ambiser
/
face-api
mirror of https://github.com/vladmandic/face-api.git
9
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity
face-api/dist/face-api.min.js

114 lines
349 KiB
JavaScript
Raw Blame History

var faceapi=(()=>{var Xs=Object.defineProperty,Sv=Object.prototype.hasOwnProperty,S=(e,t)=>()=>(t||(t={exports:{}},e(t.exports,t)),t.exports),lf=e=>Xs(e,"__esModule",{value:!0}),Oe=(e,t)=>{lf(e);for(var n in t)Xs(e,n,{get:t[n],enumerable:!0})},Tv=(e,t)=>{if(lf(e),typeof t=="object"||typeof t=="function")for(let n in t)!Sv.call(e,n)&&n!=="default"&&Xs(e,n,{get:()=>t[n],enumerable:!0});return e},Iv=e=>e&&e.__esModule?e:Tv(Xs({},"default",{value:e,enumerable:!0}),e);var Be=S(Av=>{Oe(Av,{__assign:()=>Zp,__asyncDelegator:()=>Wv,__asyncGenerator:()=>Uv,__asyncValues:()=>qv,__await:()=>hi,__awaiter:()=>Ov,__classPrivateFieldGet:()=>$v,__classPrivateFieldSet:()=>Hv,__createBinding:()=>Qp,__decorate:()=>Rv,__exportStar:()=>Dv,__extends:()=>Nv,__generator:()=>kv,__importDefault:()=>Gv,__importStar:()=>jv,__makeTemplateObject:()=>Bv,__metadata:()=>Ev,__param:()=>Cv,__read:()=>uf,__rest:()=>_v,__spread:()=>Fv,__spreadArrays:()=>Mv,__values:()=>el});var Xp=function(e,t){return Xp=Object.setPrototypeOf||{__proto__:[]}instanceof Array&&function(n,r){n.__proto__=r}||function(n,r){for(var i in r)Object.prototype.hasOwnProperty.call(r,i)&&(n[i]=r[i])},Xp(e,t)};function Nv(e,t){Xp(e,t);function n(){this.constructor=e}e.prototype=t===null?Object.create(t):(n.prototype=t.prototype,new n())}var Zp=function(){return Zp=Object.assign||function(t){for(var n,r=1,i=arguments.length;r<i;r++){n=arguments[r];for(var o in n)Object.prototype.hasOwnProperty.call(n,o)&&(t[o]=n[o])}return t},Zp.apply(this,arguments)};function _v(e,t){var n={};for(var r in e)Object.prototype.hasOwnProperty.call(e,r)&&t.indexOf(r)<0&&(n[r]=e[r]);if(e!=null&&typeof Object.getOwnPropertySymbols=="function")for(var i=0,r=Object.getOwnPropertySymbols(e);i<r.length;i++)t.indexOf(r[i])<0&&Object.prototype.propertyIsEnumerable.call(e,r[i])&&(n[r[i]]=e[r[i]]);return n}function Rv(e,t,n,r){var i=arguments.length,o=i<3?t:r===null?r=Object.getOwnPropertyDescriptor(t,n):r,a;if(typeof Reflect=="object"&&typeof Reflect.decorate=="function")o=Reflect.decorate(e,t,n,r);else for(var s=e.length-1;s>=0;s--)(a=e[s])&&(o=(i<3?a(o):i>3?a(t,n,o):a(t,n))||o);return i>3&&o&&Object.defineProperty(t,n,o),o}function Cv(e,t){return function(n,r){t(n,r,e)}}function Ev(e,t){if(typeof Reflect=="object"&&typeof Reflect.metadata=="function")return Reflect.metadata(e,t)}function Ov(e,t,n,r){function i(o){return o instanceof n?o:new n(function(a){a(o)})}return new(n||(n=Promise))(function(o,a){function s(l){try{p(r.next(l))}catch(u){a(u)}}function c(l){try{p(r.throw(l))}catch(u){a(u)}}function p(l){l.done?o(l.value):i(l.value).then(s,c)}p((r=r.apply(e,t||[])).next())})}function kv(e,t){var n={label:0,sent:function(){if(o[0]&1)throw o[1];return o[1]},trys:[],ops:[]},r,i,o,a;return a={next:s(0),throw:s(1),return:s(2)},typeof Symbol=="function"&&(a[Symbol.iterator]=function(){return this}),a;function s(p){return function(l){return c([p,l])}}function c(p){if(r)throw new TypeError("Generator is already executing.");for(;n;)try{if(r=1,i&&(o=p[0]&2?i.return:p[0]?i.throw||((o=i.return)&&o.call(i),0):i.next)&&!(o=o.call(i,p[1])).done)return o;(i=0,o)&&(p=[p[0]&2,o.value]);switch(p[0]){case 0:case 1:o=p;break;case 4:return n.label++,{value:p[1],done:!1};case 5:n.label++,i=p[1],p=[0];continue;case 7:p=n.ops.pop(),n.trys.pop();continue;default:if(!(o=n.trys,o=o.length>0&&o[o.length-1])&&(p[0]===6||p[0]===2)){n=0;continue}if(p[0]===3&&(!o||p[1]>o[0]&&p[1]<o[3])){n.label=p[1];break}if(p[0]===6&&n.label<o[1]){n.label=o[1],o=p;break}if(o&&n.label<o[2]){n.label=o[2],n.ops.push(p);break}o[2]&&n.ops.pop(),n.trys.pop();continue}p=t.call(e,n)}catch(l){p=[6,l],i=0}finally{r=o=0}if(p[0]&5)throw p[1];return{value:p[0]?p[1]:void 0,done:!0}}}var Qp=Object.create?function(e,t,n,r){r===void 0&&(r=n),Object.defineProperty(e,r,{enumerable:!0,get:function(){return t[n]}})}:function(e,t,n,r){r===void 0&&(r=n),e[r]=t[n]};function Dv(e,t){for(var n in e)n!=="default"&&!Object.prototype.hasOwnProperty.call(t,n)&&Qp(t,e,n)}function el(e){var t=typeof Symbol=="function"&&Symbol.iterator,n=t&&e[t],r=0;if(n)return n.call(e);if(e&&typeof e.length=="number")return{next:function(){return e&&r>=e.length&&(e=void 0),{value:e&&e[r++],done:!e}}};throw new TypeError(t?"Object is not iterable.":"Symbol.iterator is not defined.")}function uf(e,t){var n=typeof Symbol=="function"&&e[Symbol.iterator];if(!n)return e;var r=n.call(e),i,o=[],a;try{for(;(t===void 0||t-- >0)&&!(i=r.next()).done;)o.push(i.value)}catch(s){a={error:s}}finally{try{i&&!i.done&&(n=r.return)&&n.call(r)}finally{if(a)throw a.error}}return o}function Fv(){for(var e=[],t=0;t<arguments.length;t++)e=e.concat(uf(arguments[t]));return e}function Mv(){for(var e=0,t=0,n=arguments.length;t<n;t++)e+=arguments[t].length;for(var r=Array(e),i=0,t=0;t<n;t++)for(var o=arguments[t],a=0,s=o.length;a<s;a++,i++)r[i]=o[a];return r}function hi(e){return this instanceof hi?(this.v=e,this):new hi(e)}function Uv(e,t,n){if(!Symbol.asyncIterator)throw new TypeError("Symbol.asyncIterator is not defined.");var r=n.apply(e,t||[]),i,o=[];return i={},a("next"),a("throw"),a("return"),i[Symbol.asyncIterator]=function(){return this},i;function a(m){r[m]&&(i[m]=function(b){return new Promise(function(y,w){o.push([m,b,y,w])>1||s(m,b)})})}function s(m,b){try{c(r[m](b))}catch(y){u(o[0][3],y)}}function c(m){m.value instanceof hi?Promise.resolve(m.value.v).then(p,l):u(o[0][2],m)}function p(m){s("next",m)}function l(m){s("throw",m)}function u(m,b){(m(b),o.shift(),o.length)&&s(o[0][0],o[0][1])}}function Wv(e){var t,n;return t={},r("next"),r("throw",function(i){throw i}),r("return"),t[Symbol.iterator]=function(){return this},t;function r(i,o){t[i]=e[i]?function(a){return(n=!n)?{value:hi(e[i](a)),done:i==="return"}:o?o(a):a}:o}}function qv(e){if(!Symbol.asyncIterator)throw new TypeError("Symbol.asyncIterator is not defined.");var t=e[Symbol.asyncIterator],n;return t?t.call(e):(e=typeof el=="function"?el(e):e[Symbol.iterator](),n={},r("next"),r("throw"),r("return"),n[Symbol.asyncIterator]=function(){return this},n);function r(o){n[o]=e[o]&&function(a){return new Promise(function(s,c){a=e[o](a),i(s,c,a.done,a.value)})}}function i(o,a,s,c){Promise.resolve(c).then(function(p){o({value:p,done:s})},a)}}function Bv(e,t){return Object.defineProperty?Object.defineProperty(e,"raw",{value:t}):e.raw=t,e}var Pv=Object.create?function(e,t){Object.defineProperty(e,"default",{enumerable:!0,value:t})}:function(e,t){e.default=t};function jv(e){if(e&&e.__esModule)return e;var t={};if(e!=null)for(var n in e)n!=="default"&&Object.prototype.hasOwnProperty.call(e,n)&&Qp(t,e,n);return Pv(t,e),t}function Gv(e){return e&&e.__esModule?e:{default:e}}function $v(e,t){if(!t.has(e))throw new TypeError("attempted to get private field on non-instance");return t.get(e)}function Hv(e,t,n){if(!t.has(e))throw new TypeError("attempted to set private field on non-instance");return t.set(e,n),n}});var df=S((hf,tl)=>{(function(e,t,n){function r(s){var c=this,p=a();c.next=function(){var l=2091639*c.s0+c.c*23283064365386963e-26;return c.s0=c.s1,c.s1=c.s2,c.s2=l-(c.c=l|0)},c.c=1,c.s0=p(" "),c.s1=p(" "),c.s2=p(" "),c.s0-=p(s),c.s0<0&&(c.s0+=1),c.s1-=p(s),c.s1<0&&(c.s1+=1),c.s2-=p(s),c.s2<0&&(c.s2+=1),p=null}function i(s,c){return c.c=s.c,c.s0=s.s0,c.s1=s.s1,c.s2=s.s2,c}function o(s,c){var p=new r(s),l=c&&c.state,u=p.next;return u.int32=function(){return p.next()*4294967296|0},u.double=function(){return u()+(u()*2097152|0)*11102230246251565e-32},u.quick=u,l&&(typeof l=="object"&&i(l,p),u.state=function(){return i(p,{})}),u}function a(){var s=4022871197,c=function(p){p=p.toString();for(var l=0;l<p.length;l++){s+=p.charCodeAt(l);var u=.02519603282416938*s;s=u>>>0,u-=s,u*=s,s=u>>>0,u-=s,s+=u*4294967296}return(s>>>0)*23283064365386963e-26};return c}t&&t.exports?t.exports=o:n&&n.amd?n(function(){return o}):this.alea=o})(hf,typeof tl=="object"&&tl,typeof define=="function"&&define)});var ff=S((mf,nl)=>{(function(e,t,n){function r(a){var s=this,c="";s.x=0,s.y=0,s.z=0,s.w=0,s.next=function(){var l=s.x^s.x<<11;return s.x=s.y,s.y=s.z,s.z=s.w,s.w^=s.w>>>19^l^l>>>8},a===(a|0)?s.x=a:c+=a;for(var p=0;p<c.length+64;p++)s.x^=c.charCodeAt(p)|0,s.next()}function i(a,s){return s.x=a.x,s.y=a.y,s.z=a.z,s.w=a.w,s}function o(a,s){var c=new r(a),p=s&&s.state,l=function(){return(c.next()>>>0)/4294967296};return l.double=function(){do var u=c.next()>>>11,m=(c.next()>>>0)/4294967296,b=(u+m)/(1<<21);while(b===0);return b},l.int32=c.next,l.quick=l,p&&(typeof p=="object"&&i(p,c),l.state=function(){return i(c,{})}),l}t&&t.exports?t.exports=o:n&&n.amd?n(function(){return o}):this.xor128=o})(mf,typeof nl=="object"&&nl,typeof define=="function"&&define)});var bf=S((gf,rl)=>{(function(e,t,n){function r(a){var s=this,c="";s.next=function(){var l=s.x^s.x>>>2;return s.x=s.y,s.y=s.z,s.z=s.w,s.w=s.v,(s.d=s.d+362437|0)+(s.v=s.v^s.v<<4^(l^l<<1))|0},s.x=0,s.y=0,s.z=0,s.w=0,s.v=0,a===(a|0)?s.x=a:c+=a;for(var p=0;p<c.length+64;p++)s.x^=c.charCodeAt(p)|0,p==c.length&&(s.d=s.x<<10^s.x>>>4),s.next()}function i(a,s){return s.x=a.x,s.y=a.y,s.z=a.z,s.w=a.w,s.v=a.v,s.d=a.d,s}function o(a,s){var c=new r(a),p=s&&s.state,l=function(){return(c.next()>>>0)/4294967296};return l.double=function(){do var u=c.next()>>>11,m=(c.next()>>>0)/4294967296,b=(u+m)/(1<<21);while(b===0);return b},l.int32=c.next,l.quick=l,p&&(typeof p=="object"&&i(p,c),l.state=function(){return i(c,{})}),l}t&&t.exports?t.exports=o:n&&n.amd?n(function(){return o}):this.xorwow=o})(gf,typeof rl=="object"&&rl,typeof define=="function"&&define)});var yf=S((wf,il)=>{(function(e,t,n){function r(a){var s=this;s.next=function(){var p=s.x,l=s.i,u,m,b;return u=p[l],u^=u>>>7,m=u^u<<24,u=p[l+1&7],m^=u^u>>>10,u=p[l+3&7],m^=u^u>>>3,u=p[l+4&7],m^=u^u<<7,u=p[l+7&7],u=u^u<<13,m^=u^u<<9,p[l]=m,s.i=l+1&7,m};function c(p,l){var u,m,b=[];if(l===(l|0))m=b[0]=l;else for(l=""+l,u=0;u<l.length;++u)b[u&7]=b[u&7]<<15^l.charCodeAt(u)+b[u+1&7]<<13;for(;b.length<8;)b.push(0);for(u=0;u<8&&b[u]===0;++u);for(u==8?m=b[7]=-1:m=b[u],p.x=b,p.i=0,u=256;u>0;--u)p.next()}c(s,a)}function i(a,s){return s.x=a.x.slice(),s.i=a.i,s}function o(a,s){a==null&&(a=+new Date());var c=new r(a),p=s&&s.state,l=function(){return(c.next()>>>0)/4294967296};return l.double=function(){do var u=c.next()>>>11,m=(c.next()>>>0)/4294967296,b=(u+m)/(1<<21);while(b===0);return b},l.int32=c.next,l.quick=l,p&&(p.x&&i(p,c),l.state=function(){return i(c,{})}),l}t&&t.exports?t.exports=o:n&&n.amd?n(function(){return o}):this.xorshift7=o})(wf,typeof il=="object"&&il,typeof define=="function"&&define)});var Lf=S((xf,ol)=>{(function(e,t,n){function r(a){var s=this;s.next=function(){var p=s.w,l=s.X,u=s.i,m,b;return s.w=p=p+1640531527|0,b=l[u+34&127],m=l[u=u+1&127],b^=b<<13,m^=m<<17,b^=b>>>15,m^=m>>>12,b=l[u]=b^m,s.i=u,b+(p^p>>>16)|0};function c(p,l){var u,m,b,y,w,L=[],T=128;for(l===(l|0)?(m=l,l=null):(l=l+"\0",m=0,T=Math.max(T,l.length)),b=0,y=-32;y<T;++y)l&&(m^=l.charCodeAt((y+32)%l.length)),y===0&&(w=m),m^=m<<10,m^=m>>>15,m^=m<<4,m^=m>>>13,y>=0&&(w=w+1640531527|0,u=L[y&127]^=m+w,b=u==0?b+1:0);for(b>=128&&(L[(l&&l.length||0)&127]=-1),b=127,y=4*128;y>0;--y)m=L[b+34&127],u=L[b=b+1&127],m^=m<<13,u^=u<<17,m^=m>>>15,u^=u>>>12,L[b]=m^u;p.w=w,p.X=L,p.i=b}c(s,a)}function i(a,s){return s.i=a.i,s.w=a.w,s.X=a.X.slice(),s}function o(a,s){a==null&&(a=+new Date());var c=new r(a),p=s&&s.state,l=function(){return(c.next()>>>0)/4294967296};return l.double=function(){do var u=c.next()>>>11,m=(c.next()>>>0)/4294967296,b=(u+m)/(1<<21);while(b===0);return b},l.int32=c.next,l.quick=l,p&&(p.X&&i(p,c),l.state=function(){return i(c,{})}),l}t&&t.exports?t.exports=o:n&&n.amd?n(function(){return o}):this.xor4096=o})(xf,typeof ol=="object"&&ol,typeof define=="function"&&define)});var Sf=S((vf,sl)=>{(function(e,t,n){function r(a){var s=this,c="";s.next=function(){var l=s.b,u=s.c,m=s.d,b=s.a;return l=l<<25^l>>>7^u,u=u-m|0,m=m<<24^m>>>8^b,b=b-l|0,s.b=l=l<<20^l>>>12^u,s.c=u=u-m|0,s.d=m<<16^u>>>16^b,s.a=b-l|0},s.a=0,s.b=0,s.c=2654435769|0,s.d=1367130551,a===Math.floor(a)?(s.a=a/4294967296|0,s.b=a|0):c+=a;for(var p=0;p<c.length+20;p++)s.b^=c.charCodeAt(p)|0,s.next()}function i(a,s){return s.a=a.a,s.b=a.b,s.c=a.c,s.d=a.d,s}function o(a,s){var c=new r(a),p=s&&s.state,l=function(){return(c.next()>>>0)/4294967296};return l.double=function(){do var u=c.next()>>>11,m=(c.next()>>>0)/4294967296,b=(u+m)/(1<<21);while(b===0);return b},l.int32=c.next,l.quick=l,p&&(typeof p=="object"&&i(p,c),l.state=function(){return i(c,{})}),l}t&&t.exports?t.exports=o:n&&n.amd?n(function(){return o}):this.tychei=o})(vf,typeof sl=="object"&&sl,typeof define=="function"&&define)});var Tf=S(()=>{});var If=S((Rk,Zs)=>{(function(e,t){var n=this,r=256,i=6,o=52,a="random",s=t.pow(r,i),c=t.pow(2,o),p=c*2,l=r-1,u;function m(_,N,C){var F=[];N=N==!0?{entropy:!0}:N||{};var E=L(w(N.entropy?[_,A(e)]:_==null?T():_,3),F),B=new b(F),q=function(){for(var P=B.g(i),te=s,ne=0;P<c;)P=(P+ne)*r,te*=r,ne=B.g(1);for(;P>=p;)P/=2,te/=2,ne>>>=1;return(P+ne)/te};return q.int32=function(){return B.g(4)|0},q.quick=function(){return B.g(4)/4294967296},q.double=q,L(A(B.S),e),(N.pass||C||function(P,te,ne,pe){return pe&&(pe.S&&y(pe,B),P.state=function(){return y(B,{})}),ne?(t[a]=P,te):P})(q,E,"global"in N?N.global:this==t,N.state)}t["seed"+a]=m;function b(_){var N,C=_.length,F=this,E=0,B=F.i=F.j=0,q=F.S=[];for(C||(_=[C++]);E<r;)q[E]=E++;for(E=0;E<r;E++)q[E]=q[B=l&B+_[E%C]+(N=q[E])],q[B]=N;(F.g=function(P){for(var te,ne=0,pe=F.i,Se=F.j,qe=F.S;P--;)te=qe[pe=l&pe+1],ne=ne*r+qe[l&(qe[pe]=qe[Se=l&Se+te])+(qe[Se]=te)];return F.i=pe,F.j=Se,ne})(r)}function y(_,N){return N.i=_.i,N.j=_.j,N.S=_.S.slice(),N}function w(_,N){var C=[],F=typeof _,E;if(N&&F=="object")for(E in _)try{C.push(w(_[E],N-1))}catch(B){}return C.length?C:F=="string"?_:_+"\0"}function L(_,N){for(var C=_+"",F,E=0;E<C.length;)N[l&E]=l&(F^=N[l&E]*19)+C.charCodeAt(E++);return A(N)}function T(){try{var _;return u&&(_=u.randomBytes)?_=_(r):(_=new Uint8Array(r),(n.crypto||n.msCrypto).getRandomValues(_)),A(_)}catch(F){var N=n.navigator,C=N&&N.plugins;return[+new Date(),n,C,n.screen,A(e)]}}function A(_){return String.fromCharCode.apply(0,_)}if(L(t.random(),e),typeof Zs=="object"&&Zs.exports){Zs.exports=m;try{u=Tf()}catch(_){}}else typeof define=="function"&&define.amd&&define(function(){return m})})([],Math)});var Nf=S((Ck,Af)=>{var Yv=df(),zv=ff(),Vv=bf(),Kv=yf(),Jv=Lf(),Xv=Sf(),An=If();An.alea=Yv;An.xor128=zv;An.xorwow=Vv;An.xorshift7=Kv;An.xor4096=Jv;An.tychei=Xv;Af.exports=An});var ee=S(Zv=>{Oe(Zv,{Abs:()=>Mi,Acos:()=>Ui,Acosh:()=>Wi,AdadeltaOptimizer:()=>kn,AdagradOptimizer:()=>On,AdamOptimizer:()=>En,AdamaxOptimizer:()=>Cn,Add:()=>Jn,AddN:()=>qi,All:()=>Wu,Any:()=>qu,ArgMax:()=>Bi,ArgMin:()=>Pi,Asin:()=>ji,Asinh:()=>Gi,Atan:()=>$i,Atan2:()=>Yi,Atanh:()=>Hi,AvgPool:()=>zi,AvgPool3D:()=>Vi,AvgPool3DBackprop:()=>Pu,AvgPoolBackprop:()=>Bu,BatchMatMul:()=>Ki,BatchToSpaceND:()=>Ji,BroadcastTo:()=>Xi,Cast:()=>Xn,Ceil:()=>Zi,ClipByValue:()=>Qi,Complex:()=>ju,Concat:()=>eo,Conv2D:()=>to,Conv2DBackpropFilter:()=>Gu,Conv2DBackpropInput:()=>no,Conv3D:()=>ro,Conv3DBackpropFilterV2:()=>$u,Conv3DBackpropInputV2:()=>Hu,Cos:()=>io,Cosh:()=>oo,CropAndResize:()=>Yu,Cumsum:()=>so,DataStorage:()=>RS,DepthToSpace:()=>zu,DepthwiseConv2dNative:()=>ao,DepthwiseConv2dNativeBackpropFilter:()=>Vu,DepthwiseConv2dNativeBackpropInput:()=>Ku,Diag:()=>Ju,Dilation2D:()=>co,Dilation2DBackpropFilter:()=>Zu,Dilation2DBackpropInput:()=>Xu,Div:()=>po,ENV:()=>Mh,Elu:()=>lo,EluGrad:()=>Qu,Environment:()=>Fh,Equal:()=>eh,Erf:()=>uo,Exp:()=>ho,Expm1:()=>mo,FFT:()=>th,Fill:()=>nh,FlipLeftRight:()=>rh,Floor:()=>fo,FloorDiv:()=>go,FromPixels:()=>ra,FusedBatchNorm:()=>bo,FusedConv2D:()=>oa,FusedDepthwiseConv2D:()=>sa,GatherNd:()=>ih,GatherV2:()=>wo,Greater:()=>oh,GreaterEqual:()=>yo,IFFT:()=>sh,Identity:()=>xo,Imag:()=>ah,IsFinite:()=>Lo,IsInf:()=>vo,IsNan:()=>So,KernelBackend:()=>CS,LRN:()=>No,LRNBackprop:()=>mh,Less:()=>ch,LessEqual:()=>ph,LinSpace:()=>lh,Log:()=>To,Log1p:()=>Io,LogSoftmax:()=>Ao,LogicalAnd:()=>uh,LogicalNot:()=>hh,LogicalOr:()=>dh,Max:()=>_o,MaxPool:()=>Co,MaxPool3D:()=>Eo,MaxPool3DBackprop:()=>gh,MaxPoolBackprop:()=>fh,MaxPoolWithArgmax:()=>bh,Maximum:()=>Ro,Mean:()=>oS,Min:()=>Oo,Minimum:()=>ko,Mod:()=>Do,MomentumOptimizer:()=>Rn,Multiply:()=>Fo,Negate:()=>Mo,NonMaxSuppressionV3:()=>yh,NonMaxSuppressionV4:()=>xh,NonMaxSuppressionV5:()=>Lh,NotEqual:()=>wh,OneHot:()=>Wo,OnesLike:()=>Uo,Optimizer:()=>Ze,PadV2:()=>qo,Pool:()=>sS,Pow:()=>Bo,Prelu:()=>Po,Prod:()=>vh,RMSPropOptimizer:()=>_n,Range:()=>Sh,Rank:()=>al,Real:()=>Th,Reciprocal:()=>jo,Reduction:()=>ue,Relu:()=>Go,Relu6:()=>zo,Reshape:()=>$o,ResizeBilinear:()=>Yo,ResizeBilinearGrad:()=>Ah,ResizeNearestNeighbor:()=>Ho,ResizeNearestNeighborGrad:()=>Ih,Reverse:()=>Vo,RotateWithOffset:()=>Oh,Round:()=>Ko,Rsqrt:()=>Jo,SGDOptimizer:()=>$t,ScatterNd:()=>Nh,SelectV2:()=>Xo,Selu:()=>Zo,Sigmoid:()=>rs,Sign:()=>ns,Sin:()=>es,Sinh:()=>ts,Slice:()=>Qo,Softmax:()=>ps,Softplus:()=>is,SpaceToBatchND:()=>as,SparseToDense:()=>Rh,SplitV:()=>cs,Sqrt:()=>os,Square:()=>_h,SquaredDifference:()=>ls,Step:()=>ys,StridedSlice:()=>Ch,Sub:()=>us,Sum:()=>ss,Tan:()=>hs,Tanh:()=>ds,Tensor:()=>de,TensorBuffer:()=>yr,Tile:()=>ms,TopK:()=>Eh,Transpose:()=>fs,Unpack:()=>gs,UnsortedSegmentSum:()=>bs,Variable:()=>Nn,ZerosLike:()=>ws,_FusedMatMul:()=>ia,abs:()=>_e,acos:()=>Mu,acosh:()=>Fu,add:()=>O,addN:()=>Du,addStrict:()=>eg,all:()=>ku,any:()=>Ou,argMax:()=>Eu,argMin:()=>Cu,asin:()=>Ru,asinh:()=>_u,atan:()=>Nu,atan2:()=>Au,atanh:()=>Iu,avgPool:()=>Di,avgPool3d:()=>Tu,backend:()=>AS,backend_util:()=>dg,basicLSTMCell:()=>Su,batchNorm:()=>Jt,batchNorm2d:()=>vu,batchNorm3d:()=>Lu,batchNorm4d:()=>xu,batchToSpaceND:()=>Kn,booleanMaskAsync:()=>Qf,broadcastTo:()=>Ir,browser:()=>Zf,buffer:()=>ke,cast:()=>R,ceil:()=>yu,clipByValue:()=>wu,clone:()=>At,complex:()=>Ge,concat:()=>re,concat1d:()=>bu,concat2d:()=>gu,concat3d:()=>fu,concat4d:()=>mu,conv1d:()=>du,conv2d:()=>It,conv2dTranspose:()=>hu,conv3d:()=>uu,conv3dTranspose:()=>lu,cos:()=>Vn,cosh:()=>ki,cosineWindow:()=>Lr,cumsum:()=>Oi,customGrad:()=>$e,deprecationWarn:()=>ge,depthToSpace:()=>pu,depthwiseConv2d:()=>zn,device_util:()=>ug,diag:()=>cu,dilation2d:()=>au,disableDeprecationWarnings:()=>dS,dispose:()=>be,disposeVariables:()=>mS,div:()=>M,divNoNan:()=>su,divStrict:()=>tg,dot:()=>ou,dropout:()=>Hf,elu:()=>Ei,enableDebugMode:()=>hS,enableProdMode:()=>uS,enclosingPowerOfTwo:()=>Qs,engine:()=>fS,env:()=>se,equal:()=>Kt,equalStrict:()=>Yf,erf:()=>iu,exp:()=>Ne,expandDims:()=>Ke,expm1:()=>ru,eye:()=>Ci,fft:()=>Yn,fill:()=>Hn,findBackend:()=>SS,findBackendFactory:()=>TS,floor:()=>$n,floorDiv:()=>Ri,fused:()=>nu,gather:()=>Gn,gatherND:()=>$f,gather_util:()=>Gf,getBackend:()=>LS,getGradient:()=>ta,getKernel:()=>Fi,getKernelsForBackend:()=>na,grad:()=>aS,grads:()=>cS,greater:()=>Ve,greaterEqual:()=>ct,greaterEqualStrict:()=>zf,greaterStrict:()=>Vf,ifft:()=>cn,imag:()=>Vt,image:()=>Wf,inTopKAsync:()=>jf,io:()=>lg,irfft:()=>_i,isFinite:()=>tu,isInf:()=>eu,isNaN:()=>Ql,keep:()=>Dh,kernel_impls:()=>hg,leakyRelu:()=>Zl,less:()=>jn,lessEqual:()=>at,lessEqualStrict:()=>Kf,lessStrict:()=>Jf,linalg:()=>qf,linspace:()=>Xl,localResponseNormalization:()=>Jl,log:()=>gt,log1p:()=>Ni,logSigmoid:()=>Kl,logSoftmax:()=>Vl,logSumExp:()=>Ai,logicalAnd:()=>ft,logicalNot:()=>Pn,logicalOr:()=>Ii,logicalXor:()=>zl,losses:()=>Bf,matMul:()=>K,math:()=>pg,max:()=>st,maxPool:()=>Ti,maxPool3d:()=>Yl,maxPoolWithArgmax:()=>Hl,maximum:()=>Tt,maximumStrict:()=>ng,mean:()=>an,memory:()=>gS,min:()=>sn,minimum:()=>Bn,minimumStrict:()=>rg,mod:()=>Si,modStrict:()=>ig,moments:()=>$l,movingAverage:()=>Pf,mul:()=>v,mulStrict:()=>og,multiRNNCell:()=>Gl,multinomial:()=>jl,neg:()=>Q,nextFrame:()=>_S,norm:()=>Tr,notEqual:()=>qn,notEqualStrict:()=>Xf,oneHot:()=>Sr,ones:()=>mt,onesLike:()=>ea,op:()=>d,outerProduct:()=>Pl,pad:()=>ot,pad1d:()=>Bl,pad2d:()=>ql,pad3d:()=>Wl,pad4d:()=>Ul,pool:()=>Ml,pow:()=>je,powStrict:()=>sg,prelu:()=>vi,print:()=>Fl,prod:()=>Dl,profile:()=>bS,rand:()=>kl,randomGamma:()=>Ol,randomNormal:()=>El,randomUniform:()=>Li,range:()=>vr,ready:()=>xS,real:()=>kt,reciprocal:()=>Cl,registerBackend:()=>IS,registerGradient:()=>Uu,registerKernel:()=>nS,relu:()=>zt,relu6:()=>xi,removeBackend:()=>vS,reshape:()=>x,reverse:()=>Pe,reverse1d:()=>Rl,reverse2d:()=>_l,reverse3d:()=>Nl,reverse4d:()=>Al,rfft:()=>Wn,round:()=>Il,rsqrt:()=>yi,scalar:()=>k,scatterND:()=>Ff,scatter_util:()=>Df,selu:()=>Tl,separableConv2d:()=>Sl,serialization:()=>Ef,setBackend:()=>yS,setPlatform:()=>NS,setdiff1dAsync:()=>vl,sigmoid:()=>St,sign:()=>Ll,signal:()=>Uf,sin:()=>wi,sinh:()=>bi,slice:()=>Y,slice1d:()=>xl,slice2d:()=>yl,slice3d:()=>wl,slice4d:()=>bl,slice_util:()=>kf,softmax:()=>gl,softplus:()=>gi,spaceToBatchND:()=>Un,sparseToDense:()=>Of,spectral:()=>Mf,split:()=>Ot,sqrt:()=>le,square:()=>G,squaredDifference:()=>Mn,squaredDifferenceStrict:()=>ag,squeeze:()=>Fn,stack:()=>dt,step:()=>vt,stridedSlice:()=>fl,sub:()=>D,subStrict:()=>cg,sum:()=>U,sumOutType:()=>eS,tan:()=>ml,tanh:()=>xr,tensor:()=>Qe,tensor1d:()=>ht,tensor2d:()=>Yt,tensor3d:()=>fi,tensor4d:()=>dl,tensor5d:()=>hl,tensor6d:()=>ul,tensor_util:()=>Cf,test_util:()=>Rf,tidy:()=>ce,tile:()=>Ht,time:()=>wS,topk:()=>ll,train:()=>tS,transpose:()=>X,truncatedNormal:()=>pl,unregisterGradient:()=>iS,unregisterKernel:()=>rS,unsortedSegmentSum:()=>mi,unstack:()=>Dn,upcastType:()=>wr,util:()=>_f,valueAndGrad:()=>pS,valueAndGrads:()=>lS,variable:()=>cl,variableGrads:()=>kh,version_core:()=>Qv,where:()=>Ae,whereAsync:()=>di,zeros:()=>Ue,zerosLike:()=>W});});var Uh=S(aa=>{"use strict";Object.defineProperty(aa,"__esModule",{value:!0});aa.drawContour=void 0;function ES(e,t,n=!1){if(e.beginPath(),t.slice(1).forEach(({x:r,y:i},o)=>{const a=t[o];e.moveTo(a.x,a.y),e.lineTo(r,i)}),n){const r=t[t.length-1],i=t[0];if(!r||!i)return;e.moveTo(r.x,r.y),e.lineTo(i.x,i.y)}e.stroke()}aa.drawContour=ES});var Ar=S(ca=>{"use strict";Object.defineProperty(ca,"__esModule",{value:!0});ca.Dimensions=void 0;const mg=me();class fg{constructor(e,t){if(!mg.isValidNumber(e)||!mg.isValidNumber(t))throw new Error(`Dimensions.constructor - expected width and height to be valid numbers, instead have ${JSON.stringify({width:e,height:t})}`);this._width=e,this._height=t}get width(){return this._width}get height(){return this._height}reverse(){return new fg(1/this.width,1/this.height)}}ca.Dimensions=fg});var me=S(oe=>{"use strict";Object.defineProperty(oe,"__esModule",{value:!0});oe.isValidProbablitiy=oe.isValidNumber=oe.range=oe.getCenterPoint=oe.computeReshapedDimensions=oe.isDimensions=oe.round=oe.isEven=oe.isFloat=oe.isTensor4D=oe.isTensor3D=oe.isTensor2D=oe.isTensor1D=oe.isTensor=void 0;const OS=ee(),gg=pt(),kS=Ar();function xs(e,t){return e instanceof OS.Tensor&&e.shape.length===t}oe.isTensor=xs;function DS(e){return xs(e,1)}oe.isTensor1D=DS;function FS(e){return xs(e,2)}oe.isTensor2D=FS;function MS(e){return xs(e,3)}oe.isTensor3D=MS;function US(e){return xs(e,4)}oe.isTensor4D=US;function WS(e){return e%1!==0}oe.isFloat=WS;function qS(e){return e%2===0}oe.isEven=qS;function BS(e,t=2){const n=Math.pow(10,t);return Math.floor(e*n)/n}oe.round=BS;function PS(e){return e&&e.width&&e.height}oe.isDimensions=PS;function jS({width:e,height:t},n){const r=n/Math.max(t,e);return new kS.Dimensions(Math.round(e*r),Math.round(t*r))}oe.computeReshapedDimensions=jS;function GS(e){return e.reduce((t,n)=>t.add(n),new gg.Point(0,0)).div(new gg.Point(e.length,e.length))}oe.getCenterPoint=GS;function $S(e,t,n){return Array(e).fill(0).map((r,i)=>t+i*n)}oe.range=$S;function bg(e){return!!e&&e!==Infinity&&e!==-Infinity&&!isNaN(e)||e===0}oe.isValidNumber=bg;function HS(e){return bg(e)&&0<=e&&e<=1}oe.isValidProbablitiy=HS});var la=S(pa=>{"use strict";Object.defineProperty(pa,"__esModule",{value:!0});pa.Point=void 0;class Zn{constructor(e,t){this._x=e,this._y=t}get x(){return this._x}get y(){return this._y}add(e){return new Zn(this.x+e.x,this.y+e.y)}sub(e){return new Zn(this.x-e.x,this.y-e.y)}mul(e){return new Zn(this.x*e.x,this.y*e.y)}div(e){return new Zn(this.x/e.x,this.y/e.y)}abs(){return new Zn(Math.abs(this.x),Math.abs(this.y))}magnitude(){return Math.sqrt(Math.pow(this.x,2)+Math.pow(this.y,2))}floor(){return new Zn(Math.floor(this.x),Math.floor(this.y))}}pa.Point=Zn});var Qn=S(ua=>{"use strict";Object.defineProperty(ua,"__esModule",{value:!0});ua.Box=void 0;const Ls=me(),ha=la();class Nt{constructor(e,t=!0){const n=e||{},r=[n.left,n.top,n.right,n.bottom].every(Ls.isValidNumber),i=[n.x,n.y,n.width,n.height].every(Ls.isValidNumber);if(!i&&!r)throw new Error(`Box.constructor - expected box to be IBoundingBox | IRect, instead have ${JSON.stringify(n)}`);const[o,a,s,c]=i?[n.x,n.y,n.width,n.height]:[n.left,n.top,n.right-n.left,n.bottom-n.top];Nt.assertIsValidBox({x:o,y:a,width:s,height:c},"Box.constructor",t),this._x=o,this._y=a,this._width=s,this._height=c}static isRect(e){return!!e&&[e.x,e.y,e.width,e.height].every(Ls.isValidNumber)}static assertIsValidBox(e,t,n=!1){if(!Nt.isRect(e))throw new Error(`${t} - invalid box: ${JSON.stringify(e)}, expected object with properties x, y, width, height`);if(!n&&(e.width<0||e.height<0))throw new Error(`${t} - width (${e.width}) and height (${e.height}) must be positive numbers`)}get x(){return this._x}get y(){return this._y}get width(){return this._width}get height(){return this._height}get left(){return this.x}get top(){return this.y}get right(){return this.x+this.width}get bottom(){return this.y+this.height}get area(){return this.width*this.height}get topLeft(){return new ha.Point(this.left,this.top)}get topRight(){return new ha.Point(this.right,this.top)}get bottomLeft(){return new ha.Point(this.left,this.bottom)}get bottomRight(){return new ha.Point(this.right,this.bottom)}round(){const[e,t,n,r]=[this.x,this.y,this.width,this.height].map(i=>Math.round(i));return new Nt({x:e,y:t,width:n,height:r})}floor(){const[e,t,n,r]=[this.x,this.y,this.width,this.height].map(i=>Math.floor(i));return new Nt({x:e,y:t,width:n,height:r})}toSquare(){let{x:e,y:t,width:n,height:r}=this;const i=Math.abs(n-r);return n<r&&(e-=i/2,n+=i),r<n&&(t-=i/2,r+=i),new Nt({x:e,y:t,width:n,height:r})}rescale(e){const t=Ls.isDimensions(e)?e.width:e,n=Ls.isDimensions(e)?e.height:e;return new Nt({x:this.x*t,y:this.y*n,width:this.width*t,height:this.height*n})}pad(e,t){let[n,r,i,o]=[this.x-e/2,this.y-t/2,this.width+e,this.height+t];return new Nt({x:n,y:r,width:i,height:o})}clipAtImageBorders(e,t){const{x:n,y:r,right:i,bottom:o}=this,a=Math.max(n,0),s=Math.max(r,0),c=i-a,p=o-s,l=Math.min(c,e-a),u=Math.min(p,t-s);return new Nt({x:a,y:s,width:l,height:u}).floor()}shift(e,t){const{width:n,height:r}=this,i=this.x+e,o=this.y+t;return new Nt({x:i,y:o,width:n,height:r})}padAtBorders(e,t){const n=this.width+1,r=this.height+1;let i=1,o=1,a=n,s=r,c=this.left,p=this.top,l=this.right,u=this.bottom;return l>t&&(a=-l+t+n,l=t),u>e&&(s=-u+e+r,u=e),c<1&&(s=2-c,c=1),p<1&&(s=2-p,p=1),{dy:o,edy:s,dx:i,edx:a,y:p,ey:u,x:c,ex:l,w:n,h:r}}calibrate(e){return new Nt({left:this.left+e.left*this.width,top:this.top+e.top*this.height,right:this.right+e.right*this.width,bottom:this.bottom+e.bottom*this.height}).toSquare().round()}}ua.Box=Nt});var Wh=S(da=>{"use strict";Object.defineProperty(da,"__esModule",{value:!0});da.BoundingBox=void 0;const YS=Qn();class zS extends YS.Box{constructor(e,t,n,r,i=!1){super({left:e,top:t,right:n,bottom:r},i)}}da.BoundingBox=zS});var fa=S(ma=>{"use strict";Object.defineProperty(ma,"__esModule",{value:!0});ma.ObjectDetection=void 0;const wg=Qn(),VS=Ar();class yg{constructor(e,t,n,r,i){this._imageDims=new VS.Dimensions(i.width,i.height),this._score=e,this._classScore=t,this._className=n,this._box=new wg.Box(r).rescale(this._imageDims)}get score(){return this._score}get classScore(){return this._classScore}get className(){return this._className}get box(){return this._box}get imageDims(){return this._imageDims}get imageWidth(){return this.imageDims.width}get imageHeight(){return this.imageDims.height}get relativeBox(){return new wg.Box(this._box).rescale(this.imageDims.reverse())}forSize(e,t){return new yg(this.score,this.classScore,this.className,this.relativeBox,{width:e,height:t})}}ma.ObjectDetection=yg});var Xt=S(ga=>{"use strict";Object.defineProperty(ga,"__esModule",{value:!0});ga.FaceDetection=void 0;const KS=fa();class xg extends KS.ObjectDetection{constructor(e,t,n){super(e,e,"",t,n)}forSize(e,t){const{score:n,relativeBox:r,imageDims:i}=super.forSize(e,t);return new xg(n,r,i)}}ga.FaceDetection=xg});var qh=S(ba=>{"use strict";Object.defineProperty(ba,"__esModule",{value:!0});ba.iou=void 0;function JS(e,t,n=!0){const r=Math.max(0,Math.min(e.right,t.right)-Math.max(e.left,t.left)),i=Math.max(0,Math.min(e.bottom,t.bottom)-Math.max(e.top,t.top)),o=r*i;return n?o/(e.area+t.area-o):o/Math.min(e.area,t.area)}ba.iou=JS});var Lg=S(wa=>{"use strict";Object.defineProperty(wa,"__esModule",{value:!0});wa.minBbox=void 0;const XS=pt();function ZS(e){const t=e.map(s=>s.x),n=e.map(s=>s.y),r=t.reduce((s,c)=>c<s?c:s,Infinity),i=n.reduce((s,c)=>c<s?c:s,Infinity),o=t.reduce((s,c)=>s<c?c:s,0),a=n.reduce((s,c)=>s<c?c:s,0);return new XS.BoundingBox(r,i,o,a)}wa.minBbox=ZS});var Bh=S(ya=>{"use strict";Object.defineProperty(ya,"__esModule",{value:!0});ya.nonMaxSuppression=void 0;const QS=qh();function eT(e,t,n,r=!0){let i=t.map((a,s)=>({score:a,boxIndex:s})).sort((a,s)=>a.score-s.score).map(a=>a.boxIndex);const o=[];for(;i.length>0;){const a=i.pop();o.push(a);const s=i,c=[];for(let p=0;p<s.length;p++){const l=s[p],u=e[a],m=e[l];c.push(QS.iou(u,m,r))}i=i.filter((p,l)=>c[l]<=n)}return o}ya.nonMaxSuppression=eT});var Ph=S(xa=>{"use strict";Object.defineProperty(xa,"__esModule",{value:!0});xa.normalize=void 0;const Nr=ee();function tT(e,t){return Nr.tidy(()=>{const[n,r,i]=t,o=Nr.fill([...e.shape.slice(0,3),1],n),a=Nr.fill([...e.shape.slice(0,3),1],r),s=Nr.fill([...e.shape.slice(0,3),1],i),c=Nr.concat([o,a,s],3);return Nr.sub(e,c)})}xa.normalize=tT});var Gh=S(La=>{"use strict";Object.defineProperty(La,"__esModule",{value:!0});La.padToSquare=void 0;const jh=ee();function nT(e,t=!1){return jh.tidy(()=>{const[n,r]=e.shape.slice(1);if(n===r)return e;const i=Math.abs(n-r),o=Math.round(i*(t?.5:1)),a=n>r?2:1,s=m=>{const b=e.shape.slice();return b[a]=m,jh.fill(b,0)},c=s(o),p=i-c.shape[a],l=t&&p?s(p):null,u=[l,e,c].filter(m=>!!m).map(m=>m.toFloat());return jh.concat(u,a)})}La.padToSquare=nT});var vg=S(va=>{"use strict";Object.defineProperty(va,"__esModule",{value:!0});va.shuffleArray=void 0;function rT(e){const t=e.slice();for(let n=t.length-1;n>0;n--){const r=Math.floor(Math.random()*(n+1)),i=t[n];t[n]=t[r],t[r]=i}return t}va.shuffleArray=rT});var pn=S(bt=>{"use strict";Object.defineProperty(bt,"__esModule",{value:!0});bt.inverseSigmoid=bt.sigmoid=void 0;const _r=Be();_r.__exportStar(qh(),bt);_r.__exportStar(Lg(),bt);_r.__exportStar(Bh(),bt);_r.__exportStar(Ph(),bt);_r.__exportStar(Gh(),bt);_r.__exportStar(vg(),bt);function iT(e){return 1/(1+Math.exp(-e))}bt.sigmoid=iT;function oT(e){return Math.log(e/(1-e))}bt.inverseSigmoid=oT});var $h=S(Sa=>{"use strict";Object.defineProperty(Sa,"__esModule",{value:!0});Sa.Rect=void 0;const sT=Qn();class aT extends sT.Box{constructor(e,t,n,r,i=!1){super({x:e,y:t,width:n,height:r},i)}}Sa.Rect=aT});var er=S(Ta=>{"use strict";Object.defineProperty(Ta,"__esModule",{value:!0});Ta.FaceLandmarks=void 0;const cT=pn(),pT=me(),lT=Qn(),uT=Ar(),hT=Xt(),vs=la(),dT=$h(),mT=.5,fT=.43,gT=.45;class bT{constructor(e,t,n=new vs.Point(0,0)){const{width:r,height:i}=t;this._imgDims=new uT.Dimensions(r,i),this._shift=n,this._positions=e.map(o=>o.mul(new vs.Point(r,i)).add(n))}get shift(){return new vs.Point(this._shift.x,this._shift.y)}get imageWidth(){return this._imgDims.width}get imageHeight(){return this._imgDims.height}get positions(){return this._positions}get relativePositions(){return this._positions.map(e=>e.sub(this._shift).div(new vs.Point(this.imageWidth,this.imageHeight)))}forSize(e,t){return new this.constructor(this.relativePositions,{width:e,height:t})}shiftBy(e,t){return new this.constructor(this.relativePositions,this._imgDims,new vs.Point(e,t))}shiftByPoint(e){return this.shiftBy(e.x,e.y)}align(e,t={}){if(e){const i=e instanceof hT.FaceDetection?e.box.floor():new lT.Box(e);return this.shiftBy(i.x,i.y).align(null,t)}const{useDlibAlignment:n,minBoxPadding:r}=Object.assign({},{useDlibAlignment:!1,minBoxPadding:.2},t);return n?this.alignDlib():this.alignMinBbox(r)}alignDlib(){const e=this.getRefPointsForAlignment(),[t,n,r]=e,i=l=>r.sub(l).magnitude(),o=(i(t)+i(n))/2,a=Math.floor(o/gT),s=pT.getCenterPoint(e),c=Math.floor(Math.max(0,s.x-mT*a)),p=Math.floor(Math.max(0,s.y-fT*a));return new dT.Rect(c,p,Math.min(a,this.imageWidth+c),Math.min(a,this.imageHeight+p))}alignMinBbox(e){const t=cT.minBbox(this.positions);return t.pad(t.width*e,t.height*e)}getRefPointsForAlignment(){throw new Error("getRefPointsForAlignment not implemented by base class")}}Ta.FaceLandmarks=bT});var Sg=S(Ia=>{"use strict";Object.defineProperty(Ia,"__esModule",{value:!0});Ia.FaceLandmarks5=void 0;const wT=me(),yT=er();class xT extends yT.FaceLandmarks{getRefPointsForAlignment(){const e=this.positions;return[e[0],e[1],wT.getCenterPoint([e[3],e[4]])]}}Ia.FaceLandmarks5=xT});var Na=S(Aa=>{"use strict";Object.defineProperty(Aa,"__esModule",{value:!0});Aa.FaceLandmarks68=void 0;const LT=me(),vT=er();class ST extends vT.FaceLandmarks{getJawOutline(){return this.positions.slice(0,17)}getLeftEyeBrow(){return this.positions.slice(17,22)}getRightEyeBrow(){return this.positions.slice(22,27)}getNose(){return this.positions.slice(27,36)}getLeftEye(){return this.positions.slice(36,42)}getRightEye(){return this.positions.slice(42,48)}getMouth(){return this.positions.slice(48,68)}getRefPointsForAlignment(){return[this.getLeftEye(),this.getRightEye(),this.getMouth()].map(LT.getCenterPoint)}}Aa.FaceLandmarks68=ST});var Hh=S(_a=>{"use strict";Object.defineProperty(_a,"__esModule",{value:!0});_a.FaceMatch=void 0;const TT=me();class IT{constructor(e,t){this._label=e,this._distance=t}get label(){return this._label}get distance(){return this._distance}toString(e=!0){return`${this.label}${e?` (${TT.round(this.distance)})`:""}`}}_a.FaceMatch=IT});var Yh=S(Ra=>{"use strict";Object.defineProperty(Ra,"__esModule",{value:!0});Ra.LabeledBox=void 0;const AT=me(),Tg=Qn();class NT extends Tg.Box{constructor(e,t){super(e);this._label=t}static assertIsValidLabeledBox(e,t){if(Tg.Box.assertIsValidBox(e,t),!AT.isValidNumber(e.label))throw new Error(`${t} - expected property label (${e.label}) to be a number`)}get label(){return this._label}}Ra.LabeledBox=NT});var zh=S(Ca=>{"use strict";Object.defineProperty(Ca,"__esModule",{value:!0});Ca.LabeledFaceDescriptors=void 0;class Ig{constructor(e,t){if(!(typeof e=="string"))throw new Error("LabeledFaceDescriptors - constructor expected label to be a string");if(!Array.isArray(t)||t.some(n=>!(n instanceof Float32Array)))throw new Error("LabeledFaceDescriptors - constructor expected descriptors to be an array of Float32Array");this._label=e,this._descriptors=t}get label(){return this._label}get descriptors(){return this._descriptors}toJSON(){return{label:this.label,descriptors:this.descriptors.map(e=>Array.from(e))}}static fromJSON(e){const t=e.descriptors.map(n=>new Float32Array(n));return new Ig(e.label,t)}}Ca.LabeledFaceDescriptors=Ig});var _g=S(Ea=>{"use strict";Object.defineProperty(Ea,"__esModule",{value:!0});Ea.PredictedBox=void 0;const Ag=me(),Ng=Yh();class _T extends Ng.LabeledBox{constructor(e,t,n,r){super(e,t);this._score=n,this._classScore=r}static assertIsValidPredictedBox(e,t){if(Ng.LabeledBox.assertIsValidLabeledBox(e,t),!Ag.isValidProbablitiy(e.score)||!Ag.isValidProbablitiy(e.classScore))throw new Error(`${t} - expected properties score (${e.score}) and (${e.classScore}) to be a number between [0, 1]`)}get score(){return this._score}get classScore(){return this._classScore}}Ea.PredictedBox=_T});var pt=S(He=>{"use strict";Object.defineProperty(He,"__esModule",{value:!0});const et=Be();et.__exportStar(Wh(),He);et.__exportStar(Qn(),He);et.__exportStar(Ar(),He);et.__exportStar(Xt(),He);et.__exportStar(er(),He);et.__exportStar(Sg(),He);et.__exportStar(Na(),He);et.__exportStar(Hh(),He);et.__exportStar(Yh(),He);et.__exportStar(zh(),He);et.__exportStar(fa(),He);et.__exportStar(la(),He);et.__exportStar(_g(),He);et.__exportStar($h(),He)});var tr=S(Rr=>{"use strict";Object.defineProperty(Rr,"__esModule",{value:!0});Rr.extendWithFaceDetection=Rr.isWithFaceDetection=void 0;const RT=Xt();function CT(e){return e.detection instanceof RT.FaceDetection}Rr.isWithFaceDetection=CT;function ET(e,t){const n={detection:t};return Object.assign({},e,n)}Rr.extendWithFaceDetection=ET});var Rg=S(Oa=>{"use strict";Object.defineProperty(Oa,"__esModule",{value:!0});Oa.createBrowserEnv=void 0;function OT(){const e=window.fetch||function(){throw new Error("fetch - missing fetch implementation for browser environment")},t=function(){throw new Error("readFile - filesystem not available for browser environment")};return{Canvas:HTMLCanvasElement,CanvasRenderingContext2D,Image:HTMLImageElement,ImageData,Video:HTMLVideoElement,createCanvasElement:()=>document.createElement("canvas"),createImageElement:()=>document.createElement("img"),fetch:e,readFile:t}}Oa.createBrowserEnv=OT});var Vh=S(ka=>{"use strict";Object.defineProperty(ka,"__esModule",{value:!0});ka.createFileSystem=void 0;function kT(e){let t="";if(!e)try{e=require("fs")}catch(r){t=r.toString()}const n=e?function(r){return new Promise((i,o)=>{e.readFile(r,function(a,s){return a?o(a):i(s)})})}:function(){throw new Error(`readFile - failed to require fs in nodejs environment with error: ${t}`)};return{readFile:n}}ka.createFileSystem=kT});var Cg=S(Da=>{"use strict";Object.defineProperty(Da,"__esModule",{value:!0});Da.createNodejsEnv=void 0;const DT=Vh();function FT(){const e=global.Canvas||global.HTMLCanvasElement,t=global.Image||global.HTMLImageElement,n=function(){if(e)return new e();throw new Error("createCanvasElement - missing Canvas implementation for nodejs environment")},r=function(){if(t)return new t();throw new Error("createImageElement - missing Image implementation for nodejs environment")},i=global.fetch||function(){throw new Error("fetch - missing fetch implementation for nodejs environment")},o=DT.createFileSystem();return{Canvas:e||class{},CanvasRenderingContext2D:global.CanvasRenderingContext2D||class{},Image:t||class{},ImageData:global.ImageData||class{},Video:global.HTMLVideoElement||class{},createCanvasElement:n,createImageElement:r,fetch:i,...o}}Da.createNodejsEnv=FT});var Eg=S(Fa=>{"use strict";Object.defineProperty(Fa,"__esModule",{value:!0});Fa.isBrowser=void 0;function MT(){return typeof window=="object"&&typeof document!="undefined"&&typeof HTMLImageElement!="undefined"&&typeof HTMLCanvasElement!="undefined"&&typeof HTMLVideoElement!="undefined"&&typeof ImageData!="undefined"&&typeof CanvasRenderingContext2D!="undefined"}Fa.isBrowser=MT});var kg=S((Ma,Og)=>{"use strict";Object.defineProperty(Ma,"__esModule",{value:!0});Ma.isNodejs=void 0;function UT(){return typeof global=="object"&&!0&&typeof Og!="undefined"&&typeof process!="undefined"&&!!process.version}Ma.isNodejs=UT});var Fg=S(Dg=>{"use strict";Object.defineProperty(Dg,"__esModule",{value:!0})});var Ye=S(Ss=>{"use strict";Object.defineProperty(Ss,"__esModule",{value:!0});Ss.env=void 0;const WT=Be(),Mg=Rg(),qT=Vh(),Ug=Cg(),Wg=Eg(),qg=kg();let Re;function BT(){if(!Re)throw new Error("getEnv - environment is not defined, check isNodejs() and isBrowser()");return Re}function Kh(e){Re=e}function Jh(){if(Wg.isBrowser())return Kh(Mg.createBrowserEnv());if(qg.isNodejs())return Kh(Ug.createNodejsEnv())}function PT(e){if(Re||Jh(),!Re)throw new Error("monkeyPatch - environment is not defined, check isNodejs() and isBrowser()");const{Canvas:t=Re.Canvas,Image:n=Re.Image}=e;Re.Canvas=t,Re.Image=n,Re.createCanvasElement=e.createCanvasElement||(()=>new t()),Re.createImageElement=e.createImageElement||(()=>new n()),Re.ImageData=e.ImageData||Re.ImageData,Re.Video=e.Video||Re.Video,Re.fetch=e.fetch||Re.fetch,Re.readFile=e.readFile||Re.readFile}Ss.env={getEnv:BT,setEnv:Kh,initialize:Jh,createBrowserEnv:Mg.createBrowserEnv,createFileSystem:qT.createFileSystem,createNodejsEnv:Ug.createNodejsEnv,monkeyPatch:PT,isBrowser:Wg.isBrowser,isNodejs:qg.isNodejs};Jh();WT.__exportStar(Fg(),Ss)});var Ts=S(Ua=>{"use strict";Object.defineProperty(Ua,"__esModule",{value:!0});Ua.resolveInput=void 0;const jT=Ye();function GT(e){return!jT.env.isNodejs()&&typeof e=="string"?document.getElementById(e):e}Ua.resolveInput=GT});var ln=S(Wa=>{"use strict";Object.defineProperty(Wa,"__esModule",{value:!0});Wa.getContext2dOrThrow=void 0;const $T=Ye(),HT=Ts();function YT(e){const{Canvas:t,CanvasRenderingContext2D:n}=$T.env.getEnv();if(e instanceof n)return e;const r=HT.resolveInput(e);if(!(r instanceof t))throw new Error("resolveContext2d - expected canvas to be of instance of Canvas");const i=r.getContext("2d");if(!i)throw new Error("resolveContext2d - canvas 2d context is null");return i}Wa.getContext2dOrThrow=YT});var qa=S(Zt=>{"use strict";Object.defineProperty(Zt,"__esModule",{value:!0});Zt.DrawTextField=Zt.DrawTextFieldOptions=Zt.AnchorPosition=void 0;const zT=ln(),VT=Ts();var Cr;(function(e){e.TOP_LEFT="TOP_LEFT",e.TOP_RIGHT="TOP_RIGHT",e.BOTTOM_LEFT="BOTTOM_LEFT",e.BOTTOM_RIGHT="BOTTOM_RIGHT"})(Cr=Zt.AnchorPosition||(Zt.AnchorPosition={}));class Bg{constructor(e={}){const{anchorPosition:t,backgroundColor:n,fontColor:r,fontSize:i,fontStyle:o,padding:a}=e;this.anchorPosition=t||Cr.TOP_LEFT,this.backgroundColor=n||"rgba(0, 0, 0, 0.5)",this.fontColor=r||"rgba(255, 255, 255, 1)",this.fontSize=i||14,this.fontStyle=o||"Georgia",this.padding=a||4}}Zt.DrawTextFieldOptions=Bg;class Pg{constructor(e,t,n={}){this.text=typeof e=="string"?[e]:e instanceof Pg?e.text:e,this.anchor=t,this.options=new Bg(n)}measureWidth(e){const{padding:t}=this.options;return this.text.map(n=>e.measureText(n).width).reduce((n,r)=>n<r?r:n,0)+2*t}measureHeight(){const{fontSize:e,padding:t}=this.options;return this.text.length*e+2*t}getUpperLeft(e,t){const{anchorPosition:n}=this.options,r=n===Cr.BOTTOM_RIGHT||n===Cr.TOP_RIGHT,i=n===Cr.BOTTOM_LEFT||n===Cr.BOTTOM_RIGHT,o=this.measureWidth(e),a=this.measureHeight(),s=r?this.anchor.x-o:this.anchor.x,c=i?this.anchor.y-a:this.anchor.y;if(t){const{width:p,height:l}=t,u=Math.max(Math.min(s,p-o),0),m=Math.max(Math.min(c,l-a),0);return{x:u,y:m}}return{x:s,y:c}}draw(e){const t=VT.resolveInput(e),n=zT.getContext2dOrThrow(t),{backgroundColor:r,fontColor:i,fontSize:o,fontStyle:a,padding:s}=this.options;n.font=`${o}px ${a}`;const c=this.measureWidth(n),p=this.measureHeight();n.fillStyle=r;const l=this.getUpperLeft(n,t);n.fillRect(l.x,l.y,c,p),n.fillStyle=i,this.text.forEach((u,m)=>{const b=s+l.x,y=s+l.y+(m+1)*o;n.fillText(u,b,y)})}}Zt.DrawTextField=Pg});var Zh=S(Er=>{"use strict";Object.defineProperty(Er,"__esModule",{value:!0});Er.DrawBox=Er.DrawBoxOptions=void 0;const KT=pt(),JT=ln(),Xh=qa();class jg{constructor(e={}){const{boxColor:t,lineWidth:n,label:r,drawLabelOptions:i}=e;this.boxColor=t||"rgba(0, 0, 255, 1)",this.lineWidth=n||2,this.label=r;const o={anchorPosition:Xh.AnchorPosition.BOTTOM_LEFT,backgroundColor:this.boxColor};this.drawLabelOptions=new Xh.DrawTextFieldOptions(Object.assign({},o,i))}}Er.DrawBoxOptions=jg;class XT{constructor(e,t={}){this.box=new KT.Box(e),this.options=new jg(t)}draw(e){const t=JT.getContext2dOrThrow(e),{boxColor:n,lineWidth:r}=this.options,{x:i,y:o,width:a,height:s}=this.box;t.strokeStyle=n,t.lineWidth=r,t.strokeRect(i,o,a,s);const{label:c}=this.options;c&&new Xh.DrawTextField([c],{x:i-r/2,y:o},this.options.drawLabelOptions).draw(e)}}Er.DrawBox=XT});var Hg=S(Ba=>{"use strict";Object.defineProperty(Ba,"__esModule",{value:!0});Ba.drawDetections=void 0;const ZT=pt(),Gg=Xt(),$g=tr(),QT=me(),eI=Zh();function tI(e,t){const n=Array.isArray(t)?t:[t];n.forEach(r=>{const i=r instanceof Gg.FaceDetection?r.score:$g.isWithFaceDetection(r)?r.detection.score:void 0,o=r instanceof Gg.FaceDetection?r.box:$g.isWithFaceDetection(r)?r.detection.box:new ZT.Box(r),a=i?`${QT.round(i)}`:void 0;new eI.DrawBox(o,{label:a}).draw(e)})}Ba.drawDetections=tI});var ja=S(Pa=>{"use strict";Object.defineProperty(Pa,"__esModule",{value:!0});Pa.isMediaLoaded=void 0;const nI=Ye();function rI(e){const{Image:t,Video:n}=nI.env.getEnv();return e instanceof t&&e.complete||e instanceof n&&e.readyState>=3}Pa.isMediaLoaded=rI});var Qh=S(Ga=>{"use strict";Object.defineProperty(Ga,"__esModule",{value:!0});Ga.awaitMediaLoaded=void 0;const iI=Ye(),oI=ja();function sI(e){return new Promise((t,n)=>{if(e instanceof iI.env.getEnv().Canvas||oI.isMediaLoaded(e))return t();function r(o){if(!o.currentTarget)return;o.currentTarget.removeEventListener("load",r),o.currentTarget.removeEventListener("error",i),t(o)}function i(o){if(!o.currentTarget)return;o.currentTarget.removeEventListener("load",r),o.currentTarget.removeEventListener("error",i),n(o)}e.addEventListener("load",r),e.addEventListener("error",i)})}Ga.awaitMediaLoaded=sI});var ed=S($a=>{"use strict";Object.defineProperty($a,"__esModule",{value:!0});$a.bufferToImage=void 0;const aI=Ye();function cI(e){return new Promise((t,n)=>{if(!(e instanceof Blob))return n("bufferToImage - expected buf to be of type: Blob");const r=new FileReader();r.onload=()=>{if(typeof r.result!="string")return n("bufferToImage - expected reader.result to be a string, in onload");const i=aI.env.getEnv().createImageElement();i.onload=()=>t(i),i.onerror=n,i.src=r.result},r.onerror=n,r.readAsDataURL(e)})}$a.bufferToImage=cI});var Is=S(Ha=>{"use strict";Object.defineProperty(Ha,"__esModule",{value:!0});Ha.getMediaDimensions=void 0;const td=Ar(),pI=Ye();function lI(e){const{Image:t,Video:n}=pI.env.getEnv();return e instanceof t?new td.Dimensions(e.naturalWidth,e.naturalHeight):e instanceof n?new td.Dimensions(e.videoWidth,e.videoHeight):new td.Dimensions(e.width,e.height)}Ha.getMediaDimensions=lI});var As=S(Or=>{"use strict";Object.defineProperty(Or,"__esModule",{value:!0});Or.createCanvasFromMedia=Or.createCanvas=void 0;const Yg=Ye(),zg=ln(),uI=Is(),hI=ja();function Vg({width:e,height:t}){const{createCanvasElement:n}=Yg.env.getEnv(),r=n();return r.width=e,r.height=t,r}Or.createCanvas=Vg;function dI(e,t){const{ImageData:n}=Yg.env.getEnv();if(!(e instanceof n)&&!hI.isMediaLoaded(e))throw new Error("createCanvasFromMedia - media has not finished loading yet");const{width:r,height:i}=t||uI.getMediaDimensions(e),o=Vg({width:r,height:i});return e instanceof n?zg.getContext2dOrThrow(o).putImageData(e,0,0):zg.getContext2dOrThrow(o).drawImage(e,0,0,r,i),o}Or.createCanvasFromMedia=dI});var nd=S(Ya=>{"use strict";Object.defineProperty(Ya,"__esModule",{value:!0});Ya.imageTensorToCanvas=void 0;const Kg=ee(),mI=Ye(),fI=me();async function gI(e,t){const n=t||mI.env.getEnv().createCanvasElement(),[r,i,o]=e.shape.slice(fI.isTensor4D(e)?1:0),a=Kg.tidy(()=>e.as3D(r,i,o).toInt());return await Kg.browser.toPixels(a,n),a.dispose(),n}Ya.imageTensorToCanvas=gI});var rd=S(za=>{"use strict";Object.defineProperty(za,"__esModule",{value:!0});za.isMediaElement=void 0;const bI=Ye();function wI(e){const{Image:t,Canvas:n,Video:r}=bI.env.getEnv();return e instanceof t||e instanceof n||e instanceof r}za.isMediaElement=wI});var id=S(Va=>{"use strict";Object.defineProperty(Va,"__esModule",{value:!0});Va.imageToSquare=void 0;const yI=Ye(),Jg=As(),xI=ln(),LI=Is();function vI(e,t,n=!1){const{Image:r,Canvas:i}=yI.env.getEnv();if(!(e instanceof r||e instanceof i))throw new Error("imageToSquare - expected arg0 to be HTMLImageElement | HTMLCanvasElement");const o=LI.getMediaDimensions(e),a=t/Math.max(o.height,o.width),s=a*o.width,c=a*o.height,p=Jg.createCanvas({width:t,height:t}),l=e instanceof i?e:Jg.createCanvasFromMedia(e),u=Math.abs(s-c)/2,m=n&&s<c?u:0,b=n&&c<s?u:0;return xI.getContext2dOrThrow(p).drawImage(l,m,b,s,c),p}Va.imageToSquare=vI});var od=S(Ka=>{"use strict";Object.defineProperty(Ka,"__esModule",{value:!0});Ka.NetInput=void 0;const Ns=ee(),Xg=Ye(),SI=Gh(),kr=me(),TI=As(),II=id();class AI{constructor(e,t=!1){if(this._imageTensors=[],this._canvases=[],this._treatAsBatchInput=!1,this._inputDimensions=[],!Array.isArray(e))throw new Error(`NetInput.constructor - expected inputs to be an Array of TResolvedNetInput or to be instanceof tf.Tensor4D, instead have ${e}`);this._treatAsBatchInput=t,this._batchSize=e.length,e.forEach((n,r)=>{if(kr.isTensor3D(n)){this._imageTensors[r]=n,this._inputDimensions[r]=n.shape;return}if(kr.isTensor4D(n)){const o=n.shape[0];if(o!==1)throw new Error(`NetInput - tf.Tensor4D with batchSize ${o} passed, but not supported in input array`);this._imageTensors[r]=n,this._inputDimensions[r]=n.shape.slice(1);return}const i=n instanceof Xg.env.getEnv().Canvas?n:TI.createCanvasFromMedia(n);this._canvases[r]=i,this._inputDimensions[r]=[i.height,i.width,3]})}get imageTensors(){return this._imageTensors}get canvases(){return this._canvases}get isBatchInput(){return this.batchSize>1||this._treatAsBatchInput}get batchSize(){return this._batchSize}get inputDimensions(){return this._inputDimensions}get inputSize(){return this._inputSize}get reshapedInputDimensions(){return kr.range(this.batchSize,0,1).map((e,t)=>this.getReshapedInputDimensions(t))}getInput(e){return this.canvases[e]||this.imageTensors[e]}getInputDimensions(e){return this._inputDimensions[e]}getInputHeight(e){return this._inputDimensions[e][0]}getInputWidth(e){return this._inputDimensions[e][1]}getReshapedInputDimensions(e){if(typeof this.inputSize!="number")throw new Error("getReshapedInputDimensions - inputSize not set, toBatchTensor has not been called yet");const t=this.getInputWidth(e),n=this.getInputHeight(e);return kr.computeReshapedDimensions({width:t,height:n},this.inputSize)}toBatchTensor(e,t=!0){return this._inputSize=e,Ns.tidy(()=>{const n=kr.range(this.batchSize,0,1).map(i=>{const o=this.getInput(i);if(o instanceof Ns.Tensor){let a=kr.isTensor4D(o)?o:o.expandDims();return a=SI.padToSquare(a,t),(a.shape[1]!==e||a.shape[2]!==e)&&(a=Ns.image.resizeBilinear(a,[e,e])),a.as3D(e,e,3)}if(o instanceof Xg.env.getEnv().Canvas)return Ns.browser.fromPixels(II.imageToSquare(o,e,t));throw new Error(`toBatchTensor - at batchIdx ${i}, expected input to be instanceof tf.Tensor or instanceof HTMLCanvasElement, instead have ${o}`)}),r=Ns.stack(n.map(i=>i.toFloat())).as4D(this.batchSize,e,e,3);return r})}}Ka.NetInput=AI});var ad=S(Ja=>{"use strict";Object.defineProperty(Ja,"__esModule",{value:!0});Ja.toNetInput=void 0;const sd=me(),NI=Qh(),Zg=rd(),Qg=od(),_I=Ts();async function RI(e){if(e instanceof Qg.NetInput)return e;let t=Array.isArray(e)?e:[e];if(!t.length)throw new Error("toNetInput - empty array passed as input");const n=i=>Array.isArray(e)?` at input index ${i}:`:"",r=t.map(_I.resolveInput);return r.forEach((i,o)=>{if(!Zg.isMediaElement(i)&&!sd.isTensor3D(i)&&!sd.isTensor4D(i))throw typeof t[o]=="string"?new Error(`toNetInput -${n(o)} string passed, but could not resolve HTMLElement for element id ${t[o]}`):new Error(`toNetInput -${n(o)} expected media to be of type HTMLImageElement | HTMLVideoElement | HTMLCanvasElement | tf.Tensor3D, or to be an element id`);if(sd.isTensor4D(i)){const a=i.shape[0];if(a!==1)throw new Error(`toNetInput -${n(o)} tf.Tensor4D with batchSize ${a} passed, but not supported in input array`)}}),await Promise.all(r.map(i=>Zg.isMediaElement(i)&&NI.awaitMediaLoaded(i))),new Qg.NetInput(r,Array.isArray(e))}Ja.toNetInput=RI});var tb=S(Xa=>{"use strict";Object.defineProperty(Xa,"__esModule",{value:!0});Xa.extractFaces=void 0;const CI=Xt(),EI=Ye(),OI=As(),eb=ln(),kI=nd(),DI=ad();async function FI(e,t){const{Canvas:n}=EI.env.getEnv();let r=e;if(!(e instanceof n)){const a=await DI.toNetInput(e);if(a.batchSize>1)throw new Error("extractFaces - batchSize > 1 not supported");const s=a.getInput(0);r=s instanceof n?s:await kI.imageTensorToCanvas(s)}const i=eb.getContext2dOrThrow(r),o=t.map(a=>a instanceof CI.FaceDetection?a.forSize(r.width,r.height).box.floor():a).map(a=>a.clipAtImageBorders(r.width,r.height));return o.map(({x:a,y:s,width:c,height:p})=>{const l=OI.createCanvas({width:c,height:p});return eb.getContext2dOrThrow(l).putImageData(i.getImageData(a,s,c,p),0,0),l})}Xa.extractFaces=FI});var rb=S(Za=>{"use strict";Object.defineProperty(Za,"__esModule",{value:!0});Za.extractFaceTensors=void 0;const nb=ee(),MI=Xt(),Qa=me();async function UI(e,t){if(!Qa.isTensor3D(e)&&!Qa.isTensor4D(e))throw new Error("extractFaceTensors - expected image tensor to be 3D or 4D");if(Qa.isTensor4D(e)&&e.shape[0]>1)throw new Error("extractFaceTensors - batchSize > 1 not supported");return nb.tidy(()=>{const[n,r,i]=e.shape.slice(Qa.isTensor4D(e)?1:0),o=t.map(s=>s instanceof MI.FaceDetection?s.forSize(r,n).box:s).map(s=>s.clipAtImageBorders(r,n)),a=o.map(({x:s,y:c,width:p,height:l})=>nb.slice3d(e.as3D(n,r,i),[c,s,0],[l,p,i]));return a})}Za.extractFaceTensors=UI});var _s=S(ec=>{"use strict";Object.defineProperty(ec,"__esModule",{value:!0});ec.fetchOrThrow=void 0;const WI=Ye();async function qI(e,t){const n=WI.env.getEnv().fetch,r=await n(e,t);if(!(r.status<400))throw new Error(`failed to fetch: (${r.status}) ${r.statusText}, from url: ${r.url}`);return r}ec.fetchOrThrow=qI});var ib=S(tc=>{"use strict";Object.defineProperty(tc,"__esModule",{value:!0});tc.fetchImage=void 0;const BI=ed(),PI=_s();async function jI(e){const t=await PI.fetchOrThrow(e),n=await t.blob();if(!n.type.startsWith("image/"))throw new Error(`fetchImage - expected blob type to be of type image/*, instead have: ${n.type}, for url: ${t.url}`);return BI.bufferToImage(n)}tc.fetchImage=jI});var cd=S(nc=>{"use strict";Object.defineProperty(nc,"__esModule",{value:!0});nc.fetchJson=void 0;const GI=_s();async function $I(e){return(await GI.fetchOrThrow(e)).json()}nc.fetchJson=$I});var ob=S(rc=>{"use strict";Object.defineProperty(rc,"__esModule",{value:!0});rc.fetchNetWeights=void 0;const HI=_s();async function YI(e){return new Float32Array(await(await HI.fetchOrThrow(e)).arrayBuffer())}rc.fetchNetWeights=YI});var oc=S(ic=>{"use strict";Object.defineProperty(ic,"__esModule",{value:!0});ic.getModelUris=void 0;function zI(e,t){const n=`${t}-weights_manifest.json`;if(!e)return{modelBaseUri:"",manifestUri:n};if(e==="/")return{modelBaseUri:"/",manifestUri:`/${n}`};const r=e.startsWith("http://")?"http://":e.startsWith("https://")?"https://":"";e=e.replace(r,"");const i=e.split("/").filter(s=>s),o=e.endsWith(".json")?i[i.length-1]:n;let a=r+(e.endsWith(".json")?i.slice(0,i.length-1):i).join("/");return a=e.startsWith("/")?`/${a}`:a,{modelBaseUri:a,manifestUri:a==="/"?`/${o}`:`${a}/${o}`}}ic.getModelUris=zI});var sb=S(sc=>{"use strict";Object.defineProperty(sc,"__esModule",{value:!0});sc.loadWeightMap=void 0;const VI=ee(),KI=oc(),JI=cd();async function XI(e,t){const{manifestUri:n,modelBaseUri:r}=KI.getModelUris(e,t),i=await JI.fetchJson(n);return VI.io.loadWeights(i,r)}sc.loadWeightMap=XI});var ab=S(ac=>{"use strict";Object.defineProperty(ac,"__esModule",{value:!0});ac.matchDimensions=void 0;const ZI=Is();function QI(e,t,n=!1){const{width:r,height:i}=n?ZI.getMediaDimensions(t):t;return e.width=r,e.height=i,{width:r,height:i}}ac.matchDimensions=QI});var pb=S(cb=>{"use strict";Object.defineProperty(cb,"__esModule",{value:!0})});var tt=S(we=>{"use strict";Object.defineProperty(we,"__esModule",{value:!0});const Te=Be();Te.__exportStar(Qh(),we);Te.__exportStar(ed(),we);Te.__exportStar(As(),we);Te.__exportStar(tb(),we);Te.__exportStar(rb(),we);Te.__exportStar(ib(),we);Te.__exportStar(cd(),we);Te.__exportStar(ob(),we);Te.__exportStar(_s(),we);Te.__exportStar(ln(),we);Te.__exportStar(Is(),we);Te.__exportStar(nd(),we);Te.__exportStar(id(),we);Te.__exportStar(rd(),we);Te.__exportStar(ja(),we);Te.__exportStar(sb(),we);Te.__exportStar(ab(),we);Te.__exportStar(od(),we);Te.__exportStar(Ts(),we);Te.__exportStar(ad(),we);Te.__exportStar(pb(),we)});var Qt=S(cc=>{"use strict";Object.defineProperty(cc,"__esModule",{value:!0});cc.NeuralNetwork=void 0;const Rs=ee(),eA=oc(),tA=tt(),nA=Ye();class rA{constructor(e){this._name=e,this._params=void 0,this._paramMappings=[]}get params(){return this._params}get paramMappings(){return this._paramMappings}get isLoaded(){return!!this.params}getParamFromPath(e){const{obj:t,objProp:n}=this.traversePropertyPath(e);return t[n]}reassignParamFromPath(e,t){const{obj:n,objProp:r}=this.traversePropertyPath(e);n[r].dispose(),n[r]=t}getParamList(){return this._paramMappings.map(({paramPath:e})=>({path:e,tensor:this.getParamFromPath(e)}))}getTrainableParams(){return this.getParamList().filter(e=>e.tensor instanceof Rs.Variable)}getFrozenParams(){return this.getParamList().filter(e=>!(e.tensor instanceof Rs.Variable))}variable(){this.getFrozenParams().forEach(({path:e,tensor:t})=>{this.reassignParamFromPath(e,t.variable())})}freeze(){this.getTrainableParams().forEach(({path:e,tensor:t})=>{const n=Rs.tensor(t.dataSync());t.dispose(),this.reassignParamFromPath(e,n)})}dispose(e=!0){this.getParamList().forEach(t=>{if(e&&t.tensor.isDisposed)throw new Error(`param tensor has already been disposed for path ${t.path}`);t.tensor.dispose()}),this._params=void 0}serializeParams(){return new Float32Array(this.getParamList().map(({tensor:e})=>Array.from(e.dataSync())).reduce((e,t)=>e.concat(t)))}async load(e){if(e instanceof Float32Array){this.extractWeights(e);return}await this.loadFromUri(e)}async loadFromUri(e){if(e&&typeof e!="string")throw new Error(`${this._name}.loadFromUri - expected model uri`);const t=await tA.loadWeightMap(e,this.getDefaultModelName());this.loadFromWeightMap(t)}async loadFromDisk(e){if(e&&typeof e!="string")throw new Error(`${this._name}.loadFromDisk - expected model file path`);const{readFile:t}=nA.env.getEnv(),{manifestUri:n,modelBaseUri:r}=eA.getModelUris(e,this.getDefaultModelName()),i=c=>Promise.all(c.map(p=>t(p).then(l=>l.buffer))),o=Rs.io.weightsLoaderFactory(i),a=JSON.parse((await t(n)).toString()),s=await o(a,r);this.loadFromWeightMap(s)}loadFromWeightMap(e){const{paramMappings:t,params:n}=this.extractParamsFromWeigthMap(e);this._paramMappings=t,this._params=n}extractWeights(e){const{paramMappings:t,params:n}=this.extractParams(e);this._paramMappings=t,this._params=n}traversePropertyPath(e){if(!this.params)throw new Error("traversePropertyPath - model has no loaded params");const t=e.split("/").reduce((i,o)=>{if(!i.nextObj.hasOwnProperty(o))throw new Error(`traversePropertyPath - object does not have property ${o}, for path ${e}`);return{obj:i.nextObj,objProp:o,nextObj:i.nextObj[o]}},{nextObj:this.params}),{obj:n,objProp:r}=t;if(!n||!r||!(n[r]instanceof Rs.Tensor))throw new Error(`traversePropertyPath - parameter is not a tensor, for path ${e}`);return{obj:n,objProp:r}}}cc.NeuralNetwork=rA});var ld=S(pc=>{"use strict";Object.defineProperty(pc,"__esModule",{value:!0});pc.depthwiseSeparableConv=void 0;const pd=ee();function iA(e,t,n){return pd.tidy(()=>{let r=pd.separableConv2d(e,t.depthwise_filter,t.pointwise_filter,n,"same");return r=pd.add(r,t.bias),r})}pc.depthwiseSeparableConv=iA});var ud=S(Dr=>{"use strict";Object.defineProperty(Dr,"__esModule",{value:!0});Dr.denseBlock4=Dr.denseBlock3=void 0;const xe=ee(),nr=ld();function oA(e,t,n=!1){return xe.tidy(()=>{const r=xe.relu(n?xe.add(xe.conv2d(e,t.conv0.filters,[2,2],"same"),t.conv0.bias):nr.depthwiseSeparableConv(e,t.conv0,[2,2])),i=nr.depthwiseSeparableConv(r,t.conv1,[1,1]),o=xe.relu(xe.add(r,i)),a=nr.depthwiseSeparableConv(o,t.conv2,[1,1]);return xe.relu(xe.add(r,xe.add(i,a)))})}Dr.denseBlock3=oA;function sA(e,t,n=!1,r=!0){return xe.tidy(()=>{const i=xe.relu(n?xe.add(xe.conv2d(e,t.conv0.filters,r?[2,2]:[1,1],"same"),t.conv0.bias):nr.depthwiseSeparableConv(e,t.conv0,r?[2,2]:[1,1])),o=nr.depthwiseSeparableConv(i,t.conv1,[1,1]),a=xe.relu(xe.add(i,o)),s=nr.depthwiseSeparableConv(a,t.conv2,[1,1]),c=xe.relu(xe.add(i,xe.add(o,s))),p=nr.depthwiseSeparableConv(c,t.conv3,[1,1]);return xe.relu(xe.add(i,xe.add(o,xe.add(s,p))))})}Dr.denseBlock4=sA});var lb=S(lc=>{"use strict";Object.defineProperty(lc,"__esModule",{value:!0});lc.convLayer=void 0;const uc=ee();function aA(e,t,n="same",r=!1){return uc.tidy(()=>{const i=uc.add(uc.conv2d(e,t.filters,[1,1],n),t.bias);return r?uc.relu(i):i})}lc.convLayer=aA});var hd=S(hc=>{"use strict";Object.defineProperty(hc,"__esModule",{value:!0});hc.disposeUnusedWeightTensors=void 0;function cA(e,t){Object.keys(e).forEach(n=>{t.some(r=>r.originalPath===n)||e[n].dispose()})}hc.disposeUnusedWeightTensors=cA});var hb=S(dc=>{"use strict";Object.defineProperty(dc,"__esModule",{value:!0});dc.extractConvParamsFactory=void 0;const ub=ee();function pA(e,t){return function(n,r,i,o){const a=ub.tensor4d(e(n*r*i*i),[i,i,n,r]),s=ub.tensor1d(e(r));return t.push({paramPath:`${o}/filters`},{paramPath:`${o}/bias`}),{filters:a,bias:s}}}dc.extractConvParamsFactory=pA});var mb=S(mc=>{"use strict";Object.defineProperty(mc,"__esModule",{value:!0});mc.extractFCParamsFactory=void 0;const db=ee();function lA(e,t){return function(n,r,i){const o=db.tensor2d(e(n*r),[n,r]),a=db.tensor1d(e(r));return t.push({paramPath:`${i}/weights`},{paramPath:`${i}/bias`}),{weights:o,bias:a}}}mc.extractFCParamsFactory=lA});var dd=S(fc=>{"use strict";Object.defineProperty(fc,"__esModule",{value:!0});fc.SeparableConvParams=void 0;class uA{constructor(e,t,n){this.depthwise_filter=e,this.pointwise_filter=t,this.bias=n}}fc.SeparableConvParams=uA});var gc=S(Fr=>{"use strict";Object.defineProperty(Fr,"__esModule",{value:!0});Fr.loadSeparableConvParamsFactory=Fr.extractSeparableConvParamsFactory=void 0;const md=ee(),fb=dd();function hA(e,t){return function(n,r,i){const o=md.tensor4d(e(3*3*n),[3,3,n,1]),a=md.tensor4d(e(n*r),[1,1,n,r]),s=md.tensor1d(e(r));return t.push({paramPath:`${i}/depthwise_filter`},{paramPath:`${i}/pointwise_filter`},{paramPath:`${i}/bias`}),new fb.SeparableConvParams(o,a,s)}}Fr.extractSeparableConvParamsFactory=hA;function dA(e){return function(t){const n=e(`${t}/depthwise_filter`,4),r=e(`${t}/pointwise_filter`,4),i=e(`${t}/bias`,1);return new fb.SeparableConvParams(n,r,i)}}Fr.loadSeparableConvParamsFactory=dA});var fd=S(bc=>{"use strict";Object.defineProperty(bc,"__esModule",{value:!0});bc.extractWeightEntryFactory=void 0;const mA=me();function fA(e,t){return function(n,r,i){const o=e[n];if(!mA.isTensor(o,r))throw new Error(`expected weightMap[${n}] to be a Tensor${r}D, instead have ${o}`);return t.push({originalPath:n,paramPath:i||n}),o}}bc.extractWeightEntryFactory=fA});var gd=S(wc=>{"use strict";Object.defineProperty(wc,"__esModule",{value:!0});wc.extractWeightsFactory=void 0;function gA(e){let t=e;function n(i){const o=t.slice(0,i);return t=t.slice(i),o}function r(){return t}return{extractWeights:n,getRemainingWeights:r}}wc.extractWeightsFactory=gA});var De=S(wt=>{"use strict";Object.defineProperty(wt,"__esModule",{value:!0});const Dt=Be();Dt.__exportStar(lb(),wt);Dt.__exportStar(ld(),wt);Dt.__exportStar(hd(),wt);Dt.__exportStar(hb(),wt);Dt.__exportStar(mb(),wt);Dt.__exportStar(gc(),wt);Dt.__exportStar(fd(),wt);Dt.__exportStar(gd(),wt);Dt.__exportStar(oc(),wt);Dt.__exportStar(dd(),wt)});var bd=S(yc=>{"use strict";Object.defineProperty(yc,"__esModule",{value:!0});yc.extractorsFactory=void 0;const gb=De();function bA(e,t){const n=gb.extractConvParamsFactory(e,t),r=gb.extractSeparableConvParamsFactory(e,t);function i(a,s,c,p=!1){const l=p?n(a,s,3,`${c}/conv0`):r(a,s,`${c}/conv0`),u=r(s,s,`${c}/conv1`),m=r(s,s,`${c}/conv2`);return{conv0:l,conv1:u,conv2:m}}function o(a,s,c,p=!1){const{conv0:l,conv1:u,conv2:m}=i(a,s,c,p),b=r(s,s,`${c}/conv3`);return{conv0:l,conv1:u,conv2:m,conv3:b}}return{extractDenseBlock3Params:i,extractDenseBlock4Params:o}}yc.extractorsFactory=bA});var bb=S(xc=>{"use strict";Object.defineProperty(xc,"__esModule",{value:!0});xc.extractParams=void 0;const wA=De(),yA=bd();function xA(e){const t=[],{extractWeights:n,getRemainingWeights:r}=wA.extractWeightsFactory(e),{extractDenseBlock4Params:i}=yA.extractorsFactory(n,t),o=i(3,32,"dense0",!0),a=i(32,64,"dense1"),s=i(64,128,"dense2"),c=i(128,256,"dense3");if(r().length!==0)throw new Error(`weights remaing after extract: ${r().length}`);return{paramMappings:t,params:{dense0:o,dense1:a,dense2:s,dense3:c}}}xc.extractParams=xA});var wd=S(Lc=>{"use strict";Object.defineProperty(Lc,"__esModule",{value:!0});Lc.loadConvParamsFactory=void 0;function LA(e){return function(t){const n=e(`${t}/filters`,4),r=e(`${t}/bias`,1);return{filters:n,bias:r}}}Lc.loadConvParamsFactory=LA});var yd=S(vc=>{"use strict";Object.defineProperty(vc,"__esModule",{value:!0});vc.loadParamsFactory=void 0;const wb=De(),vA=wd();function SA(e,t){const n=wb.extractWeightEntryFactory(e,t),r=vA.loadConvParamsFactory(n),i=wb.loadSeparableConvParamsFactory(n);function o(s,c=!1){const p=c?r(`${s}/conv0`):i(`${s}/conv0`),l=i(`${s}/conv1`),u=i(`${s}/conv2`);return{conv0:p,conv1:l,conv2:u}}function a(s,c=!1){const p=c?r(`${s}/conv0`):i(`${s}/conv0`),l=i(`${s}/conv1`),u=i(`${s}/conv2`),m=i(`${s}/conv3`);return{conv0:p,conv1:l,conv2:u,conv3:m}}return{extractDenseBlock3Params:o,extractDenseBlock4Params:a}}vc.loadParamsFactory=SA});var yb=S(Sc=>{"use strict";Object.defineProperty(Sc,"__esModule",{value:!0});Sc.extractParamsFromWeigthMap=void 0;const TA=De(),IA=yd();function AA(e){const t=[],{extractDenseBlock4Params:n}=IA.loadParamsFactory(e,t),r={dense0:n("dense0",!0),dense1:n("dense1"),dense2:n("dense2"),dense3:n("dense3")};return TA.disposeUnusedWeightTensors(e,t),{params:r,paramMappings:t}}Sc.extractParamsFromWeigthMap=AA});var Ld=S(Tc=>{"use strict";Object.defineProperty(Tc,"__esModule",{value:!0});Tc.FaceFeatureExtractor=void 0;const xd=ee(),NA=tt(),_A=Qt(),RA=pn(),Ic=ud(),CA=bb(),EA=yb();class OA extends _A.NeuralNetwork{constructor(){super("FaceFeatureExtractor")}forwardInput(e){const{params:t}=this;if(!t)throw new Error("FaceFeatureExtractor - load model before inference");return xd.tidy(()=>{const n=e.toBatchTensor(112,!0),r=[122.782,117.001,104.298],i=RA.normalize(n,r).div(xd.scalar(255));let o=Ic.denseBlock4(i,t.dense0,!0);return o=Ic.denseBlock4(o,t.dense1),o=Ic.denseBlock4(o,t.dense2),o=Ic.denseBlock4(o,t.dense3),o=xd.avgPool(o,[7,7],[2,2],"valid"),o})}async forward(e){return this.forwardInput(await NA.toNetInput(e))}getDefaultModelName(){return"face_feature_extractor_model"}extractParamsFromWeigthMap(e){return EA.extractParamsFromWeigthMap(e)}extractParams(e){return CA.extractParams(e)}}Tc.FaceFeatureExtractor=OA});var Sd=S(Ac=>{"use strict";Object.defineProperty(Ac,"__esModule",{value:!0});Ac.fullyConnectedLayer=void 0;const vd=ee();function kA(e,t){return vd.tidy(()=>vd.add(vd.matMul(e,t.weights),t.bias))}Ac.fullyConnectedLayer=kA});var Lb=S(Nc=>{"use strict";Object.defineProperty(Nc,"__esModule",{value:!0});Nc.extractParams=void 0;const xb=De();function DA(e,t,n){const r=[],{extractWeights:i,getRemainingWeights:o}=xb.extractWeightsFactory(e),a=xb.extractFCParamsFactory(i,r),s=a(t,n,"fc");if(o().length!==0)throw new Error(`weights remaing after extract: ${o().length}`);return{paramMappings:r,params:{fc:s}}}Nc.extractParams=DA});var Sb=S(_c=>{"use strict";Object.defineProperty(_c,"__esModule",{value:!0});_c.extractParamsFromWeigthMap=void 0;const vb=De();function FA(e){const t=[],n=vb.extractWeightEntryFactory(e,t);function r(o){const a=n(`${o}/weights`,2),s=n(`${o}/bias`,1);return{weights:a,bias:s}}const i={fc:r("fc")};return vb.disposeUnusedWeightTensors(e,t),{params:i,paramMappings:t}}_c.extractParamsFromWeigthMap=FA});var Td=S(Rc=>{"use strict";Object.defineProperty(Rc,"__esModule",{value:!0});Rc.seperateWeightMaps=void 0;function MA(e){const t={},n={};return Object.keys(e).forEach(r=>{const i=r.startsWith("fc")?n:t;i[r]=e[r]}),{featureExtractorMap:t,classifierMap:n}}Rc.seperateWeightMaps=MA});var Id=S(Cc=>{"use strict";Object.defineProperty(Cc,"__esModule",{value:!0});Cc.FaceProcessor=void 0;const UA=ee(),WA=Sd(),qA=tt(),BA=Qt(),PA=Lb(),jA=Sb(),GA=Td();class $A extends BA.NeuralNetwork{constructor(e,t){super(e);this._faceFeatureExtractor=t}get faceFeatureExtractor(){return this._faceFeatureExtractor}runNet(e){const{params:t}=this;if(!t)throw new Error(`${this._name} - load model before inference`);return UA.tidy(()=>{const n=e instanceof qA.NetInput?this.faceFeatureExtractor.forwardInput(e):e;return WA.fullyConnectedLayer(n.as2D(n.shape[0],-1),t.fc)})}dispose(e=!0){this.faceFeatureExtractor.dispose(e),super.dispose(e)}loadClassifierParams(e){const{params:t,paramMappings:n}=this.extractClassifierParams(e);this._params=t,this._paramMappings=n}extractClassifierParams(e){return PA.extractParams(e,this.getClassifierChannelsIn(),this.getClassifierChannelsOut())}extractParamsFromWeigthMap(e){const{featureExtractorMap:t,classifierMap:n}=GA.seperateWeightMaps(e);return this.faceFeatureExtractor.loadFromWeightMap(t),jA.extractParamsFromWeigthMap(n)}extractParams(e){const t=this.getClassifierChannelsIn(),n=this.getClassifierChannelsOut(),r=n*t+n,i=e.slice(0,e.length-r),o=e.slice(e.length-r);return this.faceFeatureExtractor.extractWeights(i),this.extractClassifierParams(o)}}Cc.FaceProcessor=$A});var Ec=S(un=>{"use strict";Object.defineProperty(un,"__esModule",{value:!0});un.FaceExpressions=un.FACE_EXPRESSION_LABELS=void 0;un.FACE_EXPRESSION_LABELS=["neutral","happy","sad","angry","fearful","disgusted","surprised"];class HA{constructor(e){if(e.length!==7)throw new Error(`FaceExpressions.constructor - expected probabilities.length to be 7, have: ${e.length}`);un.FACE_EXPRESSION_LABELS.forEach((t,n)=>{this[t]=e[n]})}asSortedArray(){return un.FACE_EXPRESSION_LABELS.map(e=>({expression:e,probability:this[e]})).sort((e,t)=>t.probability-e.probability)}}un.FaceExpressions=HA});var Nd=S(Oc=>{"use strict";Object.defineProperty(Oc,"__esModule",{value:!0});Oc.FaceExpressionNet=void 0;const Ad=ee(),Tb=tt(),YA=Ld(),zA=Id(),VA=Ec();class KA extends zA.FaceProcessor{constructor(e=new YA.FaceFeatureExtractor()){super("FaceExpressionNet",e)}forwardInput(e){return Ad.tidy(()=>Ad.softmax(this.runNet(e)))}async forward(e){return this.forwardInput(await Tb.toNetInput(e))}async predictExpressions(e){const t=await Tb.toNetInput(e),n=await this.forwardInput(t),r=await Promise.all(Ad.unstack(n).map(async o=>{const a=await o.data();return o.dispose(),a}));n.dispose();const i=r.map(o=>new VA.FaceExpressions(o));return t.isBatchInput?i:i[0]}getDefaultModelName(){return"face_expression_model"}getClassifierChannelsIn(){return 256}getClassifierChannelsOut(){return 7}}Oc.FaceExpressionNet=KA});var _d=S(kc=>{"use strict";Object.defineProperty(kc,"__esModule",{value:!0});const Ib=Be();Ib.__exportStar(Nd(),kc);Ib.__exportStar(Ec(),kc)});var Dc=S(Mr=>{"use strict";Object.defineProperty(Mr,"__esModule",{value:!0});Mr.extendWithFaceExpressions=Mr.isWithFaceExpressions=void 0;const JA=Ec();function XA(e){return e.expressions instanceof JA.FaceExpressions}Mr.isWithFaceExpressions=XA;function ZA(e,t){const n={expressions:t};return Object.assign({},e,n)}Mr.extendWithFaceExpressions=ZA});var Ab=S(Fc=>{"use strict";Object.defineProperty(Fc,"__esModule",{value:!0});Fc.drawFaceExpressions=void 0;const QA=pt(),eN=_d(),tN=tr(),nN=Dc(),rN=me(),iN=qa();function oN(e,t,n=.1,r){const i=Array.isArray(t)?t:[t];i.forEach(o=>{const a=o instanceof eN.FaceExpressions?o:nN.isWithFaceExpressions(o)?o.expressions:void 0;if(!a)throw new Error("drawFaceExpressions - expected faceExpressions to be FaceExpressions | WithFaceExpressions<{}> or array thereof");const s=a.asSortedArray(),c=s.filter(u=>u.probability>n),p=tN.isWithFaceDetection(o)?o.detection.box.bottomLeft:r||new QA.Point(0,0),l=new iN.DrawTextField(c.map(u=>`${u.expression} (${rN.round(u.probability)})`),p);l.draw(e)})}Fc.drawFaceExpressions=oN});var Wr=S(Ur=>{"use strict";Object.defineProperty(Ur,"__esModule",{value:!0});Ur.extendWithFaceLandmarks=Ur.isWithFaceLandmarks=void 0;const Nb=Xt(),_b=er(),sN=tr();function aN(e){return sN.isWithFaceDetection(e)&&e.landmarks instanceof _b.FaceLandmarks&&e.unshiftedLandmarks instanceof _b.FaceLandmarks&&e.alignedRect instanceof Nb.FaceDetection}Ur.isWithFaceLandmarks=aN;function cN(e,t){const{box:n}=e.detection,r=t.shiftBy(n.x,n.y),i=r.align(),{imageDims:o}=e.detection,a=new Nb.FaceDetection(e.detection.score,i.rescale(o.reverse()),o),s={landmarks:r,unshiftedLandmarks:t,alignedRect:a};return Object.assign({},e,s)}Ur.extendWithFaceLandmarks=cN});var Eb=S(hn=>{"use strict";Object.defineProperty(hn,"__esModule",{value:!0});hn.drawFaceLandmarks=hn.DrawFaceLandmarks=hn.DrawFaceLandmarksOptions=void 0;const pN=er(),lN=Na(),uN=ln(),hN=Wr(),rr=Uh();class Rb{constructor(e={}){const{drawLines:t=!0,drawPoints:n=!0,lineWidth:r,lineColor:i,pointSize:o,pointColor:a}=e;this.drawLines=t,this.drawPoints=n,this.lineWidth=r||1,this.pointSize=o||2,this.lineColor=i||"rgba(0, 255, 255, 1)",this.pointColor=a||"rgba(255, 0, 255, 1)"}}hn.DrawFaceLandmarksOptions=Rb;class Cb{constructor(e,t={}){this.faceLandmarks=e,this.options=new Rb(t)}draw(e){const t=uN.getContext2dOrThrow(e),{drawLines:n,drawPoints:r,lineWidth:i,lineColor:o,pointSize:a,pointColor:s}=this.options;if(n&&this.faceLandmarks instanceof lN.FaceLandmarks68&&(t.strokeStyle=o,t.lineWidth=i,rr.drawContour(t,this.faceLandmarks.getJawOutline()),rr.drawContour(t,this.faceLandmarks.getLeftEyeBrow()),rr.drawContour(t,this.faceLandmarks.getRightEyeBrow()),rr.drawContour(t,this.faceLandmarks.getNose()),rr.drawContour(t,this.faceLandmarks.getLeftEye(),!0),rr.drawContour(t,this.faceLandmarks.getRightEye(),!0),rr.drawContour(t,this.faceLandmarks.getMouth(),!0)),r){t.strokeStyle=s,t.fillStyle=s;const c=p=>{t.beginPath(),t.arc(p.x,p.y,a,0,2*Math.PI),t.fill()};this.faceLandmarks.positions.forEach(c)}}}hn.DrawFaceLandmarks=Cb;function dN(e,t){const n=Array.isArray(t)?t:[t];n.forEach(r=>{const i=r instanceof pN.FaceLandmarks?r:hN.isWithFaceLandmarks(r)?r.landmarks:void 0;if(!i)throw new Error("drawFaceLandmarks - expected faceExpressions to be FaceLandmarks | WithFaceLandmarks<WithFaceDetection<{}>> or array thereof");new Cb(i).draw(e)})}hn.drawFaceLandmarks=dN});var Ob=S(dn=>{"use strict";Object.defineProperty(dn,"__esModule",{value:!0});const qr=Be();qr.__exportStar(Uh(),dn);qr.__exportStar(Hg(),dn);qr.__exportStar(Ab(),dn);qr.__exportStar(Zh(),dn);qr.__exportStar(Eb(),dn);qr.__exportStar(qa(),dn)});var kb=S(Mc=>{"use strict";Object.defineProperty(Mc,"__esModule",{value:!0});Mc.extractParams=void 0;const Rd=De(),mN=me();function fN(e,t){const n=Rd.extractConvParamsFactory(e,t),r=Rd.extractSeparableConvParamsFactory(e,t);function i(a,s,c){const p=r(a,s,`${c}/separable_conv0`),l=r(s,s,`${c}/separable_conv1`),u=n(a,s,1,`${c}/expansion_conv`);return{separable_conv0:p,separable_conv1:l,expansion_conv:u}}function o(a,s){const c=r(a,a,`${s}/separable_conv0`),p=r(a,a,`${s}/separable_conv1`),l=r(a,a,`${s}/separable_conv2`);return{separable_conv0:c,separable_conv1:p,separable_conv2:l}}return{extractConvParams:n,extractSeparableConvParams:r,extractReductionBlockParams:i,extractMainBlockParams:o}}function gN(e,t){const n=[],{extractWeights:r,getRemainingWeights:i}=Rd.extractWeightsFactory(e),{extractConvParams:o,extractSeparableConvParams:a,extractReductionBlockParams:s,extractMainBlockParams:c}=fN(r,n),p=o(3,32,3,"entry_flow/conv_in"),l=s(32,64,"entry_flow/reduction_block_0"),u=s(64,128,"entry_flow/reduction_block_1"),m={conv_in:p,reduction_block_0:l,reduction_block_1:u},b={};mN.range(t,0,1).forEach(T=>{b[`main_block_${T}`]=c(128,`middle_flow/main_block_${T}`)});const y=s(128,256,"exit_flow/reduction_block"),w=a(256,512,"exit_flow/separable_conv"),L={reduction_block:y,separable_conv:w};if(i().length!==0)throw new Error(`weights remaing after extract: ${i().length}`);return{paramMappings:n,params:{entry_flow:m,middle_flow:b,exit_flow:L}}}Mc.extractParams=gN});var Db=S(Uc=>{"use strict";Object.defineProperty(Uc,"__esModule",{value:!0});Uc.extractParamsFromWeigthMap=void 0;const Cd=De(),bN=wd(),wN=me();function yN(e,t){const n=Cd.extractWeightEntryFactory(e,t),r=bN.loadConvParamsFactory(n),i=Cd.loadSeparableConvParamsFactory(n);function o(s){const c=i(`${s}/separable_conv0`),p=i(`${s}/separable_conv1`),l=r(`${s}/expansion_conv`);return{separable_conv0:c,separable_conv1:p,expansion_conv:l}}function a(s){const c=i(`${s}/separable_conv0`),p=i(`${s}/separable_conv1`),l=i(`${s}/separable_conv2`);return{separable_conv0:c,separable_conv1:p,separable_conv2:l}}return{extractConvParams:r,extractSeparableConvParams:i,extractReductionBlockParams:o,extractMainBlockParams:a}}function xN(e,t){const n=[],{extractConvParams:r,extractSeparableConvParams:i,extractReductionBlockParams:o,extractMainBlockParams:a}=yN(e,n),s=r("entry_flow/conv_in"),c=o("entry_flow/reduction_block_0"),p=o("entry_flow/reduction_block_1"),l={conv_in:s,reduction_block_0:c,reduction_block_1:p},u={};wN.range(t,0,1).forEach(w=>{u[`main_block_${w}`]=a(`middle_flow/main_block_${w}`)});const m=o("exit_flow/reduction_block"),b=i("exit_flow/separable_conv"),y={reduction_block:m,separable_conv:b};return Cd.disposeUnusedWeightTensors(e,n),{params:{entry_flow:l,middle_flow:u,exit_flow:y},paramMappings:n}}Uc.extractParamsFromWeigthMap=xN});var Mb=S(Wc=>{"use strict";Object.defineProperty(Wc,"__esModule",{value:!0});Wc.TinyXception=void 0;const nt=ee(),Br=De(),LN=tt(),vN=Qt(),SN=pn(),TN=me(),IN=kb(),AN=Db();function Fb(e,t,n){return nt.add(nt.conv2d(e,t.filters,n,"same"),t.bias)}function Ed(e,t,n=!0){let r=n?nt.relu(e):e;return r=Br.depthwiseSeparableConv(r,t.separable_conv0,[1,1]),r=Br.depthwiseSeparableConv(nt.relu(r),t.separable_conv1,[1,1]),r=nt.maxPool(r,[3,3],[2,2],"same"),r=nt.add(r,Fb(e,t.expansion_conv,[2,2])),r}function NN(e,t){let n=Br.depthwiseSeparableConv(nt.relu(e),t.separable_conv0,[1,1]);return n=Br.depthwiseSeparableConv(nt.relu(n),t.separable_conv1,[1,1]),n=Br.depthwiseSeparableConv(nt.relu(n),t.separable_conv2,[1,1]),n=nt.add(n,e),n}class _N extends vN.NeuralNetwork{constructor(e){super("TinyXception");this._numMainBlocks=e}forwardInput(e){const{params:t}=this;if(!t)throw new Error("TinyXception - load model before inference");return nt.tidy(()=>{const n=e.toBatchTensor(112,!0),r=[122.782,117.001,104.298],i=SN.normalize(n,r).div(nt.scalar(256));let o=nt.relu(Fb(i,t.entry_flow.conv_in,[2,2]));return o=Ed(o,t.entry_flow.reduction_block_0,!1),o=Ed(o,t.entry_flow.reduction_block_1),TN.range(this._numMainBlocks,0,1).forEach(a=>{o=NN(o,t.middle_flow[`main_block_${a}`])}),o=Ed(o,t.exit_flow.reduction_block),o=nt.relu(Br.depthwiseSeparableConv(o,t.exit_flow.separable_conv,[1,1])),o})}async forward(e){return this.forwardInput(await LN.toNetInput(e))}getDefaultModelName(){return"tiny_xception_model"}extractParamsFromWeigthMap(e){return AN.extractParamsFromWeigthMap(e,this._numMainBlocks)}extractParams(e){return IN.extractParams(e,this._numMainBlocks)}}Wc.TinyXception=_N});var Wb=S(qc=>{"use strict";Object.defineProperty(qc,"__esModule",{value:!0});qc.extractParams=void 0;const Ub=De();function RN(e){const t=[],{extractWeights:n,getRemainingWeights:r}=Ub.extractWeightsFactory(e),i=Ub.extractFCParamsFactory(n,t),o=i(512,1,"fc/age"),a=i(512,2,"fc/gender");if(r().length!==0)throw new Error(`weights remaing after extract: ${r().length}`);return{paramMappings:t,params:{fc:{age:o,gender:a}}}}qc.extractParams=RN});var Bb=S(Bc=>{"use strict";Object.defineProperty(Bc,"__esModule",{value:!0});Bc.extractParamsFromWeigthMap=void 0;const qb=De();function CN(e){const t=[],n=qb.extractWeightEntryFactory(e,t);function r(o){const a=n(`${o}/weights`,2),s=n(`${o}/bias`,1);return{weights:a,bias:s}}const i={fc:{age:r("fc/age"),gender:r("fc/gender")}};return qb.disposeUnusedWeightTensors(e,t),{params:i,paramMappings:t}}Bc.extractParamsFromWeigthMap=CN});var Pc=S(Cs=>{"use strict";Object.defineProperty(Cs,"__esModule",{value:!0});Cs.Gender=void 0;var EN;(function(e){e.FEMALE="female",e.MALE="male"})(EN=Cs.Gender||(Cs.Gender={}))});var kd=S(jc=>{"use strict";Object.defineProperty(jc,"__esModule",{value:!0});jc.AgeGenderNet=void 0;const Pr=ee(),Pb=Sd(),ON=Td(),kN=Mb(),DN=Wb(),FN=Bb(),jb=Pc(),MN=Qt(),Od=tt();class UN extends MN.NeuralNetwork{constructor(e=new kN.TinyXception(2)){super("AgeGenderNet");this._faceFeatureExtractor=e}get faceFeatureExtractor(){return this._faceFeatureExtractor}runNet(e){const{params:t}=this;if(!t)throw new Error(`${this._name} - load model before inference`);return Pr.tidy(()=>{const n=e instanceof Od.NetInput?this.faceFeatureExtractor.forwardInput(e):e,r=Pr.avgPool(n,[7,7],[2,2],"valid").as2D(n.shape[0],-1),i=Pb.fullyConnectedLayer(r,t.fc.age).as1D(),o=Pb.fullyConnectedLayer(r,t.fc.gender);return{age:i,gender:o}})}forwardInput(e){return Pr.tidy(()=>{const{age:t,gender:n}=this.runNet(e);return{age:t,gender:Pr.softmax(n)}})}async forward(e){return this.forwardInput(await Od.toNetInput(e))}async predictAgeAndGender(e){const t=await Od.toNetInput(e),n=await this.forwardInput(t),r=Pr.unstack(n.age),i=Pr.unstack(n.gender),o=r.map((s,c)=>({ageTensor:s,genderTensor:i[c]})),a=await Promise.all(o.map(async({ageTensor:s,genderTensor:c})=>{const p=(await s.data())[0],l=(await c.data())[0],u=l>.5,m=u?jb.Gender.MALE:jb.Gender.FEMALE,b=u?l:1-l;return s.dispose(),c.dispose(),{age:p,gender:m,genderProbability:b}}));return n.age.dispose(),n.gender.dispose(),t.isBatchInput?a:a[0]}getDefaultModelName(){return"age_gender_model"}dispose(e=!0){this.faceFeatureExtractor.dispose(e),super.dispose(e)}loadClassifierParams(e){const{params:t,paramMappings:n}=this.extractClassifierParams(e);this._params=t,this._paramMappings=n}extractClassifierParams(e){return DN.extractParams(e)}extractParamsFromWeigthMap(e){const{featureExtractorMap:t,classifierMap:n}=ON.seperateWeightMaps(e);return this.faceFeatureExtractor.loadFromWeightMap(t),FN.extractParamsFromWeigthMap(n)}extractParams(e){const t=512*1+1+(512*2+2),n=e.slice(0,e.length-t),r=e.slice(e.length-t);return this.faceFeatureExtractor.extractWeights(n),this.extractClassifierParams(r)}}jc.AgeGenderNet=UN});var $b=S(Gc=>{"use strict";Object.defineProperty(Gc,"__esModule",{value:!0});const Gb=Be();Gb.__exportStar(kd(),Gc);Gb.__exportStar(Pc(),Gc)});var Dd=S($c=>{"use strict";Object.defineProperty($c,"__esModule",{value:!0});$c.FaceLandmark68NetBase=void 0;const Ft=ee(),WN=pt(),qN=Na(),Hb=tt(),BN=Id(),Yb=me();class PN extends BN.FaceProcessor{postProcess(e,t,n){const r=n.map(({width:o,height:a})=>{const s=t/Math.max(a,o);return{width:o*s,height:a*s}}),i=r.length;return Ft.tidy(()=>{const o=(l,u)=>Ft.stack([Ft.fill([68],l),Ft.fill([68],u)],1).as2D(1,136).as1D(),a=(l,u)=>{const{width:m,height:b}=r[l];return u(m,b)?Math.abs(m-b)/2:0},s=l=>a(l,(u,m)=>u<m),c=l=>a(l,(u,m)=>m<u),p=e.mul(Ft.fill([i,136],t)).sub(Ft.stack(Array.from(Array(i),(l,u)=>o(s(u),c(u))))).div(Ft.stack(Array.from(Array(i),(l,u)=>o(r[u].width,r[u].height))));return p})}forwardInput(e){return Ft.tidy(()=>{const t=this.runNet(e);return this.postProcess(t,e.inputSize,e.inputDimensions.map(([n,r])=>({height:n,width:r})))})}async forward(e){return this.forwardInput(await Hb.toNetInput(e))}async detectLandmarks(e){const t=await Hb.toNetInput(e),n=Ft.tidy(()=>Ft.unstack(this.forwardInput(t))),r=await Promise.all(n.map(async(i,o)=>{const a=Array.from(await i.data()),s=a.filter((p,l)=>Yb.isEven(l)),c=a.filter((p,l)=>!Yb.isEven(l));return new qN.FaceLandmarks68(Array(68).fill(0).map((p,l)=>new WN.Point(s[l],c[l])),{height:t.getInputHeight(o),width:t.getInputWidth(o)})}));return n.forEach(i=>i.dispose()),t.isBatchInput?r:r[0]}getClassifierChannelsOut(){return 136}}$c.FaceLandmark68NetBase=PN});var Yc=S(Hc=>{"use strict";Object.defineProperty(Hc,"__esModule",{value:!0});Hc.FaceLandmark68Net=void 0;const jN=Ld(),GN=Dd();class $N extends GN.FaceLandmark68NetBase{constructor(e=new jN.FaceFeatureExtractor()){super("FaceLandmark68Net",e)}getDefaultModelName(){return"face_landmark_68_model"}getClassifierChannelsIn(){return 256}}Hc.FaceLandmark68Net=$N});var zb=S(zc=>{"use strict";Object.defineProperty(zc,"__esModule",{value:!0});zc.extractParamsFromWeigthMapTiny=void 0;const HN=De(),YN=yd();function zN(e){const t=[],{extractDenseBlock3Params:n}=YN.loadParamsFactory(e,t),r={dense0:n("dense0",!0),dense1:n("dense1"),dense2:n("dense2")};return HN.disposeUnusedWeightTensors(e,t),{params:r,paramMappings:t}}zc.extractParamsFromWeigthMapTiny=zN});var Vb=S(Vc=>{"use strict";Object.defineProperty(Vc,"__esModule",{value:!0});Vc.extractParamsTiny=void 0;const VN=De(),KN=bd();function JN(e){const t=[],{extractWeights:n,getRemainingWeights:r}=VN.extractWeightsFactory(e),{extractDenseBlock3Params:i}=KN.extractorsFactory(n,t),o=i(3,32,"dense0",!0),a=i(32,64,"dense1"),s=i(64,128,"dense2");if(r().length!==0)throw new Error(`weights remaing after extract: ${r().length}`);return{paramMappings:t,params:{dense0:o,dense1:a,dense2:s}}}Vc.extractParamsTiny=JN});var Kb=S(Kc=>{"use strict";Object.defineProperty(Kc,"__esModule",{value:!0});Kc.TinyFaceFeatureExtractor=void 0;const Fd=ee(),XN=tt(),ZN=Qt(),QN=pn(),Md=ud(),e2=zb(),t2=Vb();class n2 extends ZN.NeuralNetwork{constructor(){super("TinyFaceFeatureExtractor")}forwardInput(e){const{params:t}=this;if(!t)throw new Error("TinyFaceFeatureExtractor - load model before inference");return Fd.tidy(()=>{const n=e.toBatchTensor(112,!0),r=[122.782,117.001,104.298],i=QN.normalize(n,r).div(Fd.scalar(255));let o=Md.denseBlock3(i,t.dense0,!0);return o=Md.denseBlock3(o,t.dense1),o=Md.denseBlock3(o,t.dense2),o=Fd.avgPool(o,[14,14],[2,2],"valid"),o})}async forward(e){return this.forwardInput(await XN.toNetInput(e))}getDefaultModelName(){return"face_feature_extractor_tiny_model"}extractParamsFromWeigthMap(e){return e2.extractParamsFromWeigthMapTiny(e)}extractParams(e){return t2.extractParamsTiny(e)}}Kc.TinyFaceFeatureExtractor=n2});var Ud=S(Jc=>{"use strict";Object.defineProperty(Jc,"__esModule",{value:!0});Jc.FaceLandmark68TinyNet=void 0;const r2=Kb(),i2=Dd();class o2 extends i2.FaceLandmark68NetBase{constructor(e=new r2.TinyFaceFeatureExtractor()){super("FaceLandmark68TinyNet",e)}getDefaultModelName(){return"face_landmark_68_tiny_model"}getClassifierChannelsIn(){return 128}}Jc.FaceLandmark68TinyNet=o2});var Xb=S(jr=>{"use strict";Object.defineProperty(jr,"__esModule",{value:!0});jr.FaceLandmarkNet=void 0;const Jb=Be(),s2=Yc();Jb.__exportStar(Yc(),jr);Jb.__exportStar(Ud(),jr);class a2 extends s2.FaceLandmark68Net{}jr.FaceLandmarkNet=a2});var Qb=S(Xc=>{"use strict";Object.defineProperty(Xc,"__esModule",{value:!0});Xc.scale=void 0;const Zb=ee();function c2(e,t){return Zb.add(Zb.mul(e,t.weights),t.biases)}Xc.scale=c2});var Bd=S(mn=>{"use strict";Object.defineProperty(mn,"__esModule",{value:!0});mn.convDown=mn.convNoRelu=mn.conv=void 0;const Wd=ee(),p2=Qb();function qd(e,t,n,r,i="same"){const{filters:o,bias:a}=t.conv;let s=Wd.conv2d(e,o,n,i);return s=Wd.add(s,a),s=p2.scale(s,t.scale),r?Wd.relu(s):s}function l2(e,t){return qd(e,t,[1,1],!0)}mn.conv=l2;function u2(e,t){return qd(e,t,[1,1],!1)}mn.convNoRelu=u2;function h2(e,t){return qd(e,t,[2,2],!0,"valid")}mn.convDown=h2});var ew=S(Zc=>{"use strict";Object.defineProperty(Zc,"__esModule",{value:!0});Zc.extractParams=void 0;const en=ee(),d2=De(),m2=me();function f2(e,t){function n(s,c,p){const l=e(s),u=l.length/(c*p*p);if(m2.isFloat(u))throw new Error(`depth has to be an integer: ${u}, weights.length: ${l.length}, numFilters: ${c}, filterSize: ${p}`);return en.tidy(()=>en.transpose(en.tensor4d(l,[c,u,p,p]),[2,3,1,0]))}function r(s,c,p,l){const u=n(s,c,p),m=en.tensor1d(e(c));return t.push({paramPath:`${l}/filters`},{paramPath:`${l}/bias`}),{filters:u,bias:m}}function i(s,c){const p=en.tensor1d(e(s)),l=en.tensor1d(e(s));return t.push({paramPath:`${c}/weights`},{paramPath:`${c}/biases`}),{weights:p,biases:l}}function o(s,c,p,l){const u=r(s,c,p,`${l}/conv`),m=i(c,`${l}/scale`);return{conv:u,scale:m}}function a(s,c,p,l,u=!1){const m=o((u?.5:1)*s,c,p,`${l}/conv1`),b=o(s,c,p,`${l}/conv2`);return{conv1:m,conv2:b}}return{extractConvLayerParams:o,extractResidualLayerParams:a}}function g2(e){const{extractWeights:t,getRemainingWeights:n}=d2.extractWeightsFactory(e),r=[],{extractConvLayerParams:i,extractResidualLayerParams:o}=f2(t,r),a=i(4704,32,7,"conv32_down"),s=o(9216,32,3,"conv32_1"),c=o(9216,32,3,"conv32_2"),p=o(9216,32,3,"conv32_3"),l=o(36864,64,3,"conv64_down",!0),u=o(36864,64,3,"conv64_1"),m=o(36864,64,3,"conv64_2"),b=o(36864,64,3,"conv64_3"),y=o(147456,128,3,"conv128_down",!0),w=o(147456,128,3,"conv128_1"),L=o(147456,128,3,"conv128_2"),T=o(589824,256,3,"conv256_down",!0),A=o(589824,256,3,"conv256_1"),_=o(589824,256,3,"conv256_2"),N=o(589824,256,3,"conv256_down_out"),C=en.tidy(()=>en.transpose(en.tensor2d(t(256*128),[128,256]),[1,0]));if(r.push({paramPath:"fc"}),n().length!==0)throw new Error(`weights remaing after extract: ${n().length}`);const F={conv32_down:a,conv32_1:s,conv32_2:c,conv32_3:p,conv64_down:l,conv64_1:u,conv64_2:m,conv64_3:b,conv128_down:y,conv128_1:w,conv128_2:L,conv256_down:T,conv256_1:A,conv256_2:_,conv256_down_out:N,fc:C};return{params:F,paramMappings:r}}Zc.extractParams=g2});var nw=S(Qc=>{"use strict";Object.defineProperty(Qc,"__esModule",{value:!0});Qc.extractParamsFromWeigthMap=void 0;const tw=De(),b2=me();function w2(e,t){const n=tw.extractWeightEntryFactory(e,t);function r(a){const s=n(`${a}/scale/weights`,1),c=n(`${a}/scale/biases`,1);return{weights:s,biases:c}}function i(a){const s=n(`${a}/conv/filters`,4),c=n(`${a}/conv/bias`,1),p=r(a);return{conv:{filters:s,bias:c},scale:p}}function o(a){return{conv1:i(`${a}/conv1`),conv2:i(`${a}/conv2`)}}return{extractConvLayerParams:i,extractResidualLayerParams:o}}function y2(e){const t=[],{extractConvLayerParams:n,extractResidualLayerParams:r}=w2(e,t),i=n("conv32_down"),o=r("conv32_1"),a=r("conv32_2"),s=r("conv32_3"),c=r("conv64_down"),p=r("conv64_1"),l=r("conv64_2"),u=r("conv64_3"),m=r("conv128_down"),b=r("conv128_1"),y=r("conv128_2"),w=r("conv256_down"),L=r("conv256_1"),T=r("conv256_2"),A=r("conv256_down_out"),_=e.fc;if(t.push({originalPath:"fc",paramPath:"fc"}),!b2.isTensor2D(_))throw new Error(`expected weightMap[fc] to be a Tensor2D, instead have ${_}`);const N={conv32_down:i,conv32_1:o,conv32_2:a,conv32_3:s,conv64_down:c,conv64_1:p,conv64_2:l,conv64_3:u,conv128_down:m,conv128_1:b,conv128_2:y,conv256_down:w,conv256_1:L,conv256_2:T,conv256_down_out:A,fc:_};return tw.disposeUnusedWeightTensors(e,t),{params:N,paramMappings:t}}Qc.extractParamsFromWeigthMap=y2});var rw=S(Gr=>{"use strict";Object.defineProperty(Gr,"__esModule",{value:!0});Gr.residualDown=Gr.residual=void 0;const _t=ee(),ep=Bd();function x2(e,t){let n=ep.conv(e,t.conv1);return n=ep.convNoRelu(n,t.conv2),n=_t.add(n,e),n=_t.relu(n),n}Gr.residual=x2;function L2(e,t){let n=ep.convDown(e,t.conv1);n=ep.convNoRelu(n,t.conv2);let r=_t.avgPool(e,2,2,"valid");const i=_t.zeros(r.shape),o=r.shape[3]!==n.shape[3],a=r.shape[1]!==n.shape[1]||r.shape[2]!==n.shape[2];if(a){const s=[...n.shape];s[1]=1;const c=_t.zeros(s);n=_t.concat([n,c],1);const p=[...n.shape];p[2]=1;const l=_t.zeros(p);n=_t.concat([n,l],2)}return r=o?_t.concat([r,i],3):r,n=_t.add(r,n),n=_t.relu(n),n}Gr.residualDown=L2});var np=S(tp=>{"use strict";Object.defineProperty(tp,"__esModule",{value:!0});tp.FaceRecognitionNet=void 0;const $r=ee(),iw=tt(),v2=Qt(),S2=pn(),T2=Bd(),I2=ew(),A2=nw(),rt=rw();class N2 extends v2.NeuralNetwork{constructor(){super("FaceRecognitionNet")}forwardInput(e){const{params:t}=this;if(!t)throw new Error("FaceRecognitionNet - load model before inference");return $r.tidy(()=>{const n=e.toBatchTensor(150,!0).toFloat(),r=[122.782,117.001,104.298],i=S2.normalize(n,r).div($r.scalar(256));let o=T2.convDown(i,t.conv32_down);o=$r.maxPool(o,3,2,"valid"),o=rt.residual(o,t.conv32_1),o=rt.residual(o,t.conv32_2),o=rt.residual(o,t.conv32_3),o=rt.residualDown(o,t.conv64_down),o=rt.residual(o,t.conv64_1),o=rt.residual(o,t.conv64_2),o=rt.residual(o,t.conv64_3),o=rt.residualDown(o,t.conv128_down),o=rt.residual(o,t.conv128_1),o=rt.residual(o,t.conv128_2),o=rt.residualDown(o,t.conv256_down),o=rt.residual(o,t.conv256_1),o=rt.residual(o,t.conv256_2),o=rt.residualDown(o,t.conv256_down_out);const a=o.mean([1,2]),s=$r.matMul(a,t.fc);return s})}async forward(e){return this.forwardInput(await iw.toNetInput(e))}async computeFaceDescriptor(e){const t=await iw.toNetInput(e),n=$r.tidy(()=>$r.unstack(this.forwardInput(t))),r=await Promise.all(n.map(i=>i.data()));return n.forEach(i=>i.dispose()),t.isBatchInput?r:r[0]}getDefaultModelName(){return"face_recognition_model"}extractParamsFromWeigthMap(e){return A2.extractParamsFromWeigthMap(e)}extractParams(e){return I2.extractParams(e)}}tp.FaceRecognitionNet=N2});var ow=S(Es=>{"use strict";Object.defineProperty(Es,"__esModule",{value:!0});Es.createFaceRecognitionNet=void 0;const _2=Be(),R2=np();_2.__exportStar(np(),Es);function C2(e){const t=new R2.FaceRecognitionNet();return t.extractWeights(e),t}Es.createFaceRecognitionNet=C2});var Pd=S(rp=>{"use strict";Object.defineProperty(rp,"__esModule",{value:!0});rp.extendWithFaceDescriptor=void 0;function E2(e,t){const n={descriptor:t};return Object.assign({},e,n)}rp.extendWithFaceDescriptor=E2});var jd=S(Hr=>{"use strict";Object.defineProperty(Hr,"__esModule",{value:!0});Hr.extendWithAge=Hr.isWithAge=void 0;function O2(e){return typeof e.age=="number"}Hr.isWithAge=O2;function k2(e,t){const n={age:t};return Object.assign({},e,n)}Hr.extendWithAge=k2});var Gd=S(Yr=>{"use strict";Object.defineProperty(Yr,"__esModule",{value:!0});Yr.extendWithGender=Yr.isWithGender=void 0;const sw=Pc(),D2=me();function F2(e){return(e.gender===sw.Gender.MALE||e.gender===sw.Gender.FEMALE)&&D2.isValidProbablitiy(e.genderProbability)}Yr.isWithGender=F2;function M2(e,t,n){const r={gender:t,genderProbability:n};return Object.assign({},e,r)}Yr.extendWithGender=M2});var aw=S(fn=>{"use strict";Object.defineProperty(fn,"__esModule",{value:!0});const zr=Be();zr.__exportStar(Pd(),fn);zr.__exportStar(tr(),fn);zr.__exportStar(Dc(),fn);zr.__exportStar(Wr(),fn);zr.__exportStar(jd(),fn);zr.__exportStar(Gd(),fn)});var ir=S(ip=>{"use strict";Object.defineProperty(ip,"__esModule",{value:!0});ip.ComposableTask=void 0;class U2{async then(e){return e(await this.run())}async run(){throw new Error("ComposableTask - run is not implemented")}}ip.ComposableTask=U2});var op=S(Vr=>{"use strict";Object.defineProperty(Vr,"__esModule",{value:!0});Vr.extractSingleFaceAndComputeResult=Vr.extractAllFacesAndComputeResults=void 0;const cw=ee(),pw=tt(),W2=Wr();async function lw(e,t,n,r,i=({alignedRect:o})=>o){const o=e.map(c=>W2.isWithFaceLandmarks(c)?i(c):c.detection),a=r||(t instanceof cw.Tensor?await pw.extractFaceTensors(t,o):await pw.extractFaces(t,o)),s=await n(a);return a.forEach(c=>c instanceof cw.Tensor&&c.dispose()),s}Vr.extractAllFacesAndComputeResults=lw;async function q2(e,t,n,r,i){return lw([e],t,async o=>n(o[0]),r,i)}Vr.extractSingleFaceAndComputeResult=q2});var $d=S(sp=>{"use strict";Object.defineProperty(sp,"__esModule",{value:!0});sp.validateConfig=void 0;const ap=e=>typeof e=="number";function B2(e){if(!e)throw new Error(`invalid config: ${e}`);if(typeof e.withSeparableConvs!="boolean")throw new Error(`config.withSeparableConvs has to be a boolean, have: ${e.withSeparableConvs}`);if(!ap(e.iouThreshold)||e.iouThreshold<0||e.iouThreshold>1)throw new Error(`config.iouThreshold has to be a number between [0, 1], have: ${e.iouThreshold}`);if(!Array.isArray(e.classes)||!e.classes.length||!e.classes.every(t=>typeof t=="string"))throw new Error(`config.classes has to be an array class names: string[], have: ${JSON.stringify(e.classes)}`);if(!Array.isArray(e.anchors)||!e.anchors.length||!e.anchors.map(t=>t||{}).every(t=>ap(t.x)&&ap(t.y)))throw new Error(`config.anchors has to be an array of { x: number, y: number }, have: ${JSON.stringify(e.anchors)}`);if(e.meanRgb&&(!Array.isArray(e.meanRgb)||e.meanRgb.length!==3||!e.meanRgb.every(ap)))throw new Error(`config.meanRgb has to be an array of shape [number, number, number], have: ${JSON.stringify(e.meanRgb)}`)}sp.validateConfig=B2});var pp=S(cp=>{"use strict";Object.defineProperty(cp,"__esModule",{value:!0});cp.leaky=void 0;const Kr=ee();function P2(e){return Kr.tidy(()=>{const t=Kr.mul(e,Kr.scalar(.10000000149011612));return Kr.add(Kr.relu(Kr.sub(e,t)),t)})}cp.leaky=P2});var uw=S(lp=>{"use strict";Object.defineProperty(lp,"__esModule",{value:!0});lp.convWithBatchNorm=void 0;const Jr=ee(),j2=pp();function G2(e,t){return Jr.tidy(()=>{let n=Jr.pad(e,[[0,0],[1,1],[1,1],[0,0]]);return n=Jr.conv2d(n,t.conv.filters,[1,1],"valid"),n=Jr.sub(n,t.bn.sub),n=Jr.mul(n,t.bn.truediv),n=Jr.add(n,t.conv.bias),j2.leaky(n)})}lp.convWithBatchNorm=G2});var hw=S(up=>{"use strict";Object.defineProperty(up,"__esModule",{value:!0});up.depthwiseSeparableConv=void 0;const hp=ee(),$2=pp();function H2(e,t){return hp.tidy(()=>{let n=hp.pad(e,[[0,0],[1,1],[1,1],[0,0]]);return n=hp.separableConv2d(n,t.depthwise_filter,t.pointwise_filter,[1,1],"valid"),n=hp.add(n,t.bias),$2.leaky(n)})}up.depthwiseSeparableConv=H2});var mw=S(dp=>{"use strict";Object.defineProperty(dp,"__esModule",{value:!0});dp.extractParams=void 0;const dw=ee(),Y2=De(),z2=gc(),V2=gd();function K2(e,t){const n=Y2.extractConvParamsFactory(e,t);function r(a,s){const c=dw.tensor1d(e(a)),p=dw.tensor1d(e(a));return t.push({paramPath:`${s}/sub`},{paramPath:`${s}/truediv`}),{sub:c,truediv:p}}function i(a,s,c){const p=n(a,s,3,`${c}/conv`),l=r(s,`${c}/bn`);return{conv:p,bn:l}}const o=z2.extractSeparableConvParamsFactory(e,t);return{extractConvParams:n,extractConvWithBatchNormParams:i,extractSeparableConvParams:o}}function J2(e,t,n,r){const{extractWeights:i,getRemainingWeights:o}=V2.extractWeightsFactory(e),a=[],{extractConvParams:s,extractConvWithBatchNormParams:c,extractSeparableConvParams:p}=K2(i,a);let l;if(t.withSeparableConvs){const[u,m,b,y,w,L,T,A,_]=r,N=t.isFirstLayerConv2d?s(u,m,3,"conv0"):p(u,m,"conv0"),C=p(m,b,"conv1"),F=p(b,y,"conv2"),E=p(y,w,"conv3"),B=p(w,L,"conv4"),q=p(L,T,"conv5"),P=A?p(T,A,"conv6"):void 0,te=_?p(A,_,"conv7"):void 0,ne=s(_||A||T,5*n,1,"conv8");l={conv0:N,conv1:C,conv2:F,conv3:E,conv4:B,conv5:q,conv6:P,conv7:te,conv8:ne}}else{const[u,m,b,y,w,L,T,A,_]=r,N=c(u,m,"conv0"),C=c(m,b,"conv1"),F=c(b,y,"conv2"),E=c(y,w,"conv3"),B=c(w,L,"conv4"),q=c(L,T,"conv5"),P=c(T,A,"conv6"),te=c(A,_,"conv7"),ne=s(_,5*n,1,"conv8");l={conv0:N,conv1:C,conv2:F,conv3:E,conv4:B,conv5:q,conv6:P,conv7:te,conv8:ne}}if(o().length!==0)throw new Error(`weights remaing after extract: ${o().length}`);return{params:l,paramMappings:a}}dp.extractParams=J2});var fw=S(mp=>{"use strict";Object.defineProperty(mp,"__esModule",{value:!0});mp.extractParamsFromWeigthMap=void 0;const X2=hd(),Z2=gc(),Q2=fd();function e_(e,t){const n=Q2.extractWeightEntryFactory(e,t);function r(s){const c=n(`${s}/sub`,1),p=n(`${s}/truediv`,1);return{sub:c,truediv:p}}function i(s){const c=n(`${s}/filters`,4),p=n(`${s}/bias`,1);return{filters:c,bias:p}}function o(s){const c=i(`${s}/conv`),p=r(`${s}/bn`);return{conv:c,bn:p}}const a=Z2.loadSeparableConvParamsFactory(n);return{extractConvParams:i,extractConvWithBatchNormParams:o,extractSeparableConvParams:a}}function t_(e,t){const n=[],{extractConvParams:r,extractConvWithBatchNormParams:i,extractSeparableConvParams:o}=e_(e,n);let a;if(t.withSeparableConvs){const s=t.filterSizes&&t.filterSizes.length||9;a={conv0:t.isFirstLayerConv2d?r("conv0"):o("conv0"),conv1:o("conv1"),conv2:o("conv2"),conv3:o("conv3"),conv4:o("conv4"),conv5:o("conv5"),conv6:s>7?o("conv6"):void 0,conv7:s>8?o("conv7"):void 0,conv8:r("conv8")}}else a={conv0:i("conv0"),conv1:i("conv1"),conv2:i("conv2"),conv3:i("conv3"),conv4:i("conv4"),conv5:i("conv5"),conv6:i("conv6"),conv7:i("conv7"),conv8:r("conv8")};return X2.disposeUnusedWeightTensors(e,n),{params:a,paramMappings:n}}mp.extractParamsFromWeigthMap=t_});var Hd=S(or=>{"use strict";Object.defineProperty(or,"__esModule",{value:!0});or.TinyYolov2Options=or.TinyYolov2SizeType=void 0;var n_;(function(e){e[e.XS=224]="XS",e[e.SM=320]="SM",e[e.MD=416]="MD",e[e.LG=608]="LG"})(n_=or.TinyYolov2SizeType||(or.TinyYolov2SizeType={}));class r_{constructor({inputSize:e,scoreThreshold:t}={}){if(this._name="TinyYolov2Options",this._inputSize=e||416,this._scoreThreshold=t||.5,typeof this._inputSize!="number"||this._inputSize%32!==0)throw new Error(`${this._name} - expected inputSize to be a number divisible by 32`);if(typeof this._scoreThreshold!="number"||this._scoreThreshold<=0||this._scoreThreshold>=1)throw new Error(`${this._name} - expected scoreThreshold to be a number between 0 and 1`)}get inputSize(){return this._inputSize}get scoreThreshold(){return this._scoreThreshold}}or.TinyYolov2Options=r_});var Kd=S(fp=>{"use strict";Object.defineProperty(fp,"__esModule",{value:!0});fp.TinyYolov2Base=void 0;const Ce=ee(),i_=Wh(),o_=fa(),Yd=De(),gw=tt(),s_=Qt(),zd=pn(),a_=Bh(),c_=Ph(),p_=$d(),gn=uw(),bn=hw(),l_=mw(),u_=fw(),h_=pp(),d_=Hd();class Vd extends s_.NeuralNetwork{constructor(e){super("TinyYolov2");p_.validateConfig(e),this._config=e}get config(){return this._config}get withClassScores(){return this.config.withClassScores||this.config.classes.length>1}get boxEncodingSize(){return 5+(this.withClassScores?this.config.classes.length:0)}runTinyYolov2(e,t){let n=gn.convWithBatchNorm(e,t.conv0);return n=Ce.maxPool(n,[2,2],[2,2],"same"),n=gn.convWithBatchNorm(n,t.conv1),n=Ce.maxPool(n,[2,2],[2,2],"same"),n=gn.convWithBatchNorm(n,t.conv2),n=Ce.maxPool(n,[2,2],[2,2],"same"),n=gn.convWithBatchNorm(n,t.conv3),n=Ce.maxPool(n,[2,2],[2,2],"same"),n=gn.convWithBatchNorm(n,t.conv4),n=Ce.maxPool(n,[2,2],[2,2],"same"),n=gn.convWithBatchNorm(n,t.conv5),n=Ce.maxPool(n,[2,2],[1,1],"same"),n=gn.convWithBatchNorm(n,t.conv6),n=gn.convWithBatchNorm(n,t.conv7),Yd.convLayer(n,t.conv8,"valid",!1)}runMobilenet(e,t){let n=this.config.isFirstLayerConv2d?h_.leaky(Yd.convLayer(e,t.conv0,"valid",!1)):bn.depthwiseSeparableConv(e,t.conv0);return n=Ce.maxPool(n,[2,2],[2,2],"same"),n=bn.depthwiseSeparableConv(n,t.conv1),n=Ce.maxPool(n,[2,2],[2,2],"same"),n=bn.depthwiseSeparableConv(n,t.conv2),n=Ce.maxPool(n,[2,2],[2,2],"same"),n=bn.depthwiseSeparableConv(n,t.conv3),n=Ce.maxPool(n,[2,2],[2,2],"same"),n=bn.depthwiseSeparableConv(n,t.conv4),n=Ce.maxPool(n,[2,2],[2,2],"same"),n=bn.depthwiseSeparableConv(n,t.conv5),n=Ce.maxPool(n,[2,2],[1,1],"same"),n=t.conv6?bn.depthwiseSeparableConv(n,t.conv6):n,n=t.conv7?bn.depthwiseSeparableConv(n,t.conv7):n,Yd.convLayer(n,t.conv8,"valid",!1)}forwardInput(e,t){const{params:n}=this;if(!n)throw new Error("TinyYolov2 - load model before inference");return Ce.tidy(()=>{let r=e.toBatchTensor(t,!1).toFloat();return r=this.config.meanRgb?c_.normalize(r,this.config.meanRgb):r,r=r.div(Ce.scalar(256)),this.config.withSeparableConvs?this.runMobilenet(r,n):this.runTinyYolov2(r,n)})}async forward(e,t){return await this.forwardInput(await gw.toNetInput(e),t)}async detect(e,t={}){const{inputSize:n,scoreThreshold:r}=new d_.TinyYolov2Options(t),i=await gw.toNetInput(e),o=await this.forwardInput(i,n),a=Ce.tidy(()=>Ce.unstack(o)[0].expandDims()),s={width:i.getInputWidth(0),height:i.getInputHeight(0)},c=await this.extractBoxes(a,i.getReshapedInputDimensions(0),r);o.dispose(),a.dispose();const p=c.map(w=>w.box),l=c.map(w=>w.score),u=c.map(w=>w.classScore),m=c.map(w=>this.config.classes[w.label]),b=a_.nonMaxSuppression(p.map(w=>w.rescale(n)),l,this.config.iouThreshold,!0),y=b.map(w=>new o_.ObjectDetection(l[w],u[w],m[w],p[w],s));return y}getDefaultModelName(){return""}extractParamsFromWeigthMap(e){return u_.extractParamsFromWeigthMap(e,this.config)}extractParams(e){const t=this.config.filterSizes||Vd.DEFAULT_FILTER_SIZES,n=t?t.length:void 0;if(n!==7&&n!==8&&n!==9)throw new Error(`TinyYolov2 - expected 7 | 8 | 9 convolutional filters, but found ${n} filterSizes in config`);return l_.extractParams(e,this.config,this.boxEncodingSize,t)}async extractBoxes(e,t,n){const{width:r,height:i}=t,o=Math.max(r,i),a=o/r,s=o/i,c=e.shape[1],p=this.config.anchors.length,[l,u,m]=Ce.tidy(()=>{const L=e.reshape([c,c,p,this.boxEncodingSize]),T=L.slice([0,0,0,0],[c,c,p,4]),A=L.slice([0,0,0,4],[c,c,p,1]),_=this.withClassScores?Ce.softmax(L.slice([0,0,0,5],[c,c,p,this.config.classes.length]),3):Ce.scalar(0);return[T,A,_]}),b=[],y=await u.array(),w=await l.array();for(let L=0;L<c;L++)for(let T=0;T<c;T++)for(let A=0;A<p;A++){const _=zd.sigmoid(y[L][T][A][0]);if(!n||_>n){const N=(T+zd.sigmoid(w[L][T][A][0]))/c*a,C=(L+zd.sigmoid(w[L][T][A][1]))/c*s,F=Math.exp(w[L][T][A][2])*this.config.anchors[A].x/c*a,E=Math.exp(w[L][T][A][3])*this.config.anchors[A].y/c*s,B=N-F/2,q=C-E/2,P={row:L,col:T,anchor:A},{classScore:te,label:ne}=this.withClassScores?await this.extractPredictedClass(m,P):{classScore:1,label:0};b.push({box:new i_.BoundingBox(B,q,B+F,q+E),score:_,classScore:_*te,label:ne,...P})}}return l.dispose(),u.dispose(),m.dispose(),b}async extractPredictedClass(e,t){const{row:n,col:r,anchor:i}=t,o=await e.array();return Array(this.config.classes.length).fill(0).map((a,s)=>o[n][r][i][s]).map((a,s)=>({classScore:a,label:s})).reduce((a,s)=>a.classScore>s.classScore?a:s)}}fp.TinyYolov2Base=Vd;Vd.DEFAULT_FILTER_SIZES=[3,16,32,64,128,256,512,1024,1024]});var bw=S(wn=>{"use strict";Object.defineProperty(wn,"__esModule",{value:!0});wn.MEAN_RGB=wn.BOX_ANCHORS=wn.IOU_THRESHOLD=void 0;const Os=pt();wn.IOU_THRESHOLD=.4;wn.BOX_ANCHORS=[new Os.Point(1.603231,2.094468),new Os.Point(6.041143,7.080126),new Os.Point(2.882459,3.518061),new Os.Point(4.266906,5.178857),new Os.Point(9.041765,10.66308)];wn.MEAN_RGB=[117.001,114.697,97.404]});var bp=S(gp=>{"use strict";Object.defineProperty(gp,"__esModule",{value:!0});gp.TinyFaceDetector=void 0;const m_=pt(),f_=Kd(),Jd=bw();class g_ extends f_.TinyYolov2Base{constructor(){const e={withSeparableConvs:!0,iouThreshold:Jd.IOU_THRESHOLD,classes:["face"],anchors:Jd.BOX_ANCHORS,meanRgb:Jd.MEAN_RGB,isFirstLayerConv2d:!0,filterSizes:[3,16,32,64,128,256,512]};super(e)}get anchors(){return this.config.anchors}async locateFaces(e,t){const n=await this.detect(e,t);return n.map(r=>new m_.FaceDetection(r.score,r.relativeBox,{width:r.imageWidth,height:r.imageHeight}))}getDefaultModelName(){return"tiny_face_detector_model"}extractParamsFromWeigthMap(e){return super.extractParamsFromWeigthMap(e)}}gp.TinyFaceDetector=g_});var ww=S(it=>{"use strict";Object.defineProperty(it,"__esModule",{value:!0});it.DEFAULT_MODEL_NAME_SEPARABLE_CONV=it.DEFAULT_MODEL_NAME=it.MEAN_RGB_SEPARABLE=it.BOX_ANCHORS_SEPARABLE=it.BOX_ANCHORS=it.IOU_THRESHOLD=void 0;const Mt=pt();it.IOU_THRESHOLD=.4;it.BOX_ANCHORS=[new Mt.Point(.738768,.874946),new Mt.Point(2.42204,2.65704),new Mt.Point(4.30971,7.04493),new Mt.Point(10.246,4.59428),new Mt.Point(12.6868,11.8741)];it.BOX_ANCHORS_SEPARABLE=[new Mt.Point(1.603231,2.094468),new Mt.Point(6.041143,7.080126),new Mt.Point(2.882459,3.518061),new Mt.Point(4.266906,5.178857),new Mt.Point(9.041765,10.66308)];it.MEAN_RGB_SEPARABLE=[117.001,114.697,97.404];it.DEFAULT_MODEL_NAME="tiny_yolov2_model";it.DEFAULT_MODEL_NAME_SEPARABLE_CONV="tiny_yolov2_separable_conv_model"});var yw=S(wp=>{"use strict";Object.defineProperty(wp,"__esModule",{value:!0});wp.TinyYolov2=void 0;const b_=pt(),Xr=ww(),w_=Kd();class y_ extends w_.TinyYolov2Base{constructor(e=!0){const t=Object.assign({},{withSeparableConvs:e,iouThreshold:Xr.IOU_THRESHOLD,classes:["face"]},e?{anchors:Xr.BOX_ANCHORS_SEPARABLE,meanRgb:Xr.MEAN_RGB_SEPARABLE}:{anchors:Xr.BOX_ANCHORS,withClassScores:!0});super(t)}get withSeparableConvs(){return this.config.withSeparableConvs}get anchors(){return this.config.anchors}async locateFaces(e,t){const n=await this.detect(e,t);return n.map(r=>new b_.FaceDetection(r.score,r.relativeBox,{width:r.imageWidth,height:r.imageHeight}))}getDefaultModelName(){return this.withSeparableConvs?Xr.DEFAULT_MODEL_NAME_SEPARABLE_CONV:Xr.DEFAULT_MODEL_NAME}extractParamsFromWeigthMap(e){return super.extractParamsFromWeigthMap(e)}}wp.TinyYolov2=y_});var Lw=S(xw=>{"use strict";Object.defineProperty(xw,"__esModule",{value:!0})});var yp=S(tn=>{"use strict";Object.defineProperty(tn,"__esModule",{value:!0});tn.createTinyYolov2=tn.TinyYolov2=void 0;const Xd=Be(),vw=yw();Object.defineProperty(tn,"TinyYolov2",{enumerable:!0,get:function(){return vw.TinyYolov2}});Xd.__exportStar(Hd(),tn);Xd.__exportStar($d(),tn);Xd.__exportStar(Lw(),tn);function x_(e,t=!0){const n=new vw.TinyYolov2(t);return n.extractWeights(e),n}tn.createTinyYolov2=x_});var sr=S(j=>{"use strict";Object.defineProperty(j,"__esModule",{value:!0});j.detectLandmarks=j.locateFaces=j.loadFaceDetectionModel=j.loadAgeGenderModel=j.loadFaceExpressionModel=j.loadFaceRecognitionModel=j.loadFaceLandmarkTinyModel=j.loadFaceLandmarkModel=j.loadTinyYolov2Model=j.loadTinyFaceDetectorModel=j.predictAgeAndGender=j.recognizeFaceExpressions=j.computeFaceDescriptor=j.detectFaceLandmarksTiny=j.detectFaceLandmarks=j.tinyFaceDetector=j.nets=void 0;const L_=kd(),v_=Nd(),S_=Yc(),T_=Ud(),I_=np(),Sw=bp(),A_=yp();j.nets={tinyFaceDetector:new Sw.TinyFaceDetector(),tinyYolov2:new A_.TinyYolov2(),faceLandmark68Net:new S_.FaceLandmark68Net(),faceLandmark68TinyNet:new T_.FaceLandmark68TinyNet(),faceRecognitionNet:new I_.FaceRecognitionNet(),faceExpressionNet:new v_.FaceExpressionNet(),ageGenderNet:new L_.AgeGenderNet()};j.tinyFaceDetector=(e,t)=>j.nets.tinyFaceDetector.locateFaces(e,t);j.detectFaceLandmarks=e=>j.nets.faceLandmark68Net.detectLandmarks(e);j.detectFaceLandmarksTiny=e=>j.nets.faceLandmark68TinyNet.detectLandmarks(e);j.computeFaceDescriptor=e=>j.nets.faceRecognitionNet.computeFaceDescriptor(e);j.recognizeFaceExpressions=e=>j.nets.faceExpressionNet.predictExpressions(e);j.predictAgeAndGender=e=>j.nets.ageGenderNet.predictAgeAndGender(e);j.loadTinyFaceDetectorModel=e=>j.nets.tinyFaceDetector.load(e);j.loadTinyYolov2Model=e=>j.nets.tinyYolov2.load(e);j.loadFaceLandmarkModel=e=>j.nets.faceLandmark68Net.load(e);j.loadFaceLandmarkTinyModel=e=>j.nets.faceLandmark68TinyNet.load(e);j.loadFaceRecognitionModel=e=>j.nets.faceRecognitionNet.load(e);j.loadFaceExpressionModel=e=>j.nets.faceExpressionNet.load(e);j.loadAgeGenderModel=e=>j.nets.ageGenderNet.load(e);j.loadFaceDetectionModel=j.loadTinyFaceDetectorModel;j.locateFaces=Sw.TinyFaceDetector;j.detectLandmarks=j.detectFaceLandmarks});var Fs=S(yt=>{"use strict";Object.defineProperty(yt,"__esModule",{value:!0});yt.PredictSingleFaceExpressionsWithFaceAlignmentTask=yt.PredictAllFaceExpressionsWithFaceAlignmentTask=yt.PredictSingleFaceExpressionsTask=yt.PredictAllFaceExpressionsTask=yt.PredictFaceExpressionsTaskBase=void 0;const Tw=Dc(),N_=ir(),Iw=ks(),Aw=op(),Nw=sr(),xp=Ds();class Zd extends N_.ComposableTask{constructor(e,t,n){super();this.parentTask=e,this.input=t,this.extractedFaces=n}}yt.PredictFaceExpressionsTaskBase=Zd;class _w extends Zd{async run(){const e=await this.parentTask,t=await Aw.extractAllFacesAndComputeResults(e,this.input,async n=>await Promise.all(n.map(r=>Nw.nets.faceExpressionNet.predictExpressions(r))),this.extractedFaces);return e.map((n,r)=>Tw.extendWithFaceExpressions(n,t[r]))}withAgeAndGender(){return new xp.PredictAllAgeAndGenderTask(this,this.input)}}yt.PredictAllFaceExpressionsTask=_w;class Rw extends Zd{async run(){const e=await this.parentTask;if(!e)return;const t=await Aw.extractSingleFaceAndComputeResult(e,this.input,n=>Nw.nets.faceExpressionNet.predictExpressions(n),this.extractedFaces);return Tw.extendWithFaceExpressions(e,t)}withAgeAndGender(){return new xp.PredictSingleAgeAndGenderTask(this,this.input)}}yt.PredictSingleFaceExpressionsTask=Rw;class __ extends _w{withAgeAndGender(){return new xp.PredictAllAgeAndGenderWithFaceAlignmentTask(this,this.input)}withFaceDescriptors(){return new Iw.ComputeAllFaceDescriptorsTask(this,this.input)}}yt.PredictAllFaceExpressionsWithFaceAlignmentTask=__;class R_ extends Rw{withAgeAndGender(){return new xp.PredictSingleAgeAndGenderWithFaceAlignmentTask(this,this.input)}withFaceDescriptor(){return new Iw.ComputeSingleFaceDescriptorTask(this,this.input)}}yt.PredictSingleFaceExpressionsWithFaceAlignmentTask=R_});var Ds=S(xt=>{"use strict";Object.defineProperty(xt,"__esModule",{value:!0});xt.PredictSingleAgeAndGenderWithFaceAlignmentTask=xt.PredictAllAgeAndGenderWithFaceAlignmentTask=xt.PredictSingleAgeAndGenderTask=xt.PredictAllAgeAndGenderTask=xt.PredictAgeAndGenderTaskBase=void 0;const Cw=jd(),Ew=Gd(),C_=ir(),Ow=ks(),kw=op(),Dw=sr(),Lp=Fs();class Qd extends C_.ComposableTask{constructor(e,t,n){super();this.parentTask=e,this.input=t,this.extractedFaces=n}}xt.PredictAgeAndGenderTaskBase=Qd;class Fw extends Qd{async run(){const e=await this.parentTask,t=await kw.extractAllFacesAndComputeResults(e,this.input,async n=>await Promise.all(n.map(r=>Dw.nets.ageGenderNet.predictAgeAndGender(r))),this.extractedFaces);return e.map((n,r)=>{const{age:i,gender:o,genderProbability:a}=t[r];return Cw.extendWithAge(Ew.extendWithGender(n,o,a),i)})}withFaceExpressions(){return new Lp.PredictAllFaceExpressionsTask(this,this.input)}}xt.PredictAllAgeAndGenderTask=Fw;class Mw extends Qd{async run(){const e=await this.parentTask;if(!e)return;const{age:t,gender:n,genderProbability:r}=await kw.extractSingleFaceAndComputeResult(e,this.input,i=>Dw.nets.ageGenderNet.predictAgeAndGender(i),this.extractedFaces);return Cw.extendWithAge(Ew.extendWithGender(e,n,r),t)}withFaceExpressions(){return new Lp.PredictSingleFaceExpressionsTask(this,this.input)}}xt.PredictSingleAgeAndGenderTask=Mw;class E_ extends Fw{withFaceExpressions(){return new Lp.PredictAllFaceExpressionsWithFaceAlignmentTask(this,this.input)}withFaceDescriptors(){return new Ow.ComputeAllFaceDescriptorsTask(this,this.input)}}xt.PredictAllAgeAndGenderWithFaceAlignmentTask=E_;class O_ extends Mw{withFaceExpressions(){return new Lp.PredictSingleFaceExpressionsWithFaceAlignmentTask(this,this.input)}withFaceDescriptor(){return new Ow.ComputeSingleFaceDescriptorTask(this,this.input)}}xt.PredictSingleAgeAndGenderWithFaceAlignmentTask=O_});var ks=S(yn=>{"use strict";Object.defineProperty(yn,"__esModule",{value:!0});yn.ComputeSingleFaceDescriptorTask=yn.ComputeAllFaceDescriptorsTask=yn.ComputeFaceDescriptorsTaskBase=void 0;const Uw=Pd(),k_=ir(),Ww=op(),qw=sr(),Bw=Ds(),Pw=Fs();class em extends k_.ComposableTask{constructor(e,t){super();this.parentTask=e,this.input=t}}yn.ComputeFaceDescriptorsTaskBase=em;class D_ extends em{async run(){const e=await this.parentTask,t=await Ww.extractAllFacesAndComputeResults(e,this.input,n=>Promise.all(n.map(r=>qw.nets.faceRecognitionNet.computeFaceDescriptor(r))),null,n=>n.landmarks.align(null,{useDlibAlignment:!0}));return t.map((n,r)=>Uw.extendWithFaceDescriptor(e[r],n))}withFaceExpressions(){return new Pw.PredictAllFaceExpressionsWithFaceAlignmentTask(this,this.input)}withAgeAndGender(){return new Bw.PredictAllAgeAndGenderWithFaceAlignmentTask(this,this.input)}}yn.ComputeAllFaceDescriptorsTask=D_;class F_ extends em{async run(){const e=await this.parentTask;if(!e)return;const t=await Ww.extractSingleFaceAndComputeResult(e,this.input,n=>qw.nets.faceRecognitionNet.computeFaceDescriptor(n),null,n=>n.landmarks.align(null,{useDlibAlignment:!0}));return Uw.extendWithFaceDescriptor(e,t)}withFaceExpressions(){return new Pw.PredictSingleFaceExpressionsWithFaceAlignmentTask(this,this.input)}withAgeAndGender(){return new Bw.PredictSingleAgeAndGenderWithFaceAlignmentTask(this,this.input)}}yn.ComputeSingleFaceDescriptorTask=F_});var Sp=S(vp=>{"use strict";Object.defineProperty(vp,"__esModule",{value:!0});vp.TinyFaceDetectorOptions=void 0;const M_=yp();class U_ extends M_.TinyYolov2Options{constructor(){super(...arguments);this._name="TinyFaceDetectorOptions"}}vp.TinyFaceDetectorOptions=U_});var nm=S(xn=>{"use strict";Object.defineProperty(xn,"__esModule",{value:!0});xn.DetectSingleFaceLandmarksTask=xn.DetectAllFaceLandmarksTask=xn.DetectFaceLandmarksTaskBase=void 0;const Tp=ee(),Ip=tt(),jw=Wr(),W_=ir(),Gw=ks(),$w=sr(),Hw=Ds(),Yw=Fs();class tm extends W_.ComposableTask{constructor(e,t,n){super();this.parentTask=e,this.input=t,this.useTinyLandmarkNet=n}get landmarkNet(){return this.useTinyLandmarkNet?$w.nets.faceLandmark68TinyNet:$w.nets.faceLandmark68Net}}xn.DetectFaceLandmarksTaskBase=tm;class q_ extends tm{async run(){const e=await this.parentTask,t=e.map(i=>i.detection),n=this.input instanceof Tp.Tensor?await Ip.extractFaceTensors(this.input,t):await Ip.extractFaces(this.input,t),r=await Promise.all(n.map(i=>this.landmarkNet.detectLandmarks(i)));return n.forEach(i=>i instanceof Tp.Tensor&&i.dispose()),e.map((i,o)=>jw.extendWithFaceLandmarks(i,r[o]))}withFaceExpressions(){return new Yw.PredictAllFaceExpressionsWithFaceAlignmentTask(this,this.input)}withAgeAndGender(){return new Hw.PredictAllAgeAndGenderWithFaceAlignmentTask(this,this.input)}withFaceDescriptors(){return new Gw.ComputeAllFaceDescriptorsTask(this,this.input)}}xn.DetectAllFaceLandmarksTask=q_;class B_ extends tm{async run(){const e=await this.parentTask;if(!e)return;const{detection:t}=e,n=this.input instanceof Tp.Tensor?await Ip.extractFaceTensors(this.input,[t]):await Ip.extractFaces(this.input,[t]),r=await this.landmarkNet.detectLandmarks(n[0]);return n.forEach(i=>i instanceof Tp.Tensor&&i.dispose()),jw.extendWithFaceLandmarks(e,r)}withFaceExpressions(){return new Yw.PredictSingleFaceExpressionsWithFaceAlignmentTask(this,this.input)}withAgeAndGender(){return new Hw.PredictSingleAgeAndGenderWithFaceAlignmentTask(this,this.input)}withFaceDescriptor(){return new Gw.ComputeSingleFaceDescriptorTask(this,this.input)}}xn.DetectSingleFaceLandmarksTask=B_});var im=S(Ln=>{"use strict";Object.defineProperty(Ln,"__esModule",{value:!0});Ln.DetectSingleFaceTask=Ln.DetectAllFacesTask=Ln.DetectFacesTaskBase=void 0;const zw=tr(),Vw=Sp(),P_=ir(),Kw=nm(),j_=sr(),Jw=Ds(),Xw=Fs();class rm extends P_.ComposableTask{constructor(e,t=new Vw.TinyFaceDetectorOptions()){super();this.input=e,this.options=t}}Ln.DetectFacesTaskBase=rm;class Zw extends rm{async run(){const{input:e,options:t}=this,n=t instanceof Vw.TinyFaceDetectorOptions?r=>j_.nets.tinyFaceDetector.locateFaces(r,t):null;if(!n)throw new Error("detectFaces - expected options to be instance of TinyFaceDetectorOptions | SsdMobilenetv1Options | MtcnnOptions | TinyYolov2Options");return n(e)}runAndExtendWithFaceDetections(){return new Promise(async e=>{const t=await this.run();return e(t.map(n=>zw.extendWithFaceDetection({},n)))})}withFaceLandmarks(e=!1){return new Kw.DetectAllFaceLandmarksTask(this.runAndExtendWithFaceDetections(),this.input,e)}withFaceExpressions(){return new Xw.PredictAllFaceExpressionsTask(this.runAndExtendWithFaceDetections(),this.input)}withAgeAndGender(){return new Jw.PredictAllAgeAndGenderTask(this.runAndExtendWithFaceDetections(),this.input)}}Ln.DetectAllFacesTask=Zw;class G_ extends rm{async run(){const e=await new Zw(this.input,this.options);let t=e[0];return e.forEach(n=>{n.score>t.score&&(t=n)}),t}runAndExtendWithFaceDetection(){return new Promise(async e=>{const t=await this.run();return e(t?zw.extendWithFaceDetection({},t):void 0)})}withFaceLandmarks(e=!1){return new Kw.DetectSingleFaceLandmarksTask(this.runAndExtendWithFaceDetection(),this.input,e)}withFaceExpressions(){return new Xw.PredictSingleFaceExpressionsTask(this.runAndExtendWithFaceDetection(),this.input)}withAgeAndGender(){return new Jw.PredictSingleAgeAndGenderTask(this.runAndExtendWithFaceDetection(),this.input)}}Ln.DetectSingleFaceTask=G_});var ty=S(Zr=>{"use strict";Object.defineProperty(Zr,"__esModule",{value:!0});Zr.detectAllFaces=Zr.detectSingleFace=void 0;const Qw=im(),ey=Sp();function $_(e,t=new ey.TinyFaceDetectorOptions()){return new Qw.DetectSingleFaceTask(e,t)}Zr.detectSingleFace=$_;function H_(e,t=new ey.TinyFaceDetectorOptions()){return new Qw.DetectAllFacesTask(e,t)}Zr.detectAllFaces=H_});var om=S(Ap=>{"use strict";Object.defineProperty(Ap,"__esModule",{value:!0});Ap.euclideanDistance=void 0;function Y_(e,t){if(e.length!==t.length)throw new Error("euclideanDistance: arr1.length !== arr2.length");const n=Array.from(e),r=Array.from(t);return Math.sqrt(n.map((i,o)=>i-r[o]).reduce((i,o)=>i+Math.pow(o,2),0))}Ap.euclideanDistance=Y_});var iy=S(Np=>{"use strict";Object.defineProperty(Np,"__esModule",{value:!0});Np.FaceMatcher=void 0;const ny=Hh(),_p=zh(),z_=om();class ry{constructor(e,t=.6){this._distanceThreshold=t;const n=Array.isArray(e)?e:[e];if(!n.length)throw new Error("FaceRecognizer.constructor - expected atleast one input");let r=1;const i=()=>`person ${r++}`;this._labeledDescriptors=n.map(o=>{if(o instanceof _p.LabeledFaceDescriptors)return o;if(o instanceof Float32Array)return new _p.LabeledFaceDescriptors(i(),[o]);if(o.descriptor&&o.descriptor instanceof Float32Array)return new _p.LabeledFaceDescriptors(i(),[o.descriptor]);throw new Error("FaceRecognizer.constructor - expected inputs to be of type LabeledFaceDescriptors | WithFaceDescriptor<any> | Float32Array | Array<LabeledFaceDescriptors | WithFaceDescriptor<any> | Float32Array>")})}get labeledDescriptors(){return this._labeledDescriptors}get distanceThreshold(){return this._distanceThreshold}computeMeanDistance(e,t){return t.map(n=>z_.euclideanDistance(n,e)).reduce((n,r)=>n+r,0)/(t.length||1)}matchDescriptor(e){return this.labeledDescriptors.map(({descriptors:t,label:n})=>new ny.FaceMatch(n,this.computeMeanDistance(e,t))).reduce((t,n)=>t.distance<n.distance?t:n)}findBestMatch(e){const t=this.matchDescriptor(e);return t.distance<this.distanceThreshold?t:new ny.FaceMatch("unknown",t.distance)}toJSON(){return{distanceThreshold:this.distanceThreshold,labeledDescriptors:this.labeledDescriptors.map(e=>e.toJSON())}}static fromJSON(e){const t=e.labeledDescriptors.map(n=>_p.LabeledFaceDescriptors.fromJSON(n));return new ry(t,e.distanceThreshold)}}Np.FaceMatcher=ry});var sy=S(oy=>{"use strict";Object.defineProperty(oy,"__esModule",{value:!0})});var ay=S(Ut=>{"use strict";Object.defineProperty(Ut,"__esModule",{value:!0});const vn=Be();vn.__exportStar(ir(),Ut);vn.__exportStar(ks(),Ut);vn.__exportStar(ty(),Ut);vn.__exportStar(im(),Ut);vn.__exportStar(nm(),Ut);vn.__exportStar(iy(),Ut);vn.__exportStar(sr(),Ut);vn.__exportStar(sy(),Ut)});var py=S(Qr=>{"use strict";Object.defineProperty(Qr,"__esModule",{value:!0});Qr.createTinyFaceDetector=void 0;const cy=Be(),V_=bp();cy.__exportStar(bp(),Qr);cy.__exportStar(Sp(),Qr);function K_(e){const t=new V_.TinyFaceDetector();return t.extractWeights(e),t}Qr.createTinyFaceDetector=K_});var hy=S(Rp=>{"use strict";Object.defineProperty(Rp,"__esModule",{value:!0});Rp.resizeResults=void 0;const J_=pt(),X_=Xt(),Z_=er(),sm=tr(),ly=Wr();function uy(e,t){const{width:n,height:r}=new J_.Dimensions(t.width,t.height);if(n<=0||r<=0)throw new Error(`resizeResults - invalid dimensions: ${JSON.stringify({width:n,height:r})}`);if(Array.isArray(e))return e.map(i=>uy(i,{width:n,height:r}));if(ly.isWithFaceLandmarks(e)){const i=e.detection.forSize(n,r),o=e.unshiftedLandmarks.forSize(i.box.width,i.box.height);return ly.extendWithFaceLandmarks(sm.extendWithFaceDetection(e,i),o)}return sm.isWithFaceDetection(e)?sm.extendWithFaceDetection(e,e.detection.forSize(n,r)):e instanceof Z_.FaceLandmarks||e instanceof X_.FaceDetection?e.forSize(n,r):e}Rp.resizeResults=uy});var dy=S(ye=>{"use strict";Object.defineProperty(ye,"__esModule",{value:!0});ye.tf=ye.utils=ye.draw=void 0;const Je=Be(),Q_=ee();ye.tf=Q_;const e0=Ob();ye.draw=e0;const t0=me();ye.utils=t0;Je.__exportStar($b(),ye);Je.__exportStar(pt(),ye);Je.__exportStar(tt(),ye);Je.__exportStar(Ye(),ye);Je.__exportStar(_d(),ye);Je.__exportStar(Xb(),ye);Je.__exportStar(ow(),ye);Je.__exportStar(aw(),ye);Je.__exportStar(ay(),ye);Je.__exportStar(pn(),ye);Je.__exportStar(py(),ye);Je.__exportStar(yp(),ye);Je.__exportStar(om(),ye);Je.__exportStar(Qt(),ye);Je.__exportStar(hy(),ye)});const my="tfjsflags";class Fh{constructor(e){this.global=e,this.flags={},this.flagRegistry={},this.urlFlags={},this.populateURLFlags()}setPlatform(e,t){this.platform!=null&&console.warn(`Platform ${this.platformName} has already been set. Overwriting the platform with ${t}.`),this.platformName=e,this.platform=t}registerFlag(e,t,n){if(this.flagRegistry[e]={evaluationFn:t,setHook:n},this.urlFlags[e]!=null){const r=this.urlFlags[e];console.warn(`Setting feature override from URL ${e}: ${r}.`),this.set(e,r)}}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];const t=this.evaluateFlag(e);if(t instanceof Promise)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;const e=n0(this.global.location.search);if(my in e){const t=e[my].split(",");t.forEach(n=>{const[r,i]=n.split(":");this.urlFlags[r]=r0(r,i)})}}}function n0(e){const t={};return e.replace(/[?&]([^=?&]+)(?:=([^&]*))?/g,(n,...r)=>(i0(t,r[0],r[1]),r.join("="))),t}function i0(e,t,n){e[decodeURIComponent(t)]=decodeURIComponent(n||"")}function r0(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 ${e}.`)}function se(){return Mh}let Mh=null;function fy(e){Mh=e}let am;function cm(){if(am==null){let e;if(typeof window!="undefined")e=window;else if(typeof global!="undefined")e=global;else if(typeof process!="undefined")e=process;else if(typeof self!="undefined")e=self;else throw new Error("Could not find a global object");am=e}return am}function o0(){const e=cm();return e._tfGlobals==null&&(e._tfGlobals=new Map()),e._tfGlobals}function pm(e,t){const n=o0();if(n.has(e))return n.get(e);{const r=t();return n.set(e,r),n.get(e)}}const Mi="Abs",Ui="Acos",Wi="Acosh",Jn="Add",qi="AddN",Wu="All",qu="Any",Bi="ArgMax",Pi="ArgMin",ji="Asin",Gi="Asinh",$i="Atan",Hi="Atanh",Yi="Atan2",zi="AvgPool",Bu="AvgPoolBackprop",Vi="AvgPool3D",Pu="AvgPool3DBackprop",Ki="BatchMatMul",Ji="BatchToSpaceND",Xi="BroadcastTo",Xn="Cast",Zi="Ceil",Qi="ClipByValue",ju="Complex",eo="Concat",to="Conv2D",Gu="Conv2DBackpropFilter",no="Conv2DBackpropInput",ro="Conv3D",$u="Conv3DBackpropFilterV2",Hu="Conv3DBackpropInputV2",io="Cos",oo="Cosh",so="Cumsum",Yu="CropAndResize",zu="DepthToSpace",ao="DepthwiseConv2dNative",Vu="DepthwiseConv2dNativeBackpropFilter",Ku="DepthwiseConv2dNativeBackpropInput",Ju="Diag",co="Dilation2D",Xu="Dilation2DBackpropInput",Zu="Dilation2DBackpropFilter",po="Div",lo="Elu",Qu="EluGrad",uo="Erf",eh="Equal",ho="Exp",mo="Expm1",th="FFT",nh="Fill",rh="FlipLeftRight",fo="Floor",go="FloorDiv",bo="FusedBatchNorm",wo="GatherV2",ih="GatherNd",oh="Greater",yo="GreaterEqual",xo="Identity",sh="IFFT",ah="Imag",Lo="IsFinite",vo="IsInf",So="IsNan",ch="Less",ph="LessEqual",lh="LinSpace",To="Log",Io="Log1p",uh="LogicalAnd",hh="LogicalNot",dh="LogicalOr",Ao="LogSoftmax",No="LRN",mh="LRNBackprop",_o="Max",Ro="Maximum",Co="MaxPool",fh="MaxPoolBackprop",Eo="MaxPool3D",gh="MaxPool3DBackprop",bh="MaxPoolWithArgmax",oS="Mean",Oo="Min",ko="Minimum",Do="Mod",Fo="Multiply",Mo="Negate",wh="NotEqual",yh="NonMaxSuppressionV3",xh="NonMaxSuppressionV4",Lh="NonMaxSuppressionV5",Uo="OnesLike",Wo="OneHot",qo="PadV2",sS="Pool",Bo="Pow",Po="Prelu",vh="Prod",Sh="Range",Th="Real",jo="Reciprocal",Go="Relu",$o="Reshape",Ho="ResizeNearestNeighbor",Ih="ResizeNearestNeighborGrad",Yo="ResizeBilinear",Ah="ResizeBilinearGrad",zo="Relu6",Vo="Reverse",Ko="Round",Jo="Rsqrt",Nh="ScatterNd",Xo="SelectV2",Zo="Selu",Qo="Slice",es="Sin",ts="Sinh",ns="Sign",rs="Sigmoid",is="Softplus",os="Sqrt",ss="Sum",as="SpaceToBatchND",cs="SplitV",ps="Softmax",ls="SquaredDifference",_h="Square",us="Sub",Rh="SparseToDense",Ch="StridedSlice",hs="Tan",ds="Tanh",ms="Tile",Eh="TopK",fs="Transpose",gs="Unpack",bs="UnsortedSegmentSum",ws="ZerosLike",ys="Step",ra="FromPixels",Oh="RotateWithOffset",ia="_FusedMatMul",oa="FusedConv2D",sa="FusedDepthwiseConv2D";const ei=pm("kernelRegistry",()=>new Map()),Ms=pm("gradRegistry",()=>new Map());function Fi(e,t){const n=lm(e,t);return ei.get(n)}function ta(e){return Ms.get(e)}function na(e){const t=ei.entries(),n=[];for(;;){const{done:r,value:i}=t.next();if(r)break;const[o,a]=i,[s]=o.split("_");s===e&&n.push(a)}return n}function nS(e){const{kernelName:t,backendName:n}=e,r=lm(t,n);ei.has(r)&&console.warn(`The kernel '${t}' for backend '${n}' is already registered`),ei.set(r,e)}function Uu(e){const{kernelName:t}=e;Ms.has(t)&&(se().getBool("DEBUG")&&console.warn(`Overriding the gradient for '${t}'`)),Ms.set(t,e)}function rS(e,t){const n=lm(e,t);if(!ei.has(n))throw new Error(`The kernel '${e}' for backend '${t}' is not registered`);ei.delete(n)}function iS(e){if(!Ms.has(e))throw new Error(`The gradient '${e}' for backend is not registered`);Ms.delete(e)}function lm(e,t){return`${t}_${e}`}const _f={};Oe(_f,{arraysEqual:()=>Le,assert:()=>f,assertNonNegativeIntegerDimensions:()=>qs,assertNonNull:()=>lt,assertShapesMatch:()=>$,bytesFromStringArray:()=>bm,bytesPerElement:()=>gm,checkConversionForErrors:()=>by,clamp:()=>ti,computeStrides:()=>Bt,createShuffledIndices:()=>d0,decodeString:()=>Ep,distSquared:()=>p0,encodeString:()=>wm,fetch:()=>b0,flatten:()=>Wt,getArrayFromDType:()=>dm,getTypedArrayFromDType:()=>ni,hasEncodingLoss:()=>fm,indexToLoc:()=>y0,inferDtype:()=>ri,inferFromImplicitShape:()=>um,isBoolean:()=>wy,isFunction:()=>qt,isInt:()=>J,isNumber:()=>yy,isScalarShape:()=>l0,isString:()=>Rt,isTypedArray:()=>Ie,isValidDtype:()=>mm,locToIndex:()=>w0,makeOnesTypedArray:()=>Ws,makeZerosNestedTypedArray:()=>g0,makeZerosTypedArray:()=>Pt,nearestDivisor:()=>ii,nearestLargerEven:()=>s0,now:()=>Cp,parseAxisParam:()=>z,randUniform:()=>c0,repeatedTry:()=>m0,rightPad:()=>ar,shuffle:()=>gy,sizeFromShape:()=>Z,sizeToSquarishShape:()=>h0,squeezeShape:()=>hm,sum:()=>a0,tanh:()=>u0,toNestedArray:()=>cr,toTypedArray:()=>Us});function gy(e){let t=e.length,n=0,r=0;for(;t>0;)r=Math.random()*t|0,t--,n=e[t],e[t]=e[r],e[r]=n}function ti(e,t,n){return Math.max(e,Math.min(t,n))}function s0(e){return e%2===0?e:e+1}function a0(e){let t=0;for(let n=0;n<e.length;n++)t+=e[n];return t}function c0(e,t){const n=Math.random();return t*n+(1-n)*e}function p0(e,t){let n=0;for(let r=0;r<e.length;r++){const i=Number(e[r])-Number(t[r]);n+=i*i}return n}function f(e,t){if(!e)throw new Error(typeof t=="string"?t:t())}function $(e,t,n=""){f(Le(e,t),()=>n+` Shapes ${e} and ${t} must match`)}function lt(e){f(e!=null,()=>"The input to the tensor constructor must be a non-null value.")}function Wt(e,t=[],n=!1){if(t==null&&(t=[]),Array.isArray(e)||Ie(e)&&!n)for(let r=0;r<e.length;++r)Wt(e[r],t,n);else t.push(e);return t}function Z(e){if(e.length===0)return 1;let t=e[0];for(let n=1;n<e.length;n++)t*=e[n];return t}function l0(e){return e.length===0}function Le(e,t){if(e===t)return!0;if(e==null||t==null)return!1;if(e.length!==t.length)return!1;for(let n=0;n<e.length;n++)if(e[n]!==t[n])return!1;return!0}function J(e){return e%1===0}function u0(e){if(Math.tanh!=null)return Math.tanh(e);if(e===Infinity)return 1;if(e===-Infinity)return-1;{const t=Math.exp(2*e);return(t-1)/(t+1)}}function h0(e){const t=Math.ceil(Math.sqrt(e));return[t,Math.ceil(e/t)]}function d0(e){const t=new Uint32Array(e);for(let n=0;n<e;++n)t[n]=n;return gy(t),t}function ar(e,t){return t<=e.length?e:e+" ".repeat(t-e.length)}function m0(e,t=r=>0,n){return new Promise((r,i)=>{let o=0;const a=()=>{if(e()){r();return}o++;const s=t(o);if(n!=null&&o>=n){i();return}setTimeout(a,s)};a()})}function um(e,t){let n=1,r=-1;for(let o=0;o<e.length;++o)if(e[o]>=0)n*=e[o];else if(e[o]===-1){if(r!==-1)throw Error(`Shapes can only have 1 implicit size. Found -1 at dim ${r} and dim ${o}`);r=o}else if(e[o]<0)throw Error(`Shapes can not be < 0. Found ${e[o]} at dim ${o}`);if(r===-1){if(t>0&&t!==n)throw Error(`Size(${t}) must match the product of shape ${e}`);return e}if(n===0)throw Error(`Cannot infer the missing size in [${e}] when there are 0 elements`);if(t%n!==0)throw Error(`The implicit shape can't be a fractional number. Got ${t} / ${n}`);const i=e.slice();return i[r]=t/n,i}function z(e,t){const n=t.length;return e=e==null?t.map((r,i)=>i):[].concat(e),f(e.every(r=>r>=-n&&r<n),()=>`All values in axis param must be in range [-${n}, ${n}) but got axis ${e}`),f(e.every(r=>J(r)),()=>`All values in axis param must be integers but got axis ${e}`),e.map(r=>r<0?n+r:r)}function hm(e,t){const n=[],r=[],i=t!=null&&Array.isArray(t)&&t.length===0,o=t==null||i?null:z(t,e).sort();let a=0;for(let s=0;s<e.length;++s){if(o!=null){if(o[a]===s&&e[s]!==1)throw new Error(`Can't squeeze axis ${s} since its dim '${e[s]}' is not 1`);(o[a]==null||o[a]>s)&&e[s]===1&&(n.push(e[s]),r.push(s)),o[a]<=s&&a++}e[s]!==1&&(n.push(e[s]),r.push(s))}return{newShape:n,keptDims:r}}function ni(e,t){let n=null;if(e==null||e==="float32")n=new Float32Array(t);else if(e==="int32")n=new Int32Array(t);else if(e==="bool")n=new Uint8Array(t);else throw new Error(`Unknown data type ${e}`);return n}function dm(e,t){let n=null;if(e==null||e==="float32")n=new Float32Array(t);else if(e==="int32")n=new Int32Array(t);else if(e==="bool")n=new Uint8Array(t);else if(e==="string")n=new Array(t);else throw new Error(`Unknown data type ${e}`);return n}function by(e,t){for(let n=0;n<e.length;n++){const r=e[n];if(isNaN(r)||!isFinite(r))throw Error(`A tensor of type ${t} being uploaded contains ${r}.`)}}function mm(e){return e==="bool"||e==="complex64"||e==="float32"||e==="int32"||e==="string"}function fm(e,t){return t==="complex64"||(t==="float32"&&e!=="complex64"||t==="int32"&&e!=="float32"&&e!=="complex64")?!1:!(t==="bool"&&e==="bool")}function Ie(e){return e instanceof Float32Array||e instanceof Int32Array||e instanceof Uint8Array}function gm(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 bm(e){if(e==null)return 0;let t=0;return e.forEach(n=>t+=n.length),t}function Rt(e){return typeof e=="string"||e instanceof String}function wy(e){return typeof e=="boolean"}function yy(e){return typeof e=="number"}function ri(e){return Array.isArray(e)?ri(e[0]):e instanceof Float32Array?"float32":e instanceof Int32Array||e instanceof Uint8Array?"int32":yy(e)?"float32":Rt(e)?"string":wy(e)?"bool":"float32"}function qt(e){return!!(e&&e.constructor&&e.call&&e.apply)}function ii(e,t){for(let n=t;n<e;++n)if(e%n===0)return n;return e}function Bt(e){const t=e.length;if(t<2)return[];const n=new Array(t-1);n[t-2]=e[t-1];for(let r=t-3;r>=0;--r)n[r]=n[r+1]*e[r+1];return n}function Us(e,t){if(t==="string")throw new Error("Cannot convert a string[] to a TypedArray");if(Array.isArray(e)&&(e=Wt(e)),se().getBool("DEBUG")&&by(e,t),f0(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"){const n=new Uint8Array(e.length);for(let r=0;r<n.length;++r)Math.round(e[r])!==0&&(n[r]=1);return n}else throw new Error(`Unknown data type ${t}`)}function xy(e,t,n){const r=new Array();if(t.length===1){const i=t[0];for(let o=0;o<i;o++)r[o]=n[e+o]}else{const i=t[0],o=t.slice(1),a=o.reduce((s,c)=>s*c);for(let s=0;s<i;s++)r[s]=xy(e+s*a,o,n)}return r}function cr(e,t){if(e.length===0)return t[0];const n=e.reduce((r,i)=>r*i);if(n===0)return[];if(n!==t.length)throw new Error(`[${e}] does not match the input size ${t.length}.`);return xy(0,e,t)}function f0(e,t){return e instanceof Float32Array&&t==="float32"||e instanceof Int32Array&&t==="int32"||e instanceof Uint8Array&&t==="bool"}function Ws(e,t){const n=Pt(e,t);for(let r=0;r<n.length;r++)n[r]=1;return n}function Pt(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 g0(e,t){const n=e.reduce((r,i)=>r*i,1);if(t==null||t==="float32")return cr(e,new Float32Array(n));if(t==="int32")return cr(e,new Int32Array(n));if(t==="bool")return cr(e,new Uint8Array(n));throw new Error(`Unknown data type ${t}`)}function Cp(){return se().platform.now()}function qs(e){e.forEach(t=>{f(Number.isInteger(t)&&t>=0,()=>`Tensor must have a shape comprised of positive integers but got shape [${e}].`)})}function b0(e,t){return se().platform.fetch(e,t)}function wm(e,t="utf-8"){return t=t||"utf-8",se().platform.encode(e,t)}function Ep(e,t="utf-8"){return t=t||"utf-8",se().platform.decode(e,t)}function w0(e,t,n){if(t===0)return 0;if(t===1)return e[0];let r=e[e.length-1];for(let i=0;i<e.length-1;++i)r+=n[i]*e[i];return r}function y0(e,t,n){if(t===0)return[];if(t===1)return[e];const r=new Array(t);for(let i=0;i<r.length-1;++i)r[i]=Math.floor(e/n[i]),e-=r[i]*n[i];return r[r.length-1]=e,r}class Ly{constructor(e,t){this.backendTimer=e,this.logger=t,t==null&&(this.logger=new L0())}profileKernel(e,t,n){let r;const i=()=>{r=n()},o=this.backendTimer.time(i);r.map(s=>{s.data().then(c=>{x0(c,s.dtype,e)})});const a={kernelName:e,outputs:r,inputs:t,timeMs:o.then(s=>s.kernelMs),extraInfo:o.then(s=>s.getExtraProfileInfo!=null?s.getExtraProfileInfo():"")};return a}logKernelProfile(e){const{kernelName:t,outputs:n,timeMs:r,inputs:i,extraInfo:o}=e;n.forEach(a=>{Promise.all([a.data(),r,o]).then(s=>{this.logger.logKernelProfile(t,a,s[0],s[1],i,s[2])})})}}function x0(e,t,n){if(t!=="float32")return!1;for(let r=0;r<e.length;r++){const i=e[r];if(isNaN(i)||!isFinite(i))return console.warn(`Found ${i} in the result of '${n}'`),!0}return!1}class L0{logKernelProfile(e,t,n,r,i,o){const a=typeof r=="number"?ar(`${r}ms`,9):r.error,s=ar(e,25),c=t.rank,p=t.size,l=ar(t.shape.toString(),14);let u="";for(const m in i){const b=i[m];if(b!=null){const y=b.shape||t.shape,w=y.length;u+=`${m}: ${w}D ${w>0?y:""} `}}console.log(`%c${s} %c${a} %c${c}D ${l} %c${p} %c${u} %c${o}`,"font-weight:bold","color:red","color:blue","color: orange","color: green","color: steelblue")}}function vy(e,t,n){const r={},i={};for(let c=0;c<t.length;c++)r[t[c].id]=!0;for(let c=0;c<e.length;c++){const p=e[c],l=p.inputs;for(const u in l){const m=l[u];let b=!1;for(let y=0;y<t.length;y++)if(r[m.id]){p.outputs.forEach(w=>r[w.id]=!0),b=!0,i[p.id]=!0;break}if(b)break}}const o={};o[n.id]=!0;const a={};for(let c=e.length-1;c>=0;c--){const p=e[c],l=p.inputs;for(let u=0;u<p.outputs.length;u++)if(o[p.outputs[u].id]){for(const m in l)o[l[m].id]=!0,a[p.id]=!0;break}}const s=[];for(let c=0;c<e.length;c++){const p=e[c];if(i[p.id]&&a[p.id]){const l={};for(const m in p.inputs){const b=p.inputs[m];r[b.id]&&(l[m]=b)}const u=Object.assign({},p);u.inputs=l,u.outputs=p.outputs,s.push(u)}}return s}function Sy(e,t,n,r){for(let i=t.length-1;i>=0;i--){const o=t[i],a=[];if(o.outputs.forEach(c=>{const p=e[c.id];p!=null?a.push(p):a.push(null)}),o.gradient==null)throw new Error(`Cannot compute gradient: gradient function not found for ${o.kernelName}.`);const s=o.gradient(a);for(const c in o.inputs){if(!(c in s))throw new Error(`Cannot backprop through input ${c}. Available gradients found: ${Object.keys(s)}.`);const p=n(()=>s[c]());if(p.dtype!=="float32")throw new Error(`Error in gradient for op ${o.kernelName}. The gradient of input ${c} must have 'float32' dtype, but has '${p.dtype}'`);const l=o.inputs[c];if(!Le(p.shape,l.shape))throw new Error(`Error in gradient for op ${o.kernelName}. The gradient of input '${c}' has shape '${p.shape}', which does not match the shape of the input '${l.shape}'`);if(e[l.id]==null)e[l.id]=p;else{const u=e[l.id];e[l.id]=r(u,p),u.dispose()}}}}const Ty=20,Bs=3,ym=7;function Iy(e,t,n,r){const i=Bt(t),o=v0(e,t,n,i),a=t.length,s=Op(e,t,n,i,o),c=["Tensor"];return r&&(c.push(` dtype: ${n}`),c.push(` rank: ${a}`),c.push(` shape: [${t}]`),c.push(" values:")),c.push(s.map(p=>" "+p).join(`
`)),c.join(`
`)}function v0(e,t,n,r){const i=Z(t),o=r[r.length-1],a=new Array(o).fill(0),s=t.length,c=n==="complex64"?js(e):e;if(s>1)for(let p=0;p<i/o;p++){const l=p*o;for(let u=0;u<o;u++)a[u]=Math.max(a[u],Ps(c[l+u],0,n).length)}return a}function Ps(e,t,n){let r;return Array.isArray(e)?r=`${parseFloat(e[0].toFixed(ym))} + ${parseFloat(e[1].toFixed(ym))}j`:Rt(e)?r=`'${e}'`:n==="bool"?r=Ay(e):r=parseFloat(e.toFixed(ym)).toString(),ar(r,t)}function Ay(e){return e===0?"false":"true"}function Op(e,t,n,r,i,o=!0){const a=n==="complex64"?2:1,s=t[0],c=t.length;if(c===0){if(n==="complex64"){const w=js(e);return[Ps(w[0],0,n)]}return n==="bool"?[Ay(e[0])]:[e[0].toString()]}if(c===1){if(s>Ty){const L=Bs*a;let T=Array.from(e.slice(0,L)),A=Array.from(e.slice((s-Bs)*a,s*a));return n==="complex64"&&(T=js(T),A=js(A)),["["+T.map((_,N)=>Ps(_,i[N],n)).join(", ")+", ..., "+A.map((_,N)=>Ps(_,i[s-Bs+N],n)).join(", ")+"]"]}const w=n==="complex64"?js(e):Array.from(e);return["["+w.map((L,T)=>Ps(L,i[T],n)).join(", ")+"]"]}const p=t.slice(1),l=r.slice(1),u=r[0]*a,m=[];if(s>Ty){for(let w=0;w<Bs;w++){const L=w*u,T=L+u;m.push(...Op(e.slice(L,T),p,n,l,i,!1))}m.push("...");for(let w=s-Bs;w<s;w++){const L=w*u,T=L+u;m.push(...Op(e.slice(L,T),p,n,l,i,w===s-1))}}else for(let w=0;w<s;w++){const L=w*u,T=L+u;m.push(...Op(e.slice(L,T),p,n,l,i,w===s-1))}const b=c===2?",":"";m[0]="["+m[0]+b;for(let w=1;w<m.length-1;w++)m[w]=" "+m[w]+b;let y=`,
`;for(let w=2;w<c;w++)y+=`
`;return m[m.length-1]=" "+m[m.length-1]+"]"+(o?"":y),m}function js(e){const t=[];for(let n=0;n<e.length;n+=2)t.push([e[n],e[n+1]]);return t}class yr{constructor(e,t,n){if(this.dtype=t,this.shape=e.slice(),this.size=Z(e),n!=null){const r=n.length;f(r===this.size,()=>`Length of values '${r}' does not match the size inferred by the shape '${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=n||dm(t,this.size),this.strides=Bt(e)}set(e,...t){t.length===0&&(t=[0]),f(t.length===this.rank,()=>`The number of provided coordinates (${t.length}) must match the rank (${this.rank})`);const n=this.locToIndex(t);this.values[n]=e}get(...e){e.length===0&&(e=[0]);let t=0;for(const r of e){if(r<0||r>=this.shape[t]){const i=`Requested out of range element at ${e}. Buffer shape=${this.shape}`;throw new Error(i)}t++}let n=e[e.length-1];for(let r=0;r<e.length-1;++r)n+=this.strides[r]*e[r];return this.values[n]}locToIndex(e){if(this.rank===0)return 0;if(this.rank===1)return e[0];let t=e[e.length-1];for(let n=0;n<e.length-1;++n)t+=this.strides[n]*e[n];return t}indexToLoc(e){if(this.rank===0)return[];if(this.rank===1)return[e];const t=new Array(this.shape.length);for(let n=0;n<t.length-1;++n)t[n]=Math.floor(e/this.strides[n]),e-=t[n]*this.strides[n];return t[t.length-1]=e,t}get rank(){return this.shape.length}toTensor(){return jt().makeTensor(this.values,this.shape,this.dtype)}}let jt=null,oi=null,S0=null;function Ny(e){jt=e}function _y(e){oi=e}function Ry(e){S0=e}class de{constructor(e,t,n,r){this.kept=!1,this.isDisposedInternal=!1,this.shape=e.slice(),this.dtype=t||"float32",this.size=Z(e),this.strides=Bt(e),this.dataId=n,this.id=r,this.rankType=this.rank<5?this.rank.toString():"higher"}get rank(){return this.shape.length}async buffer(){const e=await this.data();return oi.buffer(this.shape,this.dtype,e)}bufferSync(){return oi.buffer(this.shape,this.dtype,this.dataSync())}async array(){const e=await this.data();return cr(this.shape,e)}arraySync(){return cr(this.shape,this.dataSync())}async data(){this.throwIfDisposed();const e=jt().read(this.dataId);if(this.dtype==="string"){const t=await e;try{return t.map(n=>Ep(n))}catch(n){throw new Error("Failed to decode the string bytes into utf-8. To get the original bytes, call tensor.bytes().")}}return e}dataSync(){this.throwIfDisposed();const e=jt().readSync(this.dataId);if(this.dtype==="string")try{return e.map(t=>Ep(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();const e=await jt().read(this.dataId);return this.dtype==="string"?e:new Uint8Array(e.buffer)}dispose(){if(this.isDisposed)return;jt().disposeTensor(this),this.isDisposedInternal=!0}get isDisposed(){return this.isDisposedInternal}throwIfDisposed(){if(this.isDisposed)throw new Error("Tensor is disposed.")}print(e=!1){return oi.print(this,e)}clone(){return this.throwIfDisposed(),oi.clone(this)}toString(e=!1){const t=this.dataSync();return Iy(t,this.shape,this.dtype,e)}cast(e){return this.throwIfDisposed(),oi.cast(this,e)}variable(e=!0,t,n){return this.throwIfDisposed(),jt().makeVariable(this,e,t,n)}}Object.defineProperty(de,Symbol.hasInstance,{value:e=>!!e&&e.dataId!=null&&e.shape!=null&&e.dtype!=null});class Nn extends de{constructor(e,t,n,r){super(e.shape,e.dtype,e.dataId,r);this.trainable=t,this.name=n}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(!Le(e.shape,this.shape))throw new Error(`shape of the new value (${e.shape}) and previous value (${this.shape}) must match`);jt().disposeTensor(this),this.dataId=e.dataId,jt().incRef(this,null)}dispose(){jt().disposeVariable(this),this.isDisposedInternal=!0}}Object.defineProperty(Nn,Symbol.hasInstance,{value:e=>e instanceof de&&e.assign!=null&&e.assign instanceof Function});var al;(function(e){e.R0="R0",e.R1="R1",e.R2="R2",e.R3="R3",e.R4="R4",e.R5="R5",e.R6="R6"})(al||(al={}));var xm;(function(e){e.float32="float32",e.int32="int32",e.bool="int32",e.complex64="complex64"})(xm||(xm={}));var Lm;(function(e){e.float32="float32",e.int32="int32",e.bool="bool",e.complex64="complex64"})(Lm||(Lm={}));var vm;(function(e){e.float32="float32",e.int32="float32",e.bool="float32",e.complex64="complex64"})(vm||(vm={}));var Sm;(function(e){e.float32="complex64",e.int32="complex64",e.bool="complex64",e.complex64="complex64"})(Sm||(Sm={}));const T0={float32:vm,int32:xm,bool:Lm,complex64:Sm};function wr(e,t){if(e==="string"||t==="string"){if(e==="string"&&t==="string")return"string";throw new Error(`Can not upcast ${e} with ${t}`)}return T0[e][t]}function eS(e){return wr(e,"int32")}const Cf={};Oe(Cf,{assertTypesMatch:()=>Tm,getTensorsInContainer:()=>Gs,isTensorInList:()=>I0,makeTypesMatch:()=>V});function V(e,t){if(e.dtype===t.dtype)return[e,t];const n=wr(e.dtype,t.dtype);return[e.cast(n),t.cast(n)]}function Tm(e,t){f(e.dtype===t.dtype,()=>`The dtypes of the first(${e.dtype}) and second(${t.dtype}) input must match`)}function I0(e,t){return t.some(n=>n.id===e.id)}function Gs(e){const t=[],n=new Set();return Cy(e,t,n),t}function Cy(e,t,n){if(e==null)return;if(e instanceof de){t.push(e);return}if(!A0(e))return;const r=e;for(const i in r){const o=r[i];n.has(o)||(n.add(o),Cy(o,t,n))}}function A0(e){return Array.isArray(e)||typeof e=="object"}class Ey{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}}dispose(){for(const e in this.registeredVariables)this.registeredVariables[e].dispose()}}class $s{constructor(e){this.ENV=e,this.registry={},this.registryFactory={},this.pendingBackendInitId=0,this.state=new Ey()}async ready(){if(this.pendingBackendInit!=null)return this.pendingBackendInit.then(()=>{});if(this.backendInstance!=null)return;const e=this.getSortedBackends();for(let t=0;t<e.length;t++){const n=e[t],r=await this.initializeBackend(n).success;if(r){await this.setBackend(n);return}}throw new Error("Could not initialize any backends, all backend initializations failed.")}get backend(){if(this.pendingBackendInit!=null)throw new Error(`Backend '${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){const{name:e,asyncInit:t}=this.initializeBackendsAndReturnBest();if(t)throw new Error(`The highest priority backend '${e}' has not yet been initialized. Make sure to await tf.ready() or await tf.setBackend() before calling other methods`);this.setBackend(e)}return this.backendInstance}backendNames(){return Object.keys(this.registryFactory)}findBackend(e){if(!(e in this.registry))if(e in this.registryFactory){const{asyncInit:t}=this.initializeBackend(e);if(t)return null}else return null;return this.registry[e]}findBackendFactory(e){return e in this.registryFactory?this.registryFactory[e].factory:null}registerBackend(e,t,n=1){return e in this.registryFactory?(console.warn(`${e} backend was already registered. Reusing existing backend factory.`),!1):(this.registryFactory[e]={factory:t,priority:n},!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;const{success:t,asyncInit:n}=this.initializeBackend(e),r=n?await t:t;if(!r)return!1}return this.backendInstance=this.registry[e],this.setupRegisteredKernels(),this.profiler=new Ly(this.backendInstance),!0}setupRegisteredKernels(){const e=na(this.backendName);e.forEach(t=>{t.setupFunc!=null&&t.setupFunc(this.backendInstance)})}disposeRegisteredKernels(e){const t=na(e);t.forEach(n=>{n.disposeFunc!=null&&n.disposeFunc(this.registry[e])})}initializeBackend(e){const t=this.registryFactory[e];if(t==null)throw new Error(`Cannot initialize backend ${e}, no registration found.`);try{const n=t.factory();if(Promise.resolve(n)===n){const r=++this.pendingBackendInitId,i=n.then(o=>r<this.pendingBackendInitId?!1:(this.registry[e]=o,this.pendingBackendInit=null,!0)).catch(o=>(r<this.pendingBackendInitId||(this.pendingBackendInit=null,console.warn(`Initialization of backend ${e} failed`),console.warn(o.stack||o.message)),!1));return this.pendingBackendInit=i,{success:i,asyncInit:!0}}else return this.registry[e]=n,{success:!0,asyncInit:!1}}catch(n){return console.warn(`Initialization of backend ${e} failed`),console.warn(n.stack||n.message),{success:!1,asyncInit:!1}}}removeBackend(e){if(!(e in this.registryFactory))throw new Error(`${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(){const e=this.getSortedBackends();for(let t=0;t<e.length;t++){const n=e[t],{success:r,asyncInit:i}=this.initializeBackend(n);if(i||r)return{name:n,asyncInit:i}}throw new Error("Could not initialize any backends, all backend initializations failed.")}moveData(e,t){const n=this.state.tensorInfo.get(t),r=n.backend,i=this.readSync(t);r.disposeData(t),n.backend=e,e.move(t,i,n.shape,n.dtype),this.shouldCheckForMemLeaks()&&this.state.numDataMovesStack[this.state.numDataMovesStack.length-1]++}tidy(e,t){let n=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");n=e}let r;return this.scopedRun(()=>this.startScope(n),()=>this.endScope(r),()=>(r=t(),r instanceof Promise&&console.error("Cannot return a Promise inside of tidy."),r))}scopedRun(e,t,n){e();try{const r=n();return t(),r}catch(r){throw t(),r}}nextTensorId(){return $s.nextTensorId++}nextVariableId(){return $s.nextVariableId++}clone(e){const t=this.makeTensorFromDataId(e.dataId,e.shape,e.dtype),n={x:e},r=o=>({x:()=>{const a="float32",s={x:o},c={dtype:a};return g.runKernelFunc(p=>p.cast(o,a),s,null,Xn,c)}}),i=[];return this.addTapeNode(this.state.activeScope.name,n,[t],r,i,{}),t}runKernel(e,t,n,r,i){const o=null,a=null;return this.runKernelFunc(o,t,a,e,n,r,i)}shouldCheckForMemLeaks(){return this.ENV.getBool("IS_TEST")}checkKernelForMemLeak(e,t,n){const r=this.backend.numDataIds();let i=0;n.forEach(s=>{i+=s.dtype==="complex64"?3:1});const o=this.state.numDataMovesStack[this.state.numDataMovesStack.length-1],a=r-t-i-o;if(a>0)throw new Error(`Backend '${this.backendName}' has an internal memory leak (${a} data ids) after running '${e}'`)}runKernelFunc(e,t,n,r,i,o,a){let s,c=[];const p=this.isTapeOn();r==null&&(r=this.state.activeScope!=null?this.state.activeScope.name:"");const l=this.state.numBytes,u=this.state.numTensors;this.shouldCheckForMemLeaks()&&this.state.numDataMovesStack.push(0);let m;const b=Fi(r,this.backendName);let y;if(b!=null)m=()=>{const L=this.backend.numDataIds();y=b.kernelFunc({inputs:t,attrs:i,backend:this.backend});const T=Array.isArray(y)?y:[y];this.shouldCheckForMemLeaks()&&this.checkKernelForMemLeak(r,L,T);const A=T.map(({dataId:_,shape:N,dtype:C})=>this.makeTensorFromDataId(_,N,C));if(p){let _=this.getTensorsForGradient(r,t,A);if(_==null){a==null&&(a=[]);const N=A.filter((C,F)=>a[F]);_=(o||[]).slice().concat(N)}c=this.saveTensorsForBackwardMode(_)}return A};else{const L=T=>{if(!p)return;c=T.map(A=>this.keep(this.clone(A)))};m=()=>{const T=this.backend.numDataIds();y=this.tidy(()=>e(this.backend,L));const A=Array.isArray(y)?y:[y];return this.shouldCheckForMemLeaks()&&this.checkKernelForMemLeak(r,T,A),A}}let w;return this.scopedRun(()=>this.state.kernelDepth++,()=>this.state.kernelDepth--,()=>{!this.ENV.getBool("DEBUG")&&!this.state.profiling?s=m():(w=this.profiler.profileKernel(r,t,()=>m()),this.ENV.getBool("DEBUG")&&this.profiler.logKernelProfile(w),s=w.outputs)}),p&&this.addTapeNode(r,t,s,n,c,i),this.state.profiling&&this.state.activeProfile.kernels.push({name:r,bytesAdded:this.state.numBytes-l,totalBytesSnapshot:this.state.numBytes,tensorsAdded:this.state.numTensors-u,totalTensorsSnapshot:this.state.numTensors,inputShapes:Object.keys(t).map(L=>t[L]!=null?t[L].shape:null),outputShapes:s.map(L=>L.shape),kernelTimeMs:w.timeMs,extraInfo:w.extraInfo}),Array.isArray(y)?s:s[0]}saveTensorsForBackwardMode(e){const t=e.map(n=>this.keep(this.clone(n)));return t}getTensorsForGradient(e,t,n){const r=ta(e);if(r!=null){const i=r.inputsToSave||[],o=r.outputsToSave||[];let a;r.saveAllInputs?(f(Array.isArray(t),()=>"saveAllInputs is true, expected inputs to be an array."),a=Object.keys(t).map(c=>t[c])):a=i.map(c=>t[c]);const s=n.filter((c,p)=>o[p]);return a.concat(s)}return null}makeTensor(e,t,n,r){if(e==null)throw new Error("Values passed to engine.makeTensor() are null");n=n||"float32",r=r||this.backend;let i=e;n==="string"&&Rt(e[0])&&(i=e.map(s=>wm(s)));const o=r.write(i,t,n),a=new de(t,n,o,this.nextTensorId());if(this.incRef(a,r),n==="string"){const s=this.state.tensorInfo.get(o),c=bm(i);this.state.numBytes+=c-s.bytes,s.bytes=c}return a}makeTensorFromDataId(e,t,n,r){n=n||"float32";const i=new de(t,n,e,this.nextTensorId());return this.incRef(i,r),i}makeVariable(e,t=!0,n,r){n=n||this.nextVariableId().toString(),r!=null&&r!==e.dtype&&(e=e.cast(r));const i=new Nn(e,t,n,this.nextTensorId());if(this.state.registeredVariables[i.name]!=null)throw new Error(`Variable with name ${i.name} was already registered`);return this.state.registeredVariables[i.name]=i,this.incRef(i,this.backend),i}incRef(e,t){const n=this.state.tensorInfo.has(e.dataId)?this.state.tensorInfo.get(e.dataId).refCount:0;if(this.state.numTensors++,e.dtype==="string"&&this.state.numStringTensors++,n===0){this.state.numDataBuffers++;let r=0;e.dtype!=="complex64"&&e.dtype!=="string"&&(r=e.size*gm(e.dtype)),this.state.tensorInfo.set(e.dataId,{backend:t||this.backend,dtype:e.dtype,shape:e.shape,bytes:r,refCount:0}),this.state.numBytes+=r}this.state.tensorInfo.get(e.dataId).refCount++,e instanceof Nn||this.track(e)}disposeTensor(e){if(!this.state.tensorInfo.has(e.dataId))return;this.state.numTensors--,e.dtype==="string"&&this.state.numStringTensors--;const t=this.state.tensorInfo.get(e.dataId),n=t.refCount;n<=1?(e.dtype!=="complex64"&&(this.state.numBytes-=t.bytes),this.state.numDataBuffers--,t.backend.disposeData(e.dataId),this.state.tensorInfo.delete(e.dataId)):this.state.tensorInfo.get(e.dataId).refCount--}disposeVariables(){for(const e in this.state.registeredVariables){const 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(){const 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;const t=this.state.numBytes,n=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(r=>r.totalBytesSnapshot)),this.state.activeProfile.newBytes=this.state.numBytes-t,this.state.activeProfile.newTensors=this.state.numTensors-n;for(const r of this.state.activeProfile.kernels)r.kernelTimeMs=await r.kernelTimeMs,r.extraInfo=await r.extraInfo;return this.state.activeProfile}isTapeOn(){return this.state.gradientDepth>0&&this.state.kernelDepth===0}addTapeNode(e,t,n,r,i,o){const a={id:this.state.nextTapeNodeId++,kernelName:e,inputs:t,outputs:n,saved:i},s=ta(e);s!=null&&(r=s.gradFunc),r!=null&&(a.gradient=c=>(c=c.map((p,l)=>{if(p==null){const u=n[l],m=Pt(u.size,u.dtype);return this.makeTensor(m,u.shape,u.dtype)}return p}),r(c.length>1?c:c[0],i,o))),this.state.activeTape.push(a)}keep(e){return e.kept=!0,e}startTape(){this.state.gradientDepth===0&&(this.state.activeTape=[]),this.state.gradientDepth++}endTape(){this.state.gradientDepth--}startScope(e){const t={track:[],name:"unnamed scope",id:this.state.nextScopeId++};e&&(t.name=e),this.state.scopeStack.push(t),this.state.activeScope=t}endScope(e){const t=Gs(e),n=new Set(t.map(i=>i.id));for(let i=0;i<this.state.activeScope.track.length;i++){const o=this.state.activeScope.track[i];!o.kept&&!n.has(o.id)&&o.dispose()}const r=this.state.scopeStack.pop();this.state.activeScope=this.state.scopeStack.length===0?null:this.state.scopeStack[this.state.scopeStack.length-1],t.forEach(i=>{!i.kept&&i.scopeId===r.id&&this.track(i)})}gradients(e,t,n,r=!1){if(f(t.length>0,()=>"gradients() received an empty list of xs."),n!=null&&n.dtype!=="float32")throw new Error(`dy must have 'float32' dtype, but has '${n.dtype}'`);const i=this.scopedRun(()=>this.startTape(),()=>this.endTape(),()=>this.tidy("forward",e));f(i instanceof de,()=>"The result y returned by f() must be a tensor.");const o=vy(this.state.activeTape,t,i);if(!r&&o.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",()=>{const a={};a[i.id]=n==null?N0(i.shape):n,Sy(a,o,c=>this.tidy(c),_0);const s=t.map(c=>a[c.id]);return this.state.gradientDepth===0&&(this.state.activeTape.forEach(c=>{for(const p of c.saved)p.dispose()}),this.state.activeTape=null),{value:i,grads:s}})}customGrad(e){return f(qt(e),()=>"The f passed in customGrad(f) must be a function."),(...t)=>{f(t.every(i=>i instanceof de),()=>"The args passed in customGrad(f)(x1, x2,...) must all be tensors");let n;const r={};return t.forEach((i,o)=>{r[o]=i}),this.runKernelFunc((i,o)=>(n=e(...t,o),f(n.value instanceof de,()=>"The function f passed in customGrad(f) must return an object where `obj.value` is a tensor"),f(qt(n.gradFunc),()=>"The function f passed in customGrad(f) must return an object where `obj.gradFunc` is a function."),n.value),r,(i,o)=>{const a=n.gradFunc(i,o),s=Array.isArray(a)?a:[a];f(s.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(...)."),f(s.every(p=>p instanceof de),()=>"The function f passed in customGrad(f) must return an object where `obj.gradFunc` is a function that returns a list of only tensors.");const c={};return s.forEach((p,l)=>{c[l]=()=>p}),c})}}readSync(e){const t=this.state.tensorInfo.get(e);return t.backend.readSync(e)}read(e){const t=this.state.tensorInfo.get(e);return t.backend.read(e)}async time(e){const t=Cp(),n=await this.backend.time(e);return n.wallMs=Cp()-t,n}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 Ey();for(const e in this.registry)this.disposeRegisteredKernels(e),this.registry[e].dispose(),delete this.registry[e];this.backendName=null,this.backendInstance=null,this.pendingBackendInit=null}}$s.nextTensorId=0;$s.nextVariableId=0;function N0(e){const t=Ws(Z(e),"float32");return g.makeTensor(t,e,"float32")}function R0(){const e=cm();if(e._tfengine==null){const t=new Fh(e);e._tfengine=new $s(t)}return fy(e._tfengine.ENV),Ny(()=>e._tfengine),e._tfengine}const g=R0();function _0(e,t){const n={a:e,b:t};return g.runKernelFunc((r,i)=>{const o=r.add(e,t);return i([e,t]),o},n,null,Jn)}const ug={};Oe(ug,{isBrowser:()=>Im,isMobile:()=>E0});function C0(){return typeof navigator!="undefined"&&navigator!=null}function E0(){if(C0()){const e=navigator.userAgent||navigator.vendor||window.opera;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(e)||/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(e.substr(0,4))}return!1}function Im(){return typeof window!="undefined"&&window.document!=null||typeof WorkerGlobalScope!="undefined"}const nn=se();nn.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.")});nn.registerFlag("IS_BROWSER",()=>Im());nn.registerFlag("IS_NODE",()=>typeof process!="undefined"&&typeof process.versions!="undefined"&&typeof process.versions.node!="undefined");nn.registerFlag("IS_CHROME",()=>typeof navigator!="undefined"&&navigator!=null&&navigator.userAgent!=null&&/Chrome/.test(navigator.userAgent)&&/Google Inc/.test(navigator.vendor));nn.registerFlag("PROD",()=>!1);nn.registerFlag("TENSORLIKE_CHECK_SHAPE_CONSISTENCY",()=>nn.getBool("DEBUG"));nn.registerFlag("DEPRECATION_WARNINGS_ENABLED",()=>!0);nn.registerFlag("IS_TEST",()=>!1);function Fe(e,t){let n=e;if(Ie(e))return t==="string"?[]:[e.length];if(!Array.isArray(e))return[];const r=[];for(;Array.isArray(n)||Ie(n)&&t!=="string";)r.push(n.length),n=n[0];return Array.isArray(e)&&se().getBool("TENSORLIKE_CHECK_SHAPE_CONSISTENCY")&&Oy(e,r,[]),r}function Oy(e,t,n){if(n=n||[],!Array.isArray(e)&&!Ie(e)){f(t.length===0,()=>`Element arr[${n.join("][")}] is a primitive, but should be an array/TypedArray of ${t[0]} elements`);return}f(t.length>0,()=>`Element arr[${n.join("][")}] should be a primitive, but is an array of ${e.length} elements`),f(e.length===t[0],()=>`Element arr[${n.join("][")}] should have ${t[0]} elements, but has ${e.length} elements`);const r=t.slice(1);for(let i=0;i<e.length;++i)Oy(e[i],r,n.concat(i))}function ky(e,t,n,r){if(e==null)return;if(e!=="numeric"&&e!==t||e==="numeric"&&t==="string")throw new Error(`Argument '${n}' passed to '${r}' must be ${e} tensor, but got ${t} tensor`)}function h(e,t,n,r="numeric"){if(e instanceof de)return ky(r,e.dtype,t,n),e;let i=ri(e);if(i!=="string"&&["bool","int32","float32"].indexOf(r)>=0&&(i=r),ky(r,i,t,n),e==null||!Ie(e)&&!Array.isArray(e)&&typeof e!="number"&&typeof e!="boolean"&&typeof e!="string"){const c=e==null?"null":e.constructor.name;throw new Error(`Argument '${t}' passed to '${n}' must be a Tensor or TensorLike, but got '${c}'`)}const o=Fe(e,i);!Ie(e)&&!Array.isArray(e)&&(e=[e]);const a=!0,s=i!=="string"?Us(e,i):Wt(e,[],a);return g.makeTensor(s,o,i)}function rn(e,t,n,r="numeric"){if(!Array.isArray(e))throw new Error(`Argument ${t} passed to ${n} must be a \`Tensor[]\` or \`TensorLike[]\``);const i=e;return i.map((o,a)=>h(o,`${t}[${a}]`,n),r)}function d(e){const 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 n=t[0];const r=e[n];n.endsWith("_")&&(n=n.substring(0,n.length-1));const i=(...o)=>{g.startScope(n);try{const a=r(...o);return a instanceof Promise&&console.error("Cannot return a Promise inside of tidy."),g.endScope(a),a}catch(a){throw g.endScope(null),a}};return Object.defineProperty(i,"name",{value:n,configurable:!0}),i}function O0(e){const t=h(e,"x","abs"),n={x:t};return g.runKernelFunc((r,i)=>(i([t]),t.dtype==="complex64"?r.complexAbs(t):r.abs(t)),n,null,Mi)}const _e=d({abs_:O0});function k0(e){const t=h(e,"x","acos"),n={x:t};return g.runKernelFunc((r,i)=>{const o=r.acos(t);return i([t]),o},n,null,Ui)}const Mu=d({acos_:k0});function D0(e){const t=h(e,"x","acosh"),n={x:t};return g.runKernelFunc((r,i)=>{const o=r.acosh(t);return i([t]),o},n,null,Wi)}const Fu=d({acosh_:D0});function F0(e,t){let n=h(e,"a","add"),r=h(t,"b","add");[n,r]=V(n,r);const i=(a,s)=>{const c=a.add(n,r);return s([n,r]),c},o={a:n,b:r};return g.runKernelFunc(i,o,null,Jn)}const O=d({add_:F0});function M0(e){f(Array.isArray(e),()=>"The argument passed to tf.addN() must be a list of tensors"),f(e.length>=1,()=>`Must pass at least one tensor to tf.addN(), but got ${e.length}`);const t=e.map((o,a)=>h(o,`tensors${a}`,"addN")),n=t[0];t.forEach(o=>{if(o.dtype!==n.dtype)throw new Error("All tensors passed to tf.addN() must have the same dtype")}),t.forEach(o=>{if(!Le(o.shape,n.shape))throw new Error("All tensors passed to tf.addN() must have the same shape")});const r=(o,a)=>{const s=o.addN(t);return a(t),s},i=t;return g.runKernelFunc(r,i,null,qi)}const Du=d({addN_:M0});function Am(e,t){for(let n=0;n<e.length;++n)if(e[e.length-n-1]!==t-1-n)return!1;return!0}function Dy(e,t,n){const r=e.length+t.length,i=[];let o=0,a=0;for(let s=0;s<r;s++)n.indexOf(s)===-1?i.push(e[o++]):i.push(t[a++]);return i}function Nm(e,t){const n=[],r=e.length;for(let o=0;o<r;o++)t.indexOf(o)===-1&&n.push(e[o]);const i=t.map(o=>e[o]);return[n,i]}function fe(e,t){const n=t.map(r=>1);return Dy(e,n,t)}function U0(e,t,n){f(Am(t,n),()=>`${e} supports only inner-most axes for now. Got axes ${t} and rank-${n} input.`)}function he(e,t){if(Am(e,t))return null;const n=[];for(let r=0;r<t;++r)e.indexOf(r)===-1&&n.push(r);return e.forEach(r=>n.push(r)),n}function pr(e){return e.map((t,n)=>[n,t]).sort((t,n)=>t[1]-n[1]).map(t=>t[0])}function Ee(e,t){const n=[];for(let r=t-e;r<t;++r)n.push(r);return n}function W0(e,t){const n=h(e,"x","reshape",null);t=um(t,n.size),f(n.size===Z(t),()=>"new shape and old shape must have the same number of elements.");const r={x:n},i={shape:t},o=(a,s)=>(s([n]),a.reshape(n,t));return g.runKernelFunc(o,r,null,$o,i)}const x=d({reshape_:W0});function q0(e,t){const n=h(e,"x","transpose");if(t==null&&(t=n.shape.map((o,a)=>a).reverse()),f(n.rank===t.length,()=>`Error in transpose: rank of input ${n.rank} must match length of perm ${t}.`),t.forEach(o=>{f(o>=0&&o<n.rank,()=>`All entries in 'perm' must be between 0 and ${n.rank-1} but got ${t}`)}),n.rank<=1)return n.clone();const r={x:n},i={perm:t};return g.runKernelFunc(o=>o.transpose(n,t),r,null,fs,i)}const X=d({transpose_:q0});function B0(e,t=null,n=!1){let r=h(e,"x","all","bool");const i=s=>{const c=z(t,r.shape);let p=c;const l=he(p,r.rank);l!=null&&(r=X(r,l),p=Ee(p.length,r.rank));const u=s.all(r,p);if(n){const m=fe(u.shape,c);return x(u,m)}return u},o={x:r},a={axis:t,keepDims:n};return g.runKernelFunc(i,o,null,Wu,a)}const ku=d({all_:B0});function P0(e,t=null,n=!1){let r=h(e,"x","any","bool");const i=s=>{const c=z(t,r.shape);let p=c;const l=he(p,r.rank);l!=null&&(r=X(r,l),p=Ee(p.length,r.rank));const u=s.any(r,p);if(n){const m=fe(u.shape,c);return x(u,m)}return u},o={x:r},a={axis:t,keepDims:n};return g.runKernelFunc(i,o,null,qu,a)}const Ou=d({any_:P0});function j0(e,t=0){let n=h(e,"x","argMax");const r=(a,s)=>{s([n]);let c=z(t,n.shape);const p=he(c,n.rank);return p!=null&&(n=X(n,p),c=Ee(c.length,n.rank)),a.argMax(n,c[0])},i={x:n},o={axis:t};return g.runKernelFunc(r,i,null,Bi,o)}const Eu=d({argMax_:j0});function G0(e,t=0){let n=h(e,"x","argMin");const r=(a,s)=>{s([n]),t==null&&(t=0);let c=z(t,n.shape);const p=he(c,n.rank);return p!=null&&(n=X(n,p),c=Ee(c.length,n.rank)),a.argMin(n,c[0])},i={x:n},o={axis:t};return g.runKernelFunc(r,i,null,Pi,o)}const Cu=d({argMin_:G0});function $0(e){const t=h(e,"x","asin"),n={x:t};return g.runKernelFunc((r,i)=>{const o=r.asin(t);return i([t]),o},n,null,ji)}const Ru=d({asin_:$0});function H0(e){const t=h(e,"x","asinh"),n={x:t};return g.runKernelFunc((r,i)=>{const o=r.asinh(t);return i([t]),o},n,null,Gi)}const _u=d({asinh_:H0});function Y0(e){const t=h(e,"x","atan"),n={x:t};return g.runKernelFunc((r,i)=>{const o=r.atan(t);return i([t]),o},n,null,$i)}const Nu=d({atan_:Y0});function z0(e,t){let n=h(e,"a","atan2"),r=h(t,"b","atan2");[n,r]=V(n,r);const i=(a,s)=>{const c=a.atan2(n,r);return s([n,r]),c},o={a:n,b:r};return g.runKernelFunc(i,o,null,Yi)}const Au=d({atan2_:z0});function V0(e){const t=h(e,"x","atanh"),n={x:t};return g.runKernelFunc((r,i)=>{const o=r.atanh(t);return i([t]),o},n,null,Hi)}const Iu=d({atanh_:V0});function K0(e,t){const n=h(e,"x","cast");if(!mm(t))throw new Error(`Failed to cast to unknown dtype ${t}`);if(t==="string"&&n.dtype!=="string"||t!=="string"&&n.dtype==="string")throw new Error("Only strings can be casted to strings");const r={x:n},i={dtype:t};return g.runKernelFunc(o=>o.cast(n,t),r,null,Xn,i)}const R=d({cast_:K0});function J0(e,t,n,r,i="NHWC",o){const a=e[3],s=[...t,a],c=Sn(i);return We(e,s,n,o,r,null,null,c)}function Ct(e,t,n,r,i,o,a="channelsLast"){const[s,c]=kp(t);let p;if(a==="channelsLast")p=[s,c,e[3],e[3]];else if(a==="channelsFirst")p=[s,c,e[1],e[1]];else throw new Error(`Unknown dataFormat ${a}`);return We(e,p,n,r,i,o,!1,a)}function on(e,t,n,r,i,o,a="NDHWC"){const[s,c,p]=_m(t);let l,u;if(a==="NDHWC")u="channelsLast",l=[s,c,p,e[4],e[4]];else if(a==="NCDHW")u="channelsFirst",l=[s,c,p,e[1],e[1]];else throw new Error(`Unknown dataFormat ${a}`);return Tn(e,l,n,r,i,!1,u,o)}function We(e,t,n,r,i,o,a=!1,s="channelsLast"){let[c,p,l,u]=[-1,-1,-1,-1];if(s==="channelsLast")[c,p,l,u]=e;else if(s==="channelsFirst")[c,u,p,l]=e;else throw new Error(`Unknown dataFormat ${s}`);const[m,b,,y]=t,[w,L]=kp(n),[T,A]=kp(r),_=si(m,T),N=si(b,A),{padInfo:C,outHeight:F,outWidth:E}=X0(i,p,l,w,L,_,N,o,s),B=a?y*u:y;let q;return s==="channelsFirst"?q=[c,B,F,E]:s==="channelsLast"&&(q=[c,F,E,B]),{batchSize:c,dataFormat:s,inHeight:p,inWidth:l,inChannels:u,outHeight:F,outWidth:E,outChannels:B,padInfo:C,strideHeight:w,strideWidth:L,filterHeight:m,filterWidth:b,effectiveFilterHeight:_,effectiveFilterWidth:N,dilationHeight:T,dilationWidth:A,inShape:e,outShape:q,filterShape:t}}function Tn(e,t,n,r,i,o=!1,a="channelsLast",s){let[c,p,l,u,m]=[-1,-1,-1,-1,-1];if(a==="channelsLast")[c,p,l,u,m]=e;else if(a==="channelsFirst")[c,m,p,l,u]=e;else throw new Error(`Unknown dataFormat ${a}`);const[b,y,w,,L]=t,[T,A,_]=_m(n),[N,C,F]=_m(r),E=si(b,N),B=si(y,C),q=si(w,F),{padInfo:P,outDepth:te,outHeight:ne,outWidth:pe}=Z0(i,p,l,u,T,A,_,E,B,q,s),Se=o?L*m:L;let qe;return a==="channelsFirst"?qe=[c,Se,te,ne,pe]:a==="channelsLast"&&(qe=[c,te,ne,pe,Se]),{batchSize:c,dataFormat:a,inDepth:p,inHeight:l,inWidth:u,inChannels:m,outDepth:te,outHeight:ne,outWidth:pe,outChannels:Se,padInfo:P,strideDepth:T,strideHeight:A,strideWidth:_,filterDepth:b,filterHeight:y,filterWidth:w,effectiveFilterDepth:E,effectiveFilterHeight:B,effectiveFilterWidth:q,dilationDepth:N,dilationHeight:C,dilationWidth:F,inShape:e,outShape:qe,filterShape:t}}function Q0(e,t,n,r,i){r==null&&(r=Rm(e,t,n));const o=e[0],a=e[1],s=lr((o-t+2*r)/n+1,i);f(J(s),()=>`The output # of rows (${s}) must be an integer. Change the stride and/or zero pad parameters`);const c=lr((a-t+2*r)/n+1,i);return f(J(c),()=>`The output # of columns (${c}) must be an integer. Change the stride and/or zero pad parameters`),[s,c]}function eR(e,t,n,r,i,o){i==null&&(i=Rm(e,t,r));const a=e[0],s=e[1],c=e[2],p=lr((a-t+2*i)/r+1,o);f(J(p),()=>`The output # of depths (${p}) must be an integer. Change the stride and/or zero pad parameters`);const l=lr((s-t+2*i)/r+1,o);f(J(l),()=>`The output # of rows (${l}) must be an integer. Change the stride and/or zero pad parameters`);const u=lr((c-t+2*i)/r+1,o);return f(J(u),()=>`The output # of columns (${u}) must be an integer. Change the stride and/or zero pad parameters`),[p,l,u,n]}function Rm(e,t,n,r=1){const i=si(t,r);return Math.floor((e[0]*(n-1)-n+i)/2)}function kp(e){return typeof e=="number"?[e,e,e]:e.length===2?[e[0],e[1],1]:e}function _m(e){return typeof e=="number"?[e,e,e]:e}function si(e,t){return t<=1?e:e+(e-1)*(t-1)}function X0(e,t,n,r,i,o,a,s,c){let p,l,u;if(typeof e=="number"){const m=e===0?"VALID":"NUMBER";p={top:e,bottom:e,left:e,right:e,type:m};const b=Q0([t,n],o,r,e,s);l=b[0],u=b[1]}else if(e==="same"){l=Math.ceil(t/r),u=Math.ceil(n/i);const m=Math.max(0,(l-1)*r+o-t),b=Math.max(0,(u-1)*i+a-n),y=Math.floor(m/2),w=m-y,L=Math.floor(b/2),T=b-L;p={top:y,bottom:w,left:L,right:T,type:"SAME"}}else if(e==="valid")p={top:0,bottom:0,left:0,right:0,type:"VALID"},l=Math.ceil((t-o+1)/r),u=Math.ceil((n-a+1)/i);else if(typeof e=="object"){const m=c==="channelsLast"?e[1][0]:e[2][0],b=c==="channelsLast"?e[1][1]:e[2][1],y=c==="channelsLast"?e[2][0]:e[3][0],w=c==="channelsLast"?e[2][1]:e[3][1],L=m===0&&b===0&&y===0&&w===0?"VALID":"EXPLICIT";p={top:m,bottom:b,left:y,right:w,type:L},l=lr((t-o+m+b)/r+1,s),u=lr((n-a+y+w)/i+1,s)}else throw Error(`Unknown padding parameter: ${e}`);return{padInfo:p,outHeight:l,outWidth:u}}function Z0(e,t,n,r,i,o,a,s,c,p,l){let u,m,b,y;if(typeof e=="number"){const w=e===0?"VALID":"NUMBER";u={top:e,bottom:e,left:e,right:e,front:e,back:e,type:w};const L=eR([t,n,r,1],s,1,i,e,l);m=L[0],b=L[1],y=L[2]}else if(e==="same"){m=Math.ceil(t/i),b=Math.ceil(n/o),y=Math.ceil(r/a);const w=(m-1)*i+s-t,L=(b-1)*o+c-n,T=(y-1)*a+p-r,A=Math.floor(w/2),_=w-A,N=Math.floor(L/2),C=L-N,F=Math.floor(T/2),E=T-F;u={top:N,bottom:C,left:F,right:E,front:A,back:_,type:"SAME"}}else if(e==="valid")u={top:0,bottom:0,left:0,right:0,front:0,back:0,type:"VALID"},m=Math.ceil((t-s+1)/i),b=Math.ceil((n-c+1)/o),y=Math.ceil((r-p+1)/a);else throw Error(`Unknown padding parameter: ${e}`);return{padInfo:u,outDepth:m,outHeight:b,outWidth:y}}function lr(e,t){if(!t)return 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 ut(e){const[t,n,r]=kp(e);return t===1&&n===1&&r===1}function ae(e,t){return ut(e)||ut(t)}function Sn(e){if(e==="NHWC")return"channelsLast";if(e==="NCHW")return"channelsFirst";throw new Error(`Unknown dataFormat ${e}`)}function tR(e,t,n,r,i){const o=h(e,"x","avgPool","float32"),a=1;f(ae(n,a),()=>`Error in avgPool: Either strides or dilations must be 1. Got strides ${n} and dilations '${a}'`);let s=o,c=!1;o.rank===3&&(c=!0,s=x(o,[1,o.shape[0],o.shape[1],o.shape[2]])),f(s.rank===4,()=>`Error in avgPool: x must be rank 4 but got rank ${s.rank}.`),i!=null&&f(J(r),()=>`Error in avgPool: pad must be an integer when using, dimRoundingMode ${i} but got pad ${r}.`);const p=(b,y)=>{const w=Ct(s.shape,t,n,1,r,i);return y([s]),w.filterWidth===1&&w.filterHeight===1&&Le(w.inShape,w.outShape)?s.clone():b.avgPool(s,w)},l={x:s},u={filterSize:t,strides:n,pad:r,dimRoundingMode:i};let m=g.runKernelFunc(p,l,null,zi,u);return m=R(m,o.dtype),c?x(m,[m.shape[1],m.shape[2],m.shape[3]]):m}const Di=d({avgPool_:tR});function uS(){se().set("PROD",!0)}function hS(){se().set("DEBUG",!0)}function dS(){se().set("DEPRECATION_WARNINGS_ENABLED",!1),console.warn("TensorFlow.js deprecation warnings have been disabled.")}function ge(e){se().getBool("DEPRECATION_WARNINGS_ENABLED")&&console.warn(e+" You can disable deprecation warnings with tf.disableDeprecationWarnings().")}Ry(ge);function mS(){g.disposeVariables()}function fS(){return g}function gS(){return g.memory()}function bS(e){return g.profile(e)}function ce(e,t){return g.tidy(e,t)}function be(e){const t=Gs(e);t.forEach(n=>n.dispose())}function Dh(e){return g.keep(e)}function wS(e){return g.time(e)}function yS(e){return g.setBackend(e)}function xS(){return g.ready()}function LS(){return g.backendName}function vS(e){g.removeBackend(e)}function SS(e){return g.findBackend(e)}function TS(e){return g.findBackendFactory(e)}function IS(e,t,n=1){return g.registerBackend(e,t,n)}function AS(){return g.backend}function NS(e,t){se().setPlatform(e,t)}function nR(e,t,n,r,i,o="NDHWC",a){a==null?a=[1,1,1]:ge("dilations is deprecated, this field will be gone in v3.0.0.");const s=h(e,"x","avgPool3d","float32");let c=s,p=!1;s.rank===4&&(p=!0,c=x(s,[1,s.shape[0],s.shape[1],s.shape[2],s.shape[3]])),f(c.rank===5,()=>`Error in avgPool3d: x must be rank 5 but got rank ${c.rank}.`),f(o==="NDHWC",()=>`Error in avgPool3d: Only NDHWC is currently supported, but got dataFormat of ${o}`),f(ae(n,a),()=>`Error in avgPool3d: Either strides or dilations must be 1. Got strides ${n} and dilations '${a}'`),i!=null&&f(J(r),()=>`Error in avgPool3d: pad must be an integer when using, dimRoundingMode ${i} but got pad ${r}.`);const l=(y,w)=>{a==null&&(a=[1,1,1]);const L=on(c.shape,t,n,a,r,i,o);return w([c]),y.avgPool3d(c,L)},u={x:c},m={filterSize:t,strides:n,pad:r,dimRoundingMode:i,dataFormat:o,dilations:a};let b=g.runKernelFunc(l,u,null,Vi,m);return b=R(b,c.dtype),p?x(b,[b.shape[1],b.shape[2],b.shape[3],b.shape[4]]):b}const Tu=d({avgPool3d_:nR});function Cm(e,t){const n=e[0].length;e.forEach((i,o)=>{f(i.length===n,()=>`Error in concat${n}D: rank of tensors[${o}] must be the same as the rank of the rest (${n})`)}),f(t>=0&&t<n,()=>`Error in concat${n}D: axis must be between 0 and ${n-1}.`);const r=e[0];e.forEach((i,o)=>{for(let a=0;a<n;a++)f(a===t||i[a]===r[a],()=>`Error in concat${n}D: Shape of tensors[${o}] (${i}) does not match the shape of the rest (${r}) along the non-concatenated axis ${o}.`)})}function Em(e,t){const n=e[0].slice();for(let r=1;r<e.length;r++)n[t]+=e[r][t];return n}function ze(e,t,n,r){if(r==null&&(r=ri(e)),r==="complex64")throw new Error("Cannot construct a complex64 tensor directly. Please use tf.complex(real, imag).");if(!Ie(e)&&!Array.isArray(e)&&typeof e!="number"&&typeof e!="boolean"&&typeof e!="string")throw new Error("values passed to tensor(values) must be a number/boolean/string or an array of numbers/booleans/strings, or a TypedArray");if(t!=null){qs(t);const i=Z(t),o=Z(n);f(i===o,()=>`Based on the provided shape, [${t}], the tensor should have ${i} values but has ${o}`);for(let a=0;a<n.length;++a){const s=n[a],c=a===n.length-1?s!==Z(t.slice(a)):!0;f(n[a]===t[a]||!c,()=>`Error creating a new Tensor. Inferred shape (${n}) does not match the provided shape (${t}). `)}}return!Ie(e)&&!Array.isArray(e)&&(e=[e]),t=t||n,e=r!=="string"?Us(e,r):Wt(e,[],!0),g.makeTensor(e,t,r)}function Qe(e,t,n){const r=Fe(e,n);return ze(e,t,r,n)}function rR(e,t=0){f(e.length>=1,()=>"Pass at least one tensor to concat");let n=rn(e,"tensors","concat");n[0].dtype==="complex64"&&n.forEach(p=>{if(p.dtype!=="complex64")throw new Error(`Cannot concatenate complex64 tensors with a tensor
with dtype ${p.dtype}. `)});const r=z(t,n[0].shape)[0],i=Em(n.map(p=>p.shape),r);if(Z(i)===0)return Qe([],i);if(n=n.filter(p=>p.size>0),n.length===1)return n[0];const o=n.map(p=>p.shape);Cm(o,r);const a=(p,l)=>{const u=p.concat(n,r);return l(n),u},s=n,c={axis:t};return g.runKernelFunc(a,s,null,eo,c)}const re=d({concat_:rR});function iR(e,t,n=!1,r=!1){let i=h(e,"a","matMul"),o=h(t,"b","matMul");[i,o]=V(i,o),f(i.rank>=2&&o.rank>=2&&i.rank===o.rank,()=>`Error in matMul: inputs must have the same rank of at least 2, got ranks ${i.rank} and ${o.rank}.`);const a=n?i.shape[i.rank-2]:i.shape[i.rank-1],s=r?o.shape[o.rank-1]:o.shape[o.rank-2],c=n?i.shape[i.rank-1]:i.shape[i.rank-2],p=r?o.shape[o.rank-2]:o.shape[o.rank-1],l=i.shape.slice(0,-2),u=o.shape.slice(0,-2),m=Z(l),b=Z(u);f(Le(l,u),()=>`Error in matMul: outer dimensions (${l}) and (${u}) of Tensors with shapes ${i.shape} and ${o.shape} must match.`),f(a===s,()=>`Error in matMul: inner shapes (${a}) and (${s}) of Tensors with shapes ${i.shape} and ${o.shape} and transposeA=${n} and transposeB=${r} must match.`);const y=i.shape.slice(0,-2).concat([c,p]),w=n?x(i,[m,a,c]):x(i,[m,c,a]),L=r?x(o,[b,p,s]):x(o,[b,s,p]),T=(C,F)=>(F([w,L]),C.batchMatMul(w,L,n,r)),A={a:w,b:L},_={transposeA:n,transposeB:r},N=g.runKernelFunc(T,A,null,Ki,_);return x(N,y)}const K=d({matMul_:iR});function oR(e,t){let n=h(e,"a","mul"),r=h(t,"b","mul");[n,r]=V(n,r);const i=(a,s)=>{const c=a.multiply(n,r);return s([n,r]),c},o={a:n,b:r};return g.runKernelFunc(i,o,null,Fo)}const v=d({mul_:oR});function sR(e){const t=h(e,"x","sigmoid"),n={x:t};return g.runKernelFunc((r,i)=>{const o=r.sigmoid(t);return i([o]),o},n,null,rs)}const St=d({sigmoid_:sR});const kf={};Oe(kf,{assertParamsValid:()=>Om,computeFlatOffset:()=>cR,computeOutShape:()=>km,isSliceContinous:()=>aR,maskToAxes:()=>Hs,parseSliceParams:()=>Ys,startForAxis:()=>Wm,startIndicesWithElidedDims:()=>Fm,stopForAxis:()=>qm,stopIndicesWithElidedDims:()=>Mm,stridesForAxis:()=>Um,stridesWithElidedDims:()=>Dm});function Om(e,t,n){f(e.rank===t.length,()=>`Error in slice${e.rank}D: Length of begin ${t} must match the rank of the array (${e.rank}).`),f(e.rank===n.length,()=>`Error in slice${e.rank}D: Length of size ${n} must match the rank of the array (${e.rank}).`);for(let r=0;r<e.rank;++r)f(t[r]+n[r]<=e.shape[r],()=>`Error in slice${e.rank}D: begin[${r}] + size[${r}] (${t[r]+n[r]}) would overflow input.shape[${r}] (${e.shape[r]})`)}function Hs(e){const t=[];let n=0;for(;e>0;)e&1&&t.push(n),e/=2,n++;return t}function km(e,t,n){const r=[];for(let i=0;i<e.length;i++)r[i]=Math.ceil((t[i]-e[i])/n[i]);return r}function Dm(e,t,n,r){const i=[...e];for(let o=i.length;o<r.length;o++)i.push(1);for(let o=0;o<n;o++)o===0?i[t]=1:(i.splice(t,0,1),i.pop());return i}function Fy(e,t,n){return n<=e?n:n-(t-1)}function My(e,t){const n=[];for(let r=0;r<e;r++)n.push(t+r);return n}function Fm(e,t,n,r,i){const o=[...i],a=My(n,t);for(let s=0;s<o.length;s++)if(a.indexOf(s)>-1)o[s]=0;else{const c=Fy(t,n,s);let p=r[c];e&1<<c&&(p=0),o[s]=p}return o}function Mm(e,t,n,r,i){const o=[...i],a=My(n,t);for(let s=0;s<o.length;s++)if(a.indexOf(s)>-1)o[s]=Number.MAX_SAFE_INTEGER;else{const c=Fy(t,n,s);let p=r[c];e&1<<c&&(p=Number.MAX_SAFE_INTEGER),o[s]=p}for(let s=0;s<o.length;s++){const c=i[s];o[s]<0&&(o[s]+=c),o[s]=ti(0,o[s],i[s])}return o}function Um(e,t,n){let r=e[t];return(n&1<<t||r==null)&&(r=1),r}function Wm(e,t,n,r,i,o){let a=t[i];const s=n[i]||1;(e&1<<i||o&1<<i||a==null)&&(s>0?a=Number.MIN_SAFE_INTEGER:a=Number.MAX_SAFE_INTEGER);const c=r[i];return a<0&&(a+=c),a=ti(0,a,c-1),a}function qm(e,t,n,r,i,o){let a=t[i];const s=n[i]||1;(e&1<<i||o&1<<i||a==null)&&(s>0?a=Number.MAX_SAFE_INTEGER:a=Number.MIN_SAFE_INTEGER);const c=r[i];return a<0&&(a+=c),s>0?a=ti(0,a,c):a=ti(-1,a,c-1),a}function aR(e,t,n){let r=n.length;for(let i=0;i<n.length;i++)if(n[i]>1){r=i;break}for(let i=r+1;i<n.length;i++)if(t[i]>0||n[i]!==e[i])return!1;return!0}function cR(e,t){let n=e.length>0?e[e.length-1]:1;for(let r=0;r<e.length-1;r++)n+=e[r]*t[r];return n}function Ys(e,t,n){let r;typeof t=="number"?r=[t,...new Array(e.rank-1).fill(0)]:t.length<e.rank?r=t.concat(new Array(e.rank-t.length).fill(0)):r=t.slice(),r.forEach(o=>{f(o!==-1,()=>"slice() does not support negative begin indexing.")});let i;return n==null?i=new Array(e.rank).fill(-1):typeof n=="number"?i=[n,...new Array(e.rank-1).fill(-1)]:n.length<e.rank?i=n.concat(new Array(e.rank-n.length).fill(-1)):i=n,i=i.map((o,a)=>o>=0?o:(f(o===-1,()=>`Negative size values should be exactly -1 but got ${o} for the slice() size at index ${a}.`),e.shape[a]-r[a])),[r,i]}function pR(e,t,n){const r=h(e,"x","slice");if(r.rank===0)throw new Error("Slicing scalar is not possible");const[i,o]=Ys(r,t,n);Om(r,i,o);const a=(p,l)=>(l([r]),p.slice(r,i,o)),s={x:r},c={begin:t,size:n};return g.runKernelFunc(a,s,null,Qo,c)}const Y=d({slice_:pR});function lR(e){const t=h(e,"x","tanh"),n={x:t};return g.runKernelFunc((r,i)=>{const o=r.tanh(t);return i([o]),o},n,null,ds)}const xr=d({tanh_:lR});function uR(e,t,n,r,i,o){const a=h(e,"forgetBias","basicLSTMCell"),s=h(t,"lstmKernel","basicLSTMCell"),c=h(n,"lstmBias","basicLSTMCell"),p=h(r,"data","basicLSTMCell"),l=h(i,"c","basicLSTMCell"),u=h(o,"h","basicLSTMCell"),m=re([p,u],1),b=K(m,s),y=O(b,c),w=y.shape[0],L=y.shape[1]/4,T=[w,L],A=Y(y,[0,0],T),_=Y(y,[0,L],T),N=Y(y,[0,L*2],T),C=Y(y,[0,L*3],T),F=O(v(St(A),xr(_)),v(l,St(O(a,N)))),E=v(xr(F),St(C));return[F,E]}const Su=d({basicLSTMCell_:uR});function hR(e,t,n){const r=h(e,"x","batchToSpaceND"),i=t.reduce((c,p)=>c*p);f(r.rank>=1+t.length,()=>`input rank is ${r.rank} but should be > than blockShape.length ${t.length}`),f(n.length===t.length,()=>`crops.length is ${n.length} but should be equal to blockShape.length ${t.length}`),f(r.shape[0]%i===0,()=>`input tensor batch is ${r.shape[0]} but is not divisible by the product of the elements of blockShape ${t.join(" * ")} === ${i}`);const o=c=>c.batchToSpaceND(r,t,n),a={x:r},s={blockShape:t,crops:n};return g.runKernelFunc(o,a,null,Ji,s)}const Kn=d({batchToSpaceND_:hR});function Uy(e){let t;return e.rank===0||e.rank===1?t=x(e,[1,1,1,e.size]):e.rank===2?t=x(e,[1,1,e.shape[0],e.shape[1]]):e.rank===3?t=x(e,[1,e.shape[0],e.shape[1],e.shape[2]]):t=e,t}function dR(e,t,n,r,i,o){o==null&&(o=.001);const a=h(e,"x","batchNorm"),s=h(t,"mean","batchNorm"),c=h(n,"variance","batchNorm");let p;i!=null&&(p=h(i,"scale","batchNorm"));let l;r!=null&&(l=h(r,"offset","batchNorm")),f(s.rank===c.rank,()=>"Batch normalization gradient requires mean and variance to have equal ranks."),f(l==null||s.rank===l.rank,()=>"Batch normalization gradient requires mean and offset to have equal ranks."),f(p==null||s.rank===p.rank,()=>"Batch normalization gradient requires mean and scale to have equal ranks.");const u=Uy(a),m=(L,T)=>(T([u,s,c,p]),L.batchNorm(u,Dp(s),Dp(c),Dp(l),Dp(p),o)),b={x:u,scale:p,offset:l,mean:s,variance:c},y={varianceEpsilon:o},w=g.runKernelFunc(m,b,null,bo,y);return x(w,a.shape)}function Dp(e){return e==null?null:e.rank===0?x(e,[e.size]):e.rank===1?e:e.rank===2?x(e,[1,1,e.shape[0],e.shape[1]]):e.rank===3?x(e,[1,e.shape[0],e.shape[1],e.shape[2]]):e}const Jt=d({batchNorm_:dR});function mR(e,t,n,r,i,o){const a=h(e,"x","batchNorm"),s=h(t,"mean","batchNorm"),c=h(n,"variance","batchNorm");let p;i!=null&&(p=h(i,"scale","batchNorm"));let l;return r!=null&&(l=h(r,"offset","batchNorm")),f(a.rank===2,()=>`Error in batchNorm2D: x must be rank 2 but got rank ${a.rank}.`),f(s.rank===2||s.rank===1,()=>`Error in batchNorm2D: mean must be rank 2 or rank 1 but got rank ${s.rank}.`),f(c.rank===2||c.rank===1,()=>`Error in batchNorm2D: variance must be rank 2 or rank 1 but got rank ${c.rank}.`),p!=null&&f(p.rank===2||p.rank===1,()=>`Error in batchNorm2D: scale must be rank 2 or rank 1 but got rank ${p.rank}.`),l!=null&&f(l.rank===2||l.rank===1,()=>`Error in batchNorm2D: offset must be rank 2 or rank 1 but got rank ${l.rank}.`),Jt(a,s,c,l,p,o)}const vu=d({batchNorm2d_:mR});function fR(e,t,n,r,i,o){const a=h(e,"x","batchNorm"),s=h(t,"mean","batchNorm"),c=h(n,"variance","batchNorm");let p;i!=null&&(p=h(i,"scale","batchNorm"));let l;return r!=null&&(l=h(r,"offset","batchNorm")),f(a.rank===3,()=>`Error in batchNorm3D: x must be rank 3 but got rank ${a.rank}.`),f(s.rank===3||s.rank===1,()=>`Error in batchNorm3D: mean must be rank 3 or rank 1 but got rank ${s.rank}.`),f(c.rank===3||c.rank===1,()=>`Error in batchNorm3D: variance must be rank 3 or rank 1 but got rank ${c.rank}.`),p!=null&&f(p.rank===3||p.rank===1,()=>`Error in batchNorm3D: scale must be rank 3 or rank 1 but got rank ${p.rank}.`),l!=null&&f(l.rank===3||l.rank===1,()=>`Error in batchNorm3D: offset must be rank 3 or rank 1 but got rank ${l.rank}.`),Jt(a,s,c,l,p,o)}const Lu=d({batchNorm3d_:fR});function gR(e,t,n,r,i,o){const a=h(e,"x","batchNorm"),s=h(t,"mean","batchNorm"),c=h(n,"variance","batchNorm");let p;i!=null&&(p=h(i,"scale","batchNorm"));let l;return r!=null&&(l=h(r,"offset","batchNorm")),f(a.rank===4,()=>`Error in batchNorm4D: x must be rank 4 but got rank ${a.rank}.`),f(s.rank===4||s.rank===1,()=>`Error in batchNorm4D: mean must be rank 4 or rank 1 but got rank ${s.rank}.`),f(c.rank===4||c.rank===1,()=>`Error in batchNorm4D: variance must be rank 4 or rank 1 but got rank ${c.rank}.`),p!=null&&f(p.rank===4||p.rank===1,()=>`Error in batchNorm4D: scale must be rank 4 or rank 1 but got rank ${p.rank}.`),l!=null&&f(l.rank===4||l.rank===1,()=>`Error in batchNorm4D: offset must be rank 4 or rank 1 but got rank ${l.rank}.`),Jt(a,s,c,l,p,o)}const xu=d({batchNorm4d_:gR});function bR(e){const t=h(e,"x","clone",null),n=()=>g.makeTensorFromDataId(t.dataId,t.shape,t.dtype),r={x:t};return g.runKernelFunc(n,r,null,xo)}const At=d({clone_:bR});function wR(e,t){let n=h(e,"broadcastTo","x");const r=n.shape;if(t.some(l=>!(l>0)||l%1!==0))throw new Error(`broadcastTo(): Invalid broadcast shape [${t}].`);if(t.length<n.rank)throw new Error(`broadcastTo(): shape.length=${t.length} < input.rank=${n.rank}.`);if(t.length>n.rank){const l=n.shape.slice();for(;l.length<t.length;)l.unshift(1);n=x(n,l)}const i=n.shape,o=Array.from(t);for(let l=t.length-1;l>=0;l--)if(i[l]===t[l])o[l]=1;else if(n.shape[l]!==1)throw new Error(`broadcastTo(): [${r}] cannot be broadcast to [${t}].`);const a=o.map((l,u)=>l>1?u:-1).filter(l=>l>=0);if(a.length===0)return At(n);const s=l=>l.tile(n,o),c={x:n},p={shape:t,inputShape:i};return g.runKernelFunc(s,c,null,Xi,p)}const Ir=d({broadcastTo_:wR});function ke(e,t="float32",n){return t=t||"float32",qs(e),new yr(e,t,n)}function yR(e){const t=h(e,"x","ceil"),n={x:t};return g.runKernelFunc(r=>r.ceil(t),n,null,Zi)}const yu=d({ceil_:yR});function xR(e,t,n){const r=h(e,"x","clipByValue");f(t<=n,()=>`Error in clip: min (${t}) must be less than or equal to max (${n}).`);const i={x:r},o={clipValueMin:t,clipValueMax:n};return g.runKernelFunc((a,s)=>{const c=a.clip(r,t,n);return s([r]),c},i,null,Qi,o)}const wu=d({clipByValue_:xR});function LR(e,t){const n=h(e,"real","complex"),r=h(t,"imag","complex");$(n.shape,r.shape,`real and imag shapes, ${n.shape} and ${r.shape}, must match in call to tf.complex().`);const i=a=>a.complex(n,r),o={real:n,imag:r};return g.runKernelFunc(i,o,null,ju)}const Ge=d({complex_:LR});function vR(e){return re(e,0)}const bu=d({concat1d_:vR});function SR(e,t){return re(e,t)}const gu=d({concat2d_:SR});function TR(e,t){return re(e,t)}const fu=d({concat3d_:TR});function IR(e,t){return re(e,t)}const mu=d({concat4d_:IR});function AR(e,t,n,r,i="NHWC",o=[1,1],a){const s=h(e,"x","conv2d"),c=h(t,"filter","conv2d");let p=s,l=!1;s.rank===3&&(l=!0,p=x(s,[1,s.shape[0],s.shape[1],s.shape[2]])),f(p.rank===4,()=>`Error in conv2d: input must be rank 4, but got rank ${p.rank}.`),f(c.rank===4,()=>`Error in conv2d: filter must be rank 4, but got rank ${c.rank}.`),a!=null&&f(J(r),()=>`Error in conv2d: pad must be an integer when using, dimRoundingMode ${a} but got pad ${r}.`);const u=i==="NHWC"?p.shape[3]:p.shape[1];f(u===c.shape[2],()=>`Error in conv2d: depth of input (${u}) must match input depth for filter ${c.shape[2]}.`),f(ae(n,o),()=>`Error in conv2D: Either strides or dilations must be 1. Got strides ${n} and dilations '${o}'`);const m=(L,T)=>{const A=Sn(i),_=We(p.shape,c.shape,n,o,r,a,!1,A),N=L.conv2d(p,c,_);return T([p,c]),N},b={x:p,filter:c},y={strides:n,pad:r,dataFormat:i,dilations:o,dimRoundingMode:a},w=g.runKernelFunc(m,b,null,to,y);return l?x(w,[w.shape[1],w.shape[2],w.shape[3]]):w}const It=d({conv2d_:AR});function NR(e,t,n,r,i="NWC",o=1,a){const s=h(e,"x","conv1d"),c=h(t,"filter","conv1d");let p=s,l=!1;s.rank===2&&(l=!0,p=x(s,[1,s.shape[0],s.shape[1]])),f(p.rank===3,()=>`Error in conv1d: input must be rank 3, but got rank ${p.rank}.`),f(c.rank===3,()=>`Error in conv1d: filter must be rank 3, but got rank ${c.rank}.`),a!=null&&f(J(r),()=>`Error in conv1d: pad must be an integer when using, dimRoundingMode ${a} but got pad ${r}.`),f(p.shape[2]===c.shape[1],()=>`Error in conv1d: depth of input (${p.shape[2]}) must match input depth for filter ${c.shape[1]}.`),f(ae(n,o),()=>`Error in conv1D: Either stride or dilation must be 1. Got stride ${n} and dilation '${o}'`),f(i==="NWC",()=>`Error in conv1d: got dataFormat of ${i} but only NWC is currently supported.`);const u=x(c,[1,c.shape[0],c.shape[1],c.shape[2]]),m=x(p,[p.shape[0],1,p.shape[1],p.shape[2]]),b=[1,n],y=[1,o],w="NHWC",L=It(m,u,b,r,w,y,a);return l?x(L,[L.shape[2],L.shape[3]]):x(L,[L.shape[0],L.shape[2],L.shape[3]])}const du=d({conv1d_:NR});function _R(e,t,n,r,i,o="NHWC",a){f(e.length===t.rank,()=>`Length of inShape (${e.length}) and rank of dy (${t.rank}) must match`);let s=e,c=t,p=!1;t.rank===3&&(p=!0,c=x(t,[1,t.shape[0],t.shape[1],t.shape[2]]),s=[1,e[0],e[1],e[2]]),f(s.length===4,()=>`Error in conv2dDerInput: inShape must be length 4, but got length ${s.length}.`),f(c.rank===4,()=>`Error in conv2dDerInput: dy must be rank 4, but got rank ${c.rank}`),f(n.rank===4,()=>`Error in conv2dDerInput: filter must be rank 4, but got rank ${n.rank}`);const l=o==="NHWC"?s[3]:s[1],u=o==="NHWC"?c.shape[3]:c.shape[1];f(l===n.shape[2],()=>`Error in conv2dDerInput: depth of input (${l}) must match input depth for filter ${n.shape[2]}.`),f(u===n.shape[3],()=>`Error in conv2dDerInput: depth of output (${u}) must match output depth for filter ${n.shape[3]}.`),a!=null&&f(J(i),()=>`Error in conv2dDerInput: pad must be an integer when using, dimRoundingMode ${a} but got pad ${i}.`);const m=(L,T)=>{const A=1,_=Sn(o),N=We(s,n.shape,r,A,i,a,!1,_),C=L.conv2dDerInput(c,n,N);return T([c,n]),C},b={dy:c,filter:n},y={strides:r,pad:i,dataFormat:o,dimRoundingMode:a,inputShape:s},w=g.runKernelFunc(m,b,null,no,y);return p?x(w,[w.shape[1],w.shape[2],w.shape[3]]):w}const ai=d({conv2DBackpropInput_:_R});function RR(e,t,n,r,i,o){const a=h(e,"x","conv2dTranspose"),s=h(t,"filter","conv2dTranspose");return ai(n,a,s,r,i,"NHWC",o)}const hu=d({conv2dTranspose_:RR});function CR(e,t,n,r,i="NDHWC",o=[1,1,1]){const a=h(e,"x","conv3d"),s=h(t,"filter","conv3d");let c=a,p=!1;a.rank===4&&(p=!0,c=x(a,[1,a.shape[0],a.shape[1],a.shape[2],a.shape[3]])),f(c.rank===5,()=>`Error in conv3d: input must be rank 5, but got rank ${c.rank}.`),f(s.rank===5,()=>`Error in conv3d: filter must be rank 5, but got rank ${s.rank}.`),f(c.shape[4]===s.shape[3],()=>`Error in conv3d: depth of input (${c.shape[4]}) must match input depth for filter ${s.shape[3]}.`),f(ae(n,o),()=>`Error in conv3D: Either strides or dilations must be 1. Got strides ${n} and dilations '${o}'`),f(i==="NDHWC",()=>`Error in conv3d: got dataFormat of ${i} but only NDHWC is currently supported.`);const l=(y,w)=>{const L=Tn(c.shape,s.shape,n,o,r),T=y.conv3d(c,s,L);return w([c,s]),T},u={x:c,filter:s},m={strides:n,pad:r,dataFormat:i,dilations:o},b=g.runKernelFunc(l,u,null,ro,m);return p?x(b,[b.shape[1],b.shape[2],b.shape[3],b.shape[4]]):b}const uu=d({conv3d_:CR});function ER(e,t,n,r,i){f(e.length===t.rank,()=>`Length of inShape (${e.length}) and rank of dy (${t.rank}) must match`);let o=e,a=t,s=!1;t.rank===4&&(s=!0,a=x(t,[1,t.shape[0],t.shape[1],t.shape[2],t.shape[3]]),o=[1,e[0],e[1],e[2],e[3]]);const c=o[4],p=a.shape[4];f(o.length===5,()=>`Error in conv3dDerInput: inShape must be length 5, but got length ${o.length}.`),f(a.rank===5,()=>`Error in conv3dDerInput: dy must be rank 5, but got rank ${a.rank}`),f(n.rank===5,()=>`Error in conv3dDerInput: filter must be rank 5, but got rank ${n.rank}`),f(c===n.shape[3],()=>`Error in conv3dDerInput: depth of input (${c}) must match input depth for filter ${n.shape[3]}.`),f(p===n.shape[4],()=>`Error in conv3dDerInput: depth of output (${p}) must match output depth for filter ${n.shape[4]}.`);const l=y=>{const w=1,L=Tn(o,n.shape,r,w,i);return y.conv3dDerInput(a,n,L)},u={dy:a},m={pad:i},b=g.runKernelFunc(l,u,null,Hu,m);return s?x(b,[b.shape[1],b.shape[2],b.shape[3],b.shape[4]]):b}const Fp=d({conv3DBackpropInput_:ER});function OR(e,t,n,r,i){const o=h(e,"x","conv3dTranspose"),a=h(t,"filter","conv3dTranspose");return Fp(n,o,a,r,i)}const lu=d({conv3dTranspose_:OR});function kR(e){const t=h(e,"x","cos"),n={x:t};return g.runKernelFunc((r,i)=>{const o=r.cos(t);return i([t]),o},n,null,io)}const Vn=d({cos_:kR});function DR(e){const t=h(e,"x","cosh"),n={x:t};return g.runKernelFunc((r,i)=>{const o=r.cosh(t);return i([t]),o},n,null,oo)}const ki=d({cosh_:DR});function FR(e,t=0,n=!1,r=!1){const i=h(e,"x","cumsum"),o=(c,p)=>{const l=he([t],i.rank);let u=i;l!=null&&(u=X(i,l));const m=Ee(1,i.rank)[0];let b=c.cumsum(u,m,n,r);if(p([i]),l!=null){const y=pr(l);b=X(b,y)}return b},a={x:i},s={axis:t,exclusive:n,reverse:r};return g.runKernelFunc(o,a,null,so,s)}const Oi=d({cumsum_:FR});function MR(e,t,n="NHWC"){const r=h(e,"x","depthToSpace"),i=n==="NHWC"?r.shape[1]:r.shape[2],o=n==="NHWC"?r.shape[2]:r.shape[3],a=n==="NHWC"?r.shape[3]:r.shape[1];f(i*t>=0,()=>`Negative dimension size caused by overflow when multiplying
${i} and ${t} for depthToSpace with input shape
${r.shape}`),f(o*t>=0,()=>`Negative dimension size caused by overflow when multiplying
${o} and ${t} for depthToSpace with input shape
${r.shape}`),f(a%(t*t)===0,()=>`Dimension size must be evenly divisible by ${t*t} but is ${a} for depthToSpace with input shape ${r.shape}`);const s=l=>l.depthToSpace(r,t,n),c={x:r},p={blockSize:t,dataFormat:n};return g.runKernelFunc(s,c,null,zu,p)}const pu=d({depthToSpace_:MR});function UR(e,t,n,r,i="NHWC",o=[1,1],a){const s=h(e,"x","depthwiseConv2d"),c=h(t,"filter","depthwiseConv2d");let p=s,l=!1;s.rank===3&&(l=!0,p=x(s,[1,s.shape[0],s.shape[1],s.shape[2]])),f(p.rank===4,()=>`Error in depthwiseConv2d: input must be rank 4, but got rank ${p.rank}.`),f(c.rank===4,()=>`Error in depthwiseConv2d: filter must be rank 4, but got rank ${c.rank}.`),f(p.shape[3]===c.shape[2],()=>`Error in depthwiseConv2d: number of input channels (${p.shape[3]}) must match the inChannels dimension in filter ${c.shape[2]}.`),a!=null&&f(J(r),()=>`Error in depthwiseConv2d: pad must be an integer when using, dimRoundingMode ${a} but got pad ${r}.`);const u=(w,L)=>{o==null&&(o=[1,1]),f(ae(n,o),()=>`Error in depthwiseConv2d: Either strides or dilations must be 1. Got strides ${n} and dilations '${o}'`);const T=We(p.shape,c.shape,n,o,r,a,!0),A=w.depthwiseConv2D(p,c,T);return L([p,c]),A},m={x:p,filter:c},b={strides:n,pad:r,dataFormat:i,dilations:o,dimRoundingMode:a},y=g.runKernelFunc(u,m,null,ao,b);return l?x(y,[y.shape[1],y.shape[2],y.shape[3]]):y}const zn=d({depthwiseConv2d_:UR});function WR(e){const t=h(e,"x","diag"),n=i=>{const o=x(t,[t.size]),a=i.diag(o),s=[...e.shape,...e.shape];return x(a,s)},r={x:t};return g.runKernelFunc(n,r,null,Ju)}const cu=d({diag_:WR});function qR(e,t,n,r,i=[1,1],o="NHWC"){const a=h(e,"x","dilation2d"),s=h(t,"filter","dilation2d");f(a.rank===3||a.rank===4,()=>`Error in dilation2d: input must be rank 3 or 4, but got rank ${a.rank}.`),f(s.rank===3,()=>`Error in dilation2d: filter must be rank 3, but got rank ${s.rank}.`),f(o==="NHWC",()=>`Error in dilation2d: Only NHWC is currently supported, but got dataFormat of ${o}`);let c=a,p=!1;a.rank===3&&(c=x(a,[1,a.shape[0],a.shape[1],a.shape[2]]),p=!0);const l={x:c,filter:s},u={strides:n,pad:r,dilations:i},m=g.runKernel(co,l,u);return p?x(m,[m.shape[1],m.shape[2],m.shape[3]]):m}const au=d({dilation2d_:qR});function BR(e,t){let n=h(e,"a","floorDiv"),r=h(t,"b","floorDiv");[n,r]=V(n,r);const i=(a,s)=>{const c=a.floorDiv(n,r);return s([n,r]),c},o={a:n,b:r};return g.runKernelFunc(i,o,null,go)}const Ri=d({floorDiv_:BR});function PR(e,t){let n=h(e,"a","div"),r=h(t,"b","div");if([n,r]=V(n,r),n.dtype==="int32"&&r.dtype==="int32")return Ri(n,r);const i=(s,c)=>{const p=s.realDivide(n,r);return c([n,r]),p},o={a:n,b:r},a={};return g.runKernelFunc(i,o,null,po,a)}const M=d({div_:PR});function jR(e,t){const n=e.length,r=[];for(let i=0;i<n;i++){const o=n-1-i,a=e[o]||1,s=t[t.length-1-i]||1;s>1&&a===1&&r.unshift(o)}return r}function ie(e,t){const n=[];for(let r=0;r<t.length;r++){const i=e[e.length-r-1],o=t.length-r-1,a=t[o];(i==null||i===1&&a>1)&&n.unshift(o)}return n}function H(e,t){const n=[],r=Math.max(e.length,t.length);for(let i=0;i<r;i++){let o=e[e.length-i-1];o==null&&(o=1);let a=t[t.length-i-1];if(a==null&&(a=1),o===1)n.unshift(a);else if(a===1)n.unshift(o);else if(o!==a){const s=`Operands could not be broadcast together with shapes ${e} and ${t}.`;throw Error(s)}else n.unshift(o)}return n}function GR(e,t){let n=h(e,"a","equal"),r=h(t,"b","equal");[n,r]=V(n,r),H(n.shape,r.shape);const i=a=>a.equal(n,r),o={a:n,b:r};return g.runKernelFunc(i,o,null,eh)}const Kt=d({equal_:GR});function $R(e,t,n){const r=h(t,"a","where"),i=h(n,"b","where"),o=h(e,"condition","where","bool"),a=H(r.shape,i.shape),s=Ir(r,a),c=Ir(i,a);o.rank===1&&f(o.shape[0]===r.shape[0],()=>"The first dimension of `a` must match the size of `condition`."),o.rank!==1&&$(o.shape,c.shape,"Error in where: ");const p=(u,m)=>{const b=u.select(o,s,c);return m([o]),b},l={condition:o,t:s,e:c};return g.runKernelFunc(p,l,null,Xo)}const Ae=d({where_:$R});function HR(e){const t=h(e,"x","zerosLike"),n={x:t};return g.runKernelFunc(r=>r.zerosLike(t),n,null,ws)}const W=d({zerosLike_:HR});function YR(e,t){let n=h(e,"a","div"),r=h(t,"b","div");[n,r]=V(n,r);const i=M(n,r),o=W(i),a=Kt(r,o);return Ae(a,o,i)}const su=d({divNoNan_:YR});function zR(e,t){const n=h(e,"t1","dot"),r=h(t,"t2","dot");f((n.rank===1||n.rank===2)&&(r.rank===1||r.rank===2),()=>`Error in dot: inputs must all be rank 1 or 2, but got ranks ${n.rank} and ${r.rank}.`);const i=n.rank===1?n.size:n.shape[1],o=r.rank===1?r.size:r.shape[0];if(f(i===o,()=>`Error in dot: inner dimensions of inputs must match, but got ${i} and ${o}.`),n.rank===1&&r.rank===1){const a=x(n,[1,-1]),s=x(r,[-1,1]),c=K(a,s);return x(c,[])}else if(n.rank===1&&r.rank===2){const a=x(n,[1,-1]),s=x(r,[r.shape[0],r.shape[1]]),c=K(a,s);return x(c,[c.size])}else if(n.rank===2&&r.rank===1){const a=x(r,[-1,1]),s=K(n,a);return x(s,[s.size])}else{const a=x(r,[r.shape[0],r.shape[1]]),s=K(n,a);return s}}const ou=d({dot_:zR});function VR(e){const t=h(e,"x","elu"),n=(i,o)=>{const a=i.elu(t);return o([a]),a},r={x:t};return g.runKernelFunc(n,r,null,lo)}const Ei=d({elu_:VR});function KR(e){let t=h(e,"x","erf");f(t.dtype==="int32"||t.dtype==="float32",()=>"Input dtype must be `int32` or `float32`."),t.dtype==="int32"&&(t=R(t,"float32"));const n={x:t};return g.runKernelFunc((r,i)=>{const o=r.erf(t);return i([t]),o},n,null,uo)}const iu=d({erf_:KR});function JR(e){const t=h(e,"x","exp"),n={x:t};return g.runKernelFunc((r,i)=>{const o=r.exp(t);return i([o]),o},n,null,ho)}const Ne=d({exp_:JR});function XR(e,t=0){const n=null,r=h(e,"x","expandDims",n);f(t<=r.rank,()=>"Axis must be <= rank of the tensor");const i=r.shape.slice();return t<0&&(f(-(r.rank+1)<=t,()=>`Axis must be in the interval [${-(r.rank+1)}, ${r.rank}]`),t=r.rank+t+1),i.splice(t,0,1),x(r,i)}const Ke=d({expandDims_:XR});function ZR(e){const t=h(e,"x","expm1"),n={x:t};return g.runKernelFunc((r,i)=>{const o=r.expm1(t);return i([t]),o},n,null,mo)}const ru=d({expm1_:ZR});function QR(e,t){const n=null,r=h(e,"x","tile",n);f(r.rank===t.length,()=>`Error in transpose: rank of input ${r.rank} must match length of reps ${t}.`);const i=(c,p)=>{const l=c.tile(r,t);return p([r]),l},o=[r],a={x:r},s={reps:t};return g.runKernelFunc(i,a,null,ms,s,o)}const Ht=d({tile_:QR});function e1(e,t,n,r="float32"){t==null&&(t=e);const i=ke([e,t],r),o=e<=t?e:t;for(let s=0;s<o;++s)i.set(1,s,s);const a=x(i.toTensor(),[e,t]);if(n==null)return a;if(n.length===1)return Ht(Ke(a,0),[n[0],1,1]);if(n.length===2)return Ht(Ke(Ke(a,0),0),[n[0],n[1],1,1]);if(n.length===3)return Ht(Ke(Ke(Ke(a,0),0),0),[n[0],n[1],n[2],1,1]);throw new Error(`eye() currently supports only 1D and 2D batchShapes, but received ${n.length}D.`)}const Ci=d({eye_:e1});function t1(e){f(e.dtype==="complex64",()=>`The dtype for tf.spectral.fft() must be complex64 but got ${e.dtype}.`);const t={input:e};return g.runKernelFunc(n=>{const r=e.shape[e.shape.length-1],i=e.size/r,o=e.as2D(i,r),a=n.fft(o);return a.reshape(e.shape)},t,null,th)}const Yn=d({fft_:t1});function Hn(e,t,n){const r={shape:e,value:t,dtype:n};return g.runKernelFunc(i=>i.fill(e,t,n),{},null,nh,r)}function n1(e){const t=h(e,"x","floor"),n={x:t};return g.runKernelFunc(r=>r.floor(t),n,null,fo)}const $n=d({floor_:n1});const Mp=30;function r1(e){return e<=Mp?e:ii(e,Math.floor(Math.sqrt(e)))}const Wy={};Oe(Wy,{collectGatherOpShapeInfo:()=>Bm,computeOutShape:()=>o1,segOpComputeOptimalWindowSize:()=>i1});function i1(e,t){let n=!1,r;for(e<=Mp?(r=e,n=!0):r=ii(e,Math.floor(Math.sqrt(e)));!n;)r>t||r===e?n=!0:r=ii(e,r+1);return r}function o1(e,t,n){const r=[],i=e.length;for(let o=0;o<i;o++)o!==t?r.push(e[o]):r.push(n);return r}function Bm(e,t,n){const r=e.shape[n],i=[];let o=1,a=1;for(let s=0;s<n;s++)i.push(e.shape[s]),o*=e.shape[s];for(let s=0;s<t.rank;s++)i.push(t.shape[s]);for(let s=n+1;s<e.rank;s++)i.push(e.shape[s]),a*=e.shape[s];return{batchSize:o,sliceSize:a,dimSize:r,outputShape:i}}function s1(e,t,n=0){const r=h(e,"x","gather"),i=h(t,"indices","gather","int32"),o={x:r,indices:i},a={axis:n},s=(c,p)=>{const l=z(n,r.shape)[0],u=Bm(r,i,l),m=c.gather(r,x(i,[i.size]),l);return p([r,i]),x(m,u.outputShape)};return g.runKernelFunc(s,o,null,wo,a)}const Gn=d({gather_:s1});function a1(e,t){let n=h(e,"a","greater"),r=h(t,"b","greater");[n,r]=V(n,r),H(n.shape,r.shape);const i=a=>a.greater(n,r),o={a:n,b:r};return g.runKernelFunc(i,o,null,oh)}const Ve=d({greater_:a1});function c1(e,t){let n=h(e,"a","greaterEqual"),r=h(t,"b","greaterEqual");[n,r]=V(n,r),H(n.shape,r.shape);const i=(a,s)=>{const c=a.greaterEqual(n,r);return s([n,r]),c},o={a:n,b:r};return g.runKernelFunc(i,o,null,yo)}const ct=d({greaterEqual_:c1});function p1(e){f(e.dtype==="complex64",()=>`The dtype for tf.spectral.ifft() must be complex64 but got ${e.dtype}.`);const t={input:e};return g.runKernelFunc(n=>{const r=e.shape[e.shape.length-1],i=e.size/r,o=x(e,[i,r]),a=n.ifft(o);return x(a,e.shape)},t,null,sh)}const cn=d({ifft_:p1});function l1(e){const t=h(e,"input","imag"),n=i=>i.imag(t),r={input:t};return g.runKernelFunc(n,r,null,ah)}const Vt=d({imag_:l1});function u1(e){const t=h(e,"input","real"),n=i=>i.real(t),r={input:t};return g.runKernelFunc(n,r,null,Th)}const kt=d({real_:u1});function h1(e,t){const n=h(e,"x","reverse"),r=a=>{const s=z(t,n.shape);if(n.rank===0)return At(n);const c=a.reverse(n,s);return x(c,n.shape)},i={x:n},o={dims:t};return g.runKernelFunc(r,i,null,Vo,o)}const Pe=d({reverse_:h1});function k(e,t){if((Ie(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"&&Ie(e)&&!(e instanceof Uint8Array))throw new Error("When making a scalar from encoded string, the value must be `Uint8Array`.");const n=[],r=[];return ze(e,n,r,t)}function d1(e){const t=e.shape[e.shape.length-1],n=e.size/t;let r;if(t<=2){const i=x(e,[n,t]);r=cn(i)}else{const i=[n,2*(t-1)],o=x(kt(e),[n,t]),a=x(Vt(e),[n,t]),s=Pe(Y(o,[0,1],[n,t-2]),1),c=v(Pe(Y(a,[0,1],[n,t-2]),1),k(-1)),p=re([o,s],1),l=re([a,c],1),u=x(Ge(p,l),[i[0],i[1]]);r=cn(u)}if(r=kt(r),e.rank===3&&e.shape[0]!==0){const i=r,o=e.shape[0];r=x(r,[o,r.shape[0]/o,r.shape[1]]),i.dispose()}return r}const _i=d({irfft_:d1});function m1(e){const t=h(e,"x","isFinite"),n={x:t};return g.runKernelFunc(r=>r.isFinite(t),n,null,Lo)}const tu=d({isFinite_:m1});function f1(e){const t=h(e,"x","isInf"),n={x:t};return g.runKernelFunc(r=>r.isInf(t),n,null,vo)}const eu=d({isInf_:f1});function g1(e){const t=h(e,"x","isNaN"),n={x:t};return g.runKernelFunc(r=>r.isNaN(t),n,null,So)}const Ql=d({isNaN_:g1});function b1(e,t){let n=h(e,"a","maximum"),r=h(t,"b","maximum");[n,r]=V(n,r),n.dtype==="bool"&&(n=R(n,"int32"),r=R(r,"int32")),H(n.shape,r.shape);const i=(a,s)=>{const c=a.maximum(n,r);return s([n,r]),c},o={a:n,b:r};return g.runKernelFunc(i,o,null,Ro)}const Tt=d({maximum_:b1});function w1(e,t=.2){const n=h(e,"x","leakyRelu");return Tt(v(k(t),n),n)}const Zl=d({leakyRelu_:w1});function y1(e,t){let n=h(e,"a","less"),r=h(t,"b","less");[n,r]=V(n,r),H(n.shape,r.shape);const i=a=>a.less(n,r),o={a:n,b:r};return g.runKernelFunc(i,o,null,ch)}const jn=d({less_:y1});function x1(e,t){let n=h(e,"a","lessEqual"),r=h(t,"b","lessEqual");[n,r]=V(n,r),H(n.shape,r.shape);const i=(a,s)=>{const c=a.lessEqual(n,r);return s([n,r]),c},o={a:n,b:r};return g.runKernelFunc(i,o,null,ph)}const at=d({lessEqual_:x1});function Xl(e,t,n){if(n<=0)throw new Error("The number of values should be positive.");const r={start:e,stop:t,num:n};return g.runKernelFunc(i=>i.linspace(e,t,n),{},null,lh,r)}function L1(e,t=5,n=1,r=1,i=.5){const o=h(e,"x","localResponseNormalization");f(o.rank===4||o.rank===3,()=>`Error in localResponseNormalization: x must be rank 3 or 4 but got
rank ${o.rank}.`),f(J(t),()=>`Error in localResponseNormalization: depthRadius must be an integer but got depthRadius ${t}.`);let a=o,s=!1;o.rank===3&&(s=!0,a=x(o,[1,o.shape[0],o.shape[1],o.shape[2]]));const c=(m,b)=>{const y=m.localResponseNormalization4D(a,t,n,r,i);return b([a,y]),y},p={x:a},l={depthRadius:t,bias:n,alpha:r,beta:i},u=g.runKernelFunc(c,p,null,No,l);return s?x(u,[u.shape[1],u.shape[2],u.shape[3]]):u}const Jl=d({localResponseNormalization_:L1});function v1(e){const t=h(e,"x","log"),n={x:t};return g.runKernelFunc((r,i)=>{const o=r.log(t);return i([t]),o},n,null,To)}const gt=d({log_:v1});function S1(e){const t=h(e,"x","log1p"),n={x:t};return g.runKernelFunc((r,i)=>{const o=r.log1p(t);return i([t]),o},n,null,Io)}const Ni=d({log1p_:S1});function aS(e){return f(qt(e),()=>"The f passed in grad(f) must be a function"),(t,n)=>{const r=h(t,"x","tf.grad",null),i=n!=null?h(n,"dy","tf.grad"):null;return g.tidy(()=>{const{value:o,grads:a}=g.gradients(()=>e(r),[r],i);return i!=null&&$(o.shape,i.shape,"The shape of dy passed in grad(f)(x, dy) must match the shape returned by f(x)"),Up(a),a[0]})}}function cS(e){return f(qt(e),()=>"The f passed in grads(f) must be a function"),(t,n)=>{f(Array.isArray(t),()=>"The args passed in grads(f)(args) must be an array of `Tensor`s or `TensorLike`s");const r=rn(t,"args","tf.grads",null),i=n!=null?h(n,"dy","tf.grads"):null;return g.tidy(()=>{const{value:o,grads:a}=g.gradients(()=>e(...r),r,i);return i!=null&&$(o.shape,i.shape,"The shape of dy passed in grads(f)([x1,...], dy) must match the shape returned by f([x1,...])"),Up(a),a})}}function pS(e){return f(qt(e),()=>"The f passed in valueAndGrad(f) must be a function"),(t,n)=>{f(t instanceof de,()=>"The x passed in valueAndGrad(f)(x) must be a tensor"),f(n==null||n instanceof de,()=>"The dy passed in valueAndGrad(f)(x, dy) must be a tensor");const{grads:r,value:i}=g.gradients(()=>e(t),[t],n);return Up(r),{grad:r[0],value:i}}}function lS(e){return f(qt(e),()=>"The f passed in valueAndGrads(f) must be a function"),(t,n)=>{f(Array.isArray(t)&&t.every(i=>i instanceof de),()=>"The args passed in valueAndGrads(f)(args) must be array of tensors"),f(n==null||n instanceof de,()=>"The dy passed in valueAndGrads(f)(args, dy) must be a tensor");const r=g.gradients(()=>e(...t),t,n);return n!=null&&$(r.value.shape,n.shape,"The shape of dy passed in valueAndGrads(f)([x1,...], dy) must match the shape returned by f([x1,...])"),Up(r.grads),r}}function kh(e,t){f(qt(e),()=>"The f passed in variableGrads(f) must be a function"),f(t==null||Array.isArray(t)&&t.every(p=>p instanceof Nn),()=>"The varList passed in variableGrads(f, varList) must be an array of variables");const n=t!=null;if(!n){t=[];for(const p in g.registeredVariables)t.push(g.registeredVariables[p])}const r=n?t.filter(p=>!p.trainable):null,i=t.length;t=t.filter(p=>p.trainable),f(t.length>0,()=>`variableGrads() expects at least one of the input variables to be trainable, but none of the ${i} variables is trainable.`);const o=!0,{value:a,grads:s}=g.gradients(e,t,null,o);f(s.some(p=>p!=null),()=>"Cannot find a connection between any variable and the result of the loss function y=f(x). Please make sure the operations that use variables are inside the function f passed to minimize()."),f(a.rank===0,()=>`The f passed in variableGrads(f) must return a scalar, but it returned a rank-${a.rank} tensor`);const c={};return t.forEach((p,l)=>{s[l]!=null&&(c[p.name]=s[l])}),r!=null&&r.forEach(p=>c[p.name]=null),{value:a,grads:c}}function $e(e){return g.customGrad(e)}function Up(e){const t=e.filter(n=>n==null).length;if(t>0)throw new Error(`Cannot compute gradient of y=f(x) with respect to x. Make sure that
the f you passed encloses all operations that lead from x to y.`)}function T1(e){const t=h(e,"x","neg"),n={x:t};return g.runKernelFunc(r=>r.neg(t),n,null,Mo)}const Q=d({neg_:T1});function I1(e){const t=h(e,"x","softplus"),n={x:t};return g.runKernelFunc((r,i)=>{const o=r.softplus(t);return i([t]),o},n,null,is)}const gi=d({softplus_:I1});function A1(e){const t=h(e,"x","logSigmoid"),n=$e(r=>{const i=Q(gi(Q(r))),o=a=>{const s=v(a,St(Q(r)));return s};return{value:i,gradFunc:o}});return n(t)}const Kl=d({logSigmoid_:A1});function N1(e,t=null,n=!1){const r=h(e,"x","max"),i=(s,c)=>{const p=z(t,r.shape);let l=p;const u=he(l,r.rank);let m=r;u!=null&&(m=X(r,u),l=Ee(l.length,m.rank));const b=s.max(m,l);u!=null&&m.dispose();let y=b;if(n){const w=fe(y.shape,z(t,r.shape));y=x(y,w),b.dispose()}return c([r,y]),y},o={x:r},a={reductionIndices:t,keepDims:n};return g.runKernelFunc(i,o,null,_o,a)}const st=d({max_:N1});function _1(e,t){let n=h(e,"a","sub"),r=h(t,"b","sub");[n,r]=V(n,r);const i=(a,s)=>{const c=a.subtract(n,r);return s([n,r]),c},o={a:n,b:r};return g.runKernelFunc(i,o,null,us)}const D=d({sub_:_1});function R1(e,t=null,n=!1){let r=h(e,"x","sum");r.dtype==="bool"&&(r=R(r,"int32"));const i=(s,c)=>{c([r]);const p=z(t,r.shape),l=he(p,r.rank);let u=p,m=r;l!=null&&(m=X(r,l),u=Ee(u.length,r.rank));let b=s.sum(m,u);if(n){const y=fe(b.shape,p);b=x(b,y)}return b},o={x:r},a={axis:t,keepDims:n};return g.runKernelFunc(i,o,null,ss,a)}const U=d({sum_:R1});function C1(e,t=-1){const n=h(e,"logits","logSoftmax");if(t===-1&&(t=n.rank-1),t!==n.rank-1)throw Error(`Log Softmax along a non-last dimension is not yet supported. Logits was rank ${n.rank} and axis was ${t}`);const r=(a,s)=>{const c=!0,p=st(e,t,!0),l=D(e,p),u=D(R(l,"float32"),gt(U(Ne(l),t,c)));return s([u]),u},i={logits:n},o={axis:t};return g.runKernelFunc(r,i,null,Ao,o)}const Vl=d({logSoftmax_:C1});function E1(e,t=null,n=!1){const r=h(e,"x","logSumExp"),i=z(t,r.shape),o=st(r,i,!0),a=D(r,o),s=Ne(a),c=U(s,i),p=gt(c),l=O(x(o,p.shape),p);if(n){const u=fe(l.shape,i);return x(l,u)}return l}const Ai=d({logSumExp_:E1});function O1(e,t){const n=h(e,"a","logicalAnd","bool"),r=h(t,"b","logicalAnd","bool");H(n.shape,r.shape);const i={a:n,b:r};return g.runKernelFunc(o=>o.logicalAnd(n,r),i,null,uh)}const ft=d({logicalAnd_:O1});function k1(e){const t=h(e,"x","logicalNot","bool"),n={x:t};return g.runKernelFunc(r=>r.logicalNot(t),n,null,hh)}const Pn=d({logicalNot_:k1});function D1(e,t){const n=h(e,"a","logicalOr","bool"),r=h(t,"b","logicalOr","bool");H(n.shape,r.shape);const i={a:n,b:r};return g.runKernelFunc(o=>o.logicalOr(n,r),i,null,dh)}const Ii=d({logicalOr_:D1});function F1(e,t){const n=h(e,"a","logicalXor","bool"),r=h(t,"b","logicalXor","bool");return H(n.shape,r.shape),ft(Ii(e,t),Pn(ft(e,t)))}const zl=d({logicalXor_:F1});function M1(e,t,n,r,i){const o=h(e,"x","maxPool"),a=1;let s=o,c=!1;o.rank===3&&(c=!0,s=x(o,[1,o.shape[0],o.shape[1],o.shape[2]])),f(s.rank===4,()=>`Error in maxPool: input must be rank 4 but got rank ${s.rank}.`),f(ae(n,a),()=>`Error in maxPool: Either strides or dilations must be 1. Got strides ${n} and dilations '${a}'`),i!=null&&f(J(r),()=>`Error in maxPool: pad must be an integer when using, dimRoundingMode ${i} but got pad ${r}.`);const p=(b,y)=>{const w=Ct(s.shape,t,n,1,r,i);let L;return w.filterWidth===1&&w.filterHeight===1&&Le(w.inShape,w.outShape)?L=s.clone():L=b.maxPool(s,w),y([s,L]),L},l={x:s},u={filterSize:t,strides:n,pad:r,dimRoundingMode:i},m=g.runKernelFunc(p,l,null,Co,u);return c?x(m,[m.shape[1],m.shape[2],m.shape[3]]):m}const Ti=d({maxPool_:M1});function U1(e,t=[1,1,1],n,r,i,o="NDHWC",a){a==null?a=[1,1,1]:ge("dilations is deprecated, this field will be gone in v3.0.0.");const s=h(e,"x","maxPool3d");let c=s,p=!1;s.rank===4&&(p=!0,c=x(s,[1,s.shape[0],s.shape[1],s.shape[2],s.shape[3]])),f(c.rank===5,()=>`Error in maxPool3d: x must be rank 5 but got rank ${c.rank}.`),f(o==="NDHWC",()=>`Error in maxPool3d: Only NDHWC is currently supported, but got dataFormat of ${o}`),f(ae(n,a),()=>`Error in maxPool3d: Either strides or dilations must be 1. Got strides ${n} and dilations '${a}'`),i!=null&&f(J(r),()=>`Error in maxPool3d: pad must be an integer when using, dimRoundingMode ${i} but got pad ${r}.`);const l=(y,w)=>{a==null&&(a=[1,1,1]);const L=on(c.shape,t,n,a,r,i,o),T=y.maxPool3d(c,L);return w([c,T]),T},u={x:c},m={filterSize:t,strides:n,pad:r,dimRoundingMode:i,dataFormat:o,dilations:a},b=g.runKernelFunc(l,u,null,Eo,m);return p?x(b,[b.shape[1],b.shape[2],b.shape[3],b.shape[4]]):b}const Yl=d({maxPool3d_:U1});function W1(e,t,n,r,i=!1){const o=h(e,"x","maxPoolWithArgmax"),a={x:o},s={filterSize:t,strides:n,pad:r,includeBatchInIndex:i},c=g.runKernel(bh,a,s);return{result:c[0],indexes:c[1]}}const Hl=d({maxPoolWithArgmax_:W1});function Ue(e,t="float32"){if(t==="complex64"){const r=Ue(e,"float32"),i=Ue(e,"float32");return Ge(r,i)}const n=Pt(Z(e),t);return g.makeTensor(n,e,t)}function mt(e,t="float32"){if(t==="complex64"){const r=mt(e,"float32"),i=Ue(e,"float32");return Ge(r,i)}const n=Ws(Z(e),t);return g.makeTensor(n,e,t)}function q1(e,t=null,n=!1){const r=h(e,"x","mean"),i=z(t,r.shape),o=Nm(r.shape,i),a=o[1],s=Z(a),c=$e(p=>{const l=k(s),u=l.dtype===p.dtype?p:R(p,l.dtype),m=M(u,l),b=U(m,t,n),y=w=>{const L=p.shape.slice();i.forEach(_=>{L[_]=1});const T=x(w,L),A=M(v(T,mt(p.shape,"float32")),s);return A};return{value:b,gradFunc:y}});return c(r)}const an=d({mean_:q1});function B1(e,t=null,n=!1){const r=h(e,"x","min"),i=(s,c)=>{const p=z(t,r.shape);let l=p;const u=he(l,r.rank);let m=r;u!=null&&(m=X(r,u),l=Ee(l.length,r.rank));const b=s.min(m,l);u!=null&&m.dispose();let y=b;if(n){const w=fe(y.shape,p);y=x(b,w),b.dispose()}return c([r,y]),y},o={x:r},a={axis:t,keepDims:n};return g.runKernelFunc(i,o,null,Oo,a)}const sn=d({min_:B1});function P1(e,t){let n=h(e,"a","minimum"),r=h(t,"b","minimum");[n,r]=V(n,r),n.dtype==="bool"&&(n=R(n,"int32"),r=R(r,"int32")),H(n.shape,r.shape);const i=(a,s)=>{const c=a.minimum(n,r);return s([n,r]),c},o={a:n,b:r};return g.runKernelFunc(i,o,null,ko)}const Bn=d({minimum_:P1});function j1(e,t){let n=h(e,"a","mod"),r=h(t,"b","mod");[n,r]=V(n,r);const i=(a,s)=>{const c=a.mod(n,r);return s([n,r]),c},o={a:n,b:r};return g.runKernelFunc(i,o,null,Do)}const Si=d({mod_:j1});function G1(e){const t=h(e,"x","square"),n={},r=[t],i=[];return g.runKernelFunc((o,a)=>(a([t]),o.square(t)),{x:t},null,"Square",n,r,i)}const G=d({square_:G1});function $1(e,t=null,n=!1){e=h(e,"x","moments");const r=z(t,e.shape),i=an(e,r,n);let o=i.shape;n||(o=fe(i.shape,r));const a=G(D(R(e,"float32"),x(i,o))),s=an(a,r,n);return{mean:i,variance:s}}const $l=d({moments_:$1});function H1(e,t,n,r){const i=h(t,"data","multiRNNCell"),o=rn(n,"c","multiRNNCell"),a=rn(r,"h","multiRNNCell");let s=i;const c=[];for(let u=0;u<e.length;u++){const m=e[u](s,o[u],a[u]);c.push(m[0]),c.push(m[1]),s=m[1]}const p=[],l=[];for(let u=0;u<c.length;u+=2)p.push(c[u]),l.push(c[u+1]);return[p,l]}const Gl=d({multiRNNCell_:H1});function Y1(e,t,n,r=!1){const i=h(e,"logits","multinomial"),o=i.size,a=i.rank;if(o<2)throw new Error(`Error in multinomial: you need at least 2 outcomes, but got ${o}.`);if(a>2)throw new Error(`Rank of probabilities must be 1 or 2, but is ${a}`);n=n||Math.random();const s=a===1?x(i,[1,-1]):i,c=g.runKernelFunc(p=>p.multinomial(s,r,t,n),{logits2D:s});return a===1?x(c,[c.size]):c}const jl=d({multinomial_:Y1});function z1(e,t){let n=h(e,"a","notEqual"),r=h(t,"b","notEqual");[n,r]=V(n,r),H(n.shape,r.shape);const i=a=>a.notEqual(n,r),o={a:n,b:r};return g.runKernelFunc(i,o,null,wh)}const qn=d({notEqual_:z1});function V1(e,t,n=1,r=0){if(t<2)throw new Error(`Error in oneHot: depth must be >=2, but it is ${t}`);const i=h(e,"indices","oneHot","int32"),o=[...i.shape,t],a=(p,l)=>(l([i]),x(p.oneHot(x(i,[i.size]),t,n,r),o)),s={indices:i},c={depth:t,onValue:n,offValue:r};return g.runKernelFunc(a,s,null,Wo,c)}const Sr=d({oneHot_:V1});function K1(e){const t=h(e,"x","onesLike"),n=(i,o)=>{if(t.dtype==="complex64"){const a=ea(kt(t)),s=W(Vt(t));return Ge(a,s)}return i.onesLike(t)},r={x:t};return g.runKernelFunc(n,r,null,Uo)}const ea=d({onesLike_:K1});function J1(e,t){const n=h(e,"v1","outerProduct"),r=h(t,"v2","outerProduct");f(n.rank===1&&r.rank===1,()=>`Error in outerProduct: inputs must be rank 1, but got ranks ${n.rank} and ${r.rank}.`);const i=x(n,[-1,1]),o=x(r,[1,-1]);return K(i,o)}const Pl=d({outerProduct_:J1});function X1(e,t,n=0){const r=h(e,"x","pad");if(r.rank===0)throw new Error("pad(scalar) is not defined. Pass non-scalar to pad");const i=(s,c)=>(c([r]),s.pad(r,t,n)),o={paddings:t,constantValue:n},a={x:r};return g.runKernelFunc(i,a,null,qo,o)}const ot=d({pad_:X1});function Z1(e,t,n=0){return f(t.length===2,()=>"Invalid number of paddings. Must be length of 2."),ot(e,[t],n)}const Bl=d({pad1d_:Z1});function Q1(e,t,n=0){return f(t.length===2&&t[0].length===2&&t[1].length===2,()=>"Invalid number of paddings. Must be length of 2 each."),ot(e,t,n)}const ql=d({pad2d_:Q1});function eC(e,t,n=0){return f(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."),ot(e,t,n)}const Wl=d({pad3d_:eC});function tC(e,t,n=0){return f(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."),ot(e,t,n)}const Ul=d({pad4d_:tC});function nC(e,t,n){const r=h(e,"x","spaceToBatchND");f(r.rank>=1+t.length,()=>`input rank ${r.rank} should be > than [blockShape] ${t.length}`),f(n.length===t.length,()=>`paddings.shape[0] ${n.length} must be equal to [blockShape] ${t.length}`),f(r.shape.reduce((s,c,p)=>p>0&&p<=t.length?s&&(c+n[p-1][0]+n[p-1][1])%t[p-1]===0:s,!0),()=>`input spatial dimensions ${r.shape.slice(1)} with paddings ${n.toString()} must be divisible by blockShapes ${t.toString()}`);const i=s=>s.spaceToBatchND(r,t,n),o={x:r},a={blockShape:t,paddings:n};return g.runKernelFunc(i,o,null,as,a)}const Un=d({spaceToBatchND_:nC});function oC(e,t,n,r,i,o){i==null&&(i=[1,1]),o==null&&(o=1),r===0&&(r="valid");const a=h(e,"x","maxPool");let s=a,c=!1;a.rank===3&&(c=!0,s=x(a,[1,a.shape[0],a.shape[1],a.shape[2]])),f(ae(o,i),()=>`Error in pool: Either strides or dilations must be 1. Got strides ${o} and dilations '${i}'`);const p=Ct(s.shape,t,o,i,r),l=[p.dilationHeight,p.dilationWidth];let u;r==="same"?u=iC([p.filterHeight,p.filterWidth],l):u=[[0,0],[0,0]];const m=l[0]===1&&l[1]===1,[b,y]=rC([p.inHeight,p.inWidth],l,u),w=m?r:"valid",L=m?s:Un(s,l,b),T=n==="avg"?()=>Di(L,t,o,w):()=>Ti(L,t,o,w),A=T(),_=m?A:Kn(A,l,y);return c?x(_,[_.shape[1],_.shape[2],_.shape[3]]):_}function rC(e,t,n){const r=n.map(l=>l[0]),i=n.map(l=>l[1]),o=e.concat(r,i),a=t.map((l,u)=>(l-o[u]%l)%l),s=i.map((l,u)=>l+a[u]),c=t.map((l,u)=>[r[u],s[u]]),p=t.map((l,u)=>[0,a[u]]);return[c,p]}function iC(e,t){const n=e.map((a,s)=>a+(a-1)*(t[s]-1)),r=n.map(a=>a-1),i=r.map(a=>Math.floor(a/2)),o=r.map((a,s)=>a-i[s]);return r.map((a,s)=>[i[s],o[s]])}const Ml=d({pool_:oC});function sC(e,t){let n=h(e,"base","pow"),r=h(t,"exp","pow");[n,r]=V(n,r);const i={a:n,b:r},o=(a,s)=>{const c=a.pow(n,r);return s([n,r,c]),c};return g.runKernelFunc(o,i,null,Bo)}const je=d({pow_:sC});function aC(e,t){const n=h(e,"x","prelu"),r=h(t,"alpha","prelu"),i=(a,s)=>{const c=a.prelu(n,r);return s([n,r]),c},o={x:n,alpha:r};return g.runKernelFunc(i,o,null,Po)}const vi=d({prelu_:aC});function Fl(e,t=!1){console.log(e.toString(t))}function cC(e,t=null,n=!1){let r=h(e,"x","prod");const i=s=>{r.dtype==="bool"&&(r=R(r,"int32"));const c=z(t,r.shape),p=he(c,r.rank);let l=c,u=r;p!=null&&(u=X(r,p),l=Ee(l.length,r.rank));let m=s.prod(u,l);if(n){const b=fe(m.shape,c);m=x(m,b)}return m},o={x:r},a={axis:t,keepDims:n};return g.runKernelFunc(i,o,null,vh,a)}const Dl=d({prod_:cC});function pC(e,t,n){const r=Z(e);let i=null;if(n==null||n==="float32")i=new Float32Array(r);else if(n==="int32")i=new Int32Array(r);else if(n==="bool")i=new Uint8Array(r);else throw new Error(`Unknown data type ${n}`);for(let o=0;o<r;o++)i[o]=t();return g.makeTensor(i,e,n)}const kl=d({rand_:pC});const Rf={};Oe(Rf,{TEST_EPSILON_FLOAT16:()=>qy,expectArrayBuffersEqual:()=>gC,expectArraysClose:()=>uC,expectArraysEqual:()=>dC,expectNumbersClose:()=>mC,expectPromiseToFail:()=>hC,expectValuesInRange:()=>fC,testEpsilon:()=>Pm});const lC=.001,qy=.1;function uC(e,t,n){return n==null&&(n=Pm()),jm(e,t,(r,i)=>Gm(r,i,n))}function Pm(){return g.backend.floatPrecision()===32?lC:qy}function jm(e,t,n){let r=!0;if((Ie(e)||Ie(t))&&(r=!1),Ie(e)&&Ie(t)&&(r=!0),r){const a=e.constructor.name,s=t.constructor.name;if(a!==s)throw new Error(`Arrays are of different type. Actual: ${a}. Expected: ${s}`)}if(Array.isArray(e)&&Array.isArray(t)){const a=Fe(e),s=Fe(t);if(!Le(a,s))throw new Error(`Arrays have different shapes. Actual: [${a}]. Expected: [${s}]`)}const i=Ie(e)?e:Wt(e),o=Ie(t)?t:Wt(t);if(i.length!==o.length)throw new Error(`Arrays have different lengths actual: ${i.length} vs expected: ${o.length}.
Actual: ${i}.
Expected: ${o}.`);for(let a=0;a<o.length;++a){const s=i[a],c=o[a];if(!n(s,c))throw new Error(`Arrays differ: actual[${a}] = ${s}, expected[${a}] = ${c}.
Actual: ${i}.
Expected: ${o}.`)}}function hC(e,t){e().then(()=>t.fail(),()=>t())}function dC(e,t){const n=typeof t=="string"||typeof t=="number"||typeof t=="boolean"?[t]:t;return Rt(e)||Rt(e[0])||Rt(t)||Rt(t[0])?jm(e,n,(r,i)=>r==i):jm(e,t,(r,i)=>Gm(r,i,0))}function mC(e,t,n){if(n==null&&(n=Pm()),!Gm(e,t,n))throw new Error(`Numbers differ: actual === ${e}, expected === ${t}`)}function Gm(e,t,n){return!isFinite(e)&&!isFinite(t)?!0:!(isNaN(e)||isNaN(t)||Math.abs(e-t)>n)}function fC(e,t,n){for(let r=0;r<e.length;r++)if(e[r]<t||e[r]>n)throw new Error(`Value out of range:${e[r]} low: ${t}, high: ${n}`)}function gC(e,t){expect(new Float32Array(e)).toEqual(new Float32Array(t))}const Wp=Iv(Nf());class zs{constructor(e,t,n,r,i){this.mean=e,this.stdDev=t,this.dtype=n,this.nextVal=NaN,this.truncated=r,this.truncated&&(this.upper=this.mean+this.stdDev*2,this.lower=this.mean-this.stdDev*2);const o=i||Math.random();this.random=Wp.alea(o.toString())}nextValue(){if(!isNaN(this.nextVal)){const r=this.nextVal;return this.nextVal=NaN,r}let e,t,n=!1;for(;!n;){let r,i,o;do r=2*this.random()-1,i=2*this.random()-1,o=r*r+i*i;while(o>=1||o===0);const a=Math.sqrt(-2*Math.log(o)/o);e=this.mean+this.stdDev*r*a,t=this.mean+this.stdDev*i*a,(!this.truncated||this.isValidTruncated(e))&&(n=!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}}class By{constructor(e,t,n,r){this.alpha=e,this.beta=1/t,this.dtype=n;const i=r||Math.random();this.randu=Wp.alea(i.toString()),this.randn=new zs(0,1,n,!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,n,r,i,o;for(;;){do r=this.randn.nextValue(),o=1+this.c*r;while(o<=0);if(o*=o*o,e=r*r,t=1-.331*e*e,n=.5*e+this.d*(1-o+Math.log(o)),i=this.randu(),i<t||Math.log(i)<n)break}return o=1/this.beta*this.d*o,this.alpha<1&&(o*=Math.pow(this.randu(),1/this.alpha)),this.convertValue(o)}convertValue(e){return this.dtype==="float32"?e:Math.round(e)}}class Py{constructor(e=0,t=1,n,r){if(this.canReturnFloat=()=>this.dtype==null||this.dtype==="float32",this.min=e,this.range=t-e,this.dtype=n,r==null&&(r=Math.random()),typeof r=="number"&&(r=r.toString()),!this.canReturnFloat()&&this.range<=1)throw new Error(`The difference between ${e} - ${t} <= 1 and dtype is not float`);this.random=Wp.alea(r)}convertValue(e){return this.canReturnFloat()?e:Math.round(e)}nextValue(){return this.convertValue(this.min+this.range*this.random())}}function bC(e,t,n=1,r="float32",i){if(n==null&&(n=1),r==null&&(r="float32"),r!=="float32"&&r!=="int32")throw new Error(`Unsupported data type ${r}`);const o=new By(t,n,r,i),a=ke(e,r);for(let s=0;s<a.values.length;s++)a.values[s]=o.nextValue();return a.toTensor()}const Ol=d({randomGamma_:bC});function wC(e,t=0,n=1,r,i){if(r!=null&&r==="bool")throw new Error(`Unsupported data type ${r}`);const o=new zs(t,n,r,!1,i),a=ke(e,r);for(let s=0;s<a.values.length;s++)a.values[s]=o.nextValue();return a.toTensor()}const El=d({randomNormal_:wC});function yC(e,t=0,n=1,r="float32",i){const o=ke(e,r),a=new Py(t,n,null,i);for(let s=0;s<o.values.length;s++)o.values[s]=a.nextValue();return o.toTensor()}const Li=d({randomUniform_:yC});function ht(e,t){lt(e);const n=Fe(e,t);if(n.length!==1)throw new Error("tensor1d() requires values to be a flat/TypedArray");const r=null;return ze(e,r,n,t)}function vr(e,t,n=1,r="float32"){if(n===0)throw new Error("Cannot have a step of zero");const i=()=>{const a=e===t,s=e<t&&n<0,c=t<e&&n>1;if(a||s||c)return Ue([0],r);const p=Math.abs(Math.ceil((t-e)/n)),l=Pt(p,r);t<e&&n===1&&(n=-1),l[0]=e;for(let u=1;u<l.length;u++)l[u]=l[u-1]+n;return ht(l,r)},o={start:e,stop:t,step:n,dtype:r};return g.runKernelFunc(i,{},null,Sh,o)}function xC(e){const t=h(e,"x","reciprocal"),n={x:t};return g.runKernelFunc((r,i)=>{const o=r.reciprocal(t);return i([t]),o},n,null,jo)}const Cl=d({reciprocal_:xC});function LC(e){const t=h(e,"x","relu"),n=(i,o)=>(o([t]),t.dtype==="bool"?R(t,"int32"):i.relu(t)),r={x:t};return g.runKernelFunc(n,r,null,Go)}const zt=d({relu_:LC});function vC(e){const t=h(e,"x","relu6"),n=(i,o)=>(o([t]),t.dtype==="bool"?R(t,"int32"):i.relu6(t)),r={x:t};return g.runKernelFunc(n,r,null,zo)}const xi=d({relu6_:vC});function SC(e){const t=h(e,"x","reverse");return f(t.rank===1,()=>`Error in reverse1D: x must be rank 1 but got rank ${t.rank}.`),Pe(t,0)}const Rl=d({reverse1d_:SC});function TC(e,t){const n=h(e,"x","reverse");return f(n.rank===2,()=>`Error in reverse2D: x must be rank 2 but got rank ${n.rank}.`),Pe(n,t)}const _l=d({reverse2d_:TC});function IC(e,t){const n=h(e,"x","reverse");return f(n.rank===3,()=>`Error in reverse3D: x must be rank 3 but got rank ${n.rank}.`),Pe(n,t)}const Nl=d({reverse3d_:IC});function AC(e,t){const n=h(e,"x","reverse");return f(n.rank===4,()=>`Error in reverse4D: x must be rank 4 but got rank ${n.rank}.`),Pe(n,t)}const Al=d({reverse4d_:AC});function $m(e,t,n=0){let r=[];if(typeof t=="number")f(e.shape[n]%t===0,()=>"Number of splits must evenly divide the axis."),r=new Array(t).fill(e.shape[n]/t);else{const i=t.reduce((a,s)=>(s===-1&&(a+=1),a),0);f(i<=1,()=>"There should be only one negative value in split array.");const o=t.indexOf(-1);if(o!==-1){const a=t.reduce((s,c)=>c>0?s+c:s);t[o]=e.shape[n]-a}f(e.shape[n]===t.reduce((a,s)=>a+s),()=>"The sum of sizes must match the size of the axis dimension."),r=t}return r}function NC(e,t,n=0){const r=h(e,"x","split"),i=(s,c)=>{const p=z(n,r.shape)[0],l=$m(r,t,p);return s.split(r,l,p)},o={x:r},a={numOrSizeSplits:t,axis:n};return g.runKernelFunc(i,o,null,cs,a)}const Ot=d({split_:NC});function _C(e,t){f(e.dtype==="float32",()=>`The dtype for rfft() must be real value but got ${e.dtype}`);let n=e.shape[e.shape.length-1];const r=e.size/n;let i;if(t!=null&&t<n){const y=e.shape.map(L=>0),w=e.shape.map(L=>L);w[e.shape.length-1]=t,i=Y(e,y,w),n=t}else if(t!=null&&t>n){const y=e.shape.map(w=>w);y[e.shape.length-1]=t-n,i=re([e,Ue(y)],e.shape.length-1),n=t}else i=e;const o=W(i),a=x(Ge(i,o),[r,n]),s=Yn(a),c=Math.floor(n/2)+1,p=kt(s),l=Vt(s),u=Ot(p,[c,n-c],p.shape.length-1),m=Ot(l,[c,n-c],l.shape.length-1),b=i.shape.slice();return b[i.shape.length-1]=c,x(Ge(u[0],m[0]),b)}const Wn=d({rfft_:_C});function RC(e){const t=h(e,"x","round"),n={x:t};return g.runKernelFunc(r=>r.round(t),n,null,Ko)}const Il=d({round_:RC});function CC(e){const t=h(e,"x","rsqrt"),n={x:t};return g.runKernelFunc((r,i)=>{const o=r.rsqrt(t);return i([t]),o},n,null,Jo)}const yi=d({rsqrt_:CC});function EC(e){const t=h(e,"x","selu"),n=(i,o)=>{const a=i.selu(t);return o([t]),a},r={x:t};return g.runKernelFunc(n,r,null,Zo)}const Tl=d({selu_:EC});function OC(e,t,n,r,i,o=[1,1],a="NHWC"){const s=h(e,"x","separableConv2d"),c=h(t,"depthwiseFilter","separableConv2d"),p=h(n,"pointwiseFilter","separableConv2d");let l=s,u=!1;if(s.rank===3&&(u=!0,l=x(s,[1,s.shape[0],s.shape[1],s.shape[2]])),a==="NCHW")throw new Error("separableConv2d currently does not support dataFormat NCHW; only NHWC is supported");f(l.rank===4,()=>`Error in separableConv2d: input must be rank 4, but got rank ${l.rank}.`),f(c.rank===4,()=>`Error in separableConv2d: depthwise filter must be rank 4, but got rank ${c.rank}.`),f(p.rank===4,()=>`Error in separableConv2d: pointwise filter must be rank 4, but got rank ${c.rank}.`),f(p.shape[0]===1,()=>`Error in separableConv2d: the first dimension of pointwise filter must be 1, but got ${p.shape[0]}.`),f(p.shape[1]===1,()=>`Error in separableConv2d: the second dimension of pointwise filter must be 1, but got ${p.shape[1]}.`);const m=c.shape[2],b=c.shape[3];f(p.shape[2]===m*b,()=>`Error in separableConv2d: the third dimension of pointwise filter must be ${m*b}, but got ${p.shape[2]}.`);const y=zn(l,c,r,i,a,o),w=1,L=It(y,p,w,"valid",a);return u?x(L,[L.shape[1],L.shape[2],L.shape[3]]):L}const Sl=d({separableConv2d_:OC});async function kC(e,t){const n=h(e,"x","setdiff1d"),r=h(t,"y","setdiff1d");f(n.dtype===r.dtype,()=>`x and y should have the same dtype, but got x (${n.dtype}) and y (${r.dtype}).`),f(n.rank===1,()=>`x should be 1D tensor, but got x (${n.shape}).`),f(r.rank===1,()=>`y should be 1D tensor, but got y (${r.shape}).`);const i=await n.data(),o=await r.data(),a=new Set(o);let s=0;for(let l=0;l<i.length;l++)a.has(i[l])||s++;const c=new yr([s],n.dtype),p=new yr([s],"int32");for(let l=0,u=0;l<i.length;l++)a.has(i[l])||(c.values[u]=i[l],p.values[u]=l,u++);return[c.toTensor(),p.toTensor()]}const vl=kC;function DC(e){const t=h(e,"x","sign"),n={x:t};return g.runKernelFunc(r=>r.sign(t),n,null,ns)}const Ll=d({sign_:DC});function FC(e){const t=h(e,"x","sin"),n={x:t};return g.runKernelFunc((r,i)=>{const o=r.sin(t);return i([t]),o},n,null,es)}const wi=d({sin_:FC});function MC(e){const t=h(e,"x","sinh"),n={x:t};return g.runKernelFunc((r,i)=>{const o=r.sinh(t);return i([t]),o},n,null,ts)}const bi=d({sinh_:MC});function UC(e,t,n){const r=h(e,"x","slice1d");return f(r.rank===1,()=>`slice1d expects a rank-1 tensor, but got a rank-${r.rank} tensor`),Y(r,[t],[n])}const xl=d({slice1d_:UC});function WC(e,t,n){const r=h(e,"x","slice2d");return f(r.rank===2,()=>`slice2d expects a rank-2 tensor, but got a rank-${r.rank} tensor`),Y(r,t,n)}const yl=d({slice2d_:WC});function qC(e,t,n){const r=h(e,"x","slice3d");return f(r.rank===3,()=>`slice3d expects a rank-3 tensor, but got a rank-${r.rank} tensor`),Y(r,t,n)}const wl=d({slice3d_:qC});function BC(e,t,n){const r=h(e,"x","slice4d");return f(r.rank===4,()=>`slice4d expects a rank-4 tensor, but got a rank-${r.rank} tensor`),Y(r,t,n)}const bl=d({slice4d_:BC});function PC(e,t=-1){const n=h(e,"logits","softmax","float32");if(t===-1&&(t=n.rank-1),t!==n.rank-1)throw Error(`Softmax along a non-last dimension is not yet supported. Logits was rank ${n.rank} and dim was ${t}`);const r={logits:n},i={dim:t};return g.runKernelFunc((o,a)=>{const s=o.softmax(n,t);return a([s]),s},r,null,ps,i)}const gl=d({softmax_:PC});function jC(e){const t=h(e,"x","sqrt"),n={x:t};return g.runKernelFunc((r,i)=>{const o=r.sqrt(t);return i([t]),o},n,null,os)}const le=d({sqrt_:jC});function GC(e,t){let n=h(e,"a","squaredDifference"),r=h(t,"b","squaredDifference");[n,r]=V(n,r),H(n.shape,r.shape);const i=(s,c)=>{const p=s.squaredDifference(n,r);return c([n,r]),p},o={a:n,b:r},a={};return g.runKernelFunc(i,o,null,ls,a)}const Mn=d({squaredDifference_:GC});function $C(e,t){const n=h(e,"x","squeeze");return x(n,hm(n.shape,t).newShape)}const Fn=d({squeeze_:$C});function HC(e,t=0){const n=rn(e,"tensors","stack");if(f(n.length>=1,()=>"Pass at least one tensor to tf.stack"),n.length===1)return Ke(n[0],t);const r=n[0].rank,i=n[0].shape,o=n[0].dtype;f(t<=r,()=>"Axis must be <= rank of the tensor"),n.forEach(s=>{$(i,s.shape,"All tensors passed to stack must have matching shapes"),f(o===s.dtype,()=>"All tensors passed to stack must have matching dtypes")});const a=n.map(s=>Ke(s,t));return re(a,t)}const dt=d({stack_:HC});function YC(e,t=0){const n=h(e,"x","step"),r={x:n},i={alpha:t};return g.runKernelFunc(o=>o.step(n,t),r,null,ys,i)}const vt=d({step_:YC});function zC(e,t,n,r,i=0,o=0,a=0,s=0,c=0){r==null&&(r=new Array(t.length));let p=h(e,"x","stridedSlice");const l=b=>{const y=Hs(a);if(y.length>1)throw new Error("Multiple ellipses in slice is not allowed.");if(a!==0&&s!==0)throw new Error("Using both ellipsisMask and newAxisMask is not yet supported.");if(a!==0&&c!==0)throw new Error("Using both ellipsisMask and shrinkAxisMask is not yet supported.");const w=p.rank-t.length,L=Hs(s),T=p.shape.slice();if(L.forEach(E=>{t[E]=0,n[E]=1,T.splice(E,0,1)}),p=x(p,T),y.length&&w>0){const E=y[0],B=w+1;t=Fm(i,E,B,t,p.shape),n=Mm(o,E,B,n,p.shape),r=Dm(r,E,B,p.shape)}else for(let E=0;E<p.rank;E++)t[E]=Wm(i,t,r,p.shape,E,a),n[E]=qm(o,n,r,p.shape,E,a),r[E]=Um(r,E,a);const A=Hs(c);A.forEach(E=>{n[E]=t[E]+1,r[E]=1});const _=km(t,n,r),N=_.filter((E,B)=>A.indexOf(B)===-1),C=r.every(E=>E===1);if(C)return x(Y(p,t,_),N);const F=b.stridedSlice(p,t,n,r);return x(F,N)},u={x:p},m={begin:t,end:n,strides:r,beginMask:i,endMask:o,ellipsisMask:a,newAxisMask:s,shrinkAxisMask:c};return g.runKernelFunc(l,u,null,Ch,m)}const fl=d({stridedSlice_:zC});function VC(e){const t=h(e,"x","tan"),n={x:t};return g.runKernelFunc((r,i)=>{const o=r.tan(t);return i([t]),o},n,null,hs)}const ml=d({tan_:VC});function Yt(e,t,n){if(lt(e),t!=null&&t.length!==2)throw new Error("tensor2d() requires shape to have two numbers");const r=Fe(e,n);if(r.length!==2&&r.length!==1)throw new Error("tensor2d() requires values to be number[][] or flat/TypedArray");if(r.length===1&&t==null)throw new Error("tensor2d() requires shape to be provided when `values` are a flat/TypedArray");return ze(e,t,r,n)}function fi(e,t,n){if(lt(e),t!=null&&t.length!==3)throw new Error("tensor3d() requires shape to have three numbers");const r=Fe(e,n);if(r.length!==3&&r.length!==1)throw new Error("tensor3d() requires values to be number[][][] or flat/TypedArray");if(r.length===1&&t==null)throw new Error("tensor3d() requires shape to be provided when `values` are a flat array");return ze(e,t,r,n)}function dl(e,t,n){if(lt(e),t!=null&&t.length!==4)throw new Error("tensor4d() requires shape to have four numbers");const r=Fe(e,n);if(r.length!==4&&r.length!==1)throw new Error("tensor4d() requires values to be number[][][][] or flat/TypedArray");if(r.length===1&&t==null)throw new Error("tensor4d() requires shape to be provided when `values` are a flat array");return ze(e,t,r,n)}function hl(e,t,n){if(lt(e),t!=null&&t.length!==5)throw new Error("tensor5d() requires shape to have five numbers");const r=Fe(e,n);if(r.length!==5&&r.length!==1)throw new Error("tensor5d() requires values to be number[][][][][] or flat/TypedArray");if(r.length===1&&t==null)throw new Error("tensor5d() requires shape to be provided when `values` are a flat array");return ze(e,t,r,n)}function ul(e,t,n){if(lt(e),t!=null&&t.length!==6)throw new Error("tensor6d() requires shape to have six numbers");const r=Fe(e,n);if(r.length!==6&&r.length!==1)throw new Error("tensor6d() requires values to be number[][][][][][] or flat/TypedArray");if(r.length===1&&t==null)throw new Error("tensor6d() requires shape to be provided when `values` are a flat array");return t=t||r,ze(e,t,r,n)}function KC(e,t=1,n=!0){const r=h(e,"x","topk");if(r.rank===0)throw new Error("topk() expects the input to be of rank 1 or higher");const i=r.shape[r.shape.length-1];if(t>i)throw new Error(`'k' passed to topk() must be <= the last dimension (${i}) but got ${t}`);const o={x:r},a={k:t,sorted:n},[s,c]=g.runKernelFunc(p=>p.topk(r,t,n),o,null,Eh,a);return{values:s,indices:c}}const ll=d({topk_:KC});function JC(e,t=0,n=1,r,i){if(r!=null&&r==="bool")throw new Error("Unsupported data type $ { dtype }");const o=new zs(t,n,r,!0,i),a=ke(e,r);for(let s=0;s<a.values.length;s++)a.values[s]=o.nextValue();return a.toTensor()}const pl=d({truncatedNormal_:JC});function XC(e,t,n){const r=h(e,"x","unsortedSegmentSum"),i=h(t,"segmentIds","unsortedSegmentSum","int32");f(J(n),()=>"numSegments must be of dtype int");const o={x:r,segmentIds:i},a={numSegments:n},s=(c,p)=>{const l=c.unsortedSegmentSum(r,i,n);return p([i]),l};return g.runKernelFunc(s,o,null,bs,a)}const mi=d({unsortedSegmentSum_:XC});function ZC(e,t=0){const n=h(e,"x","unstack");f(t>=-n.shape.length&&t<n.shape.length,()=>`Axis = ${t} is not in [-${n.shape.length}, ${n.shape.length})`),t<0&&(t+=n.shape.length);const r={value:n},i={axis:t},o=a=>a.unstack(n,t);return g.runKernelFunc(o,r,null,gs,i)}const Dn=d({unstack_:ZC});function cl(e,t=!0,n,r){return g.makeVariable(e,t,n,r)}function qp(e,t){const n=[];for(let o=0;o<t.length;o++)t[o]&&n.push(o);const r=ke(e,"int32"),i=ke([n.length,e.length],"int32");for(let o=0;o<n.length;o++){const a=r.indexToLoc(n[o]),s=o*e.length;i.values.set(a,s)}return i.toTensor()}async function QC(e){const t=h(e,"condition","whereAsync","bool"),n=await t.data(),r=qp(t.shape,n);return e!==t&&t.dispose(),r}const di=QC;async function eE(e,t,n){const r=h(e,"tensor","boolMask"),i=h(t,"mask","boolMask","bool"),o=n==null?0:n,a=i.rank,s=r.shape;f(a>0,()=>"mask cannot be scalar"),$(s.slice(o,o+a),i.shape,"mask's shape must match the first K dimensions of tensor's shape,");let c=1;for(let w=o;w<o+a;w++)c*=s[w];const p=s.slice(0,o).concat([c],s.slice(o+a)),l=x(r,p),u=x(i,[-1]),m=await di(u),b=Fn(m,[1]),y=Gn(l,b,o);return e!==r&&r.dispose(),t!==i&&i.dispose(),b.dispose(),l.dispose(),u.dispose(),m.dispose(),y}const Qf=eE;function tE(e,t){ge("strict variants of ops have been deprecated and will be removed in future");const n=h(e,"a","notEqualStrict"),r=h(t,"b","notEqualStrict");return $(n.shape,r.shape,"Error in notEqualStrict: "),qn(n,r)}function nE(e,t){ge("strict variants of ops have been deprecated and will be removed in future");const n=h(e,"a","lessStrict"),r=h(t,"b","lessStrict");return $(n.shape,r.shape,"Error in lessStrict: "),jn(n,r)}function rE(e,t){ge("strict variants of ops have been deprecated and will be removed in future");const n=h(e,"a","equalStrict"),r=h(t,"b","equalStrict");return $(n.shape,r.shape,"Error in equalStrict: "),Kt(n,r)}function iE(e,t){ge("strict variants of ops have been deprecated and will be removed in future");const n=h(e,"a","lessEqualStrict"),r=h(t,"b","lessEqualStrict");return $(n.shape,r.shape,"Error in lessEqualStrict: "),at(n,r)}function oE(e,t){ge("strict variants of ops have been deprecated and will be removed in future");const n=h(e,"a","greaterStrict"),r=h(t,"b","greaterStrict");return $(n.shape,r.shape,"Error in greaterStrict: "),Ve(n,r)}function sE(e,t){ge("strict variants of ops have been deprecated and will be removed in future");const n=h(e,"a","greaterEqualStrict"),r=h(t,"b","greaterEqualStrict");return $(n.shape,r.shape,"Error in greaterEqualStrict: "),ct(n,r)}const Yf=d({equalStrict_:rE}),zf=d({greaterEqualStrict_:sE}),Vf=d({greaterStrict_:oE}),Kf=d({lessEqualStrict_:iE}),Jf=d({lessStrict_:nE}),Xf=d({notEqualStrict_:tE});function aE(e,t){ge("strict variants of ops have been deprecated and will be removed in future");const n=h(e,"a","addStrict"),r=h(t,"b","addStrict");return $(n.shape,r.shape,"Error in addStrict: "),O(n,r)}function cE(e,t){ge("strict variants of ops have been deprecated and will be removed in future");const n=h(e,"a","subStrict"),r=h(t,"b","subStrict");return $(n.shape,r.shape,"Error in subStrict: "),D(n,r)}function pE(e,t){return ge("strict variants of ops have been deprecated and will be removed in future"),$(e.shape,t.shape,"Error in powStrict: "),je(e,t)}function lE(e,t){ge("strict variants of ops have been deprecated and will be removed in future");const n=h(e,"a","mul"),r=h(t,"b","mul");return $(n.shape,r.shape,"Error in multiplyStrict: "),v(n,r)}function uE(e,t){ge("strict variants of ops have been deprecated and will be removed in future");const n=h(e,"a","div"),r=h(t,"b","div");return $(n.shape,r.shape,"Error in divideStrict: "),M(n,r)}function hE(e,t){ge("strict variants of ops have been deprecated and will be removed in future");const n=h(e,"a","modStrict"),r=h(t,"b","modStrict");return $(n.shape,r.shape,"Error in modStrict: "),Si(n,r)}function dE(e,t){ge("strict variants of ops have been deprecated and will be removed in future");const n=h(e,"a","minimumStrict"),r=h(t,"b","minimumStrict");return $(n.shape,r.shape,"Error in minimumStrict: "),Bn(n,r)}function mE(e,t){ge("strict variants of ops have been deprecated and will be removed in future");const n=h(e,"a","maximumStrict"),r=h(t,"b","maximumStrict");return $(n.shape,r.shape,"Error in maximumStrict: "),Tt(n,r)}function fE(e,t){ge("strict variants of ops have been deprecated and will be removed in future");const n=h(e,"a","squaredDifferenceStrict"),r=h(t,"b","squaredDifferenceStrict");return $(n.shape,r.shape,"Error in squaredDifferenceStrict: "),Mn(n,r)}const eg=d({addStrict_:aE}),tg=d({divStrict_:uE}),ng=d({maximumStrict_:mE}),rg=d({minimumStrict_:dE}),ig=d({modStrict_:hE}),og=d({mulStrict_:lE}),sg=d({powStrict_:pE}),ag=d({squaredDifferenceStrict_:fE}),cg=d({subStrict_:cE});function gE(e,t="euclidean",n=null,r=!1){e=h(e,"x","norm");const i=jy(e,t,n);let o=i.shape;if(r){const a=z(n,e.shape);o=fe(i.shape,a)}return x(i,o)}function jy(e,t,n=null){if(e.rank===0)return _e(e);if(e.rank!==1&&n===null)return jy(x(e,[-1]),t,n);if(e.rank===1||typeof n=="number"||Array.isArray(n)&&n.length===1){if(t===1)return U(_e(e),n);if(t===Infinity)return st(_e(e),n);if(t===-Infinity)return sn(_e(e),n);if(t==="euclidean"||t===2)return le(U(je(_e(e),k(2,"int32")),n));throw new Error(`Error in norm: invalid ord value: ${t}`)}if(Array.isArray(n)&&n.length===2){if(t===1)return st(U(_e(e),n[0]),n[1]-1);if(t===Infinity)return st(U(_e(e),n[1]),n[0]);if(t===-Infinity)return sn(U(_e(e),n[1]),n[0]);if(t==="fro"||t==="euclidean")return le(U(G(e),n));throw new Error(`Error in norm: invalid ord value: ${t}`)}throw new Error(`Error in norm: invalid axis: ${n}`)}const Tr=d({norm_:gE});function bE(e,t,n,r,i=!0){const o=h(e,"v","movingAverage"),a=h(t,"x","movingAverage"),s=h(n,"decay","movingAverage");Tm(o,a),f(Le(o.shape,a.shape),()=>"Shape mismatch in v and x");const c=k(1),p=D(c,s);let l=v(D(a,o),p);if(i){f(r!=null,()=>"When using zeroDebias: true, step is required.");const u=h(r,"step","movingAverage");l=M(l,D(c,je(s,u)))}return O(o,l)}const Pf=d({movingAverage_:bE});const Df={};Oe(Df,{calculateShapes:()=>Gy,validateInput:()=>Bp,validateUpdateShape:()=>Hm});function Hm(e,t,n){const r=t.rank>1?t.shape[t.rank-1]:1,i=t.rank>1?t.rank-1:1,o=`Must have updates.shape = indices.shape[:batchDim] + shape[sliceDim:], got updates.shape: ${n.shape}, indices.shape: ${t.shape}, shape: ${e}, sliceDim: ${r}, and batchDim: ${i}.`;if(n.rank<i)throw new Error(o+` update.rank < ${i}. `);if(e.length<r+(n.rank-i))throw new Error(o+` Output shape length < ${r+(n.rank-i)}`);if(n.rank!==i+e.length-r)throw new Error(o+` update.rank != ${i+e.length-r}`);for(let a=0;a<i;++a)if(n.shape[a]!==t.shape[a])throw new Error(o+` updates.shape[${a}] (${n.shape[a]}) != indices.shape[${a}] (${t.shape[a]}).`);for(let a=0;a<n.rank-i;++a)if(n.shape[a+i]!==e[a+r])throw new Error(o+` updates.shape[${a+i}] (${n.shape[a+i]}) != shape[${a+i}] (${e[a+i]})`)}function Bp(e,t,n){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(n.length<1)throw new Error(`Output rank must be greater or equal to 1, but got shape: ${n}`);if(n.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}`)}Hm(n,t,e)}function Gy(e,t,n){const r=t.shape.length,i=r>1?t.shape[r-1]:1,o=n.length;let a=1;for(let u=i;u<o;++u)a*=n[u];const s=i<1?1:i,c=Z(t.shape)/s,p=[...Bt(n.slice(0,i)),1],l=Z(n);return{sliceRank:i,numUpdates:c,sliceSize:a,strides:p,outputSize:l}}function wE(e,t,n){const r=h(e,"indices","scatterND","int32"),i=h(t,"updates","scatterND");Bp(i,r,n);const o=c=>c.scatterND(r,i,n),a={indices:r,updates:i},s={shape:n};return g.runKernelFunc(o,a,null,Nh,s)}const Ff=d({scatterND_:wE});function $y(e,t,n,r){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}.`);const i=e.rank>0?e.shape[0]:1,o=e.rank>1?e.shape[1]:1;if(n.length!==o)throw new Error(`outputShape has incorrect number of elements:, ${n.length}, should be: ${o}.`);const a=t.size;if(!(t.rank===0||t.rank===1&&a===i))throw new Error(`sparseValues has incorrect shape ${t.shape}, should be [] or [${i}]`);if(t.dtype!==r.dtype)throw new Error("sparseValues.dtype must match defaultValues.dtype")}function yE(e,t,n,r=0){const i=h(e,"sparseIndices","sparseToDense","int32"),o=h(t,"sparseValues","sparseToDense"),a=h(r,"defaultValue","sparseToDense",o.dtype);$y(i,o,n,a);const s={sparseIndices:i,sparseValues:o,defaultValue:a},c={outputShape:n};return g.runKernelFunc(p=>p.sparseToDense(i,o,n,a),s,null,Rh,c)}const Of=d({sparseToDense_:yE});function xE(e,t){const n=h(t,"indices","gatherND","int32"),r=h(e,"x","gatherND"),i=a=>a.gatherND(r,n),o={params:r,indices:n};return g.runKernelFunc(i,o,null,ih)}const $f=d({gatherND_:xE});function Hy(e,t){if(t==null)return e.shape.slice();if(Le(e.shape,t))return t;if(e.shape.length===t.length){const n=[];for(let r=0;r<e.shape.length;r++)t[r]==null&&e.shape[r]!=null?n.push(e.shape[r]):n.push(t[r]);return n}return t}function LE(e,t,n,r){const i=h(e,"x","dropout");if(f(i.dtype==="float32",()=>`x has to be a floating point tensor since it's going to be scaled, but got a ${i.dtype} tensor instead.`),f(t>=0&&t<1,()=>`rate must be a float in the range [0, 1), but got ${t}.`),t===0)return e instanceof de?i.clone():i;const o=Hy(i,n),a=1-t,s=M($n(O(Li(o,0,1,"float32",r),a)),a);return v(i,s)}const Hf=d({dropout_:LE});function Qs(e){return Math.floor(Math.pow(2,Math.ceil(Math.log(e)/Math.log(2))))}function Lr(e,t,n){const r=1-e%2,i=new Float32Array(e);for(let o=0;o<e;++o){const a=2*Math.PI*o/(e+r-1);i[o]=t-n*Math.cos(a)}return ht(i,"float32")}async function vE(e,t,n=1){const r=h(e,"predictions","inTopK"),i=h(t,"targets","inTopK");f(r.rank>1,()=>`inTopK() expects the predictions to be of rank 2 or higher, but got ${r.rank}`),f(r.rank-1===i.rank,()=>`predictions rank should be 1 larger than targets rank, but got predictions rank ${r.rank} and targets rank ${i.rank}`),$(r.shape.slice(0,r.shape.length-1),i.shape,"predictions's shape should be align with the targets' shape, except the last dimension.");const o=r.shape[r.shape.length-1];f(n>0&&n<=o,()=>`'k' passed to inTopK() must be > 0 && <= the predictions last dimension (${o}), but got ${n}`);const a=await r.data(),s=await i.data(),[c,p]=[a.length/o,o],l=ni("bool",c);for(let u=0;u<c;u++){const m=u*p,b=a.subarray(m,m+p),y=[];for(let w=0;w<b.length;w++)y.push({value:b[w],index:w});y.sort((w,L)=>L.value-w.value),l[u]=0;for(let w=0;w<n;w++)if(y[w].index===s[u]){l[u]=1;break}}return e!==r&&r.dispose(),t!==i&&i.dispose(),Qe(l,i.shape,"bool")}const jf=vE;function SE(e,t,n,r,i,o="NHWC",a){let s=e;e.rank===3&&(s=x(e,[1,e.shape[0],e.shape[1],e.shape[2]]));let c=t;c.rank===3&&(c=x(t,[1,t.shape[0],t.shape[1],t.shape[2]])),f(s.rank===4,()=>`Error in conv2dDerFilter: input must be rank 4, but got shape ${s.shape}.`),f(c.rank===4,()=>`Error in conv2dDerFilter: dy must be rank 4, but got shape ${c.shape}.`),f(n.length===4,()=>`Error in conv2dDerFilter: filterShape must be length 4, but got ${n}.`);const p=o==="NHWC"?s.shape[3]:s.shape[1],l=o==="NHWC"?c.shape[3]:c.shape[1];f(p===n[2],()=>`Error in conv2dDerFilter: depth of input ${p}) must match input depth in filter (${n[2]}.`),f(l===n[3],()=>`Error in conv2dDerFilter: depth of dy (${l}) must match output depth for filter (${n[3]}).`),a!=null&&f(J(i),()=>`Error in conv2dDerFilter: pad must be an integer when using, dimRoundingMode ${a} but got pad ${i}.`);const u=y=>{const w=1,L=Sn(o),T=We(s.shape,n,r,w,i,a,!1,L);return y.conv2dDerFilter(s,c,T)},m={x:s,dy:c},b={strides:r,pad:i,dataFormat:o,dimRoundingMode:a};return g.runKernelFunc(u,m,null,Gu,b)}const ci=d({conv2DBackpropFilter_:SE});function ur(e,t,n){if(n==null||n==="linear")return e;if(n==="relu")return v(e,vt(t));throw new Error(`Cannot compute gradient for fused activation ${n}.`)}function hr(e,t){let n=t;const r=ie(e.shape,t.shape);return r.length>0&&(n=U(n,r)),x(n,e.shape)}function dr(e,t,n){if(t==="linear")return e;if(t==="relu")return zt(e);if(t==="elu")return Ei(e);if(t==="relu6")return xi(e);if(t==="prelu")return vi(e,n);throw new Error(`Unknown fused activation ${t}.`)}const mr=(e,t)=>{const n=e>0;return!n||t==="linear"};function TE({x:e,filter:t,strides:n,pad:r,dataFormat:i="NHWC",dilations:o=[1,1],dimRoundingMode:a,bias:s,activation:c="linear",preluActivationWeights:p}){if(c=c||"linear",mr(g.state.gradientDepth,c)===!1){let C=It(e,t,n,r,i,o,a);return s!=null&&(C=O(C,s)),dr(C,c,p)}const l=h(e,"x","conv2d"),u=h(t,"filter","conv2d");let m=l,b=!1;l.rank===3&&(b=!0,m=x(l,[1,l.shape[0],l.shape[1],l.shape[2]])),f(m.rank===4,()=>`Error in fused conv2d: input must be rank 4, but got rank ${m.rank}.`),f(u.rank===4,()=>`Error in fused conv2d: filter must be rank 4, but got rank ${u.rank}.`),a!=null&&f(J(r),()=>`Error in fused conv2d: pad must be an integer when using, dimRoundingMode ${a} but got pad ${r}.`),f(m.shape[3]===u.shape[2],()=>`Error in conv2d: depth of input (${m.shape[3]}) must match input depth for filter ${u.shape[2]}.`),f(ae(n,o),()=>`Error in conv2D: Either strides or dilations must be 1. Got strides ${n} and dilations '${o}'`),f(i==="NHWC",()=>`Error in conv2d: got dataFormat of ${i} but only NHWC is currently supported.`);const y=We(m.shape,u.shape,n,o,r,a);let w;s!=null&&(w=h(s,"bias","fused conv2d"),[w]=V(w,l),H(y.outShape,w.shape));let L;p!=null&&(L=h(p,"prelu weights","fused conv2d"));const T=(C,F)=>{const[E,B,q,P]=F,te=ur(C,q,c);f(ut(o),()=>`Error in gradient of fused conv2D: dilation rates greater than 1 are not yet supported in gradients. Got dilations '${o}'`);const ne=ai(B.shape,te,E,n,r),pe=ci(B,te,E.shape,n,r),Se=[ne,pe];if(P!=null){const qe=hr(P,te);Se.push(qe)}return Se},A=C=>{const F=C.fusedConv2d({input:m,filter:u,convInfo:y,bias:w,activation:c,preluActivationWeights:L});return F},_={x:m,filter:u,bias:w,preluActivationWeights:L},N={strides:n,pad:r,dataFormat:i,dilations:o,dimRoundingMode:a,activation:c};if(s==null){const C=$e((F,E,B)=>{let q=g.runKernelFunc(A,_,null,oa,N);return B([E,F,q]),b&&(q=x(q,[q.shape[1],q.shape[2],q.shape[3]])),{value:q,gradFunc:T}});return C(m,u)}else{const C=$e((F,E,B,q)=>{let P=g.runKernelFunc(A,_,null,oa,N);return q([E,F,P,B]),b&&(P=x(P,[P.shape[1],P.shape[2],P.shape[3]])),{value:P,gradFunc:T}});return C(m,u,w)}}const IE=d({fusedConv2d_:TE});function AE(e,t,n,r){let i=e;e.rank===3&&(i=x(e,[1,e.shape[0],e.shape[1],e.shape[2]]));let o=t;o.rank===3&&(o=x(t,[1,t.shape[0],t.shape[1],t.shape[2]]));const a=c=>c.depthwiseConv2DDerFilter(i,o,r),s={x:i,dy:o};return g.runKernelFunc(a,s,null,Vu)}const Pp=d({depthwiseConv2dNativeBackpropFilter_:AE});function NE(e,t,n,r){let i=t,o=!1;t.rank===3&&(o=!0,i=x(t,[1,t.shape[0],t.shape[1],t.shape[2]]));const a=p=>p.depthwiseConv2DDerInput(i,n,r),s={dy:i},c=g.runKernelFunc(a,s,null,Ku);return o?x(c,[c.shape[1],c.shape[2],c.shape[3]]):c}const jp=d({depthwiseConv2dNativeBackpropInput_:NE});function _E({x:e,filter:t,strides:n,pad:r,dataFormat:i="NHWC",dilations:o=[1,1],dimRoundingMode:a,bias:s,activation:c="linear",preluActivationWeights:p}){if(mr(g.state.gradientDepth,c)===!1){let C=zn(e,t,n,r,i,o,a);return s!=null&&(C=O(C,s)),dr(C,c,p)}const l=h(e,"x","depthwiseConv2d"),u=h(t,"filter","depthwiseConv2d");let m=l,b=!1;l.rank===3&&(b=!0,m=x(l,[1,l.shape[0],l.shape[1],l.shape[2]])),f(m.rank===4,()=>`Error in fused depthwiseConv2d: input must be rank 4, but got rank ${m.rank}.`),f(u.rank===4,()=>`Error in fused depthwiseConv2d: filter must be rank 4, but got rank ${u.rank}.`),f(m.shape[3]===u.shape[2],()=>`Error in fused depthwiseConv2d: number of input channels (${m.shape[3]}) must match the inChannels dimension in filter ${u.shape[2]}.`),o==null&&(o=[1,1]),f(ae(n,o),()=>`Error in fused depthwiseConv2d: Either strides or dilations must be 1. Got strides ${n} and dilations '${o}'`),a!=null&&f(J(r),()=>`Error in fused depthwiseConv2d: pad must be an integer when using dimRoundingMode ${a} but got pad ${r}.`);const y=We(m.shape,u.shape,n,o,r,a,!0);let w;s!=null&&(w=h(s,"bias","fused conv2d"),[w]=V(w,l),H(y.outShape,w.shape));let L;p!=null&&(L=h(p,"prelu weights","fused depthwiseConv2d"));const T=(C,F)=>{f(ut(o),()=>`Error in gradient of fused depthwiseConv2d: dilation rates greater than 1 are not yet supported. Got dilations '${o}'`);const[E,B,q,P]=F,te=ur(C,q,c),ne=jp(B.shape,te,E,y),pe=Pp(B,te,E.shape,y);if(P!=null){const Se=hr(w,te);return[ne,pe,Se]}return[ne,pe]},A=C=>{const F=C.fusedDepthwiseConv2D({input:m,filter:u,convInfo:y,bias:w,activation:c,preluActivationWeights:L});return F},_={x:m,filter:u,bias:w,preluActivationWeights:L},N={strides:n,pad:r,dataFormat:i,dilations:o,dimRoundingMode:a,activation:c};if(s==null){const C=$e((F,E,B)=>{let q=g.runKernelFunc(A,_,null,sa,N);return B([E,F,q]),b&&(q=x(q,[q.shape[1],q.shape[2],q.shape[3]])),{value:q,gradFunc:T}});return C(m,u)}else{const C=$e((F,E,B,q)=>{let P=g.runKernelFunc(A,_,null,sa,N);return q([E,F,P,B]),b&&(P=x(P,[P.shape[1],P.shape[2],P.shape[3]])),{value:P,gradFunc:T}});return C(m,u,w)}}const RE=d({fusedDepthwiseConv2d_:_E});function CE({a:e,b:t,transposeA:n=!1,transposeB:r=!1,bias:i,activation:o="linear",preluActivationWeights:a}){if(mr(g.state.gradientDepth,o)===!1){let P=K(e,t,n,r);return i!=null&&(P=O(P,i)),dr(P,o,a)}let s=h(e,"a","fused matMul"),c=h(t,"b","fused matMul");[s,c]=V(s,c);const p=n?s.shape[s.rank-2]:s.shape[s.rank-1],l=r?c.shape[c.rank-1]:c.shape[c.rank-2],u=n?s.shape[s.rank-1]:s.shape[s.rank-2],m=r?c.shape[c.rank-2]:c.shape[c.rank-1],b=s.shape.slice(0,-2),y=c.shape.slice(0,-2),w=Z(b),L=Z(y);f(s.rank>=2&&c.rank>=2&&s.rank===c.rank,()=>`Error in fused matMul: inputs must have the same rank of at least 2, got ranks ${s.rank} and ${c.rank}.`),f(Le(b,y),()=>`Error in fused matMul: outer dimensions (${b}) and (${y}) of Tensors with shapes ${s.shape} and ${c.shape} must match.`),f(p===l,()=>`Error in fused matMul: inner shapes (${p}) and (${l}) of Tensors with shapes ${s.shape} and ${c.shape} and transposeA=${n} and transposeB=${r} must match.`);const T=s.shape.slice(0,-2).concat([u,m]),A=n?x(s,[w,p,u]):x(s,[w,u,p]),_=r?x(c,[L,m,l]):x(c,[L,l,m]);let N;i!=null&&(N=h(i,"bias","fused matMul"),[N]=V(N,s),H(T,N.shape));let C;a!=null&&(C=h(a,"prelu weights","fused matMul"));const F=(P,te)=>{const[ne,pe,Se,qe]=te,Gt=ur(x(P,Se.shape),Se,o);let gr,br;if(!n&&!r?(gr=K(Gt,pe,!1,!0),br=K(ne,Gt,!0,!1)):!n&&r?(gr=K(Gt,pe,!1,!1),br=K(Gt,ne,!0,!1)):n&&!r?(gr=K(pe,Gt,!1,!0),br=K(ne,Gt,!1,!1)):(gr=K(pe,Gt,!0,!0),br=K(Gt,ne,!0,!0)),i!=null){const vv=hr(qe,Gt);return[gr,br,vv]}else return[gr,br]},E=P=>{const te=P.fusedBatchMatMul({a:A,b:_,transposeA:n,transposeB:r,bias:N,activation:o,preluActivationWeights:C});return te},B={a:A,b:_,bias:N,preluActivationWeights:C},q={transposeA:n,transposeB:r,activation:o};if(i==null){const P=$e((te,ne,pe)=>{const Se=g.runKernelFunc(E,B,null,ia,q);return pe([te,ne,Se]),{value:x(Se,T),gradFunc:F}});return P(A,_)}else{const P=$e((te,ne,pe,Se)=>{const qe=g.runKernelFunc(E,B,null,ia,q);return Se([te,ne,qe,pe]),{value:x(qe,T),gradFunc:F}});return P(A,_,N)}}const EE=d({fusedMatMul_:CE});const nu={};Oe(nu,{conv2d:()=>IE,depthwiseConv2d:()=>RE,matMul:()=>EE});function OE(e){return Lr(e,.54,.46)}const Yy=d({hammingWindow_:OE});function kE(e){return Lr(e,.5,.5)}const Gp=d({hannWindow_:kE});function DE(e,t,n,r=!1,i=0){let o=0;const a=[];for(;o+t<=e.size;)a.push(Y(e,o,t)),o+=n;if(r)for(;o<e.size;){const s=o+t-e.size,c=re([Y(e,o,t-s),Hn([s],i)]);a.push(c),o+=n}return a.length===0?Yt([],[0,t]):x(re(a),[a.length,t])}const $p=d({frame_:DE});function FE(e,t,n,r,i=Gp){r==null&&(r=Qs(t));const o=$p(e,t,n),a=v(o,i(t)),s=[];for(let c=0;c<o.shape[0];c++)s.push(Wn(Y(a,[c,0],[1,t]),r));return re(s)}const zy=d({stft_:FE});function ME(e,t,n,r,i,o){const a=h(e,"image","cropAndResize"),s=h(t,"boxes","cropAndResize","float32"),c=h(n,"boxInd","cropAndResize","int32");i=i||"bilinear",o=o||0;const p=s.shape[0];f(a.rank===4,()=>`Error in cropAndResize: image must be rank 4,but got rank ${a.rank}.`),f(s.rank===2&&s.shape[1]===4,()=>`Error in cropAndResize: boxes must be have size [${p},4] but had shape ${s.shape}.`),f(c.rank===1&&c.shape[0]===p,()=>`Error in cropAndResize: boxInd must be have size [${p}] but had shape ${s.shape}.`),f(r.length===2,()=>`Error in cropAndResize: cropSize must be of length 2, but got length ${r.length}.`),f(r[0]>=1&&r[1]>=1,()=>`cropSize must be atleast [1,1], but was ${r}`),f(i==="bilinear"||i==="nearest",()=>`method must be bilinear or nearest, but was ${i}`);const l=y=>y.cropAndResize(a,s,c,r,i,o),u={image:a,boxes:s,boxInd:c},m={method:i,extrapolationValue:o,cropSize:r},b=g.runKernelFunc(l,u,null,Yu,m);return b}const Vy=d({cropAndResize_:ME});function UE(e){const t=h(e,"image","flipLeftRight","float32");f(t.rank===4,()=>`Error in flipLeftRight: image must be rank 4,but got rank ${t.rank}.`);const n={image:t},r=g.runKernel(rh,n,{});return r}const Ky=d({flipLeftRight_:UE});function WE(e,t,n=0,r=.5){const i=h(e,"image","rotateWithOffset","float32");f(i.rank===4,()=>`Error in rotateWithOffset: image must be rank 4,but got rank ${i.rank}.`);const o={image:i},a={radians:t,fillValue:n,center:r},s=g.runKernel(Oh,o,a);return s}const Jy=d({rotateWithOffset_:WE});function Lt(e,t,n,r,i,o){r==null&&(r=.5),i==null&&(i=Number.NEGATIVE_INFINITY),o==null&&(o=0);const a=e.shape[0];return n=Math.min(n,a),f(0<=r&&r<=1,()=>`iouThreshold must be in [0, 1], but was '${r}'`),f(e.rank===2,()=>`boxes must be a 2D tensor, but was of rank '${e.rank}'`),f(e.shape[1]===4,()=>`boxes must have 4 columns, but 2nd dimension was ${e.shape[1]}`),f(t.rank===1,()=>"scores must be a 1D tensor"),f(t.shape[0]===a,()=>`scores has incompatible shape with boxes. Expected ${a}, but was ${t.shape[0]}`),f(0<=o&&o<=1,()=>`softNmsSigma must be in [0, 1], but was '${o}'`),{maxOutputSize:n,iouThreshold:r,scoreThreshold:i,softNmsSigma:o}}function qE(e,t,n,r=.5,i=Number.NEGATIVE_INFINITY){const o=h(e,"boxes","nonMaxSuppression"),a=h(t,"scores","nonMaxSuppression"),s=Lt(o,a,n,r,i);n=s.maxOutputSize,r=s.iouThreshold,i=s.scoreThreshold;const c={maxOutputSize:n,iouThreshold:r,scoreThreshold:i};return g.runKernelFunc(p=>p.nonMaxSuppression(o,a,n,r,i),{boxes:o,scores:a},null,yh,c)}const Xy=d({nonMaxSuppression_:qE});function Zy(e,t,n){const r=BE(e,t,n),i=r<0?-(r+1):r;e.splice(i,0,t)}function BE(e,t,n){return jE(e,t,n||PE)}function PE(e,t){return e>t?1:e<t?-1:0}function jE(e,t,n){let r=0,i=e.length,o=0,a=!1;for(;r<i;){o=r+(i-r>>>1);const s=n(t,e[o]);s>0?r=o+1:(i=o,a=!s)}return a?r:-r-1}function Hp(e,t,n,r,i){return Ym(e,t,n,r,i,0).selectedIndices}function Yp(e,t,n,r,i,o){return Ym(e,t,n,r,i,0,!1,o,!0)}function zp(e,t,n,r,i,o){return Ym(e,t,n,r,i,o,!0)}function Ym(e,t,n,r,i,o,a=!1,s=!1,c=!1){const p=[];for(let L=0;L<t.length;L++)t[L]>i&&p.push({score:t[L],boxIndex:L,suppressBeginIndex:0});p.sort(Qy);const l=o>0?-.5/o:0,u=[],m=[];for(;u.length<n&&p.length>0;){const L=p.pop(),{score:T,boxIndex:A,suppressBeginIndex:_}=L;if(T<i)break;let N=!1;for(let C=u.length-1;C>=_;--C){const F=GE(e,A,u[C]);if(F>=r){N=!0;break}if(L.score=L.score*$E(r,l,F),L.score<=i)break}L.suppressBeginIndex=u.length,N||(L.score===T?(u.push(A),m.push(L.score)):L.score>i&&Zy(p,L,Qy))}const b=u.length,y=n-b;s&&y>0&&(u.push(...new Array(y).fill(0)),m.push(...new Array(y).fill(0)));const w={selectedIndices:ht(u,"int32")};return a&&(w.selectedScores=ht(m,"float32")),c&&(w.validOutputs=k(b,"int32")),w}function GE(e,t,n){const r=e.subarray(t*4,t*4+4),i=e.subarray(n*4,n*4+4),o=Math.min(r[0],r[2]),a=Math.min(r[1],r[3]),s=Math.max(r[0],r[2]),c=Math.max(r[1],r[3]),p=Math.min(i[0],i[2]),l=Math.min(i[1],i[3]),u=Math.max(i[0],i[2]),m=Math.max(i[1],i[3]),b=(s-o)*(c-a),y=(u-p)*(m-l);if(b<=0||y<=0)return 0;const w=Math.max(o,p),L=Math.max(a,l),T=Math.min(s,u),A=Math.min(c,m),_=Math.max(T-w,0)*Math.max(A-L,0);return _/(b+y-_)}function $E(e,t,n){const r=Math.exp(t*n*n);return n<=e?r:0}function Qy(e,t){return e.score-t.score||e.score===t.score&&t.boxIndex-e.boxIndex}async function HE(e,t,n,r=.5,i=Number.NEGATIVE_INFINITY){const o=h(e,"boxes","nonMaxSuppressionAsync"),a=h(t,"scores","nonMaxSuppressionAsync"),s=Lt(o,a,n,r,i);n=s.maxOutputSize,r=s.iouThreshold,i=s.scoreThreshold;const c=await Promise.all([o.data(),a.data()]),p=c[0],l=c[1],u=Hp(p,l,n,r,i);return o!==e&&o.dispose(),a!==t&&a.dispose(),u}const ex=HE;function YE(e,t,n,r=.5,i=Number.NEGATIVE_INFINITY,o=0){const a=h(e,"boxes","nonMaxSuppression"),s=h(t,"scores","nonMaxSuppression"),c=Lt(a,s,n,r,i,o);n=c.maxOutputSize,r=c.iouThreshold,i=c.scoreThreshold,o=c.softNmsSigma;const p={boxes:a,scores:s},l={maxOutputSize:n,iouThreshold:r,scoreThreshold:i,softNmsSigma:o},u=g.runKernel(Lh,p,l);return{selectedIndices:u[0],selectedScores:u[1]}}const tx=d({nonMaxSuppressionWithScore_:YE});async function zE(e,t,n,r=.5,i=Number.NEGATIVE_INFINITY,o=0){const a=h(e,"boxes","nonMaxSuppressionAsync"),s=h(t,"scores","nonMaxSuppressionAsync"),c=Lt(a,s,n,r,i,o);n=c.maxOutputSize,r=c.iouThreshold,i=c.scoreThreshold,o=c.softNmsSigma;const p=await Promise.all([a.data(),s.data()]),l=p[0],u=p[1],m=zp(l,u,n,r,i,o);return a!==e&&a.dispose(),s!==t&&s.dispose(),m}const nx=zE;function VE(e,t,n,r=.5,i=Number.NEGATIVE_INFINITY,o=!1){const a=h(e,"boxes","nonMaxSuppression"),s=h(t,"scores","nonMaxSuppression"),c=Lt(a,s,n,r,i,null),p=c.maxOutputSize,l=c.iouThreshold,u=c.scoreThreshold,m={boxes:a,scores:s},b={maxOutputSize:p,iouThreshold:l,scoreThreshold:u,padToMaxOutputSize:o},y=g.runKernel(xh,m,b);return{selectedIndices:y[0],validOutputs:y[1]}}const rx=d({nonMaxSuppressionPadded_:VE});async function KE(e,t,n,r=.5,i=Number.NEGATIVE_INFINITY,o=!1){const a=h(e,"boxes","nonMaxSuppressionAsync"),s=h(t,"scores","nonMaxSuppressionAsync"),c=Lt(a,s,n,r,i,null),p=c.maxOutputSize,l=c.iouThreshold,u=c.scoreThreshold,[m,b]=await Promise.all([a.data(),s.data()]),y=Yp(m,b,p,l,u,o);return a!==e&&a.dispose(),s!==t&&s.dispose(),y}const ix=KE;function JE(e,t,n=!1){const r=h(e,"images","resizeBilinear");f(r.rank===3||r.rank===4,()=>`Error in resizeBilinear: x must be rank 3 or 4, but got rank ${r.rank}.`),f(t.length===2,()=>`Error in resizeBilinear: new shape must 2D, but got shape ${t}.`);let i=r,o=!1;r.rank===3&&(o=!0,i=x(r,[1,r.shape[0],r.shape[1],r.shape[2]]));const[a,s]=t,c=(m,b)=>(b([i]),m.resizeBilinear(i,a,s,n)),p={images:i},l={alignCorners:n,size:t},u=g.runKernelFunc(c,p,null,Yo,l);return o?x(u,[u.shape[1],u.shape[2],u.shape[3]]):u}const ox=d({resizeBilinear_:JE});function XE(e,t,n=!1){const r=h(e,"images","resizeNearestNeighbor");f(r.rank===3||r.rank===4,()=>`Error in resizeNearestNeighbor: x must be rank 3 or 4, but got rank ${r.rank}.`),f(t.length===2,()=>`Error in resizeNearestNeighbor: new shape must 2D, but got shape ${t}.`),f(r.dtype==="float32"||r.dtype==="int32",()=>"`images` must have `int32` or `float32` as dtype");let i=r,o=!1;r.rank===3&&(o=!0,i=x(r,[1,r.shape[0],r.shape[1],r.shape[2]]));const[a,s]=t,c={images:i},p={alignCorners:n,size:t},l=(m,b)=>(b([i]),m.resizeNearestNeighbor(i,a,s,n)),u=g.runKernelFunc(l,c,null,Ho,p);return o?x(u,[u.shape[1],u.shape[2],u.shape[3]]):u}const sx=d({resizeNearestNeighbor_:XE});function ZE(e,t,n){f(t%1===0,()=>`bandPart(): numLower must be an integer, got ${t}.`),f(n%1===0,()=>`bandPart(): numUpper must be an integer, got ${n}.`);const r=h(e,"a","bandPart");f(r.rank>=2,()=>`bandPart(): Rank must be at least 2, got ${r.rank}.`);const i=r.shape,[o,a]=r.shape.slice(-2);if(!(t<=o))throw new Error(`bandPart(): numLower (${t}) must not be greater than the number of rows (${o}).`);if(!(n<=a))throw new Error(`bandPart(): numUpper (${n}) must not be greater than the number of columns (${a}).`);t<0&&(t=o),n<0&&(n=a);const s=x(vr(0,o,1,"int32"),[-1,1]),c=vr(0,a,1,"int32"),p=D(s,c),l=ft(at(p,k(+t,"int32")),ct(p,k(-n,"int32"))),u=Ue([o,a],r.dtype);return x(dt(Dn(x(r,[-1,o,a])).map(m=>Ae(l,m,u))),i)}const ax=d({bandPart_:ZE});function QE(e){let t;if(Array.isArray(e)){t=!1,f(e!=null&&e.length>0,()=>"Gram-Schmidt process: input must not be null, undefined, or empty");const i=e[0].shape[0];for(let o=1;o<e.length;++o)f(e[o].shape[0]===i,()=>`Gram-Schmidt: Non-unique lengths found in the input vectors: (${e[o].shape[0]} vs. ${i})`)}else t=!0,e=Ot(e,e.shape[0],0).map(i=>Fn(i,[0]));f(e.length<=e[0].shape[0],()=>`Gram-Schmidt: Number of vectors (${e.length}) exceeds number of dimensions (${e[0].shape[0]}).`);const n=[],r=e;for(let i=0;i<e.length;++i)n.push(g.tidy(()=>{let o=r[i];if(i>0)for(let a=0;a<i;++a){const s=v(U(v(n[a],o)),n[a]);o=D(o,s)}return M(o,Tr(o,"euclidean"))}));return t?dt(n,0):n}const cx=d({gramSchmidt_:QE});function eO(e,t=!1){if(f(e.rank>=2,()=>`qr() requires input tensor to have a rank >= 2, but got rank ${e.rank}`),e.rank===2)return px(e,t);{const n=e.shape.slice(0,e.shape.length-2).reduce((c,p)=>c*p),r=Dn(x(e,[n,e.shape[e.shape.length-2],e.shape[e.shape.length-1]]),0),i=[],o=[];r.forEach(c=>{const[p,l]=px(c,t);i.push(p),o.push(l)});const a=x(dt(i,0),e.shape),s=x(dt(o,0),e.shape);return[a,s]}}function px(e,t=!1){return g.tidy(()=>{f(e.shape.length===2,()=>`qr2d() requires a 2D Tensor, but got a ${e.shape.length}D Tensor.`);const n=e.shape[0],r=e.shape[1];let i=Ci(n),o=At(e);const a=Yt([[1]],[1,1]);let s=At(a);const c=n>=r?r:n;for(let p=0;p<c;++p){const l=o,u=s,m=i;[s,o,i]=g.tidy(()=>{const b=Y(o,[p,p],[n-p,1]),y=Tr(b),w=Y(o,[p,p],[1,1]),L=Ae(Ve(w,0),Yt([[-1]]),Yt([[1]])),T=D(w,v(L,y)),A=M(b,T);A.shape[0]===1?s=At(a):s=re([a,Y(A,[1,0],[A.shape[0]-1,A.shape[1]])],0);const _=Q(M(K(L,T),y)),N=Y(o,[p,0],[n-p,r]),C=v(_,s),F=X(s);if(p===0)o=D(N,K(C,K(F,N)));else{const q=D(N,K(C,K(F,N)));o=re([Y(o,[0,0],[p,r]),q],0)}const E=X(C),B=Y(i,[0,p],[n,i.shape[1]-p]);if(p===0)i=D(B,K(K(B,s),E));else{const q=D(B,K(K(B,s),E));i=re([Y(i,[0,0],[n,p]),q],1)}return[s,o,i]}),be([l,u,m])}return!t&&n>r&&(i=Y(i,[0,0],[n,r]),o=Y(o,[0,0],[r,r])),[i,o]})}const lx=d({qr_:eO});var ue;(function(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"})(ue||(ue={}));function tO(e,t,n=ue.SUM_BY_NONZERO_WEIGHTS){const r=h(e,"losses","computeWeightedLoss");let i=null;t!=null&&(i=h(t,"weights","computeWeightedLoss"));const o=i==null?r:v(r,i);if(n===ue.NONE)return o;if(n===ue.SUM)return U(o);if(n===ue.MEAN){if(i==null)return an(o);{const a=r.size/i.size,s=M(U(o),U(i));return a>1?M(s,k(a)):s}}if(n===ue.SUM_BY_NONZERO_WEIGHTS){if(i==null)return M(U(o),k(r.size));{const a=v(i,mt(r.shape)),s=R(U(qn(a,k(0))),"float32");return M(U(o),s)}}throw Error(`Unknown reduction: ${n}`)}const Me=d({computeWeightedLoss_:tO});function nO(e,t,n,r=ue.SUM_BY_NONZERO_WEIGHTS){const i=h(e,"labels","absoluteDifference"),o=h(t,"predictions","absoluteDifference");let a=null;n!=null&&(a=h(n,"weights","absoluteDifference")),$(i.shape,o.shape,"Error in absoluteDifference: ");const s=_e(D(i,o));return Me(s,a,r)}const ux=d({absoluteDifference_:nO});function rO(e,t,n,r,i=ue.SUM_BY_NONZERO_WEIGHTS){const o=h(e,"labels","cosineDistance"),a=h(t,"predictions","cosineDistance");let s=null;r!=null&&(s=h(r,"weights","cosineDistance")),$(o.shape,a.shape,"Error in cosineDistance: ");const c=k(1),p=D(c,U(v(o,a),n,!0));return Me(p,s,i)}const hx=d({cosineDistance_:rO});function iO(e,t,n,r=ue.SUM_BY_NONZERO_WEIGHTS){let i=h(e,"labels","hingeLoss");const o=h(t,"predictions","hingeLoss");let a=null;n!=null&&(a=h(n,"weights","hingeLoss")),$(i.shape,o.shape,"Error in hingeLoss: ");const s=k(1);i=D(v(k(2),i),s);const c=zt(D(s,v(i,o)));return Me(c,a,r)}const dx=d({hingeLoss_:iO});function oO(e,t,n,r=1,i=ue.SUM_BY_NONZERO_WEIGHTS){const o=h(e,"labels","huberLoss"),a=h(t,"predictions","huberLoss");let s=null;n!=null&&(s=h(n,"weights","huberLoss")),$(o.shape,a.shape,"Error in huberLoss: ");const c=k(r),p=_e(D(a,o)),l=Bn(p,c),u=D(p,l),m=O(v(k(.5),G(l)),v(c,u));return Me(m,s,i)}const mx=d({huberLoss_:oO});function sO(e,t,n,r=1e-7,i=ue.SUM_BY_NONZERO_WEIGHTS){const o=h(e,"labels","logLoss"),a=h(t,"predictions","logLoss");let s=null;n!=null&&(s=h(n,"weights","logLoss")),$(o.shape,a.shape,"Error in logLoss: ");const c=k(1),p=k(r),l=Q(v(o,gt(O(a,p)))),u=v(D(c,o),gt(O(D(c,a),p))),m=D(l,u);return Me(m,s,i)}const fx=d({logLoss_:sO});function aO(e,t,n,r=ue.SUM_BY_NONZERO_WEIGHTS){const i=h(e,"labels","meanSquaredError"),o=h(t,"predictions","meanSquaredError");let a=null;n!=null&&(a=h(n,"weights","meanSquaredError")),$(i.shape,o.shape,"Error in meanSquaredError: ");const s=Mn(i,o);return Me(s,a,r)}const gx=d({meanSquaredError_:aO});function cO(e,t){const n=h(e,"labels","sigmoidCrossEntropyWithLogits"),r=h(t,"logits","sigmoidCrossEntropyWithLogits");$(n.shape,r.shape,"Error in sigmoidCrossEntropyWithLogits: ");const i=zt(r),o=v(r,n),a=Ni(Ne(Q(_e(r))));return O(D(i,o),a)}function pO(e,t,n,r=0,i=ue.SUM_BY_NONZERO_WEIGHTS){let o=h(e,"multiClassLabels","sigmoidCrossEntropy");const a=h(t,"logits","sigmoidCrossEntropy");let s=null;if(n!=null&&(s=h(n,"weights","sigmoidCrossEntropy")),$(o.shape,a.shape,"Error in sigmoidCrossEntropy: "),r>0){const p=k(r),l=k(1),u=k(.5);o=O(v(o,D(l,p)),v(u,p))}const c=cO(o,a);return Me(c,s,i)}const bx=d({sigmoidCrossEntropy_:pO});function lO(e,t,n=-1){if(n===-1&&(n=t.rank-1),n!==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 ${n}`);const r=$e((i,o,a)=>{const s=!0,c=Ai(o,[n],s),p=D(R(o,"float32"),c);a([i,p]);const l=Q(v(p,i)),u=U(l,[n]),m=(b,y)=>{const[w,L]=y,T=fe(b.shape,[n]);return[v(x(b,T),D(R(w,"float32"),Ne(L))),v(x(b,T),D(Ne(L),R(w,"float32")))]};return{value:u,gradFunc:m}});return r(e,t)}function uO(e,t,n,r=0,i=ue.SUM_BY_NONZERO_WEIGHTS){let o=h(e,"onehotLabels","softmaxCrossEntropy");const a=h(t,"logits","softmaxCrossEntropy");let s=null;if(n!=null&&(s=h(n,"weights","softmaxCrossEntropy")),$(o.shape,a.shape,"Error in softmaxCrossEntropy: "),r>0){const p=k(r),l=k(1),u=k(o.shape[1]);o=O(v(o,D(l,p)),M(p,u))}const c=lO(o,a);return Me(c,s,i)}const wx=d({softmaxCrossEntropy_:uO});const zm={};Oe(zm,{abs:()=>_e,acos:()=>Mu,acosh:()=>Fu,add:()=>O,addN:()=>Du,addStrict:()=>eg,all:()=>ku,any:()=>Ou,argMax:()=>Eu,argMin:()=>Cu,asin:()=>Ru,asinh:()=>_u,atan:()=>Nu,atan2:()=>Au,atanh:()=>Iu,avgPool:()=>Di,avgPool3d:()=>Tu,basicLSTMCell:()=>Su,batchNorm:()=>Jt,batchNorm2d:()=>vu,batchNorm3d:()=>Lu,batchNorm4d:()=>xu,batchToSpaceND:()=>Kn,booleanMaskAsync:()=>Qf,broadcastTo:()=>Ir,buffer:()=>ke,cast:()=>R,ceil:()=>yu,clipByValue:()=>wu,clone:()=>At,complex:()=>Ge,concat:()=>re,concat1d:()=>bu,concat2d:()=>gu,concat3d:()=>fu,concat4d:()=>mu,conv1d:()=>du,conv2d:()=>It,conv2dTranspose:()=>hu,conv3d:()=>uu,conv3dTranspose:()=>lu,cos:()=>Vn,cosh:()=>ki,cosineWindow:()=>Lr,cumsum:()=>Oi,depthToSpace:()=>pu,depthwiseConv2d:()=>zn,diag:()=>cu,dilation2d:()=>au,div:()=>M,divNoNan:()=>su,divStrict:()=>tg,dot:()=>ou,dropout:()=>Hf,elu:()=>Ei,enclosingPowerOfTwo:()=>Qs,equal:()=>Kt,equalStrict:()=>Yf,erf:()=>iu,exp:()=>Ne,expandDims:()=>Ke,expm1:()=>ru,eye:()=>Ci,fft:()=>Yn,fill:()=>Hn,floor:()=>$n,floorDiv:()=>Ri,fused:()=>nu,gather:()=>Gn,gatherND:()=>$f,greater:()=>Ve,greaterEqual:()=>ct,greaterEqualStrict:()=>zf,greaterStrict:()=>Vf,ifft:()=>cn,imag:()=>Vt,image:()=>Wf,inTopKAsync:()=>jf,irfft:()=>_i,isFinite:()=>tu,isInf:()=>eu,isNaN:()=>Ql,leakyRelu:()=>Zl,less:()=>jn,lessEqual:()=>at,lessEqualStrict:()=>Kf,lessStrict:()=>Jf,linalg:()=>qf,linspace:()=>Xl,localResponseNormalization:()=>Jl,log:()=>gt,log1p:()=>Ni,logSigmoid:()=>Kl,logSoftmax:()=>Vl,logSumExp:()=>Ai,logicalAnd:()=>ft,logicalNot:()=>Pn,logicalOr:()=>Ii,logicalXor:()=>zl,losses:()=>Bf,matMul:()=>K,max:()=>st,maxPool:()=>Ti,maxPool3d:()=>Yl,maxPoolWithArgmax:()=>Hl,maximum:()=>Tt,maximumStrict:()=>ng,mean:()=>an,min:()=>sn,minimum:()=>Bn,minimumStrict:()=>rg,mod:()=>Si,modStrict:()=>ig,moments:()=>$l,movingAverage:()=>Pf,mul:()=>v,mulStrict:()=>og,multiRNNCell:()=>Gl,multinomial:()=>jl,neg:()=>Q,norm:()=>Tr,notEqual:()=>qn,notEqualStrict:()=>Xf,oneHot:()=>Sr,ones:()=>mt,onesLike:()=>ea,op:()=>d,outerProduct:()=>Pl,pad:()=>ot,pad1d:()=>Bl,pad2d:()=>ql,pad3d:()=>Wl,pad4d:()=>Ul,pool:()=>Ml,pow:()=>je,powStrict:()=>sg,prelu:()=>vi,print:()=>Fl,prod:()=>Dl,rand:()=>kl,randomGamma:()=>Ol,randomNormal:()=>El,randomUniform:()=>Li,range:()=>vr,real:()=>kt,reciprocal:()=>Cl,relu:()=>zt,relu6:()=>xi,reshape:()=>x,reverse:()=>Pe,reverse1d:()=>Rl,reverse2d:()=>_l,reverse3d:()=>Nl,reverse4d:()=>Al,rfft:()=>Wn,round:()=>Il,rsqrt:()=>yi,scalar:()=>k,scatterND:()=>Ff,selu:()=>Tl,separableConv2d:()=>Sl,setdiff1dAsync:()=>vl,sigmoid:()=>St,sign:()=>Ll,signal:()=>Uf,sin:()=>wi,sinh:()=>bi,slice:()=>Y,slice1d:()=>xl,slice2d:()=>yl,slice3d:()=>wl,slice4d:()=>bl,softmax:()=>gl,softplus:()=>gi,spaceToBatchND:()=>Un,sparseToDense:()=>Of,spectral:()=>Mf,split:()=>Ot,sqrt:()=>le,square:()=>G,squaredDifference:()=>Mn,squaredDifferenceStrict:()=>ag,squeeze:()=>Fn,stack:()=>dt,step:()=>vt,stridedSlice:()=>fl,sub:()=>D,subStrict:()=>cg,sum:()=>U,tan:()=>ml,tanh:()=>xr,tensor:()=>Qe,tensor1d:()=>ht,tensor2d:()=>Yt,tensor3d:()=>fi,tensor4d:()=>dl,tensor5d:()=>hl,tensor6d:()=>ul,tile:()=>Ht,topk:()=>ll,transpose:()=>X,truncatedNormal:()=>pl,unsortedSegmentSum:()=>mi,unstack:()=>Dn,variable:()=>cl,where:()=>Ae,whereAsync:()=>di,zeros:()=>Ue,zerosLike:()=>W});const Mf={fft:Yn,ifft:cn,rfft:Wn,irfft:_i},Uf={hammingWindow:Yy,hannWindow:Gp,frame:$p,stft:zy},Wf={flipLeftRight:Ky,resizeNearestNeighbor:sx,resizeBilinear:ox,rotateWithOffset:Jy,cropAndResize:Vy,nonMaxSuppression:Xy,nonMaxSuppressionAsync:ex,nonMaxSuppressionWithScore:tx,nonMaxSuppressionWithScoreAsync:nx,nonMaxSuppressionPadded:rx,nonMaxSuppressionPaddedAsync:ix},qf={bandPart:ax,gramSchmidt:cx,qr:lx},Bf={absoluteDifference:ux,computeWeightedLoss:Me,cosineDistance:hx,hingeLoss:dx,huberLoss:mx,logLoss:fx,meanSquaredError:gx,sigmoidCrossEntropy:bx,softmaxCrossEntropy:wx};_y(zm);const Vs={float32:4,float16:2,int32:4,uint16:2,uint8:1,bool:1,complex64:8};const Vp=4;async function dO(e,t){const n=[],r=[],i=Array.isArray(e)?e.map(a=>a.name):Object.keys(e);for(let a=0;a<i.length;++a){const s=i[a],c=Array.isArray(e)?e[a].tensor:e[s];if(c.dtype!=="float32"&&c.dtype!=="int32"&&c.dtype!=="bool"&&c.dtype!=="string"&&c.dtype!=="complex64")throw new Error(`Unsupported dtype in weight '${s}': ${c.dtype}`);const p={name:s,shape:c.shape,dtype:c.dtype};if(c.dtype==="string"){const l=new Promise(async u=>{const m=await c.bytes(),b=m.reduce((L,T)=>L+T.length,0)+Vp*m.length,y=new Uint8Array(b);let w=0;for(let L=0;L<m.length;L++){const T=m[L],A=new Uint8Array(new Uint32Array([T.length]).buffer);y.set(A,w),w+=Vp,y.set(T,w),w+=T.length}u(y)});r.push(l)}else r.push(c.data());t!=null&&(p.group=t),n.push(p)}const o=await Promise.all(r);return{data:hO(o),specs:n}}function Vm(e,t){const n={};let r,i=0;for(const o of t){const a=o.name,s=o.dtype,c=o.shape,p=Z(c);let l;if("quantization"in o){const u=o.quantization;if(u.dtype==="uint8"||u.dtype==="uint16"){if(!("min"in u&&"scale"in u))throw new Error(`Weight ${o.name} with quantization ${u.dtype} doesn't have corresponding metadata min and scale.`)}else if(u.dtype==="float16"){if(s!=="float32")throw new Error(`Weight ${o.name} is quantized with ${u.dtype} which only supports weights of type float32 not ${s}.`)}else throw new Error(`Weight ${o.name} has unknown quantization dtype ${u.dtype}. Supported quantization dtypes are: 'uint8', 'uint16', and 'float16'.`);const m=Vs[u.dtype],b=e.slice(i,i+p*m),y=u.dtype==="uint8"?new Uint8Array(b):new Uint16Array(b);if(s==="float32")if(u.dtype==="uint8"||u.dtype==="uint16"){l=new Float32Array(y.length);for(let w=0;w<y.length;w++){const L=y[w];l[w]=L*u.scale+u.min}}else if(u.dtype==="float16")r===void 0&&(r=mO()),l=r(y);else throw new Error(`Unsupported quantization type ${u.dtype} for weight type float32.`);else if(s==="int32"){if(u.dtype!=="uint8"&&u.dtype!=="uint16")throw new Error(`Unsupported quantization type ${u.dtype} for weight type int32.`);l=new Int32Array(y.length);for(let w=0;w<y.length;w++){const L=y[w];l[w]=Math.round(L*u.scale+u.min)}}else throw new Error(`Unsupported dtype in weight '${a}': ${s}`);i+=p*m}else if(s==="string"){const u=Z(o.shape);l=[];for(let m=0;m<u;m++){const b=new Uint32Array(e.slice(i,i+Vp))[0];i+=Vp;const y=new Uint8Array(e.slice(i,i+b));l.push(y),i+=b}}else{const u=Vs[s],m=e.slice(i,i+p*u);if(s==="float32")l=new Float32Array(m);else if(s==="int32")l=new Int32Array(m);else if(s==="bool")l=new Uint8Array(m);else if(s==="complex64"){l=new Float32Array(m);const b=new Float32Array(l.length/2),y=new Float32Array(l.length/2);for(let T=0;T<b.length;T++)b[T]=l[T*2],y[T]=l[T*2+1];const w=Qe(b,c,"float32"),L=Qe(y,c,"float32");n[a]=Ge(w,L)}else throw new Error(`Unsupported dtype in weight '${a}': ${s}`);i+=p*u}s!=="complex64"&&(n[a]=Qe(l,c,s))}return n}function hO(e){if(e===null)throw new Error(`Invalid input value: ${JSON.stringify(e)}`);let t=0;const n=[];e.forEach(o=>{if(t+=o.byteLength,n.push(o.byteLength===o.buffer.byteLength?o:new o.constructor(o)),!(o instanceof Float32Array||o instanceof Int32Array||o instanceof Uint8Array))throw new Error(`Unsupported TypedArray subtype: ${o.constructor.name}`)});const r=new Uint8Array(t);let i=0;return n.forEach(o=>{r.set(new Uint8Array(o.buffer),i),i+=o.byteLength}),r.buffer}const fO=typeof Buffer!="undefined"&&(typeof Blob=="undefined"||typeof atob=="undefined"||typeof btoa=="undefined");function yx(e){return fO?Buffer.byteLength(e):new Blob([e]).size}function Ks(e){if(e.length===1)return e[0];let t=0;e.forEach(i=>{t+=i.byteLength});const n=new Uint8Array(t);let r=0;return e.forEach(i=>{n.set(new Uint8Array(i),r),r+=i.byteLength}),n.buffer}function Km(e){const t="/";for(e=e.trim();e.endsWith(t);)e=e.slice(0,e.length-1);const n=e.split(t);return n[n.length-1]}function Js(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:yx(JSON.stringify(e.modelTopology)),weightSpecsBytes:e.weightSpecs==null?0:yx(JSON.stringify(e.weightSpecs)),weightDataBytes:e.weightData==null?0:e.weightData.byteLength}}function gO(){const e=n=>{let r=n<<13,i=0;for(;(r&8388608)===0;)i-=8388608,r<<=1;return r&=~8388608,i+=947912704,r|i},t=new Uint32Array(2048);t[0]=0;for(let n=1;n<1024;n++)t[n]=e(n);for(let n=1024;n<2048;n++)t[n]=939524096+(n-1024<<13);return t}function bO(){const 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 wO(){const e=new Uint32Array(64);for(let t=0;t<64;t++)e[t]=1024;return e[0]=e[32]=0,e}function mO(){const e=gO(),t=bO(),n=wO();return r=>{const i=new ArrayBuffer(4*r.length),o=new Uint32Array(i);for(let a=0;a<r.length;a++){const s=r[a],c=e[n[s>>10]+(s&1023)]+t[s>>10];o[a]=c}return new Float32Array(i)}}class ve{constructor(){this.saveRouters=[],this.loadRouters=[]}static getInstance(){return ve.instance==null&&(ve.instance=new ve()),ve.instance}static registerSaveRouter(e){ve.getInstance().saveRouters.push(e)}static registerLoadRouter(e){ve.getInstance().loadRouters.push(e)}static getSaveHandlers(e){return ve.getHandlers(e,"save")}static getLoadHandlers(e,t){return ve.getHandlers(e,"load",t)}static getHandlers(e,t,n){const r=[],i=t==="load"?ve.getInstance().loadRouters:ve.getInstance().saveRouters;return i.forEach(o=>{const a=o(e,n);a!==null&&r.push(a)}),r}}const yO=e=>ve.registerSaveRouter(e),xO=e=>ve.registerLoadRouter(e),LO=e=>ve.getSaveHandlers(e),vO=(e,t)=>ve.getLoadHandlers(e,t);const pi="://";class Et{constructor(){this.managers={}}static getInstance(){return Et.instance==null&&(Et.instance=new Et()),Et.instance}static registerManager(e,t){f(e!=null,()=>"scheme must not be undefined or null."),e.endsWith(pi)&&(e=e.slice(0,e.indexOf(pi))),f(e.length>0,()=>"scheme must not be an empty string.");const n=Et.getInstance();f(n.managers[e]==null,()=>`A model store manager is already registered for scheme '${e}'.`),n.managers[e]=t}static getManager(e){const 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 Kp(e){if(e.indexOf(pi)===-1)throw new Error(`The url string provided does not contain a scheme. Supported schemes are: ${Et.getSchemes().join(",")}`);return{scheme:e.split(pi)[0],path:e.split(pi)[1]}}async function xx(e,t,n=!1){f(e!==t,()=>`Old path and new path are the same: '${e}'`);const r=ve.getLoadHandlers(e);f(r.length>0,()=>`Copying failed because no load handler is found for source URL ${e}.`),f(r.length<2,()=>`Copying failed because more than one (${r.length}) load handlers for source URL ${e}.`);const i=r[0],o=ve.getSaveHandlers(t);f(o.length>0,()=>`Copying failed because no save handler is found for destination URL ${t}.`),f(o.length<2,()=>`Copying failed because more than one (${r.length}) save handlers for destination URL ${t}.`);const a=o[0],s=Kp(e).scheme,c=Kp(e).path,p=s===Kp(e).scheme,l=await i.load();n&&p&&await Et.getManager(s).removeModel(c);const u=await a.save(l);return n&&!p&&await Et.getManager(s).removeModel(c),u.modelArtifactsInfo}async function Lx(){const e=Et.getSchemes(),t={};for(const n of e){const r=await Et.getManager(n).listModels();for(const i in r){const o=n+pi+i;t[o]=r[i]}}return t}async function vx(e){const t=Kp(e),n=Et.getManager(t.scheme);return n.removeModel(t.path)}async function Sx(e,t){const n=!1;return xx(e,t,n)}async function Tx(e,t){const n=!0;return xx(e,t,n)}const SO="model",TO=".json",IO=".weights.bin";function Ix(e){return new Promise(t=>setTimeout(t)).then(e)}class li{constructor(e){if(!se().getBool("IS_BROWSER"))throw new Error("browserDownloads() cannot proceed because the current environment is not a browser.");e.startsWith(li.URL_SCHEME)&&(e=e.slice(li.URL_SCHEME.length)),(e==null||e.length===0)&&(e=SO),this.modelTopologyFileName=e+TO,this.weightDataFileName=e+IO}async save(e){if(typeof document=="undefined")throw new Error("Browser downloads are not supported in this environment since `document` is not present");const 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.");{const n=[{paths:["./"+this.weightDataFileName],weights:e.weightSpecs}],r={modelTopology:e.modelTopology,format:e.format,generatedBy:e.generatedBy,convertedBy:e.convertedBy,weightsManifest:n},i=window.URL.createObjectURL(new Blob([JSON.stringify(r)],{type:"application/json"})),o=this.jsonAnchor==null?document.createElement("a"):this.jsonAnchor;if(o.download=this.modelTopologyFileName,o.href=i,await Ix(()=>o.dispatchEvent(new MouseEvent("click"))),e.weightData!=null){const a=this.weightDataAnchor==null?document.createElement("a"):this.weightDataAnchor;a.download=this.weightDataFileName,a.href=t,await Ix(()=>a.dispatchEvent(new MouseEvent("click")))}return{modelArtifactsInfo:Js(e)}}}}li.URL_SCHEME="downloads://";class AO{constructor(e){if(e==null||e.length<1)throw new Error(`When calling browserFiles, at least 1 file is required, but received ${e}`);this.files=e}async load(){const e=this.files[0],t=this.files.slice(1);return new Promise((n,r)=>{const i=new FileReader();i.onload=o=>{const a=JSON.parse(o.target.result),s=a.modelTopology;if(s==null){r(new Error(`modelTopology field is missing from file ${e.name}`));return}t.length===0&&n({modelTopology:s});const c=a.weightsManifest;if(c==null){r(new Error(`weightManifest field is missing from file ${e.name}`));return}let p;try{p=this.checkManifestAndWeightFiles(c,t)}catch(b){r(b);return}const l=[],u=[],m=[];c.forEach(b=>{b.paths.forEach(y=>{u.push(y),m.push(null)}),l.push(...b.weights)}),c.forEach(b=>{b.paths.forEach(y=>{const w=new FileReader();w.onload=L=>{const T=L.target.result,A=u.indexOf(y);m[A]=T,m.indexOf(null)===-1&&n({modelTopology:s,weightSpecs:l,weightData:Ks(m),format:a.format,generatedBy:a.generatedBy,convertedBy:a.convertedBy,userDefinedMetadata:a.userDefinedMetadata})},w.onerror=L=>r(`Failed to weights data from file of path '${y}'.`),w.readAsArrayBuffer(p[y])})})},i.onerror=o=>r(`Failed to read model topology and weights manifest JSON from file '${e.name}'. BrowserFiles supports loading Keras-style tf.Model artifacts only.`),i.readAsText(e)})}checkManifestAndWeightFiles(e,t){const n=[],r=t.map(o=>Km(o.name)),i={};for(const o of e)o.paths.forEach(a=>{const s=Km(a);if(n.indexOf(s)!==-1)throw new Error(`Duplicate file basename found in weights manifest: '${s}'`);if(n.push(s),r.indexOf(s)===-1)throw new Error(`Weight file with basename '${s}' is not provided.`);i[a]=t[r.indexOf(s)]});if(n.length!==t.length)throw new Error(`Mismatch in the number of files in weights manifest (${n.length}) and the number of weight files provided (${t.length}).`);return i}}const _O=e=>se().getBool("IS_BROWSER")&&(!Array.isArray(e)&&e.startsWith(li.URL_SCHEME))?NO(e.slice(li.URL_SCHEME.length)):null;ve.registerSaveRouter(_O);function NO(e="model"){return new li(e)}function RO(e){return new AO(e)}function Jm(e,t,n,r){a(e),n=n==null?0:n,r=r==null?1:r,s(n,r);let i=0;const o=c=>(c.then(p=>{const l=n+ ++i/e.length*(r-n);return t(l),p}),c);function a(c){f(c!=null&&Array.isArray(c)&&c.length>0,()=>"promises must be a none empty array")}function s(c,p){f(c>=0&&c<=1,()=>`Progress fraction must be in range [0, 1], but got startFraction ${c}`),f(p>=0&&p<=1,()=>`Progress fraction must be in range [0, 1], but got endFraction ${p}`),f(p>=c,()=>`startFraction must be no more than endFraction, but got startFraction ${c} and endFraction ${p}`)}return Promise.all(e.map(o))}async function Xm(e,t){t==null&&(t={});const n=t.fetchFunc==null?se().platform.fetch:t.fetchFunc,r=e.map(u=>n(u,t.requestInit,{isBinary:!0})),i=0,o=.5,a=t.onProgress==null?await Promise.all(r):await Jm(r,t.onProgress,i,o),s=a.map(u=>u.arrayBuffer()),c=.5,p=1,l=t.onProgress==null?await Promise.all(s):await Jm(s,t.onProgress,c,p);return l}async function CO(e,t="",n,r){const i=a=>Xm(a,{requestInit:r}),o=Ax(i);return o(e,t,n)}function Ax(e){return async(t,n="",r)=>{const i=t.map(()=>!1),o={},a=r!=null?r.map(()=>!1):[],s=[];if(t.forEach((b,y)=>{let w=0;b.weights.forEach(L=>{const T="quantization"in L?L.quantization.dtype:L.dtype,A=Vs[T]*Z(L.shape),_=()=>{i[y]=!0,o[y]==null&&(o[y]=[]),o[y].push({manifestEntry:L,groupOffset:w,sizeBytes:A})};r!=null?r.forEach((N,C)=>{N===L.name&&(_(),a[C]=!0)}):_(),s.push(L.name),w+=A})}),!a.every(b=>b)){const b=r.filter((y,w)=>!a[w]);throw new Error(`Could not find weights in manifest with names: ${b.join(", ")}.
Manifest JSON has weights with names: ${s.join(", ")}.`)}const c=i.reduce((b,y,w)=>(y&&b.push(w),b),[]),p=[];c.forEach(b=>{t[b].paths.forEach(y=>{const w=n+(n.endsWith("/")?"":"/")+y;p.push(w)})});const l=await e(p),u={};let m=0;return c.forEach(b=>{const y=t[b].paths.length;let w=0;for(let N=0;N<y;N++)w+=l[m+N].byteLength;const L=new ArrayBuffer(w),T=new Uint8Array(L);let A=0;for(let N=0;N<y;N++){const C=new Uint8Array(l[m+N]);T.set(C,A),A+=C.byteLength}const _=o[b];_.forEach(N=>{const C=L.slice(N.groupOffset,N.groupOffset+N.sizeBytes),F=Vm(C,[N.manifestEntry]);for(const E in F)u[E]=F[E]}),m+=y}),u}}const EO="application/octet-stream",OO="application/json";class Zm{constructor(e,t){if(this.DEFAULT_METHOD="POST",t==null&&(t={}),this.weightPathPrefix=t.weightPathPrefix,this.onProgress=t.onProgress,t.fetchFunc!=null?(f(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=se().platform.fetch,f(e!=null&&e.length>0,()=>"URL path for http must not be null, undefined or empty."),Array.isArray(e)&&f(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.");const t=Object.assign({method:this.DEFAULT_METHOD},this.requestInit);t.body=new FormData();const n=[{paths:["./model.weights.bin"],weights:e.weightSpecs}],r={modelTopology:e.modelTopology,format:e.format,generatedBy:e.generatedBy,convertedBy:e.convertedBy,userDefinedMetadata:e.userDefinedMetadata,weightsManifest:n};t.body.append("model.json",new Blob([JSON.stringify(r)],{type:OO}),"model.json"),e.weightData!=null&&t.body.append("model.weights.bin",new Blob([e.weightData],{type:EO}),"model.weights.bin");const i=await this.fetch(this.path,t);if(i.ok)return{modelArtifactsInfo:Js(e),responses:[i]};throw new Error(`BrowserHTTPRequest.save() failed due to HTTP response status ${i.status}.`)}async load(){const 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(l){let u=`Failed to parse model JSON of response from ${this.path}.`;throw this.path.endsWith(".pb")?u+=" 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.":u+=" Please make sure the server is serving valid JSON for this request.",new Error(u)}const n=t.modelTopology,r=t.weightsManifest,i=t.generatedBy,o=t.convertedBy,a=t.format,s=t.userDefinedMetadata;if(n==null&&r==null)throw new Error(`The JSON from HTTP path ${this.path} contains neither model topology or manifest for weights.`);let c,p;if(r!=null){const l=await this.loadWeights(r);[c,p]=l}return{modelTopology:n,weightSpecs:c,weightData:p,userDefinedMetadata:s,generatedBy:i,convertedBy:o,format:a}}async loadWeights(e){const t=Array.isArray(this.path)?this.path[1]:this.path,[n,r]=kO(t),i=this.weightPathPrefix||n,o=[];for(const c of e)o.push(...c.weights);const a=[];e.forEach(c=>{c.paths.forEach(p=>{a.push(i+p+r)})});const s=await Xm(a,{requestInit:this.requestInit,fetchFunc:this.fetch,onProgress:this.onProgress});return[o,Ks(s)]}}Zm.URL_SCHEME_REGEX=/^https?:\/\//;function kO(e){const t=e.lastIndexOf("/"),n=e.lastIndexOf("?"),r=e.substring(0,t),i=n>t?e.substring(n):"";return[r+"/",i]}function Qm(e){return e.match(Zm.URL_SCHEME_REGEX)!=null}const Nx=(e,t)=>{if(typeof fetch=="undefined"&&(t==null||t.fetchFunc==null))return null;{let n=!0;if(Array.isArray(e)?n=e.every(r=>Qm(r)):n=Qm(e),n)return ef(e,t)}return null};ve.registerSaveRouter(Nx);ve.registerLoadRouter(Nx);function ef(e,t){return new Zm(e,t)}function DO(e,t){return ef(e,t)}class tf{constructor(e){this.modelArtifacts=e}async load(){return this.modelArtifacts}}class FO{constructor(e){this.saveHandler=e}async save(e){return this.saveHandler(e)}}function MO(e,t,n,r){if(arguments.length===1){const i=e.modelTopology!=null||e.weightSpecs!=null;return i?new tf(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 tf({modelTopology:e}))}else return 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 tf({modelTopology:e,weightSpecs:t,weightData:n,trainingConfig:r})}function UO(e){return new FO(e)}const lg={};Oe(lg,{browserFiles:()=>RO,browserHTTPRequest:()=>DO,concatenateArrayBuffers:()=>Ks,copyModel:()=>Sx,decodeWeights:()=>Vm,encodeWeights:()=>dO,fromMemory:()=>MO,getLoadHandlers:()=>vO,getModelArtifactsInfoForJSON:()=>Js,getSaveHandlers:()=>LO,http:()=>ef,isHTTPScheme:()=>Qm,listModels:()=>Lx,loadWeights:()=>CO,moveModel:()=>Tx,registerLoadRouter:()=>xO,registerSaveRouter:()=>yO,removeModel:()=>vx,weightsLoaderFactory:()=>Ax,withSaveHandler:()=>UO});function WO(e,t,n){const r=h(e,"labels","confusionMatrix"),i=h(t,"predictions","confusionMatrix");f(n==null||n>0&&Number.isInteger(n),()=>`If provided, numClasses must be a positive integer, but got ${n}`),f(r.rank===1,()=>`Expected the rank of labels to be 1, but got ${r.rank}`),f(i.rank===1,()=>`Expected the rank of predictions to be 1, but got ${i.rank}`),f(r.shape[0]===i.shape[0],()=>`Mismatch in the number of examples: ${r.shape[0]} vs. ${i.shape[0]}. Labels and predictions should have the same number of elements.`),f(n>0&&Number.isInteger(n),()=>`numClasses is required to be a positive integer, but got ${n}`);const o=Sr(R(r,"int32"),n),a=Sr(R(i,"int32"),n),s=X(o);return R(K(s,a),"int32")}const qO=d({confusionMatrix_:WO});const pg={};Oe(pg,{confusionMatrix:()=>qO});const Zf={};Oe(Zf,{fromPixels:()=>jO,toPixels:()=>PO});let ui;function BO(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 n=!1,r=!1,i=!1,o=!1,a=!1;if(e.data instanceof Uint8Array)n=!0;else if(typeof ImageData!="undefined"&&e instanceof ImageData)r=!0;else if(typeof HTMLVideoElement!="undefined"&&e instanceof HTMLVideoElement)i=!0;else if(typeof HTMLImageElement!="undefined"&&e instanceof HTMLImageElement)o=!0;else if(e.getContext!=null)a=!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(i){const b=2;if(i&&e.readyState<b)throw new Error("The video element has not loaded data yet. Please wait for `loadeddata` event on the <video> element.")}const s=Fi(ra,g.backendName);if(s!=null){const b={pixels:e},y={numChannels:t};return g.runKernel(ra,b,y)}const[c,p]=i?[e.videoWidth,e.videoHeight]:[e.width,e.height];let l;a?l=e.getContext("2d").getImageData(0,0,c,p).data:r||n?l=e.data:(o||i)&&(ui==null&&(ui=document.createElement("canvas").getContext("2d")),ui.canvas.width=c,ui.canvas.height=p,ui.drawImage(e,0,0,c,p),l=ui.getImageData(0,0,c,p).data);let u;if(t===4)u=new Int32Array(l);else{const b=c*p;u=new Int32Array(b*t);for(let y=0;y<b;y++)for(let w=0;w<t;++w)u[y*t+w]=l[y*4+w]}const m=[p,c,t];return fi(u,m,"int32")}async function PO(e,t){let n=h(e,"img","toPixels");if(!(e instanceof de)){const L=n;n=R(L,"int32"),L.dispose()}if(n.rank!==2&&n.rank!==3)throw new Error(`toPixels only supports rank 2 or 3 tensors, got rank ${n.rank}.`);const[r,i]=n.shape.slice(0,2),o=n.rank===2?1:n.shape[2];if(o>4||o===2)throw new Error(`toPixels only supports depth of size 1, 3 or 4 but got ${o}`);const a=await n.data(),s=sn(n),c=st(n),p=await Promise.all([s.data(),c.data()]),l=p[0],u=p[1],m=l[0],b=u[0];if(s.dispose(),c.dispose(),n.dtype==="float32"){if(m<0||b>1)throw new Error(`Tensor values for a float32 Tensor must be in the range [0 - 1] but got range [${m} - ${b}].`)}else if(n.dtype==="int32"){if(m<0||b>255)throw new Error(`Tensor values for a int32 Tensor must be in the range [0 - 255] but got range [${m} - ${b}].`)}else throw new Error(`Unsupported type for toPixels: ${n.dtype}. Please use float32 or int32 tensors.`);const y=n.dtype==="float32"?255:1,w=new Uint8ClampedArray(i*r*4);for(let L=0;L<r*i;++L){let T,A,_,N;o===1?(T=a[L]*y,A=a[L]*y,_=a[L]*y,N=255):o===3?(T=a[L*3]*y,A=a[L*3+1]*y,_=a[L*3+2]*y,N=255):o===4&&(T=a[L*4]*y,A=a[L*4+1]*y,_=a[L*4+2]*y,N=a[L*4+3]*y);const C=L*4;w[C+0]=Math.round(T),w[C+1]=Math.round(A),w[C+2]=Math.round(_),w[C+3]=Math.round(N)}if(t!=null){t.width=i,t.height=r;const L=t.getContext("2d"),T=new ImageData(w,i,r);L.putImageData(T,0,0)}return n!==e&&n.dispose(),w}const jO=d({fromPixels_:BO});const Gf={};Oe(Gf,{prepareAndValidate:()=>_x});function _x(e,t){if(e.rank<1)throw new Error(`tf.gatherND() expects the input to be rank 1 or higher, but the rank was ${e.rank}.`);if(t.rank<1)throw new Error(`tf.gatherND() expects the indices to be rank 1 or higher, but the rank was ${t.rank}.`);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[t.rank-1]>e.rank)throw new Error(`index innermost dimension length must be <= tensor rank; saw: ${t.shape[t.rank-1]} vs. ${e.rank}`);if(e.size===0)throw new Error(`Requested more than 0 entries, but input is empty. Input shape: ${e.shape}.`);const n=t.shape,r=n[n.length-1];let i=1;for(let p=0;p<n.length-1;++p)i*=n[p];const o=e.shape,a=n.slice();a.pop();let s=1;for(let p=r;p<e.rank;++p)s*=o[p],a.push(o[p]);const c=[...Bt(e.shape).map(p=>p/s),1].slice(0,r);return[a,i,s,c]}const Ef={};Oe(Ef,{Serializable:()=>nf,SerializationMap:()=>fr,registerClass:()=>Xe});class nf{getClassName(){return this.constructor.className}static fromConfig(e,t){return new e(t)}}class fr{constructor(){this.classNameMap={}}static getMap(){return fr.instance==null&&(fr.instance=new fr()),fr.instance}static register(e){fr.getMap().classNameMap[e.className]=[e,e.fromConfig]}}function Xe(e){f(e.className!=null,()=>"Class being registered does not have the static className property defined."),f(typeof e.className=="string",()=>"className is required to be a string, but got type "+typeof e.className),f(e.className.length>0,()=>"Class being registered has an empty-string as its className, which is disallowed."),fr.register(e)}const Qv="2.3.0";class Ze extends nf{minimize(e,t=!1,n){const{value:r,grads:i}=this.computeGradients(e,n);if(n!=null){const o=n.map(a=>({name:a.name,tensor:i[a.name]}));this.applyGradients(o)}else this.applyGradients(i);return be(i),t?r:(r.dispose(),null)}get iterations(){return this.iterations_==null&&(this.iterations_=0),this.iterations_}incrementIterations(){this.iterations_=this.iterations+1}computeGradients(e,t){return kh(e,t)}dispose(){this.iterations_!=null&&be(this.iterations_)}async saveIterations(){return this.iterations_==null&&(this.iterations_=0),{name:"iter",tensor:k(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 ${this.getClassName()}`)}async extractIterations(e){return this.iterations_=(await e[0].tensor.data())[0],e.slice(1)}}Object.defineProperty(Ze,Symbol.hasInstance,{value:e=>e.minimize!=null&&e.computeGradients!=null&&e.applyGradients!=null});class kn extends Ze{constructor(e,t,n=null){super();this.learningRate=e,this.rho=t,this.epsilon=n,this.accumulatedGrads=[],this.accumulatedUpdates=[],n==null&&(this.epsilon=g.backend.epsilon())}applyGradients(e){const t=Array.isArray(e)?e.map(n=>n.name):Object.keys(e);t.forEach((n,r)=>{const i=g.registeredVariables[n],o=!1;this.accumulatedGrads[r]==null&&(this.accumulatedGrads[r]={originalName:`${n}/accum_grad`,variable:ce(()=>W(i).variable(o))}),this.accumulatedUpdates[r]==null&&(this.accumulatedUpdates[r]={originalName:`${n}/accum_var`,variable:ce(()=>W(i).variable(o))});const a=Array.isArray(e)?e[r].tensor:e[n];if(a==null)return;const s=this.accumulatedGrads[r].variable,c=this.accumulatedUpdates[r].variable;ce(()=>{const p=O(v(s,this.rho),v(G(a),1-this.rho)),l=v(M(le(O(c,this.epsilon)),le(O(s,this.epsilon))),a),u=O(v(c,this.rho),v(G(l),1-this.rho));s.assign(p),c.assign(u);const m=O(v(l,-this.learningRate),i);i.assign(m)})}),this.incrementIterations()}dispose(){this.accumulatedUpdates!=null&&(be(this.accumulatedGrads.map(e=>e.variable)),be(this.accumulatedUpdates.map(e=>e.variable)))}async getWeights(){const e=[...this.accumulatedGrads,...this.accumulatedUpdates];return[await this.saveIterations()].concat(e.map(t=>({name:t.originalName,tensor:t.variable})))}async setWeights(e){e=await this.extractIterations(e);const t=e.length/2,n=!1;this.accumulatedGrads=e.slice(0,t).map(r=>({originalName:r.name,variable:r.tensor.variable(n)})),this.accumulatedUpdates=e.slice(t,t*2).map(r=>({originalName:r.name,variable:r.tensor.variable(n)}))}getConfig(){return{learningRate:this.learningRate,rho:this.rho,epsilon:this.epsilon}}static fromConfig(e,t){return new e(t.learningRate,t.rho,t.epsilon)}}kn.className="Adadelta";Xe(kn);class On extends Ze{constructor(e,t=.1){super();this.learningRate=e,this.initialAccumulatorValue=t,this.accumulatedGrads=[]}applyGradients(e){const t=Array.isArray(e)?e.map(n=>n.name):Object.keys(e);t.forEach((n,r)=>{const i=g.registeredVariables[n];if(this.accumulatedGrads[r]==null){const s=!1;this.accumulatedGrads[r]={originalName:`${n}/accumulator`,variable:ce(()=>Hn(i.shape,this.initialAccumulatorValue).variable(s))}}const o=Array.isArray(e)?e[r].tensor:e[n];if(o==null)return;const a=this.accumulatedGrads[r].variable;ce(()=>{const s=O(a,G(o));a.assign(s);const c=O(v(M(o,le(O(s,g.backend.epsilon()))),-this.learningRate),i);i.assign(c)})}),this.incrementIterations()}dispose(){this.accumulatedGrads!=null&&be(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);const t=!1;this.accumulatedGrads=e.map(n=>({originalName:n.name,variable:n.tensor.variable(t)}))}getConfig(){return{learningRate:this.learningRate,initialAccumulatorValue:this.initialAccumulatorValue}}static fromConfig(e,t){return new e(t.learningRate,t.initialAccumulatorValue)}}On.className="Adagrad";Xe(On);class En extends Ze{constructor(e,t,n,r=null){super();this.learningRate=e,this.beta1=t,this.beta2=n,this.epsilon=r,this.accumulatedFirstMoment=[],this.accumulatedSecondMoment=[],ce(()=>{this.accBeta1=k(t).variable(),this.accBeta2=k(n).variable()}),r==null&&(this.epsilon=g.backend.epsilon())}applyGradients(e){const t=Array.isArray(e)?e.map(n=>n.name):Object.keys(e);ce(()=>{const n=D(1,this.accBeta1),r=D(1,this.accBeta2);t.forEach((i,o)=>{const a=g.registeredVariables[i],s=!1;this.accumulatedFirstMoment[o]==null&&(this.accumulatedFirstMoment[o]={originalName:`${i}/m`,variable:ce(()=>W(a).variable(s))}),this.accumulatedSecondMoment[o]==null&&(this.accumulatedSecondMoment[o]={originalName:`${i}/v`,variable:ce(()=>W(a).variable(s))});const c=Array.isArray(e)?e[o].tensor:e[i];if(c==null)return;const p=this.accumulatedFirstMoment[o].variable,l=this.accumulatedSecondMoment[o].variable,u=O(v(p,this.beta1),v(c,1-this.beta1)),m=O(v(l,this.beta2),v(G(c),1-this.beta2)),b=M(u,n),y=M(m,r);p.assign(u),l.assign(m);const w=O(v(M(b,O(le(y),this.epsilon)),-this.learningRate),a);a.assign(w)}),this.accBeta1.assign(v(this.accBeta1,this.beta1)),this.accBeta2.assign(v(this.accBeta2,this.beta2))}),this.incrementIterations()}dispose(){this.accBeta1.dispose(),this.accBeta2.dispose(),this.accumulatedFirstMoment!=null&&be(this.accumulatedFirstMoment.map(e=>e.variable)),this.accumulatedSecondMoment!=null&&be(this.accumulatedSecondMoment.map(e=>e.variable))}async getWeights(){const 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),ce(()=>{this.accBeta1.assign(je(this.beta1,this.iterations_+1)),this.accBeta2.assign(je(this.beta2,this.iterations_+1))});const t=e.length/2,n=!1;this.accumulatedFirstMoment=e.slice(0,t).map(r=>({originalName:r.name,variable:r.tensor.variable(n)})),this.accumulatedSecondMoment=e.slice(t,t*2).map(r=>({originalName:r.name,variable:r.tensor.variable(n)}))}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)}}En.className="Adam";Xe(En);class Cn extends Ze{constructor(e,t,n,r=null,i=0){super();this.learningRate=e,this.beta1=t,this.beta2=n,this.epsilon=r,this.decay=i,this.accumulatedFirstMoment=[],this.accumulatedWeightedInfNorm=[],ce(()=>{this.iteration=k(0).variable(),this.accBeta1=k(t).variable()}),r==null&&(this.epsilon=g.backend.epsilon())}applyGradients(e){const t=Array.isArray(e)?e.map(n=>n.name):Object.keys(e);ce(()=>{const n=D(1,this.accBeta1),r=M(-this.learningRate,O(v(this.iteration,this.decay),1));t.forEach((i,o)=>{const a=g.registeredVariables[i],s=!1;this.accumulatedFirstMoment[o]==null&&(this.accumulatedFirstMoment[o]={originalName:`${i}/m`,variable:W(a).variable(s)}),this.accumulatedWeightedInfNorm[o]==null&&(this.accumulatedWeightedInfNorm[o]={originalName:`${i}/v`,variable:W(a).variable(s)});const c=Array.isArray(e)?e[o].tensor:e[i];if(c==null)return;const p=this.accumulatedFirstMoment[o].variable,l=this.accumulatedWeightedInfNorm[o].variable,u=O(v(p,this.beta1),v(c,1-this.beta1)),m=v(l,this.beta2),b=_e(c),y=Tt(m,b);p.assign(u),l.assign(y);const w=O(v(M(r,n),M(u,O(y,this.epsilon))),a);a.assign(w)}),this.iteration.assign(O(this.iteration,1)),this.accBeta1.assign(v(this.accBeta1,this.beta1))}),this.incrementIterations()}dispose(){this.accBeta1.dispose(),this.iteration.dispose(),this.accumulatedFirstMoment!=null&&be(this.accumulatedFirstMoment.map(e=>e.variable)),this.accumulatedWeightedInfNorm!=null&&be(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)}}Cn.className="Adamax";Xe(Cn);class $t extends Ze{constructor(e){super();this.learningRate=e,this.setLearningRate(e)}applyGradients(e){const t=Array.isArray(e)?e.map(n=>n.name):Object.keys(e);t.forEach((n,r)=>{const i=Array.isArray(e)?e[r].tensor:e[n];if(i==null)return;const o=g.registeredVariables[n];ce(()=>{const a=O(v(this.c,i),o);o.assign(a)})}),this.incrementIterations()}setLearningRate(e){this.learningRate=e,this.c!=null&&this.c.dispose(),this.c=Dh(k(-e))}dispose(){this.c.dispose()}async getWeights(){return[await this.saveIterations()]}async setWeights(e){if(e=await this.extractIterations(e),e.length!==0)throw new Error("SGD optimizer does not have settable weights.")}getConfig(){return{learningRate:this.learningRate}}static fromConfig(e,t){return new e(t.learningRate)}}$t.className="SGD";Xe($t);class Rn extends $t{constructor(e,t,n=!1){super(e);this.learningRate=e,this.momentum=t,this.useNesterov=n,this.accumulations=[],this.m=k(this.momentum)}applyGradients(e){const t=Array.isArray(e)?e.map(n=>n.name):Object.keys(e);t.forEach((n,r)=>{const i=g.registeredVariables[n];if(this.accumulations[r]==null){const s=!1;this.accumulations[r]={originalName:`${n}/momentum`,variable:ce(()=>W(i).variable(s))}}const o=this.accumulations[r].variable,a=Array.isArray(e)?e[r].tensor:e[n];if(a==null)return;ce(()=>{let s;const c=O(v(this.m,o),a);this.useNesterov?s=O(v(this.c,O(a,v(c,this.m))),i):s=O(v(this.c,c),i),o.assign(c),i.assign(s)})}),this.incrementIterations()}dispose(){this.m.dispose(),this.accumulations!=null&&be(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);const t=!1;this.accumulations=e.map(n=>({originalName:n.name,variable:n.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)}}Rn.className="Momentum";Xe(Rn);class _n extends Ze{constructor(e,t=.9,n=0,r=null,i=!1){super();if(this.learningRate=e,this.decay=t,this.momentum=n,this.epsilon=r,this.accumulatedMeanSquares=[],this.accumulatedMoments=[],this.accumulatedMeanGrads=[],this.centered=i,r==null&&(this.epsilon=g.backend.epsilon()),e==null)throw new Error("learningRate for RMSPropOptimizer must be defined.")}applyGradients(e){const t=Array.isArray(e)?e.map(n=>n.name):Object.keys(e);t.forEach((n,r)=>{const i=g.registeredVariables[n],o=!1;this.accumulatedMeanSquares[r]==null&&(this.accumulatedMeanSquares[r]={originalName:`${n}/rms`,variable:ce(()=>W(i).variable(o))}),this.accumulatedMoments[r]==null&&(this.accumulatedMoments[r]={originalName:`${n}/momentum`,variable:ce(()=>W(i).variable(o))}),this.accumulatedMeanGrads[r]==null&&this.centered&&(this.accumulatedMeanGrads[r]={originalName:`${n}/mg`,variable:ce(()=>W(i).variable(o))});const a=Array.isArray(e)?e[r].tensor:e[n];if(a==null)return;const s=this.accumulatedMeanSquares[r].variable,c=this.accumulatedMoments[r].variable;ce(()=>{const p=O(v(s,this.decay),v(G(a),1-this.decay));if(this.centered){const l=this.accumulatedMeanGrads[r].variable,u=O(v(l,this.decay),v(a,1-this.decay)),m=M(v(a,this.learningRate),le(D(p,O(G(u),this.epsilon)))),b=O(v(c,this.momentum),m);s.assign(p),l.assign(u),c.assign(b);const y=D(i,b);i.assign(y)}else{const l=O(v(s,this.decay),v(G(a),1-this.decay)),u=O(v(c,this.momentum),M(v(a,this.learningRate),le(O(l,this.epsilon))));s.assign(l),c.assign(u);const m=D(i,u);i.assign(m)}})}),this.incrementIterations()}dispose(){this.accumulatedMeanSquares!=null&&be(this.accumulatedMeanSquares.map(e=>e.variable)),this.accumulatedMeanGrads!=null&&this.centered&&be(this.accumulatedMeanGrads.map(e=>e.variable)),this.accumulatedMoments!=null&&be(this.accumulatedMoments.map(e=>e.variable))}async getWeights(){const 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);const t=this.centered?e.length/3:e.length/2,n=!1;this.accumulatedMeanSquares=e.slice(0,t).map(r=>({originalName:r.name,variable:r.tensor.variable(n)})),this.accumulatedMoments=e.slice(t,t*2).map(r=>({originalName:r.name,variable:r.tensor.variable(n)})),this.centered&&(this.accumulatedMeanGrads=e.slice(t*2,t*3).map(r=>({originalName:r.name,variable:r.tensor.variable(n)})))}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)}}_n.className="RMSProp";Xe(_n);class In{static sgd(e){return new $t(e)}static momentum(e,t,n=!1){return new Rn(e,t,n)}static rmsprop(e,t=.9,n=0,r=null,i=!1){return new _n(e,t,n,r,i)}static adam(e=.001,t=.9,n=.999,r=null){return new En(e,t,n,r)}static adadelta(e=.001,t=.95,n=null){return new kn(e,t,n)}static adamax(e=.002,t=.9,n=.999,r=null,i=0){return new Cn(e,t,n,r,i)}static adagrad(e,t=.1){return new On(e,t)}}Rn,$t,kn,On,_n,Cn,En;const tS={sgd:In.sgd,momentum:In.momentum,adadelta:In.adadelta,adagrad:In.adagrad,rmsprop:In.rmsprop,adamax:In.adamax,adam:In.adam};const GO=(()=>typeof requestAnimationFrame!="undefined"?requestAnimationFrame:typeof setImmediate!="undefined"?setImmediate:e=>e())();function _S(){return new Promise(e=>GO(()=>e()))}function $O(e,t,n){const r=n*(typeof e=="number"?e:e[0]),i=t*(typeof e=="number"?e:e[1]);return[r,i]}function HO(e,t,n,r=!0){let i=[];if(r)i=i.concat(t.slice(0)),i.push(e[0]/n),i=i.concat(e.slice(1));else{i=i.concat(e[0]);const o=t.length;for(let a=0;a<o;++a)i=i.concat([e[a+1]/t[a],t[a]]);i=i.concat(e.slice(o+1))}return i}function YO(e,t,n=!0){const r=[];if(n){r.push(t);for(let i=t+1;i<e;++i)i<=2*t?(r.push(i),r.push(i-(t+1))):r.push(i)}else{const i=[],o=[];for(let a=1;a<e;++a)a>=t*2+1||a%2===1?o.push(a):i.push(a);r.push(...i),r.push(0),r.push(...o)}return r}function zO(e,t,n,r=!0){const i=[];r?i.push(e[0]/n):i.push(e[0]*n);for(let o=1;o<e.length;++o)o<=t.length?r?i.push(t[o-1]*e[o]):i.push(e[o]/t[o-1]):i.push(e[o]);return i}function VO(e,t){const n=[0];for(let r=0;r<t;++r)n.push(e[r][0]);return n}function KO(e,t,n){const r=e.slice(0,1);for(let i=0;i<n;++i)r.push(e[i+1]-t[i][0]-t[i][1]);return r}const rf=1.7580993408473768,of=1.0507009873554805;const JO=.3275911,XO=.254829592,ZO=-.284496736,QO=1.421413741,ek=-1.453152027,tk=1.061405429;function nk(...e){se().getBool("IS_TEST")||console.warn(...e)}function rk(...e){se().getBool("IS_TEST")||console.log(...e)}function ik(e,t){if(e.length!==t.length)throw new Error(`Cannot merge real and imag arrays of different lengths. real:${e.length}, imag: ${t.length}.`);const n=new Float32Array(e.length*2);for(let r=0;r<n.length;r+=2)n[r]=e[r/2],n[r+1]=t[r/2];return n}function ok(e){const t=new Float32Array(e.length/2),n=new Float32Array(e.length/2);for(let r=0;r<e.length;r+=2)t[r/2]=e[r],n[r/2]=e[r+1];return{real:t,imag:n}}function sk(e){const t=Math.ceil(e.length/4),n=new Float32Array(t),r=new Float32Array(t);for(let i=0;i<e.length;i+=4)n[Math.floor(i/4)]=e[i],r[Math.floor(i/4)]=e[i+1];return{real:n,imag:r}}function ak(e){const t=Math.floor(e.length/4),n=new Float32Array(t),r=new Float32Array(t);for(let i=2;i<e.length;i+=4)n[Math.floor(i/4)]=e[i],r[Math.floor(i/4)]=e[i+1];return{real:n,imag:r}}function ck(e,t){const n=e[t*2],r=e[t*2+1];return{real:n,imag:r}}function pk(e,t,n,r){e[r*2]=t,e[r*2+1]=n}function lk(e,t){const n=new Float32Array(e/2),r=new Float32Array(e/2);for(let i=0;i<Math.ceil(e/2);i++){const o=(t?2:-2)*Math.PI*(i/e);n[i]=Math.cos(o),r[i]=Math.sin(o)}return{real:n,imag:r}}function uk(e,t,n){const r=(n?2:-2)*Math.PI*(e/t),i=Math.cos(r),o=Math.sin(r);return{real:i,imag:o}}const dg={};Oe(dg,{ERF_A1:()=>XO,ERF_A2:()=>ZO,ERF_A3:()=>QO,ERF_A4:()=>ek,ERF_A5:()=>tk,ERF_P:()=>JO,PARALLELIZE_THRESHOLD:()=>Mp,SELU_SCALE:()=>of,SELU_SCALEALPHA:()=>rf,applyActivation:()=>dr,assertAndGetBroadcastShape:()=>H,assertAxesAreInnerMostDims:()=>U0,assertParamsConsistent:()=>Cm,assignToTypedArray:()=>pk,axesAreInnerMostDims:()=>Am,calculateShapes:()=>Gy,castTensor:()=>hk,combineLocations:()=>Dy,complexWithEvenIndex:()=>sk,complexWithOddIndex:()=>ak,computeConv2DInfo:()=>We,computeConv3DInfo:()=>Tn,computeDefaultPad:()=>Rm,computeDilation2DInfo:()=>J0,computeOptimalWindowSize:()=>r1,computeOutAndReduceShapes:()=>Nm,computeOutShape:()=>Em,computePool2DInfo:()=>Ct,computePool3DInfo:()=>on,convertConv2DDataFormat:()=>Sn,eitherStridesOrDilationsAreOne:()=>ae,expandShapeToKeepDim:()=>fe,exponent:()=>uk,exponents:()=>lk,getAxesPermutation:()=>he,getBroadcastDims:()=>jR,getComplexWithIndex:()=>ck,getFusedBiasGradient:()=>hr,getFusedDyActivation:()=>ur,getImageCenter:()=>$O,getInnerMostAxes:()=>Ee,getPermuted:()=>YO,getReductionAxes:()=>ie,getReshaped:()=>HO,getReshapedPermuted:()=>zO,getSliceBeginCoords:()=>VO,getSliceSize:()=>KO,getUndoAxesPermutation:()=>pr,linspaceImpl:()=>mk,log:()=>rk,mergeRealAndImagArrays:()=>ik,prepareAndValidate:()=>_x,prepareSplitSize:()=>$m,reshapeTensor:()=>dk,segment_util:()=>Wy,shouldFuse:()=>mr,splitRealAndImagArrays:()=>ok,tupleValuesAreOne:()=>ut,upcastType:()=>wr,validateInput:()=>Bp,validateUpdateShape:()=>Hm,warn:()=>nk});function hk(e,t,n){if(t==="complex64"){if(e.dtype==="complex64")return e.clone();const r=Ue(e.shape),i=R(e,"float32"),o=n.complex(i,r);return r.dispose(),i.dispose(),o}if(!fm(e.dtype,t))return g.makeTensorFromDataId(e.dataId,e.shape,t);if(e.dtype==="complex64"){const r=n.real(e),i=R(r,t);return r.dispose(),i}if(t==="int32")return n.int(e);if(t==="bool"){const r=k(0,e.dtype),i=n.notEqual(e,r);return r.dispose(),i}else throw new Error(`Error in Cast: failed to cast ${e.dtype} to ${t}`)}function dk(e,t){return g.makeTensorFromDataId(e.dataId,t,e.dtype)}function mk(e,t,n){const r=(t-e)/(n-1),i=Pt(n,"float32");i[0]=e;for(let o=1;o<i.length;o++)i[o]=i[o-1]+r;return ht(i,"float32")}function Rx(e,t,n){const r=new Array(e.rank).fill(0),i=e.shape.slice();return t.map(o=>{const a=[...i];a[n]=o;const s=Y(e,r,a);return r[n]+=o,s})}function Cx(e,t){const n=new Array(e.rank);for(let i=0;i<n.length;i++)n[i]=e.shape[i]*t[i];const r=ke(n,e.dtype);for(let i=0;i<r.values.length;++i){const o=r.indexToLoc(i),a=new Array(e.rank);for(let c=0;c<a.length;c++)a[c]=o[c]%e.shape[c];const s=e.locToIndex(a);r.values[i]=e.values[s]}return r.toTensor()}function Ex(e,t,n,r,i){const o=t[t.length-1],[a,s]=[e.length/o,o],c=ni(n,a*r),p=ni("int32",a*r);for(let u=0;u<a;u++){const m=u*s,b=e.subarray(m,m+s),y=[];for(let A=0;A<b.length;A++)y.push({value:b[A],index:A});y.sort((A,_)=>_.value-A.value);const w=u*r,L=c.subarray(w,w+r),T=p.subarray(w,w+r);for(let A=0;A<r;A++)L[A]=y[A].value,T[A]=y[A].index}const l=t.slice();return l[l.length-1]=r,[Qe(c,l,n),Qe(p,l,"int32")]}const hg={};Oe(hg,{nonMaxSuppressionV3Impl:()=>Hp,nonMaxSuppressionV4Impl:()=>Yp,nonMaxSuppressionV5Impl:()=>zp,split:()=>Rx,tile:()=>Cx,topkImpl:()=>Ex,whereImpl:()=>qp});const fk=1e-7,gk=1e-4;class RS{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}}class CS{time(e){return I("time")}read(e){return I("read")}readSync(e){return I("readSync")}numDataIds(){return I("numDataIds")}disposeData(e){return I("disposeData")}write(e,t,n){return I("write")}move(e,t,n,r){return I("move")}memory(){return I("memory")}floatPrecision(){return I("floatPrecision")}epsilon(){return this.floatPrecision()===32?fk:gk}batchMatMul(e,t,n,r){return I("batchMatMul")}fusedBatchMatMul({a:e,b:t,transposeA:n,transposeB:r,bias:i,activation:o,preluActivationWeights:a}){return I("fusedBatchMatMul")}slice(e,t,n){return I("slice")}stridedSlice(e,t,n,r){return I("stridedSlice")}unstack(e,t){return I("unstack")}reverse(e,t){return I("reverse")}concat(e,t){return I("concat")}neg(e){return I("neg")}add(e,t){return I("add")}addN(e){return I("addN")}subtract(e,t){return I("subtract")}multiply(e,t){return I("multiply")}realDivide(e,t){return I("realDivide")}floorDiv(e,t){return I("floorDiv")}sum(e,t){return I("sum")}prod(e,t){return I("prod")}unsortedSegmentSum(e,t,n){return I("unsortedSegmentSum")}argMin(e,t){return I("argMin")}argMax(e,t){return I("argMax")}equal(e,t){return I("equal")}notEqual(e,t){return I("notEqual")}less(e,t){return I("less")}lessEqual(e,t){return I("lessEqual")}greater(e,t){return I("greater")}greaterEqual(e,t){return I("greaterEqual")}logicalNot(e){return I("logicalNot")}logicalAnd(e,t){return I("logicalAnd")}logicalOr(e,t){return I("logicalOr")}where(e){return I("where")}select(e,t,n){return I("select")}topk(e,t,n){return I("topk")}min(e,t){return I("min")}minimum(e,t){return I("minimum")}mod(e,t){return I("mod")}max(e,t){return I("max")}maximum(e,t){return I("maximum")}all(e,t){return I("all")}any(e,t){return I("any")}squaredDifference(e,t){return I("squaredDifference")}ceil(e){return I("ceil")}floor(e){return I("floor")}round(e){return I("round")}sign(e){return I("sign")}isNaN(e){return I("isNaN")}isInf(e){return I("isInf")}isFinite(e){return I("isFinite")}pow(e,t){return I("pow")}exp(e){return I("exp")}expm1(e){return I("expm1")}softmax(e,t){return I("softmax")}log(e){return I("log")}log1p(e){return I("log1p")}sqrt(e){return I("sqrt")}rsqrt(e){return I("rsqrt")}square(e){return I("square")}reciprocal(e){return I("reciprocal")}relu(e){return I("relu")}relu6(e){return I("relu6")}prelu(e,t){return I("prelu")}elu(e){return I("elu")}eluDer(e,t){return I("eluDer")}selu(e){return I("selu")}int(e){return I("int")}clip(e,t,n){return I("clip")}abs(e){return I("abs")}complexAbs(e){return I("complexAbs")}sigmoid(e){return I("sigmoid")}softplus(e){return I("softplus")}sin(e){return I("sin")}cos(e){return I("cos")}tan(e){return I("tan")}asin(e){return I("asin")}acos(e){return I("acos")}atan(e){return I("atan")}atan2(e,t){return I("atan2")}sinh(e){return I("sinh")}cosh(e){return I("cosh")}tanh(e){return I("tanh")}asinh(e){return I("asinh")}acosh(e){return I("acosh")}atanh(e){return I("atanh")}erf(e){return I("erf")}step(e,t){return I("step")}fusedConv2d({input:e,filter:t,convInfo:n,bias:r,activation:i,preluActivationWeights:o}){return I("fusedConv2d")}conv2d(e,t,n){return I("conv2d")}conv2dDerInput(e,t,n){return I("conv2dDerInput")}conv2dDerFilter(e,t,n){return I("conv2dDerFilter")}fusedDepthwiseConv2D({input:e,filter:t,convInfo:n,bias:r,activation:i,preluActivationWeights:o}){return I("fusedDepthwiseConv2D")}depthwiseConv2D(e,t,n){return I("depthwiseConv2D")}depthwiseConv2DDerInput(e,t,n){return I("depthwiseConv2DDerInput")}depthwiseConv2DDerFilter(e,t,n){return I("depthwiseConv2DDerFilter")}conv3d(e,t,n){return I("conv3d")}conv3dDerInput(e,t,n){return I("conv3dDerInput")}conv3dDerFilter(e,t,n){return I("conv3dDerFilter")}maxPool(e,t){return I("maxPool")}maxPoolBackprop(e,t,n,r){return I("maxPoolBackprop")}avgPool(e,t){return I("avgPool")}avgPoolBackprop(e,t,n){return I("avgPoolBackprop")}avgPool3d(e,t){return I("avgPool3d")}avgPool3dBackprop(e,t,n){return I("avgPool3dBackprop")}maxPool3d(e,t){return I("maxPool3d")}maxPool3dBackprop(e,t,n,r){return I("maxPool3dBackprop")}reshape(e,t){return I("reshape")}cast(e,t){return I("cast")}tile(e,t){return I("tile")}pad(e,t,n){return I("pad")}transpose(e,t){return I("transpose")}gather(e,t,n){return I("gather")}gatherND(e,t){return I("gatherND")}scatterND(e,t,n){return I("scatterND")}batchToSpaceND(e,t,n){return I("batchToSpaceND")}spaceToBatchND(e,t,n){return I("spaceToBatchND")}resizeBilinear(e,t,n,r){return I("resizeBilinear")}resizeBilinearBackprop(e,t,n){return I("resizeBilinearBackprop")}resizeNearestNeighbor(e,t,n,r){return I("resizeNearestNeighbor")}resizeNearestNeighborBackprop(e,t,n){return I("resizeNearestNeighborBackprop")}batchNorm(e,t,n,r,i,o){return I("batchNorm")}localResponseNormalization4D(e,t,n,r,i){return I("localResponseNormalization4D")}LRNGrad(e,t,n,r,i,o,a){return I("LRNGrad")}multinomial(e,t,n,r){return I("multinomial")}oneHot(e,t,n,r){return I("oneHot")}cumsum(e,t,n,r){return I("cumsum")}nonMaxSuppression(e,t,n,r,i){return I("nonMaxSuppression")}fft(e){return I("fft")}ifft(e){return I("ifft")}complex(e,t){return I("complex")}real(e){return I("real")}imag(e){return I("imag")}cropAndResize(e,t,n,r,i,o){return I("cropAndResize")}depthToSpace(e,t,n){return I("depthToSpace")}split(e,t,n){return I("split")}sparseToDense(e,t,n,r){return I("sparseToDense")}diag(e){return I("diag")}fill(e,t,n){return I("fill")}onesLike(e){return I("onesLike")}zerosLike(e){return I("zerosLike")}linspace(e,t,n){return I("linspace")}dispose(){return I("dispose")}}function I(e){throw new Error(`'${e}' not yet implemented or not found in the registry. Did you forget to import the kernel?`)}const Ox={kernelName:Mi,inputsToSave:["x"],gradFunc:(e,t)=>{const[n]=t;return{x:()=>v(e,vt(R(n,"float32"),-1))}}};const kx={kernelName:Ui,inputsToSave:["x"],gradFunc:(e,t)=>{const[n]=t;return{x:()=>{const r=G(R(n,"float32")),i=le(D(k(1),r));return Q(M(e,i))}}}};const Dx={kernelName:Wi,inputsToSave:["x"],gradFunc:(e,t)=>{const[n]=t;return{x:()=>{const r=le(D(G(R(n,"float32")),1));return M(e,r)}}}};const Fx={kernelName:Jn,inputsToSave:["a","b"],gradFunc:(e,t)=>{const[n,r]=t,i=H(n.shape,r.shape),o=()=>{let s=e;const c=ie(n.shape,i);return c.length>0&&(s=U(s,c)),x(s,n.shape)},a=()=>{let s=e;const c=ie(r.shape,i);return c.length>0&&(s=U(s,c)),x(s,r.shape)};return{a:o,b:a}}};const Mx={kernelName:qi,saveAllInputs:!0,gradFunc:(e,t)=>{const n={};return t.forEach((r,i)=>{n[i]=()=>e.clone()}),n}};const Ux={kernelName:Bi,inputsToSave:["x"],gradFunc:(e,t)=>{const[n]=t;return{x:()=>W(n)}}};const Wx={kernelName:Pi,inputsToSave:["x"],gradFunc:(e,t)=>{const[n]=t;return{x:()=>W(n)}}};const qx={kernelName:ji,inputsToSave:["x"],gradFunc:(e,t)=>{const[n]=t;return{x:()=>M(e,le(D(k(1),G(R(n,"float32")))))}}};const Bx={kernelName:Gi,inputsToSave:["x"],gradFunc:(e,t)=>{const[n]=t;return{x:()=>{const r=le(O(k(1),G(R(n,"float32"))));return M(e,r)}}}};const Px={kernelName:Yi,inputsToSave:["a","b"],gradFunc:(e,t)=>{const[n,r]=t,i=H(n.shape,r.shape),o=()=>{const s=O(G(n),G(r));let c=v(e,M(r,s));const p=ie(n.shape,i);return p.length>0&&(c=U(c,p)),x(c,n.shape)},a=()=>{const s=O(G(n),G(r));let c=Q(v(e,M(n,s)));const p=ie(r.shape,i);return p.length>0&&(c=U(c,p)),x(c,r.shape)};return{a:o,b:a}}};const jx={kernelName:$i,inputsToSave:["x"],gradFunc:(e,t)=>{const[n]=t;return{x:()=>M(e,O(G(R(n,"float32")),1))}}};const Gx={kernelName:Hi,inputsToSave:["x"],gradFunc:(e,t)=>{const[n]=t;return{x:()=>M(e,D(k(1),G(R(n,"float32"))))}}};function bk(e,t,n,r,i=[1,1,1],o,a){const s=h(e,"dy","avgPool3dBackprop"),c=h(t,"input","avgPool3dBackprop");let p=s,l=c,u=!1;c.rank===4&&(u=!0,p=x(s,[1,s.shape[0],s.shape[1],s.shape[2],s.shape[3]]),l=x(c,[1,c.shape[0],c.shape[1],c.shape[2],c.shape[3]])),f(p.rank===5,()=>`Error in avgPool3dBackprop: dy must be rank 5 but got rank ${p.rank}.`),f(l.rank===5,()=>`Error in avgPool3dBackprop: input must be rank 5 but got rank ${l.rank}.`),f(ae(r,i),()=>`Error in avgPool3dBackprop: Either strides or dilations must be 1. Got strides ${r} and dilations '${i}'`),a!=null&&f(J(o),()=>`Error in maxPool3dBackprop: pad must be an integer when using, dimRoundingMode ${a} but got pad ${o}.`);const m=L=>{const T=on(l.shape,n,r,i,o,a);return L.avgPool3dBackprop(p,l,T)},b={dy:p,input:l},y={filterSize:n,strides:r,dilations:i,pad:o,dimRoundingMode:a},w=g.runKernelFunc(m,b,null,Pu,y);return u?x(w,[w.shape[1],w.shape[2],w.shape[3],w.shape[4]]):w}const $x=d({avgPool3dBackprop_:bk});const Hx={kernelName:Vi,inputsToSave:["x"],gradFunc:(e,t,n)=>{const[r]=t,{filterSize:i,strides:o,dilations:a,pad:s,dimRoundingMode:c}=n,p=a==null?[1,1,1]:a;return{x:()=>$x(e,r,i,o,p,s,c)}}};function wk(e,t,n,r,i){const o=h(e,"dy","avgPoolBackprop"),a=h(t,"input","avgPoolBackprop");f(a.rank===o.rank,()=>`Rank of input (${a.rank}) does not match rank of dy (${o.rank})`);let s=a,c=o,p=!1;a.rank===3&&(p=!0,s=x(a,[1,a.shape[0],a.shape[1],a.shape[2]]),c=x(o,[1,o.shape[0],o.shape[1],o.shape[2]])),f(c.rank===4,()=>`Error in avgPoolBackprop: dy must be rank 4 but got rank ${c.rank}.`),f(s.rank===4,()=>`Error in avgPoolBackprop: input must be rank 4 but got rank ${s.rank}.`);const l=y=>{const w=Ct(s.shape,n,r,1,i);return y.avgPoolBackprop(c,s,w)},u={dy:c,input:s},m={filterSize:n,strides:r,pad:i},b=g.runKernelFunc(l,u,null,Bu,m);return p?x(b,[b.shape[1],b.shape[2],b.shape[3]]):b}const Yx=d({avgPoolBackprop_:wk});const zx={kernelName:zi,inputsToSave:["x"],gradFunc:(e,t,n)=>{const[r]=t,{filterSize:i,strides:o,pad:a}=n;return{x:()=>Yx(e,r,i,o,a)}}};const Vx={kernelName:Ki,inputsToSave:["a","b"],gradFunc:(e,t,n)=>{const[r,i]=t,{transposeA:o,transposeB:a}=n;return!o&&!a?{a:()=>K(e,i,!1,!0),b:()=>K(r,e,!0,!1)}:!o&&a?{a:()=>K(e,i,!1,!1),b:()=>K(e,r,!0,!1)}:o&&!a?{a:()=>K(i,e,!1,!0),b:()=>K(r,e,!1,!1)}:{a:()=>K(i,e,!0,!0),b:()=>K(e,r,!0,!0)}}};const Kx={kernelName:Ji,gradFunc:(e,t,n)=>{const{blockShape:r,crops:i}=n;return{x:()=>Un(e,r,i)}}};const Jx={kernelName:Xi,gradFunc:(e,t,n)=>{const r=n,i=r.inputShape,o=r.shape,a=Array.from(o);for(let c=i.length-1;c>=0;c--)if(i[c]===o[c])a[c]=1;else if(i[c]!==1)throw new Error(`broadcastTo(): [${i}] cannot be broadcast to [${o}].`);const s=[];for(let c=0;c<a.length;c++)a[c]>1&&s.push(c);return{x:()=>U(e,s,!0)}}};const Xx={kernelName:Xn,gradFunc:e=>({x:()=>e.clone()})};const Zx={kernelName:Zi,gradFunc:e=>({x:()=>W(e)})};const Qx={kernelName:Qi,inputsToSave:["x"],gradFunc:(e,t,n)=>{const[r]=t,{clipValueMin:i,clipValueMax:o}=n;return{x:()=>Ae(ft(ct(r,i),at(r,o)),e,W(e))}}};const eL={kernelName:eo,saveAllInputs:!0,gradFunc:(e,t,n)=>{const r=t.map(c=>c.shape),{axis:i}=n,o=z(i,t[0].shape)[0],a=r.map(c=>c[o]),s=Ot(e,a,o);return s.map(c=>()=>c)}};const tL={kernelName:to,inputsToSave:["x","filter"],gradFunc:(e,t,n)=>{const[r,i]=t,{dilations:o,strides:a,pad:s,dataFormat:c}=n;return f(ut(o),()=>`Error in gradient of conv2D: dilation rates greater than 1 are not yet supported in gradients. Got dilations '${o}'`),{x:()=>ai(r.shape,e,i,a,s,c),filter:()=>ci(r,e,i.shape,a,s,c)}}};const nL={kernelName:no,inputsToSave:["dy","filter"],gradFunc:(e,t,n)=>{const[r,i]=t,{strides:o,pad:a,dataFormat:s,dimRoundingMode:c}=n;return{dy:()=>It(e,i,o,a,s,1,c),filter:()=>ci(e,r,i.shape,o,a,s,c)}}};function yk(e,t,n,r,i){let o=e;e.rank===4&&(o=x(e,[1,e.shape[0],e.shape[1],e.shape[2],e.shape[3]]));let a=t;a.rank===4&&(a=x(t,[1,t.shape[0],t.shape[1],t.shape[2],t.shape[3]])),f(o.rank===5,()=>`Error in conv3dDerFilter: input must be rank 5, but got shape ${o.shape}.`),f(a.rank===5,()=>`Error in conv3dDerFilter: dy must be rank 5, but got shape ${a.shape}.`),f(n.length===5,()=>`Error in conv3dDerFilter: filterShape must be length 5, but got ${n}.`),f(o.shape[4]===n[3],()=>`Error in conv3dDerFilter: depth of input ${o.shape[4]}) must match input depth in filter (${n[3]}.`),f(a.shape[4]===n[4],()=>`Error in conv3dDerFilter: depth of dy (${a.shape[4]}) must match output depth for filter (${n[4]}).`);const s=l=>{const u=1,m=Tn(o.shape,n,r,u,i);return l.conv3dDerFilter(o,a,m)},c={x:o,y:a},p={strides:r,pad:i};return g.runKernelFunc(s,c,null,$u,p)}const rL=d({conv3DBackpropFilter_:yk});const iL={kernelName:ro,inputsToSave:["x","filter"],gradFunc:(e,t,n)=>{const{dilations:r,strides:i,pad:o}=n;f(ut(r),()=>`Error in gradient of conv3D: dilation rates greater than 1 are not yet supported in gradients. Got dilations '${r}'`);const[a,s]=t;return{x:()=>Fp(a.shape,e,s,i,o),filter:()=>rL(a,e,s.shape,i,o)}}};const oL={kernelName:io,inputsToSave:["x"],gradFunc:(e,t)=>{const[n]=t;return{x:()=>v(Q(wi(R(n,"float32"))),e)}}};const sL={kernelName:oo,inputsToSave:["x"],gradFunc:(e,t)=>{const[n]=t;return{x:()=>v(bi(R(n,"float32")),e)}}};const aL={kernelName:so,inputsToSave:["x"],gradFunc:(e,t,n)=>{const[r]=t,{axis:i,exclusive:o,reverse:a}=n;return{x:()=>{const s=he([i],r.rank);let c=Oi(e,i,o,!a);return s!=null&&(c=X(c,s)),c}}}};const cL={kernelName:ao,inputsToSave:["x","filter"],gradFunc:(e,t,n)=>{const{dilations:r,strides:i,pad:o,dimRoundingMode:a}=n,s=r==null?[1,1]:r;f(ut(s),()=>`Error in gradient of depthwiseConv2dNative: dilation rates greater than 1 are not yet supported. Got dilations '${s}'`);const[c,p]=t;f(c.rank===4,()=>`Error in gradient of depthwiseConv2dNative: input must be rank 4, but got rank ${c.rank}.`),f(p.rank===4,()=>`Error in gradient of depthwiseConv2dNative: filter must be rank 4, but got rank ${p.rank}.`),f(c.shape[3]===p.shape[2],()=>`Error in gradient of depthwiseConv2d: number of input channels (${c.shape[3]}) must match the inChannels dimension in filter ${p.shape[2]}.`),f(ae(i,s),()=>`Error in gradient of depthwiseConv2d: Either strides or dilations must be 1. Got strides ${i} and dilations '${s}'.`),a!=null&&f(J(o),()=>`Error in depthwiseConv2d: pad must be an integer when using, dimRoundingMode ${a} but got pad ${o}.`);const l=We(c.shape,p.shape,i,s,o,a,!0);return{x:()=>jp(c.shape,e,p,l),filter:()=>Pp(c,e,p.shape,l)}}};const pL={kernelName:co,inputsToSave:["x","filter"],gradFunc:(e,t,n)=>{const[r,i]=t,o={x:r,filter:i,dy:e},a={x:r,filter:i,dy:e};return{x:()=>g.runKernel(Xu,o,n),filter:()=>g.runKernel(Zu,a,n)}}};const lL={kernelName:po,inputsToSave:["a","b"],gradFunc:(e,t)=>{const[n,r]=t,i=H(n.shape,r.shape),o=()=>{const s=M(e,R(r,"float32")),c=ie(n.shape,i);return c.length>0?x(U(s,c),n.shape):s},a=()=>{let s=v(e,R(n,"float32"));const c=ie(r.shape,i);c.length>0&&(s=x(U(s,c),r.shape));const p=G(r);return Q(M(s,R(p,"float32")))};return{a:o,b:a}}};const uL={kernelName:lo,outputsToSave:[!0],gradFunc:(e,t)=>{const[n]=t,r=o=>o.eluDer(e,n),i={dy:e,y:n};return{x:()=>g.runKernelFunc(r,i,null,Qu)}}};const hL={kernelName:uo,inputsToSave:["x"],gradFunc:(e,t)=>{const[n]=t,r=v(Ne(Q(G(n))),2/Math.sqrt(Math.PI));return{x:()=>v(e,r)}}};const dL={kernelName:ho,outputsToSave:[!0],gradFunc:(e,t)=>{const[n]=t;return{x:()=>v(e,n)}}};const mL={kernelName:mo,inputsToSave:["x"],gradFunc:(e,t)=>{const[n]=t;return{x:()=>v(e,Ne(n))}}};const fL={kernelName:fo,gradFunc:e=>({x:()=>W(e)})};const gL={kernelName:go,inputsToSave:["a","b"],gradFunc:(e,t)=>{const[n,r]=t,i=H(n.shape,r.shape),o=()=>{const s=M(e,R(r,"float32")),c=ie(n.shape,i);return c.length>0?x(U(s,c),n.shape):s},a=()=>{let s=v(e,R(n,"float32"));const c=ie(r.shape,i);c.length>0&&(s=x(U(s,c),r.shape));const p=G(r);return Q(M(s,R(p,"float32")))};return{a:o,b:a}}};const bL={kernelName:bo,inputsToSave:["x","mean","variance","scale"],gradFunc:(e,t,n)=>{const{varianceEpsilon:r}=n,[i,o,a,s]=t,c=s==null?k(1):s,p=ie(o.shape,i.shape),l=[];if(o.rank===1){for(let N=0;N<i.shape.length-1;++N)l.push(i.shape[N]);l.push(1)}const u=D(i,o),m=v(e,c),b=yi(O(a,k(r))),y=v(v(v(b,b),b),k(-.5)),w=()=>o.rank===1?x(v(v(e,Ht(x(b,[1,1,1,o.shape[0]]),l)),c),i.shape):x(v(v(e,b),c),i.shape),L=()=>{let N=v(v(b,k(-1)),m);return o.rank===1&&(N=U(N,p)),x(N,o.shape)},T=()=>{let N=v(v(y,u),m);return o.rank===1&&(N=U(N,p)),x(N,o.shape)},A=()=>{const N=v(u,b);let C=v(e,N);return o.rank===1&&(C=U(C,p)),x(C,o.shape)},_=()=>{let N=e;return o.rank===1&&(N=U(N,p)),x(N,o.shape)};return{x:w,mean:L,variance:T,scale:A,offset:_}}};const xL={kernelName:wo,inputsToSave:["x","indices"],gradFunc:(e,t,n)=>{const[r,i]=t,{axis:o}=n,a=z(o,r.shape)[0],s=()=>{const c=r.shape,p=i.size,l=c.slice(0,a),u=l.length,m=c.slice(o,c.length).slice(1),b=m.length,y=wL(0,u),w=wL(u+1,u+1+b),L=yL([l,[p],m]),T=x(e,L),A=x(i,[p]),_=yL([[u],y,w]),N=X(T,_);let C=mi(N,A,r.shape[a]);const F=pr(_);return C=X(C,F),C};return{x:s,indices:()=>i}}};function wL(e,t){const n=[];for(let r=e;r<t;++r)n.push(r);return n}function yL(e){const t=[];for(let n=0;n<e.length;++n)for(let r=0;r<e[n].length;++r)t.push(e[n][r]);return t}const LL={kernelName:yo,inputsToSave:["a","b"],gradFunc:(e,t)=>{const[n,r]=t;return{a:()=>W(n),b:()=>W(r)}}};const vL={kernelName:xo,gradFunc:e=>({x:()=>R(e,"float32")})};const SL={kernelName:Lo,gradFunc:e=>({x:()=>W(e)})};const TL={kernelName:vo,gradFunc:e=>({x:()=>W(e)})};const IL={kernelName:So,gradFunc:e=>({x:()=>W(e)})};const AL={kernelName:Io,inputsToSave:["x"],gradFunc:(e,t)=>{const[n]=t;return{x:()=>M(e,O(n,1))}}};const NL={kernelName:To,inputsToSave:["x"],gradFunc:(e,t)=>{const[n]=t;return{x:()=>M(e,R(n,"float32"))}}};const _L={kernelName:Ao,inputsToSave:[],outputsToSave:[!0],gradFunc:(e,t,n)=>{const[r]=t,{axis:i}=n;return{logits:()=>{const o=!0,a=Ne(r);return D(e,v(U(e,i,o),a))}}}};function xk(e,t,n,r=5,i=1,o=1,a=.5){const s=l=>l.LRNGrad(n,e,t,r,i,o,a),c={x:e,y:t,dy:n},p={depthRadius:r,bias:i,alpha:o,beta:a};return g.runKernelFunc(s,c,null,mh,p)}const RL=d({localResponseNormalizationBackprop_:xk});const CL={kernelName:No,inputsToSave:["x"],outputsToSave:[!0],gradFunc:(e,t,n)=>{const[r,i]=t,{depthRadius:o,bias:a,alpha:s,beta:c}=n;return{x:()=>RL(r,i,e,o,a,s,c)}}};function Jp(e,t,n,r,i){return t.rank<n.rank&&(t=x(t,fe(t.shape,r))),e.rank<n.rank&&(e=x(e,fe(e.shape,r))),{x:()=>{const o=v(e,R(Kt(n,t),e.dtype));return i==null?o:X(o,i)}}}const sf={kernelName:_o,inputsToSave:["x"],outputsToSave:[!0],gradFunc:(e,t,n)=>{const r=n,{reductionIndices:i}=r,[o,a]=t,s=z(i,o.shape),c=he(s,o.rank),p=Jp(e,a,o,s,c);return{x:()=>{let l=p.x();return c!=null&&(l=X(l)),l}}}};const EL={kernelName:Ro,inputsToSave:["a","b"],gradFunc:(e,t)=>{const[n,r]=t,i=()=>v(e,R(ct(n,r),"float32")),o=()=>v(e,R(jn(n,r),"float32"));return{a:i,b:o}}};function Lk(e,t,n,r,i,o=[1,1,1],a,s){const c=h(e,"dy","maxPool3dBackprop"),p=h(t,"input","maxPool3dBackprop"),l=h(n,"output","maxPool3dBackprop");let u=c,m=p,b=l,y=!1;p.rank===4&&(y=!0,u=x(c,[1,c.shape[0],c.shape[1],c.shape[2],c.shape[3]]),m=x(p,[1,p.shape[0],p.shape[1],p.shape[2],p.shape[3]]),b=x(l,[1,l.shape[0],l.shape[1],l.shape[2],l.shape[3]])),f(u.rank===5,()=>`Error in maxPool3dBackprop: dy must be rank 5 but got rank ${u.rank}.`),f(m.rank===5,()=>`Error in maxPool3dBackprop: input must be rank 5 but got rank ${m.rank}.`),f(b.rank===5,()=>`Error in maxPool3dBackprop: output must be rank 5 but got rank ${b.rank}.`),f(ae(i,o),()=>`Error in maxPool3dBackprop: Either strides or dilations must be 1. Got strides ${i} and dilations '${o}'`),s!=null&&f(J(a),()=>`Error in maxPool3dBackprop: pad must be an integer when using, dimRoundingMode ${s} but got pad ${a}.`);const w=_=>{const N=on(m.shape,r,i,o,a,s);return _.maxPool3dBackprop(u,m,b,N)},L={dy:u,input:m,output:b},T={filterSize:r,strides:i,dilations:o,pad:a,dimRoundingMode:s},A=g.runKernelFunc(w,L,null,gh,T);return y?x(A,[A.shape[1],A.shape[2],A.shape[3],A.shape[4]]):A}const OL=d({maxPool3dBackprop_:Lk});const kL={kernelName:Eo,inputsToSave:["x"],outputsToSave:[!0],gradFunc:(e,t,n)=>{const[r,i]=t,{filterSize:o,strides:a,dilations:s,pad:c,dimRoundingMode:p}=n,l=s==null?[1,1,1]:s;return{x:()=>OL(e,r,i,o,a,l,c,p)}}};function vk(e,t,n,r,i,o,a){const s=h(e,"dy","maxPoolBackprop"),c=h(t,"input","maxPoolBackprop"),p=h(n,"output","maxPoolBackprop");f(c.rank===s.rank,()=>`Rank of input (${c.rank}) does not match rank of dy (${s.rank})`),f(s.rank===4,()=>`Error in maxPoolBackprop: dy must be rank 4 but got rank ${s.rank}.`),f(c.rank===4,()=>`Error in maxPoolBackprop: input must be rank 4 but got rank ${c.rank}.`),a!=null&&f(J(o),()=>`Error in maxPoolBackprop: pad must be an integer when using, dimRoundingMode ${a} but got pad ${o}.`);const l=b=>{const y=Ct(c.shape,r,i,1,o,a);return b.maxPoolBackprop(s,c,p,y)},u={dy:s,input:c,output:p},m={filterSize:r,strides:i,pad:o,dimRoundingMode:a};return g.runKernelFunc(l,u,null,fh,m)}const DL=d({maxPoolBackprop_:vk});const FL={kernelName:Co,inputsToSave:["x"],outputsToSave:[!0],gradFunc:(e,t,n)=>{const[r,i]=t,{filterSize:o,strides:a,pad:s}=n;return{x:()=>DL(e,r,i,o,a,s)}}};const ML={kernelName:Oo,inputsToSave:["x"],outputsToSave:[!0],gradFunc:(e,t,n)=>{const r=n,{axis:i}=r,[o,a]=t,s=z(i,o.shape),c=he(s,o.rank),p=Jp(e,a,o,s,c);return{x:()=>{let l=p.x();return c!=null&&(l=X(l)),l}}}};const UL={kernelName:ko,inputsToSave:["a","b"],gradFunc:(e,t)=>{const[n,r]=t,i=()=>v(e,R(at(n,r),"float32")),o=()=>v(e,R(Ve(n,r),"float32"));return{a:i,b:o}}};const WL={kernelName:Do,inputsToSave:["a","b"],gradFunc:(e,t)=>{const[n,r]=t,i=H(n.shape,r.shape),o=()=>{const s=ie(n.shape,i);return s.length>0?x(U(e,s),n.shape):e},a=()=>{const s=v(e,Q($n(M(n,r)))),c=ie(r.shape,i);return c.length>0?x(U(s,c),r.shape):s};return{a:o,b:a}}};const qL={kernelName:Fo,inputsToSave:["a","b"],gradFunc:(e,t)=>{const[n,r]=t,i=H(n.shape,r.shape),o=()=>{const s=v(e,R(r,"float32")),c=ie(n.shape,i);return c.length>0?x(U(s,c),n.shape):s},a=()=>{const s=v(e,R(n,"float32")),c=ie(r.shape,i);return c.length>0?x(U(s,c),r.shape):s};return{a:o,b:a}}};const BL={kernelName:Mo,gradFunc:e=>({x:()=>Q(e)})};const PL={kernelName:Wo,inputsToSave:["indices"],gradFunc:(e,t)=>{const n=t[0];return{indices:()=>Ue(n.shape,"float32")}}};const jL={kernelName:Uo,gradFunc:e=>({x:()=>W(e)})};const af={kernelName:qo,inputsToSave:["x"],gradFunc:(e,t,n)=>{const r=t[0],{paddings:i}=n,o=i.map(a=>a[0]);return{x:()=>Y(e,o,r.shape)}}};const GL={kernelName:Bo,inputsToSave:["a","b"],outputsToSave:[!0],gradFunc:(e,t)=>{const[n,r,i]=t,o=n,a=r,s=H(o.shape,a.shape),c=()=>{const l=R(a,"float32");let u=v(e,v(l,je(o,D(l,k(1)))));const m=ie(o.shape,s);return m.length>0&&(u=U(u,m)),x(u,o.shape)},p=()=>{const l=Ve(o,0),u=Ae(l,gt(o),W(o));let m=v(e,v(i,u));const b=ie(a.shape,s);return b.length>0&&(m=U(m,b)),x(m,a.shape)};return{a:c,b:p}}};const $L={kernelName:Po,inputsToSave:["x","alpha"],gradFunc:(e,t)=>{const[n,r]=t,i=Ve(n,0);return{x:()=>Ae(i,e,v(e,r)),alpha:()=>{let o=Ae(i,W(e),v(e,n));const a=ie(r.shape,e.shape);return a.length>0&&(o=U(o,a)),x(o,r.shape)}}}};const HL={kernelName:jo,inputsToSave:["x"],gradFunc:(e,t)=>{const[n]=t;return{x:()=>M(e,Q(G(n)))}}};const YL={kernelName:zo,inputsToSave:["x"],gradFunc:(e,t)=>{const[n]=t,r=v(at(n,6),vt(n));return{x:()=>v(e,R(r,"float32"))}}};const zL={kernelName:Go,inputsToSave:["x"],gradFunc:(e,t)=>{const[n]=t;return{x:()=>v(e,R(vt(n),"float32"))}}};const VL={kernelName:$o,inputsToSave:["x"],gradFunc:(e,t)=>{const[n]=t;return{x:()=>x(e,n.shape)}}};const KL={kernelName:Yo,inputsToSave:["images"],gradFunc:(e,t,n)=>{const[r]=t,i=s=>{const{alignCorners:c}=n;return s.resizeBilinearBackprop(e,r,c)},o={images:r},a=()=>g.runKernelFunc(i,o,null,Ah,n);return{images:a}}};const JL={kernelName:Ho,inputsToSave:["images"],gradFunc:(e,t,n)=>{const[r]=t,i=s=>{const{alignCorners:c}=n;return s.resizeNearestNeighborBackprop(e,r,c)},o={images:r},a=()=>g.runKernelFunc(i,o,null,Ih,n);return{images:a}}};const XL={kernelName:Vo,gradFunc:(e,t,n)=>{const{dims:r}=n,i=z(r,e.shape);return{x:()=>Pe(e,i)}}};const ZL={kernelName:Ko,gradFunc:e=>({x:()=>W(e)})};const QL={kernelName:Jo,inputsToSave:["x"],gradFunc:(e,t)=>{const[n]=t;return{x:()=>Q(M(e,v(je(n,1.5),2)))}}};const ev={kernelName:Xo,inputsToSave:["condition"],gradFunc:(e,t)=>{const[n]=t;return{condition:()=>R(W(n),"float32"),t:()=>v(e,R(n,e.dtype)),e:()=>v(e,R(Pn(n),e.dtype))}}};const tv={kernelName:Zo,inputsToSave:["x"],gradFunc:(e,t)=>{const[n]=t;return{x:()=>{const r=Ve(n,k(0)),i=k(rf),o=k(of),a=v(e,o),s=v(v(e,i),Ne(R(n,"float32")));return Ae(r,a,s)}}}};const nv={kernelName:rs,outputsToSave:[!0],gradFunc:(e,t)=>{const[n]=t;return{x:()=>v(e,v(n,D(k(1),n)))}}};const rv={kernelName:ns,gradFunc:e=>({x:()=>W(e)})};const iv={kernelName:es,inputsToSave:["x"],gradFunc:(e,t)=>{const[n]=t;return{x:()=>v(Vn(R(n,"float32")),e)}}};const ov={kernelName:ts,inputsToSave:["x"],gradFunc:(e,t)=>{const[n]=t;return{x:()=>v(ki(R(n,"float32")),e)}}};const sv={kernelName:Qo,inputsToSave:["x"],gradFunc:(e,t,n)=>{const[r]=t,{begin:i,size:o}=n,a=r.shape,[s,c]=Ys(r,i,o),p=[];for(let l=0;l<e.rank;l++)p.push([s[l],a[l]-s[l]-c[l]]);return{x:()=>ot(e,p)}}};const av={kernelName:ps,outputsToSave:[!0],gradFunc:(e,t,n)=>{const[r]=t,{dim:i}=n,o=!0,a=v(e,r);return{logits:()=>D(a,v(U(a,[i],o),r))}}};const cv={kernelName:is,inputsToSave:["x"],gradFunc:(e,t)=>{const[n]=t;return{x:()=>v(e,St(n))}}};const cf={kernelName:as,gradFunc:(e,t,n)=>{const{blockShape:r,paddings:i}=n;return{x:()=>Kn(e,r,i)}}};const pf={kernelName:cs,gradFunc:(e,t,n)=>{const{axis:r}=n;return{x:()=>re(e,r)}}};const pv={kernelName:os,inputsToSave:["x"],gradFunc:(e,t)=>{const[n]=t;return{x:()=>M(e,v(le(R(n,"float32")),2))}}};const lv={kernelName:_h,inputsToSave:["x"],gradFunc:(e,t)=>{const[n]=t;return{x:()=>v(e,v(R(n,"float32"),2))}}};const uv={kernelName:ls,inputsToSave:["a","b"],gradFunc:(e,t)=>{const[n,r]=t,i=k(2),o=()=>v(e,v(i,D(n,r))),a=()=>v(e,v(i,D(r,n)));return{a:o,b:a}}};const hv={kernelName:ys,gradFunc:e=>({x:()=>W(e)})};const dv={kernelName:us,inputsToSave:["a","b"],gradFunc:(e,t)=>{const[n,r]=t,i=H(n.shape,r.shape),o=()=>{let s=e;const c=ie(n.shape,i);return c.length>0&&(s=U(s,c)),x(s,n.shape)},a=()=>{let s=e;const c=ie(r.shape,i);return c.length>0&&(s=U(s,c)),x(Q(s),r.shape)};return{a:o,b:a}}};const mv={kernelName:ss,inputsToSave:["x"],gradFunc:(e,t,n)=>{const[r]=t,i=r.shape.slice(),{axis:o}=n,a=z(o,r.shape);a.forEach(p=>{i[p]=1});const s=x(e,i),c=v(s,mt(r.shape,"float32"));return{x:()=>c}}};const fv={kernelName:hs,inputsToSave:["x"],gradFunc:(e,t)=>{const[n]=t;return{x:()=>M(e,G(Vn(n)))}}};const gv={kernelName:ds,outputsToSave:[!0],gradFunc:(e,t)=>{const[n]=t;return{x:()=>v(D(k(1),G(n)),e)}}};const bv={kernelName:ms,inputsToSave:["x"],gradFunc:(e,t,n)=>{const[r]=t,{reps:i}=n,o=()=>{let a=W(r);if(r.rank===1)for(let s=0;s<i[0];++s)a=O(a,Y(e,[s*r.shape[0]],[r.shape[0]]));else if(r.rank===2)for(let s=0;s<i[0];++s)for(let c=0;c<i[1];++c)a=O(a,Y(e,[s*r.shape[0],c*r.shape[1]],[r.shape[0],r.shape[1]]));else if(r.rank===3)for(let s=0;s<i[0];++s)for(let c=0;c<i[1];++c)for(let p=0;p<i[2];++p)a=O(a,Y(e,[s*r.shape[0],c*r.shape[1],p*r.shape[2]],[r.shape[0],r.shape[1],r.shape[2]]));else if(r.rank===4)for(let s=0;s<i[0];++s)for(let c=0;c<i[1];++c)for(let p=0;p<i[2];++p)for(let l=0;l<i[3];++l)a=O(a,Y(e,[s*r.shape[0],c*r.shape[1],p*r.shape[2],l*r.shape[3]],[r.shape[0],r.shape[1],r.shape[2],r.shape[3]]));else throw new Error(`Gradient for tile operation is not implemented for rank-${r.rank} tensors yet.`);return a};return{x:o}}};const wv={kernelName:fs,gradFunc:(e,t,n)=>{const r=n,{perm:i}=r,o=pr(i);return{x:()=>X(e,o)}}};const yv={kernelName:gs,gradFunc:(e,t,n)=>{const r=n,{axis:i}=r;return{value:()=>dt(e,i)}}};const xv={kernelName:bs,inputsToSave:["segmentIds"],gradFunc:(e,t)=>{const[n]=t,r=()=>Sk(e,n);return{x:r}}};function Sk(e,t){const n=Tt(t,W(t)),r=Gn(e,n);let i=ct(t,k(0,"int32"));const o=r.rank-i.rank;for(let s=0;s<o;++s)i=Ke(i,s+1);i=ft(i,mt(r.shape,"bool"));const a=W(r);return Ae(i,r,a)}const Lv={kernelName:ws,gradFunc:e=>({x:()=>W(e)})};const Tk=[Ox,kx,Dx,Fx,Mx,Ux,Wx,qx,Bx,Px,jx,Gx,Hx,zx,Vx,Kx,Jx,Xx,Zx,Qx,eL,nL,tL,iL,oL,sL,aL,cL,pL,lL,uL,hL,dL,mL,gL,fL,bL,xL,LL,vL,SL,TL,IL,AL,NL,_L,CL,sf,sf,EL,kL,FL,ML,UL,WL,qL,BL,PL,jL,af,af,GL,$L,HL,YL,zL,VL,KL,JL,XL,ZL,QL,ev,tv,nv,rv,iv,ov,sv,av,cv,cf,cf,pf,pf,pv,uv,lv,hv,dv,mv,fv,gv,bv,wv,yv,xv,Lv];for(const e of Tk)Uu(e);return dy();})();
/*! *****************************************************************************
Copyright (c) Microsoft Corporation.
Permission to use, copy, modify, and/or distribute this software for any
purpose with or without fee is hereby granted.
THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES WITH
REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY
AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY SPECIAL, DIRECT,
INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM
LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR
OTHER TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR
PERFORMANCE OF THIS SOFTWARE.
***************************************************************************** */
/**
* @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. 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. 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. */
//# sourceMappingURL=face-api.min.js.map
Reference in New Issue View Git Blame Copy Permalink