistance_:oD});function uD(e,t,r,n=3){let a=$(e,"labels","hingeLoss"),s=$(t,"predictions","hingeLoss"),i=null;r!=null&&(i=$(r,"weights","hingeLoss")),Ur(a.shape,s.shape,"Error in hingeLoss: ");let o=Se(1);a=ce(L(Se(2),a),o);let l=Oa(ce(o,L(a,s)));return rs(l,i,n)}var dD=W({hingeLoss_:uD});function pD(e,t,r,n=1,a=3){let s=$(e,"labels","huberLoss"),i=$(t,"predictions","huberLoss"),o=null;r!=null&&(o=$(r,"weights","huberLoss")),Ur(s.shape,i.shape,"Error in huberLoss: ");let l=Se(n),u=ar(ce(i,s)),d=Ih(u,l),h=ce(u,d),p=le(L(Se(.5),At(d)),L(l,h));return rs(p,o,a)}var hD=W({huberLoss_:pD});function cD(e,t,r,n=1e-7,a=3){let s=$(e,"labels","logLoss"),i=$(t,"predictions","logLoss"),o=null;r!=null&&(o=$(r,"weights","logLoss")),Ur(s.shape,i.shape,"Error in logLoss: ");let l=Se(1),u=Se(n),d=Dt(L(s,Fn(le(i,u)))),h=L(ce(l,s),Fn(le(ce(l,i),u))),p=ce(d,h);return rs(p,o,a)}var fD=W({logLoss_:cD});function mD(e,t,r,n=3){let a=$(e,"labels","meanSquaredError"),s=$(t,"predictions","meanSquaredError"),i=null;r!=null&&(i=$(r,"weights","meanSquaredError")),Ur(a.shape,s.shape,"Error in meanSquaredError: ");let o=cA(a,s);return rs(o,i,n)}var gD=W({meanSquaredError_:mD});function yD(e,t){let r=$(e,"labels","sigmoidCrossEntropyWithLogits"),n=$(t,"logits","sigmoidCrossEntropyWithLogits");Ur(r.shape,n.shape,"Error in sigmoidCrossEntropyWithLogits: ");let a=Oa(n),s=L(n,r),i=Mf(Mn(Dt(ar(n))));return le(ce(a,s),i)}function AD(e,t,r,n=0,a=3){let s=$(e,"multiClassLabels","sigmoidCrossEntropy"),i=$(t,"logits","sigmoidCrossEntropy"),o=null;if(r!=null&&(o=$(r,"weights","sigmoidCrossEntropy")),Ur(s.shape,i.shape,"Error in sigmoidCrossEntropy: "),n>0){let u=Se(n),d=Se(1),h=Se(.5);s=le(L(s,ce(d,u)),L(h,u))}let l=yD(s,i);return rs(l,o,a)}var xD=W({sigmoidCrossEntropy_:AD});function bD(e,t,r=-1){if(r===-1&&(r=t.rank-1),r!==t.rank-1)throw Error(`Softmax cross entropy along a non-last dimension is not yet supported. Labels / logits was rank ${t.rank} and dim was ${r}`);return Pa((n,a,s)=>{let i=fw(a,[r],!0),o=ce(me(a,"float32"),i);s([n,o]);let l=Dt(L(o,n));return{value:ke(l,[r]),gradFunc:(u,d)=>{let[h,p]=d,c=zo(u.shape,[r]);return[L(U(u,c),ce(me(h,"float32"),Mn(p))),L(U(u,c),ce(Mn(p),me(h,"float32")))]}}})(e,t)}function vD(e,t,r,n=0,a=3){let s=$(e,"onehotLabels","softmaxCrossEntropy"),i=$(t,"logits","softmaxCrossEntropy"),o=null;if(r!=null&&(o=$(r,"weights","softmaxCrossEntropy")),Ur(s.shape,i.shape,"Error in softmaxCrossEntropy: "),n>0){let u=Se(n),d=Se(1),h=Se(s.shape[1]);s=le(L(s,ce(d,u)),pe(u,h))}let l=bD(s,i);return rs(l,o,a)}var wD=W({softmaxCrossEntropy_:vD});function kD(e,t,r,n){let a=$(e,"indices","sparseFillEmptyRows","int32"),s=$(t,"values","sparseFillEmptyRows"),i=$(r,"denseShape","sparseFillEmptyRows","int32"),o=$(n,"defaultValue","sparseFillEmptyRows",s.dtype);if(a.rank!==2)throw new Error(`Indices should be Tensor2D but received shape
|
|
${a.shape}`);if(s.rank!==1)throw new Error(`Values should be Tensor1D but received shape ${s.shape}`);if(i.rank!==1)throw new Error(`Dense shape should be Tensor1D but received shape ${i.shape}`);if(o.rank!==0)throw new Error(`Default value should be a scalar but received shape ${o.shape}`);let l={indices:a,values:s,denseShape:i,defaultValue:o},u=B.runKernel(lh,l);return{outputIndices:u[0],outputValues:u[1],emptyRowIndicator:u[2],reverseIndexMap:u[3]}}var ID=W({sparseFillEmptyRows_:kD});function SD(e,t,r){let n=$(e,"inputIndices","sparseReshape","int32"),a=$(t,"inputShape","sparseReshape","int32"),s=$(r,"newShape","sparseReshape","int32");if(n.rank!==2)throw new Error(`Input indices should be Tensor2D but received shape
|
|
${n.shape}`);if(a.rank!==1)throw new Error(`Input shape should be Tensor1D but received shape ${a.shape}`);if(s.rank!==1)throw new Error(`New shape should be Tensor1D but received shape ${s.shape}`);let i={inputIndices:n,inputShape:a,newShape:s},o=B.runKernel(ud,i);return{outputIndices:o[0],outputShape:o[1]}}var CD=W({sparseReshape_:SD});function TD(e,t,r){let n=$(e,"data","sparseSegmentMean"),a=$(t,"indices","sparseSegmentMean","int32"),s=$(r,"segmentIds","sparseSegmentMean","int32");if(n.rank<1)throw new Error("Data should be at least 1 dimensional but received scalar");if(a.rank!==1)throw new Error(`Indices should be Tensor1D but received shape
|
|
${a.shape}`);if(s.rank!==1)throw new Error(`Segment ids should be Tensor1D but received shape
|
|
${s.shape}`);let i={data:n,indices:a,segmentIds:s};return B.runKernel(uh,i)}var ND=W({sparseSegmentMean_:TD});function ED(e,t,r){let n=$(e,"data","sparseSegmentSum"),a=$(t,"indices","sparseSegmentSum","int32"),s=$(r,"segmentIds","sparseSegmentSum","int32");if(n.rank<1)throw new Error("Data should be at least 1 dimensional but received scalar");if(a.rank!==1)throw new Error(`Indices should be Tensor1D but received shape
|
|
${a.shape}`);if(s.rank!==1)throw new Error(`Segment ids should be Tensor1D but received shape
|
|
${s.shape}`);let i={data:n,indices:a,segmentIds:s};return B.runKernel(dh,i)}var RD=W({sparseSegmentSum_:ED});function $D(e,t,r,n,a,s,i,o){let l=$(e,"data","stringNGrams","string");if(l.dtype!=="string")throw new Error("Data must be of datatype string");if(l.shape.length!==1)throw new Error(`Data must be a vector, saw: ${l.shape}`);let u=$(t,"dataSplits","stringNGrams");if(u.dtype!=="int32")throw new Error("Data splits must be of datatype int32");let d={separator:r,nGramWidths:n,leftPad:a,rightPad:s,padWidth:i,preserveShortSequences:o},h={data:l,dataSplits:u},p=B.runKernel(hh,h,d);return{nGrams:p[0],nGramsSplits:p[1]}}var MD=W({stringNGrams_:$D});function FD(e,t,r=!0){let n=$(e,"input","stringSplit","string"),a=$(t,"delimiter","stringSplit","string");if(n.rank!==1)throw new Error(`Input should be Tensor1D but received shape ${n.shape}`);if(a.rank!==0)throw new Error(`Delimiter should be a scalar but received shape ${a.shape}`);let s={skipEmpty:r},i={input:n,delimiter:a},o=B.runKernel(wf,i,s);return{indices:o[0],values:o[1],shape:o[2]}}var PD=W({stringSplit_:FD});function _D(e,t){let r=$(e,"input","stringToHashBucketFast","string"),n={numBuckets:t};if(t<=0)throw new Error("Number of buckets must be at least 1");let a={input:r};return B.runKernel(kf,a,n)}var zD=W({stringToHashBucketFast_:_D}),OD={fft:Bf,ifft:Gp,rfft:Wf,irfft:hA},DD={hammingWindow:hO,hannWindow:Dw,frame:Lw,stft:gO},Ie={flipLeftRight:bO,grayscaleToRGB:wO,resizeNearestNeighbor:HO,resizeBilinear:GO,rotateWithOffset:IO,cropAndResize:AO,nonMaxSuppression:CO,nonMaxSuppressionAsync:PO,nonMaxSuppressionWithScore:zO,nonMaxSuppressionWithScoreAsync:DO,nonMaxSuppressionPadded:BO,nonMaxSuppressionPaddedAsync:VO,threshold:KO,transform:YO},Gw={bandPart:QO,gramSchmidt:tD,qr:nD},LD={absoluteDifference:iD,computeWeightedLoss:rs,cosineDistance:lD,hingeLoss:dD,huberLoss:hD,logLoss:fD,meanSquaredError:gD,sigmoidCrossEntropy:xD,softmaxCrossEntropy:wD},kp={sparseFillEmptyRows:ID,sparseReshape:CD,sparseSegmentMean:ND,sparseSegmentSum:RD},i0={stringNGrams:MD,stringSplit:PD,stringToHashBucketFast:zD},ns=class extends E7{minimize(e,t=!1,r){let{value:n,grads:a}=this.computeGradients(e,r);if(r!=null){let s=r.map(i=>({name:i.name,tensor:a[i.name]}));this.applyGradients(s)}else this.applyGradients(a);return te(a),t?n:(n.dispose(),null)}get iterations(){return this.iterations_==null&&(this.iterations_=0),this.iterations_}incrementIterations(){this.iterations_=this.iterations+1}computeGradients(e,t){return dw(e,t)}dispose(){this.iterations_!=null&&te(this.iterations_)}async saveIterations(){return this.iterations_==null&&(this.iterations_=0),{name:"iter",tensor:Se(this.iterations_,"int32")}}async getWeights(){throw new Error("getWeights() is not implemented for this optimizer yet.")}async setWeights(e){throw new Error(`setWeights() is not implemented for this optimizer class ${this.getClassName()}`)}async extractIterations(e){return this.iterations_=(await e[0].tensor.data())[0],e.slice(1)}};Object.defineProperty(ns,Symbol.hasInstance,{value:e=>e.minimize!=null&&e.computeGradients!=null&&e.applyGradients!=null});var qf=class extends ns{constructor(e,t,r=null){super(),this.learningRate=e,this.rho=t,this.epsilon=r,this.accumulatedGrads=[],this.accumulatedUpdates=[],r==null&&(this.epsilon=B.backend.epsilon())}applyGradients(e){(Array.isArray(e)?e.map(t=>t.name):Object.keys(e)).forEach((t,r)=>{let n=B.registeredVariables[t],a=!1;this.accumulatedGrads[r]==null&&(this.accumulatedGrads[r]={originalName:`${t}/accum_grad`,variable:X(()=>at(n).variable(a))}),this.accumulatedUpdates[r]==null&&(this.accumulatedUpdates[r]={originalName:`${t}/accum_var`,variable:X(()=>at(n).variable(a))});let s=Array.isArray(e)?e[r].tensor:e[t];if(s==null)return;let i=this.accumulatedGrads[r].variable,o=this.accumulatedUpdates[r].variable;X(()=>{let l=le(L(i,this.rho),L(At(s),1-this.rho)),u=L(pe(Er(le(o,this.epsilon)),Er(le(i,this.epsilon))),s),d=le(L(o,this.rho),L(At(u),1-this.rho));i.assign(l),o.assign(d);let h=le(L(u,-this.learningRate),n);n.assign(h)})}),this.incrementIterations()}dispose(){this.accumulatedUpdates!=null&&(te(this.accumulatedGrads.map(e=>e.variable)),te(this.accumulatedUpdates.map(e=>e.variable)))}async getWeights(){let e=[...this.accumulatedGrads,...this.accumulatedUpdates];return[await this.saveIterations()].concat(e.map(t=>({name:t.originalName,tensor:t.variable})))}async setWeights(e){e=await this.extractIterations(e);let t=e.length/2,r=!1;this.accumulatedGrads=e.slice(0,t).map(n=>({originalName:n.name,variable:n.tensor.variable(r)})),this.accumulatedUpdates=e.slice(t,t*2).map(n=>({originalName:n.name,variable:n.tensor.variable(r)}))}getConfig(){return{learningRate:this.learningRate,rho:this.rho,epsilon:this.epsilon}}static fromConfig(e,t){return new e(t.learningRate,t.rho,t.epsilon)}};qf.className="Adadelta";Hi(qf);var Xf=class extends ns{constructor(e,t=.1){super(),this.learningRate=e,this.initialAccumulatorValue=t,this.accumulatedGrads=[]}applyGradients(e){(Array.isArray(e)?e.map(t=>t.name):Object.keys(e)).forEach((t,r)=>{let n=B.registeredVariables[t];this.accumulatedGrads[r]==null&&(this.accumulatedGrads[r]={originalName:`${t}/accumulator`,variable:X(()=>cd(n.shape,this.initialAccumulatorValue).variable(!1))});let a=Array.isArray(e)?e[r].tensor:e[t];if(a==null)return;let s=this.accumulatedGrads[r].variable;X(()=>{let i=le(s,At(a));s.assign(i);let o=le(L(pe(a,Er(le(i,B.backend.epsilon()))),-this.learningRate),n);n.assign(o)})}),this.incrementIterations()}dispose(){this.accumulatedGrads!=null&&te(this.accumulatedGrads.map(e=>e.variable))}async getWeights(){return[await this.saveIterations()].concat(this.accumulatedGrads.map(e=>({name:e.originalName,tensor:e.variable})))}async setWeights(e){e=await this.extractIterations(e);let t=!1;this.accumulatedGrads=e.map(r=>({originalName:r.name,variable:r.tensor.variable(t)}))}getConfig(){return{learningRate:this.learningRate,initialAccumulatorValue:this.initialAccumulatorValue}}static fromConfig(e,t){return new e(t.learningRate,t.initialAccumulatorValue)}};Xf.className="Adagrad";Hi(Xf);var Kf=class extends ns{constructor(e,t,r,n=null){super(),this.learningRate=e,this.beta1=t,this.beta2=r,this.epsilon=n,this.accumulatedFirstMoment=[],this.accumulatedSecondMoment=[],X(()=>{this.accBeta1=Se(t).variable(),this.accBeta2=Se(r).variable()}),n==null&&(this.epsilon=B.backend.epsilon())}applyGradients(e){let t=Array.isArray(e)?e.map(r=>r.name):Object.keys(e);X(()=>{let r=ce(1,this.accBeta1),n=ce(1,this.accBeta2);t.forEach((a,s)=>{let i=B.registeredVariables[a],o=!1;this.accumulatedFirstMoment[s]==null&&(this.accumulatedFirstMoment[s]={originalName:`${a}/m`,variable:X(()=>at(i).variable(o))}),this.accumulatedSecondMoment[s]==null&&(this.accumulatedSecondMoment[s]={originalName:`${a}/v`,variable:X(()=>at(i).variable(o))});let l=Array.isArray(e)?e[s].tensor:e[a];if(l==null)return;let u=this.accumulatedFirstMoment[s].variable,d=this.accumulatedSecondMoment[s].variable,h=le(L(u,this.beta1),L(l,1-this.beta1)),p=le(L(d,this.beta2),L(At(l),1-this.beta2)),c=pe(h,r),f=pe(p,n);u.assign(h),d.assign(p);let m=le(L(pe(c,le(Er(f),this.epsilon)),-this.learningRate),i);i.assign(m)}),this.accBeta1.assign(L(this.accBeta1,this.beta1)),this.accBeta2.assign(L(this.accBeta2,this.beta2))}),this.incrementIterations()}dispose(){this.accBeta1.dispose(),this.accBeta2.dispose(),this.accumulatedFirstMoment!=null&&te(this.accumulatedFirstMoment.map(e=>e.variable)),this.accumulatedSecondMoment!=null&&te(this.accumulatedSecondMoment.map(e=>e.variable))}async getWeights(){let e=[...this.accumulatedFirstMoment,...this.accumulatedSecondMoment];return[await this.saveIterations()].concat(e.map(t=>({name:t.originalName,tensor:t.variable})))}async setWeights(e){e=await this.extractIterations(e),X(()=>{this.accBeta1.assign(Ws(this.beta1,this.iterations_+1)),this.accBeta2.assign(Ws(this.beta2,this.iterations_+1))});let t=e.length/2,r=!1;this.accumulatedFirstMoment=e.slice(0,t).map(n=>({originalName:n.name,variable:n.tensor.variable(r)})),this.accumulatedSecondMoment=e.slice(t,t*2).map(n=>({originalName:n.name,variable:n.tensor.variable(r)}))}getConfig(){return{learningRate:this.learningRate,beta1:this.beta1,beta2:this.beta2,epsilon:this.epsilon}}static fromConfig(e,t){return new e(t.learningRate,t.beta1,t.beta2,t.epsilon)}};Kf.className="Adam";Hi(Kf);var Zf=class extends ns{constructor(e,t,r,n=null,a=0){super(),this.learningRate=e,this.beta1=t,this.beta2=r,this.epsilon=n,this.decay=a,this.accumulatedFirstMoment=[],this.accumulatedWeightedInfNorm=[],X(()=>{this.iteration=Se(0).variable(),this.accBeta1=Se(t).variable()}),n==null&&(this.epsilon=B.backend.epsilon())}applyGradients(e){let t=Array.isArray(e)?e.map(r=>r.name):Object.keys(e);X(()=>{let r=ce(1,this.accBeta1),n=pe(-this.learningRate,le(L(this.iteration,this.decay),1));t.forEach((a,s)=>{let i=B.registeredVariables[a],o=!1;this.accumulatedFirstMoment[s]==null&&(this.accumulatedFirstMoment[s]={originalName:`${a}/m`,variable:at(i).variable(o)}),this.accumulatedWeightedInfNorm[s]==null&&(this.accumulatedWeightedInfNorm[s]={originalName:`${a}/v`,variable:at(i).variable(o)});let l=Array.isArray(e)?e[s].tensor:e[a];if(l==null)return;let u=this.accumulatedFirstMoment[s].variable,d=this.accumulatedWeightedInfNorm[s].variable,h=le(L(u,this.beta1),L(l,1-this.beta1)),p=L(d,this.beta2),c=ar(l),f=ts(p,c);u.assign(h),d.assign(f);let m=le(L(pe(n,r),pe(h,le(f,this.epsilon))),i);i.assign(m)}),this.iteration.assign(le(this.iteration,1)),this.accBeta1.assign(L(this.accBeta1,this.beta1))}),this.incrementIterations()}dispose(){this.accBeta1.dispose(),this.iteration.dispose(),this.accumulatedFirstMoment!=null&&te(this.accumulatedFirstMoment.map(e=>e.variable)),this.accumulatedWeightedInfNorm!=null&&te(this.accumulatedWeightedInfNorm.map(e=>e.variable))}async getWeights(){throw new Error("getWeights() is not implemented for Adamax yet.")}async setWeights(e){throw new Error("setWeights() is not implemented for Adamax yet.")}getConfig(){return{learningRate:this.learningRate,beta1:this.beta1,beta2:this.beta2,epsilon:this.epsilon,decay:this.decay}}static fromConfig(e,t){return new e(t.learningRate,t.beta1,t.beta2,t.epsilon,t.decay)}};Zf.className="Adamax";Hi(Zf);var Ch=class extends ns{constructor(e){super(),this.learningRate=e,this.setLearningRate(e)}applyGradients(e){(Array.isArray(e)?e.map(t=>t.name):Object.keys(e)).forEach((t,r)=>{let n=Array.isArray(e)?e[r].tensor:e[t];if(n==null)return;let a=B.registeredVariables[t];X(()=>{let s=le(L(this.c,n),a);a.assign(s)})}),this.incrementIterations()}setLearningRate(e){this.learningRate=e,this.c!=null&&this.c.dispose(),this.c=gr(Se(-e))}dispose(){this.c.dispose()}async getWeights(){return[await this.saveIterations()]}async setWeights(e){if(e=await this.extractIterations(e),e.length!==0)throw new Error("SGD optimizer does not have settable weights.")}getConfig(){return{learningRate:this.learningRate}}static fromConfig(e,t){return new e(t.learningRate)}};Ch.className="SGD";Hi(Ch);var Yf=class extends Ch{constructor(e,t,r=!1){super(e),this.learningRate=e,this.momentum=t,this.useNesterov=r,this.accumulations=[],this.m=Se(this.momentum)}applyGradients(e){(Array.isArray(e)?e.map(t=>t.name):Object.keys(e)).forEach((t,r)=>{let n=B.registeredVariables[t];this.accumulations[r]==null&&(this.accumulations[r]={originalName:`${t}/momentum`,variable:X(()=>at(n).variable(!1))});let a=this.accumulations[r].variable,s=Array.isArray(e)?e[r].tensor:e[t];s!=null&&X(()=>{let i,o=le(L(this.m,a),s);this.useNesterov?i=le(L(this.c,le(s,L(o,this.m))),n):i=le(L(this.c,o),n),a.assign(o),n.assign(i)})}),this.incrementIterations()}dispose(){this.m.dispose(),this.accumulations!=null&&te(this.accumulations.map(e=>e.variable))}setMomentum(e){this.momentum=e}async getWeights(){return[await this.saveIterations()].concat(this.accumulations.map(e=>({name:e.originalName,tensor:e.variable})))}async setWeights(e){e=await this.extractIterations(e);let t=!1;this.accumulations=e.map(r=>({originalName:r.name,variable:r.tensor.variable(t)}))}getConfig(){return{learningRate:this.learningRate,momentum:this.momentum,useNesterov:this.useNesterov}}static fromConfig(e,t){return new e(t.learningRate,t.momentum,t.useNesterov)}};Yf.className="Momentum";Hi(Yf);var Jf=class extends ns{constructor(e,t=.9,r=0,n=null,a=!1){if(super(),this.learningRate=e,this.decay=t,this.momentum=r,this.epsilon=n,this.accumulatedMeanSquares=[],this.accumulatedMoments=[],this.accumulatedMeanGrads=[],this.centered=a,n==null&&(this.epsilon=B.backend.epsilon()),e==null)throw new Error("learningRate for RMSPropOptimizer must be defined.")}applyGradients(e){(Array.isArray(e)?e.map(t=>t.name):Object.keys(e)).forEach((t,r)=>{let n=B.registeredVariables[t],a=!1;this.accumulatedMeanSquares[r]==null&&(this.accumulatedMeanSquares[r]={originalName:`${t}/rms`,variable:X(()=>at(n).variable(a))}),this.accumulatedMoments[r]==null&&(this.accumulatedMoments[r]={originalName:`${t}/momentum`,variable:X(()=>at(n).variable(a))}),this.accumulatedMeanGrads[r]==null&&this.centered&&(this.accumulatedMeanGrads[r]={originalName:`${t}/mg`,variable:X(()=>at(n).variable(a))});let s=Array.isArray(e)?e[r].tensor:e[t];if(s==null)return;let i=this.accumulatedMeanSquares[r].variable,o=this.accumulatedMoments[r].variable;X(()=>{let l=le(L(i,this.decay),L(At(s),1-this.decay));if(this.centered){let u=this.accumulatedMeanGrads[r].variable,d=le(L(u,this.decay),L(s,1-this.decay)),h=pe(L(s,this.learningRate),Er(ce(l,le(At(d),this.epsilon)))),p=le(L(o,this.momentum),h);i.assign(l),u.assign(d),o.assign(p);let c=ce(n,p);n.assign(c)}else{let u=le(L(i,this.decay),L(At(s),1-this.decay)),d=le(L(o,this.momentum),pe(L(s,this.learningRate),Er(le(u,this.epsilon))));i.assign(u),o.assign(d);let h=ce(n,d);n.assign(h)}})}),this.incrementIterations()}dispose(){this.accumulatedMeanSquares!=null&&te(this.accumulatedMeanSquares.map(e=>e.variable)),this.accumulatedMeanGrads!=null&&this.centered&&te(this.accumulatedMeanGrads.map(e=>e.variable)),this.accumulatedMoments!=null&&te(this.accumulatedMoments.map(e=>e.variable))}async getWeights(){let e=[...this.accumulatedMeanSquares,...this.accumulatedMoments];return this.centered&&e.push(...this.accumulatedMeanGrads),[await this.saveIterations()].concat(e.map(t=>({name:t.originalName,tensor:t.variable})))}async setWeights(e){e=await this.extractIterations(e);let t=this.centered?e.length/3:e.length/2,r=!1;this.accumulatedMeanSquares=e.slice(0,t).map(n=>({originalName:n.name,variable:n.tensor.variable(r)})),this.accumulatedMoments=e.slice(t,t*2).map(n=>({originalName:n.name,variable:n.tensor.variable(r)})),this.centered&&(this.accumulatedMeanGrads=e.slice(t*2,t*3).map(n=>({originalName:n.name,variable:n.tensor.variable(r)})))}getConfig(){return{learningRate:this.learningRate,decay:this.decay,momentum:this.momentum,epsilon:this.epsilon,centered:this.centered}}static fromConfig(e,t){return new e(t.learningRate,t.decay,t.momentum,t.epsilon,t.centered)}};Jf.className="RMSProp";Hi(Jf);var Is=class{static sgd(e){return new Ch(e)}static momentum(e,t,r=!1){return new Yf(e,t,r)}static rmsprop(e,t=.9,r=0,n=null,a=!1){return new Jf(e,t,r,n,a)}static adam(e=.001,t=.9,r=.999,n=null){return new Kf(e,t,r,n)}static adadelta(e=.001,t=.95,r=null){return new qf(e,t,r)}static adamax(e=.002,t=.9,r=.999,n=null,a=0){return new Zf(e,t,r,n,a)}static adagrad(e,t=.1){return new Xf(e,t)}},bo={sgd:Is.sgd,momentum:Is.momentum,adadelta:Is.adadelta,adagrad:Is.adagrad,rmsprop:Is.rmsprop,adamax:Is.adamax,adam:Is.adam},BD=(()=>typeof requestAnimationFrame!="undefined"?requestAnimationFrame:typeof setImmediate!="undefined"?setImmediate:e=>e())();function bA(){return new Promise(e=>BD(()=>e()))}var C={};Le(C,{ERF_A1:()=>YD,ERF_A2:()=>JD,ERF_A3:()=>QD,ERF_A4:()=>eL,ERF_A5:()=>tL,ERF_P:()=>ZD,PARALLELIZE_THRESHOLD:()=>vA,SELU_SCALE:()=>Hw,SELU_SCALEALPHA:()=>jw,applyActivation:()=>jf,assertAndGetBroadcastShape:()=>bt,assertAxesAreInnerMostDims:()=>YP,assertParamsConsistent:()=>WD,assignToTypedArray:()=>oL,axesAreInnerMostDims:()=>Yy,calculateShapes:()=>x7,checkEinsumDimSizes:()=>cL,checkPadOnDimRoundingMode:()=>jr,combineLocations:()=>pw,complexWithEvenIndex:()=>aL,complexWithOddIndex:()=>sL,computeConv2DInfo:()=>bh,computeConv3DInfo:()=>V7,computeDefaultPad:()=>Ly,computeDilation2DInfo:()=>gF,computeOptimalWindowSize:()=>UD,computeOutAndReduceShapes:()=>hw,computeOutShape:()=>VD,computePool2DInfo:()=>W7,computePool3DInfo:()=>yF,convertConv2DDataFormat:()=>U7,decodeEinsumEquation:()=>pL,eitherStridesOrDilationsAreOne:()=>za,expandShapeToKeepDim:()=>zo,exponent:()=>uL,exponents:()=>lL,fromStringArrayToUint8:()=>PL,fromUint8ToStringArray:()=>FL,getAxesPermutation:()=>cw,getBroadcastDims:()=>m7,getComplexWithIndex:()=>iL,getEinsumComputePath:()=>fL,getEinsumPermutation:()=>hL,getFusedBiasGradient:()=>Gf,getFusedDyActivation:()=>Uf,getImageCenter:()=>GD,getInnerMostAxes:()=>JP,getPermuted:()=>HD,getReductionAxes:()=>Jt,getReshaped:()=>jD,getReshapedPermuted:()=>qD,getSliceBeginCoords:()=>XD,getSliceSize:()=>KD,getSparseFillEmptyRowsIndicesDenseShapeMismatch:()=>AL,getSparseFillEmptyRowsNegativeIndexErrorMessage:()=>xL,getSparseFillEmptyRowsOutOfRangeIndexErrorMessage:()=>bL,getSparseReshapeEmptyTensorZeroOutputDimErrorMessage:()=>kL,getSparseReshapeInputOutputMismatchErrorMessage:()=>SL,getSparseReshapeInputOutputMultipleErrorMessage:()=>IL,getSparseReshapeMultipleNegativeOneOutputDimErrorMessage:()=>vL,getSparseReshapeNegativeOutputDimErrorMessage:()=>wL,getSparseSegmentReductionIndicesOutOfRangeErrorMessage:()=>EL,getSparseSegmentReductionNegativeSegmentIdsErrorMessage:()=>CL,getSparseSegmentReductionNonIncreasingSegmentIdsErrorMessage:()=>TL,getSparseSegmentReductionSegmentIdOutOfRangeErrorMessage:()=>NL,getUndoAxesPermutation:()=>Jy,isIdentityPermutation:()=>mL,log:()=>GR,mergeRealAndImagArrays:()=>rL,prepareAndValidate:()=>A7,prepareSplitSize:()=>yL,segment_util:()=>qw,shouldFuse:()=>Hf,slice_util:()=>Ot,splitRealAndImagArrays:()=>nL,tupleValuesAreOne:()=>Ds,upcastType:()=>Nr,validateInput:()=>Ry,validateUpdateShape:()=>Ey,warn:()=>Cs});function WD(e,t){let r=e[0].length;e.forEach((a,s)=>{P(a.length===r,()=>`Error in concat${r}D: rank of tensors[${s}] must be the same as the rank of the rest (${r})`)}),P(t>=0&&t<r,()=>`Error in concat${r}D: axis must be between 0 and ${r-1}.`);let n=e[0];e.forEach((a,s)=>{for(let i=0;i<r;i++)P(i===t||a[i]===n[i],()=>`Error in concat${r}D: Shape of tensors[${s}] (${a}) does not match the shape of the rest (${n}) along the non-concatenated axis ${s}.`)})}function VD(e,t){let r=e[0].slice();for(let n=1;n<e.length;n++)r[t]+=e[n][t];return r}var vA=30;function UD(e){return e<=vA?e:A0(e,Math.floor(Math.sqrt(e)))}function GD(e,t,r){let n=r*(typeof e=="number"?e:e[0]),a=t*(typeof e=="number"?e:e[1]);return[n,a]}function jD(e,t,r,n=!0){let a=[];if(n)a=a.concat(t.slice(0)),a.push(e[0]/r),a=a.concat(e.slice(1));else{a=a.concat(e[0]);let s=t.length;for(let i=0;i<s;++i)a=a.concat([e[i+1]/t[i],t[i]]);a=a.concat(e.slice(s+1))}return a}function HD(e,t,r=!0){let n=[];if(r){n.push(t);for(let a=t+1;a<e;++a)a<=2*t?(n.push(a),n.push(a-(t+1))):n.push(a)}else{let a=[],s=[];for(let i=1;i<e;++i)i>=t*2+1||i%2===1?s.push(i):a.push(i);n.push(...a),n.push(0),n.push(...s)}return n}function qD(e,t,r,n=!0){let a=[];n?a.push(e[0]/r):a.push(e[0]*r);for(let s=1;s<e.length;++s)s<=t.length?n?a.push(t[s-1]*e[s]):a.push(e[s]/t[s-1]):a.push(e[s]);return a}function XD(e,t){let r=[0];for(let n=0;n<t;++n)r.push(e[n][0]);return r}function KD(e,t,r){let n=e.slice(0,1);for(let a=0;a<r;++a)n.push(e[a+1]-t[a][0]-t[a][1]);return n}var jw=1.7580993408473768,Hw=1.0507009873554805,ZD=.3275911,YD=.254829592,JD=-.284496736,QD=1.421413741,eL=-1.453152027,tL=1.061405429;function rL(e,t){if(e.length!==t.length)throw new Error(`Cannot merge real and imag arrays of different lengths. real:${e.length}, imag: ${t.length}.`);let r=new Float32Array(e.length*2);for(let n=0;n<r.length;n+=2)r[n]=e[n/2],r[n+1]=t[n/2];return r}function nL(e){let t=new Float32Array(e.length/2),r=new Float32Array(e.length/2);for(let n=0;n<e.length;n+=2)t[n/2]=e[n],r[n/2]=e[n+1];return{real:t,imag:r}}function aL(e){let t=Math.ceil(e.length/4),r=new Float32Array(t),n=new Float32Array(t);for(let a=0;a<e.length;a+=4)r[Math.floor(a/4)]=e[a],n[Math.floor(a/4)]=e[a+1];return{real:r,imag:n}}function sL(e){let t=Math.floor(e.length/4),r=new Float32Array(t),n=new Float32Array(t);for(let a=2;a<e.length;a+=4)r[Math.floor(a/4)]=e[a],n[Math.floor(a/4)]=e[a+1];return{real:r,imag:n}}function iL(e,t){let r=e[t*2],n=e[t*2+1];return{real:r,imag:n}}function oL(e,t,r,n){e[n*2]=t,e[n*2+1]=r}function lL(e,t){let r=new Float32Array(e/2),n=new Float32Array(e/2);for(let a=0;a<Math.ceil(e/2);a++){let s=(t?2:-2)*Math.PI*(a/e);r[a]=Math.cos(s),n[a]=Math.sin(s)}return{real:r,imag:n}}function uL(e,t,r){let n=(r?2:-2)*Math.PI*(e/t),a=Math.cos(n),s=Math.sin(n);return{real:a,imag:s}}var sg="->",dL=/->/g,Z3=",",Y3="...";function pL(e,t){e=e.replace(/\s/g,"");let r=(e.length-e.replace(dL,"").length)/sg.length;if(r<1)throw new Error("Equations without an arrow are not supported.");if(r>1)throw new Error(`Equation must contain exactly one arrow ("${sg}").`);let[n,a]=e.split(sg);P(n.indexOf(Y3)===-1,()=>`The ellipsis notation ("${Y3}") is not supported yet.`);let s=n.split(Z3),i=s.length;if(t!==i)throw new Error(`Expected ${i} input tensors, received ${t}`);if(i>2)throw new Error("Support for more than 2 input tensors is not implemented yet.");let o=[];for(let p=0;p<a.length;++p){let c=a[p];if(!s.some(f=>f.indexOf(c)!==-1))throw new Error(`Output subscripts contain the label ${c} not present in the input subscripts.`);o.indexOf(c)===-1&&o.push(c)}for(let p=0;p<n.length;++p){let c=n[p];o.indexOf(c)===-1&&c!==Z3&&o.push(c)}let l=new Array(s.length);for(let p=0;p<i;++p){if(new Set(s[p].split("")).size!==s[p].length)throw new Error(`Found duplicate axes in input component ${s[p]}. Support for duplicate axes in input is not implemented yet.`);l[p]=[];for(let c=0;c<s[p].length;++c)l[p].push(o.indexOf(s[p][c]))}let u=o.length,d=a.length,h=[];for(let p=d;p<u;++p)h.push(p);return{allDims:o,summedDims:h,idDims:l}}function hL(e,t){let r=new Array(e);r.fill(-1);for(let a=0;a<t.length;++a)r[t[a]]=a;let n=[];for(let a=0;a<e;++a)r[a]===-1&&n.push(a);return r=r.filter(a=>a!==-1),{permutationIndices:r,expandDims:n}}function cL(e,t,r){let n=new Array(e);for(let a=0;a<r.length;++a){let s=r[a].shape;for(let i=0;i<t[a].length;++i)n[t[a][i]]===void 0?n[t[a][i]]=s[i]:P(n[t[a][i]]===s[i],()=>`Expected dimension ${n[t[a][i]]} at axis ${i} of input shaped ${JSON.stringify(s)}, but got dimension ${s[i]}`)}}function fL(e,t){let r=e,n=[],a=0;e.length===0&&r.push(-1),a=e.length+1;for(let i=0;i<a;++i)n.push([]);let s=[];for(let i=0;i<r.length;++i){let o=r[i],l=gL(t,o);for(let u of l)s.indexOf(u)===-1&&(n[i].push(u),s.push(u))}return{path:r,steps:n}}function mL(e){return e.every((t,r)=>t===r)}function gL(e,t){let r=[];for(let n=0;n<e.length;++n)(e[n].length===0||e[n].indexOf(t)!==-1||t===-1)&&r.push(n);return r}function yL(e,t,r=0){let n=[];if(typeof t=="number")P(e.shape[r]%t===0,()=>"Number of splits must evenly divide the axis."),n=new Array(t).fill(e.shape[r]/t);else{let a=t.reduce((i,o)=>(o===-1&&(i+=1),i),0);P(a<=1,()=>"There should be only one negative value in split array.");let s=t.indexOf(-1);if(s!==-1){let i=t.reduce((o,l)=>l>0?o+l:o);t[s]=e.shape[r]-i}P(e.shape[r]===t.reduce((i,o)=>i+o),()=>"The sum of sizes must match the size of the axis dimension."),n=t}return n}function AL(e){return`Received SparseTensor with denseShape[0] = 0 but
|
|
indices.shape[0] = ${e}`}function xL(e,t){return`indices(${e}, 0) is invalid: ${t} < 0`}function bL(e,t,r){return`indices(${e}, 0) is invalid: ${t} >= ${r}`}function vL(e,t){return`only one output dimension may be -1, not both ${e} and ${t}`}function wL(e,t){return`size ${e} must be non-negative, not ${t}`}function kL(){return"reshape cannot infer the missing input size for an empty tensor unless all specified input sizes are non-zero"}function IL(e,t){let r=It(e),n=It(t);return`Input to reshape is a SparseTensor with ${r}
|
|
dense values, but the requested shape requires a multiple of ${n}. inputShape=${e} outputShape= ${t}`}function SL(e,t){let r=It(e),n=It(t);return`Input to reshape is a tensor with ${r} dense values, but the requested shape has ${n}. inputShape=${e} outputShape=${t}`}function CL(){return"segment ids must be >= 0"}function TL(){return"segment ids are not increasing"}function NL(e,t){return`Segment id ${e} out of range [0, ${t}), possibly because segmentIds input is not sorted.`}function EL(e,t,r){return`Bad: indices[${e}] == ${t} out of range [0, ${r})`}var qw={};Le(qw,{collectGatherOpShapeInfo:()=>ML,computeOutShape:()=>$L,segOpComputeOptimalWindowSize:()=>RL});function RL(e,t){let r=!1,n;for(e<=vA?(n=e,r=!0):n=A0(e,Math.floor(Math.sqrt(e)));!r;)n>t||n===e?r=!0:n=A0(e,n+1);return n}function $L(e,t,r){let n=[],a=e.length;for(let s=0;s<a;s++)s!==t?n.push(e[s]):n.push(r);return n}function ML(e,t,r,n){let a=t.shape.length,s=e.shape.length;if(n!==0&&(n<-a||n>a))throw new Error(`Expect batchDims in the range of [-${a}, ${a}], but got ${n}`);if(n<0&&(n+=a),n>s)throw new Error(`batchDims (${n}) must be less than rank(x) (
|
|
${s}).`);if(r<n)throw new Error(`batchDims (${n}) must be less than or equal to axis (${r}).`);for(let h=0;h<n;++h)if(e.shape[h]!==t.shape[h])throw new Error(`x.shape[${h}]: ${e.shape[h]} should be equal to indices.shape[${h}]: ${t.shape[h]}.`);let i=e.shape[r],o=[],l=1,u=1,d=1;for(let h=0;h<n;++h)o.push(e.shape[h]),l*=e.shape[h];for(let h=n;h<r;h++)o.push(e.shape[h]),u*=e.shape[h];for(let h=n;h<a;h++)o.push(t.shape[h]);for(let h=r+1;h<s;h++)o.push(e.shape[h]),d*=e.shape[h];return{batchSize:l,sliceSize:d,outerSize:u,dimSize:i,outputShape:o}}function FL(e){try{return e.map(t=>w0(t))}catch(t){throw new Error(`Failed to decode encoded string bytes into utf-8, error: ${t}`)}}function PL(e){return e.map(t=>mh(t))}var Kn={};Le(Kn,{nonMaxSuppressionV3Impl:()=>Bw,nonMaxSuppressionV4Impl:()=>Ww,nonMaxSuppressionV5Impl:()=>Vw,whereImpl:()=>Rw});var Ga=class extends Error{constructor(e){super(e),Object.setPrototypeOf(this,Ga.prototype)}},da=class extends Error{constructor(e){super(e),Object.setPrototypeOf(this,da.prototype)}},H=class extends Error{constructor(e){super(e),Object.setPrototypeOf(this,H.prototype)}},Ve=class extends Error{constructor(e){super(e),Object.setPrototypeOf(this,Ve.prototype)}},Xw=class extends Error{constructor(e){super(e),Object.setPrototypeOf(this,Xw.prototype)}},Kw=class{constructor(e){this.maxEntries=e||100,this.cache=new Map}get(e){let t;return this.cache.has(e)&&(t=this.cache.get(e),this.cache.delete(e),this.cache.set(e,t)),t}put(e,t){if(this.cache.has(e))this.cache.delete(e);else if(this.cache.size>=this.maxEntries){let r=this.cache.keys().next().value;this.cache.delete(r)}this.cache.set(e,t)}getMaxEntries(){return this.maxEntries}setMaxEntries(e){if(e<0)throw new Error(`The maxEntries of LRU caches must be at least 0, but got ${e}.`);if(this.maxEntries>e)for(let t=0;t<this.maxEntries-e;t++){let r=this.cache.keys().next().value;this.cache.delete(r)}this.maxEntries=e}};function Do(e,t){if(Array.isArray(e)){let r=[];for(let n=0;n<t;n++)r=r.concat(e);return r}else{let r=new Array(t);return r.fill(e),r}}function Na(e,t){if(!e)throw new Xw(t)}function J3(e,t){let r=0;for(let n of e)n===t&&r++;return r}function tn(e){return e.length===1?e[0]:e}function Ct(e){return Array.isArray(e)?e:[e]}function ja(e){let t=e.replace(/(.)([A-Z][a-z0-9]+)/g,"$1_$2").replace(/([a-z])([A-Z])/g,"$1_$2").toLowerCase();return t[0]!=="_"?t:"private"+t}function Io(e){return e.length<=1||e.indexOf("_")===-1?e:e.replace(/[_]+(\w|$)/g,(t,r)=>r.toUpperCase())}var Vn={};function wA(e){if(e==null)return null;let t={};return t.className=e.getClassName(),t.config=e.getConfig(),t}function Eg(e){if(!(e==null||typeof e!="object"))if(Array.isArray(e))e.forEach(t=>Eg(t));else{let t=Object.keys(e);for(let r of t){let n=e[r];n!=null&&typeof n=="object"&&(!Array.isArray(n)&&n.type==="ndarray"&&typeof n.value=="number"?e[r]=n.value:Eg(n))}}}function Th(e,t={},r={},n="object",a=!1){if(typeof e=="string"){let s=e,i;if(s in r)i=r[s];else if(s in Vn)i=Vn[s];else if(i=t[s],i==null)throw new H(`Unknown ${n}: ${e}. This may be due to one of the following reasons:
|
|
1. The ${n} is defined in Python, in which case it needs to be ported to TensorFlow.js or your JavaScript code.
|
|
2. The custom ${n} is defined in JavaScript, but is not registered properly with tf.serialization.registerClass().`);return i}else{let s=e;if(s.className==null||s.config==null)throw new H(`${n}: Improper config format: ${JSON.stringify(s)}.
|
|
'className' and 'config' must set.`);let i=s.className,o,l;if(i in r?[o,l]=r[i]:i in Vn?[o,l]=Vn.className:i in t&&([o,l]=t[i]),o==null)throw new H(`Unknown ${n}: ${i}. This may be due to one of the following reasons:
|
|
1. The ${n} is defined in Python, in which case it needs to be ported to TensorFlow.js or your JavaScript code.
|
|
2. The custom ${n} is defined in JavaScript, but is not registered properly with tf.serialization.registerClass().`);if(l!=null){let u={};for(let c of Object.keys(Vn))u[c]=Vn[c];for(let c of Object.keys(r))u[c]=r[c];let d=s.config;d.customObjects=u;let h={...Vn};for(let c of Object.keys(r))Vn[c]=r[c];Eg(s.config);let p=l(o,s.config,r,a);return Vn={...h},p}else{let u={...Vn};for(let h of Object.keys(r))Vn[h]=r[h];let d=new o(s.config);return Vn={...u},d}}}function _L(e,t){return e<t?-1:e>t?1:0}function Xc(e,t){return-1*_L(e,t)}function Rs(e){if(e==null)return e;let t=[];for(let r of e)t.indexOf(r)===-1&&t.push(r);return t}function zL(e){if(e==null)throw new H(`Invalid value in obj: ${JSON.stringify(e)}`);for(let t in e)if(e.hasOwnProperty(t))return!1;return!0}function _l(e,t,r){if(r!=null&&e.indexOf(r)<0)throw new H(`${r} is not a valid ${t}. Valid values are ${e} or null/undefined.`)}function kA(e,t,r=0,n=1/0){return Na(r>=0),Na(n>=r),Array.isArray(e)&&e.length>=r&&e.length<=n&&e.every(a=>typeof a===t)}function yr(e,t){Array.isArray(e)?(v.assert(e.length>0,()=>`${t} is unexpectedly an empty array.`),e.forEach((r,n)=>yr(r,`element ${n+1} of ${t}`))):v.assert(Number.isInteger(e)&&e>0,()=>`Expected ${t} to be a positive integer, but got ${Zw(e)}.`)}function Zw(e){return e===null?"null":Array.isArray(e)?"["+e.map(t=>Zw(t)).join(",")+"]":typeof e=="string"?`"${e}"`:`${e}`}function OL(e,t,r){let n=r!=null?r():v.now(),a;return(...s)=>{let i=r!=null?r():v.now();return i-n<t||(n=i,a=e(...s)),a}}function Yw(e){return e==="relu"?"relu":e==="linear"?"linear":e==="elu"?"elu":null}var DL=0;function Jw(){return DL++}var Kc={};function Qf(e=""){return e in Kc||(Kc[e]=0),Kc[e]+=1,e+Kc[e].toString()}var LL=["channelsFirst","channelsLast"],BL=["nearest","bilinear"],WL=["valid","same","causal"],VL=["max","avg"],UL=["sum","mul","concat","ave"],uu=new Map;function jt(e){_l(LL,"DataFormat",e)}function GL(e){_l(BL,"InterpolationFormat",e)}function Ln(e){_l(WL,"PaddingMode",e)}function Qw(e){_l(VL,"PoolMode",e)}var Fp=[],Q3="/";function Ro(e,t){Fp.push(e);try{let r=t();return Fp.pop(),r}catch(r){throw Fp.pop(),r}}function jL(){return Fp.length===0?"":Fp.join(Q3)+Q3}function e6(e){if(!r6(e))throw new Error("Not a valid tensor name: '"+e+"'");return jL()+e}function t6(e){if(!r6(e))throw new Error("Not a valid tensor name: '"+e+"'");uu.has(e)||uu.set(e,0);let t=uu.get(e);if(uu.set(e,uu.get(e)+1),t>0){let r=`${e}_${t}`;return uu.set(r,1),r}else return e}var HL=new RegExp(/^[A-Za-z0-9][-A-Za-z0-9\._\/]*$/);function r6(e){return!!e.match(HL)}function qL(e){return e===parseInt(e.toString(),10)}function $s(e,t,r){t==null&&(t=0),r==null&&(r=e.length);let n=1;for(let a=t;a<r;++a)n*=e[a];return n}function Eu(e){if(e.length===0)return Number.NaN;let t=Number.POSITIVE_INFINITY;for(let r=0;r<e.length;r++){let n=e[r];n<t&&(t=n)}return t}function Us(e){if(e.length===0)return Number.NaN;let t=Number.NEGATIVE_INFINITY;for(let r=0;r<e.length;r++){let n=e[r];n>t&&(t=n)}return t}function ya(e,t){if(t<e)throw new H(`end (${t}) < begin (${e}) is forbidden.`);let r=[];for(let n=e;n<t;++n)r.push(n);return r}var ig;function sr(){return ig==null&&(ig=Dn().epsilon()),ig}function Aa(){return"channelsLast"}function em(e,t){return me(e,t)}function Nh(e,t=-1){let r=e.shape.slice();return t<0&&(t=r.length+t+1),r.splice(t,0,1),U(e,r)}function XL(e,t){return X(()=>{if(e.shape.length!==2)throw new H(`repeat() expects a rank-2 tensor, but received a rank-${e.shape.length} tensor.`);let r=Nh(e,1);return Rg(r,[1,t,1])})}function KL(e){let t=[$s(e.shape)];return U(e,t)}function ZL(e){if(e.rank<=1)throw new H(`batchFlatten requires a minimum rank of 2. Got rank: ${e.rank}.`);let t=[e.shape[0],$s(e.shape,1)];return U(e,t)}function $o(e,t,r){return X(()=>{switch(e.rank){case 1:return Lf(e,t,r);case 2:return pA(e,[t,0],[r,e.shape[1]]);case 3:return Pl(e,[t,0,0],[r,e.shape[1],e.shape[2]]);case 4:return Oo(e,[t,0,0,0],[r,e.shape[1],e.shape[2],e.shape[3]]);case 5:return Pe(e,[t,0,0,0,0],[r,e.shape[1],e.shape[2],e.shape[3],e.shape[4]]);case 6:return Pe(e,[t,0,0,0,0,0],[r,e.shape[1],e.shape[2],e.shape[3],e.shape[4],e.shape[5]]);default:throw new H(`sliceAlongFirstAxis() received an unsupported tensor rank: ${e.rank}`)}})}function og(e,t,r){return X(()=>{switch(e.rank){case 1:return Lf(e,t,r);case 2:return pA(e,[0,t],[e.shape[0],r]);case 3:return Pl(e,[0,0,t],[e.shape[0],e.shape[1],r]);case 4:return Oo(e,[0,0,0,t],[e.shape[0],e.shape[1],e.shape[2],r]);default:throw new H(`sliceAlongLastAxis() received an unsupported tensor rank: ${e.rank}`)}})}function Zc(e,t,r,n){return X(()=>{switch(e.rank){case 1:return Lf(e,t,r);case 2:switch(n){case 1:return $o(e,t,r);case 2:return og(e,t,r);default:throw new H(`The axis is not within the rank of the tensor ${n}`)}case 3:switch(n){case 1:return $o(e,t,r);case 2:return Pl(e,[0,t,0],[e.shape[0],r,e.shape[2]]);case 3:return og(e,t,r);default:throw new H(`The axis is not within the rank of the tensor ${n}`)}case 4:switch(n){case 1:return $o(e,t,r);case 2:return Oo(e,[0,t,0,0],[e.shape[0],r,e.shape[2],e.shape[3]]);case 3:return Oo(e,[0,0,t,0],[e.shape[0],e.shape[1],r,e.shape[3]]);case 4:return og(e,t,r);default:throw new H(`The axis is not within the rank of the tensor ${n}`)}default:throw new H(`sliceAlongLastAxis() received an unsupported tensor rank: ${e.rank}`)}})}function IA(e,t=-1){let r;return t<0&&(r=e[0].rank,r!==0?t=r:t=0),t===e[0].rank&&(t=-1),St(e,t)}function e4(e,t){switch(e.rank){case 1:return K7([e,t]);case 2:return hd([e,t],0);case 3:return Z7([e,t],0);case 4:return Y7([e,t],0);default:throw new H(`concatAlongFirstAxis() received an unsupported tensor rank: ${e.rank}`)}}function Rg(e,t){if(Array.isArray(t)||(t=[t]),e.rank!==t.length)throw new H(`The length of input n (${t.length}) does not match the number of dimensions in input x (${e.rank})`);return Gn(e,t)}function tm(e,t=0,r=1,n,a){return xw(e,t,r,n,a)}function $a(e,t,r,n){if(e.rank<2||t.rank<2)throw new Ve(`dot requires both inputs to be rank >= 2 but got x shape = ${e.shape} and y shape = ${t.shape}`);if(t.rank>=3){let a=e.shape.slice(-1)[0],s=t.shape.slice(-2)[0];if(a!==s)throw new Ve(`If rank y >= 3, then the second last dim of y must equal the last dim of x but got x shape = ${e.shape} and y shape = ${t.shape}`)}if(e.rank===2&&t.rank===2)return Vs.matMul({a:e,b:t,transposeA:!1,transposeB:!1,bias:n?$g(e.rank,n,Aa()):null,activation:r});{let a=e.shape.slice(),s=a.pop();e=U(e,[-1,s]);let i=t.shape.slice(),o=i.pop(),l=i.pop(),u=[...i,o],d=Array.from({length:t.rank},(f,m)=>m===0?t.rank-2:m<=t.rank-2?m-1:m);t=U(et(t,d),[l,-1]);let h=[...a,...u],p=!1,c=!1;return U(Vs.matMul({a:e,b:t,transposeA:p,transposeB:c,bias:n?$g(e.rank,n,Aa()):null,activation:r}),h)}}function n6(e,t,r){return X(()=>(Array.isArray(t)?t=Tt(t,"int32"):t=me(t,"int32"),Cu(e,t,r)))}function Eh(e){return L(e,e)}function $g(e,t,r){let n=t.shape;if(t.rank!==1&&t.rank!==e)throw new H(`Unexpected bias dimensions: ${t.rank}; expected it to be 1 or ${e}`);if(e===5){if(r==="channelsFirst")return n.length===1?U(t,[1,n[0],1,1,1]):U(t,[1,n[3],n[0],n[1],n[2]]);if(r==="channelsLast")return n.length===1?U(t,[1,1,1,1,n[0]]):U(t,[1].concat(n))}else if(e===4){if(r==="channelsFirst")return n.length===1?U(t,[1,n[0],1,1]):U(t,[1,n[2],n[0],n[1]]);if(r==="channelsLast")return n.length===1?U(t,[1,1,1,n[0]]):U(t,[1].concat(n))}else if(e===3){if(r==="channelsFirst")return n.length===1?U(t,[1,n[0],1]):U(t,[1,n[1],n[0]]);if(r==="channelsLast")return n.length===1?U(t,[1,1,n[0]]):U(t,[1].concat(n))}else if(e<3)return t;throw new H(`Unsupported input rank by biasAdd: ${t.rank}`)}function va(e,t,r){return X(()=>(r==null&&(r=Aa()),jt(r),le(e,$g(e.rank,t,r))))}function YL(e,t=1){if(t!==1)throw new Ve(`Support for alpha values other than 1 (${t}) is not implemented yet.`);return wh(e)}function JL(e){return X(()=>pe(e,le(ar(e),1)))}function a6(e,t,r,n){return X(()=>Pw(e,t,r,n))}function QL(e){return X(()=>{let t=le(.5,L(.2,e));return cn(t,0,1)})}function Rh(e,t,r=!1){return r?e():t()}var eB=["fanIn","fanOut","fanAvg"],tB=["normal","uniform","truncatedNormal"];function rB(e){_l(eB,"FanMode",e)}function nB(e){_l(tB,"Distribution",e)}var Zn=class extends ue.Serializable{fromConfigUsesCustomObjects(){return!1}getConfig(){return{}}},SA=class extends Zn{apply(e,t){return _t(e,t)}};SA.className="Zeros";ue.registerClass(SA);var rm=class extends Zn{apply(e,t){return hn(e,t)}};rm.className="Ones";ue.registerClass(rm);var CA=class extends Zn{constructor(e){if(super(),typeof e!="object")throw new H(`Expected argument of type ConstantConfig but got ${e}`);if(e.value===void 0)throw new H(`config must have value set but got ${e}`);this.value=e.value}apply(e,t){return X(()=>L(Se(this.value),hn(e,t)))}getConfig(){return{value:this.value}}};CA.className="Constant";ue.registerClass(CA);var TA=class extends Zn{constructor(e){super(),this.DEFAULT_MINVAL=-.05,this.DEFAULT_MAXVAL=.05,this.minval=e.minval||this.DEFAULT_MINVAL,this.maxval=e.maxval||this.DEFAULT_MAXVAL,this.seed=e.seed}apply(e,t){return gd(e,this.minval,this.maxval,t)}getConfig(){return{minval:this.minval,maxval:this.maxval,seed:this.seed}}};TA.className="RandomUniform";ue.registerClass(TA);var NA=class extends Zn{constructor(e){super(),this.DEFAULT_MEAN=0,this.DEFAULT_STDDEV=.05,this.mean=e.mean||this.DEFAULT_MEAN,this.stddev=e.stddev||this.DEFAULT_STDDEV,this.seed=e.seed}apply(e,t){if(t=t||"float32",t!=="float32"&&t!=="int32")throw new Ve(`randomNormal does not support dType ${t}.`);return tm(e,this.mean,this.stddev,t,this.seed)}getConfig(){return{mean:this.mean,stddev:this.stddev,seed:this.seed}}};NA.className="RandomNormal";ue.registerClass(NA);var EA=class extends Zn{constructor(e){super(),this.DEFAULT_MEAN=0,this.DEFAULT_STDDEV=.05,this.mean=e.mean||this.DEFAULT_MEAN,this.stddev=e.stddev||this.DEFAULT_STDDEV,this.seed=e.seed}apply(e,t){if(t=t||"float32",t!=="float32"&&t!=="int32")throw new Ve(`truncatedNormal does not support dType ${t}.`);return Vf(e,this.mean,this.stddev,t,this.seed)}getConfig(){return{mean:this.mean,stddev:this.stddev,seed:this.seed}}};EA.className="TruncatedNormal";ue.registerClass(EA);var RA=class extends Zn{constructor(e){super(),this.gain=e.gain!=null?e.gain:1}apply(e,t){return X(()=>{if(e.length!==2||e[0]!==e[1])throw new H("Identity matrix initializer can only be used for 2D square matrices.");return L(this.gain,Xy(e[0]))})}getConfig(){return{gain:this.gain}}};RA.className="Identity";ue.registerClass(RA);function aB(e,t="channelsLast"){let r,n;if(jt(t),e.length===2)r=e[0],n=e[1];else if([3,4,5].indexOf(e.length)!==-1){if(t==="channelsFirst"){let a=$s(e,2);r=e[1]*a,n=e[0]*a}else if(t==="channelsLast"){let a=$s(e,0,e.length-2);r=e[e.length-2]*a,n=e[e.length-1]*a}}else{let a=$s(e);r=Math.sqrt(a),n=Math.sqrt(a)}return[r,n]}var an=class extends Zn{constructor(e){if(super(),e.scale<0)throw new H(`scale must be a positive float. Got: ${e.scale}`);this.scale=e.scale==null?1:e.scale,this.mode=e.mode==null?"fanIn":e.mode,rB(this.mode),this.distribution=e.distribution==null?"normal":e.distribution,nB(this.distribution),this.seed=e.seed}apply(e,t){let r=aB(e),n=r[0],a=r[1],s=this.scale;if(this.mode==="fanIn"?s/=Math.max(1,n):this.mode==="fanOut"?s/=Math.max(1,a):s/=Math.max(1,(n+a)/2),this.distribution==="normal"){let i=Math.sqrt(s);if(t=t||"float32",t!=="float32"&&t!=="int32")throw new Ve(`${this.getClassName()} does not support dType ${t}.`);return Vf(e,0,i,t,this.seed)}else{let i=Math.sqrt(3*s);return gd(e,-i,i,t)}}getConfig(){return{scale:this.scale,mode:this.mode,distribution:this.distribution,seed:this.seed}}};an.className="VarianceScaling";ue.registerClass(an);var nm=class extends an{constructor(e){super({scale:1,mode:"fanAvg",distribution:"uniform",seed:e==null?null:e.seed})}getClassName(){return an.className}};nm.className="GlorotUniform";ue.registerClass(nm);var am=class extends an{constructor(e){super({scale:1,mode:"fanAvg",distribution:"normal",seed:e==null?null:e.seed})}getClassName(){return an.className}};am.className="GlorotNormal";ue.registerClass(am);var sm=class extends an{constructor(e){super({scale:2,mode:"fanIn",distribution:"normal",seed:e==null?null:e.seed})}getClassName(){return an.className}};sm.className="HeNormal";ue.registerClass(sm);var im=class extends an{constructor(e){super({scale:2,mode:"fanIn",distribution:"uniform",seed:e==null?null:e.seed})}getClassName(){return an.className}};im.className="HeUniform";ue.registerClass(im);var om=class extends an{constructor(e){super({scale:1,mode:"fanIn",distribution:"normal",seed:e==null?null:e.seed})}getClassName(){return an.className}};om.className="LeCunNormal";ue.registerClass(om);var lm=class extends an{constructor(e){super({scale:1,mode:"fanIn",distribution:"uniform",seed:e==null?null:e.seed})}getClassName(){return an.className}};lm.className="LeCunNormal";ue.registerClass(lm);var $A=class extends Zn{constructor(e){if(super(),this.DEFAULT_GAIN=1,this.gain=e.gain==null?this.DEFAULT_GAIN:e.gain,this.seed=e.seed,this.seed!=null)throw new Ve("Random seed is not implemented for Orthogonal Initializer yet.")}apply(e,t){return X(()=>{if(e.length<2)throw new Ve("Shape must be at least 2D.");e[0]*e[1]>2e3&&console.warn(`Orthogonal initializer is being called on a matrix with more than 2000 (${e[0]*e[1]}) elements: Slowness may result.`);let r=e[0]>e[1]?[e[1],e[0]]:e,n=tm(r,0,1,"float32"),a=Gw.gramSchmidt(n);return e[0]>e[1]&&(a=et(a)),L(this.gain,a)})}getConfig(){return{gain:this.gain,seed:this.seed}}};$A.className="Orthogonal";ue.registerClass($A);var t4={constant:"Constant",glorotNormal:"GlorotNormal",glorotUniform:"GlorotUniform",heNormal:"HeNormal",heUniform:"HeUniform",identity:"Identity",leCunNormal:"LeCunNormal",leCunUniform:"LeCunUniform",ones:"Ones",orthogonal:"Orthogonal",randomNormal:"RandomNormal",randomUniform:"RandomUniform",truncatedNormal:"TruncatedNormal",varianceScaling:"VarianceScaling",zeros:"Zeros"};function r4(e,t={}){return Th(e,ue.SerializationMap.getMap().classNameMap,t,"initializer")}function zt(e){return wA(e)}function Rt(e){if(typeof e=="string"){let t=e in t4?t4[e]:e;if(t==="GlorotNormal")return new am;if(t==="GlorotUniform")return new nm;if(t==="HeNormal")return new sm;if(t==="HeUniform")return new im;if(t==="LeCunNormal")return new om;if(t==="LeCunUniform")return new lm;{let r={};return r.className=t,r.config={},r4(r)}}else return e instanceof Zn?e:r4(e)}function Mg(e){return Array.isArray(e)&&Array.isArray(e[0])}function E0(e){return e.length===0?[]:Array.isArray(e[0])?e:[e]}function je(e){let t;if(Array.isArray(e)){if(e.length!==1)throw new H(`Expected Tensor length to be 1; got ${e.length}`);t=e[0]}else t=e;return t}function mt(e){if(Array.isArray(e)&&Array.isArray(e[0])){if(e.length===1)return e=e,e[0];throw new H(`Expected exactly 1 Shape; got ${e.length}`)}else return e}function R0(e){let t=0;for(let r of e)r.shape.length===0?t+=1:t+=r.shape.reduce((n,a)=>n*a);return t}var n4="Variable",s6=class{constructor(e,t="float32",r=n4,n=!0,a=null){this.dtype=t==null?"float32":t,this.shape=e.shape,this.id=Jw(),r=r==null?n4:r,this.originalName=e6(r),this.name=t6(this.originalName),this.trainable_=n,this.constraint=a,this.val=Ew(e,this.trainable_,this.name,this.dtype)}read(){return this.assertNotDisposed(),this.val}write(e){return this.assertNotDisposed(),sB(this.val,e),this.val.id!==e.id&&(this.val.assign(e),this.constraint!=null&&this.val.assign(this.constraint.apply(this.val))),this}dispose(){this.assertNotDisposed(),this.val.dispose()}assertNotDisposed(){if(this.val.isDisposed)throw new Error(`LayersVariable ${this.name} is already disposed.`)}get trainable(){return this.trainable_}set trainable(e){this.trainable_=e,this.val.trainable=e}};function sB(e,t){if(e.shape.toString()!==t.shape.toString())throw new Error("Shape mismatch: "+JSON.stringify(e.shape)+" vs. "+JSON.stringify(t.shape))}function Fg(e){return e.map(t=>t.read())}function MA(e){e.forEach(t=>{t[0].write(t[1])})}var Zt=class{constructor(e){this.dtype=e.dtype,this.shape=e.shape,e.shape!=null?this.ndim=e.shape.length:this.ndim=e.ndim,this.maxNDim=e.maxNDim,this.minNDim=e.minNDim,this.axes=e.axes||{}}},pa=class{constructor(e,t,r,n,a,s,i){this.dtype=e,this.shape=t,this.sourceLayer=r,this.inputs=n,this.callArgs=a,this.outputTensorIndex=i,this.id=Jw(),s!=null&&(this.originalName=e6(s),this.name=t6(this.originalName)),this.rank=t.length}},iB=0,um=class{constructor(e,t){this.callArgs=t,this.id=iB++,this.outboundLayer=e.outboundLayer,this.inboundLayers=e.inboundLayers,this.nodeIndices=e.nodeIndices,this.tensorIndices=e.tensorIndices,this.inputTensors=e.inputTensors,this.outputTensors=e.outputTensors,this.inputMasks=e.inputMasks,this.outputMasks=e.outputMasks,this.inputShapes=e.inputShapes,this.outputShapes=e.outputShapes;for(let r of e.inboundLayers)r!=null&&r.outboundNodes.push(this);e.outboundLayer.inboundNodes.push(this)}getConfig(){let e=[];for(let t of this.inboundLayers)t!=null?e.push(t.name):e.push(null);return{outboundLayer:this.outboundLayer?this.outboundLayer.name:null,inboundLayers:e,nodeIndices:this.nodeIndices,tensorIndices:this.tensorIndices}}},oB=0,st=class extends ue.Serializable{constructor(e={}){super(),this._callHook=null,this._addedWeightNames=[],this._stateful=!1,this.id=oB++,this.activityRegularizer=null,this.inputSpec=null,this.supportsMasking=!1,this._trainableWeights=[],this._nonTrainableWeights=[],this._losses=[],this._updates=[],this._built=!1,this.inboundNodes=[],this.outboundNodes=[];let t=e.name;if(!t){let r=this.getClassName();t=ja(r)+"_"+Qf(r)}if(this.name=t,this.trainable_=e.trainable==null?!0:e.trainable,e.inputShape!=null||e.batchInputShape!=null){let r;if(e.batchInputShape!=null)r=e.batchInputShape;else if(e.inputShape!=null){let a=null;e.batchSize!=null&&(a=e.batchSize),r=[a].concat(e.inputShape)}this.batchInputShape=r;let n=e.dtype;n==null&&(n=e.inputDType),n==null&&(n="float32"),this.dtype=n}e.weights!=null?this.initialWeights=e.weights:this.initialWeights=null,this._refCount=null,this.fastWeightInitDuringBuild=!1}static nodeKey(e,t){return e.name+"_ib-"+t.toString()}getNodeAtIndex(e,t){if(this.inboundNodes.length===0)throw new da(`The layer has never been called and thus has no defined ${t}.`);if(this.inboundNodes.length<=e)throw new H(`Asked to get ${t} at node ${e}, but the layer has only ${this.inboundNodes.length} inbound nodes.`);return this.inboundNodes[e]}getInputAt(e){return tn(this.getNodeAtIndex(e,"input").inputTensors)}getOutputAt(e){return tn(this.getNodeAtIndex(e,"output").outputTensors)}get input(){if(this.inboundNodes.length>1)throw new Ga(`Layer ${this.name} has multiple inbound nodes, hence the notion of "layer input" is ill-defined. Use \`getInputAt(nodeIndex)\` instead.`);if(this.inboundNodes.length===0)throw new Ga(`Layer ${this.name} is not connected, no input to return.`);return tn(this.getNodeAtIndex(0,"input").inputTensors)}get output(){if(this.inboundNodes.length===0)throw new Ga(`Layer ${this.name} has no inbound nodes.`);if(this.inboundNodes.length>1)throw new Ga(`Layer ${this.name} has multiple inbound nodes, hence the notion of "layer output" is ill-defined. Use \`getOutputAt(nodeIndex)\` instead.`);return tn(this.getNodeAtIndex(0,"output").outputTensors)}get losses(){return this._losses}calculateLosses(){return this.losses.map(e=>e())}get updates(){return this._updates}get built(){return this._built}set built(e){this._built=e}get trainable(){return this.trainable_}set trainable(e){this._trainableWeights.forEach(t=>t.trainable=e),this.trainable_=e}get trainableWeights(){return this.trainable_?this._trainableWeights.filter(e=>e.trainable):[]}set trainableWeights(e){this._trainableWeights=e}get nonTrainableWeights(){return this.trainable?this._trainableWeights.filter(e=>!e.trainable).concat(this._nonTrainableWeights):this._trainableWeights.concat(this._nonTrainableWeights)}set nonTrainableWeights(e){this._nonTrainableWeights=e}get weights(){return this.trainableWeights.concat(this.nonTrainableWeights)}get stateful(){return this._stateful}resetStates(){if(!this.stateful)throw new Error("Cannot call the resetStates() method of a non-stateful Layer object.")}assertInputCompatibility(e){if(e=Ct(e),this.inputSpec==null||this.inputSpec.length===0)return;let t=Ct(this.inputSpec);if(e.length!==t.length)throw new H(`Layer ${this.name} expects ${t.length} inputs, but it received ${e.length} input tensors. Input received: ${e}`);for(let r=0;r<e.length;r++){let n=e[r],a=t[r];if(a==null)continue;let s=n.rank;if(a.ndim!=null&&s!==a.ndim)throw new H(`Input ${r} is incompatible with layer ${this.name}: expected ndim=${a.ndim}, found ndim=${s}`);if(a.maxNDim!=null&&s>a.maxNDim)throw new H(`Input ${r} is incompatible with layer ${this.name}: expected max_ndim=${a.maxNDim}, found ndim=${s}`);if(a.minNDim!=null&&s<a.minNDim)throw new H(`Input ${r} is incompatible with layer ${this.name}: expected min_ndim=${a.minNDim}, found ndim=${s}.`);if(a.dtype!=null&&n.dtype!==a.dtype)throw new H(`Input ${r} is incompatible with layer ${this.name} : expected dtype=${a.dtype}, found dtype=${n.dtype}.`);if(a.axes){let i=n.shape;for(let o in a.axes){let l=Number(o),u=a.axes[o],d=l>=0?i[l]:i[i.length+l];if(u!=null&&[u,null].indexOf(d)===-1)throw new H(`Input ${r} is incompatible with layer ${this.name}: expected axis ${l} of input shape to have value ${u} but got shape ${i}.`)}}if(a.shape!=null)for(let i=0;i<a.shape.length;++i){let o=a.shape[i],l=n.shape[i];if(o!=null&&l!=null&&o!==l)throw new H(`Input ${r} is incompatible with layer ${this.name}: expected shape=${a.shape}, found shape=${n.shape}.`)}}}call(e,t){return e}invokeCallHook(e,t){this._callHook!=null&&this._callHook(e,t)}setCallHook(e){this._callHook=e}clearCallHook(){this._callHook=null}apply(e,t){t=t||{},this.assertNotDisposed();let r=Ct(e),n=!0;for(let s of r)if(!(s instanceof pa)){n=!1;break}let a=!0;for(let s of r)if(s instanceof pa){a=!1;break}if(n===a)throw new H("Arguments to apply() must be all SymbolicTensors or all Tensors");return Ro(this.name,()=>{if(!this.built){this.assertInputCompatibility(e);let s=[];for(let i of Ct(e))s.push(i.shape);this.build(tn(s)),this.built=!0,this.initialWeights&&this.setWeights(this.initialWeights),this._refCount===null&&a&&(this._refCount=1)}if(this.assertInputCompatibility(e),a){let s=this.call(e,t),i=Ct(s),o=[];for(let l of i)r.indexOf(l)!==-1&&(l=l.clone()),o.push(l);if(s=tn(o),this.activityRegularizer!=null)throw new Ve("Layer invocation in the presence of activity regularizer(s) is not supported yet.");return s}else{let s=lB(e),i=this.computeOutputShape(s),o,l=uB(e);if(this.warnOnIncompatibleInputShape(Array.isArray(e)?s[0]:s),i!=null&&i.length>0&&Array.isArray(i[0])?o=i.map((u,d)=>new pa(l,u,this,Ct(e),t,this.name,d)):o=new pa(l,i,this,Ct(e),t,this.name),this.addInboundNode(e,o,null,null,s,i,t),this._refCount++,this.activityRegularizer!=null)throw new Ve("Layer invocation in the presence of activity regularizer(s) is not supported yet.");return o}})}warnOnIncompatibleInputShape(e){if(this.batchInputShape!=null)if(e.length!==this.batchInputShape.length)console.warn(`The rank of the input tensor provided (shape: ${JSON.stringify(e)}) does not match that of the batchInputShape (${JSON.stringify(this.batchInputShape)}) of the layer ${this.name}`);else{let t=!1;this.batchInputShape.forEach((r,n)=>{r!=null&&e[n]!=null&&e[n]!==r&&(t=!0)}),t&&console.warn(`The shape of the input tensor (${JSON.stringify(e)}) does not match the expectation of layer ${this.name}: ${JSON.stringify(this.batchInputShape)}`)}}get outputShape(){if(this.inboundNodes==null||this.inboundNodes.length===0)throw new Ga(`The layer ${this.name} has never been called and thus has no defined output shape.`);let e=[];for(let t of this.inboundNodes){let r=JSON.stringify(t.outputShapes);e.indexOf(r)===-1&&e.push(r)}if(e.length===1){let t=this.inboundNodes[0].outputShapes;return Array.isArray(t)&&Array.isArray(t[0])&&t.length===1?t[0]:t}else throw new Ga(`The layer ${this.name} has multiple inbound nodes with different output shapes. Hence the notion of "output shape" is ill-defined for the layer.`)}countParams(){if(!this.built)throw new da(`You tried to call countParams() on ${this.name}, but the layer is not built yet. Build it first by calling build(batchInputShape).`);return R0(this.weights)}build(e){this.built=!0}getWeights(e=!1){return Fg(e?this.trainableWeights:this.weights)}setWeights(e){X(()=>{let t=this.weights;if(t.length!==e.length)throw new H(`You called setWeights(weights) on layer "${this.name}" with a weight list of length ${e.length}, but the layer was expecting ${t.length} weights. Provided weights: ${e}...`);if(t.length===0)return;let r=[],n=Fg(t);for(let a=0;a<n.length;++a){let s=n[a],i=t[a],o=e[a];if(!v.arraysEqual(s.shape,o.shape))throw new H(`Layer weight shape ${s.shape} not compatible with provided weight shape ${o.shape}`);r.push([i,o])}MA(r)})}addWeight(e,t,r,n,a,s,i,o){if(this._addedWeightNames.indexOf(e)!==-1)throw new H(`Duplicate weight name ${e} for layer ${this.name}`);this._addedWeightNames.push(e),r==null&&(r="float32"),this.fastWeightInitDuringBuild&&(n=o!=null?o():Rt("zeros"));let l=n.apply(t,r),u=new s6(l,r,e,s,i);return l.dispose(),a!=null&&this.addLoss(()=>a.apply(u.read())),s==null&&(s=!0),s?this._trainableWeights.push(u):this._nonTrainableWeights.push(u),u}setFastWeightInitDuringBuild(e){this.fastWeightInitDuringBuild=e}addLoss(e){e==null||Array.isArray(e)&&e.length===0||(e=Ct(e),this._losses!==void 0&&this._losses!==null&&this.losses.push(...e))}computeOutputShape(e){return e}computeMask(e,t){if(!this.supportsMasking){if(t!=null)if(Array.isArray(t))t.forEach(r=>{if(r!=null)throw new TypeError(`Layer ${this.name} does not support masking, but was passed an inputMask.`)});else throw new TypeError(`Layer ${this.name} does not support masking, but was passed an inputMask.`);return null}return t}addInboundNode(e,t,r,n,a,s,i=null){let o=Ct(e);t=Ct(t),r=Ct(r),n=Ct(n),a=E0(a),s=E0(s);let l=[],u=[],d=[];for(let h of o)l.push(h.sourceLayer),u.push(h.nodeIndex),d.push(h.tensorIndex);new um({outboundLayer:this,inboundLayers:l,nodeIndices:u,tensorIndices:d,inputTensors:o,outputTensors:t,inputMasks:r,outputMasks:n,inputShapes:a,outputShapes:s},i);for(let h=0;h<t.length;h++)t[h].sourceLayer=this,t[h].nodeIndex=this.inboundNodes.length-1,t[h].tensorIndex=h}getConfig(){let e={name:this.name,trainable:this.trainable};return this.batchInputShape!=null&&(e.batchInputShape=this.batchInputShape),this.dtype!=null&&(e.dtype=this.dtype),e}disposeWeights(){return this.weights.forEach(e=>e.dispose()),this.weights.length}assertNotDisposed(){if(this._refCount===0)throw new Error(`Layer '${this.name}' is already disposed.`)}dispose(){if(!this.built)throw new Error(`Cannot dispose Layer ${this.name} because it has not been built yet.`);if(this._refCount===null)throw new Error(`Cannot dispose Layer ${this.name} because it has not been used yet.`);this.assertNotDisposed();let e=0;return--this._refCount===0&&(e=this.disposeWeights()),{refCountAfterDispose:this._refCount,numDisposedVariables:e}}};function lB(e){e=Ct(e);let t=[];for(let r of e)t.push(r.shape);return tn(t)}function uB(e){return"float32"}function i6(e,t,r){if((t==null||r!=null&&r>0)&&(t=e.sourceLayer,r=e.nodeIndex),t.inboundNodes.length===0)return[e];{let n=t.inboundNodes[r];if(n.inboundLayers.length===0)return n.inputTensors;{let a=[];for(let s=0;s<n.inboundLayers.length;s++){let i=n.inputTensors[s],o=n.inboundLayers[s],l=n.nodeIndices[s],u=i6(i,o,l);for(let d of u)a.indexOf(d)===-1&&a.push(d)}return a}}}var xd=class extends st{constructor(e){if(super({dtype:e.dtype,name:e.name!=null?e.name:Qf("input").toString()}),e.batchSize==null&&(e.batchSize=null),e.sparse==null&&(e.sparse=!1),this.trainable=!1,this.built=!0,this.sparse=e.sparse,e.inputShape!=null&&e.batchInputShape!=null)throw new H("Only provide the inputShape OR batchInputShape argument to inputLayer, not both at the same time.");let t=e.batchInputShape;if(t==null){if(e.inputShape==null)throw new H("An InputLayer should be passed either a `batchInputShape` or an `inputShape`.");t=[e.batchSize].concat(e.inputShape)}else if(e.batchSize!=null)throw new H("Cannot specify batchSize if batchInputShape is specified when creating an InputLayer.");let r=e.dtype||"float32";this.batchInputShape=t,this.dtype=r,this.inputSpec=[{shape:t}];let n=new pa(this.dtype,this.batchInputShape,this,[],{},this.name);n.nodeIndex=0,n.tensorIndex=0,new um({outboundLayer:this,inboundLayers:[],nodeIndices:[],tensorIndices:[],inputTensors:[n],outputTensors:[n],inputMasks:[null],outputMasks:[null],inputShapes:[t],outputShapes:[t]})}apply(e,t){throw new H(`Cannot pass any input to an InputLayer's apply() method. InputLayer name: ${this.name}`)}dispose(){return{refCountAfterDispose:this._refCount,numDisposedVariables:0}}getConfig(){return{batchInputShape:this.batchInputShape,dtype:this.dtype,sparse:this.sparse,name:this.name}}};xd.className="InputLayer";ue.registerClass(xd);function o6(e){if(e.batchShape==null&&e.shape==null)throw new Error("Please provide to Input either a `shape` or a `batchShape` argument. Note that `shape` does not include the batch dimension.");if(e.batchShape!=null&&e.shape!=null)throw new H("Please provide either a `shape` or `batchShape` argument to Input, but not both.");let t=e.batchShape;e.shape!=null&&t==null&&(t=[null].concat(e.shape));let r=e.dtype;return r==null&&(r="float32"),new xd({batchInputShape:t,name:e.name,dtype:r,sparse:e.sparse}).inboundNodes[0].outputTensors[0]}function dB(e,t){if(e.dtype==null||e.dtype===t.dtype)return t;try{return me(t,e.dtype)}catch(r){throw new H(`The dtype of the feed (${t.dtype}) can not be cast to the dtype of the key '${e.name}' (${e.dtype}).`)}}var To=class{constructor(e){if(this.id2Value={},this.id2Mask={},this.name2Id={},e instanceof To)for(let t in e.id2Value)this.id2Value[t]=e.id2Value[t],t in e.id2Mask&&(this.id2Mask[t]=e.id2Mask[t]);else{if(e==null)return;for(let t of e)this.add(t.key,t.value)}}add(e,t,r){if(this.id2Value[e.id]==null)this.id2Value[e.id]=dB(e,t),this.name2Id[e.name]=e.id,r!=null&&(this.id2Mask[e.id]=r);else throw new H(`Duplicate key: name=${e.name}, id=${e.id}`);return this}addFeed(e){this.add(e.key,e.value)}hasKey(e){return this.id2Value[e.id]!=null}names(){return Object.keys(this.name2Id)}getValue(e){if(e instanceof pa){if(this.id2Value[e.id]==null)throw new H(`Nonexistent key: ${e.name}`);return this.id2Value[e.id]}else{let t=this.name2Id[e];if(t==null)throw new H(`Feed dict has no SymbolicTensor name: ${e}`);return this.id2Value[t]}}getMask(e){if(e instanceof pa){if(this.id2Value[e.id]==null)throw new H(`Nonexistent key: ${e.name}`);return this.id2Mask[e.id]}else{let t=this.name2Id[e];if(t==null)throw new H(`Feed dict has no SymbolicTensor name: ${e}`);return this.id2Mask[t]}}disposeMasks(){this.id2Mask!=null&&te(this.id2Mask)}},$0=new Kw,M0=new Kw;function pB(e){$0!=null&&$0.setMaxEntries(e),M0!=null&&M0.setMaxEntries(e)}function Ip(e,t,r,n){let a=r==null?!1:r.training,s=Array.isArray(e),i=s?e:[e],o=i.map(f=>f.name),l=[],u=t.names();for(let f of o)u.indexOf(f)!==-1?l.push(t.getValue(f)):l.push(null);n!=null&&(n.maxNumTensors=-1/0,n.minNumTensors=1/0);let d=o.join(",")+"|"+t.names().sort().join(","),h=$0.get(d),p;if(h==null){let f=hB(i,t);h=f.sorted,p=f.recipientCounts,$0.put(d,h),M0.put(d,p)}p={},a||Object.assign(p,M0.get(d));let c=new To(t);for(let f=0;f<h.length;++f){if(n!=null){let R=S0().numTensors;R>n.maxNumTensors&&(n.maxNumTensors=R),R<n.minNumTensors&&(n.minNumTensors=R)}let m=h[f],g=m.sourceLayer;if(g instanceof xd)continue;let y=[],A=[],x=[],b=!1;for(let R of m.inputs){let z=c.getValue(R),M=c.getMask(R);y.push(z),A.push(M),M!=null&&(b=!0),a||(p[R.name]--,p[R.name]===0&&!t.hasKey(R)&&o.indexOf(R.name)===-1&&!z.isDisposed&&R.sourceLayer.stateful!==!0&&x.push(z))}b&&(r=r||{},r.mask=A[0]);let w=Ct(g.apply(y,r)),S=null;g.supportsMasking&&(S=g.computeMask(y,A));let N=fB(m),E=Array.isArray(N)?N:[N];for(let R=0;R<E.length;++R){c.hasKey(E[R])||c.add(E[R],w[R],Array.isArray(S)?S[0]:S);let z=o.indexOf(E[R].name);z!==-1&&(l[z]=w[R])}a||te(x)}return c.disposeMasks(),s?l:l[0]}function hB(e,t){v.assert(e!=null&&e.length>0,()=>"Expected at least one fetch, got none");let r=[],n={};if(e.length===1){let a=a4(e[0],t);r=a.sorted,n=a.recipientMap}else{let a=new Set;for(let s of e){let{sorted:i,recipientMap:o}=a4(s,t);for(let l of i)a.has(l.name)||(r.push(l),a.add(l.name));for(let l in o)n[l]==null&&(n[l]=new Set),o[l].forEach(u=>n[l].add(u))}}return{sorted:r,recipientCounts:cB(n)}}function cB(e){let t={};for(let r in e)t[r]=e[r].size;return t}function a4(e,t){let r=new Set,n=[],a={};for(let o of t.names())r.add(o);let s=[],i=[];for(s.push(e);s.length>0;){let o=s[s.length-1];if(r.has(o.name)){s.pop();continue}let l=i[i.length-1]===s.length-1;if(o.inputs.length===0||l)s.pop(),n.push(o),r.add(o.name),l&&i.pop();else{i.push(s.length-1);for(let u of o.inputs)a[u.name]==null&&(a[u.name]=new Set),a[u.name].add(o.name),!r.has(u.name)&&s.push(u)}}return{sorted:n,recipientMap:a}}function fB(e){let t;if(e.sourceLayer.inboundNodes.length===1)t=e.sourceLayer.output;else{let r=null;for(let n=0;n<e.sourceLayer.inboundNodes.length;++n)for(let a of e.sourceLayer.inboundNodes[n].outputTensors)if(a.id===e.id){r=n;break}t=e.sourceLayer.getOutputAt(r)}return t}var mB=Z();mB.registerFlag("TOPOLOGICAL_SORT_CACHE_MAX_ENTRIES",()=>100,pB);var l6={kernelName:Ho,inputsToSave:["x"],gradFunc:(e,t)=>{let[r]=t;return{x:()=>L(e,Sh(me(r,"float32"),-1))}}},gB={kernelName:Du,inputsToSave:["x"],gradFunc:(e,t)=>{let[r]=t;return{x:()=>{let n=At(me(r,"float32")),a=Er(ce(Se(1),n));return Dt(pe(e,a))}}}},yB={kernelName:Lu,inputsToSave:["x"],gradFunc:(e,t)=>{let[r]=t;return{x:()=>{let n=Er(ce(At(me(r,"float32")),1));return pe(e,n)}}}},AB={kernelName:Ja,inputsToSave:["a","b"],gradFunc:(e,t)=>{let[r,n]=t,a=bt(r.shape,n.shape);return{a:()=>{let s=e,i=Jt(r.shape,a);return i.length>0&&(s=ke(s,i)),U(s,r.shape)},b:()=>{let s=e,i=Jt(n.shape,a);return i.length>0&&(s=ke(s,i)),U(s,n.shape)}}}},xB={kernelName:Zs,saveAllInputs:!0,gradFunc:(e,t)=>{let r={};return t.forEach((n,a)=>{r[a]=()=>e.clone()}),r}},bB={kernelName:Ys,inputsToSave:["x"],gradFunc:(e,t)=>{let[r]=t;return{x:()=>at(r)}}},vB={kernelName:Vu,inputsToSave:["x"],gradFunc:(e,t)=>{let[r]=t;return{x:()=>at(r)}}},wB={kernelName:Uu,inputsToSave:["x"],gradFunc:(e,t)=>{let[r]=t;return{x:()=>pe(e,Er(ce(Se(1),At(me(r,"float32")))))}}},kB={kernelName:Gu,inputsToSave:["x"],gradFunc:(e,t)=>{let[r]=t;return{x:()=>{let n=Er(le(Se(1),At(me(r,"float32"))));return pe(e,n)}}}},IB={kernelName:qu,inputsToSave:["a","b"],gradFunc:(e,t)=>{let[r,n]=t,a=bt(r.shape,n.shape);return{a:()=>{let s=le(At(r),At(n)),i=L(e,pe(n,s)),o=Jt(r.shape,a);return o.length>0&&(i=ke(i,o)),U(i,r.shape)},b:()=>{let s=le(At(r),At(n)),i=Dt(L(e,pe(r,s))),o=Jt(n.shape,a);return o.length>0&&(i=ke(i,o)),U(i,n.shape)}}}},SB={kernelName:ju,inputsToSave:["x"],gradFunc:(e,t)=>{let[r]=t;return{x:()=>pe(e,le(At(me(r,"float32")),1))}}},CB={kernelName:Hu,inputsToSave:["x"],gradFunc:(e,t)=>{let[r]=t;return{x:()=>pe(e,ce(Se(1),At(me(r,"float32"))))}}};function TB(e,t,r,n,a,s){let i=$(e,"dy","avgPool3dGrad"),o=$(t,"input","avgPool3dGrad"),l=i,u=o,d=!1;o.rank===4&&(d=!0,l=U(i,[1,i.shape[0],i.shape[1],i.shape[2],i.shape[3]]),u=U(o,[1,o.shape[0],o.shape[1],o.shape[2],o.shape[3]])),P(l.rank===5,()=>`Error in avgPool3dGrad: dy must be rank 5 but got rank ${l.rank}.`),P(u.rank===5,()=>`Error in avgPool3dGrad: input must be rank 5 but got rank ${u.rank}.`),jr("avgPool3dGrad",a,s);let h={dy:l,input:u},p={filterSize:r,strides:n,pad:a,dimRoundingMode:s},c=B.runKernel(Q0,h,p);return d?U(c,[c.shape[1],c.shape[2],c.shape[3],c.shape[4]]):c}var NB=W({avgPool3dGrad_:TB}),EB={kernelName:Yp,inputsToSave:["x"],gradFunc:(e,t,r)=>{let[n]=t,{filterSize:a,strides:s,pad:i,dimRoundingMode:o}=r;return{x:()=>NB(e,n,a,s,i,o)}}};function RB(e,t,r,n,a){let s=$(e,"dy","avgPoolGrad"),i=$(t,"input","avgPoolGrad");P(i.rank===s.rank,()=>`Rank of input (${i.rank}) does not match rank of dy (${s.rank})`);let o=i,l=s,u=!1;i.rank===3&&(u=!0,o=U(i,[1,i.shape[0],i.shape[1],i.shape[2]]),l=U(s,[1,s.shape[0],s.shape[1],s.shape[2]])),P(l.rank===4,()=>`Error in avgPoolGrad: dy must be rank 4 but got rank ${l.rank}.`),P(o.rank===4,()=>`Error in avgPoolGrad: input must be rank 4 but got rank ${o.rank}.`);let d={dy:l,input:o},h={filterSize:r,strides:n,pad:a},p=B.runKernel(J0,d,h);return u?U(p,[p.shape[1],p.shape[2],p.shape[3]]):p}var $B=W({avgPoolGrad_:RB}),MB={kernelName:Js,inputsToSave:["x"],gradFunc:(e,t,r)=>{let[n]=t,{filterSize:a,strides:s,pad:i}=r;return{x:()=>$B(e,n,a,s,i)}}},FB={kernelName:Qs,inputsToSave:["a","b"],gradFunc:(e,t,r)=>{let[n,a]=t,{transposeA:s,transposeB:i}=r;return!s&&!i?{a:()=>Ze(e,a,!1,!0),b:()=>Ze(n,e,!0,!1)}:!s&&i?{a:()=>Ze(e,a,!1,!1),b:()=>Ze(e,n,!0,!1)}:s&&!i?{a:()=>Ze(a,e,!1,!0),b:()=>Ze(n,e,!1,!1)}:{a:()=>Ze(a,e,!0,!0),b:()=>Ze(e,n,!0,!0)}}},PB={kernelName:qo,gradFunc:(e,t,r)=>{let{blockShape:n,crops:a}=r;return{x:()=>Of(e,n,a)}}},_B={kernelName:Dv,gradFunc:(e,t,r)=>{let n=r,a=n.inputShape,s=n.shape,i=Array.from(s);for(let l=a.length-1;l>=0;l--)if(a[l]===s[l])i[l]=1;else if(a[l]!==1)throw new Error(`broadcastTo(): [${a}] cannot be broadcast to [${s}].`);let o=[];for(let l=0;l<i.length;l++)i[l]>1&&o.push(l);return{x:()=>ke(e,o,!0)}}},zB={kernelName:ei,gradFunc:e=>({x:()=>e.clone()})},OB={kernelName:ti,gradFunc:e=>({x:()=>at(e)})},DB={kernelName:Qa,inputsToSave:["x"],gradFunc:(e,t,r)=>{let[n]=t,{clipValueMin:a,clipValueMax:s}=r;return{x:()=>Vr(ga(Ml(n,a),Fl(n,s)),e,at(e))}}},LB={kernelName:Qp,inputsToSave:["x"],gradFunc:l6.gradFunc},BB={kernelName:Xo,saveAllInputs:!0,gradFunc:(e,t,r)=>{let n=t.map(o=>o.shape),{axis:a}=r,s=Hn(a,t[0].shape)[0],i=n.map(o=>o[s]);return Yt(e,i,s).map(o=>()=>o)}},WB={kernelName:ri,inputsToSave:["x","filter"],gradFunc:(e,t,r)=>{let[n,a]=t,{dilations:s,strides:i,pad:o,dataFormat:l}=r;return P(Ds(s),()=>`Error in gradient of conv2D: dilation rates greater than 1 are not yet supported in gradients. Got dilations '${s}'`),{x:()=>Uy(n.shape,e,a,i,o,l),filter:()=>AA(n,e,a.shape,i,o,l)}}},VB={kernelName:ni,inputsToSave:["dy","filter"],gradFunc:(e,t,r)=>{let[n,a]=t,{strides:s,pad:i,dataFormat:o,dimRoundingMode:l}=r;return{dy:()=>Ls(e,a,s,i,o,1,l),filter:()=>AA(e,n,a.shape,s,i,o,l)}}};function UB(e,t,r,n,a){let s=e;e.rank===4&&(s=U(e,[1,e.shape[0],e.shape[1],e.shape[2],e.shape[3]]));let i=t;i.rank===4&&(i=U(t,[1,t.shape[0],t.shape[1],t.shape[2],t.shape[3]])),P(s.rank===5,()=>`Error in conv3dDerFilter: input must be rank 5, but got shape ${s.shape}.`),P(i.rank===5,()=>`Error in conv3dDerFilter: dy must be rank 5, but got shape ${i.shape}.`),P(r.length===5,()=>`Error in conv3dDerFilter: filterShape must be length 5, but got ${r}.`),P(s.shape[4]===r[3],()=>`Error in conv3dDerFilter: depth of input ${s.shape[4]}) must match input depth in filter (${r[3]}.`),P(i.shape[4]===r[4],()=>`Error in conv3dDerFilter: depth of dy (${i.shape[4]}) must match output depth for filter (${r[4]}).`);let o={x:s,dy:i},l={strides:n,pad:a,filterShape:r};return B.runKernel(nf,o,l)}var GB=W({conv3DBackpropFilter_:UB}),jB={kernelName:eh,inputsToSave:["x","filter"],gradFunc:(e,t,r)=>{let{dilations:n,strides:a,pad:s}=r;P(Ds(n),()=>`Error in gradient of conv3D: dilation rates greater than 1 are not yet supported in gradients. Got dilations '${n}'`);let[i,o]=t;return{x:()=>J7(i.shape,e,o,a,s),filter:()=>GB(i,e,o.shape,a,s)}}},HB={kernelName:ai,inputsToSave:["x"],gradFunc:(e,t)=>{let[r]=t;return{x:()=>L(Dt(uA(me(r,"float32"))),e)}}},qB={kernelName:si,inputsToSave:["x"],gradFunc:(e,t)=>{let[r]=t;return{x:()=>L(dA(me(r,"float32")),e)}}},XB={kernelName:ii,inputsToSave:["x"],gradFunc:(e,t,r)=>{let[n]=t,{axis:a,exclusive:s,reverse:i}=r;return{x:()=>{let o=cw([a],n.rank),l=qy(e,a,s,!i);return o!=null&&(l=et(l,o)),l}}}},KB={kernelName:oi,inputsToSave:["x","filter"],gradFunc:(e,t,r)=>{let{dilations:n,strides:a,pad:s,dimRoundingMode:i}=r,o=n==null?[1,1]:n;P(Ds(o),()=>`Error in gradient of depthwiseConv2dNative: dilation rates greater than 1 are not yet supported. Got dilations '${o}'`);let[l,u]=t;return P(l.rank===4,()=>`Error in gradient of depthwiseConv2dNative: input must be rank 4, but got rank ${l.rank}.`),P(u.rank===4,()=>`Error in gradient of depthwiseConv2dNative: filter must be rank 4, but got rank ${u.rank}.`),P(l.shape[3]===u.shape[2],()=>`Error in gradient of depthwiseConv2d: number of input channels (${l.shape[3]}) must match the inChannels dimension in filter ${u.shape[2]}.`),P(za(a,o),()=>`Error in gradient of depthwiseConv2d: Either strides or dilations must be 1. Got strides ${a} and dilations '${o}'.`),jr("depthwiseConv2d",s,i),{x:()=>Ow(l.shape,e,u,a,s,o,i),filter:()=>zw(l,e,u.shape,a,s,o,i)}}},ZB={kernelName:th,inputsToSave:["x","filter"],gradFunc:(e,t,r)=>{let[n,a]=t,s={x:n,filter:a,dy:e},i={x:n,filter:a,dy:e};return{x:()=>B.runKernel(x0,s,r),filter:()=>B.runKernel(b0,i,r)}}},YB={kernelName:ui,outputsToSave:[!0],gradFunc:(e,t)=>{let[r]=t,n={dy:e,y:r};return{x:()=>B.runKernel(df,n)}}},JB={kernelName:Xu,inputsToSave:["x"],gradFunc:(e,t)=>{let[r]=t,n=L(Mn(Dt(At(r))),2/Math.sqrt(Math.PI));return{x:()=>L(e,n)}}},QB={kernelName:di,outputsToSave:[!0],gradFunc:(e,t)=>{let[r]=t;return{x:()=>L(e,r)}}},eW={kernelName:Qo,inputsToSave:["input"],gradFunc:(e,t)=>{let[r]=t;return{input:()=>U(e,r.shape)}}},tW={kernelName:el,inputsToSave:["x"],gradFunc:(e,t)=>{let[r]=t;return{x:()=>L(e,Mn(r))}}},rW={kernelName:pi,gradFunc:e=>({x:()=>at(e)})},nW={kernelName:hi,inputsToSave:["a","b"],gradFunc:(e,t)=>{let[r,n]=t,a=bt(r.shape,n.shape);return{a:()=>{let s=pe(e,me(n,"float32")),i=Jt(r.shape,a);return i.length>0?U(ke(s,i),r.shape):s},b:()=>{let s=L(e,me(r,"float32")),i=Jt(n.shape,a);i.length>0&&(s=U(ke(s,i),n.shape));let o=At(n);return Dt(pe(s,me(o,"float32")))}}}},aW={kernelName:ci,inputsToSave:["x","mean","variance","scale"],gradFunc:(e,t,r)=>{let{varianceEpsilon:n}=r,[a,s,i,o]=t,l=o==null?Se(1):o,u=Jt(s.shape,a.shape),d=[];if(s.rank===1){for(let m=0;m<a.shape.length-1;++m)d.push(a.shape[m]);d.push(1)}let h=ce(a,s),p=L(e,l),c=oA(le(i,Se(n))),f=L(L(L(c,c),c),Se(-.5));return{x:()=>s.rank===1?U(L(L(e,Gn(U(c,[1,1,1,s.shape[0]]),d)),l),a.shape):U(L(L(e,c),l),a.shape),mean:()=>{let m=L(L(c,Se(-1)),p);return s.rank===1&&(m=ke(m,u)),U(m,s.shape)},variance:()=>{let m=L(L(f,h),p);return s.rank===1&&(m=ke(m,u)),U(m,s.shape)},scale:()=>{let m=L(h,c),g=L(e,m);return s.rank===1&&(g=ke(g,u)),U(g,s.shape)},offset:()=>{let m=e;return s.rank===1&&(m=ke(m,u)),U(m,s.shape)}}}},sW={kernelName:rl,inputsToSave:["x","indices"],gradFunc:(e,t,r)=>{let[n,a]=t,{axis:s}=r,i=Hn(s,n.shape)[0];return{x:()=>{let o=n.shape,l=a.size,u=o.slice(0,i),d=u.length,h=o.slice(s,o.length).slice(1),p=h.length,c=s4(0,d),f=s4(d+1,d+1+p),m=i4([u,[l],h]),g=U(e,m),y=U(a,[l]),A=i4([[d],c,f]),x=et(g,A),b=Tw(x,y,n.shape[i]),w=Jy(A);return b=et(b,w),b},indices:()=>a}}};function s4(e,t){let r=[];for(let n=e;n<t;++n)r.push(n);return r}function i4(e){let t=[];for(let r=0;r<e.length;++r)for(let n=0;n<e[r].length;++n)t.push(e[r][n]);return t}var iW={kernelName:fi,inputsToSave:["a","b"],gradFunc:(e,t)=>{let[r,n]=t;return{a:()=>at(r),b:()=>at(n)}}},oW={kernelName:mi,gradFunc:e=>({x:()=>me(e,"float32")})},lW={kernelName:Zu,gradFunc:e=>({x:()=>at(e)})},uW={kernelName:Yu,gradFunc:e=>({x:()=>at(e)})},dW={kernelName:Ju,gradFunc:e=>({x:()=>at(e)})},pW={kernelName:gi,inputsToSave:["x"],gradFunc:(e,t,r)=>{let[n]=t,{alpha:a}=r,s=mn(n,0);return{x:()=>Vr(s,e,L(e,a))}}},hW={kernelName:Qu,inputsToSave:["x"],gradFunc:(e,t)=>{let[r]=t;return{x:()=>pe(e,le(r,1))}}},cW={kernelName:yi,inputsToSave:["x"],gradFunc:(e,t)=>{let[r]=t;return{x:()=>pe(e,me(r,"float32"))}}},fW={kernelName:Lv,inputsToSave:[],outputsToSave:[!0],gradFunc:(e,t,r)=>{let[n]=t,{axis:a}=r;return{logits:()=>{let s=Mn(n);return ce(e,L(ke(e,a,!0),s))}}}};function mW(e,t,r,n=5,a=1,s=1,i=.5){let o={x:e,y:t,dy:r},l={depthRadius:n,bias:a,alpha:s,beta:i};return B.runKernel(ff,o,l)}var gW=W({localResponseNormalizationBackprop_:mW}),yW={kernelName:sh,inputsToSave:["x"],outputsToSave:[!0],gradFunc:(e,t,r)=>{let[n,a]=t,{depthRadius:s,bias:i,alpha:o,beta:l}=r;return{x:()=>gW(n,a,e,s,i,o,l)}}};function u6(e,t,r,n){return t.rank<r.rank&&(t=U(t,zo(t.shape,n))),e.rank<r.rank&&(e=U(e,zo(e.shape,n))),{x:()=>L(e,me($n(r,t),e.dtype))}}var o4={kernelName:Ai,inputsToSave:["x"],outputsToSave:[!0],gradFunc:(e,t,r)=>{let n=r,{reductionIndices:a}=n,s=t[0],i=t[1],o=Hn(a,s.shape),l=u6(e,i,s,o);return{x:()=>l.x()}}},AW={kernelName:xi,inputsToSave:["a","b"],gradFunc:(e,t)=>{let[r,n]=t;return{a:()=>L(e,me(Ml(r,n),"float32")),b:()=>L(e,me(Ky(r,n),"float32"))}}};function xW(e,t,r,n,a,s,i){let o=$(e,"dy","maxPool3dGrad"),l=$(t,"input","maxPool3dGrad"),u=$(r,"output","maxPool3dGrad"),d=o,h=l,p=u,c=!1;l.rank===4&&(c=!0,d=U(o,[1,o.shape[0],o.shape[1],o.shape[2],o.shape[3]]),h=U(l,[1,l.shape[0],l.shape[1],l.shape[2],l.shape[3]]),p=U(u,[1,u.shape[0],u.shape[1],u.shape[2],u.shape[3]])),P(d.rank===5,()=>`Error in maxPool3dGrad: dy must be rank 5 but got rank ${d.rank}.`),P(h.rank===5,()=>`Error in maxPool3dGrad: input must be rank 5 but got rank ${h.rank}.`),P(p.rank===5,()=>`Error in maxPool3dGrad: output must be rank 5 but got rank ${p.rank}.`),jr("maxPool3dGrad",s,i);let f={dy:d,input:h,output:p},m={filterSize:n,strides:a,pad:s,dimRoundingMode:i},g=B.runKernel(gf,f,m);return c?U(g,[g.shape[1],g.shape[2],g.shape[3],g.shape[4]]):g}var bW=W({maxPool3dGrad_:xW}),vW={kernelName:ih,inputsToSave:["x"],outputsToSave:[!0],gradFunc:(e,t,r)=>{let[n,a]=t,{filterSize:s,strides:i,pad:o,dimRoundingMode:l}=r;return{x:()=>bW(e,n,a,s,i,o,l)}}};function wW(e,t,r,n,a,s,i){let o=$(e,"dy","maxPoolGrad"),l=$(t,"input","maxPoolGrad"),u=$(r,"output","maxPoolGrad");P(l.rank===o.rank,()=>`Rank of input (${l.rank}) does not match rank of dy (${o.rank})`),P(o.rank===4,()=>`Error in maxPoolGrad: dy must be rank 4 but got rank ${o.rank}.`),P(l.rank===4,()=>`Error in maxPoolGrad: input must be rank 4 but got rank ${l.rank}.`),jr("maxPoolGrad",s,i);let d={dy:o,input:l,output:u},h={filterSize:n,strides:a,pad:s,dimRoundingMode:i};return B.runKernel(mf,d,h)}var kW=W({maxPoolGrad_:wW}),IW={kernelName:bi,inputsToSave:["x"],outputsToSave:[!0],gradFunc:(e,t,r)=>{let[n,a]=t,{filterSize:s,strides:i,pad:o}=r;return{x:()=>kW(e,n,a,s,i,o)}}},SW={kernelName:vi,inputsToSave:["x"],gradFunc:(e,t,r)=>{let[n]=t,{axis:a}=r,s=Hn(a,n.shape),i=hw(n.shape,s)[1],o=It(i);return{x:()=>{let l=n.shape.slice();s.forEach(d=>{l[d]=1});let u=U(e,l);return pe(L(u,hn(n.shape,"float32")),o)}}}},CW={kernelName:wi,inputsToSave:["x"],outputsToSave:[!0],gradFunc:(e,t,r)=>{let n=r,{axis:a}=n,[s,i]=t,o=Hn(a,s.shape),l=u6(e,i,s,o);return{x:()=>l.x()}}},TW={kernelName:ki,inputsToSave:["a","b"],gradFunc:(e,t)=>{let[r,n]=t;return{a:()=>L(e,me(Fl(r,n),"float32")),b:()=>L(e,me(mn(r,n),"float32"))}}},NW={kernelName:Ii,inputsToSave:["x"],gradFunc:(e,t,r)=>{let n=t[0],{paddings:a}=r,s=a.map(i=>i[0]);return{x:()=>Pe(e,s,n.shape)}}},EW={kernelName:td,inputsToSave:["a","b"],gradFunc:(e,t)=>{let[r,n]=t,a=bt(r.shape,n.shape);return{a:()=>{let s=Jt(r.shape,a);return s.length>0?U(ke(e,s),r.shape):e},b:()=>{let s=L(e,Dt(kh(pe(r,n)))),i=Jt(n.shape,a);return i.length>0?U(ke(s,i),n.shape):s}}}},RW={kernelName:Si,inputsToSave:["a","b"],gradFunc:(e,t)=>{let[r,n]=t,a=bt(r.shape,n.shape);return{a:()=>{let s=L(e,me(n,"float32")),i=Jt(r.shape,a);return i.length>0?U(ke(s,i),r.shape):s},b:()=>{let s=L(e,me(r,"float32")),i=Jt(n.shape,a);return i.length>0?U(ke(s,i),n.shape):s}}}},$W={kernelName:ll,gradFunc:e=>({x:()=>Dt(e)})},MW={kernelName:cl,inputsToSave:["indices"],gradFunc:(e,t)=>{let r=t[0];return{indices:()=>_t(r.shape,"float32")}}},FW={kernelName:hl,gradFunc:e=>({x:()=>at(e)})},PW={kernelName:fl,saveAllInputs:!0,gradFunc:(e,t,r)=>{let{axis:n}=r;return nn(e,n).map(a=>()=>a)}},l4={kernelName:Ci,inputsToSave:["x"],gradFunc:(e,t,r)=>{let n=t[0],{paddings:a}=r,s=a.map(i=>i[0]);return{x:()=>Pe(e,s,n.shape)}}},_W={kernelName:Ti,inputsToSave:["a","b"],outputsToSave:[!0],gradFunc:(e,t)=>{let[r,n,a]=t,s=r,i=n,o=bt(s.shape,i.shape);return{a:()=>{let l=me(i,"float32"),u=L(e,L(l,Ws(s,ce(l,Se(1))))),d=Jt(s.shape,o);return d.length>0&&(u=ke(u,d)),U(u,s.shape)},b:()=>{let l=mn(s,0),u=Vr(l,Fn(s),at(s)),d=L(e,L(a,u)),h=Jt(i.shape,o);return h.length>0&&(d=ke(d,h)),U(d,i.shape)}}}},zW={kernelName:Ni,inputsToSave:["x","alpha"],gradFunc:(e,t)=>{let[r,n]=t,a=mn(r,0);return{x:()=>Vr(a,e,L(e,n)),alpha:()=>{let s=Vr(a,at(e),L(e,r)),i=Jt(n.shape,e.shape);return i.length>0&&(s=ke(s,i)),U(s,n.shape)}}}};function OW(e,t,r){let n=e.shape.slice();n[r]=1;let a=U(t,n),s=N0(e,r,!0,!1),i=N0(e,r,!0,!0),o=L(s,i);return L(a,o)}function DW(e,t,r){let n=e.shape.length,a=n-r.length,s=C.getAxesPermutation(r,n),i=e;s!=null&&(i=et(e,s));let o=i.shape.slice(),l=o.splice(n-r.length,r.length).reduce((h,p)=>h*p,1);o.push(l);let u=i.reshape(o),d=OW(u,t,a);if(d=d.reshape(i.shape),s!=null){let h=C.getUndoAxesPermutation(s);d=et(d,h)}return d}var LW={kernelName:Ei,inputsToSave:["x"],gradFunc:(e,t,r)=>{let[n]=t,{axis:a}=r,s=[];return a==null?s=n.shape.map((i,o)=>o):typeof a=="number"?s=[a]:s=a,{x:()=>DW(n,e,s)}}},BW={kernelName:li,inputsToSave:["a","b"],gradFunc:(e,t)=>{let[r,n]=t,a=bt(r.shape,n.shape);return{a:()=>{let s=pe(e,me(n,"float32")),i=Jt(r.shape,a);return i.length>0?U(ke(s,i),r.shape):s},b:()=>{let s=L(e,me(r,"float32")),i=Jt(n.shape,a);i.length>0&&(s=U(ke(s,i),n.shape));let o=At(n);return Dt(pe(s,me(o,"float32")))}}}},WW={kernelName:ad,inputsToSave:["x"],gradFunc:(e,t)=>{let[r]=t;return{x:()=>pe(e,Dt(At(r)))}}},VW={kernelName:Mi,inputsToSave:["x"],gradFunc:(e,t)=>{let[r]=t,n=L(Fl(r,6),Sh(r));return{x:()=>L(e,me(n,"float32"))}}},UW={kernelName:Ri,inputsToSave:["x"],gradFunc:(e,t)=>{let[r]=t;return{x:()=>L(e,me(Sh(r),"float32"))}}},GW={kernelName:ml,inputsToSave:["x"],gradFunc:(e,t)=>{let[r]=t;return{x:()=>U(e,r.shape)}}},jW={kernelName:$i,inputsToSave:["images"],gradFunc:(e,t,r)=>{let[n]=t,a={dy:e,images:n};return{images:()=>B.runKernel(bf,a,r)}}},HW={kernelName:sd,inputsToSave:["images"],gradFunc:(e,t,r)=>{let[n]=t,a={dy:e,images:n};return{images:()=>B.runKernel(xf,a,r)}}},qW={kernelName:gl,gradFunc:(e,t,r)=>{let{dims:n}=r,a=Hn(n,e.shape);return{x:()=>_n(e,a)}}},XW={kernelName:yl,gradFunc:e=>({x:()=>at(e)})},KW={kernelName:Fi,inputsToSave:["x"],gradFunc:(e,t)=>{let[r]=t;return{x:()=>Dt(pe(e,L(Ws(r,1.5),2)))}}},ZW={kernelName:xl,inputsToSave:["condition"],gradFunc:(e,t)=>{let[r]=t;return{condition:()=>me(at(r),"float32"),t:()=>L(e,me(r,e.dtype)),e:()=>L(e,me(Pf(r),e.dtype))}}},YW={kernelName:id,inputsToSave:["x"],gradFunc:(e,t)=>{let[r]=t;return{x:()=>{let n=mn(r,Se(0)),a=Se(jw),s=Se(Hw),i=L(e,s),o=L(L(e,a),Mn(me(r,"float32")));return Vr(n,i,o)}}}},JW={kernelName:_i,outputsToSave:[!0],gradFunc:(e,t)=>{let[r]=t;return{x:()=>L(e,L(r,ce(Se(1),r)))}}},QW={kernelName:od,gradFunc:e=>({x:()=>at(e)})},eV={kernelName:Pi,inputsToSave:["x"],gradFunc:(e,t)=>{let[r]=t;return{x:()=>L(Ef(me(r,"float32")),e)}}},tV={kernelName:vl,inputsToSave:["x"],gradFunc:(e,t)=>{let[r]=t;return{x:()=>L(Hy(me(r,"float32")),e)}}},rV={kernelName:bl,inputsToSave:["x"],gradFunc:(e,t,r)=>{let[n]=t,{begin:a,size:s}=r,i=n.shape,[o,l]=N7(n,a,s),u=[];for(let d=0;d<e.rank;d++)u.push([o[d],i[d]-o[d]-l[d]]);return{x:()=>Xn(e,u)}}},nV={kernelName:Di,outputsToSave:[!0],gradFunc:(e,t,r)=>{let[n]=t,{dim:a}=r,s=!0,i=L(e,n);return{logits:()=>ce(i,L(ke(i,[a],s),n))}}},aV={kernelName:ld,inputsToSave:["x"],gradFunc:(e,t)=>{let[r]=t;return{x:()=>L(e,Tr(r))}}},u4={kernelName:wl,gradFunc:(e,t,r)=>{let{blockShape:n,paddings:a}=r;return{x:()=>Nf(e,n,a)}}},d4={kernelName:kl,gradFunc:(e,t,r)=>{let{axis:n}=r;return{x:()=>St(e,n)}}},sV={kernelName:zi,inputsToSave:["x"],gradFunc:(e,t)=>{let[r]=t;return{x:()=>pe(e,L(Er(me(r,"float32")),2))}}},iV={kernelName:dd,inputsToSave:["x"],gradFunc:(e,t)=>{let[r]=t;return{x:()=>L(e,L(me(r,"float32"),2))}}},oV={kernelName:Li,inputsToSave:["a","b"],gradFunc:(e,t)=>{let[r,n]=t,a=Se(2);return{a:()=>L(e,L(a,ce(r,n))),b:()=>L(e,L(a,ce(n,r)))}}},lV={kernelName:Ui,gradFunc:e=>({x:()=>at(e)})},uV={kernelName:Bi,inputsToSave:["a","b"],gradFunc:(e,t)=>{let[r,n]=t,a=bt(r.shape,n.shape);return{a:()=>{let s=e,i=Jt(r.shape,a);return i.length>0&&(s=ke(s,i)),U(s,r.shape)},b:()=>{let s=e,i=Jt(n.shape,a);return i.length>0&&(s=ke(s,i)),U(Dt(s),n.shape)}}}},dV={kernelName:Oi,inputsToSave:["x"],gradFunc:(e,t,r)=>{let[n]=t,a=n.shape.slice(),{axis:s}=r;Hn(s,n.shape).forEach(l=>{a[l]=1});let i=U(e,a),o=L(i,hn(n.shape,"float32"));return{x:()=>o}}},pV={kernelName:Sl,inputsToSave:["x"],gradFunc:(e,t)=>{let[r]=t;return{x:()=>pe(e,At(Ef(r)))}}},hV={kernelName:Wi,outputsToSave:[!0],gradFunc:(e,t)=>{let[r]=t;return{x:()=>L(ce(Se(1),At(r)),e)}}},cV={kernelName:es,inputsToSave:["x"],gradFunc:(e,t,r)=>{let[n]=t,{reps:a}=r;return{x:()=>{let s=at(n);if(n.rank===1)for(let i=0;i<a[0];++i)s=le(s,Pe(e,[i*n.shape[0]],[n.shape[0]]));else if(n.rank===2)for(let i=0;i<a[0];++i)for(let o=0;o<a[1];++o)s=le(s,Pe(e,[i*n.shape[0],o*n.shape[1]],[n.shape[0],n.shape[1]]));else if(n.rank===3)for(let i=0;i<a[0];++i)for(let o=0;o<a[1];++o)for(let l=0;l<a[2];++l)s=le(s,Pe(e,[i*n.shape[0],o*n.shape[1],l*n.shape[2]],[n.shape[0],n.shape[1],n.shape[2]]));else if(n.rank===4)for(let i=0;i<a[0];++i)for(let o=0;o<a[1];++o)for(let l=0;l<a[2];++l)for(let u=0;u<a[3];++u)s=le(s,Pe(e,[i*n.shape[0],o*n.shape[1],l*n.shape[2],u*n.shape[3]],[n.shape[0],n.shape[1],n.shape[2],n.shape[3]]));else throw new Error(`Gradient for tile operation is not implemented for rank-${n.rank} tensors yet.`);return s}}}},fV={kernelName:Vi,gradFunc:(e,t,r)=>{let n=r,{perm:a}=n,s=Jy(a);return{x:()=>et(e,s)}}},mV={kernelName:Nl,gradFunc:(e,t,r)=>{let n=r,{axis:a}=n;return{value:()=>ur(e,a)}}},gV={kernelName:ch,inputsToSave:["segmentIds"],gradFunc:(e,t)=>{let[r]=t;return{x:()=>yV(e,r)}}};function yV(e,t){let r=ts(t,at(t)),n=Cu(e,r),a=Ml(t,Se(0,"int32")),s=n.rank-a.rank;for(let o=0;o<s;++o)a=Kt(a,o+1);a=ga(a,hn(n.shape,"bool"));let i=at(n);return Vr(a,n,i)}var AV={kernelName:El,gradFunc:e=>({x:()=>at(e)})},xV=[l6,gB,yB,AB,xB,bB,vB,wB,kB,IB,SB,CB,EB,MB,FB,PB,_B,zB,OB,DB,LB,BB,VB,WB,jB,HB,qB,XB,KB,ZB,BW,YB,JB,QB,eW,tW,nW,rW,aW,sW,iW,oW,lW,uW,dW,pW,hW,cW,fW,yW,o4,o4,AW,vW,IW,SW,CW,TW,NW,EW,RW,$W,MW,FW,PW,l4,l4,_W,zW,LW,WW,VW,UW,GW,jW,HW,qW,XW,KW,ZW,YW,JW,QW,eV,tV,rV,nV,aV,u4,u4,d4,d4,sV,oV,iV,lV,uV,dV,pV,hV,cV,fV,mV,gV,AV];for(let e of xV)Bv(e);var d6={};Le(d6,{maxNorm:()=>bV,minMaxNorm:()=>kV,nonNeg:()=>wV,unitNorm:()=>vV});function FA(e,t){return X(()=>Er(ke(L(e,e),t,!0)))}var $h=class extends ue.Serializable{getConfig(){return{}}},PA=class extends $h{constructor(e){super(),this.defaultMaxValue=2,this.defaultAxis=0,this.maxValue=e.maxValue!=null?e.maxValue:this.defaultMaxValue,this.axis=e.axis!=null?e.axis:this.defaultAxis}apply(e){return X(()=>{let t=FA(e,this.axis),r=cn(t,0,this.maxValue);return L(e,pe(r,le(sr(),t)))})}getConfig(){return{maxValue:this.maxValue,axis:this.axis}}};PA.className="MaxNorm";ue.registerClass(PA);var _A=class extends $h{constructor(e){super(),this.defaultAxis=0,this.axis=e.axis!=null?e.axis:this.defaultAxis}apply(e){return X(()=>pe(e,le(sr(),FA(e,this.axis))))}getConfig(){return{axis:this.axis}}};_A.className="UnitNorm";ue.registerClass(_A);var zA=class extends $h{apply(e){return Oa(e)}};zA.className="NonNeg";ue.registerClass(zA);var OA=class extends $h{constructor(e){super(),this.defaultMinValue=0,this.defaultMaxValue=1,this.defaultRate=1,this.defaultAxis=0,this.minValue=e.minValue!=null?e.minValue:this.defaultMinValue,this.maxValue=e.maxValue!=null?e.maxValue:this.defaultMaxValue,this.rate=e.rate!=null?e.rate:this.defaultRate,this.axis=e.axis!=null?e.axis:this.defaultAxis}apply(e){return X(()=>{let t=FA(e,this.axis),r=le(L(this.rate,cn(t,this.minValue,this.maxValue)),L(1-this.rate,t));return L(e,pe(r,le(sr(),t)))})}getConfig(){return{minValue:this.minValue,maxValue:this.maxValue,rate:this.rate,axis:this.axis}}};OA.className="MinMaxNorm";ue.registerClass(OA);var p4={maxNorm:"MaxNorm",minMaxNorm:"MinMaxNorm",nonNeg:"NonNeg",unitNorm:"UnitNorm"};function or(e){return wA(e)}function h4(e,t={}){return Th(e,ue.SerializationMap.getMap().classNameMap,t,"constraint")}function lr(e){if(e==null)return null;if(typeof e=="string"){let t={className:e in p4?p4[e]:e,config:{}};return h4(t)}else return e instanceof $h?e:h4(e)}function bV(e){return new PA(e)}function vV(e){return new _A(e)}function wV(){return new zA}function kV(e){return new OA(e)}var p6={};Le(p6,{constant:()=>CV,glorotNormal:()=>FV,glorotUniform:()=>MV,heNormal:()=>PV,heUniform:()=>_V,identity:()=>RV,leCunNormal:()=>zV,leCunUniform:()=>OV,ones:()=>SV,orthogonal:()=>DV,randomNormal:()=>NV,randomUniform:()=>TV,truncatedNormal:()=>EV,varianceScaling:()=>$V,zeros:()=>IV});function IV(){return new SA}function SV(){return new rm}function CV(e){return new CA(e)}function TV(e){return new TA(e)}function NV(e){return new NA(e)}function EV(e){return new EA(e)}function RV(e){return new RA(e)}function $V(e){return new an(e)}function MV(e){return new nm(e)}function FV(e){return new am(e)}function PV(e){return new sm(e)}function _V(e){return new im(e)}function zV(e){return new om(e)}function OV(e){return new lm(e)}function DV(e){return new $A(e)}var h6={};Le(h6,{Layer:()=>st,RNN:()=>as,RNNCell:()=>Ph,activation:()=>uG,add:()=>AG,alphaDropout:()=>tj,average:()=>xG,averagePooling1d:()=>Hx,averagePooling2d:()=>qx,averagePooling3d:()=>Xx,avgPool1d:()=>NG,avgPool2d:()=>RG,avgPool3d:()=>MG,avgPooling1d:()=>EG,avgPooling2d:()=>$G,avgPooling3d:()=>FG,batchNormalization:()=>SG,bidirectional:()=>qG,concatenate:()=>bG,conv1d:()=>eG,conv2d:()=>tG,conv2dTranspose:()=>rG,conv3d:()=>nG,conv3dTranspose:()=>aG,convLstm2d:()=>UG,convLstm2dCell:()=>GG,cropping2D:()=>iG,dense:()=>dG,depthwiseConv2d:()=>lG,dot:()=>IG,dropout:()=>pG,elu:()=>XU,embedding:()=>yG,flatten:()=>cG,gaussianDropout:()=>ej,gaussianNoise:()=>QG,globalAveragePooling1d:()=>PG,globalAveragePooling2d:()=>_G,globalMaxPool1d:()=>KG,globalMaxPool2d:()=>ZG,globalMaxPooling1d:()=>ok,globalMaxPooling2d:()=>lk,gru:()=>OG,gruCell:()=>DG,input:()=>R6,inputLayer:()=>qU,layerNormalization:()=>CG,leakyReLU:()=>ZU,lstm:()=>LG,lstmCell:()=>BG,masking:()=>rj,maxPool1d:()=>YG,maxPool2d:()=>JG,maxPooling1d:()=>uk,maxPooling2d:()=>dk,maxPooling3d:()=>zG,maximum:()=>vG,minimum:()=>wG,multiply:()=>kG,permute:()=>gG,prelu:()=>YU,reLU:()=>KU,repeatVector:()=>fG,reshape:()=>mG,rnn:()=>jG,separableConv2d:()=>sG,simpleRNN:()=>WG,simpleRNNCell:()=>VG,softmax:()=>JU,spatialDropout1d:()=>hG,stackedRNNCells:()=>HG,thresholdedReLU:()=>QU,timeDistributed:()=>XG,upSampling2d:()=>oG,zeroPadding2d:()=>TG});async function Ss(e){if(e==null)return;let t=[],r=[],n=[];for(let a in e){let s=e[a];if(typeof s!="number"){let i=s;t.push(i.data()),r.push(a),n.push(i)}}if(t.length>0){let a=await Promise.all(t);for(let s=0;s<a.length;++s)e[r[s]]=a[s][0];te(n)}}function c6(e){if(e!=null)for(let t in e){let r=e[t];typeof r!="number"&&r.dispose()}}var LV=125,Ru=class{constructor(){this.validationData=null}setParams(e){this.params=e}async onEpochBegin(e,t){}async onEpochEnd(e,t){}async onBatchBegin(e,t){}async onBatchEnd(e,t){}async onTrainBegin(e){}async onTrainEnd(e){}setModel(e){}},f6=class{constructor(e,t=10){e==null&&(e=[]),this.callbacks=e,this.queueLength=t}append(e){this.callbacks.push(e)}setParams(e){for(let t of this.callbacks)t.setParams(e)}setModel(e){for(let t of this.callbacks)t.setModel(e)}async onEpochBegin(e,t){t==null&&(t={});for(let r of this.callbacks)await r.onEpochBegin(e,t)}async onEpochEnd(e,t){t==null&&(t={});for(let r of this.callbacks)await r.onEpochEnd(e,t)}async onBatchBegin(e,t){t==null&&(t={});for(let r of this.callbacks)await r.onBatchBegin(e,t)}async onBatchEnd(e,t){t==null&&(t={});for(let r of this.callbacks)await r.onBatchEnd(e,t)}async onTrainBegin(e){e==null&&(e={});for(let t of this.callbacks)await t.onTrainBegin(e)}async onTrainEnd(e){e==null&&(e={});for(let t of this.callbacks)await t.onTrainEnd(e)}},BV=class extends Ru{constructor(){super()}async onEpochBegin(e){this.seen=0,this.totals={}}async onBatchEnd(e,t){t==null&&(t={});let r=t.size==null?0:t.size;this.seen+=r;for(let n in t){let a=t[n];if(typeof a=="number")this.totals.hasOwnProperty(n)||(this.totals[n]=0),this.totals[n]=this.totals[n]+a*r;else{let s;n in this.totals?s=this.totals[n]:this.totals[n]=0;let i=X(()=>le(this.totals[n],L(a,r)));this.totals[n]=i,s!=null&&s.dispose()}}}async onEpochEnd(e,t){if(t!=null)for(let r of this.params.metrics)this.totals[r]!=null&&(typeof this.totals[r]=="number"?t[r]=this.totals[r]/this.seen:X(()=>{let n=L(pe(1,this.seen),this.totals[r]);t[r]=n,this.totals[r].dispose(),gr(t[r])}))}},m6=class extends Ru{async onTrainBegin(e){this.epoch=[],this.history={}}async onEpochEnd(e,t){t==null&&(t={}),this.epoch.push(e);for(let r in t)this.history[r]==null&&(this.history[r]=[]),this.history[r].push(t[r])}async syncData(){let e=[],t=[],r=[];for(let a in this.history){let s=this.history[a];for(let i=0;i<s.length;++i)if(typeof s[i]!="number"){let o=s[i];e.push(o.data()),t.push(a),r.push(i)}}let n=await Promise.all(e);for(let a=0;a<n.length;++a)this.history[t[a]][r[a]].dispose(),this.history[t[a]][r[a]]=n[a][0]}},g6=class extends Ru{constructor(e,t){if(super(),this.currentEpoch=0,this.nowFunc=e.nowFunc,this.nextFrameFunc=e.nextFrameFunc||bA,this.yieldEvery=t||"auto",this.yieldEvery==="auto"&&(this.yieldEvery=LV),this.yieldEvery==="never"&&e.onYield!=null)throw new Error("yieldEvery is `never` but you provided an `onYield` callback. Either change `yieldEvery` or remove the callback");v.isNumber(this.yieldEvery)&&(this.maybeWait=OL(this.maybeWait.bind(this),this.yieldEvery,this.nowFunc)),this.trainBegin=e.onTrainBegin,this.trainEnd=e.onTrainEnd,this.epochBegin=e.onEpochBegin,this.epochEnd=e.onEpochEnd,this.batchBegin=e.onBatchBegin,this.batchEnd=e.onBatchEnd,this.yield=e.onYield}async maybeWait(e,t,r){let n=[];this.yield!=null&&(await Ss(r),n.push(this.yield(e,t,r))),n.push(this.nextFrameFunc()),await Promise.all(n)}async onEpochBegin(e,t){this.currentEpoch=e,this.epochBegin!=null&&(await Ss(t),await this.epochBegin(e,t))}async onEpochEnd(e,t){let r=[];this.epochEnd!=null&&(await Ss(t),r.push(this.epochEnd(e,t))),this.yieldEvery==="epoch"&&r.push(this.nextFrameFunc()),await Promise.all(r)}async onBatchBegin(e,t){this.batchBegin!=null&&(await Ss(t),await this.batchBegin(e,t))}async onBatchEnd(e,t){let r=[];this.batchEnd!=null&&(await Ss(t),r.push(this.batchEnd(e,t))),this.yieldEvery==="batch"?r.push(this.nextFrameFunc()):v.isNumber(this.yieldEvery)&&r.push(this.maybeWait(this.currentEpoch,e,t)),await Promise.all(r)}async onTrainBegin(e){this.trainBegin!=null&&(await Ss(e),await this.trainBegin(e))}async onTrainEnd(e){this.trainEnd!=null&&(await Ss(e),await this.trainEnd(e))}};function y6(e,t){return e==null&&(e={}),e instanceof Ru?[e]:Array.isArray(e)&&e[0]instanceof Ru?e:Ct(e).map(r=>new g6(r,t))}var Ca=class{constructor(){}static registerCallbackConstructor(e,t){v.assert(e>=0&&Number.isInteger(e),()=>`Verbosity level is expected to be an integer >= 0, but got ${e}`),Ca.checkForDuplicate(t),Ca.constructors[e]==null&&(Ca.constructors[e]=[]),Ca.constructors[e].push(t)}static checkForDuplicate(e){for(let t in Ca.constructors)Ca.constructors[+t].forEach(r=>{if(r===e)throw new H("Duplicate callback constructor.")})}static clear(){Ca.constructors={}}static createCallbacks(e){let t=[];for(let r in Ca.constructors){let n=+r;e>=n&&t.push(...Ca.constructors[n])}return t.map(r=>new r)}},DA=Ca;DA.constructors={};function A6(e,t,r,n,a,s,i,o,l){let u=new m6,d=[new BV,...DA.createCallbacks(t)];e!=null&&d.push(...e),d.push(u);let h=new f6(d);return h.setParams({epochs:r,initialEpoch:n,samples:a,steps:s,batchSize:i,verbose:t,doValidation:o,metrics:l}),{callbackList:h,history:u}}function fa(e,t={},r=!1){return Th(e,ue.SerializationMap.getMap().classNameMap,t,"layer",r)}function F0(e,t){return X(()=>{e.dtype!=="float32"&&(e=me(e,"float32"));let r=ke(Eh(e),t,!0),n=cd(r.shape,sr()),a=Er(ts(r,n));return pe(e,a)})}function zl(e,t){return X(()=>Vt(Eh(ce(t,e)),-1))}function dm(e,t){return X(()=>Vt(ar(ce(t,e)),-1))}function bd(e,t){return X(()=>{let r=ce(e,t),n=cn(ar(e),sr(),Number.MAX_VALUE),a=ar(pe(r,n));return L(100,Vt(a,-1))})}function WV(e,t){return X(()=>{let r=cn(t,sr(),Number.MAX_VALUE),n=Fn(le(1,r)),a=cn(e,sr(),Number.MAX_VALUE),s=Fn(le(1,a));return Vt(Eh(ce(n,s)),-1)})}function VV(e,t){return X(()=>{let r=ts(0,ce(1,L(e,t)));return Vt(Eh(r),-1)})}function UV(e,t){return X(()=>{let r=ts(0,ce(1,L(e,t)));return Vt(r,-1)})}function GV(e,t){return X(()=>{let r=ke(L(e,t),-1),n=Ar(L(ce(1,e),t),-1);return ts(0,le(1,ce(n,r)))})}function jV(e,t){return X(()=>{let r=Math.log(2),n=ce(t,e),a=ce(le(n,fd(L(-2,n))),r);return Vt(a,-1)})}function jp(e,t,r=!1){return X(()=>{if(r)t=yd(t);else{let n=ke(t,t.shape.length-1,!0);t=pe(t,n)}return t=cn(t,sr(),1-sr()),Dt(ke(L(me(e,"float32"),Fn(t)),t.shape.length-1))})}function P0(e,t,r=!1){return X(()=>{let n=me(kh(KL(e)),"int32");t=cn(t,sr(),1-sr());let a=t.shape,s=U(Vp(n,a[a.length-1]),a);return jp(s,t,r)})}function HV(e,t){if(!v.arraysEqual(e.shape,t.shape))throw new H(`logits and labels must have the same shape, but got shapes ${JSON.stringify(e.shape)} and ${JSON.stringify(t.shape)}`);return X(()=>{let r=Oa(t),n=Dt(ar(t));return le(ce(r,L(t,e)),Mf(Mn(n)))})}function pm(e,t){return X(()=>{let r;return r=cn(t,sr(),1-sr()),r=Fn(pe(r,ce(1,r))),Vt(HV(e,r),-1)})}function qV(e,t){return X(()=>{let r=cn(e,sr(),1),n=cn(t,sr(),1);return ke(L(e,Fn(pe(r,n))),-1)})}function XV(e,t){return X(()=>{let r=Fn(le(sr(),t));return Vt(ce(t,L(e,r)),-1)})}function LA(e,t){return X(()=>{let r=F0(e,-1),n=F0(t,-1),a=L(r,n);return Dt(ke(a,-1))})}var _0={meanSquaredError:zl,meanAbsoluteError:dm,meanAbsolutePercentageError:bd,meanSquaredLogarithmicError:WV,squaredHinge:VV,hinge:UV,categoricalHinge:GV,logcosh:jV,categoricalCrossentropy:jp,sparseCategoricalCrossentropy:P0,binaryCrossentropy:pm,kullbackLeiblerDivergence:qV,poisson:XV,cosineProximity:LA};function lg(e){if(typeof e=="string"){if(e in _0)return _0[e];let t=`Unknown loss ${e}`;throw e.toLowerCase().includes("softmaxcrossentropy")&&(t=`Unknown loss ${e}. Use "categoricalCrossentropy" as the string name for tf.losses.softmaxCrossEntropy`),new H(t)}else return e}function BA(e,t){return X(()=>{let r=L(.5,Pn(t)),n=em(mn(t,r),e.dtype);return Vt($n(e,n),-1)})}function WA(e,t){return X(()=>em($n(Rn(e,-1),Rn(t,-1)),"float32"))}function x6(e,t){return X(()=>me(ke(ga($n(e,1),$n(t,1))),"float32"))}function KV(e,t){return X(()=>me(ke(ga($n(e,1),$n(t,0))),"float32"))}function ZV(e,t){return X(()=>me(ke(ga($n(e,0),$n(t,1))),"float32"))}function b6(e,t){return X(()=>{let r=x6(e,t),n=ZV(e,t),a=le(r,n);return me(Vr(mn(a,0),pe(r,a),0),"float32")})}function YV(e,t){return X(()=>{let r=x6(e,t),n=KV(e,t),a=le(r,n);return me(Vr(mn(a,0),pe(r,a),0),"float32")})}function v6(e,t){return pm(e,t)}function w6(e,t){return e.rank===t.rank&&(e=tt(e,[e.rank-1])),t=Rn(t,-1),t.dtype!==e.dtype&&(t=me(t,e.dtype)),me($n(e,t),"float32")}var JV=zl,QV=zl,eU=dm,tU=dm,rU=bd,nU=bd,VA=jp,aU=LA,k6=P0,z0={binaryAccuracy:BA,categoricalAccuracy:WA,precision:b6,categoricalCrossentropy:VA,sparseCategoricalCrossentropy:k6,mse:JV,MSE:QV,mae:eU,MAE:tU,mape:rU,MAPE:nU,cosine:aU};function sU(e){if(typeof e=="string"&&e in z0)return z0[e];if(typeof e!="string"&&e!=null)return e;throw new H(`Unknown metric ${e}`)}function Yc(e){if(Na(e!==null,`Unknown LossOrMetricFn ${e}`),typeof e=="string")return e;{let t;for(let r of Object.keys(_0))if(_0[r]===e){t=r;break}if(t!==void 0)return t;for(let r of Object.keys(z0))if(z0[r]===e){t=r;break}return t!==void 0?t:e.name}}function iU(e){let t={Adagrad:()=>bo.adagrad(.01),Adadelta:()=>bo.adadelta(1,.95,sr()),Adam:()=>bo.adam(.001,.9,.999,sr()),Adamax:()=>bo.adamax(.002,.9,.999,sr(),0),RMSProp:()=>bo.rmsprop(.001,.9,0,sr()),SGD:()=>bo.sgd(.01)};if(t.adagrad=t.Adagrad,t.adadelta=t.Adadelta,t.adam=t.Adam,t.adamax=t.Adamax,t.rmsprop=t.RMSProp,t.sgd=t.SGD,e in t)return t[e]();throw new H(`Unknown Optimizer ${e}`)}var c4=1*1024*1024;function f4(e,t,r=!1){if(e==null||typeof e!="object"||Object.getPrototypeOf(e)!==Object.prototype||!Pg(e))throw new Error("User-defined metadata is expected to be a JSON object, but is not.");if(r){let n=JSON.stringify(e);n.length>c4&&console.warn(`User-defined metadata of model "${t}" is too large in size (length=${n.length} when serialized). It is not recommended to store such large objects in user-defined metadata. Please make sure its serialized length is <= ${c4}.`)}}function Pg(e){if(e===null)return!0;if(typeof e=="object")if(Object.getPrototypeOf(e)===Object.prototype){let t=Object.keys(e);for(let r of t)if(typeof r!="string"||!Pg(e[r]))return!1;return!0}else if(Array.isArray(e)){for(let t of e)if(!Pg(t))return!1;return!0}else return!1;else{let t=typeof e;return t==="string"||t==="number"||t==="boolean"}}function oU(e,t,r,n=console.log){let a=uU(e),s=["Layer (type)","Input Shape","Output shape","Param #"];a?(t=t||90,r=r||[.32,.61,.89,1]):(t=t||115,r=r||[.24,.48,.7,.8,1]),r[r.length-1]<=1&&(r=r.map(d=>Math.floor(t*d)));let i;if(!a){s.push("Receives inputs"),i=[];for(let d in e.nodesByDepth)i.push(...e.nodesByDepth[d])}n("_".repeat(t)),O0(s,r,n),n("=".repeat(t));let o=e.layers;for(let d=0;d<o.length;++d)a?dU(o[d],r,n):pU(o[d],r,i,n),n((d===o.length-1?"=":"_").repeat(t));e.checkTrainableWeightsConsistency();let l=lU(e),u=R0(e.nonTrainableWeights);n(`Total params: ${l+u}`),n(`Trainable params: ${l}`),n(`Non-trainable params: ${u}`),n("_".repeat(t))}function lU(e){let t;return e.collectedTrainableWeights!=null?t=R0(e.collectedTrainableWeights):t=R0(e.trainableWeights),t}function uU(e){let t=!0,r=[],n=[];for(let a in e.nodesByDepth)r.push(e.nodesByDepth[a]);for(let a of r){if(a.length>1||a.length===1&&a[0].inboundLayers.length>1){t=!1;break}n.push(...a)}if(t)for(let a of e.layers){let s=!1;for(let i of a.inboundNodes)if(n.indexOf(i)!==-1)if(s){t=!1;break}else s=!0;if(!t)break}return t}function O0(e,t,r=console.log){let n="";for(let a=0;a<e.length;++a)a>0&&(n=n.slice(0,n.length-1)+" "),n+=e[a],n=n.slice(0,t[a]),n+=" ".repeat(t[a]-n.length);r(n)}function dU(e,t,r){let n,a;try{a=e.inboundNodes.map(l=>JSON.stringify(l.inputShapes)).join(",")}catch(l){a="multiple"}try{n=JSON.stringify(e.outputShape)}catch(l){n="multiple"}let s=e.name,i=e.getClassName(),o=[`${s} (${i})`,a,n,e.countParams().toString()];O0(o,t,r)}function pU(e,t,r,n){let a,s;try{s=e.inboundNodes.map(h=>JSON.stringify(h.inputShapes)).join(",")}catch(h){s="multiple"}try{a=JSON.stringify(e.outputShape)}catch(h){a="multiple"}let i=[];for(let h of e.inboundNodes)if(!(r!=null&&r.length>0&&r.indexOf(h)===-1))for(let p=0;p<h.inboundLayers.length;++p){let c=h.inboundLayers[p].name,f=h.nodeIndices[p],m=h.tensorIndices[p];i.push(`${c}[${f}][${m}]`)}let o=e.name,l=e.getClassName(),u=i.length===0?"":i[0],d=[`${o} (${l})`,s,a,e.countParams().toString(),u];O0(d,t,n);for(let h=1;h<i.length;++h)O0(["","","","",i[h]],t,n)}function I6(e,t,r){return(e==="inboundNodes"||e==="outputLayers"||e==="inputLayers")&&t===0&&typeof r=="string"}function Hp(e,t){if(e===null)return null;if(typeof e=="string")return Io(e);if(typeof e=="number"||typeof e=="boolean")return e;if(e instanceof Array){let r=[],n=e.length;for(let a=0;a<n;++a){let s=e[a];I6(t,a,s)?r.push(s):r.push(Hp(s,t))}return r}else{let r={};for(let n of Object.keys(e)){let a=e[n];if(n==="name"&&typeof a=="string")r[n]=a;else{let s=Io(n);r[s]=Hp(a,s)}}return r}}function _g(e,t){if(e==null)return null;if(typeof e=="string")return ja(e);if(typeof e=="number"||typeof e=="boolean")return e;if(e instanceof Array){let r=[],n=e.length;for(let a=0;a<n;++a){let s=e[a];I6(t,a,s)?r.push(s):r.push(_g(s,t))}return r}else{let r={};for(let n of Object.keys(e)){let a=e[n],s=ja(n);(n==="name"||n==="className")&&typeof a=="string"?r[s]=a:r[s]=_g(a,n)}return r}}var UA="0.0.0",Ta=class extends st{constructor(e){if(super({}),this.containerNodes=new Set,this.name=e.name,this.name==null){let y=this.getClassName().toLowerCase();this.name=Qf(y)}if(this.supportsMasking=!1,this.trainable_=!0,Array.isArray(e.inputs)?this.inputs=e.inputs.slice():this.inputs=[e.inputs],Array.isArray(e.outputs)?this.outputs=e.outputs.slice():this.outputs=[e.outputs],Rs(this.inputs).length!==this.inputs.length)throw new H(`The list of inputs passed to the model is redundant. All inputs should only appear once. Found: ${this.inputs.map(y=>y.name)}`);Rs(this.outputs).length!==this.outputs.length&&console.warn(`The list of outputs passed to the model is redundant. All outputs should only appear once. Found: ${this.outputs.map(y=>y.name)}`),this.inputLayers=[],this.inputLayersNodeIndices=[],this.inputLayersTensorIndices=[],this.outputLayers=[],this.outputLayersNodeIndices=[],this.outputLayersTensorIndices=[],this.layers=[],this.internalContainerRefs=[];for(let y of this.outputs){let A=y.sourceLayer,x=y.nodeIndex,b=y.tensorIndex;this.outputLayers.push(A),this.outputLayersNodeIndices.push(x),this.outputLayersTensorIndices.push(b)}for(let y of this.inputs){let A=y.sourceLayer,x=y.nodeIndex,b=y.tensorIndex;Na(x===0,"input layer has >1 nodes"),Na(b===0,"input layer has >1 tensors"),this.inputLayers.push(A),this.inputLayersNodeIndices.push(x),this.inputLayersTensorIndices.push(b)}this.inputNames=[],this.outputNames=[],this.feedInputShapes=[],this.feedInputNames=[],this.feedOutputNames=[];for(let y=0;y<this.inputLayers.length;y++){let A=this.inputLayers[y];if(!(A instanceof xd))throw new TypeError(`Input layers to a LayersModel must be InputLayer objects. Received inputs: ${e.inputs}. Input ${y} (0-based) originates from layer type ${A.getClassName()}.`);this.inputNames.push(A.name),this.feedInputShapes.push(A.batchInputShape),this.feedInputNames.push(A.name)}for(let y of this.outputLayers)this.outputNames.push(y.name);this.internalInputShapes=this.inputs.map(y=>y.shape),this.internalOutputShapes=this.outputs.map(y=>y.shape);let t={},r={},n={},a={},s={},i=[],o=(y,A,x,b,w,S)=>{(b==null||w==null||S==null)&&(b=y.sourceLayer,w=y.nodeIndex,S=y.tensorIndex);let N=b.inboundNodes[w];if(x.indexOf(N)!==-1)throw new da(`The tensor ${y.name} at layer "${b.name}" is part of a cycle.`);if(A.indexOf(N)!==-1)return;this.containerNodes.add(Ta.nodeKey(b,w)),b.id in s||(s[b.id]=Object.keys(s).length),x.indexOf(N)===-1&&x.push(N);let E=N.inboundLayers.length;for(let R=0;R<E;R++){let z=N.inputTensors[R],M=N.inboundLayers[R],I=N.nodeIndices[R],_=N.tensorIndices[R];o(z,A,x,M,I,_)}for(A.push(N);x.indexOf(N)>=0;)x.splice(x.indexOf(N),1);i.push(N)},l=[],u=[];for(let y of this.outputs)o(y,l,u);let d=i.slice().reverse();for(let y of d){r[y.id]=y,y.id in t||(t[y.id]=0);let A=t[y.id],x=n[y.outboundLayer.id]==null?0:n[y.outboundLayer.id];A=Math.max(A,x),n[y.outboundLayer.id]=A,a[y.outboundLayer.id]=y.outboundLayer,t[y.id]=A;for(let b=0;b<y.inboundLayers.length;b++){let w=y.inboundLayers[b],S=y.nodeIndices[b],N=w.inboundNodes[S],E=t[N.id]==null?0:t[N.id];t[N.id]=Math.max(A+1,E),r[N.id]=N}}let h={};for(let y in t){let A=t[y];A in h||(h[A]=[]),h[A].push(r[y])}let p={};for(let y in n){let A=n[y];A in p||(p[A]=[]),p[A].push(a[y])}let c=Object.keys(p).map(y=>parseInt(y,10)).sort(Xc);this.layers=[];for(let y of c){let A=p[y];A.sort((x,b)=>{let w=s[x.id],S=s[b.id];return w<S?-1:w>S?1:0});for(let x of A)x instanceof Ta&&this.internalContainerRefs.push(x),this.layers.push(x)}this.layersByDepth=p,c=Object.keys(h).map(y=>parseInt(y,10)).sort(Xc);let f=this.inputs.slice(),m=[];for(let y of c)for(let A of h[y]){let x=A.outboundLayer;if(x!=null){for(let b of A.inputTensors)if(f.indexOf(b)===-1)throw new da(`Graph disconnected: cannot obtain value for tensor ${b} at layer "${x.name}". The following previous layers were accessed without issue: ${m}`);for(let b of A.outputTensors)f.push(b);m.push(x.name)}}this.nodesByDepth=h;let g=this.layers.map(y=>y.name);for(let y of g){let A=g.filter(x=>x===y).length;if(A!==1)throw new da(`The name "${y}" is used ${A} times in the model. All layer names should be unique. Layer names: `+JSON.stringify(g))}this.outboundNodes=[],this.inboundNodes=[],new um({outboundLayer:this,inboundLayers:[],nodeIndices:[],tensorIndices:[],inputTensors:this.inputs,outputTensors:this.outputs,inputMasks:this.inputs.map(y=>null),outputMasks:this.outputs.map(y=>null),inputShapes:this.inputs.map(y=>y.shape),outputShapes:this.outputs.map(y=>y.shape)}),this.built=!0,this._refCount=1}assertNotDisposed(){if(this._refCount===0)throw new Error(`Container '${this.name}' is already disposed.`)}dispose(){this.assertNotDisposed();let e={refCountAfterDispose:null,numDisposedVariables:0};if(--this._refCount===0){for(let t of this.layers)e.numDisposedVariables+=t.dispose().numDisposedVariables;for(let t of this.internalContainerRefs)e.numDisposedVariables+=t.dispose().numDisposedVariables}return e.refCountAfterDispose=this._refCount,e}get trainable(){return this.trainable_}set trainable(e){this.layers.forEach(t=>{t._trainableWeights.forEach(r=>r.trainable=e)}),this.trainable_=e}get trainableWeights(){if(this._trainableWeights.length>0)throw new H("Container instance unexpectedly contains _trainableWeights.The trainable weights of a Container are a union of the trainable weights of its consituent Layers. Its own _trainableWeights must remain an empty Array.");if(!this.trainable)return[];let e=[];for(let t of this.layers)e=e.concat(t.trainableWeights);return e}get nonTrainableWeights(){let e=[];for(let t of this.layers)e.push(...t.nonTrainableWeights);if(!this.trainable){let t=[];for(let r of this.layers)t.push(...r.trainableWeights);return t.concat(e)}return e}get weights(){return this.trainableWeights.concat(this.nonTrainableWeights)}loadWeights(e,t=!0){let r={},n=0;for(let s of this.layers)for(let i of s.weights){if(r[i.originalName]!=null)throw new H(`Duplicate weight name: ${i.originalName}`);r[i.originalName]=i,n++}let a=[];for(let s in e){let i=s;if(r[s]==null){let o=s.split("/");i=o.slice(0,-2).concat([o[o.length-1]]).join("/")}if(r[i]!=null)a.push([r[i],e[s]]);else if(t)throw new H(`Provided weight data has no target variable: ${s}`);delete r[i]}if(t){let s=[];for(let i in r)s.push(i);if(s.length>0)throw new H(`${s.length} of ${n} weights are not set: ${s}`)}MA(a)}updatedConfig(){let e=this.getConfig(),t={};return t.className=this.getClassName(),t.config=e,t.kerasVersion=`tfjs-layers ${UA}`,t.backend="TensorFlow.js",t}toJSON(e,t=!0){let r=_g(this.updatedConfig());return t?JSON.stringify(r):r}call(e,t){return X(()=>{e=Ct(e);let r=new To;for(let n=0;n<this.inputs.length;++n)r.add(this.inputs[n],e[n]);return Ip(this.outputs,r,t)})}computeMask(e,t){return X(()=>{e=Ct(e);let r;return t==null?r=Do(null,e.length):r=Ct(t),this.runInternalGraph(e,r)[1]})}computeOutputShape(e){let t=E0(e);if(t.length!==this.inputLayers.length)throw new H(`Invalid inputShape argument ${e}: model has ${this.inputLayers.length} tensor inputs.`);let r={};for(let i=0;i<t.length;i++){let o=this.inputLayers[i],l=t[i],u=o.name+"_0_0";r[u]=l}let n=Object.keys(this.nodesByDepth).map(i=>parseInt(i,10)).sort(Xc);if(n.length>1)for(let i of n){let o=this.nodesByDepth[i];for(let l of o){let u=l.outboundLayer;if(this.inputLayers.map(f=>f.id).indexOf(u.id)!==-1)continue;let d=[];for(let f=0;f<l.inboundLayers.length;f++){let m=l.inboundLayers[f],g=l.nodeIndices[f],y=l.tensorIndices[f],A=`${m.name}_${g}_${y}`,x=r[A];d.push(x)}let h=u.computeOutputShape(tn(d)),p=E0(h),c=u.inboundNodes.indexOf(l);for(let f=0;f<p.length;f++){let m=`${u.name}_${c}_${f}`;r[m]=p[f]}}}let a=[],s=[];for(let i=0;i<this.outputLayers.length;i++){let o=this.outputLayers[i],l=this.outputLayersNodeIndices[i],u=this.outputLayersTensorIndices[i],d=`${o.name}_${l}_${u}`;s.push(d)}for(let i=0;i<s.length;i++){let o=s[i];Na(o in r),a.push(r[o])}return tn(a)}runInternalGraph(e,t){t==null&&(t=Do(null,e.length));let r={};for(let o=0;o<this.inputs.length;++o){let l=this.inputs[o],u=e[o],d=t[o];r[l.id]=[u,d]}let n=Object.keys(this.nodesByDepth).map(o=>parseInt(o,10)).sort(Xc);for(let o of n){let l=this.nodesByDepth[o];for(let u of l){let d=u.outboundLayer,h=u.inputTensors,p=u.outputTensors,c=new Array;for(let f of h)f.id in r&&c.push(r[f.id]);if(c.length===h.length){let f={},m,g,y,A;if(u.callArgs!=null&&(f=u.callArgs),c.length===1){let[x,b]=c[0];f.mask==null&&(f.mask=b),y=Ct(d.call(x,f)),A=Ct(d.computeMask(x,b)),m=[x],g=[b]}else m=c.map(x=>x[0]),g=c.map(x=>x[1]),f.mask==null&&(f.mask=g),y=Ct(d.call(m,f)),A=Ct(d.computeMask(m,g));if(d.activityRegularizer)throw new Ve("LayersModel invocation with concrete Tensor value(s) in the presence of activity regularizer(s) is not supported yet.");for(let x=0;x<p.length;++x){let b=p[x],w=y[x],S=A[x];r[b.id]=[w,S]}}}}let a=[],s=[],i=[];for(let o of this.outputs){Na(o.id in r,`Could not compute output ${o.name} : ${o.id}`);let[l,u]=r[o.id];i.push(l.shape),a.push(l),s.push(u)}return[a,s,i]}buildNodeConversionMap(e){let t={},r;for(let n of this.layers){r=n instanceof Ta?1:0;for(let a=0;a<n.inboundNodes.length;a++){let s=Ta.nodeKey(n,a);this.containerNodes.has(s)&&(t[s]=r,r+=1)}}return t}getLayer(e,t){if(t!=null){if(this.layers.length<=t)throw new H(`Was asked to retrieve layer at index ${t}, but model only has ${this.layers.length} layer(s).`);return this.layers[t]}else if(e==null)throw new H("Provide either a layer name or layer index");for(let r of this.layers)if(r.name===e)return r;throw new H(`No such layer: ${e}`)}calculateLosses(){return X(()=>{let e=[];for(let t of this.layers)for(let r=0;r<t.inboundNodes.length;++r){let n=Ta.nodeKey(t,r);this.containerNodes.has(n)&&e.push(...t.calculateLosses())}return e})}getConfig(){let e={name:this.name},t=this.buildNodeConversionMap(this.layers),r=[];for(let s of this.layers){let i=s.getClassName(),o=s.getConfig(),l=[];for(let d=0;d<s.inboundNodes.length;d++){let h=s.inboundNodes[d],p=Ta.nodeKey(s,d),c={};if(this.containerNodes.has(p)){if(h.callArgs)try{JSON.stringify(h.callArgs),c=h.callArgs}catch(f){console.warn(`Layer ${s.name} was passed non-serializable keyword arguments: ${h.callArgs}. They will not be included in the serialized model (and thus will be missing at deserialization time).`),c={}}if(h.inboundLayers.length>0){let f=[];for(let m=0;m<h.inboundLayers.length;m++){let g=h.inboundLayers[m],y=h.nodeIndices[m],A=h.tensorIndices[m],x=Ta.nodeKey(g,y),b=t[x];b==null&&(b=0),f.push([g.name,b,A,c])}l.push(f)}}}let u={};u.name=s.name,u.className=i,u.config=o,u.inboundNodes=l,r.push(u)}e.layers=r;let n=[];for(let s=0;s<this.inputLayers.length;s++){let i=this.inputLayers[s],o=this.inputLayersNodeIndices[s],l=Ta.nodeKey(i,o);if(!this.containerNodes.has(l))continue;let u=t[l];u==null&&(u=0);let d=this.inputLayersTensorIndices[s];n.push([i.name,u,d])}e.inputLayers=n;let a=[];for(let s=0;s<this.outputLayers.length;s++){let i=this.outputLayers[s],o=this.outputLayersNodeIndices[s],l=Ta.nodeKey(i,o);if(!this.containerNodes.has(l))continue;let u=t[l];u==null&&(u=0);let d=this.outputLayersTensorIndices[s];a.push([i.name,u,d])}return e.outputLayers=a,e}static fromConfig(e,t,r={},n=!1){let a={},s={};function i(m,g){m.name in s?s[m.name].push(g):s[m.name]=[g]}function o(m,g){let y=[],A;for(let x of g){let b=x[0],w=x[1],S=x[2];if(A=x[3]==null?{}:x[3],!(b in a)){i(m,g);return}let N=a[b];if(N.inboundNodes.length<=w){i(m,g);return}let E=N.inboundNodes[w];y.push(E.outputTensors[S])}y.length>0&&m.apply(tn(y),A)}function l(m){let g=m.name,y=fa(m,t.customObjects!=null?t.customObjects:{});y.setFastWeightInitDuringBuild(n),a[g]=y,m.inboundNodes.forEach(A=>{if(!(A instanceof Array))throw new H(`Corrupted configuration, expected array for nodeData: ${A}`);i(y,A)})}let u=t.name,d=t.layers;for(let m of d)l(m);for(;!zL(s);)for(let m of d){let g=a[m.name];if(g.name in s){let y=s[g.name];delete s[g.name];for(let A of y)o(g,A)}}let h=[],p=[],c=t.inputLayers;for(let m of c){let g=m[0],y=m[1],A=m[2];Na(g in a);let x=a[g].inboundNodes[y].outputTensors;h.push(x[A])}let f=t.outputLayers;for(let m of f){let g=m[0],y=m[1],A=m[2];Na(g in a);let x=a[g].inboundNodes[y].outputTensors;p.push(x[A])}return new e({inputs:h,outputs:p,name:u})}get stateful(){if(this._stateful)throw new H("Container instance unexpectedly has _stateful = true. The statefulness of a Container is determined by the Layers it contains. Its _stateful property must remain the default false.");for(let e of this.layers)if(e.stateful)return!0;return!1}resetStates(){X(()=>{this.layers.forEach(e=>{e.stateful&&e.resetStates()})})}};function hU(e,t,r){let n=t.length;if(e==null||Array.isArray(e)&&e.length===0)return t.map(a=>null);if(n===1)return Array.isArray(e)&&e.length===1?e:typeof e=="object"&&t[0]in e?[e[t[0]]]:[e];if(Array.isArray(e)){if(e.length!==n)throw new Error(`Provided ${r} is an array of ${e.length} element(s), but the model has ${n} outputs. Make sure a set of weights is provided for each model output.`);return e}else if(typeof e=="object"&&Object.keys(e).length>0&&typeof e[Object.keys(e)[0]]=="object"){let a=[];return t.forEach(s=>{s in e?a.push(e[s]):a.push(null)}),a}else throw new Error(`The model has multiple (${n}) outputs, so ${r} must be either an array with ${n} elements or an object with ${t} keys. Provided ${r} not understood: ${JSON.stringify(e)}`)}function S6(e,t){return hU(e,t,"classWeight")}async function C6(e,t,r,n){if(t!=null||n!=null)throw new Error("Support sampleWeight is not implemented yet");if(r!=null){let a=X(()=>{if(e.shape.length===1)return Wr(e);if(e.shape.length===2){if(e.shape[1]>1)return Rn(e,1);if(e.shape[1]===1)return U(e,[e.shape[0]]);throw new Error(`Encountered unexpected last-dimension size (${e.shape[1]}) during handling of class weights. The size is expected to be >= 1.`)}else throw new Error(`Unexpected rank of target (y) tensor (${e.rank}) during handling of class weights. The rank is expected to be 1 or 2.`)}),s=Array.from(await a.data());te(a);let i=[];return s.forEach(o=>{if(r[o]==null)throw new Error(`classWeight must contain all classes in the training data. The class ${o} exists in the data but not in classWeight`);i.push(r[o])}),Tt(i,"float32")}else return null}function cU(e,t){return L(e,t)}var fU=32;function T6(e,t){let r,n,a=t;r=a.xs,n=a.ys,v.assert(r!=null&&n!=null,()=>`A Dataset iterator for fitDataset() is expected to generate objects of the form \`{xs: xVal, ys: yVal}\`, where the two values may be \`tf.Tensor\`, an array of Tensors, or a map of string to Tensor. The provided Dataset instead generates ${t}`);let s=m4("input",e.inputNames,r),i=m4("output",e.outputNames,n),o=s[0].shape[0];v.assert(s.length===e.inputs.length,()=>`LayersModel has ${e.inputs.length} inputs, but the dataset provides ${s.length} inputs. (Expected input keys: ${JSON.stringify(e.inputNames)})`),v.assert(i.length===e.outputs.length,()=>`LayersModel has ${e.outputs.length} outputs, but the dataset provides ${i.length} outputs. (Expected output keys: ${JSON.stringify(e.outputNames)})`);for(let l=0;l<s.length;l++)v.assert(s[l].shape[0]===o,()=>`Batch size mismatch: input ${e.inputNames[l]} has ${s[l].shape[0]}; expected ${o} based on input ${e.inputNames[0]}.`);for(let l=0;l<i.length;l++)v.assert(i[l].shape[0]===o,()=>`Batch size mismatch: output ${e.outputNames[l]} has ${i[l].shape[0]}; expected ${o} based on input ${e.inputNames[0]}.`);return{xs:s,ys:i}}function m4(e,t,r){if(r instanceof nt)return[r];if(Array.isArray(r))return v.assert(r.length===t.length,()=>`Received an array of ${r.length} Tensors, but expected ${t.length} to match the ${e} keys ${t}.`),r;{let n=[];for(let a of t){if(r[a]==null)throw new H(`The feature data generated by the dataset lacks the required ${e} key '${a}'.`);n.push(r[a])}return n}}function mU(e){if(e.length===3)throw new Ve("Validation with sample weights is not implemented yet.");return{xs:e[0],ys:e[1]}}async function gU(e,t,r){let n=r.batchesPerEpoch!=null;if(v.assert(e.optimizer!=null,()=>"You must compile a model before training/testing. Use LayersModel.compile(modelCompileConfig)."),v.assert(r!=null,()=>"For fitDataset(), the 2nd argument (config) is required, but it is not provided in this call."),v.assert(r.epochs!=null&&r.epochs>0&&Number.isInteger(r.epochs),()=>`For fitDataset(), config.epochs is expected to be a positive integer, but got ${r.epochs}`),v.assert(!n||r.batchesPerEpoch>0&&Number.isInteger(r.batchesPerEpoch),()=>`For fitDataset(), config.batchesPerEpoch is expected to be a positive integer if specified, but got ${r.batchesPerEpoch}`),v.assert(r.validationSplit==null,()=>"`validationSplit` is not supported by `fitDataset()`. Use validationData instead."),e.isTraining)throw new Error("Cannot start training because another fit() call is ongoing.");e.isTraining=!0;try{let a=r.validationData!=null,s,i;if(a)if(g4(r.validationData))v.assert(r.validationBatches==null||r.validationBatches>0&&Number.isInteger(r.validationBatches),()=>`For fitDataset() with dataset-based validation, config.validationBatches is expected not to be provided, or to be a positive integer, but got ${r.validationBatches}`);else{let g=mU(r.validationData);s=g.xs,i=g.ys}let o=e.makeTrainFunction(),l=e.getDedupedMetricsNames(),u;a?u=l.slice().concat(l.map(g=>"val_"+g)):u=l.slice();let d=y6(r.callbacks,r.yieldEvery),h=r.verbose==null?1:r.verbose,{callbackList:p,history:c}=A6(d,h,r.epochs,null,null,yU(t,r),null,a,u);p.setModel(e),e.history=c,await p.onTrainBegin(),e.stopTraining_=!1;let f=r.initialEpoch==null?0:r.initialEpoch,m=await t.iterator();for(;f<r.epochs;){let g={};await p.onEpochBegin(f);let y=0,A=0;for(n||(m=await t.iterator());!n||y<r.batchesPerEpoch;){let x=await m.next();if(n&&x.done){console.warn(`You provided \`batchesPerEpoch\` as ${r.batchesPerEpoch}, but your dataset iterator ran out of data after ${y} batches; interrupting training. Make sure that your dataset can generate at least \`batchesPerEpoch * epochs\` batches (in this case, ${r.batchesPerEpoch*r.epochs} batches). You may need to use the repeat() function when building your dataset.`);break}if(x.value!=null){let{xs:b,ys:w}=T6(e,x.value),S={};S.batch=A,S.size=b[0].shape[0],await p.onBatchBegin(A,S);let N=[];if(r.classWeight!=null){let z=S6(r.classWeight,e.outputNames);for(let M=0;M<z.length;++M)N.push(await C6(w[M],null,z[M]))}let E=b.concat(w).concat(N),R=o(E);te(E);for(let z=0;z<l.length;++z){let M=l[z],I=R[z];S[M]=I,gr(I)}await p.onBatchEnd(A,S),c6(S),A++,y++}if(n?y>=r.batchesPerEpoch:x.done){if(a){let b;g4(r.validationData)?b=Ct(await e.evaluateDataset(r.validationData,{batches:r.validationBatches})):b=Ct(e.evaluate(s,i,{batchSize:r.validationBatchSize==null?fU:r.validationBatchSize,verbose:0}));for(let w=0;w<e.metricsNames.length;++w)g[`val_${e.metricsNames[w]}`]=b[w]}break}if(e.stopTraining_)break}if(await p.onEpochEnd(f,g),f++,e.stopTraining_)break}return await p.onTrainEnd(),await e.history.syncData(),e.history}finally{e.isTraining=!1}}function yU(e,t){let r=null;return t.batchesPerEpoch!=null?r=t.batchesPerEpoch:Number.isFinite(e.size)&&(r=e.size),r}function g4(e){return typeof e.iterator=="function"}function AU(e){return typeof e.next=="function"}async function xU(e,t,r){r=r||{};let n=r.batches!=null,a=e.testFunction,s=[];if(r.verbose>0)throw new Ve("Verbose mode is not implemented yet.");v.assert(!n||r.batches>0&&Number.isInteger(r.batches),()=>`Test loop expects \`batches\` to be a positive integer, but received ${JSON.stringify(r.batches)}`);let i=AU(t)?t:await t.iterator(),o=0,l=0;for(;!n||l<r.batches;){let u=await i.next();if(s=X(()=>{if(u.value){let{xs:d,ys:h}=T6(e,u.value),p=d.concat(h),c=X(()=>a(p));if(te(p),l===0)for(let m=0;m<c.length;++m)s.push(Se(0));let f=p[0].shape[0];for(let m=0;m<c.length;++m){let g=c[m],y=s[m];s[m]=X(()=>le(s[m],L(f,g))),l>0&&te(y)}te(c),o+=f,++l}return s}),u.done){n&&console.warn(`Your dataset iterator ran out of data during evaluateDataset(). Interrupting evalution. Make sure that your dataset can generate at least \`batches\` batches (in this case, ${r.batches} batches). You may need to use the repeat() function when building your dataset.`);break}}for(let u=0;u<s.length;++u){let d=s[u];s[u]=pe(s[u],o),te(d)}return tn(s)}function zg(e){v.assert(e>0&&Number.isInteger(e),()=>`batchSize is required to be a positive integer, but got ${e}`)}function Sp(e,t,r){return e==null?[null]:Array.isArray(e)?e.map(n=>$o(n,t,r-t)):$o(e,t,r-t)}function GA(e,t){return X(()=>e==null?null:Array.isArray(e)?e.map(r=>GA(r,t)):n6(e,t.dtype==="int32"?t:me(t,"int32")))}function Og(e,t){let r=[],n=0,a=null;for(;n<e;)a=n+t,a>=e&&(a=e),r.push([n,a]),n=a;return r}async function bU(e,t,r,n,a,s,i,o,l,u,d,h,p,c,f){a==null&&(a=32),s==null&&(s=1),d==null&&(d=!0),p==null&&(p=0);let m=!1;if(l!=null&&u!=null&&(m=!0),f!=null&&(m=!0,c==null))throw new H("Can only use `validationSteps` when doing step-wise training, i.e., `stepsPerEpoch` must be set.");let g=e.checkNumSamples(r,a,c,"steps_per_epoch"),y;g!=null&&(y=ya(0,g)),i==null&&(i=1);let{callbackList:A,history:x}=A6(o,i,s,p,g,c,a,m,h);A.setModel(e),e.history=x,await A.onTrainBegin(),e.stopTraining_=!1;for(let b=p;b<s;++b){await A.onEpochBegin(b);let w={};if(c!=null)throw new Ve("stepsPerEpoch mode is not implemented yet.");{if(d==="batch")throw new Ve("batch shuffling is not implemneted yet");d&&v.shuffle(y);let S=Tt(y),N=Og(g,a);for(let E=0;E<N.length;++E){let R={};if(await A.onBatchBegin(E,R),X(()=>{let z=N[E][0],M=N[E][1],I=$o(S,z,M-z);R.batch=E,R.size=M-z;let _=GA(r,I),O=t(_);for(let q=0;q<n.length;++q){let K=n[q],D=O[q];R[K]=D,gr(D)}if(E===N.length-1&&m){let q=e.testLoop(l,u,a);for(let K=0;K<n.length;++K){let D=n[K],Q=q[K];gr(Q),w["val_"+D]=Q}}}),await A.onBatchEnd(E,R),c6(R),e.stopTraining_)break}S.dispose()}if(await A.onEpochEnd(b,w),e.stopTraining_)break}return await A.onTrainEnd(),await e.history.syncData(),e.history}async function vU(e,t,r,n={}){if(e.isTraining)throw new Error("Cannot start training because another fit() call is ongoing.");e.isTraining=!0;let a,s,i,o,l,u,d,h,p;try{let c=n.batchSize==null?32:n.batchSize;zg(c);let f=!1,m=await e.standardizeUserData(t,r,n.sampleWeight,n.classWeight,f,c);a=m[0],s=m[1],p=m[2];let g=!1,y;if(n.validationData!=null&&n.validationData.length>0){if(g=!0,n.validationData.length===2)l=n.validationData[0],u=n.validationData[1];else throw n.validationData.length===3?new Ve("validationData including sample weights is not supported yet."):new H(`When passing validation data, it must contain 2 (valX, valY) or 3 (valX, valY, valSampleWeight) items; ${n.validationData} is invalid.`);let E=!0,R=await e.standardizeUserData(l,u,null,null,E,c);d=R[0],h=R[1],y=d.concat(h)}else if(n.validationSplit!=null&&n.validationSplit>0&&n.validationSplit<1){g=!0;let E=Math.floor(a[0].shape[0]*(1-n.validationSplit)),R=a[0].shape[0];d=Sp(a,E,R),i=a,a=Sp(a,0,E),h=Sp(s,E,R),o=s,s=Sp(s,0,E),y=d.concat(h)}else n.validationSteps!=null&&(g=!0);let A=a.concat(s).concat(p);e.checkTrainableWeightsConsistency();let x=e.makeTrainFunction(),b=e.getDedupedMetricsNames(),w,S;g?(e.makeTestFunction(),w=e.testFunction,S=b.slice().concat(b.map(E=>"val_"+E))):(w=null,y=[],S=b.slice());let N=y6(n.callbacks,n.yieldEvery);return await bU(e,x,A,b,c,n.epochs,n.verbose,N,w,y,n.shuffle,S,n.initialEpoch,null,null)}finally{e.isTraining=!1,ua(a,t),ua(s,r),ua(i,t),ua(o,r),ua(d,l),ua(h,u),p!=null&&te(p)}}function N6(e){let t=[];e instanceof nt&&(e=[e]);for(let r=0;r<e.length;++r){let n=e[r];if(n.rank===1)t.push(Nh(n,1));else{if(n.rank===0)throw new Error("Expected tensor to be at least 1D, but received a 0D tensor (scalar).");t.push(n)}}return t}function ua(e,t){if(e==null)return;let r=[];if(t instanceof nt)r.push(t.id);else if(Array.isArray(t))t.forEach(a=>r.push(a.id));else if(t!=null)for(let a in t){let s=t[a];r.push(s.id)}let n=[];if(e instanceof nt)r.indexOf(e.id)===-1&&n.push(e);else if(Array.isArray(e))e.forEach(a=>{r.indexOf(a.id)===-1&&n.push(a)});else if(e!=null)for(let a in e){let s=e[a];r.indexOf(s.id)===-1&&n.push(s)}n.forEach(a=>{a.isDisposed||a.dispose()})}function wU(e){return e instanceof nt}function Dg(e){return Array.isArray(e)}function y4(e){return!wU(e)&&!Dg(e)}function A4(e,t,r,n=!0,a=""){if(t==null||t.length===0){if(e!=null){let i=!1;if(Dg(e)&&e.length>0)i=!0;else if(y4(e)){for(let o in e)if(e.hasOwnProperty(o)){i=!0;break}}else i=!0;if(i)throw new H(`Error when checking model ${a} expected no data, but got ${e}`)}return[]}if(e==null)return t.map(i=>null);let s;if(y4(e)){e=e,s=[];for(let i of t){if(e[i]==null)throw new H(`No data provided for "${i}". Need data for each key in: ${t}`);s.push(e[i])}}else if(Dg(e)){if(e=e,e.length!==t.length)throw new H(`Error when checking model ${a}: the Array of Tensors that you are passing to your model is not the size the model expected. Expected to see ${t.length} Tensor(s), but instead got the following list of Tensor(s): ${e}`);s=e}else{if(e=e,t.length>1)throw new H(`The model ${a} expects ${t.length} Tensor(s), but only received one Tensor. Found: Tensor with shape ${e.shape}`);s=[e]}if(s=N6(s),r!=null)for(let i=0;i<t.length;++i){if(r[i]==null)continue;let o=s[i];if(o.shape.length!==r[i].length)throw new H(`Error when checking ${a}: expected ${t[i]} to have ${r[i].length} dimension(s). but got array with shape ${o.shape}`);for(let l=0;l<r[i].length;++l){if(l===0&&!n)continue;let u=o.shape[l],d=r[i][l];if(d!=null&&d>=0&&u!==d)throw new H(`${a} expected a batch of elements where each example has shape [${r[i].slice(1,r[i].length)}] (i.e.,tensor shape [*,${r[i].slice(1,r[i].length)}]) but the ${a} received an input with ${o.shape[0]} examples, each with shape [${o.shape.slice(1,o.shape.length)}] (tensor shape [${o.shape}])`)}}return s}function kU(e,t,r){let n=Rs(e.map(s=>s.shape[0]));n.sort();let a=Rs(t.map(s=>s.shape[0]));if(a.sort(),n.length>1)throw new H(`All input Tensors (x) should have the same number of samples. Got array shapes: ${JSON.stringify(e.map(s=>s.shape))}`);if(a.length>1)throw new H(`All target Tensors (y) should have the same number of samples. Got array shapes: ${JSON.stringify(t.map(s=>s.shape))}`);if(n.length>0&&a.length>0&&!v.arraysEqual(n,a))throw new H(`Input Tensors should have the same number of samples as target Tensors. Found ${n[0]} input sample(s) and ${a[0]} target sample(s).`)}function IU(e,t,r){let n=[zl,pm,jp];for(let a=0;a<e.length;++a){let s=e[a],i=t[a],o=r[a];if(i!=null){if(i===jp&&s.shape[s.shape.length-1]===1)throw new H(`You are passing a target array of shape ${s.shape} while using a loss 'categorical_crossentropy'. 'categorical_crossentropy'expects targets to be binary matrices (1s and 0s) of shape [samples, classes].`);if(n.indexOf(i)!==-1){let l=s.shape.slice(1),u=o.slice(1);for(let d=0;d<l.length;++d){let h=l[d],p=u[d];if(p!=null&&h!==p)throw new H(`A target Tensor with shape ${s.shape} was passed for an output of shape ${o}, while using a loss function that expects targets to have the same shape as the output.`)}}}}}function x4(e,t,r,n=!0,a=""){let s;if(Array.isArray(e)){if(e.length!==t.length)throw new H(`Error when checking model ${a}: the Array of Tensors that you are passing to your model is not the size the the model expected. Expected to see ${t.length} Tensor(s), but instead got ${e.length} Tensors(s).`);s=e}else{if(t.length>1)throw new H(`The model expects ${t.length} ${a} Tensors, but only received one Tensor. Found: array with shape ${JSON.stringify(e.shape)}.`);s=[e]}if(r!=null)for(let i=0;i<t.length;++i){if(r[i]==null)continue;let o=s[i];if(o.shape.length!==r[i].length)throw new H(`Error when checking ${a}: expected ${t[i]} to have ${r[i].length} dimension(s), but got array with shape ${JSON.stringify(o.shape)}`);for(let l=0;l<r[i].length;++l){if(l===0&&!n)continue;let u=o.shape[l],d=r[i][l];if(d!=null&&d!==u)throw new H(`Error when checking ${a}: expected ${t[i]} to have shape ${JSON.stringify(r[i])} but got array with shape ${JSON.stringify(o.shape)}.`)}}}function SU(e,t){if(e==null||Array.isArray(e)&&e.length===0)return t.map(n=>[]);let r;if(typeof e=="string"||typeof e=="function")r=[e];else if(Array.isArray(e)||typeof e=="object")r=e;else throw new TypeError(`Type of metrics argument not understood. Expected an string,function, Array, or Object, found: ${e}`);if(Array.isArray(r))return t.map(n=>r);{let n=[];for(let a of t){let s=r.hasOwnProperty(a)?r[a]:[];Array.isArray(s)||(s=[s]),n.push(s)}return n}}var CU="layers-model",Ka=class extends Ta{constructor(e){super(e),this.isTraining=!1}summary(e,t,r=console.log){if(!this.built)throw new H("This model has never been called, thus its weights have not been created yet. So no summary can be displayed. Build the model first (e.g., by calling it on some test data).");oU(this,e,t,r)}compile(e){if(e.loss==null&&(e.loss=[]),this.loss=e.loss,typeof e.optimizer=="string")this.optimizer_=iU(e.optimizer),this.isOptimizerOwned=!0;else{if(!(e.optimizer instanceof ns))throw new H("User-defined optimizer must be an instance of tf.Optimizer.");this.optimizer_=e.optimizer,this.isOptimizerOwned=!1}let t=[];if(!Array.isArray(e.loss)&&typeof e.loss!="string"&&typeof e.loss!="function"){e.loss=e.loss;for(let s in e.loss)if(this.outputNames.indexOf(s)===-1)throw new H(`Unknown entry in loss dictionary: "${s}". Only expected the following keys: ${this.outputNames}`);for(let s of this.outputNames)e.loss[s]==null&&console.warn(`Output "${s}" is missing from loss dictionary. We assume this was done on purpose, and we will not be expecting data to be passed to ${s} during training`),t.push(lg(e.loss[s]))}else if(Array.isArray(e.loss)){if(e.loss.length!==this.outputs.length)throw new H(`When passing an Array as loss, it should have one entry per model output. The model has ${this.outputs.length} output(s), but you passed loss=${e.loss}.`);t=e.loss.map(s=>lg(s))}else{let s=lg(e.loss);this.outputs.forEach(i=>{t.push(s)})}this.lossFunctions=t,this.feedOutputNames=[],this.feedOutputShapes=[],this.feedLossFns=[];for(let s=0;s<this.outputs.length;++s){let i=this.internalOutputShapes[s],o=this.outputNames[s];this.feedOutputNames.push(o),this.feedOutputShapes.push(i),this.feedLossFns.push(this.lossFunctions[s])}let r=[];this.metrics=e.metrics,this.metricsNames=["loss"],this.metricsTensors=[],Ro("loss",()=>{for(let s=0;s<this.outputs.length;++s){if(r.indexOf(s)!==-1)continue;let i=this.lossFunctions[s];this.outputs.length>1&&(this.metricsTensors.push([i,s]),this.metricsNames.push(this.outputNames[s]+"_loss"))}});let n=SU(e.metrics,this.outputNames),a=(s,i,o)=>{this.outputNames.length>1&&(i=this.outputNames[s]+"_"+i),this.metricsNames.push(i),this.metricsTensors.push([o,s])};Ro("metric",()=>{for(let s=0;s<this.outputs.length;++s){if(r.indexOf(s)!==-1)continue;let i=n[s];(o=>{let l="",u,d,h;for(let p of o){if(typeof p=="string"&&["accuracy","acc","crossentropy","ce"].indexOf(p)!==-1){let f=this.internalOutputShapes[s];f[f.length-1]===1||this.lossFunctions[s]===pm?["accuracy","acc"].indexOf(p)!==-1?d=BA:["crossentropy","ce"].indexOf(p)!==-1&&(d=v6):this.lossFunctions[s]===P0?["accuracy","acc"].indexOf(p)!==-1?d=w6:["crossentropy","ce"].indexOf(p)!==-1&&(d=k6):["accuracy","acc"].indexOf(p)!==-1?d=WA:["crossentropy","ce"].indexOf(p)!==-1&&(d=VA);let m;["accuracy","acc"].indexOf(p)!==-1?m="acc":["crossentropy","ce"].indexOf(p)!==-1&&(m="ce"),h=d,u=l+m}else h=sU(p),u=l+Yc(p);let c;Ro(u,()=>{c=h}),a(s,u,c)}})(i)}}),this.collectedTrainableWeights=this.trainableWeights}checkTrainableWeightsConsistency(){this.collectedTrainableWeights!=null&&this.trainableWeights.length!==this.collectedTrainableWeights.length&&console.warn("Discrepancy between trainableweights and collected trainable weights. Did you set `model.trainable` without calling `model.compile()` afterwards?")}evaluate(e,t,r={}){let n=r.batchSize==null?32:r.batchSize;zg(n);let a=!0,s=this.standardizeUserDataXY(e,t,a,n);try{let i=s[0].concat(s[1]);this.makeTestFunction();let o=this.testFunction,l=this.testLoop(o,i,n,r.verbose,r.steps);return tn(l)}finally{ua(s[0],e),ua(s[1],t)}}async evaluateDataset(e,t){return this.makeTestFunction(),xU(this,e,t)}checkNumSamples(e,t,r,n="steps"){let a;if(r!=null){if(a=null,t!=null)throw new H(`If ${n} is set, batchSize must be null or undefined.Got batchSize = ${t}`)}else if(e!=null)Array.isArray(e)?a=e[0].shape[0]:a=e.shape[0];else throw new H(`Either the input data should have a defined shape, or ${n} shoud be specified.`);return a}execute(e,t){if(Array.isArray(t)&&t.length===0)throw new H("`outputs` is an empty Array, which is not allowed.");let r=Array.isArray(t),n=r?t:[t],a=this.retrieveSymbolicTensors(n),s=new To;if(e instanceof nt&&(e=[e]),Array.isArray(e)){if(e.length!==this.inputs.length)throw new H(`The number of inputs provided (${e.length}) does not match the number of inputs of this model (${this.inputs.length}).`);for(let o=0;o<this.inputs.length;++o)s.add(this.inputs[o],e[o])}else for(let o of this.inputs){let l=e[o.name];if(l==null)throw new H(`No value is provided for the model's input ${o.name}`);s.add(o,l)}let i=Ip(a,s);return r?i:i[0]}retrieveSymbolicTensors(e){let t=Do(null,e.length),r=e.length;for(let n of this.layers){let a=Array.isArray(n.output)?n.output:[n.output],s=a.map(i=>i.name);for(let i=0;i<e.length;++i){let o=s.indexOf(e[i]);if(o!==-1&&(t[i]=a[o],r--),r===0)break}if(r===0)break}if(r>0){let n=[];throw t.forEach((a,s)=>{a==null&&n.push(e[s])}),new H(`Cannot find SymbolicTensors for output name(s): ${JSON.stringify(n)}`)}return t}predictLoop(e,t=32,r=!1){return X(()=>{let n=this.checkNumSamples(e);if(r)throw new Ve("Verbose predictLoop() is not implemented yet.");let a=Og(n,t),s=this.outputs.map(i=>[]);for(let i=0;i<a.length;++i)X(()=>{let o=a[i][0],l=a[i][1],u=Sp(e,o,l),d=[];if(Array.isArray(u))for(let p=0;p<u.length;++p)d.push({key:this.inputs[p],value:u[p]});else d.push({key:this.inputs[0],value:u});let h=new To(d);return Ip(this.outputs,h)}).forEach((o,l)=>s[l].push(o));return tn(s.map(i=>St(i,0)))})}predict(e,t={}){let r=N6(e);x4(r,this.inputNames,this.feedInputShapes,!1);try{let n=t.batchSize==null?32:t.batchSize;return zg(n),this.predictLoop(r,n)}finally{ua(r,e)}}predictOnBatch(e){x4(e,this.inputNames,this.feedInputShapes,!0);let t=(Array.isArray(e)?e[0]:e).shape[0];return this.predictLoop(e,t)}standardizeUserDataXY(e,t,r=!0,n){if(this.optimizer_==null)throw new da("You must compile a model before training/testing. Use LayersModel.compile(modelCompileArgs).");let a=[];for(let s=0;s<this.feedOutputShapes.length;++s){let i=this.feedOutputShapes[s];this.feedLossFns[s]===P0?a.push(i.slice(0,i.length-1).concat([1])):a.push(i)}if(e=A4(e,this.feedInputNames,this.feedInputShapes,!1,"input"),t=A4(t,this.feedOutputNames,a,!1,"target"),kU(e,t,null),IU(t,this.feedLossFns,this.feedOutputShapes),this.stateful&&n!=null&&n>0&&e[0].shape[0]%n!==0)throw new H(`In a stateful network, you should only pass inputs with a number of samples that is divisible by the batch size ${n}. Found: ${e[0].shape[0]} sample(s).`);return[e,t]}async standardizeUserData(e,t,r,n,a=!0,s){let[i,o]=this.standardizeUserDataXY(e,t,a,s);if(r!=null)throw new Error("sample weight is not supported yet.");let l=null;if(n!=null){let u=S6(n,this.outputNames);l=[];for(let d=0;d<u.length;++d)l.push(await C6(o[d],null,u[d]))}return[i,o,l]}testLoop(e,t,r,n=0,a){return X(()=>{let s=this.checkNumSamples(t,r,a,"steps"),i=[];if(n>0)throw new Ve("Verbose mode is not implemented yet.");if(a!=null)throw new Ve("steps mode in testLoop() is not implemented yet");{let o=Og(s,r),l=Tt(ya(0,s));for(let u=0;u<o.length;++u){let d=o[u][0],h=o[u][1],p=$o(l,d,h-d),c=GA(t,p),f=e(c);if(u===0)for(let m=0;m<f.length;++m)i.push(Se(0));for(let m=0;m<f.length;++m){let g=f[m];i[m]=le(i[m],L(h-d,g))}}for(let u=0;u<i.length;++u)i[u]=pe(i[u],s)}return i})}getDedupedMetricsNames(){let e=this.metricsNames,t=[];for(let r=0;r<e.length;++r){let n=e[r],a=n;J3(e,n)>1&&(a+=`_${J3(e.slice(0,r),n)}`),t.push(a)}return t}makeTrainFunction(){return e=>{let t=[],r=e.slice(0,this.inputs.length),n=e.slice(this.inputs.length,this.inputs.length+this.outputs.length),a=e.slice(this.inputs.length+this.outputs.length,this.inputs.length+this.outputs.length*2),s=[],i=()=>{let u=[];for(let c=0;c<this.inputs.length;++c)u.push({key:this.inputs[c],value:r[c]});let d=new To(u),h=Ip(this.outputs,d,{training:!0}),p;for(let c=0;c<this.lossFunctions.length;++c){let f=this.lossFunctions[c](n[c],h[c]);a[c]!=null&&(f=cU(f,a[c]));let m=Vt(f);t.push(m),c===0?p=f:p=le(p,f)}for(let c=0;c<this.metricsTensors.length;++c){let f;if(this.outputs.length>1&&c<this.outputs.length)f=t[c];else{let m=this.metricsTensors[c][0],g=this.metricsTensors[c][1];f=Vt(m(n[g],h[g]))}gr(f),s.push(f)}return p=Vt(p),this.calculateLosses().forEach(c=>{p=le(p,c)}),p},o=this.collectedTrainableWeights.map(u=>u.read()),l=!0;return[this.optimizer_.minimize(i,l,o)].concat(s)}}makeTestFunction(){this.testFunction=e=>X(()=>{let t=[],r,n=e.slice(0,this.inputs.length),a=e.slice(this.inputs.length,this.inputs.length+this.outputs.length),s=[];for(let l=0;l<this.inputs.length;++l)s.push({key:this.inputs[l],value:n[l]});let i=new To(s),o=Ip(this.outputs,i);for(let l=0;l<this.lossFunctions.length;++l){let u=this.lossFunctions[l],d=Vt(u(a[l],o[l]));l===0?r=d:r=le(r,d),t.push(r)}for(let l=0;l<this.metricsTensors.length;++l){let u=this.metricsTensors[l][0],d=this.metricsTensors[l][1],h=Vt(u(a[d],o[d]));t.push(h)}return t})}async fit(e,t,r={}){return vU(this,e,t,r)}async fitDataset(e,t){return gU(this,e,t)}async trainOnBatch(e,t){let r=await this.standardizeUserData(e,t),n=r[0],a=r[1],s=this.makeTrainFunction()(n.concat(a)),i=[];for(let o of s){let l=await o.data();i.push(l[0])}return te(s),ua(r[0],e),ua(r[1],t),tn(i)}getNamedWeights(e){let t=[],r=e!=null&&e.trainableOnly,n=r?this.trainableWeights:this.weights,a=this.getWeights(r);for(let s=0;s<n.length;++s)r&&!n[s].trainable||t.push({name:n[s].originalName,tensor:a[s]});return t}set stopTraining(e){this.stopTraining_=e}get stopTraining(){return this.stopTraining_}get optimizer(){return this.optimizer_}set optimizer(e){this.optimizer_!==e&&(this.optimizer_=e,this.isOptimizerOwned=!1)}dispose(){let e=super.dispose();if(e.refCountAfterDispose===0&&this.optimizer!=null&&this.isOptimizerOwned){let t=S0().numTensors;this.optimizer_.dispose(),e.numDisposedVariables+=t-S0().numTensors}return e}getLossIdentifiers(){let e;if(typeof this.loss=="string")e=ja(this.loss);else if(Array.isArray(this.loss)){for(let t of this.loss)if(typeof t!="string")throw new Error("Serialization of non-string loss is not supported.");e=this.loss.map(t=>ja(t))}else{let t=Object.keys(this.loss);e={};let r=this.loss;for(let n of t)if(typeof r[n]=="string")e[n]=ja(r[n]);else throw new Error("Serialization of non-string loss is not supported.")}return e}getMetricIdentifiers(){if(typeof this.metrics=="string"||typeof this.metrics=="function")return[ja(Yc(this.metrics))];if(Array.isArray(this.metrics))return this.metrics.map(e=>ja(Yc(e)));{let e={};for(let t in this.metrics)e[t]=ja(Yc(this.metrics[t]));return e}}getTrainingConfig(){return{loss:this.getLossIdentifiers(),metrics:this.getMetricIdentifiers(),optimizer_config:{class_name:this.optimizer.getClassName(),config:this.optimizer.getConfig()}}}loadTrainingConfig(e){if(e.weighted_metrics!=null)throw new Error("Loading weight_metrics is not supported yet.");if(e.loss_weights!=null)throw new Error("Loading loss_weights is not supported yet.");if(e.sample_weight_mode!=null)throw new Error("Loading sample_weight_mode is not supported yet.");let t=Hp(e.optimizer_config),r=fa(t),n;if(typeof e.loss=="string")n=Io(e.loss);else if(Array.isArray(e.loss))n=e.loss.map(s=>Io(s));else if(e.loss!=null){n={};for(let s in e.loss)n[s]=Io(e.loss[s])}let a;if(Array.isArray(e.metrics))a=e.metrics.map(s=>Io(s));else if(e.metrics!=null){a={};for(let s in e.metrics)a[s]=Io(e.metrics[s])}this.compile({loss:n,metrics:a,optimizer:r})}async save(e,t){if(typeof e=="string"){let i=Cr.getSaveHandlers(e);if(i.length===0)throw new H(`Cannot find any save handlers for URL '${e}'`);if(i.length>1)throw new H(`Found more than one (${i.length}) save handlers for URL '${e}'`);e=i[0]}if(e.save==null)throw new H("LayersModel.save() cannot proceed because the IOHandler provided does not have the `save` attribute defined.");let r=await Cr.encodeWeights(this.getNamedWeights(t)),n=!1,a=null,s={modelTopology:this.toJSON(a,n),format:CU,generatedBy:`TensorFlow.js tfjs-layers v${UA}`,convertedBy:null};if((t==null?!1:t.includeOptimizer)&&this.optimizer!=null){s.trainingConfig=this.getTrainingConfig();let i="optimizer",{data:o,specs:l}=await Cr.encodeWeights(await this.optimizer.getWeights(),i);r.specs.push(...l),r.data=Cr.concatenateArrayBuffers([r.data,o])}return this.userDefinedMetadata!=null&&(f4(this.userDefinedMetadata,this.name,!0),s.userDefinedMetadata=this.userDefinedMetadata),s.weightData=r.data,s.weightSpecs=r.specs,e.save(s)}setUserDefinedMetadata(e){f4(e,this.name),this.userDefinedMetadata=e}getUserDefinedMetadata(){return this.userDefinedMetadata}};Ka.className="Model";ue.registerClass(Ka);var E6=class extends Ka{};E6.className="Functional";ue.registerClass(E6);async function TU(e,t){"modelTopology"in e||(e={modelTopology:e}),e=e;let r=e.modelTopology;r.model_config!=null&&(r=r.model_config);let n=Hp(r),a=fa(n,t);if(e.weightsManifest!=null){let s=await Cr.loadWeights(e.weightsManifest,e.pathPrefix,a.weights.map(o=>o.originalName)),i={};for(let o of a.weights)i[o.originalName]=s[o.originalName];a.loadWeights(i),te(s)}return a}async function NU(e,t){if(t==null&&(t={}),typeof e=="string"){let r=Cr.getLoadHandlers(e,t);if(r.length===0)r.push(Cr.browserHTTPRequest(e,t));else if(r.length>1)throw new H(`Found more than one (${r.length}) load handlers for URL '${e}'`);e=r[0]}return EU(e,void 0,t)}async function EU(e,t,r){if(r==null&&(r={}),e.load==null)throw new H("Cannot proceed with model loading because the IOHandler provided does not have the `load` method implemented.");let n=await e.load(),a=n.modelTopology;a.model_config!=null&&(a=a.model_config);let s=r.strict==null?!0:r.strict,i=n.weightData!=null&&n.weightSpecs!=null&&s,o=fa(Hp(a),t,i),l=n.trainingConfig;if(l!=null&&o.loadTrainingConfig(l),n.userDefinedMetadata!=null&&o.setUserDefinedMetadata(n.userDefinedMetadata),n.weightData!=null){if(n.weightSpecs==null)throw new H("LayersModel artifacts contains weight data, but not weight specs. Therefore loading of weights cannot proceed.");let{modelWeights:u,optimizerWeights:d}=RU(n.weightData,n.weightSpecs);o.loadWeights(u,s),o.optimizer!=null&&d.length>0&&await o.optimizer.setWeights(d),te(u),te(d.map(h=>h.tensor))}return o}function RU(e,t){let r=Cr.decodeWeights(e,t),n={},a=[];return t.forEach(s=>{s.group==="optimizer"?a.push({name:s.name,tensor:r[s.name]}):n[s.name]=r[s.name]}),{modelWeights:n,optimizerWeights:a}}var Lg=class extends Ka{constructor(e){if(super({inputs:[],outputs:[]}),e=e||{},this.trainable=!0,this.built=!1,this.name=e.name!=null?e.name:Qf("sequential_"),e.layers!=null)for(let t of e.layers)this.add(t)}checkShape(e){if(e.inboundNodes[0].outputTensors[0].shape.some(t=>t<0))throw new H(`Negative dimension size caused by adding layer ${e.name} with input shape [${e.inboundNodes[0].inputTensors[0].shape}]`)}add(e){let t=e instanceof Lg||e instanceof Ka,r;if(t){if(r=e,r.outputs.length!==1)throw new H("All layers in a Sequential model should have a single output tensor. For multi-output layers, use the functional API.");if(r.inputs.length!==1)throw new H("All layers in a Sequential model should have a single input tensor. For multi-input layers, use the functional API.")}if(this.outputs.length===0){if(e.inboundNodes.length===0){if(e.batchInputShape==null)throw new H("The first layer in a Sequential model must get an `inputShape` or `batchInputShape` argument.");let n=o6({batchShape:e.batchInputShape,dtype:e.dtype,name:e.name+"_input"});e.apply(n)}if(t)this.outputs=r.outputs,this.inputs=r.inputs;else{if(e.inboundNodes.length!==1)throw new H(`A layer added to a Sequential model must not already be connected somewhere else. LayersModel received layer ${e.name} which has ${e.inboundNodes.length} pre-existing inbound connections.`);if(e.inboundNodes[0].outputTensors.length!==1)throw new H("All layers in a Sequential model should have a single output tensor. For multi-output layers, use the functional API.");this.checkShape(e),this.outputs=[e.inboundNodes[0].outputTensors[0]],this.inputs=i6(this.outputs[0])}this.inboundNodes=[],new um({outboundLayer:this,inboundLayers:[],nodeIndices:[],tensorIndices:[],inputTensors:this.inputs,outputTensors:this.outputs,inputMasks:Do(null,this.inputs.length),outputMasks:[null],inputShapes:this.inputs.map(n=>n.shape),outputShapes:this.outputs[0].shape})}else{let n=e.apply(this.outputs[0]);if(Array.isArray(n))throw new TypeError("All layers in a Sequential model should have a single output tensor. For multi-output layers, use the functional API.");this.checkShape(e),this.outputs=[n],this.inboundNodes[0].outputTensors=this.outputs,this.inboundNodes[0].outputShapes=[this.outputs[0].shape]}this.layers.push(e),this.built=!1}pop(){if(this.layers.length===0)throw new TypeError("There are no layers in the model.");if(this.layers.pop(),this.layers.length===0)this.outputs=[],this.inboundNodes=[],this.outboundNodes=[];else{let e=this.layers.length-1;this.layers[e].outboundNodes=[],this.outputs=[this.layers[e].output],this.inboundNodes[0].outputTensors=this.outputs,this.inboundNodes[0].outputShapes=[this.outputs[0].shape]}}call(e,t){return this.model==null&&this.build(),this.model.call(e,t)}build(e){if(mt(e),this.inputs.length===0||this.outputs.length===0)throw new TypeError("Sequential model cannot be built: model is empty. Add some layers first.");this.model=new Ka({inputs:this.inputs,outputs:this.outputs[0],name:this.name+"_model"}),this.model.trainable=this.trainable,this.supportsMasking=this.model.supportsMasking,this.inputLayers=this.model.inputLayers,this.inputLayersNodeIndices=this.model.inputLayersNodeIndices,this.inputLayersTensorIndices=this.model.inputLayersTensorIndices,this.outputLayers=this.model.outputLayers,this.outputLayersNodeIndices=this.model.outputLayersNodeIndices,this.outputLayersTensorIndices=this.model.outputLayersTensorIndices,this.nodesByDepth=this.model.nodesByDepth,this.containerNodes=this.model.containerNodes,this.outputNames=this.model.outputNames,this.inputNames=this.model.inputNames,this.built=!0}countParams(){return this.built||this.build(),super.countParams()}summary(e,t,r=console.log){this.built||this.build(),super.summary(e,t,r)}setWeights(e){this.model==null&&this.build(),this.model.setWeights(e)}evaluate(e,t,r={}){if(!this.built)throw new da("The model needs to be compiled before being used.");return this.model.evaluate(e,t,r)}async evaluateDataset(e,t){if(!this.built)throw new da("The model needs to be compiled before being used.");return this.model.evaluateDataset(e,t)}predict(e,t={}){return this.model==null&&this.build(),this.model.predict(e,t)}predictOnBatch(e){return this.model==null&&this.build(),this.model.predictOnBatch(e)}compile(e){this.build(),this.model.compile(e),this.optimizer_=this.model.optimizer,this.isOptimizerOwned=this.model.isOptimizerOwned,this.loss=this.model.loss,this.metrics=this.model.metrics,this.metricsTensors=this.model.metricsTensors,this.metricsNames=this.model.metricsNames}get optimizer(){return this.model==null?void 0:this.model.optimizer}set optimizer(e){this.model.optimizer=e}async fit(e,t,r={}){if(!this.built)throw new da("The model needs to be compiled before being used.");return this.model.fit(e,t,r)}async fitDataset(e,t){if(!this.built)throw new da("The model needs to be compiled before being used.");return this.model.fitDataset(e,t)}async trainOnBatch(e,t){return this.model.trainOnBatch(e,t)}static fromConfig(e,t,r={},n=!1){let a,s={};if(t instanceof Array){if(t[0].className==null||t[0].className==="Merge")throw new H("Legacy serialization format not supported yet.");a=t}else v.assert(t.layers!=null,()=>"When the config data for a Sequential model is not an Array, it must be an Object that contains the 'layers' field."),a=t.layers,delete t.layers,s=t;let i=new e(s);if(!(i instanceof Lg))throw new Ve(`Sequential.fromConfig called on non-Sequential input: ${i}`);for(let o of a){let l=fa(o,void 0,n);n&&l.setFastWeightInitDuringBuild(!0),i.add(l)}return i}set stopTraining(e){if(this.model==null)throw new H("Cannot set the stopTraining property of a sequential model before it is compiled.");this.model.stopTraining=e}get stopTraining(){if(this.model==null)throw new H("Cannot get the stopTraining property of a sequential model before it is compiled.");return this.model.stopTraining}getConfig(){let e=[];for(let t of this.layers){let r={};r.className=t.getClassName(),r.config=t.getConfig(),e.push(r)}return{name:this.name,layers:e}}},hm=Lg;hm.className="Sequential";ue.registerClass(hm);function $U(e){return new Ka(e)}function MU(e){return new hm(e)}function FU(e,t){return t==null&&(t={}),NU(e,t)}function R6(e){return o6(e)}function PU(e,t){DA.registerCallbackConstructor(e,t)}var on=class extends ue.Serializable{getConfig(){return{}}},$6=class extends on{apply(e,t=1){return YL(e,t)}};$6.className="elu";ue.registerClass($6);var M6=class extends on{apply(e){return lA(e)}};M6.className="selu";ue.registerClass(M6);var F6=class extends on{apply(e){return Oa(e)}};F6.className="relu";ue.registerClass(F6);var P6=class extends on{apply(e){return X(()=>Ih(6,Oa(e)))}};P6.className="relu6";ue.registerClass(P6);var _6=class extends on{apply(e){return e}};_6.className="linear";ue.registerClass(_6);var z6=class extends on{apply(e){return Tr(e)}};z6.className="sigmoid";ue.registerClass(z6);var O6=class extends on{apply(e){return QL(e)}};O6.className="hardSigmoid";ue.registerClass(O6);var D6=class extends on{apply(e){return fd(e)}};D6.className="softplus";ue.registerClass(D6);var L6=class extends on{apply(e){return JL(e)}};L6.className="softsign";ue.registerClass(L6);var B6=class extends on{apply(e){return Iu(e)}};B6.className="tanh";ue.registerClass(B6);var jA=class extends on{apply(e,t=-1){return yd(e,t)}};jA.className="softmax";ue.registerClass(jA);var W6=class extends on{apply(e,t=-1){return Zy(e,t)}};W6.className="logSoftmax";ue.registerClass(W6);var V6=class extends on{apply(e,t=1){return X(()=>L(Tr(L(e,t)),e))}};V6.className="swish";ue.registerClass(V6);var U6=class extends on{apply(e){return X(()=>L(e,Iu(fd(e))))}};U6.className="mish";ue.registerClass(U6);function Gs(e){return e.getClassName()}function ug(e,t={}){return Th(e,ue.SerializationMap.getMap().classNameMap,t,"activation")}function js(e){if(e==null){let t={};return t.className="linear",t.config={},ug(t)}if(typeof e=="string"){let t={};return t.className=e,t.config={},ug(t)}else return e instanceof on?e:ug(e)}function HA(e){if(e!=null&&typeof e!="object")throw new Error(`Argument to L1L2 regularizer's constructor is expected to be an object, but received: ${e}`)}var G6=class extends ue.Serializable{},Mh=class extends G6{constructor(e){super(),HA(e),this.l1=e==null||e.l1==null?.01:e.l1,this.l2=e==null||e.l2==null?.01:e.l2,this.hasL1=this.l1!==0,this.hasL2=this.l2!==0}apply(e){return X(()=>{let t=_t([1]);return this.hasL1&&(t=le(t,ke(L(this.l1,ar(e))))),this.hasL2&&(t=le(t,ke(L(this.l2,Eh(e))))),U(t,[])})}getConfig(){return{l1:this.l1,l2:this.l2}}static fromConfig(e,t){return new e({l1:t.l1,l2:t.l2})}};Mh.className="L1L2";ue.registerClass(Mh);function _U(e){return HA(e),new Mh({l1:e!=null?e.l1:null,l2:0})}function zU(e){return HA(e),new Mh({l2:e!=null?e.l2:null,l1:0})}var b4={l1l2:"L1L2"};function xt(e){return wA(e)}function v4(e,t={}){return Th(e,ue.SerializationMap.getMap().classNameMap,t,"regularizer")}function $t(e){if(e==null)return null;if(typeof e=="string"){let t={className:e in b4?b4[e]:e,config:{}};return v4(t)}else return e instanceof G6?e:v4(e)}var qA=class extends st{constructor(e){super(e==null?{}:e),this.supportsMasking=!0,e!=null&&(this.maxValue=e.maxValue)}call(e,t){e=je(e);let r=Oa(e);return this.maxValue!=null&&(r=cn(r,0,this.maxValue)),r}computeOutputShape(e){return e}getConfig(){let e={maxValue:this.maxValue},t=super.getConfig();return Object.assign(e,t),e}};qA.className="ReLU";ue.registerClass(qA);var XA=class extends st{constructor(e){super(e==null?{}:e),this.DEFAULT_ALPHA=.3,e==null&&(e={}),this.alpha=e.alpha==null?this.DEFAULT_ALPHA:e.alpha}call(e,t){let r=je(e);return $f(r,this.alpha)}computeOutputShape(e){return e}getConfig(){let e={alpha:this.alpha},t=super.getConfig();return Object.assign(e,t),e}};XA.className="LeakyReLU";ue.registerClass(XA);var KA=class extends st{constructor(e){if(super(e==null?{}:e),this.DEFAULT_ALPHA_INITIALIZER="zeros",e==null&&(e={}),this.supportsMasking=!0,this.alphaInitializer=Rt(e.alphaInitializer||this.DEFAULT_ALPHA_INITIALIZER),this.alphaRegularizer=$t(e.alphaRegularizer),this.alphaConstraint=lr(e.alphaConstraint),e.sharedAxes==null)this.sharedAxes=null;else if(Array.isArray(e.sharedAxes))this.sharedAxes=e.sharedAxes;else if(typeof e.sharedAxes=="number")this.sharedAxes=[e.sharedAxes];else throw new H(`Expected sharedAxes to be a number or an array of numbers, but got ${e.sharedAxes}`)}build(e){e=mt(e);let t=e.slice(1);if(this.sharedAxes!=null)for(let n of this.sharedAxes)t[n-1]=1;this.alpha=this.addWeight("alpha",t,"float32",this.alphaInitializer,this.alphaRegularizer,!0,this.alphaConstraint);let r={};if(this.sharedAxes!=null)for(let n=1;n<e.length;++n)r[n]=e[n];this.inputSpec=[new Zt({ndim:e.length,axes:r})],this.built=!0}call(e,t){return e=je(e),Df(e,this.alpha.read())}getConfig(){let e={alphaInitializer:zt(this.alphaInitializer),alphaRegularizer:xt(this.alphaRegularizer),alphaConstraint:or(this.alphaConstraint),sharedAxes:this.sharedAxes},t=super.getConfig();return Object.assign(e,t),e}};KA.className="PReLU";ue.registerClass(KA);var ZA=class extends st{constructor(e){if(super(e==null?{}:e),this.DEFAULT_ALPHA=1,e==null&&(e={}),e.alpha!=null&&e.alpha!==this.DEFAULT_ALPHA)throw new Ve(`Non-default alpha value (${e.alpha}) is not supported by the ELU layer yet.`);this.alpha=e.alpha==null?this.DEFAULT_ALPHA:e.alpha}call(e,t){let r=je(e);return wh(r)}computeOutputShape(e){return e}getConfig(){let e={alpha:this.alpha},t=super.getConfig();return Object.assign(e,t),e}};ZA.className="ELU";ue.registerClass(ZA);var YA=class extends st{constructor(e){super(e==null?{}:e),this.DEFAULT_THETA=1,e==null&&(e={}),this.theta=e.theta==null?this.DEFAULT_THETA:e.theta}call(e,t){let r=je(e);return L(r,me(mn(r,this.theta),"float32"))}computeOutputShape(e){return e}getConfig(){let e={theta:this.theta},t=super.getConfig();return Object.assign(e,t),e}};YA.className="ThresholdedReLU";ue.registerClass(YA);var JA=class extends st{constructor(e){super(e==null?{}:e),this.DEFAULT_AXIS=1,e==null&&(e={}),this.softmax=new jA().apply,this.axis=e.axis==null?this.DEFAULT_AXIS:e.axis}call(e,t){let r=je(e);return this.softmax(r,this.axis)}computeOutputShape(e){return e}getConfig(){let e={axis:this.axis},t=super.getConfig();return Object.assign(e,t),e}};JA.className="Softmax";ue.registerClass(JA);function bu(e,t,r){if(typeof e=="number")return Do(e,t);if(e.length!==t)throw new H(`The ${r} argument must be an integer or tuple of ${t} integers. Received: ${e.length} elements.`);for(let n=0;n<t;++n){let a=e[n];if(!qL(a))throw new H(`The ${r} argument must be an integer or tuple of ${t} integers. Received: ${JSON.stringify(e)} including a non-integer number ${a}`)}return e}function ma(e,t,r,n,a=1){if(e==null)return e;let s=t+(t-1)*(a-1),i;return r==="same"?i=e:i=e-s+1,Math.floor((i+n-1)/n)}function Ea(e,t,r,n){if(e==null)return null;if(n==="valid")e=e*t+Us([r-t,0]);else if(n==="same")e=e*t;else throw new H(`Unsupport padding mode: ${n}.`);return e}function QA(e,t){return X(()=>(jt(t),t==="channelsFirst"?et(e,[0,2,3,1]):e))}function j6(e,t){return X(()=>(jt(t),t==="channelsFirst"?et(e,[0,2,3,4,1]):e))}function OU(e,t,r,n=1,a="valid",s,i=1){return X(()=>{if(s==null&&(s=Aa()),jt(s),e.shape.length!==3)throw new H(`The input of a conv1dWithBias operation should be 3, but is ${e.shape.length} instead.`);if(t.shape.length!==3)throw new H(`The kernel for a conv1dWithBias operation should be 3, but is ${t.shape.length} instead`);if(r!=null&&r.shape.length!==1)throw new H(`The bias for a conv1dWithBias operation should be 1, but is ${t.shape.length} instead`);if(s==="channelsFirst"&&(e=et(e,[0,2,1])),a==="causal")throw new Ve("The support for CAUSAL padding mode in conv1dWithBias is not implemented yet.");let o=Vy(e,t,n,a==="same"?"same":"valid","NWC",i);return r!=null&&(o=va(o,r)),o})}function w4(e,t,r,n=[1,1],a="valid",s,i,o=null){return X(()=>{if(s==null&&(s=Aa()),jt(s),e.rank!==3&&e.rank!==4)throw new H(`conv2dWithBiasActivation expects input to be of rank 3 or 4, but received ${e.rank}.`);if(t.rank!==3&&t.rank!==4)throw new H(`conv2dWithBiasActivation expects kernel to be of rank 3 or 4, but received ${e.rank}.`);let l=QA(e,s);if(a==="causal")throw new Ve("The support for CAUSAL padding mode in conv1dWithBias is not implemented yet.");return l=Vs.conv2d({x:l,filter:t,strides:n,pad:a==="same"?"same":"valid",dilations:i,dataFormat:"NHWC",bias:r,activation:o}),s==="channelsFirst"&&(l=et(l,[0,3,1,2])),l})}function DU(e,t,r,n=[1,1,1],a="valid",s,i){return X(()=>{if(s==null&&(s=Aa()),jt(s),e.rank!==4&&e.rank!==5)throw new H(`conv3dWithBias expects input to be of rank 4 or 5, but received ${e.rank}.`);if(t.rank!==4&&t.rank!==5)throw new H(`conv3dWithBias expects kernel to be of rank 4 or 5, but received ${e.rank}.`);let o=j6(e,s);if(a==="causal")throw new Ve("The support for CAUSAL padding mode in conv3dWithBias is not implemented yet.");return o=jy(o,t,n,a==="same"?"same":"valid","NDHWC",i),r!=null&&(o=va(o,r)),s==="channelsFirst"&&(o=et(o,[0,4,1,2,3])),o})}var ex=class extends st{constructor(e,t){if(super(t),this.bias=null,this.DEFAULT_KERNEL_INITIALIZER="glorotNormal",this.DEFAULT_BIAS_INITIALIZER="zeros",ex.verifyArgs(t),this.rank=e,yr(this.rank,"rank"),this.rank!==1&&this.rank!==2&&this.rank!==3)throw new Ve(`Convolution layer for rank other than 1, 2, or 3 (${this.rank}) is not implemented yet.`);if(this.kernelSize=bu(t.kernelSize,e,"kernelSize"),this.strides=bu(t.strides==null?1:t.strides,e,"strides"),this.padding=t.padding==null?"valid":t.padding,Ln(this.padding),this.dataFormat=t.dataFormat==null?"channelsLast":t.dataFormat,jt(this.dataFormat),this.activation=js(t.activation),this.useBias=t.useBias==null?!0:t.useBias,this.biasInitializer=Rt(t.biasInitializer||this.DEFAULT_BIAS_INITIALIZER),this.biasConstraint=lr(t.biasConstraint),this.biasRegularizer=$t(t.biasRegularizer),this.activityRegularizer=$t(t.activityRegularizer),this.dilationRate=bu(t.dilationRate==null?1:t.dilationRate,e,"dilationRate"),this.rank===1&&Array.isArray(this.dilationRate)&&this.dilationRate.length!==1)throw new H(`dilationRate must be a number or an array of a single number for 1D convolution, but received ${JSON.stringify(this.dilationRate)}`);if(this.rank===2){if(typeof this.dilationRate=="number")this.dilationRate=[this.dilationRate,this.dilationRate];else if(this.dilationRate.length!==2)throw new H(`dilationRate must be a number or array of two numbers for 2D convolution, but received ${JSON.stringify(this.dilationRate)}`)}else if(this.rank===3){if(typeof this.dilationRate=="number")this.dilationRate=[this.dilationRate,this.dilationRate,this.dilationRate];else if(this.dilationRate.length!==3)throw new H(`dilationRate must be a number or array of three numbers for 3D convolution, but received ${JSON.stringify(this.dilationRate)}`)}}static verifyArgs(e){if(Na("kernelSize"in e,"required key 'kernelSize' not in config"),typeof e.kernelSize!="number"&&!kA(e.kernelSize,"number",1,3))throw new H(`BaseConv expects config.kernelSize to be number or number[] with length 1, 2, or 3, but received ${JSON.stringify(e.kernelSize)}.`)}getConfig(){let e={kernelSize:this.kernelSize,strides:this.strides,padding:this.padding,dataFormat:this.dataFormat,dilationRate:this.dilationRate,activation:Gs(this.activation),useBias:this.useBias,biasInitializer:zt(this.biasInitializer),biasRegularizer:xt(this.biasRegularizer),activityRegularizer:xt(this.activityRegularizer),biasConstraint:or(this.biasConstraint)},t=super.getConfig();return Object.assign(e,t),e}},Fh=class extends ex{constructor(e,t){super(e,t),this.kernel=null,Fh.verifyArgs(t),this.filters=t.filters,yr(this.filters,"filters"),this.kernelInitializer=Rt(t.kernelInitializer||this.DEFAULT_KERNEL_INITIALIZER),this.kernelConstraint=lr(t.kernelConstraint),this.kernelRegularizer=$t(t.kernelRegularizer)}build(e){e=mt(e);let t=this.dataFormat==="channelsFirst"?1:e.length-1;if(e[t]==null)throw new H(`The channel dimension of the input should be defined. Found ${e[t]}`);let r=e[t],n=this.kernelSize.concat([r,this.filters]);this.kernel=this.addWeight("kernel",n,null,this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint),this.useBias&&(this.bias=this.addWeight("bias",[this.filters],null,this.biasInitializer,this.biasRegularizer,!0,this.biasConstraint)),this.inputSpec=[{ndim:this.rank+2,axes:{[t]:r}}],this.built=!0}call(e,t){return X(()=>{e=je(e);let r,n=this.bias==null?null:this.bias.read(),a=Yw(this.activation.getClassName());if(a!=null&&this.rank===2)r=w4(e,this.kernel.read(),n,this.strides,this.padding,this.dataFormat,this.dilationRate,a);else{if(this.rank===1)r=OU(e,this.kernel.read(),n,this.strides[0],this.padding,this.dataFormat,this.dilationRate[0]);else if(this.rank===2)r=w4(e,this.kernel.read(),n,this.strides,this.padding,this.dataFormat,this.dilationRate);else if(this.rank===3)r=DU(e,this.kernel.read(),n,this.strides,this.padding,this.dataFormat,this.dilationRate);else throw new Ve("convolutions greater than 3D are not implemented yet.");this.activation!=null&&(r=this.activation.apply(r))}return r})}computeOutputShape(e){e=mt(e);let t=[],r=this.dataFormat==="channelsLast"?e.slice(1,e.length-1):e.slice(2);for(let a=0;a<r.length;++a){let s=ma(r[a],this.kernelSize[a],this.padding,this.strides[a],typeof this.dilationRate=="number"?this.dilationRate:this.dilationRate[a]);t.push(s)}let n=[e[0]];return this.dataFormat==="channelsLast"?(n=n.concat(t),n.push(this.filters)):(n.push(this.filters),n=n.concat(t)),n}getConfig(){let e={filters:this.filters,kernelInitializer:zt(this.kernelInitializer),kernelRegularizer:xt(this.kernelRegularizer),kernelConstraint:or(this.kernelConstraint)},t=super.getConfig();return Object.assign(e,t),e}static verifyArgs(e){if(!("filters"in e)||typeof e.filters!="number"||e.filters<1)throw new H(`Convolution layer expected config.filters to be a 'number' > 0 but got ${JSON.stringify(e.filters)}`)}},H6=class extends Fh{constructor(e){super(2,e),H6.verifyArgs(e)}getConfig(){let e=super.getConfig();return delete e.rank,e}static verifyArgs(e){if(typeof e.kernelSize!="number"&&!kA(e.kernelSize,"number",1,2))throw new H(`Conv2D expects config.kernelSize to be number or number[] with length 1 or 2, but received ${JSON.stringify(e.kernelSize)}.`)}},cm=H6;cm.className="Conv2D";ue.registerClass(cm);var q6=class extends Fh{constructor(e){super(3,e),q6.verifyArgs(e)}getConfig(){let e=super.getConfig();return delete e.rank,e}static verifyArgs(e){if(typeof e.kernelSize!="number"&&!(Array.isArray(e.kernelSize)&&(e.kernelSize.length===1||e.kernelSize.length===3)))throw new H(`Conv3D expects config.kernelSize to be number or [number, number, number], but received ${JSON.stringify(e.kernelSize)}.`)}},fm=q6;fm.className="Conv3D";ue.registerClass(fm);var tx=class extends cm{constructor(e){if(super(e),this.inputSpec=[new Zt({ndim:4})],this.padding!=="same"&&this.padding!=="valid")throw new H(`Conv2DTranspose currently supports only padding modes 'same' and 'valid', but received padding mode ${this.padding}`)}build(e){if(e=mt(e),e.length!==4)throw new H("Input should have rank 4; Received input shape: "+JSON.stringify(e));let t=this.dataFormat==="channelsFirst"?1:e.length-1;if(e[t]==null)throw new H("The channel dimension of the inputs should be defined. Found `None`.");let r=e[t],n=this.kernelSize.concat([this.filters,r]);this.kernel=this.addWeight("kernel",n,"float32",this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint),this.useBias&&(this.bias=this.addWeight("bias",[this.filters],"float32",this.biasInitializer,this.biasRegularizer,!0,this.biasConstraint)),this.inputSpec=[new Zt({ndim:4,axes:{[t]:r}})],this.built=!0}call(e,t){return X(()=>{let r=je(e);if(r.shape.length!==4)throw new H(`Conv2DTranspose.call() expects input tensor to be rank-4, but received a tensor of rank-${r.shape.length}`);let n=r.shape,a=n[0],s,i;this.dataFormat==="channelsFirst"?(s=2,i=3):(s=1,i=2);let o=n[s],l=n[i],u=this.kernelSize[0],d=this.kernelSize[1],h=this.strides[0],p=this.strides[1],c=Ea(o,h,u,this.padding),f=Ea(l,p,d,this.padding),m=[a,c,f,this.filters];this.dataFormat!=="channelsLast"&&(r=et(r,[0,2,3,1]));let g=Gy(r,this.kernel.read(),m,this.strides,this.padding);return this.dataFormat!=="channelsLast"&&(g=et(g,[0,3,1,2])),this.bias!=null&&(g=va(g,this.bias.read(),this.dataFormat)),this.activation!=null&&(g=this.activation.apply(g)),g})}computeOutputShape(e){e=mt(e);let t=e.slice(),r,n,a;this.dataFormat==="channelsFirst"?(r=1,n=2,a=3):(r=3,n=1,a=2);let s=this.kernelSize[0],i=this.kernelSize[1],o=this.strides[0],l=this.strides[1];return t[r]=this.filters,t[n]=Ea(t[n],o,s,this.padding),t[a]=Ea(t[a],l,i,this.padding),t}getConfig(){let e=super.getConfig();return delete e.dilationRate,e}};tx.className="Conv2DTranspose";ue.registerClass(tx);var rx=class extends fm{constructor(e){if(super(e),this.inputSpec=[new Zt({ndim:5})],this.padding!=="same"&&this.padding!=="valid")throw new H(`Conv3DTranspose currently supports only padding modes 'same' and 'valid', but received padding mode ${this.padding}`)}build(e){if(e=mt(e),e.length!==5)throw new H("Input should have rank 5; Received input shape: "+JSON.stringify(e));let t=this.dataFormat==="channelsFirst"?1:e.length-1;if(e[t]==null)throw new H("The channel dimension of the inputs should be defined. Found `None`.");let r=e[t],n=this.kernelSize.concat([this.filters,r]);this.kernel=this.addWeight("kernel",n,"float32",this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint),this.useBias&&(this.bias=this.addWeight("bias",[this.filters],"float32",this.biasInitializer,this.biasRegularizer,!0,this.biasConstraint)),this.inputSpec=[new Zt({ndim:5,axes:{[t]:r}})],this.built=!0}call(e,t){return X(()=>{let r=je(e);if(r.shape.length!==5)throw new H(`Conv3DTranspose.call() expects input tensor to be rank-4, but received a tensor of rank-${r.shape.length}`);let n=r.shape,a=n[0],s,i,o;this.dataFormat==="channelsFirst"?(o=2,s=3,i=4):(o=1,s=2,i=3);let l=n[o],u=n[s],d=n[i],h=this.kernelSize[0],p=this.kernelSize[1],c=this.kernelSize[2],f=this.strides[0],m=this.strides[1],g=this.strides[2],y=Ea(l,f,h,this.padding),A=Ea(u,m,p,this.padding),x=Ea(d,g,c,this.padding),b=[a,y,A,x,this.filters];this.dataFormat!=="channelsLast"&&(r=et(r,[0,2,3,4,1]));let w=Q7(r,this.kernel.read(),b,this.strides,this.padding);return this.dataFormat!=="channelsLast"&&(w=et(w,[0,4,1,2,3])),this.bias!==null&&(w=va(w,this.bias.read(),this.dataFormat)),this.activation!==null&&(w=this.activation.apply(w)),w})}computeOutputShape(e){e=mt(e);let t=e.slice(),r,n,a,s;this.dataFormat==="channelsFirst"?(r=1,n=2,a=3,s=4):(r=4,n=1,a=2,s=3);let i=this.kernelSize[0],o=this.kernelSize[1],l=this.kernelSize[2],u=this.strides[0],d=this.strides[1],h=this.strides[2];return t[r]=this.filters,t[n]=Ea(t[n],u,i,this.padding),t[a]=Ea(t[a],d,o,this.padding),t[s]=Ea(t[s],h,l,this.padding),t}getConfig(){let e=super.getConfig();return delete e.dilationRate,e}};rx.className="Conv3DTranspose";ue.registerClass(rx);var X6=class extends Fh{constructor(e,t){if(super(e,t),this.DEFAULT_DEPTHWISE_INITIALIZER="glorotUniform",this.DEFAULT_POINTWISE_INITIALIZER="glorotUniform",this.depthwiseKernel=null,this.pointwiseKernel=null,t.filters==null)throw new H("The `filters` configuration field is required by SeparableConv, but is unspecified.");if(t.kernelInitializer!=null||t.kernelRegularizer!=null||t.kernelConstraint!=null)throw new H("Fields kernelInitializer, kernelRegularizer and kernelConstraint are invalid for SeparableConv2D. Use depthwiseInitializer, depthwiseRegularizer, depthwiseConstraint, pointwiseInitializer, pointwiseRegularizer and pointwiseConstraint instead.");if(t.padding!=null&&t.padding!=="same"&&t.padding!=="valid")throw new H(`SeparableConv${this.rank}D supports only padding modes: 'same' and 'valid', but received ${JSON.stringify(t.padding)}`);this.depthMultiplier=t.depthMultiplier==null?1:t.depthMultiplier,this.depthwiseInitializer=Rt(t.depthwiseInitializer||this.DEFAULT_DEPTHWISE_INITIALIZER),this.depthwiseRegularizer=$t(t.depthwiseRegularizer),this.depthwiseConstraint=lr(t.depthwiseConstraint),this.pointwiseInitializer=Rt(t.depthwiseInitializer||this.DEFAULT_POINTWISE_INITIALIZER),this.pointwiseRegularizer=$t(t.pointwiseRegularizer),this.pointwiseConstraint=lr(t.pointwiseConstraint)}build(e){if(e=mt(e),e.length<this.rank+2)throw new H(`Inputs to SeparableConv${this.rank}D should have rank ${this.rank+2}, but received input shape: ${JSON.stringify(e)}`);let t=this.dataFormat==="channelsFirst"?1:e.length-1;if(e[t]==null||e[t]<0)throw new H(`The channel dimension of the inputs should be defined, but found ${JSON.stringify(e[t])}`);let r=e[t],n=this.kernelSize.concat([r,this.depthMultiplier]),a=[];for(let i=0;i<this.rank;++i)a.push(1);a.push(r*this.depthMultiplier,this.filters);let s=!0;this.depthwiseKernel=this.addWeight("depthwise_kernel",n,"float32",this.depthwiseInitializer,this.depthwiseRegularizer,s,this.depthwiseConstraint),this.pointwiseKernel=this.addWeight("pointwise_kernel",a,"float32",this.pointwiseInitializer,this.pointwiseRegularizer,s,this.pointwiseConstraint),this.useBias?this.bias=this.addWeight("bias",[this.filters],"float32",this.biasInitializer,this.biasRegularizer,s,this.biasConstraint):this.bias=null,this.inputSpec=[new Zt({ndim:this.rank+2,axes:{[t]:r}})],this.built=!0}call(e,t){return X(()=>{e=je(e);let r;if(this.rank===1)throw new Ve("1D separable convolution is not implemented yet.");return this.rank===2&&(this.dataFormat==="channelsFirst"&&(e=et(e,[0,2,3,1])),r=vw(e,this.depthwiseKernel.read(),this.pointwiseKernel.read(),this.strides,this.padding,this.dilationRate,"NHWC")),this.useBias&&(r=va(r,this.bias.read(),this.dataFormat)),this.activation!=null&&(r=this.activation.apply(r)),this.dataFormat==="channelsFirst"&&(r=et(r,[0,3,1,2])),r})}getConfig(){let e=super.getConfig();return delete e.rank,delete e.kernelInitializer,delete e.kernelRegularizer,delete e.kernelConstraint,e.depthwiseInitializer=zt(this.depthwiseInitializer),e.pointwiseInitializer=zt(this.pointwiseInitializer),e.depthwiseRegularizer=xt(this.depthwiseRegularizer),e.pointwiseRegularizer=xt(this.pointwiseRegularizer),e.depthwiseConstraint=or(this.depthwiseConstraint),e.pointwiseConstraint=or(this.pointwiseConstraint),e}};X6.className="SeparableConv";var nx=class extends X6{constructor(e){super(2,e)}};nx.className="SeparableConv2D";ue.registerClass(nx);var K6=class extends Fh{constructor(e){super(1,e),K6.verifyArgs(e),this.inputSpec=[{ndim:3}]}getConfig(){let e=super.getConfig();return delete e.rank,delete e.dataFormat,e}static verifyArgs(e){if(typeof e.kernelSize!="number"&&!kA(e.kernelSize,"number",1,1))throw new H(`Conv1D expects config.kernelSize to be number or number[] with length 1, but received ${JSON.stringify(e.kernelSize)}.`)}},ax=K6;ax.className="Conv1D";ue.registerClass(ax);var sx=class extends st{constructor(e){super(e),typeof e.cropping=="number"?this.cropping=[[e.cropping,e.cropping],[e.cropping,e.cropping]]:typeof e.cropping[0]=="number"?this.cropping=[[e.cropping[0],e.cropping[0]],[e.cropping[1],e.cropping[1]]]:this.cropping=e.cropping,this.dataFormat=e.dataFormat===void 0?"channelsLast":e.dataFormat,this.inputSpec=[{ndim:4}]}computeOutputShape(e){return this.dataFormat==="channelsFirst"?[e[0],e[1],e[2]-this.cropping[0][0]-this.cropping[0][1],e[3]-this.cropping[1][0]-this.cropping[1][1]]:[e[0],e[1]-this.cropping[0][0]-this.cropping[0][1],e[2]-this.cropping[1][0]-this.cropping[1][1],e[3]]}call(e,t){return X(()=>{if(e=je(e),this.dataFormat==="channelsLast"){let r=Zc(e,this.cropping[0][0],e.shape[1]-this.cropping[0][0]-this.cropping[0][1],2);return Zc(r,this.cropping[1][0],e.shape[2]-this.cropping[1][1]-this.cropping[1][0],3)}else{let r=Zc(e,this.cropping[0][0],e.shape[2]-this.cropping[0][0]-this.cropping[0][1],3);return Zc(r,this.cropping[1][0],e.shape[3]-this.cropping[1][1]-this.cropping[1][0],4)}})}getConfig(){let e={cropping:this.cropping,dataFormat:this.dataFormat},t=super.getConfig();return Object.assign(e,t),e}};sx.className="Cropping2D";ue.registerClass(sx);var ix=class extends st{constructor(e){super(e),this.DEFAULT_SIZE=[2,2],this.inputSpec=[{ndim:4}],this.size=e.size==null?this.DEFAULT_SIZE:e.size,this.dataFormat=e.dataFormat==null?"channelsLast":e.dataFormat,jt(this.dataFormat),this.interpolation=e.interpolation==null?"nearest":e.interpolation,GL(this.interpolation)}computeOutputShape(e){if(this.dataFormat==="channelsFirst"){let t=e[2]==null?null:this.size[0]*e[2],r=e[3]==null?null:this.size[1]*e[3];return[e[0],e[1],t,r]}else{let t=e[1]==null?null:this.size[0]*e[1],r=e[2]==null?null:this.size[1]*e[2];return[e[0],t,r,e[3]]}}call(e,t){return X(()=>{let r=je(e),n=r.shape;if(this.dataFormat==="channelsFirst"){r=et(r,[0,2,3,1]);let a=this.size[0]*n[2],s=this.size[1]*n[3],i=this.interpolation==="nearest"?Ie.resizeNearestNeighbor(r,[a,s]):Ie.resizeBilinear(r,[a,s]);return et(i,[0,3,1,2])}else{let a=this.size[0]*n[1],s=this.size[1]*n[2];return this.interpolation==="nearest"?Ie.resizeNearestNeighbor(r,[a,s]):Ie.resizeBilinear(r,[a,s])}})}getConfig(){let e={size:this.size,dataFormat:this.dataFormat,interpolation:this.interpolation},t=super.getConfig();return Object.assign(e,t),e}};ix.className="UpSampling2D";ue.registerClass(ix);function LU(e,t,r=[1,1],n="valid",a,s){return X(()=>{a==null&&(a=Aa()),jt(a);let i=QA(e,a);if(e.rank!==4)throw new H(`Input for depthwiseConv2d is required to be 4-D, but is instead ${e.rank}-D`);if(t.rank!==4)throw new H(`depthwiseKernel is required to be 4-D, but is instead ${t.rank}-D`);return i=vh(i,t,r,n==="same"?"same":"valid","NHWC",s),a==="channelsFirst"&&(i=et(i,[0,3,1,2])),i})}var ox=class extends ex{constructor(e){super(2,e),this.depthwiseKernel=null,this.depthMultiplier=e.depthMultiplier==null?1:e.depthMultiplier,this.depthwiseInitializer=Rt(e.depthwiseInitializer||this.DEFAULT_KERNEL_INITIALIZER),this.depthwiseConstraint=lr(e.depthwiseConstraint),this.depthwiseRegularizer=$t(e.depthwiseRegularizer)}build(e){if(e=mt(e),e.length<4)throw new H(`Inputs to DepthwiseConv2D should have rank 4. Received input shape: ${JSON.stringify(e)}.`);let t=this.dataFormat==="channelsFirst"?1:3;if(e[t]==null||e[t]<0)throw new H(`The channel dimension of the inputs to DepthwiseConv2D should be defined, but is not (${e[t]}).`);let r=e[t],n=[this.kernelSize[0],this.kernelSize[1],r,this.depthMultiplier];this.depthwiseKernel=this.addWeight("depthwise_kernel",n,null,this.depthwiseInitializer,this.depthwiseRegularizer,!0,this.depthwiseConstraint),this.useBias?this.bias=this.addWeight("bias",[r*this.depthMultiplier],null,this.biasInitializer,this.biasRegularizer,!0,this.biasConstraint):this.bias=null,this.built=!0}call(e,t){return X(()=>{e=je(e);let r=LU(e,this.depthwiseKernel.read(),this.strides,this.padding,this.dataFormat,null);return this.useBias&&(r=va(r,this.bias.read(),this.dataFormat)),this.activation!=null&&(r=this.activation.apply(r)),r})}computeOutputShape(e){e=mt(e);let t=this.dataFormat==="channelsFirst"?e[2]:e[1],r=this.dataFormat==="channelsFirst"?e[3]:e[2],n=this.dataFormat==="channelsFirst"?e[1]*this.depthMultiplier:e[3]*this.depthMultiplier,a=ma(t,this.kernelSize[0],this.padding,this.strides[0]),s=ma(r,this.kernelSize[1],this.padding,this.strides[1]);return this.dataFormat==="channelsFirst"?[e[0],n,a,s]:[e[0],a,s,n]}getConfig(){let e=super.getConfig();return e.depthMultiplier=this.depthMultiplier,e.depthwiseInitializer=zt(this.depthwiseInitializer),e.depthwiseRegularizer=xt(this.depthwiseRegularizer),e.depthwiseConstraint=or(this.depthwiseRegularizer),e}};ox.className="DepthwiseConv2D";ue.registerClass(ox);function Z6(e,t,r,n){if(Array.isArray(e)){if(t!=null||r!=null)throw new H("When inputs is an array, neither initialState or constants should be provided");n!=null&&(r=e.slice(e.length-n,e.length),e=e.slice(0,e.length-n)),e.length>1&&(t=e.slice(1,e.length)),e=e[0]}function a(s){return s==null||Array.isArray(s)?s:[s]}return t=a(t),r=a(r),{inputs:e,initialState:t,constants:r}}function Y6(e,t,r,n=!1,a,s,i=!1,o=!1){return X(()=>{let l=t.shape.length;if(l<3)throw new H(`Input should be at least 3D, but is ${l}D.`);let u=[1,0].concat(ya(2,l));if(t=et(t,u),s!=null)throw new Ve("The rnn() functoin of the deeplearn.js backend does not support constants yet.");i&&console.warn("Backend rnn(): the unroll = true option is not applicable to the imperative deeplearn.js backend."),a!=null&&(a=me(me(a,"bool"),"float32"),a.rank===l-1&&(a=Kt(a,-1)),a=et(a,u)),n&&(t=_n(t,0),a!=null&&(a=_n(a,0)));let d=[],h,p=r,c=t.shape[0],f=nn(t),m;a!=null&&(m=nn(a));for(let y=0;y<c;++y){let A=f[y],x=X(()=>e(A,p));if(a==null)h=x[0],p=x[1];else{let b=X(()=>{let w=m[y],S=ce(Pn(w),w),N=le(L(x[0],w),L(p[0],S)),E=p.map((R,z)=>le(L(x[1][z],w),L(R,S)));return{output:N,newStates:E}});h=b.output,p=b.newStates}o&&d.push(h)}let g;return o&&(g=ur(d,1)),[h,g,p]})}var J6=class extends st{constructor(e){super(e);let t;if(e.cell==null)throw new H("cell property is missing for the constructor of RNN.");if(Array.isArray(e.cell)?t=new ym({cells:e.cell}):t=e.cell,t.stateSize==null)throw new H("The RNN cell should have an attribute `stateSize` (tuple of integers, one integer per RNN state).");this.cell=t,this.returnSequences=e.returnSequences==null?!1:e.returnSequences,this.returnState=e.returnState==null?!1:e.returnState,this.goBackwards=e.goBackwards==null?!1:e.goBackwards,this._stateful=e.stateful==null?!1:e.stateful,this.unroll=e.unroll==null?!1:e.unroll,this.supportsMasking=!0,this.inputSpec=[new Zt({ndim:3})],this.stateSpec=null,this.states_=null,this.numConstants=null,this.keptStates=[]}getStates(){if(this.states_==null){let e=Array.isArray(this.cell.stateSize)?this.cell.stateSize.length:1;return ya(0,e).map(t=>null)}else return this.states_}setStates(e){this.states_=e}computeOutputShape(e){Mg(e)&&(e=e[0]),e=e;let t=this.cell.stateSize;Array.isArray(t)||(t=[t]);let r=t[0],n;if(this.returnSequences?n=[e[0],e[1],r]:n=[e[0],r],this.returnState){let a=[];for(let s of t)a.push([e[0],s]);return[n].concat(a)}else return n}computeMask(e,t){return X(()=>{Array.isArray(t)&&(t=t[0]);let r=this.returnSequences?t:null;if(this.returnState){let n=this.states.map(a=>null);return[r].concat(n)}else return r})}get states(){if(this.states_==null){let e=Array.isArray(this.cell.stateSize)?this.cell.stateSize.length:1,t=[];for(let r=0;r<e;++r)t.push(null);return t}else return this.states_}set states(e){this.states_=e}build(e){if(this.numConstants!=null)throw new Ve("Constants support is not implemented in RNN yet.");Mg(e)&&(e=e[0]),e=e;let t=this.stateful?e[0]:null,r=e.slice(2);this.inputSpec[0]=new Zt({shape:[t,null,...r]});let n=[e[0]].concat(e.slice(2));this.cell.build(n);let a;if(Array.isArray(this.cell.stateSize)?a=this.cell.stateSize:a=[this.cell.stateSize],this.stateSpec!=null){if(!v.arraysEqual(this.stateSpec.map(s=>s.shape[s.shape.length-1]),a))throw new H(`An initialState was passed that is not compatible with cell.stateSize. Received stateSpec=${this.stateSpec}; However cell.stateSize is ${this.cell.stateSize}`)}else this.stateSpec=a.map(s=>new Zt({shape:[null,s]}));this.stateful&&this.resetStates()}resetStates(e,t=!1){X(()=>{if(!this.stateful)throw new Ga("Cannot call resetStates() on an RNN Layer that is not stateful.");let r=this.inputSpec[0].shape[0];if(r==null)throw new H("If an RNN is stateful, it needs to know its batch size. Specify the batch size of your input tensors: \n- If using a Sequential model, specify the batch size by passing a `batchInputShape` option to your first layer.\n- If using the functional API, specify the batch size by passing a `batchShape` option to your Input layer.");if(this.states_==null)Array.isArray(this.cell.stateSize)?this.states_=this.cell.stateSize.map(n=>_t([r,n])):this.states_=[_t([r,this.cell.stateSize])];else if(e==null)te(this.states_),this.keptStates!=null&&(te(this.keptStates),this.keptStates=[]),Array.isArray(this.cell.stateSize)?this.states_=this.cell.stateSize.map(n=>_t([r,n])):this.states_[0]=_t([r,this.cell.stateSize]);else{if(Array.isArray(e)||(e=[e]),e.length!==this.states_.length)throw new H(`Layer ${this.name} expects ${this.states_.length} state(s), but it received ${e.length} state value(s). Input received: ${e}`);t===!0?this.keptStates.push(this.states_.slice()):te(this.states_);for(let n=0;n<this.states_.length;++n){let a=e[n],s=Array.isArray(this.cell.stateSize)?this.cell.stateSize[n]:this.cell.stateSize,i=[r,s];if(!v.arraysEqual(a.shape,i))throw new H(`State ${n} is incompatible with layer ${this.name}: expected shape=${i}, received shape=${a.shape}`);this.states_[n]=a}}this.states_=this.states_.map(n=>gr(n.clone()))})}apply(e,t){let r=t==null?null:t.initialState,n=t==null?null:t.constants;t==null&&(t={});let a=Z6(e,r,n,this.numConstants);e=a.inputs,r=a.initialState,n=a.constants;let s=[],i=[];if(r!=null){t.initialState=r,s=s.concat(r),this.stateSpec=[];for(let o of r)this.stateSpec.push(new Zt({shape:o.shape}));i=i.concat(this.stateSpec)}if(n!=null&&(t.constants=n,s=s.concat(n),this.numConstants=n.length),s[0]instanceof pa){let o=[e].concat(s),l=this.inputSpec.concat(i),u=this.inputSpec;this.inputSpec=l;let d=super.apply(o,t);return this.inputSpec=u,d}else return super.apply(e,t)}call(e,t){return X(()=>{let r=t==null?null:t.mask,n=t==null?null:t.training,a=t==null?null:t.initialState;e=je(e),a==null&&(this.stateful?a=this.states_:a=this.getInitialState(e));let s=Array.isArray(this.cell.stateSize)?this.cell.stateSize.length:1;if(a.length!==s)throw new H(`RNN Layer has ${s} state(s) but was passed ${a.length} initial state(s).`);this.unroll&&console.warn("Ignoring unroll = true for RNN layer, due to imperative backend.");let i={training:n},o=Y6((p,c)=>{let f=this.cell.call([p].concat(c),i);return[f[0],f.slice(1)]},e,a,this.goBackwards,r,null,this.unroll,this.returnSequences),l=o[0],u=o[1],d=o[2];this.stateful&&this.resetStates(d,n);let h=this.returnSequences?u:l;return this.returnState?[h].concat(d):h})}getInitialState(e){return X(()=>{let t=_t(e.shape);return t=ke(t,[1,2]),t=Nh(t),Array.isArray(this.cell.stateSize)?this.cell.stateSize.map(r=>r>1?Rg(t,[1,r]):t):this.cell.stateSize>1?[Rg(t,[1,this.cell.stateSize])]:[t]})}get trainableWeights(){return this.trainable?this.cell.trainableWeights:[]}get nonTrainableWeights(){return this.trainable?this.cell.nonTrainableWeights:this.cell.weights}setFastWeightInitDuringBuild(e){super.setFastWeightInitDuringBuild(e),this.cell!=null&&this.cell.setFastWeightInitDuringBuild(e)}getConfig(){let e=super.getConfig(),t={returnSequences:this.returnSequences,returnState:this.returnState,goBackwards:this.goBackwards,stateful:this.stateful,unroll:this.unroll};this.numConstants!=null&&(t.numConstants=this.numConstants);let r=this.cell.getConfig();return this.getClassName()===J6.className&&(t.cell={className:this.cell.getClassName(),config:r}),{...r,...e,...t}}static fromConfig(e,t,r={}){let n=t.cell,a=fa(n,r);return new e(Object.assign(t,{cell:a}))}},as=J6;as.className="RNN";ue.registerClass(as);var Ph=class extends st{},mm=class extends Ph{constructor(e){super(e),this.DEFAULT_ACTIVATION="tanh",this.DEFAULT_KERNEL_INITIALIZER="glorotNormal",this.DEFAULT_RECURRENT_INITIALIZER="orthogonal",this.DEFAULT_BIAS_INITIALIZER="zeros",this.units=e.units,yr(this.units,"units"),this.activation=js(e.activation==null?this.DEFAULT_ACTIVATION:e.activation),this.useBias=e.useBias==null?!0:e.useBias,this.kernelInitializer=Rt(e.kernelInitializer||this.DEFAULT_KERNEL_INITIALIZER),this.recurrentInitializer=Rt(e.recurrentInitializer||this.DEFAULT_RECURRENT_INITIALIZER),this.biasInitializer=Rt(e.biasInitializer||this.DEFAULT_BIAS_INITIALIZER),this.kernelRegularizer=$t(e.kernelRegularizer),this.recurrentRegularizer=$t(e.recurrentRegularizer),this.biasRegularizer=$t(e.biasRegularizer),this.kernelConstraint=lr(e.kernelConstraint),this.recurrentConstraint=lr(e.recurrentConstraint),this.biasConstraint=lr(e.biasConstraint),this.dropout=Eu([1,Us([0,e.dropout==null?0:e.dropout])]),this.recurrentDropout=Eu([1,Us([0,e.recurrentDropout==null?0:e.recurrentDropout])]),this.dropoutFunc=e.dropoutFunc,this.stateSize=this.units,this.dropoutMask=null,this.recurrentDropoutMask=null}build(e){e=mt(e),this.kernel=this.addWeight("kernel",[e[e.length-1],this.units],null,this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint),this.recurrentKernel=this.addWeight("recurrent_kernel",[this.units,this.units],null,this.recurrentInitializer,this.recurrentRegularizer,!0,this.recurrentConstraint),this.useBias?this.bias=this.addWeight("bias",[this.units],null,this.biasInitializer,this.biasRegularizer,!0,this.biasConstraint):this.bias=null,this.built=!0}call(e,t){return X(()=>{if(e=e,e.length!==2)throw new H(`SimpleRNNCell expects 2 input Tensors, got ${e.length}.`);let r=e[1];e=e[0];let n=t.training==null?!1:t.training;0<this.dropout&&this.dropout<1&&this.dropoutMask==null&&(this.dropoutMask=Hs({ones:()=>Pn(e),rate:this.dropout,training:n,dropoutFunc:this.dropoutFunc})),0<this.recurrentDropout&&this.recurrentDropout<1&&this.recurrentDropoutMask==null&&(this.recurrentDropoutMask=Hs({ones:()=>Pn(r),rate:this.recurrentDropout,training:n,dropoutFunc:this.dropoutFunc}));let a,s=this.dropoutMask,i=this.recurrentDropoutMask;s!=null?a=$a(L(e,s),this.kernel.read()):a=$a(e,this.kernel.read()),this.bias!=null&&(a=va(a,this.bias.read())),i!=null&&(r=L(r,i));let o=le(a,$a(r,this.recurrentKernel.read()));return this.activation!=null&&(o=this.activation.apply(o)),[o,o]})}getConfig(){let e=super.getConfig(),t={units:this.units,activation:Gs(this.activation),useBias:this.useBias,kernelInitializer:zt(this.kernelInitializer),recurrentInitializer:zt(this.recurrentInitializer),biasInitializer:zt(this.biasInitializer),kernelRegularizer:xt(this.kernelRegularizer),recurrentRegularizer:xt(this.recurrentRegularizer),biasRegularizer:xt(this.biasRegularizer),activityRegularizer:xt(this.activityRegularizer),kernelConstraint:or(this.kernelConstraint),recurrentConstraint:or(this.recurrentConstraint),biasConstraint:or(this.biasConstraint),dropout:this.dropout,recurrentDropout:this.recurrentDropout};return{...e,...t}}};mm.className="SimpleRNNCell";ue.registerClass(mm);var lx=class extends as{constructor(e){e.cell=new mm(e),super(e)}call(e,t){return X(()=>{this.cell.dropoutMask!=null&&(te(this.cell.dropoutMask),this.cell.dropoutMask=null),this.cell.recurrentDropoutMask!=null&&(te(this.cell.recurrentDropoutMask),this.cell.recurrentDropoutMask=null);let r=t==null?null:t.mask,n=t==null?null:t.training,a=t==null?null:t.initialState;return super.call(e,{mask:r,training:n,initialState:a})})}static fromConfig(e,t){return new e(t)}};lx.className="SimpleRNN";ue.registerClass(lx);var gm=class extends Ph{constructor(e){if(super(e),this.DEFAULT_ACTIVATION="tanh",this.DEFAULT_RECURRENT_ACTIVATION="hardSigmoid",this.DEFAULT_KERNEL_INITIALIZER="glorotNormal",this.DEFAULT_RECURRENT_INITIALIZER="orthogonal",this.DEFAULT_BIAS_INITIALIZER="zeros",e.resetAfter)throw new H("GRUCell does not support reset_after parameter set to true.");this.units=e.units,yr(this.units,"units"),this.activation=js(e.activation===void 0?this.DEFAULT_ACTIVATION:e.activation),this.recurrentActivation=js(e.recurrentActivation===void 0?this.DEFAULT_RECURRENT_ACTIVATION:e.recurrentActivation),this.useBias=e.useBias==null?!0:e.useBias,this.kernelInitializer=Rt(e.kernelInitializer||this.DEFAULT_KERNEL_INITIALIZER),this.recurrentInitializer=Rt(e.recurrentInitializer||this.DEFAULT_RECURRENT_INITIALIZER),this.biasInitializer=Rt(e.biasInitializer||this.DEFAULT_BIAS_INITIALIZER),this.kernelRegularizer=$t(e.kernelRegularizer),this.recurrentRegularizer=$t(e.recurrentRegularizer),this.biasRegularizer=$t(e.biasRegularizer),this.kernelConstraint=lr(e.kernelConstraint),this.recurrentConstraint=lr(e.recurrentConstraint),this.biasConstraint=lr(e.biasConstraint),this.dropout=Eu([1,Us([0,e.dropout==null?0:e.dropout])]),this.recurrentDropout=Eu([1,Us([0,e.recurrentDropout==null?0:e.recurrentDropout])]),this.dropoutFunc=e.dropoutFunc,this.implementation=e.implementation,this.stateSize=this.units,this.dropoutMask=null,this.recurrentDropoutMask=null}build(e){e=mt(e);let t=e[e.length-1];this.kernel=this.addWeight("kernel",[t,this.units*3],null,this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint),this.recurrentKernel=this.addWeight("recurrent_kernel",[this.units,this.units*3],null,this.recurrentInitializer,this.recurrentRegularizer,!0,this.recurrentConstraint),this.useBias?this.bias=this.addWeight("bias",[this.units*3],null,this.biasInitializer,this.biasRegularizer,!0,this.biasConstraint):this.bias=null,this.built=!0}call(e,t){return X(()=>{if(e=e,e.length!==2)throw new H(`GRUCell expects 2 input Tensors (inputs, h, c), got ${e.length}.`);let r=t.training==null?!1:t.training,n=e[1];e=e[0],0<this.dropout&&this.dropout<1&&this.dropoutMask==null&&(this.dropoutMask=Hs({ones:()=>Pn(e),rate:this.dropout,training:r,count:3,dropoutFunc:this.dropoutFunc})),0<this.recurrentDropout&&this.recurrentDropout<1&&this.recurrentDropoutMask==null&&(this.recurrentDropoutMask=Hs({ones:()=>Pn(n),rate:this.recurrentDropout,training:r,count:3,dropoutFunc:this.dropoutFunc}));let a=this.dropoutMask,s=this.recurrentDropoutMask,i,o,l;0<this.dropout&&this.dropout<1&&(e=L(e,a[0]));let u=$a(e,this.kernel.read());this.useBias&&(u=va(u,this.bias.read())),0<this.recurrentDropout&&this.recurrentDropout<1&&(n=L(n,s[0]));let d=this.recurrentKernel.read(),[h,p]=Yt(d,[2*this.units,this.units],d.rank-1),c=$a(n,h),[f,m,g]=Yt(u,3,u.rank-1),[y,A]=Yt(c,2,c.rank-1);i=this.recurrentActivation.apply(le(f,y)),o=this.recurrentActivation.apply(le(m,A));let x=$a(L(o,n),p);l=this.activation.apply(le(g,x));let b=le(L(i,n),L(le(1,Dt(i)),l));return[b,b]})}getConfig(){let e=super.getConfig(),t={units:this.units,activation:Gs(this.activation),recurrentActivation:Gs(this.recurrentActivation),useBias:this.useBias,kernelInitializer:zt(this.kernelInitializer),recurrentInitializer:zt(this.recurrentInitializer),biasInitializer:zt(this.biasInitializer),kernelRegularizer:xt(this.kernelRegularizer),recurrentRegularizer:xt(this.recurrentRegularizer),biasRegularizer:xt(this.biasRegularizer),activityRegularizer:xt(this.activityRegularizer),kernelConstraint:or(this.kernelConstraint),recurrentConstraint:or(this.recurrentConstraint),biasConstraint:or(this.biasConstraint),dropout:this.dropout,recurrentDropout:this.recurrentDropout,implementation:this.implementation,resetAfter:!1};return{...e,...t}}};gm.className="GRUCell";ue.registerClass(gm);var ux=class extends as{constructor(e){e.implementation===0&&console.warn("`implementation=0` has been deprecated, and now defaults to `implementation=1`. Please update your layer call."),e.cell=new gm(e),super(e)}call(e,t){return X(()=>{this.cell.dropoutMask!=null&&(te(this.cell.dropoutMask),this.cell.dropoutMask=null),this.cell.recurrentDropoutMask!=null&&(te(this.cell.recurrentDropoutMask),this.cell.recurrentDropoutMask=null);let r=t==null?null:t.mask,n=t==null?null:t.training,a=t==null?null:t.initialState;return super.call(e,{mask:r,training:n,initialState:a})})}static fromConfig(e,t){return t.implmentation===0&&(t.implementation=1),new e(t)}};ux.className="GRU";ue.registerClass(ux);var _h=class extends Ph{constructor(e){super(e),this.DEFAULT_ACTIVATION="tanh",this.DEFAULT_RECURRENT_ACTIVATION="hardSigmoid",this.DEFAULT_KERNEL_INITIALIZER="glorotNormal",this.DEFAULT_RECURRENT_INITIALIZER="orthogonal",this.DEFAULT_BIAS_INITIALIZER="zeros",this.units=e.units,yr(this.units,"units"),this.activation=js(e.activation===void 0?this.DEFAULT_ACTIVATION:e.activation),this.recurrentActivation=js(e.recurrentActivation===void 0?this.DEFAULT_RECURRENT_ACTIVATION:e.recurrentActivation),this.useBias=e.useBias==null?!0:e.useBias,this.kernelInitializer=Rt(e.kernelInitializer||this.DEFAULT_KERNEL_INITIALIZER),this.recurrentInitializer=Rt(e.recurrentInitializer||this.DEFAULT_RECURRENT_INITIALIZER),this.biasInitializer=Rt(e.biasInitializer||this.DEFAULT_BIAS_INITIALIZER),this.unitForgetBias=e.unitForgetBias,this.kernelRegularizer=$t(e.kernelRegularizer),this.recurrentRegularizer=$t(e.recurrentRegularizer),this.biasRegularizer=$t(e.biasRegularizer),this.kernelConstraint=lr(e.kernelConstraint),this.recurrentConstraint=lr(e.recurrentConstraint),this.biasConstraint=lr(e.biasConstraint),this.dropout=Eu([1,Us([0,e.dropout==null?0:e.dropout])]),this.recurrentDropout=Eu([1,Us([0,e.recurrentDropout==null?0:e.recurrentDropout])]),this.dropoutFunc=e.dropoutFunc,this.implementation=e.implementation,this.stateSize=[this.units,this.units],this.dropoutMask=null,this.recurrentDropoutMask=null}build(e){var t;e=mt(e);let r=e[e.length-1];this.kernel=this.addWeight("kernel",[r,this.units*4],null,this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint),this.recurrentKernel=this.addWeight("recurrent_kernel",[this.units,this.units*4],null,this.recurrentInitializer,this.recurrentRegularizer,!0,this.recurrentConstraint);let n;if(this.useBias){if(this.unitForgetBias){let a=this.biasInitializer,s=this.units;n=new(t=class extends Zn{apply(i,o){let l=a.apply([s]),u=new rm().apply([s]),d=a.apply([s*2]);return e4(e4(l,u),d)}},t.className="CustomInit",t)}else n=this.biasInitializer;this.bias=this.addWeight("bias",[this.units*4],null,n,this.biasRegularizer,!0,this.biasConstraint)}else this.bias=null;this.built=!0}call(e,t){return X(()=>{let r=t.training==null?!1:t.training;if(e=e,e.length!==3)throw new H(`LSTMCell expects 3 input Tensors (inputs, h, c), got ${e.length}.`);let n=e[1],a=e[2];e=e[0],0<this.dropout&&this.dropout<1&&this.dropoutMask==null&&(this.dropoutMask=Hs({ones:()=>Pn(e),rate:this.dropout,training:r,count:4,dropoutFunc:this.dropoutFunc})),0<this.recurrentDropout&&this.recurrentDropout<1&&this.recurrentDropoutMask==null&&(this.recurrentDropoutMask=Hs({ones:()=>Pn(n),rate:this.recurrentDropout,training:r,count:4,dropoutFunc:this.dropoutFunc}));let s=this.dropoutMask,i=this.recurrentDropoutMask,o,l,u,d;0<this.dropout&&this.dropout<1&&(e=L(e,s[0]));let h=$a(e,this.kernel.read());0<this.recurrentDropout&&this.recurrentDropout<1&&(n=L(n,i[0])),h=le(h,$a(n,this.recurrentKernel.read())),this.useBias&&(h=va(h,this.bias.read()));let[p,c,f,m]=Yt(h,4,h.rank-1);o=this.recurrentActivation.apply(p),l=this.recurrentActivation.apply(c),u=le(L(l,a),L(o,this.activation.apply(f))),d=this.recurrentActivation.apply(m);let g=L(d,this.activation.apply(u));return[g,g,u]})}getConfig(){let e=super.getConfig(),t={units:this.units,activation:Gs(this.activation),recurrentActivation:Gs(this.recurrentActivation),useBias:this.useBias,kernelInitializer:zt(this.kernelInitializer),recurrentInitializer:zt(this.recurrentInitializer),biasInitializer:zt(this.biasInitializer),unitForgetBias:this.unitForgetBias,kernelRegularizer:xt(this.kernelRegularizer),recurrentRegularizer:xt(this.recurrentRegularizer),biasRegularizer:xt(this.biasRegularizer),activityRegularizer:xt(this.activityRegularizer),kernelConstraint:or(this.kernelConstraint),recurrentConstraint:or(this.recurrentConstraint),biasConstraint:or(this.biasConstraint),dropout:this.dropout,recurrentDropout:this.recurrentDropout,implementation:this.implementation};return{...e,...t}}};_h.className="LSTMCell";ue.registerClass(_h);var dx=class extends as{constructor(e){e.implementation===0&&console.warn("`implementation=0` has been deprecated, and now defaults to `implementation=1`. Please update your layer call."),e.cell=new _h(e),super(e)}call(e,t){return X(()=>{this.cell.dropoutMask!=null&&(te(this.cell.dropoutMask),this.cell.dropoutMask=null),this.cell.recurrentDropoutMask!=null&&(te(this.cell.recurrentDropoutMask),this.cell.recurrentDropoutMask=null);let r=t==null?null:t.mask,n=t==null?null:t.training,a=t==null?null:t.initialState;return super.call(e,{mask:r,training:n,initialState:a})})}static fromConfig(e,t){return t.implmentation===0&&(t.implementation=1),new e(t)}};dx.className="LSTM";ue.registerClass(dx);var ym=class extends Ph{constructor(e){super(e),this.cells=e.cells}get stateSize(){let e=[];for(let t of this.cells.slice().reverse())Array.isArray(t.stateSize)?e.push(...t.stateSize):e.push(t.stateSize);return e}call(e,t){return X(()=>{e=e;let r=e.slice(1),n=[];for(let i of this.cells.slice().reverse())Array.isArray(i.stateSize)?n.push(r.splice(0,i.stateSize.length)):n.push(r.splice(0,1));n.reverse();let a=[],s;for(let i=0;i<this.cells.length;++i){let o=this.cells[i];r=n[i],i===0?s=[e[0]].concat(r):s=[s[0]].concat(r),s=o.call(s,t),a.push(s.slice(1))}r=[];for(let i of a.slice().reverse())r.push(...i);return[s[0]].concat(r)})}build(e){Mg(e)&&(e=e[0]),e=e;let t;this.cells.forEach((r,n)=>{Ro(`RNNCell_${n}`,()=>{r.build(e),Array.isArray(r.stateSize)?t=r.stateSize[0]:t=r.stateSize,e=[e[0],t]})}),this.built=!0}getConfig(){let e=super.getConfig(),t=n=>({className:n.getClassName(),config:n.getConfig()}),r={cells:this.cells.map(t)};return{...e,...r}}static fromConfig(e,t,r={}){let n=[];for(let a of t.cells)n.push(fa(a,r));return new e({cells:n})}get trainableWeights(){if(!this.trainable)return[];let e=[];for(let t of this.cells)e.push(...t.trainableWeights);return e}get nonTrainableWeights(){let e=[];for(let t of this.cells)e.push(...t.nonTrainableWeights);if(!this.trainable){let t=[];for(let r of this.cells)t.push(...r.trainableWeights);return t.concat(e)}return e}getWeights(){let e=[];for(let t of this.cells)e.push(...t.weights);return Fg(e)}setWeights(e){let t=[];for(let r of this.cells){let n=r.weights.length,a=e.splice(n);for(let s=0;s<r.weights.length;++s)t.push([r.weights[s],a[s]])}MA(t)}};ym.className="StackedRNNCells";ue.registerClass(ym);function Hs(e){let{ones:t,rate:r,training:n=!1,count:a=1,dropoutFunc:s}=e,i=()=>s!=null?s(t(),r):a6(t(),r),o=()=>Rh(i,t,n);return!a||a<=1?gr(o().clone()):Array(a).fill(void 0).map(o).map(l=>gr(l.clone()))}var Q6=class extends as{constructor(e){if(e.unroll)throw new Ve("Unrolling is not possible with convolutional RNNs.");if(Array.isArray(e.cell))throw new Ve("It is not possible at the moment to stack convolutional cells.");super(e),this.inputSpec=[new Zt({ndim:5})]}call(e,t){return X(()=>{if(this.cell.dropoutMask!=null&&(te(this.cell.dropoutMask),this.cell.dropoutMask=null),this.cell.recurrentDropoutMask!=null&&(te(this.cell.recurrentDropoutMask),this.cell.recurrentDropoutMask=null),t&&t.constants)throw new H("ConvRNN2D cell does not support constants");let r=t==null?null:t.mask,n=t==null?null:t.training,a=t==null?null:t.initialState;return super.call(e,{mask:r,training:n,initialState:a})})}computeOutputShape(e){let t=this.computeSingleOutputShape(e);return this.returnSequences||(t=[t[0],...t.slice(2)]),this.returnState&&(t=[t,...Array(2).fill([e[0],...t.slice(-3)])]),t}getInitialState(e){return X(()=>{let{stateSize:t}=this.cell,r=e.shape,n=this.computeSingleOutputShape(r),a=[n[0],...n.slice(2)],s=_t(a);return Array.isArray(t)?Array(t.length).fill(s):[s]})}resetStates(e,t=!1){X(()=>{if(!this.stateful)throw new Ga("Cannot call resetStates() on an RNN Layer that is not stateful.");let r=this.inputSpec[0].shape,n=this.computeSingleOutputShape(r),a=[n[0],...n.slice(2)];if(r[0]==null)throw new H("If an RNN is stateful, it needs to know its batch size. Specify the batch size of your input tensors: \n- If using a Sequential model, specify the batch size by passing a `batchInputShape` option to your first layer.\n- If using the functional API, specify the batch size by passing a `batchShape` option to your Input layer.");if(this.getStates()==null)Array.isArray(this.cell.stateSize)?this.states_=this.cell.stateSize.map(()=>_t(a)):this.states_=[_t(a)];else if(e==null)te(this.states_),this.keptStates!=null&&(te(this.keptStates),this.keptStates=[]),Array.isArray(this.cell.stateSize)?this.states_=this.cell.stateSize.map(()=>_t(a)):this.states_[0]=_t(a);else{if(Array.isArray(e)||(e=[e]),e.length!==this.states_.length)throw new H(`Layer ${this.name} expects ${this.states_.length} state(s), but it received ${e.length} state value(s). Input received: ${e}`);t?this.keptStates.push(this.states_.slice()):te(this.states_);for(let s=0;s<this.states_.length;++s){let i=e[s],o=a;if(!v.arraysEqual(i.shape,o))throw new H(`State ${s} is incompatible with layer ${this.name}: expected shape=${o}, received shape=${i.shape}`);this.states_[s]=i}}this.states_=this.states_.map(s=>gr(s.clone()))})}computeSingleOutputShape(e){let{dataFormat:t,filters:r,kernelSize:n,padding:a,strides:s,dilationRate:i}=this.cell,o=t==="channelsFirst",l=e[o?3:2],u=e[o?4:3],d=ma(l,n[0],a,s[0],i[0]),h=ma(u,n[1],a,s[1],i[1]);return[...e.slice(0,2),...o?[r,d,h]:[d,h,r]]}};Q6.className="ConvRNN2D";var Am=class extends _h{constructor(e){let{filters:t,kernelSize:r,strides:n,padding:a,dataFormat:s,dilationRate:i}=e;super({...e,units:t}),this.filters=t,yr(this.filters,"filters"),this.kernelSize=bu(r,2,"kernelSize"),this.kernelSize.forEach(o=>yr(o,"kernelSize")),this.strides=bu(n||1,2,"strides"),this.strides.forEach(o=>yr(o,"strides")),this.padding=a||"valid",Ln(this.padding),this.dataFormat=s||"channelsLast",jt(this.dataFormat),this.dilationRate=bu(i||1,2,"dilationRate"),this.dilationRate.forEach(o=>yr(o,"dilationRate"))}build(e){var t;e=mt(e);let r=this.dataFormat==="channelsFirst"?1:e.length-1;if(e[r]==null)throw new H(`The channel dimension of the input should be defined. Found ${e[r]}`);let n=e[r],a=4,s=this.kernelSize.concat([n,this.filters*a]);this.kernel=this.addWeight("kernel",s,null,this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint);let i=this.kernelSize.concat([this.filters,this.filters*a]);if(this.recurrentKernel=this.addWeight("recurrent_kernel",i,null,this.recurrentInitializer,this.recurrentRegularizer,!0,this.recurrentConstraint),this.useBias){let o;if(this.unitForgetBias){let l=this.biasInitializer,u=this.filters;o=new(t=class extends Zn{apply(d,h){let p=l.apply([u]),c=hn([u]),f=l.apply([u*2]);return IA([p,c,f])}},t.className="CustomInit",t)}else o=this.biasInitializer;this.bias=this.addWeight("bias",[this.filters*a],null,o,this.biasRegularizer,!0,this.biasConstraint)}this.built=!0}call(e,t){return X(()=>{if(e.length!==3)throw new H(`ConvLSTM2DCell expects 3 input Tensors (inputs, h, c), got ${e.length}.`);let r=t.training||!1,n=e[0],a=e[1],s=e[2],i=4;0<this.dropout&&this.dropout<1&&this.dropoutMask==null&&(this.dropoutMask=Hs({ones:()=>Pn(n),rate:this.dropout,training:r,count:i,dropoutFunc:this.dropoutFunc}));let o=this.dropoutMask,l=(V,re,Y)=>!re||!re[Y]?V:L(re[Y],V),u=l(n,o,0),d=l(n,o,1),h=l(n,o,2),p=l(n,o,3);0<this.recurrentDropout&&this.recurrentDropout<1&&this.recurrentDropoutMask==null&&(this.recurrentDropoutMask=Hs({ones:()=>Pn(a),rate:this.recurrentDropout,training:r,count:i,dropoutFunc:this.dropoutFunc}));let c=this.recurrentDropoutMask,f=l(a,c,0),m=l(a,c,1),g=l(a,c,2),y=l(a,c,3),A=3,[x,b,w,S]=Yt(this.kernel.read(),i,A),[N,E,R,z]=this.useBias?Yt(this.bias.read(),i):[null,null,null,null];u=this.inputConv(u,x,N,this.padding),d=this.inputConv(d,b,E,this.padding),h=this.inputConv(h,w,R,this.padding),p=this.inputConv(p,S,z,this.padding);let[M,I,_,O]=Yt(this.recurrentKernel.read(),i,A);f=this.recurrentConv(f,M),m=this.recurrentConv(m,I),g=this.recurrentConv(g,_),y=this.recurrentConv(y,O);let q=this.recurrentActivation.apply(le(u,f)),K=this.recurrentActivation.apply(le(d,m)),D=le(L(K,s),L(q,this.activation.apply(le(h,g)))),Q=L(this.recurrentActivation.apply(le(p,y)),this.activation.apply(D));return[Q,Q,D]})}getConfig(){let{units:e,...t}=super.getConfig(),r={filters:this.filters,kernelSize:this.kernelSize,padding:this.padding,dataFormat:this.dataFormat,dilationRate:this.dilationRate,strides:this.strides};return{...t,...r}}inputConv(e,t,r,n){let a=Ls(e,t,this.strides,n||"valid",this.dataFormat==="channelsFirst"?"NCHW":"NHWC",this.dilationRate);return r?va(a,r,this.dataFormat):a}recurrentConv(e,t){return Ls(e,t,1,"same",this.dataFormat==="channelsFirst"?"NCHW":"NHWC")}};Am.className="ConvLSTM2DCell";ue.registerClass(Am);var px=class extends Q6{constructor(e){let t=new Am(e);super({...e,cell:t})}static fromConfig(e,t){return new e(t)}};px.className="ConvLSTM2D";ue.registerClass(px);var xm=class extends st{constructor(e){super(e),this.rate=Math.max(Math.min(e.rate,1),0),this.noiseShape=e.noiseShape,this.seed=e.seed,this.supportsMasking=!0}getNoiseShape(e){if(this.noiseShape==null)return this.noiseShape;let t=e.shape,r=[];for(let n=0;n<this.noiseShape.length;++n)r.push(this.noiseShape[n]==null?t[n]:this.noiseShape[n]);return r}call(e,t){return X(()=>{this.invokeCallHook(e,t);let r=je(e);if(0<this.rate&&this.rate<1){let n=t.training==null?!1:t.training,a=this.getNoiseShape(r);return Rh(()=>a6(r,this.rate,a,this.seed),()=>r,n)}return e})}getConfig(){let e={rate:this.rate,noiseShape:this.noiseShape,seed:this.seed},t=super.getConfig();return Object.assign(e,t),e}dispose(){return super.dispose()}};xm.className="Dropout";ue.registerClass(xm);var hx=class extends xm{constructor(e){super(e),this.inputSpec=[{ndim:3}]}getNoiseShape(e){let t=e.shape;return[t[0],1,t[2]]}};hx.className="SpatialDropout1D";ue.registerClass(hx);var cx=class extends st{constructor(e){if(super(e),this.activation=null,this.useBias=!0,this.kernel=null,this.bias=null,this.DEFAULT_KERNEL_INITIALIZER="glorotNormal",this.DEFAULT_BIAS_INITIALIZER="zeros",e.batchInputShape==null&&e.inputShape==null&&e.inputDim!=null){let t=null;e.batchSize!=null&&(t=e.batchSize),this.batchInputShape=[t,e.inputDim]}this.units=e.units,yr(this.units,"units"),this.activation=js(e.activation),e.useBias!=null&&(this.useBias=e.useBias),this.kernelInitializer=Rt(e.kernelInitializer||this.DEFAULT_KERNEL_INITIALIZER),this.biasInitializer=Rt(e.biasInitializer||this.DEFAULT_BIAS_INITIALIZER),this.kernelConstraint=lr(e.kernelConstraint),this.biasConstraint=lr(e.biasConstraint),this.kernelRegularizer=$t(e.kernelRegularizer),this.biasRegularizer=$t(e.biasRegularizer),this.activityRegularizer=$t(e.activityRegularizer),this.supportsMasking=!0,this.inputSpec=[{minNDim:2}]}build(e){e=mt(e);let t=e[e.length-1];this.kernel==null&&(this.kernel=this.addWeight("kernel",[t,this.units],null,this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint),this.useBias&&(this.bias=this.addWeight("bias",[this.units],null,this.biasInitializer,this.biasRegularizer,!0,this.biasConstraint))),this.inputSpec=[{minNDim:2,axes:{[-1]:t}}],this.built=!0}computeOutputShape(e){e=mt(e);let t=e.slice();return t[t.length-1]=this.units,t}call(e,t){return X(()=>{this.invokeCallHook(e,t);let r=je(e),n=Yw(this.activation.getClassName()),a;return n!=null?a=$a(r,this.kernel.read(),n,this.bias?this.bias.read():null):(a=$a(r,this.kernel.read()),this.bias!=null&&(a=va(a,this.bias.read())),this.activation!=null&&(a=this.activation.apply(a))),a})}getConfig(){let e={units:this.units,activation:Gs(this.activation),useBias:this.useBias,kernelInitializer:zt(this.kernelInitializer),biasInitializer:zt(this.biasInitializer),kernelRegularizer:xt(this.kernelRegularizer),biasRegularizer:xt(this.biasRegularizer),activityRegularizer:xt(this.activityRegularizer),kernelConstraint:or(this.kernelConstraint),biasConstraint:or(this.biasConstraint)},t=super.getConfig();return Object.assign(e,t),e}};cx.className="Dense";ue.registerClass(cx);var fx=class extends st{constructor(e){e=e||{},super(e),this.inputSpec=[{minNDim:3}],this.dataFormat=e.dataFormat}computeOutputShape(e){e=mt(e);for(let t of e.slice(1))if(t==null)throw new H(`The shape of the input to "Flatten" is not fully defined (got ${e.slice(1)}). Make sure to pass a complete "input_shape" or "batch_input_shape" argument to the first layer in your model.`);return[e[0],$s(e,1)]}call(e,t){return X(()=>{this.invokeCallHook(e,t);let r=je(e);if(this.dataFormat==="channelsFirst"&&r.rank>1){let n=[0];for(let a=2;a<r.rank;++a)n.push(a);n.push(1),r=et(r,n)}return ZL(r)})}getConfig(){let e={};this.dataFormat!=null&&(e.dataFormat=this.dataFormat);let t=super.getConfig();return Object.assign(e,t),e}};fx.className="Flatten";ue.registerClass(fx);var mx=class extends st{constructor(e){super(e),this.supportsMasking=!0,this.activation=js(e.activation)}call(e,t){return X(()=>{this.invokeCallHook(e,t);let r=je(e);return this.activation.apply(r)})}getConfig(){let e={activation:Gs(this.activation)},t=super.getConfig();return Object.assign(e,t),e}};mx.className="Activation";ue.registerClass(mx);var gx=class extends st{constructor(e){super(e),this.n=e.n,this.inputSpec=[{ndim:2}]}computeOutputShape(e){return[e[0],this.n,e[1]]}call(e,t){return X(()=>(e=je(e),XL(e,this.n)))}getConfig(){let e={n:this.n},t=super.getConfig();return Object.assign(e,t),e}};gx.className="RepeatVector";ue.registerClass(gx);var yx=class extends st{constructor(e){super(e),this.targetShape=e.targetShape;for(let t=0;t<this.targetShape.length;++t)this.isUnknown(this.targetShape[t])&&(this.targetShape[t]=null)}isUnknown(e){return e<0||e==null}fixUnknownDimension(e,t){let r="Total size of new array must be unchanged.",n=t.slice(),a=1,s=null;for(let o=0;o<n.length;++o){let l=n[o];if(this.isUnknown(l))if(s===null)s=o;else throw new H("Can only specifiy one unknown dimension.");else a*=l}let i=$s(e);if(s!==null){if(a===0||i%a!==0)throw new H(r);n[s]=i/a}else if(i!==a)throw new H(r);return n}computeOutputShape(e){let t=!1;for(let r=0;r<e.length;++r)if(this.isUnknown(e[r])){t=!0;break}return t?e.slice(0,1).concat(this.targetShape):e.slice(0,1).concat(this.fixUnknownDimension(e.slice(1),this.targetShape))}call(e,t){return X(()=>{this.invokeCallHook(e,t);let r=je(e),n=r.shape,a=n.slice(0,1).concat(this.fixUnknownDimension(n.slice(1),this.targetShape));return U(r,a)})}getConfig(){let e={targetShape:this.targetShape},t=super.getConfig();return Object.assign(e,t),e}};yx.className="Reshape";ue.registerClass(yx);var Ax=class extends st{constructor(e){if(super(e),e.dims==null)throw new Error("Required configuration field `dims` is missing during Permute constructor call.");if(!Array.isArray(e.dims))throw new Error(`Permute constructor requires \`dims\` to be an Array, but received ${e.dims} instead.`);let t=ya(1,e.dims.length+1);if(!v.arraysEqual(e.dims.slice().sort(),t))throw new Error("Invalid permutation `dims`: "+JSON.stringify(e.dims)+" `dims` must contain consecutive integers starting from 1.");this.dims=e.dims,this.dimsIncludingBatch=[0].concat(this.dims),this.inputSpec=[new Zt({ndim:this.dims.length+1})]}computeOutputShape(e){e=mt(e);let t=e.slice();return this.dims.forEach((r,n)=>{t[n+1]=e[r]}),t}call(e,t){return et(je(e),this.dimsIncludingBatch)}getConfig(){let e={dims:this.dims},t=super.getConfig();return Object.assign(e,t),e}};Ax.className="Permute";ue.registerClass(Ax);var xx=class extends st{constructor(e){super(e==null?{}:e),this.supportsMasking=!0,e!=null?this.maskValue=e.maskValue==null?0:e.maskValue:this.maskValue=0}computeOutputShape(e){return e}getConfig(){let e=super.getConfig(),t={maskValue:this.maskValue};return Object.assign(t,e),t}computeMask(e,t){let r=je(e),n=-1;return C0(Tu(r,this.maskValue),n)}call(e,t){return X(()=>{this.invokeCallHook(e,t);let r=je(e),n=-1,a=!0,s=C0(Tu(r,this.maskValue),n,a);return L(r,me(s,r.dtype))})}};xx.className="Masking";ue.registerClass(xx);var bx=class extends st{constructor(e){if(super(e),this.embeddings=null,this.DEFAULT_EMBEDDINGS_INITIALIZER="randomUniform",e.batchInputShape==null&&e.inputShape==null){let t=null;e.batchSize!=null&&(t=e.batchSize),e.inputLength==null?this.batchInputShape=[t,null]:this.batchInputShape=[t].concat(Ct(e.inputLength))}this.inputDim=e.inputDim,yr(this.inputDim,"inputDim"),this.outputDim=e.outputDim,yr(this.outputDim,"outputDim"),this.embeddingsInitializer=Rt(e.embeddingsInitializer||this.DEFAULT_EMBEDDINGS_INITIALIZER),this.embeddingsRegularizer=$t(e.embeddingsRegularizer),this.activityRegularizer=$t(e.activityRegularizer),this.embeddingsConstraint=lr(e.embeddingsConstraint),this.maskZero=e.maskZero,this.supportsMasking=e.maskZero,this.inputLength=e.inputLength}build(e){this.embeddings=this.addWeight("embeddings",[this.inputDim,this.outputDim],this.dtype,this.embeddingsInitializer,this.embeddingsRegularizer,!0,this.embeddingsConstraint),this.built=!0}warnOnIncompatibleInputShape(e){}computeMask(e,t){return X(()=>this.maskZero?(e=je(e),Tu(e,at(e))):null)}computeOutputShape(e){if(e=mt(e),this.inputLength==null)return[...e,this.outputDim];let t=Ct(this.inputLength);if(t.length!==e.length-1)throw new H(`"inputLength" is ${this.inputLength}, but received input shape has shape ${e}`);{let r=0;for(let n=0;n<t.length;++n){let a=t[n],s=e[n+1];if(a!=null&&s!=null&&a!==s)throw new H(`"inputLength" is ${this.inputLength}, but received input shape has shape ${e}`);a==null&&(t[r]=s),r++}}return[e[0],...t,this.outputDim]}call(e,t){return X(()=>{this.invokeCallHook(e,t);let r=je(e);r.dtype!=="int32"&&(r=em(r,"int32"));let n=n6(this.embeddings.read(),U(r,[r.size]));return U(n,mt(this.computeOutputShape(r.shape)))})}getConfig(){let e={inputDim:this.inputDim,outputDim:this.outputDim,embeddingsInitializer:zt(this.embeddingsInitializer),embeddingsRegularizer:xt(this.embeddingsRegularizer),activityRegularizer:xt(this.activityRegularizer),embeddingsConstraint:or(this.embeddingsConstraint),maskZero:this.maskZero,inputLength:this.inputLength},t=super.getConfig();return Object.assign(e,t),e}};bx.className="Embedding";ue.registerClass(bx);var Ol=class extends st{constructor(e){super(e||{}),this.supportsMasking=!0}mergeFunction(e){throw new Ve}computeElementwiseOpOutputShape(e,t){if(e==null||t==null)return null;if(e.length<t.length)return this.computeElementwiseOpOutputShape(t,e);if(t.length===0)return e;let r=e.slice(0,e.length-t.length);for(let n=0;n<t.length;++n){let a=e[e.length-t.length+n],s=t[n];if(a==null||s==null||a<0||s<0)r.push(null);else if(a===1)r.push(s);else if(s===1)r.push(a);else{if(a!==s)throw new H("Operands could not be broadcast together with shapes "+JSON.stringify(e)+" "+JSON.stringify(t));r.push(a)}}return r}build(e){if(Array.isArray(e)&&!Array.isArray(e[0])&&(e=[mt(e)]),e=e,e.length<2)throw new H(`A merge layer should be called on an Array of at least 2 inputs. Got ${e.length} input(s).`);let t=[];for(let a of e)a!=null&&a[0]!==null&&t.push(a[0]);if(t=Rs(t),t.length>1)throw new H(`Can not merge tensors with different batch sizes. Got tensors with shapes: ${JSON.stringify(e)}.`);let r=e[0]==null?null:e[0].slice(1);for(let a=1;a<e.length;++a){let s=e[a]==null?null:e[a].slice(1);r=this.computeElementwiseOpOutputShape(r,s)}let n=e.map(a=>a.length);e.indexOf(null)===-1&&Rs(n).length===1?this.reshapeRequired=!1:this.reshapeRequired=!0}call(e,t){return X(()=>{if(e=e,this.reshapeRequired){let r=[],n=e.map(a=>a.rank);if(n.indexOf(null)===-1){let a=Us(n);for(let s of e){let i=s.rank;for(let o=0;o<a-i;++o)s=Nh(s,1);r.push(s)}return this.mergeFunction(r)}else{let a=!1;for(let o of e){let l=o.rank;if(l==null){let u=o.shape,d=u[0],h=u.slice(1).concat([d]),p=U(o,[d].concat($s(u.slice(1))));p=et(p,[1,0]),p=U(p,h),r.push(p),a=!0}else if(l>1){let u=ya(1,l).concat([0]);r.push(et(o,u)),a=!0}else r.push(o)}let s=this.mergeFunction(r),i=s.rank;if(a){if(i==null){let o=s.shape,l=o.length,u=o[l-1],d=[u].concat(o.slice(0,o.length-1));s=U(et(U(s,[-1,u]),[1,0]),d)}else if(i>1){let o=[i-1].concat(ya(0,i-1));s=et(s,o)}}return s}}else return this.mergeFunction(e)})}computeOutputShape(e){e=e;let t;e[0]==null?t=null:t=e[0].slice(1);for(let n=1;n<e.length;++n){let a=e[n]==null?null:e[n].slice(1);t=this.computeElementwiseOpOutputShape(t,a)}let r=[];for(let n of e)n!=null&&n[0]!==null&&r.push(n[0]);return r=Rs(r),r.length===1?t=r.concat(t):t=[null].concat(t),t}computeMask(e,t){return X(()=>{if(t==null)return null;if(!Array.isArray(t))throw new H("`mask` should be an Array");if(!Array.isArray(e))throw new H("`inputs` should be an Array");if(t.length!==e.length)throw new H(`The Array 'inputs' and 'mask' are expected to have the same length, but have different lengths (${e.length} vs ${t.length})`);if(t.every(n=>n==null))return null;t=t.map(n=>n==null?n:Kt(n,0));let r=t[0];for(let n=1;n<t.length-1;++n)r=ga(r,t[n]);return r})}},vx=class extends Ol{constructor(e){super(e)}mergeFunction(e){return X(()=>{let t=e[0].clone();for(let r=1;r<e.length;++r)t=le(t,e[r]);return t})}};vx.className="Add";ue.registerClass(vx);var wx=class extends Ol{constructor(e){super(e)}mergeFunction(e){return X(()=>{let t=e[0].clone();for(let r=1;r<e.length;++r)t=L(t,e[r]);return t})}};wx.className="Multiply";ue.registerClass(wx);var kx=class extends Ol{constructor(e){super(e)}mergeFunction(e){return X(()=>{let t=e[0].clone();for(let r=1;r<e.length;++r)t=le(t,e[r]);return L(1/e.length,t)})}};kx.className="Average";ue.registerClass(kx);var Ix=class extends Ol{constructor(e){super(e)}mergeFunction(e){return X(()=>{let t=e[0];for(let r=1;r<e.length;++r)t=ts(t,e[r]);return t})}};Ix.className="Maximum";ue.registerClass(Ix);var Sx=class extends Ol{constructor(e){super(e)}mergeFunction(e){return X(()=>{let t=e[0];for(let r=1;r<e.length;++r)t=Ih(t,e[r]);return t})}};Sx.className="Minimum";ue.registerClass(Sx);var Cx=class extends Ol{constructor(e){super(e),this.DEFAULT_AXIS=-1,e==null&&(e={}),this.axis=e.axis==null?this.DEFAULT_AXIS:e.axis,this.supportsMasking=!0,this.reshapeRequired=!1}build(e){if(!(Array.isArray(e)&&Array.isArray(e[0]))||e.length===1)throw new H("A `Concatenate` layer should be called on a list of at least 2 inputs");e=e;let t=!0;for(let n of e)if(n!=null){t=!1;break}if(t)return;let r=[];for(let n=0;n<e.length;++n){let a=e[n].slice();a.splice(this.axis,1);let s=!1;for(let i of r)if(v.arraysEqual(i,a)){s=!0;break}s||r.push(a)}if(r.length>1)throw new H("A `Concatenate` layer requires inputs with matching shapes except for the concat axis. Got input shapes: "+JSON.stringify(e))}mergeFunction(e){return X(()=>IA(e,this.axis))}computeOutputShape(e){if(!(Array.isArray(e)&&Array.isArray(e[0])))throw new H("A `Concatenate` layer should be called on a list of inputs.");let t=e,r=t[0].slice(),n=this.axis<0?r.length+this.axis:this.axis;for(let a of t.slice(1)){if(r[n]==null||a[n]==null){r[n]=null;break}r[n]+=a[n]}return r}computeMask(e,t){if(t==null)return null;if(!Array.isArray(t))throw new H("`mask` should be an array for Concatenate");if(!Array.isArray(e))throw new H("`inputs` should be an array for Concatenate");if(t.length!==e.length)throw new H(`Mismatch in the length of mask (${t.length}) and the legnth of inputs (${e.length})`);return X(()=>{let r=!0;if(t.forEach(s=>{if(s!=null){r=!1;return}}),r)return null;let n=[];for(let s=0;s<e.length;++s)t[s]==null?n.push(me(Pn(e[s]),"bool")):t[s].rank<e[s].rank?n.push(Kt(t[s],-1)):n.push(t[s]);let a=St(n,this.axis);return Dy(a,-1,!1)})}getConfig(){let e={axis:this.axis},t=super.getConfig();return Object.assign(e,t),e}};Cx.className="Concatenate";ue.registerClass(Cx);function Ap(e,t){for(;e<0;)e+=t;return e}function BU(e,t,r){if(e.shape.length>3||t.shape.length>3)throw new Ve("batchDot is not implemented for tensors of 4D or higher rank yet");if(v.assert(e.shape.length>=2,()=>`batchDot requires the rank of x to be >= 2, but got ${e.shape.length}`),v.assert(e.shape.length>=2,()=>`batchDot requires the rank of y to be >= 2, but got ${t.shape.length}`),typeof r=="number"&&(r=[r,r]),e.dtype==="complex64"||t.dtype==="complex64")throw new Ve("batchDot is not implemented for complex64-type Tensors yet.");let n=e.shape.length,a=t.shape.length;r==null&&(r=[n-1,a-2]);let s=r;return X(()=>{let i;if(n>a){i=n-a;let l=[];for(let u=0;u<i;++u)l.push(1);t=U(t,t.shape.concat(l))}else if(a>n){i=a-n;let l=[];for(let u=0;u<i;++u)l.push(1);e=U(e,e.shape.concat(l))}else i=0;let o;if(e.shape.length===2&&t.shape.length===2)s[0]===s[1]?o=ke(L(e,t),s[0]):o=ke(L(et(e,[1,0]),t),s[1]);else{let l=s[0]!==e.shape.length-1,u=s[1]===t.shape.length-1;o=Ze(e,t,l,u)}if(i>0){let l;n>a?l=n+a-3:l=n-1;let u=[];for(let d=l;d<l+i;++d)u.push(d);o=tt(o,u)}return o.shape.length===1&&(o=Kt(o,1)),o})}var Tx=class extends Ol{constructor(e){super(e),this.axes=e.axes,this.normalize=e.normalize==null?!1:e.normalize,this.supportsMasking=!0,this.reshapeRequired=!1}build(e){v.assert(Array.isArray(e)&&e.length===2&&Array.isArray(e[0])&&Array.isArray(e[1]),()=>"A `Dot` layer should be called on a list of exactly 2 inputs.");let t=e[0],r=e[1];if(t.length>3||r.length>3)throw new Ve("Dot layer does not support tensors of 4D or higher rank yet.");let n=this.interpretAxes(t,r);if(t[n[0]]!==r[n[1]])throw new H(`Dimension incompatibility: ${t[n[0]]} !== ${r[n[1]]}`)}mergeFunction(e){if(e.length!==2)throw new H(`A \`Dot\` layer must be called on exactly 2 inputs, but received ${e.length} input(s).`);let t=e[0],r=e[1],n;return Array.isArray(this.axes)?n=this.axes.map((a,s)=>Ap(a,e[s].shape.length)):n=[Ap(this.axes,t.shape.length),Ap(this.axes,r.shape.length)],this.normalize&&(t=F0(t,n[0]),r=F0(r,n[1])),BU(t,r,n)}interpretAxes(e,t){let r;return Array.isArray(this.axes)?r=this.axes:r=[Ap(this.axes,e.length),Ap(this.axes,t.length)],r}computeOutputShape(e){v.assert(Array.isArray(e)&&e.length===2&&Array.isArray(e[0])&&Array.isArray(e[1]),()=>"A `Dot` layer should be called on a list of exactly 2 inputs.");let t=e[0].slice(),r=e[1].slice();if(t.length>3||r.length>3)throw new Ve("Dot layer does not support tensors of 4D or higher rank yet.");let n=this.interpretAxes(t,r);t.splice(n[0],1),r.splice(n[1],1),r.splice(0,1);let a=t.concat(r);return a.length===1&&a.push(1),a}computeMask(e,t){return null}getConfig(){let e={axes:this.axes,normalize:this.normalize},t=super.getConfig();return Object.assign(e,t),e}};Tx.className="Dot";ue.registerClass(Tx);var Nx=class extends st{constructor(e){super(e),this.supportsMasking=!0,this.stddev=e.stddev}computeOutputShape(e){return e}getConfig(){let e=super.getConfig(),t={stddev:this.stddev};return Object.assign(t,e),t}call(e,t){return X(()=>{this.invokeCallHook(e,t);let r=je(e);return Rh(()=>le(tm(r.shape,0,this.stddev),r),()=>r,t.training||!1)})}};Nx.className="GaussianNoise";ue.registerClass(Nx);var Ex=class extends st{constructor(e){super(e),this.supportsMasking=!0,this.rate=e.rate}computeOutputShape(e){return e}getConfig(){let e=super.getConfig(),t={rate:this.rate};return Object.assign(t,e),t}call(e,t){return X(()=>{this.invokeCallHook(e,t);let r=je(e);return this.rate>0&&this.rate<1?Rh(()=>{let n=Math.sqrt(this.rate/(1-this.rate));return L(r,tm(r.shape,1,n))},()=>r,t.training||!1):r})}};Ex.className="GaussianDropout";ue.registerClass(Ex);var Rx=class extends st{constructor(e){super(e),this.supportsMasking=!0,this.rate=e.rate,this.noiseShape=e.noiseShape}_getNoiseShape(e){return this.noiseShape||je(e).shape}computeOutputShape(e){return e}getConfig(){let e=super.getConfig(),t={rate:this.rate};return Object.assign(t,e),t}call(e,t){return X(()=>{if(this.rate<1&&this.rate>0){let r=this._getNoiseShape(e);return Rh(()=>{let n=je(e),a=1.6732632423543772,s=1.0507009873554805,i=-a*s,o=Ml(gd(r),this.rate);o=em(o,"float32");let l=((1-this.rate)*(1+this.rate*i**2))**-.5,u=-l*i*this.rate,d=le(L(n,o),L(le(o,-1),i));return le(L(d,l),u)},()=>je(e),t.training||!1)}return e})}};Rx.className="AlphaDropout";ue.registerClass(Rx);function qp(e,t,r,n,a,s=.001){let i;if(e.rank===2)i=G7(e,t,r,n,a,s);else if(e.rank===3)i=j7(e,t,r,n,a,s);else if(e.rank===4)i=H7(e,t,r,n,a,s);else throw new Ve(`batchNormalization is not implemented for array of rank ${e.rank} yet`);return i}function WU(e,t,r,n,a=.001){return X(()=>{let s=zf(e,n),i=s.mean,o=s.variance;return[qp(e,i,o,r,t,a),i,o]})}function VU(e,t,r,n,a=.001){return X(()=>{let s=zf(e,n),i=s.mean,o=s.variance,l=[];for(let c of ya(0,e.rank))n.indexOf(c)!==-1?l.push(1):l.push(e.shape[c]);let u=U(i,l),d=U(o,l),h=t==null?null:U(t,l),p=r==null?null:U(r,l);return[qp(e,u,d,p,h,a),i,o]})}function UU(e,t,r,n,a=.001){return v.arraysEqual(n.slice().sort(),ya(0,e.rank-1))?WU(e,t,r,n,a):VU(e,t,r,n,a)}var $x=class extends st{constructor(e){e==null&&(e={}),super(e),this.supportsMasking=!0,this.axis=e.axis==null?-1:e.axis,this.momentum=e.momentum==null?.99:e.momentum,this.epsilon=e.epsilon==null?.001:e.epsilon,this.center=e.center==null?!0:e.center,this.scale=e.scale==null?!0:e.scale,this.betaInitializer=Rt(e.betaInitializer||"zeros"),this.gammaInitializer=Rt(e.gammaInitializer||"ones"),this.movingMeanInitializer=Rt(e.movingMeanInitializer||"zeros"),this.movingVarianceInitializer=Rt(e.movingVarianceInitializer||"ones"),this.betaConstraint=lr(e.betaConstraint),this.gammaConstraint=lr(e.gammaConstraint),this.betaRegularizer=$t(e.betaRegularizer),this.gammaRegularizer=$t(e.gammaRegularizer)}build(e){e=mt(e);let t=this.axis>=0?this.axis:this.axis+e.length,r=e[t];if(r==null)throw new H(`Axis ${t} of input tensor should have a defined dimension but the layer received an input with shape ${JSON.stringify(e)}.`);this.inputSpec=[new Zt({ndim:e.length,axes:{[t]:r}})];let n=[r];this.scale&&(this.gamma=this.addWeight("gamma",n,null,this.gammaInitializer,this.gammaRegularizer,!0,this.gammaConstraint)),this.center&&(this.beta=this.addWeight("beta",n,null,this.betaInitializer,this.betaRegularizer,!0,this.betaConstraint)),this.movingMean=this.addWeight("moving_mean",n,null,this.movingMeanInitializer,null,!1),this.movingVariance=this.addWeight("moving_variance",n,null,this.movingVarianceInitializer,null,!1),this.built=!0}call(e,t){return X(()=>{let r=t.training==null?!1:t.training,n=je(e),a=n.shape,s=a.length,i=ya(0,s),o=this.axis>=0?this.axis:this.axis+s;i.splice(o,1);let l=Do(1,s);l[o]=a[o];let u=i.slice();u.sort();let d=!v.arraysEqual(u,ya(0,s).slice(0,s-1)),h=()=>{if(d){let g=U(this.movingMean.read(),l),y=U(this.movingVariance.read(),l),A=this.center?U(this.beta.read(),l):null,x=this.scale?U(this.gamma.read(),l):null;return qp(n,g,y,A,x,this.epsilon)}else return qp(n,this.movingMean.read(),this.movingVariance.read(),this.beta==null?null:this.beta.read(),this.gamma==null?null:this.gamma.read(),this.epsilon)};if(!r)return h();let[p,c,f]=UU(n,this.gamma.read(),this.beta.read(),i,this.epsilon),m=(g,y,A)=>{X(()=>{let x=1-A,b=g.read(),w=L(ce(b,y),x);g.write(ce(b,w))})};return m(this.movingMean,c,this.momentum),m(this.movingVariance,f,this.momentum),p})}getConfig(){let e={axis:this.axis,momentum:this.momentum,epsilon:this.epsilon,center:this.center,scale:this.scale,betaInitializer:zt(this.betaInitializer),gammaInitializer:zt(this.gammaInitializer),movingMeanInitializer:zt(this.movingMeanInitializer),movingVarianceInitializer:zt(this.movingVarianceInitializer),betaRegularizer:xt(this.betaRegularizer),gammaRegularizer:xt(this.gammaRegularizer),betaConstraint:or(this.betaConstraint),gammaConstraint:or(this.gammaConstraint)},t=super.getConfig();return Object.assign(e,t),e}};$x.className="BatchNormalization";ue.registerClass($x);var Mx=class extends st{constructor(e){if(e==null&&(e={}),super(e),this.axis=e.axis==null?-1:e.axis,typeof this.axis=="number"){if(!Number.isInteger(this.axis))throw new Error(`Expected axis to be an integer, but received ${this.axis}`)}else if(Array.isArray(this.axis)){for(let t of this.axis)if(!Number.isInteger(t))throw new Error(`Expected axis to be an array of integers, but received ${JSON.stringify(this.axis)}`)}else throw new Error(`Expected axis to be an integer or an array of integers, but received ${JSON.stringify(this.axis)}`);this.epsilon=e.epsilon==null?.001:e.epsilon,this.center=e.center==null?!0:e.center,this.scale=e.scale==null?!0:e.scale,this.betaInitializer=Rt(e.betaInitializer||"zeros"),this.gammaInitializer=Rt(e.gammaInitializer||"ones"),this.betaRegularizer=$t(e.betaRegularizer),this.gammaRegularizer=$t(e.gammaRegularizer),this.supportsMasking=!0}build(e){e=mt(e);let t=e.length;typeof this.axis=="number"&&(this.axis=[this.axis]);for(let a=0;a<this.axis.length;++a)this.axis[a]<0&&(this.axis[a]+=t);for(let a of this.axis)if(a<0||a>=t)throw new Error(`Invalid axis: ${a}`);if(this.axis.length!==Rs(this.axis).length)throw new Error(`Found duplicate axes in: ${this.axis}`);let r=this.axis.map(a=>e[a]),n=!0;this.scale?this.gamma=this.addWeight("gamma",r,"float32",this.gammaInitializer,this.gammaRegularizer,n):this.gamma=null,this.center?this.beta=this.addWeight("beta",r,"float32",this.betaInitializer,this.betaRegularizer,n):this.beta=null,this.built=!0}call(e,t){let r=je(e),n=r.shape,a=n.length;return X(()=>{let{mean:s,variance:i}=zf(r,this.axis,!0),o=Do(1,a);for(let c of this.axis)o[c]=n[c];let l=c=>c!=null&&c.shape.length!==a?U(c,o):c,u=l(this.gamma.read()),d=l(this.beta.read()),h=[],p=[];for(let c=0;c<a;++c)this.axis.indexOf(c)!==-1?(h.push(n[c]),p.push(1)):(h.push(1),p.push(n[c]));return s=Gn(s,h),i=Gn(i,h),u=Gn(u,p),d=Gn(d,p),qp(r,s,i,d,u,this.epsilon)})}getConfig(){let e={axis:this.axis,epsilon:this.epsilon,center:this.center,scale:this.scale,betaInitializer:zt(this.betaInitializer),gammaInitializer:zt(this.gammaInitializer),betaRegularizer:xt(this.betaRegularizer),gammaRegularizer:xt(this.gammaRegularizer)},t=super.getConfig();return Object.assign(e,t),e}};Mx.className="LayerNormalization";ue.registerClass(Mx);function GU(e,t,r){return X(()=>{if(e.rank!==4)throw new H(`temporalPadding expects input tensor to be 4-D, but received a ${e.rank}-D tensor.`);if(t==null&&(t=[[1,1],[1,1]]),t.length!==2||t[0].length!==2||t[1].length!==2)throw new H("spatial2dPadding expects `padding` to be an Array of two Arrays, each of which is an Array of two integers.");if(r==null&&(r=Aa()),r!=="channelsLast"&&r!=="channelsFirst")throw new H(`Unknown data format: ${r}. Supported data formats are 'channelsLast' and 'channelsFirst.`);let n;return r==="channelsFirst"?n=[[0,0],[0,0],t[0],t[1]]:n=[[0,0],t[0],t[1],[0,0]],Xn(e,n)})}var Fx=class extends st{constructor(e){if(e==null&&(e={}),super(e),this.dataFormat=e.dataFormat==null?Aa():e.dataFormat,e.padding==null)this.padding=[[1,1],[1,1]];else if(typeof e.padding=="number")this.padding=[[e.padding,e.padding],[e.padding,e.padding]];else{if(e.padding=e.padding,e.padding.length!==2)throw new H(`ZeroPadding2D expects padding to be a length-2 array, but received a length-${e.padding.length} array.`);let t,r;if(typeof e.padding[0]=="number")t=[e.padding[0],e.padding[0]],r=[e.padding[1],e.padding[1]];else{if(e.padding=e.padding,e.padding[0].length!==2)throw new H(`ZeroPadding2D expects height padding to be a length-2 array, but received a length-${e.padding[0].length} array.`);if(t=e.padding[0],e.padding[1].length!==2)throw new H(`ZeroPadding2D expects width padding to be a length-2 array, but received a length-${e.padding[1].length} array.`);r=e.padding[1]}this.padding=[t,r]}this.inputSpec=[new Zt({ndim:4})]}computeOutputShape(e){e=mt(e);let t,r;return this.dataFormat==="channelsFirst"?(e[2]!=null&&e[2]>=0?t=e[2]+this.padding[0][0]+this.padding[0][1]:t=null,e[3]!=null&&e[3]>=0?r=e[3]+this.padding[1][0]+this.padding[1][1]:r=null,[e[0],e[1],t,r]):(e[1]!=null&&e[1]>=0?t=e[1]+this.padding[0][0]+this.padding[0][1]:t=null,e[2]!=null&&e[2]>=0?r=e[2]+this.padding[1][0]+this.padding[1][1]:r=null,[e[0],t,r,e[3]])}call(e,t){return X(()=>GU(je(e),this.padding,this.dataFormat))}getConfig(){let e={padding:this.padding,dataFormat:this.dataFormat},t=super.getConfig();return Object.assign(e,t),e}};Fx.className="ZeroPadding2D";ue.registerClass(Fx);function bm(e,t,r,n,a,s){return X(()=>{jt(a),Qw(s),Ln(n),r==null&&(r=[1,1]),n==null&&(n="valid"),a==null&&(a=Aa()),s==null&&(s="max"),e=QA(e,a);let i,o=n==="same"?"same":"valid";return s==="max"?i=_f(e,t,r,o):i=Tf(e,t,r,o),a==="channelsFirst"&&(i=et(i,[0,3,1,2])),i})}function ek(e,t,r,n,a,s){return X(()=>{jt(a),Qw(s),Ln(n),r==null&&(r=[1,1,1]),n==null&&(n="valid"),a==null&&(a=Aa()),s==null&&(s="max"),e=j6(e,a);let i,o=n==="same"?"same":"valid";return s==="max"?i=tA(e,t,r,o):i=By(e,t,r,o),a==="channelsFirst"&&(i=et(i,[0,4,1,2,3])),i})}var tk=class extends st{constructor(e){if(e.poolSize==null&&(e.poolSize=2),super(e),typeof e.poolSize=="number")this.poolSize=[e.poolSize];else if(Array.isArray(e.poolSize)&&e.poolSize.length===1&&typeof e.poolSize[0]=="number")this.poolSize=e.poolSize;else throw new H(`poolSize for 1D convolutional layer must be a number or an Array of a single number, but received ${JSON.stringify(e.poolSize)}`);if(yr(this.poolSize,"poolSize"),e.strides==null)this.strides=this.poolSize;else if(typeof e.strides=="number")this.strides=[e.strides];else if(Array.isArray(e.strides)&&e.strides.length===1&&typeof e.strides[0]=="number")this.strides=e.strides;else throw new H(`strides for 1D convolutional layer must be a number or an Array of a single number, but received ${JSON.stringify(e.strides)}`);yr(this.strides,"strides"),this.padding=e.padding==null?"valid":e.padding,Ln(this.padding),this.inputSpec=[new Zt({ndim:3})]}computeOutputShape(e){e=mt(e);let t=ma(e[1],this.poolSize[0],this.padding,this.strides[0]);return[e[0],t,e[2]]}call(e,t){return X(()=>{this.invokeCallHook(e,t),e=Nh(je(e),2);let r=this.poolingFunction(je(e),[this.poolSize[0],1],[this.strides[0],1],this.padding,"channelsLast");return tt(r,[2])})}getConfig(){let e={poolSize:this.poolSize,padding:this.padding,strides:this.strides},t=super.getConfig();return Object.assign(e,t),e}},Px=class extends tk{constructor(e){super(e)}poolingFunction(e,t,r,n,a){return jt(a),Ln(n),bm(e,t,r,n,a,"max")}};Px.className="MaxPooling1D";ue.registerClass(Px);var _x=class extends tk{constructor(e){super(e)}poolingFunction(e,t,r,n,a){return jt(a),Ln(n),bm(e,t,r,n,a,"avg")}};_x.className="AveragePooling1D";ue.registerClass(_x);var rk=class extends st{constructor(e){if(e.poolSize==null&&(e.poolSize=[2,2]),super(e),this.poolSize=Array.isArray(e.poolSize)?e.poolSize:[e.poolSize,e.poolSize],e.strides==null)this.strides=this.poolSize;else if(Array.isArray(e.strides)){if(e.strides.length!==2)throw new H(`If the strides property of a 2D pooling layer is an Array, it is expected to have a length of 2, but received length ${e.strides.length}.`);this.strides=e.strides}else this.strides=[e.strides,e.strides];yr(this.poolSize,"poolSize"),yr(this.strides,"strides"),this.padding=e.padding==null?"valid":e.padding,this.dataFormat=e.dataFormat==null?"channelsLast":e.dataFormat,jt(this.dataFormat),Ln(this.padding),this.inputSpec=[new Zt({ndim:4})]}computeOutputShape(e){e=mt(e);let t=this.dataFormat==="channelsFirst"?e[2]:e[1],r=this.dataFormat==="channelsFirst"?e[3]:e[2];return t=ma(t,this.poolSize[0],this.padding,this.strides[0]),r=ma(r,this.poolSize[1],this.padding,this.strides[1]),this.dataFormat==="channelsFirst"?[e[0],e[1],t,r]:[e[0],t,r,e[3]]}call(e,t){return X(()=>(this.invokeCallHook(e,t),this.poolingFunction(je(e),this.poolSize,this.strides,this.padding,this.dataFormat)))}getConfig(){let e={poolSize:this.poolSize,padding:this.padding,strides:this.strides,dataFormat:this.dataFormat},t=super.getConfig();return Object.assign(e,t),e}},zx=class extends rk{constructor(e){super(e)}poolingFunction(e,t,r,n,a){return jt(a),Ln(n),bm(e,t,r,n,a,"max")}};zx.className="MaxPooling2D";ue.registerClass(zx);var Ox=class extends rk{constructor(e){super(e)}poolingFunction(e,t,r,n,a){return jt(a),Ln(n),bm(e,t,r,n,a,"avg")}};Ox.className="AveragePooling2D";ue.registerClass(Ox);var nk=class extends st{constructor(e){if(e.poolSize==null&&(e.poolSize=[2,2,2]),super(e),this.poolSize=Array.isArray(e.poolSize)?e.poolSize:[e.poolSize,e.poolSize,e.poolSize],e.strides==null)this.strides=this.poolSize;else if(Array.isArray(e.strides)){if(e.strides.length!==3)throw new H(`If the strides property of a 3D pooling layer is an Array, it is expected to have a length of 3, but received length ${e.strides.length}.`);this.strides=e.strides}else this.strides=[e.strides,e.strides,e.strides];yr(this.poolSize,"poolSize"),yr(this.strides,"strides"),this.padding=e.padding==null?"valid":e.padding,this.dataFormat=e.dataFormat==null?"channelsLast":e.dataFormat,jt(this.dataFormat),Ln(this.padding),this.inputSpec=[new Zt({ndim:5})]}computeOutputShape(e){e=mt(e);let t=this.dataFormat==="channelsFirst"?e[2]:e[1],r=this.dataFormat==="channelsFirst"?e[3]:e[2],n=this.dataFormat==="channelsFirst"?e[4]:e[3];return t=ma(t,this.poolSize[0],this.padding,this.strides[0]),r=ma(r,this.poolSize[1],this.padding,this.strides[1]),n=ma(n,this.poolSize[2],this.padding,this.strides[2]),this.dataFormat==="channelsFirst"?[e[0],e[1],t,r,n]:[e[0],t,r,n,e[4]]}call(e,t){return X(()=>(this.invokeCallHook(e,t),this.poolingFunction(je(e),this.poolSize,this.strides,this.padding,this.dataFormat)))}getConfig(){let e={poolSize:this.poolSize,padding:this.padding,strides:this.strides,dataFormat:this.dataFormat},t=super.getConfig();return Object.assign(e,t),e}},Dx=class extends nk{constructor(e){super(e)}poolingFunction(e,t,r,n,a){return jt(a),Ln(n),ek(e,t,r,n,a,"max")}};Dx.className="MaxPooling3D";ue.registerClass(Dx);var Lx=class extends nk{constructor(e){super(e)}poolingFunction(e,t,r,n,a){return jt(a),Ln(n),ek(e,t,r,n,a,"avg")}};Lx.className="AveragePooling3D";ue.registerClass(Lx);var ak=class extends st{constructor(e){super(e),this.inputSpec=[new Zt({ndim:3})]}computeOutputShape(e){return[e[0],e[2]]}call(e,t){throw new Ve}},Bx=class extends ak{constructor(e){super(e||{})}call(e,t){return X(()=>{let r=je(e);return Vt(r,1)})}};Bx.className="GlobalAveragePooling1D";ue.registerClass(Bx);var Wx=class extends ak{constructor(e){super(e||{})}call(e,t){return X(()=>{let r=je(e);return Ar(r,1)})}};Wx.className="GlobalMaxPooling1D";ue.registerClass(Wx);var sk=class extends st{constructor(e){super(e),this.dataFormat=e.dataFormat==null?"channelsLast":e.dataFormat,jt(this.dataFormat),this.inputSpec=[new Zt({ndim:4})]}computeOutputShape(e){return e=e,this.dataFormat==="channelsLast"?[e[0],e[3]]:[e[0],e[1]]}call(e,t){throw new Ve}getConfig(){let e={dataFormat:this.dataFormat},t=super.getConfig();return Object.assign(e,t),e}},Vx=class extends sk{call(e,t){return X(()=>{let r=je(e);return this.dataFormat==="channelsLast"?Vt(r,[1,2]):Vt(r,[2,3])})}};Vx.className="GlobalAveragePooling2D";ue.registerClass(Vx);var Ux=class extends sk{call(e,t){return X(()=>{let r=je(e);return this.dataFormat==="channelsLast"?Ar(r,[1,2]):Ar(r,[2,3])})}};Ux.className="GlobalMaxPooling2D";ue.registerClass(Ux);var ik=class extends st{constructor(e){super(e),this.layer=e.layer}build(e){this.built=!0}get trainable(){return this.layer!=null?this.layer.trainable:!1}set trainable(e){this.layer!=null&&(this.layer.trainable=e)}get trainableWeights(){return this.layer.trainableWeights}get nonTrainableWeights(){return this.layer.nonTrainableWeights}get updates(){return this.layer._updates}get losses(){return this.layer.losses}getWeights(){return this.layer.getWeights()}setWeights(e){this.layer.setWeights(e)}getConfig(){let e={layer:{className:this.layer.getClassName(),config:this.layer.getConfig()}},t=super.getConfig();return Object.assign(e,t),e}setFastWeightInitDuringBuild(e){super.setFastWeightInitDuringBuild(e),this.layer!=null&&this.layer.setFastWeightInitDuringBuild(e)}static fromConfig(e,t,r={}){let n=t.layer,a=fa(n,r);delete t.layer;let s={layer:a};return Object.assign(s,t),new e(s)}},Gx=class extends ik{constructor(e){super(e),this.supportsMasking=!0}build(e){if(e=mt(e),e.length<3)throw new H(`TimeDistributed layer expects an input shape >= 3D, but received input shape ${JSON.stringify(e)}`);this.inputSpec=[{shape:e}];let t=[e[0]].concat(e.slice(2));this.layer.built||(this.layer.build(t),this.layer.built=!0),super.build(e)}computeOutputShape(e){e=mt(e);let t=[e[0]].concat(e.slice(2)),r=this.layer.computeOutputShape(t),n=e[1];return[r[0],n].concat(r.slice(1))}call(e,t){return X(()=>(e=je(e),Y6((r,n)=>[je(this.layer.call(r,t)),[]],e,[],!1,null,null,!1,!0)[1]))}};Gx.className="TimeDistributed";ue.registerClass(Gx);function jU(e){_l(UL,"BidirectionalMergeMode",e)}var HU="concat",jx=class extends ik{constructor(e){super(e);let t=e.layer.getConfig(),r={};r.className=e.layer.getClassName(),r.config=t,this.forwardLayer=fa(r),t.goBackwards=t.goBackwards!==!0;let n={};if(n.className=e.layer.getClassName(),n.config=t,this.backwardLayer=fa(n),this.forwardLayer.name="forward_"+this.forwardLayer.name,this.backwardLayer.name="backward_"+this.backwardLayer.name,this.mergeMode=e.mergeMode===void 0?HU:e.mergeMode,jU(this.mergeMode),e.weights)throw new Ve("weights support is not implemented for Bidirectional layer yet.");this._stateful=e.layer.stateful,this.returnSequences=e.layer.returnSequences,this.returnState=e.layer.returnState,this.supportsMasking=!0,this._trainable=!0,this.inputSpec=e.layer.inputSpec,this.numConstants=null}get trainable(){return this._trainable}set trainable(e){this._trainable=e,this.forwardLayer!=null&&(this.forwardLayer.trainable=e),this.backwardLayer!=null&&(this.backwardLayer.trainable=e)}getWeights(){return this.forwardLayer.getWeights().concat(this.backwardLayer.getWeights())}setWeights(e){let t=e.length,r=Math.floor(t/2);this.forwardLayer.setWeights(e.slice(0,r)),this.backwardLayer.setWeights(e.slice(r))}computeOutputShape(e){let t=this.forwardLayer.computeOutputShape(e);Array.isArray(t)&&Array.isArray(t[0])||(t=[t]),t=t;let r,n,a;return this.returnState&&(a=t.slice(1)),r=t[0],r=r,this.mergeMode==="concat"?(r[r.length-1]*=2,n=[r]):this.mergeMode==null?n=[r,r.slice()]:n=[r],this.returnState?this.mergeMode==null?n.concat(a).concat(a.slice()):[r].concat(a).concat(a.slice()):tn(n)}apply(e,t){let r=t==null?null:t.initialState,n=t==null?null:t.constants;t==null&&(t={});let a=Z6(e,r,n,this.numConstants);if(e=a.inputs,r=a.initialState,n=a.constants,Array.isArray(e)&&(r=e.slice(1),e=e[0]),(r==null||r.length===0)&&n==null)return super.apply(e,t);let s=[],i=[];if(r!=null){let l=r.length;if(l%2>0)throw new H("When passing `initialState` to a Bidrectional RNN, the state should be an Array containing the states of the underlying RNNs.");t.initialState=r,s.push(...r);let u=r.map(d=>new Zt({shape:d.shape}));this.forwardLayer.stateSpec=u.slice(0,l/2),this.backwardLayer.stateSpec=u.slice(l/2),i.push(...u)}if(n!=null)throw new Ve("Support for constants in Bidirectional layers is not implemented yet.");let o=s[0]instanceof pa;for(let l of s)if(l instanceof pa!==o)throw new H("The initial state of a Bidirectional layer cannot be specified as a mix of symbolic and non-symbolic tensors");if(o){let l=[e].concat(s),u=this.inputSpec.concat(i),d=this.inputSpec;this.inputSpec=u;let h=super.apply(l,t);return this.inputSpec=d,h}else return super.apply(e,t)}call(e,t){return X(()=>{let r=t.initialState,n,a;if(r==null)n=this.forwardLayer.call(e,t),a=this.backwardLayer.call(e,t);else{let o=r.slice(0,r.length/2),l=r.slice(r.length/2);n=this.forwardLayer.call(e,Object.assign(t,{initialState:o})),a=this.backwardLayer.call(e,Object.assign(t,{initialState:l}))}let s;this.returnState&&(Array.isArray(n)&&(s=n.slice(1).concat(a.slice(1))),n=n[0],a=a[0]),this.returnSequences&&(a=_n(a,1));let i;return this.mergeMode==="concat"?i=IA([n,a]):this.mergeMode==="sum"?i=le(n,a):this.mergeMode==="ave"?i=L(.5,le(n,a)):this.mergeMode==="mul"?i=L(n,a):this.mergeMode==null&&(i=[n,a]),this.returnState?this.mergeMode==null?i.concat(s):[i].concat(s):i})}resetStates(e){this.forwardLayer.resetStates(),this.backwardLayer.resetStates()}build(e){Ro(this.forwardLayer.name,()=>{this.forwardLayer.build(e)}),Ro(this.backwardLayer.name,()=>{this.backwardLayer.build(e)}),this.built=!0}computeMask(e,t){Array.isArray(t)&&(t=t[0]);let r;if(this.returnSequences?this.mergeMode==null?r=[t,t]:r=t:this.mergeMode==null?r=[null,null]:r=null,this.returnState){let n=this.forwardLayer.states.map(a=>null);return Array.isArray(r)?r.concat(n).concat(n):[r].concat(n).concat(n)}else return r}get trainableWeights(){return this.forwardLayer.trainableWeights.concat(this.backwardLayer.trainableWeights)}get nonTrainableWeights(){return this.forwardLayer.nonTrainableWeights.concat(this.backwardLayer.nonTrainableWeights)}setFastWeightInitDuringBuild(e){super.setFastWeightInitDuringBuild(e),this.forwardLayer!=null&&this.forwardLayer.setFastWeightInitDuringBuild(e),this.backwardLayer!=null&&this.backwardLayer.setFastWeightInitDuringBuild(e)}getConfig(){let e={mergeMode:this.mergeMode},t=super.getConfig();return Object.assign(e,t),e}static fromConfig(e,t){let r=fa(t.layer);if(delete t.layer,t.numConstants!=null)throw new Ve("Deserialization of a Bidirectional layer with numConstants present is not supported yet.");let n=t;return n.layer=r,new e(n)}};jx.className="Bidirectional";ue.registerClass(jx);function qU(e){return new xd(e)}function XU(e){return new ZA(e)}function KU(e){return new qA(e)}function ZU(e){return new XA(e)}function YU(e){return new KA(e)}function JU(e){return new JA(e)}function QU(e){return new YA(e)}function eG(e){return new ax(e)}function tG(e){return new cm(e)}function rG(e){return new tx(e)}function nG(e){return new fm(e)}function aG(e){return new rx(e)}function sG(e){return new nx(e)}function iG(e){return new sx(e)}function oG(e){return new ix(e)}function lG(e){return new ox(e)}function uG(e){return new mx(e)}function dG(e){return new cx(e)}function pG(e){return new xm(e)}function hG(e){return new hx(e)}function cG(e){return new fx(e)}function fG(e){return new gx(e)}function mG(e){return new yx(e)}function gG(e){return new Ax(e)}function yG(e){return new bx(e)}function AG(e){return new vx(e)}function xG(e){return new kx(e)}function bG(e){return new Cx(e)}function vG(e){return new Ix(e)}function wG(e){return new Sx(e)}function kG(e){return new wx(e)}function IG(e){return new Tx(e)}function SG(e){return new $x(e)}function CG(e){return new Mx(e)}function TG(e){return new Fx(e)}function Hx(e){return new _x(e)}function NG(e){return Hx(e)}function EG(e){return Hx(e)}function qx(e){return new Ox(e)}function RG(e){return qx(e)}function $G(e){return qx(e)}function Xx(e){return new Lx(e)}function MG(e){return Xx(e)}function FG(e){return Xx(e)}function PG(e){return new Bx(e)}function _G(e){return new Vx(e)}function ok(e){return new Wx(e)}function lk(e){return new Ux(e)}function uk(e){return new Px(e)}function dk(e){return new zx(e)}function zG(e){return new Dx(e)}function OG(e){return new ux(e)}function DG(e){return new gm(e)}function LG(e){return new dx(e)}function BG(e){return new _h(e)}function WG(e){return new lx(e)}function VG(e){return new mm(e)}function UG(e){return new px(e)}function GG(e){return new Am(e)}function jG(e){return new as(e)}function HG(e){return new ym(e)}function qG(e){return new jx(e)}function XG(e){return new Gx(e)}var KG=ok,ZG=lk,YG=uk,JG=dk;function QG(e){return new Nx(e)}function ej(e){return new Ex(e)}function tj(e){return new Rx(e)}function rj(e){return new xx(e)}var pk={};Le(pk,{MAPE:()=>cj,MSE:()=>gj,binaryAccuracy:()=>nj,binaryCrossentropy:()=>aj,categoricalAccuracy:()=>ij,categoricalCrossentropy:()=>oj,cosineProximity:()=>dj,mape:()=>fj,meanAbsoluteError:()=>pj,meanAbsolutePercentageError:()=>hj,meanSquaredError:()=>mj,mse:()=>yj,precision:()=>lj,recall:()=>uj,sparseCategoricalAccuracy:()=>sj});function nj(e,t){return BA(e,t)}function aj(e,t){return v6(e,t)}function sj(e,t){return w6(e,t)}function ij(e,t){return WA(e,t)}function oj(e,t){return VA(e,t)}function lj(e,t){return b6(e,t)}function uj(e,t){return YV(e,t)}function dj(e,t){return LA(e,t)}function pj(e,t){return dm(e,t)}function hj(e,t){return bd(e,t)}function cj(e,t){return bd(e,t)}function fj(e,t){return bd(e,t)}function mj(e,t){return zl(e,t)}function gj(e,t){return zl(e,t)}function yj(e,t){return zl(e,t)}var hk={};Le(hk,{modelFromJSON:()=>TU});var ck={};Le(ck,{l1:()=>xj,l1l2:()=>Aj,l2:()=>bj});function Aj(e){return new Mh(e)}function xj(e){return _U(e)}function bj(e){return zU(e)}var fk=class extends Ru{constructor(){super(...arguments),this.model=null}setModel(e){if(!(e instanceof Ka))throw new Error("model must be a LayersModel, not some other Container");this.model=e}};function Jc(e,t){return e<t}function k4(e,t){return e>t}var mk=class extends fk{constructor(e){if(super(),e==null&&(e={}),e.restoreBestWeights)throw new Ve("restoreBestWeights = True is not implemented in EarlyStopping yet.");this.monitor=e.monitor||"val_loss",this.minDelta=Math.abs(e.minDelta||0),this.patience=e.patience||0,this.verbose=e.verbose||0,this.mode=e.mode||"auto",this.baseline=e.baseline,["auto","min","max"].indexOf(this.mode)===-1&&(console.warn(`EarlyStopping mode '${this.mode}' is invalid. Falling back to mode 'auto'.`),this.mode="auto"),this.mode==="min"?this.monitorFunc=Jc:this.mode==="max"?this.monitorFunc=k4:this.monitor.indexOf("acc")!==-1?this.monitorFunc=k4:this.monitorFunc=Jc,this.monitorFunc===Jc&&(this.minDelta*=-1)}async onTrainBegin(e){this.wait=0,this.stoppedEpoch=0,this.baseline!=null?this.best=this.baseline:this.best=this.monitorFunc===Jc?1/0:-1/0}async onEpochEnd(e,t){await Ss(t);let r=this.getMonitorValue(t);r!=null&&(this.monitorFunc(r-this.minDelta,this.best)?(this.best=r,this.wait=0):(this.wait++,this.wait>=this.patience&&(this.stoppedEpoch=e,this.model.stopTraining=!0)))}async onTrainEnd(e){this.stoppedEpoch>0&&this.verbose&&console.log(`Epoch ${this.stoppedEpoch}: early stopping.`)}getMonitorValue(e){e==null&&(e={});let t=e[this.monitor];return t==null&&console.warn(`Metric for EarlyStopping ${this.monitor} is not available. Available metrics are: ${Object.keys(e)}`),t}};function vj(e){return new mk(e)}var wj={earlyStopping:vj},kj=Z();kj.registerFlag("KEEP_INTERMEDIATE_TENSORS",()=>!1,e=>{e&&console.warn("Keep intermediate tensors is ON. This will print the values of all intermediate tensors during model inference. Not all models support this mode. For details, check e2e/benchmarks/ model_config.js. This significantly impacts performance.")});var gk=(e=>(e[e.DT_INVALID=0]="DT_INVALID",e[e.DT_FLOAT=1]="DT_FLOAT",e[e.DT_DOUBLE=2]="DT_DOUBLE",e[e.DT_INT32=3]="DT_INT32",e[e.DT_UINT8=4]="DT_UINT8",e[e.DT_INT16=5]="DT_INT16",e[e.DT_INT8=6]="DT_INT8",e[e.DT_STRING=7]="DT_STRING",e[e.DT_COMPLEX64=8]="DT_COMPLEX64",e[e.DT_INT64=9]="DT_INT64",e[e.DT_BOOL=10]="DT_BOOL",e[e.DT_QINT8=11]="DT_QINT8",e[e.DT_QUINT8=12]="DT_QUINT8",e[e.DT_QINT32=13]="DT_QINT32",e[e.DT_BFLOAT16=14]="DT_BFLOAT16",e[e.DT_QINT16=15]="DT_QINT16",e[e.DT_QUINT16=16]="DT_QUINT16",e[e.DT_UINT16=17]="DT_UINT16",e[e.DT_COMPLEX128=18]="DT_COMPLEX128",e[e.DT_HALF=19]="DT_HALF",e[e.DT_RESOURCE=20]="DT_RESOURCE",e[e.DT_VARIANT=21]="DT_VARIANT",e[e.DT_UINT32=22]="DT_UINT32",e[e.DT_UINT64=23]="DT_UINT64",e[e.DT_FLOAT_REF=101]="DT_FLOAT_REF",e[e.DT_DOUBLE_REF=102]="DT_DOUBLE_REF",e[e.DT_INT32_REF=103]="DT_INT32_REF",e[e.DT_UINT8_REF=104]="DT_UINT8_REF",e[e.DT_INT16_REF=105]="DT_INT16_REF",e[e.DT_INT8_REF=106]="DT_INT8_REF",e[e.DT_STRING_REF=107]="DT_STRING_REF",e[e.DT_COMPLEX64_REF=108]="DT_COMPLEX64_REF",e[e.DT_INT64_REF=109]="DT_INT64_REF",e[e.DT_BOOL_REF=110]="DT_BOOL_REF",e[e.DT_QINT8_REF=111]="DT_QINT8_REF",e[e.DT_QUINT8_REF=112]="DT_QUINT8_REF",e[e.DT_QINT32_REF=113]="DT_QINT32_REF",e[e.DT_BFLOAT16_REF=114]="DT_BFLOAT16_REF",e[e.DT_QINT16_REF=115]="DT_QINT16_REF",e[e.DT_QUINT16_REF=116]="DT_QUINT16_REF",e[e.DT_UINT16_REF=117]="DT_UINT16_REF",e[e.DT_COMPLEX128_REF=118]="DT_COMPLEX128_REF",e[e.DT_HALF_REF=119]="DT_HALF_REF",e[e.DT_RESOURCE_REF=120]="DT_RESOURCE_REF",e[e.DT_VARIANT_REF=121]="DT_VARIANT_REF",e[e.DT_UINT32_REF=122]="DT_UINT32_REF",e[e.DT_UINT64_REF=123]="DT_UINT64_REF",e))(gk||{}),I4;(e=>{let t;(r=>{r[r.LEGACY=0]="LEGACY",r[r.V1=1]="V1",r[r.V2=2]="V2"})(t=e.CheckpointFormatVersion||(e.CheckpointFormatVersion={}))})(I4||(I4={}));var Kx={};function Ij(e,t){let r={tfOpName:e,category:"custom",inputs:[],attrs:[],customExecutor:t};Kx[e]=r}function yk(e){return Kx[e]}function Sj(e){delete Kx[e]}function k(e,t,r,n,a){let s=t.inputParams[e];if(s&&s.inputIndexStart!==void 0){let o=s.inputIndexStart,l=s.inputIndexEnd===0?void 0:s.inputIndexEnd===void 0?o+1:s.inputIndexEnd;if(s.type==="tensor")return Lr(t.inputNames[s.inputIndexStart],r,n,a);if(s.type==="tensors")return t.inputNames.slice(o,l).map(h=>Lr(h,r,n,a));let u=Lr(t.inputNames.slice(o)[0],r,n,a),d=u.dataSync();return s.type==="number"?d[0]:v.toNestedArray(u.shape,d)}let i=t.attrParams[e];return i&&i.value}function Lr(e,t,r,n){let[a,s]=dn(e);if(n!=null){let o=n.getHashTableHandleByName(a);if(o!=null)return o}let i=r.currentContextIds.find(o=>!!t[D0(a,o)]);return i!==void 0?t[D0(a,i)][s]:void 0}function Cj(e,t,r){return t[D0(e,r.currentContextId)]}function Ra(e,t){let[r,n,a]=dn(e);return[D0(r,t&&t.currentContextId),n,a]}function D0(e,t){return t?`${e}-${t}`:e}function dn(e){let t=e.split(":");if(t.length===1)return[e,0,void 0];let r=t[0],n=t.length===3?t[1]:void 0,a=Number(t[t.length-1]);return[r,a,n]}function o0(e,t,r){let n=k("pad",e,t,r);if(n==="explicit"){n=k("explicitPaddings",e,t,r);let a=[[0,0],[0,0],[0,0],[0,0]];for(let s=0;s<4;s++)a[s][0]=n[s*2],a[s][1]=n[s*2+1];return a}return n}function Ha(e){return e.kept?e:Wr(e)}var Ak={};Le(Ak,{json:()=>Tj});var Tj=[{tfOpName:"Add",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"AddV2",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"AddN",category:"arithmetic",inputs:[{start:0,end:0,name:"tensors",type:"tensors"}]},{tfOpName:"BiasAdd",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0}]},{tfOpName:"Sub",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"RealDiv",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Div",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"DivNoNan",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"FloorDiv",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Mul",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Maximum",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Minimum",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Pow",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"SquaredDifference",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Mod",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"FloorMod",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]}],xk={};Le(xk,{json:()=>Nj});var Nj=[{tfOpName:"Abs",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Acos",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Asin",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Atan",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Atan2",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"y",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Ceil",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"ClipByValue",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"clipValueMin",type:"number"},{start:2,name:"clipValueMax",type:"number"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Complex",category:"basic_math",inputs:[{start:0,name:"real",type:"tensor"},{start:1,name:"imag",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"ComplexAbs",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Cos",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Cosh",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Elu",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Exp",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Floor",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Log",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Imag",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"Tout",name:"outputType",type:"dtype",notSupported:!0}]},{tfOpName:"Neg",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Real",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"Tout",name:"outputType",type:"dtype",notSupported:!0}]},{tfOpName:"Prelu",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"alpha",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Relu",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Relu6",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Selu",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Sigmoid",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Sin",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Sinh",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Sqrt",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Rsqrt",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Square",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Tan",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Tanh",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Sign",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Round",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Expm1",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Log1p",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Reciprocal",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Softplus",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Asinh",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Acosh",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Atanh",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Erf",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Prod",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axes",type:"number[]"}],attrs:[{tfName:"keep_dims",name:"keepDims",type:"bool",notSupported:!0},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"LeakyRelu",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"alpha",name:"alpha",type:"number",defaultValue:.2},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"IsNan",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]}],bk={};Le(bk,{json:()=>Ej});var Ej=[{tfOpName:"EmptyTensorList",category:"control",inputs:[{start:0,name:"elementShape",type:"shape"},{start:1,name:"maxNumElements",type:"number"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"LoopCond",category:"control",inputs:[{start:0,name:"pred",type:"tensor"}]},{tfOpName:"Switch",category:"control",inputs:[{start:0,name:"data",type:"tensor"},{start:1,name:"pred",type:"tensor"}]},{tfOpName:"Merge",category:"control",inputs:[{start:0,end:0,name:"tensors",type:"tensors"}]},{tfOpName:"Enter",category:"control",inputs:[{start:0,name:"tensor",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"frame_name",name:"frameName",type:"string"},{tfName:"is_constant",name:"isConstant",type:"bool"}]},{tfOpName:"Exit",category:"control",inputs:[{start:0,name:"tensor",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"NextIteration",category:"control",inputs:[{start:0,name:"tensor",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"TensorArrayV3",category:"control",inputs:[{start:0,name:"size",type:"number"}],attrs:[{tfName:"dtype",name:"dtype",type:"dtype"},{tfName:"element_shape",name:"elementShape",type:"shape"},{tfName:"dynamic_size",name:"dynamicSize",type:"bool"},{tfName:"clear_after_read",name:"clearAfterRead",type:"bool"},{tfName:"identical_element_shapes",name:"identicalElementShapes",type:"bool"},{tfName:"tensor_array_name",name:"name",type:"string"}]},{tfOpName:"TensorArrayWriteV3",category:"control",inputs:[{start:0,name:"tensorArrayId",type:"tensor"},{start:1,name:"index",type:"number"},{start:2,name:"tensor",type:"tensor"},{start:3,name:"flowIn",type:"number"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"TensorArrayReadV3",category:"control",inputs:[{start:0,name:"tensorArrayId",type:"tensor"},{start:1,name:"index",type:"number"},{start:2,name:"flowIn",type:"number"}],attrs:[{tfName:"dtype",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"TensorArrayGatherV3",category:"control",inputs:[{start:0,name:"tensorArrayId",type:"tensor"},{start:1,name:"indices",type:"number[]"},{start:2,name:"flowIn",type:"number"}],attrs:[{tfName:"dtype",name:"dtype",type:"dtype"},{tfName:"element_shape",name:"elementShape",type:"shape"}]},{tfOpName:"TensorArrayScatterV3",category:"control",inputs:[{start:0,name:"tensorArrayId",type:"tensor"},{start:1,name:"indices",type:"number[]"},{start:2,name:"tensor",type:"tensor"},{start:3,name:"flowIn",type:"number"}],attrs:[{tfName:"T",name:"dtype",type:"dtype"}]},{tfOpName:"TensorArrayConcatV3",category:"control",inputs:[{start:0,name:"tensorArrayId",type:"tensor"},{start:1,name:"flowIn",type:"number"}],attrs:[{tfName:"dtype",name:"dtype",type:"dtype"},{tfName:"element_shape_except0",name:"elementShapeExcept0",type:"shape",notSupported:!0}]},{tfOpName:"TensorArraySplitV3",category:"control",inputs:[{start:0,name:"tensorArrayId",type:"tensor"},{start:1,name:"tensor",type:"tensor"},{start:2,name:"lengths",type:"number[]"},{start:3,name:"flowIn",type:"number"}],attrs:[{tfName:"T",name:"dtype",type:"dtype"}]},{tfOpName:"TensorArraySizeV3",category:"control",inputs:[{start:0,name:"tensorArrayId",type:"tensor"},{start:1,name:"flowIn",type:"number"}]},{tfOpName:"TensorArrayCloseV3",category:"control",inputs:[{start:0,name:"tensorArrayId",type:"tensor"}]},{tfOpName:"StatelessIf",category:"control",inputs:[{start:0,name:"cond",type:"tensor"},{start:1,end:0,name:"args",type:"tensors"}],attrs:[{tfName:"then_branch",name:"thenBranch",type:"func"},{tfName:"else_branch",name:"elseBranch",type:"func"}]},{tfOpName:"If",category:"control",inputs:[{start:0,name:"cond",type:"tensor"},{start:1,end:0,name:"args",type:"tensors"}],attrs:[{tfName:"then_branch",name:"thenBranch",type:"func"},{tfName:"else_branch",name:"elseBranch",type:"func"}]},{tfOpName:"StatelessWhile",category:"control",inputs:[{start:0,end:0,name:"args",type:"tensors"}],attrs:[{tfName:"cond",name:"cond",type:"func"},{tfName:"body",name:"body",type:"func"}]},{tfOpName:"While",category:"control",inputs:[{start:0,end:0,name:"args",type:"tensors"}],attrs:[{tfName:"cond",name:"cond",type:"func"},{tfName:"body",name:"body",type:"func"}]},{tfOpName:"TensorListScatter",category:"control",inputs:[{start:0,name:"tensor",type:"tensor"},{start:1,name:"indices",type:"number[]"},{start:2,name:"elementShape",type:"shape"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListScatterV2",category:"control",inputs:[{start:0,name:"tensor",type:"tensor"},{start:1,name:"indices",type:"number[]"},{start:2,name:"elementShape",type:"shape"},{start:3,name:"numElements",type:"number"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListGather",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"},{start:1,name:"indices",type:"number[]"},{start:2,name:"elementShape",type:"shape"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListGetItem",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"},{start:1,name:"index",type:"number"},{start:2,name:"elementShape",type:"shape"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListSetItem",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"},{start:1,name:"index",type:"number"},{start:2,name:"tensor",type:"tensor"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListReserve",category:"control",inputs:[{start:0,name:"elementShape",type:"shape"},{start:1,name:"numElements",type:"number"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListFromTensor",category:"control",inputs:[{start:0,name:"tensor",type:"tensor"},{start:1,name:"elementShape",type:"shape"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListStack",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"},{start:1,name:"elementShape",type:"shape"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"},{tfName:"num_elements",name:"numElements",type:"dtype"}]},{tfOpName:"TensorListSplit",category:"control",inputs:[{start:0,name:"tensor",type:"tensor"},{start:1,name:"elementShape",type:"shape"},{start:2,name:"lengths",type:"number[]"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListConcat",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"}],attrs:[{tfName:"element_shape",name:"elementShape",type:"shape"},{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListConcatV2",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"}],attrs:[{tfName:"element_shape",name:"elementShape",type:"shape"},{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListPopBack",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"},{start:1,name:"elementShape",type:"shape"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListPushBack",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"},{start:1,name:"tensor",type:"tensor"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListLength",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"}]},{tfOpName:"TensorListResize",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"},{start:1,name:"size",type:"number"}]}],vk={};Le(vk,{json:()=>Rj});var Rj=[{tfOpName:"AvgPool",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0},{tfName:"ksize",name:"kernelSize",type:"number[]"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"MaxPool",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0},{tfName:"ksize",name:"kernelSize",type:"number[]"},{tfName:"explicit_paddings",name:"explicitPaddings",type:"number[]",defaultValue:[],notSupported:!0},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"MaxPoolWithArgmax",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"ksize",name:"kernelSize",type:"number[]"},{tfName:"include_batch_in_index",name:"includeBatchInIndex",type:"bool"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"AvgPool3D",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0},{tfName:"ksize",name:"kernelSize",type:"number[]"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"MaxPool3D",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0},{tfName:"ksize",name:"kernelSize",type:"number[]"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Conv1D",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"filter",type:"tensor"}],attrs:[{tfName:"stride",name:"stride",type:"number"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",defaultValue:"NWC"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"dilation",name:"dilation",type:"number",defaultValue:1}]},{tfOpName:"Conv2D",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"filter",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"useCudnnOnGpu",name:"useCudnnOnGpu",type:"bool"},{tfName:"data_format",name:"dataFormat",type:"string",defaultValue:"NHWC"},{tfName:"explicit_paddings",name:"explicitPaddings",type:"number[]",defaultValue:[]},{tfName:"dilations",name:"dilations",type:"number[]"}]},{tfOpName:"_FusedConv2D",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"filter",type:"tensor"},{start:2,end:0,name:"args",type:"tensors"}],attrs:[{tfName:"num_args",name:"numArgs",type:"number"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"explicit_paddings",name:"explicitPaddings",type:"number[]",defaultValue:[]},{tfName:"use_cudnn_on_gpu",name:"useCudnnOnGpu",type:"bool",defaultValue:!0},{tfName:"data_format",name:"dataFormat",type:"string",defaultValue:"NHWC"},{tfName:"dilations",name:"dilations",type:"number[]",defaultValue:[1,1,1,1]},{tfName:"fused_ops",name:"fusedOps",type:"string[]",defaultValue:[]},{tfName:"epsilon",name:"epsilon",type:"number",defaultValue:1e-4},{tfName:"leakyrelu_alpha",name:"leakyreluAlpha",type:"number"}]},{tfOpName:"Conv2DBackpropInput",category:"convolution",inputs:[{start:2,name:"x",type:"tensor"},{start:1,name:"filter",type:"tensor"},{start:0,name:"outputShape",type:"number[]"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0},{tfName:"explicit_paddings",name:"explicitPaddings",type:"number[]",defaultValue:[]},{tfName:"dilations",name:"dilations",type:"number[]",notSupported:!0}]},{tfOpName:"DepthwiseConv2d",category:"convolution",inputs:[{start:0,name:"input",type:"tensor"},{start:1,name:"filter",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",defaultValue:"NHWC"},{tfName:"explicit_paddings",name:"explicitPaddings",type:"number[]",defaultValue:[]},{tfName:"dilations",name:"dilations",type:"number[]"}]},{tfOpName:"DepthwiseConv2dNative",category:"convolution",inputs:[{start:0,name:"input",type:"tensor"},{start:1,name:"filter",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",defaultValue:"NHWC"},{tfName:"explicit_paddings",name:"explicitPaddings",type:"number[]",defaultValue:[]},{tfName:"dilations",name:"dilations",type:"number[]"}]},{tfOpName:"FusedDepthwiseConv2dNative",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"filter",type:"tensor"},{start:2,end:0,name:"args",type:"tensors"}],attrs:[{tfName:"num_args",name:"numArgs",type:"number"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",defaultValue:"NHWC"},{tfName:"dilations",name:"dilations",type:"number[]",defaultValue:[1,1,1,1]},{tfName:"fused_ops",name:"fusedOps",type:"string[]",defaultValue:[]},{tfName:"explicit_paddings",name:"explicitPaddings",type:"number[]",defaultValue:[]}]},{tfOpName:"Conv3D",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"filter",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",defaultValue:"NHWC"},{tfName:"dilations",name:"dilations",type:"number[]"}]},{tfOpName:"Dilation2D",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"filter",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"rates",name:"dilations",type:"number[]"},{tfName:"padding",name:"pad",type:"string"}]}],wk={};Le(wk,{json:()=>$j});var $j=[{tfOpName:"Fill",category:"creation",inputs:[{start:0,name:"shape",type:"number[]"},{start:1,name:"value",type:"number"}],attrs:[{tfName:"T",name:"dtype",type:"dtype"}]},{tfOpName:"LinSpace",category:"creation",inputs:[{start:0,name:"start",type:"number"},{start:1,name:"stop",type:"number"},{start:2,name:"num",type:"number"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"OneHot",category:"creation",inputs:[{start:0,name:"indices",type:"tensor"},{start:1,name:"depth",type:"number"},{start:2,name:"onValue",type:"number",defaultValue:1},{start:3,name:"offValue",type:"number",defaultValue:0}],attrs:[{tfName:"axis",name:"axis",type:"number",notSupported:!0},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Ones",category:"creation",inputs:[{start:0,name:"shape",type:"number[]"}],attrs:[{tfName:"T",name:"dtype",type:"dtype"}]},{tfOpName:"OnesLike",category:"creation",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"dtype",name:"dtype",type:"dtype"}]},{tfOpName:"RandomUniform",category:"creation",inputs:[{start:0,name:"shape",type:"number[]"}],attrs:[{tfName:"minval",name:"minval",type:"number",defaultValue:0},{tfName:"maxval",name:"maxval",type:"number",defaultValue:1},{tfName:"dtype",name:"dtype",type:"dtype"},{tfName:"seed",name:"seed",type:"number",defaultValue:0},{tfName:"seed2",name:"seed2",type:"number",defaultValue:0,notSupported:!0},{tfName:"T",name:"T",type:"number",notSupported:!0}]},{tfOpName:"Range",category:"creation",inputs:[{start:0,name:"start",type:"number"},{start:1,name:"stop",type:"number"},{start:2,name:"step",type:"number",defaultValue:0}],attrs:[{tfName:"Tidx",name:"dtype",type:"dtype"}]},{tfOpName:"TruncatedNormal",category:"creation",inputs:[{start:0,name:"shape",type:"number[]"}],attrs:[{tfName:"means",name:"mean",type:"number",defaultValue:0},{tfName:"stddev",name:"stdDev",type:"number",defaultValue:1},{tfName:"seed",name:"seed",type:"number"},{tfName:"seed2",name:"seed2",type:"number",defaultValue:0,notSupported:!0},{tfName:"dtype",name:"dtype",type:"dtype"},{tfName:"T",name:"T",type:"number",notSupported:!0}]},{tfOpName:"Zeros",category:"creation",inputs:[{start:0,name:"shape",type:"number[]"}],attrs:[{tfName:"T",name:"dtype",type:"dtype"}]},{tfOpName:"ZerosLike",category:"creation",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype"}]},{tfOpName:"Multinomial",category:"creation",inputs:[{start:0,name:"logits",type:"tensor"},{start:1,name:"numSamples",type:"number"}],attrs:[{tfName:"seed",name:"seed",type:"number"},{tfName:"seed2",name:"seed2",type:"number"},{tfName:"T",name:"dtype",type:"dtype"},{tfName:"output_dtype",name:"output_dtype",type:"dtype"}]}],kk={};Le(kk,{json:()=>Mj});var Mj=[{tfOpName:"NonMaxSuppressionV2",category:"dynamic",inputs:[{start:0,name:"boxes",type:"tensor"},{start:1,name:"scores",type:"tensor"},{start:2,name:"maxOutputSize",type:"number"},{start:3,name:"iouThreshold",type:"number"}]},{tfOpName:"NonMaxSuppressionV3",category:"dynamic",inputs:[{start:0,name:"boxes",type:"tensor"},{start:1,name:"scores",type:"tensor"},{start:2,name:"maxOutputSize",type:"number"},{start:3,name:"iouThreshold",type:"number"},{start:4,name:"scoreThreshold",type:"number"}]},{tfOpName:"NonMaxSuppressionV4",category:"dynamic",inputs:[{start:0,name:"boxes",type:"tensor"},{start:1,name:"scores",type:"tensor"},{start:2,name:"maxOutputSize",type:"number"},{start:3,name:"iouThreshold",type:"number"},{start:4,name:"scoreThreshold",type:"number"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"T_threshold",name:"threshold",type:"dtype",notSupported:!0},{tfName:"pad_to_max_output_size",name:"padToMaxOutputSize",type:"bool"}]},{tfOpName:"NonMaxSuppressionV5",category:"dynamic",inputs:[{start:0,name:"boxes",type:"tensor"},{start:1,name:"scores",type:"tensor"},{start:2,name:"maxOutputSize",type:"number"},{start:3,name:"iouThreshold",type:"number"},{start:4,name:"scoreThreshold",type:"number"},{start:5,name:"softNmsSigma",type:"number"}]},{tfOpName:"Where",category:"dynamic",inputs:[{start:0,name:"condition",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"ListDiff",category:"dynamic",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"y",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]}],Ik={};Le(Ik,{json:()=>Fj});var Fj=[{tfOpName:"LowerBound",category:"evaluation",inputs:[{start:0,name:"sortedSequence",type:"tensor"},{start:1,name:"values",type:"tensor"}]},{tfOpName:"TopKV2",category:"evaluation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"k",type:"number"}],attrs:[{tfName:"sorted",name:"sorted",type:"bool"}]},{tfOpName:"UpperBound",category:"evaluation",inputs:[{start:0,name:"sortedSequence",type:"tensor"},{start:1,name:"values",type:"tensor"}]},{tfOpName:"Unique",category:"evaluation",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"UniqueV2",category:"evaluation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number"}]}],Sk={};Le(Sk,{json:()=>Pj});var Pj=[{tfOpName:"PlaceholderWithDefault",category:"graph",inputs:[{start:0,name:"default",type:"tensor"}],attrs:[{tfName:"shape",name:"shape",type:"shape"},{tfName:"dtype",name:"dtype",type:"dtype"}]},{tfOpName:"Placeholder",category:"graph",attrs:[{tfName:"shape",name:"shape",type:"shape"},{tfName:"dtype",name:"dtype",type:"dtype"}]},{tfOpName:"Const",category:"graph"},{tfOpName:"Identity",category:"graph",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"IdentityN",category:"graph",inputs:[{start:0,end:0,name:"x",type:"tensors"}]},{tfOpName:"Snapshot",category:"graph",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"Rank",category:"graph",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"Size",category:"graph",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"Shape",category:"graph",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"ShapeN",category:"graph",inputs:[{start:0,end:0,name:"x",type:"tensors"}]},{tfOpName:"Print",category:"graph",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"data",type:"tensors"}],attrs:[{tfName:"message",name:"message",type:"string"},{tfName:"first_n",name:"firstN",type:"number",notSupported:!0},{tfName:"summarize",name:"summarize",type:"number",defaultValue:3}]},{tfOpName:"NoOp",category:"graph",inputs:[]},{tfOpName:"StopGradient",category:"graph",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"FakeQuantWithMinMaxVars",category:"graph",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"min",name:"min",type:"number"},{tfName:"max",name:"max",type:"number"}]}],Ck={};Le(Ck,{json:()=>_j});var _j=[{tfOpName:"HashTable",category:"hash_table",inputs:[],attrs:[{tfName:"shared_name",name:"sharedName",type:"string"},{tfName:"use_node_name_sharing",name:"useNodeNameSharing",type:"bool"},{tfName:"key_dtype",name:"keyDType",type:"dtype"},{tfName:"value_dtype",name:"valueDType",type:"dtype"}]},{tfOpName:"HashTableV2",category:"hash_table",inputs:[],attrs:[{tfName:"shared_name",name:"sharedName",type:"string"},{tfName:"use_node_name_sharing",name:"useNodeNameSharing",type:"bool"},{tfName:"key_dtype",name:"keyDType",type:"dtype"},{tfName:"value_dtype",name:"valueDType",type:"dtype"}]},{tfOpName:"LookupTableImport",category:"hash_table",inputs:[{start:0,name:"tableHandle",type:"tensor"},{start:1,name:"keys",type:"tensor"},{start:2,name:"values",type:"tensor"}],attrs:[{tfName:"Tin",name:"tIn",type:"dtype",notSupported:!0},{tfName:"Tout",name:"tOut",type:"dtype",notSupported:!0}]},{tfOpName:"LookupTableImportV2",category:"hash_table",inputs:[{start:0,name:"tableHandle",type:"tensor"},{start:1,name:"keys",type:"tensor"},{start:2,name:"values",type:"tensor"}],attrs:[{tfName:"Tin",name:"tIn",type:"dtype",notSupported:!0},{tfName:"Tout",name:"tOut",type:"dtype",notSupported:!0}]},{tfOpName:"LookupTableFind",category:"hash_table",inputs:[{start:0,name:"tableHandle",type:"tensor"},{start:1,name:"keys",type:"tensor"},{start:2,name:"defaultValue",type:"tensor"}],attrs:[{tfName:"Tin",name:"tIn",type:"dtype",notSupported:!0},{tfName:"Tout",name:"tOut",type:"dtype",notSupported:!0}]},{tfOpName:"LookupTableFindV2",category:"hash_table",inputs:[{start:0,name:"tableHandle",type:"tensor"},{start:1,name:"keys",type:"tensor"},{start:2,name:"defaultValue",type:"tensor"}],attrs:[{tfName:"Tin",name:"tIn",type:"dtype",notSupported:!0},{tfName:"Tout",name:"tOut",type:"dtype",notSupported:!0}]},{tfOpName:"LookupTableSize",category:"hash_table",inputs:[{start:0,name:"tableHandle",type:"tensor"}]},{tfOpName:"LookupTableSizeV2",category:"hash_table",inputs:[{start:0,name:"tableHandle",type:"tensor"}]}],Tk={};Le(Tk,{json:()=>zj});var zj=[{tfOpName:"ResizeBilinear",category:"image",inputs:[{start:0,name:"images",type:"tensor"},{start:1,name:"size",type:"number[]"}],attrs:[{tfName:"align_corners",name:"alignCorners",type:"bool"},{tfName:"half_pixel_centers",name:"halfPixelCenters",type:"bool"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"ResizeNearestNeighbor",category:"image",inputs:[{start:0,name:"images",type:"tensor"},{start:1,name:"size",type:"number[]"}],attrs:[{tfName:"align_corners",name:"alignCorners",type:"bool"},{tfName:"half_pixel_centers",name:"halfPixelCenters",type:"bool"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"CropAndResize",category:"image",inputs:[{start:0,name:"image",type:"tensor"},{start:1,name:"boxes",type:"tensor"},{start:2,name:"boxInd",type:"tensor"},{start:3,name:"cropSize",type:"number[]"}],attrs:[{tfName:"method",name:"method",type:"string"},{tfName:"extrapolation_value",name:"extrapolationValue",type:"number"}]},{tfOpName:"ImageProjectiveTransformV3",category:"image",inputs:[{start:0,name:"images",type:"tensor"},{start:1,name:"transforms",type:"tensor"},{start:2,name:"outputShape",type:"number[]"},{start:3,name:"fillValue",type:"number"}],attrs:[{tfName:"interpolation",name:"interpolation",type:"string"},{tfName:"fill_mode",name:"fillMode",type:"string"}]}],Nk={};Le(Nk,{json:()=>Oj});var Oj=[{tfOpName:"Equal",category:"logical",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"NotEqual",category:"logical",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Greater",category:"logical",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"GreaterEqual",category:"logical",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Less",category:"logical",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"LessEqual",category:"logical",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"LogicalAnd",category:"logical",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"LogicalNot",category:"logical",inputs:[{start:0,name:"a",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"LogicalOr",category:"logical",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Select",category:"logical",inputs:[{start:0,name:"condition",type:"tensor"},{start:1,name:"a",type:"tensor"},{start:2,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"SelectV2",category:"logical",inputs:[{start:0,name:"condition",type:"tensor"},{start:1,name:"a",type:"tensor"},{start:2,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]}],Ek={};Le(Ek,{json:()=>Dj});var Dj=[{tfOpName:"_FusedMatMul",category:"matrices",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"},{start:2,end:0,name:"args",type:"tensors"}],attrs:[{tfName:"num_args",name:"numArgs",type:"number"},{tfName:"fused_ops",name:"fusedOps",type:"string[]",defaultValue:[]},{tfName:"epsilon",name:"epsilon",type:"number",defaultValue:1e-4},{tfName:"transpose_a",name:"transposeA",type:"bool",defaultValue:!1},{tfName:"transpose_b",name:"transposeB",type:"bool",defaultValue:!1},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"MatMul",category:"matrices",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"transpose_a",name:"transposeA",type:"bool",defaultValue:!1},{tfName:"transpose_b",name:"transposeB",type:"bool",defaultValue:!1},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"BatchMatMul",category:"matrices",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"adj_x",name:"transposeA",type:"bool",defaultValue:!1},{tfName:"adj_y",name:"transposeB",type:"bool",defaultValue:!1},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"BatchMatMulV2",category:"matrices",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"adj_x",name:"transposeA",type:"bool",defaultValue:!1},{tfName:"adj_y",name:"transposeB",type:"bool",defaultValue:!1},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Transpose",category:"matrices",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"perm",type:"number[]"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Einsum",category:"matrices",inputs:[{start:0,end:0,name:"tensors",type:"tensors"}],attrs:[{tfName:"equation",name:"equation",type:"string"},{tfName:"N",name:"n",type:"number",defaultValue:2},{tfName:"T",name:"dtype",type:"dtype"}]}],Rk={};Le(Rk,{json:()=>Lj});var Lj=[{tfOpName:"FusedBatchNorm",category:"normalization",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"scale",type:"tensor"},{start:2,name:"offset",type:"tensor"},{start:3,name:"mean",type:"tensor"},{start:4,name:"variance",type:"tensor"}],attrs:[{tfName:"epsilon",name:"epsilon",type:"number",defaultValue:.001},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0}]},{tfOpName:"FusedBatchNormV2",category:"normalization",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"scale",type:"tensor"},{start:2,name:"offset",type:"tensor"},{start:3,name:"mean",type:"tensor"},{start:4,name:"variance",type:"tensor"}],attrs:[{tfName:"epsilon",name:"epsilon",type:"number",defaultValue:.001},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0}]},{tfOpName:"FusedBatchNormV3",category:"normalization",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"scale",type:"tensor"},{start:2,name:"offset",type:"tensor"},{start:3,name:"mean",type:"tensor"},{start:4,name:"variance",type:"tensor"}],attrs:[{tfName:"epsilon",name:"epsilon",type:"number",defaultValue:.001},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0}]},{tfOpName:"LRN",category:"normalization",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"depth_radius",name:"radius",type:"number",defaultValue:5},{tfName:"bias",name:"bias",type:"number",defaultValue:1},{tfName:"alpha",name:"alpha",type:"number",defaultValue:1},{tfName:"beta",name:"beta",type:"number",defaultValue:.5}]},{tfOpName:"Softmax",category:"normalization",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"LogSoftmax",category:"normalization",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"SparseToDense",category:"normalization",inputs:[{start:0,name:"sparseIndices",type:"tensor"},{start:1,name:"outputShape",type:"number[]"},{start:2,name:"sparseValues",type:"tensor"},{start:3,name:"defaultValue",type:"tensor"}],attrs:[{tfName:"validate_indices",name:"validateIndices",type:"bool",defaultValue:!0,notSupported:!0}]}],$k={};Le($k,{json:()=>Bj});var Bj=[{tfOpName:"Bincount",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"size",type:"number"},{start:2,name:"weights",type:"tensor"}]},{tfOpName:"DenseBincount",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"size",type:"number"},{start:2,name:"weights",type:"tensor"}],attrs:[{tfName:"binary_output",name:"binaryOutput",type:"bool"}]},{tfOpName:"Max",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number[]"}],attrs:[{tfName:"keep_dims",name:"keepDims",type:"bool"}]},{tfOpName:"Mean",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number[]"}],attrs:[{tfName:"keep_dims",name:"keepDims",type:"bool"}]},{tfOpName:"Min",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number[]"}],attrs:[{tfName:"keep_dims",name:"keepDims",type:"bool"}]},{tfOpName:"Sum",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number[]"}],attrs:[{tfName:"keep_dims",name:"keepDims",type:"bool"}]},{tfOpName:"All",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number[]"}],attrs:[{tfName:"keep_dims",name:"keepDims",type:"bool"}]},{tfOpName:"Any",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number[]"}],attrs:[{tfName:"keep_dims",name:"keepDims",type:"bool"}]},{tfOpName:"ArgMax",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number"}]},{tfOpName:"ArgMin",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number"}]},{tfOpName:"Prod",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number[]"}],attrs:[{tfName:"keep_dims",name:"keepDims",type:"bool"}]},{tfOpName:"Cumprod",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number"}],attrs:[{tfName:"exclusive",name:"exclusive",type:"bool"},{tfName:"reverse",name:"reverse",type:"bool"}]},{tfOpName:"Cumsum",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number"}],attrs:[{tfName:"exclusive",name:"exclusive",type:"bool"},{tfName:"reverse",name:"reverse",type:"bool"}]}],Mk={};Le(Mk,{json:()=>Wj});var Wj=[{tfOpName:"ConcatV2",category:"slice_join",inputs:[{start:0,end:-1,name:"tensors",type:"tensors"},{start:-1,name:"axis",type:"number"}],attrs:[{tfName:"N",name:"n",type:"number",defaultValue:2}]},{tfOpName:"Concat",category:"slice_join",inputs:[{start:1,end:0,name:"tensors",type:"tensors"},{start:0,name:"axis",type:"number"}],attrs:[{tfName:"N",name:"n",type:"number",defaultValue:2}]},{tfOpName:"GatherV2",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"indices",type:"tensor"},{start:2,name:"axis",type:"number",defaultValue:0}],attrs:[{tfName:"batch_dims",name:"batchDims",type:"number",defaultValue:0}]},{tfOpName:"Gather",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"indices",type:"tensor"}],attrs:[{tfName:"validate_indices",name:"validateIndices",type:"bool",notSupported:!0}]},{tfOpName:"Reverse",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"dims",type:"bool[]"}]},{tfOpName:"ReverseV2",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number[]"}]},{tfOpName:"Slice",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"begin",type:"number[]"},{start:2,name:"size",type:"number[]"}]},{tfOpName:"StridedSlice",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"begin",type:"number[]"},{start:2,name:"end",type:"number[]"},{start:3,name:"strides",type:"number[]"}],attrs:[{tfName:"begin_mask",name:"beginMask",type:"number",defaultValue:0},{tfName:"end_mask",name:"endMask",type:"number",defaultValue:0},{tfName:"new_axis_mask",name:"newAxisMask",type:"number",defaultValue:0},{tfName:"ellipsis_mask",name:"ellipsisMask",type:"number",defaultValue:0},{tfName:"shrink_axis_mask",name:"shrinkAxisMask",type:"number",defaultValue:0}]},{tfOpName:"Pack",category:"slice_join",inputs:[{start:0,end:0,name:"tensors",type:"tensors"}],attrs:[{tfName:"axis",name:"axis",type:"number",defaultValue:0}]},{tfOpName:"Unpack",category:"slice_join",inputs:[{start:0,name:"tensor",type:"tensor"}],attrs:[{tfName:"axis",name:"axis",type:"number",defaultValue:0},{tfName:"num",name:"num",type:"number",defaultValue:0,notSupported:!0}]},{tfOpName:"Tile",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"reps",type:"number[]"}]},{tfOpName:"Split",category:"slice_join",inputs:[{start:0,name:"axis",type:"number",defaultValue:0},{start:1,name:"x",type:"tensor"}],attrs:[{tfName:"num_split",name:"numOrSizeSplits",type:"number",defaultValue:1}]},{tfOpName:"SplitV",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"numOrSizeSplits",type:"number[]"},{start:2,name:"axis",type:"number",defaultValue:0}]},{tfOpName:"ScatterNd",category:"slice_join",inputs:[{start:0,name:"indices",type:"tensor"},{start:1,name:"values",type:"tensor"},{start:2,name:"shape",type:"number[]"}]},{tfOpName:"GatherNd",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"indices",type:"tensor"}]},{tfOpName:"SparseToDense",category:"slice_join",inputs:[{start:0,name:"sparseIndices",type:"tensor"},{start:1,name:"outputShape",type:"number[]"},{start:2,name:"sparseValues",type:"tensor"},{start:3,name:"defaultValue",type:"tensor"}],attrs:[{tfName:"validate_indices",name:"validateIndices",type:"bool",defaultValue:!1,notSupported:!0}]}],Fk={};Le(Fk,{json:()=>Vj});var Vj=[{tfOpName:"SparseFillEmptyRows",category:"sparse",inputs:[{start:0,name:"indices",type:"tensor"},{start:1,name:"values",type:"tensor"},{start:2,name:"denseShape",type:"tensor"},{start:3,name:"defaultValue",type:"tensor"}]},{tfOpName:"SparseReshape",category:"sparse",inputs:[{start:0,name:"inputIndices",type:"tensor"},{start:1,name:"inputShape",type:"tensor"},{start:2,name:"newShape",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"SparseSegmentMean",category:"sparse",inputs:[{start:0,name:"data",type:"tensor"},{start:1,name:"indices",type:"tensor"},{start:2,name:"segmentIds",type:"tensor"}]},{tfOpName:"SparseSegmentSum",category:"sparse",inputs:[{start:0,name:"data",type:"tensor"},{start:1,name:"indices",type:"tensor"},{start:2,name:"segmentIds",type:"tensor"}]}],Pk={};Le(Pk,{json:()=>Uj});var Uj=[{tfOpName:"FFT",category:"spectral",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"IFFT",category:"spectral",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"RFFT",category:"spectral",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"fft_length",type:"number",notSupported:!0}]},{tfOpName:"IRFFT",category:"spectral",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"fft_length",type:"number",notSupported:!0}]}],_k={};Le(_k,{json:()=>Gj});var Gj=[{tfOpName:"StringNGrams",category:"string",inputs:[{start:0,name:"data",type:"tensor"},{start:1,name:"dataSplits",type:"tensor"}],attrs:[{tfName:"separator",name:"separator",type:"string"},{tfName:"ngram_widths",name:"nGramWidths",type:"number[]"},{tfName:"left_pad",name:"leftPad",type:"string"},{tfName:"right_pad",name:"rightPad",type:"string"},{tfName:"pad_width",name:"padWidth",type:"number"},{tfName:"preserve_short_sequences",name:"preserveShortSequences",type:"bool"}],outputs:["ngrams","ngrams_splits"]},{tfOpName:"StringSplit",category:"string",inputs:[{start:0,name:"input",type:"tensor"},{start:1,name:"delimiter",type:"tensor"}],attrs:[{tfName:"skip_empty",name:"skipEmpty",type:"bool"}],outputs:["indices","values","shape"]},{tfOpName:"StringToHashBucketFast",category:"string",inputs:[{start:0,name:"input",type:"tensor"}],attrs:[{tfName:"num_buckets",name:"numBuckets",type:"number"}]}],zk={};Le(zk,{json:()=>jj});var jj=[{tfOpName:"Cast",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"SrcT",name:"sdtype",type:"dtype",notSupported:!0},{tfName:"DstT",name:"dtype",type:"dtype"}]},{tfOpName:"ExpandDims",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number"}]},{tfOpName:"MirrorPad",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"padding",type:"number[]"}],attrs:[{tfName:"mode",name:"mode",type:"string"}]},{tfOpName:"Pad",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"padding",type:"number[]"}],attrs:[{tfName:"constant_value",name:"constantValue",type:"number",defaultValue:0}]},{tfOpName:"PadV2",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"padding",type:"number[]"},{start:2,name:"constantValue",type:"number",defaultValue:0}]},{tfOpName:"Reshape",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"shape",type:"number[]"}]},{tfOpName:"Squeeze",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"axis",tfDeprecatedName:"squeeze_dims",name:"axis",type:"number[]"}]},{tfOpName:"SpaceToBatchND",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"blockShape",type:"number[]"},{start:2,name:"paddings",type:"number[]"}]},{tfOpName:"BatchToSpaceND",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"blockShape",type:"number[]"},{start:2,name:"crops",type:"number[]"}]},{tfOpName:"DepthToSpace",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"block_size",name:"blockSize",type:"number"},{tfName:"data_format",name:"dataFormat",type:"string"}]},{tfOpName:"BroadcastTo",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"shape",type:"number[]"}],attrs:[]},{tfOpName:"BroadcastArgs",category:"transformation",inputs:[{start:0,name:"s0",type:"tensor"},{start:1,name:"s1",type:"tensor"}],attrs:[]}],S4=class{static get Instance(){return this._instance||(this._instance=new this)}constructor(){let e=[Ak,xk,bk,vk,wk,kk,Ik,Sk,Ck,Tk,Nk,Ek,Rk,$k,Mk,Fk,Pk,_k,zk],t=[].concat(...e.map(r=>r.json));this.opMappers=t.reduce((r,n)=>(r[n.tfOpName]=n,r),{})}transformGraph(e,t={}){let r=e.node,n=[],a=[],s=[],i=r.reduce((f,m)=>(f[m.name]=this.mapNode(m),m.op.startsWith("Placeholder")?n.push(f[m.name]):m.op==="Const"?a.push(f[m.name]):(m.input==null||m.input.length===0)&&s.push(f[m.name]),f),{}),o=[],l=[],u={},d={};t!=null&&(u=this.mapSignatureEntries(t.inputs),d=this.mapSignatureEntries(t.outputs));let h=Object.keys(i);h.forEach(f=>{let m=i[f];m.inputNames.forEach((g,y)=>{let[A,,x]=Ra(g),b=i[A];if(b.outputs!=null){let w=b.outputs.indexOf(x);if(w!==-1){let S=`${A}:${w}`;m.inputNames[y]=S}}m.inputs.push(b),b.children.push(m)})}),Object.keys(d).length===0?h.forEach(f=>{let m=i[f];m.children.length===0&&l.push(m)}):Object.keys(d).forEach(f=>{let[m]=Ra(f),g=i[m];g!=null&&(g.signatureKey=d[f],l.push(g))}),Object.keys(u).length>0?Object.keys(u).forEach(f=>{let[m]=Ra(f),g=i[m];g&&(g.signatureKey=u[f],o.push(g))}):o=n;let p={};e.library!=null&&e.library.function!=null&&(p=e.library.function.reduce((f,m)=>(f[m.signature.name]=this.mapFunction(m),f),{}));let c={nodes:i,inputs:o,outputs:l,weights:a,placeholders:n,signature:t,functions:p};return s.length>0&&(c.initNodes=s),c}mapSignatureEntries(e){return Object.keys(e||{}).reduce((t,r)=>(t[e[r].name]=r,t),{})}mapNode(e){let t=yk(e.op)||this.opMappers[e.op]||{};e.attr==null&&(e.attr={});let r={name:e.name,op:e.op,category:t.category,inputNames:(e.input||[]).map(n=>n.startsWith("^")?n.slice(1):n),inputs:[],children:[],inputParams:{},attrParams:{},rawAttrs:e.attr,outputs:t.outputs};return t.inputs!=null&&(r.inputParams=t.inputs.reduce((n,a)=>(n[a.name]={type:a.type,inputIndexStart:a.start,inputIndexEnd:a.end},n),{})),t.attrs!=null&&(r.attrParams=t.attrs.reduce((n,a)=>{let s=a.type,i;switch(a.type){case"string":i=Bg(e.attr,a.tfName,a.defaultValue),i===void 0&&!!a.tfDeprecatedName&&(i=Bg(e.attr,a.tfDeprecatedName,a.defaultValue));break;case"string[]":i=qg(e.attr,a.tfName,a.defaultValue),i===void 0&&!!a.tfDeprecatedName&&(i=qg(e.attr,a.tfDeprecatedName,a.defaultValue));break;case"number":i=Vg(e.attr,a.tfName,a.defaultValue||0),i===void 0&&!!a.tfDeprecatedName&&(i=Vg(e.attr,a.tfDeprecatedName,a.defaultValue));break;case"number[]":i=Hg(e.attr,a.tfName,a.defaultValue),i===void 0&&!!a.tfDeprecatedName&&(i=Hg(e.attr,a.tfDeprecatedName,a.defaultValue));break;case"bool":i=Wg(e.attr,a.tfName,a.defaultValue),i===void 0&&!!a.tfDeprecatedName&&(i=Wg(e.attr,a.tfDeprecatedName,a.defaultValue));break;case"bool[]":i=Kg(e.attr,a.tfName,a.defaultValue),i===void 0&&!!a.tfDeprecatedName&&(i=Kg(e.attr,a.tfDeprecatedName,a.defaultValue));break;case"shape":i=jg(e.attr,a.tfName,a.defaultValue),i===void 0&&!!a.tfDeprecatedName&&(i=jg(e.attr,a.tfDeprecatedName,a.defaultValue));break;case"shape[]":i=Xg(e.attr,a.tfName,a.defaultValue),i===void 0&&!!a.tfDeprecatedName&&(i=Xg(e.attr,a.tfDeprecatedName,a.defaultValue));break;case"dtype":i=Ug(e.attr,a.tfName,a.defaultValue),i===void 0&&!!a.tfDeprecatedName&&(i=Ug(e.attr,a.tfDeprecatedName,a.defaultValue));break;case"dtype[]":i=Gg(e.attr,a.tfName,a.defaultValue),i===void 0&&!!a.tfDeprecatedName&&(i=Gg(e.attr,a.tfDeprecatedName,a.defaultValue));break;case"func":i=C4(e.attr,a.tfName,a.defaultValue),i===void 0&&!!a.tfDeprecatedName&&(i=C4(e.attr,a.tfDeprecatedName,a.defaultValue));break;case"tensor":case"tensors":break;default:throw new Error(`Unsupported param type: ${a.type} for op: ${e.op}`)}return n[a.name]={value:i,type:s},n},{})),r}mapFunction(e){let t=e.nodeDef,r=[],n=[],a={};t!=null&&(a=t.reduce((u,d)=>(u[d.name]=this.mapNode(d),d.op==="Const"&&n.push(u[d.name]),u),{}));let s=[],i=[];e.signature.inputArg.forEach(u=>{let[d]=Ra(u.name),h={name:d,op:"Placeholder",inputs:[],inputNames:[],category:"graph",inputParams:{},attrParams:{dtype:{value:Zx(u.type),type:"dtype"}},children:[]};h.signatureKey=u.name,s.push(h),a[d]=h}),Object.keys(a).forEach(u=>{let d=a[u];d.inputNames.forEach((h,p)=>{let[c,,f]=Ra(h),m=a[c];if(m.outputs!=null){let g=m.outputs.indexOf(f);if(g!==-1){let y=`${c}:${g}`;d.inputNames[p]=y}}d.inputs.push(m),m.children.push(d)})});let o=e.ret;e.signature.outputArg.forEach(u=>{let[d,h]=Ra(o[u.name]),p=a[d];p!=null&&(p.defaultOutput=h,i.push(p))});let l=this.mapArgsToSignature(e);return{nodes:a,inputs:s,outputs:i,weights:n,placeholders:r,signature:l}}mapArgsToSignature(e){return{methodName:e.signature.name,inputs:e.signature.inputArg.reduce((t,r)=>(t[r.name]=this.mapArgToTensorInfo(r),t),{}),outputs:e.signature.outputArg.reduce((t,r)=>(t[r.name]=this.mapArgToTensorInfo(r,e.ret),t),{})}}mapArgToTensorInfo(e,t){let r=e.name;return t!=null&&(r=t[r]),{name:r,dtype:e.type}}};function Hj(e){let t=Z().global;if(typeof t.atob!="undefined")return t.atob(e);if(typeof Buffer!="undefined")return new Buffer(e,"base64").toString();throw new Error("Unable to decode base64 in this environment. Missing built-in atob() or Buffer()")}function Ok(e,t){let r=Array.isArray(e)?String.fromCharCode.apply(null,e):Hj(e);return t?r:r.toLowerCase()}function Bg(e,t,r,n=!1){let a=e[t];return a!=null?Ok(a.s,n):r}function Wg(e,t,r){let n=e[t];return n?n.b:r}function Vg(e,t,r){let n=e[t]||{},a=n.i!=null?n.i:n.f!=null?n.f:r;return typeof a=="number"?a:parseInt(a,10)}function Zx(e){switch(typeof e=="string"&&(e=gk[e]),e){case 1:case 19:return"float32";case 3:case 9:case 6:case 4:return"int32";case 10:return"bool";case 2:return"float32";case 7:return"string";default:return null}}function C4(e,t,r){let n=e[t];return n&&n.func?n.func.name:r}function Ug(e,t,r){let n=e[t];return n&&n.type?Zx(n.type):r}function Gg(e,t,r){let n=e[t];return n&&n.list&&n.list.type?n.list.type.map(a=>Zx(a)):r}function Dk(e){if(!e.unknownRank)return e.dim!=null?e.dim.map(t=>typeof t.size=="number"?t.size:parseInt(t.size,10)):[]}function jg(e,t,r){let n=e[t];return n&&n.shape?Dk(n.shape):r}function Hg(e,t,r){let n=e[t];return n?((n.list.f&&n.list.f.length?n.list.f:n.list.i)||[]).map(a=>typeof a=="number"?a:parseInt(a,10)):r}function qg(e,t,r,n=!1){let a=e[t];return a&&a.list&&a.list.s?a.list.s.map(s=>Ok(s,n)):r}function Xg(e,t,r){let n=e[t];return n&&n.list&&n.list.shape?n.list.shape.map(a=>Dk(a)):r}function Kg(e,t,r){let n=e[t];return n&&n.list&&n.list.b?n.list.b:r}var qj=class{constructor(e,t,r){this.node=e,this.tensorMap=t,this.context=r,this.inputs=[],this.attrs={},this.inputs=e.inputNames.map(n=>this.getInput(n)),e.rawAttrs!=null&&(this.attrs=Object.keys(e.rawAttrs).reduce((n,a)=>(n[a]=this.getAttr(a),n),{}))}getInput(e){return Lr(e,this.tensorMap,this.context)}getAttr(e,t){let r=this.node.rawAttrs[e];if(r.tensor!=null)return Lr(e,this.tensorMap,this.context);if(r.i!=null||r.f!=null)return Vg(this.node.rawAttrs,e,t);if(r.s!=null)return Bg(this.node.rawAttrs,e,t);if(r.b!=null)return Wg(this.node.rawAttrs,e,t);if(r.shape!=null)return jg(this.node.rawAttrs,e,t);if(r.type!=null)return Ug(this.node.rawAttrs,e,t);if(r.list!=null){if(r.list.i!=null||r.list.f!=null)return Hg(this.node.rawAttrs,e,t);if(r.list.s!=null)return qg(this.node.rawAttrs,e,t);if(r.list.shape!=null)return Xg(this.node.rawAttrs,e,t);if(r.list.b!=null)return Kg(this.node.rawAttrs,e,t);if(r.list.type!=null)return Gg(this.node.rawAttrs,e,t)}return t}},Xj=(e,t,r)=>{switch(e.op){case"BiasAdd":case"AddV2":case"Add":return[le(k("a",e,t,r),k("b",e,t,r))];case"AddN":return[Cf(k("tensors",e,t,r))];case"FloorMod":case"Mod":return[md(k("a",e,t,r),k("b",e,t,r))];case"Mul":return[L(k("a",e,t,r),k("b",e,t,r))];case"RealDiv":case"Div":return[pe(k("a",e,t,r),k("b",e,t,r))];case"DivNoNan":return[nw(k("a",e,t,r),k("b",e,t,r))];case"FloorDiv":return[xh(k("a",e,t,r),k("b",e,t,r))];case"Sub":return[ce(k("a",e,t,r),k("b",e,t,r))];case"Minimum":return[Ih(k("a",e,t,r),k("b",e,t,r))];case"Maximum":return[ts(k("a",e,t,r),k("b",e,t,r))];case"Pow":return[Ws(k("a",e,t,r),k("b",e,t,r))];case"SquaredDifference":return[cA(k("a",e,t,r),k("b",e,t,r))];default:throw TypeError(`Node type ${e.op} is not implemented`)}},Kj=(e,t,r)=>{switch(e.op){case"Abs":case"ComplexAbs":return[ar(k("x",e,t,r))];case"Acos":return[F7(k("x",e,t,r))];case"Acosh":return[P7(k("x",e,t,r))];case"Asin":return[z7(k("x",e,t,r))];case"Asinh":return[O7(k("x",e,t,r))];case"Atan":return[D7(k("x",e,t,r))];case"Atan2":return[L7(k("x",e,t,r),k("y",e,t,r))];case"Atanh":return[B7(k("x",e,t,r))];case"Ceil":return[X7(k("x",e,t,r))];case"Complex":return[Os(k("real",e,t,r),k("imag",e,t,r))];case"Cos":return[Ef(k("x",e,t,r))];case"Cosh":return[Hy(k("x",e,t,r))];case"Elu":return[wh(k("x",e,t,r))];case"Erf":return[sw(k("x",e,t,r))];case"Exp":return[Mn(k("x",e,t,r))];case"Expm1":return[iw(k("x",e,t,r))];case"Floor":return[kh(k("x",e,t,r))];case"Log":return[Fn(k("x",e,t,r))];case"Log1p":return[Mf(k("x",e,t,r))];case"Imag":return[Rf(k("x",e,t,r))];case"Neg":return[Dt(k("x",e,t,r))];case"Reciprocal":return[bw(k("x",e,t,r))];case"Real":return[Up(k("x",e,t,r))];case"Relu":return[Oa(k("x",e,t,r))];case"Round":return[iA(k("x",e,t,r))];case"Selu":return[lA(k("x",e,t,r))];case"Sigmoid":return[Tr(k("x",e,t,r))];case"Sin":return[uA(k("x",e,t,r))];case"Sign":return[kw(k("x",e,t,r))];case"Sinh":return[dA(k("x",e,t,r))];case"Softplus":return[fd(k("x",e,t,r))];case"Sqrt":return[Er(k("x",e,t,r))];case"Square":return[At(k("x",e,t,r))];case"Tanh":return[Iu(k("x",e,t,r))];case"Tan":return[Sw(k("x",e,t,r))];case"ClipByValue":return[cn(k("x",e,t,r),k("clipValueMin",e,t,r),k("clipValueMax",e,t,r))];case"Relu6":return[sA(k("x",e,t,r))];case"Rsqrt":return[oA(Lr(e.inputNames[0],t,r))];case"Prod":return[rA(k("x",e,t,r),k("axes",e,t,r))];case"LeakyRelu":return[$f(k("x",e,t,r),k("alpha",e,t,r))];case"Prelu":return[Df(k("x",e,t,r),k("alpha",e,t,r))];case"IsNan":return[ow(Lr(e.inputNames[0],t,r))];default:throw TypeError(`Node type ${e.op} is not implemented`)}};function Un(e,t,r=""){if(!(typeof e=="number"||typeof t=="number")){v.assert(e.length===t.length,()=>r+` Shapes ${e} and ${t} must match`);for(let n=0;n<e.length;n++){let a=e[n],s=t[n];v.assert(a<0||s<0||a===s,()=>r+` Shapes ${e} and ${t} must match`)}}}function T4(e){return!(typeof e=="number"||e.some(t=>t<0))}function xp(e,t,r){let n=Zg(e,r),a=!T4(n);if(a&&t.length===0)throw new Error(`Tried to calculate elements of an empty list with non-fully-defined elementShape: ${n}`);if(a&&t.forEach(s=>{n=Zg(s.shape,n)}),!T4(n))throw new Error(`Non-fully-defined elementShape: ${n}`);return n}function Zg(e,t){if(typeof e=="number")return t;if(typeof t=="number")return e;if(e.length!==t.length)throw new Error(`Incompatible ranks during merge: ${e} vs. ${t}`);let r=[];for(let n=0;n<e.length;++n){let a=e[n],s=t[n];if(a>=0&&s>=0&&a!==s)throw new Error(`Incompatible shape during merge: ${e} vs. ${t}`);r[n]=a>=0?a:s}return r}var Zj=class{constructor(e,t,r,n,a,s,i){this.name=e,this.dtype=t,this.maxSize=r,this.elementShape=n,this.identicalElementShapes=a,this.dynamicSize=s,this.clearAfterRead=i,this.tensors=[],this.closed_=!1,this.idTensor=Se(0),gr(this.idTensor)}get id(){return this.idTensor.id}get closed(){return this.closed_}clearAndClose(e){this.tensors.forEach(t=>{(e==null||!e.has(t.tensor.id))&&t.tensor.dispose()}),this.tensors=[],this.closed_=!0,this.idTensor.dispose()}size(){return this.tensors.length}read(e){if(this.closed_)throw new Error(`TensorArray ${this.name} has already been closed.`);if(e<0||e>=this.size())throw new Error(`Tried to read from index ${e}, but array size is: ${this.size()}`);let t=this.tensors[e];if(t.cleared)throw new Error(`TensorArray ${this.name}: Could not read index ${e} twice because it was cleared after a previous read (perhaps try setting clear_after_read = false?).`);return this.clearAfterRead&&(t.cleared=!0),t.read=!0,t.tensor}readMany(e){return e.map(t=>this.read(t))}write(e,t){if(this.closed_)throw new Error(`TensorArray ${this.name} has already been closed.`);if(e<0||!this.dynamicSize&&e>=this.maxSize)throw new Error(`Tried to write to index ${e}, but array is not resizeable and size is: ${this.maxSize}`);let r=this.tensors[e]||{};if(t.dtype!==this.dtype)throw new Error(`TensorArray ${this.name}: Could not write to TensorArray index ${e},
|
|
because the value dtype is ${t.dtype}, but TensorArray dtype is ${this.dtype}.`);if(this.size()===0&&(this.elementShape==null||this.elementShape.length===0)&&(this.elementShape=t.shape),Un(this.elementShape,t.shape,`TensorArray ${this.name}: Could not write to TensorArray index ${e}.`),r.read)throw new Error(`TensorArray ${this.name}: Could not write to TensorArray index ${e}, because it has already been read.`);if(r.written)throw new Error(`TensorArray ${this.name}: Could not write to TensorArray index ${e}, because it has already been written.`);r.tensor=t,gr(t),r.written=!0,this.tensors[e]=r}writeMany(e,t){if(e.length!==t.length)throw new Error(`TensorArray ${this.name}: could not write multiple tensors,because the index size: ${e.length} is not the same as tensors size: ${t.length}.`);e.forEach((r,n)=>this.write(r,t[n]))}gather(e,t){if(!!t&&t!==this.dtype)throw new Error(`TensorArray dtype is ${this.dtype} but gather requested dtype ${t}`);if(e)e=e.slice(0,this.size());else{e=[];for(let n=0;n<this.size();n++)e.push(n)}if(e.length===0)return ft([],[0].concat(this.elementShape));let r=this.readMany(e);return Un(this.elementShape,r[0].shape,"TensorArray shape mismatch: "),ur(r,0)}concat(e){if(!!e&&e!==this.dtype)throw new Error(`TensorArray dtype is ${this.dtype} but concat requested dtype ${e}`);if(this.size()===0)return ft([],[0].concat(this.elementShape));let t=[];for(let n=0;n<this.size();n++)t.push(n);let r=this.readMany(t);return Un(this.elementShape,r[0].shape,`TensorArray shape mismatch: tensor array shape (${this.elementShape}) vs first tensor shape (${r[0].shape})`),St(r,0)}scatter(e,t){if(t.dtype!==this.dtype)throw new Error(`TensorArray dtype is ${this.dtype} but tensor has dtype ${t.dtype}`);if(e.length!==t.shape[0])throw new Error(`Expected len(indices) == tensor.shape[0], but saw: ${e.length} vs. ${t.shape[0]}`);let r=Math.max(...e);if(!this.dynamicSize&&r>=this.maxSize)throw new Error(`Max index must be < array size (${r} vs. ${this.maxSize})`);this.writeMany(e,nn(t,0))}split(e,t){if(t.dtype!==this.dtype)throw new Error(`TensorArray dtype is ${this.dtype} but tensor has dtype ${t.dtype}`);let r=0,n=e.map(o=>(r+=o,r));if(r!==t.shape[0])throw new Error(`Expected sum of lengths to be equal to
|
|
tensor.shape[0], but sum of lengths is
|
|
${r}, and tensor's shape is: ${t.shape}`);if(!this.dynamicSize&&e.length!==this.maxSize)throw new Error(`TensorArray's size is not equal to the size of lengths (${this.maxSize} vs. ${e.length}), and the TensorArray is not marked as dynamically resizeable`);let a=r===0?0:t.size/r,s=[];X(()=>{t=U(t,[1,r,a]);for(let o=0;o<e.length;++o){let l=o===0?0:n[o-1],u=[0,l,0],d=[1,e[o],a];s[o]=U(Pe(t,u,d),this.elementShape)}return s});let i=[];for(let o=0;o<e.length;o++)i[o]=o;this.writeMany(i,s)}},$u=class{constructor(e,t,r,n=-1){this.tensors=e,this.elementShape=t,this.elementDtype=r,e!=null&&e.forEach(a=>{if(r!==a.dtype)throw new Error(`Invalid data types; op elements ${r}, but list elements ${a.dtype}`);Un(t,a.shape,"TensorList shape mismatch: "),gr(a)}),this.idTensor=Se(0),this.maxNumElements=n,gr(this.idTensor)}get id(){return this.idTensor.id}copy(){return new $u([...this.tensors],this.elementShape,this.elementDtype)}clearAndClose(e){this.tensors.forEach(t=>{(e==null||!e.has(t.id))&&t.dispose()}),this.tensors.length=0,this.idTensor.dispose()}size(){return this.tensors.length}stack(e,t,r=-1){if(t!==this.elementDtype)throw new Error(`Invalid data types; op elements ${t}, but list elements ${this.elementDtype}`);if(r!==-1&&this.tensors.length!==r)throw new Error(`Operation expected a list with ${r} elements but got a list with ${this.tensors.length} elements.`);Un(e,this.elementShape,"TensorList shape mismatch: ");let n=xp(this.elementShape,this.tensors,e);return X(()=>{let a=this.tensors.map(s=>U(s,n));return ur(a,0)})}popBack(e,t){if(t!==this.elementDtype)throw new Error(`Invalid data types; op elements ${t}, but list elements ${this.elementDtype}`);if(this.size()===0)throw new Error("Trying to pop from an empty list.");let r=xp(this.elementShape,this.tensors,e),n=this.tensors.pop();return Un(n.shape,e,"TensorList shape mismatch: "),U(n,r)}pushBack(e){if(e.dtype!==this.elementDtype)throw new Error(`Invalid data types; op elements ${e.dtype}, but list elements ${this.elementDtype}`);if(Un(e.shape,this.elementShape,"TensorList shape mismatch: "),this.maxNumElements===this.size())throw new Error("Trying to push element into a full list.");gr(e),this.tensors.push(e)}resize(e){if(e<0)throw new Error(`TensorListResize expects size to be non-negative. Got: ${e}`);if(this.maxNumElements!==-1&&e>this.maxNumElements)throw new Error(`TensorListResize input size ${e} is greater maxNumElement ${this.maxNumElements}.`);let t=new $u([],this.elementShape,this.elementDtype,this.maxNumElements);t.tensors.length=e;for(let r=0;r<Math.min(this.tensors.length,e);++r)t.tensors[r]=this.tensors[r];return t}getItem(e,t,r){if(r!==this.elementDtype)throw new Error(`Invalid data types; op elements ${r}, but list elements ${this.elementDtype}`);if(e<0||e>this.tensors.length)throw new Error(`Trying to access element ${e} in a list with ${this.tensors.length} elements.`);if(this.tensors[e]==null)throw new Error(`element at index ${e} is null.`);Un(this.tensors[e].shape,t,"TensorList shape mismatch: ");let n=xp(this.elementShape,this.tensors,t);return U(this.tensors[e],n)}setItem(e,t){if(t.dtype!==this.elementDtype)throw new Error(`Invalid data types; op elements ${t.dtype}, but list elements ${this.elementDtype}`);if(e<0||this.maxNumElements!==-1&&e>=this.maxNumElements)throw new Error(`Trying to set element ${e} in a list with max ${this.maxNumElements} elements.`);Un(this.elementShape,t.shape,"TensorList shape mismatch: "),gr(t),this.tensors[e]=t}gather(e,t,r){if(t!==this.elementDtype)throw new Error(`Invalid data types; op elements ${t}, but list elements ${this.elementDtype}`);Un(this.elementShape,r,"TensorList shape mismatch: "),e=e.slice(0,this.size());let n=xp(this.elementShape,this.tensors,r);return e.length===0?ft([],[0].concat(n)):X(()=>{let a=e.map(s=>U(this.tensors[s],n));return ur(a,0)})}concat(e,t){if(!!e&&e!==this.elementDtype)throw new Error(`TensorList dtype is ${this.elementDtype} but concat requested dtype ${e}`);Un(this.elementShape,t,"TensorList shape mismatch: ");let r=xp(this.elementShape,this.tensors,t);return this.size()===0?ft([],[0].concat(r)):X(()=>{let n=this.tensors.map(a=>U(a,r));return St(n,0)})}};function Yj(e,t,r){let n=e.dtype;if(e.shape.length<1)throw new Error(`Tensor must be at least a vector, but saw shape: ${e.shape}`);if(e.dtype!==r)throw new Error(`Invalid data types; op elements ${e.dtype}, but list elements ${r}`);let a=e.shape.slice(1);Un(a,t,"TensorList shape mismatch: ");let s=nn(e);return new $u(s,t,n)}function Jj(e,t,r){return new $u([],e,t,r)}function Qj(e,t,r,n){if(t.length!==e.shape[0])throw new Error(`Expected len(indices) == tensor.shape[0], but saw: ${t.length} vs. ${e.shape[0]}`);let a=Math.max(...t);if(n!=null&&n!==-1&&a>=n)throw new Error(`Max index must be < array size (${a} vs. ${n})`);let s=new $u([],r,e.dtype,n),i=nn(e,0);return t.forEach((o,l)=>{s.setItem(o,i[l])}),s}function eH(e,t,r){let n=0,a=t.map(d=>(n+=d,n));if(n!==e.shape[0])throw new Error(`Expected sum of lengths to be equal to
|
|
tensor.shape[0], but sum of lengths is
|
|
${n}, and tensor's shape is: ${e.shape}`);let s=e.shape.slice(1),i=Zg(s,r),o=n===0?0:e.size/n,l=X(()=>{let d=[];e=U(e,[1,n,o]);for(let h=0;h<t.length;++h){let p=h===0?0:a[h-1],c=[0,p,0],f=[1,t[h],o];d[h]=U(Pe(e,c,f),i)}return e.dispose(),d}),u=new $u([],r,e.dtype,t.length);for(let d=0;d<l.length;d++)u.setItem(d,l[d]);return u}var tH=async(e,t,r)=>{switch(e.op){case"If":case"StatelessIf":{let n=k("thenBranch",e,t,r),a=k("elseBranch",e,t,r),s=k("cond",e,t,r),i=k("args",e,t,r);return(await s.data())[0]?r.functionMap[n].executeFunctionAsync(i,r.tensorArrayMap,r.tensorListMap):r.functionMap[a].executeFunctionAsync(i,r.tensorArrayMap,r.tensorListMap)}case"While":case"StatelessWhile":{let n=k("body",e,t,r),a=k("cond",e,t,r),s=k("args",e,t,r),i=await r.functionMap[a].executeFunctionAsync(s,r.tensorArrayMap,r.tensorListMap),o=s.map(d=>d.id),l=await i[0].data();i.forEach(d=>{!d.kept&&o.indexOf(d.id)===-1&&d.dispose()});let u=s;for(;l[0];){let d=u;u=await r.functionMap[n].executeFunctionAsync(u,r.tensorArrayMap,r.tensorListMap);let h=u.map(c=>c.id);d.forEach(c=>{!c.kept&&o.indexOf(c.id)===-1&&h.indexOf(c.id)===-1&&c.dispose()});let p=await r.functionMap[a].executeFunctionAsync(u,r.tensorArrayMap,r.tensorListMap);l=await p[0].data(),p.forEach(c=>{!c.kept&&o.indexOf(c.id)===-1&&h.indexOf(c.id)===-1&&c.dispose()})}return u}case"LoopCond":{let n=k("pred",e,t,r);return[Ha(n)]}case"Switch":{let n=k("pred",e,t,r),a=k("data",e,t,r);return a.kept||(a=Ha(a)),(await n.data())[0]?[void 0,a]:[a,void 0]}case"Merge":{let n=e.inputNames.find(a=>Lr(a,t,r)!==void 0);if(n){let a=Lr(n,t,r);return[Ha(a)]}return}case"Enter":{let n=k("frameName",e,t,r),a=k("tensor",e,t,r);return r.enterFrame(n),[Ha(a)]}case"Exit":{let n=k("tensor",e,t,r);return r.exitFrame(),[Ha(n)]}case"NextIteration":{let n=k("tensor",e,t,r);return r.nextIteration(),[Ha(n)]}case"TensorArrayV3":{let n=k("size",e,t,r),a=k("dtype",e,t,r),s=k("elementShape",e,t,r),i=k("dynamicSize",e,t,r),o=k("clearAfterRead",e,t,r),l=k("identicalElementShapes",e,t,r),u=k("name",e,t,r),d=new Zj(u,a,n,s,l,i,o);return r.addTensorArray(d),[d.idTensor,Se(1)]}case"TensorArrayWriteV3":{let n=k("tensorArrayId",e,t,r),a=k("index",e,t,r),s=k("tensor",e,t,r),i=r.getTensorArray(n.id);return i.write(a,s),[i.idTensor]}case"TensorArrayReadV3":{let n=k("tensorArrayId",e,t,r),a=k("index",e,t,r);return[r.getTensorArray(n.id).read(a)]}case"TensorArrayGatherV3":{let n=k("tensorArrayId",e,t,r),a=k("indices",e,t,r),s=k("dtype",e,t,r);return[r.getTensorArray(n.id).gather(a,s)]}case"TensorArrayScatterV3":{let n=k("tensorArrayId",e,t,r),a=k("indices",e,t,r),s=k("tensor",e,t,r),i=r.getTensorArray(n.id);return i.scatter(a,s),[i.idTensor]}case"TensorArrayConcatV3":{let n=k("tensorArrayId",e,t,r),a=r.getTensorArray(n.id),s=k("dtype",e,t,r);return[a.concat(s)]}case"TensorArraySplitV3":{let n=k("tensorArrayId",e,t,r),a=k("tensor",e,t,r),s=k("lengths",e,t,r),i=r.getTensorArray(n.id);return i.split(s,a),[i.idTensor]}case"TensorArraySizeV3":{let n=k("tensorArrayId",e,t,r),a=r.getTensorArray(n.id);return[Se(a.size(),"int32")]}case"TensorArrayCloseV3":{let n=k("tensorArrayId",e,t,r),a=r.getTensorArray(n.id);return a.clearAndClose(),[a.idTensor]}case"TensorListSetItem":{let n=k("tensorListId",e,t,r),a=k("index",e,t,r),s=k("tensor",e,t,r),i=r.getTensorList(n.id);return i.setItem(a,s),[i.idTensor]}case"TensorListGetItem":{let n=k("tensorListId",e,t,r),a=k("index",e,t,r),s=k("elementShape",e,t,r),i=k("elementDType",e,t,r);return[r.getTensorList(n.id).getItem(a,s,i)]}case"TensorListScatterV2":case"TensorListScatter":{let n=k("indices",e,t,r),a=k("tensor",e,t,r),s=k("elementShape",e,t,r),i=k("numElements",e,t,r),o=Qj(a,n,s,i);return r.addTensorList(o),[o.idTensor]}case"TensorListReserve":case"EmptyTensorList":{let n=k("elementShape",e,t,r),a=k("elementDType",e,t,r),s;e.op==="TensorListReserve"?s="numElements":s="maxNumElements";let i=k(s,e,t,r),o=Jj(n,a,i);return r.addTensorList(o),[o.idTensor]}case"TensorListGather":{let n=k("tensorListId",e,t,r),a=k("indices",e,t,r),s=k("elementShape",e,t,r),i=k("elementDType",e,t,r);return[r.getTensorList(n.id).gather(a,i,s)]}case"TensorListStack":{let n=k("tensorListId",e,t,r),a=k("elementShape",e,t,r),s=k("elementDType",e,t,r),i=k("numElements",e,t,r);return[r.getTensorList(n.id).stack(a,s,i)]}case"TensorListFromTensor":{let n=k("tensor",e,t,r),a=k("elementShape",e,t,r),s=k("elementDType",e,t,r),i=Yj(n,a,s);return r.addTensorList(i),[i.idTensor]}case"TensorListConcat":case"TensorListConcatV2":{let n=k("tensorListId",e,t,r),a=r.getTensorList(n.id),s=k("dtype",e,t,r),i=k("elementShape",e,t,r);return[a.concat(s,i)]}case"TensorListPushBack":{let n=k("tensorListId",e,t,r),a=k("tensor",e,t,r),s=r.getTensorList(n.id);return s.pushBack(a),[s.idTensor]}case"TensorListPopBack":{let n=k("tensorListId",e,t,r),a=k("elementShape",e,t,r),s=k("elementDType",e,t,r);return[r.getTensorList(n.id).popBack(a,s)]}case"TensorListSplit":{let n=k("tensor",e,t,r),a=k("elementShape",e,t,r),s=k("lengths",e,t,r),i=eH(n,s,a);return r.addTensorList(i),[i.idTensor]}case"TensorListLength":{let n=k("tensorListId",e,t,r),a=r.getTensorList(n.id);return[Se(a.size(),"int32")]}case"TensorListResize":{let n=k("tensorListId",e,t,r),a=k("size",e,t,r),s=r.getTensorList(n.id).resize(a);return r.addTensorList(s),[s.idTensor]}default:throw TypeError(`Node type ${e.op} is not implemented`)}};function N4(e,t,r){let[n,a]=k("fusedOps",e,t,r),s=n==="biasadd",i=!s,o=a==="prelu",l=n==="fusedbatchnorm",u=k("numArgs",e,t,r);if(s){if(o&&u!==2)throw new Error("FusedConv2d and DepthwiseConv2d with BiasAdd and Prelu must have two extra arguments: bias and alpha.");if(!o&&s&&u!==1)throw new Error("FusedConv2d and DepthwiseConv2d with BiasAdd must have one extra argument: bias.")}if(l)throw new Error("FusedConv2d and DepthwiseConv2d with FusedBatchNorm is not supported");let d=k("strides",e,t,r),h=o0(e,t,r),p=k("dataFormat",e,t,r).toUpperCase(),c=k("dilations",e,t,r),[f,m]=k("args",e,t,r);i&&(m=f,f=void 0);let g=k("leakyreluAlpha",e,t,r);return{stride:d,pad:h,dataFormat:p,dilations:c,biasArg:f,preluArg:m,activationFunc:a,leakyreluAlpha:g}}var rH=(e,t,r)=>{switch(e.op){case"Conv1D":{let n=k("stride",e,t,r),a=k("pad",e,t,r),s=k("dataFormat",e,t,r).toUpperCase(),i=k("dilation",e,t,r);return[Vy(k("x",e,t,r),k("filter",e,t,r),n,a,s,i)]}case"Conv2D":{let n=k("strides",e,t,r),a=o0(e,t,r),s=k("dataFormat",e,t,r).toUpperCase(),i=k("dilations",e,t,r);return[Ls(k("x",e,t,r),k("filter",e,t,r),[n[1],n[2]],a,s,[i[1],i[2]])]}case"_FusedConv2D":{let{stride:n,pad:a,dataFormat:s,dilations:i,biasArg:o,preluArg:l,activationFunc:u,leakyreluAlpha:d}=N4(e,t,r);return[Vs.conv2d({x:k("x",e,t,r),filter:k("filter",e,t,r),strides:[n[1],n[2]],pad:a,dataFormat:s,dilations:[i[1],i[2]],bias:o,activation:u,preluActivationWeights:l,leakyreluAlpha:d})]}case"FusedDepthwiseConv2dNative":{let{stride:n,pad:a,dataFormat:s,dilations:i,biasArg:o,preluArg:l,activationFunc:u,leakyreluAlpha:d}=N4(e,t,r);return[Vs.depthwiseConv2d({x:k("x",e,t,r),filter:k("filter",e,t,r),strides:[n[1],n[2]],pad:a,dataFormat:s,dilations:[i[1],i[2]],bias:o,activation:u,preluActivationWeights:l,leakyreluAlpha:d})]}case"Conv2DBackpropInput":case"Conv2dTranspose":{let n=k("outputShape",e,t,r),a=k("strides",e,t,r),s=o0(e,t,r);return[Gy(k("x",e,t,r),k("filter",e,t,r),n,[a[1],a[2]],s)]}case"DepthwiseConv2dNative":case"DepthwiseConv2d":{let n=k("strides",e,t,r),a=o0(e,t,r),s=k("dilations",e,t,r),i=k("dataFormat",e,t,r).toUpperCase();return[vh(k("input",e,t,r),k("filter",e,t,r),[n[1],n[2]],a,i,[s[1],s[2]])]}case"Conv3D":{let n=k("strides",e,t,r),a=k("pad",e,t,r),s=k("dataFormat",e,t,r).toUpperCase(),i=k("dilations",e,t,r);return[jy(k("x",e,t,r),k("filter",e,t,r),[n[1],n[2],n[3]],a,s,[i[1],i[2],i[3]])]}case"AvgPool":{let n=k("strides",e,t,r),a=k("pad",e,t,r),s=k("kernelSize",e,t,r);return[Tf(k("x",e,t,r),[s[1],s[2]],[n[1],n[2]],a)]}case"MaxPool":{let n=k("strides",e,t,r),a=k("pad",e,t,r),s=k("kernelSize",e,t,r);return[_f(k("x",e,t,r),[s[1],s[2]],[n[1],n[2]],a)]}case"MaxPoolWithArgmax":{let n=k("strides",e,t,r),a=k("pad",e,t,r),s=k("kernelSize",e,t,r),i=k("includeBatchInIndex",e,t,r),{result:o,indexes:l}=gw(k("x",e,t,r),[s[1],s[2]],[n[1],n[2]],a,i);return[o,l]}case"AvgPool3D":{let n=k("strides",e,t,r),a=k("pad",e,t,r),s=k("kernelSize",e,t,r);return[By(k("x",e,t,r),[s[1],s[2],s[3]],[n[1],n[2],n[3]],a)]}case"MaxPool3D":{let n=k("strides",e,t,r),a=k("pad",e,t,r),s=k("kernelSize",e,t,r);return[tA(k("x",e,t,r),[s[1],s[2],s[3]],[n[1],n[2],n[3]],a)]}case"Dilation2D":{let n=k("strides",e,t,r),a=k("pad",e,t,r),s=k("dilations",e,t,r),i=n[1],o=n[2],l=s[1],u=s[2];return[rw(k("x",e,t,r),k("filter",e,t,r),[i,o],a,[l,u],"NHWC")]}default:throw TypeError(`Node type ${e.op} is not implemented`)}},nH=(e,t,r)=>{switch(e.op){case"Fill":{let n=k("shape",e,t,r),a=k("dtype",e,t,r),s=k("value",e,t,r);return[cd(n,s,a)]}case"LinSpace":{let n=k("start",e,t,r),a=k("stop",e,t,r),s=k("num",e,t,r);return[lw(n,a,s)]}case"Multinomial":{let n=k("logits",e,t,r),a=k("numSamples",e,t,r),s=k("seed",e,t,r);return[Aw(n,a,s)]}case"OneHot":{let n=k("indices",e,t,r),a=k("depth",e,t,r),s=k("onValue",e,t,r),i=k("offValue",e,t,r);return[Vp(n,a,s,i)]}case"Ones":return[hn(k("shape",e,t,r),k("dtype",e,t,r))];case"OnesLike":return[Pn(k("x",e,t,r))];case"RandomUniform":return[gd(k("shape",e,t,r),k("minval",e,t,r),k("maxval",e,t,r),k("dtype",e,t,r))];case"Range":{let n=k("start",e,t,r),a=k("stop",e,t,r),s=k("step",e,t,r);return[Nu(n,a,s,k("dtype",e,t,r))]}case"TruncatedNormal":{let n=k("shape",e,t,r),a=k("mean",e,t,r),s=k("stdDev",e,t,r),i=k("seed",e,t,r);return[Vf(n,a,s,k("dtype",e,t,r),i)]}case"Zeros":return[_t(k("shape",e,t,r),k("dtype",e,t,r))];case"ZerosLike":return[at(k("x",e,t,r))];default:throw TypeError(`Node type ${e.op} is not implemented`)}};function dg(e,t,r){let n=k("boxes",e,t,r),a=k("scores",e,t,r),s=k("maxOutputSize",e,t,r),i=k("iouThreshold",e,t,r),o=k("scoreThreshold",e,t,r),l=k("softNmsSigma",e,t,r);return{boxes:n,scores:a,maxOutputSize:s,iouThreshold:i,scoreThreshold:o,softNmsSigma:l}}var aH=async(e,t,r)=>{switch(e.op){case"NonMaxSuppressionV5":{let{boxes:n,scores:a,maxOutputSize:s,iouThreshold:i,scoreThreshold:o,softNmsSigma:l}=dg(e,t,r),u=await Ie.nonMaxSuppressionWithScoreAsync(n,a,s,i,o,l);return[u.selectedIndices,u.selectedScores]}case"NonMaxSuppressionV4":{let{boxes:n,scores:a,maxOutputSize:s,iouThreshold:i,scoreThreshold:o}=dg(e,t,r),l=k("padToMaxOutputSize",e,t,r),u=await Ie.nonMaxSuppressionPaddedAsync(n,a,s,i,o,l);return[u.selectedIndices,u.validOutputs]}case"NonMaxSuppressionV3":case"NonMaxSuppressionV2":{let{boxes:n,scores:a,maxOutputSize:s,iouThreshold:i,scoreThreshold:o}=dg(e,t,r);return[await Ie.nonMaxSuppressionAsync(n,a,s,i,o)]}case"Where":{let n=me(k("condition",e,t,r),"bool"),a=[await fA(n)];return n.dispose(),a}case"ListDiff":return ww(k("x",e,t,r),k("y",e,t,r));default:throw TypeError(`Node type ${e.op} is not implemented`)}},sH=(e,t,r)=>{switch(e.op){case"LowerBound":{let n=k("sortedSequence",e,t,r),a=k("values",e,t,r);return[mw(n,a)]}case"TopKV2":{let n=k("x",e,t,r),a=k("k",e,t,r),s=k("sorted",e,t,r),i=Cw(n,a,s);return[i.values,i.indices]}case"UpperBound":{let n=k("sortedSequence",e,t,r),a=k("values",e,t,r);return[Nw(n,a)]}case"Unique":{let n=k("x",e,t,r),a=Ng(n);return[a.values,a.indices]}case"UniqueV2":{let n=k("x",e,t,r),a=k("axis",e,t,r),s=Ng(n,a);return[s.values,s.indices]}default:throw TypeError(`Node type ${e.op} is not implemented`)}},iH=(e,t,r)=>{switch(e.op){case"Const":return t[e.name];case"PlaceholderWithDefault":let n=k("default",e,t,r);return[Lr(e.name,t,r)||n];case"Placeholder":return[Lr(e.name,t,r)];case"Identity":case"StopGradient":case"FakeQuantWithMinMaxVars":{let u=k("x",e,t,r);return[Ha(u)]}case"IdentityN":return k("x",e,t,r).map(u=>Ha(u));case"Snapshot":let a=k("x",e,t,r);return[Ha(a)];case"Shape":return[Tt(k("x",e,t,r).shape,"int32")];case"ShapeN":return k("x",e,t,r).map(u=>Tt(u.shape));case"Size":return[Se(k("x",e,t,r).size,"int32")];case"Rank":return[Se(k("x",e,t,r).rank,"int32")];case"NoOp":return[Se(1)];case"Print":let s=k("x",e,t,r),i=k("data",e,t,r),o=k("message",e,t,r),l=k("summarize",e,t,r);console.warn("The graph has a tf.print() operation,usually used for debugging, which slows down performance."),console.log(o);for(let u=0;u<i.length;u++)console.log(Array.prototype.slice.call(i[u].dataSync()).slice(0,l));return[s];default:throw TypeError(`Node type ${e.op} is not implemented`)}},oH=class{constructor(e,t){this.keyDType=e,this.valueDType=t,this.handle=Se(0),this.tensorMap=new Map,gr(this.handle)}get id(){return this.handle.id}clearAndClose(){this.tensorMap.forEach(e=>e.dispose()),this.tensorMap.clear(),this.handle.dispose()}size(){return this.tensorMap.size}tensorSize(){return Se(this.size(),"int32")}async import(e,t){this.checkKeyAndValueTensor(e,t);let r=await e.data();return this.tensorMap.forEach(n=>n.dispose()),this.tensorMap.clear(),X(()=>{let n=nn(t),a=r.length,s=n.length;v.assert(a===s,()=>`The number of elements doesn't match, keys has ${a} elements, the values has ${s} elements.`);for(let i=0;i<a;i++){let o=r[i],l=n[i];gr(l),this.tensorMap.set(o,l)}return this.handle})}async find(e,t){this.checkKeyAndValueTensor(e,t);let r=await e.data();return X(()=>{let n=[];for(let a=0;a<r.length;a++){let s=r[a],i=this.findWithDefault(s,t);n.push(i)}return ur(n)})}findWithDefault(e,t){let r=this.tensorMap.get(e);return r!=null?r:t}checkKeyAndValueTensor(e,t){if(e.dtype!==this.keyDType)throw new Error(`Expect key dtype ${this.keyDType}, but got ${e.dtype}`);if(t.dtype!==this.valueDType)throw new Error(`Expect value dtype ${this.valueDType}, but got ${t.dtype}`)}},lH=async(e,t,r,n)=>{switch(e.op){case"HashTable":case"HashTableV2":{let a=k("keyDType",e,t,r),s=k("valueDType",e,t,r),i=new oH(a,s);return n.addHashTable(e.name,i),[i.handle]}case"LookupTableImport":case"LookupTableImportV2":{let a=k("tableHandle",e,t,r,n),s=k("keys",e,t,r),i=k("values",e,t,r);return[await n.getHashTableById(a.id).import(s,i)]}case"LookupTableFind":case"LookupTableFindV2":{let a=k("tableHandle",e,t,r,n),s=k("keys",e,t,r),i=k("defaultValue",e,t,r);return[await n.getHashTableById(a.id).find(s,i)]}case"LookupTableSize":case"LookupTableSizeV2":{let a=k("tableHandle",e,t,r,n);return[n.getHashTableById(a.id).tensorSize()]}default:throw TypeError(`Node type ${e.op} is not implemented`)}},uH=(e,t,r)=>{switch(e.op){case"ResizeBilinear":{let n=k("images",e,t,r),a=k("size",e,t,r),s=k("alignCorners",e,t,r),i=k("halfPixelCenters",e,t,r);return[Ie.resizeBilinear(n,[a[0],a[1]],s,i)]}case"ResizeNearestNeighbor":{let n=k("images",e,t,r),a=k("size",e,t,r),s=k("alignCorners",e,t,r),i=k("halfPixelCenters",e,t,r);return[Ie.resizeNearestNeighbor(n,[a[0],a[1]],s,i)]}case"CropAndResize":{let n=k("image",e,t,r),a=k("boxes",e,t,r),s=k("boxInd",e,t,r),i=k("cropSize",e,t,r),o=k("method",e,t,r),l=k("extrapolationValue",e,t,r);return[Ie.cropAndResize(n,a,s,i,o,l)]}case"ImageProjectiveTransformV3":{let n=k("images",e,t,r),a=k("transforms",e,t,r),s=k("outputShape",e,t,r),i=k("fillValue",e,t,r),o=k("interpolation",e,t,r),l=k("fillMode",e,t,r);return[Ie.transform(n,a,o.toLowerCase(),l.toLowerCase(),i,s)]}default:throw TypeError(`Node type ${e.op} is not implemented`)}},dH=(e,t,r)=>{switch(e.op){case"Equal":return[$n(k("a",e,t,r),k("b",e,t,r))];case"NotEqual":return[Tu(k("a",e,t,r),k("b",e,t,r))];case"Greater":return[mn(k("a",e,t,r),k("b",e,t,r))];case"GreaterEqual":return[Ml(k("a",e,t,r),k("b",e,t,r))];case"Less":return[Ky(k("a",e,t,r),k("b",e,t,r))];case"LessEqual":return[Fl(k("a",e,t,r),k("b",e,t,r))];case"LogicalAnd":return[ga(k("a",e,t,r),k("b",e,t,r))];case"LogicalNot":return[Pf(k("a",e,t,r))];case"LogicalOr":return[Qy(k("a",e,t,r),k("b",e,t,r))];case"Select":case"SelectV2":return[Vr(k("condition",e,t,r),k("a",e,t,r),k("b",e,t,r))];default:throw TypeError(`Node type ${e.op} is not implemented`)}},pH=(e,t,r)=>{switch(e.op){case"BatchMatMul":case"BatchMatMulV2":case"MatMul":return[Ze(k("a",e,t,r),k("b",e,t,r),k("transposeA",e,t,r),k("transposeB",e,t,r))];case"Einsum":return[aw(k("equation",e,t,r),...k("tensors",e,t,r))];case"Transpose":return[et(k("x",e,t,r),k("perm",e,t,r))];case"_FusedMatMul":let[n,a]=k("fusedOps",e,t,r),s=n==="biasadd",i=a==="prelu",o=k("numArgs",e,t,r),l=k("leakyreluAlpha",e,t,r);if(s){if(i&&o!==2)throw new Error("Fused MatMul with BiasAdd and Prelu must have two extra arguments: bias and alpha.");if(!i&&o!==1)throw new Error("Fused MatMul with BiasAdd must have one extra argument: bias.")}let[u,d]=k("args",e,t,r);return[Vs.matMul({a:k("a",e,t,r),b:k("b",e,t,r),transposeA:k("transposeA",e,t,r),transposeB:k("transposeB",e,t,r),bias:u,activation:a,preluActivationWeights:d,leakyreluAlpha:l})];default:throw TypeError(`Node type ${e.op} is not implemented`)}},hH=(e,t,r)=>{switch(e.op){case"FusedBatchNorm":case"FusedBatchNormV2":return[Su(k("x",e,t,r),k("mean",e,t,r),k("variance",e,t,r),k("offset",e,t,r),k("scale",e,t,r),k("epsilon",e,t,r))];case"FusedBatchNormV3":return[Su(k("x",e,t,r),k("mean",e,t,r),k("variance",e,t,r),k("offset",e,t,r),k("scale",e,t,r),k("epsilon",e,t,r))];case"LRN":return[uw(k("x",e,t,r),k("radius",e,t,r),k("bias",e,t,r),k("alpha",e,t,r),k("beta",e,t,r))];case"Softmax":return[yd(k("x",e,t,r))];case"LogSoftmax":return[Zy(k("x",e,t,r))];case"SparseToDense":return[gA(k("sparseIndices",e,t,r),k("outputShape",e,t,r),k("sparseValues",e,t,r),k("defaultValue",e,t,r))];default:throw TypeError(`Node type ${e.op} is not implemented`)}},cH=(e,t,r)=>{switch(e.op){case"Max":{let i=k("axis",e,t,r),o=k("keepDims",e,t,r);return[Ar(k("x",e,t,r),i,o)]}case"Mean":{let i=k("axis",e,t,r),o=k("keepDims",e,t,r);return[Vt(k("x",e,t,r),i,o)]}case"Min":{let i=k("axis",e,t,r),o=k("keepDims",e,t,r);return[Bs(k("x",e,t,r),i,o)]}case"Sum":{let i=k("axis",e,t,r),o=k("keepDims",e,t,r);return[ke(k("x",e,t,r),i,o)]}case"All":{let i=k("axis",e,t,r),o=k("keepDims",e,t,r);return[Dy(k("x",e,t,r),i,o)]}case"Any":{let i=k("axis",e,t,r),o=k("keepDims",e,t,r);return[C0(k("x",e,t,r),i,o)]}case"ArgMax":{let i=k("axis",e,t,r);return[Rn(k("x",e,t,r),i)]}case"ArgMin":{let i=k("axis",e,t,r);return[_7(k("x",e,t,r),i)]}case"Prod":{let i=k("axis",e,t,r),o=k("keepDims",e,t,r);return[rA(k("x",e,t,r),i,o)]}case"Cumprod":{let i=k("axis",e,t,r),o=k("exclusive",e,t,r),l=k("reverse",e,t,r);return[N0(k("x",e,t,r),i,o,l)]}case"Cumsum":{let i=k("axis",e,t,r),o=k("exclusive",e,t,r),l=k("reverse",e,t,r);return[qy(k("x",e,t,r),i,o,l)]}case"Bincount":let n=k("x",e,t,r),a=k("weights",e,t,r),s=k("size",e,t,r);return[Wy(n,a,s)];case"DenseBincount":{let i=k("x",e,t,r),o=k("weights",e,t,r),l=k("size",e,t,r),u=k("binaryOutput",e,t,r);return[ew(i,o,l,u)]}default:throw TypeError(`Node type ${e.op} is not implemented`)}},fH=(e,t,r)=>{switch(e.op){case"ConcatV2":case"Concat":{let n=k("n",e,t,r),a=k("axis",e,t,r),s=k("tensors",e,t,r);return s=s.slice(0,n),[St(s,a)]}case"Gather":{let n=k("x",e,t,r),a=k("indices",e,t,r);return[Cu(n,me(a,"int32"),0)]}case"GatherV2":{let n=k("axis",e,t,r),a=k("batchDims",e,t,r),s=k("x",e,t,r),i=k("indices",e,t,r);return[Cu(s,me(i,"int32"),n,a)]}case"Reverse":{let n=k("dims",e,t,r),a=[];for(let i=0;i<n.length;i++)n[i]&&a.push(i);let s=k("x",e,t,r);return[_n(s,a)]}case"ReverseV2":{let n=k("axis",e,t,r),a=k("x",e,t,r);return[_n(a,n)]}case"Slice":{let n=k("begin",e,t,r),a=k("size",e,t,r);return[Pe(k("x",e,t,r),n,a)]}case"StridedSlice":{let n=k("begin",e,t,r),a=k("end",e,t,r),s=k("strides",e,t,r),i=k("beginMask",e,t,r),o=k("endMask",e,t,r),l=k("ellipsisMask",e,t,r),u=k("newAxisMask",e,t,r),d=k("shrinkAxisMask",e,t,r),h=k("x",e,t,r);return[Iw(h,n,a,s,i,o,l,u,d)]}case"Pack":return X(()=>{let n=k("axis",e,t,r),a=k("tensors",e,t,r),s=a[0].shape,i=tt(a[0]).shape,o=a.map(l=>{let u=v.arraysEqual(l.shape,s);if(!u&&!v.arraysEqual(tt(l).shape,i))throw new Error("the input tensors shape does not match");return u?l:U(l,s)});return[ur(o,n)]});case"Unpack":{let n=k("axis",e,t,r),a=k("tensor",e,t,r);return nn(a,n)}case"Tile":{let n=k("reps",e,t,r);return[Gn(k("x",e,t,r),n)]}case"Split":case"SplitV":{let n=k("axis",e,t,r),a=k("numOrSizeSplits",e,t,r),s=k("x",e,t,r);return Yt(s,a,n)}case"ScatterNd":{let n=k("indices",e,t,r),a=k("values",e,t,r),s=k("shape",e,t,r);return[Mw(n,a,s)]}case"GatherNd":{let n=k("x",e,t,r),a=k("indices",e,t,r);return[Fw(n,a)]}case"SparseToDense":{let n=k("sparseIndices",e,t,r),a=k("outputShape",e,t,r),s=k("sparseValues",e,t,r),i=k("defaultValue",e,t,r);return[gA(n,s,a,s.dtype===i.dtype?i:me(i,s.dtype))]}default:throw TypeError(`Node type ${e.op} is not implemented`)}},mH=(e,t,r)=>{switch(e.op){case"SparseFillEmptyRows":{let{outputIndices:n,outputValues:a,emptyRowIndicator:s,reverseIndexMap:i}=kp.sparseFillEmptyRows(k("indices",e,t,r),k("values",e,t,r),k("denseShape",e,t,r),k("defaultValue",e,t,r));return[n,a,s,i]}case"SparseReshape":{let{outputIndices:n,outputShape:a}=kp.sparseReshape(k("inputIndices",e,t,r),k("inputShape",e,t,r),k("newShape",e,t,r));return[n,a]}case"SparseSegmentMean":return[kp.sparseSegmentMean(k("data",e,t,r),k("indices",e,t,r),k("segmentIds",e,t,r))];case"SparseSegmentSum":return[kp.sparseSegmentSum(k("data",e,t,r),k("indices",e,t,r),k("segmentIds",e,t,r))];default:throw TypeError(`Node type ${e.op} is not implemented`)}},gH=(e,t,r)=>{switch(e.op){case"FFT":return[Bf(k("x",e,t,r))];case"IFFT":return[Gp(k("x",e,t,r))];case"RFFT":return[Wf(k("x",e,t,r))];case"IRFFT":return[hA(k("x",e,t,r))];default:throw TypeError(`Node type ${e.op} is not implemented`)}},yH=(e,t,r)=>{switch(e.op){case"StringNGrams":{let{nGrams:n,nGramsSplits:a}=i0.stringNGrams(k("data",e,t,r),k("dataSplits",e,t,r),k("separator",e,t,r),k("nGramWidths",e,t,r),k("leftPad",e,t,r),k("rightPad",e,t,r),k("padWidth",e,t,r),k("preserveShortSequences",e,t,r));return[n,a]}case"StringSplit":{let{indices:n,values:a,shape:s}=i0.stringSplit(k("input",e,t,r),k("delimiter",e,t,r),k("skipEmpty",e,t,r));return[n,a,s]}case"StringToHashBucketFast":return[i0.stringToHashBucketFast(k("input",e,t,r),k("numBuckets",e,t,r))];default:throw TypeError(`Node type ${e.op} is not implemented`)}},AH=(e,t,r)=>{switch(e.op){case"Cast":return[me(k("x",e,t,r),k("dtype",e,t,r))];case"ExpandDims":{let n=k("axis",e,t,r);return[Kt(k("x",e,t,r),n)]}case"Squeeze":{let n=k("axis",e,t,r);return[tt(k("x",e,t,r),n)]}case"Reshape":return[U(k("x",e,t,r),k("shape",e,t,r))];case"MirrorPad":return[yw(k("x",e,t,r),k("padding",e,t,r),k("mode",e,t,r))];case"PadV2":case"Pad":return[Xn(k("x",e,t,r),k("padding",e,t,r),k("constantValue",e,t,r))];case"SpaceToBatchND":{let n=k("blockShape",e,t,r),a=k("paddings",e,t,r);return[Of(k("x",e,t,r),n,a)]}case"BatchToSpaceND":{let n=k("blockShape",e,t,r),a=k("crops",e,t,r);return[Nf(k("x",e,t,r),n,a)]}case"DepthToSpace":{let n=k("blockSize",e,t,r),a=k("dataFormat",e,t,r).toUpperCase();return[tw(k("x",e,t,r),n,a)]}case"BroadcastTo":return[Mp(k("x",e,t,r),k("shape",e,t,r))];case"BroadcastArgs":return[q7(k("s0",e,t,r),k("s1",e,t,r))];default:throw TypeError(`Node type ${e.op} is not implemented`)}};function E4(e,t,r,n){let a=((s,i,o)=>{switch(s.category){case"arithmetic":return X(()=>Xj(s,i,o));case"basic_math":return X(()=>Kj(s,i,o));case"control":return tH(s,i,o);case"convolution":return X(()=>rH(s,i,o));case"creation":return X(()=>nH(s,i,o));case"dynamic":return aH(s,i,o);case"evaluation":return X(()=>sH(s,i,o));case"image":return X(()=>uH(s,i,o));case"graph":return X(()=>iH(s,i,o));case"logical":return X(()=>dH(s,i,o));case"matrices":return X(()=>pH(s,i,o));case"normalization":return X(()=>hH(s,i,o));case"reduction":return X(()=>cH(s,i,o));case"slice_join":return X(()=>fH(s,i,o));case"sparse":return X(()=>mH(s,i,o));case"spectral":return X(()=>gH(s,i,o));case"string":return X(()=>yH(s,i,o));case"transformation":return X(()=>AH(s,i,o));case"hash_table":return lH(s,i,o,n);case"custom":let l=yk(s.op);if(l&&l.customExecutor)return l.customExecutor(new qj(s,i,o));throw TypeError(`Custom op ${s.op} is not registered.`);default:throw TypeError(`Unknown op '${s.op}'. File an issue at https://github.com/tensorflow/tfjs/issues so we can add it, or register a custom execution with tf.registerOp()`)}})(e,t,r);return v.isPromise(a)?a.then(s=>[].concat(s)):[].concat(a)}var R4=class{constructor(e={},t={},r={},n={}){this.weightMap=e,this.tensorArrayMap=t,this.tensorListMap=r,this.functionMap=n,this.rootContext={id:0,frameName:"",iterationId:0},this.contexts=[this.rootContext],this.lastId=0,this.generateCurrentContextIds()}newFrame(e,t){return{id:e,frameName:t,iterationId:0}}set currentContext(e){this.contexts!==e&&(this.contexts=e,this.generateCurrentContextIds())}get currentContext(){return this.contexts}get currentContextId(){return this._currentContextIds[0]}get currentContextIds(){return this._currentContextIds}generateCurrentContextIds(){let e=[];for(let t=0;t<this.contexts.length-1;t++){let r=this.contexts.slice(0,this.contexts.length-t);e.push(this.contextIdforContexts(r))}e.push(""),this._currentContextIds=e}contextIdforContexts(e){return e?e.map(t=>t.id===0&&t.iterationId===0?"":`${t.frameName}-${t.iterationId}`).join("/"):""}enterFrame(e){this.contexts&&(this.lastId++,this.contexts=this.contexts.slice(),this.contexts.push(this.newFrame(this.lastId,e)),this._currentContextIds.unshift(this.contextIdforContexts(this.contexts)))}exitFrame(){if(this.contexts&&this.contexts.length>1)this.contexts=this.contexts.slice(),this.contexts.splice(-1),this.currentContextIds.shift();else throw new Error("Cannot exit frame, the context is empty")}nextIteration(){if(this.contexts&&this.contexts.length>0){this.contexts=this.contexts.slice(),this.lastId++;let e=Object.assign({},this.contexts[this.contexts.length-1]);e.iterationId+=1,e.id=this.lastId,this.contexts.splice(-1,1,e),this._currentContextIds.splice(0,1,this.contextIdforContexts(this.contexts))}else throw new Error("Cannot increase frame iteration, the context is empty")}getWeight(e){return this.weightMap[e]}addTensorArray(e){this.tensorArrayMap[e.id]=e}getTensorArray(e){return this.tensorArrayMap[e]}addTensorList(e){this.tensorListMap[e.id]=e}getTensorList(e){return this.tensorListMap[e]}dispose(e){for(let t in this.tensorArrayMap)this.tensorArrayMap[t].clearAndClose(e);for(let t in this.tensorListMap)this.tensorListMap[t].clearAndClose(e)}};function $4(e,t,r,n){let a=new Set,s=[],i=null,o=null,l=new Set,u=Object.keys(e).map(p=>dn(p)[0]),d=[];n!=null&&(d=n.map(p=>dn(p.name)[0]));let h=[...t];for(;h.length>0;){let p=h.pop();if((Lk(p)||kH(p)||IH(p))&&i==null&&(i=p,o=i.children.map(c=>c.name).filter(c=>a.has(c))),a.add(p.name),r[p.name]==null&&u.indexOf(p.name)===-1&&d.indexOf(p.name)===-1){if(p.inputs.length===0){s.push(p.name);continue}p.inputs.forEach(c=>{l.has(c.name)||(l.add(c.name),h.push(c))})}}return{inputs:e,outputs:t,usedNodes:a,missingInputs:s,dynamicNode:i,syncInputs:o}}function xH(e,t,r){let{usedNodes:n,inputs:a}=r,s=[],i=Object.keys(a).map(d=>dn(d)[0]).map(d=>e.nodes[d]),o=e.initNodes;i.forEach(d=>{n.has(d.name)&&s.push(d)}),e.weights.forEach(d=>{n.has(d.name)&&s.push(d)}),o!=null&&o.forEach(d=>{n.has(d.name)&&s.push(d)});let l=new Set,u=[];for(;s.length>0;){let d=s.pop();l.add(d.name),t[d.name]||u.push(d),d.children.forEach(h=>{!l.has(h.name)&&n.has(h.name)&&h.inputs.every(p=>l.has(p.name))&&s.push(h)})}return u}var bH=["Switch","Merge","Enter","Exit","NextIteration","StatelessIf","StatelessWhile","if","While"],vH=["NonMaxSuppressionV2","NonMaxSuppressionV3","NonMaxSuppressionV5","Where"],wH=["HashTable","HashTableV2","LookupTableImport","LookupTableImportV2","LookupTableFind","LookupTableFindV2","LookupTableSize","LookupTableSizeV2"];function Lk(e){return bH.indexOf(e.op)>=0}function kH(e){return vH.indexOf(e.op)>=0}function IH(e){return wH.indexOf(e.op)>=0}var Yg=class{constructor(e,t){this.graph=e,this.parent=t,this.compiledMap=new Map,this._weightMap={},this.SEPERATOR=",",this._functions={},this._functionExecutorMap={},this.intermediateTensors={},this.keepTensorForDebug=!1,this._outputs=e.outputs,this._inputs=e.inputs,this._initNodes=e.initNodes,this._signature=e.signature,this._functions=e.functions,e.functions!=null&&Object.keys(e.functions).forEach(r=>{this._functionExecutorMap[r]=new Yg(e.functions[r],this)})}get weightIds(){return this.parent?this.parent.weightIds:this._weightIds}get functionExecutorMap(){return this.parent?this.parent.functionExecutorMap:this._functionExecutorMap}get weightMap(){return this.parent?this.parent.weightMap:this._weightMap}set weightMap(e){let t=Object.keys(e).map(r=>e[r].map(n=>n.id));this._weightIds=[].concat(...t),this._weightMap=e}set resourceManager(e){this._resourceManager=e}get inputs(){return this._inputs.map(e=>({name:e.name,shape:e.attrParams.shape?e.attrParams.shape.value:void 0,dtype:e.attrParams.dtype?e.attrParams.dtype.value:void 0}))}get outputs(){return this._outputs.map(e=>({name:e.name,shape:e.attrParams.shape?e.attrParams.shape.value:void 0,dtype:e.attrParams.dtype?e.attrParams.dtype.value:void 0}))}get inputNodes(){return this._inputs.map(e=>e.signatureKey||e.name)}get outputNodes(){return this._outputs.map(e=>{let t=e.signatureKey||e.name;return e.defaultOutput?`${t}:${e.defaultOutput}`:t})}get functions(){return Object.keys(this._functions).reduce((e,t)=>(e[t]=this._functions[t].signature,e),{})}getCompilationKey(e,t){let r=e.map(a=>a.name).sort(),n=t.map(a=>a.name).sort();return r.join(this.SEPERATOR)+"--"+n.join(this.SEPERATOR)}compile(e,t){let r=$4(e,t,this.weightMap,this._initNodes),{missingInputs:n,dynamicNode:a,syncInputs:s}=r;if(a!=null)throw new Error(`This execution contains the node '${a.name}', which has the dynamic op '${a.op}'. Please use model.executeAsync() instead. Alternatively, to avoid the dynamic ops, specify the inputs [${s}]`);if(n.length>0){let i=t.map(l=>l.name),o=Object.keys(e);throw new Error(`Cannot compute the outputs [${i}] from the provided inputs [${o}]. Missing the following inputs: [${n}]`)}return xH(this.graph,this.weightMap,r)}execute(e,t){e=this.mapInputs(e);let r=Object.keys(e).sort();this.checkInputs(e),this.checkInputShapeAndType(e),t=this.mapOutputs(t),this.checkOutputs(t);let n=r.map(d=>this.graph.nodes[dn(d)[0]]),a=t.map(d=>dn(d)[0]),s=a.map(d=>this.graph.nodes[d]);this.resetIntermediateTensors(),s.length===0&&(s=this._outputs);let i=this.getCompilationKey(n,s),o=this.compiledMap.get(i);o==null&&(o=this.compile(e,s),this.compiledMap.set(i,o));let l={},u={};return X(()=>{let d=new R4(this.weightMap,l,u,this.functionExecutorMap),h={...this.weightMap};Object.keys(e).forEach(f=>{let[m,g]=dn(f),y=[];y[g]=e[f],h[m]=y});let p=this.getFrozenTensorIds(h),c={};for(let f=0;f<o.length;f++){let m=o[f];if(!h[m.name]){let g=E4(m,h,d,this._resourceManager);if(v.isPromise(g))throw new Error(`The execution of the op '${m.op}' returned a promise. Please use model.executeAsync() instead.`);h[m.name]=g,this.checkTensorForDisposal(m.name,m,h,d,p,a,c)}}return this.parent==null&&d.dispose(p),t.map(f=>Lr(f,h,d))})}getFrozenTensorIds(e){let t=[].concat.apply([],Object.keys(e).map(r=>e[r]).map(r=>r.map(n=>n.id)));return new Set(t)}checkTensorForDisposal(e,t,r,n,a,s,i){t.category==="control"||s.indexOf(e)!==-1||(r[e].forEach(o=>{o!=null&&(i[o.id]=(i[o.id]||0)+t.children.length)}),t.inputs.forEach(o=>{if(o.category!=="control"){let l=Cj(o.name,r,n);l!=null&&l.forEach(u=>{if(u&&!u.kept&&!a.has(u.id)){let d=i[u.id];if(d===1){if(!this.keepTensorForDebug)u.dispose();else{let[h,p]=Ra(t.name,n);this.intermediateTensors[h]?this.intermediateTensors[h][p]=u:(this.intermediateTensors[h]=[],this.intermediateTensors[h][p]=u)}delete i[u.id]}else d!=null&&i[u.id]--}})}}))}async executeAsync(e,t){return this._executeAsync(e,t)}disposeIntermediateTensors(){!this.intermediateTensors||(Object.keys(this.intermediateTensors).forEach(e=>this.intermediateTensors[e].forEach(t=>t.dispose())),this.disposeTensorsMap())}disposeTensorsMap(){!this.tensorsMap||Object.keys(this.tensorsMap).forEach(e=>{this.tensorsMap[e].forEach(t=>{t&&!t.kept&&!t.isDisposed&&!this.keepIds.has(t.id)&&t.dispose()})})}getIntermediateTensors(){return this.tensorsMap}resetIntermediateTensors(){for(let e in this.intermediateTensors)this.intermediateTensors[e].forEach(t=>t.dispose()),delete this.intermediateTensors[e]}async _executeAsync(e,t,r=!1,n={},a={}){r||(e=this.mapInputs(e),this.checkInputs(e),this.checkInputShapeAndType(e),t=this.mapOutputs(t),this.checkOutputs(t));try{this.keepTensorForDebug=Z().getBool("KEEP_INTERMEDIATE_TENSORS")}catch(u){console.warn(u.message)}this.resetIntermediateTensors();let s=new R4(this.weightMap,n,a,this.functionExecutorMap);this.tensorsMap=await this.executeWithControlFlow(e,s,t,r);let i=t.map(u=>Lr(u,this.tensorsMap,s)),o=i.map(u=>u.id),l=Object.keys(e).map(u=>e[u].id);return this.keepIds=new Set([...o,...l,...this.weightIds]),this.keepTensorForDebug||this.disposeTensorsMap(),this.parent==null&&s.dispose(this.keepIds),i}async executeFunctionAsync(e,t,r){let n=e.reduce((a,s,i)=>(a[this.inputs[i].name]=s,a),{});return this._executeAsync(n,this.outputNodes,!0,t,r)}async executeWithControlFlow(e,t,r,n){let a=Object.keys(e),s=a.map(A=>this.graph.nodes[dn(A)[0]]),i=r.map(A=>dn(A)[0]),o=i.map(A=>this.graph.nodes[A]);o.length===0&&(o=this._outputs);let{usedNodes:l,missingInputs:u,dynamicNode:d,syncInputs:h}=$4(e,o,this.weightMap,this._initNodes),p=[...s,...this.graph.weights,...this._initNodes||[]].map(A=>({node:A,contexts:t.currentContext})),c={...this.weightMap};Object.keys(e).forEach(A=>{let[x,b]=dn(A),w=[];w[b]=e[A],c[x]=w});let f={},m=this.getFrozenTensorIds(c),g={};for(;p.length>0;){let A=this.processStack(s,p,t,c,g,m,i,f,l);await Promise.all(A)}d==null&&!n&&console.warn("This model execution did not contain any nodes with control flow or dynamic output shapes. You can use model.execute() instead.");let y=o.filter(A=>!Lk(A)&&!Lr(A.name,c,t)).map(A=>A.name);if(y.length>0){let A="";throw d!=null&&(A=`Alternatively, to avoid the dynamic ops, use model.execute() and specify the inputs [${h}]`),new Error(`Cannot compute the outputs [${y}] from the provided inputs [${a}]. Consider providing the following inputs: [${u}]. ${A}`)}return c}processStack(e,t,r,n,a,s,i,o,l){let u=[];for(;t.length>0;){let d=t.pop();r.currentContext=d.contexts;let h="";if(d.node.op==="Enter"&&k("isConstant",d.node,n,r)&&([h]=Ra(d.node.name,r)),n[d.node.name]==null){let p=E4(d.node,n,r,this._resourceManager);h||([h]=Ra(d.node.name,r));let c=r.currentContext;v.isPromise(p)?u.push(p.then(f=>(n[h]=f,r.currentContext=c,this.checkTensorForDisposal(h,d.node,n,r,s,i,o),this.processChildNodes(d.node,t,r,n,a,l),f))):(n[h]=p,this.checkTensorForDisposal(h,d.node,n,r,s,i,o),this.processChildNodes(d.node,t,r,n,a,l))}else this.processChildNodes(d.node,t,r,n,a,l)}return u}processChildNodes(e,t,r,n,a,s){e.children.forEach(i=>{let[o]=Ra(i.name,r);a[o]||!s.has(i.name)||(i.op==="Merge"?i.inputNames.some(l=>!!Lr(l,n,r))&&(a[o]=!0,t.push({contexts:r.currentContext,node:i})):i.inputNames.every(l=>!!Lr(l,n,r))&&(a[o]=!0,t.push({contexts:r.currentContext,node:i})))})}dispose(){Object.keys(this.weightMap).forEach(e=>this.weightMap[e].forEach(t=>t.dispose()))}checkInputShapeAndType(e){Object.keys(e).forEach(t=>{let r=e[t],[n]=dn(t),a=this.graph.nodes[n];if(a.attrParams.shape&&a.attrParams.shape.value){let s=a.attrParams.shape.value,i=s.length===r.shape.length&&r.shape.every((o,l)=>s[l]===-1||s[l]===o);v.assert(i,()=>`The shape of dict['${a.name}'] provided in model.execute(dict) must be [${s}], but was [${r.shape}]`)}a.attrParams.dtype&&a.attrParams.dtype.value&&v.assert(r.dtype===a.attrParams.dtype.value,()=>`The dtype of dict['${a.name}'] provided in model.execute(dict) must be ${a.attrParams.dtype.value}, but was ${r.dtype}`)})}mapInputs(e){let t={};for(let r in e)if(this._signature!=null&&this._signature.inputs!=null&&this._signature.inputs[r]!=null){let n=this._signature.inputs[r];t[n.name]=e[r]}else t[r]=e[r];return t}checkInputs(e){let t=Object.keys(e).filter(r=>{let[n]=dn(r);return this.graph.nodes[n]==null});if(t.length>0)throw new Error(`The dict provided in model.execute(dict) has keys: [${t}] that are not part of graph`)}mapOutputs(e){return e.map(t=>this._signature!=null&&this._signature.outputs!=null&&this._signature.outputs[t]!=null?this._signature.outputs[t].name:t,{})}checkOutputs(e){e.forEach(t=>{let[r]=dn(t);if(!this.graph.nodes[r])throw new Error(`The output '${t}' is not found in the graph`)})}},SH=class{constructor(e={},t={}){this.hashTableNameToHandle=e,this.hashTableMap=t}addHashTable(e,t){this.hashTableNameToHandle[e]=t.handle,this.hashTableMap[t.id]=t}getHashTableHandleByName(e){return this.hashTableNameToHandle[e]}getHashTableById(e){return this.hashTableMap[e]}dispose(){for(let e in this.hashTableMap)this.hashTableMap[e].clearAndClose(),delete this.hashTableMap[e];for(let e in this.hashTableNameToHandle)this.hashTableNameToHandle[e].dispose(),delete this.hashTableNameToHandle[e]}},CH="?tfjs-format=file",TH="model.json",vm=class{constructor(e,t={}){this.modelUrl=e,this.loadOptions=t,this.version="n/a",t==null&&(this.loadOptions={}),this.resourceManager=new SH}get modelVersion(){return this.version}get inputNodes(){return this.executor.inputNodes}get outputNodes(){return this.executor.outputNodes}get inputs(){return this.executor.inputs}get outputs(){return this.executor.outputs}get weights(){return this.executor.weightMap}get metadata(){return this.artifacts.userDefinedMetadata}get modelSignature(){return this.signature}findIOHandler(){let e=this.modelUrl;if(e.load!=null)this.handler=e;else if(this.loadOptions.requestInit!=null)this.handler=Cr.browserHTTPRequest(e,this.loadOptions);else{let t=Cr.getLoadHandlers(e,this.loadOptions);if(t.length===0)t.push(Cr.browserHTTPRequest(e,this.loadOptions));else if(t.length>1)throw new Error(`Found more than one (${t.length}) load handlers for URL '${[e]}'`);this.handler=t[0]}}async load(){if(this.findIOHandler(),this.handler.load==null)throw new Error("Cannot proceed with model loading because the IOHandler provided does not have the `load` method implemented.");let e=await this.handler.load();return this.loadSync(e)}loadSync(e){this.artifacts=e;let t=this.artifacts.modelTopology,r;this.artifacts.userDefinedMetadata!=null&&this.artifacts.userDefinedMetadata.signature!=null?r=this.artifacts.userDefinedMetadata.signature:r=this.artifacts.signature,this.signature=r,this.version=`${t.versions.producer}.${t.versions.minConsumer}`;let n=Cr.decodeWeights(this.artifacts.weightData,this.artifacts.weightSpecs);if(this.executor=new Yg(S4.Instance.transformGraph(t,this.signature)),this.executor.weightMap=this.convertTensorMapToTensorsMap(n),this.executor.resourceManager=this.resourceManager,e.modelInitializer!=null&&e.modelInitializer.node!=null){let a=S4.Instance.transformGraph(e.modelInitializer);this.initializer=new Yg(a),this.initializer.weightMap=this.executor.weightMap,this.initializer.resourceManager=this.resourceManager,this.initializer.executeAsync({},[])}return!0}async save(e,t){if(typeof e=="string"){let r=Cr.getSaveHandlers(e);if(r.length===0)throw new Error(`Cannot find any save handlers for URL '${e}'`);if(r.length>1)throw new Error(`Found more than one (${r.length}) save handlers for URL '${e}'`);e=r[0]}if(e.save==null)throw new Error("GraphModel.save() cannot proceed because the IOHandler provided does not have the `save` attribute defined.");return e.save(this.artifacts)}predict(e,t){return this.execute(e,this.outputNodes)}normalizeInputs(e){if(!(e instanceof nt)&&!Array.isArray(e))return e;if(e=Array.isArray(e)?e:[e],e.length!==this.inputNodes.length)throw new Error(`Input tensor count mismatch,the graph model has ${this.inputNodes.length} placeholders, while there are ${e.length} input tensors.`);return this.inputNodes.reduce((t,r,n)=>(t[r]=e[n],t),{})}normalizeOutputs(e){return e=e||this.outputNodes,Array.isArray(e)?e:[e]}execute(e,t){e=this.normalizeInputs(e),t=this.normalizeOutputs(t);let r=this.executor.execute(e,t);return r.length>1?r:r[0]}async executeAsync(e,t){e=this.normalizeInputs(e),t=this.normalizeOutputs(t);let r=await this.executor.executeAsync(e,t);return r.length>1?r:r[0]}getIntermediateTensors(){return this.executor.getIntermediateTensors()}disposeIntermediateTensors(){this.executor.disposeIntermediateTensors()}convertTensorMapToTensorsMap(e){return Object.keys(e).reduce((t,r)=>(t[r]=[e[r]],t),{})}dispose(){this.executor.dispose(),this.initializer&&this.initializer.dispose(),this.resourceManager.dispose()}};async function NH(e,t={}){if(e==null)throw new Error("modelUrl in loadGraphModel() cannot be null. Please provide a url or an IOHandler that loads the model");t==null&&(t={}),t.fromTFHub&&e.load==null&&(e.endsWith("/")||(e=e+"/"),e=`${e}${TH}${CH}`);let r=new vm(e,t);return await r.load(),r}var EH="0.0.0",Bk={};Le(Bk,{CSVDataset:()=>Jk,Dataset:()=>vd,FileDataSource:()=>s8,TextLineDataset:()=>Yk,URLDataSource:()=>i8,array:()=>JH,csv:()=>uq,func:()=>dq,generator:()=>pq,microphone:()=>cq,version_data:()=>fq,webcam:()=>hq,zip:()=>QH});var RH=Go(K0()),$H=Go(K0());function MH(e,t){return L0(e,t)}function L0(e,t,r=new Map,n=new Set){if(e==null)return null;if(typeof Blob=="function"&&e instanceof Blob)return e.slice();if(n.has(e))throw new Error("Circular references are not supported.");if(r.has(e))return r.get(e);let a=t(e);if(a.recurse&&a.value!==null)throw new Error("A deep map function may not return both a value and recurse=true.");if(a.recurse)if(Mu(e)){let s=Array.isArray(e)?[]:{};n.add(e);for(let i in e){let o=e[i],l=L0(o,t,r,n);s[i]=l}return n.delete(e),e.__proto__&&(s.__proto__=e.__proto__),s}else throw new Error(`Can't recurse into non-iterable type: ${e}`);else return r.set(e,a.value),a.value}function FH(e,t=Vk){return Wk(e,t)}function Wk(e,t,r=new Set){let n=e[0];if(r.has(n))throw new Error("Circular references are not supported.");let a=t(e);if(a.recurse&&a.value!==null)throw new Error("A deep zip function may not return both a value and recurse=true.");if(a.recurse)if(Mu(n)){let s=Array.isArray(n)?[]:{};r.add(n);for(let i in n){let o=e.map(u=>u[i]),l=Wk(o,t,r);s[i]=l}return r.delete(n),s}else throw new Error(`Can't recurse into non-iterable type: ${n}`);else return a.value}function Vk(e){return e===null?null:Mu(e[0])?{value:null,recurse:!0}:{value:e,recurse:!1}}async function Uk(e,t){let r=new Map;L0(e,t,r);for(let n of Array.from(r.keys())){let a=r.get(n);if(v.isPromise(a)){let s=await a;r.set(n,s)}}return L0(e,t,r)}function Mu(e){let t=!1;if(Z().get("IS_BROWSER"))t=e instanceof TextDecoder;else{let{StringDecoder:r}=Sv();t=e instanceof r}return e!=null&&!ArrayBuffer.isView(e)&&(Array.isArray(e)||typeof e=="object"&&!(e instanceof nt)&&!(e instanceof Promise)&&!t)}function PH(e){return e==null||_H(e)||Array.isArray(e)||typeof e=="object"&&e instanceof nt||v.isTypedArray(e)}function _H(e){return e===null||typeof e!="object"&&typeof e!="function"}function zH(e){return MH(e,OH)}function OH(e){return e instanceof nt?{value:e.clone(),recurse:!1}:Mu(e)?{value:null,recurse:!0}:{value:e,recurse:!1}}var Gk=class{constructor(e){if(this.capacity=e,this.begin=0,this.end=0,e==null)throw new RangeError("Can't create a ring buffer of unknown capacity.");if(e<1)throw new RangeError("Can't create ring buffer of capacity < 1.");this.data=new Array(e),this.doubledCapacity=2*e}wrap(e){for(;e<0;)e+=this.doubledCapacity;return e%this.doubledCapacity}get(e){if(e<0)throw new RangeError("Can't get item at a negative index.");return this.data[e%this.capacity]}set(e,t){if(e<0)throw new RangeError("Can't set item at a negative index.");this.data[e%this.capacity]=t}length(){let e=this.end-this.begin;return e<0&&(e=this.doubledCapacity+e),e}isFull(){return this.length()===this.capacity}isEmpty(){return this.length()===0}push(e){if(this.isFull())throw new RangeError("Ring buffer is full.");this.set(this.end,e),this.end=this.wrap(this.end+1)}pushAll(e){for(let t of e)this.push(t)}pop(){if(this.isEmpty())throw new RangeError("Ring buffer is empty.");this.end=this.wrap(this.end-1);let e=this.get(this.end);return this.set(this.end,void 0),e}unshift(e){if(this.isFull())throw new RangeError("Ring buffer is full.");this.begin=this.wrap(this.begin-1),this.set(this.begin,e)}shift(){if(this.isEmpty())throw new RangeError("Ring buffer is empty.");let e=this.get(this.begin);return this.set(this.begin,void 0),this.begin=this.wrap(this.begin+1),e}shuffleExcise(e){if(this.isEmpty())throw new RangeError("Ring buffer is empty.");let t=this.wrap(this.begin+e),r=this.get(t);return this.set(t,this.pop()),r}},jk=class extends Gk{constructor(){super(jk.INITIAL_CAPACITY)}isFull(){return!1}push(e){super.isFull()&&this.expand(),super.push(e)}unshift(e){super.isFull()&&this.expand(),super.unshift(e)}expand(){let e=this.capacity*2,t=new Array(e),r=this.length();for(let n=0;n<r;n++)t[n]=this.get(this.wrap(this.begin+n));this.data=t,this.capacity=e,this.doubledCapacity=2*this.capacity,this.begin=0,this.end=r}},Hk=jk;Hk.INITIAL_CAPACITY=32;function qk(e){return new BH(e)}function Yx(e){return new WH(e)}function DH(e,t){return new Xk(e,t)}function LH(e,t=Kk.FAIL){return new ZH(e,t)}var br=class{async toArray(){let e=[],t=await this.next();for(;!t.done;)e.push(t.value),t=await this.next();return e}async toArrayForTest(){let e=this.prefetch(100),t=[],r=await e.next();for(;!r.done;)t.push(r.value),r=await e.next();return t}async resolveFully(){let e=await this.next();for(;!e.done;)e=await this.next()}async resolveWhile(e){let t=await this.next(),r=e(t.value);for(;!t.done&&r;)t=await this.next(),r=e(t.value)}handleErrors(e){return new XH(this,e)}filter(e){return new HH(this,e)}map(e){return new qH(this,e)}mapAsync(e){return new M4(this,e)}serialMapAsync(e){return new M4(this,e).serial()}flatmap(e){return new KH(this,e)}async forEachAsync(e){return this.map(e).resolveFully()}async serialForEach(e){return this.serialMapAsync(e).resolveWhile(t=>t===!0)}rowMajorBatch(e,t=!0){return new jH(this,e,t)}columnMajorBatch(e,t=!0,r=Vk){return this.rowMajorBatch(e,t).map(n=>FH(n,r))}concatenate(e,t){return new Xk(qk([this,e]),t)}take(e){return e<0||e==null?this:new GH(this,e)}skip(e){return e<0||e==null?this:new UH(this,e)}prefetch(e){return new Zk(this,e)}shuffle(e,t){return new YH(this,e,t)}serial(){return new VH(this)}},BH=class extends br{constructor(e){super(),this.items=e,this.trav=0}summary(){return`Array of ${this.items.length} items`}async next(){if(this.trav>=this.items.length)return{value:null,done:!0};let e=this.items[this.trav];return this.trav++,{value:zH(e),done:!1}}},WH=class extends br{constructor(e){super(),this.nextFn=e}summary(){return"Function call"}async next(){try{return this.nextFn()}catch(e){throw e.message=`Error thrown while iterating through a dataset: ${e.message}`,e}}},VH=class extends br{constructor(e){super(),this.upstream=e,this.lastRead=Promise.resolve({value:null,done:!1})}summary(){return`${this.upstream.summary()} -> Serial`}async next(){return this.lastRead=this.lastRead.then(()=>this.serialNext()),this.lastRead}async serialNext(){return this.upstream.next()}},UH=class extends br{constructor(e,t){super(),this.upstream=e,this.maxCount=t,this.count=0,this.lastRead=Promise.resolve({value:null,done:!1})}summary(){return`${this.upstream.summary()} -> Skip`}async next(){return this.lastRead=this.lastRead.then(()=>this.serialNext()),this.lastRead}async serialNext(){for(;this.count++<this.maxCount;){let e=await this.upstream.next();if(e.done)return e;te(e.value)}return this.upstream.next()}},GH=class extends br{constructor(e,t){super(),this.upstream=e,this.maxCount=t,this.count=0}summary(){return`${this.upstream.summary()} -> Take`}async next(){return this.count++>=this.maxCount?{value:null,done:!0}:this.upstream.next()}},jH=class extends br{constructor(e,t,r=!0){super(),this.upstream=e,this.batchSize=t,this.enableSmallLastBatch=r,this.lastRead=Promise.resolve({value:null,done:!1})}summary(){return`${this.upstream.summary()} -> RowMajorBatch`}async next(){return this.lastRead=this.lastRead.then(()=>this.serialNext()),this.lastRead}async serialNext(){let e=[];for(;e.length<this.batchSize;){let t=await this.upstream.next();if(t.done)return this.enableSmallLastBatch&&e.length>0?{value:e,done:!1}:{value:null,done:!0};e.push(t.value)}return{value:e,done:!1}}},HH=class extends br{constructor(e,t){super(),this.upstream=e,this.predicate=t,this.lastRead=Promise.resolve({value:null,done:!1})}summary(){return`${this.upstream.summary()} -> Filter`}async next(){return this.lastRead=this.lastRead.then(()=>this.serialNext()),this.lastRead}async serialNext(){for(;;){let e=await this.upstream.next();if(e.done||this.predicate(e.value))return e;te(e.value)}}},qH=class extends br{constructor(e,t){super(),this.upstream=e,this.transform=t}summary(){return`${this.upstream.summary()} -> Map`}async next(){let e=await this.upstream.next();if(e.done)return{value:null,done:!0};let t=ha.getTensorsInContainer(e.value),r=this.transform(e.value),n=ha.getTensorsInContainer(r);for(let a of t)ha.isTensorInList(a,n)||a.dispose();return{value:r,done:!1}}},XH=class extends br{constructor(e,t){super(),this.upstream=e,this.handler=t,this.count=0,this.lastRead=Promise.resolve({value:null,done:!1})}summary(){return`${this.upstream.summary()} -> handleErrors`}async next(){return this.lastRead=this.lastRead.then(()=>this.serialNext()),this.lastRead}async serialNext(){for(;;)try{return await this.upstream.next()}catch(e){if(!this.handler(e))return{value:null,done:!0}}}},M4=class extends br{constructor(e,t){super(),this.upstream=e,this.transform=t}summary(){return`${this.upstream.summary()} -> AsyncMap`}async next(){let e=await this.upstream.next();if(e.done)return{value:null,done:!0};let t=ha.getTensorsInContainer(e.value),r=await this.transform(e.value),n=ha.getTensorsInContainer(r);for(let a of t)ha.isTensorInList(a,n)||a.dispose();return{value:r,done:!1}}},Jx=class extends br{constructor(){super(),this.outputQueue=new Hk,this.lastRead=Promise.resolve({value:null,done:!1})}async next(){return this.lastRead=this.lastRead.then(()=>this.serialNext()),this.lastRead}async serialNext(){for(;this.outputQueue.length()===0;)if(!await this.pump())return{value:null,done:!0};return{value:this.outputQueue.shift(),done:!1}}},KH=class extends Jx{constructor(e,t){super(),this.upstream=e,this.transform=t}summary(){return`${this.upstream.summary()} -> Flatmap`}async pump(){let e=await this.upstream.next();if(e.done)return!1;let t=ha.getTensorsInContainer(e.value),r=this.transform(e.value),n=ha.getTensorsInContainer(r);this.outputQueue.pushAll(r);for(let a of t)ha.isTensorInList(a,n)||a.dispose();return!0}},Xk=class extends br{constructor(e,t){super(),this.baseErrorHandler=t,this.lastRead=null,this.iterator=null,this.moreIterators=e}summary(){return"TODO: fill in upstream of chained summaries -> Chained"}async next(){return this.lastRead=this.readFromChain(this.lastRead),this.lastRead}async readFromChain(e){if(await e,this.iterator==null){let r=await this.moreIterators.next();if(r.done)return{value:null,done:!0};this.iterator=r.value,this.baseErrorHandler!=null&&(this.iterator=this.iterator.handleErrors(this.baseErrorHandler))}let t=await this.iterator.next();return t.done?(this.iterator=null,this.readFromChain(e)):t}},Kk=(e=>(e[e.FAIL=0]="FAIL",e[e.SHORTEST=1]="SHORTEST",e[e.LONGEST=2]="LONGEST",e))(Kk||{}),ZH=class extends br{constructor(e,t=0){super(),this.iterators=e,this.mismatchMode=t,this.count=0,this.currentPromise=null}summary(){return"{TODO: fill in upstream of zip summaries} -> Zip"}async nextState(e){await e;let t=0,r=0;function n(s){return s instanceof br?{value:s.next().then(i=>(t++,i.done&&r++,i.value)),recurse:!1}:{value:null,recurse:!0}}let a=await Uk(this.iterators,n);if(t===r)return{value:null,done:!0};if(r>0)switch(this.mismatchMode){case 0:throw new Error(`Zipped streams should have the same length. Mismatched at element ${this.count}.`);case 1:return{value:null,done:!0};case 2:default:}return this.count++,{value:a,done:!1}}async next(){return this.currentPromise=this.nextState(this.currentPromise),this.currentPromise}},Zk=class extends br{constructor(e,t){super(),this.upstream=e,this.bufferSize=t,this.buffer=new Gk(t)}summary(){return`${this.upstream.summary()} -> Prefetch`}refill(){for(;!this.buffer.isFull();){let e=this.upstream.next();this.buffer.push(e)}}next(){return this.refill(),this.buffer.shift()}},YH=class extends Zk{constructor(e,t,r){super(e,t),this.upstream=e,this.windowSize=t,this.upstreamExhausted=!1,this.random=$H.alea(r||v.now().toString()),this.lastRead=Promise.resolve({value:null,done:!1})}async next(){return this.lastRead=this.lastRead.then(()=>this.serialNext()),this.lastRead}randomInt(e){return Math.floor(this.random()*e)}chooseIndex(){return this.randomInt(this.buffer.length())}async serialNext(){for(this.upstreamExhausted||this.refill();!this.buffer.isEmpty();){let e=this.chooseIndex(),t=await this.buffer.shuffleExcise(e);if(t.done)this.upstreamExhausted=!0;else return this.refill(),t}return{value:null,done:!0}}},vd=class{constructor(){this.size=null}batch(e,t=!0){let r=this;v.assert(e>0,()=>`batchSize needs to be positive, but it is
|
|
${e}`);let n;return this.size===1/0||this.size==null?n=this.size:t?n=Math.ceil(this.size/e):n=Math.floor(this.size/e),un(async()=>(await r.iterator()).columnMajorBatch(e,t,eq),n)}concatenate(e){let t=this,r;return this.size===1/0||e.size===1/0?r=1/0:this.size!=null&&e.size!=null?r=this.size+e.size:r=null,un(async()=>(await t.iterator()).concatenate(await e.iterator()),r)}filter(e){let t=this,r;return this.size===1/0?r=1/0:r=null,un(async()=>(await t.iterator()).filter(n=>X(()=>e(n))),r)}async forEachAsync(e){return(await this.iterator()).forEachAsync(e)}map(e){let t=this;return un(async()=>(await t.iterator()).map(r=>X(()=>e(r))),this.size)}mapAsync(e){let t=this;return un(async()=>(await t.iterator()).mapAsync(e),this.size)}prefetch(e){if(e==null)throw new RangeError("`Dataset.prefetch()` requires bufferSize to be specified.");let t=this;return un(async()=>(await t.iterator()).prefetch(e),this.size)}repeat(e){let t=this,r;return this.size!=null&&e>0?r=this.size*e:e===0?r=0:this.size!=null&&(e===void 0||e<0)?r=1/0:r=null,un(async()=>{let n=Yx(async()=>({value:await t.iterator(),done:!1}));return DH(n.take(e))},r)}skip(e){let t=this,r;return this.size!=null&&e>=0&&this.size>=e?r=this.size-e:this.size!=null&&(this.size<e||e===void 0||e<0)?r=0:r=null,un(async()=>(await t.iterator()).skip(e),r)}shuffle(e,t,r=!0){if(e==null||e<0)throw this.size==null?new RangeError("`Dataset.shuffle()` requires bufferSize to be specified."):new RangeError(`\`Dataset.shuffle()\` requires bufferSize to be specified. If your data fits in main memory (for regular JS objects), and/or GPU memory (for \`tf.Tensor\`s), consider setting bufferSize to the dataset size (${this.size} elements)`);let n=this,a=RH.alea(t||v.now().toString());return un(async()=>{let s=a.int32();return r&&(s+=a.int32()),(await n.iterator()).shuffle(e,s.toString())},this.size)}take(e){let t=this,r;return this.size!=null&&this.size>e?r=e:this.size!=null&&this.size<=e?r=this.size:r=null,un(async()=>(await t.iterator()).take(e),r)}async toArray(){if(this.size===1/0)throw new Error("Can not convert infinite data stream to array.");return(await this.iterator()).toArray()}async toArrayForTest(){if(this.size===1/0)throw new Error("Can not convert infinite data stream to array.");return(await this.iterator()).toArrayForTest()}};vd.MAX_BUFFER_SIZE=1e4;function un(e,t=null){return new class extends vd{constructor(){super(...arguments),this.size=t}async iterator(){return e()}}}function JH(e){return un(async()=>qk(e),e.length)}function QH(e){if(!Mu(e))throw new Error("The argument to zip() must be an object or array.");let t;if(Array.isArray(e))for(let r=0;r<e.length;r++)t=t==null?e[r].size:Math.min(t,e[r].size);else if(e instanceof Object)for(let r in e)t=t==null?e[r].size:Math.min(t,e[r].size);return un(async()=>{let r=await Uk(e,n=>{if(n instanceof vd)return{value:n.iterator(),recurse:!1};if(Mu(n))return{value:null,recurse:!0};throw new Error("Leaves of the structure passed to zip() must be Datasets, not primitives.")});return LH(r,1)},t)}function eq(e){if(e===null)return null;let t=e[0];return PH(t)?{value:tq(e),recurse:!1}:{value:null,recurse:!0}}function tq(e){if(e.length===0)throw new Error("Can't make a batch of zero elements.");return e[0]instanceof nt?ur(e):ft(e)}var Yk=class extends vd{constructor(e){super(),this.input=e}async iterator(){return(await this.input.iterator()).decodeUTF8().split(`
|
|
`).map(e=>(e.endsWith("\r")&&(e=e.slice(0,-1)),e))}},Qc='"',bp=Symbol("out"),F4=Symbol("field"),e0=Symbol("quote"),pg=Symbol("quoteafterquote"),P4=Symbol("quoteinquote"),Jk=class extends vd{constructor(e,t){super(),this.input=e,this.hasHeader=!0,this.fullColumnNames=null,this.columnNamesValidated=!1,this.columnConfigs=null,this.configuredColumnsOnly=!1,this.delimiter=",",this.delimWhitespace=!1,this.base=new Yk(e),t||(t={}),this.hasHeader=t.hasHeader!==!1,this.fullColumnNames=t.columnNames,this.columnConfigs=t.columnConfigs,this.configuredColumnsOnly=t.configuredColumnsOnly,t.delimWhitespace?(v.assert(t.delimiter==null,()=>"Delimiter should not be provided when delimWhitespace is true."),this.delimWhitespace=!0,this.delimiter=" "):this.delimiter=t.delimiter?t.delimiter:","}async columnNames(){return this.columnNamesValidated||await this.setColumnNames(),this.configuredColumnsOnly?Object.keys(this.columnConfigs):this.fullColumnNames}async setColumnNames(){let e=await this.maybeReadHeaderLine();if(!this.fullColumnNames&&!e)throw new Error("Column names must be provided if there is no header line.");this.fullColumnNames&&e&&v.assert(e.length===this.fullColumnNames.length,()=>"The length of provided columnNames ("+this.fullColumnNames.length.toString()+") does not match the length of the header line read from file ("+e.length.toString()+")."),this.fullColumnNames||(this.fullColumnNames=e);let t=this.fullColumnNames.reduce((n,a)=>(n[a]=n[a]+1||1,n),{}),r=Object.keys(t).filter(n=>t[n]>1);if(v.assert(r.length===0,()=>"Duplicate column names found: "+r.toString()),this.columnConfigs){for(let n of Object.keys(this.columnConfigs))if(this.fullColumnNames.indexOf(n)===-1)throw new Error('The key "'+n+'" provided in columnConfigs does not match any of the column names ('+this.fullColumnNames.toString()+").")}this.columnNamesValidated=!0}async maybeReadHeaderLine(){if(this.hasHeader){let e=await(await this.base.iterator()).next();if(e.done)throw new Error("No data was found for CSV parsing.");let t=e.value;return this.parseRow(t,!1)}else return null}async iterator(){this.columnNamesValidated||await this.setColumnNames();let e=await this.base.iterator();return this.hasHeader&&(e=e.skip(1)),e.map(t=>this.makeDataElement(t))}makeDataElement(e){let t=this.parseRow(e),r={},n={};for(let a=0;a<this.fullColumnNames.length;a++){let s=this.fullColumnNames[a],i=this.columnConfigs?this.columnConfigs[s]:null;if(!(this.configuredColumnsOnly&&!i)){let o=t[a],l=null;if(o==="")if(i&&i.default!==void 0)l=i.default;else{if(i&&(i.required||i.isLabel))throw new Error(`Required column ${s} is empty in this line: ${e}`);l=void 0}else{let u=Number(o);if(isNaN(u))i&&i.dtype==="bool"?l=this.getBoolean(o):l=o;else if(!i||!i.dtype)l=u;else switch(i.dtype){case"float32":l=u;break;case"int32":l=Math.floor(u);break;case"bool":l=this.getBoolean(o);break;default:l=u}}i&&i.isLabel?n[s]=l:r[s]=l}}return Object.keys(n).length===0?r:{xs:r,ys:n}}getBoolean(e){return e==="1"||e.toLowerCase()==="true"?1:0}parseRow(e,t=!0){let r=[],n=0,a=e.length,s=bp;for(let i=0;i<a;i++)switch(s){case bp:switch(e.charAt(i)){case Qc:n=i+1,s=e0;break;case this.delimiter:if(n=i+1,this.delimiter===" "&&this.delimWhitespace)break;r.push(""),s=bp;break;default:s=F4,n=i;break}break;case F4:switch(e.charAt(i)){case this.delimiter:r.push(e.substring(n,i)),s=bp,n=i+1;break;default:}break;case e0:switch(e.charAt(i)){case Qc:s=pg;break;default:}break;case pg:switch(e.charAt(i)){case this.delimiter:r.push(e.substring(n,i-1)),s=bp,n=i+1;break;case Qc:s=e0;break;default:s=P4;break}break;case P4:switch(e.charAt(i)){case Qc:s=e0;break;default:}break;default:}if(s===pg?r.push(e.substring(n,a-1)):r.push(e.substring(n)),t&&r.length!==this.fullColumnNames.length)throw new Error(`Invalid row in csv file. Should have ${this.fullColumnNames.length} elements in a row, but got ${r}`);return r}},Qk=class extends br{constructor(e){super(),this.microphoneConfig=e,this.isClosed=!1,this.fftSize=e.fftSize||1024;let t=Math.log2(this.fftSize);if(this.fftSize<0||t<4||t>14||!Number.isInteger(t))throw new Error(`Invalid fftSize: it must be a power of 2 between 2 to 4 and 2 to 14, but got ${this.fftSize}`);if(this.numFrames=e.numFramesPerSpectrogram||43,this.sampleRateHz=e.sampleRateHz,this.columnTruncateLength=e.columnTruncateLength||this.fftSize,this.audioTrackConstraints=e.audioTrackConstraints,this.smoothingTimeConstant=e.smoothingTimeConstant||0,this.includeSpectrogram=e.includeSpectrogram!==!1,this.includeWaveform=e.includeWaveform===!0,!this.includeSpectrogram&&!this.includeWaveform)throw new Error("Both includeSpectrogram and includeWaveform are false. At least one type of data should be returned.")}summary(){return"microphone"}static async create(e={}){if(!Z().get("IS_BROWSER"))throw new Error("microphone API is only supported in browser environment.");let t=new Qk(e);return await t.start(),t}async start(){try{this.stream=await navigator.mediaDevices.getUserMedia({audio:this.audioTrackConstraints==null?!0:this.audioTrackConstraints,video:!1})}catch(r){throw new Error(`Error thrown while initializing video stream: ${r.message}`)}if(!this.stream)throw new Error("Could not obtain audio from microphone.");let e=window.AudioContext||window.webkitAudioContext;if(this.audioContext=new e,!this.sampleRateHz)this.sampleRateHz=this.audioContext.sampleRate;else if(this.audioContext.sampleRate!==this.sampleRateHz)throw new Error(`Mismatch in sampling rate: Expected: ${this.sampleRateHz}; Actual: ${this.audioContext.sampleRate}`);let t=this.audioContext.createMediaStreamSource(this.stream);this.analyser=this.audioContext.createAnalyser(),this.analyser.fftSize=this.fftSize*2,this.analyser.smoothingTimeConstant=this.smoothingTimeConstant,t.connect(this.analyser),this.freqData=new Float32Array(this.fftSize),this.timeData=new Float32Array(this.fftSize)}async next(){if(this.isClosed)return{value:null,done:!0};let e,t,r=await this.getAudioData();if(this.includeSpectrogram){let n=this.flattenQueue(r.freqDataQueue);e=this.getTensorFromAudioDataArray(n,[this.numFrames,this.columnTruncateLength,1])}if(this.includeWaveform){let n=this.flattenQueue(r.timeDataQueue);t=this.getTensorFromAudioDataArray(n,[this.numFrames*this.fftSize,1])}return{value:{spectrogram:e,waveform:t},done:!1}}async capture(){return(await this.next()).value}async getAudioData(){let e=[],t=[],r=0;return new Promise(n=>{let a=setInterval(()=>{this.includeSpectrogram&&(this.analyser.getFloatFrequencyData(this.freqData),this.freqData[0]===-1/0&&n({freqDataQueue:e,timeDataQueue:t}),e.push(this.freqData.slice(0,this.columnTruncateLength))),this.includeWaveform&&(this.analyser.getFloatTimeDomainData(this.timeData),t.push(this.timeData.slice())),++r===this.numFrames&&(clearInterval(a),n({freqDataQueue:e,timeDataQueue:t}))},this.fftSize/this.sampleRateHz*1e3)})}stop(){this.isClosed||(this.isClosed=!0,this.analyser.disconnect(),this.audioContext.close(),this.stream!=null&&this.stream.getTracks().length>0&&this.stream.getTracks()[0].stop())}toArray(){throw new Error("Can not convert infinite audio stream to array.")}getSampleRate(){return this.sampleRateHz}flattenQueue(e){let t=e[0].length,r=new Float32Array(e.length*t);return e.forEach((n,a)=>r.set(n,a*t)),r}getTensorFromAudioDataArray(e,t){let r=new Float32Array(v.sizeFromShape(t));return r.set(e,r.length-e.length),ft(r,t)}},e8=class extends br{constructor(e,t){if(super(),this.webcamVideoElement=e,this.webcamConfig=t,this.isClosed=!0,this.resize=!1,this.needToResize())if(this.resize=!0,this.cropSize=[this.webcamConfig.resizeHeight,this.webcamConfig.resizeWidth],this.cropBoxInd=Tt([0],"int32"),this.webcamConfig.centerCrop){let r=this.webcamConfig.resizeWidth*1/this.webcamVideoElement.width,n=this.webcamConfig.resizeHeight*1/this.webcamVideoElement.height,a=(1-r)/2,s=(1-n)/2,i=a+r,o=n+s;this.cropBox=ca([s,a,o,i],[1,4])}else this.cropBox=ca([0,0,1,1],[1,4])}summary(){return"webcam"}static async create(e,t={}){if(!Z().get("IS_BROWSER"))throw new Error("tf.data.webcam is only supported in browser environment.");if(!e){if(e=document.createElement("video"),!t.resizeWidth||!t.resizeHeight)throw new Error("Please provide webcam video element, or resizeWidth and resizeHeight to create a hidden video element.");e.width=t.resizeWidth,e.height=t.resizeHeight}let r=new e8(e,t);return await r.start(),r}async start(){this.webcamConfig.facingMode&&v.assert(this.webcamConfig.facingMode==="user"||this.webcamConfig.facingMode==="environment",()=>`Invalid webcam facing mode: ${this.webcamConfig.facingMode}. Please provide 'user' or 'environment'`);try{this.stream=await navigator.mediaDevices.getUserMedia({video:{deviceId:this.webcamConfig.deviceId,facingMode:this.webcamConfig.facingMode?this.webcamConfig.facingMode:"user",width:this.webcamVideoElement.width,height:this.webcamVideoElement.height}})}catch(e){throw e.message=`Error thrown while initializing video stream: ${e.message}`,e}if(!this.stream)throw new Error("Could not obtain video from webcam.");try{this.webcamVideoElement.srcObject=this.stream}catch(e){console.log(e),this.webcamVideoElement.src=window.URL.createObjectURL(this.stream)}return this.webcamVideoElement.play(),this.isClosed=!1,new Promise(e=>{this.webcamVideoElement.onloadedmetadata=()=>{e()}})}async next(){if(this.isClosed)return{value:null,done:!0};let e;try{e=On.fromPixels(this.webcamVideoElement)}catch(t){throw new Error(`Error thrown converting video to pixels: ${JSON.stringify(t)}`)}if(this.resize)try{return{value:this.cropAndResizeFrame(e),done:!1}}catch(t){throw new Error(`Error thrown cropping the video: ${t.message}`)}finally{e.dispose()}else return{value:e,done:!1}}needToResize(){return!!(this.webcamConfig.resizeWidth&&this.webcamConfig.resizeHeight&&(this.webcamVideoElement.width!==this.webcamConfig.resizeWidth||this.webcamVideoElement.height!==this.webcamConfig.resizeHeight))}cropAndResizeFrame(e){return X(()=>{let t=Kt(me(e,"float32"),0),r;r=Ie.cropAndResize(t,this.cropBox,this.cropBoxInd,this.cropSize,"bilinear");let n=r.shape;return U(r,n.slice(1))})}async capture(){return(await this.next()).value}stop(){this.stream.getTracks().forEach(e=>e.stop());try{this.webcamVideoElement.srcObject=null}catch(e){console.log(e),this.webcamVideoElement.src=null}this.isClosed=!0}toArray(){throw new Error("Can not convert infinite video stream to array.")}},t8=class{},r8=class extends br{split(e){return new rq(this,e)}},rq=class extends r8{constructor(e,t){super(),this.upstream=e,this.impl=new nq(e,t)}summary(){return this.impl.summary()}async next(){return this.impl.next()}},nq=class extends Jx{constructor(e,t){super(),this.upstream=e,this.separator=t,this.carryover=""}summary(){return`${this.upstream.summary()} -> Split('${this.separator}')`}async pump(){let e=await this.upstream.next();if(e.done)return this.carryover===""?!1:(this.outputQueue.push(this.carryover),this.carryover="",!0);let t=e.value.split(this.separator);t[0]=this.carryover+t[0];for(let r of t.slice(0,-1))this.outputQueue.push(r);return this.carryover=t[t.length-1],!0}},aq=class extends br{decodeUTF8(){return new sq(this)}},sq=class extends r8{constructor(e){super(),this.upstream=e,this.impl=new iq(e)}summary(){return this.impl.summary()}async next(){return this.impl.next()}},iq=class extends Jx{constructor(e){if(super(),this.upstream=e,Z().get("IS_BROWSER"))this.decoder=new TextDecoder("utf-8");else{let{StringDecoder:t}=Sv();this.decoder=new t("utf8")}}summary(){return`${this.upstream.summary()} -> Utf8`}async pump(){let e=await this.upstream.next(),t;if(e.done)return!1;t=e.value;let r;return Z().get("IS_BROWSER")?r=this.decoder.decode(t,{stream:!0}):r=this.decoder.write(Buffer.from(t.buffer)),this.outputQueue.push(r),!0}},n8=class extends aq{constructor(e,t={}){super(),this.file=e,this.options=t,v.assert(e instanceof Uint8Array||(Z().get("IS_BROWSER")?e instanceof File||e instanceof Blob:!1),()=>"FileChunkIterator only supports File, Blob and Uint8Array right now."),this.offset=t.offset||0,this.chunkSize=t.chunkSize||1024*1024}summary(){return`FileChunks ${this.file}`}async next(){return this.offset>=(this.file instanceof Uint8Array?this.file.byteLength:this.file.size)?{value:null,done:!0}:{value:await new Promise((e,t)=>{let r=this.offset+this.chunkSize;if(this.file instanceof Uint8Array)e(new Uint8Array(this.file.slice(this.offset,r)));else{let n=new FileReader;n.onload=s=>{let i=n.result;if(i instanceof ArrayBuffer&&(i=new Uint8Array(i)),!(i instanceof Uint8Array))return t(new TypeError("FileReader returned unknown type."));e(i)},n.onabort=s=>t(new Error("Aborted")),n.onerror=s=>t(new Error(s.type));let a=this.file.slice(this.offset,r);n.readAsArrayBuffer(a)}this.offset=r}),done:!1}}};async function oq(e,t={},r){let n,a;typeof e=="string"?n=e:(n=e.url,a=lq(e));let s=await(r||v.fetch)(n,a);if(s.ok){let i=new Uint8Array(await s.arrayBuffer());return new n8(i,t)}else throw new Error(s.statusText)}var lq=e=>({method:e.method,headers:e.headers,body:e.body,mode:e.mode,credentials:e.credentials,cache:e.cache,redirect:e.redirect,referrer:e.referrer,integrity:e.integrity});function a8(e){return typeof e=="string"&&e.slice(0,7)==="file://"}var s8=class extends t8{constructor(e,t={}){super(),this.input=e,this.options=t}async iterator(){if(a8(this.input)&&Z().get("IS_NODE")){let e=fy();this.input=e.readFileSync(this.input.slice(7))}return new n8(this.input,this.options)}},i8=class extends t8{constructor(e,t={}){super(),this.url=e,this.fileOptions=t}async iterator(){return a8(this.url)?new s8(this.url,this.fileOptions).iterator():oq(this.url,this.fileOptions)}};function uq(e,t={}){return new Jk(new i8(e),t)}function dq(e){let t=Yx(e);return un(async()=>t)}function pq(e){return un(async()=>{let t=await e();return Yx(()=>t.next())})}async function hq(e,t){return e8.create(e,t)}async function cq(e){return Qk.create(e)}var fq="0.0.0";function Ce(e,t){Array.isArray(e)||(e=[e]),e.forEach(r=>{r!=null&&v.assert(r.dtype!=="complex64",()=>`${t} does not support complex64 tensors in the CPU backend.`)})}var mq=Kn.whereImpl,o8=class extends zu{constructor(){super(),this.blockSize=48,this.firstUse=!0,this.data=new Zp(this,Xt())}nextDataId(){return o8.nextDataId++}write(e,t,r){this.firstUse&&(this.firstUse=!1,Z().get("IS_NODE")&&C.warn(`
|
|
============================
|
|
Hi there \u{1F44B}. Looks like you are running TensorFlow.js in Node.js. To speed things up dramatically, install our node backend, which binds to TensorFlow C++, by running npm i @tensorflow/tfjs-node, or npm i @tensorflow/tfjs-node-gpu if you have CUDA. Then call require('@tensorflow/tfjs-node'); (-gpu suffix for CUDA) at the start of your program. Visit https://github.com/tensorflow/tfjs-node for more details.
|
|
============================`));let n={id:this.nextDataId()};return this.data.set(n,{values:e,dtype:r,refCount:1}),n}makeTensorInfo(e,t,r){let n;if(t==="string"&&r!=null&&r.length>0&&v.isString(r[0])){let a=r.map(s=>v.encodeString(s));n=this.write(a,e,t)}else n=this.write(r,e,t);return{dataId:n,shape:e,dtype:t}}refCount(e){return this.data.has(e)?this.data.get(e).refCount:0}incRef(e){let t=this.data.get(e);t.refCount++}decRef(e){if(this.data.has(e)){let t=this.data.get(e);t.refCount--}}move(e,t,r,n,a){this.data.set(e,{values:t,dtype:n,refCount:a})}numDataIds(){return this.data.numDataIds()}async read(e){return this.readSync(e)}readSync(e){let{dtype:t,complexTensorInfos:r}=this.data.get(e);if(t==="complex64"){let n=this.readSync(r.real.dataId),a=this.readSync(r.imag.dataId);return C.mergeRealAndImagArrays(n,a)}return this.data.get(e).values}bufferSync(e){let t=this.readSync(e.dataId),r=t;if(e.dtype==="string")try{r=t.map(n=>v.decodeString(n))}catch(n){throw new Error("Failed to decode encoded string bytes into utf-8")}return We(e.shape,e.dtype,r)}makeOutput(e,t,r){return Xt().makeTensorFromTensorInfo(this.makeTensorInfo(t,r,e),this)}disposeData(e,t=!1){if(this.data.has(e)){if(this.data.get(e).refCount--,!t&&this.data.get(e).refCount>0)return!1;let{complexTensorInfos:r}=this.data.get(e);r!=null&&(this.disposeData(r.real.dataId,!0),this.disposeData(r.imag.dataId,!0)),this.data.delete(e)}return!0}disposeIntermediateTensorInfo(e){this.disposeData(e.dataId)}async time(e){let t=v.now();return e(),{kernelMs:v.now()-t}}memory(){return{unreliable:!0,reasons:["The reported memory is an upper bound. Due to automatic garbage collection, the true allocated memory may be less."]}}where(e){Ce([e],"where");let t=this.readSync(e.dataId);return mq(e.shape,t)}dispose(){}floatPrecision(){return 32}epsilon(){return super.epsilon()}},Qx=o8;Qx.nextDataId=0;var wm={};Le(wm,{addImpl:()=>u8,bincountImpl:()=>t5,bincountReduceImpl:()=>d8,ceilImpl:()=>p8,concatImpl:()=>r5,equalImpl:()=>h8,expImpl:()=>f8,expm1Impl:()=>g8,floorImpl:()=>y8,gatherNdImpl:()=>A8,gatherV2Impl:()=>x8,greaterEqualImpl:()=>v8,greaterImpl:()=>b8,lessEqualImpl:()=>k8,lessImpl:()=>w8,linSpaceImpl:()=>I8,logImpl:()=>S8,maxImpl:()=>C8,maximumImpl:()=>T8,minimumImpl:()=>N8,multiplyImpl:()=>n5,negImpl:()=>E8,notEqualImpl:()=>R8,prodImpl:()=>$8,rangeImpl:()=>s5,rsqrtImpl:()=>M8,sigmoidImpl:()=>nX,simpleAbsImpl:()=>l8,sliceImpl:()=>W0,sparseFillEmptyRowsImpl:()=>P8,sparseReshapeImpl:()=>_8,sparseSegmentReductionImpl:()=>i5,sqrtImpl:()=>iX,squaredDifferenceImpl:()=>z8,stridedSliceImpl:()=>O8,stringNGramsImpl:()=>D8,stringSplitImpl:()=>L8,stringToHashBucketFastImpl:()=>B8,subImpl:()=>W8,tileImpl:()=>V8,topKImpl:()=>G8,transposeImpl:()=>a5,uniqueImpl:()=>j8});function l8(e){let t=new Float32Array(e.length);for(let r=0;r<e.length;++r)t[r]=Math.abs(e[r]);return t}var gq=e=>{let{x:t}=e.inputs,r=e.backend;Ce(t,"abs");let n=new Float32Array(v.sizeFromShape(t.shape)),a=r.data.get(t.dataId).values;return n=l8(a),r.makeOutput(n,t.shape,t.dtype)},yq={kernelName:Ho,backendName:"cpu",kernelFunc:gq};function Qt(e){return(t,r,n,a,s)=>{let i=C.assertAndGetBroadcastShape(t,r),o=i.length,l=v.computeStrides(i),u=v.sizeFromShape(i),d=v.getTypedArrayFromDType(s,u),h=t.length,p=r.length,c=v.computeStrides(t),f=v.computeStrides(r),m=C.getBroadcastDims(t,i),g=C.getBroadcastDims(r,i);if(m.length+g.length===0)for(let y=0;y<d.length;++y)d[y]=e(n[y%n.length],a[y%a.length]);else for(let y=0;y<d.length;++y){let A=v.indexToLoc(y,o,l),x=A.slice(-h);m.forEach(N=>x[N]=0);let b=v.locToIndex(x,h,c),w=A.slice(-p);g.forEach(N=>w[N]=0);let S=v.locToIndex(w,p,f);d[y]=e(n[b],a[S])}return[d,i]}}function pn(e){let{inputs:t,backend:r}=e,{real:n,imag:a}=t,s=r.data.get(n.dataId).values,i=r.data.get(a.dataId).values,o=r.makeTensorInfo(n.shape,"complex64"),l=r.data.get(o.dataId);return l.complexTensorInfos={real:r.makeTensorInfo(n.shape,"float32",s),imag:r.makeTensorInfo(a.shape,"float32",i)},o}var Aq={kernelName:Jp,backendName:"cpu",kernelFunc:pn};function B0(e,t,r="float32"){if(r==="complex64"){let a=B0(e,t,"float32"),s=B0(e,t,"float32");return pn({inputs:{real:a,imag:s},backend:e})}let n=v.makeZerosTypedArray(v.sizeFromShape(t),r);return e.makeTensorInfo(t,r,n)}function _a(e){let{inputs:t,backend:r}=e,{x:n}=t;return r.incRef(n.dataId),{dataId:n.dataId,shape:n.shape,dtype:n.dtype}}var xq={kernelName:mi,backendName:"cpu",kernelFunc:_a};function Lo(e){let{inputs:t,backend:r}=e,{input:n}=t,a=r.data.get(n.dataId).complexTensorInfos.real,s=r.data.get(a.dataId).values;return r.makeTensorInfo(a.shape,a.dtype,s)}var bq={kernelName:oh,backendName:"cpu",kernelFunc:Lo};function qs(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{dtype:s}=n;if(s==="complex64"){if(a.dtype==="complex64")return _a({inputs:{x:a},backend:r});let i=B0(r,a.shape,a.dtype),o=qs({inputs:{x:a},backend:r,attrs:{dtype:"float32"}}),l=pn({inputs:{real:o,imag:i},backend:r});return r.disposeIntermediateTensorInfo(i),r.disposeIntermediateTensorInfo(o),l}if(a.dtype==="complex64"){let i=Lo({inputs:{input:a},backend:r}),o=qs({inputs:{x:i},backend:r,attrs:{dtype:s}});return r.disposeIntermediateTensorInfo(i),o}if(!v.hasEncodingLoss(a.dtype,s)){let i=_a({inputs:{x:a},backend:r});return{dataId:i.dataId,shape:i.shape,dtype:s}}if(s==="int32"){let i=r.data.get(a.dataId).values,o=Int32Array.from(i);return r.makeTensorInfo(a.shape,"int32",o)}if(s==="bool"){let i=r.data.get(a.dataId).values,o=v.toTypedArray([0],a.dtype),[l,u]=Qt((d,h)=>d!==h?1:0)(a.shape,[],i,o,"bool");return r.makeTensorInfo(u,"bool",l)}throw new Error(`Error in Cast: failed to cast ${a.dtype} to ${s}`)}var vq={kernelName:ei,backendName:"cpu",kernelFunc:qs};function vr(e,t,r,n){return r==null?({inputs:a,backend:s})=>{let{a:i,b:o}=a,l=s;Ce([i,o],e);let u=l.data.get(i.dataId).values,d=l.data.get(o.dataId).values,h=i.dtype==="string"?C.fromUint8ToStringArray(u):u,p=i.dtype==="string"?C.fromUint8ToStringArray(d):d,c=n||i.dtype,[f,m]=t(i.shape,o.shape,h,p,c);return l.makeTensorInfo(m,c,f)}:({inputs:a,backend:s})=>{let{a:i,b:o}=a,l=s;if(i.dtype==="complex64"||o.dtype==="complex64"){let u=qs({inputs:{x:i},backend:l,attrs:{dtype:"complex64"}}),d=l.data.get(u.dataId),h=d.complexTensorInfos.real,p=d.complexTensorInfos.imag,c=l.data.get(h.dataId).values,f=l.data.get(p.dataId).values,m=qs({inputs:{x:o},backend:l,attrs:{dtype:"complex64"}}),g=l.data.get(m.dataId),y=g.complexTensorInfos.real,A=g.complexTensorInfos.imag,x=l.data.get(y.dataId).values,b=l.data.get(A.dataId).values,[w,S,N]=r(i.shape,o.shape,c,f,x,b),E=l.makeTensorInfo(N,"float32",w),R=l.makeTensorInfo(N,"float32",S),z=pn({inputs:{real:E,imag:R},backend:l});return l.disposeIntermediateTensorInfo(u),l.disposeIntermediateTensorInfo(m),l.disposeIntermediateTensorInfo(E),l.disposeIntermediateTensorInfo(R),z}else{let u=l.data.get(i.dataId).values,d=l.data.get(o.dataId).values,h=n||i.dtype,[p,c]=t(i.shape,o.shape,u,d,h);return l.makeTensorInfo(c,h,p)}}}function e5(e){return(t,r,n,a,s,i)=>{let o=C.assertAndGetBroadcastShape(t,r),l=v.sizeFromShape(o),u=o.length,d=v.computeStrides(o),h=v.getTypedArrayFromDType("float32",l),p=v.getTypedArrayFromDType("float32",l),c=C.getBroadcastDims(t,o),f=C.getBroadcastDims(r,o),m=C.mergeRealAndImagArrays(n,a),g=C.mergeRealAndImagArrays(s,i),y=t.length,A=v.computeStrides(t),x=r.length,b=v.computeStrides(r);if(c.length+f.length===0)for(let w=0;w<h.length;w++){let S=w%m.length,N=w%g.length,E=e(m[S*2],m[S*2+1],g[N*2],g[N*2+1]);h[w]=E.real,p[w]=E.imag}else for(let w=0;w<h.length;w++){let S=v.indexToLoc(w,u,d),N=S.slice(-y);c.forEach(I=>N[I]=0);let E=v.locToIndex(N,y,A),R=S.slice(-x);f.forEach(I=>R[I]=0);let z=v.locToIndex(R,x,b),M=e(m[E*2],m[E*2+1],g[z*2],g[z*2+1]);h[w]=M.real,p[w]=M.imag}return[h,p,o]}}var u8=Qt((e,t)=>e+t),wq=e5((e,t,r,n)=>({real:e+r,imag:t+n})),zh=vr(Ja,u8,wq),kq={kernelName:Ja,backendName:"cpu",kernelFunc:zh};function t5(e,t,r,n,a){let s=v.sizeFromShape(n),i=v.makeZerosTypedArray(a,r);for(let o=0;o<e.length;o++){let l=e[o];if(l<0)throw new Error("Input x must be non-negative!");l>=a||(s>0?i[l]+=t[o]:i[l]+=1)}return i}function d8(e,t,r,n=!1){let a=e.shape[0],s=e.shape[1],i=We([a,r],t.dtype);for(let o=0;o<a;o++)for(let l=0;l<s;l++){let u=e.get(o,l);if(u<0)throw new Error("Input x must be non-negative!");u>=r||(n?i.set(1,o,u):t.size>0?i.set(i.get(o,u)+t.get(o,l),o,u):i.set(i.get(o,u)+1,o,u))}return i}function qi(e){return(t,r,n)=>{let a=v.getTypedArrayFromDType(r,t.length);for(let s=0;s<t.length;++s)a[s]=e(t[s],n);return a}}function gt(e,t,r){return({inputs:n,attrs:a,backend:s})=>{let{x:i}=n;if(Ce(i,e),i.dtype==="string"||r==="string")throw new Error("unaryKernelFunc does not support string input/output");let o=s,l=o.data.get(i.dataId).values,u=v.sizeFromShape(i.shape),d=r||i.dtype,h=v.getArrayFromDType(d,u);for(let p=0;p<u;++p)h[p]=t(l[p],a);return o.makeTensorInfo(i.shape,d,h)}}function wd(e,t,r){return({inputs:n,attrs:a,backend:s})=>{let{x:i}=n;if(Ce(i,e),i.dtype==="string"||r==="string")throw new Error("unaryKernelFunc does not support string input/output");let o=s,l=o.data.get(i.dataId).values,u=r||i.dtype,d=t(l,u,a);return o.makeTensorInfo(i.shape,u,d)}}var p8=qi(e=>Math.ceil(e)),Iq=wd(ti,p8),Sq={kernelName:ti,backendName:"cpu",kernelFunc:Iq};function r5(e,t,r,n){let a=v.getArrayFromDType(r,v.sizeFromShape(t));if(n&&r!=="string"){let s=0;e.forEach(i=>{let o=v.sizeFromShape(i.shape);a.set(i.vals,s),s+=o})}else{let s=0;e.forEach(i=>{let o=r==="string"?C.fromUint8ToStringArray(i.vals):i.vals,l=0;for(let u=0;u<i.shape[0];++u){let d=u*t[1]+s;for(let h=0;h<i.shape[1];++h)a[d+h]=o[l++]}s+=i.shape[1]})}return a}var h8=Qt((e,t)=>e===t?1:0),c8=vr(Jo,h8,null,"bool"),Cq={kernelName:Jo,backendName:"cpu",kernelFunc:c8},f8=qi(e=>Math.exp(e)),m8=wd(di,f8,"float32"),Tq={kernelName:di,backendName:"cpu",kernelFunc:m8},g8=qi(e=>Math.expm1(e)),Nq=wd(el,g8),Eq={kernelName:el,backendName:"cpu",kernelFunc:Nq},y8=qi(e=>Math.floor(e)),Rq=wd(pi,y8),$q={kernelName:pi,backendName:"cpu",kernelFunc:Rq};function A8(e,t,r,n,a,s,i,o,l){let u=We([n,s],r);for(let d=0;d<n;d++){let h=[],p=0;for(let c=0;c<a;c++){let f=e[d*a+c];p+=f*i[c],h.push(f)}if(p<0||p>=l/s)throw new Error(`Invalid indices: ${h} does not index into ${o}`);for(let c=0;c<s;c++)u.values[d*s+c]=t.get(...t.indexToLoc(p*s+c))}return u}function x8(e,t,r){let n=We(r,e.dtype);for(let a=0;a<n.size;++a){let s=n.indexToLoc(a).slice(),i=s[0],o=s[2],l=t.locToIndex([i,o]);s[2]=t.values[l];let u=e.locToIndex(s);0<=u&&u<e.values.length&&(n.values[a]=e.values[u])}return n}var b8=Qt((e,t)=>e>t?1:0),Mq=vr(al,b8,null,"bool"),Fq={kernelName:al,backendName:"cpu",kernelFunc:Mq},v8=Qt((e,t)=>e>=t?1:0),Pq=vr(fi,v8,null,"bool"),_q={kernelName:fi,backendName:"cpu",kernelFunc:Pq},w8=Qt((e,t)=>e<t?1:0),zq=vr(sl,w8,null,"bool"),Oq={kernelName:sl,backendName:"cpu",kernelFunc:zq},k8=Qt((e,t)=>e<=t?1:0),Dq=vr(il,k8,null,"bool"),Lq={kernelName:il,backendName:"cpu",kernelFunc:Dq};function I8(e,t,r){let n=(t-e)/(r-1),a=v.makeZerosTypedArray(r,"float32");a[0]=e;for(let s=1;s<a.length;s++)a[s]=a[s-1]+n;return a}var S8=qi(e=>Math.log(e)),Bq=wd(yi,S8),Wq={kernelName:yi,backendName:"cpu",kernelFunc:Bq};function C8(e,t,r,n){let a=v.getTypedArrayFromDType(n,v.sizeFromShape(r));for(let s=0;s<a.length;++s){let i=s*t,o=e[i];for(let l=0;l<t;++l){let u=e[i+l];(Number.isNaN(u)||u>o)&&(o=u)}a[s]=o}return a}var T8=Qt((e,t)=>Math.max(e,t)),Vq=vr(xi,T8),Uq={kernelName:xi,backendName:"cpu",kernelFunc:Vq},N8=Qt((e,t)=>Math.min(e,t)),Gq=vr(ki,N8),jq={kernelName:ki,backendName:"cpu",kernelFunc:Gq},n5=Qt((e,t)=>e*t),Hq=e5((e,t,r,n)=>({real:e*r-t*n,imag:e*n+t*r})),km=vr(Si,n5,Hq),qq={kernelName:Si,backendName:"cpu",kernelFunc:km};function E8(e,t,r){let n=v.createScalarValue(-1,r);return n5([],t,n,e,r)}function Xq(e){let{inputs:t,backend:r}=e,{x:n}=t;Ce(n,"neg");let a=r.data.get(n.dataId).values,[s,i]=E8(a,n.shape,n.dtype);return r.makeTensorInfo(i,n.dtype,s)}var Kq={kernelName:ll,backendName:"cpu",kernelFunc:Xq},R8=Qt((e,t)=>e!==t?1:0),Zq=vr(ul,R8,null,"bool"),Yq={kernelName:ul,backendName:"cpu",kernelFunc:Zq};function a5(e,t,r,n,a){let s=t.length,i=v.sizeFromShape(t),o=v.computeStrides(t),l=v.computeStrides(a),u=v.getTypedArrayFromDType(r,v.sizeFromShape(a));for(let d=0;d<i;++d){let h=v.indexToLoc(d,s,o),p=new Array(h.length);for(let f=0;f<p.length;f++)p[f]=h[n[f]];let c=v.locToIndex(p,s,l);u[c]=e[d]}return u}function sn(e){let{inputs:t,attrs:r,backend:n}=e,{x:a}=t,{perm:s}=r;Ce(a,"transpose");let i=a.shape.length,o=new Array(i);for(let d=0;d<o.length;d++)o[d]=a.shape[s[d]];let l=n.data.get(a.dataId).values,u=a5(l,a.shape,a.dtype,s,o);return{dataId:n.write(u,o,a.dtype),shape:o,dtype:a.dtype}}var Jq={kernelName:Vi,backendName:"cpu",kernelFunc:sn};function $8(e,t,r,n){let[a,s]=C.computeOutAndReduceShapes(e,n),i=Nr(t,"int32"),o=v.makeZerosTypedArray(v.sizeFromShape(a),i),l=v.sizeFromShape(s);for(let u=0;u<o.length;++u){let d=u*l,h=1;for(let p=0;p<l;++p)h*=r[d+p];o[u]=h}return{outVals:o,outShape:a,outDtype:i}}function Qq(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{axis:s,keepDims:i}=n;Ce(a,"prod");let o=a.shape.length,l=v.parseAxisParam(s,a.shape),u=C.getAxesPermutation(l,o),d=l,h=a,p=[];u!=null&&(h=sn({inputs:{x:a},backend:r,attrs:{perm:u}}),p.push(h),d=C.getInnerMostAxes(d.length,o));let c=r.data.get(h.dataId).values,{outVals:f,outShape:m,outDtype:g}=$8(h.shape,h.dtype,c,d),y=m;return i&&(y=C.expandShapeToKeepDim(m,l)),p.forEach(A=>r.disposeIntermediateTensorInfo(A)),r.makeTensorInfo(y,g,f)}var eX={kernelName:Ei,backendName:"cpu",kernelFunc:Qq};function s5(e,t,r,n){let a=e===t,s=e<t&&r<0,i=t<e&&r>1;if(a||s||i)return v.makeZerosTypedArray(0,n);let o=Math.abs(Math.ceil((t-e)/r)),l=v.makeZerosTypedArray(o,n);t<e&&r===1&&(r=-1),l[0]=e;for(let u=1;u<l.length;u++)l[u]=l[u-1]+r;return l}var M8=qi(e=>1/Math.sqrt(e)),tX=wd(Fi,M8),rX={kernelName:Fi,backendName:"cpu",kernelFunc:tX},nX=qi(e=>1/(1+Math.exp(-e))),F8=gt(_i,e=>1/(1+Math.exp(-e))),aX={kernelName:_i,backendName:"cpu",kernelFunc:F8};function W0(e,t,r,n,a){let s=Ot.isSliceContinous(n,t,r),i=v.sizeFromShape(r),o=v.computeStrides(n);if(s){let h=Ot.computeFlatOffset(t,o);return a==="string"?e.slice(h,h+i):e.subarray(h,h+i)}let l=a==="string"?C.fromUint8ToStringArray(e):e,u=We(n,a,l),d=We(r,a);for(let h=0;h<d.size;++h){let p=d.indexToLoc(h),c=p.map((f,m)=>f+t[m]);d.set(u.get(...c),...p)}return a==="string"?C.fromStringArrayToUint8(d.values):d.values}function Bo(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{begin:s,size:i}=n;Ce(a,"slice");let[o,l]=Ot.parseSliceParams(a,s,i);Ot.assertParamsValid(a,o,l);let u=r.data.get(a.dataId).values,d=W0(u,o,l,a.shape,a.dtype);return r.makeTensorInfo(l,a.dtype,d)}var sX={kernelName:bl,backendName:"cpu",kernelFunc:Bo};function P8(e,t,r,n,a,s,i){let o=t[0],l=s[0],u=new Array(l),d=new Array(o),h=t[1];if(l===0){if(o!==0)throw new Error(C.getSparseFillEmptyRowsIndicesDenseShapeMismatch(o));let g=v.getArrayFromDType(r,0),y=v.getArrayFromDType(a,0);return[g,[0,h],y,u,d]}let p=!0,c=0,f=new Array(l).fill(0);for(let g=0;g<o;++g){let y=e[g*h];if(y<0)throw new Error(C.getSparseFillEmptyRowsNegativeIndexErrorMessage(g,y));if(y>=l)throw new Error(C.getSparseFillEmptyRowsOutOfRangeIndexErrorMessage(g,y,l));++f[y],p=p&&y>=c,c=y}let m=!0;for(let g=0;g<l;++g){let y=f[g]===0;u[g]=y,m=m&&!y,f[g]=Math.max(f[g],1),g>0&&(f[g]+=f[g-1])}if(m&&p){let g=e,y=n;for(let A=0;A<o;++A)d[A]=A;return[g,[o,h],y,u,d]}else{let g=f[l-1],y=v.getArrayFromDType(r,g*h),A=v.getArrayFromDType(a,g),x=new Array(l).fill(0);for(let b=0;b<o;++b){let w=e[b*h],S=x[w],N=(w===0?0:f[w-1])+S;x[w]++;for(let E=0;E<h;++E)y[N*h+E]=e[b*h+E];A[N]=n[b],d[b]=N}for(let b=0;b<l;++b)if(x[b]===0){let w=b===0?0:f[b-1];y[w*h+0]=b;for(let S=1;S<h;++S)y[w*h+S]=0;A[w]=i}return[y,[g,h],A,u,d]}}function _8(e,t,r,n,a){let s=v.sizeFromShape(n),i=t[0],o=a.length,l=[],u=1,d=-1;for(let m=0;m<o;++m){let g=a[m];if(g===-1){if(d!==-1)throw new Error(C.getSparseReshapeMultipleNegativeOneOutputDimErrorMessage(d,m));d=m,l.push(1)}else{if(g<0)throw new Error(C.getSparseReshapeNegativeOutputDimErrorMessage(m,g));u*=g,l.push(g)}}if(d!==-1){if(u<=0)throw new Error(C.getSparseReshapeEmptyTensorZeroOutputDimErrorMessage());let m=Math.trunc(s/u);if(u*m!==s)throw new Error(C.getSparseReshapeInputOutputMultipleErrorMessage(n,l));l[d]=m}if(v.sizeFromShape(l)!==s)throw new Error(C.getSparseReshapeInputOutputMismatchErrorMessage(n,l));let h=n.length,p=[];if(h>0){p[h-1]=1;for(let m=h-2;m>=0;--m)p[m]=p[m+1]*n[m+1]}let c=[];if(o>0){c[o-1]=1;for(let m=o-2;m>=0;--m)c[m]=c[m+1]*l[m+1]}let f=v.getArrayFromDType(r,i*o);for(let m=0;m<i;++m){let g=0;for(let y=0;y<h;++y)g+=e[m*h+y]*p[y];for(let y=0;y<o;++y)f[m*o+y]=Math.trunc(g/c[y]),g%=c[y]}return[f,[i,o],l]}function i5(e,t,r,n,a,s=!1,i=0){let o=n.length,l=[t[0],e.length/t[0]],u=l[1],d=o>0?a[o-1]+1:0;if(d<0)throw new Error(C.getSparseSegmentReductionNegativeSegmentIdsErrorMessage());let h=t.slice();h[0]=d;let p=h.reduce((A,x)=>A*x,1),c=v.getArrayFromDType(r,p);if(o===0)return d>0&&c.fill(i),[c,h];if(d<=0)throw new Error(C.getSparseSegmentReductionNegativeSegmentIdsErrorMessage());let f=0,m=1,g=0,y=a[f];for(;;){let A=0;if(m<o){if(A=a[m],y===A){++m;continue}if(y>=A)throw new Error(C.getSparseSegmentReductionNonIncreasingSegmentIdsErrorMessage())}if(y<0||y>=d)throw new Error(C.getSparseSegmentReductionSegmentIdOutOfRangeErrorMessage(y,d));y>g&&c.fill(i,g*u,y*u);for(let x=f;x<m;++x){let b=n[x];if(b<0||b>=l[0])throw new Error(C.getSparseSegmentReductionIndicesOutOfRangeErrorMessage(x,n[x],l[0]));for(let w=0;w<u;w++)c[y*u+w]+=e[b*u+w]}if(s)for(let x=0;x<u;x++)c[y*u+x]/=m-f;if(f=m,++m,g=y+1,y=A,m>o)break}return g<d&&c.fill(i,g*u,d*u),[c,h]}var iX=qi(e=>Math.sqrt(e)),oX=gt(zi,e=>Math.sqrt(e)),lX={kernelName:zi,backendName:"cpu",kernelFunc:oX},z8=Qt((e,t)=>{let r=e-t;return r*r}),uX=vr(Li,z8),dX={kernelName:Li,backendName:"cpu",kernelFunc:uX};function O8(e,t,r,n){let a=We(e,t.dtype);for(let s=0;s<a.size;s++){let i=a.indexToLoc(s),o=new Array(i.length);for(let l=0;l<o.length;l++)o[l]=i[l]*r[l]+n[l];a.set(t.get(...o),...i)}return a}var pX=class{constructor(e,t,r,n,a,s){this.separator=v.encodeString(e),this.nGramWidths=t,this.leftPad=v.encodeString(r),this.rightPad=v.encodeString(n),this.padWidth=a,this.preserveShort=s}getPadWidth(e){return Math.min(this.padWidth<0?e-1:this.padWidth,e-1)}getNumNGrams(e,t){let r=this.getPadWidth(t);return Math.max(0,e+2*r-t+1)}createNGrams(e,t,r,n,a,s){for(let i=0;i<a;++i){let o=this.getPadWidth(s),l=Math.max(0,o-i),u=Math.max(0,o-(a-(i+1))),d=s-(l+u),h=t+(l>0?0:i-o),p=0;p+=l*this.leftPad.length;for(let g=0;g<d;++g)p+=e[h+g].length;p+=u*this.rightPad.length,p+=(l+u+d-1)*this.separator.length,r[n+i]=new Uint8Array(p);let c=r[n+i],f=0,m=g=>g.forEach(y=>c[f++]=y);for(let g=0;g<l;++g)m(this.leftPad),m(this.separator);for(let g=0;g<d-1;++g)m(e[h+g]),m(this.separator);if(d>0){m(e[h+d-1]);for(let g=0;g<u;++g)m(this.separator),m(this.rightPad)}else{for(let g=0;g<u-1;++g)m(this.rightPad),m(this.separator);m(this.rightPad)}}}compute(e,t){let r=e.length,n=t.length;if(n>0){let o=t[0];if(o!==0)throw new Error(`First split value must be 0, got ${o}`);for(let l=1;l<n;++l){let u=t[l]>=o;if(u=u&&t[l]<=r,!u)throw new Error(`Invalid split value ${t[l]}, must be in [${o}, ${r}]`);o=t[l]}if(o!==r)throw new Error(`Last split value must be data size. Expected ${r}, got ${o}`)}let a=n-1,s=v.getArrayFromDType("int32",n);if(r===0||n===0){let o=new Array(r);for(let l=0;l<=a;++l)s[l]=0;return[o,s]}s[0]=0;for(let o=1;o<=a;++o){let l=t[o]-t[o-1],u=0;this.nGramWidths.forEach(d=>{u+=this.getNumNGrams(l,d)}),this.preserveShort&&l>0&&u===0&&(u=1),s[o]=s[o-1]+u}let i=new Array(s[a]);for(let o=0;o<a;++o){let l=t[o],u=s[o];if(this.nGramWidths.forEach(d=>{let h=t[o+1]-t[o],p=this.getNumNGrams(h,d);this.createNGrams(e,l,i,u,p,d),u+=p}),this.preserveShort&&u===s[o]){let d=t[o+1]-t[o];if(d===0)continue;let h=d+2*this.padWidth,p=1;this.createNGrams(e,l,i,u,p,h)}}return[i,s]}};function D8(e,t,r,n,a,s,i,o){return new pX(r,n,a,s,i,o).compute(e,t)}function hX(e,t,r,n){if(!e.length)return;if(t.length===0){for(let s=0;s<e.length;++s)n.push(e.subarray(s,s+1));return}if(t.length===1){let s=t[0],i=e.indexOf(s);for(;i!==-1;){let o=e.subarray(0,i);(!r||o.length!==0)&&n.push(o),e=e.subarray(i+1),i=e.indexOf(s)}(!r||e.length!==0)&&n.push(e);return}let a=0;for(let s=0;s<e.length+1;s++)if(s===e.length||t.indexOf(e[s])!==-1){let i=e.subarray(a,s);(!r||i.length!==0)&&n.push(i),a=s+1}}function L8(e,t,r){let n=e.length,a=[],s=0,i=0,o=new Array(n);for(let p=0;p<n;++p){let c=a.length;hX(e[p],t,r,a);let f=a.length-c;o[p]=f,s+=f,i=Math.max(i,f)}let l=v.getArrayFromDType("int32",s*2),u=new Array(s),d=[n,i],h=0;for(let p=0;p<n;++p)for(let c=0;c<o[p];++c)l[h*2]=p,l[h*2+1]=c,u[h]=a[h],++h;return[l,u,d]}function B8(e,t){let r=v.getArrayFromDType("int32",e.length);for(let n=0;n<e.length;++n)r[n]=v.fingerPrint64(e[n]).modulo(t).getLowBitsUnsigned();return r}var W8=Qt((e,t)=>e-t),cX=e5((e,t,r,n)=>({real:e-r,imag:t-n})),o5=vr(Bi,W8,cX),fX={kernelName:Bi,backendName:"cpu",kernelFunc:o5};function V8(e,t){let r=new Array(e.rank);for(let a=0;a<r.length;a++)r[a]=e.shape[a]*t[a];let n=We(r,e.dtype);for(let a=0;a<n.values.length;++a){let s=n.indexToLoc(a),i=new Array(e.rank);for(let l=0;l<i.length;l++)i[l]=s[l]%e.shape[l];let o=e.locToIndex(i);n.values[a]=e.values[o]}return n}var Cp=(e,t)=>{let r=t.value-e.value;return r===0?e.index-t.index:r};function U8(e,t,r=0,n=e.length-1){for(;n>r;){if(n-r>600){let o=n-r+1,l=t-r+1,u=Math.log(o),d=.5*Math.exp(2*u/3),h=.5*Math.sqrt(u*d*(o-d)/o)*Math.sign(l-o/2),p=Math.max(r,Math.floor(t-l*d/o+h)),c=Math.min(n,Math.floor(t+(o-l)*d/o+h));U8(e,t,p,c)}let a=e[t],s=r,i=n;for(v.swap(e,r,t),Cp(e[n],a)>0&&v.swap(e,r,n);s<i;){for(v.swap(e,s,i),s++,i--;Cp(e[s],a)<0;)s=s+1;for(;Cp(e[i],a)>0;)i=i-1}Cp(e[r],a)===0?v.swap(e,r,i):(i=i+1,v.swap(e,i,n)),i<=t&&(r=i+1),t<=i&&(n=i-1)}}function G8(e,t,r,n,a){let s=t[t.length-1],[i,o]=[e.length/s,s],l=v.getTypedArrayFromDType(r,i*n),u=v.getTypedArrayFromDType("int32",i*n);for(let h=0;h<i;h++){let p=h*o,c=e.subarray(p,p+o),f=new Array(c.length);c.forEach((A,x)=>f[x]={value:A,index:x}),n<f.length&&(U8(f,n),f=f.slice(0,n)),a&&f.sort(Cp);let m=h*n,g=l.subarray(m,m+n),y=u.subarray(m,m+n);for(let A=0;A<n;A++)g[A]=f[A].value,y[A]=f[A].index}let d=t.slice();return d[d.length-1]=n,[We(d,r,l),We(d,"int32",u)]}function j8(e,t,r,n){let a=v.parseAxisParam(t,r)[0],s=[1,r[0],1];for(let f=0;f<a;f++)s[0]*=r[f];s[1]=r[a];for(let f=a+1;f<r.length;f++)s[2]*=r[f];let i={},o=new Int32Array(r[a]),l=new ir(s,n,e),u=[],d=s[0]===1&&s[2]===1;for(let f=0;f<r[a];f++){let m;if(d)m=e[f].toString();else{let g=[];for(let y=0;y<s[0];y++)for(let A=0;A<s[2];A++)g.push(l.get(y,f,A));m=g.join(",")}if(i[m]!==void 0)o[f]=i[m];else{let g=Object.keys(i).length;i[m]=g,o[f]=g,u.push(f)}}let h=s.slice();h[1]=Object.keys(i).length;let p=new ir(h,n);u.forEach((f,m)=>{for(let g=0;g<s[0];g++)for(let y=0;y<s[2];y++)p.set(l.get(g,f,y),g,m,y)});let c=r.slice();return c[a]=h[1],{outputValues:p.values,outputShape:c,indices:o}}var mX="0.0.0";$l("cpu",()=>new Qx,1);var H8=gt(ui,e=>e>=0?e:Math.exp(e)-1),gX={kernelName:ui,backendName:"cpu",kernelFunc:H8};function q8(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{alpha:s}=n;Ce([a],"leakyRelu");let i=v.sizeFromShape(a.shape),o=r.data.get(a.dataId).values,l=v.getTypedArrayFromDType("float32",i);for(let u=0;u<o.length;u++)l[u]=o[u]<0?s*o[u]:o[u];return r.makeTensorInfo(a.shape,"float32",l)}var yX={kernelName:gi,backendName:"cpu",kernelFunc:q8},AX=Qt((e,t)=>e<0?t*e:e);function X8(e){let{inputs:t,backend:r}=e,{x:n,alpha:a}=t;Ce([n,a],"prelu");let s=r.data.get(n.dataId).values,i=r.data.get(a.dataId).values,[o,l]=AX(n.shape,a.shape,s,i,"float32");return r.makeTensorInfo(l,"float32",o)}var xX={kernelName:Ni,backendName:"cpu",kernelFunc:X8},K8=gt(Ri,e=>Math.max(0,e)),bX={kernelName:Ri,backendName:"cpu",kernelFunc:K8},Z8=gt(Mi,e=>Math.min(Math.max(0,e),6)),vX={kernelName:Mi,backendName:"cpu",kernelFunc:Z8};function l5(e,t,r,n,a){if(r==="linear")return _a({inputs:{x:t},backend:e});if(r==="relu")return K8({inputs:{x:t},backend:e});if(r==="elu")return H8({inputs:{x:t},backend:e});if(r==="relu6")return Z8({inputs:{x:t},backend:e});if(r==="prelu")return X8({inputs:{x:t,alpha:n},backend:e});if(r==="leakyrelu")return q8({inputs:{x:t},backend:e,attrs:{alpha:a}});if(r==="sigmoid")return F8({inputs:{x:t},backend:e});throw new Error(`Activation ${r} has not been implemented for the CPU backend.`)}function Mt(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{shape:s}=n,i=v.sizeFromShape(a.shape),o=v.inferFromImplicitShape(s,i),l=v.sizeFromShape(o);v.assert(i===l,()=>`The new shape (${o}) has ${l} elements and the old shape (${a.shape}) has ${i} elements. The new shape and old shape must have the same number of elements.`),r.incRef(a.dataId);let u=r.data.get(a.dataId);if(u.complexTensorInfos!=null){let d=u.complexTensorInfos.real,h=u.complexTensorInfos.imag;d.shape=o,h.shape=o}return{dataId:a.dataId,shape:o,dtype:a.dtype}}var wX={kernelName:ml,backendName:"cpu",kernelFunc:Mt};function Y8(e){let{inputs:t,backend:r,attrs:n}=e,{a,b:s}=t,{transposeA:i,transposeB:o}=n;Ce([a,s],"matMul");let l=a.shape.length,u=s.shape.length,d=i?a.shape[l-2]:a.shape[l-1],h=o?s.shape[u-1]:s.shape[u-2],p=i?a.shape[l-1]:a.shape[l-2],c=o?s.shape[u-2]:s.shape[u-1],f=a.shape.slice(0,-2),m=s.shape.slice(0,-2),g=v.sizeFromShape(f),y=v.sizeFromShape(m),A=ji.assertAndGetBroadcastShape(a.shape.slice(0,-2),s.shape.slice(0,-2)).concat([p,c]);v.assert(d===h,()=>`Error in matMul: inner shapes (${d}) and (${h}) of Tensors with shapes ${a.shape} and ${s.shape} and transposeA=${i} and transposeB=${o} must match.`);let x=i?[g,d,p]:[g,p,d],b=o?[y,c,h]:[y,h,c],w=Mt({inputs:{x:a},backend:r,attrs:{shape:x}}),S=Mt({inputs:{x:s},backend:r,attrs:{shape:b}}),N=i?w.shape[1]:w.shape[2],E=i?w.shape[2]:w.shape[1],R=o?S.shape[1]:S.shape[2],z=Math.max(g,y),M=r.data.get(w.dataId).values,I=r.data.get(S.dataId).values,_=v.computeStrides(w.shape),O=v.computeStrides(S.shape),[q,K,D]=i?[_[0],1,_[1]]:[_[0],_[1],1],[Q,V,re]=o?[1,O[1],O[0]]:[O[1],1,O[0]],Y=E*R,ie=We([z,E,R],w.dtype),J=ie.values,ae=r.blockSize;for(let de=0;de<z;de++)for(let xe=0;xe<E;xe+=ae)for(let be=0;be<R;be+=ae)for(let Ee=0;Ee<N;Ee+=ae){let $e=Math.min(xe+ae,E),De=Math.min(be+ae,R),Be=Math.min(Ee+ae,N);for(let Xe=xe;Xe<$e;Xe++)for(let ot=be;ot<De;ot++){let pt=0;for(let ht=Ee;ht<Be;ht++){let Fe=Math.min(de,g-1)*q,wt=Math.min(de,y-1)*re,yt=M[Fe+Xe*K+ht*D],Pr=I[ht*Q+ot*V+wt];pt+=yt*Pr}J[de*Y+(Xe*R+ot)]+=pt}}return r.disposeIntermediateTensorInfo(w),r.disposeIntermediateTensorInfo(S),r.makeTensorInfo(A,ie.dtype,ie.values)}var kX={kernelName:Qs,backendName:"cpu",kernelFunc:Y8};function IX(e){let{inputs:t,backend:r,attrs:n}=e,{a,b:s,bias:i,preluActivationWeights:o}=t,{transposeA:l,transposeB:u,activation:d,leakyreluAlpha:h}=n,p,c,f,m=[];p=Y8({inputs:{a,b:s},attrs:{transposeA:l,transposeB:u},backend:r}),i&&(c=zh({inputs:{a:p,b:i},backend:r}),m.push(p),p=c),d&&(f=l5(r,p,d,o,h),m.push(p),p=f);for(let g of m)r.disposeIntermediateTensorInfo(g);return p}var SX={kernelName:Ps,backendName:"cpu",kernelFunc:IX},CX=gt(Du,e=>Math.acos(e)),TX={kernelName:Du,backendName:"cpu",kernelFunc:CX},NX=gt(Lu,e=>Math.acosh(e)),EX={kernelName:Lu,backendName:"cpu",kernelFunc:NX};function RX(e){let{inputs:t,backend:r}=e,n=t;Ce(t,"addN");let a=n.map(o=>r.data.get(o.dataId).values),s=We(n[0].shape,n[0].dtype),i=s.values;for(let o=0;o<n.length;o++){let l=a[o];for(let u=0;u<i.length;u++)i[u]+=l[u]}return r.makeTensorInfo(s.shape,s.dtype,s.values)}var $X={kernelName:Zs,backendName:"cpu",kernelFunc:RX};function MX(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{axis:s,keepDims:i}=n;Ce(a,"all");let o=v.parseAxisParam(s,a.shape),l=o,u=C.getAxesPermutation(l,a.shape.length),d=a;u!=null&&(d=sn({inputs:{x:a},backend:r,attrs:{perm:u}}),l=C.getInnerMostAxes(l.length,a.shape.length)),C.assertAxesAreInnerMostDims("all",l,d.shape.length);let[h,p]=C.computeOutAndReduceShapes(d.shape,l),c=v.sizeFromShape(p),f=v.makeZerosTypedArray(v.sizeFromShape(h),d.dtype),m=r.data.get(d.dataId).values;for(let y=0;y<f.length;++y){let A=y*c,x=m[A];for(let b=0;b<c;++b){let w=m[A+b];x=x&&w}f[y]=x}u!=null&&r.disposeIntermediateTensorInfo(d);let g=r.makeTensorInfo(h,d.dtype,f);if(i){let y=C.expandShapeToKeepDim(h,o),A=Mt({inputs:{x:g},backend:r,attrs:{shape:y}});return r.disposeIntermediateTensorInfo(g),A}return g}var FX={kernelName:Bu,backendName:"cpu",kernelFunc:MX};function PX(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{axis:s,keepDims:i}=n;Ce(a,"any");let o=v.parseAxisParam(s,a.shape),l=o,u=C.getAxesPermutation(l,a.shape.length),d=a;u!=null&&(d=sn({inputs:{x:a},backend:r,attrs:{perm:u}}),l=C.getInnerMostAxes(l.length,a.shape.length)),C.assertAxesAreInnerMostDims("any",l,d.shape.length);let[h,p]=C.computeOutAndReduceShapes(d.shape,l),c=v.sizeFromShape(p),f=v.makeZerosTypedArray(v.sizeFromShape(h),d.dtype),m=r.data.get(d.dataId).values;for(let y=0;y<f.length;++y){let A=y*c,x=m[A];for(let b=0;b<c;++b){let w=m[A+b];x=x||w}f[y]=x}u!=null&&r.disposeIntermediateTensorInfo(d);let g=r.makeTensorInfo(h,d.dtype,f);if(i){let y=C.expandShapeToKeepDim(h,o),A=Mt({inputs:{x:g},backend:r,attrs:{shape:y}});return r.disposeIntermediateTensorInfo(g),A}return g}var _X={kernelName:Wu,backendName:"cpu",kernelFunc:PX};function zX(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{axis:s}=n;Ce(a,"argMax");let i=v.parseAxisParam(s,a.shape),o=C.getAxesPermutation(i,a.shape.length),l=a,u=[];o!=null&&(l=sn({inputs:{x:a},backend:r,attrs:{perm:o}}),u.push(l),i=C.getInnerMostAxes(i.length,l.shape.length)),i=[i[0]],C.assertAxesAreInnerMostDims("argMax",i,l.shape.length);let[d,h]=C.computeOutAndReduceShapes(l.shape,i),p=v.sizeFromShape(d),c=v.makeZerosTypedArray(p,"int32"),f=v.sizeFromShape(h),m=r.data.get(l.dataId).values;for(let g=0;g<c.length;++g){let y=g*f,A=m[y],x=0;for(let b=0;b<f;++b){let w=m[y+b];w>A&&(A=w,x=b)}c[g]=x}return u.forEach(g=>r.disposeIntermediateTensorInfo(g)),r.makeTensorInfo(d,"int32",c)}var OX={kernelName:Ys,backendName:"cpu",kernelFunc:zX};function DX(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{axis:s}=n;Ce(a,"argMin");let i=v.parseAxisParam(s,a.shape),o=C.getAxesPermutation(i,a.shape.length),l=a,u=[];o!=null&&(l=sn({inputs:{x:a},backend:r,attrs:{perm:o}}),u.push(l),i=C.getInnerMostAxes(i.length,l.shape.length)),i=[i[0]],C.assertAxesAreInnerMostDims("argMin",i,l.shape.length);let[d,h]=C.computeOutAndReduceShapes(l.shape,i),p=v.sizeFromShape(d),c=v.makeZerosTypedArray(p,"int32"),f=v.sizeFromShape(h),m=r.data.get(l.dataId).values;for(let g=0;g<c.length;++g){let y=g*f,A=m[y],x=0;for(let b=0;b<f;++b){let w=m[y+b];w<A&&(A=w,x=b)}c[g]=x}return u.forEach(g=>r.disposeIntermediateTensorInfo(g)),r.makeTensorInfo(d,"int32",c)}var LX={kernelName:Vu,backendName:"cpu",kernelFunc:DX},BX=gt(Uu,e=>Math.asin(e)),WX={kernelName:Uu,backendName:"cpu",kernelFunc:BX},VX=gt(Gu,e=>Math.asinh(e)),UX={kernelName:Gu,backendName:"cpu",kernelFunc:VX},GX=gt(ju,e=>Math.atan(e)),jX={kernelName:ju,backendName:"cpu",kernelFunc:GX},HX=Qt((e,t)=>Math.atan2(e,t)),qX=vr(qu,HX),XX={kernelName:qu,backendName:"cpu",kernelFunc:qX},KX=gt(Hu,e=>Math.atanh(e)),ZX={kernelName:Hu,backendName:"cpu",kernelFunc:KX};function u5(e,t,r,n,a,s){let i=a.strideHeight,o=a.strideWidth,l=a.dilationHeight,u=a.dilationWidth,d=a.effectiveFilterHeight,h=a.effectiveFilterWidth,p=a.padInfo.top,c=a.padInfo.left,f=s==="max"?Number.NEGATIVE_INFINITY:Number.POSITIVE_INFINITY,m=We(a.outShape,r),g=m.values,y=a.outShape[1]*a.outShape[2]*a.outShape[3],A=a.outShape[2]*a.outShape[3],x=a.outShape[3];for(let b=0;b<a.batchSize;++b){let w=b*y,S=b*n[0];for(let N=0;N<a.inChannels;++N)for(let E=0;E<a.outHeight;++E){let R=E*i-p,z=Math.max(0,R),M=Math.min(a.inHeight,d+R),I=w+E*A;for(let _=0;_<a.outWidth;++_){let O=_*o-c,q=Math.max(0,O),K=Math.min(a.inWidth,h+O),D=f,Q=0,V=0;for(let Y=z;Y<M;Y+=l){let ie=S+Y*n[1];for(let J=q;J<K;J+=u){let ae=ie+J*n[2],de=e[ae+N];s==="max"&&de>D?D=de:s==="avg"&&(Q+=de,V++)}if(isNaN(D))break}let re=I+_*x+N;g[re]=s==="avg"?Q/V:D}}}return m}function J8(e,t,r,n,a=!1,s=!1){let i=We(n.outShape,"int32"),o=n.strideHeight,l=n.strideWidth,u=n.dilationHeight,d=n.dilationWidth,h=n.effectiveFilterHeight,p=n.effectiveFilterWidth,c=n.padInfo.top,f=n.padInfo.left,m=We(t,r,e);for(let g=0;g<n.batchSize;++g)for(let y=0;y<n.inChannels;++y)for(let A=0;A<n.outHeight;++A){let x=A*o-c,b=x;for(;b<0;)b+=u;let w=Math.min(n.inHeight,h+x);for(let S=0;S<n.outWidth;++S){let N=S*l-f,E=N;for(;E<0;)E+=d;let R=Math.min(n.inWidth,p+N),z=Number.NEGATIVE_INFINITY,M=-1;for(let I=b;I<w;I+=u){let _=I-x;for(let O=E;O<R;O+=d){let q=O-N,K=m.get(g,I,O,y);K>z&&(z=K,a?M=s?((g*n.inHeight+I)*n.inWidth+O)*n.inChannels+y:(I*n.inWidth+O)*n.inChannels+y:M=_*p+q)}}i.set(M,g,A,S,y)}}return i}function Q8(e,t,r,n,a,s){let i=a.strideDepth,o=a.strideHeight,l=a.strideWidth,u=a.dilationDepth,d=a.dilationHeight,h=a.dilationWidth,p=a.effectiveFilterDepth,c=a.effectiveFilterHeight,f=a.effectiveFilterWidth,m=a.padInfo.front,g=a.padInfo.top,y=a.padInfo.left,A=s==="max"?Number.NEGATIVE_INFINITY:Number.POSITIVE_INFINITY,x=We(a.outShape,r),b=x.values,w=a.outShape[1]*a.outShape[2]*a.outShape[3]*a.outShape[4],S=a.outShape[2]*a.outShape[3]*a.outShape[4],N=a.outShape[3]*a.outShape[4],E=a.outShape[4];for(let R=0;R<a.batchSize;++R){let z=R*w,M=R*n[0];for(let I=0;I<a.inChannels;++I)for(let _=0;_<a.outDepth;++_){let O=_*i-m,q=O;for(;q<0;)q+=u;let K=Math.min(a.inDepth,p+O),D=z+_*S;for(let Q=0;Q<a.outHeight;++Q){let V=Q*o-g,re=V;for(;re<0;)re+=d;let Y=Math.min(a.inHeight,c+V),ie=D+Q*N;for(let J=0;J<a.outWidth;++J){let ae=J*l-y,de=ae;for(;de<0;)de+=h;let xe=Math.min(a.inWidth,f+ae),be=ie+J*E,Ee=A,$e=0,De=0;for(let Xe=q;Xe<K;Xe+=u){let ot=M+Xe*n[1];for(let pt=re;pt<Y;pt+=d){let ht=ot+pt*n[2];for(let Fe=de;Fe<xe;Fe+=h){let wt=ht+Fe*n[3],yt=e[wt+I];if(s==="max"&&yt>Ee?Ee=yt:s==="avg"&&($e+=yt,De++),isNaN(Ee))break}if(isNaN(Ee))break}if(isNaN(Ee))break}let Be=be+I;b[Be]=s==="avg"?$e/De:Ee}}}}return x}function YX(e,t){let r=We(t.outShape,"int32"),n=t.strideDepth,a=t.strideHeight,s=t.strideWidth,i=t.dilationDepth,o=t.dilationHeight,l=t.dilationWidth,u=t.effectiveFilterDepth,d=t.effectiveFilterHeight,h=t.effectiveFilterWidth,p=t.padInfo.front,c=t.padInfo.top,f=t.padInfo.left;for(let m=0;m<t.batchSize;++m)for(let g=0;g<t.inChannels;++g)for(let y=0;y<t.outDepth;++y){let A=y*n-p,x=A;for(;x<0;)x+=i;let b=Math.min(t.inDepth,u+A);for(let w=0;w<t.outHeight;++w){let S=w*a-c,N=S;for(;N<0;)N+=o;let E=Math.min(t.inHeight,d+S);for(let R=0;R<t.outWidth;++R){let z=R*s-f,M=z;for(;M<0;)M+=l;let I=Math.min(t.inWidth,h+z),_=Number.NEGATIVE_INFINITY,O=-1;for(let q=x;q<b;q+=i){let K=q-A;for(let D=N;D<E;D+=o){let Q=D-S;for(let V=M;V<I;V+=l){let re=V-z,Y=e.get(m,q,D,V,g);Y>=_&&(_=Y,O=K*d*h+Q*d+re)}}}r.set(O,m,y,w,R,g)}}}return r}function JX(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t;Ce(a,"avgPool");let{filterSize:s,strides:i,pad:o,dimRoundingMode:l}=n,u=1;v.assert(C.eitherStridesOrDilationsAreOne(i,u),()=>`Error in avgPool: Either strides or dilations must be 1. Got strides ${i} and dilations '${u}'`);let d=C.computePool2DInfo(a.shape,s,i,u,o,l),h;if(d.filterWidth===1&&d.filterHeight===1&&v.arraysEqual(d.inShape,d.outShape))h=_a({inputs:{x:a},backend:r});else{let p=r.data.get(a.dataId).values,c=v.computeStrides(a.shape),f=u5(p,a.shape,a.dtype,c,d,"avg");h=r.makeTensorInfo(d.outShape,a.dtype,f.values)}return h}var QX={kernelName:Js,backendName:"cpu",kernelFunc:JX};function eK(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{filterSize:s,strides:i,pad:o,dimRoundingMode:l,dataFormat:u}=n;Ce(a,"avgPool3d");let d=C.computePool3DInfo(a.shape,s,i,1,o,l,u),h=r.data.get(a.dataId).values,p=Q8(h,a.shape,a.dtype,v.computeStrides(a.shape),d,"avg");return r.makeTensorInfo(p.shape,"float32",p.values)}var tK={kernelName:Yp,backendName:"cpu",kernelFunc:eK};function rK(e){let{inputs:t,backend:r,attrs:n}=e,{dy:a,input:s}=t,{filterSize:i,strides:o,pad:l,dimRoundingMode:u}=n;Ce([a,s],"avgPool3DGrad");let d=C.computePool3DInfo(s.shape,i,o,1,l,u),h=d.strideDepth,p=d.strideHeight,c=d.strideWidth,f=d.filterDepth,m=d.filterHeight,g=d.filterWidth,y=d.dilationDepth,A=d.dilationHeight,x=d.dilationWidth,b=d.effectiveFilterDepth,w=d.effectiveFilterHeight,S=d.effectiveFilterWidth,N=b-1-d.padInfo.front,E=S-1-d.padInfo.left,R=w-1-d.padInfo.top,z=We(s.shape,"float32"),M=1/(f*m*g),I=r.bufferSync(a);for(let _=0;_<d.batchSize;++_)for(let O=0;O<d.inChannels;++O)for(let q=0;q<d.inDepth;++q)for(let K=0;K<d.inHeight;++K)for(let D=0;D<d.inWidth;++D){let Q=q-N,V=K-R,re=D-E,Y=0;for(let ie=0;ie<b;ie+=y){let J=(Q+ie)/h;if(!(J<0||J>=d.outDepth||Math.floor(J)!==J))for(let ae=0;ae<w;ae+=A){let de=(V+ae)/p;if(!(de<0||de>=d.outHeight||Math.floor(de)!==de))for(let xe=0;xe<S;xe+=x){let be=(re+xe)/c;be<0||be>=d.outWidth||Math.floor(be)!==be||(Y+=I.get(_,J,de,be,O))}}}z.set(Y*M,_,q,K,D,O)}return r.makeTensorInfo(z.shape,z.dtype,z.values)}var nK={kernelName:Q0,backendName:"cpu",kernelFunc:rK};function aK(e){let{inputs:t,backend:r,attrs:n}=e,{dy:a,input:s}=t,i=s;Ce([a,s],"avgPoolGrad");let{filterSize:o,strides:l,pad:u}=n,d=C.computePool2DInfo(i.shape,o,l,1,u),h=d.strideHeight,p=d.strideWidth,c=d.filterHeight,f=d.filterWidth,m=d.dilationHeight,g=d.dilationWidth,y=d.effectiveFilterHeight,A=d.effectiveFilterWidth,x=A-1-d.padInfo.left,b=y-1-d.padInfo.top,w=We(i.shape,"float32"),S=1/(c*f),N=r.data.get(a.dataId).values,E=We(a.shape,"float32",N);for(let R=0;R<d.batchSize;++R)for(let z=0;z<d.inChannels;++z)for(let M=0;M<d.inHeight;++M)for(let I=0;I<d.inWidth;++I){let _=M-b,O=I-x,q=0;for(let K=0;K<y;K+=m){let D=(_+K)/h;if(!(D<0||D>=d.outHeight||Math.floor(D)!==D))for(let Q=0;Q<A;Q+=g){let V=(O+Q)/p;V<0||V>=d.outWidth||Math.floor(V)!==V||(q+=E.get(R,D,V,z))}}w.set(q*S,R,M,I,z)}return r.makeTensorInfo(w.shape,w.dtype,w.values)}var sK={kernelName:J0,backendName:"cpu",kernelFunc:aK};function iK(e){let{inputs:t,backend:r,attrs:n}=e,{x:a,scale:s,offset:i,mean:o,variance:l}=t;v.assert(o.shape.length===l.shape.length,()=>"Batch normalization gradient requires mean and variance to have equal ranks."),v.assert(i==null||o.shape.length===i.shape.length,()=>"Batch normalization gradient requires mean and offset to have equal ranks."),v.assert(s==null||o.shape.length===s.shape.length,()=>"Batch normalization gradient requires mean and scale to have equal ranks."),Ce([a,o,l,s,i],"batchNorm");let{varianceEpsilon:u}=n;u==null&&(u=.001);let d=r.data.get(a.dataId).values,h=r.data.get(o.dataId).values,p=r.data.get(l.dataId).values,c=s?r.data.get(s.dataId).values:new Float32Array([1]),f=i?r.data.get(i.dataId).values:new Float32Array([0]),m=new Float32Array(d.length),g=f.length,y=c.length,A=p.length,x=h.length,b=0,w=0,S=0,N=0;for(let E=0;E<d.length;++E)m[E]=f[b++]+(d[E]-h[w++])*c[S++]/Math.sqrt(p[N++]+u),b>=g&&(b=0),w>=x&&(w=0),S>=y&&(S=0),N>=A&&(N=0);return r.makeTensorInfo(a.shape,a.dtype,m)}var oK={kernelName:ci,backendName:"cpu",kernelFunc:iK};function lK(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{blockShape:s,crops:i}=n;Ce([a],"batchToSpaceND");let o=s.reduce((y,A)=>y*A),l=C.getReshaped(a.shape,s,o),u=C.getPermuted(l.length,s.length),d=C.getReshapedPermuted(a.shape,s,o),h=C.getSliceBeginCoords(i,s.length),p=C.getSliceSize(d,i,s.length),c=Mt({inputs:{x:a},backend:r,attrs:{shape:l}}),f=sn({inputs:{x:c},backend:r,attrs:{perm:u}}),m=Mt({inputs:{x:f},backend:r,attrs:{shape:d}}),g=Bo({inputs:{x:m},backend:r,attrs:{begin:h,size:p}});return r.disposeIntermediateTensorInfo(c),r.disposeIntermediateTensorInfo(f),r.disposeIntermediateTensorInfo(m),g}var uK={kernelName:qo,backendName:"cpu",kernelFunc:lK};function dK(e){let{inputs:t,backend:r,attrs:n}=e,{x:a,weights:s}=t,{size:i}=n,o=r.data.get(a.dataId).values,l=r.data.get(s.dataId).values,u=t5(o,l,s.dtype,s.shape,i);return r.makeTensorInfo([i],s.dtype,u)}var pK={kernelName:ef,backendName:"cpu",kernelFunc:dK};function hK(e){let{inputs:t,backend:r}=e,{s0:n,s1:a}=t,s=r.data.get(n.dataId).values,i=r.data.get(a.dataId).values,o=C.assertAndGetBroadcastShape(Array.from(s),Array.from(i));return r.makeTensorInfo([o.length],"int32",Int32Array.from(o))}var cK={kernelName:tf,backendName:"cpu",kernelFunc:hK},fK=gt(Qa,(e,t)=>{let r=t;return e>r.clipValueMax?r.clipValueMax:e<r.clipValueMin?r.clipValueMin:e}),mK={kernelName:Qa,backendName:"cpu",kernelFunc:fK},gK=e=>{let{x:t}=e.inputs,r=e.backend,n=new Float32Array(v.sizeFromShape(t.shape)),a=r.data.get(t.dataId),s=a.complexTensorInfos.real,i=a.complexTensorInfos.imag,o=r.data.get(s.dataId).values,l=r.data.get(i.dataId).values;for(let u=0;u<o.length;u++){let d=o[u],h=l[u];n[u]=Math.hypot(d,h)}return r.makeOutput(n,t.shape,"float32")},yK={kernelName:Qp,backendName:"cpu",kernelFunc:gK};function Fu(e){let{inputs:t,backend:r}=e,{input:n}=t,a=r.data.get(n.dataId).complexTensorInfos.imag,s=r.data.get(a.dataId).values;return r.makeTensorInfo(a.shape,a.dtype,s)}var AK={kernelName:nh,backendName:"cpu",kernelFunc:Fu};function Pu(e){let{inputs:t,backend:r,attrs:n}=e,{axis:a}=n,s=v.parseAxisParam(a,t[0].shape)[0],i=C.computeOutShape(t.map(m=>m.shape),s);if(v.sizeFromShape(i)===0)return r.makeTensorInfo(i,t[0].dtype,[]);let o=t.filter(m=>v.sizeFromShape(m.shape)>0);if(o.length===1)return _a({inputs:{x:o[0]},backend:r});let l=o.map(m=>m.shape);if(C.assertParamsConsistent(l,s),o[0].dtype==="complex64"){let m=o.map(b=>Lo({inputs:{input:b},backend:r})),g=o.map(b=>Fu({inputs:{input:b},backend:r})),y=Pu({inputs:m,backend:r,attrs:{axis:s}}),A=Pu({inputs:g,backend:r,attrs:{axis:s}}),x=pn({inputs:{real:y,imag:A},backend:r});return m.forEach(b=>r.disposeIntermediateTensorInfo(b)),g.forEach(b=>r.disposeIntermediateTensorInfo(b)),r.disposeIntermediateTensorInfo(y),r.disposeIntermediateTensorInfo(A),x}let u=o.map(m=>{let g=v.sizeFromShape(m.shape.slice(s));return Mt({inputs:{x:m},backend:r,attrs:{shape:[-1,g]}})}),d=u.map(m=>({vals:r.data.get(m.dataId).values,shape:m.shape}));i=C.computeOutShape(u.map(m=>m.shape),1);let h=u[0].shape[0]===1,p=r5(d,i,t[0].dtype,h),c=C.computeOutShape(o.map(m=>m.shape),s),f=r.makeTensorInfo(c,t[0].dtype,p);return u.forEach(m=>r.disposeIntermediateTensorInfo(m)),f}var xK={kernelName:Xo,backendName:"cpu",kernelFunc:Pu};function eI(e){let{inputs:t,backend:r,attrs:n}=e,{x:a,filter:s}=t,{strides:i,pad:o,dataFormat:l,dilations:u,dimRoundingMode:d}=n;Ce([a,s],"conv2d");let h=C.convertConv2DDataFormat(l),p=C.computeConv2DInfo(a.shape,s.shape,i,u,o,d,!1,h),c=p.filterHeight,f=p.filterWidth,m=p.dilationHeight,g=p.dilationWidth,y=p.padInfo.left,A=p.padInfo.top,x=p.dataFormat==="channelsLast",b=new ir(p.outShape,a.dtype),w=v.computeStrides(a.shape),S=v.computeStrides(s.shape),N=w[0],E=x?w[1]:w[2],R=x?w[2]:1,z=x?1:w[1],M=b.strides[0],I=x?b.strides[1]:b.strides[2],_=x?b.strides[2]:1,O=x?1:b.strides[1],q=r.data.get(a.dataId).values,K=r.data.get(s.dataId).values,D=b.values;for(let Q=0;Q<p.batchSize;++Q){let V=Q*N,re=Q*M;for(let Y=0;Y<p.outHeight;++Y){let ie=re+Y*I,J=Y*p.strideHeight-A;for(let ae=0;ae<c;++ae){let de=J+ae*m;if(de<0||de>=p.inHeight)continue;let xe=ae*S[0],be=V+de*E;for(let Ee=0;Ee<p.outWidth;++Ee){let $e=ie+Ee*_,De=Ee*p.strideWidth-y;for(let Be=0;Be<f;++Be){let Xe=De+Be*g;if(Xe<0||Xe>=p.inWidth)continue;let ot=xe+Be*S[1],pt=be+Xe*R,ht=ot;for(let Fe=0;Fe<p.inChannels;++Fe){let wt=q[pt+Fe*z];for(let yt=0;yt<p.outChannels;++yt)D[$e+yt*O]+=wt*K[ht+yt];ht+=p.outChannels}}}}}}return r.makeTensorInfo(b.shape,b.dtype,D)}var bK={kernelName:ri,backendName:"cpu",kernelFunc:eI};function vK(e){let{inputs:t,backend:r,attrs:n}=e,{x:a,dy:s}=t,{strides:i,pad:o,dataFormat:l,dimRoundingMode:u,filterShape:d}=n;Ce([a,s],"conv2dBackpropFilter");let h=C.convertConv2DDataFormat(l),p=C.computeConv2DInfo(a.shape,d,i,1,o,u,!1,h),{strideHeight:c,strideWidth:f,filterHeight:m,filterWidth:g}=p,y=p.dataFormat==="channelsLast",A=new ir(p.filterShape,"float32"),x=p.padInfo.left,b=p.padInfo.top,w=r.data.get(a.dataId).values,S=r.data.get(s.dataId).values,N=new ir(a.shape,a.dtype,w),E=new ir(s.shape,s.dtype,S);for(let R=0;R<m;++R){let z=Math.max(0,Math.ceil((b-R)/c)),M=Math.min(p.outHeight,(p.inHeight+b-R)/c);for(let I=0;I<g;++I){let _=Math.max(0,Math.ceil((x-I)/f)),O=Math.min(p.outWidth,(p.inWidth+x-I)/f);for(let q=0;q<p.inChannels;++q)for(let K=0;K<p.outChannels;++K){let D=0;for(let Q=0;Q<p.batchSize;++Q)for(let V=z;V<M;++V){let re=R+V*c-b;for(let Y=_;Y<O;++Y){let ie=I+Y*f-x;y?D+=N.get(Q,re,ie,q)*E.get(Q,V,Y,K):D+=N.get(Q,q,re,ie)*E.get(Q,K,V,Y)}}A.set(D,R,I,q,K)}}}return r.makeTensorInfo(A.shape,A.dtype,A.values)}var wK={kernelName:rf,backendName:"cpu",kernelFunc:vK};function kK(e){let{inputs:t,backend:r,attrs:n}=e,{dy:a,filter:s}=t,{inputShape:i,strides:o,pad:l,dataFormat:u,dimRoundingMode:d}=n;Ce([a,s],"conv2dBackpropInput");let h=v.computeStrides(s.shape),p=v.computeStrides(a.shape),c=C.convertConv2DDataFormat(u),f=C.computeConv2DInfo(i,s.shape,o,1,l,d,!1,c),m=new ir(f.inShape,"float32"),g=m.values,y=r.data.get(a.dataId).values,A=r.data.get(s.dataId).values,[x,b,w]=h,{batchSize:S,filterHeight:N,filterWidth:E,inChannels:R,inHeight:z,inWidth:M,outChannels:I,outHeight:_,outWidth:O,strideHeight:q,strideWidth:K}=f;c=f.dataFormat;let D=N-1-f.padInfo.top,Q=E-1-f.padInfo.left,V=c==="channelsLast",re=m.strides[0],Y=V?m.strides[1]:m.strides[2],ie=V?m.strides[2]:1,J=V?1:m.strides[1],ae=p[0],de=V?p[1]:p[2],xe=V?p[2]:1,be=V?1:p[1];for(let Ee=0;Ee<S;++Ee)for(let $e=0;$e<R;++$e)for(let De=0;De<z;++De){let Be=De-D,Xe=Math.max(0,Math.ceil(Be/q)),ot=Math.min(_,(N+Be)/q);for(let pt=0;pt<M;++pt){let ht=pt-Q,Fe=Math.max(0,Math.ceil(ht/K)),wt=Math.min(O,(E+ht)/K),yt=0;for(let cr=Xe;cr<ot;++cr){let Jr=cr*q-Be;for(let rr=Fe;rr<wt;++rr){let fr=rr*K-ht,ta=ae*Ee+de*cr+xe*rr,Qr=x*(N-1-Jr)+b*(E-1-fr)+w*$e;for(let nr=0;nr<I;++nr){let kn=y[ta+be*nr],In=A[Qr+nr];yt+=kn*In}}}let Pr=re*Ee+Y*De+ie*pt+J*$e;g[Pr]=yt}}return r.makeTensorInfo(m.shape,m.dtype,m.values)}var IK={kernelName:ni,backendName:"cpu",kernelFunc:kK};function SK(e){let{inputs:t,backend:r,attrs:n}=e,{x:a,filter:s}=t,{strides:i,pad:o,dilations:l}=n;Ce([a,s],"conv3d");let u=C.computeConv3DInfo(a.shape,s.shape,i,l,o),{filterDepth:d,filterHeight:h,filterWidth:p,dilationDepth:c,dilationHeight:f,dilationWidth:m,padInfo:g}=u,y=g.front,A=g.left,x=g.top,b=new ir(u.outShape,a.dtype),w=r.data.get(a.dataId).values,S=r.data.get(s.dataId).values,N=b.values,E=v.computeStrides(a.shape),R=v.computeStrides(s.shape);for(let z=0;z<u.batchSize;++z){let M=z*E[0],I=z*b.strides[0];for(let _=0;_<u.outDepth;++_){let O=I+_*b.strides[1],q=_*u.strideDepth-y;for(let K=0;K<d;++K){let D=q+K*c;if(D<0||D>=u.inDepth)continue;let Q=K*R[0],V=M+D*E[1];for(let re=0;re<u.outHeight;++re){let Y=O+re*b.strides[2],ie=re*u.strideHeight-x;for(let J=0;J<h;++J){let ae=ie+J*f;if(ae<0||ae>=u.inHeight)continue;let de=Q+J*R[1],xe=V+ae*E[2];for(let be=0;be<u.outWidth;++be){let Ee=Y+be*u.outChannels,$e=be*u.strideWidth-A;for(let De=0;De<p;++De){let Be=$e+De*m;if(Be<0||Be>=u.inWidth)continue;let Xe=de+De*R[2],ot=xe+Be*u.inChannels,pt=Xe;for(let ht=0;ht<u.inChannels;++ht){let Fe=w[ot+ht];for(let wt=0;wt<u.outChannels;++wt)N[Ee+wt]+=Fe*S[pt+wt];pt+=u.outChannels}}}}}}}}return r.makeTensorInfo(b.shape,b.dtype,b.values)}var CK={kernelName:eh,backendName:"cpu",kernelFunc:SK};function TK(e){let{inputs:t,backend:r,attrs:n}=e,{x:a,dy:s}=t,{strides:i,pad:o,filterShape:l}=n;Ce([a,s],"conv3dBackpropFilterV2");let u=v.computeStrides(a.shape),d=v.computeStrides(s.shape),h=C.computeConv3DInfo(a.shape,l,i,1,o),p=h.strideDepth,c=h.strideHeight,f=h.strideWidth,m=h.filterDepth,g=h.filterHeight,y=h.filterWidth,A=new ir(h.filterShape,"float32"),x=A.values,[b,w,S,N]=A.strides,E=r.data.get(s.dataId).values,[R,z,M,I]=d,_=r.data.get(a.dataId).values,[O,q,K,D]=u,Q=h.padInfo.front,V=h.padInfo.left,re=h.padInfo.top;for(let Y=0;Y<m;++Y){let ie=Math.max(0,Math.ceil((Q-Y)/p)),J=Math.min(h.outDepth,(h.inDepth+Q-Y)/p),ae=Y*b;for(let de=0;de<g;++de){let xe=Math.max(0,Math.ceil((re-de)/c)),be=Math.min(h.outHeight,(h.inHeight+re-de)/c),Ee=de*w+ae;for(let $e=0;$e<y;++$e){let De=Math.max(0,Math.ceil((V-$e)/f)),Be=Math.min(h.outWidth,(h.inWidth+V-$e)/f),Xe=$e*S+Ee;for(let ot=0;ot<h.inChannels;++ot){let pt=ot*N+Xe;for(let ht=0;ht<h.outChannels;++ht){let Fe=0;for(let wt=0;wt<h.batchSize;++wt){let yt=wt*O,Pr=wt*R;for(let cr=ie;cr<J;++cr){let Jr=(Y+cr*p-Q)*q+yt,rr=cr*z+Pr;for(let fr=xe;fr<be;++fr){let ta=(de+fr*c-re)*K+Jr,Qr=fr*M+rr;for(let nr=De;nr<Be;++nr){let kn=($e+nr*f-V)*D+ta,In=nr*I+Qr;Fe+=_[kn+ot]*E[In+ht]}}}}x[pt+ht]=Fe}}}}}return r.makeTensorInfo(A.shape,A.dtype,A.values)}var NK={kernelName:nf,backendName:"cpu",kernelFunc:TK};function EK(e){let{inputs:t,backend:r,attrs:n}=e,{dy:a,filter:s}=t,{pad:i,strides:o,inputShape:l}=n;Ce([a],"conv3dBackpropInputV2");let u=v.computeStrides(a.shape),d=v.computeStrides(s.shape),h=C.computeConv3DInfo(l,s.shape,o,1,i),p=new ir(h.inShape,"float32"),c=p.values,[f,m,g,y]=p.strides,A=r.data.get(a.dataId).values,[x,b,w,S]=u,N=r.data.get(s.dataId).values,[E,R,z,M]=d,{batchSize:I,filterDepth:_,filterHeight:O,filterWidth:q,inChannels:K,inDepth:D,inHeight:Q,inWidth:V,outChannels:re,outDepth:Y,outHeight:ie,outWidth:J,strideDepth:ae,strideHeight:de,strideWidth:xe}=h,be=_-1-h.padInfo.front,Ee=O-1-h.padInfo.top,$e=q-1-h.padInfo.left;for(let De=0;De<I;++De)for(let Be=0;Be<K;++Be)for(let Xe=0;Xe<D;++Xe){let ot=Xe-be,pt=Math.max(0,Math.ceil(ot/ae)),ht=Math.min(Y,(_+ot)/ae);for(let Fe=0;Fe<Q;++Fe){let wt=Fe-Ee,yt=Math.max(0,Math.ceil(wt/de)),Pr=Math.min(ie,(O+wt)/de);for(let cr=0;cr<V;++cr){let Jr=cr-$e,rr=Math.max(0,Math.ceil(Jr/xe)),fr=Math.min(J,(q+Jr)/xe),ta=0;for(let Qr=pt;Qr<ht;++Qr){let nr=Qr*ae-ot;for(let kn=yt;kn<Pr;++kn){let In=kn*de-wt;for(let gs=rr;gs<fr;++gs){let uo=gs*xe-Jr,uc=x*De+b*Qr+w*kn+S*gs,ys=E*(_-1-nr)+R*(O-1-In)+z*(q-1-uo)+M*Be;for(let Ua=0;Ua<re;++Ua){let rp=A[uc+Ua],Zl=N[ys+Ua];ta+=rp*Zl}}}}c[f*De+m*Xe+g*Fe+y*cr+Be]=ta}}}return r.makeTensorInfo(p.shape,p.dtype,p.values)}var RK={kernelName:af,backendName:"cpu",kernelFunc:EK},$K=gt(ai,e=>Math.cos(e)),MK={kernelName:ai,backendName:"cpu",kernelFunc:$K},FK=gt(si,e=>Math.cosh(e)),PK={kernelName:si,backendName:"cpu",kernelFunc:FK};function _K(e){let{inputs:t,backend:r,attrs:n}=e,{image:a,boxes:s,boxInd:i}=t,{cropSize:o,method:l,extrapolationValue:u}=n,[d,h,p,c]=a.shape,f=s.shape[0],[m,g]=o,y=We([f,m,g,c],"float32"),A=r.data.get(s.dataId).values,x=r.data.get(i.dataId).values,b=r.data.get(a.dataId).values,w=v.computeStrides(a.shape),S=v.computeStrides(y.shape);for(let N=0;N<f;N++){let E=N*4,R=A[E],z=A[E+1],M=A[E+2],I=A[E+3],_=x[N];if(_>=d)continue;let O=m>1?(M-R)*(h-1)/(m-1):0,q=g>1?(I-z)*(p-1)/(g-1):0;for(let K=0;K<m;K++){let D=m>1?R*(h-1)+K*O:.5*(R+M)*(h-1);if(D<0||D>h-1){for(let Q=0;Q<g;Q++)for(let V=0;V<c;V++){let re=V+Q*S[2]+K*S[1]+N*S[0];y.values[re]=u}continue}if(l==="bilinear"){let Q=Math.floor(D),V=Math.ceil(D),re=D-Q;for(let Y=0;Y<g;Y++){let ie=g>1?z*(p-1)+Y*q:.5*(z+I)*(p-1);if(ie<0||ie>p-1){for(let xe=0;xe<c;xe++){let be=xe+Y*S[2]+K*S[1]+N*S[0];y.values[be]=u}continue}let J=Math.floor(ie),ae=Math.ceil(ie),de=ie-J;for(let xe=0;xe<c;xe++){let be=xe+J*w[2]+Q*w[1]+_*w[0],Ee=b[be];be=xe+ae*w[2]+Q*w[1]+_*w[0];let $e=b[be];be=xe+J*w[2]+V*w[1]+_*w[0];let De=b[be];be=xe+ae*w[2]+V*w[1]+_*w[0];let Be=b[be],Xe=Ee+($e-Ee)*de,ot=De+(Be-De)*de;be=xe+Y*S[2]+K*S[1]+N*S[0],y.values[be]=Xe+(ot-Xe)*re}}}else for(let Q=0;Q<g;++Q){let V=g>1?z*(p-1)+Q*q:.5*(z+I)*(p-1);if(V<0||V>p-1){for(let ie=0;ie<c;ie++){let J=ie+Q*S[2]+K*S[1]+N*S[0];y.values[J]=u}continue}let re=Math.round(V),Y=Math.round(D);for(let ie=0;ie<c;ie++){let J=ie+re*w[2]+Y*w[1]+_*w[0],ae=ie+Q*S[2]+K*S[1]+N*S[0];y.values[ae]=b[J]}}}}return r.makeTensorInfo(y.shape,y.dtype,y.values)}var zK={kernelName:Zo,backendName:"cpu",kernelFunc:_K};function OK(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{axis:s,exclusive:i,reverse:o}=n;Ce(a,"cumprod");let l=C.getAxesPermutation([s],a.shape.length),u=a;l!=null&&(u=sn({inputs:{x:a},backend:r,attrs:{perm:l}}));let d=C.getInnerMostAxes(1,a.shape.length)[0];if(d!==u.shape.length-1)throw new Error(`backend.cumprod in CPU expects an inner-most axis=${u.shape.length-1} but got axis=${d}`);let h=Nr(u.dtype,"int32"),p=v.makeOnesTypedArray(v.sizeFromShape(u.shape),h),c=r.data.get(u.dataId).values,f=u.shape[u.shape.length-1],m=o?(y,A)=>y+f-A-1:(y,A)=>y+A;for(let y=0;y<c.length;y+=f)for(let A=0;A<f;A++){let x=m(y,A);if(A===0)p[x]=i?1:c[x];else{let b=m(y,A-1);p[x]=i?c[b]*p[b]:c[x]*p[b]}}let g=r.makeTensorInfo(u.shape,h,p);if(l!=null){let y=C.getUndoAxesPermutation(l),A=sn({inputs:{x:g},backend:r,attrs:{perm:y}});return r.disposeIntermediateTensorInfo(g),r.disposeIntermediateTensorInfo(u),A}return g}var DK={kernelName:Ko,backendName:"cpu",kernelFunc:OK};function LK(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{axis:s,exclusive:i,reverse:o}=n;Ce(a,"cumsum");let l=C.getAxesPermutation([s],a.shape.length),u=a;l!=null&&(u=sn({inputs:{x:a},backend:r,attrs:{perm:l}}));let d=C.getInnerMostAxes(1,a.shape.length)[0];if(d!==u.shape.length-1)throw new Error(`backend.cumsum in CPU expects an inner-most axis=${u.shape.length-1} but got axis=${d}`);let h=Nr(u.dtype,"int32"),p=v.makeZerosTypedArray(v.sizeFromShape(u.shape),h),c=r.data.get(u.dataId).values,f=u.shape[u.shape.length-1],m=o?(y,A)=>y+f-A-1:(y,A)=>y+A;for(let y=0;y<c.length;y+=f)for(let A=0;A<f;A++){let x=m(y,A);if(A===0)p[x]=i?0:c[x];else{let b=m(y,A-1);p[x]=i?c[b]+p[b]:c[x]+p[b]}}let g=r.makeTensorInfo(u.shape,h,p);if(l!=null){let y=C.getUndoAxesPermutation(l),A=sn({inputs:{x:g},backend:r,attrs:{perm:y}});return r.disposeIntermediateTensorInfo(g),r.disposeIntermediateTensorInfo(u),A}return g}var BK={kernelName:ii,backendName:"cpu",kernelFunc:LK};function WK(e){let{inputs:t,backend:r,attrs:n}=e,{x:a,weights:s}=t,{size:i,binaryOutput:o}=n;if(a.shape.length===1){let l=r.data.get(a.dataId).values,u=r.data.get(s.dataId).values,d=t5(l,u,s.dtype,s.shape,i);return r.makeTensorInfo([i],s.dtype,d)}else if(a.shape.length===2){let l=r.bufferSync(a),u=r.bufferSync(s),d=d8(l,u,i,o);return r.makeTensorInfo(d.shape,s.dtype,d.values)}throw new Error(`Error in denseBincount: input must be at most rank 2, but got rank${a.shape.length}.`)}var VK={kernelName:sf,backendName:"cpu",kernelFunc:WK};function UK(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{blockSize:s,dataFormat:i}=n;v.assert(i==="NHWC",()=>`Only NHWC dataFormat supported on CPU for depthToSpace. Got ${i}`);let o=a.shape[0],l=a.shape[1],u=a.shape[2],d=a.shape[3],h=l*s,p=u*s,c=d/(s*s),f=r.data.get(a.dataId).values,m=new Float32Array(o*h*p*c),g=0;for(let y=0;y<o;++y)for(let A=0;A<h;++A){let x=Math.floor(A/s),b=A%s;for(let w=0;w<p;++w){let S=Math.floor(w/s),N=w%s,E=(b*s+N)*c;for(let R=0;R<c;++R){let z=R+E+d*(S+u*(x+l*y));m[g++]=f[z]}}}return r.makeTensorInfo([o,h,p,c],a.dtype,m)}var GK={kernelName:Yo,backendName:"cpu",kernelFunc:UK};function tI(e){let{inputs:t,backend:r,attrs:n}=e,{x:a,filter:s}=t,{strides:i,pad:o,dilations:l,dimRoundingMode:u}=n;Ce([a,s],"depthwiseConv2DNative");let d=v.computeStrides(a.shape),h=v.computeStrides(s.shape),p=l;p==null&&(p=[1,1]),v.assert(C.eitherStridesOrDilationsAreOne(i,p),()=>`Error in depthwiseConv2d: Either strides or dilations must be 1. Got strides ${i} and dilations '${p}'`);let c=C.computeConv2DInfo(a.shape,s.shape,i,p,o,u,!0),{filterHeight:f,filterWidth:m,dilationHeight:g,dilationWidth:y,padInfo:A}=c,x=A.left,b=A.top,w=c.outChannels/c.inChannels,S=new ir(c.outShape,a.dtype),N=r.data.get(a.dataId).values,E=r.data.get(s.dataId).values,R=S.values;for(let z=0;z<c.batchSize;++z){let M=z*d[0],I=z*S.strides[0];for(let _=0;_<c.outHeight;++_){let O=I+_*S.strides[1],q=_*c.strideHeight-b;for(let K=0;K<f;++K){let D=q+K*g;if(D<0||D>=c.inHeight)continue;let Q=K*h[0],V=M+D*d[1];for(let re=0;re<c.outWidth;++re){let Y=O+re*S.strides[2],ie=re*c.strideWidth-x;for(let J=0;J<m;++J){let ae=ie+J*y;if(ae<0||ae>=c.inWidth)continue;let de=Q+J*h[1],xe=V+ae*c.inChannels,be=Y,Ee=de;for(let $e=0;$e<c.inChannels;++$e){let De=N[xe+$e];for(let Be=0;Be<w;++Be)R[be+Be]+=De*E[Ee+Be];be+=w,Ee+=w}}}}}}return r.makeTensorInfo(S.shape,S.dtype,S.values)}var jK={kernelName:oi,backendName:"cpu",kernelFunc:tI};function HK(e){let{inputs:t,backend:r,attrs:n}=e,{x:a,dy:s}=t,{strides:i,dilations:o,pad:l,dimRoundingMode:u,filterShape:d}=n;Ce([a,s],"depthwiseConv2dNativeBackpropFilter");let h=C.computeConv2DInfo(a.shape,d,i,o,l,u,!0),{strideHeight:p,strideWidth:c,filterHeight:f,filterWidth:m}=h,g=new ir(h.filterShape,"float32"),y=h.padInfo.left,A=h.padInfo.top,x=h.outChannels/h.inChannels,b=r.data.get(a.dataId).values,w=new ir(a.shape,a.dtype,b),S=r.data.get(s.dataId).values,N=new ir(s.shape,s.dtype,S);for(let E=0;E<f;++E){let R=Math.max(0,Math.ceil((A-E)/p)),z=Math.min(h.outHeight,(h.inHeight+A-E)/p);for(let M=0;M<m;++M){let I=Math.max(0,Math.ceil((y-M)/c)),_=Math.min(h.outWidth,(h.inWidth+y-M)/c);for(let O=0;O<h.outChannels;++O){let q=Math.trunc(O/x),K=O%x,D=0;for(let Q=0;Q<h.batchSize;++Q)for(let V=R;V<z;++V){let re=E+V*p-A;for(let Y=I;Y<_;++Y){let ie=M+Y*c-y;D+=w.get(Q,re,ie,q)*N.get(Q,V,Y,O)}}g.set(D,E,M,q,K)}}}return r.makeTensorInfo(g.shape,g.dtype,g.values)}var qK={kernelName:of,backendName:"cpu",kernelFunc:HK};function XK(e){let{inputs:t,backend:r,attrs:n}=e,{dy:a,filter:s}=t,{strides:i,dilations:o,pad:l,dimRoundingMode:u,inputShape:d}=n;Ce([a,s],"depthwiseConv2DNativeBackpropInput");let h=v.computeStrides(a.shape),p=v.computeStrides(s.shape),c=C.computeConv2DInfo(d,s.shape,i,o,l,u,!0),f=new ir(c.inShape,"float32"),m=f.values,[g,y,A]=f.strides,x=r.data.get(a.dataId).values,[b,w,S]=h,N=r.data.get(s.dataId).values,[E,R,z]=p,{batchSize:M,filterHeight:I,filterWidth:_,inChannels:O,inHeight:q,inWidth:K,outChannels:D,outHeight:Q,outWidth:V,strideHeight:re,strideWidth:Y}=c,ie=I-1-c.padInfo.top,J=_-1-c.padInfo.left,ae=D/O;for(let de=0;de<M;++de)for(let xe=0;xe<O;++xe)for(let be=0;be<q;++be){let Ee=be-ie,$e=Math.max(0,Math.ceil(Ee/re)),De=Math.min(Q,(I+Ee)/re);for(let Be=0;Be<K;++Be){let Xe=Be-J,ot=Math.max(0,Math.ceil(Xe/Y)),pt=Math.min(V,(_+Xe)/Y),ht=0;for(let Fe=$e;Fe<De;++Fe){let wt=Fe*re-Ee;for(let yt=ot;yt<pt;++yt){let Pr=yt*Y-Xe,cr=b*de+w*Fe+S*yt,Jr=E*(I-1-wt)+R*(_-1-Pr)+z*xe;for(let rr=0;rr<ae;++rr){let fr=xe*ae+rr,ta=x[cr+fr],Qr=N[Jr+rr];ht+=ta*Qr}}}m[g*de+y*be+A*Be+xe]=ht}}return r.makeTensorInfo(f.shape,f.dtype,f.values)}var KK={kernelName:lf,backendName:"cpu",kernelFunc:XK};function ZK(e){let{inputs:t,backend:r}=e,{x:n}=t,a=v.sizeFromShape(n.shape),s=r.data.get(n.dataId).values,i=We([a,a],n.dtype),o=i.values;for(let u=0;u<s.length;u++)o[u*a+u]=s[u];let l=[...n.shape,...n.shape];return r.makeTensorInfo(l,i.dtype,i.values)}var YK={kernelName:uf,backendName:"cpu",kernelFunc:ZK},JK={kernelName:th,backendName:"cpu",kernelFunc:({inputs:e,backend:t,attrs:r})=>{let{x:n,filter:a}=e,{strides:s,pad:i,dilations:o}=r,l=t,u=l.data.get(n.dataId).values,d=n.shape.length,h=l.data.get(a.dataId).values,p=a.shape.length,{batchSize:c,inHeight:f,inWidth:m,inChannels:g,outHeight:y,outWidth:A,padInfo:x,strideHeight:b,strideWidth:w,filterHeight:S,filterWidth:N,dilationHeight:E,dilationWidth:R,outShape:z}=C.computeDilation2DInfo(n.shape,a.shape,s,i,"NHWC",o),M=v.sizeFromShape(z),I=z.length,_=v.getArrayFromDType(n.dtype,M);for(let O=0;O<c;++O)for(let q=0;q<y;++q){let K=q*b-x.top;for(let D=0;D<A;++D){let Q=D*w-x.left;for(let V=0;V<g;++V){let re=Number.MIN_SAFE_INTEGER;for(let ie=0;ie<S;++ie){let J=K+ie*E;if(J>=0&&J<f)for(let ae=0;ae<N;++ae){let de=Q+ae*R;if(de>=0&&de<m){let xe=v.locToIndex([O,J,de,V],d,v.computeStrides(n.shape)),be=v.locToIndex([ie,ae,V],p,v.computeStrides(a.shape)),Ee=u[xe]+h[be];Ee>re&&(re=Ee)}}}let Y=v.locToIndex([O,q,D,V],I,v.computeStrides(z));_[Y]=re}}}return{dataId:l.write(v.toTypedArray(_,n.dtype),z,n.dtype),shape:z,dtype:n.dtype}}},QK={kernelName:b0,backendName:"cpu",kernelFunc:({inputs:e,backend:t,attrs:r})=>{let{x:n,filter:a,dy:s}=e,{strides:i,pad:o,dilations:l}=r,u=t,d=v.toNestedArray(n.shape,u.data.get(n.dataId).values),h=v.toNestedArray(a.shape,u.data.get(a.dataId).values),{batchSize:p,inHeight:c,inWidth:f,inChannels:m,outHeight:g,outWidth:y,padInfo:A,strideHeight:x,strideWidth:b,filterHeight:w,filterWidth:S,dilationHeight:N,dilationWidth:E,outShape:R}=C.computeDilation2DInfo(n.shape,a.shape,i,o,"NHWC",l);v.assert(s.rank===R.length,()=>`Error in ${b0}, dy must have the same rank as output ${R.length}, but got ${s.rank}`);let z=v.toNestedArray(R,u.data.get(s.dataId).values),M=v.makeZerosNestedTypedArray(a.shape,a.dtype);for(let I=0;I<p;++I)for(let _=0;_<g;++_){let O=_*x-A.top;for(let q=0;q<y;++q){let K=q*b-A.left;for(let D=0;D<m;++D){let Q=Number.MIN_SAFE_INTEGER,V=0,re=0;for(let Y=0;Y<w;++Y){let ie=O+Y*N;if(ie>=0&&ie<c)for(let J=0;J<S;++J){let ae=K+J*E;if(ae>=0&&ae<f){let de=d[I][ie][ae][D]+h[Y][J][D];de>Q&&(Q=de,V=Y,re=J)}}}M[V][re][D]+=z[I][_][q][D]}}}return{dataId:u.write(v.toTypedArray(M,n.dtype),a.shape,a.dtype),shape:a.shape,dtype:a.dtype}}},eZ={kernelName:x0,backendName:"cpu",kernelFunc:({inputs:e,backend:t,attrs:r})=>{let{x:n,filter:a,dy:s}=e,{strides:i,pad:o,dilations:l}=r,u=t,d=v.toNestedArray(n.shape,u.data.get(n.dataId).values),h=v.toNestedArray(a.shape,u.data.get(a.dataId).values),{batchSize:p,inHeight:c,inWidth:f,inChannels:m,outHeight:g,outWidth:y,padInfo:A,strideHeight:x,strideWidth:b,filterHeight:w,filterWidth:S,dilationHeight:N,dilationWidth:E,outShape:R}=C.computeDilation2DInfo(n.shape,a.shape,i,o,"NHWC",l);v.assert(s.rank===R.length,()=>`Error in ${x0}, dy must have the same rank as output ${R.length}, but got ${s.rank}`);let z=v.toNestedArray(R,u.data.get(s.dataId).values),M=v.makeZerosNestedTypedArray(n.shape,n.dtype);for(let I=0;I<p;++I)for(let _=0;_<g;++_){let O=_*x-A.top;for(let q=0;q<y;++q){let K=q*b-A.left;for(let D=0;D<m;++D){let Q=Number.MIN_SAFE_INTEGER,V=O<0?0:O,re=K<0?0:K;for(let Y=0;Y<w;++Y){let ie=O+Y*N;if(ie>=0&&ie<c)for(let J=0;J<S;++J){let ae=K+J*E;if(ae>=0&&ae<f){let de=d[I][ie][ae][D]+h[Y][J][D];de>Q&&(Q=de,V=ie,re=ae)}}}M[I][V][re][D]+=z[I][_][q][D]}}}return{dataId:u.write(v.toTypedArray(M,n.dtype),n.shape,n.dtype),shape:n.shape,dtype:n.dtype}}};function Oh(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{axis:s,keepDims:i}=n;Ce(a,"sum");let o;a.dtype==="bool"?o=qs({inputs:{x:a},backend:r,attrs:{dtype:"int32"}}):o=_a({inputs:{x:a},backend:r});let l=o.shape.length,u=v.parseAxisParam(s,o.shape),d=C.getAxesPermutation(u,l),h=u,p=o;d!=null&&(p=sn({inputs:{x:o},backend:r,attrs:{perm:d}}),h=C.getInnerMostAxes(h.length,l)),C.assertAxesAreInnerMostDims("sum",h,p.shape.length);let[c,f]=C.computeOutAndReduceShapes(p.shape,h),m=C.upcastType(p.dtype,"int32"),g=B0(r,c,m),y=v.sizeFromShape(f),A=r.data.get(g.dataId).values,x=r.data.get(p.dataId).values;for(let b=0;b<A.length;++b){let w=b*y,S=0;for(let N=0;N<y;++N)S+=x[w+N];A[b]=S}if(i){let b=C.expandShapeToKeepDim(g.shape,u),w=g;g=Mt({inputs:{x:g},backend:r,attrs:{shape:b}}),r.disposeIntermediateTensorInfo(w)}return r.disposeIntermediateTensorInfo(o),d!=null&&r.disposeIntermediateTensorInfo(p),g}var tZ={kernelName:Oi,backendName:"cpu",kernelFunc:Oh};function rZ(e){let{inputs:t,backend:r,attrs:n}=e,{equation:a}=n,s=t,{allDims:i,summedDims:o,idDims:l}=C.decodeEinsumEquation(a,s.length);C.checkEinsumDimSizes(i.length,l,s);let{path:u,steps:d}=C.getEinsumComputePath(o,l),h=d.length,p=null,c=i.length,f=[];for(let m=0;m<h;++m){for(let g of d[m]){let{permutationIndices:y,expandDims:A}=C.getEinsumPermutation(c,l[g]),x;C.isIdentityPermutation(y)?x=s[g]:(x=sn({inputs:{x:s[g]},backend:r,attrs:{perm:y}}),f.push(x));let b=x.shape.slice();for(let w=0;w<A.length;++w)b.splice(A[w],0,1);v.arraysEqual(x.shape,b)||(x=Mt({inputs:{x},backend:r,attrs:{shape:b}}),f.push(x)),p===null?p=x:(p=km({inputs:{a:x,b:p},backend:r}),f.push(p))}m<h-1&&(u[m]>=0&&(p=Oh({inputs:{x:p},backend:r,attrs:{axis:u[m]-(i.length-c),keepDims:!1}}),f.push(p)),c--)}for(let m of f)m!==p&&r.disposeIntermediateTensorInfo(m);return p}var nZ={kernelName:rh,backendName:"cpu",kernelFunc:rZ};function aZ(e){let{inputs:t,backend:r}=e,{dy:n,y:a}=t;Ce([n,a],"eluGrad");let s=new Float32Array(v.sizeFromShape(a.shape)),i=r.data.get(a.dataId).values,o=r.data.get(n.dataId).values;for(let l=0;l<i.length;++l){let u=i[l];u>=1?s[l]=o[l]:s[l]=o[l]*(u+1)}return r.makeTensorInfo(a.shape,"float32",s)}var sZ={kernelName:df,backendName:"cpu",kernelFunc:aZ},iZ=C.ERF_P,oZ=C.ERF_A1,lZ=C.ERF_A2,uZ=C.ERF_A3,dZ=C.ERF_A4,pZ=C.ERF_A5,hZ=gt(Xu,e=>{let t=Math.sign(e),r=Math.abs(e),n=1/(1+iZ*r);return t*(1-((((pZ*n+dZ)*n+uZ)*n+lZ)*n+oZ)*n*Math.exp(-r*r))}),cZ={kernelName:Xu,backendName:"cpu",kernelFunc:hZ};function V0(e){let{inputs:t,backend:r,attrs:n}=e,{input:a}=t,{dim:s}=n,i=a.shape.length,o=a.shape.slice(),l=s;return s<0&&(v.assert(-(i+1)<=s,()=>`Axis must be in the interval [${-(i+1)}, ${i}]`),l=i+s+1),o.splice(l,0,1),Mt({inputs:{x:a},backend:r,attrs:{shape:o}})}var fZ={kernelName:Qo,backendName:"cpu",kernelFunc:V0},mZ=Qt((e,t)=>e/t),d5=vr(li,mZ),Jg={kernelName:li,backendName:"cpu",kernelFunc:d5};function rI(e,t,r){let n=e.shape,a=n[0],s=n[1],i=r.data.get(e.dataId),o=i.complexTensorInfos.real,l=i.complexTensorInfos.imag,u=[a,s],d=v.sizeFromShape(u),h=v.getTypedArrayFromDType("float32",d),p=v.getTypedArrayFromDType("float32",d);for(let g=0;g<a;g++){let y=Bo({inputs:{x:o},backend:r,attrs:{begin:[g,0],size:[1,s]}}),A=Bo({inputs:{x:l},backend:r,attrs:{begin:[g,0],size:[1,s]}}),x=pn({inputs:{real:y,imag:A},backend:r}),{real:b,imag:w}=gZ(x,t,r),S=C.mergeRealAndImagArrays(b,w);for(let N=0;N<s;N++){let E=C.getComplexWithIndex(S,N);h[g*s+N]=E.real,p[g*s+N]=E.imag}r.disposeIntermediateTensorInfo(y),r.disposeIntermediateTensorInfo(A),r.disposeIntermediateTensorInfo(x)}let c=r.makeTensorInfo(u,"float32",h),f=r.makeTensorInfo(u,"float32",p),m=pn({inputs:{real:c,imag:f},backend:r});return r.disposeIntermediateTensorInfo(c),r.disposeIntermediateTensorInfo(f),m}function gZ(e,t,r){let n=v.sizeFromShape(e.shape),a=r.data.get(e.dataId),s=r.data.get(a.complexTensorInfos.real.dataId).values,i=r.data.get(a.complexTensorInfos.imag.dataId).values;if(yZ(n)){let o=Qg(s,i,n,t,r),l=[e.shape[0],e.shape[1]];if(t){let u=r.makeTensorInfo(l,"float32",o.real),d=r.makeTensorInfo(l,"float32",o.imag),h=r.makeTensorInfo([],"float32",v.createScalarValue(n,"float32")),p=_a({inputs:{x:h},backend:r}),c=Jg.kernelFunc({inputs:{a:u,b:h},backend:r}),f=Jg.kernelFunc({inputs:{a:d,b:p},backend:r}),m=r.data.get(c.dataId).values,g=r.data.get(f.dataId).values;return r.disposeIntermediateTensorInfo(u),r.disposeIntermediateTensorInfo(d),r.disposeIntermediateTensorInfo(h),r.disposeIntermediateTensorInfo(p),r.disposeIntermediateTensorInfo(c),r.disposeIntermediateTensorInfo(f),{real:m,imag:g}}return o}else{let o=C.mergeRealAndImagArrays(s,i),l=AZ(o,n,t);return C.splitRealAndImagArrays(l)}}function yZ(e){return(e&e-1)===0}function Qg(e,t,r,n,a){if(r===1)return{real:e,imag:t};let s=C.mergeRealAndImagArrays(e,t),i=r/2,o=C.complexWithEvenIndex(s),l=o.real,u=o.imag,d=[l.length],h=a.makeTensorInfo(d,"float32",l),p=a.makeTensorInfo(d,"float32",u),c=pn({inputs:{real:h,imag:p},backend:a}),f=C.complexWithOddIndex(s),m=f.real,g=f.imag,y=[m.length],A=a.makeTensorInfo(y,"float32",m),x=a.makeTensorInfo(y,"float32",g),b=pn({inputs:{real:A,imag:x},backend:a}),w=Qg(l,u,i,n,a),S=w.real,N=w.imag,E=[S.length],R=a.makeTensorInfo(E,"float32",S),z=a.makeTensorInfo(E,"float32",N),M=pn({inputs:{real:R,imag:z},backend:a}),I=Qg(m,g,i,n,a),_=I.real,O=I.imag,q=[_.length],K=a.makeTensorInfo(q,"float32",_),D=a.makeTensorInfo(q,"float32",O),Q=pn({inputs:{real:K,imag:D},backend:a}),V=C.exponents(r,n),re=[V.real.length],Y=a.makeTensorInfo(re,"float32",V.real),ie=a.makeTensorInfo(re,"float32",V.imag),J=pn({inputs:{real:Y,imag:ie},backend:a}),ae=km({inputs:{a:J,b:Q},backend:a}),de=zh({inputs:{a:M,b:ae},backend:a}),xe=o5({inputs:{a:M,b:ae},backend:a}),be=Lo({inputs:{input:de},backend:a}),Ee=Lo({inputs:{input:xe},backend:a}),$e=Fu({inputs:{input:de},backend:a}),De=Fu({inputs:{input:xe},backend:a}),Be=Pu({inputs:[be,Ee],backend:a,attrs:{axis:0}}),Xe=Pu({inputs:[$e,De],backend:a,attrs:{axis:0}}),ot=a.data.get(Be.dataId).values,pt=a.data.get(Xe.dataId).values;return a.disposeIntermediateTensorInfo(h),a.disposeIntermediateTensorInfo(p),a.disposeIntermediateTensorInfo(c),a.disposeIntermediateTensorInfo(A),a.disposeIntermediateTensorInfo(x),a.disposeIntermediateTensorInfo(b),a.disposeIntermediateTensorInfo(R),a.disposeIntermediateTensorInfo(z),a.disposeIntermediateTensorInfo(M),a.disposeIntermediateTensorInfo(K),a.disposeIntermediateTensorInfo(D),a.disposeIntermediateTensorInfo(Q),a.disposeIntermediateTensorInfo(Y),a.disposeIntermediateTensorInfo(ie),a.disposeIntermediateTensorInfo(J),a.disposeIntermediateTensorInfo(ae),a.disposeIntermediateTensorInfo(de),a.disposeIntermediateTensorInfo(xe),a.disposeIntermediateTensorInfo(be),a.disposeIntermediateTensorInfo($e),a.disposeIntermediateTensorInfo(Ee),a.disposeIntermediateTensorInfo(De),a.disposeIntermediateTensorInfo(Be),a.disposeIntermediateTensorInfo(Xe),{real:ot,imag:pt}}function AZ(e,t,r){let n=new Float32Array(t*2);for(let a=0;a<t;a++){let s=0,i=0;for(let o=0;o<t;o++){let l=C.exponent(a*o,t,r),u=C.getComplexWithIndex(e,o);s+=u.real*l.real-u.imag*l.imag,i+=u.real*l.imag+u.imag*l.real}r&&(s/=t,i/=t),C.assignToTypedArray(n,s,i,a)}return n}function xZ(e){let{inputs:t,backend:r}=e,{input:n}=t,a=v.sizeFromShape(n.shape),s=n.shape[n.shape.length-1],i=a/s,o=Mt({inputs:{x:n},backend:r,attrs:{shape:[i,s]}}),l=rI(o,!1,r),u=Mt({inputs:{x:l},backend:r,attrs:{shape:n.shape}});return r.disposeIntermediateTensorInfo(o),r.disposeIntermediateTensorInfo(l),u}var bZ={kernelName:pf,backendName:"cpu",kernelFunc:xZ};function p5(e){let{backend:t,attrs:r}=e,{shape:n,value:a,dtype:s}=r,i=s||v.inferDtype(a),o=v.getArrayFromDType(i,v.sizeFromShape(n));return wZ(o,a,i),t.makeTensorInfo(n,i,o)}var vZ={kernelName:Ku,backendName:"cpu",kernelFunc:p5};function wZ(e,t,r){e.fill(t)}var kZ={kernelName:tl,backendName:"cpu",kernelFunc:({inputs:e,attrs:t,backend:r})=>{let{image:n}=e,a=r,s=v.getTypedArrayFromDType(n.dtype,v.sizeFromShape(n.shape)),[i,o,l,u]=n.shape,d=a.data.get(n.dataId).values;for(let h=0;h<i;h++){let p=h*l*o*u;for(let c=0;c<o;c++){let f=c*(l*u);for(let m=0;m<l;m++){let g=m*u;for(let y=0;y<u;y++){let A=Math.round(l-m-1),x=p+f+g+y,b=d[x];if(A>=0&&A<l){let w=A*u,S=p+f+w+y;b=d[S]}s[x]=b}}}}return{dataId:a.write(s,n.shape,n.dtype),shape:n.shape,dtype:n.dtype}}},IZ=Qt((e,t)=>Math.floor(e/t)),SZ=vr(hi,IZ,null,"int32"),CZ={kernelName:hi,backendName:"cpu",kernelFunc:SZ};function TZ(e){let{inputs:t,backend:r,attrs:n}=e,{x:a,filter:s,bias:i,preluActivationWeights:o}=t,{strides:l,pad:u,dataFormat:d,dilations:h,dimRoundingMode:p,activation:c,leakyreluAlpha:f}=n,m=eI({inputs:{x:a,filter:s},backend:r,attrs:{strides:l,pad:u,dataFormat:d,dilations:h,dimRoundingMode:p}});if(i){let g=m;m=zh({inputs:{a:m,b:i},backend:r}),r.disposeIntermediateTensorInfo(g)}if(c){let g=m;m=l5(r,m,c,o,f),r.disposeIntermediateTensorInfo(g)}return m}var NZ={kernelName:_s,backendName:"cpu",kernelFunc:TZ};function EZ(e){let{inputs:t,backend:r,attrs:n}=e,{x:a,filter:s,bias:i,preluActivationWeights:o}=t,{strides:l,pad:u,dataFormat:d,dilations:h,dimRoundingMode:p,activation:c,leakyreluAlpha:f}=n,m=tI({inputs:{x:a,filter:s},backend:r,attrs:{strides:l,pad:u,dataFormat:d,dilations:h,dimRoundingMode:p}});if(i){let g=m;m=zh({inputs:{a:m,b:i},backend:r}),r.disposeIntermediateTensorInfo(g)}if(c){let g=m;m=l5(r,m,c,o,f),r.disposeIntermediateTensorInfo(g)}return m}var RZ={kernelName:zs,backendName:"cpu",kernelFunc:EZ};function $Z(e){let{inputs:t,backend:r}=e,{params:n,indices:a}=t,s=v.sizeFromShape(n.shape),i=a.shape,o=i[i.length-1],[l,u,d,h]=C.prepareAndValidate(n,a);if(u===0)return r.makeTensorInfo(l,n.dtype,[]);let p=r.data.get(a.dataId).values,c=r.bufferSync(n),f=A8(p,c,n.dtype,u,o,d,h,n.shape,s);return r.makeTensorInfo(l,n.dtype,f.values)}var MZ={kernelName:nl,backendName:"cpu",kernelFunc:$Z};function FZ(e){let{inputs:t,backend:r,attrs:n}=e,{x:a,indices:s}=t,{axis:i,batchDims:o}=n;Ce([a,s],"gatherV2");let l=v.parseAxisParam(i,a.shape)[0],u=r.data.get(s.dataId).values,d=a.shape[l];for(let b=0;b<u.length;++b){let w=u[b];v.assert(w<=d-1&&w>=0,()=>`GatherV2: the index value ${w} is not in [0, ${d-1}]`)}let h=o;o==null&&(h=0);let p=v.sizeFromShape(s.shape),c=C.segment_util.collectGatherOpShapeInfo(a,s,l,h),f=Mt({inputs:{x:a},backend:r,attrs:{shape:[c.batchSize,c.outerSize,c.dimSize,c.sliceSize]}}),m=Mt({inputs:{x:s},backend:r,attrs:{shape:[c.batchSize,p/c.batchSize]}}),g=[c.batchSize,c.outerSize,p/c.batchSize,c.sliceSize],y=r.bufferSync(m),A=r.bufferSync(f),x=x8(A,y,g);return r.disposeIntermediateTensorInfo(f),r.disposeIntermediateTensorInfo(m),r.makeTensorInfo(c.outputShape,x.dtype,x.values)}var PZ={kernelName:rl,backendName:"cpu",kernelFunc:FZ};function _Z(e){let{inputs:t,backend:r}=e,{input:n}=t,a=v.sizeFromShape(n.shape),s=n.shape[n.shape.length-1],i=a/s,o=Mt({inputs:{x:n},backend:r,attrs:{shape:[i,s]}}),l=rI(o,!0,r),u=Mt({inputs:{x:l},backend:r,attrs:{shape:n.shape}});return r.disposeIntermediateTensorInfo(o),r.disposeIntermediateTensorInfo(l),u}var zZ={kernelName:hf,backendName:"cpu",kernelFunc:_Z},OZ=gt(Zu,e=>Number.isFinite(e)?1:0,"bool"),DZ={kernelName:Zu,backendName:"cpu",kernelFunc:OZ},LZ=gt(Yu,e=>Math.abs(e)===1/0?1:0,"bool"),BZ={kernelName:Yu,backendName:"cpu",kernelFunc:LZ},WZ=gt(Ju,e=>Number.isNaN(e)?1:0,"bool"),VZ={kernelName:Ju,backendName:"cpu",kernelFunc:WZ};function UZ(e){let{backend:t,attrs:r}=e,{start:n,stop:a,num:s}=r,i=I8(n,a,s);return t.makeTensorInfo([i.length],"float32",i)}var GZ={kernelName:cf,backendName:"cpu",kernelFunc:UZ},jZ=gt(Qu,e=>Math.log1p(e)),HZ={kernelName:Qu,backendName:"cpu",kernelFunc:jZ},qZ=Qt((e,t)=>e&&t),XZ=vr(ol,qZ,null,"bool"),KZ={kernelName:ol,backendName:"cpu",kernelFunc:XZ},ZZ=gt(ed,e=>e?0:1,"bool"),YZ={kernelName:ed,backendName:"cpu",kernelFunc:ZZ},JZ=Qt((e,t)=>e||t),QZ=vr(ah,JZ,null,"bool"),eY={kernelName:ah,backendName:"cpu",kernelFunc:QZ};function tY(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{depthRadius:s,bias:i,alpha:o,beta:l}=n;Ce(a,"LRN");let u=a.shape[3],d=u-1,h=r.data.get(a.dataId).values,p=v.sizeFromShape(a.shape),c=new Float32Array(p);function f(m){let g=m%u,y=m-g+Math.max(0,g-s),A=m-g+Math.min(g+s,d),x=0;for(;y<=A;y++){let b=h[y];x+=b*b}return x}for(let m=0;m<p;m++){let g=f(m),y=h[m]*Math.pow(i+o*g,-l);c[m]=y}return r.makeTensorInfo(a.shape,a.dtype,c)}var rY={kernelName:sh,backendName:"cpu",kernelFunc:tY};function nY(e){let{inputs:t,backend:r,attrs:n}=e,{x:a,y:s,dy:i}=t,{depthRadius:o,bias:l,alpha:u,beta:d}=n;Ce(i,"LRNGrad");let h=v.sizeFromShape(i.shape),p=i.shape[3],c=r.data.get(i.dataId).values,f=r.data.get(a.dataId).values,m=r.data.get(s.dataId).values,g=new Float32Array(h),y=h;for(let A=0;A<y;A++){let x=A%p,b=A-x+Math.max(0,x-o),w=A-x+Math.min(p,x+o+1),S=0;for(let N=b;N<w;N++)S+=Math.pow(f[N],2);S=u*S+l;for(let N=b;N<w;N++){let E=-2*u*d*f[N]*m[A]/S;A===N&&(E+=Math.pow(S,-d)),E*=c[A],g[N]+=E}}return r.makeTensorInfo(i.shape,a.dtype,g)}var aY={kernelName:ff,backendName:"cpu",kernelFunc:nY};function nI(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{reductionIndices:s,keepDims:i}=n,o=r,l=a.shape,u=l.length,d=v.parseAxisParam(s,l),h=d,p=C.getAxesPermutation(h,u),c=o.data.get(a.dataId).values;if(p!=null){let b=new Array(u);for(let w=0;w<b.length;w++)b[w]=l[p[w]];c=a5(c,l,a.dtype,p,b),h=C.getInnerMostAxes(h.length,u),l=b}Ce(a,"max"),C.assertAxesAreInnerMostDims("max",h,u);let[f,m]=C.computeOutAndReduceShapes(l,h),g=v.sizeFromShape(m),y=C8(c,g,f,a.dtype),A=o.write(y,f,a.dtype),x=f;return i&&(x=C.expandShapeToKeepDim(f,d)),{dataId:A,shape:x,dtype:a.dtype}}var sY={kernelName:Ai,backendName:"cpu",kernelFunc:nI};function iY(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t;Ce(a,"maxPool");let{filterSize:s,strides:i,pad:o,dimRoundingMode:l}=n,u=1;v.assert(C.eitherStridesOrDilationsAreOne(i,u),()=>`Error in maxPool: Either strides or dilations must be 1. Got strides ${i} and dilations '${u}'`);let d=C.computePool2DInfo(a.shape,s,i,u,o,l),h;if(d.filterWidth===1&&d.filterHeight===1&&v.arraysEqual(d.inShape,d.outShape))h=_a({inputs:{x:a},backend:r});else{let p=r.data.get(a.dataId).values,c=v.computeStrides(a.shape),f=u5(p,a.shape,a.dtype,c,d,"max");h=r.makeTensorInfo(d.outShape,a.dtype,f.values)}return h}var oY={kernelName:bi,backendName:"cpu",kernelFunc:iY};function lY(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{filterSize:s,strides:i,pad:o,dimRoundingMode:l,dataFormat:u}=n;Ce(a,"maxPool3d");let d=C.computePool3DInfo(a.shape,s,i,1,o,l,u),h=r.data.get(a.dataId).values,p=Q8(h,a.shape,a.dtype,v.computeStrides(a.shape),d,"max");return r.makeTensorInfo(p.shape,"float32",p.values)}var uY={kernelName:ih,backendName:"cpu",kernelFunc:lY};function dY(e){let{inputs:t,backend:r,attrs:n}=e,{dy:a,input:s}=t,{filterSize:i,strides:o,pad:l,dimRoundingMode:u}=n;Ce([a,s],"maxPool3DGrad");let d=C.computePool3DInfo(s.shape,i,o,1,l,u),h=r.bufferSync(s),p=YX(h,d),c=d.strideDepth,f=d.strideHeight,m=d.strideWidth,g=d.dilationDepth,y=d.dilationHeight,A=d.dilationWidth,x=d.effectiveFilterDepth,b=d.effectiveFilterHeight,w=d.effectiveFilterWidth,S=x-1-d.padInfo.front,N=w-1-d.padInfo.left,E=b-1-d.padInfo.top,R=We(s.shape,"float32"),z=r.bufferSync(a);for(let M=0;M<d.batchSize;++M)for(let I=0;I<d.inChannels;++I)for(let _=0;_<d.inDepth;++_)for(let O=0;O<d.inHeight;++O)for(let q=0;q<d.inWidth;++q){let K=_-S,D=O-E,Q=q-N,V=0;for(let re=0;re<x;re+=g){let Y=(K+re)/c;if(!(Y<0||Y>=d.outDepth||Math.floor(Y)!==Y))for(let ie=0;ie<b;ie+=y){let J=(D+ie)/f;if(!(J<0||J>=d.outHeight||Math.floor(J)!==J))for(let ae=0;ae<w;ae+=A){let de=(Q+ae)/m;if(de<0||de>=d.outWidth||Math.floor(de)!==de)continue;let xe=x*b*w-1-p.get(M,Y,J,de,I),be=re*b*w+ie*w+ae,Ee=xe===be?1:0;Ee!==0&&(V+=z.get(M,Y,J,de,I)*Ee)}}}R.set(V,M,_,O,q,I)}return r.makeTensorInfo(R.shape,R.dtype,R.values)}var pY={kernelName:gf,backendName:"cpu",kernelFunc:dY};function hY(e){let{inputs:t,backend:r,attrs:n}=e,{dy:a,input:s,output:i}=t,o=s;Ce([s,i],"maxPoolGrad");let{filterSize:l,strides:u,pad:d,dimRoundingMode:h}=n,p=C.computePool2DInfo(o.shape,l,u,1,d,h),c=r.data.get(o.dataId).values,f=We(p.outShape,o.dtype,J8(c,o.shape,o.dtype,p).values),m=p.strideHeight,g=p.strideWidth,y=p.dilationHeight,A=p.dilationWidth,x=p.effectiveFilterHeight,b=p.effectiveFilterWidth,w=b-1-p.padInfo.left,S=x-1-p.padInfo.top,N=We(o.shape,"float32"),E=r.data.get(a.dataId).values,R=We(a.shape,"float32",E);for(let z=0;z<p.batchSize;++z)for(let M=0;M<p.inChannels;++M)for(let I=0;I<p.inHeight;++I)for(let _=0;_<p.inWidth;++_){let O=I-S,q=_-w,K=0;for(let D=0;D<x;D+=y){let Q=(O+D)/m;if(!(Q<0||Q>=p.outHeight||Math.floor(Q)!==Q))for(let V=0;V<b;V+=A){let re=(q+V)/g;if(re<0||re>=p.outWidth||Math.floor(re)!==re)continue;let Y=x*b-1-f.get(z,Q,re,M),ie=D*b+V,J=Y===ie?1:0;J!==0&&(K+=R.get(z,Q,re,M)*J)}}N.set(K,z,I,_,M)}return r.makeTensorInfo(N.shape,N.dtype,N.values)}var cY={kernelName:mf,backendName:"cpu",kernelFunc:hY};function fY(e,t,r,n,a){let s=v.computeStrides(t),i=u5(e,t,r,s,a,"max"),o=J8(e,t,r,a,!0,n);return[i.values,o.values]}var mY={kernelName:yf,backendName:"cpu",kernelFunc:({inputs:e,attrs:t,backend:r})=>{let{x:n}=e,{filterSize:a,strides:s,pad:i,includeBatchInIndex:o}=t,l=r;Ce(n,"MaxPoolWithArgmax");let u=l.data.get(n.dataId).values,d=C.computePool2DInfo(n.shape,a,s,[1,1],i),[h,p]=fY(u,n.shape,n.dtype,o,d),c=l.write(h,d.outShape,n.dtype),f=l.write(p,d.outShape,n.dtype);return[{dataId:c,shape:d.outShape,dtype:n.dtype},{dataId:f,shape:d.outShape,dtype:"int32"}]}};function gY(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{axis:s,keepDims:i}=n,o=v.parseAxisParam(s,a.shape),l=C.computeOutAndReduceShapes(a.shape,o)[1],u=v.sizeFromShape(l),d=[],h=r.makeTensorInfo([],"float32",new Float32Array([u]));d.push(h);let p=qs({inputs:{x:a},backend:r,attrs:{dtype:"float32"}});d.push(p);let c=d5({inputs:{a:p,b:h},backend:r});d.push(c);let f=Oh({inputs:{x:c},backend:r,attrs:{axis:s,keepDims:i}});return d.forEach(m=>r.disposeIntermediateTensorInfo(m)),f}var yY={kernelName:vi,backendName:"cpu",kernelFunc:gY};function AY(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{axis:s,keepDims:i}=n;Ce(a,"min");let o=v.parseAxisParam(s,a.shape),l=o,u=C.getAxesPermutation(l,a.shape.length),d=a;u!=null&&(d=sn({inputs:{x:a},backend:r,attrs:{perm:u}}),l=C.getInnerMostAxes(l.length,a.shape.length)),C.assertAxesAreInnerMostDims("min",l,d.shape.length);let[h,p]=C.computeOutAndReduceShapes(d.shape,l),c=v.sizeFromShape(p),f=v.makeZerosTypedArray(v.sizeFromShape(h),d.dtype),m=r.data.get(d.dataId).values;for(let y=0;y<f.length;++y){let A=y*c,x=m[A];for(let b=0;b<c;++b){let w=m[A+b];(Number.isNaN(w)||w<x)&&(x=w)}f[y]=x}u!=null&&r.disposeIntermediateTensorInfo(d);let g=r.makeTensorInfo(h,d.dtype,f);if(i){let y=C.expandShapeToKeepDim(h,o),A=Mt({inputs:{x:g},backend:r,attrs:{shape:y}});return r.disposeIntermediateTensorInfo(g),A}return g}var xY={kernelName:wi,backendName:"cpu",kernelFunc:AY};function bY(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{paddings:s,mode:i}=n;Ce(a,"mirrorPad");let o=s.map((A,x)=>A[0]+a.shape[x]+A[1]),l=s.map(A=>A[0]),u=s.map((A,x)=>A[0]+a.shape[x]),d=i==="reflect"?0:1,h=r.data.get(a.dataId).values,p=a.shape.length,c=v.computeStrides(a.shape),f=v.sizeFromShape(o),m=o.length,g=v.computeStrides(o),y=v.getTypedArrayFromDType(a.dtype,f);for(let A=0;A<f;A++){let x=v.indexToLoc(A,m,g);for(let w=0;w<m;w++)x[w]<l[w]?x[w]=l[w]*2-x[w]-d:x[w]>=u[w]&&(x[w]=(u[w]-1)*2-x[w]+d);x=x.map((w,S)=>w-l[S]);let b=v.locToIndex(x,p,c);y[A]=h[b]}return{dataId:r.write(y,o,a.dtype),shape:o,dtype:a.dtype}}var vY={kernelName:Ii,backendName:"cpu",kernelFunc:bY},wY=Qt((e,t)=>{let r=e%t;return e<0&&t<0||e>=0&&t>=0?r:(r+t)%t}),kY=vr(td,wY),IY={kernelName:td,backendName:"cpu",kernelFunc:kY},SY=Go(K0());function aI(e){let{inputs:t,backend:r,attrs:n}=e,{logits:a}=t,{dim:s}=n,i=a.shape.length,o=s;if(o===-1&&(o=i-1),o!==i-1)throw Error(`Softmax along a non-last dimension is not yet supported. Logits was rank ${i} and dim was ${o}`);let l=v.parseAxisParam([o],a.shape),u=nI({inputs:{x:a},backend:r,attrs:{reductionIndices:l,keepDims:!1}}),d=C.expandShapeToKeepDim(u.shape,l),h=Mt({inputs:{x:u},backend:r,attrs:{shape:d}}),p=o5({inputs:{a,b:h},backend:r}),c=m8({inputs:{x:p},backend:r}),f=Oh({inputs:{x:c},backend:r,attrs:{axis:l,keepDims:!1}}),m=Mt({inputs:{x:f},backend:r,attrs:{shape:d}}),g=d5({inputs:{a:c,b:m},backend:r});return r.disposeIntermediateTensorInfo(u),r.disposeIntermediateTensorInfo(h),r.disposeIntermediateTensorInfo(p),r.disposeIntermediateTensorInfo(c),r.disposeIntermediateTensorInfo(f),r.disposeIntermediateTensorInfo(m),g}var CY={kernelName:Di,backendName:"cpu",kernelFunc:aI};function TY(e){let{inputs:t,backend:r,attrs:n}=e,{logits:a}=t,{numSamples:s,seed:i,normalized:o}=n;Ce(a,"multinomial");let l=o?a:aI({inputs:{logits:a},backend:r,attrs:{dim:-1}}),u=l.shape[0],d=l.shape[1],h=r.data.get(l.dataId).values,p=[u,s],c=v.makeZerosTypedArray(v.sizeFromShape(p),"int32");for(let f=0;f<u;++f){let m=f*d,g=new Float32Array(d-1);g[0]=h[m];for(let x=1;x<g.length;++x)g[x]=g[x-1]+h[m+x];let y=SY.alea(i.toString()),A=f*s;for(let x=0;x<s;++x){let b=y();c[A+x]=g.length;for(let w=0;w<g.length;w++)if(b<g[w]){c[A+x]=w;break}}}return o||r.disposeIntermediateTensorInfo(l),r.makeTensorInfo(p,"int32",c)}var NY={kernelName:Af,backendName:"cpu",kernelFunc:TY},EY=Kn.nonMaxSuppressionV3Impl;function RY(e){let{inputs:t,backend:r,attrs:n}=e,{boxes:a,scores:s}=t,{maxOutputSize:i,iouThreshold:o,scoreThreshold:l}=n;Ce(a,"NonMaxSuppression");let u=r.data.get(a.dataId).values,d=r.data.get(s.dataId).values,{selectedIndices:h}=EY(u,d,i,o,l);return r.makeTensorInfo([h.length],"int32",new Int32Array(h))}var $Y={kernelName:dl,backendName:"cpu",kernelFunc:RY},MY=Kn.nonMaxSuppressionV4Impl;function FY(e){let{inputs:t,backend:r,attrs:n}=e,{boxes:a,scores:s}=t,{maxOutputSize:i,iouThreshold:o,scoreThreshold:l,padToMaxOutputSize:u}=n;Ce(a,"NonMaxSuppressionPadded");let d=r.data.get(a.dataId).values,h=r.data.get(s.dataId).values,{selectedIndices:p,validOutputs:c}=MY(d,h,i,o,l,u);return[r.makeTensorInfo([p.length],"int32",new Int32Array(p)),r.makeTensorInfo([],"int32",new Int32Array([c]))]}var PY={kernelName:rd,backendName:"cpu",kernelFunc:FY},_Y=Kn.nonMaxSuppressionV5Impl;function zY(e){let{inputs:t,backend:r,attrs:n}=e,{boxes:a,scores:s}=t,{maxOutputSize:i,iouThreshold:o,scoreThreshold:l,softNmsSigma:u}=n;Ce(a,"NonMaxSuppressionWithScore");let d=r.data.get(a.dataId).values,h=r.data.get(s.dataId).values,p=i,c=o,f=l,m=u,{selectedIndices:g,selectedScores:y}=_Y(d,h,p,c,f,m);return[r.makeTensorInfo([g.length],"int32",new Int32Array(g)),r.makeTensorInfo([y.length],"float32",new Float32Array(y))]}var OY={kernelName:pl,backendName:"cpu",kernelFunc:zY};function DY(e){let{inputs:t,backend:r,attrs:n}=e,{indices:a}=t,{depth:s,onValue:i,offValue:o}=n;Ce(a,"oneHot");let l=v.sizeFromShape(a.shape),u=new Float32Array(l*s);u.fill(o);let d=r.data.get(a.dataId).values;for(let h=0;h<l;++h)d[h]>=0&&d[h]<s&&(u[h*s+d[h]]=i);return r.makeTensorInfo([...a.shape,s],"int32",u)}var LY={kernelName:cl,backendName:"cpu",kernelFunc:DY};function U0(e){let{inputs:t,backend:r}=e,{x:n}=t;if(n.dtype==="string")throw new Error("zerosLike is not supported for string tensors");if(n.dtype==="complex64"){let a=Lo({inputs:{input:n},backend:r}),s=U0({inputs:{x:a},backend:r}),i=Fu({inputs:{input:n},backend:r}),o=U0({inputs:{x:i},backend:r}),l=pn({inputs:{real:s,imag:o},backend:r});return r.disposeIntermediateTensorInfo(a),r.disposeIntermediateTensorInfo(s),r.disposeIntermediateTensorInfo(i),r.disposeIntermediateTensorInfo(o),l}else return p5({backend:r,attrs:{shape:n.shape,value:0,dtype:n.dtype}})}var BY={kernelName:El,backendName:"cpu",kernelFunc:U0};function sI(e){let{inputs:t,backend:r}=e,{x:n}=t;if(n.dtype==="string")throw new Error("onesLike is not supported for string tensors");if(n.dtype==="complex64"){let a=Lo({inputs:{input:n},backend:r}),s=sI({inputs:{x:a},backend:r}),i=Fu({inputs:{input:n},backend:r}),o=U0({inputs:{x:i},backend:r}),l=pn({inputs:{real:s,imag:o},backend:r});return r.disposeIntermediateTensorInfo(a),r.disposeIntermediateTensorInfo(s),r.disposeIntermediateTensorInfo(i),r.disposeIntermediateTensorInfo(o),l}else return p5({backend:r,attrs:{shape:n.shape,value:1,dtype:n.dtype}})}var WY={kernelName:hl,backendName:"cpu",kernelFunc:sI};function iI(e){let{inputs:t,backend:r,attrs:n}=e,{axis:a}=n;if(t.length===1)return V0({inputs:{input:t[0]},backend:r,attrs:{dim:a}});let s=t[0].shape,i=t[0].dtype;t.forEach(d=>{v.assertShapesMatch(s,d.shape,"All tensors passed to stack must have matching shapes"),v.assert(i===d.dtype,()=>"All tensors passed to stack must have matching dtypes")});let o=[],l=t.map(d=>{let h=V0({inputs:{input:d},backend:r,attrs:{dim:a}});return o.push(h),h}),u=Pu({inputs:l,backend:r,attrs:{axis:a}});return o.forEach(d=>r.disposeIntermediateTensorInfo(d)),u}var VY={kernelName:fl,backendName:"cpu",kernelFunc:iI};function UY(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{paddings:s,constantValue:i}=n;Ce(a,"pad");let o=s.map((y,A)=>y[0]+a.shape[A]+y[1]),l=s.map(y=>y[0]),u=r.data.get(a.dataId).values,d=v.sizeFromShape(a.shape),h=a.shape.length,p=v.computeStrides(a.shape),c=v.sizeFromShape(o),f=o.length,m=v.computeStrides(o),g=v.getTypedArrayFromDType(a.dtype,c);i!==0&&g.fill(i);for(let y=0;y<d;y++){let A=v.indexToLoc(y,h,p).map((b,w)=>b+l[w]),x=v.locToIndex(A,f,m);g[x]=u[y]}return{dataId:r.write(g,o,a.dtype),shape:o,dtype:a.dtype}}var oI={kernelName:Ci,backendName:"cpu",kernelFunc:UY},GY=Qt((e,t)=>Math.pow(e,t)),jY=vr(Ti,GY),HY={kernelName:Ti,backendName:"cpu",kernelFunc:jY};function qY(e){let{backend:t,attrs:r}=e,{start:n,stop:a,dtype:s,step:i}=r,o=s5(n,a,i,s);return t.makeTensorInfo([o.length],s,o)}var XY={kernelName:nd,backendName:"cpu",kernelFunc:qY},KY=gt(ad,e=>1/e),ZY={kernelName:ad,backendName:"cpu",kernelFunc:KY};function YY(e){let{inputs:t,backend:r,attrs:n}=e,{images:a}=t,{alignCorners:s,halfPixelCenters:i,size:o}=n;Ce(a,"resizeBilinear");let l=v.computeStrides(a.shape),[u,d]=o,[h,p,c,f]=a.shape,m=r.data.get(a.dataId).values,g=new Float32Array(v.sizeFromShape([h,u,d,f])),y=[s&&u>1?p-1:p,s&&d>1?c-1:c],A=[s&&u>1?u-1:u,s&&d>1?d-1:d],x=0,b=y[0]/A[0],w=y[1]/A[1];for(let S=0;S<h;S++)for(let N=0;N<u;N++){let E;i?E=b*(N+.5)-.5:E=b*N;let R=Math.max(0,Math.floor(E)),z=E-R,M=Math.min(p-1,Math.ceil(E)),I=S*l[0]+R*l[1],_=S*l[0]+M*l[1];for(let O=0;O<d;O++){let q;i?q=w*(O+.5)-.5:q=w*O;let K=Math.max(0,Math.floor(q)),D=q-K,Q=Math.min(c-1,Math.ceil(q)),V=I+K*l[2],re=_+K*l[2],Y=I+Q*l[2],ie=_+Q*l[2];for(let J=0;J<f;J++){let ae=m[V+J],de=m[re+J],xe=m[Y+J],be=m[ie+J],Ee=ae+(xe-ae)*D,$e=de+(be-de)*D,De=Ee+($e-Ee)*z;g[x++]=De}}}return r.makeTensorInfo([h,u,d,f],"float32",g)}var JY={kernelName:$i,backendName:"cpu",kernelFunc:YY};function QY(e){let{inputs:t,backend:r,attrs:n}=e,{images:a,dy:s}=t,{alignCorners:i}=n;Ce([s,a],"resizeBilinearGrad");let o=v.computeStrides(a.shape),[l,u,d,h]=a.shape,[,p,c]=s.shape,f=new Float32Array(l*u*d*h),m=[i&&p>1?u-1:u,i&&c>1?d-1:d],g=[i&&p>1?p-1:p,i&&c>1?c-1:c],y=m[0]/g[0],A=m[1]/g[1],x=r.data.get(s.dataId).values,b=0;for(let w=0;w<l;w++){let S=w*o[0];for(let N=0;N<p;N++){let E=N*y,R=Math.floor(E),z=Math.min(Math.ceil(E),u-1),M=S+R*o[1],I=S+z*o[1],_=E-R,O=1-_;for(let q=0;q<c;q++){let K=q*A,D=Math.floor(K),Q=Math.min(Math.ceil(K),d-1),V=K-D,re=1-V,Y=M+D*o[2],ie=M+Q*o[2],J=I+D*o[2],ae=I+Q*o[2],de=O*re,xe=O*V,be=_*re,Ee=_*V;for(let $e=0;$e<h;$e++){let De=x[b++];f[Y+$e]+=De*de,f[ie+$e]+=De*xe,f[J+$e]+=De*be,f[ae+$e]+=De*Ee}}}}return r.makeTensorInfo([l,d,u,h],"float32",f)}var eJ={kernelName:bf,backendName:"cpu",kernelFunc:QY};function tJ(e){let{inputs:t,backend:r,attrs:n}=e,{images:a}=t,{alignCorners:s,halfPixelCenters:i,size:o}=n;Ce(a,"resizeNearestNeighbor");let l=v.computeStrides(a.shape),[u,d]=o,[h,p,c,f]=a.shape,m=r.data.get(a.dataId).values,g=new Float32Array(h*u*d*f),y=[s&&u>1?p-1:p,s&&d>1?c-1:c],A=[s&&u>1?u-1:u,s&&d>1?d-1:d],x=y[0]/A[0],b=y[1]/A[1],w=0;for(let S=0;S<h;S++){let N=S*l[0];for(let E=0;E<u;E++){let R=i?x*(E+.5):x*E,z=Math.min(p-1,s?Math.round(R):Math.floor(R));i&&(z=Math.max(0,z));let M=N+z*l[1];for(let I=0;I<d;I++){let _=i?b*(I+.5):b*I,O=Math.min(c-1,s?Math.round(_):Math.floor(_));i&&(O=Math.max(0,O));let q=M+O*l[2];for(let K=0;K<f;K++){let D=m[q+K];g[w++]=D}}}}return r.makeTensorInfo([h,u,d,f],a.dtype,g)}var rJ={kernelName:sd,backendName:"cpu",kernelFunc:tJ};function nJ(e){let{inputs:t,backend:r,attrs:n}=e,{images:a,dy:s}=t,{alignCorners:i}=n;Ce([s,a],"resizeNearestNeighborGrad");let o=v.computeStrides(a.shape),l=v.computeStrides(s.shape),[u,d,h,p]=a.shape,[,c,f]=s.shape,m=new Float32Array(u*d*h*p),g=r.data.get(s.dataId).values,y=[i&&c>1?d-1:d,i&&f>1?h-1:h],A=[i&&c>1?c-1:c,i&&f>1?f-1:f],x=y[0]/A[0],b=y[1]/A[1],w=1/x,S=1/b,N=Math.ceil(w)*2+2,E=Math.ceil(S)*2+2;for(let R=0;R<u;R++){let z=R*o[0];for(let M=0;M<d;M++){let I=z+M*o[1],_=Math.floor(M*w),O=Math.floor(_-N/2);for(let q=0;q<h;q++){let K=I+q*o[2],D=Math.floor(q*S),Q=Math.floor(D-E/2);for(let V=0;V<p;V++){let re=0;for(let Y=0;Y<N;Y++){let ie=Y+O;if(ie<0||ie>=c)continue;let J=z+ie*l[1],ae=ie*x,de=Math.min(d-1,i?Math.round(ae):Math.floor(ae));if(M===de)for(let xe=0;xe<E;xe++){let be=xe+Q;if(be<0||be>=f)continue;let Ee=J+be*l[2],$e=be*b,De=Math.min(h-1,i?Math.round($e):Math.floor($e));q===De&&(re+=g[Ee+V])}}m[K+V]=re}}}}return r.makeTensorInfo(a.shape,a.dtype,m)}var aJ={kernelName:xf,backendName:"cpu",kernelFunc:nJ};function sJ(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{dims:s}=n;Ce(a,"reverse");let i=a.shape.length,o=v.parseAxisParam(s,a.shape);if(i===0)return _a({inputs:{x:a},backend:r});let l=new ir(a.shape,a.dtype),u=r.bufferSync(a);for(let d=0;d<l.size;d++){let h=l.indexToLoc(d),p=h.slice();o.forEach(c=>p[c]=a.shape[c]-1-p[c]),l.set(u.get(...p),...h)}return r.makeTensorInfo(l.shape,l.dtype,l.values)}var iJ={kernelName:gl,backendName:"cpu",kernelFunc:sJ},oJ={kernelName:Rl,backendName:"cpu",kernelFunc:({inputs:e,attrs:t,backend:r})=>{let{image:n}=e,{radians:a,fillValue:s,center:i}=t,o=r,l=v.getTypedArrayFromDType(n.dtype,v.sizeFromShape(n.shape)),[u,d,h,p]=n.shape,[c,f]=C.getImageCenter(i,d,h),m=255,g=Math.sin(a),y=Math.cos(a),A=o.data.get(n.dataId).values;for(let x=0;x<u;x++){let b=x*h*d*p;for(let w=0;w<d;w++){let S=w*(h*p);for(let N=0;N<h;N++){let E=N*p;for(let R=0;R<p;R++){let z=[u,w,N,R],M=z[2],I=z[1],_=(M-c)*y-(I-f)*g,O=(M-c)*g+(I-f)*y;_=Math.round(_+c),O=Math.round(O+f);let q=s;if(typeof s!="number"&&(R===3?q=m:q=s[R]),_>=0&&_<h&&O>=0&&O<d){let D=O*(h*p),Q=_*p,V=b+D+Q+R;q=A[V]}let K=b+S+E+R;l[K]=q}}}}return{dataId:o.write(l,n.shape,n.dtype),shape:n.shape,dtype:n.dtype}}},lJ=gt(yl,e=>{let t=Math.floor(e);return e-t<.5?Math.floor(e):e-t>.5?Math.ceil(e):t%2===0?t:t+1}),uJ={kernelName:yl,backendName:"cpu",kernelFunc:lJ};function lI(e,t,r,n,a,s,i,o,l,u){let d=[n/a,a],h=e.values,p=t.values;if(n===0)return We(r,t.dtype);let c=We(d,t.dtype);c.values.fill(l);for(let f=0;f<s;f++){let m=[],g=0;for(let y=0;y<i;y++){let A=h[f*i+y];m.push(A),g+=A*o[y]}if(g<0||g>=n/a)throw new Error(`Invalid indices: ${m} does not index into ${r}`);for(let y=0;y<a;y++)u?c.values[g*a+y]+=p[f*a+y]:c.values[g*a+y]=t.rank===0?p[0]:p[f*a+y]}return c}function dJ(e){let{inputs:t,backend:r,attrs:n}=e,{indices:a,updates:s}=t,{shape:i}=n,{sliceRank:o,numUpdates:l,sliceSize:u,strides:d,outputSize:h}=C.calculateShapes(s,a,i),p=!0,c=r.bufferSync(a),f=r.bufferSync(s),m=lI(c,f,i,h,u,l,o,d,0,p);return r.makeTensorInfo(i,m.dtype,m.values)}var pJ={kernelName:Al,backendName:"cpu",kernelFunc:dJ};function hJ(e,t){let r=0,n=e.length,a=0;for(;r<n;)a=Math.floor((r+n)/2),e[a]<t?r=a+1:n=a;return n}function cJ(e,t){let r=0,n=e.length,a=0;for(;r<n;)a=Math.floor((r+n)/2),e[a]<=t?r=a+1:n=a;return n}function fJ(e,t,r,n,a,s){let i=v.getArrayFromDType("int32",r*a);for(let o=0;o<r;++o){let l=e.slice(o*n,(o+1)*n),u=o*a;for(let d=0;d<a;++d)i[u+d]=s==="left"?hJ(l,t[d+u]):cJ(l,t[d+u])}return i}function mJ(e){let{inputs:t,backend:r,attrs:n}=e,{sortedSequence:a,values:s}=t,{side:i}=n,o=r.data.get(a.dataId).values,l=r.data.get(s.dataId).values,u=fJ(o,l,a.shape[0],a.shape[1],s.shape[1],i);return r.makeTensorInfo(s.shape,"int32",u)}var gJ={kernelName:vf,backendName:"cpu",kernelFunc:mJ};function yJ(e){let{inputs:t,backend:r}=e,{condition:n,t:a,e:s}=t;Ce([n,a,s],"select");let i=n.shape.length,o=r.data.get(n.dataId).values,l=r.data.get(a.dataId).values,u=r.data.get(s.dataId).values,d=Nr(a.dtype,s.dtype),h=v.makeZerosTypedArray(v.sizeFromShape(a.shape),d),p=0,c=i===0||i>1||a.shape.length===1?1:v.sizeFromShape(a.shape.slice(1));for(let f=0;f<o.length;f++)for(let m=0;m<c;m++)o[f]===1?h[p++]=l[f]:h[p++]=u[f];return r.makeTensorInfo(a.shape,d,h)}var AJ={kernelName:xl,backendName:"cpu",kernelFunc:yJ},xJ=C.SELU_SCALEALPHA,bJ=C.SELU_SCALE,vJ=gt(id,e=>e>=0?bJ*e:xJ*(Math.exp(e)-1)),wJ={kernelName:id,backendName:"cpu",kernelFunc:vJ},kJ=gt(od,e=>e<0?-1:e>0?1:0),IJ={kernelName:od,backendName:"cpu",kernelFunc:kJ},SJ=gt(Pi,e=>Math.sin(e)),CJ={kernelName:Pi,backendName:"cpu",kernelFunc:SJ},TJ=gt(vl,e=>Math.sinh(e)),NJ={kernelName:vl,backendName:"cpu",kernelFunc:TJ},EJ=11920928955078125e-23,_4=Math.log(EJ)+2,RJ=gt(ld,e=>{let t=e>-_4,r=e<_4,n=Math.exp(e),a;return r?a=n:t?a=e:a=Math.log(1+n),a}),$J={kernelName:ld,backendName:"cpu",kernelFunc:RJ};function MJ(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{blockShape:s,paddings:i}=n;Ce([a],"spaceToBatchND");let o=v.sizeFromShape(s),l=[[0,0]];l.push(...i);for(let g=1+s.length;g<a.shape.length;++g)l.push([0,0]);let u=oI.kernelFunc({inputs:{x:a},backend:r,attrs:{paddings:l,constantValue:0}}),d=C.getReshaped(u.shape,s,o,!1),h=C.getPermuted(d.length,s.length,!1),p=C.getReshapedPermuted(u.shape,s,o,!1),c=Mt({inputs:{x:u},backend:r,attrs:{shape:d}}),f=sn({inputs:{x:c},backend:r,attrs:{perm:h}}),m=Mt({inputs:{x:f},backend:r,attrs:{shape:p}});return r.disposeIntermediateTensorInfo(u),r.disposeIntermediateTensorInfo(c),r.disposeIntermediateTensorInfo(f),m}var FJ={kernelName:wl,backendName:"cpu",kernelFunc:MJ};function PJ(e){let{inputs:t,backend:r}=e,{indices:n,values:a,denseShape:s,defaultValue:i}=t;if(s.shape.length!==1)throw new Error(`Dense shape must be a vector, saw:
|
|
${s.shape}`);if(n.shape.length!==2)throw new Error(`Indices must be a matrix, saw:
|
|
${n.shape}`);if(a.shape.length!==1)throw new Error(`Values must be a vector, saw:
|
|
${a.shape}`);if(i.shape.length!==0)throw new Error(`Default value must be a scalar, saw:
|
|
${i.shape}`);let o=r.data.get(n.dataId).values,l=r.data.get(a.dataId).values,u=r.data.get(s.dataId).values,d=r.data.get(i.dataId).values[0],[h,p,c,f,m]=P8(o,n.shape,n.dtype,l,a.dtype,u,d);return[r.makeTensorInfo(p,n.dtype,h),r.makeTensorInfo([p[0]],a.dtype,c),r.makeTensorInfo([f.length],"bool",new Uint8Array(f.map(g=>Number(g)))),r.makeTensorInfo([m.length],n.dtype,new Int32Array(m))]}var _J={kernelName:lh,backendName:"cpu",kernelFunc:PJ};function zJ(e){let{inputs:t,backend:r}=e,{inputIndices:n,inputShape:a,newShape:s}=t;if(n.shape.length!==2)throw new Error(`Input indices should be a matrix but received shape
|
|
${n.shape}`);if(a.shape.length!==1)throw new Error(`Input shape should be a vector but received shape
|
|
${a.shape}`);if(s.shape.length!==1)throw new Error(`Target shape should be a vector but received shape ${s.shape}`);let i=Array.from(r.data.get(a.dataId).values),o=r.data.get(n.dataId).values,l=Array.from(r.data.get(s.dataId).values),[u,d,h]=_8(o,n.shape,n.dtype,i,l);return[r.makeTensorInfo(d,n.dtype,u),r.makeTensorInfo([h.length],s.dtype,new Int32Array(h))]}var OJ={kernelName:ud,backendName:"cpu",kernelFunc:zJ};function DJ(e){let{inputs:t,backend:r}=e,{data:n,indices:a,segmentIds:s}=t;if(n.shape.length<1)throw new Error("Data should be at least 1 dimensional but received scalar");if(a.shape.length!==1)throw new Error(`Indices should be a vector but received shape
|
|
${a.shape}`);if(s.shape.length!==1)throw new Error(`Segment ids should be a vector but received shape
|
|
${s.shape}`);if(a.shape[0]!==s.shape[0])throw new Error("segmentIds and indices should have same size.");let i=r.data.get(n.dataId).values,o=r.data.get(a.dataId).values,l=r.data.get(s.dataId).values,[u,d]=i5(i,n.shape,n.dtype,o,l,!0);return r.makeTensorInfo(d,n.dtype,u)}var LJ={kernelName:uh,backendName:"cpu",kernelFunc:DJ};function BJ(e){let{inputs:t,backend:r}=e,{data:n,indices:a,segmentIds:s}=t;if(n.shape.length<1)throw new Error("Data should be at least 1 dimensional but received scalar");if(a.shape.length!==1)throw new Error(`Indices should be a vector but received shape
|
|
${a.shape}`);if(s.shape.length!==1)throw new Error(`Segment ids should be a vector but received shape
|
|
${s.shape}`);if(a.shape[0]!==s.shape[0])throw new Error("segmentIds and indices should have same size.");let i=r.data.get(n.dataId).values,o=r.data.get(a.dataId).values,l=r.data.get(s.dataId).values,[u,d]=i5(i,n.shape,n.dtype,o,l);return r.makeTensorInfo(d,n.dtype,u)}var WJ={kernelName:dh,backendName:"cpu",kernelFunc:BJ};function VJ(e){let{inputs:t,backend:r,attrs:n}=e,{sparseIndices:a,sparseValues:s,defaultValue:i}=t,{outputShape:o}=n,{sliceRank:l,numUpdates:u,sliceSize:d,strides:h,outputSize:p}=C.calculateShapes(s,a,o),c=!1,f=r.bufferSync(a),m=r.bufferSync(s),g=r.data.get(i.dataId).values[0],y=lI(f,m,o,p,d,u,l,h,g,c);return r.makeTensorInfo(o,y.dtype,y.values)}var UJ={kernelName:ph,backendName:"cpu",kernelFunc:VJ};function GJ(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{numOrSizeSplits:s,axis:i}=n,o=v.parseAxisParam(i,a.shape)[0],l=C.prepareSplitSize(a,s,o),u=new Array(a.shape.length).fill(0),d=a.shape.slice();return l.map(h=>{let p=[...d];p[o]=h;let c=Bo({inputs:{x:a},backend:r,attrs:{begin:u,size:p}});return u[o]+=h,c})}var jJ={kernelName:kl,backendName:"cpu",kernelFunc:GJ},HJ={kernelName:dd,backendName:"cpu",kernelFunc:({inputs:e,backend:t})=>{let{x:r}=e,n=t;Ce(r,"square");let a=n.data.get(r.dataId).values,s=new Float32Array(a.length);for(let i=0;i<a.length;++i){let o=a[i];s[i]=o*o}return{dataId:n.write(s,r.shape,r.dtype),shape:r.shape,dtype:r.dtype}}},qJ=gt(Ui,(e,t)=>{let r=t;return isNaN(e)?NaN:e>0?1:r.alpha}),XJ={kernelName:Ui,backendName:"cpu",kernelFunc:qJ};function KJ(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{begin:s,end:i,strides:o,beginMask:l,endMask:u,ellipsisMask:d,newAxisMask:h,shrinkAxisMask:p}=n;Ce(a,"stridedSlice");let{finalShapeSparse:c,finalShape:f,isIdentity:m,sliceDim0:g,isSimpleSlice:y,begin:A,end:x,strides:b}=Ot.sliceInfo(a.shape,s,i,o,l,u,d,h,p),w;if(m)w=Mt({inputs:{x:a},backend:r,attrs:{shape:f}});else if(g||y){v.assert(a.shape.length>=1,()=>`Input must have rank at least 1, got: ${a.shape.length}`);let S=Ot.computeOutShape(A,x,b),N=Bo({inputs:{x:a},backend:r,attrs:{begin:A,size:S}});w=Mt({inputs:{x:N},backend:r,attrs:{shape:f}}),r.disposeIntermediateTensorInfo(N)}else{let S=r.bufferSync(a),N=O8(c,S,b,A);w=r.makeTensorInfo(f,N.dtype,N.values)}return w}var ZJ={kernelName:Il,backendName:"cpu",kernelFunc:KJ};function YJ(e){let{inputs:t,backend:r,attrs:n}=e,{separator:a,nGramWidths:s,leftPad:i,rightPad:o,padWidth:l,preserveShortSequences:u}=n,{data:d,dataSplits:h}=t,p=r.data.get(d.dataId).values,c=r.data.get(h.dataId).values,[f,m]=D8(p,c,a,s,i,o,l,u);return[r.makeTensorInfo([f.length],"string",f),r.makeTensorInfo(h.shape,"int32",m)]}var JJ={kernelName:hh,backendName:"cpu",kernelFunc:YJ};function QJ(e){let{inputs:t,backend:r,attrs:n}=e,{skipEmpty:a}=n,{input:s,delimiter:i}=t;if(s.dtype!=="string")throw new Error("Input must be of datatype string");if(s.shape.length!==1)throw new Error(`Input must be a vector, got shape: ${s.shape}`);if(i.shape.length!==0)throw new Error(`Delimiter must be a scalar, got shape: ${i.shape}`);let o=r.data.get(s.dataId).values,l=r.data.get(i.dataId).values[0],[u,d,h]=L8(o,l,a),p=d.length;return[r.makeTensorInfo([p,2],"int32",u),r.makeTensorInfo([p],"string",d),r.makeTensorInfo([2],"int32",new Int32Array(h))]}var eQ={kernelName:wf,backendName:"cpu",kernelFunc:QJ};function tQ(e){let{inputs:t,backend:r,attrs:n}=e,{numBuckets:a}=n,{input:s}=t;if(s.dtype!=="string")throw new Error("Input must be of datatype string");if(a<=0)throw new Error("Number of buckets must be at least 1");let i=r.data.get(s.dataId).values,o=B8(i,a);return r.makeTensorInfo(s.shape,"int32",o)}var rQ={kernelName:kf,backendName:"cpu",kernelFunc:tQ},nQ=gt(Sl,e=>Math.tan(e)),aQ={kernelName:Sl,backendName:"cpu",kernelFunc:nQ},sQ=gt(Wi,e=>Math.tanh(e)),iQ={kernelName:Wi,backendName:"cpu",kernelFunc:sQ};function oQ(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{reps:s}=n;Ce(a,"tile");let i=V8(r.bufferSync(a),s);return r.makeTensorInfo(i.shape,i.dtype,i.values)}var lQ={kernelName:es,backendName:"cpu",kernelFunc:oQ};function uQ(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{k:s,sorted:i}=n;Ce(a,"topk");let o=r.data.get(a.dataId).values,[l,u]=G8(o,a.shape,a.dtype,s,i);return[r.makeTensorInfo(l.shape,l.dtype,l.values),r.makeTensorInfo(u.shape,u.dtype,u.values)]}var dQ={kernelName:Cl,backendName:"cpu",kernelFunc:uQ};function pQ(e){let{inputs:t,attrs:r,backend:n}=e,{image:a,transforms:s}=t,{interpolation:i,fillMode:o,fillValue:l,outputShape:u}=r,[d,h,p,c]=a.shape,[f,m]=u!=null?u:[h,p],g=[d,f,m,c],y=v.computeStrides(a.shape),A=y[0],x=y[1],b=y[2],w=v.getTypedArrayFromDType(a.dtype,v.sizeFromShape(g));w.fill(l);let S=n.data.get(a.dataId).values,N=n.data.get(s.dataId).values;for(let E=0;E<d;++E){let R=s.shape[0]===1?N:N.subarray(E*8,E*8+8);for(let z=0;z<f;++z)for(let M=0;M<m;++M)for(let I=0;I<c;++I){let _,O=R[6]*M+R[7]*z+1;if(O===0)continue;let q=(R[0]*M+R[1]*z+R[2])/O,K=(R[3]*M+R[4]*z+R[5])/O,D=z4(q,p,o),Q=z4(K,h,o);switch(i){case"nearest":_=yQ(S,h,p,A,x,b,E,Q,D,I,l);break;case"bilinear":_=AQ(S,h,p,A,x,b,E,Q,D,I,l);break;default:throw new Error(`Error in Transform: Expect 'nearest' or 'bilinear', but got ${i}`)}let V=E*A+z*x+M*b+I;w[V]=_}return n.makeTensorInfo(g,a.dtype,w)}return{dataId:n.write(w,g,a.dtype),shape:a.shape,dtype:a.dtype}}var hQ={kernelName:Tl,backendName:"cpu",kernelFunc:pQ};function z4(e,t,r){switch(r){case"reflect":return cQ(e,t);case"wrap":return fQ(e,t);case"nearest":return gQ(e,t);case"constant":default:return mQ(e,t)}}function cQ(e,t){let r=e;if(r<0)if(t<=1)r=0;else{let n=2*t;r<n&&(r=n*Math.trunc(-r/n)+r),r=r<-t?r+n:-r-1}else if(r>t-1)if(t<=1)r=0;else{let n=2*t;r-=n*Math.trunc(r/n),r>=t&&(r=n-r-1)}return v.clamp(0,r,t-1)}function fQ(e,t){let r=e;if(r<0)if(t<=1)r=0;else{let n=t-1;r+=t*(Math.trunc(-r/n)+1)}else if(r>t-1)if(t<=1)r=0;else{let n=t-1;r-=t*Math.trunc(r/n)}return v.clamp(0,r,t-1)}function mQ(e,t){return e}function gQ(e,t){return v.clamp(0,e,t-1)}function Tp(e,t,r,n,a,s,i,o,l,u,d){let h=i*n+o*a+l*s+u;return 0<=o&&o<t&&0<=l&&l<r?e[h]:d}function yQ(e,t,r,n,a,s,i,o,l,u,d){let h=Math.round(o),p=Math.round(l);return Tp(e,t,r,n,a,s,i,h,p,u,d)}function AQ(e,t,r,n,a,s,i,o,l,u,d){let h=Math.floor(o),p=Math.floor(l),c=h+1,f=p+1,m=(f-l)*Tp(e,t,r,n,a,s,i,h,p,u,d)+(l-p)*Tp(e,t,r,n,a,s,i,h,f,u,d),g=(f-l)*Tp(e,t,r,n,a,s,i,c,p,u,d)+(l-p)*Tp(e,t,r,n,a,s,i,c,f,u,d);return(c-o)*m+(o-h)*g}function xQ(e){let{inputs:t,attrs:r,backend:n}=e,{axis:a}=r,{x:s}=t;Ce(s,"unique");let i=n.data.get(s.dataId).values,{outputValues:o,outputShape:l,indices:u}=j8(i,a,s.shape,s.dtype);return[n.makeTensorInfo(l,s.dtype,o),n.makeTensorInfo([u.length],"int32",u)]}var bQ={kernelName:If,backendName:"cpu",kernelFunc:xQ};function vQ(e){let{inputs:t,backend:r,attrs:n}=e,{value:a}=t,{axis:s}=n;s<0&&(s+=a.shape.length);let i=a.shape.length,o=a.shape[s],l=new Array(i-1),u=0;for(let c=0;c<i;c++)c!==s&&(l[u++]=a.shape[c]);let d=new Array(i).fill(0),h=a.shape.slice();h[s]=1;let p=new Array(o);for(let c=0;c<p.length;c++){d[s]=c;let f=Bo({inputs:{x:a},backend:r,attrs:{begin:d,size:h}});p[c]=Mt({inputs:{x:f},backend:r,attrs:{shape:l}}),r.disposeIntermediateTensorInfo(f)}return p}var wQ={kernelName:Nl,backendName:"cpu",kernelFunc:vQ};function kQ(e){let{inputs:t,backend:r,attrs:n}=e,{x:a,segmentIds:s}=t,{numSegments:i}=n;Ce(a,"unsortedSegmentSum");let o=a.shape.length,l=s.shape.length,u=[],d=[],h=o-l,p=s;for(let f=0;f<h;++f){let m=V0({inputs:{input:p},backend:r,attrs:{dim:f+1}});p=m,d.push(m)}for(let f=0;f<i;++f){let m=v.createScalarValue(f,"int32"),g=r.makeTensorInfo([],"int32",m),y=c8({inputs:{a:g,b:p},backend:r}),A=qs({inputs:{x:y},backend:r,attrs:{dtype:"float32"}}),x=km({inputs:{a:A,b:a},backend:r}),b=Oh({inputs:{x},backend:r,attrs:{axis:0,keepDims:!1}});u.push(b),d.push(g),d.push(y),d.push(A),d.push(x),d.push(b)}let c=iI({inputs:u,backend:r,attrs:{axis:0}});return d.forEach(f=>r.disposeIntermediateTensorInfo(f)),c}var IQ={kernelName:ch,backendName:"cpu",kernelFunc:kQ},SQ=[SX,yq,TX,EX,kq,$X,FX,_X,OX,LX,WX,UX,jX,XX,ZX,QX,tK,nK,sK,kX,oK,uK,pK,cK,vq,Sq,mK,Aq,yK,xK,bK,wK,IK,CK,NK,RK,MK,PK,zK,DK,BK,VK,GK,jK,qK,KK,YK,JK,QK,eZ,nZ,gX,sZ,Cq,cZ,Tq,fZ,Eq,bZ,vZ,kZ,$q,CZ,NZ,RZ,MZ,PZ,Fq,_q,xq,zZ,AK,DZ,BZ,VZ,yX,Oq,Lq,GZ,Wq,HZ,KZ,YZ,eY,rY,aY,sY,Uq,oY,uY,pY,cY,mY,yY,xY,jq,vY,IY,NY,qq,Kq,$Y,PY,OY,Yq,LY,WY,VY,oI,HY,xX,eX,XY,bq,Jg,ZY,bX,vX,wX,JY,eJ,rJ,aJ,iJ,oJ,uJ,rX,pJ,gJ,AJ,wJ,aX,IJ,CJ,NJ,sX,CY,$J,FJ,_J,OJ,LJ,WJ,UJ,jJ,lX,HJ,dX,XJ,ZJ,JJ,eQ,rQ,fX,tZ,aQ,iQ,lQ,dQ,hQ,Jq,bQ,wQ,IQ,BY];for(let e of SQ)qn(e);var uI={};Le(uI,{assertNotComplex:()=>Id,bindCanvasToFramebuffer:()=>zQ,bindColorTextureToFramebuffer:()=>u0,bindTextureToProgramUniformSampler:()=>II,bindTextureUnit:()=>vI,bindVertexBufferToProgramAttribute:()=>ey,callAndCheck:()=>we,canBeRepresented:()=>dI,createFragmentShader:()=>cI,createFramebuffer:()=>bI,createProgram:()=>fI,createStaticIndexBuffer:()=>yI,createStaticVertexBuffer:()=>gI,createTexture:()=>AI,createVertexShader:()=>hI,getBatchDim:()=>Wo,getExtensionOrThrow:()=>Np,getFramebufferErrorMessage:()=>SI,getMaxTexturesInShader:()=>EI,getNumChannels:()=>PQ,getProgramUniformLocation:()=>kI,getProgramUniformLocationOrThrow:()=>wI,getRowsCols:()=>Vo,getShapeAs3D:()=>d0,getTextureShapeFromLogicalShape:()=>TI,getWebGLDisjointQueryTimerVersion:()=>RI,getWebGLErrorMessage:()=>pI,getWebGLMaxTextureSize:()=>NI,hasExtension:()=>En,isCapableOfRenderingToFloatTexture:()=>$I,isDownloadFloatTextureEnabled:()=>MI,isReshapeFree:()=>Xp,isWebGLFenceEnabled:()=>FI,isWebGLVersionEnabled:()=>ry,linkProgram:()=>mI,logShaderSourceAndInfoLog:()=>c5,resetMaxTextureSize:()=>OQ,resetMaxTexturesInShader:()=>DQ,unbindColorTextureFromFramebuffer:()=>ty,unbindTextureUnit:()=>_Q,validateFramebuffer:()=>Ep,validateProgram:()=>l0,validateTextureSize:()=>xI});var So={},hg={alpha:!1,antialias:!1,premultipliedAlpha:!1,preserveDrawingBuffer:!1,depth:!1,stencil:!1,failIfMajorPerformanceCaveat:!0};function Im(e,t){So[e]=t}function xa(e,t){if(!(e in So)||t!=null){let n=TQ(e,t);if(n!==null)So[e]=n;else return console.log("Could not get context for WebGL version",e),null}let r=So[e];return r==null||r.isContextLost()?(delete So[e],xa(e)):(r.disable(r.DEPTH_TEST),r.disable(r.STENCIL_TEST),r.disable(r.BLEND),r.disable(r.DITHER),r.disable(r.POLYGON_OFFSET_FILL),r.disable(r.SAMPLE_COVERAGE),r.enable(r.SCISSOR_TEST),r.enable(r.CULL_FACE),r.cullFace(r.BACK),So[e])}function CQ(e){if(typeof OffscreenCanvas!="undefined"&&e===2)return new OffscreenCanvas(300,150);if(typeof document!="undefined")return document.createElement("canvas");throw new Error("Cannot create a canvas in this context")}function TQ(e,t){if(e!==1&&e!==2)throw new Error("Cannot get WebGL rendering context, WebGL is disabled.");let r=t==null?CQ(e):t;return r.addEventListener("webglcontextlost",n=>{n.preventDefault(),delete So[e]},!1),e===1?r.getContext("webgl",hg)||r.getContext("experimental-webgl",hg):r.getContext("webgl2",hg)}function Dh(e,t){return[t,e]}function NQ(e,t){return e*t}function t0(e){let t=v.sizeFromShape(e),r=Math.ceil(t/4);return v.sizeToSquarishShape(r)}function kd(e,t){return[Math.max(1,Math.ceil(t/2)),Math.max(1,Math.ceil(e/2))]}function EQ(e,t){let[r,n]=kd(e,t);return r*n*4}function h5(e,t){let r=e,n,a,s,i,o,l,u,d,h,p;return Z().getNumber("WEBGL_VERSION")===2?(n=r.R32F,a=r.R16F,s=r.RGBA16F,i=r.RGBA32F,o=r.RED,u=4,d=1,h=r.HALF_FLOAT,p=r.FLOAT,l=r.RGBA8):(n=e.RGBA,a=e.RGBA,s=e.RGBA,i=r.RGBA,o=e.RGBA,u=4,d=4,h=t!=null?t.HALF_FLOAT_OES:null,p=e.FLOAT,l=e.RGBA),{internalFormatFloat:n,internalFormatHalfFloat:a,internalFormatPackedHalfFloat:s,internalFormatPackedFloat:i,textureFormatFloat:o,downloadTextureFormat:l,downloadUnpackNumChannels:u,defaultNumChannels:d,textureTypeHalfFloat:h,textureTypeFloat:p}}function we(e,t){let r=t();return Z().getBool("DEBUG")&&RQ(e),r}function RQ(e){let t=e.getError();if(t!==e.NO_ERROR)throw new Error("WebGL Error: "+pI(e,t))}var $Q=596e-10,MQ=65504;function dI(e){return!!(Z().getBool("WEBGL_RENDER_FLOAT32_ENABLED")||e===0||$Q<Math.abs(e)&&Math.abs(e)<MQ)}function pI(e,t){switch(t){case e.NO_ERROR:return"NO_ERROR";case e.INVALID_ENUM:return"INVALID_ENUM";case e.INVALID_VALUE:return"INVALID_VALUE";case e.INVALID_OPERATION:return"INVALID_OPERATION";case e.INVALID_FRAMEBUFFER_OPERATION:return"INVALID_FRAMEBUFFER_OPERATION";case e.OUT_OF_MEMORY:return"OUT_OF_MEMORY";case e.CONTEXT_LOST_WEBGL:return"CONTEXT_LOST_WEBGL";default:return`Unknown error code ${t}`}}function Np(e,t){return ss(e,()=>e.getExtension(t),'Extension "'+t+'" not supported on this browser.')}function hI(e,t){let r=ss(e,()=>e.createShader(e.VERTEX_SHADER),"Unable to create vertex WebGLShader.");if(we(e,()=>e.shaderSource(r,t)),we(e,()=>e.compileShader(r)),e.getShaderParameter(r,e.COMPILE_STATUS)===!1)throw console.log(e.getShaderInfoLog(r)),new Error("Failed to compile vertex shader.");return r}function cI(e,t){let r=ss(e,()=>e.createShader(e.FRAGMENT_SHADER),"Unable to create fragment WebGLShader.");if(we(e,()=>e.shaderSource(r,t)),we(e,()=>e.compileShader(r)),Z().get("ENGINE_COMPILE_ONLY"))return r;if(e.getShaderParameter(r,e.COMPILE_STATUS)===!1)throw c5(t,e.getShaderInfoLog(r)),new Error("Failed to compile fragment shader.");return r}var FQ=/ERROR: [0-9]+:([0-9]+):/g;function c5(e,t){let r=FQ.exec(t);if(r==null){console.log(`Couldn't parse line number in error: ${t}`),console.log(e);return}let n=+r[1],a=e.split(`
|
|
`),s=a.length.toString().length+2,i=a.map((h,p)=>v.rightPad((p+1).toString(),s)+h),o=0;for(let h=0;h<i.length;h++)o=Math.max(i[h].length,o);let l=i.slice(0,n-1),u=i.slice(n-1,n),d=i.slice(n);console.log(l.join(`
|
|
`)),console.log(t.split(`
|
|
`)[0]),console.log(`%c ${v.rightPad(u[0],o)}`,"border:1px solid red; background-color:#e3d2d2; color:#a61717"),console.log(d.join(`
|
|
`))}function fI(e){return ss(e,()=>e.createProgram(),"Unable to create WebGLProgram.")}function mI(e,t){if(we(e,()=>e.linkProgram(t)),!Z().get("ENGINE_COMPILE_ONLY")&&e.getProgramParameter(t,e.LINK_STATUS)===!1)throw console.log(e.getProgramInfoLog(t)),new Error("Failed to link vertex and fragment shaders.")}function l0(e,t){if(we(e,()=>e.validateProgram(t)),e.getProgramParameter(t,e.VALIDATE_STATUS)===!1)throw console.log(e.getProgramInfoLog(t)),new Error("Shader program validation failed.")}function gI(e,t){let r=ss(e,()=>e.createBuffer(),"Unable to create WebGLBuffer");return we(e,()=>e.bindBuffer(e.ARRAY_BUFFER,r)),we(e,()=>e.bufferData(e.ARRAY_BUFFER,t,e.STATIC_DRAW)),r}function yI(e,t){let r=ss(e,()=>e.createBuffer(),"Unable to create WebGLBuffer");return we(e,()=>e.bindBuffer(e.ELEMENT_ARRAY_BUFFER,r)),we(e,()=>e.bufferData(e.ELEMENT_ARRAY_BUFFER,t,e.STATIC_DRAW)),r}function PQ(){return Z().getNumber("WEBGL_VERSION")===2?1:4}function AI(e){return ss(e,()=>e.createTexture(),"Unable to create WebGLTexture.")}function xI(e,t){let r=Z().getNumber("WEBGL_MAX_TEXTURE_SIZE");if(e<=0||t<=0){let n=`[${e}x${t}]`;throw new Error("Requested texture size "+n+" is invalid.")}if(e>r||t>r){let n=`[${e}x${t}]`,a=`[${r}x${r}]`;throw new Error("Requested texture size "+n+" greater than WebGL maximum on this browser / GPU "+a+".")}}function bI(e){return ss(e,()=>e.createFramebuffer(),"Unable to create WebGLFramebuffer.")}function ey(e,t,r,n,a,s,i){let o=e.getAttribLocation(t,r);return o===-1?!1:(we(e,()=>e.bindBuffer(e.ARRAY_BUFFER,n)),we(e,()=>e.vertexAttribPointer(o,a,e.FLOAT,!1,s,i)),we(e,()=>e.enableVertexAttribArray(o)),!0)}function vI(e,t,r){CI(e,r),we(e,()=>e.activeTexture(e.TEXTURE0+r)),we(e,()=>e.bindTexture(e.TEXTURE_2D,t))}function _Q(e,t){CI(e,t),we(e,()=>e.activeTexture(e.TEXTURE0+t)),we(e,()=>e.bindTexture(e.TEXTURE_2D,null))}function wI(e,t,r){return ss(e,()=>e.getUniformLocation(t,r),'uniform "'+r+'" not present in program.')}function kI(e,t,r){return e.getUniformLocation(t,r)}function II(e,t,r,n){we(e,()=>vI(e,t,n)),we(e,()=>e.uniform1i(r,n))}function zQ(e){we(e,()=>e.bindFramebuffer(e.FRAMEBUFFER,null)),we(e,()=>e.viewport(0,0,e.canvas.width,e.canvas.height)),we(e,()=>e.scissor(0,0,e.canvas.width,e.canvas.height))}function u0(e,t,r){we(e,()=>e.bindFramebuffer(e.FRAMEBUFFER,r)),we(e,()=>e.framebufferTexture2D(e.FRAMEBUFFER,e.COLOR_ATTACHMENT0,e.TEXTURE_2D,t,0))}function ty(e,t){we(e,()=>e.bindFramebuffer(e.FRAMEBUFFER,t)),we(e,()=>e.framebufferTexture2D(e.FRAMEBUFFER,e.COLOR_ATTACHMENT0,e.TEXTURE_2D,null,0))}function Ep(e){let t=e.checkFramebufferStatus(e.FRAMEBUFFER);if(t!==e.FRAMEBUFFER_COMPLETE)throw new Error("Error binding framebuffer: "+SI(e,t))}function SI(e,t){switch(t){case e.FRAMEBUFFER_INCOMPLETE_ATTACHMENT:return"FRAMEBUFFER_INCOMPLETE_ATTACHMENT";case e.FRAMEBUFFER_INCOMPLETE_MISSING_ATTACHMENT:return"FRAMEBUFFER_INCOMPLETE_MISSING_ATTACHMENT";case e.FRAMEBUFFER_INCOMPLETE_DIMENSIONS:return"FRAMEBUFFER_INCOMPLETE_DIMENSIONS";case e.FRAMEBUFFER_UNSUPPORTED:return"FRAMEBUFFER_UNSUPPORTED";default:return`unknown error ${t}`}}function ss(e,t,r){let n=we(e,()=>t());if(n==null)throw new Error(r);return n}function CI(e,t){let r=e.MAX_COMBINED_TEXTURE_IMAGE_UNITS-1,n=t+e.TEXTURE0;if(n<e.TEXTURE0||n>r){let a=`[gl.TEXTURE0, gl.TEXTURE${r}]`;throw new Error(`textureUnit must be in ${a}.`)}}function Wo(e,t=2){return v.sizeFromShape(e.slice(0,e.length-t))}function Vo(e){if(e.length===0)throw Error("Cannot get rows and columns of an empty shape array.");return[e.length>1?e[e.length-2]:1,e[e.length-1]]}function d0(e){let t=[1,1,1];return e.length===0||e.length===1&&e[0]===1||(t=[Wo(e),...Vo(e)]),t}function TI(e,t=!1){let r=Z().getNumber("WEBGL_MAX_TEXTURE_SIZE");t&&(r=r*2,e=e.map((a,s)=>s>=e.length-2?v.nearestLargerEven(e[s]):e[s]),e.length===1&&(e=[2,e[0]])),e.length!==2&&(e=v.squeezeShape(e).newShape);let n=v.sizeFromShape(e);if(e.length<=1&&n<=r)return[1,n];if(e.length===2&&e[0]<=r&&e[1]<=r)return e;if(e.length===3&&e[0]*e[1]<=r&&e[2]<=r)return[e[0]*e[1],e[2]];if(e.length===3&&e[0]<=r&&e[1]*e[2]<=r)return[e[0],e[1]*e[2]];if(e.length===4&&e[0]*e[1]*e[2]<=r&&e[3]<=r)return[e[0]*e[1]*e[2],e[3]];if(e.length===4&&e[0]<=r&&e[1]*e[2]*e[3]<=r)return[e[0],e[1]*e[2]*e[3]];if(t){let a=Wo(e),s=2,i=2;return e.length&&([s,i]=Vo(e)),n=a*(s/2)*(i/2),v.sizeToSquarishShape(n).map(o=>o*2)}return v.sizeToSquarishShape(n)}function r0(e){return e%2===0}function Xp(e,t){if(e=e.slice(-2),t=t.slice(-2),v.arraysEqual(e,t)||!e.length||!t.length||e[0]===0||e[1]===0||t[0]===0||t[1]===0)return!0;if(e.length!==t.length){let r=e.slice(-1)[0],n=t.slice(-1)[0];if(r===n||r0(r)&&r0(n)&&(e[0]===1||t[0]===1))return!0}return e[1]===t[1]&&r0(e[0])&&r0(t[0])}var p0,h0;function NI(e){if(p0==null){let t=xa(e);p0=t.getParameter(t.MAX_TEXTURE_SIZE)}return p0}function OQ(){p0=null}function DQ(){h0=null}function EI(e){if(h0==null){let t=xa(e);h0=t.getParameter(t.MAX_TEXTURE_IMAGE_UNITS)}return Math.min(16,h0)}function RI(e){if(e===0)return 0;let t,r=xa(e);return En(r,"EXT_disjoint_timer_query_webgl2")&&e===2?t=2:En(r,"EXT_disjoint_timer_query")?t=1:t=0,t}function En(e,t){return e.getExtension(t)!=null}function ry(e){try{if(xa(e)!=null)return!0}catch(t){return console.log("Error when getting WebGL context: ",t),!1}return!1}function $I(e){if(e===0)return!1;let t=xa(e);if(e===1){if(!En(t,"OES_texture_float"))return!1}else if(!En(t,"EXT_color_buffer_float"))return!1;return ny(t)}function MI(e){if(e===0)return!1;let t=xa(e);if(e===1){if(!En(t,"OES_texture_float")||!En(t,"WEBGL_color_buffer_float"))return!1}else{if(En(t,"EXT_color_buffer_float"))return ny(t);let r="EXT_color_buffer_half_float";if(En(t,r)){let n=t.getExtension(r);return LQ(t,n)}return!1}return ny(t)}function ny(e){let t=h5(e),r=e.createTexture();e.bindTexture(e.TEXTURE_2D,r);let n=1,a=1;e.texImage2D(e.TEXTURE_2D,0,t.internalFormatFloat,n,a,0,t.textureFormatFloat,t.textureTypeFloat,null);let s=e.createFramebuffer();e.bindFramebuffer(e.FRAMEBUFFER,s),e.framebufferTexture2D(e.FRAMEBUFFER,e.COLOR_ATTACHMENT0,e.TEXTURE_2D,r,0);let i=e.checkFramebufferStatus(e.FRAMEBUFFER)===e.FRAMEBUFFER_COMPLETE;return e.bindTexture(e.TEXTURE_2D,null),e.bindFramebuffer(e.FRAMEBUFFER,null),e.deleteTexture(r),e.deleteFramebuffer(s),i}function LQ(e,t){let r=h5(e,t),n=e.createTexture();e.bindTexture(e.TEXTURE_2D,n);let a=1,s=1;e.texImage2D(e.TEXTURE_2D,0,r.internalFormatHalfFloat,a,s,0,r.textureFormatFloat,r.textureTypeHalfFloat,null);let i=e.createFramebuffer();e.bindFramebuffer(e.FRAMEBUFFER,i),e.framebufferTexture2D(e.FRAMEBUFFER,e.COLOR_ATTACHMENT0,e.TEXTURE_2D,n,0);let o=e.checkFramebufferStatus(e.FRAMEBUFFER)===e.FRAMEBUFFER_COMPLETE;return e.bindTexture(e.TEXTURE_2D,null),e.bindFramebuffer(e.FRAMEBUFFER,null),e.deleteTexture(n),e.deleteFramebuffer(i),o}function FI(e){return e!==2?!1:xa(e).fenceSync!=null}function Id(e,t){Array.isArray(e)||(e=[e]),e.forEach(r=>{r!=null&&v.assert(r.dtype!=="complex64",()=>`${t} does not support complex64 tensors in the WebGL backend.`)})}var Me=Z();Me.registerFlag("HAS_WEBGL",()=>Me.getNumber("WEBGL_VERSION")>0);Me.registerFlag("WEBGL_VERSION",()=>ry(2)?2:ry(1)?1:0);Me.registerFlag("WEBGL_CHECK_NUMERICAL_PROBLEMS",()=>!1);Me.registerFlag("WEBGL_BUFFER_SUPPORTED",()=>Me.get("WEBGL_VERSION")===2);Me.registerFlag("WEBGL_CPU_FORWARD",()=>!0);Me.registerFlag("WEBGL_FORCE_F16_TEXTURES",()=>!1);Me.registerFlag("WEBGL_PACK",()=>Me.getBool("HAS_WEBGL"));Me.registerFlag("WEBGL_PACK_NORMALIZATION",()=>Me.getBool("WEBGL_PACK"));Me.registerFlag("WEBGL_PACK_CLIP",()=>Me.getBool("WEBGL_PACK"));Me.registerFlag("WEBGL_PACK_DEPTHWISECONV",()=>Me.getBool("WEBGL_PACK"));Me.registerFlag("WEBGL_PACK_BINARY_OPERATIONS",()=>Me.getBool("WEBGL_PACK"));Me.registerFlag("WEBGL_PACK_UNARY_OPERATIONS",()=>Me.getBool("WEBGL_PACK"));Me.registerFlag("WEBGL_PACK_ARRAY_OPERATIONS",()=>Me.getBool("WEBGL_PACK"));Me.registerFlag("WEBGL_PACK_IMAGE_OPERATIONS",()=>Me.getBool("WEBGL_PACK"));Me.registerFlag("WEBGL_PACK_REDUCE",()=>Me.getBool("WEBGL_PACK"));Me.registerFlag("WEBGL_LAZILY_UNPACK",()=>Me.getBool("WEBGL_PACK"));Me.registerFlag("WEBGL_CONV_IM2COL",()=>Me.getBool("WEBGL_PACK"));Me.registerFlag("WEBGL_MAX_TEXTURE_SIZE",()=>NI(Me.getNumber("WEBGL_VERSION")));Me.registerFlag("WEBGL_MAX_TEXTURES_IN_SHADER",()=>EI(Me.getNumber("WEBGL_VERSION")));Me.registerFlag("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION",()=>{let e=Me.getNumber("WEBGL_VERSION");return e===0?0:RI(e)});Me.registerFlag("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE",()=>Me.getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION")>0&&!yh.isMobile());Me.registerFlag("WEBGL_RENDER_FLOAT32_CAPABLE",()=>$I(Me.getNumber("WEBGL_VERSION")));Me.registerFlag("WEBGL_RENDER_FLOAT32_ENABLED",()=>Me.getBool("WEBGL_FORCE_F16_TEXTURES")?!1:Me.getBool("WEBGL_RENDER_FLOAT32_CAPABLE"));Me.registerFlag("WEBGL_DOWNLOAD_FLOAT_ENABLED",()=>MI(Me.getNumber("WEBGL_VERSION")));Me.registerFlag("WEBGL_FENCE_API_ENABLED",()=>FI(Me.getNumber("WEBGL_VERSION")));Me.registerFlag("WEBGL_SIZE_UPLOAD_UNIFORM",()=>Me.getBool("WEBGL_RENDER_FLOAT32_ENABLED")?4:0);Me.registerFlag("WEBGL_DELETE_TEXTURE_THRESHOLD",()=>-1,e=>{if(e<0&&e!==-1)throw new Error(`WEBGL_DELETE_TEXTURE_THRESHOLD must be -1 (indicating never delete) or at least 0, but got ${e}.`)});Me.registerFlag("WEBGL_FLUSH_THRESHOLD",()=>yh.isMobile()?1:-1,e=>{if(e<0&&e!==-1)throw new Error(`WEBGL_FLUSH_THRESHOLD must be -1 (indicating never manual flush) or at least 0, but got ${e}.`)});Me.registerFlag("CPU_HANDOFF_SIZE_THRESHOLD",()=>128);Me.registerFlag("WEBGL_USE_SHAPES_UNIFORMS",()=>!1);Me.registerFlag("TOPK_LAST_DIM_CPU_HANDOFF_SIZE_THRESHOLD",()=>1e5);Me.registerFlag("TOPK_K_CPU_HANDOFF_THRESHOLD",()=>128);function Hr(){let e,t,r,n,a,s,i,o,l,u;return Z().getNumber("WEBGL_VERSION")===2?(e="#version 300 es",t="in",r="out",n="in",a="texture",s="outputColor",i="out vec4 outputColor;",o=`
|
|
bool isnan_custom(float val) {
|
|
uint floatToUint = floatBitsToUint(val);
|
|
return (floatToUint & 0x7fffffffu) > 0x7f800000u;
|
|
}
|
|
|
|
bvec4 isnan_custom(vec4 val) {
|
|
return bvec4(isnan_custom(val.x),
|
|
isnan_custom(val.y), isnan_custom(val.z), isnan_custom(val.w));
|
|
}
|
|
|
|
#define isnan(value) isnan_custom(value)
|
|
`,l="",u=`
|
|
#define round(value) newRound(value)
|
|
int newRound(float value) {
|
|
return int(floor(value + 0.5));
|
|
}
|
|
|
|
ivec4 newRound(vec4 value) {
|
|
return ivec4(floor(value + vec4(0.5)));
|
|
}
|
|
`):(e="",t="attribute",r="varying",n="varying",a="texture2D",s="gl_FragColor",i="",o=`
|
|
#define isnan(value) isnan_custom(value)
|
|
bool isnan_custom(float val) {
|
|
return (val > 0. || val < 1. || val == 0.) ? false : true;
|
|
}
|
|
bvec4 isnan_custom(vec4 val) {
|
|
return bvec4(isnan(val.x), isnan(val.y), isnan(val.z), isnan(val.w));
|
|
}
|
|
`,l=`
|
|
uniform float INFINITY;
|
|
|
|
bool isinf(float val) {
|
|
return abs(val) == INFINITY;
|
|
}
|
|
bvec4 isinf(vec4 val) {
|
|
return equal(abs(val), vec4(INFINITY));
|
|
}
|
|
`,u=`
|
|
int round(float value) {
|
|
return int(floor(value + 0.5));
|
|
}
|
|
|
|
ivec4 round(vec4 value) {
|
|
return ivec4(floor(value + vec4(0.5)));
|
|
}
|
|
`),{version:e,attribute:t,varyingVs:r,varyingFs:n,texture2D:a,output:s,defineOutput:i,defineSpecialNaN:o,defineSpecialInf:l,defineRound:u}}function Dl(e,t,r="index"){let n=v.computeStrides(t);return n.map((a,s)=>{let i=`int ${e[s]} = ${r} / ${a}`,o=s===n.length-1?`int ${e[s+1]} = ${r} - ${e[s]} * ${a}`:`index -= ${e[s]} * ${a}`;return`${i}; ${o};`}).join("")}function Sm(e,t,r="index"){let n=v.computeStrides(t);return n.map((a,s)=>{let i=`int ${e[s]} = ${r} / outShapeStrides[${s}]`,o=s===n.length-1?`int ${e[s+1]} = ${r} - ${e[s]} * outShapeStrides[${s}]`:`index -= ${e[s]} * outShapeStrides[${s}]`;return`${i}; ${o};`}).join("")}function BQ(e,t){let r=e.length,n=e.map(s=>`${t}[${s}]`),a=new Array(r-1);a[r-2]=n[r-1];for(let s=r-3;s>=0;--s)a[s]=`(${a[s+1]} * ${n[s+1]})`;return a}function WQ(e,t,r="index"){let n=e.map((s,i)=>i),a=BQ(n,t);return a.map((s,i)=>{let o=`int ${e[i]} = ${r} / ${a[i]}`,l=i===a.length-1?`int ${e[i+1]} = ${r} - ${e[i]} * ${a[i]}`:`index -= ${e[i]} * ${a[i]}`;return`${o}; ${l};`}).join("")}function f5(e){let t=v.computeStrides(e).map(r=>r.toString());return`
|
|
int getFlatIndex(ivec3 coords) {
|
|
return coords.x * ${t[0]} + coords.y * ${t[1]} + coords.z;
|
|
}
|
|
`}function m5(){return`
|
|
int getFlatIndex(ivec3 coords) {
|
|
return coords.x * outShapeStrides[0] + coords.y * outShapeStrides[1] + coords.z;
|
|
}
|
|
`}var PI=`
|
|
const float FLOAT_MAX = 1.70141184e38;
|
|
const float FLOAT_MIN = 1.17549435e-38;
|
|
|
|
lowp vec4 encode_float(highp float v) {
|
|
if (isnan(v)) {
|
|
return vec4(255, 255, 255, 255);
|
|
}
|
|
|
|
highp float av = abs(v);
|
|
|
|
if(av < FLOAT_MIN) {
|
|
return vec4(0.0, 0.0, 0.0, 0.0);
|
|
} else if(v > FLOAT_MAX) {
|
|
return vec4(0.0, 0.0, 128.0, 127.0) / 255.0;
|
|
} else if(v < -FLOAT_MAX) {
|
|
return vec4(0.0, 0.0, 128.0, 255.0) / 255.0;
|
|
}
|
|
|
|
highp vec4 c = vec4(0,0,0,0);
|
|
|
|
highp float e = floor(log2(av));
|
|
highp float m = exp2(fract(log2(av))) - 1.0;
|
|
|
|
c[2] = floor(128.0 * m);
|
|
m -= c[2] / 128.0;
|
|
c[1] = floor(32768.0 * m);
|
|
m -= c[1] / 32768.0;
|
|
c[0] = floor(8388608.0 * m);
|
|
|
|
highp float ebias = e + 127.0;
|
|
c[3] = floor(ebias / 2.0);
|
|
ebias -= c[3] * 2.0;
|
|
c[2] += floor(ebias) * 128.0;
|
|
|
|
c[3] += 128.0 * step(0.0, -v);
|
|
|
|
return c / 255.0;
|
|
}
|
|
`,{getBroadcastDims:_I}=C;function VQ(e,t,r){let n=[];if(e.forEach(p=>{let c=v.sizeFromShape(p.shapeInfo.logicalShape);if(p.shapeInfo.isUniform?n.push(`uniform float ${p.name}${c>1?`[${c}]`:""};`):(n.push(`uniform sampler2D ${p.name};`),n.push(`uniform int offset${p.name};`)),r.enableShapeUniforms){let{uniformShape:f}=g5(r.packedInputs,p.shapeInfo.logicalShape,p.shapeInfo.texShape);switch(f.length){case 1:n.push(`uniform int ${p.name}Shape;`);break;case 2:n.push(`uniform ivec2 ${p.name}Shape;`);break;case 3:n.push(`uniform ivec3 ${p.name}Shape;`);break;case 4:n.push(`uniform ivec4 ${p.name}Shape;`);break;default:break}n.push(`uniform ivec2 ${p.name}TexShape;`)}}),r.enableShapeUniforms){switch(t.logicalShape.length){case 1:n.push("uniform int outShape;");break;case 2:n.push("uniform ivec2 outShape;"),n.push("uniform int outShapeStrides;");break;case 3:n.push("uniform ivec3 outShape;"),n.push("uniform ivec2 outShapeStrides;");break;case 4:n.push("uniform ivec4 outShape;"),n.push("uniform ivec3 outShapeStrides;");break;default:break}n.push("uniform ivec2 outTexShape;")}r.customUniforms&&r.customUniforms.forEach(p=>{n.push(`uniform ${p.type} ${p.name}${p.arrayIndex?`[${p.arrayIndex}]`:""};`)});let a=n.join(`
|
|
`),s=e.map(p=>UQ(p,t,r.packedInputs,r.enableShapeUniforms)).join(`
|
|
`),i=t.texShape,o=Hr(),l=HQ(o),u,d,h=KQ(o);return t.isPacked?(u=GQ(t.logicalShape,i,r.enableShapeUniforms),d=XQ(o)):(u=jQ(t.logicalShape,i,r.enableShapeUniforms),d=qQ(o)),r.packedInputs&&(h+=QQ),[h,l,d,a,u,s,r.userCode].join(`
|
|
`)}function Sd(e,t=!1){let r=e.shapeInfo.logicalShape;switch(r.length){case 0:return pee(e,t);case 1:return cee(e,t);case 2:return mee(e,t);case 3:return yee(e,t);case 4:return xee(e,t);case 5:return bee(e);case 6:return vee(e);default:throw new Error(`${r.length}-D input sampling is not yet supported`)}}function zI(e,t){switch(e.shapeInfo.logicalShape.length){case 0:return dee(e);case 1:return hee(e,t);case 2:return fee(e,t);case 3:return gee(e,t);default:return Aee(e,t)}}function UQ(e,t,r=!1,n){let a="";r?a+=zI(e,n):a+=Sd(e,n);let s=e.shapeInfo.logicalShape,i=t.logicalShape;return s.length<=i.length&&(r?a+=wee(e,t):a+=kee(e,t)),a}function GQ(e,t,r){switch(e.length){case 0:return OI();case 1:return eee(e,t,r);case 2:return lee(e,t,r);case 3:return ree(e,t,r);default:return aee(e,t,r)}}function jQ(e,t,r){switch(e.length){case 0:return OI();case 1:return tee(e,t,r);case 2:return uee(e,t,r);case 3:return nee(e,t,r);case 4:return see(e,t,r);case 5:return iee(e,t);case 6:return oee(e,t);default:throw new Error(`${e.length}-D output sampling is not yet supported`)}}function HQ(e){return`
|
|
float sampleTexture(sampler2D textureSampler, vec2 uv) {
|
|
return ${e.texture2D}(textureSampler, uv).r;
|
|
}
|
|
`}function qQ(e){return`
|
|
void setOutput(float val) {
|
|
${e.output} = vec4(val, 0, 0, 0);
|
|
}
|
|
`}function XQ(e){return`
|
|
void setOutput(vec4 val) {
|
|
${e.output} = val;
|
|
}
|
|
`}function KQ(e){return`${e.version}
|
|
precision highp float;
|
|
precision highp int;
|
|
precision highp sampler2D;
|
|
${e.varyingFs} vec2 resultUV;
|
|
${e.defineOutput}
|
|
const vec2 halfCR = vec2(0.5, 0.5);
|
|
|
|
struct ivec5
|
|
{
|
|
int x;
|
|
int y;
|
|
int z;
|
|
int w;
|
|
int u;
|
|
};
|
|
|
|
struct ivec6
|
|
{
|
|
int x;
|
|
int y;
|
|
int z;
|
|
int w;
|
|
int u;
|
|
int v;
|
|
};
|
|
|
|
uniform float NAN;
|
|
${e.defineSpecialNaN}
|
|
${e.defineSpecialInf}
|
|
${e.defineRound}
|
|
|
|
int imod(int x, int y) {
|
|
return x - y * (x / y);
|
|
}
|
|
|
|
int idiv(int a, int b, float sign) {
|
|
int res = a / b;
|
|
int mod = imod(a, b);
|
|
if (sign < 0. && mod != 0) {
|
|
res -= 1;
|
|
}
|
|
return res;
|
|
}
|
|
|
|
//Based on the work of Dave Hoskins
|
|
//https://www.shadertoy.com/view/4djSRW
|
|
#define HASHSCALE1 443.8975
|
|
float random(float seed){
|
|
vec2 p = resultUV * seed;
|
|
vec3 p3 = fract(vec3(p.xyx) * HASHSCALE1);
|
|
p3 += dot(p3, p3.yzx + 19.19);
|
|
return fract((p3.x + p3.y) * p3.z);
|
|
}
|
|
|
|
${ZQ}
|
|
${YQ}
|
|
${JQ}
|
|
`}var ZQ=`
|
|
vec2 uvFromFlat(int texNumR, int texNumC, int index) {
|
|
int texR = index / texNumC;
|
|
int texC = index - texR * texNumC;
|
|
return (vec2(texC, texR) + halfCR) / vec2(texNumC, texNumR);
|
|
}
|
|
vec2 packedUVfrom1D(int texNumR, int texNumC, int index) {
|
|
int texelIndex = index / 2;
|
|
int texR = texelIndex / texNumC;
|
|
int texC = texelIndex - texR * texNumC;
|
|
return (vec2(texC, texR) + halfCR) / vec2(texNumC, texNumR);
|
|
}
|
|
`,YQ=`
|
|
vec2 packedUVfrom2D(int texelsInLogicalRow, int texNumR,
|
|
int texNumC, int row, int col) {
|
|
int texelIndex = (row / 2) * texelsInLogicalRow + (col / 2);
|
|
int texR = texelIndex / texNumC;
|
|
int texC = texelIndex - texR * texNumC;
|
|
return (vec2(texC, texR) + halfCR) / vec2(texNumC, texNumR);
|
|
}
|
|
`,JQ=`
|
|
vec2 packedUVfrom3D(int texNumR, int texNumC,
|
|
int texelsInBatch, int texelsInLogicalRow, int b,
|
|
int row, int col) {
|
|
int index = b * texelsInBatch + (row / 2) * texelsInLogicalRow + (col / 2);
|
|
int texR = index / texNumC;
|
|
int texC = index - texR * texNumC;
|
|
return (vec2(texC, texR) + halfCR) / vec2(texNumC, texNumR);
|
|
}
|
|
`,QQ=`
|
|
float getChannel(vec4 frag, vec2 innerDims) {
|
|
vec2 modCoord = mod(innerDims, 2.);
|
|
return modCoord.x == 0. ?
|
|
(modCoord.y == 0. ? frag.r : frag.g) :
|
|
(modCoord.y == 0. ? frag.b : frag.a);
|
|
}
|
|
float getChannel(vec4 frag, int dim) {
|
|
float modCoord = mod(float(dim), 2.);
|
|
return modCoord == 0. ? frag.r : frag.g;
|
|
}
|
|
`;function OI(){return`
|
|
int getOutputCoords() {
|
|
return 0;
|
|
}
|
|
`}function eee(e,t,r){let n=[Math.ceil(t[0]/2),Math.ceil(t[1]/2)];return n[0]===1?r?`
|
|
int getOutputCoords() {
|
|
return 2 * int(resultUV.x * ceil(float(outTexShape[1]) / 2.0));
|
|
}
|
|
`:`
|
|
int getOutputCoords() {
|
|
return 2 * int(resultUV.x * ${n[1]}.0);
|
|
}
|
|
`:n[1]===1?r?`
|
|
int getOutputCoords() {
|
|
return 2 * int(resultUV.y * ceil(float(outTexShape[0]) / 2.0));
|
|
}
|
|
`:`
|
|
int getOutputCoords() {
|
|
return 2 * int(resultUV.y * ${n[0]}.0);
|
|
}
|
|
`:r?`
|
|
int getOutputCoords() {
|
|
ivec2 packedTexShape = ivec2(ceil(float(outTexShape[0]) / 2.0), ceil(float(outTexShape[1]) / 2.0));
|
|
ivec2 resTexRC = ivec2(resultUV.yx *
|
|
vec2(packedTexShape[0], packedTexShape[1]));
|
|
return 2 * (resTexRC.x * packedTexShape[1] + resTexRC.y);
|
|
}
|
|
`:`
|
|
int getOutputCoords() {
|
|
ivec2 resTexRC = ivec2(resultUV.yx *
|
|
vec2(${n[0]}, ${n[1]}));
|
|
return 2 * (resTexRC.x * ${n[1]} + resTexRC.y);
|
|
}
|
|
`}function tee(e,t,r){return t[0]===1?r?`
|
|
int getOutputCoords() {
|
|
return int(resultUV.x * float(outTexShape[1]));
|
|
}
|
|
`:`
|
|
int getOutputCoords() {
|
|
return int(resultUV.x * ${t[1]}.0);
|
|
}
|
|
`:t[1]===1?r?`
|
|
int getOutputCoords() {
|
|
return int(resultUV.y * float(outTexShape[0]));
|
|
}
|
|
`:`
|
|
int getOutputCoords() {
|
|
return int(resultUV.y * ${t[0]}.0);
|
|
}
|
|
`:r?`
|
|
int getOutputCoords() {
|
|
ivec2 resTexRC = ivec2(resultUV.yx *
|
|
vec2(outTexShape[0], outTexShape[1]));
|
|
return resTexRC.x * outTexShape[1] + resTexRC.y;
|
|
}
|
|
`:`
|
|
int getOutputCoords() {
|
|
ivec2 resTexRC = ivec2(resultUV.yx *
|
|
vec2(${t[0]}, ${t[1]}));
|
|
return resTexRC.x * ${t[1]} + resTexRC.y;
|
|
}
|
|
`}function ree(e,t,r){if(r)return`
|
|
ivec3 getOutputCoords() {
|
|
ivec2 packedTexShape = ivec2(ceil(float(outTexShape[0]) / 2.0), ceil(float(outTexShape[1]) / 2.0));
|
|
int texelsInLogicalRow = int(ceil(float(outShape[2]) / 2.0));
|
|
int texelsInBatch = texelsInLogicalRow * int(ceil(float(outShape[1]) / 2.0));
|
|
ivec2 resTexRC = ivec2(resultUV.yx *
|
|
vec2(packedTexShape[0], packedTexShape[1]));
|
|
int index = resTexRC.x * packedTexShape[1] + resTexRC.y;
|
|
|
|
int b = index / texelsInBatch;
|
|
index -= b * texelsInBatch;
|
|
|
|
int r = 2 * (index / texelsInLogicalRow);
|
|
int c = imod(index, texelsInLogicalRow) * 2;
|
|
|
|
return ivec3(b, r, c);
|
|
}
|
|
`;let n=[Math.ceil(t[0]/2),Math.ceil(t[1]/2)],a=Math.ceil(e[2]/2),s=a*Math.ceil(e[1]/2);return`
|
|
ivec3 getOutputCoords() {
|
|
ivec2 resTexRC = ivec2(resultUV.yx *
|
|
vec2(${n[0]}, ${n[1]}));
|
|
int index = resTexRC.x * ${n[1]} + resTexRC.y;
|
|
|
|
int b = index / ${s};
|
|
index -= b * ${s};
|
|
|
|
int r = 2 * (index / ${a});
|
|
int c = imod(index, ${a}) * 2;
|
|
|
|
return ivec3(b, r, c);
|
|
}
|
|
`}function nee(e,t,r){if(r)return`
|
|
ivec3 getOutputCoords() {
|
|
ivec2 resTexRC = ivec2(resultUV.yx *
|
|
vec2(outTexShape[0], outTexShape[1]));
|
|
int index = resTexRC.x * outTexShape[1] + resTexRC.y;
|
|
${Sm(["r","c","d"],e)}
|
|
return ivec3(r, c, d);
|
|
}
|
|
`;let n=Dl(["r","c","d"],e);return`
|
|
ivec3 getOutputCoords() {
|
|
ivec2 resTexRC = ivec2(resultUV.yx *
|
|
vec2(${t[0]}, ${t[1]}));
|
|
int index = resTexRC.x * ${t[1]} + resTexRC.y;
|
|
${n}
|
|
return ivec3(r, c, d);
|
|
}
|
|
`}function aee(e,t,r){if(r)return`
|
|
ivec4 getOutputCoords() {
|
|
ivec2 packedTexShape = ivec2(ceil(float(outTexShape[0]) / 2.0), ceil(float(outTexShape[1]) / 2.0));
|
|
ivec2 resTexRC = ivec2(resultUV.yx *
|
|
vec2(packedTexShape[0], packedTexShape[1]));
|
|
int index = resTexRC.x * packedTexShape[1] + resTexRC.y;
|
|
|
|
int texelsInLogicalRow = int(ceil(float(outShape[3]) / 2.0));
|
|
int texelsInBatch = texelsInLogicalRow * int(ceil(float(outShape[2]) / 2.0));
|
|
int texelsInBatchN = texelsInBatch * outShape[1];
|
|
|
|
int b2 = index / texelsInBatchN;
|
|
index -= b2 * texelsInBatchN;
|
|
|
|
int b = index / texelsInBatch;
|
|
index -= b * texelsInBatch;
|
|
|
|
int r = 2 * (index / texelsInLogicalRow);
|
|
int c = imod(index, texelsInLogicalRow) * 2;
|
|
|
|
return ivec4(b2, b, r, c);
|
|
}
|
|
`;let n=[Math.ceil(t[0]/2),Math.ceil(t[1]/2)],a=Math.ceil(e[e.length-1]/2),s=a*Math.ceil(e[e.length-2]/2),i=s,o="",l="b, r, c";for(let u=2;u<e.length-1;u++)i*=e[e.length-u-1],o=`
|
|
int b${u} = index / ${i};
|
|
index -= b${u} * ${i};
|
|
`+o,l=`b${u}, `+l;return`
|
|
ivec${e.length} getOutputCoords() {
|
|
ivec2 resTexRC = ivec2(resultUV.yx *
|
|
vec2(${n[0]}, ${n[1]}));
|
|
int index = resTexRC.x * ${n[1]} + resTexRC.y;
|
|
|
|
${o}
|
|
|
|
int b = index / ${s};
|
|
index -= b * ${s};
|
|
|
|
int r = 2 * (index / ${a});
|
|
int c = imod(index, ${a}) * 2;
|
|
|
|
return ivec${e.length}(${l});
|
|
}
|
|
`}function see(e,t,r){if(r)return`
|
|
ivec4 getOutputCoords() {
|
|
ivec2 resTexRC = ivec2(resultUV.yx *
|
|
vec2(outTexShape[0], outTexShape[1]));
|
|
int index = resTexRC.x * outTexShape[1] + resTexRC.y;
|
|
${Sm(["r","c","d","d2"],e)}
|
|
return ivec4(r, c, d, d2);
|
|
}
|
|
`;let n=Dl(["r","c","d","d2"],e);return`
|
|
ivec4 getOutputCoords() {
|
|
ivec2 resTexRC = ivec2(resultUV.yx *
|
|
vec2(${t[0]}, ${t[1]}));
|
|
int index = resTexRC.x * ${t[1]} + resTexRC.y;
|
|
${n}
|
|
return ivec4(r, c, d, d2);
|
|
}
|
|
`}function iee(e,t){let r=Dl(["r","c","d","d2","d3"],e);return`
|
|
ivec5 getOutputCoords() {
|
|
ivec2 resTexRC = ivec2(resultUV.yx * vec2(${t[0]},
|
|
${t[1]}));
|
|
|
|
int index = resTexRC.x * ${t[1]} + resTexRC.y;
|
|
|
|
${r}
|
|
|
|
ivec5 outShape = ivec5(r, c, d, d2, d3);
|
|
return outShape;
|
|
}
|
|
`}function oee(e,t){let r=Dl(["r","c","d","d2","d3","d4"],e);return`
|
|
ivec6 getOutputCoords() {
|
|
ivec2 resTexRC = ivec2(resultUV.yx *
|
|
vec2(${t[0]}, ${t[1]}));
|
|
int index = resTexRC.x * ${t[1]} + resTexRC.y;
|
|
|
|
${r}
|
|
|
|
ivec6 result = ivec6(r, c, d, d2, d3, d4);
|
|
return result;
|
|
}
|
|
`}function lee(e,t,r){let n=[Math.ceil(t[0]/2),Math.ceil(t[1]/2)];if(v.arraysEqual(e,t))return r?`
|
|
ivec2 getOutputCoords() {
|
|
ivec2 packedTexShape = ivec2(ceil(float(outTexShape[0]) / 2.0), ceil(float(outTexShape[1]) / 2.0));
|
|
return 2 * ivec2(resultUV.yx * vec2(packedTexShape[0], packedTexShape[1]));
|
|
}
|
|
`:`
|
|
ivec2 getOutputCoords() {
|
|
return 2 * ivec2(resultUV.yx * vec2(${n[0]}, ${n[1]}));
|
|
}
|
|
`;let a=Math.ceil(e[1]/2);return r?`
|
|
ivec2 getOutputCoords() {
|
|
ivec2 packedTexShape = ivec2(ceil(float(outTexShape[0]) / 2.0), ceil(float(outTexShape[1]) / 2.0));
|
|
int texelsInLogicalRow = int(ceil(float(outShape[1]) / 2.0));
|
|
ivec2 resTexRC = ivec2(resultUV.yx *
|
|
vec2(packedTexShape[0], packedTexShape[1]));
|
|
|
|
int index = resTexRC.x * packedTexShape[1] + resTexRC.y;
|
|
int r = 2 * (index / texelsInLogicalRow);
|
|
int c = imod(index, texelsInLogicalRow) * 2;
|
|
|
|
return ivec2(r, c);
|
|
}
|
|
`:`
|
|
ivec2 getOutputCoords() {
|
|
ivec2 resTexRC = ivec2(resultUV.yx *
|
|
vec2(${n[0]}, ${n[1]}));
|
|
|
|
int index = resTexRC.x * ${n[1]} + resTexRC.y;
|
|
int r = 2 * (index / ${a});
|
|
int c = imod(index, ${a}) * 2;
|
|
|
|
return ivec2(r, c);
|
|
}
|
|
`}function uee(e,t,r){return v.arraysEqual(e,t)?r?`
|
|
ivec2 getOutputCoords() {
|
|
return ivec2(resultUV.yx * vec2(outTexShape[0], outTexShape[1]));
|
|
}
|
|
`:`
|
|
ivec2 getOutputCoords() {
|
|
return ivec2(resultUV.yx * vec2(${t[0]}, ${t[1]}));
|
|
}
|
|
`:e[1]===1?r?`
|
|
ivec2 getOutputCoords() {
|
|
ivec2 resTexRC = ivec2(resultUV.yx *
|
|
vec2(outTexShape[0], outTexShape[1]));
|
|
int index = resTexRC.x * outTexShape[1] + resTexRC.y;
|
|
return ivec2(index, 0);
|
|
}
|
|
`:`
|
|
ivec2 getOutputCoords() {
|
|
ivec2 resTexRC = ivec2(resultUV.yx *
|
|
vec2(${t[0]}, ${t[1]}));
|
|
int index = resTexRC.x * ${t[1]} + resTexRC.y;
|
|
return ivec2(index, 0);
|
|
}
|
|
`:e[0]===1?r?`
|
|
ivec2 getOutputCoords() {
|
|
ivec2 resTexRC = ivec2(resultUV.yx *
|
|
vec2(outTexShape[0], outTexShape[1]));
|
|
int index = resTexRC.x * outTexShape[1] + resTexRC.y;
|
|
return ivec2(0, index);
|
|
}
|
|
`:`
|
|
ivec2 getOutputCoords() {
|
|
ivec2 resTexRC = ivec2(resultUV.yx *
|
|
vec2(${t[0]}, ${t[1]}));
|
|
int index = resTexRC.x * ${t[1]} + resTexRC.y;
|
|
return ivec2(0, index);
|
|
}
|
|
`:r?`
|
|
ivec2 getOutputCoords() {
|
|
ivec2 resTexRC = ivec2(resultUV.yx *
|
|
vec2(outTexShape[0], outTexShape[1]));
|
|
int index = resTexRC.x * outTexShape[1] + resTexRC.y;
|
|
int r = index / outShape[1];
|
|
int c = index - r * outShape[1];
|
|
return ivec2(r, c);
|
|
}
|
|
`:`
|
|
ivec2 getOutputCoords() {
|
|
ivec2 resTexRC = ivec2(resultUV.yx *
|
|
vec2(${t[0]}, ${t[1]}));
|
|
int index = resTexRC.x * ${t[1]} + resTexRC.y;
|
|
int r = index / ${e[1]};
|
|
int c = index - r * ${e[1]};
|
|
return ivec2(r, c);
|
|
}
|
|
`}function Ll(e){return`offset${e}`}function dee(e){let t=e.name,r="get"+t.charAt(0).toUpperCase()+t.slice(1),n=Hr();return`
|
|
vec4 ${r}() {
|
|
return ${n.texture2D}(${t}, halfCR);
|
|
}
|
|
`}function pee(e,t){let r=e.name,n="get"+r.charAt(0).toUpperCase()+r.slice(1);if(e.shapeInfo.isUniform)return`float ${n}() {return ${r};}`;let[a,s]=e.shapeInfo.texShape;if(a===1&&s===1)return`
|
|
float ${n}() {
|
|
return sampleTexture(${r}, halfCR);
|
|
}
|
|
`;let i=Ll(r);if(t)return`
|
|
float ${n}() {
|
|
vec2 uv = uvFromFlat(${r}TexShape[0], ${r}TexShape[1], ${i});
|
|
return sampleTexture(${r}, uv);
|
|
}
|
|
`;let[o,l]=e.shapeInfo.texShape;return`
|
|
float ${n}() {
|
|
vec2 uv = uvFromFlat(${o}, ${l}, ${i});
|
|
return sampleTexture(${r}, uv);
|
|
}
|
|
`}function hee(e,t){let r=e.name,n="get"+r.charAt(0).toUpperCase()+r.slice(1),a=e.shapeInfo.texShape,s=Hr();if(t)return`
|
|
vec4 ${n}(int index) {
|
|
ivec2 packedTexShape = ivec2(ceil(float(${r}TexShape[0]) / 2.0), ceil(float(${r}TexShape[1]) / 2.0));
|
|
vec2 uv = packedUVfrom1D(
|
|
packedTexShape[0], packedTexShape[1], index);
|
|
return ${s.texture2D}(${r}, uv);
|
|
}
|
|
`;let i=[Math.ceil(a[0]/2),Math.ceil(a[1]/2)];return`
|
|
vec4 ${n}(int index) {
|
|
vec2 uv = packedUVfrom1D(
|
|
${i[0]}, ${i[1]}, index);
|
|
return ${s.texture2D}(${r}, uv);
|
|
}
|
|
`}function cee(e,t){let r=e.name,n="get"+r.charAt(0).toUpperCase()+r.slice(1);if(e.shapeInfo.isUniform)return`
|
|
float ${n}(int index) {
|
|
${Cd(e)}
|
|
}
|
|
`;let a=e.shapeInfo.texShape,s=a[0],i=a[1];if(i===1&&s===1)return`
|
|
float ${n}(int index) {
|
|
return sampleTexture(${r}, halfCR);
|
|
}
|
|
`;let o=Ll(r);return i===1?t?`
|
|
float ${n}(int index) {
|
|
vec2 uv = vec2(0.5, (float(index + ${o}) + 0.5) / float(${r}TexShape[0]));
|
|
return sampleTexture(${r}, uv);
|
|
}
|
|
`:`
|
|
float ${n}(int index) {
|
|
vec2 uv = vec2(0.5, (float(index + ${o}) + 0.5) / ${s}.0);
|
|
return sampleTexture(${r}, uv);
|
|
}
|
|
`:s===1?t?`
|
|
float ${n}(int index) {
|
|
vec2 uv = vec2((float(index + ${o}) + 0.5) / float(${r}TexShape[1]), 0.5);
|
|
return sampleTexture(${r}, uv);
|
|
}
|
|
`:`
|
|
float ${n}(int index) {
|
|
vec2 uv = vec2((float(index + ${o}) + 0.5) / ${i}.0, 0.5);
|
|
return sampleTexture(${r}, uv);
|
|
}
|
|
`:t?`
|
|
float ${n}(int index) {
|
|
vec2 uv = uvFromFlat(${r}TexShape[0], ${r}TexShape[1], index + ${o});
|
|
return sampleTexture(${r}, uv);
|
|
}
|
|
`:`
|
|
float ${n}(int index) {
|
|
vec2 uv = uvFromFlat(${s}, ${i}, index + ${o});
|
|
return sampleTexture(${r}, uv);
|
|
}
|
|
`}function fee(e,t){let r=e.shapeInfo.logicalShape,n=e.name,a="get"+n.charAt(0).toUpperCase()+n.slice(1),s=e.shapeInfo.texShape,i=s[0],o=s[1],l=Hr();if(s!=null&&v.arraysEqual(r,s))return t?`
|
|
vec4 ${a}(int row, int col) {
|
|
vec2 uv = (vec2(col, row) + halfCR) / vec2(${n}TexShape[1], ${n}TexShape[0]);
|
|
|
|
return ${l.texture2D}(${n}, uv);
|
|
}
|
|
`:`
|
|
vec4 ${a}(int row, int col) {
|
|
vec2 uv = (vec2(col, row) + halfCR) / vec2(${o}.0, ${i}.0);
|
|
|
|
return ${l.texture2D}(${n}, uv);
|
|
}
|
|
`;if(t)return`
|
|
vec4 ${a}(int row, int col) {
|
|
ivec2 packedTexShape = ivec2(ceil(float(${n}TexShape[0]) / 2.0), ceil(float(${n}TexShape[1]) / 2.0));
|
|
int valuesPerRow = int(ceil(float(${n}Shape[1]) / 2.0));
|
|
vec2 uv = packedUVfrom2D(valuesPerRow, packedTexShape[0], packedTexShape[1], row, col);
|
|
return ${l.texture2D}(${n}, uv);
|
|
}
|
|
`;let u=[Math.ceil(s[0]/2),Math.ceil(s[1]/2)],d=Math.ceil(r[1]/2);return`
|
|
vec4 ${a}(int row, int col) {
|
|
vec2 uv = packedUVfrom2D(${d}, ${u[0]}, ${u[1]}, row, col);
|
|
return ${l.texture2D}(${n}, uv);
|
|
}
|
|
`}function mee(e,t){let r=e.shapeInfo.logicalShape,n=e.name,a="get"+n.charAt(0).toUpperCase()+n.slice(1),s=e.shapeInfo.texShape;if(s!=null&&v.arraysEqual(r,s)){if(t)return`
|
|
float ${a}(int row, int col) {
|
|
vec2 uv = (vec2(col, row) + halfCR) / vec2(${n}TexShape[1], ${n}TexShape[0]);
|
|
return sampleTexture(${n}, uv);
|
|
}
|
|
`;let p=s[0],c=s[1];return`
|
|
float ${a}(int row, int col) {
|
|
vec2 uv = (vec2(col, row) + halfCR) / vec2(${c}.0, ${p}.0);
|
|
return sampleTexture(${n}, uv);
|
|
}
|
|
`}let{newShape:i,keptDims:o}=v.squeezeShape(r),l=i;if(l.length<r.length){let p=Td(e,l),c=["row","col"];return`
|
|
${Sd(p,t)}
|
|
float ${a}(int row, int col) {
|
|
return ${a}(${Nd(c,o)});
|
|
}
|
|
`}if(e.shapeInfo.isUniform)return`
|
|
float ${a}(int row, int col) {
|
|
int index = round(dot(vec2(row, col), vec2(${r[1]}, 1)));
|
|
${Cd(e)}
|
|
}
|
|
`;let u=s[0],d=s[1],h=Ll(n);return d===1?t?`
|
|
float ${a}(int row, int col) {
|
|
float index = dot(vec3(row, col, ${h}), vec3(${n}Shape[1], 1, 1));
|
|
vec2 uv = vec2(0.5, (index + 0.5) / float(${n}TexShape[0]));
|
|
return sampleTexture(${n}, uv);
|
|
}
|
|
`:`
|
|
float ${a}(int row, int col) {
|
|
float index = dot(vec3(row, col, ${h}), vec3(${r[1]}, 1, 1));
|
|
vec2 uv = vec2(0.5, (index + 0.5) / ${u}.0);
|
|
return sampleTexture(${n}, uv);
|
|
}
|
|
`:u===1?t?`
|
|
float ${a}(int row, int col) {
|
|
float index = dot(vec3(row, col, ${h}), vec3(${n}Shape[1], 1, 1));
|
|
vec2 uv = vec2((index + 0.5) / float(${n}TexShape[1]), 0.5);
|
|
return sampleTexture(${n}, uv);
|
|
}
|
|
`:`
|
|
float ${a}(int row, int col) {
|
|
float index = dot(vec3(row, col, ${h}), vec3(${r[1]}, 1, 1));
|
|
vec2 uv = vec2((index + 0.5) / ${d}.0, 0.5);
|
|
return sampleTexture(${n}, uv);
|
|
}
|
|
`:t?`
|
|
float ${a}(int row, int col) {
|
|
// Explicitly use integer operations as dot() only works on floats.
|
|
int index = row * ${n}Shape[1] + col + ${h};
|
|
vec2 uv = uvFromFlat(${n}TexShape[0], ${n}TexShape[1], index);
|
|
return sampleTexture(${n}, uv);
|
|
}
|
|
`:`
|
|
float ${a}(int row, int col) {
|
|
// Explicitly use integer operations as dot() only works on floats.
|
|
int index = row * ${r[1]} + col + ${h};
|
|
vec2 uv = uvFromFlat(${u}, ${d}, index);
|
|
return sampleTexture(${n}, uv);
|
|
}
|
|
`}function gee(e,t){let r=e.shapeInfo.logicalShape,n=e.name,a="get"+n.charAt(0).toUpperCase()+n.slice(1),s=e.shapeInfo.texShape,i=[Math.ceil(s[0]/2),Math.ceil(s[1]/2)];if(r[0]===1){let p=r.slice(1),c=[1,2],f=Td(e,p),m=["b","row","col"];return`
|
|
${zI(f,t)}
|
|
vec4 ${a}(int b, int row, int col) {
|
|
return ${a}(${Nd(m,c)});
|
|
}
|
|
`}let o=Hr();if(t)return`
|
|
vec4 ${a}(int b, int row, int col) {
|
|
ivec2 packedTexShape = ivec2(ceil(float(${n}TexShape[0]) / 2.0), ceil(float(${n}TexShape[1]) / 2.0));
|
|
int valuesPerRow = int(ceil(float(${n}Shape[2]) / 2.0));
|
|
int texelsInBatch = valuesPerRow * int(ceil(float(${n}Shape[1]) / 2.0));
|
|
vec2 uv = packedUVfrom3D(
|
|
packedTexShape[0], packedTexShape[1], texelsInBatch, valuesPerRow, b, row, col);
|
|
return ${o.texture2D}(${n}, uv);
|
|
}
|
|
`;let l=i[0],u=i[1],d=Math.ceil(r[2]/2),h=d*Math.ceil(r[1]/2);return`
|
|
vec4 ${a}(int b, int row, int col) {
|
|
vec2 uv = packedUVfrom3D(
|
|
${l}, ${u}, ${h}, ${d}, b, row, col);
|
|
return ${o.texture2D}(${n}, uv);
|
|
}
|
|
`}function yee(e,t){let r=e.shapeInfo.logicalShape,n=e.name,a="get"+n.charAt(0).toUpperCase()+n.slice(1),s=r[1]*r[2],i=r[2],{newShape:o,keptDims:l}=v.squeezeShape(r),u=o;if(u.length<r.length){let m=Td(e,u),g=["row","col","depth"];return`
|
|
${Sd(m,t)}
|
|
float ${a}(int row, int col, int depth) {
|
|
return ${a}(${Nd(g,l)});
|
|
}
|
|
`}if(e.shapeInfo.isUniform)return`
|
|
float ${a}(int row, int col, int depth) {
|
|
int index = round(dot(vec3(row, col, depth),
|
|
vec3(${s}, ${i}, 1)));
|
|
${Cd(e)}
|
|
}
|
|
`;let d=e.shapeInfo.texShape,h=d[0],p=d[1],c=e.shapeInfo.flatOffset;if(p===s&&c==null)return t?`
|
|
float ${a}(int row, int col, int depth) {
|
|
int stride1 = ${n}Shape[2];
|
|
float texR = float(row);
|
|
float texC = dot(vec2(col, depth), vec2(stride1, 1));
|
|
vec2 uv = (vec2(texC, texR) + halfCR) /
|
|
vec2(${n}TexShape[1], ${n}TexShape[0]);
|
|
return sampleTexture(${n}, uv);
|
|
}
|
|
`:`
|
|
float ${a}(int row, int col, int depth) {
|
|
float texR = float(row);
|
|
float texC = dot(vec2(col, depth), vec2(${i}, 1));
|
|
vec2 uv = (vec2(texC, texR) + halfCR) /
|
|
vec2(${p}.0, ${h}.0);
|
|
return sampleTexture(${n}, uv);
|
|
}
|
|
`;if(p===i&&c==null)return t?`
|
|
float ${a}(int row, int col, int depth) {
|
|
float texR = dot(vec2(row, col), vec2(${n}Shape[1], 1));
|
|
float texC = float(depth);
|
|
vec2 uv = (vec2(texC, texR) + halfCR) / vec2(${n}TexShape[1], ${n}TexShape[0]);
|
|
return sampleTexture(${n}, uv);
|
|
}
|
|
`:`
|
|
float ${a}(int row, int col, int depth) {
|
|
float texR = dot(vec2(row, col), vec2(${r[1]}, 1));
|
|
float texC = float(depth);
|
|
vec2 uv = (vec2(texC, texR) + halfCR) / vec2(${p}.0, ${h}.0);
|
|
return sampleTexture(${n}, uv);
|
|
}
|
|
`;let f=Ll(n);return t?`
|
|
float ${a}(int row, int col, int depth) {
|
|
// Explicitly use integer operations as dot() only works on floats.
|
|
int stride0 = ${n}Shape[1] * ${n}Shape[2];
|
|
int stride1 = ${n}Shape[2];
|
|
int index = row * ${s} + col * ${i} + depth + ${f};
|
|
vec2 uv = uvFromFlat(${n}TexShape[0], ${n}TexShape[1], index);
|
|
return sampleTexture(${n}, uv);
|
|
}
|
|
`:`
|
|
float ${a}(int row, int col, int depth) {
|
|
// Explicitly use integer operations as dot() only works on floats.
|
|
int index = row * ${s} + col * ${i} + depth + ${f};
|
|
vec2 uv = uvFromFlat(${h}, ${p}, index);
|
|
return sampleTexture(${n}, uv);
|
|
}
|
|
`}function Aee(e,t){let r=e.name,n="get"+r.charAt(0).toUpperCase()+r.slice(1),a=Hr();if(t)return`
|
|
vec4 ${n}(int b2, int b, int row, int col) {
|
|
int valuesPerRow = int(ceil(float(${r}Shape[3]) / 2.0));
|
|
int texelsInBatch = valuesPerRow * int(ceil(float(${r}Shape[2]) / 2.0));
|
|
int index = b * texelsInBatch + (row / 2) * valuesPerRow + (col / 2);
|
|
texelsInBatch *= ${r}Shape[1];
|
|
index = b2 * texelsInBatch + index;
|
|
ivec2 packedTexShape = ivec2(ceil(float(${r}TexShape[0]) / 2.0), ceil(float(${r}TexShape[1]) / 2.0));
|
|
int texR = index / packedTexShape[1];
|
|
int texC = index - texR * packedTexShape[1];
|
|
vec2 uv = (vec2(texC, texR) + halfCR) / vec2(packedTexShape[1], packedTexShape[0]); return ${a.texture2D}(${r}, uv);
|
|
}
|
|
`;let s=e.shapeInfo.logicalShape,i=s.length,o=e.shapeInfo.texShape,l=[Math.ceil(o[0]/2),Math.ceil(o[1]/2)],u=l[0],d=l[1],h=Math.ceil(s[i-1]/2),p=h*Math.ceil(s[i-2]/2),c="int b, int row, int col",f=`b * ${p} + (row / 2) * ${h} + (col / 2)`;for(let m=2;m<i-1;m++)c=`int b${m}, `+c,p*=s[i-m-1],f=`b${m} * ${p} + `+f;return`
|
|
vec4 ${n}(${c}) {
|
|
int index = ${f};
|
|
int texR = index / ${d};
|
|
int texC = index - texR * ${d};
|
|
vec2 uv = (vec2(texC, texR) + halfCR) / vec2(${d}, ${u});
|
|
return ${a.texture2D}(${r}, uv);
|
|
}
|
|
`}function xee(e,t){let r=e.shapeInfo.logicalShape,n=e.name,a="get"+n.charAt(0).toUpperCase()+n.slice(1),s=r[3],i=r[2]*s,o=r[1]*i,{newShape:l,keptDims:u}=v.squeezeShape(r);if(l.length<r.length){let A=Td(e,l),x=["row","col","depth","depth2"];return`
|
|
${Sd(A,t)}
|
|
float ${a}(int row, int col, int depth, int depth2) {
|
|
return ${a}(${Nd(x,u)});
|
|
}
|
|
`}if(e.shapeInfo.isUniform)return`
|
|
float ${a}(int row, int col, int depth, int depth2) {
|
|
int index = round(dot(vec4(row, col, depth, depth2),
|
|
vec4(${o}, ${i}, ${s}, 1)));
|
|
${Cd(e)}
|
|
}
|
|
`;let d=e.shapeInfo.flatOffset,h=e.shapeInfo.texShape,p=h[0],c=h[1],f=`int stride2 = ${n}Shape[3];`,m=`int stride1 = ${n}Shape[2] * stride2;`,g=`int stride0 = ${n}Shape[1] * stride1;`;if(c===o&&d==null)return t?`
|
|
float ${a}(int row, int col, int depth, int depth2) {
|
|
${f}
|
|
${m}
|
|
float texR = float(row);
|
|
float texC =
|
|
dot(vec3(col, depth, depth2),
|
|
vec3(stride1, stride2, 1));
|
|
vec2 uv = (vec2(texC, texR) + halfCR) /
|
|
vec2(${n}TexShape[1], ${n}TexShape[0]);
|
|
return sampleTexture(${n}, uv);
|
|
}
|
|
`:`
|
|
float ${a}(int row, int col, int depth, int depth2) {
|
|
float texR = float(row);
|
|
float texC =
|
|
dot(vec3(col, depth, depth2),
|
|
vec3(${i}, ${s}, 1));
|
|
vec2 uv = (vec2(texC, texR) + halfCR) /
|
|
vec2(${c}.0, ${p}.0);
|
|
return sampleTexture(${n}, uv);
|
|
}
|
|
`;if(c===s&&d==null)return t?`
|
|
float ${a}(int row, int col, int depth, int depth2) {
|
|
float texR = dot(vec3(row, col, depth),
|
|
vec3(${n}Shape[1] * ${n}Shape[2], ${n}Shape[2], 1));
|
|
float texC = float(depth2);
|
|
vec2 uv = (vec2(texC, texR) + halfCR) /
|
|
vec2(${n}TexShape[1], ${n}TexShape[0]);
|
|
return sampleTexture(${n}, uv);
|
|
}
|
|
`:`
|
|
float ${a}(int row, int col, int depth, int depth2) {
|
|
float texR = dot(vec3(row, col, depth),
|
|
vec3(${r[1]*r[2]}, ${r[2]}, 1));
|
|
float texC = float(depth2);
|
|
vec2 uv = (vec2(texC, texR) + halfCR) /
|
|
vec2(${c}.0, ${p}.0);
|
|
return sampleTexture(${n}, uv);
|
|
}
|
|
`;let y=Ll(n);return t?`
|
|
float ${a}(int row, int col, int depth, int depth2) {
|
|
// Explicitly use integer operations as dot() only works on floats.
|
|
${f}
|
|
${m}
|
|
${g}
|
|
int index = row * stride0 + col * stride1 +
|
|
depth * stride2 + depth2;
|
|
vec2 uv = uvFromFlat(${n}TexShape[0], ${n}TexShape[1], index + ${y});
|
|
return sampleTexture(${n}, uv);
|
|
}
|
|
`:`
|
|
float ${a}(int row, int col, int depth, int depth2) {
|
|
// Explicitly use integer operations as dot() only works on floats.
|
|
int index = row * ${o} + col * ${i} +
|
|
depth * ${s} + depth2;
|
|
vec2 uv = uvFromFlat(${p}, ${c}, index + ${y});
|
|
return sampleTexture(${n}, uv);
|
|
}
|
|
`}function bee(e){let t=e.shapeInfo.logicalShape,r=e.name,n="get"+r.charAt(0).toUpperCase()+r.slice(1),a=t[4],s=t[3]*a,i=t[2]*s,o=t[1]*i,{newShape:l,keptDims:u}=v.squeezeShape(t);if(l.length<t.length){let m=Td(e,l),g=["row","col","depth","depth2","depth3"];return`
|
|
${Sd(m)}
|
|
float ${n}(int row, int col, int depth, int depth2, int depth3) {
|
|
return ${n}(${Nd(g,u)});
|
|
}
|
|
`}if(e.shapeInfo.isUniform)return`
|
|
float ${n}(int row, int col, int depth, int depth2, int depth3) {
|
|
float index = dot(
|
|
vec4(row, col, depth, depth2),
|
|
vec4(${o}, ${i}, ${s}, ${a})) +
|
|
depth3;
|
|
${Cd(e)}
|
|
}
|
|
`;let d=e.shapeInfo.flatOffset,h=e.shapeInfo.texShape,p=h[0],c=h[1];if(c===o&&d==null)return`
|
|
float ${n}(int row, int col, int depth, int depth2, int depth3) {
|
|
int texR = row;
|
|
float texC = dot(vec4(col, depth, depth2, depth3),
|
|
vec4(${i}, ${s}, ${a}, 1));
|
|
vec2 uv = (vec2(texC, texR) + halfCR) /
|
|
vec2(${c}.0, ${p}.0);
|
|
return sampleTexture(${r}, uv);
|
|
}
|
|
`;if(c===a&&d==null)return`
|
|
float ${n}(int row, int col, int depth, int depth2, int depth3) {
|
|
float texR = dot(
|
|
vec4(row, col, depth, depth2),
|
|
vec4(${t[1]*t[2]*t[3]},
|
|
${t[2]*t[3]}, ${t[3]}, 1));
|
|
int texC = depth3;
|
|
vec2 uv = (vec2(texC, texR) + halfCR) /
|
|
vec2(${c}.0, ${p}.0);
|
|
return sampleTexture(${r}, uv);
|
|
}
|
|
`;let f=Ll(r);return`
|
|
float ${n}(int row, int col, int depth, int depth2, int depth3) {
|
|
// Explicitly use integer operations as dot() only works on floats.
|
|
int index = row * ${o} + col * ${i} + depth * ${s} +
|
|
depth2 * ${a} + depth3 + ${f};
|
|
vec2 uv = uvFromFlat(${p}, ${c}, index);
|
|
return sampleTexture(${r}, uv);
|
|
}
|
|
`}function vee(e){let t=e.shapeInfo.logicalShape,r=e.name,n="get"+r.charAt(0).toUpperCase()+r.slice(1),{newShape:a,keptDims:s}=v.squeezeShape(t);if(a.length<t.length){let g=Td(e,a),y=["row","col","depth","depth2","depth3","depth4"];return`
|
|
${Sd(g)}
|
|
float ${n}(int row, int col, int depth,
|
|
int depth2, int depth3, int depth4) {
|
|
return ${n}(${Nd(y,s)});
|
|
}
|
|
`}let i=t[5],o=t[4]*i,l=t[3]*o,u=t[2]*l,d=t[1]*u;if(e.shapeInfo.isUniform)return`
|
|
float ${n}(int row, int col, int depth,
|
|
int depth2, int depth3, int depth4) {
|
|
int index = round(dot(
|
|
vec4(row, col, depth, depth2),
|
|
vec4(${d}, ${u}, ${l}, ${o})) +
|
|
dot(
|
|
vec2(depth3, depth4),
|
|
vec2(${i}, 1)));
|
|
${Cd(e)}
|
|
}
|
|
`;let h=e.shapeInfo.flatOffset,p=e.shapeInfo.texShape,c=p[0],f=p[1];if(f===d&&h==null)return`
|
|
float ${n}(int row, int col, int depth,
|
|
int depth2, int depth3, int depth4) {
|
|
int texR = row;
|
|
float texC = dot(vec4(col, depth, depth2, depth3),
|
|
vec4(${u}, ${l}, ${o}, ${i})) +
|
|
float(depth4);
|
|
vec2 uv = (vec2(texC, texR) + halfCR) /
|
|
vec2(${f}.0, ${c}.0);
|
|
return sampleTexture(${r}, uv);
|
|
}
|
|
`;if(f===i&&h==null)return`
|
|
float ${n}(int row, int col, int depth,
|
|
int depth2, int depth3, int depth4) {
|
|
float texR = dot(vec4(row, col, depth, depth2),
|
|
vec4(${t[1]*t[2]*t[3]*t[4]},
|
|
${t[2]*t[3]*t[4]},
|
|
${t[3]*t[4]},
|
|
${t[4]})) + float(depth3);
|
|
int texC = depth4;
|
|
vec2 uv = (vec2(texC, texR) + halfCR) /
|
|
vec2(${f}.0, ${c}.0);
|
|
return sampleTexture(${r}, uv);
|
|
}
|
|
`;let m=Ll(r);return`
|
|
float ${n}(int row, int col, int depth,
|
|
int depth2, int depth3, int depth4) {
|
|
// Explicitly use integer operations as dot() only works on floats.
|
|
int index = row * ${d} + col * ${u} + depth * ${l} +
|
|
depth2 * ${o} + depth3 * ${i} + depth4 + ${m};
|
|
vec2 uv = uvFromFlat(${c}, ${f}, index);
|
|
return sampleTexture(${r}, uv);
|
|
}
|
|
`}function Cd(e){let t=e.name,r=v.sizeFromShape(e.shapeInfo.logicalShape);return r<2?`return ${t};`:`
|
|
for (int i = 0; i < ${r}; i++) {
|
|
if (i == index) {
|
|
return ${t}[i];
|
|
}
|
|
}
|
|
`}function wee(e,t){let r=e.name,n=r.charAt(0).toUpperCase()+r.slice(1),a="get"+n+"AtOutCoords",s=e.shapeInfo.logicalShape.length,i=t.logicalShape.length,o=_I(e.shapeInfo.logicalShape,t.logicalShape),l=vt(i),u=i-s,d,h=["x","y","z","w","u","v"];s===0?d="":i<2&&o.length>=1?d="coords = 0;":d=o.map(g=>`coords.${h[g+u]} = 0;`).join(`
|
|
`);let p="";i<2&&s>0?p="coords":p=e.shapeInfo.logicalShape.map((g,y)=>`coords.${h[y+u]}`).join(", ");let c="return outputValue;",f=v.sizeFromShape(e.shapeInfo.logicalShape)===1,m=v.sizeFromShape(t.logicalShape)===1;if(s===1&&!f&&!m)c=`
|
|
return vec4(outputValue.xy, outputValue.xy);
|
|
`;else if(f&&!m)i===1?c=`
|
|
return vec4(outputValue.x, outputValue.x, 0., 0.);
|
|
`:c=`
|
|
return vec4(outputValue.x);
|
|
`;else if(o.length){let g=s-2,y=s-1;o.indexOf(g)>-1&&o.indexOf(y)>-1?c="return vec4(outputValue.x);":o.indexOf(g)>-1?c="return vec4(outputValue.x, outputValue.y, outputValue.x, outputValue.y);":o.indexOf(y)>-1&&(c="return vec4(outputValue.xx, outputValue.zz);")}return`
|
|
vec4 ${a}() {
|
|
${l} coords = getOutputCoords();
|
|
${d}
|
|
vec4 outputValue = get${n}(${p});
|
|
${c}
|
|
}
|
|
`}function kee(e,t){let r=e.name,n=r.charAt(0).toUpperCase()+r.slice(1),a="get"+n+"AtOutCoords",s=t.texShape,i=e.shapeInfo.texShape,o=e.shapeInfo.logicalShape.length,l=t.logicalShape.length;if(!e.shapeInfo.isUniform&&o===l&&e.shapeInfo.flatOffset==null&&v.arraysEqual(i,s))return`
|
|
float ${a}() {
|
|
return sampleTexture(${r}, resultUV);
|
|
}
|
|
`;let u=vt(l),d=_I(e.shapeInfo.logicalShape,t.logicalShape),h=l-o,p,c=["x","y","z","w","u","v"];o===0?p="":l<2&&d.length>=1?p="coords = 0;":p=d.map(m=>`coords.${c[m+h]} = 0;`).join(`
|
|
`);let f="";return l<2&&o>0?f="coords":f=e.shapeInfo.logicalShape.map((m,g)=>`coords.${c[g+h]}`).join(", "),`
|
|
float ${a}() {
|
|
${u} coords = getOutputCoords();
|
|
${p}
|
|
return get${n}(${f});
|
|
}
|
|
`}function vt(e){if(e<=1)return"int";if(e===2)return"ivec2";if(e===3)return"ivec3";if(e===4)return"ivec4";if(e===5)return"ivec5";if(e===6)return"ivec6";throw Error(`GPU for rank ${e} is not yet supported`)}function g5(e,t,r){let{newShape:n,keptDims:a}=v.squeezeShape(t),s=t.length,i=e&&s===3&&t[0]===1,o=i?t.slice(1):n,l=!e&&s>1&&!v.arraysEqual(t,r)&&n.length<s||i;return{useSqueezeShape:l,uniformShape:l?o:t,keptDims:a}}function Td(e,t){let r=JSON.parse(JSON.stringify(e));return r.shapeInfo.logicalShape=t,r}function Nd(e,t){return t.map(r=>e[r]).join(", ")}function Iee(e,t,r,n){let a=r.map((d,h)=>{let p={logicalShape:d.shape,texShape:d.isUniform?null:d.texData.texShape,isUniform:d.isUniform,isPacked:d.isUniform?!1:d.texData.isPacked,flatOffset:null};return d.texData!=null&&d.texData.slice!=null&&d.texData.slice.flatOffset>0&&(p.flatOffset=d.texData.slice.flatOffset),{name:t.variableNames[h],shapeInfo:p}}),s=a.map(d=>d.shapeInfo),i={logicalShape:n.shape,texShape:n.texData.texShape,isUniform:!1,isPacked:n.texData.isPacked,flatOffset:null},o=VQ(a,i,t),l=cI(e.gl,o),u=e.createProgram(l);return Z().get("ENGINE_COMPILE_ONLY")?{program:t,fragmentShader:l,source:o,webGLProgram:u,inShapeInfos:s,outShapeInfo:i,uniformLocations:null,customUniformLocations:null,infLoc:null,nanLoc:null,inShapesLocations:null,inTexShapesLocations:null,outShapeLocation:null,outShapeStridesLocation:null,outTexShapeLocation:null}:{program:t,fragmentShader:l,source:o,webGLProgram:u,inShapeInfos:s,outShapeInfo:i,...DI(e,t,u)}}function DI(e,t,r){let n={},a={},s={},i=[],o,l,u,d=null,h=null;h=e.getUniformLocation(r,"NAN",!1),Z().getNumber("WEBGL_VERSION")===1&&(d=e.getUniformLocation(r,"INFINITY",!1));let p=!1;for(let c=0;c<t.variableNames.length;c++){let f=t.variableNames[c];n[f]=e.getUniformLocation(r,f,p),n[`offset${f}`]=e.getUniformLocation(r,`offset${f}`,p),t.enableShapeUniforms&&(a[`${f}Shape`]=e.getUniformLocation(r,`${f}Shape`,p),s[`${f}TexShape`]=e.getUniformLocation(r,`${f}TexShape`,p))}return t.enableShapeUniforms&&(o=e.getUniformLocation(r,"outShape",p),u=e.getUniformLocation(r,"outShapeStrides",p),l=e.getUniformLocation(r,"outTexShape",p)),t.customUniforms&&t.customUniforms.forEach((c,f)=>{i[f]=e.getUniformLocation(r,c.name,p)}),{uniformLocations:n,customUniformLocations:i,infLoc:d,nanLoc:h,inShapesLocations:a,inTexShapesLocations:s,outShapeLocation:o,outShapeStridesLocation:u,outTexShapeLocation:l}}function O4(e,t){if(e.length!==t.length)throw Error(`Binary was compiled with ${e.length} inputs, but was executed with ${t.length} inputs`);e.forEach((r,n)=>{let a=r.logicalShape,s=t[n],i=s.shape;if(!v.arraysEqual(a,i))throw Error(`Binary was compiled with different shapes than the current args. Shapes ${a} and ${i} must match`);if(r.isUniform&&s.isUniform)return;let o=r.texShape,l=s.isUniform?null:s.texData.texShape;if(!v.arraysEqual(o,l))throw Error(`Binary was compiled with different texture shapes than the current args. Shape ${o} and ${l} must match`)})}function See(e,t,r,n,a){t.program.enableShapeUniforms||(O4(t.inShapeInfos,r),O4([t.outShapeInfo],[n]));let s=n.texData.texture,i=n.texData.texShape;n.texData.isPacked?e.setOutputPackedMatrixTexture(s.texture,i[0],i[1]):e.setOutputMatrixTexture(s.texture,i[0],i[1]),e.setProgram(t.webGLProgram),Z().getNumber("WEBGL_VERSION")===1&&t.infLoc!==null&&e.gl.uniform1f(t.infLoc,1/0),t.nanLoc!==null&&e.gl.uniform1f(t.nanLoc,NaN),r.forEach((l,u)=>{let d=t.program.variableNames[u],h=t.uniformLocations[d],p=t.uniformLocations[`offset${d}`],c=t.inShapesLocations[`${d}Shape`],f=t.inTexShapesLocations[`${d}TexShape`];if(c){let{uniformShape:m}=g5(t.program.packedInputs,l.shape,l.texData.texShape);switch(m.length){case 1:e.gl.uniform1iv(c,new Int32Array(m));break;case 2:e.gl.uniform2iv(c,new Int32Array(m));break;case 3:e.gl.uniform3iv(c,new Int32Array(m));break;case 4:e.gl.uniform4iv(c,new Int32Array(m));break;default:break}}if(f&&e.gl.uniform2i(f,l.texData.texShape[0],l.texData.texShape[1]),h!=null){if(l.isUniform){if(v.sizeFromShape(l.shape)<2)e.gl.uniform1f(h,l.uniformValues[0]);else{let m=l.uniformValues;m instanceof Float32Array||(m=new Float32Array(m)),e.gl.uniform1fv(h,m)}return}l.texData.slice!=null&&p!=null&&e.gl.uniform1i(p,l.texData.slice.flatOffset),e.setInputMatrixTexture(l.texData.texture.texture,h,u)}});let o=t.outShapeLocation;if(o)switch(n.shape.length){case 1:e.gl.uniform1iv(o,new Int32Array(n.shape));break;case 2:e.gl.uniform2iv(o,new Int32Array(n.shape));break;case 3:e.gl.uniform3iv(o,new Int32Array(n.shape));break;case 4:e.gl.uniform4iv(o,new Int32Array(n.shape));break;default:break}if(t.outShapeStridesLocation){let l=v.computeStrides(n.shape);switch(n.shape.length){case 2:e.gl.uniform1iv(t.outShapeStridesLocation,new Int32Array(l));break;case 3:e.gl.uniform2iv(t.outShapeStridesLocation,new Int32Array(l));break;case 4:e.gl.uniform3iv(t.outShapeStridesLocation,new Int32Array(l));break;default:break}}t.outTexShapeLocation&&e.gl.uniform2i(t.outTexShapeLocation,n.texData.texShape[0],n.texData.texShape[1]),t.program.customUniforms&&a&&t.program.customUniforms.forEach((l,u)=>{let d=t.customUniformLocations[u],h=a[u];if(l.type==="float")e.gl.uniform1fv(d,h);else if(l.type==="vec2")e.gl.uniform2fv(d,h);else if(l.type==="vec3")e.gl.uniform3fv(d,h);else if(l.type==="vec4")e.gl.uniform4fv(d,h);else if(l.type==="int")e.gl.uniform1iv(d,h);else if(l.type==="ivec2")e.gl.uniform2iv(d,h);else if(l.type==="ivec3")e.gl.uniform3iv(d,h);else if(l.type==="ivec4")e.gl.uniform4iv(d,h);else throw Error(`uniform type ${l.type} is not supported yet.`)}),e.executeProgram()}function Cee(e,t,r){let n="";t.concat(r).forEach(i=>{let o=i.texData!=null&&i.texData.slice!=null&&i.texData.slice.flatOffset>0;if(e.enableShapeUniforms&&!i.isUniform){let l=i.texData.texShape,{useSqueezeShape:u,uniformShape:d,keptDims:h}=g5(e.packedInputs,i.shape,l),p="",c="",f="";if(d.length===1&&e.packedInputs){let w=[Math.ceil(l[0]/2),Math.ceil(l[1]/2)];p=`${w[0]>1}_${w[1]>1}`}else if(d.length===2&&!e.packedInputs)c=`${d[0]>1}_${d[1]>1}`;else if(d.length>2&&!e.packedInputs){let w=v.computeStrides(d);f=`${w[0]===l[1]}_${w[w.length-1]===l[1]}`}let m=i.shape.length,g=d.length===2&&v.arraysEqual(i.shape,l),y=v.sizeFromShape(i.shape)===1,A=C.getBroadcastDims(i.shape,r.shape),x=!e.packedInputs&&m===r.shape.length&&v.arraysEqual(l,r.texData.texShape),b=e.packedInputs||d.length>2?"":`${l[0]>1}_${l[1]>1}`;n+=`${m}_${x}_${u?h:""}_${d.length}_${y}_${A}_${g}_${p}_${c}_${f}_${b}_${o}`}else{let l=i.isUniform?"uniform":i.texData.texShape;n+=`${i.shape}_${l}_${o}`}});let a=e.userCode,s=e.constructor.name;return s+="_"+n+"_"+a+`${Z().getNumber("WEBGL_VERSION")}`,s}function ln(e){return Z().getBool("WEBGL_USE_SHAPES_UNIFORMS")&&e<=4}var Tee=class{constructor(e){this.variableNames=["A"],this.packedInputs=!1,this.packedOutput=!0,this.outPackingScheme=0,this.customUniforms=[{name:"texShape",type:"ivec2"}];let t=Hr();this.outputShape=e,this.enableShapeUniforms=ln(this.outputShape.length),this.userCode=`
|
|
ivec3 outCoordsFromFlatIndex(int index) {
|
|
${this.enableShapeUniforms?Sm(["r","c","d"],e):Dl(["r","c","d"],e)}
|
|
return ivec3(r, c, d);
|
|
}
|
|
|
|
void main() {
|
|
ivec2 resTexRC = ivec2(resultUV.yx * vec2(texShape[0], texShape[1]));
|
|
int index = 4 * (resTexRC.x * texShape[1] + resTexRC.y);
|
|
|
|
vec4 result = vec4(0.);
|
|
|
|
for (int i=0; i<4; i++) {
|
|
int flatIndex = index + i;
|
|
ivec3 rc = outCoordsFromFlatIndex(flatIndex);
|
|
result[i] = getA(rc.x, rc.y, rc.z);
|
|
}
|
|
|
|
${t.output} = result;
|
|
}
|
|
`}},Nee=class{constructor(e){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0,this.outPackingScheme=0,this.customUniforms=[{name:"texShape",type:"ivec2"}];let t=Hr();this.outputShape=e,this.enableShapeUniforms=ln(this.outputShape.length),this.userCode=`
|
|
ivec3 outCoordsFromFlatIndex(int index) {
|
|
${this.enableShapeUniforms?Sm(["r","c","d"],e):Dl(["r","c","d"],e)}
|
|
return ivec3(r, c, d);
|
|
}
|
|
|
|
void main() {
|
|
ivec2 resTexRC = ivec2(resultUV.yx * vec2(texShape[0], texShape[1]));
|
|
int index = 4 * (resTexRC.x * texShape[1] + resTexRC.y);
|
|
|
|
vec4 result = vec4(0.);
|
|
|
|
for (int i=0; i<4; i++) {
|
|
int flatIndex = index + i;
|
|
ivec3 rc = outCoordsFromFlatIndex(flatIndex);
|
|
result[i] = getChannel(getA(rc.x, rc.y, rc.z), vec2(rc.y, rc.z));
|
|
}
|
|
|
|
${t.output} = result;
|
|
}
|
|
`}},Eee=class{constructor(e){this.variableNames=["A"],this.outTexUsage=3;let t=Hr();this.outputShape=e,this.userCode=`
|
|
${PI}
|
|
|
|
void main() {
|
|
float x = getAAtOutCoords();
|
|
${t.output} = encode_float(x);
|
|
}
|
|
`}},Ree=class{constructor(e){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!1,this.outTexUsage=3;let t=Hr();this.outputShape=e,this.userCode=`
|
|
${PI}
|
|
|
|
void main() {
|
|
ivec3 coords = getOutputCoords();
|
|
float x = getChannel(getAAtOutCoords(), vec2(coords.y, coords.z));
|
|
${t.output} = encode_float(x);
|
|
}
|
|
`}},$ee=class{constructor(e,t=!1){this.variableNames=["A"],this.customUniforms=[{name:"texShape",type:"ivec2"}];let r=Hr();this.outputShape=e,this.enableShapeUniforms=ln(this.outputShape.length);let n="result";t&&(n="floor(result * 255. + 0.5)"),this.userCode=`
|
|
${this.enableShapeUniforms?m5():f5(e)}
|
|
|
|
void main() {
|
|
ivec3 coords = getOutputCoords();
|
|
|
|
int flatIndex = getFlatIndex(coords);
|
|
int offset = imod(flatIndex, 4);
|
|
|
|
flatIndex = idiv(flatIndex, 4, 1.);
|
|
|
|
int r = flatIndex / texShape[1];
|
|
int c = imod(flatIndex, texShape[1]);
|
|
vec2 uv = (vec2(c, r) + halfCR) / vec2(texShape[1], texShape[0]);
|
|
vec4 values = ${r.texture2D}(A, uv);
|
|
|
|
float result;
|
|
|
|
if(offset == 0) {
|
|
result = values[0];
|
|
} else if(offset == 1) {
|
|
result = values[1];
|
|
} else if(offset == 2) {
|
|
result = values[2];
|
|
} else {
|
|
result = values[3];
|
|
}
|
|
|
|
${r.output} = vec4(${n}, 0., 0., 0.);
|
|
}
|
|
`}},Mee=class{constructor(e,t=!1){this.variableNames=["A"],this.packedInputs=!1,this.packedOutput=!0,this.customUniforms=[{name:"texShape",type:"ivec2"}];let r=Hr();this.outputShape=e,this.enableShapeUniforms=ln(this.outputShape.length);let n="",a="result";t&&(a="floor(result * 255. + 0.5)");for(let s=0;s<=1;s++)for(let i=0;i<=1;i++){let o=s*2+i;n+=`
|
|
localCoords = coords;
|
|
if(localCoords[2] + ${i} < ${this.enableShapeUniforms?"outShape[2]":`${e[2]}`}) {
|
|
localCoords[2] += ${i};
|
|
if (localCoords[1] + ${s} < ${this.enableShapeUniforms?"outShape[1]":`${e[1]}`}) {
|
|
localCoords[1] += ${s};
|
|
|
|
flatIndex = getFlatIndex(localCoords);
|
|
offset = imod(flatIndex, 4);
|
|
|
|
flatIndex = idiv(flatIndex, 4, 1.);
|
|
|
|
int r = flatIndex / texShape[1];
|
|
int c = imod(flatIndex, texShape[1]);
|
|
vec2 uv = (vec2(c, r) + halfCR) / vec2(texShape[1], texShape[0]);
|
|
values = ${r.texture2D}(A, uv);
|
|
|
|
if (offset == 0) {
|
|
result[${o}] = values[0];
|
|
} else if (offset == 1) {
|
|
result[${o}] = values[1];
|
|
} else if (offset == 2) {
|
|
result[${o}] = values[2];
|
|
} else {
|
|
result[${o}] = values[3];
|
|
}
|
|
}
|
|
}
|
|
`}this.userCode=`
|
|
${this.enableShapeUniforms?m5():f5(e)}
|
|
|
|
void main() {
|
|
ivec3 coords = getOutputCoords();
|
|
|
|
vec4 result = vec4(0.);
|
|
int flatIndex, r, c, offset;
|
|
ivec3 localCoords;
|
|
vec2 uv;
|
|
vec4 values;
|
|
|
|
${n}
|
|
|
|
${r.output} = ${a};
|
|
}
|
|
`}},LI={};Le(LI,{bindVertexProgramAttributeStreams:()=>XI,createBufferFromOutputTexture:()=>YI,createFloat16MatrixTexture:()=>GI,createFloat16PackedMatrixTexture:()=>qI,createFloat32MatrixTexture:()=>UI,createIndexBuffer:()=>VI,createPackedMatrixTexture:()=>HI,createUnsignedBytesMatrixTexture:()=>jI,createVertexBuffer:()=>WI,createVertexShader:()=>BI,downloadByteEncodedFloatMatrixFromOutputTexture:()=>QI,downloadFloat32MatrixFromBuffer:()=>JI,downloadMatrixFromPackedOutputTexture:()=>t9,downloadPackedMatrixFromBuffer:()=>e9,getInternalFormatForFloat16MatrixTexture:()=>A5,getInternalFormatForFloat16PackedMatrixTexture:()=>v5,getInternalFormatForFloat32MatrixTexture:()=>y5,getInternalFormatForPackedMatrixTexture:()=>b5,getInternalFormatForUnsignedBytesMatrixTexture:()=>x5,uploadDenseMatrixToTexture:()=>KI,uploadPixelDataToTexture:()=>ZI});function BI(e){let t=Hr(),r=`${t.version}
|
|
precision highp float;
|
|
${t.attribute} vec3 clipSpacePos;
|
|
${t.attribute} vec2 uv;
|
|
${t.varyingVs} vec2 resultUV;
|
|
|
|
void main() {
|
|
gl_Position = vec4(clipSpacePos, 1);
|
|
resultUV = uv;
|
|
}`;return hI(e,r)}function WI(e){let t=new Float32Array([-1,1,0,0,1,-1,-1,0,0,0,1,1,0,1,1,1,-1,0,1,0]);return gI(e,t)}function VI(e){let t=new Uint16Array([0,1,2,2,1,3]);return yI(e,t)}function Lh(e,t,r,n,a,s){xI(t,r);let i=AI(e),o=e.TEXTURE_2D;return we(e,()=>e.bindTexture(o,i)),we(e,()=>e.texParameteri(o,e.TEXTURE_WRAP_S,e.CLAMP_TO_EDGE)),we(e,()=>e.texParameteri(o,e.TEXTURE_WRAP_T,e.CLAMP_TO_EDGE)),we(e,()=>e.texParameteri(o,e.TEXTURE_MIN_FILTER,e.NEAREST)),we(e,()=>e.texParameteri(o,e.TEXTURE_MAG_FILTER,e.NEAREST)),Z().getNumber("WEBGL_VERSION")===1?we(e,()=>e.texImage2D(o,0,n,t,r,0,a,s,null)):we(e,()=>e.texStorage2D(o,1,n,t,r)),we(e,()=>e.bindTexture(e.TEXTURE_2D,null)),{texture:i,texShape:[r,t]}}function y5(e){return e.internalFormatFloat}function UI(e,t,r,n){let[a,s]=Dh(t,r);return Lh(e,a,s,y5(n),n.textureFormatFloat,e.FLOAT)}function A5(e){return e.internalFormatHalfFloat}function GI(e,t,r,n){let[a,s]=Dh(t,r);return Lh(e,a,s,A5(n),n.textureFormatFloat,n.textureTypeHalfFloat)}function x5(e){return e.downloadTextureFormat}function jI(e,t,r,n){let[a,s]=Dh(t,r);return Lh(e,a,s,x5(n),e.RGBA,e.UNSIGNED_BYTE)}function b5(e){return e.internalFormatPackedFloat}function HI(e,t,r,n){let[a,s]=kd(t,r);return Lh(e,a,s,b5(n),e.RGBA,e.FLOAT)}function v5(e){return e.internalFormatPackedHalfFloat}function qI(e,t,r,n){let[a,s]=kd(t,r);return Lh(e,a,s,v5(n),e.RGBA,n.textureTypeHalfFloat)}function XI(e,t,r){return we(e,()=>e.bindBuffer(e.ARRAY_BUFFER,r)),ey(e,t,"clipSpacePos",r,3,20,0)&&ey(e,t,"uv",r,2,20,12)}function KI(e,t,r,n,a,s){we(e,()=>e.bindTexture(e.TEXTURE_2D,t));let i,o,l;a instanceof Uint8Array?(i=new Uint8Array(r*n*4),o=e.UNSIGNED_BYTE,l=e.RGBA):(i=new Float32Array(r*n*4),o=e.FLOAT,l=s.internalFormatPackedFloat),i.set(a),Z().getNumber("WEBGL_VERSION")===2?we(e,()=>e.texSubImage2D(e.TEXTURE_2D,0,0,0,r,n,e.RGBA,o,i)):we(e,()=>e.texImage2D(e.TEXTURE_2D,0,l,r,n,0,e.RGBA,o,i)),we(e,()=>e.bindTexture(e.TEXTURE_2D,null))}function ZI(e,t,r){we(e,()=>e.bindTexture(e.TEXTURE_2D,t)),r.data instanceof Uint8Array?Z().getNumber("WEBGL_VERSION")===2?we(e,()=>e.texSubImage2D(e.TEXTURE_2D,0,0,0,r.width,r.height,e.RGBA,e.UNSIGNED_BYTE,r.data)):we(e,()=>e.texImage2D(e.TEXTURE_2D,0,e.RGBA,r.width,r.height,0,e.RGBA,e.UNSIGNED_BYTE,r.data)):Z().getNumber("WEBGL_VERSION")===2?we(e,()=>e.texSubImage2D(e.TEXTURE_2D,0,0,0,e.RGBA,e.UNSIGNED_BYTE,r)):we(e,()=>e.texImage2D(e.TEXTURE_2D,0,e.RGBA,e.RGBA,e.UNSIGNED_BYTE,r)),we(e,()=>e.bindTexture(e.TEXTURE_2D,null))}function YI(e,t,r,n){let a=e.createBuffer();we(e,()=>e.bindBuffer(e.PIXEL_PACK_BUFFER,a));let s=4*4*t*r;return we(e,()=>e.bufferData(e.PIXEL_PACK_BUFFER,s,e.STREAM_READ)),we(e,()=>e.readPixels(0,0,r,t,e.RGBA,e.FLOAT,0)),we(e,()=>e.bindBuffer(e.PIXEL_PACK_BUFFER,null)),a}function JI(e,t,r){let n=e,a=new Float32Array(r);return n.bindBuffer(n.PIXEL_PACK_BUFFER,t),n.getBufferSubData(n.PIXEL_PACK_BUFFER,0,a),n.bindBuffer(n.PIXEL_PACK_BUFFER,null),a}function QI(e,t,r,n){let[a,s]=Dh(t,r),i=4,o=new Uint8Array(NQ(t*r,i));return we(e,()=>e.readPixels(0,0,a,s,n.downloadTextureFormat,e.UNSIGNED_BYTE,o)),new Float32Array(o.buffer)}function e9(e,t,r,n,a,s,i,o){let l=e,u=new Float32Array(EQ(s,i));return l.bindBuffer(l.PIXEL_PACK_BUFFER,t),l.getBufferSubData(l.PIXEL_PACK_BUFFER,0,u),l.bindBuffer(l.PIXEL_PACK_BUFFER,null),u}function t9(e,t,r){let n=new Float32Array(t*r*4);return we(e,()=>e.readPixels(0,0,r,t,e.RGBA,e.FLOAT,n)),n}var vu=class{constructor(e){this.outputTexture=null,this.program=null,this.disposed=!1,this.vertexAttrsAreBound=!1,this.itemsToPoll=[];let t=Z().getNumber("WEBGL_VERSION");e!=null?(this.gl=e,Im(t,e)):this.gl=xa(t);let r="WEBGL_color_buffer_float",n="EXT_color_buffer_half_float";if(this.parallelCompilationExtension=this.gl.getExtension("KHR_parallel_shader_compile"),Z().getNumber("WEBGL_VERSION")===1){let a="OES_texture_float",s="OES_texture_half_float";if(this.textureFloatExtension=Np(this.gl,a),En(this.gl,s))this.textureHalfFloatExtension=Np(this.gl,s);else if(Z().get("WEBGL_FORCE_F16_TEXTURES"))throw new Error("GL context does not support half float textures, yet the environment flag WEBGL_FORCE_F16_TEXTURES is set to true.");if(this.colorBufferFloatExtension=this.gl.getExtension(r),En(this.gl,n))this.colorBufferHalfFloatExtension=Np(this.gl,n);else if(Z().get("WEBGL_FORCE_F16_TEXTURES"))throw new Error("GL context does not support color renderable half floats, yet the environment flag WEBGL_FORCE_F16_TEXTURES is set to true.")}else if(r="EXT_color_buffer_float",En(this.gl,r))this.colorBufferFloatExtension=this.gl.getExtension(r);else if(En(this.gl,n))this.colorBufferHalfFloatExtension=this.gl.getExtension(n);else throw new Error("GL context does not support color renderable floats");this.vertexBuffer=WI(this.gl),this.indexBuffer=VI(this.gl),this.framebuffer=bI(this.gl),this.textureConfig=h5(this.gl,this.textureHalfFloatExtension)}get debug(){return Z().getBool("DEBUG")}dispose(){if(this.disposed)return;this.program!=null&&console.warn("Disposing a GPGPUContext that still has a bound WebGLProgram. This is probably a resource leak, delete the program with GPGPUContext.deleteProgram before disposing."),this.outputTexture!=null&&console.warn("Disposing a GPGPUContext that still has a bound output matrix texture. This is probably a resource leak, delete the output matrix texture with GPGPUContext.deleteMatrixTexture before disposing.");let e=this.gl;we(e,()=>e.finish()),we(e,()=>e.bindFramebuffer(e.FRAMEBUFFER,null)),we(e,()=>e.deleteFramebuffer(this.framebuffer)),we(e,()=>e.bindBuffer(e.ARRAY_BUFFER,null)),we(e,()=>e.bindBuffer(e.ELEMENT_ARRAY_BUFFER,null)),we(e,()=>e.deleteBuffer(this.indexBuffer)),this.disposed=!0}createFloat32MatrixTexture(e,t){return this.throwIfDisposed(),UI(this.gl,e,t,this.textureConfig)}createFloat16MatrixTexture(e,t){return this.throwIfDisposed(),GI(this.gl,e,t,this.textureConfig)}createUnsignedBytesMatrixTexture(e,t){return this.throwIfDisposed(),jI(this.gl,e,t,this.textureConfig)}uploadPixelDataToTexture(e,t){this.throwIfDisposed(),ZI(this.gl,e,t)}uploadDenseMatrixToTexture(e,t,r,n){this.throwIfDisposed(),KI(this.gl,e,t,r,n,this.textureConfig)}createFloat16PackedMatrixTexture(e,t){return this.throwIfDisposed(),qI(this.gl,e,t,this.textureConfig)}createPackedMatrixTexture(e,t){return this.throwIfDisposed(),HI(this.gl,e,t,this.textureConfig)}deleteMatrixTexture(e){this.throwIfDisposed(),this.outputTexture===e&&(ty(this.gl,this.framebuffer),this.outputTexture=null),we(this.gl,()=>this.gl.deleteTexture(e))}downloadByteEncodedFloatMatrixFromOutputTexture(e,t,r){return this.downloadMatrixDriver(e,()=>QI(this.gl,t,r,this.textureConfig))}downloadPackedMatrixFromBuffer(e,t,r,n,a,s){return e9(this.gl,e,t,r,n,a,s,this.textureConfig)}downloadFloat32MatrixFromBuffer(e,t){return JI(this.gl,e,t)}createBufferFromTexture(e,t,r){this.bindTextureToFrameBuffer(e);let n=YI(this.gl,t,r,this.textureConfig);return this.unbindTextureToFrameBuffer(),n}createAndWaitForFence(){let e=this.createFence(this.gl);return this.pollFence(e)}createFence(e){let t,r;if(Z().getBool("WEBGL_FENCE_API_ENABLED")){let n=e,a=n.fenceSync(n.SYNC_GPU_COMMANDS_COMPLETE,0);e.flush(),r=()=>{let s=n.clientWaitSync(a,0,0);return s===n.ALREADY_SIGNALED||s===n.CONDITION_SATISFIED},t=a}else Z().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION")>0?(t=this.beginQuery(),this.endQuery(),r=()=>this.isQueryAvailable(t,Z().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION"))):r=()=>!0;return{query:t,isFencePassed:r}}downloadMatrixFromPackedTexture(e,t,r){return this.downloadMatrixDriver(e,()=>t9(this.gl,t,r))}createProgram(e){this.throwIfDisposed();let t=this.gl;this.vertexShader==null&&(this.vertexShader=BI(t));let r=fI(t);return we(t,()=>t.attachShader(r,this.vertexShader)),we(t,()=>t.attachShader(r,e)),mI(t,r),this.debug&&l0(t,r),this.vertexAttrsAreBound||(this.setProgram(r),this.vertexAttrsAreBound=XI(t,this.program,this.vertexBuffer)),r}deleteProgram(e){this.throwIfDisposed(),e===this.program&&(this.program=null),e!=null&&we(this.gl,()=>this.gl.deleteProgram(e))}setProgram(e){this.throwIfDisposed(),this.program=e,this.program!=null&&this.debug&&l0(this.gl,this.program),we(this.gl,()=>this.gl.useProgram(e))}getUniformLocation(e,t,r=!0){return this.throwIfDisposed(),r?wI(this.gl,e,t):kI(this.gl,e,t)}getAttributeLocation(e,t){return this.throwIfDisposed(),we(this.gl,()=>this.gl.getAttribLocation(e,t))}getUniformLocationNoThrow(e,t){return this.throwIfDisposed(),this.gl.getUniformLocation(e,t)}setInputMatrixTexture(e,t,r){this.throwIfDisposed(),this.throwIfNoProgram(),II(this.gl,e,t,r)}setOutputMatrixTexture(e,t,r){this.setOutputMatrixTextureDriver(e,r,t)}setOutputPackedMatrixTexture(e,t,r){this.throwIfDisposed();let[n,a]=kd(t,r);this.setOutputMatrixTextureDriver(e,n,a)}setOutputMatrixWriteRegion(e,t,r,n){this.setOutputMatrixWriteRegionDriver(r,e,n,t)}setOutputPackedMatrixWriteRegion(e,t,r,n){throw new Error("setOutputPackedMatrixWriteRegion not implemented.")}debugValidate(){this.program!=null&&l0(this.gl,this.program),Ep(this.gl)}executeProgram(){this.throwIfDisposed(),this.throwIfNoProgram();let e=this.gl;this.debug&&this.debugValidate(),we(e,()=>e.drawElements(e.TRIANGLES,6,e.UNSIGNED_SHORT,0))}blockUntilAllProgramsCompleted(){this.throwIfDisposed(),we(this.gl,()=>this.gl.finish())}getQueryTimerExtension(){return this.disjointQueryTimerExtension==null&&(this.disjointQueryTimerExtension=Np(this.gl,Z().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION")===2?"EXT_disjoint_timer_query_webgl2":"EXT_disjoint_timer_query")),this.disjointQueryTimerExtension}getQueryTimerExtensionWebGL2(){return this.getQueryTimerExtension()}getQueryTimerExtensionWebGL1(){return this.getQueryTimerExtension()}beginQuery(){if(Z().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION")===2){let r=this.gl,n=this.getQueryTimerExtensionWebGL2(),a=r.createQuery();return r.beginQuery(n.TIME_ELAPSED_EXT,a),a}let e=this.getQueryTimerExtensionWebGL1(),t=e.createQueryEXT();return e.beginQueryEXT(e.TIME_ELAPSED_EXT,t),t}endQuery(){if(Z().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION")===2){let t=this.gl,r=this.getQueryTimerExtensionWebGL2();t.endQuery(r.TIME_ELAPSED_EXT);return}let e=this.getQueryTimerExtensionWebGL1();e.endQueryEXT(e.TIME_ELAPSED_EXT)}async waitForQueryAndGetTime(e){return await v.repeatedTry(()=>this.disposed||this.isQueryAvailable(e,Z().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION"))),this.getQueryTime(e,Z().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION"))}getQueryTime(e,t){if(t===0)return null;if(t===2){let r=this.gl;return r.getQueryParameter(e,r.QUERY_RESULT)/1e6}else{let r=this.getQueryTimerExtensionWebGL1();return r.getQueryObjectEXT(e,r.QUERY_RESULT_EXT)/1e6}}isQueryAvailable(e,t){if(t===0)return!0;if(t===2){let r=this.gl,n=this.getQueryTimerExtensionWebGL2(),a=r.getQueryParameter(e,r.QUERY_RESULT_AVAILABLE);return this.disjoint==null&&(this.disjoint=this.gl.getParameter(n.GPU_DISJOINT_EXT)),a&&!this.disjoint}else{let r=this.getQueryTimerExtensionWebGL1(),n=r.getQueryObjectEXT(e,r.QUERY_RESULT_AVAILABLE_EXT);return this.disjoint==null&&(this.disjoint=this.gl.getParameter(r.GPU_DISJOINT_EXT)),n&&!this.disjoint}}pollFence(e){return new Promise(t=>{this.addItemToPoll(()=>e.isFencePassed(),()=>t())})}pollItems(){let e=Fee(this.itemsToPoll.map(t=>t.isDoneFn));for(let t=0;t<=e;++t){let{resolveFn:r}=this.itemsToPoll[t];r()}this.itemsToPoll=this.itemsToPoll.slice(e+1)}addItemToPoll(e,t){this.itemsToPoll.push({isDoneFn:e,resolveFn:t}),!(this.itemsToPoll.length>1)&&v.repeatedTry(()=>(this.pollItems(),this.itemsToPoll.length===0))}bindTextureToFrameBuffer(e){this.throwIfDisposed(),u0(this.gl,e,this.framebuffer),this.debug&&Ep(this.gl)}unbindTextureToFrameBuffer(){this.outputTexture!=null?(u0(this.gl,this.outputTexture,this.framebuffer),this.debug&&Ep(this.gl)):ty(this.gl,this.framebuffer)}downloadMatrixDriver(e,t){this.bindTextureToFrameBuffer(e);let r=t();return this.unbindTextureToFrameBuffer(),r}setOutputMatrixTextureDriver(e,t,r){this.throwIfDisposed();let n=this.gl;u0(n,e,this.framebuffer),this.debug&&Ep(n),this.outputTexture=e,we(n,()=>n.viewport(0,0,t,r)),we(n,()=>n.scissor(0,0,t,r))}setOutputMatrixWriteRegionDriver(e,t,r,n){this.throwIfDisposed(),we(this.gl,()=>this.gl.scissor(e,t,r,n))}throwIfDisposed(){if(this.disposed)throw new Error("Attempted to use disposed GPGPUContext.")}throwIfNoProgram(){if(this.program==null)throw new Error("No GPU program is currently set.")}};function Fee(e){let t=0;for(;t<e.length&&e[t]();++t);return t-1}var{addImpl:Pee,bincountImpl:r9,bincountReduceImpl:_ee,ceilImpl:zee,concatImpl:Oee,equalImpl:Dee,expImpl:Lee,expm1Impl:Bee,floorImpl:Wee,gatherNdImpl:Vee,gatherV2Impl:Uee,greaterImpl:Gee,greaterEqualImpl:jee,lessImpl:Hee,lessEqualImpl:qee,linSpaceImpl:Xee,logImpl:Kee,maxImpl:Zee,maximumImpl:Yee,minimumImpl:Jee,multiplyImpl:Qee,negImpl:ete,notEqualImpl:tte,prodImpl:rte,rangeImpl:nte,rsqrtImpl:ate,sigmoidImpl:ste,simpleAbsImpl:n9,sliceImpl:ite,sparseFillEmptyRowsImpl:ote,sparseReshapeImpl:lte,sparseSegmentReductionImpl:a9,sqrtImpl:ute,stridedSliceImpl:dte,stringNGramsImpl:pte,stringSplitImpl:hte,stringToHashBucketFastImpl:cte,subImpl:fte,tileImpl:mte,topKImpl:gte,transposeImpl:w5,uniqueImpl:yte}=wm;function s9(e,t){return["x","y","z","w","u","v"].slice(0,t).map(r=>`${e}.${r}`)}function Br(e,t){return t===1?[e]:s9(e,t)}function Ate(e,t){if(e===1)return"rc";let r="";for(let n=0;n<e;n++)r+=t[n],n<e-1&&(r+=",");return r}var xte=class{constructor(e){if(this.variableNames=["A"],this.packedInputs=!1,this.packedOutput=!0,this.outputShape=e,this.rank=e.length,this.enableShapeUniforms=ln(this.outputShape.length),this.rank===0)this.userCode=`
|
|
void main() {
|
|
setOutput(vec4(getA(), 0., 0., 0.));
|
|
}
|
|
`;else{let t=Br("rc",this.rank),r=vt(this.rank),n=this.getOutOfBoundsCondition(t),a=this.getSetup(t),s=this.getOutput(t);this.userCode=`
|
|
void main() {
|
|
${r} rc = getOutputCoords();
|
|
|
|
if(${n}) {
|
|
setOutput(vec4(0));
|
|
} else {
|
|
${a}
|
|
|
|
setOutput(vec4(${s}));
|
|
}
|
|
}
|
|
`}}getSourceCoordsArr(e){let t=[];for(let r=0;r<=1;r++)for(let n=0;n<=1;n++){let a=`${r===0?"r":"rp1"}, ${n===0?"c":"cp1"}`;for(let s=2;s<this.rank;s++)a=`${e[e.length-1-s]},`+a;t.push(a)}return t}getOutOfBoundsCondition(e){if(this.rank===1)return`rc > ${this.enableShapeUniforms?"outShape":this.outputShape[0]}`;let t="";for(let r=this.rank-2;r<this.rank;r++)t+=`${e[r]} >= ${this.enableShapeUniforms?`outShape[${r}]`:this.outputShape[r]}`,r<this.rank-1&&(t+="||");return t}getSetup(e){if(this.rank===1)return"";let t=e.slice(-2),r=this.enableShapeUniforms?`outShape[${this.rank} - 1]`:this.outputShape[this.rank-1],n=this.enableShapeUniforms?`outShape[${this.rank} - 2]`:this.outputShape[this.rank-2];return`
|
|
int r = ${t[0]};
|
|
int c = ${t[1]};
|
|
int rp1 = r + 1;
|
|
int cp1 = c + 1;
|
|
|
|
bool cEdge = cp1 >= ${r};
|
|
bool rEdge = rp1 >= ${n};
|
|
`}getOutput(e){let t=this.getSourceCoordsArr(e);return this.rank===1?`getA(rc), (rc + 1 >= ${this.enableShapeUniforms?"outShape":this.outputShape[0]} ? 0. : getA(rc + 1)), 0, 0`:`getA(${t[0]}),
|
|
cEdge ? 0. : getA(${t[1]}),
|
|
rEdge ? 0. : getA(${t[2]}),
|
|
rEdge || cEdge ? 0. : getA(${t[3]})`}},i9=class{constructor(e,t){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0,this.customUniforms=[{name:"inputShape",type:"ivec3"}],this.outputShape=e,this.enableShapeUniforms=ln(this.outputShape.length);let r="";for(let n=0;n<4;n++){let a="thisRC = rc;";n%2===1&&(a+="thisRC.z += 1;"),n>1&&(a+="thisRC.y += 1;"),r+=`
|
|
${a}
|
|
${n>0?"if(thisRC.y < rows && thisRC.z < cols){":""}
|
|
int flatIndex = getFlatIndex(thisRC);
|
|
|
|
ivec3 inputRC = inputCoordsFromReshapedOutCoords(flatIndex);
|
|
vec2 inputRCInnerDims = vec2(float(inputRC.y),float(inputRC.z));
|
|
|
|
result[${n}] =
|
|
getChannel(getA(inputRC.x, inputRC.y, inputRC.z), inputRCInnerDims);
|
|
${n>0?"}":""}
|
|
`}this.userCode=`
|
|
${bte(t,this.enableShapeUniforms)}
|
|
${this.enableShapeUniforms?m5():f5(e)}
|
|
|
|
void main() {
|
|
ivec3 rc = getOutputCoords();
|
|
|
|
vec4 result = vec4(0.);
|
|
|
|
ivec3 thisRC;
|
|
int rows = ${this.enableShapeUniforms?"outShape[1]":e[1]};
|
|
int cols = ${this.enableShapeUniforms?"outShape[2]":e[2]};
|
|
|
|
${r}
|
|
|
|
setOutput(result);
|
|
}
|
|
`}};function bte(e,t){return`
|
|
ivec3 inputCoordsFromReshapedOutCoords(int index) {
|
|
${t?WQ(["r","c","d"],"inputShape"):Dl(["r","c","d"],e)}
|
|
return ivec3(r, c, d);
|
|
}
|
|
`}var vte=class{constructor(e){this.gpgpu=e,this.numUsedTextures=0,this.numFreeTextures=0,this._numBytesAllocated=0,this._numBytesFree=0,this.freeTextures={},this.logEnabled=!1,this.usedTextures={}}acquireTexture(e,t,r){let n=L4(t,r),a=B4(e,n,r);a in this.freeTextures||(this.freeTextures[a]=[]),a in this.usedTextures||(this.usedTextures[a]=[]);let s=D4(e,n,this.gpgpu.gl,this.gpgpu.textureConfig,r);if(this.freeTextures[a].length>0){this.numFreeTextures--,this.numUsedTextures++,this._numBytesFree-=s,this.log();let o=this.freeTextures[a].shift();return this.usedTextures[a].push(o),o}let i;return n===3?i=this.gpgpu.createPackedMatrixTexture(e[0],e[1]):n===4?i=this.gpgpu.createFloat16PackedMatrixTexture(e[0],e[1]):n===1?i=this.gpgpu.createFloat32MatrixTexture(e[0],e[1]):n===0?i=this.gpgpu.createFloat16MatrixTexture(e[0],e[1]):n===2&&(i=this.gpgpu.createUnsignedBytesMatrixTexture(e[0],e[1])),this.usedTextures[a].push(i),this.numUsedTextures++,this._numBytesAllocated+=s,this.log(),i}releaseTexture(e,t,r,n){if(this.freeTextures==null)return;let a=L4(r,n),s=B4(t,a,n);s in this.freeTextures||(this.freeTextures[s]=[]);let i=D4(t,a,this.gpgpu.gl,this.gpgpu.textureConfig,n),o=Z().get("WEBGL_DELETE_TEXTURE_THRESHOLD");o!==-1&&this._numBytesAllocated>o?(this.gpgpu.deleteMatrixTexture(e.texture),this._numBytesAllocated-=i):(this.freeTextures[s].push(e),this.numFreeTextures++,this._numBytesFree+=i),this.numUsedTextures--;let l=this.usedTextures[s],u=l.indexOf(e);if(u<0)throw new Error("Cannot release a texture that was never provided by this texture manager");l.splice(u,1),this.log()}log(){if(!this.logEnabled)return;let e=this.numFreeTextures+this.numUsedTextures;console.log("Free/Used",`${this.numFreeTextures} / ${this.numUsedTextures}`,`(${e})`);let t=this._numBytesFree/this._numBytesAllocated;console.log(`Bytes allocated: ${this._numBytesAllocated}`),console.log(`Bytes unused: ${this._numBytesFree} (${Math.round(100*t)}%)`)}get numBytesAllocated(){return this._numBytesAllocated}get numBytesFree(){return this._numBytesFree}getNumUsedTextures(){return this.numUsedTextures}getNumFreeTextures(){return this.numFreeTextures}dispose(){if(this.freeTextures!=null){for(let e in this.freeTextures)this.freeTextures[e].forEach(t=>{this.gpgpu.deleteMatrixTexture(t.texture)});for(let e in this.usedTextures)this.usedTextures[e].forEach(t=>{this.gpgpu.deleteMatrixTexture(t.texture)});this.freeTextures=null,this.usedTextures=null,this.numUsedTextures=0,this.numFreeTextures=0,this._numBytesAllocated=0,this._numBytesFree=0}}};function wte(e,t){let r=e;if(t===r.R32F)return 4;if(t===r.R16F)return 2;if(t===r.RGBA32F||t===e.RGBA)return 16;if(t===r.RGBA16F)return 8;if(t===r.RGBA8)return 4;throw new Error(`Unknown internal format ${t}`)}function D4(e,t,r,n,a){let s=kte(t,n),i;if(a){let[l,u]=kd(e[0],e[1]);i=l*u}else{let[l,u]=Dh(e[0],e[1]);i=l*u}let o=wte(r,s);return i*o}function kte(e,t){switch(e){case 3:return b5(t);case 4:return v5(t);case 1:return y5(t);case 0:return A5(t);case 2:return x5(t);default:throw new Error(`Unknown physical texture type ${e}`)}}function Ite(e){return Z().getBool("WEBGL_RENDER_FLOAT32_ENABLED")?e?3:1:e?4:0}function L4(e,t){if(e===1)return 3;if(e===0||e==null)return Ite(t);if(e===3||e===2)return 2;throw new Error(`Unknown logical texture type ${e}`)}function B4(e,t,r){return`${e[0]}_${e[1]}_${t}_${r}`}var Xa=class{constructor(e,t){this.variableNames=["A"],this.outputShape=e,this.enableShapeUniforms=ln(this.outputShape.length),this.userCode=`
|
|
float unaryOperation(float x) {
|
|
${t}
|
|
}
|
|
|
|
void main() {
|
|
float x = getAAtOutCoords();
|
|
float y = unaryOperation(x);
|
|
|
|
setOutput(y);
|
|
}
|
|
`}},Yn="if (isnan(x)) return x;",Ste="return x;",W4="return abs(x);",Cte="return (x >= 0.0) ? x : (exp(x) - 1.0);",Tte=Yn+`
|
|
return (x < 0.0) ? 0.0 : x;
|
|
`,Nte=Yn+`
|
|
return (x < 0.0) ? 0.0 : min(6.0, x);
|
|
`,du="return x;",Ete="return 1.0 / (1.0 + exp(-1.0 * x));",Rte="return x;",$te=`
|
|
vec4 result;
|
|
|
|
result.r = (x.r >= 0.0) ? x.r : (exp(x.r) - 1.0);
|
|
result.g = (x.g >= 0.0) ? x.g : (exp(x.g) - 1.0);
|
|
result.b = (x.b >= 0.0) ? x.b : (exp(x.b) - 1.0);
|
|
result.a = (x.a >= 0.0) ? x.a : (exp(x.a) - 1.0);
|
|
|
|
return result;
|
|
`,Mte=`
|
|
vec4 result = x * vec4(greaterThanEqual(x, vec4(0.0)));
|
|
bvec4 isNaN = isnan(x);
|
|
|
|
result.r = isNaN.r ? x.r : result.r;
|
|
result.g = isNaN.g ? x.g : result.g;
|
|
result.b = isNaN.b ? x.b : result.b;
|
|
result.a = isNaN.a ? x.a : result.a;
|
|
|
|
return result;
|
|
`,Fte=`
|
|
vec4 result = min(x, vec4(6.)) * vec4(greaterThanEqual(x, vec4(0.0)));
|
|
bvec4 isNaN = isnan(x);
|
|
|
|
result.r = isNaN.r ? x.r : result.r;
|
|
result.g = isNaN.g ? x.g : result.g;
|
|
result.b = isNaN.b ? x.b : result.b;
|
|
result.a = isNaN.a ? x.a : result.a;
|
|
|
|
return result;
|
|
`,Pte="return 1.0 / (1.0 + exp(-1.0 * x));",No=class{constructor(e,t){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0,this.outputShape=e,this.enableShapeUniforms=ln(this.outputShape.length),this.userCode=`
|
|
vec4 unaryOperation(vec4 x) {
|
|
${t}
|
|
}
|
|
|
|
void main() {
|
|
vec4 x = getAAtOutCoords();
|
|
vec4 y = unaryOperation(x);
|
|
|
|
setOutput(y);
|
|
}
|
|
`}},_te=class{constructor(e){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!1,this.outputShape=e,this.enableShapeUniforms=ln(this.outputShape.length);let t=e.length,r=Br("rc",t),n=vt(t),a=Ate(t,r),s=r.slice(-2),i=t<=1?"rc":`vec2(${s.join(",")})`;this.userCode=`
|
|
void main() {
|
|
${n} rc = getOutputCoords();
|
|
vec4 packedInput = getA(${a});
|
|
|
|
setOutput(getChannel(packedInput, ${i}));
|
|
}
|
|
`}},zte=Kn.whereImpl,Ote=1e-7,Dte=1e-4,cg={};function Lte(e){return e in cg||(cg[e]={}),cg[e]}var Bte=Z().getNumber("CPU_HANDOFF_SIZE_THRESHOLD"),Wte=600;function Vte(){return Z().global.screen==null?1024:Z().global.screen.height*Z().global.screen.width*window.devicePixelRatio*Wte/1024/1024}var o9=class extends zu{constructor(e){if(super(),this.pendingRead=new WeakMap,this.pendingDisposal=new WeakSet,this.dataRefCount=new WeakMap,this.numBytesInGPU=0,this.uploadWaitMs=0,this.downloadWaitMs=0,this.lastGlFlushTime=0,this.warnedAboutMemory=!1,this.pendingDeletes=0,this.disposed=!1,!Z().getBool("HAS_WEBGL"))throw new Error("WebGL is not supported on this device");let t;if(e!=null){if(e instanceof vu)t=e;else{let r=xa(Z().getNumber("WEBGL_VERSION"),e);t=new vu(r)}this.binaryCache={},this.gpgpuCreatedLocally=!1}else{let r=xa(Z().getNumber("WEBGL_VERSION"));t=new vu(r),this.binaryCache=Lte(Z().getNumber("WEBGL_VERSION")),this.gpgpuCreatedLocally=!0}this.gpgpu=t,this.canvas=this.gpgpu.gl.canvas,this.textureManager=new vte(this.gpgpu),this.numMBBeforeWarning=Vte(),this.texData=new Zp(this,Xt())}nextDataId(){return o9.nextDataId++}numDataIds(){return this.texData.numDataIds()-this.pendingDeletes}write(e,t,r){if((Z().getBool("WEBGL_CHECK_NUMERICAL_PROBLEMS")||Z().getBool("DEBUG"))&&this.checkNumericalProblems(e),r==="complex64"&&e!=null)throw new Error("Cannot write to a complex64 dtype. Please use tf.complex(real, imag).");let n={id:this.nextDataId()};return this.texData.set(n,{shape:t,dtype:r,values:e,usage:1,refCount:1}),n}refCount(e){return this.texData.has(e)?this.texData.get(e).refCount:0}incRef(e){let t=this.texData.get(e);t.refCount++}decRef(e){if(this.texData.has(e)){let t=this.texData.get(e);t.refCount--}}move(e,t,r,n,a){if(Z().getBool("DEBUG")&&this.checkNumericalProblems(t),n==="complex64")throw new Error("Cannot write to a complex64 dtype. Please use tf.complex(real, imag).");this.texData.set(e,{shape:r,dtype:n,values:t,usage:1,refCount:a})}disposeIntermediateTensorInfo(e){this.disposeData(e.dataId)}readSync(e){let t=this.texData.get(e),{values:r,dtype:n,complexTensorInfos:a,slice:s,shape:i,isPacked:o}=t;if(s!=null){let h;o?h=new No(i,du):h=new Xa(i,du);let p=this.runWebGLProgram(h,[{dataId:e,shape:i,dtype:n}],n),c=this.readSync(p.dataId);return this.disposeIntermediateTensorInfo(p),c}if(r!=null)return this.convertAndCacheOnCPU(e);if(n==="string")return r;let l=this.activeTimers!=null,u;l&&(u=v.now());let d;if(n==="complex64"){let h=this.readSync(a.real.dataId),p=this.readSync(a.imag.dataId);d=C.mergeRealAndImagArrays(h,p)}else d=this.getValuesFromTexture(e);return l&&(this.downloadWaitMs+=v.now()-u),this.convertAndCacheOnCPU(e,d)}async read(e){if(this.pendingRead.has(e)){let c=this.pendingRead.get(e);return new Promise(f=>c.push(f))}let t=this.texData.get(e),{values:r,shape:n,slice:a,dtype:s,complexTensorInfos:i,isPacked:o}=t;if(a!=null){let c;o?c=new No(n,du):c=new Xa(n,du);let f=this.runWebGLProgram(c,[{dataId:e,shape:n,dtype:s}],s),m=this.read(f.dataId);return this.disposeIntermediateTensorInfo(f),m}if(r!=null)return this.convertAndCacheOnCPU(e);if(Z().getBool("DEBUG")&&!Z().getBool("WEBGL_DOWNLOAD_FLOAT_ENABLED")&&Z().getNumber("WEBGL_VERSION")===2)throw new Error("tensor.data() with WEBGL_DOWNLOAD_FLOAT_ENABLED=false and WEBGL_VERSION=2 not yet supported.");let l=null,u;if(s!=="complex64"&&Z().get("WEBGL_BUFFER_SUPPORTED")){u=this.decode(e);let c=this.texData.get(u.dataId);l=this.gpgpu.createBufferFromTexture(c.texture.texture,...t0(n))}this.pendingRead.set(e,[]),s!=="complex64"&&await this.gpgpu.createAndWaitForFence();let d;if(s==="complex64"){let c=await Promise.all([this.read(i.real.dataId),this.read(i.imag.dataId)]),f=c[0],m=c[1];d=C.mergeRealAndImagArrays(f,m)}else if(l==null)d=this.getValuesFromTexture(e);else{let c=v.sizeFromShape(n);d=this.gpgpu.downloadFloat32MatrixFromBuffer(l,c)}if(u!=null&&this.disposeIntermediateTensorInfo(u),l!=null){let c=this.gpgpu.gl;we(c,()=>c.deleteBuffer(l))}let h=this.convertAndCacheOnCPU(e,d),p=this.pendingRead.get(e);return this.pendingRead.delete(e),p.forEach(c=>c(h)),this.pendingDisposal.has(e)&&(this.pendingDisposal.delete(e),this.disposeData(e)&&Xt().removeDataId(e,this),this.pendingDeletes--),h}readToGPU(e,t={}){let r=this.texData.get(e),{values:n,shape:a,slice:s,dtype:i,isPacked:o,texture:l}=r;if(i==="complex64")throw new Error("Does not support reading texture for complex64 dtype.");if(s!=null){let p;o?p=new No(a,du):p=new Xa(a,du);let c=this.runWebGLProgram(p,[{dataId:e,shape:a,dtype:i}],i),f=this.readToGPU(c,t);return this.disposeIntermediateTensorInfo(c),f}if(l==null)throw n!=null?new Error("Data is not on GPU but on CPU."):new Error("There is no data on GPU or CPU.");let u=this.decode(e,t.customTexShape),d=Xt().makeTensorFromTensorInfo(u),h=this.texData.get(u.dataId);return{tensorRef:d,...h.texture}}bufferSync(e){let t=this.readSync(e.dataId),r=t;if(e.dtype==="string")try{r=t.map(n=>v.decodeString(n))}catch(n){throw new Error("Failed to decode encoded string bytes into utf-8")}return We(e.shape,e.dtype,r)}checkNumericalProblems(e){if(e!=null)for(let t=0;t<e.length;t++){let r=e[t];if(!dI(r))throw Z().getBool("WEBGL_RENDER_FLOAT32_CAPABLE")?Error(`The value ${r} cannot be represented with your current settings. Consider enabling float32 rendering: 'tf.env().set('WEBGL_RENDER_FLOAT32_ENABLED', true);'`):Error(`The value ${r} cannot be represented on this device.`)}}getValuesFromTexture(e){let{shape:t,dtype:r,isPacked:n}=this.texData.get(e),a=v.sizeFromShape(t);if(Z().getBool("WEBGL_DOWNLOAD_FLOAT_ENABLED")){let h=this.decode(e),p=this.texData.get(h.dataId),c=this.gpgpu.downloadMatrixFromPackedTexture(p.texture.texture,...t0(t)).subarray(0,a);return this.disposeIntermediateTensorInfo(h),c}let s=Z().getBool("WEBGL_PACK")&&n===!0,i=s?d0(t):t,o=s?new Ree(i):new Eee(i),l=this.runWebGLProgram(o,[{shape:i,dtype:r,dataId:e}],"float32"),u=this.texData.get(l.dataId),d=this.gpgpu.downloadByteEncodedFloatMatrixFromOutputTexture(u.texture.texture,u.texShape[0],u.texShape[1]).subarray(0,a);return this.disposeIntermediateTensorInfo(l),d}timerAvailable(){return Z().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE")>0}time(e){let t=this.activeTimers,r=[],n=!1;this.programTimersStack==null?(this.programTimersStack=r,n=!0):this.activeTimers.push(r),this.activeTimers=r,e();let a=v.flatten(this.activeTimers.map(o=>o.query)).filter(o=>o!=null),s=v.flatten(this.activeTimers.map(o=>o.name)).filter(o=>o!=null);this.activeTimers=t,n&&(this.programTimersStack=null);let i={uploadWaitMs:this.uploadWaitMs,downloadWaitMs:this.downloadWaitMs,kernelMs:null,wallMs:null};return(async()=>{if(Z().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE")>0){let o=await Promise.all(a);i.kernelMs=v.sum(o),i.getExtraProfileInfo=()=>o.map((l,u)=>({name:s[u],ms:l})).map(l=>`${l.name}: ${l.ms}`).join(", ")}else i.kernelMs={error:"WebGL query timers are not supported in this environment."};return this.uploadWaitMs=0,this.downloadWaitMs=0,i})()}memory(){return{unreliable:!1,numBytesInGPU:this.numBytesInGPU,numBytesInGPUAllocated:this.textureManager.numBytesAllocated,numBytesInGPUFree:this.textureManager.numBytesFree}}startTimer(){return Z().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE")>0?this.gpgpu.beginQuery():{startMs:v.now(),endMs:null}}endTimer(e){return Z().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE")>0?(this.gpgpu.endQuery(),e):(e.endMs=v.now(),e)}async getQueryTime(e){if(Z().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE")>0)return this.gpgpu.waitForQueryAndGetTime(e);let t=e;return t.endMs-t.startMs}disposeData(e,t=!1){if(this.pendingDisposal.has(e))return!1;if(!this.texData.has(e))return!0;if(t?this.texData.get(e).refCount=0:this.texData.get(e).refCount--,!t&&this.texData.get(e).refCount>0)return!1;if(this.pendingRead.has(e))return this.pendingDisposal.add(e),this.pendingDeletes++,!1;this.releaseGPUData(e);let{complexTensorInfos:r}=this.texData.get(e);return r!=null&&(this.disposeData(r.real.dataId,t),this.disposeData(r.imag.dataId,t)),this.texData.delete(e),!0}releaseGPUData(e){let{texture:t,dtype:r,texShape:n,usage:a,isPacked:s,slice:i}=this.texData.get(e),o=i&&i.origDataId||e,l=this.dataRefCount.get(o);l>1?this.dataRefCount.set(o,l-1):(this.dataRefCount.delete(o),t!=null&&(this.numBytesInGPU-=this.computeBytes(n,r),this.textureManager.releaseTexture(t,n,a,s)));let u=this.texData.get(e);u.texture=null,u.texShape=null,u.isPacked=!1,u.slice=null}getTexture(e){return this.uploadToGPU(e),this.texData.get(e).texture.texture}getDataInfo(e){return this.texData.get(e)}shouldExecuteOnCPU(e,t=Bte){return Z().getBool("WEBGL_CPU_FORWARD")&&e.every(r=>this.texData.get(r.dataId).texture==null&&v.sizeFromShape(r.shape)<t)}getGPGPUContext(){return this.gpgpu}where(e){C.warn("tf.where() in webgl locks the UI thread. Call tf.whereAsync() instead");let t=e.dataSync();return zte(e.shape,t)}packedUnaryOp(e,t,r){let n=new No(e.shape,t),a=this.compileAndRun(n,[e],r);return Xt().makeTensorFromTensorInfo(a)}abs(e){if(this.shouldExecuteOnCPU([e])&&e.dtype!=="complex64"){let n=n9(this.texData.get(e.dataId).values);return this.makeOutput(e.shape,e.dtype,n)}if(Z().getBool("WEBGL_PACK_UNARY_OPERATIONS"))return this.packedUnaryOp(e,W4,e.dtype);let t=new Xa(e.shape,W4),r=this.compileAndRun(t,[e]);return Xt().makeTensorFromTensorInfo(r)}makeTensorInfo(e,t,r){let n;if(t==="string"&&r!=null&&r.length>0&&v.isString(r[0])){let a=r.map(s=>v.encodeString(s));n=this.write(a,e,t)}else n=this.write(r,e,t);return this.texData.get(n).usage=null,{dataId:n,shape:e,dtype:t}}makeOutput(e,t,r){return Xt().makeTensorFromTensorInfo(this.makeTensorInfo(e,t,r),this)}unpackTensor(e){let t=new _te(e.shape);return this.runWebGLProgram(t,[e],e.dtype)}packTensor(e){let t=new xte(e.shape),r=!0;return this.runWebGLProgram(t,[e],e.dtype,null,r)}packedReshape(e,t){let r=[Wo(e.shape),...Vo(e.shape)],n={dtype:e.dtype,shape:r,dataId:e.dataId},a=[Wo(t),...Vo(t)],s=new i9(a,r),i=!0,o=[r],l=this.runWebGLProgram(s,[n],e.dtype,o,i);return{dataId:l.dataId,shape:t,dtype:l.dtype}}decode(e,t){let r=this.texData.get(e),{isPacked:n,shape:a,dtype:s}=r;if(t!=null){let h=v.sizeFromShape(a),p=t[0]*t[1]*4;v.assert(h<=p,()=>"customTexShape is too small. Row * Column * 4 should be equal or larger than the size of the tensor data.")}let i=d0(a),o;n?o=new Nee(i):o=new Tee(i);let l=!0,u=[t!=null?t:t0(i)],d=this.runWebGLProgram(o,[{shape:i,dtype:s,dataId:e}],s,u,l,t);return{dtype:s,shape:a,dataId:d.dataId}}runWebGLProgram(e,t,r,n,a=!1,s){let i=this.makeTensorInfo(e.outputShape,r),o=this.texData.get(i.dataId);if(e.packedOutput&&(o.isPacked=!0),e.outPackingScheme===0){let g=s!=null?s:t0(e.outputShape);o.texShape=g.map(y=>y*2)}if(e.outTexUsage!=null&&(o.usage=e.outTexUsage),v.sizeFromShape(i.shape)===0)return o.values=v.getTypedArrayFromDType(i.dtype,0),i;let l=[],u=t.map(g=>{if(g.dtype==="complex64")throw new Error("GPGPUProgram does not support complex64 input. For complex64 dtypes, please separate the program into real and imaginary parts.");let y=this.texData.get(g.dataId);if(y.texture==null){if(!e.packedInputs&&v.sizeFromShape(g.shape)<=Z().getNumber("WEBGL_SIZE_UPLOAD_UNIFORM"))return{shape:g.shape,texData:null,isUniform:!0,uniformValues:y.values};e.packedInputs&&(y.isPacked=!0,y.shape=g.shape)}if(this.uploadToGPU(g.dataId),!!y.isPacked!=!!e.packedInputs)g=y.isPacked?this.unpackTensor(g):this.packTensor(g),l.push(g),y=this.texData.get(g.dataId);else if(y.isPacked&&!Xp(y.shape,g.shape)){let A=g,x=g.shape;g.shape=y.shape,g=this.packedReshape(g,x),l.push(g),y=this.texData.get(g.dataId),A.shape=x}return{shape:g.shape,texData:y,isUniform:!1}});this.uploadToGPU(i.dataId);let d={shape:i.shape,texData:o,isUniform:!1},h=Cee(e,u,d),p=this.getAndSaveBinary(h,()=>Iee(this.gpgpu,e,u,d)),c=this.activeTimers!=null,f;c&&(f=this.startTimer()),Z().get("ENGINE_COMPILE_ONLY")||See(this.gpgpu,p,u,d,n),l.forEach(g=>this.disposeIntermediateTensorInfo(g)),c&&(f=this.endTimer(f),this.activeTimers.push({name:e.constructor.name,query:this.getQueryTime(f)}));let m=Z().get("WEBGL_FLUSH_THRESHOLD");if(m>0){let g=v.now();g-this.lastGlFlushTime>m&&(this.gpgpu.gl.flush(),this.lastGlFlushTime=g)}if(!Z().getBool("WEBGL_LAZILY_UNPACK")&&o.isPacked&&a===!1){let g=this.unpackTensor(i);return this.disposeIntermediateTensorInfo(i),g}return i}compileAndRun(e,t,r,n,a=!1){return r=r||t[0].dtype,this.runWebGLProgram(e,t,r,n,a)}getAndSaveBinary(e,t){return e in this.binaryCache||(this.binaryCache[e]=t()),this.binaryCache[e]}getTextureManager(){return this.textureManager}dispose(){this.disposed||(Z().getBool("IS_TEST")||Object.keys(this.binaryCache).forEach(e=>{this.gpgpu.deleteProgram(this.binaryCache[e].webGLProgram),delete this.binaryCache[e]}),this.textureManager.dispose(),this.canvas!=null&&typeof HTMLCanvasElement!="undefined"&&this.canvas instanceof HTMLCanvasElement?this.canvas.remove():this.canvas=null,this.gpgpuCreatedLocally&&(this.gpgpu.program=null,this.gpgpu.dispose()),this.disposed=!0)}floatPrecision(){return this.floatPrecisionValue==null&&(this.floatPrecisionValue=X(()=>{if(!Z().get("WEBGL_RENDER_FLOAT32_ENABLED")){let e=Z().getBool("DEBUG");Z().set("DEBUG",!1);let t=this.abs(Se(1e-8)).dataSync()[0];if(Z().set("DEBUG",e),t>0)return 32}return 16})),this.floatPrecisionValue}epsilon(){return this.floatPrecision()===32?Ote:Dte}uploadToGPU(e){let t=this.texData.get(e),{shape:r,dtype:n,values:a,texture:s,usage:i,isPacked:o}=t;if(s!=null)return;let l=this.activeTimers!=null,u;l&&(u=v.now());let d=t.texShape;if(d==null&&(d=TI(r,o),t.texShape=d),a!=null){let h=d0(r),p,c=d[1],f=d[0],m=a instanceof Uint8Array||a instanceof Uint8ClampedArray;(o||!m)&&([c,f]=kd(d[0],d[1])),o?p=new Mee(h,m):p=new $ee(h,m);let g=m?[f,c]:d,y=this.makeTensorInfo(g,n),A=this.texData.get(y.dataId);m?A.usage=2:A.usage=1,A.texShape=g,this.gpgpu.uploadDenseMatrixToTexture(this.getTexture(y.dataId),c,f,a);let x=[[f,c]],b=!0,w=this.runWebGLProgram(p,[y],n,x,b),S=this.texData.get(w.dataId);t.texShape=S.texShape,t.isPacked=S.isPacked,t.usage=S.usage,Z().get("ENGINE_COMPILE_ONLY")?this.disposeData(w.dataId):(t.texture=S.texture,t.values=null,this.texData.delete(w.dataId)),this.disposeIntermediateTensorInfo(y),l&&(this.uploadWaitMs+=v.now()-u)}else{let h=this.acquireTexture(d,i,n,o);t.texture=h}}convertAndCacheOnCPU(e,t){let r=this.texData.get(e),{dtype:n}=r;return this.releaseGPUData(e),t!=null&&(r.values=Ute(t,n)),r.values}acquireTexture(e,t,r,n){if(this.numBytesInGPU+=this.computeBytes(e,r),!this.warnedAboutMemory&&this.numBytesInGPU>this.numMBBeforeWarning*1024*1024){let a=(this.numBytesInGPU/1024/1024).toFixed(2);this.warnedAboutMemory=!0,console.warn(`High memory usage in GPU: ${a} MB, most likely due to a memory leak`)}return this.textureManager.acquireTexture(e,t,n)}computeBytes(e,t){return e[0]*e[1]*v.bytesPerElement(t)}checkCompileCompletion(){for(let[,e]of Object.entries(this.binaryCache))this.checkCompletion_(e)}async checkCompileCompletionAsync(){let e=[];if(this.gpgpu.parallelCompilationExtension){for(let[,t]of Object.entries(this.binaryCache))e.push(this.checkCompletionAsync_(t));return Promise.all(e)}else{for(let[,t]of Object.entries(this.binaryCache)){let r=new Promise(n=>{try{this.checkCompletion_(t),n(!0)}catch(a){throw a}});e.push(r)}return Promise.all(e)}}async checkCompletionAsync_(e){return this.gpgpu.gl.getProgramParameter(e.webGLProgram,this.gpgpu.parallelCompilationExtension.COMPLETION_STATUS_KHR)?this.checkCompletion_(e):(await bA(),this.checkCompletionAsync_(e))}checkCompletion_(e){if(this.gpgpu.gl.getProgramParameter(e.webGLProgram,this.gpgpu.gl.LINK_STATUS)===!1)throw console.log(this.gpgpu.gl.getProgramInfoLog(e.webGLProgram)),this.gpgpu.gl.getShaderParameter(e.fragmentShader,this.gpgpu.gl.COMPILE_STATUS)===!1?(c5(e.source,this.gpgpu.gl.getShaderInfoLog(e.fragmentShader)),new Error("Failed to compile fragment shader.")):new Error("Failed to link vertex and fragment shaders.");return!0}getUniformLocations(){for(let[,e]of Object.entries(this.binaryCache)){let{uniformLocations:t,customUniformLocations:r,infLoc:n,nanLoc:a,inShapesLocations:s,inTexShapesLocations:i,outShapeLocation:o,outShapeStridesLocation:l,outTexShapeLocation:u}=DI(this.gpgpu,e.program,e.webGLProgram);e.uniformLocations=t,e.customUniformLocations=r,e.infLoc=n,e.nanLoc=a,e.inShapesLocations=s,e.inTexShapesLocations=i,e.outShapeLocation=o,e.outShapeStridesLocation=l,e.outTexShapeLocation=u}}},Bh=o9;Bh.nextDataId=0;function Ute(e,t){if(t==="float32"||t==="complex64")return e;if(t==="int32"||t==="bool"){let r=t==="int32"?new Int32Array(e.length):new Uint8Array(e.length);for(let n=0;n<r.length;++n)r[n]=Math.round(e[n]);return r}else throw new Error(`Unknown dtype ${t}`)}var Gte="0.0.0";function l9(){Z().set("WEBGL_FORCE_F16_TEXTURES",!0)}yh.isBrowser()&&$l("webgl",()=>new Bh,2);var jte={forceHalfFloat:l9},u9=`
|
|
if (isnan(a)) return a;
|
|
if (isnan(b)) return b;
|
|
`,_u=class{constructor(e,t,r){this.variableNames=["A","B"],this.outputShape=C.assertAndGetBroadcastShape(t,r),this.enableShapeUniforms=ln(this.outputShape.length),this.userCode=`
|
|
float binaryOperation(float a, float b) {
|
|
${e}
|
|
}
|
|
|
|
void main() {
|
|
float a = getAAtOutCoords();
|
|
float b = getBAtOutCoords();
|
|
setOutput(binaryOperation(a, b));
|
|
}
|
|
`}},Cm=`
|
|
result.r = isNaN.r > 0. ? NAN : result.r;
|
|
result.g = isNaN.g > 0. ? NAN : result.g;
|
|
result.b = isNaN.b > 0. ? NAN : result.b;
|
|
result.a = isNaN.a > 0. ? NAN : result.a;
|
|
`,Wh=class{constructor(e,t,r,n=!1){this.variableNames=["A","B"],this.supportsBroadcasting=!0,this.packedInputs=!0,this.packedOutput=!0,this.outputShape=C.assertAndGetBroadcastShape(t,r);let a=this.outputShape.length;this.enableShapeUniforms=ln(a);let s="";if(n)if(a===0||v.sizeFromShape(this.outputShape)===1)s=`
|
|
result.y = 0.;
|
|
result.z = 0.;
|
|
result.w = 0.;
|
|
`;else if(s=`
|
|
${vt(a)} coords = getOutputCoords();
|
|
`,a===1)this.enableShapeUniforms?s+=`
|
|
result.y = (coords + 1) >= outShape ? 0. : result.y;
|
|
result.z = 0.;
|
|
result.w = 0.;
|
|
`:s+=`
|
|
result.y = (coords + 1) >= ${this.outputShape[0]} ? 0. : result.y;
|
|
result.z = 0.;
|
|
result.w = 0.;
|
|
`;else{let i=Br("coords",a);this.enableShapeUniforms?s+=`
|
|
bool nextRowOutOfBounds =
|
|
(${i[a-2]} + 1) >= outShape[${a} - 2];
|
|
bool nextColOutOfBounds =
|
|
(${i[a-1]} + 1) >= outShape[${a} - 1];
|
|
result.y = nextColOutOfBounds ? 0. : result.y;
|
|
result.z = nextRowOutOfBounds ? 0. : result.z;
|
|
result.w = nextColOutOfBounds || nextRowOutOfBounds ? 0. : result.w;
|
|
`:s+=`
|
|
bool nextRowOutOfBounds =
|
|
(${i[a-2]} + 1) >= ${this.outputShape[a-2]};
|
|
bool nextColOutOfBounds =
|
|
(${i[a-1]} + 1) >= ${this.outputShape[a-1]};
|
|
result.y = nextColOutOfBounds ? 0. : result.y;
|
|
result.z = nextRowOutOfBounds ? 0. : result.z;
|
|
result.w = nextColOutOfBounds || nextRowOutOfBounds ? 0. : result.w;
|
|
`}this.userCode=`
|
|
vec4 binaryOperation(vec4 a, vec4 b) {
|
|
${e}
|
|
}
|
|
|
|
void main() {
|
|
vec4 a = getAAtOutCoords();
|
|
vec4 b = getBAtOutCoords();
|
|
|
|
vec4 result = binaryOperation(a, b);
|
|
${s}
|
|
|
|
setOutput(result);
|
|
}
|
|
`}};function fn(e){let{inputs:t,backend:r}=e,{x:n}=t;return r.incRef(n.dataId),{dataId:n.dataId,shape:n.shape,dtype:n.dtype}}var Hte={kernelName:mi,backendName:"webgl",kernelFunc:fn};function Xi(e){let{inputs:t,backend:r}=e,{real:n,imag:a}=t,s=r.makeTensorInfo(n.shape,"complex64"),i=r.texData.get(s.dataId),o=fn({inputs:{x:n},backend:r}),l=fn({inputs:{x:a},backend:r});return i.complexTensorInfos={real:o,imag:l},s}var qte={kernelName:Jp,backendName:"webgl",kernelFunc:Xi},d9="return (a < 0.) ? b * a : a;",p9=`
|
|
vec4 aLessThanZero = vec4(lessThan(a, vec4(0.)));
|
|
return (aLessThanZero * (b * a)) + ((vec4(1.0) - aLessThanZero) * a);
|
|
`;function Xte(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{alpha:s}=n,i=r.makeTensorInfo([],"float32",v.createScalarValue(s,"float32")),o=Z().getBool("WEBGL_PACK_BINARY_OPERATIONS")?new Wh(p9,a.shape,i.shape):new _u(d9,a.shape,i.shape),l=r.runWebGLProgram(o,[a,i],"float32");return r.disposeIntermediateTensorInfo(i),l}var Kte={kernelName:gi,backendName:"webgl",kernelFunc:Xte},h9="return (a < 0.) ? b * a : a;",c9=`
|
|
vec4 aLessThanZero = vec4(lessThan(a, vec4(0.)));
|
|
return (aLessThanZero * (b * a)) + ((vec4(1.0) - aLessThanZero) * a);
|
|
`;function Zte(e){let{inputs:t,backend:r}=e,{x:n,alpha:a}=t,s=Z().getBool("WEBGL_PACK_BINARY_OPERATIONS")?new Wh(c9,n.shape,a.shape):new _u(h9,n.shape,a.shape);return r.runWebGLProgram(s,[n,a],"float32")}var Yte={kernelName:Ni,backendName:"webgl",kernelFunc:Zte},Ed="if (isnan(x)) return x;",Jte=`
|
|
if (isnan(a)) return a;
|
|
if (isnan(b)) return b;
|
|
`,Qte=`
|
|
result.r = isNaN.r > 0. ? NAN : result.r;
|
|
result.g = isNaN.g > 0. ? NAN : result.g;
|
|
result.b = isNaN.b > 0. ? NAN : result.b;
|
|
result.a = isNaN.a > 0. ? NAN : result.a;
|
|
`;function it({opSnippet:e,packedOpSnippet:t,cpuKernelImpl:r,dtype:n}){return({inputs:a,backend:s})=>{let{x:i}=a,o=s,l=n||i.dtype;if(o.shouldExecuteOnCPU([i])&&r!=null){let h=o.texData.get(i.dataId),p=r(h.values,l);return o.makeTensorInfo(i.shape,l,p)}let u=Z().getBool("WEBGL_PACK_UNARY_OPERATIONS")&&t!=null,d;return u?d=new No(i.shape,t):d=new Xa(i.shape,e),o.runWebGLProgram(d,[i],l)}}function wr({opSnippet:e,packedOpSnippet:t,checkOutOfBounds:r=!1,supportsComplex:n=!1,cpuKernelImpl:a,dtype:s}){return({inputs:i,backend:o})=>{let{a:l,b:u}=i,d=o;if(n&&l.dtype==="complex64"){let f=d.texData.get(l.dataId),m=d.texData.get(u.dataId),[g,y]=[[f.complexTensorInfos.real,m.complexTensorInfos.real],[f.complexTensorInfos.imag,m.complexTensorInfos.imag]].map(x=>{let[b,w]=x,S={dataId:b.dataId,dtype:b.dtype,shape:l.shape},N={dataId:w.dataId,dtype:w.dtype,shape:u.shape},E=new _u(e,l.shape,u.shape);return d.runWebGLProgram(E,[S,N],Nr(b.dtype,w.dtype))}),A=Xi({inputs:{real:g,imag:y},backend:d});return d.disposeIntermediateTensorInfo(g),d.disposeIntermediateTensorInfo(y),A}let h=s||Nr(l.dtype,u.dtype);if((l.dtype==="string"||u.dtype==="string"||d.shouldExecuteOnCPU([l,u]))&&a!=null){let f=d.texData.get(l.dataId).values,m=d.texData.get(u.dataId).values,g=l.dtype==="string"?C.fromUint8ToStringArray(f):f,y=l.dtype==="string"?C.fromUint8ToStringArray(m):m,[A,x]=a(l.shape,u.shape,g,y,h),b=d.makeTensorInfo(x,h),w=d.texData.get(b.dataId);return w.values=A,b}let p=Z().getBool("WEBGL_PACK_BINARY_OPERATIONS")&&t!=null,c;return p?c=new Wh(t,l.shape,u.shape,r):c=new _u(e,l.shape,u.shape),d.runWebGLProgram(c,[l,u],h)}}function Tm(e,t=!1){if(e==="linear")return t?Rte:Ste;if(e==="relu")return t?Mte:Tte;if(e==="elu")return t?$te:Cte;if(e==="relu6")return t?Fte:Nte;if(e==="prelu")return t?c9:h9;if(e==="leakyrelu")return t?p9:d9;if(e==="sigmoid")return t?Pte:Ete;throw new Error(`Activation ${e} has not been implemented for the WebGL backend.`)}var f9=class{constructor(e,t,r,n=!1,a=!1,s=!1,i=null,o=!1,l=!1){this.variableNames=["matrixA","matrixB"],this.packedInputs=!0,this.packedOutput=!0,this.outputShape=r,this.enableShapeUniforms=ln(this.outputShape.length);let u=n?e[1]:e[2],d=Math.ceil(u/2),h=n?"i * 2, rc.y":"rc.y, i * 2",p=a?"rc.z, i * 2":"i * 2, rc.z",c=n?["a.xxyy","a.zzww"]:["a.xxzz","a.yyww"],f=a?["b.xzxz","b.ywyw"]:["b.xyxy","b.zwzw"],m="",g="";i&&(o?m=`vec4 activation(vec4 a) {
|
|
vec4 b = getPreluActivationWeightsAtOutCoords();
|
|
${i}
|
|
}`:l?m=`vec4 activation(vec4 a) {
|
|
vec4 b = getLeakyreluAlphaAtOutCoords();
|
|
${i}
|
|
}`:m=`vec4 activation(vec4 x) {
|
|
${i}
|
|
}`,g="result = activation(result);");let y=s?"result += getBiasAtOutCoords();":"";s&&this.variableNames.push("bias"),o&&this.variableNames.push("preluActivationWeights"),l&&this.variableNames.push("leakyreluAlpha");let A="rc.x",x="rc.x";e[0]<t[0]?A=`int(min(float(rc.x), ${e[0]-1}.))`:t[0]<e[0]&&(x=`int(min(float(rc.x), ${t[0]-1}.))`),this.userCode=`
|
|
${m}
|
|
// Don't use uniform for sharedDimensionPacked for performance.
|
|
const float sharedDimension = ${d}.0;
|
|
|
|
vec4 dot2x2ARowBCol(ivec3 rc) {
|
|
vec4 result = vec4(0);
|
|
for (int i = 0; i < ${d}; i++) {
|
|
int batchA = ${A};
|
|
int batchB = ${x};
|
|
vec4 a = getMatrixA(batchA, ${h});
|
|
vec4 b = getMatrixB(batchB, ${p});
|
|
|
|
// These swizzled products need to be separately added.
|
|
// See: https://github.com/tensorflow/tfjs/issues/1735
|
|
result += (${c[0]} * ${f[0]});
|
|
result += (${c[1]} * ${f[1]});
|
|
}
|
|
return result;
|
|
}
|
|
|
|
void main() {
|
|
ivec3 rc = getOutputCoords();
|
|
vec4 result = dot2x2ARowBCol(rc);
|
|
|
|
${y}
|
|
|
|
${g}
|
|
|
|
setOutput(result);
|
|
}
|
|
`}},V4={REAL:"return areal * breal - aimag * bimag;",IMAG:"return areal * bimag + aimag * breal;"},U4=class{constructor(e,t,r){this.variableNames=["AReal","AImag","BReal","BImag"],this.outputShape=C.assertAndGetBroadcastShape(t,r),this.userCode=`
|
|
float binaryOpComplex(
|
|
float areal, float aimag, float breal, float bimag) {
|
|
${e}
|
|
}
|
|
|
|
void main() {
|
|
float areal = getARealAtOutCoords();
|
|
float aimag = getAImagAtOutCoords();
|
|
float breal = getBRealAtOutCoords();
|
|
float bimag = getBImagAtOutCoords();
|
|
setOutput(binaryOpComplex(areal, aimag, breal, bimag));
|
|
}
|
|
`}},G4="return a * b;";function k5(e){let{inputs:t,backend:r}=e,{a:n,b:a}=t,s=C.upcastType(n.dtype,a.dtype);if(n.dtype==="complex64"){let o=r.texData.get(n.dataId),l=r.texData.get(a.dataId),u=new U4(V4.REAL,n.shape,a.shape),d=new U4(V4.IMAG,n.shape,a.shape),h=[{dataId:o.complexTensorInfos.real.dataId,dtype:o.complexTensorInfos.real.dtype,shape:n.shape},{dataId:o.complexTensorInfos.imag.dataId,dtype:o.complexTensorInfos.imag.dtype,shape:n.shape},{dataId:l.complexTensorInfos.real.dataId,dtype:l.complexTensorInfos.real.dtype,shape:a.shape},{dataId:l.complexTensorInfos.imag.dataId,dtype:l.complexTensorInfos.imag.dtype,shape:a.shape}],p=r.runWebGLProgram(u,h,"float32"),c=r.runWebGLProgram(d,h,"float32"),f=Xi({inputs:{real:p,imag:c},backend:r});return r.disposeIntermediateTensorInfo(p),r.disposeIntermediateTensorInfo(c),f}if(r.shouldExecuteOnCPU([n,a])){let o=r.texData.get(n.dataId),l=r.texData.get(a.dataId),[u,d]=Qee(n.shape,a.shape,o.values,l.values,s),h=r.makeTensorInfo(d,s),p=r.texData.get(h.dataId);return p.values=u,h}let i;return Z().getBool("WEBGL_PACK_BINARY_OPERATIONS")?i=new Wh(G4,n.shape,a.shape):i=new _u(G4,n.shape,a.shape),r.runWebGLProgram(i,[n,a],s)}var ere={kernelName:Si,backendName:"webgl",kernelFunc:k5};function tre(e,t,r){let n=[Wo(e.shape),...Vo(e.shape)],a={dtype:e.dtype,shape:n,dataId:e.dataId},s=[Wo(t),...Vo(t)],i=new i9(s,n),o=!0,l=[n],u=r.runWebGLProgram(i,[a],e.dtype,l,o);return{dataId:u.dataId,shape:t,dtype:u.dtype}}function ve(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{shape:s}=n,i=r,o=v.sizeFromShape(a.shape),l=v.inferFromImplicitShape(s,o),u=v.sizeFromShape(l);v.assert(o===u,()=>`The new shape (${l}) has ${u} elements and the old shape (${a.shape}) has ${o} elements. The new shape and old shape must have the same number of elements.`);let d=i.texData.get(a.dataId);return d.isPacked&&!Xp(a.shape,l)&&!(d.texture!==null&&Xp(d.shape,l))?tre(a,l,i):(i.incRef(a.dataId),{dataId:a.dataId,shape:l,dtype:a.dtype})}var rre={kernelName:ml,backendName:"webgl",kernelFunc:ve},j4=class{constructor(e,t){this.variableNames=["x"];let{windowSize:r,batchSize:n,inSize:a,outSize:s}=e;this.outputShape=[n,s];let i=Math.floor(r/4)*4,o=r%4,l="sumValue += dot(values, ones);";if(t!=null){let d=1/t;l=`sumValue += dot(values * ${v.isInt(d)?d.toPrecision(2):d}, ones);`}let u="";a%r>0&&(u=`
|
|
if (inIdx < 0 || inIdx >= ${a}) {
|
|
return 0.0;
|
|
}
|
|
`),this.userCode=`
|
|
const vec4 ones = vec4(1.0, 1.0, 1.0, 1.0);
|
|
|
|
float getValue(int batch, int inIdx) {
|
|
${u}
|
|
return getX(batch, inIdx);
|
|
}
|
|
|
|
void main() {
|
|
ivec2 coords = getOutputCoords();
|
|
int batch = coords[0];
|
|
int outIdx = coords[1];
|
|
int inOffset = outIdx * ${r};
|
|
|
|
float sumValue = 0.0;
|
|
|
|
for (int i = 0; i < ${i}; i += 4) {
|
|
int inIdx = inOffset + i;
|
|
vec4 values = vec4(
|
|
getValue(batch, inIdx),
|
|
getValue(batch, inIdx + 1),
|
|
getValue(batch, inIdx + 2),
|
|
getValue(batch, inIdx + 3)
|
|
);
|
|
|
|
${l}
|
|
}
|
|
|
|
int inIdx = inOffset + ${i};
|
|
if (${o===1}) {
|
|
vec4 values = vec4(getValue(batch, inIdx), 0.0, 0.0, 0.0);
|
|
|
|
${l}
|
|
} else if (${o===2}) {
|
|
vec4 values = vec4(
|
|
getValue(batch, inIdx),
|
|
getValue(batch, inIdx + 1), 0.0, 0.0);
|
|
|
|
${l}
|
|
} else if (${o===3}) {
|
|
vec4 values = vec4(
|
|
getValue(batch, inIdx),
|
|
getValue(batch, inIdx + 1),
|
|
getValue(batch, inIdx + 2), 0.0);
|
|
|
|
${l}
|
|
}
|
|
setOutput(sumValue);
|
|
}
|
|
`}},nre=class{constructor(e,t){this.variableNames=["x"];let{windowSize:r,batchSize:n,inSize:a,outSize:s}=e;this.outputShape=[n,s];let i="0.0",o="";t==="prod"?i="1.0":t==="min"?(i="1.0 / 1e-20",o="min"):t==="max"&&(i="-1.0 / 1e-20",o="max");let l=`${t}(${t}(${t}(minMaxValue[0], minMaxValue[1]), minMaxValue[2]), minMaxValue[3])`;t==="sum"?l="sumValue":t==="prod"?l="prodValue":t==="all"?l="allValue":t==="any"&&(l="anyValue");let u=Math.floor(r/4)*4,d=r%4,h=`
|
|
if (${t==="sum"}) {
|
|
sumValue += dot(values, ones);
|
|
} else if (${t==="prod"}) {
|
|
vec2 tmp = vec2(values[0], values[1]) * vec2(values[2], values[3]);
|
|
prodValue *= tmp[0] * tmp[1];
|
|
} else {
|
|
minMaxValue = ${o}(values, minMaxValue);
|
|
if (${t==="min"} || ${t==="max"}) {
|
|
minMaxValue = ${o}(values, minMaxValue);
|
|
bvec4 isNaN = isnan(values);
|
|
if (isNaN.r || isNaN.g || isNaN.b || isNaN.a) {
|
|
minMaxValue = vec4(NAN);
|
|
}
|
|
}
|
|
}
|
|
`,p="vec4";t==="all"?(i="1.0",h=`
|
|
bool reducedAllValue = all(values);
|
|
float floatedReducedAllValue = float(reducedAllValue);
|
|
allValue = float(allValue >= 1.0 && floatedReducedAllValue >= 1.0);
|
|
`,p="bvec4"):t==="any"&&(i="0.0",h=`
|
|
bool reducedAnyValue = any(values);
|
|
float floatedReducedAnyValue = float(reducedAnyValue);
|
|
anyValue = float(anyValue >= 1.0 || floatedReducedAnyValue >= 1.0);
|
|
`,p="bvec4");let c="";a%r>0&&(c=`
|
|
if (inIdx < 0 || inIdx >= ${a}) {
|
|
return initializationValue;
|
|
}
|
|
`),this.userCode=`
|
|
const float initializationValue = ${i};
|
|
const vec4 ones = vec4(1.0, 1.0, 1.0, 1.0);
|
|
|
|
float getValue(int batch, int inIdx) {
|
|
${c}
|
|
return getX(batch, inIdx);
|
|
}
|
|
|
|
void main() {
|
|
ivec2 coords = getOutputCoords();
|
|
int batch = coords[0];
|
|
int outIdx = coords[1];
|
|
int inOffset = outIdx * ${r};
|
|
|
|
vec4 minMaxValue = vec4(${i});
|
|
float prodValue = 1.0;
|
|
float sumValue = 0.0;
|
|
float allValue = 1.0;
|
|
float anyValue = 0.0;
|
|
|
|
for (int i = 0; i < ${u}; i += 4) {
|
|
int inIdx = inOffset + i;
|
|
${p} values = ${p}(
|
|
getValue(batch, inIdx),
|
|
getValue(batch, inIdx + 1),
|
|
getValue(batch, inIdx + 2),
|
|
getValue(batch, inIdx + 3)
|
|
);
|
|
|
|
${h}
|
|
}
|
|
|
|
int inIdx = inOffset + ${u};
|
|
if (${d===1}) {
|
|
${p} values = ${p}(
|
|
getValue(batch, inIdx),
|
|
initializationValue,
|
|
initializationValue,
|
|
initializationValue
|
|
);
|
|
|
|
${h}
|
|
} else if (${d===2}) {
|
|
${p} values = ${p}(
|
|
getValue(batch, inIdx),
|
|
getValue(batch, inIdx + 1),
|
|
initializationValue,
|
|
initializationValue
|
|
);
|
|
|
|
${h}
|
|
} else if (${d===3}) {
|
|
${p} values = ${p}(
|
|
getValue(batch, inIdx),
|
|
getValue(batch, inIdx + 1),
|
|
getValue(batch, inIdx + 2),
|
|
initializationValue
|
|
);
|
|
|
|
${h}
|
|
}
|
|
setOutput(${l});
|
|
}
|
|
`}};function are(e){let t=[];for(;t.length===0||t[t.length-1].outSize!==1;){let r=t.length?t[t.length-1].outSize:e[1],n=C.computeOptimalWindowSize(r);t.push({inSize:r,windowSize:n,outSize:Math.ceil(r/n)})}return t}function Bl(e,t,r,n){let a=are(e.shape),s=e;for(let i=0;i<a.length;i++){let{inSize:o,windowSize:l,outSize:u}=a[i],d,h;r==="mean"?d=i===0?new j4({windowSize:l,inSize:o,batchSize:e.shape[0],outSize:u},o):new j4({windowSize:l,inSize:o,batchSize:e.shape[0],outSize:u}):d=new nre({windowSize:l,inSize:o,batchSize:e.shape[0],outSize:u},r),h=s,s=n.runWebGLProgram(d,[s],t),h.dataId!==e.dataId&&n.disposeIntermediateTensorInfo(h)}return s}var sre=class{constructor(e,t){this.variableNames=["A"];let r=new Array(e.length);for(let s=0;s<r.length;s++)r[s]=e[t[s]];this.outputShape=r,this.rank=r.length;let n=vt(this.rank),a=ire(t);this.userCode=`
|
|
void main() {
|
|
${n} resRC = getOutputCoords();
|
|
setOutput(getA(${a}));
|
|
}
|
|
`}};function ire(e){let t=e.length;if(t>6)throw Error(`Transpose for rank ${t} is not yet supported`);let r=["resRC.x","resRC.y","resRC.z","resRC.w","resRC.u","resRC.v"],n=new Array(t);for(let a=0;a<e.length;a++)n[e[a]]=r[a];return n.join()}var ore=class{constructor(e,t){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0;let r=new Array(e.length);for(let u=0;u<r.length;u++)r[u]=e[t[u]];if(this.outputShape=r,this.rank=r.length,this.rank>6)throw Error(`Packed transpose for rank ${this.rank} is not yet supported.`);let n=vt(this.rank),a=s9("rc",this.rank),s=new Array(this.rank);for(let u=0;u<t.length;u++)s[t[u]]=a[u];let i=`vec2(${s.slice(-2).join()})`,o=`++${a[this.rank-1]} < ${r[this.rank-1]}`,l=`getChannel(getA(${s.join()}), ${i})`;this.userCode=`
|
|
void main() {
|
|
${n} rc = getOutputCoords();
|
|
vec4 result = vec4(0.);
|
|
result[0] = ${l};
|
|
if(${o}) {
|
|
result[1] = ${l};
|
|
}
|
|
--${a[this.rank-1]};
|
|
if(++${a[this.rank-2]} < ${r[this.rank-2]}) {
|
|
result[2] = ${l};
|
|
if(${o}) {
|
|
result[3] = ${l};
|
|
}
|
|
}
|
|
setOutput(result);
|
|
}
|
|
`}};function Nm(e,t,r){let n=Z().getBool("WEBGL_PACK_ARRAY_OPERATIONS")?new ore(e.shape,t):new sre(e.shape,t);return r.runWebGLProgram(n,[e],e.dtype)}function lre(e,t,r,n){let a=t,s=e.shape.length,i=v.parseAxisParam(a,e.shape),o=i,l=C.getAxesPermutation(o,s),u=l!=null,d=e;u&&(d=Nm(e,l,n),o=C.getInnerMostAxes(o.length,s)),C.assertAxesAreInnerMostDims("sum",o,s);let[h,p]=C.computeOutAndReduceShapes(d.shape,o),c=h;r&&(c=C.expandShapeToKeepDim(h,i));let f=v.sizeFromShape(p),m=v.sizeFromShape(e.shape)/f,g=ve({inputs:{x:d},attrs:{shape:[m,f]},backend:n}),y=gh(e.dtype),A=Bl(g,y,"sum",n),x=ve({inputs:{x:A},attrs:{shape:c},backend:n});return n.disposeIntermediateTensorInfo(g),n.disposeIntermediateTensorInfo(A),u&&n.disposeIntermediateTensorInfo(d),x}function Em(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{axis:s,keepDims:i}=n;return lre(a,s,i,r)}var ure={kernelName:Oi,backendName:"webgl",kernelFunc:Em};function dr(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{perm:s}=n,i=r,o=a.shape.length,l=new Array(o);for(let d=0;d<l.length;d++)l[d]=a.shape[s[d]];let u;if(i.shouldExecuteOnCPU([a])){let d=i.texData.get(a.dataId).values,h=w5(d,a.shape,a.dtype,s,l);u=i.makeTensorInfo(l,a.dtype);let p=i.texData.get(u.dataId);p.values=h}else u=Nm(a,s,i);return u}var dre={kernelName:Vi,backendName:"webgl",kernelFunc:dr},m9=1e3;function G0({a:e,b:t,transposeA:r,transposeB:n,backend:a,bias:s=null,preluActivationWeights:i=null,leakyreluAlpha:o=0,activation:l=null}){let u=e.shape.length,d=t.shape.length,h=r?e.shape[u-2]:e.shape[u-1],p=n?t.shape[d-1]:t.shape[d-2],c=r?e.shape[u-1]:e.shape[u-2],f=n?t.shape[d-2]:t.shape[d-1],m=e.shape.slice(0,-2),g=t.shape.slice(0,-2),y=v.sizeFromShape(m),A=v.sizeFromShape(g),x=ji.assertAndGetBroadcastShape(e.shape.slice(0,-2),t.shape.slice(0,-2)).concat([c,f]);v.assert(h===p,()=>`Error in matMul: inner shapes (${h}) and (${p}) of Tensors with shapes ${e.shape} and ${t.shape} and transposeA=${r} and transposeB=${n} must match.`);let b=r?[y,h,c]:[y,c,h],w=n?[A,f,p]:[A,p,f],S=ve({inputs:{x:e},backend:a,attrs:{shape:b}}),N=ve({inputs:{x:t},backend:a,attrs:{shape:w}}),E=[S,N],R=Math.max(y,A),z=r?S.shape[1]:S.shape[2],M=s!=null,I=i!=null,_=l==="leakyrelu",O=l!=null?Tm(l,!0):null,q=M||I||_||O!=null,K;if((c===1||f===1)&&z>m9&&q===!1){let Q=S,V=N;r&&(Q=dr({inputs:{x:S},backend:a,attrs:{perm:[0,2,1]}}),E.push(Q)),n&&(V=dr({inputs:{x:N},backend:a,attrs:{perm:[0,2,1]}}),E.push(V));let re=f!==1,Y=f===1,ie=Q;re&&(ie=ve({inputs:{x:Q},backend:a,attrs:{shape:[R,z,1]}}),E.push(ie));let J=f===1?2:1,ae=V;Y&&(ae=ve({inputs:{x:V},backend:a,attrs:{shape:[R,1,z]}}),E.push(ae));let de=k5({inputs:{a:ie,b:ae},backend:a});K=Em({inputs:{x:de},backend:a,attrs:{axis:J,keepDims:!0}}),E.push(de)}else{let Q=Nr(e.dtype,t.dtype),V=new f9(b,w,[R,c,f],r,n,M,O,I,_),re=[S,N];if(s!=null&&re.push(s),I&&re.push(i),_){let Y=a.makeTensorInfo([],"float32",v.createScalarValue(o,"float32"));re.push(Y),E.push(Y)}K=a.runWebGLProgram(V,re,Q)}let D=ve({inputs:{x:K},backend:a,attrs:{shape:x}});E.push(K);for(let Q of E)a.disposeIntermediateTensorInfo(Q);return D}function pre(e){let{inputs:t,backend:r,attrs:n}=e,{a,b:s,bias:i,preluActivationWeights:o}=t,{transposeA:l,transposeB:u,activation:d,leakyreluAlpha:h}=n;return G0({a,b:s,transposeA:l,transposeB:u,backend:r,bias:i,preluActivationWeights:o,leakyreluAlpha:h,activation:d})}var hre={kernelName:Ps,backendName:"webgl",kernelFunc:pre},H4="return abs(x);";function cre(e){let{inputs:t,backend:r}=e,{x:n}=t;if(r.shouldExecuteOnCPU([n])&&n.dtype!=="complex64"){let s=r.texData.get(n.dataId),i=n9(s.values);return r.makeTensorInfo(n.shape,n.dtype,i)}let a;return Z().getBool("WEBGL_PACK_UNARY_OPERATIONS")?a=new No(n.shape,H4):a=new Xa(n.shape,H4),r.runWebGLProgram(a,[n],n.dtype)}var fre={kernelName:Ho,backendName:"webgl",kernelFunc:cre},mre=Yn+`
|
|
if (abs(x) > 1.) {
|
|
return NAN;
|
|
}
|
|
return acos(x);
|
|
`,gre=it({opSnippet:mre}),yre={kernelName:Du,backendName:"webgl",kernelFunc:gre},Are=Yn+`
|
|
if (x < 1.0) return NAN;
|
|
return log(x + sqrt(x * x - 1.0));`,xre=it({opSnippet:Are}),bre={kernelName:Lu,backendName:"webgl",kernelFunc:xre},q4="return a + b;",vre=wr({opSnippet:q4,packedOpSnippet:q4,supportsComplex:!0,cpuKernelImpl:Pee}),wre={kernelName:Ja,backendName:"webgl",kernelFunc:vre},kre=class{constructor(e,t){this.outputShape=[],this.outputShape=e,this.variableNames=t.map((a,s)=>`T${s}`);let r=[];this.variableNames.forEach(a=>{r.push(`float v${a} = get${a}AtOutCoords();`)});let n=this.variableNames.map(a=>`v${a}`).join(" + ");this.userCode=`
|
|
void main() {
|
|
${r.join(`
|
|
`)}
|
|
|
|
float result = ${n};
|
|
setOutput(result);
|
|
}
|
|
`}},Ire=class{constructor(e,t){this.outputShape=[],this.packedInputs=!0,this.packedOutput=!0,this.outputShape=e,this.variableNames=t.map((a,s)=>`T${s}`);let r=[];this.variableNames.forEach(a=>{r.push(`vec4 v${a} = get${a}AtOutCoords();`)});let n=this.variableNames.map(a=>`v${a}`).join(" + ");this.userCode=`
|
|
void main() {
|
|
${r.join(`
|
|
`)}
|
|
|
|
vec4 result = ${n};
|
|
setOutput(result);
|
|
}
|
|
`}};function c0(e){let{inputs:t,backend:r}=e,n=t;if(n.length===1)return fn({inputs:{x:n[0]},backend:r});if(n.length>Z().get("WEBGL_MAX_TEXTURES_IN_SHADER")){let o=Math.floor(n.length/2),l=c0({inputs:n.slice(0,o),backend:r}),u=c0({inputs:n.slice(o),backend:r});return c0({inputs:[l,u],backend:r})}let a=n.map(o=>o.dtype).reduce((o,l)=>Nr(o,l)),s=n.map(o=>o.shape),i=Z().getBool("WEBGL_PACK")?new Ire(n[0].shape,s):new kre(n[0].shape,s);return r.runWebGLProgram(i,n,a)}var Sre={kernelName:Zs,backendName:"webgl",kernelFunc:c0};function Cre(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{axis:s,keepDims:i}=n,o=a.shape.length,l=v.parseAxisParam(s,a.shape),u=l,d=C.getAxesPermutation(u,o),h=a;d!=null&&(h=dr({inputs:{x:a},backend:r,attrs:{perm:d}}),u=C.getInnerMostAxes(u.length,o)),C.assertAxesAreInnerMostDims("all",u,o);let[p,c]=C.computeOutAndReduceShapes(h.shape,u),f=v.sizeFromShape(c),m=ve({inputs:{x:h},backend:r,attrs:{shape:[-1,f]}}),g=Bl(m,m.dtype,"all",r),y;if(i){let A=C.expandShapeToKeepDim(p,l);y=ve({inputs:{x:g},backend:r,attrs:{shape:A}})}else y=ve({inputs:{x:g},backend:r,attrs:{shape:p}});return r.disposeIntermediateTensorInfo(m),r.disposeIntermediateTensorInfo(g),d!=null&&r.disposeIntermediateTensorInfo(h),y}var Tre={kernelName:Bu,backendName:"webgl",kernelFunc:Cre};function Nre(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{axis:s,keepDims:i}=n,o=a.shape.length,l=v.parseAxisParam(s,a.shape),u=l,d=C.getAxesPermutation(u,o),h=a;d!=null&&(h=dr({inputs:{x:a},backend:r,attrs:{perm:d}}),u=C.getInnerMostAxes(u.length,o)),C.assertAxesAreInnerMostDims("any",u,o);let[p,c]=C.computeOutAndReduceShapes(h.shape,u),f=v.sizeFromShape(c),m=ve({inputs:{x:h},backend:r,attrs:{shape:[-1,f]}}),g=Bl(m,m.dtype,"any",r),y;if(i){let A=C.expandShapeToKeepDim(p,l);y=ve({inputs:{x:g},backend:r,attrs:{shape:A}})}else y=ve({inputs:{x:g},backend:r,attrs:{shape:p}});return r.disposeIntermediateTensorInfo(m),r.disposeIntermediateTensorInfo(g),d!=null&&r.disposeIntermediateTensorInfo(h),y}var Ere={kernelName:Wu,backendName:"webgl",kernelFunc:Nre},Rre=class{constructor(e,t,r){this.variableNames=["A"];let{windowSize:n,batchSize:a,outSize:s}=e;r||this.variableNames.push("bestIndicesA"),this.outputShape=[a,s];let i=t==="max"?">":"<",o=r?"inOffset + i;":"round(getBestIndicesA(batch, inOffset + i));";this.userCode=`
|
|
void main() {
|
|
ivec2 coords = getOutputCoords();
|
|
int batch = coords[0];
|
|
int outIdx = coords[1];
|
|
int inOffset = outIdx * ${n};
|
|
|
|
int bestIndex = inOffset;
|
|
float bestValue = getA(batch, bestIndex);
|
|
|
|
for (int i = 0; i < ${n}; i++) {
|
|
int inIdx = ${o};
|
|
float candidate = getA(batch, inIdx);
|
|
if (candidate ${i} bestValue) {
|
|
bestValue = candidate;
|
|
bestIndex = inIdx;
|
|
}
|
|
}
|
|
setOutput(float(bestIndex));
|
|
}
|
|
`}},$re=class{constructor(e,t,r,n){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0,v.assert(e.length>2,()=>`Packed arg${r.charAt(0).toUpperCase()+r.slice(1)} supports only inputs with rank above 2.`);let a=e[e.length-1],s=Math.ceil(a/t);this.outputShape=e.slice(0,-1),s>1&&this.outputShape.push(s),n||this.variableNames.push("bestIndicesA");let i=this.outputShape,o=i.length,l=vt(o),u=Br("coords",o),d,h;if(s===1){h=o+1;let N=vt(h);d=`
|
|
${N} sourceLocR = ${N}(${u.join()}, 0);
|
|
++${u[o-1]};
|
|
${N} sourceLocG = ${N}(${u.join()}, 0);
|
|
++${u[o-2]};
|
|
${N} sourceLocA = ${N}(${u.join()}, 0);
|
|
--${u[o-1]};
|
|
${N} sourceLocB = ${N}(${u.join()}, 0);
|
|
--${u[o-2]};`}else h=o,d=`
|
|
${l} sourceLocR = coords;
|
|
++${u[o-1]};
|
|
${l} sourceLocG = coords;
|
|
++${u[o-2]};
|
|
${l} sourceLocA = coords;
|
|
--${u[o-1]};
|
|
${l} sourceLocB = coords;
|
|
--${u[o-2]};`;let p=["x","y","z","w","u","v"].slice(0,h),c="."+p[h-1],f=p.map(N=>"int "+N),m=Br("sourceLocR",h-1).concat("inIdx.r"),g=Br("sourceLocG",h-1).concat("inIdx.g"),y=Br("sourceLocB",h-1).concat("inIdx.b"),A=Br("sourceLocA",h-1).concat("inIdx.a"),x=r==="max"?"greaterThan":"lessThan",b=n?"":`
|
|
inIdx = round(vec4(getBestIndicesAChannel(${m.join()}),
|
|
getBestIndicesAChannel(${g.join()}),
|
|
getBestIndicesAChannel(${y.join()}),
|
|
getBestIndicesAChannel(${A.join()})));`,w=`vec4(
|
|
getAChannel(${m.join()}),
|
|
hasNextCol ? getAChannel(${g.join()}) : 0.,
|
|
hasNextRow ? getAChannel(${y.join()}) : 0.,
|
|
hasNextRow && hasNextCol ? getAChannel(${A.join()}) : 0.)`,S=n?"":`
|
|
float getBestIndicesAChannel(${f.join()}) {
|
|
return getChannel(getBestIndicesA(${p.join()}),
|
|
vec2(${p.slice(-2).join()}));
|
|
}`;this.userCode=`
|
|
float getAChannel(${f.join()}) {
|
|
return getChannel(getA(${p.join()}),
|
|
vec2(${p.slice(-2).join()}));
|
|
}
|
|
${S}
|
|
void main() {
|
|
${l} coords = getOutputCoords();
|
|
bool hasNextCol = ${u[o-1]} < ${i[o-1]-1};
|
|
bool hasNextRow = ${u[o-2]} < ${i[o-2]-1};
|
|
${d}
|
|
ivec4 srcIdx = ivec4(sourceLocR${c}, sourceLocG${c},
|
|
sourceLocB${c}, sourceLocA${c}) * ${t};
|
|
ivec4 inIdx = srcIdx;
|
|
vec4 bestIndex = vec4(inIdx);
|
|
vec4 bestValue = ${w};
|
|
|
|
for (int i = 0; i < ${t}; i++) {
|
|
inIdx = srcIdx;
|
|
${b}
|
|
vec4 candidate = ${w};
|
|
bvec4 nan = isnan(candidate);
|
|
bvec4 replace = bvec4(
|
|
vec4(${x}(candidate, bestValue)) * (vec4(1.0) - vec4(nan)));
|
|
|
|
bestValue = vec4(replace.x ? candidate.x : bestValue.x,
|
|
replace.y ? candidate.y : bestValue.y,
|
|
replace.z ? candidate.z : bestValue.z,
|
|
replace.w ? candidate.w : bestValue.w);
|
|
bestIndex = mix(bestIndex, vec4(inIdx), vec4(replace));
|
|
srcIdx++;
|
|
}
|
|
setOutput(bestIndex);
|
|
}
|
|
`}};function g9(e,t,r,n=null){let a=t.shape[0],s=t.shape[1];n!=null&&(a=n.shape[0],s=n.shape[1]);let i=C.computeOptimalWindowSize(s),o={windowSize:i,inSize:s,batchSize:a,outSize:Math.ceil(s/i)},l=new Rre(o,r,n==null),u=[t];n!=null&&u.push(n);let d=e.runWebGLProgram(l,u,"int32");if(d.shape[1]===1)return d;let h=g9(e,t,r,d);return e.disposeIntermediateTensorInfo(d),h}function y9(e,t,r,n=null){let a=n!=null?n.shape:t.shape,s=a[a.length-1],i=C.computeOptimalWindowSize(s),o=new $re(a,i,r,n==null),l=n==null?[t]:[t,n],u=e.runWebGLProgram(o,l,"int32");if(u.shape.length===t.shape.length){let d=y9(e,t,r,u);return e.disposeIntermediateTensorInfo(u),d}return u}function A9(e,t,r,n){let a=[r];if(C.assertAxesAreInnerMostDims("arg"+n.charAt(0).toUpperCase()+n.slice(1),a,t.shape.length),!Z().getBool("WEBGL_PACK_REDUCE")||t.shape.length<=2){let s=[],i=e.texData.get(t.dataId),o=i!==null&&i.isPacked,l=t;o&&(l=e.unpackTensor(t),s.push(l));let[u,d]=C.computeOutAndReduceShapes(l.shape,a),h=v.sizeFromShape(d),p=ve({inputs:{x:l},backend:e,attrs:{shape:[-1,h]}});s.push(p);let c=g9(e,p,n);s.push(c);let f=ve({inputs:{x:c},backend:e,attrs:{shape:u}});return s.forEach(m=>e.disposeIntermediateTensorInfo(m)),f}return y9(e,t,n)}function Mre(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{axis:s}=n,i=v.parseAxisParam(s,a.shape),o=C.getAxesPermutation(i,a.shape.length),l=a,u=[];o!=null&&(l=dr({inputs:{x:a},backend:r,attrs:{perm:o}}),u.push(l),i=C.getInnerMostAxes(i.length,l.shape.length)),C.assertAxesAreInnerMostDims("argMax",[i[0]],l.shape.length);let d=A9(r,l,i[0],"max");return u.forEach(h=>r.disposeIntermediateTensorInfo(h)),d}var Fre={kernelName:Ys,backendName:"webgl",kernelFunc:Mre};function Pre(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{axis:s}=n,i=v.parseAxisParam(s,a.shape),o=C.getAxesPermutation(i,a.shape.length),l=a,u=[];o!=null&&(l=dr({inputs:{x:a},backend:r,attrs:{perm:o}}),u.push(l),i=C.getInnerMostAxes(i.length,l.shape.length)),C.assertAxesAreInnerMostDims("argMin",[i[0]],l.shape.length);let d=A9(r,l,i[0],"min");return u.forEach(h=>r.disposeIntermediateTensorInfo(h)),d}var _re={kernelName:Vu,backendName:"webgl",kernelFunc:Pre},zre=Yn+`
|
|
if (abs(x) > 1.) {
|
|
return NAN;
|
|
}
|
|
return asin(x);
|
|
`,Ore=it({opSnippet:zre}),Dre={kernelName:Uu,backendName:"webgl",kernelFunc:Ore},Lre=Yn+"return log(x + sqrt(x * x + 1.0));",Bre=it({opSnippet:Lre}),Wre={kernelName:Gu,backendName:"webgl",kernelFunc:Bre},Vre=Yn+`
|
|
return atan(x);
|
|
`,Ure=it({opSnippet:Vre}),Gre={kernelName:ju,backendName:"webgl",kernelFunc:Ure},jre=Jte+`
|
|
return atan(a, b);
|
|
`,Hre=`
|
|
vec4 result = atan(a, b);
|
|
vec4 isNaN = min(vec4(isnan(a)) + vec4(isnan(b)), vec4(1.0));
|
|
`+Qte+`
|
|
return result;
|
|
`,qre=wr({opSnippet:jre,packedOpSnippet:Hre}),Xre={kernelName:qu,backendName:"webgl",kernelFunc:qre},Kre=Yn+`
|
|
if ((x < -1.0) || (x > 1.0)) return NAN;
|
|
return (log(1.0 + x) - log(1.0 - x)) / 2.0;`,Zre=it({opSnippet:Kre}),Yre={kernelName:Hu,backendName:"webgl",kernelFunc:Zre},Kp=class{constructor(e,t,r,n=!1,a=!1){if(this.variableNames=["x"],t==="avg"&&r)throw new Error("Cannot compute positions for average pool.");let s=e.filterWidth,i=e.strideHeight,o=e.strideWidth,l=e.dilationHeight,u=e.dilationWidth,d=e.effectiveFilterHeight,h=e.effectiveFilterWidth,p=e.padInfo.top,c=e.padInfo.left;this.outputShape=e.outShape;let f=t==="avg",m=`((batch * ${e.inHeight} + xR) * ${e.inWidth} + xC) * ${e.inChannels} + d`,g=`(xR * ${e.inWidth} + xC) * ${e.inChannels} + d`,y="0.0";if(f||(y="-1.0 / 1e-20"),r){let N=">=";this.userCode=`
|
|
const ivec2 strides = ivec2(${i}, ${o});
|
|
const ivec2 pads = ivec2(${p}, ${c});
|
|
|
|
void main() {
|
|
ivec4 coords = getOutputCoords();
|
|
int batch = coords[0];
|
|
int d = coords[3];
|
|
|
|
ivec2 xRCCorner = coords.yz * strides - pads;
|
|
int xRCorner = xRCCorner.x;
|
|
int xCCorner = xRCCorner.y;
|
|
|
|
// max/min x(?, ?, d) to get y(yR, yC, d).
|
|
// ? = to be determined
|
|
float minMaxValue = 0.0;
|
|
float minMaxValueFound = 0.0;
|
|
int minMaxPosition = 0;
|
|
float avgValue = 0.0;
|
|
|
|
for (int wR = 0; wR < ${d};
|
|
wR += ${l}) {
|
|
int xR = xRCorner + wR;
|
|
|
|
if (xR < 0 || xR >= ${e.inHeight}) {
|
|
continue;
|
|
}
|
|
|
|
for (int wC = 0; wC < ${h};
|
|
wC += ${u}) {
|
|
int xC = xCCorner + wC;
|
|
|
|
if (xC < 0 || xC >= ${e.inWidth}) {
|
|
continue;
|
|
}
|
|
|
|
float value = getX(batch, xR, xC, d);
|
|
|
|
// If a min / max value has already been found, use it. If not,
|
|
// use the current value.
|
|
float currMinMaxValue = mix(
|
|
value, minMaxValue, minMaxValueFound);
|
|
if (value ${N} currMinMaxValue) {
|
|
minMaxValue = value;
|
|
minMaxValueFound = 1.0;
|
|
minMaxPosition = ${n?a?m:g:`wR * ${h} + wC`};
|
|
}
|
|
}
|
|
}
|
|
setOutput(float(minMaxPosition));
|
|
}
|
|
`;return}let A="max",x=`${t}(${t}(${t}(minMaxValue[0], minMaxValue[1]), minMaxValue[2]), minMaxValue[3])`;t==="avg"&&(x="avgValue / count");let b=Math.floor(s/4)*4,w=s%4,S=`
|
|
if (${f}) {
|
|
avgValue += dot(values, ones);
|
|
} else {
|
|
minMaxValue = ${A}(values, minMaxValue);
|
|
}
|
|
`;this.userCode=`
|
|
const ivec2 strides = ivec2(${i}, ${o});
|
|
const ivec2 pads = ivec2(${p}, ${c});
|
|
const float initializationValue = ${y};
|
|
const vec4 ones = vec4(1.0, 1.0, 1.0, 1.0);
|
|
|
|
float count = 0.0;
|
|
|
|
float getValue(int batch, int xR, int xC, int d) {
|
|
if (xC < 0 || xC >= ${e.inWidth}) {
|
|
return initializationValue;
|
|
}
|
|
count += 1.0;
|
|
return getX(batch, xR, xC, d);
|
|
}
|
|
|
|
void main() {
|
|
ivec4 coords = getOutputCoords();
|
|
int batch = coords[0];
|
|
int d = coords[3];
|
|
|
|
ivec2 xRCCorner = coords.yz * strides - pads;
|
|
int xRCorner = xRCCorner.x;
|
|
int xCCorner = xRCCorner.y;
|
|
|
|
// max/min x(?, ?, d) to get y(yR, yC, d).
|
|
// ? = to be determined
|
|
vec4 minMaxValue = vec4(${y});
|
|
float avgValue = 0.0;
|
|
count = 0.0;
|
|
|
|
for (int wR = 0; wR < ${d};
|
|
wR += ${l}) {
|
|
int xR = xRCorner + wR;
|
|
|
|
if (xR < 0 || xR >= ${e.inHeight}) {
|
|
continue;
|
|
}
|
|
|
|
for (int wC = 0; wC < ${b}; wC += 4) {
|
|
int xC = xCCorner + wC * ${u};
|
|
|
|
vec4 values = vec4(
|
|
getValue(batch, xR, xC, d),
|
|
getValue(batch, xR, xC + ${u}, d),
|
|
getValue(batch, xR, xC + 2 * ${u}, d),
|
|
getValue(batch, xR, xC + 3 * ${u}, d)
|
|
);
|
|
|
|
${S}
|
|
}
|
|
|
|
int xC = xCCorner + ${b};
|
|
if (${w===1}) {
|
|
vec4 values = vec4(
|
|
getValue(batch, xR, xC, d),
|
|
initializationValue,
|
|
initializationValue,
|
|
initializationValue
|
|
);
|
|
|
|
${S}
|
|
} else if (${w===2}) {
|
|
vec4 values = vec4(
|
|
getValue(batch, xR, xC, d),
|
|
getValue(batch, xR, xC + ${u}, d),
|
|
initializationValue,
|
|
initializationValue
|
|
);
|
|
|
|
${S}
|
|
} else if (${w===3}) {
|
|
vec4 values = vec4(
|
|
getValue(batch, xR, xC, d),
|
|
getValue(batch, xR, xC + ${u}, d),
|
|
getValue(batch, xR, xC + 2 * ${u}, d),
|
|
initializationValue
|
|
);
|
|
|
|
${S}
|
|
}
|
|
}
|
|
setOutput(${x});
|
|
}
|
|
`}},I5=class{constructor(e,t,r,n=!1,a=!1){if(this.variableNames=["x"],t==="avg"&&r)throw new Error("Cannot compute positions for average pool.");let s=e.filterWidth,i=e.strideDepth,o=e.strideHeight,l=e.strideWidth,u=e.dilationDepth,d=e.dilationHeight,h=e.dilationWidth,p=e.effectiveFilterDepth,c=e.effectiveFilterHeight,f=e.effectiveFilterWidth,m=e.padInfo.front,g=e.padInfo.top,y=e.padInfo.left;this.outputShape=e.outShape;let A=t==="avg",x="0.0";if(A||(x="-1.0 / 1e-20"),r){let R=">=";this.userCode=`
|
|
const ivec3 strides =
|
|
ivec3(${i}, ${o}, ${l});
|
|
const ivec3 pads = ivec3(${m}, ${g}, ${y});
|
|
|
|
void main() {
|
|
ivec5 coords = getOutputCoords();
|
|
int batch = coords.x;
|
|
int ch = coords.u;
|
|
|
|
ivec3 xCorner = ivec3(coords.y, coords.z, coords.w) * strides - pads;
|
|
int xDCorner = xCorner.x;
|
|
int xRCorner = xCorner.y;
|
|
int xCCorner = xCorner.z;
|
|
|
|
// max/min x(?, ?, ?, ch) to get y(yD, yR, yC, ch).
|
|
// ? = to be determined
|
|
float minMaxValue = 0.0;
|
|
float minMaxValueFound = 0.0;
|
|
int minMaxPosition = 0;
|
|
|
|
for (int wD = 0; wD < ${p};
|
|
wD += ${u}) {
|
|
int xD = xDCorner + wD;
|
|
|
|
if (xD < 0 || xD >= ${e.inDepth}) {
|
|
continue;
|
|
}
|
|
|
|
for (int wR = 0; wR < ${c};
|
|
wR += ${d}) {
|
|
int xR = xRCorner + wR;
|
|
|
|
if (xR < 0 || xR >= ${e.inHeight}) {
|
|
continue;
|
|
}
|
|
|
|
for (int wC = 0; wC < ${f};
|
|
wC += ${h}) {
|
|
int xC = xCCorner + wC;
|
|
|
|
if (xC < 0 || xC >= ${e.inWidth}) {
|
|
continue;
|
|
}
|
|
|
|
float value = getX(batch, xD, xR, xC, ch);
|
|
|
|
// If a min / max value has already been found, use it. If not,
|
|
// use the current value.
|
|
float currMinMaxValue = mix(
|
|
value, minMaxValue, minMaxValueFound);
|
|
if (value ${R} currMinMaxValue) {
|
|
minMaxValue = value;
|
|
minMaxValueFound = 1.0;
|
|
minMaxPosition = ${n?a?`(((batch * ${e.inDepth} + xD) * ${e.inHeight} + xR) * ${e.inWidth} + xC) * ${e.inChannels} + ch`:`((xD * ${e.inHeight} + xR) * ${e.inWidth} + xC) * ${e.inChannels} + ch`:`wD * ${c} * ${f} +
|
|
wR * ${f} + wC`};
|
|
}
|
|
}
|
|
}
|
|
}
|
|
setOutput(float(minMaxPosition));
|
|
}
|
|
`;return}let b="max",w=`${t}(${t}(${t}(minMaxValue[0], minMaxValue[1]), minMaxValue[2]), minMaxValue[3])`;t==="avg"&&(w="avgValue / count");let S=Math.floor(s/4)*4,N=s%4,E=`
|
|
if (${A}) {
|
|
avgValue += dot(values, ones);
|
|
} else {
|
|
minMaxValue = ${b}(values, minMaxValue);
|
|
}
|
|
`;this.userCode=`
|
|
const ivec3 strides =
|
|
ivec3(${i}, ${o}, ${l});
|
|
const ivec3 pads = ivec3(${m}, ${g}, ${y});
|
|
const float initializationValue = ${x};
|
|
const vec4 ones = vec4(1.0, 1.0, 1.0, 1.0);
|
|
|
|
float count = 0.0;
|
|
|
|
float getValue(int batch, int xD, int xR, int xC, int ch) {
|
|
if (xC < 0 || xC >= ${e.inWidth}) {
|
|
return initializationValue;
|
|
}
|
|
count += 1.0;
|
|
return getX(batch, xD, xR, xC, ch);
|
|
}
|
|
|
|
void main() {
|
|
ivec5 coords = getOutputCoords();
|
|
int batch = coords.x;
|
|
int ch = coords.u;
|
|
|
|
ivec3 xCorner = ivec3(coords.y, coords.z, coords.w) * strides - pads;
|
|
int xDCorner = xCorner.x;
|
|
int xRCorner = xCorner.y;
|
|
int xCCorner = xCorner.z;
|
|
|
|
// max/min x(?, ?, ?, d) to get y(yD, yR, yC, ch).
|
|
// ? = to be determined
|
|
vec4 minMaxValue = vec4(${x});
|
|
float avgValue = 0.0;
|
|
count = 0.0;
|
|
|
|
for (int wD = 0; wD < ${p};
|
|
wD += ${u}) {
|
|
int xD = xDCorner + wD;
|
|
|
|
if (xD < 0 || xD >= ${e.inDepth}) {
|
|
continue;
|
|
}
|
|
|
|
for (int wR = 0; wR < ${c};
|
|
wR += ${d}) {
|
|
int xR = xRCorner + wR;
|
|
|
|
if (xR < 0 || xR >= ${e.inHeight}) {
|
|
continue;
|
|
}
|
|
|
|
for (int wC = 0; wC < ${S}; wC += 4) {
|
|
int xC = xCCorner + wC * ${h};
|
|
|
|
vec4 values = vec4(
|
|
getValue(batch, xD, xR, xC, ch),
|
|
getValue(batch, xD, xR, xC + ${h}, ch),
|
|
getValue(batch, xD, xR, xC + 2 * ${h}, ch),
|
|
getValue(batch, xD, xR, xC + 3 * ${h}, ch)
|
|
);
|
|
|
|
${E}
|
|
}
|
|
|
|
int xC = xCCorner + ${S};
|
|
if (${N===1}) {
|
|
vec4 values = vec4(
|
|
getValue(batch, xD, xR, xC, ch),
|
|
initializationValue,
|
|
initializationValue,
|
|
initializationValue
|
|
);
|
|
|
|
${E}
|
|
} else if (${N===2}) {
|
|
vec4 values = vec4(
|
|
getValue(batch, xD, xR, xC, ch),
|
|
getValue(batch, xD, xR, xC + ${h}, ch),
|
|
initializationValue,
|
|
initializationValue
|
|
);
|
|
|
|
${E}
|
|
} else if (${N===3}) {
|
|
vec4 values = vec4(
|
|
getValue(batch, xD, xR, xC, ch),
|
|
getValue(batch, xD, xR, xC + ${h}, ch),
|
|
getValue(batch, xD, xR, xC + 2 * ${h}, ch),
|
|
initializationValue
|
|
);
|
|
|
|
${E}
|
|
}
|
|
}
|
|
setOutput(${w});
|
|
}
|
|
}
|
|
`}};function Jre(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t;Id(a,"avgPool");let{filterSize:s,strides:i,pad:o,dimRoundingMode:l}=n,u=1;v.assert(C.eitherStridesOrDilationsAreOne(i,u),()=>`Error in avgPool: Either strides or dilations must be 1. Got strides ${i} and dilations '${u}'`);let d=C.computePool2DInfo(a.shape,s,i,u,o,l);if(d.filterWidth===1&&d.filterHeight===1&&v.arraysEqual(d.inShape,d.outShape))return fn({inputs:{x:a},backend:r});let h=new Kp(d,"avg",!1);return r.runWebGLProgram(h,[a],"float32")}var Qre={kernelName:Js,backendName:"webgl",kernelFunc:Jre};function ene(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{filterSize:s,strides:i,pad:o,dimRoundingMode:l,dataFormat:u}=n,d=[1,1,1],h=C.computePool3DInfo(a.shape,s,i,d,o,l,u),p=new I5(h,"avg",!1);return r.runWebGLProgram(p,[a],"float32")}var tne={kernelName:Yp,backendName:"webgl",kernelFunc:ene},rne=class{constructor(e){this.variableNames=["dy"],this.outputShape=e.inShape;let t=e.filterHeight,r=e.filterWidth,n=e.strideHeight,a=e.strideWidth,s=e.dilationHeight,i=e.dilationWidth,o=e.effectiveFilterHeight,l=e.effectiveFilterWidth,u=o-1-e.padInfo.top,d=l-1-e.padInfo.left,h=1/(t*r);this.userCode=`
|
|
const ivec2 pads = ivec2(${u}, ${d});
|
|
const float avgMultiplier = float(${h});
|
|
|
|
void main() {
|
|
ivec4 coords = getOutputCoords();
|
|
int b = coords[0];
|
|
int d = coords[3];
|
|
|
|
ivec2 dyRCCorner = coords.yz - pads;
|
|
int dyRCorner = dyRCCorner.x;
|
|
int dyCCorner = dyRCCorner.y;
|
|
|
|
// Convolve dy(?, ?, d) with pos mask(:, :, d) to get dx(xR, xC, d).
|
|
// ? = to be determined. : = across all values in that axis.
|
|
float dotProd = 0.0;
|
|
for (int wR = 0; wR < ${o};
|
|
wR += ${s}) {
|
|
float dyR = float(dyRCorner + wR) / ${n}.0;
|
|
|
|
if (dyR < 0.0 || dyR >= ${e.outHeight}.0 || fract(dyR) > 0.0) {
|
|
continue;
|
|
}
|
|
int idyR = int(dyR);
|
|
|
|
for (int wC = 0; wC < ${l};
|
|
wC+= ${i}) {
|
|
float dyC = float(dyCCorner + wC) / ${a}.0;
|
|
|
|
if (dyC < 0.0 || dyC >= ${e.outWidth}.0 ||
|
|
fract(dyC) > 0.0) {
|
|
continue;
|
|
}
|
|
int idyC = int(dyC);
|
|
|
|
float dyValue = getDy(b, idyR, idyC, d);
|
|
|
|
dotProd += dyValue * avgMultiplier;
|
|
}
|
|
}
|
|
setOutput(dotProd);
|
|
}
|
|
`}},nne=class{constructor(e){this.variableNames=["dy"],this.outputShape=e.inShape;let t=e.filterDepth,r=e.filterHeight,n=e.filterWidth,a=e.strideDepth,s=e.strideHeight,i=e.strideWidth,o=e.dilationDepth,l=e.dilationHeight,u=e.dilationWidth,d=e.effectiveFilterDepth,h=e.effectiveFilterHeight,p=e.effectiveFilterWidth,c=d-1-e.padInfo.front,f=h-1-e.padInfo.top,m=p-1-e.padInfo.left,g=1/(t*r*n);this.userCode=`
|
|
const ivec3 pads = ivec3(${c}, ${f}, ${m});
|
|
const float avgMultiplier = float(${g});
|
|
|
|
void main() {
|
|
ivec5 coords = getOutputCoords();
|
|
int batch = coords.x;
|
|
int ch = coords.u;
|
|
|
|
ivec3 dyCorner = ivec3(coords.y, coords.z, coords.w) - pads;
|
|
int dyDCorner = dyCorner.x;
|
|
int dyRCorner = dyCorner.y;
|
|
int dyCCorner = dyCorner.z;
|
|
|
|
// Convolve dy(?, ?, ?, d) with pos mask(:, :, :, ch) to get
|
|
// dx(xD, xR, xC, ch).
|
|
// ? = to be determined. : = across all values in that axis.
|
|
float dotProd = 0.0;
|
|
|
|
for (int wD = 0; wD < ${d};
|
|
wD += ${o}) {
|
|
float dyD = float(dyDCorner + wD) / ${a}.0;
|
|
|
|
if (dyD < 0.0 || dyD >= ${e.outDepth}.0 || fract(dyD) > 0.0) {
|
|
continue;
|
|
}
|
|
int idyD = int(dyD);
|
|
|
|
for (int wR = 0; wR < ${h};
|
|
wR += ${l}) {
|
|
float dyR = float(dyRCorner + wR) / ${s}.0;
|
|
|
|
if (dyR < 0.0 || dyR >= ${e.outHeight}.0 ||
|
|
fract(dyR) > 0.0) {
|
|
continue;
|
|
}
|
|
int idyR = int(dyR);
|
|
|
|
for (int wC = 0; wC < ${p};
|
|
wC += ${u}) {
|
|
float dyC = float(dyCCorner + wC) / ${i}.0;
|
|
|
|
if (dyC < 0.0 || dyC >= ${e.outWidth}.0 ||
|
|
fract(dyC) > 0.0) {
|
|
continue;
|
|
}
|
|
int idyC = int(dyC);
|
|
|
|
float dyValue = getDy(batch, idyD, idyR, idyC, ch);
|
|
|
|
dotProd += dyValue * avgMultiplier;
|
|
}
|
|
}
|
|
}
|
|
setOutput(dotProd);
|
|
}
|
|
`}};function ane(e){let{inputs:t,backend:r,attrs:n}=e,{dy:a,input:s}=t,i=s,{filterSize:o,strides:l,pad:u,dimRoundingMode:d}=n,h=[1,1,1],p=C.computePool3DInfo(i.shape,o,l,h,u,d),c=new nne(p);return r.runWebGLProgram(c,[a],i.dtype)}var sne={kernelName:Q0,backendName:"webgl",kernelFunc:ane};function ine(e){let{inputs:t,backend:r,attrs:n}=e,{dy:a,input:s}=t,i=s;Id([a,s],"avgPoolGrad");let{filterSize:o,strides:l,pad:u}=n,d=C.computePool2DInfo(i.shape,o,l,1,u),h=new rne(d);return r.runWebGLProgram(h,[a],i.dtype)}var one={kernelName:J0,backendName:"webgl",kernelFunc:ine};function lne(e){let{inputs:t,backend:r,attrs:n}=e,{a,b:s}=t,{transposeA:i,transposeB:o}=n;return G0({a,b:s,transposeA:i,transposeB:o,backend:r})}var une={kernelName:Qs,backendName:"webgl",kernelFunc:lne},dne=class{constructor(e,t,r,n,a,s){this.outputShape=[],this.variableNames=["x","mean","variance"],C.assertAndGetBroadcastShape(e,t),C.assertAndGetBroadcastShape(e,r);let i="0.0";n!=null&&(C.assertAndGetBroadcastShape(e,n),this.variableNames.push("offset"),i="getOffsetAtOutCoords()");let o="1.0";a!=null&&(C.assertAndGetBroadcastShape(e,a),this.variableNames.push("scale"),o="getScaleAtOutCoords()"),this.outputShape=e,this.userCode=`
|
|
void main() {
|
|
float x = getXAtOutCoords();
|
|
float mean = getMeanAtOutCoords();
|
|
float variance = getVarianceAtOutCoords();
|
|
float offset = ${i};
|
|
float scale = ${o};
|
|
float inv = scale * inversesqrt(variance + float(${s}));
|
|
setOutput(dot(vec3(x, -mean, offset), vec3(inv, inv, 1)));
|
|
}
|
|
`}},pne=class{constructor(e,t,r,n,a,s){this.packedInputs=!0,this.packedOutput=!0,this.variableNames=["x","mean","variance"],C.assertAndGetBroadcastShape(e,t),C.assertAndGetBroadcastShape(e,r);let i="vec4(0.0)";n!=null&&(C.assertAndGetBroadcastShape(e,n),this.variableNames.push("offset"),i="getOffsetAtOutCoords()");let o="vec4(1.0)";a!=null&&(C.assertAndGetBroadcastShape(e,a),this.variableNames.push("scale"),o="getScaleAtOutCoords()"),this.outputShape=e,this.userCode=`
|
|
void main() {
|
|
vec4 offset = ${i};
|
|
vec4 scale = ${o};
|
|
|
|
vec4 x = getXAtOutCoords();
|
|
vec4 mean = getMeanAtOutCoords();
|
|
vec4 variance = getVarianceAtOutCoords();
|
|
|
|
vec4 inv = scale * inversesqrt(variance + vec4(${s}));
|
|
|
|
setOutput((x - mean) * inv + offset);
|
|
}
|
|
`}},hne=({inputs:e,backend:t,attrs:r})=>{let{x:n,mean:a,variance:s,offset:i,scale:o}=e;v.assert(a.shape.length===s.shape.length,()=>"Batch normalization gradient requires mean and variance to have equal ranks."),v.assert(i==null||a.shape.length===i.shape.length,()=>"Batch normalization gradient requires mean and offset to have equal ranks."),v.assert(o==null||a.shape.length===o.shape.length,()=>"Batch normalization gradient requires mean and scale to have equal ranks.");let{varianceEpsilon:l}=r;l==null&&(l=.001);let u=[n,a,s],d=null;i!=null&&(d=i.shape,u.push(i));let h=null;o!=null&&(h=o.shape,u.push(o));let p=Z().getBool("WEBGL_PACK_NORMALIZATION")?new pne(n.shape,a.shape,s.shape,d,h,l):new dne(n.shape,a.shape,s.shape,d,h,l);return t.runWebGLProgram(p,u,u[0].dtype)},cne={kernelName:ci,backendName:"webgl",kernelFunc:hne},fne=class{constructor(e){this.variableNames=["source"],this.outputShape=e,this.rank=e.length;let t=vt(this.rank);this.customUniforms=[{name:"start",arrayIndex:this.rank,type:"int"}];let r=mne(this.rank),n,a=e.map((s,i)=>`sourceLoc.${ay[i]} = start[${i}] + coords.${ay[i]};`);n=`
|
|
${t} sourceLoc;
|
|
${t} coords = getOutputCoords();
|
|
${a.join(`
|
|
`)}
|
|
`,this.userCode=`
|
|
void main() {
|
|
${n}
|
|
setOutput(getSource(${r}));
|
|
}
|
|
`}},ay=["x","y","z","w","u","v"];function mne(e){if(e===1)return"sourceLoc";if(e<=6)return ay.slice(0,e).map(t=>"sourceLoc."+t).join(",");throw Error(`Slicing for rank ${e} is not yet supported`)}var gne=class{constructor(e){this.variableNames=["source"],this.packedInputs=!0,this.packedOutput=!0,this.outputShape=e,this.rank=e.length,this.customUniforms=[{name:"start",arrayIndex:this.rank,type:"int"}];let t=vt(this.rank),r=Br("coords",this.rank),n=Br("sourceLoc",this.rank),a=this.rank===1?"sourceLoc":`vec2(${n.slice(-2).join()})`,s=`getChannel(getSource(${n.join()}), ${a})`,i=`
|
|
result.x = ${s};
|
|
if (++${r[this.rank-1]} < ${e[this.rank-1]}) {
|
|
++${n[this.rank-1]};
|
|
result.y = ${s};
|
|
--${n[this.rank-1]};
|
|
}
|
|
`,o=this.rank===1?"":`
|
|
--${r[this.rank-1]};
|
|
if (++${r[this.rank-2]} < ${e[this.rank-2]}) {
|
|
++${n[this.rank-2]};
|
|
result.z = ${s};
|
|
if (++${r[this.rank-1]} < ${e[this.rank-1]}) {
|
|
++${n[this.rank-1]};
|
|
result.w = ${s};
|
|
}
|
|
}
|
|
`,l=this.rank<=4?`sourceLoc = coords +
|
|
${t}(${e.map((u,d)=>`start[${d}]`).join()});`:e.map((u,d)=>`${n[d]} = ${r[d]} + start[${d}];`).join(`
|
|
`);this.userCode=`
|
|
void main() {
|
|
${t} coords = getOutputCoords();
|
|
${t} sourceLoc;
|
|
${l}
|
|
vec4 result = vec4(0.);
|
|
${i}
|
|
${o}
|
|
setOutput(result);
|
|
}
|
|
`}};function yne(e,t,r,n){let a=n.texData.get(e.dataId),s=n.makeTensorInfo(r,e.dtype),i=n.texData.get(s.dataId);Object.assign(i,a),i.refCount=1,i.shape=r,i.dtype=e.dtype;let o=Ot.computeFlatOffset(t,v.computeStrides(e.shape));a.slice&&(o+=a.slice.flatOffset),i.slice={flatOffset:o,origDataId:a.slice&&a.slice.origDataId||e.dataId};let l=n.dataRefCount.get(i.slice.origDataId)||1;return n.dataRefCount.set(i.slice.origDataId,l+1),s}function Rd(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{begin:s,size:i}=n,[o,l]=Ot.parseSliceParams(a,s,i);if(Ot.assertParamsValid(a,o,l),v.sizeFromShape(l)===0)return r.makeTensorInfo(l,a.dtype,[]);if(r.shouldExecuteOnCPU([a])||a.dtype==="string"){let h=r.texData.get(a.dataId),p=ite(h.values,o,l,a.shape,a.dtype);return r.makeTensorInfo(l,a.dtype,p)}let{isPacked:u}=r.texData.get(a.dataId),d=Ot.isSliceContinous(a.shape,o,l);if(u||!d){let h=Z().getBool("WEBGL_PACK_ARRAY_OPERATIONS")?new gne(l):new fne(l),p=[o];return r.runWebGLProgram(h,[a],a.dtype,p)}return r.uploadToGPU(a.dataId),yne(a,o,l,r)}var Ane={kernelName:bl,backendName:"webgl",kernelFunc:Rd},xne=e=>{let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{blockShape:s,crops:i}=n;v.assert(a.shape.length<=4,()=>"batchToSpaceND for rank > 4 with a WebGL backend not implemented yet");let o=s.reduce((A,x)=>A*x),l=C.getReshaped(a.shape,s,o),u=C.getPermuted(l.length,s.length),d=C.getReshapedPermuted(a.shape,s,o),h=C.getSliceBeginCoords(i,s.length),p=C.getSliceSize(d,i,s.length),c=[],f=ve({inputs:{x:a},backend:r,attrs:{shape:l}}),m=dr({inputs:{x:f},backend:r,attrs:{perm:u}}),g=ve({inputs:{x:m},backend:r,attrs:{shape:d}}),y=Rd({inputs:{x:g},backend:r,attrs:{begin:h,size:p}});return c.push(f),c.push(m),c.push(g),c.forEach(A=>r.disposeIntermediateTensorInfo(A)),y},bne={kernelName:qo,backendName:"webgl",kernelFunc:xne};function vne(e){let{inputs:t,backend:r,attrs:n}=e,{x:a,weights:s}=t,{size:i}=n,o=r.readSync(a.dataId),l=r.readSync(s.dataId),u=r9(o,l,s.dtype,s.shape,i);return r.makeTensorInfo([i],s.dtype,u)}var wne={kernelName:ef,backendName:"webgl",kernelFunc:vne};function kne(e){let{inputs:t,backend:r}=e,{s0:n,s1:a}=t,s=r.readSync(n.dataId),i=r.readSync(a.dataId),o=C.assertAndGetBroadcastShape(Array.from(s),Array.from(i));return r.makeTensorInfo([o.length],"int32",Int32Array.from(o))}var Ine={kernelName:tf,backendName:"webgl",kernelFunc:kne},Sne="return float(a != b);",x9=wr({opSnippet:Sne,cpuKernelImpl:tte,dtype:"bool"}),Cne={kernelName:ul,backendName:"webgl",kernelFunc:x9};function Vh(e){let{inputs:t,backend:r}=e,{input:n}=t,a=r.texData.get(n.dataId);return fn({inputs:{x:a.complexTensorInfos.real},backend:r})}var Tne={kernelName:oh,backendName:"webgl",kernelFunc:Vh},Nne="return float(int(x));";function Ene(e,t){let r=new Xa(e.shape,Nne),n=t.runWebGLProgram(r,[e],"int32");return{dataId:n.dataId,shape:n.shape,dtype:n.dtype}}function sy(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{dtype:s}=n;if(s==="complex64"){if(a.dtype==="complex64")return fn({inputs:{x:a},backend:r});let i=_t(a.shape),o=sy({inputs:{x:a},backend:r,attrs:{dtype:"float32"}}),l=Xi({inputs:{real:o,imag:i},backend:r});return i.dispose(),r.disposeIntermediateTensorInfo(o),l}if(a.dtype==="complex64"){let i=Vh({inputs:{input:a},backend:r}),o=sy({inputs:{x:i},backend:r,attrs:{dtype:s}});return r.disposeIntermediateTensorInfo(i),o}if(!v.hasEncodingLoss(a.dtype,s)){let i=fn({inputs:{x:a},backend:r});return{dataId:i.dataId,shape:i.shape,dtype:s}}if(s==="int32")return Ene(a,r);if(s==="bool"){let i=r.makeTensorInfo([],"bool",v.getTypedArrayFromDType("bool",1)),o=x9({inputs:{a,b:i},backend:r});return r.disposeIntermediateTensorInfo(i),o}throw new Error(`Error in Cast: failed to cast ${a.dtype} to ${s}`)}var Rne={kernelName:ei,backendName:"webgl",kernelFunc:sy},X4="return ceil(x);",$ne=it({opSnippet:X4,packedOpSnippet:X4,cpuKernelImpl:zee}),Mne={kernelName:ti,backendName:"webgl",kernelFunc:$ne},Fne=class{constructor(e){this.variableNames=["A"],this.customUniforms=[{name:"minVal",type:"float"},{name:"maxVal",type:"float"}],this.outputShape=e,this.userCode=`
|
|
|
|
void main() {
|
|
float value = getAAtOutCoords();
|
|
if (isnan(value)) {
|
|
setOutput(value);
|
|
return;
|
|
}
|
|
|
|
setOutput(clamp(value, minVal, maxVal));
|
|
}
|
|
`}},Pne=class{constructor(e){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0,this.customUniforms=[{name:"minVal",type:"float"},{name:"maxVal",type:"float"}],this.outputShape=e,this.userCode=`
|
|
void main() {
|
|
vec4 value = getAAtOutCoords();
|
|
|
|
if (any(isnan(value))) {
|
|
setOutput(value);
|
|
return;
|
|
}
|
|
|
|
setOutput(clamp(value, vec4(minVal), vec4(maxVal)));
|
|
}
|
|
`}};function _ne(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{clipValueMin:s,clipValueMax:i}=n,o;Z().getBool("WEBGL_PACK_CLIP")?o=new Pne(a.shape):o=new Fne(a.shape);let l=[[s],[i]];return r.runWebGLProgram(o,[a],a.dtype,l)}var zne={kernelName:Qa,backendName:"webgl",kernelFunc:_ne},One=class{constructor(e){this.variableNames=["real","imag"],this.outputShape=e,this.userCode=`
|
|
void main() {
|
|
float re = abs(getRealAtOutCoords());
|
|
float im = abs(getImagAtOutCoords());
|
|
float mx = max(re, im);
|
|
|
|
// sadly the length function in glsl is not underflow-safe
|
|
// (at least not on Intel GPUs). So the safe solution is
|
|
// to ensure underflow-safety in all cases.
|
|
setOutput(
|
|
mx == 0.0 ? 0.0 : mx * length(vec2(1, min(re, im)/mx))
|
|
);
|
|
}
|
|
`}};function K4(e,t){return{dataId:t.dataId,dtype:t.dtype,shape:e.shape}}function Dne(e){let{inputs:t,backend:r}=e,{x:n}=t,a=r.texData.get(n.dataId),s=new One(n.shape),i=[K4(n,a.complexTensorInfos.real),K4(n,a.complexTensorInfos.imag)];return r.runWebGLProgram(s,i,i[0].dtype)}var Lne={kernelName:Qp,backendName:"webgl",kernelFunc:Dne},Bne=class{constructor(e){this.outputShape=[],this.outputShape=C.computeOutShape(e,1),this.variableNames=e.map((s,i)=>`T${i}`);let t=new Array(e.length-1);t[0]=e[0][1];for(let s=1;s<t.length;s++)t[s]=t[s-1]+e[s][1];let r=[`if (yC < ${t[0]}) setOutput(getT0(yR, yC));`];for(let s=1;s<t.length;s++){let i=t[s-1];r.push(`else if (yC < ${t[s]}) setOutput(getT${s}(yR, yC-${i}));`)}let n=t.length,a=t[t.length-1];r.push(`else setOutput(getT${n}(yR, yC-${a}));`),this.userCode=`
|
|
void main() {
|
|
ivec2 coords = getOutputCoords();
|
|
int yR = coords.x;
|
|
int yC = coords.y;
|
|
|
|
${r.join(`
|
|
`)}
|
|
}
|
|
`}},Wne=class{constructor(e,t){this.packedInputs=!0,this.packedOutput=!0,this.outputShape=[],this.outputShape=C.computeOutShape(e,t);let r=this.outputShape,n=r.length,a=vt(n),s=Br("coords",n),i=["x","y","z","w","u","v"].slice(0,n);this.variableNames=e.map((f,m)=>`T${m}`);let o=new Array(e.length-1);o[0]=e[0][t];for(let f=1;f<o.length;f++)o[f]=o[f-1]+e[f][t];let l=i[t],u=i.slice(-2),d=i.join(),h=`if (${l} < ${o[0]}) {
|
|
return getChannel(
|
|
getT0(${d}), vec2(${u.join()}));
|
|
}`;for(let f=1;f<o.length;f++){let m=o[f-1];h+=`
|
|
if (${l} < ${o[f]} && ${l} >= ${o[f-1]}) {
|
|
return getChannel(
|
|
getT${f}(${n0(i,l,m)}),
|
|
vec2(${n0(u,l,m)}));
|
|
}`}let p=o.length,c=o[o.length-1];h+=`
|
|
return getChannel(
|
|
getT${p}(${n0(i,l,c)}),
|
|
vec2(${n0(u,l,c)}));`,this.userCode=`
|
|
float getValue(${i.map(f=>"int "+f)}) {
|
|
${h}
|
|
}
|
|
|
|
void main() {
|
|
${a} coords = getOutputCoords();
|
|
vec4 result = vec4(getValue(${s}), 0., 0., 0.);
|
|
|
|
${s[n-1]} = ${s[n-1]} + 1;
|
|
if (${s[n-1]} < ${r[n-1]}) {
|
|
result.g = getValue(${s});
|
|
}
|
|
|
|
${s[n-2]} = ${s[n-2]} + 1;
|
|
if (${s[n-2]} < ${r[n-2]}) {
|
|
result.a = getValue(${s});
|
|
}
|
|
|
|
${s[n-1]} = ${s[n-1]} - 1;
|
|
if (${s[n-2]} < ${r[n-2]} &&
|
|
${s[n-1]} < ${r[n-1]}) {
|
|
result.b = getValue(${s});
|
|
}
|
|
setOutput(result);
|
|
}
|
|
`}};function n0(e,t,r){let n=e.indexOf(t);return e.map((a,s)=>s===n?`${a} - ${r}`:a).join()}function Rm(e){let{inputs:t,backend:r}=e,{input:n}=t,a=r.texData.get(n.dataId);return fn({inputs:{x:a.complexTensorInfos.imag},backend:r})}var Vne={kernelName:nh,backendName:"webgl",kernelFunc:Rm};function gu(e,t,r){let n=e[0].dtype;if(n==="complex64"){let d=e.map(m=>Vh({inputs:{input:m},backend:r})),h=e.map(m=>Rm({inputs:{input:m},backend:r})),p=gu(d,t,r),c=gu(h,t,r),f=Xi({inputs:{real:p,imag:c},backend:r});return d.forEach(m=>r.disposeIntermediateTensorInfo(m)),h.forEach(m=>r.disposeIntermediateTensorInfo(m)),r.disposeIntermediateTensorInfo(p),r.disposeIntermediateTensorInfo(c),f}let a=r.shouldExecuteOnCPU(e);if(n==="string"&&(a=!0),a){let d=e.map(y=>{let A=v.sizeFromShape(y.shape.slice(t));return ve({inputs:{x:y},backend:r,attrs:{shape:[-1,A]}})}),h=d.map(y=>({vals:r.readSync(y.dataId),shape:y.shape})),p=C.computeOutShape(d.map(y=>y.shape),1),c=d[0].shape[0]===1,f=Oee(h,p,n,c),m=C.computeOutShape(e.map(y=>y.shape),t),g=r.makeTensorInfo(m,n,f);return d.forEach(y=>r.disposeIntermediateTensorInfo(y)),g}if(e.length>Z().getNumber("WEBGL_MAX_TEXTURES_IN_SHADER")){let d=Math.floor(e.length/2),h=gu(e.slice(0,d),t,r),p=gu(e.slice(d),t,r),c=gu([h,p],t,r);return r.disposeIntermediateTensorInfo(h),r.disposeIntermediateTensorInfo(p),c}if(Z().getBool("WEBGL_PACK_ARRAY_OPERATIONS")&&e[0].shape.length>1){let d=new Wne(e.map(h=>h.shape),t);return r.runWebGLProgram(d,e,n)}let{tensors2D:s,outShape:i}=Une(e,t,r),o=new Bne(s.map(d=>d.shape)),l=r.runWebGLProgram(o,s,n);s.forEach(d=>r.disposeIntermediateTensorInfo(d));let u=ve({inputs:{x:l},attrs:{shape:i},backend:r});return r.disposeIntermediateTensorInfo(l),u}function Une(e,t,r){let n=C.computeOutShape(e.map(a=>a.shape),t);return{tensors2D:e.map(a=>ve({inputs:{x:a},attrs:{shape:[-1,v.sizeFromShape(a.shape.slice(t))]},backend:r})),outShape:n}}function b9(e){let{inputs:t,backend:r,attrs:n}=e,{axis:a}=n,s=v.parseAxisParam(a,t[0].shape)[0],i=C.computeOutShape(t.map(u=>u.shape),s);if(v.sizeFromShape(i)===0)return r.makeTensorInfo(i,t[0].dtype,[]);let o=t.filter(u=>v.sizeFromShape(u.shape)>0);if(o.length===1)return fn({inputs:{x:o[0]},backend:r});let l=o.map(u=>u.shape);return C.assertParamsConsistent(l,s),gu(o,s,r)}var Gne={kernelName:Xo,backendName:"webgl",kernelFunc:b9},v9=class{constructor(e,t=!1,r=null,n=!1,a=!1){this.variableNames=["x","W"],this.outputShape=e.outShape;let s=e.padInfo.top,i=e.padInfo.left,o=e.strideHeight,l=e.strideWidth,u=e.dilationHeight,d=e.dilationWidth,h=e.filterHeight,p=e.filterWidth,c=Math.floor(e.inChannels/4)*4,f=e.inChannels%4,m=e.dataFormat==="channelsLast",g=m?1:2,y=m?2:3,A=m?3:1,x="",b="";r&&(n?x=`float activation(float a) {
|
|
float b = getPreluActivationWeightsAtOutCoords();
|
|
${r}
|
|
}`:a?x=`float activation(float a) {
|
|
float b = getLeakyreluAlphaAtOutCoords();
|
|
${r}
|
|
}`:x=`
|
|
float activation(float x) {
|
|
${r}
|
|
}
|
|
`,b="result = activation(result);");let w=t?"result += getBiasAtOutCoords();":"";t&&this.variableNames.push("bias"),n&&this.variableNames.push("preluActivationWeights"),a&&this.variableNames.push("leakyreluAlpha"),this.userCode=`
|
|
${x}
|
|
|
|
const ivec2 strides = ivec2(${o}, ${l});
|
|
const ivec2 pads = ivec2(${s}, ${i});
|
|
|
|
void main() {
|
|
ivec4 coords = getOutputCoords();
|
|
int batch = coords[0];
|
|
int d2 = coords[${A}];
|
|
|
|
ivec2 xRCCorner =
|
|
ivec2(coords[${g}], coords[${y}]) * strides - pads;
|
|
int xRCorner = xRCCorner.x;
|
|
int xCCorner = xRCCorner.y;
|
|
|
|
// Convolve x(?, ?, d1) with w(:, :, d1, d2) to get y(yR, yC, d2).
|
|
// ? = to be determined. : = across all values in that axis.
|
|
float dotProd = 0.0;
|
|
for (int wR = 0; wR < ${h}; wR++) {
|
|
int xR = xRCorner + wR * ${u};
|
|
|
|
if (xR < 0 || xR >= ${e.inHeight}) {
|
|
continue;
|
|
}
|
|
|
|
for (int wC = 0; wC < ${p}; wC++) {
|
|
int xC = xCCorner + wC * ${d};
|
|
|
|
if (xC < 0 || xC >= ${e.inWidth}) {
|
|
continue;
|
|
}
|
|
|
|
for (int d1 = 0; d1 < ${c}; d1 += 4) {
|
|
vec4 wValues = vec4(
|
|
getW(wR, wC, d1, d2),
|
|
getW(wR, wC, d1 + 1, d2),
|
|
getW(wR, wC, d1 + 2, d2),
|
|
getW(wR, wC, d1 + 3, d2)
|
|
);
|
|
|
|
if (${m}) {
|
|
vec4 xValues = vec4(
|
|
getX(batch, xR, xC, d1),
|
|
getX(batch, xR, xC, d1 + 1),
|
|
getX(batch, xR, xC, d1 + 2),
|
|
getX(batch, xR, xC, d1 + 3)
|
|
);
|
|
dotProd += dot(xValues, wValues);
|
|
} else {
|
|
vec4 xValues = vec4(
|
|
getX(batch, d1, xR, xC),
|
|
getX(batch, d1 + 1, xR, xC),
|
|
getX(batch, d1 + 2, xR, xC),
|
|
getX(batch, d1 + 3, xR, xC)
|
|
);
|
|
dotProd += dot(xValues, wValues);
|
|
}
|
|
}
|
|
|
|
if (${f===1}) {
|
|
|
|
if (${m}) {
|
|
dotProd +=
|
|
getX(batch, xR, xC, ${c}) *
|
|
getW(wR, wC, ${c}, d2);
|
|
} else {
|
|
dotProd +=
|
|
getX(batch, ${c}, xR, xC) *
|
|
getW(wR, wC, ${c}, d2);
|
|
}
|
|
|
|
} else if (${f===2}) {
|
|
vec2 wValues = vec2(
|
|
getW(wR, wC, ${c}, d2),
|
|
getW(wR, wC, ${c} + 1, d2)
|
|
);
|
|
|
|
if (${m}) {
|
|
vec2 xValues = vec2(
|
|
getX(batch, xR, xC, ${c}),
|
|
getX(batch, xR, xC, ${c} + 1)
|
|
);
|
|
dotProd += dot(xValues, wValues);
|
|
} else {
|
|
vec2 xValues = vec2(
|
|
getX(batch, ${c}, xR, xC),
|
|
getX(batch, ${c} + 1, xR, xC)
|
|
);
|
|
dotProd += dot(xValues, wValues);
|
|
}
|
|
|
|
} else if (${f===3}) {
|
|
vec3 wValues = vec3(
|
|
getW(wR, wC, ${c}, d2),
|
|
getW(wR, wC, ${c} + 1, d2),
|
|
getW(wR, wC, ${c} + 2, d2)
|
|
);
|
|
|
|
if (${m}) {
|
|
vec3 xValues = vec3(
|
|
getX(batch, xR, xC, ${c}),
|
|
getX(batch, xR, xC, ${c} + 1),
|
|
getX(batch, xR, xC, ${c} + 2)
|
|
);
|
|
dotProd += dot(xValues, wValues);
|
|
} else {
|
|
vec3 xValues = vec3(
|
|
getX(batch, ${c}, xR, xC),
|
|
getX(batch, ${c} + 1, xR, xC),
|
|
getX(batch, ${c} + 2, xR, xC)
|
|
);
|
|
dotProd += dot(xValues, wValues);
|
|
}
|
|
|
|
}
|
|
}
|
|
}
|
|
|
|
float result = dotProd;
|
|
${w}
|
|
${b}
|
|
setOutput(result);
|
|
}
|
|
`}},jne=class{constructor(e){this.variableNames=["x","W"],this.outputShape=e.outShape;let t=e.padInfo.front,r=e.padInfo.top,n=e.padInfo.left,a=e.strideDepth,s=e.strideHeight,i=e.strideWidth,o=e.dilationDepth,l=e.dilationHeight,u=e.dilationWidth,d=e.filterDepth,h=e.filterHeight,p=e.filterWidth,c=Math.floor(e.inChannels/4)*4,f=e.inChannels%4;this.userCode=`
|
|
const ivec3 strides = ivec3(${a}, ${s}, ${i});
|
|
const ivec3 pads = ivec3(${t}, ${r}, ${n});
|
|
|
|
void main() {
|
|
ivec5 coords = getOutputCoords();
|
|
int batch = coords.x;
|
|
int d2 = coords.u;
|
|
|
|
ivec3 xFRCCorner = ivec3(coords.y, coords.z, coords.w) * strides - pads;
|
|
int xFCorner = xFRCCorner.x;
|
|
int xRCorner = xFRCCorner.y;
|
|
int xCCorner = xFRCCorner.z;
|
|
|
|
// Convolve x(?, ?, ?, d1) with w(:, :, :, d1, d2) to get
|
|
// y(yF, yR, yC, d2). ? = to be determined. : = across all
|
|
// values in that axis.
|
|
float dotProd = 0.0;
|
|
for (int wF = 0; wF < ${d}; wF++) {
|
|
int xF = xFCorner + wF * ${o};
|
|
|
|
if (xF < 0 || xF >= ${e.inDepth}) {
|
|
continue;
|
|
}
|
|
|
|
for (int wR = 0; wR < ${h}; wR++) {
|
|
int xR = xRCorner + wR * ${l};
|
|
|
|
if (xR < 0 || xR >= ${e.inHeight}) {
|
|
continue;
|
|
}
|
|
|
|
for (int wC = 0; wC < ${p}; wC++) {
|
|
int xC = xCCorner + wC * ${u};
|
|
|
|
if (xC < 0 || xC >= ${e.inWidth}) {
|
|
continue;
|
|
}
|
|
|
|
for (int d1 = 0; d1 < ${c}; d1 += 4) {
|
|
vec4 xValues = vec4(
|
|
getX(batch, xF, xR, xC, d1),
|
|
getX(batch, xF, xR, xC, d1 + 1),
|
|
getX(batch, xF, xR, xC, d1 + 2),
|
|
getX(batch, xF, xR, xC, d1 + 3)
|
|
);
|
|
vec4 wValues = vec4(
|
|
getW(wF, wR, wC, d1, d2),
|
|
getW(wF, wR, wC, d1 + 1, d2),
|
|
getW(wF, wR, wC, d1 + 2, d2),
|
|
getW(wF, wR, wC, d1 + 3, d2)
|
|
);
|
|
|
|
dotProd += dot(xValues, wValues);
|
|
}
|
|
|
|
if (${f===1}) {
|
|
dotProd +=
|
|
getX(batch, xF, xR, xC, ${c}) *
|
|
getW(wF, wR, wC, ${c}, d2);
|
|
} else if (${f===2}) {
|
|
vec2 xValues = vec2(
|
|
getX(batch, xF, xR, xC, ${c}),
|
|
getX(batch, xF, xR, xC, ${c} + 1)
|
|
);
|
|
vec2 wValues = vec2(
|
|
getW(wF, wR, wC, ${c}, d2),
|
|
getW(wF, wR, wC, ${c} + 1, d2)
|
|
);
|
|
dotProd += dot(xValues, wValues);
|
|
} else if (${f===3}) {
|
|
vec3 xValues = vec3(
|
|
getX(batch, xF, xR, xC, ${c}),
|
|
getX(batch, xF, xR, xC, ${c} + 1),
|
|
getX(batch, xF, xR, xC, ${c} + 2)
|
|
);
|
|
vec3 wValues = vec3(
|
|
getW(wF, wR, wC, ${c}, d2),
|
|
getW(wF, wR, wC, ${c} + 1, d2),
|
|
getW(wF, wR, wC, ${c} + 2, d2)
|
|
);
|
|
dotProd += dot(xValues, wValues);
|
|
}
|
|
}
|
|
}
|
|
}
|
|
setOutput(dotProd);
|
|
}
|
|
`}},Hne=class{constructor(e,t){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0,this.customUniforms=[{name:"inputShape",type:"ivec3"},{name:"pad",type:"ivec2"},{name:"stride",type:"ivec2"},{name:"dilation",type:"ivec2"},{name:"inChannels",type:"int"},{name:"itemsPerBlockRow",type:"int"},{name:"outWidth",type:"int"}],this.outputShape=e,this.enableShapeUniforms=ln(this.outputShape.length);let{dataFormat:r}=t,n=Hr(),a=r==="channelsLast",s=a?0:1,i=a?1:2,o=this.enableShapeUniforms?"if(blockIndex < outShape[1] && pos < outShape[0]) {":`if(blockIndex < ${e[1]} && pos < ${e[0]}) {`,l="";for(let u=0;u<=1;u++)for(let d=0;d<=1;d++)l+=`
|
|
blockIndex = rc.y + ${d};
|
|
pos = rc.x + ${u};
|
|
|
|
${o}
|
|
offsetY = int(blockIndex / outWidth) * stride[0] - pad[0];
|
|
d0 = offsetY + dilation[0] * (pos / itemsPerBlockRow);
|
|
|
|
if(d0 < inputShape[${s}] && d0 >= 0) {
|
|
// Use custom imod instead mod. On Intel GPU, mod may generate
|
|
// unexpected value.
|
|
// https://github.com/tensorflow/tfjs/issues/5447
|
|
offsetX = imod(blockIndex, outWidth) * stride[1] - pad[1];
|
|
d1 = offsetX + dilation[1] * (imod(pos, itemsPerBlockRow) /
|
|
inChannels);
|
|
|
|
if(d1 < inputShape[${i}] && d1 >= 0) {
|
|
|
|
ch = imod(pos, inChannels);
|
|
|
|
if (${a}) {
|
|
innerDims = vec2(d1, ch);
|
|
result[${u*2+d}] = getChannel(
|
|
getA(d0, int(innerDims.x),
|
|
int(innerDims.y)), innerDims);
|
|
} else {
|
|
innerDims = vec2(d0, d1);
|
|
result[${u*2+d}] = getChannel(
|
|
getA(ch, int(innerDims.x),
|
|
int(innerDims.y)), innerDims);
|
|
}
|
|
}
|
|
}
|
|
}
|
|
`;this.userCode=`
|
|
void main() {
|
|
ivec2 rc = getOutputCoords();
|
|
|
|
vec4 result = vec4(0);
|
|
|
|
int blockIndex, pos, offsetY, d0, offsetX, d1, ch;
|
|
vec2 innerDims;
|
|
|
|
${l}
|
|
|
|
${n.output} = result;
|
|
}
|
|
`}};function w9(e,t,r,n){let a=e.shape;if(v.assert(a.length<=1||a.length===3,()=>`WebGL conv2d only supports scalar, 1-D Tensor or 3-D Tensor PReLU activation weights but got a tensor of rank-${a.length}.`),a.length===1){let s=r?t[3]:t[1];v.assert(a[0]===1||a[0]===s,()=>`WebGL conv2d PReLU activation weights (${a}) is not compatible with the number of output channels (${s}).`)}else if(a.length===3){try{ji.assertAndGetBroadcastShape(a,t)}catch(s){let i=`WebGL conv2d PReLU activation weights (${a}) is not compatible with the output shape of the conv2d (${t}).`;throw Error(i)}if(!r)return dr({inputs:{x:e},backend:n,attrs:{perm:[1,2,0]}})}return e}function k9({x:e,filter:t,convInfo:r,backend:n,bias:a=null,preluActivationWeights:s=null,leakyreluAlpha:i=0,activation:o=null}){let l=e.shape,u=n.texData.get(e.dataId),d=r.inChannels,h=l[0]*l[1]*l[2],p=r.outChannels,c=r.dataFormat==="channelsLast",f=!1,m=!1,g,y=[];if(a!=null&&(v.assert(a.shape.length<=1,()=>`WebGL conv2dByMatMul only supports 1-D Tensor bias but got the bias of rank-${a.shape.length}.`),v.assert(a.shape.length===0||a.shape[0]===r.outChannels,()=>`WebGL conv2dByMatMul bias shape (${a.shape}) is not compatible with the number of output channels (${r.outChannels})`)),s!=null){let A=w9(s,r.outShape,c,n);A.dataId!==s.dataId&&(y.push(A),s=A)}if(!((h===1||p===1)&&d>m9)&&u.isPacked&&c&&u.texture!=null&&l[2]%2!==0&&v.arraysEqual(u.shape.slice(-3),l.slice(-3))){let A=l[0]*l[1]*(l[2]+1),x={dataId:e.dataId,shape:[1,A,r.inChannels],dtype:e.dtype},b=u.shape;u.shape=u.shape.slice(),u.shape[u.shape.length-2]++,v.assert(Xp(u.shape,x.shape),()=>`packed reshape ${u.shape} to ${x.shape} isn't free`);let w=ve({inputs:{x:t},backend:n,attrs:{shape:[1,r.inChannels,r.outChannels]}});y.push(w);let S=G0({a:x,b:w,backend:n,transposeA:f,transposeB:m,bias:a,activation:o,preluActivationWeights:s,leakyreluAlpha:i}),N=n.texData.get(S.dataId);v.assert(N.isPacked,()=>"batchMatMul result is expected to be packed"),u.shape=b,N.shape=r.outShape,g=fn({inputs:{x:S},backend:n}),g.shape=r.outShape,y.push(S)}else{let A=c?e:dr({inputs:{x:e},backend:n,attrs:{perm:[0,2,3,1]}}),x=A.shape,b=x[0]*x[1]*x[2],w=ve({inputs:{x:A},backend:n,attrs:{shape:[1,b,r.inChannels]}}),S=ve({inputs:{x:t},backend:n,attrs:{shape:[1,r.inChannels,r.outChannels]}}),N=G0({a:w,b:S,transposeA:f,transposeB:m,backend:n,bias:a,activation:o,preluActivationWeights:s,leakyreluAlpha:i}),E=[r.batchSize,r.outHeight,r.outWidth,r.outChannels],R=ve({inputs:{x:N},backend:n,attrs:{shape:E}});g=c?R:dr({inputs:{x:R},backend:n,attrs:{perm:[0,3,1,2]}}),c||(y.push(A),y.push(R)),y.push(w),y.push(S),y.push(N)}for(let A of y)n.disposeIntermediateTensorInfo(A);return g}function I9({x:e,filter:t,convInfo:r,backend:n,bias:a=null,preluActivationWeights:s=null,leakyreluAlpha:i=0,activation:o=null}){let{filterWidth:l,filterHeight:u,inChannels:d,outWidth:h,outHeight:p,dataFormat:c}=r,f=c==="channelsLast",m=l*u*d,g=p*h,y=[m,g],A=!0,x=!1,b=[];if(a!=null&&(v.assert(a.shape.length<=1,()=>`WebGL conv2dWithIm2Row only supports 1-D Tensor bias but got the bias of rank-${a.shape.length}.`),v.assert(a.shape.length===0||a.shape[0]===r.outChannels,()=>`WebGL conv2dWithIm2Row bias shape (${a.shape}) is not compatible with the number of output channels (${r.outChannels})`)),s!=null){let Y=w9(s,r.outShape,f,n);Y.dataId!==s.dataId&&(b.push(Y),s=Y)}let w=ve({inputs:{x:e},backend:n,attrs:{shape:e.shape.slice(1)}}),S=ve({inputs:{x:t},backend:n,attrs:{shape:[1,m,v.sizeFromShape(t.shape)/m]}});b.push(w),b.push(S);let N=new Hne(y,r),E=[w.shape,[r.padInfo.top,r.padInfo.left],[r.strideHeight,r.strideWidth],[r.dilationHeight,r.dilationWidth],[r.inChannels],[r.filterWidth*r.inChannels],[r.outWidth]],R=n.runWebGLProgram(N,[w],"float32",E),z=ve({inputs:{x:R},backend:n,attrs:{shape:[1,y[0],y[1]]}});b.push(R),b.push(z);let M=a!=null,I=s!=null,_=o==="leakyrelu",O=o?Tm(o,!0):null,q=new f9(z.shape,S.shape,[1,g,r.outChannels],A,x,M,O,I,_),K=[z,S];if(a&&K.push(a),I&&K.push(s),_){let Y=n.makeTensorInfo([],"float32",v.createScalarValue(i,"float32"));K.push(Y),b.push(Y)}let D=n.runWebGLProgram(q,K,"float32"),Q=[1,p,h,r.outChannels],V=ve({inputs:{x:D},backend:n,attrs:{shape:Q}}),re=f?V:dr({inputs:{x:V},backend:n,attrs:{perm:[0,3,1,2]}});f||b.push(V),b.push(D);for(let Y of b)n.disposeIntermediateTensorInfo(Y);return re}function qne(e){let{inputs:t,backend:r,attrs:n}=e,{x:a,filter:s}=t,{strides:i,pad:o,dataFormat:l,dilations:u,dimRoundingMode:d}=n,h=C.convertConv2DDataFormat(l),p=C.computeConv2DInfo(a.shape,s.shape,i,u,o,d,!1,h),c;if(p.filterHeight===1&&p.filterWidth===1&&p.dilationHeight===1&&p.dilationWidth===1&&p.strideHeight===1&&p.strideWidth===1&&(p.padInfo.type==="SAME"||p.padInfo.type==="VALID"))c=k9({x:a,filter:s,convInfo:p,backend:r});else if(Z().getBool("WEBGL_CONV_IM2COL")&&a.shape[0]===1)c=I9({x:a,filter:s,convInfo:p,backend:r});else{let m=new v9(p);c=r.runWebGLProgram(m,[a,s],"float32")}let f=ve({inputs:{x:c},backend:r,attrs:{shape:p.outShape}});return r.disposeIntermediateTensorInfo(c),f}var Xne={kernelName:ri,backendName:"webgl",kernelFunc:qne},Kne=class{constructor(e){this.variableNames=["x","dy"],this.outputShape=e.filterShape;let t=e.strideHeight,r=e.strideWidth,n=e.padInfo.top,a=e.padInfo.left,s=e.dataFormat==="channelsLast";this.userCode=`
|
|
void main() {
|
|
ivec4 coords = getOutputCoords();
|
|
int wR = coords.x;
|
|
int wC = coords.y;
|
|
int d1 = coords.z;
|
|
int d2 = coords.w;
|
|
|
|
// Convolve x(?, ?, d1) with dy(:, :, d2) to get dw(wR, wC, d1, d2).
|
|
// ? = to be determined. : = across all values in that axis.
|
|
float dotProd = 0.0;
|
|
|
|
for (int b = 0; b < ${e.batchSize}; b++) {
|
|
for (int yR = 0; yR < ${e.outHeight}; yR++) {
|
|
int xR = wR + yR * ${t} - ${n};
|
|
|
|
if (xR < 0 || xR >= ${e.inHeight}) {
|
|
continue;
|
|
}
|
|
|
|
for (int yC = 0; yC < ${e.outWidth}; yC++) {
|
|
int xC = wC + yC * ${r} - ${a};
|
|
|
|
if (xC < 0 || xC >= ${e.inWidth}) {
|
|
continue;
|
|
}
|
|
|
|
if (${s}) {
|
|
float dyValue = getDy(b, yR, yC, d2);
|
|
float xValue = getX(b, xR, xC, d1);
|
|
dotProd += (xValue * dyValue);
|
|
} else {
|
|
float dyValue = getDy(b, d2, yR, yC);
|
|
float xValue = getX(b, d1, xR, xC);
|
|
dotProd += (xValue * dyValue);
|
|
}
|
|
|
|
}
|
|
}
|
|
}
|
|
setOutput(dotProd);
|
|
}
|
|
`}},Zne=class{constructor(e){this.variableNames=["dy","W"],this.outputShape=e.inShape;let t=e.filterHeight,r=e.filterWidth,n=e.strideHeight,a=e.strideWidth,s=e.dataFormat==="channelsLast",i=t-1-e.padInfo.top,o=r-1-e.padInfo.left,l=s?1:2,u=s?2:3,d=s?3:1;this.userCode=`
|
|
const ivec2 pads = ivec2(${i}, ${o});
|
|
|
|
void main() {
|
|
ivec4 coords = getOutputCoords();
|
|
int batch = coords[0];
|
|
int d1 = coords[${d}];
|
|
|
|
ivec2 dyCorner = ivec2(coords[${l}], coords[${u}]) - pads;
|
|
int dyRCorner = dyCorner.x;
|
|
int dyCCorner = dyCorner.y;
|
|
|
|
// Convolve dy(?, ?, d2) with w(:, :, d1, d2) to compute dx(xR, xC, d1).
|
|
// ? = to be determined. : = across all values in that axis.
|
|
float dotProd = 0.0;
|
|
for (int wR = 0; wR < ${t}; wR++) {
|
|
float dyR = float(dyRCorner + wR) / ${n}.0;
|
|
|
|
if (dyR < 0.0 || dyR >= ${e.outHeight}.0 || fract(dyR) > 0.0) {
|
|
continue;
|
|
}
|
|
int idyR = int(dyR);
|
|
|
|
int wRPerm = ${t} - 1 - wR;
|
|
|
|
for (int wC = 0; wC < ${r}; wC++) {
|
|
float dyC = float(dyCCorner + wC) / ${a}.0;
|
|
|
|
if (dyC < 0.0 || dyC >= ${e.outWidth}.0 ||
|
|
fract(dyC) > 0.0) {
|
|
continue;
|
|
}
|
|
int idyC = int(dyC);
|
|
|
|
int wCPerm = ${r} - 1 - wC;
|
|
|
|
for (int d2 = 0; d2 < ${e.outChannels}; d2++) {
|
|
|
|
if (${s}) {
|
|
float xValue = getDy(batch, idyR, idyC, d2);
|
|
float wValue = getW(wRPerm, wCPerm, d1, d2);
|
|
dotProd += xValue * wValue;
|
|
} else {
|
|
float xValue = getDy(batch, d2, idyR, idyC);
|
|
float wValue = getW(wRPerm, wCPerm, d1, d2);
|
|
dotProd += xValue * wValue;
|
|
}
|
|
|
|
}
|
|
}
|
|
}
|
|
setOutput(dotProd);
|
|
}
|
|
`}},Yne=class{constructor(e){this.variableNames=["x","dy"],this.outputShape=e.filterShape;let t=e.strideDepth,r=e.strideHeight,n=e.strideWidth,a=e.padInfo.front,s=e.padInfo.top,i=e.padInfo.left;this.userCode=`
|
|
void main() {
|
|
ivec5 coords = getOutputCoords();
|
|
int wF = coords.x;
|
|
int wR = coords.y;
|
|
int wC = coords.z;
|
|
int d1 = coords.w;
|
|
int d2 = coords.u;
|
|
|
|
float dotProd = 0.0;
|
|
|
|
for (int b = 0; b < ${e.batchSize}; b++) {
|
|
for (int yF = 0; yF < ${e.outDepth}; yF++) {
|
|
int xF = wF + yF * ${t} - ${a};
|
|
|
|
if (xF < 0 || xF >= ${e.inDepth}) {
|
|
continue;
|
|
}
|
|
|
|
for (int yR = 0; yR < ${e.outHeight}; yR++) {
|
|
int xR = wR + yR * ${r} - ${s};
|
|
|
|
if (xR < 0 || xR >= ${e.inHeight}) {
|
|
continue;
|
|
}
|
|
|
|
for (int yC = 0; yC < ${e.outWidth}; yC++) {
|
|
int xC = wC + yC * ${n} - ${i};
|
|
|
|
if (xC < 0 || xC >= ${e.inWidth}) {
|
|
continue;
|
|
}
|
|
|
|
float dyValue = getDy(b, yF, yR, yC, d2);
|
|
float xValue = getX(b, xF, xR, xC, d1);
|
|
dotProd += (xValue * dyValue);
|
|
}
|
|
}
|
|
}
|
|
}
|
|
setOutput(dotProd);
|
|
}
|
|
`}},Jne=class{constructor(e){this.variableNames=["dy","W"],this.outputShape=e.inShape;let t=e.filterDepth,r=e.filterHeight,n=e.filterWidth,a=e.strideDepth,s=e.strideHeight,i=e.strideWidth,o=t-1-e.padInfo.front,l=r-1-e.padInfo.top,u=n-1-e.padInfo.left;this.userCode=`
|
|
const ivec3 pads = ivec3(${o}, ${l}, ${u});
|
|
|
|
void main() {
|
|
ivec5 coords = getOutputCoords();
|
|
int batch = coords.x;
|
|
int d1 = coords.u;
|
|
|
|
|
|
ivec3 dyCorner = ivec3(coords.y, coords.z, coords.w) - pads;
|
|
int dyFCorner = dyCorner.x;
|
|
int dyRCorner = dyCorner.y;
|
|
int dyCCorner = dyCorner.z;
|
|
|
|
float dotProd = 0.0;
|
|
for (int wF = 0; wF < ${t}; wF++) {
|
|
float dyF = float(dyFCorner + wF) / ${a}.0;
|
|
|
|
if (dyF < 0.0 || dyF >= ${e.outDepth}.0 || fract(dyF) > 0.0) {
|
|
continue;
|
|
}
|
|
int idyF = int(dyF);
|
|
|
|
int wFPerm = ${t} - 1 - wF;
|
|
|
|
for (int wR = 0; wR < ${r}; wR++) {
|
|
float dyR = float(dyRCorner + wR) / ${s}.0;
|
|
|
|
if (dyR < 0.0 || dyR >= ${e.outHeight}.0 ||
|
|
fract(dyR) > 0.0) {
|
|
continue;
|
|
}
|
|
int idyR = int(dyR);
|
|
|
|
int wRPerm = ${r} - 1 - wR;
|
|
|
|
for (int wC = 0; wC < ${n}; wC++) {
|
|
float dyC = float(dyCCorner + wC) / ${i}.0;
|
|
|
|
if (dyC < 0.0 || dyC >= ${e.outWidth}.0 ||
|
|
fract(dyC) > 0.0) {
|
|
continue;
|
|
}
|
|
int idyC = int(dyC);
|
|
|
|
int wCPerm = ${n} - 1 - wC;
|
|
|
|
for (int d2 = 0; d2 < ${e.outChannels}; d2++) {
|
|
float xValue = getDy(batch, idyF, idyR, idyC, d2);
|
|
float wValue = getW(wFPerm, wRPerm, wCPerm, d1, d2);
|
|
dotProd += xValue * wValue;
|
|
}
|
|
}
|
|
}
|
|
}
|
|
setOutput(dotProd);
|
|
}
|
|
`}};function Qne(e){let{inputs:t,backend:r,attrs:n}=e,{x:a,dy:s}=t,{strides:i,pad:o,dataFormat:l,dimRoundingMode:u,filterShape:d}=n,h=C.convertConv2DDataFormat(l),p=C.computeConv2DInfo(a.shape,d,i,1,o,u,!1,h),c=new Kne(p);return r.runWebGLProgram(c,[a,s],"float32")}var eae={kernelName:rf,backendName:"webgl",kernelFunc:Qne};function tae(e){let{inputs:t,backend:r,attrs:n}=e,{dy:a,filter:s}=t,{inputShape:i,strides:o,pad:l,dataFormat:u,dimRoundingMode:d}=n,h=C.convertConv2DDataFormat(u),p=C.computeConv2DInfo(i,s.shape,o,1,l,d,!1,h),c=new Zne(p);return r.runWebGLProgram(c,[a,s],"float32")}var rae={kernelName:ni,backendName:"webgl",kernelFunc:tae};function nae(e){let{inputs:t,backend:r,attrs:n}=e,{x:a,filter:s}=t,{strides:i,pad:o,dilations:l}=n,u=C.computeConv3DInfo(a.shape,s.shape,i,l,o),d=new jne(u);return r.runWebGLProgram(d,[a,s],"float32")}var aae={kernelName:eh,backendName:"webgl",kernelFunc:nae};function sae(e){let{inputs:t,backend:r,attrs:n}=e,{x:a,dy:s}=t,{strides:i,pad:o,filterShape:l}=n,u=C.computeConv3DInfo(a.shape,l,i,1,o),d=new Yne(u);return r.runWebGLProgram(d,[a,s],"float32")}var iae={kernelName:nf,backendName:"webgl",kernelFunc:sae};function oae(e){let{inputs:t,backend:r,attrs:n}=e,{dy:a,filter:s}=t,{pad:i,strides:o,inputShape:l}=n,u=C.computeConv3DInfo(l,s.shape,o,1,i),d=new Jne(u);return r.runWebGLProgram(d,[a,s],"float32")}var lae={kernelName:af,backendName:"webgl",kernelFunc:oae},uae=Ed+`
|
|
return cos(x);
|
|
`,dae=it({opSnippet:uae}),pae={kernelName:ai,backendName:"webgl",kernelFunc:dae},hae=`
|
|
float e2x = exp(-x);
|
|
return (e2x + 1.0 / e2x) / 2.0;
|
|
`,cae=it({opSnippet:hae}),fae={kernelName:si,backendName:"webgl",kernelFunc:cae},mae=class{constructor(e,t,r,n,a){this.variableNames=["Image","Boxes","BoxInd"],this.outputShape=[];let[s,i,o,l]=e,[u]=t,[d,h]=r;this.outputShape=[u,d,h,l];let p=n==="bilinear"?1:0,[c,f]=[`${i-1}.0`,`${o-1}.0`],[m,g,y]=d>1?[`${(i-1)/(d-1)}`,"(y2-y1) * height_ratio",`y1*${c} + float(y)*(height_scale)`]:["0.0","0.0",`0.5 * (y1+y2) * ${c}`],[A,x,b]=h>1?[`${(o-1)/(h-1)}`,"(x2-x1) * width_ratio",`x1*${f} + float(x)*(width_scale)`]:["0.0","0.0",`0.5 * (x1+x2) * ${f}`];this.userCode=`
|
|
const float height_ratio = float(${m});
|
|
const float width_ratio = float(${A});
|
|
void main() {
|
|
ivec4 coords = getOutputCoords();
|
|
int b = coords[0];
|
|
int y = coords[1];
|
|
int x = coords[2];
|
|
int d = coords[3];
|
|
|
|
// get box vals
|
|
float y1 = getBoxes(b,0);
|
|
float x1 = getBoxes(b,1);
|
|
float y2 = getBoxes(b,2);
|
|
float x2 = getBoxes(b,3);
|
|
|
|
// get image in batch index
|
|
int bInd = round(getBoxInd(b));
|
|
if(bInd < 0 || bInd >= ${s}) {
|
|
return;
|
|
}
|
|
|
|
float height_scale = ${g};
|
|
float width_scale = ${x};
|
|
|
|
float in_y = ${y};
|
|
if( in_y < 0.0 || in_y > ${c} ) {
|
|
setOutput(float(${a}));
|
|
return;
|
|
}
|
|
float in_x = ${b};
|
|
if( in_x < 0.0 || in_x > ${f} ) {
|
|
setOutput(float(${a}));
|
|
return;
|
|
}
|
|
|
|
vec2 sourceFracIndexCR = vec2(in_x,in_y);
|
|
if(${p} == 1) {
|
|
// Compute the four integer indices.
|
|
ivec2 sourceFloorCR = ivec2(sourceFracIndexCR);
|
|
ivec2 sourceCeilCR = ivec2(ceil(sourceFracIndexCR));
|
|
|
|
float topLeft = getImage(b, sourceFloorCR.y, sourceFloorCR.x, d);
|
|
float bottomLeft = getImage(b, sourceCeilCR.y, sourceFloorCR.x, d);
|
|
float topRight = getImage(b, sourceFloorCR.y, sourceCeilCR.x, d);
|
|
float bottomRight = getImage(b, sourceCeilCR.y, sourceCeilCR.x, d);
|
|
|
|
vec2 fracCR = sourceFracIndexCR - vec2(sourceFloorCR);
|
|
|
|
float top = topLeft + (topRight - topLeft) * fracCR.x;
|
|
float bottom = bottomLeft + (bottomRight - bottomLeft) * fracCR.x;
|
|
float newValue = top + (bottom - top) * fracCR.y;
|
|
setOutput(newValue);
|
|
} else {
|
|
// Compute the coordinators of nearest neighbor point.
|
|
ivec2 sourceNearestCR = ivec2(floor(
|
|
sourceFracIndexCR + vec2(0.5,0.5)));
|
|
float newValue = getImage(b, sourceNearestCR.y, sourceNearestCR.x, d);
|
|
setOutput(newValue);
|
|
}
|
|
}
|
|
`}},gae=e=>{let{inputs:t,backend:r,attrs:n}=e,{image:a,boxes:s,boxInd:i}=t,{cropSize:o,method:l,extrapolationValue:u}=n,d=new mae(a.shape,s.shape,o,l,u);return r.runWebGLProgram(d,[a,s,i],"float32")},yae={kernelName:Zo,backendName:"webgl",kernelFunc:gae},Z4=class{constructor(e,t,r,n){this.variableNames=["x"],this.customUniforms=[{name:"index",type:"float"}],this.op=e,this.outputShape=t;let a=t.length,s=this.op==="*"?"1.0":"0.0",i=r?s:`getX(${Y4(a,"coords",this.op)})`,o=t[t.length-1],l="",u="";r?(l=n?`end != ${o-1}`:"end != 0",u=n?"end + 1":"end - 1"):(l=n?`end + pow2 < ${o}`:"end >= pow2",u=n?"end + pow2":"end - pow2"),this.userCode=`
|
|
void main() {
|
|
${vt(a)} coords = getOutputCoords();
|
|
int end = ${J4(a,"coords",this.op)};
|
|
float val = ${i};
|
|
int pow2 = int(pow(2.0, index));
|
|
if (${l}) {
|
|
int idx = ${u};
|
|
${J4(a,"coords",this.op)} = idx;
|
|
val ${this.op}= getX(${Y4(a,"coords",this.op)});
|
|
}
|
|
setOutput(val);
|
|
}
|
|
`}};function Y4(e,t,r){if(e===1)return`${t}`;if(e===2)return`${t}.x, ${t}.y`;if(e===3)return`${t}.x, ${t}.y, ${t}.z`;if(e===4)return`${t}.x, ${t}.y, ${t}.z, ${t}.w`;throw Error(`Cumulative ${r} for rank ${e} is not yet supported`)}function J4(e,t,r){if(e===1)return`${t}`;if(e===2)return`${t}.y`;if(e===3)return`${t}.z`;if(e===4)return`${t}.w`;throw Error(`Cumulative ${r} for rank ${e} is not yet supported`)}function S9(e,t,r,n,a,s){let i=t.shape.length,o=C.getAxesPermutation([n],i),l=t;o!=null&&(l=dr({inputs:{x:t},backend:r,attrs:{perm:o}}));let u=C.getInnerMostAxes(1,i)[0];if(u!==i-1)throw new Error(`WebGL cumprod shader expects an inner-most axis=${t.shape.length-1} but got axis=${n}`);let d=l.shape[u],h=fn({inputs:{x:l},backend:r});for(let p=0;p<=Math.ceil(Math.log2(d))-1;p++){let c=new Z4(e,l.shape,!1,s),f=[[p]],m=h;h=r.runWebGLProgram(c,[h],h.dtype,f),r.disposeIntermediateTensorInfo(m)}if(a){let p=new Z4(e,l.shape,a,s),c=h;h=r.runWebGLProgram(p,[h],h.dtype),r.disposeIntermediateTensorInfo(c)}if(o!=null){let p=C.getUndoAxesPermutation(o),c=dr({inputs:{x:h},backend:r,attrs:{perm:p}});return r.disposeIntermediateTensorInfo(h),r.disposeIntermediateTensorInfo(l),c}return h}function Aae(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{axis:s,exclusive:i,reverse:o}=n;return S9("*",a,r,s,i,o)}var xae={kernelName:Ko,backendName:"webgl",kernelFunc:Aae};function bae(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{axis:s,exclusive:i,reverse:o}=n;return S9("+",a,r,s,i,o)}var vae={kernelName:ii,backendName:"webgl",kernelFunc:bae};function wae(e){let{inputs:t,backend:r,attrs:n}=e,{x:a,weights:s}=t,{size:i,binaryOutput:o}=n;if(a.shape.length===1){let l=r.readSync(a.dataId),u=r.readSync(s.dataId),d=r9(l,u,s.dtype,s.shape,i);return r.makeTensorInfo([i],s.dtype,d)}else if(a.shape.length===2){let l=r.bufferSync(a),u=r.bufferSync(s),d=_ee(l,u,i,o);return r.makeTensorInfo(d.shape,s.dtype,d.values)}throw new Error(`Error in denseBincount: input must be at most rank 2, but got rank${a.shape.length}.`)}var kae={kernelName:sf,backendName:"webgl",kernelFunc:wae},Iae=class{constructor(e,t,r){this.variableNames=["x"],this.outputShape=[],this.outputShape=e,this.blockSize=t,this.dataFormat=r,this.userCode=`
|
|
void main() {
|
|
ivec4 coords = getOutputCoords();
|
|
int b = coords[0];
|
|
int h = ${this.getHeightCoordString()};
|
|
int w = ${this.getWidthCoordString()};
|
|
int d = ${this.getDepthCoordString()};
|
|
|
|
int in_h = h / ${t};
|
|
int offset_h = imod(h, ${t});
|
|
int in_w = w / ${t};
|
|
int offset_w = imod(w, ${t});
|
|
int offset_d = (offset_h * ${t} + offset_w) *
|
|
${this.getOutputDepthSize()};
|
|
int in_d = d + offset_d;
|
|
|
|
float result = ${this.getInputSamplingString()};
|
|
setOutput(result);
|
|
}
|
|
`}getHeightCoordString(){return this.dataFormat==="NHWC"?"coords[1]":"coords[2]"}getWidthCoordString(){return this.dataFormat==="NHWC"?"coords[2]":"coords[3]"}getDepthCoordString(){return this.dataFormat==="NHWC"?"coords[3]":"coords[1]"}getOutputDepthSize(){return this.dataFormat==="NHWC"?this.outputShape[3]:this.outputShape[1]}getInputSamplingString(){return this.dataFormat==="NHWC"?"getX(b, in_h, in_w, in_d)":"getX(b, in_d, in_h, in_w)"}};function Sae(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{blockSize:s,dataFormat:i}=n,o=a.shape[0],l=i==="NHWC"?a.shape[1]:a.shape[2],u=i==="NHWC"?a.shape[2]:a.shape[3],d=i==="NHWC"?a.shape[3]:a.shape[1],h=l*s,p=u*s,c=d/(s*s),f=i==="NHWC"?[o,h,p,c]:[o,c,h,p],m=new Iae(f,s,i);return r.runWebGLProgram(m,[a],a.dtype)}var Cae={kernelName:Yo,backendName:"webgl",kernelFunc:Sae},C9=class{constructor(e,t=!1,r=null,n=!1,a=!1){this.variableNames=["x","W"],this.customUniforms=[{name:"pads",type:"ivec2"},{name:"strides",type:"ivec2"},{name:"dilations",type:"ivec2"},{name:"inDims",type:"ivec2"}],this.outputShape=e.outShape,this.enableShapeUniforms=ln(this.outputShape.length);let s=e.filterHeight,i=e.filterWidth,o=e.outChannels/e.inChannels,l="",u="";r&&(n?l=`float activation(float a) {
|
|
float b = getPreluActivationWeightsAtOutCoords();
|
|
${r}
|
|
}`:a?l=`float activation(float a) {
|
|
float b = getLeakyreluAlphaAtOutCoords();
|
|
${r}
|
|
}`:l=`
|
|
float activation(float x) {
|
|
${r}
|
|
}
|
|
`,u="result = activation(result);");let d=t?"result += getBiasAtOutCoords();":"";t&&this.variableNames.push("bias"),n&&this.variableNames.push("preluActivationWeights"),a&&this.variableNames.push("leakyreluAlpha"),this.userCode=`
|
|
${l}
|
|
|
|
void main() {
|
|
ivec4 coords = getOutputCoords();
|
|
int batch = coords.x;
|
|
ivec2 xRCCorner = coords.yz * strides - pads;
|
|
int d2 = coords.w;
|
|
int d1 = d2 / ${o};
|
|
int q = d2 - d1 * ${o};
|
|
|
|
int xRCorner = xRCCorner.x;
|
|
int xCCorner = xRCCorner.y;
|
|
|
|
// Convolve x(?, ?, d1) with w(:, :, d1, q) to get y(yR, yC, d2).
|
|
// ? = to be determined. : = across all values in that axis.
|
|
float dotProd = 0.0;
|
|
// TO DO(dsmilkov): Flatten the two for loops and vec4 the operations.
|
|
for (int wR = 0; wR < ${s}; wR++) {
|
|
int xR = xRCorner + wR * dilations[0];
|
|
|
|
if (xR < 0 || xR >= inDims[0]) {
|
|
continue;
|
|
}
|
|
|
|
for (int wC = 0; wC < ${i}; wC++) {
|
|
int xC = xCCorner + wC * dilations[1];
|
|
|
|
if (xC < 0 || xC >= inDims[1]) {
|
|
continue;
|
|
}
|
|
|
|
float xVal = getX(batch, xR, xC, d1);
|
|
float wVal = getW(wR, wC, d1, q);
|
|
dotProd += xVal * wVal;
|
|
}
|
|
}
|
|
|
|
float result = dotProd;
|
|
${d}
|
|
${u}
|
|
setOutput(result);
|
|
}
|
|
`}},T9=class{constructor(e,t=!1,r=null,n=!1,a=!1){this.variableNames=["x","W"],this.packedInputs=!0,this.packedOutput=!0,this.customUniforms=[{name:"pads",type:"ivec2"},{name:"strides",type:"ivec2"},{name:"dilations",type:"ivec2"},{name:"inDims",type:"ivec2"}],this.outputShape=e.outShape,this.enableShapeUniforms=ln(this.outputShape.length);let s=e.outChannels/e.inChannels,i=e.padInfo.left,o=e.strideWidth,l=e.dilationWidth,u=e.filterHeight,d=e.filterWidth,h=d,p=`
|
|
int xR; int xC; int xCOffset;
|
|
vec4 wTexel; vec4 previous; vec4 final;`;for(let g=0;g<d;g++)p+=`
|
|
vec4 xTexelC${g*2};
|
|
int xTexelC${g*2}Ready;
|
|
vec4 xTexelC${g*2+1};
|
|
int xTexelC${g*2+1}Ready;
|
|
vec4 xC${g};`;p+=`
|
|
for (int r = 0; r < ${u}; r++) {
|
|
`;for(let g=0;g<d;g++)p+=`
|
|
xTexelC${g*2} = vec4(0.0);
|
|
xTexelC${g*2}Ready = 0;
|
|
xTexelC${g*2+1} = vec4(0.0);
|
|
xTexelC${g*2+1}Ready = 0;
|
|
xC${g} = vec4(0.0);`;p+=`
|
|
xR = xRCorner + r * dilations[0];
|
|
if (xR >=0 && xR < inDims[0]) {
|
|
`;for(let g=0;g<(h+1)/2;g++){let y=g*2;if(p+=`
|
|
xC = xCCorner + ${y*l};
|
|
`,o===1){if(y<d&&(i%2===1?(p+=`
|
|
xCOffset = xC + 1;
|
|
if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${y}Ready == 0) {
|
|
xTexelC${y} = getX(batch, xR, xCOffset, d1);
|
|
|
|
// Need to manually clear unused channels in case
|
|
// we're reading from recycled texture.
|
|
if (xCOffset + 1 >= inDims[1]) {
|
|
xTexelC${y}.zw = vec2(0.0);
|
|
}
|
|
xTexelC${y}Ready = 1;
|
|
}
|
|
`,l===1&&y>0?p+=`
|
|
xC${y} = vec4(xTexelC${y-2}.zw, xTexelC${y}.xy);
|
|
`:p+=`
|
|
xCOffset = xC + 1 - 2;
|
|
|
|
if (xCOffset >= 0 && xCOffset < inDims[1]) {
|
|
previous = getX(batch, xR, xCOffset, d1);
|
|
|
|
// Need to manually clear unused channels in case
|
|
// we're reading from recycled texture.
|
|
if (xCOffset + 1 >= inDims[1]) {
|
|
previous.zw = vec2(0.0);
|
|
}
|
|
|
|
xC${y} = vec4(previous.zw, xTexelC${y}.xy);
|
|
} else {
|
|
xC${y} = vec4(0.0, 0.0, xTexelC${y}.xy);
|
|
}
|
|
`):p+=`
|
|
if (xC >= 0 && xC < inDims[1] && xTexelC${y}Ready == 0) {
|
|
xTexelC${y} = getX(batch, xR, xC, d1);
|
|
if (xC + 1 >= inDims[1]) {
|
|
xTexelC${y}.zw = vec2(0.0);
|
|
}
|
|
xTexelC${y}Ready = 1;
|
|
}
|
|
|
|
xC${y} = xTexelC${y};
|
|
`,y+1<d)){let A=i%2===0?v.nearestLargerEven(l):l;l%2===0&&i%2===1||l%2!==0&&i%2!==1?(p+=`
|
|
xCOffset = xC + imod(pads[1], 2) + ${A};
|
|
|
|
if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${y+1}Ready == 0) {
|
|
xTexelC${y+1} = getX(batch, xR, xCOffset, d1);
|
|
|
|
// Need to manually clear unused channels in case
|
|
// we're reading from recycled texture.
|
|
if (xCOffset + 1 >= inDims[1]) {
|
|
xTexelC${y+1}.zw = vec2(0.0);
|
|
}
|
|
xTexelC${y+1}Ready = 1;
|
|
}
|
|
`,l>1&&(p+=`
|
|
xCOffset -= 2;
|
|
if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${y}Ready == 0) {
|
|
xTexelC${y} = getX(batch, xR, xCOffset, d1);
|
|
xTexelC${y}Ready = 1;
|
|
}
|
|
`),p+=`
|
|
xC${y+1} = vec4(xTexelC${y}.zw, xTexelC${y+1}.xy);
|
|
`):A===1?p+=`
|
|
xC${y+1} = xTexelC${y};
|
|
`:p+=`
|
|
xCOffset = xC + ${A};
|
|
|
|
if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${y+1}Ready == 0) {
|
|
xTexelC${y+1} = getX(batch, xR, xCOffset, d1);
|
|
if (xCOffset + 1 >= inDims[1]) {
|
|
xTexelC${y+1}.zw = vec2(0.0);
|
|
}
|
|
xTexelC${y+1}Ready = 1;
|
|
}
|
|
|
|
xC${y+1} = xTexelC${y+1};
|
|
`}}else y<d&&(i%2===1?(p+=`
|
|
xCOffset = xC + 1 - strides[1];
|
|
if(xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${y}Ready == 0) {
|
|
xTexelC${y} = getX(batch, xR, xCOffset, d1);
|
|
// Need to manually clear unused channels in case
|
|
// we're reading from recycled texture.
|
|
if (xCOffset + 1 >= inDims[1]) {
|
|
xTexelC${y}.zw = vec2(0.0);
|
|
}
|
|
xTexelC${y}Ready = 1;
|
|
}
|
|
|
|
if(xC + 1 >= 0 && xC + 1 < inDims[1] && xTexelC${y+1}Ready == 0) {
|
|
xTexelC${y+1} = getX(batch, xR, xC + 1, d1);
|
|
// Need to manually clear unused channels in case
|
|
// we're reading from recycled texture.
|
|
if (xC + 2 >= inDims[1]) {
|
|
xTexelC${y+1}.zw = vec2(0.0);
|
|
}
|
|
xTexelC${y+1}Ready = 1;
|
|
}
|
|
|
|
xC${y} = vec4(xTexelC${y}.zw, xTexelC${y+1}.zw);
|
|
`,y+1<d&&(p+=`
|
|
final = vec4(0.0);
|
|
xCOffset = xC + 1 + strides[1];
|
|
if(xCOffset >= 0 && xCOffset < inDims[1]) {
|
|
final = getX(batch, xR, xCOffset, d1);
|
|
}
|
|
xC${y+1} = vec4(xTexelC${y+1}.xy, final.xy);
|
|
`)):(p+=`
|
|
if(xC >= 0 && xC < inDims[1] && xTexelC${y}Ready == 0) {
|
|
xTexelC${y} = getX(batch, xR, xC, d1);
|
|
if (xC + 1 >= inDims[1]) {
|
|
xTexelC${y}.zw = vec2(0.0);
|
|
}
|
|
xTexelC${y}Ready = 1;
|
|
}
|
|
|
|
xCOffset = xC + strides[1];
|
|
if(xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${y+1}Ready == 0) {
|
|
xTexelC${y+1} = getX(batch, xR, xCOffset, d1);
|
|
if (xCOffset + 1 >= inDims[1]) {
|
|
xTexelC${y+1}.zw = vec2(0.);
|
|
}
|
|
xTexelC${y+1}Ready = 1;
|
|
}
|
|
|
|
xC${y} = vec4(
|
|
xTexelC${y}.xy, xTexelC${y+1}.xy);
|
|
`,y+1<d&&(p+=`
|
|
xC${y+1} = vec4(xTexelC${y}.zw, xTexelC${y+1}.zw);
|
|
`)));y<d&&(p+=`
|
|
wTexel = getW(r, ${y}, d1, q);
|
|
dotProd += xC${y} * vec4(wTexel.xz, wTexel.xz);
|
|
`,y+1<d&&(p+=`
|
|
wTexel = getW(r, ${y+1}, d1, q);
|
|
dotProd += xC${y+1} * vec4(wTexel.xz, wTexel.xz);
|
|
`))}p+=`
|
|
}
|
|
`,p+=`
|
|
}
|
|
`;let c="",f="";r&&(n?c=`vec4 activation(vec4 a) {
|
|
vec4 b = getPreluActivationWeightsAtOutCoords();
|
|
${r}
|
|
}`:a?c=`vec4 activation(vec4 a) {
|
|
vec4 b = getLeakyreluAlphaAtOutCoords();
|
|
${r}
|
|
}`:c=`vec4 activation(vec4 x) {
|
|
${r}
|
|
}`,f="result = activation(result);");let m=t?"result += getBiasAtOutCoords();":"";t&&this.variableNames.push("bias"),n&&this.variableNames.push("preluActivationWeights"),a&&this.variableNames.push("leakyreluAlpha"),this.userCode=`
|
|
${c}
|
|
|
|
void main() {
|
|
ivec4 coords = getOutputCoords();
|
|
int batch = coords.x;
|
|
ivec2 xRCCorner = coords.yz * strides - pads;
|
|
int d2 = coords.w;
|
|
int d1 = d2 / ${s};
|
|
int q = d2 - d1 * ${s};
|
|
int xRCorner = xRCCorner.x;
|
|
int xCCorner = xRCCorner.y;
|
|
|
|
//intialize dotProd with a small epsilon seems to reduce GPU accuracy loss.
|
|
vec4 dotProd = vec4(0.000000000000001);
|
|
|
|
${p}
|
|
|
|
vec4 result = dotProd - vec4(0.000000000000001);
|
|
${m}
|
|
${f}
|
|
setOutput(result);
|
|
}
|
|
`}};function Tae(e){let{inputs:t,backend:r,attrs:n}=e,{x:a,filter:s}=t,{strides:i,pad:o,dilations:l,dimRoundingMode:u}=n,d=l;d==null&&(d=[1,1]),v.assert(C.eitherStridesOrDilationsAreOne(i,d),()=>`Error in depthwiseConv2d: Either strides or dilations must be 1. Got strides ${i} and dilations '${d}'`);let h=C.computeConv2DInfo(a.shape,s.shape,i,d,o,u,!0),p;Z().getBool("WEBGL_PACK_DEPTHWISECONV")&&h.strideWidth<=2&&h.outChannels/h.inChannels===1?p=new T9(h):p=new C9(h);let c=[[h.padInfo.top,h.padInfo.left],[h.strideHeight,h.strideWidth],[h.dilationHeight,h.dilationWidth],[h.inHeight,h.inWidth]];return r.runWebGLProgram(p,[a,s],"float32",c)}var Nae={kernelName:oi,backendName:"webgl",kernelFunc:Tae},Eae=class{constructor(e){this.variableNames=["x","dy"],this.outputShape=e.filterShape;let t=e.strideHeight,r=e.strideWidth,n=e.padInfo.top,a=e.padInfo.left,s=e.outChannels/e.inChannels;this.userCode=`
|
|
void main() {
|
|
ivec4 coords = getOutputCoords();
|
|
int wR = coords.x;
|
|
int wC = coords.y;
|
|
int d1 = coords.z;
|
|
int dm = coords.w;
|
|
int d2 = d1 * ${s} + dm;
|
|
|
|
float dotProd = 0.0;
|
|
|
|
// TO DO: Vec4 over the batch size
|
|
for (int b = 0; b < ${e.batchSize}; b++) {
|
|
for (int yR = 0; yR < ${e.outHeight}; yR++) {
|
|
int xR = wR + yR * ${t} - ${n};
|
|
|
|
if (xR < 0 || xR >= ${e.inHeight}) {
|
|
continue;
|
|
}
|
|
|
|
for (int yC = 0; yC < ${e.outWidth}; yC++) {
|
|
int xC = wC + yC * ${r} - ${a};
|
|
|
|
if (xC < 0 || xC >= ${e.inWidth}) {
|
|
continue;
|
|
}
|
|
|
|
float dyValue = getDy(b, yR, yC, d2);
|
|
float xValue = getX(b, xR, xC, d1);
|
|
dotProd += (xValue * dyValue);
|
|
}
|
|
}
|
|
}
|
|
setOutput(dotProd);
|
|
}
|
|
`}},Rae=class{constructor(e){this.variableNames=["dy","W"],this.outputShape=e.inShape;let t=e.filterHeight,r=e.filterWidth,n=e.strideHeight,a=e.strideWidth,s=t-1-e.padInfo.top,i=r-1-e.padInfo.left,o=e.outChannels/e.inChannels;this.userCode=`
|
|
const ivec2 pads = ivec2(${s}, ${i});
|
|
|
|
void main() {
|
|
ivec4 coords = getOutputCoords();
|
|
int batch = coords[0];
|
|
int d1 = coords[3];
|
|
ivec2 dyCorner = coords.yz - pads;
|
|
int dyRCorner = dyCorner.x;
|
|
int dyCCorner = dyCorner.y;
|
|
|
|
float dotProd = 0.0;
|
|
|
|
for (int wR = 0; wR < ${t}; wR++) {
|
|
float dyR = float(dyRCorner + wR) / ${n}.0;
|
|
|
|
if (dyR < 0.0 || dyR >= ${e.outHeight}.0 || fract(dyR) > 0.0) {
|
|
continue;
|
|
}
|
|
int idyR = int(dyR);
|
|
|
|
int wRPerm = ${t} - 1 - wR;
|
|
|
|
for (int wC = 0; wC < ${r}; wC++) {
|
|
float dyC = float(dyCCorner + wC) / ${a}.0;
|
|
|
|
if (dyC < 0.0 || dyC >= ${e.outWidth}.0 ||
|
|
fract(dyC) > 0.0) {
|
|
continue;
|
|
}
|
|
int idyC = int(dyC);
|
|
|
|
int wCPerm = ${r} - 1 - wC;
|
|
|
|
// TO DO: Vec4 over the channelMul
|
|
for (int dm = 0; dm < ${o}; dm++) {
|
|
int d2 = d1 * ${o} + dm;
|
|
float xValue = getDy(batch, idyR, idyC, d2);
|
|
float wValue = getW(wRPerm, wCPerm, d1, dm);
|
|
dotProd += xValue * wValue;
|
|
}
|
|
}
|
|
}
|
|
setOutput(dotProd);
|
|
}
|
|
`}};function $ae(e){let{inputs:t,backend:r,attrs:n}=e,{x:a,dy:s}=t,{strides:i,dilations:o,pad:l,dimRoundingMode:u,filterShape:d}=n,h=C.computeConv2DInfo(a.shape,d,i,o,l,u,!0),p=new Eae(h);return r.runWebGLProgram(p,[a,s],"float32")}var Mae={kernelName:of,backendName:"webgl",kernelFunc:$ae};function Fae(e){let{inputs:t,backend:r,attrs:n}=e,{dy:a,filter:s}=t,{strides:i,dilations:o,pad:l,dimRoundingMode:u,inputShape:d}=n,h=C.computeConv2DInfo(d,s.shape,i,o,l,u,!0),p=new Rae(h);return r.runWebGLProgram(p,[a,s],"float32")}var Pae={kernelName:lf,backendName:"webgl",kernelFunc:Fae},_ae=class{constructor(e){this.variableNames=["X"],this.outputShape=[e,e],this.userCode=`
|
|
void main() {
|
|
ivec2 coords = getOutputCoords();
|
|
float val = coords[0] == coords[1] ? getX(coords[0]) : 0.0;
|
|
setOutput(val);
|
|
}
|
|
`}};function zae(e){let{inputs:t,backend:r}=e,{x:n}=t,a=[...n.shape,...n.shape],s=v.sizeFromShape(n.shape),i=ve({inputs:{x:n},backend:r,attrs:{shape:[s]}}),o=new _ae(s),l=r.runWebGLProgram(o,[i],i.dtype),u=ve({inputs:{x:l},backend:r,attrs:{shape:a}});return r.disposeIntermediateTensorInfo(i),r.disposeIntermediateTensorInfo(l),u}var Oae={kernelName:uf,backendName:"webgl",kernelFunc:zae},Dae=class{constructor(e){this.variableNames=["x","W"],this.outputShape=e.outShape;let{inHeight:t,inWidth:r,padInfo:n,strideHeight:a,strideWidth:s,filterHeight:i,filterWidth:o,dilationHeight:l,dilationWidth:u}=e,{top:d,left:h}=n;this.userCode=`
|
|
const ivec2 strides = ivec2(${a}, ${s});
|
|
const ivec2 pads = ivec2(${d}, ${h});
|
|
const float neg_infinity = -3.4e38;
|
|
|
|
void main() {
|
|
ivec4 coords = getOutputCoords();
|
|
int batch = coords.x;
|
|
int d1 = coords.w;
|
|
ivec2 outTopLeftCorner =
|
|
coords.yz * strides - pads;
|
|
int hBeg = outTopLeftCorner.x;
|
|
int wBeg = outTopLeftCorner.y;
|
|
|
|
float curVal = neg_infinity;
|
|
for (int h = 0; h < ${i}; h++) {
|
|
int hIn = hBeg + h * ${l};
|
|
|
|
if (hIn >= 0 && hIn < ${t}) {
|
|
for (int w = 0; w < ${o}; w++) {
|
|
int wIn = wBeg + w * ${u};
|
|
|
|
if (wIn >= 0 && wIn < ${r}) {
|
|
float xVal = getX(batch, hIn, wIn, d1);
|
|
float wVal = getW(h, w, d1);
|
|
|
|
float val = xVal + wVal;
|
|
if (val > curVal) {
|
|
curVal = val;
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
float result = curVal;
|
|
setOutput(result);
|
|
}
|
|
`}};function Lae(e){let{inputs:t,backend:r,attrs:n}=e,{x:a,filter:s}=t,{strides:i,pad:o,dilations:l}=n,u=C.computeDilation2DInfo(a.shape,s.shape,i,o,"NHWC",l),d,h=new Dae(u);d=r.runWebGLProgram(h,[a,s],"float32");let p=ve({inputs:{x:d},backend:r,attrs:{shape:u.outShape}});return r.disposeIntermediateTensorInfo(d),p}var Bae={kernelName:th,backendName:"webgl",kernelFunc:Lae};function Wae(e){let{inputs:t,backend:r,attrs:n}=e,{equation:a}=n,s=t,{allDims:i,summedDims:o,idDims:l}=C.decodeEinsumEquation(a,s.length);C.checkEinsumDimSizes(i.length,l,s);let{path:u,steps:d}=C.getEinsumComputePath(o,l),h=d.length,p=null,c=i.length,f=[];for(let m=0;m<h;++m){for(let g of d[m]){let{permutationIndices:y,expandDims:A}=C.getEinsumPermutation(c,l[g]),x;C.isIdentityPermutation(y)?x=s[g]:(x=dr({inputs:{x:s[g]},backend:r,attrs:{perm:y}}),f.push(x));let b=x.shape.slice();for(let w=0;w<A.length;++w)b.splice(A[w],0,1);v.arraysEqual(x.shape,b)||(x=ve({inputs:{x},backend:r,attrs:{shape:b}}),f.push(x)),p===null?p=x:(p=k5({inputs:{a:x,b:p},backend:r}),f.push(p))}m<h-1&&(u[m]>=0&&(p=Em({inputs:{x:p},backend:r,attrs:{axis:u[m]-(i.length-c),keepDims:!1}}),f.push(p)),c--)}for(let m of f)m!==p&&r.disposeIntermediateTensorInfo(m);return p}var Vae={kernelName:rh,backendName:"webgl",kernelFunc:Wae},Uae="return (x >= 0.0) ? x : (exp(x) - 1.0);",Gae=`
|
|
vec4 result;
|
|
|
|
result.r = (x.r >= 0.0) ? x.r : (exp(x.r) - 1.0);
|
|
result.g = (x.g >= 0.0) ? x.g : (exp(x.g) - 1.0);
|
|
result.b = (x.b >= 0.0) ? x.b : (exp(x.b) - 1.0);
|
|
result.a = (x.a >= 0.0) ? x.a : (exp(x.a) - 1.0);
|
|
|
|
return result;
|
|
`,jae=it({opSnippet:Uae,packedOpSnippet:Gae}),Hae={kernelName:ui,backendName:"webgl",kernelFunc:jae},qae="return (b >= 1.0) ? a : a * (b + 1.0);",Xae=`
|
|
vec4 bGTEZero = vec4(greaterThanEqual(b, vec4(0.)));
|
|
return (bGTEZero * a) + ((vec4(1.0) - bGTEZero) * (a * (b + vec4(1.0))));
|
|
`,Kae=e=>{let{inputs:t,backend:r}=e,{dy:n,y:a}=t,s=Z().getBool("WEBGL_PACK_BINARY_OPERATIONS")?new Wh(Xae,n.shape,a.shape):new _u(qae,n.shape,a.shape);return r.runWebGLProgram(s,[n,a],n.dtype)},Zae={kernelName:df,backendName:"webgl",kernelFunc:Kae},Yae=`
|
|
return vec4(equal(a, b));
|
|
`,Jae="return float(a == b);",Qae=wr({opSnippet:Jae,packedOpSnippet:Yae,dtype:"bool",cpuKernelImpl:Dee}),ese={kernelName:Jo,backendName:"webgl",kernelFunc:Qae},tse=`
|
|
// Error function is calculated approximately with elementary function.
|
|
// See "Handbook of Mathematical Functions with Formulas,
|
|
// Graphs, and Mathematical Tables", Abramowitz and Stegun.
|
|
float p = ${C.ERF_P};
|
|
float a1 = ${C.ERF_A1};
|
|
float a2 = ${C.ERF_A2};
|
|
float a3 = ${C.ERF_A3};
|
|
float a4 = ${C.ERF_A4};
|
|
float a5 = ${C.ERF_A5};
|
|
|
|
float sign = sign(x);
|
|
x = abs(x);
|
|
float t = 1.0 / (1.0 + p * x);
|
|
return sign * (1.0 - (((((a5*t + a4)*t) + a3)*t + a2)*t + a1)*t*exp(-x*x));
|
|
`,rse=it({opSnippet:tse}),nse={kernelName:Xu,backendName:"webgl",kernelFunc:rse},ase=Ed+`
|
|
return exp(x);
|
|
`,sse=`
|
|
vec4 result = exp(x);
|
|
bvec4 isNaN = isnan(x);
|
|
result.r = isNaN.r ? x.r : result.r;
|
|
result.g = isNaN.g ? x.g : result.g;
|
|
result.b = isNaN.b ? x.b : result.b;
|
|
result.a = isNaN.a ? x.a : result.a;
|
|
|
|
return result;
|
|
`,N9=it({opSnippet:ase,packedOpSnippet:sse,cpuKernelImpl:Lee,dtype:"float32"}),ise={kernelName:di,backendName:"webgl",kernelFunc:N9};function iy(e){let{inputs:t,attrs:r,backend:n}=e,{dim:a}=r,{input:s}=t,i=s.shape.length,o=s.shape.slice(),l=a;return a<0&&(v.assert(-(i+1)<=a,()=>`Axis must be in the interval [${-(i+1)}, ${i}]`),l=i+a+1),o.splice(l,0,1),ve({inputs:{x:s},backend:n,attrs:{shape:o}})}var ose={kernelName:Qo,backendName:"webgl",kernelFunc:iy},Q4="return exp(x) - 1.0;",lse=it({opSnippet:Q4,packedOpSnippet:Q4,cpuKernelImpl:Bee}),use={kernelName:el,backendName:"webgl",kernelFunc:lse},ev=class{constructor(e,t,r){this.variableNames=["real","imag"];let n=t[1];this.outputShape=t;let a=r?`2.0 * ${Math.PI}`:`-2.0 * ${Math.PI}`,s=r?`${n}.0`:"1.0",i;if(e==="real")i="return real * expR - imag * expI;";else if(e==="imag")i="return real * expI + imag * expR;";else throw new Error(`FFT component must be either "real" or "imag", got ${e}.`);this.userCode=`
|
|
const float exponentMultiplier = ${a};
|
|
|
|
float unaryOpComplex(float real, float expR, float imag, float expI) {
|
|
${i}
|
|
}
|
|
|
|
float mulMatDFT(int batch, int index) {
|
|
float indexRatio = float(index) / float(${n});
|
|
float exponentMultiplierTimesIndexRatio =
|
|
exponentMultiplier * indexRatio;
|
|
|
|
float result = 0.0;
|
|
|
|
for (int i = 0; i < ${n}; i++) {
|
|
// x = (-2|2 * PI / N) * index * i;
|
|
float x = exponentMultiplierTimesIndexRatio * float(i);
|
|
float expR = cos(x);
|
|
float expI = sin(x);
|
|
float real = getReal(batch, i);
|
|
float imag = getImag(batch, i);
|
|
|
|
result +=
|
|
unaryOpComplex(real, expR, imag, expI) / ${s};
|
|
}
|
|
|
|
return result;
|
|
}
|
|
|
|
void main() {
|
|
ivec2 coords = getOutputCoords();
|
|
setOutput(mulMatDFT(coords[0], coords[1]));
|
|
}
|
|
`}};function E9(e,t,r){let n=r.texData.get(e.dataId),a=v.sizeFromShape(e.shape),s=e.shape[e.shape.length-1],i=a/s,o=ve({inputs:{x:e},backend:r,attrs:{shape:[i,s]}}),l=o.shape,u=new ev("real",l,t),d=new ev("imag",l,t),h=[{dataId:n.complexTensorInfos.real.dataId,dtype:n.complexTensorInfos.real.dtype,shape:l},{dataId:n.complexTensorInfos.imag.dataId,dtype:n.complexTensorInfos.imag.dtype,shape:l}],p=r.runWebGLProgram(u,h,"float32"),c=r.runWebGLProgram(d,h,"float32"),f=Xi({inputs:{real:p,imag:c},backend:r});r.disposeIntermediateTensorInfo(p),r.disposeIntermediateTensorInfo(c);let m=ve({inputs:{x:f},backend:r,attrs:{shape:e.shape}});return r.disposeIntermediateTensorInfo(o),r.disposeIntermediateTensorInfo(f),m}function dse(e){let{inputs:t,backend:r}=e,{input:n}=t;return E9(n,!1,r)}var pse={kernelName:pf,backendName:"webgl",kernelFunc:dse},hse=class{constructor(e,t){this.outputShape=[],this.customUniforms=[{name:"value",type:"float"}],this.variableNames=["x"],this.outputShape=e,this.userCode=`
|
|
void main() {
|
|
// Input can be obtained from uniform value.
|
|
setOutput(value);
|
|
}
|
|
`}};function Uh(e){let{backend:t,attrs:r}=e,{shape:n,value:a}=r,{dtype:s}=r;if(s=s||v.inferDtype(a),s==="string"){let i=v.getArrayFromDType(s,v.sizeFromShape(n));return i.fill(a),t.makeTensorInfo(n,s,i)}else{let i=new hse(n,a),o=[[a]];return t.runWebGLProgram(i,[],s,o)}}var cse={kernelName:Ku,backendName:"webgl",kernelFunc:Uh},fse=class{constructor(e){this.variableNames=["Image"],this.outputShape=[];let t=e[2];this.outputShape=e,this.userCode=`
|
|
void main() {
|
|
ivec4 coords = getOutputCoords();
|
|
int x = coords[2];
|
|
|
|
int coordX = ${t} - x - 1;
|
|
float outputValue;
|
|
if(coordX >= 0 && coordX < ${t}) {
|
|
outputValue = getImage(coords[0], coords[1], coordX, coords[3]);
|
|
} else {
|
|
outputValue = getImage(coords[0], coords[1], coords[2], coords[3]);
|
|
}
|
|
setOutput(outputValue);
|
|
}
|
|
`}},mse={kernelName:tl,backendName:"webgl",kernelFunc:({inputs:e,backend:t})=>{let{image:r}=e,n=t,a=new fse(r.shape);return n.runWebGLProgram(a,[r],r.dtype)}},tv="return floor(x);",gse=it({opSnippet:tv,packedOpSnippet:tv,cpuKernelImpl:Wee}),yse={kernelName:pi,backendName:"webgl",kernelFunc:gse},Ase=`
|
|
float s = sign(a) * sign(b);
|
|
int ia = round(a);
|
|
int ib = round(b);
|
|
if (ib != 0) {
|
|
// Windows (D3D) wants guaranteed non-zero int division at compile-time.
|
|
return float(idiv(ia, ib, s));
|
|
} else {
|
|
return NAN;
|
|
}
|
|
`,xse=`
|
|
ivec4 ia = round(a);
|
|
ivec4 ib = round(b);
|
|
bvec4 cond = notEqual(ib, ivec4(0));
|
|
ivec4 result = ivec4(0);
|
|
vec4 s = sign(a) * sign(b);
|
|
|
|
// Windows (D3D) wants guaranteed non-zero int division at compile-time.
|
|
if (cond[0]) {
|
|
result[0] = idiv(ia[0], ib[0], s[0]);
|
|
}
|
|
if (cond[1]) {
|
|
result[1] = idiv(ia[1], ib[1], s[1]);
|
|
}
|
|
if (cond[2]) {
|
|
result[2] = idiv(ia[2], ib[2], s[2]);
|
|
}
|
|
if (cond[3]) {
|
|
result[3] = idiv(ia[3], ib[3], s[3]);
|
|
}
|
|
return vec4(result);
|
|
`,bse=wr({opSnippet:Ase,packedOpSnippet:xse,dtype:"int32"}),vse={kernelName:hi,backendName:"webgl",kernelFunc:bse},wse=class{constructor(e){this.variableNames=["A"];let t=Hr(),[r,n]=e;this.outputShape=e,this.userCode=`
|
|
void main() {
|
|
ivec3 coords = getOutputCoords();
|
|
int texR = coords[0];
|
|
int texC = coords[1];
|
|
int depth = coords[2];
|
|
vec2 uv = (vec2(texC, texR) + halfCR) / vec2(${n}.0, ${r}.0);
|
|
|
|
vec4 values = ${t.texture2D}(A, uv);
|
|
float value;
|
|
if (depth == 0) {
|
|
value = values.r;
|
|
} else if (depth == 1) {
|
|
value = values.g;
|
|
} else if (depth == 2) {
|
|
value = values.b;
|
|
} else if (depth == 3) {
|
|
value = values.a;
|
|
}
|
|
|
|
setOutput(floor(value * 255.0 + 0.5));
|
|
}
|
|
`}},kse=class{constructor(e){this.variableNames=["A"],this.packedInputs=!1,this.packedOutput=!0;let t=Hr(),[r,n]=e;this.outputShape=e,this.userCode=`
|
|
void main() {
|
|
ivec3 coords = getOutputCoords();
|
|
int texR = coords[0];
|
|
int texC = coords[1];
|
|
int depth = coords[2];
|
|
|
|
vec4 result = vec4(0.);
|
|
|
|
for(int row=0; row<=1; row++) {
|
|
for(int col=0; col<=1; col++) {
|
|
texC = coords[1] + row;
|
|
depth = coords[2] + col;
|
|
|
|
vec2 uv = (vec2(texC, texR) + halfCR) /
|
|
vec2(${n}.0, ${r}.0);
|
|
vec4 values = ${t.texture2D}(A, uv);
|
|
float value;
|
|
if (depth == 0) {
|
|
value = values.r;
|
|
} else if (depth == 1) {
|
|
value = values.g;
|
|
} else if (depth == 2) {
|
|
value = values.b;
|
|
} else if (depth == 3) {
|
|
value = values.a;
|
|
}
|
|
|
|
result[row * 2 + col] = floor(value * 255.0 + 0.5);
|
|
}
|
|
}
|
|
|
|
${t.output} = result;
|
|
}
|
|
`}},Ise={kernelName:Op,backendName:"webgl",kernelFunc:Sse},pu;function Sse(e){let{inputs:t,backend:r,attrs:n}=e,{pixels:a}=t,{numChannels:s}=n,i=typeof HTMLVideoElement!="undefined"&&a instanceof HTMLVideoElement,o=typeof HTMLImageElement!="undefined"&&a instanceof HTMLImageElement,[l,u]=i?[a.videoWidth,a.videoHeight]:[a.width,a.height],d=[u,l],h=[u,l,s];(o||i)&&(pu==null&&(pu=document.createElement("canvas").getContext("2d")),pu.canvas.width=l,pu.canvas.height=u,pu.drawImage(a,0,0,l,u),a=pu.canvas);let p=r.makeTensorInfo(d,"int32");r.texData.get(p.dataId).usage=2,r.gpgpu.uploadPixelDataToTexture(r.getTexture(p.dataId),a);let c=Z().getBool("WEBGL_PACK")?new kse(h):new wse(h),f=r.runWebGLProgram(c,[p],"int32");return r.disposeData(p.dataId),f}function Cse(e){let{inputs:t,backend:r,attrs:n}=e,{x:a,filter:s,bias:i,preluActivationWeights:o}=t,{strides:l,pad:u,dataFormat:d,dilations:h,dimRoundingMode:p,activation:c,leakyreluAlpha:f}=n,m=C.convertConv2DDataFormat(d),g=C.computeConv2DInfo(a.shape,s.shape,l,h,u,p,!1,m),y,A=[];if(g.filterHeight===1&&g.filterWidth===1&&g.dilationHeight===1&&g.dilationWidth===1&&g.strideHeight===1&&g.strideWidth===1&&(g.padInfo.type==="SAME"||g.padInfo.type==="VALID"))y=k9({x:a,filter:s,convInfo:g,backend:r,bias:i,activation:c,preluActivationWeights:o,leakyreluAlpha:f});else if(Z().getBool("WEBGL_CONV_IM2COL")&&a.shape[0]===1)y=I9({x:a,filter:s,convInfo:g,backend:r,bias:i,activation:c,preluActivationWeights:o,leakyreluAlpha:f});else{let b=i!=null,w=o!=null,S=c==="leakyrelu",N=c?Tm(c,!1):null,E=new v9(g,b,N,w,S),R=[a,s];if(i&&R.push(i),o&&R.push(o),S){let z=r.makeTensorInfo([],"float32",v.createScalarValue(f,"float32"));R.push(z),A.push(z)}y=r.runWebGLProgram(E,R,"float32")}let x=ve({inputs:{x:y},backend:r,attrs:{shape:g.outShape}});return A.push(y),A.forEach(b=>r.disposeIntermediateTensorInfo(b)),x}var Tse={kernelName:_s,backendName:"webgl",kernelFunc:Cse};function Nse(e){let{inputs:t,backend:r,attrs:n}=e,{x:a,filter:s,bias:i,preluActivationWeights:o}=t,{strides:l,pad:u,dilations:d,dimRoundingMode:h,activation:p,leakyreluAlpha:c}=n,f=[],m=d;m==null&&(m=[1,1]),v.assert(C.eitherStridesOrDilationsAreOne(l,m),()=>`Error in depthwiseConv2d: Either strides or dilations must be 1. Got strides ${l} and dilations '${m}'`);let g=C.computeConv2DInfo(a.shape,s.shape,l,m,u,h,!0),y=Z().getBool("WEBGL_PACK_DEPTHWISECONV")&&g.strideWidth<=2&&g.outChannels/g.inChannels===1,A=p?Tm(p,y):null,x=[a,s],b=i!=null,w=o!=null,S=p==="leakyrelu";if(b&&x.push(i),w&&x.push(o),S){let z=r.makeTensorInfo([],"float32",v.createScalarValue(c,"float32"));x.push(z),f.push(z)}let N;y?N=new T9(g,b,A,w,S):N=new C9(g,b,A,w,S);let E=[[g.padInfo.top,g.padInfo.left],[g.strideHeight,g.strideWidth],[g.dilationHeight,g.dilationWidth],[g.inHeight,g.inWidth]],R=r.runWebGLProgram(N,x,"float32",E);return f.forEach(z=>r.disposeIntermediateTensorInfo(z)),R}var Ese={kernelName:zs,backendName:"webgl",kernelFunc:Nse},Rse=class{constructor(e,t,r){this.sliceDim=e,this.strides=t,this.variableNames=["x","indices"],this.outputShape=r;let n=vt(t.length),a=vt(r.length),s=this.sliceDim>1?"strides[j]":"strides";this.userCode=`
|
|
${n} strides = ${n}(${this.strides});
|
|
void main() {
|
|
${a} coords = getOutputCoords();
|
|
int flattenIndex = 0;
|
|
for (int j = 0; j < ${this.sliceDim}; j++) {
|
|
int index = round(getIndices(coords[0], j));
|
|
flattenIndex += index * ${s};
|
|
}
|
|
setOutput(getX(flattenIndex, coords[1]));
|
|
}
|
|
`}};function $se(e){let{inputs:t,backend:r}=e,{params:n,indices:a}=t,s=a.shape,i=s[s.length-1],o=v.sizeFromShape(n.shape),[l,u,d,h]=C.prepareAndValidate(n,a),p=ve({inputs:{x:a},backend:r,attrs:{shape:[u,i]}}),c=ve({inputs:{x:n},backend:r,attrs:{shape:[v.sizeFromShape(n.shape)/d,d]}});if(r.shouldExecuteOnCPU([n,a])||n.dtype==="string"){let y=r.readSync(a.dataId),A=r.bufferSync(n),x=Vee(y,A,n.dtype,u,i,d,h,n.shape,o);return r.makeTensorInfo(l,n.dtype,x.values)}let f=new Rse(i,h,[u,d]),m=r.runWebGLProgram(f,[c,p],c.dtype),g=ve({inputs:{x:m},backend:r,attrs:{shape:l}});return r.disposeIntermediateTensorInfo(p),r.disposeIntermediateTensorInfo(c),r.disposeIntermediateTensorInfo(m),g}var Mse={kernelName:nl,backendName:"webgl",kernelFunc:$se},Fse=class{constructor(e,t){this.variableNames=["A","indices"],this.outputShape=t,this.rank=t.length;let r=vt(this.rank),n=Pse(e,2);this.userCode=`
|
|
void main() {
|
|
${r} resRC = getOutputCoords();
|
|
int index = int(getIndices(resRC.x, resRC.z));
|
|
float inBounds = (index >= 0) && (index < ${e[2]}) ? 1.0 : 0.0;
|
|
setOutput(inBounds * getA(${n}));
|
|
}
|
|
`}};function Pse(e,t){let r=["resRC.x","resRC.y","resRC.z","resRC.w"],n=[];for(let a=0;a<e.length;a++)a===2?n.push("index"):n.push(`${r[a]}`);return n.join()}function R9(e){let{inputs:t,backend:r,attrs:n}=e,{x:a,indices:s}=t,{axis:i,batchDims:o}=n,l=v.parseAxisParam(i,a.shape)[0];if(Z().get("DEBUG")){let A=r.readSync(s.dataId),x=a.shape[l];for(let b=0;b<A.length;++b){let w=A[b];v.assert(w<=x-1&&w>=0,()=>`GatherV2: the index value ${w} is not in [0, ${x-1}]`)}}let u=C.segment_util.collectGatherOpShapeInfo(a,s,l,o),d=v.sizeFromShape(s.shape),h=[],p=ve({inputs:{x:a},backend:r,attrs:{shape:[u.batchSize,u.outerSize,u.dimSize,u.sliceSize]}}),c=ve({inputs:{x:s},backend:r,attrs:{shape:[u.batchSize,d/u.batchSize]}});h.push(p),h.push(c);let f=[u.batchSize,u.outerSize,d/u.batchSize,u.sliceSize];if(r.shouldExecuteOnCPU([a,s])||a.dtype==="string"){let A=r.bufferSync(c),x=r.bufferSync(p),b=Uee(x,A,f);return h.forEach(w=>r.disposeIntermediateTensorInfo(w)),r.makeTensorInfo(u.outputShape,b.dtype,b.values)}let m=new Fse(p.shape,f),g=r.runWebGLProgram(m,[p,c],p.dtype);h.push(g);let y=ve({inputs:{x:g},backend:r,attrs:{shape:u.outputShape}});return h.forEach(A=>r.disposeIntermediateTensorInfo(A)),y}var _se={kernelName:rl,backendName:"webgl",kernelFunc:R9},zse="return float(a > b);",Ose=`
|
|
return vec4(greaterThan(a, b));
|
|
`,Dse=wr({opSnippet:zse,packedOpSnippet:Ose,cpuKernelImpl:Gee,dtype:"bool"}),Lse={kernelName:al,backendName:"webgl",kernelFunc:Dse},Bse="return float(a >= b);",Wse=`
|
|
return vec4(greaterThanEqual(a, b));
|
|
`,Vse=wr({opSnippet:Bse,packedOpSnippet:Wse,dtype:"bool",cpuKernelImpl:jee}),Use={kernelName:fi,backendName:"webgl",kernelFunc:Vse};function Gse(e){let{inputs:t,backend:r}=e,{input:n}=t;return E9(n,!0,r)}var jse={kernelName:hf,backendName:"webgl",kernelFunc:Gse},Hse="return float(!isnan(x) && !isinf(x));",qse=it({opSnippet:Hse,dtype:"bool"}),Xse={kernelName:Zu,backendName:"webgl",kernelFunc:qse},Kse="return float(isinf(x));",Zse=it({opSnippet:Kse,dtype:"bool"}),Yse={kernelName:Yu,backendName:"webgl",kernelFunc:Zse},Jse="return float(isnan(x));",Qse=it({opSnippet:Jse,dtype:"bool"}),eie={kernelName:Ju,backendName:"webgl",kernelFunc:Qse},tie="return float(a < b);",rie=`
|
|
return vec4(lessThan(a, b));
|
|
`,nie=wr({opSnippet:tie,packedOpSnippet:rie,cpuKernelImpl:Hee,dtype:"bool"}),aie={kernelName:sl,backendName:"webgl",kernelFunc:nie},sie="return float(a <= b);",iie=`
|
|
return vec4(lessThanEqual(a, b));
|
|
`,oie=wr({opSnippet:sie,packedOpSnippet:iie,cpuKernelImpl:qee,dtype:"bool"}),lie={kernelName:il,backendName:"webgl",kernelFunc:oie};function uie(e){let{backend:t,attrs:r}=e,{start:n,stop:a,num:s}=r,i=Xee(n,a,s);return t.makeTensorInfo([i.length],"float32",i)}var die={kernelName:cf,backendName:"webgl",kernelFunc:uie},pie=Ed+`
|
|
return x < 0.0 ? 0./0. : log(x);
|
|
`,hie=`
|
|
vec4 result = log(x);
|
|
bvec4 isNaN = isnan(x);
|
|
result.r = isNaN.r ? x.r : (x.r < 0.0 ? 0./0. : result.r);
|
|
result.g = isNaN.g ? x.g : (x.g < 0.0 ? 0./0. : result.g);
|
|
result.b = isNaN.b ? x.b : (x.b < 0.0 ? 0./0. : result.b);
|
|
result.a = isNaN.a ? x.a : (x.a < 0.0 ? 0./0. : result.a);
|
|
return result;
|
|
`,cie=it({opSnippet:pie,packedOpSnippet:hie,cpuKernelImpl:Kee}),fie={kernelName:yi,backendName:"webgl",kernelFunc:cie},mie=Ed+`
|
|
return log(1.0 + x);
|
|
`,gie=it({opSnippet:mie}),yie={kernelName:Qu,backendName:"webgl",kernelFunc:gie},Aie="return float(a >= 1.0 && b >= 1.0);",xie=`
|
|
return vec4(
|
|
vec4(greaterThanEqual(a, vec4(1.0))) *
|
|
vec4(greaterThanEqual(b, vec4(1.0))));
|
|
`,bie=wr({opSnippet:Aie,packedOpSnippet:xie,dtype:"bool"}),vie={kernelName:ol,backendName:"webgl",kernelFunc:bie},wie="return float(!(x >= 1.0));",kie=it({opSnippet:wie}),Iie={kernelName:ed,backendName:"webgl",kernelFunc:kie},Sie="return float(a >= 1.0 || b >= 1.0);",Cie=`
|
|
return min(
|
|
vec4(greaterThanEqual(a, vec4(1.0))) +
|
|
vec4(greaterThanEqual(b, vec4(1.0))),
|
|
vec4(1.0));
|
|
`,Tie=wr({opSnippet:Sie,packedOpSnippet:Cie,dtype:"bool"}),Nie={kernelName:ah,backendName:"webgl",kernelFunc:Tie},Eie=class{constructor(e,t,r,n,a){this.variableNames=["x"],this.outputShape=[];let s=t,i=e[3]-1;this.outputShape=e;let o,l=`float(${r}) + float(${n}) * sum`;a===.5?o=`inversesqrt(${l})`:a===1?o=`1.0/(${l})`:o=`exp(log(${l}) * float(-${a}));`,this.userCode=`
|
|
void main() {
|
|
ivec4 coords = getOutputCoords();
|
|
int b = coords[0];
|
|
int r = coords[1];
|
|
int c = coords[2];
|
|
int d = coords[3];
|
|
float x = getX(b, r, c, d);
|
|
float sum = 0.0;
|
|
for (int j = -${s}; j <= ${s}; j++) {
|
|
int idx = d + j;
|
|
if (idx >= 0 && idx <= ${i}) {
|
|
float z = getX(b, r, c, idx);
|
|
sum += z * z;
|
|
}
|
|
}
|
|
float val = x * ${o};
|
|
setOutput(val);
|
|
}
|
|
`}},Rie=class{constructor(e,t,r,n,a){this.variableNames=["x"],this.outputShape=[],this.packedInputs=!0,this.packedOutput=!0;let s=t,i=e[3]-1;this.outputShape=e;let o,l=`float(${r}) + float(${n}) * sum`;a===.5?o=`inversesqrt(${l})`:a===1?o=`1.0/(${l})`:o=`exp(log(${l}) * float(-${a}));`,this.userCode=`
|
|
void main() {
|
|
ivec4 coords = getOutputCoords();
|
|
int b = coords.x;
|
|
int r = coords.y;
|
|
int c = coords.z;
|
|
int d = coords.w;
|
|
|
|
bool hasNextCol = d < ${this.outputShape[3]};
|
|
bool hasNextRow = c < ${this.outputShape[2]};
|
|
|
|
vec4 sum = vec4(0.);
|
|
vec4 xFragAtOutputCoords = getX(b, r, c, d);
|
|
|
|
vec4 xAtOutputCoords = vec4(
|
|
getChannel(xFragAtOutputCoords, vec2(c, d)),
|
|
hasNextCol ?
|
|
getChannel(xFragAtOutputCoords, vec2(c, d + 1)) : 0.0,
|
|
hasNextRow ?
|
|
getChannel(xFragAtOutputCoords , vec2(c + 1, d)) : 0.0,
|
|
(hasNextRow && hasNextCol) ?
|
|
getChannel(xFragAtOutputCoords, vec2(c + 1, d + 1)) : 0.0
|
|
);
|
|
|
|
int firstChannel = d - ${s};
|
|
vec2 cache = vec2(0.);
|
|
if(firstChannel >= 0){
|
|
vec4 firstChannelFrag = getX(b, r, c, firstChannel);
|
|
cache.x = getChannel(firstChannelFrag, vec2(c, firstChannel));
|
|
if(hasNextRow){
|
|
cache.y = getChannel(firstChannelFrag, vec2(c + 1, firstChannel));
|
|
}
|
|
}
|
|
|
|
ivec2 depth = ivec2(d, d + 1);
|
|
for (int j = - ${s}; j <= ${s}; j++) {
|
|
ivec2 idx = depth + j;
|
|
bvec2 aboveLowerBound = greaterThanEqual(idx, ivec2(0));
|
|
bvec2 belowUpperBound = lessThanEqual(idx, ivec2(${i}));
|
|
|
|
bool depthInRange = aboveLowerBound.x && belowUpperBound.x;
|
|
bool depthPlusOneInRange = aboveLowerBound.y && belowUpperBound.y;
|
|
|
|
if(depthInRange || depthPlusOneInRange){
|
|
vec4 z = vec4(0.);
|
|
vec4 xFragAtCurrentDepth;
|
|
z.xz = cache.xy;
|
|
if(depthPlusOneInRange && hasNextCol){
|
|
xFragAtCurrentDepth = idx.y != d ?
|
|
getX(b, r, c, idx.y) : xFragAtOutputCoords;
|
|
z.y = getChannel(xFragAtCurrentDepth, vec2(c, idx.y));
|
|
if(hasNextRow){
|
|
z.w = getChannel(xFragAtCurrentDepth, vec2(c + 1, idx.y));
|
|
}
|
|
}
|
|
cache.xy = z.yw;
|
|
sum += z * z;
|
|
}
|
|
}
|
|
vec4 result = xAtOutputCoords * ${o};
|
|
setOutput(result);
|
|
}
|
|
`}},$ie=e=>{let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{depthRadius:s,bias:i,alpha:o,beta:l}=n,u=Z().getBool("WEBGL_PACK_NORMALIZATION")?new Rie(a.shape,s,i,o,l):new Eie(a.shape,s,i,o,l);return r.runWebGLProgram(u,[a],a.dtype)},Mie={kernelName:sh,backendName:"webgl",kernelFunc:$ie},Fie=class{constructor(e,t,r,n,a){this.variableNames=["inputImage","outputImage","dy"],this.outputShape=[],this.outputShape=e,this.depth=e[3],this.depthRadius=t,this.bias=r,this.alpha=n,this.beta=a,this.userCode=`
|
|
void main() {
|
|
ivec4 coords = getOutputCoords();
|
|
int b = coords[0];
|
|
int r = coords[1];
|
|
int c = coords[2];
|
|
|
|
float result = 0.0;
|
|
for (int d = 0; d < ${this.depth}; ++d) {
|
|
int depthBegin = int(max(0.0, float(d - ${t})));
|
|
int depthEnd = int(min(float(${this.depth}),
|
|
float(d + ${t} + 1)));
|
|
|
|
const int MIN_DEPTH_BEGIN = 0;
|
|
const int MAX_DEPTH_END = ${this.depth};
|
|
|
|
float norm = 0.0;
|
|
for (int k = MIN_DEPTH_BEGIN; k < MAX_DEPTH_END; ++k) {
|
|
if (k < depthBegin){
|
|
continue;
|
|
}
|
|
else if (k >= depthBegin && k < depthEnd) {
|
|
norm += getInputImage(b, r, c, k) * getInputImage(b, r, c, k);
|
|
}
|
|
else {
|
|
break;
|
|
}
|
|
}
|
|
|
|
norm = float(${n}) * norm + float(${r});
|
|
|
|
for(int k = MIN_DEPTH_BEGIN; k < MAX_DEPTH_END; ++k){
|
|
if (k < depthBegin){
|
|
continue;
|
|
}
|
|
else if (k >= depthBegin && k < depthEnd){
|
|
float dyi = -2.0 * float(${n})
|
|
* float(${a})
|
|
* getInputImage(b ,r ,c, k) * getOutputImage(b, r, c, d)
|
|
/ norm;
|
|
if (k == d) {
|
|
dyi += pow(norm, -1.0 * ${a});
|
|
}
|
|
if (k == coords[3]) {
|
|
dyi *= getDy(b, r, c, d);
|
|
result += dyi;
|
|
}
|
|
}
|
|
else {
|
|
break;
|
|
}
|
|
}
|
|
}
|
|
setOutput(result);
|
|
}
|
|
`}},Pie=e=>{let{inputs:t,backend:r,attrs:n}=e,{x:a,y:s,dy:i}=t,{depthRadius:o,bias:l,alpha:u,beta:d}=n,h=new Fie(a.shape,o,l,u,d);return r.runWebGLProgram(h,[a,s,i],a.dtype)},_ie={kernelName:ff,backendName:"webgl",kernelFunc:Pie};function zie(e,t,r,n){let a=v.sizeFromShape(t),s=v.sizeFromShape(e.shape)/a,i=ve({inputs:{x:e},attrs:{shape:[s,a]},backend:n}),o=Bl(i,e.dtype,"max",n),l=ve({inputs:{x:o},attrs:{shape:r},backend:n});return n.disposeIntermediateTensorInfo(i),n.disposeIntermediateTensorInfo(o),l}function $9(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{reductionIndices:s,keepDims:i}=n,o=a.shape.length,l=v.parseAxisParam(s,a.shape),u=l,d=C.getAxesPermutation(u,o),h=d!=null,p=r.shouldExecuteOnCPU([a]),c=a;if(h){if(p){let A=r.texData.get(c.dataId).values,x=new Array(o);for(let S=0;S<x.length;S++)x[S]=a.shape[d[S]];let b=w5(A,a.shape,a.dtype,d,x);c=r.makeTensorInfo(x,a.dtype);let w=r.texData.get(c.dataId);w.values=b}else c=Nm(a,d,r);u=C.getInnerMostAxes(u.length,o)}C.assertAxesAreInnerMostDims("max",u,o);let[f,m]=C.computeOutAndReduceShapes(c.shape,u),g=f;i&&(g=C.expandShapeToKeepDim(f,l));let y;if(p){let A=r.texData.get(c.dataId).values,x=Zee(A,v.sizeFromShape(m),g,a.dtype);y=r.makeTensorInfo(g,a.dtype);let b=r.texData.get(y.dataId);b.values=x}else y=zie(c,m,g,r);return h&&r.disposeIntermediateTensorInfo(c),y}var Oie={kernelName:Ai,backendName:"webgl",kernelFunc:$9},Die=u9+`
|
|
return max(a, b);
|
|
`,Lie=`
|
|
vec4 result = vec4(max(a, b));
|
|
vec4 isNaN = min(vec4(isnan(a)) + vec4(isnan(b)), vec4(1.0));
|
|
`+Cm+`
|
|
return result;
|
|
`,Bie=wr({opSnippet:Die,packedOpSnippet:Lie,cpuKernelImpl:Yee}),Wie={kernelName:xi,backendName:"webgl",kernelFunc:Bie};function Vie(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t;Id(a,"maxPool");let{filterSize:s,strides:i,pad:o,dimRoundingMode:l}=n,u=1;v.assert(C.eitherStridesOrDilationsAreOne(i,u),()=>`Error in maxPool: Either strides or dilations must be 1. Got strides ${i} and dilations '${u}'`);let d=C.computePool2DInfo(a.shape,s,i,u,o,l);if(d.filterWidth===1&&d.filterHeight===1&&v.arraysEqual(d.inShape,d.outShape))return fn({inputs:{x:a},backend:r});let h=new Kp(d,"max",!1);return r.runWebGLProgram(h,[a],a.dtype)}var Uie={kernelName:bi,backendName:"webgl",kernelFunc:Vie};function Gie(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{filterSize:s,strides:i,pad:o,dataFormat:l,dimRoundingMode:u}=n,d=[1,1,1],h=C.computePool3DInfo(a.shape,s,i,d,o,u,l),p=new I5(h,"max",!1);return r.runWebGLProgram(p,[a],a.dtype)}var jie={kernelName:ih,backendName:"webgl",kernelFunc:Gie},Hie=class{constructor(e){this.variableNames=["dy","maxPos"],this.outputShape=e.inShape;let t=e.strideHeight,r=e.strideWidth,n=e.dilationHeight,a=e.effectiveFilterHeight,s=e.effectiveFilterWidth,i=a-1-e.padInfo.top,o=s-1-e.padInfo.left,l=a*s-1;this.userCode=`
|
|
const ivec2 pads = ivec2(${i}, ${o});
|
|
|
|
void main() {
|
|
ivec4 coords = getOutputCoords();
|
|
int b = coords[0];
|
|
int d = coords[3];
|
|
|
|
ivec2 dyRCCorner = coords.yz - pads;
|
|
int dyRCorner = dyRCCorner.x;
|
|
int dyCCorner = dyRCCorner.y;
|
|
|
|
// Convolve dy(?, ?, d) with pos mask(:, :, d) to get dx(xR, xC, d).
|
|
// ? = to be determined. : = across all values in that axis.
|
|
float dotProd = 0.0;
|
|
for (int wR = 0; wR < ${a};
|
|
wR += ${n}) {
|
|
float dyR = float(dyRCorner + wR) / ${t}.0;
|
|
|
|
if (dyR < 0.0 || dyR >= ${e.outHeight}.0 || fract(dyR) > 0.0) {
|
|
continue;
|
|
}
|
|
int idyR = int(dyR);
|
|
|
|
for (int wC = 0; wC < ${s}; wC++) {
|
|
float dyC = float(dyCCorner + wC) / ${r}.0;
|
|
|
|
if (dyC < 0.0 || dyC >= ${e.outWidth}.0 ||
|
|
fract(dyC) > 0.0) {
|
|
continue;
|
|
}
|
|
int idyC = int(dyC);
|
|
|
|
float dyValue = getDy(b, idyR, idyC, d);
|
|
int maxPosValue = ${l} - int(getMaxPos(b, idyR, idyC, d));
|
|
|
|
// Get the current value, check it against the value from the
|
|
// position matrix.
|
|
int curPosValue = wR * ${s} + wC;
|
|
float mask = float(maxPosValue == curPosValue ? 1.0 : 0.0);
|
|
|
|
dotProd += dyValue * mask;
|
|
}
|
|
}
|
|
setOutput(dotProd);
|
|
}
|
|
`}},qie=class{constructor(e){this.variableNames=["dy","maxPos"],this.outputShape=e.inShape;let t=e.strideDepth,r=e.strideHeight,n=e.strideWidth,a=e.dilationDepth,s=e.dilationHeight,i=e.dilationWidth,o=e.effectiveFilterDepth,l=e.effectiveFilterHeight,u=e.effectiveFilterWidth,d=o-1-e.padInfo.front,h=l-1-e.padInfo.top,p=u-1-e.padInfo.left,c=o*l*u-1;this.userCode=`
|
|
const ivec3 pads = ivec3(${d}, ${h}, ${p});
|
|
|
|
void main() {
|
|
ivec5 coords = getOutputCoords();
|
|
int batch = coords.x;
|
|
int ch = coords.u;
|
|
|
|
ivec3 dyCorner = ivec3(coords.y, coords.z, coords.w) - pads;
|
|
int dyDCorner = dyCorner.x;
|
|
int dyRCorner = dyCorner.y;
|
|
int dyCCorner = dyCorner.z;
|
|
|
|
// Convolve dy(?, ?, ?, ch) with pos mask(:, :, :, d) to get
|
|
// dx(xD, xR, xC, ch).
|
|
// ? = to be determined. : = across all values in that axis.
|
|
float dotProd = 0.0;
|
|
|
|
for (int wD = 0; wD < ${o};
|
|
wD += ${a}) {
|
|
float dyD = float(dyDCorner + wD) / ${t}.0;
|
|
|
|
if (dyD < 0.0 || dyD >= ${e.outDepth}.0 || fract(dyD) > 0.0) {
|
|
continue;
|
|
}
|
|
int idyD = int(dyD);
|
|
|
|
for (int wR = 0; wR < ${l};
|
|
wR += ${s}) {
|
|
float dyR = float(dyRCorner + wR) / ${r}.0;
|
|
|
|
if (dyR < 0.0 || dyR >= ${e.outHeight}.0 ||
|
|
fract(dyR) > 0.0) {
|
|
continue;
|
|
}
|
|
int idyR = int(dyR);
|
|
|
|
for (int wC = 0; wC < ${u};
|
|
wC += ${i}) {
|
|
float dyC = float(dyCCorner + wC) / ${n}.0;
|
|
|
|
if (dyC < 0.0 || dyC >= ${e.outWidth}.0 ||
|
|
fract(dyC) > 0.0) {
|
|
continue;
|
|
}
|
|
int idyC = int(dyC);
|
|
|
|
float dyValue = getDy(batch, idyD, idyR, idyC, ch);
|
|
int maxPosValue = ${c} -
|
|
int(getMaxPos(batch, idyD, idyR, idyC, ch));
|
|
|
|
// Get the current value, check it against the value from the
|
|
// position matrix.
|
|
int curPosValue =
|
|
wD * ${l} * ${u} +
|
|
wR * ${u} + wC;
|
|
float mask = float(maxPosValue == curPosValue ? 1.0 : 0.0);
|
|
|
|
dotProd += dyValue * mask;
|
|
}
|
|
}
|
|
}
|
|
setOutput(dotProd);
|
|
}
|
|
`}};function Xie(e){let{inputs:t,backend:r,attrs:n}=e,{dy:a,input:s}=t,i=s,{filterSize:o,strides:l,pad:u,dimRoundingMode:d}=n,h=[1,1,1],p=C.computePool3DInfo(i.shape,o,l,h,u,d),c=new I5(p,"max",!0),f=r.runWebGLProgram(c,[i],i.dtype),m=new qie(p),g=r.runWebGLProgram(m,[a,f],i.dtype);return r.disposeIntermediateTensorInfo(f),g}var Kie={kernelName:gf,backendName:"webgl",kernelFunc:Xie};function Zie(e){let{inputs:t,backend:r,attrs:n}=e,{dy:a,input:s,output:i}=t,o=s;Id([s,i],"maxPoolGrad");let{filterSize:l,strides:u,pad:d,dimRoundingMode:h}=n,p=C.computePool2DInfo(o.shape,l,u,1,d,h),c=!0,f=new Kp(p,"max",c),m=r.runWebGLProgram(f,[o],o.dtype),g=new Hie(p),y=r.runWebGLProgram(g,[a,m],o.dtype);return r.disposeIntermediateTensorInfo(m),y}var Yie={kernelName:mf,backendName:"webgl",kernelFunc:Zie};function Jie(e,t,r,n){let a=new Kp(r,"max",!1),s=n.runWebGLProgram(a,[e],"float32");a=new Kp(r,"max",!0,!0,t);let i=n.runWebGLProgram(a,[e],"float32");return[s,i]}var Qie={kernelName:yf,backendName:"webgl",kernelFunc:({inputs:e,attrs:t,backend:r})=>{let{x:n}=e,{filterSize:a,strides:s,pad:i,includeBatchInIndex:o}=t,l=r;v.assert(n.shape.length===4,()=>`Error in maxPool: input must be rank 4 but got rank ${n.shape.length}.`);let u=[1,1];v.assert(C.eitherStridesOrDilationsAreOne(s,u),()=>`Error in maxPool: Either strides or dilations must be 1. Got strides ${s} and dilations '${u}'`);let d=C.computePool2DInfo(n.shape,a,s,u,i),[h,p]=Jie(n,o,d,l);return[h,p]}};function eoe(e,t,r,n){let a=v.sizeFromShape(t),s=v.sizeFromShape(e.shape)/a,i=ve({inputs:{x:e},attrs:{shape:[s,a]},backend:n}),o=Bl(i,"float32","mean",n),l=ve({inputs:{x:o},attrs:{shape:r},backend:n});return n.disposeIntermediateTensorInfo(i),n.disposeIntermediateTensorInfo(o),l}var toe={kernelName:vi,backendName:"webgl",kernelFunc:({inputs:e,attrs:t,backend:r})=>{let{x:n}=e,{keepDims:a,axis:s}=t,i=r,o=n.shape.length,l=v.parseAxisParam(s,n.shape),u=l,d=C.getAxesPermutation(u,o),h=d!=null,p=i.shouldExecuteOnCPU([n]),c=[],f=n;if(h){if(p){let x=i.texData.get(f.dataId).values,b=new Array(o);for(let N=0;N<b.length;N++)b[N]=n.shape[d[N]];let w=w5(x,n.shape,n.dtype,d,b);f=i.makeTensorInfo(b,n.dtype);let S=i.texData.get(f.dataId);S.values=w}else f=Nm(n,d,i);c.push(f),u=C.getInnerMostAxes(u.length,o)}C.assertAxesAreInnerMostDims("sum",u,o);let[m,g]=C.computeOutAndReduceShapes(f.shape,u),y=m;a&&(y=C.expandShapeToKeepDim(m,l));let A=eoe(f,g,y,i);for(let x of c)i.disposeIntermediateTensorInfo(x);return A}};function roe(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{axis:s,keepDims:i}=n,o=a.shape.length,l=v.parseAxisParam(s,a.shape),u=l,d=C.getAxesPermutation(u,o),h=a;d!=null&&(h=dr({inputs:{x:a},backend:r,attrs:{perm:d}}),u=C.getInnerMostAxes(u.length,a.shape.length)),C.assertAxesAreInnerMostDims("min",u,o);let[p,c]=C.computeOutAndReduceShapes(h.shape,u),f=v.sizeFromShape(c),m=ve({inputs:{x:h},backend:r,attrs:{shape:[-1,f]}}),g=Bl(m,m.dtype,"min",r),y;if(i){let A=C.expandShapeToKeepDim(p,l);y=ve({inputs:{x:g},backend:r,attrs:{shape:A}})}else y=ve({inputs:{x:g},backend:r,attrs:{shape:p}});return r.disposeIntermediateTensorInfo(m),r.disposeIntermediateTensorInfo(g),d!=null&&r.disposeIntermediateTensorInfo(h),y}var noe={kernelName:wi,backendName:"webgl",kernelFunc:roe},aoe=u9+`
|
|
return min(a, b);
|
|
`,soe=`
|
|
vec4 result = vec4(min(a, b));
|
|
vec4 isNaN = min(vec4(isnan(a)) + vec4(isnan(b)), vec4(1.0));
|
|
`+Cm+`
|
|
return result;
|
|
`,ioe=wr({opSnippet:aoe,packedOpSnippet:soe,cpuKernelImpl:Jee}),ooe={kernelName:ki,backendName:"webgl",kernelFunc:ioe},loe=class{constructor(e,t,r){this.variableNames=["x"],this.outputShape=t.map((u,d)=>u[0]+e[d]+u[1]);let n=e.length,a=vt(n),s=t.map(u=>u[0]).join(","),i=t.map((u,d)=>u[0]+e[d]).join(","),o=["coords[0]","coords[1]","coords[2]","coords[3]"].slice(0,n),l=r==="reflect"?0:1;if(n===1){this.userCode=`
|
|
int start = ${s};
|
|
int end = ${i};
|
|
|
|
void main() {
|
|
int outC = getOutputCoords();
|
|
if (outC < start) {
|
|
outC = start * 2 - outC - ${l};
|
|
} else if(outC >= end) {
|
|
outC = (end - 1) * 2 - outC + ${l};
|
|
}
|
|
setOutput(getX(outC - start));
|
|
}
|
|
`;return}this.userCode=`
|
|
${a} start = ${a}(${s});
|
|
${a} end = ${a}(${i});
|
|
|
|
void main() {
|
|
${a} outC = getOutputCoords();
|
|
for (int i = 0; i < ${n}; i++) {
|
|
if (outC[i] < start[i]) {
|
|
outC[i] = start[i] * 2 - outC[i] - ${l};
|
|
} else if(outC[i] >= end[i]) {
|
|
outC[i] = (end[i] - 1) * 2 - outC[i] + ${l};
|
|
}
|
|
}
|
|
${a} coords = outC - start;
|
|
setOutput(getX(${o}));
|
|
}
|
|
`}},uoe=class{constructor(e,t,r){this.variableNames=["x"],this.packedInputs=!0,this.packedOutput=!0,this.outputShape=t.map((c,f)=>c[0]+e[f]+c[1]);let n=e.length,a=vt(n),s=t.map(c=>c[0]).join(","),i=t.map((c,f)=>c[0]+e[f]).join(","),o=Br("rc",n),l=Br("source",n),u=`${o[n-1]} < ${this.outputShape[n-1]}`,d=n===1?"source":`vec2(${l.slice(-2).join()})`,h=r==="reflect"?0:1,p="";if(n===1){let c=`
|
|
${a} source = rc;
|
|
if (source < start) {
|
|
source = start * 2 - source - ${h};
|
|
} else if (source >= end) {
|
|
source = (end - 1) * 2 - source + ${h};
|
|
}
|
|
source -= start;
|
|
`;p=`
|
|
${a} rc = outputLoc;
|
|
${c}
|
|
result[0] = getChannel(getX(${l.join()}), ${d});
|
|
${o[n-1]} += 1;
|
|
if(${u}) {
|
|
${c}
|
|
result[1] = getChannel(getX(${l.join()}), ${d});
|
|
}
|
|
`}else{let c=`
|
|
${a} source = rc;
|
|
${a} lt = ${a}(lessThan(source, start));
|
|
${a} gte = ${a}(greaterThanEqual(source, end));
|
|
${a} orig = 1 - (lt + gte);
|
|
source = orig * source +
|
|
lt * (start * 2 - source - ${h}) +
|
|
gte * ((end - 1) * 2 - source + ${h});
|
|
source -= start;
|
|
`;p=`
|
|
${a} rc = outputLoc;
|
|
${c}
|
|
result[0] = getChannel(getX(${l.join()}), ${d});
|
|
${o[n-1]} += 1;
|
|
if(${u}) {
|
|
${c}
|
|
result[1] = getChannel(getX(${l.join()}), ${d});
|
|
}
|
|
rc = outputLoc;
|
|
${o[n-2]} += 1;
|
|
if(${o[n-2]} < ${this.outputShape[n-2]}) {
|
|
${c}
|
|
result[2] = getChannel(getX(${l.join()}), ${d});
|
|
${o[n-1]} += 1;
|
|
if(${u}) {
|
|
${c}
|
|
result[3] = getChannel(getX(${l.join()}), ${d});
|
|
}
|
|
}
|
|
`}this.userCode=`
|
|
const ${a} start = ${a}(${s});
|
|
const ${a} end = ${a}(${i});
|
|
|
|
void main() {
|
|
${a} outputLoc = getOutputCoords();
|
|
vec4 result = vec4(0.);
|
|
${p}
|
|
setOutput(result);
|
|
}
|
|
`}},doe=({inputs:e,backend:t,attrs:r})=>{let{x:n}=e,{paddings:a,mode:s}=r,i=Z().getBool("WEBGL_PACK_ARRAY_OPERATIONS")?new uoe(n.shape,a,s):new loe(n.shape,a,s);return t.runWebGLProgram(i,[n],n.dtype)},poe={kernelName:Ii,backendName:"webgl",kernelFunc:doe},hoe=`if (b == 0.0) return NAN;
|
|
return mod(a, b);`,coe=`
|
|
vec4 result = mod(a, b);
|
|
vec4 isNaN = vec4(equal(b, vec4(0.0)));
|
|
`+Cm+`
|
|
return result;
|
|
`,foe=wr({opSnippet:hoe,packedOpSnippet:coe}),moe={kernelName:td,backendName:"webgl",kernelFunc:foe},goe=class{constructor(e,t,r){this.variableNames=["probs"],this.customUniforms=[{name:"seed",type:"float"}],this.outputShape=[e,r],this.userCode=`
|
|
void main() {
|
|
ivec2 coords = getOutputCoords();
|
|
int batch = coords[0];
|
|
|
|
float r = random(seed);
|
|
float cdf = 0.0;
|
|
|
|
for (int i = 0; i < ${t-1}; i++) {
|
|
cdf += getProbs(batch, i);
|
|
|
|
if (r < cdf) {
|
|
setOutput(float(i));
|
|
return;
|
|
}
|
|
}
|
|
|
|
// If no other event happened, last event happened.
|
|
setOutput(float(${t-1}));
|
|
}
|
|
`}},yoe=`
|
|
if (a == b) {
|
|
return 1.0;
|
|
};
|
|
return a / b;`,Aoe=`
|
|
// vec4 one = vec4(equal(a, b));
|
|
// return one + (vec4(1.0) - one) * a / b;
|
|
vec4 result = a / b;
|
|
if(a.x == b.x) {
|
|
result.x = 1.;
|
|
}
|
|
if(a.y == b.y) {
|
|
result.y = 1.;
|
|
}
|
|
if(a.z == b.z) {
|
|
result.z = 1.;
|
|
}
|
|
if(a.w == b.w) {
|
|
result.w = 1.;
|
|
}
|
|
|
|
return result;
|
|
`,M9=wr({opSnippet:yoe,packedOpSnippet:Aoe,checkOutOfBounds:!0}),xoe={kernelName:li,backendName:"webgl",kernelFunc:M9},rv="return a - b;",F9=wr({opSnippet:rv,packedOpSnippet:rv,supportsComplex:!0,cpuKernelImpl:fte}),boe={kernelName:Bi,backendName:"webgl",kernelFunc:F9};function P9(e){let{inputs:t,backend:r,attrs:n}=e,{logits:a}=t,{dim:s}=n,i=v.parseAxisParam([s],a.shape),o=$9({inputs:{x:a},backend:r,attrs:{reductionIndices:i,keepDims:!1}}),l=C.expandShapeToKeepDim(o.shape,i),u=ve({inputs:{x:o},backend:r,attrs:{shape:l}}),d=F9({inputs:{a,b:u},backend:r}),h=N9({inputs:{x:d},backend:r}),p=Em({inputs:{x:h},backend:r,attrs:{axis:i,keepDims:!1}}),c=ve({inputs:{x:p},backend:r,attrs:{shape:l}}),f=M9({inputs:{a:h,b:c},backend:r});return r.disposeIntermediateTensorInfo(o),r.disposeIntermediateTensorInfo(u),r.disposeIntermediateTensorInfo(d),r.disposeIntermediateTensorInfo(h),r.disposeIntermediateTensorInfo(p),r.disposeIntermediateTensorInfo(c),f}var voe={kernelName:Di,backendName:"webgl",kernelFunc:P9};function woe(e){let{inputs:t,backend:r,attrs:n}=e,{logits:a}=t,{numSamples:s,seed:i,normalized:o}=n,l=o?a:P9({inputs:{logits:a},backend:r,attrs:{dim:a.shape.length-1}}),u=l.shape[0],d=l.shape[1],h=new goe(u,d,s),p=[[i]],c=r.runWebGLProgram(h,[l],"int32",p);return o||r.disposeIntermediateTensorInfo(l),c}var koe={kernelName:Af,backendName:"webgl",kernelFunc:woe},Ioe=Yn+`
|
|
return -x;
|
|
`,Soe=`
|
|
vec4 result = -x;
|
|
bvec4 isNaN = isnan(x);
|
|
|
|
result.r = isNaN.r ? x.r : result.r;
|
|
result.g = isNaN.g ? x.g : result.g;
|
|
result.b = isNaN.b ? x.b : result.b;
|
|
result.a = isNaN.a ? x.a : result.a;
|
|
|
|
return result;
|
|
`;function Coe(e){let{inputs:t,backend:r}=e,{x:n}=t;if(r.shouldExecuteOnCPU([n])){let s=r.texData.get(n.dataId),[i,o]=ete(s.values,n.shape,n.dtype);return r.makeTensorInfo(o,n.dtype,i)}let a;return Z().getBool("WEBGL_PACK_UNARY_OPERATIONS")?a=new No(n.shape,Soe):a=new Xa(n.shape,Ioe),r.runWebGLProgram(a,[n],n.dtype)}var Toe={kernelName:ll,backendName:"webgl",kernelFunc:Coe},Noe=Kn.nonMaxSuppressionV3Impl;function Eoe(e){C.warn("tf.nonMaxSuppression() in webgl locks the UI thread. Call tf.nonMaxSuppressionAsync() instead");let{inputs:t,backend:r,attrs:n}=e,{boxes:a,scores:s}=t,{maxOutputSize:i,iouThreshold:o,scoreThreshold:l}=n,u=r.readSync(a.dataId),d=r.readSync(s.dataId),{selectedIndices:h}=Noe(u,d,i,o,l);return r.makeTensorInfo([h.length],"int32",new Int32Array(h))}var Roe={kernelName:dl,backendName:"webgl",kernelFunc:Eoe},$oe=Kn.nonMaxSuppressionV4Impl;function Moe(e){C.warn("tf.nonMaxSuppression() in webgl locks the UI thread. Call tf.nonMaxSuppressionAsync() instead");let{inputs:t,backend:r,attrs:n}=e,{boxes:a,scores:s}=t,{maxOutputSize:i,iouThreshold:o,scoreThreshold:l,padToMaxOutputSize:u}=n,d=r.readSync(a.dataId),h=r.readSync(s.dataId),{selectedIndices:p,validOutputs:c}=$oe(d,h,i,o,l,u);return[r.makeTensorInfo([p.length],"int32",new Int32Array(p)),r.makeTensorInfo([],"int32",new Int32Array([c]))]}var Foe={kernelName:rd,backendName:"webgl",kernelFunc:Moe},Poe=Kn.nonMaxSuppressionV5Impl;function _oe(e){C.warn("tf.nonMaxSuppression() in webgl locks the UI thread. Call tf.nonMaxSuppressionAsync() instead");let{inputs:t,backend:r,attrs:n}=e,{boxes:a,scores:s}=t,{maxOutputSize:i,iouThreshold:o,scoreThreshold:l,softNmsSigma:u}=n,d=r.readSync(a.dataId),h=r.readSync(s.dataId),p=i,c=o,f=l,m=u,{selectedIndices:g,selectedScores:y}=Poe(d,h,p,c,f,m);return[r.makeTensorInfo([g.length],"int32",new Int32Array(g)),r.makeTensorInfo([y.length],"float32",new Float32Array(y))]}var zoe={kernelName:pl,backendName:"webgl",kernelFunc:_oe},Ooe=class{constructor(e,t,r,n){this.variableNames=["indices"],this.outputShape=[e,t],this.userCode=`
|
|
void main() {
|
|
ivec2 coords = getOutputCoords();
|
|
int index = round(getIndices(coords.x));
|
|
setOutput(mix(float(${n}), float(${r}),
|
|
float(index == coords.y)));
|
|
}
|
|
`}},Doe=e=>{let{inputs:t,backend:r,attrs:n}=e,{indices:a}=t,{depth:s,onValue:i,offValue:o}=n,l=v.sizeFromShape(a.shape),u=new Ooe(l,s,i,o),d=ve({inputs:{x:a},backend:r,attrs:{shape:[l]}}),h=r.runWebGLProgram(u,[d],a.dtype);r.disposeIntermediateTensorInfo(d);let p=[...a.shape,s],c=ve({inputs:{x:h},backend:r,attrs:{shape:p}});return r.disposeIntermediateTensorInfo(h),c},Loe={kernelName:cl,backendName:"webgl",kernelFunc:Doe};function j0(e){let{inputs:t,backend:r}=e,{x:n}=t;if(n.dtype==="complex64"){let a=Vh({inputs:{input:n},backend:r}),s=j0({inputs:{x:a},backend:r}),i=Rm({inputs:{input:n},backend:r}),o=j0({inputs:{x:i},backend:r}),l=Xi({inputs:{real:s,imag:o},backend:r});return r.disposeIntermediateTensorInfo(a),r.disposeIntermediateTensorInfo(s),r.disposeIntermediateTensorInfo(i),r.disposeIntermediateTensorInfo(o),l}else return Uh({attrs:{shape:n.shape,dtype:n.dtype,value:n.dtype==="string"?"":0},backend:r})}var Boe={kernelName:El,backendName:"webgl",kernelFunc:j0};function _9(e){let{inputs:t,backend:r}=e,{x:n}=t;if(n.dtype==="string")throw new Error("onesLike is not supported under string dtype");if(n.dtype==="complex64"){let a=Vh({inputs:{input:n},backend:r}),s=_9({inputs:{x:a},backend:r}),i=Rm({inputs:{input:n},backend:r}),o=j0({inputs:{x:i},backend:r}),l=Xi({inputs:{real:s,imag:o},backend:r});return r.disposeIntermediateTensorInfo(a),r.disposeIntermediateTensorInfo(s),r.disposeIntermediateTensorInfo(i),r.disposeIntermediateTensorInfo(o),l}else return Uh({attrs:{shape:n.shape,dtype:n.dtype,value:1},backend:r})}var Woe={kernelName:hl,backendName:"webgl",kernelFunc:_9};function Voe(e){let{inputs:t,backend:r,attrs:n}=e,{axis:a}=n;if(t.length===1)return iy({inputs:{input:t[0]},backend:r,attrs:{dim:a}});let s=t[0].shape,i=t[0].dtype;t.forEach(d=>{v.assertShapesMatch(s,d.shape,"All tensors passed to stack must have matching shapes"),v.assert(i===d.dtype,()=>"All tensors passed to stack must have matching dtypes")});let o=[],l=t.map(d=>{let h=iy({inputs:{input:d},backend:r,attrs:{dim:a}});return o.push(h),h}),u=b9({inputs:l,backend:r,attrs:{axis:a}});return o.forEach(d=>r.disposeIntermediateTensorInfo(d)),u}var Uoe={kernelName:fl,backendName:"webgl",kernelFunc:Voe},Goe=class{constructor(e,t,r){this.variableNames=["x"],this.customUniforms=[{name:"value",type:"float"}],this.outputShape=t.map((l,u)=>l[0]+e[u]+l[1]);let n=e.length,a=vt(n),s=t.map(l=>l[0]).join(","),i=t.map((l,u)=>l[0]+e[u]).join(","),o=["coords[0]","coords[1]","coords[2]","coords[3]"].slice(0,n);if(n===1){this.userCode=`
|
|
int start = ${s};
|
|
int end = ${i};
|
|
|
|
void main() {
|
|
int outC = getOutputCoords();
|
|
if (outC < start || outC >= end) {
|
|
setOutput(value);
|
|
} else {
|
|
setOutput(getX(outC - start));
|
|
}
|
|
}
|
|
`;return}this.userCode=`
|
|
${a} start = ${a}(${s});
|
|
${a} end = ${a}(${i});
|
|
|
|
void main() {
|
|
${a} outC = getOutputCoords();
|
|
if (any(lessThan(outC, start)) || any(greaterThanEqual(outC, end))) {
|
|
setOutput(value);
|
|
} else {
|
|
${a} coords = outC - start;
|
|
setOutput(getX(${o}));
|
|
}
|
|
}
|
|
`}},joe=class{constructor(e,t,r){this.variableNames=["x"],this.packedInputs=!0,this.packedOutput=!0,this.customUniforms=[{name:"value",type:"float"}],this.outputShape=t.map((f,m)=>f[0]+e[m]+f[1]);let n=e.length,a=vt(n),s=t.map(f=>f[0]).join(","),i=t.map((f,m)=>f[0]+e[m]).join(","),o=Br("rc",n),l=Br("source",n),u=`${o[n-1]} < ${this.outputShape[n-1]}`,d=n===1?"source":`vec2(${l.slice(-2).join()})`,h=[`${a} rc = outputLoc;`,`${o[n-1]} += 1;
|
|
if(${u}) {
|
|
`,n===1?"":`}
|
|
rc = outputLoc;
|
|
${o[n-2]} += 1;
|
|
if(${o[n-2]} < ${this.outputShape[n-2]}) {`,n===1?"":` ${o[n-1]} += 1;
|
|
if(${u}) {`],p=n===1?"rc < start || rc >= end":"any(lessThan(rc, start)) || any(greaterThanEqual(rc, end))",c="";for(let f=0,m=n===1?2:4;f<m;f++)c+=`
|
|
${h[f]}
|
|
if (${p}) {
|
|
result[${f}] = float(value);
|
|
} else {
|
|
${a} source = rc - start;
|
|
result[${f}] = getChannel(getX(${l.join()}), ${d});
|
|
}
|
|
`;c+=n===1?"} ":"}}",this.userCode=`
|
|
const ${a} start = ${a}(${s});
|
|
const ${a} end = ${a}(${i});
|
|
|
|
void main() {
|
|
${a} outputLoc = getOutputCoords();
|
|
vec4 result = vec4(0.);
|
|
${c}
|
|
setOutput(result);
|
|
}
|
|
`}},z9=e=>{let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{paddings:s,constantValue:i}=n;if(v.sizeFromShape(a.shape)===0){let u=s.map((d,h)=>d[0]+a.shape[h]+d[1]);return Uh({backend:r,attrs:{shape:u,value:i,dtype:a.dtype}})}let o=Z().getBool("WEBGL_PACK_ARRAY_OPERATIONS")?new joe(a.shape,s,i):new Goe(a.shape,s,i),l=[[i]];return r.runWebGLProgram(o,[a],a.dtype,l)},Hoe={kernelName:Ci,backendName:"webgl",kernelFunc:z9},qoe=`
|
|
if(a < 0.0 && floor(b) < b){
|
|
return NAN;
|
|
}
|
|
if (b == 0.0) {
|
|
return 1.0;
|
|
}
|
|
return (round(mod(b, 2.0)) != 1) ?
|
|
pow(abs(a), b) : sign(a) * pow(abs(a), b);
|
|
`,Xoe=`
|
|
// isModRound1 has 1 for components with round(mod(b, 2.0)) == 1, 0 otherwise.
|
|
vec4 isModRound1 = vec4(equal(round(mod(b, 2.0)), ivec4(1)));
|
|
vec4 multiplier = sign(a) * isModRound1 + (vec4(1.0) - isModRound1);
|
|
vec4 result = multiplier * pow(abs(a), b);
|
|
|
|
// Ensure that a^0 = 1, including 0^0 = 1 as this correspond to TF and JS
|
|
bvec4 isExpZero = equal(b, vec4(0.0));
|
|
result.r = isExpZero.r ? 1.0 : result.r;
|
|
result.g = isExpZero.g ? 1.0 : result.g;
|
|
result.b = isExpZero.b ? 1.0 : result.b;
|
|
result.a = isExpZero.a ? 1.0 : result.a;
|
|
|
|
vec4 isNaN = vec4(lessThan(a, vec4(0.0))) * vec4(lessThan(floor(b), b));
|
|
`+Cm+`
|
|
return result;
|
|
`,Koe=wr({opSnippet:qoe,packedOpSnippet:Xoe}),Zoe={kernelName:Ti,backendName:"webgl",kernelFunc:Koe};function Yoe(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{axis:s,keepDims:i}=n,o=a.shape.length,l=[],u=v.parseAxisParam(s,a.shape),d=u,h=C.getAxesPermutation(d,o),p=a;h!=null&&(p=dr({inputs:{x:a},backend:r,attrs:{perm:h}}),d=C.getInnerMostAxes(d.length,o),l.push(p)),C.assertAxesAreInnerMostDims("prod",d,o);let c;if(r.shouldExecuteOnCPU([p])){let f=r.texData.get(p.dataId).values,{outVals:m,outShape:g,outDtype:y}=rte(p.shape,p.dtype,f,d);c=r.makeTensorInfo(g,y,m)}else{let[f,m]=C.computeOutAndReduceShapes(p.shape,d),g=v.sizeFromShape(m),y=ve({inputs:{x:p},backend:r,attrs:{shape:[-1,g]}}),A=gh(a.dtype),x=Bl(y,A,"prod",r);c=ve({inputs:{x},backend:r,attrs:{shape:f}}),l.push(y),l.push(x)}if(i){l.push(c);let f=C.expandShapeToKeepDim(c.shape,u);c=ve({inputs:{x:c},backend:r,attrs:{shape:f}})}return l.forEach(f=>r.disposeIntermediateTensorInfo(f)),c}var Joe={kernelName:Ei,backendName:"webgl",kernelFunc:Yoe},O9=e=>{let{backend:t,attrs:r}=e,{start:n,stop:a,step:s,dtype:i}=r,o=nte(n,a,s,i);return t.makeTensorInfo([o.length],i,o)},Qoe={kernelName:nd,backendName:"webgl",kernelFunc:O9},ele="return 1.0 / x;",tle=it({opSnippet:ele}),rle={kernelName:ad,backendName:"webgl",kernelFunc:tle},nle=Yn+`
|
|
return (x < 0.0) ? 0.0 : x;
|
|
`,ale=`
|
|
vec4 result = x * vec4(greaterThanEqual(x, vec4(0.0)));
|
|
bvec4 isNaN = isnan(x);
|
|
|
|
result.r = isNaN.r ? x.r : result.r;
|
|
result.g = isNaN.g ? x.g : result.g;
|
|
result.b = isNaN.b ? x.b : result.b;
|
|
result.a = isNaN.a ? x.a : result.a;
|
|
|
|
return result;
|
|
`,sle=it({opSnippet:nle,packedOpSnippet:ale}),ile={kernelName:Ri,backendName:"webgl",kernelFunc:sle},ole=Yn+`
|
|
return (x < 0.0) ? 0.0 : min(6.0, x);
|
|
`,lle=`
|
|
vec4 result = min(x, vec4(6.)) * vec4(greaterThanEqual(x, vec4(0.0)));
|
|
bvec4 isNaN = isnan(x);
|
|
|
|
result.r = isNaN.r ? x.r : result.r;
|
|
result.g = isNaN.g ? x.g : result.g;
|
|
result.b = isNaN.b ? x.b : result.b;
|
|
result.a = isNaN.a ? x.a : result.a;
|
|
|
|
return result;
|
|
`,ule=it({opSnippet:ole,packedOpSnippet:lle}),dle={kernelName:Mi,backendName:"webgl",kernelFunc:ule},ple=class{constructor(e,t,r,n,a){this.variableNames=["A"],this.outputShape=[];let[s,i,o,l]=e;this.outputShape=[s,t,r,l];let u=[n&&t>1?i-1:i,n&&r>1?o-1:o],d=[n&&t>1?t-1:t,n&&r>1?r-1:r],h;a?h="(vec2(yRC) + vec2(0.5)) * effectiveInputOverOutputRatioRC - vec2(0.5)":h="vec2(yRC) * effectiveInputOverOutputRatioRC",this.userCode=`
|
|
const vec2 effectiveInputOverOutputRatioRC = vec2(
|
|
${u[0]/d[0]},
|
|
${u[1]/d[1]});
|
|
const vec2 inputShapeRC = vec2(${i}.0, ${o}.0);
|
|
|
|
void main() {
|
|
ivec4 coords = getOutputCoords();
|
|
int b = coords[0];
|
|
int d = coords[3];
|
|
ivec2 yRC = coords.yz;
|
|
|
|
// Fractional source index.
|
|
vec2 sourceFracIndexRC = ${h};
|
|
|
|
// Compute the four integer indices.
|
|
ivec2 sourceFloorRC = ivec2(max(sourceFracIndexRC, vec2(0.0)));
|
|
ivec2 sourceCeilRC = ivec2(
|
|
min(inputShapeRC - 1.0, ceil(sourceFracIndexRC)));
|
|
|
|
float topLeft = getA(b, sourceFloorRC.x, sourceFloorRC.y, d);
|
|
float bottomLeft = getA(b, sourceCeilRC.x, sourceFloorRC.y, d);
|
|
float topRight = getA(b, sourceFloorRC.x, sourceCeilRC.y, d);
|
|
float bottomRight = getA(b, sourceCeilRC.x, sourceCeilRC.y, d);
|
|
|
|
vec2 fracRC = sourceFracIndexRC - vec2(sourceFloorRC);
|
|
|
|
float top = topLeft + (topRight - topLeft) * fracRC.y;
|
|
float bottom = bottomLeft + (bottomRight - bottomLeft) * fracRC.y;
|
|
float newValue = top + (bottom - top) * fracRC.x;
|
|
|
|
setOutput(newValue);
|
|
}
|
|
`}},hle=class{constructor(e,t,r,n,a){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0,this.outputShape=[];let[s,i,o,l]=e;this.outputShape=[s,t,r,l];let u=[n&&t>1?i-1:i,n&&r>1?o-1:o],d=[n&&t>1?t-1:t,n&&r>1?r-1:r],h;a?h="(vec3(yRC) + vec3(0.5)) * effectiveInputOverOutputRatioRC - vec3(0.5)":h="vec3(yRC) * effectiveInputOverOutputRatioRC",this.userCode=`
|
|
const vec3 effectiveInputOverOutputRatioRC = vec3(
|
|
${u[0]/d[0]},
|
|
${u[1]/d[1]},
|
|
${u[1]/d[1]});
|
|
const vec3 inputShapeRC = vec3(${i}.0, ${o}.0,
|
|
${o}.0);
|
|
|
|
float getAValue(int b, int r, int c, int d) {
|
|
return getChannel(getA(b, r, c, d), vec2(c, d));
|
|
}
|
|
|
|
void main() {
|
|
ivec4 coords = getOutputCoords();
|
|
int b = coords[0];
|
|
int d = coords[3];
|
|
// Calculate values for next column in yRC.z.
|
|
ivec3 yRC = coords.yzz + ivec3(0, 0, 1);
|
|
|
|
// Fractional source index.
|
|
vec3 sourceFracIndexRC = ${h};
|
|
|
|
// Compute the four integer indices.
|
|
ivec3 sourceFloorRC = ivec3(max(sourceFracIndexRC, vec3(0.0)));
|
|
ivec3 sourceCeilRC = ivec3(
|
|
min(inputShapeRC - 1.0, ceil(sourceFracIndexRC)));
|
|
|
|
// Should we calculate next column and row elements in 2x2 packed cell.
|
|
bool hasNextCol = d < ${l-1};
|
|
bool hasNextRow = coords.z < ${r-1};
|
|
|
|
// In parallel, construct four corners for all four components in
|
|
// packed 2x2 cell.
|
|
vec4 topLeft = vec4(
|
|
getAValue(b, sourceFloorRC.x, sourceFloorRC.y, d),
|
|
hasNextCol ? getAValue(b, sourceFloorRC.x, sourceFloorRC.y, d + 1)
|
|
: 0.0,
|
|
hasNextRow ? getAValue(b, sourceFloorRC.x, sourceFloorRC.z, d)
|
|
: 0.0,
|
|
(hasNextRow && hasNextCol) ?
|
|
getAValue(b, sourceFloorRC.x, sourceFloorRC.z, d + 1) : 0.0);
|
|
|
|
vec4 bottomLeft = vec4(
|
|
getAValue(b, sourceCeilRC.x, sourceFloorRC.y, d),
|
|
hasNextCol ? getAValue(b, sourceCeilRC.x, sourceFloorRC.y, d + 1)
|
|
: 0.0,
|
|
hasNextRow ? getAValue(b, sourceCeilRC.x, sourceFloorRC.z, d)
|
|
: 0.0,
|
|
(hasNextRow && hasNextCol) ?
|
|
getAValue(b, sourceCeilRC.x, sourceFloorRC.z, d + 1) : 0.0);
|
|
|
|
vec4 topRight = vec4(
|
|
getAValue(b, sourceFloorRC.x, sourceCeilRC.y, d),
|
|
hasNextCol ? getAValue(b, sourceFloorRC.x, sourceCeilRC.y, d + 1)
|
|
: 0.0,
|
|
hasNextRow ? getAValue(b, sourceFloorRC.x, sourceCeilRC.z, d)
|
|
: 0.0,
|
|
(hasNextRow && hasNextCol) ?
|
|
getAValue(b, sourceFloorRC.x, sourceCeilRC.z, d + 1) : 0.0);
|
|
|
|
vec4 bottomRight = vec4(
|
|
getAValue(b, sourceCeilRC.x, sourceCeilRC.y, d),
|
|
hasNextCol ? getAValue(b, sourceCeilRC.x, sourceCeilRC.y, d + 1)
|
|
: 0.0,
|
|
hasNextRow ? getAValue(b, sourceCeilRC.x, sourceCeilRC.z, d)
|
|
: 0.0,
|
|
(hasNextRow && hasNextCol) ?
|
|
getAValue(b, sourceCeilRC.x, sourceCeilRC.z, d + 1) : 0.0);
|
|
|
|
vec3 fracRC = sourceFracIndexRC - vec3(sourceFloorRC);
|
|
|
|
vec4 top = mix(topLeft, topRight, fracRC.yyzz);
|
|
vec4 bottom = mix(bottomLeft, bottomRight, fracRC.yyzz);
|
|
vec4 newValue = mix(top, bottom, fracRC.x);
|
|
|
|
setOutput(newValue);
|
|
}
|
|
`}};function cle(e){let{inputs:t,backend:r,attrs:n}=e,{images:a}=t,{alignCorners:s,halfPixelCenters:i,size:o}=n,[l,u]=o,d=Z().getBool("WEBGL_PACK_IMAGE_OPERATIONS")?new hle(a.shape,l,u,s,i):new ple(a.shape,l,u,s,i);return r.runWebGLProgram(d,[a],"float32")}var fle={kernelName:$i,backendName:"webgl",kernelFunc:cle},mle=class{constructor(e,t,r){this.variableNames=["dy"],this.outputShape=[],this.outputShape=t;let[,n,a]=t,[,s,i]=e,o=[r&&s>1?n-1:n,r&&i>1?a-1:a],l=[r&&s>1?s-1:s,r&&i>1?i-1:i],u=o[0]/l[0],d=o[1]/l[1],h=1/u,p=1/d,c=Math.ceil(h)*2+2,f=Math.ceil(p)*2+2;this.userCode=`
|
|
void main() {
|
|
ivec4 coords = getOutputCoords();
|
|
int b = coords[0];
|
|
int d = coords[3];
|
|
int r = coords[1];
|
|
int c = coords[2];
|
|
|
|
float accumulator = 0.0;
|
|
|
|
const float heightScale = float(${u});
|
|
const float widthScale = float(${d});
|
|
|
|
const float invHeightScale = float(${h});
|
|
const float invWidthScale = float(${p});
|
|
|
|
const int winHeight = int(${c});
|
|
const int winWidth = int(${f});
|
|
|
|
// Compute bounds for where in dy we will look
|
|
float startRLerp = floor(float(r) * invHeightScale);
|
|
int startDyR = int(startRLerp - float(winHeight / 2));
|
|
|
|
float startCLerp = floor(float(c) * invWidthScale);
|
|
int startDyC = int(startCLerp - float(winWidth / 2));
|
|
|
|
// Loop over dy
|
|
for (int dyROffset = 0; dyROffset < winHeight; dyROffset++) {
|
|
int dyR = dyROffset + startDyR;
|
|
|
|
// Guard against the window exceeding the bounds of dy
|
|
if (dyR < 0 || dyR >= ${s}) {
|
|
continue;
|
|
}
|
|
|
|
for (int dyCOffset = 0; dyCOffset < winWidth; dyCOffset++) {
|
|
int dyC = dyCOffset + startDyC;
|
|
|
|
// Guard against the window exceeding the bounds of dy
|
|
if (dyC < 0 || dyC >= ${i}) {
|
|
continue;
|
|
}
|
|
|
|
float dxR = float(dyR) * heightScale;
|
|
int topDxRIndex = int(floor(dxR));
|
|
int bottomDxRIndex = int(min(ceil(dxR), ${n-1}.0));
|
|
float dxRLerp = dxR - float(topDxRIndex);
|
|
float inverseDxRLerp = 1.0 - dxRLerp;
|
|
|
|
float dxC = float(dyC) * widthScale;
|
|
int leftDxCIndex = int(floor(dxC));
|
|
int rightDxCIndex = int(min(ceil(dxC), ${a-1}.0));
|
|
float dxCLerp = dxC - float(leftDxCIndex);
|
|
float inverseDxCLerp = 1.0 - dxCLerp;
|
|
|
|
if (r == topDxRIndex && c == leftDxCIndex) {
|
|
// topLeft
|
|
accumulator +=
|
|
getDy(b, dyR, dyC, d) * inverseDxRLerp * inverseDxCLerp;
|
|
}
|
|
|
|
if (r == topDxRIndex && c == rightDxCIndex) {
|
|
// topRight
|
|
accumulator += getDy(b, dyR, dyC, d) * inverseDxRLerp * dxCLerp;
|
|
}
|
|
|
|
if (r == bottomDxRIndex && c == leftDxCIndex) {
|
|
// bottomLeft
|
|
accumulator += getDy(b, dyR, dyC, d) * dxRLerp * inverseDxCLerp;
|
|
}
|
|
|
|
if (r == bottomDxRIndex && c == rightDxCIndex) {
|
|
// bottomRight
|
|
accumulator += getDy(b, dyR, dyC, d) * dxRLerp * dxCLerp;
|
|
}
|
|
}
|
|
}
|
|
// End loop over dy
|
|
|
|
setOutput(accumulator);
|
|
}
|
|
`}};function gle(e){let{inputs:t,backend:r,attrs:n}=e,{images:a,dy:s}=t,{alignCorners:i}=n,o=new mle(s.shape,a.shape,i);return r.runWebGLProgram(o,[s],s.dtype)}var yle={kernelName:bf,backendName:"webgl",kernelFunc:gle},Ale=class{constructor(e,t,r,n,a){this.variableNames=["A"],this.outputShape=[];let[s,i,o,l]=e;this.outputShape=[s,t,r,l];let u=[n&&t>1?i-1:i,n&&r>1?o-1:o],d=[n&&t>1?t-1:t,n&&r>1?r-1:r],h=n?"0.5":"0.0",p;a?p="max((vec2(yRC) + vec2(0.5)) * effectiveInputOverOutputRatioRC, vec2(0.0))":p="vec2(yRC) * effectiveInputOverOutputRatioRC",this.userCode=`
|
|
const vec2 effectiveInputOverOutputRatioRC = vec2(
|
|
${u[0]/d[0]},
|
|
${u[1]/d[1]});
|
|
const vec2 inputShapeRC = vec2(${i}.0, ${o}.0);
|
|
|
|
void main() {
|
|
ivec4 coords = getOutputCoords();
|
|
int b = coords[0];
|
|
int d = coords[3];
|
|
ivec2 yRC = coords.yz;
|
|
|
|
// Fractional source index.
|
|
vec2 sourceFracIndexRC = ${p};
|
|
|
|
// Compute the coordinators of nearest neighbor point.
|
|
ivec2 sourceNearestRC = ivec2(
|
|
min(inputShapeRC - 1.0, floor(sourceFracIndexRC + ${h})));
|
|
float newValue = getA(b, sourceNearestRC.x, sourceNearestRC.y, d);
|
|
|
|
setOutput(newValue);
|
|
}
|
|
`}},xle=class{constructor(e,t,r,n,a){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0,this.outputShape=[];let[s,i,o,l]=e;this.outputShape=[s,t,r,l];let u=[n&&t>1?i-1:i,n&&r>1?o-1:o],d=[n&&t>1?t-1:t,n&&r>1?r-1:r],h=n?"0.5":"0.0",p;a?p="max((vec3(yRC) + vec3(0.5)) * effectiveInputOverOutputRatioRC, vec3(0.0))":p="vec3(yRC) * effectiveInputOverOutputRatioRC",this.userCode=`
|
|
const vec3 effectiveInputOverOutputRatioRC = vec3(
|
|
${u[0]/d[0]},
|
|
${u[1]/d[1]},
|
|
${u[1]/d[1]});
|
|
const vec3 inputShapeRC = vec3(${i}.0, ${o}.0,
|
|
${o}.0);
|
|
|
|
float getAValue(int b, int r, int c, int d) {
|
|
return getChannel(getA(b, r, c, d), vec2(c, d));
|
|
}
|
|
|
|
void main() {
|
|
ivec4 coords = getOutputCoords();
|
|
int b = coords[0];
|
|
int d = coords[3];
|
|
// Calculate values for next column in yRC.z.
|
|
ivec3 yRC = coords.yzz + ivec3(0, 0, 1);
|
|
|
|
// Fractional source index.
|
|
vec3 sourceFracIndexRC = ${p};
|
|
|
|
// Compute the coordinators of nearest neighbor point.
|
|
ivec3 sourceNearestRC = ivec3(
|
|
min(inputShapeRC - 1.0, floor(sourceFracIndexRC + ${h})));
|
|
|
|
// Should we calculate next column and row elements in 2x2 packed cell.
|
|
bool hasNextCol = d < ${l-1};
|
|
bool hasNextRow = coords.z < ${r-1};
|
|
|
|
vec4 newValue = vec4(
|
|
getAValue(b, sourceNearestRC.x, sourceNearestRC.y, d),
|
|
hasNextCol ? getAValue(b, sourceNearestRC.x, sourceNearestRC.y, d + 1)
|
|
: 0.0,
|
|
hasNextRow ? getAValue(b, sourceNearestRC.x, sourceNearestRC.z, d)
|
|
: 0.0,
|
|
(hasNextRow && hasNextCol) ?
|
|
getAValue(b, sourceNearestRC.x, sourceNearestRC.z, d + 1) : 0.0);
|
|
|
|
setOutput(newValue);
|
|
}
|
|
`}};function ble(e){let{inputs:t,backend:r,attrs:n}=e,{images:a}=t,{alignCorners:s,halfPixelCenters:i,size:o}=n,[l,u]=o,d=Z().getBool("WEBGL_PACK_IMAGE_OPERATIONS")?new xle(a.shape,l,u,s,i):new Ale(a.shape,l,u,s,i);return r.runWebGLProgram(d,[a],a.dtype)}var vle={kernelName:sd,backendName:"webgl",kernelFunc:ble},wle=class{constructor(e,t,r){this.variableNames=["dy"],this.outputShape=[],this.outputShape=t;let[,n,a]=t,[,s,i]=e,o=[r&&s>1?n-1:n,r&&i>1?a-1:a],l=[r&&s>1?s-1:s,r&&i>1?i-1:i],u=o[0]/l[0],d=o[1]/l[1],h=1/u,p=1/d,c=Math.ceil(h)*2+2,f=Math.ceil(p)*2+2;this.userCode=`
|
|
void main() {
|
|
ivec4 coords = getOutputCoords();
|
|
int b = coords[0];
|
|
int d = coords[3];
|
|
int r = coords[1];
|
|
int c = coords[2];
|
|
|
|
float accumulator = 0.0;
|
|
|
|
const float heightScale = float(${u});
|
|
const float widthScale = float(${d});
|
|
|
|
const float invHeightScale = float(${h});
|
|
const float invWidthScale = float(${p});
|
|
|
|
const int winHeight = int(${c});
|
|
const int winWidth = int(${f});
|
|
|
|
// Compute bounds for where in dy we will look
|
|
float startRLerp = floor(float(r) * invHeightScale);
|
|
int startDyR = int(floor(startRLerp - float(winHeight / 2)));
|
|
|
|
float startCLerp = floor(float(c) * invWidthScale);
|
|
int startDyC = int(floor(startCLerp - float(winWidth / 2)));
|
|
|
|
// Loop over dy
|
|
for (int dyROffset = 0; dyROffset < winHeight; dyROffset++) {
|
|
int dyR = dyROffset + startDyR;
|
|
|
|
// Guard against the window exceeding the bounds of dy
|
|
if (dyR < 0 || dyR >= ${s}) {
|
|
continue;
|
|
}
|
|
|
|
for (int dyCOffset = 0; dyCOffset < winWidth; dyCOffset++) {
|
|
int dyC = dyCOffset + startDyC;
|
|
|
|
// Guard against the window exceeding the bounds of dy
|
|
if (dyC < 0 || dyC >= ${i}) {
|
|
continue;
|
|
}
|
|
|
|
float sourceFracRow =
|
|
float(${o[0]}) *
|
|
(float(dyR) / float(${l[0]}));
|
|
|
|
float sourceFracCol =
|
|
float(${o[1]}) *
|
|
(float(dyC) / float(${l[1]}));
|
|
|
|
int sourceNearestRow = int(min(
|
|
float(int(${n}) - 1),
|
|
${r} ? float(round(sourceFracRow)) :
|
|
float(floor(sourceFracRow))));
|
|
|
|
int sourceNearestCol = int(min(
|
|
float(int(${a}) - 1),
|
|
${r} ? float(round(sourceFracCol)) :
|
|
float(floor(sourceFracCol))));
|
|
|
|
if (r == sourceNearestRow && c == sourceNearestCol) {
|
|
accumulator += getDy(b, dyR, dyC, d);
|
|
}
|
|
}
|
|
}
|
|
// End loop over dy
|
|
|
|
setOutput(accumulator);
|
|
}
|
|
`}};function kle(e){let{inputs:t,backend:r,attrs:n}=e,{images:a,dy:s}=t,{alignCorners:i}=n,o=new wle(s.shape,a.shape,i);return r.runWebGLProgram(o,[s],s.dtype)}var Ile={kernelName:xf,backendName:"webgl",kernelFunc:kle},Sle=class{constructor(e,t){this.variableNames=["x"];let r=e.length;if(r>4)throw new Error(`WebGL backend: Reverse of rank-${r} tensor is not yet supported`);if(this.outputShape=e,r===1){this.userCode=`
|
|
void main() {
|
|
int coord = getOutputCoords();
|
|
setOutput(getX(${e[0]} - coord - 1));
|
|
}
|
|
`;return}let n=i=>t.indexOf(i)!==-1&&e[i]!==1?`${e[i]} - coords[${i}] - 1`:`coords[${i}]`,a=e.map((i,o)=>n(o)).join(","),s=vt(r);this.userCode=`
|
|
void main() {
|
|
${s} coords = getOutputCoords();
|
|
setOutput(getX(${a}));
|
|
}
|
|
`}},Cle=class{constructor(e,t){this.variableNames=["x"],this.packedInputs=!0,this.packedOutput=!0;let r=e.length;if(r>4)throw new Error(`WebGL backend: Reverse of rank-${r} tensor is not yet supported`);this.outputShape=e;let n=Br("rc",r),a=`${n[r-1]} + 1 < ${this.outputShape[r-1]}`,s=`${n[r-2]} + 1 < ${this.outputShape[r-2]}`,i=vt(r);r===1?this.userCode=`
|
|
void main(){
|
|
int rc = getOutputCoords();
|
|
vec4 result = vec4(0.);
|
|
result.r = getChannel(getX(${e[0]} - rc - 1),
|
|
${e[0]} - rc - 1);
|
|
if(${a}){
|
|
result.g = getChannel(getX(${e[0]} - (rc + 1) - 1),
|
|
${e[0]} - (rc + 1) - 1);
|
|
}
|
|
setOutput(result);
|
|
}
|
|
`:this.userCode=`
|
|
void main() {
|
|
${i} rc = getOutputCoords();
|
|
vec4 result = vec4(0.);
|
|
result.r = ${o(n.slice())};
|
|
if(${a}){
|
|
result.g = ${l(n.slice())};
|
|
}
|
|
if(${s}) {
|
|
result.b = ${u(n.slice())};
|
|
if(${a}) {
|
|
result.a = ${d(n.slice())};
|
|
}
|
|
}
|
|
setOutput(result);
|
|
}
|
|
`;function o(c){return h(c)}function l(c){return c[r-1]="("+c[r-1]+" + 1)",h(c)}function u(c){return c[r-2]="("+c[r-2]+" + 1)",h(c)}function d(c){return c[r-1]="("+c[r-1]+" + 1)",c[r-2]="("+c[r-2]+" + 1)",h(c)}function h(c){let f=e.map((y,A)=>p(A,c)),m=f.join(","),g=f.slice(-2).join(",");return`getChannel(getX(${m}), vec2(${g}))`}function p(c,f){return t.indexOf(c)!==-1&&e[c]!==1?`${e[c]} - ${f[c]} - 1`:`${f[c]}`}}};function Tle(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{dims:s}=n,i=a.shape.length,o=v.parseAxisParam(s,a.shape);if(i===0)return fn({inputs:{x:a},backend:r});let l=Z().getBool("WEBGL_PACK_ARRAY_OPERATIONS")?new Cle(a.shape,o):new Sle(a.shape,o);return r.runWebGLProgram(l,[a],a.dtype)}var Nle={kernelName:gl,backendName:"webgl",kernelFunc:Tle},Ele=class{constructor(e,t){this.variableNames=["Image"],this.outputShape=[],this.customUniforms=[{name:"params",type:"vec4"}];let r=e[1],n=e[2];this.outputShape=e;let a="";typeof t=="number"?a=`float outputValue = ${t.toFixed(2)};`:a=`
|
|
vec3 fill = vec3(${t.join(",")});
|
|
float outputValue = fill[coords[3]];`,this.userCode=`
|
|
void main() {
|
|
ivec4 coords = getOutputCoords();
|
|
int x = coords[2];
|
|
int y = coords[1];
|
|
float coordXFloat = (float(x) - params[0]) * params[3] -
|
|
(float(y) - params[1]) * params[2];
|
|
float coordYFloat = (float(x) - params[0]) * params[2] +
|
|
(float(y) - params[1]) * params[3];
|
|
int coordX = int(round(coordXFloat + params[0]));
|
|
int coordY = int(round(coordYFloat + params[1]));
|
|
${a}
|
|
if(coordX >= 0 && coordX < ${n} && coordY >= 0 && coordY < ${r}) {
|
|
outputValue = getImage(coords[0], coordY, coordX, coords[3]);
|
|
}
|
|
setOutput(outputValue);
|
|
}
|
|
`}},Rle={kernelName:Rl,backendName:"webgl",kernelFunc:({inputs:e,attrs:t,backend:r})=>{let{image:n}=e,{radians:a,fillValue:s,center:i}=t,o=r,l=new Ele(n.shape,s),[u,d]=C.getImageCenter(i,n.shape[1],n.shape[2]),h=[[u,d,Math.sin(a),Math.cos(a)]];return o.runWebGLProgram(l,[n],n.dtype,h)}},$le=`
|
|
// OpenGL ES does not support round function.
|
|
// The algorithm is based on banker's rounding.
|
|
float base = floor(x);
|
|
if ((x - base) < 0.5) {
|
|
return floor(x);
|
|
} else if ((x - base) > 0.5) {
|
|
return ceil(x);
|
|
} else {
|
|
if (mod(base, 2.0) == 0.0) {
|
|
return base;
|
|
} else {
|
|
return base + 1.0;
|
|
}
|
|
}
|
|
`,Mle=it({opSnippet:$le}),Fle={kernelName:yl,backendName:"webgl",kernelFunc:Mle},Ple="return inversesqrt(x);",_le=it({opSnippet:Ple,cpuKernelImpl:ate}),zle={kernelName:Fi,backendName:"webgl",kernelFunc:_le},D9=class{constructor(e,t,r,n,a,s,i=!0){this.variableNames=["updates","indices","defaultValue"],this.outputShape=s;let o=vt(a.length),l=vt(s.length),u="";r===1?u="i":r===2&&(u="i, j");let d=`getIndices(${u})`,h="";n===1?h="i":n===2&&(h="i, coords[1]");let p=`getUpdates(${h})`,c=t>1?"strides[j]":"strides";this.userCode=`
|
|
${o} strides = ${o}(${a});
|
|
|
|
void main() {
|
|
${l} coords = getOutputCoords();
|
|
float sum = 0.0;
|
|
bool found = false;
|
|
for (int i = 0; i < ${e}; i++) {
|
|
int flattenedIndex = 0;
|
|
for (int j = 0; j < ${t}; j++) {
|
|
int index = round(${d});
|
|
flattenedIndex += index * ${c};
|
|
}
|
|
if (flattenedIndex == coords[0]) {
|
|
sum += ${p};
|
|
found = true;
|
|
}
|
|
}
|
|
setOutput(mix(getDefaultValue(), sum, float(found)));
|
|
}
|
|
`}};function Ole(e){let{inputs:t,backend:r,attrs:n}=e,{indices:a,updates:s}=t,{shape:i}=n,{sliceRank:o,numUpdates:l,sliceSize:u,strides:d,outputSize:h}=C.calculateShapes(s,a,i),p=[h/u,u];if(h===0)return r.makeTensorInfo(i,a.dtype);let c=ve({inputs:{x:a},backend:r,attrs:{shape:[l,o]}}),f=ve({inputs:{x:s},backend:r,attrs:{shape:[l,u]}}),m=r.makeTensorInfo([],"float32",new Float32Array([0])),g=new D9(l,o,c.shape.length,f.shape.length,d,p),y=r.runWebGLProgram(g,[f,c,m],f.dtype),A=ve({inputs:{x:y},backend:r,attrs:{shape:i}});return r.disposeIntermediateTensorInfo(c),r.disposeIntermediateTensorInfo(f),r.disposeIntermediateTensorInfo(y),r.disposeIntermediateTensorInfo(m),A}var Dle={kernelName:Al,backendName:"webgl",kernelFunc:Ole},Lle=class{constructor(e,t,r,n){this.variableNames=["sortedSequence","values"],this.customUniforms=[{name:"numInputs",type:"int"}],this.outputShape=[e,r];let a="while (left < right) {",s=`for (int i = 0; i < ${Math.ceil(Math.log2(t+1))}; ++i) { if (left >= right) break;`,i=Z().getNumber("WEBGL_VERSION")===2?a:s,o=n==="left"?"<":"<=";this.userCode=`
|
|
int findBound(int batch, float value) {
|
|
int left = 0;
|
|
int right = numInputs;
|
|
int mid;
|
|
${i}
|
|
mid = (left + right) / 2;
|
|
if (getSortedSequence(batch, mid) ${o} value) {
|
|
left = mid + 1;
|
|
} else {
|
|
right = mid;
|
|
}
|
|
}
|
|
return right;
|
|
}
|
|
|
|
void main() {
|
|
ivec2 coords = getOutputCoords();
|
|
int batch = coords[0];
|
|
int valueIndex = coords[1];
|
|
|
|
float value = getValues(batch, valueIndex);
|
|
|
|
setOutput(float(findBound(batch, value)));
|
|
}
|
|
`}};function Ble(e){let{inputs:t,backend:r,attrs:n}=e,{sortedSequence:a,values:s}=t,{side:i}=n,o=new Lle(a.shape[0],a.shape[1],s.shape[1],i),l=[[a.shape[1]]];return r.runWebGLProgram(o,[a,s],"int32",l)}var Wle={kernelName:vf,backendName:"webgl",kernelFunc:Ble},Vle=class{constructor(e,t,r){this.variableNames=["c","a","b"],this.outputShape=t;let n,a;if(r>4)throw Error(`Where for rank ${r} is not yet supported`);if(r===1)a="resRC",n="resRC";else{let i=["resRC.x","resRC.y","resRC.z","resRC.w"],o=[],l=[];for(let u=0;u<t.length;u++)l.push(`${i[u]}`),u<e&&o.push(`${i[u]}`);n=o.join(),a=l.join()}let s=vt(r);this.userCode=`
|
|
void main() {
|
|
${s} resRC = getOutputCoords();
|
|
float cVal = getC(${n});
|
|
if (cVal >= 1.0) {
|
|
setOutput(getA(${a}));
|
|
} else {
|
|
setOutput(getB(${a}));
|
|
}
|
|
}
|
|
`}};function Ule(e){let{inputs:t,backend:r}=e,{condition:n,t:a,e:s}=t,i=new Vle(n.shape.length,a.shape,a.shape.length);return r.runWebGLProgram(i,[n,a,s],Nr(a.dtype,s.dtype))}var Gle={kernelName:xl,backendName:"webgl",kernelFunc:Ule},jle=`
|
|
// Stable and Attracting Fixed Point (0, 1) for Normalized Weights.
|
|
// see: https://arxiv.org/abs/1706.02515
|
|
float scaleAlpha = ${C.SELU_SCALEALPHA};
|
|
float scale = ${C.SELU_SCALE};
|
|
return (x >= 0.0) ? scale * x : scaleAlpha * (exp(x) - 1.0);
|
|
`,Hle=it({opSnippet:jle}),qle={kernelName:id,backendName:"webgl",kernelFunc:Hle},Xle=Ed+`
|
|
return 1.0 / (1.0 + exp(-1.0 * x));
|
|
`,Kle=`
|
|
vec4 result = 1.0 / (1.0 + exp(-1.0 * x));
|
|
bvec4 isNaN = isnan(x);
|
|
|
|
result.r = isNaN.r ? x.r : result.r;
|
|
result.g = isNaN.g ? x.g : result.g;
|
|
result.b = isNaN.b ? x.b : result.b;
|
|
result.a = isNaN.a ? x.a : result.a;
|
|
|
|
return result;
|
|
`,Zle=it({opSnippet:Xle,packedOpSnippet:Kle,cpuKernelImpl:ste}),Yle={kernelName:_i,backendName:"webgl",kernelFunc:Zle},Jle=`
|
|
if (isnan(x)) { return 0.0; }
|
|
return sign(x);
|
|
`,Qle=it({opSnippet:Jle}),eue={kernelName:od,backendName:"webgl",kernelFunc:Qle},tue=Ed+`
|
|
return sin(x);
|
|
`,rue=it({opSnippet:tue}),nue={kernelName:Pi,backendName:"webgl",kernelFunc:rue},aue=`
|
|
float e2x = exp(x);
|
|
return (e2x - 1.0 / e2x) / 2.0;
|
|
`,sue=it({opSnippet:aue}),iue={kernelName:vl,backendName:"webgl",kernelFunc:sue},oue=`
|
|
float epsilon = 1.1920928955078125e-7;
|
|
float threshold = log(epsilon) + 2.0;
|
|
|
|
bool too_large = x > -threshold;
|
|
bool too_small = x < threshold;
|
|
|
|
float result;
|
|
float exp_x = exp(x);
|
|
|
|
if (too_large){
|
|
result = x;
|
|
}
|
|
else if (too_small){
|
|
result = exp_x;
|
|
}
|
|
else{
|
|
result = log(exp_x + 1.0);
|
|
}
|
|
return result;
|
|
`,lue=it({opSnippet:oue}),uue={kernelName:ld,backendName:"webgl",kernelFunc:lue},due=e=>{let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{blockShape:s,paddings:i}=n;v.assert(a.shape.length<=4,()=>"spaceToBatchND for rank > 4 with a WebGL backend not implemented yet");let o=s.reduce((y,A)=>y*A),l=[[0,0]];l.push(...i);for(let y=1+s.length;y<a.shape.length;++y)l.push([0,0]);let u=[],d=z9({inputs:{x:a},backend:r,attrs:{paddings:l,constantValue:0}}),h=C.getReshaped(d.shape,s,o,!1),p=C.getPermuted(h.length,s.length,!1),c=C.getReshapedPermuted(d.shape,s,o,!1),f=ve({inputs:{x:d},backend:r,attrs:{shape:h}}),m=dr({inputs:{x:f},backend:r,attrs:{perm:p}}),g=ve({inputs:{x:m},backend:r,attrs:{shape:c}});return u.push(d),u.push(f),u.push(m),u.forEach(y=>r.disposeIntermediateTensorInfo(y)),g},pue={kernelName:wl,backendName:"webgl",kernelFunc:due};function hue(e){let{inputs:t,backend:r}=e,{indices:n,values:a,denseShape:s,defaultValue:i}=t;if(s.shape.length!==1)throw new Error(`Dense shape must be a vector, saw:
|
|
${s.shape}`);if(n.shape.length!==2)throw new Error(`Indices must be a matrix, saw:
|
|
${n.shape}`);if(a.shape.length!==1)throw new Error(`Values must be a vector, saw:
|
|
${a.shape}`);if(i.shape.length!==0)throw new Error(`Default value must be a scalar, saw:
|
|
${i.shape}`);let o=r.readSync(n.dataId),l=r.readSync(a.dataId),u=r.readSync(s.dataId),d=r.readSync(i.dataId)[0],[h,p,c,f,m]=ote(o,n.shape,n.dtype,l,a.dtype,u,d);return[r.makeTensorInfo(p,n.dtype,h),r.makeTensorInfo([p[0]],a.dtype,c),r.makeTensorInfo([f.length],"bool",new Uint8Array(f.map(g=>Number(g)))),r.makeTensorInfo([m.length],n.dtype,new Int32Array(m))]}var cue={kernelName:lh,backendName:"webgl",kernelFunc:hue};function fue(e){let{inputs:t,backend:r}=e,{inputIndices:n,inputShape:a,newShape:s}=t;if(n.shape.length!==2)throw new Error(`Input indices should be a matrix but received shape ${n.shape}`);if(a.shape.length!==1)throw new Error(`Input shape should be a vector but received shape ${a.shape}`);if(s.shape.length!==1)throw new Error(`Target shape should be a vector but received shape ${s.shape}`);let i=Array.from(r.readSync(a.dataId)),o=r.readSync(n.dataId),l=Array.from(r.readSync(s.dataId)),[u,d,h]=lte(o,n.shape,n.dtype,i,l);return[r.makeTensorInfo(d,n.dtype,u),r.makeTensorInfo([h.length],s.dtype,new Int32Array(h))]}var mue={kernelName:ud,backendName:"webgl",kernelFunc:fue};function gue(e){let{inputs:t,backend:r}=e,{data:n,indices:a,segmentIds:s}=t;if(n.shape.length<1)throw new Error("Data should be at least 1 dimensional but received scalar");if(a.shape.length!==1)throw new Error(`Indices should be a vector but received shape
|
|
${a.shape}`);if(s.shape.length!==1)throw new Error(`Segment ids should be a vector but received shape
|
|
${s.shape}`);let i=r.readSync(n.dataId),o=r.readSync(a.dataId),l=r.readSync(s.dataId),[u,d]=a9(i,n.shape,n.dtype,o,l,!0);return r.makeTensorInfo(d,n.dtype,u)}var yue={kernelName:uh,backendName:"webgl",kernelFunc:gue};function Aue(e){let{inputs:t,backend:r}=e,{data:n,indices:a,segmentIds:s}=t;if(n.shape.length<1)throw new Error("Data should be at least 1 dimensional but received scalar");if(a.shape.length!==1)throw new Error(`Indices should be a vector but received shape
|
|
${a.shape}`);if(s.shape.length!==1)throw new Error(`Segment ids should be a vector but received shape
|
|
${s.shape}`);let i=r.readSync(n.dataId),o=r.readSync(a.dataId),l=r.readSync(s.dataId),[u,d]=a9(i,n.shape,n.dtype,o,l);return r.makeTensorInfo(d,n.dtype,u)}var xue={kernelName:dh,backendName:"webgl",kernelFunc:Aue};function bue(e){let{inputs:t,backend:r,attrs:n}=e,{sparseIndices:a,sparseValues:s,defaultValue:i}=t,{outputShape:o}=n,{sliceRank:l,numUpdates:u,strides:d,outputSize:h}=C.calculateShapes(s,a,o),p=!1,c=new D9(u,l,a.shape.length,s.shape.length,d,[h,1],p),f=r.runWebGLProgram(c,[s,a,i],s.dtype),m=ve({inputs:{x:f},backend:r,attrs:{shape:o}});return r.disposeIntermediateTensorInfo(f),m}var vue={kernelName:ph,backendName:"webgl",kernelFunc:bue};function wue(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{numOrSizeSplits:s,axis:i}=n,o=v.parseAxisParam(i,a.shape)[0],l=C.prepareSplitSize(a,s,o),u=a.shape.length,d=new Array(u).fill(0),h=a.shape.slice();return l.map(p=>{let c=[...h];c[o]=p;let f=Rd({inputs:{x:a},backend:r,attrs:{begin:d,size:c}});return d[o]+=p,f})}var kue={kernelName:kl,backendName:"webgl",kernelFunc:wue},nv="return sqrt(x);",Iue=it({opSnippet:nv,packedOpSnippet:nv,cpuKernelImpl:ute}),Sue={kernelName:zi,backendName:"webgl",kernelFunc:Iue},Cue="return x * x;",Tue=it({opSnippet:Cue}),Nue={kernelName:dd,backendName:"webgl",kernelFunc:Tue},av="return (a - b) * (a - b);",Eue=wr({opSnippet:av,packedOpSnippet:av}),Rue={kernelName:Li,backendName:"webgl",kernelFunc:Eue};function $ue({inputs:e,attrs:t,backend:r}){let{x:n}=e,a=Yn+`
|
|
return x > 0.0 ? 1.0 : float(${t.alpha});
|
|
`,s=new Xa(n.shape,a);return r.runWebGLProgram(s,[n],n.dtype)}var Mue={kernelName:Ui,backendName:"webgl",kernelFunc:$ue},Fue=class{constructor(e,t,r){this.variableNames=["x"],this.outputShape=r;let n=r.length,a=vt(r.length),s=vt(r.length),i="";if(n===1)i="coords * strides + begin";else{let o=0;i=r.map((l,u)=>(o++,r.length===1?`coords * strides[${u}] + begin[${u}]`:`coords[${o-1}] * strides[${u}] + begin[${u}]`)).join(",")}this.userCode=`
|
|
${a} begin = ${a}(${e});
|
|
${a} strides = ${a}(${t});
|
|
|
|
void main() {
|
|
${s} coords = getOutputCoords();
|
|
setOutput(getX(${i}));
|
|
}
|
|
`}};function Pue(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{begin:s,end:i,strides:o,beginMask:l,endMask:u,ellipsisMask:d,newAxisMask:h,shrinkAxisMask:p}=n,{finalShapeSparse:c,finalShape:f,isIdentity:m,sliceDim0:g,isSimpleSlice:y,begin:A,end:x,strides:b}=Ot.sliceInfo(a.shape,s,i,o,l,u,d,h,p),w;if(m)w=ve({inputs:{x:a},backend:r,attrs:{shape:f}});else if(g||y){v.assert(a.shape.length>=1,()=>`Input must have rank at least 1, got: ${a.shape.length}`);let N=Ot.computeOutShape(A,x,b),E=Rd({inputs:{x:a},backend:r,attrs:{begin:A,size:N}});w=ve({inputs:{x:E},backend:r,attrs:{shape:f}}),r.disposeIntermediateTensorInfo(E)}else if(r.shouldExecuteOnCPU([a])){let N=r.readSync(a.dataId),E=We(a.shape,a.dtype,N),R=dte(c,E,b,A);w=r.makeTensorInfo(f,a.dtype,R.values)}else{let N=new Fue(A,b,c);w=r.runWebGLProgram(N,[a],a.dtype)}let S=ve({inputs:{x:w},backend:r,attrs:{shape:f}});return r.disposeIntermediateTensorInfo(w),S}var _ue={kernelName:Il,backendName:"webgl",kernelFunc:Pue};function zue(e){let{inputs:t,backend:r,attrs:n}=e,{separator:a,nGramWidths:s,leftPad:i,rightPad:o,padWidth:l,preserveShortSequences:u}=n,{data:d,dataSplits:h}=t,p=r.readSync(d.dataId),c=r.readSync(h.dataId),[f,m]=pte(p,c,a,s,i,o,l,u);return[r.makeTensorInfo([f.length],"string",f),r.makeTensorInfo(h.shape,"int32",m)]}var Oue={kernelName:hh,backendName:"webgl",kernelFunc:zue};function Due(e){let{inputs:t,backend:r,attrs:n}=e,{skipEmpty:a}=n,{input:s,delimiter:i}=t;if(s.dtype!=="string")throw new Error("Input must be of datatype string");if(s.shape.length!==1)throw new Error(`Input must be a vector, got shape: ${s.shape}`);if(i.shape.length!==0)throw new Error(`Delimiter must be a scalar, got shape: ${i.shape}`);let o=r.readSync(s.dataId),l=r.readSync(i.dataId)[0],[u,d,h]=hte(o,l,a),p=d.length;return[r.makeTensorInfo([p,2],"int32",u),r.makeTensorInfo([p],"string",d),r.makeTensorInfo([2],"int32",new Int32Array(h))]}var Lue={kernelName:wf,backendName:"webgl",kernelFunc:Due};function Bue(e){let{inputs:t,backend:r,attrs:n}=e,{numBuckets:a}=n,{input:s}=t;if(s.dtype!=="string")throw new Error("Input must be of datatype string");if(a<=0)throw new Error("Number of buckets must be at least 1");let i=r.readSync(s.dataId),o=cte(i,a);return r.makeTensorInfo(s.shape,"int32",o)}var Wue={kernelName:kf,backendName:"webgl",kernelFunc:Bue},Vue="return tan(x);",Uue=it({opSnippet:Vue}),Gue={kernelName:Sl,backendName:"webgl",kernelFunc:Uue},jue=`
|
|
float e2x = exp(-2.0 * abs(x));
|
|
return sign(x) * (1.0 - e2x) / (1.0 + e2x);
|
|
`,Hue=it({opSnippet:jue}),que={kernelName:Wi,backendName:"webgl",kernelFunc:Hue},Xue=class{constructor(e,t){this.variableNames=["A"];let r=new Array(e.length);for(let s=0;s<r.length;s++)r[s]=e[s]*t[s];this.outputShape=r,this.rank=r.length;let n=vt(this.rank),a=Kue(e);this.userCode=`
|
|
void main() {
|
|
${n} resRC = getOutputCoords();
|
|
setOutput(getA(${a}));
|
|
}
|
|
`}};function Kue(e){let t=e.length;if(t>5)throw Error(`Tile for rank ${t} is not yet supported`);if(t===1)return`imod(resRC, ${e[0]})`;let r=["resRC.x","resRC.y","resRC.z","resRC.w","resRC.u"],n=[];for(let a=0;a<e.length;a++)n.push(`imod(${r[a]}, ${e[a]})`);return n.join()}function L9(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{reps:s}=n;if(a.dtype==="string"||a.shape.length>5){let o=r.readSync(a.dataId),l=a.dtype==="string"?o.map(h=>v.decodeString(h)):o,u=We(a.shape,a.dtype,l),d=mte(u,s);return r.makeTensorInfo(d.shape,d.dtype,d.values)}let i=new Xue(a.shape,s);return r.runWebGLProgram(i,[a],a.dtype)}var Zue={kernelName:es,backendName:"webgl",kernelFunc:L9},Yue=class{constructor(e){this.variableNames=["x","indices"],this.customUniforms=[{name:"n",type:"int"},{name:"firstPass",type:"int"},{name:"negativeInf",type:"float"},{name:"dir",type:"int"},{name:"inc",type:"int"}],this.outputShape=e,this.userCode=`
|
|
void main() {
|
|
ivec2 coords = getOutputCoords();
|
|
int batch = coords[0];
|
|
int elemIdx = coords[1];
|
|
|
|
// We compare elements pair-wise within a group of size 2 * inc.
|
|
// The comparing rule for each group alternates between ascending
|
|
// and descending. Within each group, we compare each pair at
|
|
// positions i and i+inc. To decide whether an element at position i
|
|
// is x0 or x1, we mod it by 2 * inc, if the result is smaller than
|
|
// inc, it is in the first half of the group, we denote it as x0,
|
|
// otherwise we denote it as x1.
|
|
// For example, as shown in the Bitonic top K paper referenced above,
|
|
// Figure5(a) shows that element[1] is in the
|
|
// second half of the group when group size is 2, but it is in the
|
|
// first half of the group when group size is 4.
|
|
|
|
bool isFirstInPair = imod(elemIdx, 2 * inc) < inc;
|
|
int i = isFirstInPair ? elemIdx : elemIdx - inc;
|
|
|
|
int i0 = firstPass == 1 ? i : int(getIndices(batch, i));
|
|
int i1 = firstPass == 1 ? i + inc : int(getIndices(batch, i + inc));
|
|
float x0 = i0 < n ? getX(batch, i0) : negativeInf;
|
|
float x1 = i1 < n ? getX(batch, i1) : negativeInf;
|
|
|
|
// Denotes which direction indices are in (ascending or descending).
|
|
bool reverse = imod(elemIdx, 2 * dir) >= dir;
|
|
bool isGreater = x0 > x1 || (x0 == x1 && i1 > i0);
|
|
if (reverse == isGreater) { // Elements in opposite order of direction
|
|
int iTemp = i0;
|
|
i0 = i1;
|
|
i1 = iTemp;
|
|
}
|
|
if (isFirstInPair) {
|
|
setOutput(float(i0));
|
|
} else {
|
|
setOutput(float(i1));
|
|
}
|
|
}
|
|
`}},Jue=class{constructor(e){this.variableNames=["x","indices"],this.customUniforms=[{name:"n",type:"int"},{name:"firstPass",type:"int"},{name:"k",type:"int"}],this.outputShape=e,this.userCode=`
|
|
void main() {
|
|
// Takes max of indices (0, k), (1, k + 1), (2, k + 2) ...
|
|
ivec2 coords = getOutputCoords();
|
|
int batch = coords[0];
|
|
int elemIdx = coords[1];
|
|
|
|
// The output size is half of the previous size.
|
|
// If the previous sequence is | | | | _ _ _ _ | | | | _ _ _ _ (k=4),
|
|
// we only need to output the indices at positions |, the indices at
|
|
// positions _ can be thrown away, see Figure5(b) After Phase 2
|
|
// (Merge phase) in the Bitonic Top K paper referenced above.
|
|
// For example, the paper shows we only need to output the orange bars.
|
|
// The output sequence should look like this | | | | | | | |.
|
|
// Because the sequence is halved, to map the output index back
|
|
// to the previous sequence to find the corresponding value,
|
|
// we need to double the index. When we double the index,
|
|
// we basically interpolate a position, so 2i looks like
|
|
// | _ | _ | _ | _ | _ | _ | _. We move the | to the first k position
|
|
// of each 2k positions by - elemIdx % k. E.g. for output at
|
|
// index 4,5,6,7, we want to get the corresponding element at
|
|
// original index 8,9,10,11, for output at index 8,9,10,11,
|
|
// we want to get the corresponding element at original index
|
|
// 16,17,18,19, so on and so forth.
|
|
|
|
int i = elemIdx < k ? elemIdx : (elemIdx * 2 - imod(elemIdx, k));
|
|
int i0 = firstPass == 1 ? i : int(getIndices(batch, i));
|
|
int i1 = firstPass == 1 ? i + k : int(getIndices(batch, i + k));
|
|
|
|
float x0 = getX(batch, i0);
|
|
float x1 = i1 < n ? getX(batch, i1) : x0;
|
|
|
|
setOutput(x0 >= x1 ? float(i0) : float(i1));
|
|
}
|
|
`}};function Ao(e,t){t!==null&&e.disposeIntermediateTensorInfo(t)}function sv(e){let t=1;for(;t<e;)t*=2;return t}function Que(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{k:s,sorted:i}=n,o=Z().getNumber("TOPK_LAST_DIM_CPU_HANDOFF_SIZE_THRESHOLD"),l=Z().getNumber("TOPK_K_CPU_HANDOFF_THRESHOLD"),u=a.shape,d=u[u.length-1];if(r.shouldExecuteOnCPU([a])||d<o||s>l){let R=r.readSync(a.dataId),[z,M]=gte(R,u,a.dtype,s,i);return[r.makeTensorInfo(z.shape,z.dtype,z.values),r.makeTensorInfo(M.shape,M.dtype,M.values)]}if(s===0)return u[u.length-1]=0,[r.makeTensorInfo(u,a.dtype,[]),r.makeTensorInfo(u,"int32",[])];if(d===1)return[a,Uh({attrs:{shape:u,dtype:"int32",value:0},backend:r})];let h=r.texData.get(a.dataId),p=h!==null&&h.isPacked,c=p?r.unpackTensor(a):a,f=v.sizeFromShape(u)/d,m=ve({inputs:{x:c},attrs:{shape:[f,d]},backend:r});p&&Ao(r,c);let g=sv(s),y=sv(d),A=null,x=()=>A===null?[m,m]:[m,A],b=(R,z,M)=>{let I=x(),_=new Yue(M),O=[[d],[A===null?1:0],[Number.NEGATIVE_INFINITY],[R],[z]],q=A;A=r.runWebGLProgram(_,I,"int32",O),Ao(r,q)};for(let R=1;R<g;R*=2){let z=R*2;for(let M=R;M>=1;M/=2)b(z,M,[f,y])}for(let R=y;R>g;R/=2){let z=x(),M=new Jue([f,R/2]),I=[[d],[A===null?1:0],[g]],_=A;A=r.runWebGLProgram(M,z,"int32",I),Ao(r,_);let O=g/2,q=O*2;for(let K=O;K>=1;K/=2)b(q,K,A.shape)}let w=A;A=Rd({inputs:{x:A},backend:r,attrs:{begin:0,size:[f,s]}}),Ao(r,w);let S=R9({inputs:{x:m,indices:A},backend:r,attrs:{axis:1,batchDims:1}});Ao(r,m);let N=u.slice(0,-1);N.push(s),w=A,A=ve({inputs:{x:A},attrs:{shape:N},backend:r}),Ao(r,w);let E=S;return S=ve({inputs:{x:S},attrs:{shape:N},backend:r}),Ao(r,E),[S,A]}var ede={kernelName:Cl,backendName:"webgl",kernelFunc:Que},tde=class{constructor(e,t,r,n,a,s){this.variableNames=["Image","Transforms"],this.outputShape=s;let i=r==="nearest"?1:2,o;switch(n){case"constant":o=1;break;case"reflect":o=2;break;case"wrap":o=3;break;case"nearest":o=4;break;default:o=1;break}this.userCode=`
|
|
float mapCoord(float outCoord, float len) {
|
|
float inCoord = outCoord;
|
|
if(${o} == 2) {
|
|
if (inCoord < 0.0) {
|
|
if (len <= 1.0) {
|
|
inCoord = 0.0;
|
|
} else {
|
|
float sz2 = 2.0 * len;
|
|
if (inCoord < sz2) {
|
|
inCoord = sz2 * float(int(float(-inCoord / sz2))) +
|
|
inCoord;
|
|
}
|
|
inCoord = inCoord < -len ? inCoord + sz2 : -inCoord - 1.0;
|
|
}
|
|
} else if (inCoord > len - 1.0) {
|
|
if (len <= 1.0) {
|
|
inCoord = 0.0;
|
|
} else {
|
|
float sz2 = 2.0 * len;
|
|
inCoord -= sz2 * float(int(float(inCoord / sz2)));
|
|
if (inCoord >= len) {
|
|
inCoord = sz2 - inCoord - 1.0;
|
|
}
|
|
}
|
|
}
|
|
return clamp(inCoord, 0.0, len - 1.0);
|
|
} else if (${o} == 3) {
|
|
if (inCoord < 0.0) {
|
|
if (len <= 1.0) {
|
|
inCoord = 0.0;
|
|
} else {
|
|
float sz = len - 1.0;
|
|
inCoord += len * (float(int(float(-inCoord / sz))) + 1.0);
|
|
}
|
|
} else if (inCoord > len - 1.0) {
|
|
if (len <= 1.0) {
|
|
inCoord = 0.0;
|
|
} else {
|
|
float sz = len - 1.0;
|
|
inCoord -= len * float(int(float(inCoord / sz)));
|
|
}
|
|
}
|
|
return clamp(inCoord, 0.0, len - 1.0);
|
|
} else if (${o} == 4) {
|
|
return clamp(outCoord, 0.0, len - 1.0);
|
|
} else {
|
|
return outCoord;
|
|
}
|
|
}
|
|
|
|
float readWithFillValue(int batch, int coordY, int coordX,
|
|
int channel) {
|
|
float outputValue;
|
|
if (0 <= coordY && coordY < ${e} && 0 <= coordX && coordX < ${t}) {
|
|
outputValue = getImage(batch, coordY, coordX, channel);
|
|
} else {
|
|
outputValue = float(${a});
|
|
}
|
|
return outputValue;
|
|
}
|
|
|
|
void main() {
|
|
ivec4 coords = getOutputCoords();
|
|
float outputValue;
|
|
int batch = coords[0];
|
|
int x = coords[2];
|
|
int y = coords[1];
|
|
int channel = coords[3];
|
|
float xf = float(x);
|
|
float yf = float(y);
|
|
float a1 = getTransforms(batch, 0);
|
|
float a2 = getTransforms(batch, 1);
|
|
float a3 = getTransforms(batch, 2);
|
|
float b1 = getTransforms(batch, 3);
|
|
float b2 = getTransforms(batch, 4);
|
|
float b3 = getTransforms(batch, 5);
|
|
float c1 = getTransforms(batch, 6);
|
|
float c2 = getTransforms(batch, 7);
|
|
float projection = c1 * xf + c2 * yf + 1.0;
|
|
if (projection == 0.0) {
|
|
outputValue = float(${a});
|
|
} else {
|
|
float inX = (a1 * xf + a2 * yf + a3) / projection;
|
|
float inY = (b1 * xf + b2 * yf + b3) / projection;
|
|
float mapX = mapCoord(inX, float(${t}));
|
|
float mapY = mapCoord(inY, float(${e}));
|
|
|
|
if (${i} == 1) {
|
|
int coordY = int(round(mapY));
|
|
int coordX = int(round(mapX));
|
|
outputValue = readWithFillValue(batch, coordY, coordX,
|
|
channel);
|
|
} else {
|
|
float yFloor = floor(mapY);
|
|
float xFloor = floor(mapX);
|
|
float yCeil = yFloor + 1.0;
|
|
float xCeil = xFloor + 1.0;
|
|
float valueYFloor = (xCeil - mapX) *
|
|
readWithFillValue(batch, int(yFloor), int(xFloor), channel) +
|
|
(mapX - xFloor) *
|
|
readWithFillValue(batch, int(yFloor), int(xCeil), channel);
|
|
float valueYCeil = (xCeil - mapX) *
|
|
readWithFillValue(batch, int(yCeil), int(xFloor), channel) +
|
|
(mapX - xFloor) *
|
|
readWithFillValue(batch, int(yCeil), int(xCeil), channel);
|
|
outputValue = (yCeil - mapY) * valueYFloor +
|
|
(mapY - yFloor) * valueYCeil;
|
|
}
|
|
}
|
|
setOutput(outputValue);
|
|
}
|
|
`}};function rde(e){let{inputs:t,backend:r,attrs:n}=e,{image:a,transforms:s}=t,{interpolation:i,fillMode:o,fillValue:l,outputShape:u}=n,[d,h,p,c]=a.shape,[f,m]=u!=null?u:[h,p],g=[d,f,m,c],y=new tde(h,p,i,o,l,g);return r.runWebGLProgram(y,[a,s],"float32")}var nde={kernelName:Tl,backendName:"webgl",kernelFunc:rde};function ade(e){let{inputs:t,attrs:r,backend:n}=e,{axis:a}=r,{x:s}=t;Id(s,"unique"),console.warn("WARNING: ","UI might be locked temporarily as data is being downloaded");let i=n.readSync(s.dataId),{outputValues:o,outputShape:l,indices:u}=yte(i,a,s.shape,s.dtype);return[n.makeTensorInfo(l,s.dtype,o),n.makeTensorInfo([u.length],"int32",u)]}var sde={kernelName:If,backendName:"webgl",kernelFunc:ade};function ide(e){let{inputs:t,backend:r,attrs:n}=e,{value:a}=t,{axis:s}=n;s<0&&(s+=a.shape.length);let i=a,o=i.shape.length,l=a.shape[s],u=new Array(o-1),d=0;for(let m=0;m<o;m++)m!==s&&(u[d++]=i.shape[m]);let h=[],p=new Array(o).fill(0),c=i.shape.slice();c[s]=1;let f=new Array(l);for(let m=0;m<f.length;m++){p[s]=m;let g=Rd({inputs:{x:i},backend:r,attrs:{begin:p,size:c}}),y=ve({inputs:{x:g},backend:r,attrs:{shape:u}});f[m]=y,h.push(g)}return h.forEach(m=>r.disposeIntermediateTensorInfo(m)),f}var ode={kernelName:Nl,backendName:"webgl",kernelFunc:ide},lde=class{constructor(e,t){this.variableNames=["x","segmentIds"];let r=e.windowSize,n=e.batchSize,a=e.inSize,s=e.numSegments,i=s*Math.ceil(a/r);this.outputShape=[n,i];let o="0.0",l="sumValue",u=Math.floor(r/4)*4,d=r%4,h=`
|
|
sumValue += dot(values, segFilter);
|
|
`,p="";a%r>0&&(p=`
|
|
if (inIdx < 0 || inIdx >= ${a}) {
|
|
return initializationValue;
|
|
}
|
|
`);let c="";a%r>0&&(c=`
|
|
if (inIdx < 0 || inIdx >= ${a}) {
|
|
return -1.0;
|
|
}
|
|
`),this.userCode=`
|
|
const float initializationValue = ${o};
|
|
|
|
float getValue(int batch, int inIdx) {
|
|
${p}
|
|
return getX(batch, inIdx);
|
|
}
|
|
|
|
float getSegmentIdAtIndex(int inIdx) {
|
|
${c}
|
|
return getSegmentIds(inIdx);
|
|
}
|
|
|
|
void main() {
|
|
ivec2 coords = getOutputCoords();
|
|
int batch = coords[0];
|
|
int outIdx = coords[1];
|
|
int inOffset = int(floor(float(outIdx) / float(
|
|
${s})) * float(${r}));
|
|
int currentSeg = int(mod(float(outIdx), float(${s})));
|
|
|
|
float sumValue = 0.0;
|
|
|
|
for (int i = 0; i < ${u}; i += 4) {
|
|
int inIdx = inOffset + i;
|
|
vec4 values = vec4(
|
|
getValue(batch, inIdx),
|
|
getValue(batch, inIdx + 1),
|
|
getValue(batch, inIdx + 2),
|
|
getValue(batch, inIdx + 3)
|
|
);
|
|
|
|
vec4 segFilter = vec4(
|
|
int(getSegmentIdAtIndex(inIdx)) == currentSeg ? 1 : 0,
|
|
int(getSegmentIdAtIndex(inIdx + 1)) == currentSeg ? 1 : 0,
|
|
int(getSegmentIdAtIndex(inIdx + 2)) == currentSeg ? 1 : 0,
|
|
int(getSegmentIdAtIndex(inIdx + 3)) == currentSeg ? 1 : 0
|
|
);
|
|
|
|
${h}
|
|
}
|
|
|
|
int inIdx = inOffset + ${u};
|
|
if (${d===1}) {
|
|
vec4 values = vec4(
|
|
getValue(batch, inIdx),
|
|
initializationValue,
|
|
initializationValue,
|
|
initializationValue
|
|
);
|
|
|
|
int inIdxSeg = int(getSegmentIdAtIndex(inIdx));
|
|
|
|
vec4 segFilter = vec4(
|
|
int(getSegmentIdAtIndex(inIdx)) == currentSeg ? 1 : 0,
|
|
0,
|
|
0,
|
|
0
|
|
);
|
|
|
|
${h}
|
|
} else if (${d===2}) {
|
|
vec4 values = vec4(
|
|
getValue(batch, inIdx),
|
|
getValue(batch, inIdx + 1),
|
|
initializationValue,
|
|
initializationValue
|
|
);
|
|
|
|
vec4 segFilter = vec4(
|
|
int(getSegmentIdAtIndex(inIdx)) == currentSeg ? 1 : 0,
|
|
int(getSegmentIdAtIndex(inIdx + 1)) == currentSeg ? 1 : 0,
|
|
0,
|
|
0
|
|
);
|
|
|
|
${h}
|
|
} else if (${d===3}) {
|
|
vec4 values = vec4(
|
|
getValue(batch, inIdx),
|
|
getValue(batch, inIdx + 1),
|
|
getValue(batch, inIdx + 2),
|
|
initializationValue
|
|
);
|
|
|
|
vec4 segFilter = vec4(
|
|
int(getSegmentIdAtIndex(inIdx)) == currentSeg ? 1 : 0,
|
|
int(getSegmentIdAtIndex(inIdx + 1)) == currentSeg ? 1 : 0,
|
|
int(getSegmentIdAtIndex(inIdx + 2)) == currentSeg ? 1 : 0,
|
|
0
|
|
);
|
|
|
|
${h}
|
|
}
|
|
setOutput(${l});
|
|
}
|
|
`}};function ude(e){let{inputs:t,backend:r,attrs:n}=e,{x:a,segmentIds:s}=t,{numSegments:i}=n,o=a.shape.length,l=[],u=0,d=C.getAxesPermutation([u],o),h=a;d!=null&&(h=dr({inputs:{x:a},backend:r,attrs:{perm:d}}),l.push(h),u=C.getInnerMostAxes(1,o)[0]);let p=C.segment_util.computeOutShape(h.shape,u,i),c=v.sizeFromShape([h.shape[u]]),f=ve({inputs:{x:h},backend:r,attrs:{shape:[-1,c]}});l.push(f);let m=gh(a.dtype),g=(b,w,S,N,E)=>{let R=b.shape[0],z=b.shape[1],M=C.segment_util.segOpComputeOptimalWindowSize(z,E),I={windowSize:M,inSize:z,batchSize:R,numSegments:E},_=new lde(I,w),O=r.compileAndRun(_,[b,S],N);if(l.push(O),O.shape[1]===E)return O;let q=O9({backend:r,attrs:{start:0,stop:E,step:1,dtype:"float32"}}),K=L9({inputs:{x:q},backend:r,attrs:{reps:[z/M]}});return l.push(q),l.push(K),g(O,w,K,N,E)},y=g(f,"unsortedSegmentSum",s,m,i),A=ve({inputs:{x:y},backend:r,attrs:{shape:p}}),x=A;if(d!=null){l.push(A);let b=C.getUndoAxesPermutation(d);x=dr({inputs:{x},backend:r,attrs:{perm:b}})}return l.forEach(b=>r.disposeIntermediateTensorInfo(b)),x}var dde={kernelName:ch,backendName:"webgl",kernelFunc:ude},pde=[hre,fre,yre,bre,wre,Sre,Tre,Ere,Fre,_re,Dre,Wre,Gre,Xre,Yre,Qre,tne,sne,one,une,cne,bne,wne,Ine,Rne,Mne,zne,qte,Lne,Gne,Xne,eae,rae,aae,iae,lae,pae,fae,yae,xae,vae,kae,Cae,Nae,Mae,Pae,Oae,Bae,Vae,Hae,Zae,ese,nse,ise,ose,use,pse,cse,mse,yse,vse,Ise,Tse,Ese,Mse,_se,Lse,Use,Hte,jse,Vne,Xse,Yse,eie,Kte,aie,lie,die,fie,yie,vie,Iie,Nie,Mie,_ie,Oie,Wie,Uie,jie,Kie,Yie,Qie,toe,noe,ooe,poe,moe,koe,ere,Toe,Roe,Foe,zoe,Cne,Loe,Woe,Uoe,Hoe,Zoe,Yte,Joe,Qoe,Tne,xoe,rle,ile,dle,rre,fle,yle,vle,Ile,Nle,Rle,Fle,zle,Dle,Wle,Gle,qle,Yle,eue,nue,iue,Ane,voe,uue,pue,cue,mue,yue,xue,vue,kue,Sue,Nue,Rue,Mue,_ue,Oue,Lue,Wue,boe,ure,Gue,que,Zue,ede,nde,dre,sde,ode,dde,Boe];for(let e of pde)qn(e);var Ki=Z();Ki.registerFlag("WEBGPU_DEFERRED_SUBMIT_BATCH_SIZE",()=>15);Ki.registerFlag("WEBGPU_CPU_FORWARD",()=>!0);Ki.registerFlag("WEBGPU_MATMUL_WORK_PER_THREAD",()=>4);Ki.registerFlag("WEBGPU_USE_NAIVE_CONV2D_TRANSPOSE",()=>!1);Ki.registerFlag("WEBGPU_USE_LOW_POWER_GPU",()=>!1);Ki.registerFlag("WEBGPU_CPU_HANDOFF_SIZE_THRESHOLD",()=>1e3);Ki.registerFlag("WEBGPU_USE_PROFILE_TOOL",()=>!1);Ki.registerFlag("WEBGPU_USE_IMPORT",()=>!1);var hde="return a + b;",cde="return areal * breal - aimag * bimag;",fde="return areal * bimag + aimag * breal;",mde="return a / b;",gde="return a * b;",yde="return (a - b) * (a - b);",Ade="return a - b;",xde="return f32(a == b);",bde="return vec4<f32>(a == b);",vde="return f32(a > b);",wde="return vec4<f32>(a > b);",kde="return f32(a >= b);",Ide="return vec4<f32>(a >= b);",Sde="return f32(a < b);",Cde="return vec4<f32>(a < b);",Tde="return f32(a <= b);",Nde="return vec4<f32>(a <= b);",Ede="return f32(f32(a) >= 1.0 && f32(b) >= 1.0);",Rde=`return (vec4<f32>(a >= vec4<f32>(1.0)) *
|
|
vec4<f32>(b >= vec4<f32>(1.0)));`,$de=`
|
|
if (isnan(a)) { return a; }
|
|
if (isnan(b)) { return b; }
|
|
`,B9=`
|
|
if (isNaN.r) {
|
|
resultTemp.r = uniforms.NAN;
|
|
}
|
|
if (isNaN.g) {
|
|
resultTemp.g = uniforms.NAN;
|
|
}
|
|
if (isNaN.b) {
|
|
resultTemp.b = uniforms.NAN;
|
|
}
|
|
if (isNaN.a) {
|
|
resultTemp.a = uniforms.NAN;
|
|
}
|
|
`,Mde=`
|
|
let s = sign(a) * sign(b);
|
|
let ia = i32(round(a));
|
|
let ib = i32(round(b));
|
|
return f32(idiv(ia, ib, s));
|
|
`,Fde=`
|
|
let ia = vec4<i32>(round(a));
|
|
let ib = vec4<i32>(round(b));
|
|
let cond = ib != vec4<i32>(0);
|
|
var resultTemp = vec4<i32>(0);
|
|
let s = sign(a) * sign(b);
|
|
|
|
// Windows (D3D) wants guaranteed non-zero int division at compile-time.
|
|
if (cond[0]) {
|
|
resultTemp[0] = idiv(ia[0], ib[0], s[0]);
|
|
}
|
|
if (cond[1]) {
|
|
resultTemp[1] = idiv(ia[1], ib[1], s[1]);
|
|
}
|
|
if (cond[2]) {
|
|
resultTemp[2] = idiv(ia[2], ib[2], s[2]);
|
|
}
|
|
if (cond[3]) {
|
|
resultTemp[3] = idiv(ia[3], ib[3], s[3]);
|
|
}
|
|
return vec4<f32>(resultTemp);
|
|
`,Pde="return f32(a != b);",_de="return vec4<f32>(a != b);",zde=`
|
|
if(a < 0.0 && floor(b) < b) {
|
|
return uniforms.NAN;
|
|
}
|
|
if (b == 0.0) {
|
|
return 1.0;
|
|
}
|
|
if (round(abs(b) % 2.0) != 1.0) {
|
|
return pow(abs(a), b);
|
|
}
|
|
return sign(a) * pow(abs(a), b);
|
|
`,Ode=`
|
|
let isModRound1Bool = vec4<i32>(round(abs(b) % vec4<f32>(2.0))) == vec4<i32>(1);
|
|
let isModRound1 = vec4<f32>(isModRound1Bool);
|
|
let multiplier = sign(a) * isModRound1 + (vec4<f32>(1.0) - isModRound1);
|
|
var resultTemp = multiplier * pow(abs(a), b);
|
|
|
|
// Ensure that a^0 = 1, including 0^0 = 1 as this correspond to TF and JS
|
|
let isExpZero = b == vec4<f32>(0.0);
|
|
if (isExpZero.r) {
|
|
resultTemp.r = 1.0;
|
|
}
|
|
if (isExpZero.g) {
|
|
resultTemp.g = 1.0;
|
|
}
|
|
if (isExpZero.b) {
|
|
resultTemp.b = 1.0;
|
|
}
|
|
if (isExpZero.a) {
|
|
resultTemp.a = 1.0;
|
|
}
|
|
let isNaN = a < vec4<f32>(0.0) & floor(b) < b;
|
|
${B9}
|
|
return resultTemp;
|
|
`,Dde="if (a < 0.0) { return b * a; } return a;",Lde=`
|
|
let aLessThanZero = vec4<f32>(a < vec4<f32>(0.0));
|
|
return (aLessThanZero * (b * a)) + ((vec4<f32>(1.0) - aLessThanZero) * a);
|
|
`;function iv(e,t){let r=t?B9:$de;return t?`
|
|
var resultTemp = vec4<f32>(${e}(a, b));
|
|
let isNaN = isnanVec4(a) | isnanVec4(b);
|
|
`+r+`
|
|
return resultTemp;
|
|
`:r+`
|
|
return ${e}(a, b);
|
|
`}function Gh(e,t){switch(e){case 0:return gde;case 1:return hde;case 2:return Ade;case 3:return mde;case 4:return t?bde:xde;case 5:return t?wde:vde;case 6:return t?Ide:kde;case 7:return t?Cde:Sde;case 8:return t?Nde:Tde;case 9:return t?Rde:Ede;case 10:return t?_de:Pde;case 11:return yde;case 12:return t?Fde:Mde;case 14:return t?Lde:Dde;case 15:return iv("max",t);case 16:return iv("min",t);case 13:return t?Ode:zde;case 17:return cde;case 18:return fde;default:throw new Error(`BinaryType ${e} is not implemented!`)}}var Bde="return abs(a);",Wde="return ceil(a);",Vde="return cos(a);",Ude=`
|
|
let e2x = exp(-a);
|
|
return (e2x + 1.0 / e2x) / 2.0;
|
|
`,Gde="return exp(a) - 1.0;",jde="if (a >= 0.0) { return a; } return (exp(a) - 1.0);",Hde=`
|
|
var resFloat = exp(a) - vec4<f32>(1.0);
|
|
if (a.r >= 0.0) {
|
|
resFloat.r = a.r;
|
|
}
|
|
if (a.g >= 0.0) {
|
|
resFloat.g = a.g;
|
|
}
|
|
if (a.b >= 0.0) {
|
|
resFloat.b = a.b;
|
|
}
|
|
if (a.a >= 0.0) {
|
|
resFloat.a = a.a;
|
|
}
|
|
return resFloat;
|
|
`,qde="return exp(a);",Xde="return floor(a);",Kde="return a;",Zde=`if (a < 0.0) { return 1.0/0.0; }
|
|
return log(a);`,Yde="return f32(!(a >= 1.0));",Jde="return -a;",Qde="if (a < 0.0) { return uniforms.alpha * a; } return a;",epe=`
|
|
let aLessThanZero = vec4<f32>(a < vec4<f32>(0.0));
|
|
return (aLessThanZero * (uniforms.alpha * a)) + ((vec4<f32>(1.0) - aLessThanZero) * a);
|
|
`,tpe="return select(a, 0.0, a < 0.0);",rpe="return clamp(a, 0.0, 6.0);",npe="return clamp(a, vec4<f32>(0.0, 0.0, 0.0, 0.0), vec4<f32>(6.0, 6.0, 6.0, 6.0));",ape=`
|
|
return select(a, vec4<f32>(0.0), a < vec4<f32>(0.0));
|
|
`,spe="return 1.0/sqrt(a);",ipe="return 1.0 / (1.0 + exp(-1.0 * a));",ope="return sin(a);",lpe=`
|
|
let e2x = exp(a);
|
|
return (e2x - 1.0 / e2x) / 2.0;
|
|
`,upe="return sqrt(a);",dpe="return a * a;",ppe=`
|
|
let e2x = exp(-2.0 * abs(a));
|
|
return sign(a) * (1.0 - e2x) / (1.0 + e2x);
|
|
`,hpe="return f32(i32((a)));";function vo(e,t){switch(e){case 0:return Bde;case 2:return Vde;case 3:return Ude;case 1:return Wde;case 4:return t?Hde:jde;case 5:return qde;case 6:return Gde;case 7:return Xde;case 8:return Kde;case 9:return Zde;case 10:return Yde;case 11:return Jde;case 14:return t?epe:Qde;case 12:return t?ape:tpe;case 13:return t?npe:rpe;case 15:return spe;case 18:return ipe;case 16:return ope;case 17:return lpe;case 19:return upe;case 20:return dpe;case 21:return ppe;case 22:return hpe;default:throw new Error(`BinaryType ${e} is not implemented!`)}}function Zi(e,t=!1){if(e===null)return null;if(e==="linear")return vo(8);if(e==="relu")return vo(12,t);if(e==="elu")return vo(4,t);if(e==="relu6")return vo(13,t);if(e==="prelu")return Gh(14,t);if(e==="sigmoid")return vo(18,t);if(e==="leakyrelu")return vo(14,t);throw new Error(`Activation ${e} has not been implemented for the WebGPU backend.`)}function cpe(e,t){if(Math.max(...e)>3)throw new Error("Cannot symbolically compute strides for rank > 4 tensor.");let r=e.length,n=e.map(s=>`${t}[${s}]`),a=new Array(r-1);a[r-2]=n[r-1];for(let s=r-3;s>=0;--s)a[s]=`(${a[s+1]} * ${n[s+1]})`;return a}function xr(e){if(e<=1)return"i32";if(e===2)return"vec2<i32>";if(e===3)return"vec3<i32>";if(e===4)return"vec4<i32>";if(e===5)return"vec5";if(e===6)return"vec6";throw Error(`GPU for rank ${e} is not yet supported`)}function Ms(e){if(e===0)return"x";if(e===1)return"y";if(e===2)return"z";if(e===3)return"w";if(e===4)return"u";if(e===5)return"v";throw Error(`Index ${e} is not yet supported`)}function f0(e,t){return e==="float32"?t?"vec4<f32>":"f32":e==="int32"||e==="bool"?t?"vec4<i32>":"i32":e}function S5(){return`
|
|
@stage(compute) @workgroup_size(workGroupSizeX, workGroupSizeY, workGroupSizeZ)
|
|
`}function Wl(){return`
|
|
${S5()}
|
|
fn main(@builtin(local_invocation_id) LocalId : vec3<u32>,
|
|
@builtin(global_invocation_id) GlobalId : vec3<u32>,
|
|
@builtin(num_workgroups) NumWorkgroups: vec3<u32>) {
|
|
localId = LocalId;
|
|
globalId = GlobalId;
|
|
numWorkgroups = NumWorkgroups;
|
|
`}function rt(){return`
|
|
${Wl()}
|
|
let index = getGlobalIndex();
|
|
`}function fpe(e,t,r,n=!1){let a=[];if(a.push(`
|
|
let workGroupSizeX = ${r.workGroupSize[0]}u;
|
|
let workGroupSizeY = ${r.workGroupSize[1]}u;
|
|
let workGroupSizeZ = ${r.workGroupSize[2]}u;
|
|
|
|
var<private> localId: vec3<u32>;
|
|
var<private> globalId: vec3<u32>;
|
|
var<private> numWorkgroups: vec3<u32>;
|
|
|
|
// Only used when the y/z dimension of workgroup size is 1.
|
|
fn getGlobalIndex() -> i32 {
|
|
if (numWorkgroups.y == 1u && numWorkgroups.z == 1u) {
|
|
return i32(globalId.x);
|
|
}
|
|
|
|
let localInvocationIndex = localId.z * workGroupSizeX * workGroupSizeY +
|
|
localId.y * workGroupSizeX + localId.x;
|
|
let workGroupID = (globalId - localId)/vec3<u32>(
|
|
workGroupSizeX, workGroupSizeY, workGroupSizeZ);
|
|
|
|
return i32((workGroupID.z * numWorkgroups.x * numWorkgroups.y +
|
|
workGroupID.y * numWorkgroups.x + workGroupID.x) *
|
|
(workGroupSizeX * workGroupSizeY * workGroupSizeZ) +
|
|
localInvocationIndex);
|
|
}
|
|
`),n===!0)return a.push(`
|
|
struct Uniform {
|
|
size : i32,
|
|
numChannels : i32,
|
|
outShapeStrides : vec2<i32>,
|
|
dispatchSize : vec3<u32>,
|
|
};
|
|
|
|
@group(0) @binding(0) var<storage, write> result: array<${f0(t.dtype,r.isVec4)}>;
|
|
@group(0) @binding(2) var<uniform> uniforms: Uniform;
|
|
`),[ov,a.join(`
|
|
`),lv(t.shape),r.getUserCode()].join(`
|
|
`);let s=!1,i=!1,o="struct Uniforms { NAN : f32, ";r.variableNames.forEach((f,m)=>{let g=xr(e[m].shape.length);(g==="vec5"||g==="vec6")&&(i=!0),(s||i)&&(o+="@align(16) "),s=i,o+=`${f.charAt(0).toLowerCase()+f.slice(1)}Shape : ${g}, `});let l=xr(t.shape.length);i=l==="vec5"||l==="vec6",(s||i)&&(o+="@align(16) "),s=i,o+=`outShape : ${l}, `;let u=t.shape.length-1,d=xr(u);i=d==="vec5"||d==="vec6",(s||i)&&(o+="@align(16) "),s=i,o+=`
|
|
outShapeStrides: ${d}, `,r.size&&(s&&(o+="@align(16) "),s=!1,o+="size : i32, "),r.uniforms&&(s&&(o+="@align(16) "),o+=r.uniforms),o+="};",a.push(o),r.atomic?a.push(`
|
|
@group(0) @binding(0) var<storage, read_write> result: array<atomic<i32>>;
|
|
`):a.push(`
|
|
@group(0) @binding(0) var<storage, write> result: array<${f0(t.dtype,r.isVec4)}>;
|
|
`),r.variableNames.forEach((f,m)=>{a.push(`
|
|
@group(0) @binding(${1+m}) var<storage, read> ${f}: array<${f0(e[m].dtype,r.isVec4)}>;
|
|
`)}),o!==""&&a.push(`
|
|
@group(0) @binding(${1+r.variableNames.length}) var<uniform> uniforms: Uniforms;
|
|
`);let[h,p]=bpe(t.shape,r.dispatchLayout),c=[ov,a.join(`
|
|
`),lv(t.shape),h,mpe(t.shape.length)];if(r.atomic||c.push(gpe(t.shape,t.dtype,r.isVec4)),p===t.shape.length){let f=e.map(m=>ype(m,t.shape,r.isVec4,r.dispatchLayout.x.length===t.shape.length)).join(`
|
|
`);c.push(f)}return c.push(r.getUserCode()),c.join(`
|
|
`)}var ov=`
|
|
struct vec5 {x: i32, y: i32, z: i32, w: i32, u: i32};
|
|
struct vec6 {x: i32, y: i32, z: i32, w: i32, u: i32, v: i32};
|
|
|
|
// Checks whether coordinates lie within the bounds of the shape.
|
|
fn coordsInBounds2D(coord : vec2<i32>, shape : vec2<i32>) -> bool {
|
|
return all(coord >= vec2<i32>(0)) && all(coord < shape);
|
|
}
|
|
fn coordsInBounds3D(coord : vec3<i32>, shape : vec3<i32>) -> bool {
|
|
return all(coord >= vec3<i32>(0)) && all(coord < shape);
|
|
}
|
|
fn coordsInBounds4D(coord : vec4<i32>, shape : vec4<i32>) -> bool {
|
|
return all(coord >= vec4<i32>(0)) && all(coord < shape);
|
|
}
|
|
|
|
fn getIndexFromCoords1D(coord : i32, shape : i32) -> i32 {
|
|
return coord;
|
|
}
|
|
fn getIndexFromCoords2D(coords : vec2<i32>, shape : vec2<i32>) -> i32 {
|
|
return dot(coords, vec2<i32>(shape.y, 1));
|
|
}
|
|
fn getIndexFromCoords3D(coords : vec3<i32>, shape : vec3<i32>) -> i32 {
|
|
return dot(coords, vec3<i32>(shape.y * shape.z, shape.z, 1));
|
|
}
|
|
fn getIndexFromCoords4D(coords : vec4<i32>, shape : vec4<i32>) -> i32 {
|
|
return dot(coords, vec4<i32>(
|
|
shape.y * shape.z * shape.w, shape.z * shape.w, shape.w, 1));
|
|
}
|
|
fn getIndexFromCoords5D(coords : vec5, shape : vec5) -> i32 {
|
|
let shapeStrides: vec5 = vec5(shape.y * shape.z * shape.w * shape.u, shape.z * shape.w * shape.u, shape.w * shape.u, shape.u, 1);
|
|
return coords.x*shapeStrides.x + coords.y*shapeStrides.y + coords.z*shapeStrides.z + coords.w*shapeStrides.w + coords.u*shapeStrides.u;
|
|
}
|
|
fn getIndexFromCoords6D(coords : vec6, shape : vec6) -> i32 {
|
|
let shapeStrides: vec6 = vec6(shape.y * shape.z * shape.w * shape.u * shape.v, shape.z * shape.w * shape.u * shape.v, shape.w * shape.u * shape.v, shape.u * shape.v, shape.v, 1);
|
|
return coords.x*shapeStrides.x + coords.y*shapeStrides.y + coords.z*shapeStrides.z + coords.w*shapeStrides.w + coords.u*shapeStrides.u + coords.v*shapeStrides.v;
|
|
}
|
|
|
|
fn idiv(a: i32, b: i32, sign: f32) -> i32 {
|
|
var res: i32 = a / b;
|
|
let mod: i32 = a % b;
|
|
if (sign < 0. && mod != 0) {
|
|
res = res - 1;
|
|
}
|
|
return res;
|
|
}
|
|
|
|
// NaN defination in IEEE 754-1985 is :
|
|
// - sign = either 0 or 1.
|
|
// - biased exponent = all 1 bits.
|
|
// - fraction = anything except all 0 bits (since all 0 bits represents infinity).
|
|
// https://en.wikipedia.org/wiki/IEEE_754-1985#Representation_of_non-numbers
|
|
fn isnan(val: f32) -> bool {
|
|
let floatToUint: u32 = bitcast<u32>(val);
|
|
return (floatToUint & 0x7fffffffu) > 0x7f800000u;
|
|
}
|
|
fn isnanVec4(val : vec4<f32>) -> vec4<bool> {
|
|
return vec4<bool>(isnan(val[0]), isnan(val[1]), isnan(val[2]), isnan(val[3]));
|
|
}
|
|
`;function mpe(e){let t="";switch(e){case 0:case 1:t+=`
|
|
fn getOutputIndexFromCoords(coords : i32) -> i32 {
|
|
return coords;
|
|
}
|
|
`;break;case 2:t+=`
|
|
fn getOutputIndexFromCoords(coords : vec2<i32>) -> i32 {
|
|
return dot(coords, vec2<i32>(uniforms.outShapeStrides, 1));
|
|
}
|
|
`;break;case 3:t+=`
|
|
fn getOutputIndexFromCoords(coords : vec3<i32>) -> i32 {
|
|
return dot(coords, vec3<i32>(uniforms.outShapeStrides.x, uniforms.outShapeStrides.y, 1));
|
|
}
|
|
`;break;case 4:t+=`
|
|
fn getOutputIndexFromCoords(coords : vec4<i32>) -> i32 {
|
|
return dot(coords, vec4<i32>(
|
|
uniforms.outShapeStrides.x, uniforms.outShapeStrides.y, uniforms.outShapeStrides.z, 1));
|
|
}
|
|
`;break;case 5:t+=`
|
|
fn getOutputIndexFromCoords(coords : vec5) -> i32 {
|
|
return coords.x * uniforms.outShapeStrides.x +
|
|
coords.y * uniforms.outShapeStrides.y +
|
|
coords.z * uniforms.outShapeStrides.z +
|
|
coords.w * uniforms.outShapeStrides.w +
|
|
coords.u;
|
|
}
|
|
`;break;case 6:t+=`
|
|
fn getOutputIndexFromCoords(coords : vec6) -> i32 {
|
|
return coords.x * uniforms.outShapeStrides.x +
|
|
coords.y * uniforms.outShapeStrides.y +
|
|
coords.z * uniforms.outShapeStrides.z +
|
|
coords.w * uniforms.outShapeStrides.w +
|
|
coords.u * uniforms.outShapeStrides.u +
|
|
coords.v;
|
|
}
|
|
`;break;default:v.assert(!1,()=>`Unsupported ${e}D shape`);break}return t}function gpe(e,t,r){let n=e.length,a=f0(t,r),s;if(r?s=`fn setOutputAtIndex(flatIndex : i32, value : vec4<f32>) {
|
|
result[flatIndex] = ${a}(value);
|
|
}
|
|
fn setOutputAtIndexI32(flatIndex : i32, value : vec4<i32>) {
|
|
result[flatIndex] = ${a}(value);
|
|
}`:s=`fn setOutputAtIndex(flatIndex : i32, value : f32) {
|
|
result[flatIndex] = ${a}(value);
|
|
}
|
|
fn setOutputAtIndexI32(flatIndex : i32, value : i32) {
|
|
result[flatIndex] = ${a}(value);
|
|
}`,n>=2){let i=["d0","d1","d2","d3","d4","d5"].slice(0,n),o=xr(n);r?s+=`
|
|
fn setOutputAtCoords(${i.map(l=>`${l} : i32`).join(", ")}, value : vec4<f32>) {
|
|
let flatIndex = getOutputIndexFromCoords(${o}(${i.join(", ")}));
|
|
setOutputAtIndex(flatIndex / 4, value);
|
|
}
|
|
fn setOutputAtCoordsI32(${i.map(l=>`${l} : i32`).join(", ")}, value : vec4<i32>) {
|
|
let flatIndex = getOutputIndexFromCoords(${o}(${i.join(", ")}));
|
|
setOutputAtIndexI32(flatIndex / 4, value);
|
|
}
|
|
`:s+=`
|
|
fn setOutputAtCoords(${i.map(l=>`${l} : i32`).join(", ")}, value : f32) {
|
|
let flatIndex = getOutputIndexFromCoords(${o}(${i.join(", ")}));
|
|
setOutputAtIndex(flatIndex, value);
|
|
}
|
|
fn setOutputAtCoordsI32(${i.map(l=>`${l} : i32`).join(", ")}, value : i32) {
|
|
let flatIndex = getOutputIndexFromCoords(${o}(${i.join(", ")}));
|
|
setOutputAtIndexI32(flatIndex, value);
|
|
}
|
|
`}return s}function ype(e,t,r,n){let a=Ape(e,r);return e.shape.length<=t.length&&(a+=xpe(e,t,r,n)),a}function Ape(e,t){let r=e.name,n=e.shape.length,a=xr(n),s="get"+r.charAt(0).toUpperCase()+r.slice(1),i=["d0","d1","d2","d3","d4","d5"].slice(0,n),o=i.map(d=>`${d} : i32`).join(", ");if(n<1)return t?`
|
|
fn ${s}() -> vec4<f32> {
|
|
return vec4<f32>(${r}[0]);
|
|
}
|
|
`:`
|
|
fn ${s}() ->f32 {
|
|
return f32(${r}[0]);
|
|
}
|
|
`;let l=`uniforms.${r.charAt(0).toLowerCase()+r.slice(1)}Shape`,u=`${n}D`;return n===0&&(u="1D"),t?`
|
|
fn ${s}(${o}) -> vec4<f32> {
|
|
return vec4<f32>(${r}[getIndexFromCoords${u}(${a}(${i.join(",")}),
|
|
${l}) / 4]);
|
|
}
|
|
`:`
|
|
fn ${s}(${o}) -> f32 {
|
|
return f32(${r}[getIndexFromCoords${u}(${a}(${i.join(",")}),
|
|
${l})]);
|
|
}
|
|
`}function xpe(e,t,r,n){let a=e.name,s=a.charAt(0).toUpperCase()+a.slice(1),i="get"+s+"ByOutput",o=e.shape.length,l=t.length,u=xr(l);if(v.arraysEqual(e.shape,t)&&n)return r?`
|
|
fn ${i}Index(globalIndex : i32) -> vec4<f32> {
|
|
return vec4<f32>(${a}[globalIndex]);
|
|
}
|
|
|
|
fn ${i}Coords(coords : ${u}) -> vec4<f32> {
|
|
return vec4<f32>(${a}[${l>1?"getOutputIndexFromCoords(coords)":"coords"} / 4]);
|
|
}
|
|
`:`
|
|
fn ${i}Index(globalIndex : i32) -> f32 {
|
|
return f32(${a}[globalIndex]);
|
|
}
|
|
|
|
fn ${i}Coords(coords : ${u}) -> f32 {
|
|
return f32(${a}[${l>1?"getOutputIndexFromCoords(coords)":"coords"}]);
|
|
}
|
|
`;let d=C.getBroadcastDims(e.shape,t),h=l-o,p="";if(o===0)return r?`
|
|
fn ${i}Index(globalIndex : i32) -> vec4<f32> {
|
|
return get${s}();
|
|
}
|
|
|
|
fn ${i}Coords(coords : ${u}) -> vec4<f32> {
|
|
return get${s}();
|
|
}
|
|
`:`
|
|
fn ${i}Index(globalIndex : i32) -> f32{
|
|
return get${s}();
|
|
}
|
|
|
|
fn ${i}Coords(coords : ${u}) -> f32{
|
|
return get${s}();
|
|
}
|
|
`;l<2&&d.length>=1?p="coords = 0;":p=d.map(g=>`coords.${Ms(g+h)} = 0;`).join(`
|
|
`);let c="";if(l<2&&o>0)c="coords";else if(l>1){let g=xr(o),y=e.shape.map((A,x)=>`coords.${Ms(x+h)}`).join(", ");c=`${g}(${y})`}else c="coords";let f=`uniforms.${a.charAt(0).toLowerCase()+a.slice(1)}Shape`,m=`${o}D`;return r?`
|
|
fn ${i}Index(globalIndex : i32) -> vec4<f32> {
|
|
var coords = getCoordsFromIndex(globalIndex);
|
|
${p}
|
|
return ${a}[getIndexFromCoords${m}(${c}, ${f}) / 4];
|
|
}
|
|
|
|
fn ${i}Coords(coordsIn : ${u}) -> vec4<f32> {
|
|
var coords = coordsIn;
|
|
${p}
|
|
return ${a}[getIndexFromCoords${m}(${c}, ${f}) / 4];
|
|
}
|
|
`:`
|
|
fn ${i}Index(globalIndex : i32) -> f32 {
|
|
var coords = getCoordsFromIndex(globalIndex);
|
|
${p}
|
|
return f32(${a}[getIndexFromCoords${m}(${c}, ${f})]);
|
|
}
|
|
|
|
fn ${i}Coords(coordsIn : ${u}) -> f32 {
|
|
var coords = coordsIn;
|
|
${p}
|
|
return f32(${a}[getIndexFromCoords${m}(${c}, ${f})]);
|
|
}
|
|
`}function bpe(e,t){let{x:r,y:n=[],z:a=[]}=t,s=e.length;if(r.length===s)return[`fn getOutputCoords() -> ${xr(s)}{
|
|
let globalIndex = getGlobalIndex();
|
|
return getCoordsFromIndex(globalIndex);
|
|
}
|
|
`,s];let i="",o=[r,n,a],l=0;for(let p=0;p<o.length;p++){let c=o[p];if(c.length!==0)if(l+=c.length,c.length===1)i+=`let d${c[0]} = i32(globalId[${p}]);`;else{let f=cpe(c,"uniforms.outShape");i+=`var index${p} = i32(globalId[${p}]);`;for(let m=0;m<f.length;m++)i+=`let d${c[m]} = index${p} / ${f[m]};`,m===f.length-1?i+=`let d${c[m+1]} = index${p} - d${c[m]} * ${f[m]};`:i+=`index${p} = index${p} - d${c[m]} * ${f[m]};`}}let u=[];for(let p=0;p<l;p++)u.push(`d${p}`);let d=xr(l),h=`fn getOutputCoords() -> ${d} {
|
|
${i}
|
|
`;return u.length===0?h+=`return ${d}(0); }`:h+=`return ${d}(${u.join(",")}); }`,[h,l]}function lv(e){let t=e.length;if(t<=1)return"fn getCoordsFromIndex(index : i32) -> i32 { return index; }";let r=v.computeStrides(e),n=xr(t),a=[];for(let i=0;i<t;i++)a.push(`d${i}`);if(r.length===1)return` fn getCoordsFromIndex(index : i32) -> vec2<i32> {
|
|
let d0 = index / uniforms.outShapeStrides; let d1 = index - d0 * uniforms.outShapeStrides;
|
|
return vec2<i32>(d0, d1);
|
|
}`;let s;return s="var index2 = index;"+r.map((i,o)=>{let l=`let ${a[o]} = index2 / uniforms.outShapeStrides.${Ms(o)}`,u=o===r.length-1?`let ${a[o+1]} = index2 - ${a[o]} * uniforms.outShapeStrides.${Ms(o)}`:`index2 = index2 - ${a[o]} * uniforms.outShapeStrides.${Ms(o)}`;return`${l}; ${u};`}).join(""),`
|
|
fn getCoordsFromIndex(index : i32) -> ${n} {
|
|
${s}
|
|
return ${n}(${a.join(",")});
|
|
}
|
|
`}var W9={};Le(W9,{ArrayBufferToTypedArray:()=>U9,GPUBytesPerElement:()=>m0,computeDispatch:()=>Oe,computeWorkGroupSizeForConv2d:()=>C5,computeWorkGroupSizeForMatMul:()=>V9,computeWorkPerThreadForConv2d:()=>T5,flatDispatchLayout:()=>Ye,isWebGPUSupported:()=>N5,tilesFitEvenlyIntoShape:()=>Za});var Mo=e=>{let t=1;for(let r=0;r<e.length;r++)t*=e[r];return t};function Za(e,t){if(e.length!==t.length)throw new Error(`Cannot compute whether rank ${e.length} tiles fit evenly into rank ${t.length} shape - ranks must match.`);return t.every((r,n)=>r%e[n]===0)}function Oe(e,t,r=[1,1,1],n=[1,1,1]){let[a,s,i]=[Math.ceil(Mo(e.x.map(o=>t[o]))/(r[0]*n[0])),e.y?Math.ceil(Mo(e.y.map(o=>t[o]))/(r[1]*n[1])):1,e.z?Math.ceil(Mo(e.z.map(o=>t[o]))/(r[2]*n[2])):1];return[a,s,i]}function C5(e,t){let r=Mo(e.x.map(a=>t[a])),n=Mo(e.y.map(a=>t[a]));return r<=4?[4,16,1]:n<=4?[16,4,1]:[16,16,1]}function V9(e,t,r){return e===1?[32,1,1]:r===1?[1,32,1]:[8,8,1]}function T5(e,t){let r=Mo(e.x.map(a=>t[a])),n=Mo(e.y.map(a=>t[a]));return r<=4?[1,2,1]:n<=4?[2,1,1]:[2,2,1]}function Ye(e){return{x:e.map((t,r)=>r)}}function m0(e){if(e==="float32"||e==="int32"||e==="bool"||e==="string")return 4;if(e==="complex64")return 8;throw new Error(`Unknown dtype ${e}`)}function U9(e,t){if(t==="float32")return new Float32Array(e);if(t==="int32")return new Int32Array(e);if(t==="bool"||t==="string")return Uint8Array.from(new Int32Array(e));throw new Error(`Unknown dtype ${t}`)}function N5(){return(typeof window!="undefined"||typeof WorkerGlobalScope!="undefined")&&!!navigator.gpu}function G9(e,t,r,n){return v.assert(n%4===0&&e[0]===4,()=>"tileInner must be divisible by 4. And ColPerThread must be 4"),`
|
|
var<workgroup> mm_Asub : array<array<vec4<f32>, ${n/e[0]}>, ${t}>;
|
|
var<workgroup> mm_Bsub : array<array<vec4<f32>, ${r/e[0]}>, ${n}>;
|
|
|
|
let RowPerThread = ${e[1]};
|
|
let ColPerThread = ${e[0]};
|
|
let TileInner = ${n};
|
|
|
|
${Wl()}
|
|
|
|
let tileRow = ${t===1?"0":"i32(localId.y) * RowPerThread"};
|
|
let tileCol = i32(localId.x);
|
|
|
|
let globalRow = ${t===1?"0":"i32(globalId.y) * RowPerThread"};
|
|
let globalCol = i32(globalId.x);
|
|
let numTiles = (uniforms.dimInner - 1) / TileInner + 1;
|
|
|
|
var acc: array<vec4<f32>, RowPerThread>;
|
|
var ACached : vec4<f32>;
|
|
var BCached : array<vec4<f32>, 4>;
|
|
|
|
// Loop over shared dimension.
|
|
var globalColA = tileCol;
|
|
let RowPerThreadB = TileInner / i32(workGroupSizeY);
|
|
let tileRowB = i32(localId.y) * RowPerThreadB;
|
|
for (var t = 0; t < numTiles; t = t + 1) {
|
|
// Load one tile of A into local memory.
|
|
for (var innerRow = 0; innerRow < RowPerThread; innerRow = innerRow + 1) {
|
|
let inputRow = tileRow + innerRow;
|
|
let inputCol = tileCol;
|
|
mm_Asub[inputRow][inputCol] = mm_readA(globalRow + innerRow, globalColA, globalId);
|
|
}
|
|
globalColA = globalColA + TileInner / ColPerThread;
|
|
|
|
// Load one tile of B into local memory.
|
|
for (var innerRow = 0; innerRow < RowPerThreadB; innerRow = innerRow + 1) {
|
|
let inputRow = tileRowB + innerRow;
|
|
let inputCol = tileCol;
|
|
mm_Bsub[inputRow][inputCol] = mm_readB(t * TileInner + inputRow, globalCol, globalId);
|
|
}
|
|
|
|
workgroupBarrier();
|
|
|
|
// Compute acc values for a single thread.
|
|
for (var k = 0; k < TileInner / ColPerThread; k = k + 1) {
|
|
BCached[0] = mm_Bsub[k * ColPerThread][tileCol];
|
|
BCached[1] = mm_Bsub[k * ColPerThread + 1][tileCol];
|
|
BCached[2] = mm_Bsub[k * ColPerThread + 2][tileCol];
|
|
BCached[3] = mm_Bsub[k * ColPerThread + 3][tileCol];
|
|
|
|
for (var i = 0; i < RowPerThread; i = i + 1) {
|
|
ACached = mm_Asub[tileRow + i][k];
|
|
acc[i] = BCached[0] * ACached.x + acc[i];
|
|
acc[i] = BCached[1] * ACached.y + acc[i];
|
|
acc[i] = BCached[2] * ACached.z + acc[i];
|
|
acc[i] = BCached[3] * ACached.w + acc[i];
|
|
}
|
|
}
|
|
|
|
workgroupBarrier();
|
|
}
|
|
|
|
for (var innerRow = 0; innerRow < RowPerThread; innerRow = innerRow + 1) {
|
|
mm_write(globalRow + innerRow,
|
|
globalCol,
|
|
acc[innerRow], globalId);
|
|
}
|
|
}`}var vpe=class{constructor(e,t,r,n,a,s=null,i=null,o=null){this.variableNames=["A","B"],this.uniforms="dimAOuter : i32, dimBOuter : i32, dimInner : i32,",this.workGroupSize=[8,8,1],this.isVec4=!0,this.outputShape=t,this.dispatchLayout={x:[2],y:[1],z:[0]},t[1]===1?this.elementsPerThread=[4,1,1]:this.elementsPerThread=[4,4,1],this.dispatch=Oe(this.dispatchLayout,this.outputShape,this.workGroupSize,this.elementsPerThread);let l=s!=null,u=o!=null;l&&this.variableNames.push("bias"),u&&this.variableNames.push("preluActivationWeights"),this.tileAOuter=t[1]===1?1:this.workGroupSize[1]*this.elementsPerThread[1],this.tileBOuter=this.workGroupSize[0]*this.elementsPerThread[0],this.tileInner=this.tileBOuter,this.aShape=e,this.addBias=l,this.activation=i,this.hasPreluActivationWeights=u,this.batchAEqualOne=n,this.batchBEqualOne=a,[this.fitA,this.fitB]=this.getShapeFit(),this.shaderKey=`matMulPackedVec4_${this.activation}_${this.fitA}_${this.fitB}_${this.elementsPerThread}_${this.batchAEqualOne}_${this.batchBEqualOne}`}getShapeFit(){let e=this.aShape[2],t=this.outputShape[2],r=[this.outputShape[0],e,t],n=[this.tileAOuter,this.tileInner],a=[this.tileInner,this.tileBOuter];return[Za(n,this.aShape.slice(1)),Za(a,r.slice(1))]}getUserCode(){let e=this.fitA?"return A[batch * batchASize + row * uniforms.dimInner / 4 + col]":`if (coordsInBounds2D(vec2<i32>(row, col * 4), vec2<i32>(uniforms.dimAOuter, uniforms.dimInner))) {
|
|
return A[batch * batchASize + row * uniforms.dimInner / 4 + col];
|
|
}
|
|
return vec4<f32>(0.0)`,t=this.fitB?"return B[batch * batchBSize + row * uniforms.dimBOuter / 4 + col]":`if(coordsInBounds2D(vec2<i32>(row, col * 4), vec2<i32>(uniforms.dimInner, uniforms.dimBOuter))) {
|
|
return B[batch * batchBSize + row * uniforms.dimBOuter / 4 + col];
|
|
}
|
|
return vec4<f32>(0.0)`,r="",n="";if(this.activation){let s=Zi(this.activation,this.isVec4);this.hasPreluActivationWeights?r=`fn activation(a : vec4<f32>, outCoord : vec3<i32>) -> vec4<f32> {
|
|
let b = getPreluActivationWeightsByOutputCoords(outCoord);
|
|
${s}
|
|
}`:r=`
|
|
fn activation(a : vec4<f32>, outCoord : vec3<i32>) -> vec4<f32> {
|
|
${s}
|
|
}`,n="value = activation(value, outCoord);"}let a=this.addBias?"value = value + getBiasByOutputCoords(outCoord);":"";return`
|
|
${r}
|
|
fn mm_readA(row : i32, col : i32, globalId : vec3<u32>) -> vec4<f32> {
|
|
${this.batchAEqualOne?`
|
|
let batchASize = 0;
|
|
let batch = 0;
|
|
`:`
|
|
let batchASize = uniforms.aShape[1] * uniforms.aShape[2] / 4;
|
|
let batch = i32(globalId.z);
|
|
`}
|
|
|
|
${e};
|
|
}
|
|
|
|
fn mm_readB(row : i32, col : i32, globalId : vec3<u32>) -> vec4<f32> {
|
|
${this.batchBEqualOne?`
|
|
let batchBSize = 0;
|
|
let batch = 0;
|
|
`:`
|
|
let batchBSize = uniforms.bShape[1] * uniforms.bShape[2] / 4;
|
|
let batch = i32(globalId.z);
|
|
`}
|
|
${t};
|
|
}
|
|
|
|
fn mm_write(row : i32, col : i32, valueIn : vec4<f32>, globalId : vec3<u32>) {
|
|
if (row < uniforms.aShape[1] && col * 4 < uniforms.bShape[2])
|
|
{
|
|
var value = valueIn;
|
|
let batch = i32(globalId.z);
|
|
let outCoord = vec3<i32>(batch, row, col * 4);
|
|
${a}
|
|
${n}
|
|
setOutputAtCoords(outCoord[0], outCoord[1], outCoord[2], value);
|
|
}
|
|
}
|
|
${G9(this.elementsPerThread,this.tileAOuter,this.tileBOuter,this.tileInner)}
|
|
`}};function E5(e,t){let r=t[1]*e[1],n=t[0]*e[0],a=r>n?r:n;return`
|
|
var<workgroup> mm_Asub : array<array<f32, ${a}>, ${r}>;
|
|
var<workgroup> mm_Bsub : array<array<f32, ${n}>, ${a}>;
|
|
${Wl()}
|
|
let tileRow = i32(localId.y) * ${e[1]};
|
|
let tileCol = i32(localId.x) * ${e[0]};
|
|
|
|
let globalRow = i32(globalId.y) * ${e[1]};
|
|
let globalCol = i32(globalId.x) * ${e[0]};
|
|
|
|
let numTiles = (uniforms.dimInner - 1) / ${a} + 1;
|
|
|
|
var acc : array<array<f32, ${e[0]}>, ${e[1]}>;
|
|
var ACached : f32;
|
|
var BCached : array<f32, ${e[0]}>;
|
|
|
|
// Without this initialization strange values show up in acc.
|
|
for (var innerRow = 0; innerRow < ${e[1]}; innerRow = innerRow + 1) {
|
|
for (var innerCol = 0; innerCol < ${e[0]}; innerCol = innerCol + 1) {
|
|
acc[innerRow][innerCol] = 0.0;
|
|
}
|
|
}
|
|
|
|
let ColPerThreadA = ${a} / ${t[0]};
|
|
let tileColA = i32(localId.x) * ColPerThreadA;
|
|
let RowPerThreadB = ${a} / ${t[1]};
|
|
let tileRowB = i32(localId.y) * RowPerThreadB;
|
|
|
|
// Loop over shared dimension.
|
|
for (var t = 0; t < numTiles; t = t + 1) {
|
|
// Load one tile of A into local memory.
|
|
for (var innerRow = 0; innerRow < ${e[1]}; innerRow = innerRow + 1) {
|
|
for (var innerCol = 0; innerCol < ColPerThreadA; innerCol = innerCol + 1) {
|
|
let inputRow = tileRow + innerRow;
|
|
let inputCol = tileColA + innerCol;
|
|
|
|
mm_Asub[inputRow][inputCol] = mm_readA(
|
|
globalRow + innerRow,
|
|
t * ${a} + inputCol, globalId);
|
|
}
|
|
}
|
|
// Load one tile of B into local memory.
|
|
for (var innerRow = 0; innerRow < RowPerThreadB; innerRow = innerRow + 1) {
|
|
for (var innerCol = 0; innerCol < ${e[0]}; innerCol = innerCol + 1) {
|
|
let inputRow = tileRowB + innerRow;
|
|
let inputCol = tileCol + innerCol;
|
|
|
|
mm_Bsub[inputRow][inputCol] = mm_readB(
|
|
t * ${a} + inputRow,
|
|
globalCol + innerCol, globalId);
|
|
}
|
|
}
|
|
|
|
workgroupBarrier();
|
|
|
|
// Compute acc values for a single thread.
|
|
for (var k = 0; k < ${a}; k = k + 1) {
|
|
for (var inner = 0; inner < ${e[0]}; inner = inner + 1) {
|
|
BCached[inner] = mm_Bsub[k][tileCol + inner];
|
|
}
|
|
|
|
for (var innerRow = 0; innerRow < ${e[1]}; innerRow = innerRow + 1) {
|
|
ACached = mm_Asub[tileRow + innerRow][k];
|
|
for (var innerCol = 0; innerCol < ${e[0]}; innerCol = innerCol + 1) {
|
|
acc[innerRow][innerCol] = acc[innerRow][innerCol] + ACached * BCached[innerCol];
|
|
}
|
|
}
|
|
}
|
|
|
|
workgroupBarrier();
|
|
}
|
|
|
|
for (var innerRow = 0; innerRow < ${e[1]}; innerRow = innerRow + 1) {
|
|
for (var innerCol = 0; innerCol < ${e[0]}; innerCol = innerCol + 1) {
|
|
|
|
if ((globalCol + innerCol) < uniforms.dimBOuter &&
|
|
(globalRow + innerRow) < uniforms.dimAOuter) {
|
|
mm_write(globalRow + innerRow,
|
|
globalCol + innerCol,
|
|
acc[innerRow][innerCol], globalId);
|
|
}
|
|
}
|
|
}
|
|
}
|
|
`}function wpe(e){return`
|
|
let TileSize = ${e[0]*4};
|
|
var<workgroup> mm_Asub : array<vec4<f32>, ${e[0]}>;
|
|
|
|
${Wl()}
|
|
let tileCol = i32(localId.x);
|
|
let globalCol = i32(globalId.x);
|
|
let globalRow = i32(globalId.y);
|
|
|
|
let numTiles = (uniforms.dimInner - 1) / TileSize + 1;
|
|
|
|
// Without this initialization strange values show up in acc.
|
|
var acc = 0.0;
|
|
|
|
// Loop over shared dimension.
|
|
for (var t = 0; t < numTiles; t = t + 1) {
|
|
// Load one tile of A into local memory.
|
|
let colA = t * TileSize + tileCol * 4;
|
|
mm_Asub[tileCol] = vec4<f32>(mm_readA(globalRow, colA, globalId),
|
|
mm_readA(globalRow, colA + 1, globalId),
|
|
mm_readA(globalRow, colA + 2, globalId),
|
|
mm_readA(globalRow, colA + 3, globalId));
|
|
workgroupBarrier();
|
|
|
|
// Compute acc values for a single thread.
|
|
for (var k = 0; k < TileSize / 4; k = k + 1) {
|
|
let rowB = t * TileSize + k * 4;
|
|
let BCached = vec4<f32>(mm_readB(rowB, globalCol, globalId),
|
|
mm_readB(rowB + 1, globalCol, globalId),
|
|
mm_readB(rowB + 2, globalCol, globalId),
|
|
mm_readB(rowB + 3, globalCol, globalId));
|
|
|
|
let ACached = mm_Asub[k];
|
|
acc = acc + dot(ACached, BCached);
|
|
}
|
|
|
|
workgroupBarrier();
|
|
}
|
|
|
|
if (globalRow < uniforms.dimAOuter && globalCol < uniforms.dimBOuter) {
|
|
mm_write(globalRow, globalCol, acc, globalId);
|
|
}
|
|
}
|
|
`}var kpe=class{constructor(e,t,r,n,a,s=!1,i=!1,o=null,l=null,u=null){this.variableNames=["A","B"],this.uniforms="dimAOuter : i32, dimBOuter : i32, dimInner : i32,",this.workGroupSize=[16,16,1],this.outputShape=t,this.dispatchLayout={x:[2],y:[1],z:[0]};let d=s?e[1]:e[2];this.workGroupSize=V9(t[1],d,t[2]),(t[1]===1||t[2]===1)&&(r=1),this.dispatch=Oe(this.dispatchLayout,this.outputShape,this.workGroupSize,[r,r,1]),v.arraysEqual(this.dispatch,[1,1,1])&&(r=1,this.dispatch=Oe(this.dispatchLayout,this.outputShape,this.workGroupSize,[r,r,1]));let h=o!=null,p=u!=null;h&&this.variableNames.push("bias"),p&&this.variableNames.push("preluActivationWeights"),this.workPerThread=r,this.aShape=e,this.transposeA=s,this.transposeB=i,this.addBias=h,this.activation=l,this.hasPreluActivationWeights=p,this.batchAEqualOne=n,this.batchBEqualOne=a;let c=this.outputShape[2],f=this.transposeB?[this.outputShape[0],c,d]:[this.outputShape[0],d,c];[this.fitA,this.fitB]=this.getShapeFit(f),this.shaderKey=`matMulPacked_${this.workPerThread}_${s}_${i}_${this.activation}_${this.fitA}_${this.fitB}_${this.outputShape[1]>1}_${this.batchAEqualOne}_${this.batchBEqualOne}`}getShapeFit(e){let t=this.workGroupSize[1]*this.workPerThread,r=this.workGroupSize[0]*this.workPerThread,n=t>r?t:r;this.outputShape[1]===1&&(n*=4),v.assert(n%this.workGroupSize[0]===0&&n%this.workGroupSize[1]===0,()=>"tileInner must be multiple of workgroupsize.x and workgroupsize.y");let a=[t,n],s=[n,r];return[Za(a,this.aShape.slice(1)),Za(s,e.slice(1))]}getUserCode(){let e;this.transposeA===!1?e=this.fitA?"return A[batch * batchASize + row * uniforms.dimInner + col];":`if(coordsInBounds2D(vec2<i32>(row, col), vec2<i32>(uniforms.dimAOuter, uniforms.dimInner))) {
|
|
return A[batch * batchASize + row * uniforms.dimInner + col];
|
|
}
|
|
return 0.0;`:e=this.fitA?"return A[batch * batchASize + col * uniforms.dimAOuter + row];":`if(coordsInBounds2D(vec2<i32>(row, col), vec2<i32>(uniforms.dimAOuter, uniforms.dimInner))) {
|
|
return A[batch* batchASize + col * uniforms.dimAOuter + row];
|
|
}
|
|
return 0.0;`;let t;this.transposeB===!1?t=this.fitB?"return B[batch * batchBSize + row * uniforms.dimBOuter + col];":`if(coordsInBounds2D(vec2<i32>(row, col), vec2<i32>(uniforms.dimInner, uniforms.dimBOuter))) {
|
|
return B[batch * batchBSize + row * uniforms.dimBOuter + col];
|
|
}
|
|
return 0.0;`:t=this.fitB?"return B[batch * batchBSize + col * uniforms.dimInner + row];":`if(coordsInBounds2D(vec2<i32>(row, col), vec2<i32>(uniforms.dimInner, uniforms.dimBOuter))) {
|
|
return B[batch * batchBSize + col * uniforms.dimInner + row];
|
|
}
|
|
return 0.0;`;let r="",n="";if(this.activation){let s=Zi(this.activation,!1);this.hasPreluActivationWeights?r=`fn activation(a : f32, outCoord : vec3<i32>) -> f32 {
|
|
let b = getPreluActivationWeightsByOutputCoords(outCoord);
|
|
${s}
|
|
}`:r=`
|
|
fn activation(a : f32, outCoord : vec3<i32>) -> f32 {
|
|
${s}
|
|
}
|
|
`,n="value = activation(value, outCoord);"}let a=this.addBias?"value = value + getBiasByOutputCoords(outCoord);":"";return`
|
|
${r}
|
|
|
|
fn mm_readA(row : i32, col : i32, globalId : vec3<u32>) -> f32 {
|
|
${this.batchAEqualOne?`
|
|
let batch = 0;
|
|
let batchASize = 0;
|
|
`:`
|
|
let batch = i32(globalId.z);
|
|
let batchASize = uniforms.aShape[1] * uniforms.aShape[2];
|
|
`}
|
|
${e}
|
|
}
|
|
|
|
fn mm_readB(row : i32, col : i32, globalId : vec3<u32>) -> f32 {
|
|
${this.batchBEqualOne?`
|
|
let batch = 0;
|
|
let batchBSize = 0;
|
|
`:`
|
|
let batch = i32(globalId.z);
|
|
let batchBSize = uniforms.bShape[1] * uniforms.bShape[2];
|
|
`}
|
|
${t}
|
|
}
|
|
|
|
fn mm_write(row : i32, col : i32, valueIn : f32, globalId : vec3<u32>) {
|
|
var value = valueIn;
|
|
let batch = i32(globalId.z);
|
|
let outCoord = vec3<i32>(batch, row, col);
|
|
${a}
|
|
${n}
|
|
setOutputAtCoords(batch, row, col, value);
|
|
}
|
|
${this.outputShape[1]>1?E5([this.workPerThread,this.workPerThread,1],this.workGroupSize):wpe(this.workGroupSize)}
|
|
`}};function Ipe(){return`
|
|
var<workgroup> sumValues : array<f32, workGroupSizeX>;
|
|
${Wl()}
|
|
let coords = getOutputCoords();
|
|
let batch = coords[0];
|
|
let row = coords[1];
|
|
let col = coords[2];
|
|
var sum = 0.0;
|
|
let Length = uniforms.dimInner;
|
|
for (var k = i32(localId.x); k < Length; k = k + i32(workGroupSizeX)) {
|
|
let dataA = mm_readA(batch, row, k);
|
|
let dataB = mm_readB(batch, k, col);
|
|
sum = sum + dataA * dataB;
|
|
}
|
|
sumValues[localId.x] = sum;
|
|
workgroupBarrier();
|
|
|
|
for(var currentSize = workGroupSizeX / 2u; currentSize > 1u;
|
|
currentSize = currentSize / 2u) {
|
|
if (localId.x < currentSize)
|
|
{
|
|
sumValues[localId.x] = sumValues[localId.x] + sumValues[localId.x + currentSize];
|
|
}
|
|
workgroupBarrier();
|
|
}
|
|
|
|
if (localId.x == 0u) {
|
|
sum = sumValues[0] + sumValues[1];
|
|
mm_write(batch, row, col, sum);
|
|
}
|
|
}
|
|
`}var Spe=class{constructor(e,t,r,n=!1,a=!1,s=null,i=null,o=null){this.variableNames=["A","B"],this.uniforms="dimAOuter : i32, dimBOuter : i32, dimInner : i32,",this.workGroupSize=[256,1,1],this.outputShape=e,this.dispatchLayout={x:[],y:[1,2],z:[0]},this.dispatch=Oe(this.dispatchLayout,this.outputShape,this.workGroupSize);let l=s!=null,u=o!=null;l&&this.variableNames.push("bias"),u&&this.variableNames.push("preluActivationWeights"),this.transposeA=n,this.transposeB=a,this.addBias=l,this.activation=i,this.hasPreluActivationWeights=u,this.batchAEqualOne=t,this.batchBEqualOne=r,this.shaderKey=`matMulReduce_${this.activation}_${n}_${a}_${this.batchAEqualOne}_${this.batchBEqualOne}`}getUserCode(){let e;this.transposeA===!1?e="return f32(A[batch * batchASize + row * uniforms.dimInner + col]);":e="return f32(A[batch * batchASize + col * uniforms.dimAOuter + row]);";let t;this.transposeB===!1?t="return f32(B[batch * batchBSize + row * uniforms.dimBOuter + col]);":t="return f32(B[batch * batchBSize + col * uniforms.dimInner + row]);";let r="",n="";if(this.activation){let s=Zi(this.activation,!1);this.hasPreluActivationWeights?r=`fn activation(a : f32, outCoord : vec3<i32>) -> f32 {
|
|
let b = getPreluActivationWeightsByOutputCoords(outCoord);
|
|
${s}
|
|
}`:r=`
|
|
fn activation(a : f32, outCoord : vec3<i32>) -> f32 {
|
|
${s}
|
|
}
|
|
`,n="value = activation(value, outCoord);"}let a=this.addBias?"value = value + getBiasByOutputCoords(outCoord);":"";return`
|
|
${r}
|
|
|
|
fn mm_readA(batchIn: i32, row : i32, col : i32) -> f32 {
|
|
${this.batchAEqualOne?`
|
|
let batchASize = 0;
|
|
let batch = 0;
|
|
`:`
|
|
let batchASize = uniforms.aShape[1] * uniforms.aShape[2];
|
|
let batch = batchIn;
|
|
`}
|
|
${e}
|
|
}
|
|
|
|
fn mm_readB(batchIn: i32, row : i32, col : i32) -> f32 {
|
|
${this.batchBEqualOne?`
|
|
let batch = 0;
|
|
let batchBSize = 0;
|
|
`:`
|
|
let batch = batchIn;
|
|
let batchBSize = uniforms.bShape[1] * uniforms.bShape[2];
|
|
`}
|
|
${t}
|
|
}
|
|
|
|
fn mm_write(batch: i32, row : i32, col : i32, valueIn : f32) {
|
|
var value = valueIn;
|
|
let outCoord = vec3<i32>(batch, row, col);
|
|
${a}
|
|
${n}
|
|
setOutputAtCoords(batch, row, col, value);
|
|
}
|
|
${Ipe()}
|
|
`}};function Cpe(e){let t=e[1]/2,r=e[0],n=t>r?t:r;return`
|
|
var<workgroup> mm_Asub1 : array<array<f32, ${n}>, ${t}>;
|
|
var<workgroup> mm_Bsub1 : array<array<f32, ${r}>, ${n}>;
|
|
var<workgroup> mm_Asub2 : array<array<f32, ${n}>, ${t}>;
|
|
var<workgroup> mm_Bsub2 : array<array<f32, ${r}>, ${n}>;
|
|
|
|
// If the output size is small for matrix multiplication, avoid to use vec4
|
|
// and handle some elements per thread to optimally utilize the ALU.
|
|
// Introduces two shared memory buffers, some logical threads could handle
|
|
// arithmetic operations and others handle IO operations between barrier api,
|
|
// makes ALUs and load/store units work simultaneously, could improves
|
|
// the performance.
|
|
${Wl()}
|
|
let tileRow = i32(localId.y);
|
|
let tileCol = i32(localId.x);
|
|
let globalRow = i32(globalId.y);
|
|
let globalCol = i32(globalId.x);
|
|
|
|
// uniforms.dimInner should be greater than 0.
|
|
let numTiles = (uniforms.dimInner - 1) / ${n} + 1;
|
|
var acc = 0.0;
|
|
|
|
var globalColA = tileCol;
|
|
var globalRowB = tileRow;
|
|
for (var t = 0; t < numTiles; t = t + 1) {
|
|
if (t == 0) {
|
|
if (tileRow < ${t}) {
|
|
// Load one tile of A and B into local memory.
|
|
// globalRow is always greater than or equal tileRow.
|
|
mm_Asub1[tileRow][tileCol] =
|
|
mm_readA((globalRow - tileRow) / 2 + tileRow, globalColA, globalId);
|
|
globalColA = globalColA + ${n};
|
|
mm_Bsub1[tileRow][tileCol] = mm_readB(globalRowB, globalCol, globalId);
|
|
globalRowB = globalRowB + ${n};
|
|
}
|
|
} else {
|
|
if (tileRow < ${t}) {
|
|
// Load one tile of A and B into local memory.
|
|
// globalRow is always greater than or equal tileRow.
|
|
mm_Asub1[tileRow][tileCol] =
|
|
mm_readA((globalRow - tileRow) / 2 + tileRow, globalColA, globalId);
|
|
globalColA = globalColA + ${n};
|
|
mm_Bsub1[tileRow][tileCol] = mm_readB(globalRowB, globalCol, globalId);
|
|
globalRowB = globalRowB + ${n};
|
|
} else {
|
|
// Compute acc values for a single thread.
|
|
for (var k = 0; k < ${n}; k = k + 1) {
|
|
let subRow = tileRow - ${t};
|
|
if (subRow < 0) {
|
|
continue;
|
|
}
|
|
acc = acc + mm_Asub2[subRow][k] * mm_Bsub2[k][tileCol];
|
|
}
|
|
}
|
|
}
|
|
workgroupBarrier();
|
|
if (t != 0) {
|
|
t = t + 1;
|
|
}
|
|
|
|
if (t < numTiles) {
|
|
if (tileRow < ${t}) {
|
|
// Load one tile of A and B into local memory.
|
|
// globalRow is always greater than or equal tileRow.
|
|
mm_Asub2[tileRow][tileCol] =
|
|
mm_readA((globalRow - tileRow) / 2 + tileRow, globalColA, globalId);
|
|
globalColA = globalColA + ${n};
|
|
mm_Bsub2[tileRow][tileCol] = mm_readB(globalRowB, globalCol, globalId);
|
|
globalRowB = globalRowB + ${n};
|
|
} else {
|
|
// Compute acc values for a single thread.
|
|
for (var k = 0; k < ${n}; k = k + 1) {
|
|
let subRow = tileRow - ${t};
|
|
if (subRow < 0) {
|
|
continue;
|
|
}
|
|
acc = acc + mm_Asub1[subRow][k] * mm_Bsub1[k][tileCol];
|
|
}
|
|
}
|
|
}
|
|
workgroupBarrier();
|
|
}
|
|
let writeCol = (globalRow - tileRow) / 2 + tileRow - ${t};
|
|
if (tileRow >= ${t} && writeCol >= 0) {
|
|
mm_write(writeCol, globalCol, acc, globalId);
|
|
}
|
|
}
|
|
`}var Tpe=class{constructor(e,t,r,n=null,a=null,s=null){this.variableNames=["A","B"],this.uniforms="dimAOuter : i32, dimBOuter : i32, dimInner : i32,",this.workGroupSize=[8,16,1],v.assert(e[1]<=16||t[2]<=16,()=>"This program can be only used when A width or B Height are small"),this.outputShape=r,this.dispatchLayout={x:[2],y:[1],z:[0]},this.dispatch=[Math.ceil(r[2]/this.workGroupSize[0]),Math.ceil(r[1]*2/this.workGroupSize[1]),r[0]];let i=n!=null;i&&this.variableNames.push("bias");let o=s!=null;o&&this.variableNames.push("preluActivationWeights"),this.addBias=i,this.activation=a,this.hasPreluActivationWeights=o,this.batchAEqualOne=e[0]===1,this.batchBEqualOne=t[0]===1,this.shaderKey=`matMulSmallOutputSize_${this.activation}_${this.batchAEqualOne}_${this.batchBEqualOne}`}getUserCode(){let e=`if (coordsInBounds2D(vec2<i32>(row, col), vec2<i32>(uniforms.dimAOuter, uniforms.dimInner))) {
|
|
return A[batch * batchASize + row * uniforms.dimInner + col];
|
|
}
|
|
return 0.0;`,t=`if (coordsInBounds2D(vec2<i32>(row, col), vec2<i32>(uniforms.dimInner, uniforms.dimBOuter))) {
|
|
return B[batch * batchBSize + row * uniforms.dimBOuter + col];
|
|
}
|
|
return 0.0;`,r="",n="";if(this.activation){let s=Zi(this.activation,!1);this.hasPreluActivationWeights?r=`fn activation(a : f32, outCoord : vec3<i32>) -> f32 {
|
|
let b = getPreluActivationWeightsByOutputCoords(outCoord);
|
|
${s}
|
|
}`:r=`fn activation(a : f32, outCoord : vec3<i32>) -> f32 {
|
|
${s}
|
|
}`,n="value = activation(value, outCoord);"}let a=this.addBias?"value = value + getBiasByOutputCoords(outCoord);":"";return`
|
|
${r}
|
|
|
|
fn mm_readA(row : i32, col : i32, globalId : vec3<u32>) -> f32 {
|
|
${this.batchAEqualOne?`
|
|
let batch = 0;
|
|
let batchASize = 0;
|
|
`:`
|
|
let batchASize = uniforms.aShape[1] * uniforms.aShape[2];
|
|
let batch = i32(globalId.z);
|
|
`}
|
|
${e}
|
|
}
|
|
fn mm_readB(row : i32, col : i32, globalId : vec3<u32>) -> f32 {
|
|
${this.batchBEqualOne?`
|
|
let batch = 0;
|
|
let batchBSize = 0;
|
|
`:`
|
|
let batch = i32(globalId.z);
|
|
let batchBSize = uniforms.bShape[1] * uniforms.bShape[2];
|
|
`}
|
|
${t}
|
|
}
|
|
fn mm_write(row : i32, col : i32, valueIn : f32, globalId : vec3<u32>) {
|
|
if (coordsInBounds2D(vec2<i32>(row, col), vec2<i32>(uniforms.dimAOuter, uniforms.dimBOuter))) {
|
|
let batch = i32(globalId.z);
|
|
let outCoord = vec3<i32>(batch, row, col);
|
|
var value = valueIn;
|
|
${a}
|
|
${n}
|
|
setOutputAtCoords(batch, row, col, value);
|
|
}
|
|
}
|
|
${Cpe(this.workGroupSize)}
|
|
`}};function Qe(e){let{inputs:t,attrs:r}=e,{x:n}=t,{shape:a}=r,s=v.sizeFromShape(n.shape),i=v.inferFromImplicitShape(a,s),o=v.sizeFromShape(i);return v.assert(s===o,()=>`The new shape (${i}) has ${o} elements and the old shape (${n.shape}) has ${s} elements. The new shape and old shape must have the same number of elements.`),e.backend.incRef(n.dataId),{dataId:n.dataId,shape:i,dtype:n.dtype}}var Npe={kernelName:ml,backendName:"webgpu",kernelFunc:Qe};function R5({a:e,b:t,transposeA:r,transposeB:n,backend:a,bias:s=null,preluActivationWeights:i=null,leakyreluAlpha:o=0,activation:l=null}){let u=e.shape.length,d=t.shape.length,h=r?e.shape[u-2]:e.shape[u-1],p=n?t.shape[d-1]:t.shape[d-2],c=r?e.shape[u-1]:e.shape[u-2],f=n?t.shape[d-2]:t.shape[d-1],m=e.shape.slice(0,-2),g=t.shape.slice(0,-2),y=v.sizeFromShape(m),A=v.sizeFromShape(g),x=ji.assertAndGetBroadcastShape(e.shape.slice(0,-2),t.shape.slice(0,-2)).concat([c,f]);v.assert(h===p,()=>`Error in matMul: inner shapes (${h}) and (${p}) of Tensors with shapes ${e.shape} and ${t.shape} and transposeA=${r} and transposeB=${n} must match.`);let b=r?[y,h,c]:[y,c,h],w=n?[A,f,p]:[A,p,f],S=Qe({inputs:{x:e},backend:a,attrs:{shape:b}}),N=Qe({inputs:{x:t},backend:a,attrs:{shape:w}}),E=[S,N],R=Math.max(y,A),z=y===1,M=A===1,I=h%4===0&&f%4===0&&!r&&!n,_;c*f<=32?_=new Spe([R,c,f],z,M,r,n,s,l,i):!r&&!n&&(c<=16&&(f<=512||p>=2*f)||f<=16&&(c<=512||h>=2*c))?_=new Tpe(b,w,[R,c,f],s,l,i):I?_=new vpe(b,[R,c,f],Z().get("WEBGPU_MATMUL_WORK_PER_THREAD"),z,M,s,l,i):_=new kpe(b,[R,c,f],Z().get("WEBGPU_MATMUL_WORK_PER_THREAD"),z,M,r,n,s,l,i);let O=[S,N];s&&O.push(s),i&&O.push(i);let q=[{type:"int32",data:[c]},{type:"int32",data:[f]},{type:"int32",data:[h]}];l==="leakyrelu"&&(q.push({type:"float32",data:[o]}),_.uniforms+=" alpha : f32,");let K=a.runWebGPUProgram(_,O,e.dtype,q),D=Qe({inputs:{x:K},backend:a,attrs:{shape:x}});E.push(K);for(let Q of E)a.disposeData(Q.dataId);return D}function Epe(e){let{inputs:t,backend:r,attrs:n}=e,{a,b:s,bias:i,preluActivationWeights:o}=t,{transposeA:l,transposeB:u,activation:d,leakyreluAlpha:h}=n;return R5({a,b:s,transposeA:l,transposeB:u,backend:r,bias:i,preluActivationWeights:o,leakyreluAlpha:h,activation:d})}var Rpe={kernelName:Ps,backendName:"webgpu",kernelFunc:Epe},uv=class{constructor(e,t,r){this.variableNames=["AReal","AImag","BReal","BImag"],this.workGroupSize=[128,1,1],this.size=!0,this.outputShape=C.assertAndGetBroadcastShape(t,r),this.dispatchLayout=Ye(this.outputShape),this.dispatch=Oe(this.dispatchLayout,this.outputShape,this.workGroupSize),this.shaderKey=`binaryOpComplex_${e}`,this.op=e}getUserCode(){return`
|
|
fn binaryOpComplex(
|
|
areal : f32, aimag : f32, breal : f32, bimag : f32) -> f32 {
|
|
${Gh(this.op,!1)}
|
|
}
|
|
|
|
${rt()}
|
|
if(index < uniforms.size) {
|
|
let areal = getARealByOutputIndex(index);
|
|
let aimag = getAImagByOutputIndex(index);
|
|
let breal = getBRealByOutputIndex(index);
|
|
let bimag = getBImagByOutputIndex(index);
|
|
setOutputAtIndex(index, binaryOpComplex(areal, aimag, breal, bimag));
|
|
}
|
|
}
|
|
`}},$pe=class{constructor(e,t,r,n){this.variableNames=["A","B"],this.size=!0;let a=256;this.workGroupSize=[a,1,1],this.outputShape=C.assertAndGetBroadcastShape(t,r),this.dispatchLayout=Ye(this.outputShape),this.lastDimensionSize=n?r[0]:t[0],this.lastDimensionSize<256?this.workPerThread=1:this.lastDimensionSize<512?this.workPerThread=2:this.workPerThread=4,this.dispatch=Oe(this.dispatchLayout,this.outputShape,this.workGroupSize,[this.workPerThread,1,1]),this.useSharedMemoryWithB=n,this.op=e,this.shaderKey=`binaryShared_${e}_${this.lastDimensionSize}_${this.useSharedMemoryWithB}`}getUserCode(){let e=this.lastDimensionSize>1?`coords[${this.outputShape.length-1}]`:"0",t=this.useSharedMemoryWithB?`let a = getAByOutputCoords(coords);
|
|
let b = sharedBuf[${e}];`:`let a = sharedBuf[${e}];
|
|
let b = getBByOutputCoords(coords);`;return`
|
|
fn binaryOperation(a : f32, b : f32) -> f32 {
|
|
${Gh(this.op,!1)}
|
|
}
|
|
var<workgroup> sharedBuf : array<f32, ${this.lastDimensionSize}>;
|
|
${rt()}
|
|
|
|
// Fill in the shared memory buffer. Here we need a loop to make sure
|
|
// that all data in A|B are uploaded when |sharedMemorySize| is larger
|
|
// than work group size.
|
|
for(var localIndex = i32(localId.x); localIndex < ${this.lastDimensionSize}; localIndex = localIndex + ${this.workGroupSize[0]}) {
|
|
sharedBuf[localIndex] = f32(${this.useSharedMemoryWithB?"B":"A"}[localIndex]);
|
|
}
|
|
workgroupBarrier();
|
|
|
|
for(var i = 0; i < ${this.workPerThread}; i = i + 1) {
|
|
let flatIndex = index * ${this.workPerThread} + i;
|
|
if(flatIndex < uniforms.size) {
|
|
let coords = getCoordsFromIndex(flatIndex);
|
|
|
|
${t}
|
|
setOutputAtIndex(flatIndex, binaryOperation(a, b));
|
|
}
|
|
}
|
|
}
|
|
`}},Mpe=class{constructor(e,t,r){this.variableNames=["A","B"],this.workPerThread=4,this.isVec4=!0,this.size=!0;let n=128;this.workGroupSize=[n,1,1],this.outputShape=C.assertAndGetBroadcastShape(t,r),this.dispatchLayout=Ye(this.outputShape),this.dispatch=Oe(this.dispatchLayout,this.outputShape,this.workGroupSize,[this.workPerThread,1,1]),this.op=e,this.shaderKey=`binaryVec4_${e}`}getUserCode(){return`
|
|
fn binaryOperation(a : vec4<f32>, b : vec4<f32>) -> vec4<f32> {
|
|
${Gh(this.op,this.isVec4)}
|
|
}
|
|
${rt()}
|
|
if (index < uniforms.size) {
|
|
let a = getAByOutputIndex(index);
|
|
let b = getBByOutputIndex(index);
|
|
setOutputAtIndex(index, binaryOperation(a, b));
|
|
}
|
|
}
|
|
`}},j9=class{constructor(e,t,r){this.variableNames=["A","B"],this.size=!0;let n=128;this.workGroupSize=[n,1,1],this.outputShape=C.assertAndGetBroadcastShape(t,r),this.dispatchLayout=Ye(this.outputShape),this.dispatch=Oe(this.dispatchLayout,this.outputShape,this.workGroupSize),this.shaderKey=`binary_${e}`,this.op=e}getUserCode(){return`
|
|
fn binaryOperation(a : f32, b : f32) -> f32 {
|
|
${Gh(this.op,!1)}
|
|
}
|
|
${rt()}
|
|
if (index < uniforms.size) {
|
|
let a = getAByOutputIndex(index);
|
|
let b = getBByOutputIndex(index);
|
|
setOutputAtIndex(index, binaryOperation(a, b));
|
|
}
|
|
}
|
|
`}};function dv(e,t,r){if(v.arraysEqual(t,r)&&v.sizeFromShape(t)%4===0)return new Mpe(e,t,r);let n=t.length===1&&r.length>1&&t[0]<1024,a=r.length===1&&t.length>1&&r[0]<1024;return n||a?new $pe(e,t,r,a):new j9(e,t,r)}function zn(e){let{inputs:t}=e,{x:r}=t;return e.backend.incRef(r.dataId),{dataId:r.dataId,shape:r.shape,dtype:r.dtype}}var Fpe={kernelName:mi,backendName:"webgpu",kernelFunc:zn};function $d(e){let{inputs:t,backend:r}=e,{real:n,imag:a}=t,s=r.makeTensorInfo(n.shape,"complex64"),i=r.tensorMap.get(s.dataId),o=zn({inputs:{x:n},backend:r}),l=zn({inputs:{x:a},backend:r});return i.complexTensorInfos={real:o,imag:l},s}var Ppe={kernelName:Jp,backendName:"webgpu",kernelFunc:$d},jh=class{constructor(e,t){this.variableNames=["A"],this.size=!0;let r=128;this.workGroupSize=[r,1,1],this.outputShape=e,this.dispatchLayout=Ye(this.outputShape),this.dispatch=Oe(this.dispatchLayout,this.outputShape,this.workGroupSize),this.op=t,this.shaderKey=`unary_${t}`}getUserCode(){return`
|
|
fn unaryOperation(a : f32) -> f32 {
|
|
${vo(this.op,!1)}
|
|
}
|
|
${rt()}
|
|
if (index < uniforms.size) {
|
|
let a = getAByOutputIndex(index);
|
|
setOutputAtIndex(index, unaryOperation(a));
|
|
}
|
|
}
|
|
`}};function kr({opType:e,cpuKernelImpl:t,dtype:r}){return({inputs:n,backend:a})=>{let{x:s}=n,i=a,o=r||s.dtype;if(i.shouldExecuteOnCPU([s])&&t!=null){let u=i.tensorMap.get(s.dataId),d=t(u.values,o);return i.makeTensorInfo(s.shape,o,d)}let l=new jh(s.shape,e);return i.runWebGPUProgram(l,[s],o)}}function qr({opSnippet:e,cpuKernelImpl:t,supportsComplex:r=!1,dtype:n}){return({inputs:a,backend:s})=>{let{a:i,b:o}=a,l=s;if(r&&i.dtype==="complex64"){let h=l.tensorMap.get(i.dataId),p=l.tensorMap.get(o.dataId),c,f;if(e!==0)[c,f]=[[h.complexTensorInfos.real,p.complexTensorInfos.real],[h.complexTensorInfos.imag,p.complexTensorInfos.imag]].map(g=>{let[y,A]=g,x={dataId:y.dataId,dtype:y.dtype,shape:i.shape},b={dataId:A.dataId,dtype:A.dtype,shape:o.shape},w=dv(e,i.shape,o.shape);return l.runWebGPUProgram(w,[x,b],Nr(y.dtype,A.dtype))});else{let g=new uv(17,i.shape,o.shape),y=new uv(18,i.shape,o.shape),A=[{dataId:h.complexTensorInfos.real.dataId,dtype:h.complexTensorInfos.real.dtype,shape:i.shape},{dataId:h.complexTensorInfos.imag.dataId,dtype:h.complexTensorInfos.imag.dtype,shape:i.shape},{dataId:p.complexTensorInfos.real.dataId,dtype:p.complexTensorInfos.real.dtype,shape:o.shape},{dataId:p.complexTensorInfos.imag.dataId,dtype:p.complexTensorInfos.imag.dtype,shape:o.shape}];c=l.runWebGPUProgram(g,A,"float32"),f=l.runWebGPUProgram(y,A,"float32")}let m=$d({inputs:{real:c,imag:f},backend:l});return l.disposeData(c.dataId),l.disposeData(f.dataId),m}let u=n||Nr(i.dtype,o.dtype);if((i.dtype==="string"||o.dtype==="string"||l.shouldExecuteOnCPU([i,o]))&&t!=null){let h=l.tensorMap.get(i.dataId).values,p=l.tensorMap.get(o.dataId).values,c=i.dtype==="string"?C.fromUint8ToStringArray(h):h,f=i.dtype==="string"?C.fromUint8ToStringArray(p):p,[m,g]=t(i.shape,o.shape,c,f,u);return l.makeTensorInfo(g,u,m)}let d=dv(e,i.shape,o.shape);return l.runWebGPUProgram(d,[i,o],u)}}var{addImpl:_pe,ceilImpl:zpe,concatImpl:Ope,equalImpl:Dpe,expImpl:Lpe,expm1Impl:Bpe,floorImpl:Wpe,gatherNdImpl:Vpe,gatherV2Impl:Upe,greaterEqualImpl:Gpe,greaterImpl:jpe,lessEqualImpl:Hpe,lessImpl:qpe,logImpl:Xpe,maxImpl:Kpe,maximumImpl:Zpe,minimumImpl:Ype,multiplyImpl:Jpe,negImpl:Qpe,notEqualImpl:ehe,prodImpl:the,rangeImpl:rhe,rsqrtImpl:nhe,simpleAbsImpl:ahe,sliceImpl:she,stridedSliceImpl:ihe,stringNGramsImpl:ohe,subImpl:lhe,tileImpl:uhe,topKImpl:dhe,transposeImpl:phe,uniqueImpl:Fxe}=wm,hhe=kr({opType:0,cpuKernelImpl:ahe}),che={kernelName:Ho,backendName:"webgpu",kernelFunc:hhe},fhe=qr({opSnippet:1,cpuKernelImpl:_pe,supportsComplex:!0}),mhe={kernelName:Ja,backendName:"webgpu",kernelFunc:fhe},ghe=class{constructor(e){this.workPerThread=4,this.workGroupSize=[64,1,1],this.size=!0,this.outputShape=e[0],this.variableNames=e.map((t,r)=>`T${r}`),this.dispatchLayout=Ye(this.outputShape),this.dispatch=Oe(this.dispatchLayout,this.outputShape,this.workGroupSize,[this.workPerThread,1,1]),this.shaderKey="addN"}getUserCode(){let e=[];this.variableNames.forEach(r=>{e.push(`let v${r} = get${r}ByOutputCoords(coords);`)});let t=this.variableNames.map(r=>`v${r}`).join(" + ");return`
|
|
${rt()}
|
|
for (var i = 0; i < ${this.workPerThread}; i = i + 1) {
|
|
let flatIndex = index * ${this.workPerThread} + i;
|
|
if (flatIndex < uniforms.size) {
|
|
let coords = getCoordsFromIndex(flatIndex);
|
|
${e.join(`
|
|
`)}
|
|
setOutputAtIndex(flatIndex, ${t});
|
|
}
|
|
}
|
|
}
|
|
`}};function yhe(e){let{inputs:t,backend:r}=e,n=t;if(n.length===1)return zn({inputs:{x:n[0]},backend:r});let a=n.map(o=>o.dtype).reduce((o,l)=>Nr(o,l)),s=n.map(o=>o.shape),i=new ghe(s);return r.runWebGPUProgram(i,n,a)}var Ahe={kernelName:Zs,backendName:"webgpu",kernelFunc:yhe},H9=class{constructor(e,t,r){this.workGroupSize=[64,1,1],this.variableNames=["x"],this.uniforms="infinityValue : f32,",this.size=!0;let n=[t];C.assertAxesAreInnerMostDims("arg"+r.charAt(0).toUpperCase()+r.slice(1),n,e.length),this.op=r==="min"?"<":">";let[a]=C.computeOutAndReduceShapes(e,n);this.outputShape=a.length===0?[1]:a,this.dispatchLayout=Ye(this.outputShape),this.dispatch=Oe(this.dispatchLayout,this.outputShape,[1,1,1]),this.inputShape=e,this.shaderKey=`argMinMax${this.op}`}getUserCode(){let e=`
|
|
var<workgroup> xBestIndices : array<i32, ${this.workGroupSize[0]}>;
|
|
var<workgroup> xBestValues : array<f32, ${this.workGroupSize[0]}>;
|
|
`,t=()=>this.inputShape.length===1?"uniforms.xShape":`uniforms.xShape.${Ms(this.inputShape.length-1)}`,r=()=>{let n="";if(this.outputShape.length===1)this.inputShape.length!==1&&(n+="outputCoords,");else for(let a=0;a<this.outputShape.length;a++)n+=`outputCoords.${Ms(a)},`;return n};return`
|
|
fn DIV_CEIL(a : u32, b : u32) -> u32 {
|
|
return ((a - 1u) / b + 1u);
|
|
}
|
|
|
|
${e}
|
|
|
|
${rt()}
|
|
let outputIndex = index / i32(workGroupSizeX);
|
|
let reduceLength = ${t()};
|
|
|
|
var bestIndex = i32(localId.x);
|
|
var bestValue = uniforms.infinityValue;
|
|
let outputCoords = getCoordsFromIndex(outputIndex);
|
|
for (var k = i32(localId.x); k < reduceLength && outputIndex < uniforms.size;
|
|
k = k + i32(workGroupSizeX)) {
|
|
let candidate = getX(${r()} k);
|
|
if (!isnan(candidate) && candidate ${this.op} bestValue) {
|
|
bestValue = candidate;
|
|
bestIndex = k;
|
|
}
|
|
}
|
|
xBestValues[localId.x] = bestValue;
|
|
xBestIndices[localId.x] = bestIndex;
|
|
workgroupBarrier();
|
|
|
|
var reduceSize = min(u32(reduceLength), workGroupSizeX);
|
|
for (var currentSize = reduceSize / 2u; reduceSize > 1u;
|
|
currentSize = reduceSize / 2u) {
|
|
let interval = DIV_CEIL(reduceSize, 2u);
|
|
if (localId.x < currentSize) {
|
|
let candidate = xBestValues[localId.x + interval];
|
|
if (candidate ${this.op} bestValue) {
|
|
bestValue = candidate;
|
|
xBestValues[localId.x] = bestValue;
|
|
xBestIndices[localId.x] = xBestIndices[localId.x + interval];
|
|
}
|
|
}
|
|
reduceSize = interval;
|
|
workgroupBarrier();
|
|
}
|
|
|
|
if (localId.x == 0u && outputIndex < uniforms.size) {
|
|
setOutputAtIndexI32(outputIndex, xBestIndices[localId.x]);
|
|
}
|
|
}
|
|
`}},xhe=class{constructor(e,t){this.variableNames=["A"],this.workGroupSize=[16,16,1];let r=new Array(e.length);for(let n=0;n<r.length;n++)r[n]=e[t[n]];this.outputShape=r,this.dispatchLayout={x:[0],y:[1]},this.dispatch=Oe(this.dispatchLayout,this.outputShape,this.workGroupSize,[1,1,1]),this.shaderKey="transposeShared"}getUserCode(){return`
|
|
let TILE_DIM = ${this.workGroupSize[0]};
|
|
var<workgroup> tile : array<array<f32, ${this.workGroupSize[0]+1}>, ${this.workGroupSize[0]}>;
|
|
${S5()}
|
|
fn main(@builtin(local_invocation_id) localId : vec3<u32>,
|
|
@builtin(workgroup_id) workgroupId : vec3<u32>) {
|
|
var x = i32(workgroupId.x) * TILE_DIM + i32(localId.x);
|
|
var y = i32(workgroupId.y) * TILE_DIM + i32(localId.y);
|
|
let width = uniforms.outShape[0];
|
|
let height = uniforms.outShape[1];
|
|
if (x < width && y < height) {
|
|
tile[localId.y][localId.x] = A[y * width + x];
|
|
}
|
|
workgroupBarrier();
|
|
|
|
x = i32(workgroupId.y) * TILE_DIM + i32(localId.x);
|
|
y = i32(workgroupId.x) * TILE_DIM + i32(localId.y);
|
|
if (x < height && y < width) {
|
|
setOutputAtIndex((y * height + x), tile[localId.x]
|
|
[localId.y]);
|
|
}
|
|
}
|
|
`}},bhe=class{constructor(e,t){this.variableNames=["A"],this.workPerThread=4,this.workGroupSize=[64,1,1],this.size=!0;let r=new Array(e.length);for(let n=0;n<r.length;n++)r[n]=e[t[n]];this.outputShape=r,this.dispatchLayout=Ye(this.outputShape),this.dispatch=Oe(this.dispatchLayout,this.outputShape,this.workGroupSize,[this.workPerThread,1,1]),this.newDim=t,this.shaderKey=`transpose_${t}`}getUserCode(){let e=xr(this.outputShape.length),t=vhe(this.newDim);return`
|
|
${rt()}
|
|
|
|
for(var i = 0; i < ${this.workPerThread}; i = i + 1) {
|
|
let flatIndex = index * ${this.workPerThread} + i;
|
|
if(flatIndex < uniforms.size) {
|
|
let resRC = getCoordsFromIndex(flatIndex);
|
|
setOutputAtIndex(flatIndex, A[getIndexFromCoords${this.outputShape.length}D(
|
|
${e}(${t}), uniforms.aShape)]);
|
|
}
|
|
}
|
|
}
|
|
`}};function vhe(e){let t=e.length;if(t>6)throw Error(`Transpose for rank ${t} is not yet supported`);let r=new Array(t);for(let n=0;n<e.length;n++)r[e[n]]=`resRC.${Ms(n)}`;return r.join()}function Ya(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{perm:s}=n,i=r,o=a.shape.length,l=new Array(o);for(let d=0;d<l.length;d++)l[d]=a.shape[s[d]];if(r.shouldExecuteOnCPU([a])){let d=i.tensorMap.get(a.dataId).values,h=phe(d,a.shape,a.dtype,s,l);return r.makeTensorInfo(l,a.dtype,h)}if(a.shape.length===2&&v.arraysEqual(s,[1,0])){let d=new xhe(a.shape,s);return i.runWebGPUProgram(d,[a],a.dtype)}let u=new bhe(a.shape,s);return i.runWebGPUProgram(u,[a],a.dtype)}var whe={kernelName:Vi,backendName:"webgpu",kernelFunc:Ya};function khe(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{axis:s}=n,i=v.parseAxisParam(s,a.shape),o=C.getAxesPermutation(i,a.shape.length),l=a,u=[];o!=null&&(l=Ya({inputs:{x:a},backend:r,attrs:{perm:o}}),u.push(l),i=C.getInnerMostAxes(i.length,l.shape.length)),C.assertAxesAreInnerMostDims("argMax",[i[0]],l.shape.length);let d=new H9(l.shape,i[0],"max"),h=[{type:"float32",data:[Number.NEGATIVE_INFINITY]}],p=r.runWebGPUProgram(d,[l],"int32",h);return u.forEach(c=>r.disposeData(c.dataId)),p}var Ihe={kernelName:Ys,backendName:"webgpu",kernelFunc:khe};function She(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{axis:s}=n,i=v.parseAxisParam(s,a.shape),o=C.getAxesPermutation(i,a.shape.length),l=a,u=[];o!=null&&(l=Ya({inputs:{x:a},backend:r,attrs:{perm:o}}),u.push(l),i=C.getInnerMostAxes(i.length,l.shape.length)),C.assertAxesAreInnerMostDims("argMin",[i[0]],l.shape.length);let d=new H9(l.shape,i[0],"min"),h=[{type:"float32",data:[Number.POSITIVE_INFINITY]}],p=r.runWebGPUProgram(d,[l],"int32",h);return u.forEach(c=>r.disposeData(c.dataId)),p}var Che={kernelName:Vu,backendName:"webgpu",kernelFunc:She},q9=class{constructor(e,t){this.variableNames=["x"],this.uniforms="stride : vec2<i32>, pad : vec2<i32>, dilation : vec2<i32>, convDims : vec2<i32>, filterDims : vec2<i32>,",this.workGroupSize=[128,1,1],this.size=!0,this.outputShape=e.outShape,this.dispatchLayout=Ye(this.outputShape),this.dispatch=Oe(this.dispatchLayout,this.outputShape,this.workGroupSize),this.shaderKey=`pool2D_${t}`,this.poolType=t}getUserCode(){let e="resultValue = max(value, resultValue);";this.poolType==="avg"&&(e="resultValue = resultValue + value; count = count + 1.0;");let t="resultValue";return this.poolType==="avg"&&(t="resultValue / count"),`
|
|
${rt()}
|
|
if (index < uniforms.size) {
|
|
let coords = getCoordsFromIndex(index);
|
|
let batch = coords[0];
|
|
let xRCCorner = vec2<i32>(coords.yz) * uniforms.stride - uniforms.pad;
|
|
let xRCorner = xRCCorner.x;
|
|
let xCCorner = xRCCorner.y;
|
|
|
|
var resultValue = ${this.poolType==="avg"?"0.0":"-1.0 / pow(10.0, -20.0)"};
|
|
var count = 0.0;
|
|
|
|
for (var wR = 0; wR < uniforms.filterDims.x; wR = wR + uniforms.dilation.x) {
|
|
let xR = xRCorner + wR;
|
|
|
|
if (xR < 0 || xR >= uniforms.convDims.x) {
|
|
continue;
|
|
}
|
|
|
|
for (var wC = 0; wC < uniforms.filterDims.y; wC = wC + uniforms.dilation.y) {
|
|
let xC = xCCorner + wC;
|
|
if (xC < 0 || xC >= uniforms.convDims.y) {
|
|
continue;
|
|
}
|
|
|
|
let value = getX(batch, xR, xC, coords[3]);
|
|
${e}
|
|
}
|
|
}
|
|
|
|
setOutputAtIndex(index, ${t});
|
|
}
|
|
}
|
|
`}},X9=class{constructor(e){this.variableNames=["x"],this.uniforms="stride : vec2<i32>,",this.workGroupSize=[256,1,1],this.size=!0,this.outputShape=e.outShape,this.dispatchLayout=Ye(this.outputShape),this.dispatch=Oe(this.dispatchLayout,this.outputShape,this.workGroupSize),this.shaderKey="poolWithFilterSizeEqualsOne"}getUserCode(){return`
|
|
${rt()}
|
|
if (index < uniforms.size) {
|
|
let coords = getCoordsFromIndex(index);
|
|
let batch = coords[0];
|
|
let d = coords[3];
|
|
|
|
let xRCCorner = coords.yz * uniforms.stride;
|
|
let xRCorner = xRCCorner.x;
|
|
let xCCorner = xRCCorner.y;
|
|
|
|
let value = getX(batch, xRCorner, xCCorner, d);
|
|
setOutputAtIndex(index, value);
|
|
}
|
|
}
|
|
`}};function The(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{filterSize:s,strides:i,pad:o,dimRoundingMode:l}=n,u=1,d=C.computePool2DInfo(a.shape,s,i,u,o,l);if(d.filterWidth===1&&d.filterHeight===1&&v.arraysEqual(d.inShape,d.outShape))return zn({inputs:{x:a},backend:r});let h,p=[{type:"int32",data:[d.strideHeight,d.strideWidth]}];return d.filterHeight===1&&d.filterWidth===1?h=new X9(d):(h=new q9(d,"avg"),p.push({type:"int32",data:[d.padInfo.top,d.padInfo.left]},{type:"int32",data:[d.dilationHeight,d.dilationWidth]},{type:"int32",data:[d.inHeight,d.inWidth]},{type:"int32",data:[d.effectiveFilterHeight,d.effectiveFilterWidth]})),r.runWebGPUProgram(h,[a],a.dtype,p)}var Nhe={kernelName:Js,backendName:"webgpu",kernelFunc:The};function Ehe(e){let{inputs:t,backend:r,attrs:n}=e,{a,b:s}=t,{transposeA:i,transposeB:o}=n;return R5({a,b:s,transposeA:i,transposeB:o,backend:r})}var Rhe={kernelName:Qs,backendName:"webgpu",kernelFunc:Ehe},$he=class{constructor(e,t){this.variableNames=["source"],this.workPerThread=1,this.workGroupSize=[64,1,1],this.size=!0,this.outputShape=t,this.rank=t.length,this.dispatchLayout=Ye(this.outputShape),this.dispatch=Oe(this.dispatchLayout,this.outputShape,this.workGroupSize,[this.workPerThread,1,1]),this.start=e,this.uniforms=`start : ${xr(e.length)}, `,this.shaderKey="slice"}getUserCode(){let e=xr(this.rank),t=Mhe(this.rank),r;return this.start.length===1?r=this.outputShape.map((n,a)=>"sourceLoc = uniforms.start + coords;"):r=this.outputShape.map((n,a)=>`sourceLoc.${oy[a]} = uniforms.start[${a}] + coords.${oy[a]};`),`
|
|
${rt()}
|
|
if (index < uniforms.size) {
|
|
var sourceLoc : ${e};
|
|
let coords = getCoordsFromIndex(index);
|
|
${r.join(`
|
|
`)}
|
|
setOutputAtIndex(index, getSource(${t}));
|
|
}
|
|
}
|
|
`}},oy=["x","y","z","w","u","v"];function Mhe(e){if(e===1)return"sourceLoc";if(e<=6)return oy.slice(0,e).map(t=>`sourceLoc.${t}`).join(",");throw Error(`Slicing for rank ${e} is not yet supported`)}function Md(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{begin:s,size:i}=n,[o,l]=Ot.parseSliceParams(a,s,i);if(Ot.assertParamsValid(a,o,l),r.shouldExecuteOnCPU([a])||a.dtype==="string"){let h=r.tensorMap.get(a.dataId),p=she(h.values,o,l,a.shape,a.dtype);return r.makeTensorInfo(l,a.dtype,p)}if(v.sizeFromShape(l)===0)return r.makeTensorInfo(l,a.dtype,[]);let u=new $he(o,l),d=[{type:"int32",data:o}];return r.runWebGPUProgram(u,[a],a.dtype,d)}var Fhe={kernelName:bl,backendName:"webgpu",kernelFunc:Md},Phe=e=>{let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{blockShape:s,crops:i}=n;v.assert(a.shape.length<=4,()=>"batchToSpaceND for rank > 4 with a WebGPU backend not implemented yet");let o=s.reduce((A,x)=>A*x),l=C.getReshaped(a.shape,s,o),u=C.getPermuted(l.length,s.length),d=C.getReshapedPermuted(a.shape,s,o),h=C.getSliceBeginCoords(i,s.length),p=C.getSliceSize(d,i,s.length),c=[],f=Qe({inputs:{x:a},backend:r,attrs:{shape:l}}),m=Ya({inputs:{x:f},backend:r,attrs:{perm:u}}),g=Qe({inputs:{x:m},backend:r,attrs:{shape:d}}),y=Md({inputs:{x:g},backend:r,attrs:{begin:h,size:p}});return c.push(f),c.push(m),c.push(g),c.forEach(A=>r.disposeData(A.dataId)),y},_he={kernelName:qo,backendName:"webgpu",kernelFunc:Phe},K9=qr({opSnippet:10,dtype:"bool",cpuKernelImpl:ehe}),zhe={kernelName:ul,backendName:"webgpu",kernelFunc:K9};function Hh(e){let{inputs:t,backend:r}=e,{input:n}=t,a=r.tensorMap.get(n.dataId);return zn({inputs:{x:a.complexTensorInfos.real},backend:r})}var Ohe={kernelName:oh,backendName:"webgpu",kernelFunc:Hh};function Dhe(e,t){let r=new jh(e.shape,22),n=t.runWebGPUProgram(r,[e],"int32");return{dataId:n.dataId,shape:n.shape,dtype:n.dtype}}function ly(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{dtype:s}=n;if(s==="complex64"){if(a.dtype==="complex64")return zn({inputs:{x:a},backend:r});let i=_t(a.shape),o=ly({inputs:{x:a},backend:r,attrs:{dtype:"float32"}}),l=$d({inputs:{real:o,imag:i},backend:r});return i.dispose(),r.disposeData(o.dataId),l}if(a.dtype==="complex64"){let i=Hh({inputs:{input:a},backend:r}),o=ly({inputs:{x:i},backend:r,attrs:{dtype:s}});return r.disposeData(i.dataId),o}if(!v.hasEncodingLoss(a.dtype,s)){let i=zn({inputs:{x:a},backend:r});return{dataId:i.dataId,shape:i.shape,dtype:s}}if(s==="int32")return Dhe(a,r);if(s==="bool"){let i=r.makeTensorInfo([],"bool",v.getTypedArrayFromDType("bool",1)),o=K9({inputs:{a,b:i},backend:r});return r.disposeData(i.dataId),o}throw new Error(`Error in Cast: failed to cast ${a.dtype} to ${s}`)}var Lhe={kernelName:ei,backendName:"webgpu",kernelFunc:ly},Bhe=kr({opType:1,cpuKernelImpl:zpe}),Whe={kernelName:ti,backendName:"webgpu",kernelFunc:Bhe},Vhe=class{constructor(e){this.variableNames=["A"],this.uniforms="minVal : f32, maxVal : f32,",this.workPerThread=4,this.workGroupSize=[64,1,1],this.isVec4=!0,this.size=!0,this.outputShape=e,this.dispatchLayout=Ye(this.outputShape),this.dispatch=Oe(this.dispatchLayout,this.outputShape,this.workGroupSize,[this.workPerThread,1,1]),this.shaderKey="clipVec4"}getUserCode(){return`
|
|
${rt()}
|
|
if(index < uniforms.size) {
|
|
let value = getAByOutputIndex(index);
|
|
var clampedValue : vec4<f32>;
|
|
for (var i = 0; i < 4; i = i + 1) {
|
|
if (isnan(value[i])) {
|
|
clampedValue[i] = value[i];
|
|
} else {
|
|
clampedValue[i] = clamp(value[i], uniforms.minVal, uniforms.maxVal);
|
|
}
|
|
}
|
|
|
|
setOutputAtIndex(index, clampedValue);
|
|
}
|
|
}
|
|
`}},Uhe=class{constructor(e){this.variableNames=["A"],this.uniforms="minVal : f32, maxVal : f32,",this.workGroupSize=[64,1,1],this.size=!0,this.outputShape=e,this.dispatchLayout=Ye(this.outputShape),this.dispatch=Oe(this.dispatchLayout,this.outputShape,this.workGroupSize),this.shaderKey="clip"}getUserCode(){return`
|
|
${rt()}
|
|
if(index < uniforms.size) {
|
|
let value = getAByOutputIndex(index);
|
|
if (isnan(value)) {
|
|
setOutputAtIndex(index, value);
|
|
return;
|
|
}
|
|
setOutputAtIndex(index, clamp(value, uniforms.minVal, uniforms.maxVal));
|
|
}
|
|
}
|
|
`}};function Ghe(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{clipValueMin:s,clipValueMax:i}=n,o,l=[{type:"float32",data:[s]},{type:"float32",data:[i]}];return v.sizeFromShape(a.shape)%4===0?o=new Vhe(a.shape):o=new Uhe(a.shape),r.runWebGPUProgram(o,[a],a.dtype,l)}var jhe={kernelName:Qa,backendName:"webgpu",kernelFunc:Ghe},Hhe=class{constructor(e){this.uniforms="",this.workPerThread=4,this.workGroupSize=[64,1,1],this.size=!0,this.outputShape=C.computeOutShape(e,1),this.variableNames=e.map((t,r)=>`T${r}`),this.dispatchLayout=Ye(this.outputShape),this.dispatch=Oe(this.dispatchLayout,this.outputShape,this.workGroupSize,[this.workPerThread,1,1]),this.offsetLength=e.length-1;for(let t=0;t<this.offsetLength;t++)this.uniforms+=`offset${t} : i32,`;this.shaderKey="concat"}getUserCode(){let e=[];if(this.offsetLength>0){e.push("if (yC < uniforms.offset0){ setOutputAtCoords(coords.x, coords.y, getT0(yR, yC)); }");for(let n=1;n<this.offsetLength;n++)e.push(`else if (yC < uniforms.offset${[n]}){ setOutputAtCoords(coords.x, coords.y, getT${n}(yR, yC - uniforms.offset${n-1})); }`);let t=this.offsetLength,r=this.offsetLength-1;e.push(`else { setOutputAtCoords(coords.x, coords.y, getT${t}(yR, yC - uniforms.offset${r})); }`)}else e.push("setOutputAtCoords(coords.x, coords.y, getT0(yR, yC));");return`
|
|
${rt()}
|
|
for(var i = 0; i < ${this.workPerThread}; i = i + 1) {
|
|
let flatIndex = index * ${this.workPerThread} + i;
|
|
if(flatIndex < uniforms.size) {
|
|
let coords = getCoordsFromIndex(flatIndex);
|
|
let yR = coords.x;
|
|
let yC = coords.y;
|
|
|
|
${e.join(`
|
|
`)}
|
|
}
|
|
}
|
|
}
|
|
`}};function $m(e){let{inputs:t,backend:r}=e,{input:n}=t,a=r.tensorMap.get(n.dataId);return zn({inputs:{x:a.complexTensorInfos.imag},backend:r})}var qhe={kernelName:nh,backendName:"webgpu",kernelFunc:$m};function uy(e,t,r){let n=e[0].dtype;if(n==="complex64"){let c=e.map(A=>Hh({inputs:{input:A},backend:r})),f=e.map(A=>$m({inputs:{input:A},backend:r})),m=uy(c,t,r),g=uy(f,t,r),y=$d({inputs:{real:m,imag:g},backend:r});return c.forEach(A=>r.disposeData(A.dataId)),f.forEach(A=>r.disposeData(A.dataId)),r.disposeData(m.dataId),r.disposeData(g.dataId),y}let a=r.shouldExecuteOnCPU(e);if(n==="string"&&(a=!0),a){let c=e.map(b=>{let w=v.sizeFromShape(b.shape.slice(t));return Qe({inputs:{x:b},backend:r,attrs:{shape:[-1,w]}})}),f=c.map(b=>({vals:r.readSync(b.dataId),shape:b.shape})),m=C.computeOutShape(c.map(b=>b.shape),1),g=c[0].shape[0]===1,y=Ope(f,m,n,g),A=C.computeOutShape(e.map(b=>b.shape),t),x=r.makeTensorInfo(A,n,y);return c.forEach(b=>r.disposeData(b.dataId)),x}let{tensors2D:s,outShape:i}=Xhe(e,t,r),o=s.map(c=>c.shape),l=new Hhe(o),u=[],d=new Array(o.length-1);if(d.length>0){d[0]=o[0][1],u.push({type:"int32",data:[d[0]]});for(let c=1;c<d.length;c++)d[c]=d[c-1]+o[c][1],u.push({type:"int32",data:[d[c]]})}let h=r.runWebGPUProgram(l,s,s[0].dtype,u);s.forEach(c=>r.disposeData(c.dataId));let p=Qe({inputs:{x:h},backend:r,attrs:{shape:i}});return r.disposeData(h.dataId),p}function Xhe(e,t,r){let n=C.computeOutShape(e.map(a=>a.shape),t);return{tensors2D:e.map(a=>Qe({inputs:{x:a},backend:r,attrs:{shape:[v.sizeFromShape(a.shape.slice(0,t)),v.sizeFromShape(a.shape.slice(t))]}})),outShape:n}}function Z9(e){let{inputs:t,backend:r,attrs:n}=e,{axis:a}=n,s=v.parseAxisParam(a,t[0].shape)[0],i=C.computeOutShape(t.map(u=>u.shape),s);if(v.sizeFromShape(i)===0)return r.makeTensorInfo(i,t[0].dtype,[]);let o=t.filter(u=>v.sizeFromShape(u.shape)>0);if(o.length===1)return zn({inputs:{x:o[0]},backend:r});let l=o.map(u=>u.shape);return C.assertParamsConsistent(l,s),uy(o,s,r)}var Khe={kernelName:Xo,backendName:"webgpu",kernelFunc:Z9},Zhe=class{constructor(e,t=!1,r=null,n=!1){this.variableNames=["x","W"],this.uniforms=`filterDims : vec2<i32>, pad : vec2<i32>, stride : vec2<i32>, dilation : vec2<i32>,
|
|
dimAOuter : i32, dimBOuter : i32, dimInner : i32,`,this.workGroupSize=[8,8,1],this.isVec4=!0,this.outputShape=e.outShape,v.assert(e.dataFormat==="channelsLast",()=>"TODO: NCHW is unimplemented"),this.dispatchLayout={x:[3],y:[1,2],z:[0]},this.outputShape[1]===1?this.elementsPerThread=[4,1,1]:this.elementsPerThread=[4,4,1],this.dispatch=Oe(this.dispatchLayout,this.outputShape,this.workGroupSize,this.elementsPerThread),this.convInfo=e,this.addBias=t,this.activation=r,this.hasPreluActivationWeights=n,this.addBias&&this.variableNames.push("bias"),this.hasPreluActivationWeights&&this.variableNames.push("preluActivationWeights"),this.tileAOuter=this.outputShape[1]===1?1:this.workGroupSize[1]*this.elementsPerThread[1],this.tileBOuter=this.workGroupSize[0]*this.elementsPerThread[0],this.tileInner=this.tileBOuter,[this.fitA,this.fitB]=this.getShapeFit(),this.remainder=this.convInfo.inChannels%4===0,this.shaderKey=`conv2DMMVec4_${this.activation}_${this.fitA}_${this.fitB}_${this.elementsPerThread}_${this.remainder}`}getShapeFit(){let e=[this.tileAOuter,this.tileInner],t=[this.tileInner,this.tileBOuter],r=this.outputShape[1]*this.outputShape[2],n=this.outputShape[3],a=this.convInfo.filterHeight*this.convInfo.filterWidth*this.convInfo.inChannels;return[Za(e,[r,a]),Za(t,[a,n])]}getSampleAWithRemainder(e){return`let flatIndex${e} = getIndexFromCoords4D(coord, uniforms.xShape);
|
|
let divBy4Remainder${e} = flatIndex${e} % 4;
|
|
let divBy4Index${e} = flatIndex${e} / 4;
|
|
let curData${e} = x[divBy4Index${e}];
|
|
if (divBy4Remainder${e} == 0) {
|
|
temp = curData${e};
|
|
} else {
|
|
// TODO: This could end up being a redundant load with another one in
|
|
// the same shader invocation. Perhaps there's an opportunity for
|
|
// optimization
|
|
let nextData${e} = x[divBy4Index${e} + 1];
|
|
if (divBy4Remainder${e} == 1) {
|
|
temp = vec4<f32>(curData${e}.yzw, nextData${e}.x);
|
|
} else if (divBy4Remainder${e} == 2) {
|
|
temp = vec4<f32>(curData${e}.zw, nextData${e}.xy);
|
|
} else if (divBy4Remainder${e} == 3) {
|
|
temp = vec4<f32>(curData${e}.w, nextData${e}.xyz);
|
|
}
|
|
}
|
|
`}getUserCode(){let e=G9(this.elementsPerThread,this.tileAOuter,this.tileBOuter,this.tileInner),t=`let outRow = r / uniforms.outShape[2];
|
|
let outCol = r % uniforms.outShape[2];
|
|
let WRow = c / (uniforms.filterDims[1] * uniforms.xShape[3]);
|
|
let WCol = c / uniforms.xShape[3] % uniforms.filterDims[1];
|
|
let inChCoord = c % uniforms.xShape[3];
|
|
var coord = vec4<i32>(
|
|
batch,
|
|
outRow * uniforms.stride[0] + uniforms.dilation[0] * WRow - uniforms.pad[0],
|
|
outCol * uniforms.stride[1] + uniforms.dilation[1] * WCol - uniforms.pad[1],
|
|
inChCoord);
|
|
var resData = vec4<f32>(0.0);
|
|
${this.remainder?`// The bounds checking is always needed since we use it to pad zero for
|
|
// the 'same' padding type.
|
|
if (coordsInBounds4D(coord, uniforms.xShape)) {
|
|
resData = x[getIndexFromCoords4D(coord, uniforms.xShape) / 4];
|
|
} else {
|
|
resData = vec4<f32>(0.0); }`:`var temp = vec4<f32>(0.0);
|
|
${this.getSampleAWithRemainder(1)}
|
|
resData = temp;
|
|
if (WCol == (uniforms.filterDims[1] - 1)) {
|
|
coord = vec4<i32>(
|
|
coord.x, coord.y + 1, coord.z + 1 - uniforms.filterDims[1], 0);
|
|
${this.getSampleAWithRemainder(2)}
|
|
if (inChCoord == 0) {
|
|
resData = vec4<f32>(resData.xyz, temp.x);
|
|
} else if (inChCoord == 1) {
|
|
resData = vec4<f32>(resData.xy, temp.xy);
|
|
} else {
|
|
resData = vec4<f32>(resData.x, temp.xyz);
|
|
}
|
|
}
|
|
`}
|
|
return resData;`,r=this.fitA?`${t}`:`if (r < uniforms.dimAOuter && c < uniforms.dimInner) {
|
|
${t}
|
|
}
|
|
return vec4<f32>(0.0);
|
|
`,n=this.fitB?"return W[row * uniforms.dimBOuter / 4 + col];":`if(coordsInBounds2D(vec2<i32>(row, col * 4), vec2<i32>(uniforms.dimInner, uniforms.dimBOuter))) {
|
|
return W[row * uniforms.dimBOuter / 4 + col];
|
|
}
|
|
return vec4<f32>(0.0);
|
|
`,a="",s="";if(this.activation){let o=Zi(this.activation,this.isVec4);this.hasPreluActivationWeights?a=`fn activation(a : vec4<f32>, outCoord : vec4<i32>) -> vec4<f32> {
|
|
let b = getPreluActivationWeightsByOutputCoords(outCoord);
|
|
${o}
|
|
}`:a=`
|
|
fn activation(a : vec4<f32>, outCoord : vec4<i32>) -> vec4<f32> {
|
|
${o}
|
|
}`,s="value = activation(value, outCoord);"}let i=this.addBias?"value = value + getBiasByOutputCoords(outCoord);":"";return`
|
|
${a}
|
|
fn mm_readA(row : i32, col : i32, globalId : vec3<u32>) -> vec4<f32> {
|
|
let r = row;
|
|
let c = col * 4;
|
|
var batch = i32(globalId.z);
|
|
${r}
|
|
}
|
|
|
|
fn mm_readB(row : i32, col : i32, globalId : vec3<u32>) -> vec4<f32> {
|
|
${n}
|
|
}
|
|
|
|
fn mm_write(row : i32, col : i32, valueInput : vec4<f32>, globalId : vec3<u32>) {
|
|
var batch = i32(globalId.z);
|
|
var value = valueInput;
|
|
if (row < uniforms.dimAOuter && col * 4 < uniforms.dimBOuter)
|
|
{
|
|
let outCoord = vec4<i32>(
|
|
batch,
|
|
row / uniforms.outShape[2],
|
|
row % uniforms.outShape[2],
|
|
col * 4);
|
|
${i}
|
|
${s}
|
|
setOutputAtCoords(outCoord[0], outCoord[1], outCoord[2], outCoord[3],
|
|
value);
|
|
}
|
|
}
|
|
${e}
|
|
`}},Yhe=class{constructor(e,t=!1,r=null,n=!1){this.variableNames=["x","W"],this.uniforms="filterDims : vec2<i32>, pad : vec2<i32>, stride : vec2<i32>, dilation : vec2<i32>, dimAOuter : i32, dimBOuter : i32, dimInner : i32,",this.outputShape=e.outShape,this.isChannelsLast=e.dataFormat==="channelsLast",this.dispatchLayout=this.isChannelsLast?{x:[3],y:[1,2],z:[0]}:{x:[1],y:[2,3],z:[0]},this.workGroupSize=C5(this.dispatchLayout,this.outputShape),this.elementsPerThread=T5(this.dispatchLayout,this.outputShape),this.dispatch=Oe(this.dispatchLayout,this.outputShape,this.workGroupSize,this.elementsPerThread),t&&this.variableNames.push("bias"),n&&this.variableNames.push("preluActivationWeights"),this.convInfo=e,this.addBias=t,this.activation=r,this.hasPreluActivationWeights=n,[this.fitA,this.fitB]=this.getShapeFit(),this.shaderKey=`conv2DMM_${this.elementsPerThread}_${this.activation}_${this.fitA}_${this.fitB}_${this.isChannelsLast}`}getShapeFit(){let e=this.workGroupSize[1]*this.elementsPerThread[1],t=this.workGroupSize[0]*this.elementsPerThread[0],r=e>t?e:t;v.assert(r%this.workGroupSize[0]===0&&r%this.workGroupSize[1]===0,()=>"tileInner must be multiple of workgroupsize.x and workgroupsize.y");let n=[e,r],a=[r,t],s=this.convInfo.outHeight*this.convInfo.outWidth,i=this.convInfo.outChannels,o=this.convInfo.filterHeight*this.convInfo.filterWidth*this.convInfo.inChannels;return[Za(n,[s,o]),Za(a,[o,i])]}getUserCode(){let e=this.isChannelsLast?`
|
|
let coord = vec4<i32>(batch, xRow, xCol, col % inChannels);
|
|
`:`
|
|
let coord = vec4<i32>(batch, col % inChannels, xRow, xCol);
|
|
`,t=this.isChannelsLast?`
|
|
let outCoord = vec4<i32>(
|
|
batch,
|
|
row / outWidth,
|
|
row % outWidth,
|
|
col);
|
|
`:`
|
|
let outCoord = vec4<i32>(
|
|
batch,
|
|
col,
|
|
row / outWidth,
|
|
row % outWidth);
|
|
`,r=E5(this.elementsPerThread,this.workGroupSize),n=`
|
|
let inChannels = uniforms.wShape[2];
|
|
let outWidth = ${this.isChannelsLast?"uniforms.outShape[2]":"uniforms.outShape[3]"};
|
|
let outRow = row / outWidth;
|
|
let outCol = row % outWidth;
|
|
|
|
let WRow = col / (uniforms.filterDims[1] * inChannels);
|
|
let WCol = col / inChannels % uniforms.filterDims[1];
|
|
let xRow = outRow * uniforms.stride[0] + uniforms.dilation[0] * WRow - uniforms.pad[0];
|
|
let xCol = outCol * uniforms.stride[1] + uniforms.dilation[1] * WCol - uniforms.pad[1];
|
|
${e}
|
|
// The bounds checking is always needed since we use it to pad zero for the
|
|
// 'same' padding type.
|
|
if(coordsInBounds4D(coord, uniforms.xShape)) {
|
|
return x[getIndexFromCoords4D(coord, uniforms.xShape)];
|
|
}
|
|
return 0.0;`,a=this.fitA?`${n}`:`if (row < uniforms.dimAOuter && col < uniforms.dimInner) {
|
|
${n}
|
|
}
|
|
return 0.0;
|
|
`,s=this.fitB?"return W[row * uniforms.dimBOuter + col];":`if(coordsInBounds2D(vec2<i32>(row, col), vec2<i32>(uniforms.dimInner, uniforms.dimBOuter))) {
|
|
return W[row * uniforms.dimBOuter + col];
|
|
}
|
|
return 0.0;
|
|
`,i="",o="";if(this.activation){let u=Zi(this.activation,!1);this.hasPreluActivationWeights?i=`fn activation(a: f32, outCoord : vec4<i32>) -> f32 {
|
|
let b = getPreluActivationWeightsByOutputCoords(outCoord);
|
|
${u}
|
|
}`:i=`
|
|
fn activation(a : f32, outCoord : vec4<i32>) -> f32 {
|
|
${u}
|
|
}
|
|
`,o="value = activation(value, outCoord);"}let l=this.addBias?"value = value + getBiasByOutputCoords(outCoord);":"";return`
|
|
${i}
|
|
fn mm_readA(row : i32, col : i32, globalId : vec3<u32>) -> f32 {
|
|
var batch = i32(globalId.z);
|
|
${a}
|
|
}
|
|
|
|
fn mm_readB(row : i32, col : i32, globalId : vec3<u32>) -> f32 {
|
|
${s}
|
|
}
|
|
|
|
fn mm_write(row : i32, col : i32, valueInput : f32, globalId : vec3<u32>) {
|
|
var batch = i32(globalId.z);
|
|
var value = valueInput;
|
|
let outWidth = ${this.isChannelsLast?"uniforms.outShape[2]":"uniforms.outShape[3]"};
|
|
${t}
|
|
${l}
|
|
${o}
|
|
result[getIndexFromCoords4D(outCoord, uniforms.outShape)] = value;
|
|
}
|
|
${r}
|
|
`}};function Jhe({x:e,filter:t,convInfo:r,backend:n,bias:a=null,preluActivationWeights:s=null,leakyreluAlpha:i=0,activation:o=null}){let l=r.dataFormat==="channelsLast",u=!l,d=!1,h=l&&r.filterHeight===r.inHeight&&r.filterWidth===r.inWidth&&r.padInfo.type==="VALID",p,c;if(h){let g=r.inHeight*r.inWidth*r.inChannels;p=Qe({inputs:{x:e},backend:n,attrs:{shape:[1,r.batchSize,g]}}),c=Qe({inputs:{x:t},backend:n,attrs:{shape:[1,g,r.outChannels]}})}else p=Qe({inputs:{x:e},backend:n,attrs:{shape:l?[r.batchSize,r.inHeight*r.inWidth,r.inChannels]:[r.batchSize,r.inChannels,r.inHeight*r.inWidth]}}),c=Qe({inputs:{x:t},backend:n,attrs:{shape:[1,r.inChannels,r.outChannels]}});let f=R5({a:l?p:c,b:l?c:p,transposeA:u,transposeB:d,backend:n,bias:a,activation:o,preluActivationWeights:s,leakyreluAlpha:i}),m=Qe({inputs:{x:f},backend:n,attrs:{shape:r.outShape}});return n.disposeData(p.dataId),n.disposeData(c.dataId),n.disposeData(f.dataId),m}function Y9({x:e,filter:t,convInfo:r,backend:n,bias:a=null,preluActivationWeights:s=null,leakyreluAlpha:i=0,activation:o=null}){let l=a!=null,u=s!=null,d=r.dataFormat==="channelsLast",h;if(d&&r.filterHeight===r.inHeight&&r.filterWidth===r.inWidth&&r.padInfo.type==="VALID"||r.filterHeight===1&&r.filterWidth===1&&r.dilationHeight===1&&r.dilationWidth===1&&r.strideHeight===1&&r.strideWidth===1&&(r.padInfo.type==="SAME"||r.padInfo.type==="VALID"))return Jhe({x:e,filter:t,convInfo:r,backend:n,bias:a,activation:o,preluActivationWeights:s,leakyreluAlpha:i});let p=(r.inChannels%4===0||r.inChannels===3&&r.padInfo.type==="VALID")&&r.outChannels%4===0&&d,c=[r.padInfo.top,r.padInfo.left],f=[{type:"int32",data:[r.filterHeight,r.filterWidth]},{type:"int32",data:[...c]},{type:"int32",data:[r.strideHeight,r.strideWidth]},{type:"int32",data:[r.dilationHeight,r.dilationWidth]}];p?h=new Zhe(r,l,o,u):h=new Yhe(r,l,o,u);let m=r.outHeight*r.outWidth,g=r.outChannels,y=r.filterHeight*r.filterWidth*r.inChannels;f.push({type:"int32",data:[m]},{type:"int32",data:[g]},{type:"int32",data:[y]});let A=[e,t];return l&&A.push(a),u&&A.push(s),o==="leakyrelu"&&(f.push({type:"float32",data:[i]}),h.uniforms+=" alpha : f32,"),n.runWebGPUProgram(h,A,e.dtype,f)}function Qhe(e){let{inputs:t,attrs:r,backend:n}=e,{x:a,filter:s}=t,{strides:i,pad:o,dataFormat:l,dilations:u,dimRoundingMode:d}=r,h=C.convertConv2DDataFormat(l),p=C.computeConv2DInfo(a.shape,s.shape,i,u,o,d,!1,h);return Y9({x:a,filter:s,convInfo:p,backend:n})}var ece={kernelName:ri,backendName:"webgpu",kernelFunc:Qhe},tce=class{constructor(e){this.variableNames=["x","W"],this.uniforms="filterDims : vec2<i32>, pads : vec2<i32>, stride : vec2<i32>, outBackprop : vec4<i32>, dimAOuter : i32, dimBOuter : i32, dimInner : i32,",this.outputShape=e.inShape,v.assert(e.dataFormat==="channelsLast",()=>"TODO: NCHW is unimplemented"),this.dispatchLayout={x:[3],y:[1,2],z:[0]},this.workGroupSize=C5(this.dispatchLayout,this.outputShape),this.elementsPerThread=T5(this.dispatchLayout,this.outputShape),this.dispatch=Oe(this.dispatchLayout,this.outputShape,this.workGroupSize,this.elementsPerThread),this.shaderKey=`conv2DDerInputMM_${this.elementsPerThread}`}getUserCode(){return`
|
|
fn mm_readA(row : i32, col : i32, globalId : vec3<u32>) -> f32 {
|
|
var batch = i32(globalId.z);
|
|
if (row < uniforms.dimAOuter && col < uniforms.dimInner) {
|
|
|
|
let outRow = row / uniforms.outShape[2];
|
|
let outCol = row % uniforms.outShape[2];
|
|
|
|
let WRow = col / (uniforms.filterDims[1] * uniforms.outBackprop[3]);
|
|
let WCol = col / uniforms.outBackprop[3] % uniforms.filterDims[1];
|
|
let xR = f32(outRow - uniforms.pads[0] + WRow) / f32(uniforms.stride[0]);
|
|
let xC = f32(outCol - uniforms.pads[1] + WCol) / f32(uniforms.stride[1]);
|
|
if (xR < 0.0 || xR >= f32(uniforms.outBackprop[1]) || fract(xR) > 0.0) {
|
|
return 0.0;
|
|
}
|
|
if (xC < 0.0 || xC >= f32(uniforms.outBackprop[2]) || fract(xC) > 0.0) {
|
|
return 0.0;
|
|
}
|
|
let coord = vec4<i32>(
|
|
batch,
|
|
i32(xR),
|
|
i32(xC),
|
|
col % uniforms.outBackprop[3]);
|
|
return x[getIndexFromCoords4D(coord, uniforms.xShape)];
|
|
}
|
|
return 0.0;
|
|
}
|
|
|
|
fn mm_readB(row : i32, col : i32, globalId : vec3<u32>) -> f32 {
|
|
let coordX = uniforms.filterDims.x - 1 -
|
|
row / (uniforms.filterDims[1] * uniforms.outBackprop[3]);
|
|
let coordY = uniforms.filterDims.y - 1 -
|
|
(row / uniforms.outBackprop[3]) % uniforms.filterDims[1];
|
|
if (row < uniforms.dimInner && col < uniforms.dimBOuter &&
|
|
coordX >= 0 && coordY >= 0) {
|
|
let coord = vec4<i32>(coordX, coordY, col,
|
|
row % uniforms.outBackprop[3]);
|
|
return W[getIndexFromCoords4D(coord, uniforms.wShape)];
|
|
}
|
|
return 0.0;
|
|
}
|
|
|
|
fn mm_write(row : i32, col : i32, valueInput : f32, globalId : vec3<u32>) {
|
|
var batch = i32(globalId.z);
|
|
var value = valueInput;
|
|
let outCoord = vec4<i32>(
|
|
batch,
|
|
row / uniforms.outShape[2],
|
|
row % uniforms.outShape[2],
|
|
col);
|
|
result[getIndexFromCoords4D(outCoord, uniforms.outShape)] = value;
|
|
}
|
|
|
|
${E5(this.elementsPerThread,this.workGroupSize)}
|
|
`}},rce=class{constructor(e){this.variableNames=["dy","W"],this.uniforms="filterDims : vec2<i32>, pads : vec2<i32>, stride : vec2<i32>, outBackprop : vec4<i32>,",this.workGroupSize=[64,1,1],this.size=!0,this.outputShape=e.inShape,this.dispatchLayout=Ye(this.outputShape),this.dispatch=Oe(this.dispatchLayout,this.outputShape,this.workGroupSize),this.isChannelsLast=e.dataFormat==="channelsLast",this.shaderKey=`conv2DDerInput_${this.isChannelsLast}`}getUserCode(){let e=this.isChannelsLast?1:2,t=this.isChannelsLast?2:3,r=this.isChannelsLast?3:1;return`
|
|
${rt()} {
|
|
if(index < uniforms.size) {
|
|
let coords = getCoordsFromIndex(index);
|
|
let batch = coords[0];
|
|
let d1 = coords[${r}];
|
|
|
|
let dyCorner = vec2<i32>(coords[${e}]), coords[${t}]) - uniforms.pads;
|
|
let dyRCorner = dyCorner.x;
|
|
let dyCCorner = dyCorner.y;
|
|
|
|
// Convolve dy(?, ?, d2) with w(:, :, d1, d2) to compute dx(xR, xC, d1).
|
|
// ? = to be determined. : = across all values in that axis.
|
|
var dotProd = 0.0;
|
|
for (var wR = 0; wR < uniforms.filterDims.x; wR = wR + 1) {
|
|
let dyR = (f32(dyRCorner) + f32(wR)) / f32(uniforms.stride.x);
|
|
let wRPerm = uniforms.filterDims.x - 1 - wR;
|
|
if (dyR < 0.0 || dyR >= f32(uniforms.outBackprop[1]) || fract(dyR) > 0.0 ||
|
|
wRPerm < 0) {
|
|
continue;
|
|
}
|
|
let idyR = dyR;
|
|
|
|
for (var wC = 0; wC < uniforms.filterDims.y; wC = wC + 1) {
|
|
let dyC = (f32(dyCCorner) + f32(wC)) / f32(uniforms.stride.y);
|
|
let wCPerm = uniforms.filterDims.y - 1 - wC;
|
|
if (dyC < 0.0 || dyC >= f32(uniforms.outBackprop[2]) ||
|
|
fract(dyC) > 0.0 || wCPerm < 0) {
|
|
continue;
|
|
}
|
|
let idyC = dyC;
|
|
|
|
for (var d2 = 0; d2 < uniforms.outBackprop[3]; d2 = d2 + 1) {
|
|
if (${this.isChannelsLast}) {
|
|
let xValue = getDy(batch, idyR, idyC, d2);
|
|
let wValue = getW(wRPerm, wCPerm, d1, d2);
|
|
dotProd = dotProd + xValue * wValue;
|
|
} else {
|
|
let xValue = getDy(batch, d2, idyR, idyC);
|
|
let wValue = getW(wRPerm, wCPerm, d1, d2);
|
|
dotProd = dotProd + xValue * wValue;
|
|
}
|
|
|
|
}
|
|
}
|
|
}
|
|
setOutputAtIndex(index, dotProd);
|
|
}
|
|
}
|
|
`}};function nce(e){let{inputs:t,backend:r,attrs:n}=e,{dy:a,filter:s}=t,{inputShape:i,strides:o,pad:l,dataFormat:u,dimRoundingMode:d}=n,h=C.convertConv2DDataFormat(u),p=C.computeConv2DInfo(i,s.shape,o,1,l,d,!1,h),c=[{type:"int32",data:[p.filterHeight,p.filterWidth]},{type:"int32",data:[p.filterHeight-1-p.padInfo.top,p.filterWidth-1-p.padInfo.left]},{type:"int32",data:[p.strideHeight,p.strideWidth]},{type:"int32",data:[p.batchSize,p.outHeight,p.outWidth,p.outChannels]}],f;if(Z().getBool("WEBGPU_USE_NAIVE_CONV2D_TRANSPOSE"))f=new rce(p);else{f=new tce(p);let m=p.inShape[1]*p.inShape[2],g=p.inShape[3],y=p.filterHeight*p.filterWidth*p.outChannels;c.push({type:"uint32",data:[m]},{type:"uint32",data:[g]},{type:"uint32",data:[y]})}return r.runWebGPUProgram(f,[a,s],"float32",c)}var ace={kernelName:ni,backendName:"webgpu",kernelFunc:nce},sce=kr({opType:2}),ice={kernelName:ai,backendName:"webgpu",kernelFunc:sce},oce=kr({opType:3}),lce={kernelName:si,backendName:"webgpu",kernelFunc:oce},uce=class{constructor(e,t,r,n){this.variableNames=["Image","Boxes","BoxInd"],this.uniforms="extrapolationValue : f32,",this.workGroupSize=[64,1,1],this.size=!0;let[a]=t;this.outputShape=[a,r[0],r[1],e],this.dispatchLayout=Ye(this.outputShape),this.dispatch=Oe(this.dispatchLayout,this.outputShape,this.workGroupSize),this.methodId=n==="bilinear"?1:0,this.cropHeightBiggerThan1=this.outputShape[1]>1,this.cropWidthBiggerThan1=this.outputShape[2]>1,this.shaderKey=`cropAndResize_${this.methodId}_${this.cropHeightBiggerThan1}_${this.cropWidthBiggerThan1}`}getUserCode(){let[e,t]=["f32(uniforms.imageShape[1] - 1)","f32(uniforms.imageShape[2] - 1)"],[r,n,a]=this.cropHeightBiggerThan1?[`(${e} / f32(uniforms.outShape[1] - 1))`,"(y2-y1) * height_ratio",`y1*${e} + f32(y)*(height_scale)`]:["0.0","0.0",`0.5 * (y1+y2) * ${e}`],[s,i,o]=this.cropWidthBiggerThan1?[`(${t} / f32(uniforms.outShape[2] - 1))`,"(x2-x1) * width_ratio",`x1*${t} + f32(x)*(width_scale)`]:["0.0","0.0",`0.5 * (x1+x2) * ${t}`];return`
|
|
${rt()}
|
|
if (index < uniforms.size) {
|
|
let coords = getCoordsFromIndex(index);
|
|
let height_ratio = f32(${r});
|
|
let width_ratio = f32(${s});
|
|
let b = coords[0];
|
|
let y = coords[1];
|
|
let x = coords[2];
|
|
let d = coords[3];
|
|
// get box vals
|
|
let y1 = getBoxes(b, 0);
|
|
let x1 = getBoxes(b, 1);
|
|
let y2 = getBoxes(b, 2);
|
|
let x2 = getBoxes(b, 3);
|
|
// get image in batch index
|
|
let bInd = i32(round(getBoxInd(b)));
|
|
if(bInd < 0 || bInd >= uniforms.outShape[0]) {
|
|
return;
|
|
}
|
|
let height_scale = ${n};
|
|
let width_scale = ${i};
|
|
let in_y = ${a};
|
|
if( in_y < 0.0 || in_y > ${e} ) {
|
|
setOutputAtIndex(index, uniforms.extrapolationValue);
|
|
return;
|
|
}
|
|
let in_x = ${o};
|
|
if( in_x < 0.0 || in_x > ${t} ) {
|
|
setOutputAtIndex(index, uniforms.extrapolationValue);
|
|
return;
|
|
}
|
|
let sourceFracIndexCR = vec2<f32>(in_x,in_y);
|
|
if(${this.methodId} == 1) {
|
|
// Compute the four integer indices.
|
|
let sourceFloorCR = vec2<i32>(sourceFracIndexCR);
|
|
let sourceCeilCR = vec2<i32>(ceil(sourceFracIndexCR));
|
|
let topLeft = getImage(bInd, sourceFloorCR.y, sourceFloorCR.x, d);
|
|
let bottomLeft = getImage(bInd, sourceCeilCR.y, sourceFloorCR.x, d);
|
|
let topRight = getImage(bInd, sourceFloorCR.y, sourceCeilCR.x, d);
|
|
let bottomRight = getImage(bInd, sourceCeilCR.y, sourceCeilCR.x, d);
|
|
let fracCR = sourceFracIndexCR - vec2<f32>(sourceFloorCR);
|
|
let top = topLeft + (topRight - topLeft) * fracCR.x;
|
|
let bottom = bottomLeft + (bottomRight - bottomLeft) * fracCR.x;
|
|
let newValue = top + (bottom - top) * fracCR.y;
|
|
setOutputAtIndex(index, newValue);
|
|
} else {
|
|
// Compute the coordinators of nearest neighbor point.
|
|
let sourceNearestCR = vec2<i32>(floor(
|
|
sourceFracIndexCR + vec2<f32>(0.5,0.5)));
|
|
let newValue = getImage(
|
|
bInd, sourceNearestCR.y, sourceNearestCR.x, d);
|
|
setOutputAtIndex(index, newValue);
|
|
}
|
|
}
|
|
}
|
|
`}},dce=e=>{let{inputs:t,backend:r,attrs:n}=e,{image:a,boxes:s,boxInd:i}=t,{cropSize:o,method:l,extrapolationValue:u}=n,d=new uce(a.shape[3],s.shape,o,l),h=[{type:"float32",data:[u]}];return r.runWebGPUProgram(d,[a,s,i],"float32",h)},pce={kernelName:Zo,backendName:"webgpu",kernelFunc:dce},pv=class{constructor(e,t,r,n){this.variableNames=["x"],this.uniforms="index : f32,",this.size=!0;let a=128;this.workGroupSize=[a,1,1],this.outputShape=t,this.dispatchLayout=Ye(this.outputShape),this.dispatch=Oe(this.dispatchLayout,this.outputShape,this.workGroupSize),this.exclusive=r,this.reverse=n,this.op=e,this.shaderKey=`cum_${this.op}_${this.exclusive}_${this.reverse}`}getUserCode(){let e=this.outputShape.length,t=this.op==="*"?"1.0":"0.0",r=this.exclusive?t:`getX(${hv(e,"coords",this.op)})`,n=this.outputShape[this.outputShape.length-1],a="",s="";return this.exclusive?(a=this.reverse?`end != ${n-1}`:"end != 0",s=this.reverse?"end + 1":"end - 1"):(a=this.reverse?`end + pow2 < ${n}`:"end >= pow2",s=this.reverse?"end + pow2":"end - pow2"),`
|
|
${rt()}
|
|
if (index < uniforms.size) {
|
|
var coords = getCoordsFromIndex(index);
|
|
|
|
let end = ${cv(e,"coords",this.op)};
|
|
var val = ${r};
|
|
let pow2 = i32(pow(2.0, uniforms.index));
|
|
if (${a}) {
|
|
let idx = ${s};
|
|
${cv(e,"coords",this.op)} = idx;
|
|
val ${this.op}= getX(${hv(e,"coords",this.op)});
|
|
}
|
|
setOutputAtIndex(index, val);
|
|
}
|
|
}
|
|
`}};function hv(e,t,r){if(e===1)return`${t}`;if(e===2)return`${t}.x, ${t}.y`;if(e===3)return`${t}.x, ${t}.y, ${t}.z`;if(e===4)return`${t}.x, ${t}.y, ${t}.z, ${t}.w`;throw Error(`Cumulative ${r} for rank ${e} is not yet supported`)}function cv(e,t,r){if(e===1)return`${t}`;if(e===2)return`${t}.y`;if(e===3)return`${t}.z`;if(e===4)return`${t}.w`;throw Error(`Cumulative ${r} for rank ${e} is not yet supported`)}function J9(e,t,r,n,a,s){let i=t.shape.length,o=C.getAxesPermutation([n],i),l=t;o!=null&&(l=Ya({inputs:{x:t},backend:r,attrs:{perm:o}}));let u=C.getInnerMostAxes(1,i)[0];if(u!==i-1)throw new Error(`WebGPU cumprod shader expects an inner-most axis=${t.shape.length-1} but got axis=${n}`);let d=l.shape[u],h=zn({inputs:{x:l},backend:r});for(let p=0;p<=Math.ceil(Math.log2(d))-1;p++){let c=new pv(e,l.shape,!1,s),f=h,m=[{type:"float32",data:[p]}];h=r.runWebGPUProgram(c,[h],h.dtype,m),r.disposeData(f.dataId)}if(a){let p=new pv(e,l.shape,a,s),c=h,f=[{type:"float32",data:[0]}];h=r.runWebGPUProgram(p,[h],h.dtype,f),r.disposeData(c.dataId)}if(o!=null){let p=C.getUndoAxesPermutation(o),c=Ya({inputs:{x:h},backend:r,attrs:{perm:p}});return r.disposeData(h.dataId),r.disposeData(l.dataId),c}return h}function hce(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{axis:s,exclusive:i,reverse:o}=n;return J9("*",a,r,s,i,o)}var cce={kernelName:Ko,backendName:"webgpu",kernelFunc:hce};function fce(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{axis:s,exclusive:i,reverse:o}=n;return J9("+",a,r,s,i,o)}var mce={kernelName:ii,backendName:"webgpu",kernelFunc:fce},gce=class{constructor(e,t){this.variableNames=["x"],this.workGroupSize=[64,1,1],this.size=!0,this.uniforms="blockSize : i32,",this.outputShape=e,this.dispatchLayout=Ye(this.outputShape),this.dispatch=Oe(this.dispatchLayout,this.outputShape,this.workGroupSize),this.shaderKey=`depthToSpace_${t}`,this.dataFormat=t}getUserCode(){return`
|
|
${rt()}
|
|
if (index < uniforms.size) {
|
|
let coords = getCoordsFromIndex(index);
|
|
let b = coords[0];
|
|
let h = ${this.getHeightCoordString()};
|
|
let w = ${this.getWidthCoordString()};
|
|
let d = ${this.getDepthCoordString()};
|
|
|
|
let in_h = h / uniforms.blockSize;
|
|
let offset_h = h % uniforms.blockSize;
|
|
let in_w = w / uniforms.blockSize;
|
|
let offset_w = w % uniforms.blockSize;
|
|
let offset_d = (offset_h * uniforms.blockSize + offset_w) *
|
|
${this.getOutputDepthSize()};
|
|
let in_d = d + offset_d;
|
|
|
|
let rlt = ${this.getInputSamplingString()};
|
|
setOutputAtIndex(index, rlt);
|
|
}
|
|
}`}getHeightCoordString(){return this.dataFormat==="NHWC"?"coords[1]":"coords[2]"}getWidthCoordString(){return this.dataFormat==="NHWC"?"coords[2]":"coords[3]"}getDepthCoordString(){return this.dataFormat==="NHWC"?"coords[3]":"coords[1]"}getOutputDepthSize(){return this.dataFormat==="NHWC"?"uniforms.outShape[3]":"uniforms.outShape[1]"}getInputSamplingString(){return this.dataFormat==="NHWC"?"getX(b, in_h, in_w, in_d)":"getX(b, in_d, in_h, in_w)"}};function yce(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{blockSize:s,dataFormat:i}=n,o=a.shape[0],l=i==="NHWC"?a.shape[1]:a.shape[2],u=i==="NHWC"?a.shape[2]:a.shape[3],d=i==="NHWC"?a.shape[3]:a.shape[1],h=l*s,p=u*s,c=d/(s*s),f=i==="NHWC"?[o,h,p,c]:[o,c,h,p],m=[{type:"int32",data:[s]}],g=new gce(f,i);return r.runWebGPUProgram(g,[a],a.dtype,m)}var Ace={kernelName:Yo,backendName:"webgpu",kernelFunc:yce},Q9=class{constructor(e,t=!1,r=null,n=!1){this.variableNames=["x","W"],this.uniforms="pad : vec2<i32>, stride : vec2<i32>, dilation : vec2<i32>, inDims : vec2<i32>,",this.workGroupSize=[4,4,4],this.isVec4=!0,this.outputShape=e.outShape,this.dispatchLayout={x:[0,1],y:[2],z:[3]},this.dispatch=Oe(this.dispatchLayout,this.outputShape,this.workGroupSize,[1,4,4]),v.assert(e.dataFormat==="channelsLast",()=>"TODO: NCHW is unimplemented"),t&&this.variableNames.push("bias"),n&&this.variableNames.push("preluActivationWeights"),this.convInfo=e,this.addBias=t,this.activation=r,this.hasPreluActivation=n,this.shaderKey=`depthwise3x3_${r}`}getUserCode(){let e="",t="";if(this.activation){let n=Zi(this.activation,this.isVec4);this.hasPreluActivation?e=`fn activation(a : vec4<f32>, outCoord : vec4<i32>) -> vec4<f32> {
|
|
let b = getPreluActivationWeightsByOutputCoords(outCoord);
|
|
${n}
|
|
}`:e=`
|
|
fn activation(a : vec4<f32>, outCoord : vec4<i32>) -> vec4<f32> {
|
|
${n}
|
|
}
|
|
`,t="dotProd[i] = activation(dotProd[i], coords);"}let r=this.addBias?"dotProd[i] = dotProd[i] + getBiasByOutputCoords(coords);":"";return`
|
|
${e}
|
|
|
|
${S5()}
|
|
fn main(@builtin(global_invocation_id) globalId: vec3<u32>) {
|
|
let batch = 0;
|
|
let r = i32(globalId.x);
|
|
let c = i32(globalId.y) * 4;
|
|
let d2 = i32(globalId.z) * 4;
|
|
let xRCCorner = vec2<i32>(r, c) * uniforms.stride - uniforms.pad;
|
|
let d1 = d2;
|
|
let q = 0;
|
|
|
|
let xRCorner = xRCCorner.x;
|
|
let xCCorner = xRCCorner.y;
|
|
|
|
var wVals : array<vec4<f32>, 9>;
|
|
wVals[0] = getW(0, 0, d1, q);
|
|
wVals[1] = getW(0, 1, d1, q);
|
|
wVals[2] = getW(0, 2, d1, q);
|
|
wVals[3] = getW(1, 0, d1, q);
|
|
wVals[4] = getW(1, 1, d1, q);
|
|
wVals[5] = getW(1, 2, d1, q);
|
|
wVals[6] = getW(2, 0, d1, q);
|
|
wVals[7] = getW(2, 1, d1, q);
|
|
wVals[8] = getW(2, 2, d1, q);
|
|
|
|
var xVals : array<array<vec4<f32>, 6>, 3>;
|
|
for (var wR = 0; wR < 3; wR = wR + 1) {
|
|
let xR = xRCorner + wR * uniforms.dilation[0];
|
|
for (var wC = 0; wC < 6; wC = wC + 1) {
|
|
let xC = xCCorner + wC * uniforms.dilation[1];
|
|
if (xR < 0 || xR >= uniforms.inDims[0] || xC < 0 || xC >= uniforms.inDims[1]) {
|
|
xVals[wR][wC] = vec4<f32>(0.0);
|
|
} else {
|
|
xVals[wR][wC] = getX(batch, xR, xC, d1);
|
|
}
|
|
}
|
|
}
|
|
|
|
var dotProd : array<vec4<f32>, 4>;
|
|
dotProd[0] = vec4<f32>(0.0);
|
|
dotProd[1] = vec4<f32>(0.0);
|
|
dotProd[2] = vec4<f32>(0.0);
|
|
dotProd[3] = vec4<f32>(0.0);
|
|
|
|
for (var wR = 0; wR < 3; wR = wR + 1) {
|
|
for (var wC = 0; wC < 3; wC = wC + 1) {
|
|
let indexW = wR * 3 + wC;
|
|
dotProd[0] = dotProd[0] + xVals[wR][0 + wC] * wVals[indexW];
|
|
dotProd[1] = dotProd[1] + xVals[wR][1 + wC] * wVals[indexW];
|
|
dotProd[2] = dotProd[2] + xVals[wR][2 + wC] * wVals[indexW];
|
|
dotProd[3] = dotProd[3] + xVals[wR][3 + wC] * wVals[indexW];
|
|
}
|
|
}
|
|
|
|
for (var i = 0; i < 4; i = i + 1) {
|
|
let coords = vec4<i32>(batch, r, c + i, d2);
|
|
if (coordsInBounds4D(coords, uniforms.outShape)) {
|
|
${r}
|
|
${t}
|
|
setOutputAtCoords(coords[0], coords[1], coords[2], coords[3], dotProd[i]);
|
|
}
|
|
}
|
|
}
|
|
`}},eS=class{constructor(e,t=!1,r=null,n=!1){this.variableNames=["x","W"],this.uniforms=`pad : vec2<i32>, stride : vec2<i32>, dilation : vec2<i32>,
|
|
inDims : vec2<i32>, filterHeight : i32, filterWidth : i32,
|
|
channelMul : i32,`,this.workGroupSize=[256,1,1],this.outputShape=e.outShape,this.dispatchLayout=Ye(this.outputShape),this.dispatch=Oe(this.dispatchLayout,this.outputShape,this.workGroupSize),v.assert(e.dataFormat==="channelsLast",()=>"TODO: NCHW is unimplemented"),t&&this.variableNames.push("bias"),n&&this.variableNames.push("preluActivationWeights"),this.convInfo=e,this.addBias=t,this.activation=r,this.hasPreluActivation=n,this.shaderKey=`depthwise_${this.activation}`}getUserCode(){let e="",t="";if(this.activation){let n=Zi(this.activation,!1);this.hasPreluActivation?e=`fn activation(a : f32, outCoord : vec4<i32>) -> f32 {
|
|
let b = getPreluActivationWeightsByOutputCoords(outCoord);
|
|
${n}
|
|
}`:e=`
|
|
fn activation(a : f32, outCoord : vec4<i32>) -> f32 {
|
|
${n}
|
|
}
|
|
`,t="dotProd = activation(dotProd, coords);"}let r=this.addBias?"dotProd = dotProd + getBiasByOutputCoords(coords);":"";return`
|
|
${e}
|
|
|
|
fn writeResult(batch : i32, row : i32, col : i32, chan : i32,
|
|
value : f32) {
|
|
let coord = vec4<i32>(batch, row, col, chan);
|
|
if (coordsInBounds4D(coord, uniforms.outShape)) {
|
|
setOutputAtCoords(batch, row, col, chan, value);
|
|
}
|
|
}
|
|
|
|
${Wl()}
|
|
let coords = getOutputCoords();
|
|
let batch = coords[0];
|
|
let xRCCorner = vec2<i32>(coords.yz) * uniforms.stride - uniforms.pad;
|
|
let d2 = coords[3];
|
|
let d1 = d2 / uniforms.channelMul;
|
|
let q = d2 - d1 * uniforms.channelMul;
|
|
|
|
let inputRowStart = xRCCorner.x;
|
|
let inputColStart = xRCCorner.y;
|
|
let inputRowEnd = inputRowStart + uniforms.filterHeight *
|
|
uniforms.dilation[0];
|
|
let inputColEnd = inputColStart + uniforms.filterWidth *
|
|
uniforms.dilation[1];
|
|
|
|
// Convolve x(?, ?, d1) with w(:, :, d1, q) to get y(yR, yC, d2).
|
|
// ? = to be determined. : = across all values in that axis.
|
|
var dotProd = 0.0;
|
|
|
|
// Extract if checking out of for loop for performance.
|
|
if (inputRowStart >= 0 && inputColStart >= 0 &&
|
|
inputRowEnd < uniforms.inDims[0] &&
|
|
inputColEnd < uniforms.inDims[1]) {
|
|
// Here using a constant value |this.convInfo.filterHeight| instead
|
|
// of uniform value is in order to loop unrolling.
|
|
for (var wR = 0; wR < uniforms.filterHeight; wR = wR + 1) {
|
|
let xR = inputRowStart + wR * uniforms.dilation[0];
|
|
|
|
for (var wC = 0; wC < uniforms.filterWidth; wC = wC + 1) {
|
|
let xC = inputColStart + wC * uniforms.dilation[1];
|
|
|
|
let xVal = getX(batch, xR, xC, d1);
|
|
let wVal = getW(wR, wC, d1, q);
|
|
dotProd = dotProd + xVal * wVal;
|
|
}
|
|
}
|
|
} else {
|
|
for (var wR = 0; wR < uniforms.filterHeight; wR = wR + 1) {
|
|
let xR = inputRowStart + wR * uniforms.dilation[0];
|
|
|
|
if (xR < 0 || xR >= uniforms.inDims[0]) {
|
|
continue;
|
|
}
|
|
|
|
for (var wC = 0; wC < uniforms.filterWidth; wC = wC + 1) {
|
|
let xC = inputColStart + wC * uniforms.dilation[1];
|
|
|
|
if (xC < 0 || xC >= uniforms.inDims[1]) {
|
|
continue;
|
|
}
|
|
|
|
let xVal = getX(batch, xR, xC, d1);
|
|
let wVal = getW(wR, wC, d1, q);
|
|
dotProd = dotProd + xVal * wVal;
|
|
}
|
|
}
|
|
}
|
|
|
|
${r}
|
|
${t}
|
|
writeResult(batch, coords[1], coords[2], d2, dotProd);
|
|
}
|
|
`}};function xce(e){let{inputs:t,backend:r,attrs:n}=e,{x:a,filter:s}=t,{strides:i,pad:o,dilations:l,dimRoundingMode:u}=n,d=l;d==null&&(d=[1,1]);let h=C.computeConv2DInfo(a.shape,s.shape,i,d,o,u,!0),p=[{type:"int32",data:[h.padInfo.top,h.padInfo.left]},{type:"int32",data:[h.strideHeight,h.strideWidth]},{type:"int32",data:[h.dilationHeight,h.dilationWidth]},{type:"int32",data:[h.inHeight,h.inWidth]}],c;return h.batchSize===1&&h.inHeight===h.outHeight&&h.inWidth===h.outWidth&&h.strideHeight===1&&h.strideWidth===1&&h.filterHeight===h.filterWidth&&h.inChannels===h.outChannels&&h.dilationHeight===1&&h.dilationWidth===1&&h.filterHeight===3&&h.inChannels%4===0?c=new Q9(h):(c=new eS(h),p.push({type:"int32",data:[h.filterHeight]},{type:"int32",data:[h.filterWidth]},{type:"int32",data:[h.outChannels/h.inChannels]})),r.runWebGPUProgram(c,[a,s],a.dtype,p)}var bce={kernelName:oi,backendName:"webgpu",kernelFunc:xce},tS=qr({opSnippet:0,cpuKernelImpl:Jpe,supportsComplex:!0}),vce={kernelName:Si,backendName:"webgpu",kernelFunc:tS},wce=class{constructor(e,t){this.workGroupSize=[64,1,1],this.variableNames=["x"],this.uniforms="reduceSize : i32,",this.size=!0,this.inputShape=[e.batchSize,e.inSize];let[r]=C.computeOutAndReduceShapes(this.inputShape,[1]);this.outputShape=r.length===0?[1]:r,this.dispatchLayout=Ye(this.outputShape),this.dispatch=Oe(this.dispatchLayout,this.outputShape,[1,1,1]),this.reduceType=t,this.shaderKey=`reduce_${t}`}getUserCode(){let e="",t="0.0";this.reduceType==="min"||this.reduceType==="max"?(e=`
|
|
if (isnan(candidate)) {
|
|
bestValue = uniforms.NAN;
|
|
} else if (!isnan(bestValue) && candidate ${this.reduceType==="min"?"<":">"} bestValue)
|
|
{ bestValue = candidate; }`,t="f32(x[offset])"):this.reduceType==="sum"||this.reduceType==="mean"?e=" bestValue = bestValue + candidate; ":this.reduceType==="prod"&&(e=" bestValue = bestValue * candidate; ",t="1.0");let r=this.reduceType==="mean"?"setOutputAtIndex(outputIndex, bestValue / f32(uniforms.reduceSize));":"setOutputAtIndex(outputIndex, bestValue);";return`
|
|
fn DIV_CEIL(a : u32, b : u32) -> u32 {
|
|
return ((a - 1u) / b + 1u);
|
|
}
|
|
|
|
${`
|
|
var<workgroup> xBestValues : array<f32, ${this.workGroupSize[0]}>;
|
|
`}
|
|
fn getOffset(outputIndex : i32) -> i32 {
|
|
let outputCoords = getCoordsFromIndex(outputIndex);
|
|
let offset = ${this.outputShape.length===1?"outputCoords":"outputCoords[0]"} * uniforms.reduceSize;
|
|
return offset;
|
|
}
|
|
${rt()}
|
|
let outputIndex = index / i32(workGroupSizeX);
|
|
let offset = getOffset(outputIndex);
|
|
var bestValue = ${t};
|
|
let Length = uniforms.reduceSize;
|
|
let WorkPerThread = DIV_CEIL(u32(Length), workGroupSizeX);
|
|
for (var k = i32(localId.x); k < Length && outputIndex < uniforms.size;
|
|
k = k + i32(workGroupSizeX)) {
|
|
let candidate = f32(x[offset + k]);
|
|
${e}
|
|
}
|
|
xBestValues[localId.x] = bestValue;
|
|
workgroupBarrier();
|
|
|
|
var reduceSize = min(u32(Length), workGroupSizeX);
|
|
for (var currentSize = reduceSize / 2u; reduceSize > 1u;
|
|
currentSize = reduceSize / 2u) {
|
|
let interval = DIV_CEIL(reduceSize, 2u);
|
|
if (localId.x < currentSize) {
|
|
let candidate = xBestValues[localId.x + interval];
|
|
${e}
|
|
xBestValues[localId.x] = bestValue;
|
|
}
|
|
reduceSize = interval;
|
|
workgroupBarrier();
|
|
}
|
|
|
|
if (localId.x == 0u && outputIndex < uniforms.size) {
|
|
${r}
|
|
}
|
|
}
|
|
`}};function qh(e,t,r,n,a){let s=e.shape.length,i=[],o=v.parseAxisParam(t,e.shape),l=o,u=C.getAxesPermutation(l,s),d=e;u!=null&&(d=Ya({inputs:{x:e},attrs:{perm:u},backend:a}),l=C.getInnerMostAxes(l.length,s),i.push(d)),C.assertAxesAreInnerMostDims(n,l,s);let[h,p]=C.computeOutAndReduceShapes(d.shape,l),c=h;r&&(c=C.expandShapeToKeepDim(h,o));let f;if((n==="max"||n==="prod")&&a.shouldExecuteOnCPU([d])){let m=a.tensorMap.get(d.dataId).values;switch(n){case"max":let g=Kpe(m,v.sizeFromShape(p),c,e.dtype);f=a.makeTensorInfo(c,e.dtype,g);break;case"prod":let{outVals:y,outShape:A,outDtype:x}=the(d.shape,d.dtype,m,l);f=a.makeTensorInfo(A,x,y);break;default:throw new Error(`${n} CPU implementation is not yet supported.`)}}else{let m=v.sizeFromShape(p),g=v.sizeFromShape(d.shape)/m,y={windowSize:m,inSize:m,batchSize:g,outSize:1},A=n==="mean"?"float32":gh(e.dtype),x=[{type:"int32",data:[m]}],b=new wce(y,n),w=a.runWebGPUProgram(b,[d],A,x);i.push(w),f=Qe({inputs:{x:w},attrs:{shape:c},backend:a})}return i.forEach(m=>a.disposeData(m.dataId)),f}function $5(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{axis:s,keepDims:i}=n;return qh(a,s,i,"sum",r)}var kce={kernelName:Oi,backendName:"webgpu",kernelFunc:$5};function Ice(e){let{inputs:t,backend:r,attrs:n}=e,{equation:a}=n,s=t,{allDims:i,summedDims:o,idDims:l}=C.decodeEinsumEquation(a,s.length);C.checkEinsumDimSizes(i.length,l,s);let{path:u,steps:d}=C.getEinsumComputePath(o,l),h=d.length,p=null,c=i.length,f=[];for(let m=0;m<h;++m){for(let g of d[m]){let{permutationIndices:y,expandDims:A}=C.getEinsumPermutation(c,l[g]),x;C.isIdentityPermutation(y)?x=s[g]:(x=Ya({inputs:{x:s[g]},backend:r,attrs:{perm:y}}),f.push(x));let b=x.shape.slice();for(let w=0;w<A.length;++w)b.splice(A[w],0,1);v.arraysEqual(x.shape,b)||(x=Qe({inputs:{x},backend:r,attrs:{shape:b}}),f.push(x)),p===null?p=x:(p=tS({inputs:{a:x,b:p},backend:r}),f.push(p))}m<h-1&&(u[m]>=0&&(p=$5({inputs:{x:p},backend:r,attrs:{axis:u[m]-(i.length-c),keepDims:!1}}),f.push(p)),c--)}for(let m of f)m!==p&&r.disposeData(m.dataId);return p}var Sce={kernelName:rh,backendName:"webgpu",kernelFunc:Ice},Cce=kr({opType:4}),Tce={kernelName:ui,backendName:"webgpu",kernelFunc:Cce},Nce=qr({opSnippet:4,dtype:"bool",cpuKernelImpl:Dpe}),Ece={kernelName:Jo,backendName:"webgpu",kernelFunc:Nce},rS=kr({opType:5,cpuKernelImpl:Lpe,dtype:"float32"}),Rce={kernelName:di,backendName:"webgpu",kernelFunc:rS};function dy(e){let{inputs:t,attrs:r,backend:n}=e,{dim:a}=r,{input:s}=t,i=s.shape.length,o=s.shape.slice(),l=a;return a<0&&(v.assert(-(i+1)<=a,()=>`Axis must be in the interval [${-(i+1)}, ${i}]`),l=i+a+1),o.splice(l,0,1),Qe({inputs:{x:s},backend:n,attrs:{shape:o}})}var $ce={kernelName:Qo,backendName:"webgpu",kernelFunc:dy},Mce=kr({opType:6,cpuKernelImpl:Bpe}),Fce={kernelName:el,backendName:"webgpu",kernelFunc:Mce},Pce=class{constructor(e){this.variableNames=[],this.outputShape=[],this.uniforms="value : f32,",this.workGroupSize=[64,1,1],this.size=!0,this.outputShape=e,this.dispatchLayout=Ye(this.outputShape),this.dispatch=Oe(this.dispatchLayout,this.outputShape,this.workGroupSize),this.shaderKey="fill"}getUserCode(){return`
|
|
${rt()}
|
|
if (index < uniforms.size) {
|
|
setOutputAtIndex(index, uniforms.value);
|
|
}
|
|
}
|
|
`}};function Fd(e){let{backend:t,attrs:r}=e,{shape:n,value:a}=r,{dtype:s}=r;if(s=s||v.inferDtype(a),s==="string"){let i=v.getArrayFromDType(s,v.sizeFromShape(n));return i.fill(a),t.makeTensorInfo(n,s,i)}else{let i=new Pce(n),o=[{type:"float32",data:[a]}];return t.runWebGPUProgram(i,[],s,o)}}var _ce={kernelName:Ku,backendName:"webgpu",kernelFunc:Fd},zce=class{constructor(e){this.outputShape=[],this.variableNames=["x"],this.workGroupSize=[64,1,1],this.size=!0,this.outputShape=e,this.dispatchLayout=Ye(this.outputShape),this.dispatch=Oe(this.dispatchLayout,this.outputShape,this.workGroupSize),this.shaderKey="flipLeftRight"}getUserCode(){return`
|
|
${rt()}
|
|
if (index < uniforms.size) {
|
|
let coords = getCoordsFromIndex(index);
|
|
let coordX = uniforms.xShape[2] - coords[2] - 1;
|
|
let outputValue = getX(coords[0], coords[1], coordX, coords[3]);
|
|
setOutputAtIndex(index, outputValue);
|
|
}
|
|
}
|
|
`}},Oce={kernelName:tl,backendName:"webgpu",kernelFunc:({inputs:e,backend:t})=>{let{image:r}=e,n=t,a=new zce(r.shape);return n.runWebGPUProgram(a,[r],r.dtype)}},Dce=kr({opType:7,cpuKernelImpl:Wpe}),Lce={kernelName:pi,backendName:"webgpu",kernelFunc:Dce},Bce=qr({opSnippet:12,dtype:"int32"}),Wce={kernelName:hi,backendName:"webgpu",kernelFunc:Bce},Vce=class{constructor(e,t=!1){this.outputShape=[0],this.variableNames=[],this.workGroupSize=[256,1,1],this.outputShape=e,this.dispatchLayout=Ye(this.outputShape),this.dispatch=Oe(this.dispatchLayout,this.outputShape,this.workGroupSize),this.useImport=t,this.shaderKey=`fromPixels_${this.useImport}`}getUserCode(){let e=this.useImport?"textureLoad(src, vec2<i32>(coords.yx));":"textureLoad(src, vec2<i32>(coords.yx), 0)";return`
|
|
@binding(1) @group(0) var src: ${this.useImport?"texture_external":"texture_2d<f32>"};
|
|
|
|
${rt()}
|
|
let flatIndexBase = index * uniforms.numChannels;
|
|
for (var i = 0; i < uniforms.numChannels; i = i + 1) {
|
|
let flatIndex = flatIndexBase + i;
|
|
if (flatIndex < uniforms.size) {
|
|
let coords = getCoordsFromIndex(flatIndexBase);
|
|
let values = ${e};
|
|
result[flatIndex] = i32(floor(255.0 * values[i]));
|
|
}
|
|
}
|
|
}
|
|
`}},Uce={kernelName:Op,backendName:"webgpu",kernelFunc:Gce},hu;function Gce(e){let{inputs:t,backend:r,attrs:n}=e,{pixels:a}=t,{numChannels:s}=n;if(a==null)throw new Error("pixels passed to tf.browser.fromPixels() can not be null");let i=typeof HTMLVideoElement!="undefined"&&a instanceof HTMLVideoElement,o=typeof HTMLImageElement!="undefined"&&a instanceof HTMLImageElement,l=typeof HTMLCanvasElement!="undefined"&&a instanceof HTMLCanvasElement||typeof OffscreenCanvas!="undefined"&&a instanceof OffscreenCanvas,u=typeof ImageBitmap!="undefined"&&a instanceof ImageBitmap,[d,h]=i?[a.videoWidth,a.videoHeight]:[a.width,a.height],p=[h,d,s];if(Z().getBool("WEBGPU_USE_IMPORT")&&i)return fv({externalImage:a,backend:r,attrs:n,outShape:p,useImport:!0});if((i||o)&&(hu==null&&(hu=document.createElement("canvas").getContext("2d")),hu.canvas.width=d,hu.canvas.height=h,hu.drawImage(a,0,0,d,h),a=hu.canvas),u||l||i||o)return fv({externalImage:a,backend:r,attrs:n,outShape:p,useImport:!1});let c=a.data,f=c;if(s!=null&&s!==4){f=new Uint8Array(a.width*a.height*s);let y=c.length,A=0;for(let x=0;x<y;x++)x%4<s&&(f[A++]=c[x])}let m=r.makeTensorInfo(p,"int32"),g=r.tensorMap.get(m.dataId);return g.values=new Int32Array(f),r.maybeReleaseBuffer(m.dataId),r.uploadToGPU(m.dataId),m}function fv(e){let{externalImage:t,backend:r,attrs:n,outShape:a,useImport:s}=e,{numChannels:i}=n,o=v.sizeFromShape(a),l=v.computeStrides(a),u=new Vce(a,s),d=[{type:"uint32",data:[o]},{type:"uint32",data:[i]},{type:"uint32",data:[...l]},{type:"uint32",data:[...u.dispatch]}];return r.runFromPixelsProgram(u,a,d,s,t)}var jce=class{constructor(e,t,r,n,a){this.uniforms="varianceEpsilon : f32,",this.workGroupSize=[128,1,1],this.size=!0,this.variableNames=["x","mean","variance"],C.assertAndGetBroadcastShape(e,t),C.assertAndGetBroadcastShape(e,r),this.outputShape=e,this.dispatchLayout=Ye(this.outputShape),this.dispatch=Oe(this.dispatchLayout,this.outputShape,this.workGroupSize),n!=null&&(C.assertAndGetBroadcastShape(e,n),this.variableNames.push("offset")),a!=null&&(C.assertAndGetBroadcastShape(e,a),this.variableNames.push("scale")),this.offsetShape=n,this.scaleShape=a,this.shaderKey="batchNorm"}getUserCode(){let e="0.0";this.offsetShape!=null&&(e="getOffsetByOutputIndex(index)");let t="1.0";return this.scaleShape!=null&&(t="getScaleByOutputIndex(index)"),`
|
|
${rt()}
|
|
if (index < uniforms.size)
|
|
{
|
|
let xValue = getXByOutputIndex(index);
|
|
let meanValue = getMeanByOutputIndex(index);
|
|
let varianValue = getVarianceByOutputIndex(index);
|
|
let offsetValue = ${e};
|
|
let scaleValue = ${t};
|
|
let inv = scaleValue * inverseSqrt(varianValue + f32(uniforms.varianceEpsilon));
|
|
setOutputAtIndex(index,dot(vec3<f32>(xValue, -meanValue, offsetValue), vec3<f32>(inv, inv, 1.0)));
|
|
}
|
|
}
|
|
`}},Hce={kernelName:ci,backendName:"webgpu",kernelFunc:({inputs:e,attrs:t,backend:r})=>{let{x:n,scale:a,offset:s,mean:i,variance:o}=e,{varianceEpsilon:l}=t,u=r,d=[n,i,o],h=null;s!=null&&(h=s.shape,d.push(s));let p=null;a!=null&&(p=a.shape,d.push(a));let c=new jce(n.shape,i.shape,o.shape,h,p),f=[{type:"float32",data:[l]}];return u.runWebGPUProgram(c,d,n.dtype,f)}};function qce(e){let{inputs:t,backend:r,attrs:n}=e,{x:a,filter:s,bias:i,preluActivationWeights:o}=t,{strides:l,pad:u,dataFormat:d,dilations:h,dimRoundingMode:p,activation:c,leakyreluAlpha:f}=n,m=C.convertConv2DDataFormat(d),g=C.computeConv2DInfo(a.shape,s.shape,l,h,u,p,!1,m);return Y9({x:a,filter:s,convInfo:g,backend:r,bias:i,preluActivationWeights:o,leakyreluAlpha:f,activation:c})}var Xce={kernelName:_s,backendName:"webgpu",kernelFunc:qce};function Kce(e){let{inputs:t,backend:r,attrs:n}=e,{x:a,filter:s,bias:i,preluActivationWeights:o}=t,{strides:l,pad:u,dilations:d,dimRoundingMode:h,activation:p,leakyreluAlpha:c}=n,f=d;f==null&&(f=[1,1]),v.assert(C.eitherStridesOrDilationsAreOne(l,f),()=>`Error in depthwiseConv2d: Either strides or dilations must be 1. Got strides ${l} and dilations '${f}'`);let m=C.computeConv2DInfo(a.shape,s.shape,l,f,u,h,!0),g=[a,s],y=i!=null,A=o!=null;y&&g.push(i),A&&g.push(o);let x=[{type:"int32",data:[m.padInfo.top,m.padInfo.left]},{type:"int32",data:[m.strideHeight,m.strideWidth]},{type:"int32",data:[m.dilationHeight,m.dilationWidth]},{type:"int32",data:[m.inHeight,m.inWidth]}],b;return m.batchSize===1&&m.inHeight===m.outHeight&&m.inWidth===m.outWidth&&m.strideHeight===1&&m.strideWidth===1&&m.filterHeight===m.filterWidth&&m.inChannels===m.outChannels&&m.dilationHeight===1&&m.dilationWidth===1&&m.filterHeight===3&&m.inChannels%4===0?b=new Q9(m,y,p,A):(b=new eS(m,y,p,A),x.push({type:"int32",data:[m.filterHeight]},{type:"int32",data:[m.filterWidth]},{type:"int32",data:[m.outChannels/m.inChannels]})),p==="leakyrelu"&&(x.push({type:"float32",data:[c]}),b.uniforms+=" alpha : f32,"),r.runWebGPUProgram(b,g,"float32",x)}var Zce={kernelName:zs,backendName:"webgpu",kernelFunc:Kce},Yce=class{constructor(e,t){this.variableNames=["A","indices"],this.workGroupSize=[64,1,1],this.size=!0,this.outputShape=t,this.dispatchLayout=Ye(this.outputShape),this.dispatch=Oe(this.dispatchLayout,this.outputShape,this.workGroupSize),this.shaderKey=`gathernd_${e}`,this.sliceDim=e,this.uniforms=`sliceDim : i32, strides : ${xr(e)},`}getUserCode(){let e;return this.sliceDim>1?e="uniforms.strides[j]":e="uniforms.strides",`
|
|
${rt()}
|
|
if (index < uniforms.size) {
|
|
let coords = getCoordsFromIndex(index);
|
|
var flattenIndex = 0;
|
|
for (var j = 0; j < uniforms.sliceDim; j = j + 1) {
|
|
let indexTemp = i32(round(getIndices(coords[0], j)));
|
|
let strideNum = ${e};
|
|
flattenIndex = flattenIndex + indexTemp * strideNum;
|
|
}
|
|
|
|
setOutputAtIndex(index, getA(flattenIndex, coords[1]));
|
|
}
|
|
}
|
|
`}};function Jce(e){let{inputs:t,backend:r}=e,{params:n,indices:a}=t,s=a.shape,i=s[s.length-1],o=v.sizeFromShape(n.shape),[l,u,d,h]=C.prepareAndValidate(n,a),p=Qe({inputs:{x:a},backend:r,attrs:{shape:[u,i]}}),c=Qe({inputs:{x:n},backend:r,attrs:{shape:[v.sizeFromShape(n.shape)/d,d]}});if(r.shouldExecuteOnCPU([n,a])||n.dtype==="string"){let A=r.readSync(a.dataId),x=r.bufferSync(n),b=Vpe(A,x,n.dtype,u,i,d,h,n.shape,o);return r.makeTensorInfo(l,n.dtype,b.values)}let f=new Yce(i,[u,d]),m=[{type:"int32",data:[i]},{type:"int32",data:h}],g=r.runWebGPUProgram(f,[c,p],c.dtype,m),y=Qe({inputs:{x:g},backend:r,attrs:{shape:l}});return r.disposeData(p.dataId),r.disposeData(c.dataId),r.disposeData(g.dataId),y}var Qce={kernelName:nl,backendName:"webgpu",kernelFunc:Jce},e0e=class{constructor(e,t){this.variableNames=["A","indices"],this.workGroupSize=[64,1,1],this.size=!0,this.outputShape=e.slice(),this.aShape=e,this.outputShape=t,this.dispatchLayout=Ye(this.outputShape),this.dispatch=Oe(this.dispatchLayout,this.outputShape,this.workGroupSize),this.shaderKey="gather"}getUserCode(){let e=t0e(this.aShape);return`
|
|
${rt()}
|
|
if (index < uniforms.size) {
|
|
let resRC = getCoordsFromIndex(index);
|
|
let indexZ = i32(getIndices(resRC.x, resRC.z));
|
|
let inBounds = select(0.0, 1.0, indexZ >= 0 && indexZ < uniforms.aShape[2]);
|
|
setOutputAtIndex(index, inBounds * getA(${e}));
|
|
}
|
|
}
|
|
`}};function t0e(e){let t=["resRC.x","resRC.y","resRC.z","resRC.w"],r=[];for(let n=0;n<e.length;n++)n===2?r.push("indexZ"):r.push(`${t[n]}`);return r.join()}function nS(e){let{inputs:t,backend:r,attrs:n}=e,{x:a,indices:s}=t,{axis:i,batchDims:o}=n,l=v.parseAxisParam(i,a.shape)[0],u=C.segment_util.collectGatherOpShapeInfo(a,s,l,o),d=v.sizeFromShape(s.shape),h=[],p=Qe({inputs:{x:a},backend:r,attrs:{shape:[u.batchSize,u.outerSize,u.dimSize,u.sliceSize]}}),c=Qe({inputs:{x:s},backend:r,attrs:{shape:[u.batchSize,d/u.batchSize]}});h.push(p),h.push(c);let f=[u.batchSize,u.outerSize,d/u.batchSize,u.sliceSize];if(r.shouldExecuteOnCPU([a,s])){let A=r.tensorMap.get(c.dataId).values,x=We(c.shape,c.dtype,A),b=r.tensorMap.get(p.dataId).values,w=We(p.shape,p.dtype,b),S=Upe(w,x,f);return h.forEach(N=>r.disposeData(N.dataId)),r.makeTensorInfo(u.outputShape,S.dtype,S.values)}let m=new e0e(p.shape,f),g=r.runWebGPUProgram(m,[p,c],p.dtype);h.push(g);let y=Qe({inputs:{x:g},backend:r,attrs:{shape:u.outputShape}});return h.forEach(A=>r.disposeData(A.dataId)),y}var r0e={kernelName:rl,backendName:"webgpu",kernelFunc:nS},n0e=qr({opSnippet:5,cpuKernelImpl:jpe,dtype:"bool"}),a0e={kernelName:al,backendName:"webgpu",kernelFunc:n0e},s0e=qr({opSnippet:6,dtype:"bool",cpuKernelImpl:Gpe}),i0e={kernelName:fi,backendName:"webgpu",kernelFunc:s0e};function o0e(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{alpha:s}=n,i=[{type:"float32",data:[s]}],o=new jh(a.shape,14);return o.uniforms="alpha : f32,",r.runWebGPUProgram(o,[a],"float32",i)}var l0e={kernelName:gi,backendName:"webgpu",kernelFunc:o0e},u0e=qr({opSnippet:7,dtype:"bool",cpuKernelImpl:qpe}),d0e={kernelName:sl,backendName:"webgpu",kernelFunc:u0e},p0e=qr({opSnippet:8,dtype:"bool",cpuKernelImpl:Hpe}),h0e={kernelName:il,backendName:"webgpu",kernelFunc:p0e},c0e=kr({opType:9,cpuKernelImpl:Xpe}),f0e={kernelName:yi,backendName:"webgpu",kernelFunc:c0e},m0e=qr({opSnippet:9,dtype:"bool"}),g0e={kernelName:ol,backendName:"webgpu",kernelFunc:m0e},y0e=kr({opType:10}),A0e={kernelName:ed,backendName:"webgpu",kernelFunc:y0e};function aS(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{reductionIndices:s,keepDims:i}=n;return qh(a,s,i,"max",r)}var x0e={kernelName:Ai,backendName:"webgpu",kernelFunc:aS},b0e=qr({opSnippet:15,cpuKernelImpl:Zpe}),v0e={kernelName:xi,backendName:"webgpu",kernelFunc:b0e};function w0e(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{filterSize:s,strides:i,pad:o,dimRoundingMode:l}=n,u=1,d=C.computePool2DInfo(a.shape,s,i,u,o,l),h,p=[];if(d.filterHeight===1&&d.filterWidth===1){if(v.arraysEqual(d.inShape,d.outShape))return zn({inputs:{x:a},backend:r});h=new X9(d),p.push({type:"int32",data:[d.strideHeight,d.strideWidth]})}else h=new q9(d,"max"),p.push({type:"int32",data:[d.strideHeight,d.strideWidth]},{type:"int32",data:[d.padInfo.top,d.padInfo.left]},{type:"int32",data:[d.dilationHeight,d.dilationWidth]},{type:"int32",data:[d.inHeight,d.inWidth]},{type:"int32",data:[d.effectiveFilterHeight,d.effectiveFilterWidth]});return r.runWebGPUProgram(h,[a],a.dtype,p)}var k0e={kernelName:bi,backendName:"webgpu",kernelFunc:w0e};function I0e(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{keepDims:s,axis:i}=n;return qh(a,i,s,"mean",r)}var S0e={kernelName:vi,backendName:"webgpu",kernelFunc:I0e};function C0e(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{axis:s,keepDims:i}=n;return qh(a,s,i,"min",r)}var T0e={kernelName:wi,backendName:"webgpu",kernelFunc:C0e},N0e=qr({opSnippet:16,cpuKernelImpl:Ype}),E0e={kernelName:ki,backendName:"webgpu",kernelFunc:N0e},R0e=class{constructor(e,t,r){this.uniforms="",this.variableNames=["x"],this.workGroupSize=[64,1,1],this.size=!0,this.outputShape=t.map((n,a)=>n[0]+e[a]+n[1]),this.dispatchLayout=Ye(this.outputShape),this.dispatch=Oe(this.dispatchLayout,this.outputShape,this.workGroupSize),this.xShape=e,t.map((n,a)=>{this.uniforms+=` pad${a} : vec2<i32>,`}),this.offset=r==="reflect"?0:1,this.shaderKey=`mirrorPad_${r}`}getUserCode(){let e=this.xShape.length,t=this.xShape.map((l,u)=>`uniforms.pad${u}[0]`).join(","),r=this.xShape.map((l,u)=>`uniforms.pad${u}[0] + uniforms.xShape${e>1?`[${u}]`:""}`).join(","),n=e===1?"start":"start[i]",a=e===1?"end":"end[i]",s=e===1?"outC":"outC[i]",i=xr(e),o=e>1?["coords[0]","coords[1]","coords[2]","coords[3]"].slice(0,e):"coords";return`
|
|
${rt()}
|
|
if (index < uniforms.size) {
|
|
let start = ${i}(${t});
|
|
let end = ${i}(${r});
|
|
var outC = getCoordsFromIndex(index);
|
|
for (var i = 0; i < ${e}; i = i + 1) {
|
|
if (${s} < ${n}) {
|
|
${s} = ${n} * 2 - ${s} - ${this.offset};
|
|
} else if(${s} >= ${a}) {
|
|
${s} = (${a} - 1) * 2 - ${s} + ${this.offset};
|
|
}
|
|
}
|
|
let coords = outC - start;
|
|
setOutputAtIndex(index, getX(${o}));
|
|
}
|
|
}
|
|
`}},$0e={kernelName:Ii,backendName:"webgpu",kernelFunc:({inputs:e,attrs:t,backend:r})=>{let{x:n}=e,{paddings:a,mode:s}=t,i=r,o=a.map(u=>({type:"int32",data:[u[0],u[1]]})),l=new R0e(n.shape,a,s);return i.runWebGPUProgram(l,[n],n.dtype,o)}};function M0e(e){let{inputs:t,backend:r}=e,{x:n}=t;if(r.shouldExecuteOnCPU([n])){let s=r.tensorMap.get(n.dataId),[i,o]=Qpe(s.values,n.shape,n.dtype);return r.makeTensorInfo(o,n.dtype,i)}let a=new jh(n.shape,11);return r.runWebGPUProgram(a,[n],n.dtype)}var F0e={kernelName:ll,backendName:"webgpu",kernelFunc:M0e};function P0e(e){console.warn("tf.nonMaxSuppression() in webgpu locks the UI thread. Call tf.nonMaxSuppressionAsync() instead");let{inputs:t,backend:r,attrs:n}=e,{boxes:a,scores:s}=t,{maxOutputSize:i,iouThreshold:o,scoreThreshold:l}=n,u=r.readSync(a.dataId),d=r.readSync(s.dataId),{selectedIndices:h}=Kn.nonMaxSuppressionV3Impl(u,d,i,o,l);return r.makeTensorInfo([h.length],"int32",new Int32Array(h))}var _0e={kernelName:dl,backendName:"webgpu",kernelFunc:P0e};function z0e(e){console.warn("tf.nonMaxSuppression() in webgpu locks the UI thread. Call tf.nonMaxSuppressionAsync() instead");let{inputs:t,backend:r,attrs:n}=e,{boxes:a,scores:s}=t,{maxOutputSize:i,iouThreshold:o,scoreThreshold:l,softNmsSigma:u}=n,d=r.readSync(a.dataId),h=r.readSync(s.dataId),p=i,c=o,f=l,m=u,{selectedIndices:g,selectedScores:y}=Kn.nonMaxSuppressionV5Impl(d,h,p,c,f,m);return[r.makeTensorInfo([g.length],"int32",new Int32Array(g)),r.makeTensorInfo([y.length],"float32",new Float32Array(y))]}var O0e={kernelName:pl,backendName:"webgpu",kernelFunc:z0e};function H0(e){let{inputs:t,backend:r}=e,{x:n}=t;if(n.dtype==="complex64"){let a=Hh({inputs:{input:n},backend:r}),s=H0({inputs:{x:a},backend:r}),i=$m({inputs:{input:n},backend:r}),o=H0({inputs:{x:i},backend:r}),l=$d({inputs:{real:s,imag:o},backend:r});return r.disposeData(a.dataId),r.disposeData(s.dataId),r.disposeData(i.dataId),r.disposeData(o.dataId),l}else return Fd({attrs:{shape:n.shape,dtype:n.dtype,value:n.dtype==="string"?"":0},backend:r})}var D0e={kernelName:El,backendName:"webgpu",kernelFunc:H0};function sS(e){let{inputs:t,backend:r}=e,{x:n}=t;if(n.dtype==="string")throw new Error("onesLike is not supported under string dtype");if(n.dtype==="complex64"){let a=Hh({inputs:{input:n},backend:r}),s=sS({inputs:{x:a},backend:r}),i=$m({inputs:{input:n},backend:r}),o=H0({inputs:{x:i},backend:r}),l=$d({inputs:{real:s,imag:o},backend:r});return r.disposeData(a.dataId),r.disposeData(s.dataId),r.disposeData(i.dataId),r.disposeData(o.dataId),l}else return Fd({attrs:{shape:n.shape,dtype:n.dtype,value:1},backend:r})}var L0e={kernelName:hl,backendName:"webgpu",kernelFunc:sS};function B0e(e){let{inputs:t,backend:r,attrs:n}=e,{axis:a}=n;if(t.length===1)return dy({inputs:{input:t[0]},backend:r,attrs:{dim:a}});let s=t[0].shape,i=t[0].dtype;t.forEach(d=>{v.assertShapesMatch(s,d.shape,"All tensors passed to stack must have matching shapes"),v.assert(i===d.dtype,()=>"All tensors passed to stack must have matching dtypes")});let o=[],l=t.map(d=>{let h=dy({inputs:{input:d},backend:r,attrs:{dim:a}});return o.push(h),h}),u=Z9({inputs:l,backend:r,attrs:{axis:a}});return o.forEach(d=>r.disposeData(d.dataId)),u}var W0e={kernelName:fl,backendName:"webgpu",kernelFunc:B0e},V0e=class{constructor(e,t){this.variableNames=["x"],this.uniforms="constantValue : f32,",this.workGroupSize=[64,1,1],this.size=!0,this.outputShape=t.map((r,n)=>r[0]+e[n]+r[1]),this.dispatchLayout=Ye(this.outputShape),this.dispatch=Oe(this.dispatchLayout,this.outputShape,this.workGroupSize),t.map((r,n)=>{this.uniforms+=` pad${n} : vec2<i32>,`}),this.xShape=e,this.shaderKey="pad"}getUserCode(){let e=this.xShape.length,t=xr(e),r=this.xShape.map((u,d)=>`uniforms.pad${d}[0]`).join(","),n=this.xShape.map((u,d)=>`uniforms.pad${d}[0] + uniforms.xShape${e>1?`[${d}]`:""}`).join(","),a=e>1?`${t}(${r})`:`${r}`,s=e>1?`${t}(${n})`:`${n}`,i=e>1?"any(outC < start)":"outC < start",o=e>1?"any(outC >= end)":"outC >= end",l=e>1?["coords[0]","coords[1]","coords[2]","coords[3]"].slice(0,e):"coords";return`
|
|
${rt()}
|
|
if (index < uniforms.size) {
|
|
let start = ${a};
|
|
let end = ${s};
|
|
let outC = getCoordsFromIndex(index);
|
|
|
|
if (${i} || ${o}) {
|
|
setOutputAtIndex(index, uniforms.constantValue);
|
|
} else {
|
|
let coords = outC - start;
|
|
setOutputAtIndex(index, getX(${l}));
|
|
}
|
|
}
|
|
}
|
|
`}},iS=e=>{let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{paddings:s,constantValue:i}=n;if(s.every(u=>v.arraysEqual(u,[0,0])))return zn({inputs:{x:a},backend:r});if(v.sizeFromShape(a.shape)===0){let u=s.map((d,h)=>d[0]+a.shape[h]+d[1]);return Fd({backend:r,attrs:{shape:u,value:i,dtype:a.dtype}})}let o=[{type:"float32",data:[i]}];s.map(u=>o.push({type:"int32",data:[u[0],u[1]]}));let l=new V0e(a.shape,s);return r.runWebGPUProgram(l,[a],a.dtype,o)},U0e={kernelName:Ci,backendName:"webgpu",kernelFunc:iS},G0e=qr({opSnippet:13}),j0e={kernelName:Ti,backendName:"webgpu",kernelFunc:G0e};function H0e(e){let{inputs:t,backend:r}=e,{x:n,alpha:a}=t,s=new j9(14,n.shape,a.shape);return r.runWebGPUProgram(s,[n,a],"float32")}var q0e={kernelName:Ni,backendName:"webgpu",kernelFunc:H0e};function X0e(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{axis:s,keepDims:i}=n;return qh(a,s,i,"prod",r)}var K0e={kernelName:Ei,backendName:"webgpu",kernelFunc:X0e},Z0e=e=>{let{backend:t,attrs:r}=e,{start:n,stop:a,step:s,dtype:i}=r,o=rhe(n,a,s,i);return t.makeTensorInfo([o.length],i,o)},Y0e={kernelName:nd,backendName:"webgpu",kernelFunc:Z0e},oS=qr({opSnippet:3}),J0e={kernelName:li,backendName:"webgpu",kernelFunc:oS},Q0e=kr({opType:12}),efe={kernelName:Ri,backendName:"webgpu",kernelFunc:Q0e},tfe=kr({opType:13}),rfe={kernelName:Mi,backendName:"webgpu",kernelFunc:tfe},nfe=class{constructor(e,t,r){this.variableNames=["x"],this.uniforms="adjustHeightWidth : vec2<f32>, halfPixelCenters : f32,",this.workGroupSize=[64,1,1],this.size=!0,this.outputShape=[e[0],t,r,e[3]],this.dispatchLayout=Ye(this.outputShape),this.dispatch=Oe(this.dispatchLayout,this.outputShape,this.workGroupSize),this.shaderKey="resizeBilinear"}getUserCode(){return`
|
|
${rt()}
|
|
if (index < uniforms.size) {
|
|
let coords = getCoordsFromIndex(index);
|
|
let b = coords[0];
|
|
let d = coords[3];
|
|
let rc = coords.yz;
|
|
|
|
let effectiveInSize = vec2<f32>(
|
|
f32(uniforms.xShape.y) - uniforms.adjustHeightWidth[0],
|
|
f32(uniforms.xShape.z) - uniforms.adjustHeightWidth[1]);
|
|
|
|
let effectiveOutSize = vec2<f32>(
|
|
f32(uniforms.outShape.y) - uniforms.adjustHeightWidth[0],
|
|
f32(uniforms.outShape.z) - uniforms.adjustHeightWidth[1]);
|
|
|
|
let effectiveInputOverOutputRatioRC =
|
|
effectiveInSize / effectiveOutSize;
|
|
|
|
// Fractional source index
|
|
let sourceFracIndexRC =
|
|
(vec2<f32>(rc) + vec2<f32>(uniforms.halfPixelCenters)) *
|
|
effectiveInputOverOutputRatioRC - vec2<f32>(uniforms.halfPixelCenters);
|
|
|
|
// Compute the four integer indices.
|
|
let sourceFloorRC = vec2<i32>(sourceFracIndexRC);
|
|
let sourceCeilRC = vec2<i32>(
|
|
min(vec2<f32>(uniforms.xShape.yz) - vec2<f32>(1.0), ceil(sourceFracIndexRC)));
|
|
|
|
let topLeft = getX(b, sourceFloorRC.x, sourceFloorRC.y, d);
|
|
let bottomLeft = getX(b, sourceCeilRC.x, sourceFloorRC.y, d);
|
|
let topRight = getX(b, sourceFloorRC.x, sourceCeilRC.y, d);
|
|
let bottomRight = getX(b, sourceCeilRC.x, sourceCeilRC.y, d);
|
|
|
|
let fracRC = sourceFracIndexRC - vec2<f32>(sourceFloorRC);
|
|
|
|
let top = topLeft + (topRight - topLeft) * fracRC.y;
|
|
let bottom = bottomLeft + (bottomRight - bottomLeft) * fracRC.y;
|
|
let newValue = top + (bottom - top) * fracRC.x;
|
|
|
|
setOutputAtIndex(index, newValue);
|
|
}
|
|
}
|
|
`}};function afe(e){let{inputs:t,backend:r,attrs:n}=e,{images:a}=t,{alignCorners:s,size:i,halfPixelCenters:o}=n,[l,u]=i,d=s&&l>1?1:0,h=s&&u>1?1:0,p=[{type:"float32",data:[d,h]},{type:"float32",data:[o?.5:0]}],c=new nfe(a.shape,l,u);return r.runWebGPUProgram(c,[a],"float32",p)}var sfe={kernelName:$i,backendName:"webgpu",kernelFunc:afe},ife=class{constructor(e,t,r,n){this.variableNames=["x"],this.uniforms="adjustHeightWidth : vec2<f32>, roundBase : f32,",this.workGroupSize=[64,1,1],this.size=!0,this.outputShape=[e[0],t,r,e[3]],this.dispatchLayout=Ye(this.outputShape),this.dispatch=Oe(this.dispatchLayout,this.outputShape,this.workGroupSize),this.halfPixelCenters=n,this.shaderKey=`resizeNearest_${n}`}getUserCode(){let e;return this.halfPixelCenters?e="max((vec2<f32>(rc) + vec2<f32>(0.5)) * effectiveInputOverOutputRatioRC, vec2<f32>(0.0))":e="vec2<f32>(rc) * effectiveInputOverOutputRatioRC",`
|
|
${rt()}
|
|
if (index < uniforms.size) {
|
|
let coords = getCoordsFromIndex(index);
|
|
let b = coords[0];
|
|
let d = coords[3];
|
|
let rc = coords.yz;
|
|
|
|
let effectiveInSize = vec2<f32>(
|
|
f32(uniforms.xShape.y) - uniforms.adjustHeightWidth[0],
|
|
f32(uniforms.xShape.z) - uniforms.adjustHeightWidth[1]);
|
|
|
|
let effectiveOutSize = vec2<f32>(
|
|
f32(uniforms.outShape.y) - uniforms.adjustHeightWidth[0],
|
|
f32(uniforms.outShape.z) - uniforms.adjustHeightWidth[1]);
|
|
|
|
let effectiveInputOverOutputRatioRC =
|
|
effectiveInSize / effectiveOutSize;
|
|
|
|
// Fractional source index
|
|
let sourceFracIndexRC = ${e};
|
|
|
|
// Compute the coordinators of nearest neighbor point.
|
|
let inputShapeRC = vec2<f32>(f32(uniforms.xShape.y), f32(uniforms.xShape.z));
|
|
let sourceNearestRC = vec2<i32>(
|
|
min(inputShapeRC - 1.0, floor(sourceFracIndexRC + uniforms.roundBase)));
|
|
let newValue = getX(b, sourceNearestRC.x, sourceNearestRC.y, d);
|
|
|
|
setOutputAtIndex(index, newValue);
|
|
}
|
|
}
|
|
`}};function ofe(e){let{inputs:t,backend:r,attrs:n}=e,{images:a}=t,{alignCorners:s,halfPixelCenters:i,size:o}=n,[l,u]=o,d=s&&l>1?1:0,h=s&&u>1?1:0,p=[{type:"float32",data:[d,h]},{type:"float32",data:[s?.5:0]}],c=new ife(a.shape,l,u,i);return r.runWebGPUProgram(c,[a],a.dtype,p)}var lfe={kernelName:sd,backendName:"webgpu",kernelFunc:ofe},ufe=class{constructor(e,t){this.outputShape=[],this.variableNames=["x"],this.workGroupSize=[64,1,1],this.size=!0,this.outputShape=e,this.dispatchLayout=Ye(this.outputShape),this.dispatch=Oe(this.dispatchLayout,this.outputShape,this.workGroupSize),this.uniforms=`centerX : f32, centerY : f32, sinRadians : f32,
|
|
cosRadians : f32,`,this.shaderKey="rotate",this.outputShape=e,typeof t=="number"?(this.uniforms+=" fillValue : f32,",this.fillSnippet="var outputValue = uniforms.fillValue;",this.shaderKey+="_float"):(this.uniforms+=" fillValue : vec3<f32>,",this.fillSnippet="var outputValue = uniforms.fillValue[coords[3]];",this.shaderKey+="_vec3")}getUserCode(){return`
|
|
${rt()}
|
|
|
|
if (index < uniforms.size) {
|
|
let coords = getCoordsFromIndex(index);
|
|
let coordXFloat = (f32(coords[2]) - uniforms.centerX) *
|
|
uniforms.cosRadians - (f32(coords[1]) - uniforms.centerY) *
|
|
uniforms.sinRadians;
|
|
let coordYFloat = (f32(coords[2]) - uniforms.centerX) *
|
|
uniforms.sinRadians + (f32(coords[1]) - uniforms.centerY) *
|
|
uniforms.cosRadians;
|
|
let coordX = i32(round(coordXFloat + uniforms.centerX));
|
|
let coordY = i32(round(coordYFloat + uniforms.centerY));
|
|
${this.fillSnippet}
|
|
if(coordX >= 0 && coordX < uniforms.xShape[2] && coordY >= 0 &&
|
|
coordY < uniforms.xShape[1]) {
|
|
outputValue = getX(coords[0], coordY, coordX, coords[3]);
|
|
}
|
|
setOutputAtIndex(index, outputValue);
|
|
}
|
|
}
|
|
`}},dfe={kernelName:Rl,backendName:"webgpu",kernelFunc:({inputs:e,attrs:t,backend:r})=>{let{image:n}=e,{radians:a,fillValue:s,center:i}=t,o=r,l=new ufe(n.shape,s),[u,d]=C.getImageCenter(i,n.shape[1],n.shape[2]),h=[{type:"float32",data:[u]},{type:"float32",data:[d]},{type:"float32",data:[Math.sin(a)]},{type:"float32",data:[Math.cos(a)]}];return typeof s=="number"?h.push({type:"float32",data:[Number.parseFloat(s.toFixed(2))]}):h.push({type:"float32",data:s}),o.runWebGPUProgram(l,[n],n.dtype,h)}},pfe=kr({opType:15,cpuKernelImpl:nhe}),hfe={kernelName:Fi,backendName:"webgpu",kernelFunc:pfe},cfe=class{constructor(e,t,r,n,a,s,i){this.variableNames=["updates","indices"],this.workGroupSize=[64,1,1],this.atomic=!0,this.outputShape=s,this.type=i,this.dispatchLayout=Ye(e),this.dispatch=Oe(this.dispatchLayout,e,this.workGroupSize),this.sliceDimGreaterThanOne=t>1,this.shaderKey=`scatter_${r}_${n}_${this.sliceDimGreaterThanOne}_${i}`;let o=xr(a.length);this.uniforms=`sliceDim : i32, strides: ${o}, size: i32,`,this.updatesRank=n,this.indicesRank=r}getUserCode(){let e="";this.indicesRank===1?e="coords[0]":this.indicesRank===2&&(e="coords[0], j");let t=`getIndices(${e})`,r=this.sliceDimGreaterThanOne?"uniforms.strides[j]":"uniforms.strides",n="",a="",s="";this.updatesRank===1?(n="coords[0]",a="flattenedIndex",s=`
|
|
fn getUpdatesCoordsFromFlatIndex(index : i32) -> i32 {
|
|
return index;
|
|
}
|
|
`):this.updatesRank===2&&(n="coords[0], coords[1]",a="vec2<i32>(flattenedIndex, coords[1])",s=`
|
|
fn getUpdatesCoordsFromFlatIndex(index : i32) -> vec2<i32> {
|
|
let d0 = index / uniforms.updatesShape[1];
|
|
let d1 = index - d0 * uniforms.updatesShape[1];
|
|
return vec2<i32>(d0, d1);
|
|
}
|
|
`);let i=`getUpdates(${n})`,o=this.type==="int32"?"atomicAdd(&(result[flatIndex]), i32(updateValue));":`
|
|
var assumed = atomicLoad(&(result[flatIndex]));
|
|
var success = 0;
|
|
for (; success == 0;) {
|
|
let new = bitcast<f32>(assumed) + updateValue;
|
|
let newI32 = bitcast<i32>(new);
|
|
let resValue = atomicCompareExchangeWeak(&(result[flatIndex]), assumed, newI32);
|
|
assumed = resValue[0];
|
|
success = resValue[1];
|
|
}
|
|
`;return`
|
|
${s}
|
|
|
|
${rt()}
|
|
|
|
if (index < uniforms.size) {
|
|
let coords = getUpdatesCoordsFromFlatIndex(index);
|
|
var flattenedIndex = 0;
|
|
for (var j = 0; j < uniforms.sliceDim; j = j + 1) {
|
|
let indexInside = i32(round(${t}));
|
|
flattenedIndex = flattenedIndex + indexInside * ${r};
|
|
}
|
|
let updateValue = ${i};
|
|
let flatIndex = getOutputIndexFromCoords(${a});
|
|
|
|
${o}
|
|
}
|
|
}`}};function ffe(e){let{inputs:t,backend:r,attrs:n}=e,{indices:a,updates:s}=t,{shape:i}=n,{sliceRank:o,numUpdates:l,sliceSize:u,strides:d,outputSize:h}=C.calculateShapes(s,a,i),p=[h/u,u];if(h===0)return r.makeTensorInfo(i,a.dtype);let c=Qe({inputs:{x:a},backend:r,attrs:{shape:[l,o]}}),f=Qe({inputs:{x:s},backend:r,attrs:{shape:[l,u]}}),m=f.dtype,g=Fd({backend:r,attrs:{shape:p,value:0,dtype:m}}),y=v.sizeFromShape(f.shape),A=[{type:"int32",data:[o]},{type:"int32",data:d},{type:"int32",data:[y]}],x=new cfe(f.shape,o,c.shape.length,f.shape.length,d,p,m),b=r.runWebGPUProgram(x,[f,c],m,A,g),w=Qe({inputs:{x:b},backend:r,attrs:{shape:i}});return r.disposeData(c.dataId),r.disposeData(f.dataId),r.disposeData(b.dataId),w}var mfe={kernelName:Al,backendName:"webgpu",kernelFunc:ffe},gfe=class{constructor(e,t,r){this.variableNames=["c","a","b"],this.workGroupSize=[64,1,1],this.size=!0,this.outputShape=t,this.dispatchLayout=Ye(this.outputShape),this.dispatch=Oe(this.dispatchLayout,this.outputShape,this.workGroupSize),this.cRank=e,this.rank=r,this.shaderKey="select"}getUserCode(){let e,t;if(this.rank>4)throw Error(`Where for rank ${this.rank} is not yet supported`);if(this.rank===1)t="resRC",e="resRC";else{let r=["resRC.x","resRC.y","resRC.z","resRC.w"],n=[],a=[];for(let s=0;s<this.outputShape.length;s++)a.push(`${r[s]}`),s<this.cRank&&n.push(`${r[s]}`);e=n.join(),t=a.join()}return`
|
|
${rt()}
|
|
if (index < uniforms.size) {
|
|
let resRC = getCoordsFromIndex(index);
|
|
let cVal = getC(${e});
|
|
if (cVal >= 1.0) {
|
|
setOutputAtIndex(index, getA(${t}));
|
|
} else {
|
|
setOutputAtIndex(index, getB(${t}));
|
|
}
|
|
}
|
|
}
|
|
`}};function yfe(e){let{inputs:t,backend:r}=e,{condition:n,t:a,e:s}=t,i=new gfe(n.shape.length,a.shape,a.shape.length);return r.runWebGPUProgram(i,[n,a,s],Nr(a.dtype,s.dtype))}var Afe={kernelName:xl,backendName:"webgpu",kernelFunc:yfe},xfe=kr({opType:18}),bfe={kernelName:_i,backendName:"webgpu",kernelFunc:xfe},vfe=kr({opType:16}),wfe={kernelName:Pi,backendName:"webgpu",kernelFunc:vfe},kfe=kr({opType:17}),Ife={kernelName:vl,backendName:"webgpu",kernelFunc:kfe},lS=qr({opSnippet:2,cpuKernelImpl:lhe,supportsComplex:!0}),Sfe={kernelName:Bi,backendName:"webgpu",kernelFunc:lS};function Cfe(e){let{inputs:t,backend:r,attrs:n}=e,{logits:a}=t,{dim:s}=n,i=v.parseAxisParam([s],a.shape),o=aS({inputs:{x:a},backend:r,attrs:{reductionIndices:i,keepDims:!1}}),l=C.expandShapeToKeepDim(o.shape,i),u=Qe({inputs:{x:o},backend:r,attrs:{shape:l}}),d=lS({inputs:{a,b:u},backend:r}),h=rS({inputs:{x:d},backend:r}),p=$5({inputs:{x:h},backend:r,attrs:{axis:i,keepDims:!1}}),c=Qe({inputs:{x:p},backend:r,attrs:{shape:l}}),f=oS({inputs:{a:h,b:c},backend:r});return r.disposeData(o.dataId),r.disposeData(u.dataId),r.disposeData(d.dataId),r.disposeData(h.dataId),r.disposeData(p.dataId),r.disposeData(c.dataId),f}var Tfe={kernelName:Di,backendName:"webgpu",kernelFunc:Cfe},Nfe=e=>{let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{blockShape:s,paddings:i}=n;v.assert(a.shape.length<=4,()=>"spaceToBatchND for rank > 4 with a WebGPU backend not implemented yet");let o=s.reduce((y,A)=>y*A),l=[[0,0]];l.push(...i);for(let y=1+s.length;y<a.shape.length;++y)l.push([0,0]);let u=[],d=iS({inputs:{x:a},backend:r,attrs:{paddings:l,constantValue:0}}),h=C.getReshaped(d.shape,s,o,!1),p=C.getPermuted(h.length,s.length,!1),c=C.getReshapedPermuted(d.shape,s,o,!1),f=Qe({inputs:{x:d},backend:r,attrs:{shape:h}}),m=Ya({inputs:{x:f},backend:r,attrs:{perm:p}}),g=Qe({inputs:{x:m},backend:r,attrs:{shape:c}});return u.push(d),u.push(f),u.push(m),u.forEach(y=>r.disposeData(y.dataId)),g},Efe={kernelName:wl,backendName:"webgpu",kernelFunc:Nfe},Rfe=class{constructor(e,t,r,n,a,s,i=!0){this.variableNames=["updates","indices","defaultValue"],this.workGroupSize=[64,1,1],this.workPerThread=4,this.size=!0,this.outputShape=s,this.dispatchLayout=Ye(this.outputShape),this.dispatch=Oe(this.dispatchLayout,this.outputShape,this.workGroupSize,[this.workPerThread,1,1]);let o=t>1;this.shaderKey=`scatter_${r}_${n}_${o}`;let l=xr(a.length);this.uniforms=`updateSize : i32, sliceDim : i32, strides: ${l},`;let u="";r===1?u="i":r===2&&(u="i, j"),this.indicesSnippet=`getIndices(${u})`;let d="";n===1?d="i":n===2&&(d="i, coords[1]"),this.updatesSnippet=`getUpdates(${d})`,this.strideString=o?"uniforms.strides[j]":"uniforms.strides"}getUserCode(){return`
|
|
${rt()}
|
|
|
|
let globalIndex = index * ${this.workPerThread};
|
|
if (globalIndex < uniforms.size) {
|
|
var sum = vec4<f32>(0.0);
|
|
var found = vec4<bool>(false);
|
|
for (var i = 0; i < uniforms.updateSize; i = i + 1) {
|
|
var flattenedIndex = 0;
|
|
for (var j = 0; j < uniforms.sliceDim; j = j + 1) {
|
|
let indexInside = i32(round(${this.indicesSnippet}));
|
|
flattenedIndex = flattenedIndex + indexInside * ${this.strideString};
|
|
}
|
|
for (var innerIndex = 0; innerIndex < ${this.workPerThread}; innerIndex = innerIndex + 1) {
|
|
let curIndex = globalIndex + innerIndex;
|
|
let coords = getCoordsFromIndex(curIndex);
|
|
if (flattenedIndex == coords[0]) {
|
|
sum[innerIndex] = sum[innerIndex] + ${this.updatesSnippet};
|
|
found[innerIndex] = true;
|
|
}
|
|
}
|
|
}
|
|
for (var innerIndex = 0; innerIndex < ${this.workPerThread}; innerIndex = innerIndex + 1) {
|
|
let curIndex = globalIndex + innerIndex;
|
|
if (curIndex < uniforms.size)
|
|
{
|
|
setOutputAtIndex(curIndex, mix(getDefaultValue(), sum[innerIndex], f32(found[innerIndex])));
|
|
}
|
|
}
|
|
}
|
|
}`}};function $fe(e){let{inputs:t,backend:r,attrs:n}=e,{sparseIndices:a,sparseValues:s,defaultValue:i}=t,{outputShape:o}=n,{sliceRank:l,numUpdates:u,strides:d,outputSize:h}=C.calculateShapes(s,a,o),p=!1,c=[{type:"int32",data:[u]},{type:"int32",data:[l]},{type:"int32",data:d}],f=new Rfe(u,l,a.shape.length,s.shape.length,d,[h,1],p),m=r.runWebGPUProgram(f,[s,a,i],s.dtype,c),g=Qe({inputs:{x:m},backend:r,attrs:{shape:o}});return r.disposeData(m.dataId),g}var Mfe={kernelName:ph,backendName:"webgpu",kernelFunc:$fe};function Ffe(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{numOrSizeSplits:s,axis:i}=n,o=v.parseAxisParam(i,a.shape)[0],l=C.prepareSplitSize(a,s,o),u=a.shape.length,d=new Array(u).fill(0),h=a.shape.slice();return l.map(p=>{let c=[...h];c[o]=p;let f=Md({inputs:{x:a},backend:r,attrs:{begin:d,size:c}});return d[o]+=p,f})}var Pfe={kernelName:kl,backendName:"webgpu",kernelFunc:Ffe},_fe=kr({opType:19}),zfe={kernelName:zi,backendName:"webgpu",kernelFunc:_fe},Ofe={kernelName:dd,backendName:"webgpu",kernelFunc:({inputs:e,backend:t})=>{let{x:r}=e,n=t,a=new jh(r.shape,20);return n.runWebGPUProgram(a,[r],r.dtype)}},Dfe=qr({opSnippet:11}),Lfe={kernelName:Li,backendName:"webgpu",kernelFunc:Dfe},Bfe=class{constructor(e){this.variableNames=["x"],this.workPerThread=1,this.workGroupSize=[64,1,1],this.size=!0,this.outputShape=e,this.dispatchLayout=Ye(this.outputShape),this.dispatch=Oe(this.dispatchLayout,this.outputShape,this.workGroupSize,[this.workPerThread,1,1]);let t=xr(this.outputShape.length);this.uniforms=`begin : ${t}, strides : ${t}, `,this.shaderKey="stridedSlice"}getUserCode(){let e=this.outputShape.length,t="";if(e===1)t="coords * uniforms.strides + uniforms.begin";else{let r=0;t=this.outputShape.map((n,a)=>(r++,this.outputShape.length===1?`coords * uniforms.strides[${a}] + uniforms.begin[${a}]`:`coords[${r-1}] * uniforms.strides[${a}] + uniforms.begin[${a}]`)).join(",")}return`
|
|
${rt()}
|
|
if (index < uniforms.size) {
|
|
let coords = getCoordsFromIndex(index);
|
|
setOutputAtIndex(index, getX(${t}));
|
|
}
|
|
}
|
|
`}};function Wfe(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{begin:s,end:i,strides:o,beginMask:l,endMask:u,ellipsisMask:d,newAxisMask:h,shrinkAxisMask:p}=n,{finalShapeSparse:c,finalShape:f,isIdentity:m,sliceDim0:g,isSimpleSlice:y,begin:A,end:x,strides:b}=Ot.sliceInfo(a.shape,s,i,o,l,u,d,h,p),w;if(m)w=Qe({inputs:{x:a},backend:r,attrs:{shape:f}});else if(g||y){v.assert(a.shape.length>=1,()=>`Input must have rank at least 1, got: ${a.shape.length}`);let S=Ot.computeOutShape(A,x,b),N=Md({inputs:{x:a},backend:r,attrs:{begin:A,size:S}});w=Qe({inputs:{x:N},backend:r,attrs:{shape:f}}),r.disposeData(N.dataId)}else if(r.shouldExecuteOnCPU([a])){let S=r.readSync(a.dataId),N=We(a.shape,a.dtype,S),E=ihe(c,N,b,A);w=r.makeTensorInfo(f,a.dtype,E.values)}else{let S=new Bfe(c),N=[{type:"int32",data:A},{type:"int32",data:b}],E=r.runWebGPUProgram(S,[a],a.dtype,N);w=Qe({inputs:{x:E},backend:r,attrs:{shape:f}}),r.disposeData(E.dataId)}return w}var Vfe={kernelName:Il,backendName:"webgpu",kernelFunc:Wfe};function Ufe(e){let{inputs:t,backend:r,attrs:n}=e,{separator:a,nGramWidths:s,leftPad:i,rightPad:o,padWidth:l,preserveShortSequences:u}=n,{data:d,dataSplits:h}=t,p=r.readSync(d.dataId),c=r.readSync(h.dataId),[f,m]=ohe(p,c,a,s,i,o,l,u);return[r.makeTensorInfo([f.length],"string",f),r.makeTensorInfo(h.shape,"int32",m)]}var Gfe={kernelName:hh,backendName:"webgpu",kernelFunc:Ufe},jfe=kr({opType:21}),Hfe={kernelName:Wi,backendName:"webgpu",kernelFunc:jfe},qfe=class{constructor(e,t){this.variableNames=["A"],this.workGroupSize=[64,1,1],this.size=!0;let r=new Array(e.length);for(let n=0;n<r.length;n++)r[n]=e[n]*t[n];this.outputShape=r,this.dispatchLayout=Ye(this.outputShape),this.dispatch=Oe(this.dispatchLayout,this.outputShape,this.workGroupSize),this.rank=this.outputShape.length,this.shaderKey="tile"}getUserCode(){let e=Xfe(this.rank,"uniforms.");return`
|
|
${rt()}
|
|
if (index < uniforms.size) {
|
|
let resRC = getCoordsFromIndex(index);
|
|
setOutputAtIndex(index, getA(${e}));
|
|
}
|
|
}
|
|
`}};function Xfe(e,t=""){if(e>=5)throw Error(`Tile for rank ${e} is not yet supported`);if(e===1)return`(resRC % ${t}aShape)`;let r=["resRC.x","resRC.y","resRC.z","resRC.w"],n=[];for(let a=0;a<e;a++)n.push(`(${r[a]} % ${t}aShape[${a}])`);return n.join()}function Kfe(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{reps:s}=n;if(r.shouldExecuteOnCPU([a])||a.dtype==="string"||a.shape.length>=5){let o=r.readSync(a.dataId),l=a.dtype==="string"?o.map(h=>v.decodeString(h)):o,u=We(a.shape,a.dtype,l),d=uhe(u,s);return r.makeTensorInfo(d.shape,d.dtype,d.values)}let i=new qfe(a.shape,s);return r.runWebGPUProgram(i,[a],a.dtype)}var Zfe={kernelName:es,backendName:"webgpu",kernelFunc:Kfe},Yfe=class{constructor(e){this.variableNames=["x","indices"],this.workGroupSize=[256,1,1],this.size=!0,this.outputShape=e,this.dispatchLayout=Ye(this.outputShape),this.dispatch=Oe(this.dispatchLayout,this.outputShape,this.workGroupSize),this.uniforms=`inputSize : i32, firstPass : i32, negativeInf : f32,
|
|
dir : i32, inc : i32,`,this.shaderKey="swap"}getUserCode(){return`
|
|
${rt()}
|
|
if (index < uniforms.size) {
|
|
let outC = getCoordsFromIndex(index);
|
|
let batch = outC[0];
|
|
let elemIdx = outC[1];
|
|
// We compare elements pair-wise within a group of size 2 * inc.
|
|
// The comparing rule for each group alternates between ascending
|
|
// and descending. Within each group, we compare each pair at
|
|
// positions i and i+inc. To decide whether an element at position i
|
|
// is x0 or x1, we mod it by 2 * inc, if the result is smaller than
|
|
// inc, it is in the first half of the group, we denote it as x0,
|
|
// otherwise we denote it as x1.
|
|
// For example, as shown in the Bitonic top K paper referenced
|
|
// above, Figure5(a) shows that element[1] is in the second half of
|
|
// the group when group size is 2, but it is in the first half of
|
|
// the group when group size is 4.
|
|
let isFirstInPair = elemIdx % (2 * uniforms.inc) < uniforms.inc;
|
|
var i = 0;
|
|
if (isFirstInPair) {
|
|
i = elemIdx;
|
|
} else {
|
|
i = elemIdx - uniforms.inc;
|
|
}
|
|
|
|
var i0 = 0;
|
|
if (uniforms.firstPass == 1) {
|
|
i0 = i;
|
|
} else {
|
|
i0 = i32(getIndices(batch, i));
|
|
}
|
|
|
|
var i1 = 0;
|
|
if (uniforms.firstPass == 1) {
|
|
i1 = i + uniforms.inc;
|
|
} else {
|
|
i1 = i32(getIndices(batch, i + uniforms.inc));
|
|
}
|
|
|
|
var x0 = f32(0.0);
|
|
var x1 = f32(0.0);
|
|
if (i0 < uniforms.inputSize) {
|
|
x0 = getX(batch, i0);
|
|
} else {
|
|
x0 = uniforms.negativeInf;
|
|
}
|
|
if (i1 < uniforms.inputSize) {
|
|
x1 = getX(batch, i1);
|
|
} else {
|
|
x1 = uniforms.negativeInf;
|
|
}
|
|
|
|
let reverse = elemIdx % (2 * uniforms.dir) >= uniforms.dir;
|
|
let isGreater = x0 > x1 || (x0 == x1 && i1 > i0);
|
|
if (reverse == isGreater) {
|
|
// Elements in opposite order of direction
|
|
let iTemp = i0;
|
|
i0 = i1;
|
|
i1 = iTemp;
|
|
}
|
|
if (isFirstInPair) {
|
|
setOutputAtIndex(index, f32(i0));
|
|
} else {
|
|
setOutputAtIndex(index, f32(i1));
|
|
}
|
|
}
|
|
}
|
|
`}},Jfe=class{constructor(e){this.variableNames=["x","indices"],this.workGroupSize=[256,1,1],this.size=!0,this.outputShape=e,this.dispatchLayout=Ye(this.outputShape),this.dispatch=Oe(this.dispatchLayout,this.outputShape,this.workGroupSize),this.uniforms="inputSize : i32, firstPass : i32, k : i32,",this.shaderKey="merge"}getUserCode(){return`
|
|
${rt()}
|
|
if (index < uniforms.size) {
|
|
let outC = getCoordsFromIndex(index);
|
|
let batch = outC[0];
|
|
let elemIdx = outC[1];
|
|
// The output size is half of the previous size.
|
|
// If the previous sequence is | | | | _ _ _ _ | | | | _ _ _ _
|
|
// (k=4), we only need to output the indices at positions |, the
|
|
// indices at positions _ can be thrown away, see Figure5(b) After
|
|
// Phase 2 (Merge phase) in the Bitonic Top K paper referenced
|
|
// above.
|
|
// For example, the paper shows we only need to output the orange
|
|
// bars. The output sequence should look like this | | | | | | | |.
|
|
// Because the sequence is halved, to map the output index back to
|
|
// the previous sequence to find the corresponding value, we need
|
|
// to double the index. When we double the index, we basically
|
|
// interpolate a position, so 2i looks like
|
|
// | _ | _ | _ | _ | _ | _ | _. We move the | to the first k
|
|
// position of each 2k positions by - elemIdx % k. E.g. for output
|
|
// at index 4,5,6,7, we want to get the corresponding element at
|
|
// original index 8,9,10,11, for output at index 8,9,10,11,
|
|
// we want to get the corresponding element at original index
|
|
// 16,17,18,19, so on and so forth.
|
|
|
|
var i = 0;
|
|
if (elemIdx < uniforms.k) {
|
|
i = elemIdx;
|
|
} else {
|
|
i = elemIdx * 2 - elemIdx % uniforms.k;
|
|
}
|
|
var i0 = 0;
|
|
if (uniforms.firstPass == 1) {
|
|
i0 = i;
|
|
} else {
|
|
i0 = i32(getIndices(batch, i));
|
|
}
|
|
var i1 = 0;
|
|
if (uniforms.firstPass == 1) {
|
|
i1 = i + uniforms.k;
|
|
} else {
|
|
i1 = i32(getIndices(batch, i + uniforms.k));
|
|
}
|
|
|
|
let x0 = getX(batch, i0);
|
|
var x1 = f32(0.0);
|
|
if (i1 < uniforms.inputSize) {
|
|
x1 = getX(batch, i1);
|
|
} else {
|
|
x1 = x0;
|
|
}
|
|
|
|
if (x0 >= x1) {
|
|
setOutputAtIndex(index, f32(i0));
|
|
} else {
|
|
setOutputAtIndex(index, f32(i1));
|
|
}
|
|
}
|
|
}
|
|
`}};function cu(e,t){t!==null&&e.disposeData(t.dataId)}function mv(e){let t=1;for(;t<e;)t*=2;return t}function Qfe(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{k:s,sorted:i}=n,o=a.shape,l=o[o.length-1];if(r.shouldExecuteOnCPU([a])){let b=r.readSync(a.dataId),[w,S]=dhe(b,o,a.dtype,s,i);return[r.makeTensorInfo(w.shape,w.dtype,w.values),r.makeTensorInfo(S.shape,S.dtype,S.values)]}if(s===0)return o[o.length-1]=0,[r.makeTensorInfo(o,a.dtype,[]),r.makeTensorInfo(o,"int32",[])];if(l===1)return[a,Fd({attrs:{shape:o,dtype:"int32",value:0},backend:r})];let u=v.sizeFromShape(o)/l,d=Qe({inputs:{x:a},attrs:{shape:[u,l]},backend:r}),h=mv(s),p=mv(l),c=null,f=()=>c===null?[d,d]:[d,c],m=(b,w,S)=>{let N=f(),E=new Yfe(S),R=[{type:"int32",data:[l]},{type:"int32",data:[c===null?1:0]},{type:"float32",data:[Number.NEGATIVE_INFINITY]},{type:"int32",data:[b]},{type:"int32",data:[w]}],z=c;c=r.runWebGPUProgram(E,N,"int32",R),cu(r,z)};for(let b=1;b<h;b*=2){let w=b*2;for(let S=b;S>=1;S/=2)m(w,S,[u,p])}for(let b=p;b>h;b/=2){let w=f(),S=new Jfe([u,b/2]),N=[{type:"int32",data:[l]},{type:"int32",data:[c===null?1:0]},{type:"int32",data:[h]}],E=c;c=r.runWebGPUProgram(S,w,"int32",N),cu(r,E);let R=h/2,z=R*2;for(let M=R;M>=1;M/=2)m(z,M,c.shape)}let g=c;c=Md({inputs:{x:c},backend:r,attrs:{begin:0,size:[u,s]}}),cu(r,g);let y=nS({inputs:{x:d,indices:c},backend:r,attrs:{axis:1,batchDims:1}});cu(r,d);let A=o.slice(0,-1);A.push(s),g=c,c=Qe({inputs:{x:c},attrs:{shape:A},backend:r}),cu(r,g);let x=y;return y=Qe({inputs:{x:y},attrs:{shape:A},backend:r}),cu(r,x),[y,c]}var eme={kernelName:Cl,backendName:"webgpu",kernelFunc:Qfe},tme=class{constructor(e){this.variableNames=["Image","Transforms"],this.uniforms="interpolationModeId : i32, fillModeId : i32, fillValue : f32,",this.workGroupSize=[64,1,1],this.size=!0,this.outputShape=e,this.dispatchLayout=Ye(this.outputShape),this.dispatch=Oe(this.dispatchLayout,this.outputShape,this.workGroupSize),this.shaderKey="transform"}getUserCode(){return`
|
|
fn mapCoord(outCoord : f32, len : f32) -> f32{
|
|
var inCoord = outCoord;
|
|
if(uniforms.fillModeId == 2) {
|
|
if (inCoord < 0.0) {
|
|
if (len <= 1.0) {
|
|
inCoord = 0.0;
|
|
} else {
|
|
let sz2 = 2.0 * len;
|
|
if (inCoord < sz2) {
|
|
inCoord = sz2 * f32(i32(f32(-inCoord / sz2))) +
|
|
inCoord;
|
|
}
|
|
if (inCoord < -len) {
|
|
inCoord = inCoord + sz2;
|
|
} else {
|
|
inCoord = -inCoord - 1.0;
|
|
}
|
|
}
|
|
} else if (inCoord > len - 1.0) {
|
|
if (len <= 1.0) {
|
|
inCoord = 0.0;
|
|
} else {
|
|
let sz2 = 2.0 * len;
|
|
inCoord = inCoord - sz2 * f32(i32(f32(inCoord / sz2)));
|
|
if (inCoord >= len) {
|
|
inCoord = sz2 - inCoord - 1.0;
|
|
}
|
|
}
|
|
}
|
|
return clamp(inCoord, 0.0, len - 1.0);
|
|
} else if (uniforms.fillModeId == 3) {
|
|
if (inCoord < 0.0) {
|
|
if (len <= 1.0) {
|
|
inCoord = 0.0;
|
|
} else {
|
|
let sz = len - 1.0;
|
|
inCoord = inCoord + len * (f32(i32(f32(-inCoord / sz))) + 1.0);
|
|
}
|
|
} else if (inCoord > len - 1.0) {
|
|
if (len <= 1.0) {
|
|
inCoord = 0.0;
|
|
} else {
|
|
let sz = len - 1.0;
|
|
inCoord = inCoord - len * f32(i32(f32(inCoord / sz)));
|
|
}
|
|
}
|
|
return clamp(inCoord, 0.0, len - 1.0);
|
|
} else if (uniforms.fillModeId == 4) {
|
|
return clamp(outCoord, 0.0, len - 1.0);
|
|
}
|
|
return outCoord;
|
|
}
|
|
fn readWithFillValue(batch : i32, coordY : i32, coordX : i32,
|
|
channel : i32) -> f32 {
|
|
var outputValue : f32;
|
|
if (0 <= coordY && coordY < uniforms.imageShape[1] && 0 <= coordX && coordX < uniforms.imageShape[2]) {
|
|
outputValue = getImage(batch, coordY, coordX, channel);
|
|
} else {
|
|
outputValue = uniforms.fillValue;
|
|
}
|
|
return outputValue;
|
|
}
|
|
|
|
${rt()}
|
|
if (index < uniforms.size) {
|
|
let coords = getCoordsFromIndex(index);
|
|
var outputValue : f32;
|
|
let batch = coords[0];
|
|
let x = coords[2];
|
|
let y = coords[1];
|
|
let channel = coords[3];
|
|
let xf = f32(x);
|
|
let yf = f32(y);
|
|
let a1 = getTransforms(batch, 0);
|
|
let a2 = getTransforms(batch, 1);
|
|
let a3 = getTransforms(batch, 2);
|
|
let b1 = getTransforms(batch, 3);
|
|
let b2 = getTransforms(batch, 4);
|
|
let b3 = getTransforms(batch, 5);
|
|
let c1 = getTransforms(batch, 6);
|
|
let c2 = getTransforms(batch, 7);
|
|
let projection = c1 * xf + c2 * yf + 1.0;
|
|
if (projection == 0.0) {
|
|
outputValue = uniforms.fillValue;
|
|
} else {
|
|
let inX = (a1 * xf + a2 * yf + a3) / projection;
|
|
let inY = (b1 * xf + b2 * yf + b3) / projection;
|
|
let mapX = mapCoord(inX, f32(uniforms.imageShape[2]));
|
|
let mapY = mapCoord(inY, f32(uniforms.imageShape[1]));
|
|
|
|
if (uniforms.interpolationModeId == 1) {
|
|
let coordY = i32(round(mapY));
|
|
let coordX = i32(round(mapX));
|
|
outputValue = readWithFillValue(batch, coordY, coordX,
|
|
channel);
|
|
} else {
|
|
let yFloor = floor(mapY);
|
|
let xFloor = floor(mapX);
|
|
let yCeil = yFloor + 1.0;
|
|
let xCeil = xFloor + 1.0;
|
|
let valueYFloor = (xCeil - mapX) *
|
|
readWithFillValue(batch, i32(yFloor), i32(xFloor), channel) +
|
|
(mapX - xFloor) *
|
|
readWithFillValue(batch, i32(yFloor), i32(xCeil), channel);
|
|
let valueYCeil = (xCeil - mapX) *
|
|
readWithFillValue(batch, i32(yCeil), i32(xFloor), channel) +
|
|
(mapX - xFloor) *
|
|
readWithFillValue(batch, i32(yCeil), i32(xCeil), channel);
|
|
outputValue = (yCeil - mapY) * valueYFloor +
|
|
(mapY - yFloor) * valueYCeil;
|
|
}
|
|
}
|
|
setOutputAtIndex(index, outputValue);
|
|
}
|
|
}
|
|
`}};function rme(e){let{inputs:t,backend:r,attrs:n}=e,{image:a,transforms:s}=t,{interpolation:i,fillMode:o,fillValue:l,outputShape:u}=n,[d,h,p,c]=a.shape,[f,m]=u!=null?u:[h,p],g=[d,f,m,c],y=new tme(g),A=i==="nearest"?1:2,x;switch(o){case"constant":x=1;break;case"reflect":x=2;break;case"wrap":x=3;break;case"nearest":x=4;break;default:x=1;break}let b=[{type:"int32",data:[A]},{type:"int32",data:[x]},{type:"float32",data:[l]}];return r.runWebGPUProgram(y,[a,s],"float32",b)}var nme={kernelName:Tl,backendName:"webgpu",kernelFunc:rme};function ame(e){let{inputs:t,backend:r,attrs:n}=e,{value:a}=t,{axis:s}=n;s<0&&(s+=a.shape.length);let i=a,o=i.shape.length,l=a.shape[s],u=new Array(o-1),d=0;for(let m=0;m<o;m++)m!==s&&(u[d++]=i.shape[m]);let h=[],p=new Array(o).fill(0),c=i.shape.slice();c[s]=1;let f=new Array(l);for(let m=0;m<f.length;m++){p[s]=m;let g=Md({inputs:{x:i},backend:r,attrs:{begin:p,size:c}}),y=Qe({inputs:{x:g},backend:r,attrs:{shape:u}});f[m]=y,h.push(g)}return h.forEach(m=>r.disposeData(m.dataId)),f}var sme={kernelName:Nl,backendName:"webgpu",kernelFunc:ame},ime=[Rpe,che,mhe,Ahe,Ihe,Che,Nhe,Rhe,_he,Lhe,Whe,jhe,Ppe,Khe,ece,ace,ice,lce,pce,cce,mce,Ace,bce,Sce,Tce,Ece,Rce,$ce,Fce,_ce,Oce,Uce,Lce,Wce,Hce,Xce,Zce,Qce,r0e,a0e,i0e,Fpe,qhe,l0e,d0e,h0e,f0e,g0e,A0e,x0e,v0e,k0e,S0e,T0e,E0e,$0e,vce,F0e,_0e,O0e,zhe,L0e,W0e,U0e,j0e,q0e,K0e,Y0e,Ohe,J0e,efe,rfe,Npe,sfe,lfe,dfe,hfe,mfe,Afe,bfe,wfe,Ife,Fhe,Vfe,Gfe,Tfe,Efe,Mfe,Pfe,zfe,Ofe,Lfe,Sfe,kce,Hfe,Zfe,eme,nme,whe,sme,D0e];for(let e of ime)qn(e);var ome=class{constructor(e){this.device=e,this.numUsedBuffers=0,this.numFreeBuffers=0,this.freeBuffers=new Map,this.usedBuffers=new Map,this.numBytesUsed=0,this.numBytesAllocated=0}acquireUploadBuffer(e,t){return this.acquireBuffer(e,t,!0)}acquireBuffer(e,t,r=!1){let n=gv(e,t);if(this.freeBuffers.has(n)||this.freeBuffers.set(n,[]),this.usedBuffers.has(n)||this.usedBuffers.set(n,[]),this.numBytesUsed+=e,this.numUsedBuffers++,this.freeBuffers.get(n).length>0){this.numFreeBuffers--;let s=this.freeBuffers.get(n).shift();return this.usedBuffers.get(n).push(s),s}this.numBytesAllocated+=e;let a=this.device.createBuffer({mappedAtCreation:r,size:e,usage:t});return this.usedBuffers.get(n).push(a),a}releaseBuffer(e,t,r){if(this.freeBuffers.size===0)return;let n=gv(t,r);this.freeBuffers.has(n)||this.freeBuffers.set(n,[]),this.freeBuffers.get(n).push(e),this.numFreeBuffers++,this.numUsedBuffers--;let a=this.usedBuffers.get(n),s=a.indexOf(e);if(s<0)throw new Error("Cannot release a buffer that was never provided by this buffer manager");a.splice(s,1),this.numBytesUsed-=t}releaseUploadBuffer(e,t,r){e.mapAsync(GPUMapMode.WRITE).then(()=>{this.releaseBuffer(e,t,r)},n=>{})}getNumUsedBuffers(){return this.numUsedBuffers}getNumFreeBuffers(){return this.numFreeBuffers}dispose(){this.freeBuffers.forEach((e,t)=>{e.forEach(r=>{r.destroy()})}),this.usedBuffers.forEach((e,t)=>{e.forEach(r=>{r.destroy()})}),this.freeBuffers=new Map,this.usedBuffers=new Map,this.numUsedBuffers=0,this.numFreeBuffers=0,this.numBytesUsed=0,this.numBytesAllocated=0}};function gv(e,t){return`${e}_${t}`}var lme=class{constructor(e){this.device=e,this.numUsedTextures=0,this.numFreeTextures=0,this.freeTextures=new Map,this.usedTextures=new Map,this.numBytesUsed=0,this.numBytesAllocated=0}acquireTexture(e,t,r,n){let a=Av(r),s=e*t*a,i=yv(e,t,r,n);if(this.freeTextures.has(i)||this.freeTextures.set(i,[]),this.usedTextures.has(i)||this.usedTextures.set(i,[]),this.numBytesUsed+=s,this.numUsedTextures++,this.freeTextures.get(i).length>0){this.numFreeTextures--;let l=this.freeTextures.get(i).shift();return this.usedTextures.get(i).push(l),l}this.numBytesAllocated+=s;let o=this.device.createTexture({size:[e,t],format:r,usage:n});return this.usedTextures.get(i).push(o),o}releaseTexture(e,t,r,n,a){if(this.freeTextures.size===0)return;let s=yv(t,r,n,a);this.freeTextures.has(s)||this.freeTextures.set(s,[]),this.freeTextures.get(s).push(e),this.numFreeTextures++,this.numUsedTextures--;let i=this.usedTextures.get(s),o=i.indexOf(e);if(o<0)throw new Error("Cannot release a texture that was never provided by this texture manager");i.splice(o,1);let l=Av(n),u=t*r*l;this.numBytesUsed-=u}getNumUsedTextures(){return this.numUsedTextures}getNumFreeTextures(){return this.numFreeTextures}dispose(){this.freeTextures.forEach((e,t)=>{e.forEach(r=>{r.destroy()})}),this.usedTextures.forEach((e,t)=>{e.forEach(r=>{r.destroy()})}),this.freeTextures=new Map,this.usedTextures=new Map,this.numUsedTextures=0,this.numFreeTextures=0,this.numBytesUsed=0,this.numBytesAllocated=0}};function yv(e,t,r,n){return`${e}_${t}_${r}_${n}`}function Av(e){if(e==="rgba8unorm")return 16;throw new Error(`${e} is not supported!`)}var ume=(e,t,r,n,a)=>{let s=[n,...r];return a&&s.push(a),e.createBindGroup({layout:t,entries:s.map((i,o)=>({binding:o,resource:i}))})},xv=(e,t,r,n,a,s=!1)=>{let i={dtype:a.dtype,shape:a.shape},o=fpe(n,i,t,s),l=e.createShaderModule({code:o,label:t.constructor.name});return e.createComputePipeline({layout:r,compute:{module:l,entryPoint:"main"},label:t.constructor.name})};function bv(e,t,r=[],n="",a=""){return e.shaderKey+"_"+(e.workGroupSize?e.workGroupSize.join(","):"")+t.map(s=>s.length).join(",")+r.join(",")+e.variableNames.join(",")+n+a}var dme=Z().getNumber("WEBGPU_CPU_HANDOFF_SIZE_THRESHOLD"),vv=(e,t)=>{let r=e.limits.maxComputeWorkgroupsPerDimension,n=t.dispatchLayout,a=t.dispatch;if(a.every(i=>i<=r))return a;v.assert(a[0]>r&&n.y===void 0&&n.z===void 0,()=>"Dispatch size exceeds WebGPU limits in Y or Z dimension.");let s=Math.ceil(Math.sqrt(a[0]));return s>r?(s=Math.ceil(Math.cbrt(a[0])),v.assert(s<=r,()=>"Total dispatch size exceeds WebGPU maximum."),[s,s,s]):[s,s,1]},uS=class extends zu{constructor(e,t=!1){if(super(),this.commandQueueOwnedIds=new WeakSet,this.tensorDisposalQueue=[],this.uniformDisposalQueue=[],this.stagingDisposalQueue=[],this.textureDisposalQueue=[],this.disposed=!1,this.uploadWaitMs=0,this.downloadWaitMs=0,this.dispatchNumberInEncoder=0,this.fromPixelTextureLayout=null,this.fromPixelImportTextureLayout=null,!N5())throw new Error("WebGPU is not supported on this device");this.layoutCache={},this.pipelineCache={},this.device=e,this.queue=e.queue,this.currentCommandEncoder=null,this.currentComputePass=null,this.supportTimeQuery=t,this.bufferManager=new ome(this.device),this.textureManager=new lme(this.device),this.tensorMap=new Zp(this,Xt()),this.supportTimeQuery&&(this.querySet=this.device.createQuerySet({type:"timestamp",count:2})),Z().getBool("WEBGPU_USE_PROFILE_TOOL")&&(this.dummyCanvas=document.createElement("canvas"),this.dummyCanvas.width=1,this.dummyCanvas.height=1,this.dummyContext=this.dummyCanvas.getContext("webgpu"),this.dummyContext.configure({device:e,format:"bgra8unorm"}),document.body.appendChild(this.dummyCanvas))}nextDataId(){return uS.nextDataId++}floatPrecision(){return 32}defaultGpuBufferUsage(){return GPUBufferUsage.STORAGE|GPUBufferUsage.COPY_SRC|GPUBufferUsage.COPY_DST}flushDisposalQueue(){this.tensorDisposalQueue.forEach(e=>{this.maybeReleaseBuffer(e),this.tensorMap.delete(e)}),this.uniformDisposalQueue.forEach(e=>this.bufferManager.releaseBuffer(e.buffer,e.byteSize,e.usage)),this.stagingDisposalQueue.forEach(e=>this.bufferManager.releaseUploadBuffer(e.buffer,e.byteSize,e.usage)),this.textureDisposalQueue.forEach(e=>this.textureManager.releaseTexture(e.texture,e.width,e.height,e.format,e.usage)),this.tensorDisposalQueue=[],this.uniformDisposalQueue=[],this.stagingDisposalQueue=[],this.textureDisposalQueue=[]}disposeData(e,t=!1){if(this.tensorMap.has(e)){let r=this.tensorMap.get(e);if(r.refCount--,!t&&r.refCount>0)return!1;if(this.commandQueueOwnedIds.has(e))return this.tensorDisposalQueue.push(e),!1;this.maybeReleaseBuffer(e);let{complexTensorInfos:n}=this.tensorMap.get(e);n!=null&&(this.disposeData(n.real.dataId,!0),this.disposeData(n.imag.dataId,!0)),this.tensorMap.delete(e)}return!0}memory(){return{numBytesInGPU:this.bufferManager.numBytesUsed,numBytesAllocatedInGPU:this.bufferManager.numBytesAllocated,unreliable:!1}}getBufferManager(){return this.bufferManager}getTextureManager(){return this.textureManager}acquireBuffer(e,t=this.defaultGpuBufferUsage()){return this.bufferManager.acquireBuffer(e,t)}maybeReleaseBuffer(e){let t=this.tensorMap.get(e);t!=null&&t.bufferInfo.buffer!=null&&(this.bufferManager.releaseBuffer(t.bufferInfo.buffer,t.bufferInfo.byteSize,t.bufferInfo.usage),t.bufferInfo.buffer=null)}refCount(e){return this.tensorMap.has(e)?this.tensorMap.get(e).refCount:0}incRef(e){let t=this.tensorMap.get(e);t.refCount++}decRef(e){if(this.tensorMap.has(e)){let t=this.tensorMap.get(e);t.refCount--}}write(e,t,r){if(r==="complex64"&&e!=null)throw new Error("Cannot write to a complex64 dtype. Please use tf.complex(real, imag).");let n={id:this.nextDataId()},a=v.sizeFromShape(t)*m0(r);return this.tensorMap.set(n,{dtype:r,shape:t,values:e,bufferInfo:{byteSize:a,usage:this.defaultGpuBufferUsage()},refCount:1}),n}move(e,t,r,n,a){if(n==="complex64")throw new Error("Cannot write to a complex64 dtype. Please use tf.complex(real, imag).");let s=v.sizeFromShape(r)*m0(n);this.tensorMap.set(e,{dtype:n,shape:r,values:t,bufferInfo:{byteSize:s,usage:this.defaultGpuBufferUsage()},refCount:a})}submitQueue(){this.ensureComputePassEnded(),this.queue.submit([this.currentCommandEncoder.finish()]),this.currentCommandEncoder=null,this.dispatchNumberInEncoder=0,this.commandQueueOwnedIds=new WeakSet,this.flushDisposalQueue()}getBuffer(e){return this.uploadToGPU(e),this.tensorMap.get(e).bufferInfo.buffer}ensureCommandEncoderReady(){this.currentCommandEncoder||(this.currentCommandEncoder=this.device.createCommandEncoder())}ensureComputePassEnded(){this.currentComputePass&&(this.currentComputePass.end(),this.currentComputePass=null)}getComputePass(){return this.currentComputePass||(this.currentComputePass=this.currentCommandEncoder.beginComputePass()),this.currentComputePass}async getBufferData(e,t){let r=this.acquireBuffer(t,GPUBufferUsage.COPY_DST|GPUBufferUsage.MAP_READ);this.ensureCommandEncoderReady(),this.ensureComputePassEnded(),this.currentCommandEncoder.copyBufferToBuffer(e,0,r,0,t),this.submitQueue(),await r.mapAsync(GPUMapMode.READ);let n=r.getMappedRange().slice(0);return r.unmap(),r!=null&&this.bufferManager.releaseBuffer(r,t,GPUBufferUsage.COPY_DST|GPUBufferUsage.MAP_READ),Z().getBool("WEBGPU_USE_PROFILE_TOOL")&&(v.assert(this.dummyContext!==void 0,()=>"Fail to get context for profiling tool"),this.dummyContext.getCurrentTexture()),n}convertAndCacheOnCPU(e,t){let r=this.tensorMap.get(e);return this.maybeReleaseBuffer(e),r.values=t,r.values}readSync(e){let t=this.tensorMap.get(e),{values:r}=t;if(r==null)throw new Error("WebGPU readSync is only available for CPU-resident tensors.");return r}async read(e){if(!this.tensorMap.has(e))throw new Error(`Tensor ${e} was not registered!`);let t=this.tensorMap.get(e),{values:r}=t;if(r!=null)return this.convertAndCacheOnCPU(e,r);let n;if(t.dtype==="complex64"){let a=await Promise.all([this.read(t.complexTensorInfos.real.dataId),this.read(t.complexTensorInfos.imag.dataId)]),s=a[0],i=a[1];n=C.mergeRealAndImagArrays(s,i)}else{let a=t.values!=null?t.values:await this.getBufferData(t.bufferInfo.buffer,t.bufferInfo.byteSize);n=U9(a,t.dtype)}return this.convertAndCacheOnCPU(e,n),n}readToGPU(e,t={}){let r=this.tensorMap.get(e),{values:n,dtype:a,shape:s,bufferInfo:i}=r;if(a==="complex64")throw new Error("Does not support reading buffer for complex64 dtype.");if(i.buffer==null)throw n!=null?new Error("Data is not on GPU but on CPU."):new Error("There is no data on GPU or CPU.");let o=v.sizeFromShape(s)*m0(a);t.customBufSize!=null&&v.assert(t.customBufSize>=o,()=>`customBufSize should be equal or larger than the source tensor size ${o} bytes.`);let l=t.customBufSize!=null?t.customBufSize:o,u=this.acquireBuffer(l);this.ensureCommandEncoderReady(),this.ensureComputePassEnded(),this.currentCommandEncoder.copyBufferToBuffer(i.buffer,0,u,0,o),this.submitQueue();let d=this.makeTensorInfo(s,a),h=Xt().makeTensorFromTensorInfo(d),p=this.tensorMap.get(d.dataId);return p.bufferInfo.buffer=u,p.bufferInfo.byteSize=l,{tensorRef:h,buffer:u,bufSize:l}}bufferSync(e){let t=this.readSync(e.dataId),r=t;if(e.dtype==="string")try{r=t.map(n=>v.decodeString(n))}catch(n){throw new Error("Failed to decode encoded string bytes into utf-8")}return We(e.shape,e.dtype,r)}async time(e){let t=this.activeTimers,r=[],n=!1;this.programTimersStack==null?(this.programTimersStack=r,n=!0):this.activeTimers.push(r),this.activeTimers=r,e();let a=v.flatten(this.activeTimers.map(l=>l.query)).filter(l=>l!=null),s=v.flatten(this.activeTimers.map(l=>l.name)).filter(l=>l!=null);this.activeTimers=t,n&&(this.programTimersStack=null);let i={uploadWaitMs:this.uploadWaitMs,downloadWaitMs:this.downloadWaitMs,kernelMs:null,wallMs:null},o=await Promise.all(a);return i.kernelMs=v.sum(o),i.getExtraProfileInfo=()=>o.map((l,u)=>({name:s[u],ms:l})).map(l=>`${l.name}: ${l.ms}`).join(", "),this.uploadWaitMs=0,this.downloadWaitMs=0,i}getAndSavePipeline(e,t){return e in this.pipelineCache||(this.pipelineCache[e]=t()),this.pipelineCache[e]}makeTensorInfo(e,t,r){let n;if(t==="string"&&r!=null&&r.length>0&&v.isString(r[0])){let a=r.map(s=>v.encodeString(s));n=this.write(a,e,t)}else n=this.write(r,e,t);return{dataId:n,shape:e,dtype:t}}tensorToBinding(e){if(!e)return null;let t=this.tensorMap.get(e.dataId);return{offset:0,size:t.bufferInfo.byteSize,buffer:t.bufferInfo.buffer}}async getQueryTime(e){return this.supportTimeQuery?this.getTimeFromQuerySet(e):0}uploadToGPU(e){let t=this.tensorMap.get(e);if(t.bufferInfo.buffer==null&&(t.bufferInfo.buffer=this.acquireBuffer(t.bufferInfo.byteSize),t.values)){let r=this.bufferManager.acquireUploadBuffer(t.bufferInfo.byteSize,GPUBufferUsage.MAP_WRITE|GPUBufferUsage.COPY_SRC),n=r.getMappedRange();t.dtype==="int32"||t.dtype==="bool"?new Int32Array(n).set(t.values):new Float32Array(n).set(t.values),r.unmap(),this.ensureCommandEncoderReady(),this.ensureComputePassEnded(),this.currentCommandEncoder.copyBufferToBuffer(r,0,t.bufferInfo.buffer,0,t.bufferInfo.byteSize);let a={byteSize:t.bufferInfo.byteSize,usage:GPUBufferUsage.MAP_WRITE|GPUBufferUsage.COPY_SRC,buffer:r};this.stagingDisposalQueue.push(a)}}makeUniforms(e){let t=0,r=0,n=[];e.forEach(o=>{o.data.length===0&&(o.data=[1]);let l;switch(o.data.length){case 1:l=4;break;case 2:l=8;break;case 3:l=16;break;case 4:l=16;break;case 5:l=16;break;case 6:l=16;break;default:v.assert(!1,()=>`Unsupported ${o.data.length}D shape`)}(r===5||r===6)&&(l=16),t=Math.ceil(t/l)*l,r=o.data.length,n.push(t),t+=o.data.length*4});let a=new ArrayBuffer(t);e.forEach((o,l)=>{let u=n[l];o.type==="int32"?new Int32Array(a,u,o.data.length).set(o.data):o.type==="uint32"?new Uint32Array(a,u,o.data.length).set(o.data):new Float32Array(a,u,o.data.length).set(o.data)});let s=this.acquireBuffer(t,GPUBufferUsage.COPY_DST|GPUBufferUsage.UNIFORM);this.queue.writeBuffer(s,0,a,0,t);let i={byteSize:t,usage:GPUBufferUsage.COPY_DST|GPUBufferUsage.UNIFORM,buffer:s};return this.uniformDisposalQueue.push(i),{offset:0,size:t,buffer:s}}createLayout(e){let t=[];t.push({binding:0,visibility:GPUShaderStage.COMPUTE,buffer:{type:"storage"}});for(let a=0;a<e;a++)t.push({binding:a+1,visibility:GPUShaderStage.COMPUTE,buffer:{type:"read-only-storage"}});t.push({binding:e+1,visibility:GPUShaderStage.COMPUTE,buffer:{type:"uniform"}});let r=this.device.createBindGroupLayout({entries:t}),n=this.device.createPipelineLayout({bindGroupLayouts:[r]});return{bindGroupLayout:r,pipelineLayout:n}}getCachedOrCreateLayout(e){return e in this.layoutCache||(this.layoutCache[e]=this.createLayout(e)),this.layoutCache[e]}runWebGPUProgram(e,t,r,n,a){if(!a){if(a=this.makeTensorInfo(e.outputShape,r),v.sizeFromShape(a.shape)===0){let S=this.tensorMap.get(a.dataId);return S.values=v.getTypedArrayFromDType(a.dtype,0),a}this.uploadToGPU(a.dataId)}e.dispatch=vv(this.device,e);let s=[{type:"float32",data:[NaN]}],i=t.concat(a).map(S=>S.shape),o="int32";i.map(S=>{s.push({type:o,data:S})});let l=v.computeStrides(a.shape);if(s.push({type:o,data:l}),e.size){let S=v.sizeFromShape(e.outputShape);s.push({type:o,data:[e.isVec4?S/4:S]})}n&&(s=[...s,...n]);let u=this.makeUniforms(s),d=t.map((S,N)=>{if(S.dtype==="complex64")throw new Error("GPGPUProgram does not support complex64 input. For complex64 dtypes, please separate the program into real and imaginary parts.");return this.uploadToGPU(S.dataId),{dtype:this.tensorMap.get(S.dataId).dtype,shape:S.shape,name:e.variableNames[N]}}),h=d.map(S=>S.dtype).concat(a.dtype),p=d.map(S=>C.getBroadcastDims(S.shape,a.shape)),c=d.map(S=>v.arraysEqual(S.shape,a.shape)).join("_"),f=p.map(S=>S.join("_")).join(";"),m=bv(e,i,h,f,c),{bindGroupLayout:g,pipelineLayout:y}=this.getCachedOrCreateLayout(e.variableNames.length),A=this.getAndSavePipeline(m,()=>xv(this.device,e,y,d,a)),x=this.activeTimers!=null,b=ume(this.device,g,t.map(S=>this.tensorToBinding(S)),this.tensorToBinding(a),u);this.ensureCommandEncoderReady();let w=this.getComputePass();return x&&this.supportTimeQuery&&w.writeTimestamp(this.querySet,0),w.setPipeline(A),w.setBindGroup(0,b),w.dispatch(e.dispatch[0],e.dispatch[1],e.dispatch[2]),x&&this.supportTimeQuery&&w.writeTimestamp(this.querySet,1),this.dispatchNumberInEncoder++,t.forEach(S=>{this.commandQueueOwnedIds.add(S.dataId)}),this.commandQueueOwnedIds.add(a.dataId),Z().get("WEBGPU_DEFERRED_SUBMIT_BATCH_SIZE")<=this.dispatchNumberInEncoder&&this.submitQueue(),x&&this.activeTimers.push({name:e.constructor.name,query:this.getQueryTime(this.querySet)}),a}getFromPixelTextureLayout(e){return e?(this.fromPixelImportTextureLayout===null&&(this.fromPixelImportTextureLayout=this.createFromPixelTextureLayout(!0)),this.fromPixelImportTextureLayout):(this.fromPixelTextureLayout===null&&(this.fromPixelTextureLayout=this.createFromPixelTextureLayout(!1)),this.fromPixelTextureLayout)}createFromPixelTextureLayout(e){let t=[];t.push({binding:0,visibility:GPUShaderStage.COMPUTE,buffer:{type:"storage"}}),e?t.push({binding:1,visibility:GPUShaderStage.COMPUTE,externalTexture:{}}):t.push({binding:1,visibility:GPUShaderStage.COMPUTE,texture:{}}),t.push({binding:2,visibility:GPUShaderStage.COMPUTE,buffer:{}});let r=this.device.createBindGroupLayout({entries:t}),n=this.device.createPipelineLayout({bindGroupLayouts:[r]});return{bindGroupLayout:r,pipelineLayout:n}}copyExternalImageToTexture(e,t){let r=GPUTextureUsage.COPY_DST|GPUTextureUsage.RENDER_ATTACHMENT|GPUTextureUsage.TEXTURE_BINDING,n="rgba8unorm",a=this.textureManager.acquireTexture(t[1],t[0],n,r),s=a.createView();this.queue.copyExternalImageToTexture({source:e},{texture:a},[t[1],t[0]]);let i={width:t[1],height:t[0],format:n,usage:r,texture:a};return this.textureDisposalQueue.push(i),s}runFromPixelsProgram(e,t,r,n,a){e.dispatch=vv(this.device,e);let s=this.makeTensorInfo(t,"int32");if(v.sizeFromShape(s.shape)===0){let m=this.tensorMap.get(s.dataId);return m.values=v.getTypedArrayFromDType(s.dtype,0),s}this.uploadToGPU(s.dataId);let i=bv(e,[s.shape]),o=this.getFromPixelTextureLayout(n),l=this.getAndSavePipeline(i,()=>xv(this.device,e,o.pipelineLayout,[],s,!0)),u;if(n){let m={source:a};u=this.device.importExternalTexture(m)}else u=this.copyExternalImageToTexture(a,s.shape);let d=this.tensorToBinding(s),h=this.makeUniforms(r),p=this.device.createBindGroup({layout:o.bindGroupLayout,entries:[{binding:0,resource:{buffer:d.buffer}},{binding:1,resource:u},{binding:2,resource:{buffer:h.buffer}}]});this.ensureCommandEncoderReady();let c=this.getComputePass(),f=this.activeTimers!=null;return f&&this.supportTimeQuery&&c.writeTimestamp(this.querySet,0),c.setPipeline(l),c.setBindGroup(0,p),c.dispatch(e.dispatch[0],e.dispatch[1],e.dispatch[2]),f&&this.supportTimeQuery&&c.writeTimestamp(this.querySet,1),this.commandQueueOwnedIds.add(s.dataId),this.dispatchNumberInEncoder++,Z().get("WEBGPU_DEFERRED_SUBMIT_BATCH_SIZE")<=this.dispatchNumberInEncoder&&this.submitQueue(),f&&this.activeTimers.push({name:e.constructor.name,query:this.getQueryTime(this.querySet)}),s}async getTimeFromQuerySet(e){let t=this.acquireBuffer(16,GPUBufferUsage.COPY_SRC|GPUBufferUsage.QUERY_RESOLVE),r=this.acquireBuffer(16,GPUBufferUsage.MAP_READ|GPUBufferUsage.COPY_DST);this.ensureCommandEncoderReady(),this.ensureComputePassEnded(),this.currentCommandEncoder.resolveQuerySet(e,0,2,t,0),this.currentCommandEncoder.copyBufferToBuffer(t,0,r,0,16),this.submitQueue(),await r.mapAsync(GPUMapMode.READ);let n=new BigUint64Array(r.getMappedRange()),a=Number(n[1]-n[0]);return r.unmap(),this.bufferManager.releaseBuffer(r,16,GPUBufferUsage.MAP_READ|GPUBufferUsage.COPY_DST),this.bufferManager.releaseBuffer(t,16,GPUBufferUsage.COPY_SRC|GPUBufferUsage.QUERY_RESOLVE),a/1e6}shouldExecuteOnCPU(e,t=dme){return Z().getBool("WEBGPU_CPU_FORWARD")&&e.every(r=>this.tensorMap.get(r.dataId).bufferInfo.buffer==null&&v.sizeFromShape(r.shape)<t)}numDataIds(){return this.tensorMap.numDataIds()-this.tensorDisposalQueue.length}dispose(){this.disposed||(this.bufferManager.dispose(),this.textureManager.dispose(),this.disposed=!0)}},M5=uS;M5.nextDataId=0;var dS={};Le(dS,{WebGPUBackend:()=>M5,webgpu_util:()=>W9});N5()&&$l("webgpu",async()=>{Z().set("CHECK_COMPUTATION_FOR_ERRORS",!1);let e={powerPreference:Z().get("WEBGPU_USE_LOW_POWER_GPU")?"low-power":"high-performance"},t=await navigator.gpu.requestAdapter(e),r=t.limits,n={},a=t.features.has("timestamp-query");n.requiredLimits={maxComputeWorkgroupStorageSize:r.maxComputeWorkgroupStorageSize,maxComputeWorkgroupsPerDimension:r.maxComputeWorkgroupsPerDimension},a?n.requiredFeatures=["timestamp-query"]:console.warn("This device doesn't support timestamp-query extension. Start Chrome browser with flag --disable-dawn-features=disallow_unsafe_apis then try again. Or zero will shown for the kernel time when profiling mode isenabled. Using performance.now is not workable for webgpu sinceit doesn't support synchronously to read data from GPU.");let s=await t.requestDevice(n);return new M5(s,a)},3);var Ut=(e=>(e[e.float32=0]="float32",e[e.int32=1]="int32",e[e.bool=2]="bool",e[e.string=3]="string",e[e.complex64=4]="complex64",e))(Ut||{}),Mm=(e=>(e[e.linear=0]="linear",e[e.relu=1]="relu",e[e.relu6=2]="relu6",e[e.prelu=3]="prelu",e[e.leakyrelu=4]="leakyrelu",e[e.sigmoid=5]="sigmoid",e[e.elu=6]="elu",e))(Mm||{}),pS;function pme(e){pS=e.wasm.cwrap(Ps,null,["number","array","number","number","array","number","number","number","number","number","number","number","number"])}function hme(e){let{inputs:t,backend:r,attrs:n}=e,{a,b:s,bias:i,preluActivationWeights:o}=t;if(a.dtype!=="float32"||s.dtype!=="float32")throw new Error("_FusedMatMul for non non-float32 tensors not yet supported.");let{transposeA:l,transposeB:u,activation:d,leakyreluAlpha:h}=n,p=r.dataIdMap.get(a.dataId).id,c=r.dataIdMap.get(s.dataId).id,f=0;if(i!=null){let E=r.dataIdMap.get(i.dataId);if(E.shape.length!==1)throw new Error(`_FusedMatMul only supports rank-1 bias but got rank ${E.shape.length}.`);f=E.id}let m=o==null?0:r.dataIdMap.get(o.dataId).id,g=Mm[d];if(g==null)throw new Error(`${d} activation not yet supported for FusedConv2D in the wasm backend.`);let y=l?a.shape[2]:a.shape[1],A=u?s.shape[1]:s.shape[2],x=ji.assertAndGetBroadcastShape(a.shape.slice(0,-2),s.shape.slice(0,-2)),b=r.makeOutput([...x,y,A],a.dtype),w=r.dataIdMap.get(b.dataId).id,S=new Uint8Array(new Int32Array(a.shape).buffer),N=new Uint8Array(new Int32Array(s.shape).buffer);return pS(p,S,a.shape.length,c,N,s.shape.length,l,u,g,f,m,h||0,w),b}var cme={kernelName:Ps,backendName:"wasm",setupFunc:pme,kernelFunc:hme};function Ir(e,t){let r;function n(s){r=s.wasm.cwrap(e,null,["number","number","number"])}function a(s){let{backend:i,inputs:{x:o}}=s,l=i.dataIdMap.get(o.dataId).id,u=i.makeOutput(o.shape,t||o.dtype),d=i.dataIdMap.get(u.dataId).id;return v.sizeFromShape(u.shape)===0||r(l,Ut[o.dtype],d),u}return{kernelName:e,backendName:"wasm",setupFunc:n,kernelFunc:a}}var fme=Ir(Ho);function Xr(e,t,r){let n;function a(i){n=i.wasm.cwrap(e,null,["number","array","number","number","array","number","number","number"])}function s(i){let{backend:o,inputs:l}=i,{a:u,b:d}=l,h=o.dataIdMap.get(u.dataId).id,p=o.dataIdMap.get(d.dataId).id,c=r!=null?r:u.dtype,f=C.assertAndGetBroadcastShape(u.shape,d.shape),m=o.makeOutput(f,c);if(v.sizeFromShape(f)===0)return m;let g=new Uint8Array(new Int32Array(u.shape).buffer),y=new Uint8Array(new Int32Array(d.shape).buffer),A=o.dataIdMap.get(m.dataId).id;return n(h,g,u.shape.length,p,y,d.shape.length,Ut[u.dtype],A),m}return{kernelName:e,backendName:"wasm",setupFunc:a,kernelFunc:s}}var mme=!0,gme=Xr(Ja,mme),hS;function yme(e){hS=e.wasm.cwrap(Zs,null,["array","number","number","number"])}function Ame(e){let{inputs:t,backend:r}=e,n=r.makeOutput(t[0].shape,t[0].dtype);if(v.sizeFromShape(n.shape)===0)return n;let a=t.map(o=>r.dataIdMap.get(o.dataId).id),s=new Uint8Array(new Int32Array(a).buffer),i=r.dataIdMap.get(n.dataId).id;return hS(s,a.length,Ut[n.dtype],i),n}var xme={kernelName:Zs,backendName:"wasm",setupFunc:yme,kernelFunc:Ame};function Fm(e){let{inputs:{x:t},backend:r}=e,n=r.makeOutput(t.shape,t.dtype),a=r.typedArrayFromHeap(t);return r.typedArrayFromHeap(n).set(a),n}var bme={kernelName:mi,backendName:"wasm",kernelFunc:Fm},cS;function vme(e){cS=e.wasm.cwrap(Vi,null,["number","array","number","number","number","array","number"])}function Xs(e){let{inputs:t,backend:r,attrs:n}=e,[a,s]=kme(t.x.shape,n.perm),i=!0;for(let f=0;f<s.length;f++)s[f]!==f&&(i=!1);let o=wme(t.x.shape,n.perm),l={dataId:t.x.dataId,shape:a,dtype:t.x.dtype};if(i){let f=Fm({inputs:t,backend:r});return f.shape=o,f}let u=r.makeOutput(o,l.dtype),d=r.dataIdMap.get(l.dataId).id,h=r.dataIdMap.get(u.dataId).id,p=new Uint8Array(new Int32Array(s).buffer),c=new Uint8Array(new Int32Array(l.shape).buffer);return cS(d,c,l.shape.length,Ut[l.dtype],h,p,s.length),u}function wme(e,t){let r=new Array(e.length);for(let n=0;n<r.length;n++)r[n]=e[t[n]];return r}function kme(e,t){let r=[],n=[];for(let a=0;a<e.length;++a)e[a]!==1&&r.push(e[a]),e[t[a]]!==1&&n.push(t[a]);for(let a=0;a<n.length;++a){let s=-1;for(let i=0;i<n.length;++i)n[i]>=a&&(s===-1||n[s]>n[i])&&(s=i);n[s]=a}return[r,n]}var Ime={kernelName:Vi,backendName:"wasm",kernelFunc:Xs,setupFunc:vme};function Yi(e,t,r){let n=e.shape,a=e.shape.length,s=v.parseAxisParam(t,n),i=s,o=C.getAxesPermutation(i,a),l=null,u=!1;if(o!=null){let d=new Array(a);for(let p=0;p<d.length;p++)d[p]=n[o[p]];i=C.getInnerMostAxes(i.length,a),l=Xs({inputs:{x:e},attrs:{perm:o},backend:r});let h=r.dataIdMap.get(e.dataId).id;r.dataIdMap.get(l.dataId).id!==h&&(u=!0)}return{transposed:l,originalAxes:s,axes:i,inputWasTransposed:u}}var fS;function Sme(e){fS=e.wasm.cwrap(Bu,null,["number, number, number"])}function Cme(e){let{backend:t,inputs:r,attrs:n}=e,{axis:a,keepDims:s}=n,{x:i}=r,o=t.dataIdMap.get(i.dataId).id,l=i,{transposed:u,axes:d,originalAxes:h,inputWasTransposed:p}=Yi(i,a,t);if(p){let A=t.dataIdMap.get(u.dataId).id;l=u,o=A}let c=l.shape.length;C.assertAxesAreInnerMostDims("all",d,c);let[f,m]=C.computeOutAndReduceShapes(l.shape,d),g=v.sizeFromShape(m),y=t.makeOutput(f,i.dtype);if(v.sizeFromShape(l.shape)!==0){let A=t.dataIdMap.get(y.dataId).id;fS(o,g,A)}if(p&&t.disposeData(u.dataId),s){let A=C.expandShapeToKeepDim(y.shape,h);y.shape=A}return y}var Tme={kernelName:Bu,backendName:"wasm",setupFunc:Sme,kernelFunc:Cme},mS;function Nme(e){mS=e.wasm.cwrap(Wu,null,["number, number, number"])}function Eme(e){let{backend:t,inputs:r,attrs:n}=e,{axis:a,keepDims:s}=n,{x:i}=r,o=t.dataIdMap.get(i.dataId).id,l=i,{transposed:u,axes:d,originalAxes:h,inputWasTransposed:p}=Yi(i,a,t);if(p){let A=t.dataIdMap.get(u.dataId).id;l=u,o=A}let c=l.shape.length;C.assertAxesAreInnerMostDims("any",d,c);let[f,m]=C.computeOutAndReduceShapes(l.shape,d),g=v.sizeFromShape(m),y=t.makeOutput(f,i.dtype);if(v.sizeFromShape(l.shape)!==0){let A=t.dataIdMap.get(y.dataId).id;mS(o,g,A)}if(p&&t.disposeData(u.dataId),s){let A=C.expandShapeToKeepDim(y.shape,h);y.shape=A}return y}var Rme={kernelName:Wu,backendName:"wasm",setupFunc:Nme,kernelFunc:Eme},gS;function $me(e){gS=e.wasm.cwrap(Ys,null,["number","number","number","number","number"])}function Mme(e){let{backend:t,inputs:r,attrs:n}=e,{axis:a}=n,{x:s}=r,i=t.dataIdMap.get(s.dataId).id,o=i,l=s,{transposed:u,axes:d,inputWasTransposed:h}=Yi(s,a,t);if(h){let y=t.dataIdMap.get(u.dataId).id;y!==i&&(l=u,o=y)}let p=l.shape.slice(0,-1),c=t.makeOutput(p,"int32"),f=t.dataIdMap.get(c.dataId).id,m=v.sizeFromShape(c.shape),g=l.shape[d[0]];return gS(o,Ut[l.dtype],m,g,f),h&&t.disposeData(u.dataId),c}var Fme={kernelName:Ys,backendName:"wasm",kernelFunc:Mme,setupFunc:$me},yS;function Pme(e){yS=e.wasm.cwrap(Js,null,["number","number","number","number","number","number","number","number","number","number","number","number","number","number"])}function _me(e){let{inputs:t,attrs:r,backend:n}=e,a=t.x,s=n.dataIdMap.get(a.dataId).id,{filterSize:i,strides:o,pad:l,dimRoundingMode:u}=r,d=C.computePool2DInfo(a.shape,i,o,1,l,u),h=d.filterHeight,p=d.filterWidth,c=d.padInfo.top,f=d.padInfo.right,m=d.padInfo.bottom,g=d.padInfo.left,y=d.strideHeight,A=d.strideWidth,x=d.inChannels;if(d.dataFormat!=="channelsLast")throw new Error(`wasm backend does not support dataFormat:'${d.dataFormat}'. Please use 'channelsLast'.`);if(d.dilationWidth!==1||d.dilationHeight!==1)throw new Error(`was backend only supports average pooling with dilation = [1, 1], got [${d.dilationHeight}, ${d.dilationWidth}].`);let b=n.makeOutput(d.outShape,"float32"),w=n.dataIdMap.get(b.dataId).id;return yS(s,a.shape[0],a.shape[1],a.shape[2],h,p,c,f,m,g,y,A,x,w),b}var zme={kernelName:Js,backendName:"wasm",setupFunc:Pme,kernelFunc:_me};function rn(e){let{inputs:t,attrs:r}=e,{x:n}=t,{shape:a}=r,s=v.sizeFromShape(n.shape),i=v.inferFromImplicitShape(a,s);return v.assert(s===v.sizeFromShape(i),()=>`new shape: ${i}, old shape: ${n.shape}. New shape and old shape must have the same number of elements.`),e.backend.incRef(n.dataId),{dataId:n.dataId,shape:i,dtype:n.dtype}}var Ome={kernelName:ml,backendName:"wasm",kernelFunc:rn},AS;function Dme(e){AS=e.wasm.cwrap(Qs,null,["number","array","number","number","array","number","number","number","number"])}function Lme(e){let{inputs:t,backend:r,attrs:n}=e,{a,b:s}=t,{transposeA:i,transposeB:o}=n;if(a.dtype!=="float32"||s.dtype!=="float32")throw new Error("BatchMatMul for non non-float32 tensors not yet supported.");let l=a.shape.length,u=s.shape.length,d=i?a.shape[l-2]:a.shape[l-1],h=o?s.shape[u-1]:s.shape[u-2],p=i?a.shape[l-1]:a.shape[l-2],c=o?s.shape[u-2]:s.shape[u-1],f=a.shape.slice(0,-2),m=s.shape.slice(0,-2),g=v.sizeFromShape(f),y=v.sizeFromShape(m),A=ji.assertAndGetBroadcastShape(a.shape.slice(0,-2),s.shape.slice(0,-2)).concat([p,c]);v.assert(d===h,()=>`Error in matMul: inner shapes (${d}) and (${h}) of Tensors with shapes ${a.shape} and ${s.shape} and transposeA=${i} and transposeB=${o} must match.`);let x=i?[g,d,p]:[g,p,d],b=o?[y,c,h]:[y,h,c],w=rn({inputs:{x:a},backend:r,attrs:{shape:x}}),S=rn({inputs:{x:s},backend:r,attrs:{shape:b}}),N=r.dataIdMap.get(w.dataId).id,E=r.dataIdMap.get(S.dataId).id,R=i?w.shape[2]:w.shape[1],z=o?S.shape[1]:S.shape[2],M=Math.max(g,y),I=r.makeOutput([M,R,z],w.dtype),_=r.dataIdMap.get(I.dataId).id,O=new Uint8Array(new Int32Array(w.shape).buffer),q=new Uint8Array(new Int32Array(S.shape).buffer);return AS(N,O,w.shape.length,E,q,S.shape.length,i,o,_),r.disposeData(w.dataId),r.disposeData(S.dataId),I.shape=A,I}var Bme={kernelName:Qs,backendName:"wasm",setupFunc:Dme,kernelFunc:Lme};function Uo(e){let{inputs:{x:t},attrs:{begin:r,size:n},backend:a}=e,[s,i]=Ot.parseSliceParams(t,r,n),o=Ot.isSliceContinous(t.shape,s,i),l=a.readSync(t.dataId),u=a.makeOutput(i,t.dtype),d=v.computeStrides(t.shape),h=a.dataIdMap.get(u.dataId);if(o){let f=Ot.computeFlatOffset(s,d);return t.dtype==="string"?h.stringBytes=l.slice(f,f+v.sizeFromShape(i)):a.typedArrayFromHeap(u).set(l.subarray(f,f+v.sizeFromShape(i))),u}if(t.dtype==="string"){let f=W0(l,s,i,t.shape,t.dtype);return h.stringBytes=f,u}let p=a.typedArrayFromHeap(u),c=t.shape.length;if(c===2)Wme(l,d[0],p,s,i);else if(c===3)Vme(l,d[0],d[1],p,s,i);else if(c===4)Ume(l,d[0],d[1],d[2],p,s,i);else{let f=W0(l,s,i,t.shape,t.dtype);p.set(f)}return u}function Wme(e,t,r,n,a){let s=0,i=n[0],o=n[1],l=i+a[0];for(let u=i;u<l;u++){let d=u*t+o;r.set(e.subarray(d,d+a[1]),s),s+=a[1]}}function Vme(e,t,r,n,a,s){let i=0,o=a[0],l=a[1],u=a[2],d=o+s[0],h=l+s[1];for(let p=o;p<d;p++)for(let c=l;c<h;c++){let f=p*t+c*r+u;n.set(e.subarray(f,f+s[2]),i),i+=s[2]}}function Ume(e,t,r,n,a,s,i){let o=0,l=s[0],u=s[1],d=s[2],h=l+i[0],p=u+i[1],c=d+i[2],f=s[3];for(let m=l;m<h;m++)for(let g=u;g<p;g++)for(let y=d;y<c;y++){let A=m*t+g*r+y*n+f;a.set(e.subarray(A,A+i[3]),o),o+=i[3]}}var Gme={kernelName:bl,backendName:"wasm",kernelFunc:Uo};function jme(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{blockShape:s,crops:i}=n,o=s.reduce((y,A)=>y*A),l=C.getReshaped(a.shape,s,o),u=C.getPermuted(l.length,s.length),d=C.getReshapedPermuted(a.shape,s,o),h=C.getSliceBeginCoords(i,s.length),p=C.getSliceSize(d,i,s.length),c=rn({inputs:{x:a},backend:r,attrs:{shape:l}}),f=Xs({inputs:{x:c},backend:r,attrs:{perm:u}}),m=rn({inputs:{x:f},backend:r,attrs:{shape:d}}),g=Uo({inputs:{x:m},backend:r,attrs:{begin:h,size:p}});return r.disposeData(c.dataId),r.disposeData(f.dataId),r.disposeData(c.dataId),g}var Hme={kernelName:qo,backendName:"wasm",kernelFunc:jme};function Xh(e){let{inputs:{x:t},attrs:{dtype:r},backend:n}=e,a=n.makeOutput(t.shape,r),s=n.typedArrayFromHeap(t);return n.typedArrayFromHeap(a).set(s),a}var qme={kernelName:ei,backendName:"wasm",kernelFunc:Xh},Xme=Ir(ti),xS;function Kme(e){xS=e.wasm.cwrap(Qa,null,["number","number","number","number"])}function Zme(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{clipValueMin:s,clipValueMax:i}=n,o=r.dataIdMap.get(a.dataId).id,l=r.makeOutput(a.shape,a.dtype),u=r.dataIdMap.get(l.dataId).id;return xS(o,s,i,u),l}var Yme={kernelName:Qa,backendName:"wasm",setupFunc:Kme,kernelFunc:Zme};function bS(e){let{inputs:t,backend:r}=e,n=v.parseAxisParam(e.attrs.axis,t[0].shape)[0],a=C.computeOutShape(t.map(c=>c.shape),n),s=t.filter(c=>v.sizeFromShape(c.shape)>0);if(s.length===1)return Fm({inputs:{x:s[0]},backend:r});let i=r.makeOutput(a,t[0].dtype);if(v.sizeFromShape(a)===0)return i;let o=s.map(c=>c.shape);if(C.assertParamsConsistent(o,n),s[0].dtype==="string"){let c=s.map(x=>{let b=v.sizeFromShape(x.shape.slice(n));return rn({inputs:{x},backend:r,attrs:{shape:[-1,b]}})}),f=c.map(x=>({vals:r.readSync(x.dataId),shape:x.shape}));a=C.computeOutShape(c.map(x=>x.shape),1);let m=c[0].shape[0]===1,g=r5(f,a,t[0].dtype,m),y=C.computeOutShape(s.map(x=>x.shape),n);i.shape=y;let A=r.dataIdMap.get(i.dataId);return A.stringBytes=C.fromStringArrayToUint8(g),c.forEach(x=>r.disposeData(x.dataId)),i}let l=v.sizeFromShape(s[0].shape.slice(0,n)),u=0,d=s.map(c=>{let f=v.sizeFromShape(c.shape.slice(n));return u+=f,f}),h=s.map(c=>r.typedArrayFromHeap(c)),p=r.typedArrayFromHeap(i);for(let c=0;c<l;c++){let f=c*u;for(let m=0;m<h.length;m++){let g=d[m],y=c*g,A=h[m].subarray(y,y+g);p.set(A,f),f+=g}}return i}var Jme={kernelName:Xo,backendName:"wasm",kernelFunc:bS},vS;function Qme(e){vS=e.wasm.cwrap(ri,null,["number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number"])}function e1e(e){let{inputs:t,attrs:r,backend:n}=e,{x:a,filter:s}=t,i=n.dataIdMap.get(a.dataId).id,o=n.dataIdMap.get(s.dataId).id,{strides:l,dilations:u,pad:d,dimRoundingMode:h,dataFormat:p}=r,c=C.convertConv2DDataFormat(p),f=C.computeConv2DInfo(a.shape,s.shape,l,u,d,h,!1,c),m=f.filterHeight,g=f.filterWidth,y=f.padInfo.top,A=f.padInfo.right,x=f.padInfo.bottom,b=f.padInfo.left,w=f.dilationHeight,S=f.dilationWidth,N=f.strideHeight,E=f.strideWidth,R=f.inChannels,z=f.outChannels,M=f.padInfo.type==="SAME"?1:0;if(f.dataFormat!=="channelsLast")throw new Error(`wasm backend Conv2D does not support dataFormat:'${f.dataFormat}'. Please use 'channelsLast'.`);let I=n.makeOutput(f.outShape,"float32"),_=n.dataIdMap.get(I.dataId).id;return vS(i,a.shape[0],a.shape[1],a.shape[2],o,m,g,y,A,x,b,M,w,S,N,E,R,z,_),I}var t1e={kernelName:ri,backendName:"wasm",setupFunc:Qme,kernelFunc:e1e},wS;function r1e(e){wS=e.wasm.cwrap(ni,null,["number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number"])}function n1e(e){let{backend:t,inputs:r,attrs:n}=e,{dy:a,filter:s}=r,{strides:i,pad:o,dataFormat:l,dimRoundingMode:u,inputShape:d}=n,h=1,p=C.convertConv2DDataFormat(l),c=C.computeConv2DInfo(d,s.shape,i,h,o,u,!1,p),{batchSize:f,filterHeight:m,filterWidth:g,inChannels:y,inHeight:A,inWidth:x,outChannels:b,outHeight:w,outWidth:S,strideHeight:N,strideWidth:E}=c,R=m-1-c.padInfo.top,z=g-1-c.padInfo.left,M=c.dataFormat==="channelsLast",I=v.computeStrides(c.inShape),_=v.computeStrides(a.shape),[O,q,K]=v.computeStrides(s.shape),D=I[0],Q=M?I[1]:I[2],V=M?I[2]:1,re=M?1:I[1],Y=_[0],ie=M?_[1]:_[2],J=M?_[2]:1,ae=M?1:_[1],de=t.makeOutput(c.inShape,"float32"),xe=t.dataIdMap.get(de.dataId).id,be=t.dataIdMap.get(a.dataId).id,Ee=t.dataIdMap.get(s.dataId).id;return wS(be,Ee,f,m,g,A,x,y,w,S,b,N,E,R,z,O,q,K,D,Q,V,re,Y,ie,J,ae,xe),de}var a1e={kernelName:ni,backendName:"wasm",setupFunc:r1e,kernelFunc:n1e},s1e=Ir(ai),i1e=Ir(si),kS=(e=>(e[e.bilinear=0]="bilinear",e[e.nearest=1]="nearest",e))(kS||{}),IS;function o1e(e){IS=e.wasm.cwrap(Zo,null,["number","number","number","number","array","number","number","number","number","number"])}function l1e(e){let{backend:t,inputs:r,attrs:n}=e,{method:a,extrapolationValue:s,cropSize:i}=n,{image:o,boxes:l,boxInd:u}=r,d=l.shape[0],[h,p]=i,c=[d,h,p,o.shape[3]],f=t.dataIdMap.get(o.dataId),m;o.dtype!=="float32"&&(m=Xh({backend:t,inputs:{x:o},attrs:{dtype:"float32"}}),f=t.dataIdMap.get(m.dataId));let g=f.id,y=t.dataIdMap.get(l.dataId).id,A=t.dataIdMap.get(u.dataId).id,x=t.makeOutput(c,"float32"),b=t.dataIdMap.get(x.dataId).id,w=new Uint8Array(new Int32Array(o.shape).buffer);return IS(g,y,A,d,w,h,p,kS[a],s,b),m!=null&&t.disposeData(m.dataId),x}var u1e={kernelName:Zo,backendName:"wasm",setupFunc:o1e,kernelFunc:l1e},SS;function d1e(e){SS=e.wasm.cwrap(Ko,null,["number","number","number","number","number","number"])}function p1e(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{axis:s,exclusive:i,reverse:o}=n,l=a.shape.length;v.assert(a.dtype==="float32"||a.dtype==="int32",()=>`cumprod does not support ${a.dtype} tensors in the WASM backend`);let u=C.getAxesPermutation([s],l),d=a;u!==null&&(d=Xs({inputs:{x:a},attrs:{perm:u},backend:r}));let h=C.getInnerMostAxes(1,l)[0];C.assertAxesAreInnerMostDims("cumprod",[h],l);let p=r.makeOutput(d.shape,d.dtype),c=d.shape[h],f=r.dataIdMap.get(d.dataId).id,m=r.dataIdMap.get(p.dataId).id;SS(f,i?1:0,o?1:0,c,m,Ut[a.dtype]);let g=p;if(u!==null){let y=C.getUndoAxesPermutation(u);g=Xs({inputs:{x:p},attrs:{perm:y},backend:r}),r.disposeData(d.dataId),r.disposeData(p.dataId)}return g}var h1e={kernelName:Ko,backendName:"wasm",setupFunc:d1e,kernelFunc:p1e},CS;function c1e(e){CS=e.wasm.cwrap(ii,null,["number","number","number","number","number","number"])}function f1e(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{axis:s,exclusive:i,reverse:o}=n,l=a.shape.length;v.assert(a.dtype==="float32"||a.dtype==="int32",()=>`cumsum does not support ${a.dtype} tensors in the WASM backend`);let u=C.getAxesPermutation([s],l),d=a;u!==null&&(d=Xs({inputs:{x:a},attrs:{perm:u},backend:r}));let h=C.getInnerMostAxes(1,l)[0];C.assertAxesAreInnerMostDims("cumsum",[h],l);let p=r.makeOutput(d.shape,d.dtype),c=d.shape[h],f=r.dataIdMap.get(d.dataId).id,m=r.dataIdMap.get(p.dataId).id;CS(f,i?1:0,o?1:0,c,m,Ut[a.dtype]);let g=p;if(u!==null){let y=C.getUndoAxesPermutation(u);g=Xs({inputs:{x:p},attrs:{perm:y},backend:r}),r.disposeData(d.dataId),r.disposeData(p.dataId)}return g}var m1e={kernelName:ii,backendName:"wasm",setupFunc:c1e,kernelFunc:f1e},TS;function g1e(e){TS=e.wasm.cwrap(Yo,null,["number","number","number","array","number","array","array","number","number"])}function y1e(e){let{backend:t,inputs:r,attrs:n}=e,{x:a}=r,{blockSize:s,dataFormat:i}=n,o=a.shape[0],l=i==="NHWC"?a.shape[1]:a.shape[2],u=i==="NHWC"?a.shape[2]:a.shape[3],d=i==="NHWC"?a.shape[3]:a.shape[1],h=l*s,p=u*s,c=d/(s*s),f=i==="NHWC"?[o,h,p,c]:[o,c,h,p],m=t.makeOutput(f,"float32"),g=t.dataIdMap.get(a.dataId).id,y=new Uint8Array(new Int32Array(v.computeStrides(a.shape)).buffer),A=new Uint8Array(new Int32Array(f).buffer),x=new Uint8Array(new Int32Array(v.computeStrides(f)).buffer),b=t.dataIdMap.get(m.dataId).id;return TS(g,s,i==="NHWC"?1:0,y,a.shape.length-1,A,x,f.length,b),m}var A1e={kernelName:Yo,backendName:"wasm",setupFunc:g1e,kernelFunc:y1e},NS;function x1e(e){NS=e.wasm.cwrap(oi,null,["number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number"])}function b1e(e){let{inputs:t,attrs:r,backend:n}=e,{x:a,filter:s}=t,i=n.dataIdMap.get(a.dataId).id,o=n.dataIdMap.get(s.dataId).id,{strides:l,dilations:u,pad:d,dimRoundingMode:h}=r,p=u==null?[1,1]:u,c=C.computeConv2DInfo(a.shape,s.shape,l,p,d,h,!0),f=c.filterHeight,m=c.filterWidth,g=c.padInfo.top,y=c.padInfo.right,A=c.padInfo.bottom,x=c.padInfo.left,b=c.dilationHeight,w=c.dilationWidth,S=c.strideHeight,N=c.strideWidth,E=c.inChannels,R=c.outChannels,z=c.padInfo.type==="SAME"?1:0;if(c.dataFormat!=="channelsLast")throw new Error(`wasm backend DepthwiseConv2dNative does not support dataFormat:'${c.dataFormat}'. Please use 'channelsLast'.`);let M=n.makeOutput(c.outShape,"float32"),I=n.dataIdMap.get(M.dataId).id;return NS(i,a.shape[0],a.shape[1],a.shape[2],o,f,m,g,y,A,x,z,b,w,S,N,E,R,I),M}var v1e={kernelName:oi,backendName:"wasm",setupFunc:x1e,kernelFunc:b1e},w1e=Ir(ui),k1e=!1,I1e=Xr(Jo,k1e,"bool"),S1e=Ir(di,"float32");function py(e){let{inputs:t,attrs:r,backend:n}=e,{input:a}=t,{dim:s}=r,i=a.shape.length,o=a.shape.slice(),l=s;return s<0&&(v.assert(-(i+1)<=s,()=>`Axis must be in the interval [${-(i+1)}, ${i}]`),l=i+s+1),o.splice(l,0,1),rn({inputs:{x:a},backend:n,attrs:{shape:o}})}var C1e={kernelName:Qo,backendName:"wasm",kernelFunc:py};function ES(e){let{attrs:{shape:t,value:r,dtype:n},backend:a}=e,s=a.makeOutput(t,n);return a.typedArrayFromHeap(s).fill(r),s}var T1e={kernelName:Ku,backendName:"wasm",kernelFunc:ES},RS;function N1e(e){RS=e.wasm.cwrap(tl,null,["number","number","number","number","number","number"])}function E1e(e){let{inputs:t,backend:r}=e,{image:n}=t,a=r.makeOutput(n.shape,n.dtype),s=r.dataIdMap.get(n.dataId).id,i=r.dataIdMap.get(a.dataId).id,[o,l,u,d]=n.shape;return RS(s,o,l,u,d,i),a}var R1e={kernelName:tl,backendName:"wasm",kernelFunc:E1e,setupFunc:N1e},$1e=Ir(pi),M1e=!1,F1e=Xr(hi,M1e),$S;function P1e(e){$S=e.wasm.cwrap(ci,null,["number","number","number","number","number","number","number"])}function _1e(e){let{backend:t,inputs:r,attrs:n}=e,{varianceEpsilon:a}=n,{x:s,mean:i,variance:o,offset:l,scale:u}=r,d=t.dataIdMap.get(s.dataId).id,h=t.dataIdMap.get(i.dataId).id,p=t.dataIdMap.get(o.dataId).id,c=l!=null?t.dataIdMap.get(l.dataId).id:0,f=u!=null?t.dataIdMap.get(u.dataId).id:0,m=t.makeOutput(s.shape,s.dtype);if(v.sizeFromShape(s.shape)===0)return m;let g=t.dataIdMap.get(m.dataId).id;return $S(d,h,p,c,f,a,g),m}var z1e={kernelName:ci,backendName:"wasm",setupFunc:P1e,kernelFunc:_1e},MS;function O1e(e){MS=e.wasm.cwrap(_s,null,["number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number"])}function D1e(e){let{inputs:t,attrs:r,backend:n}=e,{x:a,filter:s,bias:i,preluActivationWeights:o}=t,{strides:l,pad:u,dilations:d,dataFormat:h,dimRoundingMode:p,activation:c,leakyreluAlpha:f}=r,m=C.computeConv2DInfo(a.shape,s.shape,l,d,u,p),g=Mm[c];if(g==null)throw new Error(`${c} activation not yet supported for FusedConv2D in the wasm backend.`);let y=n.dataIdMap.get(a.dataId).id,A=n.dataIdMap.get(s.dataId).id,x=m.outChannels,b=0;if(i!=null){let J=n.dataIdMap.get(i.dataId);if(J.shape.length!==1)throw new Error(`FusedConv2D only supports rank-1 bias but got rank ${J.shape.length}.`);if(J.shape[0]!==x)throw new Error(`FusedConv2D bias shape (${J.shape}) does not match the number of output channels (${x})`);b=J.id}let w=m.filterHeight,S=m.filterWidth,N=m.padInfo.top,E=m.padInfo.right,R=m.padInfo.bottom,z=m.padInfo.left,M=m.dilationHeight,I=m.dilationWidth,_=m.strideHeight,O=m.strideWidth,q=m.inChannels,K=m.padInfo.type==="SAME"?1:0,D=m.batchSize,Q=m.inHeight,V=m.inWidth;if(h!=="NHWC")throw new Error(`wasm backend FusedConv2D does not support dataFormat:'${h}'. Please use 'NHWC'.`);let re=n.makeOutput(m.outShape,"float32"),Y=n.dataIdMap.get(re.dataId).id,ie=o==null?0:n.dataIdMap.get(o.dataId).id;return MS(y,D,Q,V,A,w,S,b,N,E,R,z,K,M,I,_,O,q,x,g,ie,f||0,Y),re}var L1e={kernelName:_s,backendName:"wasm",setupFunc:O1e,kernelFunc:D1e},FS;function B1e(e){FS=e.wasm.cwrap(zs,null,["number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number"])}function W1e(e){let{inputs:t,attrs:r,backend:n}=e,{x:a,filter:s,bias:i,preluActivationWeights:o}=t,{strides:l,pad:u,dilations:d,dataFormat:h,dimRoundingMode:p,activation:c,leakyreluAlpha:f}=r,m=C.computeConv2DInfo(a.shape,s.shape,l,d,u,p,!0),g=Mm[c];if(g==null)throw new Error(`${c} activation not yet supported for FusedDepthwiseConv2D in the wasm backend.`);let y=n.dataIdMap.get(a.dataId).id,A=n.dataIdMap.get(s.dataId).id,x=m.outChannels,b=0;if(i!=null){let J=n.dataIdMap.get(i.dataId);if(J.shape.length!==1)throw new Error(`FusedDepthwiseConv2D only supports rank-1 bias but got rank ${J.shape.length}.`);if(J.shape[0]!==x)throw new Error(`FusedDepthwiseConv2D bias shape (${J.shape}) does not match the number of output channels (${x})`);b=J.id}let w=m.filterHeight,S=m.filterWidth,N=m.padInfo.top,E=m.padInfo.right,R=m.padInfo.bottom,z=m.padInfo.left,M=m.dilationHeight,I=m.dilationWidth,_=m.strideHeight,O=m.strideWidth,q=m.inChannels,K=m.padInfo.type==="SAME"?1:0,D=m.batchSize,Q=m.inHeight,V=m.inWidth;if(h!=="NHWC")throw new Error(`wasm backend FusedDepthwiseConv2D does not support dataFormat:'${h}'. Please use 'NHWC'.`);let re=n.makeOutput(m.outShape,"float32"),Y=n.dataIdMap.get(re.dataId).id,ie=o==null?0:n.dataIdMap.get(o.dataId).id;return FS(y,D,Q,V,A,w,S,b,N,E,R,z,K,M,I,_,O,q,x,g,ie,f||0,Y),re}var V1e={kernelName:zs,backendName:"wasm",setupFunc:B1e,kernelFunc:W1e},PS;function U1e(e){PS=e.wasm.cwrap(nl,null,["number","number","number","number","number","number","array","number"])}function G1e(e){let{backend:t,inputs:r}=e,{params:n,indices:a}=r,[s,i,o,l]=Ty.prepareAndValidate(n,a),u=t.makeOutput(s,n.dtype);if(i===0)return u;let d=a.shape,h=d[d.length-1],p=t.dataIdMap.get(n.dataId).id,c=t.dataIdMap.get(a.dataId).id,f=new Uint8Array(new Int32Array(l).buffer),m=t.dataIdMap.get(u.dataId).id;return PS(p,Ut[n.dtype],c,i,h,o,f,m),u}var j1e={kernelName:nl,backendName:"wasm",setupFunc:U1e,kernelFunc:G1e},_S;function H1e(e){_S=e.wasm.cwrap("Gather",null,["number","number","array","number","number","number","array","number"])}function q1e(e){let{backend:t,inputs:r,attrs:n}=e,{x:a,indices:s}=r,{axis:i,batchDims:o}=n,l=v.parseAxisParam(i,a.shape)[0],u=t.readSync(s.dataId),d=a.shape[l];for(let N=0;N<u.length;++N){let E=u[N];v.assert(E<=d-1&&E>=0,()=>`GatherV2: the index value ${E} is not in [0, ${d-1}]`)}let h=C.segment_util.collectGatherOpShapeInfo(a,s,l,o),p=rn({inputs:{x:a},attrs:{shape:[h.batchSize,h.outerSize,h.dimSize,h.sliceSize]},backend:t}),c=v.sizeFromShape(s.shape),f=rn({inputs:{x:s},attrs:{shape:[h.batchSize,c/h.batchSize]},backend:t}),m=[h.batchSize,h.outerSize,c/h.batchSize,h.sliceSize],g=t.makeOutput(m,a.dtype);if(v.sizeFromShape(a.shape)===0)return g;let y=p.shape.length-1,A=t.dataIdMap.get(p.dataId).id,x=t.dataIdMap.get(f.dataId).id,b=t.dataIdMap.get(g.dataId).id,w=new Uint8Array(new Int32Array(v.computeStrides(p.shape)).buffer),S=new Uint8Array(new Int32Array(v.computeStrides(m)).buffer);return _S(A,Ut[a.dtype],w,y,x,h.batchSize,S,b),t.disposeData(p.dataId),t.disposeData(f.dataId),g.shape=h.outputShape,g}var X1e={kernelName:rl,backendName:"wasm",setupFunc:H1e,kernelFunc:q1e},K1e=!1,Z1e=Xr(al,K1e,"bool"),Y1e=!1,J1e=Xr(fi,Y1e,"bool"),zS;function Q1e(e){zS=e.wasm.cwrap(gi,null,["number","number","number","number"])}function e2e(e){let{inputs:{x:t},attrs:{alpha:r},backend:n}=e,a=n.dataIdMap.get(t.dataId).id,s=n.makeOutput(t.shape,"float32");if(v.sizeFromShape(t.shape)!==0){let i=n.dataIdMap.get(s.dataId).id;zS(a,Ut[t.dtype],r,i)}return s}var t2e={kernelName:gi,backendName:"wasm",setupFunc:Q1e,kernelFunc:e2e},r2e=!1,n2e=Xr(sl,r2e,"bool"),a2e=!1,s2e=Xr(il,a2e,"bool"),i2e=Ir(yi),o2e=!1,l2e=Xr(ol,o2e,"bool"),OS;function u2e(e){OS=e.wasm.cwrap(Ai,null,["number","number","number","number"])}function d2e(e){let{backend:t,inputs:r,attrs:n}=e,{reductionIndices:a,keepDims:s}=n,{x:i}=r,o=t.dataIdMap.get(i.dataId).id,l=i,{transposed:u,axes:d,originalAxes:h,inputWasTransposed:p}=Yi(i,a,t);if(p){let A=t.dataIdMap.get(u.dataId).id;l=u,o=A}let c=l.shape.length;C.assertAxesAreInnerMostDims("max",d,c);let[f,m]=C.computeOutAndReduceShapes(l.shape,d),g=v.sizeFromShape(m),y=t.makeOutput(f,i.dtype);if(v.sizeFromShape(l.shape)!==0){let A=t.dataIdMap.get(y.dataId).id;OS(o,Ut[i.dtype],g,A)}if(p&&t.disposeData(u.dataId),s){let A=C.expandShapeToKeepDim(y.shape,h);y.shape=A}return y}var p2e={kernelName:Ai,backendName:"wasm",setupFunc:u2e,kernelFunc:d2e},h2e=!1,c2e=Xr(xi,h2e),DS;function f2e(e){DS=e.wasm.cwrap(bi,null,["number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number"])}function m2e(e){let{inputs:t,attrs:r,backend:n}=e,a=t.x,s=n.dataIdMap.get(a.dataId).id;v.assert(a.dtype==="float32",()=>`Error in MaxPool: only float32 input is supported. Got ${a.dtype}.`);let{filterSize:i,strides:o,pad:l,dimRoundingMode:u}=r,d=C.computePool2DInfo(a.shape,i,o,1,l,u),h=d.filterHeight,p=d.filterWidth,c=d.padInfo.top,f=d.padInfo.right,m=d.padInfo.bottom,g=d.padInfo.left,y=d.dilationHeight,A=d.dilationWidth,x=d.strideHeight,b=d.strideWidth,w=d.inChannels,S=d.outChannels;if(d.dataFormat!=="channelsLast")throw new Error(`wasm backend does not support dataFormat:'${d.dataFormat}'. Please use 'channelsLast'.`);let N=n.makeOutput(d.outShape,"float32"),E=n.dataIdMap.get(N.dataId).id;return DS(s,a.shape[0],a.shape[1],a.shape[2],h,p,c,f,m,g,y,A,x,b,w,S,E),N}var g2e={kernelName:bi,backendName:"wasm",setupFunc:f2e,kernelFunc:m2e},LS;function y2e(e){LS=e.wasm.cwrap(vi,null,["number, number, number"])}function A2e(e){let{backend:t,inputs:r,attrs:n}=e,{axis:a,keepDims:s}=n,{x:i}=r,o=t.dataIdMap.get(i.dataId).id,l=o,u=i,{transposed:d,axes:h,originalAxes:p,inputWasTransposed:c}=Yi(i,a,t),f=h;if(c){let b=t.dataIdMap.get(d.dataId).id;b!==o&&(u=d,l=b,f=C.getInnerMostAxes(f.length,u.shape.length))}C.assertAxesAreInnerMostDims("mean",f,u.shape.length);let[m,g]=C.computeOutAndReduceShapes(u.shape,f),y=v.sizeFromShape(g),A=u;u.dtype!=="float32"&&(A=Xh({backend:t,inputs:{x:u},attrs:{dtype:"float32"}}),l=t.dataIdMap.get(A.dataId).id);let x=t.makeOutput(m,"float32");if(v.sizeFromShape(u.shape)!==0){let b=t.dataIdMap.get(x.dataId).id;LS(l,y,b)}if(c&&t.disposeData(d.dataId),s){let b=C.expandShapeToKeepDim(x.shape,p);x.shape=b}return u.dtype!=="float32"&&t.disposeData(A.dataId),x}var x2e={kernelName:vi,backendName:"wasm",setupFunc:y2e,kernelFunc:A2e},BS;function b2e(e){BS=e.wasm.cwrap(wi,null,["number","number","number","number"])}function v2e(e){let{backend:t,inputs:r,attrs:n}=e,{axis:a,keepDims:s}=n,{x:i}=r,o=t.dataIdMap.get(i.dataId).id,l=o,u=i,{transposed:d,axes:h,originalAxes:p,inputWasTransposed:c}=Yi(i,a,t);if(c){let x=t.dataIdMap.get(d.dataId).id;x!==o&&(u=d,l=x)}let f=u.shape.length;C.assertAxesAreInnerMostDims("min",h,f);let[m,g]=C.computeOutAndReduceShapes(u.shape,h),y=v.sizeFromShape(g),A=t.makeOutput(m,u.dtype);if(v.sizeFromShape(u.shape)!==0){let x=t.dataIdMap.get(A.dataId).id;BS(l,Ut[i.dtype],y,x)}if(c&&t.disposeData(d.dataId),s){let x=C.expandShapeToKeepDim(A.shape,p);A.shape=x}return A}var w2e={kernelName:wi,backendName:"wasm",setupFunc:b2e,kernelFunc:v2e},k2e=!1,I2e=Xr(ki,k2e),WS=(e=>(e[e.reflect=0]="reflect",e[e.symmetric=1]="symmetric",e))(WS||{}),VS;function S2e(e){VS=e.wasm.cwrap(Ii,null,["number","array","number","number","array","array","number","number"])}function C2e(e){let{inputs:{x:t},backend:r,attrs:{paddings:n,mode:a}}=e,s=n.map((f,m)=>f[0]+t.shape[m]+f[1]),i=r.dataIdMap.get(t.dataId).id,o=r.makeOutput(s,t.dtype),l=r.dataIdMap.get(o.dataId).id,u=new Uint8Array(new Int32Array(t.shape).buffer),d=n.map(f=>f[0]),h=n.map(f=>f[1]),p=new Uint8Array(new Int32Array(d).buffer),c=new Uint8Array(new Int32Array(h).buffer);return VS(i,u,t.shape.length,Ut[t.dtype],p,c,WS[a],l),o}var T2e={kernelName:Ii,backendName:"wasm",kernelFunc:C2e,setupFunc:S2e},N2e=!0,E2e=Xr(Si,N2e),R2e=Ir(ll);function F5(e,t){let r=new Int32Array(e.wasm.HEAPU8.buffer,t,4),n=r[0],a=r[1],s=r[2],i=r[3];return e.wasm._free(t),{pSelectedIndices:n,selectedSize:a,pSelectedScores:s,pValidOutputs:i}}var US;function $2e(e){US=e.wasm.cwrap(dl,"number",["number","number","number","number","number"])}function M2e(e){let{backend:t,inputs:r,attrs:n}=e,{iouThreshold:a,maxOutputSize:s,scoreThreshold:i}=n,{boxes:o,scores:l}=r,u=t.dataIdMap.get(o.dataId).id,d=t.dataIdMap.get(l.dataId).id,h=US(u,d,s,a,i),{pSelectedIndices:p,selectedSize:c,pSelectedScores:f,pValidOutputs:m}=F5(t,h);return t.wasm._free(f),t.wasm._free(m),t.makeOutput([c],"int32",p)}var F2e={kernelName:dl,backendName:"wasm",setupFunc:$2e,kernelFunc:M2e},GS;function P2e(e){GS=e.wasm.cwrap(rd,"number",["number","number","number","number","number","bool"])}function _2e(e){let{backend:t,inputs:r,attrs:n}=e,{iouThreshold:a,maxOutputSize:s,scoreThreshold:i,padToMaxOutputSize:o}=n,{boxes:l,scores:u}=r,d=t.dataIdMap.get(l.dataId).id,h=t.dataIdMap.get(u.dataId).id,p=GS(d,h,s,a,i,o),{pSelectedIndices:c,selectedSize:f,pSelectedScores:m,pValidOutputs:g}=F5(t,p);t.wasm._free(m);let y=t.makeOutput([f],"int32",c),A=t.makeOutput([],"int32",g);return[y,A]}var z2e={kernelName:rd,backendName:"wasm",setupFunc:P2e,kernelFunc:_2e},jS;function O2e(e){jS=e.wasm.cwrap(pl,"number",["number","number","number","number","number","number"])}function D2e(e){let{backend:t,inputs:r,attrs:n}=e,{iouThreshold:a,maxOutputSize:s,scoreThreshold:i,softNmsSigma:o}=n,{boxes:l,scores:u}=r,d=t.dataIdMap.get(l.dataId).id,h=t.dataIdMap.get(u.dataId).id,p=jS(d,h,s,a,i,o),{pSelectedIndices:c,selectedSize:f,pSelectedScores:m,pValidOutputs:g}=F5(t,p);t.wasm._free(g);let y=t.makeOutput([f],"int32",c),A=t.makeOutput([f],"float32",m);return[y,A]}var L2e={kernelName:pl,backendName:"wasm",setupFunc:O2e,kernelFunc:D2e},B2e=!1,W2e=Xr(ul,B2e,"bool"),HS;function V2e(e){HS=e.wasm.cwrap(cl,null,["number","number","number","number","number"])}function U2e(e){let{inputs:t,backend:r,attrs:n}=e,{indices:a}=t,{depth:s,onValue:i,offValue:o}=n,l=r.makeOutput([...a.shape,s],"int32"),u=r.dataIdMap.get(l.dataId).id,d=r.dataIdMap.get(a.dataId).id;return HS(d,s,i,o,u),l}var G2e={kernelName:cl,backendName:"wasm",setupFunc:V2e,kernelFunc:U2e};function j2e(e){let{inputs:{x:t},backend:r}=e,n=r.makeOutput(t.shape,t.dtype);return r.typedArrayFromHeap(n).fill(1),n}var H2e={kernelName:hl,backendName:"wasm",kernelFunc:j2e};function q2e(e){let{inputs:t,backend:r,attrs:n}=e,{axis:a}=n;if(t.length===1)return py({inputs:{input:t[0]},backend:r,attrs:{dim:a}});let s=t[0].shape,i=t[0].dtype;t.forEach(d=>{v.assertShapesMatch(s,d.shape,"All tensors passed to stack must have matching shapes"),v.assert(i===d.dtype,()=>"All tensors passed to stack must have matching dtypes")});let o=[],l=t.map(d=>{let h=py({inputs:{input:d},backend:r,attrs:{dim:a}});return o.push(h),h}),u=bS({inputs:l,backend:r,attrs:{axis:a}});return o.forEach(d=>r.disposeData(d.dataId)),u}var X2e={kernelName:fl,backendName:"wasm",kernelFunc:q2e},qS;function K2e(e){qS=e.wasm.cwrap(Ci,null,["number","array","number","number","array","array","number","number"])}function Z2e(e){let{inputs:{x:t},backend:r,attrs:{paddings:n,constantValue:a}}=e,s=n.map((f,m)=>f[0]+t.shape[m]+f[1]);if(v.sizeFromShape(t.shape)===0)return ES({backend:r,attrs:{shape:s,value:a,dtype:t.dtype}});let i=r.dataIdMap.get(t.dataId).id,o=r.makeOutput(s,t.dtype),l=r.dataIdMap.get(o.dataId).id,u=new Uint8Array(new Int32Array(t.shape).buffer),d=n.map(f=>f[0]),h=n.map(f=>f[1]),p=new Uint8Array(new Int32Array(d).buffer),c=new Uint8Array(new Int32Array(h).buffer);return qS(i,u,t.shape.length,Ut[t.dtype],p,c,a,l),o}var XS={kernelName:Ci,backendName:"wasm",kernelFunc:Z2e,setupFunc:K2e},Y2e=!1,J2e=Xr(Ti,Y2e),KS;function Q2e(e){KS=e.wasm.cwrap(Ni,null,["number","number","number"])}function ege(e){let{inputs:t,backend:r}=e,{x:n,alpha:a}=t,s=r.dataIdMap.get(n.dataId).id,i=r.dataIdMap.get(a.dataId).id,o=s,l=n,u=l;l.dtype!=="float32"&&(u=Xh({backend:r,inputs:{x:n},attrs:{dtype:"float32"}}),o=r.dataIdMap.get(u.dataId).id);let d=r.makeOutput(n.shape,"float32"),h=r.dataIdMap.get(d.dataId).id;return KS(o,i,h),l.dtype!=="float32"&&r.disposeData(u.dataId),d}var tge={kernelName:Ni,backendName:"wasm",setupFunc:Q2e,kernelFunc:ege},ZS;function rge(e){ZS=e.wasm.cwrap(Ei,null,["number","number","number","number"])}function nge(e){let{backend:t,inputs:r,attrs:n}=e,{axis:a,keepDims:s}=n,{x:i}=r,o=t.dataIdMap.get(i.dataId).id,l=o,u=i,{transposed:d,axes:h,originalAxes:p,inputWasTransposed:c}=Yi(i,a,t),f=h;if(c){let x=t.dataIdMap.get(d.dataId).id;x!==o&&(u=d,l=x,f=C.getInnerMostAxes(f.length,u.shape.length))}C.assertAxesAreInnerMostDims("prod",f,u.shape.length);let[m,g]=C.computeOutAndReduceShapes(u.shape,f),y=v.sizeFromShape(g),A=t.makeOutput(m,u.dtype);if(v.sizeFromShape(u.shape)!==0){let x=t.dataIdMap.get(A.dataId).id;ZS(l,y,Ut[A.dtype],x)}if(c&&t.disposeData(d.dataId),s){let x=C.expandShapeToKeepDim(A.shape,p);A.shape=x}return A}var age={kernelName:Ei,backendName:"wasm",setupFunc:rge,kernelFunc:nge},sge=e=>{let{backend:t,attrs:r}=e,{start:n,stop:a,step:s,dtype:i}=r,o=s5(n,a,s,i),l=t.makeOutput([o.length],i);return t.typedArrayFromHeap(l).set(o),l},ige={kernelName:nd,backendName:"wasm",kernelFunc:sge},oge=!0,lge=Xr(li,oge),uge=Ir(Ri),dge=Ir(Mi),YS;function pge(e){YS=e.wasm.cwrap($i,null,["number","number","number","number","number","number","number","number","number","number"])}function hge(e){let{backend:t,inputs:r,attrs:n}=e,{images:a}=r,{alignCorners:s,halfPixelCenters:i,size:o}=n,[l,u]=o,[d,h,p,c]=a.shape,f=[d,l,u,c],m=t.dataIdMap.get(a.dataId),g;m.dtype!=="float32"&&(g=Xh({backend:t,inputs:{x:a},attrs:{dtype:"float32"}}),m=t.dataIdMap.get(g.dataId));let y=m.id,A=t.makeOutput(f,"float32");if(v.sizeFromShape(a.shape)===0)return A;let x=t.dataIdMap.get(A.dataId).id;return YS(y,d,h,p,c,l,u,s?1:0,i?1:0,x),g!=null&&t.disposeData(g.dataId),A}var cge={kernelName:$i,backendName:"wasm",setupFunc:pge,kernelFunc:hge},JS;function fge(e){JS=e.wasm.cwrap(gl,null,["number","array","number","array","number","number"])}function mge(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{dims:s}=n,i=v.parseAxisParam(s,a.shape);if(a.shape.length===0)return Fm({inputs:{x:a},backend:r});let o=r.makeOutput(a.shape,a.dtype),l=r.dataIdMap.get(a.dataId).id,u=r.dataIdMap.get(o.dataId).id,d=new Uint8Array(new Int32Array(i).buffer),h=new Uint8Array(new Int32Array(a.shape).buffer);JS(l,d,i.length,h,a.shape.length,u);let p=rn({inputs:{x:o},attrs:{shape:a.shape},backend:r});return r.disposeData(o.dataId),p}var gge={kernelName:gl,backendName:"wasm",kernelFunc:mge,setupFunc:fge},QS;function yge(e){QS=e.wasm.cwrap(Rl,null,["number","number","number","number","number","number","number","number","array","number","number"])}function Age(e){let{inputs:t,backend:r,attrs:n}=e,{image:a}=t,{radians:s,fillValue:i,center:o}=n,l=r.makeOutput(a.shape,a.dtype),u=r.dataIdMap.get(a.dataId).id,d=r.dataIdMap.get(l.dataId).id,[h,p,c,f]=a.shape,[m,g]=C.getImageCenter(o,p,c),y=i===0,A=255,x=typeof i=="number"?[i,i,i,y?0:A]:[...i,A],b=new Uint8Array(new Int32Array(x).buffer);return QS(u,h,p,c,f,s,m,g,b,x.length,d),l}var xge={kernelName:Rl,backendName:"wasm",kernelFunc:Age,setupFunc:yge},bge=Ir(yl),vge=Ir(Fi),eC;function wge(e){eC=e.wasm.cwrap(Al,null,["number","number","number","number","number","number","array","number","number"])}function kge(e){let{backend:t,inputs:r,attrs:n}=e,{indices:a,updates:s}=r,{shape:i}=n,o=t.makeOutput(i,s.dtype);if(v.sizeFromShape(i)===0)return o;let{sliceRank:l,numUpdates:u,sliceSize:d,strides:h,outputSize:p}=Ny.calculateShapes(s,a,i),c=t.dataIdMap.get(a.dataId).id,f=t.dataIdMap.get(s.dataId).id,m=new Uint8Array(new Int32Array(h).buffer),g=t.dataIdMap.get(o.dataId).id;return eC(c,f,Ut[s.dtype],l,u,d,m,p,g),o}var Ige={kernelName:Al,backendName:"wasm",setupFunc:wge,kernelFunc:kge},tC;function Sge(e){tC=e.wasm.cwrap("SelectV2",null,["number","number","number","number","number"])}function Cge(e){let{inputs:t,backend:r}=e,{condition:n,t:a,e:s}=t,i=r.dataIdMap.get(n.dataId).id,o=r.dataIdMap.get(a.dataId).id,l=r.dataIdMap.get(s.dataId).id,u=r.makeOutput(a.shape,a.dtype),d=r.dataIdMap.get(u.dataId).id,h=n.shape.length,p=a.shape.length,c=h===0||h>1||p===1?1:v.sizeFromShape(a.shape.slice(1));return tC(i,o,l,c,d),u}var Tge={kernelName:xl,backendName:"wasm",kernelFunc:Cge,setupFunc:Sge},rC;function Nge(e){rC=e.wasm.cwrap(_i,null,["number","number"])}function Ege(e){let{backend:t,inputs:{x:r}}=e,n=t.dataIdMap.get(r.dataId).id,a=t.makeOutput(r.shape,r.dtype),s=t.dataIdMap.get(a.dataId).id;return v.sizeFromShape(a.shape)===0||rC(n,s),a}var Rge={kernelName:"Sigmoid",backendName:"wasm",setupFunc:Nge,kernelFunc:Ege},$ge=Ir(Pi),nC;function Mge(e){nC=e.wasm.cwrap(Di,null,["number","number","number","number"])}function Fge(e){let{backend:t,inputs:{logits:r},attrs:{dim:n}}=e,a=t.dataIdMap.get(r.dataId).id,s=t.makeOutput(r.shape,r.dtype),i=t.dataIdMap.get(s.dataId).id,o=r.shape[n],l=v.sizeFromShape(r.shape)/o;return v.sizeFromShape(s.shape)===0||nC(a,i,o,l),s}var Pge={kernelName:Di,backendName:"wasm",setupFunc:Mge,kernelFunc:Fge};function _ge(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,{blockShape:s,paddings:i}=n,o=v.sizeFromShape(s),l=[[0,0]];l.push(...i);for(let g=1+s.length;g<a.shape.length;++g)l.push([0,0]);let u=XS.kernelFunc({inputs:{x:a},backend:r,attrs:{paddings:l,constantValue:0}}),d=C.getReshaped(u.shape,s,o,!1),h=C.getPermuted(d.length,s.length,!1),p=C.getReshapedPermuted(u.shape,s,o,!1),c=rn({inputs:{x:u},backend:r,attrs:{shape:d}}),f=Xs({inputs:{x:c},backend:r,attrs:{perm:h}}),m=rn({inputs:{x:f},backend:r,attrs:{shape:p}});return r.disposeData(u.dataId),r.disposeData(c.dataId),r.disposeData(f.dataId),m}var zge={kernelName:wl,backendName:"wasm",kernelFunc:_ge},aC;function Oge(e){aC=e.wasm.cwrap("SparseFillEmptyRows","number",["number","number","number","number","number","number","number","number","number","number","number","number"])}function Dge(e){let{backend:t,inputs:r}=e,{indices:n,values:a,denseShape:s,defaultValue:i}=r,o=n.shape[0],l=n.shape[1],u=t.readSync(s.dataId)[0],d=[o+u,l],h=t.dataIdMap.get(n.dataId).id,p=t.dataIdMap.get(a.dataId).id,c=t.dataIdMap.get(i.dataId).id,f=t.makeOutput(d,n.dtype),m=t.dataIdMap.get(f.dataId).id,g=t.makeOutput(d.slice(0,1),a.dtype),y=t.dataIdMap.get(g.dataId).id,A=t.makeOutput([u],"bool"),x=t.dataIdMap.get(A.dataId).id,b=t.makeOutput([o],n.dtype),w=t.dataIdMap.get(b.dataId).id,S=t.makeOutput([4],"int32"),N=t.dataIdMap.get(S.dataId).id,E=aC(h,p,Ut[a.dtype],o,u,l,c,m,y,x,w,N),R=t.readSync(S.dataId),z;switch(R[0]){case 1:{z=C.getSparseFillEmptyRowsIndicesDenseShapeMismatch(R[1]);break}case 2:{z=C.getSparseFillEmptyRowsNegativeIndexErrorMessage(R[1],R[2]);break}case 3:z=C.getSparseFillEmptyRowsOutOfRangeIndexErrorMessage(R[1],R[2],R[3]);break;default:z=""}if(t.disposeData(S.dataId),z)throw t.disposeData(f.dataId),t.disposeData(g.dataId),t.disposeData(A.dataId),t.disposeData(b.dataId),new Error(z);let M=f,I=g;return E!==d[0]&&(M=Uo({inputs:{x:f},attrs:{begin:0,size:[E,l]},backend:t}),I=Uo({inputs:{x:g},attrs:{begin:0,size:E},backend:t}),t.disposeData(f.dataId),t.disposeData(g.dataId)),[M,I,A,b]}var Lge={kernelName:lh,backendName:"wasm",setupFunc:Oge,kernelFunc:Dge},sC;function Bge(e){sC=e.wasm.cwrap(ud,null,["number","number","number","number","number","number","number"])}function Wge(e){let{backend:t,inputs:r}=e,{inputIndices:n,inputShape:a,newShape:s}=r;if(n.shape.length!==2)throw new Error(`Input indices should be a matrix but received shape
|
|
${n.shape}`);if(a.shape.length!==1)throw new Error(`Input shape should be a vector but received shape
|
|
${a.shape}`);if(s.shape.length!==1)throw new Error(`Target shape should be a vector but received shape ${s.shape}`);let i=t.dataIdMap.get(n.dataId).id,o=t.dataIdMap.get(a.dataId).id,l=t.dataIdMap.get(s.dataId).id,u=n.shape[0],d=v.sizeFromShape(s.shape),h=t.makeOutput([u,d],n.dtype),p=t.dataIdMap.get(h.dataId).id,c=t.makeOutput([d],s.dtype),f=t.dataIdMap.get(c.dataId).id,m=t.makeOutput([3],"int32"),g=t.dataIdMap.get(m.dataId).id;sC(i,o,l,u,p,f,g);let y=t.readSync(m.dataId),A;switch(y[0]){case 0:{A=C.getSparseReshapeMultipleNegativeOneOutputDimErrorMessage(y[1],y[2]);break}case 1:{A=C.getSparseReshapeNegativeOutputDimErrorMessage(y[1],y[2]);break}case 2:A=C.getSparseReshapeEmptyTensorZeroOutputDimErrorMessage();break;case 3:{let x=Array.from(t.readSync(a.dataId)),b=Array.from(t.readSync(c.dataId));A=C.getSparseReshapeInputOutputMultipleErrorMessage(x,b);break}case 4:{let x=Array.from(t.readSync(a.dataId)),b=Array.from(t.readSync(c.dataId));A=C.getSparseReshapeInputOutputMismatchErrorMessage(x,b);break}default:A=""}if(t.disposeData(m.dataId),A)throw t.disposeData(h.dataId),t.disposeData(c.dataId),new Error(A);return[h,c]}var Vge={kernelName:ud,backendName:"wasm",setupFunc:Bge,kernelFunc:Wge},iC;function oC(e){iC=e.wasm.cwrap("SparseSegmentReduction",null,["number","number","number","number","number","number","number","number","number"])}function lC(e,t){let{backend:r,inputs:n}=e,{data:a,indices:s,segmentIds:i}=n,o=s.shape[0],l=r.readSync(i.dataId,o-1,o)[0],u=o>0?l+1:0;if(u<0)throw new Error(C.getSparseSegmentReductionNegativeSegmentIdsErrorMessage());let d=a.shape.slice();d[0]=u;let h=r.dataIdMap.get(a.dataId).id,p=r.dataIdMap.get(s.dataId).id,c=r.dataIdMap.get(i.dataId).id,f=r.makeOutput(d,a.dtype),m=r.dataIdMap.get(f.dataId).id,g=r.makeOutput([4],"int32"),y=r.dataIdMap.get(g.dataId).id;iC(h,Ut[a.dtype],a.shape[0],p,c,m,y,t,0);let A=r.readSync(g.dataId),x;switch(A[0]){case 0:{x=C.getSparseSegmentReductionNegativeSegmentIdsErrorMessage();break}case 1:{x=C.getSparseSegmentReductionNonIncreasingSegmentIdsErrorMessage();break}case 2:x=C.getSparseSegmentReductionSegmentIdOutOfRangeErrorMessage(A[1],A[2]);break;case 3:x=C.getSparseSegmentReductionIndicesOutOfRangeErrorMessage(A[1],A[2],A[3]);break;default:x=""}if(r.disposeData(g.dataId),x)throw r.disposeData(f.dataId),new Error(x);return f}function Uge(e){return lC(e,!0)}var Gge={kernelName:uh,backendName:"wasm",setupFunc:oC,kernelFunc:Uge};function jge(e){return lC(e,!1)}var Hge={kernelName:dh,backendName:"wasm",setupFunc:oC,kernelFunc:jge};function qge(e){let{inputs:t,attrs:r,backend:n}=e,{x:a}=t,{numOrSizeSplits:s,axis:i}=r,o=v.parseAxisParam(i,a.shape)[0],l=C.prepareSplitSize(a,s,o),u=new Array(a.shape.length).fill(0),d=a.shape.slice();return l.map(h=>{let p=[...d];p[o]=h;let c=Uo({inputs:{x:a},attrs:{begin:u,size:p},backend:n});return u[o]+=h,c})}var Xge={kernelName:kl,backendName:"wasm",kernelFunc:qge},Kge=Ir(zi),Zge=Ir(dd),Yge=!0,Jge=Xr(Li,Yge),uC;function Qge(e){uC=e.wasm.cwrap(Ui,null,["number","number","number","number"])}function eye(e){let{backend:t,inputs:r,attrs:n}=e,{alpha:a}=n,{x:s}=r,i=t.dataIdMap.get(s.dataId).id,o=t.makeOutput(s.shape,s.dtype),l=t.dataIdMap.get(o.dataId).id;return uC(i,a,Ut[s.dtype],l),o}var tye={kernelName:Ui,backendName:"wasm",setupFunc:Qge,kernelFunc:eye},dC;function rye(e){dC=e.wasm.cwrap(Il,null,["number","array","number","array","array","array","array","array","number","number"])}function nye(e){let{backend:t,inputs:r,attrs:n}=e,{x:a}=r,{begin:s,end:i,strides:o,beginMask:l,endMask:u,ellipsisMask:d,newAxisMask:h,shrinkAxisMask:p}=n,{finalShapeSparse:c,finalShape:f,isIdentity:m,sliceDim0:g,isSimpleSlice:y,begin:A,end:x,strides:b}=Ot.sliceInfo(a.shape,s,i,o,l,u,d,h,p),w;if(m)w=rn({inputs:{x:a},backend:t,attrs:{shape:f}});else if(g||y){v.assert(a.shape.length>=1,()=>`Input must have rank at least 1, got: ${a.shape.length}`);let S=Ot.computeOutShape(A,x,b),N=Uo({inputs:{x:a},backend:t,attrs:{begin:A,size:S}});w=rn({inputs:{x:N},backend:t,attrs:{shape:f}}),t.disposeData(N.dataId)}else{let S=t.makeOutput(c,"float32"),N=t.dataIdMap.get(a.dataId).id,E=new Uint8Array(new Int32Array(v.computeStrides(a.shape)).buffer),R=new Uint8Array(new Int32Array(A).buffer),z=new Uint8Array(new Int32Array(x).buffer),M=new Uint8Array(new Int32Array(b).buffer),I=new Uint8Array(new Int32Array(c).buffer),_=new Uint8Array(new Int32Array(v.computeStrides(c)).buffer),O=t.dataIdMap.get(S.dataId).id;dC(N,E,a.shape.length,R,z,M,I,_,c.length,O),w=rn({inputs:{x:S},backend:t,attrs:{shape:f}}),t.disposeData(S.dataId)}return w}var aye={kernelName:Il,backendName:"wasm",setupFunc:rye,kernelFunc:nye},sye=!0,iye=Xr(Bi,sye),pC;function oye(e){pC=e.wasm.cwrap(Oi,null,["number","number","number","number"])}function lye(e){let{backend:t,inputs:r,attrs:n}=e,{axis:a,keepDims:s}=n,{x:i}=r,o=t.dataIdMap.get(i.dataId).id,l=o,u=i,{transposed:d,axes:h,originalAxes:p,inputWasTransposed:c}=Yi(i,a,t),f=h;if(c){let x=t.dataIdMap.get(d.dataId).id;x!==o&&(u=d,l=x,f=C.getInnerMostAxes(f.length,u.shape.length))}C.assertAxesAreInnerMostDims("sum",f,u.shape.length);let[m,g]=C.computeOutAndReduceShapes(u.shape,f),y=v.sizeFromShape(g),A=t.makeOutput(m,u.dtype);if(v.sizeFromShape(u.shape)!==0){let x=t.dataIdMap.get(A.dataId).id;pC(l,y,Ut[A.dtype],x)}if(c&&t.disposeData(d.dataId),s){let x=C.expandShapeToKeepDim(A.shape,p);A.shape=x}return A}var uye={kernelName:Oi,backendName:"wasm",setupFunc:oye,kernelFunc:lye},dye=Ir(Sl),pye=Ir(Wi),hC;function hye(e){hC=e.wasm.cwrap(es,null,["number","array","number","array","number","number"])}function cye(e){let{inputs:t,backend:r,attrs:n}=e,{x:a}=t,s=r.dataIdMap.get(a.dataId).id,{reps:i}=n,o=new Array(a.shape.length);for(let p=0;p<o.length;p++)o[p]=a.shape[p]*i[p];let l=new Uint8Array(new Int32Array(a.shape).buffer),u=new Uint8Array(new Int32Array(o).buffer),d=r.makeOutput(o,a.dtype),h=r.dataIdMap.get(d.dataId).id;return hC(s,l,a.shape.length,u,o.length,Ut[d.dtype],h),d}var fye={kernelName:es,backendName:"wasm",setupFunc:hye,kernelFunc:cye},cC;function mye(e){cC=e.wasm.cwrap(Cl,null,["number","array","number","number","number","bool","number","number"])}var gye=({inputs:e,backend:t,attrs:r})=>{let{x:n}=e,{k:a,sorted:s}=r,i=t.dataIdMap.get(n.dataId).id,o=new Uint8Array(new Int32Array(n.shape).buffer),l=n.shape.slice();l[l.length-1]=a;let u=t.makeOutput(l,n.dtype),d=t.dataIdMap.get(u.dataId).id,h=t.makeOutput(l,"int32"),p=t.dataIdMap.get(h.dataId).id;return cC(i,o,n.shape.length,Ut[n.dtype],a,s,d,p),[u,h]},yye={kernelName:Cl,backendName:"wasm",setupFunc:mye,kernelFunc:gye},fC;function Aye(e){fC=e.wasm.cwrap(Tl,null,["number","number","bool","number","number","number","number","number","number","array","number","number","number","number","number"])}function xye(e){let{backend:t,inputs:r,attrs:n}=e,{image:a,transforms:s}=r,{interpolation:i,fillMode:o,fillValue:l,outputShape:u}=n,[d,h,p,c]=a.shape,[f,m]=u!=null?u:[h,p],g=[d,f,m,c],y=new Uint8Array(new Int32Array(v.computeStrides(a.shape)).buffer),A=t.makeOutput(g,a.dtype),x=t.dataIdMap.get(A.dataId).id,b=t.dataIdMap.get(a.dataId).id,w=t.dataIdMap.get(s.dataId).id,S=i==="nearest"?1:2,N;switch(o){case"constant":N=1;break;case"reflect":N=2;break;case"wrap":N=3;break;case"nearest":N=4;break;default:N=1;break}return fC(b,w,s.shape[0]>1,d,f,m,c,p,h,y,a.shape.length-1,S,N,l,x),A}var bye={kernelName:Tl,backendName:"wasm",setupFunc:Aye,kernelFunc:xye};function vye(e){let{inputs:t,backend:r,attrs:n}=e,{value:a}=t,{axis:s}=n;s<0&&(s+=a.shape.length);let i=a.shape[s],o=a.shape.length,l=new Array(o-1),u=0;for(let c=0;c<o;c++)c!==s&&(l[u++]=a.shape[c]);let d=new Array(i),h=new Array(o).fill(0),p=a.shape.slice();p[s]=1;for(let c=0;c<d.length;c++)h[s]=c,d[c]=Uo({inputs:{x:a},attrs:{begin:h,size:p},backend:r});return d.map(({dataId:c,dtype:f})=>({dataId:c,dtype:f,shape:l}))}var wye={kernelName:Nl,backendName:"wasm",kernelFunc:vye};function kye(e){let{inputs:{x:t},backend:r}=e,n=r.makeOutput(t.shape,t.dtype);return r.typedArrayFromHeap(n).fill(0),n}var Iye={kernelName:El,backendName:"wasm",kernelFunc:kye},Sye=[cme,fme,gme,xme,Tme,Rme,Fme,zme,Bme,Hme,qme,Xme,Yme,Jme,t1e,a1e,s1e,i1e,u1e,h1e,m1e,A1e,v1e,w1e,I1e,S1e,C1e,T1e,R1e,$1e,F1e,z1e,L1e,V1e,j1e,X1e,Z1e,J1e,bme,t2e,n2e,s2e,i2e,l2e,p2e,c2e,g2e,x2e,w2e,I2e,T2e,E2e,R2e,F2e,z2e,L2e,W2e,G2e,H2e,X2e,XS,J2e,tge,age,ige,lge,uge,dge,Ome,cge,gge,xge,bge,vge,Ige,Tge,Rge,$ge,Gme,Pge,zge,Lge,Vge,Gge,Hge,Xge,Kge,Zge,Jge,tye,aye,iye,uye,dye,pye,fye,yye,bye,Ime,wye,Iye];for(let e of Sye)qn(e);var hy=Z();hy.registerFlag("WASM_HAS_SIMD_SUPPORT",async()=>WebAssembly.validate(new Uint8Array([0,97,115,109,1,0,0,0,1,4,1,96,0,0,3,2,1,0,10,9,1,7,0,65,0,253,15,26,11])));hy.registerFlag("WASM_HAS_MULTITHREAD_SUPPORT",async()=>{if(hy.get("IS_NODE"))return!1;try{return new MessageChannel().port1.postMessage(new SharedArrayBuffer(1)),WebAssembly.validate(new Uint8Array([0,97,115,109,1,0,0,0,1,4,1,96,0,0,3,2,1,0,5,4,1,3,1,1,10,11,1,9,0,65,0,254,16,2,0,26,11]))}catch(e){return!1}});var wv=Go(yR()),Cye=`"use strict";var Module={};var ENVIRONMENT_IS_NODE=typeof process==="object"&&typeof process.versions==="object"&&typeof process.versions.node==="string";if(ENVIRONMENT_IS_NODE){var nodeWorkerThreads=require("worker_threads");var parentPort=nodeWorkerThreads.parentPort;parentPort.on("message",function(data){onmessage({data:data})});var fs=require("fs");Object.assign(global,{self:global,require:require,Module:Module,location:{href:__filename},Worker:nodeWorkerThreads.Worker,importScripts:function(f){(0,eval)(fs.readFileSync(f,"utf8"))},postMessage:function(msg){parentPort.postMessage(msg)},performance:global.performance||{now:function(){return Date.now()}}})}function threadPrintErr(){var text=Array.prototype.slice.call(arguments).join(" ");if(ENVIRONMENT_IS_NODE){fs.writeSync(2,text+"
|
|
");return}console.error(text)}function threadAlert(){var text=Array.prototype.slice.call(arguments).join(" ");postMessage({cmd:"alert",text:text,threadId:Module["_pthread_self"]()})}var err=threadPrintErr;self.alert=threadAlert;Module["instantiateWasm"]=((info,receiveInstance)=>{var instance=new WebAssembly.Instance(Module["wasmModule"],info);receiveInstance(instance);Module["wasmModule"]=null;return instance.exports});self.onmessage=(e=>{try{if(e.data.cmd==="load"){Module["wasmModule"]=e.data.wasmModule;Module["wasmMemory"]=e.data.wasmMemory;Module["buffer"]=Module["wasmMemory"].buffer;Module["ENVIRONMENT_IS_PTHREAD"]=true;if(typeof e.data.urlOrBlob==="string"){importScripts(e.data.urlOrBlob)}else{var objectUrl=URL.createObjectURL(e.data.urlOrBlob);importScripts(objectUrl);URL.revokeObjectURL(objectUrl)}WasmBackendModuleThreadedSimd(Module).then(function(instance){Module=instance})}else if(e.data.cmd==="run"){Module["__performance_now_clock_drift"]=performance.now()-e.data.time;Module["__emscripten_thread_init"](e.data.threadInfoStruct,0,0,1);Module["establishStackSpace"]();Module["PThread"].receiveObjectTransfer(e.data);Module["PThread"].threadInit();try{var result=Module["invokeEntryPoint"](e.data.start_routine,e.data.arg);if(Module["keepRuntimeAlive"]()){Module["PThread"].setExitStatus(result)}else{Module["__emscripten_thread_exit"](result)}}catch(ex){if(ex!="unwind"){if(ex instanceof Module["ExitStatus"]){if(Module["keepRuntimeAlive"]()){}else{Module["__emscripten_thread_exit"](ex.status)}}else{throw ex}}}}else if(e.data.cmd==="cancel"){if(Module["_pthread_self"]()){Module["__emscripten_thread_exit"](-1)}}else if(e.data.target==="setimmediate"){}else if(e.data.cmd==="processThreadQueue"){if(Module["_pthread_self"]()){Module["_emscripten_current_thread_process_queued_calls"]()}}else if(e.data.cmd==="processProxyingQueue"){if(Module["_pthread_self"]()){Module["_emscripten_proxy_execute_queue"](e.data.queue)}}else{err("worker.js received unknown command "+e.data.cmd);err(e.data)}}catch(ex){err("worker.js onmessage() captured an uncaught exception: "+ex);if(ex&&ex.stack)err(ex.stack);if(Module["__emscripten_thread_crashed"]){Module["__emscripten_thread_crashed"]()}throw ex}});`,Tye=Go(AR()),mC=class extends zu{constructor(e){super(),this.wasm=e,this.dataIdNextNumber=1,this.wasm.tfjs.initWithThreadsCount(gC),cy=this.wasm.tfjs.getThreadsCount(),this.dataIdMap=new Zp(this,Xt())}write(e,t,r){let n={id:this.dataIdNextNumber++};return this.move(n,e,t,r,1),n}numDataIds(){return this.dataIdMap.numDataIds()}async time(e){let t=v.now();return e(),{kernelMs:v.now()-t}}move(e,t,r,n,a){let s=this.dataIdNextNumber++;if(n==="string"){let u=t;this.dataIdMap.set(e,{id:s,stringBytes:u,shape:r,dtype:n,memoryOffset:null,refCount:a});return}let i=v.sizeFromShape(r),o=i*v.bytesPerElement(n),l=this.wasm._malloc(o);this.dataIdMap.set(e,{id:s,memoryOffset:l,shape:r,dtype:n,refCount:a}),this.wasm.tfjs.registerTensor(s,i,l),t!=null&&this.wasm.HEAPU8.set(new Uint8Array(t.buffer,t.byteOffset,o),l)}async read(e){return this.readSync(e)}readSync(e,t,r){let{memoryOffset:n,dtype:a,shape:s,stringBytes:i}=this.dataIdMap.get(e);if(a==="string")return(t==null||t===0)&&(r==null||r>=i.length)?i:i.slice(t,r);t=t||0,r=r||v.sizeFromShape(s);let o=v.bytesPerElement(a),l=this.wasm.HEAPU8.slice(n+t*o,n+r*o);return Rye(l.buffer,a)}disposeData(e,t=!1){if(this.dataIdMap.has(e)){let r=this.dataIdMap.get(e);if(r.refCount--,!t&&r.refCount>0)return!1;this.wasm._free(r.memoryOffset),this.wasm.tfjs.disposeData(r.id),this.dataIdMap.delete(e)}return!0}refCount(e){return this.dataIdMap.has(e)?this.dataIdMap.get(e).refCount:0}incRef(e){let t=this.dataIdMap.get(e);t!=null&&t.refCount++}floatPrecision(){return 32}getMemoryOffset(e){return this.dataIdMap.get(e).memoryOffset}dispose(){this.wasm.tfjs.dispose(),"PThread"in this.wasm&&this.wasm.PThread.terminateAllThreads(),this.wasm=null}memory(){return{unreliable:!1}}makeOutput(e,t,r){let n;if(r==null)n=this.write(null,e,t);else{let a=this.dataIdNextNumber++;n={id:a},this.dataIdMap.set(n,{id:a,memoryOffset:r,shape:e,dtype:t,refCount:1});let s=v.sizeFromShape(e);this.wasm.tfjs.registerTensor(a,s,r)}return{dataId:n,shape:e,dtype:t}}typedArrayFromHeap({shape:e,dtype:t,dataId:r}){let n=this.wasm.HEAPU8.buffer,{memoryOffset:a}=this.dataIdMap.get(r),s=v.sizeFromShape(e);switch(t){case"float32":return new Float32Array(n,a,s);case"int32":return new Int32Array(n,a,s);case"bool":return new Uint8Array(n,a,s);default:throw new Error(`Unknown dtype ${t}`)}}};function Nye(e){return(t,r)=>(v.fetch(e,{credentials:"same-origin"}).then(n=>{n.ok||t.env.a(`failed to load wasm binary file at '${e}'`),n.arrayBuffer().then(a=>{WebAssembly.instantiate(a,t).then(s=>{r(s.instance,s.module)})})}),{})}function kv(e,t,r){if(q0!=null)return q0;let n="tfjs-backend-wasm.wasm";return e&&t?n="tfjs-backend-wasm-threaded-simd.wasm":e&&(n="tfjs-backend-wasm-simd.wasm"),Pp!=null&&Pp[n]!=null?Pp[n]:r+n}async function Eye(){let[e,t]=await Promise.all([Z().getAsync("WASM_HAS_SIMD_SUPPORT"),Z().getAsync("WASM_HAS_MULTITHREAD_SUPPORT")]);return new Promise((r,n)=>{let a={};a.locateFile=(o,l)=>{if(o.endsWith(".worker.js")){let u=Cye.replace(/\n/g,"\\n"),d=new Blob([u],{type:"application/javascript"});return URL.createObjectURL(d)}return o.endsWith(".wasm")?kv(e,t,Rp!=null?Rp:l):l+o},P5&&(a.instantiateWasm=Nye(kv(e,t,Rp!=null?Rp:"")));let s=!1;a.onAbort=()=>{s||_p||(_p=!0,n({message:"Make sure the server can serve the `.wasm` file relative to the bundled js file. For more details see https://github.com/tensorflow/tfjs/blob/master/tfjs-backend-wasm/README.md#using-bundlers"}))};let i;t&&e&&q0==null?(a.mainScriptUrlOrBlob=new Blob(["var WasmBackendModuleThreadedSimd = "+wv.default.toString()],{type:"text/javascript"}),i=(0,wv.default)(a)):i=(0,Tye.default)(a),i.then(o=>{s=!0,_p=!1;let l=null;o.tfjs={init:o.cwrap("init",null,[]),initWithThreadsCount:o.cwrap("init_with_threads_count",null,["number"]),getThreadsCount:o.cwrap("get_threads_count","number",[]),registerTensor:o.cwrap("register_tensor",null,["number","number","number"]),disposeData:o.cwrap("dispose_data",l,["number"]),dispose:o.cwrap("dispose",l,[])},r({wasm:o})})})}function Rye(e,t){switch(t){case"float32":return new Float32Array(e);case"int32":return new Int32Array(e);case"bool":return new Uint8Array(e);default:throw new Error(`Unknown dtype ${t}`)}}var $ye=["tfjs-backend-wasm.wasm","tfjs-backend-wasm-simd.wasm","tfjs-backend-wasm-threaded-simd.wasm"],q0=null,Rp=null,Pp={},_p=!1,P5=!1;function Mye(e,t=!1){if(_y("setWasmPath has been deprecated in favor of setWasmPaths and will be removed in a future release."),_p)throw new Error("The WASM backend was already initialized. Make sure you call `setWasmPath()` before you call `tf.setBackend()` or `tf.ready()`");q0=e,P5=t}function _5(e,t=!1){if(_p)throw new Error("The WASM backend was already initialized. Make sure you call `setWasmPaths()` before you call `tf.setBackend()` or `tf.ready()`");if(typeof e=="string")Rp=e;else{Pp=e;let r=$ye.filter(n=>Pp[n]==null);if(r.length>0)throw new Error(`There were no entries found for the following binaries: ${r.join(",")}. Please either call setWasmPaths with a map providing a path for each binary, or with a string indicating the directory where all the binaries can be found.`)}P5=t}var gC=-1,cy=-1;function Fye(e){gC=e}function Pye(){if(cy===-1)throw new Error("WASM backend not initialized.");return cy}var _ye="0.0.0",zye=2;$l("wasm",async()=>{let{wasm:e}=await Eye();return new mC(e)},zye);var ks="3.16.0-20220509",Kh={tfjs:ks,"tfjs-core":ks,"tfjs-data":ks,"tfjs-layers":ks,"tfjs-converter":ks,"tfjs-backend-cpu":ks,"tfjs-backend-webgl":ks,"tfjs-backend-wasm":ks};var yC=`
|
|
precision highp float;
|
|
attribute vec2 pos;
|
|
attribute vec2 uv;
|
|
varying vec2 vUv;
|
|
uniform float flipY;
|
|
void main(void) {
|
|
vUv = uv;
|
|
gl_Position = vec4(pos.x, pos.y*flipY, 0.0, 1.);
|
|
}
|
|
`;var AC=`
|
|
precision highp float;
|
|
varying vec2 vUv;
|
|
uniform sampler2D texture;
|
|
uniform float m[20];
|
|
void main(void) {
|
|
vec4 c = texture2D(texture, vUv);
|
|
gl_FragColor.r = m[0] * c.r + m[1] * c.g + m[2] * c.b + m[3] * c.a + m[4];
|
|
gl_FragColor.g = m[5] * c.r + m[6] * c.g + m[7] * c.b + m[8] * c.a + m[9];
|
|
gl_FragColor.b = m[10] * c.r + m[11] * c.g + m[12] * c.b + m[13] * c.a + m[14];
|
|
gl_FragColor.a = m[15] * c.r + m[16] * c.g + m[17] * c.b + m[18] * c.a + m[19];
|
|
}
|
|
`,xC=`
|
|
precision highp float;
|
|
varying vec2 vUv;
|
|
uniform sampler2D texture;
|
|
uniform float m[20];
|
|
void main(void) {
|
|
vec4 c = texture2D(texture, vUv);
|
|
gl_FragColor.r = m[0] * c.r + m[1] * c.g + m[2] * c.b + m[4];
|
|
gl_FragColor.g = m[5] * c.r + m[6] * c.g + m[7] * c.b + m[9];
|
|
gl_FragColor.b = m[10] * c.r + m[11] * c.g + m[12] * c.b + m[14];
|
|
gl_FragColor.a = c.a;
|
|
}
|
|
`,bC=`
|
|
precision highp float;
|
|
varying vec2 vUv;
|
|
uniform vec2 size;
|
|
uniform sampler2D texture;
|
|
vec2 pixelate(vec2 coord, vec2 size) {
|
|
return floor( coord / size ) * size;
|
|
}
|
|
void main(void) {
|
|
gl_FragColor = vec4(0.0);
|
|
vec2 coord = pixelate(vUv, size);
|
|
gl_FragColor += texture2D(texture, coord);
|
|
}
|
|
`,vC=`
|
|
precision highp float;
|
|
varying vec2 vUv;
|
|
uniform sampler2D texture;
|
|
uniform vec2 px;
|
|
void main(void) {
|
|
gl_FragColor = vec4(0.0);
|
|
gl_FragColor += texture2D(texture, vUv + vec2(-7.0*px.x, -7.0*px.y))*0.0044299121055113265;
|
|
gl_FragColor += texture2D(texture, vUv + vec2(-6.0*px.x, -6.0*px.y))*0.00895781211794;
|
|
gl_FragColor += texture2D(texture, vUv + vec2(-5.0*px.x, -5.0*px.y))*0.0215963866053;
|
|
gl_FragColor += texture2D(texture, vUv + vec2(-4.0*px.x, -4.0*px.y))*0.0443683338718;
|
|
gl_FragColor += texture2D(texture, vUv + vec2(-3.0*px.x, -3.0*px.y))*0.0776744219933;
|
|
gl_FragColor += texture2D(texture, vUv + vec2(-2.0*px.x, -2.0*px.y))*0.115876621105;
|
|
gl_FragColor += texture2D(texture, vUv + vec2(-1.0*px.x, -1.0*px.y))*0.147308056121;
|
|
gl_FragColor += texture2D(texture, vUv )*0.159576912161;
|
|
gl_FragColor += texture2D(texture, vUv + vec2( 1.0*px.x, 1.0*px.y))*0.147308056121;
|
|
gl_FragColor += texture2D(texture, vUv + vec2( 2.0*px.x, 2.0*px.y))*0.115876621105;
|
|
gl_FragColor += texture2D(texture, vUv + vec2( 3.0*px.x, 3.0*px.y))*0.0776744219933;
|
|
gl_FragColor += texture2D(texture, vUv + vec2( 4.0*px.x, 4.0*px.y))*0.0443683338718;
|
|
gl_FragColor += texture2D(texture, vUv + vec2( 5.0*px.x, 5.0*px.y))*0.0215963866053;
|
|
gl_FragColor += texture2D(texture, vUv + vec2( 6.0*px.x, 6.0*px.y))*0.00895781211794;
|
|
gl_FragColor += texture2D(texture, vUv + vec2( 7.0*px.x, 7.0*px.y))*0.0044299121055113265;
|
|
}
|
|
`,wC=`
|
|
precision highp float;
|
|
varying vec2 vUv;
|
|
uniform sampler2D texture;
|
|
uniform vec2 px;
|
|
uniform float m[9];
|
|
void main(void) {
|
|
vec4 c11 = texture2D(texture, vUv - px); // top left
|
|
vec4 c12 = texture2D(texture, vec2(vUv.x, vUv.y - px.y)); // top center
|
|
vec4 c13 = texture2D(texture, vec2(vUv.x + px.x, vUv.y - px.y)); // top right
|
|
vec4 c21 = texture2D(texture, vec2(vUv.x - px.x, vUv.y) ); // mid left
|
|
vec4 c22 = texture2D(texture, vUv); // mid center
|
|
vec4 c23 = texture2D(texture, vec2(vUv.x + px.x, vUv.y) ); // mid right
|
|
vec4 c31 = texture2D(texture, vec2(vUv.x - px.x, vUv.y + px.y) ); // bottom left
|
|
vec4 c32 = texture2D(texture, vec2(vUv.x, vUv.y + px.y) ); // bottom center
|
|
vec4 c33 = texture2D(texture, vUv + px ); // bottom right
|
|
gl_FragColor =
|
|
c11 * m[0] + c12 * m[1] + c22 * m[2] +
|
|
c21 * m[3] + c22 * m[4] + c23 * m[5] +
|
|
c31 * m[6] + c32 * m[7] + c33 * m[8];
|
|
gl_FragColor.a = c22.a;
|
|
}
|
|
`;var z5=(e,t,r)=>{let n=new RegExp("\\b"+t+" \\w+ (\\w+)","ig");e.replace(n,(a,s)=>(r[s]=0,a))},O5=class{constructor(t,r,n){fe(this,"uniform",{});fe(this,"attribute",{});fe(this,"gl");fe(this,"id");fe(this,"compile",(t,r)=>{let n=this.gl.createShader(r);return n?(this.gl.shaderSource(n,t),this.gl.compileShader(n),this.gl.getShaderParameter(n,this.gl.COMPILE_STATUS)?n:(se(`filter: gl compile failed: ${this.gl.getShaderInfoLog(n)}`),null)):(se("filter: could not create shader"),null)});this.gl=t;let a=this.compile(r,this.gl.VERTEX_SHADER),s=this.compile(n,this.gl.FRAGMENT_SHADER);if(this.id=this.gl.createProgram(),!(!a||!s)){if(!this.id){se("filter: could not create webgl program");return}if(this.gl.attachShader(this.id,a),this.gl.attachShader(this.id,s),this.gl.linkProgram(this.id),!this.gl.getProgramParameter(this.id,this.gl.LINK_STATUS)){se(`filter: gl link failed: ${this.gl.getProgramInfoLog(this.id)}`);return}this.gl.useProgram(this.id),z5(r,"attribute",this.attribute);for(let i in this.attribute)this.attribute[i]=this.gl.getAttribLocation(this.id,i);z5(r,"uniform",this.uniform),z5(n,"uniform",this.uniform);for(let i in this.uniform)this.uniform[i]=this.gl.getUniformLocation(this.id,i)}}};function kC(){let e=0,t=null,r=!1,n=-1,a=[null,null],s=[],i=null,o=null,l=Kr(100,100),u={},d={INTERMEDIATE:1},h=l.getContext("webgl");if(!h){se("filter: cannot get webgl context");return}this.gl=h;function p(A,x){if(!(A===l.width&&x===l.height)){if(l.width=A,l.height=x,!i){let b=new Float32Array([-1,-1,0,1,1,-1,1,1,-1,1,0,0,-1,1,0,0,1,-1,1,1,1,1,1,0]);i=h.createBuffer(),h.bindBuffer(h.ARRAY_BUFFER,i),h.bufferData(h.ARRAY_BUFFER,b,h.STATIC_DRAW),h.pixelStorei(h.UNPACK_PREMULTIPLY_ALPHA_WEBGL,!0)}h.viewport(0,0,l.width,l.height),a=[null,null]}}function c(A,x){let b=h.createFramebuffer();h.bindFramebuffer(h.FRAMEBUFFER,b);let w=h.createRenderbuffer();h.bindRenderbuffer(h.RENDERBUFFER,w);let S=h.createTexture();return h.bindTexture(h.TEXTURE_2D,S),h.texImage2D(h.TEXTURE_2D,0,h.RGBA,A,x,0,h.RGBA,h.UNSIGNED_BYTE,null),h.texParameteri(h.TEXTURE_2D,h.TEXTURE_MAG_FILTER,h.LINEAR),h.texParameteri(h.TEXTURE_2D,h.TEXTURE_MIN_FILTER,h.LINEAR),h.texParameteri(h.TEXTURE_2D,h.TEXTURE_WRAP_S,h.CLAMP_TO_EDGE),h.texParameteri(h.TEXTURE_2D,h.TEXTURE_WRAP_T,h.CLAMP_TO_EDGE),h.framebufferTexture2D(h.FRAMEBUFFER,h.COLOR_ATTACHMENT0,h.TEXTURE_2D,S,0),h.bindTexture(h.TEXTURE_2D,null),h.bindFramebuffer(h.FRAMEBUFFER,null),{fbo:b,texture:S}}function f(A){return a[A]=a[A]||c(l.width,l.height),a[A]}function m(A=0){if(!o)return;let x=null,b=null,w=!1;e===0?x=t:x=f(n).texture||null,e++,r&&!(A&d.INTERMEDIATE)?(b=null,w=e%2===0):(n=(n+1)%2,b=f(n).fbo||null),h.bindTexture(h.TEXTURE_2D,x),h.bindFramebuffer(h.FRAMEBUFFER,b),h.uniform1f(o.uniform.flipY,w?-1:1),h.drawArrays(h.TRIANGLES,0,6)}function g(A){if(u[A])return o=u[A],h.useProgram((o?o.id:null)||null),o;if(o=new O5(h,yC,A),!o)return se("filter: could not get webgl program"),null;let x=Float32Array.BYTES_PER_ELEMENT,b=4*x;return h.enableVertexAttribArray(o.attribute.pos),h.vertexAttribPointer(o.attribute.pos,2,h.FLOAT,!1,b,0*x),h.enableVertexAttribArray(o.attribute.uv),h.vertexAttribPointer(o.attribute.uv,2,h.FLOAT,!1,b,2*x),u[A]=o,o}let y={colorMatrix:A=>{let x=new Float32Array(A);x[4]/=255,x[9]/=255,x[14]/=255,x[19]/=255;let b=x[18]===1&&x[3]===0&&x[8]===0&&x[13]===0&&x[15]===0&&x[16]===0&&x[17]===0&&x[19]===0?xC:AC,w=g(b);!w||(h.uniform1fv(w.uniform.m,x),m())},brightness:A=>{let x=(A||0)+1;y.colorMatrix([x,0,0,0,0,0,x,0,0,0,0,0,x,0,0,0,0,0,1,0])},saturation:A=>{let x=(A||0)*2/3+1,b=(x-1)*-.5;y.colorMatrix([x,b,b,0,0,b,x,b,0,0,b,b,x,0,0,0,0,0,1,0])},desaturate:()=>{y.saturation(-1)},contrast:A=>{let x=(A||0)+1,b=-128*(x-1);y.colorMatrix([x,0,0,0,b,0,x,0,0,b,0,0,x,0,b,0,0,0,1,0])},negative:()=>{y.contrast(-2)},hue:A=>{A=(A||0)/180*Math.PI;let x=Math.cos(A),b=Math.sin(A),w=.213,S=.715,N=.072;y.colorMatrix([w+x*(1-w)+b*-w,S+x*-S+b*-S,N+x*-N+b*(1-N),0,0,w+x*-w+b*.143,S+x*(1-S)+b*.14,N+x*-N+b*-.283,0,0,w+x*-w+b*-(1-w),S+x*-S+b*S,N+x*(1-N)+b*N,0,0,0,0,0,1,0])},desaturateLuminance:()=>{y.colorMatrix([.2764723,.929708,.0938197,0,-37.1,.2764723,.929708,.0938197,0,-37.1,.2764723,.929708,.0938197,0,-37.1,0,0,0,1,0])},sepia:()=>{y.colorMatrix([.393,.7689999,.18899999,0,0,.349,.6859999,.16799999,0,0,.272,.5339999,.13099999,0,0,0,0,0,1,0])},brownie:()=>{y.colorMatrix([.5997023498159715,.34553243048391263,-.2708298674538042,0,47.43192855600873,-.037703249837783157,.8609577587992641,.15059552388459913,0,-36.96841498319127,.24113635128153335,-.07441037908422492,.44972182064877153,0,-7.562075277591283,0,0,0,1,0])},vintagePinhole:()=>{y.colorMatrix([.6279345635605994,.3202183420819367,-.03965408211312453,0,9.651285835294123,.02578397704808868,.6441188644374771,.03259127616149294,0,7.462829176470591,.0466055556782719,-.0851232987247891,.5241648018700465,0,5.159190588235296,0,0,0,1,0])},kodachrome:()=>{y.colorMatrix([1.1285582396593525,-.3967382283601348,-.03992559172921793,0,63.72958762196502,-.16404339962244616,1.0835251566291304,-.05498805115633132,0,24.732407896706203,-.16786010706155763,-.5603416277695248,1.6014850761964943,0,35.62982807460946,0,0,0,1,0])},technicolor:()=>{y.colorMatrix([1.9125277891456083,-.8545344976951645,-.09155508482755585,0,11.793603434377337,-.3087833385928097,1.7658908555458428,-.10601743074722245,0,-70.35205161461398,-.231103377548616,-.7501899197440212,1.847597816108189,0,30.950940869491138,0,0,0,1,0])},polaroid:()=>{y.colorMatrix([1.438,-.062,-.062,0,0,-.122,1.378,-.122,0,0,-.016,-.016,1.483,0,0,0,0,0,1,0])},shiftToBGR:()=>{y.colorMatrix([0,0,1,0,0,0,1,0,0,0,1,0,0,0,0,0,0,0,1,0])},convolution:A=>{let x=new Float32Array(A),b=1/l.width,w=1/l.height,S=g(wC);!S||(h.uniform1fv(S.uniform.m,x),h.uniform2f(S.uniform.px,b,w),m())},detectEdges:()=>{y.convolution.call(this,[0,1,0,1,-4,1,0,1,0])},sobelX:()=>{y.convolution.call(this,[-1,0,1,-2,0,2,-1,0,1])},sobelY:()=>{y.convolution.call(this,[-1,-2,-1,0,0,0,1,2,1])},sharpen:A=>{let x=A||1;y.convolution.call(this,[0,-1*x,0,-1*x,1+4*x,-1*x,0,-1*x,0])},emboss:A=>{let x=A||1;y.convolution.call(this,[-2*x,-1*x,0,-1*x,1,1*x,0,1*x,2*x])},blur:A=>{let x=A/7/l.width,b=A/7/l.height,w=g(vC);!w||(h.uniform2f(w.uniform.px,0,b),m(d.INTERMEDIATE),h.uniform2f(w.uniform.px,x,0),m())},pixelate:A=>{let x=A/l.width,b=A/l.height,w=g(bC);!w||(h.uniform2f(w.uniform.size,x,b),m())}};this.add=function(A){let x=Array.prototype.slice.call(arguments,1),b=y[A];s.push({func:b,args:x})},this.reset=function(){s=[]},this.get=function(){return s},this.apply=function(A){p(A.width,A.height),e=0,t||(t=h.createTexture()),h.bindTexture(h.TEXTURE_2D,t),h.texParameteri(h.TEXTURE_2D,h.TEXTURE_WRAP_S,h.CLAMP_TO_EDGE),h.texParameteri(h.TEXTURE_2D,h.TEXTURE_WRAP_T,h.CLAMP_TO_EDGE),h.texParameteri(h.TEXTURE_2D,h.TEXTURE_MIN_FILTER,h.NEAREST),h.texParameteri(h.TEXTURE_2D,h.TEXTURE_MAG_FILTER,h.NEAREST),h.texImage2D(h.TEXTURE_2D,0,h.RGBA,h.RGBA,h.UNSIGNED_BYTE,A);for(let x=0;x<s.length;x++){r=x===s.length-1;let b=s[x];b.func.apply(this,b.args||[])}return l},this.draw=function(A){return this.add("brightness",0),this.apply(A)}}async function Pm(e){let t=e.shape.length===4?tt(e):e,r=Yt(t,3,2),n=[Bs(r[0]),Bs(r[1]),Bs(r[2])],a=[Ar(r[0]),Ar(r[1]),Ar(r[2])],s=await Promise.all(a.map(c=>c.data())),i=.99*Math.max(s[0][0],s[1][0],s[2][0]),o=[ce(r[0],n[0]),ce(r[1],n[1]),ce(r[2],n[2])],l=[ce(a[0],n[0]),ce(a[1],n[1]),ce(a[2],n[2])],u=[pe(i,l[0]),pe(i,l[1]),pe(i,l[2])],d=[L(o[0],u[0]),L(o[1],u[1]),L(o[2],u[2])],h=ur([d[0],d[1],d[2]],2),p=U(h,[1,t.shape[0],t.shape[1],3]);return te([...r,...n,...a,...o,...l,...u,...d,h,t]),p}var _m=3840,ut=null,er=null,Pd=null,Nt,is={inputSum:0,cacheDiff:1,sumMethod:0,inputTensor:void 0};function Kr(e,t){let r;if(he.browser)if(he.worker){if(typeof OffscreenCanvas=="undefined")throw new Error("canvas error: attempted to run in web worker but OffscreenCanvas is not supported");r=new OffscreenCanvas(e,t)}else{if(typeof document=="undefined")throw new Error("canvas error: attempted to run in browser but DOM is not defined");r=document.createElement("canvas"),r.width=e,r.height=t}else typeof he.Canvas!="undefined"?r=new he.Canvas(e,t):typeof globalThis.Canvas!="undefined"&&(r=new globalThis.Canvas(e,t));return r}function zm(e,t){let r=t||Kr(e.width,e.height);return r.getContext("2d").drawImage(e,0,0),r}async function _d(e,t,r=!0){if(!e)return t.debug&&se("input error: input is missing"),{tensor:null,canvas:null};if(!(e instanceof nt)&&!(typeof Image!="undefined"&&e instanceof Image)&&!(typeof he.Canvas!="undefined"&&e instanceof he.Canvas)&&!(typeof globalThis.Canvas!="undefined"&&e instanceof globalThis.Canvas)&&!(typeof ImageData!="undefined"&&e instanceof ImageData)&&!(typeof ImageBitmap!="undefined"&&e instanceof ImageBitmap)&&!(typeof HTMLImageElement!="undefined"&&e instanceof HTMLImageElement)&&!(typeof HTMLMediaElement!="undefined"&&e instanceof HTMLMediaElement)&&!(typeof HTMLVideoElement!="undefined"&&e instanceof HTMLVideoElement)&&!(typeof HTMLCanvasElement!="undefined"&&e instanceof HTMLCanvasElement)&&!(typeof OffscreenCanvas!="undefined"&&e instanceof OffscreenCanvas))throw new Error("input error: type is not recognized");if(e instanceof nt){let n=null;if(e.isDisposedInternal)throw new Error("input error: attempted to use tensor but it is disposed");if(!e.shape)throw new Error("input error: attempted to use tensor without a shape");if(e.shape.length===3){if(e.shape[2]===3)n=Kt(e,0);else if(e.shape[2]===4){let a=Pl(e,[0,0,0],[-1,-1,3]);n=Kt(a,0),te(a)}}else e.shape.length===4&&(e.shape[3]===3?n=Wr(e):e.shape[3]===4&&(n=Oo(e,[0,0,0,0],[-1,-1,-1,3])));if(n==null||n.shape.length!==4||n.shape[0]!==1||n.shape[3]!==3)throw new Error(`input error: attempted to use tensor with unrecognized shape: ${e.shape}`);if(n.dtype==="int32"){let a=me(n,"float32");te(n),n=a}return{tensor:n,canvas:t.filter.return?er:null}}else{if(typeof e.readyState!="undefined"&&e.readyState<=2)return t.debug&&se("input stream is not ready"),{tensor:null,canvas:ut};let n=e.naturalWidth||e.videoWidth||e.width||e.shape&&e.shape[1]>0,a=e.naturalHeight||e.videoHeight||e.height||e.shape&&e.shape[2]>0;if(!n||!a)return t.debug&&se("cannot determine input dimensions"),{tensor:null,canvas:ut};let s=n,i=a;if(s>_m&&(s=_m,i=Math.trunc(s*a/n)),i>_m&&(i=_m,s=Math.trunc(i*n/a)),(t.filter.width||0)>0?s=t.filter.width:(t.filter.height||0)>0&&(s=n*((t.filter.height||0)/a)),(t.filter.height||0)>0?i=t.filter.height:(t.filter.width||0)>0&&(i=a*((t.filter.width||0)/n)),!s||!i)throw new Error("input error: cannot determine dimension");(!ut||(ut==null?void 0:ut.width)!==s||(ut==null?void 0:ut.height)!==i)&&(ut=Kr(s,i));let o=ut.getContext("2d");if(typeof ImageData!="undefined"&&e instanceof ImageData?o.putImageData(e,0,0):t.filter.flip&&typeof o.translate!="undefined"?(o.translate(n,0),o.scale(-1,1),o.drawImage(e,0,0,n,a,0,0,ut==null?void 0:ut.width,ut==null?void 0:ut.height),o.setTransform(1,0,0,1,0,0)):o.drawImage(e,0,0,n,a,0,0,ut==null?void 0:ut.width,ut==null?void 0:ut.height),(!er||ut.width!==er.width||(ut==null?void 0:ut.height)!==(er==null?void 0:er.height))&&(er=Kr(ut.width,ut.height)),t.filter.enabled&&he.webgl.supported?(Nt||(Nt=he.browser?new kC:null),he.filter=!!Nt,!Nt||!Nt.add?(t.debug&&se("input process error: cannot initialize filters"),he.webgl.supported=!1,t.filter.enabled=!1,zm(ut,er)):(Nt.reset(),t.filter.brightness!==0&&Nt.add("brightness",t.filter.brightness),t.filter.contrast!==0&&Nt.add("contrast",t.filter.contrast),t.filter.sharpness!==0&&Nt.add("sharpen",t.filter.sharpness),t.filter.blur!==0&&Nt.add("blur",t.filter.blur),t.filter.saturation!==0&&Nt.add("saturation",t.filter.saturation),t.filter.hue!==0&&Nt.add("hue",t.filter.hue),t.filter.negative&&Nt.add("negative"),t.filter.sepia&&Nt.add("sepia"),t.filter.vintage&&Nt.add("brownie"),t.filter.sepia&&Nt.add("sepia"),t.filter.kodachrome&&Nt.add("kodachrome"),t.filter.technicolor&&Nt.add("technicolor"),t.filter.polaroid&&Nt.add("polaroid"),t.filter.pixelate!==0&&Nt.add("pixelate",t.filter.pixelate),Nt.get()>0?er=Nt.apply(ut):er=Nt.draw(ut))):(zm(ut,er),Nt&&(Nt=null),he.filter=!!Nt),!r)return{tensor:null,canvas:er};if(!er)throw new Error("canvas error: cannot create output");let l,u=3;if(typeof ImageData!="undefined"&&e instanceof ImageData||e.data&&e.width&&e.height)if(he.browser&&On)l=On?On.fromPixels(e):null;else{u=e.data.length/e.height/e.width;let p=new Uint8Array(e.data.buffer);l=ft(p,[e.height,e.width,u],"int32")}else if((!Pd||er.width!==Pd.width||er.height!==Pd.height)&&(Pd=Kr(er.width,er.height)),On&&he.browser)t.backend==="webgl"||t.backend==="humangl"||t.backend==="webgpu"?l=On.fromPixels(er):(Pd=zm(er),l=On.fromPixels(Pd));else{let f=zm(er).getContext("2d").getImageData(0,0,s,i);u=f.data.length/s/i;let m=new Uint8Array(f.data.buffer);l=ft(m,[s,i,u])}if(u===4){let p=Pl(l,[0,0,0],[-1,-1,3]);te(l),l=p}if(!l)throw new Error("input error: cannot create tensor");let d=me(l,"float32"),h=t.filter.equalization?await Pm(d):Kt(d,0);return te([l,d]),{tensor:h,canvas:t.filter.return?er:null}}}async function IC(e,t){let r=!1;if(e.cacheSensitivity===0||!t.shape||t.shape.length!==4||t.shape[1]>2048||t.shape[2]>2048)return r;if(!is.inputTensor)is.inputTensor=Wr(t);else if(is.inputTensor.shape[1]!==t.shape[1]||is.inputTensor.shape[2]!==t.shape[2])te(is.inputTensor),is.inputTensor=Wr(t);else{let n={};n.diff=ce(t,is.inputTensor),n.squared=L(n.diff,n.diff),n.sum=ke(n.squared);let s=(await n.sum.data())[0]/(t.shape[1]||1)/(t.shape[2]||1)/255/3;te([is.inputTensor,n.diff,n.squared,n.sum]),is.inputTensor=Wr(t),r=s<=(e.cacheSensitivity||0)}return r}async function SC(e,t,r){let n={};if(!t||!r||t.shape.length!==4||t.shape.length!==r.shape.length)return e.debug||se("invalid input tensor or tensor shapes do not match:",t.shape,r.shape),0;if(t.shape[0]!==1||r.shape[0]!==1||t.shape[3]!==3||r.shape[3]!==3)return e.debug||se("input tensors must be of shape [1, height, width, 3]:",t.shape,r.shape),0;n.input1=Wr(t),n.input2=t.shape[1]!==r.shape[1]||t.shape[2]!==r.shape[2]?Ie.resizeBilinear(r,[t.shape[1],t.shape[2]]):Wr(r),n.diff=ce(n.input1,n.input2),n.squared=L(n.diff,n.diff),n.sum=ke(n.squared);let s=(await n.sum.data())[0]/(t.shape[1]||1)/(t.shape[2]||1)/255/3;return te([n.input1,n.input2,n.diff,n.squared,n.sum]),s}var D5=class{constructor(){fe(this,"browser");fe(this,"node");fe(this,"worker");fe(this,"platform","");fe(this,"agent","");fe(this,"backends",[]);fe(this,"initial");fe(this,"filter");fe(this,"tfjs");fe(this,"offscreen");fe(this,"perfadd",!1);fe(this,"wasm",{supported:void 0,backend:void 0,simd:void 0,multithread:void 0});fe(this,"webgl",{supported:void 0,backend:void 0,version:void 0,renderer:void 0});fe(this,"webgpu",{supported:void 0,backend:void 0,adapter:void 0});fe(this,"cpu",{model:void 0,flags:[]});fe(this,"kernels",[]);fe(this,"Canvas");fe(this,"Image");fe(this,"ImageData");if(this.browser=typeof navigator!="undefined",this.node=typeof process!="undefined"&&typeof process.versions!="undefined"&&typeof process.versions.node!="undefined",this.tfjs={version:Kh["tfjs-core"]},this.offscreen=typeof OffscreenCanvas!="undefined",this.initial=!0,this.worker=this.browser&&this.offscreen?typeof WorkerGlobalScope!="undefined":void 0,typeof navigator!="undefined"){let t=navigator.userAgent.match(/\(([^()]+)\)/g);if(t&&t[0]){let r=t[0].match(/\(([^()]+)\)/g);this.platform=r&&r[0]?r[0].replace(/\(|\)/g,""):"",this.agent=navigator.userAgent.replace(t[0],""),this.platform[1]&&(this.agent=this.agent.replace(t[1],"")),this.agent=this.agent.replace(/ /g," ")}}else typeof process!="undefined"&&(this.platform=`${process.platform} ${process.arch}`,this.agent=`NodeJS ${process.version}`)}async updateBackend(){this.backends=Object.keys(Xt().registryFactory),this.wasm.supported=typeof WebAssembly!="undefined",this.wasm.backend=this.backends.includes("wasm"),this.wasm.supported&&this.wasm.backend&&Gr()==="wasm"&&(this.wasm.simd=await Z().getAsync("WASM_HAS_SIMD_SUPPORT"),this.wasm.multithread=await Z().getAsync("WASM_HAS_MULTITHREAD_SUPPORT"));let t=Kr(100,100),r=t?t.getContext("webgl2"):void 0;if(this.webgl.supported=typeof r!="undefined",this.webgl.backend=this.backends.includes("webgl"),this.webgl.supported&&this.webgl.backend&&(Gr()==="webgl"||Gr()==="humangl")){let n=Dn().gpgpu!=="undefined"?await Dn().getGPGPUContext().gl:null;n&&(this.webgl.version=n.getParameter(n.VERSION),this.webgl.renderer=n.getParameter(n.RENDERER))}this.webgpu.supported=this.browser&&typeof navigator.gpu!="undefined",this.webgpu.backend=this.backends.includes("webgpu");try{this.webgpu.supported&&(this.webgpu.adapter=(await navigator.gpu.requestAdapter()).name)}catch(n){this.webgpu.supported=!1}try{this.kernels=Ma(Gr()).map(n=>n.kernelName.toLowerCase())}catch(n){}}async updateCPU(){let t={model:"",flags:[]};this.node&&this.platform.startsWith("linux"),this.cpu?this.cpu=t:Object.defineProperty(this,"cpu",{value:t})}},he=new D5;var os={cacheModels:!1,verbose:!0,debug:!1,modelBasePath:""};async function Bye(e,t){return os.debug&&se("load model fetch:",e,t),fetch(e,t)}function CC(e){os.cacheModels=e.cacheModels,os.verbose=e.debug,os.modelBasePath=e.modelBasePath}async function Ge(e){let t=z3(os.modelBasePath,e||""),r=t.split("/"),n="indexeddb://"+r[r.length-1].replace(".json",""),a=await Cr.listModels(),s=os.cacheModels&&Object.keys(a).includes(n),i=typeof fetch=="undefined"?{}:{fetchFunc:(u,d)=>Bye(u,d)},o=new vm(s?n:t,i),l=!1;try{o.findIOHandler(),os.debug&&se("model load handler:",o.handler);let u=await o.handler.load();o.loadSync(u),os.verbose&&se("load model:",o.modelUrl),l=!0}catch(u){se("error loading model:",t,u)}if(l&&os.cacheModels&&!s)try{let u=await o.save(n);se("model saved:",n,u)}catch(u){se("error saving model:",t,u)}return o}var L5="2.7.1";var y1={};vs(y1,{Models:()=>sc,load:()=>Jb,reset:()=>g1,validate:()=>Qb});var Jn,B5=[],Uye=["white","black","asian","indian","other"],Gye=[15,23,28,35.5,45.5,55.5,65],TC=0,NC=0,W5=Number.MAX_SAFE_INTEGER;async function EC(e){return he.initial&&(Jn=null),Jn?e.debug&&se("cached model:",Jn.modelUrl):Jn=await Ge(e.face.gear),Jn}async function V5(e,t,r,n){var i,o;if(!Jn)return{age:0,gender:"unknown",genderScore:0,race:[]};let a=W5<(((i=t.face.gear)==null?void 0:i.skipFrames)||0),s=(((o=t.face.gear)==null?void 0:o.skipTime)||0)>oe()-NC;return t.skipAllowed&&s&&a&&TC===n&&B5[r]?(W5++,B5[r]):(W5=0,new Promise(async l=>{var y,A;if(!(Jn!=null&&Jn.inputs[0].shape))return;let u={},d=[[0,.1,.9,.9]];u.resize=Ie.cropAndResize(e,d,[0],[Jn.inputs[0].shape[2],Jn.inputs[0].shape[1]]);let h={age:0,gender:"unknown",genderScore:0,race:[]};(y=t.face.gear)!=null&&y.enabled&&([u.age,u.gender,u.race]=Jn.execute(u.resize,["age_output","gender_output","race_output"]));let p=await u.gender.data();h.gender=p[0]>p[1]?"male":"female",h.genderScore=Math.round(100*(p[0]>p[1]?p[0]:p[1]))/100;let c=await u.race.data();for(let x=0;x<c.length;x++)c[x]>(((A=t.face.gear)==null?void 0:A.minConfidence)||.2)&&h.race.push({score:Math.round(100*c[x])/100,race:Uye[x]});h.race.sort((x,b)=>b.score-x.score);let m=Array.from(await u.age.data()).map((x,b)=>[Gye[b],x]).sort((x,b)=>b[1]-x[1]),g=m[0][0];for(let x=1;x<m.length;x++)g+=m[x][1]*(m[x][0]-g);h.age=Math.round(10*g)/10,Object.keys(u).forEach(x=>te(u[x])),B5[r]=h,TC=n,NC=oe(),l(h)}))}var Je={tf255:255,tf1:1,tf2:2,tf05:.5,tf127:127.5,rgb:[.2989,.587,.114]};function $C(){Je.tf255=Se(255,"float32"),Je.tf1=Se(1,"float32"),Je.tf2=Se(2,"float32"),Je.tf05=Se(.5,"float32"),Je.tf127=Se(127.5,"float32"),Je.rgb=Tt([.2989,.587,.114],"float32")}var gn,Om=[],MC=0,FC=0,U5=Number.MAX_SAFE_INTEGER;async function PC(e){return he.initial&&(gn=null),gn?e.debug&&se("cached model:",gn.modelUrl):gn=await Ge(e.face.ssrnet.modelPathAge),gn}async function G5(e,t,r,n){var i,o,l,u;if(!gn)return{age:0};let a=U5<(((i=t.face.ssrnet)==null?void 0:i.skipFrames)||0),s=(((o=t.face.ssrnet)==null?void 0:o.skipTime)||0)>oe()-FC;return t.skipAllowed&&a&&s&&MC===n&&((l=Om[r])==null?void 0:l.age)&&((u=Om[r])==null?void 0:u.age)>0?(U5++,Om[r]):(U5=0,new Promise(async d=>{if(!(gn!=null&&gn.inputs)||!gn.inputs[0]||!gn.inputs[0].shape)return;let h={};h.resize=Ie.resizeBilinear(e,[gn.inputs[0].shape[2],gn.inputs[0].shape[1]],!1),h.enhance=L(h.resize,Je.tf255);let p={age:0};if(t.face.ssrnet.enabled&&(h.age=gn.execute(h.enhance)),h.age){let c=await h.age.data();p.age=Math.trunc(10*c[0])/10}Object.keys(h).forEach(c=>te(h[c])),Om[r]=p,MC=n,FC=oe(),d(p)}))}var Qn,Dm=[],zC=0,OC=0,j5=Number.MAX_SAFE_INTEGER,H5=[.2989,.587,.114];async function DC(e){return he.initial&&(Qn=null),Qn?e.debug&&se("cached model:",Qn.modelUrl):Qn=await Ge(e.face.ssrnet.modelPathGender),Qn}async function q5(e,t,r,n){var i,o,l,u;if(!Qn)return{gender:"unknown",genderScore:0};let a=j5<(((i=t.face.ssrnet)==null?void 0:i.skipFrames)||0),s=(((o=t.face.ssrnet)==null?void 0:o.skipTime)||0)>oe()-OC;return t.skipAllowed&&a&&s&&zC===n&&((l=Dm[r])==null?void 0:l.gender)&&((u=Dm[r])==null?void 0:u.genderScore)>0?(j5++,Dm[r]):(j5=0,new Promise(async d=>{if(!(Qn!=null&&Qn.inputs[0].shape))return;let h={};h.resize=Ie.resizeBilinear(e,[Qn.inputs[0].shape[2],Qn.inputs[0].shape[1]],!1),h.enhance=X(()=>{let[f,m,g]=Yt(h.resize,3,3),y=L(f,H5[0]),A=L(m,H5[1]),x=L(g,H5[2]),b=Cf([y,A,x]);return L(ce(b,Je.tf05),2)});let p={gender:"unknown",genderScore:0};t.face.ssrnet.enabled&&(h.gender=Qn.execute(h.enhance));let c=await h.gender.data();p.gender=c[0]>c[1]?"female":"male",p.genderScore=c[0]>c[1]?Math.trunc(100*c[0])/100:Math.trunc(100*c[1])/100,Object.keys(h).forEach(f=>te(h[f])),Dm[r]=p,zC=n,OC=oe(),d(p)}))}var Rr,Lm=[],X5=Number.MAX_SAFE_INTEGER,BC=0,WC=0;async function VC(e){var t;return he.initial&&(Rr=null),Rr?e.debug&&se("cached model:",Rr.modelUrl):Rr=await Ge((t=e.face.antispoof)==null?void 0:t.modelPath),Rr}async function K5(e,t,r,n){var i,o;if(!Rr)return 0;let a=(((i=t.face.antispoof)==null?void 0:i.skipTime)||0)>oe()-WC,s=X5<(((o=t.face.antispoof)==null?void 0:o.skipFrames)||0);return t.skipAllowed&&a&&s&&BC===n&&Lm[r]?(X5++,Lm[r]):(X5=0,new Promise(async l=>{let u=Ie.resizeBilinear(e,[Rr!=null&&Rr.inputs[0].shape?Rr.inputs[0].shape[2]:0,Rr!=null&&Rr.inputs[0].shape?Rr.inputs[0].shape[1]:0],!1),d=Rr==null?void 0:Rr.execute(u),h=(await d.data())[0];Lm[r]=Math.round(100*h)/100,BC=n,WC=oe(),te([u,d]),l(Lm[r])}))}var ea={silhouette:[10,338,297,332,284,251,389,356,454,323,361,288,397,365,379,378,400,377,152,148,176,149,150,136,172,58,132,93,234,127,162,21,54,103,67,109],lipsUpperOuter:[61,185,40,39,37,0,267,269,270,409,291],lipsLowerOuter:[146,91,181,84,17,314,405,321,375,291],lipsUpperInner:[78,191,80,81,82,13,312,311,310,415,308],lipsLowerInner:[78,95,88,178,87,14,317,402,318,324,308],rightEyeUpper0:[246,161,160,159,158,157,173],rightEyeLower0:[33,7,163,144,145,153,154,155,133],rightEyeUpper1:[247,30,29,27,28,56,190],rightEyeLower1:[130,25,110,24,23,22,26,112,243],rightEyeUpper2:[113,225,224,223,222,221,189],rightEyeLower2:[226,31,228,229,230,231,232,233,244],rightEyeLower3:[143,111,117,118,119,120,121,128,245],rightEyebrowUpper:[156,70,63,105,66,107,55,193],rightEyebrowLower:[35,124,46,53,52,65],rightEyeIris:[473,474,475,476,477],leftEyeUpper0:[466,388,387,386,385,384,398],leftEyeLower0:[263,249,390,373,374,380,381,382,362],leftEyeUpper1:[467,260,259,257,258,286,414],leftEyeLower1:[359,255,339,254,253,252,256,341,463],leftEyeUpper2:[342,445,444,443,442,441,413],leftEyeLower2:[446,261,448,449,450,451,452,453,464],leftEyeLower3:[372,340,346,347,348,349,350,357,465],leftEyebrowUpper:[383,300,293,334,296,336,285,417],leftEyebrowLower:[265,353,276,283,282,295],leftEyeIris:[468,469,470,471,472],midwayBetweenEyes:[168],noseTip:[1],noseBottom:[2],noseRightCorner:[98],noseLeftCorner:[327],rightCheek:[205],leftCheek:[425]},Z5={count:468,mouth:13,symmetryLine:[13,ea.midwayBetweenEyes[0]]},Yh={leftEye:0,rightEye:1,nose:2,mouth:3,leftEar:4,rightEar:5,symmetryLine:[3,2]},Y5=[{key:"EyeUpper0",indices:[9,10,11,12,13,14,15]},{key:"EyeUpper1",indices:[25,26,27,28,29,30,31]},{key:"EyeUpper2",indices:[41,42,43,44,45,46,47]},{key:"EyeLower0",indices:[0,1,2,3,4,5,6,7,8]},{key:"EyeLower1",indices:[16,17,18,19,20,21,22,23,24]},{key:"EyeLower2",indices:[32,33,34,35,36,37,38,39,40]},{key:"EyeLower3",indices:[54,55,56,57,58,59,60,61,62]},{key:"EyebrowUpper",indices:[63,64,65,66,67,68,69,70]},{key:"EyebrowLower",indices:[48,49,50,51,52,53]}],Jh=[[.499976992607117,.652534008026123],[.500025987625122,.547487020492554],[.499974012374878,.602371990680695],[.482113003730774,.471979022026062],[.500150978565216,.527155995368958],[.499909996986389,.498252987861633],[.499523013830185,.40106201171875],[.289712011814117,.380764007568359],[.499954998493195,.312398016452789],[.499987006187439,.269918978214264],[.500023007392883,.107050001621246],[.500023007392883,.666234016418457],[.5000159740448,.679224014282227],[.500023007392883,.692348003387451],[.499976992607117,.695277988910675],[.499976992607117,.70593398809433],[.499976992607117,.719385027885437],[.499976992607117,.737019002437592],[.499967992305756,.781370997428894],[.499816000461578,.562981009483337],[.473773002624512,.573909997940063],[.104906998574734,.254140973091125],[.365929991006851,.409575998783112],[.338757991790771,.41302502155304],[.311120003461838,.409460008144379],[.274657994508743,.389131009578705],[.393361985683441,.403706014156342],[.345234006643295,.344011008739471],[.370094001293182,.346076011657715],[.319321990013123,.347265005111694],[.297903001308441,.353591024875641],[.24779200553894,.410809993743896],[.396889001131058,.842755019664764],[.280097991228104,.375599980354309],[.106310002505779,.399955987930298],[.2099249958992,.391353011131287],[.355807989835739,.534406006336212],[.471751004457474,.65040397644043],[.474155008792877,.680191993713379],[.439785003662109,.657229006290436],[.414617002010345,.66654098033905],[.450374007225037,.680860996246338],[.428770989179611,.682690978050232],[.374971002340317,.727805018424988],[.486716985702515,.547628998756409],[.485300987958908,.527395009994507],[.257764995098114,.314490020275116],[.401223003864288,.455172002315521],[.429818987846375,.548614978790283],[.421351999044418,.533740997314453],[.276895999908447,.532056987285614],[.483370006084442,.499586999416351],[.33721199631691,.282882988452911],[.296391993761063,.293242990970612],[.169294998049736,.193813979625702],[.447580009698868,.302609980106354],[.392390012741089,.353887975215912],[.354490011930466,.696784019470215],[.067304998636246,.730105042457581],[.442739009857178,.572826027870178],[.457098007202148,.584792017936707],[.381974011659622,.694710969924927],[.392388999462128,.694203019142151],[.277076005935669,.271932005882263],[.422551989555359,.563233017921448],[.385919004678726,.281364023685455],[.383103013038635,.255840003490448],[.331431001424789,.119714021682739],[.229923993349075,.232002973556519],[.364500999450684,.189113974571228],[.229622006416321,.299540996551514],[.173287004232407,.278747975826263],[.472878992557526,.666198015213013],[.446828007698059,.668527007102966],[.422762006521225,.673889994621277],[.445307999849319,.580065965652466],[.388103008270264,.693961024284363],[.403039008378983,.706539988517761],[.403629004955292,.693953037261963],[.460041999816895,.557139039039612],[.431158006191254,.692366003990173],[.452181994915009,.692366003990173],[.475387006998062,.692366003990173],[.465828001499176,.779190003871918],[.472328990697861,.736225962638855],[.473087012767792,.717857003211975],[.473122000694275,.704625964164734],[.473033010959625,.695277988910675],[.427942007780075,.695277988910675],[.426479011774063,.703539967536926],[.423162013292313,.711845993995667],[.4183090031147,.720062971115112],[.390094995498657,.639572978019714],[.013953999616206,.560034036636353],[.499913990497589,.58014702796936],[.413199990987778,.69539999961853],[.409626007080078,.701822996139526],[.468080013990402,.601534962654114],[.422728985548019,.585985004901886],[.463079988956451,.593783974647522],[.37211999297142,.47341400384903],[.334562003612518,.496073007583618],[.411671012639999,.546965003013611],[.242175996303558,.14767599105835],[.290776997804642,.201445996761322],[.327338010072708,.256527006626129],[.399509996175766,.748921036720276],[.441727995872498,.261676013469696],[.429764986038208,.187834024429321],[.412198007106781,.108901023864746],[.288955003023148,.398952007293701],[.218936994671822,.435410976409912],[.41278201341629,.398970007896423],[.257135003805161,.355440020561218],[.427684992551804,.437960982322693],[.448339998722076,.536936044692993],[.178560003638268,.45755398273468],[.247308000922203,.457193970680237],[.286267012357712,.467674970626831],[.332827985286713,.460712015628815],[.368755996227264,.447206974029541],[.398963987827301,.432654976844788],[.476410001516342,.405806005001068],[.189241006970406,.523923993110657],[.228962004184723,.348950982093811],[.490725994110107,.562400996685028],[.404670000076294,.485132992267609],[.019469000399113,.401564002037048],[.426243007183075,.420431017875671],[.396993011236191,.548797011375427],[.266469985246658,.376977026462555],[.439121007919312,.51895797252655],[.032313998788595,.644356966018677],[.419054001569748,.387154996395111],[.462783008813858,.505746960639954],[.238978996872902,.779744982719421],[.198220998048782,.831938028335571],[.107550002634525,.540755033493042],[.183610007166862,.740257024765015],[.134409993886948,.333683013916016],[.385764002799988,.883153975009918],[.490967005491257,.579378008842468],[.382384985685349,.508572995662689],[.174399003386497,.397670984268188],[.318785011768341,.39623498916626],[.343364000320435,.400596976280212],[.396100014448166,.710216999053955],[.187885001301765,.588537991046906],[.430987000465393,.944064974784851],[.318993002176285,.898285031318665],[.266247987747192,.869701027870178],[.500023007392883,.190576016902924],[.499976992607117,.954452991485596],[.366169989109039,.398822009563446],[.393207013607025,.39553701877594],[.410373002290726,.391080021858215],[.194993004202843,.342101991176605],[.388664990663528,.362284004688263],[.365961998701096,.355970978736877],[.343364000320435,.355356991291046],[.318785011768341,.35834002494812],[.301414996385574,.363156020641327],[.058132998645306,.319076001644135],[.301414996385574,.387449026107788],[.499987989664078,.618434011936188],[.415838003158569,.624195992946625],[.445681989192963,.566076993942261],[.465844005346298,.620640993118286],[.49992299079895,.351523995399475],[.288718998432159,.819945991039276],[.335278987884521,.852819979190826],[.440512001514435,.902418971061707],[.128294005990028,.791940987110138],[.408771991729736,.373893976211548],[.455606997013092,.451801002025604],[.499877005815506,.908990025520325],[.375436991453171,.924192011356354],[.11421000212431,.615022003650665],[.448662012815475,.695277988910675],[.4480200111866,.704632043838501],[.447111994028091,.715808033943176],[.444831997156143,.730794012546539],[.430011987686157,.766808986663818],[.406787008047104,.685672998428345],[.400738000869751,.681069016456604],[.392399996519089,.677703022956848],[.367855995893478,.663918972015381],[.247923001646996,.601333022117615],[.452769994735718,.420849978923798],[.43639200925827,.359887003898621],[.416164010763168,.368713974952698],[.413385987281799,.692366003990173],[.228018000721931,.683571994304657],[.468268007040024,.352671027183533],[.411361992359161,.804327011108398],[.499989002943039,.469825029373169],[.479153990745544,.442654013633728],[.499974012374878,.439637005329132],[.432112008333206,.493588984012604],[.499886006116867,.866917014122009],[.49991300702095,.821729004383087],[.456548988819122,.819200992584229],[.344549000263214,.745438992977142],[.37890899181366,.574010014533997],[.374292999505997,.780184984207153],[.319687992334366,.570737957954407],[.357154995203018,.604269981384277],[.295284003019333,.621580958366394],[.447750002145767,.862477004528046],[.410986006259918,.508723020553589],[.31395098567009,.775308012962341],[.354128003120422,.812552988529205],[.324548006057739,.703992962837219],[.189096003770828,.646299958229065],[.279776990413666,.71465802192688],[.1338230073452,.682700991630554],[.336768001317978,.644733011722565],[.429883986711502,.466521978378296],[.455527991056442,.548622965812683],[.437114000320435,.558896005153656],[.467287987470627,.529924988746643],[.414712011814117,.335219979286194],[.37704598903656,.322777986526489],[.344107985496521,.320150971412659],[.312875986099243,.32233202457428],[.283526003360748,.333190023899078],[.241245999932289,.382785975933075],[.102986000478268,.468762993812561],[.267612010240555,.424560010433197],[.297879010438919,.433175981044769],[.333433985710144,.433878004550934],[.366427004337311,.426115989685059],[.396012008190155,.416696012020111],[.420121014118195,.41022801399231],[.007561000064015,.480777025222778],[.432949006557465,.569517970085144],[.458638995885849,.479089021682739],[.473466008901596,.545744001865387],[.476087987422943,.563830018043518],[.468472003936768,.555056989192963],[.433990985155106,.582361996173859],[.483518004417419,.562983989715576],[.482482999563217,.57784903049469],[.42645001411438,.389798998832703],[.438998997211456,.39649498462677],[.450067013502121,.400434017181396],[.289712011814117,.368252992630005],[.276670008897781,.363372981548309],[.517862021923065,.471948027610779],[.710287988185883,.380764007568359],[.526226997375488,.573909997940063],[.895093023777008,.254140973091125],[.634069979190826,.409575998783112],[.661242008209229,.41302502155304],[.688880026340485,.409460008144379],[.725341975688934,.389131009578705],[.606630027294159,.40370500087738],[.654766023159027,.344011008739471],[.629905998706818,.346076011657715],[.680678009986877,.347265005111694],[.702096998691559,.353591024875641],[.75221198797226,.410804986953735],[.602918028831482,.842862963676453],[.719901978969574,.375599980354309],[.893692970275879,.399959981441498],[.790081977844238,.391354024410248],[.643998026847839,.534487962722778],[.528249025344849,.65040397644043],[.525849997997284,.680191040039062],[.560214996337891,.657229006290436],[.585384011268616,.66654098033905],[.549625992774963,.680860996246338],[.57122802734375,.682691991329193],[.624852001667023,.72809898853302],[.513050019741058,.547281980514526],[.51509702205658,.527251958847046],[.742246985435486,.314507007598877],[.598631024360657,.454979002475739],[.570338010787964,.548575043678284],[.578631997108459,.533622980117798],[.723087012767792,.532054007053375],[.516445994377136,.499638974666595],[.662801027297974,.282917976379395],[.70362401008606,.293271005153656],[.830704987049103,.193813979625702],[.552385985851288,.302568018436432],[.607609987258911,.353887975215912],[.645429015159607,.696707010269165],[.932694971561432,.730105042457581],[.557260990142822,.572826027870178],[.542901992797852,.584792017936707],[.6180260181427,.694710969924927],[.607590973377228,.694203019142151],[.722943007946014,.271963000297546],[.577413976192474,.563166975975037],[.614082992076874,.281386971473694],[.616907000541687,.255886018276215],[.668509006500244,.119913995265961],[.770092010498047,.232020974159241],[.635536015033722,.189248979091644],[.77039098739624,.299556016921997],[.826722025871277,.278755009174347],[.527121007442474,.666198015213013],[.553171992301941,.668527007102966],[.577238023281097,.673889994621277],[.554691970348358,.580065965652466],[.611896991729736,.693961024284363],[.59696102142334,.706539988517761],[.596370995044708,.693953037261963],[.539958000183105,.557139039039612],[.568841993808746,.692366003990173],[.547818005084991,.692366003990173],[.52461302280426,.692366003990173],[.534089982509613,.779141008853912],[.527670979499817,.736225962638855],[.526912987232208,.717857003211975],[.526877999305725,.704625964164734],[.526966989040375,.695277988910675],[.572058022022247,.695277988910675],[.573521018028259,.703539967536926],[.57683801651001,.711845993995667],[.581691026687622,.720062971115112],[.609944999217987,.639909982681274],[.986046016216278,.560034036636353],[.5867999792099,.69539999961853],[.590372025966644,.701822996139526],[.531915009021759,.601536989212036],[.577268004417419,.585934996604919],[.536915004253387,.593786001205444],[.627542972564697,.473352015018463],[.665585994720459,.495950996875763],[.588353991508484,.546862006187439],[.757824003696442,.14767599105835],[.709249973297119,.201507985591888],[.672684013843536,.256581008434296],[.600408971309662,.74900496006012],[.55826598405838,.261672019958496],[.570303976535797,.187870979309082],[.588165998458862,.109044015407562],[.711045026779175,.398952007293701],[.781069993972778,.435405015945435],[.587247014045715,.398931980133057],[.742869973182678,.355445981025696],[.572156012058258,.437651991844177],[.55186802148819,.536570012569427],[.821442008018494,.457556009292603],[.752701997756958,.457181990146637],[.71375697851181,.467626988887787],[.66711300611496,.460672974586487],[.631101012229919,.447153985500336],[.6008620262146,.432473003864288],[.523481011390686,.405627012252808],[.810747981071472,.523926019668579],[.771045982837677,.348959028720856],[.509127020835876,.562718033790588],[.595292985439301,.485023975372314],[.980530977249146,.401564002037048],[.573499977588654,.420000016689301],[.602994978427887,.548687994480133],[.733529984951019,.376977026462555],[.560611009597778,.519016981124878],[.967685997486115,.644356966018677],[.580985009670258,.387160003185272],[.537728011608124,.505385041236877],[.760966002941132,.779752969741821],[.801778972148895,.831938028335571],[.892440974712372,.54076099395752],[.816350996494293,.740260004997253],[.865594983100891,.333687007427216],[.614073991775513,.883246004581451],[.508952975273132,.579437971115112],[.617941975593567,.508316040039062],[.825608015060425,.397674977779388],[.681214988231659,.39623498916626],[.656635999679565,.400596976280212],[.603900015354156,.710216999053955],[.81208598613739,.588539004325867],[.56801301240921,.944564998149872],[.681007981300354,.898285031318665],[.733752012252808,.869701027870178],[.633830010890961,.398822009563446],[.606792986392975,.39553701877594],[.589659988880157,.391062021255493],[.805015981197357,.342108011245728],[.611334979534149,.362284004688263],[.634037971496582,.355970978736877],[.656635999679565,.355356991291046],[.681214988231659,.35834002494812],[.698584973812103,.363156020641327],[.941866993904114,.319076001644135],[.698584973812103,.387449026107788],[.584177017211914,.624107003211975],[.554318010807037,.566076993942261],[.534153997898102,.62064003944397],[.711217999458313,.819975018501282],[.664629995822906,.852871000766754],[.559099972248077,.902631998062134],[.871706008911133,.791940987110138],[.591234028339386,.373893976211548],[.544341027736664,.451583981513977],[.624562978744507,.924192011356354],[.88577002286911,.615028977394104],[.551338016986847,.695277988910675],[.551980018615723,.704632043838501],[.552887976169586,.715808033943176],[.555167973041534,.730794012546539],[.569944024085999,.767035007476807],[.593203008174896,.685675978660583],[.599261999130249,.681069016456604],[.607599973678589,.677703022956848],[.631937980651855,.663500010967255],[.752032995223999,.601315021514893],[.547226011753082,.420395016670227],[.563543975353241,.359827995300293],[.583841025829315,.368713974952698],[.586614012718201,.692366003990173],[.771915018558502,.683578014373779],[.531597018241882,.352482974529266],[.588370978832245,.804440975189209],[.52079701423645,.442565023899078],[.567984998226166,.493479013442993],[.543282985687256,.819254994392395],[.655317008495331,.745514988899231],[.621008992195129,.574018001556396],[.625559985637665,.78031200170517],[.680198013782501,.570719003677368],[.64276397228241,.604337990283966],[.704662978649139,.621529996395111],[.552012026309967,.862591981887817],[.589071989059448,.508637011051178],[.685944974422455,.775357007980347],[.645735025405884,.812640011310577],[.675342977046967,.703978002071381],[.810858011245728,.646304965019226],[.72012197971344,.714666962623596],[.866151988506317,.682704985141754],[.663187026977539,.644596993923187],[.570082008838654,.466325998306274],[.544561982154846,.548375964164734],[.562758982181549,.558784961700439],[.531987011432648,.530140042304993],[.585271000862122,.335177004337311],[.622952997684479,.32277899980545],[.655896008014679,.320163011550903],[.687132000923157,.322345972061157],[.716481983661652,.333200991153717],[.758756995201111,.382786989212036],[.897013008594513,.468769013881683],[.732392013072968,.424547016620636],[.70211398601532,.433162987232208],[.66652500629425,.433866024017334],[.633504986763,.426087975502014],[.603875994682312,.416586995124817],[.579657971858978,.409945011138916],[.992439985275269,.480777025222778],[.567192018032074,.569419980049133],[.54136598110199,.478899002075195],[.526564002037048,.546118021011353],[.523913025856018,.563830018043518],[.531529009342194,.555056989192963],[.566035985946655,.582329034805298],[.51631098985672,.563053965568542],[.5174720287323,.577877044677734],[.573594987392426,.389806985855103],[.560697972774506,.395331978797913],[.549755990505219,.399751007556915],[.710287988185883,.368252992630005],[.723330020904541,.363372981548309]],Vl=[127,34,139,11,0,37,232,231,120,72,37,39,128,121,47,232,121,128,104,69,67,175,171,148,157,154,155,118,50,101,73,39,40,9,151,108,48,115,131,194,204,211,74,40,185,80,42,183,40,92,186,230,229,118,202,212,214,83,18,17,76,61,146,160,29,30,56,157,173,106,204,194,135,214,192,203,165,98,21,71,68,51,45,4,144,24,23,77,146,91,205,50,187,201,200,18,91,106,182,90,91,181,85,84,17,206,203,36,148,171,140,92,40,39,193,189,244,159,158,28,247,246,161,236,3,196,54,68,104,193,168,8,117,228,31,189,193,55,98,97,99,126,47,100,166,79,218,155,154,26,209,49,131,135,136,150,47,126,217,223,52,53,45,51,134,211,170,140,67,69,108,43,106,91,230,119,120,226,130,247,63,53,52,238,20,242,46,70,156,78,62,96,46,53,63,143,34,227,173,155,133,123,117,111,44,125,19,236,134,51,216,206,205,154,153,22,39,37,167,200,201,208,36,142,100,57,212,202,20,60,99,28,158,157,35,226,113,160,159,27,204,202,210,113,225,46,43,202,204,62,76,77,137,123,116,41,38,72,203,129,142,64,98,240,49,102,64,41,73,74,212,216,207,42,74,184,169,170,211,170,149,176,105,66,69,122,6,168,123,147,187,96,77,90,65,55,107,89,90,180,101,100,120,63,105,104,93,137,227,15,86,85,129,102,49,14,87,86,55,8,9,100,47,121,145,23,22,88,89,179,6,122,196,88,95,96,138,172,136,215,58,172,115,48,219,42,80,81,195,3,51,43,146,61,171,175,199,81,82,38,53,46,225,144,163,110,246,33,7,52,65,66,229,228,117,34,127,234,107,108,69,109,108,151,48,64,235,62,78,191,129,209,126,111,35,143,163,161,246,117,123,50,222,65,52,19,125,141,221,55,65,3,195,197,25,7,33,220,237,44,70,71,139,122,193,245,247,130,33,71,21,162,153,158,159,170,169,150,188,174,196,216,186,92,144,160,161,2,97,167,141,125,241,164,167,37,72,38,12,145,159,160,38,82,13,63,68,71,226,35,111,158,153,154,101,50,205,206,92,165,209,198,217,165,167,97,220,115,218,133,112,243,239,238,241,214,135,169,190,173,133,171,208,32,125,44,237,86,87,178,85,86,179,84,85,180,83,84,181,201,83,182,137,93,132,76,62,183,61,76,184,57,61,185,212,57,186,214,207,187,34,143,156,79,239,237,123,137,177,44,1,4,201,194,32,64,102,129,213,215,138,59,166,219,242,99,97,2,94,141,75,59,235,24,110,228,25,130,226,23,24,229,22,23,230,26,22,231,112,26,232,189,190,243,221,56,190,28,56,221,27,28,222,29,27,223,30,29,224,247,30,225,238,79,20,166,59,75,60,75,240,147,177,215,20,79,166,187,147,213,112,233,244,233,128,245,128,114,188,114,217,174,131,115,220,217,198,236,198,131,134,177,132,58,143,35,124,110,163,7,228,110,25,356,389,368,11,302,267,452,350,349,302,303,269,357,343,277,452,453,357,333,332,297,175,152,377,384,398,382,347,348,330,303,304,270,9,336,337,278,279,360,418,262,431,304,408,409,310,415,407,270,409,410,450,348,347,422,430,434,313,314,17,306,307,375,387,388,260,286,414,398,335,406,418,364,367,416,423,358,327,251,284,298,281,5,4,373,374,253,307,320,321,425,427,411,421,313,18,321,405,406,320,404,405,315,16,17,426,425,266,377,400,369,322,391,269,417,465,464,386,257,258,466,260,388,456,399,419,284,332,333,417,285,8,346,340,261,413,441,285,327,460,328,355,371,329,392,439,438,382,341,256,429,420,360,364,394,379,277,343,437,443,444,283,275,440,363,431,262,369,297,338,337,273,375,321,450,451,349,446,342,467,293,334,282,458,461,462,276,353,383,308,324,325,276,300,293,372,345,447,382,398,362,352,345,340,274,1,19,456,248,281,436,427,425,381,256,252,269,391,393,200,199,428,266,330,329,287,273,422,250,462,328,258,286,384,265,353,342,387,259,257,424,431,430,342,353,276,273,335,424,292,325,307,366,447,345,271,303,302,423,266,371,294,455,460,279,278,294,271,272,304,432,434,427,272,407,408,394,430,431,395,369,400,334,333,299,351,417,168,352,280,411,325,319,320,295,296,336,319,403,404,330,348,349,293,298,333,323,454,447,15,16,315,358,429,279,14,15,316,285,336,9,329,349,350,374,380,252,318,402,403,6,197,419,318,319,325,367,364,365,435,367,397,344,438,439,272,271,311,195,5,281,273,287,291,396,428,199,311,271,268,283,444,445,373,254,339,263,466,249,282,334,296,449,347,346,264,447,454,336,296,299,338,10,151,278,439,455,292,407,415,358,371,355,340,345,372,390,249,466,346,347,280,442,443,282,19,94,370,441,442,295,248,419,197,263,255,359,440,275,274,300,383,368,351,412,465,263,467,466,301,368,389,380,374,386,395,378,379,412,351,419,436,426,322,373,390,388,2,164,393,370,462,461,164,0,267,302,11,12,374,373,387,268,12,13,293,300,301,446,261,340,385,384,381,330,266,425,426,423,391,429,355,437,391,327,326,440,457,438,341,382,362,459,457,461,434,430,394,414,463,362,396,369,262,354,461,457,316,403,402,315,404,403,314,405,404,313,406,405,421,418,406,366,401,361,306,408,407,291,409,408,287,410,409,432,436,410,434,416,411,264,368,383,309,438,457,352,376,401,274,275,4,421,428,262,294,327,358,433,416,367,289,455,439,462,370,326,2,326,370,305,460,455,254,449,448,255,261,446,253,450,449,252,451,450,256,452,451,341,453,452,413,464,463,441,413,414,258,442,441,257,443,442,259,444,443,260,445,444,467,342,445,459,458,250,289,392,290,290,328,460,376,433,435,250,290,392,411,416,433,341,463,464,453,464,465,357,465,412,343,412,399,360,363,440,437,399,456,420,456,363,401,435,288,372,383,353,339,255,249,448,261,255,133,243,190,133,155,112,33,246,247,33,130,25,398,384,286,362,398,414,362,463,341,263,359,467,263,249,255,466,467,260,75,60,166,238,239,79,162,127,139,72,11,37,121,232,120,73,72,39,114,128,47,233,232,128,103,104,67,152,175,148,173,157,155,119,118,101,74,73,40,107,9,108,49,48,131,32,194,211,184,74,185,191,80,183,185,40,186,119,230,118,210,202,214,84,83,17,77,76,146,161,160,30,190,56,173,182,106,194,138,135,192,129,203,98,54,21,68,5,51,4,145,144,23,90,77,91,207,205,187,83,201,18,181,91,182,180,90,181,16,85,17,205,206,36,176,148,140,165,92,39,245,193,244,27,159,28,30,247,161,174,236,196,103,54,104,55,193,8,111,117,31,221,189,55,240,98,99,142,126,100,219,166,218,112,155,26,198,209,131,169,135,150,114,47,217,224,223,53,220,45,134,32,211,140,109,67,108,146,43,91,231,230,120,113,226,247,105,63,52,241,238,242,124,46,156,95,78,96,70,46,63,116,143,227,116,123,111,1,44,19,3,236,51,207,216,205,26,154,22,165,39,167,199,200,208,101,36,100,43,57,202,242,20,99,56,28,157,124,35,113,29,160,27,211,204,210,124,113,46,106,43,204,96,62,77,227,137,116,73,41,72,36,203,142,235,64,240,48,49,64,42,41,74,214,212,207,183,42,184,210,169,211,140,170,176,104,105,69,193,122,168,50,123,187,89,96,90,66,65,107,179,89,180,119,101,120,68,63,104,234,93,227,16,15,85,209,129,49,15,14,86,107,55,9,120,100,121,153,145,22,178,88,179,197,6,196,89,88,96,135,138,136,138,215,172,218,115,219,41,42,81,5,195,51,57,43,61,208,171,199,41,81,38,224,53,225,24,144,110,105,52,66,118,229,117,227,34,234,66,107,69,10,109,151,219,48,235,183,62,191,142,129,126,116,111,143,7,163,246,118,117,50,223,222,52,94,19,141,222,221,65,196,3,197,45,220,44,156,70,139,188,122,245,139,71,162,145,153,159,149,170,150,122,188,196,206,216,92,163,144,161,164,2,167,242,141,241,0,164,37,11,72,12,144,145,160,12,38,13,70,63,71,31,226,111,157,158,154,36,101,205,203,206,165,126,209,217,98,165,97,237,220,218,237,239,241,210,214,169,140,171,32,241,125,237,179,86,178,180,85,179,181,84,180,182,83,181,194,201,182,177,137,132,184,76,183,185,61,184,186,57,185,216,212,186,192,214,187,139,34,156,218,79,237,147,123,177,45,44,4,208,201,32,98,64,129,192,213,138,235,59,219,141,242,97,97,2,141,240,75,235,229,24,228,31,25,226,230,23,229,231,22,230,232,26,231,233,112,232,244,189,243,189,221,190,222,28,221,223,27,222,224,29,223,225,30,224,113,247,225,99,60,240,213,147,215,60,20,166,192,187,213,243,112,244,244,233,245,245,128,188,188,114,174,134,131,220,174,217,236,236,198,134,215,177,58,156,143,124,25,110,7,31,228,25,264,356,368,0,11,267,451,452,349,267,302,269,350,357,277,350,452,357,299,333,297,396,175,377,381,384,382,280,347,330,269,303,270,151,9,337,344,278,360,424,418,431,270,304,409,272,310,407,322,270,410,449,450,347,432,422,434,18,313,17,291,306,375,259,387,260,424,335,418,434,364,416,391,423,327,301,251,298,275,281,4,254,373,253,375,307,321,280,425,411,200,421,18,335,321,406,321,320,405,314,315,17,423,426,266,396,377,369,270,322,269,413,417,464,385,386,258,248,456,419,298,284,333,168,417,8,448,346,261,417,413,285,326,327,328,277,355,329,309,392,438,381,382,256,279,429,360,365,364,379,355,277,437,282,443,283,281,275,363,395,431,369,299,297,337,335,273,321,348,450,349,359,446,467,283,293,282,250,458,462,300,276,383,292,308,325,283,276,293,264,372,447,346,352,340,354,274,19,363,456,281,426,436,425,380,381,252,267,269,393,421,200,428,371,266,329,432,287,422,290,250,328,385,258,384,446,265,342,386,387,257,422,424,430,445,342,276,422,273,424,306,292,307,352,366,345,268,271,302,358,423,371,327,294,460,331,279,294,303,271,304,436,432,427,304,272,408,395,394,431,378,395,400,296,334,299,6,351,168,376,352,411,307,325,320,285,295,336,320,319,404,329,330,349,334,293,333,366,323,447,316,15,315,331,358,279,317,14,316,8,285,9,277,329,350,253,374,252,319,318,403,351,6,419,324,318,325,397,367,365,288,435,397,278,344,439,310,272,311,248,195,281,375,273,291,175,396,199,312,311,268,276,283,445,390,373,339,295,282,296,448,449,346,356,264,454,337,336,299,337,338,151,294,278,455,308,292,415,429,358,355,265,340,372,388,390,466,352,346,280,295,442,282,354,19,370,285,441,295,195,248,197,457,440,274,301,300,368,417,351,465,251,301,389,385,380,386,394,395,379,399,412,419,410,436,322,387,373,388,326,2,393,354,370,461,393,164,267,268,302,12,386,374,387,312,268,13,298,293,301,265,446,340,380,385,381,280,330,425,322,426,391,420,429,437,393,391,326,344,440,438,458,459,461,364,434,394,428,396,262,274,354,457,317,316,402,316,315,403,315,314,404,314,313,405,313,421,406,323,366,361,292,306,407,306,291,408,291,287,409,287,432,410,427,434,411,372,264,383,459,309,457,366,352,401,1,274,4,418,421,262,331,294,358,435,433,367,392,289,439,328,462,326,94,2,370,289,305,455,339,254,448,359,255,446,254,253,449,253,252,450,252,256,451,256,341,452,414,413,463,286,441,414,286,258,441,258,257,442,257,259,443,259,260,444,260,467,445,309,459,250,305,289,290,305,290,460,401,376,435,309,250,392,376,411,433,453,341,464,357,453,465,343,357,412,437,343,399,344,360,440,420,437,456,360,420,363,361,401,288,265,372,353,390,339,249,339,448,255];var Hye=[127,234,132,58,172,150,149,148,152,377,378,379,397,288,361,454,356,70,63,105,66,107,336,296,334,293,300,168,6,195,4,98,97,2,326,327,33,160,158,133,153,144,362,385,387,263,373,380,57,40,37,0,267,270,287,321,314,17,84,91,78,81,13,311,308,402,14,178],qye=[33,133,362,263,1,62,308,159,145,386,374,6,102,331,2,13,14,70,105,107,336,334,300,54,10,284,50,280,234,454,58,288,152],Xye=[33,133,362,263,1,78,308],Bwe=Hye.map(e=>Jh[e]),Wwe=qye.map(e=>Jh[e]),Vwe=Xye.map(e=>Jh[e]);function Ji(e){let t=e.map(r=>r[0]);return t.push(e[e.length-1][1]),t}var Kye=[[61,146],[146,91],[91,181],[181,84],[84,17],[17,314],[314,405],[405,321],[321,375],[375,291],[61,185],[185,40],[40,39],[39,37],[37,0],[0,267],[267,269],[269,270],[270,409],[409,291],[78,95],[95,88],[88,178],[178,87],[87,14],[14,317],[317,402],[402,318],[318,324],[324,308],[78,191],[191,80],[80,81],[81,82],[82,13],[13,312],[312,311],[311,310],[310,415],[415,308]],Zye=[[263,249],[249,390],[390,373],[373,374],[374,380],[380,381],[381,382],[382,362],[263,466],[466,388],[388,387],[387,386],[386,385],[385,384],[384,398],[398,362]],Yye=[[276,283],[283,282],[282,295],[295,285],[300,293],[293,334],[334,296],[296,336]],Jye=[[474,475],[475,476],[476,477],[477,474]],Qye=[[33,7],[7,163],[163,144],[144,145],[145,153],[153,154],[154,155],[155,133],[33,246],[246,161],[161,160],[160,159],[159,158],[158,157],[157,173],[173,133]],eAe=[[46,53],[53,52],[52,65],[65,55],[70,63],[63,105],[105,66],[66,107]],tAe=[[469,470],[470,471],[471,472],[472,469]],rAe=[[10,338],[338,297],[297,332],[332,284],[284,251],[251,389],[389,356],[356,454],[454,323],[323,361],[361,288],[288,397],[397,365],[365,379],[379,378],[378,400],[400,377],[377,152],[152,148],[148,176],[176,149],[149,150],[150,136],[136,172],[172,58],[58,132],[132,93],[93,234],[234,127],[127,162],[162,21],[21,54],[54,103],[103,67],[67,109],[109,10]],Uwe={lips:Ji(Kye),leftEye:Ji(Zye),leftEyebrow:Ji(Yye),leftIris:Ji(Jye),rightEye:Ji(Qye),rightEyebrow:Ji(eAe),rightIris:Ji(tAe),faceOval:Ji(rAe)};var zd=e=>[Math.abs(e.endPoint[0]-e.startPoint[0]),Math.abs(e.endPoint[1]-e.startPoint[1])],Bm=e=>[e.startPoint[0]+(e.endPoint[0]-e.startPoint[0])/2,e.startPoint[1]+(e.endPoint[1]-e.startPoint[1])/2,1],tb=(e,t)=>e?[Math.trunc(Math.max(0,e.startPoint[0])),Math.trunc(Math.max(0,e.startPoint[1])),Math.trunc(Math.min(t.shape[2]||0,e.endPoint[0])-Math.max(0,e.startPoint[0])),Math.trunc(Math.min(t.shape[1]||0,e.endPoint[1])-Math.max(0,e.startPoint[1]))]:[0,0,0,0],rb=(e,t)=>e?[e.startPoint[0]/(t.shape[2]||0),e.startPoint[1]/(t.shape[1]||0),(e.endPoint[0]-e.startPoint[0])/(t.shape[2]||0),(e.endPoint[1]-e.startPoint[1])/(t.shape[1]||0)]:[0,0,0,0],HC=(e,t)=>{let r=[e.startPoint[0]*t[0],e.startPoint[1]*t[1]],n=[e.endPoint[0]*t[0],e.endPoint[1]*t[1]];return{startPoint:r,endPoint:n,landmarks:e.landmarks,confidence:e.confidence}},Q5=(e,t,r)=>{let n=t.shape[1],a=t.shape[2],s=[e.startPoint[1]/n,e.startPoint[0]/a,e.endPoint[1]/n,e.endPoint[0]/a],i=Ie.cropAndResize(t,[s],[0],r),o=pe(i,Je.tf255);return te(i),o},Wm=(e,t)=>{let r=Bm(e),n=zd(e),a=[t*n[0]/2,t*n[1]/2];return{startPoint:[r[0]-a[0],r[1]-a[1]],endPoint:[r[0]+a[0],r[1]+a[1]],landmarks:e.landmarks,confidence:e.confidence}},Vm=e=>{let t=Bm(e),r=zd(e),n=Math.max(...r)/2;return{startPoint:[Math.round(t[0]-n),Math.round(t[1]-n)],endPoint:[Math.round(t[0]+n),Math.round(t[1]+n)],landmarks:e.landmarks,confidence:e.confidence}},qC=e=>{let t=e.map(n=>n[0]),r=e.map(n=>n[1]);return{startPoint:[Math.min(...t),Math.min(...r)],endPoint:[Math.max(...t),Math.max(...r)],landmarks:e}},eb=[[1,0,0],[0,1,0],[0,0,1]],nAe=e=>e-2*Math.PI*Math.floor((e+Math.PI)/(2*Math.PI)),aAe=(e,t)=>nAe(Math.PI/2-Math.atan2(-(t[1]-e[1]),t[0]-e[0]));var GC=(e,t)=>[[1,0,e],[0,1,t],[0,0,1]],Ul=(e,t)=>{let r=0;for(let n=0;n<e.length;n++)r+=e[n]*t[n];return r},sAe=(e,t)=>{let r=[];for(let n=0;n<e.length;n++)r.push(e[n][t]);return r},jC=(e,t)=>{let r=[],n=e.length;for(let a=0;a<n;a++){r.push([]);for(let s=0;s<n;s++)r[a].push(Ul(e[a],sAe(t,s)))}return r},XC=(e,t)=>{let r=Math.cos(e),n=Math.sin(e),a=[[r,-n,0],[n,r,0],[0,0,1]],s=GC(t[0],t[1]),i=jC(s,a),o=GC(-t[0],-t[1]);return jC(i,o)},iAe=e=>{let t=[[e[0][0],e[1][0]],[e[0][1],e[1][1]]],r=[e[0][2],e[1][2]],n=[-Ul(t[0],r),-Ul(t[1],r)];return[t[0].concat(n[0]),t[1].concat(n[1]),[0,0,1]]},oAe=(e,t)=>[Ul(e,t[0]),Ul(e,t[1])];function KC(e){let t={strides:[e/16,e/8],anchors:[2,6]},r=[];for(let n=0;n<t.strides.length;n++){let a=t.strides[n],s=Math.floor((e+a-1)/a),i=Math.floor((e+a-1)/a),o=t.anchors[n];for(let l=0;l<s;l++){let u=a*(l+.5);for(let d=0;d<i;d++){let h=a*(d+.5);for(let p=0;p<o;p++)r.push([h,u])}}}return r}function ZC(e,t,r,n,a){let s=zd(t),i=e.map(c=>[s[0]/a*(c[0]-a/2),s[1]/a*(c[1]-a/2),c[2]||0]),o=r&&r!==0&&Math.abs(r)>.2,l=o?XC(r,[0,0]):eb,u=o?i.map(c=>[...oAe(c,l),c[2]]):i,d=o?iAe(n):eb,h=Bm(t),p=[Ul(h,d[0]),Ul(h,d[1])];return u.map(c=>[Math.trunc(c[0]+p[0]),Math.trunc(c[1]+p[1]),Math.trunc(c[2]||0)])}function YC(e,t,r,n){let a=t.landmarks.length>=Z5.count?Z5.symmetryLine:Yh.symmetryLine,s=0,i=eb,o;if(e&&he.kernels.includes("rotatewithoffset"))if(s=aAe(t.landmarks[a[0]],t.landmarks[a[1]]),s&&s!==0&&Math.abs(s)>.2){let u=Bm(t),d=[u[0]/r.shape[2],u[1]/r.shape[1]],h=Ie.rotateWithOffset(r,s,0,d);i=XC(-s,u),o=Q5(t,h,[n,n]),te(h)}else o=Q5(t,r,[n,n]);else o=Q5(t,r,[n,n]);return[s,i,o]}var lAe=e=>{let t=e.map(n=>n[0]),r=e.map(n=>n[1]);return[Math.min(...t)+(Math.max(...t)-Math.min(...t))/2,Math.min(...r)+(Math.max(...r)-Math.min(...r))/2]},JC=(e,t)=>{let r=lAe(e),n=zd(t);return{startPoint:[r[0]-n[0]/2,r[1]-n[1]/2],endPoint:[r[0]+n[0]/2,r[1]+n[1]/2]}};var QC=6,uAe=1.4,Da,eT=null,Qi=0,Qh=null,Um=()=>Qi;async function tT(e){var t;return he.initial&&(Da=null),Da?e.debug&&se("cached model:",Da.modelUrl):Da=await Ge((t=e.face.detector)==null?void 0:t.modelPath),Qi=Da.inputs[0].shape?Da.inputs[0].shape[2]:0,Qh=Se(Qi,"int32"),eT=ca(KC(Qi)),Da}function dAe(e){let t={};t.boxStarts=Pe(e,[0,1],[-1,2]),t.centers=le(t.boxStarts,eT),t.boxSizes=Pe(e,[0,3],[-1,2]),t.boxSizesNormalized=pe(t.boxSizes,Qh),t.centersNormalized=pe(t.centers,Qh),t.halfBoxSize=pe(t.boxSizesNormalized,Je.tf2),t.starts=ce(t.centersNormalized,t.halfBoxSize),t.ends=le(t.centersNormalized,t.halfBoxSize),t.startNormalized=L(t.starts,Qh),t.endNormalized=L(t.ends,Qh);let r=hd([t.startNormalized,t.endNormalized],1);return Object.keys(t).forEach(n=>te(t[n])),r}async function rT(e,t){var o,l,u,d;if(!e||e.isDisposedInternal||e.shape.length!==4||e.shape[1]<1||e.shape[2]<1)return[];let r={};r.resized=Ie.resizeBilinear(e,[Qi,Qi]),r.div=pe(r.resized,Je.tf127),r.normalized=ce(r.div,Je.tf05);let n=Da==null?void 0:Da.execute(r.normalized);if(Array.isArray(n)){let h=n.sort((p,c)=>p.size-c.size);r.concat384=St([h[0],h[2]],2),r.concat512=St([h[1],h[3]],2),r.concat=St([r.concat512,r.concat384],1),r.batch=tt(r.concat,0)}else r.batch=tt(n);te(n),r.boxes=dAe(r.batch),r.logits=Pe(r.batch,[0,0],[-1,1]),r.sigmoid=Tr(r.logits),r.scores=tt(r.sigmoid),r.nms=await Ie.nonMaxSuppressionAsync(r.boxes,r.scores,((o=t.face.detector)==null?void 0:o.maxDetected)||0,((l=t.face.detector)==null?void 0:l.iouThreshold)||0,((u=t.face.detector)==null?void 0:u.minConfidence)||0);let a=await r.nms.array(),s=[],i=await r.scores.data();for(let h=0;h<a.length;h++){let p=i[a[h]];if(p>(((d=t.face.detector)==null?void 0:d.minConfidence)||0)){let c={};c.bbox=Pe(r.boxes,[a[h],0],[1,-1]),c.slice=Pe(r.batch,[a[h],QC-1],[1,-1]),c.squeeze=tt(c.slice),c.landmarks=U(c.squeeze,[QC,-1]);let f=await c.bbox.data(),m={startPoint:[f[0],f[1]],endPoint:[f[2],f[3]],landmarks:await c.landmarks.array(),confidence:p},g=HC(m,[(e.shape[2]||0)/Qi,(e.shape[1]||0)/Qi]),y=Wm(g,t.face.scale||uAe),A=Vm(y);s.push(A),Object.keys(c).forEach(x=>te(c[x]))}}return Object.keys(r).forEach(h=>te(r[h])),s}var Gm={};vs(Gm,{connected:()=>sb,kpt:()=>ab});var ab=["nose","leftEyeInside","leftEye","leftEyeOutside","rightEyeInside","rightEye","rightEyeOutside","leftEar","rightEar","leftMouth","rightMouth","leftShoulder","rightShoulder","leftElbow","rightElbow","leftWrist","rightWrist","leftPinky","rightPinky","leftIndex","rightIndex","leftThumb","rightThumb","leftHip","rightHip","leftKnee","rightKnee","leftAnkle","rightAnkle","leftHeel","rightHeel","leftFoot","rightFoot","bodyCenter","bodyTop","leftPalm","leftHand","rightPalm","rightHand"],sb={shoulders:["leftShoulder","rightShoulder"],hips:["rightHip","leftHip"],mouth:["leftMouth","rightMouth"],leftLegUpper:["leftHip","leftKnee"],leftLegLower:["leftKnee","leftAnkle"],leftFoot:["leftAnkle","leftHeel","leftFoot"],leftTorso:["leftShoulder","leftHip"],leftArmUpper:["leftShoulder","leftElbow"],leftArmLower:["leftElbow","leftWrist"],leftHand:["leftWrist","leftPalm"],leftHandPinky:["leftPalm","leftPinky"],leftHandIndex:["leftPalm","leftIndex"],leftHandThumb:["leftPalm","leftThumb"],leftEyeOutline:["leftEyeInside","leftEyeOutside"],rightLegUpper:["rightHip","rightKnee"],rightLegLower:["rightKnee","rightAnkle"],rightFoot:["rightAnkle","rightHeel","rightFoot"],rightTorso:["rightShoulder","rightHip"],rightArmUpper:["rightShoulder","rightElbow"],rightArmLower:["rightElbow","rightWrist"],rightHand:["rightWrist","rightPalm"],rightHandPinky:["rightPalm","rightPinky"],rightHandIndex:["rightPalm","rightIndex"],rightHandThumb:["rightPalm","rightThumb"],rightEyeOutline:["rightEyeInside","rightEyeOutside"]};var aT=224,pAe,hAe=5,jm=[8,16,32,32,32];async function sT(){let e=[],t=0;for(;t<hAe;){let r=0,n=t;for(;n<jm.length&&jm[n]===jm[t];)r+=2,n++;let a=jm[t],s=Math.ceil(aT/a),i=Math.ceil(aT/a);for(let o=0;o<s;++o)for(let l=0;l<i;++l)for(let u=0;u<r;++u)e.push({x:(l+.5)/i,y:(o+.5)/s});t=n}pAe={x:Tt(e.map(r=>r.x)),y:Tt(e.map(r=>r.y))}}function ls(e,t=[1,1]){let r=[e.map(o=>o[0]),e.map(o=>o[1])],n=[Math.min(...r[0]),Math.min(...r[1])],a=[Math.max(...r[0]),Math.max(...r[1])],s=[n[0],n[1],a[0]-n[0],a[1]-n[1]],i=[s[0]/t[0],s[1]/t[1],s[2]/t[0],s[3]/t[1]];return{box:s,boxRaw:i}}function iT(e,t=[1,1]){let r=[e.map(u=>u[0]),e.map(u=>u[1])],n=[Math.min(...r[0]),Math.min(...r[1])],a=[Math.max(...r[0]),Math.max(...r[1])],s=[(n[0]+a[0])/2,(n[1]+a[1])/2],i=Math.max(s[0]-n[0],s[1]-n[1],-s[0]+a[0],-s[1]+a[1]),o=[Math.trunc(s[0]-i),Math.trunc(s[1]-i),Math.trunc(2*i),Math.trunc(2*i)],l=[o[0]/t[0],o[1]/t[1],o[2]/t[0],o[3]/t[1]];return{box:o,boxRaw:l}}function Hm(e,t){let r=[e[2]*t,e[3]*t];return[e[0]-(r[0]-e[2])/2,e[1]-(r[1]-e[3])/2,r[0],r[1]]}var uT={initial:!0},yn={detector:null,landmarks:null},Od={detector:[224,224],landmarks:[256,256]},ib=Number.MAX_SAFE_INTEGER,fAe={landmarks:["ld_3d","activation_segmentation","activation_heatmap","world_3d","output_poseflag"],detector:[]},Xm=null,ec,eo=[[0,0],[0,0],[0,0],[0,0]],oT=0,lT=e=>1-1/(1+Math.exp(e));async function dT(e){if(uT.initial&&(yn.detector=null),!yn.detector&&e.body.detector&&e.body.detector.modelPath){yn.detector=await Ge(e.body.detector.modelPath);let t=Object.values(yn.detector.modelSignature.inputs);Od.detector[0]=Array.isArray(t)?parseInt(t[0].tensorShape.dim[1].size):0,Od.detector[1]=Array.isArray(t)?parseInt(t[0].tensorShape.dim[2].size):0}else e.debug&&yn.detector&&se("cached model:",yn.detector.modelUrl);return await sT(),yn.detector}async function pT(e){if(uT.initial&&(yn.landmarks=null),yn.landmarks)e.debug&&se("cached model:",yn.landmarks.modelUrl);else{yn.landmarks=await Ge(e.body.modelPath);let t=Object.values(yn.landmarks.modelSignature.inputs);Od.landmarks[0]=Array.isArray(t)?parseInt(t[0].tensorShape.dim[1].size):0,Od.landmarks[1]=Array.isArray(t)?parseInt(t[0].tensorShape.dim[2].size):0}return yn.landmarks}async function mAe(e,t){let r={};if(!e.shape||!e.shape[1]||!e.shape[2])return e;let n;if(ec&&(r.cropped=Ie.cropAndResize(e,[ec],[0],[e.shape[1],e.shape[2]])),e.shape[1]!==e.shape[2]){let a=[e.shape[2]>e.shape[1]?Math.trunc((e.shape[2]-e.shape[1])/2):0,e.shape[2]>e.shape[1]?Math.trunc((e.shape[2]-e.shape[1])/2):0],s=[e.shape[1]>e.shape[2]?Math.trunc((e.shape[1]-e.shape[2])/2):0,e.shape[1]>e.shape[2]?Math.trunc((e.shape[1]-e.shape[2])/2):0];eo=[[0,0],a,s,[0,0]],r.pad=Xn(r.cropped||e,eo),r.resize=Ie.resizeBilinear(r.pad,[t,t]),n=pe(r.resize,Je.tf255)}else e.shape[1]!==t?(r.resize=Ie.resizeBilinear(r.cropped||e,[t,t]),n=pe(r.resize,Je.tf255)):n=pe(r.cropped||e,Je.tf255);return Object.keys(r).forEach(a=>te(r[a])),n}function gAe(e,t){for(let r of e)r.position=[Math.trunc(r.position[0]*(t[0]+eo[2][0]+eo[2][1])/t[0]-eo[2][0]),Math.trunc(r.position[1]*(t[1]+eo[1][0]+eo[1][1])/t[1]-eo[1][0]),r.position[2]],r.positionRaw=[r.position[0]/t[0],r.position[1]/t[1],2*r.position[2]/(t[0]+t[1])];if(ec)for(let r of e)r.positionRaw=[r.positionRaw[0]+ec[1],r.positionRaw[1]+ec[0],r.positionRaw[2]],r.position=[Math.trunc(r.positionRaw[0]*t[0]),Math.trunc(r.positionRaw[1]*t[1]),r.positionRaw[2]];return e}async function yAe(e){let t=e.find(o=>o.part==="leftPalm"),r=e.find(o=>o.part==="leftWrist"),n=e.find(o=>o.part==="leftIndex");t.position[2]=((r.position[2]||0)+(n.position[2]||0))/2;let a=e.find(o=>o.part==="rightPalm"),s=e.find(o=>o.part==="rightWrist"),i=e.find(o=>o.part==="rightIndex");a.position[2]=((s.position[2]||0)+(i.position[2]||0))/2}async function AAe(e,t,r){var f;let n={};[n.ld,n.segmentation,n.heatmap,n.world,n.poseflag]=(f=yn.landmarks)==null?void 0:f.execute(e,fAe.landmarks);let a=(await n.poseflag.data())[0],s=await n.ld.data(),i=await n.world.data();Object.keys(n).forEach(m=>te(n[m]));let o=[],l=5;for(let m=0;m<s.length/l;m++){let g=lT(s[l*m+3]),y=lT(s[l*m+4]),A=Math.trunc(100*g*y*a)/100,x=[s[l*m+0]/Od.landmarks[0],s[l*m+1]/Od.landmarks[1],s[l*m+2]+0],b=[Math.trunc(r[0]*x[0]),Math.trunc(r[1]*x[1]),x[2]],w=[i[l*m+0],i[l*m+1],i[l*m+2]+0];o.push({part:ab[m],positionRaw:x,position:b,distance:w,score:A})}if(a<(t.body.minConfidence||0))return null;yAe(o);let u=gAe(o,r),d=u.map(m=>m.position),h=ls(d,[r[0],r[1]]),p={};for(let[m,g]of Object.entries(sb)){let y=[];for(let A=0;A<g.length-1;A++){let x=u.find(w=>w.part===g[A]),b=u.find(w=>w.part===g[A+1]);x&&b&&y.push([x.position,b.position])}p[m]=y}return{id:0,score:Math.trunc(100*a)/100,box:h.box,boxRaw:h.boxRaw,keypoints:u,annotations:p}}async function ob(e,t){let r=[e.shape[2]||0,e.shape[1]||0],n=(t.body.skipTime||0)>oe()-oT,a=ib<(t.body.skipFrames||0);if(t.skipAllowed&&n&&a&&Xm!==null)ib++;else{let s={};s.landmarks=await mAe(e,256),Xm=await AAe(s.landmarks,t,r),Object.keys(s).forEach(i=>te(s[i])),oT=oe(),ib=0}return Xm?[Xm]:[]}var Dd=[{class:1,label:"person"},{class:2,label:"bicycle"},{class:3,label:"car"},{class:4,label:"motorcycle"},{class:5,label:"airplane"},{class:6,label:"bus"},{class:7,label:"train"},{class:8,label:"truck"},{class:9,label:"boat"},{class:10,label:"traffic light"},{class:11,label:"fire hydrant"},{class:12,label:"stop sign"},{class:13,label:"parking meter"},{class:14,label:"bench"},{class:15,label:"bird"},{class:16,label:"cat"},{class:17,label:"dog"},{class:18,label:"horse"},{class:19,label:"sheep"},{class:20,label:"cow"},{class:21,label:"elephant"},{class:22,label:"bear"},{class:23,label:"zebra"},{class:24,label:"giraffe"},{class:25,label:"backpack"},{class:26,label:"umbrella"},{class:27,label:"handbag"},{class:28,label:"tie"},{class:29,label:"suitcase"},{class:30,label:"frisbee"},{class:31,label:"skis"},{class:32,label:"snowboard"},{class:33,label:"sports ball"},{class:34,label:"kite"},{class:35,label:"baseball bat"},{class:36,label:"baseball glove"},{class:37,label:"skateboard"},{class:38,label:"surfboard"},{class:39,label:"tennis racket"},{class:40,label:"bottle"},{class:41,label:"wine glass"},{class:42,label:"cup"},{class:43,label:"fork"},{class:44,label:"knife"},{class:45,label:"spoon"},{class:46,label:"bowl"},{class:47,label:"banana"},{class:48,label:"apple"},{class:49,label:"sandwich"},{class:50,label:"orange"},{class:51,label:"broccoli"},{class:52,label:"carrot"},{class:53,label:"hot dog"},{class:54,label:"pizza"},{class:55,label:"donut"},{class:56,label:"cake"},{class:57,label:"chair"},{class:58,label:"couch"},{class:59,label:"potted plant"},{class:60,label:"bed"},{class:61,label:"dining table"},{class:62,label:"toilet"},{class:63,label:"tv"},{class:64,label:"laptop"},{class:65,label:"mouse"},{class:66,label:"remote"},{class:67,label:"keyboard"},{class:68,label:"cell phone"},{class:69,label:"microwave"},{class:70,label:"oven"},{class:71,label:"toaster"},{class:72,label:"sink"},{class:73,label:"refrigerator"},{class:74,label:"book"},{class:75,label:"clock"},{class:76,label:"vase"},{class:77,label:"scissors"},{class:78,label:"teddy bear"},{class:79,label:"hair drier"},{class:80,label:"toothbrush"}];var us,Gl=0,lb=[],cT=0,ub=Number.MAX_SAFE_INTEGER;async function fT(e){if(he.initial&&(us=null),us)e.debug&&se("cached model:",us.modelUrl);else{us=await Ge(e.object.modelPath);let t=Object.values(us.modelSignature.inputs);Gl=Array.isArray(t)?parseInt(t[0].tensorShape.dim[2].size):0}return us}async function xAe(e,t,r){if(!e)return[];let n={},a=[],s=await e.array();n.squeeze=tt(e);let i=Yt(n.squeeze,6,1);n.stack=ur([i[1],i[0],i[3],i[2]],1),n.boxes=tt(n.stack),n.scores=tt(i[4]),n.classes=tt(i[5]),te([e,...i]),n.nms=await Ie.nonMaxSuppressionAsync(n.boxes,n.scores,r.object.maxDetected,r.object.iouThreshold,r.object.minConfidence||0);let o=await n.nms.data(),l=0;for(let u of Array.from(o)){let d=Math.trunc(100*s[0][u][4])/100,h=s[0][u][5],p=Dd[h].label,[c,f]=[s[0][u][0]/Gl,s[0][u][1]/Gl],m=[c,f,s[0][u][2]/Gl-c,s[0][u][3]/Gl-f],g=[Math.trunc(m[0]*t[0]),Math.trunc(m[1]*t[1]),Math.trunc(m[2]*t[0]),Math.trunc(m[3]*t[1])];a.push({id:l++,score:d,class:h,label:p,box:g,boxRaw:m})}return Object.keys(n).forEach(u=>te(n[u])),a}async function db(e,t){let r=(t.object.skipTime||0)>oe()-cT,n=ub<(t.object.skipFrames||0);return t.skipAllowed&&r&&n&&lb.length>0?(ub++,lb):(ub=0,new Promise(async a=>{let s=[e.shape[2]||0,e.shape[1]||0],i=Ie.resizeBilinear(e,[Gl,Gl]),o=t.object.enabled?us==null?void 0:us.execute(i,["tower_0/detections"]):null;cT=oe(),te(i);let l=await xAe(o,s,t);lb=l,a(l)}))}var Km={};vs(Km,{connected:()=>hb,kpt:()=>pb});var pb=["head","neck","rightShoulder","rightElbow","rightWrist","chest","leftShoulder","leftElbow","leftWrist","bodyCenter","rightHip","rightKnee","rightAnkle","leftHip","leftKnee","leftAnkle"],hb={leftLeg:["leftHip","leftKnee","leftAnkle"],rightLeg:["rightHip","rightKnee","rightAnkle"],torso:["leftShoulder","rightShoulder","rightHip","leftHip","leftShoulder"],leftArm:["leftShoulder","leftElbow","leftWrist"],rightArm:["rightShoulder","rightElbow","rightWrist"],head:[]};var $r,gT=0,Zr={id:0,keypoints:[],box:[0,0,0,0],boxRaw:[0,0,0,0],score:0,annotations:{}},cb=Number.MAX_SAFE_INTEGER;async function yT(e){return he.initial&&($r=null),$r?e.debug&&se("cached model:",$r.modelUrl):$r=await Ge(e.body.modelPath),$r}async function bAe(e,t){let[r,n]=e.shape,a=U(e,[n*r]),s=Ar(a,0),i=(await s.data())[0];if(te([a,s]),i>t){let o=Rn(a,0),l=md(o,r),u=(await l.data())[0],d=pe(o,Se(r,"int32")),h=(await d.data())[0];return te([l,d]),[u,h,i]}return[0,0,i]}async function fb(e,t){let r=(t.body.skipTime||0)>oe()-gT,n=cb<(t.body.skipFrames||0);return t.skipAllowed&&r&&n&&Object.keys(Zr.keypoints).length>0?(cb++,[Zr]):(cb=0,new Promise(async a=>{var h;let s=X(()=>{if(!($r!=null&&$r.inputs[0].shape))return null;let p=Ie.resizeBilinear(e,[$r.inputs[0].shape[2],$r.inputs[0].shape[1]],!1),c=L(p,Je.tf2);return ce(c,Je.tf1)}),i;if(t.body.enabled&&(i=$r==null?void 0:$r.execute(s)),gT=oe(),te(s),i){Zr.keypoints.length=0;let p=i.squeeze();te(i);let c=p.unstack(2);te(p);for(let f=0;f<c.length;f++){let[m,g,y]=await bAe(c[f],t.body.minConfidence);y>(((h=t.body)==null?void 0:h.minConfidence)||0)&&Zr.keypoints.push({score:Math.round(100*y)/100,part:pb[f],positionRaw:[m/$r.inputs[0].shape[2],g/$r.inputs[0].shape[1]],position:[Math.round(e.shape[2]*m/$r.inputs[0].shape[2]),Math.round(e.shape[1]*g/$r.inputs[0].shape[1])]})}c.forEach(f=>te(f))}Zr.score=Zr.keypoints.reduce((p,c)=>c.score>p?c.score:p,0);let o=Zr.keypoints.map(p=>p.position[0]),l=Zr.keypoints.map(p=>p.position[1]);Zr.box=[Math.min(...o),Math.min(...l),Math.max(...o)-Math.min(...o),Math.max(...l)-Math.min(...l)];let u=Zr.keypoints.map(p=>p.positionRaw[0]),d=Zr.keypoints.map(p=>p.positionRaw[1]);Zr.boxRaw=[Math.min(...u),Math.min(...d),Math.max(...u)-Math.min(...u),Math.max(...d)-Math.min(...d)];for(let[p,c]of Object.entries(hb)){let f=[];for(let m=0;m<c.length-1;m++){let g=Zr.keypoints.find(A=>A.part===c[m]),y=Zr.keypoints.find(A=>A.part===c[m+1]);g&&y&&g.score>(t.body.minConfidence||0)&&y.score>(t.body.minConfidence||0)&&f.push([g.position,y.position])}Zr.annotations[p]=f}a([Zr])}))}var vAe=["angry","disgust","fear","happy","sad","surprise","neutral"],Bn,Zm=[],xT=0,bT=0,mb=Number.MAX_SAFE_INTEGER;async function vT(e){var t;return he.initial&&(Bn=null),Bn?e.debug&&se("cached model:",Bn.modelUrl):Bn=await Ge((t=e.face.emotion)==null?void 0:t.modelPath),Bn}async function gb(e,t,r,n){var i,o;if(!Bn)return[];let a=mb<(((i=t.face.emotion)==null?void 0:i.skipFrames)||0),s=(((o=t.face.emotion)==null?void 0:o.skipTime)||0)>oe()-bT;return t.skipAllowed&&s&&a&&xT===n&&Zm[r]&&Zm[r].length>0?(mb++,Zm[r]):(mb=0,new Promise(async l=>{var d,h;let u=[];if((d=t.face.emotion)!=null&&d.enabled){let p={},c=Bn!=null&&Bn.inputs[0].shape?Bn.inputs[0].shape[2]:0;p.resize=Ie.resizeBilinear(e,[c,c],!1),p.channels=L(p.resize,Je.rgb),p.grayscale=ke(p.channels,3,!0),p.grayscaleSub=ce(p.grayscale,Je.tf05),p.grayscaleMul=L(p.grayscaleSub,Je.tf2),p.emotion=Bn==null?void 0:Bn.execute(p.grayscaleMul),bT=oe();let f=await p.emotion.data();for(let m=0;m<f.length;m++)f[m]>(((h=t.face.emotion)==null?void 0:h.minConfidence)||0)&&u.push({score:Math.min(.99,Math.trunc(100*f[m])/100),emotion:vAe[m]});u.sort((m,g)=>g.score-m.score),Object.keys(p).forEach(m=>te(p[m]))}Zm[r]=u,xT=n,l(u)}))}var An,yb=[],kT=0,IT=0,ST=Number.MAX_SAFE_INTEGER;async function CT(e){return he.initial&&(An=null),An?e.debug&&se("cached model:",An.modelUrl):An=await Ge(e.face.mobilefacenet.modelPath),An}async function Ab(e,t,r,n){var i,o;if(!An)return[];let a=ST<(((i=t.face.embedding)==null?void 0:i.skipFrames)||0),s=(((o=t.face.embedding)==null?void 0:o.skipTime)||0)>oe()-IT;return t.skipAllowed&&s&&a&&kT===n&&yb[r]?(ST++,yb[r]):new Promise(async l=>{var d;let u=[];if(((d=t.face.embedding)==null?void 0:d.enabled)&&(An==null?void 0:An.inputs[0].shape)){let h={};h.crop=Ie.resizeBilinear(e,[An.inputs[0].shape[2],An.inputs[0].shape[1]],!1),h.data=An==null?void 0:An.execute(h.crop);let p=await h.data.data();u=Array.from(p)}yb[r]=u,kT=n,IT=oe(),l(u)})}var ds,to=0,wAe=2.3,xb=ea.leftEyeLower0,bb=ea.rightEyeLower0,Ld={leftBounds:[xb[0],xb[xb.length-1]],rightBounds:[bb[0],bb[bb.length-1]]},Bd={upperCenter:3,lowerCenter:4,index:71,numCoordinates:76};async function $T(e){var t;return he.initial&&(ds=null),ds?e.debug&&se("cached model:",ds.modelUrl):ds=await Ge((t=e.face.iris)==null?void 0:t.modelPath),to=ds.inputs[0].shape?ds.inputs[0].shape[2]:0,to===-1&&(to=64),ds}function Ym(e,t,r,n){for(let a=0;a<Y5.length;a++){let{key:s,indices:i}=Y5[a],o=ea[`${r}${s}`];if(!n||n.includes(s))for(let l=0;l<i.length;l++){let u=i[l];e[o[l]]=[t[u][0],t[u][1],(t[u][2]+e[o[l]][2])/2]}}}var kAe=e=>{let t=e[Ld.leftBounds[0]][2],r=e[Ld.rightBounds[0]][2];return t-r},NT=(e,t,r,n,a,s=!1)=>{let i=Vm(Wm(qC([e[r],e[n]]),wAe)),o=zd(i),l=Ie.cropAndResize(t,[[i.startPoint[1]/a,i.startPoint[0]/a,i.endPoint[1]/a,i.endPoint[0]/a]],[0],[to,to]);if(s&&he.kernels.includes("flipleftright")){let u=Ie.flipLeftRight(l);te(l),l=u}return{box:i,boxSize:o,crop:l}},ET=(e,t,r,n=!1)=>{let a=[];for(let s=0;s<Bd.numCoordinates;s++){let i=e[s*3],o=e[s*3+1],l=e[s*3+2];a.push([(n?1-i/to:i/to)*r[0]+t.startPoint[0],o/to*r[1]+t.startPoint[1],l])}return{rawCoords:a,iris:a.slice(Bd.index)}},RT=(e,t,r)=>{let n=e[ea[`${r}EyeUpper0`][Bd.upperCenter]][2],a=e[ea[`${r}EyeLower0`][Bd.lowerCenter]][2],s=(n+a)/2;return t.map((i,o)=>{let l=s;return o===2?l=n:o===4&&(l=a),[i[0],i[1],l]})};async function MT(e,t,r,n){if(!ds)return r.debug&&se("face mesh iris detection requested, but model is not loaded"),e;let{box:a,boxSize:s,crop:i}=NT(e,t,Ld.leftBounds[0],Ld.leftBounds[1],n,!0),{box:o,boxSize:l,crop:u}=NT(e,t,Ld.rightBounds[0],Ld.rightBounds[1],n,!0),d=St([i,u]);te(i),te(u);let h=ds.execute(d);te(d);let p=await h.data();te(h);let c=p.slice(0,Bd.numCoordinates*3),{rawCoords:f,iris:m}=ET(c,a,s,!0),g=p.slice(Bd.numCoordinates*3),{rawCoords:y,iris:A}=ET(g,o,l,!1),x=kAe(e);Math.abs(x)<30?(Ym(e,f,"left",null),Ym(e,y,"right",null)):x<1?Ym(e,f,"left",["EyeUpper0","EyeLower0"]):Ym(e,y,"right",["EyeUpper0","EyeLower0"]);let b=RT(e,m,"left"),w=RT(e,A,"right");return e.concat(b).concat(w)}var xn={eyeLLower:[33,7,163,144,145,153,154,155,133],eyeRLower:[263,249,390,373,374,380,381,382,362],lips:[185,96,90,181,84,17,314,405,320,307,409,40,39,73,37,0,267,269,270,409,40,88,178,178,87,14,268,402,318,324,409,80,41,38,87,12,268,303,318,324,185,95,80,81,85,16,315,404,319,325,409,40,39,73,72,0,302,303,270,408,185,88,88,81,82,15,316,403,319,324,409,80,41,38,87,12,268,303,318,324],eyeL:[33,7,163,144,145,153,154,155,133,246,161,160,159,158,157,173,130,25,110,24,23,22,26,112,243,247,30,29,27,28,56,190,226,31,228,229,230,231,232,233,244,113,225,224,223,222,221,189,35,124,46,53,52,65,143,111,117,118,119,120,121,128,245,156,70,63,105,66,107,55,193],eyeR:[263,249,390,373,374,380,381,382,362,466,388,387,386,385,384,398,359,255,339,254,253,252,256,341,463,467,260,259,257,258,286,414,446,261,448,449,450,451,452,453,464,342,445,444,443,442,441,413,265,353,276,283,282,295,372,340,346,347,348,349,350,357,465,383,300,293,334,296,336,285,417]};async function PT(e,t){let r={irisL:t[3].dataSync(),irisR:t[1].dataSync(),eyeL:t[0].dataSync(),eyeR:t[6].dataSync(),lips:t[5].dataSync()},n=xn.eyeRLower.reduce((s,i)=>s+=e[i][2],0)/xn.eyeRLower.length;for(let s=0;s<r.irisR.length/2;s++)e.push([r.irisR[2*s+0],r.irisR[2*s+1],n]);let a=xn.eyeLLower.reduce((s,i)=>s+=e[i][2],0)/xn.eyeLLower.length;for(let s=0;s<r.irisL.length/2;s++)e.push([r.irisL[2*s+0],r.irisL[2*s+1],a]);for(let s=0;s<r.eyeL.length/2;s++)e[xn.eyeL[s]]=[r.eyeL[2*s+0],r.eyeL[2*s+1],e[xn.eyeL[s]][2]];for(let s=0;s<r.eyeR.length/2;s++)e[xn.eyeR[s]]=[r.eyeR[2*s+0],r.eyeR[2*s+1],e[xn.eyeR[s]][2]];for(let s=0;s<r.lips.length/2;s++)e[xn.lips[s]]=[r.lips[2*s+0],r.lips[2*s+1],e[xn.lips[s]][2]];return e}var La={boxes:[],skipped:Number.MAX_SAFE_INTEGER,timestamp:0},Ba=null,Wd=0;async function _T(e,t){var o,l,u,d,h,p,c,f,m,g;let r=(((o=t.face.detector)==null?void 0:o.skipTime)||0)>oe()-La.timestamp,n=La.skipped<(((l=t.face.detector)==null?void 0:l.skipFrames)||0);!t.skipAllowed||!r||!n||La.boxes.length===0?(La.boxes=await rT(e,t),La.timestamp=oe(),La.skipped=0):La.skipped++;let a=[],s=[],i=0;for(let y=0;y<La.boxes.length;y++){let A=La.boxes[y],x=0,b,w={id:i++,mesh:[],meshRaw:[],box:[0,0,0,0],boxRaw:[0,0,0,0],score:0,boxScore:0,faceScore:0,annotations:{}};if([x,b,w.tensor]=YC((u=t.face.detector)==null?void 0:u.rotation,A,e,(d=t.face.mesh)!=null&&d.enabled?Wd:Um()),(h=t==null?void 0:t.filter)!=null&&h.equalization){let S=await Pm(w.tensor);te(w.tensor),w.tensor=S}if(w.boxScore=Math.round(100*A.confidence)/100,(p=t.face.mesh)!=null&&p.enabled)if(!Ba)t.debug&&se("face mesh detection requested, but model is not loaded");else{let S=Ba.execute(w.tensor),N=S.find(I=>I.shape[I.shape.length-1]===1),E=S.find(I=>I.shape[I.shape.length-1]===1404),R=await N.data();w.faceScore=Math.round(100*R[0])/100;let z=U(E,[-1,3]),M=await z.array();if(w.faceScore<(((c=t.face.detector)==null?void 0:c.minConfidence)||1))A.confidence=w.faceScore;else{(f=t.face.attention)!=null&&f.enabled?M=await PT(M,S):(m=t.face.iris)!=null&&m.enabled&&(M=await MT(M,w.tensor,t,Wd)),w.mesh=ZC(M,A,x,b,Wd),w.meshRaw=w.mesh.map(_=>[_[0]/(e.shape[2]||0),_[1]/(e.shape[1]||0),(_[2]||0)/Wd]);for(let _ of Object.keys(ea))w.annotations[_]=ea[_].map(O=>w.mesh[O]);w.score=w.faceScore;let I={...JC(w.mesh,A),confidence:A.confidence,landmarks:A.landmarks};w.box=tb(I,e),w.boxRaw=rb(I,e),s.push(I)}te([...S,z])}else{w.box=tb(A,e),w.boxRaw=rb(A,e),w.score=w.boxScore,w.mesh=A.landmarks.map(S=>[(A.startPoint[0]+A.endPoint[0])/2+(A.endPoint[0]+A.startPoint[0])*S[0]/Um(),(A.startPoint[1]+A.endPoint[1])/2+(A.endPoint[1]+A.startPoint[1])*S[1]/Um()]),w.meshRaw=w.mesh.map(S=>[S[0]/(e.shape[2]||0),S[1]/(e.shape[1]||0),(S[2]||0)/Wd]);for(let S of Object.keys(Yh))w.annotations[S]=[w.mesh[Yh[S]]]}w.score>(((g=t.face.detector)==null?void 0:g.minConfidence)||1)?a.push(w):te(w.tensor)}return La.boxes=s,a}async function zT(e){var t,r,n;return he.initial&&(Ba=null),Ba?e.debug&&se("cached model:",Ba.modelUrl):(t=e.face.attention)!=null&&t.enabled?Ba=await Ge((r=e.face.attention)==null?void 0:r.modelPath):Ba=await Ge((n=e.face.mesh)==null?void 0:n.modelPath),Wd=Ba.inputs[0].shape?Ba.inputs[0].shape[2]:0,Ba}var OT=Vl,DT=Jh;var bn,Jm=[],LT=0,BT=0,wb=Number.MAX_SAFE_INTEGER;async function WT(e){var t;return he.initial&&(bn=null),bn?e.debug&&se("cached model:",bn.modelUrl):bn=await Ge((t=e.face.description)==null?void 0:t.modelPath),bn}function kb(e){let t=e.image||e.tensor||e;if(!(bn!=null&&bn.inputs[0].shape))return t;let r=Ie.resizeBilinear(t,[bn.inputs[0].shape[2],bn.inputs[0].shape[1]],!1),n=L(r,Je.tf255);return te(r),n}async function Ib(e,t,r,n){var i,o,l,u;if(!bn)return{age:0,gender:"unknown",genderScore:0,descriptor:[]};let a=wb<(((i=t.face.description)==null?void 0:i.skipFrames)||0),s=(((o=t.face.description)==null?void 0:o.skipTime)||0)>oe()-LT;return t.skipAllowed&&a&&s&&BT===n&&((l=Jm[r])==null?void 0:l.age)&&((u=Jm[r])==null?void 0:u.age)>0?(wb++,Jm[r]):(wb=0,new Promise(async d=>{var p,c;let h={age:0,gender:"unknown",genderScore:0,descriptor:[]};if((p=t.face.description)!=null&&p.enabled){let f=kb(e),m=bn==null?void 0:bn.execute(f);LT=oe(),te(f);let y=await(await m.find(R=>R.shape[1]===1)).data(),A=Math.trunc(200*Math.abs(y[0]-.5))/100;A>(((c=t.face.description)==null?void 0:c.minConfidence)||0)&&(h.gender=y[0]<=.5?"female":"male",h.genderScore=Math.min(.99,A));let x=Rn(m.find(R=>R.shape[1]===100),1),b=(await x.data())[0];te(x);let S=await m.find(R=>R.shape[1]===100).data();h.age=Math.round(S[b-1]>S[b+1]?10*b-100*S[b-1]:10*b+100*S[b+1])/10;let N=m.find(R=>R.shape[1]===1024),E=N?await N.data():[];h.descriptor=Array.from(E),m.forEach(R=>te(R))}Jm[r]=h,BT=n,d(h)}))}function Qm(e){return[Math.abs(e.endPoint[0]-e.startPoint[0]),Math.abs(e.endPoint[1]-e.startPoint[1])]}function tc(e){return[e.startPoint[0]+(e.endPoint[0]-e.startPoint[0])/2,e.startPoint[1]+(e.endPoint[1]-e.startPoint[1])/2]}function GT(e,t,r){let n=t.shape[1],a=t.shape[2],s=[[e.startPoint[1]/n,e.startPoint[0]/a,e.endPoint[1]/n,e.endPoint[0]/a]];return Ie.cropAndResize(t,s,[0],r)}function jT(e,t){let r=[e.startPoint[0]*t[0],e.startPoint[1]*t[1]],n=[e.endPoint[0]*t[0],e.endPoint[1]*t[1]],a=e.palmLandmarks.map(s=>[s[0]*t[0],s[1]*t[1]]);return{startPoint:r,endPoint:n,palmLandmarks:a,confidence:e.confidence}}function e1(e,t=1.5){let r=tc(e),n=Qm(e),a=[t*n[0]/2,t*n[1]/2],s=[r[0]-a[0],r[1]-a[1]],i=[r[0]+a[0],r[1]+a[1]];return{startPoint:s,endPoint:i,palmLandmarks:e.palmLandmarks}}function t1(e){let t=tc(e),r=Qm(e),a=Math.max(...r)/2,s=[t[0]-a,t[1]-a],i=[t[0]+a,t[1]+a];return{startPoint:s,endPoint:i,palmLandmarks:e.palmLandmarks}}function SAe(e){return e-2*Math.PI*Math.floor((e+Math.PI)/(2*Math.PI))}function HT(e,t){let r=Math.PI/2-Math.atan2(-(t[1]-e[1]),t[0]-e[0]);return SAe(r)}var VT=(e,t)=>[[1,0,e],[0,1,t],[0,0,1]];function ro(e,t){let r=0;for(let n=0;n<e.length;n++)r+=e[n]*t[n];return r}function CAe(e,t){let r=[];for(let n=0;n<e.length;n++)r.push(e[n][t]);return r}function UT(e,t){let r=[],n=e.length;for(let a=0;a<n;a++){r.push([]);for(let s=0;s<n;s++)r[a].push(ro(e[a],CAe(t,s)))}return r}function Cb(e,t){let r=Math.cos(e),n=Math.sin(e),a=[[r,-n,0],[n,r,0],[0,0,1]],s=VT(t[0],t[1]),i=UT(s,a),o=VT(-t[0],-t[1]);return UT(i,o)}function qT(e){let t=[[e[0][0],e[1][0]],[e[0][1],e[1][1]]],r=[e[0][2],e[1][2]],n=[-ro(t[0],r),-ro(t[1],r)];return[t[0].concat(n[0]),t[1].concat(n[1]),[0,0,1]]}function Tb(e,t){return[ro(e,t[0]),ro(e,t[1])]}var KT=[{x:.015625,y:.015625},{x:.015625,y:.015625},{x:.046875,y:.015625},{x:.046875,y:.015625},{x:.078125,y:.015625},{x:.078125,y:.015625},{x:.109375,y:.015625},{x:.109375,y:.015625},{x:.140625,y:.015625},{x:.140625,y:.015625},{x:.171875,y:.015625},{x:.171875,y:.015625},{x:.203125,y:.015625},{x:.203125,y:.015625},{x:.234375,y:.015625},{x:.234375,y:.015625},{x:.265625,y:.015625},{x:.265625,y:.015625},{x:.296875,y:.015625},{x:.296875,y:.015625},{x:.328125,y:.015625},{x:.328125,y:.015625},{x:.359375,y:.015625},{x:.359375,y:.015625},{x:.390625,y:.015625},{x:.390625,y:.015625},{x:.421875,y:.015625},{x:.421875,y:.015625},{x:.453125,y:.015625},{x:.453125,y:.015625},{x:.484375,y:.015625},{x:.484375,y:.015625},{x:.515625,y:.015625},{x:.515625,y:.015625},{x:.546875,y:.015625},{x:.546875,y:.015625},{x:.578125,y:.015625},{x:.578125,y:.015625},{x:.609375,y:.015625},{x:.609375,y:.015625},{x:.640625,y:.015625},{x:.640625,y:.015625},{x:.671875,y:.015625},{x:.671875,y:.015625},{x:.703125,y:.015625},{x:.703125,y:.015625},{x:.734375,y:.015625},{x:.734375,y:.015625},{x:.765625,y:.015625},{x:.765625,y:.015625},{x:.796875,y:.015625},{x:.796875,y:.015625},{x:.828125,y:.015625},{x:.828125,y:.015625},{x:.859375,y:.015625},{x:.859375,y:.015625},{x:.890625,y:.015625},{x:.890625,y:.015625},{x:.921875,y:.015625},{x:.921875,y:.015625},{x:.953125,y:.015625},{x:.953125,y:.015625},{x:.984375,y:.015625},{x:.984375,y:.015625},{x:.015625,y:.046875},{x:.015625,y:.046875},{x:.046875,y:.046875},{x:.046875,y:.046875},{x:.078125,y:.046875},{x:.078125,y:.046875},{x:.109375,y:.046875},{x:.109375,y:.046875},{x:.140625,y:.046875},{x:.140625,y:.046875},{x:.171875,y:.046875},{x:.171875,y:.046875},{x:.203125,y:.046875},{x:.203125,y:.046875},{x:.234375,y:.046875},{x:.234375,y:.046875},{x:.265625,y:.046875},{x:.265625,y:.046875},{x:.296875,y:.046875},{x:.296875,y:.046875},{x:.328125,y:.046875},{x:.328125,y:.046875},{x:.359375,y:.046875},{x:.359375,y:.046875},{x:.390625,y:.046875},{x:.390625,y:.046875},{x:.421875,y:.046875},{x:.421875,y:.046875},{x:.453125,y:.046875},{x:.453125,y:.046875},{x:.484375,y:.046875},{x:.484375,y:.046875},{x:.515625,y:.046875},{x:.515625,y:.046875},{x:.546875,y:.046875},{x:.546875,y:.046875},{x:.578125,y:.046875},{x:.578125,y:.046875},{x:.609375,y:.046875},{x:.609375,y:.046875},{x:.640625,y:.046875},{x:.640625,y:.046875},{x:.671875,y:.046875},{x:.671875,y:.046875},{x:.703125,y:.046875},{x:.703125,y:.046875},{x:.734375,y:.046875},{x:.734375,y:.046875},{x:.765625,y:.046875},{x:.765625,y:.046875},{x:.796875,y:.046875},{x:.796875,y:.046875},{x:.828125,y:.046875},{x:.828125,y:.046875},{x:.859375,y:.046875},{x:.859375,y:.046875},{x:.890625,y:.046875},{x:.890625,y:.046875},{x:.921875,y:.046875},{x:.921875,y:.046875},{x:.953125,y:.046875},{x:.953125,y:.046875},{x:.984375,y:.046875},{x:.984375,y:.046875},{x:.015625,y:.078125},{x:.015625,y:.078125},{x:.046875,y:.078125},{x:.046875,y:.078125},{x:.078125,y:.078125},{x:.078125,y:.078125},{x:.109375,y:.078125},{x:.109375,y:.078125},{x:.140625,y:.078125},{x:.140625,y:.078125},{x:.171875,y:.078125},{x:.171875,y:.078125},{x:.203125,y:.078125},{x:.203125,y:.078125},{x:.234375,y:.078125},{x:.234375,y:.078125},{x:.265625,y:.078125},{x:.265625,y:.078125},{x:.296875,y:.078125},{x:.296875,y:.078125},{x:.328125,y:.078125},{x:.328125,y:.078125},{x:.359375,y:.078125},{x:.359375,y:.078125},{x:.390625,y:.078125},{x:.390625,y:.078125},{x:.421875,y:.078125},{x:.421875,y:.078125},{x:.453125,y:.078125},{x:.453125,y:.078125},{x:.484375,y:.078125},{x:.484375,y:.078125},{x:.515625,y:.078125},{x:.515625,y:.078125},{x:.546875,y:.078125},{x:.546875,y:.078125},{x:.578125,y:.078125},{x:.578125,y:.078125},{x:.609375,y:.078125},{x:.609375,y:.078125},{x:.640625,y:.078125},{x:.640625,y:.078125},{x:.671875,y:.078125},{x:.671875,y:.078125},{x:.703125,y:.078125},{x:.703125,y:.078125},{x:.734375,y:.078125},{x:.734375,y:.078125},{x:.765625,y:.078125},{x:.765625,y:.078125},{x:.796875,y:.078125},{x:.796875,y:.078125},{x:.828125,y:.078125},{x:.828125,y:.078125},{x:.859375,y:.078125},{x:.859375,y:.078125},{x:.890625,y:.078125},{x:.890625,y:.078125},{x:.921875,y:.078125},{x:.921875,y:.078125},{x:.953125,y:.078125},{x:.953125,y:.078125},{x:.984375,y:.078125},{x:.984375,y:.078125},{x:.015625,y:.109375},{x:.015625,y:.109375},{x:.046875,y:.109375},{x:.046875,y:.109375},{x:.078125,y:.109375},{x:.078125,y:.109375},{x:.109375,y:.109375},{x:.109375,y:.109375},{x:.140625,y:.109375},{x:.140625,y:.109375},{x:.171875,y:.109375},{x:.171875,y:.109375},{x:.203125,y:.109375},{x:.203125,y:.109375},{x:.234375,y:.109375},{x:.234375,y:.109375},{x:.265625,y:.109375},{x:.265625,y:.109375},{x:.296875,y:.109375},{x:.296875,y:.109375},{x:.328125,y:.109375},{x:.328125,y:.109375},{x:.359375,y:.109375},{x:.359375,y:.109375},{x:.390625,y:.109375},{x:.390625,y:.109375},{x:.421875,y:.109375},{x:.421875,y:.109375},{x:.453125,y:.109375},{x:.453125,y:.109375},{x:.484375,y:.109375},{x:.484375,y:.109375},{x:.515625,y:.109375},{x:.515625,y:.109375},{x:.546875,y:.109375},{x:.546875,y:.109375},{x:.578125,y:.109375},{x:.578125,y:.109375},{x:.609375,y:.109375},{x:.609375,y:.109375},{x:.640625,y:.109375},{x:.640625,y:.109375},{x:.671875,y:.109375},{x:.671875,y:.109375},{x:.703125,y:.109375},{x:.703125,y:.109375},{x:.734375,y:.109375},{x:.734375,y:.109375},{x:.765625,y:.109375},{x:.765625,y:.109375},{x:.796875,y:.109375},{x:.796875,y:.109375},{x:.828125,y:.109375},{x:.828125,y:.109375},{x:.859375,y:.109375},{x:.859375,y:.109375},{x:.890625,y:.109375},{x:.890625,y:.109375},{x:.921875,y:.109375},{x:.921875,y:.109375},{x:.953125,y:.109375},{x:.953125,y:.109375},{x:.984375,y:.109375},{x:.984375,y:.109375},{x:.015625,y:.140625},{x:.015625,y:.140625},{x:.046875,y:.140625},{x:.046875,y:.140625},{x:.078125,y:.140625},{x:.078125,y:.140625},{x:.109375,y:.140625},{x:.109375,y:.140625},{x:.140625,y:.140625},{x:.140625,y:.140625},{x:.171875,y:.140625},{x:.171875,y:.140625},{x:.203125,y:.140625},{x:.203125,y:.140625},{x:.234375,y:.140625},{x:.234375,y:.140625},{x:.265625,y:.140625},{x:.265625,y:.140625},{x:.296875,y:.140625},{x:.296875,y:.140625},{x:.328125,y:.140625},{x:.328125,y:.140625},{x:.359375,y:.140625},{x:.359375,y:.140625},{x:.390625,y:.140625},{x:.390625,y:.140625},{x:.421875,y:.140625},{x:.421875,y:.140625},{x:.453125,y:.140625},{x:.453125,y:.140625},{x:.484375,y:.140625},{x:.484375,y:.140625},{x:.515625,y:.140625},{x:.515625,y:.140625},{x:.546875,y:.140625},{x:.546875,y:.140625},{x:.578125,y:.140625},{x:.578125,y:.140625},{x:.609375,y:.140625},{x:.609375,y:.140625},{x:.640625,y:.140625},{x:.640625,y:.140625},{x:.671875,y:.140625},{x:.671875,y:.140625},{x:.703125,y:.140625},{x:.703125,y:.140625},{x:.734375,y:.140625},{x:.734375,y:.140625},{x:.765625,y:.140625},{x:.765625,y:.140625},{x:.796875,y:.140625},{x:.796875,y:.140625},{x:.828125,y:.140625},{x:.828125,y:.140625},{x:.859375,y:.140625},{x:.859375,y:.140625},{x:.890625,y:.140625},{x:.890625,y:.140625},{x:.921875,y:.140625},{x:.921875,y:.140625},{x:.953125,y:.140625},{x:.953125,y:.140625},{x:.984375,y:.140625},{x:.984375,y:.140625},{x:.015625,y:.171875},{x:.015625,y:.171875},{x:.046875,y:.171875},{x:.046875,y:.171875},{x:.078125,y:.171875},{x:.078125,y:.171875},{x:.109375,y:.171875},{x:.109375,y:.171875},{x:.140625,y:.171875},{x:.140625,y:.171875},{x:.171875,y:.171875},{x:.171875,y:.171875},{x:.203125,y:.171875},{x:.203125,y:.171875},{x:.234375,y:.171875},{x:.234375,y:.171875},{x:.265625,y:.171875},{x:.265625,y:.171875},{x:.296875,y:.171875},{x:.296875,y:.171875},{x:.328125,y:.171875},{x:.328125,y:.171875},{x:.359375,y:.171875},{x:.359375,y:.171875},{x:.390625,y:.171875},{x:.390625,y:.171875},{x:.421875,y:.171875},{x:.421875,y:.171875},{x:.453125,y:.171875},{x:.453125,y:.171875},{x:.484375,y:.171875},{x:.484375,y:.171875},{x:.515625,y:.171875},{x:.515625,y:.171875},{x:.546875,y:.171875},{x:.546875,y:.171875},{x:.578125,y:.171875},{x:.578125,y:.171875},{x:.609375,y:.171875},{x:.609375,y:.171875},{x:.640625,y:.171875},{x:.640625,y:.171875},{x:.671875,y:.171875},{x:.671875,y:.171875},{x:.703125,y:.171875},{x:.703125,y:.171875},{x:.734375,y:.171875},{x:.734375,y:.171875},{x:.765625,y:.171875},{x:.765625,y:.171875},{x:.796875,y:.171875},{x:.796875,y:.171875},{x:.828125,y:.171875},{x:.828125,y:.171875},{x:.859375,y:.171875},{x:.859375,y:.171875},{x:.890625,y:.171875},{x:.890625,y:.171875},{x:.921875,y:.171875},{x:.921875,y:.171875},{x:.953125,y:.171875},{x:.953125,y:.171875},{x:.984375,y:.171875},{x:.984375,y:.171875},{x:.015625,y:.203125},{x:.015625,y:.203125},{x:.046875,y:.203125},{x:.046875,y:.203125},{x:.078125,y:.203125},{x:.078125,y:.203125},{x:.109375,y:.203125},{x:.109375,y:.203125},{x:.140625,y:.203125},{x:.140625,y:.203125},{x:.171875,y:.203125},{x:.171875,y:.203125},{x:.203125,y:.203125},{x:.203125,y:.203125},{x:.234375,y:.203125},{x:.234375,y:.203125},{x:.265625,y:.203125},{x:.265625,y:.203125},{x:.296875,y:.203125},{x:.296875,y:.203125},{x:.328125,y:.203125},{x:.328125,y:.203125},{x:.359375,y:.203125},{x:.359375,y:.203125},{x:.390625,y:.203125},{x:.390625,y:.203125},{x:.421875,y:.203125},{x:.421875,y:.203125},{x:.453125,y:.203125},{x:.453125,y:.203125},{x:.484375,y:.203125},{x:.484375,y:.203125},{x:.515625,y:.203125},{x:.515625,y:.203125},{x:.546875,y:.203125},{x:.546875,y:.203125},{x:.578125,y:.203125},{x:.578125,y:.203125},{x:.609375,y:.203125},{x:.609375,y:.203125},{x:.640625,y:.203125},{x:.640625,y:.203125},{x:.671875,y:.203125},{x:.671875,y:.203125},{x:.703125,y:.203125},{x:.703125,y:.203125},{x:.734375,y:.203125},{x:.734375,y:.203125},{x:.765625,y:.203125},{x:.765625,y:.203125},{x:.796875,y:.203125},{x:.796875,y:.203125},{x:.828125,y:.203125},{x:.828125,y:.203125},{x:.859375,y:.203125},{x:.859375,y:.203125},{x:.890625,y:.203125},{x:.890625,y:.203125},{x:.921875,y:.203125},{x:.921875,y:.203125},{x:.953125,y:.203125},{x:.953125,y:.203125},{x:.984375,y:.203125},{x:.984375,y:.203125},{x:.015625,y:.234375},{x:.015625,y:.234375},{x:.046875,y:.234375},{x:.046875,y:.234375},{x:.078125,y:.234375},{x:.078125,y:.234375},{x:.109375,y:.234375},{x:.109375,y:.234375},{x:.140625,y:.234375},{x:.140625,y:.234375},{x:.171875,y:.234375},{x:.171875,y:.234375},{x:.203125,y:.234375},{x:.203125,y:.234375},{x:.234375,y:.234375},{x:.234375,y:.234375},{x:.265625,y:.234375},{x:.265625,y:.234375},{x:.296875,y:.234375},{x:.296875,y:.234375},{x:.328125,y:.234375},{x:.328125,y:.234375},{x:.359375,y:.234375},{x:.359375,y:.234375},{x:.390625,y:.234375},{x:.390625,y:.234375},{x:.421875,y:.234375},{x:.421875,y:.234375},{x:.453125,y:.234375},{x:.453125,y:.234375},{x:.484375,y:.234375},{x:.484375,y:.234375},{x:.515625,y:.234375},{x:.515625,y:.234375},{x:.546875,y:.234375},{x:.546875,y:.234375},{x:.578125,y:.234375},{x:.578125,y:.234375},{x:.609375,y:.234375},{x:.609375,y:.234375},{x:.640625,y:.234375},{x:.640625,y:.234375},{x:.671875,y:.234375},{x:.671875,y:.234375},{x:.703125,y:.234375},{x:.703125,y:.234375},{x:.734375,y:.234375},{x:.734375,y:.234375},{x:.765625,y:.234375},{x:.765625,y:.234375},{x:.796875,y:.234375},{x:.796875,y:.234375},{x:.828125,y:.234375},{x:.828125,y:.234375},{x:.859375,y:.234375},{x:.859375,y:.234375},{x:.890625,y:.234375},{x:.890625,y:.234375},{x:.921875,y:.234375},{x:.921875,y:.234375},{x:.953125,y:.234375},{x:.953125,y:.234375},{x:.984375,y:.234375},{x:.984375,y:.234375},{x:.015625,y:.265625},{x:.015625,y:.265625},{x:.046875,y:.265625},{x:.046875,y:.265625},{x:.078125,y:.265625},{x:.078125,y:.265625},{x:.109375,y:.265625},{x:.109375,y:.265625},{x:.140625,y:.265625},{x:.140625,y:.265625},{x:.171875,y:.265625},{x:.171875,y:.265625},{x:.203125,y:.265625},{x:.203125,y:.265625},{x:.234375,y:.265625},{x:.234375,y:.265625},{x:.265625,y:.265625},{x:.265625,y:.265625},{x:.296875,y:.265625},{x:.296875,y:.265625},{x:.328125,y:.265625},{x:.328125,y:.265625},{x:.359375,y:.265625},{x:.359375,y:.265625},{x:.390625,y:.265625},{x:.390625,y:.265625},{x:.421875,y:.265625},{x:.421875,y:.265625},{x:.453125,y:.265625},{x:.453125,y:.265625},{x:.484375,y:.265625},{x:.484375,y:.265625},{x:.515625,y:.265625},{x:.515625,y:.265625},{x:.546875,y:.265625},{x:.546875,y:.265625},{x:.578125,y:.265625},{x:.578125,y:.265625},{x:.609375,y:.265625},{x:.609375,y:.265625},{x:.640625,y:.265625},{x:.640625,y:.265625},{x:.671875,y:.265625},{x:.671875,y:.265625},{x:.703125,y:.265625},{x:.703125,y:.265625},{x:.734375,y:.265625},{x:.734375,y:.265625},{x:.765625,y:.265625},{x:.765625,y:.265625},{x:.796875,y:.265625},{x:.796875,y:.265625},{x:.828125,y:.265625},{x:.828125,y:.265625},{x:.859375,y:.265625},{x:.859375,y:.265625},{x:.890625,y:.265625},{x:.890625,y:.265625},{x:.921875,y:.265625},{x:.921875,y:.265625},{x:.953125,y:.265625},{x:.953125,y:.265625},{x:.984375,y:.265625},{x:.984375,y:.265625},{x:.015625,y:.296875},{x:.015625,y:.296875},{x:.046875,y:.296875},{x:.046875,y:.296875},{x:.078125,y:.296875},{x:.078125,y:.296875},{x:.109375,y:.296875},{x:.109375,y:.296875},{x:.140625,y:.296875},{x:.140625,y:.296875},{x:.171875,y:.296875},{x:.171875,y:.296875},{x:.203125,y:.296875},{x:.203125,y:.296875},{x:.234375,y:.296875},{x:.234375,y:.296875},{x:.265625,y:.296875},{x:.265625,y:.296875},{x:.296875,y:.296875},{x:.296875,y:.296875},{x:.328125,y:.296875},{x:.328125,y:.296875},{x:.359375,y:.296875},{x:.359375,y:.296875},{x:.390625,y:.296875},{x:.390625,y:.296875},{x:.421875,y:.296875},{x:.421875,y:.296875},{x:.453125,y:.296875},{x:.453125,y:.296875},{x:.484375,y:.296875},{x:.484375,y:.296875},{x:.515625,y:.296875},{x:.515625,y:.296875},{x:.546875,y:.296875},{x:.546875,y:.296875},{x:.578125,y:.296875},{x:.578125,y:.296875},{x:.609375,y:.296875},{x:.609375,y:.296875},{x:.640625,y:.296875},{x:.640625,y:.296875},{x:.671875,y:.296875},{x:.671875,y:.296875},{x:.703125,y:.296875},{x:.703125,y:.296875},{x:.734375,y:.296875},{x:.734375,y:.296875},{x:.765625,y:.296875},{x:.765625,y:.296875},{x:.796875,y:.296875},{x:.796875,y:.296875},{x:.828125,y:.296875},{x:.828125,y:.296875},{x:.859375,y:.296875},{x:.859375,y:.296875},{x:.890625,y:.296875},{x:.890625,y:.296875},{x:.921875,y:.296875},{x:.921875,y:.296875},{x:.953125,y:.296875},{x:.953125,y:.296875},{x:.984375,y:.296875},{x:.984375,y:.296875},{x:.015625,y:.328125},{x:.015625,y:.328125},{x:.046875,y:.328125},{x:.046875,y:.328125},{x:.078125,y:.328125},{x:.078125,y:.328125},{x:.109375,y:.328125},{x:.109375,y:.328125},{x:.140625,y:.328125},{x:.140625,y:.328125},{x:.171875,y:.328125},{x:.171875,y:.328125},{x:.203125,y:.328125},{x:.203125,y:.328125},{x:.234375,y:.328125},{x:.234375,y:.328125},{x:.265625,y:.328125},{x:.265625,y:.328125},{x:.296875,y:.328125},{x:.296875,y:.328125},{x:.328125,y:.328125},{x:.328125,y:.328125},{x:.359375,y:.328125},{x:.359375,y:.328125},{x:.390625,y:.328125},{x:.390625,y:.328125},{x:.421875,y:.328125},{x:.421875,y:.328125},{x:.453125,y:.328125},{x:.453125,y:.328125},{x:.484375,y:.328125},{x:.484375,y:.328125},{x:.515625,y:.328125},{x:.515625,y:.328125},{x:.546875,y:.328125},{x:.546875,y:.328125},{x:.578125,y:.328125},{x:.578125,y:.328125},{x:.609375,y:.328125},{x:.609375,y:.328125},{x:.640625,y:.328125},{x:.640625,y:.328125},{x:.671875,y:.328125},{x:.671875,y:.328125},{x:.703125,y:.328125},{x:.703125,y:.328125},{x:.734375,y:.328125},{x:.734375,y:.328125},{x:.765625,y:.328125},{x:.765625,y:.328125},{x:.796875,y:.328125},{x:.796875,y:.328125},{x:.828125,y:.328125},{x:.828125,y:.328125},{x:.859375,y:.328125},{x:.859375,y:.328125},{x:.890625,y:.328125},{x:.890625,y:.328125},{x:.921875,y:.328125},{x:.921875,y:.328125},{x:.953125,y:.328125},{x:.953125,y:.328125},{x:.984375,y:.328125},{x:.984375,y:.328125},{x:.015625,y:.359375},{x:.015625,y:.359375},{x:.046875,y:.359375},{x:.046875,y:.359375},{x:.078125,y:.359375},{x:.078125,y:.359375},{x:.109375,y:.359375},{x:.109375,y:.359375},{x:.140625,y:.359375},{x:.140625,y:.359375},{x:.171875,y:.359375},{x:.171875,y:.359375},{x:.203125,y:.359375},{x:.203125,y:.359375},{x:.234375,y:.359375},{x:.234375,y:.359375},{x:.265625,y:.359375},{x:.265625,y:.359375},{x:.296875,y:.359375},{x:.296875,y:.359375},{x:.328125,y:.359375},{x:.328125,y:.359375},{x:.359375,y:.359375},{x:.359375,y:.359375},{x:.390625,y:.359375},{x:.390625,y:.359375},{x:.421875,y:.359375},{x:.421875,y:.359375},{x:.453125,y:.359375},{x:.453125,y:.359375},{x:.484375,y:.359375},{x:.484375,y:.359375},{x:.515625,y:.359375},{x:.515625,y:.359375},{x:.546875,y:.359375},{x:.546875,y:.359375},{x:.578125,y:.359375},{x:.578125,y:.359375},{x:.609375,y:.359375},{x:.609375,y:.359375},{x:.640625,y:.359375},{x:.640625,y:.359375},{x:.671875,y:.359375},{x:.671875,y:.359375},{x:.703125,y:.359375},{x:.703125,y:.359375},{x:.734375,y:.359375},{x:.734375,y:.359375},{x:.765625,y:.359375},{x:.765625,y:.359375},{x:.796875,y:.359375},{x:.796875,y:.359375},{x:.828125,y:.359375},{x:.828125,y:.359375},{x:.859375,y:.359375},{x:.859375,y:.359375},{x:.890625,y:.359375},{x:.890625,y:.359375},{x:.921875,y:.359375},{x:.921875,y:.359375},{x:.953125,y:.359375},{x:.953125,y:.359375},{x:.984375,y:.359375},{x:.984375,y:.359375},{x:.015625,y:.390625},{x:.015625,y:.390625},{x:.046875,y:.390625},{x:.046875,y:.390625},{x:.078125,y:.390625},{x:.078125,y:.390625},{x:.109375,y:.390625},{x:.109375,y:.390625},{x:.140625,y:.390625},{x:.140625,y:.390625},{x:.171875,y:.390625},{x:.171875,y:.390625},{x:.203125,y:.390625},{x:.203125,y:.390625},{x:.234375,y:.390625},{x:.234375,y:.390625},{x:.265625,y:.390625},{x:.265625,y:.390625},{x:.296875,y:.390625},{x:.296875,y:.390625},{x:.328125,y:.390625},{x:.328125,y:.390625},{x:.359375,y:.390625},{x:.359375,y:.390625},{x:.390625,y:.390625},{x:.390625,y:.390625},{x:.421875,y:.390625},{x:.421875,y:.390625},{x:.453125,y:.390625},{x:.453125,y:.390625},{x:.484375,y:.390625},{x:.484375,y:.390625},{x:.515625,y:.390625},{x:.515625,y:.390625},{x:.546875,y:.390625},{x:.546875,y:.390625},{x:.578125,y:.390625},{x:.578125,y:.390625},{x:.609375,y:.390625},{x:.609375,y:.390625},{x:.640625,y:.390625},{x:.640625,y:.390625},{x:.671875,y:.390625},{x:.671875,y:.390625},{x:.703125,y:.390625},{x:.703125,y:.390625},{x:.734375,y:.390625},{x:.734375,y:.390625},{x:.765625,y:.390625},{x:.765625,y:.390625},{x:.796875,y:.390625},{x:.796875,y:.390625},{x:.828125,y:.390625},{x:.828125,y:.390625},{x:.859375,y:.390625},{x:.859375,y:.390625},{x:.890625,y:.390625},{x:.890625,y:.390625},{x:.921875,y:.390625},{x:.921875,y:.390625},{x:.953125,y:.390625},{x:.953125,y:.390625},{x:.984375,y:.390625},{x:.984375,y:.390625},{x:.015625,y:.421875},{x:.015625,y:.421875},{x:.046875,y:.421875},{x:.046875,y:.421875},{x:.078125,y:.421875},{x:.078125,y:.421875},{x:.109375,y:.421875},{x:.109375,y:.421875},{x:.140625,y:.421875},{x:.140625,y:.421875},{x:.171875,y:.421875},{x:.171875,y:.421875},{x:.203125,y:.421875},{x:.203125,y:.421875},{x:.234375,y:.421875},{x:.234375,y:.421875},{x:.265625,y:.421875},{x:.265625,y:.421875},{x:.296875,y:.421875},{x:.296875,y:.421875},{x:.328125,y:.421875},{x:.328125,y:.421875},{x:.359375,y:.421875},{x:.359375,y:.421875},{x:.390625,y:.421875},{x:.390625,y:.421875},{x:.421875,y:.421875},{x:.421875,y:.421875},{x:.453125,y:.421875},{x:.453125,y:.421875},{x:.484375,y:.421875},{x:.484375,y:.421875},{x:.515625,y:.421875},{x:.515625,y:.421875},{x:.546875,y:.421875},{x:.546875,y:.421875},{x:.578125,y:.421875},{x:.578125,y:.421875},{x:.609375,y:.421875},{x:.609375,y:.421875},{x:.640625,y:.421875},{x:.640625,y:.421875},{x:.671875,y:.421875},{x:.671875,y:.421875},{x:.703125,y:.421875},{x:.703125,y:.421875},{x:.734375,y:.421875},{x:.734375,y:.421875},{x:.765625,y:.421875},{x:.765625,y:.421875},{x:.796875,y:.421875},{x:.796875,y:.421875},{x:.828125,y:.421875},{x:.828125,y:.421875},{x:.859375,y:.421875},{x:.859375,y:.421875},{x:.890625,y:.421875},{x:.890625,y:.421875},{x:.921875,y:.421875},{x:.921875,y:.421875},{x:.953125,y:.421875},{x:.953125,y:.421875},{x:.984375,y:.421875},{x:.984375,y:.421875},{x:.015625,y:.453125},{x:.015625,y:.453125},{x:.046875,y:.453125},{x:.046875,y:.453125},{x:.078125,y:.453125},{x:.078125,y:.453125},{x:.109375,y:.453125},{x:.109375,y:.453125},{x:.140625,y:.453125},{x:.140625,y:.453125},{x:.171875,y:.453125},{x:.171875,y:.453125},{x:.203125,y:.453125},{x:.203125,y:.453125},{x:.234375,y:.453125},{x:.234375,y:.453125},{x:.265625,y:.453125},{x:.265625,y:.453125},{x:.296875,y:.453125},{x:.296875,y:.453125},{x:.328125,y:.453125},{x:.328125,y:.453125},{x:.359375,y:.453125},{x:.359375,y:.453125},{x:.390625,y:.453125},{x:.390625,y:.453125},{x:.421875,y:.453125},{x:.421875,y:.453125},{x:.453125,y:.453125},{x:.453125,y:.453125},{x:.484375,y:.453125},{x:.484375,y:.453125},{x:.515625,y:.453125},{x:.515625,y:.453125},{x:.546875,y:.453125},{x:.546875,y:.453125},{x:.578125,y:.453125},{x:.578125,y:.453125},{x:.609375,y:.453125},{x:.609375,y:.453125},{x:.640625,y:.453125},{x:.640625,y:.453125},{x:.671875,y:.453125},{x:.671875,y:.453125},{x:.703125,y:.453125},{x:.703125,y:.453125},{x:.734375,y:.453125},{x:.734375,y:.453125},{x:.765625,y:.453125},{x:.765625,y:.453125},{x:.796875,y:.453125},{x:.796875,y:.453125},{x:.828125,y:.453125},{x:.828125,y:.453125},{x:.859375,y:.453125},{x:.859375,y:.453125},{x:.890625,y:.453125},{x:.890625,y:.453125},{x:.921875,y:.453125},{x:.921875,y:.453125},{x:.953125,y:.453125},{x:.953125,y:.453125},{x:.984375,y:.453125},{x:.984375,y:.453125},{x:.015625,y:.484375},{x:.015625,y:.484375},{x:.046875,y:.484375},{x:.046875,y:.484375},{x:.078125,y:.484375},{x:.078125,y:.484375},{x:.109375,y:.484375},{x:.109375,y:.484375},{x:.140625,y:.484375},{x:.140625,y:.484375},{x:.171875,y:.484375},{x:.171875,y:.484375},{x:.203125,y:.484375},{x:.203125,y:.484375},{x:.234375,y:.484375},{x:.234375,y:.484375},{x:.265625,y:.484375},{x:.265625,y:.484375},{x:.296875,y:.484375},{x:.296875,y:.484375},{x:.328125,y:.484375},{x:.328125,y:.484375},{x:.359375,y:.484375},{x:.359375,y:.484375},{x:.390625,y:.484375},{x:.390625,y:.484375},{x:.421875,y:.484375},{x:.421875,y:.484375},{x:.453125,y:.484375},{x:.453125,y:.484375},{x:.484375,y:.484375},{x:.484375,y:.484375},{x:.515625,y:.484375},{x:.515625,y:.484375},{x:.546875,y:.484375},{x:.546875,y:.484375},{x:.578125,y:.484375},{x:.578125,y:.484375},{x:.609375,y:.484375},{x:.609375,y:.484375},{x:.640625,y:.484375},{x:.640625,y:.484375},{x:.671875,y:.484375},{x:.671875,y:.484375},{x:.703125,y:.484375},{x:.703125,y:.484375},{x:.734375,y:.484375},{x:.734375,y:.484375},{x:.765625,y:.484375},{x:.765625,y:.484375},{x:.796875,y:.484375},{x:.796875,y:.484375},{x:.828125,y:.484375},{x:.828125,y:.484375},{x:.859375,y:.484375},{x:.859375,y:.484375},{x:.890625,y:.484375},{x:.890625,y:.484375},{x:.921875,y:.484375},{x:.921875,y:.484375},{x:.953125,y:.484375},{x:.953125,y:.484375},{x:.984375,y:.484375},{x:.984375,y:.484375},{x:.015625,y:.515625},{x:.015625,y:.515625},{x:.046875,y:.515625},{x:.046875,y:.515625},{x:.078125,y:.515625},{x:.078125,y:.515625},{x:.109375,y:.515625},{x:.109375,y:.515625},{x:.140625,y:.515625},{x:.140625,y:.515625},{x:.171875,y:.515625},{x:.171875,y:.515625},{x:.203125,y:.515625},{x:.203125,y:.515625},{x:.234375,y:.515625},{x:.234375,y:.515625},{x:.265625,y:.515625},{x:.265625,y:.515625},{x:.296875,y:.515625},{x:.296875,y:.515625},{x:.328125,y:.515625},{x:.328125,y:.515625},{x:.359375,y:.515625},{x:.359375,y:.515625},{x:.390625,y:.515625},{x:.390625,y:.515625},{x:.421875,y:.515625},{x:.421875,y:.515625},{x:.453125,y:.515625},{x:.453125,y:.515625},{x:.484375,y:.515625},{x:.484375,y:.515625},{x:.515625,y:.515625},{x:.515625,y:.515625},{x:.546875,y:.515625},{x:.546875,y:.515625},{x:.578125,y:.515625},{x:.578125,y:.515625},{x:.609375,y:.515625},{x:.609375,y:.515625},{x:.640625,y:.515625},{x:.640625,y:.515625},{x:.671875,y:.515625},{x:.671875,y:.515625},{x:.703125,y:.515625},{x:.703125,y:.515625},{x:.734375,y:.515625},{x:.734375,y:.515625},{x:.765625,y:.515625},{x:.765625,y:.515625},{x:.796875,y:.515625},{x:.796875,y:.515625},{x:.828125,y:.515625},{x:.828125,y:.515625},{x:.859375,y:.515625},{x:.859375,y:.515625},{x:.890625,y:.515625},{x:.890625,y:.515625},{x:.921875,y:.515625},{x:.921875,y:.515625},{x:.953125,y:.515625},{x:.953125,y:.515625},{x:.984375,y:.515625},{x:.984375,y:.515625},{x:.015625,y:.546875},{x:.015625,y:.546875},{x:.046875,y:.546875},{x:.046875,y:.546875},{x:.078125,y:.546875},{x:.078125,y:.546875},{x:.109375,y:.546875},{x:.109375,y:.546875},{x:.140625,y:.546875},{x:.140625,y:.546875},{x:.171875,y:.546875},{x:.171875,y:.546875},{x:.203125,y:.546875},{x:.203125,y:.546875},{x:.234375,y:.546875},{x:.234375,y:.546875},{x:.265625,y:.546875},{x:.265625,y:.546875},{x:.296875,y:.546875},{x:.296875,y:.546875},{x:.328125,y:.546875},{x:.328125,y:.546875},{x:.359375,y:.546875},{x:.359375,y:.546875},{x:.390625,y:.546875},{x:.390625,y:.546875},{x:.421875,y:.546875},{x:.421875,y:.546875},{x:.453125,y:.546875},{x:.453125,y:.546875},{x:.484375,y:.546875},{x:.484375,y:.546875},{x:.515625,y:.546875},{x:.515625,y:.546875},{x:.546875,y:.546875},{x:.546875,y:.546875},{x:.578125,y:.546875},{x:.578125,y:.546875},{x:.609375,y:.546875},{x:.609375,y:.546875},{x:.640625,y:.546875},{x:.640625,y:.546875},{x:.671875,y:.546875},{x:.671875,y:.546875},{x:.703125,y:.546875},{x:.703125,y:.546875},{x:.734375,y:.546875},{x:.734375,y:.546875},{x:.765625,y:.546875},{x:.765625,y:.546875},{x:.796875,y:.546875},{x:.796875,y:.546875},{x:.828125,y:.546875},{x:.828125,y:.546875},{x:.859375,y:.546875},{x:.859375,y:.546875},{x:.890625,y:.546875},{x:.890625,y:.546875},{x:.921875,y:.546875},{x:.921875,y:.546875},{x:.953125,y:.546875},{x:.953125,y:.546875},{x:.984375,y:.546875},{x:.984375,y:.546875},{x:.015625,y:.578125},{x:.015625,y:.578125},{x:.046875,y:.578125},{x:.046875,y:.578125},{x:.078125,y:.578125},{x:.078125,y:.578125},{x:.109375,y:.578125},{x:.109375,y:.578125},{x:.140625,y:.578125},{x:.140625,y:.578125},{x:.171875,y:.578125},{x:.171875,y:.578125},{x:.203125,y:.578125},{x:.203125,y:.578125},{x:.234375,y:.578125},{x:.234375,y:.578125},{x:.265625,y:.578125},{x:.265625,y:.578125},{x:.296875,y:.578125},{x:.296875,y:.578125},{x:.328125,y:.578125},{x:.328125,y:.578125},{x:.359375,y:.578125},{x:.359375,y:.578125},{x:.390625,y:.578125},{x:.390625,y:.578125},{x:.421875,y:.578125},{x:.421875,y:.578125},{x:.453125,y:.578125},{x:.453125,y:.578125},{x:.484375,y:.578125},{x:.484375,y:.578125},{x:.515625,y:.578125},{x:.515625,y:.578125},{x:.546875,y:.578125},{x:.546875,y:.578125},{x:.578125,y:.578125},{x:.578125,y:.578125},{x:.609375,y:.578125},{x:.609375,y:.578125},{x:.640625,y:.578125},{x:.640625,y:.578125},{x:.671875,y:.578125},{x:.671875,y:.578125},{x:.703125,y:.578125},{x:.703125,y:.578125},{x:.734375,y:.578125},{x:.734375,y:.578125},{x:.765625,y:.578125},{x:.765625,y:.578125},{x:.796875,y:.578125},{x:.796875,y:.578125},{x:.828125,y:.578125},{x:.828125,y:.578125},{x:.859375,y:.578125},{x:.859375,y:.578125},{x:.890625,y:.578125},{x:.890625,y:.578125},{x:.921875,y:.578125},{x:.921875,y:.578125},{x:.953125,y:.578125},{x:.953125,y:.578125},{x:.984375,y:.578125},{x:.984375,y:.578125},{x:.015625,y:.609375},{x:.015625,y:.609375},{x:.046875,y:.609375},{x:.046875,y:.609375},{x:.078125,y:.609375},{x:.078125,y:.609375},{x:.109375,y:.609375},{x:.109375,y:.609375},{x:.140625,y:.609375},{x:.140625,y:.609375},{x:.171875,y:.609375},{x:.171875,y:.609375},{x:.203125,y:.609375},{x:.203125,y:.609375},{x:.234375,y:.609375},{x:.234375,y:.609375},{x:.265625,y:.609375},{x:.265625,y:.609375},{x:.296875,y:.609375},{x:.296875,y:.609375},{x:.328125,y:.609375},{x:.328125,y:.609375},{x:.359375,y:.609375},{x:.359375,y:.609375},{x:.390625,y:.609375},{x:.390625,y:.609375},{x:.421875,y:.609375},{x:.421875,y:.609375},{x:.453125,y:.609375},{x:.453125,y:.609375},{x:.484375,y:.609375},{x:.484375,y:.609375},{x:.515625,y:.609375},{x:.515625,y:.609375},{x:.546875,y:.609375},{x:.546875,y:.609375},{x:.578125,y:.609375},{x:.578125,y:.609375},{x:.609375,y:.609375},{x:.609375,y:.609375},{x:.640625,y:.609375},{x:.640625,y:.609375},{x:.671875,y:.609375},{x:.671875,y:.609375},{x:.703125,y:.609375},{x:.703125,y:.609375},{x:.734375,y:.609375},{x:.734375,y:.609375},{x:.765625,y:.609375},{x:.765625,y:.609375},{x:.796875,y:.609375},{x:.796875,y:.609375},{x:.828125,y:.609375},{x:.828125,y:.609375},{x:.859375,y:.609375},{x:.859375,y:.609375},{x:.890625,y:.609375},{x:.890625,y:.609375},{x:.921875,y:.609375},{x:.921875,y:.609375},{x:.953125,y:.609375},{x:.953125,y:.609375},{x:.984375,y:.609375},{x:.984375,y:.609375},{x:.015625,y:.640625},{x:.015625,y:.640625},{x:.046875,y:.640625},{x:.046875,y:.640625},{x:.078125,y:.640625},{x:.078125,y:.640625},{x:.109375,y:.640625},{x:.109375,y:.640625},{x:.140625,y:.640625},{x:.140625,y:.640625},{x:.171875,y:.640625},{x:.171875,y:.640625},{x:.203125,y:.640625},{x:.203125,y:.640625},{x:.234375,y:.640625},{x:.234375,y:.640625},{x:.265625,y:.640625},{x:.265625,y:.640625},{x:.296875,y:.640625},{x:.296875,y:.640625},{x:.328125,y:.640625},{x:.328125,y:.640625},{x:.359375,y:.640625},{x:.359375,y:.640625},{x:.390625,y:.640625},{x:.390625,y:.640625},{x:.421875,y:.640625},{x:.421875,y:.640625},{x:.453125,y:.640625},{x:.453125,y:.640625},{x:.484375,y:.640625},{x:.484375,y:.640625},{x:.515625,y:.640625},{x:.515625,y:.640625},{x:.546875,y:.640625},{x:.546875,y:.640625},{x:.578125,y:.640625},{x:.578125,y:.640625},{x:.609375,y:.640625},{x:.609375,y:.640625},{x:.640625,y:.640625},{x:.640625,y:.640625},{x:.671875,y:.640625},{x:.671875,y:.640625},{x:.703125,y:.640625},{x:.703125,y:.640625},{x:.734375,y:.640625},{x:.734375,y:.640625},{x:.765625,y:.640625},{x:.765625,y:.640625},{x:.796875,y:.640625},{x:.796875,y:.640625},{x:.828125,y:.640625},{x:.828125,y:.640625},{x:.859375,y:.640625},{x:.859375,y:.640625},{x:.890625,y:.640625},{x:.890625,y:.640625},{x:.921875,y:.640625},{x:.921875,y:.640625},{x:.953125,y:.640625},{x:.953125,y:.640625},{x:.984375,y:.640625},{x:.984375,y:.640625},{x:.015625,y:.671875},{x:.015625,y:.671875},{x:.046875,y:.671875},{x:.046875,y:.671875},{x:.078125,y:.671875},{x:.078125,y:.671875},{x:.109375,y:.671875},{x:.109375,y:.671875},{x:.140625,y:.671875},{x:.140625,y:.671875},{x:.171875,y:.671875},{x:.171875,y:.671875},{x:.203125,y:.671875},{x:.203125,y:.671875},{x:.234375,y:.671875},{x:.234375,y:.671875},{x:.265625,y:.671875},{x:.265625,y:.671875},{x:.296875,y:.671875},{x:.296875,y:.671875},{x:.328125,y:.671875},{x:.328125,y:.671875},{x:.359375,y:.671875},{x:.359375,y:.671875},{x:.390625,y:.671875},{x:.390625,y:.671875},{x:.421875,y:.671875},{x:.421875,y:.671875},{x:.453125,y:.671875},{x:.453125,y:.671875},{x:.484375,y:.671875},{x:.484375,y:.671875},{x:.515625,y:.671875},{x:.515625,y:.671875},{x:.546875,y:.671875},{x:.546875,y:.671875},{x:.578125,y:.671875},{x:.578125,y:.671875},{x:.609375,y:.671875},{x:.609375,y:.671875},{x:.640625,y:.671875},{x:.640625,y:.671875},{x:.671875,y:.671875},{x:.671875,y:.671875},{x:.703125,y:.671875},{x:.703125,y:.671875},{x:.734375,y:.671875},{x:.734375,y:.671875},{x:.765625,y:.671875},{x:.765625,y:.671875},{x:.796875,y:.671875},{x:.796875,y:.671875},{x:.828125,y:.671875},{x:.828125,y:.671875},{x:.859375,y:.671875},{x:.859375,y:.671875},{x:.890625,y:.671875},{x:.890625,y:.671875},{x:.921875,y:.671875},{x:.921875,y:.671875},{x:.953125,y:.671875},{x:.953125,y:.671875},{x:.984375,y:.671875},{x:.984375,y:.671875},{x:.015625,y:.703125},{x:.015625,y:.703125},{x:.046875,y:.703125},{x:.046875,y:.703125},{x:.078125,y:.703125},{x:.078125,y:.703125},{x:.109375,y:.703125},{x:.109375,y:.703125},{x:.140625,y:.703125},{x:.140625,y:.703125},{x:.171875,y:.703125},{x:.171875,y:.703125},{x:.203125,y:.703125},{x:.203125,y:.703125},{x:.234375,y:.703125},{x:.234375,y:.703125},{x:.265625,y:.703125},{x:.265625,y:.703125},{x:.296875,y:.703125},{x:.296875,y:.703125},{x:.328125,y:.703125},{x:.328125,y:.703125},{x:.359375,y:.703125},{x:.359375,y:.703125},{x:.390625,y:.703125},{x:.390625,y:.703125},{x:.421875,y:.703125},{x:.421875,y:.703125},{x:.453125,y:.703125},{x:.453125,y:.703125},{x:.484375,y:.703125},{x:.484375,y:.703125},{x:.515625,y:.703125},{x:.515625,y:.703125},{x:.546875,y:.703125},{x:.546875,y:.703125},{x:.578125,y:.703125},{x:.578125,y:.703125},{x:.609375,y:.703125},{x:.609375,y:.703125},{x:.640625,y:.703125},{x:.640625,y:.703125},{x:.671875,y:.703125},{x:.671875,y:.703125},{x:.703125,y:.703125},{x:.703125,y:.703125},{x:.734375,y:.703125},{x:.734375,y:.703125},{x:.765625,y:.703125},{x:.765625,y:.703125},{x:.796875,y:.703125},{x:.796875,y:.703125},{x:.828125,y:.703125},{x:.828125,y:.703125},{x:.859375,y:.703125},{x:.859375,y:.703125},{x:.890625,y:.703125},{x:.890625,y:.703125},{x:.921875,y:.703125},{x:.921875,y:.703125},{x:.953125,y:.703125},{x:.953125,y:.703125},{x:.984375,y:.703125},{x:.984375,y:.703125},{x:.015625,y:.734375},{x:.015625,y:.734375},{x:.046875,y:.734375},{x:.046875,y:.734375},{x:.078125,y:.734375},{x:.078125,y:.734375},{x:.109375,y:.734375},{x:.109375,y:.734375},{x:.140625,y:.734375},{x:.140625,y:.734375},{x:.171875,y:.734375},{x:.171875,y:.734375},{x:.203125,y:.734375},{x:.203125,y:.734375},{x:.234375,y:.734375},{x:.234375,y:.734375},{x:.265625,y:.734375},{x:.265625,y:.734375},{x:.296875,y:.734375},{x:.296875,y:.734375},{x:.328125,y:.734375},{x:.328125,y:.734375},{x:.359375,y:.734375},{x:.359375,y:.734375},{x:.390625,y:.734375},{x:.390625,y:.734375},{x:.421875,y:.734375},{x:.421875,y:.734375},{x:.453125,y:.734375},{x:.453125,y:.734375},{x:.484375,y:.734375},{x:.484375,y:.734375},{x:.515625,y:.734375},{x:.515625,y:.734375},{x:.546875,y:.734375},{x:.546875,y:.734375},{x:.578125,y:.734375},{x:.578125,y:.734375},{x:.609375,y:.734375},{x:.609375,y:.734375},{x:.640625,y:.734375},{x:.640625,y:.734375},{x:.671875,y:.734375},{x:.671875,y:.734375},{x:.703125,y:.734375},{x:.703125,y:.734375},{x:.734375,y:.734375},{x:.734375,y:.734375},{x:.765625,y:.734375},{x:.765625,y:.734375},{x:.796875,y:.734375},{x:.796875,y:.734375},{x:.828125,y:.734375},{x:.828125,y:.734375},{x:.859375,y:.734375},{x:.859375,y:.734375},{x:.890625,y:.734375},{x:.890625,y:.734375},{x:.921875,y:.734375},{x:.921875,y:.734375},{x:.953125,y:.734375},{x:.953125,y:.734375},{x:.984375,y:.734375},{x:.984375,y:.734375},{x:.015625,y:.765625},{x:.015625,y:.765625},{x:.046875,y:.765625},{x:.046875,y:.765625},{x:.078125,y:.765625},{x:.078125,y:.765625},{x:.109375,y:.765625},{x:.109375,y:.765625},{x:.140625,y:.765625},{x:.140625,y:.765625},{x:.171875,y:.765625},{x:.171875,y:.765625},{x:.203125,y:.765625},{x:.203125,y:.765625},{x:.234375,y:.765625},{x:.234375,y:.765625},{x:.265625,y:.765625},{x:.265625,y:.765625},{x:.296875,y:.765625},{x:.296875,y:.765625},{x:.328125,y:.765625},{x:.328125,y:.765625},{x:.359375,y:.765625},{x:.359375,y:.765625},{x:.390625,y:.765625},{x:.390625,y:.765625},{x:.421875,y:.765625},{x:.421875,y:.765625},{x:.453125,y:.765625},{x:.453125,y:.765625},{x:.484375,y:.765625},{x:.484375,y:.765625},{x:.515625,y:.765625},{x:.515625,y:.765625},{x:.546875,y:.765625},{x:.546875,y:.765625},{x:.578125,y:.765625},{x:.578125,y:.765625},{x:.609375,y:.765625},{x:.609375,y:.765625},{x:.640625,y:.765625},{x:.640625,y:.765625},{x:.671875,y:.765625},{x:.671875,y:.765625},{x:.703125,y:.765625},{x:.703125,y:.765625},{x:.734375,y:.765625},{x:.734375,y:.765625},{x:.765625,y:.765625},{x:.765625,y:.765625},{x:.796875,y:.765625},{x:.796875,y:.765625},{x:.828125,y:.765625},{x:.828125,y:.765625},{x:.859375,y:.765625},{x:.859375,y:.765625},{x:.890625,y:.765625},{x:.890625,y:.765625},{x:.921875,y:.765625},{x:.921875,y:.765625},{x:.953125,y:.765625},{x:.953125,y:.765625},{x:.984375,y:.765625},{x:.984375,y:.765625},{x:.015625,y:.796875},{x:.015625,y:.796875},{x:.046875,y:.796875},{x:.046875,y:.796875},{x:.078125,y:.796875},{x:.078125,y:.796875},{x:.109375,y:.796875},{x:.109375,y:.796875},{x:.140625,y:.796875},{x:.140625,y:.796875},{x:.171875,y:.796875},{x:.171875,y:.796875},{x:.203125,y:.796875},{x:.203125,y:.796875},{x:.234375,y:.796875},{x:.234375,y:.796875},{x:.265625,y:.796875},{x:.265625,y:.796875},{x:.296875,y:.796875},{x:.296875,y:.796875},{x:.328125,y:.796875},{x:.328125,y:.796875},{x:.359375,y:.796875},{x:.359375,y:.796875},{x:.390625,y:.796875},{x:.390625,y:.796875},{x:.421875,y:.796875},{x:.421875,y:.796875},{x:.453125,y:.796875},{x:.453125,y:.796875},{x:.484375,y:.796875},{x:.484375,y:.796875},{x:.515625,y:.796875},{x:.515625,y:.796875},{x:.546875,y:.796875},{x:.546875,y:.796875},{x:.578125,y:.796875},{x:.578125,y:.796875},{x:.609375,y:.796875},{x:.609375,y:.796875},{x:.640625,y:.796875},{x:.640625,y:.796875},{x:.671875,y:.796875},{x:.671875,y:.796875},{x:.703125,y:.796875},{x:.703125,y:.796875},{x:.734375,y:.796875},{x:.734375,y:.796875},{x:.765625,y:.796875},{x:.765625,y:.796875},{x:.796875,y:.796875},{x:.796875,y:.796875},{x:.828125,y:.796875},{x:.828125,y:.796875},{x:.859375,y:.796875},{x:.859375,y:.796875},{x:.890625,y:.796875},{x:.890625,y:.796875},{x:.921875,y:.796875},{x:.921875,y:.796875},{x:.953125,y:.796875},{x:.953125,y:.796875},{x:.984375,y:.796875},{x:.984375,y:.796875},{x:.015625,y:.828125},{x:.015625,y:.828125},{x:.046875,y:.828125},{x:.046875,y:.828125},{x:.078125,y:.828125},{x:.078125,y:.828125},{x:.109375,y:.828125},{x:.109375,y:.828125},{x:.140625,y:.828125},{x:.140625,y:.828125},{x:.171875,y:.828125},{x:.171875,y:.828125},{x:.203125,y:.828125},{x:.203125,y:.828125},{x:.234375,y:.828125},{x:.234375,y:.828125},{x:.265625,y:.828125},{x:.265625,y:.828125},{x:.296875,y:.828125},{x:.296875,y:.828125},{x:.328125,y:.828125},{x:.328125,y:.828125},{x:.359375,y:.828125},{x:.359375,y:.828125},{x:.390625,y:.828125},{x:.390625,y:.828125},{x:.421875,y:.828125},{x:.421875,y:.828125},{x:.453125,y:.828125},{x:.453125,y:.828125},{x:.484375,y:.828125},{x:.484375,y:.828125},{x:.515625,y:.828125},{x:.515625,y:.828125},{x:.546875,y:.828125},{x:.546875,y:.828125},{x:.578125,y:.828125},{x:.578125,y:.828125},{x:.609375,y:.828125},{x:.609375,y:.828125},{x:.640625,y:.828125},{x:.640625,y:.828125},{x:.671875,y:.828125},{x:.671875,y:.828125},{x:.703125,y:.828125},{x:.703125,y:.828125},{x:.734375,y:.828125},{x:.734375,y:.828125},{x:.765625,y:.828125},{x:.765625,y:.828125},{x:.796875,y:.828125},{x:.796875,y:.828125},{x:.828125,y:.828125},{x:.828125,y:.828125},{x:.859375,y:.828125},{x:.859375,y:.828125},{x:.890625,y:.828125},{x:.890625,y:.828125},{x:.921875,y:.828125},{x:.921875,y:.828125},{x:.953125,y:.828125},{x:.953125,y:.828125},{x:.984375,y:.828125},{x:.984375,y:.828125},{x:.015625,y:.859375},{x:.015625,y:.859375},{x:.046875,y:.859375},{x:.046875,y:.859375},{x:.078125,y:.859375},{x:.078125,y:.859375},{x:.109375,y:.859375},{x:.109375,y:.859375},{x:.140625,y:.859375},{x:.140625,y:.859375},{x:.171875,y:.859375},{x:.171875,y:.859375},{x:.203125,y:.859375},{x:.203125,y:.859375},{x:.234375,y:.859375},{x:.234375,y:.859375},{x:.265625,y:.859375},{x:.265625,y:.859375},{x:.296875,y:.859375},{x:.296875,y:.859375},{x:.328125,y:.859375},{x:.328125,y:.859375},{x:.359375,y:.859375},{x:.359375,y:.859375},{x:.390625,y:.859375},{x:.390625,y:.859375},{x:.421875,y:.859375},{x:.421875,y:.859375},{x:.453125,y:.859375},{x:.453125,y:.859375},{x:.484375,y:.859375},{x:.484375,y:.859375},{x:.515625,y:.859375},{x:.515625,y:.859375},{x:.546875,y:.859375},{x:.546875,y:.859375},{x:.578125,y:.859375},{x:.578125,y:.859375},{x:.609375,y:.859375},{x:.609375,y:.859375},{x:.640625,y:.859375},{x:.640625,y:.859375},{x:.671875,y:.859375},{x:.671875,y:.859375},{x:.703125,y:.859375},{x:.703125,y:.859375},{x:.734375,y:.859375},{x:.734375,y:.859375},{x:.765625,y:.859375},{x:.765625,y:.859375},{x:.796875,y:.859375},{x:.796875,y:.859375},{x:.828125,y:.859375},{x:.828125,y:.859375},{x:.859375,y:.859375},{x:.859375,y:.859375},{x:.890625,y:.859375},{x:.890625,y:.859375},{x:.921875,y:.859375},{x:.921875,y:.859375},{x:.953125,y:.859375},{x:.953125,y:.859375},{x:.984375,y:.859375},{x:.984375,y:.859375},{x:.015625,y:.890625},{x:.015625,y:.890625},{x:.046875,y:.890625},{x:.046875,y:.890625},{x:.078125,y:.890625},{x:.078125,y:.890625},{x:.109375,y:.890625},{x:.109375,y:.890625},{x:.140625,y:.890625},{x:.140625,y:.890625},{x:.171875,y:.890625},{x:.171875,y:.890625},{x:.203125,y:.890625},{x:.203125,y:.890625},{x:.234375,y:.890625},{x:.234375,y:.890625},{x:.265625,y:.890625},{x:.265625,y:.890625},{x:.296875,y:.890625},{x:.296875,y:.890625},{x:.328125,y:.890625},{x:.328125,y:.890625},{x:.359375,y:.890625},{x:.359375,y:.890625},{x:.390625,y:.890625},{x:.390625,y:.890625},{x:.421875,y:.890625},{x:.421875,y:.890625},{x:.453125,y:.890625},{x:.453125,y:.890625},{x:.484375,y:.890625},{x:.484375,y:.890625},{x:.515625,y:.890625},{x:.515625,y:.890625},{x:.546875,y:.890625},{x:.546875,y:.890625},{x:.578125,y:.890625},{x:.578125,y:.890625},{x:.609375,y:.890625},{x:.609375,y:.890625},{x:.640625,y:.890625},{x:.640625,y:.890625},{x:.671875,y:.890625},{x:.671875,y:.890625},{x:.703125,y:.890625},{x:.703125,y:.890625},{x:.734375,y:.890625},{x:.734375,y:.890625},{x:.765625,y:.890625},{x:.765625,y:.890625},{x:.796875,y:.890625},{x:.796875,y:.890625},{x:.828125,y:.890625},{x:.828125,y:.890625},{x:.859375,y:.890625},{x:.859375,y:.890625},{x:.890625,y:.890625},{x:.890625,y:.890625},{x:.921875,y:.890625},{x:.921875,y:.890625},{x:.953125,y:.890625},{x:.953125,y:.890625},{x:.984375,y:.890625},{x:.984375,y:.890625},{x:.015625,y:.921875},{x:.015625,y:.921875},{x:.046875,y:.921875},{x:.046875,y:.921875},{x:.078125,y:.921875},{x:.078125,y:.921875},{x:.109375,y:.921875},{x:.109375,y:.921875},{x:.140625,y:.921875},{x:.140625,y:.921875},{x:.171875,y:.921875},{x:.171875,y:.921875},{x:.203125,y:.921875},{x:.203125,y:.921875},{x:.234375,y:.921875},{x:.234375,y:.921875},{x:.265625,y:.921875},{x:.265625,y:.921875},{x:.296875,y:.921875},{x:.296875,y:.921875},{x:.328125,y:.921875},{x:.328125,y:.921875},{x:.359375,y:.921875},{x:.359375,y:.921875},{x:.390625,y:.921875},{x:.390625,y:.921875},{x:.421875,y:.921875},{x:.421875,y:.921875},{x:.453125,y:.921875},{x:.453125,y:.921875},{x:.484375,y:.921875},{x:.484375,y:.921875},{x:.515625,y:.921875},{x:.515625,y:.921875},{x:.546875,y:.921875},{x:.546875,y:.921875},{x:.578125,y:.921875},{x:.578125,y:.921875},{x:.609375,y:.921875},{x:.609375,y:.921875},{x:.640625,y:.921875},{x:.640625,y:.921875},{x:.671875,y:.921875},{x:.671875,y:.921875},{x:.703125,y:.921875},{x:.703125,y:.921875},{x:.734375,y:.921875},{x:.734375,y:.921875},{x:.765625,y:.921875},{x:.765625,y:.921875},{x:.796875,y:.921875},{x:.796875,y:.921875},{x:.828125,y:.921875},{x:.828125,y:.921875},{x:.859375,y:.921875},{x:.859375,y:.921875},{x:.890625,y:.921875},{x:.890625,y:.921875},{x:.921875,y:.921875},{x:.921875,y:.921875},{x:.953125,y:.921875},{x:.953125,y:.921875},{x:.984375,y:.921875},{x:.984375,y:.921875},{x:.015625,y:.953125},{x:.015625,y:.953125},{x:.046875,y:.953125},{x:.046875,y:.953125},{x:.078125,y:.953125},{x:.078125,y:.953125},{x:.109375,y:.953125},{x:.109375,y:.953125},{x:.140625,y:.953125},{x:.140625,y:.953125},{x:.171875,y:.953125},{x:.171875,y:.953125},{x:.203125,y:.953125},{x:.203125,y:.953125},{x:.234375,y:.953125},{x:.234375,y:.953125},{x:.265625,y:.953125},{x:.265625,y:.953125},{x:.296875,y:.953125},{x:.296875,y:.953125},{x:.328125,y:.953125},{x:.328125,y:.953125},{x:.359375,y:.953125},{x:.359375,y:.953125},{x:.390625,y:.953125},{x:.390625,y:.953125},{x:.421875,y:.953125},{x:.421875,y:.953125},{x:.453125,y:.953125},{x:.453125,y:.953125},{x:.484375,y:.953125},{x:.484375,y:.953125},{x:.515625,y:.953125},{x:.515625,y:.953125},{x:.546875,y:.953125},{x:.546875,y:.953125},{x:.578125,y:.953125},{x:.578125,y:.953125},{x:.609375,y:.953125},{x:.609375,y:.953125},{x:.640625,y:.953125},{x:.640625,y:.953125},{x:.671875,y:.953125},{x:.671875,y:.953125},{x:.703125,y:.953125},{x:.703125,y:.953125},{x:.734375,y:.953125},{x:.734375,y:.953125},{x:.765625,y:.953125},{x:.765625,y:.953125},{x:.796875,y:.953125},{x:.796875,y:.953125},{x:.828125,y:.953125},{x:.828125,y:.953125},{x:.859375,y:.953125},{x:.859375,y:.953125},{x:.890625,y:.953125},{x:.890625,y:.953125},{x:.921875,y:.953125},{x:.921875,y:.953125},{x:.953125,y:.953125},{x:.953125,y:.953125},{x:.984375,y:.953125},{x:.984375,y:.953125},{x:.015625,y:.984375},{x:.015625,y:.984375},{x:.046875,y:.984375},{x:.046875,y:.984375},{x:.078125,y:.984375},{x:.078125,y:.984375},{x:.109375,y:.984375},{x:.109375,y:.984375},{x:.140625,y:.984375},{x:.140625,y:.984375},{x:.171875,y:.984375},{x:.171875,y:.984375},{x:.203125,y:.984375},{x:.203125,y:.984375},{x:.234375,y:.984375},{x:.234375,y:.984375},{x:.265625,y:.984375},{x:.265625,y:.984375},{x:.296875,y:.984375},{x:.296875,y:.984375},{x:.328125,y:.984375},{x:.328125,y:.984375},{x:.359375,y:.984375},{x:.359375,y:.984375},{x:.390625,y:.984375},{x:.390625,y:.984375},{x:.421875,y:.984375},{x:.421875,y:.984375},{x:.453125,y:.984375},{x:.453125,y:.984375},{x:.484375,y:.984375},{x:.484375,y:.984375},{x:.515625,y:.984375},{x:.515625,y:.984375},{x:.546875,y:.984375},{x:.546875,y:.984375},{x:.578125,y:.984375},{x:.578125,y:.984375},{x:.609375,y:.984375},{x:.609375,y:.984375},{x:.640625,y:.984375},{x:.640625,y:.984375},{x:.671875,y:.984375},{x:.671875,y:.984375},{x:.703125,y:.984375},{x:.703125,y:.984375},{x:.734375,y:.984375},{x:.734375,y:.984375},{x:.765625,y:.984375},{x:.765625,y:.984375},{x:.796875,y:.984375},{x:.796875,y:.984375},{x:.828125,y:.984375},{x:.828125,y:.984375},{x:.859375,y:.984375},{x:.859375,y:.984375},{x:.890625,y:.984375},{x:.890625,y:.984375},{x:.921875,y:.984375},{x:.921875,y:.984375},{x:.953125,y:.984375},{x:.953125,y:.984375},{x:.984375,y:.984375},{x:.984375,y:.984375},{x:.03125,y:.03125},{x:.03125,y:.03125},{x:.09375,y:.03125},{x:.09375,y:.03125},{x:.15625,y:.03125},{x:.15625,y:.03125},{x:.21875,y:.03125},{x:.21875,y:.03125},{x:.28125,y:.03125},{x:.28125,y:.03125},{x:.34375,y:.03125},{x:.34375,y:.03125},{x:.40625,y:.03125},{x:.40625,y:.03125},{x:.46875,y:.03125},{x:.46875,y:.03125},{x:.53125,y:.03125},{x:.53125,y:.03125},{x:.59375,y:.03125},{x:.59375,y:.03125},{x:.65625,y:.03125},{x:.65625,y:.03125},{x:.71875,y:.03125},{x:.71875,y:.03125},{x:.78125,y:.03125},{x:.78125,y:.03125},{x:.84375,y:.03125},{x:.84375,y:.03125},{x:.90625,y:.03125},{x:.90625,y:.03125},{x:.96875,y:.03125},{x:.96875,y:.03125},{x:.03125,y:.09375},{x:.03125,y:.09375},{x:.09375,y:.09375},{x:.09375,y:.09375},{x:.15625,y:.09375},{x:.15625,y:.09375},{x:.21875,y:.09375},{x:.21875,y:.09375},{x:.28125,y:.09375},{x:.28125,y:.09375},{x:.34375,y:.09375},{x:.34375,y:.09375},{x:.40625,y:.09375},{x:.40625,y:.09375},{x:.46875,y:.09375},{x:.46875,y:.09375},{x:.53125,y:.09375},{x:.53125,y:.09375},{x:.59375,y:.09375},{x:.59375,y:.09375},{x:.65625,y:.09375},{x:.65625,y:.09375},{x:.71875,y:.09375},{x:.71875,y:.09375},{x:.78125,y:.09375},{x:.78125,y:.09375},{x:.84375,y:.09375},{x:.84375,y:.09375},{x:.90625,y:.09375},{x:.90625,y:.09375},{x:.96875,y:.09375},{x:.96875,y:.09375},{x:.03125,y:.15625},{x:.03125,y:.15625},{x:.09375,y:.15625},{x:.09375,y:.15625},{x:.15625,y:.15625},{x:.15625,y:.15625},{x:.21875,y:.15625},{x:.21875,y:.15625},{x:.28125,y:.15625},{x:.28125,y:.15625},{x:.34375,y:.15625},{x:.34375,y:.15625},{x:.40625,y:.15625},{x:.40625,y:.15625},{x:.46875,y:.15625},{x:.46875,y:.15625},{x:.53125,y:.15625},{x:.53125,y:.15625},{x:.59375,y:.15625},{x:.59375,y:.15625},{x:.65625,y:.15625},{x:.65625,y:.15625},{x:.71875,y:.15625},{x:.71875,y:.15625},{x:.78125,y:.15625},{x:.78125,y:.15625},{x:.84375,y:.15625},{x:.84375,y:.15625},{x:.90625,y:.15625},{x:.90625,y:.15625},{x:.96875,y:.15625},{x:.96875,y:.15625},{x:.03125,y:.21875},{x:.03125,y:.21875},{x:.09375,y:.21875},{x:.09375,y:.21875},{x:.15625,y:.21875},{x:.15625,y:.21875},{x:.21875,y:.21875},{x:.21875,y:.21875},{x:.28125,y:.21875},{x:.28125,y:.21875},{x:.34375,y:.21875},{x:.34375,y:.21875},{x:.40625,y:.21875},{x:.40625,y:.21875},{x:.46875,y:.21875},{x:.46875,y:.21875},{x:.53125,y:.21875},{x:.53125,y:.21875},{x:.59375,y:.21875},{x:.59375,y:.21875},{x:.65625,y:.21875},{x:.65625,y:.21875},{x:.71875,y:.21875},{x:.71875,y:.21875},{x:.78125,y:.21875},{x:.78125,y:.21875},{x:.84375,y:.21875},{x:.84375,y:.21875},{x:.90625,y:.21875},{x:.90625,y:.21875},{x:.96875,y:.21875},{x:.96875,y:.21875},{x:.03125,y:.28125},{x:.03125,y:.28125},{x:.09375,y:.28125},{x:.09375,y:.28125},{x:.15625,y:.28125},{x:.15625,y:.28125},{x:.21875,y:.28125},{x:.21875,y:.28125},{x:.28125,y:.28125},{x:.28125,y:.28125},{x:.34375,y:.28125},{x:.34375,y:.28125},{x:.40625,y:.28125},{x:.40625,y:.28125},{x:.46875,y:.28125},{x:.46875,y:.28125},{x:.53125,y:.28125},{x:.53125,y:.28125},{x:.59375,y:.28125},{x:.59375,y:.28125},{x:.65625,y:.28125},{x:.65625,y:.28125},{x:.71875,y:.28125},{x:.71875,y:.28125},{x:.78125,y:.28125},{x:.78125,y:.28125},{x:.84375,y:.28125},{x:.84375,y:.28125},{x:.90625,y:.28125},{x:.90625,y:.28125},{x:.96875,y:.28125},{x:.96875,y:.28125},{x:.03125,y:.34375},{x:.03125,y:.34375},{x:.09375,y:.34375},{x:.09375,y:.34375},{x:.15625,y:.34375},{x:.15625,y:.34375},{x:.21875,y:.34375},{x:.21875,y:.34375},{x:.28125,y:.34375},{x:.28125,y:.34375},{x:.34375,y:.34375},{x:.34375,y:.34375},{x:.40625,y:.34375},{x:.40625,y:.34375},{x:.46875,y:.34375},{x:.46875,y:.34375},{x:.53125,y:.34375},{x:.53125,y:.34375},{x:.59375,y:.34375},{x:.59375,y:.34375},{x:.65625,y:.34375},{x:.65625,y:.34375},{x:.71875,y:.34375},{x:.71875,y:.34375},{x:.78125,y:.34375},{x:.78125,y:.34375},{x:.84375,y:.34375},{x:.84375,y:.34375},{x:.90625,y:.34375},{x:.90625,y:.34375},{x:.96875,y:.34375},{x:.96875,y:.34375},{x:.03125,y:.40625},{x:.03125,y:.40625},{x:.09375,y:.40625},{x:.09375,y:.40625},{x:.15625,y:.40625},{x:.15625,y:.40625},{x:.21875,y:.40625},{x:.21875,y:.40625},{x:.28125,y:.40625},{x:.28125,y:.40625},{x:.34375,y:.40625},{x:.34375,y:.40625},{x:.40625,y:.40625},{x:.40625,y:.40625},{x:.46875,y:.40625},{x:.46875,y:.40625},{x:.53125,y:.40625},{x:.53125,y:.40625},{x:.59375,y:.40625},{x:.59375,y:.40625},{x:.65625,y:.40625},{x:.65625,y:.40625},{x:.71875,y:.40625},{x:.71875,y:.40625},{x:.78125,y:.40625},{x:.78125,y:.40625},{x:.84375,y:.40625},{x:.84375,y:.40625},{x:.90625,y:.40625},{x:.90625,y:.40625},{x:.96875,y:.40625},{x:.96875,y:.40625},{x:.03125,y:.46875},{x:.03125,y:.46875},{x:.09375,y:.46875},{x:.09375,y:.46875},{x:.15625,y:.46875},{x:.15625,y:.46875},{x:.21875,y:.46875},{x:.21875,y:.46875},{x:.28125,y:.46875},{x:.28125,y:.46875},{x:.34375,y:.46875},{x:.34375,y:.46875},{x:.40625,y:.46875},{x:.40625,y:.46875},{x:.46875,y:.46875},{x:.46875,y:.46875},{x:.53125,y:.46875},{x:.53125,y:.46875},{x:.59375,y:.46875},{x:.59375,y:.46875},{x:.65625,y:.46875},{x:.65625,y:.46875},{x:.71875,y:.46875},{x:.71875,y:.46875},{x:.78125,y:.46875},{x:.78125,y:.46875},{x:.84375,y:.46875},{x:.84375,y:.46875},{x:.90625,y:.46875},{x:.90625,y:.46875},{x:.96875,y:.46875},{x:.96875,y:.46875},{x:.03125,y:.53125},{x:.03125,y:.53125},{x:.09375,y:.53125},{x:.09375,y:.53125},{x:.15625,y:.53125},{x:.15625,y:.53125},{x:.21875,y:.53125},{x:.21875,y:.53125},{x:.28125,y:.53125},{x:.28125,y:.53125},{x:.34375,y:.53125},{x:.34375,y:.53125},{x:.40625,y:.53125},{x:.40625,y:.53125},{x:.46875,y:.53125},{x:.46875,y:.53125},{x:.53125,y:.53125},{x:.53125,y:.53125},{x:.59375,y:.53125},{x:.59375,y:.53125},{x:.65625,y:.53125},{x:.65625,y:.53125},{x:.71875,y:.53125},{x:.71875,y:.53125},{x:.78125,y:.53125},{x:.78125,y:.53125},{x:.84375,y:.53125},{x:.84375,y:.53125},{x:.90625,y:.53125},{x:.90625,y:.53125},{x:.96875,y:.53125},{x:.96875,y:.53125},{x:.03125,y:.59375},{x:.03125,y:.59375},{x:.09375,y:.59375},{x:.09375,y:.59375},{x:.15625,y:.59375},{x:.15625,y:.59375},{x:.21875,y:.59375},{x:.21875,y:.59375},{x:.28125,y:.59375},{x:.28125,y:.59375},{x:.34375,y:.59375},{x:.34375,y:.59375},{x:.40625,y:.59375},{x:.40625,y:.59375},{x:.46875,y:.59375},{x:.46875,y:.59375},{x:.53125,y:.59375},{x:.53125,y:.59375},{x:.59375,y:.59375},{x:.59375,y:.59375},{x:.65625,y:.59375},{x:.65625,y:.59375},{x:.71875,y:.59375},{x:.71875,y:.59375},{x:.78125,y:.59375},{x:.78125,y:.59375},{x:.84375,y:.59375},{x:.84375,y:.59375},{x:.90625,y:.59375},{x:.90625,y:.59375},{x:.96875,y:.59375},{x:.96875,y:.59375},{x:.03125,y:.65625},{x:.03125,y:.65625},{x:.09375,y:.65625},{x:.09375,y:.65625},{x:.15625,y:.65625},{x:.15625,y:.65625},{x:.21875,y:.65625},{x:.21875,y:.65625},{x:.28125,y:.65625},{x:.28125,y:.65625},{x:.34375,y:.65625},{x:.34375,y:.65625},{x:.40625,y:.65625},{x:.40625,y:.65625},{x:.46875,y:.65625},{x:.46875,y:.65625},{x:.53125,y:.65625},{x:.53125,y:.65625},{x:.59375,y:.65625},{x:.59375,y:.65625},{x:.65625,y:.65625},{x:.65625,y:.65625},{x:.71875,y:.65625},{x:.71875,y:.65625},{x:.78125,y:.65625},{x:.78125,y:.65625},{x:.84375,y:.65625},{x:.84375,y:.65625},{x:.90625,y:.65625},{x:.90625,y:.65625},{x:.96875,y:.65625},{x:.96875,y:.65625},{x:.03125,y:.71875},{x:.03125,y:.71875},{x:.09375,y:.71875},{x:.09375,y:.71875},{x:.15625,y:.71875},{x:.15625,y:.71875},{x:.21875,y:.71875},{x:.21875,y:.71875},{x:.28125,y:.71875},{x:.28125,y:.71875},{x:.34375,y:.71875},{x:.34375,y:.71875},{x:.40625,y:.71875},{x:.40625,y:.71875},{x:.46875,y:.71875},{x:.46875,y:.71875},{x:.53125,y:.71875},{x:.53125,y:.71875},{x:.59375,y:.71875},{x:.59375,y:.71875},{x:.65625,y:.71875},{x:.65625,y:.71875},{x:.71875,y:.71875},{x:.71875,y:.71875},{x:.78125,y:.71875},{x:.78125,y:.71875},{x:.84375,y:.71875},{x:.84375,y:.71875},{x:.90625,y:.71875},{x:.90625,y:.71875},{x:.96875,y:.71875},{x:.96875,y:.71875},{x:.03125,y:.78125},{x:.03125,y:.78125},{x:.09375,y:.78125},{x:.09375,y:.78125},{x:.15625,y:.78125},{x:.15625,y:.78125},{x:.21875,y:.78125},{x:.21875,y:.78125},{x:.28125,y:.78125},{x:.28125,y:.78125},{x:.34375,y:.78125},{x:.34375,y:.78125},{x:.40625,y:.78125},{x:.40625,y:.78125},{x:.46875,y:.78125},{x:.46875,y:.78125},{x:.53125,y:.78125},{x:.53125,y:.78125},{x:.59375,y:.78125},{x:.59375,y:.78125},{x:.65625,y:.78125},{x:.65625,y:.78125},{x:.71875,y:.78125},{x:.71875,y:.78125},{x:.78125,y:.78125},{x:.78125,y:.78125},{x:.84375,y:.78125},{x:.84375,y:.78125},{x:.90625,y:.78125},{x:.90625,y:.78125},{x:.96875,y:.78125},{x:.96875,y:.78125},{x:.03125,y:.84375},{x:.03125,y:.84375},{x:.09375,y:.84375},{x:.09375,y:.84375},{x:.15625,y:.84375},{x:.15625,y:.84375},{x:.21875,y:.84375},{x:.21875,y:.84375},{x:.28125,y:.84375},{x:.28125,y:.84375},{x:.34375,y:.84375},{x:.34375,y:.84375},{x:.40625,y:.84375},{x:.40625,y:.84375},{x:.46875,y:.84375},{x:.46875,y:.84375},{x:.53125,y:.84375},{x:.53125,y:.84375},{x:.59375,y:.84375},{x:.59375,y:.84375},{x:.65625,y:.84375},{x:.65625,y:.84375},{x:.71875,y:.84375},{x:.71875,y:.84375},{x:.78125,y:.84375},{x:.78125,y:.84375},{x:.84375,y:.84375},{x:.84375,y:.84375},{x:.90625,y:.84375},{x:.90625,y:.84375},{x:.96875,y:.84375},{x:.96875,y:.84375},{x:.03125,y:.90625},{x:.03125,y:.90625},{x:.09375,y:.90625},{x:.09375,y:.90625},{x:.15625,y:.90625},{x:.15625,y:.90625},{x:.21875,y:.90625},{x:.21875,y:.90625},{x:.28125,y:.90625},{x:.28125,y:.90625},{x:.34375,y:.90625},{x:.34375,y:.90625},{x:.40625,y:.90625},{x:.40625,y:.90625},{x:.46875,y:.90625},{x:.46875,y:.90625},{x:.53125,y:.90625},{x:.53125,y:.90625},{x:.59375,y:.90625},{x:.59375,y:.90625},{x:.65625,y:.90625},{x:.65625,y:.90625},{x:.71875,y:.90625},{x:.71875,y:.90625},{x:.78125,y:.90625},{x:.78125,y:.90625},{x:.84375,y:.90625},{x:.84375,y:.90625},{x:.90625,y:.90625},{x:.90625,y:.90625},{x:.96875,y:.90625},{x:.96875,y:.90625},{x:.03125,y:.96875},{x:.03125,y:.96875},{x:.09375,y:.96875},{x:.09375,y:.96875},{x:.15625,y:.96875},{x:.15625,y:.96875},{x:.21875,y:.96875},{x:.21875,y:.96875},{x:.28125,y:.96875},{x:.28125,y:.96875},{x:.34375,y:.96875},{x:.34375,y:.96875},{x:.40625,y:.96875},{x:.40625,y:.96875},{x:.46875,y:.96875},{x:.46875,y:.96875},{x:.53125,y:.96875},{x:.53125,y:.96875},{x:.59375,y:.96875},{x:.59375,y:.96875},{x:.65625,y:.96875},{x:.65625,y:.96875},{x:.71875,y:.96875},{x:.71875,y:.96875},{x:.78125,y:.96875},{x:.78125,y:.96875},{x:.84375,y:.96875},{x:.84375,y:.96875},{x:.90625,y:.96875},{x:.90625,y:.96875},{x:.96875,y:.96875},{x:.96875,y:.96875},{x:.0625,y:.0625},{x:.0625,y:.0625},{x:.0625,y:.0625},{x:.0625,y:.0625},{x:.0625,y:.0625},{x:.0625,y:.0625},{x:.1875,y:.0625},{x:.1875,y:.0625},{x:.1875,y:.0625},{x:.1875,y:.0625},{x:.1875,y:.0625},{x:.1875,y:.0625},{x:.3125,y:.0625},{x:.3125,y:.0625},{x:.3125,y:.0625},{x:.3125,y:.0625},{x:.3125,y:.0625},{x:.3125,y:.0625},{x:.4375,y:.0625},{x:.4375,y:.0625},{x:.4375,y:.0625},{x:.4375,y:.0625},{x:.4375,y:.0625},{x:.4375,y:.0625},{x:.5625,y:.0625},{x:.5625,y:.0625},{x:.5625,y:.0625},{x:.5625,y:.0625},{x:.5625,y:.0625},{x:.5625,y:.0625},{x:.6875,y:.0625},{x:.6875,y:.0625},{x:.6875,y:.0625},{x:.6875,y:.0625},{x:.6875,y:.0625},{x:.6875,y:.0625},{x:.8125,y:.0625},{x:.8125,y:.0625},{x:.8125,y:.0625},{x:.8125,y:.0625},{x:.8125,y:.0625},{x:.8125,y:.0625},{x:.9375,y:.0625},{x:.9375,y:.0625},{x:.9375,y:.0625},{x:.9375,y:.0625},{x:.9375,y:.0625},{x:.9375,y:.0625},{x:.0625,y:.1875},{x:.0625,y:.1875},{x:.0625,y:.1875},{x:.0625,y:.1875},{x:.0625,y:.1875},{x:.0625,y:.1875},{x:.1875,y:.1875},{x:.1875,y:.1875},{x:.1875,y:.1875},{x:.1875,y:.1875},{x:.1875,y:.1875},{x:.1875,y:.1875},{x:.3125,y:.1875},{x:.3125,y:.1875},{x:.3125,y:.1875},{x:.3125,y:.1875},{x:.3125,y:.1875},{x:.3125,y:.1875},{x:.4375,y:.1875},{x:.4375,y:.1875},{x:.4375,y:.1875},{x:.4375,y:.1875},{x:.4375,y:.1875},{x:.4375,y:.1875},{x:.5625,y:.1875},{x:.5625,y:.1875},{x:.5625,y:.1875},{x:.5625,y:.1875},{x:.5625,y:.1875},{x:.5625,y:.1875},{x:.6875,y:.1875},{x:.6875,y:.1875},{x:.6875,y:.1875},{x:.6875,y:.1875},{x:.6875,y:.1875},{x:.6875,y:.1875},{x:.8125,y:.1875},{x:.8125,y:.1875},{x:.8125,y:.1875},{x:.8125,y:.1875},{x:.8125,y:.1875},{x:.8125,y:.1875},{x:.9375,y:.1875},{x:.9375,y:.1875},{x:.9375,y:.1875},{x:.9375,y:.1875},{x:.9375,y:.1875},{x:.9375,y:.1875},{x:.0625,y:.3125},{x:.0625,y:.3125},{x:.0625,y:.3125},{x:.0625,y:.3125},{x:.0625,y:.3125},{x:.0625,y:.3125},{x:.1875,y:.3125},{x:.1875,y:.3125},{x:.1875,y:.3125},{x:.1875,y:.3125},{x:.1875,y:.3125},{x:.1875,y:.3125},{x:.3125,y:.3125},{x:.3125,y:.3125},{x:.3125,y:.3125},{x:.3125,y:.3125},{x:.3125,y:.3125},{x:.3125,y:.3125},{x:.4375,y:.3125},{x:.4375,y:.3125},{x:.4375,y:.3125},{x:.4375,y:.3125},{x:.4375,y:.3125},{x:.4375,y:.3125},{x:.5625,y:.3125},{x:.5625,y:.3125},{x:.5625,y:.3125},{x:.5625,y:.3125},{x:.5625,y:.3125},{x:.5625,y:.3125},{x:.6875,y:.3125},{x:.6875,y:.3125},{x:.6875,y:.3125},{x:.6875,y:.3125},{x:.6875,y:.3125},{x:.6875,y:.3125},{x:.8125,y:.3125},{x:.8125,y:.3125},{x:.8125,y:.3125},{x:.8125,y:.3125},{x:.8125,y:.3125},{x:.8125,y:.3125},{x:.9375,y:.3125},{x:.9375,y:.3125},{x:.9375,y:.3125},{x:.9375,y:.3125},{x:.9375,y:.3125},{x:.9375,y:.3125},{x:.0625,y:.4375},{x:.0625,y:.4375},{x:.0625,y:.4375},{x:.0625,y:.4375},{x:.0625,y:.4375},{x:.0625,y:.4375},{x:.1875,y:.4375},{x:.1875,y:.4375},{x:.1875,y:.4375},{x:.1875,y:.4375},{x:.1875,y:.4375},{x:.1875,y:.4375},{x:.3125,y:.4375},{x:.3125,y:.4375},{x:.3125,y:.4375},{x:.3125,y:.4375},{x:.3125,y:.4375},{x:.3125,y:.4375},{x:.4375,y:.4375},{x:.4375,y:.4375},{x:.4375,y:.4375},{x:.4375,y:.4375},{x:.4375,y:.4375},{x:.4375,y:.4375},{x:.5625,y:.4375},{x:.5625,y:.4375},{x:.5625,y:.4375},{x:.5625,y:.4375},{x:.5625,y:.4375},{x:.5625,y:.4375},{x:.6875,y:.4375},{x:.6875,y:.4375},{x:.6875,y:.4375},{x:.6875,y:.4375},{x:.6875,y:.4375},{x:.6875,y:.4375},{x:.8125,y:.4375},{x:.8125,y:.4375},{x:.8125,y:.4375},{x:.8125,y:.4375},{x:.8125,y:.4375},{x:.8125,y:.4375},{x:.9375,y:.4375},{x:.9375,y:.4375},{x:.9375,y:.4375},{x:.9375,y:.4375},{x:.9375,y:.4375},{x:.9375,y:.4375},{x:.0625,y:.5625},{x:.0625,y:.5625},{x:.0625,y:.5625},{x:.0625,y:.5625},{x:.0625,y:.5625},{x:.0625,y:.5625},{x:.1875,y:.5625},{x:.1875,y:.5625},{x:.1875,y:.5625},{x:.1875,y:.5625},{x:.1875,y:.5625},{x:.1875,y:.5625},{x:.3125,y:.5625},{x:.3125,y:.5625},{x:.3125,y:.5625},{x:.3125,y:.5625},{x:.3125,y:.5625},{x:.3125,y:.5625},{x:.4375,y:.5625},{x:.4375,y:.5625},{x:.4375,y:.5625},{x:.4375,y:.5625},{x:.4375,y:.5625},{x:.4375,y:.5625},{x:.5625,y:.5625},{x:.5625,y:.5625},{x:.5625,y:.5625},{x:.5625,y:.5625},{x:.5625,y:.5625},{x:.5625,y:.5625},{x:.6875,y:.5625},{x:.6875,y:.5625},{x:.6875,y:.5625},{x:.6875,y:.5625},{x:.6875,y:.5625},{x:.6875,y:.5625},{x:.8125,y:.5625},{x:.8125,y:.5625},{x:.8125,y:.5625},{x:.8125,y:.5625},{x:.8125,y:.5625},{x:.8125,y:.5625},{x:.9375,y:.5625},{x:.9375,y:.5625},{x:.9375,y:.5625},{x:.9375,y:.5625},{x:.9375,y:.5625},{x:.9375,y:.5625},{x:.0625,y:.6875},{x:.0625,y:.6875},{x:.0625,y:.6875},{x:.0625,y:.6875},{x:.0625,y:.6875},{x:.0625,y:.6875},{x:.1875,y:.6875},{x:.1875,y:.6875},{x:.1875,y:.6875},{x:.1875,y:.6875},{x:.1875,y:.6875},{x:.1875,y:.6875},{x:.3125,y:.6875},{x:.3125,y:.6875},{x:.3125,y:.6875},{x:.3125,y:.6875},{x:.3125,y:.6875},{x:.3125,y:.6875},{x:.4375,y:.6875},{x:.4375,y:.6875},{x:.4375,y:.6875},{x:.4375,y:.6875},{x:.4375,y:.6875},{x:.4375,y:.6875},{x:.5625,y:.6875},{x:.5625,y:.6875},{x:.5625,y:.6875},{x:.5625,y:.6875},{x:.5625,y:.6875},{x:.5625,y:.6875},{x:.6875,y:.6875},{x:.6875,y:.6875},{x:.6875,y:.6875},{x:.6875,y:.6875},{x:.6875,y:.6875},{x:.6875,y:.6875},{x:.8125,y:.6875},{x:.8125,y:.6875},{x:.8125,y:.6875},{x:.8125,y:.6875},{x:.8125,y:.6875},{x:.8125,y:.6875},{x:.9375,y:.6875},{x:.9375,y:.6875},{x:.9375,y:.6875},{x:.9375,y:.6875},{x:.9375,y:.6875},{x:.9375,y:.6875},{x:.0625,y:.8125},{x:.0625,y:.8125},{x:.0625,y:.8125},{x:.0625,y:.8125},{x:.0625,y:.8125},{x:.0625,y:.8125},{x:.1875,y:.8125},{x:.1875,y:.8125},{x:.1875,y:.8125},{x:.1875,y:.8125},{x:.1875,y:.8125},{x:.1875,y:.8125},{x:.3125,y:.8125},{x:.3125,y:.8125},{x:.3125,y:.8125},{x:.3125,y:.8125},{x:.3125,y:.8125},{x:.3125,y:.8125},{x:.4375,y:.8125},{x:.4375,y:.8125},{x:.4375,y:.8125},{x:.4375,y:.8125},{x:.4375,y:.8125},{x:.4375,y:.8125},{x:.5625,y:.8125},{x:.5625,y:.8125},{x:.5625,y:.8125},{x:.5625,y:.8125},{x:.5625,y:.8125},{x:.5625,y:.8125},{x:.6875,y:.8125},{x:.6875,y:.8125},{x:.6875,y:.8125},{x:.6875,y:.8125},{x:.6875,y:.8125},{x:.6875,y:.8125},{x:.8125,y:.8125},{x:.8125,y:.8125},{x:.8125,y:.8125},{x:.8125,y:.8125},{x:.8125,y:.8125},{x:.8125,y:.8125},{x:.9375,y:.8125},{x:.9375,y:.8125},{x:.9375,y:.8125},{x:.9375,y:.8125},{x:.9375,y:.8125},{x:.9375,y:.8125},{x:.0625,y:.9375},{x:.0625,y:.9375},{x:.0625,y:.9375},{x:.0625,y:.9375},{x:.0625,y:.9375},{x:.0625,y:.9375},{x:.1875,y:.9375},{x:.1875,y:.9375},{x:.1875,y:.9375},{x:.1875,y:.9375},{x:.1875,y:.9375},{x:.1875,y:.9375},{x:.3125,y:.9375},{x:.3125,y:.9375},{x:.3125,y:.9375},{x:.3125,y:.9375},{x:.3125,y:.9375},{x:.3125,y:.9375},{x:.4375,y:.9375},{x:.4375,y:.9375},{x:.4375,y:.9375},{x:.4375,y:.9375},{x:.4375,y:.9375},{x:.4375,y:.9375},{x:.5625,y:.9375},{x:.5625,y:.9375},{x:.5625,y:.9375},{x:.5625,y:.9375},{x:.5625,y:.9375},{x:.5625,y:.9375},{x:.6875,y:.9375},{x:.6875,y:.9375},{x:.6875,y:.9375},{x:.6875,y:.9375},{x:.6875,y:.9375},{x:.6875,y:.9375},{x:.8125,y:.9375},{x:.8125,y:.9375},{x:.8125,y:.9375},{x:.8125,y:.9375},{x:.8125,y:.9375},{x:.8125,y:.9375},{x:.9375,y:.9375},{x:.9375,y:.9375},{x:.9375,y:.9375},{x:.9375,y:.9375},{x:.9375,y:.9375},{x:.9375,y:.9375}];var r1=class{constructor(t){fe(this,"model");fe(this,"anchors");fe(this,"anchorsTensor");fe(this,"inputSize");fe(this,"inputSizeTensor");fe(this,"doubleInputSizeTensor");this.model=t,this.anchors=KT.map(r=>[r.x,r.y]),this.anchorsTensor=ca(this.anchors),this.inputSize=this.model&&this.model.inputs&&this.model.inputs[0].shape?this.model.inputs[0].shape[2]:0,this.inputSizeTensor=Tt([this.inputSize,this.inputSize]),this.doubleInputSizeTensor=Tt([this.inputSize*2,this.inputSize*2])}normalizeBoxes(t){let r={};r.boxOffsets=Pe(t,[0,0],[-1,2]),r.boxSizes=Pe(t,[0,2],[-1,2]),r.div=pe(r.boxOffsets,this.inputSizeTensor),r.boxCenterPoints=le(r.div,this.anchorsTensor),r.halfBoxSizes=pe(r.boxSizes,this.doubleInputSizeTensor),r.sub=ce(r.boxCenterPoints,r.halfBoxSizes),r.startPoints=L(r.sub,this.inputSizeTensor),r.add=le(r.boxCenterPoints,r.halfBoxSizes),r.endPoints=L(r.add,this.inputSizeTensor);let n=hd([r.startPoints,r.endPoints],1);return Object.keys(r).forEach(a=>te(r[a])),n}normalizeLandmarks(t,r){let n={};n.reshape=U(t,[-1,7,2]),n.div=pe(n.reshape,this.inputSizeTensor),n.landmarks=le(n.div,this.anchors[r]);let a=L(n.landmarks,this.inputSizeTensor);return Object.keys(n).forEach(s=>te(n[s])),a}async predict(t,r){let n={};n.resize=Ie.resizeBilinear(t,[this.inputSize,this.inputSize]),n.div=pe(n.resize,Je.tf127),n.image=ce(n.div,Je.tf1),n.batched=this.model.execute(n.image),n.predictions=tt(n.batched),n.slice=Pe(n.predictions,[0,0],[-1,1]),n.sigmoid=Tr(n.slice),n.scores=tt(n.sigmoid);let a=await n.scores.data();n.boxes=Pe(n.predictions,[0,1],[-1,4]),n.norm=this.normalizeBoxes(n.boxes),n.nms=await Ie.nonMaxSuppressionAsync(n.norm,n.scores,3*r.hand.maxDetected,r.hand.iouThreshold,r.hand.minConfidence);let s=await n.nms.array(),i=[];for(let o of s){let l={};l.box=Pe(n.norm,[o,0],[1,-1]),l.slice=Pe(n.predictions,[o,5],[1,14]),l.norm=this.normalizeLandmarks(l.slice,o),l.palmLandmarks=U(l.norm,[-1,2]);let u=await l.box.data(),d=u.slice(0,2),h=u.slice(2,4),p=await l.palmLandmarks.array(),c={startPoint:d,endPoint:h,palmLandmarks:p,confidence:a[o]},f=jT(c,[t.shape[2]/this.inputSize,t.shape[1]/this.inputSize]);i.push(f),Object.keys(l).forEach(m=>te(l[m]))}return Object.keys(n).forEach(o=>te(n[o])),i}};var EAe=5,ZT=1.65,YT=[0,5,9,13,17,1,2],RAe=0,$Ae=2,JT=0,n1=class{constructor(t,r){fe(this,"handDetector");fe(this,"handPoseModel");fe(this,"inputSize");fe(this,"storedBoxes");fe(this,"skipped");fe(this,"detectedHands");this.handDetector=t,this.handPoseModel=r,this.inputSize=this.handPoseModel&&this.handPoseModel.inputs[0].shape?this.handPoseModel.inputs[0].shape[2]:0,this.storedBoxes=[],this.skipped=Number.MAX_SAFE_INTEGER,this.detectedHands=0}calculateLandmarksBoundingBox(t){let r=t.map(i=>i[0]),n=t.map(i=>i[1]),a=[Math.min(...r),Math.min(...n)],s=[Math.max(...r),Math.max(...n)];return{startPoint:a,endPoint:s}}getBoxForPalmLandmarks(t,r){let n=t.map(s=>Tb([...s,1],r)),a=this.calculateLandmarksBoundingBox(n);return e1(t1(a),EAe)}getBoxForHandLandmarks(t){let r=this.calculateLandmarksBoundingBox(t),n=e1(t1(r),ZT);n.palmLandmarks=[];for(let a=0;a<YT.length;a++)n.palmLandmarks.push(t[YT[a]].slice(0,2));return n}transformRawCoords(t,r,n,a){let s=Qm(r),i=[s[0]/this.inputSize,s[1]/this.inputSize,(s[0]+s[1])/this.inputSize/2],o=t.map(c=>[i[0]*(c[0]-this.inputSize/2),i[1]*(c[1]-this.inputSize/2),i[2]*c[2]]),l=Cb(n,[0,0]),u=o.map(c=>[...Tb(c,l),c[2]]),d=qT(a),h=[...tc(r),1],p=[ro(h,d[0]),ro(h,d[1])];return u.map(c=>[Math.trunc(c[0]+p[0]),Math.trunc(c[1]+p[1]),Math.trunc(c[2])])}async estimateHands(t,r){let n=!1,a,s=(r.hand.skipTime||0)>oe()-JT,i=this.skipped<(r.hand.skipFrames||0);r.skipAllowed&&s&&i&&(a=await this.handDetector.predict(t,r),this.skipped=0),r.skipAllowed&&this.skipped++,a&&a.length>0&&(a.length!==this.detectedHands&&this.detectedHands!==r.hand.maxDetected||!r.hand.landmarks)&&(this.detectedHands=0,this.storedBoxes=[...a],this.storedBoxes.length>0&&(n=!0));let o=[];for(let l=0;l<this.storedBoxes.length;l++){let u=this.storedBoxes[l];if(!!u)if(r.hand.landmarks){let d=r.hand.rotation?HT(u.palmLandmarks[RAe],u.palmLandmarks[$Ae]):0,h=tc(u),p=[h[0]/t.shape[2],h[1]/t.shape[1]],c=r.hand.rotation&&he.kernels.includes("rotatewithoffset")?Ie.rotateWithOffset(t,d,0,p):t.clone(),f=Cb(-d,h),m=n?this.getBoxForPalmLandmarks(u.palmLandmarks,f):u,g=GT(m,c,[this.inputSize,this.inputSize]),y=pe(g,Je.tf255);te(g),te(c);let[A,x]=this.handPoseModel.execute(y);JT=oe(),te(y);let b=(await A.data())[0];if(te(A),b>=r.hand.minConfidence/4){let w=U(x,[-1,3]),S=await w.array();te(x),te(w);let N=this.transformRawCoords(S,m,d,f),E=this.getBoxForHandLandmarks(N);this.storedBoxes[l]={...E,confidence:b};let R={landmarks:N,confidence:b,boxConfidence:u.confidence,fingerConfidence:b,box:{topLeft:E.startPoint,bottomRight:E.endPoint}};o.push(R)}else this.storedBoxes[l]=null;te(x)}else{let d=e1(t1(u),ZT),h={confidence:u.confidence,boxConfidence:u.confidence,fingerConfidence:0,box:{topLeft:d.startPoint,bottomRight:d.endPoint},landmarks:[]};o.push(h)}}return this.storedBoxes=this.storedBoxes.filter(l=>l!==null),this.detectedHands=o.length,o.length>r.hand.maxDetected&&(o.length=r.hand.maxDetected),o}};var Yr={thumb:0,index:1,middle:2,ring:3,pinky:4,all:[0,1,2,3,4],nameMapping:{0:"thumb",1:"index",2:"middle",3:"ring",4:"pinky"},pointsMapping:{0:[[0,1],[1,2],[2,3],[3,4]],1:[[0,5],[5,6],[6,7],[7,8]],2:[[0,9],[9,10],[10,11],[11,12]],3:[[0,13],[13,14],[14,15],[15,16]],4:[[0,17],[17,18],[18,19],[19,20]]},getName:e=>Yr.nameMapping[e],getPoints:e=>Yr.pointsMapping[e]},ao={none:0,half:1,full:2,nameMapping:{0:"none",1:"half",2:"full"},getName:e=>ao.nameMapping[e]},Bt={verticalUp:0,verticalDown:1,horizontalLeft:2,horizontalRight:3,diagonalUpRight:4,diagonalUpLeft:5,diagonalDownRight:6,diagonalDownLeft:7,nameMapping:{0:"verticalUp",1:"verticalDown",2:"horizontalLeft",3:"horizontalRight",4:"diagonalUpRight",5:"diagonalUpLeft",6:"diagonalDownRight",7:"diagonalDownLeft"},getName:e=>Bt.nameMapping[e]},no=class{constructor(t){fe(this,"name");fe(this,"curls");fe(this,"directions");fe(this,"weights");fe(this,"weightsRelative");this.name=t,this.curls={},this.directions={},this.weights=[1,1,1,1,1],this.weightsRelative=[1,1,1,1,1]}curl(t,r,n){typeof this.curls[t]=="undefined"&&(this.curls[t]=[]),this.curls[t].push([r,n])}direction(t,r,n){this.directions[t]||(this.directions[t]=[]),this.directions[t].push([r,n])}weight(t,r){this.weights[t]=r;let n=this.weights.reduce((a,s)=>a+s,0);this.weightsRelative=this.weights.map(a=>a*5/n)}matchAgainst(t,r){let n=0;for(let a in t){let s=t[a],i=this.curls[a];if(typeof i=="undefined"){n+=this.weightsRelative[a];continue}for(let[o,l]of i)if(s===o){n+=l*this.weightsRelative[a];break}}for(let a in r){let s=r[a],i=this.directions[a];if(typeof i=="undefined"){n+=this.weightsRelative[a];continue}for(let[o,l]of i)if(s===o){n+=l*this.weightsRelative[a];break}}return n/10}};var{thumb:wa,index:ps,middle:hs,ring:jl,pinky:Hl}=Yr,{none:ka,half:FAe,full:Ia}=ao,{verticalUp:Vd,verticalDown:M6e,horizontalLeft:Nb,horizontalRight:PAe,diagonalUpRight:_Ae,diagonalUpLeft:Ud,diagonalDownRight:F6e,diagonalDownLeft:P6e}=Bt,so=new no("thumbs up");so.curl(wa,ka,1);so.direction(wa,Vd,1);so.direction(wa,Ud,.25);so.direction(wa,_Ae,.25);for(let e of[Yr.index,Yr.middle,Yr.ring,Yr.pinky])so.curl(e,Ia,1),so.direction(e,Nb,1),so.direction(e,PAe,1);var tr=new no("victory");tr.curl(wa,FAe,.5);tr.curl(wa,ka,.5);tr.direction(wa,Vd,1);tr.direction(wa,Ud,1);tr.curl(ps,ka,1);tr.direction(ps,Vd,.75);tr.direction(ps,Ud,1);tr.curl(hs,ka,1);tr.direction(hs,Vd,1);tr.direction(hs,Ud,.75);tr.curl(jl,Ia,1);tr.direction(jl,Vd,.2);tr.direction(jl,Ud,1);tr.direction(jl,Nb,.2);tr.curl(Hl,Ia,1);tr.direction(Hl,Vd,.2);tr.direction(Hl,Ud,1);tr.direction(Hl,Nb,.2);tr.weight(ps,2);tr.weight(hs,2);var io=new no("point");io.curl(wa,Ia,1);io.curl(ps,ka,.5);io.curl(hs,Ia,.5);io.curl(jl,Ia,.5);io.curl(Hl,Ia,.5);io.weight(ps,2);io.weight(hs,2);var oo=new no("middle finger");oo.curl(wa,ka,1);oo.curl(ps,Ia,.5);oo.curl(hs,Ia,.5);oo.curl(jl,Ia,.5);oo.curl(Hl,Ia,.5);oo.weight(ps,2);oo.weight(hs,2);var Gd=new no("open palm");Gd.curl(wa,ka,.75);Gd.curl(ps,ka,.75);Gd.curl(hs,ka,.75);Gd.curl(jl,ka,.75);Gd.curl(Hl,ka,.75);var QT=[so,tr,io,oo,Gd];var zAe=.7,ql={HALF_CURL_START_LIMIT:60,NO_CURL_START_LIMIT:130,DISTANCE_VOTE_POWER:1.1,SINGLE_ANGLE_VOTE_POWER:.9,TOTAL_ANGLE_VOTE_POWER:1.6};function eN(e,t,r,n){let a=(t-n)/(e-r),s=Math.atan(a)*180/Math.PI;return s<=0?s=-s:s>0&&(s=180-s),s}function rN(e,t){if(!e||!t)return[0,0];let r=eN(e[0],e[1],t[0],t[1]);if(e.length===2)return r;let n=eN(e[1],e[2],t[1],t[2]);return[r,n]}function tN(e,t=1){let r=0,n=0,a=0;return e>=75&&e<=105?r=1*t:e>=25&&e<=155?n=1*t:a=1*t,[r,n,a]}function OAe(e,t,r){let n=e[0]-t[0],a=e[0]-r[0],s=t[0]-r[0],i=e[1]-t[1],o=e[1]-r[1],l=t[1]-r[1],u=e[2]-t[2],d=e[2]-r[2],h=t[2]-r[2],p=Math.sqrt(n*n+i*i+u*u),c=Math.sqrt(a*a+o*o+d*d),f=Math.sqrt(s*s+l*l+h*h),m=(f*f+p*p-c*c)/(2*f*p);m>1?m=1:m<-1&&(m=-1);let g=Math.acos(m);g=57.2958*g%180;let y;return g>ql.NO_CURL_START_LIMIT?y=ao.none:g>ql.HALF_CURL_START_LIMIT?y=ao.half:y=ao.full,y}function nN(e,t,r,n){let a;return n===Math.abs(e)?e>0?a=Bt.horizontalLeft:a=Bt.horizontalRight:n===Math.abs(t)?t>0?a=Bt.horizontalLeft:a=Bt.horizontalRight:r>0?a=Bt.horizontalLeft:a=Bt.horizontalRight,a}function aN(e,t,r,n){let a;return n===Math.abs(e)?e<0?a=Bt.verticalDown:a=Bt.verticalUp:n===Math.abs(t)?t<0?a=Bt.verticalDown:a=Bt.verticalUp:r<0?a=Bt.verticalDown:a=Bt.verticalUp,a}function DAe(e,t,r,n,a,s,i,o){let l,u=aN(e,t,r,n),d=nN(a,s,i,o);return u===Bt.verticalUp?d===Bt.horizontalLeft?l=Bt.diagonalUpLeft:l=Bt.diagonalUpRight:d===Bt.horizontalLeft?l=Bt.diagonalDownLeft:l=Bt.diagonalDownRight,l}function LAe(e,t,r,n){let a=e[0]-t[0],s=e[0]-r[0],i=t[0]-r[0],o=e[1]-t[1],l=e[1]-r[1],u=t[1]-r[1],d=Math.max(Math.abs(a),Math.abs(s),Math.abs(i)),h=Math.max(Math.abs(o),Math.abs(l),Math.abs(u)),p=0,c=0,f=0,m=h/(d+1e-5);m>1.5?p+=ql.DISTANCE_VOTE_POWER:m>.66?c+=ql.DISTANCE_VOTE_POWER:f+=ql.DISTANCE_VOTE_POWER;let g=Math.sqrt(a*a+o*o),y=Math.sqrt(s*s+l*l),A=Math.sqrt(i*i+u*u),x=Math.max(g,y,A),b=e[0],w=e[1],S=r[0],N=r[1];x===g?(S=r[0],N=r[1]):x===A&&(b=t[0],w=t[1]);let z=rN([b,w],[S,N]),M=tN(z,ql.TOTAL_ANGLE_VOTE_POWER);p+=M[0],c+=M[1],f+=M[2];for(let _ of n){let O=tN(_,ql.SINGLE_ANGLE_VOTE_POWER);p+=O[0],c+=O[1],f+=O[2]}let I;return p===Math.max(p,c,f)?I=aN(l,o,u,h):f===Math.max(c,f)?I=nN(s,a,i,d):I=DAe(l,o,u,h,s,a,i,d),I}function sN(e){let t=[],r=[],n=[],a=[];if(!e)return{curls:n,directions:a};for(let s of Yr.all){let i=Yr.getPoints(s),o=[],l=[];for(let u of i){let d=e[u[0]],h=e[u[1]],p=rN(d,h),c=p[0],f=p[1];o.push(c),l.push(f)}t.push(o),r.push(l)}for(let s of Yr.all){let i=s===Yr.thumb?1:0,o=Yr.getPoints(s),l=e[o[i][0]],u=e[o[i+1][1]],d=e[o[3][1]],h=OAe(l,u,d),p=LAe(l,u,d,t[s].slice(i));n[s]=h,a[s]=p}return{curls:n,directions:a}}function a1(e){if(!e||e.length===0)return null;let t=sN(e),r={};for(let n of Yr.all)r[Yr.getName(n)]={curl:ao.getName(t.curls[n]),direction:Bt.getName(t.directions[n])};return r}function iN(e){let t=[];if(!e||e.length===0)return t;let r=sN(e);for(let n of QT){let a=n.matchAgainst(r.curls,r.directions);a>=zAe&&t.push({name:n.name,confidence:a})}return t}var oN={thumb:[1,2,3,4],index:[5,6,7,8],middle:[9,10,11,12],ring:[13,14,15,16],pinky:[17,18,19,20],palm:[0]},jd,Hd,lN;async function Rb(e,t){let r=await lN.estimateHands(e,t);if(!r)return[];let n=[];for(let a=0;a<r.length;a++){let s={};if(r[a].landmarks)for(let d of Object.keys(oN))s[d]=oN[d].map(h=>r[a].landmarks[h]);let i=r[a].landmarks,o=[Number.MAX_SAFE_INTEGER,Number.MAX_SAFE_INTEGER,0,0],l=[0,0,0,0];if(i&&i.length>0){for(let d of i)d[0]<o[0]&&(o[0]=d[0]),d[1]<o[1]&&(o[1]=d[1]),d[0]>o[2]&&(o[2]=d[0]),d[1]>o[3]&&(o[3]=d[1]);o[2]-=o[0],o[3]-=o[1],l=[o[0]/(e.shape[2]||0),o[1]/(e.shape[1]||0),o[2]/(e.shape[2]||0),o[3]/(e.shape[1]||0)]}else o=r[a].box?[Math.trunc(Math.max(0,r[a].box.topLeft[0])),Math.trunc(Math.max(0,r[a].box.topLeft[1])),Math.trunc(Math.min(e.shape[2]||0,r[a].box.bottomRight[0])-Math.max(0,r[a].box.topLeft[0])),Math.trunc(Math.min(e.shape[1]||0,r[a].box.bottomRight[1])-Math.max(0,r[a].box.topLeft[1]))]:[0,0,0,0],l=[r[a].box.topLeft[0]/(e.shape[2]||0),r[a].box.topLeft[1]/(e.shape[1]||0),(r[a].box.bottomRight[0]-r[a].box.topLeft[0])/(e.shape[2]||0),(r[a].box.bottomRight[1]-r[a].box.topLeft[1])/(e.shape[1]||0)];let u=a1(i);n.push({id:a,score:Math.round(100*r[a].confidence)/100,boxScore:Math.round(100*r[a].boxConfidence)/100,fingerScore:Math.round(100*r[a].fingerConfidence)/100,label:"hand",box:o,boxRaw:l,keypoints:i,annotations:s,landmarks:u})}return n}async function $b(e){var r,n;he.initial&&(jd=null,Hd=null),!jd||!Hd?[jd,Hd]=await Promise.all([e.hand.enabled?Ge((r=e.hand.detector)==null?void 0:r.modelPath):null,e.hand.landmarks?Ge((n=e.hand.skeleton)==null?void 0:n.modelPath):null]):(e.debug&&se("cached model:",jd.modelUrl),e.debug&&se("cached model:",Hd.modelUrl));let t=new r1(jd);return lN=new n1(t,Hd),[jd,Hd]}var hr=[null,null],BAe=["StatefulPartitionedCall/Postprocessor/Slice","StatefulPartitionedCall/Postprocessor/ExpandDims_1"],lo=[[0,0],[0,0]],WAe=["hand","fist","pinch","point","face","tip","pinchtip"],dN=4,pN=1.6,VAe=512,UAe=1.4,s1=Number.MAX_SAFE_INTEGER,Mb=0,cs=[0,0],Ht={boxes:[],hands:[]},hN={thumb:[1,2,3,4],index:[5,6,7,8],middle:[9,10,11,12],ring:[13,14,15,16],pinky:[17,18,19,20],base:[0],palm:[0,17,13,9,5,1,0]};async function cN(e){var t;if(he.initial&&(hr[0]=null),hr[0])e.debug&&se("cached model:",hr[0].modelUrl);else{i1(["tensorlistreserve","enter","tensorlistfromtensor","merge","loopcond","switch","exit","tensorliststack","nextiteration","tensorlistsetitem","tensorlistgetitem","reciprocal","shape","split","where"],e),hr[0]=await Ge((t=e.hand.detector)==null?void 0:t.modelPath);let r=Object.values(hr[0].modelSignature.inputs);lo[0][0]=Array.isArray(r)?parseInt(r[0].tensorShape.dim[1].size):0,lo[0][1]=Array.isArray(r)?parseInt(r[0].tensorShape.dim[2].size):0}return hr[0]}async function fN(e){var t;if(he.initial&&(hr[1]=null),hr[1])e.debug&&se("cached model:",hr[1].modelUrl);else{hr[1]=await Ge((t=e.hand.skeleton)==null?void 0:t.modelPath);let r=Object.values(hr[1].modelSignature.inputs);lo[1][0]=Array.isArray(r)?parseInt(r[0].tensorShape.dim[1].size):0,lo[1][1]=Array.isArray(r)?parseInt(r[0].tensorShape.dim[2].size):0}return hr[1]}async function GAe(e,t){let r=[];if(!e||!hr[0])return r;let n={},a=(e.shape[2]||1)/(e.shape[1]||1),s=Math.min(Math.round((e.shape[1]||0)/8)*8,VAe),i=Math.round(s*a/8)*8;n.resize=Ie.resizeBilinear(e,[s,i]),n.cast=me(n.resize,"int32"),[n.rawScores,n.rawBoxes]=await hr[0].executeAsync(n.cast,BAe),n.boxes=tt(n.rawBoxes,[0,2]),n.scores=tt(n.rawScores,[0]);let o=nn(n.scores,1);te(o[dN]),o.splice(dN,1),n.filtered=ur(o,1),te(o),n.max=Ar(n.filtered,1),n.argmax=Rn(n.filtered,1);let l=0;n.nms=await Ie.nonMaxSuppressionAsync(n.boxes,n.max,(t.hand.maxDetected||0)+1,t.hand.iouThreshold||0,t.hand.minConfidence||1);let u=await n.nms.data(),d=await n.max.data(),h=await n.argmax.data();for(let p of Array.from(u)){let c=Pe(n.boxes,p,1),f=await c.data();te(c);let m=[f[1],f[0],f[3]-f[1],f[2]-f[0]],g=Hm(m,UAe),y=[Math.trunc(m[0]*cs[0]),Math.trunc(m[1]*cs[1]),Math.trunc(m[2]*cs[0]),Math.trunc(m[3]*cs[1])],A=d[p],x=WAe[h[p]],b={id:l++,score:A,box:y,boxRaw:g,label:x};r.push(b)}return Object.keys(n).forEach(p=>te(n[p])),r.sort((p,c)=>c.score-p.score),r.length>(t.hand.maxDetected||1)&&(r.length=t.hand.maxDetected||1),r}async function Fb(e,t,r){let n={id:t.id,score:Math.round(100*t.score)/100,boxScore:Math.round(100*t.score)/100,fingerScore:0,box:t.box,boxRaw:t.boxRaw,label:t.label,keypoints:[],landmarks:{},annotations:{}};if(e&&hr[1]&&r.hand.landmarks&&t.score>(r.hand.minConfidence||0)){let a={},s=[t.boxRaw[1],t.boxRaw[0],t.boxRaw[3]+t.boxRaw[1],t.boxRaw[2]+t.boxRaw[0]];a.crop=Ie.cropAndResize(e,[s],[0],[lo[1][0],lo[1][1]],"bilinear"),a.div=pe(a.crop,Je.tf255),[a.score,a.keypoints]=hr[1].execute(a.div,["Identity_1","Identity"]);let i=(await a.score.data())[0],o=(100-Math.trunc(100/(1+Math.exp(i))))/100;if(o>=(r.hand.minConfidence||0)){n.fingerScore=o,a.reshaped=U(a.keypoints,[-1,3]);let d=(await a.reshaped.array()).map(h=>[h[0]/lo[1][1],h[1]/lo[1][0],h[2]||0]).map(h=>[h[0]*t.boxRaw[2],h[1]*t.boxRaw[3],h[2]||0]);n.keypoints=d.map(h=>[cs[0]*(h[0]+t.boxRaw[0]),cs[1]*(h[1]+t.boxRaw[1]),h[2]||0]),n.landmarks=a1(n.keypoints);for(let h of Object.keys(hN))n.annotations[h]=hN[h].map(p=>n.landmarks&&n.keypoints[p]?n.keypoints[p]:null)}Object.keys(a).forEach(l=>te(a[l]))}return n}async function Pb(e,t){var a,s;if(!hr[0]||!hr[1]||!((a=hr[0])!=null&&a.inputs[0].shape)||!((s=hr[1])!=null&&s.inputs[0].shape))return[];cs=[e.shape[2]||0,e.shape[1]||0],s1++;let r=(t.hand.skipTime||0)>oe()-Mb,n=s1<(t.hand.skipFrames||0);return t.skipAllowed&&r&&n?Ht.hands:new Promise(async i=>{let o=3*(t.hand.skipTime||0)>oe()-Mb,l=s1<3*(t.hand.skipFrames||0);t.skipAllowed&&Ht.hands.length===t.hand.maxDetected?Ht.hands=await Promise.all(Ht.boxes.map(d=>Fb(e,d,t))):t.skipAllowed&&o&&l&&Ht.hands.length>0?Ht.hands=await Promise.all(Ht.boxes.map(d=>Fb(e,d,t))):(Ht.boxes=await GAe(e,t),Mb=oe(),Ht.hands=await Promise.all(Ht.boxes.map(d=>Fb(e,d,t))),s1=0);let u=[...Ht.boxes];if(Ht.boxes.length=0,t.cacheSensitivity>0)for(let d=0;d<Ht.hands.length;d++){let h=iT(Ht.hands[d].keypoints,cs);if(h.box[2]/(e.shape[2]||1)>.05&&h.box[3]/(e.shape[1]||1)>.05&&Ht.hands[d].fingerScore&&Ht.hands[d].fingerScore>(t.hand.minConfidence||0)){let p=Hm(h.box,pN),c=Hm(h.boxRaw,pN);Ht.boxes.push({...u[d],box:p,boxRaw:c})}}for(let d=0;d<Ht.hands.length;d++){let h=ls(Ht.hands[d].keypoints,cs);Ht.hands[d].box=h.box,Ht.hands[d].boxRaw=h.boxRaw}i(Ht.hands)})}var Mr,o1=[],_b=Number.MAX_SAFE_INTEGER,gN=0,yN=0;async function AN(e){var t;return he.initial&&(Mr=null),Mr?e.debug&&se("cached model:",Mr.modelUrl):Mr=await Ge((t=e.face.liveness)==null?void 0:t.modelPath),Mr}async function zb(e,t,r,n){var i,o;if(!Mr)return 0;let a=(((i=t.face.liveness)==null?void 0:i.skipTime)||0)>oe()-yN,s=_b<(((o=t.face.liveness)==null?void 0:o.skipFrames)||0);return t.skipAllowed&&a&&s&&gN===n&&o1[r]?(_b++,o1[r]):(_b=0,new Promise(async l=>{let u=Ie.resizeBilinear(e,[Mr!=null&&Mr.inputs[0].shape?Mr.inputs[0].shape[2]:0,Mr!=null&&Mr.inputs[0].shape?Mr.inputs[0].shape[1]:0],!1),d=Mr==null?void 0:Mr.execute(u),h=(await d.data())[0];o1[r]=Math.round(100*h)/100,gN=n,yN=oe(),te([u,d]),l(o1[r])}))}var rc={};vs(rc,{connected:()=>u1,horizontal:()=>Ob,kpt:()=>l1,relative:()=>Lb,vertical:()=>Db});var l1=["nose","leftEye","rightEye","leftEar","rightEar","leftShoulder","rightShoulder","leftElbow","rightElbow","leftWrist","rightWrist","leftHip","rightHip","leftKnee","rightKnee","leftAnkle","rightAnkle"],Ob=[["leftEye","rightEye"],["leftEar","rightEar"],["leftShoulder","rightShoulder"],["leftElbow","rightElbow"],["leftWrist","rightWrist"],["leftHip","rightHip"],["leftKnee","rightKnee"],["leftAnkle","rightAnkle"]],Db=[["leftKnee","leftShoulder"],["rightKnee","rightShoulder"],["leftAnkle","leftKnee"],["rightAnkle","rightKnee"]],Lb=[[["leftHip","rightHip"],["leftShoulder","rightShoulder"]],[["leftElbow","rightElbow"],["leftShoulder","rightShoulder"]]],u1={leftLeg:["leftHip","leftKnee","leftAnkle"],rightLeg:["rightHip","rightKnee","rightAnkle"],torso:["leftShoulder","rightShoulder","rightHip","leftHip","leftShoulder"],leftArm:["leftShoulder","leftElbow","leftWrist"],rightArm:["rightShoulder","rightElbow","rightWrist"],head:[]};var bN=.005,vn={keypoints:[],padding:[[0,0],[0,0],[0,0],[0,0]]};function Bb(e){for(let t of Ob){let r=e.keypoints.findIndex(a=>a.part===t[0]),n=e.keypoints.findIndex(a=>a.part===t[1]);if(e.keypoints[r]&&e.keypoints[n]&&e.keypoints[r].position[0]<e.keypoints[n].position[0]){let a=e.keypoints[r];e.keypoints[r]=e.keypoints[n],e.keypoints[n]=a}}for(let t of Db){let r=e.keypoints.findIndex(a=>a&&a.part===t[0]),n=e.keypoints.findIndex(a=>a&&a.part===t[1]);e.keypoints[r]&&e.keypoints[n]&&e.keypoints[r].position[1]<e.keypoints[n].position[1]&&e.keypoints.splice(r,1)}for(let[t,r]of Lb){let n=e.keypoints.findIndex(u=>u&&u.part===t[0]),a=e.keypoints.findIndex(u=>u&&u.part===t[1]),s=e.keypoints.findIndex(u=>u&&u.part===r[0]),i=e.keypoints.findIndex(u=>u&&u.part===r[1]);if(!e.keypoints[s]||!e.keypoints[i])continue;let o=e.keypoints[n]?[Math.abs(e.keypoints[s].position[0]-e.keypoints[n].position[0]),Math.abs(e.keypoints[i].position[0]-e.keypoints[n].position[0])]:[0,0],l=e.keypoints[a]?[Math.abs(e.keypoints[i].position[0]-e.keypoints[a].position[0]),Math.abs(e.keypoints[s].position[0]-e.keypoints[a].position[0])]:[0,0];if(o[0]>o[1]||l[0]>l[1]){let u=e.keypoints[n];e.keypoints[n]=e.keypoints[a],e.keypoints[a]=u}}}function vN(e){for(let t=0;t<e.length;t++)if(e[t]&&vn.keypoints[t]){let r=[Math.abs(e[t].positionRaw[0]-vn.keypoints[t].positionRaw[0]),Math.abs(e[t].positionRaw[1]-vn.keypoints[t].positionRaw[1])];r[0]<bN&&r[1]<bN?e[t]=vn.keypoints[t]:vn.keypoints[t]=e[t]}else vn.keypoints[t]=e[t];return e}function wN(e,t){let r={};if(!e.shape||!e.shape[1]||!e.shape[2])return e;vn.padding=[[0,0],[e.shape[2]>e.shape[1]?Math.trunc((e.shape[2]-e.shape[1])/2):0,e.shape[2]>e.shape[1]?Math.trunc((e.shape[2]-e.shape[1])/2):0],[e.shape[1]>e.shape[2]?Math.trunc((e.shape[1]-e.shape[2])/2):0,e.shape[1]>e.shape[2]?Math.trunc((e.shape[1]-e.shape[2])/2):0],[0,0]],r.pad=Xn(e,vn.padding),r.resize=Ie.resizeBilinear(r.pad,[t,t]);let n=me(r.resize,"int32");return Object.keys(r).forEach(a=>te(r[a])),n}function kN(e,t){e.keypoints=e.keypoints.filter(n=>n&&n.position);for(let n of e.keypoints)n.position=[n.position[0]*(t[0]+vn.padding[2][0]+vn.padding[2][1])/t[0]-vn.padding[2][0],n.position[1]*(t[1]+vn.padding[1][0]+vn.padding[1][1])/t[1]-vn.padding[1][0]],n.positionRaw=[n.position[0]/t[0],n.position[1]/t[1]];let r=ls(e.keypoints.map(n=>n.position),t);return e.box=r.box,e.boxRaw=r.boxRaw,e}var wn,d1=0,Wb=Number.MAX_SAFE_INTEGER,Xl={boxes:[],bodies:[],last:0};async function IN(e){return he.initial&&(wn=null),wn?e.debug&&se("cached model:",wn.modelUrl):(i1(["size"],e),wn=await Ge(e.body.modelPath)),d1=wn.inputs[0].shape?wn.inputs[0].shape[2]:0,d1<64&&(d1=256),wn}async function HAe(e,t,r){let n=e[0][0],a=[],s=0;for(let d=0;d<n.length;d++)if(s=n[d][2],s>t.body.minConfidence){let h=[n[d][1],n[d][0]];a.push({score:Math.round(100*s)/100,part:l1[d],positionRaw:h,position:[Math.round((r.shape[2]||0)*h[0]),Math.round((r.shape[1]||0)*h[1])]})}s=a.reduce((d,h)=>h.score>d?h.score:d,0);let i=[],o=ls(a.map(d=>d.position),[r.shape[2],r.shape[1]]),l={};for(let[d,h]of Object.entries(u1)){let p=[];for(let c=0;c<h.length-1;c++){let f=a.find(g=>g.part===h[c]),m=a.find(g=>g.part===h[c+1]);f&&m&&f.score>(t.body.minConfidence||0)&&m.score>(t.body.minConfidence||0)&&p.push([f.position,m.position])}l[d]=p}let u={id:0,score:s,box:o.box,boxRaw:o.boxRaw,keypoints:a,annotations:l};return Bb(u),i.push(u),i}async function qAe(e,t,r){let n=[];for(let a=0;a<e[0].length;a++){let s=e[0][a],i=Math.round(100*s[51+4])/100;if(i>t.body.minConfidence){let o=[];for(let h=0;h<17;h++){let p=s[3*h+2];if(p>t.body.minConfidence){let c=[s[3*h+1],s[3*h+0]];o.push({part:l1[h],score:Math.round(100*p)/100,positionRaw:c,position:[Math.round((r.shape[2]||0)*c[0]),Math.round((r.shape[1]||0)*c[1])]})}}let l=ls(o.map(h=>h.position),[r.shape[2],r.shape[1]]),u={};for(let[h,p]of Object.entries(u1)){let c=[];for(let f=0;f<p.length-1;f++){let m=o.find(y=>y.part===p[f]),g=o.find(y=>y.part===p[f+1]);m&&g&&m.score>(t.body.minConfidence||0)&&g.score>(t.body.minConfidence||0)&&c.push([m.position,g.position])}u[h]=c}let d={id:a,score:i,box:l.box,boxRaw:l.boxRaw,keypoints:[...o],annotations:u};Bb(d),n.push(d)}}return n.sort((a,s)=>s.score-a.score),n.length>t.body.maxDetected&&(n.length=t.body.maxDetected),n}async function Vb(e,t){if(!wn||!(wn!=null&&wn.inputs[0].shape))return[];t.skipAllowed||(Xl.boxes.length=0),Wb++;let r=(t.body.skipTime||0)>oe()-Xl.last,n=Wb<(t.body.skipFrames||0);return t.skipAllowed&&r&&n?Xl.bodies:new Promise(async a=>{let s={};Wb=0,s.input=wN(e,d1),s.res=wn==null?void 0:wn.execute(s.input),Xl.last=oe();let i=await s.res.array();Xl.bodies=s.res.shape[2]===17?await HAe(i,t,e):await qAe(i,t,e);for(let o of Xl.bodies)kN(o,[e.shape[2]||1,e.shape[1]||1]),vN(o.keypoints);Object.keys(s).forEach(o=>te(s[o])),a(Xl.bodies)})}var qd,p1=[],CN=0,Ub=Number.MAX_SAFE_INTEGER,c1=0,h1=2.5;async function TN(e){if(!qd||he.initial){qd=await Ge(e.object.modelPath);let t=Object.values(qd.modelSignature.inputs);c1=Array.isArray(t)?parseInt(t[0].tensorShape.dim[2].size):0}else e.debug&&se("cached model:",qd.modelUrl);return qd}async function XAe(e,t,r){let n=0,a=[];for(let l of[1,2,4])X(async()=>{let u=l*13,d=tt(e.find(m=>m.shape[1]===u**2&&(m.shape[2]||0)===Dd.length)),h=tt(e.find(m=>m.shape[1]===u**2&&(m.shape[2]||0)<Dd.length)),c=await h.reshape([-1,4,h.shape[1]/4]).argMax(2).array(),f=await d.array();for(let m=0;m<d.shape[0];m++)for(let g=0;g<d.shape[1];g++){let y=f[m][g];if(y>(r.object.minConfidence||0)&&g!==61){let A=(.5+Math.trunc(m%u))/u,x=(.5+Math.trunc(m/u))/u,b=c[m].map(I=>I*(u/l/c1)),[w,S]=[A-h1/l*b[0],x-h1/l*b[1]],[N,E]=[A+h1/l*b[2]-w,x+h1/l*b[3]-S],R=[w,S,N,E];R=R.map(I=>Math.max(0,Math.min(I,1)));let z=[R[0]*t[0],R[1]*t[1],R[2]*t[0],R[3]*t[1]],M={id:n++,score:Math.round(100*y)/100,class:g+1,label:Dd[g].label,box:z.map(I=>Math.trunc(I)),boxRaw:R};a.push(M)}}});e.forEach(l=>te(l));let s=a.map(l=>[l.boxRaw[1],l.boxRaw[0],l.boxRaw[3],l.boxRaw[2]]),i=a.map(l=>l.score),o=[];if(s&&s.length>0){let l=await Ie.nonMaxSuppressionAsync(s,i,r.object.maxDetected,r.object.iouThreshold,r.object.minConfidence);o=await l.data(),te(l)}return a=a.filter((l,u)=>o.includes(u)).sort((l,u)=>u.score-l.score),a}async function Gb(e,t){let r=(t.object.skipTime||0)>oe()-CN,n=Ub<(t.object.skipFrames||0);return t.skipAllowed&&r&&n&&p1.length>0?(Ub++,p1):(Ub=0,!he.kernels.includes("mod")||!he.kernels.includes("sparsetodense")?p1:new Promise(async a=>{let s=[e.shape[2]||0,e.shape[1]||0],i=Ie.resizeBilinear(e,[c1,c1],!1),o=pe(i,Je.tf255),l=o.transpose([0,3,1,2]);te(o),te(i);let u;t.object.enabled&&(u=qd.execute(l)),CN=oe(),te(l);let d=await XAe(u,s,t);p1=d,a(d)}))}var ac=["nose","leftEye","rightEye","leftEar","rightEar","leftShoulder","rightShoulder","leftElbow","rightElbow","leftWrist","rightWrist","leftHip","rightHip","leftKnee","rightKnee","leftAnkle","rightAnkle"],KAe=ac.length,nc=ac.reduce((e,t,r)=>(e[t]=r,e),{}),ZAe=[["leftHip","leftShoulder"],["leftElbow","leftShoulder"],["leftElbow","leftWrist"],["leftHip","leftKnee"],["leftKnee","leftAnkle"],["rightHip","rightShoulder"],["rightElbow","rightShoulder"],["rightElbow","rightWrist"],["rightHip","rightKnee"],["rightKnee","rightAnkle"],["leftShoulder","rightShoulder"],["leftHip","rightHip"]],ake=ZAe.map(([e,t])=>[nc[e],nc[t]]),EN=[["nose","leftEye"],["leftEye","leftEar"],["nose","rightEye"],["rightEye","rightEar"],["nose","leftShoulder"],["leftShoulder","leftElbow"],["leftElbow","leftWrist"],["leftShoulder","leftHip"],["leftHip","leftKnee"],["leftKnee","leftAnkle"],["nose","rightShoulder"],["rightShoulder","rightElbow"],["rightElbow","rightWrist"],["rightShoulder","rightHip"],["rightHip","rightKnee"],["rightKnee","rightAnkle"]];function RN(e){let t=e.reduce(({maxX:r,maxY:n,minX:a,minY:s},{position:{x:i,y:o}})=>({maxX:Math.max(r,i),maxY:Math.max(n,o),minX:Math.min(a,i),minY:Math.min(s,o)}),{maxX:Number.NEGATIVE_INFINITY,maxY:Number.NEGATIVE_INFINITY,minX:Number.POSITIVE_INFINITY,minY:Number.POSITIVE_INFINITY});return[t.minX,t.minY,t.maxX-t.minX,t.maxY-t.minY]}function $N(e,[t,r],[n,a]){let s=t/n,i=r/a,o=(u,d)=>({id:d,score:u.score,boxRaw:[u.box[0]/a,u.box[1]/n,u.box[2]/a,u.box[3]/n],box:[Math.trunc(u.box[0]*i),Math.trunc(u.box[1]*s),Math.trunc(u.box[2]*i),Math.trunc(u.box[3]*s)],keypoints:u.keypoints.map(({score:h,part:p,position:c})=>({score:h,part:p,position:[Math.trunc(c.x*i),Math.trunc(c.y*s)],positionRaw:[c.x/n,c.y/n]})),annotations:{}});return e.map((u,d)=>o(u,d))}var f1=class{constructor(t,r){fe(this,"priorityQueue");fe(this,"numberOfElements");fe(this,"getElementValue");this.priorityQueue=new Array(t),this.numberOfElements=-1,this.getElementValue=r}enqueue(t){this.priorityQueue[++this.numberOfElements]=t,this.swim(this.numberOfElements)}dequeue(){let t=this.priorityQueue[0];return this.exchange(0,this.numberOfElements--),this.sink(0),this.priorityQueue[this.numberOfElements+1]=null,t}empty(){return this.numberOfElements===-1}size(){return this.numberOfElements+1}all(){return this.priorityQueue.slice(0,this.numberOfElements+1)}max(){return this.priorityQueue[0]}swim(t){for(;t>0&&this.less(Math.floor(t/2),t);)this.exchange(t,Math.floor(t/2)),t=Math.floor(t/2)}sink(t){for(;2*t<=this.numberOfElements;){let r=2*t;if(r<this.numberOfElements&&this.less(r,r+1)&&r++,!this.less(t,r))break;this.exchange(t,r),t=r}}getValueAt(t){return this.getElementValue(this.priorityQueue[t])}less(t,r){return this.getValueAt(t)<this.getValueAt(r)}exchange(t,r){let n=this.priorityQueue[t];this.priorityQueue[t]=this.priorityQueue[r],this.priorityQueue[r]=n}};function jb(e,t,r,n){return{y:n.get(e,t,r),x:n.get(e,t,r+KAe)}}function Hb(e,t,r){let{heatmapY:n,heatmapX:a,id:s}=e,{y:i,x:o}=jb(n,a,s,r);return{x:e.heatmapX*t+o,y:e.heatmapY*t+i}}function qb(e,t,r){return e<t?t:e>r?r:e}function MN(e,t,r,n){let a=r-e,s=n-t;return a*a+s*s}function Xb(e,t){return{x:e.x+t.x,y:e.y+t.y}}var Sa,JAe=["MobilenetV1/offset_2/BiasAdd","MobilenetV1/heatmap_2/BiasAdd","MobilenetV1/displacement_fwd_2/BiasAdd","MobilenetV1/displacement_bwd_2/BiasAdd"],m1=1,Xd=16,QAe=50**2;function FN(e,t,r,n,a,s,i=2){let o=y=>({y:s.get(y.y,y.x,e),x:s.get(y.y,y.x,s.shape[2]/2+e)}),l=(y,A,x)=>({y:qb(Math.round(y.y/Xd),0,A-1),x:qb(Math.round(y.x/Xd),0,x-1)}),[u,d]=n.shape,h=l(t.position,u,d),p=o(h),f=Xb(t.position,p);for(let y=0;y<i;y++){let A=l(f,u,d),x=jb(A.y,A.x,r,a);f=Xb({x:A.x*Xd,y:A.y*Xd},{x:x.x,y:x.y})}let m=l(f,u,d),g=n.get(m.y,m.x,r);return{position:f,part:ac[r],score:g}}function exe(e,t,r,n,a){let s=EN.map(([p,c])=>[nc[p],nc[c]]),i=s.map(([,p])=>p),o=s.map(([p])=>p),l=t.shape[2],u=i.length,d=new Array(l),h=Hb(e.part,Xd,r);d[e.part.id]={score:e.score,part:ac[e.part.id],position:h};for(let p=u-1;p>=0;--p){let c=i[p],f=o[p];d[c]&&!d[f]&&(d[f]=FN(p,d[c],f,t,r,a))}for(let p=0;p<u;++p){let c=o[p],f=i[p];d[c]&&!d[f]&&(d[f]=FN(p,d[c],f,t,r,n))}return d}function txe(e,t,r,n,a){let[s,i]=a.shape,o=!0,l=Math.max(r-m1,0),u=Math.min(r+m1+1,s);for(let d=l;d<u;++d){let h=Math.max(n-m1,0),p=Math.min(n+m1+1,i);for(let c=h;c<p;++c)if(a.get(d,c,e)>t){o=!1;break}if(!o)break}return o}function rxe(e,t){let[r,n,a]=t.shape,s=new f1(r*n*a,({score:i})=>i);for(let i=0;i<r;++i)for(let o=0;o<n;++o)for(let l=0;l<a;++l){let u=t.get(i,o,l);u<e||txe(l,u,i,o,t)&&s.enqueue({score:u,part:{heatmapY:i,heatmapX:o,id:l}})}return s}function PN(e,{x:t,y:r},n){return e.some(({keypoints:a})=>{var i;let s=(i=a[n])==null?void 0:i.position;return s?MN(r,t,s.y,s.x)<=QAe:!1})}function nxe(e,t){return t.reduce((n,{position:a,score:s},i)=>(PN(e,a,i)||(n+=s),n),0)/t.length}function axe(e,t,r,n,a,s){let i=[],o=rxe(s,t);for(;i.length<a&&!o.empty();){let l=o.dequeue(),u=Hb(l.part,Xd,e);if(PN(i,u,l.part.id))continue;let d=exe(l,t,e,r,n);d=d.filter(c=>c.score>s);let h=nxe(i,d),p=RN(d);h>s&&i.push({keypoints:d,box:p,score:Math.round(100*h)/100})}return i}async function Kb(e,t){let r=X(()=>{if(!Sa.inputs[0].shape)return[];let i=Ie.resizeBilinear(e,[Sa.inputs[0].shape[2],Sa.inputs[0].shape[1]]),o=ce(pe(me(i,"float32"),127.5),1),u=Sa.execute(o,JAe).map(d=>tt(d,[0]));return u[1]=Tr(u[1]),u}),n=await Promise.all(r.map(i=>i.buffer()));for(let i of r)te(i);let a=await axe(n[0],n[1],n[2],n[3],t.body.maxDetected,t.body.minConfidence);return Sa.inputs[0].shape?$N(a,[e.shape[1],e.shape[2]],[Sa.inputs[0].shape[2],Sa.inputs[0].shape[1]]):[]}async function _N(e){return!Sa||he.initial?Sa=await Ge(e.body.modelPath):e.debug&&se("cached model:",Sa.modelUrl),Sa}var Wa,Zb=!1;async function Yb(e){return!Wa||he.initial?Wa=await Ge(e.segmentation.modelPath):e.debug&&se("cached model:",Wa.modelUrl),Wa}async function ON(e,t,r){var m,g;if(Zb)return{data:[],canvas:null,alpha:null};Zb=!0,Wa||await Yb(r);let n=await _d(e,r),a=((m=n.tensor)==null?void 0:m.shape[2])||0,s=((g=n.tensor)==null?void 0:g.shape[1])||0;if(!n.tensor)return{data:[],canvas:null,alpha:null};let i={};i.resize=Ie.resizeBilinear(n.tensor,[Wa.inputs[0].shape?Wa.inputs[0].shape[1]:0,Wa.inputs[0].shape?Wa.inputs[0].shape[2]:0],!1),te(n.tensor),i.norm=pe(i.resize,Je.tf255),i.res=Wa.execute(i.norm),i.squeeze=tt(i.res,0),i.squeeze.shape[2]===2?(i.softmax=yd(i.squeeze),[i.bg,i.fg]=nn(i.softmax,2),i.expand=Kt(i.fg,2),i.pad=Kt(i.expand,0),i.crop=Ie.cropAndResize(i.pad,[[0,0,.5,.5]],[0],[a,s]),i.data=tt(i.crop,0)):i.data=Ie.resizeBilinear(i.squeeze,[s,a]);let o=Array.from(await i.data.data());if(he.node&&!he.Canvas&&typeof ImageData=="undefined")return r.debug&&se("canvas support missing"),Object.keys(i).forEach(y=>te(i[y])),{data:o,canvas:null,alpha:null};let l=Kr(a,s);On&&await On.toPixels(i.data,l);let u=l.getContext("2d");r.segmentation.blur&&r.segmentation.blur>0&&(u.filter=`blur(${r.segmentation.blur}px)`);let d=u.getImageData(0,0,a,s),h=Kr(a,s),p=h.getContext("2d");n.canvas&&p.drawImage(n.canvas,0,0),p.globalCompositeOperation="darken",r.segmentation.blur&&r.segmentation.blur>0&&(p.filter=`blur(${r.segmentation.blur}px)`),p.drawImage(l,0,0),p.globalCompositeOperation="source-over",p.filter="none";let c=p.getImageData(0,0,a,s);for(let y=0;y<a*s;y++)c.data[4*y+3]=d.data[4*y+0];p.putImageData(c,0,0);let f=null;if(t&&h){f=Kr(a,s);let y=await _d(t,r);te(y.tensor);let A=f.getContext("2d");A.drawImage(y.canvas,0,0,f.width,f.height),A.drawImage(h,0,0)}return Object.keys(i).forEach(y=>te(i[y])),Zb=!1,{data:o,canvas:h,alpha:l}}var sc=class{constructor(){fe(this,"ssrnetage",null);fe(this,"gear",null);fe(this,"blazeposedetect",null);fe(this,"blazepose",null);fe(this,"centernet",null);fe(this,"efficientpose",null);fe(this,"mobilefacenet",null);fe(this,"emotion",null);fe(this,"facedetect",null);fe(this,"faceiris",null);fe(this,"facemesh",null);fe(this,"faceres",null);fe(this,"ssrnetgender",null);fe(this,"handpose",null);fe(this,"handskeleton",null);fe(this,"handtrack",null);fe(this,"liveness",null);fe(this,"movenet",null);fe(this,"nanodet",null);fe(this,"posenet",null);fe(this,"segmentation",null);fe(this,"antispoof",null)}};function g1(e){for(let t of Object.keys(e.models))e.models[t]=null}async function Jb(e){var t,r,n,a,s,i,o,l,u,d,h,p,c,f,m,g,y,A,x,b,w,S,N,E,R,z,M,I,_,O,q;he.initial&&g1(e),e.config.hand.enabled&&(!e.models.handpose&&((r=(t=e.config.hand.detector)==null?void 0:t.modelPath)==null?void 0:r.includes("handdetect"))&&([e.models.handpose,e.models.handskeleton]=await $b(e.config)),!e.models.handskeleton&&e.config.hand.landmarks&&((a=(n=e.config.hand.detector)==null?void 0:n.modelPath)==null?void 0:a.includes("handdetect"))&&([e.models.handpose,e.models.handskeleton]=await $b(e.config))),e.config.body.enabled&&!e.models.blazepose&&((i=(s=e.config.body)==null?void 0:s.modelPath)==null?void 0:i.includes("blazepose"))&&(e.models.blazepose=pT(e.config)),e.config.body.enabled&&!e.models.blazeposedetect&&e.config.body.detector&&e.config.body.detector.modelPath&&(e.models.blazeposedetect=dT(e.config)),e.config.body.enabled&&!e.models.efficientpose&&((l=(o=e.config.body)==null?void 0:o.modelPath)==null?void 0:l.includes("efficientpose"))&&(e.models.efficientpose=yT(e.config)),e.config.body.enabled&&!e.models.movenet&&((d=(u=e.config.body)==null?void 0:u.modelPath)==null?void 0:d.includes("movenet"))&&(e.models.movenet=IN(e.config)),e.config.body.enabled&&!e.models.posenet&&((p=(h=e.config.body)==null?void 0:h.modelPath)==null?void 0:p.includes("posenet"))&&(e.models.posenet=_N(e.config)),e.config.face.enabled&&!e.models.facedetect&&(e.models.facedetect=tT(e.config)),e.config.face.enabled&&((c=e.config.face.antispoof)==null?void 0:c.enabled)&&!e.models.antispoof&&(e.models.antispoof=VC(e.config)),e.config.face.enabled&&((f=e.config.face.liveness)==null?void 0:f.enabled)&&!e.models.liveness&&(e.models.liveness=AN(e.config)),e.config.face.enabled&&((m=e.config.face.description)==null?void 0:m.enabled)&&!e.models.faceres&&(e.models.faceres=WT(e.config)),e.config.face.enabled&&((g=e.config.face.emotion)==null?void 0:g.enabled)&&!e.models.emotion&&(e.models.emotion=vT(e.config)),e.config.face.enabled&&((y=e.config.face.iris)==null?void 0:y.enabled)&&!((A=e.config.face.attention)!=null&&A.enabled)&&!e.models.faceiris&&(e.models.faceiris=$T(e.config)),e.config.face.enabled&&((x=e.config.face.mesh)==null?void 0:x.enabled)&&!e.models.facemesh&&(e.models.facemesh=zT(e.config)),e.config.face.enabled&&((b=e.config.face.gear)==null?void 0:b.enabled)&&!e.models.gear&&(e.models.gear=EC(e.config)),e.config.face.enabled&&((w=e.config.face.ssrnet)==null?void 0:w.enabled)&&!e.models.ssrnetage&&(e.models.ssrnetage=PC(e.config)),e.config.face.enabled&&((S=e.config.face.ssrnet)==null?void 0:S.enabled)&&!e.models.ssrnetgender&&(e.models.ssrnetgender=DC(e.config)),e.config.face.enabled&&((N=e.config.face.mobilefacenet)==null?void 0:N.enabled)&&!e.models.mobilefacenet&&(e.models.mobilefacenet=CT(e.config)),e.config.hand.enabled&&!e.models.handtrack&&((R=(E=e.config.hand.detector)==null?void 0:E.modelPath)==null?void 0:R.includes("handtrack"))&&(e.models.handtrack=cN(e.config)),e.config.hand.enabled&&e.config.hand.landmarks&&!e.models.handskeleton&&((M=(z=e.config.hand.detector)==null?void 0:z.modelPath)==null?void 0:M.includes("handtrack"))&&(e.models.handskeleton=fN(e.config)),e.config.object.enabled&&!e.models.centernet&&((_=(I=e.config.object)==null?void 0:I.modelPath)==null?void 0:_.includes("centernet"))&&(e.models.centernet=fT(e.config)),e.config.object.enabled&&!e.models.nanodet&&((q=(O=e.config.object)==null?void 0:O.modelPath)==null?void 0:q.includes("nanodet"))&&(e.models.nanodet=TN(e.config)),e.config.segmentation.enabled&&!e.models.segmentation&&(e.models.segmentation=Yb(e.config));for await(let K of Object.keys(e.models))e.models[K]&&typeof e.models[K]!="undefined"&&(e.models[K]=await e.models[K])}async function Qb(e){let t=["const","placeholder","noop","pad","squeeze","add","sub","mul","div"];for(let r of Object.keys(e.models)){let n=e.models[r];if(!n)continue;let a=[],s=n==null?void 0:n.executor;if(s&&s.graph.nodes)for(let o of Object.values(s.graph.nodes)){let l=o.op.toLowerCase();a.includes(l)||a.push(l)}else!s&&e.config.debug&&se("model signature not determined:",r);let i=[];for(let o of a)!t.includes(o)&&!e.env.kernels.includes(o)&&!e.env.kernels.includes(o.replace("_",""))&&!e.env.kernels.includes(o.replace("native",""))&&!e.env.kernels.includes(o.replace("v2",""))&&i.push(o);e.config.debug&&i.length>0&&se("model validation failed:",r,i)}}var Et={name:"humangl",priority:999,canvas:null,gl:null,extensions:[],webGLattr:{alpha:!1,antialias:!1,premultipliedAlpha:!1,preserveDrawingBuffer:!1,depth:!1,stencil:!1,failIfMajorPerformanceCaveat:!1,desynchronized:!0}};function sxe(){let e=Et.gl;!e||(Et.extensions=e.getSupportedExtensions())}async function LN(e){var t;if(e.config.backend==="humangl"&&(Et.name in Xt().registry&&(!Et.gl||!Et.gl.getParameter(Et.gl.VERSION))&&(se("error: humangl backend invalid context"),g1(e)),!Oy(Et.name))){try{Et.canvas=await Kr(100,100)}catch(n){se("error: cannot create canvas:",n);return}try{if(Et.gl=(t=Et.canvas)==null?void 0:t.getContext("webgl2",Et.webGLattr),!Et.gl.getParameter(Et.gl.VERSION).includes("2.0")){se("override: using fallback webgl backend as webgl 2.0 is not detected"),e.config.backend="webgl";return}Et.canvas&&(Et.canvas.addEventListener("webglcontextlost",async a=>{throw se("error: humangl:",a.type),se("possible browser memory leak using webgl or conflict with multiple backend registrations"),e.emit("error"),new Error("backend error: webgl context lost")}),Et.canvas.addEventListener("webglcontextrestored",a=>{se("error: humangl context restored:",a)}),Et.canvas.addEventListener("webglcontextcreationerror",a=>{se("error: humangl context create:",a)}))}catch(n){se("error: cannot get WebGL context:",n);return}try{Im(2,Et.gl)}catch(n){se("error: cannot set WebGL context:",n);return}try{let n=new vu(Et.gl);$l(Et.name,()=>new Bh(n),Et.priority)}catch(n){se("error: cannot register WebGL backend:",n);return}try{Ma("webgl").forEach(a=>{let s={...a,backendName:Et.name};qn(s)})}catch(n){se("error: cannot update WebGL backend registration:",n);return}let r=Dn().getGPGPUContext?Dn().getGPGPUContext().gl:null;if(r)se(`humangl webgl version:${r.getParameter(r.VERSION)} renderer:${r.getParameter(r.RENDERER)}`);else{se("error: no current gl context:",r,Et.gl);return}try{ba.set("WEBGL_VERSION",2)}catch(n){se("error: cannot set WebGL backend flags:",n);return}sxe(),se("backend registered:",Et.name)}}function ixe(){if(!he.kernels.includes("mod")){let e={kernelName:"Mod",backendName:Gr(),kernelFunc:t=>X(()=>ce(t.inputs.a,L(pe(t.inputs.a,t.inputs.b),t.inputs.b)))};qn(e),he.kernels.push("mod")}if(!he.kernels.includes("floormod")){let e={kernelName:"FloorMod",backendName:Gr(),kernelFunc:t=>X(()=>xh(t.inputs.a/t.inputs.b)*t.inputs.b+md(t.inputs.a,t.inputs.b))};qn(e),he.kernels.push("floormod")}}async function A1(e,t=!1){if(e.state="backend",t||he.initial||e.config.backend&&e.config.backend.length>0&&Gr()!==e.config.backend){let r=oe();if(e.config.backend&&e.config.backend.length>0){if(typeof window=="undefined"&&typeof WorkerGlobalScope!="undefined"&&e.config.debug&&e.config.debug&&se("running inside web worker"),he.browser&&e.config.backend==="tensorflow"&&(e.config.debug&&se("override: backend set to tensorflow while running in browser"),e.config.backend="humangl"),he.node&&(e.config.backend==="webgl"||e.config.backend==="humangl")&&(e.config.debug&&se(`override: backend set to ${e.config.backend} while running in nodejs`),e.config.backend="tensorflow"),he.browser&&e.config.backend==="webgpu")if(typeof navigator=="undefined"||typeof navigator.gpu=="undefined")se("override: backend set to webgpu but browser does not support webgpu"),e.config.backend="humangl";else{let a=await navigator.gpu.requestAdapter();e.config.debug&&se("enumerated webgpu adapter:",a)}e.config.backend==="humangl"&&await LN(e);let n=Object.keys(Xt().registryFactory);if(e.config.debug&&se("available backends:",n),n.includes(e.config.backend)||(se(`error: backend ${e.config.backend} not found in registry`),e.config.backend=he.node?"tensorflow":"webgl",e.config.debug&&se(`override: setting backend ${e.config.backend}`)),e.config.debug&&se("setting backend:",e.config.backend),e.config.backend==="wasm"){if(e.config.debug&&se("wasm path:",e.config.wasmPath),typeof(Ue==null?void 0:Ue.setWasmPaths)!="undefined")await _5(e.config.wasmPath,e.config.wasmPlatformFetch);else throw new Error("backend error: attempting to use wasm backend but wasm path is not set");let a=await Z().getAsync("WASM_HAS_SIMD_SUPPORT"),s=await Z().getAsync("WASM_HAS_MULTITHREAD_SUPPORT");e.config.debug&&se(`wasm execution: ${a?"SIMD":"no SIMD"} ${s?"multithreaded":"singlethreaded"}`),e.config.debug&&!a&&se("warning: wasm simd support is not enabled")}try{await zy(e.config.backend),await pd(),$C()}catch(a){return se("error: cannot set backend:",e.config.backend,a),!1}}if(Gr()==="humangl"&&(ba.set("CHECK_COMPUTATION_FOR_ERRORS",!1),ba.set("WEBGL_CPU_FORWARD",!0),ba.set("WEBGL_USE_SHAPES_UNIFORMS",!0),ba.set("CPU_HANDOFF_SIZE_THRESHOLD",256),typeof e.config.deallocate!="undefined"&&e.config.deallocate&&(se("changing webgl: WEBGL_DELETE_TEXTURE_THRESHOLD:",!0),ba.set("WEBGL_DELETE_TEXTURE_THRESHOLD",0)),Dn().getGPGPUContext)){let n=await Dn().getGPGPUContext().gl;e.config.debug&&se(`gl version:${n.getParameter(n.VERSION)} renderer:${n.getParameter(n.RENDERER)}`)}Gr(),Py(),await pd(),e.performance.initBackend=Math.trunc(oe()-r),e.config.backend=Gr(),await he.updateBackend(),ixe()}return!0}function i1(e,t){for(let r of e){let n={kernelName:r,backendName:t.backend,kernelFunc:()=>{t.debug&&se("kernelFunc",r,t.backend)}};qn(n)}he.kernels=Ma(Gr()).map(r=>r.kernelName.toLowerCase())}var i3={};vs(i3,{all:()=>s3,body:()=>Zd,canvas:()=>a3,face:()=>Kd,gesture:()=>Qd,hand:()=>Yd,object:()=>Jd,options:()=>Fr,person:()=>n3});var Wn=e=>{if(!e)se("draw error: invalid canvas");else if(!e.getContext)se("draw error: canvas context not defined");else{let t=e.getContext("2d");if(!t)se("draw error: cannot get canvas context");else return t}return null},Kl=e=>Math.round(e*180/Math.PI),fs=(e,t)=>{if(!t.useDepth||typeof e=="undefined")return t.color;let r=Uint8ClampedArray.from([127+2*e,127-2*e,255]);return`rgba(${r[0]}, ${r[1]}, ${r[2]}, ${t.alpha})`};function ms(e,t,r,n,a){e.fillStyle=fs(n,a),e.beginPath(),e.arc(t,r,a.pointSize,0,2*Math.PI),e.fill()}function Va(e,t,r,n,a,s){if(e.beginPath(),e.lineWidth=s.lineWidth,s.useCurves){let i=(t+t+n)/2,o=(r+r+a)/2;e.ellipse(i,o,n/2,a/2,0,0,2*Math.PI)}else e.moveTo(t+s.roundRect,r),e.lineTo(t+n-s.roundRect,r),e.quadraticCurveTo(t+n,r,t+n,r+s.roundRect),e.lineTo(t+n,r+a-s.roundRect),e.quadraticCurveTo(t+n,r+a,t+n-s.roundRect,r+a),e.lineTo(t+s.roundRect,r+a),e.quadraticCurveTo(t,r+a,t,r+a-s.roundRect),e.lineTo(t,r+s.roundRect),e.quadraticCurveTo(t,r,t+s.roundRect,r),e.closePath();e.stroke()}function e3(e,t,r){if(!(t.length<2)){e.beginPath(),e.moveTo(t[0][0],t[0][1]);for(let n of t)e.strokeStyle=fs(n[2],r),e.lineTo(Math.trunc(n[0]),Math.trunc(n[1]));e.stroke(),r.fillPolygons&&(e.closePath(),e.fill())}}function WN(e,t,r){if(!(t.length<2)){if(e.lineWidth=r.lineWidth,!r.useCurves||t.length<=2){e3(e,t,r);return}e.moveTo(t[0][0],t[0][1]);for(let n=0;n<t.length-2;n++){let a=(t[n][0]+t[n+1][0])/2,s=(t[n][1]+t[n+1][1])/2;e.quadraticCurveTo(t[n][0],t[n][1],a,s)}e.quadraticCurveTo(t[t.length-2][0],t[t.length-2][1],t[t.length-1][0],t[t.length-1][1]),e.stroke(),r.fillPolygons&&(e.closePath(),e.fill())}}function t3(e,t,r,n=5){let a,s,i;e.beginPath(),e.moveTo(t[0],t[1]),e.lineTo(r[0],r[1]),a=Math.atan2(r[1]-t[1],r[0]-t[0]),s=n*Math.cos(a)+r[0],i=n*Math.sin(a)+r[1],e.moveTo(s,i),a+=1/3*(2*Math.PI),s=n*Math.cos(a)+r[0],i=n*Math.sin(a)+r[1],e.lineTo(s,i),a+=1/3*(2*Math.PI),s=n*Math.cos(a)+r[0],i=n*Math.sin(a)+r[1],e.lineTo(s,i),e.closePath(),e.stroke(),e.fill()}var Fr={color:"rgba(173, 216, 230, 0.6)",labelColor:"rgba(173, 216, 230, 1)",shadowColor:"black",alpha:.5,font:'small-caps 16px "Segoe UI"',lineHeight:18,lineWidth:4,pointSize:2,roundRect:8,drawPoints:!1,drawLabels:!0,drawBoxes:!0,drawAttention:!0,drawGestures:!0,drawPolygons:!0,drawGaze:!0,fillPolygons:!1,useDepth:!0,useCurves:!1};var dt;function lxe(e,t){if(dt.drawLabels){let r=[];if(r.push(`face: ${Math.trunc(100*e.score)}%`),e.genderScore&&r.push(`${e.gender||""} ${Math.trunc(100*e.genderScore)}%`),e.age&&r.push(`age: ${e.age||""}`),e.iris&&r.push(`distance: ${e.iris}`),e.real&&r.push(`real: ${Math.trunc(100*e.real)}%`),e.live&&r.push(`live: ${Math.trunc(100*e.live)}%`),e.emotion&&e.emotion.length>0){let n=e.emotion.map(a=>`${Math.trunc(100*a.score)}% ${a.emotion}`);n.length>3&&(n.length=3),r.push(n.join(" "))}e.rotation&&e.rotation.angle&&e.rotation.gaze&&(e.rotation.angle.roll&&r.push(`roll: ${Kl(e.rotation.angle.roll)}\xB0 yaw:${Kl(e.rotation.angle.yaw)}\xB0 pitch:${Kl(e.rotation.angle.pitch)}\xB0`),e.rotation.gaze.bearing&&r.push(`gaze: ${Kl(e.rotation.gaze.bearing)}\xB0`)),r.length===0&&r.push("face"),t.fillStyle=dt.color;for(let n=r.length-1;n>=0;n--){let a=Math.max(e.box[0],0),s=n*dt.lineHeight+e.box[1];dt.shadowColor&&dt.shadowColor!==""&&(t.fillStyle=dt.shadowColor,t.fillText(r[n],a+5,s+16)),t.fillStyle=dt.labelColor,t.fillText(r[n],a+4,s+15)}}}function uxe(e,t){if(e.annotations&&e.annotations.leftEyeIris&&e.annotations.leftEyeIris[0]){t.strokeStyle=dt.useDepth?"rgba(255, 200, 255, 0.3)":dt.color,t.beginPath();let r=Math.abs(e.annotations.leftEyeIris[3][0]-e.annotations.leftEyeIris[1][0])/2,n=Math.abs(e.annotations.leftEyeIris[4][1]-e.annotations.leftEyeIris[2][1])/2;t.ellipse(e.annotations.leftEyeIris[0][0],e.annotations.leftEyeIris[0][1],r,n,0,0,2*Math.PI),t.stroke(),dt.fillPolygons&&(t.fillStyle=dt.useDepth?"rgba(255, 255, 200, 0.3)":dt.color,t.fill())}if(e.annotations&&e.annotations.rightEyeIris&&e.annotations.rightEyeIris[0]){t.strokeStyle=dt.useDepth?"rgba(255, 200, 255, 0.3)":dt.color,t.beginPath();let r=Math.abs(e.annotations.rightEyeIris[3][0]-e.annotations.rightEyeIris[1][0])/2,n=Math.abs(e.annotations.rightEyeIris[4][1]-e.annotations.rightEyeIris[2][1])/2;t.ellipse(e.annotations.rightEyeIris[0][0],e.annotations.rightEyeIris[0][1],r,n,0,0,2*Math.PI),t.stroke(),dt.fillPolygons&&(t.fillStyle=dt.useDepth?"rgba(255, 255, 200, 0.3)":dt.color,t.fill())}}function dxe(e,t){var r;if(dt.drawGaze&&((r=e.rotation)==null?void 0:r.angle)&&typeof Path2D!="undefined"){t.strokeStyle="pink";let n=e.box[0]+e.box[2]/2-e.box[3]*Kl(e.rotation.angle.yaw)/90,a=e.box[1]+e.box[3]/2+e.box[2]*Kl(e.rotation.angle.pitch)/90,s=new Path2D(`
|
|
M ${e.box[0]+e.box[2]/2} ${e.box[1]}
|
|
C
|
|
${n} ${e.box[1]},
|
|
${n} ${e.box[1]+e.box[3]},
|
|
${e.box[0]+e.box[2]/2} ${e.box[1]+e.box[3]}
|
|
`),i=new Path2D(`
|
|
M ${e.box[0]} ${e.box[1]+e.box[3]/2}
|
|
C
|
|
${e.box[0]} ${a},
|
|
${e.box[0]+e.box[2]} ${a},
|
|
${e.box[0]+e.box[2]} ${e.box[1]+e.box[3]/2}
|
|
`);t.stroke(i),t.stroke(s)}}function pxe(e,t){var r,n,a,s;if(dt.drawGaze&&((n=(r=e.rotation)==null?void 0:r.gaze)==null?void 0:n.strength)&&((s=(a=e.rotation)==null?void 0:a.gaze)==null?void 0:s.bearing)&&e.annotations.leftEyeIris&&e.annotations.rightEyeIris&&e.annotations.leftEyeIris[0]&&e.annotations.rightEyeIris[0]){t.strokeStyle="pink",t.fillStyle="pink";let i=[e.annotations.leftEyeIris[0][0]+Math.sin(e.rotation.gaze.bearing)*e.rotation.gaze.strength*e.box[3],e.annotations.leftEyeIris[0][1]+Math.cos(e.rotation.gaze.bearing)*e.rotation.gaze.strength*e.box[2]];t3(t,[e.annotations.leftEyeIris[0][0],e.annotations.leftEyeIris[0][1]],[i[0],i[1]],4);let o=[e.annotations.rightEyeIris[0][0]+Math.sin(e.rotation.gaze.bearing)*e.rotation.gaze.strength*e.box[3],e.annotations.rightEyeIris[0][1]+Math.cos(e.rotation.gaze.bearing)*e.rotation.gaze.strength*e.box[2]];t3(t,[e.annotations.rightEyeIris[0][0],e.annotations.rightEyeIris[0][1]],[o[0],o[1]],4)}}function hxe(e,t){if(dt.drawPolygons&&e.mesh.length>=468){t.lineWidth=1;for(let r=0;r<Vl.length/3;r++){let n=[Vl[r*3+0],Vl[r*3+1],Vl[r*3+2]].map(a=>e.mesh[a]);e3(t,n,dt)}uxe(e,t)}}function cxe(e,t){if(dt.drawPoints&&e.mesh.length>=468)for(let r=0;r<e.mesh.length;r++)ms(t,e.mesh[r][0],e.mesh[r][1],e.mesh[r][2],dt),dt.drawAttention&&(xn.lips.includes(r)&&ms(t,e.mesh[r][0],e.mesh[r][1],e.mesh[r][2]+127,dt),xn.eyeL.includes(r)&&ms(t,e.mesh[r][0],e.mesh[r][1],e.mesh[r][2]-127,dt),xn.eyeR.includes(r)&&ms(t,e.mesh[r][0],e.mesh[r][1],e.mesh[r][2]-127,dt))}function fxe(e,t){dt.drawBoxes&&Va(t,e.box[0],e.box[1],e.box[2],e.box[3],dt)}async function Kd(e,t,r){if(dt=Gt(Fr,r),!t||!e)return;let n=Wn(e);if(!!n){n.font=dt.font,n.strokeStyle=dt.color,n.fillStyle=dt.color;for(let a of t)fxe(a,n),lxe(a,n),a.mesh&&a.mesh.length>0&&(cxe(a,n),hxe(a,n),dxe(a,n),pxe(a,n))}}async function Zd(e,t,r){var s;let n=Gt(Fr,r);if(!t||!e)return;let a=Wn(e);if(!!a){a.lineJoin="round";for(let i=0;i<t.length;i++){if(a.strokeStyle=n.color,a.fillStyle=n.color,a.lineWidth=n.lineWidth,a.font=n.font,n.drawBoxes&&t[i].box&&((s=t[i].box)==null?void 0:s.length)===4&&(Va(a,t[i].box[0],t[i].box[1],t[i].box[2],t[i].box[3],n),n.drawLabels&&(n.shadowColor&&n.shadowColor!==""&&(a.fillStyle=n.shadowColor,a.fillText(`body ${100*t[i].score}%`,t[i].box[0]+3,1+t[i].box[1]+n.lineHeight,t[i].box[2])),a.fillStyle=n.labelColor,a.fillText(`body ${100*t[i].score}%`,t[i].box[0]+2,0+t[i].box[1]+n.lineHeight,t[i].box[2]))),n.drawPoints&&t[i].keypoints)for(let o=0;o<t[i].keypoints.length;o++)!t[i].keypoints[o].score||t[i].keypoints[o].score===0||(a.fillStyle=fs(t[i].keypoints[o].position[2],n),ms(a,t[i].keypoints[o].position[0],t[i].keypoints[o].position[1],0,n));if(n.drawLabels&&t[i].keypoints){a.font=n.font;for(let o of t[i].keypoints)!o.score||o.score===0||(a.fillStyle=fs(o.position[2],n),a.fillText(`${o.part} ${Math.trunc(100*o.score)}%`,o.position[0]+4,o.position[1]+4))}if(n.drawPolygons&&t[i].keypoints&&t[i].annotations)for(let o of Object.values(t[i].annotations))for(let l of o)WN(a,l,n)}}}async function Yd(e,t,r){let n=Gt(Fr,r);if(!t||!e)return;let a=Wn(e);if(!!a){a.lineJoin="round",a.font=n.font;for(let s of t){if(n.drawBoxes&&(a.strokeStyle=n.color,a.fillStyle=n.color,Va(a,s.box[0],s.box[1],s.box[2],s.box[3],n),n.drawLabels&&(n.shadowColor&&n.shadowColor!==""&&(a.fillStyle=n.shadowColor,a.fillText(`hand:${Math.trunc(100*s.score)}%`,s.box[0]+3,1+s.box[1]+n.lineHeight,s.box[2])),a.fillStyle=n.labelColor,a.fillText(`hand:${Math.trunc(100*s.score)}%`,s.box[0]+2,0+s.box[1]+n.lineHeight,s.box[2])),a.stroke()),n.drawPoints&&s.keypoints&&s.keypoints.length>0)for(let i of s.keypoints)a.fillStyle=fs(i[2],n),ms(a,i[0],i[1],0,n);if(n.drawLabels&&s.annotations){let i=(o,l)=>{if(!o||o.length===0||!o[0])return;let u=o[o.length-1][2]||-256;a.fillStyle=fs(u,n),a.fillText(l,o[o.length-1][0]+4,o[o.length-1][1]+4)};a.font=n.font,i(s.annotations.index,"index"),i(s.annotations.middle,"middle"),i(s.annotations.ring,"ring"),i(s.annotations.pinky,"pinky"),i(s.annotations.thumb,"thumb"),i(s.annotations.palm,"palm")}if(n.drawPolygons&&s.annotations){let i=o=>{if(!(!o||o.length===0||!o[0]))for(let l=0;l<o.length;l++){a.beginPath();let u=o[l][2]||0;a.strokeStyle=fs(l*u,n),a.moveTo(o[l>0?l-1:0][0],o[l>0?l-1:0][1]),a.lineTo(o[l][0],o[l][1]),a.stroke()}};a.lineWidth=n.lineWidth,i(s.annotations.index),i(s.annotations.middle),i(s.annotations.ring),i(s.annotations.pinky),i(s.annotations.thumb)}}}}async function Jd(e,t,r){let n=Gt(Fr,r);if(!t||!e)return;let a=Wn(e);if(!!a){a.lineJoin="round",a.font=n.font;for(let s of t)if(n.drawBoxes){if(a.strokeStyle=n.color,a.fillStyle=n.color,Va(a,s.box[0],s.box[1],s.box[2],s.box[3],n),n.drawLabels){let i=`${s.label} ${Math.round(100*s.score)}%`;n.shadowColor&&n.shadowColor!==""&&(a.fillStyle=n.shadowColor,a.fillText(i,s.box[0]+3,1+s.box[1]+n.lineHeight,s.box[2])),a.fillStyle=n.labelColor,a.fillText(i,s.box[0]+2,0+s.box[1]+n.lineHeight,s.box[2])}a.stroke()}}}async function Qd(e,t,r){let n=Gt(Fr,r);if(!(!t||!e)&&n.drawGestures){let a=Wn(e);if(!a)return;a.font=n.font,a.fillStyle=n.color;let s=1;for(let i=0;i<t.length;i++){let o=[],l=[];if([o,l]=Object.entries(t[i]),l.length>1&&l[1].length>0){let u=o[1]>0?`#${o[1]}`:"",d=`${o[0]} ${u}: ${l[1]}`;n.shadowColor&&n.shadowColor!==""&&(a.fillStyle=n.shadowColor,a.fillText(d,8,2+s*n.lineHeight)),a.fillStyle=n.labelColor,a.fillText(d,6,0+s*n.lineHeight),s+=1}}}}var r3=0;async function n3(e,t,r){let n=Gt(Fr,r);if(!t||!e)return;let a=Wn(e);if(!!a){a.lineJoin="round",a.font=n.font;for(let s=0;s<t.length;s++)if(n.drawBoxes){if(a.strokeStyle=n.color,a.fillStyle=n.color,Va(a,t[s].box[0],t[s].box[1],t[s].box[2],t[s].box[3],n),n.drawLabels){let i=`person #${s}`;n.shadowColor&&n.shadowColor!==""&&(a.fillStyle=n.shadowColor,a.fillText(i,t[s].box[0]+3,1+t[s].box[1]+n.lineHeight,t[s].box[2])),a.fillStyle=n.labelColor,a.fillText(i,t[s].box[0]+2,0+t[s].box[1]+n.lineHeight,t[s].box[2])}a.stroke()}}}async function a3(e,t){if(!e||!t)return;let r=Wn(t);!r||r.drawImage(e,0,0)}async function s3(e,t,r){if(!t||!t.performance||!t||!e)return null;let n=oe(),a=Gt(Fr,r),s=Promise.all([Kd(e,t.face,a),Zd(e,t.body,a),Yd(e,t.hand,a),Jd(e,t.object,a),Qd(e,t.gesture,a)]);return r3=he.perfadd?r3+Math.round(oe()-n):Math.round(oe()-n),t.performance.draw=r3,s}var ep=.1,o3=.5;function mxe(e,t,r){let n=!1,a=r.length-1;for(let s=0;s<r.length;a=s++)r[s].y>t!=r[a].y>t&&e<(r[a].x-r[s].x)*(t-r[s].y)/(r[a].y-r[s].y)+r[s].x&&(n=!n);return n}async function VN(e){if(!e.tensor||!e.mesh||e.mesh.length<100)return e.tensor;let t=e.tensor.shape[2]||0,r=e.tensor.shape[1]||0,n=await e.tensor.buffer(),a=[];for(let i of ea.silhouette)a.push({x:(e.mesh[i][0]-e.box[0])/e.box[2],y:(e.mesh[i][1]-e.box[1])/e.box[3]});ep&&ep>0&&(a=a.map(i=>({x:i.x>.5?i.x+ep:i.x-ep,y:i.y>.5?i.y+ep:i.y-ep})));for(let i=0;i<t;i++)for(let o=0;o<r;o++)mxe(i/t,o/t,a)||(n.set(o3*n.get(0,o,i,0),0,o,i,0),n.set(o3*n.get(0,o,i,1),0,o,i,1),n.set(o3*n.get(0,o,i,2),0,o,i,2));let s=n.toTensor();return te(n),s}var yxe=e=>{let t=(h,p)=>Math.atan2(h[1]-p[1],h[0]-p[0]);if(!e.annotations.rightEyeIris||!e.annotations.leftEyeIris)return{bearing:0,strength:0};let r=[0,-.1],n=1,a=(e.mesh[33][2]||0)>(e.mesh[263][2]||0),s=a?e.mesh[473]:e.mesh[468],i=a?[(e.mesh[133][0]+e.mesh[33][0])/2,(e.mesh[133][1]+e.mesh[33][1])/2]:[(e.mesh[263][0]+e.mesh[362][0])/2,(e.mesh[263][1]+e.mesh[362][1])/2],o=a?[e.mesh[133][0]-e.mesh[33][0],e.mesh[23][1]-e.mesh[27][1]]:[e.mesh[263][0]-e.mesh[362][0],e.mesh[253][1]-e.mesh[257][1]],l=[(i[0]-s[0])/o[0]-r[0],n*(s[1]-i[1])/o[1]-r[1]],u=Math.sqrt(l[0]*l[0]+l[1]*l[1]);return u=Math.min(u,e.boxRaw[2]/2,e.boxRaw[3]/2),{bearing:(t([0,0],l)+Math.PI/2)%Math.PI,strength:u}},UN=(e,t)=>{let r=m=>{let g=Math.sqrt(m[0]*m[0]+m[1]*m[1]+m[2]*m[2]);return m[0]/=g,m[1]/=g,m[2]/=g,m},n=(m,g)=>{let y=m[0]-g[0],A=m[1]-g[1],x=m[2]-g[2];return[y,A,x]},a=(m,g)=>{let y=m[1]*g[2]-m[2]*g[1],A=m[2]*g[0]-m[0]*g[2],x=m[0]*g[1]-m[1]*g[0];return[y,A,x]},s=m=>{let[g,y,A,x,b,w,S,N,E]=m,R,z,M;return x<1?x>-1?(M=Math.asin(x),z=Math.atan2(-S,g),R=Math.atan2(-w,b)):(M=-Math.PI/2,z=-Math.atan2(N,E),R=0):(M=Math.PI/2,z=Math.atan2(N,E),R=0),isNaN(R)&&(R=0),isNaN(z)&&(z=0),isNaN(M)&&(M=0),{pitch:2*-R,yaw:2*-z,roll:2*-M}},i=e.meshRaw;if(!i||i.length<300)return{angle:{pitch:0,yaw:0,roll:0},matrix:[1,0,0,0,1,0,0,0,1],gaze:{bearing:0,strength:0}};let o=Math.max(e.boxRaw[2]*t[0],e.boxRaw[3]*t[1])/1.5,l=[i[10],i[152],i[234],i[454]].map(m=>[m[0]*t[0]/o,m[1]*t[1]/o,m[2]]),u=r(n(l[1],l[0])),d=r(n(l[3],l[2])),h=r(a(d,u));d=a(u,h);let p=[d[0],d[1],d[2],u[0],u[1],u[2],h[0],h[1],h[2]],c=s(p),f=i.length===478?yxe(e):{bearing:0,strength:0};return{angle:c,matrix:p,gaze:f}};var l3=async(e,t)=>{var c,f,m,g,y,A,x,b,w,S,N,E,R,z,M,I,_,O,q,K,D,Q;let r=oe(),n,a,s,i,o,l,u,d,h=[];e.state="run:face";let p=await _T(t,e.config);if(e.performance.face=he.perfadd?(e.performance.face||0)+Math.trunc(oe()-r):Math.trunc(oe()-r),!t.shape||t.shape.length!==4)return[];if(!p)return[];for(let V=0;V<p.length;V++){if(e.analyze("Get Face"),!p[V].tensor||p[V].tensor.isDisposedInternal){se("Face object is disposed:",p[V].tensor);continue}if((c=e.config.face.detector)!=null&&c.mask){let ae=await VN(p[V]);te(p[V].tensor),p[V].tensor=ae}let re=p[V].mesh&&p[V].mesh.length>200?UN(p[V],[t.shape[2],t.shape[1]]):null;e.analyze("Start Emotion:"),e.config.async?i=(f=e.config.face.emotion)!=null&&f.enabled?gb(p[V].tensor||ft([]),e.config,V,p.length):[]:(e.state="run:emotion",r=oe(),i=(m=e.config.face.emotion)!=null&&m.enabled?await gb(p[V].tensor||ft([]),e.config,V,p.length):[],e.performance.emotion=he.perfadd?(e.performance.emotion||0)+Math.trunc(oe()-r):Math.trunc(oe()-r)),e.analyze("End Emotion:"),e.analyze("Start AntiSpoof:"),e.config.async?l=(g=e.config.face.antispoof)!=null&&g.enabled?K5(p[V].tensor||ft([]),e.config,V,p.length):0:(e.state="run:antispoof",r=oe(),l=(y=e.config.face.antispoof)!=null&&y.enabled?await K5(p[V].tensor||ft([]),e.config,V,p.length):0,e.performance.antispoof=he.perfadd?(e.performance.antispoof||0)+Math.trunc(oe()-r):Math.trunc(oe()-r)),e.analyze("End AntiSpoof:"),e.analyze("Start Liveness:"),e.config.async?u=(A=e.config.face.liveness)!=null&&A.enabled?zb(p[V].tensor||ft([]),e.config,V,p.length):0:(e.state="run:liveness",r=oe(),u=(x=e.config.face.liveness)!=null&&x.enabled?await zb(p[V].tensor||ft([]),e.config,V,p.length):0,e.performance.liveness=he.perfadd?(e.performance.antispoof||0)+Math.trunc(oe()-r):Math.trunc(oe()-r)),e.analyze("End Liveness:"),e.analyze("Start GEAR:"),e.config.async?a=(b=e.config.face.gear)!=null&&b.enabled?V5(p[V].tensor||ft([]),e.config,V,p.length):null:(e.state="run:gear",r=oe(),a=(w=e.config.face.gear)!=null&&w.enabled?await V5(p[V].tensor||ft([]),e.config,V,p.length):null,e.performance.gear=Math.trunc(oe()-r)),e.analyze("End GEAR:"),e.analyze("Start SSRNet:"),e.config.async?(n=(S=e.config.face.ssrnet)!=null&&S.enabled?G5(p[V].tensor||ft([]),e.config,V,p.length):null,s=(N=e.config.face.ssrnet)!=null&&N.enabled?q5(p[V].tensor||ft([]),e.config,V,p.length):null):(e.state="run:ssrnet",r=oe(),n=(E=e.config.face.ssrnet)!=null&&E.enabled?await G5(p[V].tensor||ft([]),e.config,V,p.length):null,s=(R=e.config.face.ssrnet)!=null&&R.enabled?await q5(p[V].tensor||ft([]),e.config,V,p.length):null,e.performance.ssrnet=Math.trunc(oe()-r)),e.analyze("End SSRNet:"),e.analyze("Start MobileFaceNet:"),e.config.async?o=(z=e.config.face.mobilefacenet)!=null&&z.enabled?Ab(p[V].tensor||ft([]),e.config,V,p.length):null:(e.state="run:mobilefacenet",r=oe(),o=(M=e.config.face.mobilefacenet)!=null&&M.enabled?await Ab(p[V].tensor||ft([]),e.config,V,p.length):null,e.performance.mobilefacenet=Math.trunc(oe()-r)),e.analyze("End MobileFaceNet:"),e.analyze("Start Description:"),e.config.async?d=(I=e.config.face.description)!=null&&I.enabled?Ib(p[V].tensor||ft([]),e.config,V,p.length):null:(e.state="run:description",r=oe(),d=(_=e.config.face.description)!=null&&_.enabled?await Ib(p[V].tensor||ft([]),e.config,V,p.length):null,e.performance.description=he.perfadd?(e.performance.description||0)+Math.trunc(oe()-r):Math.trunc(oe()-r)),e.analyze("End Description:"),e.config.async&&([n,s,i,o,d,a,l,u]=await Promise.all([n,s,i,o,d,a,l,u])),e.analyze("Finish Face:"),((O=e.config.face.ssrnet)==null?void 0:O.enabled)&&n&&s&&(d={...d,age:n.age,gender:s.gender,genderScore:s.genderScore}),((q=e.config.face.gear)==null?void 0:q.enabled)&&a&&(d={...d,age:a.age,gender:a.gender,genderScore:a.genderScore,race:a.race}),((K=e.config.face.mobilefacenet)==null?void 0:K.enabled)&&o&&(d.descriptor=o),(D=e.config.face.iris)!=null&&D.enabled;let Y=p[V].annotations&&p[V].annotations.leftEyeIris&&p[V].annotations.leftEyeIris[0]&&p[V].annotations.rightEyeIris&&p[V].annotations.rightEyeIris[0]&&p[V].annotations.leftEyeIris.length>0&&p[V].annotations.rightEyeIris.length>0&&p[V].annotations.leftEyeIris[0]!==null&&p[V].annotations.rightEyeIris[0]!==null?Math.max(Math.abs(p[V].annotations.leftEyeIris[3][0]-p[V].annotations.leftEyeIris[1][0]),Math.abs(p[V].annotations.rightEyeIris[4][1]-p[V].annotations.rightEyeIris[2][1]))/t.shape[2]:0,ie=(Q=e.config.face.detector)!=null&&Q.return?tt(p[V].tensor):null;te(p[V].tensor),p[V].tensor&&delete p[V].tensor;let J={...p[V],id:V};d!=null&&d.age&&(J.age=d.age),d!=null&&d.gender&&(J.gender=d.gender),d!=null&&d.genderScore&&(J.genderScore=d==null?void 0:d.genderScore),d!=null&&d.descriptor&&(J.embedding=d==null?void 0:d.descriptor),d!=null&&d.race&&(J.race=d==null?void 0:d.race),i&&(J.emotion=i),l&&(J.real=l),u&&(J.live=u),Y&&Y!==0&&(J.iris=Math.trunc(500/Y/11.7)/100),re&&(J.rotation=re),ie&&(J.tensor=ie),h.push(J),e.analyze("End Face")}return e.analyze("End FaceMesh:"),e.config.async&&(e.performance.face&&delete e.performance.face,e.performance.age&&delete e.performance.age,e.performance.gender&&delete e.performance.gender,e.performance.emotion&&delete e.performance.emotion),h};var GN=e=>{if(!e)return[];let t=[];for(let r=0;r<e.length;r++){let n=e[r].keypoints.find(l=>l.part==="leftWrist"),a=e[r].keypoints.find(l=>l.part==="rightWrist"),s=e[r].keypoints.find(l=>l.part==="nose");s&&n&&a&&n.position[1]<s.position[1]&&a.position[1]<s.position[1]?t.push({body:r,gesture:"i give up"}):s&&n&&n.position[1]<s.position[1]?t.push({body:r,gesture:"raise left hand"}):s&&a&&a.position[1]<s.position[1]&&t.push({body:r,gesture:"raise right hand"});let i=e[r].keypoints.find(l=>l.part==="leftShoulder"),o=e[r].keypoints.find(l=>l.part==="rightShoulder");i&&o&&Math.abs(i.positionRaw[1]-o.positionRaw[1])>.1&&t.push({body:r,gesture:`leaning ${i.position[1]>o.position[1]?"left":"right"}`})}return t},jN=e=>{if(!e)return[];let t=[];for(let r=0;r<e.length;r++)if(e[r].mesh&&e[r].mesh.length>450){let n=(e[r].mesh[33][2]||0)-(e[r].mesh[263][2]||0),a=e[r].mesh[33][0]-e[r].mesh[263][0];Math.abs(n/a)<=.15?t.push({face:r,gesture:"facing center"}):t.push({face:r,gesture:`facing ${n<0?"left":"right"}`}),Math.abs(e[r].mesh[374][1]-e[r].mesh[386][1])/Math.abs(e[r].mesh[443][1]-e[r].mesh[450][1])<.2&&t.push({face:r,gesture:"blink left eye"}),Math.abs(e[r].mesh[145][1]-e[r].mesh[159][1])/Math.abs(e[r].mesh[223][1]-e[r].mesh[230][1])<.2&&t.push({face:r,gesture:"blink right eye"});let o=Math.min(100,500*Math.abs(e[r].mesh[13][1]-e[r].mesh[14][1])/Math.abs(e[r].mesh[10][1]-e[r].mesh[152][1]));o>10&&t.push({face:r,gesture:`mouth ${Math.trunc(o)}% open`});let l=e[r].mesh[152][2]||0;Math.abs(l)>10&&t.push({face:r,gesture:`head ${l<0?"up":"down"}`})}return t},HN=e=>{if(!e)return[];let t=[];for(let r=0;r<e.length;r++){if(!e[r].annotations||!e[r].annotations.leftEyeIris||!e[r].annotations.leftEyeIris[0]||!e[r].annotations.rightEyeIris||!e[r].annotations.rightEyeIris[0])continue;let n=e[r].annotations.leftEyeIris[3][0]-e[r].annotations.leftEyeIris[1][0],a=e[r].annotations.leftEyeIris[4][1]-e[r].annotations.leftEyeIris[2][1],s=Math.abs(n*a),i=e[r].annotations.rightEyeIris[3][0]-e[r].annotations.rightEyeIris[1][0],o=e[r].annotations.rightEyeIris[4][1]-e[r].annotations.rightEyeIris[2][1],l=Math.abs(i*o),u=!1;Math.abs(s-l)/Math.max(s,l)<.25&&(u=!0,t.push({iris:r,gesture:"facing center"}));let h=Math.abs(e[r].mesh[263][0]-e[r].annotations.leftEyeIris[0][0])/e[r].box[2],p=Math.abs(e[r].mesh[33][0]-e[r].annotations.rightEyeIris[0][0])/e[r].box[2];(h>.06||p>.06)&&(u=!1),h>p?h>.05&&t.push({iris:r,gesture:"looking right"}):p>.05&&t.push({iris:r,gesture:"looking left"});let c=Math.abs(e[r].mesh[145][1]-e[r].annotations.rightEyeIris[0][1])/e[r].box[3],f=Math.abs(e[r].mesh[374][1]-e[r].annotations.leftEyeIris[0][1])/e[r].box[3];(f<.01||c<.01||f>.022||c>.022)&&(u=!1),(f<.01||c<.01)&&t.push({iris:r,gesture:"looking down"}),(f>.022||c>.022)&&t.push({iris:r,gesture:"looking up"}),u&&t.push({iris:r,gesture:"looking center"})}return t},qN=e=>{if(!e)return[];let t=[];for(let r=0;r<e.length;r++){let n=[];if(e[r].annotations)for(let[a,s]of Object.entries(e[r].annotations))a!=="palmBase"&&Array.isArray(s)&&s[0]&&n.push({name:a.toLowerCase(),position:s[0]});if(n&&n.length>0){let a=n.reduce((i,o)=>(i.position[2]||0)<(o.position[2]||0)?i:o);t.push({hand:r,gesture:`${a.name} forward`});let s=n.reduce((i,o)=>i.position[1]<o.position[1]?i:o);t.push({hand:r,gesture:`${s.name} up`})}if(e[r].keypoints){let a=iN(e[r].keypoints);for(let s of a)t.push({hand:r,gesture:s.name})}}return t};var Ne={face:[],body:[],hand:[],gesture:[],object:[],persons:[],performance:{},timestamp:0,error:null},u3=0;function XN(e,t){var i,o,l,u,d,h,p,c,f,m,g,y,A,x,b,w,S,N,E,R,z,M,I,_,O,q,K;let r=oe();if(!e)return{face:[],body:[],hand:[],gesture:[],object:[],persons:[],performance:{},timestamp:0,error:null};let n=Date.now()-e.timestamp,a=n<1e3?8-Math.log(n+1):1;if(e.canvas&&(Ne.canvas=e.canvas),e.error&&(Ne.error=e.error),!Ne.body||e.body.length!==Ne.body.length)Ne.body=JSON.parse(JSON.stringify(e.body));else for(let D=0;D<e.body.length;D++){let Q=e.body[D].box.map((J,ae)=>((a-1)*Ne.body[D].box[ae]+J)/a),V=e.body[D].boxRaw.map((J,ae)=>((a-1)*Ne.body[D].boxRaw[ae]+J)/a),re=e.body[D].keypoints.map((J,ae)=>{var de,xe,be,Ee,$e,De,Be,Xe,ot;return{score:J.score,part:J.part,position:[Ne.body[D].keypoints[ae]?((a-1)*(Ne.body[D].keypoints[ae].position[0]||0)+(J.position[0]||0))/a:J.position[0],Ne.body[D].keypoints[ae]?((a-1)*(Ne.body[D].keypoints[ae].position[1]||0)+(J.position[1]||0))/a:J.position[1],Ne.body[D].keypoints[ae]?((a-1)*(Ne.body[D].keypoints[ae].position[2]||0)+(J.position[2]||0))/a:J.position[2]],positionRaw:[Ne.body[D].keypoints[ae]?((a-1)*(Ne.body[D].keypoints[ae].positionRaw[0]||0)+(J.positionRaw[0]||0))/a:J.positionRaw[0],Ne.body[D].keypoints[ae]?((a-1)*(Ne.body[D].keypoints[ae].positionRaw[1]||0)+(J.positionRaw[1]||0))/a:J.positionRaw[1],Ne.body[D].keypoints[ae]?((a-1)*(Ne.body[D].keypoints[ae].positionRaw[2]||0)+(J.positionRaw[2]||0))/a:J.positionRaw[2]],distance:[Ne.body[D].keypoints[ae]?((a-1)*(((de=Ne.body[D].keypoints[ae].distance)==null?void 0:de[0])||0)+(((xe=J.distance)==null?void 0:xe[0])||0))/a:(be=J.distance)==null?void 0:be[0],Ne.body[D].keypoints[ae]?((a-1)*(((Ee=Ne.body[D].keypoints[ae].distance)==null?void 0:Ee[1])||0)+((($e=J.distance)==null?void 0:$e[1])||0))/a:(De=J.distance)==null?void 0:De[1],Ne.body[D].keypoints[ae]?((a-1)*(((Be=Ne.body[D].keypoints[ae].distance)==null?void 0:Be[2])||0)+(((Xe=J.distance)==null?void 0:Xe[2])||0))/a:(ot=J.distance)==null?void 0:ot[2]]}}),Y={},ie={connected:{}};(o=(i=t.body)==null?void 0:i.modelPath)!=null&&o.includes("efficientpose")?ie=Km:(u=(l=t.body)==null?void 0:l.modelPath)!=null&&u.includes("blazepose")?ie=Gm:(h=(d=t.body)==null?void 0:d.modelPath)!=null&&h.includes("movenet")&&(ie=rc);for(let[J,ae]of Object.entries(ie.connected)){let de=[];for(let xe=0;xe<ae.length-1;xe++){let be=re.find($e=>$e.part===ae[xe]),Ee=re.find($e=>$e.part===ae[xe+1]);be&&Ee&&de.push([be.position,Ee.position])}Y[J]=de}Ne.body[D]={...e.body[D],box:Q,boxRaw:V,keypoints:re,annotations:Y}}if(!Ne.hand||e.hand.length!==Ne.hand.length)Ne.hand=JSON.parse(JSON.stringify(e.hand));else for(let D=0;D<e.hand.length;D++){let Q=e.hand[D].box.map((ie,J)=>((a-1)*Ne.hand[D].box[J]+ie)/a),V=e.hand[D].boxRaw.map((ie,J)=>((a-1)*Ne.hand[D].boxRaw[J]+ie)/a);Ne.hand[D].keypoints.length!==e.hand[D].keypoints.length&&(Ne.hand[D].keypoints=e.hand[D].keypoints);let re=e.hand[D].keypoints&&e.hand[D].keypoints.length>0?e.hand[D].keypoints.map((ie,J)=>ie.map((ae,de)=>((a-1)*(Ne.hand[D].keypoints[J][de]||1)+(ae||0))/a)):[],Y={};if(Object.keys(Ne.hand[D].annotations).length!==Object.keys(e.hand[D].annotations).length)Ne.hand[D].annotations=e.hand[D].annotations,Y=Ne.hand[D].annotations;else if(e.hand[D].annotations)for(let ie of Object.keys(e.hand[D].annotations))Y[ie]=e.hand[D].annotations[ie]&&e.hand[D].annotations[ie][0]?e.hand[D].annotations[ie].map((J,ae)=>J.map((de,xe)=>((a-1)*Ne.hand[D].annotations[ie][ae][xe]+de)/a)):null;Ne.hand[D]={...e.hand[D],box:Q,boxRaw:V,keypoints:re,annotations:Y}}if(!Ne.face||e.face.length!==Ne.face.length)Ne.face=JSON.parse(JSON.stringify(e.face));else for(let D=0;D<e.face.length;D++){let Q=e.face[D].box.map((re,Y)=>((a-1)*Ne.face[D].box[Y]+re)/a),V=e.face[D].boxRaw.map((re,Y)=>((a-1)*Ne.face[D].boxRaw[Y]+re)/a);if(e.face[D].rotation){let re={matrix:[0,0,0,0,0,0,0,0,0],angle:{roll:0,yaw:0,pitch:0},gaze:{bearing:0,strength:0}};re.matrix=(p=e.face[D].rotation)==null?void 0:p.matrix,re.angle={roll:((a-1)*(((f=(c=Ne.face[D].rotation)==null?void 0:c.angle)==null?void 0:f.roll)||0)+(((g=(m=e.face[D].rotation)==null?void 0:m.angle)==null?void 0:g.roll)||0))/a,yaw:((a-1)*(((A=(y=Ne.face[D].rotation)==null?void 0:y.angle)==null?void 0:A.yaw)||0)+(((b=(x=e.face[D].rotation)==null?void 0:x.angle)==null?void 0:b.yaw)||0))/a,pitch:((a-1)*(((S=(w=Ne.face[D].rotation)==null?void 0:w.angle)==null?void 0:S.pitch)||0)+(((E=(N=e.face[D].rotation)==null?void 0:N.angle)==null?void 0:E.pitch)||0))/a},re.gaze={bearing:((a-1)*(((z=(R=Ne.face[D].rotation)==null?void 0:R.gaze)==null?void 0:z.bearing)||0)+(((I=(M=e.face[D].rotation)==null?void 0:M.gaze)==null?void 0:I.bearing)||0))/a,strength:((a-1)*(((O=(_=Ne.face[D].rotation)==null?void 0:_.gaze)==null?void 0:O.strength)||0)+(((K=(q=e.face[D].rotation)==null?void 0:q.gaze)==null?void 0:K.strength)||0))/a},Ne.face[D]={...e.face[D],rotation:re,box:Q,boxRaw:V}}Ne.face[D]={...e.face[D],box:Q,boxRaw:V}}if(!Ne.object||e.object.length!==Ne.object.length)Ne.object=JSON.parse(JSON.stringify(e.object));else for(let D=0;D<e.object.length;D++){let Q=e.object[D].box.map((re,Y)=>((a-1)*Ne.object[D].box[Y]+re)/a),V=e.object[D].boxRaw.map((re,Y)=>((a-1)*Ne.object[D].boxRaw[Y]+re)/a);Ne.object[D]={...e.object[D],box:Q,boxRaw:V}}if(e.persons){let D=e.persons;if(!Ne.persons||D.length!==Ne.persons.length)Ne.persons=JSON.parse(JSON.stringify(D));else for(let Q=0;Q<D.length;Q++)Ne.persons[Q].box=D[Q].box.map((V,re)=>((a-1)*Ne.persons[Q].box[re]+V)/a)}e.gesture&&(Ne.gesture=e.gesture);let s=oe();return u3=he.perfadd?u3+Math.round(s-r):Math.round(s-r),e.performance&&(Ne.performance={...e.performance,interpolate:u3}),Ne}var h3={};vs(h3,{distance:()=>ic,match:()=>p3,similarity:()=>d3});function ic(e,t,r={order:2,multiplier:25}){let n=0;for(let a=0;a<e.length;a++){let s=!r.order||r.order===2?e[a]-t[a]:Math.abs(e[a]-t[a]);n+=!r.order||r.order===2?s*s:s**r.order}return(r.multiplier||20)*n}var KN=(e,t,r,n)=>{if(e===0)return 1;let a=t===2?Math.sqrt(e):e**(1/t),s=(1-a/100-r)/(n-r);return Math.max(Math.min(s,1),0)};function d3(e,t,r={order:2,multiplier:25,min:.2,max:.8}){let n=ic(e,t,r);return KN(n,r.order||2,r.min||0,r.max||1)}function p3(e,t,r={order:2,multiplier:25,threshold:0,min:.2,max:.8}){if(!Array.isArray(e)||!Array.isArray(t)||e.length<64||t.length===0||e.length!==t[0].length)return{index:-1,distance:Number.POSITIVE_INFINITY,similarity:0};let n=Number.MAX_SAFE_INTEGER,a=-1;for(let i=0;i<t.length;i++){let o=ic(e,t[i],r);if(o<n&&(n=o,a=i),n<(r.threshold||0))break}let s=KN(n,r.order||2,r.min||0,r.max||1);return{index:a,distance:n,similarity:s}}function ZN(e,t,r,n,a){var o,l,u,d,h,p,c,f,m,g,y,A,x,b,w,S;let s=0,i=[];for(let N of e){let E={id:s++,face:N,body:null,hands:{left:null,right:null},gestures:[],box:[0,0,0,0]};for(let O of t)N.box[0]>O.box[0]&&N.box[0]<O.box[0]+O.box[2]&&N.box[1]+N.box[3]>O.box[1]&&N.box[1]+N.box[3]<O.box[1]+O.box[3]&&(E.body=O);if(E.body)for(let O of r)O.box[0]+O.box[2]>E.body.box[0]&&O.box[0]+O.box[2]<E.body.box[0]+E.body.box[2]&&O.box[1]+O.box[3]>E.body.box[1]&&O.box[1]+O.box[3]<E.body.box[1]+E.body.box[3]&&E.hands&&(E.hands.left=O),O.box[0]<E.body.box[0]+E.body.box[2]&&O.box[0]>E.body.box[0]&&O.box[1]+O.box[3]>E.body.box[1]&&O.box[1]+O.box[3]<E.body.box[1]+E.body.box[3]&&E.hands&&(E.hands.right=O);for(let O of n)O.face!==void 0&&O.face===N.id?(o=E.gestures)==null||o.push(O):O.iris!==void 0&&O.iris===N.id?(l=E.gestures)==null||l.push(O):O.body!==void 0&&O.body===((u=E.body)==null?void 0:u.id)?(d=E.gestures)==null||d.push(O):O.hand!==void 0&&O.hand===((p=(h=E.hands)==null?void 0:h.left)==null?void 0:p.id)?(c=E.gestures)==null||c.push(O):O.hand!==void 0&&O.hand===((m=(f=E.hands)==null?void 0:f.right)==null?void 0:m.id)&&((g=E.gestures)==null||g.push(O));let R=[],z=[],M=O=>{O&&O.length===4&&(R.push(O[0],O[0]+O[2]),z.push(O[1],O[1]+O[3]))};M((y=E.face)==null?void 0:y.box),M((A=E.body)==null?void 0:A.box),M((b=(x=E.hands)==null?void 0:x.left)==null?void 0:b.box),M((S=(w=E.hands)==null?void 0:w.right)==null?void 0:S.box);let I=Math.min(...R),_=Math.min(...z);E.box=[I,_,Math.max(...R)-I,Math.max(...z)-_],a&&a[1]&&a[2]&&(E.boxRaw=[E.box[0]/a[2],E.box[1]/a[1],E.box[2]/a[2],E.box[3]/a[1]]),i.push(E)}return i}var x1=`
|
|
/9j/4AAQSkZJRgABAQEAYABgAAD/4QBoRXhpZgAATU0AKgAAAAgABAEaAAUAAAABAAAAPgEbAAUA
|
|
AAABAAAARgEoAAMAAAABAAIAAAExAAIAAAARAAAATgAAAAAAAABgAAAAAQAAAGAAAAABcGFpbnQu
|
|
bmV0IDQuMi4xMwAA/9sAQwAGBAUGBQQGBgUGBwcGCAoQCgoJCQoUDg8MEBcUGBgXFBYWGh0lHxob
|
|
IxwWFiAsICMmJykqKRkfLTAtKDAlKCko/9sAQwEHBwcKCAoTCgoTKBoWGigoKCgoKCgoKCgoKCgo
|
|
KCgoKCgoKCgoKCgoKCgoKCgoKCgoKCgoKCgoKCgoKCgoKCgo/8AAEQgBAAEAAwEhAAIRAQMRAf/E
|
|
AB8AAAEFAQEBAQEBAAAAAAAAAAABAgMEBQYHCAkKC//EALUQAAIBAwMCBAMFBQQEAAABfQECAwAE
|
|
EQUSITFBBhNRYQcicRQygZGhCCNCscEVUtHwJDNicoIJChYXGBkaJSYnKCkqNDU2Nzg5OkNERUZH
|
|
SElKU1RVVldYWVpjZGVmZ2hpanN0dXZ3eHl6g4SFhoeIiYqSk5SVlpeYmZqio6Slpqeoqaqys7S1
|
|
tre4ubrCw8TFxsfIycrS09TV1tfY2drh4uPk5ebn6Onq8fLz9PX29/j5+v/EAB8BAAMBAQEBAQEB
|
|
AQEAAAAAAAABAgMEBQYHCAkKC//EALURAAIBAgQEAwQHBQQEAAECdwABAgMRBAUhMQYSQVEHYXET
|
|
IjKBCBRCkaGxwQkjM1LwFWJy0QoWJDThJfEXGBkaJicoKSo1Njc4OTpDREVGR0hJSlNUVVZXWFla
|
|
Y2RlZmdoaWpzdHV2d3h5eoKDhIWGh4iJipKTlJWWl5iZmqKjpKWmp6ipqrKztLW2t7i5usLDxMXG
|
|
x8jJytLT1NXW19jZ2uLj5OXm5+jp6vLz9PX29/j5+v/aAAwDAQACEQMRAD8A+qaKACigApGOKAML
|
|
Xp8xlF5A7V4X8RtYs7PzfNImnx8sa8Kp9z3q2tEgp6angWs62ZZ5CTGoJ6DArGNz5p+UrID6EUrF
|
|
PUlW1EuN0XNW7PQ2L5j3JnoKXN0KijqNP0eYoqXBdgPuuo+ZPeupisWn2Jd4+0r924XgsQOCff3/
|
|
AJ1FzRKxDqGii6m3siiQ8F1XGfXI6YNWLfRbiRQMkcZI9fpTDluT2/h6Qy8gDPbtmtG38JeY480Z
|
|
5zSLUTZg8M28YwYxjAArXtdPt402qgHbpSaLWhma3o0Uqk7Nx9DWLaaVblgPs6qRyds2M/gRSQp9
|
|
zZOni2iWS2hlQ+kjYz9OMGrdjq89vIPPVhj+8M/lQyDq9P1WOYBlMZz1AOD+VdDaTiReOKulK0jO
|
|
tHmi0WDTlr0TyxRVhT8tJjIX+9SUxHXUV553BRQAVBcPhSBTSuxPY86+IGti0s5I7dsORy9fM3i6
|
|
8e8mfDO5P90ZrWWiJicNPpZZtxV/xrW0jQt4DOv6Vk2dEEdTY6BHuB25rpbPSo0QARjP0qTRI17W
|
|
wA/hFaMWmoQMgflQXYsDS142rU9tpqqenfNA7GgtihxkdKuRW6qMY/GkDZY8sY4Ap4hXbyB+VArk
|
|
EtuH4wPyrk/EGkOm+a3jw3suRQLc5i38SX9hJ9nnY+XnBUdPyNdFY6pa3KkkAE9l6f8AfJ/pSJT6
|
|
GhDmI+Zb4ZRycdv6ium0nUhKFydrelTsNnS2829RnrVgV6NKXNG55lWPLIM81Op+WrZkRMfmNNzT
|
|
A7GivPO4KKAEY4XNYWt3vkwPg4OK0giJdjw/xrqhm87Zs8tc7pX5A+leSajf6aHYJ50kn4AZpTep
|
|
rBWRm2Vobm4BXfyehPFdnpmnBFUY5rI2SN63tlToK0YI+KZpFF+3QdavwoKTLtoW0Toaswpk5pCb
|
|
LCxipAhoIuP2dKevHXoaYDylRyxhlwRQI4nxVoCXWZI1GfpXGtbSWjYPGP73+NIGupt6TqMsLruZ
|
|
ih4xnP5V09mQ+JLd8gn0xSYJnVaVdkook69K34zuUGunDS3Rx4qOzHVIp4rrOMY3NJQI7GivPO8K
|
|
KAILt9kZrz3xlebYiu8KCCWb0XvW0NFch6ysfO3jLVjfXLIn+pQkKorl7WxNxIPl71g2dUUdpo+l
|
|
pBGvHPet23iC8ihFosrxirkHQUFo0IF4FXI1O726CpKLacCrMJoJLYHAPpTwucHpSRJJ5e4AZI9x
|
|
UqpxzVpCuOC8cUpQUMRnXttuB4rjNdsYyeVwfXpmpGmcvcQyafMCFJjPY10eg34BUg4DcZP8jUO4
|
|
HaRq3lLNF+IHet7R7jz7c56rwa2wz9+xhiVeFy/T1PFegeaNPWigDsc0ZrzzvDNIaAM7VpNqdegr
|
|
xL4l6kywyRhseZ19lrdfAZL4jxYg3Fw20d63tJsdrDI5rm3Z3R0R0Mce1eKnQYAplIkWrMJ45oZS
|
|
NO3PHbNXIyfpSGWowSOasxLUiZdjFSqtNEMkUemKlAGKsRJjAppFAiORMjmsTVrNZEO4cfSoZSOD
|
|
1eJ7WXBUzQZ+7nkfSo7e2Ei+ZaMzxntjBX2NSU1Y6/wxqojiEFzkA8KTXYaUoWRyv3W5rSjpNHPX
|
|
+BmpSg8V6J5gUUAdhRXnneFFAGHrTfu5PpXzj8S70/aZtxzztXFbv4DKHxHI+H4GZiz9zxXXW8G3
|
|
GBXMjvLRXAx0oPGPSmMVeOnWrMTYpFI0bcg1fh54xmgovRcD3qxETSIZcRvzp+/BpEkqsBUqsM9K
|
|
q4Em4Gkxk0yRGXrVW6i8yFhkg+tJjRxGsWrxllkUMh9eK5uMz6bcebbnfG33kPcVkay2OntPKuo0
|
|
nhXI67c8qa7Lw3c+adjcEDGK1paSRhVV4s6A0or0jyRRQ1AHX0V553hRQBz+vNtt5z3xXzX8Qbdm
|
|
uic5YnOMdK3l8JnTXvlbwpYl+WySOgrp5YfLOOB9O1c62O7qQkc+9RsKChFPWp4DluOlSykaNruH
|
|
ArUgHShFNF2NT1qxGO3NBmyxGcE1N2560CFzjrUysO9JAPDDjFOVuKoQuSRTWouBkazbCa3cd8cV
|
|
wF7IISQccHBzUSWpV9C3o1x5b5GAjdQD1rs9DjC3kckbEhqKfxIzn8LOupRXqnkPccBSkUAzraK8
|
|
87wooA5rxMSI3HqK8B8bQl9Q8sffY5b/AAraXwkUviNrw9pH2W1ViMMRTdRjw4HpWNtDti9TPc4P
|
|
FQs2M5qdyyMHLcfjV63HTAoBGtap0wK0YxigpsuRDtVhVYd6GQydVwwIqdRnqKCR23I5pCMUW6gD
|
|
YNKuetAEise9KTxQBWuFyhrznxNZkXjFeN3I+tTIZg2OqmzmxNF0PO3vXp/g2+hukVl4zyPanTXv
|
|
JmVR+60dpThXpnlPceopWFAbnV0V553hSGgRynjC5FujOey14Ssp1HxNmTnc+a3kvcIpv37HoEYQ
|
|
QmMdVHSsnVbYJF5jVk0dsNzlruVIsl2wKxbjWrVHILjg1CRbZJb+ILHPzyhfStODWLQgFJFYd+el
|
|
UJM27HUIXxhga1Y5lLVLKLkMnoauxnPPrSEx7ShF+Y/n2qrc6xBbhizDAqkK1zJuvG9nbg8ZA681
|
|
ly/Ei052RO3uKAsZlx8QGd8xxvt9Aa1NH8dK7AXMcip64zigdkdrZX8F7EJLdwwNXMkrz1qRMRly
|
|
CK4TxmpidWI49felPYSOMmi80NIoOV6qRzXYeA5SskYPfirpfEjGr8LPWVHyD6U4CvQPL3ZItOYc
|
|
UDOoNFeed4Uhpks4H4iE/Z5MeleMeGULeLgjds10S+BGdL+Jc9OSBU2Huc5Nc74yvUtrcDBrJnZF
|
|
63PJdXvLy/lKWw46bvQVz82jXhkLO5Y+9ZlsYthcRnbIjY9R3q3awTRkEM3WmJI6C0ea3dGRsr1x
|
|
XY6TqW9FLHnjrUs0izpLK5DDjofSta3ckH09KRUkZuuTvFGdvPauE1Y3U6Mqbssf/rUxHPTaJPK2
|
|
ZmJPbBqzY6DCZh5xJC9s9aBJHU6dpemJjfEmfetJtI0+VPkUr/unFOxdiextHs33W07YHQHk11mk
|
|
Xb3KbZ1xIvcd6LEyWho4Nct41sTPYb16ipexCPPZN+wYGCvH1rrPAEJmvkPoc1VL4kZVvgZ6yFwK
|
|
cBXoHkkqinFaVyzo80GuE7WJRQSziPiGdthK5HQV4x4J/wBI8WPIewNdEvgRNL42emO/yj1UHNef
|
|
eNpRczbC+I17DvWT2OqJxc0sMK4TCisy41q0hfEkqj8aixdwTXNOlwvmqD9anS9tXH7uVG+hosO4
|
|
/wC0oOhrR0+6G4YNIEzsNEuCxAPNdjZruA4xxUmjINSjURksOlcbqFykbnjFA1sYGoassaknCqO5
|
|
rl7rxhGm7yBnBxuJq0rkSlYpw+NLlsfd5P8AerVsvHEqSBHwPVgcgVpyMyVXU3rXxcHYETAk+hru
|
|
/DWti6ZSTyOKzZqndHaxvvUGq2rQ+dYyqR24qWI8dvbr7LqDxyDAzXpvw6FvIxePGSM06Xxoyr/A
|
|
zviKFHNegeX1J41zUhXioGbuaSuM6wpCaBHG/EcA6HN/exxXjXw2jL67cv8A3Qa6H8CFR+NnoWpO
|
|
I4XI44rxLxrqjQzSEsQM1gdSPM9U1uR1YbmWIdXHf2rmpIb67YS28UrRlsLI3c/jW0VZGUpO5pW1
|
|
jfLNOjahawzwReYI5cjzMkDavHJ5/SrVv9uhtPtVxCPLBwzxnlT9KGghLU3tKvvPjHzbl7EGuisJ
|
|
GRxWLOg7nRXJEbDjmvSNK+aFSfSoZr0KutRkphc4NcRrdkVjL9aVio7Hk3iqS8ubhrWzUlsZY9kG
|
|
cZNc5D4aee5MclzJIFTzHAO0MfatqSOWu7bFS1srDUZEis0vIZoUxPvfcC+4/dx2xjr712XiTwXb
|
|
WmlQ6hol3cRhoFd4rlg3zY5wR0GelavQwjq7GD4etdVvSnk2wAB+9v8A8mvcfA2kXiRo0/UdcDis
|
|
ZnTTulqeoWqbUAJqWUb42X1FZlnjfjSwlGrr5S/eNdD4RkvLAAQ4yRyaUZcruVKl7TQ9I0G+mnzH
|
|
ckFwM8VuIK7ac3KF2eXiKapz5UWYxipNtMyNejNch0jSar3cjR27uoyQCRVRWom9DxTx54gu5fMi
|
|
lbKdMVjfCZPNlv5v9rFbVHpYqjGzbOn8SzFI9o715L4u0r7arYzk+lYdTqSujy7U/C0u4vHk+WwO
|
|
xuh9q3J9dgvbdVukMV1EwbDDgn04rZMwlHoZ+orZ6hfQ3RWVnQYCgZAq+8U0ln5NtBsV2yxYcfgK
|
|
JtW0CnB31LlroVwJ1nQLGDjeP7w+lb0dsFxjrWB0tHS6NuWPJ6A16ToUm63T3Gallr4S7cxiTjrX
|
|
PaxaF7dlVeSMUhxZ5jd+H7qCa4eF3DSE5x3zXN3Wk6jbyeaiFWUY6ZyPStYS5SalPmVipFbX0E4c
|
|
W0alvmPHJrag0rVvEE6LdljGpG2NRtQD+tW5XMI0uU9M8NeFo9PiQhecDIIrtrOMIoG3H4VlJm9t
|
|
C6CB06VPGM1IHLeItGS6uw+ORT7e3jsbQvj7gzUNam0JaWE+HN7NqOqX80n3FO1RXo8YzXdS+BHk
|
|
4z+KyzGPapcU2YIv7qQtiuaxvcaWqG4O6FwfSrS1JbPnrxoxkv7qIfejcitj4V2f2exumI+8+aKn
|
|
xHTT+G5d8Txlm4rjLxMsQwzWT3OiK0Mm6sEkVsAcjFc1d+FEmlGwEDPQVopaEuOpr6f4ZWNAu3tW
|
|
vHpAj5ZQcUFIWaDjGMVUMQ3cVDBmvbhY7QAV2nh+T/R1yeKhlrY31+b61FcQK6nIoJMi401WblRi
|
|
qr6PCw5UYq9y+YgOgWzNkRrx3xWjp+nx2v3FQcelAbmko9anQ4GBUNisPHWr1qMrQhS2K11HvmYV
|
|
hamcxSRZ5xRIqluS/DKAQQXZxyXrvo2FdlL4EeZjH+/ZbjNSZpswLNBrE1Gt7VE4ODVIlnh/j61F
|
|
j4lmeTGyUbq6LwdEqWbeX0YbhSqfEddP4Bddj4JIrhL5d8h7VjI6oLQqKNzelWre3yc4/ClFjaL6
|
|
wqBxxUUxwCKu5BmXRA6c+9ZjP83FSBoQuPs4BrsNBlUW659KmRrDY6G1lyQtW3Hy0lqQ1qVJnAbm
|
|
oy3b9KYJCqRj3o4zRctIlhjLHmpSuOBRbQOpLGpPFaES7UqkZzKN1KsEc87/AHUUmvPLTVGv72aQ
|
|
k7WJwKmRrQ3ud74Ltilgz4++2a6iNDXdS0gjyMU71my7GpqTbxSbMki3SViajTTHqkSeR/GeyZmg
|
|
nQHkEE1S+F+oPPavBL96I4/Cia1udVF+4dVrkW+Fq8+v4tjMDWUkdVJ6WM0cNV+F+MVmjUcZgqnP
|
|
1qpNNnkcVRLiZtxIS1UzzIF7mghlxUZpVQdq6nTVdAoAOKzkbQWhvwM6gMM1twOJYx3NOJE11Kt1
|
|
H1/pVVlwBkk+9NocXoOQ45FPj+fkUJFF2NSB700v/hTEty5ZpkjvVyUgcCq6GM9zC14/8Se6GcZQ
|
|
1574Xs5WkI2HBPHFQ1dm1KSSZ7Rotn9l0+KPHIHNacae1dy0Vjxaj5ptlhVp+2s2CJ9ppCKzuWNx
|
|
zSFc1SYrHNeNdIGpaYw25ZeRXmvheyk0jVpEdcLJ0q3ZxNKTa0O3vQHg/DNcHrsJDmsmjspnNzNt
|
|
fFIJ24GazOhC+azDmgZIOOKBsp3J2qSaZodubq58yQ4QAnmhGT3NO18pb7BORmu205LfYpyKVkWp
|
|
Oxr5gKYWoIZWgfGfloFq1qTPLubnGO1RPtxg4P0oBAkY/hBz6VNDDkZ6AU0W2WSdqkdKr9ZOaGSj
|
|
VtcLHmnOcgmmYvcz7mBLy3MbdD1q9ouiRK6bUAVeelOC1InPlidSsWMDFOCEdq3uefykqrinYqGy
|
|
rFvApMVka2DAowKAsMkRXQqwyDXn/iWyitNQ3qPl6itIvRoF8RXinW4tQ6HI6GuW8SIVBPalc6qe
|
|
5x9x97r3qruwTjrWZ0ksZ9TUmcDNAmZ9/wAoao63rR0+w22MLPtAzt6mghmfofiB76LdJBJBIp5D
|
|
d/oa7bSdWLIPnpDi9TM8TeKdas51XTbIyxd3J/pXS+E/EFxqNoFu7do5OmD60maHWrnZyDRkn/69
|
|
MlEyOR0xntVoNx+FUgYjPxg4FLCuWDZyKQr2RoRnP0qO+nEFpJITgAUzLqZnhu6+0rknOTXpOmwJ
|
|
Fbrt5yMmnHYyr6Oxb2ijaKLnPYMClwKQWK3n0hn+lachHOJ9pNNN0apQFzsY10a4v4hXQh0xpieQ
|
|
MA1XLZNjhK80cT8OdV+3Wl3A7ZZJCw+hrR1qLcjZ/CsbnfHRnFXseHJArOYYbrUs1uPhYbuatqFP
|
|
ByfSkMq3UIINYkto+87Tx6GkSxfsDbflGD7CtTw/pk4nzITtPIFMFudsukh4Rxz71paTpKwP5jcn
|
|
0qTRy0NORMDgVCqewoJTJgAoxjntTiTu7fWmFxAcnn1q3EPl+X8KZMi4gKqB1Peob/Tv7Us5bfeU
|
|
yOoq4R5nYxqT5I8xieH9J1DTbvyJELRg8ODwa9Ms5mSFV9BWiptbnNVrKdmif7Q1KLg96XIZc5Is
|
|
pNL5pqeUrmMtZs0jzV08phchaY00zH1p2ZNxjS1g+LdJOt6U9ssmxjyGp2urDjLlaZzng/wUPDqz
|
|
TSTmWeTrjpVjVk3Rvjr2rnqQ5dDvo1XUd2cTqSNk9OKxXGCeKxZ1DAxHTr2q5C/y8GokUhsz54qu
|
|
uCxzSQjQ0+FZblR2ro4bZYiMVQ0dBb7Qi5x0qzuG5QOh71LYErDufpSeWrHnimIXbjkUjLkH1Hem
|
|
gGxryc+tXI19KYmWegq9YLiLJ7mtqS945cS7QsWehqxA9dEjz4krPSxyZqbFFhGxUm6smjRM55Lk
|
|
HvSvNxXTY57kLT+9MNwKdhXGm5FIbkU7Bca1wMEVhaiuQcVhXWiZ14R6tHGanGBI2OtYkqEHjgVy
|
|
s9ErEeo6UBsHipKEZs5qpPdRxcbhx70NCSuybTNWihc5brW9Fq6vjMnFSdEIdDRi8RRKygZbHFbu
|
|
m6nb3RA3gMegNJhOm0jbXGOoxTuCc1Rz3FyoGKawz9KaAVcZqeMgCmIkB4FaUTbYwB6V00Fuzixb
|
|
0SFMuDU8Mlbs4UPeXHeiOXkUrDuXYnyKk3cVk0ap6HMxxketSMhrcwRC0dMMZFMQ3yzSeVQAeUaz
|
|
9Vj8uPd271nVV4m+GdpnHX67pCeKyLtBtNcR6xlk9RVeWTb3qRnO6trgttyIfm71z7ai8j7/AJmN
|
|
DNqUVa5Yi1AnjynHuBV+11YJhWWXcP8AZNSzqgmaEerSsf3NtIQP4mGKtRavdRgMIpVI9KjU0a7n
|
|
R6T43uYQI7qN2Tpkqciu503VVuQGAYZHQjFVc4alPlZrpKGAznpTwxOc9+lWjIlUACnM4XApiLNk
|
|
nmvnsK0NvpXZRVonmYqV52GsmanhXitTmFkSiJTSAvwrxUxXIrJ7miOfjf1pzNWxkRlqYWpgJupu
|
|
6gQbuahvIxPA6eo4pNXVioS5WmefakGhndH4INZs5DJXA10PaTurmLO21uKpSZqGMoXGnRzBiyjd
|
|
9Kx5rcQS428fSkjanLoaOliHGZFB56VswW+mtPufcBsGOAfmxz+tFkd8HpoaUx09FAtFY8DO71qb
|
|
Sms/Nb7RbecG6AEjFLS5c78t+p0djpVs9wsyQiJAdyr1rW+zqjErzSe559Sbk9S3C+MA1bjbgE1S
|
|
MSXzMVG0vNUI2tPKrAuCMnrVzNd0PhR49W/O2xrHmp4TxVMzQshpIzzQBehqesnuaI5VGzT2bitz
|
|
FEbNTC1ADS1JupgG6l3UAc14s04yR/aYRll+8BXCtLncDXFWjys9TCz5oW7GddH5qqNzWDOgQnC8
|
|
VSuo1kHzAGkPYopEY2+RWxV23Vzj5G/Kg3jWaNazhZuqNXS6TaKhB2c0jR1nJWOlhOxRxU4YkCgx
|
|
Y0OQatQyDbyaaFYe8uF4NY3iC9ltbVGj43NTIL3h7WzMihjzXVQXYYDdW9Cf2WcOJpfaRZ3g9KsQ
|
|
mupnCLIabGeaAL0LcVY3cVmzRHIxtUhetzEjZqjLUAIWpN1ArhupwagAfDKQ3Q1594v0c2bm6tx+
|
|
5Y8j+6ayrR5onThp8s7dzkZjuqAAmuBnqC7c0iwgtzSA0rWzjfGRW3ZadDu4AoNYo2rfS4v7orSh
|
|
05UA2r0pDbsTm29KRottBNyJ0wpJ9KhD7f6U0ikNWffIFBz60zVUW52ow4UcUN6EPcx44WsbgOmd
|
|
ua7TT5Bd24KHnFKnLlZFSN4koluLdueRWvp14swweG9DXoxldHlTjYtzGoo25qzEvwtUxas2jRPQ
|
|
5CNqkLVsYoYzUzdQA3dSFqBBmnqaBhuqhriCXTpVIzxUz+Fl03aSPI9QTypW2/dz0qKNw3SvOPZR
|
|
Mqin8VLKRcs3O4Cuk0w/MDjt1NBtHY6O2IIHY1pxgFaETIRwMkjtVSUEk4570MlFW5bap6dKzWm8
|
|
1tqH8aY+hp2FvGoGayNevVt7/ap4xzUvYjqTLtvLPcvJxSaVcyWsxTnFZlnT2t15xHmCtOBYwQy4
|
|
B9q7cPO+jPPxFO2qLEj5HWo42+aus4HpoX4W4FTF+KlotbHII9SFuK0MUNZqiLUDE3UbqBBupwag
|
|
Bc1DefPbyD/ZND2KjujyPWlKzuPesRZjHJXms9lMuw3StjnmphKDSLTJ7OfE3JrpbO4GQc9qlnRA
|
|
3LO82k5NbFvdADkjBoCSHyXIIIzgVQvdRigT7wzjgUzO1jHknlvG7qnp61etYFQDIpCZoqVijzXn
|
|
3iC8EmsOuaCGb/heR/s0ijkVv6fbxy3QMg5xmsnuX0Ldzut3+UYTPWk+2GJSe+M1pFtamcldalmx
|
|
1eO4XaThhWnC+TXqR2PHqL3maUJ4qRjxSEjj42qXdxVmaGs1MJoATfSbqBAG5p6mgAzTJTmNvpQU
|
|
tzzHXY83D/U1zF5FhjgV5r3Pa6FMsV5HWnLe7RhqBRdmTwagN2d2K2rPU1C5LAnPrUs6Iysbdrq6
|
|
f3gK0BrUKj/WClY05iM6xLOcQAj3NT29uznfKSzHuadzNu7NSBFjHNSm5VO9IRnajqoWMhTzXFtA
|
|
bvUfMduSeg702Qz0rS7FbTToQFwzjJqaGTFyfK5PQViyzUuFmuIdgGABya5u/vTaN5cnUHFUmLoZ
|
|
zyskwlgJweSK6zQdUEwVJeGr0aUrxPLxEfe0OrhPAqVjxWhznGRtUwatDK4jNxURbmkAm6jNABup
|
|
6tQAFqhupNtu59qUnZFwV5JHnWsHdIx96w5lz15rzT2uhRmt85xWbcxMnUGmZlB0bdxmrNvFIcfM
|
|
350mWjbs7YkDJY/jW5ZWW4jikWkdNp9mqYJFaJdEHHakUULu/VB1rLn1Ld/FgetMGYd/qWSQmSa0
|
|
/AemS32pfa7piLeLkg9z6UmQtz0W7uQ2cZx0A9BVzR7cAea6j2rPqX0L99KRat5A6Dk1wOoKZ52a
|
|
YfMORTYRLujiGWEq6/NWza2yKQVHNdOHerRy4laJo6TTnbbtb8KuM3Fdh5z3OJjbmpt3FaMxAtUZ
|
|
agBN1GaQBzTwaAAms3VbjERUGsa07RsdeFpuUuY4jUjljWTKK4j02RE4IpJYFk6imQkVl0xWarsO
|
|
mAEcUi0bNnZBR0rWtoguMCkUi21wI161mXuocEKaYXMS4u+pY/hVCSWSY4HT0pEmlouiSahdpEBl
|
|
mOceleiwWcNjClvHgJH97Hc1EmVFFi3Czy7mwIl/WtJbjP7uLgd/apQ2VNVvtsBhiPzdK5S4nAuR
|
|
nqOCaTGi9pcytPlU+XpmumtWII44rah8ZjiNIXRuWeNvvViQ/LXpJWPJbu7nCRvVkNxVsxBmqJmo
|
|
EPiXca0YLMuOlJsuKuPlsSi5IrNuG8s4HWs5VEkbwoOTKsk+FJY4rC1K53k1xTk5O7PSpwVNWRzt
|
|
4cms+WpKICtSLTETQj5q0YeBSGiys23pUguGxQMq3E59ayrm4x3yaAKiRtO2WPHcmhruKFxFajzZ
|
|
ScA44qRHoXhuMaLpxaUg6hcDLMf4F9KlhuDeXGASIl+8azZslYma68y48m1+7nFW5rtbRNhb5z1p
|
|
iMKbUg0zuW4A4rPgb7VdKXOMmpA7HRbMS7nUYiUda0lkQOBngVrS+JGdbWLRt2bAx5BqeQ/LXpnj
|
|
PQ4GJ+ashuK0MhWaoWcA0AaOmASMK7jRNPWYBmHyiuepO2x10qfcv6vYxCzYqoGK4HVYVTJrmb5l
|
|
c6oaM5TUJ8EgGsG4kLNUHT0M64OaqMMikSRsuKbnFMRLG3zVehOaGNE445NNlnVFpDMu6uie9Vo1
|
|
8z5mOAOST2pDK91cNN+5tsrH3PrW54a06KxT7fdrlh/q1Pc+tJ6IUdZGvHPLezMcnBOWbsPap5r3
|
|
ylFtbdT1xUWNWzU0/Zbwlgfmx8zGsHWtRHmMqE59aAMyNifvHPc1f0gtPdqkY5JosJHeNci2tktY
|
|
euPnNY+oXWZEVJNrZ9aun8SIq/CzodHuriIokhDIR1ronbKZr0o6o8ipoz//2Q==`,b1=`
|
|
/9j/4AAQSkZJRgABAQAAAQABAAD/2wBDAAsICAoIBwsKCQoNDAsNERwSEQ8PESIZGhQcKSQrKigk
|
|
JyctMkA3LTA9MCcnOEw5PUNFSElIKzZPVU5GVEBHSEX/2wBDAQwNDREPESESEiFFLicuRUVFRUVF
|
|
RUVFRUVFRUVFRUVFRUVFRUVFRUVFRUVFRUVFRUVFRUVFRUVFRUVFRUVFRUX/wAARCASwBLADASIA
|
|
AhEBAxEB/8QAGwABAAIDAQEAAAAAAAAAAAAAAAEDAgQFBgf/xABDEAEAAgECBAMECQIDBgUFAQAA
|
|
AQIDBBEFEiExE0FRBiJhcRQjMkJSgZGhsWLBJDNyFSVTY3OSNEPR4fAHFjWCokT/xAAYAQEAAwEA
|
|
AAAAAAAAAAAAAAAAAQIDBP/EACARAQEBAQADAQEBAQEBAAAAAAABAhEDITFBEjJRIhP/2gAMAwEA
|
|
AhEDEQA/APqYAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAKNTq8OkxzfNkisQC8eb1XtRNbzXT4q7eU2nu0MntRq/D8StMccvW29ZmdvgjsTyvZjxOLj
|
|
+s8WLxn8TFPXs6Oj9oct7c14rkxz22nrB2I49KOdTjelmszfmpMeUxv/AA28OqwZ4icWWtt/SUi4
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAmdo3nsPNe0Pt
|
|
Fh09Z0+DNWL7+9O/7A3eJcZppsV5raI27esvH6jX5ddM25p79Ilo59VbUZOe2Tm/PeGvfPfT2iKR
|
|
PLv1+DO678XmW/a97U6TtOyzTbTF538/T9WjTNecm9a7126tqk3rSYxY5ta1plRZqZNXGjyZcPXl
|
|
mZmsx+qjBrsuO16xM7eXRt04JrdTltk5OWJnfaWf0a2lty5MdZnfzSn+WOHiOutFpjHa9e8bQ2fp
|
|
+alYy462pk7zXbuxjPesbRS0f6ZZV1ET1tErzXFLHo+A+1ddZf6NrI8PJHa1vN6iJi0bxMTHwfOa
|
|
zhzd61v1846utwniM6DUdb3nBaNrVmd9vjC/ZVePYirBqMWppz4rxaPgtEAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAItaK1m09ojcHnvarjM8P0vh49+a/eY8ng9D
|
|
h1fGM1rxjtGPfvbzdbjuTJxHX48cTPNltM/KsS9Dw7S49Jp6UpHaGe2vjz1y9J7LYK13vHWe7bj2
|
|
ex1tvM80ekuxW3RnW3Vm6P5jRx8H0+OYmMcb+bapo8GKPdpC6bQwtdHU8JpWkdJ/JweL6e23iU67
|
|
d4dubSqyVi9Zi0bwIs68XGp36TtEq7ZJmZmevzdbifCKWtbJinkt6eTgZPFw32t+sRurbWVzxs1y
|
|
Rv6T8V1NZNPtfq0seTm+Kevr+SZuxXjvaPiV8N4viycto9HseG6+uu08W6Rkj7UPmFck1tE1nlmP
|
|
Ld3eA8V8HVVi1pjq6Ma/pnqce/ERMTETHaUrKgAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAADW19+TQ5p/p2bLS4v04Zmt5VjeQeJ4bjnLqsupv+Ka1+ERLv4reTmcNxcuC
|
|
vy3l0qdI2hlr66sT02ot0ZV7qqrInruzrVZLGSZ37JjqgYTG0K5lbaFVhDT1Ub456RPweY4hixWi
|
|
eSdpjvD1eWejz3FNHWYtkpvFo9EIseb3tS3SerOms22rfpPqZKzvvHSYUz70TExG6Gdbs2rljeJ/
|
|
Mx5L0vEzPaelnOi98c9J2bFNTFpit47+a+PVUvx9T9nOIfT+GV5p3yY/ds67wvsXqpxau+G09Lx+
|
|
r3TqrEAQAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAADV4ljnLw3U0jvO
|
|
O0fs2lWqyUw6XLkyfYrWZkHldBEV09eveG3Fq1mI3jd4vPrOIaid8G9MP3Y38k6fNrt/rMk9Ou8s
|
|
tfXXn49rGWInuy8SO/k5Gl1E3rG/fzbOe94wTy99mbRvTrMOOvNfJWsesywniukrG/jU6fF43WYN
|
|
TmtEeJtEQ06aSmK2+bNtEd+qfSO17unF9Hmvy1y13XWyVmN4tExLxVK8PmNq5NrT58zawam+m/yc
|
|
0Xj8NpRYSvQZ7xEOdqI3rPozxayNRXe0ct/ON03jmrKB5nV4q1yTO20Obmv4c+cx8HoeI6WZpNoj
|
|
q83niYmYscU0r8aJ6T1n49zeJ+Meqm1drb9J+Kd5p136StGVem9l9TbHxLDFp7W7+sS+q1nesT6w
|
|
+PcAzVjiGHftzQ+v4f8AJpv6On8jH9ZgIAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAABp8VrW/C9TW0ztOO3b5Nxp8VmI4bn37TWYB8f1HFtTfUfR9FWJmsdZ9I7MtJxDX5s
|
|
d8ta1y0xzteaR2277rcuhycP12SceLxMeWNpjttHwlu8I0mfQ1y+D7k5YmJmY36T36Ka43z/AF1t
|
|
cI1ds+qxVj7/AEej19PCw9HJ4NoK4OIU5Y35YmZdzVTGebVZabx5jJS+Tmns81rNLm1Wrzc9rVw4
|
|
Yibbem72mXTTS0w0M3BvEta1bWrM95ie5EanY87wXgNOL6XPfxraXLhra/W28bR/dzYzarBqJxRe
|
|
bzE7Rt5vWU9n8mPHOGmS0Ypnea1naJb+k9ncNLR7u2y/WcxXO4TOoyUrN6zD0FaW5Y3hu49FiwUi
|
|
KxCvLMR0hlW0jn6ukWw3iXjOJzbDlneOj3GaN6zDzfFOH+LE7SRGo83XNSZ2lbG2/WfdlvaT2cy6
|
|
rNFInlrv1mfJ37cK4PwTTxOoidRm2+/2/KFuyMp47XB4LivXiunrH2b2iH2qn2K/J8x4fGDNxTSZ
|
|
9Nh8OviRvTyfT6xtWI+DeXs9MNZubypASqAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAOZx6/LoOWPvWiHTcf2hiZ0e8fc2mf1E5+vP/AEeuSd7RC2uKtI6QjHfeINTfwtPf
|
|
Jvty9WPfbt/lucP03gxfJf7d/wBoReYpm97zaNeLb4Ims9Nt94auDjem1Wo5PFi1onylS+1o7l8V
|
|
bxvtupjDMdNkYtXS1+Stt+m63xImEJ4xjHER2ZxMUjeUTO3VRmydBbjLJqPi08mbeVOXJPq1sl5Q
|
|
Vbkz9+rRy35rxHqzmZlVEe/Ez5LRlW5iyfR6zffaIjq1OSNZps2a21rZInafSPJhxGMl9LStLRWM
|
|
lorM/A4dkrWbYfLZC2W/7K6eubX6b4RzT+W76K8b7G6X62cu3Sten59nsm3j+OXz3/0ANGIAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA0OIYfpOHPijvNNo+fdvtXJO18k/
|
|
/OwPFYbz2ls3jx8VqW6xMdWPEdP9D4lkx/dt79flLLHbkxTPwY6nt2512ORTRzE2x4/dpE7cvkme
|
|
E4IrW3hRMxO8THRtU1FKWtvtvK2upx22rzRCtXkqzh2jtF7ZbT122b01ndnpuWuP3Z3+Ky20qDVv
|
|
fauzVy3mejZzNK8dVjqi87KLRLYtXruqvXzkQp7Qoid88R6rcl+WGlW0/Sa22mfhCZOq2x082ix6
|
|
jkm822pO8VrPdr4dNObVeDo8XW3uzMbzK+mvxT7szE27cvnu9j7PcNjSaXx8mOIzZevbrEeic5tN
|
|
+SZnpt8J4fHD9HXHO3PPW0x/DeBtJxx29vaAJQAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAKNRim9Z5e89Nl4DzXtVh5babURHrSf7f3ec1+qnDorWrvvt5Pccb0n0zhmWk
|
|
Rvevv1+cPE2rGTFNZU26PFfxwa5dVkjelI2772nZnX6bbrEUq3o0d678u8wmuDL2ittvVjXdneeK
|
|
cGv4jpJ6U56+kS7+j118+GLXpakzHaWlp9NNY3tv+bbiYiNoQy1y30uyZJlrWmZnuym6q1iIJnop
|
|
yW2Te8bdWnnypQqzZOadokiIpSZntWN5lrxki19vNRxrUeBwnNNd+fJEY6/OejXLn3Xe/wDp9wyn
|
|
E8uo4lqqxblv7lJ26T6vpD5X7G8QycKzeBMbzMRM1/FH/wA/h9QwZ6ajDXLitvWzRgsAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAeL45w+dDrZvWv1OWd4+E+j2jX
|
|
12jx67TWw5Y6T2nzifU+rZ1y9eHwzDYxxEy18+DJodXfT5o96vafWPVbjyxDn1OOzHudbM0rt2UW
|
|
iI69mVtRXZq5tREb9VUoy2iIlRbJ0UX1VZ6btTLrI7V6yk62M2oisT1c7JmtkttVMUyZp6x0beDS
|
|
RWOvdKijDimvWd3G9pNRMfRcNfvZOb9Hpb0itJeP47k/3hgjaZnbaP1XxWW3T0movbNS0W645nbf
|
|
0nrMPpXs3xamoxdJiLbe/X1n8Uf3fKsOTw4jbaXo+EarJhtGTHMxeJ6xH7Sti9Zaj6x3HM4NxXFx
|
|
DS1mtoi8dJrv2l011QAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AGjxLhODieOIye7kr9m8d4eM4to9RwjPXFa0ZIvG9bR0fQXmPbDFvTTZPOJmEWS/V8bs9R43NxLL
|
|
G8eFbePg1bajU5/s0l1ceKLx1hbjwRE9mOpx0y2uRTSZsm3PMw2aaKtIjo6kYo9EXpET0hVLXxYK
|
|
xC6MZvyx1lFs0RHfaPiCnU12pLyHGNDbUajBekWma2npWN3p8+opa20e9LSyZLxExTlpM+vdOdcZ
|
|
a9tPS8MyUvFrzWlI6727u1pYxYrbVmb7x+TQx6au3Nqcl7/0rcmW9axGnwZJj1novmxnZXV0fFp4
|
|
ZxLBPgTGK8xzXr5fOH0bFlpmxVyY7Rato3iYfNuG2x56Wrqa8s2jz+7Lu8O12bS6jkwzN6THNNI6
|
|
tvrN68Y4rxlx1vHa0bskAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAA4XtTTm0OKfTJ/aXdcL2pyRGjwU362yb7fkJz9eTxxyZJjyltRXzUZK7TFtl9Lbwy06YzrHwa+
|
|
fJFd/wCVt8m0bQ0eS2qzcm+1K/an+zNZFL5M1pjFXeI72ky48eGnPkvNp27+TPU6nHpMfLXaIjpE
|
|
erk5dRMxOfN1mPeisfshW1ne1a1577Y6x5R3U0zze31FOWI6ze0byU098kRlzbxM9qrMlPDpyRMR
|
|
Md5Vt/Ihp5898mWZm1pjftE91uCt7fCI7dWeHDEW3t723l6rslqxWZnasR+SYhFbzhnfxJ2jyeq9
|
|
lcGXWZcmW0zWKxHLaI7794eJx5fpfEKabT8t8l5isddo3l9S4VjrwrRUwzSJt3tav3pdOL6Y6dXD
|
|
j8HFWm+/KsU4NRXPvtWazHquWVAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAa+fXYNP9u8b+kdZBsDkZOO135cWOZn4y5Wu4xqctbe9y19Kp4njt6vi+PDm8DFMWybbzPlV
|
|
5PiGtz67UxbNbeKTtWIjaIXYpnwuaftT5tXJT3vmi1pMsrU5qIrG1V1a+5DCa7b9GFbRr5J6Wnbt
|
|
Cu+Wmk0m8956z8ZWZNorbfzcbX5rZslazPux3hUt41NTntktObJ13+zX1bek01r4/HzVm0bxPXy/
|
|
+bNfDgjVa2uOY92kdfg6ufJOKvLXtttVVSqbcta2vM7zXtHpLQy5ZtMd+vWd+7Zy3mdJHXra3f0c
|
|
vUarw7zFY5rT2hH1Lavnrgx81p3U49Pk4nE5L35MO/StfNRXR5tXnrS8W67WvfyiPSPi7uLHFK1p
|
|
jrtSsbR5Lc4RzsXBaYreP4l45esRD2HD9fnw6evvWvO3Tfr0aGk0U55ra0TFInv6uzgrXFXlx0i0
|
|
77RPlC83Yj+JW7oddqr6vHzTTw9/f6dod+L1t9m0T8pcbFSmPHER3892W0zPuz+jSbVvidkcqmfP
|
|
Sel7bekrI4n4dZnPWIrHeYnZee2Wpy8dEaml4npNZblw5qzb8M9JbYgAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAABEzFYmZnaI7yCXL1XGa0jJXT0571nbee27DiXEprp8nhbxG20W8
|
|
5cbD0ikfnKO+urTPvjoZdXqctdsmTaPSvRpWmsdZ6yztfaGplvv3lWW1tyRlz1x0vkn7Vo5atTNe
|
|
Y0+1o79V2KsZsvX7Ne5mwxnyTNvsx2iGneM/rCdRSuOsTasTt5kRFtpjqmOH4t4nk7estiMNa97R
|
|
Hwhna0iuKTEdmGWa4672nZtRele1N59Zlq6vLOSsYorEc07qcW65euzRvtXvPZy52naZ7ujr6fXV
|
|
rWdukREK8+njHgmZmPc67bq6ivVWhxxgxZLztNrT1mZ/SP4VZs0zaOvfp84WUtNsXLvtv3699+rU
|
|
z7+Jtt5qURqMnPpctaR1rMSw4ZoK57eNk6xHaJRh97Ltt7lo5Z+L1HAPZvVauZ2nFTSzMTzeJEz8
|
|
to6xPfvsZntPZ9rXxabmxzefdrv0j1dXh/BcmstW1qxTHHasR3+b0GPhGl+kWmd64dNEVjf73T7X
|
|
y8vy+Ddx6O3iRakxTH5RXrMw1/lX+3Itw2MFIraN48qRHdZi0cUjmmPen9noox1iO0fNzdXEYrTt
|
|
stcmd9aX0bJ+HePmiKTitO8TMLZ1cVjrMfqpz6ys4pjfrPRWZ9rXXptUit6zO+23VyaRHEc05L1/
|
|
w9J9ys/en1ljqdVbwYw452tlnl3jyjzbmmiMeKtYjpEbLeTXPUU8ee/+qjJpsV5rbkrFqzE1tEbT
|
|
DpYNbW21Mnu29fKWna0KbqTdjXXjld0cvQ63ltGHNPSfs2n+HUbS9c2s2UASqAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAAAOVxPWe99HpP8ArmP4b+r1EabT3yT3iOkesvMVtN7za07zad5l
|
|
XV5GmM9vVfEstvDx0jtaVVMlq+UJ18b5cMRvPeSuK87bUt+i2Z3PtG7zXpjkzXt6R+TXyTMzvM7t
|
|
ydHqZ+zhv1+Cv/ZuqvPTHMfOYaTMil1a1K2vHSLTELq2v+KWzThGo84rH5rq8JzedqR+ZeI7WnOS
|
|
34pYTafWXR/2Pln/AMyrKOCWnvmiPyR6O1y9585lhWJvl557Q6eo4T4dYiMvW3b3UanhldHpJtGX
|
|
e09unmjsT7eb1l4trI2t0hsZfrdNO0bzy+nzU20/+NmkzO9esz+TZxWis9dttvPv+Tn21jjaW8zn
|
|
26bTG3mp1M/Wzv3t0jyWXiKZJmsTERaZhXXDbNl8WaztWenxZLstPp5pau8frDtVrNMM5cfTfpMf
|
|
3aunxxbes9d/R09Dp8ebJi09ptFr3jtt2WyrW9wy1Jx132mK+Xq9PotT0iIU19ntLtExa3T47T+q
|
|
6nBaYvsZstZ+cT/LeMnUi0TXffo1s2m8Ws2/OIMWk5Jib5L328rS2t94Sh5TV4ppklpW6PT6rh+P
|
|
NbebTHyas8E081mZy5P2W6OFhjxNTE/hr/LoRO0Kvo9dPqctKzMxEx1la5t3tdnjnMs4noievcrO
|
|
yZjeFF1OSnNV0OG62cn1GWffj7Mz5w05joovzY7xes7TE7w0xrjPeex6Ua+j1UarBFu1o6Wj0lsN
|
|
3JfQAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAACrU5o0+nvlt92P3BxuM6nxNRGCs+7Tv8
|
|
2hToxm1r3m9utrTvMsonqyt7XTmcja0u3O6FMfi5t/u0/lzdJM81p9O3zdvHTwsUR5+bfPqOfX1h
|
|
dqV+3O7bs1+T31oqmI3TEM4rvCdkDGIIhlFd2daboS0NXG2bD6bufxXU1vlmu/u4us/N0+L1tTSx
|
|
kr9qk7w89j1FNZMV3jxLzvaJ8mer+LSOZqK2xZotbvljfr/89U453rXt9lse081xZtNjx7TGKu0t
|
|
DHlrevSevaN5Y6+tJ8c7VRNMt63n3ub+6/R54rERMztDYy4a5omclYmfxKcenrjtHLvtPrCnVmdb
|
|
eFe3JXmjy6eS/DrMuLVYsta9Mdt++6qLxO+0dEc8UmInr18iUfReHcXrqccb9Z27Q61Lb13eJ9nc
|
|
1Z35rTvE9avY4bTkpG8xEfB05vYxqybc07R281naGMREdoT5JQqy9mply7Q3bV3iXG1eXw7TWSka
|
|
c258t7+tpT5/BjT7MfHqndz12Z+M4lMMKyziUJJiN1WSu9fku23RaOgKNJqbaTU1t9yelo+D0cTE
|
|
xEx1iXmM1Nt3W4PqvFweDaffx9vjDbGvxz+TP66QDRiAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAOJxzU73rp6z296zsZMkYsdr2naKxvLyObNOfNfJbvad1dXkaeOdpvsc2yuZVzfbfqybutwu
|
|
s5s8R92J3dvJb3tnO4HSMegtmt3nfZvYp8SZl0z45NfSK7onH1bNcfRFqnUKJr0Y7dVtq7prjEsK
|
|
0XVpEM6028mW20IHK41aPo3J6zs4ODhdcvPnvExFevNXpMOrxi/PlrTee7PLX6Pwa09uaNlKtHg9
|
|
dM3z5d7ReOu02nu0JzZMfblrv5R5uvrcdImZ26T1mYhxs1Os7RH93PZ7axuafNfLitvbaYU3yZYt
|
|
PXs9NwHhui1HBa5LVicsb81onrEuVqNNSuS8Y67dZ6xPZa59Il9uX41vEitImZme3q2Kxbxora0T
|
|
Md/ROSa4Ztkj7c9OafL5LuGYubmyX3iu/TfbdSfVnpvZLT/XZK233+Mbbva1xRXyiPk8pwbH4N6T
|
|
adq5a71n0tD1WDL4tPe6Xr0tDpz8YVnJHWEXYxbqlBedoef4tW0XraO09HdyztSZcbUz43C+ee9b
|
|
SVMaeOfqq7+jGckQ1Yz7+7v2RN/WXPXZPjci2+2yyJaVMuy+uSJlA2d+pNoVRbeDcSxyTE+TDDlt
|
|
pdRXLTynrHrDOyiyZeVFnY9TjvXJjres71tG8MnJ4Nqt4tp7T1jrV1nRL1x2cvABKAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAHJ49qfD09cNZ97JPX5PPw2uI6j6Vrsl/ux7tfk1mWr7dOM8iLdm
|
|
vfebREefRsWldw7SxqNbWbR7lPesrn3Vteo7dYjDpMGCvfbeXQ0uLlxRLRxROfUc34p6fCHYrXlr
|
|
EejqrjY8uzCYW7MZjdVKqK9VlaxCYrsnYExBMRMJRPZA8/xPHtmpP9W2xx76vhWOInvt/C7ike7N
|
|
vwzE9kcapGfhlevTaFbFo8RqJ5vy8/RoW09ek0msxHfp3dzNoLzp4zUmZpMbT8HJyYJi20X2n0lh
|
|
ZY1li/RaidBF4w2mK3jrHaFGp1lN+tptPp5IjBkid5mIp16TKu0abBPv33vPlM7z+iPdFNcWXU5I
|
|
tkrNce/b1W5db1nTaf3ax9q0fxDW1ebNk2phty1mOu09VOm8W19orEz23j1TwfSeERFuEYMddptW
|
|
d43dvBn21eKJ75KbW+cf/JcTgMxXTb3nbljz+TpcPmc2uyZO1KRtVtGVdi0bx07qJnllsRO6rNTe
|
|
N4XVamsy8mnvPwc3R2jPwe8TPbdlxXNOPSZfhWWpwO85OFzv57qrODkzeHntSe8Sn6Rv0a3EZ218
|
|
8nXekfr1a0ZLVnqx19dWb6demXybOO7lYMvNMdW9S/VVLo0us7tPHdtUtEwJiZU3jq2Jhham8CVG
|
|
PNODNTJXvWd3qcWSubFXJWd4tG8PK3pPd1OB6veLaa89Y61/u2xfxh5c/rsgNHOAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAANLimq+i6O0xPv392rdeZ4rq/pOqnlnelOkIt5F8Z7Wj27I2I6sb25YY
|
|
V1ImY3dbQ08LRc23vZp2j5OJG+XJWle9p2h6HHtbJXFT7OOIpX+7TxT31j5rycdTh+Dpz+XaG/sw
|
|
w18PHWseULN2trBE9UcrJKBhFU7JAQi0dEomegNDUYovM7x3jb5tO1ZvpbaTLtzRExWfWPJ08kbT
|
|
Ex5NXWYYyV5omYtHWJieyeDzuizfRs19Jn6TM7Ru1uMcJxZqTkw+5f4ebqa7SV1MR4tdrx2vEfy1
|
|
axqsNOTLjnLXytVXi3Xj8+nmsxTLM16d5npPyUzpekTtSK+U7vS6vQ/SYmK1vWPS1HOn2dvvvvE/
|
|
tDO5XlcO+LbfHSd/W3o6/BdDOXPTnj3Kz38rS6Wm4FNrRyRzTH3p6RH/AKvR8L4dXSzE3jmtHn5I
|
|
mbfqLV+m4dbLSsZInHjr3iI6zLpYaxS01rHuxHRHiT9mv6s67Vj1aqL6326MrWiYa+/Q54BxPaGe
|
|
XRZpj8MquB4+Xg8zPnB7SX30to379GxpK1xcHiKz5IS8xr8PLPixH2bftLTy05o6dHYyVjLhy0t1
|
|
izjZa3pMVv3iO/qz1G2L+NbSajbNyW7xLsY8kTDz+fJXFqKZN4iZnafi6WHL0iYlStI7OO+7axW2
|
|
crFl7dW9jvE9ULN+J3ZbdFGOy+AYWpEqN7afNXLj+1Wd23KrJVMvCzseh0+auow1yU7WhY4fCdV4
|
|
OadPefcvPuz6S7jol649Tl4AJVAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAV581NPhtkvO0R+4NPi2
|
|
r8DB4dJ9+/7Q83Po2NTqLanNbLfvPaPSFDHV66sZ5ET0hRknyW2lTtMyouz0c8usx2n7s7vScKwx
|
|
zc1vu/y85p+maJh6Th+SOWeveXR4/wDLm8v+nX5mUWa9bbrInolmu5jdTNkxYFk2Isr3TuCzeGMz
|
|
+THdEyDDJO9Ja823rt2XWnya946pGvktDXta0ztWu/ybvLE9dkcoOf4GbJPWK1j49VmLh9JtE33v
|
|
Mevb9G7WsW8l1ccREISophiJ2jpDYpijbaOjOuOJ8ujOdqxsgVcsUjaETYvbaFFrgu5lVsm0yUtu
|
|
ryg43H5m+GIj1XcJzePoL4pnrWGtxmfchr8JvfHS1622if3QljzTTLes+qrNjrkiYtCzPMxnm095
|
|
YZJ6boS5teB49Tqscza97VtvWvlv8V/FOF34RrIxTM2xXjelp/eHoeA6XnzReY3ivX/0dfivDcfE
|
|
9HbDbaLx1pb0lOs+jO7K8Lis3cN+0NKcd9PmthzV5clJ2mF9J9GHHVL108dm1SznYr/Ft0tuhLb8
|
|
mNohFbMhLWy0mJ3rPXvDvcO1karBG8/WV6Wj+7kWrvDDBlvpdRGSnbzj1hpjX4z8mOx6UYYstc2O
|
|
uSk71tG7Ns5AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAACZ2jeXneJ62dVl5KT9VTt8Z9W9xbWclPo+O
|
|
fft9qfSHEU1pv48ftYST23ZTDC/p0YtlVuvVjMbM5+LCZjYGWGdrTPxiHY4ffaf3cjTxz1v6xMS6
|
|
Olty2iXVj/Dk8n+ndrkhnGRo1v8AFdW3RCrZ5uiYsqrboncSu508yjmZRYQt50TfowYTbYGVrKrT
|
|
uTZjvukQnYhMIGVY2ZxPVWyrHVCWzXpVXkt3TE7Va+W4K7X3jv1auTNy3jdba0RZpamfroQN7Hk3
|
|
6wr1GTaN2OOJiu6Mu98NvgDi8Wy74d/yZ8PiPAiO2zU4nb6qIn1bugjfFE/ASp1ke9u15mbbRDZ1
|
|
Mb823kx0Ontn1OOkedoJCvT8I03gaKsz9q/WW+isRWsVjtHRKyrhe0XCfpWL6Vgr9fjjrEfeh5fF
|
|
feH0V5Dj3DPoOo+k4a/U5J6xH3ZZ7z3228evytOk7NvFbo0cdols47bSybt7HbddHVqUs2aW3Qnq
|
|
xVeu8LILR3SlZw3V/R8nhXn6u0/pLuPMXjeHT4Zruf6jLPvR9mZ8/g1xrvpz+TH7HUAaMAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAABRq9VXSYJyW79qx6yvmdo3l5viGs+maqYrO+OnSvx+KLeLZz2te1rZL2v
|
|
ed7WneZYWnZl5K72YV1xEyxmeqJljzIEWlVkszvbZp5soN3h2SJz3pP3odCnuWmPRxuERfJrZmtZ
|
|
mtY96fR28kbX3dXj/wAuTyf6bmK+9YX1s0cNtm3Sd4LFY2K23W1s16StiUJW7bp22RW3RluBuruz
|
|
mWEgrmCGWyNkoExKE1QlPmsqRDKeyBjaejWy2W3ttDUyz1QKslvehVqKTNosyyTvELabXptIJpaP
|
|
B39Ia2mz+JGpr51jdZefDx2hzuHZObNq58poJaGtjxJ2+LoaKP8ADRPo5+T3skx5OhpOmC0fBNQ0
|
|
5yTbn+bt8A0u9raiY6RHLVwY62mI6zMvaaHBGn0mPHt1iN5+aYVsACBXqMFNTgviyxvW0bSsAeE1
|
|
mkvw7V2w5Ote9besJx2er4rw2nEdNNekZa9aW9JeQjnxZLYskTW9Z2mJY7zz26fHrrdpbZsY7NGt
|
|
mxjvso1b9NmUwpx33XRO4K7VUTE1nmrvEx1bVo2VWiJE/XY4frY1WPlt0y17x6/FuPM0m+HJGTHO
|
|
1qu9pNVXVYt46Xj7VfRtnXXL5MfzexsALsgAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAHM4jxOMFJphmJv529Dq
|
|
ZLfjDjPEIx450+K3v2+1MeUOHSOWFc3nJkmZnf4yujpVlqunOeFpV2nctLCZUXRM7MJtsWlRkv3Q
|
|
ky5NmpWt9RnrixVm17TtEQnJabXisRMzPSIew9n+CRoccajURvqLx5/chfOest642OGcIpoOG2w7
|
|
ROW9d72+LQvXevyejcPUU5M+SvpLeOataraw2a0dLbLqTtK1G3Es4lVWWUSoldFtmcXUbpidgXzK
|
|
GEW3TuCUSncnsDFMMLSms9EC6J6FpVzbZE5ALy0809ZbFr9GtfrEoFMzuuwz0Ueey3HbaBLDXe7i
|
|
tMOfwWnP9I+NZbuttvhs1uBRtXPb4SDm3iIvf57N7Dbl0VrS5+XrltEd+Z1Jx7cNms9N4TURRw3T
|
|
+PrcO3WszEvZOD7P6aYiMlvu16S7y1QAIAABxOPcLnUY/pWCv1tI96I+9DtgmXl68Biy7/NtUu3+
|
|
O8HnFa2s0tfd75KR5fFyMWTdhrPHVnX9R0cd21S3Rzsdm1iuqs256wrmGcT0RYSx5d047X02SMmO
|
|
esd49YRE9WcdSXhZ2O1p89NRji9J+cei1xMc3wXi+KZj1j1dTTaqmor06WjvWW+ddcu8XK8BZmAA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAMMmWmKu952UZ9XFZmuP3revlDTtzWnmvO8q3XGmfHb9ZanV3yxtWeWn7y4es
|
|
vPNtDqZJ6Ts5mppvdl/XXRMyfGvSNlu/RVvtOzLfoipLT1VTKbSpvfogRkvtDVyZOhkyvQcA4Dzz
|
|
XV6yvTvTHMfvK+c9U3rkW+zvA/D21urr789cdZ8vi9KDb45rejl8Rry6iJ/FV1HP4vXbBTJEfYt1
|
|
+UpiHM295bXsqrO9l8QkZ0lZEqqLeyBZHZLGvZkhIndADKJ3TMoqWQMZ6pjsxll2jsCLSrmU2lFY
|
|
36gieyu0LJk3jbsga0wdqzK20QpyztQGprL/AFMrOE05NLkt6qdVWZxNrSe5o9vWBLiUjnzXn0vL
|
|
q555dHt8HOwV928/1z/LpzXxbYccRvzTB+jucOwxh0dI22mY3ltIrHLWIjyjZKyoAAAAACJiJjaY
|
|
3iXleM8InR5J1GniZw2n3oj7s/8Ao9Wi9a3rNbRE1mNpifNFnVs65XhcWTdt47bnFuF24dm8TFEz
|
|
p7T0/pn0a+HJux1OOrOux08d1ndqY7tillVkzExLOk7yd4YxGwluViJhE45raL0na0dtlWO0+bZr
|
|
1TKi+2zptZGTamT3b/tLacvJjiY3XaTWdYxZZ6/dtPm1zrv1z78fPcbwC7EAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAABhkyV
|
|
xUm152iAZWtFazNp2iGhm1Vss8uP3aevnKrNntqLdelI7VRHRnrX/HRjx/tZREVjZXeybW6KbWZt
|
|
pCZ6S08tN7Nmbb7zCrJtyoS5145bSx5mWafelr3tsKmS/o08uXyhlly7RPV2+AcBnPNdZrK+53pS
|
|
fP4ytnPVda4y4BwHxOXV6uvu96Unz+MvVxG0bQRG0bR2G0nHLb2gCUDX12LxtFmpHeazt82wT1gH
|
|
mMN4tWs+rcr2aEV8DU5sM/cvO3yb+O0csLUTSdrLphRE8tlkZI7Atr2ZMazDJVKTYSCawi7Ksq7z
|
|
1QERvLK3ZGPrKbyCrbdnMcsbeaa18/RhvvM7oGEwTG0JmYYTIML22a2e28xELM19oURPNO4lOem+
|
|
n3ZY5+prVnMc2GYU4/L4A0a15cNf6rz/AC6fC6+NxCPOuOu/5tHJTbHj+F5/l1+BYumXJMd9o3/d
|
|
MRXYASgAAAAAAABhlxUz4rY8lYtS0bTEvH8R4ffhmo6bzhtPu29Pg9mq1Gnx6rDbFmrzVsizq2df
|
|
zXkMWTeIbNL7tbXaHLwzUctvexWn3bmPL8WFnHVL326VZ91MfFVjvvVlz79kLrcf2m7j7bNHH3bl
|
|
J2SirLQoy4t1++7G0dBC/RanxI8PJPv18/WG241+alovSdrV6w6mDNGfFF4/OPSW2b1zeTPL1aAs
|
|
zAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAVZ9RXBTe3WZ7R6iZOpzZq4ac1p+UermZMl89+a/byj0Ra9815ted59PQ32hlrXXRjH
|
|
DpCLX6ML5NlNsm/ZRqstfdXzbsZt06sLZNvNB1Za8RDWyZdo7q8udq5Mu/mIMt4md2lmy7JzZuWJ
|
|
dHgfBL8RvGo1MTXTxPSPx/8AstJ1XWpIs4BwSdbeNVqq/URPu0n73/s9hEREbRG0QUpWlYrWIisR
|
|
tER5JbSccur2gCUAAAAPM8Sry8Uyz67fwuxbzVPGsE49XGbvF42V4M0TEL33ERnktsxpk3sumK2j
|
|
admFdPFZ33VS2Mdui2J3UU6LYlFSsN2O5NkCyJ6K7T1TEsbAsxdpReerKkTFGMxvYEz0rsqtbbpC
|
|
b2VT1QEzuwtbaGUxspuJU3neWdKoiu8rq12gCI92YatLcublnzbEz1aOptyZqTuDHLfxN6R0+t5X
|
|
qdJhjBp6UiPLeXl9NSMnEKxHa1+bb8nrlvxUAAAAAAAAAAABTqtNj1eC2LLXeto/R43VabJw/VTh
|
|
ydY+7b1h7ho8V4dXiGlmvbJXrS3xRZ1fGv5rzeHN02bEW3cys3xZJx5ImtqztMS3MeTeGFjqlb2O
|
|
8btql3NpbZtYsnSBLeiWfdTjtutid+ghherHS5p0+f3vsX6T8Fkw181d4lMvEWdnHaGnw/UeNh5L
|
|
T7+PpPxbjdyWcvAAQAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAo1Oprgr63ntAmTqdRqK4K9etp7Q5d7Wy2m953lNrWyWm953mVd77R0
|
|
Za1104xxlN9lV8qnJl2a9s3xUXX2ybsJyRDWtl3YWydEC+2VRkzeW6q+T4tbJm+KRdfK1cmWZnlr
|
|
vNp7RC/R6HU8SycmCk7ed57Q9ZwvgOn4fEXtHi5/O9o7fJaZ6z1uRyOEezVstq6jiEbV71xevzer
|
|
rWtKxWsRFY6REeSRrJxz22gCUAAAAAANbX6aNVpL0npMRvWfSXlKamsRMVvXm+EvZXjmpaPWHzfL
|
|
oNRjzXicfWJ8phfPxFejx72x7xMzK+sXiNoiXlq+Pi6fWV/VfTNqfLJl/WTg9Pji8R70LqvMV1Gq
|
|
j/zcv6yz+lanzzZP1lWpelTET6S81Gp1P/Gyf90s412rjtnyfqql6asREdWM9+jz9eJ6yP8Az7uh
|
|
odZqMt458tpB1JvEViI3/RhzRt13/R1MNaziiZiJn5K9ZNceKZiIiQcu/WekT+iYrWI3lzdTrs+8
|
|
8uW0fJzcur1Np/zsn6g79phVaIeetqNR/wAXJ/3SwnUaj/i5P+6UD0ldonum161h5mNRqP8Ai5P1
|
|
lNtRqJjacuT9Qd22WN5aGeZyZd/KHJy59RHbLf8AVq31Gp/4uT9ZEvS8Lr/vSs2npzRtL1z53wK+
|
|
oza/HW2XJNd99pmX0Rb8VAAAAAAAAAAAAAAcHj/C5yV+l4I9+v24jzj1cLFk8nu5jeNpeW41wmdL
|
|
knU6ev1Vp96sfdn/ANFdTrXG+eq1q5F2LLtbZoY8m8d11bbSydErsYsm+zZrO/zcnBm226uhiyRK
|
|
EtrvCrJDOJTeu8A1MWX6Lqq5N/dnpb5O5ExMbx2cPNTeJb/DM/iYPDtPvY+nzhri/jDy5/W6AuwA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAa2p1UYo5adbz+xbxMlvqJ1OqjDHLXree0ejmzNrWm953tPmTPWbWneZ7yoy5YhjrXXTjH8s75N
|
|
mtkyxt0VZM2/m175N1V03yTKubMLXVXybeYLLX2VXy7eam+b0bOg4VquJW+rry4/O9uyZOq3UjVm
|
|
9r25axMzPaIdvhns1kzbZddM0p5Y47z8/R2+HcF03Doi1a8+Xzvbv+TotJnjDXkt+K8ODHp8cY8N
|
|
IpSO0RCwF2YAAAAAAAAACvUZYw6fJkntWN3k8dfHz2vLucdz8mkjFE9bz1+UOZosX1UzPm0nqI/W
|
|
MYo9FlcPNklfFGeH/NshLGun+Cz6PtHZtVZWlRLS+jxPkRpIn7rdoupHTdA5s6SI+7H6Mfo+32Y2
|
|
+To3neSIiZ7A0IjPXpXLePlMotGW3272t85datKzHZjbTVnsDj+FG/2Y/RlGP4R+jo20u7H6N1Ql
|
|
o+H8I/REY957R+jpfReiK6eOYHLtj2tttH6KrY/6Y/R2c+kjeJiFVtLG24hxpw7/AHY/RRkw9O37
|
|
O99Hrt1YX0tfOBLjcGp4XF8c+u8fs9c4dcVcGemSI61nd3IneN1orQAAAAAAAAAAAAABFqxes1tE
|
|
TE9JiUgPKcX4RbRXnNgiZwWnrH4XPi28PdXpW9JraImsxtMS8pxXhF9DecuGJtgmf+1TWW2N/la1
|
|
L7N7T5e3Vy6W3hsYcvLbqzbO9jvvCzvDR0+XeO7crO6FmGSvRThy/RtVXJ92elvk2rRvDUzU7pl4
|
|
izsd2J3jeBpcNz+Lg5LT7+Pp+Xk3W7js5eAAgAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAADs0NTrN96Yp6edkW8Wzm6+LNTq4pvTHO9vOfRoWtt
|
|
1mes95YWvs1s2fZldddOczLPLn2ju0MmebT3YZc2/mpm3qqllN1drsbZIhr3yzvtHf4AsvlYYseb
|
|
V5Yx4KTe0+UQ6nDvZ3UazbJqd8OKeu33peq0eh0+hxcmnxxWPOfOfm0mP+steT/ji8N9mKY9suum
|
|
L37+HHaPm9DSlaVitKxWsdohI0Y22gAgAAAAAAAAAABXnyRhw3yT92Nwef4xm8bVzET0rPJH5d12
|
|
CvLhho3rN9RWs9Z23n5y6O21YhrVYbdGOCfrrLPJRpv863zVS6FS09SvZj3lVZZRdPSqmnSWdrIE
|
|
ebOkK4ldTsgW1WKqd1oMZhEVZyRAImOjGI6rJ7IiATNd46qL02bHkiaxaoNGY2n4ImPgtyV2n0Vo
|
|
Gvlx7x2beiyTk08RPevSVUxux00+Fn2n7N+n5rRFb4AAAAAAAAAAAAAAACLVres1tETWekxKQHlu
|
|
L8InR2nPp43wz3j8P/s5dLveWrFqzW0bxPeJeV4xwmdFec+CJnDM9Y/CrY1xv8qvTZ+WYdbDk5oh
|
|
5zHk283U0eo3jaZZ2N5XYjrCnLSJhOK+8d1kxvCqzSwZvousrb7k9LfJ3nB1OLeJdLhufx9LEWn3
|
|
6e7LXN9Ofy5/W4AuxAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAETaKxMzO0Qi9646Ta07RDmZ9VbPbaOlI7Qi3i+c3TPUaqcu9adKfy0722ZXvFa9
|
|
XO1OrjrESxt66ZJmcjPUanlidmhkzTZVfLN5VWvsC2b7R3U3yqrZZtO1esz2h2+F+zWTUcuXXTNM
|
|
feKR3n5+iZLVbqRzNJo9TxHLyaekz62ntD1fDOA6fQbZL7Zc/wCKY6R8odLBgxabFGPDSKUjyiFj
|
|
SZkYa3aALKAAAAAAAAAAAAAADQ4pl2pTFH3p3n5Q33E12Tn1eSfKscsLZ+orS00eJqbW+Lfnu1tF
|
|
XaJnZsz3WpCfsyp00fWSvmPdVYOmSUDd8kR3InoQosy7JmUX7MdwZ17ro7KKT1XRPRAsrO0rYndr
|
|
79V1ZBaQiJ6JgCSIJASwrO07MpV2nqBlrv1a1o2bf2qtfLXaQUTO0sb05o3jv3ZXhjS20xEphW5h
|
|
yeJjjf7UdJWNKLziyRePsz0lux1SgAQAAAAAAAAAAAAAADG9K5KTS8Rato2mJZAPIcU4ZbQZuekT
|
|
OC3afT4NXFkmlntc2GmoxWx5K71tG0vHa/RX0GpmlutJ61t6wrY2xr8dXS5uesN+tt4ef0eaa223
|
|
2dnHk3juyreM81OaFGiy/RtZET9jJ7s/2bdutd2jqKeic3iNTsd8a2h1H0jTVtP2o6W+bZbOO+gA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAABje9cdJt
|
|
adohGTLXFTmvO0fy52bJfU23t0pHaqLeL5xdK9Rnvqb+cUjtCi94xxvK3JetKuHrdZvaa1ljb10y
|
|
cnIs1Wt3naJc++TmVWvMz1YWybfMGdsm3eWek0mo4jm8PT0mfW3lDf4V7P5tdMZdRviwfvZ6/TaX
|
|
DpMMYsFIpWPTzXmf+steT8jn8L4Dp+HxF77Zc/4pjpHydYGjC3oAAAAAAAAAAAAAAAAADG9opS1p
|
|
7RG7zszN6WtPe0zLua+3Joss/wBOzhzG2OsL5+IrY09dsSyYRijbHEMvOChb7KjF0yS2LQ169Mso
|
|
S24noyrPVXWejNVKbTuw3T3REdQWU6LYlVvsyiUDPfqupPRr79VuOQX1lZEqoZxIMksd0gT2VT0l
|
|
bPZVbuCaW8i8bwr32WxbcGnkjaZa9p2ndv5qbw5+aNugLItF6TEtvTX5sMb969HMpfazc0d9stqe
|
|
vVZDdAQAAAAAAAAAAAAAAAADV1+iprtPOO/2u9bektoB4TJTJpNRbHkja1Z6uto8viVht+0HDvpG
|
|
H6Tjj6zHHvbecONw7Ltfkmeqmo6Ma69DXbbZTkr1mGWO3RneOaGbZRoM30fVzSelMnT83aef1FZ7
|
|
x3h1tBqfpGnjmn369LNc3sc3kzy9bQCzIAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAAa+q1dNNXr7157VhGp1Xh70x+9f9ocy283m1p5rz3mVbrjXHjt91lz
|
|
5c9+fJ1nyjyhdM8lZlOOIiqrUXikd+kMreunnI5XEdX4dZiZcG+XmtNl/F83PeeWWHDOGanieSKY
|
|
q+5H2rz2hMzWd1Iqx1yajJXHhrNrW6REeb1nCPZumn2z62Ivl7xTyr/6uhwzhGn4Zj2xxzZJ+1kn
|
|
vLoNJnjHW7TbbsAszAAAAAAAAAAAAAAAAAAAAaPFrbaSK/itEOXt0rDf4xb/ACa/GZacRvaF58Q2
|
|
IjasQnzPIhCU92tMbZGzHmotG10C6nZkwpPRmipIllEbMIZIE7solgmJBnCyk9VMM6z1BtVllEqK
|
|
z0WRILYlluriWcSDJVbusV27gwInaSWM9ECyZ3hqamnSWxFmOSOaqRx725bNnSZNs9J+OynVY+WZ
|
|
YYr7TE+nVaIr0Ais81Yn1hKAAAAAAAAAAAAAAAAAABExvG09peU4nov9n66L0j6q/WPg9Y1OJaON
|
|
ZpL0+9HWs/EWzeVz9PbmrEtnyc3h9reHy26TWdnSr2YX6657ijLXpLX0+onSamL/AHJ6W+Tbv2aW
|
|
ekTv16JzeI1Ox6KJiYiY7Slz+E6jxdN4dp3vj6fl5Og2clnKACAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAACZ2jeQRMxEbzO0Q08uqtkma4ulfO3r8lefUePMxWf
|
|
cjy9WvlzVxV6T1Z61/x0Y8f7Wc7Ur1lqVy+LqOWJ2hp6rXddon5rOF1tfmz5OkT0qzb8dWbxjp1c
|
|
biuuilJ5Z6r+IcQrixzEy8zl1E6rNt1tMztFY81sztU1eRucN4ffi2p5esRM72n0h7rS6XFo8FcO
|
|
CkVpX082nwXh3+z9FWLxHi36328vg6TZyW9ABAAAAAAAAAAAAAAAAAAAAAADj8Unm1tK/hqppHvw
|
|
y1k8/EMk+m0GOPeafiFpCZYwolnXspvHvLa9mF46gmnZmwozRUiUCBKYYsoBLOFbKAX0llEqqyzi
|
|
QXRLOJVRLOOwLIljZMEgrlhKyYYTAK5nZPN0RZjugUanHzVlz6xtLq361c+9eXItPpXX0dubTU+E
|
|
bL2lw2++O1fSW6m/VYAISAAAAAAAAAAAAAAAAAp1GbwcfTreelYEydcuMcRrM/L9nnlsV6wqpi2r
|
|
tv133mfWVkRyRtEdGFva7MzkYZNoamWN4bV4mYa9qztKIujhVppxGI8r1mJegeZpknBqKZY+7L0t
|
|
LRekWrO8TG8Ns/HJ5ZypAWZAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAADS12fp4VJ6z9qVuq1HgUiI+3bpDl589cOKZmevqprXPTbx477rDJlrhr1nq4+s182tMRP
|
|
RqaziXiZJrWekNG17ZbxWJ336M5LXRbI3dLTJrs07RMY6fan1dHLrowY+X7MVjt6N3R6Kul0EbWm
|
|
s7bz8Z+LnabQX43r7Y53php/mXj+Dnv0f1JO1x/8ZxbUzj02O15mfLtD13AvZqnDds+pmMmo26el
|
|
XX0Wh0/D8EYtNjilY7+s/NstpOOTW7QBKgAAAAAAAAAAAAAAAAAAAAAADG88tLW9I3BwJtz6nNf1
|
|
vK/DHVqYJ3pzT5y3MPZeojOWMQylEKpTVjZnDCwkqzYQyRRICATCITAJZQxhMAshnEq4ZQC2srKq
|
|
qrIBZCWNZZgwswmFloVyCu0dFcx1WyrtCBhv5NTPHXds2U5o3hIz4ffbPt+KHUcTSW5c9Jme0u2v
|
|
VYAKpAAAAAAAAAAAAAAAAYZctcVOa35R6tLrltN795/YvknNqrfhpPLH92V5isd9mWq6fHjk6rn0
|
|
ZxG8KK5Jm/wbVZiYZtqrmkqL023bkxvCiY3lJHNyRG81mHS4Rn5sNsNp64+3yaWaNrzOzHBl+i6q
|
|
mT7s9J+S+ay8mex6EIneN47SNXKAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAImYiJme0JafEs3h6fkidrZOn5eaLeJk7eOdm1Hi2vmtPTry/CHmOJcUvmvOPF1n09Pm
|
|
6HF9ZGm01qxO3R5vSY7XwzmzTy47zzTEd7en5Mfvt2/PURWdo3tvPrPlKymbktFqTtMTvHzbOLDG
|
|
f63JXbFX7FdnoODcDprZpq9TjiMMTvSn4vj8l5fxnrk91saPSa7i2hpOfbTVt5x1m0fLydzR6PDo
|
|
dPGHBXasd585n1lsRERG0dIF5OOe6tAEqgAAAAAAAAAAAAAAAAAAAAAAADX11+TRZrf0y2Gjxe22
|
|
gtH4piP3TPpXKwxtjhuYo9xq442iIblI2pC1RET2ILd9kxCqRjZmwlCSEohIJAQAAJZISDKGUd2M
|
|
MoBnVbVVCyAWVWeSuqyOwIlXZZKue4MJV2WWYT2QKbKL9YlfdRdIo35b7/Hd3KTzUrPrDh27uxpb
|
|
c2mpPwX/ABX9XAKpAAAAAAAAAAAAAACekTIp1eTwtJmv+GkyJn1oafeazbfpMzLR4jq/o8b823zX
|
|
6XNF8ERCvTcNpxLV5LauvPhx9Irv3lhztdtv8TtaWLicXrt03jzjzb2k1nid56ty3s/w+a7Uwzjn
|
|
1raejlarhmbhl/FpbxMO/fzj5p/ixSeXOvTtRfeI280ZI26tfDm3pWe63LaZx7qtGvniJ6tPLvOK
|
|
fOa9WzbJvTbza02jl3n5SSljscK1MajSxWZ96nSW88xw/VfQ9XMT9nfa3yemid43jtLeXsce88qQ
|
|
EqAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAADia3UTm1l4j7OP3Y/u
|
|
7Vp2rM+kPJW1PhYcmS0+9MzKm/jbwz31weMzbV8UppazPL9q0/BF4rk1GLDSNqxPWPhCnHmnNrtT
|
|
qPKteWPm6U6OdHaZvO+SaRNvhv12Ub/q3FhtrNVj0uKOt56z6R5y9zix1w4qY6RtWsREOJ7L6OKa
|
|
S2rvX6zNM7T6Vh3mmZyOfya7eACzIAAAAAAAAAAAAAAAAAAAAAAAAAAczjVvqMVfW/8AZ03I41bf
|
|
Lp6/OVs/UVrY47NyOzUxd4bUJpEbb3Z7IiOrKIVSjZhMLJYyhKIgmGUQSDESIEbJEgQmCITEAmGU
|
|
IiGUAyhZVhDOoM4Wx2VQtqBKuyyWEgqlhKyyuyBVaGtkbNmvk7A15l1eH2300R6TMORPSXT4ZO+O
|
|
8fFefEX63gEAAAAAAAAAAAAAAAq1WPxdLlp+Kkx+y1Fvsz8gjhaDauGK8sx07y3OE3m1tT6RaP4c
|
|
vU6yMNKUx73zT0ilY3l2eF6a+m0kRl/zbzz3+Ez5M8z26fJruW6wzYq5sV8d43raNpZjRzPPaTmx
|
|
5b6bJ9rHO3zb2WJ8GWPEscY9bgzxH2t62n19GWW0eHOzHU5XbjXZ1x8WTnz2iZ7S2M1IjH2+LX0V
|
|
KTqs8zO9ot0j8nUthi1J3UaOFMTfLFo6xMbS9BwHWTqdHOO8+/hnln5eTjYMFo1WTH5VnePzXcIm
|
|
2k4zlpPSmXy/hfF5eMfJns69OA2cgAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAADG/2LfJ874rW845mubliY7bPoto5qzHrDz0+yePNF41OotaJ7RWNtpV1OtfHqZ715fhu
|
|
j8adNpcVfeyzE2/vLuanhOu1nEctIxTTFa/+ZPbZ3eHcF0vDbTfFE2yzG03t32+DokynXl9+leDB
|
|
TTYKYccbUpWIhYCzEAAAAAAAAAAAAAAAAAAAAAAAAAAAAcXjE/4zDH9M/wAu04XF5/3jj/0f3Wz9
|
|
RUYmzDWxS2I7FSyjuzY1ZKpRKEygEwiWUIkGIk2QJNhKQhMIhkCYZQxhlAMoZwwZwgWQshVCyATL
|
|
CWc9ldpBhZXLOVdpQK7NfJPRdaWvknoDVvPvOnwuel4+TlXn3nS4VPvXj4QtEV0wAAAAAAAAAAAA
|
|
AAAAAVV02CmTxK4qRf8AFFeq0AAAanEsfPpZmO9Ji0NDLfkwdOsulrumiyzHlVzJrz4Ovoy26vB8
|
|
cTBa9NffLtMY77Rv8Yegx5ImkKdJoY1HC81Y+3OSbVn0mGGkmbY45u6tnrrTOu2xGO0RxCd+nNVj
|
|
qKxTV1vH2pjaGtnyzXXYdo96ZmGXEMk15b7/AGZiVerWPTYckZcNbx5wzc7hGbnxXxzPWk7x8pdF
|
|
0S9jh1OXgAlUAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAcPjEf4/FP9H93ccXjMf4vDP9Mx+62fqKrx+S+GvibEFSsqyYwlVK
|
|
ZYsmIMoRKYJQIPIEiQ2ATCUQygCGUIhMAyhnDCGUIFkLIV1ZxIMpVWWSrsCuyqyyyq09ECq8tfJK
|
|
66jJ2Bp5J6upwn7dv9Lk5J951uE/av8AJaIrqAAAAAAAAAAAAAAAAAAAAAAq1Mc2myxPnWf4cmtu
|
|
XT9fR0tffk0WSe28bfq5Wbamm3326MtunwfK6PCv/AxPraZ/dz9PO97/AOqf5dHhdZrw7Dv3mOb9
|
|
XOxRFM+avpe38mvkPHf/AFWlrKba7Tzt99ZxKkfR7euyNXMTrtPHfa0z+zPiM/UR8Zj+Wbdu8HpN
|
|
M2bfzrV13M4dO2pyR61dNvj44/J/oAWZgAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAADj8bj63BPzdhyeNx0wz8ZWz9RWri7Nmv
|
|
VrYu0NmqaRZHZlDGGSiwxZSgCEkCBCQSCQBMJRCYgEsoYx3Z17AlMIhlCBnDOGEM4AlhZZKq4KrK
|
|
7LLKrIFN2vdfZReAaObu6/CO9vk5OePR1uEd7fJeIrqAIAAAAAAAAAAAAAAAAAAAAGtxCk5NFliI
|
|
3mI32+XVyNTyZOHTee946PQKPoeDffw4777eW/yVs60xv+ZxOnr4Okx1t05KRv8Ao41Z5q3yed5m
|
|
XY1szXRZ5jvFJ/hxItP0aOSN9q7yrtr4f2tHFM5+KT16Yq/vK/iGSbXw4vO14UcPx5MGfNbPG18m
|
|
1oj4THRsTw7VanPXVYpi3gzMcnrvCnG11JOupwuN8+a3pEQ6jT4divjxWnJExa09pbjbM5HHu90A
|
|
JUAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAHM41H1GOf6nTc/jEf4Ws+lls/UX45uGekNujTwdm5RNIthKIZKLDFlsiQIShIC
|
|
EgCUJ7AmGTGO7IDzZQhMSDJMMYZQgZwzhhDOATuqssmVdgVWVWWyqtCBTeVF19lF+wNLNG7q8I+9
|
|
8nLyupwnt+S8RXUAQAAAAAAAAAAAAAAAAAAAAAAItWL1mto3iY2lyrcLyUxzix2ia2nvPeK+jrCL
|
|
OrTVnxpanhuPPemSs8l6RtE7dJj0ldpNP9GwRSZ3neZmV4cR/Vs4AJQAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAANHi1d9H
|
|
M+kt5ra+vPoskfDdOfqK4mn7Q3aNHBPZu0W0RdDOGFWcKLCJZeTGQQlCQSgASBsCYZQxhlAJTAmA
|
|
TsmAgGcM4YQyjsgRLC3VnaVcgwsrt3Z2V2QK7tbJ1bN5a9waeWO7p8Knt8nNyebpcK8vkvlFdQBA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAK9RXmwZI+ErEWjesx6wQeZwejeo0cccuW8
|
|
elpblJaaRGxVnCuss4ZrMvJEgCAASISCQIBlCYYpieoM0wx8k7gzIRueYM4Z79FcSy3QEsLJmWFp
|
|
BjaVVpZWlXMoGNmvkXXlr3kGtknu6XCf7OXkl1OEdl8orqgIAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAHmskcmtzV/rls0U62OXiWX4zErcc9GmkRfWVkSqqziWayxCPIANwBIhIJSxS
|
|
CRG6dwZwlhEs4BluMdzfqgZxLLdXuy3AmVdpZTKuZBjaVVpWWV2QlhZRdfZRcGpl7urwfrzfJy8r
|
|
rcH61vPyWitdMAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAHA4nHLxKZ9awnH2ZcY
|
|
jbW459aq8fZpfiI2IZwrqzhmsz3Ebm4JN0AMhCQSIASndiAziWUSriWcAyRujc80DM3RCfIETLCW
|
|
UsZEsJYSslXZAwlTddPZTkBp5e7r8Gj6rJPxhx8k9Xa4PG2C8/FaK10QAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAcfjcbZMFvnDWx9m5x2PqcNvS+zSxT7sNPxH62YZQwqzhRZO6UCB
|
|
KUAJTux3SDIRuAncQAmJZRLBMSgZ7iIAZRKd2DICUSlAljLCYWMLIFVukNfI2bNbIDTyT7zu8Ijb
|
|
Sz/qcG/2nf4T/wCE/wD2WnxWt4AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAHL9oL
|
|
+Hw2cm28VvEuPptfgyVj6yIn0no7/FtJfW8NzYMe3PaPd39d3iMug1WktNc2C9dvPbeP1aZ9xF+v
|
|
T471tHu2iflK2HkqWmvaZj5Surqc9Ps5bx+alTHqYHm68S1Vf/NmfnC2vGNTXvyT84Ql6A3cSvHM
|
|
sfaxVn5Ssrxyv3sM/lKB1xza8bwT3pePyWV4tpZ+/MfOEjfGrXiGlt2zV/PotrqcN/s5aT/+wLRj
|
|
FontMSlAlKEgndO6IAZQljDIEgeQljLCzOVdkCu/SGrkbF56NPNeKxMzMRHxENe0+89DwuNtHHzl
|
|
5PJr8NcnLW3Pbf7r1nCZm2gpae8zMrz4i/W6AgAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAETETG0xukB4HVaeMHEtRi26RedvkyjBSfX9W77QYvC4xz7dMlYlrU7M929dWJLFc6aPK0q
|
|
7YLxPS0S22FlP6q38Zac0yR92s/KVc3tHfFf8tpbcsLRvB/dR/8ALLVnU0r9uL1+dZI1mnmdvGpv
|
|
6TOy6ym+Oto2tWJ+cJ/tW+KLK5KW+zes/KU7tG+h01p64qx8Y6NXNo6Y+uPJlp8rLf0rfG7MXtHa
|
|
0x8pZxqs9e2a8f8A7Oj7HaTHn0+f6RWM23LETfr6vRW4PoL99NT8ui7F4+vEdXXtnt+fVbXjGsr/
|
|
AOZE/OsPS29nuH27YrV+VpeV9pdPXhOtw49NG9Mld55+vXcTPd42I47qo7xSfyWV9oM8d8VJ/VxM
|
|
d8l46xWF9cV7en6o/qLfxp2I9ob+eCv/AHMo9op89P8A/wBORGmyT5R+qfo2X8P7n9Q/jTsx7RR5
|
|
6ef+4/8AuHftg/8A6cWcOSO9J/WEbWr3pY7Efzp2Lcfv5YK/9zWy8d1E/ZpSv5Oba1/+Hb9lc+LP
|
|
bFt87I7E/wAabWbiurvEx4nL/pjZzc2bJkn372t85ZXx55/BX85lucC0vPxnTxlnnjm32mOiZqUu
|
|
LJ2p4TwnVavNWaYbRTfre0bQ99pcH0bT0xb78vmtiIiNojaErMwAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAHnfarF7umzRHaZrLjYrdIen9ocPi8JyTt1xzF4eUw23rCm3R4r6bMy
|
|
wt6kdTaWLdjswmNoZontsCm0K5XWjopnuDC0dGpqG5bs08/daKV672MjbSaif6oh6Z5f2LtvptRX
|
|
0tEvUN3Jfo8f7cYve0eX4zV7B5z20xc/C8eSPuZIRficfXlcPaG7ino08HWIbePpLF2NuiyOyrHK
|
|
3fZFSwuovHVfaVF4QK5YWTM9UT0EKry6Ps1Tn4zjn8NZn9nOtLseydObiWW34cf918fWfk+PYANn
|
|
KAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAq1WKM+ly4p+/WYeBxTNd6zG0xO0
|
|
vobw3FcP0bi2em20Tbmj5Srr418V9sa2Z7qKyzi07MXUylhaU7yjqhLCeiq3ddaFNxFYW7NLNG8t
|
|
zya+WO6Va9J7FW66mvwidnrXiPY3Ny8RyUn71Jj9Ht3RPjk19HK9pMHj8D1ER3rHN+jqqtTjjNps
|
|
uOe16zAifXzfTz7kNyndpYazS9qT0mszDdoxrsi6m8LazMq6zDOsq1ZEyrt1WWlXaUCqyq0rbKbi
|
|
Fdp6PReyFd8uqv8ACsfy83aXrPZHHto89/xX2/SP/dpj6y8vx6EBq5gAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAAB5n2q03LfDqqx39y39npmlxbS/TOG5se29tuavzgWzeV4mtui2
|
|
O3RRSY2hdVhqO2MvI36iu9lUsrSrvDHn6spnmSiq5jooyV6tq1VV69RC32byTh43h8otMx+r6I+Z
|
|
aK/g8TwX7bXh9Mid4iW+fjl8n1ICWb57xLBOm4zqse20Tbmj8+qKdnS9q8PhcTw5tumSm0/OHMxz
|
|
0Za+uzx3sX1t0Zxurr1ZxvspWiZYWZbsbT0QK7KLrZVZJFaqt5vbezNOTg9J/FaZeJns93wCvLwb
|
|
T/GJn92uGHldIBowAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAADuAPA67F9H4l
|
|
qMW20VvO3yRWW97T4fC4rXJHSMtI/WGhVlue3b473K2KzMML4+62tujG9pnozXaOSOVFMnVbmq1t
|
|
trJRW5E7wwvUxTvCyY6CHOt7moxz6Wh9PxTzYaT61h8x1MbZK/OH0zTf+Fxf6I/htj45vL9WgLMn
|
|
mvbPFvocGWO9L7fq85p5maw9d7VYvE4JkmPu2if3eW0+PasdFNOnxfF1Y2hlykRsmY+LJ0MZjZXa
|
|
eq2eyi8oQTO0KLdZWzPRjWu6VaqtHR73g0bcI0sf0Q8Nkq93wqNuFaWP+XDTDDytwBowAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAef9q8HNpcGaI60vtPyl56k9Iew49j8ThGe
|
|
PwxFv0l4zH2U26fDfTYiyJljvsjf4sm6vJ1hrXjq2MkqLdZEVbgbMx0auGdmzNt6iHN1Ub5af6of
|
|
TdPG2nxx6Vj+HzaaTm1+nx/iyVj930ysbViPRrj45vL9SAuyc7j1efguqj+jd4/T33rD3HEcPj8O
|
|
1GP8WOY/Z4TTT7sKadHhbcsZnaCJ3TPZk6VdrKbTutmP0U2nqgrGOsr8deiuI2X09EqKM1dt3uuG
|
|
f/jdN/06/wAPE546S9rwud+Gaaf+XH8NMMPK2wGjAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAABrcRp4nDtRWPPHP8PCYusPoWSvNjtX1iYfPuWaXtX8MzCuvjfw32siu8ptXoxi
|
|
0wy5t4YulReqmazu2skbquURWFInddM7VYRGyL291KFnCcfj8e0le/Lbmn8n0N4b2Ur4nHLWmPsY
|
|
5e5a5+OXyXugBZmiY3iY9Xz7NjnTa3Ph/BeYj5PoTxftFg8Hjk2iOmWkW/Psrr418V5WrWd2faFc
|
|
V2jdnEMXWxntupmN7NiYU27iWML6dVMVnddjgVqMsdHr+CW5uE6f4Rt+7yuSsTDv+zWXn0WTHP3L
|
|
/tK+GHl+O0A1c4AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA8Dn93W56/wDM
|
|
t/L3z59qp24jn+OS38lnpr4r7ZxHQ2TEstt3PXUrt27K57rr1VT0BjKnJPRbMqMs7QlV2fYvHvrd
|
|
VknyrEfu9m8f7FZI8fVU85iJewbT45NfQBKo817W4eulzxHaZrL0rje09ItwqbfhtBVs3leai8RD
|
|
KLw1sduesL606dWFdsZT1jdhNeq6K9DlhCVUU6s4jZnt1YzAhnM71dH2bycmszY/K1d/0c6OzY4R
|
|
fwuK4p8rTstn6z8k7HrwGzkAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAHz3
|
|
Vxvr80/8y38voTwGpj/F5/8AqT/JfjTx/WVeyY6FPspc9dZPVXaOq2WEwIUTVRmjo2rNfLHRI3vZ
|
|
DJycXtX8dZh7t879nsnhcbwz23tt+r6I2nxyb+gCVBzuPY/E4PqI9K7ui19fTxNBnp60n+Aj5/pJ
|
|
3jZu1aOnnltMNussdfXbm+l3ZM9URHREdZVXTuT1Nk7boQiOkJw28PU47/htEp5eivJPLMTCZ9Vv
|
|
x7mJ3iJ9UqNHk8XR4b+tIXuhxAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAD
|
|
weqjbWZ4/wCZP8vePCaz/wDIaiP+Zb+UX408f0r9lOxWOifJhXWjfyYWllPRXYQxnrCrJHRd3YZI
|
|
6A1NJecHEsN/S0T+76bE7xE+r5dk93LW3pL6ZpMni6PDf8VIn9m2fjm8s9rgFmQxvHNS0esbMiew
|
|
PnHLyai9fS0w2aNfUTtrs3+uf5bGPqy068fF227KtSsdFlKqNGMV6myyY6sbdIQI8tlOWOi6Jhhk
|
|
j3RD0vA8nicMx9etZmHRcT2Zyb6XNT8N9/2dt0T449T2AJVAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAHhdfG3E9TH9cvdPEcXjk4zqI/q3L8aeP6xr2TsxpLOekMK6mFo6qpXSrm
|
|
OqBixvHSVmzC4OfqK7S9/wAByeLwbTW9K7fo8Fqo6Paeyl+fglI/Da0NcMPK7QC7AAB8313TiOf/
|
|
AKk/y2MHWrX4jG3E9R/1Lfyv0/aFNOrHxuU7LI7MMayGTVlHWUXhNe6Z6wIUsb9d1m20q7dkDpez
|
|
N9tRqKT5xEvRvKez9+Xis1/FSYerb5+OTyf6AFlAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAB43j9eXjN/jWJ/Z7J5L2mry8Upb8VIF8f6aGOey2eynHvOy7bowrrYSxZSwQJ2YXZ
|
|
92N4BoanrEvVexmTm4blr+HJ/aHltRHSXofYm/1Wrp5RaJaYY+X49WA0c4AD51xONuKan/qW/lbp
|
|
+0MOLRtxbU/9SU4J7KadWPjep2WQrr2WRPRk1TvsndXMpiRCb9FNu0rbTuqvKBscCjfi9PhWZeue
|
|
V9n434rafTHL1TfPxy+T/QAszAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAHmv
|
|
avHtfTZfnV6VxPajHzcNrf8ABeJFs/XnMcr4no18c+6vr2YadkY2YM57sEDLyY37Mo7MMnYGlqO0
|
|
vQ+xNfqNVb1tEfs87qZ2rL0/sVX/AHdnt65P7Q0wx8vx6UBo5wAHz/jUbcX1PT78qtO2vaCnJxjP
|
|
8Zif2amnnspp04+OjWejKJ6MKdmcMmyJn4m5ZHzEVPMwtJv0VZLbQDqezcb8RzT6Y/7vUPM+ytZt
|
|
n1OTyiIh6Ztn45N/6AFlAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAABocbxeLw
|
|
nUR5xXm/Rvq8+OMuDJjntaswEeBxT0bNZ6NatZpNqz3rO0rqsdO3PxlaWEMpY+aqWXkryT0ZT2V3
|
|
7A0dVPuy9f7G124NM/iyT/Z4zWT7sw957MYfB4Fp4/FE2/WWmGHldcBowAAeM9qKcvFeb8VIly9P
|
|
0nq7ntbTbVYL+tJj93CwT76unR4/jo0nozhhTsy3Y1sWljM9Ce7HyQIm3RRlttVbaWrnt0Sh6n2U
|
|
x8vD8mSfv3/h3XN4Bi8Lg2nj8Uc36y6TeOPXugCUAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAPD8RxeBxXUU26Tbmj8+quro+02Lw+I4ssdslNvzhzazvDPbq8d7GW7Dfqz2VzG
|
|
0s2qd+iu/Zn5Ksk9BVztX1mI8930zh2LwOHabH+HHWP2fNYp4+vwYvxXiP3fUqxtWIjyjZtj45/L
|
|
faQFmQADzftfj3w6fJ6WmHmsP23rvaqnNwqLfhvEvIYZ+sV038bo0noy36MK9oZQxrdMyrlnMbMZ
|
|
QKrS1M07zEestq/RRjr4utwY/wAV4j91p9V18fQdJj8LR4ccfdpEfsuREbREJbuMAAAAAAAAAAAA
|
|
BAJAAAAEAJEAJQAJQAJEAJQAJQAJEACUJAQlAJEAJQAJQJAAAEAJEAJBAAAJAABAJEJAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAABwvanDzaPFmjvjv8A
|
|
tLztJ3h7HjGHx+FainnFeaPnHV4vFbeIU038VbHeGF+kso7Mb9mTdhKnLK3dRm7SIrHhGPxeP6Sv
|
|
9cT/AHfSnz72Zx+J7Q45/BWZ/Z9BbZ+OXyfQBZQABzeP4/E4NqI9Ii36S8Ng/wAx9C4jTxOH6ivr
|
|
jn+Hz3B/mQi/GvjdCnWNlsdI2V07LIlg6USrt2ZzZXMoFV+zPhGLxeOaavpbm/RVltEN72Yx+Jxm
|
|
b7dKUmf7L5+s9/HtRA2cqRACRACRACRACUAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAACQAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAQCQQCRACRACRCQBCQBCQB
|
|
ACRACRACRACRACL1i9LVntMbPATTwdRkxT3pea/u+gPE8Xx+DxrPHlaYt+qNfGvjvtXXsi0dOrKk
|
|
dEXjZg6VMtbP2bMtXUdpEV0/Y2nNxbNf8OP+727xvsXH+N1U/wBEfy9k3nxyb+gCVQAGOWvNivX1
|
|
rMPnGGOXNNfOJ2fSZ6w+dZKeHxDPX8N7R+6L8a+L63KdoZ7q6zvEMpnowdKJ6ywmWUyqvIKM0vQ+
|
|
x+D6rU55+9aKx+TzWa36vbezmDwODYenW+95/Nphj5L6dQBo5wAAAAAAAAAAAAAAAAAAAAAAAAAA
|
|
AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAEgAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAACAEiAAAEoA
|
|
AAAAAAAAAAAAAEAkEAkRuAkQbgkQAkQAkQAkQAl5T2nx8nEMOT8dNv0l6pwfarHvpcGWPu32/WCr
|
|
YvK4mOem6b9mGKd4Z3idmFdka0y1c892zfpMtLPaNpEV6D2Kj/Eauf6YeweQ9ieuTVz8K/3evbT4
|
|
5NfQBKoAA8FxCvJxrUx/XMvevD8Zry8fz/Haf2RfjTx/6RSOnRMyypHu9kXjowrqVSrvPRnZVl6V
|
|
kK0775MsUjvadn0nT4ow6bFijtSsVfPuFYvpPGtNTy54mfy6vorXDm8l9pEC7JIgBIgBIgBIgBIg
|
|
BIgBIhIAgBIhIAgBIgBIIBIAAhIAhIAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAJAAAAAAAAAAAAAAA
|
|
AAAAAAAAABAJQkAEAAAAAAAAAAjc3BIjdG4Mkbo5kcwMjdhzHMDPc3V8xzAs3N1fMjmBZubq+Y5g
|
|
Wbm6vmOYFm5ur5jmBZubq+Y5gWbm6vmOYFm5ur5jmBZubq+Y5gWbm6vmTzAz3N2HMnmBlu5ftFTx
|
|
OEZJ/DMW/d0t2rxKni8N1FPWkiZ9eS08e7Cy8dGGn6UhZaJljXZGnmc3UT3dPP2cnUT78xCIV6j2
|
|
H/8A9c/6f7vXPI+w8bU1U+vL/d63du5NfUiDcVSIAS8b7RV5eOb/AIqRL2TyXtNX/e2KfXH/AHlF
|
|
+NPH/pr4+2xcxx0hFpY11K7R16KM32ZWz3UaidqSgrc9kcPicWyZJjfw6T+727y3sXh2xarN+K0V
|
|
h6lvPjj3e0ASqAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAJQAAAAAkQAkQAkAAAAAAAAAAAAAAA
|
|
EgAAAAAAAAAAAAAAAAAAAAAgAAABKDcAN0bgkY8xzAyRux5kcwM9zdXNkTcFm6OZXzMeYFvMibKu
|
|
ZHMC2bo51U2RuC2bom6rc3BZzom6sBZzI52ADPnOdggFnMc6skFnMc6rc3BbznOp3RzAv50c6nml
|
|
HMC/nOf4qOY5wX85zqOc5wbHOc7X5znBsc6edr85zg2ec52vzpi4NjmY5bROG+/bllVzsNTk5dLl
|
|
n0pP8BHmMHWNmzt0aum8obm08vVjfrtnxztR0mXHzTvaZdjVRMTLkZo6yiFen9iZ2pqY/wBP93rN
|
|
3kPY+/LfPX1rE/u9XzN3HfqzdO6vmTuIZ7m7Hc3Bnu8t7TR/vHBP9E/y9Pu837SV31umn+if5Rfi
|
|
/j/01MMb1hjkrtKzBG0bMsmOZY11tOYamr6Und0LUc7XT7u3rJPqL8er9lcPhcFpbzyWm39v7O00
|
|
+FYvA4Zpsc94xxu227jv1IAgAAAAAAAAABKAAAASgASgBIgBIgBIgBIhIAAAAAAAAAAAAAAAAAAC
|
|
UACUJAAAAAAAAAAAABIAAAAAAAAAAAAAAAAAAAAg3AEbomQZbo3YzLGbAz3RNlc3YzcFs2YzdVN2
|
|
M2Bdzom6nmNwW86JurTAMuY3REJ2BB1ZRVMVBhsbSsiqeUFXLucq3lTygp5TlXcpygp5TlXcpygp
|
|
5TlXcqOUFXKjlXcrGYBXysdlswiYBVMdUTCyY6sZBWxlnMMZgGLGZZSwkDdHMiWO4MuY5mEyjcFn
|
|
N1OdVzHMC3nTzqeY5gX85zqOZPMC+Lqdbk20eb/RKOZr8QybaK/XvtH7iZ9aGlp2luzT3fg19NHS
|
|
OjbmPcYX67XH1XSZ9XIzRvMuzrK7zLkZYmYnciunb9lZ5dTk+OP+71cXeP8AZnJ/ip2nf3J/l6iL
|
|
/Fu5L9bMWZczXi6YuIbEWTzKIuyiwLt3nuO25uI4a/hx7/rLuczg8TicvFLbfdpEK6+NPH/phhjo
|
|
stLGkctUWnoxrrU3j1cnWTzZq1jzl1clo5Zcu8c+txR63iP3Tn6pv4+g4o5cVI9IiGe7CJ2iE7t3
|
|
GyN2O6dwSINwSISAlAAlACRAAlAAlACRACRCQAAAAAAAAAASgASISAAAAAAAAAAAAACQAAAAAAAA
|
|
AAAAAASAAAAAAAAAAAAAAAAIAAAQCAJljuljsCJlhMs9mOwMJYys5TkBVsjZdyHICrZPKt5E8oK4
|
|
qmKrOVOwMIqyirPY2Bjyp2ZbAI2NmSARsbMgEbI2ZAMdjZICNkbMkSCNmOzJEgx2YyzljMAwlhKy
|
|
WEwCuWErJhhMArlhLOWEgxljMpljIImWMyTKJA3N0IBO5vux3NwZbnMx3NwZczT4jf3MdPW27a3a
|
|
fJOq1XNP2KdIRfi+J2trSYfcjeF+Wm1OicVeWIiN9kai8xjY12ORqultnI1Ecsujq79XP1FovWYI
|
|
rTgeq+j8QrWZ+3Mx+r2UXeC0WG2Ti2kiN5mL807eUREvbzbaejefHJv62Iv8WUXa0WTFhVtRdlF2
|
|
rz9WUXBtc7jR9dqc2T1ttHyhvZMvJitb0jdq6XHNcNenWVN3028U99WRj6Kb02be3Tq18/SN2Lpc
|
|
3UdN9nOmZrqKX/DaJ/d0svvTLRzV3jomK6+Pd1vvWJj0ZczT0mXxNJht60hfFnQ4qu3N1cWTEgs3
|
|
Tur5k7gz3N2O5uDM3Y7m4MtxBuCQASIASIASAAAAAAACRCQAAAAAAAAEoSAAAAAAAAAAAlAAlCQA
|
|
AAAAAAAAAAASAAAAAAAAAAAAIASgAAAEJAQJQCNkbMgGOyOVnsAw5TlZ7GwMOVPKy2NgY7GzIBGx
|
|
skA2AAAAAAAAAAQkBAEghEskAxYzDPZGwK5hjMLJhjMAqmGEwumrCagomFcw2JqqtUFEsLLrV82F
|
|
o7gqljKyYYTGwMZRKUSCAQAboJnaN5Bjkneu0d5W4ccViIiOzHFWbTzNumP1Zarr8eeRMbxDW1Mx
|
|
NO67NbkhzNVnmInqzaOZrL93JyZeV0M1++7S02jvxDWxhxx033tPpC8Z6rrezWjmZyazJG2/u03h
|
|
2vFibTHoqvamiwVwY+nLGzV0+SZ1Mx8G0/45tOhzJ5lXMc3UVXRdlF1HP+iYsDPLPPy49/tz1+Te
|
|
pSIr0ho6ak5Ms5J8o2q6NImOrHV7XX488ypzTtHXo0s9t6zG7c1G1qz6ubeZiZ3UatXJG3yauSO7
|
|
cvMTEx5tPLb3prPRMVr0HB8vicNxf0+7+kt+LOJwTJyY/Bnz3tH93X36N58cWvq6LSyiyndMSlC7
|
|
mZcymLJiwLosmJVRLKLAtiU7q4lMSCzc3YxJuDMRuAlKAEgAAAlAkAAAAAABKAEgAAAAAJAAAAAA
|
|
AAAAAAAEgAAAAAAAAAAAAAkAAAAAAAAEAAAAAAAAAAAAAAAAAAAAAhIAAACAAAASgAAAAAAEAAAA
|
|
hGzJAImGMwzQDDZjNVuyNgUTVhNGxysZqDVmiu1G5NN2M4waM0+DCaN2cbGcQNGaMZq3JxMJxA1J
|
|
qx2bU4kU09slorWNwa20z02RXHbJbl26QvtFovbHWkxEdJt5y2MOHlr2U1W3jx+1hiw8vSO63lmI
|
|
XRTaEWmtY6snRHO1VpmJ+DjavpSZl2s8b7y4HFcnh0n0gha5ebJN55KRM2mdoiPN6fh+kpwXh0Wy
|
|
RHj5Otp/s5Ps1p62y31+em9aTMYt/OfVfxTiPjZ52naI7fBrI5t66xz5+a1rW7yx0eSL6iZjtEOX
|
|
qNbSletom3lENjh2fbHzbbWt3iVozruc+5ztWubf4M4ybpQ2Oboyrva0Vjza8WdDR4OkXt3n9ldX
|
|
kaePP9VtYqctYhdvt5oivTeCZ2YOxXk6ubqMfV0b9mrljfqlFcq88k7z2U5axeItDa1OPessuC8P
|
|
ya7XRWYnwqdbT/ZMilvIu4dpslNdixXja8Y5tt85djZdbDWnGOesRtXFtuw6T27No5Kx2OrKYQlC
|
|
ExKJgBnEpiyvdlEgsizKLKollFgWxLKJVRLKJBbEp3VxLKJBnuMWQJEbpBIAAAJAAAABIAAAAAAA
|
|
lAJAAAAAAAAAAAAAASAAAAAAAAAAAAAJAAAABAJABAlAAAAAAAAAAAAAAAAAAAAAAAAIAAAAAAAA
|
|
AAABAJQAAAAgAABAAI2EoBGyJhkgGPKxmqxAKpownHC+YRMdN5BrTj67R3bOn01o7p01Iv71u89o
|
|
b9a7LfBTfS1vWI2jf12VfQPSW8KX2mas+NC2iv6xMNfJpMnLtEbuuxtMRCtzF55NR5rPps1N/ctP
|
|
y6uHreE6nXZ4pak48X3rT06fB7fNeI33cbX6mI32R/MWu7XF116aDSRhxbRERs8f499bkyZeeKae
|
|
kzE2mdon81/tfxDLGOunwbzlzbx08oaHBvZHJlx48mrvaa94pu04y617576rNGLRRM0397JEd/lu
|
|
9Dw/S3x4qxffo6mm4NjwUiKY4iI9Ib1dHFY6QIaNabbrYrLfrpJtaK1rMzPZb/s+05IpP59OyLeJ
|
|
k7eNfRaOc1ue32I7fGXYpi5Y77M8OGMeOKxHSFsU3Y29deZMzirl6dlVvhLatCjJHeYQv1rXnps1
|
|
8k9/VsW6qLVmZIi1rzitlvFKRvaZ2h6TSaenC9FFY+3brM+sqeG8Prp4+kZ+lvuxPkr1mqm95nfp
|
|
DXM459676a2q1dsV7XietvNno78+CJn1cjX6mOeIm0bR33dfRU5NJjidt9t5afjG/V6JZ7I2QMNh
|
|
nyo2BhsMuVG3wAhMSbbQRAMolnE+iuGUSCyJZRKuGUSCyJZK4llEgyZMYTuCUsYSCQASISAAAlCQ
|
|
AAAAAAEoASCASAAAAAAAAAAAAlACRACQAAAAAAAAAEgCEoASCAAAAAAAAAAAAAAAAAAAAAAABAAA
|
|
AAAAAAAISAIAAAAAAQAAACASgAAAQJAQAAhIDHZhln3do7z0WS18mWsajHjmes7pg3dNi5aRMNqO
|
|
yvDHTpPRaigHZhN4hHRlaVN59JY3zRENLUavaO+yq0iNVlitJ6vNcR1MVi0zO0era1/Ea0rPvbz5
|
|
PM5MWp45qvo2GZrhmfrsnpHpHzTCseEcM/2vrr8Q1Eb4qzy44nziPN63HpYiIiI7LNHoqabBTFii
|
|
IpSNohuVxrKtWMEejPwY9G1FFmHB4mWJn7MdfnIM9JpIx15to5pbUaas/a6rqViI7MxPxqX0UT1r
|
|
O3wVzpbR2hviP5i03Y5s6a879FNtHljydhExCv8AMTPJXBnRZbz0iG5ptFjwe/l96zctMVamTJtE
|
|
yTMibu1VrdTzRMR0j0ed4lr64MVpm0RERvMz5NvX62uOJ69XhOKX1HH9bHDtFvNYnfJeOy0Z2ojX
|
|
6jjnEq6fRUmccTvN/J9H0eKcOnx45neaxEbubwHgOHg+milI3vP2resu3Wu0JQmITsmISDHZHKz2
|
|
JgFc1RMLJhGwK9iIZ7MZgEdgmAEwyiWCdwWRLKJVxKYsC2JTuriWUSDNlEsIlMAySx3SCRCQSIAS
|
|
AAACRACQAAAAAAASIASAAAAAAAAAAAAAAACRACRACQASIAAAAAAAAAAAAAAAAAAAAAAAAQCUAAAA
|
|
AAAAAAIAAAAAAAAQAAAAAACBICBICAAEJAQJQCJcLjuS2ny6fPG/LWdpd1o8T0X07SXx/e7wCdJx
|
|
Wa0jmneHQpxPDMdZmJfNtZm49weZrh0/j4o7VtSZ2+Uw0/8A7o49k92vBLc/ntFohFW9PqGXimOI
|
|
6Tu1L8T3eCx6r2t1O3JwvHjifO99v7t/Bwf2l1PXU6rS6eJ8qUm8x+so5TsekzcSjbvs4mt4rzW5
|
|
K2mbT0itesy2cHsvbvqtbmyz5xERWP2jd1tJwrTaONsOKtZ8585+cnDrzmn4Rq+IZObUROHD32n7
|
|
Vv8A0ej0uhxaXFGPFSK1j0bkY4jyZRVZVXFGUVWbGwKsk8mObekNrSW3pWf1a2aYjHbm7bNnQ1id
|
|
PW0TvuDdhJEbQABMsLW2R0ZTMQrvfbz2YWzVhpanUxEd0dWkW5c8R5uXxDX1w4pnfr5Q19XxKuOJ
|
|
2neXltVqtVxbV/RdJ715+1bypANfiOu1HENV9C0MTfNeesx2rD1PAeBYuE6aKx72W3W9/WVnBuB4
|
|
eF4dqRzZbdb5J72l160WVK02ZxCYhOwI23TsnY2BGxsnYBjsiYZsZBjMMZZSgGEolMsQDdG6NwZ7
|
|
piVe6YkFsSziVMWZRILolMSriWUSCyJTuwhMSDMRCQSI3SAlACRCQAAEoAEoASAAAAAAAAACUACR
|
|
ACQAAAAAAAAAAAAASAAAAAAAAAAAAAAAAAAACAAAAAAAAAAAAAABAAAAAAAAAAAAACBKAAAAAAAQ
|
|
JQAAAhICEbJAYTWJ7wx8KvpC0BV4ceieWGewDHlNmWwCNjZICNhIDmcZredBecdpiY69FXCOLW+i
|
|
UiZidukulmxxlx2paN4mNng+K4+I8Hy2yaTfl37TXetoCPfRxfp1qi3F48ofKMvtvxak8s6LDv61
|
|
rZji9rPaLUf5PC+bfttS0q8q3p9W/wBrRMdpUZuKdN99nzvFqPbTVz7nD8OKs+do2/mW3h4D7Xaq
|
|
ZnPrtNpqz35aRaYOHY9Zk4pNt9rR+rl6zi+OnS+WN57Rv1lXp/YrNaYtruL6zNPnGO3hxP6O5w/2
|
|
f0HDuun09Yv55Le9afznqcOvO4tBreMTHu30unnva0bWt8on+70nDuE4OHYYx4Kbesz3tPrMuhGO
|
|
IjpDOKrK9YVpsyiGUQnYGOyUgI2SlAIEmwMWMs9kTAMJYzDOYRMArmGErZhhMArlHmzmGMwDE3Ts
|
|
bAbs4swj5pgFkSziVcM4BZEsolXDKAZwyhjCYBkACQhIAAAAAAAJAAAAAAAAAAAAAAAAAAAShIAA
|
|
AAAAAAJAAAAAAAAAAAAAABAJEAAAAAAAAAAAAAAAIEoBKAAAAAAAAAAAAAAABAlAAAAAAAIAAAAA
|
|
BAkBAkBAkBAlACEgMZjdjbFW8bWrEx8YWANb6Fp+bfwab+vLDKMFK9qxH5L0bAr8OPRPKz2AY7J2
|
|
SbAjYZAI2E7AIEgIEgIEgMdkSy2NgY7MdlmyNoBXsxmFuyNgVTVjNV3KjlBRNTlXTVHKCrlIqt5T
|
|
lBhEMohlFerLlBjEMohMVTEARDKCITsAk2AEgAAAkAAAAAAAAAAAAAAAAAAAAAAAASAAAAAAAAD/
|
|
2Q==`;async function kxe(e){let t=(a,s="application/octet-stream")=>fetch(`data:${s};base64,${a}`).then(i=>i.blob()),r,n;switch(e.config.warmup){case"face":r=await t(x1);break;case"body":case"full":r=await t(b1);break;default:r=null}if(r){let a=await createImageBitmap(r);n=await e.detect(a,e.config),a.close()}return n}async function Ixe(e){return new Promise(t=>{let r;switch(e.config.warmup){case"face":r="data:image/jpeg;base64,"+x1;break;case"full":case"body":r="data:image/jpeg;base64,"+b1;break;default:r=null}let n;if(typeof Image!="undefined")n=new Image;else if(he.Image)n=new he.Image;else return;n.onload=async()=>{let a=Kr(n.naturalWidth,n.naturalHeight);if(!a)se("Warmup: Canvas not found"),t(void 0);else{let s=a.getContext("2d");s&&s.drawImage(n,0,0);let i=await e.image(a),o=await e.detect(i.tensor,e.config);t(o)}},r?n.src=r:t(void 0)})}async function Sxe(e){let t=a=>Buffer.from(a,"base64"),r;e.config.warmup==="face"?r=t(x1):r=t(b1);let n;if("node"in Ue){let a=(void 0).decodeJpeg(r),s=a.expandDims(0);e.tf.dispose(a),n=await e.detect(s,e.config),e.tf.dispose(s)}else e.config.debug&&se("Warmup tfjs-node not loaded");return n}async function Cxe(e){let t;return typeof createImageBitmap=="function"?t=await kxe(e):typeof Image!="undefined"||he.Canvas!==void 0?t=await Ixe(e):t=await Sxe(e),t}async function Txe(e){let t=Gr(),r=Dn();if(t!=="webgl"&&t!=="humangl"||!r||!r.checkCompileCompletion)return;Z().set("ENGINE_COMPILE_ONLY",!0);let n=Xt().state.numTensors,a=[];for(let[o,l]of Object.entries(e).filter(([u,d])=>u!==null&&d!==null)){let u=l.inputs&&l.inputs[0]&&l.inputs[0].shape?[...l.inputs[0].shape]:[1,64,64,3],d=l.inputs&&l.inputs[0]&&l.inputs[0].dtype?l.inputs[0].dtype:"float32";for(let p=0;p<u.length;p++)u[p]===-1&&(u[p]=p===0?1:64);let h=_t(u,d);try{let p=l.execute(h);a.push(o),Array.isArray(p)?p.forEach(c=>te(c)):te(p)}catch(p){se("compile fail model:",o)}te(h)}let s=await r.checkCompileCompletionAsync();r.getUniformLocations(),se("compile pass models:",a),se("compile pass kernels:",s.length),Z().set("ENGINE_COMPILE_ONLY",!1);let i=Xt().state.numTensors;i-n>0&&se("tensor leak:",i-n)}async function YN(e,t){let r=oe();return e.state="warmup",t&&(e.config=Gt(e.config,t)),!e.config.warmup||e.config.warmup.length===0||e.config.warmup==="none"?{face:[],body:[],hand:[],gesture:[],object:[],performance:e.performance,timestamp:oe(),persons:[],error:null}:new Promise(async n=>{await Txe(e.models);let a=await Cxe(e),s=oe();e.config.debug&&se("warmup",e.config.warmup,Math.round(s-r),"ms"),e.emit("warmup"),n(a)})}var tp,oc,lc,v1,c3=class{constructor(t){fe(this,"version");fe(this,"config");fe(this,"result");fe(this,"state");fe(this,"process");fe(this,"tf");fe(this,"env");fe(this,"draw");fe(this,"models");fe(this,"events");fe(this,"faceTriangulation");fe(this,"faceUVMap");fe(this,"performance");mp(this,tp,void 0);mp(this,oc,void 0);mp(this,lc,void 0);fe(this,"gl");fe(this,"analyze",(...t)=>{if(!fp(this,oc))return;let r=this.tf.engine().state.numTensors,n=fp(this,tp);gp(this,tp,r);let a=r-n;a!==0&&se(...t,a)});mp(this,v1,t=>{if(!fp(this,lc))return null;if(!t)return"input is not defined";if(this.env.node&&!(t instanceof nt))return"input must be a tensor";try{this.tf.getBackend()}catch(r){return"backend not loaded"}return null});fe(this,"similarity",d3);fe(this,"distance",ic);fe(this,"match",p3);fe(this,"emit",t=>{var r;this.events&&this.events.dispatchEvent&&((r=this.events)==null||r.dispatchEvent(new Event(t)))});this.env=he,ws.wasmPath=Kh["tfjs-core"].includes("-")?"https://vladmandic.github.io/tfjs/dist/":`https://cdn.jsdelivr.net/npm/@tensorflow/tfjs-backend-wasm@${Fy}/dist/`,ws.modelBasePath=he.browser?"../models/":"file://models/",ws.backend=he.browser?"humangl":"tensorflow",this.version=L5,Object.defineProperty(this,"version",{value:L5}),this.config=JSON.parse(JSON.stringify(ws)),Object.seal(this.config),this.config.cacheModels=typeof indexedDB!="undefined",t&&(this.config=Gt(this.config,t)),CC(this.config),this.tf=Ue,this.state="idle",gp(this,tp,0),gp(this,oc,!1),gp(this,lc,!1),this.performance={},this.events=typeof EventTarget!="undefined"?new EventTarget:void 0,this.models=new sc,this.draw={options:Fr,canvas:(r,n)=>a3(r,n),face:(r,n,a)=>Kd(r,n,a),body:(r,n,a)=>Zd(r,n,a),hand:(r,n,a)=>Yd(r,n,a),gesture:(r,n,a)=>Qd(r,n,a),object:(r,n,a)=>Jd(r,n,a),person:(r,n,a)=>n3(r,n,a),all:(r,n,a)=>s3(r,n,a)},this.result={face:[],body:[],hand:[],gesture:[],object:[],performance:{},timestamp:0,persons:[],error:null},this.process={tensor:null,canvas:null},this.faceTriangulation=OT,this.faceUVMap=DT,this.gl=Et,this.emit("create")}reset(){let t=this.config.backend;this.config=JSON.parse(JSON.stringify(ws)),this.config.backend=t}validate(t){return Q2(ws,t||this.config)}now(){return oe()}image(t,r=!0){return _d(t,this.config,r)}async segmentation(t,r){return ON(t,r,this.config)}enhance(t){return kb(t)}compare(t,r){return SC(this.config,t,r)}async init(){await A1(this,!0),await this.tf.ready()}async load(t){this.state="load";let r=oe(),n=Object.values(this.models).filter(i=>i).length;t&&(this.config=Gt(this.config,t)),this.env.initial&&(this.config.debug&&se(`version: ${this.version}`),this.config.debug&&se(`tfjs version: ${this.tf.version["tfjs-core"]}`),await A1(this)||se("error: backend check failed"),await pd(),this.env.browser&&(this.config.debug&&se("configuration:",this.config),this.config.debug&&se("environment:",this.env),this.config.debug&&se("tf flags:",this.tf.ENV.flags))),await Jb(this),this.env.initial&&this.config.debug&&se("tf engine state:",this.tf.engine().state.numBytes,"bytes",this.tf.engine().state.numTensors,"tensors"),this.env.initial=!1,Object.values(this.models).filter(i=>i).length!==n&&(await Qb(this),this.emit("load"));let s=Math.trunc(oe()-r);s>(this.performance.loadModels||0)&&(this.performance.loadModels=this.env.perfadd?(this.performance.loadModels||0)+s:s)}next(t=this.result){return XN(t,this.config)}async warmup(t){let r=oe(),n=await YN(this,t),a=oe();return this.performance.warmup=Math.trunc(a-r),n}async profile(t,r){let n=await this.tf.profile(()=>this.detect(t,r)),a={};for(let o of n.kernels)a[o.name]?a[o.name]+=o.kernelTimeMs:a[o.name]=o.kernelTimeMs;let s=[];Object.entries(a).forEach(o=>s.push({name:o[0],ms:o[1]})),s.sort((o,l)=>l.ms-o.ms),s.length=20;let i={};for(let o of s)i[o.name]=o.ms;return i}async detect(t,r){return this.state="detect",new Promise(async n=>{var g,y,A,x,b,w,S,N,E,R,z,M,I,_,O,q,K,D,Q,V,re,Y;this.state="config";let a;this.config=Gt(this.config,r),this.state="check";let s=fp(this,v1).call(this,t);s&&(se(s,t),this.emit("error"),n({face:[],body:[],hand:[],gesture:[],object:[],performance:this.performance,timestamp:oe(),persons:[],error:s}));let i=oe();await A1(this),await this.load(),a=oe(),this.state="image";let o=await _d(t,this.config);if(this.process=o,this.performance.inputProcess=this.env.perfadd?(this.performance.inputProcess||0)+Math.trunc(oe()-a):Math.trunc(oe()-a),this.analyze("Get Image:"),!o.tensor){this.config.debug&&se("could not convert input to tensor"),this.emit("error"),n({face:[],body:[],hand:[],gesture:[],object:[],performance:this.performance,timestamp:oe(),persons:[],error:"could not convert input to tensor"});return}this.emit("image"),a=oe(),this.config.skipAllowed=await IC(this.config,o.tensor),this.performance.totalFrames||(this.performance.totalFrames=0),this.performance.cachedFrames||(this.performance.cachedFrames=0),this.performance.totalFrames++,this.config.skipAllowed&&this.performance.cachedFrames++,this.performance.cacheCheck=this.env.perfadd?(this.performance.cacheCheck||0)+Math.trunc(oe()-a):Math.trunc(oe()-a),this.analyze("Check Changed:");let l=[],u=[],d=[],h=[];this.state="detect:face",this.config.async?(l=this.config.face.enabled?l3(this,o.tensor):[],this.performance.face&&delete this.performance.face):(a=oe(),l=this.config.face.enabled?await l3(this,o.tensor):[],this.performance.face=this.env.perfadd?(this.performance.face||0)+Math.trunc(oe()-a):Math.trunc(oe()-a)),this.config.async&&(this.config.body.maxDetected===-1||this.config.hand.maxDetected===-1)&&(l=await l),this.analyze("Start Body:"),this.state="detect:body";let p=this.config.body.maxDetected===-1?Gt(this.config,{body:{maxDetected:this.config.face.enabled?1*l.length:1}}):this.config;this.config.async?((g=this.config.body.modelPath)!=null&&g.includes("posenet")?u=this.config.body.enabled?Kb(o.tensor,p):[]:(y=this.config.body.modelPath)!=null&&y.includes("blazepose")?u=this.config.body.enabled?ob(o.tensor,p):[]:(A=this.config.body.modelPath)!=null&&A.includes("efficientpose")?u=this.config.body.enabled?fb(o.tensor,p):[]:(x=this.config.body.modelPath)!=null&&x.includes("movenet")&&(u=this.config.body.enabled?Vb(o.tensor,p):[]),this.performance.body&&delete this.performance.body):(a=oe(),(b=this.config.body.modelPath)!=null&&b.includes("posenet")?u=this.config.body.enabled?await Kb(o.tensor,p):[]:(w=this.config.body.modelPath)!=null&&w.includes("blazepose")?u=this.config.body.enabled?await ob(o.tensor,p):[]:(S=this.config.body.modelPath)!=null&&S.includes("efficientpose")?u=this.config.body.enabled?await fb(o.tensor,p):[]:(N=this.config.body.modelPath)!=null&&N.includes("movenet")&&(u=this.config.body.enabled?await Vb(o.tensor,p):[]),this.performance.body=this.env.perfadd?(this.performance.body||0)+Math.trunc(oe()-a):Math.trunc(oe()-a)),this.analyze("End Body:"),this.analyze("Start Hand:"),this.state="detect:hand";let c=this.config.hand.maxDetected===-1?Gt(this.config,{hand:{maxDetected:this.config.face.enabled?2*l.length:1}}):this.config;this.config.async?((R=(E=this.config.hand.detector)==null?void 0:E.modelPath)!=null&&R.includes("handdetect")?d=this.config.hand.enabled?Rb(o.tensor,c):[]:(M=(z=this.config.hand.detector)==null?void 0:z.modelPath)!=null&&M.includes("handtrack")&&(d=this.config.hand.enabled?Pb(o.tensor,c):[]),this.performance.hand&&delete this.performance.hand):(a=oe(),(_=(I=this.config.hand.detector)==null?void 0:I.modelPath)!=null&&_.includes("handdetect")?d=this.config.hand.enabled?await Rb(o.tensor,c):[]:(q=(O=this.config.hand.detector)==null?void 0:O.modelPath)!=null&&q.includes("handtrack")&&(d=this.config.hand.enabled?await Pb(o.tensor,c):[]),this.performance.hand=this.env.perfadd?(this.performance.hand||0)+Math.trunc(oe()-a):Math.trunc(oe()-a)),this.analyze("End Hand:"),this.analyze("Start Object:"),this.state="detect:object",this.config.async?((K=this.config.object.modelPath)!=null&&K.includes("nanodet")?h=this.config.object.enabled?Gb(o.tensor,this.config):[]:(D=this.config.object.modelPath)!=null&&D.includes("centernet")&&(h=this.config.object.enabled?db(o.tensor,this.config):[]),this.performance.object&&delete this.performance.object):(a=oe(),(Q=this.config.object.modelPath)!=null&&Q.includes("nanodet")?h=this.config.object.enabled?await Gb(o.tensor,this.config):[]:(V=this.config.object.modelPath)!=null&&V.includes("centernet")&&(h=this.config.object.enabled?await db(o.tensor,this.config):[]),this.performance.object=this.env.perfadd?(this.performance.object||0)+Math.trunc(oe()-a):Math.trunc(oe()-a)),this.analyze("End Object:"),this.state="detect:await",this.config.async&&([l,u,d,h]=await Promise.all([l,u,d,h])),this.state="detect:gesture";let f=[];this.config.gesture.enabled&&(a=oe(),f=[...jN(l),...GN(u),...qN(d),...HN(l)],this.config.async?this.performance.gesture&&delete this.performance.gesture:this.performance.gesture=this.env.perfadd?(this.performance.gesture||0)+Math.trunc(oe()-a):Math.trunc(oe()-a)),this.performance.total=this.env.perfadd?(this.performance.total||0)+Math.trunc(oe()-i):Math.trunc(oe()-i);let m=((Y=(re=this.process)==null?void 0:re.tensor)==null?void 0:Y.shape)||[];this.result={face:l,body:u,hand:d,gesture:f,object:h,performance:this.performance,canvas:this.process.canvas,timestamp:Date.now(),error:null,get persons(){return ZN(l,u,d,f,m)}},te(o.tensor),this.emit("detect"),this.state="idle",n(this.result)})}};tp=new WeakMap,oc=new WeakMap,lc=new WeakMap,v1=new WeakMap;return ZE(Exe);})();
|
|
/**
|
|
* @license
|
|
* Copyright 2017 Google LLC. All Rights Reserved.
|
|
* Licensed under the Apache License, Version 2.0 (the "License");
|
|
* you may not use this file except in compliance with the License.
|
|
* You may obtain a copy of the License at
|
|
*
|
|
* http://www.apache.org/licenses/LICENSE-2.0
|
|
*
|
|
* Unless required by applicable law or agreed to in writing, software
|
|
* distributed under the License is distributed on an "AS IS" BASIS,
|
|
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
* See the License for the specific language governing permissions and
|
|
* limitations under the License.
|
|
* =============================================================================
|
|
*/
|
|
/**
|
|
* @license
|
|
* Copyright 2018 Google LLC
|
|
*
|
|
* Use of this source code is governed by an MIT-style
|
|
* license that can be found in the LICENSE file or at
|
|
* https://opensource.org/licenses/MIT.
|
|
* =============================================================================
|
|
*/
|
|
/**
|
|
* @license
|
|
* Copyright 2018 Google LLC. All Rights Reserved.
|
|
* Licensed under the Apache License, Version 2.0 (the "License");
|
|
* you may not use this file except in compliance with the License.
|
|
* You may obtain a copy of the License at
|
|
*
|
|
* http://www.apache.org/licenses/LICENSE-2.0
|
|
*
|
|
* Unless required by applicable law or agreed to in writing, software
|
|
* distributed under the License is distributed on an "AS IS" BASIS,
|
|
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
* See the License for the specific language governing permissions and
|
|
* limitations under the License.
|
|
*
|
|
* =============================================================================
|
|
*/
|
|
/**
|
|
* @license
|
|
* Copyright 2018 Google LLC. All Rights Reserved.
|
|
* Licensed under the Apache License, Version 2.0 (the "License");
|
|
* you may not use this file except in compliance with the License.
|
|
* You may obtain a copy of the License at
|
|
*
|
|
* http://www.apache.org/licenses/LICENSE-2.0
|
|
*
|
|
* Unless required by applicable law or agreed to in writing, software
|
|
* distributed under the License is distributed on an "AS IS" BASIS,
|
|
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
* See the License for the specific language governing permissions and
|
|
* limitations under the License.
|
|
* =============================================================================
|
|
*/
|
|
/**
|
|
* @license
|
|
* Copyright 2019 Google LLC
|
|
*
|
|
* Use of this source code is governed by an MIT-style
|
|
* license that can be found in the LICENSE file or at
|
|
* https://opensource.org/licenses/MIT.
|
|
* =============================================================================
|
|
*/
|
|
/**
|
|
* @license
|
|
* Copyright 2019 Google LLC. All Rights Reserved.
|
|
* Licensed under the Apache License, Version 2.0 (the "License");
|
|
* you may not use this file except in compliance with the License.
|
|
* You may obtain a copy of the License at
|
|
*
|
|
* http://www.apache.org/licenses/LICENSE-2.0
|
|
*
|
|
* Unless required by applicable law or agreed to in writing, software
|
|
* distributed under the License is distributed on an "AS IS" BASIS,
|
|
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
* See the License for the specific language governing permissions and
|
|
* limitations under the License.
|
|
*
|
|
* =============================================================================
|
|
*/
|
|
/**
|
|
* @license
|
|
* Copyright 2019 Google LLC. All Rights Reserved.
|
|
* Licensed under the Apache License, Version 2.0 (the "License");
|
|
* you may not use this file except in compliance with the License.
|
|
* You may obtain a copy of the License at
|
|
*
|
|
* http://www.apache.org/licenses/LICENSE-2.0
|
|
*
|
|
* Unless required by applicable law or agreed to in writing, software
|
|
* distributed under the License is distributed on an "AS IS" BASIS,
|
|
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
* See the License for the specific language governing permissions and
|
|
* limitations under the License.
|
|
* =============================================================================
|
|
*/
|
|
/**
|
|
* @license
|
|
* Copyright 2020 Google Inc. All Rights Reserved.
|
|
* Licensed under the Apache License, Version 2.0 (the "License");
|
|
* you may not use this file except in compliance with the License.
|
|
* You may obtain a copy of the License at
|
|
*
|
|
* http://www.apache.org/licenses/LICENSE-2.0
|
|
*
|
|
* Unless required by applicable law or agreed to in writing, software
|
|
* distributed under the License is distributed on an "AS IS" BASIS,
|
|
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
* See the License for the specific language governing permissions and
|
|
* limitations under the License.
|
|
* =============================================================================
|
|
*/
|
|
/**
|
|
* @license
|
|
* Copyright 2020 Google LLC
|
|
*
|
|
* Use of this source code is governed by an MIT-style
|
|
* license that can be found in the LICENSE file or at
|
|
* https://opensource.org/licenses/MIT.
|
|
* =============================================================================
|
|
*/
|
|
/**
|
|
* @license
|
|
* Copyright 2020 Google LLC. All Rights Reserved.
|
|
* Licensed under the Apache License, Version 2.0 (the "License");
|
|
* you may not use backend file except in compliance with the License.
|
|
* You may obtain a copy of the License at
|
|
*
|
|
* http://www.apache.org/licenses/LICENSE-2.0
|
|
*
|
|
* Unless required by applicable law or agreed to in writing, software
|
|
* distributed under the License is distributed on an "AS IS" BASIS,
|
|
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
* See the License for the specific language governing permissions and
|
|
* limitations under the License.
|
|
* =============================================================================
|
|
*/
|
|
/**
|
|
* @license
|
|
* Copyright 2020 Google LLC. All Rights Reserved.
|
|
* Licensed under the Apache License, Version 2.0 (the "License");
|
|
* you may not use this file except in compliance with the License.
|
|
* You may obtain a copy of the License at
|
|
*
|
|
* http://www.apache.org/licenses/LICENSE-2.0
|
|
*
|
|
* Unless required by applicable law or agreed to in writing, software
|
|
* distributed under the License is distributed on an "AS IS" BASIS,
|
|
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
* See the License for the specific language governing permissions and
|
|
* limitations under the License.
|
|
* =============================================================================
|
|
*/
|
|
/**
|
|
* @license
|
|
* Copyright 2020 Google LLC. All Rights Reserved.
|
|
* Licensed under the Apache License, Version 2.0 (the License);
|
|
* you may not use this file except in compliance with the License.
|
|
* You may obtain a copy of the License at
|
|
*
|
|
* http://www.apache.org/licenses/LICENSE-2.0
|
|
*
|
|
* Unless required by applicable law or agreed to in writing, software
|
|
* distributed under the License is distributed on an AS IS BASIS,
|
|
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
* See the License for the specific language governing permissions and
|
|
* limitations under the License.
|
|
* =============================================================================
|
|
*/
|
|
/**
|
|
* @license
|
|
* Copyright 2021 Google LLC. All Rights Reserved.
|
|
* Licensed under the Apache License, Version 2.0 (the "License");
|
|
* you may not use this file except in compliance with the License.
|
|
* You may obtain a copy of the License at
|
|
*
|
|
* http://www.apache.org/licenses/LICENSE-2.0
|
|
*
|
|
* Unless required by applicable law or agreed to in writing, software
|
|
* distributed under the License is distributed on an "AS IS" BASIS,
|
|
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
* See the License for the specific language governing permissions and
|
|
* limitations under the License.
|
|
* =============================================================================
|
|
*/
|
|
/**
|
|
* @license
|
|
* Copyright 2021 Google LLC. All Rights Reserved.
|
|
* Licensed under the Apache License, Version 2.0 (the "License");
|
|
* you may not use this file except in compliance with the License.
|
|
* You may obtain a copy of the License at
|
|
*
|
|
* https://www.apache.org/licenses/LICENSE-2.0
|
|
*
|
|
* Unless required by applicable law or agreed to in writing, software
|
|
* distributed under the License is distributed on an "AS IS" BASIS,
|
|
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
* See the License for the specific language governing permissions and
|
|
* limitations under the License.
|
|
* =============================================================================
|
|
*/
|
|
/**
|
|
* @license
|
|
* Copyright 2021 Google LLC. All Rights Reserved.
|
|
* Licensed under the Apache License, Version 2.0 (the License);
|
|
* you may not use this file except in compliance with the License.
|
|
* You may obtain a copy of the License at
|
|
*
|
|
* http://www.apache.org/licenses/LICENSE-2.0
|
|
*
|
|
* Unless required by applicable law or agreed to in writing, software
|
|
* distributed under the License is distributed on an AS IS BASIS,
|
|
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
* See the License for the specific language governing permissions and
|
|
* limitations under the License.
|
|
* =============================================================================
|
|
*/
|
|
/**
|
|
* @license
|
|
* Copyright 2022 Google Inc. All Rights Reserved.
|
|
* Licensed under the Apache License, Version 2.0 (the "License");
|
|
* you may not use this file except in compliance with the License.
|
|
* You may obtain a copy of the License at
|
|
*
|
|
* http://www.apache.org/licenses/LICENSE-2.0
|
|
*
|
|
* Unless required by applicable law or agreed to in writing, software
|
|
* distributed under the License is distributed on an "AS IS" BASIS,
|
|
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
* See the License for the specific language governing permissions and
|
|
* limitations under the License.
|
|
* =============================================================================
|
|
*/
|
|
/**
|
|
* @license
|
|
* Copyright 2022 Google LLC
|
|
*
|
|
* Use of this source code is governed by an MIT-style
|
|
* license that can be found in the LICENSE file or at
|
|
* https://opensource.org/licenses/MIT.
|
|
* =============================================================================
|
|
*/
|
|
/**
|
|
* @license
|
|
* Copyright 2022 Google LLC. All Rights Reserved.
|
|
* Licensed under the Apache License, Version 2.0 (the "License");
|
|
* you may not use this file except in compliance with the License.
|
|
* You may obtain a copy of the License at
|
|
*
|
|
* http://www.apache.org/licenses/LICENSE-2.0
|
|
*
|
|
* Unless required by applicable law or agreed to in writing, software
|
|
* distributed under the License is distributed on an "AS IS" BASIS,
|
|
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
* See the License for the specific language governing permissions and
|
|
* limitations under the License.
|
|
* =============================================================================
|
|
*/
|
|
/**
|
|
* @license
|
|
* Copyright 2022 Google LLC. All Rights Reserved.
|
|
* Licensed under the Apache License, Version 2.0 (the 'License');
|
|
* you may not use this file except in compliance with the License.
|
|
* You may obtain a copy of the License at
|
|
*
|
|
* http://www.apache.org/licenses/LICENSE-2.0
|
|
*
|
|
* Unless required by applicable law or agreed to in writing, software
|
|
* distributed under the License is distributed on an 'AS IS' BASIS,
|
|
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
* See the License for the specific language governing permissions and
|
|
* limitations under the License.
|
|
* =============================================================================
|
|
*/
|
|
/**
|
|
* Human main module
|
|
* @default Human Library
|
|
* @summary <https://github.com/vladmandic/human>
|
|
* @author <https://github.com/vladmandic>
|
|
* @copyright <https://github.com/vladmandic>
|
|
* @license MIT
|
|
*/
|
|
/**
|
|
* @license
|
|
* Copyright 2018 Google LLC. All Rights Reserved.
|
|
* Licensed under the Apache License, Version 2.0 (the "License");
|
|
* you may not use this file except in compliance with the License.
|
|
* You may obtain a copy of the License at
|
|
*
|
|
* http://www.apache.org/licenses/LICENSE-2.0
|
|
*
|
|
* Unless required by applicable law or agreed to in writing, software
|
|
* distributed under the License is distributed on an "AS IS" BASIS,
|
|
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
* See the License for the specific language governing permissions and
|
|
* limitations under the License.
|
|
* =============================================================================
|
|
*/
|
|
/** @license See the LICENSE file. */
|