/*! For license information please see 3919.411019e5423960c4130a.js.LICENSE.txt */
(self.webpackChunkfindagrave_frontend=self.webpackChunkfindagrave_frontend||[]).push([[3919,4236,6165,9474],{40263(e){function t(e){if(e=Object.assign({},e),!(this instanceof t))throw new Error("MapboxLanguage needs to be called with the new keyword");this.setLanguage=this.setLanguage.bind(this),this._initialStyleUpdate=this._initialStyleUpdate.bind(this),this._defaultLanguage=e.defaultLanguage,this._isLanguageField=e.languageField||/^name_/,this._getLanguageField=e.getLanguageField||function(e){return"mul"===e?"name":`name_${e}`},this._languageSource=e.languageSource||null,this._languageTransform=e.languageTransform,this._excludedLayerIds=e.excludedLayerIds||[],this.supportedLanguages=e.supportedLanguages||["ar","de","en","es","fr","it","ja","ko","mul","pt","ru","vi","zh-Hans","zh-Hant"]}const n=/^\{name/;function r(e,t){const r=Array.isArray(t)&&"get"===t[0];return r&&n.test(t[1])&&console.warn("This plugin no longer supports the use of token syntax (e.g. {name}). Please use a get expression. See https://docs.mapbox.com/mapbox-gl-js/style-spec/expressions/ for more details."),r&&e.test(t[1])}function a(e,t,n){if(Array.isArray(t))for(let s=1;s<t.length;s++)Array.isArray(t[s])&&(r(e,t[s])&&(t[s][1]=n),a(e,t[s],n))}function s(e,t,n){if(r(e,t)&&(t[1]=n),a(e,t,n),"get"===t[0]&&"name"===t[1]){const e=t.slice();t=["coalesce",["get",n],e]}return t}t.prototype.setLanguage=function(e,t){if(this.supportedLanguages.indexOf(t)<0)throw new Error(`Language ${t} is not supported`);const n=this._languageSource||function(e){const t=Object.keys(e.sources).filter(t=>{const n=e.sources[t].url;return n&&n.indexOf("mapbox.mapbox-streets-v8")>-1||/mapbox-streets-v[1-9][1-9]/.test(n)});if(!t.length)throw new Error('If using MapboxLanguage with a Mapbox style, the style must be based on vector tile version 8, e.g. "streets-v11"');return t[0]}(e);if(!n)return e;const r=this._getLanguageField(t),a=this._isLanguageField,i=this._excludedLayerIds,o=e.layers.map(e=>e.source===n?function(e,t,n,r){return t.layout&&t.layout["text-field"]&&-1===r.indexOf(t.id)?Object.assign({},t,{layout:Object.assign({},t.layout,{"text-field":s(e,t.layout["text-field"],n)})}):t}(a,e,r,i):e),l=Object.assign({},e,{layers:o});return this._languageTransform?this._languageTransform(l,t):l},t.prototype._initialStyleUpdate=function(){const e=this._map.getStyle(),t=this._defaultLanguage||function(e){const t=navigator.languages?navigator.languages[0]:navigator.language||navigator.userLanguage,n=t&&t.split("-");let r=t;n.length>1&&(r=n[0]);if(e.indexOf(r)>-1)return r;return null}(this.supportedLanguages);this._map.setStyle(this.setLanguage(e,t))},t.prototype.onAdd=function(e){return this._map=e,this._map.on("style.load",this._initialStyleUpdate),this._container=document.createElement("div"),this._container},t.prototype.onRemove=function(){this._map.off("style.load",this._initialStyleUpdate),this._map=void 0},void 0!==e.exports?e.exports=t:window.MapboxLanguage=t},81909(e,t,n){"use strict";var r="/index.js";n.d(t,{$2:()=>NY,B0:()=>n0,Lz:()=>g0,Np:()=>MQ,R:()=>m0,ex:()=>ZQ});var a=Object.defineProperty,s=n(67291),i=(e,t)=>{for(var n in t)a(e,n,{get:t[n],enumerable:!0})};i({},{Abs:()=>Ve,Acos:()=>Ue,Acosh:()=>Ge,AdadeltaOptimizer:()=>Ac,AdagradOptimizer:()=>$c,AdamOptimizer:()=>Rc,AdamaxOptimizer:()=>Fc,Add:()=>He,AddN:()=>je,All:()=>qe,Any:()=>Ke,ArgMax:()=>Xe,ArgMin:()=>Ze,Asin:()=>Ye,Asinh:()=>Je,Atan:()=>Qe,Atan2:()=>tt,Atanh:()=>et,AvgPool:()=>nt,AvgPool3D:()=>at,AvgPool3DGrad:()=>st,AvgPoolGrad:()=>rt,BackendWasm:()=>lY,BatchMatMul:()=>it,BatchToSpaceND:()=>ot,Bincount:()=>lt,BitwiseAnd:()=>ut,BroadcastArgs:()=>dt,BroadcastTo:()=>ht,Callback:()=>iN,CallbackList:()=>Wv,Cast:()=>pt,Ceil:()=>ct,ClipByValue:()=>ft,Complex:()=>mt,ComplexAbs:()=>gt,Concat:()=>yt,Conv2D:()=>bt,Conv2DBackpropFilter:()=>xt,Conv2DBackpropInput:()=>vt,Conv3D:()=>wt,Conv3DBackpropFilterV2:()=>kt,Conv3DBackpropInputV2:()=>It,Cos:()=>St,Cosh:()=>_t,CropAndResize:()=>Ct,Cumprod:()=>Nt,Cumsum:()=>Tt,CustomCallback:()=>Gv,DataStorage:()=>L,DenseBincount:()=>Et,DepthToSpace:()=>At,DepthwiseConv2dNative:()=>$t,DepthwiseConv2dNativeBackpropFilter:()=>Rt,DepthwiseConv2dNativeBackpropInput:()=>Ft,Diag:()=>Dt,Dilation2D:()=>Mt,Dilation2DBackpropFilter:()=>Lt,Dilation2DBackpropInput:()=>Ot,Draw:()=>Pt,ENV:()=>ze,EarlyStopping:()=>uN,Einsum:()=>Bt,Elu:()=>Wt,EluGrad:()=>Vt,Environment:()=>Me,Equal:()=>Gt,Erf:()=>Ut,Exp:()=>Ht,ExpandDims:()=>jt,Expm1:()=>qt,FFT:()=>Kt,Fill:()=>Xt,FlipLeftRight:()=>Zt,Floor:()=>Yt,FloorDiv:()=>Jt,FromPixels:()=>na,FusedBatchNorm:()=>Qt,FusedConv2D:()=>sa,FusedDepthwiseConv2D:()=>ia,GPGPUContext:()=>iP,GatherNd:()=>tn,GatherV2:()=>en,GraphModel:()=>YT,Greater:()=>nn,GreaterEqual:()=>rn,History:()=>Uv,IFFT:()=>sn,Identity:()=>an,Imag:()=>on,InputSpec:()=>Ux,IsFinite:()=>ln,IsInf:()=>un,IsNan:()=>hn,KernelBackend:()=>P,LRN:()=>In,LRNGrad:()=>Sn,LayerVariable:()=>Bx,LayersModel:()=>Hw,LeakyRelu:()=>dn,Less:()=>pn,LessEqual:()=>cn,LinSpace:()=>fn,Log:()=>mn,Log1p:()=>gn,LogSoftmax:()=>wn,LogicalAnd:()=>yn,LogicalNot:()=>bn,LogicalOr:()=>xn,LogicalXor:()=>vn,LowerBound:()=>kn,MathBackendCPU:()=>gE,MathBackendWebGL:()=>wz,MatrixBandPart:()=>_n,Max:()=>Nn,MaxPool:()=>Cn,MaxPool3D:()=>An,MaxPool3DGrad:()=>$n,MaxPoolGrad:()=>En,MaxPoolWithArgmax:()=>Rn,Maximum:()=>Tn,Mean:()=>Fn,Min:()=>Dn,Minimum:()=>Mn,MirrorPad:()=>On,Mod:()=>Ln,MomentumOptimizer:()=>Mc,Multinomial:()=>Pn,Multiply:()=>zn,Neg:()=>Bn,NonMaxSuppressionV3:()=>Vn,NonMaxSuppressionV4:()=>Un,NonMaxSuppressionV5:()=>Gn,NotEqual:()=>Wn,OP_SCOPE_SUFFIX:()=>Ps,OneHot:()=>jn,OnesLike:()=>Hn,Optimizer:()=>Ec,OptimizerConstructors:()=>zf,Pack:()=>qn,PadV2:()=>Kn,Pool:()=>Xn,Pow:()=>Zn,Prelu:()=>Yn,Prod:()=>Jn,RMSPropOptimizer:()=>Oc,RNN:()=>Qk,RaggedGather:()=>Qn,RaggedRange:()=>er,RaggedTensorToTensor:()=>tr,Range:()=>nr,Rank:()=>as,Real:()=>rr,RealDiv:()=>zt,Reciprocal:()=>ar,Reduction:()=>Zp,Relu:()=>sr,Relu6:()=>dr,Reshape:()=>ir,ResizeBilinear:()=>ur,ResizeBilinearGrad:()=>hr,ResizeNearestNeighbor:()=>or,ResizeNearestNeighborGrad:()=>lr,Reverse:()=>pr,RotateWithOffset:()=>ra,Round:()=>cr,Rsqrt:()=>fr,SGDOptimizer:()=>Dc,ScatterNd:()=>mr,SearchSorted:()=>yr,Select:()=>br,Selu:()=>xr,Sequential:()=>Xw,Sigmoid:()=>Sr,Sign:()=>Ir,Sin:()=>wr,Sinh:()=>kr,Slice:()=>vr,Softmax:()=>Ar,Softplus:()=>_r,SpaceToBatchND:()=>Cr,SparseFillEmptyRows:()=>$r,SparseReshape:()=>Rr,SparseSegmentMean:()=>Fr,SparseSegmentSum:()=>Dr,SparseToDense:()=>Mr,SplitV:()=>Er,Sqrt:()=>Nr,Square:()=>Lr,SquaredDifference:()=>Or,StaticRegexReplace:()=>Pr,Step:()=>ta,StridedSlice:()=>zr,StringNGrams:()=>Br,StringSplit:()=>Wr,StringToHashBucketFast:()=>Vr,Sub:()=>Ur,Sum:()=>Tr,SymbolicTensor:()=>Gx,Tan:()=>Gr,Tanh:()=>Hr,Tensor:()=>ts,TensorBuffer:()=>Ja,TensorScatterUpdate:()=>gr,Tile:()=>jr,TopK:()=>qr,Transform:()=>Kr,Transpose:()=>Xr,Unique:()=>Zr,Unpack:()=>Yr,UnsortedSegmentSum:()=>Jr,UpperBound:()=>Qr,Variable:()=>rs,ZerosLike:()=>ea,_FusedMatMul:()=>aa,abs:()=>bo,acos:()=>xo,acosh:()=>vo,add:()=>fo,addN:()=>wo,all:()=>ko,any:()=>Io,argMax:()=>So,argMin:()=>_o,asin:()=>No,asinh:()=>To,atan:()=>Co,atan2:()=>Eo,atanh:()=>Ao,avgPool:()=>qo,avgPool3d:()=>Ko,backend:()=>hi,backend_util:()=>Gf,basicLSTMCell:()=>el,batchNorm:()=>rl,batchNorm2d:()=>al,batchNorm3d:()=>sl,batchNorm4d:()=>il,batchToSpaceND:()=>tl,bincount:()=>ol,bitwiseAnd:()=>ll,booleanMaskAsync:()=>Yd,broadcastArgs:()=>ul,broadcastTo:()=>hl,broadcast_util:()=>Dl,browser:()=>lf,buffer:()=>uo,callbacks:()=>pN,cast:()=>ho,ceil:()=>dl,clipByValue:()=>cl,clone:()=>po,complex:()=>Bs,concat:()=>Xo,concat1d:()=>fl,concat2d:()=>ml,concat3d:()=>gl,concat4d:()=>yl,constraints:()=>sv,conv1d:()=>xl,conv2d:()=>bl,conv2dTranspose:()=>wl,conv3d:()=>kl,conv3dTranspose:()=>Sl,copyRegisteredKernels:()=>ba,cos:()=>_l,cosh:()=>Nl,cosineWindow:()=>sp,cumprod:()=>Tl,cumsum:()=>Cl,customGrad:()=>Ou,data:()=>tC,denseBincount:()=>El,deprecationWarn:()=>Ks,depthToSpace:()=>Al,depthwiseConv2d:()=>$l,deregisterOp:()=>gN,device_util:()=>Cs,diag:()=>Rl,dilation2d:()=>Fl,disableDeprecationWarnings:()=>qs,dispose:()=>ei,disposeVariables:()=>Xs,div:()=>go,divNoNan:()=>Wl,dot:()=>Vl,dropout:()=>rp,einsum:()=>Ul,elu:()=>Gl,enableDebugMode:()=>js,enableProdMode:()=>Hs,enclosingPowerOfTwo:()=>ap,engine:()=>Zs,ensureShape:()=>Hl,env:()=>Le,equal:()=>Pl,erf:()=>jl,euclideanNorm:()=>hu,exp:()=>du,expandDims:()=>pu,expm1:()=>cu,eye:()=>mu,fft:()=>vd,fill:()=>pl,findBackend:()=>oi,findBackendFactory:()=>li,floor:()=>gu,floorDiv:()=>mo,forceHalfFloat:()=>Iz,fused:()=>op,gather:()=>yu,gatherND:()=>np,gather_util:()=>bf,getBackend:()=>si,getGradient:()=>pa,getKernel:()=>da,getKernelsForBackend:()=>ca,getThreadsCount:()=>kY,gpgpu_util:()=>LL,grad:()=>$u,grads:()=>Ru,greater:()=>bu,greaterEqual:()=>xu,ifft:()=>wd,imag:()=>vu,image:()=>yc,inTopKAsync:()=>ip,initializers:()=>vv,input:()=>Jw,io:()=>Pc,irfft:()=>kd,isFinite:()=>wu,isInf:()=>ku,isNaN:()=>Iu,keep:()=>ti,kernel_impls:()=>Km,layers:()=>Lv,leakyRelu:()=>Su,less:()=>_u,lessEqual:()=>Nu,linalg:()=>bc,linspace:()=>Tu,loadGraphModel:()=>JT,loadGraphModelSync:()=>QT,loadLayersModel:()=>Kw,localResponseNormalization:()=>Cu,log:()=>Eu,log1p:()=>Au,logSigmoid:()=>Bu,logSoftmax:()=>Vu,logSumExp:()=>Uu,logicalAnd:()=>Gu,logicalNot:()=>Hu,logicalOr:()=>ju,logicalXor:()=>qu,losses:()=>xc,lowerBound:()=>Zu,matMul:()=>Zo,math:()=>sf,max:()=>tu,maxPool:()=>Yu,maxPool3d:()=>Ju,maxPoolWithArgmax:()=>Qu,maximum:()=>eh,mean:()=>th,memory:()=>Ys,meshgrid:()=>ah,metrics:()=>P_,min:()=>nu,minimum:()=>sh,mirrorPad:()=>ih,mod:()=>oh,model:()=>Zw,models:()=>tN,moments:()=>lh,movingAverage:()=>Qd,mul:()=>yo,multiRNNCell:()=>uh,multinomial:()=>hh,neg:()=>Pu,nextFrame:()=>Vf,norm:()=>uu,notEqual:()=>dh,oneHot:()=>ph,ones:()=>rh,onesLike:()=>ch,op:()=>zs,outerProduct:()=>fh,pad:()=>mh,pad1d:()=>gh,pad2d:()=>yh,pad3d:()=>bh,pad4d:()=>xh,pool:()=>wh,pow:()=>ru,prelu:()=>kh,print:()=>co,prod:()=>Ih,profile:()=>Js,raggedGather:()=>Sh,raggedRange:()=>_h,raggedTensorToTensor:()=>Nh,rand:()=>Th,randomGamma:()=>jh,randomNormal:()=>qh,randomStandardNormal:()=>Kh,randomUniform:()=>Xh,randomUniformInt:()=>Zh,range:()=>Yh,ready:()=>ai,real:()=>Jh,reciprocal:()=>Qh,registerBackend:()=>ui,registerCallbackConstructor:()=>Qw,registerGradient:()=>ma,registerKernel:()=>fa,registerOp:()=>fN,regularizers:()=>nN,relu:()=>ed,relu6:()=>td,removeBackend:()=>ii,reshape:()=>jo,reverse:()=>nd,reverse1d:()=>rd,reverse2d:()=>ad,reverse3d:()=>sd,reverse4d:()=>id,rfft:()=>Sd,round:()=>od,rsqrt:()=>ld,scalar:()=>au,scatterND:()=>ep,scatter_util:()=>Ld,searchSorted:()=>Xu,selu:()=>ud,separableConv2d:()=>hd,sequential:()=>Yw,serialization:()=>kc,setBackend:()=>ri,setPlatform:()=>di,setThreadsCount:()=>wY,setWasmPath:()=>yY,setWasmPaths:()=>bY,setWebGLContext:()=>rO,setdiff1dAsync:()=>dd,shared:()=>yE,sigmoid:()=>Yo,sign:()=>pd,signal:()=>gc,sin:()=>cd,sinh:()=>fd,slice:()=>Jo,slice1d:()=>md,slice2d:()=>gd,slice3d:()=>yd,slice4d:()=>bd,slice_util:()=>vf,softmax:()=>xd,softplus:()=>zu,spaceToBatchND:()=>vh,sparse:()=>vc,sparseToDense:()=>tp,spectral:()=>mc,split:()=>Id,sqrt:()=>su,square:()=>iu,squaredDifference:()=>_d,squeeze:()=>Nd,stack:()=>Td,step:()=>Cd,stridedSlice:()=>Ed,string:()=>wc,sub:()=>Wu,sum:()=>ou,sumOutType:()=>cs,tan:()=>Ad,tanh:()=>Qo,tensor:()=>Vs,tensor1d:()=>$d,tensor2d:()=>Rd,tensor3d:()=>Fd,tensor4d:()=>Dd,tensor5d:()=>Md,tensor6d:()=>Od,tensorScatterUpdate:()=>Wd,tensor_util:()=>hs,test_util:()=>Eh,tidy:()=>Qs,tile:()=>fu,time:()=>ni,topk:()=>Vd,train:()=>Bf,transpose:()=>Jd,truncatedNormal:()=>Ud,unique:()=>Gd,unregisterGradient:()=>ya,unregisterKernel:()=>ga,unsortedSegmentSum:()=>Hd,unstack:()=>jd,upcastType:()=>ps,upperBound:()=>qd,util:()=>va,valueAndGrad:()=>Fu,valueAndGrads:()=>Du,variable:()=>Kd,variableGrads:()=>Mu,version:()=>_Y,version_converter:()=>eC,version_core:()=>Pf,version_cpu:()=>G$,version_layers:()=>Nw,version_wasm:()=>IY,version_webgl:()=>kz,webgl:()=>Sz,webgl_util:()=>YM,where:()=>zl,whereAsync:()=>Zd,zeros:()=>nh,zerosLike:()=>Bl});var o=Object.create,l=Object.defineProperty,u=Object.getOwnPropertyDescriptor,h=Object.getOwnPropertyNames,d=Object.getPrototypeOf,p=Object.prototype.hasOwnProperty,c=(e,t)=>()=>(t||e((t={exports:{}}).exports,t),t.exports),f=(e,t)=>{for(var n in t)l(e,n,{get:t[n],enumerable:!0})},m=(e,t,n)=>(n=null!=e?o(d(e)):{},((e,t,n,r)=>{if(t&&"object"==typeof t||"function"==typeof t)for(let a of h(t))!p.call(e,a)&&a!==n&&l(e,a,{get:()=>t[a],enumerable:!(r=u(t,a))||r.enumerable});return e})(!t&&e&&e.__esModule?n:l(n,"default",{value:e,enumerable:!0}),e)),g=c((e,t)=>{t.exports=r;var n=null;try{n=new WebAssembly.Instance(new WebAssembly.Module(new Uint8Array([0,97,115,109,1,0,0,0,1,13,2,96,0,1,127,96,4,127,127,127,127,1,127,3,7,6,0,1,1,1,1,1,6,6,1,127,1,65,0,11,7,50,6,3,109,117,108,0,1,5,100,105,118,95,115,0,2,5,100,105,118,95,117,0,3,5,114,101,109,95,115,0,4,5,114,101,109,95,117,0,5,8,103,101,116,95,104,105,103,104,0,0,10,191,1,6,4,0,35,0,11,36,1,1,126,32,0,173,32,1,173,66,32,134,132,32,2,173,32,3,173,66,32,134,132,126,34,4,66,32,135,167,36,0,32,4,167,11,36,1,1,126,32,0,173,32,1,173,66,32,134,132,32,2,173,32,3,173,66,32,134,132,127,34,4,66,32,135,167,36,0,32,4,167,11,36,1,1,126,32,0,173,32,1,173,66,32,134,132,32,2,173,32,3,173,66,32,134,132,128,34,4,66,32,135,167,36,0,32,4,167,11,36,1,1,126,32,0,173,32,1,173,66,32,134,132,32,2,173,32,3,173,66,32,134,132,129,34,4,66,32,135,167,36,0,32,4,167,11,36,1,1,126,32,0,173,32,1,173,66,32,134,132,32,2,173,32,3,173,66,32,134,132,130,34,4,66,32,135,167,36,0,32,4,167,11])),{}).exports}catch(e){}function r(e,t,n){this.low=0|e,this.high=0|t,this.unsigned=!!n}function a(e){return!0===(e&&e.__isLong__)}r.prototype.__isLong__,Object.defineProperty(r.prototype,"__isLong__",{value:!0}),r.isLong=a;var s={},i={};function o(e,t){var n,r,a;return t?(a=0<=(e>>>=0)&&e<256)&&(r=i[e])?r:(n=u(e,(0|e)<0?-1:0,!0),a&&(i[e]=n),n):(a=-128<=(e|=0)&&e<128)&&(r=s[e])?r:(n=u(e,e<0?-1:0,!1),a&&(s[e]=n),n)}function l(e,t){if(isNaN(e))return t?b:y;if(t){if(e<0)return b;if(e>=f)return I}else{if(e<=-m)return S;if(e+1>=m)return k}return e<0?l(-e,t).neg():u(e%c|0,e/c|0,t)}function u(e,t,n){return new r(e,t,n)}r.fromInt=o,r.fromNumber=l,r.fromBits=u;var h=Math.pow;function d(e,t,n){if(0===e.length)throw Error("empty string");if("NaN"===e||"Infinity"===e||"+Infinity"===e||"-Infinity"===e)return y;if("number"==typeof t?(n=t,t=!1):t=!!t,(n=n||10)<2||36<n)throw RangeError("radix");var r;if((r=e.indexOf("-"))>0)throw Error("interior hyphen");if(0===r)return d(e.substring(1),t,n).neg();for(var a=l(h(n,8)),s=y,i=0;i<e.length;i+=8){var o=Math.min(8,e.length-i),u=parseInt(e.substring(i,i+o),n);if(o<8){var p=l(h(n,o));s=s.mul(p).add(l(u))}else s=(s=s.mul(a)).add(l(u))}return s.unsigned=t,s}function p(e,t){return"number"==typeof e?l(e,t):"string"==typeof e?d(e,t):u(e.low,e.high,"boolean"==typeof t?t:e.unsigned)}r.fromString=d,r.fromValue=p;var c=4294967296,f=c*c,m=f/2,g=o(1<<24),y=o(0);r.ZERO=y;var b=o(0,!0);r.UZERO=b;var x=o(1);r.ONE=x;var v=o(1,!0);r.UONE=v;var w=o(-1);r.NEG_ONE=w;var k=u(-1,2147483647,!1);r.MAX_VALUE=k;var I=u(-1,-1,!0);r.MAX_UNSIGNED_VALUE=I;var S=u(0,-2147483648,!1);r.MIN_VALUE=S;var _=r.prototype;_.toInt=function(){return this.unsigned?this.low>>>0:this.low},_.toNumber=function(){return this.unsigned?(this.high>>>0)*c+(this.low>>>0):this.high*c+(this.low>>>0)},_.toString=function(e){if((e=e||10)<2||36<e)throw RangeError("radix");if(this.isZero())return"0";if(this.isNegative()){if(this.eq(S)){var t=l(e),n=this.div(t),r=n.mul(t).sub(this);return n.toString(e)+r.toInt().toString(e)}return"-"+this.neg().toString(e)}for(var a=l(h(e,6),this.unsigned),s=this,i="";;){var o=s.div(a),u=(s.sub(o.mul(a)).toInt()>>>0).toString(e);if((s=o).isZero())return u+i;for(;u.length<6;)u="0"+u;i=""+u+i}},_.getHighBits=function(){return this.high},_.getHighBitsUnsigned=function(){return this.high>>>0},_.getLowBits=function(){return this.low},_.getLowBitsUnsigned=function(){return this.low>>>0},_.getNumBitsAbs=function(){if(this.isNegative())return this.eq(S)?64:this.neg().getNumBitsAbs();for(var e=0!=this.high?this.high:this.low,t=31;t>0&&!(e&1<<t);t--);return 0!=this.high?t+33:t+1},_.isZero=function(){return 0===this.high&&0===this.low},_.eqz=_.isZero,_.isNegative=function(){return!this.unsigned&&this.high<0},_.isPositive=function(){return this.unsigned||this.high>=0},_.isOdd=function(){return!(1&~this.low)},_.isEven=function(){return!(1&this.low)},_.equals=function(e){return a(e)||(e=p(e)),(this.unsigned===e.unsigned||this.high>>>31!=1||e.high>>>31!=1)&&(this.high===e.high&&this.low===e.low)},_.eq=_.equals,_.notEquals=function(e){return!this.eq(e)},_.neq=_.notEquals,_.ne=_.notEquals,_.lessThan=function(e){return this.comp(e)<0},_.lt=_.lessThan,_.lessThanOrEqual=function(e){return this.comp(e)<=0},_.lte=_.lessThanOrEqual,_.le=_.lessThanOrEqual,_.greaterThan=function(e){return this.comp(e)>0},_.gt=_.greaterThan,_.greaterThanOrEqual=function(e){return this.comp(e)>=0},_.gte=_.greaterThanOrEqual,_.ge=_.greaterThanOrEqual,_.compare=function(e){if(a(e)||(e=p(e)),this.eq(e))return 0;var t=this.isNegative(),n=e.isNegative();return t&&!n?-1:!t&&n?1:this.unsigned?e.high>>>0>this.high>>>0||e.high===this.high&&e.low>>>0>this.low>>>0?-1:1:this.sub(e).isNegative()?-1:1},_.comp=_.compare,_.negate=function(){return!this.unsigned&&this.eq(S)?S:this.not().add(x)},_.neg=_.negate,_.add=function(e){a(e)||(e=p(e));var t=this.high>>>16,n=65535&this.high,r=this.low>>>16,s=65535&this.low,i=e.high>>>16,o=65535&e.high,l=e.low>>>16,h=0,d=0,c=0,f=0;return c+=(f+=s+(65535&e.low))>>>16,d+=(c+=r+l)>>>16,h+=(d+=n+o)>>>16,h+=t+i,u((c&=65535)<<16|(f&=65535),(h&=65535)<<16|(d&=65535),this.unsigned)},_.subtract=function(e){return a(e)||(e=p(e)),this.add(e.neg())},_.sub=_.subtract,_.multiply=function(e){if(this.isZero())return y;if(a(e)||(e=p(e)),n)return u(n.mul(this.low,this.high,e.low,e.high),n.get_high(),this.unsigned);if(e.isZero())return y;if(this.eq(S))return e.isOdd()?S:y;if(e.eq(S))return this.isOdd()?S:y;if(this.isNegative())return e.isNegative()?this.neg().mul(e.neg()):this.neg().mul(e).neg();if(e.isNegative())return this.mul(e.neg()).neg();if(this.lt(g)&&e.lt(g))return l(this.toNumber()*e.toNumber(),this.unsigned);var t=this.high>>>16,r=65535&this.high,s=this.low>>>16,i=65535&this.low,o=e.high>>>16,h=65535&e.high,d=e.low>>>16,c=65535&e.low,f=0,m=0,b=0,x=0;return b+=(x+=i*c)>>>16,m+=(b+=s*c)>>>16,b&=65535,m+=(b+=i*d)>>>16,f+=(m+=r*c)>>>16,m&=65535,f+=(m+=s*d)>>>16,m&=65535,f+=(m+=i*h)>>>16,f+=t*c+r*d+s*h+i*o,u((b&=65535)<<16|(x&=65535),(f&=65535)<<16|(m&=65535),this.unsigned)},_.mul=_.multiply,_.divide=function(e){if(a(e)||(e=p(e)),e.isZero())throw Error("division by zero");var t,r,s;if(n)return this.unsigned||-2147483648!==this.high||-1!==e.low||-1!==e.high?u((this.unsigned?n.div_u:n.div_s)(this.low,this.high,e.low,e.high),n.get_high(),this.unsigned):this;if(this.isZero())return this.unsigned?b:y;if(this.unsigned){if(e.unsigned||(e=e.toUnsigned()),e.gt(this))return b;if(e.gt(this.shru(1)))return v;s=b}else{if(this.eq(S))return e.eq(x)||e.eq(w)?S:e.eq(S)?x:(t=this.shr(1).div(e).shl(1)).eq(y)?e.isNegative()?x:w:(r=this.sub(e.mul(t)),s=t.add(r.div(e)));if(e.eq(S))return this.unsigned?b:y;if(this.isNegative())return e.isNegative()?this.neg().div(e.neg()):this.neg().div(e).neg();if(e.isNegative())return this.div(e.neg()).neg();s=y}for(r=this;r.gte(e);){t=Math.max(1,Math.floor(r.toNumber()/e.toNumber()));for(var i=Math.ceil(Math.log(t)/Math.LN2),o=i<=48?1:h(2,i-48),d=l(t),c=d.mul(e);c.isNegative()||c.gt(r);)c=(d=l(t-=o,this.unsigned)).mul(e);d.isZero()&&(d=x),s=s.add(d),r=r.sub(c)}return s},_.div=_.divide,_.modulo=function(e){return a(e)||(e=p(e)),n?u((this.unsigned?n.rem_u:n.rem_s)(this.low,this.high,e.low,e.high),n.get_high(),this.unsigned):this.sub(this.div(e).mul(e))},_.mod=_.modulo,_.rem=_.modulo,_.not=function(){return u(~this.low,~this.high,this.unsigned)},_.and=function(e){return a(e)||(e=p(e)),u(this.low&e.low,this.high&e.high,this.unsigned)},_.or=function(e){return a(e)||(e=p(e)),u(this.low|e.low,this.high|e.high,this.unsigned)},_.xor=function(e){return a(e)||(e=p(e)),u(this.low^e.low,this.high^e.high,this.unsigned)},_.shiftLeft=function(e){return a(e)&&(e=e.toInt()),0==(e&=63)?this:e<32?u(this.low<<e,this.high<<e|this.low>>>32-e,this.unsigned):u(0,this.low<<e-32,this.unsigned)},_.shl=_.shiftLeft,_.shiftRight=function(e){return a(e)&&(e=e.toInt()),0==(e&=63)?this:e<32?u(this.low>>>e|this.high<<32-e,this.high>>e,this.unsigned):u(this.high>>e-32,this.high>=0?0:-1,this.unsigned)},_.shr=_.shiftRight,_.shiftRightUnsigned=function(e){if(a(e)&&(e=e.toInt()),0===(e&=63))return this;var t=this.high;return e<32?u(this.low>>>e|t<<32-e,t>>>e,this.unsigned):u(32===e?t:t>>>e-32,0,this.unsigned)},_.shru=_.shiftRightUnsigned,_.shr_u=_.shiftRightUnsigned,_.toSigned=function(){return this.unsigned?u(this.low,this.high,!1):this},_.toUnsigned=function(){return this.unsigned?this:u(this.low,this.high,!0)},_.toBytes=function(e){return e?this.toBytesLE():this.toBytesBE()},_.toBytesLE=function(){var e=this.high,t=this.low;return[255&t,t>>>8&255,t>>>16&255,t>>>24,255&e,e>>>8&255,e>>>16&255,e>>>24]},_.toBytesBE=function(){var e=this.high,t=this.low;return[e>>>24,e>>>16&255,e>>>8&255,255&e,t>>>24,t>>>16&255,t>>>8&255,255&t]},r.fromBytes=function(e,t,n){return n?r.fromBytesLE(e,t):r.fromBytesBE(e,t)},r.fromBytesLE=function(e,t){return new r(e[0]|e[1]<<8|e[2]<<16|e[3]<<24,e[4]|e[5]<<8|e[6]<<16|e[7]<<24,t)},r.fromBytesBE=function(e,t){return new r(e[4]<<24|e[5]<<16|e[6]<<8|e[7],e[0]<<24|e[1]<<16|e[2]<<8|e[3],t)}}),y=c(()=>{}),b=c(()=>{}),x=c((e,t)=>{!function(e,t,n){function r(e){var t=this,n=function(){var e=4022871197,t=function(t){t=String(t);for(var n=0;n<t.length;n++){var r=.02519603282416938*(e+=t.charCodeAt(n));r-=e=r>>>0,e=(r*=e)>>>0,e+=4294967296*(r-=e)}return 2.3283064365386963e-10*(e>>>0)};return t}();t.next=function(){var e=2091639*t.s0+2.3283064365386963e-10*t.c;return t.s0=t.s1,t.s1=t.s2,t.s2=e-(t.c=0|e)},t.c=1,t.s0=n(" "),t.s1=n(" "),t.s2=n(" "),t.s0-=n(e),t.s0<0&&(t.s0+=1),t.s1-=n(e),t.s1<0&&(t.s1+=1),t.s2-=n(e),t.s2<0&&(t.s2+=1),n=null}function a(e,t){return t.c=e.c,t.s0=e.s0,t.s1=e.s1,t.s2=e.s2,t}function s(e,t){var n=new r(e),s=t&&t.state,i=n.next;return i.int32=function(){return 4294967296*n.next()|0},i.double=function(){return i()+11102230246251565e-32*(2097152*i()|0)},i.quick=i,s&&("object"==typeof s&&a(s,n),i.state=function(){return a(n,{})}),i}t&&t.exports?t.exports=s:n&&n.amd?n(function(){return s}):this.alea=s}(0,"object"==typeof t&&t,"function"==typeof define&&define)}),v=c((e,t)=>{!function(e,t,n){function r(e){var t=this,n="";t.x=0,t.y=0,t.z=0,t.w=0,t.next=function(){var e=t.x^t.x<<11;return t.x=t.y,t.y=t.z,t.z=t.w,t.w^=t.w>>>19^e^e>>>8},e===(0|e)?t.x=e:n+=e;for(var r=0;r<n.length+64;r++)t.x^=0|n.charCodeAt(r),t.next()}function a(e,t){return t.x=e.x,t.y=e.y,t.z=e.z,t.w=e.w,t}function s(e,t){var n=new r(e),s=t&&t.state,i=function(){return(n.next()>>>0)/4294967296};return i.double=function(){do{var e=((n.next()>>>11)+(n.next()>>>0)/4294967296)/(1<<21)}while(0===e);return e},i.int32=n.next,i.quick=i,s&&("object"==typeof s&&a(s,n),i.state=function(){return a(n,{})}),i}t&&t.exports?t.exports=s:n&&n.amd?n(function(){return s}):this.xor128=s}(0,"object"==typeof t&&t,"function"==typeof define&&define)}),w=c((e,t)=>{!function(e,t,n){function r(e){var t=this,n="";t.next=function(){var e=t.x^t.x>>>2;return t.x=t.y,t.y=t.z,t.z=t.w,t.w=t.v,(t.d=t.d+362437|0)+(t.v=t.v^t.v<<4^e^e<<1)|0},t.x=0,t.y=0,t.z=0,t.w=0,t.v=0,e===(0|e)?t.x=e:n+=e;for(var r=0;r<n.length+64;r++)t.x^=0|n.charCodeAt(r),r==n.length&&(t.d=t.x<<10^t.x>>>4),t.next()}function a(e,t){return t.x=e.x,t.y=e.y,t.z=e.z,t.w=e.w,t.v=e.v,t.d=e.d,t}function s(e,t){var n=new r(e),s=t&&t.state,i=function(){return(n.next()>>>0)/4294967296};return i.double=function(){do{var e=((n.next()>>>11)+(n.next()>>>0)/4294967296)/(1<<21)}while(0===e);return e},i.int32=n.next,i.quick=i,s&&("object"==typeof s&&a(s,n),i.state=function(){return a(n,{})}),i}t&&t.exports?t.exports=s:n&&n.amd?n(function(){return s}):this.xorwow=s}(0,"object"==typeof t&&t,"function"==typeof define&&define)}),k=c((e,t)=>{!function(e,t,n){function r(e){var t=this;t.next=function(){var e,n,r=t.x,a=t.i;return e=r[a],n=(e^=e>>>7)^e<<24,n^=(e=r[a+1&7])^e>>>10,n^=(e=r[a+3&7])^e>>>3,n^=(e=r[a+4&7])^e<<7,e=r[a+7&7],n^=(e^=e<<13)^e<<9,r[a]=n,t.i=a+1&7,n},function(e,t){var n,r=[];if(t===(0|t))r[0]=t;else for(t=""+t,n=0;n<t.length;++n)r[7&n]=r[7&n]<<15^t.charCodeAt(n)+r[n+1&7]<<13;for(;r.length<8;)r.push(0);for(n=0;n<8&&0===r[n];++n);for(8==n?r[7]=-1:r[n],e.x=r,e.i=0,n=256;n>0;--n)e.next()}(t,e)}function a(e,t){return t.x=e.x.slice(),t.i=e.i,t}function s(e,t){null==e&&(e=+new Date);var n=new r(e),s=t&&t.state,i=function(){return(n.next()>>>0)/4294967296};return i.double=function(){do{var e=((n.next()>>>11)+(n.next()>>>0)/4294967296)/(1<<21)}while(0===e);return e},i.int32=n.next,i.quick=i,s&&(s.x&&a(s,n),i.state=function(){return a(n,{})}),i}t&&t.exports?t.exports=s:n&&n.amd?n(function(){return s}):this.xorshift7=s}(0,"object"==typeof t&&t,"function"==typeof define&&define)}),I=c((e,t)=>{!function(e,t,n){function r(e){var t=this;t.next=function(){var e,n,r=t.w,a=t.X,s=t.i;return t.w=r=r+1640531527|0,n=a[s+34&127],e=a[s=s+1&127],n^=n<<13,e^=e<<17,n^=n>>>15,e^=e>>>12,n=a[s]=n^e,t.i=s,n+(r^r>>>16)|0},function(e,t){var n,r,a,s,i,o=[],l=128;for(t===(0|t)?(r=t,t=null):(t+="\0",r=0,l=Math.max(l,t.length)),a=0,s=-32;s<l;++s)t&&(r^=t.charCodeAt((s+32)%t.length)),0===s&&(i=r),r^=r<<10,r^=r>>>15,r^=r<<4,r^=r>>>13,s>=0&&(i=i+1640531527|0,a=0==(n=o[127&s]^=r+i)?a+1:0);for(a>=128&&(o[127&(t&&t.length||0)]=-1),a=127,s=512;s>0;--s)r=o[a+34&127],n=o[a=a+1&127],r^=r<<13,n^=n<<17,r^=r>>>15,n^=n>>>12,o[a]=r^n;e.w=i,e.X=o,e.i=a}(t,e)}function a(e,t){return t.i=e.i,t.w=e.w,t.X=e.X.slice(),t}function s(e,t){null==e&&(e=+new Date);var n=new r(e),s=t&&t.state,i=function(){return(n.next()>>>0)/4294967296};return i.double=function(){do{var e=((n.next()>>>11)+(n.next()>>>0)/4294967296)/(1<<21)}while(0===e);return e},i.int32=n.next,i.quick=i,s&&(s.X&&a(s,n),i.state=function(){return a(n,{})}),i}t&&t.exports?t.exports=s:n&&n.amd?n(function(){return s}):this.xor4096=s}(0,"object"==typeof t&&t,"function"==typeof define&&define)}),S=c((e,t)=>{!function(e,t,n){function r(e){var t=this,n="";t.next=function(){var e=t.b,n=t.c,r=t.d,a=t.a;return e=e<<25^e>>>7^n,n=n-r|0,r=r<<24^r>>>8^a,a=a-e|0,t.b=e=e<<20^e>>>12^n,t.c=n=n-r|0,t.d=r<<16^n>>>16^a,t.a=a-e|0},t.a=0,t.b=0,t.c=-1640531527,t.d=1367130551,e===Math.floor(e)?(t.a=e/4294967296|0,t.b=0|e):n+=e;for(var r=0;r<n.length+20;r++)t.b^=0|n.charCodeAt(r),t.next()}function a(e,t){return t.a=e.a,t.b=e.b,t.c=e.c,t.d=e.d,t}function s(e,t){var n=new r(e),s=t&&t.state,i=function(){return(n.next()>>>0)/4294967296};return i.double=function(){do{var e=((n.next()>>>11)+(n.next()>>>0)/4294967296)/(1<<21)}while(0===e);return e},i.int32=n.next,i.quick=i,s&&("object"==typeof s&&a(s,n),i.state=function(){return a(n,{})}),i}t&&t.exports?t.exports=s:n&&n.amd?n(function(){return s}):this.tychei=s}(0,"object"==typeof t&&t,"function"==typeof define&&define)}),_=c(()=>{}),N=c((e,t)=>{!function(e,r,a){var s,i=256,o="random",l=a.pow(i,6),u=a.pow(2,52),h=2*u,d=255;function p(t,n,d){var p=[],b=g(m((n=1==n?{entropy:!0}:n||{}).entropy?[t,y(r)]:t??function(){try{var t;return s&&(t=s.randomBytes)?t=t(i):(t=new Uint8Array(i),(e.crypto||e.msCrypto).getRandomValues(t)),y(t)}catch(t){var n=e.navigator,a=n&&n.plugins;return[+new Date,e,a,e.screen,y(r)]}}(),3),p),x=new c(p),v=function(){for(var e=x.g(6),t=l,n=0;e<u;)e=(e+n)*i,t*=i,n=x.g(1);for(;e>=h;)e/=2,t/=2,n>>>=1;return(e+n)/t};return v.int32=function(){return 0|x.g(4)},v.quick=function(){return x.g(4)/4294967296},v.double=v,g(y(x.S),r),(n.pass||d||function(e,t,n,r){return r&&(r.S&&f(r,x),e.state=function(){return f(x,{})}),n?(a[o]=e,t):e})(v,b,"global"in n?n.global:this==a,n.state)}function c(e){var t,n=e.length,r=this,a=0,s=r.i=r.j=0,o=r.S=[];for(n||(e=[n++]);a<i;)o[a]=a++;for(a=0;a<i;a++)o[a]=o[s=d&s+e[a%n]+(t=o[a])],o[s]=t;(r.g=function(e){for(var t,n=0,a=r.i,s=r.j,o=r.S;e--;)t=o[a=d&a+1],n=n*i+o[d&(o[a]=o[s=d&s+t])+(o[s]=t)];return r.i=a,r.j=s,n})(i)}function f(e,t){return t.i=e.i,t.j=e.j,t.S=e.S.slice(),t}function m(e,t){var n,r=[],a=typeof e;if(t&&"object"==a)for(n in e)try{r.push(m(e[n],t-1))}catch(e){}return r.length?r:"string"==a?e:e+"\0"}function g(e,t){for(var n,r=e+"",a=0;a<r.length;)t[d&a]=d&(n^=19*t[d&a])+r.charCodeAt(a++);return y(t)}function y(e){return String.fromCharCode.apply(0,e)}if(g(a.random(),r),"object"==typeof t&&t.exports){t.exports=p;try{s=_()}catch(e){}}else"function"==typeof define&&n.amdO?define(function(){return p}):a["seed"+o]=p}("undefined"!=typeof self?self:e,[],Math)}),T=c((e,t)=>{var n=x(),r=v(),a=w(),s=k(),i=I(),o=S(),l=N();l.alea=n,l.xor128=r,l.xorwow=a,l.xorshift7=s,l.xor4096=i,l.tychei=o,t.exports=l}),C=c(()=>{}),E=c(()=>{}),A=c(()=>{}),$=c(()=>{}),R=c(()=>{}),F=c(()=>{}),D=c((e,t)=>{var a,s=(a=(a="undefined"!=typeof document&&document.currentScript?document.currentScript.src:void 0)||r,function(e){function t(){return O.buffer!=z&&X(O.buffer),W}function r(){return O.buffer!=z&&X(O.buffer),V}function s(){return O.buffer!=z&&X(O.buffer),U}function i(){return O.buffer!=z&&X(O.buffer),G}var o,l,u,h=void 0!==(e=e||{})?e:{};h.ready=new Promise(function(e,t){o=e,l=t}),"undefined"!=typeof process&&process.listeners&&(u={uncaughtException:process.listeners("uncaughtException"),unhandledRejection:process.listeners("unhandledRejection")});var d,p,c,f=Object.assign({},h),m=[],g=(e,t)=>{throw t},y="object"==typeof window,b="function"==typeof importScripts,x="object"==typeof process&&"object"==typeof process.versions&&"string"==typeof process.versions.node,v=h.ENVIRONMENT_IS_PTHREAD||!1,w="";function k(e){return h.locateFile?h.locateFile(e,w):w+e}if(x){var I=E(),S=A();let e;w=b?S.dirname(w)+"/":"//",d=(e,t)=>(e=de(e)?new URL(e):S.normalize(e),I.readFileSync(e,t?void 0:"utf8")),c=e=>{var t=d(e,!0);return t.buffer||(t=new Uint8Array(t)),t},p=(e,t,n)=>{e=de(e)?new URL(e):S.normalize(e),I.readFile(e,function(e,r){e?n(e):t(r.buffer)})},process.argv.length>1&&process.argv[1].replace(/\\/g,"/"),m=process.argv.slice(2),process.on("uncaughtException",function(e){if(!(e instanceof fe))throw e}),process.on("unhandledRejection",function(e){throw e}),g=(e,t)=>{if(te())throw process.exitCode=e,t;(function(e){e instanceof fe||D("exiting due to exception: "+e)})(t),process.exit(e)},h.inspect=function(){return"[Emscripten Module object]"};try{e=$()}catch(e){throw console.error('The "worker_threads" module is not supported in this node.js build - perhaps a newer version is needed?'),e}n.g.Worker=e.Worker}else(y||b)&&(b?w=self.location.href:"undefined"!=typeof document&&document.currentScript&&(w=document.currentScript.src),void 0!==a&&a&&(w=a),w=0!==w.indexOf("blob:")?w.substr(0,w.replace(/[?#].*/,"").lastIndexOf("/")+1):"",x||(d=e=>{var t=new XMLHttpRequest;return t.open("GET",e,!1),t.send(null),t.responseText},b&&(c=e=>{var t=new XMLHttpRequest;return t.open("GET",e,!1),t.responseType="arraybuffer",t.send(null),new Uint8Array(t.response)}),p=(e,t,n)=>{var r=new XMLHttpRequest;r.open("GET",e,!0),r.responseType="arraybuffer",r.onload=()=>{200==r.status||0==r.status&&r.response?t(r.response):n()},r.onerror=n,r.send(null)}));x&&"undefined"==typeof performance&&(n.g.performance=R().performance);var _=console.log.bind(console),N=console.warn.bind(console);x&&(_=e=>I.writeSync(1,e+"\n"),N=e=>I.writeSync(2,e+"\n"));var T,C=h.print||_,D=h.printErr||N;Object.assign(h,f),f=null,h.arguments&&(m=h.arguments),h.thisProgram&&h.thisProgram,h.quit&&(g=h.quit),Atomics.load,Atomics.store,Atomics.compareExchange,h.wasmBinary&&(T=h.wasmBinary);var M=h.noExitRuntime||!0;"object"!=typeof WebAssembly&&ue("no native wasm support detected");var O,L,P,z,B,W,V,U,G,H=!1,j="undefined"!=typeof TextDecoder?new TextDecoder("utf8"):void 0;function q(e,t,n){for(var r=(t>>>=0)+n,a=t;e[a]&&!(a>=r);)++a;if(a-t>16&&e.buffer&&j)return j.decode(e.buffer instanceof SharedArrayBuffer?e.slice(t,a):e.subarray(t,a));for(var s="";t<a;){var i=e[t++];if(128&i){var o=63&e[t++];if(192!=(224&i)){var l=63&e[t++];if((i=224==(240&i)?(15&i)<<12|o<<6|l:(7&i)<<18|o<<12|l<<6|63&e[t++])<65536)s+=String.fromCharCode(i);else{var u=i-65536;s+=String.fromCharCode(55296|u>>10,56320|1023&u)}}else s+=String.fromCharCode((31&i)<<6|o)}else s+=String.fromCharCode(i)}return s}function K(e,n){return(e>>>=0)?q(t(),e,n):""}function X(e){z=e,h.HEAP8=B=new Int8Array(e),h.HEAP16=new Int16Array(e),h.HEAP32=V=new Int32Array(e),h.HEAPU8=W=new Uint8Array(e),h.HEAPU16=new Uint16Array(e),h.HEAPU32=U=new Uint32Array(e),h.HEAPF32=new Float32Array(e),h.HEAPF64=G=new Float64Array(e)}v&&(z=h.buffer);var Z=h.INITIAL_MEMORY||16777216;if(v)O=h.wasmMemory,z=h.buffer;else if(h.wasmMemory)O=h.wasmMemory;else if(!((O=new WebAssembly.Memory({initial:Z/65536,maximum:65536,shared:!0})).buffer instanceof SharedArrayBuffer))throw D("requested a shared WebAssembly.Memory but the returned buffer is not a SharedArrayBuffer, indicating that while the browser has SharedArrayBuffer it does not have WebAssembly threads support - you may need to set a flag"),x&&D("(on node you may need: --experimental-wasm-threads --experimental-wasm-bulk-memory and/or recent version)"),Error("bad memory");O&&(z=O.buffer),Z=z.byteLength,X(z);var Y,J=[],Q=[],ee=[];function te(){return M}function ne(){!v&&ve(Q)}function re(e){J.unshift(e)}function ae(e){ee.unshift(e)}var se,ie=0,oe=null,le=null;function ue(e){h.onAbort&&h.onAbort(e),D(e="Aborted("+e+")"),H=!0,P=1,e+=". Build with -sASSERTIONS for more info.";var t=new WebAssembly.RuntimeError(e);throw l(t),t}function he(e){return e.startsWith("data:application/octet-stream;base64,")}function de(e){return e.startsWith("file://")}function pe(e){try{if(e==se&&T)return new Uint8Array(T);if(c)return c(e);throw"both async and sync fetching of the wasm failed"}catch(e){ue(e)}}he(se="tfjs-backend-wasm-threaded-simd.wasm")||(se=k(se));var ce={};function fe(e){this.name="ExitStatus",this.message="Program terminated with exit("+e+")",this.status=e}function me(e){var t=xe.pthreads[e];(function(e,t){e||ue(t)})(t),xe.returnWorkerToPool(t)}function ge(e){var t=xe.getNewWorker();if(!t)return 6;xe.runningWorkers.push(t),xe.pthreads[e.pthread_ptr]=t,t.pthread_ptr=e.pthread_ptr;var n={cmd:"run",start_routine:e.startRoutine,arg:e.arg,pthread_ptr:e.pthread_ptr};return t.runPthread=()=>{x&&t.ref(),t.postMessage(n,e.transferList),delete t.runPthread},t.loaded&&t.runPthread(),0}function ye(e){if(v)return Ce(1,1,e);P=e,te()||(xe.terminateAllThreads(),h.onExit&&h.onExit(e),H=!0),g(e,new fe(e))}var be=function(e,t){if(P=e,!t&&v)throw we(e),"unwind";ye(e)},xe={unusedWorkers:[],runningWorkers:[],tlsInitFunctions:[],pthreads:{},init:function(){v?xe.initWorker():xe.initMainThread()},initMainThread:function(){for(var e=8;e--;)xe.allocateUnusedWorker()},initWorker:function(){M=!1},setExitStatus:function(e){P=e},terminateAllThreads:function(){for(var e of Object.values(xe.pthreads))xe.returnWorkerToPool(e);for(var e of xe.unusedWorkers)e.terminate();xe.unusedWorkers=[]},returnWorkerToPool:function(e){var t=e.pthread_ptr;delete xe.pthreads[t],xe.unusedWorkers.push(e),xe.runningWorkers.splice(xe.runningWorkers.indexOf(e),1),e.pthread_ptr=0,x&&e.unref(),Ke(t)},receiveObjectTransfer:function(e){},threadInitTLS:function(){xe.tlsInitFunctions.forEach(e=>e())},loadWasmModuleToWorker:function(e,t){e.onmessage=n=>{var r=n.data,a=r.cmd;if(e.pthread_ptr&&(xe.currentProxiedOperationCallerThread=e.pthread_ptr),r.targetThread&&r.targetThread!=Ge()){var s=xe.pthreads[r.targetThread];return s?s.postMessage(r,r.transferList):D('Internal error! Worker sent a message "'+a+'" to target pthread '+r.targetThread+", but that thread no longer exists!"),void(xe.currentProxiedOperationCallerThread=void 0)}"processProxyingQueue"===a?Ne(r.queue):"spawnThread"===a?ge(r):"cleanupThread"===a?me(r.thread):"killThread"===a?function(e){var t=xe.pthreads[e];delete xe.pthreads[e],t.terminate(),Ke(e),xe.runningWorkers.splice(xe.runningWorkers.indexOf(t),1),t.pthread_ptr=0}(r.thread):"cancelThread"===a?function(e){xe.pthreads[e].postMessage({cmd:"cancel"})}(r.thread):"loaded"===a?(e.loaded=!0,x&&e.unref(),t&&t(e),e.runPthread&&e.runPthread()):"print"===a?C("Thread "+r.threadId+": "+r.text):"printErr"===a?D("Thread "+r.threadId+": "+r.text):"alert"===a?alert("Thread "+r.threadId+": "+r.text):"setimmediate"===r.target?e.postMessage(r):"callHandler"===a?h[r.handler](...r.args):a&&D("worker sent an unknown command "+a),xe.currentProxiedOperationCallerThread=void 0},e.onerror=e=>{throw D("worker sent an error! "+e.filename+":"+e.lineno+": "+e.message),e},x&&(e.on("message",function(t){e.onmessage({data:t})}),e.on("error",function(t){e.onerror(t)}),e.on("detachedExit",function(){}));var n=[];for(var r of["onExit","onAbort","print","printErr"])h.hasOwnProperty(r)&&n.push(r);e.postMessage({cmd:"load",handlers:n,urlOrBlob:h.mainScriptUrlOrBlob||a,wasmMemory:O,wasmModule:L})},allocateUnusedWorker:function(){var e,t=k("tfjs-backend-wasm-threaded-simd.worker.js");e=new Worker(t),xe.unusedWorkers.push(e)},getNewWorker:function(){return 0==xe.unusedWorkers.length&&(xe.allocateUnusedWorker(),xe.loadWasmModuleToWorker(xe.unusedWorkers[0])),xe.unusedWorkers.pop()}};function ve(e){for(;e.length>0;)e.shift()(h)}function we(e){if(v)return Ce(2,0,e);try{be(e)}catch(e){!function(e){if(e instanceof fe||"unwind"==e)return P;g(1,e)}(e)}}h.PThread=xe,h.establishStackSpace=function(){var e=Ge(),t=r()[e+52>>>2],n=r()[e+56>>>2];Ze(t,t-n),Je(t)};var ke,Ie=[];function Se(e,t,n,r){return v?Ce(3,1,e,t,n,r):_e(e,t,n,r)}function _e(e,t,n,r){if("undefined"==typeof SharedArrayBuffer)return D("Current environment does not support SharedArrayBuffer, pthreads are not available!"),6;var a=[];if(v&&0===a.length)return Se(e,t,n,r);var s={startRoutine:n,pthread_ptr:e,arg:r,transferList:a};return v?(s.cmd="spawnThread",postMessage(s,a),0):ge(s)}function Ne(e){Atomics.store(r(),e>>2,1),Ge()&&qe(e),Atomics.compareExchange(r(),e>>2,1,0)}function Te(e){Te.shown||(Te.shown={}),Te.shown[e]||(Te.shown[e]=1,x&&(e="warning: "+e),D(e))}function Ce(e,t){var n=arguments.length-2,r=arguments;return function(e){var t=Ye(),n=e();return Je(t),n}(()=>{for(var a=n,s=Qe(8*a),o=s>>3,l=0;l<n;l++){var u=r[2+l];i()[o+l>>>0]=u}return je(e,a,s,t)})}h.invokeEntryPoint=function(e,t){var n=function(e){var t=Ie[e];return t||(e>=Ie.length&&(Ie.length=e+1),Ie[e]=t=Y.get(e)),t}(e)(t);te()?xe.setExitStatus(n):Xe(n)},h.executeNotifiedProxyingQueue=Ne,ke=x?()=>{var e=process.hrtime();return 1e3*e[0]+e[1]/1e6}:()=>performance.timeOrigin+performance.now();var Ee=[];function Ae(e){try{return O.grow(e-z.byteLength+65535>>>16),X(O.buffer),1}catch(e){}}function $e(e){return v?Ce(4,1,e):52}function Re(e,t,n,r,a){return v?Ce(5,1,e,t,n,r,a):70}var Fe=[null,[],[]];function De(e,t){var n=Fe[e];0===t||10===t?((1===e?C:D)(q(n,0)),n.length=0):n.push(t)}function Me(e,n,r,a){if(v)return Ce(6,1,e,n,r,a);for(var i=0,o=0;o<r;o++){var l=s()[n>>>2],u=s()[n+4>>>2];n+=8;for(var h=0;h<u;h++)De(e,t()[l+h>>>0]);i+=u}return s()[a>>>2]=i,0}function Oe(e){return h["_"+e]}function Le(e,t){(O.buffer!=z&&X(O.buffer),B).set(e,t>>>0)}function Pe(e,n,r,a,s){var i={string:e=>{var n=0;if(null!=e&&0!==e){var r=1+(e.length<<2);!function(e,n,r){!function(e,t,n,r){if(!(r>0))return 0;n>>>=0;for(var a=n+r-1,s=0;s<e.length;++s){var i=e.charCodeAt(s);if(i>=55296&&i<=57343&&(i=65536+((1023&i)<<10)|1023&e.charCodeAt(++s)),i<=127){if(n>=a)break;t[n++>>>0]=i}else if(i<=2047){if(n+1>=a)break;t[n++>>>0]=192|i>>6,t[n++>>>0]=128|63&i}else if(i<=65535){if(n+2>=a)break;t[n++>>>0]=224|i>>12,t[n++>>>0]=128|i>>6&63,t[n++>>>0]=128|63&i}else{if(n+3>=a)break;t[n++>>>0]=240|i>>18,t[n++>>>0]=128|i>>12&63,t[n++>>>0]=128|i>>6&63,t[n++>>>0]=128|63&i}}t[n>>>0]=0}(e,t(),n,r)}(e,n=Qe(r),r)}return n},array:e=>{var t=Qe(e.length);return Le(e,t),t}},o=Oe(e),l=[],u=0;if(a)for(var h=0;h<a.length;h++){var d=i[r[h]];d?(0===u&&(u=Ye()),l[h]=d(a[h])):l[h]=a[h]}var p,c=o.apply(null,l);return p=c,0!==u&&Je(u),c=function(e){return"string"===n?K(e):"boolean"===n?!!e:e}(p),c}xe.init();var ze,Be,We,Ve=[null,ye,we,Se,$e,Re,Me],Ue={__emscripten_init_main_thread_js:function(e){He(e,!b,1,!y),xe.threadInitTLS()},__emscripten_thread_cleanup:function(e){v?postMessage({cmd:"cleanupThread",thread:e}):me(e)},__pthread_create_js:_e,_emscripten_default_pthread_stack_size:function(){return 65536},_emscripten_get_now_is_monotonic:function(){return!0},_emscripten_notify_task_queue:function(e,t,n,r){if(e==t)setTimeout(()=>Ne(r));else if(v)postMessage({targetThread:e,cmd:"processProxyingQueue",queue:r});else{var a=xe.pthreads[e];if(!a)return;a.postMessage({cmd:"processProxyingQueue",queue:r})}return 1},_emscripten_set_offscreencanvas_size:function(e,t,n){return-1},abort:function(){ue("")},emscripten_check_blocking_allowed:function(){x||b||Te("Blocking on the main thread is very dangerous, see https://emscripten.org/docs/porting/pthreads.html#blocking-on-the-main-browser-thread")},emscripten_date_now:function(){return Date.now()},emscripten_get_heap_max:function(){return 4294901760},emscripten_get_now:ke,emscripten_memcpy_big:function(e,n,r){t().copyWithin(e>>>0,n>>>0,n+r>>>0)},emscripten_num_logical_cores:function(){return x?F().cpus().length:navigator.hardwareConcurrency},emscripten_receive_on_main_thread_js:function(e,t,n){Ee.length=t;for(var r=n>>3,a=0;a<t;a++)Ee[a]=i()[r+a>>>0];return(e<0?ce[-e-1]:Ve[e]).apply(null,Ee)},emscripten_resize_heap:function(e){var n=t().length;if((e>>>=0)<=n)return!1;var r=4294901760;if(e>r)return!1;let a=(e,t)=>e+(t-e%t)%t;for(var s=1;s<=4;s*=2){var i=n*(1+.2/s);if(i=Math.min(i,e+100663296),Ae(Math.min(r,a(Math.max(e,i),65536))))return!0}return!1},emscripten_unwind_to_js_event_loop:function(){throw"unwind"},exit:be,fd_close:$e,fd_seek:Re,fd_write:Me,memory:O||h.wasmMemory},Ge=(function(){var e={env:Ue,wasi_snapshot_preview1:Ue};function t(e,t){var n=e.exports;if(h.asm=n,function(e){xe.tlsInitFunctions.push(e)}(h.asm._emscripten_tls_init),Y=h.asm.__indirect_function_table,function(e){Q.unshift(e)}(h.asm.__wasm_call_ctors),L=t,!v){var r=xe.unusedWorkers.length;xe.unusedWorkers.forEach(function(e){xe.loadWasmModuleToWorker(e,function(){--r||function(){if(ie--,h.monitorRunDependencies&&h.monitorRunDependencies(ie),0==ie&&(null!==oe&&(clearInterval(oe),oe=null),le)){var e=le;le=null,e()}}()})})}}function n(e){t(e.instance,e.module)}function r(t){return function(){if(!T&&(y||b)){if("function"==typeof fetch&&!de(se))return fetch(se,{credentials:"same-origin"}).then(function(e){if(!e.ok)throw"failed to load wasm binary file at '"+se+"'";return e.arrayBuffer()}).catch(function(){return pe(se)});if(p)return new Promise(function(e,t){p(se,function(t){e(new Uint8Array(t))},t)})}return Promise.resolve().then(function(){return pe(se)})}().then(function(t){return WebAssembly.instantiate(t,e)}).then(function(e){return e}).then(t,function(e){D("failed to asynchronously prepare wasm: "+e),ue(e)})}if(v||(ie++,h.monitorRunDependencies&&h.monitorRunDependencies(ie)),h.instantiateWasm)try{return h.instantiateWasm(e,t)}catch(e){D("Module.instantiateWasm callback failed with error: "+e),l(e)}(T||"function"!=typeof WebAssembly.instantiateStreaming||he(se)||de(se)||x||"function"!=typeof fetch?r(n):fetch(se,{credentials:"same-origin"}).then(function(t){return WebAssembly.instantiateStreaming(t,e).then(n,function(e){return D("wasm streaming compile failed: "+e),D("falling back to ArrayBuffer instantiation"),r(n)})})).catch(l)}(),h.___wasm_call_ctors=function(){return(h.___wasm_call_ctors=h.asm.__wasm_call_ctors).apply(null,arguments)},h._init=function(){return(h._init=h.asm.init).apply(null,arguments)},h._init_with_threads_count=function(){return(h._init_with_threads_count=h.asm.init_with_threads_count).apply(null,arguments)},h._get_threads_count=function(){return(h._get_threads_count=h.asm.get_threads_count).apply(null,arguments)},h._register_tensor=function(){return(h._register_tensor=h.asm.register_tensor).apply(null,arguments)},h._dispose_data=function(){return(h._dispose_data=h.asm.dispose_data).apply(null,arguments)},h._dispose=function(){return(h._dispose=h.asm.dispose).apply(null,arguments)},h._Abs=function(){return(h._Abs=h.asm.Abs).apply(null,arguments)},h._Acos=function(){return(h._Acos=h.asm.Acos).apply(null,arguments)},h._Acosh=function(){return(h._Acosh=h.asm.Acosh).apply(null,arguments)},h._Add=function(){return(h._Add=h.asm.Add).apply(null,arguments)},h._AddN=function(){return(h._AddN=h.asm.AddN).apply(null,arguments)},h._All=function(){return(h._All=h.asm.All).apply(null,arguments)},h._Any=function(){return(h._Any=h.asm.Any).apply(null,arguments)},h._ArgMax=function(){return(h._ArgMax=h.asm.ArgMax).apply(null,arguments)},h._ArgMin=function(){return(h._ArgMin=h.asm.ArgMin).apply(null,arguments)},h._Asin=function(){return(h._Asin=h.asm.Asin).apply(null,arguments)},h._Asinh=function(){return(h._Asinh=h.asm.Asinh).apply(null,arguments)},h._Atan=function(){return(h._Atan=h.asm.Atan).apply(null,arguments)},h._Atan2=function(){return(h._Atan2=h.asm.Atan2).apply(null,arguments)},h._Atanh=function(){return(h._Atanh=h.asm.Atanh).apply(null,arguments)},h._AvgPool=function(){return(h._AvgPool=h.asm.AvgPool).apply(null,arguments)},h._AvgPool3D=function(){return(h._AvgPool3D=h.asm.AvgPool3D).apply(null,arguments)},h._AvgPool3DGrad=function(){return(h._AvgPool3DGrad=h.asm.AvgPool3DGrad).apply(null,arguments)},h._AvgPoolGrad=function(){return(h._AvgPoolGrad=h.asm.AvgPoolGrad).apply(null,arguments)},h._BatchMatMul=function(){return(h._BatchMatMul=h.asm.BatchMatMul).apply(null,arguments)},h._Bincount=function(){return(h._Bincount=h.asm.Bincount).apply(null,arguments)},h._BitwiseAnd=function(){return(h._BitwiseAnd=h.asm.BitwiseAnd).apply(null,arguments)},h._Ceil=function(){return(h._Ceil=h.asm.Ceil).apply(null,arguments)},h._ClipByValue=function(){return(h._ClipByValue=h.asm.ClipByValue).apply(null,arguments)},h._Conv2D=function(){return(h._Conv2D=h.asm.Conv2D).apply(null,arguments)},h._Conv2DBackpropInput=function(){return(h._Conv2DBackpropInput=h.asm.Conv2DBackpropInput).apply(null,arguments)},h._Conv3D=function(){return(h._Conv3D=h.asm.Conv3D).apply(null,arguments)},h._Conv3DBackpropFilterV2=function(){return(h._Conv3DBackpropFilterV2=h.asm.Conv3DBackpropFilterV2).apply(null,arguments)},h._Conv3DBackpropInputV2=function(){return(h._Conv3DBackpropInputV2=h.asm.Conv3DBackpropInputV2).apply(null,arguments)},h._Cos=function(){return(h._Cos=h.asm.Cos).apply(null,arguments)},h._Cosh=function(){return(h._Cosh=h.asm.Cosh).apply(null,arguments)},h._CropAndResize=function(){return(h._CropAndResize=h.asm.CropAndResize).apply(null,arguments)},h._Cumprod=function(){return(h._Cumprod=h.asm.Cumprod).apply(null,arguments)},h._Cumsum=function(){return(h._Cumsum=h.asm.Cumsum).apply(null,arguments)},h._DenseBincount=function(){return(h._DenseBincount=h.asm.DenseBincount).apply(null,arguments)},h._DepthToSpace=function(){return(h._DepthToSpace=h.asm.DepthToSpace).apply(null,arguments)},h._DepthwiseConv2dNative=function(){return(h._DepthwiseConv2dNative=h.asm.DepthwiseConv2dNative).apply(null,arguments)},h._Diag=function(){return(h._Diag=h.asm.Diag).apply(null,arguments)},h._Dilation2D=function(){return(h._Dilation2D=h.asm.Dilation2D).apply(null,arguments)},h._Dilation2DBackpropFilter=function(){return(h._Dilation2DBackpropFilter=h.asm.Dilation2DBackpropFilter).apply(null,arguments)},h._Dilation2DBackpropInput=function(){return(h._Dilation2DBackpropInput=h.asm.Dilation2DBackpropInput).apply(null,arguments)},h._Elu=function(){return(h._Elu=h.asm.Elu).apply(null,arguments)},h._EluGrad=function(){return(h._EluGrad=h.asm.EluGrad).apply(null,arguments)},h._Equal=function(){return(h._Equal=h.asm.Equal).apply(null,arguments)},h._Erf=function(){return(h._Erf=h.asm.Erf).apply(null,arguments)},h._Exp=function(){return(h._Exp=h.asm.Exp).apply(null,arguments)},h._Expm1=function(){return(h._Expm1=h.asm.Expm1).apply(null,arguments)},h._FlipLeftRight=function(){return(h._FlipLeftRight=h.asm.FlipLeftRight).apply(null,arguments)},h._Floor=function(){return(h._Floor=h.asm.Floor).apply(null,arguments)},h._FloorDiv=function(){return(h._FloorDiv=h.asm.FloorDiv).apply(null,arguments)},h._FusedBatchNorm=function(){return(h._FusedBatchNorm=h.asm.FusedBatchNorm).apply(null,arguments)},h._FusedConv2D=function(){return(h._FusedConv2D=h.asm.FusedConv2D).apply(null,arguments)},h._FusedDepthwiseConv2D=function(){return(h._FusedDepthwiseConv2D=h.asm.FusedDepthwiseConv2D).apply(null,arguments)},h._Gather=function(){return(h._Gather=h.asm.Gather).apply(null,arguments)},h._GatherNd=function(){return(h._GatherNd=h.asm.GatherNd).apply(null,arguments)},h._Greater=function(){return(h._Greater=h.asm.Greater).apply(null,arguments)},h._GreaterEqual=function(){return(h._GreaterEqual=h.asm.GreaterEqual).apply(null,arguments)},h._IsFinite=function(){return(h._IsFinite=h.asm.IsFinite).apply(null,arguments)},h._IsInf=function(){return(h._IsInf=h.asm.IsInf).apply(null,arguments)},h._IsNan=function(){return(h._IsNan=h.asm.IsNan).apply(null,arguments)},h._LRN=function(){return(h._LRN=h.asm.LRN).apply(null,arguments)},h._LRNGrad=function(){return(h._LRNGrad=h.asm.LRNGrad).apply(null,arguments)},h._LeakyRelu=function(){return(h._LeakyRelu=h.asm.LeakyRelu).apply(null,arguments)},h._Less=function(){return(h._Less=h.asm.Less).apply(null,arguments)},h._LessEqual=function(){return(h._LessEqual=h.asm.LessEqual).apply(null,arguments)},h._LinSpace=function(){return(h._LinSpace=h.asm.LinSpace).apply(null,arguments)},h._Log=function(){return(h._Log=h.asm.Log).apply(null,arguments)},h._Log1p=function(){return(h._Log1p=h.asm.Log1p).apply(null,arguments)},h._LogicalAnd=function(){return(h._LogicalAnd=h.asm.LogicalAnd).apply(null,arguments)},h._LogicalNot=function(){return(h._LogicalNot=h.asm.LogicalNot).apply(null,arguments)},h._LogicalOr=function(){return(h._LogicalOr=h.asm.LogicalOr).apply(null,arguments)},h._LogicalXor=function(){return(h._LogicalXor=h.asm.LogicalXor).apply(null,arguments)},h._Max=function(){return(h._Max=h.asm.Max).apply(null,arguments)},h._MaxPool=function(){return(h._MaxPool=h.asm.MaxPool).apply(null,arguments)},h._MaxPool3D=function(){return(h._MaxPool3D=h.asm.MaxPool3D).apply(null,arguments)},h._MaxPool3DGrad=function(){return(h._MaxPool3DGrad=h.asm.MaxPool3DGrad).apply(null,arguments)},h._MaxPoolGrad=function(){return(h._MaxPoolGrad=h.asm.MaxPoolGrad).apply(null,arguments)},h._MaxPoolWithArgmax=function(){return(h._MaxPoolWithArgmax=h.asm.MaxPoolWithArgmax).apply(null,arguments)},h._Maximum=function(){return(h._Maximum=h.asm.Maximum).apply(null,arguments)},h._Mean=function(){return(h._Mean=h.asm.Mean).apply(null,arguments)},h._Min=function(){return(h._Min=h.asm.Min).apply(null,arguments)},h._Minimum=function(){return(h._Minimum=h.asm.Minimum).apply(null,arguments)},h._MirrorPad=function(){return(h._MirrorPad=h.asm.MirrorPad).apply(null,arguments)},h._Mod=function(){return(h._Mod=h.asm.Mod).apply(null,arguments)},h._Multinomial=function(){return(h._Multinomial=h.asm.Multinomial).apply(null,arguments)},h._Multiply=function(){return(h._Multiply=h.asm.Multiply).apply(null,arguments)},h._Neg=function(){return(h._Neg=h.asm.Neg).apply(null,arguments)},h._NonMaxSuppressionV3=function(){return(h._NonMaxSuppressionV3=h.asm.NonMaxSuppressionV3).apply(null,arguments)},h._NonMaxSuppressionV4=function(){return(h._NonMaxSuppressionV4=h.asm.NonMaxSuppressionV4).apply(null,arguments)},h._NonMaxSuppressionV5=function(){return(h._NonMaxSuppressionV5=h.asm.NonMaxSuppressionV5).apply(null,arguments)},h._NotEqual=function(){return(h._NotEqual=h.asm.NotEqual).apply(null,arguments)},h._OneHot=function(){return(h._OneHot=h.asm.OneHot).apply(null,arguments)},h._PadV2=function(){return(h._PadV2=h.asm.PadV2).apply(null,arguments)},h._Pow=function(){return(h._Pow=h.asm.Pow).apply(null,arguments)},h._Prelu=function(){return(h._Prelu=h.asm.Prelu).apply(null,arguments)},h._Prod=function(){return(h._Prod=h.asm.Prod).apply(null,arguments)},h._RealDiv=function(){return(h._RealDiv=h.asm.RealDiv).apply(null,arguments)},h._Reciprocal=function(){return(h._Reciprocal=h.asm.Reciprocal).apply(null,arguments)},h._Relu=function(){return(h._Relu=h.asm.Relu).apply(null,arguments)},h._Relu6=function(){return(h._Relu6=h.asm.Relu6).apply(null,arguments)},h._ResizeBilinear=function(){return(h._ResizeBilinear=h.asm.ResizeBilinear).apply(null,arguments)},h._ResizeBilinearGrad=function(){return(h._ResizeBilinearGrad=h.asm.ResizeBilinearGrad).apply(null,arguments)},h._ResizeNearestNeighbor=function(){return(h._ResizeNearestNeighbor=h.asm.ResizeNearestNeighbor).apply(null,arguments)},h._ResizeNearestNeighborGrad=function(){return(h._ResizeNearestNeighborGrad=h.asm.ResizeNearestNeighborGrad).apply(null,arguments)},h._Reverse=function(){return(h._Reverse=h.asm.Reverse).apply(null,arguments)},h._RotateWithOffset=function(){return(h._RotateWithOffset=h.asm.RotateWithOffset).apply(null,arguments)},h._Round=function(){return(h._Round=h.asm.Round).apply(null,arguments)},h._Rsqrt=function(){return(h._Rsqrt=h.asm.Rsqrt).apply(null,arguments)},h._ScatterNd=function(){return(h._ScatterNd=h.asm.ScatterNd).apply(null,arguments)},h._SearchSorted=function(){return(h._SearchSorted=h.asm.SearchSorted).apply(null,arguments)},h._SelectV2=function(){return(h._SelectV2=h.asm.SelectV2).apply(null,arguments)},h._Selu=function(){return(h._Selu=h.asm.Selu).apply(null,arguments)},h._Sigmoid=function(){return(h._Sigmoid=h.asm.Sigmoid).apply(null,arguments)},h._Sign=function(){return(h._Sign=h.asm.Sign).apply(null,arguments)},h._Sin=function(){return(h._Sin=h.asm.Sin).apply(null,arguments)},h._Sinh=function(){return(h._Sinh=h.asm.Sinh).apply(null,arguments)},h._Softmax=function(){return(h._Softmax=h.asm.Softmax).apply(null,arguments)},h._Softplus=function(){return(h._Softplus=h.asm.Softplus).apply(null,arguments)},h._SparseFillEmptyRows=function(){return(h._SparseFillEmptyRows=h.asm.SparseFillEmptyRows).apply(null,arguments)},h._SparseReshape=function(){return(h._SparseReshape=h.asm.SparseReshape).apply(null,arguments)},h._SparseSegmentReduction=function(){return(h._SparseSegmentReduction=h.asm.SparseSegmentReduction).apply(null,arguments)},h._SparseToDense=function(){return(h._SparseToDense=h.asm.SparseToDense).apply(null,arguments)},h._Sqrt=function(){return(h._Sqrt=h.asm.Sqrt).apply(null,arguments)},h._Square=function(){return(h._Square=h.asm.Square).apply(null,arguments)},h._SquaredDifference=function(){return(h._SquaredDifference=h.asm.SquaredDifference).apply(null,arguments)},h._Step=function(){return(h._Step=h.asm.Step).apply(null,arguments)},h._StridedSlice=function(){return(h._StridedSlice=h.asm.StridedSlice).apply(null,arguments)},h._Sub=function(){return(h._Sub=h.asm.Sub).apply(null,arguments)},h._Sum=function(){return(h._Sum=h.asm.Sum).apply(null,arguments)},h._Tan=function(){return(h._Tan=h.asm.Tan).apply(null,arguments)},h._Tanh=function(){return(h._Tanh=h.asm.Tanh).apply(null,arguments)},h._TensorScatterUpdate=function(){return(h._TensorScatterUpdate=h.asm.TensorScatterUpdate).apply(null,arguments)},h._Tile=function(){return(h._Tile=h.asm.Tile).apply(null,arguments)},h._TopK=function(){return(h._TopK=h.asm.TopK).apply(null,arguments)},h._Transform=function(){return(h._Transform=h.asm.Transform).apply(null,arguments)},h._Transpose=function(){return(h._Transpose=h.asm.Transpose).apply(null,arguments)},h.__FusedMatMul=function(){return(h.__FusedMatMul=h.asm._FusedMatMul).apply(null,arguments)},h._malloc=function(){return(h._malloc=h.asm.malloc).apply(null,arguments)},h._free=function(){return(h._free=h.asm.free).apply(null,arguments)},h.__emscripten_tls_init=function(){return(h.__emscripten_tls_init=h.asm._emscripten_tls_init).apply(null,arguments)},h._pthread_self=function(){return(Ge=h._pthread_self=h.asm.pthread_self).apply(null,arguments)}),He=(h.___errno_location=function(){return(h.___errno_location=h.asm.__errno_location).apply(null,arguments)},h.__emscripten_thread_init=function(){return(He=h.__emscripten_thread_init=h.asm._emscripten_thread_init).apply(null,arguments)}),je=(h.__emscripten_thread_crashed=function(){return(h.__emscripten_thread_crashed=h.asm._emscripten_thread_crashed).apply(null,arguments)},h._emscripten_main_thread_process_queued_calls=function(){return(h._emscripten_main_thread_process_queued_calls=h.asm.emscripten_main_thread_process_queued_calls).apply(null,arguments)},h._emscripten_main_browser_thread_id=function(){return(h._emscripten_main_browser_thread_id=h.asm.emscripten_main_browser_thread_id).apply(null,arguments)},h._emscripten_run_in_main_runtime_thread_js=function(){return(je=h._emscripten_run_in_main_runtime_thread_js=h.asm.emscripten_run_in_main_runtime_thread_js).apply(null,arguments)}),qe=(h._emscripten_dispatch_to_thread_=function(){return(h._emscripten_dispatch_to_thread_=h.asm.emscripten_dispatch_to_thread_).apply(null,arguments)},h.__emscripten_proxy_execute_task_queue=function(){return(qe=h.__emscripten_proxy_execute_task_queue=h.asm._emscripten_proxy_execute_task_queue).apply(null,arguments)}),Ke=h.__emscripten_thread_free_data=function(){return(Ke=h.__emscripten_thread_free_data=h.asm._emscripten_thread_free_data).apply(null,arguments)},Xe=h.__emscripten_thread_exit=function(){return(Xe=h.__emscripten_thread_exit=h.asm._emscripten_thread_exit).apply(null,arguments)},Ze=h._emscripten_stack_set_limits=function(){return(Ze=h._emscripten_stack_set_limits=h.asm.emscripten_stack_set_limits).apply(null,arguments)},Ye=h.stackSave=function(){return(Ye=h.stackSave=h.asm.stackSave).apply(null,arguments)},Je=h.stackRestore=function(){return(Je=h.stackRestore=h.asm.stackRestore).apply(null,arguments)},Qe=h.stackAlloc=function(){return(Qe=h.stackAlloc=h.asm.stackAlloc).apply(null,arguments)};function et(e){if(e=e||m,!(ie>0)){if(v)return o(h),ne(),void startWorker(h);(function(){if(h.preRun)for("function"==typeof h.preRun&&(h.preRun=[h.preRun]);h.preRun.length;)re(h.preRun.shift());ve(J)})(),ie>0||(h.setStatus?(h.setStatus("Running..."),setTimeout(function(){setTimeout(function(){h.setStatus("")},1),t()},1)):t())}function t(){ze||(ze=!0,h.calledRun=!0,!H&&(ne(),o(h),h.onRuntimeInitialized&&h.onRuntimeInitialized(),function(){if(!v){if(h.postRun)for("function"==typeof h.postRun&&(h.postRun=[h.postRun]);h.postRun.length;)ae(h.postRun.shift());ve(ee)}}()))}}if(h.dynCall_iijjiiii=function(){return(h.dynCall_iijjiiii=h.asm.dynCall_iijjiiii).apply(null,arguments)},h.dynCall_jiji=function(){return(h.dynCall_jiji=h.asm.dynCall_jiji).apply(null,arguments)},h.keepRuntimeAlive=te,h.wasmMemory=O,h.cwrap=function(e,t,n,r){var a=(n=n||[]).every(e=>"number"===e||"boolean"===e);return"string"!==t&&a&&!r?Oe(e):function(){return Pe(e,t,n,arguments)}},h.ExitStatus=fe,h.PThread=xe,le=function e(){ze||et(),ze||(le=e)},h.preInit)for("function"==typeof h.preInit&&(h.preInit=[h.preInit]);h.preInit.length>0;)h.preInit.pop()();if(et(),u&&(Be={uncaughtException:process.listeners("uncaughtException").filter(function(e){return!u.uncaughtException.indexOf(e)>-1}),unhandledRejection:process.listeners("unhandledRejection").filter(function(e){return!u.unhandledRejection.indexOf(e)>-1})}),"undefined"!=typeof WasmBackendModule)We=WasmBackendModule;else{if(void 0===e)throw new Error("Could not find wasm module in post.js");We=e}if(Be){var tt=We._dispose;We._dispose=function(){tt(),Be.uncaughtException.forEach(function(e){process.removeListener("uncaughtException",e)}),Be.unhandledRejection.forEach(function(e){process.removeListener("unhandledRejection",e)})}}return e.ready});"object"==typeof e&&"object"==typeof t?t.exports=s:"function"==typeof define&&n.amdO?define([],function(){return s}):"object"==typeof e&&(e.WasmBackendModuleThreadedSimd=s)}),M=c((e,t)=>{t.exports.wasmWorkerContents='"use strict";var Module={};var ENVIRONMENT_IS_NODE=typeof process=="object"&&typeof process.versions=="object"&&typeof process.versions.node=="string";if(ENVIRONMENT_IS_NODE){var nodeWorkerThreads=require("worker_threads");var parentPort=nodeWorkerThreads.parentPort;parentPort.on("message",data=>onmessage({data:data}));var fs=require("fs");Object.assign(global,{self:global,require:require,Module:Module,location:{href:__filename},Worker:nodeWorkerThreads.Worker,importScripts:function(f){(0,eval)(fs.readFileSync(f,"utf8")+"//# sourceURL="+f)},postMessage:function(msg){parentPort.postMessage(msg)},performance:global.performance||{now:function(){return Date.now()}}})}var initializedJS=false;var pendingNotifiedProxyingQueues=[];function threadPrintErr(){var text=Array.prototype.slice.call(arguments).join(" ");if(ENVIRONMENT_IS_NODE){fs.writeSync(2,text+"\n");return}console.error(text)}function threadAlert(){var text=Array.prototype.slice.call(arguments).join(" ");postMessage({cmd:"alert",text:text,threadId:Module["_pthread_self"]()})}var err=threadPrintErr;self.alert=threadAlert;Module["instantiateWasm"]=(info,receiveInstance)=>{var instance=new WebAssembly.Instance(Module["wasmModule"],info);receiveInstance(instance);Module["wasmModule"]=null;return instance.exports};self.onunhandledrejection=e=>{throw e.reason??e};self.startWorker=instance=>{Module=instance;postMessage({"cmd":"loaded"})};self.onmessage=e=>{try{if(e.data.cmd==="load"){Module["wasmModule"]=e.data.wasmModule;for(const handler of e.data.handlers){Module[handler]=function(){postMessage({cmd:"callHandler",handler:handler,args:[...arguments]})}}Module["wasmMemory"]=e.data.wasmMemory;Module["buffer"]=Module["wasmMemory"].buffer;Module["ENVIRONMENT_IS_PTHREAD"]=true;if(typeof e.data.urlOrBlob=="string"){importScripts(e.data.urlOrBlob)}else{var objectUrl=URL.createObjectURL(e.data.urlOrBlob);importScripts(objectUrl);URL.revokeObjectURL(objectUrl)}WasmBackendModuleThreadedSimd(Module)}else if(e.data.cmd==="run"){Module["__emscripten_thread_init"](e.data.pthread_ptr,0,0,1);Module["establishStackSpace"]();Module["PThread"].receiveObjectTransfer(e.data);Module["PThread"].threadInitTLS();if(!initializedJS){pendingNotifiedProxyingQueues.forEach(queue=>{Module["executeNotifiedProxyingQueue"](queue)});pendingNotifiedProxyingQueues=[];initializedJS=true}try{Module["invokeEntryPoint"](e.data.start_routine,e.data.arg)}catch(ex){if(ex!="unwind"){if(ex instanceof Module["ExitStatus"]){if(Module["keepRuntimeAlive"]()){}else{Module["__emscripten_thread_exit"](ex.status)}}else{throw ex}}}}else if(e.data.cmd==="cancel"){if(Module["_pthread_self"]()){Module["__emscripten_thread_exit"](-1)}}else if(e.data.target==="setimmediate"){}else if(e.data.cmd==="processProxyingQueue"){if(initializedJS){Module["executeNotifiedProxyingQueue"](e.data.queue)}else{pendingNotifiedProxyingQueues.push(e.data.queue)}}else if(e.data.cmd){err("worker.js received unknown command "+e.data.cmd);err(e.data)}}catch(ex){if(Module["__emscripten_thread_crashed"]){Module["__emscripten_thread_crashed"]()}throw ex}};'}),O=c((e,t)=>{var a,s=(a=(a="undefined"!=typeof document&&document.currentScript?document.currentScript.src:void 0)||r,function(e){var t,n,r,s=void 0!==(e=e||{})?e:{};s.ready=new Promise(function(e,r){t=e,n=r}),"undefined"!=typeof process&&process.listeners&&(r={uncaughtException:process.listeners("uncaughtException"),unhandledRejection:process.listeners("unhandledRejection")});var i,o,l,u=Object.assign({},s),h=[],d="object"==typeof window,p="function"==typeof importScripts,c="object"==typeof process&&"object"==typeof process.versions&&"string"==typeof process.versions.node,f="";if(c){var m=E(),g=A();f=p?g.dirname(f)+"/":"//",i=(e,t)=>(e=G(e)?new URL(e):g.normalize(e),m.readFileSync(e,t?void 0:"utf8")),l=e=>{var t=i(e,!0);return t.buffer||(t=new Uint8Array(t)),t},o=(e,t,n)=>{e=G(e)?new URL(e):g.normalize(e),m.readFile(e,function(e,r){e?n(e):t(r.buffer)})},process.argv.length>1&&process.argv[1].replace(/\\/g,"/"),h=process.argv.slice(2),process.on("uncaughtException",function(e){if(!(e instanceof j))throw e}),process.on("unhandledRejection",function(e){throw e}),s.inspect=function(){return"[Emscripten Module object]"}}else(d||p)&&(p?f=self.location.href:"undefined"!=typeof document&&document.currentScript&&(f=document.currentScript.src),a&&(f=a),f=0!==f.indexOf("blob:")?f.substr(0,f.replace(/[?#].*/,"").lastIndexOf("/")+1):"",i=e=>{var t=new XMLHttpRequest;return t.open("GET",e,!1),t.send(null),t.responseText},p&&(l=e=>{var t=new XMLHttpRequest;return t.open("GET",e,!1),t.responseType="arraybuffer",t.send(null),new Uint8Array(t.response)}),o=(e,t,n)=>{var r=new XMLHttpRequest;r.open("GET",e,!0),r.responseType="arraybuffer",r.onload=()=>{200==r.status||0==r.status&&r.response?t(r.response):n()},r.onerror=n,r.send(null)});var y,b=s.print||console.log.bind(console),x=s.printErr||console.warn.bind(console);Object.assign(s,u),u=null,s.arguments&&(h=s.arguments),s.thisProgram&&s.thisProgram,s.quit&&s.quit,s.wasmBinary&&(y=s.wasmBinary);s.noExitRuntime;"object"!=typeof WebAssembly&&V("no native wasm support detected");var v,w,k,I,S,_=!1,N="undefined"!=typeof TextDecoder?new TextDecoder("utf8"):void 0;function T(e,t,n){for(var r=(t>>>=0)+n,a=t;e[a]&&!(a>=r);)++a;if(a-t>16&&e.buffer&&N)return N.decode(e.subarray(t,a));for(var s="";t<a;){var i=e[t++];if(128&i){var o=63&e[t++];if(192!=(224&i)){var l=63&e[t++];if((i=224==(240&i)?(15&i)<<12|o<<6|l:(7&i)<<18|o<<12|l<<6|63&e[t++])<65536)s+=String.fromCharCode(i);else{var u=i-65536;s+=String.fromCharCode(55296|u>>10,56320|1023&u)}}else s+=String.fromCharCode((31&i)<<6|o)}else s+=String.fromCharCode(i)}return s}function C(e,t){return(e>>>=0)?T(I,e,t):""}function $(e){w=e,s.HEAP8=k=new Int8Array(e),s.HEAP16=new Int16Array(e),s.HEAP32=new Int32Array(e),s.HEAPU8=I=new Uint8Array(e),s.HEAPU16=new Uint16Array(e),s.HEAPU32=S=new Uint32Array(e),s.HEAPF32=new Float32Array(e),s.HEAPF64=new Float64Array(e)}s.INITIAL_MEMORY;var R=[],F=[],D=[];function M(e){R.unshift(e)}function O(e){D.unshift(e)}var L,P,z=0,B=null,W=null;function V(e){s.onAbort&&s.onAbort(e),x(e="Aborted("+e+")"),_=!0,e+=". Build with -sASSERTIONS for more info.";var t=new WebAssembly.RuntimeError(e);throw n(t),t}function U(e){return e.startsWith("data:application/octet-stream;base64,")}function G(e){return e.startsWith("file://")}function H(e){try{if(e==L&&y)return new Uint8Array(y);if(l)return l(e);throw"both async and sync fetching of the wasm failed"}catch(e){V(e)}}function j(e){this.name="ExitStatus",this.message="Program terminated with exit("+e+")",this.status=e}function q(e){for(;e.length>0;)e.shift()(s)}function K(e){try{return v.grow(e-w.byteLength+65535>>>16),$(v.buffer),1}catch(e){}}U(L="tfjs-backend-wasm.wasm")||(P=L,L=s.locateFile?s.locateFile(P,f):f+P);var X=[null,[],[]];function Z(e,t){var n=X[e];0===t||10===t?((1===e?b:x)(T(n,0)),n.length=0):n.push(t)}function Y(e){return s["_"+e]}function J(e,t,n,r,a){var s={string:e=>{var t=0;if(null!=e&&0!==e){var n=1+(e.length<<2);!function(e,t,n){!function(e,t,n,r){if(!(r>0))return 0;n>>>=0;for(var a=n+r-1,s=0;s<e.length;++s){var i=e.charCodeAt(s);if(i>=55296&&i<=57343&&(i=65536+((1023&i)<<10)|1023&e.charCodeAt(++s)),i<=127){if(n>=a)break;t[n++>>>0]=i}else if(i<=2047){if(n+1>=a)break;t[n++>>>0]=192|i>>6,t[n++>>>0]=128|63&i}else if(i<=65535){if(n+2>=a)break;t[n++>>>0]=224|i>>12,t[n++>>>0]=128|i>>6&63,t[n++>>>0]=128|63&i}else{if(n+3>=a)break;t[n++>>>0]=240|i>>18,t[n++>>>0]=128|i>>12&63,t[n++>>>0]=128|i>>6&63,t[n++>>>0]=128|63&i}}t[n>>>0]=0}(e,I,t,n)}(e,t=se(n),n)}return t},array:e=>{var t=se(e.length);return function(e,t){k.set(e,t>>>0)}(e,t),t}},i=Y(e),o=[],l=0;if(r)for(var u=0;u<r.length;u++){var h=s[n[u]];h?(0===l&&(l=re()),o[u]=h(r[u])):o[u]=r[u]}var d,p=i.apply(null,o);return d=p,0!==l&&ae(l),p=function(e){return"string"===t?C(e):"boolean"===t?!!e:e}(d),p}var Q,ee,te,ne={abort:function(){V("")},emscripten_get_heap_max:function(){return 4294901760},emscripten_memcpy_big:function(e,t,n){I.copyWithin(e>>>0,t>>>0,t+n>>>0)},emscripten_resize_heap:function(e){var t=I.length,n=4294901760;if((e>>>=0)>n)return!1;let r=(e,t)=>e+(t-e%t)%t;for(var a=1;a<=4;a*=2){var s=t*(1+.2/a);if(s=Math.min(s,e+100663296),K(Math.min(n,r(Math.max(e,s),65536))))return!0}return!1},fd_close:function(e){return 52},fd_seek:function(e,t,n,r,a){return 70},fd_write:function(e,t,n,r){for(var a=0,s=0;s<n;s++){var i=S[t>>>2],o=S[t+4>>>2];t+=8;for(var l=0;l<o;l++)Z(e,I[i+l>>>0]);a+=o}return S[r>>>2]=a,0}},re=(function(){var e={env:ne,wasi_snapshot_preview1:ne};function t(e,t){var n=e.exports;s.asm=n,$((v=s.asm.memory).buffer),s.asm.__indirect_function_table,function(e){F.unshift(e)}(s.asm.__wasm_call_ctors),function(){if(z--,s.monitorRunDependencies&&s.monitorRunDependencies(z),0==z&&(null!==B&&(clearInterval(B),B=null),W)){var e=W;W=null,e()}}()}function r(e){t(e.instance)}function a(t){return function(){if(!y&&(d||p)){if("function"==typeof fetch&&!G(L))return fetch(L,{credentials:"same-origin"}).then(function(e){if(!e.ok)throw"failed to load wasm binary file at '"+L+"'";return e.arrayBuffer()}).catch(function(){return H(L)});if(o)return new Promise(function(e,t){o(L,function(t){e(new Uint8Array(t))},t)})}return Promise.resolve().then(function(){return H(L)})}().then(function(t){return WebAssembly.instantiate(t,e)}).then(function(e){return e}).then(t,function(e){x("failed to asynchronously prepare wasm: "+e),V(e)})}if(z++,s.monitorRunDependencies&&s.monitorRunDependencies(z),s.instantiateWasm)try{return s.instantiateWasm(e,t)}catch(e){x("Module.instantiateWasm callback failed with error: "+e),n(e)}(y||"function"!=typeof WebAssembly.instantiateStreaming||U(L)||G(L)||c||"function"!=typeof fetch?a(r):fetch(L,{credentials:"same-origin"}).then(function(t){return WebAssembly.instantiateStreaming(t,e).then(r,function(e){return x("wasm streaming compile failed: "+e),x("falling back to ArrayBuffer instantiation"),a(r)})})).catch(n)}(),s.___wasm_call_ctors=function(){return(s.___wasm_call_ctors=s.asm.__wasm_call_ctors).apply(null,arguments)},s._init=function(){return(s._init=s.asm.init).apply(null,arguments)},s._init_with_threads_count=function(){return(s._init_with_threads_count=s.asm.init_with_threads_count).apply(null,arguments)},s._get_threads_count=function(){return(s._get_threads_count=s.asm.get_threads_count).apply(null,arguments)},s._register_tensor=function(){return(s._register_tensor=s.asm.register_tensor).apply(null,arguments)},s._dispose_data=function(){return(s._dispose_data=s.asm.dispose_data).apply(null,arguments)},s._dispose=function(){return(s._dispose=s.asm.dispose).apply(null,arguments)},s._Abs=function(){return(s._Abs=s.asm.Abs).apply(null,arguments)},s._Acos=function(){return(s._Acos=s.asm.Acos).apply(null,arguments)},s._Acosh=function(){return(s._Acosh=s.asm.Acosh).apply(null,arguments)},s._Add=function(){return(s._Add=s.asm.Add).apply(null,arguments)},s._AddN=function(){return(s._AddN=s.asm.AddN).apply(null,arguments)},s._All=function(){return(s._All=s.asm.All).apply(null,arguments)},s._Any=function(){return(s._Any=s.asm.Any).apply(null,arguments)},s._ArgMax=function(){return(s._ArgMax=s.asm.ArgMax).apply(null,arguments)},s._ArgMin=function(){return(s._ArgMin=s.asm.ArgMin).apply(null,arguments)},s._Asin=function(){return(s._Asin=s.asm.Asin).apply(null,arguments)},s._Asinh=function(){return(s._Asinh=s.asm.Asinh).apply(null,arguments)},s._Atan=function(){return(s._Atan=s.asm.Atan).apply(null,arguments)},s._Atan2=function(){return(s._Atan2=s.asm.Atan2).apply(null,arguments)},s._Atanh=function(){return(s._Atanh=s.asm.Atanh).apply(null,arguments)},s._AvgPool=function(){return(s._AvgPool=s.asm.AvgPool).apply(null,arguments)},s._AvgPool3D=function(){return(s._AvgPool3D=s.asm.AvgPool3D).apply(null,arguments)},s._AvgPool3DGrad=function(){return(s._AvgPool3DGrad=s.asm.AvgPool3DGrad).apply(null,arguments)},s._AvgPoolGrad=function(){return(s._AvgPoolGrad=s.asm.AvgPoolGrad).apply(null,arguments)},s._BatchMatMul=function(){return(s._BatchMatMul=s.asm.BatchMatMul).apply(null,arguments)},s._Bincount=function(){return(s._Bincount=s.asm.Bincount).apply(null,arguments)},s._BitwiseAnd=function(){return(s._BitwiseAnd=s.asm.BitwiseAnd).apply(null,arguments)},s._Ceil=function(){return(s._Ceil=s.asm.Ceil).apply(null,arguments)},s._ClipByValue=function(){return(s._ClipByValue=s.asm.ClipByValue).apply(null,arguments)},s._Conv2D=function(){return(s._Conv2D=s.asm.Conv2D).apply(null,arguments)},s._Conv2DBackpropInput=function(){return(s._Conv2DBackpropInput=s.asm.Conv2DBackpropInput).apply(null,arguments)},s._Conv3D=function(){return(s._Conv3D=s.asm.Conv3D).apply(null,arguments)},s._Conv3DBackpropFilterV2=function(){return(s._Conv3DBackpropFilterV2=s.asm.Conv3DBackpropFilterV2).apply(null,arguments)},s._Conv3DBackpropInputV2=function(){return(s._Conv3DBackpropInputV2=s.asm.Conv3DBackpropInputV2).apply(null,arguments)},s._Cos=function(){return(s._Cos=s.asm.Cos).apply(null,arguments)},s._Cosh=function(){return(s._Cosh=s.asm.Cosh).apply(null,arguments)},s._CropAndResize=function(){return(s._CropAndResize=s.asm.CropAndResize).apply(null,arguments)},s._Cumprod=function(){return(s._Cumprod=s.asm.Cumprod).apply(null,arguments)},s._Cumsum=function(){return(s._Cumsum=s.asm.Cumsum).apply(null,arguments)},s._DenseBincount=function(){return(s._DenseBincount=s.asm.DenseBincount).apply(null,arguments)},s._DepthToSpace=function(){return(s._DepthToSpace=s.asm.DepthToSpace).apply(null,arguments)},s._DepthwiseConv2dNative=function(){return(s._DepthwiseConv2dNative=s.asm.DepthwiseConv2dNative).apply(null,arguments)},s._Diag=function(){return(s._Diag=s.asm.Diag).apply(null,arguments)},s._Dilation2D=function(){return(s._Dilation2D=s.asm.Dilation2D).apply(null,arguments)},s._Dilation2DBackpropFilter=function(){return(s._Dilation2DBackpropFilter=s.asm.Dilation2DBackpropFilter).apply(null,arguments)},s._Dilation2DBackpropInput=function(){return(s._Dilation2DBackpropInput=s.asm.Dilation2DBackpropInput).apply(null,arguments)},s._Elu=function(){return(s._Elu=s.asm.Elu).apply(null,arguments)},s._EluGrad=function(){return(s._EluGrad=s.asm.EluGrad).apply(null,arguments)},s._Equal=function(){return(s._Equal=s.asm.Equal).apply(null,arguments)},s._Erf=function(){return(s._Erf=s.asm.Erf).apply(null,arguments)},s._Exp=function(){return(s._Exp=s.asm.Exp).apply(null,arguments)},s._Expm1=function(){return(s._Expm1=s.asm.Expm1).apply(null,arguments)},s._FlipLeftRight=function(){return(s._FlipLeftRight=s.asm.FlipLeftRight).apply(null,arguments)},s._Floor=function(){return(s._Floor=s.asm.Floor).apply(null,arguments)},s._FloorDiv=function(){return(s._FloorDiv=s.asm.FloorDiv).apply(null,arguments)},s._FusedBatchNorm=function(){return(s._FusedBatchNorm=s.asm.FusedBatchNorm).apply(null,arguments)},s._FusedConv2D=function(){return(s._FusedConv2D=s.asm.FusedConv2D).apply(null,arguments)},s._FusedDepthwiseConv2D=function(){return(s._FusedDepthwiseConv2D=s.asm.FusedDepthwiseConv2D).apply(null,arguments)},s._Gather=function(){return(s._Gather=s.asm.Gather).apply(null,arguments)},s._GatherNd=function(){return(s._GatherNd=s.asm.GatherNd).apply(null,arguments)},s._Greater=function(){return(s._Greater=s.asm.Greater).apply(null,arguments)},s._GreaterEqual=function(){return(s._GreaterEqual=s.asm.GreaterEqual).apply(null,arguments)},s._IsFinite=function(){return(s._IsFinite=s.asm.IsFinite).apply(null,arguments)},s._IsInf=function(){return(s._IsInf=s.asm.IsInf).apply(null,arguments)},s._IsNan=function(){return(s._IsNan=s.asm.IsNan).apply(null,arguments)},s._LRN=function(){return(s._LRN=s.asm.LRN).apply(null,arguments)},s._LRNGrad=function(){return(s._LRNGrad=s.asm.LRNGrad).apply(null,arguments)},s._LeakyRelu=function(){return(s._LeakyRelu=s.asm.LeakyRelu).apply(null,arguments)},s._Less=function(){return(s._Less=s.asm.Less).apply(null,arguments)},s._LessEqual=function(){return(s._LessEqual=s.asm.LessEqual).apply(null,arguments)},s._LinSpace=function(){return(s._LinSpace=s.asm.LinSpace).apply(null,arguments)},s._Log=function(){return(s._Log=s.asm.Log).apply(null,arguments)},s._Log1p=function(){return(s._Log1p=s.asm.Log1p).apply(null,arguments)},s._LogicalAnd=function(){return(s._LogicalAnd=s.asm.LogicalAnd).apply(null,arguments)},s._LogicalNot=function(){return(s._LogicalNot=s.asm.LogicalNot).apply(null,arguments)},s._LogicalOr=function(){return(s._LogicalOr=s.asm.LogicalOr).apply(null,arguments)},s._LogicalXor=function(){return(s._LogicalXor=s.asm.LogicalXor).apply(null,arguments)},s._Max=function(){return(s._Max=s.asm.Max).apply(null,arguments)},s._MaxPool=function(){return(s._MaxPool=s.asm.MaxPool).apply(null,arguments)},s._MaxPool3D=function(){return(s._MaxPool3D=s.asm.MaxPool3D).apply(null,arguments)},s._MaxPool3DGrad=function(){return(s._MaxPool3DGrad=s.asm.MaxPool3DGrad).apply(null,arguments)},s._MaxPoolGrad=function(){return(s._MaxPoolGrad=s.asm.MaxPoolGrad).apply(null,arguments)},s._MaxPoolWithArgmax=function(){return(s._MaxPoolWithArgmax=s.asm.MaxPoolWithArgmax).apply(null,arguments)},s._Maximum=function(){return(s._Maximum=s.asm.Maximum).apply(null,arguments)},s._Mean=function(){return(s._Mean=s.asm.Mean).apply(null,arguments)},s._Min=function(){return(s._Min=s.asm.Min).apply(null,arguments)},s._Minimum=function(){return(s._Minimum=s.asm.Minimum).apply(null,arguments)},s._MirrorPad=function(){return(s._MirrorPad=s.asm.MirrorPad).apply(null,arguments)},s._Mod=function(){return(s._Mod=s.asm.Mod).apply(null,arguments)},s._Multinomial=function(){return(s._Multinomial=s.asm.Multinomial).apply(null,arguments)},s._Multiply=function(){return(s._Multiply=s.asm.Multiply).apply(null,arguments)},s._Neg=function(){return(s._Neg=s.asm.Neg).apply(null,arguments)},s._NonMaxSuppressionV3=function(){return(s._NonMaxSuppressionV3=s.asm.NonMaxSuppressionV3).apply(null,arguments)},s._NonMaxSuppressionV4=function(){return(s._NonMaxSuppressionV4=s.asm.NonMaxSuppressionV4).apply(null,arguments)},s._NonMaxSuppressionV5=function(){return(s._NonMaxSuppressionV5=s.asm.NonMaxSuppressionV5).apply(null,arguments)},s._NotEqual=function(){return(s._NotEqual=s.asm.NotEqual).apply(null,arguments)},s._OneHot=function(){return(s._OneHot=s.asm.OneHot).apply(null,arguments)},s._PadV2=function(){return(s._PadV2=s.asm.PadV2).apply(null,arguments)},s._Pow=function(){return(s._Pow=s.asm.Pow).apply(null,arguments)},s._Prelu=function(){return(s._Prelu=s.asm.Prelu).apply(null,arguments)},s._Prod=function(){return(s._Prod=s.asm.Prod).apply(null,arguments)},s._RealDiv=function(){return(s._RealDiv=s.asm.RealDiv).apply(null,arguments)},s._Reciprocal=function(){return(s._Reciprocal=s.asm.Reciprocal).apply(null,arguments)},s._Relu=function(){return(s._Relu=s.asm.Relu).apply(null,arguments)},s._Relu6=function(){return(s._Relu6=s.asm.Relu6).apply(null,arguments)},s._ResizeBilinear=function(){return(s._ResizeBilinear=s.asm.ResizeBilinear).apply(null,arguments)},s._ResizeBilinearGrad=function(){return(s._ResizeBilinearGrad=s.asm.ResizeBilinearGrad).apply(null,arguments)},s._ResizeNearestNeighbor=function(){return(s._ResizeNearestNeighbor=s.asm.ResizeNearestNeighbor).apply(null,arguments)},s._ResizeNearestNeighborGrad=function(){return(s._ResizeNearestNeighborGrad=s.asm.ResizeNearestNeighborGrad).apply(null,arguments)},s._Reverse=function(){return(s._Reverse=s.asm.Reverse).apply(null,arguments)},s._RotateWithOffset=function(){return(s._RotateWithOffset=s.asm.RotateWithOffset).apply(null,arguments)},s._Round=function(){return(s._Round=s.asm.Round).apply(null,arguments)},s._Rsqrt=function(){return(s._Rsqrt=s.asm.Rsqrt).apply(null,arguments)},s._ScatterNd=function(){return(s._ScatterNd=s.asm.ScatterNd).apply(null,arguments)},s._SearchSorted=function(){return(s._SearchSorted=s.asm.SearchSorted).apply(null,arguments)},s._SelectV2=function(){return(s._SelectV2=s.asm.SelectV2).apply(null,arguments)},s._Selu=function(){return(s._Selu=s.asm.Selu).apply(null,arguments)},s._Sigmoid=function(){return(s._Sigmoid=s.asm.Sigmoid).apply(null,arguments)},s._Sign=function(){return(s._Sign=s.asm.Sign).apply(null,arguments)},s._Sin=function(){return(s._Sin=s.asm.Sin).apply(null,arguments)},s._Sinh=function(){return(s._Sinh=s.asm.Sinh).apply(null,arguments)},s._Softmax=function(){return(s._Softmax=s.asm.Softmax).apply(null,arguments)},s._Softplus=function(){return(s._Softplus=s.asm.Softplus).apply(null,arguments)},s._SparseFillEmptyRows=function(){return(s._SparseFillEmptyRows=s.asm.SparseFillEmptyRows).apply(null,arguments)},s._SparseReshape=function(){return(s._SparseReshape=s.asm.SparseReshape).apply(null,arguments)},s._SparseSegmentReduction=function(){return(s._SparseSegmentReduction=s.asm.SparseSegmentReduction).apply(null,arguments)},s._SparseToDense=function(){return(s._SparseToDense=s.asm.SparseToDense).apply(null,arguments)},s._Sqrt=function(){return(s._Sqrt=s.asm.Sqrt).apply(null,arguments)},s._Square=function(){return(s._Square=s.asm.Square).apply(null,arguments)},s._SquaredDifference=function(){return(s._SquaredDifference=s.asm.SquaredDifference).apply(null,arguments)},s._Step=function(){return(s._Step=s.asm.Step).apply(null,arguments)},s._StridedSlice=function(){return(s._StridedSlice=s.asm.StridedSlice).apply(null,arguments)},s._Sub=function(){return(s._Sub=s.asm.Sub).apply(null,arguments)},s._Sum=function(){return(s._Sum=s.asm.Sum).apply(null,arguments)},s._Tan=function(){return(s._Tan=s.asm.Tan).apply(null,arguments)},s._Tanh=function(){return(s._Tanh=s.asm.Tanh).apply(null,arguments)},s._TensorScatterUpdate=function(){return(s._TensorScatterUpdate=s.asm.TensorScatterUpdate).apply(null,arguments)},s._Tile=function(){return(s._Tile=s.asm.Tile).apply(null,arguments)},s._TopK=function(){return(s._TopK=s.asm.TopK).apply(null,arguments)},s._Transform=function(){return(s._Transform=s.asm.Transform).apply(null,arguments)},s._Transpose=function(){return(s._Transpose=s.asm.Transpose).apply(null,arguments)},s.__FusedMatMul=function(){return(s.__FusedMatMul=s.asm._FusedMatMul).apply(null,arguments)},s._malloc=function(){return(s._malloc=s.asm.malloc).apply(null,arguments)},s._free=function(){return(s._free=s.asm.free).apply(null,arguments)},s.___errno_location=function(){return(s.___errno_location=s.asm.__errno_location).apply(null,arguments)},s.stackSave=function(){return(re=s.stackSave=s.asm.stackSave).apply(null,arguments)}),ae=s.stackRestore=function(){return(ae=s.stackRestore=s.asm.stackRestore).apply(null,arguments)},se=s.stackAlloc=function(){return(se=s.stackAlloc=s.asm.stackAlloc).apply(null,arguments)};function ie(e){function n(){Q||(Q=!0,s.calledRun=!0,!_&&(q(F),t(s),s.onRuntimeInitialized&&s.onRuntimeInitialized(),function(){if(s.postRun)for("function"==typeof s.postRun&&(s.postRun=[s.postRun]);s.postRun.length;)O(s.postRun.shift());q(D)}()))}e=e||h,z>0||(function(){if(s.preRun)for("function"==typeof s.preRun&&(s.preRun=[s.preRun]);s.preRun.length;)M(s.preRun.shift());q(R)}(),z>0)||(s.setStatus?(s.setStatus("Running..."),setTimeout(function(){setTimeout(function(){s.setStatus("")},1),n()},1)):n())}if(s.dynCall_iijjiiii=function(){return(s.dynCall_iijjiiii=s.asm.dynCall_iijjiiii).apply(null,arguments)},s.dynCall_jiji=function(){return(s.dynCall_jiji=s.asm.dynCall_jiji).apply(null,arguments)},s.cwrap=function(e,t,n,r){var a=(n=n||[]).every(e=>"number"===e||"boolean"===e);return"string"!==t&&a&&!r?Y(e):function(){return J(e,t,n,arguments)}},W=function e(){Q||ie(),Q||(W=e)},s.preInit)for("function"==typeof s.preInit&&(s.preInit=[s.preInit]);s.preInit.length>0;)s.preInit.pop()();if(ie(),r&&(ee={uncaughtException:process.listeners("uncaughtException").filter(function(e){return!r.uncaughtException.indexOf(e)>-1}),unhandledRejection:process.listeners("unhandledRejection").filter(function(e){return!r.unhandledRejection.indexOf(e)>-1})}),void 0!==e)te=e;else{if("undefined"==typeof WasmBackendModuleThreadedSimd)throw new Error("Could not find wasm module in post.js");te=WasmBackendModuleThreadedSimd}if(ee){var oe=te._dispose;te._dispose=function(){oe(),ee.uncaughtException.forEach(function(e){process.removeListener("uncaughtException",e)}),ee.unhandledRejection.forEach(function(e){process.removeListener("unhandledRejection",e)})}}return e.ready});"object"==typeof e&&"object"==typeof t?t.exports=s:"function"==typeof define&&n.amdO?define([],function(){return s}):"object"==typeof e&&(e.WasmBackendModule=s)}),L=class{constructor(e,t){this.backend=e,this.dataMover=t,this.data=new WeakMap,this.dataIdsCount=0}get(e){return this.data.has(e)||this.dataMover.moveData(this.backend,e),this.data.get(e)}set(e,t){this.dataIdsCount++,this.data.set(e,t)}has(e){return this.data.has(e)}delete(e){return this.dataIdsCount--,this.data.delete(e)}numDataIds(){return this.dataIdsCount}},P=class{refCount(e){return z("refCount")}incRef(e){return z("incRef")}timerAvailable(){return!0}time(e){return z("time")}read(e){return z("read")}readSync(e){return z("readSync")}readToGPU(e,t){return z("readToGPU")}numDataIds(){return z("numDataIds")}disposeData(e,t){return z("disposeData")}write(e,t,n){return z("write")}move(e,t,n,r,a){return z("move")}createTensorFromGPUData(e,t,n){return z("createTensorFromGPUData")}memory(){return z("memory")}floatPrecision(){return z("floatPrecision")}epsilon(){return 32===this.floatPrecision()?1e-7:1e-4}dispose(){return z("dispose")}};function z(e){throw new Error(`'${e}' not yet implemented or not found in the registry. This kernel may not be supported by the tfjs backend you have chosen`)}function B(e){let t=e.length,n=0;for(;t>0;)n=Math.random()*t|0,t--,G(e,t,n)}function W(e,t){if(e.length!==t.length)throw new Error(`Array sizes must match to be shuffled together First array length was ${e.length}Second array length was ${t.length}`);let n=e.length,r=0;for(;n>0;)r=Math.random()*n|0,n--,G(e,n,r),G(t,n,r)}function V(e,t,n){return Math.max(e,Math.min(t,n))}function U(e){return e%2==0?e:e+1}function G(e,t,n){let r=e[t];e[t]=e[n],e[n]=r}function H(e){let t=0;for(let n=0;n<e.length;n++)t+=e[n];return t}function j(e,t){let n=Math.random();return t*n+(1-n)*e}function q(e,t){let n=0;for(let r=0;r<e.length;r++){let a=Number(e[r])-Number(t[r]);n+=a*a}return n}function K(e,t){if(!e)throw new Error("string"==typeof t?t:t())}function X(e,t,n=""){K(ee(e,t),()=>n+` Shapes ${e} and ${t} must match`)}function Z(e){K(null!=e,()=>"The input to the tensor constructor must be a non-null value.")}function Y(e){if(0===e.length)return 1;let t=e[0];for(let n=1;n<e.length;n++)t*=e[n];return t}function J(e){return 0===e.length}function Q(e,t){if(e===t)return!0;if(null==e||null==t||e.length!==t.length)return!1;for(let n=0;n<e.length;n++)if(null!==e[n]&&null!==t[n]&&e[n]!==t[n])return!1;return!0}function ee(e,t){if(e===t)return!0;if(null==e||null==t||e.length!==t.length)return!1;for(let n=0;n<e.length;n++)if(e[n]!==t[n])return!1;return!0}function te(e){return e%1==0}function ne(e){if(null!=Math.tanh)return Math.tanh(e);if(e===1/0)return 1;if(e===-1/0)return-1;{let t=Math.exp(2*e);return(t-1)/(t+1)}}function re(e){let t=Math.ceil(Math.sqrt(e));return[t,Math.ceil(e/t)]}function ae(e){let t=new Uint32Array(e);for(let n=0;n<e;++n)t[n]=n;return B(t),t}function se(e,t){return t<=e.length?e:e+" ".repeat(t-e.length)}function ie(e,t=e=>0,n,r){return new Promise((a,s)=>{let i=0,o=()=>{if(e())return void a();i++;let l=t(i);null!=n&&i>=n?s():null!=r?r(o,l):setTimeout(o,l)};o()})}function oe(e,t){let n=1,r=-1;for(let t=0;t<e.length;++t)if(e[t]>=0)n*=e[t];else if(-1===e[t]){if(-1!==r)throw Error(`Shapes can only have 1 implicit size. Found -1 at dim ${r} and dim ${t}`);r=t}else if(e[t]<0)throw Error(`Shapes can not be < 0. Found ${e[t]} at dim ${t}`);if(-1===r){if(t>0&&t!==n)throw Error(`Size(${t}) must match the product of shape ${e}`);return e}if(0===n)throw Error(`Cannot infer the missing size in [${e}] when there are 0 elements`);if(t%n!==0)throw Error(`The implicit shape can't be a fractional number. Got ${t} / ${n}`);let a=e.slice();return a[r]=t/n,a}function le(e,t){let n=t.length;return K((e=null==e?t.map((e,t)=>t):[].concat(e)).every(e=>e>=-n&&e<n),()=>`All values in axis param must be in range [-${n}, ${n}) but got axis ${e}`),K(e.every(e=>te(e)),()=>`All values in axis param must be integers but got axis ${e}`),e.map(e=>e<0?n+e:e)}function ue(e,t){let n=[],r=[],a=null!=t&&Array.isArray(t)&&0===t.length,s=null==t||a?null:le(t,e).sort(),i=0;for(let t=0;t<e.length;++t){if(null!=s){if(s[i]===t&&1!==e[t])throw new Error(`Can't squeeze axis ${t} since its dim '${e[t]}' is not 1`);(null==s[i]||s[i]>t)&&1===e[t]&&(n.push(e[t]),r.push(t)),s[i]<=t&&i++}1!==e[t]&&(n.push(e[t]),r.push(t))}return{newShape:n,keptDims:r}}function he(e,t){return de(e,t)}function de(e,t){let n=null;if(null==e||"float32"===e)n=new Float32Array(t);else if("int32"===e)n=new Int32Array(t);else if("bool"===e)n=new Uint8Array(t);else{if("string"!==e)throw new Error(`Unknown data type ${e}`);n=new Array(t)}return n}function pe(e,t){for(let n=0;n<e.length;n++){let r=e[n];if(isNaN(r)||!isFinite(r))throw Error(`A tensor of type ${t} being uploaded contains ${r}.`)}}function ce(e){return"bool"===e||"complex64"===e||"float32"===e||"int32"===e||"string"===e}function fe(e,t){return!("complex64"===t||"float32"===t&&"complex64"!==e||"int32"===t&&"float32"!==e&&"complex64"!==e||"bool"===t&&"bool"===e)}function me(e){if("float32"===e||"int32"===e)return 4;if("complex64"===e)return 8;if("bool"===e)return 1;throw new Error(`Unknown dtype ${e}`)}function ge(e){if(null==e)return 0;let t=0;return e.forEach(e=>t+=e.length),t}function ye(e){return"string"==typeof e||e instanceof String}function be(e){return"boolean"==typeof e}function xe(e){return"number"==typeof e}function ve(e){return Array.isArray(e)?ve(e[0]):e instanceof Float32Array?"float32":e instanceof Int32Array||e instanceof Uint8Array||e instanceof Uint8ClampedArray?"int32":xe(e)?"float32":ye(e)?"string":be(e)?"bool":"float32"}function we(e){return!!(e&&e.constructor&&e.call&&e.apply)}function ke(e,t){for(let n=t;n<e;++n)if(e%n===0)return n;return e}function Ie(e){let t=e.length;if(t<2)return[];let n=new Array(t-1);n[t-2]=e[t-1];for(let r=t-3;r>=0;--r)n[r]=n[r+1]*e[r+1];return n}function Se(e,t,n,r=!1){let a=new Array;if(1===t.length){let s=t[0]*(r?2:1);for(let t=0;t<s;t++)a[t]=n[e+t]}else{let s=t[0],i=t.slice(1),o=i.reduce((e,t)=>e*t)*(r?2:1);for(let t=0;t<s;t++)a[t]=Se(e+t*o,i,n,r)}return a}function _e(e,t,n=!1){if(0===e.length)return t[0];let r=e.reduce((e,t)=>e*t)*(n?2:1);if(0===r)return[];if(r!==t.length)throw new Error(`[${e}] does not match the input size ${t.length}${n?" for a complex tensor":""}.`);return Se(0,e,t,n)}function Ne(e,t){if(Array.isArray(e))return e;if("float32"===t)return e instanceof Float32Array?e:new Float32Array(e);if("int32"===t)return e instanceof Int32Array?e:new Int32Array(e);if("bool"===t||"string"===t)return Uint8Array.from(new Int32Array(e));throw new Error(`Unknown dtype ${t}`)}function Te(e,t){let n=Ce(e,t);for(let e=0;e<n.length;e++)n[e]=1;return n}function Ce(e,t){if(null==t||"float32"===t||"complex64"===t)return new Float32Array(e);if("int32"===t)return new Int32Array(e);if("bool"===t)return new Uint8Array(e);throw new Error(`Unknown data type ${t}`)}function Ee(e,t){let n=e.reduce((e,t)=>e*t,1);if(null==t||"float32"===t)return _e(e,new Float32Array(n));if("int32"===t)return _e(e,new Int32Array(n));if("bool"===t)return _e(e,new Uint8Array(n));throw new Error(`Unknown data type ${t}`)}function Ae(e){e.forEach(t=>{K(Number.isInteger(t)&&t>=0,()=>`Tensor must have a shape comprised of positive integers but got shape [${e}].`)})}function $e(e,t,n){if(0===t)return 0;if(1===t)return e[0];let r=e[e.length-1];for(let t=0;t<e.length-1;++t)r+=n[t]*e[t];return r}function Re(e,t,n){if(0===t)return[];if(1===t)return[e];let r=new Array(t);for(let t=0;t<r.length-1;++t)r[t]=Math.floor(e/n[t]),e-=r[t]*n[t];return r[r.length-1]=e,r}function Fe(e){return e&&e.then&&"function"==typeof e.then}var De="tfjsflags",Me=class{constructor(e){this.global=e,this.flags={},this.flagRegistry={},this.urlFlags={},this.getQueryParams=Oe,this.populateURLFlags()}setPlatform(e,t){null!=this.platform&&(Le().getBool("IS_TEST")||Le().getBool("PROD")||console.warn(`Platform ${this.platformName} has already been set. Overwriting the platform with ${e}.`)),this.platformName=e,this.platform=t}registerFlag(e,t,n){if(this.flagRegistry[e]={evaluationFn:t,setHook:n},null!=this.urlFlags[e]){let t=this.urlFlags[e];Le().getBool("IS_TEST")||Le().getBool("PROD")||console.warn(`Setting feature override from URL ${e}: ${t}.`),this.set(e,t)}}async getAsync(e){return e in this.flags||(this.flags[e]=await this.evaluateFlag(e)),this.flags[e]}get(e){if(e in this.flags)return this.flags[e];let t=this.evaluateFlag(e);if(Fe(t))throw new Error(`Flag ${e} cannot be synchronously evaluated. Please use getAsync() instead.`);return this.flags[e]=t,this.flags[e]}getNumber(e){return this.get(e)}getBool(e){return this.get(e)}getString(e){return this.get(e)}getFlags(){return this.flags}get features(){return this.flags}set(e,t){if(null==this.flagRegistry[e])throw new Error(`Cannot set flag ${e} as it has not been registered.`);this.flags[e]=t,null!=this.flagRegistry[e].setHook&&this.flagRegistry[e].setHook(t)}evaluateFlag(e){if(null==this.flagRegistry[e])throw new Error(`Cannot evaluate flag '${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(void 0===this.global||void 0===this.global.location||void 0===this.global.location.search)return;let e=this.getQueryParams(this.global.location.search);De in e&&e[De].split(",").forEach(e=>{let[t,n]=e.split(":");this.urlFlags[t]=function(e,t){let n=t.toLowerCase();return"true"===n||"false"===n?"true"===n:""+ +n===n?+n:t}(0,n)})}};function Oe(e){let t={};return e.replace(/[?&]([^=?&]+)(?:=([^&]*))?/g,(e,...n)=>(function(e,t,n){e[decodeURIComponent(t)]=decodeURIComponent(n||"")}(t,n[0],n[1]),n.join("="))),t}function Le(){return ze}var Pe,ze=null;function Be(){if(null==Pe){let e;if("undefined"!=typeof window)e=window;else if(void 0!==n.g)e=n.g;else if("undefined"!=typeof process)e=process;else{if("undefined"==typeof self)throw new Error("Could not find a global object");e=self}Pe=e}return Pe}function We(e,t){let n=function(){let e=Be();return null==e._tfGlobals&&(e._tfGlobals=new Map),e._tfGlobals}();if(n.has(e))return n.get(e);{let r=t();return n.set(e,r),n.get(e)}}var Ve="Abs",Ue="Acos",Ge="Acosh",He="Add",je="AddN",qe="All",Ke="Any",Xe="ArgMax",Ze="ArgMin",Ye="Asin",Je="Asinh",Qe="Atan",et="Atanh",tt="Atan2",nt="AvgPool",rt="AvgPoolGrad",at="AvgPool3D",st="AvgPool3DGrad",it="BatchMatMul",ot="BatchToSpaceND",lt="Bincount",ut="BitwiseAnd",ht="BroadcastTo",dt="BroadcastArgs",pt="Cast",ct="Ceil",ft="ClipByValue",mt="Complex",gt="ComplexAbs",yt="Concat",bt="Conv2D",xt="Conv2DBackpropFilter",vt="Conv2DBackpropInput",wt="Conv3D",kt="Conv3DBackpropFilterV2",It="Conv3DBackpropInputV2",St="Cos",_t="Cosh",Nt="Cumprod",Tt="Cumsum",Ct="CropAndResize",Et="DenseBincount",At="DepthToSpace",$t="DepthwiseConv2dNative",Rt="DepthwiseConv2dNativeBackpropFilter",Ft="DepthwiseConv2dNativeBackpropInput",Dt="Diag",Mt="Dilation2D",Ot="Dilation2DBackpropInput",Lt="Dilation2DBackpropFilter",Pt="Draw",zt="RealDiv",Bt="Einsum",Wt="Elu",Vt="EluGrad",Ut="Erf",Gt="Equal",Ht="Exp",jt="ExpandDims",qt="Expm1",Kt="FFT",Xt="Fill",Zt="FlipLeftRight",Yt="Floor",Jt="FloorDiv",Qt="FusedBatchNorm",en="GatherV2",tn="GatherNd",nn="Greater",rn="GreaterEqual",an="Identity",sn="IFFT",on="Imag",ln="IsFinite",un="IsInf",hn="IsNan",dn="LeakyRelu",pn="Less",cn="LessEqual",fn="LinSpace",mn="Log",gn="Log1p",yn="LogicalAnd",bn="LogicalNot",xn="LogicalOr",vn="LogicalXor",wn="LogSoftmax",kn="LowerBound",In="LRN",Sn="LRNGrad",_n="MatrixBandPart",Nn="Max",Tn="Maximum",Cn="MaxPool",En="MaxPoolGrad",An="MaxPool3D",$n="MaxPool3DGrad",Rn="MaxPoolWithArgmax",Fn="Mean",Dn="Min",Mn="Minimum",On="MirrorPad",Ln="Mod",Pn="Multinomial",zn="Multiply",Bn="Neg",Wn="NotEqual",Vn="NonMaxSuppressionV3",Un="NonMaxSuppressionV4",Gn="NonMaxSuppressionV5",Hn="OnesLike",jn="OneHot",qn="Pack",Kn="PadV2",Xn="Pool",Zn="Pow",Yn="Prelu",Jn="Prod",Qn="RaggedGather",er="RaggedRange",tr="RaggedTensorToTensor",nr="Range",rr="Real",ar="Reciprocal",sr="Relu",ir="Reshape",or="ResizeNearestNeighbor",lr="ResizeNearestNeighborGrad",ur="ResizeBilinear",hr="ResizeBilinearGrad",dr="Relu6",pr="Reverse",cr="Round",fr="Rsqrt",mr="ScatterNd",gr="TensorScatterUpdate",yr="SearchSorted",br="Select",xr="Selu",vr="Slice",wr="Sin",kr="Sinh",Ir="Sign",Sr="Sigmoid",_r="Softplus",Nr="Sqrt",Tr="Sum",Cr="SpaceToBatchND",Er="SplitV",Ar="Softmax",$r="SparseFillEmptyRows",Rr="SparseReshape",Fr="SparseSegmentMean",Dr="SparseSegmentSum",Mr="SparseToDense",Or="SquaredDifference",Lr="Square",Pr="StaticRegexReplace",zr="StridedSlice",Br="StringNGrams",Wr="StringSplit",Vr="StringToHashBucketFast",Ur="Sub",Gr="Tan",Hr="Tanh",jr="Tile",qr="TopK",Kr="Transform",Xr="Transpose",Zr="Unique",Yr="Unpack",Jr="UnsortedSegmentSum",Qr="UpperBound",ea="ZerosLike",ta="Step",na="FromPixels",ra="RotateWithOffset",aa="_FusedMatMul",sa="FusedConv2D",ia="FusedDepthwiseConv2D";function oa(...e){Le().getBool("IS_TEST")||Le().getBool("PROD")||console.warn(...e)}function la(...e){Le().getBool("IS_TEST")||Le().getBool("PROD")||console.log(...e)}var ua=We("kernelRegistry",()=>new Map),ha=We("gradRegistry",()=>new Map);function da(e,t){let n=xa(e,t);return ua.get(n)}function pa(e){return ha.get(e)}function ca(e){let t=ua.entries(),n=[];for(;;){let{done:r,value:a}=t.next();if(r)break;let[s,i]=a,[o]=s.split("_");o===e&&n.push(i)}return n}function fa(e){let{kernelName:t,backendName:n}=e,r=xa(t,n);ua.has(r)&&oa(`The kernel '${t}' for backend '${n}' is already registered`),ua.set(r,e)}function ma(e){let{kernelName:t}=e;ha.has(t)&&Le().getBool("DEBUG")&&oa(`Overriding the gradient for '${t}'`),ha.set(t,e)}function ga(e,t){let n=xa(e,t);if(!ua.has(n))throw new Error(`The kernel '${e}' for backend '${t}' is not registered`);ua.delete(n)}function ya(e){if(!ha.has(e))throw new Error(`The gradient '${e}' for backend is not registered`);ha.delete(e)}function ba(e,t){ca(e).forEach(e=>{fa(Object.assign({},e,{backendName:t}))})}function xa(e,t){return`${t}_${e}`}var va={};function wa(e){return e instanceof Float32Array||e instanceof Int32Array||e instanceof Uint8Array||e instanceof Uint8ClampedArray}f(va,{arraysEqual:()=>ee,arraysEqualWithNull:()=>Q,assert:()=>K,assertNonNegativeIntegerDimensions:()=>Ae,assertNonNull:()=>Z,assertShapesMatch:()=>X,bytesFromStringArray:()=>ge,bytesPerElement:()=>me,checkConversionForErrors:()=>pe,clamp:()=>V,computeStrides:()=>Ie,convertBackendValuesAndArrayBuffer:()=>Ne,createScalarValue:()=>Oa,createShuffledIndices:()=>ae,decodeString:()=>Wa,distSquared:()=>q,encodeString:()=>Ba,fetch:()=>za,fingerPrint64:()=>Ma,flatten:()=>Ua,getArrayFromDType:()=>de,getTypedArrayFromDType:()=>he,hasEncodingLoss:()=>fe,hexToLong:()=>Sa,indexToLoc:()=>Re,inferDtype:()=>ve,inferFromImplicitShape:()=>oe,isBoolean:()=>be,isFunction:()=>we,isInt:()=>te,isNumber:()=>xe,isPromise:()=>Fe,isScalarShape:()=>J,isString:()=>ye,isTypedArray:()=>Va,isValidDtype:()=>ce,locToIndex:()=>$e,makeOnesTypedArray:()=>Te,makeZerosNestedTypedArray:()=>Ee,makeZerosTypedArray:()=>Ce,nearestDivisor:()=>ke,nearestLargerEven:()=>U,now:()=>Pa,parseAxisParam:()=>le,randUniform:()=>j,repeatedTry:()=>ie,rightPad:()=>se,shuffle:()=>B,shuffleCombo:()=>W,sizeFromShape:()=>Y,sizeToSquarishShape:()=>re,squeezeShape:()=>ue,sum:()=>H,swap:()=>G,tanh:()=>ne,toNestedArray:()=>_e,toTypedArray:()=>La});var ka=m(g()),Ia=ka.default||ka;function Sa(e){return Ia.fromString(e,!0,16)}var _a=Sa("c3a5c85c97cb3127"),Na=Sa("b492b66fbe98f273"),Ta=Sa("9ae16a3b2f90404f");function Ca(e){return e.xor(e.shru(47))}function Ea(e,t,n){let r=e.slice(t,t+n);return Ia.fromBytes(Array.from(r),!0,!0)}function Aa(e,t){return Ea(e,t,8)}function $a(e,t){return Ea(e,t,4)}function Ra(e,t){return 0===t?e:e.shru(t).or(e.shl(64-t))}function Fa(e,t,n=Sa("9ddfea08eb382d69")){let r=e.xor(t).mul(n);r=r.xor(r.shru(47));let a=t.xor(r).mul(n);return a=a.xor(a.shru(47)),a=a.mul(n),a}function Da(e,t,n,r){return function(e,t,n,r,a,s){a=a.add(e),s=Ra(s.add(a).add(r),21);let i=a;return a=(a=a.add(t)).add(n),s=s.add(Ra(a,44)),[a.add(r),s.add(i)]}(Aa(e,t),Aa(e,t+8),Aa(e,t+16),Aa(e,t+24),n,r)}function Ma(e,t=e.length){let n=Ia.fromNumber(81,!0);if(t<=32)return t<=16?function(e,t=e.length){if(t>=8){let n=Ta.add(2*t),r=Aa(e,0).add(Ta),a=Aa(e,t-8);return Fa(Ra(a,37).mul(n).add(r),Ra(r,25).add(a).mul(n),n)}if(t>=4){let n=Ta.add(2*t);return Fa($a(e,0).shl(3).add(t),$a(e,t-4),n)}if(t>0){let n=e[0]+(e[t>>1]<<8),r=t+(e[t-1]<<2);return Ca(Ta.mul(n).xor(_a.mul(r))).mul(Ta)}return Ta}(e,t):function(e,t=e.length){let n=Ta.add(2*t),r=Aa(e,0).mul(Na),a=Aa(e,8),s=Aa(e,t-8).mul(n),i=Aa(e,t-16).mul(Ta);return Fa(Ra(r.add(a),43).add(Ra(s,30)).add(i),r.add(Ra(a.add(Ta),18)).add(s),n)}(e,t);if(t<=64)return function(e,t=e.length){let n=Ta.add(2*t),r=Aa(e,0).mul(Ta),a=Aa(e,8),s=Aa(e,t-8).mul(n),i=Aa(e,t-16).mul(Ta),o=Ra(r.add(a),43).add(Ra(s,30)).add(i),l=Fa(o,r.add(Ra(a.add(Ta),18)).add(s),n),u=Aa(e,16).mul(n),h=Aa(e,24),d=o.add(Aa(e,t-32)).mul(n),p=l.add(Aa(e,t-24)).mul(n);return Fa(Ra(u.add(h),43).add(Ra(d,30)).add(p),u.add(Ra(h.add(r),18)).add(d),n)}(e,t);let r=n,a=n.mul(Na).add(113),s=Ca(a.mul(Ta).add(113)).mul(Ta),i=[Ia.UZERO,Ia.UZERO],o=[Ia.UZERO,Ia.UZERO];r=r.mul(Ta).add(Aa(e,0));let l=0,u=64*(t-1>>6),h=u+(t-1&63)-63;do{r=Ra(r.add(a).add(i[0]).add(Aa(e,l+8)),37).mul(Na),a=Ra(a.add(i[1]).add(Aa(e,l+48)),42).mul(Na),r=r.xor(o[1]),a=a.add(i[0]).add(Aa(e,l+40)),s=Ra(s.add(o[0]),33).mul(Na),i=Da(e,l,i[1].mul(Na),r.add(o[0])),o=Da(e,l+32,s.add(o[1]),a.add(Aa(e,l+16))),[s,r]=[r,s],l+=64}while(l!==u);let d=Na.add(s.and(255).shl(1));return l=h,o[0]=o[0].add(t-1&63),i[0]=i[0].add(o[0]),o[0]=o[0].add(i[0]),r=Ra(r.add(a).add(i[0]).add(Aa(e,l+8)),37).mul(d),a=Ra(a.add(i[1]).add(Aa(e,l+48)),42).mul(d),r=r.xor(o[1].mul(9)),a=a.add(i[0].mul(9).add(Aa(e,l+40))),s=Ra(s.add(o[0]),33).mul(d),i=Da(e,l,i[1].mul(d),r.add(o[0])),o=Da(e,l+32,s.add(o[1]),a.add(Aa(e,l+16))),[s,r]=[r,s],Fa(Fa(i[0],o[0],d).add(Ca(a).mul(_a)).add(s),Fa(i[1],o[1],d).add(r),d)}function Oa(e,t){return"string"===t?Ba(e):La([e],t)}function La(e,t){if("string"===t)throw new Error("Cannot convert a string[] to a TypedArray");if(Array.isArray(e)&&(e=Ua(e)),Le().getBool("DEBUG")&&pe(e,t),function(e,t){return e instanceof Float32Array&&"float32"===t||e instanceof Int32Array&&"int32"===t||e instanceof Uint8Array&&"bool"===t}(e,t))return e;if(null==t||"float32"===t||"complex64"===t)return new Float32Array(e);if("int32"===t)return new Int32Array(e);if("bool"===t){let t=new Uint8Array(e.length);for(let n=0;n<t.length;++n)0!==Math.round(e[n])&&(t[n]=1);return t}throw new Error(`Unknown data type ${t}`)}function Pa(){return Le().platform.now()}function za(e,t){return Le().platform.fetch(e,t)}function Ba(e,t="utf-8"){return t=t||"utf-8",Le().platform.encode(e,t)}function Wa(e,t="utf-8"){return t=t||"utf-8",Le().platform.decode(e,t)}function Va(e){return null!=Le().platform.isTypedArray?Le().platform.isTypedArray(e):wa(e)}function Ua(e,t=[],n=!1){if(null==t&&(t=[]),"boolean"==typeof e||"number"==typeof e||"string"==typeof e||Fe(e)||null==e||Va(e)&&n)t.push(e);else if(Array.isArray(e)||Va(e))for(let r=0;r<e.length;++r)Ua(e[r],t,n);else{let r=-1;for(let t of Object.keys(e))/^([1-9]+[0-9]*|0)$/.test(t)&&(r=Math.max(r,Number(t)));for(let a=0;a<=r;a++)Ua(e[a],t,n)}return t}var Ga=class{constructor(e,t){this.backendTimer=e,this.logger=t,null==t&&(this.logger=new ja)}profileKernel(e,t,n){let r,a,s=()=>{r=n()},i=Pa();if(this.backendTimer.timerAvailable())a=this.backendTimer.time(s);else{s();for(let e of r)e.dataSync();a=Promise.resolve({kernelMs:Pa()-i})}if(Le().getBool("CHECK_COMPUTATION_FOR_ERRORS"))for(let t=0;t<r.length;t++){let n=r[t];n.data().then(t=>{Ha(t,n.dtype,e)})}return{kernelName:e,outputs:r,inputs:t,timeMs:a.then(e=>e.kernelMs),extraInfo:a.then(e=>null!=e.getExtraProfileInfo?e.getExtraProfileInfo():"")}}logKernelProfile(e){let{kernelName:t,outputs:n,timeMs:r,inputs:a,extraInfo:s}=e;n.forEach(e=>{Promise.all([e.data(),r,s]).then(n=>{this.logger.logKernelProfile(t,e,n[0],n[1],a,n[2])})})}};function Ha(e,t,n){if("float32"!==t)return!1;for(let t=0;t<e.length;t++){let r=e[t];if(isNaN(r)||!isFinite(r))return console.warn(`Found ${r} in the result of '${n}'`),!0}return!1}var ja=class{logKernelProfile(e,t,n,r,a,s){let i="number"==typeof r?se(`${r}ms`,9):r.error,o=se(e,25),l=t.rank,u=t.size,h=se(t.shape.toString(),14),d="";for(let e in a){let n=a[e];if(null!=n){let r=n.shape||t.shape,a=r.length;d+=`${e}: ${a}D ${a>0?r:""} `}}console.log(`%c${o}\t%c${i}\t%c${l}D ${h}\t%c${u}\t%c${d}\t%c${s}`,"font-weight:bold","color:red","color:blue","color: orange","color: green","color: steelblue")}};function qa(e,t,n,r){let a=Ie(t),s=function(e,t,n,r){let a=Y(t),s=r[r.length-1],i=new Array(s).fill(0),o=t.length,l="complex64"===n?Ya(e):e;if(o>1)for(let e=0;e<a/s;e++){let t=e*s;for(let e=0;e<s;e++)i[e]=Math.max(i[e],Ka(l[t+e],0,n).length)}return i}(e,t,n,a),i=t.length,o=Za(e,t,n,a,s),l=["Tensor"];return r&&(l.push(`  dtype: ${n}`),l.push(`  rank: ${i}`),l.push(`  shape: [${t}]`),l.push("  values:")),l.push(o.map(e=>"    "+e).join("\n")),l.join("\n")}function Ka(e,t,n){let r;return r=Array.isArray(e)?`${parseFloat(e[0].toFixed(7))} + ${parseFloat(e[1].toFixed(7))}j`:ye(e)?`'${e}'`:"bool"===n?Xa(e):parseFloat(e.toFixed(7)).toString(),se(r,t)}function Xa(e){return 0===e?"false":"true"}function Za(e,t,n,r,a,s=!0){let i="complex64"===n?2:1,o=t[0],l=t.length;if(0===l){if("complex64"===n){return[Ka(Ya(e)[0],0,n)]}return"bool"===n?[Xa(e[0])]:[e[0].toString()]}if(1===l){if(o>20){let t=3*i,r=Array.from(e.slice(0,t)),s=Array.from(e.slice((o-3)*i,o*i));return"complex64"===n&&(r=Ya(r),s=Ya(s)),["["+r.map((e,t)=>Ka(e,a[t],n)).join(", ")+", ..., "+s.map((e,t)=>Ka(e,a[o-3+t],n)).join(", ")+"]"]}return["["+("complex64"===n?Ya(e):Array.from(e)).map((e,t)=>Ka(e,a[t],n)).join(", ")+"]"]}let u=t.slice(1),h=r.slice(1),d=r[0]*i,p=[];if(o>20){for(let t=0;t<3;t++){let r=t*d,s=r+d;p.push(...Za(e.slice(r,s),u,n,h,a,!1))}p.push("...");for(let t=o-3;t<o;t++){let r=t*d,s=r+d;p.push(...Za(e.slice(r,s),u,n,h,a,t===o-1))}}else for(let t=0;t<o;t++){let r=t*d,s=r+d;p.push(...Za(e.slice(r,s),u,n,h,a,t===o-1))}let c=2===l?",":"";p[0]="["+(o>0?p[0]+c:"");for(let e=1;e<p.length-1;e++)p[e]=" "+p[e]+c;let f=",\n";for(let e=2;e<l;e++)f+="\n";return p[p.length-1]=" "+p[p.length-1]+"]"+(s?"":f),p}function Ya(e){let t=[];for(let n=0;n<e.length;n+=2)t.push([e[n],e[n+1]]);return t}var Ja=class{constructor(e,t,n){if(this.dtype=t,this.shape=e.slice(),this.size=Y(e),null!=n){let e=n.length;K(e===this.size,()=>`Length of values '${e}' does not match the size inferred by the shape '${this.size}'.`)}if("complex64"===t)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||de(t,this.size),this.strides=Ie(e)}set(e,...t){0===t.length&&(t=[0]),K(t.length===this.rank,()=>`The number of provided coordinates (${t.length}) must match the rank (${this.rank})`);let n=this.locToIndex(t);this.values[n]=e}get(...e){0===e.length&&(e=[0]);let t=0;for(let n of e){if(n<0||n>=this.shape[t]){let t=`Requested out of range element at ${e}.   Buffer shape=${this.shape}`;throw new Error(t)}t++}let n=e[e.length-1];for(let t=0;t<e.length-1;++t)n+=this.strides[t]*e[t];return this.values[n]}locToIndex(e){if(0===this.rank)return 0;if(1===this.rank)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(0===this.rank)return[];if(1===this.rank)return[e];let 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 Qa().makeTensor(this.values,this.shape,this.dtype)}},Qa=null,es=null;var ts=class{constructor(e,t,n,r){this.kept=!1,this.isDisposedInternal=!1,this.shape=e.slice(),this.dtype=t||"float32",this.size=Y(e),this.strides=Ie(e),this.dataId=n,this.id=r,this.rankType=this.rank<5?this.rank.toString():"higher"}get rank(){return this.shape.length}async buffer(){let e=await this.data();return es.buffer(this.shape,this.dtype,e)}bufferSync(){return es.buffer(this.shape,this.dtype,this.dataSync())}async array(){let e=await this.data();return _e(this.shape,e,"complex64"===this.dtype)}arraySync(){return _e(this.shape,this.dataSync(),"complex64"===this.dtype)}async data(){this.throwIfDisposed();let e=Qa().read(this.dataId);if("string"===this.dtype){let t=await e;try{return t.map(e=>Wa(e))}catch(e){throw new Error("Failed to decode the string bytes into utf-8. To get the original bytes, call tensor.bytes().")}}return e}dataToGPU(e){return this.throwIfDisposed(),Qa().readToGPU(this.dataId,e)}dataSync(){this.throwIfDisposed();let e=Qa().readSync(this.dataId);if("string"===this.dtype)try{return e.map(e=>Wa(e))}catch(e){throw new Error("Failed to decode the string bytes into utf-8. To get the original bytes, call tensor.bytes().")}return e}async bytes(){this.throwIfDisposed();let e=await Qa().read(this.dataId);return"string"===this.dtype?e:new Uint8Array(e.buffer)}dispose(){this.isDisposed||(this.kerasMask&&this.kerasMask.dispose(),Qa().disposeTensor(this),this.isDisposedInternal=!0)}get isDisposed(){return this.isDisposedInternal}throwIfDisposed(){if(this.isDisposed)throw new Error("Tensor is disposed.")}print(e=!1){return es.print(this,e)}clone(){return this.throwIfDisposed(),es.clone(this)}toString(e=!1){return qa(this.dataSync(),this.shape,this.dtype,e)}cast(e){return this.throwIfDisposed(),es.cast(this,e)}variable(e=!0,t,n){return this.throwIfDisposed(),Qa().makeVariable(this,e,t,n)}};function ns(){return We("Tensor",()=>ts)}Object.defineProperty(ts,Symbol.hasInstance,{value:e=>!!e&&null!=e.data&&null!=e.dataSync&&null!=e.throwIfDisposed}),ns();var rs=class extends ts{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(!ee(e.shape,this.shape))throw new Error(`shape of the new value (${e.shape}) and previous value (${this.shape}) must match`);Qa().disposeTensor(this),this.dataId=e.dataId,Qa().incRef(this,null)}dispose(){Qa().disposeVariable(this),this.isDisposedInternal=!0}};Object.defineProperty(rs,Symbol.hasInstance,{value:e=>e instanceof ts&&null!=e.assign&&e.assign instanceof Function});var as,ss,is,os,ls,us,hs={};f(hs,{assertTypesMatch:()=>ys,getTensorsInContainer:()=>xs,isTensorInList:()=>bs,makeTypesMatch:()=>gs}),(ss=as||(as={})).R0="R0",ss.R1="R1",ss.R2="R2",ss.R3="R3",ss.R4="R4",ss.R5="R5",ss.R6="R6",function(e){e.float32="float32",e.int32="int32",e.bool="int32",e.complex64="complex64"}(is||(is={})),function(e){e.float32="float32",e.int32="int32",e.bool="bool",e.complex64="complex64"}(os||(os={})),function(e){e.float32="float32",e.int32="float32",e.bool="float32",e.complex64="complex64"}(ls||(ls={})),function(e){e.float32="complex64",e.int32="complex64",e.bool="complex64",e.complex64="complex64"}(us||(us={}));var ds={float32:ls,int32:is,bool:os,complex64:us};function ps(e,t){if("string"===e||"string"===t){if("string"===e&&"string"===t)return"string";throw new Error(`Can not upcast ${e} with ${t}`)}return ds[e][t]}function cs(e){return ps(e,"int32")}function fs(e){return null!=e&&"object"==typeof e&&"texture"in e&&e.texture instanceof WebGLTexture}function ms(e){return"undefined"!=typeof GPUBuffer&&null!=e&&"object"==typeof e&&"buffer"in e&&e.buffer instanceof GPUBuffer}function gs(e,t){if(e.dtype===t.dtype)return[e,t];let n=ps(e.dtype,t.dtype);return[e.cast(n),t.cast(n)]}function ys(e,t){K(e.dtype===t.dtype,()=>`The dtypes of the first(${e.dtype}) and second(${t.dtype}) input must match`)}function bs(e,t){return t.some(t=>t.id===e.id)}function xs(e){let t=[];return vs(e,t,new Set),t}function vs(e,t,n){if(null==e)return;if(e instanceof ts)return void t.push(e);if(!function(e){return Array.isArray(e)||"object"==typeof e}(e))return;let r=e;for(let e in r){let a=r[e];n.has(a)||(n.add(a),vs(a,t,n))}}function ws(e){return null!=e.kernelName}var ks=class{constructor(){this.registeredVariables={},this.nextTapeNodeId=0,this.numBytes=0,this.numTensors=0,this.numStringTensors=0,this.numDataBuffers=0,this.gradientDepth=0,this.kernelDepth=0,this.scopeStack=[],this.numDataMovesStack=[],this.nextScopeId=0,this.tensorInfo=new WeakMap,this.profiling=!1,this.activeProfile={newBytes:0,newTensors:0,peakBytes:0,kernels:[],result:null,get kernelNames(){return Array.from(new Set(this.kernels.map(e=>e.name)))}}}dispose(){for(let e in this.registeredVariables)this.registeredVariables[e].dispose()}},Is=class e{constructor(e){this.ENV=e,this.registry={},this.registryFactory={},this.pendingBackendInitId=0,this.state=new ks}async ready(){if(null!=this.pendingBackendInit)return this.pendingBackendInit.then(()=>{});if(null!=this.backendInstance)return;let e=this.getSortedBackends();for(let t=0;t<e.length;t++){let n=e[t];if(await this.initializeBackend(n).success)return void await this.setBackend(n)}throw new Error("Could not initialize any backends, all backend initializations failed.")}get backend(){if(null!=this.pendingBackendInit)throw new Error(`Backend '${this.backendName}' has not yet been initialized. Make sure to await tf.ready() or await tf.setBackend() before calling other methods`);if(null==this.backendInstance){let{name:e,asyncInit:t}=this.initializeBackendsAndReturnBest();if(t)throw new Error(`The highest priority backend '${e}' has not yet been initialized. Make sure to await tf.ready() or await tf.setBackend() before calling other methods`);this.setBackend(e)}return this.backendInstance}backendNames(){return Object.keys(this.registryFactory)}findBackend(e){if(!(e in this.registry)){if(!(e in this.registryFactory))return null;{let{asyncInit:t}=this.initializeBackend(e);if(t)return null}}return this.registry[e]}findBackendFactory(e){return e in this.registryFactory?this.registryFactory[e].factory:null}registerBackend(e,t,n=1){return e in this.registryFactory?(oa(`${e} backend was already registered. Reusing existing backend factory.`),!1):(this.registryFactory[e]={factory:t,priority:n},!0)}async setBackend(e){if(null==this.registryFactory[e])throw new Error(`Backend name '${e}' not found in registry`);if(this.backendName=e,null==this.registry[e]){this.backendInstance=null;let{success:t,asyncInit:n}=this.initializeBackend(e);if(!(n?await t:t))return!1}return this.backendInstance=this.registry[e],this.setupRegisteredKernels(),this.profiler=new Ga(this.backendInstance),!0}setupRegisteredKernels(){ca(this.backendName).forEach(e=>{null!=e.setupFunc&&e.setupFunc(this.backendInstance)})}disposeRegisteredKernels(e){ca(e).forEach(t=>{null!=t.disposeFunc&&t.disposeFunc(this.registry[e])})}initializeBackend(e){let t=this.registryFactory[e];if(null==t)throw new Error(`Cannot initialize backend ${e}, no registration found.`);try{let n=t.factory();if(!n||n instanceof P||"function"!=typeof n.then)return this.registry[e]=n,{success:!0,asyncInit:!1};{let t=++this.pendingBackendInitId,r=n.then(n=>!(t<this.pendingBackendInitId)&&(this.registry[e]=n,this.pendingBackendInit=null,!0)).catch(n=>(t<this.pendingBackendInitId||(this.pendingBackendInit=null,oa(`Initialization of backend ${e} failed`),oa(n.stack||n.message)),!1));return this.pendingBackendInit=r,{success:r,asyncInit:!0}}}catch(t){return oa(`Initialization of backend ${e} failed`),oa(t.stack||t.message),{success:!1,asyncInit:!1}}}removeBackend(e){if(!(e in this.registryFactory))throw new Error(`${e} backend not found in registry`);this.backendName===e&&null!=this.pendingBackendInit&&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(0===Object.keys(this.registryFactory).length)throw new Error("No backend found in registry.");return Object.keys(this.registryFactory).sort((e,t)=>this.registryFactory[t].priority-this.registryFactory[e].priority)}initializeBackendsAndReturnBest(){let e=this.getSortedBackends();for(let t=0;t<e.length;t++){let n=e[t],{success:r,asyncInit:a}=this.initializeBackend(n);if(a||r)return{name:n,asyncInit:a}}throw new Error("Could not initialize any backends, all backend initializations failed.")}moveData(e,t){let n=this.state.tensorInfo.get(t),r=n.backend,a=this.readSync(t),s=r.refCount(t);r.disposeData(t,!0),n.backend=e,e.move(t,a,n.shape,n.dtype,s),this.shouldCheckForMemLeaks()&&this.state.numDataMovesStack[this.state.numDataMovesStack.length-1]++}tidy(e,t){let n,r=null;if(null==t){if("function"!=typeof e)throw new Error("Please provide a function to tidy()");t=e}else{if("string"!=typeof e&&!(e instanceof String))throw new Error("When calling with two arguments, the first argument to tidy() must be a string");if("function"!=typeof t)throw new Error("When calling with two arguments, the 2nd argument to tidy() must be a function");r=e}return this.scopedRun(()=>this.startScope(r),()=>this.endScope(n),()=>(n=t(),n instanceof Promise&&console.error("Cannot return a Promise inside of tidy."),n))}scopedRun(e,t,n){e();try{let e=n();return t(),e}catch(e){throw t(),e}}nextTensorId(){return e.nextTensorId++}nextVariableId(){return e.nextVariableId++}clone(e){let t=_s.runKernel(an,{x:e}),n={x:e};return this.addTapeNode(this.state.activeScope.name,n,[t],e=>({x:()=>{let t={x:e},n={dtype:"float32"};return _s.runKernel(pt,t,n)}}),[],{}),t}runKernel(e,t,n){if(null==this.backendName&&this.backend,null==da(e,this.backendName))throw new Error(`Kernel '${e}' not registered for backend '${this.backendName}'`);return this.runKernelFunc({kernelName:e,inputs:t,attrs:n})}shouldCheckForMemLeaks(){return this.ENV.getBool("IS_TEST")}checkKernelForMemLeak(e,t,n){let r=this.backend.numDataIds(),a=0;n.forEach(e=>{a+="complex64"===e.dtype?3:1});let s=this.state.numDataMovesStack[this.state.numDataMovesStack.length-1],i=r-t-a-s;if(i>0)throw new Error(`Backend '${this.backendName}' has an internal memory leak (${i} data ids) after running '${e}'`)}runKernelFunc(e){let t,n,r=[],a=this.isTapeOn(),s=this.state.numBytes,i=this.state.numTensors;this.shouldCheckForMemLeaks()&&this.state.numDataMovesStack.push(0),null==this.backendName&&this.backend;let o,l=ws(e)?e.kernelName:null!=this.state.activeScope?this.state.activeScope.name:"";if(ws(e)){let{kernelName:t,inputs:s,attrs:i}=e;null==this.backendName&&this.backend;let l=da(t,this.backendName);K(null!=l,()=>`Cannot find registered kernel '${t}' for backend '${this.backendName}'`),n=()=>{let e=this.backend.numDataIds();o=l.kernelFunc({inputs:s,attrs:i,backend:this.backend});let n=Array.isArray(o)?o:[o];this.shouldCheckForMemLeaks()&&this.checkKernelForMemLeak(t,e,n);let u=n.map(e=>null!=e.rank?e:this.makeTensorFromTensorInfo(e));if(a){let e=this.getTensorsForGradient(t,s,u);r=this.saveTensorsForBackwardMode(e)}return u}}else{let{forwardFunc:t}=e,s=e=>{a&&(r=e.map(e=>this.keep(this.clone(e))))};n=()=>{let e=this.backend.numDataIds();o=this.tidy(()=>t(this.backend,s));let n=Array.isArray(o)?o:[o];return this.shouldCheckForMemLeaks()&&this.checkKernelForMemLeak(l,e,n),n}}let u,{inputs:h,attrs:d}=e,p=ws(e)?null:e.backwardsFunc;return this.scopedRun(()=>this.state.kernelDepth++,()=>this.state.kernelDepth--,()=>{this.ENV.getBool("DEBUG")||this.state.profiling?(u=this.profiler.profileKernel(l,h,()=>n()),this.ENV.getBool("DEBUG")&&this.profiler.logKernelProfile(u),t=u.outputs):t=n()}),a&&this.addTapeNode(l,h,t,p,r,d),this.state.profiling&&this.state.activeProfile.kernels.push({name:l,bytesAdded:this.state.numBytes-s,totalBytesSnapshot:this.state.numBytes,tensorsAdded:this.state.numTensors-i,totalTensorsSnapshot:this.state.numTensors,inputShapes:Object.keys(h).map(e=>null!=h[e]?h[e].shape:null),outputShapes:t.map(e=>e.shape),kernelTimeMs:u.timeMs,extraInfo:u.extraInfo}),Array.isArray(o)?t:t[0]}saveTensorsForBackwardMode(e){return e.map(e=>this.keep(this.clone(e)))}getTensorsForGradient(e,t,n){let r=pa(e);if(null!=r){let e,a=r.inputsToSave||[],s=r.outputsToSave||[];r.saveAllInputs?(K(Array.isArray(t),()=>"saveAllInputs is true, expected inputs to be an array."),e=Object.keys(t).map(e=>t[e])):e=a.map(e=>t[e]);let i=n.filter((e,t)=>s[t]);return e.concat(i)}return[]}makeTensor(e,t,n,r){if(null==e)throw new Error("Values passed to engine.makeTensor() are null");n=n||"float32",r=r||this.backend;let a=e;"string"===n&&ye(e[0])&&(a=e.map(e=>Ba(e)));let s=r.write(a,t,n),i=new ts(t,n,s,this.nextTensorId());if(this.trackTensor(i,r),"string"===n){let e=this.state.tensorInfo.get(s),t=ge(a);this.state.numBytes+=t-e.bytes,e.bytes=t}return i}makeTensorFromDataId(e,t,n,r){let a={dataId:e,shape:t,dtype:n=n||"float32"};return this.makeTensorFromTensorInfo(a,r)}makeTensorFromTensorInfo(e,t){let{dataId:n,shape:r,dtype:a}=e,s=new ts(r,a,n,this.nextTensorId());return this.trackTensor(s,t),s}makeVariable(e,t=!0,n,r){n=n||this.nextVariableId().toString(),null!=r&&r!==e.dtype&&(e=e.cast(r));let a=new rs(e,t,n,this.nextTensorId());if(null!=this.state.registeredVariables[a.name])throw new Error(`Variable with name ${a.name} was already registered`);return this.state.registeredVariables[a.name]=a,this.incRef(a,this.backend),a}trackTensor(e,t){this.state.numTensors++,"string"===e.dtype&&this.state.numStringTensors++;let n=0;"complex64"!==e.dtype&&"string"!==e.dtype&&(n=e.size*me(e.dtype)),this.state.numBytes+=n,this.state.tensorInfo.has(e.dataId)||(this.state.numDataBuffers++,this.state.tensorInfo.set(e.dataId,{backend:t||this.backend,dtype:e.dtype,shape:e.shape,bytes:n})),e instanceof rs||this.track(e)}incRef(e,t){this.trackTensor(e,t),this.backend.incRef(e.dataId)}removeDataId(e,t){this.state.tensorInfo.has(e)&&this.state.tensorInfo.get(e).backend===t&&(this.state.tensorInfo.delete(e),this.state.numDataBuffers--)}disposeTensor(e){if(!this.state.tensorInfo.has(e.dataId))return;let t=this.state.tensorInfo.get(e.dataId);if(this.state.numTensors--,"string"===e.dtype&&(this.state.numStringTensors--,this.state.numBytes-=t.bytes),"complex64"!==e.dtype&&"string"!==e.dtype){let t=e.size*me(e.dtype);this.state.numBytes-=t}t.backend.disposeData(e.dataId)&&this.removeDataId(e.dataId,t.backend)}disposeVariables(){for(let e in this.state.registeredVariables){let t=this.state.registeredVariables[e];this.disposeVariable(t)}}disposeVariable(e){this.disposeTensor(e),null!=this.state.registeredVariables[e.name]&&delete this.state.registeredVariables[e.name]}memory(){let e=this.backend.memory();return e.numTensors=this.state.numTensors,e.numDataBuffers=this.state.numDataBuffers,e.numBytes=this.state.numBytes,this.state.numStringTensors>0&&(e.unreliable=!0,null==e.reasons&&(e.reasons=[]),e.reasons.push("Memory usage by string tensors is approximate (2 bytes per character)")),e}async profile(e){this.state.profiling=!0;let t=this.state.numBytes,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(e=>e.totalBytesSnapshot)),this.state.activeProfile.newBytes=this.state.numBytes-t,this.state.activeProfile.newTensors=this.state.numTensors-n;for(let e of this.state.activeProfile.kernels)e.kernelTimeMs=await e.kernelTimeMs,e.extraInfo=await e.extraInfo;return this.state.activeProfile}isTapeOn(){return this.state.gradientDepth>0&&0===this.state.kernelDepth}addTapeNode(e,t,n,r,a,s){let i={id:this.state.nextTapeNodeId++,kernelName:e,inputs:t,outputs:n,saved:a},o=pa(e);null!=o&&(r=o.gradFunc),null!=r&&(i.gradient=e=>(e=e.map((e,t)=>{if(null==e){let e=n[t],r=Ce(e.size,e.dtype);return this.makeTensor(r,e.shape,e.dtype)}return e}),r(e.length>1?e:e[0],a,s))),this.state.activeTape.push(i)}keep(e){return e.kept=!0,e}startTape(){0===this.state.gradientDepth&&(this.state.activeTape=[]),this.state.gradientDepth++}endTape(){this.state.gradientDepth--}startScope(e){let t={track:[],name:"unnamed scope",id:this.state.nextScopeId++};e&&(t.name=e),this.state.scopeStack.push(t),this.state.activeScope=t}endScope(e){let t=xs(e),n=new Set(t.map(e=>e.id));for(let e=0;e<this.state.activeScope.track.length;e++){let t=this.state.activeScope.track[e];!t.kept&&!n.has(t.id)&&t.dispose()}let r=this.state.scopeStack.pop();this.state.activeScope=0===this.state.scopeStack.length?null:this.state.scopeStack[this.state.scopeStack.length-1],t.forEach(e=>{!e.kept&&e.scopeId===r.id&&this.track(e)})}gradients(e,t,n,r=!1){if(K(t.length>0,()=>"gradients() received an empty list of xs."),null!=n&&"float32"!==n.dtype)throw new Error(`dy must have 'float32' dtype, but has '${n.dtype}'`);let a=this.scopedRun(()=>this.startTape(),()=>this.endTape(),()=>this.tidy("forward",e));K(a instanceof ts,()=>"The result y returned by f() must be a tensor.");let s=function(e,t,n){let r={},a={};for(let e=0;e<t.length;e++)r[t[e].id]=!0;for(let n=0;n<e.length;n++){let s=e[n],i=s.inputs;for(let e in i){let n=i[e],o=!1;for(let e=0;e<t.length;e++)if(r[n.id]){s.outputs.forEach(e=>r[e.id]=!0),o=!0,a[s.id]=!0;break}if(o)break}}let s={};s[n.id]=!0;let i={};for(let t=e.length-1;t>=0;t--){let n=e[t],r=n.inputs;for(let e=0;e<n.outputs.length;e++)if(s[n.outputs[e].id]){for(let e in r)s[r[e].id]=!0,i[n.id]=!0;break}}let o=[];for(let t=0;t<e.length;t++){let n=e[t];if(a[n.id]&&i[n.id]){let e={};for(let t in n.inputs){let a=n.inputs[t];r[a.id]&&(e[t]=a)}let t=Object.assign({},n);t.inputs=e,t.outputs=n.outputs,o.push(t)}}return o}(this.state.activeTape,t,a);if(!r&&0===s.length&&t.length>0)throw new Error("Cannot compute gradient of y=f(x) with respect to x. Make sure that the f you passed encloses all operations that lead from x to y.");return this.tidy("backward",()=>{let e={};e[a.id]=n??function(e){let t=Te(Y(e),"float32");return _s.makeTensor(t,e,"float32")}(a.shape),function(e,t,n,r){for(let a=t.length-1;a>=0;a--){let s=t[a],i=[];if(s.outputs.forEach(t=>{let n=e[t.id];null!=n?i.push(n):i.push(null)}),null==s.gradient)throw new Error(`Cannot compute gradient: gradient function not found for ${s.kernelName}.`);let o=s.gradient(i);for(let t in s.inputs){if(!(t in o))throw new Error(`Cannot backprop through input ${t}. Available gradients found: ${Object.keys(o)}.`);let a=n(()=>o[t]());if("float32"!==a.dtype)throw new Error(`Error in gradient for op ${s.kernelName}. The gradient of input ${t} must have 'float32' dtype, but has '${a.dtype}'`);let i=s.inputs[t];if(!ee(a.shape,i.shape))throw new Error(`Error in gradient for op ${s.kernelName}. The gradient of input '${t}' has shape '${a.shape}', which does not match the shape of the input '${i.shape}'`);if(null==e[i.id])e[i.id]=a;else{let t=e[i.id];e[i.id]=r(t,a),t.dispose()}}}}(e,s,e=>this.tidy(e),Ns);let r=t.map(t=>e[t.id]);return 0===this.state.gradientDepth&&(this.state.activeTape.forEach(e=>{for(let t of e.saved)t.dispose()}),this.state.activeTape=null),{value:a,grads:r}})}customGrad(e){return K(we(e),()=>"The f passed in customGrad(f) must be a function."),(...t)=>{K(t.every(e=>e instanceof ts),()=>"The args passed in customGrad(f)(x1, x2,...) must all be tensors");let n,r={};t.forEach((e,t)=>{r[t]=e});return this.runKernelFunc({forwardFunc:(r,a)=>(n=e(...t,a),K(n.value instanceof ts,()=>"The function f passed in customGrad(f) must return an object where `obj.value` is a tensor"),K(we(n.gradFunc),()=>"The function f passed in customGrad(f) must return an object where `obj.gradFunc` is a function."),n.value),backwardsFunc:(e,r)=>{let a=n.gradFunc(e,r),s=Array.isArray(a)?a:[a];K(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(...)."),K(s.every(e=>e instanceof ts),()=>"The function f passed in customGrad(f) must return an object where `obj.gradFunc` is a function that returns a list of only tensors.");let i={};return s.forEach((e,t)=>{i[t]=()=>e}),i},inputs:r})}}readSync(e){return this.state.tensorInfo.get(e).backend.readSync(e)}read(e){return this.state.tensorInfo.get(e).backend.read(e)}readToGPU(e,t){return this.state.tensorInfo.get(e).backend.readToGPU(e,t)}async time(e){let t=Pa(),n=await this.backend.time(e);return n.wallMs=Pa()-t,n}track(e){return null!=this.state.activeScope&&(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 ks;for(let e in this.registry)this.disposeRegisteredKernels(e),this.registry[e].dispose(),delete this.registry[e];this.backendName=null,this.backendInstance=null,this.pendingBackendInit=null}};function Ss(){let e=Be();if(null==e._tfengine){let t=new Me(e);e._tfengine=new Is(t)}return function(e){ze=e}(e._tfengine.ENV),function(e){Qa=e}(()=>e._tfengine),e._tfengine}Is.nextTensorId=0,Is.nextVariableId=0;var _s=Ss();function Ns(e,t){let n={a:e,b:t};return _s.runKernel(He,n)}var Ts,Cs={};function Es(e){Ts=e}function As(e){if(void 0!==Ts)return Ts;if(e||"undefined"!=typeof navigator&&null!=navigator){if(e||(e=navigator),"ReactNative"===e.product)return!0;let t=e.userAgent||e.vendor||("undefined"!=typeof window?window.opera:"");if(!t){let t=e;return t.userAgentData&&t.userAgentData.mobile}return/(android|bb\d+|meego).+mobile|avantgo|bada\/|blackberry|blazer|compal|elaine|fennec|hiptop|iemobile|ip(hone|od)|iris|kindle|lge |maemo|midp|mmp|mobile.+firefox|netfront|opera m(ob|in)i|palm( os)?|phone|p(ixi|re)\/|plucker|pocket|psp|series(4|6)0|symbian|treo|up\.(browser|link)|vodafone|wap|windows ce|xda|xiino/i.test(t)||/1207|6310|6590|3gso|4thp|50[1-6]i|770s|802s|a wa|abac|ac(er|oo|s\-)|ai(ko|rn)|al(av|ca|co)|amoi|an(ex|ny|yw)|aptu|ar(ch|go)|as(te|us)|attw|au(di|\-m|r |s )|avan|be(ck|ll|nq)|bi(lb|rd)|bl(ac|az)|br(e|v)w|bumb|bw\-(n|u)|c55\/|capi|ccwa|cdm\-|cell|chtm|cldc|cmd\-|co(mp|nd)|craw|da(it|ll|ng)|dbte|dc\-s|devi|dica|dmob|do(c|p)o|ds(12|\-d)|el(49|ai)|em(l2|ul)|er(ic|k0)|esl8|ez([4-7]0|os|wa|ze)|fetc|fly(\-|_)|g1 u|g560|gene|gf\-5|g\-mo|go(\.w|od)|gr(ad|un)|haie|hcit|hd\-(m|p|t)|hei\-|hi(pt|ta)|hp( i|ip)|hs\-c|ht(c(\-| |_|a|g|p|s|t)|tp)|hu(aw|tc)|i\-(20|go|ma)|i230|iac( |\-|\/)|ibro|idea|ig01|ikom|im1k|inno|ipaq|iris|ja(t|v)a|jbro|jemu|jigs|kddi|keji|kgt( |\/)|klon|kpt |kwc\-|kyo(c|k)|le(no|xi)|lg( g|\/(k|l|u)|50|54|\-[a-w])|libw|lynx|m1\-w|m3ga|m50\/|ma(te|ui|xo)|mc(01|21|ca)|m\-cr|me(rc|ri)|mi(o8|oa|ts)|mmef|mo(01|02|bi|de|do|t(\-| |o|v)|zz)|mt(50|p1|v )|mwbp|mywa|n10[0-2]|n20[2-3]|n30(0|2)|n50(0|2|5)|n7(0(0|1)|10)|ne((c|m)\-|on|tf|wf|wg|wt)|nok(6|i)|nzph|o2im|op(ti|wv)|oran|owg1|p800|pan(a|d|t)|pdxg|pg(13|\-([1-8]|c))|phil|pire|pl(ay|uc)|pn\-2|po(ck|rt|se)|prox|psio|pt\-g|qa\-a|qc(07|12|21|32|60|\-[2-7]|i\-)|qtek|r380|r600|raks|rim9|ro(ve|zo)|s55\/|sa(ge|ma|mm|ms|ny|va)|sc(01|h\-|oo|p\-)|sdk\/|se(c(\-|0|1)|47|mc|nd|ri)|sgh\-|shar|sie(\-|m)|sk\-0|sl(45|id)|sm(al|ar|b3|it|t5)|so(ft|ny)|sp(01|h\-|v\-|v )|sy(01|mb)|t2(18|50)|t6(00|10|18)|ta(gt|lk)|tcl\-|tdg\-|tel(i|m)|tim\-|t\-mo|to(pl|sh)|ts(70|m\-|m3|m5)|tx\-9|up(\.b|g1|si)|utst|v400|v750|veri|vi(rg|te)|vk(40|5[0-3]|\-v)|vm40|voda|vulc|vx(52|53|60|61|70|80|81|83|85|98)|w3c(\-| )|webc|whit|wi(g |nc|nw)|wmlb|wonu|x700|yas\-|your|zeto|zte\-/i.test(t.substr(0,4))}return!1}function $s(){return"undefined"!=typeof window&&null!=window.document||"undefined"!=typeof WorkerGlobalScope}f(Cs,{isBrowser:()=>$s,isMobile:()=>As,mockIsMobile:()=>Es});var Rs=Le();function Fs(e,t){let n=e;if(Va(e))return"string"===t?[]:[e.length];if(fs(e)){let t=e.channels||"RGBA";return[e.height,e.width*t.length]}if(ms(e))return[e.buffer.size/(null==t?4:me(t))];if(!Array.isArray(e))return[];let r=[];for(;Array.isArray(n)||Va(n)&&"string"!==t;)r.push(n.length),n=n[0];return Array.isArray(e)&&Le().getBool("TENSORLIKE_CHECK_SHAPE_CONSISTENCY")&&Ds(e,r,[]),r}function Ds(e,t,n){if(n=n||[],!Array.isArray(e)&&!Va(e))return void K(0===t.length,()=>`Element arr[${n.join("][")}] is a primitive, but should be an array/TypedArray of ${t[0]} elements`);K(t.length>0,()=>`Element arr[${n.join("][")}] should be a primitive, but is an array of ${e.length} elements`),K(e.length===t[0],()=>`Element arr[${n.join("][")}] should have ${t[0]} elements, but has ${e.length} elements`);let r=t.slice(1);for(let t=0;t<e.length;++t)Ds(e[t],r,n.concat(t))}function Ms(e,t,n,r){if("string_or_numeric"!==e){if(null==e)throw new Error("Expected dtype cannot be null.");if("numeric"!==e&&e!==t||"numeric"===e&&"string"===t)throw new Error(`Argument '${n}' passed to '${r}' must be ${e} tensor, but got ${t} tensor`)}}function Os(e,t,n,r="numeric"){if(e instanceof ns())return Ms(r,e.dtype,t,n),e;let a=ve(e);if("string"!==a&&["bool","int32","float32"].indexOf(r)>=0&&(a=r),Ms(r,a,t,n),null==e||!Va(e)&&!Array.isArray(e)&&"number"!=typeof e&&"boolean"!=typeof e&&"string"!=typeof e){let r=null==e?"null":e.constructor.name;throw new Error(`Argument '${t}' passed to '${n}' must be a Tensor or TensorLike, but got '${r}'`)}let s=Fs(e,a);!Va(e)&&!Array.isArray(e)&&(e=[e]);let i="string"!==a?La(e,a):Ua(e,[],!0);return _s.makeTensor(i,s,a)}function Ls(e,t,n,r="numeric"){if(!Array.isArray(e))throw new Error(`Argument ${t} passed to ${n} must be a \`Tensor[]\` or \`TensorLike[]\``);return e.map((e,a)=>Os(e,`${t}[${a}]`,n,r))}Rs.registerFlag("DEBUG",()=>!1,e=>{e&&console.warn("Debugging mode is ON. The output of every math call will be downloaded to CPU and checked for NaNs. This significantly impacts performance.")}),Rs.registerFlag("IS_BROWSER",()=>$s()),Rs.registerFlag("IS_NODE",()=>"undefined"!=typeof process&&void 0!==process.versions&&void 0!==process.versions.node),Rs.registerFlag("IS_CHROME",()=>"undefined"!=typeof navigator&&null!=navigator&&null!=navigator.userAgent&&/Chrome/.test(navigator.userAgent)&&/Google Inc/.test(navigator.vendor)),Rs.registerFlag("IS_SAFARI",()=>"undefined"!=typeof navigator&&null!=navigator&&null!=navigator.userAgent&&/Safari/.test(navigator.userAgent)&&/Apple/.test(navigator.vendor)),Rs.registerFlag("PROD",()=>!1),Rs.registerFlag("TENSORLIKE_CHECK_SHAPE_CONSISTENCY",()=>Rs.getBool("DEBUG")),Rs.registerFlag("DEPRECATION_WARNINGS_ENABLED",()=>!0),Rs.registerFlag("IS_TEST",()=>!1),Rs.registerFlag("CHECK_COMPUTATION_FOR_ERRORS",()=>Rs.getBool("DEBUG")),Rs.registerFlag("WRAP_TO_IMAGEBITMAP",()=>!1),Rs.registerFlag("CANVAS2D_WILL_READ_FREQUENTLY_FOR_GPU",()=>!1),Rs.registerFlag("USE_SETTIMEOUTCUSTOM",()=>!1);var Ps="__op";function zs(e){let t=Object.keys(e);if(1!==t.length)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],r=e[n];n.endsWith("_")&&(n=n.substring(0,n.length-1)),n+=Ps;let a=(...e)=>{_s.startScope(n);try{let t=r(...e);return Fe(t)&&console.error("Cannot return a Promise inside of tidy."),_s.endScope(t),t}catch(e){throw _s.endScope(null),e}};return Object.defineProperty(a,"name",{value:n,configurable:!0}),a}var Bs=zs({complex_:function(e,t){let n=Os(e,"real","complex"),r=Os(t,"imag","complex");X(n.shape,r.shape,`real and imag shapes, ${n.shape} and ${r.shape}, must match in call to tf.complex().`);let a={real:n,imag:r};return _s.runKernel(mt,a)}});function Ws(e,t,n,r){if(null==r)r=ve(e);else if("complex64"===r)throw new Error("Cannot construct a complex64 tensor directly. Please use tf.complex(real, imag).");if(ms(e)||fs(e)){if("float32"!==r&&"int32"!==r)throw new Error(`Creating tensor from GPU data only supports 'float32'|'int32' dtype, while the dtype is ${r}.`);return _s.backend.createTensorFromGPUData(e,t||n,r)}if(!Va(e)&&!Array.isArray(e)&&"number"!=typeof e&&"boolean"!=typeof e&&"string"!=typeof e)throw new Error("values passed to tensor(values) must be a number/boolean/string or an array of numbers/booleans/strings, or a TypedArray");if(null!=t){Ae(t);let e=Y(t),r=Y(n);K(e===r,()=>`Based on the provided shape, [${t}], the tensor should have ${e} values but has ${r}`);for(let e=0;e<n.length;++e){let r=n[e],a=e!==n.length-1||r!==Y(t.slice(e));K(n[e]===t[e]||!a,()=>`Error creating a new Tensor. Inferred shape (${n}) does not match the provided shape (${t}). `)}}return!Va(e)&&!Array.isArray(e)&&(e=[e]),t=t||n,e="string"!==r?La(e,r):Ua(e,[],!0),_s.makeTensor(e,t,r)}function Vs(e,t,n){return Ws(e,t,Fs(e,n),n)}var Us={float32:4,float16:2,int32:4,uint16:2,uint8:1,bool:1,complex64:8},Gs=class e{static join(t){return new e(t).slice()}constructor(e){if(this.shards=[],this.previousShardIndex=0,null==e||(e instanceof Array||(e=[e]),0===(e=e.map(e=>Va(e)?e.buffer:e)).length))return;this.bufferUniformSize=e[0].byteLength;let t=0;for(let n=0;n<e.length;n++){let r=e[n];n!==e.length-1&&r.byteLength!==this.bufferUniformSize&&(this.bufferUniformSize=void 0);let a=t+r.byteLength;this.shards.push({buffer:r,start:t,end:a}),t=a}0===this.shards.length&&(this.byteLength=0),this.byteLength=this.shards[this.shards.length-1].end}slice(e=0,t=this.byteLength){if(0===this.shards.length)return new ArrayBuffer(0);if(e=isNaN(Number(e))?0:e,t=isNaN(Number(t))?0:t,e=Math.max(0,e),(t=Math.min(this.byteLength,t))<=e)return new ArrayBuffer(0);let n=this.findShardForByte(e);if(-1===n)throw new Error(`Could not find start shard for byte ${e}`);let r=new ArrayBuffer(t-e),a=new Uint8Array(r),s=0;for(let r=n;r<this.shards.length;r++){let n=this.shards[r],i=e+s-n.start,o=s,l=Math.min(t,n.end)-n.start,u=new Uint8Array(n.buffer,i,l-i);if(a.set(u,o),s+=u.length,t<n.end)break}return r}findShardForByte(e){if(0===this.shards.length||e<0||e>=this.byteLength)return-1;if(null!=this.bufferUniformSize)return this.previousShardIndex=Math.floor(e/this.bufferUniformSize),this.previousShardIndex;function t(t){return e<t.start?-1:e>=t.end?1:0}if(0===t(this.shards[this.previousShardIndex]))return this.previousShardIndex;let n=function(e,t){let n=0,r=e.length;for(;n<=r;){let a=Math.floor((r-n)/2)+n,s=t(e[a]);if(0===s)return a;s<0?r=a:n=a+1}return-1}(this.shards,t);return-1===n?-1:(this.previousShardIndex=n,this.previousShardIndex)}};function Hs(){Le().set("PROD",!0)}function js(){Le().set("DEBUG",!0)}function qs(){Le().set("DEPRECATION_WARNINGS_ENABLED",!1),console.warn("TensorFlow.js deprecation warnings have been disabled.")}function Ks(e){Le().getBool("DEPRECATION_WARNINGS_ENABLED")&&console.warn(e+" You can disable deprecation warnings with tf.disableDeprecationWarnings().")}function Xs(){_s.disposeVariables()}function Zs(){return _s}function Ys(){return _s.memory()}function Js(e){return _s.profile(e)}function Qs(e,t){return _s.tidy(e,t)}function ei(e){xs(e).forEach(e=>e.dispose())}function ti(e){return _s.keep(e)}function ni(e){return _s.time(e)}function ri(e){return _s.setBackend(e)}function ai(){return _s.ready()}function si(){return _s.backendName}function ii(e){_s.removeBackend(e)}function oi(e){return _s.findBackend(e)}function li(e){return _s.findBackendFactory(e)}function ui(e,t,n=1){return _s.registerBackend(e,t,n)}function hi(){return _s.backend}function di(e,t){Le().setPlatform(e,t)}!function(e){0}(Ks);async function pi(e,t){let n=[],r=[],a=Array.isArray(e)?e.map(e=>e.name):Object.keys(e);for(let s=0;s<a.length;++s){let i=a[s],o=Array.isArray(e)?e[s].tensor:e[i];if("float32"!==o.dtype&&"int32"!==o.dtype&&"bool"!==o.dtype&&"string"!==o.dtype&&"complex64"!==o.dtype)throw new Error(`Unsupported dtype in weight '${i}': ${o.dtype}`);let l={name:i,shape:o.shape,dtype:o.dtype};if("string"===o.dtype){let e=new Promise(async e=>{let t=await o.bytes(),n=t.reduce((e,t)=>e+t.length,0)+4*t.length,r=new Uint8Array(n),a=0;for(let e=0;e<t.length;e++){let n=t[e],s=new Uint8Array(new Uint32Array([n.length]).buffer);r.set(s,a),a+=4,r.set(n,a),a+=n.length}e(r)});r.push(e)}else r.push(o.data());null!=t&&(l.group=t),n.push(l)}return{data:xi(await Promise.all(r)),specs:n}}function ci(e,t){let n=new Gs(e),r={},a=0;for(let e of t){let t=fi(e,(e,t)=>n.slice(a+e,a+t));r[e.name]=gi(e,n.slice(a,a+t)),a+=t}return r}function fi(e,t){let n,r=Y(e.shape);if("quantization"in e){let t=e.quantization;n=Us[t.dtype]}else{if("string"===e.dtype){let e=0;for(let n=0;n<r;n++)e+=4+new Uint32Array(t(e,e+4))[0];return e}n=Us[e.dtype]}return r*n}async function mi(e,t){let n,r=Y(e.shape);if("quantization"in e){let t=e.quantization;n=Us[t.dtype]}else{if("string"===e.dtype){let e=0;for(let n=0;n<r;n++)e+=4+new Uint32Array(await t(e,e+4))[0];return e}n=Us[e.dtype]}return r*n}function gi(e,t){let n,r=e.name,a=e.dtype,s=e.shape,i=Y(s),o=0;if("quantization"in e){let s=e.quantization;if("uint8"===s.dtype||"uint16"===s.dtype){if(!("min"in s)||!("scale"in s))throw new Error(`Weight ${e.name} with quantization ${s.dtype} doesn't have corresponding metadata min and scale.`)}else{if("float16"!==s.dtype)throw new Error(`Weight ${e.name} has unknown quantization dtype ${s.dtype}. Supported quantization dtypes are: 'uint8', 'uint16', and 'float16'.`);if("float32"!==a)throw new Error(`Weight ${e.name} is quantized with ${s.dtype} which only supports weights of type float32 not ${a}.`)}let l=Us[s.dtype],u="uint8"===s.dtype?new Uint8Array(t):new Uint16Array(t);if("float32"===a)if("uint8"===s.dtype||"uint16"===s.dtype){n=new Float32Array(u.length);for(let e=0;e<u.length;e++){let t=u[e];n[e]=t*s.scale+s.min}}else{if("float16"!==s.dtype)throw new Error(`Unsupported quantization type ${s.dtype} for weight type float32.`);n=function(){let e=function(){let e=e=>{let t=e<<13,n=0;for(;!(8388608&t);)n-=8388608,t<<=1;return t&=-8388609,n+=947912704,t|n},t=new Uint32Array(2048);t[0]=0;for(let n=1;n<1024;n++)t[n]=e(n);for(let e=1024;e<2048;e++)t[e]=939524096+(e-1024<<13);return t}(),t=function(){let e=new Uint32Array(64);e[0]=0,e[31]=1199570944,e[32]=2147483648,e[63]=3347054592;for(let t=1;t<31;t++)e[t]=t<<23;for(let t=33;t<63;t++)e[t]=2147483648+(t-32<<23);return e}(),n=function(){let e=new Uint32Array(64);for(let t=0;t<64;t++)e[t]=1024;return e[0]=e[32]=0,e}();return r=>{let a=new ArrayBuffer(4*r.length),s=new Uint32Array(a);for(let a=0;a<r.length;a++){let i=r[a],o=e[n[i>>10]+(1023&i)]+t[i>>10];s[a]=o}return new Float32Array(a)}}()(u)}else{if("int32"!==a)throw new Error(`Unsupported dtype in weight '${r}': ${a}`);if("uint8"!==s.dtype&&"uint16"!==s.dtype)throw new Error(`Unsupported quantization type ${s.dtype} for weight type int32.`);n=new Int32Array(u.length);for(let e=0;e<u.length;e++){let t=u[e];n[e]=Math.round(t*s.scale+s.min)}}o+=i*l}else if("string"===a){let r=Y(e.shape);n=[];for(let e=0;e<r;e++){let e=new Uint32Array(t.slice(o,o+4))[0];o+=4;let r=new Uint8Array(t.slice(o,o+e));n.push(r),o+=e}}else{let e=Us[a];if("float32"===a)n=new Float32Array(t);else if("int32"===a)n=new Int32Array(t);else{if("bool"!==a){if("complex64"===a){n=new Float32Array(t);let e=new Float32Array(n.length/2),r=new Float32Array(n.length/2);for(let t=0;t<e.length;t++)e[t]=n[2*t],r[t]=n[2*t+1];let a=Vs(e,s,"float32"),i=Vs(r,s,"float32"),o=Bs(a,i);return a.dispose(),i.dispose(),o}throw new Error(`Unsupported dtype in weight '${r}': ${a}`)}n=new Uint8Array(t)}o+=i*e}return Vs(n,s,a)}async function yi(e,t,n){let r=new Uint8Array(t);for(;r.byteLength<n;){let{done:t,value:a}=await e.read();if(t&&null==a){let e=n-r.byteLength;throw new Error(`Reader is done but ${e} bytes are still expected`)}let s=new Uint8Array(r.length+a.byteLength);s.set(r,0),s.set(new Uint8Array(a),r.length),r=s}return r.buffer}async function bi(e,t){let n={},r=e.getReader(),a=new ArrayBuffer(0);for(let e of t){let t=await mi(e,async(e,t)=>(a=await yi(r,a,t),a.slice(e,t)));a=await yi(r,a,t);let s=a.slice(0,t);a=a.slice(t);let i=gi(e,s);if(n[e.name]=i,"webgpu"===si()){let e=hi();"uploadToGPU"in e&&Y(i.shape)>=Le().get("WEBGPU_CPU_HANDOFF_SIZE_THRESHOLD")&&e.uploadToGPU(i.dataId)}}return n}function xi(e){if(null===e)throw new Error(`Invalid input value: ${JSON.stringify(e)}`);let t=0,n=[];e.forEach(e=>{if(t+=e.byteLength,n.push(e.byteLength===e.buffer.byteLength?e:new e.constructor(e)),!(e instanceof Float32Array||e instanceof Int32Array||e instanceof Uint8Array))throw new Error(`Unsupported TypedArray subtype: ${e.constructor.name}`)});let r=new Uint8Array(t),a=0;return n.forEach(e=>{r.set(new Uint8Array(e.buffer),a),a+=e.byteLength}),r.buffer}var vi="undefined"!=typeof Buffer&&("undefined"==typeof Blob||"undefined"==typeof atob||"undefined"==typeof btoa);function wi(e){return vi?Buffer.byteLength(e,"utf8"):new Blob([e]).size}function ki(e){return Gs.join(e)}function Ii(e){for(e=e.trim();e.endsWith("/");)e=e.slice(0,e.length-1);let t=e.split("/");return t[t.length-1]}function Si(e,t){let n={modelTopology:e.modelTopology,format:e.format,generatedBy:e.generatedBy,convertedBy:e.convertedBy,weightsManifest:t};return null!=e.signature&&(n.signature=e.signature),null!=e.userDefinedMetadata&&(n.userDefinedMetadata=e.userDefinedMetadata),null!=e.modelInitializer&&(n.modelInitializer=e.modelInitializer),null!=e.initializerSignature&&(n.initializerSignature=e.initializerSignature),null!=e.trainingConfig&&(n.trainingConfig=e.trainingConfig),n}function _i(e,t,n){let r={modelTopology:e.modelTopology,format:e.format,generatedBy:e.generatedBy,convertedBy:e.convertedBy};if(null!=e.trainingConfig&&(r.trainingConfig=e.trainingConfig),null!=e.weightsManifest){if(!t)throw new Error("modelJSON has weightsManifest but weightSpecs is null");if(!n)throw new Error("modelJSON has weightsManifest but weightData is null");r.weightSpecs=t,r.weightData=n}return null!=e.signature&&(r.signature=e.signature),null!=e.userDefinedMetadata&&(r.userDefinedMetadata=e.userDefinedMetadata),null!=e.modelInitializer&&(r.modelInitializer=e.modelInitializer),null!=e.initializerSignature&&(r.initializerSignature=e.initializerSignature),r}async function Ni(e,t){let n,r;return null!=e.weightsManifest&&([n,r]=await t(e.weightsManifest)),_i(e,n,r)}function Ti(e){if(e.modelTopology instanceof ArrayBuffer)throw new Error("Expected JSON model topology, received ArrayBuffer.");return{dateSaved:new Date,modelTopologyType:"JSON",modelTopologyBytes:null==e.modelTopology?0:wi(JSON.stringify(e.modelTopology)),weightSpecsBytes:null==e.weightSpecs?0:wi(JSON.stringify(e.weightSpecs)),weightDataBytes:null==e.weightData?0:new Gs(e.weightData).byteLength}}function Ci(e){let t=[];for(let n of e)t.push(...n.weights);return t}var Ei=class e{constructor(){this.saveRouters=[],this.loadRouters=[]}static getInstance(){return null==e.instance&&(e.instance=new e),e.instance}static registerSaveRouter(t){e.getInstance().saveRouters.push(t)}static registerLoadRouter(t){e.getInstance().loadRouters.push(t)}static getSaveHandlers(t){return e.getHandlers(t,"save")}static getLoadHandlers(t,n){return e.getHandlers(t,"load",n)}static getHandlers(t,n,r){let a=[];return("load"===n?e.getInstance().loadRouters:e.getInstance().saveRouters).forEach(e=>{let n=e(t,r);null!==n&&a.push(n)}),a}},Ai=e=>Ei.registerSaveRouter(e),$i=e=>Ei.registerLoadRouter(e),Ri=e=>Ei.getSaveHandlers(e),Fi=(e,t)=>Ei.getLoadHandlers(e,t),Di="tensorflowjs",Mi="models_store",Oi="model_info_store";function Li(){if(!Le().getBool("IS_BROWSER"))throw new Error("Failed to obtain IndexedDB factory because the current environmentis not a web browser.");let e="undefined"==typeof window?self:window,t=e.indexedDB||e.mozIndexedDB||e.webkitIndexedDB||e.msIndexedDB||e.shimIndexedDB;if(null==t)throw new Error("The current browser does not appear to support IndexedDB.");return t}function Pi(e){let t=e.result;t.createObjectStore(Mi,{keyPath:"modelPath"}),t.createObjectStore(Oi,{keyPath:"modelPath"})}var zi=class{constructor(e){if(this.indexedDB=Li(),null==e||!e)throw new Error("For IndexedDB, modelPath must not be null, undefined or empty.");this.modelPath=e}async save(e){if(e.modelTopology instanceof ArrayBuffer)throw new Error("BrowserLocalStorage.save() does not support saving model topology in binary formats yet.");return this.databaseAction(this.modelPath,e)}async load(){return this.databaseAction(this.modelPath)}databaseAction(e,t){return new Promise((e,n)=>{let r=this.indexedDB.open(Di,1);r.onupgradeneeded=()=>Pi(r),r.onsuccess=()=>{let a=r.result;if(null==t){let t=a.transaction(Mi,"readonly"),r=t.objectStore(Mi).get(this.modelPath);r.onsuccess=()=>{if(null==r.result)return a.close(),n(new Error(`Cannot find model with path '${this.modelPath}' in IndexedDB.`));e(r.result.modelArtifacts)},r.onerror=e=>(a.close(),n(r.error)),t.oncomplete=()=>a.close()}else{t.weightData=Gs.join(t.weightData);let r,s,i=Ti(t),o=a.transaction(Oi,"readwrite"),l=o.objectStore(Oi);try{r=l.put({modelPath:this.modelPath,modelArtifactsInfo:i})}catch(e){return n(e)}r.onsuccess=()=>{s=a.transaction(Mi,"readwrite");let r,u=s.objectStore(Mi);try{r=u.put({modelPath:this.modelPath,modelArtifacts:t,modelArtifactsInfo:i})}catch(e){return n(e)}r.onsuccess=()=>e({modelArtifactsInfo:i}),r.onerror=e=>{l=o.objectStore(Oi);let t=l.delete(this.modelPath);t.onsuccess=()=>(a.close(),n(r.error)),t.onerror=e=>(a.close(),n(r.error))}},r.onerror=e=>(a.close(),n(r.error)),o.oncomplete=()=>{null==s?a.close():s.oncomplete=()=>a.close()}}},r.onerror=e=>n(r.error)})}};zi.URL_SCHEME="indexeddb://";var Bi=e=>Le().getBool("IS_BROWSER")&&!Array.isArray(e)&&e.startsWith(zi.URL_SCHEME)?function(e){return new zi(e)}(e.slice(zi.URL_SCHEME.length)):null;Ei.registerSaveRouter(Bi),Ei.registerLoadRouter(Bi);var Wi="/",Vi="tensorflowjs_models",Ui="info",Gi="model_topology",Hi="weight_specs",ji="weight_data",qi="model_metadata";function Ki(e){return{info:[Vi,e,Ui].join(Wi),topology:[Vi,e,Gi].join(Wi),weightSpecs:[Vi,e,Hi].join(Wi),weightData:[Vi,e,ji].join(Wi),modelMetadata:[Vi,e,qi].join(Wi)}}function Xi(e){for(let t of Object.values(e))window.localStorage.removeItem(t)}function Zi(e){let t=e.split(Wi);if(t.length<3)throw new Error(`Invalid key format: ${e}`);return t.slice(1,t.length-1).join(Wi)}var Yi=class{constructor(e){if(!Le().getBool("IS_BROWSER")||"undefined"==typeof window||void 0===window.localStorage)throw new Error("The current environment does not support local storage.");if(this.LS=window.localStorage,null==e||!e)throw new Error("For local storage, modelPath must not be null, undefined or empty.");this.modelPath=e,this.keys=Ki(this.modelPath)}async save(e){if(e.modelTopology instanceof ArrayBuffer)throw new Error("BrowserLocalStorage.save() does not support saving model topology in binary formats yet.");{let t=JSON.stringify(e.modelTopology),n=JSON.stringify(e.weightSpecs),r=Ti(e),a=Gs.join(e.weightData);try{this.LS.setItem(this.keys.info,JSON.stringify(r)),this.LS.setItem(this.keys.topology,t),this.LS.setItem(this.keys.weightSpecs,n),this.LS.setItem(this.keys.weightData,function(e){if(vi)return Buffer.from(e).toString("base64");let t=new Uint8Array(e),n="";for(let e=0,r=t.length;e<r;e++)n+=String.fromCharCode(t[e]);return btoa(n)}(a));let s={format:e.format,generatedBy:e.generatedBy,convertedBy:e.convertedBy,signature:null!=e.signature?e.signature:void 0,userDefinedMetadata:null!=e.userDefinedMetadata?e.userDefinedMetadata:void 0,modelInitializer:null!=e.modelInitializer?e.modelInitializer:void 0,initializerSignature:null!=e.initializerSignature?e.initializerSignature:void 0,trainingConfig:null!=e.trainingConfig?e.trainingConfig:void 0};return this.LS.setItem(this.keys.modelMetadata,JSON.stringify(s)),{modelArtifactsInfo:r}}catch(e){throw Xi(this.keys),new Error(`Failed to save model '${this.modelPath}' to local storage: size quota being exceeded is a possible cause of this failure: modelTopologyBytes=${r.modelTopologyBytes}, weightSpecsBytes=${r.weightSpecsBytes}, weightDataBytes=${r.weightDataBytes}.`)}}}async load(){let e=JSON.parse(this.LS.getItem(this.keys.info));if(null==e)throw new Error(`In local storage, there is no model with name '${this.modelPath}'`);if("JSON"!==e.modelTopologyType)throw new Error("BrowserLocalStorage does not support loading non-JSON model topology yet.");let t={},n=JSON.parse(this.LS.getItem(this.keys.topology));if(null==n)throw new Error(`In local storage, the topology of model '${this.modelPath}' is missing.`);t.modelTopology=n;let r=JSON.parse(this.LS.getItem(this.keys.weightSpecs));if(null==r)throw new Error(`In local storage, the weight specs of model '${this.modelPath}' are missing.`);t.weightSpecs=r;let a=this.LS.getItem(this.keys.modelMetadata);if(null!=a){let e=JSON.parse(a);t.format=e.format,t.generatedBy=e.generatedBy,t.convertedBy=e.convertedBy,null!=e.signature&&(t.signature=e.signature),null!=e.userDefinedMetadata&&(t.userDefinedMetadata=e.userDefinedMetadata),null!=e.modelInitializer&&(t.modelInitializer=e.modelInitializer),null!=e.initializerSignature&&(t.initializerSignature=e.initializerSignature),null!=e.trainingConfig&&(t.trainingConfig=e.trainingConfig)}let s=this.LS.getItem(this.keys.weightData);if(null==s)throw new Error(`In local storage, the binary weight values of model '${this.modelPath}' are missing.`);return t.weightData=function(e){if(vi){let t=Buffer.from(e,"base64");return t.buffer.slice(t.byteOffset,t.byteOffset+t.byteLength)}let t=atob(e),n=new Uint8Array(t.length);for(let e=0;e<t.length;++e)n.set([t.charCodeAt(e)],e);return n.buffer}(s),t}};Yi.URL_SCHEME="localstorage://";var Ji=e=>Le().getBool("IS_BROWSER")&&!Array.isArray(e)&&e.startsWith(Yi.URL_SCHEME)?function(e){return new Yi(e)}(e.slice(Yi.URL_SCHEME.length)):null;Ei.registerSaveRouter(Ji),Ei.registerLoadRouter(Ji);var Qi="://",eo=class e{constructor(){this.managers={}}static getInstance(){return null==e.instance&&(e.instance=new e),e.instance}static registerManager(t,n){K(null!=t,()=>"scheme must not be undefined or null."),t.endsWith(Qi)&&(t=t.slice(0,t.indexOf(Qi))),K(t.length>0,()=>"scheme must not be an empty string.");let r=e.getInstance();K(null==r.managers[t],()=>`A model store manager is already registered for scheme '${t}'.`),r.managers[t]=n}static getManager(t){let n=e.getInstance().managers[t];if(null==n)throw new Error(`Cannot find model manager for scheme '${t}'`);return n}static getSchemes(){return Object.keys(e.getInstance().managers)}};function to(e){if(-1===e.indexOf(Qi))throw new Error(`The url string provided does not contain a scheme. Supported schemes are: ${eo.getSchemes().join(",")}`);return{scheme:e.split(Qi)[0],path:e.split(Qi)[1]}}async function no(e,t,n=!1){K(e!==t,()=>`Old path and new path are the same: '${e}'`);let r=Ei.getLoadHandlers(e);K(r.length>0,()=>`Copying failed because no load handler is found for source URL ${e}.`),K(r.length<2,()=>`Copying failed because more than one (${r.length}) load handlers for source URL ${e}.`);let a=r[0],s=Ei.getSaveHandlers(t);K(s.length>0,()=>`Copying failed because no save handler is found for destination URL ${t}.`),K(s.length<2,()=>`Copying failed because more than one (${r.length}) save handlers for destination URL ${t}.`);let i=s[0],o=to(e).scheme,l=to(e).path,u=o===to(e).scheme,h=await a.load();n&&u&&await eo.getManager(o).removeModel(l);let d=await i.save(h);return n&&!u&&await eo.getManager(o).removeModel(l),d.modelArtifactsInfo}async function ro(){let e=eo.getSchemes(),t={};for(let n of e){let e=await eo.getManager(n).listModels();for(let r in e){t[n+Qi+r]=e[r]}}return t}async function ao(e){let t=to(e);return eo.getManager(t.scheme).removeModel(t.path)}async function so(e,t){return no(e,t,!1)}async function io(e,t){return no(e,t,!0)}if(Le().get("IS_BROWSER")){Le().setPlatform("browser",new class{constructor(){this.messageName="setTimeoutCustom",this.functionRefs=[],this.handledMessageCount=0,this.hasEventListener=!1}fetch(e,t){return fetch(e,t)}now(){return performance.now()}encode(e,t){if("utf-8"!==t&&"utf8"!==t)throw new Error(`Browser's encoder only supports utf-8, but got ${t}`);return null==this.textEncoder&&(this.textEncoder=new TextEncoder),this.textEncoder.encode(e)}decode(e,t){return new TextDecoder(t).decode(e)}setTimeoutCustom(e,t){"undefined"!=typeof window&&Le().getBool("USE_SETTIMEOUTCUSTOM")?(this.functionRefs.push(e),setTimeout(()=>{window.postMessage({name:this.messageName,index:this.functionRefs.length-1},"*")},t),this.hasEventListener||(this.hasEventListener=!0,window.addEventListener("message",e=>{e.source===window&&e.data.name===this.messageName&&(e.stopPropagation(),(0,this.functionRefs[e.data.index])(),this.handledMessageCount++,this.handledMessageCount===this.functionRefs.length&&(this.functionRefs=[],this.handledMessageCount=0))},!0))):setTimeout(e,t)}isTypedArray(e){return wa(e)}});try{eo.registerManager(Yi.URL_SCHEME,new class{constructor(){K(Le().getBool("IS_BROWSER"),()=>"Current environment is not a web browser"),K("undefined"==typeof window||void 0!==window.localStorage,()=>"Current browser does not appear to support localStorage"),this.LS=window.localStorage}async listModels(){let e={},t=Vi+Wi,n=Wi+Ui;for(let r=0;r<this.LS.length;++r){let a=this.LS.key(r);a.startsWith(t)&&a.endsWith(n)&&(e[Zi(a)]=JSON.parse(this.LS.getItem(a)))}return e}async removeModel(e){let t=Ki(e=function(e){return e.startsWith(Yi.URL_SCHEME)?e.slice(Yi.URL_SCHEME.length):e}(e));if(null==this.LS.getItem(t.info))throw new Error(`Cannot find model at path '${e}'`);let n=JSON.parse(this.LS.getItem(t.info));return Xi(t),n}})}catch(ss){}try{eo.registerManager(zi.URL_SCHEME,new class{constructor(){this.indexedDB=Li()}async listModels(){return new Promise((e,t)=>{let n=this.indexedDB.open(Di,1);n.onupgradeneeded=()=>Pi(n),n.onsuccess=()=>{let r=n.result,a=r.transaction(Oi,"readonly"),s=a.objectStore(Oi).getAll();s.onsuccess=()=>{let t={};for(let e of s.result)t[e.modelPath]=e.modelArtifactsInfo;e(t)},s.onerror=e=>(r.close(),t(s.error)),a.oncomplete=()=>r.close()},n.onerror=e=>t(n.error)})}async removeModel(e){return e=function(e){return e.startsWith(zi.URL_SCHEME)?e.slice(zi.URL_SCHEME.length):e}(e),new Promise((t,n)=>{let r=this.indexedDB.open(Di,1);r.onupgradeneeded=()=>Pi(r),r.onsuccess=()=>{let a,s=r.result,i=s.transaction(Oi,"readwrite"),o=i.objectStore(Oi),l=o.get(e);l.onsuccess=()=>{if(null==l.result)return s.close(),n(new Error(`Cannot find model with path '${e}' in IndexedDB.`));{let r=o.delete(e),i=()=>{a=s.transaction(Mi,"readwrite");let r=a.objectStore(Mi).delete(e);r.onsuccess=()=>t(l.result.modelArtifactsInfo),r.onerror=e=>n(l.error)};r.onsuccess=i,r.onerror=e=>(i(),s.close(),n(l.error))}},l.onerror=e=>(s.close(),n(l.error)),i.oncomplete=()=>{null==a?s.close():a.oncomplete=()=>s.close()}},r.onerror=e=>n(r.error)})}})}catch(ss){}}var oo,lo=()=>y();function uo(e,t="float32",n){return t=t||"float32",Ae(e),new Ja(e,t,n)}Le().get("IS_NODE")&&!Le().get("IS_BROWSER")&&Le().setPlatform("node",new class{constructor(){this.util=b(),this.textEncoder=new this.util.TextEncoder}fetch(e,t){return null!=Le().global.fetch?Le().global.fetch(e,t):(null==oo&&(oo=lo()),oo(e,t))}now(){let e=process.hrtime();return 1e3*e[0]+e[1]/1e6}encode(e,t){if("utf-8"!==t&&"utf8"!==t)throw new Error(`Node built-in encoder only supports utf-8, but got ${t}`);return this.textEncoder.encode(e)}decode(e,t){return 0===e.length?"":new this.util.TextDecoder(t).decode(e)}isTypedArray(e){return this.util.types.isFloat32Array(e)||this.util.types.isInt32Array(e)||this.util.types.isUint8Array(e)||this.util.types.isUint8ClampedArray(e)}});var ho=zs({cast_:function(e,t){let n=Os(e,"x","cast");if(!ce(t))throw new Error(`Failed to cast to unknown dtype ${t}`);if("string"===t&&"string"!==n.dtype||"string"!==t&&"string"===n.dtype)throw new Error("Only strings can be casted to strings");let r={x:n},a={dtype:t};return _s.runKernel(pt,r,a)}});var po=zs({clone_:function(e){let t={x:Os(e,"x","clone","string_or_numeric")};return _s.runKernel(an,t)}});function co(e,t=!1){console.log(e.toString(t))}Ss(),function(e){es=e}({buffer:uo,cast:ho,clone:po,print:co});var fo=zs({add_:function(e,t){let n=Os(e,"a","add"),r=Os(t,"b","add");[n,r]=gs(n,r);let a={a:n,b:r};return _s.runKernel(He,a)}});var mo=zs({floorDiv_:function(e,t){let n=Os(e,"a","floorDiv"),r=Os(t,"b","floorDiv");[n,r]=gs(n,r);let a={a:n,b:r};return _s.runKernel(Jt,a)}});var go=zs({div_:function(e,t){let n=Os(e,"a","div"),r=Os(t,"b","div");if([n,r]=gs(n,r),"int32"===n.dtype&&"int32"===r.dtype)return mo(n,r);let a={a:n,b:r};return _s.runKernel(zt,a,{})}});var yo=zs({mul_:function(e,t){let n=Os(e,"a","mul"),r=Os(t,"b","mul");[n,r]=gs(n,r);let a={a:n,b:r};return _s.runKernel(zn,a)}});var bo=zs({abs_:function(e){let t=Os(e,"x","abs");if("complex64"===t.dtype){let e={x:t};return _s.runKernel(gt,e)}{let e={x:t};return _s.runKernel(Ve,e)}}});var xo=zs({acos_:function(e){let t={x:Os(e,"x","acos")};return _s.runKernel(Ue,t)}});var vo=zs({acosh_:function(e){let t={x:Os(e,"x","acosh")};return _s.runKernel(Ge,t)}});var wo=zs({addN_:function(e){K(Array.isArray(e),()=>"The argument passed to tf.addN() must be a list of tensors"),K(e.length>=1,()=>`Must pass at least one tensor to tf.addN(), but got ${e.length}`);let t=e.map((e,t)=>Os(e,`tensors${t}`,"addN")),n=t[0];t.forEach(e=>{if(e.dtype!==n.dtype)throw new Error("All tensors passed to tf.addN() must have the same dtype")}),t.forEach(e=>{if(!ee(e.shape,n.shape))throw new Error("All tensors passed to tf.addN() must have the same shape")});let r=t;return _s.runKernel(je,r)}});var ko=zs({all_:function(e,t=null,n=!1){let r={x:Os(e,"x","all","bool")},a={axis:t,keepDims:n};return _s.runKernel(qe,r,a)}});var Io=zs({any_:function(e,t=null,n=!1){let r={x:Os(e,"x","any","bool")},a={axis:t,keepDims:n};return _s.runKernel(Ke,r,a)}});var So=zs({argMax_:function(e,t=0){let n={x:Os(e,"x","argMax")},r={axis:t};return _s.runKernel(Xe,n,r)}});var _o=zs({argMin_:function(e,t=0){let n={x:Os(e,"x","argMin")},r={axis:t};return _s.runKernel(Ze,n,r)}});var No=zs({asin_:function(e){let t={x:Os(e,"x","asin")};return _s.runKernel(Ye,t)}});var To=zs({asinh_:function(e){let t={x:Os(e,"x","asinh")};return _s.runKernel(Je,t)}});var Co=zs({atan_:function(e){let t={x:Os(e,"x","atan")};return _s.runKernel(Qe,t)}});var Eo=zs({atan2_:function(e,t){let n=Os(e,"a","atan2"),r=Os(t,"b","atan2");[n,r]=gs(n,r);let a={a:n,b:r};return _s.runKernel(tt,a)}});var Ao=zs({atanh_:function(e){let t={x:Os(e,"x","atanh")};return _s.runKernel(et,t)}});function $o(e,t,n,r,a="NHWC",s){return Do(e,[...t,e[3]],n,s,r,null,null,Go(a))}function Ro(e,t,n,r,a,s,i="channelsLast"){let o,[l,u]=Lo(t);if("channelsLast"===i)o=[l,u,e[3],e[3]];else{if("channelsFirst"!==i)throw new Error(`Unknown dataFormat ${i}`);o=[l,u,e[1],e[1]]}return Do(e,o,n,r,a,s,!1,i)}function Fo(e,t,n,r,a,s,i="NDHWC"){let o,l,[u,h,d]=Po(t);if("NDHWC"===i)l="channelsLast",o=[u,h,d,e[4],e[4]];else{if("NCDHW"!==i)throw new Error(`Unknown dataFormat ${i}`);l="channelsFirst",o=[u,h,d,e[1],e[1]]}return Mo(e,o,n,r,a,!1,l,s)}function Do(e,t,n,r,a,s,i=!1,o="channelsLast"){let[l,u,h,d]=[-1,-1,-1,-1];if("channelsLast"===o)[l,u,h,d]=e;else{if("channelsFirst"!==o)throw new Error(`Unknown dataFormat ${o}`);[l,d,u,h]=e}let p,[c,f,,m]=t,[g,y]=Lo(n),[b,x]=Lo(r),v=zo(c,b),w=zo(f,x),{padInfo:k,outHeight:I,outWidth:S}=function(e,t,n,r,a,s,i,o,l){let u,h,d;if("number"==typeof e){u={top:e,bottom:e,left:e,right:e,type:0===e?"VALID":"NUMBER"};let a=function(e,t,n,r,a){null==r&&(r=Oo(e,t,n));let s=e[0],i=e[1],o=Bo((s-t+2*r)/n+1,a),l=Bo((i-t+2*r)/n+1,a);return[o,l]}([t,n],s,r,e,o);h=a[0],d=a[1]}else if("same"===e){h=Math.ceil(t/r),d=Math.ceil(n/a);let e=Math.max(0,(h-1)*r+s-t),o=Math.max(0,(d-1)*a+i-n),l=Math.floor(e/2),p=e-l,c=Math.floor(o/2);u={top:l,bottom:p,left:c,right:o-c,type:"SAME"}}else if("valid"===e)u={top:0,bottom:0,left:0,right:0,type:"VALID"},h=Math.ceil((t-s+1)/r),d=Math.ceil((n-i+1)/a);else{if("object"!=typeof e)throw Error(`Unknown padding parameter: ${e}`);{let p="channelsLast"===l?e[1][0]:e[2][0],c="channelsLast"===l?e[1][1]:e[2][1],f="channelsLast"===l?e[2][0]:e[3][0],m="channelsLast"===l?e[2][1]:e[3][1];u={top:p,bottom:c,left:f,right:m,type:0===p&&0===c&&0===f&&0===m?"VALID":"EXPLICIT"},h=Bo((t-s+p+c)/r+1,o),d=Bo((n-i+f+m)/a+1,o)}}return{padInfo:u,outHeight:h,outWidth:d}}(a,u,h,g,y,v,w,s,o),_=i?m*d:m;return"channelsFirst"===o?p=[l,_,I,S]:"channelsLast"===o&&(p=[l,I,S,_]),{batchSize:l,dataFormat:o,inHeight:u,inWidth:h,inChannels:d,outHeight:I,outWidth:S,outChannels:_,padInfo:k,strideHeight:g,strideWidth:y,filterHeight:c,filterWidth:f,effectiveFilterHeight:v,effectiveFilterWidth:w,dilationHeight:b,dilationWidth:x,inShape:e,outShape:p,filterShape:t}}function Mo(e,t,n,r,a,s=!1,i="channelsLast",o){let[l,u,h,d,p]=[-1,-1,-1,-1,-1];if("channelsLast"===i)[l,u,h,d,p]=e;else{if("channelsFirst"!==i)throw new Error(`Unknown dataFormat ${i}`);[l,p,u,h,d]=e}let c,[f,m,g,,y]=t,[b,x,v]=Po(n),[w,k,I]=Po(r),S=zo(f,w),_=zo(m,k),N=zo(g,I),{padInfo:T,outDepth:C,outHeight:E,outWidth:A}=function(e,t,n,r,a,s,i,o,l,u,h){let d,p,c,f;if("valid"===e&&(e=0),"number"==typeof e){d={top:e,bottom:e,left:e,right:e,front:e,back:e,type:0===e?"VALID":"NUMBER"};let m=function(e,t,n,r,a,s){null==a&&(a=Oo(e,t[0],r[0]));let i=[0,0,0,n];for(let n=0;n<3;n++)e[n]+2*a>=t[n]&&(i[n]=Bo((e[n]-t[n]+2*a)/r[n]+1,s));return i}([t,n,r,1],[o,l,u],1,[a,s,i],e,h);p=m[0],c=m[1],f=m[2]}else{if("same"!==e)throw Error(`Unknown padding parameter: ${e}`);{p=Math.ceil(t/a),c=Math.ceil(n/s),f=Math.ceil(r/i);let e=(p-1)*a+o-t,h=(c-1)*s+l-n,m=(f-1)*i+u-r,g=Math.floor(e/2),y=e-g,b=Math.floor(h/2),x=h-b,v=Math.floor(m/2);d={top:b,bottom:x,left:v,right:m-v,front:g,back:y,type:"SAME"}}}return{padInfo:d,outDepth:p,outHeight:c,outWidth:f}}(a,u,h,d,b,x,v,S,_,N,o),$=s?y*p:y;return"channelsFirst"===i?c=[l,$,C,E,A]:"channelsLast"===i&&(c=[l,C,E,A,$]),{batchSize:l,dataFormat:i,inDepth:u,inHeight:h,inWidth:d,inChannels:p,outDepth:C,outHeight:E,outWidth:A,outChannels:$,padInfo:T,strideDepth:b,strideHeight:x,strideWidth:v,filterDepth:f,filterHeight:m,filterWidth:g,effectiveFilterDepth:S,effectiveFilterHeight:_,effectiveFilterWidth:N,dilationDepth:w,dilationHeight:k,dilationWidth:I,inShape:e,outShape:c,filterShape:t}}function Oo(e,t,n,r=1){let a=zo(t,r);return Math.floor((e[0]*(n-1)-n+a)/2)}function Lo(e){return"number"==typeof e?[e,e,e]:2===e.length?[e[0],e[1],1]:e}function Po(e){return"number"==typeof e?[e,e,e]:e}function zo(e,t){return t<=1?e:e+(e-1)*(t-1)}function Bo(e,t){if(!t)return Math.trunc(e);switch(t){case"round":return Math.round(e);case"ceil":return Math.ceil(e);case"floor":return Math.floor(e);default:throw new Error(`Unknown roundingMode ${t}`)}}function Wo(e){let[t,n,r]=Lo(e);return 1===t&&1===n&&1===r}function Vo(e,t){return Wo(e)||Wo(t)}function Uo(e){return Lo(e).every(e=>e>0)}function Go(e){if("NHWC"===e)return"channelsLast";if("NCHW"===e)return"channelsFirst";throw new Error(`Unknown dataFormat ${e}`)}function Ho(e,t,n){if(null!=n){if("string"==typeof t)throw Error(`Error in ${e}: pad must be an integer when using dimRoundingMode ${n} but got pad ${t}.`);if("number"==typeof t)K(te(t),()=>`Error in ${e}: pad must be an integer when using dimRoundingMode ${n} but got pad ${t}.`);else{if("object"!=typeof t)throw Error(`Error in ${e}: Unknown padding parameter: ${t}`);t.forEach(t=>{t.forEach(t=>{K(te(t),()=>`Error in ${e}: pad must be an integer when using dimRoundingMode ${n} but got pad ${t}.`)})})}}}var jo=zs({reshape_:function(e,t){let n={x:Os(e,"x","reshape","string_or_numeric")},r={shape:t};return _s.runKernel(ir,n,r)}});var qo=zs({avgPool_:function(e,t,n,r,a){let s=Os(e,"x","avgPool","float32");K(Vo(n,1),()=>`Error in avgPool: Either strides or dilations must be 1. Got strides ${n} and dilations '1'`);let i=s,o=!1;3===s.rank&&(o=!0,i=jo(s,[1,s.shape[0],s.shape[1],s.shape[2]])),K(4===i.rank,()=>`Error in avgPool: x must be rank 4 but got rank ${i.rank}.`),Ho("avgPool",r,a);let l={x:i},u={filterSize:t,strides:n,pad:r,dimRoundingMode:a},h=_s.runKernel(nt,l,u);return h=ho(h,s.dtype),o?jo(h,[h.shape[1],h.shape[2],h.shape[3]]):h}});var Ko=zs({avgPool3d_:function(e,t,n,r,a,s="NDHWC"){let i=Os(e,"x","avgPool3d","float32"),o=i,l=!1;4===i.rank&&(l=!0,o=jo(i,[1,i.shape[0],i.shape[1],i.shape[2],i.shape[3]])),K(5===o.rank,()=>`Error in avgPool3d: x must be rank 5 but got rank ${o.rank}.`),K("NDHWC"===s,()=>`Error in avgPool3d: Only NDHWC is currently supported, but got dataFormat of ${s}`),K("number"==typeof n&&n>0||Array.isArray(n)&&n[0]>0&&n[1]>0&&n[2]>0,()=>`Error in avgPool3d: Stride must be > 0, but got '${n}'`),Ho("avgPool3d",r,a);let u={x:o},h={filterSize:t,strides:n,pad:r,dimRoundingMode:a,dataFormat:s},d=_s.runKernel(at,u,h);return d=ho(d,o.dtype),l?jo(d,[d.shape[1],d.shape[2],d.shape[3],d.shape[4]]):d}});var Xo=zs({concat_:function(e,t=0){K(e.length>=1,()=>"Pass at least one tensor to concat");let n=Ls(e,"tensors","concat","string_or_numeric");if("complex64"===n[0].dtype&&n.forEach(e=>{if("complex64"!==e.dtype)throw new Error(`Cannot concatenate complex64 tensors with a tensor\n          with dtype ${e.dtype}. `)}),1===n.length)return po(n[0]);let r=n,a={axis:t};return _s.runKernel(yt,r,a)}});var Zo=zs({matMul_:function(e,t,n=!1,r=!1){let a=Os(e,"a","matMul"),s=Os(t,"b","matMul");[a,s]=gs(a,s);let i={a,b:s},o={transposeA:n,transposeB:r};return _s.runKernel(it,i,o)}});var Yo=zs({sigmoid_:function(e){let t={x:Os(e,"x","sigmoid","float32")};return _s.runKernel(Sr,t)}});var Jo=zs({slice_:function(e,t,n){let r=Os(e,"x","slice","string_or_numeric");if(0===r.rank)throw new Error("Slicing scalar is not possible");let a={x:r},s={begin:t,size:n};return _s.runKernel(vr,a,s)}});var Qo=zs({tanh_:function(e){let t={x:Os(e,"x","tanh","float32")};return _s.runKernel(Hr,t)}});var el=zs({basicLSTMCell_:function(e,t,n,r,a,s){let i=Os(e,"forgetBias","basicLSTMCell"),o=Os(t,"lstmKernel","basicLSTMCell"),l=Os(n,"lstmBias","basicLSTMCell"),u=Os(r,"data","basicLSTMCell"),h=Os(a,"c","basicLSTMCell"),d=Os(s,"h","basicLSTMCell"),p=Xo([u,d],1),c=Zo(p,o),f=fo(c,l),m=f.shape[0],g=f.shape[1]/4,y=[m,g],b=Jo(f,[0,0],y),x=Jo(f,[0,g],y),v=Jo(f,[0,2*g],y),w=Jo(f,[0,3*g],y),k=fo(yo(Yo(b),Qo(x)),yo(h,Yo(fo(i,v))));return[k,yo(Qo(k),Yo(w))]}});var tl=zs({batchToSpaceND_:function(e,t,n){let r=Os(e,"x","batchToSpaceND"),a=t.reduce((e,t)=>e*t);K(r.rank>=1+t.length,()=>`input rank is ${r.rank} but should be > than blockShape.length ${t.length}`),K(n.length===t.length,()=>`crops.length is ${n.length} but should be equal to blockShape.length  ${t.length}`),K(r.shape[0]%a===0,()=>`input tensor batch is ${r.shape[0]} but is not divisible by the product of the elements of blockShape ${t.join(" * ")} === ${a}`);let s={x:r},i={blockShape:t,crops:n};return _s.runKernel(ot,s,i)}});function nl(e){let t;return t=0===e.rank||1===e.rank?jo(e,[1,1,1,e.size]):2===e.rank?jo(e,[1,1,e.shape[0],e.shape[1]]):3===e.rank?jo(e,[1,e.shape[0],e.shape[1],e.shape[2]]):e,t}var rl=zs({batchNorm_:function(e,t,n,r,a,s){null==s&&(s=.001);let i,o,l=Os(e,"x","batchNorm"),u=Os(t,"mean","batchNorm"),h=Os(n,"variance","batchNorm");null!=a&&(i=Os(a,"scale","batchNorm")),null!=r&&(o=Os(r,"offset","batchNorm")),K(u.rank===h.rank,()=>"Batch normalization gradient requires mean and variance to have equal ranks."),K(null==o||u.rank===o.rank,()=>"Batch normalization gradient requires mean and offset to have equal ranks."),K(null==i||u.rank===i.rank,()=>"Batch normalization gradient requires mean and scale to have equal ranks.");let d={x:nl(l),scale:i,offset:o,mean:u,variance:h},p={varianceEpsilon:s},c=_s.runKernel(Qt,d,p);return jo(c,l.shape)}});var al=zs({batchNorm2d_:function(e,t,n,r,a,s){let i,o,l=Os(e,"x","batchNorm"),u=Os(t,"mean","batchNorm"),h=Os(n,"variance","batchNorm");return null!=a&&(i=Os(a,"scale","batchNorm")),null!=r&&(o=Os(r,"offset","batchNorm")),K(2===l.rank,()=>`Error in batchNorm2D: x must be rank 2 but got rank ${l.rank}.`),K(2===u.rank||1===u.rank,()=>`Error in batchNorm2D: mean must be rank 2 or rank 1 but got rank ${u.rank}.`),K(2===h.rank||1===h.rank,()=>`Error in batchNorm2D: variance must be rank 2 or rank 1 but got rank ${h.rank}.`),null!=i&&K(2===i.rank||1===i.rank,()=>`Error in batchNorm2D: scale must be rank 2 or rank 1 but got rank ${i.rank}.`),null!=o&&K(2===o.rank||1===o.rank,()=>`Error in batchNorm2D: offset must be rank 2 or rank 1 but got rank ${o.rank}.`),rl(l,u,h,o,i,s)}});var sl=zs({batchNorm3d_:function(e,t,n,r,a,s){let i,o,l=Os(e,"x","batchNorm"),u=Os(t,"mean","batchNorm"),h=Os(n,"variance","batchNorm");return null!=a&&(i=Os(a,"scale","batchNorm")),null!=r&&(o=Os(r,"offset","batchNorm")),K(3===l.rank,()=>`Error in batchNorm3D: x must be rank 3 but got rank ${l.rank}.`),K(3===u.rank||1===u.rank,()=>`Error in batchNorm3D: mean must be rank 3 or rank 1 but got rank ${u.rank}.`),K(3===h.rank||1===h.rank,()=>`Error in batchNorm3D: variance must be rank 3 or rank 1 but got rank ${h.rank}.`),null!=i&&K(3===i.rank||1===i.rank,()=>`Error in batchNorm3D: scale must be rank 3 or rank 1 but got rank ${i.rank}.`),null!=o&&K(3===o.rank||1===o.rank,()=>`Error in batchNorm3D: offset must be rank 3 or rank 1 but got rank ${o.rank}.`),rl(l,u,h,o,i,s)}});var il=zs({batchNorm4d_:function(e,t,n,r,a,s){let i,o,l=Os(e,"x","batchNorm"),u=Os(t,"mean","batchNorm"),h=Os(n,"variance","batchNorm");return null!=a&&(i=Os(a,"scale","batchNorm")),null!=r&&(o=Os(r,"offset","batchNorm")),K(4===l.rank,()=>`Error in batchNorm4D: x must be rank 4 but got rank ${l.rank}.`),K(4===u.rank||1===u.rank,()=>`Error in batchNorm4D: mean must be rank 4 or rank 1 but got rank ${u.rank}.`),K(4===h.rank||1===h.rank,()=>`Error in batchNorm4D: variance must be rank 4 or rank 1 but got rank ${h.rank}.`),null!=i&&K(4===i.rank||1===i.rank,()=>`Error in batchNorm4D: scale must be rank 4 or rank 1 but got rank ${i.rank}.`),null!=o&&K(4===o.rank||1===o.rank,()=>`Error in batchNorm4D: offset must be rank 4 or rank 1 but got rank ${o.rank}.`),rl(l,u,h,o,i,s)}});var ol=zs({bincount_:function(e,t,n){let r=Os(e,"x","bincount"),a=Os(t,"weights","bincount");K("int32"===r.dtype,()=>`Error in bincount: input dtype must be int32, but got ${r.dtype}`),K(n>=0,()=>`size must be non-negative, but got ${n}.`),K(a.size===r.size||0===a.size,()=>`Error in bincount: weights must have the same size as input or0-length, but got input shape: ${r.shape}, weights shape: ${a.shape}.`);let s={x:r,weights:a},i={size:n};return _s.runKernel(lt,s,i)}});var ll=zs({bitwiseAnd_:function(e,t){let n=Os(e,"x","bitwiseAnd"),r=Os(t,"y","bitwiseAnd");if(!ee(n.shape,r.shape))throw new Error(`BitwiseAnd: Tensors must have the same shape. x: ${n.shape}, y: ${r.shape}`);if("int32"!==n.dtype||"int32"!==r.dtype)throw new Error(`BitwiseAnd: Only supports 'int32' values in tensor, found type of x: ${n.dtype} and type of y: ${r.dtype}`);let a={a:n,b:r};return _s.runKernel(ut,a)}});var ul=zs({broadcastArgs_:function(e,t){let n=Os(e,"s0","broadcastArgs","int32"),r=Os(t,"s1","broadcastArgs","int32");if(1!==n.rank)throw new Error(`broadcastArgs(): first input must be a vector (rank=1). Has rank ${n.rank}`);if(1!==r.rank)throw new Error(`broadcastArgs(): second input must be a vector (rank=1). Has rank ${r.rank}`);let a={s0:n,s1:r};return _s.runKernel(dt,a)}});var hl=zs({broadcastTo_:function(e,t){let n=Os(e,"broadcastTo","x"),r=n.shape;if(Ae(t),t.length<n.rank)throw new Error(`broadcastTo(): shape.length=${t.length} < input.rank=${n.rank}.`);if(t.length>n.rank){let e=n.shape.slice();for(;e.length<t.length;)e.unshift(1);n=jo(n,e)}let a=n.shape,s=Array.from(t);for(let e=t.length-1;e>=0;e--)if(a[e]===t[e])s[e]=1;else if(1!==n.shape[e])throw new Error(`broadcastTo(): [${r}] cannot be broadcast to [${t}].`);if(0===s.map((e,t)=>e>1?t:-1).filter(e=>e>=0).length)return po(n);let i={x:n},o={reps:s};return _s.runKernel(jr,i,o)}});var dl=zs({ceil_:function(e){let t={x:Os(e,"x","ceil","float32")};return _s.runKernel(ct,t)}});function pl(e,t,n){Ae(e);let r={shape:e,value:t,dtype:n=n||ve(t)};return _s.runKernel(Xt,{},r)}var cl=zs({clipByValue_:function(e,t,n){let r=Os(e,"x","clipByValue");if(K(t<=n,()=>`Error in clip: min (${t}) must be less than or equal to max (${n}).`),t===n)return pl(r.shape,t,r.dtype);let a={x:r},s={clipValueMin:t,clipValueMax:n};return _s.runKernel(ft,a,s)}});var fl=zs({concat1d_:function(e){return Xo(e,0)}});var ml=zs({concat2d_:function(e,t){return Xo(e,t)}});var gl=zs({concat3d_:function(e,t){return Xo(e,t)}});var yl=zs({concat4d_:function(e,t){return Xo(e,t)}});var bl=zs({conv2d_:function(e,t,n,r,a="NHWC",s=[1,1],i){let o=Os(e,"x","conv2d","float32"),l=Os(t,"filter","conv2d","float32"),u=o,h=!1;3===o.rank&&(h=!0,u=jo(o,[1,o.shape[0],o.shape[1],o.shape[2]])),K(4===u.rank,()=>`Error in conv2d: input must be rank 4, but got rank ${u.rank}.`),K(4===l.rank,()=>`Error in conv2d: filter must be rank 4, but got rank ${l.rank}.`),Ho("conv2d",r,i);let d="NHWC"===a?u.shape[3]:u.shape[1];K(d===l.shape[2],()=>`Error in conv2d: depth of input (${d}) must match input depth for filter ${l.shape[2]}.`),K(Vo(n,s),()=>`Error in conv2D: Either strides or dilations must be 1. Got strides ${n} and dilations '${s}'`),K(Uo(s),()=>"Error in conv2D: Dilated rates should be larger than 0."),K(Uo(n),()=>"Error in conv2D: Strides should be larger than 0.");let p={x:u,filter:l},c={strides:n,pad:r,dataFormat:a,dilations:s,dimRoundingMode:i},f=_s.runKernel(bt,p,c);return h?jo(f,[f.shape[1],f.shape[2],f.shape[3]]):f}});var xl=zs({conv1d_:function(e,t,n,r,a="NWC",s=1,i){let o=Os(e,"x","conv1d"),l=Os(t,"filter","conv1d"),u=o,h=!1;2===o.rank&&(h=!0,u=jo(o,[1,o.shape[0],o.shape[1]])),K(3===u.rank,()=>`Error in conv1d: input must be rank 3, but got rank ${u.rank}.`),K(3===l.rank,()=>`Error in conv1d: filter must be rank 3, but got rank ${l.rank}.`),Ho("conv1d",r,i),K(u.shape[2]===l.shape[1],()=>`Error in conv1d: depth of input (${u.shape[2]}) must match input depth for filter ${l.shape[1]}.`),K(Vo(n,s),()=>`Error in conv1D: Either stride or dilation must be 1. Got stride ${n} and dilation '${s}'`),K(Uo(s),()=>"Error in conv1D: Dilated rates should be larger than 0."),K(Uo(n),()=>"Error in conv1D: Stride should be larger than 0."),K("NWC"===a,()=>`Error in conv1d: got dataFormat of ${a} but only NWC is currently supported.`);let d=jo(l,[1,l.shape[0],l.shape[1],l.shape[2]]),p=jo(u,[u.shape[0],1,u.shape[1],u.shape[2]]),c=bl(p,d,[1,n],r,"NHWC",[1,s],i);return jo(c,h?[c.shape[2],c.shape[3]]:[c.shape[0],c.shape[2],c.shape[3]])}});var vl=zs({conv2DBackpropInput_:function(e,t,n,r,a,s="NHWC",i){K(e.length===t.rank,()=>`Length of inShape (${e.length}) and rank of dy (${t.rank}) must match`);let o=e,l=t,u=!1;3===t.rank&&(u=!0,l=jo(t,[1,t.shape[0],t.shape[1],t.shape[2]]),o=[1,e[0],e[1],e[2]]),K(4===o.length,()=>`Error in conv2dDerInput: inShape must be length 4, but got length ${o.length}.`),K(4===l.rank,()=>`Error in conv2dDerInput: dy must be rank 4, but got rank ${l.rank}`),K(4===n.rank,()=>`Error in conv2dDerInput: filter must be rank 4, but got rank ${n.rank}`);let h="NHWC"===s?o[3]:o[1],d="NHWC"===s?l.shape[3]:l.shape[1];K(h===n.shape[2],()=>`Error in conv2dDerInput: depth of input (${h}) must match input depth for filter ${n.shape[2]}.`),K(d===n.shape[3],()=>`Error in conv2dDerInput: depth of output (${d}) must match output depth for filter ${n.shape[3]}.`),Ho("conv2dDerInput",a,i);let p={dy:l,filter:n},c={strides:r,pad:a,dataFormat:s,dimRoundingMode:i,inputShape:o},f=_s.runKernel(vt,p,c);return u?jo(f,[f.shape[1],f.shape[2],f.shape[3]]):f}});var wl=zs({conv2dTranspose_:function(e,t,n,r,a,s){let i=Os(e,"x","conv2dTranspose"),o=Os(t,"filter","conv2dTranspose");return vl(n,i,o,r,a,"NHWC",s)}});var kl=zs({conv3d_:function(e,t,n,r,a="NDHWC",s=[1,1,1]){let i=Os(e,"x","conv3d"),o=Os(t,"filter","conv3d"),l=i,u=!1;4===i.rank&&(u=!0,l=jo(i,[1,i.shape[0],i.shape[1],i.shape[2],i.shape[3]])),K(5===l.rank,()=>`Error in conv3d: input must be rank 5, but got rank ${l.rank}.`),K(5===o.rank,()=>`Error in conv3d: filter must be rank 5, but got rank ${o.rank}.`),K(l.shape[4]===o.shape[3],()=>`Error in conv3d: depth of input (${l.shape[4]}) must match input depth for filter ${o.shape[3]}.`),K(Vo(n,s),()=>`Error in conv3D: Either strides or dilations must be 1. Got strides ${n} and dilations '${s}'`),K("NDHWC"===a,()=>`Error in conv3d: got dataFormat of ${a} but only NDHWC is currently supported.`),K(Uo(s),()=>"Error in conv3D: Dilated rates should be larger than 0."),K(Uo(n),()=>"Error in conv3D: Strides should be larger than 0.");let h={x:l,filter:o},d={strides:n,pad:r,dataFormat:a,dilations:s},p=_s.runKernel(wt,h,d);return u?jo(p,[p.shape[1],p.shape[2],p.shape[3],p.shape[4]]):p}});var Il=zs({conv3DBackpropInput_:function(e,t,n,r,a){K(e.length===t.rank,()=>`Length of inShape (${e.length}) and rank of dy (${t.rank}) must match`);let s=e,i=t,o=!1;4===t.rank&&(o=!0,i=jo(t,[1,t.shape[0],t.shape[1],t.shape[2],t.shape[3]]),s=[1,e[0],e[1],e[2],e[3]]);let l=s[4],u=i.shape[4];K(5===s.length,()=>`Error in conv3dDerInput: inShape must be length 5, but got length ${s.length}.`),K(5===i.rank,()=>`Error in conv3dDerInput: dy must be rank 5, but got rank ${i.rank}`),K(5===n.rank,()=>`Error in conv3dDerInput: filter must be rank 5, but got rank ${n.rank}`),K(l===n.shape[3],()=>`Error in conv3dDerInput: depth of input (${l}) must match input depth for filter ${n.shape[3]}.`),K(u===n.shape[4],()=>`Error in conv3dDerInput: depth of output (${u}) must match output depth for filter ${n.shape[4]}.`);let h={dy:i,filter:n},d={pad:a,strides:r,inputShape:s},p=_s.runKernel(It,h,d);return o?jo(p,[p.shape[1],p.shape[2],p.shape[3],p.shape[4]]):p}});var Sl=zs({conv3dTranspose_:function(e,t,n,r,a){let s=Os(e,"x","conv3dTranspose"),i=Os(t,"filter","conv3dTranspose");return Il(n,s,i,r,a)}});var _l=zs({cos_:function(e){let t={x:Os(e,"x","cos","float32")};return _s.runKernel(St,t)}});var Nl=zs({cosh_:function(e){let t={x:Os(e,"x","cosh","float32")};return _s.runKernel(_t,t)}});var Tl=zs({cumprod_:function(e,t=0,n=!1,r=!1){let a={x:Os(e,"x","cumprod")},s={axis:t,exclusive:n,reverse:r};return _s.runKernel(Nt,a,s)}});var Cl=zs({cumsum_:function(e,t=0,n=!1,r=!1){let a={x:Os(e,"x","cumsum")},s={axis:t,exclusive:n,reverse:r};return _s.runKernel(Tt,a,s)}});var El=zs({denseBincount_:function(e,t,n,r=!1){let a=Os(e,"x","denseBincount"),s=Os(t,"weights","denseBincount");K("int32"===a.dtype,()=>`Error in denseBincount: input dtype must be int32, but got ${a.dtype}`),K(a.rank<=2,()=>`Error in denseBincount: input must be at most rank 2, but got rank ${a.rank}.`),K(n>=0,()=>`size must be non-negative, but got ${n}.`),K(s.size===a.size||0===s.size,()=>`Error in denseBincount: weights must have the same shape as x or 0-length, but got x shape: ${a.shape}, weights shape: ${s.shape}.`);let i={x:a,weights:s},o={size:n,binaryOutput:r};return _s.runKernel(Et,i,o)}});var Al=zs({depthToSpace_:function(e,t,n="NHWC"){let r=Os(e,"x","depthToSpace","float32"),a="NHWC"===n?r.shape[1]:r.shape[2],s="NHWC"===n?r.shape[2]:r.shape[3],i="NHWC"===n?r.shape[3]:r.shape[1];K(t>1,()=>`blockSize should be > 1 for depthToSpace, but was: ${t}`),K(a*t>=0,()=>`Negative dimension size caused by overflow when multiplying\n    ${a} and ${t}  for depthToSpace with input shape\n    ${r.shape}`),K(s*t>=0,()=>`Negative dimension size caused by overflow when multiplying\n    ${s} and ${t} for depthToSpace with input shape\n        ${r.shape}`),K(i%(t*t)===0,()=>`Dimension size must be evenly divisible by ${t*t} but is ${i} for depthToSpace with input shape ${r.shape}`);let o={x:r},l={blockSize:t,dataFormat:n};return _s.runKernel(At,o,l)}});var $l=zs({depthwiseConv2d_:function(e,t,n,r,a="NHWC",s=[1,1],i){let o=Os(e,"x","depthwiseConv2d","float32"),l=Os(t,"filter","depthwiseConv2d","float32"),u=o,h=!1;3===o.rank&&(h=!0,u=jo(o,[1,o.shape[0],o.shape[1],o.shape[2]])),K(4===u.rank,()=>`Error in depthwiseConv2d: input must be rank 4, but got rank ${u.rank}.`),K(4===l.rank,()=>`Error in depthwiseConv2d: filter must be rank 4, but got rank ${l.rank}.`);let d="NHWC"===a?u.shape[3]:u.shape[1];K(d===l.shape[2],()=>`Error in depthwiseConv2d: number of input channels (${d}) must match the inChannels dimension in filter ${l.shape[2]}.`),Ho("depthwiseConv2d",r,i);let p={x:u,filter:l},c={strides:n,pad:r,dataFormat:a,dilations:s,dimRoundingMode:i},f=_s.runKernel($t,p,c);return h?jo(f,[f.shape[1],f.shape[2],f.shape[3]]):f}});var Rl=zs({diag_:function(e){let t={x:Os(e,"x","diag")};return _s.runKernel(Dt,t)}});var Fl=zs({dilation2d_:function(e,t,n,r,a=[1,1],s="NHWC"){let i=Os(e,"x","dilation2d"),o=Os(t,"filter","dilation2d");K(3===i.rank||4===i.rank,()=>`Error in dilation2d: input must be rank 3 or 4, but got rank ${i.rank}.`),K(3===o.rank,()=>`Error in dilation2d: filter must be rank 3, but got rank ${o.rank}.`),K("NHWC"===s,()=>`Error in dilation2d: Only NHWC is currently supported, but got dataFormat of ${s}`);let l=i,u=!1;3===i.rank&&(l=jo(i,[1,i.shape[0],i.shape[1],i.shape[2]]),u=!0),K(l.shape[3]===o.shape[2],()=>`Error in dilation2d:  input and filter must have the same depth: ${l.shape[3]} vs ${o.shape[2]}`);let h={x:l,filter:o},d={strides:n,pad:r,dilations:a},p=_s.runKernel(Mt,h,d);return u?jo(p,[p.shape[1],p.shape[2],p.shape[3]]):p}}),Dl={};function Ml(e,t){let n=e.length,r=[];for(let a=0;a<n;a++){let s=n-1-a,i=e[s]||1;(t[t.length-1-a]||1)>1&&1===i&&r.unshift(s)}return r}function Ol(e,t){let n=[];for(let r=0;r<t.length;r++){let a=e[e.length-r-1],s=t.length-r-1,i=t[s];(null==a||1===a&&i>1)&&n.unshift(s)}return n}function Ll(e,t){let n=Math.max(e.length,t.length),r=new Array(n);for(let a=0;a<n;a++){let s=e[e.length-a-1];null==s&&(s=1);let i=t[t.length-a-1];if(null==i&&(i=1),1===s)r[n-a-1]=i;else if(1===i)r[n-a-1]=s;else{if(s!==i){throw Error(`Operands could not be broadcast together with shapes ${e} and ${t}.`)}r[n-a-1]=s}}return r}f(Dl,{assertAndGetBroadcastShape:()=>Ll,getBroadcastDims:()=>Ml,getReductionAxes:()=>Ol});var Pl=zs({equal_:function(e,t){let n=Os(e,"a","equal","string_or_numeric"),r=Os(t,"b","equal","string_or_numeric");[n,r]=gs(n,r),Ll(n.shape,r.shape);let a={a:n,b:r};return _s.runKernel(Gt,a)}});var zl=zs({where_:function(e,t,n){let r=Os(t,"a","where"),a=Os(n,"b","where"),s=Os(e,"condition","where","bool"),i=Ll(Ll(s.shape,r.shape),a.shape),o={condition:hl(s,i),t:hl(r,i),e:hl(a,i)};return _s.runKernel(br,o)}});var Bl=zs({zerosLike_:function(e){let t={x:Os(e,"x","zerosLike")};return _s.runKernel(ea,t)}});var Wl=zs({divNoNan_:function(e,t){let n=Os(e,"a","div"),r=Os(t,"b","div");[n,r]=gs(n,r);let a=go(n,r),s=Bl(a),i=Pl(r,s);return zl(i,s,a)}});var Vl=zs({dot_:function(e,t){let n=Os(e,"t1","dot"),r=Os(t,"t2","dot");K(!(1!==n.rank&&2!==n.rank||1!==r.rank&&2!==r.rank),()=>`Error in dot: inputs must all be rank 1 or 2, but got ranks ${n.rank} and ${r.rank}.`);let a=1===n.rank?n.size:n.shape[1],s=1===r.rank?r.size:r.shape[0];if(K(a===s,()=>`Error in dot: inner dimensions of inputs must match, but got ${a} and ${s}.`),1===n.rank&&1===r.rank){let e=jo(n,[1,-1]),t=jo(r,[-1,1]),a=Zo(e,t);return jo(a,[])}if(1===n.rank&&2===r.rank){let e=jo(n,[1,-1]),t=jo(r,[r.shape[0],r.shape[1]]),a=Zo(e,t);return jo(a,[a.size])}if(2===n.rank&&1===r.rank){let e=jo(r,[-1,1]),t=Zo(n,e);return jo(t,[t.size])}{let e=jo(r,[r.shape[0],r.shape[1]]);return Zo(n,e)}}});var Ul=zs({einsum_:function(e,...t){let n=t.map((e,t)=>Os(e,`tensors${t}`,"einsum")),r={equation:e};return _s.runKernel(Bt,n,r)}});var Gl=zs({elu_:function(e){let t={x:Os(e,"x","elu","float32")};return _s.runKernel(Wt,t)}});var Hl=zs({ensureShape_:function(e,t){let n=Os(e,"x","ensureShape","string_or_numeric");if(!Q(n.shape,t))throw new Error(`EnsureShape: Shape of tensor ${n.shape} is not compatible with expected shape ${t}`);return e}});var jl=zs({erf_:function(e){let t=Os(e,"x","erf");K("int32"===t.dtype||"float32"===t.dtype,()=>"Input dtype must be `int32` or `float32`."),"int32"===t.dtype&&(t=ho(t,"float32"));let n={x:t};return _s.runKernel(Ut,n)}});function ql(e,t){for(let n=0;n<e.length;++n)if(e[e.length-n-1]!==t-1-n)return!1;return!0}function Kl(e,t,n){let r=e.length+t.length,a=[],s=0,i=0;for(let o=0;o<r;o++)-1===n.indexOf(o)?a.push(e[s++]):a.push(t[i++]);return a}function Xl(e,t){let n=[],r=e.length;for(let a=0;a<r;a++)-1===t.indexOf(a)&&n.push(e[a]);return[n,t.map(t=>e[t])]}function Zl(e,t){return Kl(e,t.map(e=>1),t)}function Yl(e,t,n){K(ql(t,n),()=>`${e} supports only inner-most axes for now. Got axes ${t} and rank-${n} input.`)}function Jl(e,t){if(ql(e,t))return null;let n=[];for(let r=0;r<t;++r)-1===e.indexOf(r)&&n.push(r);return e.forEach(e=>n.push(e)),n}function Ql(e){return e.map((e,t)=>[t,e]).sort((e,t)=>e[1]-t[1]).map(e=>e[0])}function eu(e,t){let n=[];for(let r=t-e;r<t;++r)n.push(r);return n}var tu=zs({max_:function(e,t=null,n=!1){let r={x:Os(e,"x","max")},a={reductionIndices:t,keepDims:n};return _s.runKernel(Nn,r,a)}});var nu=zs({min_:function(e,t=null,n=!1){let r={x:Os(e,"x","min")},a={axis:t,keepDims:n};return _s.runKernel(Dn,r,a)}});var ru=zs({pow_:function(e,t){let n=Os(e,"base","pow"),r=Os(t,"exp","pow");[n,r]=gs(n,r);let a={a:n,b:r};return _s.runKernel(Zn,a)}});function au(e,t){if((Va(e)&&"string"!==t||Array.isArray(e))&&"complex64"!==t)throw new Error("Error creating a new Scalar: value must be a primitive (number|boolean|string)");if("string"===t&&Va(e)&&!(e instanceof Uint8Array))throw new Error("When making a scalar from encoded string, the value must be `Uint8Array`.");return Ws(e,[],[],t)}var su=zs({sqrt_:function(e){let t={x:Os(e,"x","sqrt","float32")};return _s.runKernel(Nr,t)}});var iu=zs({square_:function(e){let t=Os(e,"x","square");return _s.runKernel("Square",{x:t},{})}});var ou=zs({sum_:function(e,t=null,n=!1){let r=Os(e,"x","sum");"bool"===r.dtype&&(r=ho(r,"int32"));let a={x:r},s={axis:t,keepDims:n};return _s.runKernel(Tr,a,s)}});function lu(e,t,n=null){if(0===e.rank)return bo(e);if(1!==e.rank&&null===n)return lu(jo(e,[-1]),t,n);if(1===e.rank||"number"==typeof n||Array.isArray(n)&&1===n.length){if(1===t)return ou(bo(e),n);if(t===1/0)return tu(bo(e),n);if(t===-1/0)return nu(bo(e),n);if("euclidean"===t||2===t)return su(ou(ru(bo(e),au(2,"int32")),n));throw new Error(`Error in norm: invalid ord value: ${t}`)}if(Array.isArray(n)&&2===n.length){if(1===t)return tu(ou(bo(e),n[0]),n[1]-1);if(t===1/0)return tu(ou(bo(e),n[1]),n[0]);if(t===-1/0)return nu(ou(bo(e),n[1]),n[0]);if("fro"===t||"euclidean"===t)return su(ou(iu(e),n));throw new Error(`Error in norm: invalid ord value: ${t}`)}throw new Error(`Error in norm: invalid axis: ${n}`)}var uu=zs({norm_:function(e,t="euclidean",n=null,r=!1){let a=lu(e=Os(e,"x","norm"),t,n),s=a.shape;if(r){let t=le(n,e.shape);s=Zl(a.shape,t)}return jo(a,s)}});var hu=zs({euclideanNorm_:function(e,t=null,n=!1){return uu(e,"euclidean",t,n)}});var du=zs({exp_:function(e){let t={x:Os(e,"x","exp")};return _s.runKernel(Ht,t)}});var pu=zs({expandDims_:function(e,t=0){let n=Os(e,"x","expandDims","string_or_numeric");K(t<=n.rank,()=>"Axis must be <= rank of the tensor");let r={input:n},a={dim:t};return _s.runKernel(jt,r,a)}});var cu=zs({expm1_:function(e){let t={x:Os(e,"x","expm1")};return _s.runKernel(qt,t)}});var fu=zs({tile_:function(e,t){let n=Os(e,"x","tile","string_or_numeric");K(n.rank===t.length,()=>`Error in transpose: rank of input ${n.rank} must match length of reps ${t}.`);let r={x:n},a={reps:t};return _s.runKernel(jr,r,a)}});var mu=zs({eye_:function(e,t,n,r="float32"){null==t&&(t=e);let a=uo([e,t],r),s=e<=t?e:t;for(let e=0;e<s;++e)a.set(1,e,e);let i=jo(a.toTensor(),[e,t]);if(null==n)return i;if(1===n.length)return fu(pu(i,0),[n[0],1,1]);if(2===n.length)return fu(pu(pu(i,0),0),[n[0],n[1],1,1]);if(3===n.length)return fu(pu(pu(pu(i,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.`)}});var gu=zs({floor_:function(e){let t={x:Os(e,"x","floor","float32")};return _s.runKernel(Yt,t)}});var yu=zs({gather_:function(e,t,n=0,r=0){let a={x:Os(e,"x","gather"),indices:Os(t,"indices","gather","int32")},s={axis:n,batchDims:r};return _s.runKernel(en,a,s)}});var bu=zs({greater_:function(e,t){let n=Os(e,"a","greater","string_or_numeric"),r=Os(t,"b","greater","string_or_numeric");[n,r]=gs(n,r),Ll(n.shape,r.shape);let a={a:n,b:r};return _s.runKernel(nn,a)}});var xu=zs({greaterEqual_:function(e,t){let n=Os(e,"a","greaterEqual","string_or_numeric"),r=Os(t,"b","greaterEqual","string_or_numeric");[n,r]=gs(n,r),Ll(n.shape,r.shape);let a={a:n,b:r};return _s.runKernel(rn,a)}});var vu=zs({imag_:function(e){let t={input:Os(e,"input","imag")};return _s.runKernel(on,t)}});var wu=zs({isFinite_:function(e){let t={x:Os(e,"x","isFinite")};return _s.runKernel(ln,t)}});var ku=zs({isInf_:function(e){let t={x:Os(e,"x","isInf")};return _s.runKernel(un,t)}});var Iu=zs({isNaN_:function(e){let t={x:Os(e,"x","isNaN")};return _s.runKernel(hn,t)}});var Su=zs({leakyRelu_:function(e,t=.2){let n={x:Os(e,"x","leakyRelu")},r={alpha:t};return _s.runKernel(dn,n,r)}});var _u=zs({less_:function(e,t){let n=Os(e,"a","less","string_or_numeric"),r=Os(t,"b","less","string_or_numeric");[n,r]=gs(n,r),Ll(n.shape,r.shape);let a={a:n,b:r};return _s.runKernel(pn,a)}});var Nu=zs({lessEqual_:function(e,t){let n=Os(e,"a","lessEqual","string_or_numeric"),r=Os(t,"b","lessEqual","string_or_numeric");[n,r]=gs(n,r),Ll(n.shape,r.shape);let a={a:n,b:r};return _s.runKernel(cn,a)}});function Tu(e,t,n){if(n<=0)throw new Error("The number of values should be positive.");let r={start:e,stop:t,num:n};return _s.runKernel(fn,{},r)}var Cu=zs({localResponseNormalization_:function(e,t=5,n=1,r=1,a=.5){let s=Os(e,"x","localResponseNormalization");K(4===s.rank||3===s.rank,()=>`Error in localResponseNormalization: x must be rank 3 or 4 but got\n               rank ${s.rank}.`),K(te(t),()=>`Error in localResponseNormalization: depthRadius must be an integer but got depthRadius ${t}.`);let i=s,o=!1;3===s.rank&&(o=!0,i=jo(s,[1,s.shape[0],s.shape[1],s.shape[2]]));let l={x:i},u={depthRadius:t,bias:n,alpha:r,beta:a},h=_s.runKernel(In,l,u);return o?jo(h,[h.shape[1],h.shape[2],h.shape[3]]):h}});var Eu=zs({log_:function(e){let t={x:Os(e,"x","log","float32")};return _s.runKernel(mn,t)}});var Au=zs({log1p_:function(e){let t={x:Os(e,"x","log1p")};return _s.runKernel(gn,t)}});function $u(e){return K(we(e),()=>"The f passed in grad(f) must be a function"),(t,n)=>{let r=Os(t,"x","tf.grad","string_or_numeric"),a=null!=n?Os(n,"dy","tf.grad"):null;return _s.tidy(()=>{let{value:t,grads:n}=_s.gradients(()=>e(r),[r],a);return null!=a&&X(t.shape,a.shape,"The shape of dy passed in grad(f)(x, dy) must match the shape returned by f(x)"),Lu(n),n[0]})}}function Ru(e){return K(we(e),()=>"The f passed in grads(f) must be a function"),(t,n)=>{K(Array.isArray(t),()=>"The args passed in grads(f)(args) must be an array of `Tensor`s or `TensorLike`s");let r=Ls(t,"args","tf.grads","string_or_numeric"),a=null!=n?Os(n,"dy","tf.grads"):null;return _s.tidy(()=>{let{value:t,grads:n}=_s.gradients(()=>e(...r),r,a);return null!=a&&X(t.shape,a.shape,"The shape of dy passed in grads(f)([x1,...], dy) must match the shape returned by f([x1,...])"),Lu(n),n})}}function Fu(e){return K(we(e),()=>"The f passed in valueAndGrad(f) must be a function"),(t,n)=>{K(t instanceof ts,()=>"The x passed in valueAndGrad(f)(x) must be a tensor"),K(null==n||n instanceof ts,()=>"The dy passed in valueAndGrad(f)(x, dy) must be a tensor");let{grads:r,value:a}=_s.gradients(()=>e(t),[t],n);return Lu(r),{grad:r[0],value:a}}}function Du(e){return K(we(e),()=>"The f passed in valueAndGrads(f) must be a function"),(t,n)=>{K(Array.isArray(t)&&t.every(e=>e instanceof ts),()=>"The args passed in valueAndGrads(f)(args) must be array of tensors"),K(null==n||n instanceof ts,()=>"The dy passed in valueAndGrads(f)(args, dy) must be a tensor");let r=_s.gradients(()=>e(...t),t,n);return null!=n&&X(r.value.shape,n.shape,"The shape of dy passed in valueAndGrads(f)([x1,...], dy) must match the shape returned by f([x1,...])"),Lu(r.grads),r}}function Mu(e,t){K(we(e),()=>"The f passed in variableGrads(f) must be a function"),K(null==t||Array.isArray(t)&&t.every(e=>e instanceof rs),()=>"The varList passed in variableGrads(f, varList) must be an array of variables");let n=null!=t;if(!n){t=[];for(let e in _s.registeredVariables)t.push(_s.registeredVariables[e])}let r=n?t.filter(e=>!e.trainable):null,a=t.length;K((t=t.filter(e=>e.trainable)).length>0,()=>`variableGrads() expects at least one of the input variables to be trainable, but none of the ${a} variables is trainable.`);let{value:s,grads:i}=_s.gradients(e,t,null,!0);K(i.some(e=>null!=e),()=>"Cannot find a connection between any variable and the result of the loss function y=f(x). Please make sure the operations that use variables are inside the function f passed to minimize()."),K(0===s.rank,()=>`The f passed in variableGrads(f) must return a scalar, but it returned a rank-${s.rank} tensor`);let o={};return t.forEach((e,t)=>{null!=i[t]&&(o[e.name]=i[t])}),null!=r&&r.forEach(e=>o[e.name]=null),{value:s,grads:o}}function Ou(e){return _s.customGrad(e)}function Lu(e){if(e.filter(e=>null==e).length>0)throw new Error("Cannot compute gradient of y=f(x) with respect to x. Make sure that\n    the f you passed encloses all operations that lead from x to y.")}var Pu=zs({neg_:function(e){let t={x:Os(e,"x","neg")};return _s.runKernel(Bn,t)}});var zu=zs({softplus_:function(e){let t={x:Os(e,"x","softplus")};return _s.runKernel(_r,t)}});var Bu=zs({logSigmoid_:function(e){let t=Os(e,"x","logSigmoid");return Ou(e=>({value:Pu(zu(Pu(e))),gradFunc:t=>yo(t,Yo(Pu(e)))}))(t)}});var Wu=zs({sub_:function(e,t){let n=Os(e,"a","sub"),r=Os(t,"b","sub");[n,r]=gs(n,r);let a={a:n,b:r};return _s.runKernel(Ur,a)}});var Vu=zs({logSoftmax_:function(e,t=-1){let n=Os(e,"logits","logSoftmax");if(-1===t&&(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}`);return Ou((e,n)=>{let r=tu(e,t,!0),a=Wu(e,r),s=Wu(ho(a,"float32"),Eu(ou(du(a),t,!0)));return n([s]),{value:s,gradFunc:(e,n)=>{let[r]=n,a=du(r);return Wu(e,yo(ou(e,t,!0),a))}}})(n)}});var Uu=zs({logSumExp_:function(e,t=null,n=!1){let r=Os(e,"x","logSumExp"),a=le(t,r.shape),s=tu(r,a,!0),i=Wu(r,s),o=du(i),l=ou(o,a),u=Eu(l),h=fo(jo(s,u.shape),u);if(n){let e=Zl(h.shape,a);return jo(h,e)}return h}});var Gu=zs({logicalAnd_:function(e,t){let n=Os(e,"a","logicalAnd","bool"),r=Os(t,"b","logicalAnd","bool");Ll(n.shape,r.shape);let a={a:n,b:r};return _s.runKernel(yn,a)}});var Hu=zs({logicalNot_:function(e){let t={x:Os(e,"x","logicalNot","bool")};return _s.runKernel(bn,t)}});var ju=zs({logicalOr_:function(e,t){let n=Os(e,"a","logicalOr","bool"),r=Os(t,"b","logicalOr","bool");Ll(n.shape,r.shape);let a={a:n,b:r};return _s.runKernel(xn,a)}});var qu=zs({logicalXor_:function(e,t){let n=Os(e,"a","logicalXor","bool"),r=Os(t,"b","logicalXor","bool");return Ll(n.shape,r.shape),Gu(ju(e,t),Hu(Gu(e,t)))}}),Ku=2147483648;var Xu=zs({searchSorted_:function(e,t,n="left"){let r=Os(e,"sortedSequence","searchSorted"),a=Os(t,"values","searchSorted"),s=r.shape[r.shape.length-1],i=a.shape[a.shape.length-1],o=jo(r,[-1,s]),l=jo(a,[-1,i]);if(o.rank<2)throw new Error("Sorted input argument must be at least 2-dimensional");if(o.shape[0]!==l.shape[0])throw new Error("Leading dimension of 'sortedSequence' and 'values' must match.");if(Y(l.shape)>=Ku)throw new Error(`values tensor size must less than ${Ku}`);if(o.shape[1]>=Ku)throw new Error(`trailing dim_size must less than ${Ku} for int32 output type, was ${o.shape[1]}`);let u={sortedSequence:o,values:l},h={side:n};return _s.runKernel(yr,u,h)}});function Zu(e,t){return Xu(e,t,"left")}var Yu=zs({maxPool_:function(e,t,n,r,a){let s=Os(e,"x","maxPool"),i=s,o=!1;3===s.rank&&(o=!0,i=jo(s,[1,s.shape[0],s.shape[1],s.shape[2]])),K(4===i.rank,()=>`Error in maxPool: input must be rank 4 but got rank ${i.rank}.`),K(Vo(n,1),()=>`Error in maxPool: Either strides or dilations must be 1. Got strides ${n} and dilations '1'`),Ho("maxPool",r,a);let l={x:i},u={filterSize:t,strides:n,pad:r,dimRoundingMode:a},h=_s.runKernel(Cn,l,u);return o?jo(h,[h.shape[1],h.shape[2],h.shape[3]]):h}});var Ju=zs({maxPool3d_:function(e,t=[1,1,1],n,r,a,s="NDHWC"){let i=Os(e,"x","maxPool3d"),o=i,l=!1;4===i.rank&&(l=!0,o=jo(i,[1,i.shape[0],i.shape[1],i.shape[2],i.shape[3]])),K(5===o.rank,()=>`Error in maxPool3d: x must be rank 5 but got rank ${o.rank}.`),K("NDHWC"===s,()=>`Error in maxPool3d: Only NDHWC is currently supported, but got dataFormat of ${s}`),Ho("maxPool3d",r,a);let u={x:o},h={filterSize:t,strides:n,pad:r,dimRoundingMode:a,dataFormat:s},d=_s.runKernel(An,u,h);return l?jo(d,[d.shape[1],d.shape[2],d.shape[3],d.shape[4]]):d}});var Qu=zs({maxPoolWithArgmax_:function(e,t,n,r,a=!1){let s={x:Os(e,"x","maxPoolWithArgmax")},i={filterSize:t,strides:n,pad:r,includeBatchInIndex:a},o=_s.runKernel(Rn,s,i);return{result:o[0],indexes:o[1]}}});var eh=zs({maximum_:function(e,t){let n=Os(e,"a","maximum"),r=Os(t,"b","maximum");[n,r]=gs(n,r),"bool"===n.dtype&&(n=ho(n,"int32"),r=ho(r,"int32")),Ll(n.shape,r.shape);let a={a:n,b:r};return _s.runKernel(Tn,a)}});var th=zs({mean_:function(e,t=null,n=!1){let r={x:Os(e,"x","mean")},a={axis:t,keepDims:n};return _s.runKernel(Fn,r,a)}});function nh(e,t="float32"){if(Ae(e),"complex64"===t){let t=nh(e,"float32"),n=nh(e,"float32");return Bs(t,n)}let n=Ce(Y(e),t);return _s.makeTensor(n,e,t)}function rh(e,t="float32"){if(Ae(e),"complex64"===t){let t=rh(e,"float32"),n=nh(e,"float32");return Bs(t,n)}let n=Te(Y(e),t);return _s.makeTensor(n,e,t)}function ah(e,t,{indexing:n="xy"}={}){if("xy"!==n&&"ij"!==n)throw new TypeError(`${n} is not a valid third argument to meshgrid`);if(void 0===e)return[];let r=Os(e,"x","meshgrid",e instanceof ts?e.dtype:"float32");if(void 0===t)return[r];let a=Os(t,"y","meshgrid",t instanceof ts?t.dtype:"float32"),s=Y(r.shape),i=Y(a.shape);return"xy"===n?(r=jo(r,[1,-1]),a=jo(a,[-1,1]),[Zo(rh([i,1],r.dtype),r),Zo(a,rh([1,s],a.dtype))]):(r=jo(r,[-1,1]),a=jo(a,[1,-1]),[Zo(r,rh([1,i],r.dtype)),Zo(rh([s,1],a.dtype),a)])}var sh=zs({minimum_:function(e,t){let n=Os(e,"a","minimum"),r=Os(t,"b","minimum");[n,r]=gs(n,r),"bool"===n.dtype&&(n=ho(n,"int32"),r=ho(r,"int32")),Ll(n.shape,r.shape);let a={a:n,b:r};return _s.runKernel(Mn,a)}});var ih=zs({mirrorPad_:function(e,t,n){K("reflect"===n||"symmetric"===n,()=>`Invalid mode. Mode must be either reflect or symmetric. Got ${n}.`);let r=Os(e,"x","mirrorPad");if(0===r.rank)throw new Error("mirrorPad(scalar) is not defined. Pass non-scalar to mirrorPad");K(t.length===r.rank,()=>`Padding doesn't match input. Must be ${r.rank}. Got ${t.length}.`);let a="reflect"===n?1:0;for(let e=0;e<r.rank;e++)K(2===t[e].length,()=>"Invalid number of paddings. Must be length of 2 each."),K(t[e][0]>=0&&t[e][0]<=r.shape[e]-a&&t[e][1]>=0&&t[e][1]<=r.shape[e]-a,()=>`Padding in dimension ${e} cannot be greater than or equal to ${r.shape[e]-a} or less than 0 for input of shape ${r.shape}`);let s={paddings:t,mode:n},i={x:r};return _s.runKernel(On,i,s)}});var oh=zs({mod_:function(e,t){let n=Os(e,"a","mod"),r=Os(t,"b","mod");[n,r]=gs(n,r);let a={a:n,b:r};return _s.runKernel(Ln,a)}});var lh=zs({moments_:function(e,t=null,n=!1){let r=le(t,(e=Os(e,"x","moments")).shape),a=th(e,r,n),s=a.shape;n||(s=Zl(a.shape,r));let i=iu(Wu(ho(e,"float32"),jo(a,s)));return{mean:a,variance:th(i,r,n)}}});var uh=zs({multiRNNCell_:function(e,t,n,r){let a=Os(t,"data","multiRNNCell"),s=Ls(n,"c","multiRNNCell"),i=Ls(r,"h","multiRNNCell"),o=a,l=[];for(let t=0;t<e.length;t++){let n=e[t](o,s[t],i[t]);l.push(n[0]),l.push(n[1]),o=n[1]}let u=[],h=[];for(let e=0;e<l.length;e+=2)u.push(l[e]),h.push(l[e+1]);return[u,h]}});var hh=zs({multinomial_:function(e,t,n,r=!1){let a=Os(e,"logits","multinomial"),s=a.size,i=a.rank;if(s<2)throw new Error(`Error in multinomial: you need at least 2 outcomes, but got ${s}.`);if(i>2)throw new Error(`Rank of probabilities must be 1 or 2, but is ${i}`);n=n||Math.random();let o={logits:1===i?jo(a,[1,-1]):a},l={numSamples:t,seed:n,normalized:r},u=_s.runKernel(Pn,o,l);return 1===i?jo(u,[u.size]):u}});var dh=zs({notEqual_:function(e,t){let n=Os(e,"a","notEqual","string_or_numeric"),r=Os(t,"b","notEqual","string_or_numeric");[n,r]=gs(n,r),Ll(n.shape,r.shape);let a={a:n,b:r};return _s.runKernel(Wn,a)}});var ph=zs({oneHot_:function(e,t,n=1,r=0,a="int32"){if(t<2)throw new Error(`Error in oneHot: depth must be >=2, but it is ${t}`);let s={indices:Os(e,"indices","oneHot","int32")},i={dtype:a,depth:t,onValue:n,offValue:r};return _s.runKernel(jn,s,i)}});var ch=zs({onesLike_:function(e){let t={x:Os(e,"x","onesLike")};return _s.runKernel(Hn,t)}});var fh=zs({outerProduct_:function(e,t){let n=Os(e,"v1","outerProduct"),r=Os(t,"v2","outerProduct");K(1===n.rank&&1===r.rank,()=>`Error in outerProduct: inputs must be rank 1, but got ranks ${n.rank} and ${r.rank}.`);let a=jo(n,[-1,1]),s=jo(r,[1,-1]);return Zo(a,s)}});var mh=zs({pad_:function(e,t,n=0){let r=Os(e,"x","pad");if(0===r.rank)throw new Error("pad(scalar) is not defined. Pass non-scalar to pad");let a={paddings:t,constantValue:n},s={x:r};return _s.runKernel(Kn,s,a)}});var gh=zs({pad1d_:function(e,t,n=0){return K(2===t.length,()=>"Invalid number of paddings. Must be length of 2."),mh(e,[t],n)}});var yh=zs({pad2d_:function(e,t,n=0){return K(2===t.length&&2===t[0].length&&2===t[1].length,()=>"Invalid number of paddings. Must be length of 2 each."),mh(e,t,n)}});var bh=zs({pad3d_:function(e,t,n=0){return K(3===t.length&&2===t[0].length&&2===t[1].length&&2===t[2].length,()=>"Invalid number of paddings. Must be length of 2 each."),mh(e,t,n)}});var xh=zs({pad4d_:function(e,t,n=0){return K(4===t.length&&2===t[0].length&&2===t[1].length&&2===t[2].length&&2===t[3].length,()=>"Invalid number of paddings. Must be length of 2 each."),mh(e,t,n)}});var vh=zs({spaceToBatchND_:function(e,t,n){let r=Os(e,"x","spaceToBatchND");K(r.rank>=1+t.length,()=>`input rank ${r.rank} should be > than [blockShape] ${t.length}`),K(n.length===t.length,()=>`paddings.shape[0] ${n.length} must be equal to [blockShape] ${t.length}`),K(r.shape.reduce((e,r,a)=>a>0&&a<=t.length?e&&(r+n[a-1][0]+n[a-1][1])%t[a-1]===0:e,!0),()=>`input spatial dimensions ${r.shape.slice(1)} with paddings ${n.toString()} must be divisible by blockShapes ${t.toString()}`);let a={x:r},s={blockShape:t,paddings:n};return _s.runKernel(Cr,a,s)}});var wh=zs({pool_:function(e,t,n,r,a,s,i){null==a&&(a=[1,1]),null==s&&(s=1),0===r&&(r="valid");let o=Os(e,"x","maxPool"),l=o,u=!1;3===o.rank&&(u=!0,l=jo(o,[1,o.shape[0],o.shape[1],o.shape[2]])),K(Vo(s,a),()=>`Error in pool: Either strides or dilations must be 1. Got strides ${s} and dilations '${a}'`);let h,d=Ro(l.shape,t,s,a,r),p=[d.dilationHeight,d.dilationWidth];h="same"===r?function(e,t){let n=e.map((e,n)=>e+(e-1)*(t[n]-1)).map(e=>e-1),r=n.map(e=>Math.floor(e/2)),a=n.map((e,t)=>e-r[t]);return n.map((e,t)=>[r[t],a[t]])}([d.filterHeight,d.filterWidth],p):[[0,0],[0,0]];let c=1===p[0]&&1===p[1],[f,m]=function(e,t,n){let r=n.map(e=>e[0]),a=n.map(e=>e[1]),s=e.concat(r,a),i=t.map((e,t)=>(e-s[t]%e)%e),o=a.map((e,t)=>e+i[t]),l=t.map((e,t)=>[r[t],o[t]]),u=t.map((e,t)=>[0,i[t]]);return[l,u]}([d.inHeight,d.inWidth],p,h),g=c?r:"valid",y=c?l:vh(l,p,f),b=("avg"===n?()=>qo(y,t,s,g,i):()=>Yu(y,t,s,g,i))(),x=c?b:tl(b,p,m);return u?jo(x,[x.shape[1],x.shape[2],x.shape[3]]):x}});var kh=zs({prelu_:function(e,t){let n={x:Os(e,"x","prelu"),alpha:Os(t,"alpha","prelu")};return _s.runKernel(Yn,n)}});var Ih=zs({prod_:function(e,t=null,n=!1){let r=Os(e,"x","prod");"bool"===r.dtype&&(r=ho(r,"int32"));let a={x:r},s={axis:t,keepDims:n};return _s.runKernel(Jn,a,s)}});var Sh=zs({raggedGather_:function(e,t,n,r){let a={paramsNestedSplits:e.map((e,t)=>Os(e,`tensors${t}`,"raggedGather","int32")),paramsDenseValues:Os(t,"paramsDenseValues","raggedGather"),indices:Os(n,"indices","raggedGather","int32")},s={outputRaggedRank:r},i=_s.runKernel(Qn,a,s);return{outputNestedSplits:i.slice(0,i.length-1),outputDenseValues:i[i.length-1]}}});var _h=zs({raggedRange_:function(e,t,n){let r=Os(e,"starts","raggedRange"),a={starts:r,limits:Os(t,"limits","raggedRange",r.dtype),deltas:Os(n,"deltas","raggedRange",r.dtype)},s=_s.runKernel(er,a);return{rtNestedSplits:s[0],rtDenseValues:s[1]}}});var Nh=zs({raggedTensorToTensor_:function(e,t,n,r,a){let s=Os(e,"shape","raggedTensorToTensor","int32"),i=Os(t,"values","raggedTensorToTensor"),o={shape:s,values:i,defaultValue:Os(n,"defaultValue","raggedTensorToTensor",i.dtype),rowPartitionTensors:r.map((e,t)=>Os(e,`tensors${t}`,"raggedTensorToTensor","int32"))},l={rowPartitionTypes:a};return _s.runKernel(tr,o,l)}});var Th=zs({rand_:function(e,t,n){Ae(e);let r=Y(e),a=null;if(null==n||"float32"===n)a=new Float32Array(r);else if("int32"===n)a=new Int32Array(r);else{if("bool"!==n)throw new Error(`Unknown data type ${n}`);a=new Uint8Array(r)}for(let e=0;e<r;e++)a[e]=t();return _s.makeTensor(a,e,n)}}),Ch=m(T()),Eh={};f(Eh,{TEST_EPSILON_FLOAT16:()=>Ah,createVideoElement:()=>Wh,encodeStrings:()=>Bh,expectArrayBuffersEqual:()=>zh,expectArraysClose:()=>$h,expectArraysEqual:()=>Mh,expectNumbersClose:()=>Oh,expectPromiseToFail:()=>Dh,expectValuesInRange:()=>Ph,play:()=>Vh,testEpsilon:()=>Rh});var Ah=.1;function $h(e,t,n){return null==n&&(n=Rh()),Fh(e,t,(e,t)=>Lh(e,t,n))}function Rh(){return 32===_s.backend.floatPrecision()?.001:Ah}function Fh(e,t,n){let r=!0;if((Va(e)||Va(t))&&(r=!1),Va(e)&&Va(t)&&(r=!0),r){let n=e.constructor.name,r=t.constructor.name;if(n!==r)throw new Error(`Arrays are of different type. Actual: ${n}. Expected: ${r}`)}if(Array.isArray(e)&&Array.isArray(t)){let n=Fs(e),r=Fs(t);if(!ee(n,r))throw new Error(`Arrays have different shapes. Actual: [${n}]. Expected: [${r}]`)}let a=Va(e)?e:Ua(e),s=Va(t)?t:Ua(t);if(a.length!==s.length)throw new Error(`Arrays have different lengths actual: ${a.length} vs expected: ${s.length}.\nActual:   ${a}.\nExpected: ${s}.`);for(let e=0;e<s.length;++e){let t=a[e],r=s[e];if(!n(t,r))throw new Error(`Arrays differ: actual[${e}] = ${t}, expected[${e}] = ${r}.\nActual:   ${a}.\nExpected: ${s}.`)}"undefined"!=typeof expect&&expect().nothing()}function Dh(e,t){e().then(()=>t.fail(),()=>t()),"undefined"!=typeof expect&&expect().nothing()}function Mh(e,t){let n="string"==typeof t||"number"==typeof t||"boolean"==typeof t?[t]:t;return ye(e)||ye(e[0])||ye(t)||ye(t[0])?Fh(e,n,(e,t)=>e==t):Fh(e,t,(e,t)=>Lh(e,t,0))}function Oh(e,t,n){if(null==n&&(n=Rh()),!Lh(e,t,n))throw new Error(`Numbers differ: actual === ${e}, expected === ${t}`);"undefined"!=typeof expect&&expect().nothing()}function Lh(e,t,n){return!isFinite(e)&&!isFinite(t)||!(isNaN(e)||isNaN(t)||Math.abs(e-t)>n)}function Ph(e,t,n){for(let r=0;r<e.length;r++)if(e[r]<t||e[r]>n)throw new Error(`Value out of range:${e[r]} low: ${t}, high: ${n}`)}function zh(e,t){let n=new Float32Array(e),r=new Float32Array(t);if(n.length!==r.length)throw new Error(`Expected ArrayBuffer to be of length ${r.length}, but it was ${n.length}`);for(let e=0;e<r.length;e++)if(n[e]!==r[e])throw new Error(`Expected ArrayBuffer value at ${e} to be ${r[e]} but got ${n[e]} instead`)}function Bh(e){for(let t=0;t<e.length;t++){let n=e[t];Array.isArray(n)?Bh(n):e[t]=Ba(n)}return e}function Wh(e){let t=document.createElement("video");return"playsInline"in t&&(t.playsInline=!0),t.muted=!0,t.loop=!0,t.style.position="fixed",t.style.left="0px",t.style.top="0px",t.preload="auto",t.appendChild(e),new Promise(e=>{t.addEventListener("loadeddata",n=>e(t)),t.load()})}async function Vh(e){await e.play(),"requestVideoFrameCallback"in e&&await new Promise(t=>{e.requestVideoFrameCallback(t)})}var Uh=class{constructor(e,t,n,r,a){this.mean=e,this.stdDev=t,this.dtype=n,this.nextVal=NaN,this.truncated=r,this.truncated&&(this.upper=this.mean+2*this.stdDev,this.lower=this.mean-2*this.stdDev);let s=a||Math.random();this.random=Ch.alea(s.toString())}nextValue(){if(!isNaN(this.nextVal)){let e=this.nextVal;return this.nextVal=NaN,e}let e,t,n=!1;for(;!n;){let r,a,s;do{r=2*this.random()-1,a=2*this.random()-1,s=r*r+a*a}while(s>=1||0===s);let i=Math.sqrt(-2*Math.log(s)/s);e=this.mean+this.stdDev*r*i,t=this.mean+this.stdDev*a*i,(!this.truncated||this.isValidTruncated(e))&&(n=!0)}return(!this.truncated||this.isValidTruncated(t))&&(this.nextVal=this.convertValue(t)),this.convertValue(e)}convertValue(e){return null==this.dtype||"float32"===this.dtype?e:Math.round(e)}isValidTruncated(e){return e<=this.upper&&e>=this.lower}},Gh=class{constructor(e,t,n,r){this.alpha=e,this.beta=1/t,this.dtype=n;let a=r||Math.random();this.randu=Ch.alea(a.toString()),this.randn=new Uh(0,1,n,!1,this.randu()),this.d=e<1?e+2/3:e-1/3,this.c=1/Math.sqrt(9*this.d)}nextValue(){let e,t,n,r,a,s;for(;;){do{r=this.randn.nextValue(),s=1+this.c*r}while(s<=0);if(s*=s*s,e=r*r,t=1-.331*e*e,n=.5*e+this.d*(1-s+Math.log(s)),a=this.randu(),a<t||Math.log(a)<n)break}return s=1/this.beta*this.d*s,this.alpha<1&&(s*=Math.pow(this.randu(),1/this.alpha)),this.convertValue(s)}convertValue(e){return"float32"===this.dtype?e:Math.round(e)}},Hh=class{constructor(e=0,t=1,n,r){if(this.canReturnFloat=()=>null==this.dtype||"float32"===this.dtype,this.min=e,this.range=t-e,this.dtype=n,null==r&&(r=Math.random()),"number"==typeof r&&(r=r.toString()),!this.canReturnFloat()&&this.range<=1)throw new Error(`The difference between ${e} - ${t} <= 1 and dtype is not float`);this.random=Ch.alea(r)}convertValue(e){return this.canReturnFloat()?e:Math.round(e)}nextValue(){return this.convertValue(this.min+this.range*this.random())}};var jh=zs({randomGamma_:function(e,t,n=1,r="float32",a){if(Ae(e),null==n&&(n=1),null==r&&(r="float32"),"float32"!==r&&"int32"!==r)throw new Error(`Unsupported data type ${r}`);let s=new Gh(t,n,r,a),i=uo(e,r);for(let e=0;e<i.values.length;e++)i.values[e]=s.nextValue();return i.toTensor()}});var qh=zs({randomNormal_:function(e,t=0,n=1,r,a){if(Ae(e),null!=r&&"bool"===r)throw new Error(`Unsupported data type ${r}`);let s=new Uh(t,n,r,!1,a),i=uo(e,r);for(let e=0;e<i.values.length;e++)i.values[e]=s.nextValue();return i.toTensor()}});var Kh=zs({randomStandardNormal_:function(e,t,n){if(null!=t&&"bool"===t)throw new Error(`Unsupported data type ${t}`);return qh(e,0,1,t,n)}});var Xh=zs({randomUniform_:function(e,t=0,n=1,r="float32",a){Ae(e);let s=uo(e,r),i=new Hh(t,n,null,a);for(let e=0;e<s.values.length;e++)s.values[e]=i.nextValue();return s.toTensor()}});var Zh=zs({randomUniformInt_:function(e,t,n,r){return Xh(e,t,n,"int32",r)}});function Yh(e,t,n=1,r="float32"){if(0===n)throw new Error("Cannot have a step of zero");let a={start:e,stop:t,step:n,dtype:r};return _s.runKernel(nr,{},a)}var Jh=zs({real_:function(e){let t={input:Os(e,"input","real")};return _s.runKernel(rr,t)}});var Qh=zs({reciprocal_:function(e){let t={x:Os(e,"x","reciprocal")};return _s.runKernel(ar,t)}});var ed=zs({relu_:function(e){let t={x:Os(e,"x","relu")};return _s.runKernel(sr,t)}});var td=zs({relu6_:function(e){let t={x:Os(e,"x","relu6")};return _s.runKernel(dr,t)}});var nd=zs({reverse_:function(e,t){let n={x:Os(e,"x","reverse")},r={dims:t};return _s.runKernel(pr,n,r)}});var rd=zs({reverse1d_:function(e){let t=Os(e,"x","reverse");return K(1===t.rank,()=>`Error in reverse1D: x must be rank 1 but got rank ${t.rank}.`),nd(t,0)}});var ad=zs({reverse2d_:function(e,t){let n=Os(e,"x","reverse");return K(2===n.rank,()=>`Error in reverse2D: x must be rank 2 but got rank ${n.rank}.`),nd(n,t)}});var sd=zs({reverse3d_:function(e,t){let n=Os(e,"x","reverse");return K(3===n.rank,()=>`Error in reverse3D: x must be rank 3 but got rank ${n.rank}.`),nd(n,t)}});var id=zs({reverse4d_:function(e,t){let n=Os(e,"x","reverse");return K(4===n.rank,()=>`Error in reverse4D: x must be rank 4 but got rank ${n.rank}.`),nd(n,t)}});var od=zs({round_:function(e){let t={x:Os(e,"x","round")};return _s.runKernel(cr,t)}});var ld=zs({rsqrt_:function(e){let t={x:Os(e,"x","rsqrt","float32")};return _s.runKernel(fr,t)}});var ud=zs({selu_:function(e){let t={x:Os(e,"x","selu")};return _s.runKernel(xr,t)}});var hd=zs({separableConv2d_:function(e,t,n,r,a,s=[1,1],i="NHWC"){let o=Os(e,"x","separableConv2d"),l=Os(t,"depthwiseFilter","separableConv2d"),u=Os(n,"pointwiseFilter","separableConv2d"),h=o,d=!1;if(3===o.rank&&(d=!0,h=jo(o,[1,o.shape[0],o.shape[1],o.shape[2]])),"NCHW"===i)throw new Error("separableConv2d currently does not support dataFormat NCHW; only NHWC is supported");K(4===h.rank,()=>`Error in separableConv2d: input must be rank 4, but got rank ${h.rank}.`),K(4===l.rank,()=>`Error in separableConv2d: depthwise filter must be rank 4, but got rank ${l.rank}.`),K(4===u.rank,()=>`Error in separableConv2d: pointwise filter must be rank 4, but got rank ${l.rank}.`),K(1===u.shape[0],()=>`Error in separableConv2d: the first dimension of pointwise filter  must be 1, but got ${u.shape[0]}.`),K(1===u.shape[1],()=>`Error in separableConv2d: the second dimension of pointwise filter must be 1, but got ${u.shape[1]}.`);let p=l.shape[2],c=l.shape[3];K(u.shape[2]===p*c,()=>`Error in separableConv2d: the third dimension of pointwise filter must be ${p*c}, but got ${u.shape[2]}.`);let f=$l(h,l,r,a,i,s),m=bl(f,u,1,"valid",i);return d?jo(m,[m.shape[1],m.shape[2],m.shape[3]]):m}});var dd=async function(e,t){let n=Os(e,"x","setdiff1d"),r=Os(t,"y","setdiff1d");K(n.dtype===r.dtype,()=>`x and y should have the same dtype, but got x (${n.dtype}) and y (${r.dtype}).`),K(1===n.rank,()=>`x should be 1D tensor, but got x (${n.shape}).`),K(1===r.rank,()=>`y should be 1D tensor, but got y (${r.shape}).`);let a=await n.data(),s=await r.data(),i=new Set(s),o=0;for(let e=0;e<a.length;e++)i.has(a[e])||o++;let l=new Ja([o],n.dtype),u=new Ja([o],"int32");for(let e=0,t=0;e<a.length;e++)i.has(a[e])||(l.values[t]=a[e],u.values[t]=e,t++);return[l.toTensor(),u.toTensor()]};var pd=zs({sign_:function(e){let t={x:Os(e,"x","sign")};return _s.runKernel(Ir,t)}});var cd=zs({sin_:function(e){let t={x:Os(e,"x","sin","float32")};return _s.runKernel(wr,t)}});var fd=zs({sinh_:function(e){let t={x:Os(e,"x","sinh")};return _s.runKernel(kr,t)}});var md=zs({slice1d_:function(e,t,n){let r=Os(e,"x","slice1d");return K(1===r.rank,()=>`slice1d expects a rank-1 tensor, but got a rank-${r.rank} tensor`),Jo(r,[t],[n])}});var gd=zs({slice2d_:function(e,t,n){let r=Os(e,"x","slice2d");return K(2===r.rank,()=>`slice2d expects a rank-2 tensor, but got a rank-${r.rank} tensor`),Jo(r,t,n)}});var yd=zs({slice3d_:function(e,t,n){let r=Os(e,"x","slice3d");return K(3===r.rank,()=>`slice3d expects a rank-3 tensor, but got a rank-${r.rank} tensor`),Jo(r,t,n)}});var bd=zs({slice4d_:function(e,t,n){let r=Os(e,"x","slice4d");return K(4===r.rank,()=>`slice4d expects a rank-4 tensor, but got a rank-${r.rank} tensor`),Jo(r,t,n)}});var xd=zs({softmax_:function(e,t=-1){let n=Os(e,"logits","softmax","float32");if(-1===t&&(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}`);let r={logits:n},a={dim:t};return _s.runKernel(Ar,r,a)}});var vd=zs({fft_:function(e){K("complex64"===e.dtype,()=>`The dtype for tf.spectral.fft() must be complex64 but got ${e.dtype}.`);let t={input:e};return _s.runKernel(Kt,t)}});var wd=zs({ifft_:function(e){K("complex64"===e.dtype,()=>`The dtype for tf.spectral.ifft() must be complex64 but got ${e.dtype}.`);let t={input:e};return _s.runKernel(sn,t)}});var kd=zs({irfft_:function(e){let t,n=e.shape[e.shape.length-1],r=e.size/n;if(n<=2){let a=jo(e,[r,n]);t=wd(a)}else{let a=[r,2*(n-1)],s=jo(Jh(e),[r,n]),i=jo(vu(e),[r,n]),o=nd(Jo(s,[0,1],[r,n-2]),1),l=yo(nd(Jo(i,[0,1],[r,n-2]),1),au(-1)),u=Xo([s,o],1),h=Xo([i,l],1),d=jo(Bs(u,h),[a[0],a[1]]);t=wd(d)}if(t=Jh(t),3===e.rank&&0!==e.shape[0]){let n=t,r=e.shape[0];t=jo(t,[r,t.shape[0]/r,t.shape[1]]),n.dispose()}return t}});var Id=zs({split_:function(e,t,n=0){let r={x:Os(e,"x","split")},a={numOrSizeSplits:t,axis:n};return _s.runKernel(Er,r,a)}});var Sd=zs({rfft_:function(e,t){K("float32"===e.dtype,()=>`The dtype for rfft() must be real value but got ${e.dtype}`);let n,r=e.shape[e.shape.length-1],a=e.size/r;if(null!=t&&t<r){let a=e.shape.map(e=>0),s=e.shape.map(e=>e);s[e.shape.length-1]=t,n=Jo(e,a,s),r=t}else if(null!=t&&t>r){let a=e.shape.map(e=>e);a[e.shape.length-1]=t-r,n=Xo([e,nh(a)],e.shape.length-1),r=t}else n=e;let s=Bl(n),i=jo(Bs(n,s),[a,r]),o=vd(i),l=Math.floor(r/2)+1,u=Jh(o),h=vu(o),d=Id(u,[l,r-l],u.shape.length-1),p=Id(h,[l,r-l],h.shape.length-1),c=n.shape.slice();return c[n.shape.length-1]=l,jo(Bs(d[0],p[0]),c)}});var _d=zs({squaredDifference_:function(e,t){let n=Os(e,"a","squaredDifference"),r=Os(t,"b","squaredDifference");[n,r]=gs(n,r),Ll(n.shape,r.shape);let a={a:n,b:r};return _s.runKernel(Or,a,{})}});var Nd=zs({squeeze_:function(e,t){let n=Os(e,"x","squeeze","string_or_numeric");return jo(n,ue(n.shape,t).newShape)}});var Td=zs({stack_:function(e,t=0){let n=Ls(e,"tensors","stack","string_or_numeric");K(n.length>=1,()=>"Pass at least one tensor to tf.stack"),n.length>0&&K(t<=n[0].rank,()=>"Axis must be <= rank of the tensor");let r=n,a={axis:t};return _s.runKernel(qn,r,a)}});var Cd=zs({step_:function(e,t=0){let n={x:Os(e,"x","step")},r={alpha:t};return _s.runKernel(ta,n,r)}});var Ed=zs({stridedSlice_:function(e,t,n,r,a=0,s=0,i=0,o=0,l=0){let u={x:Os(e,"x","stridedSlice","string_or_numeric")},h={begin:t,end:n,strides:r,beginMask:a,endMask:s,ellipsisMask:i,newAxisMask:o,shrinkAxisMask:l};return _s.runKernel(zr,u,h)}});var Ad=zs({tan_:function(e){let t={x:Os(e,"x","tan","float32")};return _s.runKernel(Gr,t)}});function $d(e,t){Z(e);let n=Fs(e,t);if(1!==n.length)throw new Error("tensor1d() requires values to be a flat/TypedArray");return Ws(e,null,n,t)}function Rd(e,t,n){if(Z(e),null!=t&&2!==t.length)throw new Error("tensor2d() requires shape to have two numbers");let r=Fs(e,n);if(2!==r.length&&1!==r.length)throw new Error("tensor2d() requires values to be number[][] or flat/TypedArray");if(1===r.length&&null==t)throw new Error("tensor2d() requires shape to be provided when `values` are a flat/TypedArray");return Ws(e,t,r,n)}function Fd(e,t,n){if(Z(e),null!=t&&3!==t.length)throw new Error("tensor3d() requires shape to have three numbers");let r=Fs(e,n);if(3!==r.length&&1!==r.length)throw new Error("tensor3d() requires values to be number[][][] or flat/TypedArray");if(1===r.length&&null==t)throw new Error("tensor3d() requires shape to be provided when `values` are a flat array");return Ws(e,t,r,n)}function Dd(e,t,n){if(Z(e),null!=t&&4!==t.length)throw new Error("tensor4d() requires shape to have four numbers");let r=Fs(e,n);if(4!==r.length&&1!==r.length)throw new Error("tensor4d() requires values to be number[][][][] or flat/TypedArray");if(1===r.length&&null==t)throw new Error("tensor4d() requires shape to be provided when `values` are a flat array");return Ws(e,t,r,n)}function Md(e,t,n){if(Z(e),null!=t&&5!==t.length)throw new Error("tensor5d() requires shape to have five numbers");let r=Fs(e,n);if(5!==r.length&&1!==r.length)throw new Error("tensor5d() requires values to be number[][][][][] or flat/TypedArray");if(1===r.length&&null==t)throw new Error("tensor5d() requires shape to be provided when `values` are a flat array");return Ws(e,t,r,n)}function Od(e,t,n){if(Z(e),null!=t&&6!==t.length)throw new Error("tensor6d() requires shape to have six numbers");let r=Fs(e,n);if(6!==r.length&&1!==r.length)throw new Error("tensor6d() requires values to be number[][][][][][] or flat/TypedArray");if(1===r.length&&null==t)throw new Error("tensor6d() requires shape to be provided when `values` are a flat array");return Ws(e,t=t||r,r,n)}var Ld={};function Pd(e,t,n){let r=t.rank>1?t.shape[t.rank-1]:1,a=t.rank>1?t.rank-1:1,s=`Must have updates.shape = indices.shape[:batchDim] + shape[sliceDim:], got updates.shape: ${n.shape}, indices.shape: ${t.shape}, shape: ${e}, sliceDim: ${r}, and batchDim: ${a}.`;if(n.rank<a)throw new Error(s+` update.rank < ${a}. `);if(e.length<r+(n.rank-a))throw new Error(s+` Output shape length < ${r+(n.rank-a)}`);if(n.rank!==a+e.length-r)throw new Error(s+" update.rank != "+(a+e.length-r));for(let e=0;e<a;++e)if(n.shape[e]!==t.shape[e])throw new Error(s+` updates.shape[${e}] (${n.shape[e]}) != indices.shape[${e}] (${t.shape[e]}).`);for(let t=0;t<n.rank-a;++t)if(n.shape[t+a]!==e[t+r])throw new Error(s+` updates.shape[${t+a}] (${n.shape[t+a]}) != shape[${t+a}] (${e[t+a]})`)}function zd(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("int32"!==t.dtype)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(0===n.length){if(0===t.size)throw new Error(`Indices specified for empty output. indices shape: ${t.shape}`);if(0===e.size)throw new Error(`Updates specified for empty output. updates shape: ${e.shape}`)}Pd(n,t,e)}function Bd(e,t,n){let r=t.shape.length,a=r>1?t.shape[r-1]:1,s=n.length,i=1;for(let e=a;e<s;++e)i*=n[e];let o=a<1?1:a;return{sliceRank:a,numUpdates:Y(t.shape)/o,sliceSize:i,strides:[...Ie(n.slice(0,a)),1],outputSize:Y(n)}}f(Ld,{calculateShapes:()=>Bd,validateInput:()=>zd,validateUpdateShape:()=>Pd});var Wd=zs({tensorScatterUpdate_:function(e,t,n){let r=Os(e,"tensor","tensorScatterupdate"),a=Os(t,"indices","tensorScatterupdate","int32"),s=Os(n,"updates","tensorScatterupdate");if(zd(s,a,r.shape),r.dtype!==s.dtype)throw new Error(`tensor and updates must have the same dtype, instead they are ${r.dtype} and ${s.dtype}.`);let i={tensor:r,indices:a,updates:s};return _s.runKernel(gr,i,{})}});var Vd=zs({topk_:function(e,t=1,n=!0){let r=Os(e,"x","topk");if(0===r.rank)throw new Error("topk() expects the input to be of rank 1 or higher");let a=r.shape[r.shape.length-1];if(t<0)throw new Error(`'k' passed to topk() must be >= 0 but got ${t}`);if(t>a)throw new Error(`'k' passed to topk() must be <= the last dimension (${a}) but got ${t}`);let s={x:r},i={k:t,sorted:n},[o,l]=_s.runKernel(qr,s,i);return{values:o,indices:l}}});var Ud=zs({truncatedNormal_:function(e,t=0,n=1,r,a){if(Ae(e),null!=r&&"bool"===r)throw new Error("Unsupported data type $ { dtype }");let s=new Uh(t,n,r,!0,a),i=uo(e,r);for(let e=0;e<i.values.length;e++)i.values[e]=s.nextValue();return i.toTensor()}});var Gd=zs({unique_:function(e,t=0){let n=Os(e,"x","unique","string_or_numeric");K(n.rank>0,()=>"The input tensor must be at least 1D");let r={x:n},a={axis:t},[s,i]=_s.runKernel(Zr,r,a);return{values:s,indices:i}}});var Hd=zs({unsortedSegmentSum_:function(e,t,n){let r=Os(e,"x","unsortedSegmentSum"),a=Os(t,"segmentIds","unsortedSegmentSum","int32");K(te(n),()=>"numSegments must be of dtype int");let s={x:r,segmentIds:a},i={numSegments:n};return _s.runKernel(Jr,s,i)}});var jd=zs({unstack_:function(e,t=0){let n=Os(e,"x","unstack","string_or_numeric");K(t>=-n.shape.length&&t<n.shape.length,()=>`Axis = ${t} is not in [-${n.shape.length}, ${n.shape.length})`);let r={value:n},a={axis:t};return _s.runKernel(Yr,r,a)}});function qd(e,t){return Xu(e,t,"right")}function Kd(e,t=!0,n,r){return _s.makeVariable(e,t,n,r)}function Xd(e,t){let n=[];for(let e=0;e<t.length;e++)t[e]&&n.push(e);let r=uo(e,"int32"),a=uo([n.length,e.length],"int32");for(let t=0;t<n.length;t++){let s=r.indexToLoc(n[t]),i=t*e.length;a.values.set(s,i)}return a.toTensor()}var Zd=async function(e){let t=Os(e,"condition","whereAsync","bool"),n=await t.data(),r=Xd(t.shape,n);return e!==t&&t.dispose(),r};var Yd=async function(e,t,n){let r=Os(e,"tensor","boolMask"),a=Os(t,"mask","boolMask","bool"),s=n??0,i=a.rank,o=r.shape;K(i>0,()=>"mask cannot be scalar"),X(o.slice(s,s+i),a.shape,"mask's shape must match the first K dimensions of tensor's shape,");let l=1;for(let e=s;e<s+i;e++)l*=o[e];let u=o.slice(0,s).concat([l],o.slice(s+i)),h=jo(r,u),d=jo(a,[-1]),p=await Zd(d),c=Nd(p,[1]),f=yu(h,c,s);return e!==r&&r.dispose(),t!==a&&a.dispose(),c.dispose(),h.dispose(),d.dispose(),p.dispose(),f};var Jd=zs({transpose_:function(e,t,n){let r=Os(e,"x","transpose");if(null==t&&(t=r.shape.map((e,t)=>t).reverse()),K(r.rank===t.length,()=>`Error in transpose: rank of input ${r.rank} must match length of perm ${t}.`),t.forEach(e=>{K(e>=0&&e<r.rank,()=>`All entries in 'perm' must be between 0 and ${r.rank-1} but got ${t}`)}),r.rank<=1)return r.clone();let a={x:r},s={perm:t};return"complex64"===r.dtype?Qs(()=>{let e=Jh(r),t=vu(r);return e=_s.runKernel(Xr,{x:e},s),t=_s.runKernel(Xr,{x:t},s),n&&(t=Pu(t)),Bs(e,t)}):_s.runKernel(Xr,a,s)}});var Qd=zs({movingAverage_:function(e,t,n,r,a=!0){let s=Os(e,"v","movingAverage"),i=Os(t,"x","movingAverage"),o=Os(n,"decay","movingAverage");ys(s,i),K(ee(s.shape,i.shape),()=>"Shape mismatch in v and x");let l=au(1),u=Wu(l,o),h=yo(Wu(i,s),u);if(a){K(null!=r,()=>"When using zeroDebias: true, step is required.");let e=Os(r,"step","movingAverage");h=go(h,Wu(l,ru(o,e)))}return fo(s,h)}});var ep=zs({scatterND_:function(e,t,n){Ae(n);let r=Os(e,"indices","scatterND","int32"),a=Os(t,"updates","scatterND");zd(a,r,n);let s={indices:r,updates:a},i={shape:n};return _s.runKernel(mr,s,i)}});var tp=zs({sparseToDense_:function(e,t,n,r=0){Ae(n);let a=Os(e,"sparseIndices","sparseToDense","int32"),s=Os(t,"sparseValues","sparseToDense","string_or_numeric"),i=Os(r,"defaultValue","sparseToDense",s.dtype);!function(e,t,n,r){if("int32"!==e.dtype)throw new Error(`tf.sparseToDense() expects the indices to be int32 type, but the dtype was ${e.dtype}.`);if(e.rank>2)throw new Error(`sparseIndices should be a scalar, vector, or matrix, but got shape ${e.shape}.`);let a=e.rank>0?e.shape[0]:1,s=e.rank>1?e.shape[1]:1;if(n.length!==s)throw new Error(`outputShape has incorrect number of elements:, ${n.length}, should be: ${s}.`);let i=t.size;if(0!==t.rank&&(1!==t.rank||i!==a))throw new Error(`sparseValues has incorrect shape ${t.shape}, should be [] or [${a}]`);if(t.dtype!==r.dtype)throw new Error("sparseValues.dtype must match defaultValues.dtype")}(a,s,n,i);let o={sparseIndices:a,sparseValues:s,defaultValue:i},l={outputShape:n};return _s.runKernel(Mr,o,l)}});var np=zs({gatherND_:function(e,t){let n=Os(t,"indices","gatherND","int32"),r={params:Os(e,"x","gatherND","string_or_numeric"),indices:n};return _s.runKernel(tn,r)}});var rp=zs({dropout_:function(e,t,n,r){let a=Os(e,"x","dropout");if(K("float32"===a.dtype,()=>`x has to be a floating point tensor since it's going to be scaled, but got a ${a.dtype} tensor instead.`),K(t>=0&&t<1,()=>`rate must be a float in the range [0, 1), but got ${t}.`),0===t)return e instanceof ts?a.clone():a;let s=function(e,t){if(null==t)return e.shape.slice();if(ee(e.shape,t))return t;if(e.shape.length===t.length){let n=[];for(let r=0;r<e.shape.length;r++)null==t[r]&&null!=e.shape[r]?n.push(e.shape[r]):n.push(t[r]);return n}return t}(a,n),i=1-t,o=go(gu(fo(Xh(s,0,1,"float32",r),i)),i);return yo(a,o)}});function ap(e){return Math.floor(Math.pow(2,Math.ceil(Math.log(e)/Math.log(2))))}function sp(e,t,n){let r=1-e%2,a=new Float32Array(e);for(let s=0;s<e;++s){let i=2*Math.PI*s/(e+r-1);a[s]=t-n*Math.cos(i)}return $d(a,"float32")}var ip=async function(e,t,n=1){let r=Os(e,"predictions","inTopK"),a=Os(t,"targets","inTopK");K(r.rank>1,()=>`inTopK() expects the predictions to be of rank 2 or higher, but got ${r.rank}`),K(r.rank-1===a.rank,()=>`predictions rank should be 1 larger than targets rank, but got predictions rank ${r.rank} and targets rank ${a.rank}`),X(r.shape.slice(0,r.shape.length-1),a.shape,"predictions's shape should be align with the targets' shape, except the last dimension.");let s=r.shape[r.shape.length-1];K(n>0&&n<=s,()=>`'k' passed to inTopK() must be > 0 && <= the predictions last dimension (${s}), but got ${n}`);let i=await r.data(),o=await a.data(),[l,u]=[i.length/s,s],h=he("bool",l);for(let e=0;e<l;e++){let t=e*u,r=i.subarray(t,t+u),a=[];for(let e=0;e<r.length;e++)a.push({value:r[e],index:e});a.sort((e,t)=>t.value-e.value),h[e]=0;for(let t=0;t<n;t++)if(a[t].index===o[e]){h[e]=1;break}}return e!==r&&r.dispose(),t!==a&&a.dispose(),Vs(h,a.shape,"bool")},op={};f(op,{conv2d:()=>cp,depthwiseConv2d:()=>gp,matMul:()=>yp});var lp=zs({conv2DBackpropFilter_:function(e,t,n,r,a,s="NHWC",i){let o=e;3===e.rank&&(o=jo(e,[1,e.shape[0],e.shape[1],e.shape[2]]));let l=t;3===l.rank&&(l=jo(t,[1,t.shape[0],t.shape[1],t.shape[2]])),K(4===o.rank,()=>`Error in conv2dDerFilter: input must be rank 4, but got shape ${o.shape}.`),K(4===l.rank,()=>`Error in conv2dDerFilter: dy must be rank 4, but got shape ${l.shape}.`),K(4===n.length,()=>`Error in conv2dDerFilter: filterShape must be length 4, but got ${n}.`);let u="NHWC"===s?o.shape[3]:o.shape[1],h="NHWC"===s?l.shape[3]:l.shape[1];K(u===n[2],()=>`Error in conv2dDerFilter: depth of input ${u}) must match input depth in filter (${n[2]}.`),K(h===n[3],()=>`Error in conv2dDerFilter: depth of dy (${h}) must match output depth for filter (${n[3]}).`),Ho("conv2dDerFilter",a,i);let d={x:o,dy:l},p={strides:r,pad:a,dataFormat:s,dimRoundingMode:i,filterShape:n};return _s.runKernel(xt,d,p)}});function up(e,t,n){if(null==n||"linear"===n)return e;if("relu"===n)return yo(e,Cd(t));throw new Error(`Cannot compute gradient for fused activation ${n}.`)}function hp(e,t){let n=t,r=Ol(e.shape,t.shape);return r.length>0&&(n=ou(n,r)),jo(n,e.shape)}function dp(e,t,n,r){if("linear"===t)return e;if("relu"===t)return ed(e);if("elu"===t)return Gl(e);if("relu6"===t)return td(e);if("prelu"===t)return kh(e,n);if("leakyrelu"===t)return Su(e,r);if("sigmoid"===t)return Yo(e);throw new Error(`Unknown fused activation ${t}.`)}var pp=(e,t)=>!(e>0)||"linear"===t;var cp=zs({fusedConv2d_:function({x:e,filter:t,strides:n,pad:r,dataFormat:a="NHWC",dilations:s=[1,1],dimRoundingMode:i,bias:o,activation:l="linear",preluActivationWeights:u,leakyreluAlpha:h}){if(l=l||"linear",!1===pp(_s.state.gradientDepth,l)){K("NHWC"===a,()=>`Error in fused conv2d: got dataFormat of ${a} but only NHWC is currently supported for the case of gradient depth is 0 and the activation is not linear.`);let d=bl(e,t,n,r,a,s,i);return null!=o&&(d=fo(d,o)),dp(d,l,u,h)}let d=Os(e,"x","conv2d","float32"),p=Os(t,"filter","conv2d","float32"),c=d,f=!1;3===d.rank&&(f=!0,c=jo(d,[1,d.shape[0],d.shape[1],d.shape[2]])),K(4===c.rank,()=>`Error in fused conv2d: input must be rank 4, but got rank ${c.rank}.`),K(4===p.rank,()=>`Error in fused conv2d: filter must be rank 4, but got rank ${p.rank}.`),Ho("fused conv2d",r,i);let m="NHWC"===a?c.shape[3]:c.shape[1];K(p.shape[2]===m,()=>`Error in conv2d: depth of input (${m}) must match input depth for filter ${p.shape[2]}.`),K(Vo(n,s),()=>`Error in conv2D: Either strides or dilations must be 1. Got strides ${n} and dilations '${s}'`);let g,y,b=Do(c.shape,p.shape,n,s,r,i);if(null!=o&&(g=Os(o,"bias","fused conv2d"),[g]=gs(g,d),"NHWC"===a?Ll(b.outShape,g.shape):(K(g.shape.length<=1,()=>`Error in fused conv2d: only supports scalar or 1-D Tensor bias for NCHW format but got the bias of rank-${g.shape.length}.`),K(0===g.shape.length||g.shape[0]===b.outChannels||1===g.shape[0],()=>`Error in fused conv2d: bias shape (${g.shape}) is not compatible with the number of output channels (${b.outChannels})`))),null!=u){let e=u.shape;if(K(e.length<=1||3===e.length,()=>`Error in fused conv2d: only supports scalar, 1-D Tensor or 3-D Tensor PReLU activation weights but got a tensor of rank-${e.length}.`),1===e.length)K(1===e[0]||e[0]===b.outChannels,()=>`Error in fused conv2d: PReLU activation weights (${e}) is not compatible with the number of output channels (${b.outChannels}).`);else if(3===e.length)try{Ll(e,b.outShape)}catch(t){let n=`Error in fused conv2d: PReLU activation weights (${e}) is not compatible with the output shape of the conv2d (${b.outShape}).`;throw Error(n)}y=Os(u,"prelu weights","fused conv2d")}let x=(e,t)=>{K("NHWC"===a,()=>`Error in gradient of fused conv2D: got dataFormat of ${a} but only NHWC is currently supported.`);let[i,o,u,h]=t,d=up(e,u,l);K(Wo(s),()=>`Error in gradient of fused conv2D: dilation rates greater than 1 are not yet supported in gradients. Got dilations '${s}'`);let p=[vl(o.shape,d,i,n,r),lp(o,d,i.shape,n,r)];if(null!=h){let e=hp(h,d);p.push(e)}return p},v={x:c,filter:p,bias:g,preluActivationWeights:y},w={strides:n,pad:r,dataFormat:a,dilations:s,dimRoundingMode:i,activation:l,leakyreluAlpha:h};return null==o?Ou((e,t,n)=>{let r=_s.runKernel(sa,v,w);return n([t,e,r]),f&&(r=jo(r,[r.shape[1],r.shape[2],r.shape[3]])),{value:r,gradFunc:x}})(c,p):Ou((e,t,n,r)=>{let a=_s.runKernel(sa,v,w);return r([t,e,a,n]),f&&(a=jo(a,[a.shape[1],a.shape[2],a.shape[3]])),{value:a,gradFunc:x}})(c,p,g)}});var fp=zs({depthwiseConv2dNativeBackpropFilter_:function(e,t,n,r,a,s=[1,1],i){let o=e;3===e.rank&&(o=jo(e,[1,e.shape[0],e.shape[1],e.shape[2]]));let l=t;3===l.rank&&(l=jo(t,[1,t.shape[0],t.shape[1],t.shape[2]]));let u={x:o,dy:l},h={strides:r,pad:a,dimRoundingMode:i,dilations:s,filterShape:n};return _s.runKernel(Rt,u,h)}});var mp=zs({depthwiseConv2dNativeBackpropInput_:function(e,t,n,r,a,s=[1,1],i){let o=t,l=!1;3===t.rank&&(l=!0,o=jo(t,[1,t.shape[0],t.shape[1],t.shape[2]]));let u={dy:o,filter:n},h={strides:r,pad:a,dimRoundingMode:i,dilations:s,inputShape:e},d=_s.runKernel(Ft,u,h);return l?jo(d,[d.shape[1],d.shape[2],d.shape[3]]):d}});var gp=zs({fusedDepthwiseConv2d_:function({x:e,filter:t,strides:n,pad:r,dataFormat:a="NHWC",dilations:s=[1,1],dimRoundingMode:i,bias:o,activation:l="linear",preluActivationWeights:u,leakyreluAlpha:h}){if(!1===pp(_s.state.gradientDepth,l)){let d=$l(e,t,n,r,a,s,i);return null!=o&&(d=fo(d,o)),dp(d,l,u,h)}let d=Os(e,"x","depthwiseConv2d","float32"),p=Os(t,"filter","depthwiseConv2d","float32"),c=d,f=!1;3===d.rank&&(f=!0,c=jo(d,[1,d.shape[0],d.shape[1],d.shape[2]])),K(4===c.rank,()=>`Error in fused depthwiseConv2d: input must be rank 4, but got rank ${c.rank}.`),K(4===p.rank,()=>`Error in fused depthwiseConv2d: filter must be rank 4, but got rank ${p.rank}.`),K(c.shape[3]===p.shape[2],()=>`Error in fused depthwiseConv2d: number of input channels (${c.shape[3]}) must match the inChannels dimension in filter ${p.shape[2]}.`),null==s&&(s=[1,1]),K(Vo(n,s),()=>`Error in fused depthwiseConv2d: Either strides or dilations must be 1. Got strides ${n} and dilations '${s}'`),Ho("fused depthwiseConv2d",r,i);let m,g,y=Do(c.shape,p.shape,n,s,r,i,!0);null!=o&&(m=Os(o,"bias","fused conv2d"),[m]=gs(m,d),Ll(y.outShape,m.shape)),null!=u&&(g=Os(u,"prelu weights","fused depthwiseConv2d"));let b=(e,t)=>{K(Wo(s),()=>`Error in gradient of fused depthwiseConv2d: dilation rates greater than 1 are not yet supported. Got dilations '${s}'`);let[a,o,u,h]=t,d=up(e,u,l),p=mp(o.shape,d,a,n,r,s,i),c=fp(o,d,a.shape,n,r,s,i);if(null!=h){return[p,c,hp(m,d)]}return[p,c]},x={x:c,filter:p,bias:m,preluActivationWeights:g},v={strides:n,pad:r,dataFormat:a,dilations:s,dimRoundingMode:i,activation:l,leakyreluAlpha:h};return null==o?Ou((e,t,n)=>{let r=_s.runKernel(ia,x,v);return n([t,e,r]),f&&(r=jo(r,[r.shape[1],r.shape[2],r.shape[3]])),{value:r,gradFunc:b}})(c,p):Ou((e,t,n,r)=>{let a=_s.runKernel(ia,x,v);return r([t,e,a,n]),f&&(a=jo(a,[a.shape[1],a.shape[2],a.shape[3]])),{value:a,gradFunc:b}})(c,p,m)}});var yp=zs({fusedMatMul_:function({a:e,b:t,transposeA:n=!1,transposeB:r=!1,bias:a,activation:s="linear",preluActivationWeights:i,leakyreluAlpha:o=.2}){if(!1===pp(_s.state.gradientDepth,s)){let l=Zo(e,t,n,r);return null!=a&&(l=fo(l,a)),dp(l,s,i,o)}let l=Os(e,"a","fused matMul"),u=Os(t,"b","fused matMul");[l,u]=gs(l,u);let h=n?l.shape[l.rank-2]:l.shape[l.rank-1],d=r?u.shape[u.rank-1]:u.shape[u.rank-2],p=n?l.shape[l.rank-1]:l.shape[l.rank-2],c=r?u.shape[u.rank-2]:u.shape[u.rank-1],f=l.shape.slice(0,-2),m=u.shape.slice(0,-2),g=Y(f),y=Y(m);K(h===d,()=>`Error in fused matMul: inner shapes (${h}) and (${d}) of Tensors with shapes ${l.shape} and ${u.shape} and transposeA=${n} and transposeB=${r} must match.`);let b,x,v=Ll(l.shape.slice(0,-2),u.shape.slice(0,-2)).concat([p,c]),w=jo(l,n?[g,h,p]:[g,p,h]),k=jo(u,r?[y,c,d]:[y,d,c]);null!=a&&(b=Os(a,"bias","fused matMul"),[b]=gs(b,l),Ll(v,b.shape)),null!=i&&(x=Os(i,"prelu weights","fused matMul"));let I=(e,t)=>{let i,o,[l,u,h,d]=t,p=up(jo(e,h.shape),h,s);if(n||r?!n&&r?(i=Zo(p,u,!1,!1),o=Zo(p,l,!0,!1)):n&&!r?(i=Zo(u,p,!1,!0),o=Zo(l,p,!1,!1)):(i=Zo(u,p,!0,!0),o=Zo(p,l,!0,!0)):(i=Zo(p,u,!1,!0),o=Zo(l,p,!0,!1)),null!=a){return[i,o,hp(d,p)]}return[i,o]},S={a:w,b:k,bias:b,preluActivationWeights:x},_={transposeA:n,transposeB:r,activation:s,leakyreluAlpha:o};return null==a?Ou((e,t,n)=>{let r=_s.runKernel(aa,S,_);return n([e,t,r]),{value:jo(r,v),gradFunc:I}})(w,k):Ou((e,t,n,r)=>{let a=_s.runKernel(aa,S,_);return r([e,t,a,n]),{value:jo(a,v),gradFunc:I}})(w,k,b)}});var bp=zs({hammingWindow_:function(e){return sp(e,.54,.46)}});var xp=zs({hannWindow_:function(e){return sp(e,.5,.5)}});var vp=zs({frame_:function(e,t,n,r=!1,a=0){let s=0,i=[];for(;s+t<=e.size;)i.push(Jo(e,s,t)),s+=n;if(r)for(;s<e.size;){let r=s+t-e.size,o=Xo([Jo(e,s,t-r),pl([r],a)]);i.push(o),s+=n}return 0===i.length?Rd([],[0,t]):jo(Xo(i),[i.length,t])}});var wp=zs({stft_:function(e,t,n,r,a=xp){null==r&&(r=ap(t));let s=vp(e,t,n),i=yo(s,a(t));return Sd(i,r)}});var kp=zs({cropAndResize_:function(e,t,n,r,a="bilinear",s=0){let i=Os(e,"image","cropAndResize"),o=Os(t,"boxes","cropAndResize","float32"),l=Os(n,"boxInd","cropAndResize","int32"),u=o.shape[0];K(4===i.rank,()=>`Error in cropAndResize: image must be rank 4,but got rank ${i.rank}.`),K(2===o.rank&&4===o.shape[1],()=>`Error in cropAndResize: boxes must be have size [${u},4] but had shape ${o.shape}.`),K(1===l.rank&&l.shape[0]===u,()=>`Error in cropAndResize: boxInd must be have size [${u}] but had shape ${o.shape}.`),K(2===r.length,()=>`Error in cropAndResize: cropSize must be of length 2, but got length ${r.length}.`),K(r[0]>=1&&r[1]>=1,()=>`cropSize must be atleast [1,1], but was ${r}`),K("bilinear"===a||"nearest"===a,()=>`method must be bilinear or nearest, but was ${a}`);let h={image:i,boxes:o,boxInd:l},d={method:a,extrapolationValue:s,cropSize:r};return _s.runKernel(Ct,h,d)}});var Ip=zs({flipLeftRight_:function(e){let t=Os(e,"image","flipLeftRight","float32");K(4===t.rank,()=>`Error in flipLeftRight: image must be rank 4,but got rank ${t.rank}.`);let n={image:t};return _s.runKernel(Zt,n,{})}});var Sp=zs({grayscaleToRGB_:function(e){let t=Os(e,"image","grayscaleToRGB"),n=t.rank-1,r=t.shape[n];K(t.rank>=2,()=>`Error in grayscaleToRGB: images must be at least rank 2, but got rank ${t.rank}.`),K(1===r,()=>`Error in grayscaleToRGB: last dimension of a grayscale image should be size 1, but got size ${r}.`);let a=new Array(t.rank);return a.fill(1,0,n),a[n]=3,fu(t,a)}});var _p=zs({rgbToGrayscale_:function(e){let t=Os(e,"image","RGBToGrayscale"),n=t.rank-1,r=t.shape[n];K(t.rank>=2,()=>`Error in RGBToGrayscale: images must be at least rank 2, but got rank ${t.rank}.`),K(3===r,()=>`Error in RGBToGrayscale: last dimension of an RGB image should be size 3, but got size ${r}.`);let a,s=t.dtype,i=ho(t,"float32"),o=$d([.2989,.587,.114]);switch(t.rank){case 2:a=Ul("ij,j->i",i,o);break;case 3:a=Ul("ijk,k->ij",i,o);break;case 4:a=Ul("ijkl,l->ijk",i,o);break;case 5:a=Ul("ijklm,m->ijkl",i,o);break;case 6:a=Ul("ijklmn,n->ijklm",i,o);break;default:throw new Error("Not a valid tensor rank.")}return a=pu(a,-1),ho(a,s)}});var Np=zs({rotateWithOffset_:function(e,t,n=0,r=.5){let a=Os(e,"image","rotateWithOffset","float32");K(4===a.rank,()=>`Error in rotateWithOffset: image must be rank 4,but got rank ${a.rank}.`);let s={image:a},i={radians:t,fillValue:n,center:r};return _s.runKernel(ra,s,i)}});function Tp(e,t,n,r,a,s){null==r&&(r=.5),null==a&&(a=Number.NEGATIVE_INFINITY),null==s&&(s=0);let i=e.shape[0];return n=Math.min(n,i),K(0<=r&&r<=1,()=>`iouThreshold must be in [0, 1], but was '${r}'`),K(2===e.rank,()=>`boxes must be a 2D tensor, but was of rank '${e.rank}'`),K(4===e.shape[1],()=>`boxes must have 4 columns, but 2nd dimension was ${e.shape[1]}`),K(1===t.rank,()=>"scores must be a 1D tensor"),K(t.shape[0]===i,()=>`scores has incompatible shape with boxes. Expected ${i}, but was ${t.shape[0]}`),K(0<=s&&s<=1,()=>`softNmsSigma must be in [0, 1], but was '${s}'`),{maxOutputSize:n,iouThreshold:r,scoreThreshold:a,softNmsSigma:s}}var Cp=zs({nonMaxSuppression_:function(e,t,n,r=.5,a=Number.NEGATIVE_INFINITY){let s=Os(e,"boxes","nonMaxSuppression","float32"),i=Os(t,"scores","nonMaxSuppression","float32"),o=Tp(s,i,n,r,a),l={maxOutputSize:n=o.maxOutputSize,iouThreshold:r=o.iouThreshold,scoreThreshold:a=o.scoreThreshold};return _s.runKernel(Vn,{boxes:s,scores:i},l)}});function Ep(e,t,n){let r=function(e,t,n){return function(e,t,n){let r=0,a=e.length,s=0,i=!1;for(;r<a;){s=r+(a-r>>>1);let o=n(t,e[s]);o>0?r=s+1:(a=s,i=!o)}return i?r:-r-1}(e,t,n||Ap)}(e,t,n),a=r<0?-(r+1):r;e.splice(a,0,t)}function Ap(e,t){return e>t?1:e<t?-1:0}function $p(e,t,n,r,a){return Dp(e,t,n,r,a,0)}function Rp(e,t,n,r,a,s){return Dp(e,t,n,r,a,0,!1,s,!0)}function Fp(e,t,n,r,a,s){return Dp(e,t,n,r,a,s,!0)}function Dp(e,t,n,r,a,s,i=!1,o=!1,l=!1){let u=[];for(let e=0;e<t.length;e++)t[e]>a&&u.push({score:t[e],boxIndex:e,suppressBeginIndex:0});u.sort(Lp);let h=s>0?-.5/s:0,d=[],p=[];for(;d.length<n&&u.length>0;){let t=u.pop(),{score:n,boxIndex:s,suppressBeginIndex:i}=t;if(n<a)break;let o=!1;for(let n=d.length-1;n>=i;--n){let i=Mp(e,s,d[n]);if(i>=r){o=!0;break}if(t.score=t.score*Op(r,h,i),t.score<=a)break}t.suppressBeginIndex=d.length,o||(t.score===n?(d.push(s),p.push(t.score)):t.score>a&&Ep(u,t,Lp))}let c=d.length,f=n-c;o&&f>0&&(d.push(...new Array(f).fill(0)),p.push(...new Array(f).fill(0)));let m={selectedIndices:d};return i&&(m.selectedScores=p),l&&(m.validOutputs=c),m}function Mp(e,t,n){let r=e.subarray(4*t,4*t+4),a=e.subarray(4*n,4*n+4),s=Math.min(r[0],r[2]),i=Math.min(r[1],r[3]),o=Math.max(r[0],r[2]),l=Math.max(r[1],r[3]),u=Math.min(a[0],a[2]),h=Math.min(a[1],a[3]),d=Math.max(a[0],a[2]),p=Math.max(a[1],a[3]),c=(o-s)*(l-i),f=(d-u)*(p-h);if(c<=0||f<=0)return 0;let m=Math.max(s,u),g=Math.max(i,h),y=Math.min(o,d),b=Math.min(l,p),x=Math.max(y-m,0)*Math.max(b-g,0);return x/(c+f-x)}function Op(e,t,n){let r=Math.exp(t*n*n);return n<=e?r:0}function Lp(e,t){return e.score-t.score||e.score===t.score&&t.boxIndex-e.boxIndex}var Pp=async function(e,t,n,r=.5,a=Number.NEGATIVE_INFINITY){let s=Os(e,"boxes","nonMaxSuppressionAsync"),i=Os(t,"scores","nonMaxSuppressionAsync"),o=Tp(s,i,n,r,a);n=o.maxOutputSize,r=o.iouThreshold,a=o.scoreThreshold;let l=await Promise.all([s.data(),i.data()]),u=l[0],h=l[1],{selectedIndices:d}=$p(u,h,n,r,a);return s!==e&&s.dispose(),i!==t&&i.dispose(),$d(d,"int32")};var zp=zs({nonMaxSuppressionWithScore_:function(e,t,n,r=.5,a=Number.NEGATIVE_INFINITY,s=0){let i=Os(e,"boxes","nonMaxSuppression"),o=Os(t,"scores","nonMaxSuppression"),l=Tp(i,o,n,r,a,s),u={boxes:i,scores:o},h={maxOutputSize:n=l.maxOutputSize,iouThreshold:r=l.iouThreshold,scoreThreshold:a=l.scoreThreshold,softNmsSigma:s=l.softNmsSigma},d=_s.runKernel(Gn,u,h);return{selectedIndices:d[0],selectedScores:d[1]}}});var Bp=async function(e,t,n,r=.5,a=Number.NEGATIVE_INFINITY,s=0){let i=Os(e,"boxes","nonMaxSuppressionAsync"),o=Os(t,"scores","nonMaxSuppressionAsync"),l=Tp(i,o,n,r,a,s);n=l.maxOutputSize,r=l.iouThreshold,a=l.scoreThreshold,s=l.softNmsSigma;let u=await Promise.all([i.data(),o.data()]),h=u[0],d=u[1],{selectedIndices:p,selectedScores:c}=Fp(h,d,n,r,a,s);return i!==e&&i.dispose(),o!==t&&o.dispose(),{selectedIndices:$d(p,"int32"),selectedScores:$d(c)}};var Wp=zs({nonMaxSuppressionPadded_:function(e,t,n,r=.5,a=Number.NEGATIVE_INFINITY,s=!1){let i=Os(e,"boxes","nonMaxSuppression"),o=Os(t,"scores","nonMaxSuppression"),l=Tp(i,o,n,r,a,null),u={boxes:i,scores:o},h={maxOutputSize:l.maxOutputSize,iouThreshold:l.iouThreshold,scoreThreshold:l.scoreThreshold,padToMaxOutputSize:s},d=_s.runKernel(Un,u,h);return{selectedIndices:d[0],validOutputs:d[1]}}});var Vp=async function(e,t,n,r=.5,a=Number.NEGATIVE_INFINITY,s=!1){let i=Os(e,"boxes","nonMaxSuppressionAsync"),o=Os(t,"scores","nonMaxSuppressionAsync"),l=Tp(i,o,n,r,a,null),u=l.maxOutputSize,h=l.iouThreshold,d=l.scoreThreshold,[p,c]=await Promise.all([i.data(),o.data()]),{selectedIndices:f,validOutputs:m}=Rp(p,c,u,h,d,s);return i!==e&&i.dispose(),o!==t&&o.dispose(),{selectedIndices:$d(f,"int32"),validOutputs:au(m,"int32")}};var Up=zs({resizeBilinear_:function(e,t,n=!1,r=!1){let a=Os(e,"images","resizeBilinear");K(3===a.rank||4===a.rank,()=>`Error in resizeBilinear: x must be rank 3 or 4, but got rank ${a.rank}.`),K(2===t.length,()=>`Error in resizeBilinear: new shape must 2D, but got shape ${t}.`),K(!1===r||!1===n,()=>"Error in resizeBilinear: If halfPixelCenters is true, alignCorners must be false.");let s=a,i=!1;3===a.rank&&(i=!0,s=jo(a,[1,a.shape[0],a.shape[1],a.shape[2]]));let[]=t,o={images:s},l={alignCorners:n,halfPixelCenters:r,size:t},u=_s.runKernel(ur,o,l);return i?jo(u,[u.shape[1],u.shape[2],u.shape[3]]):u}});var Gp=zs({resizeNearestNeighbor_:function(e,t,n=!1,r=!1){let a=Os(e,"images","resizeNearestNeighbor");K(3===a.rank||4===a.rank,()=>`Error in resizeNearestNeighbor: x must be rank 3 or 4, but got rank ${a.rank}.`),K(2===t.length,()=>`Error in resizeNearestNeighbor: new shape must 2D, but got shape ${t}.`),K("float32"===a.dtype||"int32"===a.dtype,()=>"`images` must have `int32` or `float32` as dtype"),K(!1===r||!1===n,()=>"Error in resizeNearestNeighbor: If halfPixelCenters is true, alignCorners must be false.");let s=a,i=!1;3===a.rank&&(i=!0,s=jo(a,[1,a.shape[0],a.shape[1],a.shape[2]]));let[]=t,o={images:s},l={alignCorners:n,halfPixelCenters:r,size:t},u=_s.runKernel(or,o,l);return i?jo(u,[u.shape[1],u.shape[2],u.shape[3]]):u}});var Hp=zs({threshold_:function(e,t="binary",n=!1,r=.5){let a,s,i,o,l=Os(e,"image","threshold"),u=l.shape[0]*l.shape[1],h=yo($d([r]),255);if(K(3===l.rank,()=>`Error in threshold: image must be rank 3,but got rank ${l.rank}.`),K(3===l.shape[2]||1===l.shape[2],()=>`Error in threshold: image color channel must be equal to 3 or 1but got ${l.shape[2]}.`),K("int32"===l.dtype||"float32"===l.dtype,()=>`Error in dtype: image dtype must be int32 or float32,but got dtype ${l.dtype}.`),K("otsu"===t||"binary"===t,()=>`Method must be binary or otsu, but was ${t}`),3===l.shape[2]){[a,s,i]=Id(l,[1,1,1],-1);let e=yo(a,.2989),t=yo(s,.587),n=yo(i,.114);o=fo(fo(e,t),n)}else o=e;if("otsu"===t){h=function(e,t){let n,r,a,s,i,o,l=$d([-1]),u=$d([0]),h=$d([0]);for(let d=0;d<e.size-1;d++){n=Jo(e,0,d+1),r=Jo(e,d+1),i=go(ou(n),t),o=go(ou(r),t);let p=ou(yo(n,Yh(0,n.size)));a=go(p,ou(n));let c=pl(r.shape,n.size),f=fo(Yh(0,r.size),c),m=yo(r,f);s=go(ou(m),ou(r));let g=Wu(a,s),y=Wu(a,s),b=yo(i,o);h=yo(yo(b,g),y);let x=bu(h,u);u=zl(x,h,u),l=zl(x,$d([d]),l)}return l}(ol(ho(od(o),"int32"),Vs([]),256),u)}let d=n?Nu(o,h):bu(o,h);return ho(yo(d,255),"int32")}});var jp=zs({transform_:function(e,t,n="nearest",r="constant",a=0,s){let i=Os(e,"image","transform","float32"),o=Os(t,"transforms","transform","float32");K(4===i.rank,()=>`Error in transform: image must be rank 4,but got rank ${i.rank}.`),K(2===o.rank&&(o.shape[0]===i.shape[0]||1===o.shape[0])&&8===o.shape[1],()=>"Error in transform: Input transform should be batch x 8 or 1 x 8"),K(null==s||2===s.length,()=>`Error in transform: outputShape must be [height, width] or null, but got ${s}.`);let l={image:i,transforms:o},u={interpolation:n,fillMode:r,fillValue:a,outputShape:s};return _s.runKernel(Kr,l,u)}});var qp=zs({bandPart_:function(e,t,n){let r=Os(e,"a","bandPart");K(r.rank>=2,()=>`bandPart(): Rank must be at least 2, got ${r.rank}.`);let a,s,i=r.shape,[o,l]=r.shape.slice(-2);"number"==typeof t?(K(t%1==0,()=>`bandPart(): numLower must be an integer, got ${t}.`),K(t<=o,()=>`bandPart(): numLower (${t}) must not be greater than the number of rows (${o}).`),a=Os(t<0?o:t,"numLower","bandPart")):(K("int32"===t.dtype,()=>"bandPart(): numLower's dtype must be an int32."),a=zl(_u(t,0),o,sh(t,o))),"number"==typeof n?(K(n%1==0,()=>`bandPart(): numUpper must be an integer, got ${n}.`),K(n<=l,()=>`bandPart(): numUpper (${n}) must not be greater than the number of columns (${l}).`),s=Os(n<0?l:n,"numUpper","bandPart")):(K("int32"===n.dtype,()=>"bandPart(): numUpper's dtype must be an int32."),s=zl(_u(n,0),l,sh(n,l)));let u=jo(Yh(0,o,1,"int32"),[-1,1]),h=Yh(0,l,1,"int32"),d=Wu(u,h),p=Gu(Nu(d,a),xu(d,Pu(s))),c=nh([o,l],r.dtype);return jo(Td(jd(jo(r,[-1,o,l])).map(e=>zl(p,e,c))),i)}});var Kp=zs({gramSchmidt_:function(e){let t;if(Array.isArray(e)){t=!1,K(null!=e&&e.length>0,()=>"Gram-Schmidt process: input must not be null, undefined, or empty");let n=e[0].shape[0];for(let t=1;t<e.length;++t)K(e[t].shape[0]===n,()=>`Gram-Schmidt: Non-unique lengths found in the input vectors: (${e[t].shape[0]} vs. ${n})`)}else t=!0,e=Id(e,e.shape[0],0).map(e=>Nd(e,[0]));K(e.length<=e[0].shape[0],()=>`Gram-Schmidt: Number of vectors (${e.length}) exceeds number of dimensions (${e[0].shape[0]}).`);let n=[],r=e;for(let t=0;t<e.length;++t)n.push(_s.tidy(()=>{let e=r[t];if(t>0)for(let r=0;r<t;++r){let t=yo(ou(yo(n[r],e)),n[r]);e=Wu(e,t)}return go(e,uu(e,"euclidean"))}));return t?Td(n,0):n}});function Xp(e,t=!1){return _s.tidy(()=>{K(2===e.shape.length,()=>`qr2d() requires a 2D Tensor, but got a ${e.shape.length}D Tensor.`);let n=e.shape[0],r=e.shape[1],a=mu(n),s=po(e),i=Rd([[1]],[1,1]),o=po(i),l=n>=r?r:n;for(let e=0;e<l;++e){let t=s,l=o,u=a;[o,s,a]=_s.tidy(()=>{let t=Jo(s,[e,e],[n-e,1]),l=uu(t),u=Jo(s,[e,e],[1,1]),h=zl(bu(u,0),Rd([[-1]]),Rd([[1]])),d=Wu(u,yo(h,l)),p=go(t,d);o=1===p.shape[0]?po(i):Xo([i,Jo(p,[1,0],[p.shape[0]-1,p.shape[1]])],0);let c=Pu(go(Zo(h,d),l)),f=Jo(s,[e,0],[n-e,r]),m=yo(c,o),g=Jd(o);if(0===e)s=Wu(f,Zo(m,Zo(g,f)));else{let t=Wu(f,Zo(m,Zo(g,f)));s=Xo([Jo(s,[0,0],[e,r]),t],0)}let y=Jd(m),b=Jo(a,[0,e],[n,a.shape[1]-e]);if(0===e)a=Wu(b,Zo(Zo(b,o),y));else{let t=Wu(b,Zo(Zo(b,o),y));a=Xo([Jo(a,[0,0],[n,e]),t],1)}return[o,s,a]}),ei([t,l,u])}return!t&&n>r&&(a=Jo(a,[0,0],[n,r]),s=Jo(s,[0,0],[r,r])),[a,s]})}var Zp,Yp=zs({qr_:function(e,t=!1){if(K(e.rank>=2,()=>`qr() requires input tensor to have a rank >= 2, but got rank ${e.rank}`),2===e.rank)return Xp(e,t);{let n=e.shape.slice(0,e.shape.length-2).reduce((e,t)=>e*t),r=jd(jo(e,[n,e.shape[e.shape.length-2],e.shape[e.shape.length-1]]),0),a=[],s=[];return r.forEach(e=>{let[n,r]=Xp(e,t);a.push(n),s.push(r)}),[jo(Td(a,0),e.shape),jo(Td(s,0),e.shape)]}}});!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"}(Zp||(Zp={}));var Jp=zs({computeWeightedLoss_:function(e,t,n=Zp.SUM_BY_NONZERO_WEIGHTS){let r=Os(e,"losses","computeWeightedLoss"),a=null;null!=t&&(a=Os(t,"weights","computeWeightedLoss"));let s=null==a?r:yo(r,a);if(n===Zp.NONE)return s;if(n===Zp.SUM)return ou(s);if(n===Zp.MEAN){if(null==a)return th(s);{let e=r.size/a.size,t=go(ou(s),ou(a));return e>1?go(t,au(e)):t}}if(n===Zp.SUM_BY_NONZERO_WEIGHTS){if(null==a)return go(ou(s),au(r.size));{let e=yo(a,rh(r.shape)),t=ho(ou(dh(e,au(0))),"float32");return go(ou(s),t)}}throw Error(`Unknown reduction: ${n}`)}});var Qp=zs({absoluteDifference_:function(e,t,n,r=Zp.SUM_BY_NONZERO_WEIGHTS){let a=Os(e,"labels","absoluteDifference"),s=Os(t,"predictions","absoluteDifference"),i=null;null!=n&&(i=Os(n,"weights","absoluteDifference")),X(a.shape,s.shape,"Error in absoluteDifference: ");let o=bo(Wu(a,s));return Jp(o,i,r)}});var ec=zs({cosineDistance_:function(e,t,n,r,a=Zp.SUM_BY_NONZERO_WEIGHTS){let s=Os(e,"labels","cosineDistance"),i=Os(t,"predictions","cosineDistance"),o=null;null!=r&&(o=Os(r,"weights","cosineDistance")),X(s.shape,i.shape,"Error in cosineDistance: ");let l=au(1),u=Wu(l,ou(yo(s,i),n,!0));return Jp(u,o,a)}});var tc=zs({hingeLoss_:function(e,t,n,r=Zp.SUM_BY_NONZERO_WEIGHTS){let a=Os(e,"labels","hingeLoss"),s=Os(t,"predictions","hingeLoss"),i=null;null!=n&&(i=Os(n,"weights","hingeLoss")),X(a.shape,s.shape,"Error in hingeLoss: ");let o=au(1);a=Wu(yo(au(2),a),o);let l=ed(Wu(o,yo(a,s)));return Jp(l,i,r)}});var nc=zs({huberLoss_:function(e,t,n,r=1,a=Zp.SUM_BY_NONZERO_WEIGHTS){let s=Os(e,"labels","huberLoss"),i=Os(t,"predictions","huberLoss"),o=null;null!=n&&(o=Os(n,"weights","huberLoss")),X(s.shape,i.shape,"Error in huberLoss: ");let l=au(r),u=bo(Wu(i,s)),h=sh(u,l),d=Wu(u,h),p=fo(yo(au(.5),iu(h)),yo(l,d));return Jp(p,o,a)}});var rc=zs({logLoss_:function(e,t,n,r=1e-7,a=Zp.SUM_BY_NONZERO_WEIGHTS){let s=Os(e,"labels","logLoss"),i=Os(t,"predictions","logLoss"),o=null;null!=n&&(o=Os(n,"weights","logLoss")),X(s.shape,i.shape,"Error in logLoss: ");let l=au(1),u=au(r),h=Pu(yo(s,Eu(fo(i,u)))),d=yo(Wu(l,s),Eu(fo(Wu(l,i),u))),p=Wu(h,d);return Jp(p,o,a)}});var ac=zs({meanSquaredError_:function(e,t,n,r=Zp.SUM_BY_NONZERO_WEIGHTS){let a=Os(e,"labels","meanSquaredError"),s=Os(t,"predictions","meanSquaredError"),i=null;null!=n&&(i=Os(n,"weights","meanSquaredError")),X(a.shape,s.shape,"Error in meanSquaredError: ");let o=_d(a,s);return Jp(o,i,r)}});var sc=zs({sigmoidCrossEntropy_:function(e,t,n,r=0,a=Zp.SUM_BY_NONZERO_WEIGHTS){let s=Os(e,"multiClassLabels","sigmoidCrossEntropy"),i=Os(t,"logits","sigmoidCrossEntropy"),o=null;if(null!=n&&(o=Os(n,"weights","sigmoidCrossEntropy")),X(s.shape,i.shape,"Error in sigmoidCrossEntropy: "),r>0){let e=au(r),t=au(1),n=au(.5);s=fo(yo(s,Wu(t,e)),yo(n,e))}let l=function(e,t){let n=Os(e,"labels","sigmoidCrossEntropyWithLogits"),r=Os(t,"logits","sigmoidCrossEntropyWithLogits");X(n.shape,r.shape,"Error in sigmoidCrossEntropyWithLogits: ");let a=ed(r),s=yo(r,n),i=Au(du(Pu(bo(r))));return fo(Wu(a,s),i)}(s,i);return Jp(l,o,a)}});var ic=zs({softmaxCrossEntropy_:function(e,t,n,r=0,a=Zp.SUM_BY_NONZERO_WEIGHTS){let s=Os(e,"onehotLabels","softmaxCrossEntropy"),i=Os(t,"logits","softmaxCrossEntropy"),o=null;if(null!=n&&(o=Os(n,"weights","softmaxCrossEntropy")),X(s.shape,i.shape,"Error in softmaxCrossEntropy: "),r>0){let e=au(r),t=au(1),n=au(s.shape[1]);s=fo(yo(s,Wu(t,e)),go(e,n))}let l=function(e,t,n=-1){if(-1===n&&(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}`);return Ou((e,t,r)=>{let a=Uu(t,[n],!0),s=Wu(ho(t,"float32"),a);r([e,s]);let i=Pu(yo(s,e));return{value:ou(i,[n]),gradFunc:(e,t)=>{let[r,a]=t,s=Zl(e.shape,[n]);return[yo(jo(e,s),Wu(ho(r,"float32"),du(a))),yo(jo(e,s),Wu(du(a),ho(r,"float32")))]}}})(e,t)}(s,i);return Jp(l,o,a)}});var oc=zs({sparseFillEmptyRows_:function(e,t,n,r){let a=Os(e,"indices","sparseFillEmptyRows","int32"),s=Os(t,"values","sparseFillEmptyRows"),i=Os(n,"denseShape","sparseFillEmptyRows","int32"),o=Os(r,"defaultValue","sparseFillEmptyRows",s.dtype);if(2!==a.rank)throw new Error(`Indices should be Tensor2D but received shape\n        ${a.shape}`);if(1!==s.rank)throw new Error(`Values should be Tensor1D but received shape ${s.shape}`);if(1!==i.rank)throw new Error(`Dense shape should be Tensor1D but received shape ${i.shape}`);if(0!==o.rank)throw new Error(`Default value should be a scalar but received shape ${o.shape}`);let l={indices:a,values:s,denseShape:i,defaultValue:o},u=_s.runKernel($r,l);return{outputIndices:u[0],outputValues:u[1],emptyRowIndicator:u[2],reverseIndexMap:u[3]}}});var lc=zs({sparseReshape_:function(e,t,n){let r=Os(e,"inputIndices","sparseReshape","int32"),a=Os(t,"inputShape","sparseReshape","int32"),s=Os(n,"newShape","sparseReshape","int32");if(2!==r.rank)throw new Error(`Input indices should be Tensor2D but received shape\n        ${r.shape}`);if(1!==a.rank)throw new Error(`Input shape should be Tensor1D but received shape ${a.shape}`);if(1!==s.rank)throw new Error(`New shape should be Tensor1D but received shape ${s.shape}`);let i={inputIndices:r,inputShape:a,newShape:s},o=_s.runKernel(Rr,i);return{outputIndices:o[0],outputShape:o[1]}}});var uc=zs({sparseSegmentMean_:function(e,t,n){let r=Os(e,"data","sparseSegmentMean"),a=Os(t,"indices","sparseSegmentMean","int32"),s=Os(n,"segmentIds","sparseSegmentMean","int32");if(r.rank<1)throw new Error("Data should be at least 1 dimensional but received scalar");if(1!==a.rank)throw new Error(`Indices should be Tensor1D but received shape\n          ${a.shape}`);if(1!==s.rank)throw new Error(`Segment ids should be Tensor1D but received shape\n          ${s.shape}`);let i={data:r,indices:a,segmentIds:s};return _s.runKernel(Fr,i)}});var hc=zs({sparseSegmentSum_:function(e,t,n){let r=Os(e,"data","sparseSegmentSum"),a=Os(t,"indices","sparseSegmentSum","int32"),s=Os(n,"segmentIds","sparseSegmentSum","int32");if(r.rank<1)throw new Error("Data should be at least 1 dimensional but received scalar");if(1!==a.rank)throw new Error(`Indices should be Tensor1D but received shape\n         ${a.shape}`);if(1!==s.rank)throw new Error(`Segment ids should be Tensor1D but received shape\n         ${s.shape}`);let i={data:r,indices:a,segmentIds:s};return _s.runKernel(Dr,i)}});var dc=zs({stringNGrams_:function(e,t,n,r,a,s,i,o){let l=Os(e,"data","stringNGrams","string");if("string"!==l.dtype)throw new Error("Data must be of datatype string");if(1!==l.shape.length)throw new Error(`Data must be a vector, saw: ${l.shape}`);let u=Os(t,"dataSplits","stringNGrams");if("int32"!==u.dtype)throw new Error("Data splits must be of datatype int32");let h={separator:n,nGramWidths:r,leftPad:a,rightPad:s,padWidth:i,preserveShortSequences:o},d={data:l,dataSplits:u},p=_s.runKernel(Br,d,h);return{nGrams:p[0],nGramsSplits:p[1]}}});var pc=zs({stringSplit_:function(e,t,n=!0){let r=Os(e,"input","stringSplit","string"),a=Os(t,"delimiter","stringSplit","string");if(1!==r.rank)throw new Error(`Input should be Tensor1D but received shape ${r.shape}`);if(0!==a.rank)throw new Error(`Delimiter should be a scalar but received shape ${a.shape}`);let s={skipEmpty:n},i={input:r,delimiter:a},o=_s.runKernel(Wr,i,s);return{indices:o[0],values:o[1],shape:o[2]}}});var cc=zs({stringToHashBucketFast_:function(e,t){let n=Os(e,"input","stringToHashBucketFast","string"),r={numBuckets:t};if(t<=0)throw new Error("Number of buckets must be at least 1");let a={input:n};return _s.runKernel(Vr,a,r)}});var fc=zs({staticRegexReplace_:function(e,t,n,r=!0){let a=Os(e,"input","staticRegexReplace","string"),s={pattern:t,rewrite:n,replaceGlobal:r};return _s.runKernel(Pr,{x:a},s)}}),mc={fft:vd,ifft:wd,rfft:Sd,irfft:kd},gc={hammingWindow:bp,hannWindow:xp,frame:vp,stft:wp},yc={flipLeftRight:Ip,grayscaleToRGB:Sp,resizeNearestNeighbor:Gp,resizeBilinear:Up,rgbToGrayscale:_p,rotateWithOffset:Np,cropAndResize:kp,nonMaxSuppression:Cp,nonMaxSuppressionAsync:Pp,nonMaxSuppressionWithScore:zp,nonMaxSuppressionWithScoreAsync:Bp,nonMaxSuppressionPadded:Wp,nonMaxSuppressionPaddedAsync:Vp,threshold:Hp,transform:jp},bc={bandPart:qp,gramSchmidt:Kp,qr:Yp},xc={absoluteDifference:Qp,computeWeightedLoss:Jp,cosineDistance:ec,hingeLoss:tc,huberLoss:nc,logLoss:rc,meanSquaredError:ac,sigmoidCrossEntropy:sc,softmaxCrossEntropy:ic},vc={sparseFillEmptyRows:oc,sparseReshape:lc,sparseSegmentMean:uc,sparseSegmentSum:hc},wc={stringNGrams:dc,stringSplit:pc,stringToHashBucketFast:cc,staticRegexReplace:fc},kc={};f(kc,{Serializable:()=>_c,SerializationMap:()=>Nc,getRegisteredName:()=>Cc,registerClass:()=>Tc});var Ic=new Map,Sc=new Map,_c=class{getClassName(){return this.constructor.className}static fromConfig(e,t){return new e(t)}},Nc=class e{constructor(){this.classNameMap={}}static getMap(){return null==e.instance&&(e.instance=new e),e.instance}static register(t){e.getMap().classNameMap[t.className]=[t,t.fromConfig]}};function Tc(e,t,n){K(null!=e.className,()=>"Class being registered does not have the static className property defined."),K("string"==typeof e.className,()=>"className is required to be a string, but got type "+typeof e.className),K(e.className.length>0,()=>"Class being registered has an empty-string as its className, which is disallowed."),void 0===t&&(t="Custom"),void 0===n&&(n=e.className);let r=t+">"+n;return Nc.register(e),Ic.set(r,e),Sc.set(e,r),e}function Cc(e){return Sc.has(e)?Sc.get(e):e.className}var Ec=class extends _c{minimize(e,t=!1,n){let{value:r,grads:a}=this.computeGradients(e,n);if(null!=n){let e=n.map(e=>({name:e.name,tensor:a[e.name]}));this.applyGradients(e)}else this.applyGradients(a);return ei(a),t?r:(r.dispose(),null)}get iterations(){return null==this.iterations_&&(this.iterations_=0),this.iterations_}incrementIterations(){this.iterations_=this.iterations+1}computeGradients(e,t){return Mu(e,t)}dispose(){null!=this.iterations_&&ei(this.iterations_)}async saveIterations(){return null==this.iterations_&&(this.iterations_=0),{name:"iter",tensor:au(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 ${this.getClassName()}`)}async extractIterations(e){return this.iterations_=(await e[0].tensor.data())[0],e.slice(1)}};Object.defineProperty(Ec,Symbol.hasInstance,{value:e=>null!=e.minimize&&null!=e.computeGradients&&null!=e.applyGradients});var Ac=class extends Ec{static get className(){return"Adadelta"}constructor(e,t,n=null){super(),this.learningRate=e,this.rho=t,this.epsilon=n,this.accumulatedGrads=[],this.accumulatedUpdates=[],null==n&&(this.epsilon=_s.backend.epsilon())}applyGradients(e){(Array.isArray(e)?e.map(e=>e.name):Object.keys(e)).forEach((t,n)=>{let r=_s.registeredVariables[t],a=!1;null==this.accumulatedGrads[n]&&(this.accumulatedGrads[n]={originalName:`${t}/accum_grad`,variable:Qs(()=>Bl(r).variable(a))}),null==this.accumulatedUpdates[n]&&(this.accumulatedUpdates[n]={originalName:`${t}/accum_var`,variable:Qs(()=>Bl(r).variable(a))});let s=Array.isArray(e)?e[n].tensor:e[t];if(null==s)return;let i=this.accumulatedGrads[n].variable,o=this.accumulatedUpdates[n].variable;Qs(()=>{let e=fo(yo(i,this.rho),yo(iu(s),1-this.rho)),t=yo(go(su(fo(o,this.epsilon)),su(fo(i,this.epsilon))),s),n=fo(yo(o,this.rho),yo(iu(t),1-this.rho));i.assign(e),o.assign(n);let a=fo(yo(t,-this.learningRate),r);r.assign(a)})}),this.incrementIterations()}dispose(){null!=this.accumulatedUpdates&&(ei(this.accumulatedGrads.map(e=>e.variable)),ei(this.accumulatedUpdates.map(e=>e.variable)))}async getWeights(){let e=[...this.accumulatedGrads,...this.accumulatedUpdates];return[await this.saveIterations()].concat(e.map(e=>({name:e.originalName,tensor:e.variable})))}async setWeights(e){let t=(e=await this.extractIterations(e)).length/2,n=!1;this.accumulatedGrads=e.slice(0,t).map(e=>({originalName:e.name,variable:e.tensor.variable(n)})),this.accumulatedUpdates=e.slice(t,2*t).map(e=>({originalName:e.name,variable:e.tensor.variable(n)}))}getConfig(){return{learningRate:this.learningRate,rho:this.rho,epsilon:this.epsilon}}static fromConfig(e,t){return new e(t.learningRate,t.rho,t.epsilon)}},$c=class extends Ec{static get className(){return"Adagrad"}constructor(e,t=.1){super(),this.learningRate=e,this.initialAccumulatorValue=t,this.accumulatedGrads=[]}applyGradients(e){(Array.isArray(e)?e.map(e=>e.name):Object.keys(e)).forEach((t,n)=>{let r=_s.registeredVariables[t];null==this.accumulatedGrads[n]&&(this.accumulatedGrads[n]={originalName:`${t}/accumulator`,variable:Qs(()=>pl(r.shape,this.initialAccumulatorValue).variable(!1))});let a=Array.isArray(e)?e[n].tensor:e[t];if(null==a)return;let s=this.accumulatedGrads[n].variable;Qs(()=>{let e=fo(s,iu(a));s.assign(e);let t=fo(yo(go(a,su(fo(e,_s.backend.epsilon()))),-this.learningRate),r);r.assign(t)})}),this.incrementIterations()}dispose(){null!=this.accumulatedGrads&&ei(this.accumulatedGrads.map(e=>e.variable))}async getWeights(){return[await this.saveIterations()].concat(this.accumulatedGrads.map(e=>({name:e.originalName,tensor:e.variable})))}async setWeights(e){e=await this.extractIterations(e);this.accumulatedGrads=e.map(e=>({originalName:e.name,variable:e.tensor.variable(false)}))}getConfig(){return{learningRate:this.learningRate,initialAccumulatorValue:this.initialAccumulatorValue}}static fromConfig(e,t){return new e(t.learningRate,t.initialAccumulatorValue)}},Rc=class extends Ec{static get className(){return"Adam"}constructor(e,t,n,r=null){super(),this.learningRate=e,this.beta1=t,this.beta2=n,this.epsilon=r,this.accumulatedFirstMoment=[],this.accumulatedSecondMoment=[],Qs(()=>{this.accBeta1=au(t).variable(),this.accBeta2=au(n).variable()}),null==r&&(this.epsilon=_s.backend.epsilon())}applyGradients(e){let t=Array.isArray(e)?e.map(e=>e.name):Object.keys(e);Qs(()=>{let n=Wu(1,this.accBeta1),r=Wu(1,this.accBeta2);t.forEach((t,a)=>{let s=_s.registeredVariables[t],i=!1;null==this.accumulatedFirstMoment[a]&&(this.accumulatedFirstMoment[a]={originalName:`${t}/m`,variable:Qs(()=>Bl(s).variable(i))}),null==this.accumulatedSecondMoment[a]&&(this.accumulatedSecondMoment[a]={originalName:`${t}/v`,variable:Qs(()=>Bl(s).variable(i))});let o=Array.isArray(e)?e[a].tensor:e[t];if(null==o)return;let l=this.accumulatedFirstMoment[a].variable,u=this.accumulatedSecondMoment[a].variable,h=fo(yo(l,this.beta1),yo(o,1-this.beta1)),d=fo(yo(u,this.beta2),yo(iu(o),1-this.beta2)),p=go(h,n),c=go(d,r);l.assign(h),u.assign(d);let f=fo(yo(go(p,fo(su(c),this.epsilon)),-this.learningRate),s);s.assign(f)}),this.accBeta1.assign(yo(this.accBeta1,this.beta1)),this.accBeta2.assign(yo(this.accBeta2,this.beta2))}),this.incrementIterations()}dispose(){this.accBeta1.dispose(),this.accBeta2.dispose(),null!=this.accumulatedFirstMoment&&ei(this.accumulatedFirstMoment.map(e=>e.variable)),null!=this.accumulatedSecondMoment&&ei(this.accumulatedSecondMoment.map(e=>e.variable))}async getWeights(){let e=[...this.accumulatedFirstMoment,...this.accumulatedSecondMoment];return[await this.saveIterations()].concat(e.map(e=>({name:e.originalName,tensor:e.variable})))}async setWeights(e){e=await this.extractIterations(e),Qs(()=>{this.accBeta1.assign(ru(this.beta1,this.iterations_+1)),this.accBeta2.assign(ru(this.beta2,this.iterations_+1))});let t=e.length/2,n=!1;this.accumulatedFirstMoment=e.slice(0,t).map(e=>({originalName:e.name,variable:e.tensor.variable(n)})),this.accumulatedSecondMoment=e.slice(t,2*t).map(e=>({originalName:e.name,variable:e.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)}},Fc=class extends Ec{static get className(){return"Adamax"}constructor(e,t,n,r=null,a=0){super(),this.learningRate=e,this.beta1=t,this.beta2=n,this.epsilon=r,this.decay=a,this.accumulatedFirstMoment=[],this.accumulatedWeightedInfNorm=[],Qs(()=>{this.iteration=au(0).variable(),this.accBeta1=au(t).variable()}),null==r&&(this.epsilon=_s.backend.epsilon())}applyGradients(e){let t=Array.isArray(e)?e.map(e=>e.name):Object.keys(e);Qs(()=>{let n=Wu(1,this.accBeta1),r=go(-this.learningRate,fo(yo(this.iteration,this.decay),1));t.forEach((t,a)=>{let s=_s.registeredVariables[t],i=!1;null==this.accumulatedFirstMoment[a]&&(this.accumulatedFirstMoment[a]={originalName:`${t}/m`,variable:Bl(s).variable(i)}),null==this.accumulatedWeightedInfNorm[a]&&(this.accumulatedWeightedInfNorm[a]={originalName:`${t}/v`,variable:Bl(s).variable(i)});let o=Array.isArray(e)?e[a].tensor:e[t];if(null==o)return;let l=this.accumulatedFirstMoment[a].variable,u=this.accumulatedWeightedInfNorm[a].variable,h=fo(yo(l,this.beta1),yo(o,1-this.beta1)),d=yo(u,this.beta2),p=bo(o),c=eh(d,p);l.assign(h),u.assign(c);let f=fo(yo(go(r,n),go(h,fo(c,this.epsilon))),s);s.assign(f)}),this.iteration.assign(fo(this.iteration,1)),this.accBeta1.assign(yo(this.accBeta1,this.beta1))}),this.incrementIterations()}dispose(){this.accBeta1.dispose(),this.iteration.dispose(),null!=this.accumulatedFirstMoment&&ei(this.accumulatedFirstMoment.map(e=>e.variable)),null!=this.accumulatedWeightedInfNorm&&ei(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)}},Dc=class extends Ec{static get className(){return"SGD"}constructor(e){super(),this.learningRate=e,this.setLearningRate(e)}applyGradients(e){(Array.isArray(e)?e.map(e=>e.name):Object.keys(e)).forEach((t,n)=>{let r=Array.isArray(e)?e[n].tensor:e[t];if(null==r)return;let a=_s.registeredVariables[t];Qs(()=>{let e=fo(yo(this.c,r),a);a.assign(e)})}),this.incrementIterations()}setLearningRate(e){this.learningRate=e,null!=this.c&&this.c.dispose(),this.c=ti(au(-e))}dispose(){this.c.dispose()}async getWeights(){return[await this.saveIterations()]}async setWeights(e){if(0!==(e=await this.extractIterations(e)).length)throw new Error("SGD optimizer does not have settable weights.")}getConfig(){return{learningRate:this.learningRate}}static fromConfig(e,t){return new e(t.learningRate)}},Mc=class extends Dc{static get className(){return"Momentum"}constructor(e,t,n=!1){super(e),this.learningRate=e,this.momentum=t,this.useNesterov=n,this.accumulations=[],this.m=au(this.momentum)}applyGradients(e){(Array.isArray(e)?e.map(e=>e.name):Object.keys(e)).forEach((t,n)=>{let r=_s.registeredVariables[t];null==this.accumulations[n]&&(this.accumulations[n]={originalName:`${t}/momentum`,variable:Qs(()=>Bl(r).variable(!1))});let a=this.accumulations[n].variable,s=Array.isArray(e)?e[n].tensor:e[t];null!=s&&Qs(()=>{let e,t=fo(yo(this.m,a),s);e=this.useNesterov?fo(yo(this.c,fo(s,yo(t,this.m))),r):fo(yo(this.c,t),r),a.assign(t),r.assign(e)})}),this.incrementIterations()}dispose(){this.m.dispose(),null!=this.accumulations&&ei(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);this.accumulations=e.map(e=>({originalName:e.name,variable:e.tensor.variable(false)}))}getConfig(){return{learningRate:this.learningRate,momentum:this.momentum,useNesterov:this.useNesterov}}static fromConfig(e,t){return new e(t.learningRate,t.momentum,t.useNesterov)}},Oc=class extends Ec{static get className(){return"RMSProp"}constructor(e,t=.9,n=0,r=null,a=!1){if(super(),this.learningRate=e,this.decay=t,this.momentum=n,this.epsilon=r,this.accumulatedMeanSquares=[],this.accumulatedMoments=[],this.accumulatedMeanGrads=[],this.centered=a,null==r&&(this.epsilon=_s.backend.epsilon()),null==e)throw new Error("learningRate for RMSPropOptimizer must be defined.")}applyGradients(e){(Array.isArray(e)?e.map(e=>e.name):Object.keys(e)).forEach((t,n)=>{let r=_s.registeredVariables[t],a=!1;null==this.accumulatedMeanSquares[n]&&(this.accumulatedMeanSquares[n]={originalName:`${t}/rms`,variable:Qs(()=>Bl(r).variable(a))}),null==this.accumulatedMoments[n]&&(this.accumulatedMoments[n]={originalName:`${t}/momentum`,variable:Qs(()=>Bl(r).variable(a))}),null==this.accumulatedMeanGrads[n]&&this.centered&&(this.accumulatedMeanGrads[n]={originalName:`${t}/mg`,variable:Qs(()=>Bl(r).variable(a))});let s=Array.isArray(e)?e[n].tensor:e[t];if(null==s)return;let i=this.accumulatedMeanSquares[n].variable,o=this.accumulatedMoments[n].variable;Qs(()=>{let e=fo(yo(i,this.decay),yo(iu(s),1-this.decay));if(this.centered){let t=this.accumulatedMeanGrads[n].variable,a=fo(yo(t,this.decay),yo(s,1-this.decay)),l=go(yo(s,this.learningRate),su(Wu(e,fo(iu(a),this.epsilon)))),u=fo(yo(o,this.momentum),l);i.assign(e),t.assign(a),o.assign(u);let h=Wu(r,u);r.assign(h)}else{let e=fo(yo(i,this.decay),yo(iu(s),1-this.decay)),t=fo(yo(o,this.momentum),go(yo(s,this.learningRate),su(fo(e,this.epsilon))));i.assign(e),o.assign(t);let n=Wu(r,t);r.assign(n)}})}),this.incrementIterations()}dispose(){null!=this.accumulatedMeanSquares&&ei(this.accumulatedMeanSquares.map(e=>e.variable)),null!=this.accumulatedMeanGrads&&this.centered&&ei(this.accumulatedMeanGrads.map(e=>e.variable)),null!=this.accumulatedMoments&&ei(this.accumulatedMoments.map(e=>e.variable))}async getWeights(){let e=[...this.accumulatedMeanSquares,...this.accumulatedMoments];return this.centered&&e.push(...this.accumulatedMeanGrads),[await this.saveIterations()].concat(e.map(e=>({name:e.originalName,tensor:e.variable})))}async setWeights(e){e=await this.extractIterations(e);let t=this.centered?e.length/3:e.length/2,n=!1;this.accumulatedMeanSquares=e.slice(0,t).map(e=>({originalName:e.name,variable:e.tensor.variable(n)})),this.accumulatedMoments=e.slice(t,2*t).map(e=>({originalName:e.name,variable:e.tensor.variable(n)})),this.centered&&(this.accumulatedMeanGrads=e.slice(2*t,3*t).map(e=>({originalName:e.name,variable:e.tensor.variable(n)})))}getConfig(){return{learningRate:this.learningRate,decay:this.decay,momentum:this.momentum,epsilon:this.epsilon,centered:this.centered}}static fromConfig(e,t){return new e(t.learningRate,t.decay,t.momentum,t.epsilon,t.centered)}},Lc=[Ac,$c,Rc,Fc,Mc,Oc,Dc];var Pc={};f(Pc,{CompositeArrayBuffer:()=>Gs,browserFiles:()=>Vc,browserHTTPRequest:()=>Yc,concatenateArrayBuffers:()=>ki,copyModel:()=>so,decodeWeights:()=>ci,decodeWeightsStream:()=>bi,encodeWeights:()=>pi,fromMemory:()=>tf,fromMemorySync:()=>nf,getLoadHandlers:()=>Fi,getModelArtifactsForJSON:()=>Ni,getModelArtifactsForJSONSync:()=>_i,getModelArtifactsInfoForJSON:()=>Ti,getSaveHandlers:()=>Ri,getWeightSpecs:()=>Ci,http:()=>Zc,isHTTPScheme:()=>Kc,listModels:()=>ro,loadWeights:()=>Hc,moveModel:()=>io,registerLoadRouter:()=>$i,registerSaveRouter:()=>Ai,removeModel:()=>ao,weightsLoaderFactory:()=>jc,withSaveHandler:()=>rf,withSaveHandlerSync:()=>af});function zc(e){return new Promise(e=>setTimeout(e)).then(e)}var Bc=class e{constructor(t){if(!Le().getBool("IS_BROWSER"))throw new Error("browserDownloads() cannot proceed because the current environment is not a browser.");t.startsWith(e.URL_SCHEME)&&(t=t.slice(e.URL_SCHEME.length)),(null==t||0===t.length)&&(t="model"),this.modelJsonFileName=t+".json",this.weightDataFileName=t+".weights.bin"}async save(e){if("undefined"==typeof document)throw new Error("Browser downloads are not supported in this environment since `document` is not present");let t=Gs.join(e.weightData),n=window.URL.createObjectURL(new Blob([t],{type:"application/octet-stream"}));if(e.modelTopology instanceof ArrayBuffer)throw new Error("BrowserDownloads.save() does not support saving model topology in binary formats yet.");{let t=Si(e,[{paths:["./"+this.weightDataFileName],weights:e.weightSpecs}]),r=window.URL.createObjectURL(new Blob([JSON.stringify(t)],{type:"application/json"})),a=null==this.modelJsonAnchor?document.createElement("a"):this.modelJsonAnchor;if(a.download=this.modelJsonFileName,a.href=r,await zc(()=>a.dispatchEvent(new MouseEvent("click"))),null!=e.weightData){let e=null==this.weightDataAnchor?document.createElement("a"):this.weightDataAnchor;e.download=this.weightDataFileName,e.href=n,await zc(()=>e.dispatchEvent(new MouseEvent("click")))}return{modelArtifactsInfo:Ti(e)}}}};Bc.URL_SCHEME="downloads://";var Wc=class{constructor(e){if(null==e||e.length<1)throw new Error(`When calling browserFiles, at least 1 file is required, but received ${e}`);this.jsonFile=e[0],this.weightsFiles=e.slice(1)}async load(){return new Promise((e,t)=>{let n=new FileReader;n.onload=n=>{let r=JSON.parse(n.target.result),a=r.modelTopology;if(null==a)return void t(new Error(`modelTopology field is missing from file ${this.jsonFile.name}`));if(null==r.weightsManifest)return void t(new Error(`weightManifest field is missing from file ${this.jsonFile.name}`));if(0===this.weightsFiles.length)return void e({modelTopology:a});let s=Ni(r,e=>this.loadWeights(e));e(s)},n.onerror=e=>t(`Failed to read model topology and weights manifest JSON from file '${this.jsonFile.name}'. BrowserFiles supports loading Keras-style tf.Model artifacts only.`),n.readAsText(this.jsonFile)})}loadWeights(e){let t=[],n=[];for(let r of e)t.push(...r.weights),n.push(...r.paths);let r=this.checkManifestAndWeightFiles(e),a=n.map(e=>this.loadWeightsFile(e,r[e]));return Promise.all(a).then(e=>[t,e])}loadWeightsFile(e,t){return new Promise((n,r)=>{let a=new FileReader;a.onload=e=>{let t=e.target.result;n(t)},a.onerror=t=>r(`Failed to weights data from file of path '${e}'.`),a.readAsArrayBuffer(t)})}checkManifestAndWeightFiles(e){let t=[],n=this.weightsFiles.map(e=>Ii(e.name)),r={};for(let a of e)a.paths.forEach(e=>{let a=Ii(e);if(-1!==t.indexOf(a))throw new Error(`Duplicate file basename found in weights manifest: '${a}'`);if(t.push(a),-1===n.indexOf(a))throw new Error(`Weight file with basename '${a}' is not provided.`);r[e]=this.weightsFiles[n.indexOf(a)]});if(t.length!==this.weightsFiles.length)throw new Error(`Mismatch in the number of files in weights manifest (${t.length}) and the number of weight files provided (${this.weightsFiles.length}).`);return r}};function Vc(e){return new Wc(e)}function Uc(e,t,n,r){var a;K(null!=(a=e)&&Array.isArray(a)&&a.length>0,()=>"promises must be a none empty array"),function(e,t){K(e>=0&&e<=1,()=>`Progress fraction must be in range [0, 1], but got startFraction ${e}`),K(t>=0&&t<=1,()=>`Progress fraction must be in range [0, 1], but got endFraction ${t}`),K(t>=e,()=>`startFraction must be no more than endFraction, but got startFraction ${e} and endFraction ${t}`)}(n=n??0,r=r??1);let s=0;return Promise.all(e.map(a=>(a.then(a=>{let i=n+ ++s/e.length*(r-n);return t(i),a}),a)))}async function Gc(e,t){null==t&&(t={});let n=null==t.fetchFunc?Le().platform.fetch:t.fetchFunc,r=e.map(e=>n(e,t.requestInit,{isBinary:!0})),a=(null==t.onProgress?await Promise.all(r):await Uc(r,t.onProgress,0,.5)).map(e=>e.arrayBuffer());return null==t.onProgress?await Promise.all(a):await Uc(a,t.onProgress,.5,1)}async function Hc(e,t="",n,r){return jc(e=>Gc(e,{requestInit:r}))(e,t,n)}function jc(e){return async(t,n="",r)=>{let a=t.map(()=>!1),s={},i=null!=r?r.map(()=>!1):[],o=[];if(t.forEach((e,t)=>{let n=0;e.weights.forEach(e=>{let l="quantization"in e?e.quantization.dtype:e.dtype,u=Us[l]*Y(e.shape),h=()=>{a[t]=!0,null==s[t]&&(s[t]=[]),s[t].push({manifestEntry:e,groupOffset:n,sizeBytes:u})};null!=r?r.forEach((t,n)=>{t===e.name&&(h(),i[n]=!0)}):h(),o.push(e.name),n+=u})}),!i.every(e=>e)){let e=r.filter((e,t)=>!i[t]);throw new Error(`Could not find weights in manifest with names: ${e.join(", ")}. \nManifest JSON has weights with names: ${o.join(", ")}.`)}let l=a.reduce((e,t,n)=>(t&&e.push(n),e),[]),u=[];l.forEach(e=>{t[e].paths.forEach(e=>{let t=n+(n.endsWith("/")?"":"/")+e;u.push(t)})});let h=await e(u),d={},p=0;return l.forEach(e=>{let n=t[e].paths.length,r=new Gs(h.slice(p,p+n));s[e].forEach(e=>{let t=ci(r.slice(e.groupOffset,e.groupOffset+e.sizeBytes),[e.manifestEntry]);for(let e in t)d[e]=t[e]}),p+=n}),d}}Ei.registerSaveRouter(e=>Le().getBool("IS_BROWSER")&&!Array.isArray(e)&&e.startsWith(Bc.URL_SCHEME)?function(e="model"){return new Bc(e)}(e.slice(Bc.URL_SCHEME.length)):null);var qc=class{constructor(e,t){if(this.DEFAULT_METHOD="POST",null==t&&(t={}),this.weightPathPrefix=t.weightPathPrefix,this.weightUrlConverter=t.weightUrlConverter,null!=t.fetchFunc?(K("function"==typeof t.fetchFunc,()=>"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=Le().platform.fetch,K(null!=e&&e.length>0,()=>"URL path for http must not be null, undefined or empty."),Array.isArray(e)&&K(2===e.length,()=>`URL paths for http must have a length of 2, (actual length is ${e.length}).`),this.path=e,null!=t.requestInit&&null!=t.requestInit.body)throw new Error("requestInit is expected to have no pre-existing body, but has one.");this.requestInit=t.requestInit||{},this.loadOptions=t}async save(e){if(e.modelTopology instanceof ArrayBuffer)throw new Error("BrowserHTTPRequest.save() does not support saving model topology in binary formats yet.");let t=Object.assign({method:this.DEFAULT_METHOD},this.requestInit);t.body=new FormData;let n=Si(e,[{paths:["./model.weights.bin"],weights:e.weightSpecs}]);if(t.body.append("model.json",new Blob([JSON.stringify(n)],{type:"application/json"}),"model.json"),null!=e.weightData){let n=Gs.join(e.weightData);t.body.append("model.weights.bin",new Blob([n],{type:"application/octet-stream"}),"model.weights.bin")}let r=await this.fetch(this.path,t);if(r.ok)return{modelArtifactsInfo:Ti(e),responses:[r]};throw new Error(`BrowserHTTPRequest.save() failed due to HTTP response status ${r.status}.`)}async loadModelJSON(){let e,t=await this.fetch(this.path,this.requestInit);if(!t.ok)throw new Error(`Request to ${this.path} failed with status code ${t.status}. Please verify this URL points to the model JSON of the model to load.`);try{e=await t.json()}catch(e){let t=`Failed to parse model JSON of response from ${this.path}.`;throw this.path.endsWith(".pb")?t+=" 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.":t+=" Please make sure the server is serving valid JSON for this request.",new Error(t)}let n=e.modelTopology,r=e.weightsManifest;if(null==n&&null==r)throw new Error(`The JSON from HTTP path ${this.path} contains neither model topology or manifest for weights.`);return e}async load(){if(this.loadOptions.streamWeights)return this.loadStream();return Ni(await this.loadModelJSON(),e=>this.loadWeights(e))}async loadStream(){let e=await this.loadModelJSON(),t=await this.getWeightUrls(e.weightsManifest),n=Ci(e.weightsManifest);return Object.assign(Object.assign({},e),{weightSpecs:n,getWeightStream:()=>function(e,t){var n;let r,a=null==t.fetchFunc?Le().platform.fetch:t.fetchFunc,s=0;return null===(n=t.onProgress)||void 0===n||n.call(t,0),new ReadableStream({pull:async n=>{for(var i;s<e.length;){r||(r=(await a(e[s],t.requestInit,{isBinary:!0})).body.getReader());let{done:o,value:l}=await r.read();if(!o)return void n.enqueue(l);s++,r=void 0,null===(i=t.onProgress)||void 0===i||i.call(t,s/e.length)}n.close()}})}(t,this.loadOptions)})}async getWeightUrls(e){let t=Array.isArray(this.path)?this.path[1]:this.path,[n,r]=function(e){let t=e.lastIndexOf("/"),n=e.lastIndexOf("?"),r=e.substring(0,t),a=n>t?e.substring(n):"";return[r+"/",a]}(t),a=this.weightPathPrefix||n,s=[],i=[];for(let t of e)for(let e of t.paths)null!=this.weightUrlConverter?i.push(this.weightUrlConverter(e)):s.push(a+e+r);return this.weightUrlConverter&&s.push(...await Promise.all(i)),s}async loadWeights(e){let t=await this.getWeightUrls(e);return[Ci(e),await Gc(t,this.loadOptions)]}};function Kc(e){return null!=e.match(qc.URL_SCHEME_REGEX)}qc.URL_SCHEME_REGEX=/^https?:\/\//;var Xc=(e,t)=>{if("undefined"==typeof fetch&&(null==t||null==t.fetchFunc))return null;{let n=!0;if(n=Array.isArray(e)?e.every(e=>Kc(e)):Kc(e),n)return Zc(e,t)}return null};function Zc(e,t){return new qc(e,t)}function Yc(e,t){return Zc(e,t)}Ei.registerSaveRouter(Xc),Ei.registerLoadRouter(Xc);var Jc=class{constructor(e){this.modelArtifacts=e}load(){return this.modelArtifacts}},Qc=class{constructor(e){this.saveHandler=e}save(e){return this.saveHandler(e)}},ef=class{constructor(e){e.load&&(this.load=()=>Promise.resolve(e.load())),e.save&&(this.save=t=>Promise.resolve(e.save(t)))}};function tf(e,t,n,r){return new ef(nf(...arguments))}function nf(e,t,n,r){return 1===arguments.length?null!=e.modelTopology||null!=e.weightSpecs?new Jc(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 Jc({modelTopology:e})):(console.warn("Please call tf.io.fromMemory() with only one argument. The argument should be of type ModelArtifacts. The multi-argument signature of tf.io.fromMemory() has been deprecated and will be removed in a future release."),new Jc({modelTopology:e,weightSpecs:t,weightData:n,trainingConfig:r}))}function rf(e){return new Qc(e)}function af(e){return new Qc(e)}var sf={};f(sf,{confusionMatrix:()=>of});var of=zs({confusionMatrix_:function(e,t,n){let r=Os(e,"labels","confusionMatrix"),a=Os(t,"predictions","confusionMatrix");K(null==n||n>0&&Number.isInteger(n),()=>`If provided, numClasses must be a positive integer, but got ${n}`),K(1===r.rank,()=>`Expected the rank of labels to be 1, but got ${r.rank}`),K(1===a.rank,()=>`Expected the rank of predictions to be 1, but got ${a.rank}`),K(r.shape[0]===a.shape[0],()=>`Mismatch in the number of examples: ${r.shape[0]} vs. ${a.shape[0]}. Labels and predictions should have the same number of elements.`),K(n>0&&Number.isInteger(n),()=>`numClasses is required to be a positive integer, but got ${n}`);let s=ph(ho(r,"int32"),n),i=ph(ho(a,"int32"),n),o=Jd(s),l=Zo(o,i);return ho(l,"int32")}}),lf={};f(lf,{draw:()=>gf,fromPixels:()=>yf,fromPixelsAsync:()=>cf,toPixels:()=>mf});var uf,hf=!1;function df(e,t=3){if(t>4)throw new Error("Cannot construct Tensor with more than 4 channels from pixels.");if(null==e)throw new Error("pixels passed to tf.browser.fromPixels() can not be null");let n=!1,r=!1,a=!1,s=!1,i=!1,o=!1;if(e.data instanceof Uint8Array)n=!0;else if("undefined"!=typeof ImageData&&e instanceof ImageData)r=!0;else if("undefined"!=typeof HTMLVideoElement&&e instanceof HTMLVideoElement)a=!0;else if("undefined"!=typeof HTMLImageElement&&e instanceof HTMLImageElement)s=!0;else if(null!=e.getContext)i=!0;else{if(!("undefined"!=typeof ImageBitmap&&e instanceof ImageBitmap))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}`);o=!0}if(null!=da(na,_s.backendName)){let n={pixels:e},r={numChannels:t};return _s.runKernel(na,n,r)}let l,u,[h,d]=a?[e.videoWidth,e.videoHeight]:[e.width,e.height];if(i)l=e.getContext("2d").getImageData(0,0,h,d).data;else if(r||n)l=e.data;else if(s||a||o){if(null==uf)if("undefined"==typeof document){if("undefined"==typeof OffscreenCanvas||"undefined"==typeof OffscreenCanvasRenderingContext2D)throw new Error("Cannot parse input in current context. Reason: OffscreenCanvas Context2D rendering is not supported.");uf=new OffscreenCanvas(1,1).getContext("2d")}else uf=document.createElement("canvas").getContext("2d",{willReadFrequently:!0});uf.canvas.width=h,uf.canvas.height=d,uf.drawImage(e,0,0,h,d),l=uf.getImageData(0,0,h,d).data}if(4===t)u=new Int32Array(l);else{let e=h*d;u=new Int32Array(e*t);for(let n=0;n<e;n++)for(let e=0;e<t;++e)u[n*t+e]=l[4*n+e]}return Fd(u,[d,h,t],"int32")}function pf(e){return"undefined"!=typeof window&&"undefined"!=typeof ImageBitmap&&window.hasOwnProperty("createImageBitmap")&&!(e instanceof ImageBitmap)&&function(e){return null!=e&&0!==e.width&&0!==e.height}(e)&&!function(e){return null!=e&&e.data instanceof Uint8Array}(e)}async function cf(e,t=3){let n=null;if(Le().getBool("WRAP_TO_IMAGEBITMAP")&&pf(e)){let t;try{t=await createImageBitmap(e,{premultiplyAlpha:"none"})}catch(e){t=null}n=null!=t&&t.width===e.width&&t.height===e.height?t:e}else n=e;return df(n,t)}function ff(e){if(2!==e.rank&&3!==e.rank)throw new Error(`toPixels only supports rank 2 or 3 tensors, got rank ${e.rank}.`);let t=2===e.rank?1:e.shape[2];if(t>4||2===t)throw new Error(`toPixels only supports depth of size 1, 3 or 4 but got ${t}`);if("float32"!==e.dtype&&"int32"!==e.dtype)throw new Error(`Unsupported type for toPixels: ${e.dtype}. Please use float32 or int32 tensors.`)}async function mf(e,t){let n=Os(e,"img","toPixels");if(!(e instanceof ts)){let e=n;n=ho(e,"int32"),e.dispose()}ff(n);let[r,a]=n.shape.slice(0,2),s=2===n.rank?1:n.shape[2],i=await n.data(),o="float32"===n.dtype?255:1,l=new Uint8ClampedArray(a*r*4);for(let e=0;e<r*a;++e){let t=[0,0,0,255];for(let r=0;r<s;r++){let a=i[e*s+r];if("float32"===n.dtype){if(a<0||a>1)throw new Error(`Tensor values for a float32 Tensor must be in the range [0 - 1] but encountered ${a}.`)}else if("int32"===n.dtype&&(a<0||a>255))throw new Error(`Tensor values for a int32 Tensor must be in the range [0 - 255] but encountered ${a}.`);1===s?(t[0]=a*o,t[1]=a*o,t[2]=a*o):t[r]=a*o}let r=4*e;l[r+0]=Math.round(t[0]),l[r+1]=Math.round(t[1]),l[r+2]=Math.round(t[2]),l[r+3]=Math.round(t[3])}if(null!=t){hf||null!=da(Pt,_s.backendName)&&(console.warn("tf.browser.toPixels is not efficient to draw tensor on canvas. Please try tf.browser.draw instead."),hf=!0),t.width=a,t.height=r;let e=t.getContext("2d"),n=new ImageData(l,a,r);e.putImageData(n,0,0)}return n!==e&&n.dispose(),l}function gf(e,t,n){let r=Os(e,"img","draw");if(!(e instanceof ts)){let e=r;r=ho(e,"int32"),e.dispose()}ff(r),function(e){let t=(null==e?void 0:e.alpha)||1;if(t>1||t<0)throw new Error(`Alpha value ${t} is suppoed to be in range [0 - 1].`)}(null==n?void 0:n.imageOptions);let a={image:r},s={canvas:t,options:n};_s.runKernel(Pt,a,s)}var yf=zs({fromPixels_:df}),bf={};function xf(e,t){let n=e.shape.length,r=t.shape.length;if(n<1)throw new Error(`tf.gatherND() expects the input to be rank 1 or higher, but the rank was ${n}.`);if(r<1)throw new Error(`tf.gatherND() expects the indices to be rank 1 or higher, but the rank was ${r}.`);if("int32"!==t.dtype)throw new Error(`tf.gatherND() expects the indices to be int32 type, but the dtype was ${t.dtype}.`);if(t.shape[r-1]>n)throw new Error(`index innermost dimension length must be <= tensor rank; saw: ${t.shape[r-1]} vs. ${n}`);if(0===Y(e.shape))throw new Error(`Requested more than 0 entries, but input is empty. Input shape: ${e.shape}.`);let a=t.shape,s=a[a.length-1],i=1;for(let e=0;e<a.length-1;++e)i*=a[e];let o=e.shape,l=a.slice();l.pop();let u=1;for(let e=s;e<n;++e)u*=o[e],l.push(o[e]);let h=[...Ie(e.shape).map(e=>e/u),1].slice(0,s);return[l,i,u,h]}f(bf,{prepareAndValidate:()=>xf});var vf={};f(vf,{assertParamsValid:()=>wf,computeFlatOffset:()=>Df,computeOutShape:()=>If,getNormalizedAxes:()=>Tf,isSliceContinous:()=>Ff,maskToAxes:()=>kf,parseSliceParams:()=>Mf,sliceInfo:()=>Of,startForAxis:()=>$f,startIndicesWithElidedDims:()=>Cf,stopForAxis:()=>Rf,stopIndicesWithElidedDims:()=>Ef,stridesForAxis:()=>Af,stridesWithElidedDims:()=>Sf});function wf(e,t,n){let r=e.shape.length;K(r===t.length,()=>`Error in slice${r}D: Length of begin ${t} must match the rank of the array (${r}).`),K(r===n.length,()=>`Error in slice${r}D: Length of size ${n} must match the rank of the array (${r}).`);for(let a=0;a<r;++a)K(t[a]+n[a]<=e.shape[a],()=>`Error in slice${r}D: begin[${a}] + size[${a}] (${t[a]+n[a]}) would overflow input.shape[${a}] (${e.shape[a]})`)}function kf(e){let t=[],n=0;for(;e>0;)1&e&&t.push(n),e/=2,n++;return t}function If(e,t,n){let r=[];for(let a=0;a<e.length;a++)r[a]=Math.ceil((t[a]-e[a])/n[a]);return r}function Sf(e,t,n,r){let a=[...e];for(let e=a.length;e<r.length;e++)a.push(1);for(let e=0;e<n;e++)0===e?a[t]=1:(a.splice(t,0,1),a.pop());return a}function _f(e,t,n){return n<=e?n:n-(t-1)}function Nf(e,t){let n=[];for(let r=0;r<e;r++)n.push(t+r);return n}function Tf(e,t,n,r,a,s,i,o,l){let u=e.length,h=new Array(u),d=new Array(u),p=new Array(u);if(t.length&&n>0){let l=t[0],u=n+1;h=Cf(i,l,u,r,e),d=Ef(o,l,u,a,e),p=Sf(s,l,u,e)}else for(let t=0;t<u;t++)h[t]=$f(i,r,s,e,t,l),d[t]=Rf(o,a,s,e,t,l),p[t]=Af(s,t,l);return{begin:h,end:d,strides:p}}function Cf(e,t,n,r,a){let s=[...a],i=Nf(n,t);for(let a=0;a<s.length;a++)if(i.indexOf(a)>-1)s[a]=0;else{let i=_f(t,n,a),o=r[i];e&1<<i&&(o=0),s[a]=o}return s}function Ef(e,t,n,r,a){let s=[...a],i=Nf(n,t);for(let a=0;a<s.length;a++)if(i.indexOf(a)>-1)s[a]=Number.MAX_SAFE_INTEGER;else{let i=_f(t,n,a),o=r[i];e&1<<i&&(o=Number.MAX_SAFE_INTEGER),s[a]=o}for(let e=0;e<s.length;e++){let t=a[e];s[e]<0&&(s[e]+=t),s[e]=V(0,s[e],a[e])}return s}function Af(e,t,n){let r=e[t];return(n&1<<t||null==r)&&(r=1),r}function $f(e,t,n,r,a,s){let i=t[a],o=n[a]||1;(e&1<<a||s&1<<a||null==i)&&(i=o>0?Number.MIN_SAFE_INTEGER:Number.MAX_SAFE_INTEGER);let l=r[a];return i<0&&(i+=l),i=V(0,i,l-1),i}function Rf(e,t,n,r,a,s){let i=t[a],o=n[a]||1;(e&1<<a||s&1<<a||null==i)&&(i=o>0?Number.MAX_SAFE_INTEGER:Number.MIN_SAFE_INTEGER);let l=r[a];return i<0&&(i+=l),i=o>0?V(0,i,l):V(-1,i,l-1),i}function Ff(e,t,n){let r=n.length;for(let e=0;e<n.length;e++)if(n[e]>1){r=e;break}for(let a=r+1;a<n.length;a++)if(t[a]>0||n[a]!==e[a])return!1;return!0}function Df(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 Mf(e,t,n){let r,a,s=e.shape.length;return r="number"==typeof t?[t,...new Array(s-1).fill(0)]:t.length<s?t.concat(new Array(s-t.length).fill(0)):t.slice(),r.forEach(e=>{K(-1!==e,()=>"slice() does not support negative begin indexing.")}),a=null==n?new Array(s).fill(-1):"number"==typeof n?[n,...new Array(s-1).fill(-1)]:n.length<s?n.concat(new Array(s-n.length).fill(-1)):n,a=a.map((t,n)=>t>=0?t:(K(-1===t,()=>`Negative size values should be exactly -1 but got ${t} for the slice() size at index ${n}.`),e.shape[n]-r[n])),[r,a]}function Of(e,t,n,r,a,s,i,o,l){let u;if(null==r?(u=new Array(t.length),u.fill(1)):u=r,null!=i&&i&i-1)throw new Error("Multiple ellipses in slice is not allowed.");let h=!1,d={dims:u.length,numAddAxisAfterEllipsis:0,begin:t.slice(),end:n.slice(),strides:u.slice(),beginMask:a,endMask:s,ellipsisMask:i,newAxisMask:o,shrinkAxisMask:l};for(let e=0;e<d.dims;e++)h&&1<<e&o&&d.numAddAxisAfterEllipsis++,1<<e&i&&(h=!0);h||(d.ellipsisMask|=1<<d.dims,d.dims++);let p={dims:e.length,beginMask:0,endMask:0,beginValid:!1,endValid:!1};!function(e,t){t.beginMask=0,t.endMask=0,t.shrinkAxisMask=0;let n=0;t.beginValid=null!=e.begin,t.endValid=null!=e.end,t.begin=new Array(t.dims),t.end=new Array(t.dims),t.strides=new Array(t.dims),t.finalShapeGatherIndices=[],t.finalShapeGatherIndicesSparse=[],t.inputShapeGatherIndicesSparse=new Array(t.dims);for(let r=0;r<e.dims;r++)if(1<<r&e.ellipsisMask){let a=Math.min(t.dims-(e.dims-r)+1+e.numAddAxisAfterEllipsis,t.dims);for(;n<a;n++)t.begin[n]=0,t.end[n]=0,t.strides[n]=1,t.beginMask|=1<<n,t.endMask|=1<<n,t.finalShapeGatherIndices.push(n),t.finalShapeGatherIndicesSparse.push(-1),t.inputShapeGatherIndicesSparse[n]=r}else if(1<<r&e.newAxisMask)t.finalShapeGatherIndices.push(-2),t.finalShapeGatherIndicesSparse.push(-1);else{if(n===t.begin.length)throw Error(`Index out of range using input dim ${n}; input has only ${t.dims} dims, ${t.begin.length}.`);null!=e.begin&&(t.begin[n]=e.begin[r]),null!=e.end&&(t.end[n]=e.end[r]),t.strides[n]=e.strides[r],e.beginMask&1<<r&&(t.beginMask|=1<<n),e.endMask&1<<r&&(t.endMask|=1<<n),e.shrinkAxisMask&1<<r?(t.finalShapeGatherIndices.push(-1),t.finalShapeGatherIndicesSparse.push(-1),t.shrinkAxisMask|=1<<n):(t.finalShapeGatherIndices.push(n),t.finalShapeGatherIndicesSparse.push(r)),t.inputShapeGatherIndicesSparse[n]=r,n++}}(d,p);let c=!0,f=!0,m=!0,g=[],y=[];for(let t=0;t<e.length;++t){if(0===p.strides[t])throw Error(`strides[${t}] must be non-zero`);let n=!!(p.shrinkAxisMask&1<<t),r=e[t];if(-1===r){g.push(n?1:-1);continue}let a=[p.beginMask&1<<t,p.endMask&1<<t],s=[p.strides[t]>0?0:-1,p.strides[t]>0?r:r-1];if(n&&p.strides[t]<=0)throw Error("only stride 1 allowed on non-range indexing.");m=m&&1===p.strides[t];let i=!!(p.beginMask&1<<t&&p.endMask&1<<t);if(p.beginValid&&p.endValid){if(n){let e=p.begin[t]<0?r+p.begin[t]:p.begin[t];if(p.begin[t]=e,p.end[t]=p.begin[t]+1,e<0||e>=r)throw Error(`slice index ${p.begin[t]} of dimension ${t} out of bounds.`)}else p.begin[t]=Lf(p.begin[t],0,p.strides[t],r,a,s),p.end[t]=Lf(p.end[t],1,p.strides[t],r,a,s);let e=1===p.strides[t]&&0===p.begin[t]&&p.end[t]===r;c=c&&e,f=f&&(0===t&&1===p.strides[t]||e)}else c=c&&1===p.strides[t]&&i,f=f&&(0===t&&1===p.strides[t]||i);let o,l=!1;if(p.beginValid&&p.endValid?(o=p.end[t]-p.begin[t],l=!0):n?(o=1,l=!0):i&&r>=0&&(o=p.strides[t]<0?-r:r,l=!0),l){let e;e=0===o||o<0!=p.strides[t]<0?0:Math.trunc(o/p.strides[t])+(o%p.strides[t]!==0?1:0),g.push(e)}else g.push(-1)}for(let e=0;e<p.finalShapeGatherIndices.length;++e){let t=p.finalShapeGatherIndices[e];t>=0?y.push(g[t]):-2===t&&y.push(1)}return{finalShapeSparse:y.filter((e,t)=>-2!==p.finalShapeGatherIndices[t]),finalShape:y,isIdentity:c,sliceDim0:f,isSimpleSlice:m,begin:p.begin,end:p.end,strides:p.strides}}function Lf(e,t,n,r,a,s){if(a[t])return n>0?s[t]:s[t+1&1];{let t=e<0?r+e:e;return t<s[0]?s[0]:t>s[1]?s[1]:t}}var Pf="4.22.0",zf=class{static sgd(e){return new Dc(e)}static momentum(e,t,n=!1){return new Mc(e,t,n)}static rmsprop(e,t=.9,n=0,r=null,a=!1){return new Oc(e,t,n,r,a)}static adam(e=.001,t=.9,n=.999,r=null){return new Rc(e,t,n,r)}static adadelta(e=.001,t=.95,n=null){return new Ac(e,t,n)}static adamax(e=.002,t=.9,n=.999,r=null,a=0){return new Fc(e,t,n,r,a)}static adagrad(e,t=.1){return new $c(e,t)}},Bf=zf,Wf="undefined"!=typeof requestAnimationFrame?requestAnimationFrame:"undefined"!=typeof setImmediate?setImmediate:e=>e();function Vf(){return new Promise(e=>Wf(()=>e()))}var Uf,Gf={};function Hf(e,t){let n=e[0].length;e.forEach((e,t)=>{K(e.length===n,()=>`Error in concat${n}D: rank of tensors[${t}] must be the same as the rank of the rest (${n})`)}),K(t>=0&&t<n,()=>`Error in concat${n}D: axis must be between 0 and ${n-1}.`);let r=e[0];e.forEach((e,a)=>{for(let s=0;s<n;s++)K(s===t||e[s]===r[s],()=>`Error in concat${n}D: Shape of tensors[${a}] (${e}) does not match the shape of the rest (${r}) along the non-concatenated axis ${a}.`)})}function jf(e,t){let n=e[0].slice();for(let r=1;r<e.length;r++)n[t]+=e[r][t];return n}function qf(e,t,n){let r=new Array;if(null==n&&null==t)return r;if(null==t)for(;r.length<e+n.length;)r.push(-1);else r=t.slice();if(null==n)return r;if(e+n.length!==r.length)throw new Error(`rt input.shape and shape=${t} are incompatible: rt input.rank = ${e+n.length}, but shape.rank = ${r.length}`);for(let a=1;a<n.length;++a){let s=n[a],i=r[r.length-n.length+a],o=r[i];if(s>=0)if(o>=0){if(o!==s)throw new Error(`rt input.shape and shape=${t} are incompatible: rt input.shape[${a+e}] = ${s} but shape[${a+e}] = ${o}`)}else r[i]=s}return r}function Kf(e){let t={FIRST_DIM_SIZE:Uf.FIRST_DIM_SIZE,VALUE_ROWIDS:Uf.VALUE_ROWIDS,ROW_LENGTHS:Uf.ROW_LENGTHS,ROW_SPLITS:Uf.ROW_SPLITS,ROW_LIMITS:Uf.ROW_LIMITS,ROW_STARTS:Uf.ROW_STARTS},n=[];for(let r of e){if(!(r in t))break;n.push(t[r])}return n}function Xf(e){return 0===e.length?0:e[0]===Uf.FIRST_DIM_SIZE?e.length-1:e.length}function Zf(e,t){if(null==e||null==t)return;let n=e.length,r=t.length;if(n>=r)throw new Error(`defaultValue.shape=${e} and ragged tensor flatValues.shape=${t}, are incompatible: defaultValue.rank = ${n} must be less than ragged tensor input flatValues.rank = ${r})`);for(let a=0;a<Math.min(n,r-1);++a){let n=e[a],r=t[a+1];if(n>=0&&r>=0&&1!==n&&n!==r)throw new Error(`defaultValue.shape=${e}, and ragged tensor input flatValues.shape=${t} are incompatible: defaultValue.shape[${a-e.length}] = ${n} but ragged tensor input.flatValues.shape[${a-e.length}] = ${r}`)}}f(Gf,{ERF_A1:()=>lm,ERF_A2:()=>um,ERF_A3:()=>hm,ERF_A4:()=>dm,ERF_A5:()=>pm,ERF_P:()=>om,PARALLELIZE_THRESHOLD:()=>Yf,RowPartitionType:()=>Uf,SELU_SCALE:()=>im,SELU_SCALEALPHA:()=>sm,applyActivation:()=>dp,assertAndGetBroadcastShape:()=>Ll,assertAxesAreInnerMostDims:()=>Yl,assertParamsConsistent:()=>Hf,assignToTypedArray:()=>bm,axesAreInnerMostDims:()=>ql,calculateShapes:()=>Bd,checkEinsumDimSizes:()=>_m,checkPadOnDimRoundingMode:()=>Ho,combineLocations:()=>Kl,combineRaggedTensorToTensorShapes:()=>qf,complexWithEvenIndex:()=>mm,complexWithOddIndex:()=>gm,computeConv2DInfo:()=>Do,computeConv3DInfo:()=>Mo,computeDefaultPad:()=>Oo,computeDilation2DInfo:()=>$o,computeOptimalWindowSize:()=>Jf,computeOutAndReduceShapes:()=>Xl,computeOutShape:()=>jf,computePool2DInfo:()=>Ro,computePool3DInfo:()=>Fo,convertConv2DDataFormat:()=>Go,decodeEinsumEquation:()=>Im,eitherStridesOrDilationsAreOne:()=>Vo,expandShapeToKeepDim:()=>Zl,exponent:()=>vm,exponents:()=>xm,fromStringArrayToUint8:()=>qm,fromUint8ToStringArray:()=>jm,getAxesPermutation:()=>Jl,getBroadcastDims:()=>Ml,getComplexWithIndex:()=>ym,getEinsumComputePath:()=>Nm,getEinsumPermutation:()=>Sm,getFusedBiasGradient:()=>hp,getFusedDyActivation:()=>up,getImageCenter:()=>Qf,getInnerMostAxes:()=>eu,getPermuted:()=>tm,getRaggedRank:()=>Xf,getReductionAxes:()=>Ol,getReshaped:()=>em,getReshapedPermuted:()=>nm,getRowPartitionTypesHelper:()=>Kf,getSliceBeginCoords:()=>rm,getSliceSize:()=>am,getSparseFillEmptyRowsIndicesDenseShapeMismatch:()=>Am,getSparseFillEmptyRowsNegativeIndexErrorMessage:()=>$m,getSparseFillEmptyRowsOutOfRangeIndexErrorMessage:()=>Rm,getSparseReshapeEmptyTensorZeroOutputDimErrorMessage:()=>Mm,getSparseReshapeInputOutputMismatchErrorMessage:()=>Lm,getSparseReshapeInputOutputMultipleErrorMessage:()=>Om,getSparseReshapeMultipleNegativeOneOutputDimErrorMessage:()=>Fm,getSparseReshapeNegativeOutputDimErrorMessage:()=>Dm,getSparseSegmentReductionIndicesOutOfRangeErrorMessage:()=>Wm,getSparseSegmentReductionNegativeSegmentIdsErrorMessage:()=>Pm,getSparseSegmentReductionNonIncreasingSegmentIdsErrorMessage:()=>zm,getSparseSegmentReductionSegmentIdOutOfRangeErrorMessage:()=>Bm,getUndoAxesPermutation:()=>Ql,isIdentityPermutation:()=>Tm,log:()=>la,mergeRealAndImagArrays:()=>cm,prepareAndValidate:()=>xf,prepareSplitSize:()=>Em,segment_util:()=>Vm,shouldFuse:()=>pp,slice_util:()=>vf,splitRealAndImagArrays:()=>fm,stridesOrDilationsArePositive:()=>Uo,tupleValuesAreOne:()=>Wo,upcastType:()=>ps,validateDefaultValueShape:()=>Zf,validateInput:()=>zd,validateUpdateShape:()=>Pd,warn:()=>oa}),function(e){e[e.FIRST_DIM_SIZE=0]="FIRST_DIM_SIZE",e[e.VALUE_ROWIDS=1]="VALUE_ROWIDS",e[e.ROW_LENGTHS=2]="ROW_LENGTHS",e[e.ROW_SPLITS=3]="ROW_SPLITS",e[e.ROW_LIMITS=4]="ROW_LIMITS",e[e.ROW_STARTS=5]="ROW_STARTS"}(Uf||(Uf={}));var Yf=30;function Jf(e){return e<=Yf?e:ke(e,Math.floor(Math.sqrt(e)))}function Qf(e,t,n){return[n*("number"==typeof e?e:e[0]),t*("number"==typeof e?e:e[1])]}function em(e,t,n,r=!0){let a=[];if(r)a=a.concat(t.slice(0)),a.push(e[0]/n),a=a.concat(e.slice(1));else{a=a.concat(e[0]);let n=t.length;for(let r=0;r<n;++r)a=a.concat([e[r+1]/t[r],t[r]]);a=a.concat(e.slice(n+1))}return a}function tm(e,t,n=!0){let r=[];if(n){r.push(t);for(let n=t+1;n<e;++n)n<=2*t?(r.push(n),r.push(n-(t+1))):r.push(n)}else{let n=[],a=[];for(let r=1;r<e;++r)r>=2*t+1||r%2==1?a.push(r):n.push(r);r.push(...n),r.push(0),r.push(...a)}return r}function nm(e,t,n,r=!0){let a=[];r?a.push(e[0]/n):a.push(e[0]*n);for(let n=1;n<e.length;++n)n<=t.length?r?a.push(t[n-1]*e[n]):a.push(e[n]/t[n-1]):a.push(e[n]);return a}function rm(e,t){let n=[0];for(let r=0;r<t;++r)n.push(e[r][0]);return n}function am(e,t,n){let r=e.slice(0,1);for(let a=0;a<n;++a)r.push(e[a+1]-t[a][0]-t[a][1]);return r}var sm=1.7580993408473768,im=1.0507009873554805,om=.3275911,lm=.254829592,um=-.284496736,hm=1.421413741,dm=-1.453152027,pm=1.061405429;function cm(e,t){if(e.length!==t.length)throw new Error(`Cannot merge real and imag arrays of different lengths. real:${e.length}, imag: ${t.length}.`);let n=new Float32Array(2*e.length);for(let r=0;r<n.length;r+=2)n[r]=e[r/2],n[r+1]=t[r/2];return n}function fm(e){let t=new Float32Array(e.length/2),n=new Float32Array(e.length/2);for(let r=0;r<e.length;r+=2)t[r/2]=e[r],n[r/2]=e[r+1];return{real:t,imag:n}}function mm(e){let t=Math.ceil(e.length/4),n=new Float32Array(t),r=new Float32Array(t);for(let t=0;t<e.length;t+=4)n[Math.floor(t/4)]=e[t],r[Math.floor(t/4)]=e[t+1];return{real:n,imag:r}}function gm(e){let t=Math.floor(e.length/4),n=new Float32Array(t),r=new Float32Array(t);for(let t=2;t<e.length;t+=4)n[Math.floor(t/4)]=e[t],r[Math.floor(t/4)]=e[t+1];return{real:n,imag:r}}function ym(e,t){return{real:e[2*t],imag:e[2*t+1]}}function bm(e,t,n,r){e[2*r]=t,e[2*r+1]=n}function xm(e,t){let n=new Float32Array(e/2),r=new Float32Array(e/2);for(let a=0;a<Math.ceil(e/2);a++){let s=(t?2:-2)*Math.PI*(a/e);n[a]=Math.cos(s),r[a]=Math.sin(s)}return{real:n,imag:r}}function vm(e,t,n){let r=(n?2:-2)*Math.PI*(e/t);return{real:Math.cos(r),imag:Math.sin(r)}}var wm="->",km=/->/g;function Im(e,t){let n=((e=e.replace(/\s/g,"")).length-e.replace(km,"").length)/2;if(n<1)throw new Error("Equations without an arrow are not supported.");if(n>1)throw new Error(`Equation must contain exactly one arrow ("${wm}").`);let[r,a]=e.split(wm);K(-1===r.indexOf("..."),()=>'The ellipsis notation ("...") is not supported yet.');let s=r.split(","),i=s.length;if(t!==i)throw new Error(`Expected ${i} input tensors, received ${t}`);if(i>2)throw new Error("Support for more than 2 input tensors is not implemented yet.");let o=[];for(let e=0;e<a.length;++e){let t=a[e];if(!s.some(e=>-1!==e.indexOf(t)))throw new Error(`Output subscripts contain the label ${t} not present in the input subscripts.`);-1===o.indexOf(t)&&o.push(t)}for(let e=0;e<r.length;++e){let t=r[e];-1===o.indexOf(t)&&","!==t&&o.push(t)}let l=new Array(s.length);for(let e=0;e<i;++e){if(new Set(s[e].split("")).size!==s[e].length)throw new Error(`Found duplicate axes in input component ${s[e]}. Support for duplicate axes in input is not implemented yet.`);l[e]=[];for(let t=0;t<s[e].length;++t)l[e].push(o.indexOf(s[e][t]))}let u=o.length,h=[];for(let e=a.length;e<u;++e)h.push(e);return{allDims:o,summedDims:h,idDims:l}}function Sm(e,t){let n=new Array(e);n.fill(-1);for(let e=0;e<t.length;++e)n[t[e]]=e;let r=[];for(let t=0;t<e;++t)-1===n[t]&&r.push(t);return n=n.filter(e=>-1!==e),{permutationIndices:n,expandDims:r}}function _m(e,t,n){let r=new Array(e);for(let e=0;e<n.length;++e){let a=n[e].shape;for(let n=0;n<t[e].length;++n)void 0===r[t[e][n]]?r[t[e][n]]=a[n]:K(r[t[e][n]]===a[n],()=>`Expected dimension ${r[t[e][n]]} at axis ${n} of input shaped ${JSON.stringify(a)}, but got dimension ${a[n]}`)}}function Nm(e,t){let n=e,r=[],a=0;0===e.length&&n.push(-1),a=e.length+1;for(let e=0;e<a;++e)r.push([]);let s=[];for(let e=0;e<n.length;++e){let a=Cm(t,n[e]);for(let t of a)-1===s.indexOf(t)&&(r[e].push(t),s.push(t))}return{path:n,steps:r}}function Tm(e){return e.every((e,t)=>e===t)}function Cm(e,t){let n=[];for(let r=0;r<e.length;++r)(0===e[r].length||-1!==e[r].indexOf(t)||-1===t)&&n.push(r);return n}function Em(e,t,n=0){let r=[];if("number"==typeof t)K(e.shape[n]%t===0,()=>"Number of splits must evenly divide the axis."),r=new Array(t).fill(e.shape[n]/t);else{K(t.reduce((e,t)=>(-1===t&&(e+=1),e),0)<=1,()=>"There should be only one negative value in split array.");let a=t.indexOf(-1);if(-1!==a){let r=t.reduce((e,t)=>t>0?e+t:e);t[a]=e.shape[n]-r}K(e.shape[n]===t.reduce((e,t)=>e+t),()=>"The sum of sizes must match the size of the axis dimension."),r=t}return r}function Am(e){return`Received SparseTensor with denseShape[0] = 0 but\n  indices.shape[0] = ${e}`}function $m(e,t){return`indices(${e}, 0) is invalid: ${t} < 0`}function Rm(e,t,n){return`indices(${e}, 0) is invalid: ${t} >= ${n}`}function Fm(e,t){return`only one output dimension may be -1, not both ${e} and ${t}`}function Dm(e,t){return`size ${e} must be non-negative, not ${t}`}function Mm(){return"reshape cannot infer the missing input size for an empty tensor unless all specified input sizes are non-zero"}function Om(e,t){return`Input to reshape is a SparseTensor with ${Y(e)}\n  dense values, but the requested shape requires a multiple of ${Y(t)}. inputShape=${e} outputShape= ${t}`}function Lm(e,t){return`Input to reshape is a tensor with ${Y(e)} dense values, but the requested shape has ${Y(t)}. inputShape=${e} outputShape=${t}`}function Pm(){return"segment ids must be >= 0"}function zm(){return"segment ids are not increasing"}function Bm(e,t){return`Segment id ${e} out of range [0, ${t}), possibly because segmentIds input is not sorted.`}function Wm(e,t,n){return`Bad: indices[${e}] == ${t} out of range [0, ${n})`}var Vm={};function Um(e,t){let n,r=!1;for(e<=Yf?(n=e,r=!0):n=ke(e,Math.floor(Math.sqrt(e)));!r;)n>t||n===e?r=!0:n=ke(e,n+1);return n}function Gm(e,t,n){let r=[],a=e.length;for(let s=0;s<a;s++)s!==t?r.push(e[s]):r.push(n);return r}function Hm(e,t,n,r){let a=t.shape.length,s=e.shape.length;if(0!==r&&(r<-a||r>a))throw new Error(`Expect batchDims in the range of [-${a}, ${a}], but got ${r}`);if(r<0&&(r+=a),r>s)throw new Error(`batchDims (${r}) must be less than rank(x) (\n    ${s}).`);if(n<r)throw new Error(`batchDims (${r}) must be less than or equal to axis (${n}).`);for(let n=0;n<r;++n)if(e.shape[n]!==t.shape[n])throw new Error(`x.shape[${n}]: ${e.shape[n]} should be equal to indices.shape[${n}]: ${t.shape[n]}.`);let i=e.shape[n],o=[],l=1,u=1,h=1;for(let t=0;t<r;++t)o.push(e.shape[t]),l*=e.shape[t];for(let t=r;t<n;t++)o.push(e.shape[t]),u*=e.shape[t];for(let e=r;e<a;e++)o.push(t.shape[e]);for(let t=n+1;t<s;t++)o.push(e.shape[t]),h*=e.shape[t];return{batchSize:l,sliceSize:h,outerSize:u,dimSize:i,outputShape:o}}function jm(e){try{return e.map(e=>Wa(e))}catch(e){throw new Error(`Failed to decode encoded string bytes into utf-8, error: ${e}`)}}function qm(e){return e.map(e=>Ba(e))}f(Vm,{collectGatherOpShapeInfo:()=>Hm,computeOutShape:()=>Gm,segOpComputeOptimalWindowSize:()=>Um});var Km={};f(Km,{nonMaxSuppressionV3Impl:()=>$p,nonMaxSuppressionV4Impl:()=>Rp,nonMaxSuppressionV5Impl:()=>Fp,whereImpl:()=>Xd}),function(){for(let e of Lc)Tc(e)}();var Xm={kernelName:Ve,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>yo(e,Cd(ho(n,"float32"),-1))}}},Zm={kernelName:Ue,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>{let t=iu(ho(n,"float32")),r=su(Wu(au(1),t));return Pu(go(e,r))}}}},Ym={kernelName:Ge,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>{let t=su(Wu(iu(ho(n,"float32")),1));return go(e,t)}}}},Jm={kernelName:He,inputsToSave:["a","b"],gradFunc:(e,t)=>{let[n,r]=t,a=Ll(n.shape,r.shape);return{a:()=>{let t=e,r=Ol(n.shape,a);return r.length>0&&(t=ou(t,r)),jo(t,n.shape)},b:()=>{let t=e,n=Ol(r.shape,a);return n.length>0&&(t=ou(t,n)),jo(t,r.shape)}}}},Qm={kernelName:je,saveAllInputs:!0,gradFunc:(e,t)=>{let n={};return t.forEach((t,r)=>{n[r]=()=>e.clone()}),n}},eg={kernelName:Xe,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>Bl(n)}}},tg={kernelName:Ze,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>Bl(n)}}},ng={kernelName:Ye,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>go(e,su(Wu(au(1),iu(ho(n,"float32")))))}}},rg={kernelName:Je,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>{let t=su(fo(au(1),iu(ho(n,"float32"))));return go(e,t)}}}},ag={kernelName:tt,inputsToSave:["a","b"],gradFunc:(e,t)=>{let[n,r]=t,a=Ll(n.shape,r.shape);return{a:()=>{let t=fo(iu(n),iu(r)),s=yo(e,go(r,t)),i=Ol(n.shape,a);return i.length>0&&(s=ou(s,i)),jo(s,n.shape)},b:()=>{let t=fo(iu(n),iu(r)),s=Pu(yo(e,go(n,t))),i=Ol(r.shape,a);return i.length>0&&(s=ou(s,i)),jo(s,r.shape)}}}},sg={kernelName:Qe,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>go(e,fo(iu(ho(n,"float32")),1))}}},ig={kernelName:et,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>go(e,Wu(au(1),iu(ho(n,"float32"))))}}};var og=zs({avgPool3dGrad_:function(e,t,n,r,a,s){let i=Os(e,"dy","avgPool3dGrad"),o=Os(t,"input","avgPool3dGrad"),l=i,u=o,h=!1;4===o.rank&&(h=!0,l=jo(i,[1,i.shape[0],i.shape[1],i.shape[2],i.shape[3]]),u=jo(o,[1,o.shape[0],o.shape[1],o.shape[2],o.shape[3]])),K(5===l.rank,()=>`Error in avgPool3dGrad: dy must be rank 5 but got rank ${l.rank}.`),K(5===u.rank,()=>`Error in avgPool3dGrad: input must be rank 5 but got rank ${u.rank}.`),Ho("avgPool3dGrad",a,s);let d={dy:l,input:u},p={filterSize:n,strides:r,pad:a,dimRoundingMode:s},c=_s.runKernel(st,d,p);return h?jo(c,[c.shape[1],c.shape[2],c.shape[3],c.shape[4]]):c}}),lg={kernelName:at,inputsToSave:["x"],gradFunc:(e,t,n)=>{let[r]=t,{filterSize:a,strides:s,pad:i,dimRoundingMode:o}=n;return{x:()=>og(e,r,a,s,i,o)}}};var ug=zs({avgPoolGrad_:function(e,t,n,r,a){let s=Os(e,"dy","avgPoolGrad"),i=Os(t,"input","avgPoolGrad");K(i.rank===s.rank,()=>`Rank of input (${i.rank}) does not match rank of dy (${s.rank})`);let o=i,l=s,u=!1;3===i.rank&&(u=!0,o=jo(i,[1,i.shape[0],i.shape[1],i.shape[2]]),l=jo(s,[1,s.shape[0],s.shape[1],s.shape[2]])),K(4===l.rank,()=>`Error in avgPoolGrad: dy must be rank 4 but got rank ${l.rank}.`),K(4===o.rank,()=>`Error in avgPoolGrad: input must be rank 4 but got rank ${o.rank}.`);let h={dy:l,input:o},d={filterSize:n,strides:r,pad:a},p=_s.runKernel(rt,h,d);return u?jo(p,[p.shape[1],p.shape[2],p.shape[3]]):p}}),hg={kernelName:nt,inputsToSave:["x"],gradFunc:(e,t,n)=>{let[r]=t,{filterSize:a,strides:s,pad:i}=n;return{x:()=>ug(e,r,a,s,i)}}},dg={kernelName:it,inputsToSave:["a","b"],gradFunc:(e,t,n)=>{let[r,a]=t,{transposeA:s,transposeB:i}=n;return s||i?!s&&i?{a:()=>Zo(e,a,!1,!1),b:()=>Zo(e,r,!0,!1)}:s&&!i?{a:()=>Zo(a,e,!1,!0),b:()=>Zo(r,e,!1,!1)}:{a:()=>Zo(a,e,!0,!0),b:()=>Zo(e,r,!0,!0)}:{a:()=>Zo(e,a,!1,!0),b:()=>Zo(r,e,!0,!1)}}},pg={kernelName:ot,gradFunc:(e,t,n)=>{let{blockShape:r,crops:a}=n;return{x:()=>vh(e,r,a)}}},cg={kernelName:ht,gradFunc:(e,t,n)=>{let r=n,a=r.inputShape,s=r.shape,i=Array.from(s);for(let e=a.length-1;e>=0;e--)if(a[e]===s[e])i[e]=1;else if(1!==a[e])throw new Error(`broadcastTo(): [${a}] cannot be broadcast to [${s}].`);let o=[];for(let e=0;e<i.length;e++)i[e]>1&&o.push(e);return{x:()=>ou(e,o,!0)}}},fg={kernelName:pt,gradFunc:e=>({x:()=>e.clone()})},mg={kernelName:ct,gradFunc:e=>({x:()=>Bl(e)})},gg={kernelName:ft,inputsToSave:["x"],gradFunc:(e,t,n)=>{let[r]=t,{clipValueMin:a,clipValueMax:s}=n;return{x:()=>zl(Gu(xu(r,a),Nu(r,s)),e,Bl(e))}}},yg={kernelName:gt,inputsToSave:["x"],gradFunc:Xm.gradFunc},bg={kernelName:yt,saveAllInputs:!0,gradFunc:(e,t,n)=>{let r=t.map(e=>e.shape),{axis:a}=n,s=le(a,t[0].shape)[0],i=r.map(e=>e[s]);return Id(e,i,s).map(e=>()=>e)}},xg={kernelName:bt,inputsToSave:["x","filter"],gradFunc:(e,t,n)=>{let[r,a]=t,{dilations:s,strides:i,pad:o,dataFormat:l}=n;return K(Wo(s),()=>`Error in gradient of conv2D: dilation rates greater than 1 are not yet supported in gradients. Got dilations '${s}'`),{x:()=>vl(r.shape,e,a,i,o,l),filter:()=>lp(r,e,a.shape,i,o,l)}}},vg={kernelName:vt,inputsToSave:["dy","filter"],gradFunc:(e,t,n)=>{let[r,a]=t,{strides:s,pad:i,dataFormat:o,dimRoundingMode:l}=n;return{dy:()=>bl(e,a,s,i,o,1,l),filter:()=>lp(e,r,a.shape,s,i,o,l)}}};var wg=zs({conv3DBackpropFilter_:function(e,t,n,r,a){let s=e;4===e.rank&&(s=jo(e,[1,e.shape[0],e.shape[1],e.shape[2],e.shape[3]]));let i=t;4===i.rank&&(i=jo(t,[1,t.shape[0],t.shape[1],t.shape[2],t.shape[3]])),K(5===s.rank,()=>`Error in conv3dDerFilter: input must be rank 5, but got shape ${s.shape}.`),K(5===i.rank,()=>`Error in conv3dDerFilter: dy must be rank 5, but got shape ${i.shape}.`),K(5===n.length,()=>`Error in conv3dDerFilter: filterShape must be length 5, but got ${n}.`),K(s.shape[4]===n[3],()=>`Error in conv3dDerFilter: depth of input ${s.shape[4]}) must match input depth in filter (${n[3]}.`),K(i.shape[4]===n[4],()=>`Error in conv3dDerFilter: depth of dy (${i.shape[4]}) must match output depth for filter (${n[4]}).`);let o={x:s,dy:i},l={strides:r,pad:a,filterShape:n};return _s.runKernel(kt,o,l)}}),kg={kernelName:wt,inputsToSave:["x","filter"],gradFunc:(e,t,n)=>{let{dilations:r,strides:a,pad:s}=n;K(Wo(r),()=>`Error in gradient of conv3D: dilation rates greater than 1 are not yet supported in gradients. Got dilations '${r}'`);let[i,o]=t;return{x:()=>Il(i.shape,e,o,a,s),filter:()=>wg(i,e,o.shape,a,s)}}},Ig={kernelName:St,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>yo(Pu(cd(ho(n,"float32"))),e)}}},Sg={kernelName:_t,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>yo(fd(ho(n,"float32")),e)}}},_g={kernelName:Tt,inputsToSave:["x"],gradFunc:(e,t,n)=>{let[r]=t,{axis:a,exclusive:s,reverse:i}=n;return{x:()=>{let t=Jl([a],r.rank),n=Cl(e,a,s,!i);return null!=t&&(n=Jd(n,t)),n}}}},Ng={kernelName:$t,inputsToSave:["x","filter"],gradFunc:(e,t,n)=>{let{dilations:r,strides:a,pad:s,dimRoundingMode:i}=n,o=r??[1,1];K(Wo(o),()=>`Error in gradient of depthwiseConv2dNative: dilation rates greater than 1 are not yet supported. Got dilations '${o}'`);let[l,u]=t;return K(4===l.rank,()=>`Error in gradient of depthwiseConv2dNative: input must be rank 4, but got rank ${l.rank}.`),K(4===u.rank,()=>`Error in gradient of depthwiseConv2dNative: filter must be rank 4, but got rank ${u.rank}.`),K(l.shape[3]===u.shape[2],()=>`Error in gradient of depthwiseConv2d: number of input channels (${l.shape[3]}) must match the inChannels dimension in filter ${u.shape[2]}.`),K(Vo(a,o),()=>`Error in gradient of depthwiseConv2d: Either strides or dilations must be  1. Got strides ${a} and dilations '${o}'.`),Ho("depthwiseConv2d",s,i),{x:()=>mp(l.shape,e,u,a,s,o,i),filter:()=>fp(l,e,u.shape,a,s,o,i)}}},Tg={kernelName:Mt,inputsToSave:["x","filter"],gradFunc:(e,t,n)=>{let[r,a]=t,s={x:r,filter:a,dy:e},i={x:r,filter:a,dy:e};return{x:()=>_s.runKernel(Ot,s,n),filter:()=>_s.runKernel(Lt,i,n)}}},Cg={kernelName:Wt,outputsToSave:[!0],gradFunc:(e,t)=>{let[n]=t,r={dy:e,y:n};return{x:()=>_s.runKernel(Vt,r)}}},Eg={kernelName:Ut,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t,r=yo(du(Pu(iu(n))),2/Math.sqrt(Math.PI));return{x:()=>yo(e,r)}}},Ag={kernelName:Ht,outputsToSave:[!0],gradFunc:(e,t)=>{let[n]=t;return{x:()=>yo(e,n)}}},$g={kernelName:jt,inputsToSave:["input"],gradFunc:(e,t)=>{let[n]=t;return{input:()=>jo(e,n.shape)}}},Rg={kernelName:qt,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>yo(e,du(n))}}},Fg={kernelName:Yt,gradFunc:e=>({x:()=>Bl(e)})},Dg={kernelName:Jt,inputsToSave:["a","b"],gradFunc:(e,t)=>{let[n,r]=t,a=Ll(n.shape,r.shape);return{a:()=>{let t=go(e,ho(r,"float32")),s=Ol(n.shape,a);return s.length>0?jo(ou(t,s),n.shape):t},b:()=>{let t=yo(e,ho(n,"float32")),s=Ol(r.shape,a);s.length>0&&(t=jo(ou(t,s),r.shape));let i=iu(r);return Pu(go(t,ho(i,"float32")))}}}},Mg={kernelName:Qt,inputsToSave:["x","mean","variance","scale"],gradFunc:(e,t,n)=>{let{varianceEpsilon:r}=n,[a,s,i,o]=t,l=o??au(1),u=Ol(s.shape,a.shape),h=[];if(1===s.rank){for(let e=0;e<a.shape.length-1;++e)h.push(a.shape[e]);h.push(1)}let d=Wu(a,s),p=yo(e,l),c=ld(fo(i,au(r))),f=yo(yo(yo(c,c),c),au(-.5));return{x:()=>1===s.rank?jo(yo(yo(e,fu(jo(c,[1,1,1,s.shape[0]]),h)),l),a.shape):jo(yo(yo(e,c),l),a.shape),mean:()=>{let e=yo(yo(c,au(-1)),p);return 1===s.rank&&(e=ou(e,u)),jo(e,s.shape)},variance:()=>{let e=yo(yo(f,d),p);return 1===s.rank&&(e=ou(e,u)),jo(e,s.shape)},scale:()=>{let t=yo(d,c),n=yo(e,t);return 1===s.rank&&(n=ou(n,u)),jo(n,s.shape)},offset:()=>{let t=e;return 1===s.rank&&(t=ou(t,u)),jo(t,s.shape)}}}},Og={kernelName:en,inputsToSave:["x","indices"],gradFunc:(e,t,n)=>{let[r,a]=t,{axis:s,batchDims:i}=n,o=le(s,r.shape)[0],l=(e,t,n)=>()=>{let r=e.shape,a=t.size,i=r.slice(0,o),l=i.length,u=r.slice(s,r.length).slice(1),h=u.length,d=Lg(0,l),p=Lg(l+1,l+1+h),c=Pg([i,[a],u]),f=jo(n,c),m=jo(t,[a]),g=Pg([[l],d,p]),y=Jd(f,g),b=Hd(y,m,e.shape[o]),x=Ql(g);return b=Jd(b,x),b};if(1===i){let t=r.shape[0],n=r.split(t,0);return{x:()=>Td(n.map((t,n)=>l(t,a.slice(n,1),e.slice(n,1))())).reshape(r.shape),indices:()=>a}}return{x:l(r,a,e),indices:()=>a}}};function Lg(e,t){let n=[];for(let r=e;r<t;++r)n.push(r);return n}function Pg(e){let 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}var zg={kernelName:rn,inputsToSave:["a","b"],gradFunc:(e,t)=>{let[n,r]=t;return{a:()=>Bl(n),b:()=>Bl(r)}}},Bg={kernelName:an,gradFunc:e=>({x:()=>ho(e,"float32")})},Wg={kernelName:ln,gradFunc:e=>({x:()=>Bl(e)})},Vg={kernelName:un,gradFunc:e=>({x:()=>Bl(e)})},Ug={kernelName:hn,gradFunc:e=>({x:()=>Bl(e)})},Gg={kernelName:dn,inputsToSave:["x"],gradFunc:(e,t,n)=>{let[r]=t,{alpha:a}=n,s=bu(r,0);return{x:()=>zl(s,e,yo(e,a))}}},Hg={kernelName:gn,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>go(e,fo(n,1))}}},jg={kernelName:mn,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>go(e,ho(n,"float32"))}}},qg={kernelName:wn,inputsToSave:[],outputsToSave:[!0],gradFunc:(e,t,n)=>{let[r]=t,{axis:a}=n;return{logits:()=>{let t=du(r);return Wu(e,yo(ou(e,a,!0),t))}}}};var Kg=zs({localResponseNormalizationBackprop_:function(e,t,n,r=5,a=1,s=1,i=.5){let o={x:e,y:t,dy:n},l={depthRadius:r,bias:a,alpha:s,beta:i};return _s.runKernel(Sn,o,l)}}),Xg={kernelName:In,inputsToSave:["x"],outputsToSave:[!0],gradFunc:(e,t,n)=>{let[r,a]=t,{depthRadius:s,bias:i,alpha:o,beta:l}=n;return{x:()=>Kg(r,a,e,s,i,o,l)}}};function Zg(e,t,n,r){return t.rank<n.rank&&(t=jo(t,Zl(t.shape,r))),e.rank<n.rank&&(e=jo(e,Zl(e.shape,r))),{x:()=>yo(e,ho(Pl(n,t),e.dtype))}}var Yg={kernelName:Nn,inputsToSave:["x"],outputsToSave:[!0],gradFunc:(e,t,n)=>{let r=n,{reductionIndices:a}=r,s=t[0],i=Zg(e,t[1],s,le(a,s.shape));return{x:()=>i.x()}}},Jg={kernelName:Tn,inputsToSave:["a","b"],gradFunc:(e,t)=>{let[n,r]=t;return{a:()=>yo(e,ho(xu(n,r),"float32")),b:()=>yo(e,ho(_u(n,r),"float32"))}}};var Qg=zs({maxPool3dGrad_:function(e,t,n,r,a,s,i){let o=Os(e,"dy","maxPool3dGrad"),l=Os(t,"input","maxPool3dGrad"),u=Os(n,"output","maxPool3dGrad"),h=o,d=l,p=u,c=!1;4===l.rank&&(c=!0,h=jo(o,[1,o.shape[0],o.shape[1],o.shape[2],o.shape[3]]),d=jo(l,[1,l.shape[0],l.shape[1],l.shape[2],l.shape[3]]),p=jo(u,[1,u.shape[0],u.shape[1],u.shape[2],u.shape[3]])),K(5===h.rank,()=>`Error in maxPool3dGrad: dy must be rank 5 but got rank ${h.rank}.`),K(5===d.rank,()=>`Error in maxPool3dGrad: input must be rank 5 but got rank ${d.rank}.`),K(5===p.rank,()=>`Error in maxPool3dGrad: output must be rank 5 but got rank ${p.rank}.`),Ho("maxPool3dGrad",s,i);let f={dy:h,input:d,output:p},m={filterSize:r,strides:a,pad:s,dimRoundingMode:i},g=_s.runKernel($n,f,m);return c?jo(g,[g.shape[1],g.shape[2],g.shape[3],g.shape[4]]):g}}),ey={kernelName:An,inputsToSave:["x"],outputsToSave:[!0],gradFunc:(e,t,n)=>{let[r,a]=t,{filterSize:s,strides:i,pad:o,dimRoundingMode:l}=n;return{x:()=>Qg(e,r,a,s,i,o,l)}}};var ty=zs({maxPoolGrad_:function(e,t,n,r,a,s,i){let o=Os(e,"dy","maxPoolGrad"),l=Os(t,"input","maxPoolGrad"),u=Os(n,"output","maxPoolGrad");K(l.rank===o.rank,()=>`Rank of input (${l.rank}) does not match rank of dy (${o.rank})`),K(4===o.rank,()=>`Error in maxPoolGrad: dy must be rank 4 but got rank ${o.rank}.`),K(4===l.rank,()=>`Error in maxPoolGrad: input must be rank 4 but got rank ${l.rank}.`),Ho("maxPoolGrad",s,i);let h={dy:o,input:l,output:u},d={filterSize:r,strides:a,pad:s,dimRoundingMode:i};return _s.runKernel(En,h,d)}}),ny={kernelName:Cn,inputsToSave:["x"],outputsToSave:[!0],gradFunc:(e,t,n)=>{let[r,a]=t,{filterSize:s,strides:i,pad:o}=n;return{x:()=>ty(e,r,a,s,i,o)}}},ry={kernelName:Fn,inputsToSave:["x"],gradFunc:(e,t,n)=>{let[r]=t,{axis:a}=n,s=le(a,r.shape),i=Y(Xl(r.shape,s)[1]);return{x:()=>{let t=r.shape.slice();s.forEach(e=>{t[e]=1});let n=jo(e,t);return go(yo(n,rh(r.shape,"float32")),i)}}}},ay={kernelName:Dn,inputsToSave:["x"],outputsToSave:[!0],gradFunc:(e,t,n)=>{let r=n,{axis:a}=r,[s,i]=t,o=Zg(e,i,s,le(a,s.shape));return{x:()=>o.x()}}},sy={kernelName:Mn,inputsToSave:["a","b"],gradFunc:(e,t)=>{let[n,r]=t;return{a:()=>yo(e,ho(Nu(n,r),"float32")),b:()=>yo(e,ho(bu(n,r),"float32"))}}},iy={kernelName:On,inputsToSave:["x"],gradFunc:(e,t,n)=>{let r=t[0],{paddings:a}=n,s=a.map(e=>e[0]);return{x:()=>Jo(e,s,r.shape)}}},oy={kernelName:Ln,inputsToSave:["a","b"],gradFunc:(e,t)=>{let[n,r]=t,a=Ll(n.shape,r.shape);return{a:()=>{let t=Ol(n.shape,a);return t.length>0?jo(ou(e,t),n.shape):e},b:()=>{let t=yo(e,Pu(gu(go(n,r)))),s=Ol(r.shape,a);return s.length>0?jo(ou(t,s),r.shape):t}}}},ly={kernelName:zn,inputsToSave:["a","b"],gradFunc:(e,t)=>{let[n,r]=t,a=Ll(n.shape,r.shape);return{a:()=>{let t=yo(e,ho(r,"float32")),s=Ol(n.shape,a);return s.length>0?jo(ou(t,s),n.shape):t},b:()=>{let t=yo(e,ho(n,"float32")),s=Ol(r.shape,a);return s.length>0?jo(ou(t,s),r.shape):t}}}},uy={kernelName:Bn,gradFunc:e=>({x:()=>Pu(e)})},hy={kernelName:jn,inputsToSave:["indices"],gradFunc:(e,t)=>{let n=t[0];return{indices:()=>nh(n.shape,"float32")}}},dy={kernelName:Hn,gradFunc:e=>({x:()=>Bl(e)})},py={kernelName:qn,saveAllInputs:!0,gradFunc:(e,t,n)=>{let{axis:r}=n;return jd(e,r).map(e=>()=>e)}},cy={kernelName:Kn,inputsToSave:["x"],gradFunc:(e,t,n)=>{let r=t[0],{paddings:a}=n,s=a.map(e=>e[0]);return{x:()=>Jo(e,s,r.shape)}}},fy={kernelName:Zn,inputsToSave:["a","b"],outputsToSave:[!0],gradFunc:(e,t)=>{let[n,r,a]=t,s=n,i=r,o=Ll(s.shape,i.shape);return{a:()=>{let t=ho(i,"float32"),n=yo(e,yo(t,ru(s,Wu(t,au(1))))),r=Ol(s.shape,o);return r.length>0&&(n=ou(n,r)),jo(n,s.shape)},b:()=>{let t=bu(s,0),n=zl(t,Eu(s),Bl(s)),r=yo(e,yo(a,n)),l=Ol(i.shape,o);return l.length>0&&(r=ou(r,l)),jo(r,i.shape)}}}},my={kernelName:Yn,inputsToSave:["x","alpha"],gradFunc:(e,t)=>{let[n,r]=t,a=bu(n,0);return{x:()=>zl(a,e,yo(e,r)),alpha:()=>{let t=zl(a,Bl(e),yo(e,n)),s=Ol(r.shape,e.shape);return s.length>0&&(t=ou(t,s)),jo(t,r.shape)}}}};function gy(e,t,n){let r=e.shape.length,a=r-n.length,s=Gf.getAxesPermutation(n,r),i=e;null!=s&&(i=Jd(e,s));let o=i.shape.slice(),l=o.splice(r-n.length,n.length).reduce((e,t)=>e*t,1);o.push(l);let u=function(e,t,n){let r=e.shape.slice();r[n]=1;let a=jo(t,r),s=Tl(e,n,!0,!1),i=Tl(e,n,!0,!0),o=yo(s,i);return yo(a,o)}(i.reshape(o),t,a);if(u=u.reshape(i.shape),null!=s){let e=Gf.getUndoAxesPermutation(s);u=Jd(u,e)}return u}var yy={kernelName:Jn,inputsToSave:["x"],gradFunc:(e,t,n)=>{let[r]=t,{axis:a}=n,s=[];return s=null==a?r.shape.map((e,t)=>t):"number"==typeof a?[a]:a,{x:()=>gy(r,e,s)}}},by={kernelName:zt,inputsToSave:["a","b"],gradFunc:(e,t)=>{let[n,r]=t,a=Ll(n.shape,r.shape);return{a:()=>{let t=go(e,ho(r,"float32")),s=Ol(n.shape,a);return s.length>0?jo(ou(t,s),n.shape):t},b:()=>{let t=yo(e,ho(n,"float32")),s=Ol(r.shape,a);s.length>0&&(t=jo(ou(t,s),r.shape));let i=iu(r);return Pu(go(t,ho(i,"float32")))}}}},xy={kernelName:ar,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>go(e,Pu(iu(n)))}}},vy={kernelName:dr,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t,r=yo(Nu(n,6),Cd(n));return{x:()=>yo(e,ho(r,"float32"))}}},wy={kernelName:sr,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>yo(e,ho(Cd(n),"float32"))}}},ky={kernelName:ir,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>jo(e,n.shape)}}},Iy={kernelName:ur,inputsToSave:["images"],gradFunc:(e,t,n)=>{let[r]=t,a={dy:e,images:r};return{images:()=>_s.runKernel(hr,a,n)}}},Sy={kernelName:or,inputsToSave:["images"],gradFunc:(e,t,n)=>{let[r]=t,a={dy:e,images:r};return{images:()=>_s.runKernel(lr,a,n)}}},_y={kernelName:pr,gradFunc:(e,t,n)=>{let{dims:r}=n,a=le(r,e.shape);return{x:()=>nd(e,a)}}},Ny={kernelName:cr,gradFunc:e=>({x:()=>Bl(e)})},Ty={kernelName:fr,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>Pu(go(e,yo(ru(n,1.5),2)))}}},Cy={kernelName:br,inputsToSave:["condition"],gradFunc:(e,t)=>{let[n]=t;return{condition:()=>ho(Bl(n),"float32"),t:()=>yo(e,ho(n,e.dtype)),e:()=>yo(e,ho(Hu(n),e.dtype))}}},Ey={kernelName:xr,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>{let t=bu(n,au(0)),r=au(sm),a=au(im),s=yo(e,a),i=yo(yo(e,r),du(ho(n,"float32")));return zl(t,s,i)}}}},Ay={kernelName:Sr,outputsToSave:[!0],gradFunc:(e,t)=>{let[n]=t;return{x:()=>yo(e,yo(n,Wu(au(1),n)))}}},$y={kernelName:Ir,gradFunc:e=>({x:()=>Bl(e)})},Ry={kernelName:wr,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>yo(_l(ho(n,"float32")),e)}}},Fy={kernelName:kr,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>yo(Nl(ho(n,"float32")),e)}}},Dy={kernelName:vr,inputsToSave:["x"],gradFunc:(e,t,n)=>{let[r]=t,{begin:a,size:s}=n,i=r.shape,[o,l]=Mf(r,a,s),u=[];for(let t=0;t<e.rank;t++)u.push([o[t],i[t]-o[t]-l[t]]);return{x:()=>mh(e,u)}}},My={kernelName:Ar,outputsToSave:[!0],gradFunc:(e,t,n)=>{let[r]=t,{dim:a}=n,s=yo(e,r);return{logits:()=>Wu(s,yo(ou(s,[a],true),r))}}},Oy={kernelName:_r,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>yo(e,Yo(n))}}},Ly={kernelName:Cr,gradFunc:(e,t,n)=>{let{blockShape:r,paddings:a}=n;return{x:()=>tl(e,r,a)}}},Py={kernelName:Er,gradFunc:(e,t,n)=>{let{axis:r}=n;return{x:()=>Xo(e,r)}}},zy={kernelName:Nr,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>go(e,yo(su(ho(n,"float32")),2))}}},By={kernelName:Lr,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>yo(e,yo(ho(n,"float32"),2))}}},Wy={kernelName:Or,inputsToSave:["a","b"],gradFunc:(e,t)=>{let[n,r]=t,a=au(2);return{a:()=>yo(e,yo(a,Wu(n,r))),b:()=>yo(e,yo(a,Wu(r,n)))}}},Vy={kernelName:ta,gradFunc:e=>({x:()=>Bl(e)})},Uy={kernelName:Ur,inputsToSave:["a","b"],gradFunc:(e,t)=>{let[n,r]=t,a=Ll(n.shape,r.shape);return{a:()=>{let t=e,r=Ol(n.shape,a);return r.length>0&&(t=ou(t,r)),jo(t,n.shape)},b:()=>{let t=e,n=Ol(r.shape,a);return n.length>0&&(t=ou(t,n)),jo(Pu(t),r.shape)}}}},Gy={kernelName:Tr,inputsToSave:["x"],gradFunc:(e,t,n)=>{let[r]=t,a=r.shape.slice(),{axis:s}=n;le(s,r.shape).forEach(e=>{a[e]=1});let i=jo(e,a),o=yo(i,rh(r.shape,"float32"));return{x:()=>o}}},Hy={kernelName:Gr,inputsToSave:["x"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>go(e,iu(_l(n)))}}},jy={kernelName:Hr,outputsToSave:[!0],gradFunc:(e,t)=>{let[n]=t;return{x:()=>yo(Wu(au(1),iu(n)),e)}}},qy={kernelName:jr,inputsToSave:["x"],gradFunc:(e,t,n)=>{let[r]=t,{reps:a}=n;return{x:()=>{let t=Bl(r);if(1===r.rank)for(let n=0;n<a[0];++n)t=fo(t,Jo(e,[n*r.shape[0]],[r.shape[0]]));else if(2===r.rank)for(let n=0;n<a[0];++n)for(let s=0;s<a[1];++s)t=fo(t,Jo(e,[n*r.shape[0],s*r.shape[1]],[r.shape[0],r.shape[1]]));else if(3===r.rank)for(let n=0;n<a[0];++n)for(let s=0;s<a[1];++s)for(let i=0;i<a[2];++i)t=fo(t,Jo(e,[n*r.shape[0],s*r.shape[1],i*r.shape[2]],[r.shape[0],r.shape[1],r.shape[2]]));else{if(4!==r.rank)throw new Error(`Gradient for tile operation is not implemented for rank-${r.rank} tensors yet.`);for(let n=0;n<a[0];++n)for(let s=0;s<a[1];++s)for(let i=0;i<a[2];++i)for(let o=0;o<a[3];++o)t=fo(t,Jo(e,[n*r.shape[0],s*r.shape[1],i*r.shape[2],o*r.shape[3]],[r.shape[0],r.shape[1],r.shape[2],r.shape[3]]))}return t}}}},Ky={kernelName:Xr,gradFunc:(e,t,n)=>{let r=n,{perm:a}=r,s=Ql(a);return{x:()=>Jd(e,s)}}},Xy={kernelName:Yr,gradFunc:(e,t,n)=>{let r=n,{axis:a}=r;return{value:()=>Td(e,a)}}},Zy={kernelName:Jr,inputsToSave:["segmentIds"],gradFunc:(e,t)=>{let[n]=t;return{x:()=>function(e,t){let n=eh(t,Bl(t)),r=yu(e,n),a=xu(t,au(0,"int32")),s=r.rank-a.rank;for(let e=0;e<s;++e)a=pu(a,e+1);a=Gu(a,rh(r.shape,"bool"));let i=Bl(r);return zl(a,r,i)}(e,n)}}};var Yy={kernelName:ea,gradFunc:e=>({x:()=>Bl(e)})},Jy=[Xm,Zm,Ym,Jm,Qm,eg,tg,ng,rg,ag,sg,ig,lg,hg,dg,pg,cg,fg,mg,gg,yg,bg,vg,xg,kg,Ig,Sg,_g,Ng,Tg,by,Cg,Eg,Ag,$g,Rg,Dg,Fg,Mg,Og,zg,Bg,Wg,Vg,Ug,Gg,Hg,jg,qg,Xg,Yg,Yg,Jg,ey,ny,ry,ay,sy,iy,oy,ly,uy,hy,dy,py,cy,cy,fy,my,yy,xy,vy,wy,ky,Iy,Sy,_y,Ny,Ty,Cy,Ey,Ay,$y,Ry,Fy,Dy,My,Oy,Ly,Ly,Py,Py,zy,Wy,By,Vy,Uy,Gy,Hy,jy,qy,Ky,Xy,Zy,Yy];for(let e of Jy)ma(e);ns().prototype.abs=function(){return this.throwIfDisposed(),bo(this)},ns().prototype.acos=function(){return this.throwIfDisposed(),xo(this)},ns().prototype.acosh=function(){return this.throwIfDisposed(),vo(this)},ns().prototype.add=function(e){return this.throwIfDisposed(),fo(this,e)},ns().prototype.all=function(e,t){return this.throwIfDisposed(),ko(this,e,t)},ns().prototype.any=function(e,t){return this.throwIfDisposed(),Io(this,e,t)},ns().prototype.argMax=function(e){return this.throwIfDisposed(),So(this,e)},ns().prototype.argMin=function(e){return this.throwIfDisposed(),_o(this,e)},ns().prototype.asScalar=function(){return this.throwIfDisposed(),K(1===this.size,()=>"The array must have only 1 element."),jo(this,[])},ns().prototype.asType=function(e){return this.throwIfDisposed(),ho(this,e)},ns().prototype.as1D=function(){return this.throwIfDisposed(),jo(this,[this.size])},ns().prototype.as2D=function(e,t){return this.throwIfDisposed(),jo(this,[e,t])},ns().prototype.as3D=function(e,t,n){return this.throwIfDisposed(),jo(this,[e,t,n])},ns().prototype.as4D=function(e,t,n,r){return this.throwIfDisposed(),jo(this,[e,t,n,r])},ns().prototype.as5D=function(e,t,n,r,a){return this.throwIfDisposed(),jo(this,[e,t,n,r,a])},ns().prototype.asin=function(){return this.throwIfDisposed(),No(this)},ns().prototype.asinh=function(){return this.throwIfDisposed(),To(this)},ns().prototype.atan=function(){return this.throwIfDisposed(),Co(this)},ns().prototype.atan2=function(e){return this.throwIfDisposed(),Eo(this,e)},ns().prototype.atanh=function(){return this.throwIfDisposed(),Ao(this)},ns().prototype.avgPool=function(e,t,n,r){return this.throwIfDisposed(),qo(this,e,t,n,r)},ns().prototype.batchToSpaceND=function(e,t){return this.throwIfDisposed(),tl(this,e,t)},ns().prototype.batchNorm=function(e,t,n,r,a){return this.throwIfDisposed(),rl(this,e,t,n,r,a)},ns().prototype.broadcastTo=function(e){return this.throwIfDisposed(),hl(this,e)},ns().prototype.cast=function(e){return this.throwIfDisposed(),ho(this,e)},ns().prototype.ceil=function(){return this.throwIfDisposed(),dl(this)},ns().prototype.clipByValue=function(e,t){return this.throwIfDisposed(),cl(this,e,t)},ns().prototype.concat=function(e,t){return this.throwIfDisposed(),e instanceof ts&&(e=[e]),Xo([this,...e],t)},ns().prototype.conv1d=function(e,t,n,r,a,s){return this.throwIfDisposed(),xl(this,e,t,n,r,a,s)},ns().prototype.conv2dTranspose=function(e,t,n,r,a){return this.throwIfDisposed(),wl(this,e,t,n,r,a)},ns().prototype.conv2d=function(e,t,n,r,a,s){return this.throwIfDisposed(),bl(this,e,t,n,r,a,s)},ns().prototype.cos=function(){return this.throwIfDisposed(),_l(this)},ns().prototype.cosh=function(){return this.throwIfDisposed(),Nl(this)},ns().prototype.cumprod=function(e,t,n){return this.throwIfDisposed(),Tl(this,e,t,n)},ns().prototype.cumsum=function(e,t,n){return this.throwIfDisposed(),Cl(this,e,t,n)},ns().prototype.depthToSpace=function(e,t){return this.throwIfDisposed(),Al(this,e,t)},ns().prototype.depthwiseConv2d=function(e,t,n,r,a,s){return this.throwIfDisposed(),$l(this,e,t,n,r,a,s)},ns().prototype.dilation2d=function(e,t,n,r,a){return this.throwIfDisposed(),Fl(this,e,t,n,r,a)},ns().prototype.divNoNan=function(e){return this.throwIfDisposed(),Wl(this,e)},ns().prototype.div=function(e){return this.throwIfDisposed(),go(this,e)},ns().prototype.dot=function(e){return this.throwIfDisposed(),Vl(this,e)},ns().prototype.elu=function(){return this.throwIfDisposed(),Gl(this)},ns().prototype.equal=function(e){return this.throwIfDisposed(),Pl(this,e)},ns().prototype.erf=function(){return this.throwIfDisposed(),jl(this)},ns().prototype.euclideanNorm=function(e,t){return this.throwIfDisposed(),hu(this,e,t)},ns().prototype.exp=function(){return this.throwIfDisposed(),du(this)},ns().prototype.expandDims=function(e){return this.throwIfDisposed(),pu(this,e)},ns().prototype.expm1=function(){return this.throwIfDisposed(),cu(this)},ns().prototype.fft=function(){return this.throwIfDisposed(),vd(this)},ns().prototype.flatten=function(){return this.throwIfDisposed(),jo(this,[this.size])},ns().prototype.floor=function(){return this.throwIfDisposed(),gu(this)},ns().prototype.floorDiv=function(e){return this.throwIfDisposed(),mo(this,e)},ns().prototype.gather=function(e,t,n){return this.throwIfDisposed(),yu(this,e,t,n)},ns().prototype.greaterEqual=function(e){return this.throwIfDisposed(),xu(this,e)},ns().prototype.greater=function(e){return this.throwIfDisposed(),bu(this,e)},ns().prototype.ifft=function(){return this.throwIfDisposed(),wd(this)},ns().prototype.irfft=function(){return this.throwIfDisposed(),kd(this)},ns().prototype.isFinite=function(){return this.throwIfDisposed(),wu(this)},ns().prototype.isInf=function(){return this.throwIfDisposed(),ku(this)},ns().prototype.isNaN=function(){return this.throwIfDisposed(),Iu(this)},ns().prototype.leakyRelu=function(e){return this.throwIfDisposed(),Su(this,e)},ns().prototype.lessEqual=function(e){return this.throwIfDisposed(),Nu(this,e)},ns().prototype.less=function(e){return this.throwIfDisposed(),_u(this,e)},ns().prototype.localResponseNormalization=function(e,t,n,r){return this.throwIfDisposed(),Cu(this,e,t,n,r)},ns().prototype.logSigmoid=function(){return this.throwIfDisposed(),Bu(this)},ns().prototype.logSoftmax=function(e){return this.throwIfDisposed(),Vu(this,e)},ns().prototype.logSumExp=function(e,t){return this.throwIfDisposed(),Uu(this,e,t)},ns().prototype.log=function(){return this.throwIfDisposed(),Eu(this)},ns().prototype.log1p=function(){return this.throwIfDisposed(),Au(this)},ns().prototype.logicalAnd=function(e){return this.throwIfDisposed(),Gu(this,e)},ns().prototype.logicalNot=function(){return this.throwIfDisposed(),Hu(this)},ns().prototype.logicalOr=function(e){return this.throwIfDisposed(),ju(this,e)},ns().prototype.logicalXor=function(e){return this.throwIfDisposed(),qu(this,e)},ns().prototype.matMul=function(e,t,n){return this.throwIfDisposed(),Zo(this,e,t,n)},ns().prototype.maxPool=function(e,t,n,r){return this.throwIfDisposed(),Yu(this,e,t,n,r)},ns().prototype.max=function(e,t){return this.throwIfDisposed(),tu(this,e,t)},ns().prototype.maximum=function(e){return this.throwIfDisposed(),eh(this,e)},ns().prototype.mean=function(e,t){return this.throwIfDisposed(),th(this,e,t)},ns().prototype.min=function(e,t){return this.throwIfDisposed(),nu(this,e,t)},ns().prototype.minimum=function(e){return this.throwIfDisposed(),sh(this,e)},ns().prototype.mirrorPad=function(e,t){return this.throwIfDisposed(),ih(this,e,t)},ns().prototype.mod=function(e){return this.throwIfDisposed(),oh(this,e)},ns().prototype.mul=function(e){return this.throwIfDisposed(),yo(this,e)},ns().prototype.neg=function(){return this.throwIfDisposed(),Pu(this)},ns().prototype.norm=function(e,t,n){return this.throwIfDisposed(),uu(this,e,t,n)},ns().prototype.notEqual=function(e){return this.throwIfDisposed(),dh(this,e)},ns().prototype.oneHot=function(e,t=1,n=0){return this.throwIfDisposed(),ph(this,e,t,n)},ns().prototype.onesLike=function(){return this.throwIfDisposed(),ch(this)},ns().prototype.pad=function(e,t){return this.throwIfDisposed(),mh(this,e,t)},ns().prototype.pool=function(e,t,n,r,a,s){return this.throwIfDisposed(),wh(this,e,t,n,r,a,s)},ns().prototype.pow=function(e){return this.throwIfDisposed(),ru(this,e)},ns().prototype.prelu=function(e){return this.throwIfDisposed(),kh(this,e)},ns().prototype.prod=function(e,t){return this.throwIfDisposed(),Ih(this,e,t)},ns().prototype.reciprocal=function(){return this.throwIfDisposed(),Qh(this)},ns().prototype.relu=function(){return this.throwIfDisposed(),ed(this)},ns().prototype.relu6=function(){return this.throwIfDisposed(),td(this)},ns().prototype.reshapeAs=function(e){return this.throwIfDisposed(),jo(this,e.shape)},ns().prototype.reshape=function(e){return this.throwIfDisposed(),jo(this,e)},ns().prototype.resizeBilinear=function(e,t,n){return this.throwIfDisposed(),Up(this,e,t,n)},ns().prototype.resizeNearestNeighbor=function(e,t,n){return this.throwIfDisposed(),Gp(this,e,t,n)},ns().prototype.reverse=function(e){return this.throwIfDisposed(),nd(this,e)},ns().prototype.rfft=function(){return this.throwIfDisposed(),Sd(this)},ns().prototype.round=function(){return this.throwIfDisposed(),od(this)},ns().prototype.rsqrt=function(){return this.throwIfDisposed(),ld(this)},ns().prototype.selu=function(){return this.throwIfDisposed(),ud(this)},ns().prototype.separableConv2d=function(e,t,n,r,a,s){return this.throwIfDisposed(),hd(this,e,t,n,r,a,s)},ns().prototype.sigmoid=function(){return this.throwIfDisposed(),Yo(this)},ns().prototype.sign=function(){return this.throwIfDisposed(),pd(this)},ns().prototype.sin=function(){return this.throwIfDisposed(),cd(this)},ns().prototype.sinh=function(){return this.throwIfDisposed(),fd(this)},ns().prototype.slice=function(e,t){return this.throwIfDisposed(),Jo(this,e,t)},ns().prototype.softmax=function(e){return this.throwIfDisposed(),xd(this,e)},ns().prototype.softplus=function(){return this.throwIfDisposed(),zu(this)},ns().prototype.spaceToBatchND=function(e,t){return this.throwIfDisposed(),vh(this,e,t)},ns().prototype.split=function(e,t){return this.throwIfDisposed(),Id(this,e,t)},ns().prototype.sqrt=function(){return this.throwIfDisposed(),su(this)},ns().prototype.square=function(){return this.throwIfDisposed(),iu(this)},ns().prototype.squaredDifference=function(e){return this.throwIfDisposed(),_d(this,e)},ns().prototype.squeeze=function(e){return this.throwIfDisposed(),Nd(this,e)},ns().prototype.stack=function(e,t){this.throwIfDisposed();let n=e instanceof ts?[this,e]:[this,...e];return Td(n,t)},ns().prototype.step=function(e){return this.throwIfDisposed(),Cd(this,e)},ns().prototype.stridedSlice=function(e,t,n,r,a,s,i,o){return this.throwIfDisposed(),Ed(this,e,t,n,r,a,s,i,o)},ns().prototype.sub=function(e){return this.throwIfDisposed(),Wu(this,e)},ns().prototype.sum=function(e,t){return this.throwIfDisposed(),ou(this,e,t)},ns().prototype.tan=function(){return this.throwIfDisposed(),Ad(this)},ns().prototype.tanh=function(){return this.throwIfDisposed(),Qo(this)},ns().prototype.tile=function(e){return this.throwIfDisposed(),fu(this,e)},ns().prototype.toBool=function(){return this.throwIfDisposed(),ho(this,"bool")},ns().prototype.toFloat=function(){return this.throwIfDisposed(),ho(this,"float32")},ns().prototype.toInt=function(){return this.throwIfDisposed(),ho(this,"int32")},ns().prototype.topk=function(e,t){return this.throwIfDisposed(),Vd(this,e,t)},ns().prototype.transpose=function(e){return this.throwIfDisposed(),Jd(this,e)},ns().prototype.unique=function(e){return this.throwIfDisposed(),Gd(this,e)},ns().prototype.unsortedSegmentSum=function(e,t){return this.throwIfDisposed(),Hd(this,e,t)},ns().prototype.unstack=function(e){return this.throwIfDisposed(),jd(this,e)},ns().prototype.where=function(e,t){return this.throwIfDisposed(),zl(e,this,t)},ns().prototype.zerosLike=function(){return this.throwIfDisposed(),Bl(this)};var Qy=class e extends Error{constructor(t){super(t),Object.setPrototypeOf(this,e.prototype)}},eb=class e extends Error{constructor(t){super(t),Object.setPrototypeOf(this,e.prototype)}},tb=class e extends Error{constructor(t){super(t),Object.setPrototypeOf(this,e.prototype)}},nb=class e extends Error{constructor(t){super(t),Object.setPrototypeOf(this,e.prototype)}},rb=class e extends Error{constructor(t){super(t),Object.setPrototypeOf(this,e.prototype)}},ab=class{constructor(e){this.maxEntries=e||100,this.cache=new Map}get(e){let t;return this.cache.has(e)&&(t=this.cache.get(e),this.cache.delete(e),this.cache.set(e,t)),t}put(e,t){if(this.cache.has(e))this.cache.delete(e);else if(this.cache.size>=this.maxEntries){let e=this.cache.keys().next().value;this.cache.delete(e)}this.cache.set(e,t)}getMaxEntries(){return this.maxEntries}setMaxEntries(e){if(e<0)throw new Error(`The maxEntries of LRU caches must be at least 0, but got ${e}.`);if(this.maxEntries>e)for(let t=0;t<this.maxEntries-e;t++){let e=this.cache.keys().next().value;this.cache.delete(e)}this.maxEntries=e}};function sb(e,t){if(Array.isArray(e)){let n=[];for(let r=0;r<t;r++)n=n.concat(e);return n}{let n=new Array(t);return n.fill(e),n}}function ib(e,t){if(!e)throw new rb(t)}function ob(e,t){let n=0;for(let r of e)r===t&&n++;return n}function lb(e){return 1===e.length?e[0]:e}function ub(e){return Array.isArray(e)?e:[e]}function hb(e){let t=e.replace(/(.)([A-Z][a-z0-9]+)/g,"$1_$2").replace(/([a-z])([A-Z])/g,"$1_$2").toLowerCase();return"_"!==t[0]?t:"private"+t}function db(e){return e.length<=1||-1===e.indexOf("_")?e:e.replace(/[_]+(\w|$)/g,(e,t)=>t.toUpperCase())}var pb={};function cb(e){if(null==e)return null;let t={};return t.className=e.getClassName(),t.config=e.getConfig(),t}function fb(e){if(null!=e&&"object"==typeof e)if(Array.isArray(e))e.forEach(e=>fb(e));else{let t=Object.keys(e);for(let n of t){let t=e[n];null!=t&&"object"==typeof t&&(Array.isArray(t)||"ndarray"!==t.type||"number"!=typeof t.value?fb(t):e[n]=t.value)}}}function mb(e,t={},n={},r="object",a=!1){if("string"==typeof e){let a,s=e;if(s in n)a=n[s];else if(s in pb)a=pb[s];else if(a=t[s],null==a)throw new tb(`Unknown ${r}: ${e}. This may be due to one of the following reasons:\n1. The ${r} is defined in Python, in which case it needs to be ported to TensorFlow.js or your JavaScript code.\n2. The custom ${r} is defined in JavaScript, but is not registered properly with tf.serialization.registerClass().`);return a}{let s=e;if(null==s.className||null==s.config)throw new tb(`${r}: Improper config format: ${JSON.stringify(s)}.\n'className' and 'config' must set.`);let i,o,l=s.className;if(l in n?[i,o]=n[l]:l in pb?[i,o]=pb.className:l in t&&([i,o]=t[l]),null==i)throw new tb(`Unknown ${r}: ${l}. This may be due to one of the following reasons:\n1. The ${r} is defined in Python, in which case it needs to be ported to TensorFlow.js or your JavaScript code.\n2. The custom ${r} is defined in JavaScript, but is not registered properly with tf.serialization.registerClass().`);if(null!=o){let e={};for(let t of Object.keys(pb))e[t]=pb[t];for(let t of Object.keys(n))e[t]=n[t];s.config.customObjects=e;let t=Object.assign({},pb);for(let e of Object.keys(n))pb[e]=n[e];fb(s.config);let r=o(i,s.config,n,a);return pb=Object.assign({},t),r}{let e=Object.assign({},pb);for(let e of Object.keys(n))pb[e]=n[e];let t=new i(s.config);return pb=Object.assign({},e),t}}}function gb(e,t){return-1*function(e,t){return e<t?-1:e>t?1:0}(e,t)}function yb(e){if(null==e)return e;let t=[];for(let n of e)-1===t.indexOf(n)&&t.push(n);return t}function bb(e){if(null==e)throw new tb(`Invalid value in obj: ${JSON.stringify(e)}`);for(let t in e)if(e.hasOwnProperty(t))return!1;return!0}function xb(e,t,n){if(null!=n&&e.indexOf(n)<0)throw new tb(`${n} is not a valid ${t}.  Valid values are ${e} or null/undefined.`)}function vb(e,t,n=0,r=1/0){return ib(n>=0),ib(r>=n),Array.isArray(e)&&e.length>=n&&e.length<=r&&e.every(e=>typeof e===t)}function wb(e,t){Array.isArray(e)?(va.assert(e.length>0,()=>`${t} is unexpectedly an empty array.`),e.forEach((e,n)=>wb(e,`element ${n+1} of ${t}`))):va.assert(Number.isInteger(e)&&e>0,()=>`Expected ${t} to be a positive integer, but got ${kb(e)}.`)}function kb(e){return null===e?"null":Array.isArray(e)?"["+e.map(e=>kb(e)).join(",")+"]":"string"==typeof e?`"${e}"`:`${e}`}function Ib(e){return"relu"===e?"relu":"linear"===e?"linear":"elu"===e?"elu":null}var Sb=0;function _b(){return Sb++}var Nb={};function Tb(e=""){return e in Nb||(Nb[e]=0),Nb[e]+=1,e+Nb[e].toString()}var Cb=["channelsFirst","channelsLast"],Eb=["nearest","bilinear"],Ab=["valid","same","causal"],$b=["max","avg"],Rb=["sum","mul","concat","ave"],Fb=new Map;function Db(e){xb(Cb,"DataFormat",e)}function Mb(e){xb(Ab,"PaddingMode",e)}function Ob(e){xb($b,"PoolMode",e)}var Lb=[];function Pb(e,t){Lb.push(e);try{let e=t();return Lb.pop(),e}catch(e){throw Lb.pop(),e}}function zb(e){if(!Ub(e))throw new Error("Not a valid tensor name: '"+e+"'");return(0===Lb.length?"":Lb.join("/")+"/")+e}function Bb(e){if(!Ub(e))throw new Error("Not a valid tensor name: '"+e+"'");Fb.has(e)||Fb.set(e,0);let t=Fb.get(e);if(Fb.set(e,Fb.get(e)+1),t>0){let n=`${e}_${t}`;return Fb.set(n,1),n}return e}var Wb,Vb=new RegExp(/^[A-Za-z0-9][-A-Za-z0-9\._\/]*$/);function Ub(e){return!!e.match(Vb)}function Gb(e){return e===parseInt(e.toString(),10)}function Hb(e,t,n){null==t&&(t=0),null==n&&(n=e.length);let r=1;for(let a=t;a<n;++a)r*=e[a];return r}function jb(e){if(0===e.length)return Number.NaN;let t=Number.POSITIVE_INFINITY;for(let n=0;n<e.length;n++){let r=e[n];r<t&&(t=r)}return t}function qb(e){if(0===e.length)return Number.NaN;let t=Number.NEGATIVE_INFINITY;for(let n=0;n<e.length;n++){let r=e[n];r>t&&(t=r)}return t}function Kb(e,t){if(t<e)throw new tb(`end (${t}) < begin (${e}) is forbidden.`);let n=[];for(let r=e;r<t;++r)n.push(r);return n}function Xb(){return null==Wb&&(Wb=hi().epsilon()),Wb}function Zb(e,t){return ho(e,t)}function Yb(e,t=-1){let n=e.shape.slice();return t<0&&(t=n.length+t+1),n.splice(t,0,1),jo(e,n)}function Jb(e,t,n){return Qs(()=>{switch(e.rank){case 1:return md(e,t,n);case 2:return gd(e,[t,0],[n,e.shape[1]]);case 3:return yd(e,[t,0,0],[n,e.shape[1],e.shape[2]]);case 4:return bd(e,[t,0,0,0],[n,e.shape[1],e.shape[2],e.shape[3]]);case 5:return Jo(e,[t,0,0,0,0],[n,e.shape[1],e.shape[2],e.shape[3],e.shape[4]]);case 6:return Jo(e,[t,0,0,0,0,0],[n,e.shape[1],e.shape[2],e.shape[3],e.shape[4],e.shape[5]]);default:throw new tb(`sliceAlongFirstAxis() received an unsupported tensor rank: ${e.rank}`)}})}function Qb(e,t,n){return Qs(()=>{switch(e.rank){case 1:return md(e,t,n);case 2:return gd(e,[0,t],[e.shape[0],n]);case 3:return yd(e,[0,0,t],[e.shape[0],e.shape[1],n]);case 4:return bd(e,[0,0,0,t],[e.shape[0],e.shape[1],e.shape[2],n]);default:throw new tb(`sliceAlongLastAxis() received an unsupported tensor rank: ${e.rank}`)}})}function ex(e,t,n,r){return Qs(()=>{switch(e.rank){case 1:return md(e,t,n);case 2:switch(r){case 1:return Jb(e,t,n);case 2:return Qb(e,t,n);default:throw new tb(`The axis is not within the rank of the tensor ${r}`)}case 3:switch(r){case 1:return Jb(e,t,n);case 2:return yd(e,[0,t,0],[e.shape[0],n,e.shape[2]]);case 3:return Qb(e,t,n);default:throw new tb(`The axis is not within the rank of the tensor ${r}`)}case 4:switch(r){case 1:return Jb(e,t,n);case 2:return bd(e,[0,t,0,0],[e.shape[0],n,e.shape[2],e.shape[3]]);case 3:return bd(e,[0,0,t,0],[e.shape[0],e.shape[1],n,e.shape[3]]);case 4:return Qb(e,t,n);default:throw new tb(`The axis is not within the rank of the tensor ${r}`)}default:throw new tb(`sliceAlongLastAxis() received an unsupported tensor rank: ${e.rank}`)}})}function tx(e,t=-1){let n;return t<0&&(n=e[0].rank,t=0!==n?n:0),t===e[0].rank&&(t=-1),Xo(e,t)}function nx(e,t){switch(e.rank){case 1:return fl([e,t]);case 2:return ml([e,t],0);case 3:return gl([e,t],0);case 4:return yl([e,t],0);default:throw new tb(`concatAlongFirstAxis() received an unsupported tensor rank: ${e.rank}`)}}function rx(e,t){if(Array.isArray(t)||(t=[t]),e.rank!==t.length)throw new tb(`The length of input n (${t.length}) does not match the number of dimensions in input x (${e.rank})`);return fu(e,t)}function ax(e,t=0,n=1,r,a){return qh(e,t,n,r,a)}function sx(e,t,n,r){if(e.rank<2||t.rank<2)throw new nb(`dot requires both inputs to be rank >= 2 but got x shape = ${e.shape} and y shape = ${t.shape}`);if(t.rank>=3){if(e.shape.slice(-1)[0]!==t.shape.slice(-2)[0])throw new nb(`If rank y >= 3, then the second last dim of y must equal the last dim of x but got x shape = ${e.shape} and  y shape = ${t.shape}`)}if(2===e.rank&&2===t.rank)return op.matMul({a:e,b:t,transposeA:!1,transposeB:!1,bias:r?lx(e.rank,r,"channelsLast"):null,activation:n});{let a=e.shape.slice(),s=a.pop();e=jo(e,[-1,s]);let i=t.shape.slice(),o=i.pop(),l=i.pop(),u=[...i,o],h=Array.from({length:t.rank},(e,n)=>0===n?t.rank-2:n<=t.rank-2?n-1:n);t=jo(Jd(t,h),[l,-1]);let d=[...a,...u];return jo(op.matMul({a:e,b:t,transposeA:!1,transposeB:!1,bias:r?lx(e.rank,r,"channelsLast"):null,activation:n}),d)}}function ix(e,t,n){return Qs(()=>(t=Array.isArray(t)?$d(t,"int32"):ho(t,"int32"),yu(e,t,n)))}function ox(e){return yo(e,e)}function lx(e,t,n){let r=t.shape;if(1!==t.rank&&t.rank!==e)throw new tb(`Unexpected bias dimensions: ${t.rank}; expected it to be 1 or ${e}`);if(5===e){if("channelsFirst"===n)return 1===r.length?jo(t,[1,r[0],1,1,1]):jo(t,[1,r[3],r[0],r[1],r[2]]);if("channelsLast"===n)return 1===r.length?jo(t,[1,1,1,1,r[0]]):jo(t,[1].concat(r))}else if(4===e){if("channelsFirst"===n)return 1===r.length?jo(t,[1,r[0],1,1]):jo(t,[1,r[2],r[0],r[1]]);if("channelsLast"===n)return 1===r.length?jo(t,[1,1,1,r[0]]):jo(t,[1].concat(r))}else if(3===e){if("channelsFirst"===n)return 1===r.length?jo(t,[1,r[0],1]):jo(t,[1,r[1],r[0]]);if("channelsLast"===n)return 1===r.length?jo(t,[1,1,r[0]]):jo(t,[1].concat(r))}else if(e<3)return t;throw new tb(`Unsupported input rank by biasAdd: ${t.rank}`)}function ux(e,t,n){return Qs(()=>(null==n&&(n="channelsLast"),Db(n),fo(e,lx(e.rank,t,n))))}function hx(e,t,n,r){return Qs(()=>rp(e,t,n,r))}function dx(e,t,n=!1){return n?e():t()}var px=["fanIn","fanOut","fanAvg"],cx=["normal","uniform","truncatedNormal"];var fx=class extends kc.Serializable{fromConfigUsesCustomObjects(){return!1}getConfig(){return{}}},mx=class extends fx{apply(e,t){return nh(e,t)}};mx.className="Zeros",kc.registerClass(mx);var gx=class extends fx{apply(e,t){return rh(e,t)}};gx.className="Ones",kc.registerClass(gx);var yx=class extends fx{constructor(e){if(super(),"object"!=typeof e)throw new tb(`Expected argument of type ConstantConfig but got ${e}`);if(void 0===e.value)throw new tb(`config must have value set but got ${e}`);this.value=e.value}apply(e,t){return Qs(()=>yo(au(this.value),rh(e,t)))}getConfig(){return{value:this.value}}};yx.className="Constant",kc.registerClass(yx);var bx=class extends fx{constructor(e){super(),this.DEFAULT_MINVAL=-.05,this.DEFAULT_MAXVAL=.05,this.minval=e.minval||this.DEFAULT_MINVAL,this.maxval=e.maxval||this.DEFAULT_MAXVAL,this.seed=e.seed}apply(e,t){return Xh(e,this.minval,this.maxval,t,this.seed)}getConfig(){return{minval:this.minval,maxval:this.maxval,seed:this.seed}}};bx.className="RandomUniform",kc.registerClass(bx);var xx=class extends fx{constructor(e){super(),this.DEFAULT_MEAN=0,this.DEFAULT_STDDEV=.05,this.mean=e.mean||this.DEFAULT_MEAN,this.stddev=e.stddev||this.DEFAULT_STDDEV,this.seed=e.seed}apply(e,t){if("float32"!==(t=t||"float32")&&"int32"!==t)throw new nb(`randomNormal does not support dType ${t}.`);return ax(e,this.mean,this.stddev,t,this.seed)}getConfig(){return{mean:this.mean,stddev:this.stddev,seed:this.seed}}};xx.className="RandomNormal",kc.registerClass(xx);var vx=class extends fx{constructor(e){super(),this.DEFAULT_MEAN=0,this.DEFAULT_STDDEV=.05,this.mean=e.mean||this.DEFAULT_MEAN,this.stddev=e.stddev||this.DEFAULT_STDDEV,this.seed=e.seed}apply(e,t){if("float32"!==(t=t||"float32")&&"int32"!==t)throw new nb(`truncatedNormal does not support dType ${t}.`);return Ud(e,this.mean,this.stddev,t,this.seed)}getConfig(){return{mean:this.mean,stddev:this.stddev,seed:this.seed}}};vx.className="TruncatedNormal",kc.registerClass(vx);var wx=class extends fx{constructor(e){super(),this.gain=null!=e.gain?e.gain:1}apply(e,t){return Qs(()=>{if(2!==e.length||e[0]!==e[1])throw new tb("Identity matrix initializer can only be used for 2D square matrices.");return yo(this.gain,mu(e[0]))})}getConfig(){return{gain:this.gain}}};wx.className="Identity",kc.registerClass(wx);var kx=class extends fx{constructor(e){if(super(),e.scale<0)throw new tb(`scale must be a positive float. Got: ${e.scale}`);this.scale=null==e.scale?1:e.scale,this.mode=null==e.mode?"fanIn":e.mode,function(e){xb(px,"FanMode",e)}(this.mode),this.distribution=null==e.distribution?"normal":e.distribution,function(e){xb(cx,"Distribution",e)}(this.distribution),this.seed=e.seed}apply(e,t){let n=function(e,t="channelsLast"){let n,r;if(Db(t),2===e.length)n=e[0],r=e[1];else if(-1!==[3,4,5].indexOf(e.length)){if("channelsFirst"===t){let t=Hb(e,2);n=e[1]*t,r=e[0]*t}else if("channelsLast"===t){let t=Hb(e,0,e.length-2);n=e[e.length-2]*t,r=e[e.length-1]*t}}else{let t=Hb(e);n=Math.sqrt(t),r=Math.sqrt(t)}return[n,r]}(e),r=n[0],a=n[1],s=this.scale;if("fanIn"===this.mode?s/=Math.max(1,r):"fanOut"===this.mode?s/=Math.max(1,a):s/=Math.max(1,(r+a)/2),"normal"===this.distribution){let n=Math.sqrt(s);if("float32"!==(t=t||"float32")&&"int32"!==t)throw new nb(`${this.getClassName()} does not support dType ${t}.`);return Ud(e,0,n,t,this.seed)}{let n=Math.sqrt(3*s);return Xh(e,-n,n,t,this.seed)}}getConfig(){return{scale:this.scale,mode:this.mode,distribution:this.distribution,seed:this.seed}}};kx.className="VarianceScaling",kc.registerClass(kx);var Ix=class extends kx{constructor(e){super({scale:1,mode:"fanAvg",distribution:"uniform",seed:null==e?null:e.seed})}getClassName(){return kx.className}};Ix.className="GlorotUniform",kc.registerClass(Ix);var Sx=class extends kx{constructor(e){super({scale:1,mode:"fanAvg",distribution:"normal",seed:null==e?null:e.seed})}getClassName(){return kx.className}};Sx.className="GlorotNormal",kc.registerClass(Sx);var _x=class extends kx{constructor(e){super({scale:2,mode:"fanIn",distribution:"normal",seed:null==e?null:e.seed})}getClassName(){return kx.className}};_x.className="HeNormal",kc.registerClass(_x);var Nx=class extends kx{constructor(e){super({scale:2,mode:"fanIn",distribution:"uniform",seed:null==e?null:e.seed})}getClassName(){return kx.className}};Nx.className="HeUniform",kc.registerClass(Nx);var Tx=class extends kx{constructor(e){super({scale:1,mode:"fanIn",distribution:"normal",seed:null==e?null:e.seed})}getClassName(){return kx.className}};Tx.className="LeCunNormal",kc.registerClass(Tx);var Cx=class extends kx{constructor(e){super({scale:1,mode:"fanIn",distribution:"uniform",seed:null==e?null:e.seed})}getClassName(){return kx.className}};Cx.className="LeCunUniform",kc.registerClass(Cx);var Ex=class extends fx{constructor(e){super(),this.DEFAULT_GAIN=1,this.ELEMENTS_WARN_SLOW=2e3,this.gain=null==e.gain?this.DEFAULT_GAIN:e.gain,this.seed=e.seed}apply(e,t){return Qs(()=>{if(e.length<2)throw new nb("Shape must be at least 2D.");if("int32"!==t&&"float32"!==t&&void 0!==t)throw new TypeError(`Unsupported data type ${t}.`);let n=va.sizeFromShape(e.slice(0,-1)),r=e[e.length-1],a=n*r;a>this.ELEMENTS_WARN_SLOW&&console.warn(`Orthogonal initializer is being called on a matrix with more than ${this.ELEMENTS_WARN_SLOW} (${a}) elements: Slowness may result.`);let s=ax([Math.max(r,n),Math.min(r,n)],0,1,t,this.seed),i=bc.qr(s,!1),o=i[0],l=i[1].flatten().stridedSlice([0],[Math.min(r,n)*Math.min(r,n)],[Math.min(r,n)+1]);return o=yo(o,l.sign()),n<r&&(o=o.transpose()),yo(au(this.gain),o.reshape(e))})}getConfig(){return{gain:this.gain,seed:this.seed}}};Ex.className="Orthogonal",kc.registerClass(Ex);var Ax={constant:"Constant",glorotNormal:"GlorotNormal",glorotUniform:"GlorotUniform",heNormal:"HeNormal",heUniform:"HeUniform",identity:"Identity",leCunNormal:"LeCunNormal",leCunUniform:"LeCunUniform",ones:"Ones",orthogonal:"Orthogonal",randomNormal:"RandomNormal",randomUniform:"RandomUniform",truncatedNormal:"TruncatedNormal",varianceScaling:"VarianceScaling",zeros:"Zeros"};function $x(e,t={}){return mb(e,kc.SerializationMap.getMap().classNameMap,t,"initializer")}function Rx(e){return cb(e)}function Fx(e){if("string"==typeof e){let t=e in Ax?Ax[e]:e;if("GlorotNormal"===t)return new Sx;if("GlorotUniform"===t)return new Ix;if("HeNormal"===t)return new _x;if("HeUniform"===t)return new Nx;if("LeCunNormal"===t)return new Tx;if("LeCunUniform"===t)return new Cx;{let e={};return e.className=t,e.config={},$x(e)}}return e instanceof fx?e:$x(e)}function Dx(e){return Array.isArray(e)&&Array.isArray(e[0])}function Mx(e){return 0===e.length?[]:Array.isArray(e[0])?e:[e]}function Ox(e){let t;if(Array.isArray(e)){if(1!==e.length)throw new tb(`Expected Tensor length to be 1; got ${e.length}`);t=e[0]}else t=e;return t}function Lx(e){if(Array.isArray(e)&&Array.isArray(e[0])){if(1===e.length)return e[0];throw new tb(`Expected exactly 1 Shape; got ${e.length}`)}return e}function Px(e){let t=0;for(let n of e)0===n.shape.length?t+=1:t+=n.shape.reduce((e,t)=>e*t);return t}var zx="Variable",Bx=class{constructor(e,t="float32",n=zx,r=!0,a=null){this.dtype=t??"float32",this.shape=e.shape,this.id=_b(),n=n??zx,this.originalName=zb(n),this.name=Bb(this.originalName),this.trainable_=r,this.constraint=a,this.val=Kd(e,this.trainable_,this.name,this.dtype)}read(){return this.assertNotDisposed(),this.val}write(e){return this.assertNotDisposed(),function(e,t){if(e.shape.toString()!==t.shape.toString())throw new Error("Shape mismatch: "+JSON.stringify(e.shape)+" vs. "+JSON.stringify(t.shape))}(this.val,e),this.val.id!==e.id&&(this.val.assign(e),null!=this.constraint&&this.val.assign(this.constraint.apply(this.val))),this}dispose(){this.assertNotDisposed(),this.val.dispose()}assertNotDisposed(){if(this.val.isDisposed)throw new Error(`LayersVariable ${this.name} is already disposed.`)}get trainable(){return this.trainable_}set trainable(e){this.trainable_=e,this.val.trainable=e}};function Wx(e){return e.map(e=>e.read())}function Vx(e){e.forEach(e=>{e[0].write(e[1])})}var Ux=class{constructor(e){this.dtype=e.dtype,this.shape=e.shape,null!=e.shape?this.ndim=e.shape.length:this.ndim=e.ndim,this.maxNDim=e.maxNDim,this.minNDim=e.minNDim,this.axes=e.axes||{}}},Gx=class{constructor(e,t,n,r,a,s,i){this.dtype=e,this.shape=t,this.sourceLayer=n,this.inputs=r,this.callArgs=a,this.outputTensorIndex=i,this.id=_b(),null!=s&&(this.originalName=zb(s),this.name=Bb(this.originalName)),this.rank=t.length}},Hx=0,jx=class{constructor(e,t){this.callArgs=t,this.id=Hx++,this.outboundLayer=e.outboundLayer,this.inboundLayers=e.inboundLayers,this.nodeIndices=e.nodeIndices,this.tensorIndices=e.tensorIndices,this.inputTensors=e.inputTensors,this.outputTensors=e.outputTensors,this.inputMasks=e.inputMasks,this.outputMasks=e.outputMasks,this.inputShapes=e.inputShapes,this.outputShapes=e.outputShapes;for(let t of e.inboundLayers)null!=t&&t.outboundNodes.push(this);e.outboundLayer.inboundNodes.push(this)}getConfig(){let e=[];for(let t of this.inboundLayers)null!=t?e.push(t.name):e.push(null);return{outboundLayer:this.outboundLayer?this.outboundLayer.name:null,inboundLayers:e,nodeIndices:this.nodeIndices,tensorIndices:this.tensorIndices}}},qx=0,Kx=class extends kc.Serializable{constructor(e={}){super(),this._callHook=null,this._addedWeightNames=[],this._stateful=!1,this.id=qx++,this.activityRegularizer=null,this.inputSpec=null,this.supportsMasking=!1,this._trainableWeights=[],this._nonTrainableWeights=[],this._losses=[],this._updates=[],this._built=!1,this.inboundNodes=[],this.outboundNodes=[];let t=e.name;if(!t){let e=this.getClassName();t=hb(e)+"_"+Tb(e)}if(this.name=t,this.trainable_=null==e.trainable||e.trainable,null!=e.inputShape||null!=e.batchInputShape){let t;if(null!=e.batchInputShape)t=e.batchInputShape;else if(null!=e.inputShape){let n=null;null!=e.batchSize&&(n=e.batchSize),t=[n].concat(e.inputShape)}this.batchInputShape=t;let n=e.dtype;null==n&&(n=e.inputDType),null==n&&(n="float32"),this.dtype=n}null!=e.weights?this.initialWeights=e.weights:this.initialWeights=null,this._refCount=null,this.fastWeightInitDuringBuild=!1}static nodeKey(e,t){return e.name+"_ib-"+t.toString()}getNodeAtIndex(e,t){if(0===this.inboundNodes.length)throw new eb(`The layer has never been called and thus has no defined ${t}.`);if(this.inboundNodes.length<=e)throw new tb(`Asked to get ${t} at node ${e}, but the layer has only ${this.inboundNodes.length} inbound nodes.`);return this.inboundNodes[e]}getInputAt(e){return lb(this.getNodeAtIndex(e,"input").inputTensors)}getOutputAt(e){return lb(this.getNodeAtIndex(e,"output").outputTensors)}get input(){if(this.inboundNodes.length>1)throw new Qy(`Layer ${this.name} has multiple inbound nodes, hence the notion of "layer input" is ill-defined. Use \`getInputAt(nodeIndex)\` instead.`);if(0===this.inboundNodes.length)throw new Qy(`Layer ${this.name} is not connected, no input to return.`);return lb(this.getNodeAtIndex(0,"input").inputTensors)}get output(){if(0===this.inboundNodes.length)throw new Qy(`Layer ${this.name} has no inbound nodes.`);if(this.inboundNodes.length>1)throw new Qy(`Layer ${this.name} has multiple inbound nodes, hence the notion of "layer output" is ill-defined. Use \`getOutputAt(nodeIndex)\` instead.`);return lb(this.getNodeAtIndex(0,"output").outputTensors)}get losses(){return this._losses}calculateLosses(){return this.losses.map(e=>e())}get updates(){return this._updates}get built(){return this._built}set built(e){this._built=e}get trainable(){return this.trainable_}set trainable(e){this._trainableWeights.forEach(t=>t.trainable=e),this.trainable_=e}get trainableWeights(){return this.trainable_?this._trainableWeights.filter(e=>e.trainable):[]}set trainableWeights(e){this._trainableWeights=e}get nonTrainableWeights(){return this.trainable?this._trainableWeights.filter(e=>!e.trainable).concat(this._nonTrainableWeights):this._trainableWeights.concat(this._nonTrainableWeights)}set nonTrainableWeights(e){this._nonTrainableWeights=e}get weights(){return this.trainableWeights.concat(this.nonTrainableWeights)}get stateful(){return this._stateful}resetStates(){if(!this.stateful)throw new Error("Cannot call the resetStates() method of a non-stateful Layer object.")}assertInputCompatibility(e){let t=ub(e);if(null==this.inputSpec||0===this.inputSpec.length)return;let n=ub(this.inputSpec);if(t.length!==n.length)throw new tb(`Layer ${this.name} expects ${n.length} inputs, but it received ${t.length} input tensors. Input received: ${e}`);for(let e=0;e<t.length;e++){let r=t[e],a=n[e];if(null==a)continue;let s=r.rank;if(null!=a.ndim&&s!==a.ndim)throw new tb(`Input ${e} is incompatible with layer ${this.name}: expected ndim=${a.ndim}, found ndim=${s}`);if(null!=a.maxNDim&&s>a.maxNDim)throw new tb(`Input ${e} is incompatible with layer ${this.name}: expected max_ndim=${a.maxNDim}, found ndim=${s}`);if(null!=a.minNDim&&s<a.minNDim)throw new tb(`Input ${e} is incompatible with layer ${this.name}: expected min_ndim=${a.minNDim}, found ndim=${s}.`);if(null!=a.dtype&&r.dtype!==a.dtype)throw new tb(`Input ${e} is incompatible with layer ${this.name} : expected dtype=${a.dtype}, found dtype=${r.dtype}.`);if(a.axes){let t=r.shape;for(let n in a.axes){let r=Number(n),s=a.axes[n],i=r>=0?t[r]:t[t.length+r];if(null!=s&&-1===[s,null].indexOf(i))throw new tb(`Input ${e} is incompatible with layer ${this.name}: expected axis ${r} of input shape to have value ${s} but got shape ${t}.`)}}if(null!=a.shape)for(let t=0;t<a.shape.length;++t){let n=a.shape[t],s=r.shape[t];if(null!=n&&null!=s&&n!==s)throw new tb(`Input ${e} is incompatible with layer ${this.name}: expected shape=${a.shape}, found shape=${r.shape}.`)}}}call(e,t){return e}invokeCallHook(e,t){null!=this._callHook&&this._callHook(e,t)}setCallHook(e){this._callHook=e}clearCallHook(){this._callHook=null}apply(e,t){t=t||{},this.assertNotDisposed();let n=ub(e),r=function(e){let t=!0;for(let n of ub(e))if(!(n instanceof Gx)){t=!1;break}return t}(e),a=function(e){let t=!0;for(let n of ub(e))if(n instanceof Gx){t=!1;break}return t}(e);if(r===a)throw new tb("Arguments to apply() must be all SymbolicTensors or all Tensors");return Pb(this.name,()=>{if(!this.built){this.assertInputCompatibility(e);let t=[];for(let n of ub(e))t.push(n.shape);this.build(lb(t)),this.built=!0,this.initialWeights&&this.setWeights(this.initialWeights),null===this._refCount&&a&&(this._refCount=1)}if(this.assertInputCompatibility(e),a){let r=this.call(e,t);this.supportsMasking&&this.setMaskMetadata(e,r);let a=ub(r),s=[];for(let e of a)-1!==n.indexOf(e)&&(e=e.clone()),s.push(e);if(r=lb(s),null!=this.activityRegularizer)throw new nb("Layer invocation in the presence of activity regularizer(s) is not supported yet.");return r}{let n,r=function(e){e=ub(e);let t=[];for(let n of e)t.push(n.shape);return lb(t)}(e),a=this.computeOutputShape(r),s="float32";if(this.warnOnIncompatibleInputShape(Array.isArray(e)?r[0]:r),n=null!=a&&a.length>0&&Array.isArray(a[0])?a.map((n,r)=>new Gx(s,n,this,ub(e),t,this.name,r)):new Gx(s,a,this,ub(e),t,this.name),this.addInboundNode(e,n,null,null,r,a,t),this._refCount++,null!=this.activityRegularizer)throw new nb("Layer invocation in the presence of activity regularizer(s) is not supported yet.");return n}})}warnOnIncompatibleInputShape(e){if(null!=this.batchInputShape)if(e.length!==this.batchInputShape.length)console.warn(`The rank of the input tensor provided (shape: ${JSON.stringify(e)}) does not match that of the batchInputShape (${JSON.stringify(this.batchInputShape)}) of the layer ${this.name}`);else{let t=!1;this.batchInputShape.forEach((n,r)=>{null!=n&&null!=e[r]&&e[r]!==n&&(t=!0)}),t&&console.warn(`The shape of the input tensor (${JSON.stringify(e)}) does not match the expectation of layer ${this.name}: ${JSON.stringify(this.batchInputShape)}`)}}get outputShape(){if(null==this.inboundNodes||0===this.inboundNodes.length)throw new Qy(`The layer ${this.name} has never been called and thus has no defined output shape.`);let e=[];for(let t of this.inboundNodes){let n=JSON.stringify(t.outputShapes);-1===e.indexOf(n)&&e.push(n)}if(1===e.length){let e=this.inboundNodes[0].outputShapes;return Array.isArray(e)&&Array.isArray(e[0])&&1===e.length?e[0]:e}throw new Qy(`The layer ${this.name} has multiple inbound nodes with different output shapes. Hence the notion of "output shape" is ill-defined for the layer.`)}countParams(){if(!this.built)throw new eb(`You tried to call countParams() on ${this.name}, but the layer is not built yet. Build it first by calling build(batchInputShape).`);return Px(this.weights)}build(e){this.built=!0}getWeights(e=!1){return Wx(e?this.trainableWeights:this.weights)}setWeights(e){Qs(()=>{let t=this.weights;if(t.length!==e.length)throw new tb(`You called setWeights(weights) on layer "${this.name}" with a weight list of length ${e.length}, but the layer was expecting ${t.length} weights. Provided weights: ${e}...`);if(0===t.length)return;let n=[],r=Wx(t);for(let a=0;a<r.length;++a){let s=r[a],i=t[a],o=e[a];if(!va.arraysEqual(s.shape,o.shape))throw new tb(`Layer weight shape ${s.shape} not compatible with provided weight shape ${o.shape}`);n.push([i,o])}Vx(n)})}addWeight(e,t,n,r,a,s,i,o){if(-1!==this._addedWeightNames.indexOf(e))throw new tb(`Duplicate weight name ${e} for layer ${this.name}`);this._addedWeightNames.push(e),null==n&&(n="float32"),this.fastWeightInitDuringBuild&&(r=null!=o?o():Fx("zeros"));let l=r.apply(t,n),u=new Bx(l,n,e,s,i);return l.dispose(),null!=a&&this.addLoss(()=>a.apply(u.read())),null==s&&(s=!0),s?this._trainableWeights.push(u):this._nonTrainableWeights.push(u),u}setFastWeightInitDuringBuild(e){this.fastWeightInitDuringBuild=e}addLoss(e){null==e||Array.isArray(e)&&0===e.length||(e=ub(e),void 0!==this._losses&&null!==this._losses&&this.losses.push(...e))}computeOutputShape(e){return e}computeMask(e,t){if(!this.supportsMasking){if(null!=t){if(!Array.isArray(t))throw new TypeError(`Layer ${this.name} does not support masking, but was passed an inputMask.`);t.forEach(e=>{if(null!=e)throw new TypeError(`Layer ${this.name} does not support masking, but was passed an inputMask.`)})}return null}return t}setMaskMetadata(e,t,n){if(!this.supportsMasking)return;let r=this.computeMask(e,n),a=ub(t),s=ub(r);if(a.length!==s.length)throw new Error(`${this.name} outputs ${a.length} tensors but ${a.length} masks for those tensors`);for(let e=0;e<a.length;e++)a[e].kerasMask=s[e]}addInboundNode(e,t,n,r,a,s,i=null){let o=ub(e);t=ub(t),n=ub(n),r=ub(r),a=Mx(a),s=Mx(s);let l=[],u=[],h=[];for(let e of o)l.push(e.sourceLayer),u.push(e.nodeIndex),h.push(e.tensorIndex);new jx({outboundLayer:this,inboundLayers:l,nodeIndices:u,tensorIndices:h,inputTensors:o,outputTensors:t,inputMasks:n,outputMasks:r,inputShapes:a,outputShapes:s},i);for(let e=0;e<t.length;e++)t[e].sourceLayer=this,t[e].nodeIndex=this.inboundNodes.length-1,t[e].tensorIndex=e}getConfig(){let e={name:this.name,trainable:this.trainable};return null!=this.batchInputShape&&(e.batchInputShape=this.batchInputShape),null!=this.dtype&&(e.dtype=this.dtype),e}disposeWeights(){return this.weights.forEach(e=>e.dispose()),this.weights.length}assertNotDisposed(){if(0===this._refCount)throw new Error(`Layer '${this.name}' is already disposed.`)}dispose(){if(!this.built)throw new Error(`Cannot dispose Layer ${this.name} because it has not been built yet.`);if(null===this._refCount)throw new Error(`Cannot dispose Layer ${this.name} because it has not been used yet.`);this.assertNotDisposed();let e=0;return 0===--this._refCount&&(e=this.disposeWeights()),{refCountAfterDispose:this._refCount,numDisposedVariables:e}}};function Xx(e,t,n){if((null==t||null!=n&&n>0)&&(t=e.sourceLayer,n=e.nodeIndex),0===t.inboundNodes.length)return[e];{let e=t.inboundNodes[n];if(0===e.inboundLayers.length)return e.inputTensors;{let t=[];for(let n=0;n<e.inboundLayers.length;n++){let r=Xx(e.inputTensors[n],e.inboundLayers[n],e.nodeIndices[n]);for(let e of r)-1===t.indexOf(e)&&t.push(e)}return t}}}var Zx=class extends Kx{constructor(e){if(super({dtype:e.dtype,name:null!=e.name?e.name:Tb("input").toString()}),null==e.batchSize&&(e.batchSize=null),null==e.sparse&&(e.sparse=!1),this.trainable=!1,this.built=!0,this.sparse=e.sparse,null!=e.inputShape&&null!=e.batchInputShape)throw new tb("Only provide the inputShape OR batchInputShape argument to inputLayer, not both at the same time.");let t=e.batchInputShape;if(null==t){if(null==e.inputShape)throw new tb("An InputLayer should be passed either a `batchInputShape` or an `inputShape`.");t=[e.batchSize].concat(e.inputShape)}else if(null!=e.batchSize)throw new tb("Cannot specify batchSize if batchInputShape is specified when creating an InputLayer.");let n=e.dtype||"float32";this.batchInputShape=t,this.dtype=n,this.inputSpec=[{shape:t}];let r=new Gx(this.dtype,this.batchInputShape,this,[],{},this.name);r.nodeIndex=0,r.tensorIndex=0,new jx({outboundLayer:this,inboundLayers:[],nodeIndices:[],tensorIndices:[],inputTensors:[r],outputTensors:[r],inputMasks:[null],outputMasks:[null],inputShapes:[t],outputShapes:[t]})}apply(e,t){throw new tb(`Cannot pass any input to an InputLayer's apply() method. InputLayer name: ${this.name}`)}dispose(){return{refCountAfterDispose:this._refCount,numDisposedVariables:0}}getConfig(){return{batchInputShape:this.batchInputShape,dtype:this.dtype,sparse:this.sparse,name:this.name}}};function Yx(e){if(null==e.batchShape&&null==e.shape)throw new Error("Please provide to Input either a `shape` or a `batchShape` argument. Note that `shape` does not include the batch dimension.");if(null!=e.batchShape&&null!=e.shape)throw new tb("Please provide either a `shape` or `batchShape` argument to Input, but not both.");let t=e.batchShape;null!=e.shape&&null==t&&(t=[null].concat(e.shape));let n=e.dtype;return null==n&&(n="float32"),new Zx({batchInputShape:t,name:e.name,dtype:n,sparse:e.sparse}).inboundNodes[0].outputTensors[0]}Zx.className="InputLayer",kc.registerClass(Zx);var Jx=class e{constructor(t){if(this.id2Value={},this.id2Mask={},this.name2Id={},t instanceof e)for(let e in t.id2Value)this.id2Value[e]=t.id2Value[e],e in t.id2Mask&&(this.id2Mask[e]=t.id2Mask[e]);else{if(null==t)return;for(let e of t)this.add(e.key,e.value)}}add(e,t,n){if(null!=this.id2Value[e.id])throw new tb(`Duplicate key: name=${e.name}, id=${e.id}`);return this.id2Value[e.id]=function(e,t){if(null==e.dtype||e.dtype===t.dtype)return t;try{return ho(t,e.dtype)}catch(n){throw new tb(`The dtype of the feed (${t.dtype}) can not be cast to the dtype of the key '${e.name}' (${e.dtype}).`)}}(e,t),this.name2Id[e.name]=e.id,null!=n&&(this.id2Mask[e.id]=n),this}addFeed(e){this.add(e.key,e.value)}hasKey(e){return null!=this.id2Value[e.id]}names(){return Object.keys(this.name2Id)}getValue(e){if(e instanceof Gx){if(null==this.id2Value[e.id])throw new tb(`Nonexistent key: ${e.name}`);return this.id2Value[e.id]}{let t=this.name2Id[e];if(null==t)throw new tb(`Feed dict has no SymbolicTensor name: ${e}`);return this.id2Value[t]}}getMask(e){if(e instanceof Gx){if(null==this.id2Value[e.id])throw new tb(`Nonexistent key: ${e.name}`);return this.id2Mask[e.id]}{let t=this.name2Id[e];if(null==t)throw new tb(`Feed dict has no SymbolicTensor name: ${e}`);return this.id2Mask[t]}}disposeMasks(){null!=this.id2Mask&&ei(this.id2Mask)}},Qx=new ab,ev=new ab;function tv(e,t,n,r){let a=null!=n&&n.training,s=Array.isArray(e),i=s?e:[e],o=i.map(e=>e.name),l=[],u=t.names();for(let e of o)-1!==u.indexOf(e)?l.push(t.getValue(e)):l.push(null);null!=r&&(r.maxNumTensors=-1/0,r.minNumTensors=1/0);let h,d=o.join(",")+"|"+t.names().sort().join(","),p=Qx.get(d);if(null==p){let e=function(e,t){va.assert(null!=e&&e.length>0,()=>"Expected at least one fetch, got none");let n=[],r={};if(1===e.length){let a=rv(e[0],t);n=a.sorted,r=a.recipientMap}else{let a=new Set;for(let s of e){let{sorted:e,recipientMap:i}=rv(s,t);for(let t of e)a.has(t.name)||(n.push(t),a.add(t.name));for(let e in i)null==r[e]&&(r[e]=new Set),i[e].forEach(t=>r[e].add(t))}}return{sorted:n,recipientCounts:nv(r)}}(i,t);p=e.sorted,h=e.recipientCounts,Qx.put(d,p),ev.put(d,h)}h={},a||Object.assign(h,ev.get(d));let c=new Jx(t);for(let e=0;e<p.length;++e){if(null!=r){let e=Ys().numTensors;e>r.maxNumTensors&&(r.maxNumTensors=e),e<r.minNumTensors&&(r.minNumTensors=e)}let s=p[e],i=s.sourceLayer;if(i instanceof Zx)continue;let u=[],d=[],f=[],m=!1;for(let e of s.inputs){let n=c.getValue(e),r=c.getMask(e);u.push(n),d.push(r),null!=r&&(m=!0),a||(h[e.name]--,0===h[e.name]&&!t.hasKey(e)&&-1===o.indexOf(e.name)&&!n.isDisposed&&!0!==e.sourceLayer.stateful&&f.push(n))}m&&((n=n||{}).mask=d[0]);let g=ub(i.apply(u,n)),y=null;i.supportsMasking&&(y=i.computeMask(u,d));let b=av(s),x=Array.isArray(b)?b:[b];for(let e=0;e<x.length;++e){c.hasKey(x[e])||c.add(x[e],g[e],Array.isArray(y)?y[0]:y);let t=o.indexOf(x[e].name);-1!==t&&(l[t]=g[e])}a||ei(f)}return c.disposeMasks(),s?l:l[0]}function nv(e){let t={};for(let n in e)t[n]=e[n].size;return t}function rv(e,t){let n=new Set,r=[],a={};for(let e of t.names())n.add(e);let s=[],i=[];for(s.push(e);s.length>0;){let e=s[s.length-1];if(n.has(e.name)){s.pop();continue}let t=i[i.length-1]===s.length-1;if(0===e.inputs.length||t)s.pop(),r.push(e),n.add(e.name),t&&i.pop();else{i.push(s.length-1);for(let t of e.inputs)null==a[t.name]&&(a[t.name]=new Set),a[t.name].add(e.name),!n.has(t.name)&&s.push(t)}}return{sorted:r,recipientMap:a}}function av(e){let t;if(1===e.sourceLayer.inboundNodes.length)t=e.sourceLayer.output;else{let n=null;for(let t=0;t<e.sourceLayer.inboundNodes.length;++t)for(let r of e.sourceLayer.inboundNodes[t].outputTensors)if(r.id===e.id){n=t;break}t=e.sourceLayer.getOutputAt(n)}return t}Le().registerFlag("TOPOLOGICAL_SORT_CACHE_MAX_ENTRIES",()=>100,function(e){null!=Qx&&Qx.setMaxEntries(e),null!=ev&&ev.setMaxEntries(e)});var sv={};function iv(e,t){return Qs(()=>su(ou(yo(e,e),t,!0)))}f(sv,{maxNorm:()=>gv,minMaxNorm:()=>xv,nonNeg:()=>bv,unitNorm:()=>yv});var ov=class extends kc.Serializable{getConfig(){return{}}},lv=class extends ov{constructor(e){super(),this.defaultMaxValue=2,this.defaultAxis=0,this.maxValue=null!=e.maxValue?e.maxValue:this.defaultMaxValue,this.axis=null!=e.axis?e.axis:this.defaultAxis}apply(e){return Qs(()=>{let t=iv(e,this.axis),n=cl(t,0,this.maxValue);return yo(e,go(n,fo(Xb(),t)))})}getConfig(){return{maxValue:this.maxValue,axis:this.axis}}};lv.className="MaxNorm",kc.registerClass(lv);var uv=class extends ov{constructor(e){super(),this.defaultAxis=0,this.axis=null!=e.axis?e.axis:this.defaultAxis}apply(e){return Qs(()=>go(e,fo(Xb(),iv(e,this.axis))))}getConfig(){return{axis:this.axis}}};uv.className="UnitNorm",kc.registerClass(uv);var hv=class extends ov{apply(e){return ed(e)}};hv.className="NonNeg",kc.registerClass(hv);var dv=class extends ov{constructor(e){super(),this.defaultMinValue=0,this.defaultMaxValue=1,this.defaultRate=1,this.defaultAxis=0,this.minValue=null!=e.minValue?e.minValue:this.defaultMinValue,this.maxValue=null!=e.maxValue?e.maxValue:this.defaultMaxValue,this.rate=null!=e.rate?e.rate:this.defaultRate,this.axis=null!=e.axis?e.axis:this.defaultAxis}apply(e){return Qs(()=>{let t=iv(e,this.axis),n=fo(yo(this.rate,cl(t,this.minValue,this.maxValue)),yo(1-this.rate,t));return yo(e,go(n,fo(Xb(),t)))})}getConfig(){return{minValue:this.minValue,maxValue:this.maxValue,rate:this.rate,axis:this.axis}}};dv.className="MinMaxNorm",kc.registerClass(dv);var pv={maxNorm:"MaxNorm",minMaxNorm:"MinMaxNorm",nonNeg:"NonNeg",unitNorm:"UnitNorm"};function cv(e){return cb(e)}function fv(e,t={}){return mb(e,kc.SerializationMap.getMap().classNameMap,t,"constraint")}function mv(e){if(null==e)return null;if("string"==typeof e){return fv({className:e in pv?pv[e]:e,config:{}})}return e instanceof ov?e:fv(e)}function gv(e){return new lv(e)}function yv(e){return new uv(e)}function bv(){return new hv}function xv(e){return new dv(e)}var vv={};function wv(){return new mx}function kv(){return new gx}function Iv(e){return new yx(e)}function Sv(e){return new bx(e)}function _v(e){return new xx(e)}function Nv(e){return new vx(e)}function Tv(e){return new wx(e)}function Cv(e){return new kx(e)}function Ev(e){return new Ix(e)}function Av(e){return new Sx(e)}function $v(e){return new _x(e)}function Rv(e){return new Nx(e)}function Fv(e){return new Tx(e)}function Dv(e){return new Cx(e)}function Mv(e){return new Ex(e)}f(vv,{constant:()=>Iv,glorotNormal:()=>Av,glorotUniform:()=>Ev,heNormal:()=>$v,heUniform:()=>Rv,identity:()=>Tv,leCunNormal:()=>Fv,leCunUniform:()=>Dv,ones:()=>kv,orthogonal:()=>Mv,randomNormal:()=>_v,randomUniform:()=>Sv,truncatedNormal:()=>Nv,varianceScaling:()=>Cv,zeros:()=>wv});var Ov,Lv={};async function Pv(e){if(null==e)return;let t=[],n=[],r=[];for(let a in e){let s=e[a];if("number"!=typeof s){let e=s;t.push(e.data()),n.push(a),r.push(e)}}if(t.length>0){let a=await Promise.all(t);for(let t=0;t<a.length;++t)e[n[t]]=a[t][0];ei(r)}}function zv(e){if(null!=e)for(let t in e){let n=e[t];"number"!=typeof n&&n.dispose()}}f(Lv,{Layer:()=>Kx,RNN:()=>Qk,RNNCell:()=>eI,activation:()=>RS,add:()=>WS,alphaDropout:()=>$_,average:()=>VS,averagePooling1d:()=>YS,averagePooling2d:()=>e_,averagePooling3d:()=>r_,avgPool1d:()=>JS,avgPool2d:()=>t_,avgPool3d:()=>a_,avgPooling1d:()=>QS,avgPooling2d:()=>n_,avgPooling3d:()=>s_,batchNormalization:()=>KS,bidirectional:()=>I_,categoryEncoding:()=>O_,centerCrop:()=>D_,concatenate:()=>US,conv1d:()=>IS,conv2d:()=>SS,conv2dTranspose:()=>_S,conv3d:()=>NS,conv3dTranspose:()=>TS,convLstm2d:()=>x_,convLstm2dCell:()=>v_,cropping2D:()=>ES,dense:()=>FS,depthwiseConv2d:()=>$S,dot:()=>qS,dropout:()=>DS,elu:()=>yS,embedding:()=>BS,flatten:()=>OS,gaussianDropout:()=>A_,gaussianNoise:()=>E_,globalAveragePooling1d:()=>i_,globalAveragePooling2d:()=>o_,globalMaxPool1d:()=>__,globalMaxPool2d:()=>N_,globalMaxPooling1d:()=>l_,globalMaxPooling2d:()=>u_,gru:()=>c_,gruCell:()=>f_,input:()=>Jw,inputLayer:()=>gS,layerNormalization:()=>XS,leakyReLU:()=>xS,lstm:()=>m_,lstmCell:()=>g_,masking:()=>R_,maxPool1d:()=>T_,maxPool2d:()=>C_,maxPooling1d:()=>h_,maxPooling2d:()=>d_,maxPooling3d:()=>p_,maximum:()=>GS,minimum:()=>HS,multiply:()=>jS,permute:()=>zS,prelu:()=>vS,randomWidth:()=>L_,reLU:()=>bS,repeatVector:()=>LS,rescaling:()=>F_,reshape:()=>PS,resizing:()=>M_,rnn:()=>w_,separableConv2d:()=>CS,simpleRNN:()=>y_,simpleRNNCell:()=>b_,softmax:()=>wS,spatialDropout1d:()=>MS,stackedRNNCells:()=>k_,thresholdedReLU:()=>kS,timeDistributed:()=>S_,upSampling2d:()=>AS,zeroPadding2d:()=>ZS}),function(e){e[e.SILENT=0]="SILENT",e[e.VERBOSE=1]="VERBOSE"}(Ov||(Ov={}));var Bv=class{constructor(){this.validationData=null}setParams(e){this.params=e}async onEpochBegin(e,t){}async onEpochEnd(e,t){}async onBatchBegin(e,t){}async onBatchEnd(e,t){}async onTrainBegin(e){}async onTrainEnd(e){}setModel(e){}},Wv=class{constructor(e,t=10){null==e&&(e=[]),this.callbacks=e,this.queueLength=t}append(e){this.callbacks.push(e)}setParams(e){for(let t of this.callbacks)t.setParams(e)}setModel(e){for(let t of this.callbacks)t.setModel(e)}async onEpochBegin(e,t){null==t&&(t={});for(let n of this.callbacks)await n.onEpochBegin(e,t)}async onEpochEnd(e,t){null==t&&(t={});for(let n of this.callbacks)await n.onEpochEnd(e,t)}async onBatchBegin(e,t){null==t&&(t={});for(let n of this.callbacks)await n.onBatchBegin(e,t)}async onBatchEnd(e,t){null==t&&(t={});for(let n of this.callbacks)await n.onBatchEnd(e,t)}async onTrainBegin(e){null==e&&(e={});for(let t of this.callbacks)await t.onTrainBegin(e)}async onTrainEnd(e){null==e&&(e={});for(let t of this.callbacks)await t.onTrainEnd(e)}},Vv=class extends Bv{constructor(){super()}async onEpochBegin(e){this.seen=0,this.totals={}}async onBatchEnd(e,t){null==t&&(t={});let n=null==t.size?0:t.size;this.seen+=n;for(let e in t){let r=t[e];if("number"==typeof r)this.totals.hasOwnProperty(e)||(this.totals[e]=0),this.totals[e]=this.totals[e]+r*n;else{let t;e in this.totals?t=this.totals[e]:this.totals[e]=0;let a=Qs(()=>fo(this.totals[e],yo(r,n)));this.totals[e]=a,null!=t&&t.dispose()}}}async onEpochEnd(e,t){if(null!=t)for(let e of this.params.metrics)null!=this.totals[e]&&("number"==typeof this.totals[e]?t[e]=this.totals[e]/this.seen:Qs(()=>{let n=yo(go(1,this.seen),this.totals[e]);t[e]=n,this.totals[e].dispose(),ti(t[e])}))}},Uv=class extends Bv{async onTrainBegin(e){this.epoch=[],this.history={}}async onEpochEnd(e,t){null==t&&(t={}),this.epoch.push(e);for(let e in t)null==this.history[e]&&(this.history[e]=[]),this.history[e].push(t[e])}async syncData(){let e=[],t=[],n=[];for(let r in this.history){let a=this.history[r];for(let s=0;s<a.length;++s)if("number"!=typeof a[s]){let i=a[s];e.push(i.data()),t.push(r),n.push(s)}}let r=await Promise.all(e);for(let e=0;e<r.length;++e)this.history[t[e]][n[e]].dispose(),this.history[t[e]][n[e]]=r[e][0]}},Gv=class extends Bv{constructor(e,t){if(super(),this.currentEpoch=0,this.nowFunc=e.nowFunc,this.nextFrameFunc=e.nextFrameFunc||Vf,this.yieldEvery=t||"auto","auto"===this.yieldEvery&&(this.yieldEvery=125),"never"===this.yieldEvery&&null!=e.onYield)throw new Error("yieldEvery is `never` but you provided an `onYield` callback. Either change `yieldEvery` or remove the callback");va.isNumber(this.yieldEvery)&&(this.maybeWait=function(e,t,n){let r,a=null!=n?n():va.now();return(...s)=>{let i=null!=n?n():va.now();return i-a<t||(a=i,r=e(...s)),r}}(this.maybeWait.bind(this),this.yieldEvery,this.nowFunc)),this.trainBegin=e.onTrainBegin,this.trainEnd=e.onTrainEnd,this.epochBegin=e.onEpochBegin,this.epochEnd=e.onEpochEnd,this.batchBegin=e.onBatchBegin,this.batchEnd=e.onBatchEnd,this.yield=e.onYield}async maybeWait(e,t,n){let r=[];null!=this.yield&&(await Pv(n),r.push(this.yield(e,t,n))),r.push(this.nextFrameFunc()),await Promise.all(r)}async onEpochBegin(e,t){this.currentEpoch=e,null!=this.epochBegin&&(await Pv(t),await this.epochBegin(e,t))}async onEpochEnd(e,t){let n=[];null!=this.epochEnd&&(await Pv(t),n.push(this.epochEnd(e,t))),"epoch"===this.yieldEvery&&n.push(this.nextFrameFunc()),await Promise.all(n)}async onBatchBegin(e,t){null!=this.batchBegin&&(await Pv(t),await this.batchBegin(e,t))}async onBatchEnd(e,t){let n=[];null!=this.batchEnd&&(await Pv(t),n.push(this.batchEnd(e,t))),"batch"===this.yieldEvery?n.push(this.nextFrameFunc()):va.isNumber(this.yieldEvery)&&n.push(this.maybeWait(this.currentEpoch,e,t)),await Promise.all(n)}async onTrainBegin(e){null!=this.trainBegin&&(await Pv(e),await this.trainBegin(e))}async onTrainEnd(e){null!=this.trainEnd&&(await Pv(e),await this.trainEnd(e))}};function Hv(e,t){return null==e&&(e={}),e instanceof Bv?[e]:Array.isArray(e)&&e[0]instanceof Bv?e:ub(e).map(e=>new Gv(e,t))}var jv=class e{constructor(){}static registerCallbackConstructor(t,n){va.assert(t>=0&&Number.isInteger(t),()=>`Verbosity level is expected to be an integer >= 0, but got ${t}`),e.checkForDuplicate(n),null==e.constructors[t]&&(e.constructors[t]=[]),e.constructors[t].push(n)}static checkForDuplicate(t){for(let n in e.constructors)e.constructors[+n].forEach(e=>{if(e===t)throw new tb("Duplicate callback constructor.")})}static clear(){e.constructors={}}static createCallbacks(t){let n=[];for(let r in e.constructors){let a=+r;t>=a&&n.push(...e.constructors[a])}return n.map(e=>new e)}};function qv(e,t,n,r,a,s,i,o,l){let u=new Uv,h=[new Vv,...jv.createCallbacks(t)];null!=e&&h.push(...e),h.push(u);let d=new Wv(h);return d.setParams({epochs:n,initialEpoch:r,samples:a,steps:s,batchSize:i,verbose:t,doValidation:o,metrics:l}),{callbackList:d,history:u}}function Kv(e,t={},n=!1){return mb(e,kc.SerializationMap.getMap().classNameMap,t,"layer",n)}function Xv(e,t){return Qs(()=>{"float32"!==e.dtype&&(e=ho(e,"float32"));let n=ou(ox(e),t,!0),r=pl(n.shape,Xb()),a=su(eh(n,r));return go(e,a)})}function Zv(e,t){return Qs(()=>th(ox(Wu(t,e)),-1))}function Yv(e,t){return Qs(()=>th(bo(Wu(t,e)),-1))}function Jv(e,t){return Qs(()=>{let n=Wu(e,t),r=cl(bo(e),Xb(),Number.MAX_VALUE),a=bo(go(n,r));return yo(100,th(a,-1))})}function Qv(e,t,n=!1){return Qs(()=>{if(n)t=xd(t);else{let e=ou(t,t.shape.length-1,!0);t=go(t,e)}return t=cl(t,Xb(),1-Xb()),Pu(ou(yo(ho(e,"float32"),Eu(t)),t.shape.length-1))})}function ew(e,t,n=!1){return Qs(()=>{let r=ho(gu(function(e){let t=[Hb(e.shape)];return jo(e,t)}(e)),"int32"),a=(t=cl(t,Xb(),1-Xb())).shape;return Qv(jo(ph(r,a[a.length-1]),a),t,n)})}function tw(e,t){return Qs(()=>{let n;return n=cl(t,Xb(),1-Xb()),n=Eu(go(n,Wu(1,n))),th(function(e,t){if(!va.arraysEqual(e.shape,t.shape))throw new tb(`logits and labels must have the same shape, but got shapes ${JSON.stringify(e.shape)} and ${JSON.stringify(t.shape)}`);return Qs(()=>{let n=ed(t),r=Pu(bo(t));return fo(Wu(n,yo(t,e)),Au(du(r)))})}(e,n),-1)})}function nw(e,t){return Qs(()=>{let n=Xv(e,-1),r=Xv(t,-1),a=yo(n,r);return Pu(ou(a,-1))})}jv.constructors={};var rw={meanSquaredError:Zv,meanAbsoluteError:Yv,meanAbsolutePercentageError:Jv,meanSquaredLogarithmicError:function(e,t){return Qs(()=>{let n=cl(t,Xb(),Number.MAX_VALUE),r=Eu(fo(1,n)),a=cl(e,Xb(),Number.MAX_VALUE),s=Eu(fo(1,a));return th(ox(Wu(r,s)),-1)})},squaredHinge:function(e,t){return Qs(()=>{let n=eh(0,Wu(1,yo(e,t)));return th(ox(n),-1)})},hinge:function(e,t){return Qs(()=>{let n=eh(0,Wu(1,yo(e,t)));return th(n,-1)})},categoricalHinge:function(e,t){return Qs(()=>{let n=ou(yo(e,t),-1),r=tu(yo(Wu(1,e),t),-1);return eh(0,fo(1,Wu(r,n)))})},logcosh:function(e,t){return Qs(()=>{let n=Math.log(2),r=Wu(t,e),a=Wu(fo(r,zu(yo(-2,r))),n);return th(a,-1)})},categoricalCrossentropy:Qv,sparseCategoricalCrossentropy:ew,binaryCrossentropy:tw,kullbackLeiblerDivergence:function(e,t){return Qs(()=>{let n=cl(e,Xb(),1),r=cl(t,Xb(),1);return ou(yo(e,Eu(go(n,r))),-1)})},poisson:function(e,t){return Qs(()=>{let n=Eu(fo(Xb(),t));return th(Wu(t,yo(e,n)),-1)})},cosineProximity:nw};function aw(e){if("string"==typeof e){if(e in rw)return rw[e];let t=`Unknown loss ${e}`;throw e.toLowerCase().includes("softmaxcrossentropy")&&(t=`Unknown loss ${e}. Use "categoricalCrossentropy" as the string name for tf.losses.softmaxCrossEntropy`),new tb(t)}return e}function sw(e,t){return Qs(()=>{let n=yo(.5,ch(t)),r=Zb(bu(t,n),e.dtype);return th(Pl(e,r),-1)})}function iw(e,t){return Qs(()=>Zb(Pl(So(e,-1),So(t,-1)),"float32"))}function ow(e,t){return Qs(()=>ho(ou(Gu(Pl(e,1),Pl(t,1))),"float32"))}function lw(e,t){return Qs(()=>{let n=ow(e,t),r=function(e,t){return Qs(()=>ho(ou(Gu(Pl(e,0),Pl(t,1))),"float32"))}(e,t),a=fo(n,r);return ho(zl(bu(a,0),go(n,a),0),"float32")})}function uw(e,t){return Qs(()=>{let n=ow(e,t),r=function(e,t){return Qs(()=>ho(ou(Gu(Pl(e,1),Pl(t,0))),"float32"))}(e,t),a=fo(n,r);return ho(zl(bu(a,0),go(n,a),0),"float32")})}function hw(e,t){return tw(e,t)}function dw(e,t){return e.rank===t.rank&&(e=Nd(e,[e.rank-1])),(t=So(t,-1)).dtype!==e.dtype&&(t=ho(t,e.dtype)),ho(Pl(e,t),"float32")}var pw=Qv,cw=ew,fw={binaryAccuracy:sw,categoricalAccuracy:iw,precision:lw,categoricalCrossentropy:pw,sparseCategoricalCrossentropy:cw,mse:Zv,MSE:Zv,mae:Yv,MAE:Yv,mape:Jv,MAPE:Jv,cosine:nw};function mw(e){if("string"==typeof e&&e in fw)return fw[e];if("string"!=typeof e&&null!=e)return e;throw new tb(`Unknown metric ${e}`)}function gw(e){if(ib(null!==e,`Unknown LossOrMetricFn ${e}`),"string"==typeof e)return e;{let t;for(let n of Object.keys(rw))if(rw[n]===e){t=n;break}if(void 0!==t)return t;for(let n of Object.keys(fw))if(fw[n]===e){t=n;break}return void 0!==t?t:e.name}}function yw(e,t,n=!1){if(null==e||"object"!=typeof e||Object.getPrototypeOf(e)!==Object.prototype||!bw(e))throw new Error("User-defined metadata is expected to be a JSON object, but is not.");if(n){let n=JSON.stringify(e);n.length>1048576&&console.warn(`User-defined metadata of model "${t}" is too large in size (length=${n.length} when serialized). It is not recommended to store such large objects in user-defined metadata. Please make sure its serialized length is <= 1048576.`)}}function bw(e){if(null===e)return!0;if("object"==typeof e){if(Object.getPrototypeOf(e)===Object.prototype){let t=Object.keys(e);for(let n of t)if("string"!=typeof n||!bw(e[n]))return!1;return!0}if(Array.isArray(e)){for(let t of e)if(!bw(t))return!1;return!0}return!1}{let t=typeof e;return"string"===t||"number"===t||"boolean"===t}}function xw(e,t,n,r=console.log){let a,s=function(e){let t=!0,n=[],r=[];for(let t in e.nodesByDepth)n.push(e.nodesByDepth[t]);for(let e of n){if(e.length>1||1===e.length&&e[0].inboundLayers.length>1){t=!1;break}r.push(...e)}if(t)for(let n of e.layers){let e=!1;for(let a of n.inboundNodes)if(-1!==r.indexOf(a)){if(e){t=!1;break}e=!0}if(!t)break}return t}(e),i=["Layer (type)","Input Shape","Output shape","Param #"];if(s?(t=t||90,n=n||[.32,.61,.89,1]):(t=t||115,n=n||[.24,.48,.7,.8,1]),n[n.length-1]<=1&&(n=n.map(e=>Math.floor(t*e))),!s){i.push("Receives inputs"),a=[];for(let t in e.nodesByDepth)a.push(...e.nodesByDepth[t])}r("_".repeat(t)),vw(i,n,r),r("=".repeat(t));let o=e.layers;for(let e=0;e<o.length;++e)s?ww(o[e],n,r):kw(o[e],n,a,r),r((e===o.length-1?"=":"_").repeat(t));e.checkTrainableWeightsConsistency();let l=function(e){let t;return t=null!=e.collectedTrainableWeights?Px(e.collectedTrainableWeights):Px(e.trainableWeights),t}(e),u=Px(e.nonTrainableWeights);r(`Total params: ${l+u}`),r(`Trainable params: ${l}`),r(`Non-trainable params: ${u}`),r("_".repeat(t))}function vw(e,t,n=console.log){let r="";for(let n=0;n<e.length;++n)n>0&&(r=r.slice(0,r.length-1)+" "),r+=e[n],r=r.slice(0,t[n]),r+=" ".repeat(t[n]-r.length);n(r)}function ww(e,t,n){let r,a;try{a=e.inboundNodes.map(e=>JSON.stringify(e.inputShapes)).join(",")}catch(e){a="multiple"}try{r=JSON.stringify(e.outputShape)}catch(e){r="multiple"}vw([`${e.name} (${e.getClassName()})`,a,r,e.countParams().toString()],t,n)}function kw(e,t,n,r){let a,s;try{s=e.inboundNodes.map(e=>JSON.stringify(e.inputShapes)).join(",")}catch(e){s="multiple"}try{a=JSON.stringify(e.outputShape)}catch(e){a="multiple"}let i=[];for(let t of e.inboundNodes)if(!(null!=n&&n.length>0&&-1===n.indexOf(t)))for(let e=0;e<t.inboundLayers.length;++e){let n=t.inboundLayers[e].name,r=t.nodeIndices[e],a=t.tensorIndices[e];i.push(`${n}[${r}][${a}]`)}let o=e.name,l=e.getClassName(),u=0===i.length?"":i[0];vw([`${o} (${l})`,s,a,e.countParams().toString(),u],t,r);for(let e=1;e<i.length;++e)vw(["","","","",i[e]],t,r)}function Iw(e,t,n){return("inboundNodes"===e||"outputLayers"===e||"inputLayers"===e)&&0===t&&"string"==typeof n}function Sw(e,t){if(null===e)return null;if("string"==typeof e)return db(e);if("number"==typeof e||"boolean"==typeof e)return e;if(e instanceof Array){let n=[],r=e.length;for(let a=0;a<r;++a){let r=e[a];Iw(t,a,r)?n.push(r):n.push(Sw(r,t))}return n}{let t={};for(let n of Object.keys(e)){let r=e[n];if("name"===n&&"string"==typeof r)t[n]=r;else{let e=db(n);t[e]=Sw(r,e)}}return t}}function _w(e,t){if(null==e)return null;if("string"==typeof e)return hb(e);if("number"==typeof e||"boolean"==typeof e)return e;if(e instanceof Array){let n=[],r=e.length;for(let a=0;a<r;++a){let r=e[a];Iw(t,a,r)?n.push(r):n.push(_w(r,t))}return n}{let t={};for(let n of Object.keys(e)){let r=e[n],a=hb(n);t[a]="name"!==n&&"className"!==n||"string"!=typeof r?_w(r,n):r}return t}}var Nw="4.22.0",Tw=class e extends Kx{constructor(t){if(super({}),this.containerNodes=new Set,this.name=t.name,null==this.name){let e=this.getClassName().toLowerCase();this.name=Tb(e)}if(this.supportsMasking=!1,this.trainable_=!0,Array.isArray(t.inputs)?this.inputs=t.inputs.slice():this.inputs=[t.inputs],Array.isArray(t.outputs)?this.outputs=t.outputs.slice():this.outputs=[t.outputs],yb(this.inputs).length!==this.inputs.length)throw new tb(`The list of inputs passed to the model is redundant. All inputs should only appear once. Found: ${this.inputs.map(e=>e.name)}`);yb(this.outputs).length!==this.outputs.length&&console.warn(`The list of outputs passed to the model is redundant. All outputs should only appear once. Found: ${this.outputs.map(e=>e.name)}`),this.inputLayers=[],this.inputLayersNodeIndices=[],this.inputLayersTensorIndices=[],this.outputLayers=[],this.outputLayersNodeIndices=[],this.outputLayersTensorIndices=[],this.layers=[],this.internalContainerRefs=[];for(let e of this.outputs){let t=e.sourceLayer,n=e.nodeIndex,r=e.tensorIndex;this.outputLayers.push(t),this.outputLayersNodeIndices.push(n),this.outputLayersTensorIndices.push(r)}for(let e of this.inputs){let t=e.sourceLayer,n=e.nodeIndex,r=e.tensorIndex;ib(0===n,"input layer has >1 nodes"),ib(0===r,"input layer has >1 tensors"),this.inputLayers.push(t),this.inputLayersNodeIndices.push(n),this.inputLayersTensorIndices.push(r)}this.inputNames=[],this.outputNames=[],this.feedInputShapes=[],this.feedInputNames=[],this.feedOutputNames=[];for(let e=0;e<this.inputLayers.length;e++){let n=this.inputLayers[e];if(!(n instanceof Zx))throw new TypeError(`Input layers to a LayersModel must be InputLayer objects. Received inputs: ${t.inputs}. Input ${e} (0-based) originates from layer type ${n.getClassName()}.`);this.inputNames.push(n.name),this.feedInputShapes.push(n.batchInputShape),this.feedInputNames.push(n.name)}for(let e of this.outputLayers)this.outputNames.push(e.name);this.internalInputShapes=this.inputs.map(e=>e.shape),this.internalOutputShapes=this.outputs.map(e=>e.shape);let n={},r={},a={},s={},i={},o=[],l=(t,n,r,a,s,u)=>{(null==a||null==s||null==u)&&(a=t.sourceLayer,s=t.nodeIndex,u=t.tensorIndex);let h=a.inboundNodes[s];if(-1!==r.indexOf(h))throw new eb(`The tensor ${t.name} at layer "${a.name}" is part of a cycle.`);if(-1!==n.indexOf(h))return;this.containerNodes.add(e.nodeKey(a,s)),a.id in i||(i[a.id]=Object.keys(i).length),-1===r.indexOf(h)&&r.push(h);let d=h.inboundLayers.length;for(let e=0;e<d;e++){let t=h.inputTensors[e],a=h.inboundLayers[e],s=h.nodeIndices[e],i=h.tensorIndices[e];l(t,n,r,a,s,i)}for(n.push(h);r.indexOf(h)>=0;)r.splice(r.indexOf(h),1);o.push(h)},u=[],h=[];for(let e of this.outputs)l(e,u,h);let d=o.slice().reverse();for(let e of d){r[e.id]=e,e.id in n||(n[e.id]=0);let t=n[e.id],i=null==a[e.outboundLayer.id]?0:a[e.outboundLayer.id];t=Math.max(t,i),a[e.outboundLayer.id]=t,s[e.outboundLayer.id]=e.outboundLayer,n[e.id]=t;for(let a=0;a<e.inboundLayers.length;a++){let s=e.inboundLayers[a],i=e.nodeIndices[a],o=s.inboundNodes[i],l=null==n[o.id]?0:n[o.id];n[o.id]=Math.max(t+1,l),r[o.id]=o}}let p={};for(let e in n){let t=n[e];t in p||(p[t]=[]),p[t].push(r[e])}let c={};for(let e in a){let t=a[e];t in c||(c[t]=[]),c[t].push(s[e])}let f=Object.keys(c).map(e=>parseInt(e,10)).sort(gb);this.layers=[];for(let t of f){let n=c[t];n.sort((e,t)=>{let n=i[e.id],r=i[t.id];return n<r?-1:n>r?1:0});for(let t of n)t instanceof e&&this.internalContainerRefs.push(t),this.layers.push(t)}this.layersByDepth=c,f=Object.keys(p).map(e=>parseInt(e,10)).sort(gb);let m=this.inputs.slice(),g=[];for(let e of f)for(let t of p[e]){let e=t.outboundLayer;if(null!=e){for(let n of t.inputTensors)if(-1===m.indexOf(n))throw new eb(`Graph disconnected: cannot obtain value for tensor ${n} at layer "${e.name}". The following previous layers were accessed without issue: ${g}`);for(let e of t.outputTensors)m.push(e);g.push(e.name)}}this.nodesByDepth=p;let y=this.layers.map(e=>e.name);for(let e of y){let t=y.filter(t=>t===e).length;if(1!==t)throw new eb(`The name "${e}" is used ${t} times in the model. All layer names should be unique. Layer names: `+JSON.stringify(y))}this.outboundNodes=[],this.inboundNodes=[],new jx({outboundLayer:this,inboundLayers:[],nodeIndices:[],tensorIndices:[],inputTensors:this.inputs,outputTensors:this.outputs,inputMasks:this.inputs.map(e=>null),outputMasks:this.outputs.map(e=>null),inputShapes:this.inputs.map(e=>e.shape),outputShapes:this.outputs.map(e=>e.shape)}),this.built=!0,this._refCount=1}assertNotDisposed(){if(0===this._refCount)throw new Error(`Container '${this.name}' is already disposed.`)}dispose(){this.assertNotDisposed();let e={refCountAfterDispose:null,numDisposedVariables:0};if(0===--this._refCount){for(let t of this.layers)e.numDisposedVariables+=t.dispose().numDisposedVariables;for(let t of this.internalContainerRefs)e.numDisposedVariables+=t.dispose().numDisposedVariables}return e.refCountAfterDispose=this._refCount,e}get trainable(){return this.trainable_}set trainable(e){this.layers.forEach(t=>{t._trainableWeights.forEach(t=>t.trainable=e)}),this.trainable_=e}get trainableWeights(){if(this._trainableWeights.length>0)throw new tb("Container instance unexpectedly contains _trainableWeights.The trainable weights of a Container are a union of the trainable weights of its consituent Layers. Its own _trainableWeights must remain an empty Array.");if(!this.trainable)return[];let e=[];for(let t of this.layers)e=e.concat(t.trainableWeights);return e}get nonTrainableWeights(){let e=[];for(let t of this.layers)e.push(...t.nonTrainableWeights);if(!this.trainable){let t=[];for(let e of this.layers)t.push(...e.trainableWeights);return t.concat(e)}return e}get weights(){return this.trainableWeights.concat(this.nonTrainableWeights)}loadWeights(e,t=!0){let n={},r=0,a=(e=>{let t=Object.keys(e);if(0===t.length)return!1;let n=t[0].split("/");return!isNaN(parseInt(n[n.length-1],10))})(e);a&&this.parseWeights(e);for(let e of this.layers)for(let[t,s]of e.weights.entries()){let e=a?`${s.name.split("/").slice(0,-1).join("/")+"/"}${t}`:s.originalName;if(null!=n[e])throw new tb(`Duplicate weight name: ${e}`);n[e]=s,r++}let s=[];for(let r in e){let a=r;if(null==n[r]){let e=r.split("/");a=e.slice(0,-2).concat([e[e.length-1]]).join("/")}if(null!=n[a])s.push([n[a],e[r]]);else if(t)throw new tb(`Provided weight data has no target variable: ${r}`);delete n[a]}if(t){let e=[];for(let t in n)e.push(t);if(e.length>0)throw new tb(`${e.length} of ${r} weights are not set: ${e}`)}Vx(s)}parseWeights(e){for(let t in Object.keys(e)){let n=t.split("/"),r=["vars","layer_checkpoint_dependencies"],a=n.map(e=>e.startsWith("_")?e.slice(1):e).filter(e=>!r.includes(e)).join("/");a!==t&&(e[a]=e[t],delete e[t])}}updatedConfig(){let e=this.getConfig(),t={};return t.className=this.getClassName(),t.config=e,t.kerasVersion=`tfjs-layers ${Nw}`,t.backend="TensorFlow.js",t}toJSON(e,t=!0){let n=_w(this.updatedConfig());return t?JSON.stringify(n):n}call(e,t){return Qs(()=>{e=ub(e);let n=new Jx;for(let t=0;t<this.inputs.length;++t)n.add(this.inputs[t],e[t]);return tv(this.outputs,n,t)})}computeMask(e,t){return Qs(()=>{let n;return e=ub(e),n=null==t?sb(null,e.length):ub(t),this.runInternalGraph(e,n)[1]})}computeOutputShape(e){let t=Mx(e);if(t.length!==this.inputLayers.length)throw new tb(`Invalid inputShape argument ${e}: model has ${this.inputLayers.length} tensor inputs.`);let n={};for(let e=0;e<t.length;e++){let r=this.inputLayers[e],a=t[e];n[r.name+"_0_0"]=a}let r=Object.keys(this.nodesByDepth).map(e=>parseInt(e,10)).sort(gb);if(r.length>1)for(let e of r){let t=this.nodesByDepth[e];for(let e of t){let t=e.outboundLayer;if(-1!==this.inputLayers.map(e=>e.id).indexOf(t.id))continue;let r=[];for(let t=0;t<e.inboundLayers.length;t++){let a=e.inboundLayers[t],s=e.nodeIndices[t],i=e.tensorIndices[t],o=n[`${a.name}_${s}_${i}`];r.push(o)}let a=Mx(t.computeOutputShape(lb(r))),s=t.inboundNodes.indexOf(e);for(let e=0;e<a.length;e++){n[`${t.name}_${s}_${e}`]=a[e]}}}let a=[],s=[];for(let e=0;e<this.outputLayers.length;e++){let t=this.outputLayers[e],n=this.outputLayersNodeIndices[e],r=this.outputLayersTensorIndices[e],a=`${t.name}_${n}_${r}`;s.push(a)}for(let e=0;e<s.length;e++){let t=s[e];ib(t in n),a.push(n[t])}return lb(a)}runInternalGraph(e,t){null==t&&(t=sb(null,e.length));let n={};for(let r=0;r<this.inputs.length;++r){let a=this.inputs[r],s=e[r],i=t[r];n[a.id]=[s,i]}let r=Object.keys(this.nodesByDepth).map(e=>parseInt(e,10)).sort(gb);for(let e of r){let t=this.nodesByDepth[e];for(let e of t){let t=e.outboundLayer,r=e.inputTensors,a=e.outputTensors,s=new Array;for(let e of r)e.id in n&&s.push(n[e.id]);if(s.length===r.length){let r,i,o,l,u={};if(null!=e.callArgs&&(u=e.callArgs),1===s.length){let[e,n]=s[0];null==u.mask&&(u.mask=n),o=ub(t.call(e,u)),l=ub(t.computeMask(e,n)),r=[e],i=[n]}else r=s.map(e=>e[0]),i=s.map(e=>e[1]),null==u.mask&&(u.mask=i),o=ub(t.call(r,u)),l=ub(t.computeMask(r,i));if(t.activityRegularizer)throw new nb("LayersModel invocation with concrete Tensor value(s) in the presence of activity regularizer(s) is not supported yet.");for(let e=0;e<a.length;++e){let t=a[e],r=o[e],s=l[e];n[t.id]=[r,s]}}}}let a=[],s=[],i=[];for(let e of this.outputs){ib(e.id in n,`Could not compute output ${e.name} : ${e.id}`);let[t,r]=n[e.id];i.push(t.shape),a.push(t),s.push(r)}return[a,s,i]}buildNodeConversionMap(t){let n,r={};for(let t of this.layers){n=t instanceof e?1:0;for(let a=0;a<t.inboundNodes.length;a++){let s=e.nodeKey(t,a);this.containerNodes.has(s)&&(r[s]=n,n+=1)}}return r}getLayer(e,t){if(null!=t)return this.findLayer(t);if(null==e)throw new tb("Provide either a layer name or layer index");if("number"==typeof e)return this.findLayer(e);for(let t of this.layers)if(t.name===e)return t;throw new tb(`No such layer: ${e}`)}findLayer(e){if(this.layers.length<=e)throw new tb(`Was asked to retrieve layer at index ${e}, but model only has ${this.layers.length} layer(s).`);return this.layers[e]}calculateLosses(){return Qs(()=>{let t=[];for(let n of this.layers)for(let r=0;r<n.inboundNodes.length;++r){let a=e.nodeKey(n,r);this.containerNodes.has(a)&&t.push(...n.calculateLosses())}return t})}getConfig(){let t={name:this.name},n=this.buildNodeConversionMap(this.layers),r=[];for(let t of this.layers){let a=t.getClassName(),s=t.getConfig(),i=[];for(let r=0;r<t.inboundNodes.length;r++){let a=t.inboundNodes[r],s=e.nodeKey(t,r),o={};if(this.containerNodes.has(s)){if(a.callArgs)try{JSON.stringify(a.callArgs),o=a.callArgs}catch(e){console.warn(`Layer ${t.name} was passed non-serializable keyword arguments: ${a.callArgs}. They will not be included in the serialized model (and thus will be missing at deserialization time).`),o={}}if(a.inboundLayers.length>0){let t=[];for(let r=0;r<a.inboundLayers.length;r++){let s=a.inboundLayers[r],i=a.nodeIndices[r],l=a.tensorIndices[r],u=n[e.nodeKey(s,i)];null==u&&(u=0),t.push([s.name,u,l,o])}i.push(t)}}}let o={};o.name=t.name,o.className=a,o.config=s,o.inboundNodes=i,r.push(o)}t.layers=r;let a=[];for(let t=0;t<this.inputLayers.length;t++){let r=this.inputLayers[t],s=this.inputLayersNodeIndices[t],i=e.nodeKey(r,s);if(!this.containerNodes.has(i))continue;let o=n[i];null==o&&(o=0);let l=this.inputLayersTensorIndices[t];a.push([r.name,o,l])}t.inputLayers=a;let s=[];for(let t=0;t<this.outputLayers.length;t++){let r=this.outputLayers[t],a=this.outputLayersNodeIndices[t],i=e.nodeKey(r,a);if(!this.containerNodes.has(i))continue;let o=n[i];null==o&&(o=0);let l=this.outputLayersTensorIndices[t];s.push([r.name,o,l])}return t.outputLayers=s,t}static fromConfig(e,t,n={},r=!1){let a={},s={};function i(e,t){e.name in s?s[e.name].push(t):s[e.name]=[t]}function o(e,t){let n,r=[];for(let s of t){let o=s[0],l=s[1],u=s[2];if(n=null==s[3]?{}:s[3],!(o in a))return void i(e,t);let h=a[o];if(h.inboundNodes.length<=l)return void i(e,t);let d=h.inboundNodes[l];r.push(d.outputTensors[u])}r.length>0&&e.apply(lb(r),n)}function l(e){let n=e.name,s=Kv(e,null!=t.customObjects?t.customObjects:{});s.setFastWeightInitDuringBuild(r),a[n]=s,e.inboundNodes.forEach(e=>{if(!(e instanceof Array))throw new tb(`Corrupted configuration, expected array for nodeData: ${e}`);i(s,e)})}let u=t.name,h=t.layers;for(let e of h)l(e);for(;!bb(s);)for(let e of h){let t=a[e.name];if(t.name in s){let e=s[t.name];delete s[t.name];for(let n of e)o(t,n)}}let d=[],p=[],c=t.inputLayers;for(let e of c){let t=e[0],n=e[1],r=e[2];ib(t in a);let s=a[t].inboundNodes[n].outputTensors;d.push(s[r])}let f=t.outputLayers;for(let e of f){let t=e[0],n=e[1],r=e[2];ib(t in a);let s=a[t].inboundNodes[n].outputTensors;p.push(s[r])}return new e({inputs:d,outputs:p,name:u})}get stateful(){if(this._stateful)throw new tb("Container instance unexpectedly has _stateful = true. The statefulness of a Container is determined by the Layers it contains. Its _stateful property must remain the default false.");for(let e of this.layers)if(e.stateful)return!0;return!1}resetStates(){Qs(()=>{this.layers.forEach(e=>{e.stateful&&e.resetStates()})})}};function Cw(e,t){return function(e,t,n){let r=t.length;if(null==e||Array.isArray(e)&&0===e.length)return t.map(e=>null);if(1===r)return Array.isArray(e)&&1===e.length?e:"object"==typeof e&&t[0]in e?[e[t[0]]]:[e];if(Array.isArray(e)){if(e.length!==r)throw new Error(`Provided ${n} is an array of ${e.length} element(s), but the model has ${r} outputs. Make sure a set of weights is provided for each model output.`);return e}if("object"==typeof e&&Object.keys(e).length>0&&"object"==typeof e[Object.keys(e)[0]]){let n=[];return t.forEach(t=>{t in e?n.push(e[t]):n.push(null)}),n}throw new Error(`The model has multiple (${r}) outputs, so ${n} must be either an array with ${r} elements or an object with ${t} keys. Provided ${n} not understood: ${JSON.stringify(e)}`)}(e,t,"classWeight")}async function Ew(e,t,n,r){if(null!=t||null!=r)throw new Error("Support sampleWeight is not implemented yet");if(null!=n){let t=Qs(()=>{if(1===e.shape.length)return po(e);if(2===e.shape.length){if(e.shape[1]>1)return So(e,1);if(1===e.shape[1])return jo(e,[e.shape[0]]);throw new Error(`Encountered unexpected last-dimension size (${e.shape[1]}) during handling of class weights. The size is expected to be >= 1.`)}throw new Error(`Unexpected rank of target (y) tensor (${e.rank}) during handling of class weights. The rank is expected to be 1 or 2.`)}),r=Array.from(await t.data());ei(t);let a=[];return r.forEach(e=>{if(null==n[e])throw new Error(`classWeight must contain all classes in the training data. The class ${e} exists in the data but not in classWeight`);a.push(n[e])}),$d(a,"float32")}return null}function Aw(e,t){return yo(e,t)}function $w(e,t){let n,r,a=t;n=a.xs,r=a.ys,va.assert(null!=n&&null!=r,()=>`A Dataset iterator for fitDataset() is expected to generate objects of the form \`{xs: xVal, ys: yVal}\`, where the two values may be \`tf.Tensor\`, an array of Tensors, or a map of string to Tensor.  The provided Dataset instead generates ${t}`);let s=Rw("input",e.inputNames,n),i=Rw("output",e.outputNames,r),o=s[0].shape[0];va.assert(s.length===e.inputs.length,()=>`LayersModel has ${e.inputs.length} inputs, but the dataset provides ${s.length} inputs.  (Expected input keys: ${JSON.stringify(e.inputNames)})`),va.assert(i.length===e.outputs.length,()=>`LayersModel has ${e.outputs.length} outputs, but the dataset provides ${i.length} outputs.  (Expected output keys: ${JSON.stringify(e.outputNames)})`);for(let t=0;t<s.length;t++)va.assert(s[t].shape[0]===o,()=>`Batch size mismatch: input ${e.inputNames[t]} has ${s[t].shape[0]}; expected  ${o} based on input ${e.inputNames[0]}.`);for(let t=0;t<i.length;t++)va.assert(i[t].shape[0]===o,()=>`Batch size mismatch: output ${e.outputNames[t]} has ${i[t].shape[0]}; expected  ${o} based on input ${e.inputNames[0]}.`);return{xs:s,ys:i}}function Rw(e,t,n){if(n instanceof ts)return[n];if(Array.isArray(n))return va.assert(n.length===t.length,()=>`Received an array of ${n.length} Tensors, but expected ${t.length} to match the ${e} keys ${t}.`),n;{let r=[];for(let a of t){if(null==n[a])throw new tb(`The feature data generated by the dataset lacks the required ${e} key '${a}'.`);r.push(n[a])}return r}}async function Fw(e,t,n){let r=null!=n.batchesPerEpoch;if(va.assert(null!=e.optimizer,()=>"You must compile a model before training/testing. Use LayersModel.compile(modelCompileConfig)."),va.assert(null!=n,()=>"For fitDataset(), the 2nd argument (config) is required, but it is not provided in this call."),va.assert(null!=n.epochs&&n.epochs>0&&Number.isInteger(n.epochs),()=>`For fitDataset(), config.epochs is expected to be a positive integer, but got ${n.epochs}`),va.assert(!r||n.batchesPerEpoch>0&&Number.isInteger(n.batchesPerEpoch),()=>`For fitDataset(), config.batchesPerEpoch is expected to be a positive integer if specified, but got ${n.batchesPerEpoch}`),va.assert(null==n.validationSplit,()=>"`validationSplit` is not supported by `fitDataset()`. Use validationData instead."),e.isTraining)throw new Error("Cannot start training because another fit() call is ongoing.");e.isTraining=!0;try{let a,s,i=null!=n.validationData;if(i)if(Dw(n.validationData))va.assert(null==n.validationBatches||n.validationBatches>0&&Number.isInteger(n.validationBatches),()=>`For fitDataset() with dataset-based validation, config.validationBatches is expected not to be provided, or to be a positive integer, but got ${n.validationBatches}`);else{let e=function(e){if(3===e.length)throw new nb("Validation with sample weights is not implemented yet.");return{xs:e[0],ys:e[1]}}(n.validationData);a=e.xs,s=e.ys}let o,l=e.makeTrainFunction(),u=e.getDedupedMetricsNames();o=i?u.slice().concat(u.map(e=>"val_"+e)):u.slice();let h=Hv(n.callbacks,n.yieldEvery),d=null==n.verbose?1:n.verbose,{callbackList:p,history:c}=qv(h,d,n.epochs,null,null,function(e,t){let n=null;return null!=t.batchesPerEpoch?n=t.batchesPerEpoch:Number.isFinite(e.size)&&(n=e.size),n}(t,n),null,i,o);p.setModel(e),e.history=c,await p.onTrainBegin(),e.stopTraining_=!1;let f=null==n.initialEpoch?0:n.initialEpoch,m=await t.iterator();for(;f<n.epochs;){let o={};await p.onEpochBegin(f);let h=0,d=0;for(r||(m=await t.iterator());!r||h<n.batchesPerEpoch;){let t=await m.next();if(r&&t.done){console.warn(`You provided \`batchesPerEpoch\` as ${n.batchesPerEpoch}, but your dataset iterator ran out of data after ${h} batches; interrupting training. Make sure that your dataset can generate at least \`batchesPerEpoch * epochs\` batches (in this case, ${n.batchesPerEpoch*n.epochs} batches). You may need to use the repeat() function when building your dataset.`);break}if(null!=t.value){let{xs:r,ys:a}=$w(e,t.value),s={};s.batch=d,s.size=r[0].shape[0],await p.onBatchBegin(d,s);let i=[];if(null!=n.classWeight){let t=Cw(n.classWeight,e.outputNames);for(let e=0;e<t.length;++e)i.push(await Ew(a[e],null,t[e]))}let o=r.concat(a).concat(i),c=l(o);ei(o);for(let e=0;e<u.length;++e){let t=u[e],n=c[e];s[t]=n,ti(n)}await p.onBatchEnd(d,s),zv(s),d++,h++}if(r?h>=n.batchesPerEpoch:t.done){if(i){let t;t=Dw(n.validationData)?ub(await e.evaluateDataset(n.validationData,{batches:n.validationBatches})):ub(e.evaluate(a,s,{batchSize:null==n.validationBatchSize?32:n.validationBatchSize,verbose:0}));for(let n=0;n<e.metricsNames.length;++n)o[`val_${e.metricsNames[n]}`]=t[n]}break}if(e.stopTraining_)break}if(await p.onEpochEnd(f,o),f++,e.stopTraining_)break}return await p.onTrainEnd(),await e.history.syncData(),e.history}finally{e.isTraining=!1}}function Dw(e){return"function"==typeof e.iterator}function Mw(e){va.assert(e>0&&Number.isInteger(e),()=>`batchSize is required to be a positive integer, but got ${e}`)}function Ow(e,t,n){return null==e?[null]:Array.isArray(e)?e.map(e=>Jb(e,t,n-t)):Jb(e,t,n-t)}function Lw(e,t){return Qs(()=>null==e?null:Array.isArray(e)?e.map(e=>Lw(e,t)):ix(e,"int32"===t.dtype?t:ho(t,"int32")))}function Pw(e,t){let n=[],r=0,a=null;for(;r<e;)a=r+t,a>=e&&(a=e),n.push([r,a]),r=a;return n}function zw(e){let t=[];e instanceof ts&&(e=[e]);for(let n=0;n<e.length;++n){let r=e[n];if(1===r.rank)t.push(Yb(r,1));else{if(0===r.rank)throw new Error("Expected tensor to be at least 1D, but received a 0D tensor (scalar).");t.push(r)}}return t}function Bw(e,t){if(null==e)return;let n=[];if(t instanceof ts)n.push(t.id);else if(Array.isArray(t))t.forEach(e=>n.push(e.id));else if(null!=t)for(let e in t){let r=t[e];n.push(r.id)}let r=[];if(e instanceof ts)-1===n.indexOf(e.id)&&r.push(e);else if(Array.isArray(e))e.forEach(e=>{-1===n.indexOf(e.id)&&r.push(e)});else if(null!=e)for(let t in e){let a=e[t];-1===n.indexOf(a.id)&&r.push(a)}r.forEach(e=>{e.isDisposed||e.dispose()})}function Ww(e){return Array.isArray(e)}function Vw(e){return!function(e){return e instanceof ts}(e)&&!Ww(e)}function Uw(e,t,n,r=!0,a=""){if(null==t||0===t.length){if(null!=e){let t=!1;if(Ww(e)&&e.length>0)t=!0;else if(Vw(e)){for(let n in e)if(e.hasOwnProperty(n)){t=!0;break}}else t=!0;if(t)throw new tb(`Error when checking model ${a} expected no data, but got ${e}`)}return[]}if(null==e)return t.map(e=>null);let s;if(Vw(e)){s=[];for(let n of t){if(null==e[n])throw new tb(`No data provided for "${n}". Need data for each key in: ${t}`);s.push(e[n])}}else if(Ww(e)){if(e.length!==t.length)throw new tb(`Error when checking model ${a}: the Array of Tensors that you are passing to your model is not the size the model expected. Expected to see ${t.length} Tensor(s), but instead got the following list of Tensor(s): ${e}`);s=e}else{if(t.length>1)throw new tb(`The model ${a} expects ${t.length} Tensor(s), but only received one Tensor. Found: Tensor with shape ${e.shape}`);s=[e]}if(s=zw(s),null!=n)for(let e=0;e<t.length;++e){if(null==n[e])continue;let i=s[e];if(i.shape.length!==n[e].length)throw new tb(`Error when checking ${a}: expected ${t[e]} to have ${n[e].length} dimension(s). but got array with shape ${i.shape}`);for(let t=0;t<n[e].length;++t){if(0===t&&!r)continue;let s=i.shape[t],o=n[e][t];if(null!=o&&o>=0&&s!==o)throw new tb(`${a} expected a batch of elements where each example has shape [${n[e].slice(1,n[e].length)}] (i.e.,tensor shape [*,${n[e].slice(1,n[e].length)}]) but the ${a} received an input with ${i.shape[0]} examples, each with shape [${i.shape.slice(1,i.shape.length)}] (tensor shape [${i.shape}])`)}}return s}function Gw(e,t,n,r=!0,a=""){let s;if(Array.isArray(e)){if(e.length!==t.length)throw new tb(`Error when checking model ${a}: the Array of Tensors that you are passing to your model is not the size the the model expected. Expected to see ${t.length} Tensor(s), but instead got ${e.length} Tensors(s).`);s=e}else{if(t.length>1)throw new tb(`The model expects ${t.length} ${a} Tensors, but only received one Tensor. Found: array with shape ${JSON.stringify(e.shape)}.`);s=[e]}if(null!=n)for(let e=0;e<t.length;++e){if(null==n[e])continue;let i=s[e];if(i.shape.length!==n[e].length)throw new tb(`Error when checking ${a}: expected ${t[e]} to have ${n[e].length} dimension(s), but got array with shape ${JSON.stringify(i.shape)}`);for(let s=0;s<n[e].length;++s){if(0===s&&!r)continue;let o=i.shape[s],l=n[e][s];if(null!=l&&l!==o)throw new tb(`Error when checking ${a}: expected ${t[e]} to have shape ${JSON.stringify(n[e])} but got array with shape ${JSON.stringify(i.shape)}.`)}}}var Hw=class extends Tw{constructor(e){super(e),this.isTraining=!1}summary(e,t,n=console.log){if(!this.built)throw new tb("This model has never been called, thus its weights have not been created yet. So no summary can be displayed. Build the model first (e.g., by calling it on some test data).");xw(this,e,t,n)}compile(e){if(null==e.loss&&(e.loss=[]),this.loss=e.loss,"string"==typeof e.optimizer)this.optimizer_=function(e){let t={Adagrad:()=>Bf.adagrad(.01),Adadelta:()=>Bf.adadelta(1,.95,Xb()),Adam:()=>Bf.adam(.001,.9,.999,Xb()),Adamax:()=>Bf.adamax(.002,.9,.999,Xb(),0),RMSProp:()=>Bf.rmsprop(.001,.9,0,Xb()),SGD:()=>Bf.sgd(.01)};if(t.adagrad=t.Adagrad,t.adadelta=t.Adadelta,t.adam=t.Adam,t.adamax=t.Adamax,t.rmsprop=t.RMSProp,t.sgd=t.SGD,e in t)return t[e]();throw new tb(`Unknown Optimizer ${e}`)}(e.optimizer),this.isOptimizerOwned=!0;else{if(!(e.optimizer instanceof Ec))throw new tb("User-defined optimizer must be an instance of tf.Optimizer.");this.optimizer_=e.optimizer,this.isOptimizerOwned=!1}let t=[];if(Array.isArray(e.loss)||"string"==typeof e.loss||"function"==typeof e.loss)if(Array.isArray(e.loss)){if(e.loss.length!==this.outputs.length)throw new tb(`When passing an Array as loss, it should have one entry per model output. The model has ${this.outputs.length} output(s), but you passed loss=${e.loss}.`);t=e.loss.map(e=>aw(e))}else{let n=aw(e.loss);this.outputs.forEach(e=>{t.push(n)})}else{e.loss=e.loss;for(let t in e.loss)if(-1===this.outputNames.indexOf(t))throw new tb(`Unknown entry in loss dictionary: "${t}". Only expected the following keys: ${this.outputNames}`);for(let n of this.outputNames)null==e.loss[n]&&console.warn(`Output "${n}" is missing from loss dictionary. We assume this was done on purpose, and we will not be expecting data to be passed to ${n} during training`),t.push(aw(e.loss[n]))}this.lossFunctions=t,this.feedOutputNames=[],this.feedOutputShapes=[],this.feedLossFns=[];for(let e=0;e<this.outputs.length;++e){let t=this.internalOutputShapes[e],n=this.outputNames[e];this.feedOutputNames.push(n),this.feedOutputShapes.push(t),this.feedLossFns.push(this.lossFunctions[e])}let n=[];this.metrics=e.metrics,this.metricsNames=["loss"],this.metricsTensors=[],Pb("loss",()=>{for(let e=0;e<this.outputs.length;++e){if(-1!==n.indexOf(e))continue;let t=this.lossFunctions[e];this.outputs.length>1&&(this.metricsTensors.push([t,e]),this.metricsNames.push(this.outputNames[e]+"_loss"))}});let r=function(e,t){if(null==e||Array.isArray(e)&&0===e.length)return t.map(e=>[]);let n;if("string"==typeof e||"function"==typeof e)n=[e];else{if(!Array.isArray(e)&&"object"!=typeof e)throw new TypeError(`Type of metrics argument not understood. Expected an string,function, Array, or Object, found: ${e}`);n=e}if(Array.isArray(n))return t.map(e=>n);{let e=[];for(let r of t){let t=n.hasOwnProperty(r)?n[r]:[];Array.isArray(t)||(t=[t]),e.push(t)}return e}}(e.metrics,this.outputNames),a=(e,t,n)=>{this.outputNames.length>1&&(t=this.outputNames[e]+"_"+t),this.metricsNames.push(t),this.metricsTensors.push([n,e])};Pb("metric",()=>{for(let e=0;e<this.outputs.length;++e){if(-1!==n.indexOf(e))continue;(t=>{let n,r,s;for(let i of t){if("string"==typeof i&&-1!==["accuracy","acc","crossentropy","ce"].indexOf(i)){let t,a=this.internalOutputShapes[e];1===a[a.length-1]||this.lossFunctions[e]===tw?-1!==["accuracy","acc"].indexOf(i)?r=sw:-1!==["crossentropy","ce"].indexOf(i)&&(r=hw):this.lossFunctions[e]===ew?-1!==["accuracy","acc"].indexOf(i)?r=dw:-1!==["crossentropy","ce"].indexOf(i)&&(r=cw):-1!==["accuracy","acc"].indexOf(i)?r=iw:-1!==["crossentropy","ce"].indexOf(i)&&(r=pw),-1!==["accuracy","acc"].indexOf(i)?t="acc":-1!==["crossentropy","ce"].indexOf(i)&&(t="ce"),s=r,n=""+t}else s=mw(i),n=""+gw(i);let t;Pb(n,()=>{t=s}),a(e,n,t)}})(r[e])}}),this.collectedTrainableWeights=this.trainableWeights}checkTrainableWeightsConsistency(){null!=this.collectedTrainableWeights&&this.trainableWeights.length!==this.collectedTrainableWeights.length&&console.warn("Discrepancy between trainableweights and collected trainable weights. Did you set `model.trainable` without calling `model.compile()` afterwards?")}evaluate(e,t,n={}){let r=null==n.batchSize?32:n.batchSize;Mw(r);let a=this.standardizeUserDataXY(e,t,!0,r);try{let e=a[0].concat(a[1]);this.makeTestFunction();let t=this.testFunction;return lb(this.testLoop(t,e,r,n.verbose,n.steps))}finally{Bw(a[0],e),Bw(a[1],t)}}async evaluateDataset(e,t){return this.makeTestFunction(),async function(e,t,n){let r=null!=(n=n||{}).batches,a=e.testFunction,s=[];if(n.verbose>0)throw new nb("Verbose mode is not implemented yet.");va.assert(!r||n.batches>0&&Number.isInteger(n.batches),()=>`Test loop expects \`batches\` to be a positive integer, but received ${JSON.stringify(n.batches)}`);let i=function(e){return"function"==typeof e.next}(t)?t:await t.iterator(),o=0,l=0;for(;!r||l<n.batches;){let t=await i.next();if(s=Qs(()=>{if(t.value){let{xs:n,ys:r}=$w(e,t.value),i=n.concat(r),u=Qs(()=>a(i));if(ei(i),0===l)for(let e=0;e<u.length;++e)s.push(au(0));let h=i[0].shape[0];for(let e=0;e<u.length;++e){let t=u[e],n=s[e];s[e]=Qs(()=>fo(s[e],yo(h,t))),l>0&&ei(n)}ei(u),o+=h,++l}return s}),t.done){r&&console.warn(`Your dataset iterator ran out of data during evaluateDataset(). Interrupting evalution. Make sure that your dataset can generate at least \`batches\` batches (in this case, ${n.batches} batches). You may need to use the repeat() function when building your dataset.`);break}}for(let e=0;e<s.length;++e){let t=s[e];s[e]=go(s[e],o),ei(t)}return lb(s)}(this,e,t)}checkNumSamples(e,t,n,r="steps"){let a;if(null!=n){if(a=null,null!=t)throw new tb(`If ${r} is set, batchSize must be null or undefined.Got batchSize = ${t}`)}else{if(null==e)throw new tb(`Either the input data should have a defined shape, or ${r} shoud be specified.`);a=Array.isArray(e)?e[0].shape[0]:e.shape[0]}return a}execute(e,t){if(Array.isArray(t)&&0===t.length)throw new tb("`outputs` is an empty Array, which is not allowed.");let n=Array.isArray(t),r=n?t:[t],a=this.retrieveSymbolicTensors(r),s=new Jx;if(e instanceof ts&&(e=[e]),Array.isArray(e)){if(e.length!==this.inputs.length)throw new tb(`The number of inputs provided (${e.length}) does not match the number of inputs of this model (${this.inputs.length}).`);for(let t=0;t<this.inputs.length;++t)s.add(this.inputs[t],e[t])}else for(let t of this.inputs){let n=e[t.name];if(null==n)throw new tb(`No value is provided for the model's input ${t.name}`);s.add(t,n)}let i=tv(a,s);return n?i:i[0]}retrieveSymbolicTensors(e){let t=sb(null,e.length),n=e.length;for(let r of this.layers){let a=Array.isArray(r.output)?r.output:[r.output],s=a.map(e=>e.name);for(let r=0;r<e.length;++r){let i=s.indexOf(e[r]);if(-1!==i&&(t[r]=a[i],n--),0===n)break}if(0===n)break}if(n>0){let n=[];throw t.forEach((t,r)=>{null==t&&n.push(e[r])}),new tb(`Cannot find SymbolicTensors for output name(s): ${JSON.stringify(n)}`)}return t}predictLoop(e,t=32,n=!1){return Qs(()=>{let r=this.checkNumSamples(e);if(n)throw new nb("Verbose predictLoop() is not implemented yet.");let a=Pw(r,t),s=this.outputs.map(e=>[]);for(let t=0;t<a.length;++t)Qs(()=>{let n=a[t][0],r=a[t][1],s=Ow(e,n,r),i=[];if(Array.isArray(s))for(let e=0;e<s.length;++e)i.push({key:this.inputs[e],value:s[e]});else i.push({key:this.inputs[0],value:s});let o=new Jx(i);return tv(this.outputs,o)}).forEach((e,t)=>s[t].push(e));return lb(s.map(e=>Xo(e,0)))})}predict(e,t={}){let n=zw(e);Gw(n,this.inputNames,this.feedInputShapes,!1);try{let e=null==t.batchSize?32:t.batchSize;return Mw(e),this.predictLoop(n,e)}finally{Bw(n,e)}}predictOnBatch(e){Gw(e,this.inputNames,this.feedInputShapes,!0);let t=(Array.isArray(e)?e[0]:e).shape[0];return this.predictLoop(e,t)}standardizeUserDataXY(e,t,n=!0,r){if(null==this.optimizer_)throw new eb("You must compile a model before training/testing. Use LayersModel.compile(modelCompileArgs).");let a=[];for(let e=0;e<this.feedOutputShapes.length;++e){let t=this.feedOutputShapes[e];this.feedLossFns[e]===ew?a.push(t.slice(0,t.length-1).concat([1])):a.push(t)}if(function(e,t){let n=yb(e.map(e=>e.shape[0]));n.sort();let r=yb(t.map(e=>e.shape[0]));if(r.sort(),n.length>1)throw new tb(`All input Tensors (x) should have the same number of samples. Got array shapes: ${JSON.stringify(e.map(e=>e.shape))}`);if(r.length>1)throw new tb(`All target Tensors (y) should have the same number of samples. Got array shapes: ${JSON.stringify(t.map(e=>e.shape))}`);if(n.length>0&&r.length>0&&!va.arraysEqual(n,r))throw new tb(`Input Tensors should have the same number of samples as target Tensors. Found ${n[0]} input sample(s) and ${r[0]} target sample(s).`)}(e=Uw(e,this.feedInputNames,this.feedInputShapes,!1,"input"),t=Uw(t,this.feedOutputNames,a,!1,"target")),function(e,t,n){let r=[Zv,tw,Qv];for(let a=0;a<e.length;++a){let s=e[a],i=t[a],o=n[a];if(null!=i){if(i===Qv&&1===s.shape[s.shape.length-1])throw new tb(`You are passing a target array of shape ${s.shape} while using a loss 'categorical_crossentropy'. 'categorical_crossentropy'expects targets to be binary matrices (1s and 0s) of shape [samples, classes].`);if(-1!==r.indexOf(i)){let e=s.shape.slice(1),t=o.slice(1);for(let n=0;n<e.length;++n){let r=e[n],a=t[n];if(null!=a&&r!==a)throw new tb(`A target Tensor with shape ${s.shape} was passed for an output of shape ${o}, while using a loss function that expects targets to have the same shape as the output.`)}}}}}(t,this.feedLossFns,this.feedOutputShapes),this.stateful&&null!=r&&r>0&&e[0].shape[0]%r!==0)throw new tb(`In a stateful network, you should only pass inputs with a number of samples that is divisible by the batch size ${r}. Found: ${e[0].shape[0]} sample(s).`);return[e,t]}async standardizeUserData(e,t,n,r,a=!0,s){let[i,o]=this.standardizeUserDataXY(e,t,a,s);if(null!=n)throw new Error("sample weight is not supported yet.");let l=null;if(null!=r){let e=Cw(r,this.outputNames);l=[];for(let t=0;t<e.length;++t)l.push(await Ew(o[t],null,e[t]))}return[i,o,l]}testLoop(e,t,n,r=0,a){return Qs(()=>{let s=this.checkNumSamples(t,n,a,"steps"),i=[];if(r>0)throw new nb("Verbose mode is not implemented yet.");if(null!=a)throw new nb("steps mode in testLoop() is not implemented yet");{let r=Pw(s,n),a=$d(Kb(0,s));for(let n=0;n<r.length;++n){let s=r[n][0],o=r[n][1],l=Jb(a,s,o-s),u=Lw(t,l),h=e(u);if(0===n)for(let e=0;e<h.length;++e)i.push(au(0));for(let e=0;e<h.length;++e){let t=h[e];i[e]=fo(i[e],yo(o-s,t))}}for(let e=0;e<i.length;++e)i[e]=go(i[e],s)}return i})}getDedupedMetricsNames(){let e=this.metricsNames,t=[];for(let n=0;n<e.length;++n){let r=e[n],a=r;if(ob(e,r)>1){a+=`_${ob(e.slice(0,n),r)}`}t.push(a)}return t}makeTrainFunction(){return e=>{let t=[],n=e.slice(0,this.inputs.length),r=e.slice(this.inputs.length,this.inputs.length+this.outputs.length),a=e.slice(this.inputs.length+this.outputs.length,this.inputs.length+2*this.outputs.length),s=[],i=this.collectedTrainableWeights.map(e=>e.read());return[this.optimizer_.minimize(()=>{let e=[];for(let t=0;t<this.inputs.length;++t)e.push({key:this.inputs[t],value:n[t]});let i,o=new Jx(e),l=tv(this.outputs,o,{training:!0});for(let e=0;e<this.lossFunctions.length;++e){let n=(0,this.lossFunctions[e])(r[e],l[e]);null!=a[e]&&(n=Aw(n,a[e]));let s=th(n);t.push(s),i=0===e?n:fo(i,n)}for(let e=0;e<this.metricsTensors.length;++e){let n;if(this.outputs.length>1&&e<this.outputs.length)n=t[e];else{let t=this.metricsTensors[e][0],a=this.metricsTensors[e][1];n=th(t(r[a],l[a]))}ti(n),s.push(n)}return i=th(i),this.calculateLosses().forEach(e=>{i=fo(i,e)}),i},!0,i)].concat(s)}}makeTestFunction(){this.testFunction=e=>Qs(()=>{let t,n=[],r=e.slice(0,this.inputs.length),a=e.slice(this.inputs.length,this.inputs.length+this.outputs.length),s=[];for(let e=0;e<this.inputs.length;++e)s.push({key:this.inputs[e],value:r[e]});let i=new Jx(s),o=tv(this.outputs,i);for(let e=0;e<this.lossFunctions.length;++e){let r=this.lossFunctions[e],s=th(r(a[e],o[e]));t=0===e?s:fo(t,s),n.push(t)}for(let e=0;e<this.metricsTensors.length;++e){let t=this.metricsTensors[e][0],r=this.metricsTensors[e][1],s=th(t(a[r],o[r]));n.push(s)}return n})}async fit(e,t,n={}){if(this.isTraining)throw new Error("Cannot start training because another fit() call is ongoing.");let r,a,s,i,o,l,u,h,d;this.isTraining=!0;try{let p=null==n.batchSize?32:n.batchSize;Mw(p);let c=await this.standardizeUserData(e,t,n.sampleWeight,n.classWeight,!1,p);r=c[0],a=c[1],d=c[2];let f,m=!1;if(null!=n.validationData&&n.validationData.length>0){if(m=!0,2!==n.validationData.length)throw 3===n.validationData.length?new nb("validationData including sample weights is not supported yet."):new tb(`When passing validation data, it must contain 2 (valX, valY) or 3 (valX, valY, valSampleWeight) items; ${n.validationData} is invalid.`);o=n.validationData[0],l=n.validationData[1];let e=await this.standardizeUserData(o,l,null,null,!0,p);u=e[0],h=e[1],f=u.concat(h)}else if(null!=n.validationSplit&&n.validationSplit>0&&n.validationSplit<1){m=!0;let e=Math.floor(r[0].shape[0]*(1-n.validationSplit)),t=r[0].shape[0];u=Ow(r,e,t),s=r,r=Ow(r,0,e),h=Ow(a,e,t),i=a,a=Ow(a,0,e),f=u.concat(h)}else null!=n.validationSteps&&(m=!0);let g=r.concat(a).concat(d);this.checkTrainableWeightsConsistency();let y,b,x=this.makeTrainFunction(),v=this.getDedupedMetricsNames();m?(this.makeTestFunction(),y=this.testFunction,b=v.slice().concat(v.map(e=>"val_"+e))):(y=null,f=[],b=v.slice());let w=Hv(n.callbacks,n.yieldEvery);return await this.fitLoop(x,g,v,p,n.epochs,n.verbose,w,y,f,n.shuffle,b,n.initialEpoch,null,null)}finally{this.isTraining=!1,Bw(r,e),Bw(a,t),Bw(s,e),Bw(i,t),Bw(u,o),Bw(h,l),null!=d&&ei(d)}}async fitLoop(e,t,n,r,a,s,i,o,l,u,h,d,p,c){null==r&&(r=32),null==a&&(a=1),null==u&&(u=!0),null==d&&(d=0);let f=!1;if(null!=o&&null!=l&&(f=!0),null!=c&&(f=!0,null==p))throw new tb("Can only use `validationSteps` when doing step-wise training, i.e., `stepsPerEpoch` must be set.");let m,g=this.checkNumSamples(t,r,p,"steps_per_epoch");null!=g&&(m=Kb(0,g)),null==s&&(s=1);let{callbackList:y,history:b}=qv(i,s,a,d,g,p,r,f,h);y.setModel(this),this.history=b,await y.onTrainBegin(),this.stopTraining_=!1;for(let s=d;s<a;++s){await y.onEpochBegin(s);let a={};if(null!=p)throw new nb("stepsPerEpoch mode is not implemented yet.");{if("batch"===u)throw new nb("batch shuffling is not implemneted yet");u&&va.shuffle(m);let s=$d(m),i=Pw(g,r);for(let u=0;u<i.length;++u){let h={};if(await y.onBatchBegin(u,h),Qs(()=>{let d=i[u][0],p=i[u][1],c=Jb(s,d,p-d);h.batch=u,h.size=p-d;let m=Lw(t,c),g=e(m);for(let e=0;e<n.length;++e){let t=n[e],r=g[e];h[t]=r,ti(r)}if(u===i.length-1&&f){let e=this.testLoop(o,l,r);for(let t=0;t<n.length;++t){let r=n[t],s=e[t];ti(s),a["val_"+r]=s}}}),await y.onBatchEnd(u,h),zv(h),this.stopTraining_)break}s.dispose()}if(await y.onEpochEnd(s,a),this.stopTraining_)break}return await y.onTrainEnd(),await this.history.syncData(),this.history}async fitDataset(e,t){return Fw(this,e,t)}async trainOnBatch(e,t){let n=await this.standardizeUserData(e,t),r=n[0],a=n[1],s=this.makeTrainFunction()(r.concat(a)),i=[];for(let e of s){let t=await e.data();i.push(t[0])}return ei(s),Bw(n[0],e),Bw(n[1],t),lb(i)}getNamedWeights(e){let t=[],n=null!=e&&e.trainableOnly,r=n?this.trainableWeights:this.weights,a=this.getWeights(n);for(let e=0;e<r.length;++e)n&&!r[e].trainable||t.push({name:r[e].originalName,tensor:a[e]});return t}set stopTraining(e){this.stopTraining_=e}get stopTraining(){return this.stopTraining_}get optimizer(){return this.optimizer_}set optimizer(e){this.optimizer_!==e&&(this.optimizer_=e,this.isOptimizerOwned=!1)}dispose(){let e=super.dispose();if(0===e.refCountAfterDispose&&null!=this.optimizer&&this.isOptimizerOwned){let t=Ys().numTensors;this.optimizer_.dispose(),e.numDisposedVariables+=t-Ys().numTensors}return e}getLossIdentifiers(){let e;if("string"==typeof this.loss)e=hb(this.loss);else if(Array.isArray(this.loss)){for(let e of this.loss)if("string"!=typeof e)throw new Error("Serialization of non-string loss is not supported.");e=this.loss.map(e=>hb(e))}else{let t=Object.keys(this.loss);e={};let n=this.loss;for(let r of t){if("string"!=typeof n[r])throw new Error("Serialization of non-string loss is not supported.");e[r]=hb(n[r])}}return e}getMetricIdentifiers(){if("string"==typeof this.metrics||"function"==typeof this.metrics)return[hb(gw(this.metrics))];if(Array.isArray(this.metrics))return this.metrics.map(e=>hb(gw(e)));{let e={};for(let t in this.metrics)e[t]=hb(gw(this.metrics[t]));return e}}getTrainingConfig(){return{loss:this.getLossIdentifiers(),metrics:this.getMetricIdentifiers(),optimizer_config:{class_name:this.optimizer.getClassName(),config:this.optimizer.getConfig()}}}loadTrainingConfig(e){if(null!=e.weighted_metrics)throw new Error("Loading weight_metrics is not supported yet.");if(null!=e.loss_weights)throw new Error("Loading loss_weights is not supported yet.");if(null!=e.sample_weight_mode)throw new Error("Loading sample_weight_mode is not supported yet.");let t,n,r=Kv(Sw(e.optimizer_config));if("string"==typeof e.loss)t=db(e.loss);else if(Array.isArray(e.loss))t=e.loss.map(e=>db(e));else if(null!=e.loss){t={};for(let n in e.loss)t[n]=db(e.loss[n])}if(Array.isArray(e.metrics))n=e.metrics.map(e=>db(e));else if(null!=e.metrics){n={};for(let t in e.metrics)n[t]=db(e.metrics[t])}this.compile({loss:t,metrics:n,optimizer:r})}async save(e,t){if("string"==typeof e){let t=Pc.getSaveHandlers(e);if(0===t.length)throw new tb(`Cannot find any save handlers for URL '${e}'`);if(t.length>1)throw new tb(`Found more than one (${t.length}) save handlers for URL '${e}'`);e=t[0]}if(null==e.save)throw new tb("LayersModel.save() cannot proceed because the IOHandler provided does not have the `save` attribute defined.");let n=await Pc.encodeWeights(this.getNamedWeights(t)),r={modelTopology:this.toJSON(null,!1),format:"layers-model",generatedBy:`TensorFlow.js tfjs-layers v${Nw}`,convertedBy:null};if(null!=t&&t.includeOptimizer&&null!=this.optimizer){r.trainingConfig=this.getTrainingConfig();let e="optimizer",{data:t,specs:a}=await Pc.encodeWeights(await this.optimizer.getWeights(),e);n.specs.push(...a),n.data=Pc.concatenateArrayBuffers([n.data,t])}return null!=this.userDefinedMetadata&&(yw(this.userDefinedMetadata,this.name,!0),r.userDefinedMetadata=this.userDefinedMetadata),r.weightData=n.data,r.weightSpecs=n.specs,e.save(r)}setUserDefinedMetadata(e){yw(e,this.name),this.userDefinedMetadata=e}getUserDefinedMetadata(){return this.userDefinedMetadata}};Hw.className="Model",kc.registerClass(Hw);var jw=class extends Hw{};async function qw(e,t){"modelTopology"in e||(e={modelTopology:e});let n=e.modelTopology;null!=n.model_config&&(n=n.model_config);let r=Kv(Sw(n),t);if(null!=e.weightsManifest){let t=await Pc.loadWeights(e.weightsManifest,e.pathPrefix,r.weights.map(e=>e.originalName)),n={};for(let e of r.weights)n[e.originalName]=t[e.originalName];r.loadWeights(n),ei(t)}return r}async function Kw(e,t){if(null==t&&(t={}),"string"==typeof e){let n=Pc.getLoadHandlers(e,t);if(0===n.length)n.push(Pc.browserHTTPRequest(e,t));else if(n.length>1)throw new tb(`Found more than one (${n.length}) load handlers for URL '${e}'`);e=n[0]}return async function(e,t,n){if(null==n&&(n={}),null==e.load)throw new tb("Cannot proceed with model loading because the IOHandler provided does not have the `load` method implemented.");let r=await e.load(),a=r.modelTopology;null!=a.model_config&&(a=a.model_config);let s=null==n.strict||n.strict,i=null!=r.weightData&&null!=r.weightSpecs&&s,o=Kv(Sw(a),t,i),l=r.trainingConfig;if(null!=l&&o.loadTrainingConfig(l),null!=r.userDefinedMetadata&&o.setUserDefinedMetadata(r.userDefinedMetadata),null!=r.weightData){if(null==r.weightSpecs)throw new tb("LayersModel artifacts contains weight data, but not weight specs. Therefore loading of weights cannot proceed.");let{modelWeights:e,optimizerWeights:t}=function(e,t){let n=Pc.decodeWeights(e,t),r={},a=[];return t.forEach(e=>{"optimizer"===e.group?a.push({name:e.name,tensor:n[e.name]}):r[e.name]=n[e.name]}),{modelWeights:r,optimizerWeights:a}}(r.weightData,r.weightSpecs);o.loadWeights(e,s),null!=o.optimizer&&t.length>0&&await o.optimizer.setWeights(t),ei(e),ei(t.map(e=>e.tensor))}return o}(e,void 0,t)}jw.className="Functional",kc.registerClass(jw);var Xw=class e extends Hw{constructor(e){if(super({inputs:[],outputs:[]}),e=e||{},this.trainable=!0,this.built=!1,this.name=null!=e.name?e.name:Tb("sequential_"),null!=e.layers)for(let t of e.layers)this.add(t)}checkShape(e){if(e.inboundNodes[0].outputTensors[0].shape.some(e=>e<0))throw new tb(`Negative dimension size caused by adding layer ${e.name} with input shape [${e.inboundNodes[0].inputTensors[0].shape}]`)}add(t){let n,r=t instanceof e||t instanceof Hw;if(r){if(n=t,1!==n.outputs.length)throw new tb("All layers in a Sequential model should have a single output tensor. For multi-output layers, use the functional API.");if(1!==n.inputs.length)throw new tb("All layers in a Sequential model should have a single input tensor. For multi-input layers, use the functional API.")}if(0===this.outputs.length){if(0===t.inboundNodes.length){if(null==t.batchInputShape)throw new tb("The first layer in a Sequential model must get an `inputShape` or `batchInputShape` argument.");let e=Yx({batchShape:t.batchInputShape,dtype:t.dtype,name:t.name+"_input"});t.apply(e)}if(r)this.outputs=n.outputs,this.inputs=n.inputs;else{if(1!==t.inboundNodes.length)throw new tb(`A layer added to a Sequential model must not already be connected somewhere else. LayersModel received layer ${t.name} which has ${t.inboundNodes.length} pre-existing inbound connections.`);if(1!==t.inboundNodes[0].outputTensors.length)throw new tb("All layers in a Sequential model should have a single output tensor. For multi-output layers, use the functional API.");this.checkShape(t),this.outputs=[t.inboundNodes[0].outputTensors[0]],this.inputs=Xx(this.outputs[0])}this.inboundNodes=[],new jx({outboundLayer:this,inboundLayers:[],nodeIndices:[],tensorIndices:[],inputTensors:this.inputs,outputTensors:this.outputs,inputMasks:sb(null,this.inputs.length),outputMasks:[null],inputShapes:this.inputs.map(e=>e.shape),outputShapes:this.outputs[0].shape})}else{let e=t.apply(this.outputs[0]);if(Array.isArray(e))throw new TypeError("All layers in a Sequential model should have a single output tensor. For multi-output layers, use the functional API.");this.checkShape(t),this.outputs=[e],this.inboundNodes[0].outputTensors=this.outputs,this.inboundNodes[0].outputShapes=[this.outputs[0].shape]}this.layers.push(t),this.built=!1}pop(){if(0===this.layers.length)throw new TypeError("There are no layers in the model.");if(this.layers.pop(),0===this.layers.length)this.outputs=[],this.inboundNodes=[],this.outboundNodes=[];else{let e=this.layers.length-1;this.layers[e].outboundNodes=[],this.outputs=[this.layers[e].output],this.inboundNodes[0].outputTensors=this.outputs,this.inboundNodes[0].outputShapes=[this.outputs[0].shape]}}call(e,t){return null==this.model&&this.build(),this.model.call(e,t)}build(e){if(Lx(e),0===this.inputs.length||0===this.outputs.length)throw new TypeError("Sequential model cannot be built: model is empty. Add some layers first.");this.model=new Hw({inputs:this.inputs,outputs:this.outputs[0],name:this.name+"_model"}),this.model.trainable=this.trainable,this.supportsMasking=this.model.supportsMasking,this.inputLayers=this.model.inputLayers,this.inputLayersNodeIndices=this.model.inputLayersNodeIndices,this.inputLayersTensorIndices=this.model.inputLayersTensorIndices,this.outputLayers=this.model.outputLayers,this.outputLayersNodeIndices=this.model.outputLayersNodeIndices,this.outputLayersTensorIndices=this.model.outputLayersTensorIndices,this.nodesByDepth=this.model.nodesByDepth,this.containerNodes=this.model.containerNodes,this.outputNames=this.model.outputNames,this.inputNames=this.model.inputNames,this.built=!0}countParams(){return this.built||this.build(),super.countParams()}summary(e,t,n=console.log){this.built||this.build(),super.summary(e,t,n)}setWeights(e){null==this.model&&this.build(),this.model.setWeights(e)}evaluate(e,t,n={}){if(!this.built)throw new eb("The model needs to be compiled before being used.");return this.model.evaluate(e,t,n)}async evaluateDataset(e,t){if(!this.built)throw new eb("The model needs to be compiled before being used.");return this.model.evaluateDataset(e,t)}predict(e,t={}){return null==this.model&&this.build(),this.model.predict(e,t)}predictOnBatch(e){return null==this.model&&this.build(),this.model.predictOnBatch(e)}compile(e){this.build(),this.model.compile(e),this.optimizer_=this.model.optimizer,this.isOptimizerOwned=this.model.isOptimizerOwned,this.loss=this.model.loss,this.metrics=this.model.metrics,this.metricsTensors=this.model.metricsTensors,this.metricsNames=this.model.metricsNames}get optimizer(){return null==this.model?void 0:this.model.optimizer}set optimizer(e){this.model.optimizer=e}async fit(e,t,n={}){if(!this.built)throw new eb("The model needs to be compiled before being used.");return this.model.fit(e,t,n)}async fitDataset(e,t){if(!this.built)throw new eb("The model needs to be compiled before being used.");return this.model.fitDataset(e,t)}async trainOnBatch(e,t){return this.model.trainOnBatch(e,t)}static fromConfig(t,n,r={},a=!1){let s,i={};if(n instanceof Array){if(null==n[0].className||"Merge"===n[0].className)throw new tb("Legacy serialization format not supported yet.");s=n}else va.assert(null!=n.layers,()=>"When the config data for a Sequential model is not an Array, it must be an Object that contains the 'layers' field."),s=n.layers,delete n.layers,i=n;let o=new t(i);if(!(o instanceof e))throw new nb(`Sequential.fromConfig called on non-Sequential input: ${o}`);for(let e of s){let t=Kv(e,void 0,a);a&&t.setFastWeightInitDuringBuild(!0),o.add(t)}return o}set stopTraining(e){if(null==this.model)throw new tb("Cannot set the stopTraining property of a sequential model before it is compiled.");this.model.stopTraining=e}get stopTraining(){if(null==this.model)throw new tb("Cannot get the stopTraining property of a sequential model before it is compiled.");return this.model.stopTraining}getConfig(){let e=[];for(let t of this.layers){let n={};n.className=t.getClassName(),n.config=t.getConfig(),e.push(n)}return{name:this.name,layers:e}}};function Zw(e){return new Hw(e)}function Yw(e){return new Xw(e)}function Jw(e){return Yx(e)}function Qw(e,t){jv.registerCallbackConstructor(e,t)}Xw.className="Sequential",kc.registerClass(Xw);var ek=class extends kc.Serializable{getConfig(){return{}}},tk=class extends ek{apply(e,t=1){return function(e,t=1){if(1!==t)throw new nb(`Support for alpha values other than 1 (${t}) is not implemented yet.`);return Gl(e)}(e,t)}};tk.className="elu",kc.registerClass(tk);var nk=class extends ek{apply(e){return ud(e)}};nk.className="selu",kc.registerClass(nk);var rk=class extends ek{apply(e){return ed(e)}};rk.className="relu",kc.registerClass(rk);var ak=class extends ek{apply(e){return Qs(()=>sh(6,ed(e)))}};ak.className="relu6",kc.registerClass(ak);var sk=class extends ek{apply(e){return e}};sk.className="linear",kc.registerClass(sk);var ik=class extends ek{apply(e){return Yo(e)}};ik.className="sigmoid",kc.registerClass(ik);var ok=class extends ek{apply(e){return function(e){return Qs(()=>{let t=fo(.5,yo(.2,e));return cl(t,0,1)})}(e)}};ok.className="hardSigmoid",kc.registerClass(ok);var lk=class extends ek{apply(e){return zu(e)}};lk.className="softplus",kc.registerClass(lk);var uk=class extends ek{apply(e){return function(e){return Qs(()=>go(e,fo(bo(e),1)))}(e)}};uk.className="softsign",kc.registerClass(uk);var hk=class extends ek{apply(e){return Qo(e)}};hk.className="tanh",kc.registerClass(hk);var dk=class extends ek{apply(e,t=-1){return xd(e,t)}};dk.className="softmax",kc.registerClass(dk);var pk=class extends ek{apply(e,t=-1){return Vu(e,t)}};pk.className="logSoftmax",kc.registerClass(pk);var ck=class extends ek{apply(e){return Qs(()=>Qs(()=>{let t=Math.sqrt(2),n=yo(.5,fo(1,jl(go(e,t))));return yo(e,n)}))}};ck.className="gelu",kc.registerClass(ck);var fk=class extends ek{apply(e){return Qs(()=>yo(.5,yo(e,fo(1,Qo(yo(su(go(2,Math.PI)),fo(e,yo(.044715,ru(e,3)))))))))}};fk.className="gelu_new",kc.registerClass(fk);var mk=class extends ek{apply(e){return Qs(()=>yo(e,Qo(zu(e))))}};mk.className="mish",kc.registerClass(mk);var gk=class extends ek{apply(e,t=1){return Qs(()=>yo(Yo(yo(e,t)),e))}};function yk(e){return e.getClassName()}function bk(e,t={}){return mb(e,kc.SerializationMap.getMap().classNameMap,t,"activation")}function xk(e){if(null==e){let e={className:"linear",config:{}};return bk(e)}if("string"==typeof e){let t={};return t.className=e,t.config={},bk(t)}return e instanceof ek?e:bk(e)}function vk(e){if(null!=e&&"object"!=typeof e)throw new Error(`Argument to L1L2 regularizer's constructor is expected to be an object, but received: ${e}`)}gk.className="swish",kc.registerClass(gk);var wk=class extends kc.Serializable{},kk=class extends wk{constructor(e){super(),vk(e),this.l1=null==e||null==e.l1?.01:e.l1,this.l2=null==e||null==e.l2?.01:e.l2,this.hasL1=0!==this.l1,this.hasL2=0!==this.l2}apply(e){return Qs(()=>{let t=nh([1]);return this.hasL1&&(t=fo(t,ou(yo(this.l1,bo(e))))),this.hasL2&&(t=fo(t,ou(yo(this.l2,ox(e))))),jo(t,[])})}getConfig(){return{l1:this.l1,l2:this.l2}}static fromConfig(e,t){return new e({l1:t.l1,l2:t.l2})}};kk.className="L1L2",kc.registerClass(kk);var Ik={l1l2:"L1L2"};function Sk(e){return cb(e)}function _k(e,t={}){return mb(e,kc.SerializationMap.getMap().classNameMap,t,"regularizer")}function Nk(e){if(null==e)return null;if("string"==typeof e){return _k({className:e in Ik?Ik[e]:e,config:{}})}return e instanceof wk?e:_k(e)}var Tk=class extends Kx{constructor(e){super(e??{}),this.supportsMasking=!0,null!=e&&(this.maxValue=e.maxValue)}call(e,t){e=Ox(e);let n=ed(e);return null!=this.maxValue&&(n=cl(n,0,this.maxValue)),n}computeOutputShape(e){return e}getConfig(){let e={maxValue:this.maxValue},t=super.getConfig();return Object.assign(e,t),e}};Tk.className="ReLU",kc.registerClass(Tk);var Ck=class extends Kx{constructor(e){super(e??{}),this.DEFAULT_ALPHA=.3,null==e&&(e={}),this.alpha=null==e.alpha?this.DEFAULT_ALPHA:e.alpha}call(e,t){let n=Ox(e);return Su(n,this.alpha)}computeOutputShape(e){return e}getConfig(){let e={alpha:this.alpha},t=super.getConfig();return Object.assign(e,t),e}};Ck.className="LeakyReLU",kc.registerClass(Ck);var Ek=class extends Kx{constructor(e){if(super(e??{}),this.DEFAULT_ALPHA_INITIALIZER="zeros",null==e&&(e={}),this.supportsMasking=!0,this.alphaInitializer=Fx(e.alphaInitializer||this.DEFAULT_ALPHA_INITIALIZER),this.alphaRegularizer=Nk(e.alphaRegularizer),this.alphaConstraint=mv(e.alphaConstraint),null==e.sharedAxes)this.sharedAxes=null;else if(Array.isArray(e.sharedAxes))this.sharedAxes=e.sharedAxes;else{if("number"!=typeof e.sharedAxes)throw new tb(`Expected sharedAxes to be a number or an array of numbers, but got ${e.sharedAxes}`);this.sharedAxes=[e.sharedAxes]}}build(e){let t=(e=Lx(e)).slice(1);if(null!=this.sharedAxes)for(let e of this.sharedAxes)t[e-1]=1;this.alpha=this.addWeight("alpha",t,"float32",this.alphaInitializer,this.alphaRegularizer,!0,this.alphaConstraint);let n={};if(null!=this.sharedAxes)for(let t=1;t<e.length;++t)n[t]=e[t];this.inputSpec=[new Ux({ndim:e.length,axes:n})],this.built=!0}call(e,t){return e=Ox(e),kh(e,this.alpha.read())}getConfig(){let e={alphaInitializer:Rx(this.alphaInitializer),alphaRegularizer:Sk(this.alphaRegularizer),alphaConstraint:cv(this.alphaConstraint),sharedAxes:this.sharedAxes},t=super.getConfig();return Object.assign(e,t),e}};Ek.className="PReLU",kc.registerClass(Ek);var Ak=class extends Kx{constructor(e){if(super(e??{}),this.DEFAULT_ALPHA=1,null==e&&(e={}),null!=e.alpha&&e.alpha!==this.DEFAULT_ALPHA)throw new nb(`Non-default alpha value (${e.alpha}) is not supported by the ELU layer yet.`);this.alpha=null==e.alpha?this.DEFAULT_ALPHA:e.alpha}call(e,t){let n=Ox(e);return Gl(n)}computeOutputShape(e){return e}getConfig(){let e={alpha:this.alpha},t=super.getConfig();return Object.assign(e,t),e}};Ak.className="ELU",kc.registerClass(Ak);var $k=class extends Kx{constructor(e){super(e??{}),this.DEFAULT_THETA=1,null==e&&(e={}),this.theta=null==e.theta?this.DEFAULT_THETA:e.theta}call(e,t){let n=Ox(e);return yo(n,ho(bu(n,this.theta),"float32"))}computeOutputShape(e){return e}getConfig(){let e={theta:this.theta},t=super.getConfig();return Object.assign(e,t),e}};$k.className="ThresholdedReLU",kc.registerClass($k);var Rk=class extends Kx{constructor(e){super(e??{}),this.DEFAULT_AXIS=1,null==e&&(e={}),this.softmax=(new dk).apply,this.axis=null==e.axis?this.DEFAULT_AXIS:e.axis}call(e,t){return Qs(()=>{let n=Ox(e),r=t.mask;if(null!=r){let e=yo(Wu(rh(n.shape),ho(r,n.dtype)),au(-1e9));n=fo(n,e)}return this.axis instanceof Array?this.axis.length>1?du(Wu(n,Uu(n,this.axis,!0))):this.softmax(n,this.axis[0]):this.softmax(n,this.axis)})}computeOutputShape(e){return e}getConfig(){let e={axis:this.axis},t=super.getConfig();return Object.assign(e,t),e}};function Fk(e,t,n){if("number"==typeof e)return sb(e,t);if(e.length!==t)throw new tb(`The ${n} argument must be an integer or tuple of ${t} integers. Received: ${e.length} elements.`);for(let r=0;r<t;++r){let a=e[r];if(!Gb(a))throw new tb(`The ${n} argument must be an integer or tuple of ${t} integers. Received: ${JSON.stringify(e)} including a non-integer number ${a}`)}return e}function Dk(e,t,n,r,a=1){if(null==e)return e;let s;return s="same"===n?e:e-(t+(t-1)*(a-1))+1,Math.floor((s+r-1)/r)}function Mk(e,t,n,r){if(null==e)return null;if("valid"===r)e=e*t+qb([n-t,0]);else{if("same"!==r)throw new tb(`Unsupport padding mode: ${r}.`);e*=t}return e}function Ok(e,t){return Qs(()=>(Db(t),"channelsFirst"===t?Jd(e,[0,2,3,1]):e))}function Lk(e,t){return Qs(()=>(Db(t),"channelsFirst"===t?Jd(e,[0,2,3,4,1]):e))}function Pk(e,t,n,r=[1,1],a="valid",s,i,o=null){return Qs(()=>{if(null==s&&(s="channelsLast"),Db(s),3!==e.rank&&4!==e.rank)throw new tb(`conv2dWithBiasActivation expects input to be of rank 3 or 4, but received ${e.rank}.`);if(3!==t.rank&&4!==t.rank)throw new tb(`conv2dWithBiasActivation expects kernel to be of rank 3 or 4, but received ${e.rank}.`);let l=Ok(e,s);if("causal"===a)throw new nb("The support for CAUSAL padding mode in conv1dWithBias is not implemented yet.");return l=op.conv2d({x:l,filter:t,strides:r,pad:"same"===a?"same":"valid",dilations:i,dataFormat:"NHWC",bias:n,activation:o}),"channelsFirst"===s&&(l=Jd(l,[0,3,1,2])),l})}Rk.className="Softmax",kc.registerClass(Rk);var zk=class e extends Kx{constructor(t,n){if(super(n),this.bias=null,this.DEFAULT_KERNEL_INITIALIZER="glorotNormal",this.DEFAULT_BIAS_INITIALIZER="zeros",e.verifyArgs(n),this.rank=t,wb(this.rank,"rank"),1!==this.rank&&2!==this.rank&&3!==this.rank)throw new nb(`Convolution layer for rank other than 1, 2, or 3 (${this.rank}) is not implemented yet.`);if(this.kernelSize=Fk(n.kernelSize,t,"kernelSize"),this.strides=Fk(null==n.strides?1:n.strides,t,"strides"),this.padding=null==n.padding?"valid":n.padding,Mb(this.padding),this.dataFormat=null==n.dataFormat?"channelsLast":n.dataFormat,Db(this.dataFormat),this.activation=xk(n.activation),this.useBias=null==n.useBias||n.useBias,this.biasInitializer=Fx(n.biasInitializer||this.DEFAULT_BIAS_INITIALIZER),this.biasConstraint=mv(n.biasConstraint),this.biasRegularizer=Nk(n.biasRegularizer),this.activityRegularizer=Nk(n.activityRegularizer),this.dilationRate=Fk(null==n.dilationRate?1:n.dilationRate,t,"dilationRate"),1===this.rank&&Array.isArray(this.dilationRate)&&1!==this.dilationRate.length)throw new tb(`dilationRate must be a number or an array of a single number for 1D convolution, but received ${JSON.stringify(this.dilationRate)}`);if(2===this.rank){if("number"==typeof this.dilationRate)this.dilationRate=[this.dilationRate,this.dilationRate];else if(2!==this.dilationRate.length)throw new tb(`dilationRate must be a number or array of two numbers for 2D convolution, but received ${JSON.stringify(this.dilationRate)}`)}else if(3===this.rank)if("number"==typeof this.dilationRate)this.dilationRate=[this.dilationRate,this.dilationRate,this.dilationRate];else if(3!==this.dilationRate.length)throw new tb(`dilationRate must be a number or array of three numbers for 3D convolution, but received ${JSON.stringify(this.dilationRate)}`)}static verifyArgs(e){if(ib("kernelSize"in e,"required key 'kernelSize' not in config"),"number"!=typeof e.kernelSize&&!vb(e.kernelSize,"number",1,3))throw new tb(`BaseConv expects config.kernelSize to be number or number[] with length 1, 2, or 3, but received ${JSON.stringify(e.kernelSize)}.`)}getConfig(){let e={kernelSize:this.kernelSize,strides:this.strides,padding:this.padding,dataFormat:this.dataFormat,dilationRate:this.dilationRate,activation:yk(this.activation),useBias:this.useBias,biasInitializer:Rx(this.biasInitializer),biasRegularizer:Sk(this.biasRegularizer),activityRegularizer:Sk(this.activityRegularizer),biasConstraint:cv(this.biasConstraint)},t=super.getConfig();return Object.assign(e,t),e}},Bk=class e extends zk{constructor(t,n){super(t,n),this.kernel=null,e.verifyArgs(n),this.filters=n.filters,wb(this.filters,"filters"),this.kernelInitializer=Fx(n.kernelInitializer||this.DEFAULT_KERNEL_INITIALIZER),this.kernelConstraint=mv(n.kernelConstraint),this.kernelRegularizer=Nk(n.kernelRegularizer)}build(e){e=Lx(e);let t="channelsFirst"===this.dataFormat?1:e.length-1;if(null==e[t])throw new tb(`The channel dimension of the input should be defined. Found ${e[t]}`);let n=e[t],r=this.kernelSize.concat([n,this.filters]);this.kernel=this.addWeight("kernel",r,null,this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint),this.useBias&&(this.bias=this.addWeight("bias",[this.filters],null,this.biasInitializer,this.biasRegularizer,!0,this.biasConstraint)),this.inputSpec=[{ndim:this.rank+2,axes:{[t]:n}}],this.built=!0}call(e,t){return Qs(()=>{e=Ox(e);let t,n=null==this.bias?null:this.bias.read(),r=Ib(this.activation.getClassName());if(null!=r&&2===this.rank)t=Pk(e,this.kernel.read(),n,this.strides,this.padding,this.dataFormat,this.dilationRate,r);else{if(1===this.rank)t=function(e,t,n,r=1,a="valid",s,i=1){return Qs(()=>{if(null==s&&(s="channelsLast"),Db(s),3!==e.shape.length)throw new tb(`The input of a conv1dWithBias operation should be 3, but is ${e.shape.length} instead.`);if(3!==t.shape.length)throw new tb(`The kernel for a conv1dWithBias operation should be 3, but is ${t.shape.length} instead`);if(null!=n&&1!==n.shape.length)throw new tb(`The bias for a conv1dWithBias operation should be 1, but is ${n.shape.length} instead`);if("channelsFirst"===s&&(e=Jd(e,[0,2,1])),"causal"===a)throw new nb("The support for CAUSAL padding mode in conv1dWithBias is not implemented yet.");let o=xl(e,t,r,"same"===a?"same":"valid","NWC",i);return null!=n&&(o=ux(o,n)),o})}(e,this.kernel.read(),n,this.strides[0],this.padding,this.dataFormat,this.dilationRate[0]);else if(2===this.rank)t=Pk(e,this.kernel.read(),n,this.strides,this.padding,this.dataFormat,this.dilationRate);else{if(3!==this.rank)throw new nb("convolutions greater than 3D are not implemented yet.");t=function(e,t,n,r=[1,1,1],a="valid",s,i){return Qs(()=>{if(null==s&&(s="channelsLast"),Db(s),4!==e.rank&&5!==e.rank)throw new tb(`conv3dWithBias expects input to be of rank 4 or 5, but received ${e.rank}.`);if(4!==t.rank&&5!==t.rank)throw new tb(`conv3dWithBias expects kernel to be of rank 4 or 5, but received ${e.rank}.`);let o=Lk(e,s);if("causal"===a)throw new nb("The support for CAUSAL padding mode in conv3dWithBias is not implemented yet.");return o=kl(o,t,r,"same"===a?"same":"valid","NDHWC",i),null!=n&&(o=ux(o,n)),"channelsFirst"===s&&(o=Jd(o,[0,4,1,2,3])),o})}(e,this.kernel.read(),n,this.strides,this.padding,this.dataFormat,this.dilationRate)}null!=this.activation&&(t=this.activation.apply(t))}return t})}computeOutputShape(e){e=Lx(e);let t=[],n="channelsLast"===this.dataFormat?e.slice(1,e.length-1):e.slice(2);for(let e=0;e<n.length;++e){let r=Dk(n[e],this.kernelSize[e],this.padding,this.strides[e],"number"==typeof this.dilationRate?this.dilationRate:this.dilationRate[e]);t.push(r)}let r=[e[0]];return"channelsLast"===this.dataFormat?(r=r.concat(t),r.push(this.filters)):(r.push(this.filters),r=r.concat(t)),r}getConfig(){let e={filters:this.filters,kernelInitializer:Rx(this.kernelInitializer),kernelRegularizer:Sk(this.kernelRegularizer),kernelConstraint:cv(this.kernelConstraint)},t=super.getConfig();return Object.assign(e,t),e}static verifyArgs(e){if(!("filters"in e)||"number"!=typeof e.filters||e.filters<1)throw new tb(`Convolution layer expected config.filters to be a 'number' > 0 but got ${JSON.stringify(e.filters)}`)}},Wk=class e extends Bk{constructor(t){super(2,t),e.verifyArgs(t)}getConfig(){let e=super.getConfig();return delete e.rank,e}static verifyArgs(e){if("number"!=typeof e.kernelSize&&!vb(e.kernelSize,"number",1,2))throw new tb(`Conv2D expects config.kernelSize to be number or number[] with length 1 or 2, but received ${JSON.stringify(e.kernelSize)}.`)}};Wk.className="Conv2D",kc.registerClass(Wk);var Vk=class e extends Bk{constructor(t){super(3,t),e.verifyArgs(t)}getConfig(){let e=super.getConfig();return delete e.rank,e}static verifyArgs(e){if("number"!=typeof e.kernelSize&&(!Array.isArray(e.kernelSize)||1!==e.kernelSize.length&&3!==e.kernelSize.length))throw new tb(`Conv3D expects config.kernelSize to be number or [number, number, number], but received ${JSON.stringify(e.kernelSize)}.`)}};Vk.className="Conv3D",kc.registerClass(Vk);var Uk=class extends Wk{constructor(e){if(super(e),this.inputSpec=[new Ux({ndim:4})],"same"!==this.padding&&"valid"!==this.padding)throw new tb(`Conv2DTranspose currently supports only padding modes 'same' and 'valid', but received padding mode ${this.padding}`)}build(e){if(4!==(e=Lx(e)).length)throw new tb("Input should have rank 4; Received input shape: "+JSON.stringify(e));let t="channelsFirst"===this.dataFormat?1:e.length-1;if(null==e[t])throw new tb("The channel dimension of the inputs should be defined. Found `None`.");let n=e[t],r=this.kernelSize.concat([this.filters,n]);this.kernel=this.addWeight("kernel",r,"float32",this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint),this.useBias&&(this.bias=this.addWeight("bias",[this.filters],"float32",this.biasInitializer,this.biasRegularizer,!0,this.biasConstraint)),this.inputSpec=[new Ux({ndim:4,axes:{[t]:n}})],this.built=!0}call(e,t){return Qs(()=>{let t=Ox(e);if(4!==t.shape.length)throw new tb(`Conv2DTranspose.call() expects input tensor to be rank-4, but received a tensor of rank-${t.shape.length}`);let n,r,a=t.shape,s=a[0];"channelsFirst"===this.dataFormat?(n=2,r=3):(n=1,r=2);let i=a[n],o=a[r],l=this.kernelSize[0],u=this.kernelSize[1],h=this.strides[0],d=this.strides[1],p=[s,Mk(i,h,l,this.padding),Mk(o,d,u,this.padding),this.filters];"channelsLast"!==this.dataFormat&&(t=Jd(t,[0,2,3,1]));let c=wl(t,this.kernel.read(),p,this.strides,this.padding);return"channelsLast"!==this.dataFormat&&(c=Jd(c,[0,3,1,2])),null!=this.bias&&(c=ux(c,this.bias.read(),this.dataFormat)),null!=this.activation&&(c=this.activation.apply(c)),c})}computeOutputShape(e){let t,n,r,a=(e=Lx(e)).slice();"channelsFirst"===this.dataFormat?(t=1,n=2,r=3):(t=3,n=1,r=2);let s=this.kernelSize[0],i=this.kernelSize[1],o=this.strides[0],l=this.strides[1];return a[t]=this.filters,a[n]=Mk(a[n],o,s,this.padding),a[r]=Mk(a[r],l,i,this.padding),a}getConfig(){let e=super.getConfig();return delete e.dilationRate,e}};Uk.className="Conv2DTranspose",kc.registerClass(Uk);var Gk=class extends Vk{constructor(e){if(super(e),this.inputSpec=[new Ux({ndim:5})],"same"!==this.padding&&"valid"!==this.padding)throw new tb(`Conv3DTranspose currently supports only padding modes 'same' and 'valid', but received padding mode ${this.padding}`)}build(e){if(5!==(e=Lx(e)).length)throw new tb("Input should have rank 5; Received input shape: "+JSON.stringify(e));let t="channelsFirst"===this.dataFormat?1:e.length-1;if(null==e[t])throw new tb("The channel dimension of the inputs should be defined. Found `None`.");let n=e[t],r=this.kernelSize.concat([this.filters,n]);this.kernel=this.addWeight("kernel",r,"float32",this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint),this.useBias&&(this.bias=this.addWeight("bias",[this.filters],"float32",this.biasInitializer,this.biasRegularizer,!0,this.biasConstraint)),this.inputSpec=[new Ux({ndim:5,axes:{[t]:n}})],this.built=!0}call(e,t){return Qs(()=>{let t=Ox(e);if(5!==t.shape.length)throw new tb(`Conv3DTranspose.call() expects input tensor to be rank-4, but received a tensor of rank-${t.shape.length}`);let n,r,a,s=t.shape,i=s[0];"channelsFirst"===this.dataFormat?(a=2,n=3,r=4):(a=1,n=2,r=3);let o=s[a],l=s[n],u=s[r],h=this.kernelSize[0],d=this.kernelSize[1],p=this.kernelSize[2],c=this.strides[0],f=this.strides[1],m=this.strides[2],g=[i,Mk(o,c,h,this.padding),Mk(l,f,d,this.padding),Mk(u,m,p,this.padding),this.filters];"channelsLast"!==this.dataFormat&&(t=Jd(t,[0,2,3,4,1]));let y=Sl(t,this.kernel.read(),g,this.strides,this.padding);return"channelsLast"!==this.dataFormat&&(y=Jd(y,[0,4,1,2,3])),null!==this.bias&&(y=ux(y,this.bias.read(),this.dataFormat)),null!==this.activation&&(y=this.activation.apply(y)),y})}computeOutputShape(e){let t,n,r,a,s=(e=Lx(e)).slice();"channelsFirst"===this.dataFormat?(t=1,n=2,r=3,a=4):(t=4,n=1,r=2,a=3);let i=this.kernelSize[0],o=this.kernelSize[1],l=this.kernelSize[2],u=this.strides[0],h=this.strides[1],d=this.strides[2];return s[t]=this.filters,s[n]=Mk(s[n],u,i,this.padding),s[r]=Mk(s[r],h,o,this.padding),s[a]=Mk(s[a],d,l,this.padding),s}getConfig(){let e=super.getConfig();return delete e.dilationRate,e}};Gk.className="Conv3DTranspose",kc.registerClass(Gk);var Hk=class extends Bk{constructor(e,t){if(super(e,t),this.DEFAULT_DEPTHWISE_INITIALIZER="glorotUniform",this.DEFAULT_POINTWISE_INITIALIZER="glorotUniform",this.depthwiseKernel=null,this.pointwiseKernel=null,null==t.filters)throw new tb("The `filters` configuration field is required by SeparableConv, but is unspecified.");if(null!=t.kernelInitializer||null!=t.kernelRegularizer||null!=t.kernelConstraint)throw new tb("Fields kernelInitializer, kernelRegularizer and kernelConstraint are invalid for SeparableConv2D. Use depthwiseInitializer, depthwiseRegularizer, depthwiseConstraint, pointwiseInitializer, pointwiseRegularizer and pointwiseConstraint instead.");if(null!=t.padding&&"same"!==t.padding&&"valid"!==t.padding)throw new tb(`SeparableConv${this.rank}D supports only padding modes: 'same' and 'valid', but received ${JSON.stringify(t.padding)}`);this.depthMultiplier=null==t.depthMultiplier?1:t.depthMultiplier,this.depthwiseInitializer=Fx(t.depthwiseInitializer||this.DEFAULT_DEPTHWISE_INITIALIZER),this.depthwiseRegularizer=Nk(t.depthwiseRegularizer),this.depthwiseConstraint=mv(t.depthwiseConstraint),this.pointwiseInitializer=Fx(t.depthwiseInitializer||this.DEFAULT_POINTWISE_INITIALIZER),this.pointwiseRegularizer=Nk(t.pointwiseRegularizer),this.pointwiseConstraint=mv(t.pointwiseConstraint)}build(e){if((e=Lx(e)).length<this.rank+2)throw new tb(`Inputs to SeparableConv${this.rank}D should have rank ${this.rank+2}, but received input shape: ${JSON.stringify(e)}`);let t="channelsFirst"===this.dataFormat?1:e.length-1;if(null==e[t]||e[t]<0)throw new tb(`The channel dimension of the inputs should be defined, but found ${JSON.stringify(e[t])}`);let n=e[t],r=this.kernelSize.concat([n,this.depthMultiplier]),a=[];for(let e=0;e<this.rank;++e)a.push(1);a.push(n*this.depthMultiplier,this.filters);let s=!0;this.depthwiseKernel=this.addWeight("depthwise_kernel",r,"float32",this.depthwiseInitializer,this.depthwiseRegularizer,s,this.depthwiseConstraint),this.pointwiseKernel=this.addWeight("pointwise_kernel",a,"float32",this.pointwiseInitializer,this.pointwiseRegularizer,s,this.pointwiseConstraint),this.useBias?this.bias=this.addWeight("bias",[this.filters],"float32",this.biasInitializer,this.biasRegularizer,s,this.biasConstraint):this.bias=null,this.inputSpec=[new Ux({ndim:this.rank+2,axes:{[t]:n}})],this.built=!0}call(e,t){return Qs(()=>{let t;if(e=Ox(e),1===this.rank)throw new nb("1D separable convolution is not implemented yet.");return 2===this.rank&&("channelsFirst"===this.dataFormat&&(e=Jd(e,[0,2,3,1])),t=hd(e,this.depthwiseKernel.read(),this.pointwiseKernel.read(),this.strides,this.padding,this.dilationRate,"NHWC")),this.useBias&&(t=ux(t,this.bias.read(),this.dataFormat)),null!=this.activation&&(t=this.activation.apply(t)),"channelsFirst"===this.dataFormat&&(t=Jd(t,[0,3,1,2])),t})}getConfig(){let e=super.getConfig();return delete e.rank,delete e.kernelInitializer,delete e.kernelRegularizer,delete e.kernelConstraint,e.depthwiseInitializer=Rx(this.depthwiseInitializer),e.pointwiseInitializer=Rx(this.pointwiseInitializer),e.depthwiseRegularizer=Sk(this.depthwiseRegularizer),e.pointwiseRegularizer=Sk(this.pointwiseRegularizer),e.depthwiseConstraint=cv(this.depthwiseConstraint),e.pointwiseConstraint=cv(this.pointwiseConstraint),e}};Hk.className="SeparableConv";var jk=class extends Hk{constructor(e){super(2,e)}};jk.className="SeparableConv2D",kc.registerClass(jk);var qk=class e extends Bk{constructor(t){super(1,t),e.verifyArgs(t),this.inputSpec=[{ndim:3}]}getConfig(){let e=super.getConfig();return delete e.rank,delete e.dataFormat,e}static verifyArgs(e){if("number"!=typeof e.kernelSize&&!vb(e.kernelSize,"number",1,1))throw new tb(`Conv1D expects config.kernelSize to be number or number[] with length 1, but received ${JSON.stringify(e.kernelSize)}.`)}};qk.className="Conv1D",kc.registerClass(qk);var Kk=class extends Kx{constructor(e){super(e),"number"==typeof e.cropping?this.cropping=[[e.cropping,e.cropping],[e.cropping,e.cropping]]:"number"==typeof e.cropping[0]?this.cropping=[[e.cropping[0],e.cropping[0]],[e.cropping[1],e.cropping[1]]]:this.cropping=e.cropping,this.dataFormat=void 0===e.dataFormat?"channelsLast":e.dataFormat,this.inputSpec=[{ndim:4}]}computeOutputShape(e){return"channelsFirst"===this.dataFormat?[e[0],e[1],e[2]-this.cropping[0][0]-this.cropping[0][1],e[3]-this.cropping[1][0]-this.cropping[1][1]]:[e[0],e[1]-this.cropping[0][0]-this.cropping[0][1],e[2]-this.cropping[1][0]-this.cropping[1][1],e[3]]}call(e,t){return Qs(()=>{if(e=Ox(e),"channelsLast"===this.dataFormat){let t=ex(e,this.cropping[0][0],e.shape[1]-this.cropping[0][0]-this.cropping[0][1],2);return ex(t,this.cropping[1][0],e.shape[2]-this.cropping[1][1]-this.cropping[1][0],3)}{let t=ex(e,this.cropping[0][0],e.shape[2]-this.cropping[0][0]-this.cropping[0][1],3);return ex(t,this.cropping[1][0],e.shape[3]-this.cropping[1][1]-this.cropping[1][0],4)}})}getConfig(){let e={cropping:this.cropping,dataFormat:this.dataFormat},t=super.getConfig();return Object.assign(e,t),e}};Kk.className="Cropping2D",kc.registerClass(Kk);var Xk=class extends Kx{constructor(e){super(e),this.DEFAULT_SIZE=[2,2],this.inputSpec=[{ndim:4}],this.size=null==e.size?this.DEFAULT_SIZE:e.size,this.dataFormat=null==e.dataFormat?"channelsLast":e.dataFormat,Db(this.dataFormat),this.interpolation=null==e.interpolation?"nearest":e.interpolation,function(e){xb(Eb,"InterpolationFormat",e)}(this.interpolation)}computeOutputShape(e){if("channelsFirst"===this.dataFormat){let t=null==e[2]?null:this.size[0]*e[2],n=null==e[3]?null:this.size[1]*e[3];return[e[0],e[1],t,n]}{let t=null==e[1]?null:this.size[0]*e[1],n=null==e[2]?null:this.size[1]*e[2];return[e[0],t,n,e[3]]}}call(e,t){return Qs(()=>{let t=Ox(e),n=t.shape;if("channelsFirst"===this.dataFormat){t=Jd(t,[0,2,3,1]);let e=this.size[0]*n[2],r=this.size[1]*n[3],a="nearest"===this.interpolation?yc.resizeNearestNeighbor(t,[e,r]):yc.resizeBilinear(t,[e,r]);return Jd(a,[0,3,1,2])}{let e=this.size[0]*n[1],r=this.size[1]*n[2];return"nearest"===this.interpolation?yc.resizeNearestNeighbor(t,[e,r]):yc.resizeBilinear(t,[e,r])}})}getConfig(){let e={size:this.size,dataFormat:this.dataFormat,interpolation:this.interpolation},t=super.getConfig();return Object.assign(e,t),e}};Xk.className="UpSampling2D",kc.registerClass(Xk);var Zk=class extends zk{constructor(e){super(2,e),this.depthwiseKernel=null,this.depthMultiplier=null==e.depthMultiplier?1:e.depthMultiplier,this.depthwiseInitializer=Fx(e.depthwiseInitializer||this.DEFAULT_KERNEL_INITIALIZER),this.depthwiseConstraint=mv(e.depthwiseConstraint),this.depthwiseRegularizer=Nk(e.depthwiseRegularizer)}build(e){if((e=Lx(e)).length<4)throw new tb(`Inputs to DepthwiseConv2D should have rank 4. Received input shape: ${JSON.stringify(e)}.`);let t="channelsFirst"===this.dataFormat?1:3;if(null==e[t]||e[t]<0)throw new tb(`The channel dimension of the inputs to DepthwiseConv2D should be defined, but is not (${e[t]}).`);let n=e[t],r=[this.kernelSize[0],this.kernelSize[1],n,this.depthMultiplier];this.depthwiseKernel=this.addWeight("depthwise_kernel",r,null,this.depthwiseInitializer,this.depthwiseRegularizer,!0,this.depthwiseConstraint),this.useBias?this.bias=this.addWeight("bias",[n*this.depthMultiplier],null,this.biasInitializer,this.biasRegularizer,!0,this.biasConstraint):this.bias=null,this.built=!0}call(e,t){return Qs(()=>{let t=function(e,t,n=[1,1],r="valid",a,s){return Qs(()=>{null==a&&(a="channelsLast"),Db(a);let i=Ok(e,a);if(4!==e.rank)throw new tb(`Input for depthwiseConv2d is required to be 4-D, but is instead ${e.rank}-D`);if(4!==t.rank)throw new tb(`depthwiseKernel is required to be 4-D, but is instead ${t.rank}-D`);return i=$l(i,t,n,"same"===r?"same":"valid","NHWC",s),"channelsFirst"===a&&(i=Jd(i,[0,3,1,2])),i})}(e=Ox(e),this.depthwiseKernel.read(),this.strides,this.padding,this.dataFormat,null);return this.useBias&&(t=ux(t,this.bias.read(),this.dataFormat)),null!=this.activation&&(t=this.activation.apply(t)),t})}computeOutputShape(e){e=Lx(e);let t="channelsFirst"===this.dataFormat?e[2]:e[1],n="channelsFirst"===this.dataFormat?e[3]:e[2],r="channelsFirst"===this.dataFormat?e[1]*this.depthMultiplier:e[3]*this.depthMultiplier,a=Dk(t,this.kernelSize[0],this.padding,this.strides[0]),s=Dk(n,this.kernelSize[1],this.padding,this.strides[1]);return"channelsFirst"===this.dataFormat?[e[0],r,a,s]:[e[0],a,s,r]}getConfig(){let e=super.getConfig();return e.depthMultiplier=this.depthMultiplier,e.depthwiseInitializer=Rx(this.depthwiseInitializer),e.depthwiseRegularizer=Sk(this.depthwiseRegularizer),e.depthwiseConstraint=cv(this.depthwiseRegularizer),e}};function Yk(e,t,n,r){if(Array.isArray(e)){if(null!=t||null!=n)throw new tb("When inputs is an array, neither initialState or constants should be provided");null!=r&&(n=e.slice(e.length-r,e.length),e=e.slice(0,e.length-r)),e.length>1&&(t=e.slice(1,e.length)),e=e[0]}function a(e){return null==e||Array.isArray(e)?e:[e]}return{inputs:e,initialState:t=a(t),constants:n=a(n)}}function Jk(e,t,n,r=!1,a,s,i=!1,o=!1){return Qs(()=>{let l=t.shape.length;if(l<3)throw new tb(`Input should be at least 3D, but is ${l}D.`);let u=[1,0].concat(Kb(2,l));if(t=Jd(t,u),null!=s)throw new nb("The rnn() functoin of the deeplearn.js backend does not support constants yet.");i&&console.warn("Backend rnn(): the unroll = true option is not applicable to the imperative deeplearn.js backend."),null!=a&&((a=ho(ho(a,"bool"),"float32")).rank===l-1&&(a=pu(a,-1)),a=Jd(a,u)),r&&(t=nd(t,0),null!=a&&(a=nd(a,0)));let h,d,p,c=[],f=n,m=t.shape[0],g=jd(t);null!=a&&(d=jd(a));for(let t=0;t<m;++t){let n=g[t],r=Qs(()=>e(n,f));if(null==a)h=r[0],f=r[1];else{let e=Qs(()=>{let e=d[t],n=Wu(ch(e),e);return{output:fo(yo(r[0],e),yo(f[0],n)),newStates:f.map((t,a)=>fo(yo(r[1][a],e),yo(t,n)))}});h=e.output,f=e.newStates}o&&c.push(h)}return o&&(p=Td(c,1)),[h,p,f]})}Zk.className="DepthwiseConv2D",kc.registerClass(Zk);var Qk=class e extends Kx{constructor(e){let t;if(super(e),null==e.cell)throw new tb("cell property is missing for the constructor of RNN.");if(t=Array.isArray(e.cell)?new oI({cells:e.cell}):e.cell,null==t.stateSize)throw new tb("The RNN cell should have an attribute `stateSize` (tuple of integers, one integer per RNN state).");this.cell=t,this.returnSequences=null!=e.returnSequences&&e.returnSequences,this.returnState=null!=e.returnState&&e.returnState,this.goBackwards=null!=e.goBackwards&&e.goBackwards,this._stateful=null!=e.stateful&&e.stateful,this.unroll=null!=e.unroll&&e.unroll,this.supportsMasking=!0,this.inputSpec=[new Ux({ndim:3})],this.stateSpec=null,this.states_=null,this.numConstants=null,this.keptStates=[]}getStates(){if(null==this.states_){return Kb(0,Array.isArray(this.cell.stateSize)?this.cell.stateSize.length:1).map(e=>null)}return this.states_}setStates(e){this.states_=e}computeOutputShape(e){Dx(e)&&(e=e[0]);let t=this.cell.stateSize;Array.isArray(t)||(t=[t]);let n,r=t[0];if(n=this.returnSequences?[e[0],e[1],r]:[e[0],r],this.returnState){let r=[];for(let n of t)r.push([e[0],n]);return[n].concat(r)}return n}computeMask(e,t){return Qs(()=>{Array.isArray(t)&&(t=t[0]);let e=this.returnSequences?t:null;if(this.returnState){let t=this.states.map(e=>null);return[e].concat(t)}return e})}get states(){if(null==this.states_){let e=Array.isArray(this.cell.stateSize)?this.cell.stateSize.length:1,t=[];for(let n=0;n<e;++n)t.push(null);return t}return this.states_}set states(e){this.states_=e}build(e){if(null!=this.numConstants)throw new nb("Constants support is not implemented in RNN yet.");Dx(e)&&(e=e[0]);let t=this.stateful?e[0]:null,n=e.slice(2);this.inputSpec[0]=new Ux({shape:[t,null,...n]});let r,a=[e[0]].concat(e.slice(2));if(this.cell.build(a),r=Array.isArray(this.cell.stateSize)?this.cell.stateSize:[this.cell.stateSize],null!=this.stateSpec){if(!va.arraysEqual(this.stateSpec.map(e=>e.shape[e.shape.length-1]),r))throw new tb(`An initialState was passed that is not compatible with cell.stateSize. Received stateSpec=${this.stateSpec}; However cell.stateSize is ${this.cell.stateSize}`)}else this.stateSpec=r.map(e=>new Ux({shape:[null,e]}));this.stateful&&this.resetStates()}resetStates(e,t=!1){Qs(()=>{if(!this.stateful)throw new Qy("Cannot call resetStates() on an RNN Layer that is not stateful.");let n=this.inputSpec[0].shape[0];if(null==n)throw new tb("If an RNN is stateful, it needs to know its batch size. Specify the batch size of your input tensors: \n- If using a Sequential model, specify the batch size by passing a `batchInputShape` option to your first layer.\n- If using the functional API, specify the batch size by passing a `batchShape` option to your Input layer.");if(null==this.states_)Array.isArray(this.cell.stateSize)?this.states_=this.cell.stateSize.map(e=>nh([n,e])):this.states_=[nh([n,this.cell.stateSize])];else if(null==e)ei(this.states_),null!=this.keptStates&&(ei(this.keptStates),this.keptStates=[]),Array.isArray(this.cell.stateSize)?this.states_=this.cell.stateSize.map(e=>nh([n,e])):this.states_[0]=nh([n,this.cell.stateSize]);else{if(Array.isArray(e)||(e=[e]),e.length!==this.states_.length)throw new tb(`Layer ${this.name} expects ${this.states_.length} state(s), but it received ${e.length} state value(s). Input received: ${e}`);!0===t?this.keptStates.push(this.states_.slice()):ei(this.states_);for(let t=0;t<this.states_.length;++t){let r=e[t],a=Array.isArray(this.cell.stateSize)?this.cell.stateSize[t]:this.cell.stateSize,s=[n,a];if(!va.arraysEqual(r.shape,s))throw new tb(`State ${t} is incompatible with layer ${this.name}: expected shape=${s}, received shape=${r.shape}`);this.states_[t]=r}}this.states_=this.states_.map(e=>ti(e.clone()))})}apply(e,t){let n=null==t?null:t.initialState,r=null==t?null:t.constants;null==t&&(t={});let a=Yk(e,n,r,this.numConstants);e=a.inputs,n=a.initialState,r=a.constants;let s=[],i=[];if(null!=n){t.initialState=n,s=s.concat(n),this.stateSpec=[];for(let e of n)this.stateSpec.push(new Ux({shape:e.shape}));i=i.concat(this.stateSpec)}if(null!=r&&(t.constants=r,s=s.concat(r),this.numConstants=r.length),s[0]instanceof Gx){let n=[e].concat(s),r=this.inputSpec.concat(i),a=this.inputSpec;this.inputSpec=r;let o=super.apply(n,t);return this.inputSpec=a,o}return super.apply(e,t)}call(e,t){return Qs(()=>{let n=null==t?null:t.mask,r=null==t?null:t.training,a=null==t?null:t.initialState;e=Ox(e),null==a&&(a=this.stateful?this.states_:this.getInitialState(e));let s=Array.isArray(this.cell.stateSize)?this.cell.stateSize.length:1;if(a.length!==s)throw new tb(`RNN Layer has ${s} state(s) but was passed ${a.length} initial state(s).`);this.unroll&&console.warn("Ignoring unroll = true for RNN layer, due to imperative backend.");let i={training:r},o=Jk((e,t)=>{let n=this.cell.call([e].concat(t),i);return[n[0],n.slice(1)]},e,a,this.goBackwards,n,null,this.unroll,this.returnSequences),l=o[0],u=o[1],h=o[2];this.stateful&&this.resetStates(h,r);let d=this.returnSequences?u:l;return this.returnState?[d].concat(h):d})}getInitialState(e){return Qs(()=>{let t=nh(e.shape);return t=ou(t,[1,2]),t=Yb(t),Array.isArray(this.cell.stateSize)?this.cell.stateSize.map(e=>e>1?rx(t,[1,e]):t):this.cell.stateSize>1?[rx(t,[1,this.cell.stateSize])]:[t]})}get trainableWeights(){return this.trainable?this.cell.trainableWeights:[]}get nonTrainableWeights(){return this.trainable?this.cell.nonTrainableWeights:this.cell.weights}setFastWeightInitDuringBuild(e){super.setFastWeightInitDuringBuild(e),null!=this.cell&&this.cell.setFastWeightInitDuringBuild(e)}getConfig(){let t=super.getConfig(),n={returnSequences:this.returnSequences,returnState:this.returnState,goBackwards:this.goBackwards,stateful:this.stateful,unroll:this.unroll};null!=this.numConstants&&(n.numConstants=this.numConstants);let r=this.cell.getConfig();return this.getClassName()===e.className&&(n.cell={className:this.cell.getClassName(),config:r}),Object.assign(Object.assign(Object.assign({},r),t),n)}static fromConfig(e,t,n={}){let r=Kv(t.cell,n);return new e(Object.assign(t,{cell:r}))}};Qk.className="RNN",kc.registerClass(Qk);var eI=class extends Kx{},tI=class extends eI{constructor(e){super(e),this.DEFAULT_ACTIVATION="tanh",this.DEFAULT_KERNEL_INITIALIZER="glorotNormal",this.DEFAULT_RECURRENT_INITIALIZER="orthogonal",this.DEFAULT_BIAS_INITIALIZER="zeros",this.units=e.units,wb(this.units,"units"),this.activation=xk(null==e.activation?this.DEFAULT_ACTIVATION:e.activation),this.useBias=null==e.useBias||e.useBias,this.kernelInitializer=Fx(e.kernelInitializer||this.DEFAULT_KERNEL_INITIALIZER),this.recurrentInitializer=Fx(e.recurrentInitializer||this.DEFAULT_RECURRENT_INITIALIZER),this.biasInitializer=Fx(e.biasInitializer||this.DEFAULT_BIAS_INITIALIZER),this.kernelRegularizer=Nk(e.kernelRegularizer),this.recurrentRegularizer=Nk(e.recurrentRegularizer),this.biasRegularizer=Nk(e.biasRegularizer),this.kernelConstraint=mv(e.kernelConstraint),this.recurrentConstraint=mv(e.recurrentConstraint),this.biasConstraint=mv(e.biasConstraint),this.dropout=jb([1,qb([0,null==e.dropout?0:e.dropout])]),this.recurrentDropout=jb([1,qb([0,null==e.recurrentDropout?0:e.recurrentDropout])]),this.dropoutFunc=e.dropoutFunc,this.stateSize=this.units,this.dropoutMask=null,this.recurrentDropoutMask=null}build(e){e=Lx(e),this.kernel=this.addWeight("kernel",[e[e.length-1],this.units],null,this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint),this.recurrentKernel=this.addWeight("recurrent_kernel",[this.units,this.units],null,this.recurrentInitializer,this.recurrentRegularizer,!0,this.recurrentConstraint),this.useBias?this.bias=this.addWeight("bias",[this.units],null,this.biasInitializer,this.biasRegularizer,!0,this.biasConstraint):this.bias=null,this.built=!0}call(e,t){return Qs(()=>{if(2!==e.length)throw new tb(`SimpleRNNCell expects 2 input Tensors, got ${e.length}.`);let n=e[1];e=e[0];let r=null!=t.training&&t.training;0<this.dropout&&this.dropout<1&&null==this.dropoutMask&&(this.dropoutMask=lI({ones:()=>ch(e),rate:this.dropout,training:r,dropoutFunc:this.dropoutFunc})),0<this.recurrentDropout&&this.recurrentDropout<1&&null==this.recurrentDropoutMask&&(this.recurrentDropoutMask=lI({ones:()=>ch(n),rate:this.recurrentDropout,training:r,dropoutFunc:this.dropoutFunc}));let a,s=this.dropoutMask,i=this.recurrentDropoutMask;a=sx(null!=s?yo(e,s):e,this.kernel.read()),null!=this.bias&&(a=ux(a,this.bias.read())),null!=i&&(n=yo(n,i));let o=fo(a,sx(n,this.recurrentKernel.read()));return null!=this.activation&&(o=this.activation.apply(o)),[o,o]})}getConfig(){let e=super.getConfig(),t={units:this.units,activation:yk(this.activation),useBias:this.useBias,kernelInitializer:Rx(this.kernelInitializer),recurrentInitializer:Rx(this.recurrentInitializer),biasInitializer:Rx(this.biasInitializer),kernelRegularizer:Sk(this.kernelRegularizer),recurrentRegularizer:Sk(this.recurrentRegularizer),biasRegularizer:Sk(this.biasRegularizer),activityRegularizer:Sk(this.activityRegularizer),kernelConstraint:cv(this.kernelConstraint),recurrentConstraint:cv(this.recurrentConstraint),biasConstraint:cv(this.biasConstraint),dropout:this.dropout,recurrentDropout:this.recurrentDropout};return Object.assign(Object.assign({},e),t)}};tI.className="SimpleRNNCell",kc.registerClass(tI);var nI=class extends Qk{constructor(e){e.cell=new tI(e),super(e)}call(e,t){return Qs(()=>{null!=this.cell.dropoutMask&&(ei(this.cell.dropoutMask),this.cell.dropoutMask=null),null!=this.cell.recurrentDropoutMask&&(ei(this.cell.recurrentDropoutMask),this.cell.recurrentDropoutMask=null);let n=null==t?null:t.mask,r=null==t?null:t.training,a=null==t?null:t.initialState;return super.call(e,{mask:n,training:r,initialState:a})})}static fromConfig(e,t){return new e(t)}};nI.className="SimpleRNN",kc.registerClass(nI);var rI=class extends eI{constructor(e){if(super(e),this.DEFAULT_ACTIVATION="tanh",this.DEFAULT_RECURRENT_ACTIVATION="hardSigmoid",this.DEFAULT_KERNEL_INITIALIZER="glorotNormal",this.DEFAULT_RECURRENT_INITIALIZER="orthogonal",this.DEFAULT_BIAS_INITIALIZER="zeros",e.resetAfter)throw new tb("GRUCell does not support reset_after parameter set to true.");this.units=e.units,wb(this.units,"units"),this.activation=xk(void 0===e.activation?this.DEFAULT_ACTIVATION:e.activation),this.recurrentActivation=xk(void 0===e.recurrentActivation?this.DEFAULT_RECURRENT_ACTIVATION:e.recurrentActivation),this.useBias=null==e.useBias||e.useBias,this.kernelInitializer=Fx(e.kernelInitializer||this.DEFAULT_KERNEL_INITIALIZER),this.recurrentInitializer=Fx(e.recurrentInitializer||this.DEFAULT_RECURRENT_INITIALIZER),this.biasInitializer=Fx(e.biasInitializer||this.DEFAULT_BIAS_INITIALIZER),this.kernelRegularizer=Nk(e.kernelRegularizer),this.recurrentRegularizer=Nk(e.recurrentRegularizer),this.biasRegularizer=Nk(e.biasRegularizer),this.kernelConstraint=mv(e.kernelConstraint),this.recurrentConstraint=mv(e.recurrentConstraint),this.biasConstraint=mv(e.biasConstraint),this.dropout=jb([1,qb([0,null==e.dropout?0:e.dropout])]),this.recurrentDropout=jb([1,qb([0,null==e.recurrentDropout?0:e.recurrentDropout])]),this.dropoutFunc=e.dropoutFunc,this.implementation=e.implementation,this.stateSize=this.units,this.dropoutMask=null,this.recurrentDropoutMask=null}build(e){let t=(e=Lx(e))[e.length-1];this.kernel=this.addWeight("kernel",[t,3*this.units],null,this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint),this.recurrentKernel=this.addWeight("recurrent_kernel",[this.units,3*this.units],null,this.recurrentInitializer,this.recurrentRegularizer,!0,this.recurrentConstraint),this.useBias?this.bias=this.addWeight("bias",[3*this.units],null,this.biasInitializer,this.biasRegularizer,!0,this.biasConstraint):this.bias=null,this.built=!0}call(e,t){return Qs(()=>{if(2!==e.length)throw new tb(`GRUCell expects 2 input Tensors (inputs, h, c), got ${e.length}.`);let n=null!=t.training&&t.training,r=e[1];e=e[0],0<this.dropout&&this.dropout<1&&null==this.dropoutMask&&(this.dropoutMask=lI({ones:()=>ch(e),rate:this.dropout,training:n,count:3,dropoutFunc:this.dropoutFunc})),0<this.recurrentDropout&&this.recurrentDropout<1&&null==this.recurrentDropoutMask&&(this.recurrentDropoutMask=lI({ones:()=>ch(r),rate:this.recurrentDropout,training:n,count:3,dropoutFunc:this.dropoutFunc}));let a,s,i,o=this.dropoutMask,l=this.recurrentDropoutMask;0<this.dropout&&this.dropout<1&&(e=yo(e,o[0]));let u=sx(e,this.kernel.read());this.useBias&&(u=ux(u,this.bias.read())),0<this.recurrentDropout&&this.recurrentDropout<1&&(r=yo(r,l[0]));let h=this.recurrentKernel.read(),[d,p]=Id(h,[2*this.units,this.units],h.rank-1),c=sx(r,d),[f,m,g]=Id(u,3,u.rank-1),[y,b]=Id(c,2,c.rank-1);a=this.recurrentActivation.apply(fo(f,y)),s=this.recurrentActivation.apply(fo(m,b));let x=sx(yo(s,r),p);i=this.activation.apply(fo(g,x));let v=fo(yo(a,r),yo(fo(1,Pu(a)),i));return[v,v]})}getConfig(){let e=super.getConfig(),t={units:this.units,activation:yk(this.activation),recurrentActivation:yk(this.recurrentActivation),useBias:this.useBias,kernelInitializer:Rx(this.kernelInitializer),recurrentInitializer:Rx(this.recurrentInitializer),biasInitializer:Rx(this.biasInitializer),kernelRegularizer:Sk(this.kernelRegularizer),recurrentRegularizer:Sk(this.recurrentRegularizer),biasRegularizer:Sk(this.biasRegularizer),activityRegularizer:Sk(this.activityRegularizer),kernelConstraint:cv(this.kernelConstraint),recurrentConstraint:cv(this.recurrentConstraint),biasConstraint:cv(this.biasConstraint),dropout:this.dropout,recurrentDropout:this.recurrentDropout,implementation:this.implementation,resetAfter:!1};return Object.assign(Object.assign({},e),t)}};rI.className="GRUCell",kc.registerClass(rI);var aI=class extends Qk{constructor(e){0===e.implementation&&console.warn("`implementation=0` has been deprecated, and now defaults to `implementation=1`. Please update your layer call."),e.cell=new rI(e),super(e)}call(e,t){return Qs(()=>{null!=this.cell.dropoutMask&&(ei(this.cell.dropoutMask),this.cell.dropoutMask=null),null!=this.cell.recurrentDropoutMask&&(ei(this.cell.recurrentDropoutMask),this.cell.recurrentDropoutMask=null);let n=null==t?null:t.mask,r=null==t?null:t.training,a=null==t?null:t.initialState;return super.call(e,{mask:n,training:r,initialState:a})})}static fromConfig(e,t){return 0===t.implmentation&&(t.implementation=1),new e(t)}};aI.className="GRU",kc.registerClass(aI);var sI=class extends eI{constructor(e){super(e),this.DEFAULT_ACTIVATION="tanh",this.DEFAULT_RECURRENT_ACTIVATION="hardSigmoid",this.DEFAULT_KERNEL_INITIALIZER="glorotNormal",this.DEFAULT_RECURRENT_INITIALIZER="orthogonal",this.DEFAULT_BIAS_INITIALIZER="zeros",this.units=e.units,wb(this.units,"units"),this.activation=xk(void 0===e.activation?this.DEFAULT_ACTIVATION:e.activation),this.recurrentActivation=xk(void 0===e.recurrentActivation?this.DEFAULT_RECURRENT_ACTIVATION:e.recurrentActivation),this.useBias=null==e.useBias||e.useBias,this.kernelInitializer=Fx(e.kernelInitializer||this.DEFAULT_KERNEL_INITIALIZER),this.recurrentInitializer=Fx(e.recurrentInitializer||this.DEFAULT_RECURRENT_INITIALIZER),this.biasInitializer=Fx(e.biasInitializer||this.DEFAULT_BIAS_INITIALIZER),this.unitForgetBias=e.unitForgetBias,this.kernelRegularizer=Nk(e.kernelRegularizer),this.recurrentRegularizer=Nk(e.recurrentRegularizer),this.biasRegularizer=Nk(e.biasRegularizer),this.kernelConstraint=mv(e.kernelConstraint),this.recurrentConstraint=mv(e.recurrentConstraint),this.biasConstraint=mv(e.biasConstraint),this.dropout=jb([1,qb([0,null==e.dropout?0:e.dropout])]),this.recurrentDropout=jb([1,qb([0,null==e.recurrentDropout?0:e.recurrentDropout])]),this.dropoutFunc=e.dropoutFunc,this.implementation=e.implementation,this.stateSize=[this.units,this.units],this.dropoutMask=null,this.recurrentDropoutMask=null}build(e){var t;let n,r=(e=Lx(e))[e.length-1];if(this.kernel=this.addWeight("kernel",[r,4*this.units],null,this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint),this.recurrentKernel=this.addWeight("recurrent_kernel",[this.units,4*this.units],null,this.recurrentInitializer,this.recurrentRegularizer,!0,this.recurrentConstraint),this.useBias){if(this.unitForgetBias){let e=this.biasInitializer,r=this.units;n=new((t=class extends fx{apply(t,n){let a=e.apply([r]),s=(new gx).apply([r]),i=e.apply([2*r]);return nx(nx(a,s),i)}}).className="CustomInit",t)}else n=this.biasInitializer;this.bias=this.addWeight("bias",[4*this.units],null,n,this.biasRegularizer,!0,this.biasConstraint)}else this.bias=null;this.built=!0}call(e,t){return Qs(()=>{let n=null!=t.training&&t.training;if(3!==e.length)throw new tb(`LSTMCell expects 3 input Tensors (inputs, h, c), got ${e.length}.`);let r=e[1],a=e[2];e=e[0],0<this.dropout&&this.dropout<1&&null==this.dropoutMask&&(this.dropoutMask=lI({ones:()=>ch(e),rate:this.dropout,training:n,count:4,dropoutFunc:this.dropoutFunc})),0<this.recurrentDropout&&this.recurrentDropout<1&&null==this.recurrentDropoutMask&&(this.recurrentDropoutMask=lI({ones:()=>ch(r),rate:this.recurrentDropout,training:n,count:4,dropoutFunc:this.dropoutFunc}));let s,i,o,l,u=this.dropoutMask,h=this.recurrentDropoutMask;0<this.dropout&&this.dropout<1&&(e=yo(e,u[0]));let d=sx(e,this.kernel.read());0<this.recurrentDropout&&this.recurrentDropout<1&&(r=yo(r,h[0])),d=fo(d,sx(r,this.recurrentKernel.read())),this.useBias&&(d=ux(d,this.bias.read()));let[p,c,f,m]=Id(d,4,d.rank-1);s=this.recurrentActivation.apply(p),i=this.recurrentActivation.apply(c),o=fo(yo(i,a),yo(s,this.activation.apply(f))),l=this.recurrentActivation.apply(m);let g=yo(l,this.activation.apply(o));return[g,g,o]})}getConfig(){let e=super.getConfig(),t={units:this.units,activation:yk(this.activation),recurrentActivation:yk(this.recurrentActivation),useBias:this.useBias,kernelInitializer:Rx(this.kernelInitializer),recurrentInitializer:Rx(this.recurrentInitializer),biasInitializer:Rx(this.biasInitializer),unitForgetBias:this.unitForgetBias,kernelRegularizer:Sk(this.kernelRegularizer),recurrentRegularizer:Sk(this.recurrentRegularizer),biasRegularizer:Sk(this.biasRegularizer),activityRegularizer:Sk(this.activityRegularizer),kernelConstraint:cv(this.kernelConstraint),recurrentConstraint:cv(this.recurrentConstraint),biasConstraint:cv(this.biasConstraint),dropout:this.dropout,recurrentDropout:this.recurrentDropout,implementation:this.implementation};return Object.assign(Object.assign({},e),t)}};sI.className="LSTMCell",kc.registerClass(sI);var iI=class extends Qk{constructor(e){0===e.implementation&&console.warn("`implementation=0` has been deprecated, and now defaults to `implementation=1`. Please update your layer call."),e.cell=new sI(e),super(e)}call(e,t){return Qs(()=>{null!=this.cell.dropoutMask&&(ei(this.cell.dropoutMask),this.cell.dropoutMask=null),null!=this.cell.recurrentDropoutMask&&(ei(this.cell.recurrentDropoutMask),this.cell.recurrentDropoutMask=null);let n=null==t?null:t.mask,r=null==t?null:t.training,a=null==t?null:t.initialState;return super.call(e,{mask:n,training:r,initialState:a})})}static fromConfig(e,t){return 0===t.implmentation&&(t.implementation=1),new e(t)}};iI.className="LSTM",kc.registerClass(iI);var oI=class extends eI{constructor(e){super(e),this.cells=e.cells}get stateSize(){let e=[];for(let t of this.cells.slice().reverse())Array.isArray(t.stateSize)?e.push(...t.stateSize):e.push(t.stateSize);return e}call(e,t){return Qs(()=>{let n=e.slice(1),r=[];for(let e of this.cells.slice().reverse())Array.isArray(e.stateSize)?r.push(n.splice(0,e.stateSize.length)):r.push(n.splice(0,1));r.reverse();let a,s=[];for(let i=0;i<this.cells.length;++i){let o=this.cells[i];n=r[i],a=0===i?[e[0]].concat(n):[a[0]].concat(n),a=o.call(a,t),s.push(a.slice(1))}n=[];for(let e of s.slice().reverse())n.push(...e);return[a[0]].concat(n)})}build(e){let t;Dx(e)&&(e=e[0]),this.cells.forEach((n,r)=>{Pb(`RNNCell_${r}`,()=>{n.build(e),t=Array.isArray(n.stateSize)?n.stateSize[0]:n.stateSize,e=[e[0],t]})}),this.built=!0}getConfig(){let e=super.getConfig(),t={cells:this.cells.map(e=>({className:e.getClassName(),config:e.getConfig()}))};return Object.assign(Object.assign({},e),t)}static fromConfig(e,t,n={}){let r=[];for(let e of t.cells)r.push(Kv(e,n));return new e({cells:r})}get trainableWeights(){if(!this.trainable)return[];let e=[];for(let t of this.cells)e.push(...t.trainableWeights);return e}get nonTrainableWeights(){let e=[];for(let t of this.cells)e.push(...t.nonTrainableWeights);if(!this.trainable){let t=[];for(let e of this.cells)t.push(...e.trainableWeights);return t.concat(e)}return e}getWeights(){let e=[];for(let t of this.cells)e.push(...t.weights);return Wx(e)}setWeights(e){let t=[];for(let n of this.cells){let r=n.weights.length,a=e.splice(r);for(let e=0;e<n.weights.length;++e)t.push([n.weights[e],a[e]])}Vx(t)}};function lI(e){let{ones:t,rate:n,training:r=!1,count:a=1,dropoutFunc:s}=e,i=()=>null!=s?s(t(),n):hx(t(),n),o=()=>dx(i,t,r);return!a||a<=1?ti(o().clone()):Array(a).fill(void 0).map(o).map(e=>ti(e.clone()))}oI.className="StackedRNNCells",kc.registerClass(oI);var uI=class extends Qk{constructor(e){if(e.unroll)throw new nb("Unrolling is not possible with convolutional RNNs.");if(Array.isArray(e.cell))throw new nb("It is not possible at the moment to stack convolutional cells.");super(e),this.inputSpec=[new Ux({ndim:5})]}call(e,t){return Qs(()=>{if(null!=this.cell.dropoutMask&&(ei(this.cell.dropoutMask),this.cell.dropoutMask=null),null!=this.cell.recurrentDropoutMask&&(ei(this.cell.recurrentDropoutMask),this.cell.recurrentDropoutMask=null),t&&t.constants)throw new tb("ConvRNN2D cell does not support constants");let n=null==t?null:t.mask,r=null==t?null:t.training,a=null==t?null:t.initialState;return super.call(e,{mask:n,training:r,initialState:a})})}computeOutputShape(e){let t=this.computeSingleOutputShape(e);return this.returnSequences||(t=[t[0],...t.slice(2)]),this.returnState&&(t=[t,...Array(2).fill([e[0],...t.slice(-3)])]),t}getInitialState(e){return Qs(()=>{let{stateSize:t}=this.cell,n=e.shape,r=this.computeSingleOutputShape(n),a=nh([r[0],...r.slice(2)]);return Array.isArray(t)?Array(t.length).fill(a):[a]})}resetStates(e,t=!1){Qs(()=>{if(!this.stateful)throw new Qy("Cannot call resetStates() on an RNN Layer that is not stateful.");let n=this.inputSpec[0].shape,r=this.computeSingleOutputShape(n),a=[r[0],...r.slice(2)];if(null==n[0])throw new tb("If an RNN is stateful, it needs to know its batch size. Specify the batch size of your input tensors: \n- If using a Sequential model, specify the batch size by passing a `batchInputShape` option to your first layer.\n- If using the functional API, specify the batch size by passing a `batchShape` option to your Input layer.");if(null==this.getStates())Array.isArray(this.cell.stateSize)?this.states_=this.cell.stateSize.map(()=>nh(a)):this.states_=[nh(a)];else if(null==e)ei(this.states_),null!=this.keptStates&&(ei(this.keptStates),this.keptStates=[]),Array.isArray(this.cell.stateSize)?this.states_=this.cell.stateSize.map(()=>nh(a)):this.states_[0]=nh(a);else{if(Array.isArray(e)||(e=[e]),e.length!==this.states_.length)throw new tb(`Layer ${this.name} expects ${this.states_.length} state(s), but it received ${e.length} state value(s). Input received: ${e}`);t?this.keptStates.push(this.states_.slice()):ei(this.states_);for(let t=0;t<this.states_.length;++t){let n=e[t],r=a;if(!va.arraysEqual(n.shape,r))throw new tb(`State ${t} is incompatible with layer ${this.name}: expected shape=${r}, received shape=${n.shape}`);this.states_[t]=n}}this.states_=this.states_.map(e=>ti(e.clone()))})}computeSingleOutputShape(e){let{dataFormat:t,filters:n,kernelSize:r,padding:a,strides:s,dilationRate:i}=this.cell,o="channelsFirst"===t,l=e[o?3:2],u=e[o?4:3],h=Dk(l,r[0],a,s[0],i[0]),d=Dk(u,r[1],a,s[1],i[1]);return[...e.slice(0,2),...o?[n,h,d]:[h,d,n]]}};uI.className="ConvRNN2D";var hI=class extends sI{constructor(e){let{filters:t,kernelSize:n,strides:r,padding:a,dataFormat:s,dilationRate:i}=e;super(Object.assign(Object.assign({},e),{units:t})),this.filters=t,wb(this.filters,"filters"),this.kernelSize=Fk(n,2,"kernelSize"),this.kernelSize.forEach(e=>wb(e,"kernelSize")),this.strides=Fk(r||1,2,"strides"),this.strides.forEach(e=>wb(e,"strides")),this.padding=a||"valid",Mb(this.padding),this.dataFormat=s||"channelsLast",Db(this.dataFormat),this.dilationRate=Fk(i||1,2,"dilationRate"),this.dilationRate.forEach(e=>wb(e,"dilationRate"))}build(e){var t;e=Lx(e);let n="channelsFirst"===this.dataFormat?1:e.length-1;if(null==e[n])throw new tb(`The channel dimension of the input should be defined. Found ${e[n]}`);let r=e[n],a=this.kernelSize.concat([r,4*this.filters]);this.kernel=this.addWeight("kernel",a,null,this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint);let s=this.kernelSize.concat([this.filters,4*this.filters]);if(this.recurrentKernel=this.addWeight("recurrent_kernel",s,null,this.recurrentInitializer,this.recurrentRegularizer,!0,this.recurrentConstraint),this.useBias){let e;if(this.unitForgetBias){let n=this.biasInitializer,r=this.filters;e=new((t=class extends fx{apply(e,t){return tx([n.apply([r]),rh([r]),n.apply([2*r])])}}).className="CustomInit",t)}else e=this.biasInitializer;this.bias=this.addWeight("bias",[4*this.filters],null,e,this.biasRegularizer,!0,this.biasConstraint)}this.built=!0}call(e,t){return Qs(()=>{if(3!==e.length)throw new tb(`ConvLSTM2DCell expects 3 input Tensors (inputs, h, c), got ${e.length}.`);let n=t.training||!1,r=e[0],a=e[1],s=e[2];0<this.dropout&&this.dropout<1&&null==this.dropoutMask&&(this.dropoutMask=lI({ones:()=>ch(r),rate:this.dropout,training:n,count:4,dropoutFunc:this.dropoutFunc}));let i=this.dropoutMask,o=(e,t,n)=>t&&t[n]?yo(t[n],e):e,l=o(r,i,0),u=o(r,i,1),h=o(r,i,2),d=o(r,i,3);0<this.recurrentDropout&&this.recurrentDropout<1&&null==this.recurrentDropoutMask&&(this.recurrentDropoutMask=lI({ones:()=>ch(a),rate:this.recurrentDropout,training:n,count:4,dropoutFunc:this.dropoutFunc}));let p=this.recurrentDropoutMask,c=o(a,p,0),f=o(a,p,1),m=o(a,p,2),g=o(a,p,3),[y,b,x,v]=Id(this.kernel.read(),4,3),[w,k,I,S]=this.useBias?Id(this.bias.read(),4):[null,null,null,null];l=this.inputConv(l,y,w,this.padding),u=this.inputConv(u,b,k,this.padding),h=this.inputConv(h,x,I,this.padding),d=this.inputConv(d,v,S,this.padding);let[_,N,T,C]=Id(this.recurrentKernel.read(),4,3);c=this.recurrentConv(c,_),f=this.recurrentConv(f,N),m=this.recurrentConv(m,T),g=this.recurrentConv(g,C);let E=this.recurrentActivation.apply(fo(l,c)),A=this.recurrentActivation.apply(fo(u,f)),$=fo(yo(A,s),yo(E,this.activation.apply(fo(h,m)))),R=yo(this.recurrentActivation.apply(fo(d,g)),this.activation.apply($));return[R,R,$]})}getConfig(){let e=super.getConfig(),{units:t}=e,n=function(e,t){var n={};for(var r in e)Object.prototype.hasOwnProperty.call(e,r)&&t.indexOf(r)<0&&(n[r]=e[r]);if(null!=e&&"function"==typeof Object.getOwnPropertySymbols){var a=0;for(r=Object.getOwnPropertySymbols(e);a<r.length;a++)t.indexOf(r[a])<0&&Object.prototype.propertyIsEnumerable.call(e,r[a])&&(n[r[a]]=e[r[a]])}return n}(e,["units"]),r={filters:this.filters,kernelSize:this.kernelSize,padding:this.padding,dataFormat:this.dataFormat,dilationRate:this.dilationRate,strides:this.strides};return Object.assign(Object.assign({},n),r)}inputConv(e,t,n,r){let a=bl(e,t,this.strides,r||"valid","channelsFirst"===this.dataFormat?"NCHW":"NHWC",this.dilationRate);return n?ux(a,n,this.dataFormat):a}recurrentConv(e,t){return bl(e,t,1,"same","channelsFirst"===this.dataFormat?"NCHW":"NHWC")}};hI.className="ConvLSTM2DCell",kc.registerClass(hI);var dI=class extends uI{constructor(e){let t=new hI(e);super(Object.assign(Object.assign({},e),{cell:t}))}static fromConfig(e,t){return new e(t)}};dI.className="ConvLSTM2D",kc.registerClass(dI);var pI=class extends Kx{constructor(e){super(e),this.rate=Math.max(Math.min(e.rate,1),0),this.noiseShape=e.noiseShape,this.seed=e.seed,this.supportsMasking=!0}getNoiseShape(e){if(null==this.noiseShape)return this.noiseShape;let t=e.shape,n=[];for(let e=0;e<this.noiseShape.length;++e)n.push(null==this.noiseShape[e]?t[e]:this.noiseShape[e]);return n}call(e,t){return Qs(()=>{this.invokeCallHook(e,t);let n=Ox(e);if(0<this.rate&&this.rate<1){let e=null!=t.training&&t.training,r=this.getNoiseShape(n);return dx(()=>hx(n,this.rate,r,this.seed),()=>n,e)}return e})}getConfig(){let e={rate:this.rate,noiseShape:this.noiseShape,seed:this.seed},t=super.getConfig();return Object.assign(e,t),e}dispose(){return super.dispose()}};pI.className="Dropout",kc.registerClass(pI);var cI=class extends pI{constructor(e){super(e),this.inputSpec=[{ndim:3}]}getNoiseShape(e){let t=e.shape;return[t[0],1,t[2]]}};cI.className="SpatialDropout1D",kc.registerClass(cI);var fI=class extends Kx{constructor(e){if(super(e),this.activation=null,this.useBias=!0,this.kernel=null,this.bias=null,this.DEFAULT_KERNEL_INITIALIZER="glorotNormal",this.DEFAULT_BIAS_INITIALIZER="zeros",null==e.batchInputShape&&null==e.inputShape&&null!=e.inputDim){let t=null;null!=e.batchSize&&(t=e.batchSize),this.batchInputShape=[t,e.inputDim]}this.units=e.units,wb(this.units,"units"),this.activation=xk(e.activation),null!=e.useBias&&(this.useBias=e.useBias),this.kernelInitializer=Fx(e.kernelInitializer||this.DEFAULT_KERNEL_INITIALIZER),this.biasInitializer=Fx(e.biasInitializer||this.DEFAULT_BIAS_INITIALIZER),this.kernelConstraint=mv(e.kernelConstraint),this.biasConstraint=mv(e.biasConstraint),this.kernelRegularizer=Nk(e.kernelRegularizer),this.biasRegularizer=Nk(e.biasRegularizer),this.activityRegularizer=Nk(e.activityRegularizer),this.supportsMasking=!0,this.inputSpec=[{minNDim:2}]}build(e){let t=(e=Lx(e))[e.length-1];null==this.kernel&&(this.kernel=this.addWeight("kernel",[t,this.units],null,this.kernelInitializer,this.kernelRegularizer,!0,this.kernelConstraint),this.useBias&&(this.bias=this.addWeight("bias",[this.units],null,this.biasInitializer,this.biasRegularizer,!0,this.biasConstraint))),this.inputSpec=[{minNDim:2,axes:{[-1]:t}}],this.built=!0}computeOutputShape(e){let t=(e=Lx(e)).slice();return t[t.length-1]=this.units,t}call(e,t){return Qs(()=>{this.invokeCallHook(e,t);let n,r=Ox(e),a=Ib(this.activation.getClassName());return null!=a?n=sx(r,this.kernel.read(),a,this.bias?this.bias.read():null):(n=sx(r,this.kernel.read()),null!=this.bias&&(n=ux(n,this.bias.read())),null!=this.activation&&(n=this.activation.apply(n))),n})}getConfig(){let e={units:this.units,activation:yk(this.activation),useBias:this.useBias,kernelInitializer:Rx(this.kernelInitializer),biasInitializer:Rx(this.biasInitializer),kernelRegularizer:Sk(this.kernelRegularizer),biasRegularizer:Sk(this.biasRegularizer),activityRegularizer:Sk(this.activityRegularizer),kernelConstraint:cv(this.kernelConstraint),biasConstraint:cv(this.biasConstraint)},t=super.getConfig();return Object.assign(e,t),e}};fI.className="Dense",kc.registerClass(fI);var mI=class extends Kx{constructor(e){super(e=e||{}),this.inputSpec=[{minNDim:3}],this.dataFormat=e.dataFormat}computeOutputShape(e){e=Lx(e);for(let t of e.slice(1))if(null==t)throw new tb(`The shape of the input to "Flatten" is not fully defined (got ${e.slice(1)}). Make sure to pass a complete "input_shape" or "batch_input_shape" argument to the first layer in your model.`);return[e[0],Hb(e,1)]}call(e,t){return Qs(()=>{this.invokeCallHook(e,t);let n=Ox(e);if("channelsFirst"===this.dataFormat&&n.rank>1){let e=[0];for(let t=2;t<n.rank;++t)e.push(t);e.push(1),n=Jd(n,e)}return function(e){if(e.rank<=1)throw new tb(`batchFlatten requires a minimum rank of 2. Got rank: ${e.rank}.`);let t=[e.shape[0],Hb(e.shape,1)];return jo(e,t)}(n)})}getConfig(){let e={};null!=this.dataFormat&&(e.dataFormat=this.dataFormat);let t=super.getConfig();return Object.assign(e,t),e}};mI.className="Flatten",kc.registerClass(mI);var gI=class extends Kx{constructor(e){super(e),this.supportsMasking=!0,this.activation=xk(e.activation)}call(e,t){return Qs(()=>{this.invokeCallHook(e,t);let n=Ox(e);return this.activation.apply(n)})}getConfig(){let e={activation:yk(this.activation)},t=super.getConfig();return Object.assign(e,t),e}};gI.className="Activation",kc.registerClass(gI);var yI=class extends Kx{constructor(e){super(e),this.n=e.n,this.inputSpec=[{ndim:2}]}computeOutputShape(e){return[e[0],this.n,e[1]]}call(e,t){return Qs(()=>function(e,t){return Qs(()=>{if(2!==e.shape.length)throw new tb(`repeat() expects a rank-2 tensor, but received a rank-${e.shape.length} tensor.`);return rx(Yb(e,1),[1,t,1])})}(e=Ox(e),this.n))}getConfig(){let e={n:this.n},t=super.getConfig();return Object.assign(e,t),e}};yI.className="RepeatVector",kc.registerClass(yI);var bI=class extends Kx{constructor(e){super(e),this.targetShape=e.targetShape;for(let e=0;e<this.targetShape.length;++e)this.isUnknown(this.targetShape[e])&&(this.targetShape[e]=null)}isUnknown(e){return e<0||null==e}fixUnknownDimension(e,t){let n="Total size of new array must be unchanged.",r=t.slice(),a=1,s=null;for(let e=0;e<r.length;++e){let t=r[e];if(this.isUnknown(t)){if(null!==s)throw new tb("Can only specifiy one unknown dimension.");s=e}else a*=t}let i=Hb(e);if(null!==s){if(0===a||i%a!==0)throw new tb(n);r[s]=i/a}else if(i!==a)throw new tb(n);return r}computeOutputShape(e){let t=!1;for(let n=0;n<e.length;++n)if(this.isUnknown(e[n])){t=!0;break}return t?e.slice(0,1).concat(this.targetShape):e.slice(0,1).concat(this.fixUnknownDimension(e.slice(1),this.targetShape))}call(e,t){return Qs(()=>{this.invokeCallHook(e,t);let n=Ox(e),r=n.shape,a=r.slice(0,1).concat(this.fixUnknownDimension(r.slice(1),this.targetShape));return jo(n,a)})}getConfig(){let e={targetShape:this.targetShape},t=super.getConfig();return Object.assign(e,t),e}};bI.className="Reshape",kc.registerClass(bI);var xI=class extends Kx{constructor(e){if(super(e),null==e.dims)throw new Error("Required configuration field `dims` is missing during Permute constructor call.");if(!Array.isArray(e.dims))throw new Error(`Permute constructor requires \`dims\` to be an Array, but received ${e.dims} instead.`);let t=Kb(1,e.dims.length+1);if(!va.arraysEqual(e.dims.slice().sort(),t))throw new Error("Invalid permutation `dims`: "+JSON.stringify(e.dims)+" `dims` must contain consecutive integers starting from 1.");this.dims=e.dims,this.dimsIncludingBatch=[0].concat(this.dims),this.inputSpec=[new Ux({ndim:this.dims.length+1})]}computeOutputShape(e){let t=(e=Lx(e)).slice();return this.dims.forEach((n,r)=>{t[r+1]=e[n]}),t}call(e,t){return Jd(Ox(e),this.dimsIncludingBatch)}getConfig(){let e={dims:this.dims},t=super.getConfig();return Object.assign(e,t),e}};xI.className="Permute",kc.registerClass(xI);var vI=class extends Kx{constructor(e){super(e??{}),this.supportsMasking=!0,this.maskValue=null!=e?null==e.maskValue?0:e.maskValue:0}computeOutputShape(e){return e}getConfig(){let e=super.getConfig(),t={maskValue:this.maskValue};return Object.assign(t,e),t}computeMask(e,t){let n=Ox(e);return Io(dh(n,this.maskValue),-1)}call(e,t){return Qs(()=>{this.invokeCallHook(e,t);let n=Ox(e),r=Io(dh(n,this.maskValue),-1,!0);return yo(n,ho(r,n.dtype))})}};vI.className="Masking",kc.registerClass(vI);var wI=class extends Kx{constructor(e){if(super(e),this.embeddings=null,this.DEFAULT_EMBEDDINGS_INITIALIZER="randomUniform",null==e.batchInputShape&&null==e.inputShape){let t=null;null!=e.batchSize&&(t=e.batchSize),null==e.inputLength?this.batchInputShape=[t,null]:this.batchInputShape=[t].concat(ub(e.inputLength))}this.inputDim=e.inputDim,wb(this.inputDim,"inputDim"),this.outputDim=e.outputDim,wb(this.outputDim,"outputDim"),this.embeddingsInitializer=Fx(e.embeddingsInitializer||this.DEFAULT_EMBEDDINGS_INITIALIZER),this.embeddingsRegularizer=Nk(e.embeddingsRegularizer),this.activityRegularizer=Nk(e.activityRegularizer),this.embeddingsConstraint=mv(e.embeddingsConstraint),this.maskZero=e.maskZero,this.supportsMasking=e.maskZero,this.inputLength=e.inputLength}build(e){this.embeddings=this.addWeight("embeddings",[this.inputDim,this.outputDim],this.dtype,this.embeddingsInitializer,this.embeddingsRegularizer,!0,this.embeddingsConstraint),this.built=!0}warnOnIncompatibleInputShape(e){}computeMask(e,t){return Qs(()=>this.maskZero?(e=Ox(e),dh(e,Bl(e))):null)}computeOutputShape(e){if(e=Lx(e),null==this.inputLength)return[...e,this.outputDim];let t=ub(this.inputLength);if(t.length!==e.length-1)throw new tb(`"inputLength" is ${this.inputLength}, but received input shape has shape ${e}`);{let n=0;for(let r=0;r<t.length;++r){let a=t[r],s=e[r+1];if(null!=a&&null!=s&&a!==s)throw new tb(`"inputLength" is ${this.inputLength}, but received input shape has shape ${e}`);null==a&&(t[n]=s),n++}}return[e[0],...t,this.outputDim]}call(e,t){return Qs(()=>{this.invokeCallHook(e,t);let n=Ox(e);"int32"!==n.dtype&&(n=Zb(n,"int32"));let r=ix(this.embeddings.read(),jo(n,[n.size]));return jo(r,Lx(this.computeOutputShape(n.shape)))})}getConfig(){let e={inputDim:this.inputDim,outputDim:this.outputDim,embeddingsInitializer:Rx(this.embeddingsInitializer),embeddingsRegularizer:Sk(this.embeddingsRegularizer),activityRegularizer:Sk(this.activityRegularizer),embeddingsConstraint:cv(this.embeddingsConstraint),maskZero:this.maskZero,inputLength:this.inputLength},t=super.getConfig();return Object.assign(e,t),e}};wI.className="Embedding",kc.registerClass(wI);var kI=class extends Kx{constructor(e){super(e||{}),this.supportsMasking=!0}mergeFunction(e){throw new nb}computeElementwiseOpOutputShape(e,t){if(null==e||null==t)return null;if(e.length<t.length)return this.computeElementwiseOpOutputShape(t,e);if(0===t.length)return e;let n=e.slice(0,e.length-t.length);for(let r=0;r<t.length;++r){let a=e[e.length-t.length+r],s=t[r];if(null==a||null==s||a<0||s<0)n.push(null);else if(1===a)n.push(s);else if(1===s)n.push(a);else{if(a!==s)throw new tb("Operands could not be broadcast together with shapes "+JSON.stringify(e)+" "+JSON.stringify(t));n.push(a)}}return n}build(e){if(Array.isArray(e)&&!Array.isArray(e[0])&&(e=[Lx(e)]),e.length<2)throw new tb(`A merge layer should be called on an Array of at least 2 inputs. Got ${e.length} input(s).`);let t=[];for(let n of e)null!=n&&null!==n[0]&&t.push(n[0]);if(t=yb(t),t.length>1)throw new tb(`Can not merge tensors with different batch sizes. Got tensors with shapes: ${JSON.stringify(e)}.`);let n=null==e[0]?null:e[0].slice(1);for(let t=1;t<e.length;++t){let r=null==e[t]?null:e[t].slice(1);n=this.computeElementwiseOpOutputShape(n,r)}let r=e.map(e=>e.length);-1===e.indexOf(null)&&1===yb(r).length?this.reshapeRequired=!1:this.reshapeRequired=!0}call(e,t){return Qs(()=>{if(this.reshapeRequired){let t=[],n=e.map(e=>e.rank);if(-1===n.indexOf(null)){let r=qb(n);for(let n of e){let e=n.rank;for(let t=0;t<r-e;++t)n=Yb(n,1);t.push(n)}return this.mergeFunction(t)}{let n=!1;for(let r of e){let e=r.rank;if(null==e){let e=r.shape,a=e[0],s=e.slice(1).concat([a]),i=jo(r,[a].concat(Hb(e.slice(1))));i=Jd(i,[1,0]),i=jo(i,s),t.push(i),n=!0}else if(e>1){let a=Kb(1,e).concat([0]);t.push(Jd(r,a)),n=!0}else t.push(r)}let r=this.mergeFunction(t),a=r.rank;if(n)if(null==a){let e=r.shape,t=e[e.length-1],n=[t].concat(e.slice(0,e.length-1));r=jo(Jd(jo(r,[-1,t]),[1,0]),n)}else if(a>1){let e=[a-1].concat(Kb(0,a-1));r=Jd(r,e)}return r}}return this.mergeFunction(e)})}computeOutputShape(e){let t;t=null==e[0]?null:e[0].slice(1);for(let n=1;n<e.length;++n){let r=null==e[n]?null:e[n].slice(1);t=this.computeElementwiseOpOutputShape(t,r)}let n=[];for(let t of e)null!=t&&null!==t[0]&&n.push(t[0]);return n=yb(n),t=1===n.length?n.concat(t):[null].concat(t),t}computeMask(e,t){return Qs(()=>{if(null==t)return null;if(!Array.isArray(t))throw new tb("`mask` should be an Array");if(!Array.isArray(e))throw new tb("`inputs` should be an Array");if(t.length!==e.length)throw new tb(`The Array 'inputs' and 'mask' are expected to have the same length, but have different lengths (${e.length} vs ${t.length})`);if(t.every(e=>null==e))return null;let n=(t=t.map(e=>null==e?e:pu(e,0)))[0];for(let e=1;e<t.length-1;++e)n=Gu(n,t[e]);return n})}},II=class extends kI{constructor(e){super(e)}mergeFunction(e){return Qs(()=>{let t=e[0].clone();for(let n=1;n<e.length;++n)t=fo(t,e[n]);return t})}};II.className="Add",kc.registerClass(II);var SI=class extends kI{constructor(e){super(e)}mergeFunction(e){return Qs(()=>{let t=e[0].clone();for(let n=1;n<e.length;++n)t=yo(t,e[n]);return t})}};SI.className="Multiply",kc.registerClass(SI);var _I=class extends kI{constructor(e){super(e)}mergeFunction(e){return Qs(()=>{let t=e[0].clone();for(let n=1;n<e.length;++n)t=fo(t,e[n]);return yo(1/e.length,t)})}};_I.className="Average",kc.registerClass(_I);var NI=class extends kI{constructor(e){super(e)}mergeFunction(e){return Qs(()=>{let t=e[0];for(let n=1;n<e.length;++n)t=eh(t,e[n]);return t})}};NI.className="Maximum",kc.registerClass(NI);var TI=class extends kI{constructor(e){super(e)}mergeFunction(e){return Qs(()=>{let t=e[0];for(let n=1;n<e.length;++n)t=sh(t,e[n]);return t})}};TI.className="Minimum",kc.registerClass(TI);var CI=class extends kI{constructor(e){super(e),this.DEFAULT_AXIS=-1,null==e&&(e={}),this.axis=null==e.axis?this.DEFAULT_AXIS:e.axis,this.supportsMasking=!0,this.reshapeRequired=!1}build(e){if(!Array.isArray(e)||!Array.isArray(e[0])||1===e.length)throw new tb("A `Concatenate` layer should be called on a list of at least 2 inputs");let t=!0;for(let n of e)if(null!=n){t=!1;break}if(t)return;let n=[];for(let t=0;t<e.length;++t){let r=e[t].slice();r.splice(this.axis,1);let a=!1;for(let e of n)if(va.arraysEqual(e,r)){a=!0;break}a||n.push(r)}if(n.length>1)throw new tb("A `Concatenate` layer requires inputs with matching shapes except for the concat axis. Got input shapes: "+JSON.stringify(e))}mergeFunction(e){return Qs(()=>tx(e,this.axis))}computeOutputShape(e){if(!Array.isArray(e)||!Array.isArray(e[0]))throw new tb("A `Concatenate` layer should be called on a list of inputs.");let t=e,n=t[0].slice(),r=this.axis<0?n.length+this.axis:this.axis;for(let e of t.slice(1)){if(null==n[r]||null==e[r]){n[r]=null;break}n[r]+=e[r]}return n}computeMask(e,t){if(null==t)return null;if(!Array.isArray(t))throw new tb("`mask` should be an array for Concatenate");if(!Array.isArray(e))throw new tb("`inputs` should be an array for Concatenate");if(t.length!==e.length)throw new tb(`Mismatch in the length of mask (${t.length}) and the legnth of inputs (${e.length})`);return Qs(()=>{let n=!0;if(t.forEach(e=>{null==e||(n=!1)}),n)return null;let r=[];for(let n=0;n<e.length;++n)null==t[n]?r.push(ho(ch(e[n]),"bool")):t[n].rank<e[n].rank?r.push(pu(t[n],-1)):r.push(t[n]);let a=Xo(r,this.axis);return ko(a,-1,!1)})}getConfig(){let e={axis:this.axis},t=super.getConfig();return Object.assign(e,t),e}};function EI(e,t){for(;e<0;)e+=t;return e}CI.className="Concatenate",kc.registerClass(CI);var AI=class extends kI{constructor(e){super(e),this.axes=e.axes,this.normalize=null!=e.normalize&&e.normalize,this.supportsMasking=!0,this.reshapeRequired=!1}build(e){va.assert(Array.isArray(e)&&2===e.length&&Array.isArray(e[0])&&Array.isArray(e[1]),()=>"A `Dot` layer should be called on a list of exactly 2 inputs.");let t=e[0],n=e[1];if(t.length>3||n.length>3)throw new nb("Dot layer does not support tensors of 4D or higher rank yet.");let r=this.interpretAxes(t,n);if(t[r[0]]!==n[r[1]])throw new tb(`Dimension incompatibility: ${t[r[0]]} !== ${n[r[1]]}`)}mergeFunction(e){if(2!==e.length)throw new tb(`A \`Dot\` layer must be called on exactly 2 inputs, but received ${e.length} input(s).`);let t,n=e[0],r=e[1];return t=Array.isArray(this.axes)?this.axes.map((t,n)=>EI(t,e[n].shape.length)):[EI(this.axes,n.shape.length),EI(this.axes,r.shape.length)],this.normalize&&(n=Xv(n,t[0]),r=Xv(r,t[1])),function(e,t,n){if(e.shape.length>3||t.shape.length>3)throw new nb("batchDot is not implemented for tensors of 4D or higher rank yet");if(va.assert(e.shape.length>=2,()=>`batchDot requires the rank of x to be >= 2, but got ${e.shape.length}`),va.assert(e.shape.length>=2,()=>`batchDot requires the rank of y to be >= 2, but got ${t.shape.length}`),"number"==typeof n&&(n=[n,n]),"complex64"===e.dtype||"complex64"===t.dtype)throw new nb("batchDot is not implemented for complex64-type Tensors yet.");let r=e.shape.length,a=t.shape.length;null==n&&(n=[r-1,a-2]);let s=n;return Qs(()=>{let n,i;if(r>a){n=r-a;let e=[];for(let t=0;t<n;++t)e.push(1);t=jo(t,t.shape.concat(e))}else if(a>r){n=a-r;let t=[];for(let e=0;e<n;++e)t.push(1);e=jo(e,e.shape.concat(t))}else n=0;if(2===e.shape.length&&2===t.shape.length)i=s[0]===s[1]?ou(yo(e,t),s[0]):ou(yo(Jd(e,[1,0]),t),s[1]);else{let n=s[0]!==e.shape.length-1,r=s[1]===t.shape.length-1;i=Zo(e,t,n,r)}if(n>0){let e;e=r>a?r+a-3:r-1;let t=[];for(let r=e;r<e+n;++r)t.push(r);i=Nd(i,t)}return 1===i.shape.length&&(i=pu(i,1)),i})}(n,r,t)}interpretAxes(e,t){let n;return n=Array.isArray(this.axes)?this.axes:[EI(this.axes,e.length),EI(this.axes,t.length)],n}computeOutputShape(e){va.assert(Array.isArray(e)&&2===e.length&&Array.isArray(e[0])&&Array.isArray(e[1]),()=>"A `Dot` layer should be called on a list of exactly 2 inputs.");let t=e[0].slice(),n=e[1].slice();if(t.length>3||n.length>3)throw new nb("Dot layer does not support tensors of 4D or higher rank yet.");let r=this.interpretAxes(t,n);t.splice(r[0],1),n.splice(r[1],1),n.splice(0,1);let a=t.concat(n);return 1===a.length&&a.push(1),a}computeMask(e,t){return null}getConfig(){let e={axes:this.axes,normalize:this.normalize},t=super.getConfig();return Object.assign(e,t),e}};AI.className="Dot",kc.registerClass(AI);var $I=class extends Kx{constructor(e){super(e),this.supportsMasking=!0,this.stddev=e.stddev}computeOutputShape(e){return e}getConfig(){let e=super.getConfig(),t={stddev:this.stddev};return Object.assign(t,e),t}call(e,t){return Qs(()=>{this.invokeCallHook(e,t);let n=Ox(e);return dx(()=>fo(ax(n.shape,0,this.stddev),n),()=>n,t.training||!1)})}};$I.className="GaussianNoise",kc.registerClass($I);var RI=class extends Kx{constructor(e){super(e),this.supportsMasking=!0,this.rate=e.rate}computeOutputShape(e){return e}getConfig(){let e=super.getConfig(),t={rate:this.rate};return Object.assign(t,e),t}call(e,t){return Qs(()=>{this.invokeCallHook(e,t);let n=Ox(e);return this.rate>0&&this.rate<1?dx(()=>{let e=Math.sqrt(this.rate/(1-this.rate));return yo(n,ax(n.shape,1,e))},()=>n,t.training||!1):n})}};RI.className="GaussianDropout",kc.registerClass(RI);var FI=class extends Kx{constructor(e){super(e),this.supportsMasking=!0,this.rate=e.rate,this.noiseShape=e.noiseShape}_getNoiseShape(e){return this.noiseShape||Ox(e).shape}computeOutputShape(e){return e}getConfig(){let e=super.getConfig(),t={rate:this.rate};return Object.assign(t,e),t}call(e,t){return Qs(()=>{if(this.rate<1&&this.rate>0){let n=this._getNoiseShape(e);return dx(()=>{let t=Ox(e),r=-1.7580993408473766,a=xu(Xh(n),this.rate);a=Zb(a,"float32");let s=((1-this.rate)*(1+this.rate*r**2))**-.5,i=-s*r*this.rate,o=fo(yo(t,a),yo(fo(a,-1),r));return fo(yo(o,s),i)},()=>Ox(e),t.training||!1)}return e})}};function DI(e,t,n,r,a,s=.001){let i;if(2===e.rank)i=al(e,t,n,r,a,s);else if(3===e.rank)i=sl(e,t,n,r,a,s);else{if(4!==e.rank)throw new nb(`batchNormalization is not implemented for array of rank ${e.rank} yet`);i=il(e,t,n,r,a,s)}return i}function MI(e,t,n,r,a=.001){return va.arraysEqual(r.slice().sort(),Kb(0,e.rank-1))?function(e,t,n,r,a=.001){return Qs(()=>{let s=lh(e,r),i=s.mean,o=s.variance;return[DI(e,i,o,n,t,a),i,o]})}(e,t,n,r,a):function(e,t,n,r,a=.001){return Qs(()=>{let s=lh(e,r),i=s.mean,o=s.variance,l=[];for(let t of Kb(0,e.rank))-1!==r.indexOf(t)?l.push(1):l.push(e.shape[t]);let u=jo(i,l),h=jo(o,l),d=null==t?null:jo(t,l),p=null==n?null:jo(n,l);return[DI(e,u,h,p,d,a),i,o]})}(e,t,n,r,a)}FI.className="AlphaDropout",kc.registerClass(FI);var OI=class extends Kx{constructor(e){null==e&&(e={}),super(e),this.supportsMasking=!0,this.axis=null==e.axis?-1:e.axis,this.momentum=null==e.momentum?.99:e.momentum,this.epsilon=null==e.epsilon?.001:e.epsilon,this.center=null==e.center||e.center,this.scale=null==e.scale||e.scale,this.betaInitializer=Fx(e.betaInitializer||"zeros"),this.gammaInitializer=Fx(e.gammaInitializer||"ones"),this.movingMeanInitializer=Fx(e.movingMeanInitializer||"zeros"),this.movingVarianceInitializer=Fx(e.movingVarianceInitializer||"ones"),this.betaConstraint=mv(e.betaConstraint),this.gammaConstraint=mv(e.gammaConstraint),this.betaRegularizer=Nk(e.betaRegularizer),this.gammaRegularizer=Nk(e.gammaRegularizer)}build(e){e=Lx(e);let t=this.axis>=0?this.axis:this.axis+e.length,n=e[t];if(null==n)throw new tb(`Axis ${t} of input tensor should have a defined dimension but the layer received an input with shape ${JSON.stringify(e)}.`);this.inputSpec=[new Ux({ndim:e.length,axes:{[t]:n}})];let r=[n];this.scale&&(this.gamma=this.addWeight("gamma",r,null,this.gammaInitializer,this.gammaRegularizer,!0,this.gammaConstraint)),this.center&&(this.beta=this.addWeight("beta",r,null,this.betaInitializer,this.betaRegularizer,!0,this.betaConstraint)),this.movingMean=this.addWeight("moving_mean",r,null,this.movingMeanInitializer,null,!1),this.movingVariance=this.addWeight("moving_variance",r,null,this.movingVarianceInitializer,null,!1),this.built=!0}call(e,t){return Qs(()=>{let n=null!=t.training&&t.training,r=Ox(e),a=r.shape,s=a.length,i=Kb(0,s),o=this.axis>=0?this.axis:this.axis+s;i.splice(o,1);let l=sb(1,s);l[o]=a[o];let u=i.slice();u.sort();let h=!va.arraysEqual(u,Kb(0,s).slice(0,s-1));if(!n)return(()=>{if(h){let e=jo(this.movingMean.read(),l),t=jo(this.movingVariance.read(),l),n=this.center?jo(this.beta.read(),l):null,a=this.scale?jo(this.gamma.read(),l):null;return DI(r,e,t,n,a,this.epsilon)}return DI(r,this.movingMean.read(),this.movingVariance.read(),null==this.beta?null:this.beta.read(),null==this.gamma?null:this.gamma.read(),this.epsilon)})();let[d,p,c]=MI(r,this.gamma.read(),this.beta.read(),i,this.epsilon),f=(e,t,n)=>{Qs(()=>{let r=1-n,a=e.read(),s=yo(Wu(a,t),r);e.write(Wu(a,s))})};return f(this.movingMean,p,this.momentum),f(this.movingVariance,c,this.momentum),d})}getConfig(){let e={axis:this.axis,momentum:this.momentum,epsilon:this.epsilon,center:this.center,scale:this.scale,betaInitializer:Rx(this.betaInitializer),gammaInitializer:Rx(this.gammaInitializer),movingMeanInitializer:Rx(this.movingMeanInitializer),movingVarianceInitializer:Rx(this.movingVarianceInitializer),betaRegularizer:Sk(this.betaRegularizer),gammaRegularizer:Sk(this.gammaRegularizer),betaConstraint:cv(this.betaConstraint),gammaConstraint:cv(this.gammaConstraint)},t=super.getConfig();return Object.assign(e,t),e}};OI.className="BatchNormalization",kc.registerClass(OI);var LI=class extends Kx{constructor(e){if(null==e&&(e={}),super(e),this.axis=null==e.axis?-1:e.axis,"number"==typeof this.axis){if(!Number.isInteger(this.axis))throw new Error(`Expected axis to be an integer, but received ${this.axis}`)}else{if(!Array.isArray(this.axis))throw new Error(`Expected axis to be an integer or an array of integers, but received ${JSON.stringify(this.axis)}`);for(let e of this.axis)if(!Number.isInteger(e))throw new Error(`Expected axis to be an array of integers, but received ${JSON.stringify(this.axis)}`)}this.epsilon=null==e.epsilon?.001:e.epsilon,this.center=null==e.center||e.center,this.scale=null==e.scale||e.scale,this.betaInitializer=Fx(e.betaInitializer||"zeros"),this.gammaInitializer=Fx(e.gammaInitializer||"ones"),this.betaRegularizer=Nk(e.betaRegularizer),this.gammaRegularizer=Nk(e.gammaRegularizer),this.supportsMasking=!0}build(e){let t=(e=Lx(e)).length;"number"==typeof this.axis&&(this.axis=[this.axis]);for(let e=0;e<this.axis.length;++e)this.axis[e]<0&&(this.axis[e]+=t);for(let e of this.axis)if(e<0||e>=t)throw new Error(`Invalid axis: ${e}`);if(this.axis.length!==yb(this.axis).length)throw new Error(`Found duplicate axes in: ${this.axis}`);let n=this.axis.map(t=>e[t]),r=!0;this.scale?this.gamma=this.addWeight("gamma",n,"float32",this.gammaInitializer,this.gammaRegularizer,r):this.gamma=null,this.center?this.beta=this.addWeight("beta",n,"float32",this.betaInitializer,this.betaRegularizer,r):this.beta=null,this.built=!0}call(e,t){let n=Ox(e),r=n.shape,a=r.length;return Qs(()=>{let{mean:e,variance:t}=lh(n,this.axis,!0),s=sb(1,a);for(let e of this.axis)s[e]=r[e];let i=e=>null!=e&&e.shape.length!==a?jo(e,s):e,o=this.scale?i(this.gamma.read()):null,l=this.center?i(this.beta.read()):null,u=[],h=[];for(let e=0;e<a;++e)-1!==this.axis.indexOf(e)?(u.push(r[e]),h.push(1)):(u.push(1),h.push(r[e]));return e=fu(e,u),t=fu(t,u),null!=o&&(o=fu(o,h)),null!=l&&(l=fu(l,h)),DI(n,e,t,l,o,this.epsilon)})}getConfig(){let e={axis:this.axis,epsilon:this.epsilon,center:this.center,scale:this.scale,betaInitializer:Rx(this.betaInitializer),gammaInitializer:Rx(this.gammaInitializer),betaRegularizer:Sk(this.betaRegularizer),gammaRegularizer:Sk(this.gammaRegularizer)},t=super.getConfig();return Object.assign(e,t),e}};LI.className="LayerNormalization",kc.registerClass(LI);var PI=class extends Kx{constructor(e){if(null==e&&(e={}),super(e),this.dataFormat=null==e.dataFormat?"channelsLast":e.dataFormat,null==e.padding)this.padding=[[1,1],[1,1]];else if("number"==typeof e.padding)this.padding=[[e.padding,e.padding],[e.padding,e.padding]];else{if(e.padding=e.padding,2!==e.padding.length)throw new tb(`ZeroPadding2D expects padding to be a length-2 array, but received a length-${e.padding.length} array.`);let t,n;if("number"==typeof e.padding[0])t=[e.padding[0],e.padding[0]],n=[e.padding[1],e.padding[1]];else{if(e.padding=e.padding,2!==e.padding[0].length)throw new tb(`ZeroPadding2D expects height padding to be a length-2 array, but received a length-${e.padding[0].length} array.`);if(t=e.padding[0],2!==e.padding[1].length)throw new tb(`ZeroPadding2D expects width padding to be a length-2 array, but received a length-${e.padding[1].length} array.`);n=e.padding[1]}this.padding=[t,n]}this.inputSpec=[new Ux({ndim:4})]}computeOutputShape(e){let t,n;return e=Lx(e),"channelsFirst"===this.dataFormat?(t=null!=e[2]&&e[2]>=0?e[2]+this.padding[0][0]+this.padding[0][1]:null,n=null!=e[3]&&e[3]>=0?e[3]+this.padding[1][0]+this.padding[1][1]:null,[e[0],e[1],t,n]):(t=null!=e[1]&&e[1]>=0?e[1]+this.padding[0][0]+this.padding[0][1]:null,n=null!=e[2]&&e[2]>=0?e[2]+this.padding[1][0]+this.padding[1][1]:null,[e[0],t,n,e[3]])}call(e,t){return Qs(()=>function(e,t,n){return Qs(()=>{if(4!==e.rank)throw new tb(`temporalPadding expects input tensor to be 4-D, but received a ${e.rank}-D tensor.`);if(null==t&&(t=[[1,1],[1,1]]),2!==t.length||2!==t[0].length||2!==t[1].length)throw new tb("spatial2dPadding expects `padding` to be an Array of two Arrays, each of which is an Array of two integers.");if(null==n&&(n="channelsLast"),"channelsLast"!==n&&"channelsFirst"!==n)throw new tb(`Unknown data format: ${n}. Supported data formats are 'channelsLast' and 'channelsFirst.`);let r;return r="channelsFirst"===n?[[0,0],[0,0],t[0],t[1]]:[[0,0],t[0],t[1],[0,0]],mh(e,r)})}(Ox(e),this.padding,this.dataFormat))}getConfig(){let e={padding:this.padding,dataFormat:this.dataFormat},t=super.getConfig();return Object.assign(e,t),e}};function zI(e,t,n,r,a,s){return Qs(()=>{Db(a),Ob(s),Mb(r),null==n&&(n=[1,1]),null==r&&(r="valid"),null==a&&(a="channelsLast"),null==s&&(s="max"),e=Ok(e,a);let i,o="same"===r?"same":"valid";return i="max"===s?Yu(e,t,n,o):qo(e,t,n,o),"channelsFirst"===a&&(i=Jd(i,[0,3,1,2])),i})}function BI(e,t,n,r,a,s){return Qs(()=>{Db(a),Ob(s),Mb(r),null==n&&(n=[1,1,1]),null==r&&(r="valid"),null==a&&(a="channelsLast"),null==s&&(s="max"),e=Lk(e,a);let i,o="same"===r?"same":"valid";return i="max"===s?Ju(e,t,n,o):Ko(e,t,n,o),"channelsFirst"===a&&(i=Jd(i,[0,4,1,2,3])),i})}PI.className="ZeroPadding2D",kc.registerClass(PI);var WI=class extends Kx{constructor(e){if(null==e.poolSize&&(e.poolSize=2),super(e),"number"==typeof e.poolSize)this.poolSize=[e.poolSize];else{if(!Array.isArray(e.poolSize)||1!==e.poolSize.length||"number"!=typeof e.poolSize[0])throw new tb(`poolSize for 1D convolutional layer must be a number or an Array of a single number, but received ${JSON.stringify(e.poolSize)}`);this.poolSize=e.poolSize}if(wb(this.poolSize,"poolSize"),null==e.strides)this.strides=this.poolSize;else if("number"==typeof e.strides)this.strides=[e.strides];else{if(!Array.isArray(e.strides)||1!==e.strides.length||"number"!=typeof e.strides[0])throw new tb(`strides for 1D convolutional layer must be a number or an Array of a single number, but received ${JSON.stringify(e.strides)}`);this.strides=e.strides}wb(this.strides,"strides"),this.padding=null==e.padding?"valid":e.padding,Mb(this.padding),this.inputSpec=[new Ux({ndim:3})]}computeOutputShape(e){let t=Dk((e=Lx(e))[1],this.poolSize[0],this.padding,this.strides[0]);return[e[0],t,e[2]]}call(e,t){return Qs(()=>{this.invokeCallHook(e,t),e=Yb(Ox(e),2);let n=this.poolingFunction(Ox(e),[this.poolSize[0],1],[this.strides[0],1],this.padding,"channelsLast");return Nd(n,[2])})}getConfig(){let e={poolSize:this.poolSize,padding:this.padding,strides:this.strides},t=super.getConfig();return Object.assign(e,t),e}},VI=class extends WI{constructor(e){super(e)}poolingFunction(e,t,n,r,a){return Db(a),Mb(r),zI(e,t,n,r,a,"max")}};VI.className="MaxPooling1D",kc.registerClass(VI);var UI=class extends WI{constructor(e){super(e)}poolingFunction(e,t,n,r,a){return Db(a),Mb(r),zI(e,t,n,r,a,"avg")}};UI.className="AveragePooling1D",kc.registerClass(UI);var GI=class extends Kx{constructor(e){if(null==e.poolSize&&(e.poolSize=[2,2]),super(e),this.poolSize=Array.isArray(e.poolSize)?e.poolSize:[e.poolSize,e.poolSize],null==e.strides)this.strides=this.poolSize;else if(Array.isArray(e.strides)){if(2!==e.strides.length)throw new tb(`If the strides property of a 2D pooling layer is an Array, it is expected to have a length of 2, but received length ${e.strides.length}.`);this.strides=e.strides}else this.strides=[e.strides,e.strides];wb(this.poolSize,"poolSize"),wb(this.strides,"strides"),this.padding=null==e.padding?"valid":e.padding,this.dataFormat=null==e.dataFormat?"channelsLast":e.dataFormat,Db(this.dataFormat),Mb(this.padding),this.inputSpec=[new Ux({ndim:4})]}computeOutputShape(e){e=Lx(e);let t="channelsFirst"===this.dataFormat?e[2]:e[1],n="channelsFirst"===this.dataFormat?e[3]:e[2];return t=Dk(t,this.poolSize[0],this.padding,this.strides[0]),n=Dk(n,this.poolSize[1],this.padding,this.strides[1]),"channelsFirst"===this.dataFormat?[e[0],e[1],t,n]:[e[0],t,n,e[3]]}call(e,t){return Qs(()=>(this.invokeCallHook(e,t),this.poolingFunction(Ox(e),this.poolSize,this.strides,this.padding,this.dataFormat)))}getConfig(){let e={poolSize:this.poolSize,padding:this.padding,strides:this.strides,dataFormat:this.dataFormat},t=super.getConfig();return Object.assign(e,t),e}},HI=class extends GI{constructor(e){super(e)}poolingFunction(e,t,n,r,a){return Db(a),Mb(r),zI(e,t,n,r,a,"max")}};HI.className="MaxPooling2D",kc.registerClass(HI);var jI=class extends GI{constructor(e){super(e)}poolingFunction(e,t,n,r,a){return Db(a),Mb(r),zI(e,t,n,r,a,"avg")}};jI.className="AveragePooling2D",kc.registerClass(jI);var qI=class extends Kx{constructor(e){if(null==e.poolSize&&(e.poolSize=[2,2,2]),super(e),this.poolSize=Array.isArray(e.poolSize)?e.poolSize:[e.poolSize,e.poolSize,e.poolSize],null==e.strides)this.strides=this.poolSize;else if(Array.isArray(e.strides)){if(3!==e.strides.length)throw new tb(`If the strides property of a 3D pooling layer is an Array, it is expected to have a length of 3, but received length ${e.strides.length}.`);this.strides=e.strides}else this.strides=[e.strides,e.strides,e.strides];wb(this.poolSize,"poolSize"),wb(this.strides,"strides"),this.padding=null==e.padding?"valid":e.padding,this.dataFormat=null==e.dataFormat?"channelsLast":e.dataFormat,Db(this.dataFormat),Mb(this.padding),this.inputSpec=[new Ux({ndim:5})]}computeOutputShape(e){e=Lx(e);let t="channelsFirst"===this.dataFormat?e[2]:e[1],n="channelsFirst"===this.dataFormat?e[3]:e[2],r="channelsFirst"===this.dataFormat?e[4]:e[3];return t=Dk(t,this.poolSize[0],this.padding,this.strides[0]),n=Dk(n,this.poolSize[1],this.padding,this.strides[1]),r=Dk(r,this.poolSize[2],this.padding,this.strides[2]),"channelsFirst"===this.dataFormat?[e[0],e[1],t,n,r]:[e[0],t,n,r,e[4]]}call(e,t){return Qs(()=>(this.invokeCallHook(e,t),this.poolingFunction(Ox(e),this.poolSize,this.strides,this.padding,this.dataFormat)))}getConfig(){let e={poolSize:this.poolSize,padding:this.padding,strides:this.strides,dataFormat:this.dataFormat},t=super.getConfig();return Object.assign(e,t),e}},KI=class extends qI{constructor(e){super(e)}poolingFunction(e,t,n,r,a){return Db(a),Mb(r),BI(e,t,n,r,a,"max")}};KI.className="MaxPooling3D",kc.registerClass(KI);var XI=class extends qI{constructor(e){super(e)}poolingFunction(e,t,n,r,a){return Db(a),Mb(r),BI(e,t,n,r,a,"avg")}};XI.className="AveragePooling3D",kc.registerClass(XI);var ZI=class extends Kx{constructor(e){super(e),this.inputSpec=[new Ux({ndim:3})]}computeOutputShape(e){return[e[0],e[2]]}call(e,t){throw new nb}},YI=class extends ZI{constructor(e){super(e||{})}call(e,t){return Qs(()=>{let t=Ox(e);return th(t,1)})}};YI.className="GlobalAveragePooling1D",kc.registerClass(YI);var JI=class extends ZI{constructor(e){super(e||{})}call(e,t){return Qs(()=>{let t=Ox(e);return tu(t,1)})}};JI.className="GlobalMaxPooling1D",kc.registerClass(JI);var QI=class extends Kx{constructor(e){super(e),this.dataFormat=null==e.dataFormat?"channelsLast":e.dataFormat,Db(this.dataFormat),this.inputSpec=[new Ux({ndim:4})]}computeOutputShape(e){return"channelsLast"===this.dataFormat?[e[0],e[3]]:[e[0],e[1]]}call(e,t){throw new nb}getConfig(){let e={dataFormat:this.dataFormat},t=super.getConfig();return Object.assign(e,t),e}},eS=class extends QI{call(e,t){return Qs(()=>{let t=Ox(e);return"channelsLast"===this.dataFormat?th(t,[1,2]):th(t,[2,3])})}};eS.className="GlobalAveragePooling2D",kc.registerClass(eS);var tS=class extends QI{call(e,t){return Qs(()=>{let t=Ox(e);return"channelsLast"===this.dataFormat?tu(t,[1,2]):tu(t,[2,3])})}};tS.className="GlobalMaxPooling2D",kc.registerClass(tS);var nS=class extends Kx{constructor(e){super(e),this.layer=e.layer}build(e){this.built=!0}get trainable(){return null!=this.layer&&this.layer.trainable}set trainable(e){null!=this.layer&&(this.layer.trainable=e)}get trainableWeights(){return this.layer.trainableWeights}get nonTrainableWeights(){return this.layer.nonTrainableWeights}get updates(){return this.layer._updates}get losses(){return this.layer.losses}getWeights(){return this.layer.getWeights()}setWeights(e){this.layer.setWeights(e)}getConfig(){let e={layer:{className:this.layer.getClassName(),config:this.layer.getConfig()}},t=super.getConfig();return Object.assign(e,t),e}setFastWeightInitDuringBuild(e){super.setFastWeightInitDuringBuild(e),null!=this.layer&&this.layer.setFastWeightInitDuringBuild(e)}static fromConfig(e,t,n={}){let r=Kv(t.layer,n);delete t.layer;let a={layer:r};return Object.assign(a,t),new e(a)}},rS=class extends nS{constructor(e){super(e),this.supportsMasking=!0}build(e){if((e=Lx(e)).length<3)throw new tb(`TimeDistributed layer expects an input shape >= 3D, but received input shape ${JSON.stringify(e)}`);this.inputSpec=[{shape:e}];let t=[e[0]].concat(e.slice(2));this.layer.built||(this.layer.build(t),this.layer.built=!0),super.build(e)}computeOutputShape(e){let t=[(e=Lx(e))[0]].concat(e.slice(2)),n=this.layer.computeOutputShape(t),r=e[1];return[n[0],r].concat(n.slice(1))}call(e,t){return Qs(()=>Jk((e,n)=>[Ox(this.layer.call(e,t)),[]],e=Ox(e),[],!1,null,null,!1,!0)[1])}};rS.className="TimeDistributed",kc.registerClass(rS);var aS=class extends nS{constructor(e){super(e);let t=e.layer.getConfig(),n={};n.className=e.layer.getClassName(),n.config=t,this.forwardLayer=Kv(n),t.goBackwards=!0!==t.goBackwards;let r={};if(r.className=e.layer.getClassName(),r.config=t,this.backwardLayer=Kv(r),this.forwardLayer.name="forward_"+this.forwardLayer.name,this.backwardLayer.name="backward_"+this.backwardLayer.name,this.mergeMode=void 0===e.mergeMode?"concat":e.mergeMode,function(e){xb(Rb,"BidirectionalMergeMode",e)}(this.mergeMode),e.weights)throw new nb("weights support is not implemented for Bidirectional layer yet.");this._stateful=e.layer.stateful,this.returnSequences=e.layer.returnSequences,this.returnState=e.layer.returnState,this.supportsMasking=!0,this._trainable=!0,this.inputSpec=e.layer.inputSpec,this.numConstants=null}get trainable(){return this._trainable}set trainable(e){this._trainable=e,null!=this.forwardLayer&&(this.forwardLayer.trainable=e),null!=this.backwardLayer&&(this.backwardLayer.trainable=e)}getWeights(){return this.forwardLayer.getWeights().concat(this.backwardLayer.getWeights())}setWeights(e){let t=e.length,n=Math.floor(t/2);this.forwardLayer.setWeights(e.slice(0,n)),this.backwardLayer.setWeights(e.slice(n))}computeOutputShape(e){let t,n,r,a=this.forwardLayer.computeOutputShape(e);return Array.isArray(a)&&Array.isArray(a[0])||(a=[a]),this.returnState&&(r=a.slice(1)),t=a[0],"concat"===this.mergeMode?(t[t.length-1]*=2,n=[t]):n=null==this.mergeMode?[t,t.slice()]:[t],this.returnState?null==this.mergeMode?n.concat(r).concat(r.slice()):[t].concat(r).concat(r.slice()):lb(n)}apply(e,t){let n=null==t?null:t.initialState,r=null==t?null:t.constants;null==t&&(t={});let a=Yk(e,n,r,this.numConstants);if(e=a.inputs,n=a.initialState,r=a.constants,Array.isArray(e)&&(n=e.slice(1),e=e[0]),(null==n||0===n.length)&&null==r)return super.apply(e,t);let s=[],i=[];if(null!=n){let e=n.length;if(e%2>0)throw new tb("When passing `initialState` to a Bidrectional RNN, the state should be an Array containing the states of the underlying RNNs.");t.initialState=n,s.push(...n);let r=n.map(e=>new Ux({shape:e.shape}));this.forwardLayer.stateSpec=r.slice(0,e/2),this.backwardLayer.stateSpec=r.slice(e/2),i.push(...r)}if(null!=r)throw new nb("Support for constants in Bidirectional layers is not implemented yet.");let o=s[0]instanceof Gx;for(let e of s)if(e instanceof Gx!==o)throw new tb("The initial state of a Bidirectional layer cannot be specified as a mix of symbolic and non-symbolic tensors");if(o){let n=[e].concat(s),r=this.inputSpec.concat(i),a=this.inputSpec;this.inputSpec=r;let o=super.apply(n,t);return this.inputSpec=a,o}return super.apply(e,t)}call(e,t){return Qs(()=>{let n,r,a,s,i=t.initialState;if(null==i)n=this.forwardLayer.call(e,t),r=this.backwardLayer.call(e,t);else{let a=i.slice(0,i.length/2),s=i.slice(i.length/2);n=this.forwardLayer.call(e,Object.assign(t,{initialState:a})),r=this.backwardLayer.call(e,Object.assign(t,{initialState:s}))}return this.returnState&&(Array.isArray(n)&&(a=n.slice(1).concat(r.slice(1))),n=n[0],r=r[0]),this.returnSequences&&(r=nd(r,1)),"concat"===this.mergeMode?s=tx([n,r]):"sum"===this.mergeMode?s=fo(n,r):"ave"===this.mergeMode?s=yo(.5,fo(n,r)):"mul"===this.mergeMode?s=yo(n,r):null==this.mergeMode&&(s=[n,r]),this.returnState?null==this.mergeMode?s.concat(a):[s].concat(a):s})}resetStates(e){this.forwardLayer.resetStates(),this.backwardLayer.resetStates()}build(e){Pb(this.forwardLayer.name,()=>{this.forwardLayer.build(e)}),Pb(this.backwardLayer.name,()=>{this.backwardLayer.build(e)}),this.built=!0}computeMask(e,t){let n;if(Array.isArray(t)&&(t=t[0]),n=this.returnSequences?null==this.mergeMode?[t,t]:t:null==this.mergeMode?[null,null]:null,this.returnState){let e=this.forwardLayer.states.map(e=>null);return Array.isArray(n)?n.concat(e).concat(e):[n].concat(e).concat(e)}return n}get trainableWeights(){return this.forwardLayer.trainableWeights.concat(this.backwardLayer.trainableWeights)}get nonTrainableWeights(){return this.forwardLayer.nonTrainableWeights.concat(this.backwardLayer.nonTrainableWeights)}setFastWeightInitDuringBuild(e){super.setFastWeightInitDuringBuild(e),null!=this.forwardLayer&&this.forwardLayer.setFastWeightInitDuringBuild(e),null!=this.backwardLayer&&this.backwardLayer.setFastWeightInitDuringBuild(e)}getConfig(){let e={mergeMode:this.mergeMode},t=super.getConfig();return Object.assign(e,t),e}static fromConfig(e,t){let n=Kv(t.layer);if(delete t.layer,null!=t.numConstants)throw new nb("Deserialization of a Bidirectional layer with numConstants present is not supported yet.");let r=t;return r.layer=n,new e(r)}};aS.className="Bidirectional",kc.registerClass(aS);var sS=class extends Kx{constructor(e){super(e),this.scale=e.scale,e.offset?this.offset=e.offset:this.offset=0}getConfig(){let e={scale:this.scale,offset:this.offset},t=super.getConfig();return Object.assign(e,t),e}call(e,t){return Qs(()=>("float32"!==(e=Ox(e)).dtype&&(e=Zb(e,"float32")),fo(yo(e,this.scale),this.offset)))}};sS.className="Rescaling",kc.registerClass(sS);var{resizeBilinear:iS,cropAndResize:oS}=yc,lS=class extends Kx{constructor(e){super(e),this.height=e.height,this.width=e.width}centerCrop(e,t,n,r,a,s,i,o){return Qs(()=>{let l,u=!1,h=[t/s,n/i,(r+t)/s,(a+n)/i],d=[];3===e.rank?(u=!0,l=Td([e])):l=e;for(let e=0;e<l.shape[0];e++)d.push(h);let p=Vs(d,[d.length,4]),c=Yh(0,d.length,1,"int32"),f=oS(l,p,c,[r,a],"nearest");return Zb(u?Ox(jd(f)):f,o)})}upsize(e,t,n,r){return Qs(()=>Zb(iS(e,[t,n]),r))}call(e,t){return Qs(()=>{let t=Ox(e),n=t.dtype,r=t.shape,a=r[r.length-3],s=r[r.length-2],i=0;a!==this.height&&(i=Math.floor((a-this.height)/2));let o=0;return s!==this.width&&(o=Math.floor((s-this.width)/2),0===o&&(o=1)),i>=0&&o>=0?this.centerCrop(t,i,o,this.height,this.width,a,s,n):this.upsize(e,this.height,this.width,n)})}getConfig(){let e={height:this.height,width:this.width},t=super.getConfig();return Object.assign(e,t),e}computeOutputShape(e){let t=(e=Lx(e)).length-3,n=e.length-2;return e[t]=this.height,e[n]=this.width,e}};lS.className="CenterCrop",kc.registerClass(lS);var uS=class extends Kx{constructor(e){super(e),this.numTokens=e.numTokens,e.outputMode?this.outputMode=e.outputMode:this.outputMode="multiHot"}getConfig(){let e={numTokens:this.numTokens,outputMode:this.outputMode},t=super.getConfig();return Object.assign(e,t),e}computeOutputShape(e){return null==(e=Lx(e))?[this.numTokens]:"oneHot"===this.outputMode&&1!==e[e.length-1]?(e.push(this.numTokens),e):(e[e.length-1]=this.numTokens,e)}call(e,t){return Qs(()=>{let n;if("int32"!==(e=Ox(e)).dtype&&(e=Zb(e,"int32")),void 0!==t.countWeights){if("count"!==this.outputMode)throw new tb(`countWeights is not used when outputMode !== count.\n              Received countWeights=${t.countWeights}`);n=Ox(t.countWeights)}let r=tu(e),a=nu(e),s=bu(this.numTokens,r).bufferSync().get(0),i=xu(a,0).bufferSync().get(0);if(!s||!i)throw new tb(`Input values must be between 0 < values <= numTokens with numTokens=${this.numTokens}`);return function(e,t,n,r){let a=Ox(e);if("int32"!==a.dtype&&(a=Zb(a,"int32")),"int"===t)return a;let s=a.shape;if(0===a.rank&&(a=pu(a,-1)),"oneHot"===t&&1!==a.shape[a.shape.length-1]&&(a=pu(a,-1)),a.rank>2)throw new tb(`When outputMode is not int, maximum output rank is 2 Received outputMode ${t} and input shape ${s} which would result in output rank ${a.rank}.`);let i,o=["multiHot","oneHot"].includes(t),l=a;if(i=El(l,void 0!==r&&"count"===t?r:[],n,o),"tfIdf"!==t)return i;if(r)return yo(i,r);throw new tb("When outputMode is 'tfIdf', weights must be provided.")}(e,this.outputMode,this.numTokens,n)})}};uS.className="CategoryEncoding",kc.registerClass(uS);var hS=new Set(["bilinear","nearest"]),dS=class extends Kx{constructor(e){if(super(e),this.height=e.height,this.width=e.width,e.interpolation){if(!hS.has(e.interpolation))throw new tb(`Invalid interpolation parameter: ${e.interpolation} is not implemented`);this.interpolation=e.interpolation}else this.interpolation="bilinear";this.cropToAspectRatio=!!e.cropToAspectRatio}computeOutputShape(e){let t=(e=Lx(e))[2];return[this.height,this.width,t]}getConfig(){let e={height:this.height,width:this.width,interpolation:this.interpolation,cropToAspectRatio:this.cropToAspectRatio},t=super.getConfig();return Object.assign(e,t),e}call(e,t){return Qs(()=>{let t=[this.height,this.width];if("bilinear"===this.interpolation)return yc.resizeBilinear(e,t,!this.cropToAspectRatio);if("nearest"===this.interpolation)return yc.resizeNearestNeighbor(e,t,!this.cropToAspectRatio);throw new Error(`Interpolation is ${this.interpolation} but only ${[...hS]} are supported`)})}};dS.className="Resizing",kc.registerClass(dS);var pS=class{constructor(e){this.seed=e}next(){if(void 0!==this.seed)return this.seed++}};pS.className="RandomSeed";var cS=class extends Kx{constructor(e){super(e),this.randomGenerator=new pS(e.seed)}getConfig(){let e={seed:this.randomGenerator.seed},t=super.getConfig();return Object.assign(e,t),e}};cS.className="BaseRandomLayer";var fS=new Set(["bilinear","nearest"]),mS=class extends cS{constructor(e){super(e);let{factor:t,interpolation:n="bilinear"}=e;if(this.factor=t,Array.isArray(this.factor)&&2===this.factor.length)this.widthLower=this.factor[0],this.widthUpper=this.factor[1];else{if(Array.isArray(this.factor)||!(this.factor>0))throw new tb(`Invalid factor: ${this.factor}. Must be positive number or tuple of 2 numbers`);this.widthLower=-this.factor,this.widthUpper=this.factor}if(this.widthLower<-1||this.widthUpper<-1)throw new tb(`factor must have values larger than -1. Got: ${this.factor}`);if(this.widthUpper<this.widthLower)throw new tb(`factor cannot have upper bound less than lower bound.\n        Got upper bound: ${this.widthUpper}.\n        Got lower bound: ${this.widthLower}\n      `);if(n){if(!fS.has(n))throw new tb(`Invalid interpolation parameter: ${n} is not implemented`);this.interpolation=n}}getConfig(){let e={factor:this.factor,interpolation:this.interpolation},t=super.getConfig();return Object.assign(e,t),e}computeOutputShape(e){let t=(e=Lx(e))[2];return[this.imgHeight,-1,t]}call(e,t){return Qs(()=>{let t=Ox(e);this.imgHeight=t.shape[t.shape.length-3];let n=t.shape[t.shape.length-2];this.widthFactor=Xh([1],1+this.widthLower,1+this.widthUpper,"float32",this.randomGenerator.next());let r=this.widthFactor.dataSync()[0]*n;r=Math.round(r);let a=[this.imgHeight,r];switch(this.interpolation){case"bilinear":return yc.resizeBilinear(e,a);case"nearest":return yc.resizeNearestNeighbor(e,a);default:throw new Error(`Interpolation is ${this.interpolation}\n          but only ${[...fS]} are supported`)}})}};function gS(e){return new Zx(e)}function yS(e){return new Ak(e)}function bS(e){return new Tk(e)}function xS(e){return new Ck(e)}function vS(e){return new Ek(e)}function wS(e){return new Rk(e)}function kS(e){return new $k(e)}function IS(e){return new qk(e)}function SS(e){return new Wk(e)}function _S(e){return new Uk(e)}function NS(e){return new Vk(e)}function TS(e){return new Gk(e)}function CS(e){return new jk(e)}function ES(e){return new Kk(e)}function AS(e){return new Xk(e)}function $S(e){return new Zk(e)}function RS(e){return new gI(e)}function FS(e){return new fI(e)}function DS(e){return new pI(e)}function MS(e){return new cI(e)}function OS(e){return new mI(e)}function LS(e){return new yI(e)}function PS(e){return new bI(e)}function zS(e){return new xI(e)}function BS(e){return new wI(e)}function WS(e){return new II(e)}function VS(e){return new _I(e)}function US(e){return new CI(e)}function GS(e){return new NI(e)}function HS(e){return new TI(e)}function jS(e){return new SI(e)}function qS(e){return new AI(e)}function KS(e){return new OI(e)}function XS(e){return new LI(e)}function ZS(e){return new PI(e)}function YS(e){return new UI(e)}function JS(e){return YS(e)}function QS(e){return YS(e)}function e_(e){return new jI(e)}function t_(e){return e_(e)}function n_(e){return e_(e)}function r_(e){return new XI(e)}function a_(e){return r_(e)}function s_(e){return r_(e)}function i_(e){return new YI(e)}function o_(e){return new eS(e)}function l_(e){return new JI(e)}function u_(e){return new tS(e)}function h_(e){return new VI(e)}function d_(e){return new HI(e)}function p_(e){return new KI(e)}function c_(e){return new aI(e)}function f_(e){return new rI(e)}function m_(e){return new iI(e)}function g_(e){return new sI(e)}function y_(e){return new nI(e)}function b_(e){return new tI(e)}function x_(e){return new dI(e)}function v_(e){return new hI(e)}function w_(e){return new Qk(e)}function k_(e){return new oI(e)}function I_(e){return new aS(e)}function S_(e){return new rS(e)}mS.className="RandomWidth",kc.registerClass(mS);var __=l_,N_=u_,T_=h_,C_=d_;function E_(e){return new $I(e)}function A_(e){return new RI(e)}function $_(e){return new FI(e)}function R_(e){return new vI(e)}function F_(e){return new sS(e)}function D_(e){return new lS(e)}function M_(e){return new dS(e)}function O_(e){return new uS(e)}function L_(e){return new mS(e)}var P_={};function z_(e,t){return sw(e,t)}function B_(e,t){return hw(e,t)}function W_(e,t){return dw(e,t)}function V_(e,t){return iw(e,t)}function U_(e,t){return pw(e,t)}function G_(e,t){return lw(e,t)}function H_(e,t){return uw(e,t)}function j_(e,t){return nw(e,t)}function q_(e,t){return Yv(e,t)}function K_(e,t){return Jv(e,t)}function X_(e,t){return Jv(e,t)}function Z_(e,t){return Jv(e,t)}function Y_(e,t){return Zv(e,t)}function J_(e,t){return Zv(e,t)}function Q_(e,t){return Zv(e,t)}function eN(e,t){return function(e,t){return Qs(()=>{let n=e.sub(t).square().sum(),r=e.sub(e.mean()).square().sum();return au(1).sub(n.div(r))})}(e,t)}f(P_,{MAPE:()=>X_,MSE:()=>J_,binaryAccuracy:()=>z_,binaryCrossentropy:()=>B_,categoricalAccuracy:()=>V_,categoricalCrossentropy:()=>U_,cosineProximity:()=>j_,mape:()=>Z_,meanAbsoluteError:()=>q_,meanAbsolutePercentageError:()=>K_,meanSquaredError:()=>Y_,mse:()=>Q_,precision:()=>G_,r2Score:()=>eN,recall:()=>H_,sparseCategoricalAccuracy:()=>W_});var tN={};f(tN,{modelFromJSON:()=>qw});var nN={};function rN(e){return new kk(e)}function aN(e){return function(e){return vk(e),new kk({l1:null!=e?e.l1:null,l2:0})}(e)}function sN(e){return function(e){return vk(e),new kk({l2:null!=e?e.l2:null,l1:0})}(e)}f(nN,{l1:()=>aN,l1l2:()=>rN,l2:()=>sN});var iN=class extends Bv{constructor(){super(...arguments),this.model=null}setModel(e){if(!(e instanceof Hw))throw new Error("model must be a LayersModel, not some other Container");this.model=e}};function oN(e,t){return e<t}function lN(e,t){return e>t}var uN=class extends iN{constructor(e){if(super(),null==e&&(e={}),e.restoreBestWeights)throw new nb("restoreBestWeights = True is not implemented in EarlyStopping yet.");this.monitor=e.monitor||"val_loss",this.minDelta=Math.abs(e.minDelta||0),this.patience=e.patience||0,this.verbose=e.verbose||0,this.mode=e.mode||"auto",this.baseline=e.baseline,-1===["auto","min","max"].indexOf(this.mode)&&(console.warn(`EarlyStopping mode '${this.mode}' is invalid. Falling back to mode 'auto'.`),this.mode="auto"),"min"===this.mode?this.monitorFunc=oN:"max"===this.mode||-1!==this.monitor.indexOf("acc")?this.monitorFunc=lN:this.monitorFunc=oN,this.monitorFunc===oN&&(this.minDelta*=-1)}async onTrainBegin(e){this.wait=0,this.stoppedEpoch=0,null!=this.baseline?this.best=this.baseline:this.best=this.monitorFunc===oN?1/0:-1/0}async onEpochEnd(e,t){await Pv(t);let n=this.getMonitorValue(t);null!=n&&(this.monitorFunc(n-this.minDelta,this.best)?(this.best=n,this.wait=0):(this.wait++,this.wait>=this.patience&&(this.stoppedEpoch=e,this.model.stopTraining=!0)))}async onTrainEnd(e){this.stoppedEpoch>0&&this.verbose&&console.log(`Epoch ${this.stoppedEpoch}: early stopping.`)}getMonitorValue(e){null==e&&(e={});let t=e[this.monitor];return null==t&&console.warn(`Metric for EarlyStopping ${this.monitor} is not available. Available metrics are: ${Object.keys(e)}`),t}};var hN,dN,pN={earlyStopping:function(e){return new uN(e)}};Le().registerFlag("KEEP_INTERMEDIATE_TENSORS",()=>!1,e=>{e&&console.warn("Keep intermediate tensors is ON. This will print the values of all intermediate tensors during model inference. Not all models support this mode. For details, check e2e/benchmarks/ model_config.js. This significantly impacts performance.")}),function(e){e[e.DT_INVALID=0]="DT_INVALID",e[e.DT_FLOAT=1]="DT_FLOAT",e[e.DT_DOUBLE=2]="DT_DOUBLE",e[e.DT_INT32=3]="DT_INT32",e[e.DT_UINT8=4]="DT_UINT8",e[e.DT_INT16=5]="DT_INT16",e[e.DT_INT8=6]="DT_INT8",e[e.DT_STRING=7]="DT_STRING",e[e.DT_COMPLEX64=8]="DT_COMPLEX64",e[e.DT_INT64=9]="DT_INT64",e[e.DT_BOOL=10]="DT_BOOL",e[e.DT_QINT8=11]="DT_QINT8",e[e.DT_QUINT8=12]="DT_QUINT8",e[e.DT_QINT32=13]="DT_QINT32",e[e.DT_BFLOAT16=14]="DT_BFLOAT16",e[e.DT_QINT16=15]="DT_QINT16",e[e.DT_QUINT16=16]="DT_QUINT16",e[e.DT_UINT16=17]="DT_UINT16",e[e.DT_COMPLEX128=18]="DT_COMPLEX128",e[e.DT_HALF=19]="DT_HALF",e[e.DT_RESOURCE=20]="DT_RESOURCE",e[e.DT_VARIANT=21]="DT_VARIANT",e[e.DT_UINT32=22]="DT_UINT32",e[e.DT_UINT64=23]="DT_UINT64",e[e.DT_FLOAT_REF=101]="DT_FLOAT_REF",e[e.DT_DOUBLE_REF=102]="DT_DOUBLE_REF",e[e.DT_INT32_REF=103]="DT_INT32_REF",e[e.DT_UINT8_REF=104]="DT_UINT8_REF",e[e.DT_INT16_REF=105]="DT_INT16_REF",e[e.DT_INT8_REF=106]="DT_INT8_REF",e[e.DT_STRING_REF=107]="DT_STRING_REF",e[e.DT_COMPLEX64_REF=108]="DT_COMPLEX64_REF",e[e.DT_INT64_REF=109]="DT_INT64_REF",e[e.DT_BOOL_REF=110]="DT_BOOL_REF",e[e.DT_QINT8_REF=111]="DT_QINT8_REF",e[e.DT_QUINT8_REF=112]="DT_QUINT8_REF",e[e.DT_QINT32_REF=113]="DT_QINT32_REF",e[e.DT_BFLOAT16_REF=114]="DT_BFLOAT16_REF",e[e.DT_QINT16_REF=115]="DT_QINT16_REF",e[e.DT_QUINT16_REF=116]="DT_QUINT16_REF",e[e.DT_UINT16_REF=117]="DT_UINT16_REF",e[e.DT_COMPLEX128_REF=118]="DT_COMPLEX128_REF",e[e.DT_HALF_REF=119]="DT_HALF_REF",e[e.DT_RESOURCE_REF=120]="DT_RESOURCE_REF",e[e.DT_VARIANT_REF=121]="DT_VARIANT_REF",e[e.DT_UINT32_REF=122]="DT_UINT32_REF",e[e.DT_UINT64_REF=123]="DT_UINT64_REF"}(hN||(hN={})),function(e){let t;var n;(n=t=e.CheckpointFormatVersion||(e.CheckpointFormatVersion={}))[n.LEGACY=0]="LEGACY",n[n.V1=1]="V1",n[n.V2=2]="V2"}(dN||(dN={}));var cN={};function fN(e,t){let n={tfOpName:e,category:"custom",inputs:[],attrs:[],customExecutor:t};cN[e]=n}function mN(e){return cN[e]}function gN(e){delete cN[e]}function yN(e,t,n,r,a){let s=t.inputParams[e];if(s&&void 0!==s.inputIndexStart){let e=s.inputIndexStart,i=0===s.inputIndexEnd?void 0:void 0===s.inputIndexEnd?e+1:s.inputIndexEnd,o=e<0?t.inputNames.length+e:e;if("tensor"===s.type)return bN(t.inputNames[o],n,r,a);if("tensors"===s.type){let s=t.inputs.slice(e,i);return t.inputNames.slice(e,i).filter((e,t)=>{var n;return"NoOp"!==(null===(n=s[t])||void 0===n?void 0:n.op)}).map(e=>bN(e,n,r,a))}let l=bN(t.inputNames[o],n,r,a),u=l.dataSync();return"number"===s.type?u[0]:va.toNestedArray(l.shape,u)}let i=t.attrParams[e];return i&&i.value}function bN(e,t,n,r){let[a,s]=kN(e,n);if(null!=r){let e=r.getHashTableHandleByName(a);if(null!=e)return e}let i=n.currentContextIds.find(e=>!!t[wN(a,e)]);return void 0!==i?t[wN(a,i)][s]:void 0}function xN(e,t,n){return t[wN(e,n.currentContextId)]}function vN(e,t){let[n,r,a]=kN(e,t);return[wN(n,t&&t.currentContextId),r,a]}function wN(e,t){return t?`${e}-${t}`:e}function kN(e,t){if(""===e)return["",0,void 0];let n=null!=t&&null!=t.parseNodeNameCache;if(n){let n=t.parseNodeNameCache.get(e);if(null!=n)return n}let r,a=e.split(":");if(1===a.length)r=[e,0,void 0];else{let e=a[0],t=3===a.length?a[1]:void 0;r=[e,Number(a[a.length-1]),t]}return n&&t.parseNodeNameCache.set(e,r),r}function IN(e,t,n){let r=yN("pad",e,t,n);if("explicit"===r){r=yN("explicitPaddings",e,t,n);let a=[[0,0],[0,0],[0,0],[0,0]];for(let e=0;e<4;e++)a[e][0]=r[2*e],a[e][1]=r[2*e+1];return a}return r}function SN(e){return e.kept?e:po(e)}var _N={};f(_N,{json:()=>NN});var NN=[{tfOpName:"Add",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"AddV2",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"AddN",category:"arithmetic",inputs:[{start:0,end:0,name:"tensors",type:"tensors"}]},{tfOpName:"BiasAdd",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0}]},{tfOpName:"Sub",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"RealDiv",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Div",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"DivNoNan",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"FloorDiv",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Mul",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Maximum",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Minimum",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Pow",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"SquaredDifference",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Mod",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"FloorMod",category:"arithmetic",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]}],TN={};f(TN,{json:()=>CN});var CN=[{tfOpName:"Abs",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Acos",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Asin",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Atan",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Atan2",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"y",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Ceil",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"ClipByValue",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"clipValueMin",type:"number"},{start:2,name:"clipValueMax",type:"number"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Complex",category:"basic_math",inputs:[{start:0,name:"real",type:"tensor"},{start:1,name:"imag",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"ComplexAbs",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Cos",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Cosh",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Elu",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Exp",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Floor",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Log",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Imag",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"Tout",name:"outputType",type:"dtype",notSupported:!0}]},{tfOpName:"Neg",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Real",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"Tout",name:"outputType",type:"dtype",notSupported:!0}]},{tfOpName:"Prelu",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"alpha",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Relu",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Relu6",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Selu",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Sigmoid",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Sin",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Sinh",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Sqrt",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Rsqrt",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Square",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Tan",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Tanh",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Sign",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Round",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Expm1",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Log1p",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Reciprocal",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Softplus",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Asinh",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Acosh",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Atanh",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Erf",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"LeakyRelu",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"alpha",name:"alpha",type:"number",defaultValue:.2},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"IsNan",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"IsFinite",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"IsInf",category:"basic_math",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]}],EN={};f(EN,{json:()=>AN});var AN=[{tfOpName:"EmptyTensorList",category:"control",inputs:[{start:0,name:"elementShape",type:"shape"},{start:1,name:"maxNumElements",type:"number"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"LoopCond",category:"control",inputs:[{start:0,name:"pred",type:"tensor"}]},{tfOpName:"Switch",category:"control",inputs:[{start:0,name:"data",type:"tensor"},{start:1,name:"pred",type:"tensor"}]},{tfOpName:"Merge",category:"control",inputs:[{start:0,end:0,name:"tensors",type:"tensors"}]},{tfOpName:"Enter",category:"control",inputs:[{start:0,name:"tensor",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"frame_name",name:"frameName",type:"string"},{tfName:"is_constant",name:"isConstant",type:"bool"}]},{tfOpName:"Exit",category:"control",inputs:[{start:0,name:"tensor",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"NextIteration",category:"control",inputs:[{start:0,name:"tensor",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"TensorArrayV3",category:"control",inputs:[{start:0,name:"size",type:"number"}],attrs:[{tfName:"dtype",name:"dtype",type:"dtype"},{tfName:"element_shape",name:"elementShape",type:"shape"},{tfName:"dynamic_size",name:"dynamicSize",type:"bool"},{tfName:"clear_after_read",name:"clearAfterRead",type:"bool"},{tfName:"identical_element_shapes",name:"identicalElementShapes",type:"bool"},{tfName:"tensor_array_name",name:"name",type:"string"}]},{tfOpName:"TensorArrayWriteV3",category:"control",inputs:[{start:0,name:"tensorArrayId",type:"tensor"},{start:1,name:"index",type:"number"},{start:2,name:"tensor",type:"tensor"},{start:3,name:"flowIn",type:"number"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"TensorArrayReadV3",category:"control",inputs:[{start:0,name:"tensorArrayId",type:"tensor"},{start:1,name:"index",type:"number"},{start:2,name:"flowIn",type:"number"}],attrs:[{tfName:"dtype",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"TensorArrayGatherV3",category:"control",inputs:[{start:0,name:"tensorArrayId",type:"tensor"},{start:1,name:"indices",type:"number[]"},{start:2,name:"flowIn",type:"number"}],attrs:[{tfName:"dtype",name:"dtype",type:"dtype"},{tfName:"element_shape",name:"elementShape",type:"shape"}]},{tfOpName:"TensorArrayScatterV3",category:"control",inputs:[{start:0,name:"tensorArrayId",type:"tensor"},{start:1,name:"indices",type:"number[]"},{start:2,name:"tensor",type:"tensor"},{start:3,name:"flowIn",type:"number"}],attrs:[{tfName:"T",name:"dtype",type:"dtype"}]},{tfOpName:"TensorArrayConcatV3",category:"control",inputs:[{start:0,name:"tensorArrayId",type:"tensor"},{start:1,name:"flowIn",type:"number"}],attrs:[{tfName:"dtype",name:"dtype",type:"dtype"},{tfName:"element_shape_except0",name:"elementShapeExcept0",type:"shape",notSupported:!0}]},{tfOpName:"TensorArraySplitV3",category:"control",inputs:[{start:0,name:"tensorArrayId",type:"tensor"},{start:1,name:"tensor",type:"tensor"},{start:2,name:"lengths",type:"number[]"},{start:3,name:"flowIn",type:"number"}],attrs:[{tfName:"T",name:"dtype",type:"dtype"}]},{tfOpName:"TensorArraySizeV3",category:"control",inputs:[{start:0,name:"tensorArrayId",type:"tensor"},{start:1,name:"flowIn",type:"number"}]},{tfOpName:"TensorArrayCloseV3",category:"control",inputs:[{start:0,name:"tensorArrayId",type:"tensor"}]},{tfOpName:"StatelessIf",category:"control",inputs:[{start:0,name:"cond",type:"tensor"},{start:1,end:0,name:"args",type:"tensors"}],attrs:[{tfName:"then_branch",name:"thenBranch",type:"func"},{tfName:"else_branch",name:"elseBranch",type:"func"}]},{tfOpName:"If",category:"control",inputs:[{start:0,name:"cond",type:"tensor"},{start:1,end:0,name:"args",type:"tensors"}],attrs:[{tfName:"then_branch",name:"thenBranch",type:"func"},{tfName:"else_branch",name:"elseBranch",type:"func"}]},{tfOpName:"StatelessWhile",category:"control",inputs:[{start:0,end:0,name:"args",type:"tensors"}],attrs:[{tfName:"cond",name:"cond",type:"func"},{tfName:"body",name:"body",type:"func"}]},{tfOpName:"While",category:"control",inputs:[{start:0,end:0,name:"args",type:"tensors"}],attrs:[{tfName:"cond",name:"cond",type:"func"},{tfName:"body",name:"body",type:"func"}]},{tfOpName:"TensorListScatter",category:"control",inputs:[{start:0,name:"tensor",type:"tensor"},{start:1,name:"indices",type:"number[]"},{start:2,name:"elementShape",type:"shape"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListScatterV2",category:"control",inputs:[{start:0,name:"tensor",type:"tensor"},{start:1,name:"indices",type:"number[]"},{start:2,name:"elementShape",type:"shape"},{start:3,name:"numElements",type:"number"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListGather",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"},{start:1,name:"indices",type:"number[]"},{start:2,name:"elementShape",type:"shape"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListGetItem",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"},{start:1,name:"index",type:"number"},{start:2,name:"elementShape",type:"shape"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListSetItem",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"},{start:1,name:"index",type:"number"},{start:2,name:"tensor",type:"tensor"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListReserve",category:"control",inputs:[{start:0,name:"elementShape",type:"shape"},{start:1,name:"numElements",type:"number"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListFromTensor",category:"control",inputs:[{start:0,name:"tensor",type:"tensor"},{start:1,name:"elementShape",type:"shape"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListStack",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"},{start:1,name:"elementShape",type:"shape"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"},{tfName:"num_elements",name:"numElements",type:"dtype"}]},{tfOpName:"TensorListSplit",category:"control",inputs:[{start:0,name:"tensor",type:"tensor"},{start:1,name:"elementShape",type:"shape"},{start:2,name:"lengths",type:"number[]"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListConcat",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"}],attrs:[{tfName:"element_shape",name:"elementShape",type:"shape"},{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListConcatV2",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"}],attrs:[{tfName:"element_shape",name:"elementShape",type:"shape"},{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListPopBack",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"},{start:1,name:"elementShape",type:"shape"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListPushBack",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"},{start:1,name:"tensor",type:"tensor"}],attrs:[{tfName:"element_dtype",name:"elementDType",type:"dtype"}]},{tfOpName:"TensorListLength",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"}]},{tfOpName:"TensorListResize",category:"control",inputs:[{start:0,name:"tensorListId",type:"tensor"},{start:1,name:"size",type:"number"}]}],$N={};f($N,{json:()=>RN});var RN=[{tfOpName:"AvgPool",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0},{tfName:"ksize",name:"kernelSize",type:"number[]"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"MaxPool",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0},{tfName:"ksize",name:"kernelSize",type:"number[]"},{tfName:"explicit_paddings",name:"explicitPaddings",type:"number[]",defaultValue:[],notSupported:!0},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"MaxPoolWithArgmax",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"ksize",name:"kernelSize",type:"number[]"},{tfName:"include_batch_in_index",name:"includeBatchInIndex",type:"bool"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"AvgPool3D",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0},{tfName:"ksize",name:"kernelSize",type:"number[]"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"MaxPool3D",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0},{tfName:"ksize",name:"kernelSize",type:"number[]"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Conv1D",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"filter",type:"tensor"}],attrs:[{tfName:"stride",name:"stride",type:"number"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",defaultValue:"NWC"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"dilation",name:"dilation",type:"number",defaultValue:1}]},{tfOpName:"Conv2D",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"filter",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"useCudnnOnGpu",name:"useCudnnOnGpu",type:"bool"},{tfName:"data_format",name:"dataFormat",type:"string",defaultValue:"NHWC"},{tfName:"explicit_paddings",name:"explicitPaddings",type:"number[]",defaultValue:[]},{tfName:"dilations",name:"dilations",type:"number[]"}]},{tfOpName:"_FusedConv2D",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"filter",type:"tensor"},{start:2,end:0,name:"args",type:"tensors"}],attrs:[{tfName:"num_args",name:"numArgs",type:"number"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"explicit_paddings",name:"explicitPaddings",type:"number[]",defaultValue:[]},{tfName:"use_cudnn_on_gpu",name:"useCudnnOnGpu",type:"bool",defaultValue:!0},{tfName:"data_format",name:"dataFormat",type:"string",defaultValue:"NHWC"},{tfName:"dilations",name:"dilations",type:"number[]",defaultValue:[1,1,1,1]},{tfName:"fused_ops",name:"fusedOps",type:"string[]",defaultValue:[]},{tfName:"epsilon",name:"epsilon",type:"number",defaultValue:1e-4},{tfName:"leakyrelu_alpha",name:"leakyreluAlpha",type:"number",defaultValue:.2}]},{tfOpName:"Conv2DBackpropInput",category:"convolution",inputs:[{start:2,name:"x",type:"tensor"},{start:1,name:"filter",type:"tensor"},{start:0,name:"outputShape",type:"number[]"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0},{tfName:"explicit_paddings",name:"explicitPaddings",type:"number[]",defaultValue:[]},{tfName:"dilations",name:"dilations",type:"number[]",notSupported:!0}]},{tfOpName:"DepthwiseConv2d",category:"convolution",inputs:[{start:0,name:"input",type:"tensor"},{start:1,name:"filter",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",defaultValue:"NHWC"},{tfName:"explicit_paddings",name:"explicitPaddings",type:"number[]",defaultValue:[]},{tfName:"dilations",name:"dilations",type:"number[]"}]},{tfOpName:"DepthwiseConv2dNative",category:"convolution",inputs:[{start:0,name:"input",type:"tensor"},{start:1,name:"filter",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",defaultValue:"NHWC"},{tfName:"explicit_paddings",name:"explicitPaddings",type:"number[]",defaultValue:[]},{tfName:"dilations",name:"dilations",type:"number[]"}]},{tfOpName:"FusedDepthwiseConv2dNative",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"filter",type:"tensor"},{start:2,end:0,name:"args",type:"tensors"}],attrs:[{tfName:"num_args",name:"numArgs",type:"number"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",defaultValue:"NHWC"},{tfName:"dilations",name:"dilations",type:"number[]",defaultValue:[1,1,1,1]},{tfName:"fused_ops",name:"fusedOps",type:"string[]",defaultValue:[]},{tfName:"explicit_paddings",name:"explicitPaddings",type:"number[]",defaultValue:[]}]},{tfOpName:"Conv3D",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"filter",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"padding",name:"pad",type:"string"},{tfName:"data_format",name:"dataFormat",type:"string",defaultValue:"NHWC"},{tfName:"dilations",name:"dilations",type:"number[]"}]},{tfOpName:"Dilation2D",category:"convolution",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"filter",type:"tensor"}],attrs:[{tfName:"strides",name:"strides",type:"number[]"},{tfName:"rates",name:"dilations",type:"number[]"},{tfName:"padding",name:"pad",type:"string"}]}],FN={};f(FN,{json:()=>DN});var DN=[{tfOpName:"Fill",category:"creation",inputs:[{start:0,name:"shape",type:"number[]"},{start:1,name:"value",type:"number"}],attrs:[{tfName:"T",name:"dtype",type:"dtype"}]},{tfOpName:"LinSpace",category:"creation",inputs:[{start:0,name:"start",type:"number"},{start:1,name:"stop",type:"number"},{start:2,name:"num",type:"number"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"OneHot",category:"creation",inputs:[{start:0,name:"indices",type:"tensor"},{start:1,name:"depth",type:"number"},{start:2,name:"onValue",type:"number",defaultValue:1},{start:3,name:"offValue",type:"number",defaultValue:0}],attrs:[{tfName:"axis",name:"axis",type:"number",notSupported:!0},{tfName:"T",name:"dtype",type:"dtype"}]},{tfOpName:"Ones",category:"creation",inputs:[{start:0,name:"shape",type:"number[]"}],attrs:[{tfName:"T",name:"dtype",type:"dtype"}]},{tfOpName:"OnesLike",category:"creation",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"dtype",name:"dtype",type:"dtype"}]},{tfOpName:"RandomStandardNormal",category:"creation",inputs:[{start:0,name:"shape",type:"number[]"}],attrs:[{tfName:"seed",name:"seed",type:"number",defaultValue:0},{tfName:"seed2",name:"seed2",type:"number",defaultValue:0,notSupported:!0},{tfName:"dtype",name:"dtype",type:"dtype"},{tfName:"T",name:"T",type:"number",notSupported:!0}]},{tfOpName:"RandomUniform",category:"creation",inputs:[{start:0,name:"shape",type:"number[]"}],attrs:[{tfName:"minval",name:"minval",type:"number",defaultValue:0},{tfName:"maxval",name:"maxval",type:"number",defaultValue:1},{tfName:"dtype",name:"dtype",type:"dtype"},{tfName:"seed",name:"seed",type:"number",defaultValue:0},{tfName:"seed2",name:"seed2",type:"number",defaultValue:0,notSupported:!0},{tfName:"T",name:"T",type:"number",notSupported:!0}]},{tfOpName:"RandomUniformInt",category:"creation",inputs:[{start:0,name:"shape",type:"number[]"}],attrs:[{tfName:"minval",name:"minval",type:"number"},{tfName:"maxval",name:"maxval",type:"number"},{tfName:"seed",name:"seed",type:"number",defaultValue:0},{tfName:"seed2",name:"seed2",type:"number",defaultValue:0,notSupported:!0}]},{tfOpName:"Range",category:"creation",inputs:[{start:0,name:"start",type:"number"},{start:1,name:"stop",type:"number"},{start:2,name:"step",type:"number",defaultValue:0}],attrs:[{tfName:"Tidx",name:"dtype",type:"dtype"}]},{tfOpName:"TruncatedNormal",category:"creation",inputs:[{start:0,name:"shape",type:"number[]"}],attrs:[{tfName:"means",name:"mean",type:"number",defaultValue:0},{tfName:"stddev",name:"stdDev",type:"number",defaultValue:1},{tfName:"seed",name:"seed",type:"number"},{tfName:"seed2",name:"seed2",type:"number",defaultValue:0,notSupported:!0},{tfName:"dtype",name:"dtype",type:"dtype"},{tfName:"T",name:"T",type:"number",notSupported:!0}]},{tfOpName:"Zeros",category:"creation",inputs:[{start:0,name:"shape",type:"number[]"}],attrs:[{tfName:"T",name:"dtype",type:"dtype"}]},{tfOpName:"ZerosLike",category:"creation",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype"}]},{tfOpName:"Multinomial",category:"creation",inputs:[{start:0,name:"logits",type:"tensor"},{start:1,name:"numSamples",type:"number"}],attrs:[{tfName:"seed",name:"seed",type:"number"},{tfName:"seed2",name:"seed2",type:"number"},{tfName:"T",name:"dtype",type:"dtype"},{tfName:"output_dtype",name:"output_dtype",type:"dtype"}]}],MN={};f(MN,{json:()=>ON});var ON=[{tfOpName:"NonMaxSuppressionV2",category:"dynamic",inputs:[{start:0,name:"boxes",type:"tensor"},{start:1,name:"scores",type:"tensor"},{start:2,name:"maxOutputSize",type:"number"},{start:3,name:"iouThreshold",type:"number"}]},{tfOpName:"NonMaxSuppressionV3",category:"dynamic",inputs:[{start:0,name:"boxes",type:"tensor"},{start:1,name:"scores",type:"tensor"},{start:2,name:"maxOutputSize",type:"number"},{start:3,name:"iouThreshold",type:"number"},{start:4,name:"scoreThreshold",type:"number"}]},{tfOpName:"NonMaxSuppressionV4",category:"dynamic",inputs:[{start:0,name:"boxes",type:"tensor"},{start:1,name:"scores",type:"tensor"},{start:2,name:"maxOutputSize",type:"number"},{start:3,name:"iouThreshold",type:"number"},{start:4,name:"scoreThreshold",type:"number"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0},{tfName:"T_threshold",name:"threshold",type:"dtype",notSupported:!0},{tfName:"pad_to_max_output_size",name:"padToMaxOutputSize",type:"bool"}]},{tfOpName:"NonMaxSuppressionV5",category:"dynamic",inputs:[{start:0,name:"boxes",type:"tensor"},{start:1,name:"scores",type:"tensor"},{start:2,name:"maxOutputSize",type:"number"},{start:3,name:"iouThreshold",type:"number"},{start:4,name:"scoreThreshold",type:"number"},{start:5,name:"softNmsSigma",type:"number"}]},{tfOpName:"Where",category:"dynamic",inputs:[{start:0,name:"condition",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"ListDiff",category:"dynamic",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"y",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]}],LN={};f(LN,{json:()=>PN});var PN=[{tfOpName:"LowerBound",category:"evaluation",inputs:[{start:0,name:"sortedSequence",type:"tensor"},{start:1,name:"values",type:"tensor"}]},{tfOpName:"TopKV2",category:"evaluation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"k",type:"number"}],attrs:[{tfName:"sorted",name:"sorted",type:"bool"}]},{tfOpName:"UpperBound",category:"evaluation",inputs:[{start:0,name:"sortedSequence",type:"tensor"},{start:1,name:"values",type:"tensor"}]},{tfOpName:"Unique",category:"evaluation",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"UniqueV2",category:"evaluation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number"}]}],zN={};f(zN,{json:()=>BN});var BN=[{tfOpName:"PlaceholderWithDefault",category:"graph",inputs:[{start:0,name:"default",type:"tensor"}],attrs:[{tfName:"shape",name:"shape",type:"shape"},{tfName:"dtype",name:"dtype",type:"dtype"}]},{tfOpName:"Placeholder",category:"graph",attrs:[{tfName:"shape",name:"shape",type:"shape"},{tfName:"dtype",name:"dtype",type:"dtype"}]},{tfOpName:"Const",category:"graph"},{tfOpName:"Identity",category:"graph",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"IdentityN",category:"graph",inputs:[{start:0,end:0,name:"x",type:"tensors"}]},{tfOpName:"Snapshot",category:"graph",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"Rank",category:"graph",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"Size",category:"graph",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"Shape",category:"graph",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"ShapeN",category:"graph",inputs:[{start:0,end:0,name:"x",type:"tensors"}]},{tfOpName:"Print",category:"graph",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"data",type:"tensors"}],attrs:[{tfName:"message",name:"message",type:"string"},{tfName:"first_n",name:"firstN",type:"number",notSupported:!0},{tfName:"summarize",name:"summarize",type:"number",defaultValue:3}]},{tfOpName:"NoOp",category:"graph",inputs:[]},{tfOpName:"StopGradient",category:"graph",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"FakeQuantWithMinMaxVars",category:"graph",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"min",name:"min",type:"number"},{tfName:"max",name:"max",type:"number"}]}],WN={};f(WN,{json:()=>VN});var VN=[{tfOpName:"HashTable",category:"hash_table",inputs:[],attrs:[{tfName:"shared_name",name:"sharedName",type:"string"},{tfName:"use_node_name_sharing",name:"useNodeNameSharing",type:"bool"},{tfName:"key_dtype",name:"keyDType",type:"dtype"},{tfName:"value_dtype",name:"valueDType",type:"dtype"}]},{tfOpName:"HashTableV2",category:"hash_table",inputs:[],attrs:[{tfName:"shared_name",name:"sharedName",type:"string"},{tfName:"use_node_name_sharing",name:"useNodeNameSharing",type:"bool"},{tfName:"key_dtype",name:"keyDType",type:"dtype"},{tfName:"value_dtype",name:"valueDType",type:"dtype"}]},{tfOpName:"LookupTableImport",category:"hash_table",inputs:[{start:0,name:"tableHandle",type:"tensor"},{start:1,name:"keys",type:"tensor"},{start:2,name:"values",type:"tensor"}],attrs:[{tfName:"Tin",name:"tIn",type:"dtype",notSupported:!0},{tfName:"Tout",name:"tOut",type:"dtype",notSupported:!0}]},{tfOpName:"LookupTableImportV2",category:"hash_table",inputs:[{start:0,name:"tableHandle",type:"tensor"},{start:1,name:"keys",type:"tensor"},{start:2,name:"values",type:"tensor"}],attrs:[{tfName:"Tin",name:"tIn",type:"dtype",notSupported:!0},{tfName:"Tout",name:"tOut",type:"dtype",notSupported:!0}]},{tfOpName:"LookupTableFind",category:"hash_table",inputs:[{start:0,name:"tableHandle",type:"tensor"},{start:1,name:"keys",type:"tensor"},{start:2,name:"defaultValue",type:"tensor"}],attrs:[{tfName:"Tin",name:"tIn",type:"dtype",notSupported:!0},{tfName:"Tout",name:"tOut",type:"dtype",notSupported:!0}]},{tfOpName:"LookupTableFindV2",category:"hash_table",inputs:[{start:0,name:"tableHandle",type:"tensor"},{start:1,name:"keys",type:"tensor"},{start:2,name:"defaultValue",type:"tensor"}],attrs:[{tfName:"Tin",name:"tIn",type:"dtype",notSupported:!0},{tfName:"Tout",name:"tOut",type:"dtype",notSupported:!0}]},{tfOpName:"LookupTableSize",category:"hash_table",inputs:[{start:0,name:"tableHandle",type:"tensor"}]},{tfOpName:"LookupTableSizeV2",category:"hash_table",inputs:[{start:0,name:"tableHandle",type:"tensor"}]},{tfOpName:"InitializeTable",category:"hash_table",inputs:[{start:0,name:"tableHandle",type:"tensor"},{start:1,name:"keys",type:"tensor"},{start:2,name:"values",type:"tensor"}]},{tfOpName:"InitializeTableV2",category:"hash_table",inputs:[{start:0,name:"tableHandle",type:"tensor"},{start:1,name:"keys",type:"tensor"},{start:2,name:"values",type:"tensor"}]}],UN={};f(UN,{json:()=>GN});var GN=[{tfOpName:"ResizeBilinear",category:"image",inputs:[{start:0,name:"images",type:"tensor"},{start:1,name:"size",type:"number[]"}],attrs:[{tfName:"align_corners",name:"alignCorners",type:"bool"},{tfName:"half_pixel_centers",name:"halfPixelCenters",type:"bool"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"ResizeNearestNeighbor",category:"image",inputs:[{start:0,name:"images",type:"tensor"},{start:1,name:"size",type:"number[]"}],attrs:[{tfName:"align_corners",name:"alignCorners",type:"bool"},{tfName:"half_pixel_centers",name:"halfPixelCenters",type:"bool"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"CropAndResize",category:"image",inputs:[{start:0,name:"image",type:"tensor"},{start:1,name:"boxes",type:"tensor"},{start:2,name:"boxInd",type:"tensor"},{start:3,name:"cropSize",type:"number[]"}],attrs:[{tfName:"method",name:"method",type:"string"},{tfName:"extrapolation_value",name:"extrapolationValue",type:"number"}]},{tfOpName:"ImageProjectiveTransformV3",category:"image",inputs:[{start:0,name:"images",type:"tensor"},{start:1,name:"transforms",type:"tensor"},{start:2,name:"outputShape",type:"number[]"},{start:3,name:"fillValue",type:"number"}],attrs:[{tfName:"interpolation",name:"interpolation",type:"string"},{tfName:"fill_mode",name:"fillMode",type:"string"}]}],HN={};f(HN,{json:()=>jN});var jN=[{tfOpName:"Equal",category:"logical",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"NotEqual",category:"logical",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Greater",category:"logical",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"GreaterEqual",category:"logical",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Less",category:"logical",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"LessEqual",category:"logical",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"LogicalAnd",category:"logical",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"LogicalNot",category:"logical",inputs:[{start:0,name:"a",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"LogicalOr",category:"logical",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Select",category:"logical",inputs:[{start:0,name:"condition",type:"tensor"},{start:1,name:"a",type:"tensor"},{start:2,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"SelectV2",category:"logical",inputs:[{start:0,name:"condition",type:"tensor"},{start:1,name:"a",type:"tensor"},{start:2,name:"b",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"BitwiseAnd",category:"logical",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"y",type:"tensor"}]}],qN={};f(qN,{json:()=>KN});var KN=[{tfOpName:"_FusedMatMul",category:"matrices",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"},{start:2,end:0,name:"args",type:"tensors"}],attrs:[{tfName:"num_args",name:"numArgs",type:"number"},{tfName:"fused_ops",name:"fusedOps",type:"string[]",defaultValue:[]},{tfName:"epsilon",name:"epsilon",type:"number",defaultValue:1e-4},{tfName:"transpose_a",name:"transposeA",type:"bool",defaultValue:!1},{tfName:"transpose_b",name:"transposeB",type:"bool",defaultValue:!1},{tfName:"leakyrelu_alpha",name:"leakyreluAlpha",type:"number",defaultValue:.2},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"MatMul",category:"matrices",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"transpose_a",name:"transposeA",type:"bool",defaultValue:!1},{tfName:"transpose_b",name:"transposeB",type:"bool",defaultValue:!1},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"BatchMatMul",category:"matrices",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"adj_x",name:"transposeA",type:"bool",defaultValue:!1},{tfName:"adj_y",name:"transposeB",type:"bool",defaultValue:!1},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"BatchMatMulV2",category:"matrices",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"b",type:"tensor"}],attrs:[{tfName:"adj_x",name:"transposeA",type:"bool",defaultValue:!1},{tfName:"adj_y",name:"transposeB",type:"bool",defaultValue:!1},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Transpose",category:"matrices",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"perm",type:"number[]"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Einsum",category:"matrices",inputs:[{start:0,end:0,name:"tensors",type:"tensors"}],attrs:[{tfName:"equation",name:"equation",type:"string"},{tfName:"N",name:"n",type:"number",defaultValue:2},{tfName:"T",name:"dtype",type:"dtype"}]},{tfOpName:"MatrixBandPart",category:"matrices",inputs:[{start:0,name:"a",type:"tensor"},{start:1,name:"numLower",type:"tensor"},{start:1,name:"numUpper",type:"tensor"}]}],XN={};f(XN,{json:()=>ZN});var ZN=[{tfOpName:"EuclideanNorm",category:"normalization",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number[]"}],attrs:[{tfName:"keep_dims",name:"keepDims",type:"bool",defaultValue:!1}]},{tfOpName:"FusedBatchNorm",category:"normalization",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"scale",type:"tensor"},{start:2,name:"offset",type:"tensor"},{start:3,name:"mean",type:"tensor"},{start:4,name:"variance",type:"tensor"}],attrs:[{tfName:"epsilon",name:"epsilon",type:"number",defaultValue:.001},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0}]},{tfOpName:"FusedBatchNormV2",category:"normalization",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"scale",type:"tensor"},{start:2,name:"offset",type:"tensor"},{start:3,name:"mean",type:"tensor"},{start:4,name:"variance",type:"tensor"}],attrs:[{tfName:"epsilon",name:"epsilon",type:"number",defaultValue:.001},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0}]},{tfOpName:"FusedBatchNormV3",category:"normalization",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"scale",type:"tensor"},{start:2,name:"offset",type:"tensor"},{start:3,name:"mean",type:"tensor"},{start:4,name:"variance",type:"tensor"}],attrs:[{tfName:"epsilon",name:"epsilon",type:"number",defaultValue:.001},{tfName:"data_format",name:"dataFormat",type:"string",notSupported:!0}]},{tfOpName:"LRN",category:"normalization",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"depth_radius",name:"radius",type:"number",defaultValue:5},{tfName:"bias",name:"bias",type:"number",defaultValue:1},{tfName:"alpha",name:"alpha",type:"number",defaultValue:1},{tfName:"beta",name:"beta",type:"number",defaultValue:.5}]},{tfOpName:"Softmax",category:"normalization",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"LogSoftmax",category:"normalization",inputs:[{start:0,name:"x",type:"tensor"}]}],YN={};f(YN,{json:()=>JN});var JN=[{tfOpName:"Bincount",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"size",type:"number"},{start:2,name:"weights",type:"tensor"}]},{tfOpName:"DenseBincount",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"size",type:"number"},{start:2,name:"weights",type:"tensor"}],attrs:[{tfName:"binary_output",name:"binaryOutput",type:"bool"}]},{tfOpName:"Max",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number[]"}],attrs:[{tfName:"keep_dims",name:"keepDims",type:"bool"}]},{tfOpName:"Mean",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number[]"}],attrs:[{tfName:"keep_dims",name:"keepDims",type:"bool"}]},{tfOpName:"Min",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number[]"}],attrs:[{tfName:"keep_dims",name:"keepDims",type:"bool"}]},{tfOpName:"Sum",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number[]"}],attrs:[{tfName:"keep_dims",name:"keepDims",type:"bool"}]},{tfOpName:"All",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number[]"}],attrs:[{tfName:"keep_dims",name:"keepDims",type:"bool"}]},{tfOpName:"Any",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number[]"}],attrs:[{tfName:"keep_dims",name:"keepDims",type:"bool"}]},{tfOpName:"ArgMax",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number"}]},{tfOpName:"ArgMin",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number"}]},{tfOpName:"Prod",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number[]"}],attrs:[{tfName:"keep_dims",name:"keepDims",type:"bool"},{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"Cumprod",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number"}],attrs:[{tfName:"exclusive",name:"exclusive",type:"bool"},{tfName:"reverse",name:"reverse",type:"bool"}]},{tfOpName:"Cumsum",category:"reduction",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number"}],attrs:[{tfName:"exclusive",name:"exclusive",type:"bool"},{tfName:"reverse",name:"reverse",type:"bool"}]}],QN={};f(QN,{json:()=>eT});var eT=[{tfOpName:"ConcatV2",category:"slice_join",inputs:[{start:0,end:-1,name:"tensors",type:"tensors"},{start:-1,name:"axis",type:"number"}],attrs:[{tfName:"N",name:"n",type:"number",defaultValue:2}]},{tfOpName:"Concat",category:"slice_join",inputs:[{start:1,end:0,name:"tensors",type:"tensors"},{start:0,name:"axis",type:"number"}],attrs:[{tfName:"N",name:"n",type:"number",defaultValue:2}]},{tfOpName:"GatherV2",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"indices",type:"tensor"},{start:2,name:"axis",type:"number",defaultValue:0}],attrs:[{tfName:"batch_dims",name:"batchDims",type:"number",defaultValue:0}]},{tfOpName:"Gather",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"indices",type:"tensor"}],attrs:[{tfName:"validate_indices",name:"validateIndices",type:"bool",notSupported:!0}]},{tfOpName:"Reverse",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"dims",type:"bool[]"}]},{tfOpName:"ReverseV2",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number[]"}]},{tfOpName:"Slice",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"begin",type:"number[]"},{start:2,name:"size",type:"number[]"}]},{tfOpName:"StridedSlice",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"begin",type:"number[]"},{start:2,name:"end",type:"number[]"},{start:3,name:"strides",type:"number[]"}],attrs:[{tfName:"begin_mask",name:"beginMask",type:"number",defaultValue:0},{tfName:"end_mask",name:"endMask",type:"number",defaultValue:0},{tfName:"new_axis_mask",name:"newAxisMask",type:"number",defaultValue:0},{tfName:"ellipsis_mask",name:"ellipsisMask",type:"number",defaultValue:0},{tfName:"shrink_axis_mask",name:"shrinkAxisMask",type:"number",defaultValue:0}]},{tfOpName:"Pack",category:"slice_join",inputs:[{start:0,end:0,name:"tensors",type:"tensors"}],attrs:[{tfName:"axis",name:"axis",type:"number",defaultValue:0}]},{tfOpName:"Unpack",category:"slice_join",inputs:[{start:0,name:"tensor",type:"tensor"}],attrs:[{tfName:"axis",name:"axis",type:"number",defaultValue:0},{tfName:"num",name:"num",type:"number",defaultValue:0,notSupported:!0}]},{tfOpName:"Tile",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"reps",type:"number[]"}]},{tfOpName:"Split",category:"slice_join",inputs:[{start:0,name:"axis",type:"number",defaultValue:0},{start:1,name:"x",type:"tensor"}],attrs:[{tfName:"num_split",name:"numOrSizeSplits",type:"number",defaultValue:1}]},{tfOpName:"SplitV",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"numOrSizeSplits",type:"number[]"},{start:2,name:"axis",type:"number",defaultValue:0}]},{tfOpName:"ScatterNd",category:"slice_join",inputs:[{start:0,name:"indices",type:"tensor"},{start:1,name:"values",type:"tensor"},{start:2,name:"shape",type:"number[]"}]},{tfOpName:"GatherNd",category:"slice_join",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"indices",type:"tensor"}]},{tfOpName:"SparseToDense",category:"slice_join",inputs:[{start:0,name:"sparseIndices",type:"tensor"},{start:1,name:"outputShape",type:"number[]"},{start:2,name:"sparseValues",type:"tensor"},{start:3,name:"defaultValue",type:"tensor"}],attrs:[{tfName:"validate_indices",name:"validateIndices",type:"bool",defaultValue:!1,notSupported:!0}]},{tfOpName:"TensorScatterUpdate",category:"slice_join",inputs:[{start:0,name:"tensor",type:"tensor"},{start:1,name:"indices",type:"tensor"},{start:2,name:"values",type:"tensor"}]}],tT={};f(tT,{json:()=>nT});var nT=[{tfOpName:"SparseFillEmptyRows",category:"sparse",inputs:[{start:0,name:"indices",type:"tensor"},{start:1,name:"values",type:"tensor"},{start:2,name:"denseShape",type:"tensor"},{start:3,name:"defaultValue",type:"tensor"}]},{tfOpName:"SparseReshape",category:"sparse",inputs:[{start:0,name:"inputIndices",type:"tensor"},{start:1,name:"inputShape",type:"tensor"},{start:2,name:"newShape",type:"tensor"}],attrs:[{tfName:"T",name:"dtype",type:"dtype",notSupported:!0}]},{tfOpName:"SparseSegmentMean",category:"sparse",inputs:[{start:0,name:"data",type:"tensor"},{start:1,name:"indices",type:"tensor"},{start:2,name:"segmentIds",type:"tensor"}]},{tfOpName:"SparseSegmentSum",category:"sparse",inputs:[{start:0,name:"data",type:"tensor"},{start:1,name:"indices",type:"tensor"},{start:2,name:"segmentIds",type:"tensor"}]}],rT={};f(rT,{json:()=>aT});var aT=[{tfOpName:"FFT",category:"spectral",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"IFFT",category:"spectral",inputs:[{start:0,name:"x",type:"tensor"}]},{tfOpName:"RFFT",category:"spectral",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"fft_length",type:"number",notSupported:!0}]},{tfOpName:"IRFFT",category:"spectral",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"fft_length",type:"number",notSupported:!0}]}],sT={};f(sT,{json:()=>iT});var iT=[{tfOpName:"StaticRegexReplace",category:"string",inputs:[{start:0,name:"input",type:"tensor"}],attrs:[{tfName:"pattern",name:"pattern",type:"string"},{tfName:"rewrite",name:"rewrite",type:"string"},{tfName:"replace_global",name:"replaceGlobal",type:"bool"}]},{tfOpName:"StringNGrams",category:"string",inputs:[{start:0,name:"data",type:"tensor"},{start:1,name:"dataSplits",type:"tensor"}],attrs:[{tfName:"separator",name:"separator",type:"string"},{tfName:"ngram_widths",name:"nGramWidths",type:"number[]"},{tfName:"left_pad",name:"leftPad",type:"string"},{tfName:"right_pad",name:"rightPad",type:"string"},{tfName:"pad_width",name:"padWidth",type:"number"},{tfName:"preserve_short_sequences",name:"preserveShortSequences",type:"bool"}],outputs:["ngrams","ngrams_splits"]},{tfOpName:"StringSplit",category:"string",inputs:[{start:0,name:"input",type:"tensor"},{start:1,name:"delimiter",type:"tensor"}],attrs:[{tfName:"skip_empty",name:"skipEmpty",type:"bool"}],outputs:["indices","values","shape"]},{tfOpName:"StringToHashBucketFast",category:"string",inputs:[{start:0,name:"input",type:"tensor"}],attrs:[{tfName:"num_buckets",name:"numBuckets",type:"number"}]}],oT={};f(oT,{json:()=>lT});var lT=[{tfOpName:"Cast",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"SrcT",name:"sdtype",type:"dtype",notSupported:!0},{tfName:"DstT",name:"dtype",type:"dtype"}]},{tfOpName:"ExpandDims",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"axis",type:"number"}]},{tfOpName:"MirrorPad",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"padding",type:"number[]"}],attrs:[{tfName:"mode",name:"mode",type:"string"}]},{tfOpName:"Pad",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"padding",type:"number[]"}],attrs:[{tfName:"constant_value",name:"constantValue",type:"number",defaultValue:0}]},{tfOpName:"PadV2",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"padding",type:"number[]"},{start:2,name:"constantValue",type:"number",defaultValue:0}]},{tfOpName:"Reshape",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"shape",type:"number[]"}]},{tfOpName:"EnsureShape",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"shape",type:"number[]"}]},{tfOpName:"Squeeze",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"axis",tfDeprecatedName:"squeeze_dims",name:"axis",type:"number[]"}]},{tfOpName:"SpaceToBatchND",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"blockShape",type:"number[]"},{start:2,name:"paddings",type:"number[]"}]},{tfOpName:"BatchToSpaceND",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"blockShape",type:"number[]"},{start:2,name:"crops",type:"number[]"}]},{tfOpName:"DepthToSpace",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"}],attrs:[{tfName:"block_size",name:"blockSize",type:"number"},{tfName:"data_format",name:"dataFormat",type:"string"}]},{tfOpName:"BroadcastTo",category:"transformation",inputs:[{start:0,name:"x",type:"tensor"},{start:1,name:"shape",type:"number[]"}],attrs:[]},{tfOpName:"BroadcastArgs",category:"transformation",inputs:[{start:0,name:"s0",type:"tensor"},{start:1,name:"s1",type:"tensor"}],attrs:[]}],uT=class{static get Instance(){return this._instance||(this._instance=new this)}constructor(){let e=[].concat(...[_N,TN,EN,$N,FN,MN,LN,zN,WN,UN,HN,qN,XN,YN,QN,tT,rT,sT,oT].map(e=>e.json));this.opMappers=e.reduce((e,t)=>(e[t.tfOpName]=t,e),{})}transformGraph(e,t={}){let n=e.node,r=[],a=[],s=[],i=n.reduce((e,t)=>(e[t.name]=this.mapNode(t),t.op.startsWith("Placeholder")?r.push(e[t.name]):"Const"===t.op?a.push(e[t.name]):(null==t.input||0===t.input.length)&&s.push(e[t.name]),e),{}),o=[],l=[],u={},h={};null!=t&&(u=this.mapSignatureEntries(t.inputs),h=this.mapSignatureEntries(t.outputs));let d=Object.keys(i);d.forEach(e=>{let t=i[e];t.inputNames.forEach((e,n)=>{let[r,,a]=vN(e),s=i[r];if(null!=s.outputs){let e=s.outputs.indexOf(a);if(-1!==e){let a=`${r}:${e}`;t.inputNames[n]=a}}t.inputs.push(s),s.children.push(t)})}),0===Object.keys(h).length?d.forEach(e=>{let t=i[e];0===t.children.length&&l.push(t)}):Object.keys(h).forEach(e=>{let[t]=vN(e),n=i[t];null!=n&&(n.signatureKey=h[e],l.push(n))}),Object.keys(u).length>0?Object.keys(u).forEach(e=>{let[t]=vN(e),n=i[t];n&&(n.signatureKey=u[e],o.push(n))}):o=r;let p={};null!=e.library&&null!=e.library.function&&(p=e.library.function.reduce((e,t)=>(e[t.signature.name]=this.mapFunction(t),e),{}));let c={nodes:i,inputs:o,outputs:l,weights:a,placeholders:r,signature:t,functions:p};return s.length>0&&(c.initNodes=s),c}mapSignatureEntries(e){return Object.keys(e||{}).reduce((t,n)=>(t[e[n].name]=n,t),{})}mapNode(e){let t=mN(e.op)||this.opMappers[e.op]||{};null==e.attr&&(e.attr={});let n={name:e.name,op:e.op,category:t.category,inputNames:(e.input||[]).map(e=>e.startsWith("^")?e.slice(1):e),inputs:[],children:[],inputParams:{},attrParams:{},rawAttrs:e.attr,outputs:t.outputs};return null!=t.inputs&&(n.inputParams=t.inputs.reduce((e,t)=>(e[t.name]={type:t.type,inputIndexStart:t.start,inputIndexEnd:t.end},e),{})),null!=t.attrs&&(n.attrParams=t.attrs.reduce((t,n)=>{let r,a=n.type;switch(n.type){case"string":r=dT(e.attr,n.tfName,n.defaultValue),void 0===r&&n.tfDeprecatedName&&(r=dT(e.attr,n.tfDeprecatedName,n.defaultValue));break;case"string[]":r=wT(e.attr,n.tfName,n.defaultValue),void 0===r&&n.tfDeprecatedName&&(r=wT(e.attr,n.tfDeprecatedName,n.defaultValue));break;case"number":r=cT(e.attr,n.tfName,n.defaultValue||0),void 0===r&&n.tfDeprecatedName&&(r=cT(e.attr,n.tfDeprecatedName,n.defaultValue));break;case"number[]":r=vT(e.attr,n.tfName,n.defaultValue),void 0===r&&n.tfDeprecatedName&&(r=vT(e.attr,n.tfDeprecatedName,n.defaultValue));break;case"bool":r=pT(e.attr,n.tfName,n.defaultValue),void 0===r&&n.tfDeprecatedName&&(r=pT(e.attr,n.tfDeprecatedName,n.defaultValue));break;case"bool[]":r=IT(e.attr,n.tfName,n.defaultValue),void 0===r&&n.tfDeprecatedName&&(r=IT(e.attr,n.tfDeprecatedName,n.defaultValue));break;case"shape":r=xT(e.attr,n.tfName,n.defaultValue),void 0===r&&n.tfDeprecatedName&&(r=xT(e.attr,n.tfDeprecatedName,n.defaultValue));break;case"shape[]":r=kT(e.attr,n.tfName,n.defaultValue),void 0===r&&n.tfDeprecatedName&&(r=kT(e.attr,n.tfDeprecatedName,n.defaultValue));break;case"dtype":r=gT(e.attr,n.tfName,n.defaultValue),void 0===r&&n.tfDeprecatedName&&(r=gT(e.attr,n.tfDeprecatedName,n.defaultValue));break;case"dtype[]":r=yT(e.attr,n.tfName,n.defaultValue),void 0===r&&n.tfDeprecatedName&&(r=yT(e.attr,n.tfDeprecatedName,n.defaultValue));break;case"func":r=mT(e.attr,n.tfName,n.defaultValue),void 0===r&&n.tfDeprecatedName&&(r=mT(e.attr,n.tfDeprecatedName,n.defaultValue));break;case"tensor":case"tensors":break;default:throw new Error(`Unsupported param type: ${n.type} for op: ${e.op}`)}return t[n.name]={value:r,type:a},t},{})),n}mapFunction(e){let t=e.nodeDef,n=[],r={};null!=t&&(r=t.reduce((e,t)=>(e[t.name]=this.mapNode(t),"Const"===t.op&&n.push(e[t.name]),e),{}));let a=[],s=[];e.signature.inputArg.forEach(e=>{let[t]=vN(e.name),n={name:t,op:"Placeholder",inputs:[],inputNames:[],category:"graph",inputParams:{},attrParams:{dtype:{value:fT(e.type),type:"dtype"}},children:[]};n.signatureKey=e.name,a.push(n),r[t]=n}),Object.keys(r).forEach(e=>{let t=r[e];t.inputNames.forEach((e,n)=>{let[a,,s]=vN(e),i=r[a];if(null!=i.outputs){let e=i.outputs.indexOf(s);if(-1!==e){let r=`${a}:${e}`;t.inputNames[n]=r}}t.inputs.push(i),i.children.push(t)})});let i=e.ret;e.signature.outputArg.forEach(e=>{let[t,n]=vN(i[e.name]),a=r[t];null!=a&&(a.defaultOutput=n,s.push(a))});let o=this.mapArgsToSignature(e);return{nodes:r,inputs:a,outputs:s,weights:n,placeholders:[],signature:o}}mapArgsToSignature(e){return{methodName:e.signature.name,inputs:e.signature.inputArg.reduce((e,t)=>(e[t.name]=this.mapArgToTensorInfo(t),e),{}),outputs:e.signature.outputArg.reduce((t,n)=>(t[n.name]=this.mapArgToTensorInfo(n,e.ret),t),{})}}mapArgToTensorInfo(e,t){let n=e.name;return null!=t&&(n=t[n]),{name:n,dtype:e.type}}};function hT(e,t){let n=Array.isArray(e)?String.fromCharCode.apply(null,e):function(e){let t=Le().global;if(void 0!==t.atob)return t.atob(e);if("undefined"!=typeof Buffer)return new Buffer(e,"base64").toString();throw new Error("Unable to decode base64 in this environment. Missing built-in atob() or Buffer()")}(e);return t?n:n.toLowerCase()}function dT(e,t,n,r=!1){let a=e[t];return null!=a?hT(a.s,r):n}function pT(e,t,n){let r=e[t];return r?r.b:n}function cT(e,t,n){let r=e[t]||{},a=null!=r.i?r.i:null!=r.f?r.f:n;return"number"==typeof a?a:parseInt(a,10)}function fT(e){switch("string"==typeof e&&(e=hN[e]),e){case hN.DT_FLOAT:case hN.DT_HALF:return"float32";case hN.DT_INT32:case hN.DT_INT64:case hN.DT_INT8:case hN.DT_UINT8:return"int32";case hN.DT_BOOL:return"bool";case hN.DT_DOUBLE:return"float32";case hN.DT_STRING:return"string";case hN.DT_COMPLEX64:case hN.DT_COMPLEX128:return"complex64";default:return null}}function mT(e,t,n){let r=e[t];return r&&r.func?r.func.name:n}function gT(e,t,n){let r=e[t];return r&&r.type?fT(r.type):n}function yT(e,t,n){let r=e[t];return r&&r.list&&r.list.type?r.list.type.map(e=>fT(e)):n}function bT(e){if(!e.unknownRank)return null!=e.dim?e.dim.map(e=>"number"==typeof e.size?e.size:parseInt(e.size,10)):[]}function xT(e,t,n){let r=e[t];return r&&r.shape?bT(r.shape):n}function vT(e,t,n){let r=e[t];return r?((r.list.f&&r.list.f.length?r.list.f:r.list.i)||[]).map(e=>"number"==typeof e?e:parseInt(e,10)):n}function wT(e,t,n,r=!1){let a=e[t];return a&&a.list&&a.list.s?a.list.s.map(e=>hT(e,r)):n}function kT(e,t,n){let r=e[t];return r&&r.list&&r.list.shape?r.list.shape.map(e=>bT(e)):n}function IT(e,t,n){let r=e[t];return r&&r.list&&r.list.b?r.list.b:n}var ST=class{constructor(e,t,n){this.node=e,this.tensorMap=t,this.context=n,this.inputs=[],this.attrs={},this.inputs=e.inputNames.map(e=>this.getInput(e)),null!=e.rawAttrs&&(this.attrs=Object.keys(e.rawAttrs).reduce((e,t)=>(e[t]=this.getAttr(t),e),{}))}getInput(e){return bN(e,this.tensorMap,this.context)}getAttr(e,t){let n=this.node.rawAttrs[e];if(null!=n.tensor)return bN(e,this.tensorMap,this.context);if(null!=n.i||null!=n.f)return cT(this.node.rawAttrs,e,t);if(null!=n.s)return dT(this.node.rawAttrs,e,t);if(null!=n.b)return pT(this.node.rawAttrs,e,t);if(null!=n.shape)return xT(this.node.rawAttrs,e,t);if(null!=n.type)return gT(this.node.rawAttrs,e,t);if(null!=n.list){if(null!=n.list.i||null!=n.list.f)return vT(this.node.rawAttrs,e,t);if(null!=n.list.s)return wT(this.node.rawAttrs,e,t);if(null!=n.list.shape)return kT(this.node.rawAttrs,e,t);if(null!=n.list.b)return IT(this.node.rawAttrs,e,t);if(null!=n.list.type)return yT(this.node.rawAttrs,e,t)}return t}},_T={};f(_T,{OP_SCOPE_SUFFIX:()=>Ps,abs:()=>bo,acos:()=>xo,acosh:()=>vo,add:()=>fo,addN:()=>wo,all:()=>ko,any:()=>Io,argMax:()=>So,argMin:()=>_o,asin:()=>No,asinh:()=>To,atan:()=>Co,atan2:()=>Eo,atanh:()=>Ao,avgPool:()=>qo,avgPool3d:()=>Ko,basicLSTMCell:()=>el,batchNorm:()=>rl,batchNorm2d:()=>al,batchNorm3d:()=>sl,batchNorm4d:()=>il,batchToSpaceND:()=>tl,bincount:()=>ol,bitwiseAnd:()=>ll,booleanMaskAsync:()=>Yd,broadcastArgs:()=>ul,broadcastTo:()=>hl,buffer:()=>uo,cast:()=>ho,ceil:()=>dl,clipByValue:()=>cl,clone:()=>po,complex:()=>Bs,concat:()=>Xo,concat1d:()=>fl,concat2d:()=>ml,concat3d:()=>gl,concat4d:()=>yl,conv1d:()=>xl,conv2d:()=>bl,conv2dTranspose:()=>wl,conv3d:()=>kl,conv3dTranspose:()=>Sl,cos:()=>_l,cosh:()=>Nl,cosineWindow:()=>sp,cumprod:()=>Tl,cumsum:()=>Cl,denseBincount:()=>El,depthToSpace:()=>Al,depthwiseConv2d:()=>$l,diag:()=>Rl,dilation2d:()=>Fl,div:()=>go,divNoNan:()=>Wl,dot:()=>Vl,dropout:()=>rp,einsum:()=>Ul,elu:()=>Gl,enclosingPowerOfTwo:()=>ap,ensureShape:()=>Hl,equal:()=>Pl,erf:()=>jl,euclideanNorm:()=>hu,exp:()=>du,expandDims:()=>pu,expm1:()=>cu,eye:()=>mu,fft:()=>vd,fill:()=>pl,floor:()=>gu,floorDiv:()=>mo,fused:()=>op,gather:()=>yu,gatherND:()=>np,greater:()=>bu,greaterEqual:()=>xu,ifft:()=>wd,imag:()=>vu,image:()=>yc,inTopKAsync:()=>ip,irfft:()=>kd,isFinite:()=>wu,isInf:()=>ku,isNaN:()=>Iu,leakyRelu:()=>Su,less:()=>_u,lessEqual:()=>Nu,linalg:()=>bc,linspace:()=>Tu,localResponseNormalization:()=>Cu,log:()=>Eu,log1p:()=>Au,logSigmoid:()=>Bu,logSoftmax:()=>Vu,logSumExp:()=>Uu,logicalAnd:()=>Gu,logicalNot:()=>Hu,logicalOr:()=>ju,logicalXor:()=>qu,losses:()=>xc,lowerBound:()=>Zu,matMul:()=>Zo,max:()=>tu,maxPool:()=>Yu,maxPool3d:()=>Ju,maxPoolWithArgmax:()=>Qu,maximum:()=>eh,mean:()=>th,meshgrid:()=>ah,min:()=>nu,minimum:()=>sh,mirrorPad:()=>ih,mod:()=>oh,moments:()=>lh,movingAverage:()=>Qd,mul:()=>yo,multiRNNCell:()=>uh,multinomial:()=>hh,neg:()=>Pu,norm:()=>uu,notEqual:()=>dh,oneHot:()=>ph,ones:()=>rh,onesLike:()=>ch,op:()=>zs,outerProduct:()=>fh,pad:()=>mh,pad1d:()=>gh,pad2d:()=>yh,pad3d:()=>bh,pad4d:()=>xh,pool:()=>wh,pow:()=>ru,prelu:()=>kh,print:()=>co,prod:()=>Ih,raggedGather:()=>Sh,raggedRange:()=>_h,raggedTensorToTensor:()=>Nh,rand:()=>Th,randomGamma:()=>jh,randomNormal:()=>qh,randomStandardNormal:()=>Kh,randomUniform:()=>Xh,randomUniformInt:()=>Zh,range:()=>Yh,real:()=>Jh,reciprocal:()=>Qh,relu:()=>ed,relu6:()=>td,reshape:()=>jo,reverse:()=>nd,reverse1d:()=>rd,reverse2d:()=>ad,reverse3d:()=>sd,reverse4d:()=>id,rfft:()=>Sd,round:()=>od,rsqrt:()=>ld,scalar:()=>au,scatterND:()=>ep,searchSorted:()=>Xu,selu:()=>ud,separableConv2d:()=>hd,setdiff1dAsync:()=>dd,sigmoid:()=>Yo,sign:()=>pd,signal:()=>gc,sin:()=>cd,sinh:()=>fd,slice:()=>Jo,slice1d:()=>md,slice2d:()=>gd,slice3d:()=>yd,slice4d:()=>bd,softmax:()=>xd,softplus:()=>zu,spaceToBatchND:()=>vh,sparse:()=>vc,sparseToDense:()=>tp,spectral:()=>mc,split:()=>Id,sqrt:()=>su,square:()=>iu,squaredDifference:()=>_d,squeeze:()=>Nd,stack:()=>Td,step:()=>Cd,stridedSlice:()=>Ed,string:()=>wc,sub:()=>Wu,sum:()=>ou,tan:()=>Ad,tanh:()=>Qo,tensor:()=>Vs,tensor1d:()=>$d,tensor2d:()=>Rd,tensor3d:()=>Fd,tensor4d:()=>Dd,tensor5d:()=>Md,tensor6d:()=>Od,tensorScatterUpdate:()=>Wd,tile:()=>fu,topk:()=>Vd,transpose:()=>Jd,truncatedNormal:()=>Ud,unique:()=>Gd,unsortedSegmentSum:()=>Hd,unstack:()=>jd,upperBound:()=>qd,variable:()=>Kd,where:()=>zl,whereAsync:()=>Zd,zeros:()=>nh,zerosLike:()=>Bl});function NT(e,t,n=""){if("number"!=typeof e&&"number"!=typeof t){va.assert(e.length===t.length,()=>n+` Shapes ${e} and ${t} must match`);for(let r=0;r<e.length;r++){let a=e[r],s=t[r];va.assert(a<0||s<0||a===s,()=>n+` Shapes ${e} and ${t} must match`)}}}function TT(e){return!("number"==typeof e||e.some(e=>e<0))}function CT(e,t,n){let r=ET(e,n),a=!TT(r);if(a&&0===t.length)throw new Error(`Tried to calculate elements of an empty list with non-fully-defined elementShape: ${r}`);if(a&&t.forEach(e=>{r=ET(e.shape,r)}),!TT(r))throw new Error(`Non-fully-defined elementShape: ${r}`);return r}function ET(e,t){if("number"==typeof e)return t;if("number"==typeof t)return e;if(e.length!==t.length)throw new Error(`Incompatible ranks during merge: ${e} vs. ${t}`);let n=[];for(let r=0;r<e.length;++r){let a=e[r],s=t[r];if(a>=0&&s>=0&&a!==s)throw new Error(`Incompatible shape during merge: ${e} vs. ${t}`);n[r]=a>=0?a:s}return n}var AT=class{constructor(e,t,n,r,a,s,i){this.name=e,this.dtype=t,this.maxSize=n,this.elementShape=r,this.identicalElementShapes=a,this.dynamicSize=s,this.clearAfterRead=i,this.tensors=[],this.closed_=!1,this.idTensor=au(0),ti(this.idTensor)}get id(){return this.idTensor.id}get closed(){return this.closed_}clearAndClose(e){this.tensors.forEach(t=>{(null==e||!e.has(t.tensor.id))&&t.tensor.dispose()}),this.tensors=[],this.closed_=!0,this.idTensor.dispose()}size(){return this.tensors.length}read(e){if(this.closed_)throw new Error(`TensorArray ${this.name} has already been closed.`);if(e<0||e>=this.size())throw new Error(`Tried to read from index ${e}, but array size is: ${this.size()}`);let t=this.tensors[e];if(t.cleared)throw new Error(`TensorArray ${this.name}: Could not read index ${e} twice because it was cleared after a previous read (perhaps try setting clear_after_read = false?).`);return this.clearAfterRead&&(t.cleared=!0),t.read=!0,t.tensor}readMany(e){return e.map(e=>this.read(e))}write(e,t){if(this.closed_)throw new Error(`TensorArray ${this.name} has already been closed.`);if(e<0||!this.dynamicSize&&e>=this.maxSize)throw new Error(`Tried to write to index ${e}, but array is not resizeable and size is: ${this.maxSize}`);let n=this.tensors[e]||{};if(t.dtype!==this.dtype)throw new Error(`TensorArray ${this.name}: Could not write to TensorArray index ${e},\n          because the value dtype is ${t.dtype}, but TensorArray dtype is ${this.dtype}.`);if(0===this.size()&&(null==this.elementShape||0===this.elementShape.length)&&(this.elementShape=t.shape),NT(this.elementShape,t.shape,`TensorArray ${this.name}: Could not write to TensorArray index ${e}.`),n.read)throw new Error(`TensorArray ${this.name}: Could not write to TensorArray index ${e}, because it has already been read.`);if(n.written)throw new Error(`TensorArray ${this.name}: Could not write to TensorArray index ${e}, because it has already been written.`);n.tensor=t,ti(t),n.written=!0,this.tensors[e]=n}writeMany(e,t){if(e.length!==t.length)throw new Error(`TensorArray ${this.name}: could not write multiple tensors,because the index size: ${e.length} is not the same as tensors size: ${t.length}.`);e.forEach((e,n)=>this.write(e,t[n]))}gather(e,t){if(t&&t!==this.dtype)throw new Error(`TensorArray dtype is ${this.dtype} but gather requested dtype ${t}`);if(e)e=e.slice(0,this.size());else{e=[];for(let t=0;t<this.size();t++)e.push(t)}if(0===e.length)return Vs([],[0].concat(this.elementShape));let n=this.readMany(e);return NT(this.elementShape,n[0].shape,"TensorArray shape mismatch: "),Td(n,0)}concat(e){if(e&&e!==this.dtype)throw new Error(`TensorArray dtype is ${this.dtype} but concat requested dtype ${e}`);if(0===this.size())return Vs([],[0].concat(this.elementShape));let t=[];for(let e=0;e<this.size();e++)t.push(e);let n=this.readMany(t);return NT(this.elementShape,n[0].shape,`TensorArray shape mismatch: tensor array shape (${this.elementShape}) vs first tensor shape (${n[0].shape})`),Xo(n,0)}scatter(e,t){if(t.dtype!==this.dtype)throw new Error(`TensorArray dtype is ${this.dtype} but tensor has dtype ${t.dtype}`);if(e.length!==t.shape[0])throw new Error(`Expected len(indices) == tensor.shape[0], but saw: ${e.length} vs. ${t.shape[0]}`);let n=Math.max(...e);if(!this.dynamicSize&&n>=this.maxSize)throw new Error(`Max index must be < array size (${n}  vs. ${this.maxSize})`);this.writeMany(e,jd(t,0))}split(e,t){if(t.dtype!==this.dtype)throw new Error(`TensorArray dtype is ${this.dtype} but tensor has dtype ${t.dtype}`);let n=0,r=e.map(e=>(n+=e,n));if(n!==t.shape[0])throw new Error(`Expected sum of lengths to be equal to\n          tensor.shape[0], but sum of lengths is\n        ${n}, and tensor's shape is: ${t.shape}`);if(!this.dynamicSize&&e.length!==this.maxSize)throw new Error(`TensorArray's size is not equal to the size of lengths (${this.maxSize} vs. ${e.length}), and the TensorArray is not marked as dynamically resizeable`);let a=0===n?0:t.size/n,s=[];Qs(()=>{t=jo(t,[1,n,a]);for(let n=0;n<e.length;++n){let i=[0,0===n?0:r[n-1],0],o=[1,e[n],a];s[n]=jo(Jo(t,i,o),this.elementShape)}return s});let i=[];for(let t=0;t<e.length;t++)i[t]=t;this.writeMany(i,s)}},$T=class e{get id(){return this.idTensor.id}constructor(e,t,n,r=-1){this.tensors=e,this.elementShape=t,this.elementDtype=n,null!=e&&e.forEach(e=>{if(n!==e.dtype)throw new Error(`Invalid data types; op elements ${n}, but list elements ${e.dtype}`);NT(t,e.shape,"TensorList shape mismatch: "),ti(e)}),this.idTensor=au(0),this.maxNumElements=r,ti(this.idTensor)}copy(){return new e([...this.tensors],this.elementShape,this.elementDtype)}clearAndClose(e){this.tensors.forEach(t=>{(null==e||!e.has(t.id))&&t.dispose()}),this.tensors.length=0,this.idTensor.dispose()}size(){return this.tensors.length}stack(e,t,n=-1){if(t!==this.elementDtype)throw new Error(`Invalid data types; op elements ${t}, but list elements ${this.elementDtype}`);if(-1!==n&&this.tensors.length!==n)throw new Error(`Operation expected a list with ${n} elements but got a list with ${this.tensors.length} elements.`);NT(e,this.elementShape,"TensorList shape mismatch: ");let r=CT(this.elementShape,this.tensors,e);return Qs(()=>{let e=this.tensors.map(e=>jo(e,r));return Td(e,0)})}popBack(e,t){if(t!==this.elementDtype)throw new Error(`Invalid data types; op elements ${t}, but list elements ${this.elementDtype}`);if(0===this.size())throw new Error("Trying to pop from an empty list.");let n=CT(this.elementShape,this.tensors,e),r=this.tensors.pop();return r.kept=!1,NT(r.shape,e,"TensorList shape mismatch: "),jo(r,n)}pushBack(e){if(e.dtype!==this.elementDtype)throw new Error(`Invalid data types; op elements ${e.dtype}, but list elements ${this.elementDtype}`);if(NT(e.shape,this.elementShape,"TensorList shape mismatch: "),this.maxNumElements===this.size())throw new Error("Trying to push element into a full list.");ti(e),this.tensors.push(e)}resize(t){if(t<0)throw new Error(`TensorListResize expects size to be non-negative. Got: ${t}`);if(-1!==this.maxNumElements&&t>this.maxNumElements)throw new Error(`TensorListResize input size ${t} is greater maxNumElement ${this.maxNumElements}.`);let n=new e([],this.elementShape,this.elementDtype,this.maxNumElements);n.tensors.length=t;for(let e=0;e<Math.min(this.tensors.length,t);++e)n.tensors[e]=this.tensors[e];return n}getItem(e,t,n){if(n!==this.elementDtype)throw new Error(`Invalid data types; op elements ${n}, but list elements ${this.elementDtype}`);if(e<0||e>this.tensors.length)throw new Error(`Trying to access element ${e} in a list with ${this.tensors.length} elements.`);if(null==this.tensors[e])throw new Error(`element at index ${e} is null.`);NT(this.tensors[e].shape,t,"TensorList shape mismatch: ");let r=CT(this.elementShape,this.tensors,t);return jo(this.tensors[e],r)}setItem(e,t){if(t.dtype!==this.elementDtype)throw new Error(`Invalid data types; op elements ${t.dtype}, but list elements ${this.elementDtype}`);if(e<0||-1!==this.maxNumElements&&e>=this.maxNumElements)throw new Error(`Trying to set element ${e} in a list with max ${this.maxNumElements} elements.`);NT(this.elementShape,t.shape,"TensorList shape mismatch: "),ti(t),null!=this.tensors[e]&&(this.tensors[e].kept=!1),this.tensors[e]=t}gather(e,t,n){if(t!==this.elementDtype)throw new Error(`Invalid data types; op elements ${t}, but list elements ${this.elementDtype}`);NT(this.elementShape,n,"TensorList shape mismatch: "),e=e.slice(0,this.size());let r=CT(this.elementShape,this.tensors,n);return 0===e.length?Vs([],[0].concat(r)):Qs(()=>{let t=e.map(e=>jo(this.tensors[e],r));return Td(t,0)})}concat(e,t){if(e&&e!==this.elementDtype)throw new Error(`TensorList dtype is ${this.elementDtype} but concat requested dtype ${e}`);NT(this.elementShape,t,"TensorList shape mismatch: ");let n=CT(this.elementShape,this.tensors,t);return 0===this.size()?Vs([],[0].concat(n)):Qs(()=>{let e=this.tensors.map(e=>jo(e,n));return Xo(e,0)})}};var RT=async(e,t,n)=>{switch(e.op){case"If":case"StatelessIf":{let r=yN("thenBranch",e,t,n),a=yN("elseBranch",e,t,n),s=yN("cond",e,t,n),i=yN("args",e,t,n);return(await s.data())[0]?n.functionMap[r].executeFunctionAsync(i,n.tensorArrayMap,n.tensorListMap):n.functionMap[a].executeFunctionAsync(i,n.tensorArrayMap,n.tensorListMap)}case"While":case"StatelessWhile":{let r=yN("body",e,t,n),a=yN("cond",e,t,n),s=yN("args",e,t,n),i=await n.functionMap[a].executeFunctionAsync(s,n.tensorArrayMap,n.tensorListMap),o=s.map(e=>e.id),l=await i[0].data();i.forEach(e=>{!e.kept&&-1===o.indexOf(e.id)&&e.dispose()});let u=s;for(;l[0];){let e=u;u=await n.functionMap[r].executeFunctionAsync(u,n.tensorArrayMap,n.tensorListMap);let t=u.map(e=>e.id);e.forEach(e=>{!e.kept&&-1===o.indexOf(e.id)&&-1===t.indexOf(e.id)&&e.dispose()});let s=await n.functionMap[a].executeFunctionAsync(u,n.tensorArrayMap,n.tensorListMap);l=await s[0].data(),s.forEach(e=>{!e.kept&&-1===o.indexOf(e.id)&&-1===t.indexOf(e.id)&&e.dispose()})}return u}case"LoopCond":return[SN(yN("pred",e,t,n))];case"Switch":{let r=yN("pred",e,t,n),a=yN("data",e,t,n);return a.kept||(a=SN(a)),(await r.data())[0]?[void 0,a]:[a,void 0]}case"Merge":{let r=e.inputNames.find(e=>void 0!==bN(e,t,n));if(r){return[SN(bN(r,t,n))]}return}case"Enter":{let r=yN("frameName",e,t,n),a=yN("tensor",e,t,n);return n.enterFrame(r),[SN(a)]}case"Exit":{let r=yN("tensor",e,t,n);return n.exitFrame(),[SN(r)]}case"NextIteration":{let r=yN("tensor",e,t,n);return n.nextIteration(),[SN(r)]}case"TensorArrayV3":{let r=yN("size",e,t,n),a=yN("dtype",e,t,n),s=yN("elementShape",e,t,n),i=yN("dynamicSize",e,t,n),o=yN("clearAfterRead",e,t,n),l=yN("identicalElementShapes",e,t,n),u=yN("name",e,t,n),h=new AT(u,a,r,s,l,i,o);return n.addTensorArray(h),[h.idTensor,au(1)]}case"TensorArrayWriteV3":{let r=yN("tensorArrayId",e,t,n),a=yN("index",e,t,n),s=yN("tensor",e,t,n),i=n.getTensorArray(r.id);return i.write(a,s),[i.idTensor]}case"TensorArrayReadV3":{let r=yN("tensorArrayId",e,t,n),a=yN("index",e,t,n);return[n.getTensorArray(r.id).read(a)]}case"TensorArrayGatherV3":{let r=yN("tensorArrayId",e,t,n),a=yN("indices",e,t,n),s=yN("dtype",e,t,n);return[n.getTensorArray(r.id).gather(a,s)]}case"TensorArrayScatterV3":{let r=yN("tensorArrayId",e,t,n),a=yN("indices",e,t,n),s=yN("tensor",e,t,n),i=n.getTensorArray(r.id);return i.scatter(a,s),[i.idTensor]}case"TensorArrayConcatV3":{let r=yN("tensorArrayId",e,t,n),a=n.getTensorArray(r.id),s=yN("dtype",e,t,n);return[a.concat(s)]}case"TensorArraySplitV3":{let r=yN("tensorArrayId",e,t,n),a=yN("tensor",e,t,n),s=yN("lengths",e,t,n),i=n.getTensorArray(r.id);return i.split(s,a),[i.idTensor]}case"TensorArraySizeV3":{let r=yN("tensorArrayId",e,t,n);return[au(n.getTensorArray(r.id).size(),"int32")]}case"TensorArrayCloseV3":{let r=yN("tensorArrayId",e,t,n),a=n.getTensorArray(r.id);return a.clearAndClose(),[a.idTensor]}case"TensorListSetItem":{let r=yN("tensorListId",e,t,n),a=yN("index",e,t,n),s=yN("tensor",e,t,n),i=n.getTensorList(r.id);return i.setItem(a,s),[i.idTensor]}case"TensorListGetItem":{let r=yN("tensorListId",e,t,n),a=yN("index",e,t,n),s=yN("elementShape",e,t,n),i=yN("elementDType",e,t,n);return[n.getTensorList(r.id).getItem(a,s,i)]}case"TensorListScatterV2":case"TensorListScatter":{let r=yN("indices",e,t,n),a=function(e,t,n,r){if(t.length!==e.shape[0])throw new Error(`Expected len(indices) == tensor.shape[0], but saw: ${t.length} vs. ${e.shape[0]}`);let a=Math.max(...t);if(null!=r&&-1!==r&&a>=r)throw new Error(`Max index must be < array size (${a}  vs. ${r})`);let s=new $T([],n,e.dtype,r),i=jd(e,0);return t.forEach((e,t)=>{s.setItem(e,i[t])}),s}(yN("tensor",e,t,n),r,yN("elementShape",e,t,n),yN("numElements",e,t,n));return n.addTensorList(a),[a.idTensor]}case"TensorListReserve":case"EmptyTensorList":{let r,a=yN("elementShape",e,t,n),s=yN("elementDType",e,t,n);r="TensorListReserve"===e.op?"numElements":"maxNumElements";let i=yN(r,e,t,n),o=function(e,t,n,r){return new $T([],e,t,r)}(a,s,0,"TensorListReserve"===e.op?-1:i);return n.addTensorList(o),[o.idTensor]}case"TensorListGather":{let r=yN("tensorListId",e,t,n),a=yN("indices",e,t,n),s=yN("elementShape",e,t,n),i=yN("elementDType",e,t,n);return[n.getTensorList(r.id).gather(a,i,s)]}case"TensorListStack":{let r=yN("tensorListId",e,t,n),a=yN("elementShape",e,t,n),s=yN("elementDType",e,t,n),i=yN("numElements",e,t,n);return[n.getTensorList(r.id).stack(a,s,i)]}case"TensorListFromTensor":{let r=function(e,t,n){let r=e.dtype;if(e.shape.length<1)throw new Error(`Tensor must be at least a vector, but saw shape: ${e.shape}`);if(e.dtype!==n)throw new Error(`Invalid data types; op elements ${e.dtype}, but list elements ${n}`);NT(e.shape.slice(1),t,"TensorList shape mismatch: ");let a=jd(e);return new $T(a,t,r)}(yN("tensor",e,t,n),yN("elementShape",e,t,n),yN("elementDType",e,t,n));return n.addTensorList(r),[r.idTensor]}case"TensorListConcat":case"TensorListConcatV2":{let r=yN("tensorListId",e,t,n),a=n.getTensorList(r.id),s=yN("dtype",e,t,n),i=yN("elementShape",e,t,n);return[a.concat(s,i)]}case"TensorListPushBack":{let r=yN("tensorListId",e,t,n),a=yN("tensor",e,t,n),s=n.getTensorList(r.id);return s.pushBack(a),[s.idTensor]}case"TensorListPopBack":{let r=yN("tensorListId",e,t,n),a=yN("elementShape",e,t,n),s=yN("elementDType",e,t,n);return[n.getTensorList(r.id).popBack(a,s)]}case"TensorListSplit":{let r=yN("tensor",e,t,n),a=yN("elementShape",e,t,n),s=function(e,t,n){let r=0,a=t.map(e=>(r+=e,r));if(r!==e.shape[0])throw new Error(`Expected sum of lengths to be equal to\n          tensor.shape[0], but sum of lengths is\n        ${r}, and tensor's shape is: ${e.shape}`);let s=ET(e.shape.slice(1),n),i=0===r?0:e.size/r,o=Qs(()=>{let n=[];e=jo(e,[1,r,i]);for(let r=0;r<t.length;++r){let o=[0,0===r?0:a[r-1],0],l=[1,t[r],i];n[r]=jo(Jo(e,o,l),s)}return e.dispose(),n}),l=new $T([],n,e.dtype,t.length);for(let e=0;e<o.length;e++)l.setItem(e,o[e]);return l}(r,yN("lengths",e,t,n),a);return n.addTensorList(s),[s.idTensor]}case"TensorListLength":{let r=yN("tensorListId",e,t,n);return[au(n.getTensorList(r.id).size(),"int32")]}case"TensorListResize":{let r=yN("tensorListId",e,t,n),a=yN("size",e,t,n),s=n.getTensorList(r.id).resize(a);return n.addTensorList(s),[s.idTensor]}default:throw TypeError(`Node type ${e.op} is not implemented`)}};function FT(e,t,n){let[r,a]=yN("fusedOps",e,t,n),s="biasadd"===r,i=!s,o="prelu"===a,l="fusedbatchnorm"===r,u=yN("numArgs",e,t,n);if(s){if(o&&2!==u)throw new Error("FusedConv2d and DepthwiseConv2d with BiasAdd and Prelu must have two extra arguments: bias and alpha.");if(!o&&s&&1!==u)throw new Error("FusedConv2d and DepthwiseConv2d with BiasAdd must have one extra argument: bias.")}if(l)throw new Error("FusedConv2d and DepthwiseConv2d with FusedBatchNorm is not supported");let h=yN("strides",e,t,n),d=IN(e,t,n),p=yN("dataFormat",e,t,n).toUpperCase(),c=yN("dilations",e,t,n),[f,m]=yN("args",e,t,n);return i&&(m=f,f=void 0),{stride:h,pad:d,dataFormat:p,dilations:c,biasArg:f,preluArg:m,activationFunc:a,leakyreluAlpha:yN("leakyreluAlpha",e,t,n)}}function DT(e,t,n){return{boxes:yN("boxes",e,t,n),scores:yN("scores",e,t,n),maxOutputSize:yN("maxOutputSize",e,t,n),iouThreshold:yN("iouThreshold",e,t,n),scoreThreshold:yN("scoreThreshold",e,t,n),softNmsSigma:yN("softNmsSigma",e,t,n)}}var MT=class{get id(){return this.handle.id}constructor(e,t){this.keyDType=e,this.valueDType=t,this.handle=au(0),this.tensorMap=new Map,ti(this.handle)}clearAndClose(){this.tensorMap.forEach(e=>e.dispose()),this.tensorMap.clear(),this.handle.dispose()}size(){return this.tensorMap.size}tensorSize(){return au(this.size(),"int32")}async import(e,t){this.checkKeyAndValueTensor(e,t);let n=await e.data();return this.tensorMap.forEach(e=>e.dispose()),this.tensorMap.clear(),Qs(()=>{let e=jd(t),r=n.length,a=e.length;va.assert(r===a,()=>`The number of elements doesn't match, keys has ${r} elements, the values has ${a} elements.`);for(let t=0;t<r;t++){let r=n[t],a=e[t];ti(a),this.tensorMap.set(r,a)}return this.handle})}async find(e,t){this.checkKeyAndValueTensor(e,t);let n=await e.data();return Qs(()=>{let e=[];for(let r=0;r<n.length;r++){let a=n[r],s=this.findWithDefault(a,t);e.push(s)}return Td(e)})}findWithDefault(e,t){let n=this.tensorMap.get(e);return null!=n?n:t}checkKeyAndValueTensor(e,t){if(e.dtype!==this.keyDType)throw new Error(`Expect key dtype ${this.keyDType}, but got ${e.dtype}`);if(t.dtype!==this.valueDType)throw new Error(`Expect value dtype ${this.valueDType}, but got ${t.dtype}`)}};function OT(e,t,n,r,a=Qs){let s=((e,t,n)=>{switch(e.category){case"arithmetic":return a(()=>((e,t,n,r=_T)=>{switch(e.op){case"BiasAdd":case"AddV2":case"Add":return[r.add(yN("a",e,t,n),yN("b",e,t,n))];case"AddN":return[r.addN(yN("tensors",e,t,n))];case"FloorMod":case"Mod":return[r.mod(yN("a",e,t,n),yN("b",e,t,n))];case"Mul":return[r.mul(yN("a",e,t,n),yN("b",e,t,n))];case"RealDiv":case"Div":return[r.div(yN("a",e,t,n),yN("b",e,t,n))];case"DivNoNan":return[r.divNoNan(yN("a",e,t,n),yN("b",e,t,n))];case"FloorDiv":return[r.floorDiv(yN("a",e,t,n),yN("b",e,t,n))];case"Sub":return[r.sub(yN("a",e,t,n),yN("b",e,t,n))];case"Minimum":return[r.minimum(yN("a",e,t,n),yN("b",e,t,n))];case"Maximum":return[r.maximum(yN("a",e,t,n),yN("b",e,t,n))];case"Pow":return[r.pow(yN("a",e,t,n),yN("b",e,t,n))];case"SquaredDifference":return[r.squaredDifference(yN("a",e,t,n),yN("b",e,t,n))];default:throw TypeError(`Node type ${e.op} is not implemented`)}})(e,t,n));case"basic_math":return a(()=>((e,t,n,r=_T)=>{switch(e.op){case"Abs":case"ComplexAbs":return[r.abs(yN("x",e,t,n))];case"Acos":return[r.acos(yN("x",e,t,n))];case"Acosh":return[r.acosh(yN("x",e,t,n))];case"Asin":return[r.asin(yN("x",e,t,n))];case"Asinh":return[r.asinh(yN("x",e,t,n))];case"Atan":return[r.atan(yN("x",e,t,n))];case"Atan2":return[r.atan2(yN("x",e,t,n),yN("y",e,t,n))];case"Atanh":return[r.atanh(yN("x",e,t,n))];case"Ceil":return[r.ceil(yN("x",e,t,n))];case"Complex":return[r.complex(yN("real",e,t,n),yN("imag",e,t,n))];case"Cos":return[r.cos(yN("x",e,t,n))];case"Cosh":return[r.cosh(yN("x",e,t,n))];case"Elu":return[r.elu(yN("x",e,t,n))];case"Erf":return[r.erf(yN("x",e,t,n))];case"Exp":return[r.exp(yN("x",e,t,n))];case"Expm1":return[r.expm1(yN("x",e,t,n))];case"Floor":return[r.floor(yN("x",e,t,n))];case"Log":return[r.log(yN("x",e,t,n))];case"Log1p":return[r.log1p(yN("x",e,t,n))];case"Imag":return[r.imag(yN("x",e,t,n))];case"Neg":return[r.neg(yN("x",e,t,n))];case"Reciprocal":return[r.reciprocal(yN("x",e,t,n))];case"Real":return[r.real(yN("x",e,t,n))];case"Relu":return[r.relu(yN("x",e,t,n))];case"Round":return[r.round(yN("x",e,t,n))];case"Selu":return[r.selu(yN("x",e,t,n))];case"Sigmoid":return[r.sigmoid(yN("x",e,t,n))];case"Sin":return[r.sin(yN("x",e,t,n))];case"Sign":return[r.sign(yN("x",e,t,n))];case"Sinh":return[r.sinh(yN("x",e,t,n))];case"Softplus":return[r.softplus(yN("x",e,t,n))];case"Sqrt":return[r.sqrt(yN("x",e,t,n))];case"Square":return[r.square(yN("x",e,t,n))];case"Tanh":return[r.tanh(yN("x",e,t,n))];case"Tan":return[r.tan(yN("x",e,t,n))];case"ClipByValue":return[r.clipByValue(yN("x",e,t,n),yN("clipValueMin",e,t,n),yN("clipValueMax",e,t,n))];case"Relu6":return[r.relu6(yN("x",e,t,n))];case"Rsqrt":return[r.rsqrt(bN(e.inputNames[0],t,n))];case"LeakyRelu":return[r.leakyRelu(yN("x",e,t,n),yN("alpha",e,t,n))];case"Prelu":return[r.prelu(yN("x",e,t,n),yN("alpha",e,t,n))];case"IsNan":return[r.isNaN(bN(e.inputNames[0],t,n))];case"IsInf":return[r.isInf(bN(e.inputNames[0],t,n))];case"IsFinite":return[r.isFinite(bN(e.inputNames[0],t,n))];default:throw TypeError(`Node type ${e.op} is not implemented`)}})(e,t,n));case"control":return RT(e,t,n);case"convolution":return a(()=>((e,t,n,r=_T)=>{switch(e.op){case"Conv1D":{let a=yN("stride",e,t,n),s=yN("pad",e,t,n),i=yN("dataFormat",e,t,n).toUpperCase(),o=yN("dilation",e,t,n);return[r.conv1d(yN("x",e,t,n),yN("filter",e,t,n),a,s,i,o)]}case"Conv2D":{let a=yN("strides",e,t,n),s=IN(e,t,n),i=yN("dataFormat",e,t,n).toUpperCase(),o=yN("dilations",e,t,n);return[r.conv2d(yN("x",e,t,n),yN("filter",e,t,n),[a[1],a[2]],s,i,[o[1],o[2]])]}case"_FusedConv2D":{let{stride:a,pad:s,dataFormat:i,dilations:o,biasArg:l,preluArg:u,activationFunc:h,leakyreluAlpha:d}=FT(e,t,n);return[r.fused.conv2d({x:yN("x",e,t,n),filter:yN("filter",e,t,n),strides:[a[1],a[2]],pad:s,dataFormat:i,dilations:[o[1],o[2]],bias:l,activation:h,preluActivationWeights:u,leakyreluAlpha:d})]}case"FusedDepthwiseConv2dNative":{let{stride:a,pad:s,dataFormat:i,dilations:o,biasArg:l,preluArg:u,activationFunc:h,leakyreluAlpha:d}=FT(e,t,n);return[r.fused.depthwiseConv2d({x:yN("x",e,t,n),filter:yN("filter",e,t,n),strides:[a[1],a[2]],pad:s,dataFormat:i,dilations:[o[1],o[2]],bias:l,activation:h,preluActivationWeights:u,leakyreluAlpha:d})]}case"Conv2DBackpropInput":case"Conv2dTranspose":{let a=yN("outputShape",e,t,n),s=yN("strides",e,t,n),i=IN(e,t,n);return[r.conv2dTranspose(yN("x",e,t,n),yN("filter",e,t,n),a,[s[1],s[2]],i)]}case"DepthwiseConv2dNative":case"DepthwiseConv2d":{let a=yN("strides",e,t,n),s=IN(e,t,n),i=yN("dilations",e,t,n),o=yN("dataFormat",e,t,n).toUpperCase();return[r.depthwiseConv2d(yN("input",e,t,n),yN("filter",e,t,n),[a[1],a[2]],s,o,[i[1],i[2]])]}case"Conv3D":{let a=yN("strides",e,t,n),s=yN("pad",e,t,n),i=yN("dataFormat",e,t,n).toUpperCase(),o=yN("dilations",e,t,n);return[r.conv3d(yN("x",e,t,n),yN("filter",e,t,n),[a[1],a[2],a[3]],s,i,[o[1],o[2],o[3]])]}case"AvgPool":{let a=yN("strides",e,t,n),s=yN("pad",e,t,n),i=yN("kernelSize",e,t,n);return[r.avgPool(yN("x",e,t,n),[i[1],i[2]],[a[1],a[2]],s)]}case"MaxPool":{let a=yN("strides",e,t,n),s=yN("pad",e,t,n),i=yN("kernelSize",e,t,n);return[r.maxPool(yN("x",e,t,n),[i[1],i[2]],[a[1],a[2]],s)]}case"MaxPoolWithArgmax":{let a=yN("strides",e,t,n),s=yN("pad",e,t,n),i=yN("kernelSize",e,t,n),o=yN("includeBatchInIndex",e,t,n),{result:l,indexes:u}=r.maxPoolWithArgmax(yN("x",e,t,n),[i[1],i[2]],[a[1],a[2]],s,o);return[l,u]}case"AvgPool3D":{let a=yN("strides",e,t,n),s=yN("pad",e,t,n),i=yN("kernelSize",e,t,n);return[r.avgPool3d(yN("x",e,t,n),[i[1],i[2],i[3]],[a[1],a[2],a[3]],s)]}case"MaxPool3D":{let a=yN("strides",e,t,n),s=yN("pad",e,t,n),i=yN("kernelSize",e,t,n);return[r.maxPool3d(yN("x",e,t,n),[i[1],i[2],i[3]],[a[1],a[2],a[3]],s)]}case"Dilation2D":{let a=yN("strides",e,t,n),s=yN("pad",e,t,n),i=yN("dilations",e,t,n),o=a[1],l=a[2],u=i[1],h=i[2];return[r.dilation2d(yN("x",e,t,n),yN("filter",e,t,n),[o,l],s,[u,h],"NHWC")]}default:throw TypeError(`Node type ${e.op} is not implemented`)}})(e,t,n));case"creation":return a(()=>((e,t,n,r=_T)=>{switch(e.op){case"Fill":{let a=yN("shape",e,t,n),s=yN("dtype",e,t,n),i=yN("value",e,t,n);return[r.fill(a,i,s)]}case"LinSpace":{let a=yN("start",e,t,n),s=yN("stop",e,t,n),i=yN("num",e,t,n);return[r.linspace(a,s,i)]}case"Multinomial":{let a=yN("logits",e,t,n),s=yN("numSamples",e,t,n),i=yN("seed",e,t,n);return[r.multinomial(a,s,i)]}case"OneHot":{let a=yN("indices",e,t,n),s=yN("depth",e,t,n),i=yN("onValue",e,t,n),o=yN("offValue",e,t,n),l=yN("dtype",e,t,n);return[r.oneHot(a,s,i,o,l)]}case"Ones":return[r.ones(yN("shape",e,t,n),yN("dtype",e,t,n))];case"OnesLike":return[r.onesLike(yN("x",e,t,n))];case"RandomStandardNormal":return[r.randomStandardNormal(yN("shape",e,t,n),yN("dtype",e,t,n),yN("seed",e,t,n))];case"RandomUniform":return[r.randomUniform(yN("shape",e,t,n),yN("minval",e,t,n),yN("maxval",e,t,n),yN("dtype",e,t,n))];case"RandomUniformInt":return[r.randomUniformInt(yN("shape",e,t,n),yN("minval",e,t,n),yN("maxval",e,t,n),yN("seed",e,t,n))];case"Range":{let a=yN("start",e,t,n),s=yN("stop",e,t,n),i=yN("step",e,t,n);return[r.range(a,s,i,yN("dtype",e,t,n))]}case"TruncatedNormal":{let a=yN("shape",e,t,n),s=yN("mean",e,t,n),i=yN("stdDev",e,t,n),o=yN("seed",e,t,n);return[r.truncatedNormal(a,s,i,yN("dtype",e,t,n),o)]}case"Zeros":return[r.zeros(yN("shape",e,t,n),yN("dtype",e,t,n))];case"ZerosLike":return[r.zerosLike(yN("x",e,t,n))];default:throw TypeError(`Node type ${e.op} is not implemented`)}})(e,t,n));case"dynamic":return(async(e,t,n,r,a=_T)=>{switch(e.op){case"NonMaxSuppressionV5":{let{boxes:r,scores:s,maxOutputSize:i,iouThreshold:o,scoreThreshold:l,softNmsSigma:u}=DT(e,t,n),h=await a.image.nonMaxSuppressionWithScoreAsync(r,s,i,o,l,u);return[h.selectedIndices,h.selectedScores]}case"NonMaxSuppressionV4":{let{boxes:r,scores:s,maxOutputSize:i,iouThreshold:o,scoreThreshold:l}=DT(e,t,n),u=yN("padToMaxOutputSize",e,t,n),h=await a.image.nonMaxSuppressionPaddedAsync(r,s,i,o,l,u);return[h.selectedIndices,h.validOutputs]}case"NonMaxSuppressionV3":case"NonMaxSuppressionV2":{let{boxes:r,scores:s,maxOutputSize:i,iouThreshold:o,scoreThreshold:l}=DT(e,t,n);return[await a.image.nonMaxSuppressionAsync(r,s,i,o,l)]}case"Where":{let r=a.cast(yN("condition",e,t,n),"bool"),s=[await a.whereAsync(r)];return r.dispose(),s}case"ListDiff":return a.setdiff1dAsync(yN("x",e,t,n),yN("y",e,t,n));default:throw TypeError(`Node type ${e.op} is not implemented`)}})(e,t,n);case"evaluation":return a(()=>((e,t,n,r=_T)=>{switch(e.op){case"LowerBound":{let a=yN("sortedSequence",e,t,n),s=yN("values",e,t,n);return[r.lowerBound(a,s)]}case"TopKV2":{let a=yN("x",e,t,n),s=yN("k",e,t,n),i=yN("sorted",e,t,n),o=r.topk(a,s,i);return[o.values,o.indices]}case"UpperBound":{let a=yN("sortedSequence",e,t,n),s=yN("values",e,t,n);return[r.upperBound(a,s)]}case"Unique":{let a=yN("x",e,t,n),s=r.unique(a);return[s.values,s.indices]}case"UniqueV2":{let a=yN("x",e,t,n),s=yN("axis",e,t,n),i=r.unique(a,s);return[i.values,i.indices]}default:throw TypeError(`Node type ${e.op} is not implemented`)}})(e,t,n));case"image":return a(()=>((e,t,n,r=_T)=>{switch(e.op){case"ResizeBilinear":{let a=yN("images",e,t,n),s=yN("size",e,t,n),i=yN("alignCorners",e,t,n),o=yN("halfPixelCenters",e,t,n);return[r.image.resizeBilinear(a,[s[0],s[1]],i,o)]}case"ResizeNearestNeighbor":{let a=yN("images",e,t,n),s=yN("size",e,t,n),i=yN("alignCorners",e,t,n),o=yN("halfPixelCenters",e,t,n);return[r.image.resizeNearestNeighbor(a,[s[0],s[1]],i,o)]}case"CropAndResize":{let a=yN("image",e,t,n),s=yN("boxes",e,t,n),i=yN("boxInd",e,t,n),o=yN("cropSize",e,t,n),l=yN("method",e,t,n),u=yN("extrapolationValue",e,t,n);return[r.image.cropAndResize(a,s,i,o,l,u)]}case"ImageProjectiveTransformV3":{let a=yN("images",e,t,n),s=yN("transforms",e,t,n),i=yN("outputShape",e,t,n),o=yN("fillValue",e,t,n),l=yN("interpolation",e,t,n),u=yN("fillMode",e,t,n);return[r.image.transform(a,s,l.toLowerCase(),u.toLowerCase(),o,i)]}default:throw TypeError(`Node type ${e.op} is not implemented`)}})(e,t,n));case"graph":return a(()=>((e,t,n,r=_T)=>{switch(e.op){case"Const":return t[e.name];case"PlaceholderWithDefault":let a=yN("default",e,t,n);return[bN(e.name,t,n)||a];case"Placeholder":return[bN(e.name,t,n)];case"Identity":case"StopGradient":case"FakeQuantWithMinMaxVars":case"Snapshot":return[SN(yN("x",e,t,n))];case"IdentityN":return yN("x",e,t,n).map(e=>SN(e));case"Shape":return[r.tensor1d(yN("x",e,t,n).shape,"int32")];case"ShapeN":return yN("x",e,t,n).map(e=>r.tensor1d(e.shape));case"Size":return[r.scalar(yN("x",e,t,n).size,"int32")];case"Rank":return[r.scalar(yN("x",e,t,n).rank,"int32")];case"NoOp":return[r.scalar(1)];case"Print":let s=yN("x",e,t,n),i=yN("data",e,t,n),o=yN("message",e,t,n),l=yN("summarize",e,t,n);console.warn("The graph has a tf.print() operation,usually used for debugging, which slows down performance."),console.log(o);for(let e=0;e<i.length;e++)console.log(Array.prototype.slice.call(i[e].dataSync()).slice(0,l));return[s];default:throw TypeError(`Node type ${e.op} is not implemented`)}})(e,t,n));case"logical":return a(()=>((e,t,n,r=_T)=>{switch(e.op){case"Equal":return[r.equal(yN("a",e,t,n),yN("b",e,t,n))];case"NotEqual":return[r.notEqual(yN("a",e,t,n),yN("b",e,t,n))];case"Greater":return[r.greater(yN("a",e,t,n),yN("b",e,t,n))];case"GreaterEqual":return[r.greaterEqual(yN("a",e,t,n),yN("b",e,t,n))];case"Less":return[r.less(yN("a",e,t,n),yN("b",e,t,n))];case"LessEqual":return[r.lessEqual(yN("a",e,t,n),yN("b",e,t,n))];case"LogicalAnd":return[r.logicalAnd(yN("a",e,t,n),yN("b",e,t,n))];case"LogicalNot":return[r.logicalNot(yN("a",e,t,n))];case"LogicalOr":return[r.logicalOr(yN("a",e,t,n),yN("b",e,t,n))];case"Select":case"SelectV2":return[r.where(yN("condition",e,t,n),yN("a",e,t,n),yN("b",e,t,n))];case"BitwiseAnd":return[r.bitwiseAnd(yN("a",e,t,n),yN("b",e,t,n))];default:throw TypeError(`Node type ${e.op} is not implemented`)}})(e,t,n));case"matrices":return a(()=>((e,t,n,r=_T)=>{switch(e.op){case"BatchMatMul":case"BatchMatMulV2":case"MatMul":return[r.matMul(yN("a",e,t,n),yN("b",e,t,n),yN("transposeA",e,t,n),yN("transposeB",e,t,n))];case"Einsum":return[r.einsum(yN("equation",e,t,n),...yN("tensors",e,t,n))];case"Transpose":return[r.transpose(yN("x",e,t,n),yN("perm",e,t,n))];case"_FusedMatMul":let[a,s]=yN("fusedOps",e,t,n),i="biasadd"===a,o="prelu"===s,l=yN("numArgs",e,t,n),u=yN("leakyreluAlpha",e,t,n);if(i){if(o&&2!==l)throw new Error("Fused MatMul with BiasAdd and Prelu must have two extra arguments: bias and alpha.");if(!o&&1!==l)throw new Error("Fused MatMul with BiasAdd must have one extra argument: bias.")}let[h,d]=yN("args",e,t,n);return[r.fused.matMul({a:yN("a",e,t,n),b:yN("b",e,t,n),transposeA:yN("transposeA",e,t,n),transposeB:yN("transposeB",e,t,n),bias:h,activation:s,preluActivationWeights:d,leakyreluAlpha:u})];case"MatrixBandPart":return[r.linalg.bandPart(yN("a",e,t,n),yN("numLower",e,t,n),yN("numUpper",e,t,n))];default:throw TypeError(`Node type ${e.op} is not implemented`)}})(e,t,n));case"normalization":return a(()=>((e,t,n,r=_T)=>{switch(e.op){case"EuclideanNorm":return[r.euclideanNorm(yN("x",e,t,n),yN("axis",e,t,n),yN("keepDims",e,t,n))];case"FusedBatchNorm":case"FusedBatchNormV2":case"FusedBatchNormV3":return[r.batchNorm(yN("x",e,t,n),yN("mean",e,t,n),yN("variance",e,t,n),yN("offset",e,t,n),yN("scale",e,t,n),yN("epsilon",e,t,n))];case"LRN":return[r.localResponseNormalization(yN("x",e,t,n),yN("radius",e,t,n),yN("bias",e,t,n),yN("alpha",e,t,n),yN("beta",e,t,n))];case"Softmax":return[r.softmax(yN("x",e,t,n))];case"LogSoftmax":return[r.logSoftmax(yN("x",e,t,n))];default:throw TypeError(`Node type ${e.op} is not implemented`)}})(e,t,n));case"ragged":return a(()=>((e,t,n,r=_T)=>{switch(e.op){case"RaggedGather":{let{outputNestedSplits:a,outputDenseValues:s}=r.raggedGather(yN("paramsNestedSplits",e,t,n),yN("paramsDenseValues",e,t,n),yN("indices",e,t,n),yN("outputRaggedRank",e,t,n));return a.concat(s)}case"RaggedRange":{let{rtNestedSplits:a,rtDenseValues:s}=r.raggedRange(yN("starts",e,t,n),yN("limits",e,t,n),yN("splits",e,t,n));return[a,s]}case"RaggedTensorToTensor":return[r.raggedTensorToTensor(yN("shape",e,t,n),yN("values",e,t,n),yN("defaultValue",e,t,n),yN("rowPartitionTensors",e,t,n),yN("rowPartitionTypes",e,t,n))];default:throw TypeError(`Node type ${e.op} is not implemented`)}})(e,t,n));case"reduction":return a(()=>((e,t,n,r=_T)=>{switch(e.op){case"Max":{let a=yN("axis",e,t,n),s=yN("keepDims",e,t,n);return[r.max(yN("x",e,t,n),a,s)]}case"Mean":{let a=yN("axis",e,t,n),s=yN("keepDims",e,t,n);return[r.mean(yN("x",e,t,n),a,s)]}case"Min":{let a=yN("axis",e,t,n),s=yN("keepDims",e,t,n);return[r.min(yN("x",e,t,n),a,s)]}case"Sum":{let a=yN("axis",e,t,n),s=yN("keepDims",e,t,n);return[r.sum(yN("x",e,t,n),a,s)]}case"All":{let a=yN("axis",e,t,n),s=yN("keepDims",e,t,n);return[r.all(yN("x",e,t,n),a,s)]}case"Any":{let a=yN("axis",e,t,n),s=yN("keepDims",e,t,n);return[r.any(yN("x",e,t,n),a,s)]}case"ArgMax":{let a=yN("axis",e,t,n);return[r.argMax(yN("x",e,t,n),a)]}case"ArgMin":{let a=yN("axis",e,t,n);return[r.argMin(yN("x",e,t,n),a)]}case"Prod":{let a=yN("axis",e,t,n),s=yN("keepDims",e,t,n);return[r.prod(yN("x",e,t,n),a,s)]}case"Cumprod":{let a=yN("axis",e,t,n),s=yN("exclusive",e,t,n),i=yN("reverse",e,t,n);return[r.cumprod(yN("x",e,t,n),a,s,i)]}case"Cumsum":{let a=yN("axis",e,t,n),s=yN("exclusive",e,t,n),i=yN("reverse",e,t,n);return[r.cumsum(yN("x",e,t,n),a,s,i)]}case"Bincount":let a=yN("x",e,t,n),s=yN("weights",e,t,n),i=yN("size",e,t,n);return[r.bincount(a,s,i)];case"DenseBincount":{let a=yN("x",e,t,n),s=yN("weights",e,t,n),i=yN("size",e,t,n),o=yN("binaryOutput",e,t,n);return[r.denseBincount(a,s,i,o)]}default:throw TypeError(`Node type ${e.op} is not implemented`)}})(e,t,n));case"slice_join":return a(()=>((e,t,n,r=_T)=>{switch(e.op){case"ConcatV2":case"Concat":{let a=yN("n",e,t,n),s=yN("axis",e,t,n),i=yN("tensors",e,t,n);return i=i.slice(0,a),[r.concat(i,s)]}case"Gather":{let a=yN("x",e,t,n),s=yN("indices",e,t,n);return[r.gather(a,r.cast(s,"int32"),0)]}case"GatherV2":{let a=yN("axis",e,t,n),s=yN("batchDims",e,t,n),i=yN("x",e,t,n),o=yN("indices",e,t,n);return[r.gather(i,r.cast(o,"int32"),a,s)]}case"Reverse":{let a=yN("dims",e,t,n),s=[];for(let e=0;e<a.length;e++)a[e]&&s.push(e);let i=yN("x",e,t,n);return[r.reverse(i,s)]}case"ReverseV2":{let a=yN("axis",e,t,n),s=yN("x",e,t,n);return[r.reverse(s,a)]}case"Slice":{let a=yN("begin",e,t,n),s=yN("size",e,t,n);return[r.slice(yN("x",e,t,n),a,s)]}case"StridedSlice":{let a=yN("begin",e,t,n),s=yN("end",e,t,n),i=yN("strides",e,t,n),o=yN("beginMask",e,t,n),l=yN("endMask",e,t,n),u=yN("ellipsisMask",e,t,n),h=yN("newAxisMask",e,t,n),d=yN("shrinkAxisMask",e,t,n),p=yN("x",e,t,n);return[r.stridedSlice(p,a,s,i,o,l,u,h,d)]}case"Pack":return Qs(()=>{let a=yN("axis",e,t,n),s=yN("tensors",e,t,n),i=s[0].shape,o=r.squeeze(s[0]).shape,l=s.map(e=>{let t=va.arraysEqual(e.shape,i);if(!t&&!va.arraysEqual(r.squeeze(e).shape,o))throw new Error("the input tensors shape does not match");return t?e:r.reshape(e,i)});return[r.stack(l,a)]});case"Unpack":{let a=yN("axis",e,t,n),s=yN("tensor",e,t,n);return r.unstack(s,a)}case"Tile":{let a=yN("reps",e,t,n);return[r.tile(yN("x",e,t,n),a)]}case"Split":case"SplitV":{let a=yN("axis",e,t,n),s=yN("numOrSizeSplits",e,t,n),i=yN("x",e,t,n);return r.split(i,s,a)}case"ScatterNd":{let a=yN("indices",e,t,n),s=yN("values",e,t,n),i=yN("shape",e,t,n);return[r.scatterND(a,s,i)]}case"GatherNd":{let a=yN("x",e,t,n),s=yN("indices",e,t,n);return[r.gatherND(a,s)]}case"SparseToDense":{let a=yN("sparseIndices",e,t,n),s=yN("outputShape",e,t,n),i=yN("sparseValues",e,t,n),o=yN("defaultValue",e,t,n);return[r.sparseToDense(a,i,s,i.dtype===o.dtype?o:r.cast(o,i.dtype))]}case"TensorScatterUpdate":{let a=yN("indices",e,t,n),s=yN("values",e,t,n),i=yN("tensor",e,t,n);return[r.tensorScatterUpdate(i,a,s)]}default:throw TypeError(`Node type ${e.op} is not implemented`)}})(e,t,n));case"sparse":return a(()=>((e,t,n,r=_T)=>{switch(e.op){case"SparseFillEmptyRows":{let{outputIndices:a,outputValues:s,emptyRowIndicator:i,reverseIndexMap:o}=r.sparse.sparseFillEmptyRows(yN("indices",e,t,n),yN("values",e,t,n),yN("denseShape",e,t,n),yN("defaultValue",e,t,n));return[a,s,i,o]}case"SparseReshape":{let{outputIndices:a,outputShape:s}=r.sparse.sparseReshape(yN("inputIndices",e,t,n),yN("inputShape",e,t,n),yN("newShape",e,t,n));return[a,s]}case"SparseSegmentMean":return[r.sparse.sparseSegmentMean(yN("data",e,t,n),yN("indices",e,t,n),yN("segmentIds",e,t,n))];case"SparseSegmentSum":return[r.sparse.sparseSegmentSum(yN("data",e,t,n),yN("indices",e,t,n),yN("segmentIds",e,t,n))];default:throw TypeError(`Node type ${e.op} is not implemented`)}})(e,t,n));case"spectral":return a(()=>((e,t,n,r=_T)=>{switch(e.op){case"FFT":return[r.fft(yN("x",e,t,n))];case"IFFT":return[r.ifft(yN("x",e,t,n))];case"RFFT":return[r.rfft(yN("x",e,t,n))];case"IRFFT":return[r.irfft(yN("x",e,t,n))];default:throw TypeError(`Node type ${e.op} is not implemented`)}})(e,t,n));case"string":return a(()=>((e,t,n,r=_T)=>{switch(e.op){case"StaticRegexReplace":return[r.string.staticRegexReplace(yN("input",e,t,n),yN("pattern",e,t,n),yN("rewrite",e,t,n),yN("replaceGlobal",e,t,n))];case"StringNGrams":{let{nGrams:a,nGramsSplits:s}=r.string.stringNGrams(yN("data",e,t,n),yN("dataSplits",e,t,n),yN("separator",e,t,n),yN("nGramWidths",e,t,n),yN("leftPad",e,t,n),yN("rightPad",e,t,n),yN("padWidth",e,t,n),yN("preserveShortSequences",e,t,n));return[a,s]}case"StringSplit":{let{indices:a,values:s,shape:i}=r.string.stringSplit(yN("input",e,t,n),yN("delimiter",e,t,n),yN("skipEmpty",e,t,n));return[a,s,i]}case"StringToHashBucketFast":return[r.string.stringToHashBucketFast(yN("input",e,t,n),yN("numBuckets",e,t,n))];default:throw TypeError(`Node type ${e.op} is not implemented`)}})(e,t,n));case"transformation":return a(()=>((e,t,n,r=_T)=>{switch(e.op){case"Cast":return[r.cast(yN("x",e,t,n),yN("dtype",e,t,n))];case"ExpandDims":{let a=yN("axis",e,t,n);return[r.expandDims(yN("x",e,t,n),a)]}case"Squeeze":{let a=yN("axis",e,t,n);return[r.squeeze(yN("x",e,t,n),a)]}case"Reshape":return[r.reshape(yN("x",e,t,n),yN("shape",e,t,n))];case"EnsureShape":return[r.ensureShape(yN("x",e,t,n),yN("shape",e,t,n))];case"MirrorPad":return[r.mirrorPad(yN("x",e,t,n),yN("padding",e,t,n),yN("mode",e,t,n))];case"PadV2":case"Pad":return[r.pad(yN("x",e,t,n),yN("padding",e,t,n),yN("constantValue",e,t,n))];case"SpaceToBatchND":{let a=yN("blockShape",e,t,n),s=yN("paddings",e,t,n);return[r.spaceToBatchND(yN("x",e,t,n),a,s)]}case"BatchToSpaceND":{let a=yN("blockShape",e,t,n),s=yN("crops",e,t,n);return[r.batchToSpaceND(yN("x",e,t,n),a,s)]}case"DepthToSpace":{let a=yN("blockSize",e,t,n),s=yN("dataFormat",e,t,n).toUpperCase();return[r.depthToSpace(yN("x",e,t,n),a,s)]}case"BroadcastTo":return[r.broadcastTo(yN("x",e,t,n),yN("shape",e,t,n))];case"BroadcastArgs":return[r.broadcastArgs(yN("s0",e,t,n),yN("s1",e,t,n))];default:throw TypeError(`Node type ${e.op} is not implemented`)}})(e,t,n));case"hash_table":return(async(e,t,n,r)=>{switch(e.op){case"HashTable":case"HashTableV2":{let a=r.getHashTableHandleByName(e.name);if(null!=a)return[a];{let a=yN("keyDType",e,t,n),s=yN("valueDType",e,t,n),i=new MT(a,s);return r.addHashTable(e.name,i),[i.handle]}}case"InitializeTable":case"InitializeTableV2":case"LookupTableImport":case"LookupTableImportV2":{let a=yN("tableHandle",e,t,n,r),s=yN("keys",e,t,n),i=yN("values",e,t,n);return[await r.getHashTableById(a.id).import(s,i)]}case"LookupTableFind":case"LookupTableFindV2":{let a=yN("tableHandle",e,t,n,r),s=yN("keys",e,t,n),i=yN("defaultValue",e,t,n);return[await r.getHashTableById(a.id).find(s,i)]}case"LookupTableSize":case"LookupTableSizeV2":{let a=yN("tableHandle",e,t,n,r);return[r.getHashTableById(a.id).tensorSize()]}default:throw TypeError(`Node type ${e.op} is not implemented`)}})(e,t,n,r);case"custom":let s=mN(e.op);if(s&&s.customExecutor)return s.customExecutor(new ST(e,t,n));throw TypeError(`Custom op ${e.op} is not registered.`);default:throw TypeError(`Unknown op '${e.op}'. File an issue at https://github.com/tensorflow/tfjs/issues so we can add it, or register a custom execution with tf.registerOp()`)}})(e,t,n);return va.isPromise(s)?s.then(e=>[].concat(e)):[].concat(s)}var LT=class{constructor(e={},t={},n={},r={},a){this.weightMap=e,this.tensorArrayMap=t,this.tensorListMap=n,this.functionMap=r,this.parseNodeNameCache=a,this.rootContext={id:0,frameName:"",iterationId:0},this.contexts=[this.rootContext],this.lastId=0,this.generateCurrentContextIds()}newFrame(e,t){return{id:e,frameName:t,iterationId:0}}set currentContext(e){this.contexts!==e&&(this.contexts=e,this.generateCurrentContextIds())}get currentContext(){return this.contexts}get currentContextId(){return this._currentContextIds[0]}get currentContextIds(){return this._currentContextIds}generateCurrentContextIds(){let e=[];for(let t=0;t<this.contexts.length-1;t++){let n=this.contexts.slice(0,this.contexts.length-t);e.push(this.contextIdforContexts(n))}e.push(""),this._currentContextIds=e}contextIdforContexts(e){return e?e.map(e=>0===e.id&&0===e.iterationId?"":`${e.frameName}-${e.iterationId}`).join("/"):""}enterFrame(e){this.contexts&&(this.lastId++,this.contexts=this.contexts.slice(),this.contexts.push(this.newFrame(this.lastId,e)),this._currentContextIds.unshift(this.contextIdforContexts(this.contexts)))}exitFrame(){if(!(this.contexts&&this.contexts.length>1))throw new Error("Cannot exit frame, the context is empty");this.contexts=this.contexts.slice(),this.contexts.splice(-1),this.currentContextIds.shift()}nextIteration(){if(!(this.contexts&&this.contexts.length>0))throw new Error("Cannot increase frame iteration, the context is empty");{this.contexts=this.contexts.slice(),this.lastId++;let e=Object.assign({},this.contexts[this.contexts.length-1]);e.iterationId+=1,e.id=this.lastId,this.contexts.splice(-1,1,e),this._currentContextIds.splice(0,1,this.contextIdforContexts(this.contexts))}}getWeight(e){return this.weightMap[e]}addTensorArray(e){this.tensorArrayMap[e.id]=e}getTensorArray(e){return this.tensorArrayMap[e]}addTensorList(e){this.tensorListMap[e.id]=e}getTensorList(e){return this.tensorListMap[e]}dispose(e){for(let t in this.tensorArrayMap)this.tensorArrayMap[t].clearAndClose(e);for(let t in this.tensorListMap)this.tensorListMap[t].clearAndClose(e)}};function PT(e,t,n,r){let a=new Set,s=[],i=null,o=null,l=new Set,u=new Set(Object.keys(e).map(e=>kN(e)[0]));r=r||[];let h=new Set(r.map(e=>kN(e.name)[0])),d=[...t];for(;d.length>0;){let e=d.pop();if((GT(e)||HT(e)||jT(e))&&null==i&&(i=e,o=i.children.map(e=>e.name).filter(e=>a.has(e))),a.add(e.name),null==n[e.name]&&!u.has(e.name)&&!h.has(e.name)){if(0===e.inputs.length){s.push(e.name);continue}e.inputs.forEach(e=>{l.has(e.name)||(l.add(e.name),d.push(e))})}}return{inputs:e,outputs:t,usedNodes:a,missingInputs:s,dynamicNode:i,syncInputs:o}}function zT(e,t){let{usedNodes:n,inputs:r}=t,a=Object.keys(r).map(e=>kN(e)[0]).map(t=>e.nodes[t]),s=e.initNodes||[],i=e=>n.has("string"==typeof e?e:e.name);function o(e){return[...new Map(e.map(e=>[e.name,e])).values()]}let l=o([...a,...e.weights,...s]).filter(i),u=o([...l,...Object.values(e.nodes)]).filter(i),h=new Map(u.map(e=>[e.name,e])),d={};for(let e of u){d[e.name]=d[e.name]||0;for(let t of e.children)i(t)||(d[t.name]=Number.POSITIVE_INFINITY),d[t.name]=(d[t.name]||0)+1}let p=Object.entries(d).filter(([,e])=>0===e).map(([e])=>e),c=[...p];for(;p.length>0;){let e=p.pop(),t=h.get(e);for(let e of t.children.filter(i))0===--d[e.name]&&(c.push(e.name),p.push(e.name))}let f=function(e,t){let n=new Map(e.map(e=>[e.name,e])),r=t.map(e=>e.name),a=new Set(r);for(;r.length>0;){let e=r.pop(),t=n.get(e);for(let e of t.children)!n.has(e.name)||a.has(e.name)||(a.add(e.name),r.push(e.name))}return e.filter(e=>a.has(e.name))}(c.map(e=>h.get(e)),l);return function(e,t){let n=new Map(e.map((e,t)=>[e.name,t])),r=new Set(t.map(e=>e.name)),a=e=>r.has("string"==typeof e?e:e.name),s=new Set(e.map(e=>e.name)),i=e=>s.has("string"==typeof e?e:e.name);for(let t of e){for(let e of t.children.filter(i)){if(!n.has(e.name))throw new BT(`Child ${e.name} of node ${t.name} is unreachable.`);if(n.get(t.name)>n.get(e.name))throw new BT(`Node ${t.name} is scheduled to run after its child ${e.name}.`)}if(!a(t))for(let e of t.inputs){if(!n.has(e.name))throw new BT(`Input ${e.name} of node ${t.name} is unreachable.`);if(n.get(e.name)>n.get(t.name))throw new BT(`Node ${t.name} is scheduled to run before its input ${e.name}.`)}}}(f,l),f}var BT=class extends Error{constructor(e){super(`NodesExecutionOrderError: ${e}`)}};var WT=new Set(["Switch","Merge","Enter","Exit","NextIteration","StatelessIf","StatelessWhile","if","While"]),VT=new Set(["NonMaxSuppressionV2","NonMaxSuppressionV3","NonMaxSuppressionV5","Where"]),UT=new Set(["HashTable","HashTableV2","LookupTableImport","LookupTableImportV2","LookupTableFind","LookupTableFindV2","LookupTableSize","LookupTableSizeV2"]);function GT(e){return WT.has(e.op)}function HT(e){return VT.has(e.op)}function jT(e){return UT.has(e.op)}var qT=class e{get weightIds(){return this.parent?this.parent.weightIds:this._weightIds}get functionExecutorMap(){return this.parent?this.parent.functionExecutorMap:this._functionExecutorMap}get weightMap(){return this.parent?this.parent.weightMap:this._weightMap}set weightMap(e){let t=Object.keys(e).map(t=>e[t].map(e=>e.id));this._weightIds=[].concat(...t),this._weightMap=e}set resourceManager(e){this._resourceManager=e}get inputs(){return this._inputs.map(e=>({name:e.name,shape:e.attrParams.shape?e.attrParams.shape.value:void 0,dtype:e.attrParams.dtype?e.attrParams.dtype.value:void 0}))}get outputs(){return this._outputs.map(e=>({name:e.name,shape:e.attrParams.shape?e.attrParams.shape.value:void 0,dtype:e.attrParams.dtype?e.attrParams.dtype.value:void 0}))}get inputNodes(){return this._inputs.map(e=>e.signatureKey||e.name)}get outputNodes(){return this._outputs.map(e=>{let t=e.signatureKey||e.name;return e.defaultOutput?`${t}:${e.defaultOutput}`:t})}get functions(){return Object.keys(this._functions).reduce((e,t)=>(e[t]=this._functions[t].signature,e),{})}constructor(t,n){this.graph=t,this.parent=n,this.compiledMap=new Map,this.parseNodeNameCache=new Map,this._weightMap={},this.SEPARATOR=",",this._functions={},this._functionExecutorMap={},this.keepIntermediateTensors=!1,this._outputs=t.outputs,this._inputs=t.inputs,this._initNodes=t.initNodes,this._signature=t.signature,this._functions=t.functions,null!=t.functions&&Object.keys(t.functions).forEach(n=>{this._functionExecutorMap[n]=new e(t.functions[n],this)})}getCompilationKey(e,t){let n=e.map(e=>e.name).sort(),r=t.map(e=>e.name).sort();return n.join(this.SEPARATOR)+"--"+r.join(this.SEPARATOR)}compile(e,t){let n=PT(e,t,this.weightMap,this._initNodes),{missingInputs:r,dynamicNode:a,syncInputs:s}=n;if(null!=a)throw new Error(`This execution contains the node '${a.name}', which has the dynamic op '${a.op}'. Please use model.executeAsync() instead. Alternatively, to avoid the dynamic ops, specify the inputs [${s}]`);if(r.length>0){let n=t.map(e=>e.name),a=Object.keys(e);throw new Error(`Cannot compute the outputs [${n}] from the provided inputs [${a}]. Missing the following inputs: [${r}]`)}let i=zT(this.graph,n),o=function(e){let t=new Map(e.map((e,t)=>[e.name,t])),n=Number.MAX_SAFE_INTEGER,r=e.map((e,t)=>GT(e)?n:t),a=e=>r[t.get(e.name)]??-1,s=e.map((e,t)=>e.children.map(a).reduce((e,t)=>Math.max(e,t),r[t])),i=new Map;for(let t=0;t<e.length;++t){let r=s[t];if(r===n)continue;let a=e[t],o=e[r];i.has(o.name)||i.set(o.name,[]),i.get(o.name).push(a)}return i}(i);return{orderedNodes:i,nodeLiveUntilMap:o}}cloneAndKeepTensor(e){if(null==e)return null;let t=e.clone();return ti(t),t}cloneTensorList(e){return e?e.map(e=>this.cloneAndKeepTensor(e)):null}cloneTensorMap(e){return Object.fromEntries(Object.entries(e).map(([e,t])=>[e,this.cloneTensorList(t)]))}execute(e,t){this.disposeIntermediateTensors(),e=this.mapInputs(e);let n=Object.keys(e).sort();this.checkInputs(e),this.checkInputShapeAndType(e),t=this.mapOutputs(t),this.checkOutputs(t);let r=n.map(e=>this.graph.nodes[kN(e)[0]]),a=t.map(e=>kN(e)[0]),s=new Set(a),i=a.map(e=>this.graph.nodes[e]);0===i.length&&(i=this._outputs);let o=this.getCompilationKey(r,i),l=this.compiledMap.get(o);null==l&&(l=this.compile(e,i),this.compiledMap.set(o,l));try{this.keepIntermediateTensors=Le().getBool("KEEP_INTERMEDIATE_TENSORS")}catch(e){this.keepIntermediateTensors=!1,console.warn(e.message)}let u={},h={};return Qs(()=>{let n=new LT(this.weightMap,u,h,this.functionExecutorMap,this.parseNodeNameCache),r=Object.assign({},this.weightMap);this.keepIntermediateTensors&&(this.clonedTensorsMap=this.cloneTensorMap(this.weightMap)),Object.keys(e).forEach(t=>{let[a,s]=kN(t,n),i=[];i[s]=e[t],r[a]=i,this.keepIntermediateTensors&&(this.clonedTensorsMap[a]=this.cloneTensorList(i))});let a=this.getFrozenTensorIds(r),{orderedNodes:i,nodeLiveUntilMap:o}=l;for(let e of i){if(r[e.name])continue;let t=OT(e,r,n,this._resourceManager);if(va.isPromise(t))throw new Error(`The execution of the op '${e.op}' returned a promise. Please use model.executeAsync() instead.`);r[e.name]=t,this.keepIntermediateTensors&&(this.clonedTensorsMap[e.name]=this.cloneTensorList(t)),this.checkTensorForDisposalWithNodeLiveUntilInfo(e,r,n,a,s,o.get(e.name))}return null==this.parent&&n.dispose(a),t.map(e=>bN(e,r,n))})}getFrozenTensorIds(e){let t=[].concat.apply([],Object.keys(e).map(t=>e[t]).map(e=>e.map(e=>e.id)));return new Set(t)}checkTensorForDisposal(e,t,n,r,a,s,i){if(!GT(t)&&!s.has(e)){for(let r of n[e])null!=r&&(i[r.id]=(i[r.id]||0)+t.children.length);for(let e of t.inputs){if(GT(e))continue;let t=xN(e.name,n,r);if(null!=t)for(let e of t){if(!e||e.kept||a.has(e.id))continue;let t=i[e.id];1===t?(e.dispose(),delete i[e.id]):null!=t&&i[e.id]--}}}}checkTensorForDisposalWithNodeLiveUntilInfo(e,t,n,r,a,s){function i(e){return GT(e)||a.has(e.name)}if(!GT(e)&&null!=s)for(let e of s){if(i(e))continue;let a=xN(e.name,t,n);for(let e of a)!e||e.kept||r.has(e.id)||e.dispose()}}async executeAsync(e,t){return this._executeAsync(e,t)}disposeIntermediateTensors(){this.clonedTensorsMap&&(Object.values(this.clonedTensorsMap).forEach(e=>{for(let t of e)t&&!t.isDisposed&&t.dispose()}),this.clonedTensorsMap=null)}getIntermediateTensors(){return this.clonedTensorsMap}async _executeAsync(e,t,n=!1,r={},a={}){this.disposeIntermediateTensors(),n||(e=this.mapInputs(e),this.checkInputs(e),this.checkInputShapeAndType(e),t=this.mapOutputs(t),this.checkOutputs(t));try{this.keepIntermediateTensors=Le().getBool("KEEP_INTERMEDIATE_TENSORS")}catch(e){this.keepIntermediateTensors=!1,console.warn(e.message)}let s=new LT(this.weightMap,r,a,this.functionExecutorMap,this.parseNodeNameCache);this.keepIntermediateTensors&&(this.clonedTensorsMap=this.cloneTensorMap(this.weightMap));let i=await this.executeWithControlFlow(e,s,t,n),o=t.map(e=>bN(e,i,s)),l=o.map(e=>e.id),u=Object.keys(e).map(t=>e[t].id),h=new Set([...l,...u,...this.weightIds]);return Object.values(i).forEach(e=>{e.forEach(e=>{e&&!e.isDisposed&&!h.has(e.id)&&e.dispose()})}),null==this.parent&&s.dispose(h),o}async executeFunctionAsync(e,t,n){let r=e.reduce((e,t,n)=>(e[this.inputs[n].name]=t,e),{});return this._executeAsync(r,this.outputNodes,!0,t,n)}async executeWithControlFlow(e,t,n,r){let a=Object.keys(e),s=a.map(e=>this.graph.nodes[kN(e)[0]]),i=n.map(e=>kN(e)[0]),o=new Set(i),l=i.map(e=>this.graph.nodes[e]);0===l.length&&(l=this._outputs);let{usedNodes:u,missingInputs:h,dynamicNode:d,syncInputs:p}=PT(e,l,this.weightMap,this._initNodes),c=[...s,...this.graph.weights,...this._initNodes||[]].map(e=>({node:e,contexts:t.currentContext})),f=Object.assign({},this.weightMap);Object.keys(e).forEach(t=>{let[n,r]=kN(t),a=[];a[r]=e[t],f[n]=a});let m={},g=this.getFrozenTensorIds(f),y={};for(;c.length>0;){let e=this.processStack(s,c,t,f,y,g,o,m,u);await Promise.all(e)}null==d&&!r&&console.warn("This model execution did not contain any nodes with control flow or dynamic output shapes. You can use model.execute() instead.");let b=l.filter(e=>!GT(e)&&!bN(e.name,f,t)).map(e=>e.name);if(b.length>0){let e="";throw null!=d&&(e=`Alternatively, to avoid the dynamic ops, use model.execute() and specify the inputs [${p}]`),new Error(`Cannot compute the outputs [${b}] from the provided inputs [${a}]. Consider providing the following inputs: [${h}]. ${e}`)}return f}processStack(e,t,n,r,a,s,i,o,l){let u=[];for(;t.length>0;){let e=t.pop();n.currentContext=e.contexts;let h="";if("Enter"===e.node.op&&yN("isConstant",e.node,r,n)&&([h]=vN(e.node.name,n)),null==r[e.node.name]){let d=OT(e.node,r,n,this._resourceManager);h||([h]=vN(e.node.name,n));let p=n.currentContext;va.isPromise(d)?u.push(d.then(u=>(r[h]=u,this.keepIntermediateTensors&&(this.clonedTensorsMap[h]=this.cloneTensorList(u)),n.currentContext=p,this.checkTensorForDisposal(h,e.node,r,n,s,i,o),this.processChildNodes(e.node,t,n,r,a,l),u))):(r[h]=d,this.keepIntermediateTensors&&(this.clonedTensorsMap[h]=this.cloneTensorList(d)),this.checkTensorForDisposal(h,e.node,r,n,s,i,o),this.processChildNodes(e.node,t,n,r,a,l))}else this.processChildNodes(e.node,t,n,r,a,l)}return u}processChildNodes(e,t,n,r,a,s){e.children.forEach(e=>{let[i]=vN(e.name,n);a[i]||!s.has(e.name)||("Merge"===e.op?e.inputNames.some(e=>!!bN(e,r,n))&&(a[i]=!0,t.push({contexts:n.currentContext,node:e})):e.inputNames.every(e=>!!bN(e,r,n))&&(a[i]=!0,t.push({contexts:n.currentContext,node:e})))})}dispose(){Object.keys(this.weightMap).forEach(e=>this.weightMap[e].forEach(e=>e.dispose()))}checkInputShapeAndType(e){Object.keys(e).forEach(t=>{let n=e[t],[r]=kN(t),a=this.graph.nodes[r];if(a.attrParams.shape&&a.attrParams.shape.value){let e=a.attrParams.shape.value,t=e.length===n.shape.length&&n.shape.every((t,n)=>-1===e[n]||e[n]===t);va.assert(t,()=>`The shape of dict['${a.name}'] provided in model.execute(dict) must be [${e}], but was [${n.shape}]`)}a.attrParams.dtype&&a.attrParams.dtype.value&&va.assert(n.dtype===a.attrParams.dtype.value,()=>`The dtype of dict['${a.name}'] provided in model.execute(dict) must be ${a.attrParams.dtype.value}, but was ${n.dtype}`)})}mapInputs(e){var t,n;let r={};for(let a in e){let s=null===(n=null===(t=this._signature)||void 0===t?void 0:t.inputs)||void 0===n?void 0:n[a];null!=s?r[s.name]=e[a]:r[a]=e[a]}return r}checkInputs(e){let t=Object.keys(e).filter(e=>{let[t]=kN(e);return null==this.graph.nodes[t]});if(t.length>0)throw new Error(`The dict provided in model.execute(dict) has keys: [${t}] that are not part of graph`)}mapOutputs(e){return e.map(e=>{var t,n;let r=null===(n=null===(t=this._signature)||void 0===t?void 0:t.outputs)||void 0===n?void 0:n[e];return null!=r?r.name:e},{})}checkOutputs(e){e.forEach(e=>{let[t]=kN(e);if(!this.graph.nodes[t])throw new Error(`The output '${e}' is not found in the graph`)})}},KT=class{constructor(e={},t={}){this.hashTableNameToHandle=e,this.hashTableMap=t}addHashTable(e,t){this.hashTableNameToHandle[e]=t.handle,this.hashTableMap[t.id]=t}getHashTableHandleByName(e){return this.hashTableNameToHandle[e]}getHashTableById(e){return this.hashTableMap[e]}dispose(){for(let e in this.hashTableMap)this.hashTableMap[e].clearAndClose(),delete this.hashTableMap[e];for(let e in this.hashTableNameToHandle)this.hashTableNameToHandle[e].dispose(),delete this.hashTableNameToHandle[e]}},XT="?tfjs-format=file",ZT="model.json",YT=class{get modelVersion(){return this.version}get inputNodes(){return this.executor.inputNodes}get outputNodes(){return this.executor.outputNodes}get inputs(){return this.executor.inputs}get outputs(){return this.executor.outputs}get weights(){return this.executor.weightMap}get metadata(){return this.artifacts.userDefinedMetadata}get modelSignature(){return this.signature}get modelStructuredOutputKeys(){return this.structuredOutputKeys}constructor(e,t={},n=Pc){this.modelUrl=e,this.loadOptions=t,this.version="n/a",this.io=n,null==t&&(this.loadOptions={}),this.resourceManager=new KT}findIOHandler(){let e=this.modelUrl;if(null!=e.load)this.handler=e;else if(null!=this.loadOptions.requestInit)this.handler=this.io.browserHTTPRequest(e,this.loadOptions);else{let t=this.io.getLoadHandlers(e,this.loadOptions);if(0===t.length)t.push(this.io.browserHTTPRequest(e,this.loadOptions));else if(t.length>1)throw new Error(`Found more than one (${t.length}) load handlers for URL '${[e]}'`);this.handler=t[0]}}load(){if(this.findIOHandler(),null==this.handler.load)throw new Error("Cannot proceed with model loading because the IOHandler provided does not have the `load` method implemented.");let e=this.handler.load();return va.isPromise(e)?e.then(e=>null==e.getWeightStream?this.loadSync(e):this.loadStreaming(e)):this.loadSync(e)}loadSync(e){let t=this.io.decodeWeights(e.weightData,e.weightSpecs);return this.loadWithWeightMap(e,t)}async loadStreaming(e){if(null==e.getWeightStream)throw new Error("Model artifacts missing streamWeights function");let t=await bi(e.getWeightStream(),e.weightSpecs);return this.loadWithWeightMap(e,t)}loadWithWeightMap(e,t){this.artifacts=e;let n=this.artifacts.modelTopology,r=this.artifacts.signature;if(null!=this.artifacts.userDefinedMetadata){let e=this.artifacts.userDefinedMetadata;null!=e.signature&&(r=e.signature),null!=e.structuredOutputKeys&&(this.structuredOutputKeys=e.structuredOutputKeys)}if(this.signature=r,this.version=`${n.versions.producer}.${n.versions.minConsumer}`,this.executor=new qT(uT.Instance.transformGraph(n,this.signature)),this.executor.weightMap=this.convertTensorMapToTensorsMap(t),this.executor.resourceManager=this.resourceManager,null!=e.modelInitializer&&null!=e.modelInitializer.node){let t=uT.Instance.transformGraph(e.modelInitializer);this.initializer=new qT(t),this.initializer.weightMap=this.executor.weightMap,this.initializer.resourceManager=this.resourceManager,this.initializerSignature=e.initializerSignature}return!0}async save(e,t){if("string"==typeof e){let t=this.io.getSaveHandlers(e);if(0===t.length)throw new Error(`Cannot find any save handlers for URL '${e}'`);if(t.length>1)throw new Error(`Found more than one (${t.length}) save handlers for URL '${e}'`);e=t[0]}if(null==e.save)throw new Error("GraphModel.save() cannot proceed because the IOHandler provided does not have the `save` attribute defined.");return e.save(this.artifacts)}addStructuredOutputNames(e){if(this.structuredOutputKeys){let t={};return(e instanceof ts?[e]:e).forEach((e,n)=>t[this.structuredOutputKeys[n]]=e),t}return e}predict(e,t){let n=this.execute(e,this.outputNodes);return this.addStructuredOutputNames(n)}async predictAsync(e,t){let n=await this.executeAsync(e,this.outputNodes);return this.addStructuredOutputNames(n)}normalizeInputs(e){var t;if(!(e instanceof ts||Array.isArray(e))){let n=null===(t=this.signature)||void 0===t?void 0:t.inputs;if(null!=n)for(let t in n){let r=n[t];null!=r.resourceId&&(e[t]=this.resourceIdToCapturedInput[r.resourceId])}return e}e=Array.isArray(e)?e:[e];let n=Object.keys(this.resourceIdToCapturedInput).length;if(e.length+n!==this.inputNodes.length)throw new Error(`Input tensor count mismatch, the graph model has ${this.inputNodes.length-n} non-resource placeholders, while there are ${e.length} input tensors provided.`);let r=0;return this.inputNodes.reduce((t,n)=>{var a,s,i;let o=null===(i=null===(s=null===(a=this.signature)||void 0===a?void 0:a.inputs)||void 0===s?void 0:s[n])||void 0===i?void 0:i.resourceId;return t[n]=null!=o?this.resourceIdToCapturedInput[o]:e[r++],t},{})}normalizeOutputs(e){return e=e||this.outputNodes,Array.isArray(e)?e:[e]}executeInitializerGraph(){return null==this.initializer?[]:null==this.initializerSignature?this.initializer.execute({},[]):this.initializer.execute({},Object.keys(this.initializerSignature.outputs))}async executeInitializerGraphAsync(){return null==this.initializer?[]:null==this.initializerSignature?this.initializer.executeAsync({},[]):this.initializer.executeAsync({},Object.keys(this.initializerSignature.outputs))}setResourceIdToCapturedInput(e){if(this.resourceIdToCapturedInput={},this.initializerSignature){let t=this.initializerSignature.outputs,n=Object.keys(t);for(let r=0;r<n.length;r++){let a=t[n[r]];this.resourceIdToCapturedInput[a.resourceId]=e[r]}}}execute(e,t){null==this.resourceIdToCapturedInput&&this.setResourceIdToCapturedInput(this.executeInitializerGraph()),e=this.normalizeInputs(e),t=this.normalizeOutputs(t);let n=this.executor.execute(e,t);return n.length>1?n:n[0]}async executeAsync(e,t){null==this.resourceIdToCapturedInput&&this.setResourceIdToCapturedInput(await this.executeInitializerGraphAsync()),e=this.normalizeInputs(e),t=this.normalizeOutputs(t);let n=await this.executor.executeAsync(e,t);return n.length>1?n:n[0]}getIntermediateTensors(){return this.executor.getIntermediateTensors()}disposeIntermediateTensors(){this.executor.disposeIntermediateTensors()}convertTensorMapToTensorsMap(e){return Object.keys(e).reduce((t,n)=>(t[n]=[e[n]],t),{})}dispose(){this.executor.dispose(),this.initializer&&(this.initializer.dispose(),this.resourceIdToCapturedInput&&ei(this.resourceIdToCapturedInput)),this.resourceManager.dispose()}};async function JT(e,t={},n=Pc){if(null==e)throw new Error("modelUrl in loadGraphModel() cannot be null. Please provide a url or an IOHandler that loads the model");null==t&&(t={}),t.fromTFHub&&"string"==typeof e&&(e=function(e){return e.endsWith("/")||(e+="/"),`${e}${ZT}${XT}`}(e));let r=new YT(e,t,n);return await r.load(),r}function QT(e){if(null==e)throw new Error("modelUrl in loadGraphModelSync() cannot be null. Please provide model artifacts or an IOHandler that loads the model");let t;if(e instanceof Array){let[n,r]=e;if(!n)throw new Error("modelJSON must be the first element of the array");if(!(r&&r instanceof ArrayBuffer))throw new Error("An ArrayBuffer of weights must be the second element of the array");if(!("modelTopology"in n))throw new Error("Model JSON is missing 'modelTopology'");if(!("weightsManifest"in n))throw new Error("Model JSON is missing 'weightsManifest'");let a=Pc.getWeightSpecs(n.weightsManifest),s=Pc.getModelArtifactsForJSONSync(n,a,r);t=Pc.fromMemorySync(s)}else if("load"in e)t=e;else{if(!("modelTopology"in e&&"weightSpecs"in e&&"weightData"in e))throw new Error("Unknown model format");t=Pc.fromMemorySync(e)}let n=new YT(t);return n.load(),n}var eC="4.22.0",tC={};f(tC,{CSVDataset:()=>qC,Dataset:()=>DC,FileDataSource:()=>iE,TextLineDataset:()=>BC,URLDataSource:()=>oE,array:()=>OC,csv:()=>lE,func:()=>uE,generator:()=>hE,microphone:()=>pE,version_data:()=>cE,webcam:()=>dE,zip:()=>LC});var nC=m(T()),rC=m(T());function aC(e,t,n=new Map,r=new Set){if(null==e)return null;if("function"==typeof Blob&&e instanceof Blob)return e.slice();if(r.has(e))throw new Error("Circular references are not supported.");if(n.has(e))return n.get(e);let a=t(e);if(a.recurse&&null!==a.value)throw new Error("A deep map function may not return both a value and recurse=true.");if(a.recurse){if(uC(e)){let a=Array.isArray(e)?[]:{};r.add(e);for(let s in e){let i=aC(e[s],t,n,r);a[s]=i}return r.delete(e),e.__proto__&&(a.__proto__=e.__proto__),a}throw new Error(`Can't recurse into non-iterable type: ${e}`)}return n.set(e,a.value),a.value}function sC(e,t=oC){return iC(e,t)}function iC(e,t,n=new Set){let r=e[0];if(n.has(r))throw new Error("Circular references are not supported.");let a=t(e);if(a.recurse&&null!==a.value)throw new Error("A deep zip function may not return both a value and recurse=true.");if(a.recurse){if(uC(r)){let a=Array.isArray(r)?[]:{};n.add(r);for(let s in r){let r=iC(e.map(e=>e[s]),t,n);a[s]=r}return n.delete(r),a}throw new Error(`Can't recurse into non-iterable type: ${r}`)}return a.value}function oC(e){return null===e?null:uC(e[0])?{value:null,recurse:!0}:{value:e,recurse:!1}}async function lC(e,t){let n=new Map;aC(e,t,n);for(let e of Array.from(n.keys())){let t=n.get(e);if(va.isPromise(t)){let r=await t;n.set(e,r)}}return aC(e,t,n)}function uC(e){let t=!1;if(Le().get("IS_BROWSER"))t=e instanceof TextDecoder;else{let{StringDecoder:n}=C();t=e instanceof n}return null!=e&&!ArrayBuffer.isView(e)&&(Array.isArray(e)||"object"==typeof e&&!(e instanceof ts)&&!(e instanceof Promise)&&!t)}function hC(e){return function(e,t){return aC(e,t)}(e,dC)}function dC(e){return e instanceof ts?{value:e.clone(),recurse:!1}:uC(e)?{value:null,recurse:!0}:{value:e,recurse:!1}}var pC=class{constructor(e){if(this.capacity=e,this.begin=0,this.end=0,null==e)throw new RangeError("Can't create a ring buffer of unknown capacity.");if(e<1)throw new RangeError("Can't create ring buffer of capacity < 1.");this.data=new Array(e),this.doubledCapacity=2*e}wrap(e){for(;e<0;)e+=this.doubledCapacity;return e%this.doubledCapacity}get(e){if(e<0)throw new RangeError("Can't get item at a negative index.");return this.data[e%this.capacity]}set(e,t){if(e<0)throw new RangeError("Can't set item at a negative index.");this.data[e%this.capacity]=t}length(){let e=this.end-this.begin;return e<0&&(e=this.doubledCapacity+e),e}isFull(){return this.length()===this.capacity}isEmpty(){return 0===this.length()}push(e){if(this.isFull())throw new RangeError("Ring buffer is full.");this.set(this.end,e),this.end=this.wrap(this.end+1)}pushAll(e){for(let t of e)this.push(t)}pop(){if(this.isEmpty())throw new RangeError("Ring buffer is empty.");this.end=this.wrap(this.end-1);let e=this.get(this.end);return this.set(this.end,void 0),e}unshift(e){if(this.isFull())throw new RangeError("Ring buffer is full.");this.begin=this.wrap(this.begin-1),this.set(this.begin,e)}shift(){if(this.isEmpty())throw new RangeError("Ring buffer is empty.");let e=this.get(this.begin);return this.set(this.begin,void 0),this.begin=this.wrap(this.begin+1),e}shuffleExcise(e){if(this.isEmpty())throw new RangeError("Ring buffer is empty.");let t=this.wrap(this.begin+e),n=this.get(t);return this.set(t,this.pop()),n}},cC=class e extends pC{constructor(){super(e.INITIAL_CAPACITY)}isFull(){return!1}push(e){super.isFull()&&this.expand(),super.push(e)}unshift(e){super.isFull()&&this.expand(),super.unshift(e)}expand(){let e=2*this.capacity,t=new Array(e),n=this.length();for(let e=0;e<n;e++)t[e]=this.get(this.wrap(this.begin+e));this.data=t,this.capacity=e,this.doubledCapacity=2*this.capacity,this.begin=0,this.end=n}};function fC(e){return new bC(e)}function mC(e){return new xC(e)}cC.INITIAL_CAPACITY=32;var gC,yC=class{async toArray(){let e=[],t=await this.next();for(;!t.done;)e.push(t.value),t=await this.next();return e}async toArrayForTest(){let e=this.prefetch(100),t=[],n=await e.next();for(;!n.done;)t.push(n.value),n=await e.next();return t}async resolveFully(){let e=await this.next();for(;!e.done;)e=await this.next()}async resolveWhile(e){let t=await this.next(),n=e(t.value);for(;!t.done&&n;)t=await this.next(),n=e(t.value)}handleErrors(e){return new NC(this,e)}filter(e){return new SC(this,e)}map(e){return new _C(this,e)}mapAsync(e){return new TC(this,e)}serialMapAsync(e){return new TC(this,e).serial()}flatmap(e){return new EC(this,e)}async forEachAsync(e){return this.map(e).resolveFully()}async serialForEach(e){return this.serialMapAsync(e).resolveWhile(e=>!0===e)}rowMajorBatch(e,t=!0){return new IC(this,e,t)}columnMajorBatch(e,t=!0,n=oC){return this.rowMajorBatch(e,t).map(e=>sC(e,n))}concatenate(e,t){return new AC(fC([this,e]),t)}take(e){return e<0||null==e?this:new kC(this,e)}skip(e){return e<0||null==e?this:new wC(this,e)}prefetch(e){return new RC(this,e)}shuffle(e,t){return new FC(this,e,t)}serial(){return new vC(this)}},bC=class extends yC{constructor(e){super(),this.items=e,this.trav=0}summary(){return`Array of ${this.items.length} items`}async next(){if(this.trav>=this.items.length)return{value:null,done:!0};let e=this.items[this.trav];return this.trav++,{value:hC(e),done:!1}}},xC=class extends yC{constructor(e){super(),this.nextFn=e}summary(){return"Function call"}async next(){try{return this.nextFn()}catch(e){throw e.message=`Error thrown while iterating through a dataset: ${e.message}`,e}}},vC=class extends yC{constructor(e){super(),this.upstream=e,this.lastRead=Promise.resolve({value:null,done:!1})}summary(){return`${this.upstream.summary()} -> Serial`}async next(){return this.lastRead=this.lastRead.then(()=>this.serialNext()),this.lastRead}async serialNext(){return this.upstream.next()}},wC=class extends yC{constructor(e,t){super(),this.upstream=e,this.maxCount=t,this.count=0,this.lastRead=Promise.resolve({value:null,done:!1})}summary(){return`${this.upstream.summary()} -> Skip`}async next(){return this.lastRead=this.lastRead.then(()=>this.serialNext()),this.lastRead}async serialNext(){for(;this.count++<this.maxCount;){let e=await this.upstream.next();if(e.done)return e;ei(e.value)}return this.upstream.next()}},kC=class extends yC{constructor(e,t){super(),this.upstream=e,this.maxCount=t,this.count=0}summary(){return`${this.upstream.summary()} -> Take`}async next(){return this.count++>=this.maxCount?{value:null,done:!0}:this.upstream.next()}},IC=class extends yC{constructor(e,t,n=!0){super(),this.upstream=e,this.batchSize=t,this.enableSmallLastBatch=n,this.lastRead=Promise.resolve({value:null,done:!1})}summary(){return`${this.upstream.summary()} -> RowMajorBatch`}async next(){return this.lastRead=this.lastRead.then(()=>this.serialNext()),this.lastRead}async serialNext(){let e=[];for(;e.length<this.batchSize;){let t=await this.upstream.next();if(t.done)return this.enableSmallLastBatch&&e.length>0?{value:e,done:!1}:{value:null,done:!0};e.push(t.value)}return{value:e,done:!1}}},SC=class extends yC{constructor(e,t){super(),this.upstream=e,this.predicate=t,this.lastRead=Promise.resolve({value:null,done:!1})}summary(){return`${this.upstream.summary()} -> Filter`}async next(){return this.lastRead=this.lastRead.then(()=>this.serialNext()),this.lastRead}async serialNext(){for(;;){let e=await this.upstream.next();if(e.done||this.predicate(e.value))return e;ei(e.value)}}},_C=class extends yC{constructor(e,t){super(),this.upstream=e,this.transform=t}summary(){return`${this.upstream.summary()} -> Map`}async next(){let e=await this.upstream.next();if(e.done)return{value:null,done:!0};let t=hs.getTensorsInContainer(e.value),n=this.transform(e.value),r=hs.getTensorsInContainer(n);for(let e of t)hs.isTensorInList(e,r)||e.dispose();return{value:n,done:!1}}},NC=class extends yC{constructor(e,t){super(),this.upstream=e,this.handler=t,this.count=0,this.lastRead=Promise.resolve({value:null,done:!1})}summary(){return`${this.upstream.summary()} -> handleErrors`}async next(){return this.lastRead=this.lastRead.then(()=>this.serialNext()),this.lastRead}async serialNext(){for(;;)try{return await this.upstream.next()}catch(e){if(!this.handler(e))return{value:null,done:!0}}}},TC=class extends yC{constructor(e,t){super(),this.upstream=e,this.transform=t}summary(){return`${this.upstream.summary()} -> AsyncMap`}async next(){let e=await this.upstream.next();if(e.done)return{value:null,done:!0};let t=hs.getTensorsInContainer(e.value),n=await this.transform(e.value),r=hs.getTensorsInContainer(n);for(let e of t)hs.isTensorInList(e,r)||e.dispose();return{value:n,done:!1}}},CC=class extends yC{constructor(){super(),this.outputQueue=new cC,this.lastRead=Promise.resolve({value:null,done:!1})}async next(){return this.lastRead=this.lastRead.then(()=>this.serialNext()),this.lastRead}async serialNext(){for(;0===this.outputQueue.length();)if(!await this.pump())return{value:null,done:!0};return{value:this.outputQueue.shift(),done:!1}}},EC=class extends CC{constructor(e,t){super(),this.upstream=e,this.transform=t}summary(){return`${this.upstream.summary()} -> Flatmap`}async pump(){let e=await this.upstream.next();if(e.done)return!1;let t=hs.getTensorsInContainer(e.value),n=this.transform(e.value),r=hs.getTensorsInContainer(n);this.outputQueue.pushAll(n);for(let e of t)hs.isTensorInList(e,r)||e.dispose();return!0}},AC=class extends yC{constructor(e,t){super(),this.baseErrorHandler=t,this.lastRead=null,this.iterator=null,this.moreIterators=e}summary(){return"TODO: fill in upstream of chained summaries -> Chained"}async next(){return this.lastRead=this.readFromChain(this.lastRead),this.lastRead}async readFromChain(e){if(await e,null==this.iterator){let e=await this.moreIterators.next();if(e.done)return{value:null,done:!0};this.iterator=e.value,null!=this.baseErrorHandler&&(this.iterator=this.iterator.handleErrors(this.baseErrorHandler))}let t=await this.iterator.next();return t.done?(this.iterator=null,this.readFromChain(e)):t}};!function(e){e[e.FAIL=0]="FAIL",e[e.SHORTEST=1]="SHORTEST",e[e.LONGEST=2]="LONGEST"}(gC||(gC={}));var $C=class extends yC{constructor(e,t=gC.FAIL){super(),this.iterators=e,this.mismatchMode=t,this.count=0,this.currentPromise=null}summary(){return"{TODO: fill in upstream of zip summaries} -> Zip"}async nextState(e){await e;let t=0,n=0;let r=await lC(this.iterators,function(e){return e instanceof yC?{value:e.next().then(e=>(t++,e.done&&n++,e.value)),recurse:!1}:{value:null,recurse:!0}});if(t===n)return{value:null,done:!0};if(n>0)switch(this.mismatchMode){case gC.FAIL:throw new Error(`Zipped streams should have the same length. Mismatched at element ${this.count}.`);case gC.SHORTEST:return{value:null,done:!0};case gC.LONGEST:}return this.count++,{value:r,done:!1}}async next(){return this.currentPromise=this.nextState(this.currentPromise),this.currentPromise}},RC=class extends yC{constructor(e,t){super(),this.upstream=e,this.bufferSize=t,this.buffer=new pC(t)}summary(){return`${this.upstream.summary()} -> Prefetch`}refill(){for(;!this.buffer.isFull();){let e=this.upstream.next();this.buffer.push(e)}}next(){return this.refill(),this.buffer.shift()}},FC=class extends RC{constructor(e,t,n){super(e,t),this.upstream=e,this.windowSize=t,this.upstreamExhausted=!1,this.random=rC.alea(n||va.now().toString()),this.lastRead=Promise.resolve({value:null,done:!1})}async next(){return this.lastRead=this.lastRead.then(()=>this.serialNext()),this.lastRead}randomInt(e){return Math.floor(this.random()*e)}chooseIndex(){return this.randomInt(this.buffer.length())}async serialNext(){for(this.upstreamExhausted||this.refill();!this.buffer.isEmpty();){let e=this.chooseIndex(),t=await this.buffer.shuffleExcise(e);if(!t.done)return this.refill(),t;this.upstreamExhausted=!0}return{value:null,done:!0}}},DC=class{constructor(){this.size=null}batch(e,t=!0){let n,r=this;return va.assert(e>0,()=>`batchSize needs to be positive, but it is\n      ${e}`),n=this.size===1/0||null==this.size?this.size:t?Math.ceil(this.size/e):Math.floor(this.size/e),MC(async()=>(await r.iterator()).columnMajorBatch(e,t,PC),n)}concatenate(e){let t,n=this;return t=this.size===1/0||e.size===1/0?1/0:null!=this.size&&null!=e.size?this.size+e.size:null,MC(async()=>(await n.iterator()).concatenate(await e.iterator()),t)}filter(e){let t,n=this;return t=this.size===1/0?1/0:null,MC(async()=>(await n.iterator()).filter(t=>Qs(()=>e(t))),t)}async forEachAsync(e){return(await this.iterator()).forEachAsync(e)}map(e){let t=this;return MC(async()=>(await t.iterator()).map(t=>Qs(()=>e(t))),this.size)}mapAsync(e){let t=this;return MC(async()=>(await t.iterator()).mapAsync(e),this.size)}prefetch(e){if(null==e)throw new RangeError("`Dataset.prefetch()` requires bufferSize to be specified.");let t=this;return MC(async()=>(await t.iterator()).prefetch(e),this.size)}repeat(e){let t,n=this;return t=null!=this.size&&e>0?this.size*e:0===e?0:null!=this.size&&(void 0===e||e<0)?1/0:null,MC(async()=>function(e,t){return new AC(e,t)}(mC(async()=>({value:await n.iterator(),done:!1})).take(e)),t)}skip(e){let t,n=this;return t=null!=this.size&&e>=0&&this.size>=e?this.size-e:null!=this.size&&(this.size<e||void 0===e||e<0)?0:null,MC(async()=>(await n.iterator()).skip(e),t)}shuffle(e,t,n=!0){if(null==e||e<0)throw null==this.size?new RangeError("`Dataset.shuffle()` requires bufferSize to be specified."):new RangeError(`\`Dataset.shuffle()\` requires bufferSize to be specified.  If your data fits in main memory (for regular JS objects), and/or GPU memory (for \`tf.Tensor\`s), consider setting bufferSize to the dataset size (${this.size} elements)`);let r=this,a=nC.alea(t||va.now().toString());return MC(async()=>{let t=a.int32();return n&&(t+=a.int32()),(await r.iterator()).shuffle(e,t.toString())},this.size)}take(e){let t,n=this;return t=null!=this.size&&this.size>e?e:null!=this.size&&this.size<=e?this.size:null,MC(async()=>(await n.iterator()).take(e),t)}async toArray(){if(this.size===1/0)throw new Error("Can not convert infinite data stream to array.");return(await this.iterator()).toArray()}async toArrayForTest(){if(this.size===1/0)throw new Error("Can not convert infinite data stream to array.");return(await this.iterator()).toArrayForTest()}};function MC(e,t=null){return new class extends DC{constructor(){super(...arguments),this.size=t}async iterator(){return e()}}}function OC(e){return MC(async()=>fC(e),e.length)}function LC(e){if(!uC(e))throw new Error("The argument to zip() must be an object or array.");let t;if(Array.isArray(e))for(let n=0;n<e.length;n++)t=null==t?e[n].size:Math.min(t,e[n].size);else if(e instanceof Object)for(let n in e)t=null==t?e[n].size:Math.min(t,e[n].size);return MC(async()=>function(e,t=gC.FAIL){return new $C(e,t)}(await lC(e,e=>{if(e instanceof DC)return{value:e.iterator(),recurse:!1};if(uC(e))return{value:null,recurse:!0};throw new Error("Leaves of the structure passed to zip() must be Datasets, not primitives.")}),gC.SHORTEST),t)}function PC(e){if(null===e)return null;return function(e){return null==e||function(e){return null===e||"object"!=typeof e&&"function"!=typeof e}(e)||Array.isArray(e)||"object"==typeof e&&e instanceof ts||va.isTypedArray(e)}(e[0])?{value:zC(e),recurse:!1}:{value:null,recurse:!0}}function zC(e){if(0===e.length)throw new Error("Can't make a batch of zero elements.");return e[0]instanceof ts?Td(e):Vs(e)}DC.MAX_BUFFER_SIZE=1e4;var BC=class extends DC{constructor(e){super(),this.input=e}async iterator(){return(await this.input.iterator()).decodeUTF8().split("\n").map(e=>(e.endsWith("\r")&&(e=e.slice(0,-1)),e))}},WC='"',VC=Symbol("out"),UC=Symbol("field"),GC=Symbol("quote"),HC=Symbol("quoteafterquote"),jC=Symbol("quoteinquote"),qC=class extends DC{async columnNames(){return this.columnNamesValidated||await this.setColumnNames(),this.configuredColumnsOnly?Object.keys(this.columnConfigs):this.fullColumnNames}async setColumnNames(){let e=await this.maybeReadHeaderLine();if(!this.fullColumnNames&&!e)throw new Error("Column names must be provided if there is no header line.");this.fullColumnNames&&e&&va.assert(e.length===this.fullColumnNames.length,()=>"The length of provided columnNames ("+this.fullColumnNames.length.toString()+") does not match the length of the header line read from file ("+e.length.toString()+")."),this.fullColumnNames||(this.fullColumnNames=e);let t=this.fullColumnNames.reduce((e,t)=>(e[t]=e[t]+1||1,e),{}),n=Object.keys(t).filter(e=>t[e]>1);if(va.assert(0===n.length,()=>"Duplicate column names found: "+n.toString()),this.columnConfigs)for(let e of Object.keys(this.columnConfigs))if(-1===this.fullColumnNames.indexOf(e))throw new Error('The key "'+e+'" provided in columnConfigs does not match any of the column names ('+this.fullColumnNames.toString()+").");this.columnNamesValidated=!0}async maybeReadHeaderLine(){if(this.hasHeader){let e=await(await this.base.iterator()).next();if(e.done)throw new Error("No data was found for CSV parsing.");let t=e.value;return this.parseRow(t,!1)}return null}constructor(e,t){super(),this.input=e,this.hasHeader=!0,this.fullColumnNames=null,this.columnNamesValidated=!1,this.columnConfigs=null,this.configuredColumnsOnly=!1,this.delimiter=",",this.delimWhitespace=!1,this.base=new BC(e),t||(t={}),this.hasHeader=!1!==t.hasHeader,this.fullColumnNames=t.columnNames,this.columnConfigs=t.columnConfigs,this.configuredColumnsOnly=t.configuredColumnsOnly,t.delimWhitespace?(va.assert(null==t.delimiter,()=>"Delimiter should not be provided when delimWhitespace is true."),this.delimWhitespace=!0,this.delimiter=" "):this.delimiter=t.delimiter?t.delimiter:","}async iterator(){this.columnNamesValidated||await this.setColumnNames();let e=await this.base.iterator();return this.hasHeader&&(e=e.skip(1)),e.map(e=>this.makeDataElement(e))}makeDataElement(e){let t=this.parseRow(e),n={},r={};for(let a=0;a<this.fullColumnNames.length;a++){let s=this.fullColumnNames[a],i=this.columnConfigs?this.columnConfigs[s]:null;if(!this.configuredColumnsOnly||i){let o=t[a],l=null;if(""===o)if(i&&void 0!==i.default)l=i.default;else{if(i&&(i.required||i.isLabel))throw new Error(`Required column ${s} is empty in this line: ${e}`);l=void 0}else{let e=Number(o);if(isNaN(e))l=i&&"bool"===i.dtype?this.getBoolean(o):o;else if(i&&i.dtype)switch(i.dtype){case"float32":default:l=e;break;case"int32":l=Math.floor(e);break;case"bool":l=this.getBoolean(o)}else l=e}i&&i.isLabel?r[s]=l:n[s]=l}}return 0===Object.keys(r).length?n:{xs:n,ys:r}}getBoolean(e){return"1"===e||"true"===e.toLowerCase()?1:0}parseRow(e,t=!0){let n=[],r=0,a=e.length,s=VC;for(let t=0;t<a;t++)switch(s){case VC:switch(e.charAt(t)){case WC:r=t+1,s=GC;break;case this.delimiter:if(r=t+1," "===this.delimiter&&this.delimWhitespace)break;n.push(""),s=VC;break;default:s=UC,r=t}break;case UC:if(e.charAt(t)===this.delimiter)n.push(e.substring(r,t)),s=VC,r=t+1;break;case GC:if(e.charAt(t)===WC)s=HC;break;case HC:switch(e.charAt(t)){case this.delimiter:n.push(e.substring(r,t-1)),s=VC,r=t+1;break;case WC:s=GC;break;default:s=jC}break;case jC:if(e.charAt(t)===WC)s=GC}if(s===HC?n.push(e.substring(r,a-1)):n.push(e.substring(r)),t&&n.length!==this.fullColumnNames.length)throw new Error(`Invalid row in csv file. Should have ${this.fullColumnNames.length} elements in a row, but got ${n}`);return n}},KC=class e extends yC{constructor(e){super(),this.microphoneConfig=e,this.isClosed=!1,this.fftSize=e.fftSize||1024;let t=Math.log2(this.fftSize);if(this.fftSize<0||t<4||t>14||!Number.isInteger(t))throw new Error(`Invalid fftSize: it must be a power of 2 between 2 to 4 and 2 to 14, but got ${this.fftSize}`);if(this.numFrames=e.numFramesPerSpectrogram||43,this.sampleRateHz=e.sampleRateHz,this.columnTruncateLength=e.columnTruncateLength||this.fftSize,this.audioTrackConstraints=e.audioTrackConstraints,this.smoothingTimeConstant=e.smoothingTimeConstant||0,this.includeSpectrogram=!1!==e.includeSpectrogram,this.includeWaveform=!0===e.includeWaveform,!this.includeSpectrogram&&!this.includeWaveform)throw new Error("Both includeSpectrogram and includeWaveform are false. At least one type of data should be returned.")}summary(){return"microphone"}static async create(t={}){if(!Le().get("IS_BROWSER"))throw new Error("microphone API is only supported in browser environment.");let n=new e(t);return await n.start(),n}async start(){try{this.stream=await navigator.mediaDevices.getUserMedia({audio:null==this.audioTrackConstraints||this.audioTrackConstraints,video:!1})}catch(e){throw new Error(`Error thrown while initializing video stream: ${e.message}`)}if(!this.stream)throw new Error("Could not obtain audio from microphone.");let e=window.AudioContext||window.webkitAudioContext;if(this.audioContext=new e,this.sampleRateHz){if(this.audioContext.sampleRate!==this.sampleRateHz)throw new Error(`Mismatch in sampling rate: Expected: ${this.sampleRateHz}; Actual: ${this.audioContext.sampleRate}`)}else this.sampleRateHz=this.audioContext.sampleRate;let t=this.audioContext.createMediaStreamSource(this.stream);this.analyser=this.audioContext.createAnalyser(),this.analyser.fftSize=2*this.fftSize,this.analyser.smoothingTimeConstant=this.smoothingTimeConstant,t.connect(this.analyser),this.freqData=new Float32Array(this.fftSize),this.timeData=new Float32Array(this.fftSize)}async next(){if(this.isClosed)return{value:null,done:!0};let e,t,n=await this.getAudioData();if(this.includeSpectrogram){let t=this.flattenQueue(n.freqDataQueue);e=this.getTensorFromAudioDataArray(t,[this.numFrames,this.columnTruncateLength,1])}if(this.includeWaveform){let e=this.flattenQueue(n.timeDataQueue);t=this.getTensorFromAudioDataArray(e,[this.numFrames*this.fftSize,1])}return{value:{spectrogram:e,waveform:t},done:!1}}async capture(){return(await this.next()).value}async getAudioData(){let e=[],t=[],n=0;return new Promise(r=>{let a=setInterval(()=>{this.includeSpectrogram&&(this.analyser.getFloatFrequencyData(this.freqData),this.freqData[0]===-1/0&&r({freqDataQueue:e,timeDataQueue:t}),e.push(this.freqData.slice(0,this.columnTruncateLength))),this.includeWaveform&&(this.analyser.getFloatTimeDomainData(this.timeData),t.push(this.timeData.slice())),++n===this.numFrames&&(clearInterval(a),r({freqDataQueue:e,timeDataQueue:t}))},this.fftSize/this.sampleRateHz*1e3)})}stop(){this.isClosed||(this.isClosed=!0,this.analyser.disconnect(),this.audioContext.close(),null!=this.stream&&this.stream.getTracks().length>0&&this.stream.getTracks()[0].stop())}toArray(){throw new Error("Can not convert infinite audio stream to array.")}getSampleRate(){return this.sampleRateHz}flattenQueue(e){let t=e[0].length,n=new Float32Array(e.length*t);return e.forEach((e,r)=>n.set(e,r*t)),n}getTensorFromAudioDataArray(e,t){let n=new Float32Array(va.sizeFromShape(t));return n.set(e,n.length-e.length),Vs(n,t)}},XC=class e extends yC{constructor(e,t){if(super(),this.webcamVideoElement=e,this.webcamConfig=t,this.isClosed=!0,this.resize=!1,this.needToResize())if(this.resize=!0,this.cropSize=[this.webcamConfig.resizeHeight,this.webcamConfig.resizeWidth],this.cropBoxInd=$d([0],"int32"),this.webcamConfig.centerCrop){let e=1*this.webcamConfig.resizeWidth/this.webcamVideoElement.width,t=1*this.webcamConfig.resizeHeight/this.webcamVideoElement.height,n=(1-e)/2,r=(1-t)/2,a=n+e,s=t+r;this.cropBox=Rd([r,n,s,a],[1,4])}else this.cropBox=Rd([0,0,1,1],[1,4])}summary(){return"webcam"}static async create(t,n={}){if(!Le().get("IS_BROWSER"))throw new Error("tf.data.webcam is only supported in browser environment.");if(!t){if(t=document.createElement("video"),!n.resizeWidth||!n.resizeHeight)throw new Error("Please provide webcam video element, or resizeWidth and resizeHeight to create a hidden video element.");t.width=n.resizeWidth,t.height=n.resizeHeight}let r=new e(t,n);return await r.start(),r}async start(){this.webcamConfig.facingMode&&va.assert("user"===this.webcamConfig.facingMode||"environment"===this.webcamConfig.facingMode,()=>`Invalid webcam facing mode: ${this.webcamConfig.facingMode}. Please provide 'user' or 'environment'`);try{this.stream=await navigator.mediaDevices.getUserMedia({video:{deviceId:this.webcamConfig.deviceId,facingMode:this.webcamConfig.facingMode?this.webcamConfig.facingMode:"user",width:this.webcamVideoElement.width,height:this.webcamVideoElement.height}})}catch(e){throw e.message=`Error thrown while initializing video stream: ${e.message}`,e}if(!this.stream)throw new Error("Could not obtain video from webcam.");try{this.webcamVideoElement.srcObject=this.stream}catch(e){console.log(e),this.webcamVideoElement.src=window.URL.createObjectURL(this.stream)}return this.webcamVideoElement.play(),this.isClosed=!1,new Promise(e=>{this.webcamVideoElement.onloadedmetadata=()=>{e()}})}async next(){if(this.isClosed)return{value:null,done:!0};let e;try{e=lf.fromPixels(this.webcamVideoElement)}catch(e){throw new Error(`Error thrown converting video to pixels: ${JSON.stringify(e)}`)}if(!this.resize)return{value:e,done:!1};try{return{value:this.cropAndResizeFrame(e),done:!1}}catch(e){throw new Error(`Error thrown cropping the video: ${e.message}`)}finally{e.dispose()}}needToResize(){return!(!this.webcamConfig.resizeWidth||!this.webcamConfig.resizeHeight||this.webcamVideoElement.width===this.webcamConfig.resizeWidth&&this.webcamVideoElement.height===this.webcamConfig.resizeHeight)}cropAndResizeFrame(e){return Qs(()=>{let t,n=pu(ho(e,"float32"),0);t=yc.cropAndResize(n,this.cropBox,this.cropBoxInd,this.cropSize,"bilinear");let r=t.shape;return jo(t,r.slice(1))})}async capture(){return(await this.next()).value}stop(){this.stream.getTracks().forEach(e=>e.stop());try{this.webcamVideoElement.srcObject=null}catch(e){console.log(e),this.webcamVideoElement.src=null}this.isClosed=!0}toArray(){throw new Error("Can not convert infinite video stream to array.")}},ZC=class{},YC=class extends yC{split(e){return new JC(this,e)}},JC=class extends YC{constructor(e,t){super(),this.upstream=e,this.impl=new QC(e,t)}summary(){return this.impl.summary()}async next(){return this.impl.next()}},QC=class extends CC{constructor(e,t){super(),this.upstream=e,this.separator=t,this.carryover=""}summary(){return`${this.upstream.summary()} -> Split('${this.separator}')`}async pump(){let e=await this.upstream.next();if(e.done)return""!==this.carryover&&(this.outputQueue.push(this.carryover),this.carryover="",!0);let t=e.value.split(this.separator);t[0]=this.carryover+t[0];for(let e of t.slice(0,-1))this.outputQueue.push(e);return this.carryover=t[t.length-1],!0}},eE=class extends yC{decodeUTF8(){return new tE(this)}},tE=class extends YC{constructor(e){super(),this.upstream=e,this.impl=new nE(e)}summary(){return this.impl.summary()}async next(){return this.impl.next()}},nE=class extends CC{constructor(e){if(super(),this.upstream=e,Le().get("IS_BROWSER"))this.decoder=new TextDecoder("utf-8");else{let{StringDecoder:e}=C();this.decoder=new e("utf8")}}summary(){return`${this.upstream.summary()} -> Utf8`}async pump(){let e,t,n=await this.upstream.next();return!n.done&&(e=n.value,t=Le().get("IS_BROWSER")?this.decoder.decode(e,{stream:!0}):this.decoder.write(Buffer.from(e.buffer)),this.outputQueue.push(t),!0)}},rE=class extends eE{constructor(e,t={}){super(),this.file=e,this.options=t,va.assert(e instanceof Uint8Array||!!Le().get("IS_BROWSER")&&(e instanceof File||e instanceof Blob),()=>"FileChunkIterator only supports File, Blob and Uint8Array right now."),this.offset=t.offset||0,this.chunkSize=t.chunkSize||1048576}summary(){return`FileChunks ${this.file}`}async next(){return this.offset>=(this.file instanceof Uint8Array?this.file.byteLength:this.file.size)?{value:null,done:!0}:{value:await new Promise((e,t)=>{let n=this.offset+this.chunkSize;if(this.file instanceof Uint8Array)e(new Uint8Array(this.file.slice(this.offset,n)));else{let r=new FileReader;r.onload=n=>{let a=r.result;if(a instanceof ArrayBuffer&&(a=new Uint8Array(a)),!(a instanceof Uint8Array))return t(new TypeError("FileReader returned unknown type."));e(a)},r.onabort=e=>t(new Error("Aborted")),r.onerror=e=>t(new Error(e.type));let a=this.file.slice(this.offset,n);r.readAsArrayBuffer(a)}this.offset=n}),done:!1}}};var aE=e=>({method:e.method,headers:e.headers,body:e.body,mode:e.mode,credentials:e.credentials,cache:e.cache,redirect:e.redirect,referrer:e.referrer,integrity:e.integrity});function sE(e){return"string"==typeof e&&"file://"===e.slice(0,7)}var iE=class extends ZC{constructor(e,t={}){super(),this.input=e,this.options=t}async iterator(){if(sE(this.input)&&Le().get("IS_NODE")){let e=E();this.input=e.readFileSync(this.input.slice(7))}return new rE(this.input,this.options)}},oE=class extends ZC{constructor(e,t={}){super(),this.url=e,this.fileOptions=t}async iterator(){return sE(this.url)?new iE(this.url,this.fileOptions).iterator():async function(e,t={},n){let r,a;"string"==typeof e?r=e:(r=e.url,a=aE(e));let s=await(n||va.fetch)(r,a);if(s.ok){let e=new Uint8Array(await s.arrayBuffer());return new rE(e,t)}throw new Error(s.statusText)}(this.url,this.fileOptions)}};function lE(e,t={}){return new qC(new oE(e),t)}function uE(e){let t=mC(e);return MC(async()=>t)}function hE(e){return MC(async()=>{let t=await e();return mC(()=>t.next())})}async function dE(e,t){return XC.create(e,t)}async function pE(e){return KC.create(e)}var cE="4.22.0";function fE(e,t){Array.isArray(e)||(e=[e]),e.forEach(e=>{null!=e&&va.assert("complex64"!==e.dtype,()=>`${t} does not support complex64 tensors in the CPU backend.`)})}var mE=Km.whereImpl,gE=class e extends P{nextDataId(){return e.nextDataId++}constructor(){super(),this.blockSize=48,this.firstUse=!0,this.data=new L(this,Zs())}write(e,t,n){this.firstUse&&(this.firstUse=!1,Le().get("IS_NODE")&&Gf.warn("\n============================\nHi, looks like you are running TensorFlow.js in Node.js. To speed things up dramatically, install our node backend, visit https://github.com/tensorflow/tfjs-node for more details. \n============================"));let r={id:this.nextDataId()};return this.data.set(r,{values:e,dtype:n,refCount:1}),r}makeTensorInfo(e,t,n){let r;if("string"===t&&null!=n&&n.length>0&&va.isString(n[0])){let a=n.map(e=>va.encodeString(e));r=this.write(a,e,t)}else r=this.write(n,e,t);return{dataId:r,shape:e,dtype:t}}refCount(e){return this.data.has(e)?this.data.get(e).refCount:0}incRef(e){this.data.get(e).refCount++}decRef(e){if(this.data.has(e)){this.data.get(e).refCount--}}move(e,t,n,r,a){this.data.set(e,{values:t,dtype:r,refCount:a})}numDataIds(){return this.data.numDataIds()}async read(e){return this.readSync(e)}readSync(e){let{dtype:t,complexTensorInfos:n}=this.data.get(e);if("complex64"===t){let e=this.readSync(n.real.dataId),t=this.readSync(n.imag.dataId);return Gf.mergeRealAndImagArrays(e,t)}return va.convertBackendValuesAndArrayBuffer(this.data.get(e).values,t)}bufferSync(e){let t=this.readSync(e.dataId);if("string"===e.dtype)try{let n=t.map(e=>va.decodeString(e));return uo(e.shape,e.dtype,n)}catch(e){throw new Error("Failed to decode encoded string bytes into utf-8")}return uo(e.shape,e.dtype,t)}makeOutput(e,t,n){return Zs().makeTensorFromTensorInfo(this.makeTensorInfo(t,n,e),this)}disposeData(e,t=!1){if(this.data.has(e)){if(this.data.get(e).refCount--,!t&&this.data.get(e).refCount>0)return!1;let{complexTensorInfos:n}=this.data.get(e);null!=n&&(this.disposeData(n.real.dataId,!0),this.disposeData(n.imag.dataId,!0)),this.data.delete(e)}return!0}disposeIntermediateTensorInfo(e){this.disposeData(e.dataId)}async time(e){let t=va.now();return e(),{kernelMs:va.now()-t}}memory(){return{unreliable:!0,reasons:["The reported memory is an upper bound. Due to automatic garbage collection, the true allocated memory may be less."]}}where(e){fE([e],"where");let t=this.readSync(e.dataId);return mE(e.shape,t)}dispose(){}floatPrecision(){return 32}epsilon(){return super.epsilon()}};gE.nextDataId=0;var yE={};function bE(e){let t=new Float32Array(e.length);for(let n=0;n<e.length;++n)t[n]=Math.abs(e[n]);return t}f(yE,{addImpl:()=>FE,bincountImpl:()=>LE,bincountReduceImpl:()=>PE,bitwiseAndImpl:()=>zE,castImpl:()=>CE,ceilImpl:()=>HE,concatImpl:()=>KE,equalImpl:()=>XE,expImpl:()=>JE,expm1Impl:()=>tA,floorDivImpl:()=>oA,floorImpl:()=>aA,gatherNdImpl:()=>hA,gatherV2Impl:()=>dA,greaterEqualImpl:()=>mA,greaterImpl:()=>pA,lessEqualImpl:()=>wA,lessImpl:()=>bA,linSpaceImpl:()=>SA,logImpl:()=>_A,maxImpl:()=>CA,maximumImpl:()=>EA,minimumImpl:()=>RA,multiplyImpl:()=>MA,negImpl:()=>zA,notEqualImpl:()=>WA,prodImpl:()=>qA,raggedGatherImpl:()=>JA,raggedRangeImpl:()=>e$,raggedTensorToTensorImpl:()=>s$,rangeImpl:()=>i$,rsqrtImpl:()=>o$,scatterImpl:()=>h$,sigmoidImpl:()=>d$,simpleAbsImpl:()=>bE,sliceImpl:()=>f$,sparseFillEmptyRowsImpl:()=>y$,sparseReshapeImpl:()=>b$,sparseSegmentReductionImpl:()=>x$,sqrtImpl:()=>v$,squaredDifferenceImpl:()=>I$,staticRegexReplaceImpl:()=>N$,stridedSliceImpl:()=>E$,stringNGramsImpl:()=>$$,stringSplitImpl:()=>F$,stringToHashBucketFastImpl:()=>D$,subImpl:()=>M$,tileImpl:()=>z$,topKImpl:()=>V$,transposeImpl:()=>GA,uniqueImpl:()=>U$});var xE={kernelName:Ve,backendName:"cpu",kernelFunc:e=>{let{x:t}=e.inputs,n=e.backend;fE(t,"abs");let r=new Float32Array(va.sizeFromShape(t.shape));return r=bE(n.data.get(t.dataId).values),n.makeOutput(r,t.shape,t.dtype)}};function vE(e){return(t,n,r,a,s)=>{let i=Gf.assertAndGetBroadcastShape(t,n),o=i.length,l=va.computeStrides(i),u=va.sizeFromShape(i),h=va.getTypedArrayFromDType(s,u),d=t.length,p=n.length,c=va.computeStrides(t),f=va.computeStrides(n),m=Gf.getBroadcastDims(t,i),g=Gf.getBroadcastDims(n,i);if(m.length+g.length===0)for(let t=0;t<h.length;++t)h[t]=e(r[t%r.length],a[t%a.length]);else for(let t=0;t<h.length;++t){let n=va.indexToLoc(t,o,l),s=n.slice(-d);m.forEach(e=>s[e]=0);let i=va.locToIndex(s,d,c),u=n.slice(-p);g.forEach(e=>u[e]=0);let y=va.locToIndex(u,p,f);h[t]=e(r[i],a[y])}return[h,i]}}function wE(e){let{inputs:t,backend:n}=e,{real:r,imag:a}=t,s=n.data.get(r.dataId).values,i=n.data.get(a.dataId).values,o=n.makeTensorInfo(r.shape,"complex64");return n.data.get(o.dataId).complexTensorInfos={real:n.makeTensorInfo(r.shape,"float32",s),imag:n.makeTensorInfo(a.shape,"float32",i)},o}var kE={kernelName:mt,backendName:"cpu",kernelFunc:wE};function IE(e,t,n="float32"){if("complex64"===n){return wE({inputs:{real:IE(e,t,"float32"),imag:IE(e,t,"float32")},backend:e})}let r=va.makeZerosTypedArray(va.sizeFromShape(t),n);return e.makeTensorInfo(t,n,r)}function SE(e){let{inputs:t,backend:n}=e,{x:r}=t;return n.incRef(r.dataId),{dataId:r.dataId,shape:r.shape,dtype:r.dtype}}var _E={kernelName:an,backendName:"cpu",kernelFunc:SE};function NE(e){let{inputs:t,backend:n}=e,{input:r}=t,a=n.data.get(r.dataId).complexTensorInfos.real,s=n.data.get(a.dataId).values;return n.makeTensorInfo(a.shape,a.dtype,s)}var TE={kernelName:rr,backendName:"cpu",kernelFunc:NE};function CE(e,t,n,r){if("int32"===r){return[t,"int32",Int32Array.from(e)]}if("bool"===r){let r=va.toTypedArray([0],n),[a,s]=vE((e,t)=>e!==t?1:0)(t,[],e,r,"bool");return[s,"bool",a]}throw new Error(`Error in Cast: failed to cast ${n} to ${r}`)}function EE(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{dtype:s}=r;if("complex64"===s){if("complex64"===a.dtype)return SE({inputs:{x:a},backend:n});let e=IE(n,a.shape,a.dtype),t=EE({inputs:{x:a},backend:n,attrs:{dtype:"float32"}}),r=wE({inputs:{real:t,imag:e},backend:n});return n.disposeIntermediateTensorInfo(e),n.disposeIntermediateTensorInfo(t),r}if("complex64"===a.dtype){let e=NE({inputs:{input:a},backend:n}),t=EE({inputs:{x:e},backend:n,attrs:{dtype:s}});return n.disposeIntermediateTensorInfo(e),t}if(!va.hasEncodingLoss(a.dtype,s)){let e=SE({inputs:{x:a},backend:n});return{dataId:e.dataId,shape:e.shape,dtype:s}}let i=n.data.get(a.dataId).values,[o,l,u]=CE(i,a.shape,a.dtype,s);return n.makeTensorInfo(o,l,u)}var AE={kernelName:pt,backendName:"cpu",kernelFunc:EE};function $E(e,t,n,r){return null==n?({inputs:n,backend:a})=>{let{a:s,b:i}=n,o=a;fE([s,i],e);let l=o.data.get(s.dataId).values,u=o.data.get(i.dataId).values,h="string"===s.dtype?Gf.fromUint8ToStringArray(l):l,d="string"===s.dtype?Gf.fromUint8ToStringArray(u):u,p=r||s.dtype,[c,f]=t(s.shape,i.shape,h,d,p);return o.makeTensorInfo(f,p,c)}:({inputs:e,backend:a})=>{let{a:s,b:i}=e,o=a;if("complex64"===s.dtype||"complex64"===i.dtype){let e=EE({inputs:{x:s},backend:o,attrs:{dtype:"complex64"}}),t=o.data.get(e.dataId),r=t.complexTensorInfos.real,a=t.complexTensorInfos.imag,l=o.data.get(r.dataId).values,u=o.data.get(a.dataId).values,h=EE({inputs:{x:i},backend:o,attrs:{dtype:"complex64"}}),d=o.data.get(h.dataId),p=d.complexTensorInfos.real,c=d.complexTensorInfos.imag,f=o.data.get(p.dataId).values,m=o.data.get(c.dataId).values,[g,y,b]=n(s.shape,i.shape,l,u,f,m),x=o.makeTensorInfo(b,"float32",g),v=o.makeTensorInfo(b,"float32",y),w=wE({inputs:{real:x,imag:v},backend:o});return o.disposeIntermediateTensorInfo(e),o.disposeIntermediateTensorInfo(h),o.disposeIntermediateTensorInfo(x),o.disposeIntermediateTensorInfo(v),w}{let e=o.data.get(s.dataId).values,n=o.data.get(i.dataId).values,a=r||s.dtype,[l,u]=t(s.shape,i.shape,e,n,a);return o.makeTensorInfo(u,a,l)}}}function RE(e){return(t,n,r,a,s,i)=>{let o=Gf.assertAndGetBroadcastShape(t,n),l=va.sizeFromShape(o),u=o.length,h=va.computeStrides(o),d=va.getTypedArrayFromDType("float32",l),p=va.getTypedArrayFromDType("float32",l),c=Gf.getBroadcastDims(t,o),f=Gf.getBroadcastDims(n,o),m=Gf.mergeRealAndImagArrays(r,a),g=Gf.mergeRealAndImagArrays(s,i),y=t.length,b=va.computeStrides(t),x=n.length,v=va.computeStrides(n);if(c.length+f.length===0)for(let t=0;t<d.length;t++){let n=t%m.length,r=t%g.length,a=e(m[2*n],m[2*n+1],g[2*r],g[2*r+1]);d[t]=a.real,p[t]=a.imag}else for(let t=0;t<d.length;t++){let n=va.indexToLoc(t,u,h),r=n.slice(-y);c.forEach(e=>r[e]=0);let a=va.locToIndex(r,y,b),s=n.slice(-x);f.forEach(e=>s[e]=0);let i=va.locToIndex(s,x,v),o=e(m[2*a],m[2*a+1],g[2*i],g[2*i+1]);d[t]=o.real,p[t]=o.imag}return[d,p,o]}}var FE=vE((e,t)=>e+t),DE=RE((e,t,n,r)=>({real:e+n,imag:t+r})),ME=$E(He,FE,DE),OE={kernelName:He,backendName:"cpu",kernelFunc:ME};function LE(e,t,n,r,a){let s=va.sizeFromShape(r),i=va.makeZerosTypedArray(a,n);for(let n=0;n<e.length;n++){let r=e[n];if(r<0)throw new Error("Input x must be non-negative!");r>=a||(i[r]+=s>0?t[n]:1)}return i}function PE(e,t,n,r=!1){let a=e.shape[0],s=e.shape[1],i=uo([a,n],t.dtype);for(let o=0;o<a;o++)for(let a=0;a<s;a++){let s=e.get(o,a);if(s<0)throw new Error("Input x must be non-negative!");s>=n||(r?i.set(1,o,s):t.size>0?i.set(i.get(o,s)+t.get(o,a),o,s):i.set(i.get(o,s)+1,o,s))}return i}var zE=vE((e,t)=>e&t),BE=$E(ut,zE),WE={kernelName:ut,backendName:"cpu",kernelFunc:BE};function VE(e){return(t,n,r)=>{let a=va.getArrayFromDType(n,t.length);for(let n=0;n<t.length;++n)a[n]=e(t[n],r);return a}}function UE(e,t,n){return GE(e,VE(t),n)}function GE(e,t,n){return({inputs:r,attrs:a,backend:s})=>{let{x:i}=r;fE(i,e);let o,l=s,u=l.data.get(i.dataId).values;if("string"===i.dtype){if(!Array.isArray(u))throw new Error("String tensor's value was not an instance of Array");o=Gf.fromUint8ToStringArray(u)}else o=u;let h=n||i.dtype,d=t(o,h,a);return l.makeTensorInfo(i.shape,h,d)}}var HE=VE(e=>Math.ceil(e)),jE=GE(ct,HE),qE={kernelName:ct,backendName:"cpu",kernelFunc:jE};function KE(e,t,n,r){let a=va.getArrayFromDType(n,va.sizeFromShape(t));if(r&&"string"!==n){let t=0;e.forEach(e=>{let n=va.sizeFromShape(e.shape);a.set(e.vals,t),t+=n})}else{let r=0;e.forEach(e=>{let s="string"===n?Gf.fromUint8ToStringArray(e.vals):e.vals,i=0;for(let n=0;n<e.shape[0];++n){let o=n*t[1]+r;for(let t=0;t<e.shape[1];++t)a[o+t]=s[i++]}r+=e.shape[1]})}return a}var XE=vE((e,t)=>e===t?1:0),ZE=$E(Gt,XE,null,"bool"),YE={kernelName:Gt,backendName:"cpu",kernelFunc:ZE},JE=VE(e=>Math.exp(e)),QE=GE(Ht,JE,"float32"),eA={kernelName:Ht,backendName:"cpu",kernelFunc:QE},tA=VE(e=>Math.expm1(e)),nA=GE(qt,tA),rA={kernelName:qt,backendName:"cpu",kernelFunc:nA},aA=VE(e=>Math.floor(e)),sA=GE(Yt,aA),iA={kernelName:Yt,backendName:"cpu",kernelFunc:sA},oA=vE((e,t)=>Math.floor(e/t)),lA=$E(Jt,oA,null,"int32"),uA={kernelName:Jt,backendName:"cpu",kernelFunc:lA};function hA(e,t,n,r,a,s,i,o,l){let u=uo([r,s],n);for(let n=0;n<r;n++){let r=[],h=0;for(let t=0;t<a;t++){let s=e[n*a+t];h+=s*i[t],r.push(s)}if(h<0||h>=l/s)throw new Error(`Invalid indices: ${r} does not index into ${o}`);for(let e=0;e<s;e++)u.values[n*s+e]=t.get(...t.indexToLoc(h*s+e))}return u}function dA(e,t,n){let r=uo(n,e.dtype);for(let n=0;n<r.size;++n){let a=r.indexToLoc(n).slice(),s=a[0],i=a[2],o=t.locToIndex([s,i]);a[2]=t.values[o];let l=e.locToIndex(a);0<=l&&l<e.values.length&&(r.values[n]=e.values[l])}return r}var pA=vE((e,t)=>e>t?1:0),cA=$E(nn,pA,null,"bool"),fA={kernelName:nn,backendName:"cpu",kernelFunc:cA},mA=vE((e,t)=>e>=t?1:0),gA=$E(rn,mA,null,"bool"),yA={kernelName:rn,backendName:"cpu",kernelFunc:gA},bA=vE((e,t)=>e<t?1:0),xA=$E(pn,bA,null,"bool"),vA={kernelName:pn,backendName:"cpu",kernelFunc:xA},wA=vE((e,t)=>e<=t?1:0),kA=$E(cn,wA,null,"bool"),IA={kernelName:cn,backendName:"cpu",kernelFunc:kA};function SA(e,t,n){let r=(t-e)/(n-1),a=va.makeZerosTypedArray(n,"float32");a[0]=e;for(let e=1;e<a.length;e++)a[e]=a[e-1]+r;return a}var _A=VE(e=>Math.log(e)),NA=GE(mn,_A),TA={kernelName:mn,backendName:"cpu",kernelFunc:NA};function CA(e,t,n,r){let a=va.getTypedArrayFromDType(r,va.sizeFromShape(n));for(let n=0;n<a.length;++n){let r=n*t,s=e[r];for(let n=0;n<t;++n){let t=e[r+n];(Number.isNaN(t)||t>s)&&(s=t)}a[n]=s}return a}var EA=vE((e,t)=>Math.max(e,t)),AA=$E(Tn,EA),$A={kernelName:Tn,backendName:"cpu",kernelFunc:AA},RA=vE((e,t)=>Math.min(e,t)),FA=$E(Mn,RA),DA={kernelName:Mn,backendName:"cpu",kernelFunc:FA},MA=vE((e,t)=>e*t),OA=RE((e,t,n,r)=>({real:e*n-t*r,imag:e*r+t*n})),LA=$E(zn,MA,OA),PA={kernelName:zn,backendName:"cpu",kernelFunc:LA};function zA(e,t,n){let r=va.createScalarValue(-1,n);return MA([],t,r,e,n)}var BA={kernelName:Bn,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n}=e,{x:r}=t;fE(r,"neg");let a=n.data.get(r.dataId).values,[s,i]=zA(a,r.shape,r.dtype);return n.makeTensorInfo(i,r.dtype,s)}},WA=vE((e,t)=>e!==t?1:0),VA=$E(Wn,WA,null,"bool"),UA={kernelName:Wn,backendName:"cpu",kernelFunc:VA};function GA(e,t,n,r,a){let s=t.length,i=va.sizeFromShape(t),o=va.computeStrides(t),l=va.computeStrides(a),u=va.getTypedArrayFromDType(n,va.sizeFromShape(a));for(let t=0;t<i;++t){let n=va.indexToLoc(t,s,o),a=new Array(n.length);for(let e=0;e<a.length;e++)a[e]=n[r[e]];u[va.locToIndex(a,s,l)]=e[t]}return u}function HA(e){let{inputs:t,attrs:n,backend:r}=e,{x:a}=t,{perm:s}=n;fE(a,"transpose");let i=a.shape.length,o=new Array(i);for(let e=0;e<o.length;e++)o[e]=a.shape[s[e]];let l=GA(r.data.get(a.dataId).values,a.shape,a.dtype,s,o);return{dataId:r.write(l,o,a.dtype),shape:o,dtype:a.dtype}}var jA={kernelName:Xr,backendName:"cpu",kernelFunc:HA};function qA(e,t,n,r){let[a,s]=Gf.computeOutAndReduceShapes(e,r),i=ps(t,"int32"),o=va.makeZerosTypedArray(va.sizeFromShape(a),i),l=va.sizeFromShape(s);for(let e=0;e<o.length;++e){let t=e*l,r=1;for(let e=0;e<l;++e)r*=n[t+e];o[e]=r}return{outVals:o,outShape:a,outDtype:i}}var KA={kernelName:Jn,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{axis:s,keepDims:i}=r;fE(a,"prod");let o=a.shape.length,l=va.parseAxisParam(s,a.shape),u=Gf.getAxesPermutation(l,o),h=l,d=a,p=[];null!=u&&(d=HA({inputs:{x:a},backend:n,attrs:{perm:u}}),p.push(d),h=Gf.getInnerMostAxes(h.length,o));let c=n.data.get(d.dataId).values,{outVals:f,outShape:m,outDtype:g}=qA(d.shape,d.dtype,c,h),y=m;return i&&(y=Gf.expandShapeToKeepDim(m,l)),p.forEach(e=>n.disposeIntermediateTensorInfo(e)),n.makeTensorInfo(y,g,f)}};function XA(e,t,n,r){let a=[],s=0,i=t.length-1+n.length,o=new Array(i).fill(null).map(()=>[0]);!function(e,t){for(let n=0;n<e.length;++n){let r=e[n],a=n===e.length-1?t:e[n+1].length;if(0===r.length)throw new Error("Ragged splits may not be empty");if(r[0]<0)throw new Error("Ragged splits must be non-negative");if(r[r.length-1]>a)throw new Error("Ragged splits must not point past values");for(let e=1;e<r.length;++e)if(r[e-1]>r[e])throw new Error("Ragged splits must be sorted in ascending order")}}(n,r);let l=1;for(let e=0;e<t.length-1;++e){l*=t[e];let n=t[e+1];for(let t=1;t<l+1;++t)o[e].push(t*n)}for(let r=0;r<e.length;++r){let i=e[r],l=e[r]+1;for(let e=0;e<n.length;++e){let r=n[e],a=e+t.length-1;if(a>=0){let e=o[a],t=e[e.length-1]-r[i];for(let e=i;e<l;++e)o[a].push(r[e+1]+t)}i=r[i],l=r[l]}l!==i&&(a.push([i,l]),s+=l-i)}return{outSplits:o,valueSlices:a,numValues:s}}function ZA(e,t){let n=e.slice(0,t);for(;n.length<t;)n.push(1);for(let r=t;r<e.length;r++)n[t-1]*=e[r];return n}function YA(e,t,n,r,a){let s=t.slice();s[0]=a;let i=va.getArrayFromDType(n,va.sizeFromShape(s)),o=e.length;return function(e,t,n,r,a,s){let i=ZA(t,2)[1],o=ZA(s,2)[1],l=0;for(let t of n)for(let n=t[0];n<t[1];++n){for(let t=0;t<r;++t)a[l*o+t]=e[n*i+t];++l}}(e,t,r,0===o?0:o/t[0],i,s),[i,s]}function JA(e,t,n,r,a,s,i,o){if(0===e.length)throw new Error("paramsNestedSplits must be non empty");if(0===t[0].length)throw new Error("Split tensors must not be scalars");if(function(e,t,n){e.forEach((e,r)=>{if(e<0||e>=n){let a=va.indexToLoc(r,t.length,va.computeStrides(t)).join(",");throw new Error(`indices[${a}] = ${e} is not in [0, ${n})`)}})}(s,i,t[0][0]-1),0===r.length)throw new Error("params.rank must be nonzero");let l=r[0],{outSplits:u,valueSlices:h,numValues:d}=XA(s,i,e,l),p=function(e){let t=[];for(let n=0;n<e.length;++n){let r=e[n].length,a=va.getArrayFromDType("int32",r);t.push(a),e[n].forEach((e,t)=>a[t]=e)}return t}(u),c=YA(n,r,a,h,d);return[p,c[0],c[1]]}var QA=2147483647;function e$(e,t,n,r,a,s,i){if(t.length>1)throw new Error("starts must be a scalar or vector");if(a.length>1)throw new Error("limits must be a scalar or vector");if(i.length>1)throw new Error("deltas must be a scalar or vector");let o=0===t.length,l=0===a.length,u=0===i.length,h=[];o||h.push(t[0]),l||h.push(a[0]),u||h.push(i[0]);for(let e=1;e<h.length;++e)if(h[e]!==h[e-1])throw new Error("starts, limits, and deltas must have the same shape");let d=0===h.length?1:h[0],p=va.getArrayFromDType("int32",d+1);p[0]=0;for(let t=0;t<d;++t){let n,a=o?e[0]:e[t],i=l?r[0]:r[t],h=u?s[0]:s[t];if(0===h)throw new Error("Requires delta != 0");if(h>0&&i<a||h<0&&i>a)n=0;else if(n=Math.ceil(Math.abs((i-a)/h)),n>QA)throw new Error(`Requires ((limit - start) / delta) <= ${QA}`);p[t+1]=p[t]+n}let c=p[d],f=va.getArrayFromDType(n,c),m=0;for(let t=0;t<d;++t){let n=p[t+1]-p[t],r=o?e[0]:e[t],a=u?s[0]:s[t];for(let e=0;e<n;++e)f[m++]=r,r+=a}return[p,f]}var t$=Gf.RowPartitionType,n$=class e{constructor(e,t,n,r,a,s,i,o,l,u){this.shape=e,this.shapeShape=t,this.values=n,this.valuesShape=r,this.valuesDType=a,this.defaultValue=s,this.defaultValueShape=i,this.rowPartitionValues=o,this.rowPartitionValuesShapes=l,this.rowPartitionTypes=Gf.getRowPartitionTypesHelper(u),this.raggedRank=Gf.getRaggedRank(this.rowPartitionTypes)}getRowPartitionTypeByDimension(e){return this.rowPartitionTypes[0]===t$.FIRST_DIM_SIZE?this.rowPartitionTypes[e+1]:this.rowPartitionTypes[e]}getRowPartitionTensor(e){return this.rowPartitionTypes[0]===t$.FIRST_DIM_SIZE?this.rowPartitionValues[e+1]:this.rowPartitionValues[e]}getMaxWidth(t){let n=this.getRowPartitionTensor(t-1);switch(this.getRowPartitionTypeByDimension(t-1)){case t$.VALUE_ROWIDS:return e.getMaxWidthValueRowID(n);case t$.ROW_SPLITS:return e.getMaxWidthRowSplit(n);default:throw new Error(`Cannot handle partition type ${t$[this.getRowPartitionTypeByDimension(t-1)]}`)}}static getMaxWidthRowSplit(e){let t=e.length;if(0===t||1===t)return 0;let n=0;for(let r=0;r<t-1;++r){let t=e[r+1]-e[r];t>n&&(n=t)}return n}static getMaxWidthValueRowID(e){let t=e.length;if(0===t)return 0;let n=0,r=e[0],a=0;for(let s=1;s<t;++s){let t=e[s];t!==r&&(r=t,a=Math.max(s-n,a),n=s)}return Math.max(t-n,a)}tensorShapeFromTensor(e,t,n=!0){if(0===t.length){if(-1===e[0])return[];throw new Error("The only valid scalar shape tensor is the fully unknown shape specified as -1.")}return a$(e,n)}calculateOutputSize(e){let t=this.valuesShape,n=this.defaultValueShape;Gf.validateDefaultValueShape(n,t);let r=this.tensorShapeFromTensor(this.shape,this.shapeShape),a=Gf.combineRaggedTensorToTensorShapes(this.raggedRank,r,t);a[0]<0&&(a[0]=e);for(let e=1;e<=this.raggedRank;++e)a[e]<0&&(a[e]=this.getMaxWidth(e));return a}calculateFirstParentOutputIndex(e,t,n){let r=Math.min(e,n),a=[],s=0;for(let e=0;e<r;++e,s+=t)a.push(s);for(let t=r;t<e;++t)a.push(-1);return va.assert(a.length===e,()=>"Final length of result must be equal to firstDimension."),a}calculateOutputIndexRowSplit(e,t,n,r){let a=e.length,s=[];for(let i=0;i<a-1;++i){let a=e[i+1]-e[i],o=Math.min(r,a),l=t[i];-1===l&&(o=0);for(let e=0;e<o;++e)s.push(l),l+=n;for(let e=0;e<a-o;++e)s.push(-1)}if(a>0&&s.length!==e[a-1])throw new Error("Invalid row split size.");return s}calculateOutputIndexValueRowID(e,t,n,r){let a=e.length,s=[];if(0===a)return[];let i=0,o=e[0];if(o>=t.length)throw new Error(`Got currentValueRowId=${o}, which is not less than ${t.length}`);let l=t[o];s.push(l);for(let u=1;u<a;++u){let a=e[u];if(a===o)l>=0&&(++i,i<r?l+=n:l=-1);else{if(i=0,o=a,a>=t.length)throw new Error(`Got nextValueRowId=${a} which is not less than ${t.length}`);l=t[a]}s.push(l)}if(s.length!==e.length)throw new Error("Invalid row ids.");return s}calculateOutputIndex(e,t,n,r){let a=this.getRowPartitionTensor(e),s=this.getRowPartitionTypeByDimension(e);switch(s){case t$.VALUE_ROWIDS:return this.calculateOutputIndexValueRowID(a,t,n,r);case t$.ROW_SPLITS:if(a.length-1>t.length)throw new Error(`Row partition size is greater than output size: ${a.length-1} > ${t.length}`);return this.calculateOutputIndexRowSplit(a,t,n,r);default:throw new Error(`Unsupported partition type: ${t$[s]}`)}}getFirstDimensionSize(){let e=this.rowPartitionValues[0];if(0===this.rowPartitionTypes.length)throw new Error("No row_partition_types given.");let t=this.rowPartitionTypes[0];switch(t){case t$.FIRST_DIM_SIZE:return e[0];case t$.VALUE_ROWIDS:throw new Error("Cannot handle VALUE_ROWIDS in first dimension.");case t$.ROW_SPLITS:return this.rowPartitionValuesShapes[0][0]-1;default:throw new Error(`Cannot handle type ${t$[t]}`)}}compute(){if(this.rowPartitionValues[0].length<=0)throw new Error("Invalid first partition input. Tensor requires at least one element.");let e=this.getFirstDimensionSize(),t=this.calculateOutputSize(e),n=new Array(this.raggedRank+1);n[n.length-1]=1;for(let e=n.length-2;e>=0;--e)n[e]=n[e+1]*t[e+1];let r=a$(t,!1),a=va.getArrayFromDType(this.valuesDType,va.sizeFromShape(r));if(n[0]*t[0]>0){let s=this.calculateFirstParentOutputIndex(e,n[0],t[0]);for(let e=1;e<=this.raggedRank;++e)s=this.calculateOutputIndex(e-1,s,n[e],t[e]);this.setOutput(this.raggedRank,s,a,r)}return[r,a]}setOutput(e,t,n,r){if(0===n.length)return;let a=this.values,s=n,i=r.slice();i=i.slice(e+1);let o=va.sizeFromShape(i),l=t.length,u=this.defaultValue;if(u.length!==o&&1!==u.length){let e=this.defaultValueShape;Qs(()=>{let t=jo(u,e);u=hl(t,i).dataSync()})}let h=0,d=0,p=0;for(let e=0;e<=l;++e){let r=e<l?t[e]:-1;if(r!==p){if(d<p){let e=a.subarray(h*o);r$(s.subarray(d*o),e,(p-d)*o)}if(e>=l){let e=n.length;r=Math.floor(e/o)}if(r>p)if(1===this.defaultValue.length)s.subarray(p*o,r*o).fill(this.defaultValue[0]),p=r;else for(;r>p;){r$(s.slice(p*o),u,o),++p}r<0?(h=e+1,d=p):(h=e,d=p,p=d+1)}else++p}}};function r$(e,t,n){for(let r=0;r<n;r++)e[r]=t[r]}function a$(e,t){let n=[];for(let r of e){if(r<0){if(!t)throw new Error(`Dimension ${r} must be >= 0`);if(r<-1)throw new Error(`Dimension ${r} must be >= -1`);r=-1}n.push(r)}return n}function s$(e,t,n,r,a,s,i,o,l,u){return new n$(e,t,n,r,a,s,i,o,l,u).compute()}function i$(e,t,n,r){if(e===t||e<t&&n<0||t<e&&n>1)return va.makeZerosTypedArray(0,r);let a=Math.abs(Math.ceil((t-e)/n)),s=va.makeZerosTypedArray(a,r);t<e&&1===n&&(n=-1),s[0]=e;for(let e=1;e<s.length;e++)s[e]=s[e-1]+n;return s}var o$=VE(e=>1/Math.sqrt(e)),l$=GE(fr,o$),u$={kernelName:fr,backendName:"cpu",kernelFunc:l$};function h$(e,t,n,r,a,s,i,o,l,u){let h=[r/a,a],d=e.values,p=t.values;if(0===r)return uo(n,t.dtype);let c=l instanceof Ja?l:uo(h,t.dtype);"string"==typeof l||"number"==typeof l?c.values.fill(l):"boolean"==typeof l&&c.values.fill(+l);for(let e=0;e<s;e++){let s=[],l=0;for(let t=0;t<i;t++){let n=d[e*i+t];s.push(n),l+=n*o[t]}if(l<0||l>=r/a)throw new Error(`Invalid indices: ${s} does not index into ${n}`);for(let n=0;n<a;n++)u?c.values[l*a+n]+=p[e*a+n]:c.values[l*a+n]=0===t.rank?p[0]:p[e*a+n]}return c}var d$=VE(e=>1/(1+Math.exp(-e))),p$=UE(Sr,e=>1/(1+Math.exp(-e))),c$={kernelName:Sr,backendName:"cpu",kernelFunc:p$};function f$(e,t,n,r,a){let s=vf.isSliceContinous(r,t,n),i=va.sizeFromShape(n),o=va.computeStrides(r);if(s){let n=vf.computeFlatOffset(t,o);return"string"===a?e.slice(n,n+i):e.subarray(n,n+i)}let l=uo(r,a,"string"===a?Gf.fromUint8ToStringArray(e):e),u=uo(n,a);for(let e=0;e<u.size;++e){let n=u.indexToLoc(e),r=n.map((e,n)=>e+t[n]);u.set(l.get(...r),...n)}return"string"===a?Gf.fromStringArrayToUint8(u.values):u.values}function m$(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{begin:s,size:i}=r;fE(a,"slice");let[o,l]=vf.parseSliceParams(a,s,i);vf.assertParamsValid(a,o,l);let u=f$(n.data.get(a.dataId).values,o,l,a.shape,a.dtype);return n.makeTensorInfo(l,a.dtype,u)}var g$={kernelName:vr,backendName:"cpu",kernelFunc:m$};function y$(e,t,n,r,a,s,i){let o=t[0],l=s[0],u=new Array(l),h=new Array(o),d=t[1];if(0===l){if(0!==o)throw new Error(Gf.getSparseFillEmptyRowsIndicesDenseShapeMismatch(o));return[va.getArrayFromDType(n,0),[0,d],va.getArrayFromDType(a,0),u,h]}let p=!0,c=0,f=new Array(l).fill(0);for(let t=0;t<o;++t){let n=e[t*d];if(n<0)throw new Error(Gf.getSparseFillEmptyRowsNegativeIndexErrorMessage(t,n));if(n>=l)throw new Error(Gf.getSparseFillEmptyRowsOutOfRangeIndexErrorMessage(t,n,l));++f[n],p=p&&n>=c,c=n}let m=!0;for(let e=0;e<l;++e){let t=0===f[e];u[e]=t,m=m&&!t,f[e]=Math.max(f[e],1),e>0&&(f[e]+=f[e-1])}if(m&&p){let t=e,n=r;for(let e=0;e<o;++e)h[e]=e;return[t,[o,d],n,u,h]}{let t=f[l-1],s=va.getArrayFromDType(n,t*d),p=va.getArrayFromDType(a,t),c=new Array(l).fill(0);for(let t=0;t<o;++t){let n=e[t*d],a=c[n],i=(0===n?0:f[n-1])+a;c[n]++;for(let n=0;n<d;++n)s[i*d+n]=e[t*d+n];p[i]=r[t],h[t]=i}for(let e=0;e<l;++e)if(0===c[e]){let t=0===e?0:f[e-1];s[t*d+0]=e;for(let e=1;e<d;++e)s[t*d+e]=0;p[t]=i}return[s,[t,d],p,u,h]}}function b$(e,t,n,r,a){let s=va.sizeFromShape(r),i=t[0],o=a.length,l=[],u=1,h=-1;for(let e=0;e<o;++e){let t=a[e];if(-1===t){if(-1!==h)throw new Error(Gf.getSparseReshapeMultipleNegativeOneOutputDimErrorMessage(h,e));h=e,l.push(1)}else{if(t<0)throw new Error(Gf.getSparseReshapeNegativeOutputDimErrorMessage(e,t));u*=t,l.push(t)}}if(-1!==h){if(u<=0)throw new Error(Gf.getSparseReshapeEmptyTensorZeroOutputDimErrorMessage());let e=Math.trunc(s/u);if(u*e!==s)throw new Error(Gf.getSparseReshapeInputOutputMultipleErrorMessage(r,l));l[h]=e}if(va.sizeFromShape(l)!==s)throw new Error(Gf.getSparseReshapeInputOutputMismatchErrorMessage(r,l));let d=r.length,p=[];if(d>0){p[d-1]=1;for(let e=d-2;e>=0;--e)p[e]=p[e+1]*r[e+1]}let c=[];if(o>0){c[o-1]=1;for(let e=o-2;e>=0;--e)c[e]=c[e+1]*l[e+1]}let f=va.getArrayFromDType(n,i*o);for(let t=0;t<i;++t){let n=0;for(let r=0;r<d;++r)n+=e[t*d+r]*p[r];for(let e=0;e<o;++e)f[t*o+e]=Math.trunc(n/c[e]),n%=c[e]}return[f,[i,o],l]}function x$(e,t,n,r,a,s=!1,i=0){let o=r.length,l=[t[0],e.length/t[0]],u=l[1],h=o>0?a[o-1]+1:0;if(h<0)throw new Error(Gf.getSparseSegmentReductionNegativeSegmentIdsErrorMessage());let d=t.slice();d[0]=h;let p=d.reduce((e,t)=>e*t,1),c=va.getArrayFromDType(n,p);if(0===o)return h>0&&c.fill(i),[c,d];if(h<=0)throw new Error(Gf.getSparseSegmentReductionNegativeSegmentIdsErrorMessage());let f=0,m=1,g=0,y=a[f];for(;;){let t=0;if(m<o){if(t=a[m],y===t){++m;continue}if(y>=t)throw new Error(Gf.getSparseSegmentReductionNonIncreasingSegmentIdsErrorMessage())}if(y<0||y>=h)throw new Error(Gf.getSparseSegmentReductionSegmentIdOutOfRangeErrorMessage(y,h));y>g&&c.fill(i,g*u,y*u);for(let t=f;t<m;++t){let n=r[t];if(n<0||n>=l[0])throw new Error(Gf.getSparseSegmentReductionIndicesOutOfRangeErrorMessage(t,r[t],l[0]));for(let t=0;t<u;t++)c[y*u+t]+=e[n*u+t]}if(s)for(let e=0;e<u;e++)c[y*u+e]/=m-f;if(f=m,++m,g=y+1,y=t,m>o)break}return g<h&&c.fill(i,g*u,h*u),[c,d]}var v$=VE(e=>Math.sqrt(e)),w$=UE(Nr,e=>Math.sqrt(e)),k$={kernelName:Nr,backendName:"cpu",kernelFunc:w$},I$=vE((e,t)=>{let n=e-t;return n*n}),S$=$E(Or,I$),_$={kernelName:Or,backendName:"cpu",kernelFunc:S$},N$=VE((e,t)=>{let{pattern:n,replaceGlobal:r,rewrite:a}=t;return e.replace(new RegExp(n,r?"g":""),a)}),T$=GE(Pr,N$),C$={kernelName:Pr,backendName:"cpu",kernelFunc:T$};function E$(e,t,n,r){let a=uo(e,t.dtype);for(let e=0;e<a.size;e++){let s=a.indexToLoc(e),i=new Array(s.length);for(let e=0;e<i.length;e++)i[e]=s[e]*n[e]+r[e];a.set(t.get(...i),...s)}return a}var A$=class{constructor(e,t,n,r,a,s){this.separator=va.encodeString(e),this.nGramWidths=t,this.leftPad=va.encodeString(n),this.rightPad=va.encodeString(r),this.padWidth=a,this.preserveShort=s}getPadWidth(e){return Math.min(this.padWidth<0?e-1:this.padWidth,e-1)}getNumNGrams(e,t){let n=this.getPadWidth(t);return Math.max(0,e+2*n-t+1)}createNGrams(e,t,n,r,a,s){for(let i=0;i<a;++i){let o=this.getPadWidth(s),l=Math.max(0,o-i),u=Math.max(0,o-(a-(i+1))),h=s-(l+u),d=t+(l>0?0:i-o),p=0;p+=l*this.leftPad.length;for(let t=0;t<h;++t)p+=e[d+t].length;p+=u*this.rightPad.length,p+=(l+u+h-1)*this.separator.length,n[r+i]=new Uint8Array(p);let c=n[r+i],f=0,m=e=>e.forEach(e=>c[f++]=e);for(let e=0;e<l;++e)m(this.leftPad),m(this.separator);for(let t=0;t<h-1;++t)m(e[d+t]),m(this.separator);if(h>0){m(e[d+h-1]);for(let e=0;e<u;++e)m(this.separator),m(this.rightPad)}else{for(let e=0;e<u-1;++e)m(this.rightPad),m(this.separator);m(this.rightPad)}}}compute(e,t){let n=e.length,r=t.length;if(r>0){let e=t[0];if(0!==e)throw new Error(`First split value must be 0, got ${e}`);for(let a=1;a<r;++a){let r=t[a]>=e;if(r=r&&t[a]<=n,!r)throw new Error(`Invalid split value ${t[a]}, must be in [${e}, ${n}]`);e=t[a]}if(e!==n)throw new Error(`Last split value must be data size. Expected ${n}, got ${e}`)}let a=r-1,s=va.getArrayFromDType("int32",r);if(0===n||0===r){let e=new Array(n);for(let e=0;e<=a;++e)s[e]=0;return[e,s]}s[0]=0;for(let e=1;e<=a;++e){let n=t[e]-t[e-1],r=0;this.nGramWidths.forEach(e=>{r+=this.getNumNGrams(n,e)}),this.preserveShort&&n>0&&0===r&&(r=1),s[e]=s[e-1]+r}let i=new Array(s[a]);for(let n=0;n<a;++n){let r=t[n],a=s[n];if(this.nGramWidths.forEach(s=>{let o=t[n+1]-t[n],l=this.getNumNGrams(o,s);this.createNGrams(e,r,i,a,l,s),a+=l}),this.preserveShort&&a===s[n]){let s=t[n+1]-t[n];if(0===s)continue;let o=s+2*this.padWidth;this.createNGrams(e,r,i,a,1,o)}}return[i,s]}};function $$(e,t,n,r,a,s,i,o){return new A$(n,r,a,s,i,o).compute(e,t)}function R$(e,t,n,r){if(!e.length)return;if(0===t.length){for(let t=0;t<e.length;++t)r.push(e.subarray(t,t+1));return}if(1===t.length){let a=t[0],s=e.indexOf(a);for(;-1!==s;){let t=e.subarray(0,s);(!n||0!==t.length)&&r.push(t),s=(e=e.subarray(s+1)).indexOf(a)}return void((!n||0!==e.length)&&r.push(e))}let a=0;for(let s=0;s<e.length+1;s++)if(s===e.length||-1!==t.indexOf(e[s])){let t=e.subarray(a,s);(!n||0!==t.length)&&r.push(t),a=s+1}}function F$(e,t,n){let r=e.length,a=[],s=0,i=0,o=new Array(r);for(let l=0;l<r;++l){let r=a.length;R$(e[l],t,n,a);let u=a.length-r;o[l]=u,s+=u,i=Math.max(i,u)}let l=va.getArrayFromDType("int32",2*s),u=new Array(s),h=[r,i],d=0;for(let e=0;e<r;++e)for(let t=0;t<o[e];++t)l[2*d]=e,l[2*d+1]=t,u[d]=a[d],++d;return[l,u,h]}function D$(e,t){let n=va.getArrayFromDType("int32",e.length);for(let r=0;r<e.length;++r)n[r]=va.fingerPrint64(e[r]).modulo(t).getLowBitsUnsigned();return n}var M$=vE((e,t)=>e-t),O$=RE((e,t,n,r)=>({real:e-n,imag:t-r})),L$=$E(Ur,M$,O$),P$={kernelName:Ur,backendName:"cpu",kernelFunc:L$};function z$(e,t){let n=new Array(e.rank);for(let r=0;r<n.length;r++)n[r]=e.shape[r]*t[r];let r=uo(n,e.dtype);for(let t=0;t<r.values.length;++t){let n=r.indexToLoc(t),a=new Array(e.rank);for(let t=0;t<a.length;t++)a[t]=n[t]%e.shape[t];let s=e.locToIndex(a);r.values[t]=e.values[s]}return r}var B$=(e,t)=>{let n=t.value-e.value;return 0===n?e.index-t.index:n};function W$(e,t,n=0,r=e.length-1){for(;r>n;){if(r-n>600){let a=r-n+1,s=t-n+1,i=Math.log(a),o=.5*Math.exp(2*i/3),l=.5*Math.sqrt(i*o*(a-o)/a)*Math.sign(s-a/2);W$(e,t,Math.max(n,Math.floor(t-s*o/a+l)),Math.min(r,Math.floor(t+(a-s)*o/a+l)))}let a=e[t],s=n,i=r;for(va.swap(e,n,t),B$(e[r],a)>0&&va.swap(e,n,r);s<i;){for(va.swap(e,s,i),s++,i--;B$(e[s],a)<0;)s+=1;for(;B$(e[i],a)>0;)i-=1}0===B$(e[n],a)?va.swap(e,n,i):(i+=1,va.swap(e,i,r)),i<=t&&(n=i+1),t<=i&&(r=i-1)}}function V$(e,t,n,r,a){let s=t[t.length-1],[i,o]=[e.length/s,s],l=va.getTypedArrayFromDType(n,i*r),u=va.getTypedArrayFromDType("int32",i*r);for(let t=0;t<i;t++){let n=t*o,s=e.subarray(n,n+o),i=new Array(s.length);s.forEach((e,t)=>i[t]={value:e,index:t}),r<i.length&&(W$(i,r),i=i.slice(0,r)),a&&i.sort(B$);let h=t*r,d=l.subarray(h,h+r),p=u.subarray(h,h+r);for(let e=0;e<r;e++)d[e]=i[e].value,p[e]=i[e].index}let h=t.slice();return h[h.length-1]=r,[uo(h,n,l),uo(h,"int32",u)]}function U$(e,t,n,r){let a=va.parseAxisParam(t,n)[0],s=[1,n[0],1];for(let e=0;e<a;e++)s[0]*=n[e];s[1]=n[a];for(let e=a+1;e<n.length;e++)s[2]*=n[e];let i=new Map,o=new Int32Array(n[a]),l=new Ja(s,r,e),u=[],h=1===s[0]&&1===s[2];for(let t=0;t<n[a];t++){let n;if(h)n=e[t].toString();else{let e=[];for(let n=0;n<s[0];n++)for(let r=0;r<s[2];r++)e.push(l.get(n,t,r));n=e.join(",")}let r=i.get(n);if(null!=r)o[t]=r;else{let e=i.size;i.set(n,e),o[t]=e,u.push(t)}}let d=s.slice();d[1]=i.size;let p=new Ja(d,r);u.forEach((e,t)=>{for(let n=0;n<s[0];n++)for(let r=0;r<s[2];r++)p.set(l.get(n,e,r),n,t,r)});let c=n.slice();return c[a]=d[1],{outputValues:p.values,outputShape:c,indices:o}}var G$="4.22.0";ui("cpu",()=>new gE,1);var H$=UE(Wt,e=>e>=0?e:Math.exp(e)-1),j$={kernelName:Wt,backendName:"cpu",kernelFunc:H$};function q$(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{alpha:s}=r;fE([a],"leakyRelu");let i=va.sizeFromShape(a.shape),o=n.data.get(a.dataId).values,l=va.getTypedArrayFromDType("float32",i);for(let e=0;e<o.length;e++)l[e]=o[e]<0?s*o[e]:o[e];return n.makeTensorInfo(a.shape,"float32",l)}var K$={kernelName:dn,backendName:"cpu",kernelFunc:q$},X$=vE((e,t)=>e<0?t*e:e);function Z$(e){let{inputs:t,backend:n}=e,{x:r,alpha:a}=t;fE([r,a],"prelu");let s=n.data.get(r.dataId).values,i=n.data.get(a.dataId).values,[o,l]=X$(r.shape,a.shape,s,i,"float32");return n.makeTensorInfo(l,"float32",o)}var Y$={kernelName:Yn,backendName:"cpu",kernelFunc:Z$},J$=UE(sr,e=>Math.max(0,e)),Q$={kernelName:sr,backendName:"cpu",kernelFunc:J$},eR=UE(dr,e=>Math.min(Math.max(0,e),6)),tR={kernelName:dr,backendName:"cpu",kernelFunc:eR};function nR(e,t,n,r,a){if("linear"===n)return SE({inputs:{x:t},backend:e});if("relu"===n)return J$({inputs:{x:t},backend:e});if("elu"===n)return H$({inputs:{x:t},backend:e});if("relu6"===n)return eR({inputs:{x:t},backend:e});if("prelu"===n)return Z$({inputs:{x:t,alpha:r},backend:e});if("leakyrelu"===n)return q$({inputs:{x:t},backend:e,attrs:{alpha:a}});if("sigmoid"===n)return p$({inputs:{x:t},backend:e});throw new Error(`Activation ${n} has not been implemented for the CPU backend.`)}function rR(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{shape:s}=r,i=va.sizeFromShape(a.shape),o=va.inferFromImplicitShape(s,i),l=va.sizeFromShape(o);va.assert(i===l,()=>`The new shape (${o}) has ${l} elements and the old shape (${a.shape}) has ${i} elements. The new shape and old shape must have the same number of elements.`),n.incRef(a.dataId);let u=n.data.get(a.dataId);if(null!=u.complexTensorInfos){let e=u.complexTensorInfos.real,t=u.complexTensorInfos.imag;e.shape=o,t.shape=o}return{dataId:a.dataId,shape:o,dtype:a.dtype}}var aR={kernelName:ir,backendName:"cpu",kernelFunc:rR};function sR(e){let{inputs:t,backend:n,attrs:r}=e,{a,b:s}=t,{transposeA:i,transposeB:o}=r;fE([a,s],"matMul");let l=a.shape.length,u=s.shape.length,h=i?a.shape[l-2]:a.shape[l-1],d=o?s.shape[u-1]:s.shape[u-2],p=i?a.shape[l-1]:a.shape[l-2],c=o?s.shape[u-2]:s.shape[u-1],f=a.shape.slice(0,-2),m=s.shape.slice(0,-2),g=va.sizeFromShape(f),y=va.sizeFromShape(m),b=Dl.assertAndGetBroadcastShape(a.shape.slice(0,-2),s.shape.slice(0,-2)).concat([p,c]);va.assert(h===d,()=>`Error in matMul: inner shapes (${h}) and (${d}) of Tensors with shapes ${a.shape} and ${s.shape} and transposeA=${i} and transposeB=${o} must match.`);let x=o?[y,c,d]:[y,d,c],v=rR({inputs:{x:a},backend:n,attrs:{shape:i?[g,h,p]:[g,p,h]}}),w=rR({inputs:{x:s},backend:n,attrs:{shape:x}}),k=i?v.shape[1]:v.shape[2],I=i?v.shape[2]:v.shape[1],S=o?w.shape[1]:w.shape[2],_=Math.max(g,y),N=n.data.get(v.dataId).values,T=n.data.get(w.dataId).values,C=va.computeStrides(v.shape),E=va.computeStrides(w.shape),[A,$,R]=i?[C[0],1,C[1]]:[C[0],C[1],1],[F,D,M]=o?[1,E[1],E[0]]:[E[1],1,E[0]],O=I*S,L=uo([_,I,S],v.dtype),P=L.values,z=n.blockSize;for(let e=0;e<_;e++){let t=e%g,n=e%y;for(let r=0;r<I;r+=z){let a=Math.min(r+z,I);for(let s=0;s<S;s+=z){let i=Math.min(s+z,S);for(let o=0;o<k;o+=z){let l=Math.min(o+z,k);for(let u=r;u<a;u++)for(let r=s;r<i;r++){let a=0;for(let e=o;e<l;e++){a+=N[t*A+u*$+e*R]*T[e*F+r*D+n*M]}P[e*O+(u*S+r)]+=a}}}}}return n.disposeIntermediateTensorInfo(v),n.disposeIntermediateTensorInfo(w),n.makeTensorInfo(b,L.dtype,L.values)}var iR={kernelName:it,backendName:"cpu",kernelFunc:sR};var oR={kernelName:aa,backendName:"cpu",kernelFunc:function(e){let t,n,r,{inputs:a,backend:s,attrs:i}=e,{a:o,b:l,bias:u,preluActivationWeights:h}=a,{transposeA:d,transposeB:p,activation:c,leakyreluAlpha:f}=i,m=[];t=sR({inputs:{a:o,b:l},attrs:{transposeA:d,transposeB:p},backend:s}),u&&(n=ME({inputs:{a:t,b:u},backend:s}),m.push(t),t=n),c&&(r=nR(s,t,c,h,f),m.push(t),t=r);for(let e of m)s.disposeIntermediateTensorInfo(e);return t}},lR=UE(Ue,e=>Math.acos(e)),uR={kernelName:Ue,backendName:"cpu",kernelFunc:lR},hR=UE(Ge,e=>Math.acosh(e)),dR={kernelName:Ge,backendName:"cpu",kernelFunc:hR};var pR={kernelName:je,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n}=e,r=t;fE(t,"addN");let a=r.map(e=>n.data.get(e.dataId).values),s=uo(r[0].shape,r[0].dtype),i=s.values;for(let e=0;e<r.length;e++){let t=a[e];for(let e=0;e<i.length;e++)i[e]+=t[e]}return n.makeTensorInfo(s.shape,s.dtype,s.values)}};var cR={kernelName:qe,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{axis:s,keepDims:i}=r;fE(a,"all");let o=va.parseAxisParam(s,a.shape),l=o,u=Gf.getAxesPermutation(l,a.shape.length),h=a;null!=u&&(h=HA({inputs:{x:a},backend:n,attrs:{perm:u}}),l=Gf.getInnerMostAxes(l.length,a.shape.length)),Gf.assertAxesAreInnerMostDims("all",l,h.shape.length);let[d,p]=Gf.computeOutAndReduceShapes(h.shape,l),c=va.sizeFromShape(p),f=va.makeZerosTypedArray(va.sizeFromShape(d),h.dtype),m=n.data.get(h.dataId).values;for(let e=0;e<f.length;++e){let t=e*c,n=m[t];for(let e=0;e<c;++e){let r=m[t+e];n=n&&r}f[e]=n}null!=u&&n.disposeIntermediateTensorInfo(h);let g=n.makeTensorInfo(d,h.dtype,f);if(i){let e=rR({inputs:{x:g},backend:n,attrs:{shape:Gf.expandShapeToKeepDim(d,o)}});return n.disposeIntermediateTensorInfo(g),e}return g}};var fR={kernelName:Ke,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{axis:s,keepDims:i}=r;fE(a,"any");let o=va.parseAxisParam(s,a.shape),l=o,u=Gf.getAxesPermutation(l,a.shape.length),h=a;null!=u&&(h=HA({inputs:{x:a},backend:n,attrs:{perm:u}}),l=Gf.getInnerMostAxes(l.length,a.shape.length)),Gf.assertAxesAreInnerMostDims("any",l,h.shape.length);let[d,p]=Gf.computeOutAndReduceShapes(h.shape,l),c=va.sizeFromShape(p),f=va.makeZerosTypedArray(va.sizeFromShape(d),h.dtype),m=n.data.get(h.dataId).values;for(let e=0;e<f.length;++e){let t=e*c,n=m[t];for(let e=0;e<c;++e){let r=m[t+e];n=n||r}f[e]=n}null!=u&&n.disposeIntermediateTensorInfo(h);let g=n.makeTensorInfo(d,h.dtype,f);if(i){let e=rR({inputs:{x:g},backend:n,attrs:{shape:Gf.expandShapeToKeepDim(d,o)}});return n.disposeIntermediateTensorInfo(g),e}return g}};var mR={kernelName:Xe,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{axis:s}=r;fE(a,"argMax");let i=va.parseAxisParam(s,a.shape),o=Gf.getAxesPermutation(i,a.shape.length),l=a,u=[];null!=o&&(l=HA({inputs:{x:a},backend:n,attrs:{perm:o}}),u.push(l),i=Gf.getInnerMostAxes(i.length,l.shape.length)),i=[i[0]],Gf.assertAxesAreInnerMostDims("argMax",i,l.shape.length);let[h,d]=Gf.computeOutAndReduceShapes(l.shape,i),p=va.sizeFromShape(h),c=va.makeZerosTypedArray(p,"int32"),f=va.sizeFromShape(d),m=n.data.get(l.dataId).values;for(let e=0;e<c.length;++e){let t=e*f,n=m[t],r=0;for(let e=0;e<f;++e){let a=m[t+e];a>n&&(n=a,r=e)}c[e]=r}return u.forEach(e=>n.disposeIntermediateTensorInfo(e)),n.makeTensorInfo(h,"int32",c)}};var gR={kernelName:Ze,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{axis:s}=r;fE(a,"argMin");let i=va.parseAxisParam(s,a.shape),o=Gf.getAxesPermutation(i,a.shape.length),l=a,u=[];null!=o&&(l=HA({inputs:{x:a},backend:n,attrs:{perm:o}}),u.push(l),i=Gf.getInnerMostAxes(i.length,l.shape.length)),i=[i[0]],Gf.assertAxesAreInnerMostDims("argMin",i,l.shape.length);let[h,d]=Gf.computeOutAndReduceShapes(l.shape,i),p=va.sizeFromShape(h),c=va.makeZerosTypedArray(p,"int32"),f=va.sizeFromShape(d),m=n.data.get(l.dataId).values;for(let e=0;e<c.length;++e){let t=e*f,n=m[t],r=0;for(let e=0;e<f;++e){let a=m[t+e];a<n&&(n=a,r=e)}c[e]=r}return u.forEach(e=>n.disposeIntermediateTensorInfo(e)),n.makeTensorInfo(h,"int32",c)}},yR=UE(Ye,e=>Math.asin(e)),bR={kernelName:Ye,backendName:"cpu",kernelFunc:yR},xR=UE(Je,e=>Math.asinh(e)),vR={kernelName:Je,backendName:"cpu",kernelFunc:xR},wR=UE(Qe,e=>Math.atan(e)),kR={kernelName:Qe,backendName:"cpu",kernelFunc:wR},IR=vE((e,t)=>Math.atan2(e,t)),SR=$E(tt,IR),_R={kernelName:tt,backendName:"cpu",kernelFunc:SR},NR=UE(et,e=>Math.atanh(e)),TR={kernelName:et,backendName:"cpu",kernelFunc:NR};function CR(e,t,n,r,a,s){let i=a.strideHeight,o=a.strideWidth,l=a.dilationHeight,u=a.dilationWidth,h=a.effectiveFilterHeight,d=a.effectiveFilterWidth,p=a.padInfo.top,c=a.padInfo.left,f="max"===s?Number.NEGATIVE_INFINITY:Number.POSITIVE_INFINITY,m=uo(a.outShape,n),g=m.values,y=a.outShape[1]*a.outShape[2]*a.outShape[3],b=a.outShape[2]*a.outShape[3],x=a.outShape[3];for(let t=0;t<a.batchSize;++t){let n=t*y,m=t*r[0];for(let t=0;t<a.inChannels;++t)for(let y=0;y<a.outHeight;++y){let v=y*i-p,w=Math.max(0,v),k=Math.min(a.inHeight,h+v),I=n+y*b;for(let n=0;n<a.outWidth;++n){let i=n*o-c,h=Math.max(0,i),p=Math.min(a.inWidth,d+i),y=f,b=0,v=0;for(let n=w;n<k;n+=l){let a=m+n*r[1];for(let n=h;n<p;n+=u){let i=e[a+n*r[2]+t];"max"===s&&i>y?y=i:"avg"===s&&(b+=i,v++)}if(isNaN(y))break}g[I+n*x+t]="avg"===s?b/v:y}}}return m}function ER(e,t,n,r,a=!1,s=!1){let i=uo(r.outShape,"int32"),o=r.strideHeight,l=r.strideWidth,u=r.dilationHeight,h=r.dilationWidth,d=r.effectiveFilterHeight,p=r.effectiveFilterWidth,c=r.padInfo.top,f=r.padInfo.left,m=uo(t,n,e);for(let e=0;e<r.batchSize;++e)for(let t=0;t<r.inChannels;++t)for(let n=0;n<r.outHeight;++n){let g=n*o-c,y=g;for(;y<0;)y+=u;let b=Math.min(r.inHeight,d+g);for(let o=0;o<r.outWidth;++o){let d=o*l-f,c=d;for(;c<0;)c+=h;let x=Math.min(r.inWidth,p+d),v=Number.NEGATIVE_INFINITY,w=-1;for(let n=y;n<b;n+=u){let i=n-g;for(let o=c;o<x;o+=h){let l=o-d,u=m.get(e,n,o,t);u>v&&(v=u,w=a?s?((e*r.inHeight+n)*r.inWidth+o)*r.inChannels+t:(n*r.inWidth+o)*r.inChannels+t:i*p+l)}}i.set(w,e,n,o,t)}}return i}function AR(e,t,n,r,a,s){let i=a.strideDepth,o=a.strideHeight,l=a.strideWidth,u=a.dilationDepth,h=a.dilationHeight,d=a.dilationWidth,p=a.effectiveFilterDepth,c=a.effectiveFilterHeight,f=a.effectiveFilterWidth,m=a.padInfo.front,g=a.padInfo.top,y=a.padInfo.left,b="max"===s?Number.NEGATIVE_INFINITY:Number.POSITIVE_INFINITY,x=uo(a.outShape,n),v=x.values,w=a.outShape[1]*a.outShape[2]*a.outShape[3]*a.outShape[4],k=a.outShape[2]*a.outShape[3]*a.outShape[4],I=a.outShape[3]*a.outShape[4],S=a.outShape[4];for(let t=0;t<a.batchSize;++t){let n=t*w,x=t*r[0];for(let t=0;t<a.inChannels;++t)for(let w=0;w<a.outDepth;++w){let _=w*i-m,N=_;for(;N<0;)N+=u;let T=Math.min(a.inDepth,p+_),C=n+w*k;for(let n=0;n<a.outHeight;++n){let i=n*o-g,p=i;for(;p<0;)p+=h;let m=Math.min(a.inHeight,c+i),w=C+n*I;for(let n=0;n<a.outWidth;++n){let i=n*l-y,o=i;for(;o<0;)o+=d;let c=Math.min(a.inWidth,f+i),g=w+n*S,k=b,I=0,_=0;for(let n=N;n<T;n+=u){let a=x+n*r[1];for(let n=p;n<m;n+=h){let i=a+n*r[2];for(let n=o;n<c;n+=d){let a=e[i+n*r[3]+t];if("max"===s&&a>k?k=a:"avg"===s&&(I+=a,_++),isNaN(k))break}if(isNaN(k))break}if(isNaN(k))break}v[g+t]="avg"===s?I/Math.max(_,1):k}}}}return x}var $R={kernelName:nt,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t;fE(a,"avgPool");let{filterSize:s,strides:i,pad:o,dimRoundingMode:l}=r;va.assert(Gf.eitherStridesOrDilationsAreOne(i,1),()=>`Error in avgPool: Either strides or dilations must be 1. Got strides ${i} and dilations '1'`);let u,h=Gf.computePool2DInfo(a.shape,s,i,1,o,l);if(1===h.filterWidth&&1===h.filterHeight&&va.arraysEqual(h.inShape,h.outShape))u=SE({inputs:{x:a},backend:n});else{let e=n.data.get(a.dataId).values,t=va.computeStrides(a.shape),r=CR(e,a.shape,a.dtype,t,h,"avg");u=n.makeTensorInfo(h.outShape,a.dtype,r.values)}return u}};var RR={kernelName:at,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{filterSize:s,strides:i,pad:o,dimRoundingMode:l,dataFormat:u}=r;fE(a,"avgPool3d");let h=Gf.computePool3DInfo(a.shape,s,i,1,o,l,u),d=AR(n.data.get(a.dataId).values,a.shape,a.dtype,va.computeStrides(a.shape),h,"avg");return n.makeTensorInfo(d.shape,"float32",d.values)}};var FR={kernelName:st,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{dy:a,input:s}=t,{filterSize:i,strides:o,pad:l,dimRoundingMode:u}=r;fE([a,s],"avgPool3DGrad");let h=Gf.computePool3DInfo(s.shape,i,o,1,l,u),d=h.strideDepth,p=h.strideHeight,c=h.strideWidth,f=h.filterDepth,m=h.filterHeight,g=h.filterWidth,y=h.dilationDepth,b=h.dilationHeight,x=h.dilationWidth,v=h.effectiveFilterDepth,w=h.effectiveFilterHeight,k=h.effectiveFilterWidth,I=v-1-h.padInfo.front,S=k-1-h.padInfo.left,_=w-1-h.padInfo.top,N=uo(s.shape,"float32"),T=1/(f*m*g),C=n.bufferSync(a);for(let e=0;e<h.batchSize;++e)for(let t=0;t<h.inChannels;++t)for(let n=0;n<h.inDepth;++n)for(let r=0;r<h.inHeight;++r)for(let a=0;a<h.inWidth;++a){let s=n-I,i=r-_,o=a-S,l=0;for(let n=0;n<v;n+=y){let r=(s+n)/d;if(!(r<0||r>=h.outDepth||Math.floor(r)!==r))for(let n=0;n<w;n+=b){let a=(i+n)/p;if(!(a<0||a>=h.outHeight||Math.floor(a)!==a))for(let n=0;n<k;n+=x){let s=(o+n)/c;s<0||s>=h.outWidth||Math.floor(s)!==s||(l+=C.get(e,r,a,s,t))}}}N.set(l*T,e,n,r,a,t)}return n.makeTensorInfo(N.shape,N.dtype,N.values)}};var DR={kernelName:rt,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{dy:a,input:s}=t,i=s;fE([a,s],"avgPoolGrad");let{filterSize:o,strides:l,pad:u}=r,h=Gf.computePool2DInfo(i.shape,o,l,1,u),d=h.strideHeight,p=h.strideWidth,c=h.filterHeight,f=h.filterWidth,m=h.dilationHeight,g=h.dilationWidth,y=h.effectiveFilterHeight,b=h.effectiveFilterWidth,x=b-1-h.padInfo.left,v=y-1-h.padInfo.top,w=uo(i.shape,"float32"),k=1/(c*f),I=n.data.get(a.dataId).values,S=uo(a.shape,"float32",I);for(let e=0;e<h.batchSize;++e)for(let t=0;t<h.inChannels;++t)for(let n=0;n<h.inHeight;++n)for(let r=0;r<h.inWidth;++r){let a=n-v,s=r-x,i=0;for(let n=0;n<y;n+=m){let r=(a+n)/d;if(!(r<0||r>=h.outHeight||Math.floor(r)!==r))for(let n=0;n<b;n+=g){let a=(s+n)/p;a<0||a>=h.outWidth||Math.floor(a)!==a||(i+=S.get(e,r,a,t))}}w.set(i*k,e,n,r,t)}return n.makeTensorInfo(w.shape,w.dtype,w.values)}};var MR={kernelName:Qt,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a,scale:s,offset:i,mean:o,variance:l}=t;va.assert(o.shape.length===l.shape.length,()=>"Batch normalization gradient requires mean and variance to have equal ranks."),va.assert(null==i||o.shape.length===i.shape.length,()=>"Batch normalization gradient requires mean and offset to have equal ranks."),va.assert(null==s||o.shape.length===s.shape.length,()=>"Batch normalization gradient requires mean and scale to have equal ranks."),fE([a,o,l,s,i],"batchNorm");let{varianceEpsilon:u}=r;null==u&&(u=.001);let h=n.data.get(a.dataId).values,d=n.data.get(o.dataId).values,p=n.data.get(l.dataId).values,c=s?n.data.get(s.dataId).values:new Float32Array([1]),f=i?n.data.get(i.dataId).values:new Float32Array([0]),m=new Float32Array(h.length),g=f.length,y=c.length,b=p.length,x=d.length,v=0,w=0,k=0,I=0;for(let e=0;e<h.length;++e)m[e]=f[v++]+(h[e]-d[w++])*c[k++]/Math.sqrt(p[I++]+u),v>=g&&(v=0),w>=x&&(w=0),k>=y&&(k=0),I>=b&&(I=0);return n.makeTensorInfo(a.shape,a.dtype,m)}};var OR={kernelName:ot,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{blockShape:s,crops:i}=r;fE([a],"batchToSpaceND");let o=s.reduce((e,t)=>e*t),l=Gf.getReshaped(a.shape,s,o),u=Gf.getPermuted(l.length,s.length),h=Gf.getReshapedPermuted(a.shape,s,o),d=Gf.getSliceBeginCoords(i,s.length),p=Gf.getSliceSize(h,i,s.length),c=rR({inputs:{x:a},backend:n,attrs:{shape:l}}),f=HA({inputs:{x:c},backend:n,attrs:{perm:u}}),m=rR({inputs:{x:f},backend:n,attrs:{shape:h}}),g=m$({inputs:{x:m},backend:n,attrs:{begin:d,size:p}});return n.disposeIntermediateTensorInfo(c),n.disposeIntermediateTensorInfo(f),n.disposeIntermediateTensorInfo(m),g}};var LR={kernelName:lt,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a,weights:s}=t,{size:i}=r,o=LE(n.data.get(a.dataId).values,n.data.get(s.dataId).values,s.dtype,s.shape,i);return n.makeTensorInfo([i],s.dtype,o)}};var PR={kernelName:dt,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n}=e,{s0:r,s1:a}=t,s=n.data.get(r.dataId).values,i=n.data.get(a.dataId).values,o=Gf.assertAndGetBroadcastShape(Array.from(s),Array.from(i));return n.makeTensorInfo([o.length],"int32",Int32Array.from(o))}},zR=UE(ft,(e,t)=>{let n=t;return e>n.clipValueMax?n.clipValueMax:e<n.clipValueMin?n.clipValueMin:e}),BR={kernelName:ft,backendName:"cpu",kernelFunc:zR},WR={kernelName:gt,backendName:"cpu",kernelFunc:e=>{let{x:t}=e.inputs,n=e.backend,r=new Float32Array(va.sizeFromShape(t.shape)),a=n.data.get(t.dataId),s=a.complexTensorInfos.real,i=a.complexTensorInfos.imag,o=n.data.get(s.dataId).values,l=n.data.get(i.dataId).values;for(let e=0;e<o.length;e++){let t=o[e],n=l[e];r[e]=Math.hypot(t,n)}return n.makeOutput(r,t.shape,"float32")}};function VR(e){let{inputs:t,backend:n}=e,{input:r}=t,a=n.data.get(r.dataId).complexTensorInfos.imag,s=n.data.get(a.dataId).values;return n.makeTensorInfo(a.shape,a.dtype,s)}var UR={kernelName:on,backendName:"cpu",kernelFunc:VR};function GR(e){let{inputs:t,backend:n,attrs:r}=e,{axis:a}=r,s=va.parseAxisParam(a,t[0].shape)[0],i=t.map(e=>e.shape);Gf.assertParamsConsistent(i,s);let o=Gf.computeOutShape(t.map(e=>e.shape),s);if(0===va.sizeFromShape(o))return n.makeTensorInfo(o,t[0].dtype,[]);let l=t.filter(e=>va.sizeFromShape(e.shape)>0);if(1===l.length)return SE({inputs:{x:l[0]},backend:n});if("complex64"===l[0].dtype){let e=l.map(e=>NE({inputs:{input:e},backend:n})),t=l.map(e=>VR({inputs:{input:e},backend:n})),r=GR({inputs:e,backend:n,attrs:{axis:s}}),a=GR({inputs:t,backend:n,attrs:{axis:s}}),i=wE({inputs:{real:r,imag:a},backend:n});return e.forEach(e=>n.disposeIntermediateTensorInfo(e)),t.forEach(e=>n.disposeIntermediateTensorInfo(e)),n.disposeIntermediateTensorInfo(r),n.disposeIntermediateTensorInfo(a),i}let u=l.map(e=>{let t=[-1,va.sizeFromShape(e.shape.slice(s))];return rR({inputs:{x:e},backend:n,attrs:{shape:t}})}),h=u.map(e=>({vals:n.data.get(e.dataId).values,shape:e.shape}));o=Gf.computeOutShape(u.map(e=>e.shape),1);let d=1===u[0].shape[0],p=KE(h,o,t[0].dtype,d),c=Gf.computeOutShape(l.map(e=>e.shape),s),f=n.makeTensorInfo(c,t[0].dtype,p);return u.forEach(e=>n.disposeIntermediateTensorInfo(e)),f}var HR={kernelName:yt,backendName:"cpu",kernelFunc:GR};function jR(e){let{inputs:t,backend:n,attrs:r}=e,{x:a,filter:s}=t,{strides:i,pad:o,dataFormat:l,dilations:u,dimRoundingMode:h}=r;fE([a,s],"conv2d");let d=Gf.convertConv2DDataFormat(l),p=Gf.computeConv2DInfo(a.shape,s.shape,i,u,o,h,!1,d),c=p.filterHeight,f=p.filterWidth,m=p.dilationHeight,g=p.dilationWidth,y=p.padInfo.left,b=p.padInfo.top,x="channelsLast"===p.dataFormat,v=new Ja(p.outShape,a.dtype),w=va.computeStrides(a.shape),k=va.computeStrides(s.shape),I=w[0],S=x?w[1]:w[2],_=x?w[2]:1,N=x?1:w[1],T=v.strides[0],C=x?v.strides[1]:v.strides[2],E=x?v.strides[2]:1,A=x?1:v.strides[1],$=n.data.get(a.dataId).values,R=n.data.get(s.dataId).values,F=v.values;for(let e=0;e<p.batchSize;++e){let t=e*I,n=e*T;for(let e=0;e<p.outHeight;++e){let r=n+e*C,a=e*p.strideHeight-b;for(let e=0;e<c;++e){let n=a+e*m;if(n<0||n>=p.inHeight)continue;let s=e*k[0],i=t+n*S;for(let e=0;e<p.outWidth;++e){let t=r+e*E,n=e*p.strideWidth-y;for(let e=0;e<f;++e){let r=n+e*g;if(r<0||r>=p.inWidth)continue;let a=i+r*_,o=s+e*k[1];for(let e=0;e<p.inChannels;++e){let n=$[a+e*N];for(let e=0;e<p.outChannels;++e)F[t+e*A]+=n*R[o+e];o+=p.outChannels}}}}}}return n.makeTensorInfo(v.shape,v.dtype,F)}var qR={kernelName:bt,backendName:"cpu",kernelFunc:jR};var KR={kernelName:xt,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a,dy:s}=t,{strides:i,pad:o,dataFormat:l,dimRoundingMode:u,filterShape:h}=r;fE([a,s],"conv2dBackpropFilter");let d=Gf.convertConv2DDataFormat(l),p=Gf.computeConv2DInfo(a.shape,h,i,1,o,u,!1,d),{strideHeight:c,strideWidth:f,filterHeight:m,filterWidth:g}=p,y="channelsLast"===p.dataFormat,b=new Ja(p.filterShape,"float32"),x=p.padInfo.left,v=p.padInfo.top,w=n.data.get(a.dataId).values,k=n.data.get(s.dataId).values,I=new Ja(a.shape,a.dtype,w),S=new Ja(s.shape,s.dtype,k);for(let e=0;e<m;++e){let t=Math.max(0,Math.ceil((v-e)/c)),n=Math.min(p.outHeight,(p.inHeight+v-e)/c);for(let r=0;r<g;++r){let a=Math.max(0,Math.ceil((x-r)/f)),s=Math.min(p.outWidth,(p.inWidth+x-r)/f);for(let i=0;i<p.inChannels;++i)for(let o=0;o<p.outChannels;++o){let l=0;for(let u=0;u<p.batchSize;++u)for(let h=t;h<n;++h){let t=e+h*c-v;for(let e=a;e<s;++e){let n=r+e*f-x;l+=y?I.get(u,t,n,i)*S.get(u,h,e,o):I.get(u,i,t,n)*S.get(u,o,h,e)}}b.set(l,e,r,i,o)}}}return n.makeTensorInfo(b.shape,b.dtype,b.values)}};var XR={kernelName:vt,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{dy:a,filter:s}=t,{inputShape:i,strides:o,pad:l,dataFormat:u,dimRoundingMode:h}=r;fE([a,s],"conv2dBackpropInput");let d=va.computeStrides(s.shape),p=va.computeStrides(a.shape),c=Gf.convertConv2DDataFormat(u),f=Gf.computeConv2DInfo(i,s.shape,o,1,l,h,!1,c),m=new Ja(f.inShape,"float32"),g=m.values,y=n.data.get(a.dataId).values,b=n.data.get(s.dataId).values,[x,v,w]=d,{batchSize:k,filterHeight:I,filterWidth:S,inChannels:_,inHeight:N,inWidth:T,outChannels:C,outHeight:E,outWidth:A,strideHeight:$,strideWidth:R}=f;c=f.dataFormat;let F=I-1-f.padInfo.top,D=S-1-f.padInfo.left,M="channelsLast"===c,O=m.strides[0],L=M?m.strides[1]:m.strides[2],P=M?m.strides[2]:1,z=M?1:m.strides[1],B=p[0],W=M?p[1]:p[2],V=M?p[2]:1,U=M?1:p[1];for(let e=0;e<k;++e)for(let t=0;t<_;++t)for(let n=0;n<N;++n){let r=n-F,a=Math.max(0,Math.ceil(r/$)),s=Math.min(E,(I+r)/$);for(let i=0;i<T;++i){let o=i-D,l=Math.max(0,Math.ceil(o/R)),u=Math.min(A,(S+o)/R),h=0;for(let n=a;n<s;++n){let a=n*$-r;for(let r=l;r<u;++r){let s=B*e+W*n+V*r,i=x*(I-1-a)+v*(S-1-(r*R-o))+w*t;for(let e=0;e<C;++e){h+=y[s+U*e]*b[i+e]}}}g[O*e+L*n+P*i+z*t]=h}}return n.makeTensorInfo(m.shape,m.dtype,m.values)}};var ZR={kernelName:wt,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a,filter:s}=t,{strides:i,pad:o,dilations:l}=r;fE([a,s],"conv3d");let u=Gf.computeConv3DInfo(a.shape,s.shape,i,l,o),{filterDepth:h,filterHeight:d,filterWidth:p,dilationDepth:c,dilationHeight:f,dilationWidth:m,padInfo:g}=u,y=g.front,b=g.left,x=g.top,v=new Ja(u.outShape,a.dtype),w=n.data.get(a.dataId).values,k=n.data.get(s.dataId).values,I=v.values,S=va.computeStrides(a.shape),_=va.computeStrides(s.shape);for(let e=0;e<u.batchSize;++e){let t=e*S[0],n=e*v.strides[0];for(let e=0;e<u.outDepth;++e){let r=n+e*v.strides[1],a=e*u.strideDepth-y;for(let e=0;e<h;++e){let n=a+e*c;if(n<0||n>=u.inDepth)continue;let s=e*_[0],i=t+n*S[1];for(let e=0;e<u.outHeight;++e){let t=r+e*v.strides[2],n=e*u.strideHeight-x;for(let e=0;e<d;++e){let r=n+e*f;if(r<0||r>=u.inHeight)continue;let a=s+e*_[1],o=i+r*S[2];for(let e=0;e<u.outWidth;++e){let n=t+e*u.outChannels,r=e*u.strideWidth-b;for(let e=0;e<p;++e){let t=r+e*m;if(t<0||t>=u.inWidth)continue;let s=a+e*_[2],i=o+t*u.inChannels,l=s;for(let e=0;e<u.inChannels;++e){let t=w[i+e];for(let e=0;e<u.outChannels;++e)I[n+e]+=t*k[l+e];l+=u.outChannels}}}}}}}}return n.makeTensorInfo(v.shape,v.dtype,v.values)}};var YR={kernelName:kt,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a,dy:s}=t,{strides:i,pad:o,filterShape:l}=r;fE([a,s],"conv3dBackpropFilterV2");let u=va.computeStrides(a.shape),h=va.computeStrides(s.shape),d=Gf.computeConv3DInfo(a.shape,l,i,1,o),p=d.strideDepth,c=d.strideHeight,f=d.strideWidth,m=d.filterDepth,g=d.filterHeight,y=d.filterWidth,b=new Ja(d.filterShape,"float32"),x=b.values,[v,w,k,I]=b.strides,S=n.data.get(s.dataId).values,[_,N,T,C]=h,E=n.data.get(a.dataId).values,[A,$,R,F]=u,D=d.padInfo.front,M=d.padInfo.left,O=d.padInfo.top;for(let e=0;e<m;++e){let t=Math.max(0,Math.ceil((D-e)/p)),n=Math.min(d.outDepth,(d.inDepth+D-e)/p),r=e*v;for(let a=0;a<g;++a){let s=Math.max(0,Math.ceil((O-a)/c)),i=Math.min(d.outHeight,(d.inHeight+O-a)/c),o=a*w+r;for(let r=0;r<y;++r){let l=Math.max(0,Math.ceil((M-r)/f)),u=Math.min(d.outWidth,(d.inWidth+M-r)/f),h=r*k+o;for(let o=0;o<d.inChannels;++o){let m=o*I+h;for(let h=0;h<d.outChannels;++h){let g=0;for(let m=0;m<d.batchSize;++m){let d=m*A,y=m*_;for(let m=t;m<n;++m){let t=(e+m*p-D)*$+d,n=m*N+y;for(let e=s;e<i;++e){let s=(a+e*c-O)*R+t,i=e*T+n;for(let e=l;e<u;++e){let t=e*C+i;g+=E[(r+e*f-M)*F+s+o]*S[t+h]}}}}x[m+h]=g}}}}}return n.makeTensorInfo(b.shape,b.dtype,b.values)}};var JR={kernelName:It,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{dy:a,filter:s}=t,{pad:i,strides:o,inputShape:l}=r;fE([a],"conv3dBackpropInputV2");let u=va.computeStrides(a.shape),h=va.computeStrides(s.shape),d=Gf.computeConv3DInfo(l,s.shape,o,1,i),p=new Ja(d.inShape,"float32"),c=p.values,[f,m,g,y]=p.strides,b=n.data.get(a.dataId).values,[x,v,w,k]=u,I=n.data.get(s.dataId).values,[S,_,N,T]=h,{batchSize:C,filterDepth:E,filterHeight:A,filterWidth:$,inChannels:R,inDepth:F,inHeight:D,inWidth:M,outChannels:O,outDepth:L,outHeight:P,outWidth:z,strideDepth:B,strideHeight:W,strideWidth:V}=d,U=E-1-d.padInfo.front,G=A-1-d.padInfo.top,H=$-1-d.padInfo.left;for(let e=0;e<C;++e)for(let t=0;t<R;++t)for(let n=0;n<F;++n){let r=n-U,a=Math.max(0,Math.ceil(r/B)),s=Math.min(L,(E+r)/B);for(let i=0;i<D;++i){let o=i-G,l=Math.max(0,Math.ceil(o/W)),u=Math.min(P,(A+o)/W);for(let h=0;h<M;++h){let d=h-H,p=Math.max(0,Math.ceil(d/V)),C=Math.min(z,($+d)/V),R=0;for(let n=a;n<s;++n){let a=n*B-r;for(let r=l;r<u;++r){let s=r*W-o;for(let i=p;i<C;++i){let o=x*e+v*n+w*r+k*i,l=S*(E-1-a)+_*(A-1-s)+N*($-1-(i*V-d))+T*t;for(let e=0;e<O;++e){R+=b[o+e]*I[l+e]}}}}c[f*e+m*n+g*i+y*h+t]=R}}}return n.makeTensorInfo(p.shape,p.dtype,p.values)}},QR=UE(St,e=>Math.cos(e)),eF={kernelName:St,backendName:"cpu",kernelFunc:QR},tF=UE(_t,e=>Math.cosh(e)),nF={kernelName:_t,backendName:"cpu",kernelFunc:tF};var rF={kernelName:Ct,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{image:a,boxes:s,boxInd:i}=t,{cropSize:o,method:l,extrapolationValue:u}=r,[h,d,p,c]=a.shape,f=s.shape[0],[m,g]=o,y=uo([f,m,g,c],"float32"),b=n.data.get(s.dataId).values,x=n.data.get(i.dataId).values,v=n.data.get(a.dataId).values,w=va.computeStrides(a.shape),k=va.computeStrides(y.shape);for(let e=0;e<f;e++){let t=4*e,n=b[t],r=b[t+1],a=b[t+2],s=b[t+3],i=x[e];if(i>=h)continue;let o=m>1?(a-n)*(d-1)/(m-1):0,f=g>1?(s-r)*(p-1)/(g-1):0;for(let t=0;t<m;t++){let h=m>1?n*(d-1)+t*o:.5*(n+a)*(d-1);if(h<0||h>d-1)for(let n=0;n<g;n++)for(let r=0;r<c;r++){let a=r+n*k[2]+t*k[1]+e*k[0];y.values[a]=u}else if("bilinear"===l){let n=Math.floor(h),a=Math.ceil(h),o=h-n;for(let l=0;l<g;l++){let h=g>1?r*(p-1)+l*f:.5*(r+s)*(p-1);if(h<0||h>p-1){for(let n=0;n<c;n++){let r=n+l*k[2]+t*k[1]+e*k[0];y.values[r]=u}continue}let d=Math.floor(h),m=Math.ceil(h),b=h-d;for(let r=0;r<c;r++){let s=r+d*w[2]+n*w[1]+i*w[0],u=v[s];s=r+m*w[2]+n*w[1]+i*w[0];let h=v[s];s=r+d*w[2]+a*w[1]+i*w[0];let p=v[s];s=r+m*w[2]+a*w[1]+i*w[0];let c=u+(h-u)*b,f=p+(v[s]-p)*b;s=r+l*k[2]+t*k[1]+e*k[0],y.values[s]=c+(f-c)*o}}}else for(let n=0;n<g;++n){let a=g>1?r*(p-1)+n*f:.5*(r+s)*(p-1);if(a<0||a>p-1){for(let r=0;r<c;r++){let a=r+n*k[2]+t*k[1]+e*k[0];y.values[a]=u}continue}let o=Math.round(a),l=Math.round(h);for(let r=0;r<c;r++){let a=r+o*w[2]+l*w[1]+i*w[0],s=r+n*k[2]+t*k[1]+e*k[0];y.values[s]=v[a]}}}}return n.makeTensorInfo(y.shape,y.dtype,y.values)}};var aF={kernelName:Nt,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{axis:s,exclusive:i,reverse:o}=r;fE(a,"cumprod");let l=Gf.getAxesPermutation([s],a.shape.length),u=a;null!=l&&(u=HA({inputs:{x:a},backend:n,attrs:{perm:l}}));let h=Gf.getInnerMostAxes(1,a.shape.length)[0];if(h!==u.shape.length-1)throw new Error(`backend.cumprod in CPU expects an inner-most axis=${u.shape.length-1} but got axis=${h}`);let d=ps(u.dtype,"int32"),p=va.makeOnesTypedArray(va.sizeFromShape(u.shape),d),c=n.data.get(u.dataId).values,f=u.shape[u.shape.length-1],m=o?(e,t)=>e+f-t-1:(e,t)=>e+t;for(let e=0;e<c.length;e+=f)for(let t=0;t<f;t++){let n=m(e,t);if(0===t)p[n]=i?1:c[n];else{let r=m(e,t-1);p[n]=i?c[r]*p[r]:c[n]*p[r]}}let g=n.makeTensorInfo(u.shape,d,p);if(null!=l){let e=HA({inputs:{x:g},backend:n,attrs:{perm:Gf.getUndoAxesPermutation(l)}});return n.disposeIntermediateTensorInfo(g),n.disposeIntermediateTensorInfo(u),e}return g}};var sF={kernelName:Tt,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{axis:s,exclusive:i,reverse:o}=r;fE(a,"cumsum");let l=Gf.getAxesPermutation([s],a.shape.length),u=a;null!=l&&(u=HA({inputs:{x:a},backend:n,attrs:{perm:l}}));let h=Gf.getInnerMostAxes(1,a.shape.length)[0];if(h!==u.shape.length-1)throw new Error(`backend.cumsum in CPU expects an inner-most axis=${u.shape.length-1} but got axis=${h}`);let d=ps(u.dtype,"int32"),p=va.makeZerosTypedArray(va.sizeFromShape(u.shape),d),c=n.data.get(u.dataId).values,f=u.shape[u.shape.length-1],m=o?(e,t)=>e+f-t-1:(e,t)=>e+t;for(let e=0;e<c.length;e+=f)for(let t=0;t<f;t++){let n=m(e,t);if(0===t)p[n]=i?0:c[n];else{let r=m(e,t-1);p[n]=i?c[r]+p[r]:c[n]+p[r]}}let g=n.makeTensorInfo(u.shape,d,p);if(null!=l){let e=HA({inputs:{x:g},backend:n,attrs:{perm:Gf.getUndoAxesPermutation(l)}});return n.disposeIntermediateTensorInfo(g),n.disposeIntermediateTensorInfo(u),e}return g}};var iF={kernelName:Et,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a,weights:s}=t,{size:i,binaryOutput:o}=r;if(1===a.shape.length){let e=LE(n.data.get(a.dataId).values,n.data.get(s.dataId).values,s.dtype,s.shape,i);return n.makeTensorInfo([i],s.dtype,e)}if(2===a.shape.length){let e=PE(n.bufferSync(a),n.bufferSync(s),i,o);return n.makeTensorInfo(e.shape,s.dtype,e.values)}throw new Error(`Error in denseBincount: input must be at most rank 2, but got rank${a.shape.length}.`)}};var oF={kernelName:At,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{blockSize:s,dataFormat:i}=r;va.assert("NHWC"===i,()=>`Only NHWC dataFormat supported on CPU for depthToSpace. Got ${i}`);let o=a.shape[0],l=a.shape[1],u=a.shape[2],h=a.shape[3],d=l*s,p=u*s,c=h/(s*s),f=n.data.get(a.dataId).values,m=new Float32Array(o*d*p*c),g=0;for(let e=0;e<o;++e)for(let t=0;t<d;++t){let n=Math.floor(t/s),r=t%s;for(let t=0;t<p;++t){let a=Math.floor(t/s),i=(r*s+t%s)*c;for(let t=0;t<c;++t){let r=t+i+h*(a+u*(n+l*e));m[g++]=f[r]}}}return n.makeTensorInfo([o,d,p,c],a.dtype,m)}};function lF(e){let{inputs:t,backend:n,attrs:r}=e,{x:a,filter:s}=t,{strides:i,pad:o,dilations:l,dimRoundingMode:u}=r;fE([a,s],"depthwiseConv2DNative");let h=va.computeStrides(a.shape),d=va.computeStrides(s.shape),p=l;null==p&&(p=[1,1]),va.assert(Gf.eitherStridesOrDilationsAreOne(i,p),()=>`Error in depthwiseConv2d: Either strides or dilations must be 1. Got strides ${i} and dilations '${p}'`);let c=Gf.computeConv2DInfo(a.shape,s.shape,i,p,o,u,!0),{filterHeight:f,filterWidth:m,dilationHeight:g,dilationWidth:y,padInfo:b}=c,x=b.left,v=b.top,w=c.outChannels/c.inChannels,k=new Ja(c.outShape,a.dtype),I=n.data.get(a.dataId).values,S=n.data.get(s.dataId).values,_=k.values;for(let e=0;e<c.batchSize;++e){let t=e*h[0],n=e*k.strides[0];for(let e=0;e<c.outHeight;++e){let r=n+e*k.strides[1],a=e*c.strideHeight-v;for(let e=0;e<f;++e){let n=a+e*g;if(n<0||n>=c.inHeight)continue;let s=e*d[0],i=t+n*h[1];for(let e=0;e<c.outWidth;++e){let t=r+e*k.strides[2],n=e*c.strideWidth-x;for(let e=0;e<m;++e){let r=n+e*y;if(r<0||r>=c.inWidth)continue;let a=s+e*d[1],o=i+r*c.inChannels,l=t,u=a;for(let e=0;e<c.inChannels;++e){let t=I[o+e];for(let e=0;e<w;++e)_[l+e]+=t*S[u+e];l+=w,u+=w}}}}}}return n.makeTensorInfo(k.shape,k.dtype,k.values)}var uF={kernelName:$t,backendName:"cpu",kernelFunc:lF};var hF={kernelName:Rt,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a,dy:s}=t,{strides:i,dilations:o,pad:l,dimRoundingMode:u,filterShape:h}=r;fE([a,s],"depthwiseConv2dNativeBackpropFilter");let d=Gf.computeConv2DInfo(a.shape,h,i,o,l,u,!0),{strideHeight:p,strideWidth:c,filterHeight:f,filterWidth:m}=d,g=new Ja(d.filterShape,"float32"),y=d.padInfo.left,b=d.padInfo.top,x=d.outChannels/d.inChannels,v=n.data.get(a.dataId).values,w=new Ja(a.shape,a.dtype,v),k=n.data.get(s.dataId).values,I=new Ja(s.shape,s.dtype,k);for(let e=0;e<f;++e){let t=Math.max(0,Math.ceil((b-e)/p)),n=Math.min(d.outHeight,(d.inHeight+b-e)/p);for(let r=0;r<m;++r){let a=Math.max(0,Math.ceil((y-r)/c)),s=Math.min(d.outWidth,(d.inWidth+y-r)/c);for(let i=0;i<d.outChannels;++i){let o=Math.trunc(i/x),l=i%x,u=0;for(let l=0;l<d.batchSize;++l)for(let h=t;h<n;++h){let t=e+h*p-b;for(let e=a;e<s;++e){let n=r+e*c-y;u+=w.get(l,t,n,o)*I.get(l,h,e,i)}}g.set(u,e,r,o,l)}}}return n.makeTensorInfo(g.shape,g.dtype,g.values)}};var dF={kernelName:Ft,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{dy:a,filter:s}=t,{strides:i,dilations:o,pad:l,dimRoundingMode:u,inputShape:h}=r;fE([a,s],"depthwiseConv2DNativeBackpropInput");let d=va.computeStrides(a.shape),p=va.computeStrides(s.shape),c=Gf.computeConv2DInfo(h,s.shape,i,o,l,u,!0),f=new Ja(c.inShape,"float32"),m=f.values,[g,y,b]=f.strides,x=n.data.get(a.dataId).values,[v,w,k]=d,I=n.data.get(s.dataId).values,[S,_,N]=p,{batchSize:T,filterHeight:C,filterWidth:E,inChannels:A,inHeight:$,inWidth:R,outChannels:F,outHeight:D,outWidth:M,strideHeight:O,strideWidth:L}=c,P=C-1-c.padInfo.top,z=E-1-c.padInfo.left,B=F/A;for(let e=0;e<T;++e)for(let t=0;t<A;++t)for(let n=0;n<$;++n){let r=n-P,a=Math.max(0,Math.ceil(r/O)),s=Math.min(D,(C+r)/O);for(let i=0;i<R;++i){let o=i-z,l=Math.max(0,Math.ceil(o/L)),u=Math.min(M,(E+o)/L),h=0;for(let n=a;n<s;++n){let a=n*O-r;for(let r=l;r<u;++r){let s=v*e+w*n+k*r,i=S*(C-1-a)+_*(E-1-(r*L-o))+N*t;for(let e=0;e<B;++e){h+=x[s+(t*B+e)]*I[i+e]}}}m[g*e+y*n+b*i+t]=h}}return n.makeTensorInfo(f.shape,f.dtype,f.values)}};var pF={kernelName:Dt,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n}=e,{x:r}=t,a=va.sizeFromShape(r.shape),s=n.data.get(r.dataId).values,i=uo([a,a],r.dtype),o=i.values;for(let e=0;e<s.length;e++)o[e*a+e]=s[e];let l=[...r.shape,...r.shape];return n.makeTensorInfo(l,i.dtype,i.values)}},cF={kernelName:Mt,backendName:"cpu",kernelFunc:({inputs:e,backend:t,attrs:n})=>{let{x:r,filter:a}=e,{strides:s,pad:i,dilations:o}=n,l=t,u=l.data.get(r.dataId).values,h=r.shape.length,d=l.data.get(a.dataId).values,p=a.shape.length,{batchSize:c,inHeight:f,inWidth:m,inChannels:g,outHeight:y,outWidth:b,padInfo:x,strideHeight:v,strideWidth:w,filterHeight:k,filterWidth:I,dilationHeight:S,dilationWidth:_,outShape:N}=Gf.computeDilation2DInfo(r.shape,a.shape,s,i,"NHWC",o),T=va.sizeFromShape(N),C=N.length,E=va.getArrayFromDType(r.dtype,T);for(let e=0;e<c;++e)for(let t=0;t<y;++t){let n=t*v-x.top;for(let s=0;s<b;++s){let i=s*w-x.left;for(let o=0;o<g;++o){let l=Number.MIN_SAFE_INTEGER;for(let t=0;t<k;++t){let s=n+t*S;if(s>=0&&s<f)for(let n=0;n<I;++n){let c=i+n*_;if(c>=0&&c<m){let i=va.locToIndex([e,s,c,o],h,va.computeStrides(r.shape)),f=va.locToIndex([t,n,o],p,va.computeStrides(a.shape)),m=u[i]+d[f];m>l&&(l=m)}}}E[va.locToIndex([e,t,s,o],C,va.computeStrides(N))]=l}}}return{dataId:l.write(va.toTypedArray(E,r.dtype),N,r.dtype),shape:N,dtype:r.dtype}}},fF={kernelName:Lt,backendName:"cpu",kernelFunc:({inputs:e,backend:t,attrs:n})=>{let{x:r,filter:a,dy:s}=e,{strides:i,pad:o,dilations:l}=n,u=t,h=va.toNestedArray(r.shape,u.data.get(r.dataId).values),d=va.toNestedArray(a.shape,u.data.get(a.dataId).values),{batchSize:p,inHeight:c,inWidth:f,inChannels:m,outHeight:g,outWidth:y,padInfo:b,strideHeight:x,strideWidth:v,filterHeight:w,filterWidth:k,dilationHeight:I,dilationWidth:S,outShape:_}=Gf.computeDilation2DInfo(r.shape,a.shape,i,o,"NHWC",l);va.assert(s.rank===_.length,()=>`Error in ${Lt}, dy must have the same rank as output ${_.length}, but got ${s.rank}`);let N=va.toNestedArray(_,u.data.get(s.dataId).values),T=va.makeZerosNestedTypedArray(a.shape,a.dtype);for(let e=0;e<p;++e)for(let t=0;t<g;++t){let n=t*x-b.top;for(let r=0;r<y;++r){let a=r*v-b.left;for(let s=0;s<m;++s){let i=Number.MIN_SAFE_INTEGER,o=0,l=0;for(let t=0;t<w;++t){let r=n+t*I;if(r>=0&&r<c)for(let n=0;n<k;++n){let u=a+n*S;if(u>=0&&u<f){let a=h[e][r][u][s]+d[t][n][s];a>i&&(i=a,o=t,l=n)}}}T[o][l][s]+=N[e][t][r][s]}}}return{dataId:u.write(va.toTypedArray(T,r.dtype),a.shape,a.dtype),shape:a.shape,dtype:a.dtype}}},mF={kernelName:Ot,backendName:"cpu",kernelFunc:({inputs:e,backend:t,attrs:n})=>{let{x:r,filter:a,dy:s}=e,{strides:i,pad:o,dilations:l}=n,u=t,h=va.toNestedArray(r.shape,u.data.get(r.dataId).values),d=va.toNestedArray(a.shape,u.data.get(a.dataId).values),{batchSize:p,inHeight:c,inWidth:f,inChannels:m,outHeight:g,outWidth:y,padInfo:b,strideHeight:x,strideWidth:v,filterHeight:w,filterWidth:k,dilationHeight:I,dilationWidth:S,outShape:_}=Gf.computeDilation2DInfo(r.shape,a.shape,i,o,"NHWC",l);va.assert(s.rank===_.length,()=>`Error in ${Ot}, dy must have the same rank as output ${_.length}, but got ${s.rank}`);let N=va.toNestedArray(_,u.data.get(s.dataId).values),T=va.makeZerosNestedTypedArray(r.shape,r.dtype);for(let e=0;e<p;++e)for(let t=0;t<g;++t){let n=t*x-b.top;for(let r=0;r<y;++r){let a=r*v-b.left;for(let s=0;s<m;++s){let i=Number.MIN_SAFE_INTEGER,o=n<0?0:n,l=a<0?0:a;for(let t=0;t<w;++t){let r=n+t*I;if(r>=0&&r<c)for(let n=0;n<k;++n){let u=a+n*S;if(u>=0&&u<f){let a=h[e][r][u][s]+d[t][n][s];a>i&&(i=a,o=r,l=u)}}}T[e][o][l][s]+=N[e][t][r][s]}}}return{dataId:u.write(va.toTypedArray(T,r.dtype),r.shape,r.dtype),shape:r.shape,dtype:r.dtype}}};var gF={kernelName:Pt,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{image:a}=t,{canvas:s,options:i}=r,{contextOptions:o,imageOptions:l}=i||{},u=(null==l?void 0:l.alpha)||1,h=(null==o?void 0:o.contextType)||"2d";if("2d"!==h)throw new Error(`Context type ${o.contextType} is not supported by the CPU backend.`);let d=s.getContext(h,(null==o?void 0:o.contextAttributes)||{});if(null==d)throw new Error(`Could not get the context with ${h} type.`);let[p,c]=a.shape.slice(0,2),f=2===a.shape.length?1:a.shape[2],m=n.data.get(a.dataId).values,g="float32"===a.dtype?255:1,y=new Uint8ClampedArray(c*p*4);for(let e=0;e<p*c;++e){let t=[0,0,0,255*u];for(let n=0;n<f;n++){let r=m[e*f+n];if("float32"===a.dtype){if(r<0||r>1)throw new Error(`Tensor values for a float32 Tensor must be in the range [0 - 1] but encountered ${r}.`)}else if("int32"===a.dtype&&(r<0||r>255))throw new Error(`Tensor values for a int32 Tensor must be in the range [0 - 255] but encountered ${r}.`);1===f?(t[0]=r*g,t[1]=r*g,t[2]=r*g):t[n]=r*g}let n=4*e;y[n+0]=Math.round(t[0]),y[n+1]=Math.round(t[1]),y[n+2]=Math.round(t[2]),y[n+3]=Math.round(t[3])}s.width=c,s.height=p;let b=new ImageData(y,c,p);return d.putImageData(b,0,0),a}};function yF(e){let t,{inputs:n,backend:r,attrs:a}=e,{x:s}=n,{axis:i,keepDims:o}=a;fE(s,"sum"),t="bool"===s.dtype?EE({inputs:{x:s},backend:r,attrs:{dtype:"int32"}}):SE({inputs:{x:s},backend:r});let l=t.shape.length,u=va.parseAxisParam(i,t.shape),h=Gf.getAxesPermutation(u,l),d=u,p=t;null!=h&&(p=HA({inputs:{x:t},backend:r,attrs:{perm:h}}),d=Gf.getInnerMostAxes(d.length,l)),Gf.assertAxesAreInnerMostDims("sum",d,p.shape.length);let[c,f]=Gf.computeOutAndReduceShapes(p.shape,d),m=IE(r,c,Gf.upcastType(p.dtype,"int32")),g=va.sizeFromShape(f),y=r.data.get(m.dataId).values,b=r.data.get(p.dataId).values;for(let e=0;e<y.length;++e){let t=e*g,n=0;for(let e=0;e<g;++e)n+=b[t+e];y[e]=n}if(o){let e=m;m=rR({inputs:{x:m},backend:r,attrs:{shape:Gf.expandShapeToKeepDim(m.shape,u)}}),r.disposeIntermediateTensorInfo(e)}return r.disposeIntermediateTensorInfo(t),null!=h&&r.disposeIntermediateTensorInfo(p),m}var bF={kernelName:Tr,backendName:"cpu",kernelFunc:yF};var xF={kernelName:Bt,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{equation:a}=r,s=t,{allDims:i,summedDims:o,idDims:l}=Gf.decodeEinsumEquation(a,s.length);Gf.checkEinsumDimSizes(i.length,l,s);let{path:u,steps:h}=Gf.getEinsumComputePath(o,l),d=h.length,p=null,c=i.length,f=[];for(let e=0;e<d;++e){for(let t of h[e]){let e,{permutationIndices:r,expandDims:a}=Gf.getEinsumPermutation(c,l[t]);Gf.isIdentityPermutation(r)?e=s[t]:(e=HA({inputs:{x:s[t]},backend:n,attrs:{perm:r}}),f.push(e));let i=e.shape.slice();for(let e=0;e<a.length;++e)i.splice(a[e],0,1);va.arraysEqual(e.shape,i)||(e=rR({inputs:{x:e},backend:n,attrs:{shape:i}}),f.push(e)),null===p?p=e:(p=LA({inputs:{a:e,b:p},backend:n}),f.push(p))}e<d-1&&(u[e]>=0&&(p=yF({inputs:{x:p},backend:n,attrs:{axis:u[e]-(i.length-c),keepDims:!1}}),f.push(p)),c--)}for(let e of f)e!==p&&n.disposeIntermediateTensorInfo(e);return p}};var vF={kernelName:Vt,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n}=e,{dy:r,y:a}=t;fE([r,a],"eluGrad");let s=new Float32Array(va.sizeFromShape(a.shape)),i=n.data.get(a.dataId).values,o=n.data.get(r.dataId).values;for(let e=0;e<i.length;++e){let t=i[e];s[e]=t>=0?o[e]:o[e]*(t+1)}return n.makeTensorInfo(a.shape,"float32",s)}},wF=Gf.ERF_P,kF=Gf.ERF_A1,IF=Gf.ERF_A2,SF=Gf.ERF_A3,_F=Gf.ERF_A4,NF=Gf.ERF_A5,TF=UE(Ut,e=>{let t=Math.sign(e),n=Math.abs(e),r=1/(1+wF*n);return t*(1-((((NF*r+_F)*r+SF)*r+IF)*r+kF)*r*Math.exp(-n*n))}),CF={kernelName:Ut,backendName:"cpu",kernelFunc:TF};function EF(e){let{inputs:t,backend:n,attrs:r}=e,{input:a}=t,{dim:s}=r,i=a.shape.length,o=a.shape.slice(),l=s;return s<0&&(va.assert(-(i+1)<=s,()=>`Axis must be in the interval [${-(i+1)}, ${i}]`),l=i+s+1),o.splice(l,0,1),rR({inputs:{x:a},backend:n,attrs:{shape:o}})}var AF={kernelName:jt,backendName:"cpu",kernelFunc:EF},$F=vE((e,t)=>e/t),RF=$E(zt,$F),FF={kernelName:zt,backendName:"cpu",kernelFunc:RF};function DF(e,t,n){let r=e.shape,a=r[0],s=r[1],i=n.data.get(e.dataId),o=i.complexTensorInfos.real,l=i.complexTensorInfos.imag,u=[a,s],h=va.sizeFromShape(u),d=va.getTypedArrayFromDType("float32",h),p=va.getTypedArrayFromDType("float32",h);for(let e=0;e<a;e++){let r=m$({inputs:{x:o},backend:n,attrs:{begin:[e,0],size:[1,s]}}),a=m$({inputs:{x:l},backend:n,attrs:{begin:[e,0],size:[1,s]}}),i=wE({inputs:{real:r,imag:a},backend:n}),{real:u,imag:h}=MF(i,t,n),c=Gf.mergeRealAndImagArrays(u,h);for(let t=0;t<s;t++){let n=Gf.getComplexWithIndex(c,t);d[e*s+t]=n.real,p[e*s+t]=n.imag}n.disposeIntermediateTensorInfo(r),n.disposeIntermediateTensorInfo(a),n.disposeIntermediateTensorInfo(i)}let c=n.makeTensorInfo(u,"float32",d),f=n.makeTensorInfo(u,"float32",p),m=wE({inputs:{real:c,imag:f},backend:n});return n.disposeIntermediateTensorInfo(c),n.disposeIntermediateTensorInfo(f),m}function MF(e,t,n){let r=va.sizeFromShape(e.shape),a=n.data.get(e.dataId),s=n.data.get(a.complexTensorInfos.real.dataId).values,i=n.data.get(a.complexTensorInfos.imag.dataId).values;if(function(e){return!(e&e-1)}(r)){let a=OF(s,i,r,t,n),o=[e.shape[0],e.shape[1]];if(t){let e=n.makeTensorInfo(o,"float32",a.real),t=n.makeTensorInfo(o,"float32",a.imag),s=n.makeTensorInfo([],"float32",va.createScalarValue(r,"float32")),i=SE({inputs:{x:s},backend:n}),l=FF.kernelFunc({inputs:{a:e,b:s},backend:n}),u=FF.kernelFunc({inputs:{a:t,b:i},backend:n}),h=n.data.get(l.dataId).values,d=n.data.get(u.dataId).values;return n.disposeIntermediateTensorInfo(e),n.disposeIntermediateTensorInfo(t),n.disposeIntermediateTensorInfo(s),n.disposeIntermediateTensorInfo(i),n.disposeIntermediateTensorInfo(l),n.disposeIntermediateTensorInfo(u),{real:h,imag:d}}return a}{let e=function(e,t,n){let r=new Float32Array(2*t);for(let a=0;a<t;a++){let s=0,i=0;for(let r=0;r<t;r++){let o=Gf.exponent(a*r,t,n),l=Gf.getComplexWithIndex(e,r);s+=l.real*o.real-l.imag*o.imag,i+=l.real*o.imag+l.imag*o.real}n&&(s/=t,i/=t),Gf.assignToTypedArray(r,s,i,a)}return r}(Gf.mergeRealAndImagArrays(s,i),r,t);return Gf.splitRealAndImagArrays(e)}}function OF(e,t,n,r,a){if(1===n)return{real:e,imag:t};let s=Gf.mergeRealAndImagArrays(e,t),i=n/2,o=Gf.complexWithEvenIndex(s),l=o.real,u=o.imag,h=[l.length],d=a.makeTensorInfo(h,"float32",l),p=a.makeTensorInfo(h,"float32",u),c=wE({inputs:{real:d,imag:p},backend:a}),f=Gf.complexWithOddIndex(s),m=f.real,g=f.imag,y=[m.length],b=a.makeTensorInfo(y,"float32",m),x=a.makeTensorInfo(y,"float32",g),v=wE({inputs:{real:b,imag:x},backend:a}),w=OF(l,u,i,r,a),k=w.real,I=w.imag,S=[k.length],_=a.makeTensorInfo(S,"float32",k),N=a.makeTensorInfo(S,"float32",I),T=wE({inputs:{real:_,imag:N},backend:a}),C=OF(m,g,i,r,a),E=C.real,A=C.imag,$=[E.length],R=a.makeTensorInfo($,"float32",E),F=a.makeTensorInfo($,"float32",A),D=wE({inputs:{real:R,imag:F},backend:a}),M=Gf.exponents(n,r),O=[M.real.length],L=a.makeTensorInfo(O,"float32",M.real),P=a.makeTensorInfo(O,"float32",M.imag),z=wE({inputs:{real:L,imag:P},backend:a}),B=LA({inputs:{a:z,b:D},backend:a}),W=ME({inputs:{a:T,b:B},backend:a}),V=L$({inputs:{a:T,b:B},backend:a}),U=NE({inputs:{input:W},backend:a}),G=NE({inputs:{input:V},backend:a}),H=VR({inputs:{input:W},backend:a}),j=VR({inputs:{input:V},backend:a}),q=GR({inputs:[U,G],backend:a,attrs:{axis:0}}),K=GR({inputs:[H,j],backend:a,attrs:{axis:0}}),X=a.data.get(q.dataId).values,Z=a.data.get(K.dataId).values;return a.disposeIntermediateTensorInfo(d),a.disposeIntermediateTensorInfo(p),a.disposeIntermediateTensorInfo(c),a.disposeIntermediateTensorInfo(b),a.disposeIntermediateTensorInfo(x),a.disposeIntermediateTensorInfo(v),a.disposeIntermediateTensorInfo(_),a.disposeIntermediateTensorInfo(N),a.disposeIntermediateTensorInfo(T),a.disposeIntermediateTensorInfo(R),a.disposeIntermediateTensorInfo(F),a.disposeIntermediateTensorInfo(D),a.disposeIntermediateTensorInfo(L),a.disposeIntermediateTensorInfo(P),a.disposeIntermediateTensorInfo(z),a.disposeIntermediateTensorInfo(B),a.disposeIntermediateTensorInfo(W),a.disposeIntermediateTensorInfo(V),a.disposeIntermediateTensorInfo(U),a.disposeIntermediateTensorInfo(H),a.disposeIntermediateTensorInfo(G),a.disposeIntermediateTensorInfo(j),a.disposeIntermediateTensorInfo(q),a.disposeIntermediateTensorInfo(K),{real:X,imag:Z}}var LF={kernelName:Kt,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n}=e,{input:r}=t,a=va.sizeFromShape(r.shape),s=r.shape[r.shape.length-1],i=rR({inputs:{x:r},backend:n,attrs:{shape:[a/s,s]}}),o=DF(i,!1,n),l=rR({inputs:{x:o},backend:n,attrs:{shape:r.shape}});return n.disposeIntermediateTensorInfo(i),n.disposeIntermediateTensorInfo(o),l}};function PF(e){let{backend:t,attrs:n}=e,{shape:r,value:a,dtype:s}=n,i=s||va.inferDtype(a),o=va.getArrayFromDType(i,va.sizeFromShape(r));return function(e,t){e.fill(t)}(o,a),t.makeTensorInfo(r,i,o)}var zF={kernelName:Xt,backendName:"cpu",kernelFunc:PF};var BF={kernelName:Zt,backendName:"cpu",kernelFunc:({inputs:e,attrs:t,backend:n})=>{let{image:r}=e,a=n,s=va.getTypedArrayFromDType(r.dtype,va.sizeFromShape(r.shape)),[i,o,l,u]=r.shape,h=a.data.get(r.dataId).values;for(let e=0;e<i;e++){let t=e*l*o*u;for(let e=0;e<o;e++){let n=e*(l*u);for(let e=0;e<l;e++){let r=e*u;for(let a=0;a<u;a++){let i=Math.round(l-e-1),o=t+n+r+a,d=h[o];if(i>=0&&i<l){d=h[t+n+i*u+a]}s[o]=d}}}}return{dataId:a.write(s,r.shape,r.dtype),shape:r.shape,dtype:r.dtype}}};var WF={kernelName:sa,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a,filter:s,bias:i,preluActivationWeights:o}=t,{strides:l,pad:u,dataFormat:h,dilations:d,dimRoundingMode:p,activation:c,leakyreluAlpha:f}=r,m=jR({inputs:{x:a,filter:s},backend:n,attrs:{strides:l,pad:u,dataFormat:h,dilations:d,dimRoundingMode:p}});if(i){let e=m;if("NCHW"===h&&1===i.shape.length&&1!==i.shape[0]){let e=rR({inputs:{x:i},backend:n,attrs:{shape:[i.shape[0],1,1]}});m=ME({inputs:{a:m,b:e},backend:n}),n.disposeIntermediateTensorInfo(e)}else m=ME({inputs:{a:m,b:i},backend:n});n.disposeIntermediateTensorInfo(e)}if(c){let e=m;if("NCHW"===h&&"prelu"===c&&1===o.shape.length&&1!==o.shape[0]){let e=rR({inputs:{x:o},backend:n,attrs:{shape:[o.shape[0],1,1]}});m=nR(n,m,c,e,f),n.disposeIntermediateTensorInfo(e)}else m=nR(n,m,c,o,f);n.disposeIntermediateTensorInfo(e)}return m}};var VF={kernelName:ia,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a,filter:s,bias:i,preluActivationWeights:o}=t,{strides:l,pad:u,dataFormat:h,dilations:d,dimRoundingMode:p,activation:c,leakyreluAlpha:f}=r,m=lF({inputs:{x:a,filter:s},backend:n,attrs:{strides:l,pad:u,dataFormat:h,dilations:d,dimRoundingMode:p}});if(i){let e=m;m=ME({inputs:{a:m,b:i},backend:n}),n.disposeIntermediateTensorInfo(e)}if(c){let e=m;m=nR(n,m,c,o,f),n.disposeIntermediateTensorInfo(e)}return m}};var UF={kernelName:tn,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n}=e,{params:r,indices:a}=t,s=va.sizeFromShape(r.shape),i=a.shape,o=i[i.length-1],[l,u,h,d]=Gf.prepareAndValidate(r,a);if(0===u)return n.makeTensorInfo(l,r.dtype,[]);let p=hA(n.data.get(a.dataId).values,n.bufferSync(r),r.dtype,u,o,h,d,r.shape,s);return n.makeTensorInfo(l,r.dtype,p.values)}};var GF={kernelName:en,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a,indices:s}=t,{axis:i,batchDims:o}=r;fE([a,s],"gatherV2");let l=va.parseAxisParam(i,a.shape)[0],u=n.data.get(s.dataId).values,h=a.shape[l];for(let e=0;e<u.length;++e){let t=u[e];va.assert(t<=h-1&&t>=0,()=>`GatherV2: the index value ${t} is not in [0, ${h-1}]`)}let d=o;null==o&&(d=0);let p=va.sizeFromShape(s.shape),c=Gf.segment_util.collectGatherOpShapeInfo(a,s,l,d),f=rR({inputs:{x:a},backend:n,attrs:{shape:[c.batchSize,c.outerSize,c.dimSize,c.sliceSize]}}),m=rR({inputs:{x:s},backend:n,attrs:{shape:[c.batchSize,p/c.batchSize]}}),g=[c.batchSize,c.outerSize,p/c.batchSize,c.sliceSize],y=n.bufferSync(m),b=dA(n.bufferSync(f),y,g);return n.disposeIntermediateTensorInfo(f),n.disposeIntermediateTensorInfo(m),n.makeTensorInfo(c.outputShape,b.dtype,b.values)}};var HF={kernelName:sn,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n}=e,{input:r}=t,a=va.sizeFromShape(r.shape),s=r.shape[r.shape.length-1],i=rR({inputs:{x:r},backend:n,attrs:{shape:[a/s,s]}}),o=DF(i,!0,n),l=rR({inputs:{x:o},backend:n,attrs:{shape:r.shape}});return n.disposeIntermediateTensorInfo(i),n.disposeIntermediateTensorInfo(o),l}},jF=UE(ln,e=>Number.isFinite(e)?1:0,"bool"),qF={kernelName:ln,backendName:"cpu",kernelFunc:jF},KF=UE(un,e=>Math.abs(e)===1/0?1:0,"bool"),XF={kernelName:un,backendName:"cpu",kernelFunc:KF},ZF=UE(hn,e=>Number.isNaN(e)?1:0,"bool"),YF={kernelName:hn,backendName:"cpu",kernelFunc:ZF};var JF={kernelName:fn,backendName:"cpu",kernelFunc:function(e){let{backend:t,attrs:n}=e,{start:r,stop:a,num:s}=n,i=SA(r,a,s);return t.makeTensorInfo([i.length],"float32",i)}},QF=UE(gn,e=>Math.log1p(e)),eD={kernelName:gn,backendName:"cpu",kernelFunc:QF},tD=vE((e,t)=>e&&t),nD=$E(yn,tD,null,"bool"),rD={kernelName:yn,backendName:"cpu",kernelFunc:nD},aD=UE(bn,e=>e?0:1,"bool"),sD={kernelName:bn,backendName:"cpu",kernelFunc:aD},iD=vE((e,t)=>e||t),oD=$E(xn,iD,null,"bool"),lD={kernelName:xn,backendName:"cpu",kernelFunc:oD};var uD={kernelName:In,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{depthRadius:s,bias:i,alpha:o,beta:l}=r;fE(a,"LRN");let u=a.shape[3],h=u-1,d=n.data.get(a.dataId).values,p=va.sizeFromShape(a.shape),c=new Float32Array(p);function f(e){let t=e%u,n=e-t+Math.max(0,t-s),r=e-t+Math.min(t+s,h),a=0;for(;n<=r;n++){let e=d[n];a+=e*e}return a}for(let e=0;e<p;e++){let t=f(e),n=d[e]*Math.pow(i+o*t,-l);c[e]=n}return n.makeTensorInfo(a.shape,a.dtype,c)}};var hD={kernelName:Sn,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a,y:s,dy:i}=t,{depthRadius:o,bias:l,alpha:u,beta:h}=r;fE(i,"LRNGrad");let d=va.sizeFromShape(i.shape),p=i.shape[3],c=n.data.get(i.dataId).values,f=n.data.get(a.dataId).values,m=n.data.get(s.dataId).values,g=new Float32Array(d),y=d;for(let e=0;e<y;e++){let t=e%p,n=e-t+Math.max(0,t-o),r=e-t+Math.min(p,t+o+1),a=0;for(let e=n;e<r;e++)a+=Math.pow(f[e],2);a=u*a+l;for(let t=n;t<r;t++){let n=-2*u*h*f[t]*m[e]/a;e===t&&(n+=Math.pow(a,-h)),n*=c[e],g[t]+=n}}return n.makeTensorInfo(i.shape,a.dtype,g)}};function dD(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{reductionIndices:s,keepDims:i}=r,o=n,l=a.shape,u=l.length,h=va.parseAxisParam(s,l),d=h,p=Gf.getAxesPermutation(d,u),c=o.data.get(a.dataId).values;if(null!=p){let e=new Array(u);for(let t=0;t<e.length;t++)e[t]=l[p[t]];c=GA(c,l,a.dtype,p,e),d=Gf.getInnerMostAxes(d.length,u),l=e}fE(a,"max"),Gf.assertAxesAreInnerMostDims("max",d,u);let[f,m]=Gf.computeOutAndReduceShapes(l,d),g=CA(c,va.sizeFromShape(m),f,a.dtype),y=o.write(g,f,a.dtype),b=f;return i&&(b=Gf.expandShapeToKeepDim(f,h)),{dataId:y,shape:b,dtype:a.dtype}}var pD={kernelName:Nn,backendName:"cpu",kernelFunc:dD};var cD={kernelName:Cn,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t;fE(a,"maxPool");let{filterSize:s,strides:i,pad:o,dimRoundingMode:l}=r;va.assert(Gf.eitherStridesOrDilationsAreOne(i,1),()=>`Error in maxPool: Either strides or dilations must be 1. Got strides ${i} and dilations '1'`);let u,h=Gf.computePool2DInfo(a.shape,s,i,1,o,l);if(1===h.filterWidth&&1===h.filterHeight&&va.arraysEqual(h.inShape,h.outShape))u=SE({inputs:{x:a},backend:n});else{let e=n.data.get(a.dataId).values,t=va.computeStrides(a.shape),r=CR(e,a.shape,a.dtype,t,h,"max");u=n.makeTensorInfo(h.outShape,a.dtype,r.values)}return u}};var fD={kernelName:An,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{filterSize:s,strides:i,pad:o,dimRoundingMode:l,dataFormat:u}=r;fE(a,"maxPool3d");let h=Gf.computePool3DInfo(a.shape,s,i,1,o,l,u),d=AR(n.data.get(a.dataId).values,a.shape,a.dtype,va.computeStrides(a.shape),h,"max");return n.makeTensorInfo(d.shape,"float32",d.values)}};var mD={kernelName:$n,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{dy:a,input:s}=t,{filterSize:i,strides:o,pad:l,dimRoundingMode:u}=r;fE([a,s],"maxPool3DGrad");let h=Gf.computePool3DInfo(s.shape,i,o,1,l,u),d=function(e,t){let n=uo(t.outShape,"int32"),r=t.strideDepth,a=t.strideHeight,s=t.strideWidth,i=t.dilationDepth,o=t.dilationHeight,l=t.dilationWidth,u=t.effectiveFilterDepth,h=t.effectiveFilterHeight,d=t.effectiveFilterWidth,p=t.padInfo.front,c=t.padInfo.top,f=t.padInfo.left;for(let m=0;m<t.batchSize;++m)for(let g=0;g<t.inChannels;++g)for(let y=0;y<t.outDepth;++y){let b=y*r-p,x=b;for(;x<0;)x+=i;let v=Math.min(t.inDepth,u+b);for(let r=0;r<t.outHeight;++r){let u=r*a-c,p=u;for(;p<0;)p+=o;let w=Math.min(t.inHeight,h+u);for(let a=0;a<t.outWidth;++a){let c=a*s-f,k=c;for(;k<0;)k+=l;let I=Math.min(t.inWidth,d+c),S=Number.NEGATIVE_INFINITY,_=-1;for(let t=x;t<v;t+=i){let n=t-b;for(let r=p;r<w;r+=o){let a=r-u;for(let s=k;s<I;s+=l){let i=s-c,o=e.get(m,t,r,s,g);o>=S&&(S=o,_=n*h*d+a*h+i)}}}n.set(_,m,y,r,a,g)}}}return n}(n.bufferSync(s),h),p=h.strideDepth,c=h.strideHeight,f=h.strideWidth,m=h.dilationDepth,g=h.dilationHeight,y=h.dilationWidth,b=h.effectiveFilterDepth,x=h.effectiveFilterHeight,v=h.effectiveFilterWidth,w=b-1-h.padInfo.front,k=v-1-h.padInfo.left,I=x-1-h.padInfo.top,S=uo(s.shape,"float32"),_=n.bufferSync(a);for(let e=0;e<h.batchSize;++e)for(let t=0;t<h.inChannels;++t)for(let n=0;n<h.inDepth;++n)for(let r=0;r<h.inHeight;++r)for(let a=0;a<h.inWidth;++a){let s=n-w,i=r-I,o=a-k,l=0;for(let n=0;n<b;n+=m){let r=(s+n)/p;if(!(r<0||r>=h.outDepth||Math.floor(r)!==r))for(let a=0;a<x;a+=g){let s=(i+a)/c;if(!(s<0||s>=h.outHeight||Math.floor(s)!==s))for(let i=0;i<v;i+=y){let u=(o+i)/f;if(u<0||u>=h.outWidth||Math.floor(u)!==u)continue;let p=b*x*v-1-d.get(e,r,s,u,t)===n*x*v+a*v+i?1:0;0!==p&&(l+=_.get(e,r,s,u,t)*p)}}}S.set(l,e,n,r,a,t)}return n.makeTensorInfo(S.shape,S.dtype,S.values)}};var gD={kernelName:En,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{dy:a,input:s,output:i}=t,o=s;fE([s,i],"maxPoolGrad");let{filterSize:l,strides:u,pad:h,dimRoundingMode:d}=r,p=Gf.computePool2DInfo(o.shape,l,u,1,h,d),c=n.data.get(o.dataId).values,f=uo(p.outShape,o.dtype,ER(c,o.shape,o.dtype,p).values),m=p.strideHeight,g=p.strideWidth,y=p.dilationHeight,b=p.dilationWidth,x=p.effectiveFilterHeight,v=p.effectiveFilterWidth,w=v-1-p.padInfo.left,k=x-1-p.padInfo.top,I=uo(o.shape,"float32"),S=n.data.get(a.dataId).values,_=uo(a.shape,"float32",S);for(let e=0;e<p.batchSize;++e)for(let t=0;t<p.inChannels;++t)for(let n=0;n<p.inHeight;++n)for(let r=0;r<p.inWidth;++r){let a=n-k,s=r-w,i=0;for(let n=0;n<x;n+=y){let r=(a+n)/m;if(!(r<0||r>=p.outHeight||Math.floor(r)!==r))for(let a=0;a<v;a+=b){let o=(s+a)/g;if(o<0||o>=p.outWidth||Math.floor(o)!==o)continue;let l=x*v-1-f.get(e,r,o,t)===n*v+a?1:0;0!==l&&(i+=_.get(e,r,o,t)*l)}}I.set(i,e,n,r,t)}return n.makeTensorInfo(I.shape,I.dtype,I.values)}};var yD={kernelName:Rn,backendName:"cpu",kernelFunc:({inputs:e,attrs:t,backend:n})=>{let{x:r}=e,{filterSize:a,strides:s,pad:i,includeBatchInIndex:o}=t,l=n;fE(r,"MaxPoolWithArgmax");let u=l.data.get(r.dataId).values,h=Gf.computePool2DInfo(r.shape,a,s,[1,1],i),[d,p]=function(e,t,n,r,a){let s=CR(e,0,n,va.computeStrides(t),a,"max"),i=ER(e,t,n,a,!0,r);return[s.values,i.values]}(u,r.shape,r.dtype,o,h),c=l.write(d,h.outShape,r.dtype),f=l.write(p,h.outShape,r.dtype);return[{dataId:c,shape:h.outShape,dtype:r.dtype},{dataId:f,shape:h.outShape,dtype:"int32"}]}};var bD={kernelName:Fn,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{axis:s,keepDims:i}=r,o=va.parseAxisParam(s,a.shape),l=Gf.computeOutAndReduceShapes(a.shape,o)[1],u=va.sizeFromShape(l),h=[],d=n.makeTensorInfo([],"float32",new Float32Array([u]));h.push(d);let p=EE({inputs:{x:a},backend:n,attrs:{dtype:"float32"}});h.push(p);let c=RF({inputs:{a:p,b:d},backend:n});h.push(c);let f=yF({inputs:{x:c},backend:n,attrs:{axis:s,keepDims:i}});return h.forEach(e=>n.disposeIntermediateTensorInfo(e)),f}};var xD={kernelName:Dn,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{axis:s,keepDims:i}=r;fE(a,"min");let o=va.parseAxisParam(s,a.shape),l=o,u=Gf.getAxesPermutation(l,a.shape.length),h=a;null!=u&&(h=HA({inputs:{x:a},backend:n,attrs:{perm:u}}),l=Gf.getInnerMostAxes(l.length,a.shape.length)),Gf.assertAxesAreInnerMostDims("min",l,h.shape.length);let[d,p]=Gf.computeOutAndReduceShapes(h.shape,l),c=va.sizeFromShape(p),f=va.makeZerosTypedArray(va.sizeFromShape(d),h.dtype),m=n.data.get(h.dataId).values;for(let e=0;e<f.length;++e){let t=e*c,n=m[t];for(let e=0;e<c;++e){let r=m[t+e];(Number.isNaN(r)||r<n)&&(n=r)}f[e]=n}null!=u&&n.disposeIntermediateTensorInfo(h);let g=n.makeTensorInfo(d,h.dtype,f);if(i){let e=rR({inputs:{x:g},backend:n,attrs:{shape:Gf.expandShapeToKeepDim(d,o)}});return n.disposeIntermediateTensorInfo(g),e}return g}};var vD={kernelName:On,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{paddings:s,mode:i}=r;fE(a,"mirrorPad");let o=s.map((e,t)=>e[0]+a.shape[t]+e[1]),l=s.map(e=>e[0]),u=s.map((e,t)=>e[0]+a.shape[t]),h="reflect"===i?0:1,d=n.data.get(a.dataId).values,p=a.shape.length,c=va.computeStrides(a.shape),f=va.sizeFromShape(o),m=o.length,g=va.computeStrides(o),y=va.getTypedArrayFromDType(a.dtype,f);for(let e=0;e<f;e++){let t=va.indexToLoc(e,m,g);for(let e=0;e<m;e++)t[e]<l[e]?t[e]=2*l[e]-t[e]-h:t[e]>=u[e]&&(t[e]=2*(u[e]-1)-t[e]+h);t=t.map((e,t)=>e-l[t]);let n=va.locToIndex(t,p,c);y[e]=d[n]}return{dataId:n.write(y,o,a.dtype),shape:o,dtype:a.dtype}}},wD=vE((e,t)=>{let n=e%t;return e<0&&t<0||e>=0&&t>=0?n:(n+t)%t}),kD=$E(Ln,wD),ID={kernelName:Ln,backendName:"cpu",kernelFunc:kD},SD=m(T());function _D(e){let{inputs:t,backend:n,attrs:r}=e,{logits:a}=t,{dim:s}=r,i=a.shape.length,o=s;if(-1===o&&(o=i-1),o!==i-1)throw Error(`Softmax along a non-last dimension is not yet supported. Logits was rank ${i} and dim was ${o}`);let l=va.parseAxisParam([o],a.shape),u=dD({inputs:{x:a},backend:n,attrs:{reductionIndices:l,keepDims:!1}}),h=Gf.expandShapeToKeepDim(u.shape,l),d=rR({inputs:{x:u},backend:n,attrs:{shape:h}}),p=L$({inputs:{a,b:d},backend:n}),c=QE({inputs:{x:p},backend:n}),f=yF({inputs:{x:c},backend:n,attrs:{axis:l,keepDims:!1}}),m=rR({inputs:{x:f},backend:n,attrs:{shape:h}}),g=RF({inputs:{a:c,b:m},backend:n});return n.disposeIntermediateTensorInfo(u),n.disposeIntermediateTensorInfo(d),n.disposeIntermediateTensorInfo(p),n.disposeIntermediateTensorInfo(c),n.disposeIntermediateTensorInfo(f),n.disposeIntermediateTensorInfo(m),g}var ND={kernelName:Ar,backendName:"cpu",kernelFunc:_D};var TD={kernelName:Pn,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{logits:a}=t,{numSamples:s,seed:i,normalized:o}=r;fE(a,"multinomial");let l=o?a:_D({inputs:{logits:a},backend:n,attrs:{dim:-1}}),u=l.shape[0],h=l.shape[1],d=n.data.get(l.dataId).values,p=[u,s],c=va.makeZerosTypedArray(va.sizeFromShape(p),"int32");for(let e=0;e<u;++e){let t=e*h,n=new Float32Array(h-1);n[0]=d[t];for(let e=1;e<n.length;++e)n[e]=n[e-1]+d[t+e];let r=SD.alea(i.toString()),a=e*s;for(let e=0;e<s;++e){let t=r();c[a+e]=n.length;for(let r=0;r<n.length;r++)if(t<n[r]){c[a+e]=r;break}}}return o||n.disposeIntermediateTensorInfo(l),n.makeTensorInfo(p,"int32",c)}},CD=Km.nonMaxSuppressionV3Impl;var ED={kernelName:Vn,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{boxes:a,scores:s}=t,{maxOutputSize:i,iouThreshold:o,scoreThreshold:l}=r;fE(a,"NonMaxSuppression");let u=n.data.get(a.dataId).values,h=n.data.get(s.dataId).values,{selectedIndices:d}=CD(u,h,i,o,l);return n.makeTensorInfo([d.length],"int32",new Int32Array(d))}},AD=Km.nonMaxSuppressionV4Impl;var $D={kernelName:Un,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{boxes:a,scores:s}=t,{maxOutputSize:i,iouThreshold:o,scoreThreshold:l,padToMaxOutputSize:u}=r;fE(a,"NonMaxSuppressionPadded");let h=n.data.get(a.dataId).values,d=n.data.get(s.dataId).values,{selectedIndices:p,validOutputs:c}=AD(h,d,i,o,l,u);return[n.makeTensorInfo([p.length],"int32",new Int32Array(p)),n.makeTensorInfo([],"int32",new Int32Array([c]))]}},RD=Km.nonMaxSuppressionV5Impl;var FD={kernelName:Gn,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{boxes:a,scores:s}=t,{maxOutputSize:i,iouThreshold:o,scoreThreshold:l,softNmsSigma:u}=r;fE(a,"NonMaxSuppressionWithScore");let h=n.data.get(a.dataId).values,d=n.data.get(s.dataId).values,p=i,c=o,f=l,m=u,{selectedIndices:g,selectedScores:y}=RD(h,d,p,c,f,m);return[n.makeTensorInfo([g.length],"int32",new Int32Array(g)),n.makeTensorInfo([y.length],"float32",new Float32Array(y))]}};var DD={kernelName:jn,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{indices:a}=t,{dtype:s,depth:i,onValue:o,offValue:l}=r;fE(a,"oneHot");let u=va.sizeFromShape(a.shape),h=new Float32Array(u*i);h.fill(l);let d=n.data.get(a.dataId).values;for(let e=0;e<u;++e)d[e]>=0&&d[e]<i&&(h[e*i+d[e]]=o);return n.makeTensorInfo([...a.shape,i],s,h)}};function MD(e){let{inputs:t,backend:n}=e,{x:r}=t;if("string"===r.dtype)throw new Error("zerosLike is not supported for string tensors");if("complex64"===r.dtype){let e=NE({inputs:{input:r},backend:n}),t=MD({inputs:{x:e},backend:n}),a=VR({inputs:{input:r},backend:n}),s=MD({inputs:{x:a},backend:n}),i=wE({inputs:{real:t,imag:s},backend:n});return n.disposeIntermediateTensorInfo(e),n.disposeIntermediateTensorInfo(t),n.disposeIntermediateTensorInfo(a),n.disposeIntermediateTensorInfo(s),i}return PF({backend:n,attrs:{shape:r.shape,value:0,dtype:r.dtype}})}var OD={kernelName:ea,backendName:"cpu",kernelFunc:MD};var LD={kernelName:Hn,backendName:"cpu",kernelFunc:function e(t){let{inputs:n,backend:r}=t,{x:a}=n;if("string"===a.dtype)throw new Error("onesLike is not supported for string tensors");if("complex64"===a.dtype){let t=NE({inputs:{input:a},backend:r}),n=e({inputs:{x:t},backend:r}),s=VR({inputs:{input:a},backend:r}),i=MD({inputs:{x:s},backend:r}),o=wE({inputs:{real:n,imag:i},backend:r});return r.disposeIntermediateTensorInfo(t),r.disposeIntermediateTensorInfo(n),r.disposeIntermediateTensorInfo(s),r.disposeIntermediateTensorInfo(i),o}return PF({backend:r,attrs:{shape:a.shape,value:1,dtype:a.dtype}})}};function PD(e){let{inputs:t,backend:n,attrs:r}=e,{axis:a}=r;if(1===t.length)return EF({inputs:{input:t[0]},backend:n,attrs:{dim:a}});let s=t[0].shape,i=t[0].dtype;t.forEach(e=>{va.assertShapesMatch(s,e.shape,"All tensors passed to stack must have matching shapes"),va.assert(i===e.dtype,()=>"All tensors passed to stack must have matching dtypes")});let o=[],l=GR({inputs:t.map(e=>{let t=EF({inputs:{input:e},backend:n,attrs:{dim:a}});return o.push(t),t}),backend:n,attrs:{axis:a}});return o.forEach(e=>n.disposeIntermediateTensorInfo(e)),l}var zD={kernelName:qn,backendName:"cpu",kernelFunc:PD};var BD={kernelName:Kn,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{paddings:s,constantValue:i}=r;fE(a,"pad");let o=s.map((e,t)=>e[0]+a.shape[t]+e[1]),l=s.map(e=>e[0]),u=n.data.get(a.dataId).values,h=va.sizeFromShape(a.shape),d=a.shape.length,p=va.computeStrides(a.shape),c=va.sizeFromShape(o),f=o.length,m=va.computeStrides(o),g=va.getTypedArrayFromDType(a.dtype,c);0!==i&&g.fill(i);for(let e=0;e<h;e++){let t=va.indexToLoc(e,d,p).map((e,t)=>e+l[t]);g[va.locToIndex(t,f,m)]=u[e]}return{dataId:n.write(g,o,a.dtype),shape:o,dtype:a.dtype}}},WD=vE((e,t)=>Math.pow(e,t)),VD=$E(Zn,WD),UD={kernelName:Zn,backendName:"cpu",kernelFunc:VD};var GD={kernelName:Qn,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{paramsNestedSplits:a,paramsDenseValues:s,indices:i}=t,{outputRaggedRank:o}=r,l=a.map(e=>n.data.get(e.dataId).values),u=a.map(e=>e.shape),h=n.data.get(s.dataId).values,d=n.data.get(i.dataId).values,[p,c,f]=JA(l,u,h,s.shape,s.dtype,d,i.shape),m=p.map(e=>n.makeTensorInfo([e.length],"int32",e)),g=n.makeTensorInfo(f,s.dtype,c);return m.concat([g])}};var HD={kernelName:er,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n}=e,{starts:r,limits:a,deltas:s}=t,i=n.data.get(r.dataId).values,o=n.data.get(a.dataId).values,l=n.data.get(s.dataId).values,[u,h]=e$(i,r.shape,r.dtype,o,a.shape,l,s.shape);return[n.makeTensorInfo([u.length],"int32",u),n.makeTensorInfo([h.length],r.dtype,h)]}};var jD={kernelName:tr,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{shape:a,values:s,defaultValue:i,rowPartitionTensors:o}=t,{rowPartitionTypes:l}=r,u=n.data.get(a.dataId).values,h=n.data.get(s.dataId).values,d=n.data.get(i.dataId).values,p=o.map(e=>n.data.get(e.dataId).values),c=o.map(e=>e.shape),[f,m]=s$(u,a.shape,h,s.shape,s.dtype,d,i.shape,p,c,l);return n.makeTensorInfo(f,s.dtype,m)}};var qD={kernelName:nr,backendName:"cpu",kernelFunc:function(e){let{backend:t,attrs:n}=e,{start:r,stop:a,dtype:s,step:i}=n,o=i$(r,a,i,s);return t.makeTensorInfo([o.length],s,o)}},KD=UE(ar,e=>1/e),XD={kernelName:ar,backendName:"cpu",kernelFunc:KD};var ZD={kernelName:ur,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{images:a}=t,{alignCorners:s,halfPixelCenters:i,size:o}=r;fE(a,"resizeBilinear");let l=va.computeStrides(a.shape),[u,h]=o,[d,p,c,f]=a.shape,m=n.data.get(a.dataId).values,g=new Float32Array(va.sizeFromShape([d,u,h,f])),y=[s&&u>1?p-1:p,s&&h>1?c-1:c],b=[s&&u>1?u-1:u,s&&h>1?h-1:h],x=0,v=y[0]/b[0],w=y[1]/b[1];for(let e=0;e<d;e++)for(let t=0;t<u;t++){let n;n=i?v*(t+.5)-.5:v*t;let r=Math.max(0,Math.floor(n)),a=n-r,s=Math.min(p-1,Math.ceil(n)),o=e*l[0]+r*l[1],u=e*l[0]+s*l[1];for(let e=0;e<h;e++){let t;t=i?w*(e+.5)-.5:w*e;let n=Math.max(0,Math.floor(t)),r=t-n,s=Math.min(c-1,Math.ceil(t)),h=o+n*l[2],d=u+n*l[2],p=o+s*l[2],y=u+s*l[2];for(let e=0;e<f;e++){let t=m[h+e],n=m[d+e],s=t+(m[p+e]-t)*r,i=s+(n+(m[y+e]-n)*r-s)*a;g[x++]=i}}}return n.makeTensorInfo([d,u,h,f],"float32",g)}};var YD={kernelName:hr,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{images:a,dy:s}=t,{alignCorners:i}=r;fE([s,a],"resizeBilinearGrad");let o=va.computeStrides(a.shape),[l,u,h,d]=a.shape,[,p,c]=s.shape,f=new Float32Array(l*u*h*d),m=[i&&p>1?u-1:u,i&&c>1?h-1:h],g=[i&&p>1?p-1:p,i&&c>1?c-1:c],y=m[0]/g[0],b=m[1]/g[1],x=n.data.get(s.dataId).values,v=0;for(let e=0;e<l;e++){let t=e*o[0];for(let e=0;e<p;e++){let n=e*y,r=Math.floor(n),a=Math.min(Math.ceil(n),u-1),s=t+r*o[1],i=t+a*o[1],l=n-r,p=1-l;for(let e=0;e<c;e++){let t=e*b,n=Math.floor(t),r=Math.min(Math.ceil(t),h-1),a=t-n,u=1-a,c=s+n*o[2],m=s+r*o[2],g=i+n*o[2],y=i+r*o[2],w=p*u,k=p*a,I=l*u,S=l*a;for(let e=0;e<d;e++){let t=x[v++];f[c+e]+=t*w,f[m+e]+=t*k,f[g+e]+=t*I,f[y+e]+=t*S}}}}return n.makeTensorInfo([l,h,u,d],"float32",f)}};var JD={kernelName:or,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{images:a}=t,{alignCorners:s,halfPixelCenters:i,size:o}=r;fE(a,"resizeNearestNeighbor");let l=va.computeStrides(a.shape),[u,h]=o,[d,p,c,f]=a.shape,m=n.data.get(a.dataId).values,g=new Float32Array(d*u*h*f),y=[s&&u>1?p-1:p,s&&h>1?c-1:c],b=[s&&u>1?u-1:u,s&&h>1?h-1:h],x=y[0]/b[0],v=y[1]/b[1],w=0;for(let e=0;e<d;e++){let t=e*l[0];for(let e=0;e<u;e++){let n=i?x*(e+.5):x*e,r=Math.min(p-1,s?Math.round(n):Math.floor(n));i&&(r=Math.max(0,r));let a=t+r*l[1];for(let e=0;e<h;e++){let t=i?v*(e+.5):v*e,n=Math.min(c-1,s?Math.round(t):Math.floor(t));i&&(n=Math.max(0,n));let r=a+n*l[2];for(let e=0;e<f;e++){let t=m[r+e];g[w++]=t}}}}return n.makeTensorInfo([d,u,h,f],a.dtype,g)}};var QD={kernelName:lr,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{images:a,dy:s}=t,{alignCorners:i}=r;fE([s,a],"resizeNearestNeighborGrad");let o=va.computeStrides(a.shape),l=va.computeStrides(s.shape),[u,h,d,p]=a.shape,[,c,f]=s.shape,m=new Float32Array(u*h*d*p),g=n.data.get(s.dataId).values,y=[i&&c>1?h-1:h,i&&f>1?d-1:d],b=[i&&c>1?c-1:c,i&&f>1?f-1:f],x=y[0]/b[0],v=y[1]/b[1],w=1/x,k=1/v,I=2*Math.ceil(w)+2,S=2*Math.ceil(k)+2;for(let e=0;e<u;e++){let t=e*o[0];for(let e=0;e<h;e++){let n=t+e*o[1],r=Math.floor(e*w),a=Math.floor(r-I/2);for(let r=0;r<d;r++){let s=n+r*o[2],u=Math.floor(r*k),y=Math.floor(u-S/2);for(let n=0;n<p;n++){let o=0;for(let s=0;s<I;s++){let u=s+a;if(u<0||u>=c)continue;let p=t+u*l[1],m=u*x;if(e===Math.min(h-1,i?Math.round(m):Math.floor(m)))for(let e=0;e<S;e++){let t=e+y;if(t<0||t>=f)continue;let a=p+t*l[2],s=t*v;r===Math.min(d-1,i?Math.round(s):Math.floor(s))&&(o+=g[a+n])}}m[s+n]=o}}}}return n.makeTensorInfo(a.shape,a.dtype,m)}};var eM={kernelName:pr,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{dims:s}=r;fE(a,"reverse");let i=a.shape.length,o=va.parseAxisParam(s,a.shape);if(0===i)return SE({inputs:{x:a},backend:n});let l=new Ja(a.shape,a.dtype),u=n.bufferSync(a);for(let e=0;e<l.size;e++){let t=l.indexToLoc(e),n=t.slice();o.forEach(e=>n[e]=a.shape[e]-1-n[e]),l.set(u.get(...n),...t)}return n.makeTensorInfo(l.shape,l.dtype,l.values)}},tM={kernelName:ra,backendName:"cpu",kernelFunc:({inputs:e,attrs:t,backend:n})=>{let{image:r}=e,{radians:a,fillValue:s,center:i}=t,o=n,l=va.getTypedArrayFromDType(r.dtype,va.sizeFromShape(r.shape)),[u,h,d,p]=r.shape,[c,f]=Gf.getImageCenter(i,h,d),m=Math.sin(a),g=Math.cos(a),y=o.data.get(r.dataId).values;for(let e=0;e<u;e++){let t=e*d*h*p;for(let e=0;e<h;e++){let n=e*(d*p);for(let r=0;r<d;r++){let a=r*p;for(let i=0;i<p;i++){let o=[u,e,r,i],b=o[2],x=o[1],v=(b-c)*g-(x-f)*m,w=(b-c)*m+(x-f)*g;v=Math.round(v+c),w=Math.round(w+f);let k=s;if("number"!=typeof s&&(k=3===i?255:s[i]),v>=0&&v<d&&w>=0&&w<h){k=y[t+w*(d*p)+v*p+i]}l[t+n+a+i]=k}}}}return{dataId:o.write(l,r.shape,r.dtype),shape:r.shape,dtype:r.dtype}}},nM=UE(cr,e=>{let t=Math.floor(e);return e-t<.5?Math.floor(e):e-t>.5?Math.ceil(e):t%2==0?t:t+1}),rM={kernelName:cr,backendName:"cpu",kernelFunc:nM};var aM={kernelName:mr,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{indices:a,updates:s}=t,{shape:i}=r,{sliceRank:o,numUpdates:l,sliceSize:u,strides:h,outputSize:d}=Gf.calculateShapes(s,a,i),p=h$(n.bufferSync(a),n.bufferSync(s),i,d,u,l,o,h,0,!0);return n.makeTensorInfo(i,p.dtype,p.values)}};function sM(e,t){let n=0,r=e.length,a=0;for(;n<r;)a=Math.floor((n+r)/2),e[a]<t?n=a+1:r=a;return r}function iM(e,t){let n=0,r=e.length,a=0;for(;n<r;)a=Math.floor((n+r)/2),e[a]<=t?n=a+1:r=a;return r}var oM={kernelName:yr,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{sortedSequence:a,values:s}=t,{side:i}=r,o=function(e,t,n,r,a,s){let i=va.getArrayFromDType("int32",n*a);for(let o=0;o<n;++o){let n=e.slice(o*r,(o+1)*r),l=o*a;for(let e=0;e<a;++e)i[l+e]="left"===s?sM(n,t[e+l]):iM(n,t[e+l])}return i}(n.data.get(a.dataId).values,n.data.get(s.dataId).values,a.shape[0],a.shape[1],s.shape[1],i);return n.makeTensorInfo(s.shape,"int32",o)}};var lM={kernelName:br,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n}=e,{condition:r,t:a,e:s}=t;fE([r,a,s],"select");let i=r.shape.length,o=n.data.get(r.dataId).values,l=n.data.get(a.dataId).values,u=n.data.get(s.dataId).values,h=ps(a.dtype,s.dtype),d=va.makeZerosTypedArray(va.sizeFromShape(a.shape),h),p=0,c=0===i||i>1||1===a.shape.length?1:va.sizeFromShape(a.shape.slice(1));for(let e=0;e<o.length;e++)for(let t=0;t<c;t++)1===o[e]?d[p++]=l[e]:d[p++]=u[e];return n.makeTensorInfo(a.shape,h,d)}},uM=Gf.SELU_SCALEALPHA,hM=Gf.SELU_SCALE,dM=UE(xr,e=>e>=0?hM*e:uM*(Math.exp(e)-1)),pM={kernelName:xr,backendName:"cpu",kernelFunc:dM},cM=UE(Ir,e=>e<0?-1:e>0?1:0),fM={kernelName:Ir,backendName:"cpu",kernelFunc:cM},mM=UE(wr,e=>Math.sin(e)),gM={kernelName:wr,backendName:"cpu",kernelFunc:mM},yM=UE(kr,e=>Math.sinh(e)),bM={kernelName:kr,backendName:"cpu",kernelFunc:yM},xM=Math.log(1.1920928955078125e-7)+2,vM=UE(_r,e=>{let t,n=e>-xM,r=e<xM,a=Math.exp(e);return t=r?a:n?e:Math.log(1+a),t}),wM={kernelName:_r,backendName:"cpu",kernelFunc:vM};var kM={kernelName:Cr,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{blockShape:s,paddings:i}=r;fE([a],"spaceToBatchND");let o=va.sizeFromShape(s),l=[[0,0]];l.push(...i);for(let e=1+s.length;e<a.shape.length;++e)l.push([0,0]);let u=BD.kernelFunc({inputs:{x:a},backend:n,attrs:{paddings:l,constantValue:0}}),h=Gf.getReshaped(u.shape,s,o,!1),d=Gf.getPermuted(h.length,s.length,!1),p=Gf.getReshapedPermuted(u.shape,s,o,!1),c=rR({inputs:{x:u},backend:n,attrs:{shape:h}}),f=HA({inputs:{x:c},backend:n,attrs:{perm:d}}),m=rR({inputs:{x:f},backend:n,attrs:{shape:p}});return n.disposeIntermediateTensorInfo(u),n.disposeIntermediateTensorInfo(c),n.disposeIntermediateTensorInfo(f),m}};var IM={kernelName:$r,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n}=e,{indices:r,values:a,denseShape:s,defaultValue:i}=t;if(1!==s.shape.length)throw new Error(`Dense shape must be a vector, saw:\n        ${s.shape}`);if(2!==r.shape.length)throw new Error(`Indices must be a matrix, saw:\n        ${r.shape}`);if(1!==a.shape.length)throw new Error(`Values must be a vector, saw:\n        ${a.shape}`);if(0!==i.shape.length)throw new Error(`Default value must be a scalar, saw:\n        ${i.shape}`);let o=n.data.get(r.dataId).values,l=n.data.get(a.dataId).values,u=n.data.get(s.dataId).values,h=n.data.get(i.dataId).values[0],[d,p,c,f,m]=y$(o,r.shape,r.dtype,l,a.dtype,u,h);return[n.makeTensorInfo(p,r.dtype,d),n.makeTensorInfo([p[0]],a.dtype,c),n.makeTensorInfo([f.length],"bool",new Uint8Array(f.map(e=>Number(e)))),n.makeTensorInfo([m.length],r.dtype,new Int32Array(m))]}};var SM={kernelName:Rr,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n}=e,{inputIndices:r,inputShape:a,newShape:s}=t;if(2!==r.shape.length)throw new Error(`Input indices should be a matrix but received shape\n        ${r.shape}`);if(1!==a.shape.length)throw new Error(`Input shape should be a vector but received shape\n        ${a.shape}`);if(1!==s.shape.length)throw new Error(`Target shape should be a vector but received shape ${s.shape}`);let i=Array.from(n.data.get(a.dataId).values),o=n.data.get(r.dataId).values,l=Array.from(n.data.get(s.dataId).values),[u,h,d]=b$(o,r.shape,r.dtype,i,l);return[n.makeTensorInfo(h,r.dtype,u),n.makeTensorInfo([d.length],s.dtype,new Int32Array(d))]}};var _M={kernelName:Fr,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n}=e,{data:r,indices:a,segmentIds:s}=t;if(r.shape.length<1)throw new Error("Data should be at least 1 dimensional but received scalar");if(1!==a.shape.length)throw new Error(`Indices should be a vector but received shape\n          ${a.shape}`);if(1!==s.shape.length)throw new Error(`Segment ids should be a vector but received shape\n          ${s.shape}`);if(a.shape[0]!==s.shape[0])throw new Error("segmentIds and indices should have same size.");let i=n.data.get(r.dataId).values,o=n.data.get(a.dataId).values,l=n.data.get(s.dataId).values,[u,h]=x$(i,r.shape,r.dtype,o,l,!0);return n.makeTensorInfo(h,r.dtype,u)}};var NM={kernelName:Dr,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n}=e,{data:r,indices:a,segmentIds:s}=t;if(r.shape.length<1)throw new Error("Data should be at least 1 dimensional but received scalar");if(1!==a.shape.length)throw new Error(`Indices should be a vector but received shape\n         ${a.shape}`);if(1!==s.shape.length)throw new Error(`Segment ids should be a vector but received shape\n         ${s.shape}`);if(a.shape[0]!==s.shape[0])throw new Error("segmentIds and indices should have same size.");let i=n.data.get(r.dataId).values,o=n.data.get(a.dataId).values,l=n.data.get(s.dataId).values,[u,h]=x$(i,r.shape,r.dtype,o,l);return n.makeTensorInfo(h,r.dtype,u)}};var TM={kernelName:Mr,backendName:"cpu",kernelFunc:function(e){let t,{inputs:n,backend:r,attrs:a}=e,{sparseIndices:s,sparseValues:i,defaultValue:o}=n,{outputShape:l}=a,{sliceRank:u,numUpdates:h,sliceSize:d,strides:p,outputSize:c}=Gf.calculateShapes(i,s,l),f=!1,m=r.bufferSync(s);switch(i.dtype){case"bool":t=h$(m,r.bufferSync(i),l,c,d,h,u,p,!!r.data.get(o.dataId).values[0],f);break;case"float32":case"int32":t=h$(m,r.bufferSync(i),l,c,d,h,u,p,r.data.get(o.dataId).values[0],f);break;case"string":t=h$(m,r.bufferSync(i),l,c,d,h,u,p,va.decodeString(r.data.get(o.dataId).values[0]),f);break;default:throw new Error(`Unsupported type ${i.dtype}`)}return r.makeTensorInfo(l,t.dtype,t.values)}};var CM={kernelName:Er,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{numOrSizeSplits:s,axis:i}=r,o=va.parseAxisParam(i,a.shape)[0],l=Gf.prepareSplitSize(a,s,o),u=new Array(a.shape.length).fill(0),h=a.shape.slice();return l.map(e=>{let t=[...h];t[o]=e;let r=m$({inputs:{x:a},backend:n,attrs:{begin:u,size:t}});return u[o]+=e,r})}},EM={kernelName:Lr,backendName:"cpu",kernelFunc:({inputs:e,backend:t})=>{let{x:n}=e,r=t;fE(n,"square");let a=r.data.get(n.dataId).values,s=new Float32Array(a.length);for(let e=0;e<a.length;++e){let t=a[e];s[e]=t*t}return{dataId:r.write(s,n.shape,n.dtype),shape:n.shape,dtype:n.dtype}}},AM=UE(ta,(e,t)=>{let n=t;return isNaN(e)?NaN:e>0?1:n.alpha}),$M={kernelName:ta,backendName:"cpu",kernelFunc:AM};var RM={kernelName:zr,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{begin:s,end:i,strides:o,beginMask:l,endMask:u,ellipsisMask:h,newAxisMask:d,shrinkAxisMask:p}=r;fE(a,"stridedSlice");let c,{finalShapeSparse:f,finalShape:m,isIdentity:g,sliceDim0:y,isSimpleSlice:b,begin:x,end:v,strides:w}=vf.sliceInfo(a.shape,s,i,o,l,u,h,d,p);if(g)c=rR({inputs:{x:a},backend:n,attrs:{shape:m}});else if(y||b){va.assert(a.shape.length>=1,()=>`Input must have rank at least 1, got: ${a.shape.length}`);let e=vf.computeOutShape(x,v,w),t=m$({inputs:{x:a},backend:n,attrs:{begin:x,size:e}});c=rR({inputs:{x:t},backend:n,attrs:{shape:m}}),n.disposeIntermediateTensorInfo(t)}else{let e=E$(f,n.bufferSync(a),w,x);c=n.makeTensorInfo(m,e.dtype,e.values)}return c}};var FM={kernelName:Br,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{separator:a,nGramWidths:s,leftPad:i,rightPad:o,padWidth:l,preserveShortSequences:u}=r,{data:h,dataSplits:d}=t,p=n.data.get(h.dataId).values,c=n.data.get(d.dataId).values,[f,m]=$$(p,c,a,s,i,o,l,u);return[n.makeTensorInfo([f.length],"string",f),n.makeTensorInfo(d.shape,"int32",m)]}};var DM={kernelName:Wr,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{skipEmpty:a}=r,{input:s,delimiter:i}=t;if("string"!==s.dtype)throw new Error("Input must be of datatype string");if(1!==s.shape.length)throw new Error(`Input must be a vector, got shape: ${s.shape}`);if(0!==i.shape.length)throw new Error(`Delimiter must be a scalar, got shape: ${i.shape}`);let o=n.data.get(s.dataId).values,l=n.data.get(i.dataId).values[0],[u,h,d]=F$(o,l,a),p=h.length;return[n.makeTensorInfo([p,2],"int32",u),n.makeTensorInfo([p],"string",h),n.makeTensorInfo([2],"int32",new Int32Array(d))]}};var MM={kernelName:Vr,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{numBuckets:a}=r,{input:s}=t;if("string"!==s.dtype)throw new Error("Input must be of datatype string");if(a<=0)throw new Error("Number of buckets must be at least 1");let i=D$(n.data.get(s.dataId).values,a);return n.makeTensorInfo(s.shape,"int32",i)}},OM=UE(Gr,e=>Math.tan(e)),LM={kernelName:Gr,backendName:"cpu",kernelFunc:OM},PM=UE(Hr,e=>Math.tanh(e));var zM={kernelName:gr,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n}=e,{tensor:r,indices:a,updates:s}=t,{sliceRank:i,numUpdates:o,sliceSize:l,strides:u,outputSize:h}=Gf.calculateShapes(s,a,r.shape),d=n.bufferSync(a),p=n.bufferSync(s),c=n.bufferSync(r),f=h$(d,p,r.shape,h,l,o,i,u,c,!1);return n.makeTensorInfo(r.shape,f.dtype,f.values)}};var BM={kernelName:jr,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{reps:s}=r;fE(a,"tile");let i=z$(n.bufferSync(a),s);return n.makeTensorInfo(i.shape,i.dtype,i.values)}};var WM={kernelName:qr,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{k:s,sorted:i}=r;fE(a,"topk");let o=n.data.get(a.dataId).values,[l,u]=V$(o,a.shape,a.dtype,s,i);return[n.makeTensorInfo(l.shape,l.dtype,l.values),n.makeTensorInfo(u.shape,u.dtype,u.values)]}};var VM={kernelName:Kr,backendName:"cpu",kernelFunc:function(e){let{inputs:t,attrs:n,backend:r}=e,{image:a,transforms:s}=t,{interpolation:i,fillMode:o,fillValue:l,outputShape:u}=n,[h,d,p,c]=a.shape,[f,m]=null!=u?u:[d,p],g=[h,f,m,c],y=va.computeStrides(a.shape),b=y[0],x=y[1],v=y[2],w=va.computeStrides(g),k=w[0],I=w[1],S=w[2],_=va.getTypedArrayFromDType(a.dtype,va.sizeFromShape(g));_.fill(l);let N=r.data.get(a.dataId).values,T=r.data.get(s.dataId).values;for(let e=0;e<h;++e){let t=1===s.shape[0]?T:T.subarray(8*e,8*e+8);for(let n=0;n<f;++n)for(let r=0;r<m;++r)for(let a=0;a<c;++a){let s,u=t[6]*r+t[7]*n+1;if(0===u)continue;let h=(t[0]*r+t[1]*n+t[2])/u,c=(t[3]*r+t[4]*n+t[5])/u,f=UM(h,p,o),m=UM(c,d,o);switch(i){case"nearest":s=HM(N,d,p,b,x,v,e,m,f,a,l);break;case"bilinear":s=jM(N,d,p,b,x,v,e,m,f,a,l);break;default:throw new Error(`Error in Transform: Expect 'nearest' or 'bilinear', but got ${i}`)}_[e*k+n*I+r*S+a]=s}return r.makeTensorInfo(g,a.dtype,_)}return{dataId:r.write(_,g,a.dtype),shape:a.shape,dtype:a.dtype}}};function UM(e,t,n){switch(n){case"reflect":return function(e,t){let n=e;if(n<0)if(t<=1)n=0;else{let e=2*t;n<e&&(n=e*Math.trunc(-n/e)+n),n=n<-t?n+e:-n-1}else if(n>t-1)if(t<=1)n=0;else{let e=2*t;n-=e*Math.trunc(n/e),n>=t&&(n=e-n-1)}return va.clamp(0,n,t-1)}(e,t);case"wrap":return function(e,t){let n=e;if(n<0)if(t<=1)n=0;else{let e=t-1;n+=t*(Math.trunc(-n/e)+1)}else if(n>t-1)if(t<=1)n=0;else{let e=t-1;n-=t*Math.trunc(n/e)}return va.clamp(0,n,t-1)}(e,t);case"nearest":return function(e,t){return va.clamp(0,e,t-1)}(e,t);default:return function(e){return e}(e)}}function GM(e,t,n,r,a,s,i,o,l,u,h){return 0<=o&&o<t&&0<=l&&l<n?e[i*r+o*a+l*s+u]:h}function HM(e,t,n,r,a,s,i,o,l,u,h){return GM(e,t,n,r,a,s,i,Math.round(o),Math.round(l),u,h)}function jM(e,t,n,r,a,s,i,o,l,u,h){let d=Math.floor(o),p=Math.floor(l),c=d+1,f=p+1;return(c-o)*((f-l)*GM(e,t,n,r,a,s,i,d,p,u,h)+(l-p)*GM(e,t,n,r,a,s,i,d,f,u,h))+(o-d)*((f-l)*GM(e,t,n,r,a,s,i,c,p,u,h)+(l-p)*GM(e,t,n,r,a,s,i,c,f,u,h))}var qM={kernelName:Zr,backendName:"cpu",kernelFunc:function(e){let{inputs:t,attrs:n,backend:r}=e,{axis:a}=n,{x:s}=t;fE(s,"unique");let i=r.data.get(s.dataId).values,{outputValues:o,outputShape:l,indices:u}=U$(i,a,s.shape,s.dtype);return[r.makeTensorInfo(l,s.dtype,o),r.makeTensorInfo([u.length],"int32",u)]}};var KM={kernelName:Yr,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{value:a}=t,{axis:s}=r;s<0&&(s+=a.shape.length);let i=a.shape.length,o=a.shape[s],l=new Array(i-1),u=0;for(let e=0;e<i;e++)e!==s&&(l[u++]=a.shape[e]);let h=new Array(i).fill(0),d=a.shape.slice();d[s]=1;let p=new Array(o);for(let e=0;e<p.length;e++){h[s]=e;let t=m$({inputs:{x:a},backend:n,attrs:{begin:h,size:d}});p[e]=rR({inputs:{x:t},backend:n,attrs:{shape:l}}),n.disposeIntermediateTensorInfo(t)}return p}};var XM={kernelName:Jr,backendName:"cpu",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a,segmentIds:s}=t,{numSegments:i}=r;fE(a,"unsortedSegmentSum");let o=[],l=[],u=a.shape.length-s.shape.length,h=s;for(let e=0;e<u;++e){let t=EF({inputs:{input:h},backend:n,attrs:{dim:e+1}});h=t,l.push(t)}for(let e=0;e<i;++e){let t=va.createScalarValue(e,"int32"),r=n.makeTensorInfo([],"int32",t),s=ZE({inputs:{a:r,b:h},backend:n}),i=EE({inputs:{x:s},backend:n,attrs:{dtype:"float32"}}),u=LA({inputs:{a:i,b:a},backend:n}),d=yF({inputs:{x:u},backend:n,attrs:{axis:0,keepDims:!1}});o.push(d),l.push(r),l.push(s),l.push(i),l.push(u),l.push(d)}let d=PD({inputs:o,backend:n,attrs:{axis:0}});return l.forEach(e=>n.disposeIntermediateTensorInfo(e)),d}},ZM=[oR,xE,uR,dR,OE,pR,cR,fR,mR,gR,bR,vR,kR,_R,TR,$R,RR,FR,DR,iR,MR,OR,LR,WE,PR,AE,qE,BR,kE,WR,HR,qR,KR,XR,ZR,YR,JR,eF,nF,rF,aF,sF,iF,oF,uF,hF,dF,pF,cF,fF,mF,gF,xF,j$,vF,YE,CF,eA,AF,rA,LF,zF,BF,iA,uA,WF,VF,UF,GF,fA,yA,_E,HF,UR,qF,XF,YF,K$,vA,IA,JF,TA,eD,rD,sD,lD,uD,hD,pD,$A,cD,fD,mD,gD,yD,bD,xD,DA,vD,ID,TD,PA,BA,ED,$D,FD,UA,DD,LD,zD,BD,UD,Y$,KA,GD,HD,jD,qD,TE,FF,XD,Q$,tR,aR,ZD,YD,JD,QD,eM,tM,rM,u$,aM,oM,lM,pM,c$,fM,gM,bM,g$,ND,wM,kM,IM,SM,_M,NM,TM,CM,k$,EM,_$,C$,$M,RM,FM,DM,MM,P$,bF,LM,{kernelName:Hr,backendName:"cpu",kernelFunc:PM},zM,BM,WM,VM,jA,qM,KM,XM,OD];for(let e of ZM)fa(e);var YM={};f(YM,{assertNotComplex:()=>aL,bindCanvasToFramebuffer:()=>DO,bindColorTextureToFramebuffer:()=>MO,bindTextureToProgramUniformSampler:()=>FO,bindTextureUnit:()=>EO,bindVertexBufferToProgramAttribute:()=>CO,callAndCheck:()=>uO,canBeRepresented:()=>hO,createFragmentShader:()=>fO,createFramebuffer:()=>TO,createProgram:()=>xO,createStaticIndexBuffer:()=>IO,createStaticVertexBuffer:()=>kO,createTexture:()=>_O,createVertexShader:()=>cO,getBatchDim:()=>WO,getExtensionOrThrow:()=>pO,getFramebufferErrorMessage:()=>PO,getMaxTexturesInShader:()=>ZO,getNumChannels:()=>SO,getProgramUniformLocation:()=>RO,getProgramUniformLocationOrThrow:()=>$O,getRowsCols:()=>VO,getShapeAs3D:()=>UO,getTextureShapeFromLogicalShape:()=>GO,getWebGLDisjointQueryTimerVersion:()=>YO,getWebGLErrorMessage:()=>dO,getWebGLMaxTextureSize:()=>qO,hasExtension:()=>JO,isCapableOfRenderingToFloatTexture:()=>eL,isDownloadFloatTextureEnabled:()=>tL,isReshapeFree:()=>jO,isWebGLFenceEnabled:()=>rL,isWebGLVersionEnabled:()=>QO,linkProgram:()=>vO,logShaderSourceAndInfoLog:()=>bO,resetMaxTextureSize:()=>KO,resetMaxTexturesInShader:()=>XO,unbindColorTextureFromFramebuffer:()=>OO,unbindTextureUnit:()=>AO,validateFramebuffer:()=>LO,validateProgram:()=>wO,validateTextureSize:()=>NO});var JM,QM,eO,tO={},nO={alpha:!1,antialias:!1,premultipliedAlpha:!1,preserveDrawingBuffer:!1,depth:!1,stencil:!1,failIfMajorPerformanceCaveat:!0};function rO(e,t){tO[e]=t}function aO(e,t){if(!(e in tO)||null!=t){let n=function(e,t){if(1!==e&&2!==e)throw new Error("Cannot get WebGL rendering context, WebGL is disabled.");let n=t??function(e){if(!Le().getBool("IS_SAFARI")&&"undefined"!=typeof OffscreenCanvas&&2===e)return new OffscreenCanvas(300,150);if("undefined"!=typeof document)return document.createElement("canvas");throw new Error("Cannot create a canvas in this context")}(e);return n.addEventListener("webglcontextlost",t=>{t.preventDefault(),delete tO[e]},!1),Le().getBool("SOFTWARE_WEBGL_ENABLED")&&(nO.failIfMajorPerformanceCaveat=!1),1===e?n.getContext("webgl",nO)||n.getContext("experimental-webgl",nO):n.getContext("webgl2",nO)}(e,t);if(null===n)return console.log("Could not get context for WebGL version",e),null;tO[e]=n}let n=tO[e];return null==n||n.isContextLost()?(delete tO[e],aO(e)):(n.disable(n.DEPTH_TEST),n.disable(n.STENCIL_TEST),n.disable(n.BLEND),n.disable(n.DITHER),n.disable(n.POLYGON_OFFSET_FILL),n.disable(n.SAMPLE_COVERAGE),n.enable(n.SCISSOR_TEST),n.enable(n.CULL_FACE),n.cullFace(n.BACK),tO[e])}function sO(e,t){return[t,e]}function iO(e){let t=va.sizeFromShape(e),n=Math.ceil(t/4);return va.sizeToSquarishShape(n)}function oO(e,t){return[Math.max(1,Math.ceil(t/2)),Math.max(1,Math.ceil(e/2))]}function lO(e,t){let n,r,a,s,i,o,l,u,h,d,p=e;return 2===Le().getNumber("WEBGL_VERSION")?(n=p.R32F,r=p.R16F,a=p.RGBA16F,s=p.RGBA32F,i=p.RED,l=4,u=1,h=p.HALF_FLOAT,d=p.FLOAT,o=p.RGBA8):(n=e.RGBA,r=e.RGBA,a=e.RGBA,s=p.RGBA,i=e.RGBA,l=4,u=4,h=null!=t?t.HALF_FLOAT_OES:null,d=e.FLOAT,o=e.RGBA),{internalFormatFloat:n,internalFormatHalfFloat:r,internalFormatPackedHalfFloat:a,internalFormatPackedFloat:s,textureFormatFloat:i,downloadTextureFormat:o,downloadUnpackNumChannels:l,defaultNumChannels:u,textureTypeHalfFloat:h,textureTypeFloat:d}}function uO(e,t){let n=t();return Le().getBool("DEBUG")&&function(e){let t=e.getError();if(t!==e.NO_ERROR)throw new Error("WebGL Error: "+dO(e,t))}(e),n}!function(e){e[e.DENSE=0]="DENSE",e[e.SHARED_BATCH=1]="SHARED_BATCH"}(JM||(JM={})),function(e){e[e.RENDER=0]="RENDER",e[e.UPLOAD=1]="UPLOAD",e[e.PIXELS=2]="PIXELS",e[e.DOWNLOAD=3]="DOWNLOAD"}(QM||(QM={})),function(e){e[e.UNPACKED_FLOAT16=0]="UNPACKED_FLOAT16",e[e.UNPACKED_FLOAT32=1]="UNPACKED_FLOAT32",e[e.PACKED_4X1_UNSIGNED_BYTE=2]="PACKED_4X1_UNSIGNED_BYTE",e[e.PACKED_2X2_FLOAT32=3]="PACKED_2X2_FLOAT32",e[e.PACKED_2X2_FLOAT16=4]="PACKED_2X2_FLOAT16"}(eO||(eO={}));function hO(e){return!!(Le().getBool("WEBGL_RENDER_FLOAT32_ENABLED")||0===e||5.96e-8<Math.abs(e)&&Math.abs(e)<65504)}function dO(e,t){switch(t){case e.NO_ERROR:return"NO_ERROR";case e.INVALID_ENUM:return"INVALID_ENUM";case e.INVALID_VALUE:return"INVALID_VALUE";case e.INVALID_OPERATION:return"INVALID_OPERATION";case e.INVALID_FRAMEBUFFER_OPERATION:return"INVALID_FRAMEBUFFER_OPERATION";case e.OUT_OF_MEMORY:return"OUT_OF_MEMORY";case e.CONTEXT_LOST_WEBGL:return"CONTEXT_LOST_WEBGL";default:return`Unknown error code ${t}`}}function pO(e,t){return zO(e,()=>e.getExtension(t),'Extension "'+t+'" not supported on this browser.')}function cO(e,t){let n=zO(e,()=>e.createShader(e.VERTEX_SHADER),"Unable to create vertex WebGLShader.");if(uO(e,()=>e.shaderSource(n,t)),uO(e,()=>e.compileShader(n)),!1===e.getShaderParameter(n,e.COMPILE_STATUS))throw console.log(e.getShaderInfoLog(n)),new Error("Failed to compile vertex shader.");return n}function fO(e,t){let n=zO(e,()=>e.createShader(e.FRAGMENT_SHADER),"Unable to create fragment WebGLShader.");if(uO(e,()=>e.shaderSource(n,t)),uO(e,()=>e.compileShader(n)),Le().get("ENGINE_COMPILE_ONLY"))return n;if(!1===e.getShaderParameter(n,e.COMPILE_STATUS))throw bO(t,e.getShaderInfoLog(n)),new Error("Failed to compile fragment shader.");return n}var mO,gO,yO=/ERROR: [0-9]+:([0-9]+):/g;function bO(e,t){let n=yO.exec(t);if(null==n)return console.log(`Couldn't parse line number in error: ${t}`),void console.log(e);let r=+n[1],a=e.split("\n"),s=a.length.toString().length+2,i=a.map((e,t)=>va.rightPad((t+1).toString(),s)+e),o=0;for(let e=0;e<i.length;e++)o=Math.max(i[e].length,o);let l=i.slice(0,r-1),u=i.slice(r-1,r),h=i.slice(r);console.log(l.join("\n")),console.log(t.split("\n")[0]),console.log(`%c ${va.rightPad(u[0],o)}`,"border:1px solid red; background-color:#e3d2d2; color:#a61717"),console.log(h.join("\n"))}function xO(e){return zO(e,()=>e.createProgram(),"Unable to create WebGLProgram.")}function vO(e,t){if(uO(e,()=>e.linkProgram(t)),!Le().get("ENGINE_COMPILE_ONLY")&&!1===e.getProgramParameter(t,e.LINK_STATUS))throw console.log(e.getProgramInfoLog(t)),new Error("Failed to link vertex and fragment shaders.")}function wO(e,t){if(uO(e,()=>e.validateProgram(t)),!1===e.getProgramParameter(t,e.VALIDATE_STATUS))throw console.log(e.getProgramInfoLog(t)),new Error("Shader program validation failed.")}function kO(e,t){let n=zO(e,()=>e.createBuffer(),"Unable to create WebGLBuffer");return uO(e,()=>e.bindBuffer(e.ARRAY_BUFFER,n)),uO(e,()=>e.bufferData(e.ARRAY_BUFFER,t,e.STATIC_DRAW)),n}function IO(e,t){let n=zO(e,()=>e.createBuffer(),"Unable to create WebGLBuffer");return uO(e,()=>e.bindBuffer(e.ELEMENT_ARRAY_BUFFER,n)),uO(e,()=>e.bufferData(e.ELEMENT_ARRAY_BUFFER,t,e.STATIC_DRAW)),n}function SO(){return 2===Le().getNumber("WEBGL_VERSION")?1:4}function _O(e){return zO(e,()=>e.createTexture(),"Unable to create WebGLTexture.")}function NO(e,t){let n=Le().getNumber("WEBGL_MAX_TEXTURE_SIZE");if(e<=0||t<=0){throw new Error("Requested texture size "+`[${e}x${t}]`+" is invalid.")}if(e>n||t>n){throw new Error("Requested texture size "+`[${e}x${t}]`+" greater than WebGL maximum on this browser / GPU "+`[${n}x${n}]`+".")}}function TO(e){return zO(e,()=>e.createFramebuffer(),"Unable to create WebGLFramebuffer.")}function CO(e,t,n,r,a,s,i){let o=e.getAttribLocation(t,n);return-1!==o&&(uO(e,()=>e.bindBuffer(e.ARRAY_BUFFER,r)),uO(e,()=>e.vertexAttribPointer(o,a,e.FLOAT,!1,s,i)),uO(e,()=>e.enableVertexAttribArray(o)),!0)}function EO(e,t,n){BO(e,n),uO(e,()=>e.activeTexture(e.TEXTURE0+n)),uO(e,()=>e.bindTexture(e.TEXTURE_2D,t))}function AO(e,t){BO(e,t),uO(e,()=>e.activeTexture(e.TEXTURE0+t)),uO(e,()=>e.bindTexture(e.TEXTURE_2D,null))}function $O(e,t,n){return zO(e,()=>e.getUniformLocation(t,n),'uniform "'+n+'" not present in program.')}function RO(e,t,n){return e.getUniformLocation(t,n)}function FO(e,t,n,r){uO(e,()=>EO(e,t,r)),uO(e,()=>e.uniform1i(n,r))}function DO(e){uO(e,()=>e.bindFramebuffer(e.FRAMEBUFFER,null)),uO(e,()=>e.viewport(0,0,e.canvas.width,e.canvas.height)),uO(e,()=>e.scissor(0,0,e.canvas.width,e.canvas.height))}function MO(e,t,n){uO(e,()=>e.bindFramebuffer(e.FRAMEBUFFER,n)),uO(e,()=>e.framebufferTexture2D(e.FRAMEBUFFER,e.COLOR_ATTACHMENT0,e.TEXTURE_2D,t,0))}function OO(e,t){uO(e,()=>e.bindFramebuffer(e.FRAMEBUFFER,t)),uO(e,()=>e.framebufferTexture2D(e.FRAMEBUFFER,e.COLOR_ATTACHMENT0,e.TEXTURE_2D,null,0))}function LO(e){let t=e.checkFramebufferStatus(e.FRAMEBUFFER);if(t!==e.FRAMEBUFFER_COMPLETE)throw new Error("Error binding framebuffer: "+PO(e,t))}function PO(e,t){switch(t){case e.FRAMEBUFFER_INCOMPLETE_ATTACHMENT:return"FRAMEBUFFER_INCOMPLETE_ATTACHMENT";case e.FRAMEBUFFER_INCOMPLETE_MISSING_ATTACHMENT:return"FRAMEBUFFER_INCOMPLETE_MISSING_ATTACHMENT";case e.FRAMEBUFFER_INCOMPLETE_DIMENSIONS:return"FRAMEBUFFER_INCOMPLETE_DIMENSIONS";case e.FRAMEBUFFER_UNSUPPORTED:return"FRAMEBUFFER_UNSUPPORTED";default:return`unknown error ${t}`}}function zO(e,t,n){let r=uO(e,()=>t());if(null==r)throw new Error(n);return r}function BO(e,t){let n=e.MAX_COMBINED_TEXTURE_IMAGE_UNITS-1,r=t+e.TEXTURE0;if(r<e.TEXTURE0||r>n){throw new Error(`textureUnit must be in ${`[gl.TEXTURE0, gl.TEXTURE${n}]`}.`)}}function WO(e,t=2){return va.sizeFromShape(e.slice(0,e.length-t))}function VO(e){if(0===e.length)throw Error("Cannot get rows and columns of an empty shape array.");return[e.length>1?e[e.length-2]:1,e[e.length-1]]}function UO(e){let t=[1,1,1];return 0===e.length||1===e.length&&1===e[0]||(t=[WO(e),...VO(e)]),t}function GO(e,t=!1){let n=Le().getNumber("WEBGL_MAX_TEXTURE_SIZE"),r=Le().getNumber("WEBGL_MAX_SIZE_FOR_NARROW_TEXTURE");r===1/0&&Le().getBool("WEBGL_AUTO_SQUARIFY_NARROW_TEXTURE_SHAPE")&&(r=n/2),t&&(n*=2,r*=2,1===(e=e.map((t,n)=>n>=e.length-2?va.nearestLargerEven(e[n]):e[n])).length&&(e=[2,e[0]])),2!==e.length&&(e=va.squeezeShape(e).newShape);let a=va.sizeFromShape(e),s=null;e.length<=1&&a<=n?s=[1,a]:2===e.length&&e[0]<=n&&e[1]<=n?s=e:3===e.length&&e[0]*e[1]<=n&&e[2]<=n?s=[e[0]*e[1],e[2]]:3===e.length&&e[0]<=n&&e[1]*e[2]<=n?s=[e[0],e[1]*e[2]]:4===e.length&&e[0]*e[1]*e[2]<=n&&e[3]<=n?s=[e[0]*e[1]*e[2],e[3]]:4===e.length&&e[0]<=n&&e[1]*e[2]*e[3]<=n&&(s=[e[0],e[1]*e[2]*e[3]]);let i=null!=s&&Math.max(...s)>r&&Math.min(...s)<=(t?2:1)&&Math.min(...s)>0;if(null==s||i)if(t){let t=WO(e),n=2,r=2;e.length&&([n,r]=VO(e)),a=t*(n/2)*(r/2),s=va.sizeToSquarishShape(a).map(e=>2*e)}else s=va.sizeToSquarishShape(a);return s}function HO(e){return e%2==0}function jO(e,t){if(e=e.slice(-2),t=t.slice(-2),va.arraysEqual(e,t)||!e.length||!t.length||0===e[0]||0===e[1]||0===t[0]||0===t[1])return!0;if(e.length!==t.length){let n=e[e.length-1],r=t[t.length-1];if(n===r||HO(n)&&HO(r)&&(1===e[0]||1===t[0]))return!0}return e[1]===t[1]&&HO(e[0])&&HO(t[0])}function qO(e){if(null==mO){let t=aO(e);mO=t.getParameter(t.MAX_TEXTURE_SIZE)}return mO}function KO(){mO=null}function XO(){gO=null}function ZO(e){if(null==gO){let t=aO(e);gO=t.getParameter(t.MAX_TEXTURE_IMAGE_UNITS)}return Math.min(16,gO)}function YO(e){if(0===e)return 0;let t,n=aO(e);return t=JO(n,"EXT_disjoint_timer_query_webgl2")&&2===e?2:JO(n,"EXT_disjoint_timer_query")?1:0,t}function JO(e,t){return null!=e.getExtension(t)}function QO(e){try{if(null!=aO(e))return!0}catch(e){return console.log("Error when getting WebGL context: ",e),!1}return!1}function eL(e){if(0===e)return!1;let t=aO(e);if(1===e){if(!JO(t,"OES_texture_float"))return!1}else if(!JO(t,"EXT_color_buffer_float"))return!1;return nL(t)}function tL(e){if(0===e)return!1;let t=aO(e);if(1!==e){if(JO(t,"EXT_color_buffer_float"))return nL(t);let e="EXT_color_buffer_half_float";if(JO(t,e)){let n=t.getExtension(e);return function(e,t){let n=lO(e,t),r=e.createTexture();e.bindTexture(e.TEXTURE_2D,r),e.texImage2D(e.TEXTURE_2D,0,n.internalFormatHalfFloat,1,1,0,n.textureFormatFloat,n.textureTypeHalfFloat,null);let a=e.createFramebuffer();e.bindFramebuffer(e.FRAMEBUFFER,a),e.framebufferTexture2D(e.FRAMEBUFFER,e.COLOR_ATTACHMENT0,e.TEXTURE_2D,r,0);let s=e.checkFramebufferStatus(e.FRAMEBUFFER)===e.FRAMEBUFFER_COMPLETE;return e.bindTexture(e.TEXTURE_2D,null),e.bindFramebuffer(e.FRAMEBUFFER,null),e.deleteTexture(r),e.deleteFramebuffer(a),s}(t,n)}return!1}return!(!JO(t,"OES_texture_float")||!JO(t,"WEBGL_color_buffer_float"))&&nL(t)}function nL(e){let t=lO(e),n=e.createTexture();e.bindTexture(e.TEXTURE_2D,n),e.texImage2D(e.TEXTURE_2D,0,t.internalFormatFloat,1,1,0,t.textureFormatFloat,t.textureTypeFloat,null);let r=e.createFramebuffer();e.bindFramebuffer(e.FRAMEBUFFER,r),e.framebufferTexture2D(e.FRAMEBUFFER,e.COLOR_ATTACHMENT0,e.TEXTURE_2D,n,0);let a=e.checkFramebufferStatus(e.FRAMEBUFFER)===e.FRAMEBUFFER_COMPLETE;return e.bindTexture(e.TEXTURE_2D,null),e.bindFramebuffer(e.FRAMEBUFFER,null),e.deleteTexture(n),e.deleteFramebuffer(r),a}function rL(e){return 2===e&&null!=aO(e).fenceSync}function aL(e,t){Array.isArray(e)||(e=[e]),e.forEach(e=>{null!=e&&va.assert("complex64"!==e.dtype,()=>`${t} does not support complex64 tensors in the WebGL backend.`)})}var sL=Le();function iL(){let e,t,n,r,a,s,i,o,l,u;return 2===Le().getNumber("WEBGL_VERSION")?(e="#version 300 es",t="in",n="out",r="in",a="texture",s="outputColor",i="out vec4 outputColor;",o=Le().getBool("WEBGL2_ISNAN_CUSTOM")?"\n      bool isnan_custom(float val) {\n        uint floatToUint = floatBitsToUint(val);\n        return (floatToUint & 0x7fffffffu) > 0x7f800000u;\n      }\n\n      bvec4 isnan_custom(vec4 val) {\n        return bvec4(isnan_custom(val.x),\n          isnan_custom(val.y), isnan_custom(val.z), isnan_custom(val.w));\n      }\n\n      #define isnan(value) isnan_custom(value)\n    ":"",l="",u="\n      #define round(value) newRound(value)\n      int newRound(float value) {\n        return int(floor(value + 0.5));\n      }\n\n      ivec4 newRound(vec4 value) {\n        return ivec4(floor(value + vec4(0.5)));\n      }\n    "):(e="",t="attribute",n="varying",r="varying",a="texture2D",s="gl_FragColor",i="",o="\n      #define isnan(value) isnan_custom(value)\n      bool isnan_custom(float val) {\n        return (val > 0. || val < 1. || val == 0.) ? false : true;\n      }\n      bvec4 isnan_custom(vec4 val) {\n        return bvec4(isnan(val.x), isnan(val.y), isnan(val.z), isnan(val.w));\n      }\n    ",l="\n      uniform float INFINITY;\n\n      bool isinf(float val) {\n        return abs(val) == INFINITY;\n      }\n      bvec4 isinf(vec4 val) {\n        return equal(abs(val), vec4(INFINITY));\n      }\n    ",u="\n      int round(float value) {\n        return int(floor(value + 0.5));\n      }\n\n      ivec4 round(vec4 value) {\n        return ivec4(floor(value + vec4(0.5)));\n      }\n    "),{version:e,attribute:t,varyingVs:n,varyingFs:r,texture2D:a,output:s,defineOutput:i,defineSpecialNaN:o,defineSpecialInf:l,defineRound:u}}function oL(e,t,n="index"){let r=va.computeStrides(t);return r.map((t,a)=>`${`int ${e[a]} = ${n} / ${t}`}; ${a===r.length-1?`int ${e[a+1]} = ${n} - ${e[a]} * ${t}`:`index -= ${e[a]} * ${t}`};`).join("")}function lL(e,t,n="index"){let r=va.computeStrides(t);return r.map((t,a)=>`${`int ${e[a]} = ${n} / outShapeStrides[${a}]`}; ${a===r.length-1?`int ${e[a+1]} = ${n} - ${e[a]} * outShapeStrides[${a}]`:`index -= ${e[a]} * outShapeStrides[${a}]`};`).join("")}function uL(e,t,n="index"){let r=function(e,t){let n=e.length,r=e.map(e=>`${t}[${e}]`),a=new Array(n-1);a[n-2]=r[n-1];for(let e=n-3;e>=0;--e)a[e]=`(${a[e+1]} * ${r[e+1]})`;return a}(e.map((e,t)=>t),t);return r.map((t,a)=>`${`int ${e[a]} = ${n} / ${r[a]}`}; ${a===r.length-1?`int ${e[a+1]} = ${n} - ${e[a]} * ${r[a]}`:`index -= ${e[a]} * ${r[a]}`};`).join("")}function hL(e){let t=va.computeStrides(e).map(e=>e.toString());return`\n  int getFlatIndex(ivec3 coords) {\n    return coords.x * ${t[0]} + coords.y * ${t[1]} + coords.z;\n  }\n`}sL.registerFlag("HAS_WEBGL",()=>sL.getNumber("WEBGL_VERSION")>0),sL.registerFlag("WEBGL_VERSION",()=>QO(2)?2:QO(1)?1:0),sL.registerFlag("WEBGL_CHECK_NUMERICAL_PROBLEMS",()=>!1),sL.registerFlag("WEBGL_BUFFER_SUPPORTED",()=>2===sL.get("WEBGL_VERSION")),sL.registerFlag("WEBGL_CPU_FORWARD",()=>!0),sL.registerFlag("WEBGL_FORCE_F16_TEXTURES",()=>!1),sL.registerFlag("WEBGL_PACK",()=>sL.getBool("HAS_WEBGL")),sL.registerFlag("WEBGL_PACK_NORMALIZATION",()=>sL.getBool("WEBGL_PACK")),sL.registerFlag("WEBGL_PACK_CLIP",()=>sL.getBool("WEBGL_PACK")),sL.registerFlag("WEBGL_PACK_DEPTHWISECONV",()=>sL.getBool("WEBGL_PACK")),sL.registerFlag("WEBGL_PACK_BINARY_OPERATIONS",()=>sL.getBool("WEBGL_PACK")),sL.registerFlag("WEBGL_PACK_UNARY_OPERATIONS",()=>sL.getBool("WEBGL_PACK")),sL.registerFlag("WEBGL_PACK_ARRAY_OPERATIONS",()=>sL.getBool("WEBGL_PACK")),sL.registerFlag("WEBGL_PACK_IMAGE_OPERATIONS",()=>sL.getBool("WEBGL_PACK")),sL.registerFlag("WEBGL_PACK_REDUCE",()=>sL.getBool("WEBGL_PACK")),sL.registerFlag("WEBGL_LAZILY_UNPACK",()=>sL.getBool("WEBGL_PACK")),sL.registerFlag("WEBGL_CONV_IM2COL",()=>sL.getBool("WEBGL_PACK")),sL.registerFlag("WEBGL_PACK_CONV2DTRANSPOSE",()=>sL.getBool("WEBGL_PACK")),sL.registerFlag("WEBGL_MAX_TEXTURE_SIZE",()=>qO(sL.getNumber("WEBGL_VERSION"))),sL.registerFlag("WEBGL_MAX_TEXTURES_IN_SHADER",()=>ZO(sL.getNumber("WEBGL_VERSION"))),sL.registerFlag("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION",()=>{let e=sL.getNumber("WEBGL_VERSION");return 0===e?0:YO(e)}),sL.registerFlag("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE",()=>sL.getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION")>0&&!Cs.isMobile()),sL.registerFlag("WEBGL_RENDER_FLOAT32_CAPABLE",()=>eL(sL.getNumber("WEBGL_VERSION"))),sL.registerFlag("WEBGL_RENDER_FLOAT32_ENABLED",()=>!sL.getBool("WEBGL_FORCE_F16_TEXTURES")&&sL.getBool("WEBGL_RENDER_FLOAT32_CAPABLE")),sL.registerFlag("WEBGL_DOWNLOAD_FLOAT_ENABLED",()=>tL(sL.getNumber("WEBGL_VERSION"))),sL.registerFlag("WEBGL_FENCE_API_ENABLED",()=>rL(sL.getNumber("WEBGL_VERSION"))),sL.registerFlag("WEBGL_SIZE_UPLOAD_UNIFORM",()=>sL.getBool("WEBGL_RENDER_FLOAT32_ENABLED")?4:0),sL.registerFlag("WEBGL_DELETE_TEXTURE_THRESHOLD",()=>-1,e=>{if("number"!=typeof e)throw new Error(`WEBGL_DELETE_TEXTURE_THRESHOLD must be a number but got ${e}.`);if(e<0&&-1!==e)throw new Error(`WEBGL_DELETE_TEXTURE_THRESHOLD must be -1 (indicating never delete) or at least 0, but got ${e}.`)}),sL.registerFlag("WEBGL_FLUSH_THRESHOLD",()=>Cs.isMobile()?1:-1,e=>{if("number"!=typeof e)throw new Error(`WEBGL_FLUSH_THRESHOLD must be a number but got ${e}.`);if(e<0&&-1!==e)throw new Error(`WEBGL_FLUSH_THRESHOLD must be -1 (indicating never manual flush) or at least 0, but got ${e}.`)}),sL.registerFlag("CPU_HANDOFF_SIZE_THRESHOLD",()=>128),sL.registerFlag("WEBGL_USE_SHAPES_UNIFORMS",()=>!1),sL.registerFlag("TOPK_LAST_DIM_CPU_HANDOFF_SIZE_THRESHOLD",()=>1e5),sL.registerFlag("TOPK_K_CPU_HANDOFF_THRESHOLD",()=>128),sL.registerFlag("WEBGL_EXP_CONV",()=>!1),sL.registerFlag("SOFTWARE_WEBGL_ENABLED",()=>sL.getBool("IS_TEST")),sL.registerFlag("WEBGL_MAX_SIZE_FOR_NARROW_TEXTURE",()=>1/0),sL.registerFlag("WEBGL_AUTO_SQUARIFY_NARROW_TEXTURE_SHAPE",()=>!1),sL.registerFlag("WEBGL2_ISNAN_CUSTOM",()=>!1),sL.registerFlag("ENGINE_COMPILE_ONLY",()=>!1);var dL="\n  const float FLOAT_MAX = 1.70141184e38;\n  const float FLOAT_MIN = 1.17549435e-38;\n\n  lowp vec4 encode_float(highp float v) {\n    if (isnan(v)) {\n      return vec4(255, 255, 255, 255);\n    }\n\n    highp float av = abs(v);\n\n    if(av < FLOAT_MIN) {\n      return vec4(0.0, 0.0, 0.0, 0.0);\n    } else if(v > FLOAT_MAX) {\n      return vec4(0.0, 0.0, 128.0, 127.0) / 255.0;\n    } else if(v < -FLOAT_MAX) {\n      return vec4(0.0, 0.0,  128.0, 255.0) / 255.0;\n    }\n\n    highp vec4 c = vec4(0,0,0,0);\n\n    highp float e = floor(log2(av));\n    highp float m = exp2(fract(log2(av))) - 1.0;\n\n    c[2] = floor(128.0 * m);\n    m -= c[2] / 128.0;\n    c[1] = floor(32768.0 * m);\n    m -= c[1] / 32768.0;\n    c[0] = floor(8388608.0 * m);\n\n    highp float ebias = e + 127.0;\n    c[3] = floor(ebias / 2.0);\n    ebias -= c[3] * 2.0;\n    c[2] += floor(ebias) * 128.0;\n\n    c[3] += 128.0 * step(0.0, -v);\n\n    return c / 255.0;\n  }\n",{getBroadcastDims:pL}=Gf;function cL(e,t,n){let r=[];if(e.forEach(e=>{let t=va.sizeFromShape(e.shapeInfo.logicalShape);if(e.shapeInfo.isUniform?r.push(`uniform float ${e.name}${t>1?`[${t}]`:""};`):(r.push(`uniform sampler2D ${e.name};`),r.push(`uniform int offset${e.name};`)),n.enableShapeUniforms){let{uniformShape:t}=SL(n.packedInputs,e.shapeInfo.logicalShape,e.shapeInfo.texShape);switch(t.length){case 1:r.push(`uniform int ${e.name}Shape;`);break;case 2:r.push(`uniform ivec2 ${e.name}Shape;`);break;case 3:r.push(`uniform ivec3 ${e.name}Shape;`);break;case 4:r.push(`uniform ivec4 ${e.name}Shape;`)}r.push(`uniform ivec2 ${e.name}TexShape;`)}}),n.enableShapeUniforms){switch(t.logicalShape.length){case 1:r.push("uniform int outShape;");break;case 2:r.push("uniform ivec2 outShape;"),r.push("uniform int outShapeStrides;");break;case 3:r.push("uniform ivec3 outShape;"),r.push("uniform ivec2 outShapeStrides;");break;case 4:r.push("uniform ivec4 outShape;"),r.push("uniform ivec3 outShapeStrides;")}r.push("uniform ivec2 outTexShape;")}n.customUniforms&&n.customUniforms.forEach(e=>{r.push(`uniform ${e.type} ${e.name}${e.arrayIndex?`[${e.arrayIndex}]`:""};`)});let a,s,i=r.join("\n"),o=e.map(e=>function(e,t,n=!1,r){let a="";a+=n?mL(e,r):fL(e,r);let s=e.shapeInfo.logicalShape,i=t.logicalShape;return s.length<=i.length&&(a+=n?function(e,t){let n,r=e.name,a=r.charAt(0).toUpperCase()+r.slice(1),s="get"+a+"AtOutCoords",i=e.shapeInfo.logicalShape.length,o=t.logicalShape.length,l=pL(e.shapeInfo.logicalShape,t.logicalShape),u=IL(o),h=o-i,d=["x","y","z","w","u","v"];n=0===i?"":o<2&&l.length>=1?"coords = 0;":l.map(e=>`coords.${d[e+h]} = 0;`).join("\n");let p="";p=o<2&&i>0?"coords":e.shapeInfo.logicalShape.map((e,t)=>`coords.${d[t+h]}`).join(", ");let c="return outputValue;",f=1===va.sizeFromShape(e.shapeInfo.logicalShape),m=1===va.sizeFromShape(t.logicalShape);if(1!==i||f||m){if(f&&!m)c=1===o?"\n        return vec4(outputValue.x, outputValue.x, 0., 0.);\n      ":"\n        return vec4(outputValue.x);\n      ";else if(l.length){let e=i-2,t=i-1;l.indexOf(e)>-1&&l.indexOf(t)>-1?c="return vec4(outputValue.x);":l.indexOf(e)>-1?c="return vec4(outputValue.x, outputValue.y, outputValue.x, outputValue.y);":l.indexOf(t)>-1&&(c="return vec4(outputValue.xx, outputValue.zz);")}}else c="\n      return vec4(outputValue.xy, outputValue.xy);\n    ";return`\n    vec4 ${s}() {\n      ${u} coords = getOutputCoords();\n      ${n}\n      vec4 outputValue = get${a}(${p});\n      ${c}\n    }\n  `}(e,t):function(e,t){let n=e.name,r=n.charAt(0).toUpperCase()+n.slice(1),a="get"+r+"AtOutCoords",s=t.texShape,i=e.shapeInfo.texShape,o=e.shapeInfo.logicalShape.length,l=t.logicalShape.length;if(!e.shapeInfo.isUniform&&o===l&&null==e.shapeInfo.flatOffset&&va.arraysEqual(i,s))return`\n      float ${a}() {\n        return sampleTexture(${n}, resultUV);\n      }\n    `;let u,h=IL(l),d=pL(e.shapeInfo.logicalShape,t.logicalShape),p=l-o,c=["x","y","z","w","u","v"];u=0===o?"":l<2&&d.length>=1?"coords = 0;":d.map(e=>`coords.${c[e+p]} = 0;`).join("\n");let f="";return f=l<2&&o>0?"coords":e.shapeInfo.logicalShape.map((e,t)=>`coords.${c[t+p]}`).join(", "),`\n    float ${a}() {\n      ${h} coords = getOutputCoords();\n      ${u}\n      return get${r}(${f});\n    }\n  `}(e,t)),a}(e,t,n.packedInputs,n.enableShapeUniforms)).join("\n"),l=t.texShape,u=iL(),h=function(e){return`\n    float sampleTexture(sampler2D textureSampler, vec2 uv) {\n      return ${e.texture2D}(textureSampler, uv).r;\n    }\n  `}(u),d=function(e){return`${e.version}\n    precision highp float;\n    precision highp int;\n    precision highp sampler2D;\n    ${e.varyingFs} vec2 resultUV;\n    ${e.defineOutput}\n    const vec2 halfCR = vec2(0.5, 0.5);\n\n    struct ivec5\n    {\n      int x;\n      int y;\n      int z;\n      int w;\n      int u;\n    };\n\n    struct ivec6\n    {\n      int x;\n      int y;\n      int z;\n      int w;\n      int u;\n      int v;\n    };\n\n    uniform float NAN;\n    ${e.defineSpecialNaN}\n    ${e.defineSpecialInf}\n    ${e.defineRound}\n\n    int imod(int x, int y) {\n      return x - y * (x / y);\n    }\n\n    int idiv(int a, int b, float sign) {\n      int res = a / b;\n      int mod = imod(a, b);\n      if (sign < 0. && mod != 0) {\n        res -= 1;\n      }\n      return res;\n    }\n\n    //Based on the work of Dave Hoskins\n    //https://www.shadertoy.com/view/4djSRW\n    #define HASHSCALE1 443.8975\n    float random(float seed){\n      vec2 p = resultUV * seed;\n      vec3 p3  = fract(vec3(p.xyx) * HASHSCALE1);\n      p3 += dot(p3, p3.yzx + 19.19);\n      return fract((p3.x + p3.y) * p3.z);\n    }\n\n    ${gL}\n    ${yL}\n    ${bL}\n  `}(u);return t.isPacked?(a=function(e,t,n){switch(e.length){case 0:return vL();case 1:return function(e,t,n){let r=[Math.ceil(t[0]/2),Math.ceil(t[1]/2)];return 1===r[0]?n?"\n      int getOutputCoords() {\n        return 2 * int(resultUV.x * ceil(float(outTexShape[1]) / 2.0));\n      }\n    ":`\n      int getOutputCoords() {\n        return 2 * int(resultUV.x * ${r[1]}.0);\n      }\n    `:1===r[1]?n?"\n      int getOutputCoords() {\n        return 2 * int(resultUV.y * ceil(float(outTexShape[0]) / 2.0));\n      }\n    ":`\n      int getOutputCoords() {\n        return 2 * int(resultUV.y * ${r[0]}.0);\n      }\n    `:n?"\n    int getOutputCoords() {\n      ivec2 packedTexShape = ivec2(ceil(float(outTexShape[0]) / 2.0), ceil(float(outTexShape[1]) / 2.0));\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(packedTexShape[0], packedTexShape[1]));\n      return 2 * (resTexRC.x * packedTexShape[1] + resTexRC.y);\n    }\n  ":`\n    int getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(${r[0]}, ${r[1]}));\n      return 2 * (resTexRC.x * ${r[1]} + resTexRC.y);\n    }\n  `}(0,t,n);case 2:return function(e,t,n){let r=[Math.ceil(t[0]/2),Math.ceil(t[1]/2)];if(va.arraysEqual(e,t))return n?"\n      ivec2 getOutputCoords() {\n        ivec2 packedTexShape = ivec2(ceil(float(outTexShape[0]) / 2.0), ceil(float(outTexShape[1]) / 2.0));\n        return 2 * ivec2(resultUV.yx * vec2(packedTexShape[0], packedTexShape[1]));\n      }\n    ":`\n      ivec2 getOutputCoords() {\n        return 2 * ivec2(resultUV.yx * vec2(${r[0]}, ${r[1]}));\n      }\n    `;let a=Math.ceil(e[1]/2);return n?"\n    ivec2 getOutputCoords() {\n      ivec2 packedTexShape = ivec2(ceil(float(outTexShape[0]) / 2.0), ceil(float(outTexShape[1]) / 2.0));\n      int texelsInLogicalRow = int(ceil(float(outShape[1]) / 2.0));\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(packedTexShape[0], packedTexShape[1]));\n\n      int index = resTexRC.x * packedTexShape[1] + resTexRC.y;\n      int r = 2 * (index / texelsInLogicalRow);\n      int c = imod(index, texelsInLogicalRow) * 2;\n\n      return ivec2(r, c);\n    }\n  ":`\n    ivec2 getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(${r[0]}, ${r[1]}));\n\n      int index = resTexRC.x * ${r[1]} + resTexRC.y;\n      int r = 2 * (index / ${a});\n      int c = imod(index, ${a}) * 2;\n\n      return ivec2(r, c);\n    }\n  `}(e,t,n);case 3:return function(e,t,n){if(n)return"\n    ivec3 getOutputCoords() {\n      ivec2 packedTexShape = ivec2(ceil(float(outTexShape[0]) / 2.0), ceil(float(outTexShape[1]) / 2.0));\n      int texelsInLogicalRow = int(ceil(float(outShape[2]) / 2.0));\n      int texelsInBatch = texelsInLogicalRow * int(ceil(float(outShape[1]) / 2.0));\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(packedTexShape[0], packedTexShape[1]));\n      int index = resTexRC.x * packedTexShape[1] + resTexRC.y;\n\n      int b = index / texelsInBatch;\n      index -= b * texelsInBatch;\n\n      int r = 2 * (index / texelsInLogicalRow);\n      int c = imod(index, texelsInLogicalRow) * 2;\n\n      return ivec3(b, r, c);\n    }\n  ";let r=[Math.ceil(t[0]/2),Math.ceil(t[1]/2)],a=Math.ceil(e[2]/2),s=a*Math.ceil(e[1]/2);return`\n    ivec3 getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(${r[0]}, ${r[1]}));\n      int index = resTexRC.x * ${r[1]} + resTexRC.y;\n\n      int b = index / ${s};\n      index -= b * ${s};\n\n      int r = 2 * (index / ${a});\n      int c = imod(index, ${a}) * 2;\n\n      return ivec3(b, r, c);\n    }\n  `}(e,t,n);default:return function(e,t,n){if(n)return"\n    ivec4 getOutputCoords() {\n      ivec2 packedTexShape = ivec2(ceil(float(outTexShape[0]) / 2.0), ceil(float(outTexShape[1]) / 2.0));\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(packedTexShape[0], packedTexShape[1]));\n      int index = resTexRC.x * packedTexShape[1] + resTexRC.y;\n\n      int texelsInLogicalRow = int(ceil(float(outShape[3]) / 2.0));\n      int texelsInBatch = texelsInLogicalRow * int(ceil(float(outShape[2]) / 2.0));\n      int texelsInBatchN = texelsInBatch * outShape[1];\n\n      int b2 = index / texelsInBatchN;\n      index -= b2 * texelsInBatchN;\n\n      int b = index / texelsInBatch;\n      index -= b * texelsInBatch;\n\n      int r = 2 * (index / texelsInLogicalRow);\n      int c = imod(index, texelsInLogicalRow) * 2;\n\n      return ivec4(b2, b, r, c);\n    }\n  ";let r=[Math.ceil(t[0]/2),Math.ceil(t[1]/2)],a=Math.ceil(e[e.length-1]/2),s=a*Math.ceil(e[e.length-2]/2),i=s,o="",l="b, r, c";for(let t=2;t<e.length-1;t++)i*=e[e.length-t-1],o=`\n      int b${t} = index / ${i};\n      index -= b${t} * ${i};\n    `+o,l=`b${t}, `+l;return`\n    ivec${e.length} getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(${r[0]}, ${r[1]}));\n      int index = resTexRC.x * ${r[1]} + resTexRC.y;\n\n      ${o}\n\n      int b = index / ${s};\n      index -= b * ${s};\n\n      int r = 2 * (index / ${a});\n      int c = imod(index, ${a}) * 2;\n\n      return ivec${e.length}(${l});\n    }\n  `}(e,t,n)}}(t.logicalShape,l,n.enableShapeUniforms),s=function(e){return`\n    void setOutput(vec4 val) {\n      ${e.output} = val;\n    }\n  `}(u)):(a=function(e,t,n){switch(e.length){case 0:return vL();case 1:return function(e,t,n){return 1===t[0]?n?"\n      int getOutputCoords() {\n        return int(resultUV.x * float(outTexShape[1]));\n      }\n    ":`\n      int getOutputCoords() {\n        return int(resultUV.x * ${t[1]}.0);\n      }\n    `:1===t[1]?n?"\n      int getOutputCoords() {\n        return int(resultUV.y * float(outTexShape[0]));\n      }\n    ":`\n      int getOutputCoords() {\n        return int(resultUV.y * ${t[0]}.0);\n      }\n    `:n?"\n    int getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(outTexShape[0], outTexShape[1]));\n      return resTexRC.x * outTexShape[1] + resTexRC.y;\n    }\n  ":`\n    int getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(${t[0]}, ${t[1]}));\n      return resTexRC.x * ${t[1]} + resTexRC.y;\n    }\n  `}(0,t,n);case 2:return function(e,t,n){return va.arraysEqual(e,t)?n?"\n      ivec2 getOutputCoords() {\n        return ivec2(resultUV.yx * vec2(outTexShape[0], outTexShape[1]));\n      }\n    ":`\n      ivec2 getOutputCoords() {\n        return ivec2(resultUV.yx * vec2(${t[0]}, ${t[1]}));\n      }\n    `:1===e[1]?n?"\n      ivec2 getOutputCoords() {\n        ivec2 resTexRC = ivec2(resultUV.yx *\n                               vec2(outTexShape[0], outTexShape[1]));\n        int index = resTexRC.x * outTexShape[1] + resTexRC.y;\n        return ivec2(index, 0);\n      }\n    ":`\n      ivec2 getOutputCoords() {\n        ivec2 resTexRC = ivec2(resultUV.yx *\n                               vec2(${t[0]}, ${t[1]}));\n        int index = resTexRC.x * ${t[1]} + resTexRC.y;\n        return ivec2(index, 0);\n      }\n    `:1===e[0]?n?"\n      ivec2 getOutputCoords() {\n        ivec2 resTexRC = ivec2(resultUV.yx *\n                               vec2(outTexShape[0], outTexShape[1]));\n        int index = resTexRC.x * outTexShape[1] + resTexRC.y;\n        return ivec2(0, index);\n      }\n    ":`\n      ivec2 getOutputCoords() {\n        ivec2 resTexRC = ivec2(resultUV.yx *\n                               vec2(${t[0]}, ${t[1]}));\n        int index = resTexRC.x * ${t[1]} + resTexRC.y;\n        return ivec2(0, index);\n      }\n    `:n?"\n    ivec2 getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(outTexShape[0], outTexShape[1]));\n      int index = resTexRC.x * outTexShape[1] + resTexRC.y;\n      int r = index / outShape[1];\n      int c = index - r * outShape[1];\n      return ivec2(r, c);\n    }\n  ":`\n    ivec2 getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(${t[0]}, ${t[1]}));\n      int index = resTexRC.x * ${t[1]} + resTexRC.y;\n      int r = index / ${e[1]};\n      int c = index - r * ${e[1]};\n      return ivec2(r, c);\n    }\n  `}(e,t,n);case 3:return function(e,t,n){if(n)return`\n  ivec3 getOutputCoords() {\n    ivec2 resTexRC = ivec2(resultUV.yx *\n                           vec2(outTexShape[0], outTexShape[1]));\n    int index = resTexRC.x * outTexShape[1] + resTexRC.y;\n    ${lL(["r","c","d"],e)}\n    return ivec3(r, c, d);\n  }\n`;let r=oL(["r","c","d"],e);return`\n    ivec3 getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx *\n                             vec2(${t[0]}, ${t[1]}));\n      int index = resTexRC.x * ${t[1]} + resTexRC.y;\n      ${r}\n      return ivec3(r, c, d);\n    }\n  `}(e,t,n);case 4:return function(e,t,n){if(n)return`\n    ivec4 getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx *\n        vec2(outTexShape[0], outTexShape[1]));\n      int index = resTexRC.x * outTexShape[1] + resTexRC.y;\n      ${lL(["r","c","d","d2"],e)}\n      return ivec4(r, c, d, d2);\n    }\n  `;let r=oL(["r","c","d","d2"],e);return`\n    ivec4 getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx *\n        vec2(${t[0]}, ${t[1]}));\n      int index = resTexRC.x * ${t[1]} + resTexRC.y;\n      ${r}\n      return ivec4(r, c, d, d2);\n    }\n  `}(e,t,n);case 5:return function(e,t){let n=oL(["r","c","d","d2","d3"],e);return`\n    ivec5 getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx * vec2(${t[0]},\n                             ${t[1]}));\n\n      int index = resTexRC.x * ${t[1]} + resTexRC.y;\n\n      ${n}\n\n      ivec5 outShape = ivec5(r, c, d, d2, d3);\n      return outShape;\n    }\n  `}(e,t);case 6:return function(e,t){let n=oL(["r","c","d","d2","d3","d4"],e);return`\n    ivec6 getOutputCoords() {\n      ivec2 resTexRC = ivec2(resultUV.yx *\n        vec2(${t[0]}, ${t[1]}));\n      int index = resTexRC.x * ${t[1]} + resTexRC.y;\n\n      ${n}\n\n      ivec6 result = ivec6(r, c, d, d2, d3, d4);\n      return result;\n    }\n  `}(e,t);default:throw new Error(`${e.length}-D output sampling is not yet supported`)}}(t.logicalShape,l,n.enableShapeUniforms),s=function(e){return`\n    void setOutput(float val) {\n      ${e.output} = vec4(val, 0, 0, 0);\n    }\n  `}(u)),n.packedInputs&&(d+=xL),[d,h,s,i,a,o,n.userCode].join("\n")}function fL(e,t=!1){let n=e.shapeInfo.logicalShape;switch(n.length){case 0:return function(e,t){let n=e.name,r="get"+n.charAt(0).toUpperCase()+n.slice(1);if(e.shapeInfo.isUniform)return`float ${r}() {return ${n};}`;let[a,s]=e.shapeInfo.texShape;if(1===a&&1===s)return`\n      float ${r}() {\n        return sampleTexture(${n}, halfCR);\n      }\n    `;let i=wL(n);if(t)return`\n    float ${r}() {\n      vec2 uv = uvFromFlat(${n}TexShape[0], ${n}TexShape[1], ${i});\n      return sampleTexture(${n}, uv);\n    }\n  `;let[o,l]=e.shapeInfo.texShape;return`\n    float ${r}() {\n      vec2 uv = uvFromFlat(${o}, ${l}, ${i});\n      return sampleTexture(${n}, uv);\n    }\n  `}(e,t);case 1:return function(e,t){let n=e.name,r="get"+n.charAt(0).toUpperCase()+n.slice(1);if(e.shapeInfo.isUniform)return`\n      float ${r}(int index) {\n        ${kL(e)}\n      }\n    `;let a=e.shapeInfo.texShape,s=a[0],i=a[1];if(1===i&&1===s)return`\n      float ${r}(int index) {\n        return sampleTexture(${n}, halfCR);\n      }\n    `;let o=wL(n);return 1===i?t?`\n      float ${r}(int index) {\n        vec2 uv = vec2(0.5, (float(index + ${o}) + 0.5) / float(${n}TexShape[0]));\n        return sampleTexture(${n}, uv);\n      }\n    `:`\n      float ${r}(int index) {\n        vec2 uv = vec2(0.5, (float(index + ${o}) + 0.5) / ${s}.0);\n        return sampleTexture(${n}, uv);\n      }\n    `:1===s?t?`\n      float ${r}(int index) {\n        vec2 uv = vec2((float(index + ${o}) + 0.5) / float(${n}TexShape[1]), 0.5);\n        return sampleTexture(${n}, uv);\n      }\n    `:`\n      float ${r}(int index) {\n        vec2 uv = vec2((float(index + ${o}) + 0.5) / ${i}.0, 0.5);\n        return sampleTexture(${n}, uv);\n      }\n    `:t?`\n    float ${r}(int index) {\n      vec2 uv = uvFromFlat(${n}TexShape[0], ${n}TexShape[1], index + ${o});\n      return sampleTexture(${n}, uv);\n    }\n  `:`\n    float ${r}(int index) {\n      vec2 uv = uvFromFlat(${s}, ${i}, index + ${o});\n      return sampleTexture(${n}, uv);\n    }\n  `}(e,t);case 2:return function(e,t){let n=e.shapeInfo.logicalShape,r=e.name,a="get"+r.charAt(0).toUpperCase()+r.slice(1),s=e.shapeInfo.texShape;if(null!=s&&va.arraysEqual(n,s)){if(t)return`\n      float ${a}(int row, int col) {\n        vec2 uv = (vec2(col, row) + halfCR) / vec2(${r}TexShape[1], ${r}TexShape[0]);\n        return sampleTexture(${r}, uv);\n      }\n    `;let e=s[0];return`\n    float ${a}(int row, int col) {\n      vec2 uv = (vec2(col, row) + halfCR) / vec2(${s[1]}.0, ${e}.0);\n      return sampleTexture(${r}, uv);\n    }\n  `}let{newShape:i,keptDims:o}=va.squeezeShape(n),l=i;if(l.length<n.length){let n=["row","col"];return`\n      ${fL(_L(e,l),t)}\n      float ${a}(int row, int col) {\n        return ${a}(${NL(n,o)});\n      }\n    `}if(e.shapeInfo.isUniform)return`\n      float ${a}(int row, int col) {\n        int index = round(dot(vec2(row, col), vec2(${n[1]}, 1)));\n        ${kL(e)}\n      }\n    `;let u=s[0],h=s[1],d=wL(r);return 1===h?t?`\n      float ${a}(int row, int col) {\n        float index = dot(vec3(row, col, ${d}), vec3(${r}Shape[1], 1, 1));\n        vec2 uv = vec2(0.5, (index + 0.5) / float(${r}TexShape[0]));\n        return sampleTexture(${r}, uv);\n      }\n    `:`\n    float ${a}(int row, int col) {\n      float index = dot(vec3(row, col, ${d}), vec3(${n[1]}, 1, 1));\n      vec2 uv = vec2(0.5, (index + 0.5) / ${u}.0);\n      return sampleTexture(${r}, uv);\n    }\n  `:1===u?t?`\n      float ${a}(int row, int col) {\n        float index = dot(vec3(row, col, ${d}), vec3(${r}Shape[1], 1, 1));\n        vec2 uv = vec2((index + 0.5) / float(${r}TexShape[1]), 0.5);\n        return sampleTexture(${r}, uv);\n      }\n    `:`\n    float ${a}(int row, int col) {\n      float index = dot(vec3(row, col, ${d}), vec3(${n[1]}, 1, 1));\n      vec2 uv = vec2((index + 0.5) / ${h}.0, 0.5);\n      return sampleTexture(${r}, uv);\n    }\n  `:t?`\n      float ${a}(int row, int col) {\n        // Explicitly use integer operations as dot() only works on floats.\n        int index = row * ${r}Shape[1] + col + ${d};\n        vec2 uv = uvFromFlat(${r}TexShape[0], ${r}TexShape[1], index);\n        return sampleTexture(${r}, uv);\n      }\n    `:`\n  float ${a}(int row, int col) {\n    // Explicitly use integer operations as dot() only works on floats.\n    int index = row * ${n[1]} + col + ${d};\n    vec2 uv = uvFromFlat(${u}, ${h}, index);\n    return sampleTexture(${r}, uv);\n  }\n`}(e,t);case 3:return function(e,t){let n=e.shapeInfo.logicalShape,r=e.name,a="get"+r.charAt(0).toUpperCase()+r.slice(1),s=n[1]*n[2],i=n[2],{newShape:o,keptDims:l}=va.squeezeShape(n),u=o;if(u.length<n.length){let n=["row","col","depth"];return`\n        ${fL(_L(e,u),t)}\n        float ${a}(int row, int col, int depth) {\n          return ${a}(${NL(n,l)});\n        }\n      `}if(e.shapeInfo.isUniform)return`\n      float ${a}(int row, int col, int depth) {\n        int index = round(dot(vec3(row, col, depth),\n                          vec3(${s}, ${i}, 1)));\n        ${kL(e)}\n      }\n    `;let h=e.shapeInfo.texShape,d=h[0],p=h[1],c=e.shapeInfo.flatOffset;if(p===s&&null==c)return t?`\n      float ${a}(int row, int col, int depth) {\n        int stride1 = ${r}Shape[2];\n        float texR = float(row);\n        float texC = dot(vec2(col, depth), vec2(stride1, 1));\n        vec2 uv = (vec2(texC, texR) + halfCR) /\n                   vec2(${r}TexShape[1], ${r}TexShape[0]);\n        return sampleTexture(${r}, uv);\n      }\n    `:`\n        float ${a}(int row, int col, int depth) {\n          float texR = float(row);\n          float texC = dot(vec2(col, depth), vec2(${i}, 1));\n          vec2 uv = (vec2(texC, texR) + halfCR) /\n                     vec2(${p}.0, ${d}.0);\n          return sampleTexture(${r}, uv);\n        }\n      `;if(p===i&&null==c)return t?`\n      float ${a}(int row, int col, int depth) {\n        float texR = dot(vec2(row, col), vec2(${r}Shape[1], 1));\n        float texC = float(depth);\n        vec2 uv = (vec2(texC, texR) + halfCR) / vec2(${r}TexShape[1], ${r}TexShape[0]);\n        return sampleTexture(${r}, uv);\n      }\n    `:`\n    float ${a}(int row, int col, int depth) {\n      float texR = dot(vec2(row, col), vec2(${n[1]}, 1));\n      float texC = float(depth);\n      vec2 uv = (vec2(texC, texR) + halfCR) / vec2(${p}.0, ${d}.0);\n      return sampleTexture(${r}, uv);\n    }\n  `;let f=wL(r);return t?`\n    float ${a}(int row, int col, int depth) {\n      // Explicitly use integer operations as dot() only works on floats.\n      int stride0 = ${r}Shape[1] * ${r}Shape[2];\n      int stride1 = ${r}Shape[2];\n      int index = row * stride0 + col * stride1 + depth + ${f};\n      vec2 uv = uvFromFlat(${r}TexShape[0], ${r}TexShape[1], index);\n      return sampleTexture(${r}, uv);\n    }\n    `:`\n      float ${a}(int row, int col, int depth) {\n        // Explicitly use integer operations as dot() only works on floats.\n        int index = row * ${s} + col * ${i} + depth + ${f};\n        vec2 uv = uvFromFlat(${d}, ${p}, index);\n        return sampleTexture(${r}, uv);\n      }\n  `}(e,t);case 4:return function(e,t){let n=e.shapeInfo.logicalShape,r=e.name,a="get"+r.charAt(0).toUpperCase()+r.slice(1),s=n[3],i=n[2]*s,o=n[1]*i,{newShape:l,keptDims:u}=va.squeezeShape(n);if(l.length<n.length){let n=["row","col","depth","depth2"];return`\n      ${fL(_L(e,l),t)}\n      float ${a}(int row, int col, int depth, int depth2) {\n        return ${a}(${NL(n,u)});\n      }\n    `}if(e.shapeInfo.isUniform)return`\n      float ${a}(int row, int col, int depth, int depth2) {\n        int index = round(dot(vec4(row, col, depth, depth2),\n                          vec4(${o}, ${i}, ${s}, 1)));\n        ${kL(e)}\n      }\n    `;let h=e.shapeInfo.flatOffset,d=e.shapeInfo.texShape,p=d[0],c=d[1],f=`int stride2 = ${r}Shape[3];`,m=`int stride1 = ${r}Shape[2] * stride2;`,g=`int stride0 = ${r}Shape[1] * stride1;`;if(c===o&&null==h)return t?`\n      float ${a}(int row, int col, int depth, int depth2) {\n        ${f}\n        ${m}\n        float texR = float(row);\n        float texC =\n            dot(vec3(col, depth, depth2),\n                vec3(stride1, stride2, 1));\n        vec2 uv = (vec2(texC, texR) + halfCR) /\n                   vec2(${r}TexShape[1], ${r}TexShape[0]);\n        return sampleTexture(${r}, uv);\n      }\n    `:`\n      float ${a}(int row, int col, int depth, int depth2) {\n        float texR = float(row);\n        float texC =\n            dot(vec3(col, depth, depth2),\n                vec3(${i}, ${s}, 1));\n        vec2 uv = (vec2(texC, texR) + halfCR) /\n                   vec2(${c}.0, ${p}.0);\n        return sampleTexture(${r}, uv);\n      }\n    `;if(c===s&&null==h)return t?`\n      float ${a}(int row, int col, int depth, int depth2) {\n        float texR = dot(vec3(row, col, depth),\n                         vec3(${r}Shape[1] * ${r}Shape[2], ${r}Shape[2], 1));\n        float texC = float(depth2);\n        vec2 uv = (vec2(texC, texR) + halfCR) /\n                  vec2(${r}TexShape[1], ${r}TexShape[0]);\n        return sampleTexture(${r}, uv);\n      }\n    `:`\n      float ${a}(int row, int col, int depth, int depth2) {\n        float texR = dot(vec3(row, col, depth),\n                         vec3(${n[1]*n[2]}, ${n[2]}, 1));\n        float texC = float(depth2);\n        vec2 uv = (vec2(texC, texR) + halfCR) /\n                  vec2(${c}.0, ${p}.0);\n        return sampleTexture(${r}, uv);\n      }\n    `;let y=wL(r);return t?`\n    float ${a}(int row, int col, int depth, int depth2) {\n      // Explicitly use integer operations as dot() only works on floats.\n      ${f}\n      ${m}\n      ${g}\n      int index = row * stride0 + col * stride1 +\n          depth * stride2 + depth2;\n      vec2 uv = uvFromFlat(${r}TexShape[0], ${r}TexShape[1], index + ${y});\n      return sampleTexture(${r}, uv);\n    }\n  `:`\n    float ${a}(int row, int col, int depth, int depth2) {\n      // Explicitly use integer operations as dot() only works on floats.\n      int index = row * ${o} + col * ${i} +\n          depth * ${s} + depth2;\n      vec2 uv = uvFromFlat(${p}, ${c}, index + ${y});\n      return sampleTexture(${r}, uv);\n    }\n  `}(e,t);case 5:return function(e){let t=e.shapeInfo.logicalShape,n=e.name,r="get"+n.charAt(0).toUpperCase()+n.slice(1),a=t[4],s=t[3]*a,i=t[2]*s,o=t[1]*i,{newShape:l,keptDims:u}=va.squeezeShape(t);if(l.length<t.length){let t=["row","col","depth","depth2","depth3"];return`\n      ${fL(_L(e,l))}\n      float ${r}(int row, int col, int depth, int depth2, int depth3) {\n        return ${r}(${NL(t,u)});\n      }\n    `}if(e.shapeInfo.isUniform)return`\n      float ${r}(int row, int col, int depth, int depth2, int depth3) {\n        float index = dot(\n          vec4(row, col, depth, depth2),\n          vec4(${o}, ${i}, ${s}, ${a})) +\n          depth3;\n        ${kL(e)}\n      }\n    `;let h=e.shapeInfo.flatOffset,d=e.shapeInfo.texShape,p=d[0],c=d[1];if(c===o&&null==h)return`\n      float ${r}(int row, int col, int depth, int depth2, int depth3) {\n        int texR = row;\n        float texC = dot(vec4(col, depth, depth2, depth3),\n                         vec4(${i}, ${s}, ${a}, 1));\n        vec2 uv = (vec2(texC, texR) + halfCR) /\n                   vec2(${c}.0, ${p}.0);\n        return sampleTexture(${n}, uv);\n      }\n    `;if(c===a&&null==h)return`\n      float ${r}(int row, int col, int depth, int depth2, int depth3) {\n        float texR = dot(\n          vec4(row, col, depth, depth2),\n          vec4(${t[1]*t[2]*t[3]},\n               ${t[2]*t[3]}, ${t[3]}, 1));\n        int texC = depth3;\n        vec2 uv = (vec2(texC, texR) + halfCR) /\n                  vec2(${c}.0, ${p}.0);\n        return sampleTexture(${n}, uv);\n      }\n    `;let f=wL(n);return`\n    float ${r}(int row, int col, int depth, int depth2, int depth3) {\n      // Explicitly use integer operations as dot() only works on floats.\n      int index = row * ${o} + col * ${i} + depth * ${s} +\n          depth2 * ${a} + depth3 + ${f};\n      vec2 uv = uvFromFlat(${p}, ${c}, index);\n      return sampleTexture(${n}, uv);\n    }\n  `}(e);case 6:return function(e){let t=e.shapeInfo.logicalShape,n=e.name,r="get"+n.charAt(0).toUpperCase()+n.slice(1),{newShape:a,keptDims:s}=va.squeezeShape(t);if(a.length<t.length){let t=["row","col","depth","depth2","depth3","depth4"];return`\n      ${fL(_L(e,a))}\n      float ${r}(int row, int col, int depth,\n                    int depth2, int depth3, int depth4) {\n        return ${r}(${NL(t,s)});\n      }\n    `}let i=t[5],o=t[4]*i,l=t[3]*o,u=t[2]*l,h=t[1]*u;if(e.shapeInfo.isUniform)return`\n      float ${r}(int row, int col, int depth,\n                  int depth2, int depth3, int depth4) {\n        int index = round(dot(\n          vec4(row, col, depth, depth2),\n          vec4(${h}, ${u}, ${l}, ${o})) +\n          dot(\n            vec2(depth3, depth4),\n            vec2(${i}, 1)));\n        ${kL(e)}\n      }\n    `;let d=e.shapeInfo.flatOffset,p=e.shapeInfo.texShape,c=p[0],f=p[1];if(f===h&&null==d)return`\n      float ${r}(int row, int col, int depth,\n                    int depth2, int depth3, int depth4) {\n        int texR = row;\n        float texC = dot(vec4(col, depth, depth2, depth3),\n          vec4(${u}, ${l}, ${o}, ${i})) +\n               float(depth4);\n        vec2 uv = (vec2(texC, texR) + halfCR) /\n                   vec2(${f}.0, ${c}.0);\n        return sampleTexture(${n}, uv);\n      }\n    `;if(f===i&&null==d)return`\n      float ${r}(int row, int col, int depth,\n                    int depth2, int depth3, int depth4) {\n        float texR = dot(vec4(row, col, depth, depth2),\n          vec4(${t[1]*t[2]*t[3]*t[4]},\n               ${t[2]*t[3]*t[4]},\n               ${t[3]*t[4]},\n               ${t[4]})) + float(depth3);\n        int texC = depth4;\n        vec2 uv = (vec2(texC, texR) + halfCR) /\n                  vec2(${f}.0, ${c}.0);\n        return sampleTexture(${n}, uv);\n      }\n    `;let m=wL(n);return`\n    float ${r}(int row, int col, int depth,\n                  int depth2, int depth3, int depth4) {\n      // Explicitly use integer operations as dot() only works on floats.\n      int index = row * ${h} + col * ${u} + depth * ${l} +\n          depth2 * ${o} + depth3 * ${i} + depth4 + ${m};\n      vec2 uv = uvFromFlat(${c}, ${f}, index);\n      return sampleTexture(${n}, uv);\n    }\n  `}(e);default:throw new Error(`${n.length}-D input sampling is not yet supported`)}}function mL(e,t){switch(e.shapeInfo.logicalShape.length){case 0:return function(e){let t=e.name,n="get"+t.charAt(0).toUpperCase()+t.slice(1),r=iL();return`\n    vec4 ${n}() {\n      return ${r.texture2D}(${t}, halfCR);\n    }\n  `}(e);case 1:return function(e,t){let n=e.name,r="get"+n.charAt(0).toUpperCase()+n.slice(1),a=e.shapeInfo.texShape,s=iL();if(t)return`\n    vec4 ${r}(int index) {\n      ivec2 packedTexShape = ivec2(ceil(float(${n}TexShape[0]) / 2.0), ceil(float(${n}TexShape[1]) / 2.0));\n      vec2 uv = packedUVfrom1D(\n        packedTexShape[0], packedTexShape[1], index);\n      return ${s.texture2D}(${n}, uv);\n    }\n  `;let i=[Math.ceil(a[0]/2),Math.ceil(a[1]/2)];return`\n    vec4 ${r}(int index) {\n      vec2 uv = packedUVfrom1D(\n        ${i[0]}, ${i[1]}, index);\n      return ${s.texture2D}(${n}, uv);\n    }\n  `}(e,t);case 2:return function(e,t){let n=e.shapeInfo.logicalShape,r=e.name,a="get"+r.charAt(0).toUpperCase()+r.slice(1),s=e.shapeInfo.texShape,i=s[0],o=s[1],l=iL();if(null!=s&&va.arraysEqual(n,s))return t?`\n      vec4 ${a}(int row, int col) {\n        vec2 uv = (vec2(col, row) + halfCR) / vec2(${r}TexShape[1], ${r}TexShape[0]);\n\n        return ${l.texture2D}(${r}, uv);\n      }\n    `:`\n      vec4 ${a}(int row, int col) {\n        vec2 uv = (vec2(col, row) + halfCR) / vec2(${o}.0, ${i}.0);\n\n        return ${l.texture2D}(${r}, uv);\n      }\n    `;if(t)return`\n    vec4 ${a}(int row, int col) {\n      ivec2 packedTexShape = ivec2(ceil(float(${r}TexShape[0]) / 2.0), ceil(float(${r}TexShape[1]) / 2.0));\n      int valuesPerRow = int(ceil(float(${r}Shape[1]) / 2.0));\n      vec2 uv = packedUVfrom2D(valuesPerRow, packedTexShape[0], packedTexShape[1], row, col);\n      return ${l.texture2D}(${r}, uv);\n    }\n  `;let u=[Math.ceil(s[0]/2),Math.ceil(s[1]/2)],h=Math.ceil(n[1]/2);return`\n    vec4 ${a}(int row, int col) {\n      vec2 uv = packedUVfrom2D(${h}, ${u[0]}, ${u[1]}, row, col);\n      return ${l.texture2D}(${r}, uv);\n    }\n  `}(e,t);case 3:return function(e,t){let n=e.shapeInfo.logicalShape,r=e.name,a="get"+r.charAt(0).toUpperCase()+r.slice(1),s=e.shapeInfo.texShape,i=[Math.ceil(s[0]/2),Math.ceil(s[1]/2)];if(1===n[0]){let r=[1,2],s=["b","row","col"];return`\n        ${mL(_L(e,n.slice(1)),t)}\n        vec4 ${a}(int b, int row, int col) {\n          return ${a}(${NL(s,r)});\n        }\n      `}let o=iL();if(t)return`\n    vec4 ${a}(int b, int row, int col) {\n      ivec2 packedTexShape = ivec2(ceil(float(${r}TexShape[0]) / 2.0), ceil(float(${r}TexShape[1]) / 2.0));\n      int valuesPerRow = int(ceil(float(${r}Shape[2]) / 2.0));\n      int texelsInBatch = valuesPerRow * int(ceil(float(${r}Shape[1]) / 2.0));\n      vec2 uv = packedUVfrom3D(\n        packedTexShape[0], packedTexShape[1], texelsInBatch, valuesPerRow, b, row, col);\n      return ${o.texture2D}(${r}, uv);\n    }\n  `;let l=i[0],u=i[1],h=Math.ceil(n[2]/2),d=h*Math.ceil(n[1]/2);return`\n    vec4 ${a}(int b, int row, int col) {\n      vec2 uv = packedUVfrom3D(\n        ${l}, ${u}, ${d}, ${h}, b, row, col);\n      return ${o.texture2D}(${r}, uv);\n    }\n  `}(e,t);default:return function(e,t){let n=e.name,r="get"+n.charAt(0).toUpperCase()+n.slice(1),a=iL();if(t)return`\n    vec4 ${r}(int b2, int b, int row, int col) {\n      int valuesPerRow = int(ceil(float(${n}Shape[3]) / 2.0));\n      int texelsInBatch = valuesPerRow * int(ceil(float(${n}Shape[2]) / 2.0));\n      int index = b * texelsInBatch + (row / 2) * valuesPerRow + (col / 2);\n      texelsInBatch *= ${n}Shape[1];\n      index = b2 * texelsInBatch + index;\n      ivec2 packedTexShape = ivec2(ceil(float(${n}TexShape[0]) / 2.0), ceil(float(${n}TexShape[1]) / 2.0));\n      int texR = index / packedTexShape[1];\n      int texC = index - texR * packedTexShape[1];\n      vec2 uv = (vec2(texC, texR) + halfCR) / vec2(packedTexShape[1], packedTexShape[0]); return ${a.texture2D}(${n}, uv);\n    }\n  `;let s=e.shapeInfo.logicalShape,i=s.length,o=e.shapeInfo.texShape,l=[Math.ceil(o[0]/2),Math.ceil(o[1]/2)],u=l[0],h=l[1],d=Math.ceil(s[i-1]/2),p=d*Math.ceil(s[i-2]/2),c="int b, int row, int col",f=`b * ${p} + (row / 2) * ${d} + (col / 2)`;for(let e=2;e<i-1;e++)c=`int b${e}, `+c,p*=s[i-e-1],f=`b${e} * ${p} + `+f;return`\n    vec4 ${r}(${c}) {\n      int index = ${f};\n      int texR = index / ${h};\n      int texC = index - texR * ${h};\n      vec2 uv = (vec2(texC, texR) + halfCR) / vec2(${h}, ${u});\n      return ${a.texture2D}(${n}, uv);\n    }\n  `}(e,t)}}var gL="\nvec2 uvFromFlat(int texNumR, int texNumC, int index) {\n  int texR = index / texNumC;\n  int texC = index - texR * texNumC;\n  return (vec2(texC, texR) + halfCR) / vec2(texNumC, texNumR);\n}\nvec2 packedUVfrom1D(int texNumR, int texNumC, int index) {\n  int texelIndex = index / 2;\n  int texR = texelIndex / texNumC;\n  int texC = texelIndex - texR * texNumC;\n  return (vec2(texC, texR) + halfCR) / vec2(texNumC, texNumR);\n}\n",yL="\nvec2 packedUVfrom2D(int texelsInLogicalRow, int texNumR,\n  int texNumC, int row, int col) {\n  int texelIndex = (row / 2) * texelsInLogicalRow + (col / 2);\n  int texR = texelIndex / texNumC;\n  int texC = texelIndex - texR * texNumC;\n  return (vec2(texC, texR) + halfCR) / vec2(texNumC, texNumR);\n}\n",bL="\nvec2 packedUVfrom3D(int texNumR, int texNumC,\n    int texelsInBatch, int texelsInLogicalRow, int b,\n    int row, int col) {\n  int index = b * texelsInBatch + (row / 2) * texelsInLogicalRow + (col / 2);\n  int texR = index / texNumC;\n  int texC = index - texR * texNumC;\n  return (vec2(texC, texR) + halfCR) / vec2(texNumC, texNumR);\n}\n",xL="\n  float getChannel(vec4 frag, vec2 innerDims) {\n    vec2 modCoord = mod(innerDims, 2.);\n    return modCoord.x == 0. ?\n      (modCoord.y == 0. ? frag.r : frag.g) :\n      (modCoord.y == 0. ? frag.b : frag.a);\n  }\n  float getChannel(vec4 frag, int dim) {\n    float modCoord = mod(float(dim), 2.);\n    return modCoord == 0. ? frag.r : frag.g;\n  }\n";function vL(){return"\n    int getOutputCoords() {\n      return 0;\n    }\n  "}function wL(e){return`offset${e}`}function kL(e){let t=e.name,n=va.sizeFromShape(e.shapeInfo.logicalShape);return n<2?`return ${t};`:`\n    for (int i = 0; i < ${n}; i++) {\n      if (i == index) {\n        return ${t}[i];\n      }\n    }\n  `}function IL(e){if(e<=1)return"int";if(2===e)return"ivec2";if(3===e)return"ivec3";if(4===e)return"ivec4";if(5===e)return"ivec5";if(6===e)return"ivec6";throw Error(`GPU for rank ${e} is not yet supported`)}function SL(e,t,n){let{newShape:r,keptDims:a}=va.squeezeShape(t),s=t.length,i=e&&3===s&&1===t[0],o=i?t.slice(1):r,l=!e&&s>1&&!va.arraysEqual(t,n)&&r.length<s||i;return{useSqueezeShape:l,uniformShape:l?o:t,keptDims:a}}function _L(e,t){let n=JSON.parse(JSON.stringify(e));return n.shapeInfo.logicalShape=t,n}function NL(e,t){return t.map(t=>e[t]).join(", ")}function TL(e,t,n){let r,a,s,i=[],o=[],l=null,u=null;u=e.getUniformLocation(n,"NAN",!1),1===Le().getNumber("WEBGL_VERSION")&&(l=e.getUniformLocation(n,"INFINITY",!1));let h=!1;for(let r of t.variableNames){let a={name:r,uniform:e.getUniformLocation(n,r,h),offset:e.getUniformLocation(n,`offset${r}`,h)};t.enableShapeUniforms&&(a.shape=e.getUniformLocation(n,`${r}Shape`,h),a.texShape=e.getUniformLocation(n,`${r}TexShape`,h)),i.push(a)}if(t.enableShapeUniforms&&(r=e.getUniformLocation(n,"outShape",h),s=e.getUniformLocation(n,"outShapeStrides",h),a=e.getUniformLocation(n,"outTexShape",h)),t.customUniforms)for(let r of t.customUniforms)o.push(e.getUniformLocation(n,r.name,h));return{variablesLocations:i,customUniformLocations:o,infLoc:l,nanLoc:u,outShapeLocation:r,outShapeStridesLocation:s,outTexShapeLocation:a}}function CL(e,t){if(e.length!==t.length)throw Error(`Binary was compiled with ${e.length} inputs, but was executed with ${t.length} inputs`);e.forEach((e,n)=>{let r=e.logicalShape,a=t[n],s=a.shape;if(!va.arraysEqual(r,s))throw Error(`Binary was compiled with different shapes than the current args. Shapes ${r} and ${s} must match`);if(e.isUniform&&a.isUniform)return;let i=e.texShape,o=a.isUniform?null:a.texData.texShape;if(!va.arraysEqual(i,o))throw Error(`Binary was compiled with different texture shapes than the current args. Shape ${i} and ${o} must match`)})}function EL(e){return Le().getBool("WEBGL_USE_SHAPES_UNIFORMS")&&e<=4}var AL=class{constructor(e){this.variableNames=["A"],this.packedInputs=!1,this.packedOutput=!0,this.outPackingScheme=JM.DENSE,this.customUniforms=[{name:"texShape",type:"ivec2"}];let t=iL();this.outputShape=e,this.enableShapeUniforms=EL(this.outputShape.length),this.userCode=`\n      ivec3 outCoordsFromFlatIndex(int index) {\n        ${this.enableShapeUniforms?lL(["r","c","d"],e):oL(["r","c","d"],e)}\n        return ivec3(r, c, d);\n      }\n\n      void main() {\n        ivec2 resTexRC = ivec2(resultUV.yx * vec2(texShape[0], texShape[1]));\n        int index = 4 * (resTexRC.x * texShape[1] + resTexRC.y);\n\n        vec4 result = vec4(0.);\n\n        for (int i=0; i<4; i++) {\n          int flatIndex = index + i;\n          ivec3 rc = outCoordsFromFlatIndex(flatIndex);\n          result[i] = getA(rc.x, rc.y, rc.z);\n        }\n\n        ${t.output} = result;\n      }\n    `}},$L=class{constructor(e){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0,this.outPackingScheme=JM.DENSE,this.customUniforms=[{name:"texShape",type:"ivec2"}];let t=iL();this.outputShape=e,this.enableShapeUniforms=EL(this.outputShape.length),this.userCode=`\n      ivec3 outCoordsFromFlatIndex(int index) {\n        ${this.enableShapeUniforms?lL(["r","c","d"],e):oL(["r","c","d"],e)}\n        return ivec3(r, c, d);\n      }\n\n      void main() {\n        ivec2 resTexRC = ivec2(resultUV.yx * vec2(texShape[0], texShape[1]));\n        int index = 4 * (resTexRC.x * texShape[1] + resTexRC.y);\n\n        vec4 result = vec4(0.);\n\n        for (int i=0; i<4; i++) {\n          int flatIndex = index + i;\n          ivec3 rc = outCoordsFromFlatIndex(flatIndex);\n          result[i] = getChannel(getA(rc.x, rc.y, rc.z), vec2(rc.y, rc.z));\n        }\n\n        ${t.output} = result;\n      }\n    `}},RL=class{constructor(e){this.variableNames=["A"],this.outTexUsage=QM.DOWNLOAD;let t=iL();this.outputShape=e,this.userCode=`\n      ${dL}\n\n      void main() {\n        float x = getAAtOutCoords();\n        ${t.output} = encode_float(x);\n      }\n    `}},FL=class{constructor(e){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!1,this.outTexUsage=QM.DOWNLOAD;let t=iL();this.outputShape=e,this.userCode=`\n      ${dL}\n\n      void main() {\n        ivec3 coords = getOutputCoords();\n        float x = getChannel(getAAtOutCoords(), vec2(coords.y, coords.z));\n        ${t.output} = encode_float(x);\n      }\n    `}},DL={R:0,G:1,B:2,A:3},ML=class{constructor(e,t=!1,n="RGBA"){this.variableNames=["A"],this.customUniforms=[{name:"texShape",type:"ivec2"}];let r=iL();this.outputShape=e,this.enableShapeUniforms=EL(this.outputShape.length);let a="result";t&&(a="floor(result * 255. + 0.5)");let s="";for(let e=0;e<n.length;e++){let t=n[e];s+=`\n          if(offset == ${e}) {\n            result = values[${DL[t]}];\n          }`}this.userCode=`\n      ${this.enableShapeUniforms?"\n  int getFlatIndex(ivec3 coords) {\n    return coords.x * outShapeStrides[0] + coords.y * outShapeStrides[1] + coords.z;\n  }\n":hL(e)}\n\n      void main() {\n        ivec3 coords = getOutputCoords();\n        int flatIndex = getFlatIndex(coords);\n        float result = 0.;\n        int offset = imod(flatIndex, ${n.length});\n\n        flatIndex = idiv(flatIndex, ${n.length}, 1.);\n\n        int r = flatIndex / texShape[1];\n        if (r < texShape[0]) {\n          int c = imod(flatIndex, texShape[1]);\n          vec2 uv = (vec2(c, r) + halfCR) / vec2(texShape[1], texShape[0]);\n          vec4 values = ${r.texture2D}(A, uv);\n          ${s}\n        }\n        ${r.output} = vec4(${a}, 0., 0., 0.);\n      }\n    `}},OL=class{constructor(e,t=!1){this.variableNames=["A"],this.packedInputs=!1,this.packedOutput=!0,this.customUniforms=[{name:"texShape",type:"ivec2"}];let n=iL();this.outputShape=e,this.enableShapeUniforms=EL(this.outputShape.length);let r="",a="result";t&&(a="floor(result * 255. + 0.5)");for(let t=0;t<=1;t++)for(let a=0;a<=1;a++){let s=2*t+a;r+=`\n          localCoords = coords;\n          if(localCoords[2] + ${a} < ${this.enableShapeUniforms?"outShape[2]":`${e[2]}`}) {\n          localCoords[2] += ${a};\n          if (localCoords[1] + ${t} < ${this.enableShapeUniforms?"outShape[1]":`${e[1]}`}) {\n            localCoords[1] += ${t};\n\n            flatIndex = getFlatIndex(localCoords);\n            offset = imod(flatIndex, 4);\n\n            flatIndex = idiv(flatIndex, 4, 1.);\n\n            int r = flatIndex / texShape[1];\n            int c = imod(flatIndex, texShape[1]);\n            vec2 uv = (vec2(c, r) + halfCR) / vec2(texShape[1], texShape[0]);\n            values = ${n.texture2D}(A, uv);\n\n            if (offset == 0) {\n              result[${s}] = values[0];\n            } else if (offset == 1) {\n              result[${s}] = values[1];\n            } else if (offset == 2) {\n              result[${s}] = values[2];\n            } else {\n              result[${s}] = values[3];\n            }\n          }\n        }\n        `}this.userCode=`\n        ${this.enableShapeUniforms?"\n  int getFlatIndex(ivec3 coords) {\n    return coords.x * outShapeStrides[0] + coords.y * outShapeStrides[1] + coords.z;\n  }\n":hL(e)}\n\n        void main() {\n          ivec3 coords = getOutputCoords();\n\n          vec4 result = vec4(0.);\n          int flatIndex, r, c, offset;\n          ivec3 localCoords;\n          vec2 uv;\n          vec4 values;\n\n          ${r}\n\n          ${n.output} = ${a};\n        }\n    `}},LL={};function PL(e){let t=iL();return cO(e,`${t.version}\n    precision highp float;\n    ${t.attribute} vec3 clipSpacePos;\n    ${t.attribute} vec2 uv;\n    ${t.varyingVs} vec2 resultUV;\n\n    void main() {\n      gl_Position = vec4(clipSpacePos, 1);\n      resultUV = uv;\n    }`)}function zL(e){return kO(e,new Float32Array([-1,1,0,0,1,-1,-1,0,0,0,1,1,0,1,1,1,-1,0,1,0]))}function BL(e){return IO(e,new Uint16Array([0,1,2,2,1,3]))}function WL(e,t,n,r,a,s){NO(t,n);let i=_O(e),o=e.TEXTURE_2D;return uO(e,()=>e.bindTexture(o,i)),uO(e,()=>e.texParameteri(o,e.TEXTURE_WRAP_S,e.CLAMP_TO_EDGE)),uO(e,()=>e.texParameteri(o,e.TEXTURE_WRAP_T,e.CLAMP_TO_EDGE)),uO(e,()=>e.texParameteri(o,e.TEXTURE_MIN_FILTER,e.NEAREST)),uO(e,()=>e.texParameteri(o,e.TEXTURE_MAG_FILTER,e.NEAREST)),1===Le().getNumber("WEBGL_VERSION")?uO(e,()=>e.texImage2D(o,0,r,t,n,0,a,s,null)):uO(e,()=>e.texStorage2D(o,1,r,t,n)),uO(e,()=>e.bindTexture(e.TEXTURE_2D,null)),{texture:i,texShape:[n,t]}}function VL(e){return e.internalFormatFloat}function UL(e,t,n,r){let[a,s]=sO(t,n);return WL(e,a,s,VL(r),r.textureFormatFloat,e.FLOAT)}function GL(e){return e.internalFormatHalfFloat}function HL(e,t,n,r){let[a,s]=sO(t,n);return WL(e,a,s,GL(r),r.textureFormatFloat,r.textureTypeHalfFloat)}function jL(e){return e.downloadTextureFormat}function qL(e,t,n,r){let[a,s]=sO(t,n);return WL(e,a,s,jL(r),e.RGBA,e.UNSIGNED_BYTE)}function KL(e){return e.internalFormatPackedFloat}function XL(e,t,n,r){let[a,s]=oO(t,n);return WL(e,a,s,KL(r),e.RGBA,e.FLOAT)}function ZL(e){return e.internalFormatPackedHalfFloat}function YL(e,t,n,r){let[a,s]=oO(t,n);return WL(e,a,s,ZL(r),e.RGBA,r.textureTypeHalfFloat)}function JL(e,t,n){return uO(e,()=>e.bindBuffer(e.ARRAY_BUFFER,n)),CO(e,t,"clipSpacePos",n,3,20,0)&&CO(e,t,"uv",n,2,20,12)}function QL(e,t,n,r,a,s){let i,o,l;uO(e,()=>e.bindTexture(e.TEXTURE_2D,t)),a instanceof Uint8Array?(i=new Uint8Array(n*r*4),o=e.UNSIGNED_BYTE,l=e.RGBA):(i=new Float32Array(n*r*4),o=e.FLOAT,l=s.internalFormatPackedFloat),i.set(a),2===Le().getNumber("WEBGL_VERSION")?uO(e,()=>e.texSubImage2D(e.TEXTURE_2D,0,0,0,n,r,e.RGBA,o,i)):uO(e,()=>e.texImage2D(e.TEXTURE_2D,0,l,n,r,0,e.RGBA,o,i)),uO(e,()=>e.bindTexture(e.TEXTURE_2D,null))}function eP(e,t,n){uO(e,()=>e.bindTexture(e.TEXTURE_2D,t)),n.data instanceof Uint8Array?2===Le().getNumber("WEBGL_VERSION")?uO(e,()=>e.texSubImage2D(e.TEXTURE_2D,0,0,0,n.width,n.height,e.RGBA,e.UNSIGNED_BYTE,n.data)):uO(e,()=>e.texImage2D(e.TEXTURE_2D,0,e.RGBA,n.width,n.height,0,e.RGBA,e.UNSIGNED_BYTE,n.data)):2===Le().getNumber("WEBGL_VERSION")?uO(e,()=>e.texSubImage2D(e.TEXTURE_2D,0,0,0,e.RGBA,e.UNSIGNED_BYTE,n)):uO(e,()=>e.texImage2D(e.TEXTURE_2D,0,e.RGBA,e.RGBA,e.UNSIGNED_BYTE,n)),uO(e,()=>e.bindTexture(e.TEXTURE_2D,null))}function tP(e,t,n,r){let a=e.createBuffer();uO(e,()=>e.bindBuffer(e.PIXEL_PACK_BUFFER,a));let s=16*t*n;return uO(e,()=>e.bufferData(e.PIXEL_PACK_BUFFER,s,e.STREAM_READ)),uO(e,()=>e.readPixels(0,0,n,t,e.RGBA,e.FLOAT,0)),uO(e,()=>e.bindBuffer(e.PIXEL_PACK_BUFFER,null)),a}function nP(e,t,n){let r=e,a=new Float32Array(n);return r.bindBuffer(r.PIXEL_PACK_BUFFER,t),r.getBufferSubData(r.PIXEL_PACK_BUFFER,0,a),r.bindBuffer(r.PIXEL_PACK_BUFFER,null),a}function rP(e,t,n,r){let[a,s]=sO(t,n),i=new Uint8Array(function(e,t){return e*t}(t*n,4));return uO(e,()=>e.readPixels(0,0,a,s,r.downloadTextureFormat,e.UNSIGNED_BYTE,i)),new Float32Array(i.buffer)}function aP(e,t,n,r,a,s,i,o){let l=e,u=new Float32Array(function(e,t){let[n,r]=oO(e,t);return n*r*4}(s,i));return l.bindBuffer(l.PIXEL_PACK_BUFFER,t),l.getBufferSubData(l.PIXEL_PACK_BUFFER,0,u),l.bindBuffer(l.PIXEL_PACK_BUFFER,null),u}function sP(e,t,n){let r=new Float32Array(t*n*4);return uO(e,()=>e.readPixels(0,0,n,t,e.RGBA,e.FLOAT,r)),r}f(LL,{bindVertexProgramAttributeStreams:()=>JL,createBufferFromOutputTexture:()=>tP,createFloat16MatrixTexture:()=>HL,createFloat16PackedMatrixTexture:()=>YL,createFloat32MatrixTexture:()=>UL,createIndexBuffer:()=>BL,createPackedMatrixTexture:()=>XL,createUnsignedBytesMatrixTexture:()=>qL,createVertexBuffer:()=>zL,createVertexShader:()=>PL,downloadByteEncodedFloatMatrixFromOutputTexture:()=>rP,downloadFloat32MatrixFromBuffer:()=>nP,downloadMatrixFromPackedOutputTexture:()=>sP,downloadPackedMatrixFromBuffer:()=>aP,getInternalFormatForFloat16MatrixTexture:()=>GL,getInternalFormatForFloat16PackedMatrixTexture:()=>ZL,getInternalFormatForFloat32MatrixTexture:()=>VL,getInternalFormatForPackedMatrixTexture:()=>KL,getInternalFormatForUnsignedBytesMatrixTexture:()=>jL,uploadDenseMatrixToTexture:()=>QL,uploadPixelDataToTexture:()=>eP});var iP=class{constructor(e){this.outputTexture=null,this.program=null,this.disposed=!1,this.itemsToPoll=[];let t=Le().getNumber("WEBGL_VERSION");if(null!=e?(this.gl=e,rO(t,e)):this.gl=aO(t),e=this.gl,2===Le().getNumber("WEBGL_VERSION")){let t=e;this.createVertexArray=()=>uO(t,()=>t.createVertexArray()),this.bindVertexArray=e=>uO(t,()=>t.bindVertexArray(e)),this.deleteVertexArray=e=>uO(t,()=>t.deleteVertexArray(e)),this.getVertexArray=()=>uO(t,()=>t.getParameter(t.VERTEX_ARRAY_BINDING))}else if(null!=e){let t=e.getExtension("OES_vertex_array_object");if(null==t)throw new Error("All WebGL1 implementations are expected to offer OES_vertex_array_object.");this.createVertexArray=()=>uO(e,()=>t.createVertexArrayOES()),this.bindVertexArray=n=>uO(e,()=>t.bindVertexArrayOES(n)),this.deleteVertexArray=n=>uO(e,()=>t.deleteVertexArrayOES(n)),this.getVertexArray=()=>uO(e,()=>e.getParameter(t.VERTEX_ARRAY_BINDING_OES))}let n="WEBGL_color_buffer_float",r="EXT_color_buffer_half_float";if(this.parallelCompilationExtension=this.gl.getExtension("KHR_parallel_shader_compile"),1===Le().getNumber("WEBGL_VERSION")){let e="OES_texture_float",t="OES_texture_half_float";if(this.textureFloatExtension=pO(this.gl,e),JO(this.gl,t))this.textureHalfFloatExtension=pO(this.gl,t);else if(Le().get("WEBGL_FORCE_F16_TEXTURES"))throw new Error("GL context does not support half float textures, yet the environment flag WEBGL_FORCE_F16_TEXTURES is set to true.");if(this.colorBufferFloatExtension=this.gl.getExtension(n),JO(this.gl,r))this.colorBufferHalfFloatExtension=pO(this.gl,r);else if(Le().get("WEBGL_FORCE_F16_TEXTURES"))throw new Error("GL context does not support color renderable half floats, yet the environment flag WEBGL_FORCE_F16_TEXTURES is set to true.")}else if(n="EXT_color_buffer_float",JO(this.gl,n))this.colorBufferFloatExtension=this.gl.getExtension(n);else{if(!JO(this.gl,r))throw new Error("GL context does not support color renderable floats");this.colorBufferHalfFloatExtension=this.gl.getExtension(r)}this.vertexBuffer=zL(this.gl),this.indexBuffer=BL(this.gl),this.framebuffer=TO(this.gl),this.textureConfig=lO(this.gl,this.textureHalfFloatExtension)}get debug(){return Le().getBool("DEBUG")}dispose(){if(this.disposed)return;null!=this.program&&console.warn("Disposing a GPGPUContext that still has a bound WebGLProgram. This is probably a resource leak, delete the program with GPGPUContext.deleteProgram before disposing."),null!=this.outputTexture&&console.warn("Disposing a GPGPUContext that still has a bound output matrix texture.  This is probably a resource leak, delete the output matrix texture with GPGPUContext.deleteMatrixTexture before disposing.");let e=this.gl;uO(e,()=>e.finish()),uO(e,()=>e.bindFramebuffer(e.FRAMEBUFFER,null)),uO(e,()=>e.deleteFramebuffer(this.framebuffer)),uO(e,()=>e.bindBuffer(e.ARRAY_BUFFER,null)),uO(e,()=>e.bindBuffer(e.ELEMENT_ARRAY_BUFFER,null)),uO(e,()=>e.deleteBuffer(this.indexBuffer)),this.disposed=!0}createFloat32MatrixTexture(e,t){return this.throwIfDisposed(),UL(this.gl,e,t,this.textureConfig)}createFloat16MatrixTexture(e,t){return this.throwIfDisposed(),HL(this.gl,e,t,this.textureConfig)}createUnsignedBytesMatrixTexture(e,t){return this.throwIfDisposed(),qL(this.gl,e,t,this.textureConfig)}uploadPixelDataToTexture(e,t){this.throwIfDisposed(),eP(this.gl,e,t)}uploadDenseMatrixToTexture(e,t,n,r){this.throwIfDisposed(),QL(this.gl,e,t,n,r,this.textureConfig)}createFloat16PackedMatrixTexture(e,t){return this.throwIfDisposed(),YL(this.gl,e,t,this.textureConfig)}createPackedMatrixTexture(e,t){return this.throwIfDisposed(),XL(this.gl,e,t,this.textureConfig)}deleteMatrixTexture(e){this.throwIfDisposed(),this.outputTexture===e&&(OO(this.gl,this.framebuffer),this.outputTexture=null),uO(this.gl,()=>this.gl.deleteTexture(e))}downloadByteEncodedFloatMatrixFromOutputTexture(e,t,n){return this.downloadMatrixDriver(e,()=>rP(this.gl,t,n,this.textureConfig))}downloadPackedMatrixFromBuffer(e,t,n,r,a,s){return aP(this.gl,e,0,0,0,a,s,this.textureConfig)}downloadFloat32MatrixFromBuffer(e,t){return nP(this.gl,e,t)}createBufferFromTexture(e,t,n){this.bindTextureToFrameBuffer(e);let r=tP(this.gl,t,n,this.textureConfig);return this.unbindTextureToFrameBuffer(),r}createAndWaitForFence(){let e=this.createFence(this.gl);return this.pollFence(e)}createFence(e){let t,n;if(Le().getBool("WEBGL_FENCE_API_ENABLED")){let r=e,a=r.fenceSync(r.SYNC_GPU_COMMANDS_COMPLETE,0);e.flush(),n=()=>{let e=r.clientWaitSync(a,0,0);return e===r.ALREADY_SIGNALED||e===r.CONDITION_SATISFIED},t=a}else Le().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION")>0?(t=this.beginQuery(),this.endQuery(),n=()=>this.isQueryAvailable(t,Le().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION"))):n=()=>!0;return{query:t,isFencePassed:n}}downloadMatrixFromPackedTexture(e,t,n){return this.downloadMatrixDriver(e,()=>sP(this.gl,t,n))}createProgram(e){this.throwIfDisposed();let t=this.gl;null==this.vertexShader&&(this.vertexShader=PL(t));let n=xO(t);uO(t,()=>t.attachShader(n,this.vertexShader)),uO(t,()=>t.attachShader(n,e)),vO(t,n);let r=Object.assign(n,{vao:this.createVertexArray()});return this.debug&&wO(t,r),r}buildVao(e){this.setProgram(e),this.bindVertexArray(e.vao);let t=this.gl;uO(t,()=>t.bindBuffer(t.ELEMENT_ARRAY_BUFFER,this.indexBuffer)),JL(t,e,this.vertexBuffer)}deleteProgram(e){this.throwIfDisposed(),e===this.program&&(this.program=null),null!=e&&(uO(this.gl,()=>this.gl.deleteProgram(e)),this.deleteVertexArray(e.vao))}setProgram(e){this.throwIfDisposed(),this.program=e,null!=this.program&&this.debug&&wO(this.gl,this.program),uO(this.gl,()=>this.gl.useProgram(e))}getUniformLocation(e,t,n=!0){return this.throwIfDisposed(),n?$O(this.gl,e,t):RO(this.gl,e,t)}getAttributeLocation(e,t){return this.throwIfDisposed(),uO(this.gl,()=>this.gl.getAttribLocation(e,t))}getUniformLocationNoThrow(e,t){return this.throwIfDisposed(),this.gl.getUniformLocation(e,t)}setInputMatrixTexture(e,t,n){this.throwIfDisposed(),this.throwIfNoProgram(),FO(this.gl,e,t,n)}setOutputMatrixTexture(e,t,n){this.setOutputMatrixTextureDriver(e,n,t)}setOutputPackedMatrixTexture(e,t,n){this.throwIfDisposed();let[r,a]=oO(t,n);this.setOutputMatrixTextureDriver(e,r,a)}setOutputMatrixWriteRegion(e,t,n,r){this.setOutputMatrixWriteRegionDriver(n,e,r,t)}setOutputPackedMatrixWriteRegion(e,t,n,r){throw new Error("setOutputPackedMatrixWriteRegion not implemented.")}debugValidate(){null!=this.program&&wO(this.gl,this.program),LO(this.gl)}executeProgram(){this.throwIfDisposed(),this.throwIfNoProgram();let e=this.gl;if(this.debug){let e=this.getVertexArray();console.assert(e===this.program.vao,"VAO changed between setProgram and executeProgram!"),this.debugValidate()}uO(e,()=>e.drawElements(e.TRIANGLES,6,e.UNSIGNED_SHORT,0))}blockUntilAllProgramsCompleted(){this.throwIfDisposed(),uO(this.gl,()=>this.gl.finish())}getQueryTimerExtension(){return null==this.disjointQueryTimerExtension&&(this.disjointQueryTimerExtension=pO(this.gl,2===Le().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION")?"EXT_disjoint_timer_query_webgl2":"EXT_disjoint_timer_query")),this.disjointQueryTimerExtension}getQueryTimerExtensionWebGL2(){return this.getQueryTimerExtension()}getQueryTimerExtensionWebGL1(){return this.getQueryTimerExtension()}beginQuery(){if(2===Le().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION")){let e=this.gl,t=this.getQueryTimerExtensionWebGL2(),n=e.createQuery();return e.beginQuery(t.TIME_ELAPSED_EXT,n),n}let e=this.getQueryTimerExtensionWebGL1(),t=e.createQueryEXT();return e.beginQueryEXT(e.TIME_ELAPSED_EXT,t),t}endQuery(){if(2===Le().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION")){let e=this.gl,t=this.getQueryTimerExtensionWebGL2();return void e.endQuery(t.TIME_ELAPSED_EXT)}let e=this.getQueryTimerExtensionWebGL1();e.endQueryEXT(e.TIME_ELAPSED_EXT)}async waitForQueryAndGetTime(e){return await va.repeatedTry(()=>this.disposed||this.isQueryAvailable(e,Le().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION"))),this.getQueryTime(e,Le().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_VERSION"))}getQueryTime(e,t){if(0===t)return null;if(2===t){let t=this.gl;return t.getQueryParameter(e,t.QUERY_RESULT)/1e6}{let t=this.getQueryTimerExtensionWebGL1();return t.getQueryObjectEXT(e,t.QUERY_RESULT_EXT)/1e6}}isQueryAvailable(e,t){if(0===t)return!0;if(2===t){let t=this.gl,n=this.getQueryTimerExtensionWebGL2(),r=t.getQueryParameter(e,t.QUERY_RESULT_AVAILABLE);return null==this.disjoint&&(this.disjoint=this.gl.getParameter(n.GPU_DISJOINT_EXT)),r&&!this.disjoint}{let t=this.getQueryTimerExtensionWebGL1(),n=t.getQueryObjectEXT(e,t.QUERY_RESULT_AVAILABLE_EXT);return null==this.disjoint&&(this.disjoint=this.gl.getParameter(t.GPU_DISJOINT_EXT)),n&&!this.disjoint}}pollFence(e){return new Promise(t=>{this.addItemToPoll(()=>e.isFencePassed(),()=>t())})}pollItems(){let e=function(e){let t=0;for(;t<e.length&&e[t]();++t);return t-1}(this.itemsToPoll.map(e=>e.isDoneFn));for(let t=0;t<=e;++t){let{resolveFn:e}=this.itemsToPoll[t];e()}this.itemsToPoll=this.itemsToPoll.slice(e+1)}addItemToPoll(e,t){if(this.itemsToPoll.push({isDoneFn:e,resolveFn:t}),this.itemsToPoll.length>1)return;let n;"setTimeoutCustom"in Le().platform&&(n=Le().platform.setTimeoutCustom.bind(Le().platform)),va.repeatedTry(()=>(this.pollItems(),0===this.itemsToPoll.length),()=>0,null,n)}bindTextureToFrameBuffer(e){this.throwIfDisposed(),MO(this.gl,e,this.framebuffer),this.debug&&LO(this.gl)}unbindTextureToFrameBuffer(){null!=this.outputTexture?(MO(this.gl,this.outputTexture,this.framebuffer),this.debug&&LO(this.gl)):OO(this.gl,this.framebuffer)}downloadMatrixDriver(e,t){this.bindTextureToFrameBuffer(e);let n=t();return this.unbindTextureToFrameBuffer(),n}setOutputMatrixTextureDriver(e,t,n){this.throwIfDisposed();let r=this.gl;MO(r,e,this.framebuffer),this.debug&&LO(r),this.outputTexture=e,uO(r,()=>r.viewport(0,0,t,n)),uO(r,()=>r.scissor(0,0,t,n))}setOutputMatrixWriteRegionDriver(e,t,n,r){this.throwIfDisposed(),uO(this.gl,()=>this.gl.scissor(e,t,n,r))}throwIfDisposed(){if(this.disposed)throw new Error("Attempted to use disposed GPGPUContext.")}throwIfNoProgram(){if(null==this.program)throw new Error("No GPU program is currently set.")}};var{addImpl:oP,bincountImpl:lP,bincountReduceImpl:uP,bitwiseAndImpl:hP,castImpl:dP,ceilImpl:pP,concatImpl:cP,equalImpl:fP,expImpl:mP,expm1Impl:gP,floorImpl:yP,gatherNdImpl:bP,gatherV2Impl:xP,greaterImpl:vP,greaterEqualImpl:wP,lessImpl:kP,lessEqualImpl:IP,linSpaceImpl:SP,logImpl:_P,maxImpl:NP,maximumImpl:TP,minimumImpl:CP,multiplyImpl:EP,negImpl:AP,notEqualImpl:$P,prodImpl:RP,raggedGatherImpl:FP,raggedRangeImpl:DP,raggedTensorToTensorImpl:MP,rangeImpl:OP,rsqrtImpl:LP,scatterImpl:PP,sigmoidImpl:zP,simpleAbsImpl:BP,sliceImpl:WP,sparseFillEmptyRowsImpl:VP,sparseReshapeImpl:UP,sparseSegmentReductionImpl:GP,sqrtImpl:HP,staticRegexReplaceImpl:jP,stridedSliceImpl:qP,stringNGramsImpl:KP,stringSplitImpl:XP,stringToHashBucketFastImpl:ZP,subImpl:YP,tileImpl:JP,topKImpl:QP,transposeImpl:ez,uniqueImpl:tz}=yE;function nz(e,t){return["x","y","z","w","u","v"].slice(0,t).map(t=>`${e}.${t}`)}function rz(e,t){return 1===t?[e]:nz(e,t)}var az=class{constructor(e){if(this.variableNames=["A"],this.packedInputs=!1,this.packedOutput=!0,this.outputShape=e,this.rank=e.length,this.enableShapeUniforms=EL(this.outputShape.length),0===this.rank)this.userCode="\n        void main() {\n          setOutput(vec4(getA(), 0., 0., 0.));\n        }\n      ";else{let e=rz("rc",this.rank),t=IL(this.rank),n=this.getOutOfBoundsCondition(e),r=this.getSetup(e),a=this.getOutput(e);this.userCode=`\n        void main() {\n          ${t} rc = getOutputCoords();\n\n          if(${n}) {\n            setOutput(vec4(0));\n          } else {\n            ${r}\n\n            setOutput(vec4(${a}));\n          }\n        }\n      `}}getSourceCoordsArr(e){let t=[];for(let n=0;n<=1;n++)for(let r=0;r<=1;r++){let a=`${0===n?"r":"rp1"}, ${0===r?"c":"cp1"}`;for(let t=2;t<this.rank;t++)a=`${e[e.length-1-t]},`+a;t.push(a)}return t}getOutOfBoundsCondition(e){if(1===this.rank)return`rc > ${this.enableShapeUniforms?"outShape":this.outputShape[0]}`;let t="";for(let n=this.rank-2;n<this.rank;n++)t+=`${e[n]} >= ${this.enableShapeUniforms?`outShape[${n}]`:this.outputShape[n]}`,n<this.rank-1&&(t+="||");return t}getSetup(e){if(1===this.rank)return"";let t=e.slice(-2),n=this.enableShapeUniforms?`outShape[${this.rank} - 1]`:this.outputShape[this.rank-1],r=this.enableShapeUniforms?`outShape[${this.rank} - 2]`:this.outputShape[this.rank-2];return`\n      int r = ${t[0]};\n      int c = ${t[1]};\n      int rp1 = r + 1;\n      int cp1 = c + 1;\n\n      bool cEdge = cp1 >= ${n};\n      bool rEdge = rp1 >= ${r};\n    `}getOutput(e){let t=this.getSourceCoordsArr(e);return 1===this.rank?`getA(rc), (rc + 1 >= ${this.enableShapeUniforms?"outShape":this.outputShape[0]} ? 0. : getA(rc + 1)), 0, 0`:`getA(${t[0]}),\n            cEdge ? 0. : getA(${t[1]}),\n            rEdge ? 0. : getA(${t[2]}),\n            rEdge || cEdge ? 0. : getA(${t[3]})`}},sz=class{constructor(e,t){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0,this.customUniforms=[{name:"inputShape",type:"ivec3"}],this.outputShape=e,this.enableShapeUniforms=EL(this.outputShape.length);let n="";for(let e=0;e<4;e++){let t="thisRC = rc;";e%2==1&&(t+="thisRC.z += 1;"),e>1&&(t+="thisRC.y += 1;"),n+=`\n        ${t}\n        ${e>0?"if(thisRC.y < rows && thisRC.z < cols){":""}\n          int flatIndex = getFlatIndex(thisRC);\n\n          ivec3 inputRC = inputCoordsFromReshapedOutCoords(flatIndex);\n          vec2 inputRCInnerDims = vec2(float(inputRC.y),float(inputRC.z));\n\n          result[${e}] =\n            getChannel(getA(inputRC.x, inputRC.y, inputRC.z), inputRCInnerDims);\n        ${e>0?"}":""}\n      `}this.userCode=`\n      ${function(e,t){return`\n    ivec3 inputCoordsFromReshapedOutCoords(int index) {\n      ${t?uL(["r","c","d"],"inputShape"):oL(["r","c","d"],e)}\n      return ivec3(r, c, d);\n    }\n  `}(t,this.enableShapeUniforms)}\n      ${this.enableShapeUniforms?"\n  int getFlatIndex(ivec3 coords) {\n    return coords.x * outShapeStrides[0] + coords.y * outShapeStrides[1] + coords.z;\n  }\n":hL(e)}\n\n      void main() {\n        ivec3 rc = getOutputCoords();\n\n        vec4 result = vec4(0.);\n\n        ivec3 thisRC;\n        int rows = ${this.enableShapeUniforms?"outShape[1]":e[1]};\n        int cols = ${this.enableShapeUniforms?"outShape[2]":e[2]};\n\n        ${n}\n\n        setOutput(result);\n      }\n    `}};var iz=class{constructor(e){this.gpgpu=e,this.numUsedTextures=0,this.numFreeTextures=0,this._numBytesAllocated=0,this._numBytesFree=0,this.freeTextures={},this.usedTextures={},this.logEnabled=!1}acquireTexture(e,t,n){let r=lz(t,n),a=uz(e,r,n);a in this.freeTextures||(this.freeTextures[a]=[]),a in this.usedTextures||(this.usedTextures[a]=[]);let s,i=oz(e,r,this.gpgpu.gl,this.gpgpu.textureConfig,n);if(this.freeTextures[a].length>0){this.numFreeTextures--,this.numUsedTextures++,this._numBytesFree-=i,this.log();let e=this.freeTextures[a].pop();return this.usedTextures[a].push(e),e}return r===eO.PACKED_2X2_FLOAT32?s=this.gpgpu.createPackedMatrixTexture(e[0],e[1]):r===eO.PACKED_2X2_FLOAT16?s=this.gpgpu.createFloat16PackedMatrixTexture(e[0],e[1]):r===eO.UNPACKED_FLOAT32?s=this.gpgpu.createFloat32MatrixTexture(e[0],e[1]):r===eO.UNPACKED_FLOAT16?s=this.gpgpu.createFloat16MatrixTexture(e[0],e[1]):r===eO.PACKED_4X1_UNSIGNED_BYTE&&(s=this.gpgpu.createUnsignedBytesMatrixTexture(e[0],e[1])),this.usedTextures[a].push(s),this.numUsedTextures++,this._numBytesAllocated+=i,this.log(),s}releaseTexture(e,t,n,r){if(null==this.freeTextures)return;let a=lz(n,r),s=uz(t,a,r);s in this.freeTextures||(this.freeTextures[s]=[]);let i=oz(t,a,this.gpgpu.gl,this.gpgpu.textureConfig,r),o=Le().getNumber("WEBGL_DELETE_TEXTURE_THRESHOLD");-1!==o&&this._numBytesAllocated>o?(this.gpgpu.deleteMatrixTexture(e.texture),this._numBytesAllocated-=i):(this.freeTextures[s].push(e),this.numFreeTextures++,this._numBytesFree+=i),this.numUsedTextures--;let l=this.usedTextures[s],u=l&&l.indexOf(e);if(null==u||u<0)throw new Error("Cannot release a texture that was never provided by this texture manager");l[u]=l[l.length-1],l.pop(),this.log()}log(){if(!this.logEnabled)return;let e=this.numFreeTextures+this.numUsedTextures;console.log("Free/Used",`${this.numFreeTextures} / ${this.numUsedTextures}`,`(${e})`);let t=this._numBytesFree/this._numBytesAllocated;console.log(`Bytes allocated: ${this._numBytesAllocated}`),console.log(`Bytes unused: ${this._numBytesFree} (${Math.round(100*t)}%)`)}get numBytesAllocated(){return this._numBytesAllocated}get numBytesFree(){return this._numBytesFree}getNumUsedTextures(){return this.numUsedTextures}getNumFreeTextures(){return this.numFreeTextures}dispose(){if(null!=this.freeTextures){for(let e in this.freeTextures)this.freeTextures[e].forEach(e=>{this.gpgpu.deleteMatrixTexture(e.texture)});for(let e in this.usedTextures)this.usedTextures[e].forEach(e=>{this.gpgpu.deleteMatrixTexture(e.texture)});this.freeTextures=null,this.usedTextures=null,this.numUsedTextures=0,this.numFreeTextures=0,this._numBytesAllocated=0,this._numBytesFree=0}}};function oz(e,t,n,r,a){let s,i=function(e,t){switch(e){case eO.PACKED_2X2_FLOAT32:return KL(t);case eO.PACKED_2X2_FLOAT16:return ZL(t);case eO.UNPACKED_FLOAT32:return VL(t);case eO.UNPACKED_FLOAT16:return GL(t);case eO.PACKED_4X1_UNSIGNED_BYTE:return jL(t);default:throw new Error(`Unknown physical texture type ${e}`)}}(t,r);if(a){let[t,n]=oO(e[0],e[1]);s=t*n}else{let[t,n]=sO(e[0],e[1]);s=t*n}let o=function(e,t){let n=e;if(t===n.R32F)return 4;if(t===n.R16F)return 2;if(t===n.RGBA32F||t===e.RGBA)return 16;if(t===n.RGBA16F)return 8;if(t===n.RGBA8)return 4;throw new Error(`Unknown internal format ${t}`)}(n,i);return s*o}function lz(e,t){if(e===QM.UPLOAD)return eO.PACKED_2X2_FLOAT32;if(e===QM.RENDER||null==e)return function(e){return Le().getBool("WEBGL_RENDER_FLOAT32_ENABLED")?e?eO.PACKED_2X2_FLOAT32:eO.UNPACKED_FLOAT32:e?eO.PACKED_2X2_FLOAT16:eO.UNPACKED_FLOAT16}(t);if(e===QM.DOWNLOAD||e===QM.PIXELS)return eO.PACKED_4X1_UNSIGNED_BYTE;throw new Error(`Unknown logical texture type ${e}`)}function uz(e,t,n){return`${e[0]}_${e[1]}_${t}_${n}`}var hz=class{constructor(e,t){this.variableNames=["A"],this.outputShape=e,this.enableShapeUniforms=EL(this.outputShape.length),this.userCode=`\n      float unaryOperation(float x) {\n        ${t}\n      }\n\n      void main() {\n        float x = getAAtOutCoords();\n        float y = unaryOperation(x);\n\n        setOutput(y);\n      }\n    `}},dz="if (isnan(x)) return x;",pz="return abs(x);",cz=dz+"\n  return (x < 0.0) ? 0.0 : x;\n",fz=dz+"\n  return (x < 0.0) ? 0.0 : min(6.0, x);\n",mz="return x;",gz=class{constructor(e,t){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0,this.outputShape=e,this.enableShapeUniforms=EL(this.outputShape.length),this.userCode=`\n      vec4 unaryOperation(vec4 x) {\n        ${t}\n      }\n\n      void main() {\n        vec4 x = getAAtOutCoords();\n        vec4 y = unaryOperation(x);\n\n        setOutput(y);\n      }\n    `}},yz=class{constructor(e){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!1,this.outputShape=e,this.enableShapeUniforms=EL(this.outputShape.length);let t=e.length,n=rz("rc",t),r=IL(t),a=function(e,t){if(1===e)return"rc";let n="";for(let r=0;r<e;r++)n+=t[r],r<e-1&&(n+=",");return n}(t,n),s=n.slice(-2),i=t<=1?"rc":`vec2(${s.join(",")})`;this.userCode=`\n      void main() {\n        ${r} rc = getOutputCoords();\n        vec4 packedInput = getA(${a});\n\n        setOutput(getChannel(packedInput, ${i}));\n      }\n    `}},bz=Km.whereImpl,xz={};var vz=Le().getNumber("CPU_HANDOFF_SIZE_THRESHOLD");var wz=class e extends P{nextDataId(){return e.nextDataId++}constructor(e){if(super(),this.pendingRead=new WeakMap,this.pendingDisposal=new WeakSet,this.dataRefCount=new WeakMap,this.numBytesInGPU=0,this.uploadWaitMs=0,this.downloadWaitMs=0,this.lastGlFlushTime=0,this.warnedAboutMemory=!1,this.pendingDeletes=0,this.disposed=!1,!Le().getBool("HAS_WEBGL"))throw new Error("WebGL is not supported on this device");let t;if(null!=e){if(e instanceof iP)t=e;else{let n=aO(Le().getNumber("WEBGL_VERSION"),e);t=new iP(n)}this.binaryCache={},this.gpgpuCreatedLocally=!1}else{let e=aO(Le().getNumber("WEBGL_VERSION"));t=new iP(e),this.binaryCache=function(e){return e in xz||(xz[e]={}),xz[e]}(Le().getNumber("WEBGL_VERSION")),this.gpgpuCreatedLocally=!0}this.gpgpu=t,this.canvas=this.gpgpu.gl.canvas,this.textureManager=new iz(this.gpgpu),this.numMBBeforeWarning=null==Le().global.screen?1024:Le().global.screen.height*Le().global.screen.width*window.devicePixelRatio*600/1024/1024,this.texData=new L(this,Zs())}numDataIds(){return this.texData.numDataIds()-this.pendingDeletes}writeTexture(e,t,n,r,a,s){let i=this.makeTensorInfo(t,n),o=this.texData.get(i.dataId);o.isPacked=!1,o.texture={texture:e,texShape:[r,a]},o.texShape=[r,a];let l=UO(t),u=new ML(l,!1,s),h=this.runWebGLProgram(u,[i],n,[[r,a]]);return h.shape=t,o.texture=null,this.disposeIntermediateTensorInfo(i),h.dataId}write(e,t,n){if((Le().getBool("WEBGL_CHECK_NUMERICAL_PROBLEMS")||Le().getBool("DEBUG"))&&this.checkNumericalProblems(e),"complex64"===n&&null!=e)throw new Error("Cannot write to a complex64 dtype. Please use tf.complex(real, imag).");let r={id:this.nextDataId()};return this.texData.set(r,{shape:t,dtype:n,values:e,usage:QM.UPLOAD,refCount:1}),r}refCount(e){return this.texData.has(e)?this.texData.get(e).refCount:0}incRef(e){this.texData.get(e).refCount++}decRef(e){if(this.texData.has(e)){this.texData.get(e).refCount--}}move(e,t,n,r,a){if(Le().getBool("DEBUG")&&this.checkNumericalProblems(t),"complex64"===r)throw new Error("Cannot write to a complex64 dtype. Please use tf.complex(real, imag).");this.texData.set(e,{shape:n,dtype:r,values:t,usage:QM.UPLOAD,refCount:a})}disposeIntermediateTensorInfo(e){this.disposeData(e.dataId)}readSync(e){let t=this.texData.get(e),{values:n,dtype:r,complexTensorInfos:a,slice:s,shape:i,isPacked:o}=t;if(null!=s){let t;t=o?new gz(i,mz):new hz(i,mz);let n=this.runWebGLProgram(t,[{dataId:e,shape:i,dtype:r}],r),a=this.readSync(n.dataId);return this.disposeIntermediateTensorInfo(n),a}if(null!=n)return this.convertAndCacheOnCPU(e);if("string"===r)return n;let l,u,h=null!=this.activeTimers;if(h&&(l=va.now()),"complex64"===r){let e=this.readSync(a.real.dataId),t=this.readSync(a.imag.dataId);u=Gf.mergeRealAndImagArrays(e,t)}else u=this.getValuesFromTexture(e);return h&&(this.downloadWaitMs+=va.now()-l),this.convertAndCacheOnCPU(e,u)}async read(e){if(this.pendingRead.has(e)){let t=this.pendingRead.get(e);return new Promise(e=>t.push(e))}let t=this.texData.get(e),{values:n,shape:r,slice:a,dtype:s,complexTensorInfos:i,isPacked:o}=t;if(null!=a){let t;t=o?new gz(r,mz):new hz(r,mz);let n=this.runWebGLProgram(t,[{dataId:e,shape:r,dtype:s}],s),a=this.read(n.dataId);return this.disposeIntermediateTensorInfo(n),a}if(null!=n)return this.convertAndCacheOnCPU(e);if(Le().getBool("DEBUG")&&!Le().getBool("WEBGL_DOWNLOAD_FLOAT_ENABLED")&&2===Le().getNumber("WEBGL_VERSION"))throw new Error("tensor.data() with WEBGL_DOWNLOAD_FLOAT_ENABLED=false and WEBGL_VERSION=2 not yet supported.");let l,u,h=null;if("complex64"!==s&&Le().get("WEBGL_BUFFER_SUPPORTED")){l=this.decode(e);let t=this.texData.get(l.dataId);h=this.gpgpu.createBufferFromTexture(t.texture.texture,...iO(r))}if(this.pendingRead.set(e,[]),"complex64"!==s&&await this.gpgpu.createAndWaitForFence(),"complex64"===s){let e=await Promise.all([this.read(i.real.dataId),this.read(i.imag.dataId)]),t=e[0],n=e[1];u=Gf.mergeRealAndImagArrays(t,n)}else if(null==h)u=this.getValuesFromTexture(e);else{let e=va.sizeFromShape(r);u=this.gpgpu.downloadFloat32MatrixFromBuffer(h,e)}if(null!=l&&this.disposeIntermediateTensorInfo(l),null!=h){let e=this.gpgpu.gl;uO(e,()=>e.deleteBuffer(h))}let d=this.convertAndCacheOnCPU(e,u),p=this.pendingRead.get(e);return this.pendingRead.delete(e),p.forEach(e=>e(d)),this.pendingDisposal.has(e)&&(this.pendingDisposal.delete(e),this.disposeData(e)&&Zs().removeDataId(e,this),this.pendingDeletes--),d}readToGPU(e,t={}){let n=this.texData.get(e),{values:r,shape:a,slice:s,dtype:i,isPacked:o,texture:l}=n;if("complex64"===i)throw new Error("Does not support reading texture for complex64 dtype.");if(null!=s){let n;n=o?new gz(a,mz):new hz(a,mz);let r=this.runWebGLProgram(n,[{dataId:e,shape:a,dtype:i}],i),s=this.readToGPU(r,t);return this.disposeIntermediateTensorInfo(r),s}if(null==l)throw null!=r?new Error("Data is not on GPU but on CPU."):new Error("There is no data on GPU or CPU.");let u=this.decode(e,t.customTexShape),h=Zs().makeTensorFromTensorInfo(u),d=this.texData.get(u.dataId);return Object.assign({tensorRef:h},d.texture)}bufferSync(e){let t=this.readSync(e.dataId);if("string"===e.dtype)try{let n=t.map(e=>va.decodeString(e));return uo(e.shape,e.dtype,n)}catch(e){throw new Error("Failed to decode encoded string bytes into utf-8")}return uo(e.shape,e.dtype,t)}checkNumericalProblems(e){if(null!=e)for(let t=0;t<e.length;t++){let n=e[t];if(!hO(n))throw Le().getBool("WEBGL_RENDER_FLOAT32_CAPABLE")?Error(`The value ${n} cannot be represented with your current settings. Consider enabling float32 rendering: 'tf.env().set('WEBGL_RENDER_FLOAT32_ENABLED', true);'`):Error(`The value ${n} cannot be represented on this device.`)}}getValuesFromTexture(e){let{shape:t,dtype:n,isPacked:r}=this.texData.get(e),a=va.sizeFromShape(t);if(Le().getBool("WEBGL_DOWNLOAD_FLOAT_ENABLED")){let n=this.decode(e),r=this.texData.get(n.dataId),s=this.gpgpu.downloadMatrixFromPackedTexture(r.texture.texture,...iO(t)).subarray(0,a);return this.disposeIntermediateTensorInfo(n),s}let s=Le().getBool("WEBGL_PACK")&&!0===r,i=s?UO(t):t,o=s?new FL(i):new RL(i),l=this.runWebGLProgram(o,[{shape:i,dtype:n,dataId:e}],"float32"),u=this.texData.get(l.dataId),h=this.gpgpu.downloadByteEncodedFloatMatrixFromOutputTexture(u.texture.texture,u.texShape[0],u.texShape[1]).subarray(0,a);return this.disposeIntermediateTensorInfo(l),h}timerAvailable(){return Le().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE")>0}time(e){let t=this.activeTimers,n=[],r=!1;null==this.programTimersStack?(this.programTimersStack=n,r=!0):this.activeTimers.push(n),this.activeTimers=n,e();let a=va.flatten(this.activeTimers.map(e=>e.query)).filter(e=>null!=e),s=va.flatten(this.activeTimers.map(e=>e.name)).filter(e=>null!=e);this.activeTimers=t,r&&(this.programTimersStack=null);let i={uploadWaitMs:this.uploadWaitMs,downloadWaitMs:this.downloadWaitMs,kernelMs:null,wallMs:null};return(async()=>{if(Le().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE")>0){let e=await Promise.all(a);i.kernelMs=va.sum(e),i.getExtraProfileInfo=()=>e.map((e,t)=>({name:s[t],ms:e})).map(e=>`${e.name}: ${e.ms}`).join(", ")}else i.kernelMs={error:"WebGL query timers are not supported in this environment."};return this.uploadWaitMs=0,this.downloadWaitMs=0,i})()}memory(){return{unreliable:!1,numBytesInGPU:this.numBytesInGPU,numBytesInGPUAllocated:this.textureManager.numBytesAllocated,numBytesInGPUFree:this.textureManager.numBytesFree}}startTimer(){return Le().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE")>0?this.gpgpu.beginQuery():{startMs:va.now(),endMs:null}}endTimer(e){return Le().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE")>0?(this.gpgpu.endQuery(),e):(e.endMs=va.now(),e)}async getQueryTime(e){if(Le().getNumber("WEBGL_DISJOINT_QUERY_TIMER_EXTENSION_RELIABLE")>0)return this.gpgpu.waitForQueryAndGetTime(e);let t=e;return t.endMs-t.startMs}disposeData(e,t=!1){if(this.pendingDisposal.has(e))return!1;if(!this.texData.has(e))return!0;if(t?this.texData.get(e).refCount=0:this.texData.get(e).refCount--,!t&&this.texData.get(e).refCount>0)return!1;if(this.pendingRead.has(e))return this.pendingDisposal.add(e),this.pendingDeletes++,!1;this.releaseGPUData(e);let{complexTensorInfos:n}=this.texData.get(e);return null!=n&&(this.disposeData(n.real.dataId,t),this.disposeData(n.imag.dataId,t)),this.texData.delete(e),!0}releaseGPUData(e){let{texture:t,dtype:n,texShape:r,usage:a,isPacked:s,slice:i}=this.texData.get(e),o=i&&i.origDataId||e,l=this.dataRefCount.get(o);l>1?this.dataRefCount.set(o,l-1):(this.dataRefCount.delete(o),null!=t&&(this.numBytesInGPU-=this.computeBytes(r,n),this.textureManager.releaseTexture(t,r,a,s)));let u=this.texData.get(e);u.texture=null,u.texShape=null,u.isPacked=!1,u.slice=null}getTexture(e){return this.uploadToGPU(e),this.texData.get(e).texture.texture}getDataInfo(e){return this.texData.get(e)}shouldExecuteOnCPU(e,t=vz){return Le().getBool("WEBGL_CPU_FORWARD")&&e.every(e=>null==this.texData.get(e.dataId).texture&&va.sizeFromShape(e.shape)<t)}getGPGPUContext(){return this.gpgpu}where(e){Gf.warn("tf.where() in webgl locks the UI thread. Call tf.whereAsync() instead");let t=e.dataSync();return bz(e.shape,t)}packedUnaryOp(e,t,n){let r=new gz(e.shape,t),a=this.compileAndRun(r,[e],n);return Zs().makeTensorFromTensorInfo(a)}abs(e){if(this.shouldExecuteOnCPU([e])&&"complex64"!==e.dtype){let t=BP(this.texData.get(e.dataId).values);return this.makeOutput(e.shape,e.dtype,t)}if(Le().getBool("WEBGL_PACK_UNARY_OPERATIONS"))return this.packedUnaryOp(e,pz,e.dtype);let t=new hz(e.shape,pz),n=this.compileAndRun(t,[e]);return Zs().makeTensorFromTensorInfo(n)}makeTensorInfo(e,t,n){let r;if("string"===t&&null!=n&&n.length>0&&va.isString(n[0])){let a=n.map(e=>va.encodeString(e));r=this.write(a,e,t)}else r=this.write(n,e,t);return this.texData.get(r).usage=null,{dataId:r,shape:e,dtype:t}}makeOutput(e,t,n){return Zs().makeTensorFromTensorInfo(this.makeTensorInfo(e,t,n),this)}unpackTensor(e){let t=new yz(e.shape);return this.runWebGLProgram(t,[e],e.dtype)}packTensor(e){let t=new az(e.shape);return this.runWebGLProgram(t,[e],e.dtype,null,!0)}packedReshape(e,t){let n=[WO(e.shape),...VO(e.shape)],r={dtype:e.dtype,shape:n,dataId:e.dataId},a=[WO(t),...VO(t)],s=new sz(a,n),i=[n],o=this.runWebGLProgram(s,[r],e.dtype,i,!0);return{dataId:o.dataId,shape:t,dtype:o.dtype}}decode(e,t){let n=this.texData.get(e),{isPacked:r,shape:a,dtype:s}=n;if(null!=t){let e=va.sizeFromShape(a),n=t[0]*t[1]*4;va.assert(e<=n,()=>"customTexShape is too small. Row * Column * 4 should be equal or larger than the size of the tensor data.")}let i,o=UO(a);i=r?new $L(o):new AL(o);let l=[null!=t?t:iO(o)];return{dtype:s,shape:a,dataId:this.runWebGLProgram(i,[{shape:o,dtype:s,dataId:e}],s,l,!0,t).dataId}}runWebGLProgram(e,t,n,r,a=!1,s){let i=this.makeTensorInfo(e.outputShape,n),o=this.texData.get(i.dataId);if(e.packedOutput&&(o.isPacked=!0),e.outPackingScheme===JM.DENSE){let t=null!=s?s:iO(e.outputShape);o.texShape=t.map(e=>2*e)}if(null!=e.outTexUsage&&(o.usage=e.outTexUsage),0===va.sizeFromShape(i.shape))return o.values=va.getTypedArrayFromDType(i.dtype,0),i;let l=[],u=t.map(t=>{if("complex64"===t.dtype)throw new Error("GPGPUProgram does not support complex64 input. For complex64 dtypes, please separate the program into real and imaginary parts.");let n=this.texData.get(t.dataId);if(null==n.texture){if(!e.packedInputs&&va.sizeFromShape(t.shape)<=Le().getNumber("WEBGL_SIZE_UPLOAD_UNIFORM"))return{shape:t.shape,texData:null,isUniform:!0,uniformValues:n.values};e.packedInputs&&(n.isPacked=!0,n.shape=t.shape)}if(this.uploadToGPU(t.dataId),!!n.isPacked!=!!e.packedInputs)t=n.isPacked?this.unpackTensor(t):this.packTensor(t),l.push(t),n=this.texData.get(t.dataId);else if(n.isPacked&&!jO(n.shape,t.shape)){let e=t,r=t.shape;t.shape=n.shape,t=this.packedReshape(t,r),l.push(t),n=this.texData.get(t.dataId),e.shape=r}return{shape:t.shape,texData:n,isUniform:!1}});this.uploadToGPU(i.dataId);let h,d={shape:i.shape,texData:o,isUniform:!1},p=function(e,t,n){let r="";t.concat(n).forEach(t=>{let a=null!=t.texData&&null!=t.texData.slice&&t.texData.slice.flatOffset>0;if(e.enableShapeUniforms&&!t.isUniform){let s=t.texData.texShape,{useSqueezeShape:i,uniformShape:o,keptDims:l}=SL(e.packedInputs,t.shape,s),u="",h="",d="";if(1===o.length&&e.packedInputs){let e=[Math.ceil(s[0]/2),Math.ceil(s[1]/2)];u=`${e[0]>1}_${e[1]>1}`}else if(2!==o.length||e.packedInputs){if(o.length>2&&!e.packedInputs){let e=va.computeStrides(o);d=`${e[0]===s[1]}_${e[e.length-1]===s[1]}`}}else h=`${o[0]>1}_${o[1]>1}`;let p=t.shape.length,c=2===o.length&&va.arraysEqual(t.shape,s),f=1===va.sizeFromShape(t.shape),m=Gf.getBroadcastDims(t.shape,n.shape),g=!e.packedInputs&&p===n.shape.length&&va.arraysEqual(s,n.texData.texShape),y=e.packedInputs||o.length>2?"":`${s[0]>1}_${s[1]>1}`;r+=`${p}_${g}_${i?l:""}_${o.length}_${f}_${m}_${c}_${u}_${h}_${d}_${y}_${a}`}else{let e=t.isUniform?"uniform":t.texData.texShape;r+=`${t.shape}_${e}_${a}`}});let a=e.userCode,s=e.constructor.name;return s+="_"+r+"_"+a+`${Le().getNumber("WEBGL_VERSION")}`,s}(e,u,d),c=this.getAndSaveBinary(p,()=>function(e,t,n,r){let a=n.map((e,n)=>{let r={logicalShape:e.shape,texShape:e.isUniform?null:e.texData.texShape,isUniform:e.isUniform,isPacked:!e.isUniform&&e.texData.isPacked,flatOffset:null};return null!=e.texData&&null!=e.texData.slice&&e.texData.slice.flatOffset>0&&(r.flatOffset=e.texData.slice.flatOffset),{name:t.variableNames[n],shapeInfo:r}}),s=a.map(e=>e.shapeInfo),i={logicalShape:r.shape,texShape:r.texData.texShape,isUniform:!1,isPacked:r.texData.isPacked,flatOffset:null},o=cL(a,i,t),l=fO(e.gl,o),u=e.createProgram(l);return Le().get("ENGINE_COMPILE_ONLY")?{program:t,fragmentShader:l,source:o,webGLProgram:u,inShapeInfos:s,outShapeInfo:i,variablesLocations:null,customUniformLocations:null,infLoc:null,nanLoc:null,outShapeLocation:null,outShapeStridesLocation:null,outTexShapeLocation:null}:(e.buildVao(u),Object.assign({program:t,fragmentShader:l,source:o,webGLProgram:u,inShapeInfos:s,outShapeInfo:i},TL(e,t,u)))}(this.gpgpu,e,u,d)),f=null!=this.activeTimers;f&&(h=this.startTimer()),Le().get("ENGINE_COMPILE_ONLY")||function(e,t,n,r,a){t.program.enableShapeUniforms||(CL(t.inShapeInfos,n),CL([t.outShapeInfo],[r]));let s=r.texData.texture,i=r.texData.texShape;r.texData.isPacked?e.setOutputPackedMatrixTexture(s.texture,i[0],i[1]):e.setOutputMatrixTexture(s.texture,i[0],i[1]),e.setProgram(t.webGLProgram),e.bindVertexArray(t.webGLProgram.vao),1===Le().getNumber("WEBGL_VERSION")&&null!==t.infLoc&&e.gl.uniform1f(t.infLoc,1/0),null!==t.nanLoc&&e.gl.uniform1f(t.nanLoc,NaN);for(let r=0;r<n.length;++r){let a=n[r],{uniform:s,offset:i,shape:o,texShape:l}=t.variablesLocations[r];if(o){let{uniformShape:n}=SL(t.program.packedInputs,a.shape,a.texData.texShape);switch(n.length){case 1:e.gl.uniform1iv(o,new Int32Array(n));break;case 2:e.gl.uniform2iv(o,new Int32Array(n));break;case 3:e.gl.uniform3iv(o,new Int32Array(n));break;case 4:e.gl.uniform4iv(o,new Int32Array(n))}}if(l&&e.gl.uniform2i(l,a.texData.texShape[0],a.texData.texShape[1]),null!=s){if(a.isUniform){if(va.sizeFromShape(a.shape)<2)e.gl.uniform1f(s,a.uniformValues[0]);else{let t=a.uniformValues;t instanceof Float32Array||(t=new Float32Array(t)),e.gl.uniform1fv(s,t)}continue}null!=a.texData.slice&&null!=i&&e.gl.uniform1i(i,a.texData.slice.flatOffset),e.setInputMatrixTexture(a.texData.texture.texture,s,r)}}let o=t.outShapeLocation;if(o)switch(r.shape.length){case 1:e.gl.uniform1iv(o,new Int32Array(r.shape));break;case 2:e.gl.uniform2iv(o,new Int32Array(r.shape));break;case 3:e.gl.uniform3iv(o,new Int32Array(r.shape));break;case 4:e.gl.uniform4iv(o,new Int32Array(r.shape))}if(t.outShapeStridesLocation){let n=va.computeStrides(r.shape);switch(r.shape.length){case 2:e.gl.uniform1iv(t.outShapeStridesLocation,new Int32Array(n));break;case 3:e.gl.uniform2iv(t.outShapeStridesLocation,new Int32Array(n));break;case 4:e.gl.uniform3iv(t.outShapeStridesLocation,new Int32Array(n))}}if(t.outTexShapeLocation&&e.gl.uniform2i(t.outTexShapeLocation,r.texData.texShape[0],r.texData.texShape[1]),t.program.customUniforms&&a)for(let n=0;n<t.program.customUniforms.length;++n){let r=t.program.customUniforms[n],s=t.customUniformLocations[n],i=a[n];if("float"===r.type)e.gl.uniform1fv(s,i);else if("vec2"===r.type)e.gl.uniform2fv(s,i);else if("vec3"===r.type)e.gl.uniform3fv(s,i);else if("vec4"===r.type)e.gl.uniform4fv(s,i);else if("int"===r.type)e.gl.uniform1iv(s,i);else if("ivec2"===r.type)e.gl.uniform2iv(s,i);else if("ivec3"===r.type)e.gl.uniform3iv(s,i);else{if("ivec4"!==r.type)throw Error(`uniform type ${r.type} is not supported yet.`);e.gl.uniform4iv(s,i)}}e.executeProgram()}(this.gpgpu,c,u,d,r),l.forEach(e=>this.disposeIntermediateTensorInfo(e)),f&&(h=this.endTimer(h),this.activeTimers.push({name:e.constructor.name,query:this.getQueryTime(h)}));let m=Le().getNumber("WEBGL_FLUSH_THRESHOLD");if(m>0){let e=va.now();e-this.lastGlFlushTime>m&&(this.gpgpu.gl.flush(),this.lastGlFlushTime=e)}if(!Le().getBool("WEBGL_LAZILY_UNPACK")&&o.isPacked&&!1===a){let e=this.unpackTensor(i);return this.disposeIntermediateTensorInfo(i),e}return i}compileAndRun(e,t,n,r,a=!1){return n=n||t[0].dtype,this.runWebGLProgram(e,t,n,r,a)}getAndSaveBinary(e,t){return e in this.binaryCache||(this.binaryCache[e]=t()),this.binaryCache[e]}getTextureManager(){return this.textureManager}dispose(){this.disposed||(Le().getBool("IS_TEST")||Object.keys(this.binaryCache).forEach(e=>{this.gpgpu.deleteProgram(this.binaryCache[e].webGLProgram),delete this.binaryCache[e]}),this.textureManager.dispose(),null!=this.canvas&&"undefined"!=typeof HTMLCanvasElement&&this.canvas instanceof HTMLCanvasElement?this.canvas.remove():this.canvas=null,this.gpgpuCreatedLocally&&(this.gpgpu.program=null,this.gpgpu.dispose()),this.disposed=!0)}floatPrecision(){return null==this.floatPrecisionValue&&(this.floatPrecisionValue=Qs(()=>{if(!Le().get("WEBGL_RENDER_FLOAT32_ENABLED")){let e=Le().getBool("DEBUG");Le().set("DEBUG",!1);let t=this.abs(au(1e-8)).dataSync()[0];if(Le().set("DEBUG",e),t>0)return 32}return 16})),this.floatPrecisionValue}epsilon(){return 32===this.floatPrecision()?1e-7:1e-4}uploadToGPU(e){let t=this.texData.get(e),{shape:n,dtype:r,values:a,texture:s,usage:i,isPacked:o}=t;if(null!=s)return;let l,u=null!=this.activeTimers;u&&(l=va.now());let h=t.texShape;if(null==h&&(h=GO(n,o),t.texShape=h),null!=a){let e,s=UO(n),i=h[1],d=h[0],p=a instanceof Uint8Array||a instanceof Uint8ClampedArray;(o||!p)&&([i,d]=oO(h[0],h[1])),e=o?new OL(s,p):new ML(s,p);let c=p?[d,i]:h,f=this.makeTensorInfo(c,r),m=this.texData.get(f.dataId);m.usage=p?QM.PIXELS:QM.UPLOAD,m.texShape=c,this.gpgpu.uploadDenseMatrixToTexture(this.getTexture(f.dataId),i,d,a);let g=[[d,i]],y=this.runWebGLProgram(e,[f],r,g,!0),b=this.texData.get(y.dataId);t.texShape=b.texShape,t.isPacked=b.isPacked,t.usage=b.usage,Le().get("ENGINE_COMPILE_ONLY")?this.disposeData(y.dataId):(t.texture=b.texture,t.values=null,this.texData.delete(y.dataId)),this.disposeIntermediateTensorInfo(f),u&&(this.uploadWaitMs+=va.now()-l)}else{let e=this.acquireTexture(h,i,r,o);t.texture=e}}convertAndCacheOnCPU(e,t){let n=this.texData.get(e),{dtype:r}=n;return null!=t&&(n.values=function(e,t){if("float32"===t||"complex64"===t)return e;if("int32"===t||"bool"===t){let n="int32"===t?new Int32Array(e.length):new Uint8Array(e.length);for(let t=0;t<n.length;++t)n[t]=Math.round(e[t]);return n}throw new Error(`Unknown dtype ${t}`)}(t,r)),n.values}acquireTexture(e,t,n,r){if(this.numBytesInGPU+=this.computeBytes(e,n),!this.warnedAboutMemory&&this.numBytesInGPU>1024*this.numMBBeforeWarning*1024){let e=(this.numBytesInGPU/1024/1024).toFixed(2);this.warnedAboutMemory=!0,console.warn(`High memory usage in GPU: ${e} MB, most likely due to a memory leak`)}return this.textureManager.acquireTexture(e,t,r)}computeBytes(e,t){return e[0]*e[1]*va.bytesPerElement(t)}checkCompileCompletion(){for(let[,e]of Object.entries(this.binaryCache))this.checkCompletion_(e)}async checkCompileCompletionAsync(){let e=[];if(this.gpgpu.parallelCompilationExtension){for(let[,t]of Object.entries(this.binaryCache))e.push(this.checkCompletionAsync_(t));return Promise.all(e)}for(let[,t]of Object.entries(this.binaryCache)){let n=new Promise(e=>{try{this.checkCompletion_(t),e(!0)}catch(e){throw e}});e.push(n)}return Promise.all(e)}async checkCompletionAsync_(e){return this.gpgpu.gl.getProgramParameter(e.webGLProgram,this.gpgpu.parallelCompilationExtension.COMPLETION_STATUS_KHR)?this.checkCompletion_(e):(await Vf(),this.checkCompletionAsync_(e))}checkCompletion_(e){if(!1===this.gpgpu.gl.getProgramParameter(e.webGLProgram,this.gpgpu.gl.LINK_STATUS))throw console.log(this.gpgpu.gl.getProgramInfoLog(e.webGLProgram)),!1===this.gpgpu.gl.getShaderParameter(e.fragmentShader,this.gpgpu.gl.COMPILE_STATUS)?(bO(e.source,this.gpgpu.gl.getShaderInfoLog(e.fragmentShader)),new Error("Failed to compile fragment shader.")):new Error("Failed to link vertex and fragment shaders.");return!0}getUniformLocations(){for(let e of Object.values(this.binaryCache)){this.gpgpu.buildVao(e.webGLProgram);let{variablesLocations:t,customUniformLocations:n,infLoc:r,nanLoc:a,outShapeLocation:s,outShapeStridesLocation:i,outTexShapeLocation:o}=TL(this.gpgpu,e.program,e.webGLProgram);e.variablesLocations=t,e.customUniformLocations=n,e.infLoc=r,e.nanLoc=a,e.outShapeLocation=s,e.outShapeStridesLocation=i,e.outTexShapeLocation=o}}createTensorFromGPUData(e,t,n){e.channels=e.channels||"RGBA";let{texture:r,height:a,width:s,channels:i}=e,o=Zs().backend;if(!o.gpgpu.gl.isTexture(r))throw new Error("The texture is invalid. Also, please make sure the texture and the TFJS WebGL backend are using the same canvas. If you want to use your own custom canvas, you have to create and use the custom TFJS WebGL backend created from the canvas through 'new tf.MathBackendWebGL(customCanvas)'.");let l=o.writeTexture(r,t,n,a,s,i);return Zs().makeTensorFromDataId(l,t,n,o)}};wz.nextDataId=0;var kz="4.22.0";function Iz(){Le().set("WEBGL_FORCE_F16_TEXTURES",!0)}Cs.isBrowser()&&ui("webgl",()=>new wz,2);var Sz={forceHalfFloat:Iz},_z="\n  if (isnan(a)) return a;\n  if (isnan(b)) return b;\n",Nz=class{constructor(e,t,n){this.variableNames=["A","B"],this.outputShape=Gf.assertAndGetBroadcastShape(t,n),this.enableShapeUniforms=EL(this.outputShape.length),this.userCode=`\n      float binaryOperation(float a, float b) {\n        ${e}\n      }\n\n      void main() {\n        float a = getAAtOutCoords();\n        float b = getBAtOutCoords();\n        setOutput(binaryOperation(a, b));\n      }\n    `}},Tz="\n  result.r = isNaN.r ? NAN : result.r;\n  result.g = isNaN.g ? NAN : result.g;\n  result.b = isNaN.b ? NAN : result.b;\n  result.a = isNaN.a ? NAN : result.a;\n",Cz=class{constructor(e,t,n,r=!1){this.variableNames=["A","B"],this.supportsBroadcasting=!0,this.packedInputs=!0,this.packedOutput=!0,this.outputShape=Gf.assertAndGetBroadcastShape(t,n);let a=this.outputShape.length;this.enableShapeUniforms=EL(a);let s="";if(r)if(0===a||1===va.sizeFromShape(this.outputShape))s="\n          result.y = 0.;\n          result.z = 0.;\n          result.w = 0.;\n        ";else if(s=`\n          ${IL(a)} coords = getOutputCoords();\n        `,1===a)this.enableShapeUniforms?s+="\n            result.y = (coords + 1) >= outShape ? 0. : result.y;\n            result.z = 0.;\n            result.w = 0.;\n          ":s+=`\n            result.y = (coords + 1) >= ${this.outputShape[0]} ? 0. : result.y;\n            result.z = 0.;\n            result.w = 0.;\n          `;else{let e=rz("coords",a);this.enableShapeUniforms?s+=`\n            bool nextRowOutOfBounds =\n              (${e[a-2]} + 1) >= outShape[${a} - 2];\n            bool nextColOutOfBounds =\n              (${e[a-1]} + 1) >= outShape[${a} - 1];\n            result.y = nextColOutOfBounds ? 0. : result.y;\n            result.z = nextRowOutOfBounds ? 0. : result.z;\n            result.w = nextColOutOfBounds || nextRowOutOfBounds ? 0. : result.w;\n          `:s+=`\n            bool nextRowOutOfBounds =\n              (${e[a-2]} + 1) >= ${this.outputShape[a-2]};\n            bool nextColOutOfBounds =\n              (${e[a-1]} + 1) >= ${this.outputShape[a-1]};\n            result.y = nextColOutOfBounds ? 0. : result.y;\n            result.z = nextRowOutOfBounds ? 0. : result.z;\n            result.w = nextColOutOfBounds || nextRowOutOfBounds ? 0. : result.w;\n          `}this.userCode=`\n      vec4 binaryOperation(vec4 a, vec4 b) {\n        ${e}\n      }\n\n      void main() {\n        vec4 a = getAAtOutCoords();\n        vec4 b = getBAtOutCoords();\n\n        vec4 result = binaryOperation(a, b);\n        ${s}\n\n        setOutput(result);\n      }\n    `}};function Ez(e){let{inputs:t,backend:n}=e,{x:r}=t;return n.incRef(r.dataId),{dataId:r.dataId,shape:r.shape,dtype:r.dtype}}var Az={kernelName:an,backendName:"webgl",kernelFunc:Ez};function $z(e){let{inputs:t,backend:n}=e,{real:r,imag:a}=t,s=n.makeTensorInfo(r.shape,"complex64"),i=n.texData.get(s.dataId),o=Ez({inputs:{x:r},backend:n}),l=Ez({inputs:{x:a},backend:n});return i.complexTensorInfos={real:o,imag:l},s}var Rz={kernelName:mt,backendName:"webgl",kernelFunc:$z},Fz="return (a < 0.) ? b * a : a;",Dz="\n  vec4 aLessThanZero = vec4(lessThan(a, vec4(0.)));\n  return (aLessThanZero * (b * a)) + ((vec4(1.0) - aLessThanZero) * a);\n";var Mz={kernelName:dn,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{alpha:s}=r,i=n.makeTensorInfo([],"float32",va.createScalarValue(s,"float32")),o=Le().getBool("WEBGL_PACK_BINARY_OPERATIONS")?new Cz(Dz,a.shape,i.shape):new Nz(Fz,a.shape,i.shape),l=n.runWebGLProgram(o,[a,i],"float32");return n.disposeIntermediateTensorInfo(i),l}},Oz="return (a < 0.) ? b * a : a;",Lz="\n  vec4 aLessThanZero = vec4(lessThan(a, vec4(0.)));\n  return (aLessThanZero * (b * a)) + ((vec4(1.0) - aLessThanZero) * a);\n";var Pz={kernelName:Yn,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n}=e,{x:r,alpha:a}=t,s=Le().getBool("WEBGL_PACK_BINARY_OPERATIONS")?new Cz(Lz,r.shape,a.shape):new Nz(Oz,r.shape,a.shape);return n.runWebGLProgram(s,[r,a],"float32")}},zz="if (isnan(x)) return x;";function Bz({opSnippet:e,packedOpSnippet:t,cpuKernelImpl:n,dtype:r}){return({inputs:a,backend:s})=>{let i,{x:o}=a,l=s,u=r||o.dtype;if(l.shouldExecuteOnCPU([o])&&null!=n){let e=l.texData.get(o.dataId),t=n(e.values,u);return l.makeTensorInfo(o.shape,u,t)}return i=Le().getBool("WEBGL_PACK_UNARY_OPERATIONS")&&null!=t?new gz(o.shape,t):new hz(o.shape,e),l.runWebGLProgram(i,[o],u)}}function Wz({opSnippet:e,packedOpSnippet:t,checkOutOfBounds:n=!1,supportsComplex:r=!1,cpuKernelImpl:a,dtype:s}){return({inputs:i,backend:o})=>{let{a:l,b:u}=i,h=o;if(r&&"complex64"===l.dtype){let t=h.texData.get(l.dataId),n=h.texData.get(u.dataId),[r,a]=[[t.complexTensorInfos.real,n.complexTensorInfos.real],[t.complexTensorInfos.imag,n.complexTensorInfos.imag]].map(t=>{let[n,r]=t,a={dataId:n.dataId,dtype:n.dtype,shape:l.shape},s={dataId:r.dataId,dtype:r.dtype,shape:u.shape},i=new Nz(e,l.shape,u.shape);return h.runWebGLProgram(i,[a,s],ps(n.dtype,r.dtype))}),s=$z({inputs:{real:r,imag:a},backend:h});return h.disposeIntermediateTensorInfo(r),h.disposeIntermediateTensorInfo(a),s}let d,p=s||ps(l.dtype,u.dtype);if(("string"===l.dtype||"string"===u.dtype||h.shouldExecuteOnCPU([l,u]))&&null!=a){let e=h.texData.get(l.dataId).values,t=h.texData.get(u.dataId).values,n="string"===l.dtype?Gf.fromUint8ToStringArray(e):e,r="string"===l.dtype?Gf.fromUint8ToStringArray(t):t,[s,i]=a(l.shape,u.shape,n,r,p),o=h.makeTensorInfo(i,p);return h.texData.get(o.dataId).values=s,o}return d=Le().getBool("WEBGL_PACK_BINARY_OPERATIONS")&&null!=t?new Cz(t,l.shape,u.shape,n):new Nz(e,l.shape,u.shape),h.runWebGLProgram(d,[l,u],p)}}function Vz(e,t=!1){if("linear"===e)return"return x;";if("relu"===e)return t?"\n  vec4 result = x * vec4(greaterThanEqual(x, vec4(0.0)));\n  bvec4 isNaN = isnan(x);\n\n  result.r = isNaN.r ? x.r : result.r;\n  result.g = isNaN.g ? x.g : result.g;\n  result.b = isNaN.b ? x.b : result.b;\n  result.a = isNaN.a ? x.a : result.a;\n\n  return result;\n":cz;if("elu"===e)return t?"\n  vec4 result;\n\n  result.r = (x.r >= 0.0) ? x.r : (exp(x.r) - 1.0);\n  result.g = (x.g >= 0.0) ? x.g : (exp(x.g) - 1.0);\n  result.b = (x.b >= 0.0) ? x.b : (exp(x.b) - 1.0);\n  result.a = (x.a >= 0.0) ? x.a : (exp(x.a) - 1.0);\n\n  return result;\n":"return (x >= 0.0) ? x : (exp(x) - 1.0);";if("relu6"===e)return t?"\n  vec4 result = min(x, vec4(6.)) * vec4(greaterThanEqual(x, vec4(0.0)));\n  bvec4 isNaN = isnan(x);\n\n  result.r = isNaN.r ? x.r : result.r;\n  result.g = isNaN.g ? x.g : result.g;\n  result.b = isNaN.b ? x.b : result.b;\n  result.a = isNaN.a ? x.a : result.a;\n\n  return result;\n":fz;if("prelu"===e)return t?Lz:Oz;if("leakyrelu"===e)return t?Dz:Fz;if("sigmoid"===e)return"return 1.0 / (1.0 + exp(-1.0 * x));";throw new Error(`Activation ${e} has not been implemented for the WebGL backend.`)}var Uz=class{constructor(e,t,n,r=!1,a=!1,s=!1,i=null,o=!1,l=!1){this.variableNames=["matrixA","matrixB"],this.packedInputs=!0,this.packedOutput=!0,this.outputShape=n,this.enableShapeUniforms=EL(this.outputShape.length);let u=r?e[1]:e[2],h=Math.ceil(u/2),d=r?"i * 2, rc.y":"rc.y, i * 2",p=a?"rc.z, i * 2":"i * 2, rc.z",c=r?["a.xxyy","a.zzww"]:["a.xxzz","a.yyww"],f=a?["b.xzxz","b.ywyw"]:["b.xyxy","b.zwzw"],m="",g="";i&&(m=o?`vec4 activation(vec4 a) {\n          vec4 b = getPreluActivationWeightsAtOutCoords();\n          ${i}\n        }`:l?`vec4 activation(vec4 a) {\n          vec4 b = getLeakyreluAlphaAtOutCoords();\n          ${i}\n        }`:`vec4 activation(vec4 x) {\n          ${i}\n        }`,g="result = activation(result);");let y=s?"result += getBiasAtOutCoords();":"";s&&this.variableNames.push("bias"),o&&this.variableNames.push("preluActivationWeights"),l&&this.variableNames.push("leakyreluAlpha");let b="rc.x",x="rc.x";e[0]<t[0]?b=`imod(rc.x, ${e[0]})`:t[0]<e[0]&&(x=`imod(rc.x, ${t[0]})`),this.userCode=`\n      ${m}\n      // Don't use uniform for sharedDimensionPacked for performance.\n      const float sharedDimension = ${h}.0;\n\n      vec4 dot2x2ARowBCol(ivec3 rc) {\n        vec4 result = vec4(0);\n        int batchA = ${b};\n        int batchB = ${x};\n        for (int i = 0; i < ${h}; i++) {\n          vec4 a = getMatrixA(batchA, ${d});\n          vec4 b = getMatrixB(batchB, ${p});\n\n          // These swizzled products need to be separately added.\n          // See: https://github.com/tensorflow/tfjs/issues/1735\n          result += (${c[0]} * ${f[0]});\n          result += (${c[1]} * ${f[1]});\n        }\n        return result;\n      }\n\n      void main() {\n        ivec3 rc = getOutputCoords();\n        vec4 result = dot2x2ARowBCol(rc);\n\n        ${y}\n\n        ${g}\n\n        setOutput(result);\n      }\n    `}},Gz="return areal * breal - aimag * bimag;",Hz="return areal * bimag + aimag * breal;",jz=class{constructor(e,t,n){this.variableNames=["AReal","AImag","BReal","BImag"],this.outputShape=Gf.assertAndGetBroadcastShape(t,n),this.userCode=`\n      float binaryOpComplex(\n          float areal, float aimag, float breal, float bimag) {\n        ${e}\n      }\n\n      void main() {\n        float areal = getARealAtOutCoords();\n        float aimag = getAImagAtOutCoords();\n        float breal = getBRealAtOutCoords();\n        float bimag = getBImagAtOutCoords();\n        setOutput(binaryOpComplex(areal, aimag, breal, bimag));\n      }\n    `}},qz="return a * b;";function Kz(e){let t,{inputs:n,backend:r}=e,{a,b:s}=n,i=Gf.upcastType(a.dtype,s.dtype);if("complex64"===a.dtype){let e=r.texData.get(a.dataId),t=r.texData.get(s.dataId),n=new jz(Gz,a.shape,s.shape),i=new jz(Hz,a.shape,s.shape),o=[{dataId:e.complexTensorInfos.real.dataId,dtype:e.complexTensorInfos.real.dtype,shape:a.shape},{dataId:e.complexTensorInfos.imag.dataId,dtype:e.complexTensorInfos.imag.dtype,shape:a.shape},{dataId:t.complexTensorInfos.real.dataId,dtype:t.complexTensorInfos.real.dtype,shape:s.shape},{dataId:t.complexTensorInfos.imag.dataId,dtype:t.complexTensorInfos.imag.dtype,shape:s.shape}],l=r.runWebGLProgram(n,o,"float32"),u=r.runWebGLProgram(i,o,"float32"),h=$z({inputs:{real:l,imag:u},backend:r});return r.disposeIntermediateTensorInfo(l),r.disposeIntermediateTensorInfo(u),h}if(r.shouldExecuteOnCPU([a,s])){let e=r.texData.get(a.dataId),t=r.texData.get(s.dataId),[n,o]=EP(a.shape,s.shape,e.values,t.values,i),l=r.makeTensorInfo(o,i);return r.texData.get(l.dataId).values=n,l}return t=Le().getBool("WEBGL_PACK_BINARY_OPERATIONS")?new Cz(qz,a.shape,s.shape):new Nz(qz,a.shape,s.shape),r.runWebGLProgram(t,[a,s],i)}var Xz={kernelName:zn,backendName:"webgl",kernelFunc:Kz};function Zz(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{shape:s}=r,i=n,o=va.sizeFromShape(a.shape),l=va.inferFromImplicitShape(s,o),u=va.sizeFromShape(l);va.assert(o===u,()=>`The new shape (${l}) has ${u} elements and the old shape (${a.shape}) has ${o} elements. The new shape and old shape must have the same number of elements.`);let h=i.texData.get(a.dataId);return!h.isPacked||jO(a.shape,l)||null!==h.texture&&jO(h.shape,l)?(i.incRef(a.dataId),{dataId:a.dataId,shape:l,dtype:a.dtype}):function(e,t,n){let r=[WO(e.shape),...VO(e.shape)],a={dtype:e.dtype,shape:r,dataId:e.dataId},s=[WO(t),...VO(t)],i=new sz(s,r),o=[r],l=n.runWebGLProgram(i,[a],e.dtype,o,!0);return{dataId:l.dataId,shape:t,dtype:l.dtype}}(a,l,i)}var Yz={kernelName:ir,backendName:"webgl",kernelFunc:Zz},Jz=class{constructor(e,t){this.variableNames=["x"];let{windowSize:n,batchSize:r,inSize:a,outSize:s}=e;this.outputShape=[r,s];let i=4*Math.floor(n/4),o=n%4,l="sumValue += dot(values, ones);";if(null!=t){let e=1/t;l=`sumValue += dot(values * ${va.isInt(e)?e.toPrecision(2):e}, ones);`}let u="";a%n>0&&(u=`\n        if (inIdx < 0 || inIdx >= ${a}) {\n          return 0.0;\n        }\n      `),this.userCode=`\n      const vec4 ones = vec4(1.0, 1.0, 1.0, 1.0);\n\n      float getValue(int batch, int inIdx) {\n        ${u}\n        return getX(batch, inIdx);\n      }\n\n      void main() {\n        ivec2 coords = getOutputCoords();\n        int batch = coords[0];\n        int outIdx = coords[1];\n        int inOffset = outIdx * ${n};\n\n        float sumValue = 0.0;\n\n        for (int i = 0; i < ${i}; i += 4) {\n          int inIdx = inOffset + i;\n          vec4 values = vec4(\n            getValue(batch, inIdx),\n            getValue(batch, inIdx + 1),\n            getValue(batch, inIdx + 2),\n            getValue(batch, inIdx + 3)\n          );\n\n          ${l}\n        }\n\n        int inIdx = inOffset + ${i};\n        if (${1===o}) {\n          vec4 values = vec4(getValue(batch, inIdx), 0.0, 0.0, 0.0);\n\n          ${l}\n        } else if (${2===o}) {\n          vec4 values = vec4(\n            getValue(batch, inIdx),\n            getValue(batch, inIdx + 1), 0.0, 0.0);\n\n          ${l}\n        } else if (${3===o}) {\n          vec4 values = vec4(\n            getValue(batch, inIdx),\n            getValue(batch, inIdx + 1),\n            getValue(batch, inIdx + 2), 0.0);\n\n          ${l}\n        }\n        setOutput(sumValue);\n      }\n    `}},Qz=class{constructor(e,t){this.variableNames=["x"];let{windowSize:n,batchSize:r,inSize:a,outSize:s}=e;this.outputShape=[r,s];let i="0.0",o="";"prod"===t?i="1.0":"min"===t?(i="1.0 / 1e-20",o="min"):"max"===t&&(i="-1.0 / 1e-20",o="max");let l=`${t}(${t}(${t}(minMaxValue[0], minMaxValue[1]), minMaxValue[2]), minMaxValue[3])`;"sum"===t?l="sumValue":"prod"===t?l="prodValue":"all"===t?l="allValue":"any"===t&&(l="anyValue");let u=4*Math.floor(n/4),h=n%4,d=`\n      if (${"sum"===t}) {\n        sumValue += dot(values, ones);\n      } else if (${"prod"===t}) {\n        vec2 tmp = vec2(values[0], values[1]) * vec2(values[2], values[3]);\n        prodValue *= tmp[0] * tmp[1];\n      } else {\n        minMaxValue = ${o}(values, minMaxValue);\n        if (${"min"===t} || ${"max"===t}) {\n          minMaxValue = ${o}(values, minMaxValue);\n          bvec4 isNaN = isnan(values);\n          if (isNaN.r || isNaN.g || isNaN.b || isNaN.a) {\n            minMaxValue = vec4(NAN);\n          }\n        }\n      }\n    `,p="vec4";"all"===t?(i="1.0",d="\n        bool reducedAllValue = all(values);\n        float floatedReducedAllValue = float(reducedAllValue);\n        allValue = float(allValue >= 1.0 && floatedReducedAllValue >= 1.0);\n      ",p="bvec4"):"any"===t&&(i="0.0",d="\n        bool reducedAnyValue = any(values);\n        float floatedReducedAnyValue = float(reducedAnyValue);\n        anyValue = float(anyValue >= 1.0 || floatedReducedAnyValue >= 1.0);\n      ",p="bvec4");let c="";a%n>0&&(c=`\n        if (inIdx < 0 || inIdx >= ${a}) {\n          return initializationValue;\n        }\n      `),this.userCode=`\n      const float initializationValue = ${i};\n      const vec4 ones = vec4(1.0, 1.0, 1.0, 1.0);\n\n      float getValue(int batch, int inIdx) {\n        ${c}\n        return getX(batch, inIdx);\n      }\n\n      void main() {\n        ivec2 coords = getOutputCoords();\n        int batch = coords[0];\n        int outIdx = coords[1];\n        int inOffset = outIdx * ${n};\n\n        vec4 minMaxValue = vec4(${i});\n        float prodValue = 1.0;\n        float sumValue = 0.0;\n        float allValue = 1.0;\n        float anyValue = 0.0;\n\n        for (int i = 0; i < ${u}; i += 4) {\n          int inIdx = inOffset + i;\n          ${p} values = ${p}(\n            getValue(batch, inIdx),\n            getValue(batch, inIdx + 1),\n            getValue(batch, inIdx + 2),\n            getValue(batch, inIdx + 3)\n          );\n\n          ${d}\n        }\n\n        int inIdx = inOffset + ${u};\n        if (${1===h}) {\n          ${p} values = ${p}(\n            getValue(batch, inIdx),\n            initializationValue,\n            initializationValue,\n            initializationValue\n          );\n\n          ${d}\n        } else if (${2===h}) {\n          ${p} values = ${p}(\n            getValue(batch, inIdx),\n            getValue(batch, inIdx + 1),\n            initializationValue,\n            initializationValue\n          );\n\n          ${d}\n        } else if (${3===h}) {\n          ${p} values = ${p}(\n            getValue(batch, inIdx),\n            getValue(batch, inIdx + 1),\n            getValue(batch, inIdx + 2),\n            initializationValue\n          );\n\n          ${d}\n        }\n        setOutput(${l});\n      }\n    `}};function eB(e,t,n,r){let a=function(e){let t=[];for(;0===t.length||1!==t[t.length-1].outSize;){let n=t.length?t[t.length-1].outSize:e[1],r=Gf.computeOptimalWindowSize(n);t.push({inSize:n,windowSize:r,outSize:Math.ceil(n/r)})}return t}(e.shape),s=e;for(let i=0;i<a.length;i++){let o,l,{inSize:u,windowSize:h,outSize:d}=a[i];o="mean"===n?0===i?new Jz({windowSize:h,inSize:u,batchSize:e.shape[0],outSize:d},u):new Jz({windowSize:h,inSize:u,batchSize:e.shape[0],outSize:d}):new Qz({windowSize:h,inSize:u,batchSize:e.shape[0],outSize:d},n),l=s,s=r.runWebGLProgram(o,[s],t),l.dataId!==e.dataId&&r.disposeIntermediateTensorInfo(l)}return s}var tB=class{constructor(e,t){this.variableNames=["A"];let n=new Array(e.length);for(let r=0;r<n.length;r++)n[r]=e[t[r]];this.outputShape=n,this.rank=n.length;let r=IL(this.rank),a=function(e){let t=e.length;if(t>6)throw Error(`Transpose for rank ${t} is not yet supported`);let n=["resRC.x","resRC.y","resRC.z","resRC.w","resRC.u","resRC.v"],r=new Array(t);for(let t=0;t<e.length;t++)r[e[t]]=n[t];return r.join()}(t);this.userCode=`\n    void main() {\n      ${r} resRC = getOutputCoords();\n      setOutput(getA(${a}));\n    }\n    `}};var nB=class{constructor(e,t){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0;let n=new Array(e.length);for(let r=0;r<n.length;r++)n[r]=e[t[r]];if(this.outputShape=n,this.rank=n.length,this.rank>6)throw Error(`Packed transpose for rank ${this.rank} is not yet supported.`);let r=IL(this.rank),a=nz("rc",this.rank),s=new Array(this.rank);for(let e=0;e<t.length;e++)s[t[e]]=a[e];let i=`vec2(${s.slice(-2).join()})`,o=`++${a[this.rank-1]} < ${n[this.rank-1]}`,l=`getChannel(getA(${s.join()}), ${i})`;this.userCode=`\n    void main() {\n      ${r} rc = getOutputCoords();\n      vec4 result = vec4(0.);\n      result[0] = ${l};\n      if(${o}) {\n        result[1] = ${l};\n      }\n      --${a[this.rank-1]};\n      if(++${a[this.rank-2]} < ${n[this.rank-2]}) {\n        result[2] = ${l};\n        if(${o}) {\n          result[3] = ${l};\n        }\n      }\n      setOutput(result);\n    }\n    `}};function rB(e,t,n){let r=Le().getBool("WEBGL_PACK_ARRAY_OPERATIONS")?new nB(e.shape,t):new tB(e.shape,t);return n.runWebGLProgram(r,[e],e.dtype)}function aB(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{axis:s,keepDims:i}=r;return function(e,t,n,r){let a=t,s=e.shape.length,i=va.parseAxisParam(a,e.shape),o=i,l=Gf.getAxesPermutation(o,s),u=null!=l,h=e;u&&(h=rB(e,l,r),o=Gf.getInnerMostAxes(o.length,s)),Gf.assertAxesAreInnerMostDims("sum",o,s);let[d,p]=Gf.computeOutAndReduceShapes(h.shape,o),c=d;n&&(c=Gf.expandShapeToKeepDim(d,i));let f=va.sizeFromShape(p),m=Zz({inputs:{x:h},attrs:{shape:[va.sizeFromShape(e.shape)/f,f]},backend:r}),g=eB(m,cs(e.dtype),"sum",r),y=Zz({inputs:{x:g},attrs:{shape:c},backend:r});return r.disposeIntermediateTensorInfo(m),r.disposeIntermediateTensorInfo(g),u&&r.disposeIntermediateTensorInfo(h),y}(a,s,i,n)}var sB={kernelName:Tr,backendName:"webgl",kernelFunc:aB};function iB(e){let t,{inputs:n,backend:r,attrs:a}=e,{x:s}=n,{perm:i}=a,o=r,l=s.shape.length,u=new Array(l);for(let e=0;e<u.length;e++)u[e]=s.shape[i[e]];if(o.shouldExecuteOnCPU([s])){let e=o.texData.get(s.dataId).values,n=ez(e,s.shape,s.dtype,i,u);t=o.makeTensorInfo(u,s.dtype),o.texData.get(t.dataId).values=n}else t=rB(s,i,o);return t}var oB={kernelName:Xr,backendName:"webgl",kernelFunc:iB};function lB({a:e,b:t,transposeA:n,transposeB:r,backend:a,bias:s=null,preluActivationWeights:i=null,leakyreluAlpha:o=0,activation:l=null}){let u=e.shape.length,h=t.shape.length,d=n?e.shape[u-2]:e.shape[u-1],p=r?t.shape[h-1]:t.shape[h-2],c=n?e.shape[u-1]:e.shape[u-2],f=r?t.shape[h-2]:t.shape[h-1],m=e.shape.slice(0,-2),g=t.shape.slice(0,-2),y=va.sizeFromShape(m),b=va.sizeFromShape(g),x=Dl.assertAndGetBroadcastShape(e.shape.slice(0,-2),t.shape.slice(0,-2)).concat([c,f]);va.assert(d===p,()=>`Error in matMul: inner shapes (${d}) and (${p}) of Tensors with shapes ${e.shape} and ${t.shape} and transposeA=${n} and transposeB=${r} must match.`);let v,w=n?[y,d,c]:[y,c,d],k=r?[b,f,p]:[b,p,f],I=Zz({inputs:{x:e},backend:a,attrs:{shape:w}}),S=Zz({inputs:{x:t},backend:a,attrs:{shape:k}}),_=[I,S],N=Math.max(y,b),T=n?I.shape[1]:I.shape[2],C=null!=s,E=null!=i,A="leakyrelu"===l,$=null!=l?Vz(l,!0):null;if((1===c||1===f)&&T>1e3&&!1===(C||E||A||null!=$)){let e=I,t=S;n&&(e=iB({inputs:{x:I},backend:a,attrs:{perm:[0,2,1]}}),_.push(e)),r&&(t=iB({inputs:{x:S},backend:a,attrs:{perm:[0,2,1]}}),_.push(t));let s=1===f,i=e;1!==f&&(i=Zz({inputs:{x:e},backend:a,attrs:{shape:[N,T,1]}}),_.push(i));let o=1===f?2:1,l=t;s&&(l=Zz({inputs:{x:t},backend:a,attrs:{shape:[N,1,T]}}),_.push(l));let u=Kz({inputs:{a:i,b:l},backend:a});v=aB({inputs:{x:u},backend:a,attrs:{axis:o,keepDims:!0}}),_.push(u)}else{let l=ps(e.dtype,t.dtype),u=new Uz(w,k,[N,c,f],n,r,C,$,E,A),h=[I,S];if(null!=s&&h.push(s),E&&h.push(i),A){let e=a.makeTensorInfo([],"float32",va.createScalarValue(o,"float32"));h.push(e),_.push(e)}v=a.runWebGLProgram(u,h,l)}let R=Zz({inputs:{x:v},backend:a,attrs:{shape:x}});_.push(v);for(let e of _)a.disposeIntermediateTensorInfo(e);return R}var uB={kernelName:aa,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{a,b:s,bias:i,preluActivationWeights:o}=t,{transposeA:l,transposeB:u,activation:h,leakyreluAlpha:d}=r;return lB({a,b:s,transposeA:l,transposeB:u,backend:n,bias:i,preluActivationWeights:o,leakyreluAlpha:d,activation:h})}},hB="return abs(x);";var dB={kernelName:Ve,backendName:"webgl",kernelFunc:function(e){let t,{inputs:n,backend:r}=e,{x:a}=n;if(r.shouldExecuteOnCPU([a])&&"complex64"!==a.dtype){let e=r.texData.get(a.dataId),t=BP(e.values);return r.makeTensorInfo(a.shape,a.dtype,t)}return t=Le().getBool("WEBGL_PACK_UNARY_OPERATIONS")?new gz(a.shape,hB):new hz(a.shape,hB),r.runWebGLProgram(t,[a],a.dtype)}},pB=Bz({opSnippet:dz+"\n  if (abs(x) > 1.) {\n    return NAN;\n  }\n  return acos(x);\n"}),cB={kernelName:Ue,backendName:"webgl",kernelFunc:pB},fB=Bz({opSnippet:dz+"\n  if (x < 1.0) return NAN;\nreturn log(x + sqrt(x * x - 1.0));"}),mB={kernelName:Ge,backendName:"webgl",kernelFunc:fB},gB="return a + b;",yB=Wz({opSnippet:gB,packedOpSnippet:gB,supportsComplex:!0,cpuKernelImpl:oP}),bB={kernelName:He,backendName:"webgl",kernelFunc:yB},xB=class{constructor(e,t){this.outputShape=[],this.outputShape=e,this.variableNames=t.map((e,t)=>`T${t}`);let n=[];this.variableNames.forEach(e=>{n.push(`float v${e} = get${e}AtOutCoords();`)});let r=this.variableNames.map(e=>`v${e}`).join(" + ");this.userCode=`\n      void main() {\n        ${n.join("\n        ")}\n\n        float result = ${r};\n        setOutput(result);\n      }\n    `}},vB=class{constructor(e,t){this.outputShape=[],this.packedInputs=!0,this.packedOutput=!0,this.outputShape=e,this.variableNames=t.map((e,t)=>`T${t}`);let n=[];this.variableNames.forEach(e=>{n.push(`vec4 v${e} = get${e}AtOutCoords();`)});let r=this.variableNames.map(e=>`v${e}`).join(" + ");this.userCode=`\n      void main() {\n        ${n.join("\n        ")}\n\n        vec4 result = ${r};\n        setOutput(result);\n      }\n    `}};var wB={kernelName:je,backendName:"webgl",kernelFunc:function e(t){let{inputs:n,backend:r}=t,a=n;if(1===a.length)return Ez({inputs:{x:a[0]},backend:r});if(a.length>Le().getNumber("WEBGL_MAX_TEXTURES_IN_SHADER")){let t=Math.floor(a.length/2),n=e({inputs:a.slice(0,t),backend:r}),s=e({inputs:a.slice(t),backend:r});return e({inputs:[n,s],backend:r})}let s=a.map(e=>e.dtype).reduce((e,t)=>ps(e,t)),i=a.map(e=>e.shape),o=Le().getBool("WEBGL_PACK")?new vB(a[0].shape,i):new xB(a[0].shape,i);return r.runWebGLProgram(o,a,s)}};var kB={kernelName:qe,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{axis:s,keepDims:i}=r,o=a.shape.length,l=va.parseAxisParam(s,a.shape),u=l,h=Gf.getAxesPermutation(u,o),d=a;null!=h&&(d=iB({inputs:{x:a},backend:n,attrs:{perm:h}}),u=Gf.getInnerMostAxes(u.length,o)),Gf.assertAxesAreInnerMostDims("all",u,o);let p,[c,f]=Gf.computeOutAndReduceShapes(d.shape,u),m=Zz({inputs:{x:d},backend:n,attrs:{shape:[-1,va.sizeFromShape(f)]}}),g=eB(m,m.dtype,"all",n);if(i){p=Zz({inputs:{x:g},backend:n,attrs:{shape:Gf.expandShapeToKeepDim(c,l)}})}else p=Zz({inputs:{x:g},backend:n,attrs:{shape:c}});return n.disposeIntermediateTensorInfo(m),n.disposeIntermediateTensorInfo(g),null!=h&&n.disposeIntermediateTensorInfo(d),p}};var IB={kernelName:Ke,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{axis:s,keepDims:i}=r,o=a.shape.length,l=va.parseAxisParam(s,a.shape),u=l,h=Gf.getAxesPermutation(u,o),d=a;null!=h&&(d=iB({inputs:{x:a},backend:n,attrs:{perm:h}}),u=Gf.getInnerMostAxes(u.length,o)),Gf.assertAxesAreInnerMostDims("any",u,o);let p,[c,f]=Gf.computeOutAndReduceShapes(d.shape,u),m=Zz({inputs:{x:d},backend:n,attrs:{shape:[-1,va.sizeFromShape(f)]}}),g=eB(m,m.dtype,"any",n);if(i){p=Zz({inputs:{x:g},backend:n,attrs:{shape:Gf.expandShapeToKeepDim(c,l)}})}else p=Zz({inputs:{x:g},backend:n,attrs:{shape:c}});return n.disposeIntermediateTensorInfo(m),n.disposeIntermediateTensorInfo(g),null!=h&&n.disposeIntermediateTensorInfo(d),p}},SB=class{constructor(e,t,n){this.variableNames=["A"];let{windowSize:r,batchSize:a,outSize:s}=e;n||this.variableNames.push("bestIndicesA"),this.outputShape=[a,s];let i="max"===t?">":"<",o=n?"inOffset + i;":"round(getBestIndicesA(batch, inOffset + i));";this.userCode=`\n      void main() {\n        ivec2 coords = getOutputCoords();\n        int batch = coords[0];\n        int outIdx = coords[1];\n        int inOffset = outIdx * ${r};\n\n        int bestIndex = inOffset;\n        float bestValue = getA(batch, bestIndex);\n\n        for (int i = 0; i < ${r}; i++) {\n          int inIdx = ${o};\n          float candidate = getA(batch, inIdx);\n          if (candidate ${i} bestValue) {\n            bestValue = candidate;\n            bestIndex = inIdx;\n          }\n        }\n        setOutput(float(bestIndex));\n      }\n    `}},_B=class{constructor(e,t,n,r){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0,va.assert(e.length>2,()=>`Packed arg${n.charAt(0).toUpperCase()+n.slice(1)} supports only inputs with rank above 2.`);let a=e[e.length-1],s=Math.ceil(a/t);this.outputShape=e.slice(0,-1),s>1&&this.outputShape.push(s),r||this.variableNames.push("bestIndicesA");let i,o,l=this.outputShape,u=l.length,h=IL(u),d=rz("coords",u);if(1===s){o=u+1;let e=IL(o);i=`\n        ${e} sourceLocR = ${e}(${d.join()}, 0);\n        ++${d[u-1]};\n        ${e} sourceLocG = ${e}(${d.join()}, 0);\n        ++${d[u-2]};\n        ${e} sourceLocA = ${e}(${d.join()}, 0);\n        --${d[u-1]};\n        ${e} sourceLocB = ${e}(${d.join()}, 0);\n        --${d[u-2]};`}else o=u,i=`\n        ${h} sourceLocR = coords;\n        ++${d[u-1]};\n        ${h} sourceLocG = coords;\n        ++${d[u-2]};\n        ${h} sourceLocA = coords;\n        --${d[u-1]};\n        ${h} sourceLocB = coords;\n        --${d[u-2]};`;let p=["x","y","z","w","u","v"].slice(0,o),c="."+p[o-1],f=p.map(e=>"int "+e),m=rz("sourceLocR",o-1).concat("inIdx.r"),g=rz("sourceLocG",o-1).concat("inIdx.g"),y=rz("sourceLocB",o-1).concat("inIdx.b"),b=rz("sourceLocA",o-1).concat("inIdx.a"),x="max"===n?"greaterThan":"lessThan",v=r?"":`\n          inIdx = round(vec4(getBestIndicesAChannel(${m.join()}),\n                             getBestIndicesAChannel(${g.join()}),\n                             getBestIndicesAChannel(${y.join()}),\n                             getBestIndicesAChannel(${b.join()})));`,w=`vec4(\n            getAChannel(${m.join()}),\n            hasNextCol ? getAChannel(${g.join()}) : 0.,\n            hasNextRow ? getAChannel(${y.join()}) : 0.,\n            hasNextRow && hasNextCol ? getAChannel(${b.join()}) : 0.)`,k=r?"":`\n      float getBestIndicesAChannel(${f.join()}) {\n        return getChannel(getBestIndicesA(${p.join()}),\n                                          vec2(${p.slice(-2).join()}));\n      }`;this.userCode=`\n      float getAChannel(${f.join()}) {\n        return getChannel(getA(${p.join()}),\n                               vec2(${p.slice(-2).join()}));\n      }\n      ${k}\n      void main() {\n        ${h} coords = getOutputCoords();\n        bool hasNextCol = ${d[u-1]} < ${l[u-1]-1};\n        bool hasNextRow = ${d[u-2]} < ${l[u-2]-1};\n        ${i}\n        ivec4 srcIdx = ivec4(sourceLocR${c}, sourceLocG${c},\n          sourceLocB${c}, sourceLocA${c}) * ${t};\n        ivec4 inIdx = srcIdx;\n        vec4 bestIndex = vec4(inIdx);\n        vec4 bestValue = ${w};\n\n        for (int i = 0; i < ${t}; i++) {\n          inIdx = srcIdx;\n          ${v}\n          vec4 candidate = ${w};\n          bvec4 nan = isnan(candidate);\n          bvec4 replace = bvec4(\n            vec4(${x}(candidate, bestValue)) * (vec4(1.0) - vec4(nan)));\n\n          bestValue = vec4(replace.x  ? candidate.x : bestValue.x,\n                           replace.y  ? candidate.y : bestValue.y,\n                           replace.z  ? candidate.z : bestValue.z,\n                           replace.w  ? candidate.w : bestValue.w);\n          bestIndex = mix(bestIndex, vec4(inIdx), vec4(replace));\n          srcIdx++;\n        }\n        setOutput(bestIndex);\n      }\n    `}};function NB(e,t,n,r=null){let a=t.shape[0],s=t.shape[1];null!=r&&(a=r.shape[0],s=r.shape[1]);let i=Gf.computeOptimalWindowSize(s),o={windowSize:i,inSize:s,batchSize:a,outSize:Math.ceil(s/i)},l=new SB(o,n,null==r),u=[t];null!=r&&u.push(r);let h=e.runWebGLProgram(l,u,"int32");if(1===h.shape[1])return h;let d=NB(e,t,n,h);return e.disposeIntermediateTensorInfo(h),d}function TB(e,t,n,r=null){let a=null!=r?r.shape:t.shape,s=a[a.length-1],i=Gf.computeOptimalWindowSize(s),o=new _B(a,i,n,null==r),l=null==r?[t]:[t,r],u=e.runWebGLProgram(o,l,"int32");if(u.shape.length===t.shape.length){let r=TB(e,t,n,u);return e.disposeIntermediateTensorInfo(u),r}return u}function CB(e,t,n,r){let a=[n];if(Gf.assertAxesAreInnerMostDims("arg"+r.charAt(0).toUpperCase()+r.slice(1),a,t.shape.length),!Le().getBool("WEBGL_PACK_REDUCE")||t.shape.length<=2){let n=[],s=e.texData.get(t.dataId),i=t;null!==s&&s.isPacked&&(i=e.unpackTensor(t),n.push(i));let[o,l]=Gf.computeOutAndReduceShapes(i.shape,a),u=va.sizeFromShape(l),h=Zz({inputs:{x:i},backend:e,attrs:{shape:[-1,u]}});n.push(h);let d=NB(e,h,r);n.push(d);let p=Zz({inputs:{x:d},backend:e,attrs:{shape:o}});return n.forEach(t=>e.disposeIntermediateTensorInfo(t)),p}return TB(e,t,r)}var EB={kernelName:Xe,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{axis:s}=r,i=va.parseAxisParam(s,a.shape),o=Gf.getAxesPermutation(i,a.shape.length),l=a,u=[];null!=o&&(l=iB({inputs:{x:a},backend:n,attrs:{perm:o}}),u.push(l),i=Gf.getInnerMostAxes(i.length,l.shape.length)),Gf.assertAxesAreInnerMostDims("argMax",[i[0]],l.shape.length);let h=CB(n,l,i[0],"max");return u.forEach(e=>n.disposeIntermediateTensorInfo(e)),h}};var AB={kernelName:Ze,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{axis:s}=r,i=va.parseAxisParam(s,a.shape),o=Gf.getAxesPermutation(i,a.shape.length),l=a,u=[];null!=o&&(l=iB({inputs:{x:a},backend:n,attrs:{perm:o}}),u.push(l),i=Gf.getInnerMostAxes(i.length,l.shape.length)),Gf.assertAxesAreInnerMostDims("argMin",[i[0]],l.shape.length);let h=CB(n,l,i[0],"min");return u.forEach(e=>n.disposeIntermediateTensorInfo(e)),h}},$B=Bz({opSnippet:dz+"\n  if (abs(x) > 1.) {\n    return NAN;\n  }\n  return asin(x);\n"}),RB={kernelName:Ye,backendName:"webgl",kernelFunc:$B},FB=Bz({opSnippet:dz+"return log(x + sqrt(x * x + 1.0));"}),DB={kernelName:Je,backendName:"webgl",kernelFunc:FB},MB=Bz({opSnippet:dz+"\n  return atan(x);\n"}),OB={kernelName:Qe,backendName:"webgl",kernelFunc:MB},LB=Wz({opSnippet:_z+"\n  return atan(a, b);\n",packedOpSnippet:"\n  vec4 result = atan(a, b);\n  bvec4 isNaNA = isnan(a);\n  bvec4 isNaNB = isnan(b);\n  bvec4 isNaN = bvec4(isNaNA.x || isNaNB.x, isNaNA.y || isNaNB.y, isNaNA.z || isNaNB.z, isNaNA.w || isNaNB.w);\n  "+Tz+"\n  return result;\n"}),PB={kernelName:tt,backendName:"webgl",kernelFunc:LB},zB=Bz({opSnippet:dz+"\n  if ((x < -1.0) || (x > 1.0)) return NAN;\nreturn (log(1.0 + x) - log(1.0 - x)) / 2.0;"}),BB={kernelName:et,backendName:"webgl",kernelFunc:zB},WB=class{constructor(e,t,n,r=!1,a=!1){if(this.variableNames=["x"],"avg"===t&&n)throw new Error("Cannot compute positions for average pool.");let s=e.filterWidth,i=e.strideHeight,o=e.strideWidth,l=e.dilationHeight,u=e.dilationWidth,h=e.effectiveFilterHeight,d=e.effectiveFilterWidth,p=e.padInfo.top,c=e.padInfo.left;this.outputShape=e.outShape;let f="avg"===t,m=`((batch  * ${e.inHeight} + xR) * ${e.inWidth} + xC) * ${e.inChannels} + d`,g=`(xR * ${e.inWidth} + xC) * ${e.inChannels} + d`,y="0.0";if(f||(y="-1.0 / 1e-20"),n){let t=">=";return void(this.userCode=`\n        const ivec2 strides = ivec2(${i}, ${o});\n        const ivec2 pads = ivec2(${p}, ${c});\n\n        void main() {\n          ivec4 coords = getOutputCoords();\n          int batch = coords[0];\n          int d = coords[3];\n\n          ivec2 xRCCorner = coords.yz * strides - pads;\n          int xRCorner = xRCCorner.x;\n          int xCCorner = xRCCorner.y;\n\n          // max/min x(?, ?, d) to get y(yR, yC, d).\n          // ? = to be determined\n          float minMaxValue = 0.0;\n          float minMaxValueFound = 0.0;\n          int minMaxPosition = 0;\n          float avgValue = 0.0;\n\n          for (int wR = 0; wR < ${h};\n              wR += ${l}) {\n            int xR = xRCorner + wR;\n\n            if (xR < 0 || xR >= ${e.inHeight}) {\n              continue;\n            }\n\n            for (int wC = 0; wC < ${d};\n                wC += ${u}) {\n              int xC = xCCorner + wC;\n\n              if (xC < 0 || xC >= ${e.inWidth}) {\n                continue;\n              }\n\n              float value = getX(batch, xR, xC, d);\n\n              // If a min / max value has already been found, use it. If not,\n              // use the current value.\n              float currMinMaxValue = mix(\n                  value, minMaxValue, minMaxValueFound);\n              if (value ${t} currMinMaxValue) {\n                minMaxValue = value;\n                minMaxValueFound = 1.0;\n                minMaxPosition = ${r?a?m:g:`wR * ${d} + wC`};\n              }\n            }\n          }\n          setOutput(float(minMaxPosition));\n        }\n      `)}let b=`${t}(${t}(${t}(minMaxValue[0], minMaxValue[1]), minMaxValue[2]), minMaxValue[3])`;"avg"===t&&(b="avgValue / max(count, 1.0)");let x=4*Math.floor(s/4),v=s%4,w=`\n      if (${f}) {\n        avgValue += dot(values, ones);\n      } else {\n        minMaxValue = max(values, minMaxValue);\n      }\n    `;this.userCode=`\n      const ivec2 strides = ivec2(${i}, ${o});\n      const ivec2 pads = ivec2(${p}, ${c});\n      const float initializationValue = ${y};\n      const vec4 ones = vec4(1.0, 1.0, 1.0, 1.0);\n\n      float count = 0.0;\n\n      float getValue(int batch, int xR, int xC, int d) {\n        if (xC < 0 || xC >= ${e.inWidth}) {\n          return initializationValue;\n        }\n        count += 1.0;\n        return getX(batch, xR, xC, d);\n      }\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int batch = coords[0];\n        int d = coords[3];\n\n        ivec2 xRCCorner = coords.yz * strides - pads;\n        int xRCorner = xRCCorner.x;\n        int xCCorner = xRCCorner.y;\n\n        // max/min x(?, ?, d) to get y(yR, yC, d).\n        // ? = to be determined\n        vec4 minMaxValue = vec4(${y});\n        float avgValue = 0.0;\n        count = 0.0;\n\n        for (int wR = 0; wR < ${h};\n            wR += ${l}) {\n          int xR = xRCorner + wR;\n\n          if (xR < 0 || xR >= ${e.inHeight}) {\n            continue;\n          }\n\n          for (int wC = 0; wC < ${x}; wC += 4) {\n            int xC = xCCorner + wC * ${u};\n\n            vec4 values = vec4(\n              getValue(batch, xR, xC, d),\n              getValue(batch, xR, xC + ${u}, d),\n              getValue(batch, xR, xC + 2 * ${u}, d),\n              getValue(batch, xR, xC + 3 * ${u}, d)\n            );\n\n            ${w}\n          }\n\n          int xC = xCCorner + ${x};\n          if (${1===v}) {\n            vec4 values = vec4(\n              getValue(batch, xR, xC, d),\n              initializationValue,\n              initializationValue,\n              initializationValue\n            );\n\n            ${w}\n          } else if (${2===v}) {\n            vec4 values = vec4(\n              getValue(batch, xR, xC, d),\n              getValue(batch, xR, xC + ${u}, d),\n              initializationValue,\n              initializationValue\n            );\n\n            ${w}\n          } else if (${3===v}) {\n            vec4 values = vec4(\n              getValue(batch, xR, xC, d),\n              getValue(batch, xR, xC + ${u}, d),\n              getValue(batch, xR, xC + 2 * ${u}, d),\n              initializationValue\n            );\n\n            ${w}\n          }\n        }\n        setOutput(${b});\n      }\n    `}},VB=class{constructor(e,t,n,r=!1,a=!1){if(this.variableNames=["x"],"avg"===t&&n)throw new Error("Cannot compute positions for average pool.");let s=e.filterWidth,i=e.strideDepth,o=e.strideHeight,l=e.strideWidth,u=e.dilationDepth,h=e.dilationHeight,d=e.dilationWidth,p=e.effectiveFilterDepth,c=e.effectiveFilterHeight,f=e.effectiveFilterWidth,m=e.padInfo.front,g=e.padInfo.top,y=e.padInfo.left;this.outputShape=e.outShape;let b="avg"===t,x="0.0";if(b||(x="-1.0 / 1e-20"),n){let t=">=";return void(this.userCode=`\n        const ivec3 strides =\n            ivec3(${i}, ${o}, ${l});\n        const ivec3 pads = ivec3(${m}, ${g}, ${y});\n\n        void main() {\n          ivec5 coords = getOutputCoords();\n          int batch = coords.x;\n          int ch = coords.u;\n\n          ivec3 xCorner = ivec3(coords.y, coords.z, coords.w) * strides - pads;\n          int xDCorner = xCorner.x;\n          int xRCorner = xCorner.y;\n          int xCCorner = xCorner.z;\n\n          // max/min x(?, ?, ?, ch) to get y(yD, yR, yC, ch).\n          // ? = to be determined\n          float minMaxValue = 0.0;\n          float minMaxValueFound = 0.0;\n          int minMaxPosition = 0;\n\n          for (int wD = 0; wD < ${p};\n              wD += ${u}) {\n            int xD = xDCorner + wD;\n\n            if (xD < 0 || xD >= ${e.inDepth}) {\n              continue;\n            }\n\n            for (int wR = 0; wR < ${c};\n                wR += ${h}) {\n              int xR = xRCorner + wR;\n\n              if (xR < 0 || xR >= ${e.inHeight}) {\n                continue;\n              }\n\n              for (int wC = 0; wC < ${f};\n                  wC += ${d}) {\n                int xC = xCCorner + wC;\n\n                if (xC < 0 || xC >= ${e.inWidth}) {\n                  continue;\n                }\n\n                float value = getX(batch, xD, xR, xC, ch);\n\n                // If a min / max value has already been found, use it. If not,\n                // use the current value.\n                float currMinMaxValue = mix(\n                    value, minMaxValue, minMaxValueFound);\n                if (value ${t} currMinMaxValue) {\n                  minMaxValue = value;\n                  minMaxValueFound = 1.0;\n                  minMaxPosition = ${r?a?`(((batch * ${e.inDepth} + xD) * ${e.inHeight} + xR) * ${e.inWidth} + xC) * ${e.inChannels} + ch`:`((xD * ${e.inHeight} + xR) * ${e.inWidth} + xC) * ${e.inChannels} + ch`:`wD * ${c} * ${f} +\n                      wR * ${f} + wC`};\n                }\n              }\n            }\n          }\n          setOutput(float(minMaxPosition));\n        }\n      `)}let v=`${t}(${t}(${t}(minMaxValue[0], minMaxValue[1]), minMaxValue[2]), minMaxValue[3])`;"avg"===t&&(v="avgValue / max(count, 1.0)");let w=4*Math.floor(s/4),k=s%4,I=`\n      if (${b}) {\n        avgValue += dot(values, ones);\n      } else {\n        minMaxValue = max(values, minMaxValue);\n      }\n    `;this.userCode=`\n      const ivec3 strides =\n        ivec3(${i}, ${o}, ${l});\n      const ivec3 pads = ivec3(${m}, ${g}, ${y});\n      const float initializationValue = ${x};\n      const vec4 ones = vec4(1.0, 1.0, 1.0, 1.0);\n\n      float count = 0.0;\n\n      float getValue(int batch, int xD, int xR, int xC, int ch) {\n        if (xC < 0 || xC >= ${e.inWidth}) {\n          return initializationValue;\n        }\n        count += 1.0;\n        return getX(batch, xD, xR, xC, ch);\n      }\n\n      void main() {\n        ivec5 coords = getOutputCoords();\n        int batch = coords.x;\n        int ch = coords.u;\n\n        ivec3 xCorner = ivec3(coords.y, coords.z, coords.w) * strides - pads;\n        int xDCorner = xCorner.x;\n        int xRCorner = xCorner.y;\n        int xCCorner = xCorner.z;\n\n        // max/min x(?, ?, ?, d) to get y(yD, yR, yC, ch).\n        // ? = to be determined\n        vec4 minMaxValue = vec4(${x});\n        float avgValue = 0.0;\n        count = 0.0;\n\n        for (int wD = 0; wD < ${p};\n            wD += ${u}) {\n          int xD = xDCorner + wD;\n\n          if (xD < 0 || xD >= ${e.inDepth}) {\n            continue;\n          }\n\n          for (int wR = 0; wR < ${c};\n            wR += ${h}) {\n            int xR = xRCorner + wR;\n\n            if (xR < 0 || xR >= ${e.inHeight}) {\n              continue;\n            }\n\n            for (int wC = 0; wC < ${w}; wC += 4) {\n              int xC = xCCorner + wC * ${d};\n\n              vec4 values = vec4(\n                getValue(batch, xD, xR, xC, ch),\n                getValue(batch, xD, xR, xC + ${d}, ch),\n                getValue(batch, xD, xR, xC + 2 * ${d}, ch),\n                getValue(batch, xD, xR, xC + 3 * ${d}, ch)\n              );\n\n              ${I}\n            }\n\n            int xC = xCCorner + ${w};\n            if (${1===k}) {\n              vec4 values = vec4(\n                getValue(batch, xD, xR, xC, ch),\n                initializationValue,\n                initializationValue,\n                initializationValue\n              );\n\n              ${I}\n            } else if (${2===k}) {\n              vec4 values = vec4(\n                getValue(batch, xD, xR, xC, ch),\n                getValue(batch, xD, xR, xC + ${d}, ch),\n                initializationValue,\n                initializationValue\n              );\n\n              ${I}\n            } else if (${3===k}) {\n              vec4 values = vec4(\n                getValue(batch, xD, xR, xC, ch),\n                getValue(batch, xD, xR, xC + ${d}, ch),\n                getValue(batch, xD, xR, xC + 2 * ${d}, ch),\n                initializationValue\n              );\n\n              ${I}\n            }\n          }\n        }\n        setOutput(${v});\n      }\n    `}};var UB={kernelName:nt,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t;aL(a,"avgPool");let{filterSize:s,strides:i,pad:o,dimRoundingMode:l}=r;va.assert(Gf.eitherStridesOrDilationsAreOne(i,1),()=>`Error in avgPool: Either strides or dilations must be 1. Got strides ${i} and dilations '1'`);let u=Gf.computePool2DInfo(a.shape,s,i,1,o,l);if(1===u.filterWidth&&1===u.filterHeight&&va.arraysEqual(u.inShape,u.outShape))return Ez({inputs:{x:a},backend:n});let h=new WB(u,"avg",!1);return n.runWebGLProgram(h,[a],"float32")}};var GB={kernelName:at,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{filterSize:s,strides:i,pad:o,dimRoundingMode:l,dataFormat:u}=r,h=Gf.computePool3DInfo(a.shape,s,i,[1,1,1],o,l,u),d=new VB(h,"avg",!1);return n.runWebGLProgram(d,[a],"float32")}},HB=class{constructor(e){this.variableNames=["dy"],this.outputShape=e.inShape;let t=e.filterHeight,n=e.filterWidth,r=e.strideHeight,a=e.strideWidth,s=e.dilationHeight,i=e.dilationWidth,o=e.effectiveFilterHeight,l=e.effectiveFilterWidth,u=o-1-e.padInfo.top,h=l-1-e.padInfo.left,d=1/(t*n);this.userCode=`\n      const ivec2 pads = ivec2(${u}, ${h});\n      const float avgMultiplier = float(${d});\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int b = coords[0];\n        int d = coords[3];\n\n        ivec2 dyRCCorner = coords.yz - pads;\n        int dyRCorner = dyRCCorner.x;\n        int dyCCorner = dyRCCorner.y;\n\n        // Convolve dy(?, ?, d) with pos mask(:, :, d) to get dx(xR, xC, d).\n        // ? = to be determined. : = across all values in that axis.\n        float dotProd = 0.0;\n        for (int wR = 0; wR < ${o};\n            wR += ${s}) {\n          float dyR = float(dyRCorner + wR) / ${r}.0;\n\n          if (dyR < 0.0 || dyR >= ${e.outHeight}.0 || fract(dyR) > 0.0) {\n            continue;\n          }\n          int idyR = int(dyR);\n\n          for (int wC = 0; wC < ${l};\n            wC+= ${i}) {\n            float dyC = float(dyCCorner + wC) / ${a}.0;\n\n            if (dyC < 0.0 || dyC >= ${e.outWidth}.0 ||\n                fract(dyC) > 0.0) {\n              continue;\n            }\n            int idyC = int(dyC);\n\n            float dyValue = getDy(b, idyR, idyC, d);\n\n            dotProd += dyValue * avgMultiplier;\n          }\n        }\n        setOutput(dotProd);\n      }\n    `}},jB=class{constructor(e){this.variableNames=["dy"],this.outputShape=e.inShape;let t=e.filterDepth,n=e.filterHeight,r=e.filterWidth,a=e.strideDepth,s=e.strideHeight,i=e.strideWidth,o=e.dilationDepth,l=e.dilationHeight,u=e.dilationWidth,h=e.effectiveFilterDepth,d=e.effectiveFilterHeight,p=e.effectiveFilterWidth,c=h-1-e.padInfo.front,f=d-1-e.padInfo.top,m=p-1-e.padInfo.left,g=1/(t*n*r);this.userCode=`\n      const ivec3 pads = ivec3(${c}, ${f}, ${m});\n      const float avgMultiplier = float(${g});\n\n      void main() {\n        ivec5 coords = getOutputCoords();\n        int batch = coords.x;\n        int ch = coords.u;\n\n        ivec3 dyCorner = ivec3(coords.y, coords.z, coords.w) - pads;\n        int dyDCorner = dyCorner.x;\n        int dyRCorner = dyCorner.y;\n        int dyCCorner = dyCorner.z;\n\n        // Convolve dy(?, ?, ?, d) with pos mask(:, :, :, ch) to get\n        // dx(xD, xR, xC, ch).\n        // ? = to be determined. : = across all values in that axis.\n        float dotProd = 0.0;\n\n        for (int wD = 0; wD < ${h};\n            wD += ${o}) {\n          float dyD = float(dyDCorner + wD) / ${a}.0;\n\n          if (dyD < 0.0 || dyD >= ${e.outDepth}.0 || fract(dyD) > 0.0) {\n            continue;\n          }\n          int idyD = int(dyD);\n\n          for (int wR = 0; wR < ${d};\n              wR += ${l}) {\n            float dyR = float(dyRCorner + wR) / ${s}.0;\n\n            if (dyR < 0.0 || dyR >= ${e.outHeight}.0 ||\n                fract(dyR) > 0.0) {\n              continue;\n            }\n            int idyR = int(dyR);\n\n            for (int wC = 0; wC < ${p};\n                wC += ${u}) {\n              float dyC = float(dyCCorner + wC) / ${i}.0;\n\n              if (dyC < 0.0 || dyC >= ${e.outWidth}.0 ||\n                  fract(dyC) > 0.0) {\n                continue;\n              }\n              int idyC = int(dyC);\n\n              float dyValue = getDy(batch, idyD, idyR, idyC, ch);\n\n              dotProd += dyValue * avgMultiplier;\n            }\n          }\n        }\n        setOutput(dotProd);\n      }\n    `}};var qB={kernelName:st,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{dy:a,input:s}=t,i=s,{filterSize:o,strides:l,pad:u,dimRoundingMode:h}=r,d=Gf.computePool3DInfo(i.shape,o,l,[1,1,1],u,h),p=new jB(d);return n.runWebGLProgram(p,[a],i.dtype)}};var KB={kernelName:rt,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{dy:a,input:s}=t,i=s;aL([a,s],"avgPoolGrad");let{filterSize:o,strides:l,pad:u}=r,h=Gf.computePool2DInfo(i.shape,o,l,1,u),d=new HB(h);return n.runWebGLProgram(d,[a],i.dtype)}};var XB={kernelName:it,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{a,b:s}=t,{transposeA:i,transposeB:o}=r;return lB({a,b:s,transposeA:i,transposeB:o,backend:n})}},ZB=class{constructor(e,t,n,r,a,s){this.outputShape=[],this.variableNames=["x","mean","variance"],Gf.assertAndGetBroadcastShape(e,t),Gf.assertAndGetBroadcastShape(e,n);let i="0.0";null!=r&&(Gf.assertAndGetBroadcastShape(e,r),this.variableNames.push("offset"),i="getOffsetAtOutCoords()");let o="1.0";null!=a&&(Gf.assertAndGetBroadcastShape(e,a),this.variableNames.push("scale"),o="getScaleAtOutCoords()"),this.outputShape=e,this.userCode=`\n      void main() {\n        float x = getXAtOutCoords();\n        float mean = getMeanAtOutCoords();\n        float variance = getVarianceAtOutCoords();\n        float offset = ${i};\n        float scale = ${o};\n        float inv = scale * inversesqrt(variance + float(${s}));\n        setOutput(dot(vec3(x, -mean, offset), vec3(inv, inv, 1)));\n      }\n    `}},YB=class{constructor(e,t,n,r,a,s){this.packedInputs=!0,this.packedOutput=!0,this.variableNames=["x","mean","variance"],Gf.assertAndGetBroadcastShape(e,t),Gf.assertAndGetBroadcastShape(e,n);let i="vec4(0.0)";null!=r&&(Gf.assertAndGetBroadcastShape(e,r),this.variableNames.push("offset"),i="getOffsetAtOutCoords()");let o="vec4(1.0)";null!=a&&(Gf.assertAndGetBroadcastShape(e,a),this.variableNames.push("scale"),o="getScaleAtOutCoords()"),this.outputShape=e,this.userCode=`\n      void main() {\n        vec4 offset = ${i};\n        vec4 scale = ${o};\n\n        vec4 x = getXAtOutCoords();\n        vec4 mean = getMeanAtOutCoords();\n        vec4 variance = getVarianceAtOutCoords();\n\n        vec4 inv = scale * inversesqrt(variance + vec4(${s}));\n\n        setOutput((x - mean) * inv + offset);\n      }\n    `}},JB={kernelName:Qt,backendName:"webgl",kernelFunc:({inputs:e,backend:t,attrs:n})=>{let{x:r,mean:a,variance:s,offset:i,scale:o}=e;va.assert(a.shape.length===s.shape.length,()=>"Batch normalization gradient requires mean and variance to have equal ranks."),va.assert(null==i||a.shape.length===i.shape.length,()=>"Batch normalization gradient requires mean and offset to have equal ranks."),va.assert(null==o||a.shape.length===o.shape.length,()=>"Batch normalization gradient requires mean and scale to have equal ranks.");let{varianceEpsilon:l}=n;null==l&&(l=.001);let u=[r,a,s],h=null;null!=i&&(h=i.shape,u.push(i));let d=null;null!=o&&(d=o.shape,u.push(o));let p=Le().getBool("WEBGL_PACK_NORMALIZATION")?new YB(r.shape,a.shape,s.shape,h,d,l):new ZB(r.shape,a.shape,s.shape,h,d,l);return t.runWebGLProgram(p,u,u[0].dtype)}},QB=class{constructor(e){this.variableNames=["source"],this.outputShape=e,this.rank=e.length;let t=IL(this.rank);this.customUniforms=[{name:"start",arrayIndex:this.rank,type:"int"}];let n,r=function(e){if(1===e)return"sourceLoc";if(e<=6)return eW.slice(0,e).map(e=>"sourceLoc."+e).join(",");throw Error(`Slicing for rank ${e} is not yet supported`)}(this.rank);n=`\n        ${t} sourceLoc;\n        ${t} coords = getOutputCoords();\n        ${e.map((e,t)=>`sourceLoc.${eW[t]} = start[${t}] + coords.${eW[t]};`).join("\n")}\n      `,this.userCode=`\n      void main() {\n        ${n}\n        setOutput(getSource(${r}));\n      }\n    `}},eW=["x","y","z","w","u","v"];var tW=class{constructor(e){this.variableNames=["source"],this.packedInputs=!0,this.packedOutput=!0,this.outputShape=e,this.rank=e.length,this.customUniforms=[{name:"start",arrayIndex:this.rank,type:"int"}];let t=IL(this.rank),n=rz("coords",this.rank),r=rz("sourceLoc",this.rank),a=1===this.rank?"sourceLoc":`vec2(${r.slice(-2).join()})`,s=`getChannel(getSource(${r.join()}), ${a})`,i=`\n      result.x = ${s};\n      if (++${n[this.rank-1]} < ${e[this.rank-1]}) {\n        ++${r[this.rank-1]};\n        result.y = ${s};\n        --${r[this.rank-1]};\n      }\n    `,o=1===this.rank?"":`\n      --${n[this.rank-1]};\n      if (++${n[this.rank-2]} < ${e[this.rank-2]}) {\n        ++${r[this.rank-2]};\n        result.z = ${s};\n        if (++${n[this.rank-1]} < ${e[this.rank-1]}) {\n          ++${r[this.rank-1]};\n          result.w = ${s};\n        }\n      }\n    `,l=this.rank<=4?`sourceLoc = coords +\n            ${t}(${e.map((e,t)=>`start[${t}]`).join()});`:e.map((e,t)=>`${r[t]} = ${n[t]} + start[${t}];`).join("\n");this.userCode=`\n      void main() {\n        ${t} coords = getOutputCoords();\n        ${t} sourceLoc;\n        ${l}\n        vec4 result = vec4(0.);\n        ${i}\n        ${o}\n        setOutput(result);\n      }\n    `}};function nW(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{begin:s,size:i}=r,[o,l]=vf.parseSliceParams(a,s,i);if(vf.assertParamsValid(a,o,l),0===va.sizeFromShape(l))return n.makeTensorInfo(l,a.dtype,[]);if(n.shouldExecuteOnCPU([a])||"string"===a.dtype){let e=n.texData.get(a.dataId),t=WP(e.values,o,l,a.shape,a.dtype);return n.makeTensorInfo(l,a.dtype,t)}let{isPacked:u}=n.texData.get(a.dataId),h=vf.isSliceContinous(a.shape,o,l);if(u||!h){let e=Le().getBool("WEBGL_PACK_ARRAY_OPERATIONS")?new tW(l):new QB(l),t=[o];return n.runWebGLProgram(e,[a],a.dtype,t)}return n.uploadToGPU(a.dataId),function(e,t,n,r){let a=r.texData.get(e.dataId),s=r.makeTensorInfo(n,e.dtype),i=r.texData.get(s.dataId);Object.assign(i,a),i.refCount=1,i.shape=n,i.dtype=e.dtype;let o=vf.computeFlatOffset(t,va.computeStrides(e.shape));a.slice&&(o+=a.slice.flatOffset),i.slice={flatOffset:o,origDataId:a.slice&&a.slice.origDataId||e.dataId};let l=r.dataRefCount.get(i.slice.origDataId)||1;return r.dataRefCount.set(i.slice.origDataId,l+1),s}(a,o,l,n)}var rW={kernelName:vr,backendName:"webgl",kernelFunc:nW},aW={kernelName:ot,backendName:"webgl",kernelFunc:e=>{let{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{blockShape:s,crops:i}=r;va.assert(a.shape.length<=4,()=>"batchToSpaceND for rank > 4 with a WebGL backend not implemented yet");let o=s.reduce((e,t)=>e*t),l=Gf.getReshaped(a.shape,s,o),u=Gf.getPermuted(l.length,s.length),h=Gf.getReshapedPermuted(a.shape,s,o),d=Gf.getSliceBeginCoords(i,s.length),p=Gf.getSliceSize(h,i,s.length),c=[],f=Zz({inputs:{x:a},backend:n,attrs:{shape:l}}),m=iB({inputs:{x:f},backend:n,attrs:{perm:u}}),g=Zz({inputs:{x:m},backend:n,attrs:{shape:h}}),y=nW({inputs:{x:g},backend:n,attrs:{begin:d,size:p}});return c.push(f),c.push(m),c.push(g),c.forEach(e=>n.disposeIntermediateTensorInfo(e)),y}};var sW={kernelName:lt,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a,weights:s}=t,{size:i}=r,o=n.readSync(a.dataId),l=n.readSync(s.dataId),u=lP(o,l,s.dtype,s.shape,i);return n.makeTensorInfo([i],s.dtype,u)}};var iW={kernelName:ut,backendName:"webgl",kernelFunc:function(e){let t,{inputs:n,backend:r}=e,{a,b:s}=n,i=Le().getBool("WEBGL_PACK_BINARY_OPERATIONS"),o=Le().getNumber("WEBGL_VERSION");if(r.shouldExecuteOnCPU([a,s])||1===o){let e=r.texData.get(a.dataId).values,t=r.texData.get(s.dataId).values,[n,i]=hP(a.shape,s.shape,e,t,a.dtype),o=r.makeTensorInfo(i,a.dtype);return r.texData.get(o.dataId).values=n,o}return t=i?new Cz("\n  int r = int(a.r) & int(b.r);\n  int g = int(a.g) & int(b.g);\n  int rb = int(a.b) & int(b.b);\n  int ra = int(a.a) & int(b.a);\n  return vec4(r, g, rb, ra);\n",a.shape,s.shape,!1):new Nz("\n  return float(int(a.r) & int(b.r));\n",a.shape,s.shape),r.runWebGLProgram(t,[a,s],a.dtype)}};var oW={kernelName:dt,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n}=e,{s0:r,s1:a}=t,s=n.readSync(r.dataId),i=n.readSync(a.dataId),o=Gf.assertAndGetBroadcastShape(Array.from(s),Array.from(i));return n.makeTensorInfo([o.length],"int32",Int32Array.from(o))}},lW=Wz({opSnippet:"return float(a != b);",cpuKernelImpl:$P,dtype:"bool"}),uW={kernelName:Wn,backendName:"webgl",kernelFunc:lW};function hW(e){let{inputs:t,backend:n}=e,{input:r}=t;return Ez({inputs:{x:n.texData.get(r.dataId).complexTensorInfos.real},backend:n})}var dW={kernelName:rr,backendName:"webgl",kernelFunc:hW};var pW={kernelName:pt,backendName:"webgl",kernelFunc:function e(t){let{inputs:n,backend:r,attrs:a}=t,{x:s}=n,{dtype:i}=a;if("complex64"===i){if("complex64"===s.dtype)return Ez({inputs:{x:s},backend:r});let t=nh(s.shape),n=e({inputs:{x:s},backend:r,attrs:{dtype:"float32"}}),a=$z({inputs:{real:n,imag:t},backend:r});return t.dispose(),r.disposeIntermediateTensorInfo(n),a}if("complex64"===s.dtype){let t=hW({inputs:{input:s},backend:r}),n=e({inputs:{x:t},backend:r,attrs:{dtype:i}});return r.disposeIntermediateTensorInfo(t),n}if(!va.hasEncodingLoss(s.dtype,i)){let e=Ez({inputs:{x:s},backend:r});return{dataId:e.dataId,shape:e.shape,dtype:i}}if(r.shouldExecuteOnCPU([s])){let e=r.texData.get(s.dataId).values,[t,n,a]=dP(e,s.shape,s.dtype,i);return r.makeTensorInfo(t,n,a)}if("int32"===i)return function(e,t){let n=new hz(e.shape,"return float(int(x));"),r=t.runWebGLProgram(n,[e],"int32");return{dataId:r.dataId,shape:r.shape,dtype:r.dtype}}(s,r);if("bool"===i){let e=r.makeTensorInfo([],"bool",va.getTypedArrayFromDType("bool",1)),t=lW({inputs:{a:s,b:e},backend:r});return r.disposeIntermediateTensorInfo(e),t}throw new Error(`Error in Cast: failed to cast ${s.dtype} to ${i}`)}},cW="return ceil(x);",fW=Bz({opSnippet:cW,packedOpSnippet:cW,cpuKernelImpl:pP}),mW={kernelName:ct,backendName:"webgl",kernelFunc:fW},gW=class{constructor(e){this.variableNames=["A"],this.customUniforms=[{name:"minVal",type:"float"},{name:"maxVal",type:"float"}],this.outputShape=e,this.userCode="\n\n      void main() {\n        float value = getAAtOutCoords();\n        if (isnan(value)) {\n          setOutput(value);\n          return;\n        }\n\n        setOutput(clamp(value, minVal, maxVal));\n      }\n    "}},yW=class{constructor(e){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0,this.customUniforms=[{name:"minVal",type:"float"},{name:"maxVal",type:"float"}],this.outputShape=e,this.userCode="\n      void main() {\n        vec4 value = getAAtOutCoords();\n\n        if (any(isnan(value))) {\n          setOutput(value);\n          return;\n        }\n\n        setOutput(clamp(value, vec4(minVal), vec4(maxVal)));\n      }\n    "}};var bW={kernelName:ft,backendName:"webgl",kernelFunc:function(e){let t,{inputs:n,backend:r,attrs:a}=e,{x:s}=n,{clipValueMin:i,clipValueMax:o}=a;t=Le().getBool("WEBGL_PACK_CLIP")?new yW(s.shape):new gW(s.shape);let l=[[i],[o]];return r.runWebGLProgram(t,[s],s.dtype,l)}},xW=class{constructor(e){this.variableNames=["real","imag"],this.outputShape=e,this.userCode="\n      void main() {\n        float re = abs(getRealAtOutCoords());\n        float im = abs(getImagAtOutCoords());\n        float mx = max(re, im);\n\n        // sadly the length function in glsl is not underflow-safe\n        // (at least not on Intel GPUs). So the safe solution is\n        // to ensure underflow-safety in all cases.\n        setOutput(\n          mx == 0.0 ? 0.0 : mx * length(vec2(1, min(re, im)/mx))\n        );\n      }\n    "}};function vW(e,t){return{dataId:t.dataId,dtype:t.dtype,shape:e.shape}}var wW={kernelName:gt,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n}=e,{x:r}=t,a=n.texData.get(r.dataId),s=new xW(r.shape),i=[vW(r,a.complexTensorInfos.real),vW(r,a.complexTensorInfos.imag)];return n.runWebGLProgram(s,i,i[0].dtype)}},kW=class{constructor(e){this.outputShape=[],this.outputShape=Gf.computeOutShape(e,1),this.variableNames=e.map((e,t)=>`T${t}`);let t=new Array(e.length-1);t[0]=e[0][1];for(let n=1;n<t.length;n++)t[n]=t[n-1]+e[n][1];let n=[`if (yC < ${t[0]}) setOutput(getT0(yR, yC));`];for(let e=1;e<t.length;e++){let r=t[e-1];n.push(`else if (yC < ${t[e]}) setOutput(getT${e}(yR, yC-${r}));`)}let r=t.length,a=t[t.length-1];n.push(`else setOutput(getT${r}(yR, yC-${a}));`),this.userCode=`\n      void main() {\n        ivec2 coords = getOutputCoords();\n        int yR = coords.x;\n        int yC = coords.y;\n\n        ${n.join("\n        ")}\n      }\n    `}},IW=class{constructor(e,t){this.packedInputs=!0,this.packedOutput=!0,this.outputShape=[],this.outputShape=Gf.computeOutShape(e,t);let n=this.outputShape,r=n.length,a=IL(r),s=rz("coords",r),i=["x","y","z","w","u","v"].slice(0,r);this.variableNames=e.map((e,t)=>`T${t}`);let o=new Array(e.length-1);o[0]=e[0][t];for(let n=1;n<o.length;n++)o[n]=o[n-1]+e[n][t];let l=i[t],u=i.slice(-2),h=i.join(),d=`if (${l} < ${o[0]}) {\n        return getChannel(\n            getT0(${h}), vec2(${u.join()}));\n        }`;for(let e=1;e<o.length;e++){let t=o[e-1];d+=`\n        if (${l} < ${o[e]}  && ${l} >= ${o[e-1]}) {\n          return getChannel(\n            getT${e}(${SW(i,l,t)}),\n            vec2(${SW(u,l,t)}));\n        }`}let p=o.length,c=o[o.length-1];d+=`\n        return getChannel(\n          getT${p}(${SW(i,l,c)}),\n          vec2(${SW(u,l,c)}));`,this.userCode=`\n      float getValue(${i.map(e=>"int "+e)}) {\n        ${d}\n      }\n\n      void main() {\n        ${a} coords = getOutputCoords();\n        vec4 result = vec4(getValue(${s}), 0., 0., 0.);\n\n        ${s[r-1]} = ${s[r-1]} + 1;\n        if (${s[r-1]} < ${n[r-1]}) {\n          result.g = getValue(${s});\n        }\n\n        ${s[r-2]} = ${s[r-2]} + 1;\n        if (${s[r-2]} < ${n[r-2]}) {\n          result.a = getValue(${s});\n        }\n\n        ${s[r-1]} = ${s[r-1]} - 1;\n        if (${s[r-2]} < ${n[r-2]} &&\n            ${s[r-1]} < ${n[r-1]}) {\n          result.b = getValue(${s});\n        }\n        setOutput(result);\n      }\n    `}};function SW(e,t,n){let r=e.indexOf(t);return e.map((e,t)=>t===r?`${e} - ${n}`:e).join()}function _W(e){let{inputs:t,backend:n}=e,{input:r}=t;return Ez({inputs:{x:n.texData.get(r.dataId).complexTensorInfos.imag},backend:n})}var NW={kernelName:on,backendName:"webgl",kernelFunc:_W};function TW(e,t,n){let r=e[0].dtype;if("complex64"===r){let r=e.map(e=>hW({inputs:{input:e},backend:n})),a=e.map(e=>_W({inputs:{input:e},backend:n})),s=TW(r,t,n),i=TW(a,t,n),o=$z({inputs:{real:s,imag:i},backend:n});return r.forEach(e=>n.disposeIntermediateTensorInfo(e)),a.forEach(e=>n.disposeIntermediateTensorInfo(e)),n.disposeIntermediateTensorInfo(s),n.disposeIntermediateTensorInfo(i),o}let a=n.shouldExecuteOnCPU(e);if("string"===r&&(a=!0),a){let a=e.map(e=>{let r=[-1,va.sizeFromShape(e.shape.slice(t))];return Zz({inputs:{x:e},backend:n,attrs:{shape:r}})}),s=a.map(e=>({vals:n.readSync(e.dataId),shape:e.shape})),i=Gf.computeOutShape(a.map(e=>e.shape),1),o=1===a[0].shape[0],l=cP(s,i,r,o),u=Gf.computeOutShape(e.map(e=>e.shape),t),h=n.makeTensorInfo(u,r,l);return a.forEach(e=>n.disposeIntermediateTensorInfo(e)),h}let s=e.filter(e=>va.sizeFromShape(e.shape)>0),i=Le().getBool("WEBGL_PACK_ARRAY_OPERATIONS")&&s[0].shape.length>1;if(1===s.length){let t=i?new hz(e[0].shape,mz):new gz(e[0].shape,mz);return n.runWebGLProgram(t,e,r)}let o=Le().getNumber("WEBGL_MAX_TEXTURES_IN_SHADER");if(s.length>o){let e=[];for(let r=0;r<s.length;r+=o){let a=s.slice(r,r+o);e.push(TW(a,t,n))}let r=TW(e,t,n);for(let t of e)n.disposeIntermediateTensorInfo(t);return r}if(i){let e=new IW(s.map(e=>e.shape),t);return n.runWebGLProgram(e,s,r)}let{tensors2D:l,outShape:u}=function(e,t,n){let r=Gf.computeOutShape(e.map(e=>e.shape),t);return{tensors2D:e.map(e=>Zz({inputs:{x:e},attrs:{shape:[-1,va.sizeFromShape(e.shape.slice(t))]},backend:n})),outShape:r}}(s,t,n),h=new kW(l.map(e=>e.shape)),d=n.runWebGLProgram(h,l,r);l.forEach(e=>n.disposeIntermediateTensorInfo(e));let p=Zz({inputs:{x:d},attrs:{shape:u},backend:n});return n.disposeIntermediateTensorInfo(d),p}function CW(e){let{inputs:t,backend:n,attrs:r}=e,{axis:a}=r,s=va.parseAxisParam(a,t[0].shape)[0],i=t.map(e=>e.shape);Gf.assertParamsConsistent(i,s);let o=Gf.computeOutShape(t.map(e=>e.shape),s);if(0===va.sizeFromShape(o))return n.makeTensorInfo(o,t[0].dtype,[]);let l=t.filter(e=>va.sizeFromShape(e.shape)>0);return 1===l.length?Ez({inputs:{x:l[0]},backend:n}):TW(l,s,n)}var EW={kernelName:yt,backendName:"webgl",kernelFunc:CW},AW=class{constructor(e,t=!1,n=null,r=!1,a=!1){this.variableNames=["x","W"],this.outputShape=e.outShape;let s=e.padInfo.top,i=e.padInfo.left,o=e.strideHeight,l=e.strideWidth,u=e.dilationHeight,h=e.dilationWidth,d=e.filterHeight,p=e.filterWidth,c=4*Math.floor(e.inChannels/4),f=e.inChannels%4,m="channelsLast"===e.dataFormat,g=m?1:2,y=m?2:3,b=m?3:1,x="",v="";n&&(x=r?`float activation(float a) {\n          float b = getPreluActivationWeightsAtOutCoords();\n          ${n}\n        }`:a?`float activation(float a) {\n          float b = getLeakyreluAlphaAtOutCoords();\n          ${n}\n        }`:`\n          float activation(float x) {\n            ${n}\n          }\n        `,v="result = activation(result);");let w=t?"result += getBiasAtOutCoords();":"";t&&this.variableNames.push("bias"),r&&this.variableNames.push("preluActivationWeights"),a&&this.variableNames.push("leakyreluAlpha"),this.userCode=`\n      ${x}\n\n      const ivec2 strides = ivec2(${o}, ${l});\n      const ivec2 pads = ivec2(${s}, ${i});\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int batch = coords[0];\n        int d2 = coords[${b}];\n\n        ivec2 xRCCorner =\n            ivec2(coords[${g}], coords[${y}]) * strides - pads;\n        int xRCorner = xRCCorner.x;\n        int xCCorner = xRCCorner.y;\n\n        // Convolve x(?, ?, d1) with w(:, :, d1, d2) to get y(yR, yC, d2).\n        // ? = to be determined. : = across all values in that axis.\n        float dotProd = 0.0;\n        for (int wR = 0; wR < ${d}; wR++) {\n          int xR = xRCorner + wR * ${u};\n\n          if (xR < 0 || xR >= ${e.inHeight}) {\n            continue;\n          }\n\n          for (int wC = 0; wC < ${p}; wC++) {\n            int xC = xCCorner + wC * ${h};\n\n            if (xC < 0 || xC >= ${e.inWidth}) {\n              continue;\n            }\n\n            for (int d1 = 0; d1 < ${c}; d1 += 4) {\n              vec4 wValues = vec4(\n                getW(wR, wC, d1, d2),\n                getW(wR, wC, d1 + 1, d2),\n                getW(wR, wC, d1 + 2, d2),\n                getW(wR, wC, d1 + 3, d2)\n              );\n\n              if (${m}) {\n                vec4 xValues = vec4(\n                  getX(batch, xR, xC, d1),\n                  getX(batch, xR, xC, d1 + 1),\n                  getX(batch, xR, xC, d1 + 2),\n                  getX(batch, xR, xC, d1 + 3)\n                );\n                dotProd += dot(xValues, wValues);\n              } else {\n                vec4 xValues = vec4(\n                  getX(batch, d1, xR, xC),\n                  getX(batch, d1 + 1, xR, xC),\n                  getX(batch, d1 + 2, xR, xC),\n                  getX(batch, d1 + 3, xR, xC)\n                );\n                dotProd += dot(xValues, wValues);\n              }\n            }\n\n            if (${1===f}) {\n\n              if (${m}) {\n                dotProd +=\n                    getX(batch, xR, xC, ${c}) *\n                    getW(wR, wC, ${c}, d2);\n              } else {\n                dotProd +=\n                    getX(batch, ${c}, xR, xC) *\n                    getW(wR, wC, ${c}, d2);\n              }\n\n            } else if (${2===f}) {\n              vec2 wValues = vec2(\n                getW(wR, wC, ${c}, d2),\n                getW(wR, wC, ${c} + 1, d2)\n              );\n\n              if (${m}) {\n                vec2 xValues = vec2(\n                  getX(batch, xR, xC, ${c}),\n                  getX(batch, xR, xC, ${c} + 1)\n                );\n                dotProd += dot(xValues, wValues);\n              } else {\n                vec2 xValues = vec2(\n                  getX(batch, ${c}, xR, xC),\n                  getX(batch, ${c} + 1, xR, xC)\n                );\n                dotProd += dot(xValues, wValues);\n              }\n\n            } else if (${3===f}) {\n              vec3 wValues = vec3(\n                getW(wR, wC, ${c}, d2),\n                getW(wR, wC, ${c} + 1, d2),\n                getW(wR, wC, ${c} + 2, d2)\n              );\n\n              if (${m}) {\n                vec3 xValues = vec3(\n                  getX(batch, xR, xC, ${c}),\n                  getX(batch, xR, xC, ${c} + 1),\n                  getX(batch, xR, xC, ${c} + 2)\n                );\n                dotProd += dot(xValues, wValues);\n              } else {\n                vec3 xValues = vec3(\n                  getX(batch, ${c}, xR, xC),\n                  getX(batch, ${c} + 1, xR, xC),\n                  getX(batch, ${c} + 2, xR, xC)\n                );\n                dotProd += dot(xValues, wValues);\n              }\n\n            }\n          }\n        }\n\n        float result = dotProd;\n        ${w}\n        ${v}\n        setOutput(result);\n      }\n    `}},$W=class{constructor(e){this.variableNames=["x","W"],this.outputShape=e.outShape;let t=e.padInfo.front,n=e.padInfo.top,r=e.padInfo.left,a=e.strideDepth,s=e.strideHeight,i=e.strideWidth,o=e.dilationDepth,l=e.dilationHeight,u=e.dilationWidth,h=e.filterDepth,d=e.filterHeight,p=e.filterWidth,c=4*Math.floor(e.inChannels/4),f=e.inChannels%4;this.userCode=`\n      const ivec3 strides = ivec3(${a}, ${s}, ${i});\n      const ivec3 pads = ivec3(${t}, ${n}, ${r});\n\n      void main() {\n        ivec5 coords = getOutputCoords();\n        int batch = coords.x;\n        int d2 = coords.u;\n\n        ivec3 xFRCCorner = ivec3(coords.y, coords.z, coords.w) * strides - pads;\n        int xFCorner = xFRCCorner.x;\n        int xRCorner = xFRCCorner.y;\n        int xCCorner = xFRCCorner.z;\n\n        // Convolve x(?, ?, ?, d1) with w(:, :, :, d1, d2) to get\n        // y(yF, yR, yC, d2). ? = to be determined. : = across all\n        // values in that axis.\n        float dotProd = 0.0;\n        for (int wF = 0; wF < ${h}; wF++) {\n          int xF = xFCorner + wF * ${o};\n\n          if (xF < 0 || xF >= ${e.inDepth}) {\n            continue;\n          }\n\n          for (int wR = 0; wR < ${d}; wR++) {\n            int xR = xRCorner + wR * ${l};\n\n            if (xR < 0 || xR >= ${e.inHeight}) {\n              continue;\n            }\n\n            for (int wC = 0; wC < ${p}; wC++) {\n              int xC = xCCorner + wC * ${u};\n\n              if (xC < 0 || xC >= ${e.inWidth}) {\n                continue;\n              }\n\n              for (int d1 = 0; d1 < ${c}; d1 += 4) {\n                vec4 xValues = vec4(\n                  getX(batch, xF, xR, xC, d1),\n                  getX(batch, xF, xR, xC, d1 + 1),\n                  getX(batch, xF, xR, xC, d1 + 2),\n                  getX(batch, xF, xR, xC, d1 + 3)\n                );\n                vec4 wValues = vec4(\n                  getW(wF, wR, wC, d1, d2),\n                  getW(wF, wR, wC, d1 + 1, d2),\n                  getW(wF, wR, wC, d1 + 2, d2),\n                  getW(wF, wR, wC, d1 + 3, d2)\n                );\n\n                dotProd += dot(xValues, wValues);\n              }\n\n              if (${1===f}) {\n                dotProd +=\n                  getX(batch, xF, xR, xC, ${c}) *\n                  getW(wF, wR, wC, ${c}, d2);\n              } else if (${2===f}) {\n                vec2 xValues = vec2(\n                  getX(batch, xF, xR, xC, ${c}),\n                  getX(batch, xF, xR, xC, ${c} + 1)\n                );\n                vec2 wValues = vec2(\n                  getW(wF, wR, wC, ${c}, d2),\n                  getW(wF, wR, wC, ${c} + 1, d2)\n                );\n                dotProd += dot(xValues, wValues);\n              } else if (${3===f}) {\n                vec3 xValues = vec3(\n                  getX(batch, xF, xR, xC, ${c}),\n                  getX(batch, xF, xR, xC, ${c} + 1),\n                  getX(batch, xF, xR, xC, ${c} + 2)\n                );\n                vec3 wValues = vec3(\n                  getW(wF, wR, wC, ${c}, d2),\n                  getW(wF, wR, wC, ${c} + 1, d2),\n                  getW(wF, wR, wC, ${c} + 2, d2)\n                );\n                dotProd += dot(xValues, wValues);\n              }\n            }\n          }\n        }\n        setOutput(dotProd);\n      }\n    `}},RW=class{constructor(e,t=!1,n=null,r=!1,a=!1){this.variableNames=["x","W"],this.packedInputs=!0,this.packedOutput=!0,this.customUniforms=[{name:"pads",type:"ivec2"},{name:"strides",type:"ivec2"},{name:"dilations",type:"ivec2"},{name:"inDims",type:"ivec2"}],this.outputShape=e.outShape,this.enableShapeUniforms=EL(this.outputShape.length);let s=e.padInfo.left,i=e.strideWidth,o=e.dilationWidth,l=e.filterHeight,u=e.filterWidth,h=u,d="\n       int xR; int xC; int xCOffset;\n       vec4 wTexel; vec4 previous; vec4 final;";for(let e=0;e<u;e++)d+=`\n           vec4 xTexelC${2*e};\n           int xTexelC${2*e}Ready;\n           vec4 xTexelC${2*e+1};\n           int xTexelC${2*e+1}Ready;\n           vec4 xC${e};`;d+=`\n     for (int r = 0; r < ${l}; r++) {\n      for (int d1 = 0; d1 < ${e.inChannels}; d1 += 2) {\n       `;for(let e=0;e<u;e++)d+=`\n           xTexelC${2*e} = vec4(0.0);\n           xTexelC${2*e}Ready = 0;\n           xTexelC${2*e+1} = vec4(0.0);\n           xTexelC${2*e+1}Ready = 0;\n           xC${e} = vec4(0.0);`;d+="\n         xR = xRCorner + r * dilations[0];\n         if (xR >=0 && xR < inDims[0]) {\n       ";for(let t=0;t<(h+1)/2;t++){let n=2*t;if(d+=`\n           xC = xCCorner + ${n*o};\n           `,1===i){if(n<u&&(s%2==1?(d+=`\n                 xCOffset = xC + 1;\n                 if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${n}Ready == 0) {\n                   xTexelC${n} = getX(batch, xR, xCOffset, d1);\n\n                   // Need to manually clear unused channels in case\n                   // we're reading from recycled texture.\n                   if (xCOffset + 1 >= inDims[1]) {\n                     xTexelC${n}.zw = vec2(0.0);\n                   }\n                   xTexelC${n}Ready = 1;\n                 }\n               `,d+=1===o&&n>0?`\n                 xC${n} = vec4(xTexelC${n-2}.zw, xTexelC${n}.xy);\n                 `:`\n                   xCOffset = xC + 1 - 2;\n\n                   if (xCOffset >= 0 && xCOffset < inDims[1]) {\n                     previous = getX(batch, xR, xCOffset, d1);\n\n                     // Need to manually clear unused channels in case\n                     // we're reading from recycled texture.\n                     if (xCOffset + 1 >= inDims[1]) {\n                       previous.zw = vec2(0.0);\n                     }\n\n                     xC${n} = vec4(previous.zw, xTexelC${n}.xy);\n                   } else {\n                     xC${n} = vec4(0.0, 0.0, xTexelC${n}.xy);\n                   }\n                   `):d+=`\n                 if (xC >= 0 && xC < inDims[1] && xTexelC${n}Ready == 0) {\n                   xTexelC${n} = getX(batch, xR, xC, d1);\n                   if (xC + 1 >= inDims[1]) {\n                     xTexelC${n}.zw = vec2(0.0);\n                   }\n                   xTexelC${n}Ready = 1;\n                 }\n\n                 xC${n} = xTexelC${n};\n                 `,n+1<u)){let e=s%2==0?va.nearestLargerEven(o):o;o%2==0&&s%2==1||o%2!=0&&s%2!=1?(d+=`\n                   xCOffset = xC + imod(pads[1], 2) + ${e};\n\n                   if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${n+1}Ready == 0) {\n                     xTexelC${n+1} = getX(batch, xR, xCOffset, d1);\n\n                     // Need to manually clear unused channels in case\n                     // we're reading from recycled texture.\n                     if (xCOffset + 1 >= inDims[1]) {\n                       xTexelC${n+1}.zw = vec2(0.0);\n                     }\n                     xTexelC${n+1}Ready = 1;\n                   }\n                   `,d+=o>1?`\n                     xCOffset -= 2;\n                     if (xCOffset >= 0 && xCOffset < inDims[1]) {\n                      previous = getX(batch, xR, xCOffset, d1);\n                      xC${n+1} = vec4(previous.zw, xTexelC${n+1}.xy);\n                     } else {\n                      xC${n+1} = vec4(0.0, 0.0, xTexelC${n+1}.xy);\n                     }\n                     `:`\n                     xC${n+1} = vec4(xTexelC${n}.zw, xTexelC${n+1}.xy);\n                     `):d+=1===e?`\n                     xC${n+1} = xTexelC${n};\n                     `:`\n                     xCOffset = xC + ${e};\n\n                     if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${n+1}Ready == 0) {\n                       xTexelC${n+1} = getX(batch, xR, xCOffset, d1);\n                       if (xCOffset + 1 >= inDims[1]) {\n                         xTexelC${n+1}.zw = vec2(0.0);\n                       }\n                       xTexelC${n+1}Ready = 1;\n                     }\n\n                     xC${n+1} = xTexelC${n+1};\n                     `}}else n<u&&(s%2==1?(d+=`\n                 xCOffset = xC + 1 - strides[1];\n                 if(xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${n}Ready == 0) {\n                   xTexelC${n} = getX(batch, xR, xCOffset, d1);\n                   // Need to manually clear unused channels in case\n                   // we're reading from recycled texture.\n                   if (xCOffset + 1 >= inDims[1]) {\n                     xTexelC${n}.zw = vec2(0.0);\n                   }\n                   xTexelC${n}Ready = 1;\n                 }\n\n                 if(xC + 1 >= 0 && xC + 1 < inDims[1] && xTexelC${n+1}Ready == 0) {\n                   xTexelC${n+1} = getX(batch, xR, xC + 1, d1);\n                   // Need to manually clear unused channels in case\n                   // we're reading from recycled texture.\n                   if (xC + 2 >= inDims[1]) {\n                     xTexelC${n+1}.zw = vec2(0.0);\n                   }\n                   xTexelC${n+1}Ready = 1;\n                 }\n\n                 xC${n} = vec4(xTexelC${n}.zw, xTexelC${n+1}.zw);\n               `,n+1<u&&(d+=`\n                   final = vec4(0.0);\n                   xCOffset = xC + 1 + strides[1];\n                   if(xCOffset >= 0 && xCOffset < inDims[1]) {\n                     final = getX(batch, xR, xCOffset, d1);\n                   }\n                   xC${n+1} = vec4(xTexelC${n+1}.xy, final.xy);\n                 `)):(d+=`\n                 if(xC >= 0 && xC < inDims[1] && xTexelC${n}Ready == 0) {\n                   xTexelC${n} = getX(batch, xR, xC, d1);\n                   if (xC + 1 >= inDims[1]) {\n                     xTexelC${n}.zw = vec2(0.0);\n                   }\n                   xTexelC${n}Ready = 1;\n                 }\n\n                 xCOffset = xC + strides[1];\n                 if(xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${n+1}Ready == 0) {\n                   xTexelC${n+1} = getX(batch, xR, xCOffset, d1);\n                   if (xCOffset + 1 >= inDims[1]) {\n                     xTexelC${n+1}.zw = vec2(0.);\n                   }\n                   xTexelC${n+1}Ready = 1;\n                 }\n\n                 xC${n} = vec4(\n                   xTexelC${n}.xy, xTexelC${n+1}.xy);\n               `,n+1<u&&(d+=`\n                   xC${n+1} = vec4(xTexelC${n}.zw, xTexelC${n+1}.zw);\n                 `)));n<u&&(d+=`\n             wTexel = getW(r, ${n}, d1, d2);\n             dotProd += xC${n}.xxzz * vec4(wTexel.xy, wTexel.xy);\n             if(d1 + 1 < ${e.inChannels}) {\n               dotProd += xC${n}.yyww * vec4(wTexel.zw, wTexel.zw);\n             }\n           `,n+1<u&&(d+=`\n               wTexel = getW(r, ${n+1}, d1, d2);\n               dotProd += xC${n+1}.xxzz * vec4(wTexel.xy, wTexel.xy);\n               if(d1 + 1 < ${e.inChannels}) {\n                 dotProd += xC${n+1}.yyww * vec4(wTexel.zw, wTexel.zw);\n               }\n             `))}d+="\n     }\n   ",d+="\n     }\n   ",d+="\n     }\n   ";let p="",c="";n&&(p=r?`vec4 activation(vec4 a) {\n           vec4 b = getPreluActivationWeightsAtOutCoords();\n           ${n}\n         }`:a?`vec4 activation(vec4 a) {\n           vec4 b = getLeakyreluAlphaAtOutCoords();\n           ${n}\n         }`:`vec4 activation(vec4 x) {\n           ${n}\n         }`,c="result = activation(result);");let f=t?"result += getBiasAtOutCoords();":"";t&&this.variableNames.push("bias"),r&&this.variableNames.push("preluActivationWeights"),a&&this.variableNames.push("leakyreluAlpha"),this.userCode=`\n       ${p}\n\n       void main() {\n         ivec4 coords = getOutputCoords();\n         int batch = coords.x;\n         ivec2 xRCCorner = coords.yz * strides - pads;\n         int d2 = coords.w;\n         int xRCorner = xRCCorner.x;\n         int xCCorner = xRCCorner.y;\n\n         //intialize dotProd with a small epsilon seems to reduce GPU accuracy loss.\n         vec4 dotProd = vec4(0.000000000000001);\n\n         ${d}\n\n         vec4 result = dotProd - vec4(0.000000000000001);\n         ${f}\n         ${c}\n         setOutput(result);\n       }\n     `}},FW=class{constructor(e,t){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0,this.customUniforms=[{name:"inputShape",type:"ivec4"},{name:"pad",type:"ivec2"},{name:"stride",type:"ivec2"},{name:"dilation",type:"ivec2"},{name:"inChannels",type:"int"},{name:"itemsPerBlockRow",type:"int"},{name:"outWidth",type:"int"}],this.outputShape=e,this.enableShapeUniforms=EL(this.outputShape.length);let{dataFormat:n}=t,r=iL(),a="channelsLast"===n,s=a?1:2,i=a?2:3,o=this.enableShapeUniforms?"if(blockIndex < outShape[2] && pos < outShape[1]) {":`if(blockIndex < ${e[2]} && pos < ${e[1]}) {`,l="";for(let e=0;e<=1;e++)for(let t=0;t<=1;t++)l+=`\n          blockIndex = rc.z + ${t};\n          pos = rc.y + ${e};\n\n          ${o}\n            offsetY = int(blockIndex / outWidth) * stride[0] - pad[0];\n            d0 = offsetY + dilation[0] * (pos / itemsPerBlockRow);\n\n            if(d0 < inputShape[${s}] && d0 >= 0) {\n              // Use custom imod instead mod. On Intel GPU, mod may generate\n              // unexpected value.\n              // https://github.com/tensorflow/tfjs/issues/5447\n              offsetX = imod(blockIndex, outWidth) * stride[1] - pad[1];\n              d1 = offsetX + dilation[1] * (imod(pos, itemsPerBlockRow) /\n                  inChannels);\n\n              if(d1 < inputShape[${i}] && d1 >= 0) {\n\n                ch = imod(pos, inChannels);\n\n                if (${a}) {\n                  innerDims = vec2(d1, ch);\n                  result[${2*e+t}] = getChannel(\n                    getA(rc.x, d0, int(innerDims.x),\n                    int(innerDims.y)), innerDims);\n                } else {\n                  innerDims = vec2(d0, d1);\n                  result[${2*e+t}] = getChannel(\n                    getA(rc.x, ch, int(innerDims.x),\n                    int(innerDims.y)), innerDims);\n                }\n              }\n            }\n          }\n        `;this.userCode=`\n      void main() {\n        ivec3 rc = getOutputCoords();\n\n        vec4 result = vec4(0);\n\n        int blockIndex, pos, offsetY, d0, offsetX, d1, ch;\n        vec2 innerDims;\n\n        ${l}\n\n        ${r.output} = result;\n      }\n    `}};function DW(e,t){let n=e.length;return n>=3?t?[...e.slice(0,-3),e[n-3]*e[n-2],e[n-1]]:[...e.slice(0,-3),e[n-3],e[n-2]*e[n-1]]:!t&&1===n&&e[0]>1?[e[0],1]:null}function MW({x:e,filter:t,convInfo:n,backend:r,bias:a=null,preluActivationWeights:s=null,leakyreluAlpha:i=0,activation:o=null}){let l,u=e.shape,h=r.texData.get(e.dataId),d=n.inChannels,p=u[0]*u[1]*u[2],c=n.outChannels,f="channelsLast"===n.dataFormat,m=!1,g=[];if(null!=s){let e=DW(s.shape,f);null!=e&&(s=Zz({inputs:{x:s},backend:r,attrs:{shape:e}}),g.push(s))}if(null!=a){let e=DW(a.shape,f);null!=e&&(a=Zz({inputs:{x:a},backend:r,attrs:{shape:e}}),g.push(a))}if((1!==p&&1!==c||!(d>1e3))&&h.isPacked&&f&&null!=h.texture&&u[2]%2!=0&&va.arraysEqual(h.shape.slice(-3),u.slice(-3))){let d=u[0]*u[1]*(u[2]+1),p={dataId:e.dataId,shape:[1,d,n.inChannels],dtype:e.dtype},c=h.shape;h.shape=h.shape.slice(),h.shape[h.shape.length-2]++,va.assert(jO(h.shape,p.shape),()=>`packed reshape ${h.shape} to ${p.shape} isn't free`);let f=Zz({inputs:{x:t},backend:r,attrs:{shape:[1,n.inChannels,n.outChannels]}});g.push(f);let y=lB({a:p,b:f,backend:r,transposeA:false,transposeB:m,bias:a,activation:o,preluActivationWeights:s,leakyreluAlpha:i}),b=r.texData.get(y.dataId);va.assert(b.isPacked,()=>"batchMatMul result is expected to be packed"),h.shape=c,b.shape=n.outShape,l=Ez({inputs:{x:y},backend:r}),l.shape=n.outShape,g.push(y)}else{let u=n.outHeight*n.outWidth,h=Zz({inputs:{x:e},backend:r,attrs:{shape:f?[n.batchSize,u,n.inChannels]:[n.batchSize,n.inChannels,u]}}),d=Zz({inputs:{x:t},backend:r,attrs:{shape:[1,n.inChannels,n.outChannels]}}),p=lB({a:f?h:d,b:f?d:h,transposeA:!f,transposeB:m,backend:r,bias:a,activation:o,preluActivationWeights:s,leakyreluAlpha:i});l=Zz({inputs:{x:p},backend:r,attrs:{shape:n.outShape}}),g.push(h),g.push(d),g.push(p)}for(let e of g)r.disposeIntermediateTensorInfo(e);return l}function OW({x:e,filter:t,convInfo:n,backend:r,bias:a=null,preluActivationWeights:s=null,leakyreluAlpha:i=0,activation:o=null}){let{filterWidth:l,filterHeight:u,inChannels:h,outWidth:d,outHeight:p,dataFormat:c}=n,f="channelsLast"===c,m=l*u*h,g=p*d,y=[n.batchSize,m,g],b=[];if(null!=s){let e=DW(s.shape,f);null!=e&&(s=Zz({inputs:{x:s},backend:r,attrs:{shape:e}}),b.push(s))}if(null!=a){let e=DW(a.shape,f);null!=e&&(a=Zz({inputs:{x:a},backend:r,attrs:{shape:e}}),b.push(a))}let x=Zz({inputs:{x:t},backend:r,attrs:{shape:[1,m,va.sizeFromShape(t.shape)/m]}});b.push(x);let v=new FW(y,n),w=[e.shape,[n.padInfo.top,n.padInfo.left],[n.strideHeight,n.strideWidth],[n.dilationHeight,n.dilationWidth],[n.inChannels],[n.filterWidth*n.inChannels],[n.outWidth]],k=r.runWebGLProgram(v,[e],"float32",w),I=Zz({inputs:{x:k},backend:r,attrs:{shape:y}});b.push(k),b.push(I);let S=null!=a,_=null!=s,N="leakyrelu"===o,T=o?Vz(o,!0):null,C=new Uz(f?I.shape:x.shape,f?x.shape:I.shape,f?[n.batchSize,g,n.outChannels]:[n.batchSize,n.outChannels,g],!0,!1,S,T,_,N),E=f?[I,x]:[x,I];if(a&&E.push(a),_&&E.push(s),N){let e=r.makeTensorInfo([],"float32",va.createScalarValue(i,"float32"));E.push(e),b.push(e)}let A=r.runWebGLProgram(C,E,"float32"),$=Zz({inputs:{x:A},backend:r,attrs:{shape:n.outShape}});b.push(A);for(let e of b)r.disposeIntermediateTensorInfo(e);return $}var LW={kernelName:bt,backendName:"webgl",kernelFunc:function(e){let t,{inputs:n,backend:r,attrs:a}=e,{x:s,filter:i}=n,{strides:o,pad:l,dataFormat:u,dilations:h,dimRoundingMode:d}=a,p=Gf.convertConv2DDataFormat(u),c=Gf.computeConv2DInfo(s.shape,i.shape,o,h,l,d,!1,p);if(1!==c.filterHeight||1!==c.filterWidth||1!==c.dilationHeight||1!==c.dilationWidth||1!==c.strideHeight||1!==c.strideWidth||"SAME"!==c.padInfo.type&&"VALID"!==c.padInfo.type)if(c.strideWidth<=2&&"channelsLast"===p&&Le().getBool("WEBGL_EXP_CONV")){let e=new RW(c),n=[[c.padInfo.top,c.padInfo.left],[c.strideHeight,c.strideWidth],[c.dilationHeight,c.dilationWidth],[c.inHeight,c.inWidth]];t=r.runWebGLProgram(e,[s,i],"float32",n)}else if(Le().getBool("WEBGL_CONV_IM2COL"))t=OW({x:s,filter:i,convInfo:c,backend:r});else{let e=new AW(c);t=r.runWebGLProgram(e,[s,i],"float32")}else t=MW({x:s,filter:i,convInfo:c,backend:r});let f=Zz({inputs:{x:t},backend:r,attrs:{shape:c.outShape}});return r.disposeIntermediateTensorInfo(t),f}},PW=class{constructor(e){this.variableNames=["x","dy"],this.outputShape=e.filterShape;let t=e.strideHeight,n=e.strideWidth,r=e.padInfo.top,a=e.padInfo.left,s="channelsLast"===e.dataFormat;this.userCode=`\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int wR = coords.x;\n        int wC = coords.y;\n        int d1 = coords.z;\n        int d2 = coords.w;\n\n        // Convolve x(?, ?, d1) with dy(:, :, d2) to get dw(wR, wC, d1, d2).\n        // ? = to be determined. : = across all values in that axis.\n        float dotProd = 0.0;\n\n        for (int b = 0; b < ${e.batchSize}; b++) {\n          for (int yR = 0; yR < ${e.outHeight}; yR++) {\n            int xR = wR + yR * ${t} - ${r};\n\n            if (xR < 0 || xR >= ${e.inHeight}) {\n              continue;\n            }\n\n            for (int yC = 0; yC < ${e.outWidth}; yC++) {\n              int xC = wC + yC * ${n} - ${a};\n\n              if (xC < 0 || xC >= ${e.inWidth}) {\n                continue;\n              }\n\n              ${s?"float dyValue = getDy(b, yR, yC, d2);\n              float xValue = getX(b, xR, xC, d1);\n              dotProd += (xValue * dyValue);":"float dyValue = getDy(b, d2, yR, yC);\n              float xValue = getX(b, d1, xR, xC);\n              dotProd += (xValue * dyValue);"}\n            }\n          }\n        }\n        setOutput(dotProd);\n      }\n    `}},zW=class{constructor(e){this.variableNames=["dy","W"],this.outputShape=e.inShape;let t=e.filterHeight,n=e.filterWidth,r=e.strideHeight,a=e.strideWidth,s="channelsLast"===e.dataFormat,i=t-1-e.padInfo.top,o=n-1-e.padInfo.left,l=s?1:2,u=s?2:3,h=s?3:1;this.userCode=`\n      const ivec2 pads = ivec2(${i}, ${o});\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int batch = coords[0];\n        int d1 = coords[${h}];\n\n        ivec2 dyCorner = ivec2(coords[${l}], coords[${u}]) - pads;\n        int dyRCorner = dyCorner.x;\n        int dyCCorner = dyCorner.y;\n\n        // Convolve dy(?, ?, d2) with w(:, :, d1, d2) to compute dx(xR, xC, d1).\n        // ? = to be determined. : = across all values in that axis.\n        float dotProd = 0.0;\n        for (int wR = 0; wR < ${t}; wR++) {\n          float dyR = float(dyRCorner + wR) / ${r}.0;\n\n          if (dyR < 0.0 || dyR >= ${e.outHeight}.0 || fract(dyR) > 0.0) {\n            continue;\n          }\n          int idyR = int(dyR);\n\n          int wRPerm = ${t} - 1 - wR;\n\n          for (int wC = 0; wC < ${n}; wC++) {\n            float dyC = float(dyCCorner + wC) / ${a}.0;\n\n            if (dyC < 0.0 || dyC >= ${e.outWidth}.0 ||\n                fract(dyC) > 0.0) {\n              continue;\n            }\n            int idyC = int(dyC);\n\n            int wCPerm = ${n} - 1 - wC;\n\n            for (int d2 = 0; d2 < ${e.outChannels}; d2++) {\n\n              if (${s}) {\n                float xValue = getDy(batch, idyR, idyC, d2);\n                float wValue = getW(wRPerm, wCPerm, d1, d2);\n                dotProd += xValue * wValue;\n              } else {\n                float xValue = getDy(batch, d2, idyR, idyC);\n                float wValue = getW(wRPerm, wCPerm, d1, d2);\n                dotProd += xValue * wValue;\n              }\n\n            }\n          }\n        }\n        setOutput(dotProd);\n      }\n    `}},BW=class{constructor(e){this.variableNames=["x","dy"],this.outputShape=e.filterShape;let t=e.strideDepth,n=e.strideHeight,r=e.strideWidth,a=e.padInfo.front,s=e.padInfo.top,i=e.padInfo.left;this.userCode=`\n      void main() {\n        ivec5 coords = getOutputCoords();\n        int wF = coords.x;\n        int wR = coords.y;\n        int wC = coords.z;\n        int d1 = coords.w;\n        int d2 = coords.u;\n\n        float dotProd = 0.0;\n\n        for (int b = 0; b < ${e.batchSize}; b++) {\n          for (int yF = 0; yF < ${e.outDepth}; yF++) {\n            int xF = wF + yF * ${t} - ${a};\n\n            if (xF < 0 || xF >= ${e.inDepth}) {\n              continue;\n            }\n\n            for (int yR = 0; yR < ${e.outHeight}; yR++) {\n              int xR = wR + yR * ${n} - ${s};\n\n              if (xR < 0 || xR >= ${e.inHeight}) {\n                continue;\n              }\n\n              for (int yC = 0; yC < ${e.outWidth}; yC++) {\n                int xC = wC + yC * ${r} - ${i};\n\n                if (xC < 0 || xC >= ${e.inWidth}) {\n                  continue;\n                }\n\n                float dyValue = getDy(b, yF, yR, yC, d2);\n                float xValue = getX(b, xF, xR, xC, d1);\n                dotProd += (xValue * dyValue);\n              }\n            }\n          }\n        }\n        setOutput(dotProd);\n      }\n    `}},WW=class{constructor(e){this.variableNames=["dy","W"],this.outputShape=e.inShape;let t=e.filterDepth,n=e.filterHeight,r=e.filterWidth,a=e.strideDepth,s=e.strideHeight,i=e.strideWidth,o=t-1-e.padInfo.front,l=n-1-e.padInfo.top,u=r-1-e.padInfo.left;this.userCode=`\n      const ivec3 pads = ivec3(${o}, ${l}, ${u});\n\n      void main() {\n        ivec5 coords = getOutputCoords();\n        int batch = coords.x;\n        int d1 = coords.u;\n\n\n        ivec3 dyCorner = ivec3(coords.y, coords.z, coords.w) - pads;\n        int dyFCorner = dyCorner.x;\n        int dyRCorner = dyCorner.y;\n        int dyCCorner = dyCorner.z;\n\n        float dotProd = 0.0;\n        for (int wF = 0; wF < ${t}; wF++) {\n          float dyF = float(dyFCorner + wF) / ${a}.0;\n\n          if (dyF < 0.0 || dyF >= ${e.outDepth}.0 || fract(dyF) > 0.0) {\n            continue;\n          }\n          int idyF = int(dyF);\n\n          int wFPerm = ${t} - 1 - wF;\n\n          for (int wR = 0; wR < ${n}; wR++) {\n            float dyR = float(dyRCorner + wR) / ${s}.0;\n\n            if (dyR < 0.0 || dyR >= ${e.outHeight}.0 ||\n              fract(dyR) > 0.0) {\n              continue;\n            }\n            int idyR = int(dyR);\n\n            int wRPerm = ${n} - 1 - wR;\n\n            for (int wC = 0; wC < ${r}; wC++) {\n              float dyC = float(dyCCorner + wC) / ${i}.0;\n\n              if (dyC < 0.0 || dyC >= ${e.outWidth}.0 ||\n                  fract(dyC) > 0.0) {\n                continue;\n              }\n              int idyC = int(dyC);\n\n              int wCPerm = ${r} - 1 - wC;\n\n              for (int d2 = 0; d2 < ${e.outChannels}; d2++) {\n                float xValue = getDy(batch, idyF, idyR, idyC, d2);\n                float wValue = getW(wFPerm, wRPerm, wCPerm, d1, d2);\n                dotProd += xValue * wValue;\n              }\n            }\n          }\n        }\n        setOutput(dotProd);\n      }\n    `}};var VW={kernelName:xt,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a,dy:s}=t,{strides:i,pad:o,dataFormat:l,dimRoundingMode:u,filterShape:h}=r,d=Gf.convertConv2DDataFormat(l),p=Gf.computeConv2DInfo(a.shape,h,i,1,o,u,!1,d),c=new PW(p);return n.runWebGLProgram(c,[a,s],"float32")}},UW=class{constructor(e){this.variableNames=["dy","W"],this.packedInputs=!0,this.packedOutput=!0,this.customUniforms=[{name:"strides",type:"vec2"}],this.outputShape=e.inShape,this.enableShapeUniforms=EL(this.outputShape.length);let t=e.filterHeight,n=e.filterWidth,r=t-1-e.padInfo.top,a=n-1-e.padInfo.left;this.userCode=`\n      const ivec2 pads = ivec2(${r}, ${a});\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int batch = coords[0];\n        int d1 = coords[3];\n\n        ivec2 dyCorner = ivec2(coords[1], coords[2]) - pads;\n        int dyRCorner = dyCorner.x;\n        int dyCCorner = dyCorner.y;\n\n        vec4 result = vec4(0.);\n        for (int wR = 0; wR < ${t}; wR++) {\n          float dyR = float(dyRCorner + wR) / strides[0];\n          if (dyR < 0.0 || dyR >= ${e.outHeight}.0 || fract(dyR) > 0.0) {\n            continue;\n          }\n          int idyR = int(dyR);\n          int wRPerm = ${t} - 1 - wR;\n\n          for (int wC = 0; wC < ${n}; wC++) {\n            int wCPerm = ${n} - 1 - wC;\n\n            float dyC = float(dyCCorner + wC) / strides[1];\n            bool idyCVal = (dyC >= 0.0) && (dyC < ${e.outWidth}.0)\n              && (fract(dyC) == 0.0);\n            int idyC = int(dyC);\n\n            float dyC2 = float(dyCCorner + wC + 1) / strides[1];\n            bool idyCVal2 = (dyC2 >= 0.0) && (dyC2 < ${e.outWidth}.0)\n              && (fract(dyC2) == 0.0);\n            int idyC2 = int(dyC2);\n\n            if (idyCVal && idyCVal2) {\n              for (int d2 = 0; d2 < ${e.outChannels}; d2 += 2) {\n                vec4 wValue = getW(wRPerm, wCPerm, d1, d2);\n                vec4 dySample = getDy(batch, idyR, idyC, d2);\n                vec4 dySample2 = (idyC / 2 == idyC2 / 2) ?\n                  dySample : getDy(batch, idyR, idyC2, d2);\n\n                vec2 dyValue = mod(float(idyC), 2.) == 0. ?\n                  dySample.xy : dySample.zw;\n                result.xy += vec2(dot(dyValue, wValue.xy),\n                  dot(dyValue, wValue.zw));\n\n                dyValue = mod(float(idyC2), 2.) == 0. ?\n                  dySample2.xy : dySample2.zw;\n                result.zw += vec2(dot(dyValue, wValue.xy),\n                  dot(dyValue, wValue.zw));\n              }\n            } else if (idyCVal) {\n              for (int d2 = 0; d2 < ${e.outChannels}; d2 += 2) {\n                vec4 wValue = getW(wRPerm, wCPerm, d1, d2);\n                vec4 dySample = getDy(batch, idyR, idyC, d2);\n                vec2 dyValue = mod(float(idyC), 2.) == 0. ?\n                  dySample.xy : dySample.zw;\n                result.xy += vec2(dot(dyValue, wValue.xy),\n                  dot(dyValue, wValue.zw));\n              }\n            } else if (idyCVal2) {\n              for (int d2 = 0; d2 < ${e.outChannels}; d2 += 2) {\n                vec4 wValue = getW(wRPerm, wCPerm, d1, d2);\n                vec4 dySample = getDy(batch, idyR, idyC2, d2);\n                vec2 dyValue = mod(float(idyC2), 2.) == 0. ?\n                  dySample.xy : dySample.zw;\n                result.zw += vec2(dot(dyValue, wValue.xy),\n                  dot(dyValue, wValue.zw));\n              }\n            }\n          }\n        }\n        setOutput(result);\n      }\n    `}};var GW={kernelName:vt,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{dy:a,filter:s}=t,{inputShape:i,strides:o,pad:l,dataFormat:u,dimRoundingMode:h}=r,d=Gf.convertConv2DDataFormat(u),p=Gf.computeConv2DInfo(i,s.shape,o,1,l,h,!1,d);if(Le().getBool("WEBGL_PACK_CONV2DTRANSPOSE")&&"channelsLast"===d){let e=[[p.strideHeight,p.strideWidth]],t=new UW(p);return n.runWebGLProgram(t,[a,s],"float32",e)}{let e=new zW(p);return n.runWebGLProgram(e,[a,s],"float32")}}};var HW={kernelName:wt,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a,filter:s}=t,{strides:i,pad:o,dilations:l}=r,u=Gf.computeConv3DInfo(a.shape,s.shape,i,l,o),h=new $W(u);return n.runWebGLProgram(h,[a,s],"float32")}};var jW={kernelName:kt,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a,dy:s}=t,{strides:i,pad:o,filterShape:l}=r,u=Gf.computeConv3DInfo(a.shape,l,i,1,o),h=new BW(u);return n.runWebGLProgram(h,[a,s],"float32")}};var qW,KW={kernelName:It,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{dy:a,filter:s}=t,{pad:i,strides:o,inputShape:l}=r,u=Gf.computeConv3DInfo(l,s.shape,o,1,i),h=new WW(u);return n.runWebGLProgram(h,[a,s],"float32")}},XW=Bz({opSnippet:zz+"\n  return cos(x);\n",packedOpSnippet:`\n  vec4 result = cos(x);\n  bvec4 isNaN = isnan(x);\n  ${Tz}\n  return result;\n`}),ZW={kernelName:St,backendName:"webgl",kernelFunc:XW},YW=Bz({opSnippet:"\n  float e2x = exp(-x);\n  return (e2x + 1.0 / e2x) / 2.0;\n"}),JW={kernelName:_t,backendName:"webgl",kernelFunc:YW},QW=class{constructor(e,t,n,r,a){this.variableNames=["Image","Boxes","BoxInd"],this.outputShape=[];let[s,i,o,l]=e,[u]=t,[h,d]=n;this.outputShape=[u,h,d,l];let p="bilinear"===r?1:0,[c,f]=[i-1+".0",o-1+".0"],[m,g,y]=h>1?[""+(i-1)/(h-1),"(y2-y1) * height_ratio",`y1*${c} + float(y)*(height_scale)`]:["0.0","0.0",`0.5 * (y1+y2) * ${c}`],[b,x,v]=d>1?[""+(o-1)/(d-1),"(x2-x1) * width_ratio",`x1*${f} + float(x)*(width_scale)`]:["0.0","0.0",`0.5 * (x1+x2) * ${f}`];this.userCode=`\n      const float height_ratio = float(${m});\n      const float width_ratio = float(${b});\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int b = coords[0];\n        int y = coords[1];\n        int x = coords[2];\n        int d = coords[3];\n\n        // get box vals\n        float y1 = getBoxes(b,0);\n        float x1 = getBoxes(b,1);\n        float y2 = getBoxes(b,2);\n        float x2 = getBoxes(b,3);\n\n        // get image in batch index\n        int bInd = round(getBoxInd(b));\n        if(bInd < 0 || bInd >= ${s}) {\n          return;\n        }\n\n        float height_scale = ${g};\n        float width_scale = ${x};\n\n        float in_y = ${y};\n        if( in_y < 0.0 || in_y > ${c} ) {\n          setOutput(float(${a}));\n          return;\n        }\n        float in_x = ${v};\n        if( in_x < 0.0 || in_x > ${f} ) {\n          setOutput(float(${a}));\n          return;\n        }\n\n        vec2 sourceFracIndexCR = vec2(in_x,in_y);\n        if(${p} == 1) {\n          // Compute the four integer indices.\n          ivec2 sourceFloorCR = ivec2(sourceFracIndexCR);\n          ivec2 sourceCeilCR = ivec2(ceil(sourceFracIndexCR));\n\n          float topLeft = getImage(b, sourceFloorCR.y, sourceFloorCR.x, d);\n          float bottomLeft = getImage(b, sourceCeilCR.y, sourceFloorCR.x, d);\n          float topRight = getImage(b, sourceFloorCR.y, sourceCeilCR.x, d);\n          float bottomRight = getImage(b, sourceCeilCR.y, sourceCeilCR.x, d);\n\n          vec2 fracCR = sourceFracIndexCR - vec2(sourceFloorCR);\n\n          float top = topLeft + (topRight - topLeft) * fracCR.x;\n          float bottom = bottomLeft + (bottomRight - bottomLeft) * fracCR.x;\n          float newValue = top + (bottom - top) * fracCR.y;\n          setOutput(newValue);\n        } else {\n          // Compute the coordinators of nearest neighbor point.\n          ivec2 sourceNearestCR = ivec2(floor(\n            sourceFracIndexCR + vec2(0.5,0.5)));\n          float newValue = getImage(b, sourceNearestCR.y, sourceNearestCR.x, d);\n          setOutput(newValue);\n        }\n      }\n    `}},eV={kernelName:Ct,backendName:"webgl",kernelFunc:e=>{let{inputs:t,backend:n,attrs:r}=e,{image:a,boxes:s,boxInd:i}=t,{cropSize:o,method:l,extrapolationValue:u}=r,h=new QW(a.shape,s.shape,o,l,u);return n.runWebGLProgram(h,[a,s,i],"float32")}};!function(e){e.Prod="*",e.Sum="+"}(qW||(qW={}));var tV=class{constructor(e,t,n,r){this.op=e,this.outputShape=t,this.variableNames=["x"],this.customUniforms=[{name:"index",type:"float"}];let a=this.outputShape.length,s=this.op===qW.Prod?"1.0":"0.0",i=n?s:`getX(${nV(a,"coords",this.op)})`,o=this.outputShape[this.outputShape.length-1],l="",u="";n?(l=r?"end != "+(o-1):"end != 0",u=r?"end + 1":"end - 1"):(l=r?`end + pow2 < ${o}`:"end >= pow2",u=r?"end + pow2":"end - pow2"),this.userCode=`\n      void main() {\n        ${IL(a)} coords = getOutputCoords();\n        int end = ${rV(a,"coords",this.op)};\n        float val = ${i};\n        int pow2 = int(pow(2.0, index));\n        if (${l}) {\n          int idx = ${u};\n          ${rV(a,"coords",this.op)} = idx;\n          val ${this.op}= getX(${nV(a,"coords",this.op)});\n        }\n        setOutput(val);\n      }\n    `}};function nV(e,t,n){if(1===e)return`${t}`;if(2===e)return`${t}.x, ${t}.y`;if(3===e)return`${t}.x, ${t}.y, ${t}.z`;if(4===e)return`${t}.x, ${t}.y, ${t}.z, ${t}.w`;throw new Error(`Cumulative ${n} for rank ${e} is not yet supported`)}function rV(e,t,n){if(1===e)return`${t}`;if(2===e)return`${t}.y`;if(3===e)return`${t}.z`;if(4===e)return`${t}.w`;throw new Error(`Cumulative ${n} for rank ${e} is not yet supported`)}function aV(e,t,n,r,a,s){let i=t.shape.length,o=Gf.getAxesPermutation([r],i),l=t;null!=o&&(l=iB({inputs:{x:t},backend:n,attrs:{perm:o}}));let u=Gf.getInnerMostAxes(1,i)[0];if(u!==i-1)throw new Error(`WebGL cumprod shader expects an inner-most axis=${t.shape.length-1} but got axis=${r}`);let h=l.shape[u],d=Ez({inputs:{x:l},backend:n});for(let t=0;t<=Math.ceil(Math.log2(h))-1;t++){let r=new tV(e,l.shape,!1,s),a=[[t]],i=d;d=n.runWebGLProgram(r,[d],d.dtype,a),n.disposeIntermediateTensorInfo(i)}if(a){let t=new tV(e,l.shape,a,s),r=d;d=n.runWebGLProgram(t,[d],d.dtype),n.disposeIntermediateTensorInfo(r)}if(null!=o){let e=iB({inputs:{x:d},backend:n,attrs:{perm:Gf.getUndoAxesPermutation(o)}});return n.disposeIntermediateTensorInfo(d),n.disposeIntermediateTensorInfo(l),e}return d}var sV={kernelName:Nt,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{axis:s,exclusive:i,reverse:o}=r;return aV(qW.Prod,a,n,s,i,o)}};var iV={kernelName:Tt,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{axis:s,exclusive:i,reverse:o}=r;return aV(qW.Sum,a,n,s,i,o)}};var oV={kernelName:Et,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a,weights:s}=t,{size:i,binaryOutput:o}=r;if(1===a.shape.length){let e=n.readSync(a.dataId),t=n.readSync(s.dataId),r=lP(e,t,s.dtype,s.shape,i);return n.makeTensorInfo([i],s.dtype,r)}if(2===a.shape.length){let e=n.bufferSync(a),t=n.bufferSync(s),r=uP(e,t,i,o);return n.makeTensorInfo(r.shape,s.dtype,r.values)}throw new Error(`Error in denseBincount: input must be at most rank 2, but got rank${a.shape.length}.`)}},lV=class{constructor(e,t,n){this.variableNames=["x"],this.outputShape=[],this.outputShape=e,this.blockSize=t,this.dataFormat=n,this.userCode=`\n    void main() {\n      ivec4 coords = getOutputCoords();\n      int b = coords[0];\n      int h = ${this.getHeightCoordString()};\n      int w = ${this.getWidthCoordString()};\n      int d = ${this.getDepthCoordString()};\n\n      int in_h = h / ${t};\n      int offset_h = imod(h, ${t});\n      int in_w = w / ${t};\n      int offset_w = imod(w, ${t});\n      int offset_d = (offset_h * ${t} + offset_w) *\n        ${this.getOutputDepthSize()};\n      int in_d = d + offset_d;\n\n      float result = ${this.getInputSamplingString()};\n      setOutput(result);\n    }\n  `}getHeightCoordString(){return"NHWC"===this.dataFormat?"coords[1]":"coords[2]"}getWidthCoordString(){return"NHWC"===this.dataFormat?"coords[2]":"coords[3]"}getDepthCoordString(){return"NHWC"===this.dataFormat?"coords[3]":"coords[1]"}getOutputDepthSize(){return"NHWC"===this.dataFormat?this.outputShape[3]:this.outputShape[1]}getInputSamplingString(){return"NHWC"===this.dataFormat?"getX(b, in_h, in_w, in_d)":"getX(b, in_d, in_h, in_w)"}};var uV={kernelName:At,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{blockSize:s,dataFormat:i}=r,o=a.shape[0],l=("NHWC"===i?a.shape[1]:a.shape[2])*s,u=("NHWC"===i?a.shape[2]:a.shape[3])*s,h=("NHWC"===i?a.shape[3]:a.shape[1])/(s*s),d=new lV("NHWC"===i?[o,l,u,h]:[o,h,l,u],s,i);return n.runWebGLProgram(d,[a],a.dtype)}},hV=class{constructor(e,t=!1,n=null,r=!1,a=!1){this.variableNames=["x","W"],this.customUniforms=[{name:"pads",type:"ivec2"},{name:"strides",type:"ivec2"},{name:"dilations",type:"ivec2"},{name:"inDims",type:"ivec2"}],this.outputShape=e.outShape,this.enableShapeUniforms=EL(this.outputShape.length);let s=e.filterHeight,i=e.filterWidth,o=e.outChannels/e.inChannels,l="",u="";n&&(l=r?`float activation(float a) {\n          float b = getPreluActivationWeightsAtOutCoords();\n          ${n}\n        }`:a?`float activation(float a) {\n          float b = getLeakyreluAlphaAtOutCoords();\n          ${n}\n        }`:`\n          float activation(float x) {\n            ${n}\n          }\n        `,u="result = activation(result);");let h=t?"result += getBiasAtOutCoords();":"";t&&this.variableNames.push("bias"),r&&this.variableNames.push("preluActivationWeights"),a&&this.variableNames.push("leakyreluAlpha"),this.userCode=`\n      ${l}\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int batch = coords.x;\n        ivec2 xRCCorner = coords.yz * strides - pads;\n        int d2 = coords.w;\n        int d1 = d2 / ${o};\n        int q = d2 - d1 * ${o};\n\n        int xRCorner = xRCCorner.x;\n        int xCCorner = xRCCorner.y;\n\n        // Convolve x(?, ?, d1) with w(:, :, d1, q) to get y(yR, yC, d2).\n        // ? = to be determined. : = across all values in that axis.\n        float dotProd = 0.0;\n        // TO DO(dsmilkov): Flatten the two for loops and vec4 the operations.\n        for (int wR = 0; wR < ${s}; wR++) {\n          int xR = xRCorner + wR * dilations[0];\n\n          if (xR < 0 || xR >= inDims[0]) {\n            continue;\n          }\n\n          for (int wC = 0; wC < ${i}; wC++) {\n            int xC = xCCorner + wC * dilations[1];\n\n            if (xC < 0 || xC >= inDims[1]) {\n              continue;\n            }\n\n            float xVal = getX(batch, xR, xC, d1);\n            float wVal = getW(wR, wC, d1, q);\n            dotProd += xVal * wVal;\n          }\n        }\n\n        float result = dotProd;\n        ${h}\n        ${u}\n        setOutput(result);\n      }\n    `}},dV=class{constructor(e,t=!1,n=null,r=!1,a=!1){this.variableNames=["x","W"],this.packedInputs=!0,this.packedOutput=!0,this.customUniforms=[{name:"pads",type:"ivec2"},{name:"strides",type:"ivec2"},{name:"dilations",type:"ivec2"},{name:"inDims",type:"ivec2"}],this.outputShape=e.outShape,this.enableShapeUniforms=EL(this.outputShape.length);let s=e.outChannels/e.inChannels,i=e.padInfo.left,o=e.strideWidth,l=e.dilationWidth,u=e.filterHeight,h=e.filterWidth,d=h,p="\n      int xR; int xC; int xCOffset;\n      vec4 wTexel; vec4 previous; vec4 final;";for(let e=0;e<h;e++)p+=`\n          vec4 xTexelC${2*e};\n          int xTexelC${2*e}Ready;\n          vec4 xTexelC${2*e+1};\n          int xTexelC${2*e+1}Ready;\n          vec4 xC${e};`;p+=`\n    for (int r = 0; r < ${u}; r++) {\n      `;for(let e=0;e<h;e++)p+=`\n          xTexelC${2*e} = vec4(0.0);\n          xTexelC${2*e}Ready = 0;\n          xTexelC${2*e+1} = vec4(0.0);\n          xTexelC${2*e+1}Ready = 0;\n          xC${e} = vec4(0.0);`;p+="\n        xR = xRCorner + r * dilations[0];\n        if (xR >=0 && xR < inDims[0]) {\n      ";for(let e=0;e<(d+1)/2;e++){let t=2*e;if(p+=`\n          xC = xCCorner + ${t*l};\n          `,1===o){if(t<h&&(i%2==1?(p+=`\n                xCOffset = xC + 1;\n                if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${t}Ready == 0) {\n                  xTexelC${t} = getX(batch, xR, xCOffset, d1);\n\n                  // Need to manually clear unused channels in case\n                  // we're reading from recycled texture.\n                  if (xCOffset + 1 >= inDims[1]) {\n                    xTexelC${t}.zw = vec2(0.0);\n                  }\n                  xTexelC${t}Ready = 1;\n                }\n              `,p+=1===l&&t>0?`\n                xC${t} = vec4(xTexelC${t-2}.zw, xTexelC${t}.xy);\n                `:`\n                  xCOffset = xC + 1 - 2;\n\n                  if (xCOffset >= 0 && xCOffset < inDims[1]) {\n                    previous = getX(batch, xR, xCOffset, d1);\n\n                    // Need to manually clear unused channels in case\n                    // we're reading from recycled texture.\n                    if (xCOffset + 1 >= inDims[1]) {\n                      previous.zw = vec2(0.0);\n                    }\n\n                    xC${t} = vec4(previous.zw, xTexelC${t}.xy);\n                  } else {\n                    xC${t} = vec4(0.0, 0.0, xTexelC${t}.xy);\n                  }\n                  `):p+=`\n                if (xC >= 0 && xC < inDims[1] && xTexelC${t}Ready == 0) {\n                  xTexelC${t} = getX(batch, xR, xC, d1);\n                  if (xC + 1 >= inDims[1]) {\n                    xTexelC${t}.zw = vec2(0.0);\n                  }\n                  xTexelC${t}Ready = 1;\n                }\n\n                xC${t} = xTexelC${t};\n                `,t+1<h)){let e=i%2==0?va.nearestLargerEven(l):l;l%2==0&&i%2==1||l%2!=0&&i%2!=1?(p+=`\n                  xCOffset = xC + imod(pads[1], 2) + ${e};\n\n                  if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${t+1}Ready == 0) {\n                    xTexelC${t+1} = getX(batch, xR, xCOffset, d1);\n\n                    // Need to manually clear unused channels in case\n                    // we're reading from recycled texture.\n                    if (xCOffset + 1 >= inDims[1]) {\n                      xTexelC${t+1}.zw = vec2(0.0);\n                    }\n                    xTexelC${t+1}Ready = 1;\n                  }\n                  `,p+=l>1?`\n                    xCOffset -= 2;\n                    if (xCOffset >= 0 && xCOffset < inDims[1]) {\n                     previous = getX(batch, xR, xCOffset, d1);\n                     xC${t+1} = vec4(previous.zw, xTexelC${t+1}.xy);\n                    } else {\n                     xC${t+1} = vec4(0.0, 0.0, xTexelC${t+1}.xy);\n                    }\n                    `:`\n                    xC${t+1} = vec4(xTexelC${t}.zw, xTexelC${t+1}.xy);\n                    `):p+=1===e?`\n                    xC${t+1} = xTexelC${t};\n                    `:`\n                    xCOffset = xC + ${e};\n\n                    if (xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${t+1}Ready == 0) {\n                      xTexelC${t+1} = getX(batch, xR, xCOffset, d1);\n                      if (xCOffset + 1 >= inDims[1]) {\n                        xTexelC${t+1}.zw = vec2(0.0);\n                      }\n                      xTexelC${t+1}Ready = 1;\n                    }\n\n                    xC${t+1} = xTexelC${t+1};\n                    `}}else t<h&&(i%2==1?(p+=`\n                xCOffset = xC + 1 - strides[1];\n                if(xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${t}Ready == 0) {\n                  xTexelC${t} = getX(batch, xR, xCOffset, d1);\n                  // Need to manually clear unused channels in case\n                  // we're reading from recycled texture.\n                  if (xCOffset + 1 >= inDims[1]) {\n                    xTexelC${t}.zw = vec2(0.0);\n                  }\n                  xTexelC${t}Ready = 1;\n                }\n\n                if(xC + 1 >= 0 && xC + 1 < inDims[1] && xTexelC${t+1}Ready == 0) {\n                  xTexelC${t+1} = getX(batch, xR, xC + 1, d1);\n                  // Need to manually clear unused channels in case\n                  // we're reading from recycled texture.\n                  if (xC + 2 >= inDims[1]) {\n                    xTexelC${t+1}.zw = vec2(0.0);\n                  }\n                  xTexelC${t+1}Ready = 1;\n                }\n\n                xC${t} = vec4(xTexelC${t}.zw, xTexelC${t+1}.zw);\n              `,t+1<h&&(p+=`\n                  final = vec4(0.0);\n                  xCOffset = xC + 1 + strides[1];\n                  if(xCOffset >= 0 && xCOffset < inDims[1]) {\n                    final = getX(batch, xR, xCOffset, d1);\n                  }\n                  xC${t+1} = vec4(xTexelC${t+1}.xy, final.xy);\n                `)):(p+=`\n                if(xC >= 0 && xC < inDims[1] && xTexelC${t}Ready == 0) {\n                  xTexelC${t} = getX(batch, xR, xC, d1);\n                  if (xC + 1 >= inDims[1]) {\n                    xTexelC${t}.zw = vec2(0.0);\n                  }\n                  xTexelC${t}Ready = 1;\n                }\n\n                xCOffset = xC + strides[1];\n                if(xCOffset >= 0 && xCOffset < inDims[1] && xTexelC${t+1}Ready == 0) {\n                  xTexelC${t+1} = getX(batch, xR, xCOffset, d1);\n                  if (xCOffset + 1 >= inDims[1]) {\n                    xTexelC${t+1}.zw = vec2(0.);\n                  }\n                  xTexelC${t+1}Ready = 1;\n                }\n\n                xC${t} = vec4(\n                  xTexelC${t}.xy, xTexelC${t+1}.xy);\n              `,t+1<h&&(p+=`\n                  xC${t+1} = vec4(xTexelC${t}.zw, xTexelC${t+1}.zw);\n                `)));t<h&&(p+=`\n            wTexel = getW(r, ${t}, d1, q);\n            dotProd += xC${t} * vec4(wTexel.xz, wTexel.xz);\n          `,t+1<h&&(p+=`\n              wTexel = getW(r, ${t+1}, d1, q);\n              dotProd += xC${t+1} * vec4(wTexel.xz, wTexel.xz);\n            `))}p+="\n    }\n  ",p+="\n      }\n    ";let c="",f="";n&&(c=r?`vec4 activation(vec4 a) {\n          vec4 b = getPreluActivationWeightsAtOutCoords();\n          ${n}\n        }`:a?`vec4 activation(vec4 a) {\n          vec4 b = getLeakyreluAlphaAtOutCoords();\n          ${n}\n        }`:`vec4 activation(vec4 x) {\n          ${n}\n        }`,f="result = activation(result);");let m=t?"result += getBiasAtOutCoords();":"";t&&this.variableNames.push("bias"),r&&this.variableNames.push("preluActivationWeights"),a&&this.variableNames.push("leakyreluAlpha"),this.userCode=`\n      ${c}\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int batch = coords.x;\n        ivec2 xRCCorner = coords.yz * strides - pads;\n        int d2 = coords.w;\n        int d1 = d2 / ${s};\n        int q = d2 - d1 * ${s};\n        int xRCorner = xRCCorner.x;\n        int xCCorner = xRCCorner.y;\n\n        //intialize dotProd with a small epsilon seems to reduce GPU accuracy loss.\n        vec4 dotProd = vec4(0.000000000000001);\n\n        ${p}\n\n        vec4 result = dotProd - vec4(0.000000000000001);\n        ${m}\n        ${f}\n        setOutput(result);\n      }\n    `}};var pV={kernelName:$t,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a,filter:s}=t,{strides:i,pad:o,dilations:l,dimRoundingMode:u}=r,h=l;null==h&&(h=[1,1]),va.assert(Gf.eitherStridesOrDilationsAreOne(i,h),()=>`Error in depthwiseConv2d: Either strides or dilations must be 1. Got strides ${i} and dilations '${h}'`);let d,p=Gf.computeConv2DInfo(a.shape,s.shape,i,h,o,u,!0);d=Le().getBool("WEBGL_PACK_DEPTHWISECONV")&&p.strideWidth<=2&&p.outChannels/p.inChannels===1?new dV(p):new hV(p);let c=[[p.padInfo.top,p.padInfo.left],[p.strideHeight,p.strideWidth],[p.dilationHeight,p.dilationWidth],[p.inHeight,p.inWidth]];return n.runWebGLProgram(d,[a,s],"float32",c)}},cV=class{constructor(e){this.variableNames=["x","dy"],this.outputShape=e.filterShape;let t=e.strideHeight,n=e.strideWidth,r=e.padInfo.top,a=e.padInfo.left,s=e.outChannels/e.inChannels;this.userCode=`\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int wR = coords.x;\n        int wC = coords.y;\n        int d1 = coords.z;\n        int dm = coords.w;\n        int d2 = d1 * ${s} + dm;\n\n        float dotProd = 0.0;\n\n        // TO DO: Vec4 over the batch size\n        for (int b = 0; b < ${e.batchSize}; b++) {\n          for (int yR = 0; yR < ${e.outHeight}; yR++) {\n            int xR = wR + yR * ${t} - ${r};\n\n            if (xR < 0 || xR >= ${e.inHeight}) {\n              continue;\n            }\n\n            for (int yC = 0; yC < ${e.outWidth}; yC++) {\n              int xC = wC + yC * ${n} - ${a};\n\n              if (xC < 0 || xC >= ${e.inWidth}) {\n                continue;\n              }\n\n              float dyValue = getDy(b, yR, yC, d2);\n              float xValue = getX(b, xR, xC, d1);\n              dotProd += (xValue * dyValue);\n            }\n          }\n        }\n        setOutput(dotProd);\n      }\n    `}},fV=class{constructor(e){this.variableNames=["dy","W"],this.outputShape=e.inShape;let t=e.filterHeight,n=e.filterWidth,r=e.strideHeight,a=e.strideWidth,s=t-1-e.padInfo.top,i=n-1-e.padInfo.left,o=e.outChannels/e.inChannels;this.userCode=`\n      const ivec2 pads = ivec2(${s}, ${i});\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int batch = coords[0];\n        int d1 = coords[3];\n        ivec2 dyCorner = coords.yz - pads;\n        int dyRCorner = dyCorner.x;\n        int dyCCorner = dyCorner.y;\n\n        float dotProd = 0.0;\n\n        for (int wR = 0; wR < ${t}; wR++) {\n          float dyR = float(dyRCorner + wR) / ${r}.0;\n\n          if (dyR < 0.0 || dyR >= ${e.outHeight}.0 || fract(dyR) > 0.0) {\n            continue;\n          }\n          int idyR = int(dyR);\n\n          int wRPerm = ${t} - 1 - wR;\n\n          for (int wC = 0; wC < ${n}; wC++) {\n            float dyC = float(dyCCorner + wC) / ${a}.0;\n\n            if (dyC < 0.0 || dyC >= ${e.outWidth}.0 ||\n                fract(dyC) > 0.0) {\n              continue;\n            }\n            int idyC = int(dyC);\n\n            int wCPerm = ${n} - 1 - wC;\n\n            // TO DO: Vec4 over the channelMul\n            for (int dm = 0; dm < ${o}; dm++) {\n              int d2 = d1 * ${o} + dm;\n              float xValue = getDy(batch, idyR, idyC, d2);\n              float wValue = getW(wRPerm, wCPerm, d1, dm);\n              dotProd += xValue * wValue;\n            }\n          }\n        }\n        setOutput(dotProd);\n      }\n    `}};var mV={kernelName:Rt,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a,dy:s}=t,{strides:i,dilations:o,pad:l,dimRoundingMode:u,filterShape:h}=r,d=Gf.computeConv2DInfo(a.shape,h,i,o,l,u,!0),p=new cV(d);return n.runWebGLProgram(p,[a,s],"float32")}};var gV={kernelName:Ft,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{dy:a,filter:s}=t,{strides:i,dilations:o,pad:l,dimRoundingMode:u,inputShape:h}=r,d=Gf.computeConv2DInfo(h,s.shape,i,o,l,u,!0),p=new fV(d);return n.runWebGLProgram(p,[a,s],"float32")}},yV=class{constructor(e){this.variableNames=["X"],this.outputShape=[e,e],this.userCode="\n      void main() {\n          ivec2 coords = getOutputCoords();\n          float val = coords[0] == coords[1] ? getX(coords[0]) : 0.0;\n          setOutput(val);\n      }\n    "}};var bV={kernelName:Dt,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n}=e,{x:r}=t,a=[...r.shape,...r.shape],s=va.sizeFromShape(r.shape),i=Zz({inputs:{x:r},backend:n,attrs:{shape:[s]}}),o=new yV(s),l=n.runWebGLProgram(o,[i],i.dtype),u=Zz({inputs:{x:l},backend:n,attrs:{shape:a}});return n.disposeIntermediateTensorInfo(i),n.disposeIntermediateTensorInfo(l),u}},xV=class{constructor(e){this.variableNames=["x","W"],this.outputShape=e.outShape;let{inHeight:t,inWidth:n,padInfo:r,strideHeight:a,strideWidth:s,filterHeight:i,filterWidth:o,dilationHeight:l,dilationWidth:u}=e,{top:h,left:d}=r;this.userCode=`\n      const ivec2 strides = ivec2(${a}, ${s});\n      const ivec2 pads = ivec2(${h}, ${d});\n      const float neg_infinity = -3.4e38;\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int batch = coords.x;\n        int d1 = coords.w;\n        ivec2 outTopLeftCorner =\n            coords.yz * strides - pads;\n        int hBeg = outTopLeftCorner.x;\n        int wBeg = outTopLeftCorner.y;\n\n        float curVal = neg_infinity;\n        for (int h = 0; h < ${i}; h++) {\n          int hIn = hBeg + h * ${l};\n\n          if (hIn >= 0 && hIn < ${t}) {\n            for (int w = 0; w < ${o}; w++) {\n              int wIn = wBeg + w * ${u};\n\n              if (wIn >= 0 && wIn < ${n}) {\n                float xVal = getX(batch, hIn, wIn, d1);\n                float wVal = getW(h, w, d1);\n\n                float val = xVal + wVal;\n                if (val > curVal) {\n                  curVal = val;\n                }\n              }\n            }\n          }\n        }\n\n        float result = curVal;\n        setOutput(result);\n      }\n    `}};var vV={kernelName:Mt,backendName:"webgl",kernelFunc:function(e){let t,{inputs:n,backend:r,attrs:a}=e,{x:s,filter:i}=n,{strides:o,pad:l,dilations:u}=a,h=Gf.computeDilation2DInfo(s.shape,i.shape,o,l,"NHWC",u),d=new xV(h);t=r.runWebGLProgram(d,[s,i],"float32");let p=Zz({inputs:{x:t},backend:r,attrs:{shape:h.outShape}});return r.disposeIntermediateTensorInfo(t),p}};var wV={kernelName:Bt,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{equation:a}=r,s=t,{allDims:i,summedDims:o,idDims:l}=Gf.decodeEinsumEquation(a,s.length);Gf.checkEinsumDimSizes(i.length,l,s);let{path:u,steps:h}=Gf.getEinsumComputePath(o,l),d=h.length,p=null,c=i.length,f=[];for(let e=0;e<d;++e){for(let t of h[e]){let e,{permutationIndices:r,expandDims:a}=Gf.getEinsumPermutation(c,l[t]);Gf.isIdentityPermutation(r)?e=s[t]:(e=iB({inputs:{x:s[t]},backend:n,attrs:{perm:r}}),f.push(e));let i=e.shape.slice();for(let e=0;e<a.length;++e)i.splice(a[e],0,1);va.arraysEqual(e.shape,i)||(e=Zz({inputs:{x:e},backend:n,attrs:{shape:i}}),f.push(e)),null===p?p=e:(p=Kz({inputs:{a:e,b:p},backend:n}),f.push(p))}e<d-1&&(u[e]>=0&&(p=aB({inputs:{x:p},backend:n,attrs:{axis:u[e]-(i.length-c),keepDims:!1}}),f.push(p)),c--)}for(let e of f)e!==p&&n.disposeIntermediateTensorInfo(e);return p}},kV=Bz({opSnippet:"return (x >= 0.0) ? x : (exp(x) - 1.0);",packedOpSnippet:"\n  vec4 result;\n\n  result.r = (x.r >= 0.0) ? x.r : (exp(x.r) - 1.0);\n  result.g = (x.g >= 0.0) ? x.g : (exp(x.g) - 1.0);\n  result.b = (x.b >= 0.0) ? x.b : (exp(x.b) - 1.0);\n  result.a = (x.a >= 0.0) ? x.a : (exp(x.a) - 1.0);\n\n  return result;\n"}),IV={kernelName:Wt,backendName:"webgl",kernelFunc:kV},SV={kernelName:Vt,backendName:"webgl",kernelFunc:e=>{let{inputs:t,backend:n}=e,{dy:r,y:a}=t,s=Le().getBool("WEBGL_PACK_BINARY_OPERATIONS")?new Cz("\n  vec4 bGTEZero = vec4(greaterThanEqual(b, vec4(0.)));\n  return (bGTEZero * a) + ((vec4(1.0) - bGTEZero) * (a * (b + vec4(1.0))));\n",r.shape,a.shape):new Nz("return (b >= 0.0) ? a : a * (b + 1.0);",r.shape,a.shape);return n.runWebGLProgram(s,[r,a],r.dtype)}},_V=Wz({opSnippet:"return float(a == b);",packedOpSnippet:"\n  return vec4(equal(a, b));\n",dtype:"bool",cpuKernelImpl:fP}),NV={kernelName:Gt,backendName:"webgl",kernelFunc:_V},TV=Bz({opSnippet:`\n  // Error function is calculated approximately with elementary function.\n  // See "Handbook of Mathematical Functions with Formulas,\n  // Graphs, and Mathematical Tables", Abramowitz and Stegun.\n  float p = ${Gf.ERF_P};\n  float a1 = ${Gf.ERF_A1};\n  float a2 = ${Gf.ERF_A2};\n  float a3 = ${Gf.ERF_A3};\n  float a4 = ${Gf.ERF_A4};\n  float a5 = ${Gf.ERF_A5};\n\n  float sign = sign(x);\n  x = abs(x);\n  float t = 1.0 / (1.0 + p * x);\n  return sign * (1.0 - (((((a5*t + a4)*t) + a3)*t + a2)*t + a1)*t*exp(-x*x));\n`}),CV={kernelName:Ut,backendName:"webgl",kernelFunc:TV},EV=Bz({opSnippet:zz+"\n  return exp(x);\n",packedOpSnippet:"\n  vec4 result = exp(x);\n  bvec4 isNaN = isnan(x);\n  result.r = isNaN.r ? x.r : result.r;\n  result.g = isNaN.g ? x.g : result.g;\n  result.b = isNaN.b ? x.b : result.b;\n  result.a = isNaN.a ? x.a : result.a;\n\n  return result;\n",cpuKernelImpl:mP,dtype:"float32"}),AV={kernelName:Ht,backendName:"webgl",kernelFunc:EV};function $V(e){let{inputs:t,attrs:n,backend:r}=e,{dim:a}=n,{input:s}=t,i=s.shape.length,o=s.shape.slice(),l=a;return a<0&&(va.assert(-(i+1)<=a,()=>`Axis must be in the interval [${-(i+1)}, ${i}]`),l=i+a+1),o.splice(l,0,1),Zz({inputs:{x:s},backend:r,attrs:{shape:o}})}var RV={kernelName:jt,backendName:"webgl",kernelFunc:$V},FV="return exp(x) - 1.0;",DV=Bz({opSnippet:FV,packedOpSnippet:FV,cpuKernelImpl:gP}),MV={kernelName:qt,backendName:"webgl",kernelFunc:DV},OV=class{constructor(e,t,n){this.variableNames=["real","imag"];let r=t[1];this.outputShape=t;let a,s=n?`2.0 * ${Math.PI}`:`-2.0 * ${Math.PI}`,i=n?`${r}.0`:"1.0";if("real"===e)a="return real * expR - imag * expI;";else{if("imag"!==e)throw new Error(`FFT component must be either "real" or "imag", got ${e}.`);a="return real * expI + imag * expR;"}this.userCode=`\n      const float exponentMultiplier = ${s};\n\n      float unaryOpComplex(float real, float expR, float imag, float expI) {\n        ${a}\n      }\n\n      float mulMatDFT(int batch, int index) {\n        float indexRatio = float(index) / float(${r});\n        float exponentMultiplierTimesIndexRatio =\n            exponentMultiplier * indexRatio;\n\n        float result = 0.0;\n\n        for (int i = 0; i < ${r}; i++) {\n          // x = (-2|2 * PI / N) * index * i;\n          float x = exponentMultiplierTimesIndexRatio * float(i);\n          float expR = cos(x);\n          float expI = sin(x);\n          float real = getReal(batch, i);\n          float imag = getImag(batch, i);\n\n          result +=\n              unaryOpComplex(real, expR, imag, expI) / ${i};\n        }\n\n        return result;\n      }\n\n      void main() {\n        ivec2 coords = getOutputCoords();\n        setOutput(mulMatDFT(coords[0], coords[1]));\n      }\n    `}};function LV(e,t,n){let r=n.texData.get(e.dataId),a=va.sizeFromShape(e.shape),s=e.shape[e.shape.length-1],i=Zz({inputs:{x:e},backend:n,attrs:{shape:[a/s,s]}}),o=i.shape,l=new OV("real",o,t),u=new OV("imag",o,t),h=[{dataId:r.complexTensorInfos.real.dataId,dtype:r.complexTensorInfos.real.dtype,shape:o},{dataId:r.complexTensorInfos.imag.dataId,dtype:r.complexTensorInfos.imag.dtype,shape:o}],d=n.runWebGLProgram(l,h,"float32"),p=n.runWebGLProgram(u,h,"float32"),c=$z({inputs:{real:d,imag:p},backend:n});n.disposeIntermediateTensorInfo(d),n.disposeIntermediateTensorInfo(p);let f=Zz({inputs:{x:c},backend:n,attrs:{shape:e.shape}});return n.disposeIntermediateTensorInfo(i),n.disposeIntermediateTensorInfo(c),f}var PV={kernelName:Kt,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n}=e,{input:r}=t;return LV(r,!1,n)}},zV=class{constructor(e,t){this.outputShape=[],this.customUniforms=[{name:"value",type:"float"}],this.variableNames=["x"],this.outputShape=e,this.userCode="\n      void main() {\n        // Input can be obtained from uniform value.\n        setOutput(value);\n      }\n    "}};function BV(e){let{backend:t,attrs:n}=e,{shape:r,value:a}=n,{dtype:s}=n;if(s=s||va.inferDtype(a),"string"===s){let e=va.getArrayFromDType(s,va.sizeFromShape(r));return e.fill(a),t.makeTensorInfo(r,s,e)}{let e=new zV(r,a),n=[[a]];return t.runWebGLProgram(e,[],s,n)}}var WV,VV={kernelName:Xt,backendName:"webgl",kernelFunc:BV},UV=class{constructor(e){this.variableNames=["Image"],this.outputShape=[];let t=e[2];this.outputShape=e,this.userCode=`\n        void main() {\n          ivec4 coords = getOutputCoords();\n          int x = coords[2];\n\n          int coordX = ${t} - x - 1;\n          float outputValue;\n          if(coordX >= 0 && coordX < ${t}) {\n            outputValue = getImage(coords[0], coords[1], coordX, coords[3]);\n          } else {\n            outputValue = getImage(coords[0], coords[1], coords[2], coords[3]);\n          }\n          setOutput(outputValue);\n        }\n    `}},GV={kernelName:Zt,backendName:"webgl",kernelFunc:({inputs:e,backend:t})=>{let{image:n}=e,r=t,a=new UV(n.shape);return r.runWebGLProgram(a,[n],n.dtype)}},HV="return floor(x);",jV=Bz({opSnippet:HV,packedOpSnippet:HV,cpuKernelImpl:yP}),qV={kernelName:Yt,backendName:"webgl",kernelFunc:jV},KV=Wz({opSnippet:"\n  float s = sign(a) * sign(b);\n  int ia = round(a);\n  int ib = round(b);\n  if (ib != 0) {\n    // Windows (D3D) wants guaranteed non-zero int division at compile-time.\n    return float(idiv(ia, ib, s));\n  } else {\n    return NAN;\n  }\n",packedOpSnippet:"\n  ivec4 ia = round(a);\n  ivec4 ib = round(b);\n  bvec4 cond = notEqual(ib, ivec4(0));\n  ivec4 result = ivec4(0);\n  vec4 s = sign(a) * sign(b);\n\n  // Windows (D3D) wants guaranteed non-zero int division at compile-time.\n  if (cond[0]) {\n    result[0] = idiv(ia[0], ib[0], s[0]);\n  }\n  if (cond[1]) {\n    result[1] = idiv(ia[1], ib[1], s[1]);\n  }\n  if (cond[2]) {\n    result[2] = idiv(ia[2], ib[2], s[2]);\n  }\n  if (cond[3]) {\n    result[3] = idiv(ia[3], ib[3], s[3]);\n  }\n  return vec4(result);\n",dtype:"int32"}),XV={kernelName:Jt,backendName:"webgl",kernelFunc:KV},ZV=class{constructor(e){this.variableNames=["A"];let t=iL(),[n,r]=e;this.outputShape=e,this.userCode=`\n      void main() {\n        ivec3 coords = getOutputCoords();\n        int texR = coords[0];\n        int texC = coords[1];\n        int depth = coords[2];\n        vec2 uv = (vec2(texC, texR) + halfCR) / vec2(${r}.0, ${n}.0);\n\n        vec4 values = ${t.texture2D}(A, uv);\n        float value;\n        if (depth == 0) {\n          value = values.r;\n        } else if (depth == 1) {\n          value = values.g;\n        } else if (depth == 2) {\n          value = values.b;\n        } else if (depth == 3) {\n          value = values.a;\n        }\n\n        setOutput(floor(value * 255.0 + 0.5));\n      }\n    `}},YV=class{constructor(e){this.variableNames=["A"],this.packedInputs=!1,this.packedOutput=!0;let t=iL(),[n,r]=e;this.outputShape=e,this.userCode=`\n      void main() {\n        ivec3 coords = getOutputCoords();\n        int texR = coords[0];\n        int texC = coords[1];\n        int depth = coords[2];\n\n        vec4 result = vec4(0.);\n\n        for(int row=0; row<=1; row++) {\n          for(int col=0; col<=1; col++) {\n            texC = coords[1] + row;\n            depth = coords[2] + col;\n\n            vec2 uv = (vec2(texC, texR) + halfCR) /\n                       vec2(${r}.0, ${n}.0);\n            vec4 values = ${t.texture2D}(A, uv);\n            float value;\n            if (depth == 0) {\n              value = values.r;\n            } else if (depth == 1) {\n              value = values.g;\n            } else if (depth == 2) {\n              value = values.b;\n            } else if (depth == 3) {\n              value = values.a;\n            }\n\n            result[row * 2 + col] = floor(value * 255.0 + 0.5);\n          }\n        }\n\n        ${t.output} = result;\n      }\n    `}},JV={kernelName:na,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{pixels:a}=t,{numChannels:s}=r,i="undefined"!=typeof HTMLVideoElement&&a instanceof HTMLVideoElement,o="undefined"!=typeof HTMLImageElement&&a instanceof HTMLImageElement,[l,u]=i?[a.videoWidth,a.videoHeight]:[a.width,a.height],h=[u,l],d=[u,l,s];if(o||i){let e=Le().getBool("CANVAS2D_WILL_READ_FREQUENTLY_FOR_GPU");(null==WV||e!==QV)&&(QV=e,WV=document.createElement("canvas").getContext("2d",{willReadFrequently:QV})),WV.canvas.width=l,WV.canvas.height=u,WV.drawImage(a,0,0,l,u),a=WV.canvas}let p=n.makeTensorInfo(h,"int32");n.texData.get(p.dataId).usage=QM.PIXELS,n.gpgpu.uploadPixelDataToTexture(n.getTexture(p.dataId),a);let c=Le().getBool("WEBGL_PACK")?new YV(d):new ZV(d),f=n.runWebGLProgram(c,[p],"int32");return n.disposeData(p.dataId),f}},QV=Le().getBool("CANVAS2D_WILL_READ_FREQUENTLY_FOR_GPU");var eU={kernelName:sa,backendName:"webgl",kernelFunc:function(e){let t,{inputs:n,backend:r,attrs:a}=e,{x:s,filter:i,bias:o,preluActivationWeights:l}=n,{strides:u,pad:h,dataFormat:d,dilations:p,dimRoundingMode:c,activation:f,leakyreluAlpha:m}=a,g=Gf.convertConv2DDataFormat(d),y=Gf.computeConv2DInfo(s.shape,i.shape,u,p,h,c,!1,g),b=[],x=null!=o,v=null!=l,w="leakyrelu"===f,k=()=>{let e=[s,i],t=(e,t)=>{if("NCHW"===t&&1===e.shape.length&&1!==e.shape[0]){let t=Zz({inputs:{x:e},backend:r,attrs:{shape:[e.shape[0],1,1]}});return b.push(t),t}return e};if(x&&e.push(t(o,d)),v&&e.push(t(l,d)),w){let t=r.makeTensorInfo([],"float32",va.createScalarValue(m,"float32"));e.push(t),b.push(t)}return e};if(1!==y.filterHeight||1!==y.filterWidth||1!==y.dilationHeight||1!==y.dilationWidth||1!==y.strideHeight||1!==y.strideWidth||"SAME"!==y.padInfo.type&&"VALID"!==y.padInfo.type)if(y.strideWidth<=2&&"channelsLast"===g&&Le().getBool("WEBGL_EXP_CONV")){let e=f?Vz(f,!0):null,n=new RW(y,x,e,v,w),a=[[y.padInfo.top,y.padInfo.left],[y.strideHeight,y.strideWidth],[y.dilationHeight,y.dilationWidth],[y.inHeight,y.inWidth]],s=k();t=r.runWebGLProgram(n,s,"float32",a)}else if(Le().getBool("WEBGL_CONV_IM2COL"))t=OW({x:s,filter:i,convInfo:y,backend:r,bias:o,activation:f,preluActivationWeights:l,leakyreluAlpha:m});else{let e=f?Vz(f,!1):null,n=new AW(y,x,e,v,w),a=k();t=r.runWebGLProgram(n,a,"float32")}else t=MW({x:s,filter:i,convInfo:y,backend:r,bias:o,activation:f,preluActivationWeights:l,leakyreluAlpha:m});let I=Zz({inputs:{x:t},backend:r,attrs:{shape:y.outShape}});return b.push(t),b.forEach(e=>r.disposeIntermediateTensorInfo(e)),I}};var tU={kernelName:ia,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a,filter:s,bias:i,preluActivationWeights:o}=t,{strides:l,pad:u,dilations:h,dimRoundingMode:d,activation:p,leakyreluAlpha:c}=r,f=[],m=h;null==m&&(m=[1,1]),va.assert(Gf.eitherStridesOrDilationsAreOne(l,m),()=>`Error in depthwiseConv2d: Either strides or dilations must be 1. Got strides ${l} and dilations '${m}'`);let g,y=Gf.computeConv2DInfo(a.shape,s.shape,l,m,u,d,!0),b=Le().getBool("WEBGL_PACK_DEPTHWISECONV")&&y.strideWidth<=2&&y.outChannels/y.inChannels===1,x=p?Vz(p,b):null,v=[a,s],w=null!=i,k=null!=o,I="leakyrelu"===p;if(w&&v.push(i),k&&v.push(o),I){let e=n.makeTensorInfo([],"float32",va.createScalarValue(c,"float32"));v.push(e),f.push(e)}g=b?new dV(y,w,x,k,I):new hV(y,w,x,k,I);let S=[[y.padInfo.top,y.padInfo.left],[y.strideHeight,y.strideWidth],[y.dilationHeight,y.dilationWidth],[y.inHeight,y.inWidth]],_=n.runWebGLProgram(g,v,"float32",S);return f.forEach(e=>n.disposeIntermediateTensorInfo(e)),_}},nU=class{constructor(e,t,n,r){this.sliceDim=e,this.strides=t,this.paramsShape=r,this.variableNames=["x","indices"],this.outputShape=n;let a=IL(n.length),s="\n    int index;";for(let e=0;e<this.sliceDim;e++)s+=`\n          index = round(getIndices(coords[0], ${e}));\n          out_of_bounds = out_of_bounds || index < 0;\n          out_of_bounds = out_of_bounds || index >= ${this.paramsShape[e]};\n          flattenIndex += index * ${this.strides[e]};`;this.userCode=`\n         void main() {\n          ${a} coords = getOutputCoords();\n          int flattenIndex = 0;\n          bool out_of_bounds = false;\n\n          ${s}\n\n          setOutput(out_of_bounds ? 0.0 : getX(flattenIndex, coords[1]));\n        }\n      `}};var rU={kernelName:tn,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n}=e,{params:r,indices:a}=t,s=a.shape,i=s[s.length-1],o=va.sizeFromShape(r.shape),[l,u,h,d]=Gf.prepareAndValidate(r,a),p=Zz({inputs:{x:a},backend:n,attrs:{shape:[u,i]}}),c=Zz({inputs:{x:r},backend:n,attrs:{shape:[va.sizeFromShape(r.shape)/h,h]}});if(n.shouldExecuteOnCPU([r,a])||"string"===r.dtype){let e=n.readSync(a.dataId),t=n.bufferSync(r),s=bP(e,t,r.dtype,u,i,h,d,r.shape,o);return n.makeTensorInfo(l,r.dtype,s.values)}let f=new nU(i,d,[u,h],r.shape),m=n.runWebGLProgram(f,[c,p],c.dtype),g=Zz({inputs:{x:m},backend:n,attrs:{shape:l}});return n.disposeIntermediateTensorInfo(p),n.disposeIntermediateTensorInfo(c),n.disposeIntermediateTensorInfo(m),g}},aU=class{constructor(e,t){this.variableNames=["A","indices"],this.outputShape=t,this.rank=t.length;let n=IL(this.rank),r=function(e){let t=["resRC.x","resRC.y","resRC.z","resRC.w"],n=[];for(let r=0;r<e.length;r++)2===r?n.push("index"):n.push(`${t[r]}`);return n.join()}(e);this.userCode=`\n      void main() {\n        ${n} resRC = getOutputCoords();\n        int index = int(getIndices(resRC.x, resRC.z));\n        float inBounds = (index >= 0) && (index < ${e[2]}) ? 1.0 : 0.0;\n        setOutput(inBounds * getA(${r}));\n      }\n    `}};function sU(e){let{inputs:t,backend:n,attrs:r}=e,{x:a,indices:s}=t,{axis:i,batchDims:o}=r,l=va.parseAxisParam(i,a.shape)[0];if(Le().get("DEBUG")){let e=n.readSync(s.dataId),t=a.shape[l];for(let n=0;n<e.length;++n){let r=e[n];va.assert(r<=t-1&&r>=0,()=>`GatherV2: the index value ${r} is not in [0, ${t-1}]`)}}let u=Gf.segment_util.collectGatherOpShapeInfo(a,s,l,o),h=va.sizeFromShape(s.shape),d=[],p=Zz({inputs:{x:a},backend:n,attrs:{shape:[u.batchSize,u.outerSize,u.dimSize,u.sliceSize]}}),c=Zz({inputs:{x:s},backend:n,attrs:{shape:[u.batchSize,h/u.batchSize]}});d.push(p),d.push(c);let f=[u.batchSize,u.outerSize,h/u.batchSize,u.sliceSize];if(n.shouldExecuteOnCPU([a,s])||"string"===a.dtype){let e=n.bufferSync(c),t=n.bufferSync(p),r=xP(t,e,f);return d.forEach(e=>n.disposeIntermediateTensorInfo(e)),n.makeTensorInfo(u.outputShape,r.dtype,r.values)}let m=new aU(p.shape,f),g=n.runWebGLProgram(m,[p,c],p.dtype);d.push(g);let y=Zz({inputs:{x:g},backend:n,attrs:{shape:u.outputShape}});return d.forEach(e=>n.disposeIntermediateTensorInfo(e)),y}var iU={kernelName:en,backendName:"webgl",kernelFunc:sU},oU=Wz({opSnippet:"return float(a > b);",packedOpSnippet:"\n  return vec4(greaterThan(a, b));\n",cpuKernelImpl:vP,dtype:"bool"}),lU={kernelName:nn,backendName:"webgl",kernelFunc:oU},uU=Wz({opSnippet:"return float(a >= b);",packedOpSnippet:"\n  return vec4(greaterThanEqual(a, b));\n",dtype:"bool",cpuKernelImpl:wP}),hU={kernelName:rn,backendName:"webgl",kernelFunc:uU};var dU={kernelName:sn,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n}=e,{input:r}=t;return LV(r,!0,n)}},pU=Bz({opSnippet:"return float(!isnan(x) && !isinf(x));",dtype:"bool"}),cU={kernelName:ln,backendName:"webgl",kernelFunc:pU},fU=Bz({opSnippet:"return float(isinf(x));",dtype:"bool"}),mU={kernelName:un,backendName:"webgl",kernelFunc:fU},gU=Bz({opSnippet:"return float(isnan(x));",dtype:"bool"}),yU={kernelName:hn,backendName:"webgl",kernelFunc:gU},bU=Wz({opSnippet:"return float(a < b);",packedOpSnippet:"\n  return vec4(lessThan(a, b));\n",cpuKernelImpl:kP,dtype:"bool"}),xU={kernelName:pn,backendName:"webgl",kernelFunc:bU},vU=Wz({opSnippet:"return float(a <= b);",packedOpSnippet:"\n  return vec4(lessThanEqual(a, b));\n",cpuKernelImpl:IP,dtype:"bool"}),wU={kernelName:cn,backendName:"webgl",kernelFunc:vU};var kU={kernelName:fn,backendName:"webgl",kernelFunc:function(e){let{backend:t,attrs:n}=e,{start:r,stop:a,num:s}=n,i=SP(r,a,s);return t.makeTensorInfo([i.length],"float32",i)}},IU=Bz({opSnippet:zz+"\n  return x < 0.0 ? 0./0. : log(x);\n",packedOpSnippet:"\n  vec4 result = log(x);\n  bvec4 isNaN = isnan(x);\n  result.r = isNaN.r ? x.r : (x.r < 0.0 ? 0./0. : result.r);\n  result.g = isNaN.g ? x.g : (x.g < 0.0 ? 0./0. : result.g);\n  result.b = isNaN.b ? x.b : (x.b < 0.0 ? 0./0. : result.b);\n  result.a = isNaN.a ? x.a : (x.a < 0.0 ? 0./0. : result.a);\n  return result;\n",cpuKernelImpl:_P}),SU={kernelName:mn,backendName:"webgl",kernelFunc:IU},_U=Bz({opSnippet:zz+"\n  return log(1.0 + x);\n"}),NU={kernelName:gn,backendName:"webgl",kernelFunc:_U},TU=Wz({opSnippet:"return float(a >= 1.0 && b >= 1.0);",packedOpSnippet:"\n  return vec4(\n    vec4(greaterThanEqual(a, vec4(1.0))) *\n    vec4(greaterThanEqual(b, vec4(1.0))));\n",dtype:"bool"}),CU={kernelName:yn,backendName:"webgl",kernelFunc:TU},EU=Bz({opSnippet:"return float(!(x >= 1.0));"}),AU={kernelName:bn,backendName:"webgl",kernelFunc:EU},$U=Wz({opSnippet:"return float(a >= 1.0 || b >= 1.0);",packedOpSnippet:"\n  return min(\n    vec4(greaterThanEqual(a, vec4(1.0))) +\n    vec4(greaterThanEqual(b, vec4(1.0))),\n    vec4(1.0));\n",dtype:"bool"}),RU={kernelName:xn,backendName:"webgl",kernelFunc:$U},FU=class{constructor(e,t,n,r,a){this.variableNames=["x"],this.outputShape=[];let s=t,i=e[3]-1;this.outputShape=e;let o,l=`float(${n}) + float(${r}) * sum`;o=.5===a?`inversesqrt(${l})`:1===a?`1.0/(${l})`:`exp(log(${l}) * float(-${a}));`,this.userCode=`\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int b = coords[0];\n        int r = coords[1];\n        int c = coords[2];\n        int d = coords[3];\n        float x = getX(b, r, c, d);\n        float sum = 0.0;\n        for (int j = -${s}; j <= ${s}; j++) {\n          int idx = d + j;\n          if (idx >= 0 && idx <=  ${i}) {\n            float z = getX(b, r, c, idx);\n            sum += z * z;\n          }\n        }\n        float val = x * ${o};\n        setOutput(val);\n      }\n    `}},DU=class{constructor(e,t,n,r,a){this.variableNames=["x"],this.outputShape=[],this.packedInputs=!0,this.packedOutput=!0;let s=t,i=e[3]-1;this.outputShape=e;let o,l=`float(${n}) + float(${r}) * sum`;o=.5===a?`inversesqrt(${l})`:1===a?`1.0/(${l})`:`exp(log(${l}) * float(-${a}));`,this.userCode=`\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int b = coords.x;\n        int r = coords.y;\n        int c = coords.z;\n        int d = coords.w;\n\n        bool hasNextCol = d < ${this.outputShape[3]};\n        bool hasNextRow = c < ${this.outputShape[2]};\n\n        vec4 sum = vec4(0.);\n        vec4 xFragAtOutputCoords = getX(b, r, c, d);\n\n        vec4 xAtOutputCoords = vec4(\n          getChannel(xFragAtOutputCoords, vec2(c, d)),\n          hasNextCol ?\n            getChannel(xFragAtOutputCoords, vec2(c, d + 1)) : 0.0,\n          hasNextRow ?\n            getChannel(xFragAtOutputCoords , vec2(c + 1, d)) : 0.0,\n          (hasNextRow && hasNextCol) ?\n            getChannel(xFragAtOutputCoords, vec2(c + 1, d + 1)) : 0.0\n        );\n\n        int firstChannel = d - ${s};\n        vec2 cache = vec2(0.);\n        if(firstChannel >= 0){\n          vec4 firstChannelFrag = getX(b, r, c, firstChannel);\n          cache.x = getChannel(firstChannelFrag, vec2(c, firstChannel));\n            if(hasNextRow){\n              cache.y = getChannel(firstChannelFrag, vec2(c + 1, firstChannel));\n            }\n        }\n\n        ivec2 depth = ivec2(d, d + 1);\n        for (int j = - ${s}; j <= ${s}; j++) {\n          ivec2 idx = depth + j;\n          bvec2 aboveLowerBound = greaterThanEqual(idx, ivec2(0));\n          bvec2 belowUpperBound = lessThanEqual(idx, ivec2(${i}));\n\n          bool depthInRange = aboveLowerBound.x && belowUpperBound.x;\n          bool depthPlusOneInRange = aboveLowerBound.y && belowUpperBound.y;\n\n          if(depthInRange || depthPlusOneInRange){\n            vec4 z = vec4(0.);\n            vec4 xFragAtCurrentDepth;\n            z.xz = cache.xy;\n            if(depthPlusOneInRange && hasNextCol){\n              xFragAtCurrentDepth = idx.y != d ?\n                getX(b, r, c, idx.y) : xFragAtOutputCoords;\n              z.y = getChannel(xFragAtCurrentDepth, vec2(c, idx.y));\n              if(hasNextRow){\n                z.w = getChannel(xFragAtCurrentDepth, vec2(c + 1, idx.y));\n              }\n            }\n            cache.xy = z.yw;\n            sum += z * z;\n          }\n        }\n        vec4 result = xAtOutputCoords * ${o};\n        setOutput(result);\n      }\n    `}},MU={kernelName:In,backendName:"webgl",kernelFunc:e=>{let{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{depthRadius:s,bias:i,alpha:o,beta:l}=r,u=Le().getBool("WEBGL_PACK_NORMALIZATION")?new DU(a.shape,s,i,o,l):new FU(a.shape,s,i,o,l);return n.runWebGLProgram(u,[a],a.dtype)}},OU=class{constructor(e,t,n,r,a){this.variableNames=["inputImage","outputImage","dy"],this.outputShape=[],this.outputShape=e,this.depth=e[3],this.depthRadius=t,this.bias=n,this.alpha=r,this.beta=a,this.userCode=`\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int b = coords[0];\n        int r = coords[1];\n        int c = coords[2];\n\n        float result = 0.0;\n        for (int d = 0; d < ${this.depth}; ++d) {\n          int depthBegin = int(max(0.0, float(d - ${t})));\n          int depthEnd = int(min(float(${this.depth}),\n              float(d + ${t} + 1)));\n\n          const int MIN_DEPTH_BEGIN = 0;\n          const int MAX_DEPTH_END = ${this.depth};\n\n          float norm = 0.0;\n          for (int k = MIN_DEPTH_BEGIN; k < MAX_DEPTH_END; ++k) {\n            if (k < depthBegin){\n              continue;\n            }\n            else if (k >= depthBegin && k < depthEnd) {\n              norm += getInputImage(b, r, c, k) * getInputImage(b, r, c, k);\n            }\n            else {\n              break;\n            }\n          }\n\n          norm = float(${r}) * norm + float(${n});\n\n          for(int k = MIN_DEPTH_BEGIN; k < MAX_DEPTH_END; ++k){\n            if (k < depthBegin){\n              continue;\n            }\n            else if (k >= depthBegin && k < depthEnd){\n              float dyi = -2.0 * float(${r})\n                * float(${a})\n                * getInputImage(b, r, c, k) * getOutputImage(b, r, c, d)\n                / norm;\n              if (k == d) {\n                dyi += pow(norm, -1.0 * ${a});\n              }\n              if (k == coords[3]) {\n                dyi *= getDy(b, r, c, d);\n                result += dyi;\n              }\n            }\n            else {\n              break;\n            }\n          }\n      }\n      setOutput(result);\n      }\n    `}},LU={kernelName:Sn,backendName:"webgl",kernelFunc:e=>{let{inputs:t,backend:n,attrs:r}=e,{x:a,y:s,dy:i}=t,{depthRadius:o,bias:l,alpha:u,beta:h}=r,d=new OU(a.shape,o,l,u,h);return n.runWebGLProgram(d,[a,s,i],a.dtype)}};function PU(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{reductionIndices:s,keepDims:i}=r,o=a.shape.length,l=va.parseAxisParam(s,a.shape),u=l,h=Gf.getAxesPermutation(u,o),d=null!=h,p=n.shouldExecuteOnCPU([a]),c=a;if(d){if(p){let e=n.texData.get(c.dataId).values,t=new Array(o);for(let e=0;e<t.length;e++)t[e]=a.shape[h[e]];let r=ez(e,a.shape,a.dtype,h,t);c=n.makeTensorInfo(t,a.dtype),n.texData.get(c.dataId).values=r}else c=rB(a,h,n);u=Gf.getInnerMostAxes(u.length,o)}Gf.assertAxesAreInnerMostDims("max",u,o);let f,[m,g]=Gf.computeOutAndReduceShapes(c.shape,u),y=m;if(i&&(y=Gf.expandShapeToKeepDim(m,l)),p){let e=n.texData.get(c.dataId).values,t=NP(e,va.sizeFromShape(g),y,a.dtype);f=n.makeTensorInfo(y,a.dtype),n.texData.get(f.dataId).values=t}else f=function(e,t,n,r){let a=va.sizeFromShape(t),s=Zz({inputs:{x:e},attrs:{shape:[va.sizeFromShape(e.shape)/a,a]},backend:r}),i=eB(s,e.dtype,"max",r),o=Zz({inputs:{x:i},attrs:{shape:n},backend:r});return r.disposeIntermediateTensorInfo(s),r.disposeIntermediateTensorInfo(i),o}(c,g,y,n);return d&&n.disposeIntermediateTensorInfo(c),f}var zU={kernelName:Nn,backendName:"webgl",kernelFunc:PU},BU=Wz({opSnippet:_z+"\n  return max(a, b);\n",packedOpSnippet:"\n  vec4 result = vec4(max(a, b));\n  bvec4 isNaNA = isnan(a);\n  bvec4 isNaNB = isnan(b);\n  bvec4 isNaN = bvec4(isNaNA.x || isNaNB.x, isNaNA.y || isNaNB.y, isNaNA.z || isNaNB.z, isNaNA.w || isNaNB.w);\n  "+Tz+"\n  return result;\n",cpuKernelImpl:TP}),WU={kernelName:Tn,backendName:"webgl",kernelFunc:BU};var VU={kernelName:Cn,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t;aL(a,"maxPool");let{filterSize:s,strides:i,pad:o,dimRoundingMode:l}=r;va.assert(Gf.eitherStridesOrDilationsAreOne(i,1),()=>`Error in maxPool: Either strides or dilations must be 1. Got strides ${i} and dilations '1'`);let u=Gf.computePool2DInfo(a.shape,s,i,1,o,l);if(1===u.filterWidth&&1===u.filterHeight&&va.arraysEqual(u.inShape,u.outShape))return Ez({inputs:{x:a},backend:n});let h=new WB(u,"max",!1);return n.runWebGLProgram(h,[a],a.dtype)}};var UU={kernelName:An,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{filterSize:s,strides:i,pad:o,dataFormat:l,dimRoundingMode:u}=r,h=Gf.computePool3DInfo(a.shape,s,i,[1,1,1],o,u,l),d=new VB(h,"max",!1);return n.runWebGLProgram(d,[a],a.dtype)}},GU=class{constructor(e){this.variableNames=["dy","maxPos"],this.outputShape=e.inShape;let t=e.strideHeight,n=e.strideWidth,r=e.dilationHeight,a=e.effectiveFilterHeight,s=e.effectiveFilterWidth,i=a-1-e.padInfo.top,o=s-1-e.padInfo.left,l=a*s-1;this.userCode=`\n      const ivec2 pads = ivec2(${i}, ${o});\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int b = coords[0];\n        int d = coords[3];\n\n        ivec2 dyRCCorner = coords.yz - pads;\n        int dyRCorner = dyRCCorner.x;\n        int dyCCorner = dyRCCorner.y;\n\n        // Convolve dy(?, ?, d) with pos mask(:, :, d) to get dx(xR, xC, d).\n        // ? = to be determined. : = across all values in that axis.\n        float dotProd = 0.0;\n        for (int wR = 0; wR < ${a};\n          wR += ${r}) {\n          float dyR = float(dyRCorner + wR) / ${t}.0;\n\n          if (dyR < 0.0 || dyR >= ${e.outHeight}.0 || fract(dyR) > 0.0) {\n            continue;\n          }\n          int idyR = int(dyR);\n\n          for (int wC = 0; wC < ${s}; wC++) {\n            float dyC = float(dyCCorner + wC) / ${n}.0;\n\n            if (dyC < 0.0 || dyC >= ${e.outWidth}.0 ||\n                fract(dyC) > 0.0) {\n              continue;\n            }\n            int idyC = int(dyC);\n\n            float dyValue = getDy(b, idyR, idyC, d);\n            int maxPosValue = ${l} - int(getMaxPos(b, idyR, idyC, d));\n\n            // Get the current value, check it against the value from the\n            // position matrix.\n            int curPosValue = wR * ${s} + wC;\n            float mask = float(maxPosValue == curPosValue ? 1.0 : 0.0);\n\n            dotProd += dyValue * mask;\n          }\n        }\n        setOutput(dotProd);\n      }\n    `}},HU=class{constructor(e){this.variableNames=["dy","maxPos"],this.outputShape=e.inShape;let t=e.strideDepth,n=e.strideHeight,r=e.strideWidth,a=e.dilationDepth,s=e.dilationHeight,i=e.dilationWidth,o=e.effectiveFilterDepth,l=e.effectiveFilterHeight,u=e.effectiveFilterWidth,h=o-1-e.padInfo.front,d=l-1-e.padInfo.top,p=u-1-e.padInfo.left,c=o*l*u-1;this.userCode=`\n      const ivec3 pads = ivec3(${h}, ${d}, ${p});\n\n      void main() {\n        ivec5 coords = getOutputCoords();\n        int batch = coords.x;\n        int ch = coords.u;\n\n        ivec3 dyCorner = ivec3(coords.y, coords.z, coords.w) - pads;\n        int dyDCorner = dyCorner.x;\n        int dyRCorner = dyCorner.y;\n        int dyCCorner = dyCorner.z;\n\n        // Convolve dy(?, ?, ?, ch) with pos mask(:, :, :, d) to get\n        // dx(xD, xR, xC, ch).\n        // ? = to be determined. : = across all values in that axis.\n        float dotProd = 0.0;\n\n        for (int wD = 0; wD < ${o};\n           wD += ${a}) {\n          float dyD = float(dyDCorner + wD) / ${t}.0;\n\n          if (dyD < 0.0 || dyD >= ${e.outDepth}.0 || fract(dyD) > 0.0) {\n            continue;\n          }\n          int idyD = int(dyD);\n\n          for (int wR = 0; wR < ${l};\n              wR += ${s}) {\n            float dyR = float(dyRCorner + wR) / ${n}.0;\n\n            if (dyR < 0.0 || dyR >= ${e.outHeight}.0 ||\n                fract(dyR) > 0.0) {\n              continue;\n            }\n            int idyR = int(dyR);\n\n            for (int wC = 0; wC < ${u};\n                wC += ${i}) {\n              float dyC = float(dyCCorner + wC) / ${r}.0;\n\n              if (dyC < 0.0 || dyC >= ${e.outWidth}.0 ||\n                  fract(dyC) > 0.0) {\n                continue;\n              }\n              int idyC = int(dyC);\n\n              float dyValue = getDy(batch, idyD, idyR, idyC, ch);\n              int maxPosValue = ${c} -\n                  int(getMaxPos(batch, idyD, idyR, idyC, ch));\n\n              // Get the current value, check it against the value from the\n              // position matrix.\n              int curPosValue =\n                  wD * ${l} * ${u} +\n                  wR * ${u} + wC;\n              float mask = float(maxPosValue == curPosValue ? 1.0 : 0.0);\n\n              dotProd += dyValue * mask;\n            }\n          }\n        }\n        setOutput(dotProd);\n      }\n    `}};var jU={kernelName:$n,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{dy:a,input:s}=t,i=s,{filterSize:o,strides:l,pad:u,dimRoundingMode:h}=r,d=Gf.computePool3DInfo(i.shape,o,l,[1,1,1],u,h),p=new VB(d,"max",!0),c=n.runWebGLProgram(p,[i],i.dtype),f=new HU(d),m=n.runWebGLProgram(f,[a,c],i.dtype);return n.disposeIntermediateTensorInfo(c),m}};var qU={kernelName:En,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{dy:a,input:s,output:i}=t,o=s;aL([s,i],"maxPoolGrad");let{filterSize:l,strides:u,pad:h,dimRoundingMode:d}=r,p=Gf.computePool2DInfo(o.shape,l,u,1,h,d),c=new WB(p,"max",!0),f=n.runWebGLProgram(c,[o],o.dtype),m=new GU(p),g=n.runWebGLProgram(m,[a,f],o.dtype);return n.disposeIntermediateTensorInfo(f),g}};var KU={kernelName:Rn,backendName:"webgl",kernelFunc:({inputs:e,attrs:t,backend:n})=>{let{x:r}=e,{filterSize:a,strides:s,pad:i,includeBatchInIndex:o}=t,l=n;va.assert(4===r.shape.length,()=>`Error in maxPool: input must be rank 4 but got rank ${r.shape.length}.`);let u=[1,1];va.assert(Gf.eitherStridesOrDilationsAreOne(s,u),()=>`Error in maxPool: Either strides or dilations must be 1. Got strides ${s} and dilations '${u}'`);let h=Gf.computePool2DInfo(r.shape,a,s,u,i),[d,p]=function(e,t,n,r){let a=new WB(n,"max",!1),s=r.runWebGLProgram(a,[e],"float32");return a=new WB(n,"max",!0,!0,t),[s,r.runWebGLProgram(a,[e],"float32")]}(r,o,h,l);return[d,p]}};var XU={kernelName:Fn,backendName:"webgl",kernelFunc:({inputs:e,attrs:t,backend:n})=>{let{x:r}=e,{keepDims:a,axis:s}=t,i=n,o=r.shape.length,l=va.parseAxisParam(s,r.shape),u=l,h=Gf.getAxesPermutation(u,o),d=null!=h,p=i.shouldExecuteOnCPU([r]),c=[],f=r;if(d){if(p){let e=i.texData.get(f.dataId).values,t=new Array(o);for(let e=0;e<t.length;e++)t[e]=r.shape[h[e]];let n=ez(e,r.shape,r.dtype,h,t);f=i.makeTensorInfo(t,r.dtype),i.texData.get(f.dataId).values=n}else f=rB(r,h,i);c.push(f),u=Gf.getInnerMostAxes(u.length,o)}Gf.assertAxesAreInnerMostDims("sum",u,o);let[m,g]=Gf.computeOutAndReduceShapes(f.shape,u),y=m;a&&(y=Gf.expandShapeToKeepDim(m,l));let b=function(e,t,n,r){let a=va.sizeFromShape(t),s=Zz({inputs:{x:e},attrs:{shape:[va.sizeFromShape(e.shape)/a,a]},backend:r}),i=eB(s,"float32","mean",r),o=Zz({inputs:{x:i},attrs:{shape:n},backend:r});return r.disposeIntermediateTensorInfo(s),r.disposeIntermediateTensorInfo(i),o}(f,g,y,i);for(let e of c)i.disposeIntermediateTensorInfo(e);return b}};var ZU={kernelName:Dn,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{axis:s,keepDims:i}=r,o=a.shape.length,l=va.parseAxisParam(s,a.shape),u=l,h=Gf.getAxesPermutation(u,o),d=a;null!=h&&(d=iB({inputs:{x:a},backend:n,attrs:{perm:h}}),u=Gf.getInnerMostAxes(u.length,a.shape.length)),Gf.assertAxesAreInnerMostDims("min",u,o);let p,[c,f]=Gf.computeOutAndReduceShapes(d.shape,u),m=Zz({inputs:{x:d},backend:n,attrs:{shape:[-1,va.sizeFromShape(f)]}}),g=eB(m,m.dtype,"min",n);if(i){p=Zz({inputs:{x:g},backend:n,attrs:{shape:Gf.expandShapeToKeepDim(c,l)}})}else p=Zz({inputs:{x:g},backend:n,attrs:{shape:c}});return n.disposeIntermediateTensorInfo(m),n.disposeIntermediateTensorInfo(g),null!=h&&n.disposeIntermediateTensorInfo(d),p}},YU=Wz({opSnippet:_z+"\n  return min(a, b);\n",packedOpSnippet:"\n  vec4 result = vec4(min(a, b));\n  bvec4 isNaNA = isnan(a);\n  bvec4 isNaNB = isnan(b);\n  bvec4 isNaN = bvec4(isNaNA.x || isNaNB.x, isNaNA.y || isNaNB.y, isNaNA.z || isNaNB.z, isNaNA.w || isNaNB.w);\n  "+Tz+"\n  return result;\n",cpuKernelImpl:CP}),JU={kernelName:Mn,backendName:"webgl",kernelFunc:YU},QU=class{constructor(e,t,n){this.variableNames=["x"],this.outputShape=t.map((t,n)=>t[0]+e[n]+t[1]);let r=e.length,a=IL(r),s=t.map(e=>e[0]).join(","),i=t.map((t,n)=>t[0]+e[n]).join(","),o=["coords[0]","coords[1]","coords[2]","coords[3]"].slice(0,r),l="reflect"===n?0:1;this.userCode=1!==r?`\n      ${a} start = ${a}(${s});\n      ${a} end = ${a}(${i});\n\n      void main() {\n        ${a} outC = getOutputCoords();\n        for (int i = 0; i < ${r}; i++) {\n          if (outC[i] < start[i]) {\n            outC[i] = start[i] * 2 - outC[i] - ${l};\n          } else if(outC[i] >= end[i]) {\n            outC[i] = (end[i] - 1) * 2 - outC[i] + ${l};\n          }\n        }\n        ${a} coords = outC - start;\n        setOutput(getX(${o}));\n      }\n    `:`\n        int start = ${s};\n        int end = ${i};\n\n        void main() {\n          int outC = getOutputCoords();\n          if (outC < start) {\n            outC = start * 2 - outC - ${l};\n          } else if(outC >= end) {\n            outC = (end - 1) * 2 - outC + ${l};\n          }\n          setOutput(getX(outC - start));\n        }\n      `}},eG=class{constructor(e,t,n){this.variableNames=["x"],this.packedInputs=!0,this.packedOutput=!0,this.outputShape=t.map((t,n)=>t[0]+e[n]+t[1]);let r=e.length,a=IL(r),s=t.map(e=>e[0]).join(","),i=t.map((t,n)=>t[0]+e[n]).join(","),o=rz("rc",r),l=rz("source",r),u=`${o[r-1]} < ${this.outputShape[r-1]}`,h=1===r?"source":`vec2(${l.slice(-2).join()})`,d="reflect"===n?0:1,p="";if(1===r){let e=`\n        ${a} source = rc;\n        if (source < start) {\n          source = start * 2 - source - ${d};\n        } else if (source >= end) {\n          source = (end - 1) * 2 - source + ${d};\n        }\n        source -= start;\n      `;p=`\n        ${a} rc = outputLoc;\n        ${e}\n        result[0] = getChannel(getX(${l.join()}), ${h});\n        ${o[r-1]} += 1;\n        if(${u}) {\n          ${e}\n          result[1] = getChannel(getX(${l.join()}), ${h});\n        }\n      `}else{let e=`\n        ${a} source = rc;\n        ${a} lt = ${a}(lessThan(source, start));\n        ${a} gte = ${a}(greaterThanEqual(source, end));\n        ${a} orig = 1 - (lt + gte);\n        source = orig * source +\n                lt * (start * 2 - source - ${d}) +\n                gte * ((end - 1) * 2 - source + ${d});\n        source -= start;\n      `;p=`\n        ${a} rc = outputLoc;\n        ${e}\n        result[0] = getChannel(getX(${l.join()}), ${h});\n        ${o[r-1]} += 1;\n        if(${u}) {\n          ${e}\n          result[1] = getChannel(getX(${l.join()}), ${h});\n        }\n        rc = outputLoc;\n        ${o[r-2]} += 1;\n        if(${o[r-2]} < ${this.outputShape[r-2]}) {\n          ${e}\n          result[2] = getChannel(getX(${l.join()}), ${h});\n          ${o[r-1]} += 1;\n          if(${u}) {\n            ${e}\n            result[3] = getChannel(getX(${l.join()}), ${h});\n          }\n        }\n      `}this.userCode=`\n      const ${a} start = ${a}(${s});\n      const ${a} end = ${a}(${i});\n\n      void main() {\n        ${a} outputLoc = getOutputCoords();\n        vec4 result = vec4(0.);\n        ${p}\n        setOutput(result);\n      }\n    `}},tG={kernelName:On,backendName:"webgl",kernelFunc:({inputs:e,backend:t,attrs:n})=>{let{x:r}=e,{paddings:a,mode:s}=n,i=Le().getBool("WEBGL_PACK_ARRAY_OPERATIONS")?new eG(r.shape,a,s):new QU(r.shape,a,s);return t.runWebGLProgram(i,[r],r.dtype)}},nG=Wz({opSnippet:"if (b == 0.0) return NAN;\n  return mod(a, b);",packedOpSnippet:"\n  vec4 result = mod(a, b);\n  bvec4 isNaN = equal(b, vec4(0.0));\n  "+Tz+"\n  return result;\n"}),rG={kernelName:Ln,backendName:"webgl",kernelFunc:nG},aG=class{constructor(e,t,n){this.variableNames=["probs"],this.customUniforms=[{name:"seed",type:"float"}],this.outputShape=[e,n],this.userCode=`\n      void main() {\n        ivec2 coords = getOutputCoords();\n        int batch = coords[0];\n\n        float r = random(seed);\n        float cdf = 0.0;\n\n        for (int i = 0; i < ${t-1}; i++) {\n          cdf += getProbs(batch, i);\n\n          if (r < cdf) {\n            setOutput(float(i));\n            return;\n          }\n        }\n\n        // If no other event happened, last event happened.\n        setOutput(float(${t-1}));\n      }\n    `}},sG=Wz({opSnippet:"\nif (a == b) {\n  return 1.0;\n};\nreturn a / b;",packedOpSnippet:"\n  // vec4 one = vec4(equal(a, b));\n  // return one + (vec4(1.0) - one) * a / b;\n  vec4 result = a / b;\n  if(a.x == b.x) {\n    result.x = 1.;\n  }\n  if(a.y == b.y) {\n    result.y = 1.;\n  }\n  if(a.z == b.z) {\n    result.z = 1.;\n  }\n  if(a.w == b.w) {\n    result.w = 1.;\n  }\n\n  return result;\n",checkOutOfBounds:!0}),iG={kernelName:zt,backendName:"webgl",kernelFunc:sG},oG="return a - b;",lG=Wz({opSnippet:oG,packedOpSnippet:oG,supportsComplex:!0,cpuKernelImpl:YP}),uG={kernelName:Ur,backendName:"webgl",kernelFunc:lG};function hG(e){let{inputs:t,backend:n,attrs:r}=e,{logits:a}=t,{dim:s}=r,i=va.parseAxisParam([s],a.shape),o=PU({inputs:{x:a},backend:n,attrs:{reductionIndices:i,keepDims:!1}}),l=Gf.expandShapeToKeepDim(o.shape,i),u=Zz({inputs:{x:o},backend:n,attrs:{shape:l}}),h=lG({inputs:{a,b:u},backend:n}),d=EV({inputs:{x:h},backend:n}),p=aB({inputs:{x:d},backend:n,attrs:{axis:i,keepDims:!1}}),c=Zz({inputs:{x:p},backend:n,attrs:{shape:l}}),f=sG({inputs:{a:d,b:c},backend:n});return n.disposeIntermediateTensorInfo(o),n.disposeIntermediateTensorInfo(u),n.disposeIntermediateTensorInfo(h),n.disposeIntermediateTensorInfo(d),n.disposeIntermediateTensorInfo(p),n.disposeIntermediateTensorInfo(c),f}var dG={kernelName:Ar,backendName:"webgl",kernelFunc:hG};var pG={kernelName:Pn,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{logits:a}=t,{numSamples:s,seed:i,normalized:o}=r,l=o?a:hG({inputs:{logits:a},backend:n,attrs:{dim:a.shape.length-1}}),u=l.shape[0],h=l.shape[1],d=new aG(u,h,s),p=[[i]],c=n.runWebGLProgram(d,[l],"int32",p);return o||n.disposeIntermediateTensorInfo(l),c}},cG=dz+"\n  return -x;\n";var fG={kernelName:Bn,backendName:"webgl",kernelFunc:function(e){let t,{inputs:n,backend:r}=e,{x:a}=n;if(r.shouldExecuteOnCPU([a])){let e=r.texData.get(a.dataId),[t,n]=AP(e.values,a.shape,a.dtype);return r.makeTensorInfo(n,a.dtype,t)}return t=Le().getBool("WEBGL_PACK_UNARY_OPERATIONS")?new gz(a.shape,"\n  vec4 result = -x;\n  bvec4 isNaN = isnan(x);\n\n  result.r = isNaN.r ? x.r : result.r;\n  result.g = isNaN.g ? x.g : result.g;\n  result.b = isNaN.b ? x.b : result.b;\n  result.a = isNaN.a ? x.a : result.a;\n\n  return result;\n"):new hz(a.shape,cG),r.runWebGLProgram(t,[a],a.dtype)}},mG=Km.nonMaxSuppressionV3Impl;var gG={kernelName:Vn,backendName:"webgl",kernelFunc:function(e){Gf.warn("tf.nonMaxSuppression() in webgl locks the UI thread. Call tf.nonMaxSuppressionAsync() instead");let{inputs:t,backend:n,attrs:r}=e,{boxes:a,scores:s}=t,{maxOutputSize:i,iouThreshold:o,scoreThreshold:l}=r,u=n.readSync(a.dataId),h=n.readSync(s.dataId),{selectedIndices:d}=mG(u,h,i,o,l);return n.makeTensorInfo([d.length],"int32",new Int32Array(d))}},yG=Km.nonMaxSuppressionV4Impl;var bG={kernelName:Un,backendName:"webgl",kernelFunc:function(e){Gf.warn("tf.nonMaxSuppression() in webgl locks the UI thread. Call tf.nonMaxSuppressionAsync() instead");let{inputs:t,backend:n,attrs:r}=e,{boxes:a,scores:s}=t,{maxOutputSize:i,iouThreshold:o,scoreThreshold:l,padToMaxOutputSize:u}=r,h=n.readSync(a.dataId),d=n.readSync(s.dataId),{selectedIndices:p,validOutputs:c}=yG(h,d,i,o,l,u);return[n.makeTensorInfo([p.length],"int32",new Int32Array(p)),n.makeTensorInfo([],"int32",new Int32Array([c]))]}},xG=Km.nonMaxSuppressionV5Impl;var vG={kernelName:Gn,backendName:"webgl",kernelFunc:function(e){Gf.warn("tf.nonMaxSuppression() in webgl locks the UI thread. Call tf.nonMaxSuppressionAsync() instead");let{inputs:t,backend:n,attrs:r}=e,{boxes:a,scores:s}=t,{maxOutputSize:i,iouThreshold:o,scoreThreshold:l,softNmsSigma:u}=r,h=n.readSync(a.dataId),d=n.readSync(s.dataId),p=i,c=o,f=l,m=u,{selectedIndices:g,selectedScores:y}=xG(h,d,p,c,f,m);return[n.makeTensorInfo([g.length],"int32",new Int32Array(g)),n.makeTensorInfo([y.length],"float32",new Float32Array(y))]}},wG=class{constructor(e,t,n,r){this.variableNames=["indices"],this.outputShape=[e,t],this.userCode=`\n      void main() {\n        ivec2 coords = getOutputCoords();\n        int index = round(getIndices(coords.x));\n        setOutput(mix(float(${r}), float(${n}),\n                      float(index == coords.y)));\n      }\n    `}},kG={kernelName:jn,backendName:"webgl",kernelFunc:e=>{let{inputs:t,backend:n,attrs:r}=e,{indices:a}=t,{dtype:s,depth:i,onValue:o,offValue:l}=r,u=va.sizeFromShape(a.shape),h=new wG(u,i,o,l),d=Zz({inputs:{x:a},backend:n,attrs:{shape:[u]}}),p=n.runWebGLProgram(h,[d],s);n.disposeIntermediateTensorInfo(d);let c=Zz({inputs:{x:p},backend:n,attrs:{shape:[...a.shape,i]}});return n.disposeIntermediateTensorInfo(p),c}};function IG(e){let{inputs:t,backend:n}=e,{x:r}=t;if("complex64"===r.dtype){let e=hW({inputs:{input:r},backend:n}),t=IG({inputs:{x:e},backend:n}),a=_W({inputs:{input:r},backend:n}),s=IG({inputs:{x:a},backend:n}),i=$z({inputs:{real:t,imag:s},backend:n});return n.disposeIntermediateTensorInfo(e),n.disposeIntermediateTensorInfo(t),n.disposeIntermediateTensorInfo(a),n.disposeIntermediateTensorInfo(s),i}return BV({attrs:{shape:r.shape,dtype:r.dtype,value:"string"===r.dtype?"":0},backend:n})}var SG={kernelName:ea,backendName:"webgl",kernelFunc:IG};var _G={kernelName:Hn,backendName:"webgl",kernelFunc:function e(t){let{inputs:n,backend:r}=t,{x:a}=n;if("string"===a.dtype)throw new Error("onesLike is not supported under string dtype");if("complex64"===a.dtype){let t=hW({inputs:{input:a},backend:r}),n=e({inputs:{x:t},backend:r}),s=_W({inputs:{input:a},backend:r}),i=IG({inputs:{x:s},backend:r}),o=$z({inputs:{real:n,imag:i},backend:r});return r.disposeIntermediateTensorInfo(t),r.disposeIntermediateTensorInfo(n),r.disposeIntermediateTensorInfo(s),r.disposeIntermediateTensorInfo(i),o}return BV({attrs:{shape:a.shape,dtype:a.dtype,value:1},backend:r})}};var NG={kernelName:qn,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{axis:a}=r;if(1===t.length)return $V({inputs:{input:t[0]},backend:n,attrs:{dim:a}});let s=t[0].shape,i=t[0].dtype;t.forEach(e=>{va.assertShapesMatch(s,e.shape,"All tensors passed to stack must have matching shapes"),va.assert(i===e.dtype,()=>"All tensors passed to stack must have matching dtypes")});let o=[],l=CW({inputs:t.map(e=>{let t=$V({inputs:{input:e},backend:n,attrs:{dim:a}});return o.push(t),t}),backend:n,attrs:{axis:a}});return o.forEach(e=>n.disposeIntermediateTensorInfo(e)),l}},TG=class{constructor(e,t,n){this.variableNames=["x"],this.customUniforms=[{name:"value",type:"float"}],this.outputShape=t.map((t,n)=>t[0]+e[n]+t[1]);let r=e.length,a=IL(r),s=t.map(e=>e[0]).join(","),i=t.map((t,n)=>t[0]+e[n]).join(","),o=["coords[0]","coords[1]","coords[2]","coords[3]"].slice(0,r);this.userCode=1!==r?`\n      ${a} start = ${a}(${s});\n      ${a} end = ${a}(${i});\n\n      void main() {\n        ${a} outC = getOutputCoords();\n        if (any(lessThan(outC, start)) || any(greaterThanEqual(outC, end))) {\n          setOutput(value);\n        } else {\n          ${a} coords = outC - start;\n          setOutput(getX(${o}));\n        }\n      }\n    `:`\n        int start = ${s};\n        int end = ${i};\n\n        void main() {\n          int outC = getOutputCoords();\n          if (outC < start || outC >= end) {\n            setOutput(value);\n          } else {\n            setOutput(getX(outC - start));\n          }\n        }\n      `}},CG=class{constructor(e,t,n){this.variableNames=["x"],this.packedInputs=!0,this.packedOutput=!0,this.customUniforms=[{name:"value",type:"float"}],this.outputShape=t.map((t,n)=>t[0]+e[n]+t[1]);let r=e.length,a=IL(r),s=t.map(e=>e[0]).join(","),i=t.map((t,n)=>t[0]+e[n]).join(","),o=rz("rc",r),l=rz("source",r),u=`${o[r-1]} < ${this.outputShape[r-1]}`,h=1===r?"source":`vec2(${l.slice(-2).join()})`,d=[`${a} rc = outputLoc;`,`${o[r-1]} += 1;\n       if(${u}) {\n      `,1===r?"":`}\n       rc = outputLoc;\n       ${o[r-2]} += 1;\n       if(${o[r-2]} < ${this.outputShape[r-2]}) {`,1===r?"":`  ${o[r-1]} += 1;\n         if(${u}) {`],p=1===r?"rc < start || rc >= end":"any(lessThan(rc, start)) || any(greaterThanEqual(rc, end))",c="";for(let e=0,t=1===r?2:4;e<t;e++)c+=`\n        ${d[e]}\n        if (${p}) {\n          result[${e}] = float(value);\n        } else {\n          ${a} source = rc - start;\n          result[${e}] = getChannel(getX(${l.join()}), ${h});\n        }\n      `;c+=1===r?"} ":"}}",this.userCode=`\n      const ${a} start = ${a}(${s});\n      const ${a} end = ${a}(${i});\n\n      void main() {\n        ${a} outputLoc = getOutputCoords();\n        vec4 result = vec4(0.);\n        ${c}\n        setOutput(result);\n      }\n    `}},EG=e=>{let{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{paddings:s,constantValue:i}=r;if(0===va.sizeFromShape(a.shape)){return BV({backend:n,attrs:{shape:s.map((e,t)=>e[0]+a.shape[t]+e[1]),value:i,dtype:a.dtype}})}let o=Le().getBool("WEBGL_PACK_ARRAY_OPERATIONS")?new CG(a.shape,s,i):new TG(a.shape,s,i),l=[[i]];return n.runWebGLProgram(o,[a],a.dtype,l)},AG={kernelName:Kn,backendName:"webgl",kernelFunc:EG},$G=Wz({opSnippet:"\n  if(a < 0.0 && floor(b) < b){\n    return NAN;\n  }\n  if (b == 0.0) {\n    return 1.0;\n  }\n  return (round(mod(b, 2.0)) != 1) ?\n      pow(abs(a), b) : sign(a) * pow(abs(a), b);\n",packedOpSnippet:"\n  // isModRound1 has 1 for components with round(mod(b, 2.0)) == 1, 0 otherwise.\n  vec4 isModRound1 = vec4(equal(round(mod(b, 2.0)), ivec4(1)));\n  vec4 multiplier = sign(a) * isModRound1 + (vec4(1.0) - isModRound1);\n  vec4 result = multiplier * pow(abs(a), b);\n\n  // Ensure that a^0 = 1, including 0^0 = 1 as this correspond to TF and JS\n  bvec4 isExpZero = equal(b, vec4(0.0));\n  result.r = isExpZero.r ? 1.0 : result.r;\n  result.g = isExpZero.g ? 1.0 : result.g;\n  result.b = isExpZero.b ? 1.0 : result.b;\n  result.a = isExpZero.a ? 1.0 : result.a;\n\n  bvec4 isNaN1 = lessThan(a, vec4(0.0));\n  bvec4 isNaN2 = lessThan(floor(b), b);\n  bvec4 isNaN = bvec4(isNaN1.x && isNaN2.x, isNaN1.y && isNaN2.y, isNaN1.z && isNaN2.z, isNaN1.w && isNaN2.w);\n  "+Tz+"\n  return result;\n"}),RG={kernelName:Zn,backendName:"webgl",kernelFunc:$G};var FG={kernelName:Jn,backendName:"webgl",kernelFunc:function(e){let t,{inputs:n,backend:r,attrs:a}=e,{x:s}=n,{axis:i,keepDims:o}=a,l=s.shape.length,u=[],h=va.parseAxisParam(i,s.shape),d=h,p=Gf.getAxesPermutation(d,l),c=s;if(null!=p&&(c=iB({inputs:{x:s},backend:r,attrs:{perm:p}}),d=Gf.getInnerMostAxes(d.length,l),u.push(c)),Gf.assertAxesAreInnerMostDims("prod",d,l),r.shouldExecuteOnCPU([c])){let e=r.texData.get(c.dataId).values,{outVals:n,outShape:a,outDtype:s}=RP(c.shape,c.dtype,e,d);t=r.makeTensorInfo(a,s,n)}else{let[e,n]=Gf.computeOutAndReduceShapes(c.shape,d),a=va.sizeFromShape(n),i=Zz({inputs:{x:c},backend:r,attrs:{shape:[-1,a]}}),o=eB(i,cs(s.dtype),"prod",r);t=Zz({inputs:{x:o},backend:r,attrs:{shape:e}}),u.push(i),u.push(o)}if(o){u.push(t);let e=Gf.expandShapeToKeepDim(t.shape,h);t=Zz({inputs:{x:t},backend:r,attrs:{shape:e}})}return u.forEach(e=>r.disposeIntermediateTensorInfo(e)),t}};var DG={kernelName:Qn,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{paramsNestedSplits:a,paramsDenseValues:s,indices:i}=t,{outputRaggedRank:o}=r,l=a.map(e=>n.readSync(e.dataId)),u=a.map(e=>e.shape),h=n.readSync(s.dataId),d=n.readSync(i.dataId),[p,c,f]=FP(l,u,h,s.shape,s.dtype,d,i.shape,o),m=p.map(e=>n.makeTensorInfo([e.length],"int32",e)),g=n.makeTensorInfo(f,s.dtype,c);return m.concat([g])}};var MG={kernelName:er,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n}=e,{starts:r,limits:a,deltas:s}=t,i=n.readSync(r.dataId),o=n.readSync(a.dataId),l=n.readSync(s.dataId),[u,h]=DP(i,r.shape,r.dtype,o,a.shape,l,s.shape);return[n.makeTensorInfo([u.length],"int32",u),n.makeTensorInfo([h.length],r.dtype,h)]}};var OG={kernelName:tr,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{shape:a,values:s,defaultValue:i,rowPartitionTensors:o}=t,{rowPartitionTypes:l}=r,u=n.readSync(a.dataId),h=n.readSync(s.dataId),d=n.readSync(i.dataId),p=o.map(e=>n.readSync(e.dataId)),c=o.map(e=>e.shape),[f,m]=MP(u,a.shape,h,s.shape,s.dtype,d,i.shape,p,c,l);return n.makeTensorInfo(f,s.dtype,m)}},LG=e=>{let{backend:t,attrs:n}=e,{start:r,stop:a,step:s,dtype:i}=n,o=OP(r,a,s,i);return t.makeTensorInfo([o.length],i,o)},PG={kernelName:nr,backendName:"webgl",kernelFunc:LG},zG=Bz({opSnippet:"return 1.0 / x;"}),BG={kernelName:ar,backendName:"webgl",kernelFunc:zG},WG=Bz({opSnippet:dz+"\n  return (x < 0.0) ? 0.0 : x;\n",packedOpSnippet:"\n  vec4 result = x * vec4(greaterThanEqual(x, vec4(0.0)));\n  bvec4 isNaN = isnan(x);\n\n  result.r = isNaN.r ? x.r : result.r;\n  result.g = isNaN.g ? x.g : result.g;\n  result.b = isNaN.b ? x.b : result.b;\n  result.a = isNaN.a ? x.a : result.a;\n\n  return result;\n"}),VG={kernelName:sr,backendName:"webgl",kernelFunc:WG},UG=Bz({opSnippet:dz+"\n  return (x < 0.0) ? 0.0 : min(6.0, x);\n",packedOpSnippet:"\n  vec4 result = min(x, vec4(6.)) * vec4(greaterThanEqual(x, vec4(0.0)));\n  bvec4 isNaN = isnan(x);\n\n  result.r = isNaN.r ? x.r : result.r;\n  result.g = isNaN.g ? x.g : result.g;\n  result.b = isNaN.b ? x.b : result.b;\n  result.a = isNaN.a ? x.a : result.a;\n\n  return result;\n"}),GG={kernelName:dr,backendName:"webgl",kernelFunc:UG},HG=class{constructor(e,t,n,r,a){this.variableNames=["A"],this.outputShape=[];let[s,i,o,l]=e;this.outputShape=[s,t,n,l];let u,h=[r&&t>1?i-1:i,r&&n>1?o-1:o],d=[r&&t>1?t-1:t,r&&n>1?n-1:n];u=a?"(vec2(yRC) + vec2(0.5)) * effectiveInputOverOutputRatioRC - vec2(0.5)":"vec2(yRC) * effectiveInputOverOutputRatioRC",this.userCode=`\n      const vec2 effectiveInputOverOutputRatioRC = vec2(\n          ${h[0]/d[0]},\n          ${h[1]/d[1]});\n      const vec2 inputShapeRC = vec2(${i}.0, ${o}.0);\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int b = coords[0];\n        int d = coords[3];\n        ivec2 yRC = coords.yz;\n\n        // Fractional source index.\n        vec2 sourceFracIndexRC = ${u};\n\n        // Compute the four integer indices.\n        ivec2 sourceFloorRC = ivec2(max(sourceFracIndexRC, vec2(0.0)));\n        ivec2 sourceCeilRC = ivec2(\n          min(inputShapeRC - 1.0, ceil(sourceFracIndexRC)));\n\n        float topLeft = getA(b, sourceFloorRC.x, sourceFloorRC.y, d);\n        float bottomLeft = getA(b, sourceCeilRC.x, sourceFloorRC.y, d);\n        float topRight = getA(b, sourceFloorRC.x, sourceCeilRC.y, d);\n        float bottomRight = getA(b, sourceCeilRC.x, sourceCeilRC.y, d);\n\n        vec2 fracRC = sourceFracIndexRC - vec2(sourceFloorRC);\n\n        float top = topLeft + (topRight - topLeft) * fracRC.y;\n        float bottom = bottomLeft + (bottomRight - bottomLeft) * fracRC.y;\n        float newValue = top + (bottom - top) * fracRC.x;\n\n        setOutput(newValue);\n      }\n    `}},jG=class{constructor(e,t,n,r,a){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0,this.outputShape=[];let[s,i,o,l]=e;this.outputShape=[s,t,n,l];let u,h=[r&&t>1?i-1:i,r&&n>1?o-1:o],d=[r&&t>1?t-1:t,r&&n>1?n-1:n];u=a?"(vec3(yRC) + vec3(0.5)) * effectiveInputOverOutputRatioRC - vec3(0.5)":"vec3(yRC) * effectiveInputOverOutputRatioRC",this.userCode=`\n      const vec3 effectiveInputOverOutputRatioRC = vec3(\n          ${h[0]/d[0]},\n          ${h[1]/d[1]},\n          ${h[1]/d[1]});\n      const vec3 inputShapeRC = vec3(${i}.0, ${o}.0,\n                                     ${o}.0);\n\n      float getAValue(int b, int r, int c, int d) {\n        return getChannel(getA(b, r, c, d), vec2(c, d));\n      }\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int b = coords[0];\n        int d = coords[3];\n        // Calculate values for next column in yRC.z.\n        ivec3 yRC = coords.yzz + ivec3(0, 0, 1);\n\n        // Fractional source index.\n        vec3 sourceFracIndexRC = ${u};\n\n        // Compute the four integer indices.\n        ivec3 sourceFloorRC = ivec3(max(sourceFracIndexRC, vec3(0.0)));\n        ivec3 sourceCeilRC = ivec3(\n          min(inputShapeRC - 1.0, ceil(sourceFracIndexRC)));\n\n        // Should we calculate next column and row elements in 2x2 packed cell.\n        bool hasNextCol = d < ${l-1};\n        bool hasNextRow = coords.z < ${n-1};\n\n        // In parallel, construct four corners for all four components in\n        // packed 2x2 cell.\n        vec4 topLeft = vec4(\n          getAValue(b, sourceFloorRC.x, sourceFloorRC.y, d),\n          hasNextCol ? getAValue(b, sourceFloorRC.x, sourceFloorRC.y, d + 1)\n                     : 0.0,\n          hasNextRow ? getAValue(b, sourceFloorRC.x, sourceFloorRC.z, d)\n                     : 0.0,\n          (hasNextRow && hasNextCol) ?\n            getAValue(b, sourceFloorRC.x, sourceFloorRC.z, d + 1) : 0.0);\n\n        vec4 bottomLeft = vec4(\n          getAValue(b, sourceCeilRC.x, sourceFloorRC.y, d),\n          hasNextCol ? getAValue(b, sourceCeilRC.x, sourceFloorRC.y, d + 1)\n                     : 0.0,\n          hasNextRow ? getAValue(b, sourceCeilRC.x, sourceFloorRC.z, d)\n                     : 0.0,\n          (hasNextRow && hasNextCol) ?\n            getAValue(b, sourceCeilRC.x, sourceFloorRC.z, d + 1) : 0.0);\n\n        vec4 topRight = vec4(\n          getAValue(b, sourceFloorRC.x, sourceCeilRC.y, d),\n          hasNextCol ? getAValue(b, sourceFloorRC.x, sourceCeilRC.y, d + 1)\n                     : 0.0,\n          hasNextRow ? getAValue(b, sourceFloorRC.x, sourceCeilRC.z, d)\n                     : 0.0,\n          (hasNextRow && hasNextCol) ?\n            getAValue(b, sourceFloorRC.x, sourceCeilRC.z, d + 1) : 0.0);\n\n        vec4 bottomRight = vec4(\n          getAValue(b, sourceCeilRC.x, sourceCeilRC.y, d),\n          hasNextCol ? getAValue(b, sourceCeilRC.x, sourceCeilRC.y, d + 1)\n                     : 0.0,\n          hasNextRow ? getAValue(b, sourceCeilRC.x, sourceCeilRC.z, d)\n                     : 0.0,\n          (hasNextRow && hasNextCol) ?\n            getAValue(b, sourceCeilRC.x, sourceCeilRC.z, d + 1) : 0.0);\n\n        vec3 fracRC = sourceFracIndexRC - vec3(sourceFloorRC);\n\n        vec4 top = mix(topLeft, topRight, fracRC.yyzz);\n        vec4 bottom = mix(bottomLeft, bottomRight, fracRC.yyzz);\n        vec4 newValue = mix(top, bottom, fracRC.x);\n\n        setOutput(newValue);\n      }\n    `}};var qG={kernelName:ur,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{images:a}=t,{alignCorners:s,halfPixelCenters:i,size:o}=r,[l,u]=o,h=Le().getBool("WEBGL_PACK_IMAGE_OPERATIONS")?new jG(a.shape,l,u,s,i):new HG(a.shape,l,u,s,i);return n.runWebGLProgram(h,[a],"float32")}},KG=class{constructor(e,t,n){this.variableNames=["dy"],this.outputShape=[],this.outputShape=t;let[,r,a]=t,[,s,i]=e,o=[n&&s>1?r-1:r,n&&i>1?a-1:a],l=[n&&s>1?s-1:s,n&&i>1?i-1:i],u=o[0]/l[0],h=o[1]/l[1],d=1/u,p=1/h,c=2*Math.ceil(d)+2,f=2*Math.ceil(p)+2;this.userCode=`\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int b = coords[0];\n        int d = coords[3];\n        int r = coords[1];\n        int c = coords[2];\n\n        float accumulator = 0.0;\n\n        const float heightScale = float(${u});\n        const float widthScale = float(${h});\n\n        const float invHeightScale = float(${d});\n        const float invWidthScale = float(${p});\n\n        const int winHeight = int(${c});\n        const int winWidth = int(${f});\n\n        // Compute bounds for where in dy we will look\n        float startRLerp = floor(float(r) * invHeightScale);\n        int startDyR = int(startRLerp - float(winHeight / 2));\n\n        float startCLerp = floor(float(c) * invWidthScale);\n        int startDyC = int(startCLerp - float(winWidth / 2));\n\n        // Loop over dy\n        for (int dyROffset = 0; dyROffset < winHeight; dyROffset++) {\n          int dyR = dyROffset + startDyR;\n\n          // Guard against the window exceeding the bounds of dy\n          if (dyR < 0 || dyR >= ${s}) {\n            continue;\n          }\n\n          for (int dyCOffset = 0; dyCOffset < winWidth; dyCOffset++) {\n            int dyC = dyCOffset + startDyC;\n\n            // Guard against the window exceeding the bounds of dy\n            if (dyC < 0 || dyC >= ${i}) {\n              continue;\n            }\n\n            float dxR = float(dyR) * heightScale;\n            int topDxRIndex = int(floor(dxR));\n            int bottomDxRIndex = int(min(ceil(dxR), ${r-1}.0));\n            float dxRLerp = dxR - float(topDxRIndex);\n            float inverseDxRLerp = 1.0 - dxRLerp;\n\n            float dxC = float(dyC) * widthScale;\n            int leftDxCIndex = int(floor(dxC));\n            int rightDxCIndex = int(min(ceil(dxC), ${a-1}.0));\n            float dxCLerp = dxC - float(leftDxCIndex);\n            float inverseDxCLerp = 1.0 - dxCLerp;\n\n            if (r == topDxRIndex && c == leftDxCIndex) {\n              // topLeft\n              accumulator +=\n                getDy(b, dyR, dyC, d) * inverseDxRLerp * inverseDxCLerp;\n            }\n\n            if (r == topDxRIndex && c == rightDxCIndex) {\n              // topRight\n              accumulator += getDy(b, dyR, dyC, d) * inverseDxRLerp * dxCLerp;\n            }\n\n            if (r == bottomDxRIndex && c == leftDxCIndex) {\n              // bottomLeft\n              accumulator += getDy(b, dyR, dyC, d) * dxRLerp * inverseDxCLerp;\n            }\n\n            if (r == bottomDxRIndex && c == rightDxCIndex) {\n              // bottomRight\n              accumulator += getDy(b, dyR, dyC, d) * dxRLerp * dxCLerp;\n            }\n          }\n        }\n        // End loop over dy\n\n        setOutput(accumulator);\n      }\n    `}};var XG={kernelName:hr,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{images:a,dy:s}=t,{alignCorners:i}=r,o=new KG(s.shape,a.shape,i);return n.runWebGLProgram(o,[s],s.dtype)}},ZG=class{constructor(e,t,n,r,a){this.variableNames=["A"],this.outputShape=[];let[s,i,o,l]=e;this.outputShape=[s,t,n,l];let u,h=[r&&t>1?i-1:i,r&&n>1?o-1:o],d=[r&&t>1?t-1:t,r&&n>1?n-1:n],p=r?"0.5":"0.0";u=a?"max((vec2(yRC) + vec2(0.5)) * effectiveInputOverOutputRatioRC, vec2(0.0))":"vec2(yRC) * effectiveInputOverOutputRatioRC",this.userCode=`\n      const vec2 effectiveInputOverOutputRatioRC = vec2(\n          ${h[0]/d[0]},\n          ${h[1]/d[1]});\n      const vec2 inputShapeRC = vec2(${i}.0, ${o}.0);\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int b = coords[0];\n        int d = coords[3];\n        ivec2 yRC = coords.yz;\n\n        // Fractional source index.\n        vec2 sourceFracIndexRC = ${u};\n\n        // Compute the coordinators of nearest neighbor point.\n        ivec2 sourceNearestRC = ivec2(\n          min(inputShapeRC - 1.0, floor(sourceFracIndexRC + ${p})));\n        float newValue = getA(b, sourceNearestRC.x, sourceNearestRC.y, d);\n\n        setOutput(newValue);\n      }\n    `}},YG=class{constructor(e,t,n,r,a){this.variableNames=["A"],this.packedInputs=!0,this.packedOutput=!0,this.outputShape=[];let[s,i,o,l]=e;this.outputShape=[s,t,n,l];let u,h=[r&&t>1?i-1:i,r&&n>1?o-1:o],d=[r&&t>1?t-1:t,r&&n>1?n-1:n],p=r?"0.5":"0.0";u=a?"max((vec3(yRC) + vec3(0.5)) * effectiveInputOverOutputRatioRC, vec3(0.0))":"vec3(yRC) * effectiveInputOverOutputRatioRC",this.userCode=`\n      const vec3 effectiveInputOverOutputRatioRC = vec3(\n          ${h[0]/d[0]},\n          ${h[1]/d[1]},\n          ${h[1]/d[1]});\n      const vec3 inputShapeRC = vec3(${i}.0, ${o}.0,\n                                     ${o}.0);\n\n      float getAValue(int b, int r, int c, int d) {\n        return getChannel(getA(b, r, c, d), vec2(c, d));\n      }\n\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int b = coords[0];\n        int d = coords[3];\n        // Calculate values for next column in yRC.z.\n        ivec3 yRC = coords.yzz + ivec3(0, 0, 1);\n\n        // Fractional source index.\n        vec3 sourceFracIndexRC = ${u};\n\n        // Compute the coordinators of nearest neighbor point.\n        ivec3 sourceNearestRC = ivec3(\n          min(inputShapeRC - 1.0, floor(sourceFracIndexRC + ${p})));\n\n        // Should we calculate next column and row elements in 2x2 packed cell.\n        bool hasNextCol = d < ${l-1};\n        bool hasNextRow = coords.z < ${n-1};\n\n        vec4 newValue = vec4(\n          getAValue(b, sourceNearestRC.x, sourceNearestRC.y, d),\n          hasNextCol ? getAValue(b, sourceNearestRC.x, sourceNearestRC.y, d + 1)\n                     : 0.0,\n          hasNextRow ? getAValue(b, sourceNearestRC.x, sourceNearestRC.z, d)\n                     : 0.0,\n          (hasNextRow && hasNextCol) ?\n            getAValue(b, sourceNearestRC.x, sourceNearestRC.z, d + 1) : 0.0);\n\n        setOutput(newValue);\n      }\n    `}};var JG={kernelName:or,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{images:a}=t,{alignCorners:s,halfPixelCenters:i,size:o}=r,[l,u]=o,h=Le().getBool("WEBGL_PACK_IMAGE_OPERATIONS")?new YG(a.shape,l,u,s,i):new ZG(a.shape,l,u,s,i);return n.runWebGLProgram(h,[a],a.dtype)}},QG=class{constructor(e,t,n){this.variableNames=["dy"],this.outputShape=[],this.outputShape=t;let[,r,a]=t,[,s,i]=e,o=[n&&s>1?r-1:r,n&&i>1?a-1:a],l=[n&&s>1?s-1:s,n&&i>1?i-1:i],u=o[0]/l[0],h=o[1]/l[1],d=1/u,p=1/h,c=2*Math.ceil(d)+2,f=2*Math.ceil(p)+2;this.userCode=`\n      void main() {\n        ivec4 coords = getOutputCoords();\n        int b = coords[0];\n        int d = coords[3];\n        int r = coords[1];\n        int c = coords[2];\n\n        float accumulator = 0.0;\n\n        const float heightScale = float(${u});\n        const float widthScale = float(${h});\n\n        const float invHeightScale = float(${d});\n        const float invWidthScale = float(${p});\n\n        const int winHeight = int(${c});\n        const int winWidth = int(${f});\n\n        // Compute bounds for where in dy we will look\n        float startRLerp = floor(float(r) * invHeightScale);\n        int startDyR = int(floor(startRLerp - float(winHeight / 2)));\n\n        float startCLerp = floor(float(c) * invWidthScale);\n        int startDyC = int(floor(startCLerp - float(winWidth / 2)));\n\n        // Loop over dy\n        for (int dyROffset = 0; dyROffset < winHeight; dyROffset++) {\n          int dyR = dyROffset + startDyR;\n\n          // Guard against the window exceeding the bounds of dy\n          if (dyR < 0 || dyR >= ${s}) {\n            continue;\n          }\n\n          for (int dyCOffset = 0; dyCOffset < winWidth; dyCOffset++) {\n            int dyC = dyCOffset + startDyC;\n\n            // Guard against the window exceeding the bounds of dy\n            if (dyC < 0 || dyC >= ${i}) {\n              continue;\n            }\n\n            float sourceFracRow =\n              float(${o[0]}) *\n                (float(dyR) / float(${l[0]}));\n\n            float sourceFracCol =\n                float(${o[1]}) *\n                  (float(dyC) / float(${l[1]}));\n\n            int sourceNearestRow = int(min(\n                float(int(${r}) - 1),\n                ${n} ? float(round(sourceFracRow)) :\n                                  float(floor(sourceFracRow))));\n\n            int sourceNearestCol = int(min(\n                float(int(${a}) - 1),\n                ${n} ? float(round(sourceFracCol)) :\n                                  float(floor(sourceFracCol))));\n\n            if (r == sourceNearestRow && c == sourceNearestCol) {\n              accumulator += getDy(b, dyR, dyC, d);\n            }\n          }\n        }\n        // End loop over dy\n\n        setOutput(accumulator);\n      }\n    `}};var eH={kernelName:lr,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{images:a,dy:s}=t,{alignCorners:i}=r,o=new QG(s.shape,a.shape,i);return n.runWebGLProgram(o,[s],s.dtype)}},tH=class{constructor(e,t){this.variableNames=["x"];let n=e.length;if(n>4)throw new Error(`WebGL backend: Reverse of rank-${n} tensor is not yet supported`);if(this.outputShape=e,1===n)return void(this.userCode=`\n        void main() {\n          int coord = getOutputCoords();\n          setOutput(getX(${e[0]} - coord - 1));\n        }\n      `);let r=e.map((n,r)=>(n=>-1!==t.indexOf(n)&&1!==e[n]?`${e[n]} - coords[${n}] - 1`:`coords[${n}]`)(r)).join(","),a=IL(n);this.userCode=`\n      void main() {\n        ${a} coords = getOutputCoords();\n        setOutput(getX(${r}));\n      }\n    `}},nH=class{constructor(e,t){this.variableNames=["x"],this.packedInputs=!0,this.packedOutput=!0;let n=e.length;if(n>4)throw new Error(`WebGL backend: Reverse of rank-${n} tensor is not yet supported`);this.outputShape=e;let r=rz("rc",n),a=`${r[n-1]} + 1 < ${this.outputShape[n-1]}`,s=`${r[n-2]} + 1 < ${this.outputShape[n-2]}`,i=IL(n);var o;function l(n){let r=e.map((r,a)=>function(n,r){return-1!==t.indexOf(n)&&1!==e[n]?`${e[n]} - ${r[n]} - 1`:`${r[n]}`}(a,n));return`getChannel(getX(${r.join(",")}), vec2(${r.slice(-2).join(",")}))`}this.userCode=1===n?`\n        void main(){\n          int rc = getOutputCoords();\n          vec4 result = vec4(0.);\n          result.r = getChannel(getX(${e[0]} - rc - 1),\n            ${e[0]} - rc - 1);\n          if(${a}){\n              result.g = getChannel(getX(${e[0]} - (rc  + 1) - 1),\n                ${e[0]} - (rc  + 1) - 1);\n          }\n          setOutput(result);\n        }\n      `:`\n        void main() {\n          ${i} rc = getOutputCoords();\n          vec4 result = vec4(0.);\n          result.r = ${o=r.slice(),l(o)};\n          if(${a}){\n            result.g = ${function(e){return e[n-1]="("+e[n-1]+" + 1)",l(e)}(r.slice())};\n          }\n          if(${s}) {\n            result.b = ${function(e){return e[n-2]="("+e[n-2]+" + 1)",l(e)}(r.slice())};\n            if(${a}) {\n              result.a = ${function(e){return e[n-1]="("+e[n-1]+" + 1)",e[n-2]="("+e[n-2]+" + 1)",l(e)}(r.slice())};\n            }\n          }\n          setOutput(result);\n        }\n    `}};var rH={kernelName:pr,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{dims:s}=r,i=a.shape.length,o=va.parseAxisParam(s,a.shape);if(0===i)return Ez({inputs:{x:a},backend:n});let l=Le().getBool("WEBGL_PACK_ARRAY_OPERATIONS")?new nH(a.shape,o):new tH(a.shape,o);return n.runWebGLProgram(l,[a],a.dtype)}},aH=class{constructor(e,t){this.variableNames=["Image"],this.outputShape=[],this.customUniforms=[{name:"params",type:"vec4"}];let n=e[1],r=e[2];this.outputShape=e;let a="";a="number"==typeof t?`float outputValue = ${t.toFixed(2)};`:`\n        vec3 fill = vec3(${t.join(",")});\n        float outputValue = fill[coords[3]];`,this.userCode=`\n        void main() {\n          ivec4 coords = getOutputCoords();\n          int x = coords[2];\n          int y = coords[1];\n          float coordXFloat = (float(x) - params[0]) * params[3] -\n            (float(y) - params[1]) * params[2];\n          float coordYFloat = (float(x) - params[0]) * params[2] +\n            (float(y) - params[1]) * params[3];\n          int coordX = int(round(coordXFloat + params[0]));\n          int coordY = int(round(coordYFloat + params[1]));\n          ${a}\n          if(coordX >= 0 && coordX < ${r} && coordY >= 0 && coordY < ${n}) {\n            outputValue = getImage(coords[0], coordY, coordX, coords[3]);\n          }\n          setOutput(outputValue);\n        }\n    `}},sH={kernelName:ra,backendName:"webgl",kernelFunc:({inputs:e,attrs:t,backend:n})=>{let{image:r}=e,{radians:a,fillValue:s,center:i}=t,o=n,l=new aH(r.shape,s),[u,h]=Gf.getImageCenter(i,r.shape[1],r.shape[2]),d=[[u,h,Math.sin(a),Math.cos(a)]];return o.runWebGLProgram(l,[r],r.dtype,d)}},iH=Bz({opSnippet:"\n  // OpenGL ES does not support round function.\n  // The algorithm is based on banker's rounding.\n  float base = floor(x);\n  if ((x - base) < 0.5) {\n    return floor(x);\n  } else if ((x - base) > 0.5) {\n    return ceil(x);\n  } else {\n    if (mod(base, 2.0) == 0.0) {\n      return base;\n    } else {\n      return base + 1.0;\n    }\n  }\n"}),oH={kernelName:cr,backendName:"webgl",kernelFunc:iH},lH=Bz({opSnippet:"return inversesqrt(x);",cpuKernelImpl:LP}),uH={kernelName:fr,backendName:"webgl",kernelFunc:lH},hH=class{constructor(e,t,n,r,a,s,i=!0,o=!1){this.variableNames=["updates","indices","defaultValue"],this.outputShape=s;let l=IL(a.length),u=IL(s.length),h="";1===n?h="i":2===n&&(h="i, j");let d=`getIndices(${h})`,p="";1===r?p="i":2===r&&(p="i, coords[1]");let c=`getUpdates(${p})`,f="";o&&(f="coords[0], coords[1]");let m=`getDefaultValue(${f})`,g=t>1?"strides[j]":"strides";this.userCode=`\n        ${l} strides = ${l}(${a});\n\n        void main() {\n          ${u} coords = getOutputCoords();\n          float sum = 0.0;\n          bool found = false;\n          for (int i = 0; i < ${e}; i++) {\n            int flattenedIndex = 0;\n            for (int j = 0; j < ${t}; j++) {\n              int index = round(${d});\n              flattenedIndex += index * ${g};\n            }\n            if (flattenedIndex == coords[0]) {\n              sum += ${c};\n              found = true;\n            }\n          }\n          setOutput(mix(${m}, sum, float(found)));\n        }\n      `}},dH=class{constructor(e,t,n,r,a,s,i=!0,o=!1){this.variableNames=["updates","indices","defaultValue"],this.packedInputs=!0,this.packedOutput=!0,this.outputShape=s;let l=IL(a.length),u=IL(s.length),h="";1===n?h="i":2===n&&(h="i, j");let d=`getIndices(${h})`,p="";1===r?p="i":2===r&&(p="i, coords[1]");let c=`getUpdates(${p})`,f="";o&&(f="coords[0], coords[1]");let m=`getDefaultValue(${f})`,g=t>1?"strides[j]":"strides",y=t>1?"strides[j + 1]":"strides";this.userCode=`\n        ${l} strides = ${l}(${a});\n\n        void main() {\n          ${u} coords = getOutputCoords();\n          vec4 sum = vec4(0.);\n          vec4 found = vec4(0.);\n          for (int i = 0; i < ${e}; i+=2) {\n            ivec2 flattenedIndex = ivec2(0);\n            for (int j = 0; j < ${t}; j+=2) {\n              ivec4 index = round(${d});\n              flattenedIndex += index.xz * ${g};\n              if (j + 1 < ${t}) {\n                flattenedIndex += index.yw * ${y};\n              }\n            }\n            if (flattenedIndex[0] == coords[0] || flattenedIndex[1] == coords[0] ||\n                flattenedIndex[0] == coords[0] + 1 || flattenedIndex[1] == coords[0] + 1) {\n              vec4 updVals = ${c};\n              if (flattenedIndex[0] == coords[0]) {\n                sum.xy += updVals.xy;\n                found.xy = vec2(1.);\n              } else if (flattenedIndex[0] == coords[0] + 1) {\n                sum.zw += updVals.xy;\n                found.zw = vec2(1.);\n              }\n              if (flattenedIndex[1] == coords[0]) {\n                sum.xy += updVals.zw;\n                found.xy = vec2(1.);\n              } else if (flattenedIndex[1] == coords[0] + 1) {\n                sum.zw += updVals.zw;\n                found.zw = vec2(1.);\n              }\n            }\n          }\n          setOutput(mix(${m}, sum, found));\n        }\n      `}};var pH={kernelName:mr,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{indices:a,updates:s}=t,{shape:i}=r,{sliceRank:o,numUpdates:l,sliceSize:u,strides:h,outputSize:d}=Gf.calculateShapes(s,a,i),p=[d/u,u];if(0===d)return n.makeTensorInfo(i,a.dtype);let c,f=Zz({inputs:{x:a},backend:n,attrs:{shape:[l,o]}}),m=Zz({inputs:{x:s},backend:n,attrs:{shape:[l,u]}}),g=n.makeTensorInfo([],"float32",new Float32Array([0]));c=Le().getBool("WEBGL_PACK")?new dH(l,o,f.shape.length,m.shape.length,h,p):new hH(l,o,f.shape.length,m.shape.length,h,p);let y=n.runWebGLProgram(c,[m,f,g],m.dtype),b=Zz({inputs:{x:y},backend:n,attrs:{shape:i}});return n.disposeIntermediateTensorInfo(f),n.disposeIntermediateTensorInfo(m),n.disposeIntermediateTensorInfo(y),n.disposeIntermediateTensorInfo(g),b}},cH=class{constructor(e,t,n,r){this.variableNames=["sortedSequence","values"],this.customUniforms=[{name:"numInputs",type:"int"}],this.outputShape=[e,n];let a=`for (int i = 0; i < ${Math.ceil(Math.log2(t+1))}; ++i) { if (left >= right) break;`,s=2===Le().getNumber("WEBGL_VERSION")?"while (left < right) {":a,i="left"===r?"<":"<=";this.userCode=`\n       int findBound(int batch, float value) {\n         int left = 0;\n         int right = numInputs;\n         int mid;\n         ${s}\n           mid = (left + right) / 2;\n           if (getSortedSequence(batch, mid) ${i} value) {\n             left = mid + 1;\n           } else {\n             right = mid;\n           }\n         }\n         return right;\n       }\n\n       void main() {\n         ivec2 coords = getOutputCoords();\n         int batch = coords[0];\n         int valueIndex = coords[1];\n\n         float value = getValues(batch, valueIndex);\n\n         setOutput(float(findBound(batch, value)));\n       }\n     `}};var fH={kernelName:yr,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{sortedSequence:a,values:s}=t,{side:i}=r,o=new cH(a.shape[0],a.shape[1],s.shape[1],i),l=[[a.shape[1]]];return n.runWebGLProgram(o,[a,s],"int32",l)}},mH=class{constructor(e,t,n){let r,a;if(this.variableNames=["c","a","b"],this.outputShape=t,n>4)throw Error(`Where for rank ${n} is not yet supported`);if(1===n)a="resRC",r="resRC";else{let n=["resRC.x","resRC.y","resRC.z","resRC.w"],s=[],i=[];for(let r=0;r<t.length;r++)i.push(`${n[r]}`),r<e&&s.push(`${n[r]}`);r=s.join(),a=i.join()}let s=IL(n);this.userCode=`\n      void main() {\n        ${s} resRC = getOutputCoords();\n        float cVal = getC(${r});\n        if (cVal >= 1.0) {\n          setOutput(getA(${a}));\n        } else {\n          setOutput(getB(${a}));\n        }\n      }\n    `}};var gH={kernelName:br,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n}=e,{condition:r,t:a,e:s}=t,i=new mH(r.shape.length,a.shape,a.shape.length);return n.runWebGLProgram(i,[r,a,s],ps(a.dtype,s.dtype))}},yH=Bz({opSnippet:`\n  // Stable and Attracting Fixed Point (0, 1) for Normalized Weights.\n  // see: https://arxiv.org/abs/1706.02515\n  float scaleAlpha = ${Gf.SELU_SCALEALPHA};\n  float scale = ${Gf.SELU_SCALE};\n  return (x >= 0.0) ? scale * x : scaleAlpha * (exp(x) - 1.0);\n`}),bH={kernelName:xr,backendName:"webgl",kernelFunc:yH},xH=Bz({opSnippet:zz+"\n  return 1.0 / (1.0 + exp(-1.0 * x));\n",packedOpSnippet:"\n  vec4 result = 1.0 / (1.0 + exp(-1.0 * x));\n  bvec4 isNaN = isnan(x);\n\n  result.r = isNaN.r ? x.r : result.r;\n  result.g = isNaN.g ? x.g : result.g;\n  result.b = isNaN.b ? x.b : result.b;\n  result.a = isNaN.a ? x.a : result.a;\n\n  return result;\n",cpuKernelImpl:zP}),vH={kernelName:Sr,backendName:"webgl",kernelFunc:xH},wH=Bz({opSnippet:"\n  if (isnan(x)) { return 0.0; }\n  return sign(x);\n"}),kH={kernelName:Ir,backendName:"webgl",kernelFunc:wH},IH=Bz({opSnippet:zz+"\n  return sin(x);\n",packedOpSnippet:`\n  vec4 result = sin(x);\n  bvec4 isNaN = isnan(x);\n  ${Tz}\n  return result;\n`}),SH={kernelName:wr,backendName:"webgl",kernelFunc:IH},_H=Bz({opSnippet:"\n  float e2x = exp(x);\n  return (e2x - 1.0 / e2x) / 2.0;\n"}),NH={kernelName:kr,backendName:"webgl",kernelFunc:_H},TH=Bz({opSnippet:"\n  float epsilon = 1.1920928955078125e-7;\n  float threshold = log(epsilon) + 2.0;\n\n  bool too_large = x > -threshold;\n  bool too_small = x < threshold;\n\n  float result;\n  float exp_x = exp(x);\n\n  if (too_large){\n    result = x;\n  }\n  else if (too_small){\n    result = exp_x;\n  }\n  else{\n    result = log(exp_x + 1.0);\n  }\n  return result;\n"}),CH={kernelName:_r,backendName:"webgl",kernelFunc:TH},EH={kernelName:Cr,backendName:"webgl",kernelFunc:e=>{let{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{blockShape:s,paddings:i}=r;va.assert(a.shape.length<=4,()=>"spaceToBatchND for rank > 4 with a WebGL backend not implemented yet");let o=s.reduce((e,t)=>e*t),l=[[0,0]];l.push(...i);for(let e=1+s.length;e<a.shape.length;++e)l.push([0,0]);let u=[],h=EG({inputs:{x:a},backend:n,attrs:{paddings:l,constantValue:0}}),d=Gf.getReshaped(h.shape,s,o,!1),p=Gf.getPermuted(d.length,s.length,!1),c=Gf.getReshapedPermuted(h.shape,s,o,!1),f=Zz({inputs:{x:h},backend:n,attrs:{shape:d}}),m=iB({inputs:{x:f},backend:n,attrs:{perm:p}}),g=Zz({inputs:{x:m},backend:n,attrs:{shape:c}});return u.push(h),u.push(f),u.push(m),u.forEach(e=>n.disposeIntermediateTensorInfo(e)),g}};var AH={kernelName:$r,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n}=e,{indices:r,values:a,denseShape:s,defaultValue:i}=t;if(1!==s.shape.length)throw new Error(`Dense shape must be a vector, saw:\n         ${s.shape}`);if(2!==r.shape.length)throw new Error(`Indices must be a matrix, saw:\n         ${r.shape}`);if(1!==a.shape.length)throw new Error(`Values must be a vector, saw:\n         ${a.shape}`);if(0!==i.shape.length)throw new Error(`Default value must be a scalar, saw:\n        ${i.shape}`);let o=n.readSync(r.dataId),l=n.readSync(a.dataId),u=n.readSync(s.dataId),h=n.readSync(i.dataId)[0],[d,p,c,f,m]=VP(o,r.shape,r.dtype,l,a.dtype,u,h);return[n.makeTensorInfo(p,r.dtype,d),n.makeTensorInfo([p[0]],a.dtype,c),n.makeTensorInfo([f.length],"bool",new Uint8Array(f.map(e=>Number(e)))),n.makeTensorInfo([m.length],r.dtype,new Int32Array(m))]}};var $H={kernelName:Rr,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n}=e,{inputIndices:r,inputShape:a,newShape:s}=t;if(2!==r.shape.length)throw new Error(`Input indices should be a matrix but received shape ${r.shape}`);if(1!==a.shape.length)throw new Error(`Input shape should be a vector but received shape ${a.shape}`);if(1!==s.shape.length)throw new Error(`Target shape should be a vector but received shape ${s.shape}`);let i=Array.from(n.readSync(a.dataId)),o=n.readSync(r.dataId),l=Array.from(n.readSync(s.dataId)),[u,h,d]=UP(o,r.shape,r.dtype,i,l);return[n.makeTensorInfo(h,r.dtype,u),n.makeTensorInfo([d.length],s.dtype,new Int32Array(d))]}};var RH={kernelName:Fr,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n}=e,{data:r,indices:a,segmentIds:s}=t;if(r.shape.length<1)throw new Error("Data should be at least 1 dimensional but received scalar");if(1!==a.shape.length)throw new Error(`Indices should be a vector but received shape\n              ${a.shape}`);if(1!==s.shape.length)throw new Error(`Segment ids should be a vector but received shape\n              ${s.shape}`);let i=n.readSync(r.dataId),o=n.readSync(a.dataId),l=n.readSync(s.dataId),[u,h]=GP(i,r.shape,r.dtype,o,l,!0);return n.makeTensorInfo(h,r.dtype,u)}};var FH={kernelName:Dr,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n}=e,{data:r,indices:a,segmentIds:s}=t;if(r.shape.length<1)throw new Error("Data should be at least 1 dimensional but received scalar");if(1!==a.shape.length)throw new Error(`Indices should be a vector but received shape\n             ${a.shape}`);if(1!==s.shape.length)throw new Error(`Segment ids should be a vector but received shape\n             ${s.shape}`);let i=n.readSync(r.dataId),o=n.readSync(a.dataId),l=n.readSync(s.dataId),[u,h]=GP(i,r.shape,r.dtype,o,l);return n.makeTensorInfo(h,r.dtype,u)}};var DH={kernelName:Mr,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{sparseIndices:a,sparseValues:s,defaultValue:i}=t,{outputShape:o}=r,{sliceRank:l,numUpdates:u,sliceSize:h,strides:d,outputSize:p}=Gf.calculateShapes(s,a,o),c=!1;if("string"===s.dtype){let e=n.bufferSync(a),t=n.bufferSync(s),r=va.decodeString(n.readSync(i.dataId)[0]),f=PP(e,t,o,p,h,u,l,d,r,c);return n.makeTensorInfo(o,f.dtype,f.values)}let f=new hH(u,l,a.shape.length,s.shape.length,d,[p,1],c),m=n.runWebGLProgram(f,[s,a,i],s.dtype),g=Zz({inputs:{x:m},backend:n,attrs:{shape:o}});return n.disposeIntermediateTensorInfo(m),g}};var MH={kernelName:Er,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{numOrSizeSplits:s,axis:i}=r,o=va.parseAxisParam(i,a.shape)[0],l=Gf.prepareSplitSize(a,s,o),u=a.shape.length,h=new Array(u).fill(0),d=a.shape.slice();return l.map(e=>{let t=[...d];t[o]=e;let r=nW({inputs:{x:a},backend:n,attrs:{begin:h,size:t}});return h[o]+=e,r})}},OH="return sqrt(x);",LH=Bz({opSnippet:OH,packedOpSnippet:OH,cpuKernelImpl:HP}),PH={kernelName:Nr,backendName:"webgl",kernelFunc:LH},zH=Bz({opSnippet:"return x * x;"}),BH={kernelName:Lr,backendName:"webgl",kernelFunc:zH},WH="return (a - b) * (a - b);",VH=Wz({opSnippet:WH,packedOpSnippet:WH}),UH={kernelName:Or,backendName:"webgl",kernelFunc:VH};var GH={kernelName:Pr,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t;if("string"!==a.dtype)throw new Error("Input must be of datatype string");let s=n.readSync(a.dataId),i=Gf.fromUint8ToStringArray(s),o=jP(i,"string",r);return n.makeTensorInfo(a.shape,"string",o)}};var HH={kernelName:ta,backendName:"webgl",kernelFunc:function({inputs:e,attrs:t,backend:n}){let{x:r}=e,a=dz+`\n    return x > 0.0 ? 1.0 : float(${t.alpha});\n  `,s=new hz(r.shape,a);return n.runWebGLProgram(s,[r],r.dtype)}},jH=class{constructor(e,t,n){this.variableNames=["x"],this.outputShape=n;let r=n.length,a=IL(n.length),s=IL(n.length),i="";if(1===r)i="coords * strides + begin";else{let e=0;i=n.map((t,r)=>(e++,1===n.length?`coords * strides[${r}] + begin[${r}]`:`coords[${e-1}] * strides[${r}] + begin[${r}]`)).join(",")}this.userCode=`\n      ${a} begin = ${a}(${e});\n      ${a} strides = ${a}(${t});\n\n      void main() {\n        ${s} coords = getOutputCoords();\n        setOutput(getX(${i}));\n      }\n    `}};var qH={kernelName:zr,backendName:"webgl",kernelFunc:function(e){let t,{inputs:n,backend:r,attrs:a}=e,{x:s}=n,{begin:i,end:o,strides:l,beginMask:u,endMask:h,ellipsisMask:d,newAxisMask:p,shrinkAxisMask:c}=a,{finalShapeSparse:f,finalShape:m,isIdentity:g,sliceDim0:y,isSimpleSlice:b,begin:x,end:v,strides:w}=vf.sliceInfo(s.shape,i,o,l,u,h,d,p,c);if(g)t=Zz({inputs:{x:s},backend:r,attrs:{shape:m}});else if(y||b){va.assert(s.shape.length>=1,()=>`Input must have rank at least 1, got: ${s.shape.length}`);let e=vf.computeOutShape(x,v,w),n=nW({inputs:{x:s},backend:r,attrs:{begin:x,size:e}});t=Zz({inputs:{x:n},backend:r,attrs:{shape:m}}),r.disposeIntermediateTensorInfo(n)}else if(r.shouldExecuteOnCPU([s])){let e=r.readSync(s.dataId),n=uo(s.shape,s.dtype,e),a=qP(f,n,w,x);t=r.makeTensorInfo(m,s.dtype,a.values)}else{let e=new jH(x,w,f);t=r.runWebGLProgram(e,[s],s.dtype)}let k=Zz({inputs:{x:t},backend:r,attrs:{shape:m}});return r.disposeIntermediateTensorInfo(t),k}};var KH={kernelName:Br,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{separator:a,nGramWidths:s,leftPad:i,rightPad:o,padWidth:l,preserveShortSequences:u}=r,{data:h,dataSplits:d}=t,p=n.readSync(h.dataId),c=n.readSync(d.dataId),[f,m]=KP(p,c,a,s,i,o,l,u);return[n.makeTensorInfo([f.length],"string",f),n.makeTensorInfo(d.shape,"int32",m)]}};var XH={kernelName:Wr,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{skipEmpty:a}=r,{input:s,delimiter:i}=t;if("string"!==s.dtype)throw new Error("Input must be of datatype string");if(1!==s.shape.length)throw new Error(`Input must be a vector, got shape: ${s.shape}`);if(0!==i.shape.length)throw new Error(`Delimiter must be a scalar, got shape: ${i.shape}`);let o=n.readSync(s.dataId),l=n.readSync(i.dataId)[0],[u,h,d]=XP(o,l,a),p=h.length;return[n.makeTensorInfo([p,2],"int32",u),n.makeTensorInfo([p],"string",h),n.makeTensorInfo([2],"int32",new Int32Array(d))]}};var ZH={kernelName:Vr,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{numBuckets:a}=r,{input:s}=t;if("string"!==s.dtype)throw new Error("Input must be of datatype string");if(a<=0)throw new Error("Number of buckets must be at least 1");let i=n.readSync(s.dataId),o=ZP(i,a);return n.makeTensorInfo(s.shape,"int32",o)}},YH=Bz({opSnippet:"return tan(x);"}),JH={kernelName:Gr,backendName:"webgl",kernelFunc:YH},QH=Bz({opSnippet:"\n  float e2x = exp(-2.0 * abs(x));\n  return sign(x) * (1.0 - e2x) / (1.0 + e2x);\n"}),ej={kernelName:Hr,backendName:"webgl",kernelFunc:QH};var tj={kernelName:gr,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{tensor:a,indices:s,updates:i}=t,{}=r,{sliceRank:o,numUpdates:l,sliceSize:u,strides:h,outputSize:d}=Gf.calculateShapes(i,s,a.shape),p=[d/u,u];if(0===d)return n.makeTensorInfo(a.shape,s.dtype);let c=Zz({inputs:{x:s},backend:n,attrs:{shape:[l,o]}}),f=Zz({inputs:{x:i},backend:n,attrs:{shape:[l,u]}}),m=Zz({inputs:{x:a},backend:n,attrs:{shape:p}}),g=new hH(l,o,c.shape.length,f.shape.length,h,p,!1,!0),y=n.runWebGLProgram(g,[f,c,m],m.dtype),b=Zz({inputs:{x:y},backend:n,attrs:{shape:a.shape}});return n.disposeIntermediateTensorInfo(c),n.disposeIntermediateTensorInfo(f),n.disposeIntermediateTensorInfo(m),n.disposeIntermediateTensorInfo(y),b}},nj=class{constructor(e,t){this.variableNames=["A"];let n=new Array(e.length);for(let r=0;r<n.length;r++)n[r]=e[r]*t[r];this.outputShape=n,this.rank=n.length;let r=IL(this.rank),a=function(e){let t=e.length;if(t>5)throw Error(`Tile for rank ${t} is not yet supported`);if(1===t)return`imod(resRC, ${e[0]})`;let n=["resRC.x","resRC.y","resRC.z","resRC.w","resRC.u"],r=[];for(let t=0;t<e.length;t++)r.push(`imod(${n[t]}, ${e[t]})`);return r.join()}(e);this.userCode=`\n      void main() {\n        ${r} resRC = getOutputCoords();\n        setOutput(getA(${a}));\n      }\n    `}};function rj(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{reps:s}=r;if("string"===a.dtype||a.shape.length>5){let e=n.readSync(a.dataId),t="string"===a.dtype?e.map(e=>va.decodeString(e)):e,r=uo(a.shape,a.dtype,t),i=JP(r,s);return n.makeTensorInfo(i.shape,i.dtype,i.values)}let i=new nj(a.shape,s);return n.runWebGLProgram(i,[a],a.dtype)}var aj={kernelName:jr,backendName:"webgl",kernelFunc:rj},sj=class{constructor(e){this.variableNames=["x","indices"],this.customUniforms=[{name:"n",type:"int"},{name:"firstPass",type:"int"},{name:"negativeInf",type:"float"},{name:"dir",type:"int"},{name:"inc",type:"int"}],this.outputShape=e,this.userCode="\n       void main() {\n         ivec2 coords = getOutputCoords();\n         int batch = coords[0];\n         int elemIdx = coords[1];\n\n         // We compare elements pair-wise within a group of size 2 * inc.\n         // The comparing rule for each group alternates between ascending\n         // and descending. Within each group, we compare each pair at\n         // positions i and i+inc. To decide whether an element at position i\n         // is x0 or x1, we mod it by 2 * inc, if the result is smaller than\n         // inc, it is in the first half of the group, we denote it as x0,\n         // otherwise we denote it as x1.\n         // For example, as shown in the Bitonic top K paper referenced above,\n         // Figure5(a) shows that element[1] is in the\n         // second half of the group when group size is 2, but it is in the\n         // first half of the group when group size is 4.\n\n         bool isFirstInPair = imod(elemIdx, 2 * inc) < inc;\n         int i = isFirstInPair ? elemIdx : elemIdx - inc;\n\n         int i0 = firstPass == 1 ? i : int(getIndices(batch, i));\n         int i1 = firstPass == 1 ? i + inc : int(getIndices(batch, i + inc));\n         float x0 = i0 < n ? getX(batch, i0) : negativeInf;\n         float x1 = i1 < n ? getX(batch, i1) : negativeInf;\n\n         // Denotes which direction indices are in (ascending or descending).\n         bool reverse = imod(elemIdx, 2 * dir) >= dir;\n         bool isGreater = x0 > x1 || (x0 == x1 && i1 > i0);\n         if (reverse == isGreater) { // Elements in opposite order of direction\n           int iTemp = i0;\n           i0 = i1;\n           i1 = iTemp;\n         }\n         if (isFirstInPair) {\n            setOutput(float(i0));\n         } else {\n            setOutput(float(i1));\n         }\n       }\n     "}},ij=class{constructor(e){this.variableNames=["x","indices"],this.customUniforms=[{name:"n",type:"int"},{name:"firstPass",type:"int"},{name:"k",type:"int"}],this.outputShape=e,this.userCode="\n    void main() {\n         // Takes max of indices (0, k), (1, k + 1), (2, k + 2) ...\n         ivec2 coords = getOutputCoords();\n         int batch = coords[0];\n         int elemIdx = coords[1];\n\n         // The output size is half of the previous size.\n         // If the previous sequence is | | | | _ _ _ _  | | | |  _ _ _ _ (k=4),\n         // we only need to output the indices at positions |, the indices at\n         // positions _ can be thrown away, see Figure5(b) After Phase 2\n         // (Merge phase) in the Bitonic Top K paper referenced above.\n         // For example, the paper shows we only need to output the orange bars.\n         // The output sequence should look like this | | | | | | | |.\n         // Because the sequence is halved, to map the output index back\n         // to the previous sequence to find the corresponding value,\n         // we need to double the index. When we double the index,\n         // we basically interpolate a position, so 2i looks like\n         // | _ | _ | _ | _ | _ | _ | _. We move the | to the first k position\n         // of each 2k positions by - elemIdx % k. E.g. for output at\n         // index 4,5,6,7, we want to get the corresponding element at\n         // original index 8,9,10,11, for output at index 8,9,10,11,\n         // we want to get the corresponding element at original index\n         // 16,17,18,19, so on and so forth.\n\n         int i = elemIdx < k ? elemIdx : (elemIdx * 2 - imod(elemIdx, k));\n         int i0 = firstPass == 1 ? i : int(getIndices(batch, i));\n         int i1 = firstPass == 1 ? i + k : int(getIndices(batch, i + k));\n\n         float x0 = getX(batch, i0);\n         float x1 = i1 < n ? getX(batch, i1) : x0;\n\n         setOutput(x0 >= x1 ? float(i0) : float(i1));\n       }\n     "}};function oj(e,t){null!==t&&e.disposeIntermediateTensorInfo(t)}function lj(e){let t=1;for(;t<e;)t*=2;return t}var uj={kernelName:qr,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{k:s,sorted:i}=r,o=Le().getNumber("TOPK_LAST_DIM_CPU_HANDOFF_SIZE_THRESHOLD"),l=Le().getNumber("TOPK_K_CPU_HANDOFF_THRESHOLD"),u=a.shape,h=u[u.length-1];if(n.shouldExecuteOnCPU([a])||h<o||s>l){let e=n.readSync(a.dataId),[t,r]=QP(e,u,a.dtype,s,i);return[n.makeTensorInfo(t.shape,t.dtype,t.values),n.makeTensorInfo(r.shape,r.dtype,r.values)]}if(0===s)return u[u.length-1]=0,[n.makeTensorInfo(u,a.dtype,[]),n.makeTensorInfo(u,"int32",[])];if(1===h)return[a,BV({attrs:{shape:u,dtype:"int32",value:0},backend:n})];let d=n.texData.get(a.dataId),p=null!==d&&d.isPacked,c=p?n.unpackTensor(a):a,f=va.sizeFromShape(u)/h,m=Zz({inputs:{x:c},attrs:{shape:[f,h]},backend:n});p&&oj(n,c);let g=lj(s),y=lj(h),b=null,x=()=>null===b?[m,m]:[m,b],v=(e,t,r)=>{let a=x(),s=new sj(r),i=[[h],[null===b?1:0],[Number.NEGATIVE_INFINITY],[e],[t]],o=b;b=n.runWebGLProgram(s,a,"int32",i),oj(n,o)};for(let e=1;e<g;e*=2){let t=2*e;for(let n=e;n>=1;n/=2)v(t,n,[f,y])}for(let e=y;e>g;e/=2){let t=x(),r=new ij([f,e/2]),a=[[h],[null===b?1:0],[g]],s=b;b=n.runWebGLProgram(r,t,"int32",a),oj(n,s);let i=g/2,o=2*i;for(let e=i;e>=1;e/=2)v(o,e,b.shape)}let w=b;b=nW({inputs:{x:b},backend:n,attrs:{begin:0,size:[f,s]}}),oj(n,w);let k=sU({inputs:{x:m,indices:b},backend:n,attrs:{axis:1,batchDims:1}});oj(n,m);let I=u.slice(0,-1);I.push(s),w=b,b=Zz({inputs:{x:b},attrs:{shape:I},backend:n}),oj(n,w);let S=k;return k=Zz({inputs:{x:k},attrs:{shape:I},backend:n}),oj(n,S),[k,b]}},hj=class{constructor(e,t,n,r,a,s){this.variableNames=["Image","Transforms"],this.outputShape=s;let i,o="nearest"===n?1:2;switch(r){case"constant":default:i=1;break;case"reflect":i=2;break;case"wrap":i=3;break;case"nearest":i=4}this.userCode=`\n            float mapCoord(float outCoord, float len) {\n              float inCoord = outCoord;\n              if(${i} == 2) {\n                if (inCoord < 0.0) {\n                  if (len <= 1.0) {\n                    inCoord = 0.0;\n                  } else {\n                    float sz2 = 2.0 * len;\n                    if (inCoord < sz2) {\n                      inCoord = sz2 * float(int(float(-inCoord / sz2))) +\n                      inCoord;\n                    }\n                    inCoord = inCoord < -len ? inCoord + sz2 : -inCoord - 1.0;\n                  }\n                } else if (inCoord > len - 1.0) {\n                  if (len <= 1.0) {\n                    inCoord = 0.0;\n                  } else {\n                    float sz2 = 2.0 * len;\n                    inCoord -= sz2 * float(int(float(inCoord / sz2)));\n                    if (inCoord >= len) {\n                      inCoord = sz2 - inCoord - 1.0;\n                    }\n                  }\n                }\n                return clamp(inCoord, 0.0, len - 1.0);\n              } else if (${i} == 3) {\n                if (inCoord < 0.0) {\n                  if (len <= 1.0) {\n                    inCoord = 0.0;\n                  } else {\n                    float sz = len - 1.0;\n                    inCoord += len * (float(int(float(-inCoord / sz))) + 1.0);\n                  }\n                } else if (inCoord > len - 1.0) {\n                  if (len <= 1.0) {\n                    inCoord = 0.0;\n                  } else {\n                    float sz = len - 1.0;\n                    inCoord -= len * float(int(float(inCoord / sz)));\n                  }\n                }\n                return clamp(inCoord, 0.0, len - 1.0);\n              } else if (${i} == 4) {\n                return clamp(outCoord, 0.0, len - 1.0);\n              } else {\n                return outCoord;\n              }\n            }\n\n            float readWithFillValue(int batch, int coordY, int coordX,\n              int channel) {\n              float outputValue;\n              if (0 <= coordY && coordY < ${e} && 0 <= coordX && coordX < ${t}) {\n                  outputValue = getImage(batch, coordY, coordX, channel);\n              } else {\n                outputValue = float(${a});\n              }\n              return outputValue;\n            }\n\n            void main() {\n              ivec4 coords = getOutputCoords();\n              float outputValue;\n              int batch = coords[0];\n              int x = coords[2];\n              int y = coords[1];\n              int channel = coords[3];\n              float xf = float(x);\n              float yf = float(y);\n              float a1 = getTransforms(batch, 0);\n              float a2 = getTransforms(batch, 1);\n              float a3 = getTransforms(batch, 2);\n              float b1 = getTransforms(batch, 3);\n              float b2 = getTransforms(batch, 4);\n              float b3 = getTransforms(batch, 5);\n              float c1 = getTransforms(batch, 6);\n              float c2 = getTransforms(batch, 7);\n              float projection = c1 * xf + c2 * yf + 1.0;\n              if (projection == 0.0) {\n                outputValue = float(${a});\n              } else {\n                float inX = (a1 * xf + a2 * yf + a3) / projection;\n                float inY = (b1 * xf + b2 * yf + b3) / projection;\n                float mapX = mapCoord(inX, float(${t}));\n                float mapY = mapCoord(inY, float(${e}));\n\n                if (${o} == 1) {\n                  int coordY = int(round(mapY));\n                  int coordX = int(round(mapX));\n                  outputValue = readWithFillValue(batch, coordY, coordX,\n                    channel);\n                } else {\n                  float yFloor = floor(mapY);\n                  float xFloor = floor(mapX);\n                  float yCeil = yFloor + 1.0;\n                  float xCeil = xFloor + 1.0;\n                  float valueYFloor = (xCeil - mapX) *\n                  readWithFillValue(batch, int(yFloor), int(xFloor), channel) +\n                  (mapX - xFloor) *\n                  readWithFillValue(batch, int(yFloor), int(xCeil), channel);\n                  float valueYCeil = (xCeil - mapX) *\n                  readWithFillValue(batch, int(yCeil), int(xFloor), channel) +\n                  (mapX - xFloor) *\n                  readWithFillValue(batch, int(yCeil), int(xCeil), channel);\n                  outputValue = (yCeil - mapY) * valueYFloor +\n                  (mapY - yFloor) * valueYCeil;\n                }\n              }\n              setOutput(outputValue);\n            }\n        `}};var dj={kernelName:Kr,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{image:a,transforms:s}=t,{interpolation:i,fillMode:o,fillValue:l,outputShape:u}=r,[h,d,p,c]=a.shape,[f,m]=null!=u?u:[d,p],g=new hj(d,p,i,o,l,[h,f,m,c]);return n.runWebGLProgram(g,[a,s],"float32")}};var pj={kernelName:Zr,backendName:"webgl",kernelFunc:function(e){let{inputs:t,attrs:n,backend:r}=e,{axis:a}=n,{x:s}=t;aL(s,"unique"),console.warn("WARNING: ","UI might be locked temporarily as data is being downloaded");let i=r.readSync(s.dataId),{outputValues:o,outputShape:l,indices:u}=tz(i,a,s.shape,s.dtype);return[r.makeTensorInfo(l,s.dtype,o),r.makeTensorInfo([u.length],"int32",u)]}};var cj={kernelName:Yr,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{value:a}=t,{axis:s}=r;s<0&&(s+=a.shape.length);let i=a,o=i.shape.length,l=a.shape[s],u=new Array(o-1),h=0;for(let e=0;e<o;e++)e!==s&&(u[h++]=i.shape[e]);let d=[],p=new Array(o).fill(0),c=i.shape.slice();c[s]=1;let f=new Array(l);for(let e=0;e<f.length;e++){p[s]=e;let t=nW({inputs:{x:i},backend:n,attrs:{begin:p,size:c}}),r=Zz({inputs:{x:t},backend:n,attrs:{shape:u}});f[e]=r,d.push(t)}return d.forEach(e=>n.disposeIntermediateTensorInfo(e)),f}},fj=class{constructor(e,t){this.variableNames=["x","segmentIds"];let n=e.windowSize,r=e.batchSize,a=e.inSize,s=e.numSegments,i=s*Math.ceil(a/n);this.outputShape=[r,i];let o=4*Math.floor(n/4),l=n%4,u="\n        sumValue += dot(values, segFilter);\n    ",h="";a%n>0&&(h=`\n        if (inIdx < 0 || inIdx >= ${a}) {\n          return initializationValue;\n        }\n      `);let d="";a%n>0&&(d=`\n        if (inIdx < 0 || inIdx >= ${a}) {\n          return -1.0;\n        }\n      `),this.userCode=`\n      const float initializationValue = 0.0;\n\n      float getValue(int batch, int inIdx) {\n        ${h}\n        return getX(batch, inIdx);\n      }\n\n      float getSegmentIdAtIndex(int inIdx) {\n        ${d}\n        return getSegmentIds(inIdx);\n      }\n\n      void main() {\n        ivec2 coords = getOutputCoords();\n        int batch = coords[0];\n        int outIdx = coords[1];\n        int inOffset = int(floor(float(outIdx) / float(\n          ${s})) * float(${n}));\n        int currentSeg = int(mod(float(outIdx), float(${s})));\n\n        float sumValue = 0.0;\n\n        for (int i = 0; i < ${o}; i += 4) {\n          int inIdx = inOffset + i;\n          vec4 values = vec4(\n            getValue(batch, inIdx),\n            getValue(batch, inIdx + 1),\n            getValue(batch, inIdx + 2),\n            getValue(batch, inIdx + 3)\n          );\n\n          vec4 segFilter = vec4(\n            int(getSegmentIdAtIndex(inIdx)) == currentSeg ? 1 : 0,\n            int(getSegmentIdAtIndex(inIdx + 1)) == currentSeg ? 1 : 0,\n            int(getSegmentIdAtIndex(inIdx + 2)) == currentSeg ? 1 : 0,\n            int(getSegmentIdAtIndex(inIdx + 3)) == currentSeg ? 1 : 0\n          );\n\n          ${u}\n        }\n\n        int inIdx = inOffset + ${o};\n        if (${1===l}) {\n          vec4 values = vec4(\n            getValue(batch, inIdx),\n            initializationValue,\n            initializationValue,\n            initializationValue\n          );\n\n          int inIdxSeg = int(getSegmentIdAtIndex(inIdx));\n\n          vec4 segFilter = vec4(\n            int(getSegmentIdAtIndex(inIdx)) == currentSeg ? 1 : 0,\n            0,\n            0,\n            0\n          );\n\n          ${u}\n        } else if (${2===l}) {\n          vec4 values = vec4(\n            getValue(batch, inIdx),\n            getValue(batch, inIdx + 1),\n            initializationValue,\n            initializationValue\n          );\n\n          vec4 segFilter = vec4(\n            int(getSegmentIdAtIndex(inIdx)) == currentSeg ? 1 : 0,\n            int(getSegmentIdAtIndex(inIdx + 1)) == currentSeg ? 1 : 0,\n              0,\n              0\n          );\n\n          ${u}\n        } else if (${3===l}) {\n          vec4 values = vec4(\n            getValue(batch, inIdx),\n            getValue(batch, inIdx + 1),\n            getValue(batch, inIdx + 2),\n            initializationValue\n          );\n\n          vec4 segFilter = vec4(\n            int(getSegmentIdAtIndex(inIdx)) == currentSeg ? 1 : 0,\n            int(getSegmentIdAtIndex(inIdx + 1)) == currentSeg ? 1 : 0,\n            int(getSegmentIdAtIndex(inIdx + 2)) == currentSeg ? 1 : 0,\n            0\n          );\n\n          ${u}\n        }\n        setOutput(sumValue);\n      }\n    `}};var mj,gj,yj,bj={kernelName:Jr,backendName:"webgl",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a,segmentIds:s}=t,{numSegments:i}=r,o=a.shape.length,l=[],u=0,h=Gf.getAxesPermutation([u],o),d=a;null!=h&&(d=iB({inputs:{x:a},backend:n,attrs:{perm:h}}),l.push(d),u=Gf.getInnerMostAxes(1,o)[0]);let p=Gf.segment_util.computeOutShape(d.shape,u,i),c=va.sizeFromShape([d.shape[u]]),f=Zz({inputs:{x:d},backend:n,attrs:{shape:[-1,c]}});l.push(f);let m=cs(a.dtype),g=(e,t,r,a,s)=>{let i=e.shape[0],o=e.shape[1],u=Gf.segment_util.segOpComputeOptimalWindowSize(o,s),h=new fj({windowSize:u,inSize:o,batchSize:i,numSegments:s},t),d=n.compileAndRun(h,[e,r],a);if(l.push(d),d.shape[1]===s)return d;let p=LG({backend:n,attrs:{start:0,stop:s,step:1,dtype:"float32"}}),c=rj({inputs:{x:p},backend:n,attrs:{reps:[o/u]}});return l.push(p),l.push(c),g(d,t,c,a,s)},y=Zz({inputs:{x:g(f,"unsortedSegmentSum",s,m,i)},backend:n,attrs:{shape:p}}),b=y;if(null!=h){l.push(y);let e=Gf.getUndoAxesPermutation(h);b=iB({inputs:{x:b},backend:n,attrs:{perm:e}})}return l.forEach(e=>n.disposeIntermediateTensorInfo(e)),b}},xj=[uB,dB,cB,mB,bB,wB,kB,IB,EB,AB,RB,DB,OB,PB,BB,UB,GB,qB,KB,XB,JB,aW,sW,iW,oW,pW,mW,bW,Rz,wW,EW,LW,VW,GW,HW,jW,KW,ZW,JW,eV,sV,iV,oV,uV,pV,mV,gV,bV,vV,wV,IV,SV,NV,CV,AV,RV,MV,PV,VV,GV,qV,XV,JV,eU,tU,rU,iU,lU,hU,Az,dU,NW,cU,mU,yU,Mz,xU,wU,kU,SU,NU,CU,AU,RU,MU,LU,zU,WU,VU,UU,jU,qU,KU,XU,ZU,JU,tG,rG,pG,Xz,fG,gG,bG,vG,uW,kG,_G,NG,AG,RG,Pz,FG,DG,MG,OG,PG,dW,iG,BG,VG,GG,Yz,qG,XG,JG,eH,rH,sH,oH,uH,pH,fH,gH,bH,vH,kH,SH,NH,rW,dG,CH,EH,AH,$H,RH,FH,DH,MH,PH,BH,UH,GH,HH,qH,KH,XH,ZH,uG,sB,JH,ej,tj,aj,uj,dj,oB,pj,cj,bj,SG];for(let e of xj)fa(e);!function(e){e[e.float32=0]="float32",e[e.int32=1]="int32",e[e.bool=2]="bool",e[e.string=3]="string",e[e.complex64=4]="complex64"}(mj||(mj={})),function(e){e[e.linear=0]="linear",e[e.relu=1]="relu",e[e.relu6=2]="relu6",e[e.prelu=3]="prelu",e[e.leakyrelu=4]="leakyrelu",e[e.sigmoid=5]="sigmoid",e[e.elu=6]="elu"}(gj||(gj={}));var vj={kernelName:aa,backendName:"wasm",setupFunc:function(e){yj=e.wasm.cwrap(aa,null,["number","array","number","number","array","number","number","number","number","number","number","number","number"])},kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{a,b:s,bias:i,preluActivationWeights:o}=t;if("float32"!==a.dtype||"float32"!==s.dtype)throw new Error("_FusedMatMul for non non-float32 tensors not yet supported.");let{transposeA:l,transposeB:u,activation:h,leakyreluAlpha:d}=r,p=n.dataIdMap.get(a.dataId).id,c=n.dataIdMap.get(s.dataId).id,f=0;if(null!=i){let e=n.dataIdMap.get(i.dataId);if(1!==e.shape.length)throw new Error(`_FusedMatMul only supports rank-1 bias but got rank ${e.shape.length}.`);f=e.id}let m=null==o?0:n.dataIdMap.get(o.dataId).id,g=gj[h];if(null==g)throw new Error(`${h} activation not yet supported for FusedConv2D in the wasm backend.`);let y=l?a.shape[2]:a.shape[1],b=u?s.shape[1]:s.shape[2],x=Dl.assertAndGetBroadcastShape(a.shape.slice(0,-2),s.shape.slice(0,-2)),v=n.makeOutput([...x,y,b],a.dtype),w=n.dataIdMap.get(v.dataId).id,k=new Uint8Array(new Int32Array(a.shape).buffer),I=new Uint8Array(new Int32Array(s.shape).buffer);return yj(p,k,a.shape.length,c,I,s.shape.length,l,u,g,f,m,d||0,w),v}};function wj(e,t){let n;return{kernelName:e,backendName:"wasm",setupFunc:function(t){n=t.wasm.cwrap(e,null,["number","number","number"])},kernelFunc:function(e){let{backend:r,inputs:{x:a}}=e,s=r.dataIdMap.get(a.dataId).id,i=r.makeOutput(a.shape,t||a.dtype),o=r.dataIdMap.get(i.dataId).id;return 0===va.sizeFromShape(i.shape)||n(s,mj[a.dtype],o),i}}}var kj=wj(Ve),Ij=wj(Ue),Sj=wj(Ge);function _j(e,t,n){let r;return{kernelName:e,backendName:"wasm",setupFunc:function(t){r=t.wasm.cwrap(e,null,["number","array","number","number","array","number","number","number"])},kernelFunc:function(e){let{backend:t,inputs:a}=e,{a:s,b:i}=a,o=t.dataIdMap.get(s.dataId).id,l=t.dataIdMap.get(i.dataId).id,u=null!=n?n:s.dtype,h=Gf.assertAndGetBroadcastShape(s.shape,i.shape),d=t.makeOutput(h,u);if(0===va.sizeFromShape(h))return d;let p=new Uint8Array(new Int32Array(s.shape).buffer),c=new Uint8Array(new Int32Array(i.shape).buffer),f=t.dataIdMap.get(d.dataId).id;return r(o,p,s.shape.length,l,c,i.shape.length,mj[s.dtype],f),d}}}var Nj,Tj=_j(He);var Cj={kernelName:je,backendName:"wasm",setupFunc:function(e){Nj=e.wasm.cwrap(je,null,["array","number","number","number"])},kernelFunc:function(e){let{inputs:t,backend:n}=e,r=n.makeOutput(t[0].shape,t[0].dtype);if(0===va.sizeFromShape(r.shape))return r;let a=t.map(e=>n.dataIdMap.get(e.dataId).id),s=new Uint8Array(new Int32Array(a).buffer),i=n.dataIdMap.get(r.dataId).id;return Nj(s,a.length,mj[r.dtype],i),r}};function Ej(e){let{inputs:{x:t},backend:n}=e;if("string"===t.dtype)return Vs(n.readSync(t.dataId),t.shape,t.dtype);let r=n.makeOutput(t.shape,t.dtype),a=n.typedArrayFromHeap(t);return n.typedArrayFromHeap(r).set(a),r}var Aj,$j={kernelName:an,backendName:"wasm",kernelFunc:Ej};function Rj(e){let{inputs:t,backend:n,attrs:r}=e,[a,s]=function(e,t){let n=[],r=[];for(let a=0;a<e.length;++a)1!==e[a]&&n.push(e[a]),1!==e[t[a]]&&r.push(t[a]);for(let e=0;e<r.length;++e){let t=-1;for(let n=0;n<r.length;++n)r[n]>=e&&(-1===t||r[t]>r[n])&&(t=n);r[t]=e}return[n,r]}(t.x.shape,r.perm),i=!0;for(let e=0;e<s.length;e++)s[e]!==e&&(i=!1);let o=function(e,t){let n=new Array(e.length);for(let r=0;r<n.length;r++)n[r]=e[t[r]];return n}(t.x.shape,r.perm),l={dataId:t.x.dataId,shape:a,dtype:t.x.dtype};if(i){let e=Ej({inputs:t,backend:n});return e.shape=o,e}let u=n.makeOutput(o,l.dtype),h=n.dataIdMap.get(l.dataId).id,d=n.dataIdMap.get(u.dataId).id,p=new Uint8Array(new Int32Array(s).buffer),c=new Uint8Array(new Int32Array(l.shape).buffer);return Aj(h,c,l.shape.length,mj[l.dtype],d,p,s.length),u}var Fj,Dj={kernelName:Xr,backendName:"wasm",kernelFunc:Rj,setupFunc:function(e){Aj=e.wasm.cwrap(Xr,null,["number","array","number","number","number","array","number"])}};function Mj(e,t,n){let r=e.shape,a=e.shape.length,s=va.parseAxisParam(t,r),i=s,o=Gf.getAxesPermutation(i,a),l=null,u=!1;if(null!=o){let t=new Array(a);for(let e=0;e<t.length;e++)t[e]=r[o[e]];i=Gf.getInnerMostAxes(i.length,a),l=Rj({inputs:{x:e},attrs:{perm:o},backend:n});let s=n.dataIdMap.get(e.dataId).id;n.dataIdMap.get(l.dataId).id!==s&&(u=!0)}return{transposed:l,originalAxes:s,axes:i,inputWasTransposed:u}}var Oj,Lj={kernelName:qe,backendName:"wasm",setupFunc:function(e){Fj=e.wasm.cwrap(qe,null,["number, number, number"])},kernelFunc:function(e){let{backend:t,inputs:n,attrs:r}=e,{axis:a,keepDims:s}=r,{x:i}=n,o=t.dataIdMap.get(i.dataId).id,l=i,{transposed:u,axes:h,originalAxes:d,inputWasTransposed:p}=Mj(i,a,t);if(p){l=u,o=t.dataIdMap.get(u.dataId).id}let c=l.shape.length;Gf.assertAxesAreInnerMostDims("all",h,c);let[f,m]=Gf.computeOutAndReduceShapes(l.shape,h),g=va.sizeFromShape(m),y=t.makeOutput(f,i.dtype);if(0!==va.sizeFromShape(l.shape)){let e=t.dataIdMap.get(y.dataId).id;Fj(o,g,e)}if(p&&t.disposeData(u.dataId),s){let e=Gf.expandShapeToKeepDim(y.shape,d);y.shape=e}return y}};var Pj={kernelName:Ke,backendName:"wasm",setupFunc:function(e){Oj=e.wasm.cwrap(Ke,null,["number, number, number"])},kernelFunc:function(e){let{backend:t,inputs:n,attrs:r}=e,{axis:a,keepDims:s}=r,{x:i}=n,o=t.dataIdMap.get(i.dataId).id,l=i,{transposed:u,axes:h,originalAxes:d,inputWasTransposed:p}=Mj(i,a,t);if(p){l=u,o=t.dataIdMap.get(u.dataId).id}let c=l.shape.length;Gf.assertAxesAreInnerMostDims("any",h,c);let[f,m]=Gf.computeOutAndReduceShapes(l.shape,h),g=va.sizeFromShape(m),y=t.makeOutput(f,i.dtype);if(0!==va.sizeFromShape(l.shape)){let e=t.dataIdMap.get(y.dataId).id;Oj(o,g,e)}if(p&&t.disposeData(u.dataId),s){let e=Gf.expandShapeToKeepDim(y.shape,d);y.shape=e}return y}};function zj(e){let t;return{kernelName:e,backendName:"wasm",setupFunc:function(n){t=n.wasm.cwrap(e,null,["number","number","number","number","number"])},kernelFunc:function(e){let{backend:n,inputs:r,attrs:a}=e,{axis:s}=a,{x:i}=r,o=n.dataIdMap.get(i.dataId).id,l=o,u=i,{transposed:h,axes:d,inputWasTransposed:p}=Mj(i,s,n);if(p){let e=n.dataIdMap.get(h.dataId).id;e!==o&&(u=h,l=e)}let c=u.shape.slice(0,-1),f=n.makeOutput(c,"int32"),m=n.dataIdMap.get(f.dataId).id,g=va.sizeFromShape(f.shape),y=u.shape[d[0]];return t(l,mj[u.dtype],g,y,m),p&&n.disposeData(h.dataId),f}}}var Bj,Wj=zj(Xe),Vj=zj(Ze),Uj=wj(Ye),Gj=wj(Je),Hj=wj(Qe),jj=_j(tt),qj=wj(et);var Kj,Xj={kernelName:nt,backendName:"wasm",setupFunc:function(e){Bj=e.wasm.cwrap(nt,null,["number","number","number","number","number","number","number","number","number","number","number","number","number","number"])},kernelFunc:function(e){let{inputs:t,attrs:n,backend:r}=e,a=t.x,s=r.dataIdMap.get(a.dataId).id,{filterSize:i,strides:o,pad:l,dimRoundingMode:u}=n,h=Gf.computePool2DInfo(a.shape,i,o,1,l,u),d=h.filterHeight,p=h.filterWidth,c=h.padInfo.top,f=h.padInfo.right,m=h.padInfo.bottom,g=h.padInfo.left,y=h.strideHeight,b=h.strideWidth,x=h.inChannels;if("channelsLast"!==h.dataFormat)throw new Error(`wasm backend does not support dataFormat:'${h.dataFormat}'. Please use 'channelsLast'.`);if(1!==h.dilationWidth||1!==h.dilationHeight)throw new Error(`was backend only supports average pooling with dilation = [1, 1], got [${h.dilationHeight}, ${h.dilationWidth}].`);let v=r.makeOutput(h.outShape,"float32"),w=r.dataIdMap.get(v.dataId).id;return Bj(s,a.shape[0],a.shape[1],a.shape[2],d,p,c,f,m,g,y,b,x,w),v}};var Zj,Yj={kernelName:at,backendName:"wasm",setupFunc:function(e){Kj=e.wasm.cwrap("AvgPool3D",null,["number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number"])},kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{filterSize:s,strides:i,pad:o,dimRoundingMode:l,dataFormat:u}=r,h=Gf.computePool3DInfo(a.shape,s,i,1,o,l,u),d=n.makeOutput(h.outShape,a.dtype);return Kj(n.dataIdMap.get(a.dataId).id,n.dataIdMap.get(d.dataId).id,h.batchSize,h.inChannels,h.inDepth,h.inHeight,h.inWidth,h.outDepth,h.outHeight,h.outWidth,h.strideDepth,h.strideHeight,h.strideWidth,h.dilationDepth,h.dilationHeight,h.dilationWidth,h.effectiveFilterDepth,h.effectiveFilterHeight,h.effectiveFilterWidth,h.padInfo.front,h.padInfo.top,h.padInfo.left),d}};var Jj,Qj={kernelName:st,backendName:"wasm",setupFunc:function(e){Zj=e.wasm.cwrap("AvgPool3DGrad",null,["number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number"])},kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{dy:a,input:s}=t,{filterSize:i,strides:o,pad:l,dimRoundingMode:u}=r,h=Gf.computePool3DInfo(s.shape,i,o,1,l,u),d=n.makeOutput(s.shape,s.dtype);return Zj(n.dataIdMap.get(a.dataId).id,n.dataIdMap.get(d.dataId).id,h.batchSize,h.inChannels,h.inDepth,h.inHeight,h.inWidth,h.outDepth,h.outHeight,h.outWidth,h.strideDepth,h.strideHeight,h.strideWidth,h.dilationDepth,h.dilationHeight,h.dilationWidth,h.effectiveFilterDepth,h.effectiveFilterHeight,h.effectiveFilterWidth,h.padInfo.front,h.padInfo.top,h.padInfo.left,h.filterDepth,h.filterHeight,h.filterWidth),d}};var eq={kernelName:rt,backendName:"wasm",setupFunc:function(e){Jj=e.wasm.cwrap("AvgPoolGrad",null,["number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number"])},kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{dy:a,input:s}=t,{filterSize:i,strides:o,pad:l}=r,u=Gf.computePool2DInfo(s.shape,i,o,1,l),h=n.makeOutput(s.shape,s.dtype);return Jj(n.dataIdMap.get(a.dataId).id,n.dataIdMap.get(h.dataId).id,u.batchSize,u.inChannels,u.inHeight,u.inWidth,u.outHeight,u.outWidth,u.strideHeight,u.strideWidth,u.dilationHeight,u.dilationWidth,u.effectiveFilterHeight,u.effectiveFilterWidth,u.padInfo.top,u.padInfo.left,u.filterHeight,u.filterWidth),h}};function tq(e){let{inputs:t,attrs:n}=e,{x:r}=t,{shape:a}=n,s=va.sizeFromShape(r.shape),i=va.inferFromImplicitShape(a,s);return va.assert(s===va.sizeFromShape(i),()=>`new shape: ${i}, old shape: ${r.shape}. New shape and old shape must have the same number of elements.`),e.backend.incRef(r.dataId),{dataId:r.dataId,shape:i,dtype:r.dtype}}var nq,rq={kernelName:ir,backendName:"wasm",kernelFunc:tq};var aq={kernelName:it,backendName:"wasm",setupFunc:function(e){nq=e.wasm.cwrap(it,null,["number","array","number","number","array","number","number","number","number"])},kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{a,b:s}=t,{transposeA:i,transposeB:o}=r;if("float32"!==a.dtype||"float32"!==s.dtype)throw new Error("BatchMatMul for non non-float32 tensors not yet supported.");let l=a.shape.length,u=s.shape.length,h=i?a.shape[l-2]:a.shape[l-1],d=o?s.shape[u-1]:s.shape[u-2],p=i?a.shape[l-1]:a.shape[l-2],c=o?s.shape[u-2]:s.shape[u-1],f=a.shape.slice(0,-2),m=s.shape.slice(0,-2),g=va.sizeFromShape(f),y=va.sizeFromShape(m),b=Dl.assertAndGetBroadcastShape(a.shape.slice(0,-2),s.shape.slice(0,-2)).concat([p,c]);va.assert(h===d,()=>`Error in matMul: inner shapes (${h}) and (${d}) of Tensors with shapes ${a.shape} and ${s.shape} and transposeA=${i} and transposeB=${o} must match.`);let x=o?[y,c,d]:[y,d,c],v=tq({inputs:{x:a},backend:n,attrs:{shape:i?[g,h,p]:[g,p,h]}}),w=tq({inputs:{x:s},backend:n,attrs:{shape:x}}),k=n.dataIdMap.get(v.dataId).id,I=n.dataIdMap.get(w.dataId).id,S=i?v.shape[2]:v.shape[1],_=o?w.shape[1]:w.shape[2],N=Math.max(g,y),T=n.makeOutput([N,S,_],v.dtype),C=n.dataIdMap.get(T.dataId).id,E=new Uint8Array(new Int32Array(v.shape).buffer),A=new Uint8Array(new Int32Array(w.shape).buffer);return nq(k,E,v.shape.length,I,A,w.shape.length,i,o,C),n.disposeData(v.dataId),n.disposeData(w.dataId),T.shape=b,T}};function sq(e){let{inputs:{x:t},attrs:{begin:n,size:r},backend:a}=e,[s,i]=vf.parseSliceParams(t,n,r),o=vf.isSliceContinous(t.shape,s,i),l=a.readSync(t.dataId),u=a.makeOutput(i,t.dtype),h=va.computeStrides(t.shape),d=a.dataIdMap.get(u.dataId);if(o){let e=vf.computeFlatOffset(s,h);return"string"===t.dtype?d.stringBytes=l.slice(e,e+va.sizeFromShape(i)):a.typedArrayFromHeap(u).set(l.subarray(e,e+va.sizeFromShape(i))),u}if("string"===t.dtype){let e=f$(l,s,i,t.shape,t.dtype);return d.stringBytes=e,u}let p=a.typedArrayFromHeap(u),c=t.shape.length;if(2===c)!function(e,t,n,r,a){let s=0,i=r[0],o=r[1],l=i+a[0];for(let r=i;r<l;r++){let i=r*t+o;n.set(e.subarray(i,i+a[1]),s),s+=a[1]}}(l,h[0],p,s,i);else if(3===c)!function(e,t,n,r,a,s){let i=0,o=a[0],l=a[1],u=a[2],h=o+s[0],d=l+s[1];for(let a=o;a<h;a++)for(let o=l;o<d;o++){let l=a*t+o*n+u;r.set(e.subarray(l,l+s[2]),i),i+=s[2]}}(l,h[0],h[1],p,s,i);else if(4===c)!function(e,t,n,r,a,s,i){let o=0,l=s[0],u=s[1],h=s[2],d=l+i[0],p=u+i[1],c=h+i[2],f=s[3];for(let s=l;s<d;s++)for(let l=u;l<p;l++)for(let u=h;u<c;u++){let h=s*t+l*n+u*r+f;a.set(e.subarray(h,h+i[3]),o),o+=i[3]}}(l,h[0],h[1],h[2],p,s,i);else{let e=f$(l,s,i,t.shape,t.dtype);p.set(e)}return u}var iq={kernelName:vr,backendName:"wasm",kernelFunc:sq};var oq,lq={kernelName:ot,backendName:"wasm",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{blockShape:s,crops:i}=r,o=s.reduce((e,t)=>e*t),l=Gf.getReshaped(a.shape,s,o),u=Gf.getPermuted(l.length,s.length),h=Gf.getReshapedPermuted(a.shape,s,o),d=Gf.getSliceBeginCoords(i,s.length),p=Gf.getSliceSize(h,i,s.length),c=tq({inputs:{x:a},backend:n,attrs:{shape:l}}),f=Rj({inputs:{x:c},backend:n,attrs:{perm:u}}),m=tq({inputs:{x:f},backend:n,attrs:{shape:h}}),g=sq({inputs:{x:m},backend:n,attrs:{begin:d,size:p}});return n.disposeData(c.dataId),n.disposeData(f.dataId),n.disposeData(m.dataId),g}};var uq={kernelName:lt,backendName:"wasm",setupFunc:function(e){oq=e.wasm.cwrap(lt,null,["number","number","boolean","number","number","number"])},kernelFunc:function(e){let{backend:t,inputs:n,attrs:r}=e,{x:a,weights:s}=n,{size:i}=r,o=0!==s.shape.reduce((e,t)=>e*t,1),l=1===a.shape.length?[i]:[a.shape[0],i],u=t.makeOutput(l,s.dtype);function h(e){return t.dataIdMap.get(e.dataId).id}return oq(h(a),i,o,h(s),mj[s.dtype],h(u)),u}},hq=_j(ut);var dq={kernelName:dt,backendName:"wasm",kernelFunc:function(e){let{inputs:t,backend:n}=e,{s0:r,s1:a}=t,s=n.typedArrayFromHeap(r),i=n.typedArrayFromHeap(a),o=Gf.assertAndGetBroadcastShape(Array.from(s),Array.from(i));return n.makeOutput([o.length],"int32",void 0,new Int32Array(o))}};function pq(e){let{inputs:{x:t},attrs:{dtype:n},backend:r}=e,a=r.makeOutput(t.shape,n),s=r.typedArrayFromHeap(t);return r.typedArrayFromHeap(a).set(s),a}var cq,fq={kernelName:pt,backendName:"wasm",kernelFunc:pq},mq=wj(ct);var gq={kernelName:ft,backendName:"wasm",setupFunc:function(e){cq=e.wasm.cwrap(ft,null,["number","number","number","number"])},kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{clipValueMin:s,clipValueMax:i}=r,o=n.dataIdMap.get(a.dataId).id,l=n.makeOutput(a.shape,a.dtype),u=n.dataIdMap.get(l.dataId).id;return cq(o,s,i,u),l}};function yq(e){let{inputs:t,backend:n}=e,r=va.parseAxisParam(e.attrs.axis,t[0].shape)[0],a=t.map(e=>e.shape);Gf.assertParamsConsistent(a,r);let s=Gf.computeOutShape(t.map(e=>e.shape),r),i=t.filter(e=>va.sizeFromShape(e.shape)>0);if(1===i.length)return Ej({inputs:{x:i[0]},backend:n});let o=n.makeOutput(s,t[0].dtype);if(0===va.sizeFromShape(s))return o;if("string"===i[0].dtype){let e=i.map(e=>{let t=[-1,va.sizeFromShape(e.shape.slice(r))];return tq({inputs:{x:e},backend:n,attrs:{shape:t}})}),a=e.map(e=>({vals:n.readSync(e.dataId),shape:e.shape}));s=Gf.computeOutShape(e.map(e=>e.shape),1);let l=1===e[0].shape[0],u=KE(a,s,t[0].dtype,l),h=Gf.computeOutShape(i.map(e=>e.shape),r);return o.shape=h,n.dataIdMap.get(o.dataId).stringBytes=Gf.fromStringArrayToUint8(u),e.forEach(e=>n.disposeData(e.dataId)),o}let l=va.sizeFromShape(i[0].shape.slice(0,r)),u=0,h=i.map(e=>{let t=va.sizeFromShape(e.shape.slice(r));return u+=t,t}),d=i.map(e=>n.typedArrayFromHeap(e)),p=n.typedArrayFromHeap(o);for(let e=0;e<l;e++){let t=e*u;for(let n=0;n<d.length;n++){let r=h[n],a=e*r,s=d[n].subarray(a,a+r);p.set(s,t),t+=r}}return o}var bq,xq={kernelName:yt,backendName:"wasm",kernelFunc:yq};var vq,wq={kernelName:bt,backendName:"wasm",setupFunc:function(e){bq=e.wasm.cwrap(bt,null,["number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number"])},kernelFunc:function(e){let{inputs:t,attrs:n,backend:r}=e,{x:a,filter:s}=t,i=r.dataIdMap.get(a.dataId).id,o=r.dataIdMap.get(s.dataId).id,{strides:l,dilations:u,pad:h,dimRoundingMode:d,dataFormat:p}=n,c=Gf.convertConv2DDataFormat(p),f=Gf.computeConv2DInfo(a.shape,s.shape,l,u,h,d,!1,c),m=f.filterHeight,g=f.filterWidth,y=f.padInfo.top,b=f.padInfo.right,x=f.padInfo.bottom,v=f.padInfo.left,w=f.dilationHeight,k=f.dilationWidth,I=f.strideHeight,S=f.strideWidth,_=f.inChannels,N=f.outChannels,T="SAME"===f.padInfo.type?1:0;if("channelsLast"!==f.dataFormat)throw new Error(`wasm backend Conv2D does not support dataFormat:'${f.dataFormat}'. Please use 'channelsLast'.`);let C=r.makeOutput(f.outShape,"float32"),E=r.dataIdMap.get(C.dataId).id;return bq(i,a.shape[0],a.shape[1],a.shape[2],o,m,g,y,b,x,v,T,w,k,I,S,_,N,E),C}};var kq,Iq={kernelName:vt,backendName:"wasm",setupFunc:function(e){vq=e.wasm.cwrap(vt,null,["number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number"])},kernelFunc:function(e){let{backend:t,inputs:n,attrs:r}=e,{dy:a,filter:s}=n,{strides:i,pad:o,dataFormat:l,dimRoundingMode:u,inputShape:h}=r,d=Gf.convertConv2DDataFormat(l),p=Gf.computeConv2DInfo(h,s.shape,i,1,o,u,!1,d),{batchSize:c,filterHeight:f,filterWidth:m,inChannels:g,inHeight:y,inWidth:b,outChannels:x,outHeight:v,outWidth:w,strideHeight:k,strideWidth:I}=p,S=f-1-p.padInfo.top,_=m-1-p.padInfo.left,N="channelsLast"===p.dataFormat,T=va.computeStrides(p.inShape),C=va.computeStrides(a.shape),[E,A,$]=va.computeStrides(s.shape),R=T[0],F=N?T[1]:T[2],D=N?T[2]:1,M=N?1:T[1],O=C[0],L=N?C[1]:C[2],P=N?C[2]:1,z=N?1:C[1],B=t.makeOutput(p.inShape,"float32"),W=t.dataIdMap.get(B.dataId).id,V=t.dataIdMap.get(a.dataId).id,U=t.dataIdMap.get(s.dataId).id;return vq(V,U,c,f,m,y,b,g,v,w,x,k,I,S,_,E,A,$,R,F,D,M,O,L,P,z,W),B}};var Sq,_q={kernelName:wt,backendName:"wasm",setupFunc:function(e){kq=e.wasm.cwrap(wt,null,["number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number"])},kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a,filter:s}=t,{strides:i,pad:o,dilations:l}=r;if("float32"!==a.dtype)throw new Error(`Tensor x must have dtype float32, got ${a.dtype}`);if("float32"!==s.dtype)throw new Error(`Tensor filter must have dtype float32, got ${s.dtype}`);let u=Gf.computeConv3DInfo(a.shape,s.shape,i,l,o),h=n.makeOutput(u.outShape,a.dtype);return kq(n.dataIdMap.get(a.dataId).id,n.dataIdMap.get(s.dataId).id,n.dataIdMap.get(h.dataId).id,u.batchSize,u.inDepth,u.inHeight,u.inWidth,u.inChannels,u.outDepth,u.outHeight,u.outWidth,u.outChannels,u.strideDepth,u.strideHeight,u.strideWidth,u.dilationDepth,u.dilationHeight,u.dilationWidth,u.filterDepth,u.filterHeight,u.filterWidth,u.padInfo.front,u.padInfo.top,u.padInfo.left),h}};var Nq,Tq={kernelName:kt,backendName:"wasm",setupFunc:function(e){Sq=e.wasm.cwrap(kt,null,["number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number"])},kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a,dy:s}=t,{strides:i,pad:o,filterShape:l}=r;if("float32"!==a.dtype)throw new Error(`Tensor dy must have dtype float32, got ${a.dtype}`);if("float32"!==s.dtype)throw new Error(`Tensor filter must have dtype float32, got ${s.dtype}`);let u=Gf.computeConv3DInfo(a.shape,l,i,1,o),h=n.makeOutput(u.filterShape,s.dtype);return Sq(n.dataIdMap.get(a.dataId).id,n.dataIdMap.get(s.dataId).id,n.dataIdMap.get(h.dataId).id,u.batchSize,u.inDepth,u.inHeight,u.inWidth,u.inChannels,u.outDepth,u.outHeight,u.outWidth,u.outChannels,u.strideDepth,u.strideHeight,u.strideWidth,u.dilationDepth,u.dilationHeight,u.dilationWidth,u.filterDepth,u.filterHeight,u.filterWidth,u.padInfo.front,u.padInfo.top,u.padInfo.left),h}};var Cq,Eq,Aq={kernelName:It,backendName:"wasm",setupFunc:function(e){Nq=e.wasm.cwrap(It,null,["number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number"])},kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{dy:a,filter:s}=t,{pad:i,strides:o,inputShape:l}=r;if("float32"!==a.dtype)throw new Error(`Tensor dy must have dtype float32, got ${a.dtype}`);if("float32"!==s.dtype)throw new Error(`Tensor filter must have dtype float32, got ${s.dtype}`);let u=Gf.computeConv3DInfo(l,s.shape,o,1,i),h=n.makeOutput(u.inShape,a.dtype);return Nq(n.dataIdMap.get(s.dataId).id,n.dataIdMap.get(a.dataId).id,n.dataIdMap.get(h.dataId).id,u.batchSize,u.inDepth,u.inHeight,u.inWidth,u.inChannels,u.outDepth,u.outHeight,u.outWidth,u.outChannels,u.strideDepth,u.strideHeight,u.strideWidth,u.dilationDepth,u.dilationHeight,u.dilationWidth,u.filterDepth,u.filterHeight,u.filterWidth,u.padInfo.front,u.padInfo.top,u.padInfo.left),h}},$q=wj(St),Rq=wj(_t);!function(e){e[e.bilinear=0]="bilinear",e[e.nearest=1]="nearest"}(Cq||(Cq={}));var Fq,Dq={kernelName:Ct,backendName:"wasm",setupFunc:function(e){Eq=e.wasm.cwrap(Ct,null,["number","number","number","number","array","number","number","number","number","number"])},kernelFunc:function(e){let t,{backend:n,inputs:r,attrs:a}=e,{method:s,extrapolationValue:i,cropSize:o}=a,{image:l,boxes:u,boxInd:h}=r,d=u.shape[0],[p,c]=o,f=[d,p,c,l.shape[3]],m=n.dataIdMap.get(l.dataId);"float32"!==l.dtype&&(t=pq({backend:n,inputs:{x:l},attrs:{dtype:"float32"}}),m=n.dataIdMap.get(t.dataId));let g=m.id,y=n.dataIdMap.get(u.dataId).id,b=n.dataIdMap.get(h.dataId).id,x=n.makeOutput(f,"float32"),v=n.dataIdMap.get(x.dataId).id,w=new Uint8Array(new Int32Array(l.shape).buffer);return Eq(g,y,b,d,w,p,c,Cq[s],i,v),null!=t&&n.disposeData(t.dataId),x}};var Mq,Oq={kernelName:Nt,backendName:"wasm",setupFunc:function(e){Fq=e.wasm.cwrap(Nt,null,["number","number","number","number","number","number"])},kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{axis:s,exclusive:i,reverse:o}=r,l=a.shape.length;va.assert("float32"===a.dtype||"int32"===a.dtype,()=>`cumprod does not support ${a.dtype} tensors in the WASM backend`);let u=Gf.getAxesPermutation([s],l),h=a;null!==u&&(h=Rj({inputs:{x:a},attrs:{perm:u},backend:n}));let d=Gf.getInnerMostAxes(1,l)[0];Gf.assertAxesAreInnerMostDims("cumprod",[d],l);let p=n.makeOutput(h.shape,h.dtype),c=h.shape[d],f=n.dataIdMap.get(h.dataId).id,m=n.dataIdMap.get(p.dataId).id;Fq(f,i?1:0,o?1:0,c,m,mj[a.dtype]);let g=p;if(null!==u){g=Rj({inputs:{x:p},attrs:{perm:Gf.getUndoAxesPermutation(u)},backend:n}),n.disposeData(h.dataId),n.disposeData(p.dataId)}return g}};var Lq,Pq={kernelName:Tt,backendName:"wasm",setupFunc:function(e){Mq=e.wasm.cwrap(Tt,null,["number","number","number","number","number","number"])},kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{axis:s,exclusive:i,reverse:o}=r,l=a.shape.length;va.assert("float32"===a.dtype||"int32"===a.dtype,()=>`cumsum does not support ${a.dtype} tensors in the WASM backend`);let u=Gf.getAxesPermutation([s],l),h=a;null!==u&&(h=Rj({inputs:{x:a},attrs:{perm:u},backend:n}));let d=Gf.getInnerMostAxes(1,l)[0];Gf.assertAxesAreInnerMostDims("cumsum",[d],l);let p=n.makeOutput(h.shape,h.dtype),c=h.shape[d],f=n.dataIdMap.get(h.dataId).id,m=n.dataIdMap.get(p.dataId).id;Mq(f,i?1:0,o?1:0,c,m,mj[a.dtype]);let g=p;if(null!==u){g=Rj({inputs:{x:p},attrs:{perm:Gf.getUndoAxesPermutation(u)},backend:n}),n.disposeData(h.dataId),n.disposeData(p.dataId)}return g}};var zq,Bq={kernelName:Et,backendName:"wasm",setupFunc:function(e){Lq=e.wasm.cwrap("DenseBincount",null,["number","array","number","number","boolean","number","number","boolean","number"])},kernelFunc:function(e){let{backend:t,inputs:n,attrs:r}=e,{x:a,weights:s}=n,{size:i,binaryOutput:o}=r,l=0!==s.shape.reduce((e,t)=>e*t,1),u=1===a.shape.length?[i]:[a.shape[0],i],h=t.makeOutput(u,s.dtype);function d(e){return t.dataIdMap.get(e.dataId).id}return Lq(d(a),new Uint8Array(new Int32Array(a.shape).buffer),a.shape.length,i,l,d(s),mj[s.dtype],o,d(h)),h}};var Wq,Vq={kernelName:At,backendName:"wasm",setupFunc:function(e){zq=e.wasm.cwrap(At,null,["number","number","number","array","number","array","array","number","number"])},kernelFunc:function(e){let{backend:t,inputs:n,attrs:r}=e,{x:a}=n,{blockSize:s,dataFormat:i}=r,o=a.shape[0],l=("NHWC"===i?a.shape[1]:a.shape[2])*s,u=("NHWC"===i?a.shape[2]:a.shape[3])*s,h=("NHWC"===i?a.shape[3]:a.shape[1])/(s*s),d="NHWC"===i?[o,l,u,h]:[o,h,l,u],p=t.makeOutput(d,"float32"),c=t.dataIdMap.get(a.dataId).id,f=new Uint8Array(new Int32Array(va.computeStrides(a.shape)).buffer),m=new Uint8Array(new Int32Array(d).buffer),g=new Uint8Array(new Int32Array(va.computeStrides(d)).buffer),y=t.dataIdMap.get(p.dataId).id;return zq(c,s,"NHWC"===i?1:0,f,a.shape.length-1,m,g,d.length,y),p}};var Uq,Gq={kernelName:$t,backendName:"wasm",setupFunc:function(e){Wq=e.wasm.cwrap($t,null,["number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number"])},kernelFunc:function(e){let{inputs:t,attrs:n,backend:r}=e,{x:a,filter:s}=t,i=r.dataIdMap.get(a.dataId).id,o=r.dataIdMap.get(s.dataId).id,{strides:l,dilations:u,pad:h,dimRoundingMode:d}=n,p=u??[1,1],c=Gf.computeConv2DInfo(a.shape,s.shape,l,p,h,d,!0),f=c.filterHeight,m=c.filterWidth,g=c.padInfo.top,y=c.padInfo.right,b=c.padInfo.bottom,x=c.padInfo.left,v=c.dilationHeight,w=c.dilationWidth,k=c.strideHeight,I=c.strideWidth,S=c.inChannels,_=c.outChannels,N="SAME"===c.padInfo.type?1:0;if("channelsLast"!==c.dataFormat)throw new Error(`wasm backend DepthwiseConv2dNative does not support dataFormat:'${c.dataFormat}'. Please use 'channelsLast'.`);let T=r.makeOutput(c.outShape,"float32"),C=r.dataIdMap.get(T.dataId).id;return Wq(i,a.shape[0],a.shape[1],a.shape[2],o,f,m,g,y,b,x,N,v,w,k,I,S,_,C),T}};var Hq,jq={kernelName:Dt,backendName:"wasm",setupFunc:function(e){Uq=e.wasm.cwrap("Diag",null,["number","number","number","number"])},kernelFunc:function(e){let{inputs:t,backend:n}=e,{x:r}=t,a=va.sizeFromShape(r.shape),s=n.makeOutput([...r.shape,...r.shape],r.dtype);return Uq(n.dataIdMap.get(r.dataId).id,mj[r.dtype],a,n.dataIdMap.get(s.dataId).id),s}};var qq,Kq={kernelName:Mt,backendName:"wasm",setupFunc:function(e){Hq=e.wasm.cwrap(Mt,null,["number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number"])},kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a,filter:s}=t,{strides:i,pad:o,dilations:l}=r;if(a.dtype!==s.dtype)throw new Error(`Dilation2D error: x must have the same dtype as filter. Got ${a.dtype} and ${s.dtype}`);let u=Gf.computeDilation2DInfo(a.shape,s.shape,i,o,"NHWC",l),h=n.makeOutput(u.outShape,a.dtype);return Hq(n.dataIdMap.get(a.dataId).id,n.dataIdMap.get(s.dataId).id,n.dataIdMap.get(h.dataId).id,mj[a.dtype],u.batchSize,u.inChannels,u.inHeight,u.inWidth,u.outHeight,u.outWidth,u.strideHeight,u.strideWidth,u.dilationHeight,u.dilationWidth,u.filterHeight,u.filterWidth,u.padInfo.top,u.padInfo.left),h}};var Xq,Zq={kernelName:Lt,backendName:"wasm",setupFunc:function(e){qq=e.wasm.cwrap(Lt,null,["number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number"])},kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a,filter:s,dy:i}=t,{strides:o,pad:l,dilations:u}=r;if(a.dtype!==s.dtype||a.dtype!==i.dtype)throw new Error(`Dilation2DBackpropFilter error: x must have the same dtype as filter and dy. Got ${a.dtype}, ${s.dtype}, and ${i.dtype}`);let h=Gf.computeDilation2DInfo(a.shape,s.shape,o,l,"NHWC",u),d=n.makeOutput(s.shape,s.dtype);return qq(n.dataIdMap.get(a.dataId).id,n.dataIdMap.get(s.dataId).id,n.dataIdMap.get(i.dataId).id,n.dataIdMap.get(d.dataId).id,mj[a.dtype],h.batchSize,h.inChannels,h.inHeight,h.inWidth,h.outHeight,h.outWidth,h.strideHeight,h.strideWidth,h.dilationHeight,h.dilationWidth,h.filterHeight,h.filterWidth,h.padInfo.top,h.padInfo.left),d}};var Yq,Jq={kernelName:Ot,backendName:"wasm",setupFunc:function(e){Xq=e.wasm.cwrap(Ot,null,["number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number"])},kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a,filter:s,dy:i}=t,{strides:o,pad:l,dilations:u}=r;if(a.dtype!==s.dtype||a.dtype!==i.dtype)throw new Error(`Dilation2DBackpropInput error: x must have the same dtype as filter and dy. Got ${a.dtype}, ${s.dtype}, and ${i.dtype}`);let h=Gf.computeDilation2DInfo(a.shape,s.shape,o,l,"NHWC",u),d=n.makeOutput(a.shape,a.dtype);return Xq(n.dataIdMap.get(a.dataId).id,n.dataIdMap.get(s.dataId).id,n.dataIdMap.get(i.dataId).id,n.dataIdMap.get(d.dataId).id,mj[a.dtype],h.batchSize,h.inChannels,h.inHeight,h.inWidth,h.outHeight,h.outWidth,h.strideHeight,h.strideWidth,h.dilationHeight,h.dilationWidth,h.filterHeight,h.filterWidth,h.padInfo.top,h.padInfo.left),d}},Qq=wj(Wt);var eK={kernelName:Vt,backendName:"wasm",setupFunc:function(e){Yq=e.wasm.cwrap(Vt,null,["number","number","number"])},kernelFunc:function(e){let{inputs:t,backend:n}=e,{dy:r,y:a}=t,s=n.makeOutput(a.shape,"float32"),i=e=>n.dataIdMap.get(e.dataId).id;return Yq(i(a),i(r),i(s)),s}},tK=_j(Gt,0,"bool"),nK=wj(Ut),rK=wj(Ht,"float32");function aK(e){let{inputs:t,attrs:n,backend:r}=e,{input:a}=t,{dim:s}=n,i=a.shape.length,o=a.shape.slice(),l=s;return s<0&&(va.assert(-(i+1)<=s,()=>`Axis must be in the interval [${-(i+1)}, ${i}]`),l=i+s+1),o.splice(l,0,1),tq({inputs:{x:a},backend:r,attrs:{shape:o}})}var sK={kernelName:jt,backendName:"wasm",kernelFunc:aK},iK=wj(qt,"float32");function oK(e){let{attrs:{shape:t,value:n},backend:r}=e,{attrs:{dtype:a}}=e;a=a||va.inferDtype(n);let s=r.makeOutput(t,a);return r.typedArrayFromHeap(s).fill(n),s}var lK,uK={kernelName:Xt,backendName:"wasm",kernelFunc:oK};var hK,dK={kernelName:Zt,backendName:"wasm",kernelFunc:function(e){let{inputs:t,backend:n}=e,{image:r}=t,a=n.makeOutput(r.shape,r.dtype),s=n.dataIdMap.get(r.dataId).id,i=n.dataIdMap.get(a.dataId).id,[o,l,u,h]=r.shape;return lK(s,o,l,u,h,i),a},setupFunc:function(e){lK=e.wasm.cwrap(Zt,null,["number","number","number","number","number","number"])}},pK=wj(Yt),cK=_j(Jt);var fK,mK={kernelName:Qt,backendName:"wasm",setupFunc:function(e){hK=e.wasm.cwrap(Qt,null,["number","number","number","number","number","number","number"])},kernelFunc:function(e){let{backend:t,inputs:n,attrs:r}=e,{varianceEpsilon:a}=r,{x:s,mean:i,variance:o,offset:l,scale:u}=n,h=t.dataIdMap.get(s.dataId).id,d=t.dataIdMap.get(i.dataId).id,p=t.dataIdMap.get(o.dataId).id,c=null!=l?t.dataIdMap.get(l.dataId).id:0,f=null!=u?t.dataIdMap.get(u.dataId).id:0,m=t.makeOutput(s.shape,s.dtype);if(0===va.sizeFromShape(s.shape))return m;let g=t.dataIdMap.get(m.dataId).id;return hK(h,d,p,c,f,a,g),m}};var gK,yK={kernelName:sa,backendName:"wasm",setupFunc:function(e){fK=e.wasm.cwrap(sa,null,["number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number"])},kernelFunc:function(e){let{inputs:t,attrs:n,backend:r}=e,{x:a,filter:s,bias:i,preluActivationWeights:o}=t,{strides:l,pad:u,dilations:h,dataFormat:d,dimRoundingMode:p,activation:c,leakyreluAlpha:f}=n,m=Gf.computeConv2DInfo(a.shape,s.shape,l,h,u,p),g=gj[c];if(null==g)throw new Error(`${c} activation not yet supported for FusedConv2D in the wasm backend.`);let y=r.dataIdMap.get(a.dataId).id,b=r.dataIdMap.get(s.dataId).id,x=m.outChannels,v=0;if(null!=i){let e=r.dataIdMap.get(i.dataId);if(1!==e.shape.length)throw new Error(`FusedConv2D only supports rank-1 bias but got rank ${e.shape.length}.`);if(e.shape[0]!==x)throw new Error(`FusedConv2D bias shape (${e.shape}) does not match the number of output channels (${x})`);v=e.id}let w=m.filterHeight,k=m.filterWidth,I=m.padInfo.top,S=m.padInfo.right,_=m.padInfo.bottom,N=m.padInfo.left,T=m.dilationHeight,C=m.dilationWidth,E=m.strideHeight,A=m.strideWidth,$=m.inChannels,R="SAME"===m.padInfo.type?1:0,F=m.batchSize,D=m.inHeight,M=m.inWidth;if("NHWC"!==d)throw new Error(`wasm backend FusedConv2D does not support dataFormat:'${d}'. Please use 'NHWC'.`);let O=r.makeOutput(m.outShape,"float32"),L=r.dataIdMap.get(O.dataId).id,P=null==o?0:r.dataIdMap.get(o.dataId).id;return fK(y,F,D,M,b,w,k,v,I,S,_,N,R,T,C,E,A,$,x,g,P,f||0,L),O}};var bK,xK={kernelName:ia,backendName:"wasm",setupFunc:function(e){gK=e.wasm.cwrap(ia,null,["number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number"])},kernelFunc:function(e){let{inputs:t,attrs:n,backend:r}=e,{x:a,filter:s,bias:i,preluActivationWeights:o}=t,{strides:l,pad:u,dilations:h,dataFormat:d,dimRoundingMode:p,activation:c,leakyreluAlpha:f}=n,m=Gf.computeConv2DInfo(a.shape,s.shape,l,h,u,p,!0),g=gj[c];if(null==g)throw new Error(`${c} activation not yet supported for FusedDepthwiseConv2D in the wasm backend.`);let y=r.dataIdMap.get(a.dataId).id,b=r.dataIdMap.get(s.dataId).id,x=m.outChannels,v=0;if(null!=i){let e=r.dataIdMap.get(i.dataId);if(1!==e.shape.length)throw new Error(`FusedDepthwiseConv2D only supports rank-1 bias but got rank ${e.shape.length}.`);if(e.shape[0]!==x)throw new Error(`FusedDepthwiseConv2D bias shape (${e.shape}) does not match the number of output channels (${x})`);v=e.id}let w=m.filterHeight,k=m.filterWidth,I=m.padInfo.top,S=m.padInfo.right,_=m.padInfo.bottom,N=m.padInfo.left,T=m.dilationHeight,C=m.dilationWidth,E=m.strideHeight,A=m.strideWidth,$=m.inChannels,R="SAME"===m.padInfo.type?1:0,F=m.batchSize,D=m.inHeight,M=m.inWidth;if("NHWC"!==d)throw new Error(`wasm backend FusedDepthwiseConv2D does not support dataFormat:'${d}'. Please use 'NHWC'.`);let O=r.makeOutput(m.outShape,"float32"),L=r.dataIdMap.get(O.dataId).id,P=null==o?0:r.dataIdMap.get(o.dataId).id;return gK(y,F,D,M,b,w,k,v,I,S,_,N,R,T,C,E,A,$,x,g,P,f||0,L),O}};var vK,wK={kernelName:tn,backendName:"wasm",setupFunc:function(e){bK=e.wasm.cwrap(tn,null,["number","number","number","number","number","number","array","number"])},kernelFunc:function(e){let{backend:t,inputs:n}=e,{params:r,indices:a}=n,[s,i,o,l]=bf.prepareAndValidate(r,a),u=t.makeOutput(s,r.dtype);if(0===i)return u;let h=a.shape,d=h[h.length-1],p=t.dataIdMap.get(r.dataId).id,c=t.dataIdMap.get(a.dataId).id,f=new Uint8Array(new Int32Array(l).buffer),m=t.dataIdMap.get(u.dataId).id;return bK(p,mj[r.dtype],c,i,d,o,f,m),u}};var kK,IK={kernelName:en,backendName:"wasm",setupFunc:function(e){vK=e.wasm.cwrap("Gather",null,["number","number","array","number","number","number","array","number"])},kernelFunc:function(e){let{backend:t,inputs:n,attrs:r}=e,{x:a,indices:s}=n,{axis:i,batchDims:o}=r,l=va.parseAxisParam(i,a.shape)[0],u=t.readSync(s.dataId),h=a.shape[l];for(let e=0;e<u.length;++e){let t=u[e];va.assert(t<=h-1&&t>=0,()=>`GatherV2: the index value ${t} is not in [0, ${h-1}]`)}let d=Gf.segment_util.collectGatherOpShapeInfo(a,s,l,o),p=tq({inputs:{x:a},attrs:{shape:[d.batchSize,d.outerSize,d.dimSize,d.sliceSize]},backend:t}),c=va.sizeFromShape(s.shape),f=tq({inputs:{x:s},attrs:{shape:[d.batchSize,c/d.batchSize]},backend:t}),m=[d.batchSize,d.outerSize,c/d.batchSize,d.sliceSize],g=t.makeOutput(m,a.dtype);if(0===va.sizeFromShape(a.shape))return g;let y=p.shape.length-1,b=t.dataIdMap.get(p.dataId).id,x=t.dataIdMap.get(f.dataId).id,v=t.dataIdMap.get(g.dataId).id,w=new Uint8Array(new Int32Array(va.computeStrides(p.shape)).buffer),k=new Uint8Array(new Int32Array(va.computeStrides(m)).buffer);return vK(b,mj[a.dtype],w,y,x,d.batchSize,k,v),t.disposeData(p.dataId),t.disposeData(f.dataId),g.shape=d.outputShape,g}},SK=_j(nn,0,"bool"),_K=_j(rn,0,"bool"),NK=wj(ln,"bool"),TK=wj(un,"bool"),CK=wj(hn,"bool");var EK,AK={kernelName:dn,backendName:"wasm",setupFunc:function(e){kK=e.wasm.cwrap(dn,null,["number","number","number","number"])},kernelFunc:function(e){let{inputs:{x:t},attrs:{alpha:n},backend:r}=e,a=r.dataIdMap.get(t.dataId).id,s=r.makeOutput(t.shape,"float32");if(0!==va.sizeFromShape(t.shape)){let e=r.dataIdMap.get(s.dataId).id;kK(a,mj[t.dtype],n,e)}return s}},$K=_j(pn,0,"bool"),RK=_j(cn,0,"bool");var FK,DK={kernelName:fn,backendName:"wasm",setupFunc:function(e){EK=e.wasm.cwrap(fn,null,["number","number","number","number"])},kernelFunc:function(e){let{attrs:t,backend:n}=e,{start:r,stop:a,num:s}=t,i=Math.floor(s),o=n.makeOutput([i],"float32");return EK(n.dataIdMap.get(o.dataId).id,r,a,i),o}},MK=wj(mn),OK=wj(gn),LK=_j(yn,0,"bool"),PK=wj(bn),zK=_j(xn,0,"bool"),BK=_j(vn,0,"bool");var WK,VK={kernelName:In,backendName:"wasm",setupFunc:function(e){FK=e.wasm.cwrap(In,null,["number","number","number","number","number","number","number"])},kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{depthRadius:s,bias:i,alpha:o,beta:l}=r;if("float32"!==a.dtype)throw new Error("LRN error: x must have dtype float32");let u=n.makeOutput(a.shape,a.dtype);return FK(n.dataIdMap.get(a.dataId).id,n.dataIdMap.get(u.dataId).id,a.shape[3],s,i,o,l),u}};var UK,GK={kernelName:Sn,backendName:"wasm",setupFunc:function(e){WK=e.wasm.cwrap(Sn,null,["number","number","number","number","number","number","number","number","number"])},kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a,y:s,dy:i}=t,{depthRadius:o,bias:l,alpha:u,beta:h}=r;if("float32"!==a.dtype||"float32"!==s.dtype||"float32"!==i.dtype)throw new Error("LRNGrad error: x, y, and dy must have dtype float32");let d=n.makeOutput(a.shape,a.dtype);return WK(n.dataIdMap.get(a.dataId).id,n.dataIdMap.get(s.dataId).id,n.dataIdMap.get(i.dataId).id,n.dataIdMap.get(d.dataId).id,i.shape[3],o,l,u,h),d}};var HK,jK={kernelName:Nn,backendName:"wasm",setupFunc:function(e){UK=e.wasm.cwrap(Nn,null,["number","number","number","number"])},kernelFunc:function(e){let{backend:t,inputs:n,attrs:r}=e,{reductionIndices:a,keepDims:s}=r,{x:i}=n,o=t.dataIdMap.get(i.dataId).id,l=i,{transposed:u,axes:h,originalAxes:d,inputWasTransposed:p}=Mj(i,a,t);if(p){l=u,o=t.dataIdMap.get(u.dataId).id}let c=l.shape.length;Gf.assertAxesAreInnerMostDims("max",h,c);let[f,m]=Gf.computeOutAndReduceShapes(l.shape,h),g=va.sizeFromShape(m),y=t.makeOutput(f,i.dtype);if(0!==va.sizeFromShape(l.shape)){let e=t.dataIdMap.get(y.dataId).id;UK(o,mj[i.dtype],g,e)}if(p&&t.disposeData(u.dataId),s){let e=Gf.expandShapeToKeepDim(y.shape,d);y.shape=e}return y}},qK=_j(Tn);var KK,XK={kernelName:Cn,backendName:"wasm",setupFunc:function(e){HK=e.wasm.cwrap(Cn,null,["number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number"])},kernelFunc:function(e){let{inputs:t,attrs:n,backend:r}=e,a=t.x,s=r.dataIdMap.get(a.dataId).id;va.assert("float32"===a.dtype,()=>`Error in MaxPool: only float32 input is supported. Got ${a.dtype}.`);let{filterSize:i,strides:o,pad:l,dimRoundingMode:u}=n,h=Gf.computePool2DInfo(a.shape,i,o,1,l,u),d=h.filterHeight,p=h.filterWidth,c=h.padInfo.top,f=h.padInfo.right,m=h.padInfo.bottom,g=h.padInfo.left,y=h.dilationHeight,b=h.dilationWidth,x=h.strideHeight,v=h.strideWidth,w=h.inChannels,k=h.outChannels;if("channelsLast"!==h.dataFormat)throw new Error(`wasm backend does not support dataFormat:'${h.dataFormat}'. Please use 'channelsLast'.`);let I=r.makeOutput(h.outShape,"float32"),S=r.dataIdMap.get(I.dataId).id;return HK(s,a.shape[0],a.shape[1],a.shape[2],d,p,c,f,m,g,y,b,x,v,w,k,S),I}};var ZK,YK={kernelName:An,backendName:"wasm",setupFunc:function(e){KK=e.wasm.cwrap("MaxPool3D",null,["number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number"])},kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{filterSize:s,strides:i,pad:o,dimRoundingMode:l,dataFormat:u}=r,h=Gf.computePool3DInfo(a.shape,s,i,1,o,l,u),d=n.makeOutput(h.outShape,a.dtype);return KK(n.dataIdMap.get(a.dataId).id,n.dataIdMap.get(d.dataId).id,h.batchSize,h.inChannels,h.inDepth,h.inHeight,h.inWidth,h.outDepth,h.outHeight,h.outWidth,h.strideDepth,h.strideHeight,h.strideWidth,h.dilationDepth,h.dilationHeight,h.dilationWidth,h.effectiveFilterDepth,h.effectiveFilterHeight,h.effectiveFilterWidth,h.padInfo.front,h.padInfo.top,h.padInfo.left),d}};var JK,QK={kernelName:$n,backendName:"wasm",setupFunc:function(e){ZK=e.wasm.cwrap("MaxPool3DGrad",null,["number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number"])},kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{dy:a,input:s}=t,{filterSize:i,strides:o,pad:l,dimRoundingMode:u}=r,h=Gf.computePool3DInfo(s.shape,i,o,1,l,u),d=n.makeOutput(s.shape,s.dtype);return ZK(n.dataIdMap.get(s.dataId).id,n.dataIdMap.get(a.dataId).id,n.dataIdMap.get(d.dataId).id,h.batchSize,h.inChannels,h.inDepth,h.inHeight,h.inWidth,h.outDepth,h.outHeight,h.outWidth,h.strideDepth,h.strideHeight,h.strideWidth,h.dilationDepth,h.dilationHeight,h.dilationWidth,h.effectiveFilterDepth,h.effectiveFilterHeight,h.effectiveFilterWidth,h.padInfo.front,h.padInfo.top,h.padInfo.left),d}};var eX,tX={kernelName:En,backendName:"wasm",setupFunc:function(e){JK=e.wasm.cwrap("MaxPoolGrad",null,["number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number","number"])},kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{dy:a,input:s}=t,{filterSize:i,strides:o,pad:l,dimRoundingMode:u}=r,h=Gf.computePool2DInfo(s.shape,i,o,1,l,u),d=n.makeOutput(s.shape,s.dtype);return JK(n.dataIdMap.get(s.dataId).id,n.dataIdMap.get(a.dataId).id,n.dataIdMap.get(d.dataId).id,h.batchSize,h.inChannels,h.inHeight,h.inWidth,h.outHeight,h.outWidth,h.strideHeight,h.strideWidth,h.dilationHeight,h.dilationWidth,h.effectiveFilterHeight,h.effectiveFilterWidth,h.padInfo.top,h.padInfo.left),d}};var nX,rX={kernelName:Rn,backendName:"wasm",setupFunc:function(e){eX=e.wasm.cwrap("MaxPoolWithArgmax",null,["number","number","number","number","boolean","number","number","number","number","number","number","number","number","number","number","number","number","number","number"])},kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{filterSize:s,strides:i,pad:o,includeBatchInIndex:l}=r;va.assert(4===a.shape.length,()=>`Error in maxPool: input must be rank 4 but got rank ${a.shape.length}.`);let u=[1,1];va.assert(Gf.eitherStridesOrDilationsAreOne(i,u),()=>`Error in maxPool: Either strides or dilations must be 1. Got strides ${i} and dilations '${u}'`);let h=Gf.computePool2DInfo(a.shape,s,i,[1,1],o),d=n.makeOutput(h.outShape,a.dtype),p=n.makeOutput(h.outShape,"int32");return eX(n.dataIdMap.get(a.dataId).id,n.dataIdMap.get(d.dataId).id,n.dataIdMap.get(p.dataId).id,mj[a.dtype],l,h.batchSize,h.inChannels,h.inHeight,h.inWidth,h.outHeight,h.outWidth,h.strideHeight,h.strideWidth,h.dilationHeight,h.dilationWidth,h.effectiveFilterHeight,h.effectiveFilterWidth,h.padInfo.top,h.padInfo.left),[d,p]}};var aX,sX={kernelName:Fn,backendName:"wasm",setupFunc:function(e){nX=e.wasm.cwrap(Fn,null,["number, number, number"])},kernelFunc:function(e){let{backend:t,inputs:n,attrs:r}=e,{axis:a,keepDims:s}=r,{x:i}=n,o=t.dataIdMap.get(i.dataId).id,l=o,u=i,{transposed:h,axes:d,originalAxes:p,inputWasTransposed:c}=Mj(i,a,t),f=d;if(c){let e=t.dataIdMap.get(h.dataId).id;e!==o&&(u=h,l=e,f=Gf.getInnerMostAxes(f.length,u.shape.length))}Gf.assertAxesAreInnerMostDims("mean",f,u.shape.length);let[m,g]=Gf.computeOutAndReduceShapes(u.shape,f),y=va.sizeFromShape(g),b=u;"float32"!==u.dtype&&(b=pq({backend:t,inputs:{x:u},attrs:{dtype:"float32"}}),l=t.dataIdMap.get(b.dataId).id);let x=t.makeOutput(m,"float32");if(0!==va.sizeFromShape(u.shape)){let e=t.dataIdMap.get(x.dataId).id;nX(l,y,e)}if(c&&t.disposeData(h.dataId),s){let e=Gf.expandShapeToKeepDim(x.shape,p);x.shape=e}return"float32"!==u.dtype&&t.disposeData(b.dataId),x}};var iX,oX,lX={kernelName:Dn,backendName:"wasm",setupFunc:function(e){aX=e.wasm.cwrap(Dn,null,["number","number","number","number"])},kernelFunc:function(e){let{backend:t,inputs:n,attrs:r}=e,{axis:a,keepDims:s}=r,{x:i}=n,o=t.dataIdMap.get(i.dataId).id,l=o,u=i,{transposed:h,axes:d,originalAxes:p,inputWasTransposed:c}=Mj(i,a,t);if(c){let e=t.dataIdMap.get(h.dataId).id;e!==o&&(u=h,l=e)}let f=u.shape.length;Gf.assertAxesAreInnerMostDims("min",d,f);let[m,g]=Gf.computeOutAndReduceShapes(u.shape,d),y=va.sizeFromShape(g),b=t.makeOutput(m,u.dtype);if(0!==va.sizeFromShape(u.shape)){let e=t.dataIdMap.get(b.dataId).id;aX(l,mj[i.dtype],y,e)}if(c&&t.disposeData(h.dataId),s){let e=Gf.expandShapeToKeepDim(b.shape,p);b.shape=e}return b}},uX=_j(Mn);!function(e){e[e.reflect=0]="reflect",e[e.symmetric=1]="symmetric"}(iX||(iX={}));var hX,dX={kernelName:On,backendName:"wasm",kernelFunc:function(e){let{inputs:{x:t},backend:n,attrs:{paddings:r,mode:a}}=e,s=r.map((e,n)=>e[0]+t.shape[n]+e[1]),i=n.dataIdMap.get(t.dataId).id,o=n.makeOutput(s,t.dtype),l=n.dataIdMap.get(o.dataId).id,u=new Uint8Array(new Int32Array(t.shape).buffer),h=r.map(e=>e[0]),d=r.map(e=>e[1]),p=new Uint8Array(new Int32Array(h).buffer),c=new Uint8Array(new Int32Array(d).buffer);return oX(i,u,t.shape.length,mj[t.dtype],p,c,iX[a],l),o},setupFunc:function(e){oX=e.wasm.cwrap(On,null,["number","array","number","number","array","array","number","number"])}};function pX(e){let{backend:t,inputs:{logits:n},attrs:{dim:r}}=e,a=t.dataIdMap.get(n.dataId).id,s=t.makeOutput(n.shape,n.dtype),i=t.dataIdMap.get(s.dataId).id,o=n.shape[r],l=va.sizeFromShape(n.shape)/o;return 0===va.sizeFromShape(s.shape)||hX(a,i,o,l),s}var cX,fX={kernelName:Ar,backendName:"wasm",setupFunc:function(e){hX=e.wasm.cwrap(Ar,null,["number","number","number","number"])},kernelFunc:pX};var mX,gX={kernelName:Pn,backendName:"wasm",setupFunc:function(e){cX=e.wasm.cwrap(Pn,null,["number","number","number","number","number","number"])},kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{logits:a}=t,{numSamples:s,seed:i,normalized:o}=r;if("float32"!==a.dtype)throw new Error(`Tensor logits must have dtype float32, got ${a.dtype}`);let l=o?a:pX({inputs:{logits:a},backend:n,attrs:{dim:a.shape.length-1}}),[u,h]=l.shape,d=n.makeOutput([u,s],"int32");return cX(n.dataIdMap.get(l.dataId).id,u,h,s,i,n.dataIdMap.get(d.dataId).id),o||n.disposeData(l.dataId),d}},yX=_j(Ln),bX=_j(zn),xX=wj(Bn);function vX(e,t){let n=new Int32Array(e.wasm.HEAPU8.buffer,t,4),r=n[0],a=n[1],s=n[2],i=n[3];return e.wasm._free(t),{pSelectedIndices:r,selectedSize:a,pSelectedScores:s,pValidOutputs:i}}var wX,kX={kernelName:Vn,backendName:"wasm",setupFunc:function(e){mX=e.wasm.cwrap(Vn,"number",["number","number","number","number","number"])},kernelFunc:function(e){let{backend:t,inputs:n,attrs:r}=e,{iouThreshold:a,maxOutputSize:s,scoreThreshold:i}=r,{boxes:o,scores:l}=n,u=t.dataIdMap.get(o.dataId).id,h=t.dataIdMap.get(l.dataId).id,d=mX(u,h,s,a,i),{pSelectedIndices:p,selectedSize:c,pSelectedScores:f,pValidOutputs:m}=vX(t,d);return t.wasm._free(f),t.wasm._free(m),t.makeOutput([c],"int32",p)}};var IX,SX={kernelName:Un,backendName:"wasm",setupFunc:function(e){wX=e.wasm.cwrap(Un,"number",["number","number","number","number","number","bool"])},kernelFunc:function(e){let{backend:t,inputs:n,attrs:r}=e,{iouThreshold:a,maxOutputSize:s,scoreThreshold:i,padToMaxOutputSize:o}=r,{boxes:l,scores:u}=n,h=t.dataIdMap.get(l.dataId).id,d=t.dataIdMap.get(u.dataId).id,p=wX(h,d,s,a,i,o),{pSelectedIndices:c,selectedSize:f,pSelectedScores:m,pValidOutputs:g}=vX(t,p);return t.wasm._free(m),[t.makeOutput([f],"int32",c),t.makeOutput([],"int32",g)]}};var _X,NX={kernelName:Gn,backendName:"wasm",setupFunc:function(e){IX=e.wasm.cwrap(Gn,"number",["number","number","number","number","number","number"])},kernelFunc:function(e){let{backend:t,inputs:n,attrs:r}=e,{iouThreshold:a,maxOutputSize:s,scoreThreshold:i,softNmsSigma:o}=r,{boxes:l,scores:u}=n,h=t.dataIdMap.get(l.dataId).id,d=t.dataIdMap.get(u.dataId).id,p=IX(h,d,s,a,i,o),{pSelectedIndices:c,selectedSize:f,pSelectedScores:m,pValidOutputs:g}=vX(t,p);return t.wasm._free(g),[t.makeOutput([f],"int32",c),t.makeOutput([f],"float32",m)]}},TX=_j(Wn,0,"bool");var CX={kernelName:jn,backendName:"wasm",setupFunc:function(e){_X=e.wasm.cwrap(jn,null,["number","number","number","number","number"])},kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{indices:a}=t,{dtype:s,depth:i,onValue:o,offValue:l}=r,u=n.makeOutput([...a.shape,i],s),h=n.dataIdMap.get(u.dataId).id,d=n.dataIdMap.get(a.dataId).id;return _X(d,i,o,l,h),u}};var EX={kernelName:Hn,backendName:"wasm",kernelFunc:function(e){let{inputs:{x:t},backend:n}=e,r=n.makeOutput(t.shape,t.dtype);return n.typedArrayFromHeap(r).fill(1),r}};var AX,$X={kernelName:qn,backendName:"wasm",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{axis:a}=r;if(1===t.length)return aK({inputs:{input:t[0]},backend:n,attrs:{dim:a}});let s=t[0].shape,i=t[0].dtype;t.forEach(e=>{va.assertShapesMatch(s,e.shape,"All tensors passed to stack must have matching shapes"),va.assert(i===e.dtype,()=>"All tensors passed to stack must have matching dtypes")});let o=[],l=yq({inputs:t.map(e=>{let t=aK({inputs:{input:e},backend:n,attrs:{dim:a}});return o.push(t),t}),backend:n,attrs:{axis:a}});return o.forEach(e=>n.disposeData(e.dataId)),l}};var RX,FX={kernelName:Kn,backendName:"wasm",kernelFunc:function(e){let{inputs:{x:t},backend:n,attrs:{paddings:r,constantValue:a}}=e,s=r.map((e,n)=>e[0]+t.shape[n]+e[1]);if(0===va.sizeFromShape(t.shape))return oK({backend:n,attrs:{shape:s,value:a,dtype:t.dtype}});let i=n.dataIdMap.get(t.dataId).id,o=n.makeOutput(s,t.dtype),l=n.dataIdMap.get(o.dataId).id,u=new Uint8Array(new Int32Array(t.shape).buffer),h=r.map(e=>e[0]),d=r.map(e=>e[1]),p=new Uint8Array(new Int32Array(h).buffer),c=new Uint8Array(new Int32Array(d).buffer);return AX(i,u,t.shape.length,mj[t.dtype],p,c,a,l),o},setupFunc:function(e){AX=e.wasm.cwrap(Kn,null,["number","array","number","number","array","array","number","number"])}},DX=_j(Zn);var MX,OX={kernelName:Yn,backendName:"wasm",setupFunc:function(e){RX=e.wasm.cwrap(Yn,null,["number","number","number"])},kernelFunc:function(e){let{inputs:t,backend:n}=e,{x:r,alpha:a}=t,s=n.dataIdMap.get(r.dataId).id,i=n.dataIdMap.get(a.dataId).id,o=s,l=r,u=l;"float32"!==l.dtype&&(u=pq({backend:n,inputs:{x:r},attrs:{dtype:"float32"}}),o=n.dataIdMap.get(u.dataId).id);let h=n.makeOutput(r.shape,"float32"),d=n.dataIdMap.get(h.dataId).id;return RX(o,i,d),"float32"!==l.dtype&&n.disposeData(u.dataId),h}};var LX,PX={kernelName:Jn,backendName:"wasm",setupFunc:function(e){MX=e.wasm.cwrap(Jn,null,["number","number","number","number"])},kernelFunc:function(e){let{backend:t,inputs:n,attrs:r}=e,{axis:a,keepDims:s}=r,{x:i}=n,o=t.dataIdMap.get(i.dataId).id,l=o,u=i,{transposed:h,axes:d,originalAxes:p,inputWasTransposed:c}=Mj(i,a,t),f=d;if(c){let e=t.dataIdMap.get(h.dataId).id;e!==o&&(u=h,l=e,f=Gf.getInnerMostAxes(f.length,u.shape.length))}Gf.assertAxesAreInnerMostDims("prod",f,u.shape.length);let[m,g]=Gf.computeOutAndReduceShapes(u.shape,f),y=va.sizeFromShape(g),b=t.makeOutput(m,u.dtype);if(0!==va.sizeFromShape(u.shape)){let e=t.dataIdMap.get(b.dataId).id;MX(l,y,mj[b.dtype],e)}if(c&&t.disposeData(h.dataId),s){let e=Gf.expandShapeToKeepDim(b.shape,p);b.shape=e}return b}},zX={kernelName:nr,backendName:"wasm",kernelFunc:e=>{let{backend:t,attrs:n}=e,{start:r,stop:a,step:s,dtype:i}=n,o=i$(r,a,s,i),l=t.makeOutput([o.length],i);return t.typedArrayFromHeap(l).set(o),l}},BX=_j(zt),WX=wj(ar),VX=wj(sr),UX=wj(dr);var GX,HX={kernelName:ur,backendName:"wasm",setupFunc:function(e){LX=e.wasm.cwrap(ur,null,["number","number","number","number","number","number","number","number","number","number"])},kernelFunc:function(e){let t,{backend:n,inputs:r,attrs:a}=e,{images:s}=r,{alignCorners:i,halfPixelCenters:o,size:l}=a,[u,h]=l,[d,p,c,f]=s.shape,m=[d,u,h,f],g=n.dataIdMap.get(s.dataId);"float32"!==g.dtype&&(t=pq({backend:n,inputs:{x:s},attrs:{dtype:"float32"}}),g=n.dataIdMap.get(t.dataId));let y=g.id,b=n.makeOutput(m,"float32");if(0===va.sizeFromShape(s.shape))return b;let x=n.dataIdMap.get(b.dataId).id;return LX(y,d,p,c,f,u,h,i?1:0,o?1:0,x),null!=t&&n.disposeData(t.dataId),b}};var jX,qX={kernelName:hr,backendName:"wasm",setupFunc:function(e){GX=e.wasm.cwrap(hr,null,["number","number","number","array","array","boolean"])},kernelFunc:function(e){let t,{inputs:n,backend:r,attrs:a}=e,{images:s,dy:i}=n,{alignCorners:o}=a,l=r.makeOutput(s.shape,"float32"),u=r.dataIdMap.get(s.dataId);return"float32"!==u.dtype&&(t=pq({backend:r,inputs:{x:s},attrs:{dtype:"float32"}}),u=r.dataIdMap.get(t.dataId)),GX(r.dataIdMap.get(s.dataId).id,r.dataIdMap.get(i.dataId).id,r.dataIdMap.get(l.dataId).id,new Uint8Array(new Int32Array(s.shape).buffer),new Uint8Array(new Int32Array(i.shape).buffer),o),null!=t&&r.disposeData(t.dataId),l}};var KX,XX={kernelName:or,backendName:"wasm",setupFunc:function(e){jX=e.wasm.cwrap(or,null,["number","number","number","number","number","number","number","number","number","number"])},kernelFunc:function(e){let{backend:t,inputs:n,attrs:r}=e,{images:a}=n,{alignCorners:s,halfPixelCenters:i,size:o}=r,[l,u]=o,[h,d,p,c]=a.shape,f=[h,l,u,c],m=t.makeOutput(f,"float32");if(0===va.sizeFromShape(a.shape))return m;let g,y=t.dataIdMap.get(a.dataId);"float32"!==y.dtype&&(g=pq({backend:t,inputs:{x:a},attrs:{dtype:"float32"}}),y=t.dataIdMap.get(g.dataId));let b=y.id,x=t.dataIdMap.get(m.dataId).id;return jX(b,h,d,p,c,l,u,s?1:0,i?1:0,x),null!=g&&t.disposeData(g.dataId),m}};var ZX,YX={kernelName:lr,backendName:"wasm",setupFunc:function(e){KX=e.wasm.cwrap(lr,null,["number","number","number","array","array","boolean"])},kernelFunc:function(e){let t,{inputs:n,backend:r,attrs:a}=e,{images:s,dy:i}=n,{alignCorners:o}=a,l=r.makeOutput(s.shape,"float32"),u=r.dataIdMap.get(s.dataId);return"float32"!==u.dtype&&(t=pq({backend:r,inputs:{x:s},attrs:{dtype:"float32"}}),u=r.dataIdMap.get(t.dataId)),KX(r.dataIdMap.get(s.dataId).id,r.dataIdMap.get(i.dataId).id,r.dataIdMap.get(l.dataId).id,new Uint8Array(new Int32Array(s.shape).buffer),new Uint8Array(new Int32Array(i.shape).buffer),o),null!=t&&r.disposeData(t.dataId),l}};var JX,QX={kernelName:pr,backendName:"wasm",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{dims:s}=r,i=va.parseAxisParam(s,a.shape);if(0===a.shape.length)return Ej({inputs:{x:a},backend:n});let o=n.makeOutput(a.shape,a.dtype),l=n.dataIdMap.get(a.dataId).id,u=n.dataIdMap.get(o.dataId).id,h=new Uint8Array(new Int32Array(i).buffer),d=new Uint8Array(new Int32Array(a.shape).buffer);ZX(l,h,i.length,d,a.shape.length,u);let p=tq({inputs:{x:o},attrs:{shape:a.shape},backend:n});return n.disposeData(o.dataId),p},setupFunc:function(e){ZX=e.wasm.cwrap(pr,null,["number","array","number","array","number","number"])}};var eZ,tZ={kernelName:ra,backendName:"wasm",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{image:a}=t,{radians:s,fillValue:i,center:o}=r,l=n.makeOutput(a.shape,a.dtype),u=n.dataIdMap.get(a.dataId).id,h=n.dataIdMap.get(l.dataId).id,[d,p,c,f]=a.shape,[m,g]=Gf.getImageCenter(o,p,c),y="number"==typeof i?[i,i,i,0===i?0:255]:[...i,255],b=new Uint8Array(new Int32Array(y).buffer);return JX(u,d,p,c,f,s,m,g,b,y.length,h),l},setupFunc:function(e){JX=e.wasm.cwrap(ra,null,["number","number","number","number","number","number","number","number","array","number","number"])}},nZ=wj(cr),rZ=wj(fr);var aZ,sZ={kernelName:mr,backendName:"wasm",setupFunc:function(e){eZ=e.wasm.cwrap(mr,null,["number","number","number","number","number","number","array","number","number"])},kernelFunc:function(e){let{backend:t,inputs:n,attrs:r}=e,{indices:a,updates:s}=n,{shape:i}=r,o=t.makeOutput(i,s.dtype);if(0===va.sizeFromShape(i))return o;let{sliceRank:l,numUpdates:u,sliceSize:h,strides:d,outputSize:p}=Ld.calculateShapes(s,a,i),c=t.dataIdMap.get(a.dataId).id,f=t.dataIdMap.get(s.dataId).id,m=new Uint8Array(new Int32Array(d).buffer),g=t.dataIdMap.get(o.dataId).id;return eZ(c,f,mj[s.dtype],l,u,h,m,p,g),o}};var iZ,oZ={kernelName:yr,backendName:"wasm",setupFunc:function(e){aZ=e.wasm.cwrap(yr,null,["number","number","number","number","number","number","bool","number"])},kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{sortedSequence:a,values:s}=t,{side:i}=r;if(a.dtype!==s.dtype)throw new Error(`SearchSorted error: sorted_sequence must have the same dtype as values. Got ${a.dtype} and ${s.dtype}`);let o=n.makeOutput(s.shape,"int32");function l(e){return n.dataIdMap.get(e.dataId).id}return aZ(l(a),l(s),a.shape[0],a.shape[1],s.shape[1],mj[a.dtype],"left"===i,l(o)),o}};var lZ,uZ={kernelName:br,backendName:"wasm",kernelFunc:function(e){let{inputs:t,backend:n}=e,{condition:r,t:a,e:s}=t,i=n.dataIdMap.get(r.dataId).id,o=n.dataIdMap.get(a.dataId).id,l=n.dataIdMap.get(s.dataId).id,u=n.makeOutput(a.shape,a.dtype),h=n.dataIdMap.get(u.dataId).id,d=r.shape.length,p=a.shape.length,c=0===d||d>1||1===p?1:va.sizeFromShape(a.shape.slice(1));return iZ(i,o,l,c,h),u},setupFunc:function(e){iZ=e.wasm.cwrap("SelectV2",null,["number","number","number","number","number"])}},hZ=wj(xr);var dZ={kernelName:"Sigmoid",backendName:"wasm",setupFunc:function(e){lZ=e.wasm.cwrap(Sr,null,["number","number"])},kernelFunc:function(e){let{backend:t,inputs:{x:n}}=e,r=t.dataIdMap.get(n.dataId).id,a=t.makeOutput(n.shape,n.dtype),s=t.dataIdMap.get(a.dataId).id;return 0===va.sizeFromShape(a.shape)||lZ(r,s),a}},pZ=wj(Ir),cZ=wj(wr),fZ=wj(kr),mZ=wj(_r);var gZ,yZ={kernelName:Cr,backendName:"wasm",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t,{blockShape:s,paddings:i}=r,o=va.sizeFromShape(s),l=[[0,0]];l.push(...i);for(let e=1+s.length;e<a.shape.length;++e)l.push([0,0]);let u=FX.kernelFunc({inputs:{x:a},backend:n,attrs:{paddings:l,constantValue:0}}),h=Gf.getReshaped(u.shape,s,o,!1),d=Gf.getPermuted(h.length,s.length,!1),p=Gf.getReshapedPermuted(u.shape,s,o,!1),c=tq({inputs:{x:u},backend:n,attrs:{shape:h}}),f=Rj({inputs:{x:c},backend:n,attrs:{perm:d}}),m=tq({inputs:{x:f},backend:n,attrs:{shape:p}});return n.disposeData(u.dataId),n.disposeData(c.dataId),n.disposeData(f.dataId),m}};var bZ,xZ={kernelName:$r,backendName:"wasm",setupFunc:function(e){gZ=e.wasm.cwrap("SparseFillEmptyRows","number",["number","number","number","number","number","number","number","number","number","number","number","number"])},kernelFunc:function(e){let t,{backend:n,inputs:r}=e,{indices:a,values:s,denseShape:i,defaultValue:o}=r,l=a.shape[0],u=a.shape[1],h=n.readSync(i.dataId)[0],d=[l+h,u],p=n.dataIdMap.get(a.dataId).id,c=n.dataIdMap.get(s.dataId).id,f=n.dataIdMap.get(o.dataId).id,m=n.makeOutput(d,a.dtype),g=n.dataIdMap.get(m.dataId).id,y=n.makeOutput(d.slice(0,1),s.dtype),b=n.dataIdMap.get(y.dataId).id,x=n.makeOutput([h],"bool"),v=n.dataIdMap.get(x.dataId).id,w=n.makeOutput([l],a.dtype),k=n.dataIdMap.get(w.dataId).id,I=n.makeOutput([4],"int32"),S=n.dataIdMap.get(I.dataId).id,_=gZ(p,c,mj[s.dtype],l,h,u,f,g,b,v,k,S),N=n.readSync(I.dataId);switch(N[0]){case 1:t=Gf.getSparseFillEmptyRowsIndicesDenseShapeMismatch(N[1]);break;case 2:t=Gf.getSparseFillEmptyRowsNegativeIndexErrorMessage(N[1],N[2]);break;case 3:t=Gf.getSparseFillEmptyRowsOutOfRangeIndexErrorMessage(N[1],N[2],N[3]);break;default:t=""}if(n.disposeData(I.dataId),t)throw n.disposeData(m.dataId),n.disposeData(y.dataId),n.disposeData(x.dataId),n.disposeData(w.dataId),new Error(t);let T=m,C=y;return _!==d[0]&&(T=sq({inputs:{x:m},attrs:{begin:0,size:[_,u]},backend:n}),C=sq({inputs:{x:y},attrs:{begin:0,size:_},backend:n}),n.disposeData(m.dataId),n.disposeData(y.dataId)),[T,C,x,w]}};var vZ,wZ={kernelName:Rr,backendName:"wasm",setupFunc:function(e){bZ=e.wasm.cwrap(Rr,null,["number","number","number","number","number","number","number"])},kernelFunc:function(e){let{backend:t,inputs:n}=e,{inputIndices:r,inputShape:a,newShape:s}=n;if(2!==r.shape.length)throw new Error(`Input indices should be a matrix but received shape\n        ${r.shape}`);if(1!==a.shape.length)throw new Error(`Input shape should be a vector but received shape\n        ${a.shape}`);if(1!==s.shape.length)throw new Error(`Target shape should be a vector but received shape ${s.shape}`);let i=t.dataIdMap.get(r.dataId).id,o=t.dataIdMap.get(a.dataId).id,l=t.dataIdMap.get(s.dataId).id,u=r.shape[0],h=va.sizeFromShape(s.shape),d=t.makeOutput([u,h],r.dtype),p=t.dataIdMap.get(d.dataId).id,c=t.makeOutput([h],s.dtype),f=t.dataIdMap.get(c.dataId).id,m=t.makeOutput([3],"int32"),g=t.dataIdMap.get(m.dataId).id;bZ(i,o,l,u,p,f,g);let y,b=t.readSync(m.dataId);switch(b[0]){case 0:y=Gf.getSparseReshapeMultipleNegativeOneOutputDimErrorMessage(b[1],b[2]);break;case 1:y=Gf.getSparseReshapeNegativeOutputDimErrorMessage(b[1],b[2]);break;case 2:y=Gf.getSparseReshapeEmptyTensorZeroOutputDimErrorMessage();break;case 3:{let e=Array.from(t.readSync(a.dataId)),n=Array.from(t.readSync(c.dataId));y=Gf.getSparseReshapeInputOutputMultipleErrorMessage(e,n);break}case 4:{let e=Array.from(t.readSync(a.dataId)),n=Array.from(t.readSync(c.dataId));y=Gf.getSparseReshapeInputOutputMismatchErrorMessage(e,n);break}default:y=""}if(t.disposeData(m.dataId),y)throw t.disposeData(d.dataId),t.disposeData(c.dataId),new Error(y);return[d,c]}};function kZ(e){vZ=e.wasm.cwrap("SparseSegmentReduction",null,["number","number","number","number","number","number","number","number","number"])}function IZ(e,t){let{backend:n,inputs:r}=e,{data:a,indices:s,segmentIds:i}=r,o=s.shape[0],l=n.readSync(i.dataId,o-1,o)[0],u=o>0?l+1:0;if(u<0)throw new Error(Gf.getSparseSegmentReductionNegativeSegmentIdsErrorMessage());let h=a.shape.slice();h[0]=u;let d=n.dataIdMap.get(a.dataId).id,p=n.dataIdMap.get(s.dataId).id,c=n.dataIdMap.get(i.dataId).id,f=n.makeOutput(h,a.dtype),m=n.dataIdMap.get(f.dataId).id,g=n.makeOutput([4],"int32"),y=n.dataIdMap.get(g.dataId).id;vZ(d,mj[a.dtype],a.shape[0],p,c,m,y,t,0);let b,x=n.readSync(g.dataId);switch(x[0]){case 0:b=Gf.getSparseSegmentReductionNegativeSegmentIdsErrorMessage();break;case 1:b=Gf.getSparseSegmentReductionNonIncreasingSegmentIdsErrorMessage();break;case 2:b=Gf.getSparseSegmentReductionSegmentIdOutOfRangeErrorMessage(x[1],x[2]);break;case 3:b=Gf.getSparseSegmentReductionIndicesOutOfRangeErrorMessage(x[1],x[2],x[3]);break;default:b=""}if(n.disposeData(g.dataId),b)throw n.disposeData(f.dataId),new Error(b);return f}var SZ={kernelName:Fr,backendName:"wasm",setupFunc:kZ,kernelFunc:function(e){return IZ(e,!0)}};var _Z,NZ={kernelName:Dr,backendName:"wasm",setupFunc:kZ,kernelFunc:function(e){return IZ(e,!1)}};var TZ={kernelName:Mr,backendName:"wasm",setupFunc:function(e){_Z=e.wasm.cwrap(Mr,null,["number","number","number","number","number","number","number","number","array","number","number"])},kernelFunc:function(e){let{backend:t,inputs:n,attrs:r}=e,{sparseIndices:a,sparseValues:s,defaultValue:i}=n,{outputShape:o}=r,l=t.makeOutput(o,i.dtype);if(0===va.sizeFromShape(o))return l;let{sliceRank:u,numUpdates:h,sliceSize:d,strides:p,outputSize:c}=Gf.calculateShapes(s,a,o),f=t.dataIdMap.get(a.dataId).id,m=t.dataIdMap.get(s.dataId).id,g=t.dataIdMap.get(i.dataId).id,y=new Uint8Array(new Int32Array(p).buffer),b=t.dataIdMap.get(l.dataId).id;return _Z(f,m,s.shape.length,g,mj[i.dtype],u,h,d,y,c,b),l}};var CZ,EZ={kernelName:Er,backendName:"wasm",kernelFunc:function(e){let{inputs:t,attrs:n,backend:r}=e,{x:a}=t,{numOrSizeSplits:s,axis:i}=n,o=va.parseAxisParam(i,a.shape)[0],l=Gf.prepareSplitSize(a,s,o),u=new Array(a.shape.length).fill(0),h=a.shape.slice();return l.map(e=>{let t=[...h];t[o]=e;let n=sq({inputs:{x:a},attrs:{begin:u,size:t},backend:r});return u[o]+=e,n})}},AZ=wj(Nr),$Z=wj(Lr),RZ=_j(Or);var FZ,DZ={kernelName:ta,backendName:"wasm",setupFunc:function(e){CZ=e.wasm.cwrap(ta,null,["number","number","number","number"])},kernelFunc:function(e){let{backend:t,inputs:n,attrs:r}=e,{alpha:a}=r,{x:s}=n,i=t.dataIdMap.get(s.dataId).id,o=t.makeOutput(s.shape,s.dtype),l=t.dataIdMap.get(o.dataId).id;return CZ(i,a,mj[s.dtype],l),o}};var MZ={kernelName:zr,backendName:"wasm",setupFunc:function(e){FZ=e.wasm.cwrap(zr,null,["number","array","number","array","array","array","array","array","number","number"])},kernelFunc:function(e){let t,{backend:n,inputs:r,attrs:a}=e,{x:s}=r,{begin:i,end:o,strides:l,beginMask:u,endMask:h,ellipsisMask:d,newAxisMask:p,shrinkAxisMask:c}=a,{finalShapeSparse:f,finalShape:m,isIdentity:g,sliceDim0:y,isSimpleSlice:b,begin:x,end:v,strides:w}=vf.sliceInfo(s.shape,i,o,l,u,h,d,p,c);if(g)t=tq({inputs:{x:s},backend:n,attrs:{shape:m}});else if(y||b){va.assert(s.shape.length>=1,()=>`Input must have rank at least 1, got: ${s.shape.length}`);let e=vf.computeOutShape(x,v,w),r=sq({inputs:{x:s},backend:n,attrs:{begin:x,size:e}});t=tq({inputs:{x:r},backend:n,attrs:{shape:m}}),n.disposeData(r.dataId)}else{let e=n.makeOutput(f,"float32"),r=n.dataIdMap.get(s.dataId).id,a=new Uint8Array(new Int32Array(va.computeStrides(s.shape)).buffer),i=new Uint8Array(new Int32Array(x).buffer),o=new Uint8Array(new Int32Array(v).buffer),l=new Uint8Array(new Int32Array(w).buffer),u=new Uint8Array(new Int32Array(f).buffer),h=new Uint8Array(new Int32Array(va.computeStrides(f)).buffer),d=n.dataIdMap.get(e.dataId).id;FZ(r,a,s.shape.length,i,o,l,u,h,f.length,d),t=tq({inputs:{x:e},backend:n,attrs:{shape:m}}),n.disposeData(e.dataId)}return t}};var OZ={kernelName:Br,backendName:"wasm",kernelFunc:function(e){let{backend:t,inputs:n,attrs:r}=e,{data:a,dataSplits:s}=n,{separator:i,nGramWidths:o,leftPad:l,rightPad:u,padWidth:h,preserveShortSequences:d}=r,p=t.readSync(a.dataId),c=t.readSync(s.dataId),[f,m]=$$(p,c,i,o,l,u,h,d),g=t.makeOutput([f.length],"string");t.dataIdMap.get(g.dataId).stringBytes=f;let y=t.makeOutput(s.shape,"int32");return t.typedArrayFromHeap(y).set(m),[g,y]}};var LZ={kernelName:Wr,backendName:"wasm",kernelFunc:function(e){let{backend:t,inputs:n,attrs:r}=e,{input:a,delimiter:s}=n,{skipEmpty:i}=r,o=t.readSync(a.dataId),l=t.readSync(s.dataId),[u,h,d]=F$(o,l[0],i),p=h.length,c=t.makeOutput([p,2],"int32");t.typedArrayFromHeap(c).set(u);let f=t.makeOutput([p],"string");t.dataIdMap.get(f.dataId).stringBytes=h;let m=t.makeOutput([2],"int32");return t.typedArrayFromHeap(m).set(d),[c,f,m]}};var PZ,zZ={kernelName:Vr,backendName:"wasm",kernelFunc:function(e){let{backend:t,inputs:n,attrs:r}=e,{input:a}=n,{numBuckets:s}=r,i=D$(t.readSync(a.dataId),s),o=t.makeOutput(a.shape,"int32");return t.typedArrayFromHeap(o).set(i),o}},BZ=_j(Ur);var WZ,VZ={kernelName:Tr,backendName:"wasm",setupFunc:function(e){PZ=e.wasm.cwrap(Tr,null,["number","number","number","number"])},kernelFunc:function(e){let{backend:t,inputs:n,attrs:r}=e,{axis:a,keepDims:s}=r,{x:i}=n,o=t.dataIdMap.get(i.dataId).id,l=o,u=i,{transposed:h,axes:d,originalAxes:p,inputWasTransposed:c}=Mj(i,a,t),f=d;if(c){let e=t.dataIdMap.get(h.dataId).id;e!==o&&(u=h,l=e,f=Gf.getInnerMostAxes(f.length,u.shape.length))}Gf.assertAxesAreInnerMostDims("sum",f,u.shape.length);let[m,g]=Gf.computeOutAndReduceShapes(u.shape,f),y=va.sizeFromShape(g),b=t.makeOutput(m,u.dtype);if(0!==va.sizeFromShape(u.shape)){let e=t.dataIdMap.get(b.dataId).id;PZ(l,y,mj[b.dtype],e)}if(c&&t.disposeData(h.dataId),s){let e=Gf.expandShapeToKeepDim(b.shape,p);b.shape=e}return b}},UZ=wj(Gr),GZ=wj(Hr);var HZ,jZ={kernelName:gr,backendName:"wasm",setupFunc:function(e){WZ=e.wasm.cwrap(gr,null,["number","number","number","number","number","number","array","number","number","number"])},kernelFunc:function(e){let{backend:t,inputs:n,attrs:r}=e,{tensor:a,indices:s,updates:i}=n,{}=r,o=t.makeOutput(a.shape,a.dtype);if(0===va.sizeFromShape(a.shape))return o;let{sliceRank:l,numUpdates:u,sliceSize:h,strides:d,outputSize:p}=Ld.calculateShapes(i,s,a.shape),c=t.dataIdMap.get(s.dataId).id,f=t.dataIdMap.get(i.dataId).id,m=t.dataIdMap.get(a.dataId).id,g=new Uint8Array(new Int32Array(d).buffer),y=t.dataIdMap.get(o.dataId).id;return WZ(c,f,mj[i.dtype],l,u,h,g,p,y,m),o}};var qZ,KZ={kernelName:jr,backendName:"wasm",setupFunc:function(e){HZ=e.wasm.cwrap(jr,null,["number","array","number","array","number","number"])},kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{x:a}=t,s=n.dataIdMap.get(a.dataId).id,{reps:i}=r,o=new Array(a.shape.length);for(let e=0;e<o.length;e++)o[e]=a.shape[e]*i[e];let l=new Uint8Array(new Int32Array(a.shape).buffer),u=new Uint8Array(new Int32Array(o).buffer),h=n.makeOutput(o,a.dtype),d=n.dataIdMap.get(h.dataId).id;return HZ(s,l,a.shape.length,u,o.length,mj[h.dtype],d),h}};var XZ,ZZ={kernelName:qr,backendName:"wasm",setupFunc:function(e){qZ=e.wasm.cwrap(qr,null,["number","array","number","number","number","bool","number","number"])},kernelFunc:({inputs:e,backend:t,attrs:n})=>{let{x:r}=e,{k:a,sorted:s}=n,i=t.dataIdMap.get(r.dataId).id,o=new Uint8Array(new Int32Array(r.shape).buffer),l=r.shape.slice();l[l.length-1]=a;let u=t.makeOutput(l,r.dtype),h=t.dataIdMap.get(u.dataId).id,d=t.makeOutput(l,"int32"),p=t.dataIdMap.get(d.dataId).id;return qZ(i,o,r.shape.length,mj[r.dtype],a,s,h,p),[u,d]}};var YZ={kernelName:Kr,backendName:"wasm",setupFunc:function(e){XZ=e.wasm.cwrap(Kr,null,["number","number","bool","number","number","number","number","number","number","array","number","array","number","number","number","number","number"])},kernelFunc:function(e){let t,{backend:n,inputs:r,attrs:a}=e,{image:s,transforms:i}=r,{interpolation:o,fillMode:l,fillValue:u,outputShape:h}=a,[d,p,c,f]=s.shape,[m,g]=null!=h?h:[p,c],y=[d,m,g,f],b=new Uint8Array(new Int32Array(va.computeStrides(s.shape)).buffer),x=new Uint8Array(new Int32Array(va.computeStrides(y)).buffer),v=n.makeOutput(y,s.dtype),w=n.dataIdMap.get(v.dataId).id,k=n.dataIdMap.get(s.dataId).id,I=n.dataIdMap.get(i.dataId).id,S="nearest"===o?1:2;switch(l){case"constant":default:t=1;break;case"reflect":t=2;break;case"wrap":t=3;break;case"nearest":t=4}return XZ(k,I,i.shape[0]>1,d,m,g,f,c,p,b,s.shape.length-1,x,y.length-1,S,t,u,w),v}};var JZ={kernelName:Zr,backendName:"wasm",kernelFunc:function(e){let{inputs:t,attrs:n,backend:r}=e,{axis:a}=n,{x:s}=t,{outputValues:i,outputShape:o,indices:l}=U$(r.readSync(s.dataId),a,s.shape,s.dtype);return[r.makeOutput(o,s.dtype,void 0,i),r.makeOutput([l.length],"int32",void 0,l)]}};var QZ={kernelName:Yr,backendName:"wasm",kernelFunc:function(e){let{inputs:t,backend:n,attrs:r}=e,{value:a}=t,{axis:s}=r;s<0&&(s+=a.shape.length);let i=a.shape[s],o=a.shape.length,l=new Array(o-1),u=0;for(let e=0;e<o;e++)e!==s&&(l[u++]=a.shape[e]);let h=new Array(i),d=new Array(o).fill(0),p=a.shape.slice();p[s]=1;for(let e=0;e<h.length;e++)d[s]=e,h[e]=sq({inputs:{x:a},attrs:{begin:d,size:p},backend:n});return h.map(({dataId:e,dtype:t})=>({dataId:e,dtype:t,shape:l}))}};var eY={kernelName:ea,backendName:"wasm",kernelFunc:function(e){let{inputs:{x:t},backend:n}=e,r=n.makeOutput(t.shape,t.dtype);return n.typedArrayFromHeap(r).fill(0),r}},tY=[vj,kj,Ij,Sj,Tj,Cj,Lj,Pj,Wj,Vj,Uj,Gj,Hj,jj,qj,Xj,eq,Yj,Qj,aq,lq,uq,hq,dq,fq,mq,gq,xq,wq,Iq,_q,Tq,Aq,$q,Rq,Dq,Oq,Pq,Bq,Vq,Gq,jq,Kq,Zq,Jq,Qq,eK,tK,nK,rK,sK,iK,uK,dK,pK,cK,mK,yK,xK,wK,IK,SK,_K,$j,NK,TK,CK,AK,$K,RK,DK,OK,MK,LK,PK,zK,BK,VK,GK,jK,qK,XK,YK,QK,tX,rX,sX,lX,uX,dX,gX,yX,bX,xX,kX,SX,NX,TX,CX,EX,$X,FX,DX,OX,PX,zX,BX,WX,VX,UX,rq,HX,qX,XX,YX,QX,tZ,nZ,rZ,sZ,oZ,uZ,hZ,dZ,pZ,cZ,fZ,iq,fX,mZ,yZ,xZ,wZ,SZ,NZ,TZ,EZ,AZ,$Z,RZ,DZ,MZ,OZ,LZ,zZ,BZ,VZ,UZ,GZ,jZ,KZ,ZZ,YZ,Dj,JZ,QZ,eY];for(let e of tY)fa(e);var nY=Le();nY.registerFlag("WASM_HAS_SIMD_SUPPORT",async()=>{try{return WebAssembly.validate(new Uint8Array([0,97,115,109,1,0,0,0,1,4,1,96,0,0,3,2,1,0,10,9,1,7,0,65,0,253,15,26,11]))}catch(e){return!1}}),nY.registerFlag("WASM_HAS_MULTITHREAD_SUPPORT",async()=>{if(nY.get("IS_NODE"))return!1;try{return(new MessageChannel).port1.postMessage(new SharedArrayBuffer(1)),WebAssembly.validate(new Uint8Array([0,97,115,109,1,0,0,0,1,4,1,96,0,0,3,2,1,0,5,4,1,3,1,1,10,11,1,9,0,65,0,254,16,2,0,26,11]))}catch(e){return!1}});var rY=m(D()),aY=m(M()),sY=m(O()),iY=rY.default||rY,oY=sY.default||sY,lY=class extends P{constructor(e){super(),this.wasm=e,this.dataIdNextNumber=1,this.wasm.tfjs.initWithThreadsCount(xY),vY=this.wasm.tfjs.getThreadsCount(),this.dataIdMap=new L(this,Zs())}write(e,t,n){let r={id:this.dataIdNextNumber++};return this.move(r,e,t,n,1),r}numDataIds(){return this.dataIdMap.numDataIds()}async time(e){let t=va.now();return e(),{kernelMs:va.now()-t}}move(e,t,n,r,a){let s=this.dataIdNextNumber++;if("string"===r){let i=t;return void this.dataIdMap.set(e,{id:s,stringBytes:i,shape:n,dtype:r,memoryOffset:null,refCount:a})}let i=va.sizeFromShape(n),o=i*va.bytesPerElement(r),l=this.wasm._malloc(o)>>>0;this.dataIdMap.set(e,{id:s,memoryOffset:l,shape:n,dtype:r,refCount:a}),this.wasm.tfjs.registerTensor(s,i,l),null!=t&&this.wasm.HEAPU8.set(new Uint8Array(t.buffer,t.byteOffset,o),l)}async read(e){return this.readSync(e)}readSync(e,t,n){let{memoryOffset:r,dtype:a,shape:s,stringBytes:i}=this.dataIdMap.get(e);if("string"===a)return null!=t&&0!==t||!(null==n||n>=i.length)?i.slice(t,n):i;t=t||0,n=n||va.sizeFromShape(s);let o=va.bytesPerElement(a);return function(e,t){switch(t){case"float32":return new Float32Array(e);case"int32":return new Int32Array(e);case"bool":return new Uint8Array(e);default:throw new Error(`Unknown dtype ${t}`)}}(this.wasm.HEAPU8.slice(r+t*o,r+n*o).buffer,a)}disposeData(e,t=!1){if(this.dataIdMap.has(e)){let n=this.dataIdMap.get(e);if(n.refCount--,!t&&n.refCount>0)return!1;this.wasm._free(n.memoryOffset),this.wasm.tfjs.disposeData(n.id),this.dataIdMap.delete(e)}return!0}refCount(e){return this.dataIdMap.has(e)?this.dataIdMap.get(e).refCount:0}incRef(e){let t=this.dataIdMap.get(e);null!=t&&t.refCount++}floatPrecision(){return 32}getMemoryOffset(e){return this.dataIdMap.get(e).memoryOffset}dispose(){this.wasm.tfjs.dispose(),"PThread"in this.wasm&&this.wasm.PThread.terminateAllThreads(),this.wasm=null}memory(){return{unreliable:!1}}makeOutput(e,t,n,r){let a;if(null==n)a=this.write(null!=r?r:null,e,t);else{let r=this.dataIdNextNumber++;a={id:r},this.dataIdMap.set(a,{id:r,memoryOffset:n,shape:e,dtype:t,refCount:1});let s=va.sizeFromShape(e);this.wasm.tfjs.registerTensor(r,s,n)}return{dataId:a,shape:e,dtype:t}}typedArrayFromHeap({shape:e,dtype:t,dataId:n}){let r=this.wasm.HEAPU8.buffer,{memoryOffset:a}=this.dataIdMap.get(n),s=va.sizeFromShape(e);switch(t){case"float32":return new Float32Array(r,a,s);case"int32":return new Int32Array(r,a,s);case"bool":return new Uint8Array(r,a,s);default:throw new Error(`Unknown dtype ${t}`)}}};function uY(e,t,n){if(null!=pY)return pY;let r="tfjs-backend-wasm.wasm";return e&&t?r="tfjs-backend-wasm-threaded-simd.wasm":e&&(r="tfjs-backend-wasm-simd.wasm"),null!=fY&&null!=fY[r]?fY[r]:n+r}async function hY(){let[e,t]=await Promise.all([Le().getAsync("WASM_HAS_SIMD_SUPPORT"),Le().getAsync("WASM_HAS_MULTITHREAD_SUPPORT")]);return new Promise((n,r)=>{let a={locateFile:(n,r)=>{if(n.endsWith(".worker.js")){let e=aY.wasmWorkerContents.replace(/\n/g,"\\n"),t=new Blob([e],{type:"application/javascript"});return URL.createObjectURL(t)}return n.endsWith(".wasm")?uY(e,t,null!=cY?cY:r):r+n}};gY&&(a.instantiateWasm=function(e){return(t,n)=>(va.fetch(e,{credentials:"same-origin"}).then(r=>{r.ok||t.env.a(`failed to load wasm binary file at '${e}'`),r.arrayBuffer().then(e=>{WebAssembly.instantiate(e,t).then(e=>{n(e.instance,e.module)})})}),{})}(uY(e,t,null!=cY?cY:"")));let s,i=!1;a.onAbort=()=>{i||mY||(mY=!0,r({message:"Make sure the server can serve the `.wasm` file relative to the bundled js file. For more details see https://github.com/tensorflow/tfjs/blob/master/tfjs-backend-wasm/README.md#using-bundlers"}))},t&&e&&null==pY?(a.mainScriptUrlOrBlob=new Blob(["var WasmBackendModuleThreadedSimd = "+iY.toString()],{type:"text/javascript"}),s=iY(a)):s=oY(a),s.then(e=>{i=!0,mY=!1;e.tfjs={init:e.cwrap("init",null,[]),initWithThreadsCount:e.cwrap("init_with_threads_count",null,["number"]),getThreadsCount:e.cwrap("get_threads_count","number",[]),registerTensor:e.cwrap("register_tensor",null,["number","number","number"]),disposeData:e.cwrap("dispose_data",null,["number"]),dispose:e.cwrap("dispose",null,[])},n({wasm:e})}).catch(r)})}var dY=["tfjs-backend-wasm.wasm","tfjs-backend-wasm-simd.wasm","tfjs-backend-wasm-threaded-simd.wasm"],pY=null,cY=null,fY={},mY=!1,gY=!1;function yY(e,t=!1){if(Ks("setWasmPath has been deprecated in favor of setWasmPaths and will be removed in a future release."),mY)throw new Error("The WASM backend was already initialized. Make sure you call `setWasmPath()` before you call `tf.setBackend()` or `tf.ready()`");pY=e,gY=t}function bY(e,t=!1){if(mY)throw new Error("The WASM backend was already initialized. Make sure you call `setWasmPaths()` before you call `tf.setBackend()` or `tf.ready()`");if("string"==typeof e)cY=e;else{fY=e;let t=dY.filter(e=>null==fY[e]);if(t.length>0)throw new Error(`There were no entries found for the following binaries: ${t.join(",")}. Please either call setWasmPaths with a map providing a path for each binary, or with a string indicating the directory where all the binaries can be found.`)}gY=t}var xY=-1,vY=-1;function wY(e){xY=e}function kY(){if(-1===vY)throw new Error("WASM backend not initialized.");return vY}var IY="4.22.0";ui("wasm",async()=>{let{wasm:e}=await hY();return new lY(e)},2);var SY="4.22.0",_Y={tfjs:SY,"tfjs-core":SY,"tfjs-converter":"4.22.0","tfjs-backend-cpu":"4.22.0","tfjs-backend-webgl":"4.22.0","tfjs-backend-wasm":"4.22.0"},NY={};function TY(e,t,n=!1){if(e.beginPath(),t.slice(1).forEach(({x:n,y:r},a)=>{let s=t[a];e.moveTo(s.x,s.y),e.lineTo(n,r)}),n){let n=t[t.length-1],r=t[0];if(!n||!r)return;e.moveTo(n.x,n.y),e.lineTo(r.x,r.y)}e.stroke()}i(NY,{AnchorPosition:()=>fJ,DrawBox:()=>bJ,DrawBoxOptions:()=>yJ,DrawFaceLandmarks:()=>sQ,DrawFaceLandmarksOptions:()=>aQ,DrawTextField:()=>gJ,DrawTextFieldOptions:()=>mJ,drawContour:()=>TY,drawDetections:()=>xJ,drawFaceExpressions:()=>tQ,drawFaceLandmarks:()=>iQ});i({},{computeReshapedDimensions:()=>PY,getCenterPoint:()=>zY,isDimensions:()=>LY,isEven:()=>MY,isFloat:()=>DY,isTensor:()=>EY,isTensor1D:()=>AY,isTensor2D:()=>$Y,isTensor3D:()=>RY,isTensor4D:()=>FY,isValidNumber:()=>WY,isValidProbablitiy:()=>VY,range:()=>BY,round:()=>OY});var CY=class e{constructor(e,t){if(!WY(e)||!WY(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 e(1/this.width,1/this.height)}};function EY(e,t){return e instanceof ts&&e.shape.length===t}function AY(e){return EY(e,1)}function $Y(e){return EY(e,2)}function RY(e){return EY(e,3)}function FY(e){return EY(e,4)}function DY(e){return e%1!=0}function MY(e){return e%2==0}function OY(e,t=2){let n=10**t;return Math.floor(e*n)/n}function LY(e){return e&&e.width&&e.height}function PY({width:e,height:t},n){let r=n/Math.max(t,e);return new CY(Math.round(e*r),Math.round(t*r))}function zY(e){return e.reduce((e,t)=>e.add(t),new UY(0,0)).div(new UY(e.length,e.length))}function BY(e,t,n){return Array(e).fill(0).map((e,r)=>t+r*n)}function WY(e){return!!e&&e!==1/0&&e!==-1/0&&!Number.isNaN(e)||0===e}function VY(e){return WY(e)&&e>=0&&e<=1}var UY=class e{constructor(e,t){this._x=e,this._y=t}get x(){return this._x}get y(){return this._y}add(t){return new e(this.x+t.x,this.y+t.y)}sub(t){return new e(this.x-t.x,this.y-t.y)}mul(t){return new e(this.x*t.x,this.y*t.y)}div(t){return new e(this.x/t.x,this.y/t.y)}abs(){return new e(Math.abs(this.x),Math.abs(this.y))}magnitude(){return Math.sqrt(this.x**2+this.y**2)}floor(){return new e(Math.floor(this.x),Math.floor(this.y))}},GY=class e{static isRect(e){return!!e&&[e.x,e.y,e.width,e.height].every(WY)}static assertIsValidBox(t,n,r=!1){if(!e.isRect(t))throw new Error(`${n} - invalid box: ${JSON.stringify(t)}, expected object with properties x, y, width, height`);if(!r&&(t.width<0||t.height<0))throw new Error(`${n} - width (${t.width}) and height (${t.height}) must be positive numbers`)}constructor(t,n=!0){let r=t||{},a=[r.left,r.top,r.right,r.bottom].every(WY),s=[r.x,r.y,r.width,r.height].every(WY);if(!s&&!a)throw new Error(`Box.constructor - expected box to be IBoundingBox | IRect, instead have ${JSON.stringify(r)}`);let[i,o,l,u]=s?[r.x,r.y,r.width,r.height]:[r.left,r.top,r.right-r.left,r.bottom-r.top];e.assertIsValidBox({x:i,y:o,width:l,height:u},"Box.constructor",n),this._x=i,this._y=o,this._width=l,this._height=u}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 UY(this.left,this.top)}get topRight(){return new UY(this.right,this.top)}get bottomLeft(){return new UY(this.left,this.bottom)}get bottomRight(){return new UY(this.right,this.bottom)}round(){let[t,n,r,a]=[this.x,this.y,this.width,this.height].map(e=>Math.round(e));return new e({x:t,y:n,width:r,height:a})}floor(){let[t,n,r,a]=[this.x,this.y,this.width,this.height].map(e=>Math.floor(e));return new e({x:t,y:n,width:r,height:a})}toSquare(){let{x:t,y:n,width:r,height:a}=this,s=Math.abs(r-a);return r<a&&(t-=s/2,r+=s),a<r&&(n-=s/2,a+=s),new e({x:t,y:n,width:r,height:a})}rescale(t){let n=LY(t)?t.width:t,r=LY(t)?t.height:t;return new e({x:this.x*n,y:this.y*r,width:this.width*n,height:this.height*r})}pad(t,n){let[r,a,s,i]=[this.x-t/2,this.y-n/2,this.width+t,this.height+n];return new e({x:r,y:a,width:s,height:i})}clipAtImageBorders(t,n){let{x:r,y:a,right:s,bottom:i}=this,o=Math.max(r,0),l=Math.max(a,0),u=s-o,h=i-l,d=Math.min(u,t-o),p=Math.min(h,n-l);return new e({x:o,y:l,width:d,height:p}).floor()}shift(t,n){let{width:r,height:a}=this,s=this.x+t,i=this.y+n;return new e({x:s,y:i,width:r,height:a})}padAtBorders(e,t){let n=this.width+1,r=this.height+1,a=n,s=r,i=this.left,o=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),i<1&&(s=2-i,i=1),o<1&&(s=2-o,o=1),{dy:1,edy:s,dx:1,edx:a,y:o,ey:u,x:i,ex:l,w:n,h:r}}calibrate(t){return new e({left:this.left+t.left*this.width,top:this.top+t.top*this.height,right:this.right+t.right*this.width,bottom:this.bottom+t.bottom*this.height}).toSquare().round()}},HY=class extends GY{constructor(e,t,n,r,a=!1){super({left:e,top:t,right:n,bottom:r},a)}},jY=class e{constructor(e,t,n,r,a){this._imageDims=new CY(a.width,a.height),this._score=e,this._classScore=t,this._className=n,this._box=new GY(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 GY(this._box).rescale(this.imageDims.reverse())}forSize(t,n){return new e(this.score,this.classScore,this.className,this.relativeBox,{width:t,height:n})}},qY=class e extends jY{constructor(e,t,n){super(e,e,"",t,n)}forSize(t,n){let{score:r,relativeBox:a,imageDims:s}=super.forSize(t,n);return new e(r,a,s)}};function KY(e,t,n=!0){let r=Math.max(0,Math.min(e.right,t.right)-Math.max(e.left,t.left))*Math.max(0,Math.min(e.bottom,t.bottom)-Math.max(e.top,t.top));return n?r/(e.area+t.area-r):r/Math.min(e.area,t.area)}function XY(e,t){return Qs(()=>{let[n,r,a]=t,s=pl([...e.shape.slice(0,3),1],n,"float32"),i=pl([...e.shape.slice(0,3),1],r,"float32"),o=pl([...e.shape.slice(0,3),1],a,"float32"),l=Xo([s,i,o],3);return Wu(e,l)})}function ZY(e){return 1/(1+Math.exp(-e))}var YY,JY=class extends GY{constructor(e,t,n,r,a=!1){super({x:e,y:t,width:n,height:r},a)}},QY=class{constructor(e,t,n=new UY(0,0)){let{width:r,height:a}=t;this._imgDims=new CY(r,a),this._shift=n,this._positions=e.map(e=>e.mul(new UY(r,a)).add(n))}get shift(){return new UY(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 UY(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 UY(e,t))}shiftByPoint(e){return this.shiftBy(e.x,e.y)}align(e,t={}){if(e){let n=e instanceof qY?e.box.floor():new GY(e);return this.shiftBy(n.x,n.y).align(null,t)}let{useDlibAlignment:n,minBoxPadding:r}={useDlibAlignment:!1,minBoxPadding:.2,...t};return n?this.alignDlib():this.alignMinBbox(r)}alignDlib(){let e=this.getRefPointsForAlignment(),[t,n,r]=e,a=e=>r.sub(e).magnitude(),s=(a(t)+a(n))/2,i=Math.floor(s/.45),o=zY(e),l=Math.floor(Math.max(0,o.x-.5*i)),u=Math.floor(Math.max(0,o.y-.43*i));return new JY(l,u,Math.min(i,this.imageWidth+l),Math.min(i,this.imageHeight+u))}alignMinBbox(e){let t=function(e){let t=e.map(e=>e.x),n=e.map(e=>e.y),r=t.reduce((e,t)=>t<e?t:e,1/0),a=n.reduce((e,t)=>t<e?t:e,1/0),s=t.reduce((e,t)=>e<t?t:e,0),i=n.reduce((e,t)=>e<t?t:e,0);return new HY(r,a,s,i)}(this.positions);return t.pad(t.width*e,t.height*e)}getRefPointsForAlignment(){throw new Error("getRefPointsForAlignment not implemented by base class")}},eJ=class extends QY{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(zY)}};function tJ(e){return e.detection instanceof qY}function nJ(e,t){return{...e,detection:t}}function rJ(){let e=window.fetch;if(!e)throw new Error("fetch - missing fetch implementation for browser environment");return{Canvas:HTMLCanvasElement,CanvasRenderingContext2D,Image:HTMLImageElement,ImageData,Video:HTMLVideoElement,createCanvasElement:()=>document.createElement("canvas"),createImageElement:()=>document.createElement("img"),createVideoElement:()=>document.createElement("video"),fetch:e,readFile:()=>{throw new Error("readFile - filesystem not available for browser environment")}}}function aJ(){return"object"==typeof n.g&&"undefined"!=typeof process&&null!=process.versions&&null!=process.versions.node}function sJ(e){let t="";if(!e&&aJ())try{e=s("fs")}catch(e){t=e.toString()}return{readFile:e?t=>new Promise((n,r)=>{e.readFile(t,(e,t)=>e?r(e):n(t))}):()=>{throw new Error(`readFile - failed to require fs in nodejs environment with error: ${t}`)}}}function iJ(){let e=n.g.Canvas||n.g.HTMLCanvasElement,t=n.g.Image||n.g.HTMLImageElement,r=n.g.Video||n.g.HTMLVideoElement,a=n.g.fetch,s=sJ();return{Canvas:e||class{},CanvasRenderingContext2D:n.g.CanvasRenderingContext2D||class{},Image:t||class{},ImageData:n.g.ImageData||class{},Video:n.g.HTMLVideoElement||class{},createCanvasElement:()=>{if(e)return new e;throw new Error("createCanvasElement - missing Canvas implementation for nodejs environment")},createImageElement:()=>{if(t)return new t;throw new Error("createImageElement - missing Image implementation for nodejs environment")},createVideoElement:()=>{if(r)return new r;throw new Error("createVideoElement - missing Video implementation for nodejs environment")},fetch:a,...s}}function oJ(){return"object"==typeof window&&"undefined"!=typeof document&&"undefined"!=typeof HTMLImageElement&&"undefined"!=typeof HTMLCanvasElement&&"undefined"!=typeof HTMLVideoElement&&"undefined"!=typeof ImageData&&"undefined"!=typeof CanvasRenderingContext2D}function lJ(e){YY=e}function uJ(){return oJ()?lJ(rJ()):aJ()?lJ(iJ()):null}var hJ={getEnv:function(){if(!YY)throw new Error("getEnv - environment is not defined, check isNodejs() and isBrowser()");return YY},setEnv:lJ,initialize:uJ,createBrowserEnv:rJ,createFileSystem:sJ,createNodejsEnv:iJ,monkeyPatch:function(e){if(YY||uJ(),!YY)throw new Error("monkeyPatch - environment is not defined, check isNodejs() and isBrowser()");let{Canvas:t=YY.Canvas,Image:n=YY.Image}=e;YY.Canvas=t,YY.Image=n,YY.createCanvasElement=e.createCanvasElement||(()=>new t),YY.createImageElement=e.createImageElement||(()=>new n),YY.ImageData=e.ImageData||YY.ImageData,YY.Video=e.Video||YY.Video,YY.fetch=e.fetch||YY.fetch,YY.readFile=e.readFile||YY.readFile},isBrowser:oJ,isNodejs:aJ};function dJ(e){return hJ.isNodejs()||"string"!=typeof e?e:document.getElementById(e)}function pJ(e){let{Canvas:t,CanvasRenderingContext2D:n}=hJ.getEnv();if(e instanceof n)return e;let r=dJ(e);if(!(r instanceof t))throw new Error("resolveContext2d - expected canvas to be of instance of Canvas");let a=r.getContext("2d",{willReadFrequently:!0});if(!a)throw new Error("resolveContext2d - canvas 2d context is null");return a}uJ();var cJ,fJ=((cJ=fJ||{}).TOP_LEFT="TOP_LEFT",cJ.TOP_RIGHT="TOP_RIGHT",cJ.BOTTOM_LEFT="BOTTOM_LEFT",cJ.BOTTOM_RIGHT="BOTTOM_RIGHT",cJ),mJ=class{constructor(e={}){let{anchorPosition:t,backgroundColor:n,fontColor:r,fontSize:a,fontStyle:s,padding:i}=e;this.anchorPosition=t||"TOP_LEFT",this.backgroundColor=n||"rgba(0, 0, 0, 0.5)",this.fontColor=r||"rgba(255, 255, 255, 1)",this.fontSize=a||14,this.fontStyle=s||"Georgia",this.padding=i||4}},gJ=class e{constructor(t,n,r={}){this.text="string"==typeof t?[t]:t instanceof e?t.text:t,this.anchor=n,this.options=new mJ(r)}measureWidth(e){let{padding:t}=this.options;return this.text.map(t=>e.measureText(t).width).reduce((e,t)=>e<t?t:e,0)+2*t}measureHeight(){let{fontSize:e,padding:t}=this.options;return this.text.length*e+2*t}getUpperLeft(e,t){let{anchorPosition:n}=this.options,r="BOTTOM_RIGHT"===n||"TOP_RIGHT"===n,a="BOTTOM_LEFT"===n||"BOTTOM_RIGHT"===n,s=this.measureWidth(e),i=this.measureHeight(),o=r?this.anchor.x-s:this.anchor.x,l=a?this.anchor.y-i:this.anchor.y;if(t){let{width:e,height:n}=t;return{x:Math.max(Math.min(o,e-s),0),y:Math.max(Math.min(l,n-i),0)}}return{x:o,y:l}}draw(e){let t=dJ(e),n=pJ(t),{backgroundColor:r,fontColor:a,fontSize:s,fontStyle:i,padding:o}=this.options;n.font=`${s}px ${i}`;let l=this.measureWidth(n),u=this.measureHeight();n.fillStyle=r;let h=this.getUpperLeft(n,t);n.fillRect(h.x,h.y,l,u),n.fillStyle=a,this.text.forEach((e,t)=>{let r=o+h.x,a=o+h.y+(t+1)*s;n.fillText(e,r,a)})}},yJ=class{constructor(e={}){let{boxColor:t,lineWidth:n,label:r,drawLabelOptions:a}=e;this.boxColor=t||"rgba(0, 0, 255, 1)",this.lineWidth=n||2,this.label=r;let s={anchorPosition:"BOTTOM_LEFT",backgroundColor:this.boxColor};this.drawLabelOptions=new mJ({...s,...a})}},bJ=class{constructor(e,t={}){this.box=new GY(e),this.options=new yJ(t)}draw(e){let t=pJ(e),{boxColor:n,lineWidth:r}=this.options,{x:a,y:s,width:i,height:o}=this.box;t.strokeStyle=n,t.lineWidth=r,t.strokeRect(a,s,i,o);let{label:l}=this.options;l&&new gJ([l],{x:a-r/2,y:s},this.options.drawLabelOptions).draw(e)}};function xJ(e,t){(Array.isArray(t)?t:[t]).forEach(t=>{let n=t instanceof qY?t.score:tJ(t)?t.detection.score:void 0,r=t instanceof qY?t.box:tJ(t)?t.detection.box:new GY(t),a=n?`${OY(n)}`:void 0;new bJ(r,{label:a}).draw(e)})}function vJ(e){let{Image:t,Video:n}=hJ.getEnv();return e instanceof t&&e.complete||e instanceof n&&e.readyState>=3}function wJ(e){let{Image:t,Video:n}=hJ.getEnv();return e instanceof t?new CY(e.naturalWidth,e.naturalHeight):e instanceof n?new CY(e.videoWidth,e.videoHeight):new CY(e.width,e.height)}function kJ({width:e,height:t}){let{createCanvasElement:n}=hJ.getEnv(),r=n();return r.width=e,r.height=t,r}function IJ(e,t){let{ImageData:n}=hJ.getEnv();if(!(e instanceof n||vJ(e)))throw new Error("createCanvasFromMedia - media has not finished loading yet");let{width:r,height:a}=t||wJ(e),s=kJ({width:r,height:a});return e instanceof n?pJ(s).putImageData(e,0,0):pJ(s).drawImage(e,0,0,r,a),s}function SJ(e){let{Image:t,Canvas:n,Video:r}=hJ.getEnv();return e instanceof t||e instanceof n||e instanceof r}var _J=class{constructor(e,t=!1){if(this._imageTensors=[],this._canvases=[],this._treatAsBatchInput=!1,this._inputDimensions=[],this._inputSize=0,!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((e,t)=>{if(RY(e))return this._imageTensors[t]=e,void(this._inputDimensions[t]=e.shape);if(FY(e)){let n=e.shape[0];if(1!==n)throw new Error(`NetInput - tf.Tensor4D with batchSize ${n} passed, but not supported in input array`);return this._imageTensors[t]=e,void(this._inputDimensions[t]=e.shape.slice(1))}let n=e instanceof hJ.getEnv().Canvas?e:IJ(e);this._canvases[t]=n,this._inputDimensions[t]=[n.height,n.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 BY(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("number"!=typeof this.inputSize)throw new Error("getReshapedInputDimensions - inputSize not set, toBatchTensor has not been called yet");return PY({width:this.getInputWidth(e),height:this.getInputHeight(e)},this.inputSize)}toBatchTensor(e,t=!0){return this._inputSize=e,Qs(()=>{let n=BY(this.batchSize,0,1).map(n=>{let r=this.getInput(n);if(r instanceof ts){let n=FY(r)?r:pu(r);return n=function(e,t=!1){return Qs(()=>{let[n,r]=e.shape.slice(1);if(n===r)return e;let a=Math.abs(n-r),s=Math.round(a*(t?.5:1)),i=n>r?2:1,o=t=>{let n=e.shape.slice();return n[i]=t,pl(n,0,"float32")},l=o(s),u=a-l.shape[i],h=[t&&u?o(u):null,e,l].filter(e=>!!e).map(e=>ho(e,"float32"));return Xo(h,i)})}(n,t),(n.shape[1]!==e||n.shape[2]!==e)&&(n=yc.resizeBilinear(n,[e,e],!1,!1)),n.as3D(e,e,3)}if(r instanceof hJ.getEnv().Canvas)return lf.fromPixels(function(e,t,n=!1){let{Image:r,Canvas:a}=hJ.getEnv();if(!(e instanceof r||e instanceof a))throw new Error("imageToSquare - expected arg0 to be HTMLImageElement | HTMLCanvasElement");if(t<=0)return kJ({width:1,height:1});let s=wJ(e),i=t/Math.max(s.height,s.width),o=i*s.width,l=i*s.height,u=kJ({width:t,height:t}),h=e instanceof a?e:IJ(e),d=Math.abs(o-l)/2,p=n&&o<l?d:0,c=n&&l<o?d:0;return h.width>0&&h.height>0&&pJ(u).drawImage(h,p,c,o,l),u}(r,e,t));throw new Error(`toBatchTensor - at batchIdx ${n}, expected input to be instanceof tf.Tensor or instanceof HTMLCanvasElement, instead have ${r}`)});return Td(n.map(e=>ho(e,"float32"))).as4D(this.batchSize,e,e,3)})}};async function NJ(e){if(e instanceof _J)return e;let t=Array.isArray(e)?e:[e];if(!t.length)throw new Error("toNetInput - empty array passed as input");let n=t=>Array.isArray(e)?` at input index ${t}:`:"",r=t.map(dJ);return r.forEach((e,r)=>{if(!SJ(e)&&!RY(e)&&!FY(e))throw"string"==typeof t[r]?new Error(`toNetInput -${n(r)} string passed, but could not resolve HTMLElement for element id ${t[r]}`):new Error(`toNetInput -${n(r)} expected media to be of type HTMLImageElement | HTMLVideoElement | HTMLCanvasElement | tf.Tensor3D, or to be an element id`);if(FY(e)){let t=e.shape[0];if(1!==t)throw new Error(`toNetInput -${n(r)} tf.Tensor4D with batchSize ${t} passed, but not supported in input array`)}}),await Promise.all(r.map(e=>SJ(e)&&function(e){return new Promise((t,n)=>{function r(e){e.currentTarget&&(e.currentTarget.removeEventListener("load",a),e.currentTarget.removeEventListener("error",r),n(e))}function a(e){e.currentTarget&&(e.currentTarget.removeEventListener("load",a),e.currentTarget.removeEventListener("error",r),t(e))}e instanceof hJ.getEnv().Canvas||vJ(e)?t(null):(e.addEventListener("load",a),e.addEventListener("error",r))})}(e))),new _J(r,Array.isArray(e))}async function TJ(e,t){let{Canvas:n}=hJ.getEnv(),r=e;if(!(e instanceof n)){let t=await NJ(e);if(t.batchSize>1)throw new Error("extractFaces - batchSize > 1 not supported");let a=t.getInput(0);r=a instanceof n?a:await async function(e,t){let n=t||hJ.getEnv().createCanvasElement(),[r,a,s]=e.shape.slice(FY(e)?1:0),i=Qs(()=>e.as3D(r,a,s).toInt());return await lf.toPixels(i,n),i.dispose(),n}(a)}let a=pJ(r);return t.map(e=>e instanceof qY?e.forSize(r.width,r.height).box.floor():e).map(e=>e.clipAtImageBorders(r.width,r.height)).map(({x:e,y:t,width:n,height:r})=>{let s=kJ({width:n,height:r});return n>0&&r>0&&pJ(s).putImageData(a.getImageData(e,t,n,r),0,0),s})}async function CJ(e,t){if(!RY(e)&&!FY(e))throw new Error("extractFaceTensors - expected image tensor to be 3D or 4D");if(FY(e)&&e.shape[0]>1)throw new Error("extractFaceTensors - batchSize > 1 not supported");return Qs(()=>{let[n,r,a]=e.shape.slice(FY(e)?1:0);return t.map(e=>e instanceof qY?e.forSize(r,n).box:e).map(e=>e.clipAtImageBorders(r,n)).filter(e=>e.width>0&&e.height>0).map(({x:t,y:s,width:i,height:o})=>yd(e.as3D(n,r,a),[s,t,0],[o,i,a]))})}async function EJ(e,t){let{fetch:n}=hJ.getEnv(),r=await n(e,t);if(!(r.status<400))throw new Error(`failed to fetch: (${r.status}) ${r.statusText}, from url: ${r.url}`);return r}function AJ(e,t){let n=`${t}-weights_manifest.json`;if(!e)return{modelBaseUri:"",manifestUri:n};if("/"===e)return{modelBaseUri:"/",manifestUri:`/${n}`};let r=e.startsWith("http://")?"http://":e.startsWith("https://")?"https://":"",a=(e=e.replace(r,"")).split("/").filter(e=>e),s=e.endsWith(".json")?a[a.length-1]:n,i=r+(e.endsWith(".json")?a.slice(0,a.length-1):a).join("/");return i=e.startsWith("/")?`/${i}`:i,{modelBaseUri:i,manifestUri:"/"===i?`/${s}`:`${i}/${s}`}}async function $J(e,t){let{manifestUri:n,modelBaseUri:r}=AJ(e,t),a=await async function(e){return(await EJ(e)).json()}(n);return Pc.loadWeights(a,r)}var RJ=class{constructor(e){this._params=void 0,this._paramMappings=[],this._name=e}get params(){return this._params}get paramMappings(){return this._paramMappings}get isLoaded(){return!!this.params}getParamFromPath(e){let{obj:t,objProp:n}=this.traversePropertyPath(e);return t[n]}reassignParamFromPath(e,t){let{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)}getFrozenParams(){return this.getParamList().filter(e=>!(e.tensor instanceof rs))}variable(){this.getFrozenParams().forEach(({path:e,tensor:t})=>{this.reassignParamFromPath(e,t.variable())})}freeze(){this.getTrainableParams().forEach(({path:e,tensor:t})=>{let n=Vs(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 ${t.path}`);t.tensor.dispose()}),this._params=void 0}serializeParams(){return new Float32Array(this.getParamList().map(({tensor:e})=>Array.from(e.dataSync())).reduce((e,t)=>e.concat(t)))}async load(e){e instanceof Float32Array?this.extractWeights(e):await this.loadFromUri(e)}async loadFromUri(e){if(e&&"string"!=typeof e)throw new Error(`${this._name}.loadFromUri - expected model uri`);let t=await $J(e,this.getDefaultModelName());this.loadFromWeightMap(t)}async loadFromDisk(e){if(e&&"string"!=typeof e)throw new Error(`${this._name}.loadFromDisk - expected model file path`);let{readFile:t}=hJ.getEnv(),{manifestUri:n,modelBaseUri:r}=AJ(e,this.getDefaultModelName()),a=Pc.weightsLoaderFactory(e=>Promise.all(e.map(e=>t(e).then(e=>"string"==typeof e?Buffer.from(e):e.buffer)))),s=JSON.parse((await t(n)).toString()),i=await a(s,r);this.loadFromWeightMap(i)}loadFromWeightMap(e){let{paramMappings:t,params:n}=this.extractParamsFromWeightMap(e);this._paramMappings=t,this._params=n}extractWeights(e){let{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");let t=e.split("/").reduce((t,n)=>{if(!t.nextObj.hasOwnProperty(n))throw new Error(`traversePropertyPath - object does not have property ${n}, for path ${e}`);return{obj:t.nextObj,objProp:n,nextObj:t.nextObj[n]}},{nextObj:this.params}),{obj:n,objProp:r}=t;if(!(n&&r&&n[r]instanceof ts))throw new Error(`traversePropertyPath - parameter is not a tensor, for path ${e}`);return{obj:n,objProp:r}}};function FJ(e,t,n){return Qs(()=>{let r=hd(e,t.depthwise_filter,t.pointwise_filter,n,"same");return r=fo(r,t.bias),r})}function DJ(e,t,n=!1){return Qs(()=>{let r=ed(n?fo(bl(e,t.conv0.filters,[2,2],"same"),t.conv0.bias):FJ(e,t.conv0,[2,2])),a=FJ(r,t.conv1,[1,1]),s=FJ(ed(fo(r,a)),t.conv2,[1,1]);return ed(fo(r,fo(a,s)))})}function MJ(e,t,n=!1,r=!0){return Qs(()=>{let a=ed(n?fo(bl(e,t.conv0.filters,r?[2,2]:[1,1],"same"),t.conv0.bias):FJ(e,t.conv0,r?[2,2]:[1,1])),s=FJ(a,t.conv1,[1,1]),i=FJ(ed(fo(a,s)),t.conv2,[1,1]),o=FJ(ed(fo(a,fo(s,i))),t.conv3,[1,1]);return ed(fo(a,fo(s,fo(i,o))))})}function OJ(e,t,n="same",r=!1){return Qs(()=>{let a=fo(bl(e,t.filters,[1,1],n),t.bias);return r?ed(a):a})}function LJ(e,t){Object.keys(e).forEach(n=>{t.some(e=>e.originalPath===n)||e[n].dispose()})}function PJ(e,t){return(n,r,a,s)=>{let i=Dd(e(n*r*a*a),[a,a,n,r]),o=$d(e(r));return t.push({paramPath:`${s}/filters`},{paramPath:`${s}/bias`}),{filters:i,bias:o}}}function zJ(e,t){return(n,r,a)=>{let s=Rd(e(n*r),[n,r]),i=$d(e(r));return t.push({paramPath:`${a}/weights`},{paramPath:`${a}/bias`}),{weights:s,bias:i}}}var BJ=class{constructor(e,t,n){this.depthwise_filter=e,this.pointwise_filter=t,this.bias=n}};function WJ(e,t){return(n,r,a)=>{let s=Dd(e(9*n),[3,3,n,1]),i=Dd(e(n*r),[1,1,n,r]),o=$d(e(r));return t.push({paramPath:`${a}/depthwise_filter`},{paramPath:`${a}/pointwise_filter`},{paramPath:`${a}/bias`}),new BJ(s,i,o)}}function VJ(e){return t=>{let n=e(`${t}/depthwise_filter`,4),r=e(`${t}/pointwise_filter`,4),a=e(`${t}/bias`,1);return new BJ(n,r,a)}}function UJ(e,t){return(n,r,a)=>{let s=e[n];if(!EY(s,r))throw new Error(`expected weightMap[${n}] to be a Tensor${r}D, instead have ${s}`);return t.push({originalPath:n,paramPath:a||n}),s}}function GJ(e){let t=e;return{extractWeights:function(e){let n=t.slice(0,e);return t=t.slice(e),n},getRemainingWeights:function(){return t}}}function HJ(e,t){let n=PJ(e,t),r=WJ(e,t);function a(e,t,a,s=!1){return{conv0:s?n(e,t,3,`${a}/conv0`):r(e,t,`${a}/conv0`),conv1:r(t,t,`${a}/conv1`),conv2:r(t,t,`${a}/conv2`)}}return{extractDenseBlock3Params:a,extractDenseBlock4Params:function(e,t,n,s=!1){let{conv0:i,conv1:o,conv2:l}=a(e,t,n,s);return{conv0:i,conv1:o,conv2:l,conv3:r(t,t,`${n}/conv3`)}}}}function jJ(e){return t=>({filters:e(`${t}/filters`,4),bias:e(`${t}/bias`,1)})}function qJ(e,t){let n=UJ(e,t),r=jJ(n),a=VJ(n);return{extractDenseBlock3Params:function(e,t=!1){return{conv0:t?r(`${e}/conv0`):a(`${e}/conv0`),conv1:a(`${e}/conv1`),conv2:a(`${e}/conv2`)}},extractDenseBlock4Params:function(e,t=!1){return{conv0:t?r(`${e}/conv0`):a(`${e}/conv0`),conv1:a(`${e}/conv1`),conv2:a(`${e}/conv2`),conv3:a(`${e}/conv3`)}}}}var KJ=class extends RJ{constructor(){super("FaceFeatureExtractor")}forwardInput(e){let{params:t}=this;if(!t)throw new Error("FaceFeatureExtractor - load model before inference");return Qs(()=>{let n=MJ(XY(ho(e.toBatchTensor(112,!0),"float32"),[122.782,117.001,104.298]).div(255),t.dense0,!0);return n=MJ(n,t.dense1),n=MJ(n,t.dense2),n=MJ(n,t.dense3),n=qo(n,[7,7],[2,2],"valid"),n})}async forward(e){return this.forwardInput(await NJ(e))}getDefaultModelName(){return"face_feature_extractor_model"}extractParamsFromWeightMap(e){return function(e){let t=[],{extractDenseBlock4Params:n}=qJ(e,t),r={dense0:n("dense0",!0),dense1:n("dense1"),dense2:n("dense2"),dense3:n("dense3")};return LJ(e,t),{params:r,paramMappings:t}}(e)}extractParams(e){return function(e){let t=[],{extractWeights:n,getRemainingWeights:r}=GJ(e),{extractDenseBlock4Params:a}=HJ(n,t),s=a(3,32,"dense0",!0),i=a(32,64,"dense1"),o=a(64,128,"dense2"),l=a(128,256,"dense3");if(0!==r().length)throw new Error(`weights remaing after extract: ${r().length}`);return{paramMappings:t,params:{dense0:s,dense1:i,dense2:o,dense3:l}}}(e)}};function XJ(e,t){return Qs(()=>fo(Zo(e,t.weights),t.bias))}function ZJ(e){let t={},n={};return Object.keys(e).forEach(r=>{(r.startsWith("fc")?n:t)[r]=e[r]}),{featureExtractorMap:t,classifierMap:n}}var YJ=class extends RJ{constructor(e,t){super(e),this._faceFeatureExtractor=t}get faceFeatureExtractor(){return this._faceFeatureExtractor}runNet(e){let{params:t}=this;if(!t)throw new Error(`${this._name} - load model before inference`);return Qs(()=>{let n=e instanceof _J?this.faceFeatureExtractor.forwardInput(e):e;return XJ(n.as2D(n.shape[0],-1),t.fc)})}dispose(e=!0){this.faceFeatureExtractor.dispose(e),super.dispose(e)}loadClassifierParams(e){let{params:t,paramMappings:n}=this.extractClassifierParams(e);this._params=t,this._paramMappings=n}extractClassifierParams(e){return function(e,t,n){let r=[],{extractWeights:a,getRemainingWeights:s}=GJ(e),i=zJ(a,r)(t,n,"fc");if(0!==s().length)throw new Error(`weights remaing after extract: ${s().length}`);return{paramMappings:r,params:{fc:i}}}(e,this.getClassifierChannelsIn(),this.getClassifierChannelsOut())}extractParamsFromWeightMap(e){let{featureExtractorMap:t,classifierMap:n}=ZJ(e);return this.faceFeatureExtractor.loadFromWeightMap(t),function(e){let t=[],n=UJ(e,t),r={fc:(a="fc",{weights:n(`${a}/weights`,2),bias:n(`${a}/bias`,1)})};var a;return LJ(e,t),{params:r,paramMappings:t}}(n)}extractParams(e){let t=this.getClassifierChannelsIn(),n=this.getClassifierChannelsOut(),r=n*t+n,a=e.slice(0,e.length-r),s=e.slice(e.length-r);return this.faceFeatureExtractor.extractWeights(a),this.extractClassifierParams(s)}},JJ=["neutral","happy","sad","angry","fearful","disgusted","surprised"],QJ=class{constructor(e){if(this.neutral=0,this.happy=0,this.sad=0,this.angry=0,this.fearful=0,this.disgusted=0,this.surprised=0,7!==e.length)throw new Error(`FaceExpressions.constructor - expected probabilities.length to be 7, have: ${e.length}`);JJ.forEach((t,n)=>{this[t]=e[n]})}asSortedArray(){return JJ.map(e=>({expression:e,probability:this[e]})).sort((e,t)=>t.probability-e.probability)}};function eQ(e,t){return{...e,expressions:t}}function tQ(e,t,n=.1,r){(Array.isArray(t)?t:[t]).forEach(t=>{let a=t instanceof QJ?t:function(e){return e.expressions instanceof QJ}(t)?t.expressions:void 0;if(!a)throw new Error("drawFaceExpressions - expected faceExpressions to be FaceExpressions | WithFaceExpressions<{}> or array thereof");let s=a.asSortedArray().filter(e=>e.probability>n),i=tJ(t)?t.detection.box.bottomLeft:r||new UY(0,0);new gJ(s.map(e=>`${e.expression} (${OY(e.probability)})`),i).draw(e)})}function nQ(e){return tJ(e)&&e.landmarks instanceof QY&&e.unshiftedLandmarks instanceof QY&&e.alignedRect instanceof qY}function rQ(e,t){let{box:n}=e.detection,r=t.shiftBy(n.x,n.y),a=r.align(),{imageDims:s}=e.detection,i=new qY(e.detection.score,a.rescale(s.reverse()),s),o=function(e){let t=e=>180*e/Math.PI,n=(e,t)=>Math.sqrt((e.x-t.x)**2+(e.y-t.y)**2),r={roll:void 0,pitch:void 0,yaw:void 0};if(!e||!e.positions||68!==e.positions.length)return r;let a=e.positions;return r.roll=((e,n)=>{let r=Math.hypot(n.x-e.x,n.y-e.y),a=n.y-e.y,s=Math.asin(a/r),i=t(s);return Math.floor(90-i)*(n.x-e.x<0?-1:1)})(a[27],a[66]),r.pitch=((e,r,a)=>{let s=n(e,a),i=new UY((e.x+a.x)/2,(e.y+a.y)/2),o=n(r,i),l=Math.atan(o/s);return Math.floor(t(l))*(i.y-r.y<0?-1:1)})(a[14],a[30],a[2]),r.yaw=(s=a[14],i=a[33],o=a[2],Math.floor(s.x-i.x)-Math.floor(i.x-o.x)),r;var s,i,o}(t);return{...e,landmarks:r,unshiftedLandmarks:t,alignedRect:i,angle:o}}var aQ=class{constructor(e={}){let{drawLines:t=!0,drawPoints:n=!0,lineWidth:r,lineColor:a,pointSize:s,pointColor:i}=e;this.drawLines=t,this.drawPoints=n,this.lineWidth=r||1,this.pointSize=s||2,this.lineColor=a||"rgba(0, 255, 255, 1)",this.pointColor=i||"rgba(255, 0, 255, 1)"}},sQ=class{constructor(e,t={}){this.faceLandmarks=e,this.options=new aQ(t)}draw(e){let t=pJ(e),{drawLines:n,drawPoints:r,lineWidth:a,lineColor:s,pointSize:i,pointColor:o}=this.options;if(n&&this.faceLandmarks instanceof eJ&&(t.strokeStyle=s,t.lineWidth=a,TY(t,this.faceLandmarks.getJawOutline()),TY(t,this.faceLandmarks.getLeftEyeBrow()),TY(t,this.faceLandmarks.getRightEyeBrow()),TY(t,this.faceLandmarks.getNose()),TY(t,this.faceLandmarks.getLeftEye(),!0),TY(t,this.faceLandmarks.getRightEye(),!0),TY(t,this.faceLandmarks.getMouth(),!0)),r){t.strokeStyle=o,t.fillStyle=o;let e=e=>{t.beginPath(),t.arc(e.x,e.y,i,0,2*Math.PI),t.fill()};this.faceLandmarks.positions.forEach(e)}}};function iQ(e,t){(Array.isArray(t)?t:[t]).forEach(t=>{let n=t instanceof QY?t:nQ(t)?t.landmarks:void 0;if(!n)throw new Error("drawFaceLandmarks - expected faceExpressions to be FaceLandmarks | WithFaceLandmarks<WithFaceDetection<{}>> or array thereof");new sQ(n).draw(e)})}function oQ(e,t){let n=[],{extractWeights:r,getRemainingWeights:a}=GJ(e),{extractConvParams:s,extractSeparableConvParams:i,extractReductionBlockParams:o,extractMainBlockParams:l}=function(e,t){let n=PJ(e,t),r=WJ(e,t);return{extractConvParams:n,extractSeparableConvParams:r,extractReductionBlockParams:function(e,t,a){return{separable_conv0:r(e,t,`${a}/separable_conv0`),separable_conv1:r(t,t,`${a}/separable_conv1`),expansion_conv:n(e,t,1,`${a}/expansion_conv`)}},extractMainBlockParams:function(e,t){return{separable_conv0:r(e,e,`${t}/separable_conv0`),separable_conv1:r(e,e,`${t}/separable_conv1`),separable_conv2:r(e,e,`${t}/separable_conv2`)}}}}(r,n),u={conv_in:s(3,32,3,"entry_flow/conv_in"),reduction_block_0:o(32,64,"entry_flow/reduction_block_0"),reduction_block_1:o(64,128,"entry_flow/reduction_block_1")},h={};BY(t,0,1).forEach(e=>{h[`main_block_${e}`]=l(128,`middle_flow/main_block_${e}`)});let d={reduction_block:o(128,256,"exit_flow/reduction_block"),separable_conv:i(256,512,"exit_flow/separable_conv")};if(0!==a().length)throw new Error(`weights remaing after extract: ${a().length}`);return{paramMappings:n,params:{entry_flow:u,middle_flow:h,exit_flow:d}}}function lQ(e,t){let n=[],{extractConvParams:r,extractSeparableConvParams:a,extractReductionBlockParams:s,extractMainBlockParams:i}=function(e,t){let n=UJ(e,t),r=jJ(n),a=VJ(n);return{extractConvParams:r,extractSeparableConvParams:a,extractReductionBlockParams:function(e){return{separable_conv0:a(`${e}/separable_conv0`),separable_conv1:a(`${e}/separable_conv1`),expansion_conv:r(`${e}/expansion_conv`)}},extractMainBlockParams:function(e){return{separable_conv0:a(`${e}/separable_conv0`),separable_conv1:a(`${e}/separable_conv1`),separable_conv2:a(`${e}/separable_conv2`)}}}}(e,n),o={conv_in:r("entry_flow/conv_in"),reduction_block_0:s("entry_flow/reduction_block_0"),reduction_block_1:s("entry_flow/reduction_block_1")},l={};BY(t,0,1).forEach(e=>{l[`main_block_${e}`]=i(`middle_flow/main_block_${e}`)});let u={reduction_block:s("exit_flow/reduction_block"),separable_conv:a("exit_flow/separable_conv")};return LJ(e,n),{params:{entry_flow:o,middle_flow:l,exit_flow:u},paramMappings:n}}function uQ(e,t,n){return fo(bl(e,t.filters,n,"same"),t.bias)}function hQ(e,t,n=!0){let r=n?ed(e):e;return r=FJ(r,t.separable_conv0,[1,1]),r=FJ(ed(r),t.separable_conv1,[1,1]),r=Yu(r,[3,3],[2,2],"same"),r=fo(r,uQ(e,t.expansion_conv,[2,2])),r}var dQ=class extends RJ{constructor(e){super("TinyXception"),this._numMainBlocks=e}forwardInput(e){let{params:t}=this;if(!t)throw new Error("TinyXception - load model before inference");return Qs(()=>{let n=XY(ho(e.toBatchTensor(112,!0),"float32"),[122.782,117.001,104.298]).div(255),r=ed(uQ(n,t.entry_flow.conv_in,[2,2]));return r=hQ(r,t.entry_flow.reduction_block_0,!1),r=hQ(r,t.entry_flow.reduction_block_1),BY(this._numMainBlocks,0,1).forEach(e=>{r=function(e,t){let n=FJ(ed(e),t.separable_conv0,[1,1]);return n=FJ(ed(n),t.separable_conv1,[1,1]),n=FJ(ed(n),t.separable_conv2,[1,1]),n=fo(n,e),n}(r,t.middle_flow[`main_block_${e}`])}),r=hQ(r,t.exit_flow.reduction_block),r=ed(FJ(r,t.exit_flow.separable_conv,[1,1])),r})}async forward(e){return this.forwardInput(await NJ(e))}getDefaultModelName(){return"tiny_xception_model"}extractParamsFromWeightMap(e){return lQ(e,this._numMainBlocks)}extractParams(e){return oQ(e,this._numMainBlocks)}};var pQ,cQ=((pQ=cQ||{}).FEMALE="female",pQ.MALE="male",pQ),fQ=class extends YJ{postProcess(e,t,n){let r=n.map(({width:e,height:n})=>{let r=t/Math.max(n,e);return{width:e*r,height:n*r}}),a=r.length;return Qs(()=>{let n=(e,t)=>Td([pl([68],e,"float32"),pl([68],t,"float32")],1).as2D(1,136).as1D(),s=(e,t)=>{let{width:n,height:a}=r[e];return t(n,a)?Math.abs(n-a)/2:0};return e.mul(pl([a,136],t,"float32")).sub(Td(Array.from(Array(a),(e,t)=>n((e=>s(e,(e,t)=>e<t))(t),(e=>s(e,(e,t)=>t<e))(t))))).div(Td(Array.from(Array(a),(e,t)=>n(r[t].width,r[t].height))))})}forwardInput(e){return Qs(()=>{let t=this.runNet(e);return this.postProcess(t,e.inputSize,e.inputDimensions.map(([e,t])=>({height:e,width:t})))})}async forward(e){return this.forwardInput(await NJ(e))}async detectLandmarks(e){let t=await NJ(e),n=Qs(()=>jd(this.forwardInput(t))),r=await Promise.all(n.map(async(e,n)=>{let r=Array.from(e.dataSync()),a=r.filter((e,t)=>MY(t)),s=r.filter((e,t)=>!MY(t));return new eJ(Array(68).fill(0).map((e,t)=>new UY(a[t],s[t])),{height:t.getInputHeight(n),width:t.getInputWidth(n)})}));return n.forEach(e=>e.dispose()),t.isBatchInput?r:r[0]}getClassifierChannelsOut(){return 136}},mQ=class extends fQ{constructor(e=new KJ){super("FaceLandmark68Net",e)}getDefaultModelName(){return"face_landmark_68_model"}getClassifierChannelsIn(){return 256}};var gQ=class extends RJ{constructor(){super("TinyFaceFeatureExtractor")}forwardInput(e){let{params:t}=this;if(!t)throw new Error("TinyFaceFeatureExtractor - load model before inference");return Qs(()=>{let n=DJ(XY(ho(e.toBatchTensor(112,!0),"float32"),[122.782,117.001,104.298]).div(255),t.dense0,!0);return n=DJ(n,t.dense1),n=DJ(n,t.dense2),n=qo(n,[14,14],[2,2],"valid"),n})}async forward(e){return this.forwardInput(await NJ(e))}getDefaultModelName(){return"face_feature_extractor_tiny_model"}extractParamsFromWeightMap(e){return function(e){let t=[],{extractDenseBlock3Params:n}=qJ(e,t),r={dense0:n("dense0",!0),dense1:n("dense1"),dense2:n("dense2")};return LJ(e,t),{params:r,paramMappings:t}}(e)}extractParams(e){return function(e){let t=[],{extractWeights:n,getRemainingWeights:r}=GJ(e),{extractDenseBlock3Params:a}=HJ(n,t),s=a(3,32,"dense0",!0),i=a(32,64,"dense1"),o=a(64,128,"dense2");if(0!==r().length)throw new Error(`weights remaing after extract: ${r().length}`);return{paramMappings:t,params:{dense0:s,dense1:i,dense2:o}}}(e)}};function yQ(e,t,n,r,a="same"){let{filters:s,bias:i}=t.conv,o=bl(e,s,n,a);return o=fo(o,i),o=function(e,t){return fo(yo(e,t.weights),t.biases)}(o,t.scale),r?ed(o):o}function bQ(e,t){return yQ(e,t,[1,1],!1)}function xQ(e,t){return yQ(e,t,[2,2],!0,"valid")}function vQ(e,t){function n(n,r,a,s){let i=function(t,n,r){let a=e(t),s=a.length/(n*r*r);if(DY(s))throw new Error(`depth has to be an integer: ${s}, weights.length: ${a.length}, numFilters: ${n}, filterSize: ${r}`);return Qs(()=>Jd(Dd(a,[n,s,r,r]),[2,3,1,0]))}(n,r,a),o=$d(e(r));return t.push({paramPath:`${s}/filters`},{paramPath:`${s}/bias`}),{filters:i,bias:o}}function r(r,a,s,i){let o=n(r,a,s,`${i}/conv`),l=function(n,r){let a=$d(e(n)),s=$d(e(n));return t.push({paramPath:`${r}/weights`},{paramPath:`${r}/biases`}),{weights:a,biases:s}}(a,`${i}/scale`);return{conv:o,scale:l}}return{extractConvLayerParams:r,extractResidualLayerParams:function(e,t,n,a,s=!1){return{conv1:r((s?.5:1)*e,t,n,`${a}/conv1`),conv2:r(e,t,n,`${a}/conv2`)}}}}function wQ(e,t){let n=UJ(e,t);function r(e){let t=n(`${e}/conv/filters`,4),r=n(`${e}/conv/bias`,1),a=function(e){return{weights:n(`${e}/scale/weights`,1),biases:n(`${e}/scale/biases`,1)}}(e);return{conv:{filters:t,bias:r},scale:a}}return{extractConvLayerParams:r,extractResidualLayerParams:function(e){return{conv1:r(`${e}/conv1`),conv2:r(`${e}/conv2`)}}}}function kQ(e,t){let n=function(e,t){return yQ(e,t,[1,1],!0)}(e,t.conv1);return n=bQ(n,t.conv2),n=fo(n,e),n=ed(n),n}function IQ(e,t){let n=xQ(e,t.conv1);n=bQ(n,t.conv2);let r=qo(e,2,2,"valid"),a=nh(r.shape),s=r.shape[3]!==n.shape[3];if(r.shape[1]!==n.shape[1]||r.shape[2]!==n.shape[2]){let e=[...n.shape];e[1]=1;let t=nh(e);n=Xo([n,t],1);let r=[...n.shape];r[2]=1;let a=nh(r);n=Xo([n,a],2)}return r=s?Xo([r,a],3):r,n=fo(r,n),n=ed(n),n}var SQ=class extends RJ{constructor(){super("FaceRecognitionNet")}forwardInput(e){let{params:t}=this;if(!t)throw new Error("FaceRecognitionNet - load model before inference");return Qs(()=>{let n=xQ(XY(ho(e.toBatchTensor(150,!0),"float32"),[122.782,117.001,104.298]).div(255),t.conv32_down);n=Yu(n,3,2,"valid"),n=kQ(n,t.conv32_1),n=kQ(n,t.conv32_2),n=kQ(n,t.conv32_3),n=IQ(n,t.conv64_down),n=kQ(n,t.conv64_1),n=kQ(n,t.conv64_2),n=kQ(n,t.conv64_3),n=IQ(n,t.conv128_down),n=kQ(n,t.conv128_1),n=kQ(n,t.conv128_2),n=IQ(n,t.conv256_down),n=kQ(n,t.conv256_1),n=kQ(n,t.conv256_2),n=IQ(n,t.conv256_down_out);let r=n.mean([1,2]);return Zo(r,t.fc)})}async forward(e){return this.forwardInput(await NJ(e))}async computeFaceDescriptor(e){var t;if(null!=(t=null==e?void 0:e.shape)&&t.some(e=>e<=0))return new Float32Array(128);let n=await NJ(e),r=Qs(()=>jd(this.forwardInput(n))),a=await Promise.all(r.map(e=>e.data()));return r.forEach(e=>e.dispose()),n.isBatchInput?a:a[0]}getDefaultModelName(){return"face_recognition_model"}extractParamsFromWeightMap(e){return function(e){let t=[],{extractConvLayerParams:n,extractResidualLayerParams:r}=wQ(e,t),a=n("conv32_down"),s=r("conv32_1"),i=r("conv32_2"),o=r("conv32_3"),l=r("conv64_down"),u=r("conv64_1"),h=r("conv64_2"),d=r("conv64_3"),p=r("conv128_down"),c=r("conv128_1"),f=r("conv128_2"),m=r("conv256_down"),g=r("conv256_1"),y=r("conv256_2"),b=r("conv256_down_out"),{fc:x}=e;if(t.push({originalPath:"fc",paramPath:"fc"}),!$Y(x))throw new Error(`expected weightMap[fc] to be a Tensor2D, instead have ${x}`);let v={conv32_down:a,conv32_1:s,conv32_2:i,conv32_3:o,conv64_down:l,conv64_1:u,conv64_2:h,conv64_3:d,conv128_down:p,conv128_1:c,conv128_2:f,conv256_down:m,conv256_1:g,conv256_2:y,conv256_down_out:b,fc:x};return LJ(e,t),{params:v,paramMappings:t}}(e)}extractParams(e){return function(e){let{extractWeights:t,getRemainingWeights:n}=GJ(e),r=[],{extractConvLayerParams:a,extractResidualLayerParams:s}=vQ(t,r),i=a(4704,32,7,"conv32_down"),o=s(9216,32,3,"conv32_1"),l=s(9216,32,3,"conv32_2"),u=s(9216,32,3,"conv32_3"),h=s(36864,64,3,"conv64_down",!0),d=s(36864,64,3,"conv64_1"),p=s(36864,64,3,"conv64_2"),c=s(36864,64,3,"conv64_3"),f=s(147456,128,3,"conv128_down",!0),m=s(147456,128,3,"conv128_1"),g=s(147456,128,3,"conv128_2"),y=s(589824,256,3,"conv256_down",!0),b=s(589824,256,3,"conv256_1"),x=s(589824,256,3,"conv256_2"),v=s(589824,256,3,"conv256_down_out"),w=Qs(()=>Jd(Rd(t(32768),[128,256]),[1,0]));if(r.push({paramPath:"fc"}),0!==n().length)throw new Error(`weights remaing after extract: ${n().length}`);return{params:{conv32_down:i,conv32_1:o,conv32_2:l,conv32_3:u,conv64_down:h,conv64_1:d,conv64_2:p,conv64_3:c,conv128_down:f,conv128_1:m,conv128_2:g,conv256_down:y,conv256_1:b,conv256_2:x,conv256_down_out:v,fc:w},paramMappings:r}}(e)}};function _Q(e,t){return{...e,descriptor:t}}function NQ(e,t){return{...e,age:t}}function TQ(e,t,n){return{...e,gender:t,genderProbability:n}}function CQ(e,t){function n(n,r,a,s,i){let o=Dd(e(n*r*a*a),[a,a,n,r]),l=$d(e(r));return t.push({paramPath:`${s}/filters`},{paramPath:`${s}/${i?"batch_norm_offset":"bias"}`}),{filters:o,bias:l}}function r(e,t,r,a){let{filters:s,bias:i}=n(e,t,r,a,!0);return{filters:s,batch_norm_offset:i}}function a(n,a,s){let i=function(n,r){let a=Dd(e(9*n),[3,3,n,1]),s=$d(e(n)),i=$d(e(n)),o=$d(e(n)),l=$d(e(n));return t.push({paramPath:`${r}/filters`},{paramPath:`${r}/batch_norm_scale`},{paramPath:`${r}/batch_norm_offset`},{paramPath:`${r}/batch_norm_mean`},{paramPath:`${r}/batch_norm_variance`}),{filters:a,batch_norm_scale:s,batch_norm_offset:i,batch_norm_mean:o,batch_norm_variance:l}}(n,`${s}/depthwise_conv`);return{depthwise_conv:i,pointwise_conv:r(n,a,1,`${s}/pointwise_conv`)}}return{extractMobilenetV1Params:function(){return{conv_0:r(3,32,3,"mobilenetv1/conv_0"),conv_1:a(32,64,"mobilenetv1/conv_1"),conv_2:a(64,128,"mobilenetv1/conv_2"),conv_3:a(128,128,"mobilenetv1/conv_3"),conv_4:a(128,256,"mobilenetv1/conv_4"),conv_5:a(256,256,"mobilenetv1/conv_5"),conv_6:a(256,512,"mobilenetv1/conv_6"),conv_7:a(512,512,"mobilenetv1/conv_7"),conv_8:a(512,512,"mobilenetv1/conv_8"),conv_9:a(512,512,"mobilenetv1/conv_9"),conv_10:a(512,512,"mobilenetv1/conv_10"),conv_11:a(512,512,"mobilenetv1/conv_11"),conv_12:a(512,1024,"mobilenetv1/conv_12"),conv_13:a(1024,1024,"mobilenetv1/conv_13")}},extractPredictionLayerParams:function(){return{conv_0:r(1024,256,1,"prediction_layer/conv_0"),conv_1:r(256,512,3,"prediction_layer/conv_1"),conv_2:r(512,128,1,"prediction_layer/conv_2"),conv_3:r(128,256,3,"prediction_layer/conv_3"),conv_4:r(256,128,1,"prediction_layer/conv_4"),conv_5:r(128,256,3,"prediction_layer/conv_5"),conv_6:r(256,64,1,"prediction_layer/conv_6"),conv_7:r(64,128,3,"prediction_layer/conv_7"),box_predictor_0:{box_encoding_predictor:n(512,12,1,"prediction_layer/box_predictor_0/box_encoding_predictor"),class_predictor:n(512,9,1,"prediction_layer/box_predictor_0/class_predictor")},box_predictor_1:{box_encoding_predictor:n(1024,24,1,"prediction_layer/box_predictor_1/box_encoding_predictor"),class_predictor:n(1024,18,1,"prediction_layer/box_predictor_1/class_predictor")},box_predictor_2:{box_encoding_predictor:n(512,24,1,"prediction_layer/box_predictor_2/box_encoding_predictor"),class_predictor:n(512,18,1,"prediction_layer/box_predictor_2/class_predictor")},box_predictor_3:{box_encoding_predictor:n(256,24,1,"prediction_layer/box_predictor_3/box_encoding_predictor"),class_predictor:n(256,18,1,"prediction_layer/box_predictor_3/class_predictor")},box_predictor_4:{box_encoding_predictor:n(256,24,1,"prediction_layer/box_predictor_4/box_encoding_predictor"),class_predictor:n(256,18,1,"prediction_layer/box_predictor_4/class_predictor")},box_predictor_5:{box_encoding_predictor:n(128,24,1,"prediction_layer/box_predictor_5/box_encoding_predictor"),class_predictor:n(128,18,1,"prediction_layer/box_predictor_5/class_predictor")}}}}}function EQ(e){let t=[],{extractMobilenetV1Params:n,extractPredictionLayerParams:r}=function(e,t){let n=UJ(e,t);function r(e,t,r){return{filters:n(`${e}/Conv2d_${t}_pointwise/weights`,4,`${r}/filters`),batch_norm_offset:n(`${e}/Conv2d_${t}_pointwise/convolution_bn_offset`,1,`${r}/batch_norm_offset`)}}function a(e){let t=`mobilenetv1/conv_${e}`,a=`MobilenetV1/Conv2d_${e}_depthwise`,s=`${t}/depthwise_conv`,i=`${t}/pointwise_conv`;return{depthwise_conv:{filters:n(`${a}/depthwise_weights`,4,`${s}/filters`),batch_norm_scale:n(`${a}/BatchNorm/gamma`,1,`${s}/batch_norm_scale`),batch_norm_offset:n(`${a}/BatchNorm/beta`,1,`${s}/batch_norm_offset`),batch_norm_mean:n(`${a}/BatchNorm/moving_mean`,1,`${s}/batch_norm_mean`),batch_norm_variance:n(`${a}/BatchNorm/moving_variance`,1,`${s}/batch_norm_variance`)},pointwise_conv:r("MobilenetV1",e,i)}}function s(e,t){return{filters:n(`${e}/weights`,4,`${t}/filters`),bias:n(`${e}/biases`,1,`${t}/bias`)}}function i(e){return{box_encoding_predictor:s(`Prediction/BoxPredictor_${e}/BoxEncodingPredictor`,`prediction_layer/box_predictor_${e}/box_encoding_predictor`),class_predictor:s(`Prediction/BoxPredictor_${e}/ClassPredictor`,`prediction_layer/box_predictor_${e}/class_predictor`)}}return{extractMobilenetV1Params:function(){return{conv_0:r("MobilenetV1",0,"mobilenetv1/conv_0"),conv_1:a(1),conv_2:a(2),conv_3:a(3),conv_4:a(4),conv_5:a(5),conv_6:a(6),conv_7:a(7),conv_8:a(8),conv_9:a(9),conv_10:a(10),conv_11:a(11),conv_12:a(12),conv_13:a(13)}},extractPredictionLayerParams:function(){return{conv_0:r("Prediction",0,"prediction_layer/conv_0"),conv_1:r("Prediction",1,"prediction_layer/conv_1"),conv_2:r("Prediction",2,"prediction_layer/conv_2"),conv_3:r("Prediction",3,"prediction_layer/conv_3"),conv_4:r("Prediction",4,"prediction_layer/conv_4"),conv_5:r("Prediction",5,"prediction_layer/conv_5"),conv_6:r("Prediction",6,"prediction_layer/conv_6"),conv_7:r("Prediction",7,"prediction_layer/conv_7"),box_predictor_0:i(0),box_predictor_1:i(1),box_predictor_2:i(2),box_predictor_3:i(3),box_predictor_4:i(4),box_predictor_5:i(5)}}}}(e,t),a=e["Output/extra_dim"];if(t.push({originalPath:"Output/extra_dim",paramPath:"output_layer/extra_dim"}),!RY(a))throw new Error(`expected weightMap['Output/extra_dim'] to be a Tensor3D, instead have ${a}`);let s={mobilenetv1:n(),prediction_layer:r(),output_layer:{extra_dim:a}};return LJ(e,t),{params:s,paramMappings:t}}function AQ(e,t,n){return Qs(()=>{let r=bl(e,t.filters,n,"same");return r=fo(r,t.batch_norm_offset),cl(r,0,6)})}function $Q(e,t){return Qs(()=>{let n,r=AQ(e,t.conv_0,[2,2]);if([t.conv_1,t.conv_2,t.conv_3,t.conv_4,t.conv_5,t.conv_6,t.conv_7,t.conv_8,t.conv_9,t.conv_10,t.conv_11,t.conv_12,t.conv_13].forEach((e,t)=>{let a=t+1,s=function(e){return[2,4,6,12].some(t=>t===e)?[2,2]:[1,1]}(a);r=function(e,t,n){return Qs(()=>{let r=$l(e,t.filters,n,"same");return r=rl(r,t.batch_norm_mean,t.batch_norm_variance,t.batch_norm_offset,t.batch_norm_scale,.0010000000474974513),cl(r,0,6)})}(r,e.depthwise_conv,s),r=AQ(r,e.pointwise_conv,[1,1]),11===a&&(n=r)}),null===n)throw new Error("mobileNetV1 - output of conv layer 11 is null");return{out:r,conv11:n}})}function RQ(e,t,n){let r=e.arraySync(),a=Math.min(r[t][0],r[t][2]),s=Math.min(r[t][1],r[t][3]),i=Math.max(r[t][0],r[t][2]),o=Math.max(r[t][1],r[t][3]),l=Math.min(r[n][0],r[n][2]),u=Math.min(r[n][1],r[n][3]),h=Math.max(r[n][0],r[n][2]),d=Math.max(r[n][1],r[n][3]),p=(i-a)*(o-s),c=(h-l)*(d-u);if(p<=0||c<=0)return 0;let f=Math.max(a,l),m=Math.max(s,u),g=Math.min(i,h),y=Math.min(o,d),b=Math.max(g-f,0)*Math.max(y-m,0);return b/(p+c-b)}function FQ(e,t){let{sizes:n,centers:r}=function(e){let t=jd(Jd(e,[1,0])),n=[Wu(t[2],t[0]),Wu(t[3],t[1])];return{sizes:n,centers:[fo(t[0],go(n[0],2)),fo(t[1],go(n[1],2))]}}(e),a=jd(Jd(t,[1,0])),s=go(yo(du(go(a[2],5)),n[0]),2),i=fo(yo(go(a[0],10),n[0]),r[0]),o=go(yo(du(go(a[3],5)),n[1]),2),l=fo(yo(go(a[1],10),n[1]),r[1]);return Jd(Td([Wu(i,s),Wu(l,o),fo(i,s),fo(l,o)]),[1,0])}function DQ(e,t){return Qs(()=>{let n=e.shape[0];return{boxPredictionEncoding:jo(OJ(e,t.box_encoding_predictor),[n,-1,1,4]),classPrediction:jo(OJ(e,t.class_predictor),[n,-1,3])}})}var MQ=class{constructor({minConfidence:e,maxResults:t}={}){if(this._name="SsdMobilenetv1Options",this._minConfidence=e||.5,this._maxResults=t||100,"number"!=typeof this._minConfidence||this._minConfidence<=0||this._minConfidence>=1)throw new Error(`${this._name} - expected minConfidence to be a number between 0 and 1`);if("number"!=typeof this._maxResults)throw new Error(`${this._name} - expected maxResults to be a number`)}get minConfidence(){return this._minConfidence}get maxResults(){return this._maxResults}},OQ=class extends RJ{constructor(){super("SsdMobilenetv1")}forwardInput(e){let{params:t}=this;if(!t)throw new Error("SsdMobilenetv1 - load model before inference");return Qs(()=>{let n=ho(e.toBatchTensor(512,!1),"float32"),r=$Q(Wu(go(n,127.5),1),t.mobilenetv1),{boxPredictions:a,classPredictions:s}=function(e,t,n){return Qs(()=>{let r=AQ(e,n.conv_0,[1,1]),a=AQ(r,n.conv_1,[2,2]),s=AQ(a,n.conv_2,[1,1]),i=AQ(s,n.conv_3,[2,2]),o=AQ(i,n.conv_4,[1,1]),l=AQ(o,n.conv_5,[2,2]),u=AQ(l,n.conv_6,[1,1]),h=AQ(u,n.conv_7,[2,2]),d=DQ(t,n.box_predictor_0),p=DQ(e,n.box_predictor_1),c=DQ(a,n.box_predictor_2),f=DQ(i,n.box_predictor_3),m=DQ(l,n.box_predictor_4),g=DQ(h,n.box_predictor_5);return{boxPredictions:Xo([d.boxPredictionEncoding,p.boxPredictionEncoding,c.boxPredictionEncoding,f.boxPredictionEncoding,m.boxPredictionEncoding,g.boxPredictionEncoding],1),classPredictions:Xo([d.classPrediction,p.classPrediction,c.classPrediction,f.classPrediction,m.classPrediction,g.classPrediction],1)}})}(r.out,r.conv11,t.prediction_layer);return function(e,t,n){return Qs(()=>{let r=e.shape[0],a=FQ(jo(fu(n.extra_dim,[r,1,1]),[-1,4]),jo(e,[-1,4]));a=jo(a,[r,a.shape[0]/r,4]);let s=Yo(Jo(t,[0,0,1],[-1,-1,-1])),i=Jo(s,[0,0,0],[-1,-1,1]);return i=jo(i,[r,i.shape[1]]),{boxes:jd(a),scores:jd(i)}})}(a,s,t.output_layer)})}async forward(e){return this.forwardInput(await NJ(e))}async locateFaces(e,t={}){let{maxResults:n,minConfidence:r}=new MQ(t),a=await NJ(e),{boxes:s,scores:i}=this.forwardInput(a),o=s[0],l=i[0];for(let e=1;e<s.length;e++)s[e].dispose(),i[e].dispose();let u=Array.from(l.dataSync()),h=function(e,t,n,r,a){let s=e.shape[0],i=Math.min(n,s),o=t.map((e,t)=>({score:e,boxIndex:t})).filter(e=>e.score>a).sort((e,t)=>t.score-e.score),l=e=>e<=r?1:0,u=[];return o.forEach(t=>{if(u.length>=i)return;let n=t.score;for(let n=u.length-1;n>=0;--n){let r=RQ(e,t.boxIndex,u[n]);if(0!==r&&(t.score*=l(r),t.score<=a))break}n===t.score&&u.push(t.boxIndex)}),u}(o,u,n,.5,r),d=a.getReshapedInputDimensions(0),p=a.inputSize,c=p/d.width,f=p/d.height,m=o.arraySync(),g=h.map(e=>{let[t,n]=[Math.max(0,m[e][0]),Math.min(1,m[e][2])].map(e=>e*f),[r,s]=[Math.max(0,m[e][1]),Math.min(1,m[e][3])].map(e=>e*c);return new qY(u[e],new JY(r,t,s-r,n-t),{height:a.getInputHeight(0),width:a.getInputWidth(0)})});return o.dispose(),l.dispose(),g}getDefaultModelName(){return"ssd_mobilenetv1_model"}extractParamsFromWeightMap(e){return EQ(e)}extractParams(e){return function(e){let t=[],{extractWeights:n,getRemainingWeights:r}=GJ(e),{extractMobilenetV1Params:a,extractPredictionLayerParams:s}=CQ(n,t),i=a(),o=s(),l={extra_dim:Fd(n(20472),[1,5118,4])};if(t.push({paramPath:"output_layer/extra_dim"}),0!==r().length)throw new Error(`weights remaing after extract: ${r().length}`);return{params:{mobilenetv1:i,prediction_layer:o,output_layer:l},paramMappings:t}}(e)}};var LQ=[new UY(.738768,.874946),new UY(2.42204,2.65704),new UY(4.30971,7.04493),new UY(10.246,4.59428),new UY(12.6868,11.8741)],PQ=[new UY(1.603231,2.094468),new UY(6.041143,7.080126),new UY(2.882459,3.518061),new UY(4.266906,5.178857),new UY(9.041765,10.66308)],zQ=[117.001,114.697,97.404],BQ=e=>"number"==typeof e;function WQ(e){return Qs(()=>{let t=yo(e,au(.10000000149011612));return fo(ed(Wu(e,t)),t)})}function VQ(e,t){return Qs(()=>{let n=mh(e,[[0,0],[1,1],[1,1],[0,0]]);return n=bl(n,t.conv.filters,[1,1],"valid"),n=Wu(n,t.bn.sub),n=yo(n,t.bn.truediv),n=fo(n,t.conv.bias),WQ(n)})}function UQ(e,t){return Qs(()=>{let n=mh(e,[[0,0],[1,1],[1,1],[0,0]]);return n=hd(n,t.depthwise_filter,t.pointwise_filter,[1,1],"valid"),n=fo(n,t.bias),WQ(n)})}function GQ(e,t){let n=PJ(e,t);let r=WJ(e,t);return{extractConvParams:n,extractConvWithBatchNormParams:function(r,a,s){let i=n(r,a,3,`${s}/conv`),o=function(n,r){let a=$d(e(n)),s=$d(e(n));return t.push({paramPath:`${r}/sub`},{paramPath:`${r}/truediv`}),{sub:a,truediv:s}}(a,`${s}/bn`);return{conv:i,bn:o}},extractSeparableConvParams:r}}function HQ(e,t){let n=UJ(e,t);function r(e){return{filters:n(`${e}/filters`,4),bias:n(`${e}/bias`,1)}}return{extractConvParams:r,extractConvWithBatchNormParams:function(e){let t=r(`${e}/conv`),a=function(e){return{sub:n(`${e}/sub`,1),truediv:n(`${e}/truediv`,1)}}(`${e}/bn`);return{conv:t,bn:a}},extractSeparableConvParams:VJ(n)}}var jQ=class{constructor({inputSize:e,scoreThreshold:t}={}){if(this._name="TinyYolov2Options",this._inputSize=e||416,this._scoreThreshold=t||.5,"number"!=typeof this._inputSize||this._inputSize%32!=0)throw new Error(`${this._name} - expected inputSize to be a number divisible by 32`);if("number"!=typeof this._scoreThreshold||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}},qQ=class e extends RJ{constructor(e){super("TinyYolov2"),function(e){if(!e)throw new Error(`invalid config: ${e}`);if("boolean"!=typeof e.withSeparableConvs)throw new Error(`config.withSeparableConvs has to be a boolean, have: ${e.withSeparableConvs}`);if(!BQ(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(e=>"string"==typeof e))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(e=>e||{}).every(e=>BQ(e.x)&&BQ(e.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)||3!==e.meanRgb.length||!e.meanRgb.every(BQ)))throw new Error(`config.meanRgb has to be an array of shape [number, number, number], have: ${JSON.stringify(e.meanRgb)}`)}(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=VQ(e,t.conv0);return n=Yu(n,[2,2],[2,2],"same"),n=VQ(n,t.conv1),n=Yu(n,[2,2],[2,2],"same"),n=VQ(n,t.conv2),n=Yu(n,[2,2],[2,2],"same"),n=VQ(n,t.conv3),n=Yu(n,[2,2],[2,2],"same"),n=VQ(n,t.conv4),n=Yu(n,[2,2],[2,2],"same"),n=VQ(n,t.conv5),n=Yu(n,[2,2],[1,1],"same"),n=VQ(n,t.conv6),n=VQ(n,t.conv7),OJ(n,t.conv8,"valid",!1)}runMobilenet(e,t){let n=this.config.isFirstLayerConv2d?WQ(OJ(e,t.conv0,"valid",!1)):UQ(e,t.conv0);return n=Yu(n,[2,2],[2,2],"same"),n=UQ(n,t.conv1),n=Yu(n,[2,2],[2,2],"same"),n=UQ(n,t.conv2),n=Yu(n,[2,2],[2,2],"same"),n=UQ(n,t.conv3),n=Yu(n,[2,2],[2,2],"same"),n=UQ(n,t.conv4),n=Yu(n,[2,2],[2,2],"same"),n=UQ(n,t.conv5),n=Yu(n,[2,2],[1,1],"same"),n=t.conv6?UQ(n,t.conv6):n,n=t.conv7?UQ(n,t.conv7):n,OJ(n,t.conv8,"valid",!1)}forwardInput(e,t){let{params:n}=this;if(!n)throw new Error("TinyYolov2 - load model before inference");return Qs(()=>{let r=ho(e.toBatchTensor(t,!1),"float32");return r=this.config.meanRgb?XY(r,this.config.meanRgb):r,r=r.div(255),this.config.withSeparableConvs?this.runMobilenet(r,n):this.runTinyYolov2(r,n)})}async forward(e,t){return this.forwardInput(await NJ(e),t)}async detect(e,t={}){let{inputSize:n,scoreThreshold:r}=new jQ(t),a=await NJ(e),s=await this.forwardInput(a,n),i=Qs(()=>jd(s)[0].expandDims()),o={width:a.getInputWidth(0),height:a.getInputHeight(0)},l=await this.extractBoxes(i,a.getReshapedInputDimensions(0),r);s.dispose(),i.dispose();let u=l.map(e=>e.box),h=l.map(e=>e.score),d=l.map(e=>e.classScore),p=l.map(e=>this.config.classes[e.label]);return function(e,t,n,r=!0){let a=t.map((e,t)=>({score:e,boxIndex:t})).sort((e,t)=>e.score-t.score).map(e=>e.boxIndex),s=[];for(;a.length>0;){let t=a.pop();s.push(t);let i=a,o=[];for(let n=0;n<i.length;n++){let a=i[n],s=e[t],l=e[a];o.push(KY(s,l,r))}a=a.filter((e,t)=>o[t]<=n)}return s}(u.map(e=>e.rescale(n)),h,this.config.iouThreshold,!0).map(e=>new jY(h[e],d[e],p[e],u[e],o))}getDefaultModelName(){return""}extractParamsFromWeightMap(e){return function(e,t){let n,r=[],{extractConvParams:a,extractConvWithBatchNormParams:s,extractSeparableConvParams:i}=HQ(e,r);if(t.withSeparableConvs){let e=t.filterSizes&&t.filterSizes.length||9;n={conv0:t.isFirstLayerConv2d?a("conv0"):i("conv0"),conv1:i("conv1"),conv2:i("conv2"),conv3:i("conv3"),conv4:i("conv4"),conv5:i("conv5"),conv6:e>7?i("conv6"):void 0,conv7:e>8?i("conv7"):void 0,conv8:a("conv8")}}else n={conv0:s("conv0"),conv1:s("conv1"),conv2:s("conv2"),conv3:s("conv3"),conv4:s("conv4"),conv5:s("conv5"),conv6:s("conv6"),conv7:s("conv7"),conv8:a("conv8")};return LJ(e,r),{params:n,paramMappings:r}}(e,this.config)}extractParams(t){let n=this.config.filterSizes||e.DEFAULT_FILTER_SIZES,r=n?n.length:void 0;if(7!==r&&8!==r&&9!==r)throw new Error(`TinyYolov2 - expected 7 | 8 | 9 convolutional filters, but found ${r} filterSizes in config`);return function(e,t,n,r){let a,{extractWeights:s,getRemainingWeights:i}=GJ(e),o=[],{extractConvParams:l,extractConvWithBatchNormParams:u,extractSeparableConvParams:h}=GQ(s,o);if(t.withSeparableConvs){let[e,s,i,o,u,d,p,c,f]=r;a={conv0:t.isFirstLayerConv2d?l(e,s,3,"conv0"):h(e,s,"conv0"),conv1:h(s,i,"conv1"),conv2:h(i,o,"conv2"),conv3:h(o,u,"conv3"),conv4:h(u,d,"conv4"),conv5:h(d,p,"conv5"),conv6:c?h(p,c,"conv6"):void 0,conv7:f?h(c,f,"conv7"):void 0,conv8:l(f||c||p,5*n,1,"conv8")}}else{let[e,t,s,i,o,h,d,p,c]=r;a={conv0:u(e,t,"conv0"),conv1:u(t,s,"conv1"),conv2:u(s,i,"conv2"),conv3:u(i,o,"conv3"),conv4:u(o,h,"conv4"),conv5:u(h,d,"conv5"),conv6:u(d,p,"conv6"),conv7:u(p,c,"conv7"),conv8:l(c,5*n,1,"conv8")}}if(0!==i().length)throw new Error(`weights remaing after extract: ${i().length}`);return{params:a,paramMappings:o}}(t,this.config,this.boxEncodingSize,n)}async extractBoxes(e,t,n){let{width:r,height:a}=t,s=Math.max(r,a),i=s/r,o=s/a,l=e.shape[1],u=this.config.anchors.length,[h,d,p]=Qs(()=>{let t=e.reshape([l,l,u,this.boxEncodingSize]);return[t.slice([0,0,0,0],[l,l,u,4]),t.slice([0,0,0,4],[l,l,u,1]),this.withClassScores?xd(t.slice([0,0,0,5],[l,l,u,this.config.classes.length]),3):au(0)]}),c=[],f=await d.array(),m=await h.array();for(let e=0;e<l;e++)for(let t=0;t<l;t++)for(let r=0;r<u;r++){let a=ZY(f[e][t][r][0]);if(!n||a>n){let n=(t+ZY(m[e][t][r][0]))/l*i,s=(e+ZY(m[e][t][r][1]))/l*o,u=Math.exp(m[e][t][r][2])*this.config.anchors[r].x/l*i,h=Math.exp(m[e][t][r][3])*this.config.anchors[r].y/l*o,d=n-u/2,f=s-h/2,g={row:e,col:t,anchor:r},{classScore:y,label:b}=this.withClassScores?await this.extractPredictedClass(p,g):{classScore:1,label:0};c.push({box:new HY(d,f,d+u,f+h),score:a,classScore:a*y,label:b,...g})}}return h.dispose(),d.dispose(),p.dispose(),c}async extractPredictedClass(e,t){let{row:n,col:r,anchor:a}=t,s=await e.array();return Array(this.config.classes.length).fill(0).map((e,t)=>s[n][r][a][t]).map((e,t)=>({classScore:e,label:t})).reduce((e,t)=>e.classScore>t.classScore?e:t)}};qQ.DEFAULT_FILTER_SIZES=[3,16,32,64,128,256,512,1024,1024];var KQ=qQ,XQ=class extends KQ{constructor(e=!0){super({withSeparableConvs:e,iouThreshold:.4,classes:["face"],...e?{anchors:PQ,meanRgb:zQ}:{anchors:LQ,withClassScores:!0}})}get withSeparableConvs(){return this.config.withSeparableConvs}get anchors(){return this.config.anchors}async locateFaces(e,t){return(await this.detect(e,t)).map(e=>new qY(e.score,e.relativeBox,{width:e.imageWidth,height:e.imageHeight}))}getDefaultModelName(){return this.withSeparableConvs?"tiny_yolov2_separable_conv_model":"tiny_yolov2_model"}extractParamsFromWeightMap(e){return super.extractParamsFromWeightMap(e)}};var ZQ=class extends jQ{constructor(){super(...arguments),this._name="TinyFaceDetectorOptions"}},YQ=class{async then(e){return e(await this.run())}async run(){throw new Error("ComposableTask - run is not implemented")}};async function JQ(e,t,n,r,a=({alignedRect:e})=>e){let s=e.map(e=>nQ(e)?a(e):e.detection),i=r||(t instanceof ts?await CJ(t,s):await TJ(t,s)),o=await n(i);return i.forEach(e=>e instanceof ts&&e.dispose()),o}var QQ=[new UY(1.603231,2.094468),new UY(6.041143,7.080126),new UY(2.882459,3.518061),new UY(4.266906,5.178857),new UY(9.041765,10.66308)],e0=[117.001,114.697,97.404],t0=class extends KQ{constructor(){super({withSeparableConvs:!0,iouThreshold:.4,classes:["face"],anchors:QQ,meanRgb:e0,isFirstLayerConv2d:!0,filterSizes:[3,16,32,64,128,256,512]})}get anchors(){return this.config.anchors}async locateFaces(e,t){return(await this.detect(e,t)).map(e=>new qY(e.score,e.relativeBox,{width:e.imageWidth,height:e.imageHeight}))}getDefaultModelName(){return"tiny_face_detector_model"}extractParamsFromWeightMap(e){return super.extractParamsFromWeightMap(e)}},n0={ssdMobilenetv1:new OQ,tinyFaceDetector:new t0,tinyYolov2:new XQ,faceLandmark68Net:new mQ,faceLandmark68TinyNet:new class extends fQ{constructor(e=new gQ){super("FaceLandmark68TinyNet",e)}getDefaultModelName(){return"face_landmark_68_tiny_model"}getClassifierChannelsIn(){return 128}},faceRecognitionNet:new SQ,faceExpressionNet:new class extends YJ{constructor(e=new KJ){super("FaceExpressionNet",e)}forwardInput(e){return Qs(()=>xd(this.runNet(e)))}async forward(e){return this.forwardInput(await NJ(e))}async predictExpressions(e){let t=await NJ(e),n=await this.forwardInput(t),r=await Promise.all(jd(n).map(async e=>{let t=e.dataSync();return e.dispose(),t}));n.dispose();let a=r.map(e=>new QJ(e));return t.isBatchInput?a:a[0]}getDefaultModelName(){return"face_expression_model"}getClassifierChannelsIn(){return 256}getClassifierChannelsOut(){return 7}},ageGenderNet:new class extends RJ{constructor(e=new dQ(2)){super("AgeGenderNet"),this._faceFeatureExtractor=e}get faceFeatureExtractor(){return this._faceFeatureExtractor}runNet(e){let{params:t}=this;if(!t)throw new Error(`${this._name} - load model before inference`);return Qs(()=>{let n=e instanceof _J?this.faceFeatureExtractor.forwardInput(e):e,r=qo(n,[7,7],[2,2],"valid").as2D(n.shape[0],-1);return{age:XJ(r,t.fc.age).as1D(),gender:XJ(r,t.fc.gender)}})}forwardInput(e){return Qs(()=>{let{age:t,gender:n}=this.runNet(e);return{age:t,gender:xd(n)}})}async forward(e){return this.forwardInput(await NJ(e))}async predictAgeAndGender(e){let t=await NJ(e),n=await this.forwardInput(t),r=jd(n.age),a=jd(n.gender),s=r.map((e,t)=>({ageTensor:e,genderTensor:a[t]})),i=await Promise.all(s.map(async({ageTensor:e,genderTensor:t})=>{let n=e.dataSync()[0],r=t.dataSync()[0],a=r>.5,s=a?"male":"female",i=a?r:1-r;return e.dispose(),t.dispose(),{age:n,gender:s,genderProbability:i}}));return n.age.dispose(),n.gender.dispose(),t.isBatchInput?i:i[0]}getDefaultModelName(){return"age_gender_model"}dispose(e=!0){this.faceFeatureExtractor.dispose(e),super.dispose(e)}loadClassifierParams(e){let{params:t,paramMappings:n}=this.extractClassifierParams(e);this._params=t,this._paramMappings=n}extractClassifierParams(e){return function(e){let t=[],{extractWeights:n,getRemainingWeights:r}=GJ(e),a=zJ(n,t),s=a(512,1,"fc/age"),i=a(512,2,"fc/gender");if(0!==r().length)throw new Error(`weights remaing after extract: ${r().length}`);return{paramMappings:t,params:{fc:{age:s,gender:i}}}}(e)}extractParamsFromWeightMap(e){let{featureExtractorMap:t,classifierMap:n}=ZJ(e);return this.faceFeatureExtractor.loadFromWeightMap(t),function(e){let t=[],n=UJ(e,t);function r(e){return{weights:n(`${e}/weights`,2),bias:n(`${e}/bias`,1)}}let a={fc:{age:r("fc/age"),gender:r("fc/gender")}};return LJ(e,t),{params:a,paramMappings:t}}(n)}extractParams(e){let t=e.slice(0,e.length-1539),n=e.slice(e.length-1539);return this.faceFeatureExtractor.extractWeights(t),this.extractClassifierParams(n)}}},r0=class extends YQ{constructor(e,t,n){super(),this.parentTask=e,this.input=t,this.extractedFaces=n}},a0=class extends r0{async run(){let e=await this.parentTask,t=await JQ(e,this.input,async e=>Promise.all(e.map(e=>n0.faceExpressionNet.predictExpressions(e))),this.extractedFaces);return e.map((e,n)=>eQ(e,t[n]))}withAgeAndGender(){return new o0(this,this.input)}},s0=class extends a0{withAgeAndGender(){return new l0(this,this.input)}withFaceDescriptors(){return new h0(this,this.input)}},i0=class extends YQ{constructor(e,t,n){super(),this.parentTask=e,this.input=t,this.extractedFaces=n}},o0=class extends i0{async run(){let e=await this.parentTask,t=await JQ(e,this.input,async e=>Promise.all(e.map(e=>n0.ageGenderNet.predictAgeAndGender(e))),this.extractedFaces);return e.map((e,n)=>{let{age:r,gender:a,genderProbability:s}=t[n];return NQ(TQ(e,a,s),r)})}withFaceExpressions(){return new a0(this,this.input)}},l0=class extends o0{withFaceExpressions(){return new s0(this,this.input)}withFaceDescriptors(){return new h0(this,this.input)}},u0=class extends YQ{constructor(e,t){super(),this.parentTask=e,this.input=t}},h0=class extends u0{async run(){let e=await this.parentTask;return(await JQ(e,this.input,e=>Promise.all(e.map(e=>n0.faceRecognitionNet.computeFaceDescriptor(e))),null,e=>e.landmarks.align(null,{useDlibAlignment:!0}))).map((t,n)=>_Q(e[n],t))}withFaceExpressions(){return new s0(this,this.input)}withAgeAndGender(){return new l0(this,this.input)}},d0=class extends YQ{constructor(e,t,n){super(),this.parentTask=e,this.input=t,this.useTinyLandmarkNet=n}get landmarkNet(){return this.useTinyLandmarkNet?n0.faceLandmark68TinyNet:n0.faceLandmark68Net}},p0=class extends d0{async run(){let e=await this.parentTask,t=e.map(e=>e.detection),n=this.input instanceof ts?await CJ(this.input,t):await TJ(this.input,t),r=await Promise.all(n.map(e=>this.landmarkNet.detectLandmarks(e)));return n.forEach(e=>e instanceof ts&&e.dispose()),e.filter((e,t)=>r[t]).map((e,t)=>rQ(e,r[t]))}withFaceExpressions(){return new s0(this,this.input)}withAgeAndGender(){return new l0(this,this.input)}withFaceDescriptors(){return new h0(this,this.input)}},c0=class extends YQ{constructor(e,t=new MQ){super(),this.input=e,this.options=t}},f0=class extends c0{async run(){let e,{input:t,options:n}=this;if(n instanceof ZQ)e=n0.tinyFaceDetector.locateFaces(t,n);else if(n instanceof MQ)e=n0.ssdMobilenetv1.locateFaces(t,n);else{if(!(n instanceof jQ))throw new Error("detectFaces - expected options to be instance of TinyFaceDetectorOptions | SsdMobilenetv1Options | TinyYolov2Options");e=n0.tinyYolov2.locateFaces(t,n)}return e}runAndExtendWithFaceDetections(){return new Promise((e,t)=>{this.run().then(t=>e(t.map(e=>nJ({},e)))).catch(e=>t(e))})}withFaceLandmarks(e=!1){return new p0(this.runAndExtendWithFaceDetections(),this.input,e)}withFaceExpressions(){return new a0(this.runAndExtendWithFaceDetections(),this.input)}withAgeAndGender(){return new o0(this.runAndExtendWithFaceDetections(),this.input)}};function m0(e,t=new MQ){return new f0(e,t)}function g0(e,t){let{width:n,height:r}=new CY(t.width,t.height);if(n<=0||r<=0)throw new Error(`resizeResults - invalid dimensions: ${JSON.stringify({width:n,height:r})}`);if(Array.isArray(e))return e.map(e=>g0(e,{width:n,height:r}));if(nQ(e)){let t=e.detection.forSize(n,r),a=e.unshiftedLandmarks.forSize(t.box.width,t.box.height);return rQ(nJ(e,t),a)}return tJ(e)?nJ(e,e.detection.forSize(n,r)):e instanceof QY||e instanceof qY?e.forSize(n,r):e}},8747(e,t,n){"use strict";n.d(t,{FN:()=>Pe,m_:()=>Gn});var r=n(33011),a=n(71576),s={};n.r(s),n.d(s,{afterMain:()=>r.EP,afterRead:()=>r.v5,afterWrite:()=>r.SE,applyStyles:()=>r.ZM,arrow:()=>r.UE,auto:()=>r.qZ,basePlacements:()=>r.OM,beforeMain:()=>r.LG,beforeRead:()=>r.ci,beforeWrite:()=>r.pA,bottom:()=>r.sQ,clippingParents:()=>r.WY,computeStyles:()=>r.KC,createPopper:()=>r.n4,createPopperBase:()=>r.eR,createPopperLite:()=>r.Ls,detectOverflow:()=>r.__,end:()=>r._N,eventListeners:()=>r.Gc,flip:()=>r.UU,hide:()=>r.jD,left:()=>r.kb,main:()=>r.iW,modifierPhases:()=>r.GM,offset:()=>r.cY,placements:()=>r.DD,popper:()=>r.xf,popperGenerator:()=>r.UD,popperOffsets:()=>r.Zo,preventOverflow:()=>r.V7,read:()=>r.LF,reference:()=>r.ir,right:()=>r.pG,start:()=>r.ni,top:()=>r.Mn,variationPlacements:()=>r.Ol,viewport:()=>r.R9,write:()=>r.M9});var i=n(71757);const o="transitionend",l=e=>null==e?`${e}`:Object.prototype.toString.call(e).match(/\s([a-z]+)/i)[1].toLowerCase(),u=e=>{let t=e.getAttribute("data-bs-target");if(!t||"#"===t){let n=e.getAttribute("href");if(!n||!n.includes("#")&&!n.startsWith("."))return null;n.includes("#")&&!n.startsWith("#")&&(n=`#${n.split("#")[1]}`),t=n&&"#"!==n?n.trim():null}return t},h=e=>{const t=u(e);return t&&document.querySelector(t)?t:null},d=e=>{const t=u(e);return t?document.querySelector(t):null},p=e=>{e.dispatchEvent(new Event(o))},c=e=>!(!e||"object"!=typeof e)&&(void 0!==e.jquery&&(e=e[0]),void 0!==e.nodeType),f=e=>c(e)?e.jquery?e[0]:e:"string"==typeof e&&e.length>0?document.querySelector(e):null,m=e=>{if(!c(e)||0===e.getClientRects().length)return!1;const t="visible"===getComputedStyle(e).getPropertyValue("visibility"),n=e.closest("details:not([open])");if(!n)return t;if(n!==e){const t=e.closest("summary");if(t&&t.parentNode!==n)return!1;if(null===t)return!1}return t},g=e=>!e||e.nodeType!==Node.ELEMENT_NODE||(!!e.classList.contains("disabled")||(void 0!==e.disabled?e.disabled:e.hasAttribute("disabled")&&"false"!==e.getAttribute("disabled"))),y=e=>{if(!document.documentElement.attachShadow)return null;if("function"==typeof e.getRootNode){const t=e.getRootNode();return t instanceof ShadowRoot?t:null}return e instanceof ShadowRoot?e:e.parentNode?y(e.parentNode):null},b=()=>{},x=e=>{e.offsetHeight},v=()=>i&&!document.body.hasAttribute("data-bs-no-jquery")?i:null,w=[],k=()=>"rtl"===document.documentElement.dir,I=e=>{var t;t=()=>{const t=v();if(t){const n=e.NAME,r=t.fn[n];t.fn[n]=e.jQueryInterface,t.fn[n].Constructor=e,t.fn[n].noConflict=()=>(t.fn[n]=r,e.jQueryInterface)}},"loading"===document.readyState?(w.length||document.addEventListener("DOMContentLoaded",()=>{for(const e of w)e()}),w.push(t)):t()},S=e=>{"function"==typeof e&&e()},_=(e,t,n=!0)=>{if(!n)return void S(e);const r=(e=>{if(!e)return 0;let{transitionDuration:t,transitionDelay:n}=window.getComputedStyle(e);const r=Number.parseFloat(t),a=Number.parseFloat(n);return r||a?(t=t.split(",")[0],n=n.split(",")[0],1e3*(Number.parseFloat(t)+Number.parseFloat(n))):0})(t)+5;let a=!1;const s=({target:n})=>{n===t&&(a=!0,t.removeEventListener(o,s),S(e))};t.addEventListener(o,s),setTimeout(()=>{a||p(t)},r)},N=(e,t,n,r)=>{const a=e.length;let s=e.indexOf(t);return-1===s?!n&&r?e[a-1]:e[0]:(s+=n?1:-1,r&&(s=(s+a)%a),e[Math.max(0,Math.min(s,a-1))])},T=/[^.]*(?=\..*)\.|.*/,C=/\..*/,E=/::\d+$/,A={};let $=1;const R={mouseenter:"mouseover",mouseleave:"mouseout"},F=new Set(["click","dblclick","mouseup","mousedown","contextmenu","mousewheel","DOMMouseScroll","mouseover","mouseout","mousemove","selectstart","selectend","keydown","keypress","keyup","orientationchange","touchstart","touchmove","touchend","touchcancel","pointerdown","pointermove","pointerup","pointerleave","pointercancel","gesturestart","gesturechange","gestureend","focus","blur","change","reset","select","submit","focusin","focusout","load","unload","beforeunload","resize","move","DOMContentLoaded","readystatechange","error","abort","scroll"]);function D(e,t){return t&&`${t}::${$++}`||e.uidEvent||$++}function M(e){const t=D(e);return e.uidEvent=t,A[t]=A[t]||{},A[t]}function O(e,t,n=null){return Object.values(e).find(e=>e.callable===t&&e.delegationSelector===n)}function L(e,t,n){const r="string"==typeof t,a=r?n:t||n;let s=W(e);return F.has(s)||(s=e),[r,a,s]}function P(e,t,n,r,a){if("string"!=typeof t||!e)return;let[s,i,o]=L(t,n,r);if(t in R){const e=e=>function(t){if(!t.relatedTarget||t.relatedTarget!==t.delegateTarget&&!t.delegateTarget.contains(t.relatedTarget))return e.call(this,t)};i=e(i)}const l=M(e),u=l[o]||(l[o]={}),h=O(u,i,s?n:null);if(h)return void(h.oneOff=h.oneOff&&a);const d=D(i,t.replace(T,"")),p=s?function(e,t,n){return function r(a){const s=e.querySelectorAll(t);for(let{target:i}=a;i&&i!==this;i=i.parentNode)for(const o of s)if(o===i)return U(a,{delegateTarget:i}),r.oneOff&&V.off(e,a.type,t,n),n.apply(i,[a])}}(e,n,i):function(e,t){return function n(r){return U(r,{delegateTarget:e}),n.oneOff&&V.off(e,r.type,t),t.apply(e,[r])}}(e,i);p.delegationSelector=s?n:null,p.callable=i,p.oneOff=a,p.uidEvent=d,u[d]=p,e.addEventListener(o,p,s)}function z(e,t,n,r,a){const s=O(t[n],r,a);s&&(e.removeEventListener(n,s,Boolean(a)),delete t[n][s.uidEvent])}function B(e,t,n,r){const a=t[n]||{};for(const s of Object.keys(a))if(s.includes(r)){const r=a[s];z(e,t,n,r.callable,r.delegationSelector)}}function W(e){return e=e.replace(C,""),R[e]||e}const V={on(e,t,n,r){P(e,t,n,r,!1)},one(e,t,n,r){P(e,t,n,r,!0)},off(e,t,n,r){if("string"!=typeof t||!e)return;const[a,s,i]=L(t,n,r),o=i!==t,l=M(e),u=l[i]||{},h=t.startsWith(".");if(void 0===s){if(h)for(const n of Object.keys(l))B(e,l,n,t.slice(1));for(const n of Object.keys(u)){const r=n.replace(E,"");if(!o||t.includes(r)){const t=u[n];z(e,l,i,t.callable,t.delegationSelector)}}}else{if(!Object.keys(u).length)return;z(e,l,i,s,a?n:null)}},trigger(e,t,n){if("string"!=typeof t||!e)return null;const r=v();let a=null,s=!0,i=!0,o=!1;t!==W(t)&&r&&(a=r.Event(t,n),r(e).trigger(a),s=!a.isPropagationStopped(),i=!a.isImmediatePropagationStopped(),o=a.isDefaultPrevented());let l=new Event(t,{bubbles:s,cancelable:!0});return l=U(l,n),o&&l.preventDefault(),i&&e.dispatchEvent(l),l.defaultPrevented&&a&&a.preventDefault(),l}};function U(e,t){for(const[n,r]of Object.entries(t||{}))try{e[n]=r}catch(t){Object.defineProperty(e,n,{configurable:!0,get:()=>r})}return e}const G=new Map,H={set(e,t,n){G.has(e)||G.set(e,new Map);const r=G.get(e);r.has(t)||0===r.size?r.set(t,n):console.error(`Bootstrap doesn't allow more than one instance per element. Bound instance: ${Array.from(r.keys())[0]}.`)},get:(e,t)=>G.has(e)&&G.get(e).get(t)||null,remove(e,t){if(!G.has(e))return;const n=G.get(e);n.delete(t),0===n.size&&G.delete(e)}};function j(e){if("true"===e)return!0;if("false"===e)return!1;if(e===Number(e).toString())return Number(e);if(""===e||"null"===e)return null;if("string"!=typeof e)return e;try{return JSON.parse(decodeURIComponent(e))}catch(t){return e}}function q(e){return e.replace(/[A-Z]/g,e=>`-${e.toLowerCase()}`)}const K={setDataAttribute(e,t,n){e.setAttribute(`data-bs-${q(t)}`,n)},removeDataAttribute(e,t){e.removeAttribute(`data-bs-${q(t)}`)},getDataAttributes(e){if(!e)return{};const t={},n=Object.keys(e.dataset).filter(e=>e.startsWith("bs")&&!e.startsWith("bsConfig"));for(const r of n){let n=r.replace(/^bs/,"");n=n.charAt(0).toLowerCase()+n.slice(1,n.length),t[n]=j(e.dataset[r])}return t},getDataAttribute:(e,t)=>j(e.getAttribute(`data-bs-${q(t)}`))};class X{static get Default(){return{}}static get DefaultType(){return{}}static get NAME(){throw new Error('You have to implement the static method "NAME", for each component!')}_getConfig(e){return e=this._mergeConfigObj(e),e=this._configAfterMerge(e),this._typeCheckConfig(e),e}_configAfterMerge(e){return e}_mergeConfigObj(e,t){const n=c(t)?K.getDataAttribute(t,"config"):{};return{...this.constructor.Default,..."object"==typeof n?n:{},...c(t)?K.getDataAttributes(t):{},..."object"==typeof e?e:{}}}_typeCheckConfig(e,t=this.constructor.DefaultType){for(const n of Object.keys(t)){const r=t[n],a=e[n],s=c(a)?"element":l(a);if(!new RegExp(r).test(s))throw new TypeError(`${this.constructor.NAME.toUpperCase()}: Option "${n}" provided type "${s}" but expected type "${r}".`)}}}class Z extends X{constructor(e,t){super(),(e=f(e))&&(this._element=e,this._config=this._getConfig(t),H.set(this._element,this.constructor.DATA_KEY,this))}dispose(){H.remove(this._element,this.constructor.DATA_KEY),V.off(this._element,this.constructor.EVENT_KEY);for(const e of Object.getOwnPropertyNames(this))this[e]=null}_queueCallback(e,t,n=!0){_(e,t,n)}_getConfig(e){return e=this._mergeConfigObj(e,this._element),e=this._configAfterMerge(e),this._typeCheckConfig(e),e}static getInstance(e){return H.get(f(e),this.DATA_KEY)}static getOrCreateInstance(e,t={}){return this.getInstance(e)||new this(e,"object"==typeof t?t:null)}static get VERSION(){return"5.2.3"}static get DATA_KEY(){return`bs.${this.NAME}`}static get EVENT_KEY(){return`.${this.DATA_KEY}`}static eventName(e){return`${e}${this.EVENT_KEY}`}}const Y=(e,t="hide")=>{const n=`click.dismiss${e.EVENT_KEY}`,r=e.NAME;V.on(document,n,`[data-bs-dismiss="${r}"]`,function(n){if(["A","AREA"].includes(this.tagName)&&n.preventDefault(),g(this))return;const a=d(this)||this.closest(`.${r}`);e.getOrCreateInstance(a)[t]()})},J=".bs.alert",Q=`close${J}`,ee=`closed${J}`;class te extends Z{static get NAME(){return"alert"}close(){if(V.trigger(this._element,Q).defaultPrevented)return;this._element.classList.remove("show");const e=this._element.classList.contains("fade");this._queueCallback(()=>this._destroyElement(),this._element,e)}_destroyElement(){this._element.remove(),V.trigger(this._element,ee),this.dispose()}static jQueryInterface(e){return this.each(function(){const t=te.getOrCreateInstance(this);if("string"==typeof e){if(void 0===t[e]||e.startsWith("_")||"constructor"===e)throw new TypeError(`No method named "${e}"`);t[e](this)}})}}Y(te,"close"),I(te);const ne='[data-bs-toggle="button"]';class re extends Z{static get NAME(){return"button"}toggle(){this._element.setAttribute("aria-pressed",this._element.classList.toggle("active"))}static jQueryInterface(e){return this.each(function(){const t=re.getOrCreateInstance(this);"toggle"===e&&t[e]()})}}V.on(document,"click.bs.button.data-api",ne,e=>{e.preventDefault();const t=e.target.closest(ne);re.getOrCreateInstance(t).toggle()}),I(re);const ae={find:(e,t=document.documentElement)=>[].concat(...Element.prototype.querySelectorAll.call(t,e)),findOne:(e,t=document.documentElement)=>Element.prototype.querySelector.call(t,e),children:(e,t)=>[].concat(...e.children).filter(e=>e.matches(t)),parents(e,t){const n=[];let r=e.parentNode.closest(t);for(;r;)n.push(r),r=r.parentNode.closest(t);return n},prev(e,t){let n=e.previousElementSibling;for(;n;){if(n.matches(t))return[n];n=n.previousElementSibling}return[]},next(e,t){let n=e.nextElementSibling;for(;n;){if(n.matches(t))return[n];n=n.nextElementSibling}return[]},focusableChildren(e){const t=["a","button","input","textarea","select","details","[tabindex]",'[contenteditable="true"]'].map(e=>`${e}:not([tabindex^="-"])`).join(",");return this.find(t,e).filter(e=>!g(e)&&m(e))}},se=".bs.swipe",ie=`touchstart${se}`,oe=`touchmove${se}`,le=`touchend${se}`,ue=`pointerdown${se}`,he=`pointerup${se}`,de={endCallback:null,leftCallback:null,rightCallback:null},pe={endCallback:"(function|null)",leftCallback:"(function|null)",rightCallback:"(function|null)"};class ce extends X{constructor(e,t){super(),this._element=e,e&&ce.isSupported()&&(this._config=this._getConfig(t),this._deltaX=0,this._supportPointerEvents=Boolean(window.PointerEvent),this._initEvents())}static get Default(){return de}static get DefaultType(){return pe}static get NAME(){return"swipe"}dispose(){V.off(this._element,se)}_start(e){this._supportPointerEvents?this._eventIsPointerPenTouch(e)&&(this._deltaX=e.clientX):this._deltaX=e.touches[0].clientX}_end(e){this._eventIsPointerPenTouch(e)&&(this._deltaX=e.clientX-this._deltaX),this._handleSwipe(),S(this._config.endCallback)}_move(e){this._deltaX=e.touches&&e.touches.length>1?0:e.touches[0].clientX-this._deltaX}_handleSwipe(){const e=Math.abs(this._deltaX);if(e<=40)return;const t=e/this._deltaX;this._deltaX=0,t&&S(t>0?this._config.rightCallback:this._config.leftCallback)}_initEvents(){this._supportPointerEvents?(V.on(this._element,ue,e=>this._start(e)),V.on(this._element,he,e=>this._end(e)),this._element.classList.add("pointer-event")):(V.on(this._element,ie,e=>this._start(e)),V.on(this._element,oe,e=>this._move(e)),V.on(this._element,le,e=>this._end(e)))}_eventIsPointerPenTouch(e){return this._supportPointerEvents&&("pen"===e.pointerType||"touch"===e.pointerType)}static isSupported(){return"ontouchstart"in document.documentElement||navigator.maxTouchPoints>0}}const fe=".bs.carousel",me=".data-api",ge="ArrowLeft",ye="ArrowRight",be="next",xe="prev",ve="left",we="right",ke=`slide${fe}`,Ie=`slid${fe}`,Se=`keydown${fe}`,_e=`mouseenter${fe}`,Ne=`mouseleave${fe}`,Te=`dragstart${fe}`,Ce=`load${fe}${me}`,Ee=`click${fe}${me}`,Ae="carousel",$e="active",Re=".active",Fe=".carousel-item",De=Re+Fe,Me={[ge]:we,[ye]:ve},Oe={interval:5e3,keyboard:!0,pause:"hover",ride:!1,touch:!0,wrap:!0},Le={interval:"(number|boolean)",keyboard:"boolean",pause:"(string|boolean)",ride:"(boolean|string)",touch:"boolean",wrap:"boolean"};class Pe extends Z{constructor(e,t){super(e,t),this._interval=null,this._activeElement=null,this._isSliding=!1,this.touchTimeout=null,this._swipeHelper=null,this._indicatorsElement=ae.findOne(".carousel-indicators",this._element),this._addEventListeners(),this._config.ride===Ae&&this.cycle()}static get Default(){return Oe}static get DefaultType(){return Le}static get NAME(){return"carousel"}next(){this._slide(be)}nextWhenVisible(){!document.hidden&&m(this._element)&&this.next()}prev(){this._slide(xe)}pause(){this._isSliding&&p(this._element),this._clearInterval()}cycle(){this._clearInterval(),this._updateInterval(),this._interval=setInterval(()=>this.nextWhenVisible(),this._config.interval)}_maybeEnableCycle(){this._config.ride&&(this._isSliding?V.one(this._element,Ie,()=>this.cycle()):this.cycle())}to(e){const t=this._getItems();if(e>t.length-1||e<0)return;if(this._isSliding)return void V.one(this._element,Ie,()=>this.to(e));const n=this._getItemIndex(this._getActive());if(n===e)return;const r=e>n?be:xe;this._slide(r,t[e])}dispose(){this._swipeHelper&&this._swipeHelper.dispose(),super.dispose()}_configAfterMerge(e){return e.defaultInterval=e.interval,e}_addEventListeners(){this._config.keyboard&&V.on(this._element,Se,e=>this._keydown(e)),"hover"===this._config.pause&&(V.on(this._element,_e,()=>this.pause()),V.on(this._element,Ne,()=>this._maybeEnableCycle())),this._config.touch&&ce.isSupported()&&this._addTouchEventListeners()}_addTouchEventListeners(){for(const e of ae.find(".carousel-item img",this._element))V.on(e,Te,e=>e.preventDefault());const e={leftCallback:()=>this._slide(this._directionToOrder(ve)),rightCallback:()=>this._slide(this._directionToOrder(we)),endCallback:()=>{"hover"===this._config.pause&&(this.pause(),this.touchTimeout&&clearTimeout(this.touchTimeout),this.touchTimeout=setTimeout(()=>this._maybeEnableCycle(),500+this._config.interval))}};this._swipeHelper=new ce(this._element,e)}_keydown(e){if(/input|textarea/i.test(e.target.tagName))return;const t=Me[e.key];t&&(e.preventDefault(),this._slide(this._directionToOrder(t)))}_getItemIndex(e){return this._getItems().indexOf(e)}_setActiveIndicatorElement(e){if(!this._indicatorsElement)return;const t=ae.findOne(Re,this._indicatorsElement);t.classList.remove($e),t.removeAttribute("aria-current");const n=ae.findOne(`[data-bs-slide-to="${e}"]`,this._indicatorsElement);n&&(n.classList.add($e),n.setAttribute("aria-current","true"))}_updateInterval(){const e=this._activeElement||this._getActive();if(!e)return;const t=Number.parseInt(e.getAttribute("data-bs-interval"),10);this._config.interval=t||this._config.defaultInterval}_slide(e,t=null){if(this._isSliding)return;const n=this._getActive(),r=e===be,a=t||N(this._getItems(),n,r,this._config.wrap);if(a===n)return;const s=this._getItemIndex(a),i=t=>V.trigger(this._element,t,{relatedTarget:a,direction:this._orderToDirection(e),from:this._getItemIndex(n),to:s});if(i(ke).defaultPrevented)return;if(!n||!a)return;const o=Boolean(this._interval);this.pause(),this._isSliding=!0,this._setActiveIndicatorElement(s),this._activeElement=a;const l=r?"carousel-item-start":"carousel-item-end",u=r?"carousel-item-next":"carousel-item-prev";a.classList.add(u),x(a),n.classList.add(l),a.classList.add(l);this._queueCallback(()=>{a.classList.remove(l,u),a.classList.add($e),n.classList.remove($e,u,l),this._isSliding=!1,i(Ie)},n,this._isAnimated()),o&&this.cycle()}_isAnimated(){return this._element.classList.contains("slide")}_getActive(){return ae.findOne(De,this._element)}_getItems(){return ae.find(Fe,this._element)}_clearInterval(){this._interval&&(clearInterval(this._interval),this._interval=null)}_directionToOrder(e){return k()?e===ve?xe:be:e===ve?be:xe}_orderToDirection(e){return k()?e===xe?ve:we:e===xe?we:ve}static jQueryInterface(e){return this.each(function(){const t=Pe.getOrCreateInstance(this,e);if("number"!=typeof e){if("string"==typeof e){if(void 0===t[e]||e.startsWith("_")||"constructor"===e)throw new TypeError(`No method named "${e}"`);t[e]()}}else t.to(e)})}}V.on(document,Ee,"[data-bs-slide], [data-bs-slide-to]",function(e){const t=d(this);if(!t||!t.classList.contains(Ae))return;e.preventDefault();const n=Pe.getOrCreateInstance(t),r=this.getAttribute("data-bs-slide-to");return r?(n.to(r),void n._maybeEnableCycle()):"next"===K.getDataAttribute(this,"slide")?(n.next(),void n._maybeEnableCycle()):(n.prev(),void n._maybeEnableCycle())}),V.on(window,Ce,()=>{const e=ae.find('[data-bs-ride="carousel"]');for(const t of e)Pe.getOrCreateInstance(t)}),I(Pe);const ze=".bs.collapse",Be=`show${ze}`,We=`shown${ze}`,Ve=`hide${ze}`,Ue=`hidden${ze}`,Ge=`click${ze}.data-api`,He="show",je="collapse",qe="collapsing",Ke=`:scope .${je} .${je}`,Xe='[data-bs-toggle="collapse"]',Ze={parent:null,toggle:!0},Ye={parent:"(null|element)",toggle:"boolean"};class Je extends Z{constructor(e,t){super(e,t),this._isTransitioning=!1,this._triggerArray=[];const n=ae.find(Xe);for(const e of n){const t=h(e),n=ae.find(t).filter(e=>e===this._element);null!==t&&n.length&&this._triggerArray.push(e)}this._initializeChildren(),this._config.parent||this._addAriaAndCollapsedClass(this._triggerArray,this._isShown()),this._config.toggle&&this.toggle()}static get Default(){return Ze}static get DefaultType(){return Ye}static get NAME(){return"collapse"}toggle(){this._isShown()?this.hide():this.show()}show(){if(this._isTransitioning||this._isShown())return;let e=[];if(this._config.parent&&(e=this._getFirstLevelChildren(".collapse.show, .collapse.collapsing").filter(e=>e!==this._element).map(e=>Je.getOrCreateInstance(e,{toggle:!1}))),e.length&&e[0]._isTransitioning)return;if(V.trigger(this._element,Be).defaultPrevented)return;for(const t of e)t.hide();const t=this._getDimension();this._element.classList.remove(je),this._element.classList.add(qe),this._element.style[t]=0,this._addAriaAndCollapsedClass(this._triggerArray,!0),this._isTransitioning=!0;const n=`scroll${t[0].toUpperCase()+t.slice(1)}`;this._queueCallback(()=>{this._isTransitioning=!1,this._element.classList.remove(qe),this._element.classList.add(je,He),this._element.style[t]="",V.trigger(this._element,We)},this._element,!0),this._element.style[t]=`${this._element[n]}px`}hide(){if(this._isTransitioning||!this._isShown())return;if(V.trigger(this._element,Ve).defaultPrevented)return;const e=this._getDimension();this._element.style[e]=`${this._element.getBoundingClientRect()[e]}px`,x(this._element),this._element.classList.add(qe),this._element.classList.remove(je,He);for(const e of this._triggerArray){const t=d(e);t&&!this._isShown(t)&&this._addAriaAndCollapsedClass([e],!1)}this._isTransitioning=!0;this._element.style[e]="",this._queueCallback(()=>{this._isTransitioning=!1,this._element.classList.remove(qe),this._element.classList.add(je),V.trigger(this._element,Ue)},this._element,!0)}_isShown(e=this._element){return e.classList.contains(He)}_configAfterMerge(e){return e.toggle=Boolean(e.toggle),e.parent=f(e.parent),e}_getDimension(){return this._element.classList.contains("collapse-horizontal")?"width":"height"}_initializeChildren(){if(!this._config.parent)return;const e=this._getFirstLevelChildren(Xe);for(const t of e){const e=d(t);e&&this._addAriaAndCollapsedClass([t],this._isShown(e))}}_getFirstLevelChildren(e){const t=ae.find(Ke,this._config.parent);return ae.find(e,this._config.parent).filter(e=>!t.includes(e))}_addAriaAndCollapsedClass(e,t){if(e.length)for(const n of e)n.classList.toggle("collapsed",!t),n.setAttribute("aria-expanded",t)}static jQueryInterface(e){const t={};return"string"==typeof e&&/show|hide/.test(e)&&(t.toggle=!1),this.each(function(){const n=Je.getOrCreateInstance(this,t);if("string"==typeof e){if(void 0===n[e])throw new TypeError(`No method named "${e}"`);n[e]()}})}}V.on(document,Ge,Xe,function(e){("A"===e.target.tagName||e.delegateTarget&&"A"===e.delegateTarget.tagName)&&e.preventDefault();const t=h(this),n=ae.find(t);for(const e of n)Je.getOrCreateInstance(e,{toggle:!1}).toggle()}),I(Je);const Qe="dropdown",et=".bs.dropdown",tt=".data-api",nt="ArrowUp",rt="ArrowDown",at=`hide${et}`,st=`hidden${et}`,it=`show${et}`,ot=`shown${et}`,lt=`click${et}${tt}`,ut=`keydown${et}${tt}`,ht=`keyup${et}${tt}`,dt="show",pt='[data-bs-toggle="dropdown"]:not(.disabled):not(:disabled)',ct=`${pt}.${dt}`,ft=".dropdown-menu",mt=k()?"top-end":"top-start",gt=k()?"top-start":"top-end",yt=k()?"bottom-end":"bottom-start",bt=k()?"bottom-start":"bottom-end",xt=k()?"left-start":"right-start",vt=k()?"right-start":"left-start",wt={autoClose:!0,boundary:"clippingParents",display:"dynamic",offset:[0,2],popperConfig:null,reference:"toggle"},kt={autoClose:"(boolean|string)",boundary:"(string|element)",display:"string",offset:"(array|string|function)",popperConfig:"(null|object|function)",reference:"(string|element|object)"};class It extends Z{constructor(e,t){super(e,t),this._popper=null,this._parent=this._element.parentNode,this._menu=ae.next(this._element,ft)[0]||ae.prev(this._element,ft)[0]||ae.findOne(ft,this._parent),this._inNavbar=this._detectNavbar()}static get Default(){return wt}static get DefaultType(){return kt}static get NAME(){return Qe}toggle(){return this._isShown()?this.hide():this.show()}show(){if(g(this._element)||this._isShown())return;const e={relatedTarget:this._element};if(!V.trigger(this._element,it,e).defaultPrevented){if(this._createPopper(),"ontouchstart"in document.documentElement&&!this._parent.closest(".navbar-nav"))for(const e of[].concat(...document.body.children))V.on(e,"mouseover",b);this._element.focus(),this._element.setAttribute("aria-expanded",!0),this._menu.classList.add(dt),this._element.classList.add(dt),V.trigger(this._element,ot,e)}}hide(){if(g(this._element)||!this._isShown())return;const e={relatedTarget:this._element};this._completeHide(e)}dispose(){this._popper&&this._popper.destroy(),super.dispose()}update(){this._inNavbar=this._detectNavbar(),this._popper&&this._popper.update()}_completeHide(e){if(!V.trigger(this._element,at,e).defaultPrevented){if("ontouchstart"in document.documentElement)for(const e of[].concat(...document.body.children))V.off(e,"mouseover",b);this._popper&&this._popper.destroy(),this._menu.classList.remove(dt),this._element.classList.remove(dt),this._element.setAttribute("aria-expanded","false"),K.removeDataAttribute(this._menu,"popper"),V.trigger(this._element,st,e)}}_getConfig(e){if("object"==typeof(e=super._getConfig(e)).reference&&!c(e.reference)&&"function"!=typeof e.reference.getBoundingClientRect)throw new TypeError(`${Qe.toUpperCase()}: Option "reference" provided type "object" without a required "getBoundingClientRect" method.`);return e}_createPopper(){let e=this._element;"parent"===this._config.reference?e=this._parent:c(this._config.reference)?e=f(this._config.reference):"object"==typeof this._config.reference&&(e=this._config.reference);const t=this._getPopperConfig();this._popper=a.n(e,this._menu,t)}_isShown(){return this._menu.classList.contains(dt)}_getPlacement(){const e=this._parent;if(e.classList.contains("dropend"))return xt;if(e.classList.contains("dropstart"))return vt;if(e.classList.contains("dropup-center"))return"top";if(e.classList.contains("dropdown-center"))return"bottom";const t="end"===getComputedStyle(this._menu).getPropertyValue("--bs-position").trim();return e.classList.contains("dropup")?t?gt:mt:t?bt:yt}_detectNavbar(){return null!==this._element.closest(".navbar")}_getOffset(){const{offset:e}=this._config;return"string"==typeof e?e.split(",").map(e=>Number.parseInt(e,10)):"function"==typeof e?t=>e(t,this._element):e}_getPopperConfig(){const e={placement:this._getPlacement(),modifiers:[{name:"preventOverflow",options:{boundary:this._config.boundary}},{name:"offset",options:{offset:this._getOffset()}}]};return(this._inNavbar||"static"===this._config.display)&&(K.setDataAttribute(this._menu,"popper","static"),e.modifiers=[{name:"applyStyles",enabled:!1}]),{...e,..."function"==typeof this._config.popperConfig?this._config.popperConfig(e):this._config.popperConfig}}_selectMenuItem({key:e,target:t}){const n=ae.find(".dropdown-menu .dropdown-item:not(.disabled):not(:disabled)",this._menu).filter(e=>m(e));n.length&&N(n,t,e===rt,!n.includes(t)).focus()}static jQueryInterface(e){return this.each(function(){const t=It.getOrCreateInstance(this,e);if("string"==typeof e){if(void 0===t[e])throw new TypeError(`No method named "${e}"`);t[e]()}})}static clearMenus(e){if(2===e.button||"keyup"===e.type&&"Tab"!==e.key)return;const t=ae.find(ct);for(const n of t){const t=It.getInstance(n);if(!t||!1===t._config.autoClose)continue;const r=e.composedPath(),a=r.includes(t._menu);if(r.includes(t._element)||"inside"===t._config.autoClose&&!a||"outside"===t._config.autoClose&&a)continue;if(t._menu.contains(e.target)&&("keyup"===e.type&&"Tab"===e.key||/input|select|option|textarea|form/i.test(e.target.tagName)))continue;const s={relatedTarget:t._element};"click"===e.type&&(s.clickEvent=e),t._completeHide(s)}}static dataApiKeydownHandler(e){const t=/input|textarea/i.test(e.target.tagName),n="Escape"===e.key,r=[nt,rt].includes(e.key);if(!r&&!n)return;if(t&&!n)return;e.preventDefault();const a=this.matches(pt)?this:ae.prev(this,pt)[0]||ae.next(this,pt)[0]||ae.findOne(pt,e.delegateTarget.parentNode),s=It.getOrCreateInstance(a);if(r)return e.stopPropagation(),s.show(),void s._selectMenuItem(e);s._isShown()&&(e.stopPropagation(),s.hide(),a.focus())}}V.on(document,ut,pt,It.dataApiKeydownHandler),V.on(document,ut,ft,It.dataApiKeydownHandler),V.on(document,lt,It.clearMenus),V.on(document,ht,It.clearMenus),V.on(document,lt,pt,function(e){e.preventDefault(),It.getOrCreateInstance(this).toggle()}),I(It);const St=".fixed-top, .fixed-bottom, .is-fixed, .sticky-top",_t=".sticky-top",Nt="padding-right",Tt="margin-right";class Ct{constructor(){this._element=document.body}getWidth(){const e=document.documentElement.clientWidth;return Math.abs(window.innerWidth-e)}hide(){const e=this.getWidth();this._disableOverFlow(),this._setElementAttributes(this._element,Nt,t=>t+e),this._setElementAttributes(St,Nt,t=>t+e),this._setElementAttributes(_t,Tt,t=>t-e)}reset(){this._resetElementAttributes(this._element,"overflow"),this._resetElementAttributes(this._element,Nt),this._resetElementAttributes(St,Nt),this._resetElementAttributes(_t,Tt)}isOverflowing(){return this.getWidth()>0}_disableOverFlow(){this._saveInitialAttribute(this._element,"overflow"),this._element.style.overflow="hidden"}_setElementAttributes(e,t,n){const r=this.getWidth();this._applyManipulationCallback(e,e=>{if(e!==this._element&&window.innerWidth>e.clientWidth+r)return;this._saveInitialAttribute(e,t);const a=window.getComputedStyle(e).getPropertyValue(t);e.style.setProperty(t,`${n(Number.parseFloat(a))}px`)})}_saveInitialAttribute(e,t){const n=e.style.getPropertyValue(t);n&&K.setDataAttribute(e,t,n)}_resetElementAttributes(e,t){this._applyManipulationCallback(e,e=>{const n=K.getDataAttribute(e,t);null!==n?(K.removeDataAttribute(e,t),e.style.setProperty(t,n)):e.style.removeProperty(t)})}_applyManipulationCallback(e,t){if(c(e))t(e);else for(const n of ae.find(e,this._element))t(n)}}const Et="backdrop",At="show",$t=`mousedown.bs.${Et}`,Rt={className:"modal-backdrop",clickCallback:null,isAnimated:!1,isVisible:!0,rootElement:"body"},Ft={className:"string",clickCallback:"(function|null)",isAnimated:"boolean",isVisible:"boolean",rootElement:"(element|string)"};class Dt extends X{constructor(e){super(),this._config=this._getConfig(e),this._isAppended=!1,this._element=null}static get Default(){return Rt}static get DefaultType(){return Ft}static get NAME(){return Et}show(e){if(!this._config.isVisible)return void S(e);this._append();const t=this._getElement();this._config.isAnimated&&x(t),t.classList.add(At),this._emulateAnimation(()=>{S(e)})}hide(e){this._config.isVisible?(this._getElement().classList.remove(At),this._emulateAnimation(()=>{this.dispose(),S(e)})):S(e)}dispose(){this._isAppended&&(V.off(this._element,$t),this._element.remove(),this._isAppended=!1)}_getElement(){if(!this._element){const e=document.createElement("div");e.className=this._config.className,this._config.isAnimated&&e.classList.add("fade"),this._element=e}return this._element}_configAfterMerge(e){return e.rootElement=f(e.rootElement),e}_append(){if(this._isAppended)return;const e=this._getElement();this._config.rootElement.append(e),V.on(e,$t,()=>{S(this._config.clickCallback)}),this._isAppended=!0}_emulateAnimation(e){_(e,this._getElement(),this._config.isAnimated)}}const Mt=".bs.focustrap",Ot=`focusin${Mt}`,Lt=`keydown.tab${Mt}`,Pt="backward",zt={autofocus:!0,trapElement:null},Bt={autofocus:"boolean",trapElement:"element"};class Wt extends X{constructor(e){super(),this._config=this._getConfig(e),this._isActive=!1,this._lastTabNavDirection=null}static get Default(){return zt}static get DefaultType(){return Bt}static get NAME(){return"focustrap"}activate(){this._isActive||(this._config.autofocus&&this._config.trapElement.focus(),V.off(document,Mt),V.on(document,Ot,e=>this._handleFocusin(e)),V.on(document,Lt,e=>this._handleKeydown(e)),this._isActive=!0)}deactivate(){this._isActive&&(this._isActive=!1,V.off(document,Mt))}_handleFocusin(e){const{trapElement:t}=this._config;if(e.target===document||e.target===t||t.contains(e.target))return;const n=ae.focusableChildren(t);0===n.length?t.focus():this._lastTabNavDirection===Pt?n[n.length-1].focus():n[0].focus()}_handleKeydown(e){"Tab"===e.key&&(this._lastTabNavDirection=e.shiftKey?Pt:"forward")}}const Vt=".bs.modal",Ut=`hide${Vt}`,Gt=`hidePrevented${Vt}`,Ht=`hidden${Vt}`,jt=`show${Vt}`,qt=`shown${Vt}`,Kt=`resize${Vt}`,Xt=`click.dismiss${Vt}`,Zt=`mousedown.dismiss${Vt}`,Yt=`keydown.dismiss${Vt}`,Jt=`click${Vt}.data-api`,Qt="modal-open",en="show",tn="modal-static",nn={backdrop:!0,focus:!0,keyboard:!0},rn={backdrop:"(boolean|string)",focus:"boolean",keyboard:"boolean"};class an extends Z{constructor(e,t){super(e,t),this._dialog=ae.findOne(".modal-dialog",this._element),this._backdrop=this._initializeBackDrop(),this._focustrap=this._initializeFocusTrap(),this._isShown=!1,this._isTransitioning=!1,this._scrollBar=new Ct,this._addEventListeners()}static get Default(){return nn}static get DefaultType(){return rn}static get NAME(){return"modal"}toggle(e){return this._isShown?this.hide():this.show(e)}show(e){if(this._isShown||this._isTransitioning)return;V.trigger(this._element,jt,{relatedTarget:e}).defaultPrevented||(this._isShown=!0,this._isTransitioning=!0,this._scrollBar.hide(),document.body.classList.add(Qt),this._adjustDialog(),this._backdrop.show(()=>this._showElement(e)))}hide(){if(!this._isShown||this._isTransitioning)return;V.trigger(this._element,Ut).defaultPrevented||(this._isShown=!1,this._isTransitioning=!0,this._focustrap.deactivate(),this._element.classList.remove(en),this._queueCallback(()=>this._hideModal(),this._element,this._isAnimated()))}dispose(){for(const e of[window,this._dialog])V.off(e,Vt);this._backdrop.dispose(),this._focustrap.deactivate(),super.dispose()}handleUpdate(){this._adjustDialog()}_initializeBackDrop(){return new Dt({isVisible:Boolean(this._config.backdrop),isAnimated:this._isAnimated()})}_initializeFocusTrap(){return new Wt({trapElement:this._element})}_showElement(e){document.body.contains(this._element)||document.body.append(this._element),this._element.style.display="block",this._element.removeAttribute("aria-hidden"),this._element.setAttribute("aria-modal",!0),this._element.setAttribute("role","dialog"),this._element.scrollTop=0;const t=ae.findOne(".modal-body",this._dialog);t&&(t.scrollTop=0),x(this._element),this._element.classList.add(en);this._queueCallback(()=>{this._config.focus&&this._focustrap.activate(),this._isTransitioning=!1,V.trigger(this._element,qt,{relatedTarget:e})},this._dialog,this._isAnimated())}_addEventListeners(){V.on(this._element,Yt,e=>{if("Escape"===e.key)return this._config.keyboard?(e.preventDefault(),void this.hide()):void this._triggerBackdropTransition()}),V.on(window,Kt,()=>{this._isShown&&!this._isTransitioning&&this._adjustDialog()}),V.on(this._element,Zt,e=>{V.one(this._element,Xt,t=>{this._element===e.target&&this._element===t.target&&("static"!==this._config.backdrop?this._config.backdrop&&this.hide():this._triggerBackdropTransition())})})}_hideModal(){this._element.style.display="none",this._element.setAttribute("aria-hidden",!0),this._element.removeAttribute("aria-modal"),this._element.removeAttribute("role"),this._isTransitioning=!1,this._backdrop.hide(()=>{document.body.classList.remove(Qt),this._resetAdjustments(),this._scrollBar.reset(),V.trigger(this._element,Ht)})}_isAnimated(){return this._element.classList.contains("fade")}_triggerBackdropTransition(){if(V.trigger(this._element,Gt).defaultPrevented)return;const e=this._element.scrollHeight>document.documentElement.clientHeight,t=this._element.style.overflowY;"hidden"===t||this._element.classList.contains(tn)||(e||(this._element.style.overflowY="hidden"),this._element.classList.add(tn),this._queueCallback(()=>{this._element.classList.remove(tn),this._queueCallback(()=>{this._element.style.overflowY=t},this._dialog)},this._dialog),this._element.focus())}_adjustDialog(){const e=this._element.scrollHeight>document.documentElement.clientHeight,t=this._scrollBar.getWidth(),n=t>0;if(n&&!e){const e=k()?"paddingLeft":"paddingRight";this._element.style[e]=`${t}px`}if(!n&&e){const e=k()?"paddingRight":"paddingLeft";this._element.style[e]=`${t}px`}}_resetAdjustments(){this._element.style.paddingLeft="",this._element.style.paddingRight=""}static jQueryInterface(e,t){return this.each(function(){const n=an.getOrCreateInstance(this,e);if("string"==typeof e){if(void 0===n[e])throw new TypeError(`No method named "${e}"`);n[e](t)}})}}V.on(document,Jt,'[data-bs-toggle="modal"]',function(e){const t=d(this);["A","AREA"].includes(this.tagName)&&e.preventDefault(),V.one(t,jt,e=>{e.defaultPrevented||V.one(t,Ht,()=>{m(this)&&this.focus()})});const n=ae.findOne(".modal.show");n&&an.getInstance(n).hide();an.getOrCreateInstance(t).toggle(this)}),Y(an),I(an);const sn=".bs.offcanvas",on=".data-api",ln=`load${sn}${on}`,un="show",hn="showing",dn="hiding",pn=".offcanvas.show",cn=`show${sn}`,fn=`shown${sn}`,mn=`hide${sn}`,gn=`hidePrevented${sn}`,yn=`hidden${sn}`,bn=`resize${sn}`,xn=`click${sn}${on}`,vn=`keydown.dismiss${sn}`,wn={backdrop:!0,keyboard:!0,scroll:!1},kn={backdrop:"(boolean|string)",keyboard:"boolean",scroll:"boolean"};class In extends Z{constructor(e,t){super(e,t),this._isShown=!1,this._backdrop=this._initializeBackDrop(),this._focustrap=this._initializeFocusTrap(),this._addEventListeners()}static get Default(){return wn}static get DefaultType(){return kn}static get NAME(){return"offcanvas"}toggle(e){return this._isShown?this.hide():this.show(e)}show(e){if(this._isShown)return;if(V.trigger(this._element,cn,{relatedTarget:e}).defaultPrevented)return;this._isShown=!0,this._backdrop.show(),this._config.scroll||(new Ct).hide(),this._element.setAttribute("aria-modal",!0),this._element.setAttribute("role","dialog"),this._element.classList.add(hn);this._queueCallback(()=>{this._config.scroll&&!this._config.backdrop||this._focustrap.activate(),this._element.classList.add(un),this._element.classList.remove(hn),V.trigger(this._element,fn,{relatedTarget:e})},this._element,!0)}hide(){if(!this._isShown)return;if(V.trigger(this._element,mn).defaultPrevented)return;this._focustrap.deactivate(),this._element.blur(),this._isShown=!1,this._element.classList.add(dn),this._backdrop.hide();this._queueCallback(()=>{this._element.classList.remove(un,dn),this._element.removeAttribute("aria-modal"),this._element.removeAttribute("role"),this._config.scroll||(new Ct).reset(),V.trigger(this._element,yn)},this._element,!0)}dispose(){this._backdrop.dispose(),this._focustrap.deactivate(),super.dispose()}_initializeBackDrop(){const e=Boolean(this._config.backdrop);return new Dt({className:"offcanvas-backdrop",isVisible:e,isAnimated:!0,rootElement:this._element.parentNode,clickCallback:e?()=>{"static"!==this._config.backdrop?this.hide():V.trigger(this._element,gn)}:null})}_initializeFocusTrap(){return new Wt({trapElement:this._element})}_addEventListeners(){V.on(this._element,vn,e=>{"Escape"===e.key&&(this._config.keyboard?this.hide():V.trigger(this._element,gn))})}static jQueryInterface(e){return this.each(function(){const t=In.getOrCreateInstance(this,e);if("string"==typeof e){if(void 0===t[e]||e.startsWith("_")||"constructor"===e)throw new TypeError(`No method named "${e}"`);t[e](this)}})}}V.on(document,xn,'[data-bs-toggle="offcanvas"]',function(e){const t=d(this);if(["A","AREA"].includes(this.tagName)&&e.preventDefault(),g(this))return;V.one(t,yn,()=>{m(this)&&this.focus()});const n=ae.findOne(pn);n&&n!==t&&In.getInstance(n).hide();In.getOrCreateInstance(t).toggle(this)}),V.on(window,ln,()=>{for(const e of ae.find(pn))In.getOrCreateInstance(e).show()}),V.on(window,bn,()=>{for(const e of ae.find("[aria-modal][class*=show][class*=offcanvas-]"))"fixed"!==getComputedStyle(e).position&&In.getOrCreateInstance(e).hide()}),Y(In),I(In);const Sn=new Set(["background","cite","href","itemtype","longdesc","poster","src","xlink:href"]),_n=/^(?:(?:https?|mailto|ftp|tel|file|sms):|[^#&/:?]*(?:[#/?]|$))/i,Nn=/^data:(?:image\/(?:bmp|gif|jpeg|jpg|png|tiff|webp)|video\/(?:mpeg|mp4|ogg|webm)|audio\/(?:mp3|oga|ogg|opus));base64,[\d+/a-z]+=*$/i,Tn=(e,t)=>{const n=e.nodeName.toLowerCase();return t.includes(n)?!Sn.has(n)||Boolean(_n.test(e.nodeValue)||Nn.test(e.nodeValue)):t.filter(e=>e instanceof RegExp).some(e=>e.test(n))},Cn={"*":["class","dir","id","lang","role",/^aria-[\w-]*$/i],a:["target","href","title","rel"],area:[],b:[],br:[],col:[],code:[],div:[],em:[],hr:[],h1:[],h2:[],h3:[],h4:[],h5:[],h6:[],i:[],img:["src","srcset","alt","title","width","height"],li:[],ol:[],p:[],pre:[],s:[],small:[],span:[],sub:[],sup:[],strong:[],u:[],ul:[]};const En={allowList:Cn,content:{},extraClass:"",html:!1,sanitize:!0,sanitizeFn:null,template:"<div></div>"},An={allowList:"object",content:"object",extraClass:"(string|function)",html:"boolean",sanitize:"boolean",sanitizeFn:"(null|function)",template:"string"},$n={entry:"(string|element|function|null)",selector:"(string|element)"};class Rn extends X{constructor(e){super(),this._config=this._getConfig(e)}static get Default(){return En}static get DefaultType(){return An}static get NAME(){return"TemplateFactory"}getContent(){return Object.values(this._config.content).map(e=>this._resolvePossibleFunction(e)).filter(Boolean)}hasContent(){return this.getContent().length>0}changeContent(e){return this._checkContent(e),this._config.content={...this._config.content,...e},this}toHtml(){const e=document.createElement("div");e.innerHTML=this._maybeSanitize(this._config.template);for(const[t,n]of Object.entries(this._config.content))this._setContent(e,n,t);const t=e.children[0],n=this._resolvePossibleFunction(this._config.extraClass);return n&&t.classList.add(...n.split(" ")),t}_typeCheckConfig(e){super._typeCheckConfig(e),this._checkContent(e.content)}_checkContent(e){for(const[t,n]of Object.entries(e))super._typeCheckConfig({selector:t,entry:n},$n)}_setContent(e,t,n){const r=ae.findOne(n,e);r&&((t=this._resolvePossibleFunction(t))?c(t)?this._putElementInTemplate(f(t),r):this._config.html?r.innerHTML=this._maybeSanitize(t):r.textContent=t:r.remove())}_maybeSanitize(e){return this._config.sanitize?function(e,t,n){if(!e.length)return e;if(n&&"function"==typeof n)return n(e);const r=(new window.DOMParser).parseFromString(e,"text/html"),a=[].concat(...r.body.querySelectorAll("*"));for(const e of a){const n=e.nodeName.toLowerCase();if(!Object.keys(t).includes(n)){e.remove();continue}const r=[].concat(...e.attributes),a=[].concat(t["*"]||[],t[n]||[]);for(const t of r)Tn(t,a)||e.removeAttribute(t.nodeName)}return r.body.innerHTML}(e,this._config.allowList,this._config.sanitizeFn):e}_resolvePossibleFunction(e){return"function"==typeof e?e(this):e}_putElementInTemplate(e,t){if(this._config.html)return t.innerHTML="",void t.append(e);t.textContent=e.textContent}}const Fn=new Set(["sanitize","allowList","sanitizeFn"]),Dn="fade",Mn="show",On=".tooltip-inner",Ln=".modal",Pn="hide.bs.modal",zn="hover",Bn="focus",Wn={AUTO:"auto",TOP:"top",RIGHT:k()?"left":"right",BOTTOM:"bottom",LEFT:k()?"right":"left"},Vn={allowList:Cn,animation:!0,boundary:"clippingParents",container:!1,customClass:"",delay:0,fallbackPlacements:["top","right","bottom","left"],html:!1,offset:[0,0],placement:"top",popperConfig:null,sanitize:!0,sanitizeFn:null,selector:!1,template:'<div class="tooltip" role="tooltip"><div class="tooltip-arrow"></div><div class="tooltip-inner"></div></div>',title:"",trigger:"hover focus"},Un={allowList:"object",animation:"boolean",boundary:"(string|element)",container:"(string|element|boolean)",customClass:"(string|function)",delay:"(number|object)",fallbackPlacements:"array",html:"boolean",offset:"(array|string|function)",placement:"(string|function)",popperConfig:"(null|object|function)",sanitize:"boolean",sanitizeFn:"(null|function)",selector:"(string|boolean)",template:"string",title:"(string|element|function)",trigger:"string"};class Gn extends Z{constructor(e,t){super(e,t),this._isEnabled=!0,this._timeout=0,this._isHovered=null,this._activeTrigger={},this._popper=null,this._templateFactory=null,this._newContent=null,this.tip=null,this._setListeners(),this._config.selector||this._fixTitle()}static get Default(){return Vn}static get DefaultType(){return Un}static get NAME(){return"tooltip"}enable(){this._isEnabled=!0}disable(){this._isEnabled=!1}toggleEnabled(){this._isEnabled=!this._isEnabled}toggle(){this._isEnabled&&(this._activeTrigger.click=!this._activeTrigger.click,this._isShown()?this._leave():this._enter())}dispose(){clearTimeout(this._timeout),V.off(this._element.closest(Ln),Pn,this._hideModalHandler),this._element.getAttribute("data-bs-original-title")&&this._element.setAttribute("title",this._element.getAttribute("data-bs-original-title")),this._disposePopper(),super.dispose()}show(){if("none"===this._element.style.display)throw new Error("Please use show on visible elements");if(!this._isWithContent()||!this._isEnabled)return;const e=V.trigger(this._element,this.constructor.eventName("show")),t=(y(this._element)||this._element.ownerDocument.documentElement).contains(this._element);if(e.defaultPrevented||!t)return;this._disposePopper();const n=this._getTipElement();this._element.setAttribute("aria-describedby",n.getAttribute("id"));const{container:r}=this._config;if(this._element.ownerDocument.documentElement.contains(this.tip)||(r.append(n),V.trigger(this._element,this.constructor.eventName("inserted"))),this._popper=this._createPopper(n),n.classList.add(Mn),"ontouchstart"in document.documentElement)for(const e of[].concat(...document.body.children))V.on(e,"mouseover",b);this._queueCallback(()=>{V.trigger(this._element,this.constructor.eventName("shown")),!1===this._isHovered&&this._leave(),this._isHovered=!1},this.tip,this._isAnimated())}hide(){if(!this._isShown())return;if(V.trigger(this._element,this.constructor.eventName("hide")).defaultPrevented)return;if(this._getTipElement().classList.remove(Mn),"ontouchstart"in document.documentElement)for(const e of[].concat(...document.body.children))V.off(e,"mouseover",b);this._activeTrigger.click=!1,this._activeTrigger[Bn]=!1,this._activeTrigger[zn]=!1,this._isHovered=null;this._queueCallback(()=>{this._isWithActiveTrigger()||(this._isHovered||this._disposePopper(),this._element.removeAttribute("aria-describedby"),V.trigger(this._element,this.constructor.eventName("hidden")))},this.tip,this._isAnimated())}update(){this._popper&&this._popper.update()}_isWithContent(){return Boolean(this._getTitle())}_getTipElement(){return this.tip||(this.tip=this._createTipElement(this._newContent||this._getContentForTemplate())),this.tip}_createTipElement(e){const t=this._getTemplateFactory(e).toHtml();if(!t)return null;t.classList.remove(Dn,Mn),t.classList.add(`bs-${this.constructor.NAME}-auto`);const n=(e=>{do{e+=Math.floor(1e6*Math.random())}while(document.getElementById(e));return e})(this.constructor.NAME).toString();return t.setAttribute("id",n),this._isAnimated()&&t.classList.add(Dn),t}setContent(e){this._newContent=e,this._isShown()&&(this._disposePopper(),this.show())}_getTemplateFactory(e){return this._templateFactory?this._templateFactory.changeContent(e):this._templateFactory=new Rn({...this._config,content:e,extraClass:this._resolvePossibleFunction(this._config.customClass)}),this._templateFactory}_getContentForTemplate(){return{[On]:this._getTitle()}}_getTitle(){return this._resolvePossibleFunction(this._config.title)||this._element.getAttribute("data-bs-original-title")}_initializeOnDelegatedTarget(e){return this.constructor.getOrCreateInstance(e.delegateTarget,this._getDelegateConfig())}_isAnimated(){return this._config.animation||this.tip&&this.tip.classList.contains(Dn)}_isShown(){return this.tip&&this.tip.classList.contains(Mn)}_createPopper(e){const t="function"==typeof this._config.placement?this._config.placement.call(this,e,this._element):this._config.placement,n=Wn[t.toUpperCase()];return a.n(this._element,e,this._getPopperConfig(n))}_getOffset(){const{offset:e}=this._config;return"string"==typeof e?e.split(",").map(e=>Number.parseInt(e,10)):"function"==typeof e?t=>e(t,this._element):e}_resolvePossibleFunction(e){return"function"==typeof e?e.call(this._element):e}_getPopperConfig(e){const t={placement:e,modifiers:[{name:"flip",options:{fallbackPlacements:this._config.fallbackPlacements}},{name:"offset",options:{offset:this._getOffset()}},{name:"preventOverflow",options:{boundary:this._config.boundary}},{name:"arrow",options:{element:`.${this.constructor.NAME}-arrow`}},{name:"preSetPlacement",enabled:!0,phase:"beforeMain",fn:e=>{this._getTipElement().setAttribute("data-popper-placement",e.state.placement)}}]};return{...t,..."function"==typeof this._config.popperConfig?this._config.popperConfig(t):this._config.popperConfig}}_setListeners(){const e=this._config.trigger.split(" ");for(const t of e)if("click"===t)V.on(this._element,this.constructor.eventName("click"),this._config.selector,e=>{this._initializeOnDelegatedTarget(e).toggle()});else if("manual"!==t){const e=t===zn?this.constructor.eventName("mouseenter"):this.constructor.eventName("focusin"),n=t===zn?this.constructor.eventName("mouseleave"):this.constructor.eventName("focusout");V.on(this._element,e,this._config.selector,e=>{const t=this._initializeOnDelegatedTarget(e);t._activeTrigger["focusin"===e.type?Bn:zn]=!0,t._enter()}),V.on(this._element,n,this._config.selector,e=>{const t=this._initializeOnDelegatedTarget(e);t._activeTrigger["focusout"===e.type?Bn:zn]=t._element.contains(e.relatedTarget),t._leave()})}this._hideModalHandler=()=>{this._element&&this.hide()},V.on(this._element.closest(Ln),Pn,this._hideModalHandler)}_fixTitle(){const e=this._element.getAttribute("title");e&&(this._element.getAttribute("aria-label")||this._element.textContent.trim()||this._element.setAttribute("aria-label",e),this._element.setAttribute("data-bs-original-title",e),this._element.removeAttribute("title"))}_enter(){this._isShown()||this._isHovered?this._isHovered=!0:(this._isHovered=!0,this._setTimeout(()=>{this._isHovered&&this.show()},this._config.delay.show))}_leave(){this._isWithActiveTrigger()||(this._isHovered=!1,this._setTimeout(()=>{this._isHovered||this.hide()},this._config.delay.hide))}_setTimeout(e,t){clearTimeout(this._timeout),this._timeout=setTimeout(e,t)}_isWithActiveTrigger(){return Object.values(this._activeTrigger).includes(!0)}_getConfig(e){const t=K.getDataAttributes(this._element);for(const e of Object.keys(t))Fn.has(e)&&delete t[e];return e={...t,..."object"==typeof e&&e?e:{}},e=this._mergeConfigObj(e),e=this._configAfterMerge(e),this._typeCheckConfig(e),e}_configAfterMerge(e){return e.container=!1===e.container?document.body:f(e.container),"number"==typeof e.delay&&(e.delay={show:e.delay,hide:e.delay}),"number"==typeof e.title&&(e.title=e.title.toString()),"number"==typeof e.content&&(e.content=e.content.toString()),e}_getDelegateConfig(){const e={};for(const t in this._config)this.constructor.Default[t]!==this._config[t]&&(e[t]=this._config[t]);return e.selector=!1,e.trigger="manual",e}_disposePopper(){this._popper&&(this._popper.destroy(),this._popper=null),this.tip&&(this.tip.remove(),this.tip=null)}static jQueryInterface(e){return this.each(function(){const t=Gn.getOrCreateInstance(this,e);if("string"==typeof e){if(void 0===t[e])throw new TypeError(`No method named "${e}"`);t[e]()}})}}I(Gn);const Hn=".popover-header",jn=".popover-body",qn={...Gn.Default,content:"",offset:[0,8],placement:"right",template:'<div class="popover" role="tooltip"><div class="popover-arrow"></div><h3 class="popover-header"></h3><div class="popover-body"></div></div>',trigger:"click"},Kn={...Gn.DefaultType,content:"(null|string|element|function)"};class Xn extends Gn{static get Default(){return qn}static get DefaultType(){return Kn}static get NAME(){return"popover"}_isWithContent(){return this._getTitle()||this._getContent()}_getContentForTemplate(){return{[Hn]:this._getTitle(),[jn]:this._getContent()}}_getContent(){return this._resolvePossibleFunction(this._config.content)}static jQueryInterface(e){return this.each(function(){const t=Xn.getOrCreateInstance(this,e);if("string"==typeof e){if(void 0===t[e])throw new TypeError(`No method named "${e}"`);t[e]()}})}}I(Xn);const Zn=".bs.scrollspy",Yn=`activate${Zn}`,Jn=`click${Zn}`,Qn=`load${Zn}.data-api`,er="active",tr="[href]",nr=".nav-link",rr=`${nr}, .nav-item > ${nr}, .list-group-item`,ar={offset:null,rootMargin:"0px 0px -25%",smoothScroll:!1,target:null,threshold:[.1,.5,1]},sr={offset:"(number|null)",rootMargin:"string",smoothScroll:"boolean",target:"element",threshold:"array"};class ir extends Z{constructor(e,t){super(e,t),this._targetLinks=new Map,this._observableSections=new Map,this._rootElement="visible"===getComputedStyle(this._element).overflowY?null:this._element,this._activeTarget=null,this._observer=null,this._previousScrollData={visibleEntryTop:0,parentScrollTop:0},this.refresh()}static get Default(){return ar}static get DefaultType(){return sr}static get NAME(){return"scrollspy"}refresh(){this._initializeTargetsAndObservables(),this._maybeEnableSmoothScroll(),this._observer?this._observer.disconnect():this._observer=this._getNewObserver();for(const e of this._observableSections.values())this._observer.observe(e)}dispose(){this._observer.disconnect(),super.dispose()}_configAfterMerge(e){return e.target=f(e.target)||document.body,e.rootMargin=e.offset?`${e.offset}px 0px -30%`:e.rootMargin,"string"==typeof e.threshold&&(e.threshold=e.threshold.split(",").map(e=>Number.parseFloat(e))),e}_maybeEnableSmoothScroll(){this._config.smoothScroll&&(V.off(this._config.target,Jn),V.on(this._config.target,Jn,tr,e=>{const t=this._observableSections.get(e.target.hash);if(t){e.preventDefault();const n=this._rootElement||window,r=t.offsetTop-this._element.offsetTop;if(n.scrollTo)return void n.scrollTo({top:r,behavior:"smooth"});n.scrollTop=r}}))}_getNewObserver(){const e={root:this._rootElement,threshold:this._config.threshold,rootMargin:this._config.rootMargin};return new IntersectionObserver(e=>this._observerCallback(e),e)}_observerCallback(e){const t=e=>this._targetLinks.get(`#${e.target.id}`),n=e=>{this._previousScrollData.visibleEntryTop=e.target.offsetTop,this._process(t(e))},r=(this._rootElement||document.documentElement).scrollTop,a=r>=this._previousScrollData.parentScrollTop;this._previousScrollData.parentScrollTop=r;for(const s of e){if(!s.isIntersecting){this._activeTarget=null,this._clearActiveClass(t(s));continue}const e=s.target.offsetTop>=this._previousScrollData.visibleEntryTop;if(a&&e){if(n(s),!r)return}else a||e||n(s)}}_initializeTargetsAndObservables(){this._targetLinks=new Map,this._observableSections=new Map;const e=ae.find(tr,this._config.target);for(const t of e){if(!t.hash||g(t))continue;const e=ae.findOne(t.hash,this._element);m(e)&&(this._targetLinks.set(t.hash,t),this._observableSections.set(t.hash,e))}}_process(e){this._activeTarget!==e&&(this._clearActiveClass(this._config.target),this._activeTarget=e,e.classList.add(er),this._activateParents(e),V.trigger(this._element,Yn,{relatedTarget:e}))}_activateParents(e){if(e.classList.contains("dropdown-item"))ae.findOne(".dropdown-toggle",e.closest(".dropdown")).classList.add(er);else for(const t of ae.parents(e,".nav, .list-group"))for(const e of ae.prev(t,rr))e.classList.add(er)}_clearActiveClass(e){e.classList.remove(er);const t=ae.find(`${tr}.${er}`,e);for(const e of t)e.classList.remove(er)}static jQueryInterface(e){return this.each(function(){const t=ir.getOrCreateInstance(this,e);if("string"==typeof e){if(void 0===t[e]||e.startsWith("_")||"constructor"===e)throw new TypeError(`No method named "${e}"`);t[e]()}})}}V.on(window,Qn,()=>{for(const e of ae.find('[data-bs-spy="scroll"]'))ir.getOrCreateInstance(e)}),I(ir);const or=".bs.tab",lr=`hide${or}`,ur=`hidden${or}`,hr=`show${or}`,dr=`shown${or}`,pr=`click${or}`,cr=`keydown${or}`,fr=`load${or}`,mr="ArrowLeft",gr="ArrowRight",yr="ArrowUp",br="ArrowDown",xr="active",vr="fade",wr="show",kr=":not(.dropdown-toggle)",Ir='[data-bs-toggle="tab"], [data-bs-toggle="pill"], [data-bs-toggle="list"]',Sr=`${`.nav-link${kr}, .list-group-item${kr}, [role="tab"]${kr}`}, ${Ir}`,_r=`.${xr}[data-bs-toggle="tab"], .${xr}[data-bs-toggle="pill"], .${xr}[data-bs-toggle="list"]`;class Nr extends Z{constructor(e){super(e),this._parent=this._element.closest('.list-group, .nav, [role="tablist"]'),this._parent&&(this._setInitialAttributes(this._parent,this._getChildren()),V.on(this._element,cr,e=>this._keydown(e)))}static get NAME(){return"tab"}show(){const e=this._element;if(this._elemIsActive(e))return;const t=this._getActiveElem(),n=t?V.trigger(t,lr,{relatedTarget:e}):null;V.trigger(e,hr,{relatedTarget:t}).defaultPrevented||n&&n.defaultPrevented||(this._deactivate(t,e),this._activate(e,t))}_activate(e,t){if(!e)return;e.classList.add(xr),this._activate(d(e));this._queueCallback(()=>{"tab"===e.getAttribute("role")?(e.removeAttribute("tabindex"),e.setAttribute("aria-selected",!0),this._toggleDropDown(e,!0),V.trigger(e,dr,{relatedTarget:t})):e.classList.add(wr)},e,e.classList.contains(vr))}_deactivate(e,t){if(!e)return;e.classList.remove(xr),e.blur(),this._deactivate(d(e));this._queueCallback(()=>{"tab"===e.getAttribute("role")?(e.setAttribute("aria-selected",!1),e.setAttribute("tabindex","-1"),this._toggleDropDown(e,!1),V.trigger(e,ur,{relatedTarget:t})):e.classList.remove(wr)},e,e.classList.contains(vr))}_keydown(e){if(![mr,gr,yr,br].includes(e.key))return;e.stopPropagation(),e.preventDefault();const t=[gr,br].includes(e.key),n=N(this._getChildren().filter(e=>!g(e)),e.target,t,!0);n&&(n.focus({preventScroll:!0}),Nr.getOrCreateInstance(n).show())}_getChildren(){return ae.find(Sr,this._parent)}_getActiveElem(){return this._getChildren().find(e=>this._elemIsActive(e))||null}_setInitialAttributes(e,t){this._setAttributeIfNotExists(e,"role","tablist");for(const e of t)this._setInitialAttributesOnChild(e)}_setInitialAttributesOnChild(e){e=this._getInnerElement(e);const t=this._elemIsActive(e),n=this._getOuterElement(e);e.setAttribute("aria-selected",t),n!==e&&this._setAttributeIfNotExists(n,"role","presentation"),t||e.setAttribute("tabindex","-1"),this._setAttributeIfNotExists(e,"role","tab"),this._setInitialAttributesOnTargetPanel(e)}_setInitialAttributesOnTargetPanel(e){const t=d(e);t&&(this._setAttributeIfNotExists(t,"role","tabpanel"),e.id&&this._setAttributeIfNotExists(t,"aria-labelledby",`#${e.id}`))}_toggleDropDown(e,t){const n=this._getOuterElement(e);if(!n.classList.contains("dropdown"))return;const r=(e,r)=>{const a=ae.findOne(e,n);a&&a.classList.toggle(r,t)};r(".dropdown-toggle",xr),r(".dropdown-menu",wr),n.setAttribute("aria-expanded",t)}_setAttributeIfNotExists(e,t,n){e.hasAttribute(t)||e.setAttribute(t,n)}_elemIsActive(e){return e.classList.contains(xr)}_getInnerElement(e){return e.matches(Sr)?e:ae.findOne(Sr,e)}_getOuterElement(e){return e.closest(".nav-item, .list-group-item")||e}static jQueryInterface(e){return this.each(function(){const t=Nr.getOrCreateInstance(this);if("string"==typeof e){if(void 0===t[e]||e.startsWith("_")||"constructor"===e)throw new TypeError(`No method named "${e}"`);t[e]()}})}}V.on(document,pr,Ir,function(e){["A","AREA"].includes(this.tagName)&&e.preventDefault(),g(this)||Nr.getOrCreateInstance(this).show()}),V.on(window,fr,()=>{for(const e of ae.find(_r))Nr.getOrCreateInstance(e)}),I(Nr);const Tr=".bs.toast",Cr=`mouseover${Tr}`,Er=`mouseout${Tr}`,Ar=`focusin${Tr}`,$r=`focusout${Tr}`,Rr=`hide${Tr}`,Fr=`hidden${Tr}`,Dr=`show${Tr}`,Mr=`shown${Tr}`,Or="hide",Lr="show",Pr="showing",zr={animation:"boolean",autohide:"boolean",delay:"number"},Br={animation:!0,autohide:!0,delay:5e3};class Wr extends Z{constructor(e,t){super(e,t),this._timeout=null,this._hasMouseInteraction=!1,this._hasKeyboardInteraction=!1,this._setListeners()}static get Default(){return Br}static get DefaultType(){return zr}static get NAME(){return"toast"}show(){if(V.trigger(this._element,Dr).defaultPrevented)return;this._clearTimeout(),this._config.animation&&this._element.classList.add("fade");this._element.classList.remove(Or),x(this._element),this._element.classList.add(Lr,Pr),this._queueCallback(()=>{this._element.classList.remove(Pr),V.trigger(this._element,Mr),this._maybeScheduleHide()},this._element,this._config.animation)}hide(){if(!this.isShown())return;if(V.trigger(this._element,Rr).defaultPrevented)return;this._element.classList.add(Pr),this._queueCallback(()=>{this._element.classList.add(Or),this._element.classList.remove(Pr,Lr),V.trigger(this._element,Fr)},this._element,this._config.animation)}dispose(){this._clearTimeout(),this.isShown()&&this._element.classList.remove(Lr),super.dispose()}isShown(){return this._element.classList.contains(Lr)}_maybeScheduleHide(){this._config.autohide&&(this._hasMouseInteraction||this._hasKeyboardInteraction||(this._timeout=setTimeout(()=>{this.hide()},this._config.delay)))}_onInteraction(e,t){switch(e.type){case"mouseover":case"mouseout":this._hasMouseInteraction=t;break;case"focusin":case"focusout":this._hasKeyboardInteraction=t}if(t)return void this._clearTimeout();const n=e.relatedTarget;this._element===n||this._element.contains(n)||this._maybeScheduleHide()}_setListeners(){V.on(this._element,Cr,e=>this._onInteraction(e,!0)),V.on(this._element,Er,e=>this._onInteraction(e,!1)),V.on(this._element,Ar,e=>this._onInteraction(e,!0)),V.on(this._element,$r,e=>this._onInteraction(e,!1))}_clearTimeout(){clearTimeout(this._timeout),this._timeout=null}static jQueryInterface(e){return this.each(function(){const t=Wr.getOrCreateInstance(this,e);if("string"==typeof e){if(void 0===t[e])throw new TypeError(`No method named "${e}"`);t[e](this)}})}}Y(Wr),I(Wr)},17334(e){function t(e,t=100,n={}){if("function"!=typeof e)throw new TypeError(`Expected the first parameter to be a function, got \`${typeof e}\`.`);if(t<0)throw new RangeError("`wait` must not be negative.");const{immediate:r}="boolean"==typeof n?{immediate:n}:n;let a,s,i,o,l;function u(){const t=a,n=s;return a=void 0,s=void 0,l=e.apply(t,n),l}function h(){const e=Date.now()-o;e<t&&e>=0?i=setTimeout(h,t-e):(i=void 0,r||(l=u()))}const d=function(...e){if(a&&this!==a&&Object.getPrototypeOf(this)===Object.getPrototypeOf(a))throw new Error("Debounced method called with different contexts of the same prototype.");a=this,s=e,o=Date.now();const n=r&&!i;return i||(i=setTimeout(h,t)),n&&(l=u()),l};return Object.defineProperty(d,"isPending",{get:()=>void 0!==i}),d.clear=()=>{i&&(clearTimeout(i),i=void 0)},d.flush=()=>{i&&d.trigger()},d.trigger=()=>{l=u(),d.clear()},d}e.exports.debounce=t,e.exports=t},84918(e){window,e.exports=function(e){var t={};function n(r){if(t[r])return t[r].exports;var a=t[r]={i:r,l:!1,exports:{}};return e[r].call(a.exports,a,a.exports,n),a.l=!0,a.exports}return n.m=e,n.c=t,n.d=function(e,t,r){n.o(e,t)||Object.defineProperty(e,t,{enumerable:!0,get:r})},n.r=function(e){"undefined"!=typeof Symbol&&Symbol.toStringTag&&Object.defineProperty(e,Symbol.toStringTag,{value:"Module"}),Object.defineProperty(e,"__esModule",{value:!0})},n.t=function(e,t){if(1&t&&(e=n(e)),8&t)return e;if(4&t&&"object"==typeof e&&e&&e.__esModule)return e;var r=Object.create(null);if(n.r(r),Object.defineProperty(r,"default",{enumerable:!0,value:e}),2&t&&"string"!=typeof e)for(var a in e)n.d(r,a,function(t){return e[t]}.bind(null,a));return r},n.n=function(e){var t=e&&e.__esModule?function(){return e.default}:function(){return e};return n.d(t,"a",t),t},n.o=function(e,t){return Object.prototype.hasOwnProperty.call(e,t)},n.p="examples",n(n.s=4)}([function(e,t,n){var r,a;!function(s){"use strict";r=[n(2)],void 0===(a=function(e){return function(e,t){var n=e.jQuery,r=e.console;function a(e,t){for(var n in t)e[n]=t[n];return e}var s=Array.prototype.slice;function i(e,t,o){if(!(this instanceof i))return new i(e,t,o);var l,u=e;"string"==typeof e&&(u=document.querySelectorAll(e)),u?(this.elements=(l=u,Array.isArray(l)?l:"object"==typeof l&&"number"==typeof l.length?s.call(l):[l]),this.options=a({},this.options),"function"==typeof t?o=t:a(this.options,t),o&&this.on("always",o),this.getImages(),n&&(this.jqDeferred=new n.Deferred),setTimeout(this.check.bind(this))):r.error("Bad element for imagesLoaded "+(u||e))}i.prototype=Object.create(t.prototype),i.prototype.options={},i.prototype.getImages=function(){this.images=[],this.elements.forEach(this.addElementImages,this)},i.prototype.addElementImages=function(e){"IMG"==e.nodeName&&this.addImage(e),!0===this.options.background&&this.addElementBackgroundImages(e);var t=e.nodeType;if(t&&o[t]){for(var n=e.querySelectorAll("img"),r=0;r<n.length;r++){var a=n[r];this.addImage(a)}if("string"==typeof this.options.background){var s=e.querySelectorAll(this.options.background);for(r=0;r<s.length;r++){var i=s[r];this.addElementBackgroundImages(i)}}}};var o={1:!0,9:!0,11:!0};function l(e){this.img=e}function u(e,t){this.url=e,this.element=t,this.img=new Image}return i.prototype.addElementBackgroundImages=function(e){var t=getComputedStyle(e);if(t)for(var n=/url\((['"])?(.*?)\1\)/gi,r=n.exec(t.backgroundImage);null!==r;){var a=r&&r[2];a&&this.addBackground(a,e),r=n.exec(t.backgroundImage)}},i.prototype.addImage=function(e){var t=new l(e);this.images.push(t)},i.prototype.addBackground=function(e,t){var n=new u(e,t);this.images.push(n)},i.prototype.check=function(){var e=this;function t(t,n,r){setTimeout(function(){e.progress(t,n,r)})}this.progressedCount=0,this.hasAnyBroken=!1,this.images.length?this.images.forEach(function(e){e.once("progress",t),e.check()}):this.complete()},i.prototype.progress=function(e,t,n){this.progressedCount++,this.hasAnyBroken=this.hasAnyBroken||!e.isLoaded,this.emitEvent("progress",[this,e,t]),this.jqDeferred&&this.jqDeferred.notify&&this.jqDeferred.notify(this,e),this.progressedCount==this.images.length&&this.complete(),this.options.debug&&r&&r.log("progress: "+n,e,t)},i.prototype.complete=function(){var e=this.hasAnyBroken?"fail":"done";if(this.isComplete=!0,this.emitEvent(e,[this]),this.emitEvent("always",[this]),this.jqDeferred){var t=this.hasAnyBroken?"reject":"resolve";this.jqDeferred[t](this)}},l.prototype=Object.create(t.prototype),l.prototype.check=function(){this.getIsImageComplete()?this.confirm(0!==this.img.naturalWidth,"naturalWidth"):(this.proxyImage=new Image,this.proxyImage.addEventListener("load",this),this.proxyImage.addEventListener("error",this),this.img.addEventListener("load",this),this.img.addEventListener("error",this),this.proxyImage.src=this.img.src)},l.prototype.getIsImageComplete=function(){return this.img.complete&&this.img.naturalWidth},l.prototype.confirm=function(e,t){this.isLoaded=e,this.emitEvent("progress",[this,this.img,t])},l.prototype.handleEvent=function(e){var t="on"+e.type;this[t]&&this[t](e)},l.prototype.onload=function(){this.confirm(!0,"onload"),this.unbindEvents()},l.prototype.onerror=function(){this.confirm(!1,"onerror"),this.unbindEvents()},l.prototype.unbindEvents=function(){this.proxyImage.removeEventListener("load",this),this.proxyImage.removeEventListener("error",this),this.img.removeEventListener("load",this),this.img.removeEventListener("error",this)},u.prototype=Object.create(l.prototype),u.prototype.check=function(){this.img.addEventListener("load",this),this.img.addEventListener("error",this),this.img.src=this.url,this.getIsImageComplete()&&(this.confirm(0!==this.img.naturalWidth,"naturalWidth"),this.unbindEvents())},u.prototype.unbindEvents=function(){this.img.removeEventListener("load",this),this.img.removeEventListener("error",this)},u.prototype.confirm=function(e,t){this.isLoaded=e,this.emitEvent("progress",[this,this.element,t])},i.makeJQueryPlugin=function(t){(t=t||e.jQuery)&&((n=t).fn.imagesLoaded=function(e,t){return new i(this,e,t).jqDeferred.promise(n(this))})},i.makeJQueryPlugin(),i}(s,e)}.apply(t,r))||(e.exports=a)}("undefined"!=typeof window?window:this)},function(e,t,n){(e.exports=n(3)(!1)).push([e.i,'.ff-container{display:inline-block;position:relative}.ff-container .ff-image{z-index:0;vertical-align:top;opacity:0}.ff-container .ff-canvas{display:inline-block;position:absolute;top:0;left:0;pointer-events:none;z-index:1;vertical-align:top;opacity:0}.ff-container .ff-canvas.ff-canvas-ready{-webkit-transition:opacity 300ms;-o-transition:opacity 300ms;-moz-transition:opacity 300ms;transition:opacity 300ms;opacity:1}.ff-container.ff-active .ff-image{opacity:1}.ff-container.ff-active .ff-canvas.ff-canvas-ready{-webkit-transition:none;-o-transition:none;-moz-transition:none;transition:none;opacity:0}.ff-container.ff-active .ff-overlay{display:none}.ff-container.ff-inactive .ff-canvas.ff-canvas-ready{-webkit-transition:opacity 300ms;-o-transition:opacity 300ms;-moz-transition:opacity 300ms;transition:opacity 300ms;opacity:1}.ff-container.ff-inactive .ff-image{-webkit-transition:opacity 300ms;-o-transition:opacity 300ms;-moz-transition:opacity 300ms;transition:opacity 300ms;-webkit-transition-delay:170ms;-moz-transition-delay:170ms;-o-transition-delay:170ms;transition-delay:170ms;opacity:0}.ff-container.ff-responsive{width:100%}.ff-container.ff-responsive .ff-image,.ff-container.ff-responsive .ff-canvas{width:100%}.ff-container.ff-loading-icon:before{content:"";position:absolute;background-image:url("data:image/svg+xml;base64,PD94bWwgdmVyc2lvbj0iMS4wIiBlbmNvZGluZz0idXRmLTgiPz48c3ZnIHdpZHRoPSc1MHB4JyBoZWlnaHQ9JzUwcHgnIHhtbG5zPSJodHRwOi8vd3d3LnczLm9yZy8yMDAwL3N2ZyIgdmlld0JveD0iMCAwIDEwMCAxMDAiIHByZXNlcnZlQXNwZWN0UmF0aW89InhNaWRZTWlkIiBjbGFzcz0idWlsLXNwaW4iPjxyZWN0IHg9IjAiIHk9IjAiIHdpZHRoPSIxMDAiIGhlaWdodD0iMTAwIiBmaWxsPSJub25lIiBjbGFzcz0iYmsiPjwvcmVjdD48ZyB0cmFuc2Zvcm09InRyYW5zbGF0ZSg1MCA1MCkiPjxnIHRyYW5zZm9ybT0icm90YXRlKDApIHRyYW5zbGF0ZSgzNCAwKSI+PGNpcmNsZSBjeD0iMCIgY3k9IjAiIHI9IjgiIGZpbGw9IiNmZmZmZmYiPjxhbmltYXRlIGF0dHJpYnV0ZU5hbWU9Im9wYWNpdHkiIGZyb209IjEiIHRvPSIwLjEiIGJlZ2luPSIwcyIgZHVyPSIxcyIgcmVwZWF0Q291bnQ9ImluZGVmaW5pdGUiPjwvYW5pbWF0ZT48YW5pbWF0ZVRyYW5zZm9ybSBhdHRyaWJ1dGVOYW1lPSJ0cmFuc2Zvcm0iIHR5cGU9InNjYWxlIiBmcm9tPSIxLjUiIHRvPSIxIiBiZWdpbj0iMHMiIGR1cj0iMXMiIHJlcGVhdENvdW50PSJpbmRlZmluaXRlIj48L2FuaW1hdGVUcmFuc2Zvcm0+PC9jaXJjbGU+PC9nPjxnIHRyYW5zZm9ybT0icm90YXRlKDQ1KSB0cmFuc2xhdGUoMzQgMCkiPjxjaXJjbGUgY3g9IjAiIGN5PSIwIiByPSI4IiBmaWxsPSIjZmZmZmZmIj48YW5pbWF0ZSBhdHRyaWJ1dGVOYW1lPSJvcGFjaXR5IiBmcm9tPSIxIiB0bz0iMC4xIiBiZWdpbj0iMC4xMnMiIGR1cj0iMXMiIHJlcGVhdENvdW50PSJpbmRlZmluaXRlIj48L2FuaW1hdGU+PGFuaW1hdGVUcmFuc2Zvcm0gYXR0cmlidXRlTmFtZT0idHJhbnNmb3JtIiB0eXBlPSJzY2FsZSIgZnJvbT0iMS41IiB0bz0iMSIgYmVnaW49IjAuMTJzIiBkdXI9IjFzIiByZXBlYXRDb3VudD0iaW5kZWZpbml0ZSI+PC9hbmltYXRlVHJhbnNmb3JtPjwvY2lyY2xlPjwvZz48ZyB0cmFuc2Zvcm09InJvdGF0ZSg5MCkgdHJhbnNsYXRlKDM0IDApIj48Y2lyY2xlIGN4PSIwIiBjeT0iMCIgcj0iOCIgZmlsbD0iI2ZmZmZmZiI+PGFuaW1hdGUgYXR0cmlidXRlTmFtZT0ib3BhY2l0eSIgZnJvbT0iMSIgdG89IjAuMSIgYmVnaW49IjAuMjVzIiBkdXI9IjFzIiByZXBlYXRDb3VudD0iaW5kZWZpbml0ZSI+PC9hbmltYXRlPjxhbmltYXRlVHJhbnNmb3JtIGF0dHJpYnV0ZU5hbWU9InRyYW5zZm9ybSIgdHlwZT0ic2NhbGUiIGZyb209IjEuNSIgdG89IjEiIGJlZ2luPSIwLjI1cyIgZHVyPSIxcyIgcmVwZWF0Q291bnQ9ImluZGVmaW5pdGUiPjwvYW5pbWF0ZVRyYW5zZm9ybT48L2NpcmNsZT48L2c+PGcgdHJhbnNmb3JtPSJyb3RhdGUoMTM1KSB0cmFuc2xhdGUoMzQgMCkiPjxjaXJjbGUgY3g9IjAiIGN5PSIwIiByPSI4IiBmaWxsPSIjZmZmZmZmIj48YW5pbWF0ZSBhdHRyaWJ1dGVOYW1lPSJvcGFjaXR5IiBmcm9tPSIxIiB0bz0iMC4xIiBiZWdpbj0iMC4zN3MiIGR1cj0iMXMiIHJlcGVhdENvdW50PSJpbmRlZmluaXRlIj48L2FuaW1hdGU+PGFuaW1hdGVUcmFuc2Zvcm0gYXR0cmlidXRlTmFtZT0idHJhbnNmb3JtIiB0eXBlPSJzY2FsZSIgZnJvbT0iMS41IiB0bz0iMSIgYmVnaW49IjAuMzdzIiBkdXI9IjFzIiByZXBlYXRDb3VudD0iaW5kZWZpbml0ZSI+PC9hbmltYXRlVHJhbnNmb3JtPjwvY2lyY2xlPjwvZz48ZyB0cmFuc2Zvcm09InJvdGF0ZSgxODApIHRyYW5zbGF0ZSgzNCAwKSI+PGNpcmNsZSBjeD0iMCIgY3k9IjAiIHI9IjgiIGZpbGw9IiNmZmZmZmYiPjxhbmltYXRlIGF0dHJpYnV0ZU5hbWU9Im9wYWNpdHkiIGZyb209IjEiIHRvPSIwLjEiIGJlZ2luPSIwLjVzIiBkdXI9IjFzIiByZXBlYXRDb3VudD0iaW5kZWZpbml0ZSI+PC9hbmltYXRlPjxhbmltYXRlVHJhbnNmb3JtIGF0dHJpYnV0ZU5hbWU9InRyYW5zZm9ybSIgdHlwZT0ic2NhbGUiIGZyb209IjEuNSIgdG89IjEiIGJlZ2luPSIwLjVzIiBkdXI9IjFzIiByZXBlYXRDb3VudD0iaW5kZWZpbml0ZSI+PC9hbmltYXRlVHJhbnNmb3JtPjwvY2lyY2xlPjwvZz48ZyB0cmFuc2Zvcm09InJvdGF0ZSgyMjUpIHRyYW5zbGF0ZSgzNCAwKSI+PGNpcmNsZSBjeD0iMCIgY3k9IjAiIHI9IjgiIGZpbGw9IiNmZmZmZmYiPjxhbmltYXRlIGF0dHJpYnV0ZU5hbWU9Im9wYWNpdHkiIGZyb209IjEiIHRvPSIwLjEiIGJlZ2luPSIwLjYycyIgZHVyPSIxcyIgcmVwZWF0Q291bnQ9ImluZGVmaW5pdGUiPjwvYW5pbWF0ZT48YW5pbWF0ZVRyYW5zZm9ybSBhdHRyaWJ1dGVOYW1lPSJ0cmFuc2Zvcm0iIHR5cGU9InNjYWxlIiBmcm9tPSIxLjUiIHRvPSIxIiBiZWdpbj0iMC42MnMiIGR1cj0iMXMiIHJlcGVhdENvdW50PSJpbmRlZmluaXRlIj48L2FuaW1hdGVUcmFuc2Zvcm0+PC9jaXJjbGU+PC9nPjxnIHRyYW5zZm9ybT0icm90YXRlKDI3MCkgdHJhbnNsYXRlKDM0IDApIj48Y2lyY2xlIGN4PSIwIiBjeT0iMCIgcj0iOCIgZmlsbD0iI2ZmZmZmZiI+PGFuaW1hdGUgYXR0cmlidXRlTmFtZT0ib3BhY2l0eSIgZnJvbT0iMSIgdG89IjAuMSIgYmVnaW49IjAuNzVzIiBkdXI9IjFzIiByZXBlYXRDb3VudD0iaW5kZWZpbml0ZSI+PC9hbmltYXRlPjxhbmltYXRlVHJhbnNmb3JtIGF0dHJpYnV0ZU5hbWU9InRyYW5zZm9ybSIgdHlwZT0ic2NhbGUiIGZyb209IjEuNSIgdG89IjEiIGJlZ2luPSIwLjc1cyIgZHVyPSIxcyIgcmVwZWF0Q291bnQ9ImluZGVmaW5pdGUiPjwvYW5pbWF0ZVRyYW5zZm9ybT48L2NpcmNsZT48L2c+PGcgdHJhbnNmb3JtPSJyb3RhdGUoMzE1KSB0cmFuc2xhdGUoMzQgMCkiPjxjaXJjbGUgY3g9IjAiIGN5PSIwIiByPSI4IiBmaWxsPSIjZmZmZmZmIj48YW5pbWF0ZSBhdHRyaWJ1dGVOYW1lPSJvcGFjaXR5IiBmcm9tPSIxIiB0bz0iMC4xIiBiZWdpbj0iMC44N3MiIGR1cj0iMXMiIHJlcGVhdENvdW50PSJpbmRlZmluaXRlIj48L2FuaW1hdGU+PGFuaW1hdGVUcmFuc2Zvcm0gYXR0cmlidXRlTmFtZT0idHJhbnNmb3JtIiB0eXBlPSJzY2FsZSIgZnJvbT0iMS41IiB0bz0iMSIgYmVnaW49IjAuODdzIiBkdXI9IjFzIiByZXBlYXRDb3VudD0iaW5kZWZpbml0ZSI+PC9hbmltYXRlVHJhbnNmb3JtPjwvY2lyY2xlPjwvZz48L2c+PC9zdmc+");background-position:center center;background-repeat:no-repeat;height:46px;width:46px;z-index:3;top:50%;left:50%;-webkit-transform:translate(-50%, -50%);-moz-transform:translate(-50%, -50%);-ms-transform:translate(-50%, -50%);-o-transform:translate(-50%, -50%);transform:translate(-50%, -50%)}.ff-container .ff-overlay{background-image:url("data:image/png;base64,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");background-repeat:no-repeat;max-width:94px;max-height:94px;position:absolute;left:0%;right:0%;top:0%;bottom:0%;margin:auto;-webkit-background-size:contain;-moz-background-size:contain;background-size:contain;background-position:center;pointer-events:none;z-index:100}',""])},function(e,t,n){var r,a;"undefined"!=typeof window&&window,void 0===(a="function"==typeof(r=function(){"use strict";function e(){}var t=e.prototype;return t.on=function(e,t){if(e&&t){var n=this._events=this._events||{},r=n[e]=n[e]||[];return-1==r.indexOf(t)&&r.push(t),this}},t.once=function(e,t){if(e&&t){this.on(e,t);var n=this._onceEvents=this._onceEvents||{};return(n[e]=n[e]||{})[t]=!0,this}},t.off=function(e,t){var n=this._events&&this._events[e];if(n&&n.length){var r=n.indexOf(t);return-1!=r&&n.splice(r,1),this}},t.emitEvent=function(e,t){var n=this._events&&this._events[e];if(n&&n.length){n=n.slice(0),t=t||[];for(var r=this._onceEvents&&this._onceEvents[e],a=0;a<n.length;a++){var s=n[a];r&&r[s]&&(this.off(e,s),delete r[s]),s.apply(this,t)}return this}},t.allOff=function(){delete this._events,delete this._onceEvents},e})?r.call(t,n,t,e):r)||(e.exports=a)},function(e,t,n){"use strict";e.exports=function(e){var t=[];return t.toString=function(){return this.map(function(t){var n=function(e,t){var n,r=e[1]||"",a=e[3];if(!a)return r;if(t&&"function"==typeof btoa){var s=(n=a,"/*# sourceMappingURL=data:application/json;charset=utf-8;base64,"+btoa(unescape(encodeURIComponent(JSON.stringify(n))))+" */"),i=a.sources.map(function(e){return"/*# sourceURL="+a.sourceRoot+e+" */"});return[r].concat(i).concat([s]).join("\n")}return[r].join("\n")}(t,e);return t[2]?"@media "+t[2]+"{"+n+"}":n}).join("")},t.i=function(e,n){"string"==typeof e&&(e=[[null,e,""]]);for(var r={},a=0;a<this.length;a++){var s=this[a][0];null!=s&&(r[s]=!0)}for(a=0;a<e.length;a++){var i=e[a];null!=i[0]&&r[i[0]]||(n&&!i[2]?i[2]=n:n&&(i[2]="("+i[2]+") and ("+n+")"),t.push(i))}},t}},function(e,t,n){"use strict";n.r(t);var r,a=n(0),s=n.n(a);!function(e){e.START="start",e.STOP="stop",e.TOGGLE="toggle"}(r||(r={}));const i=e=>`✨Freezeframe: ${e}✨`,o=(e,...t)=>{console.error(i(e),...t)},l=e=>"string"==typeof e?document.querySelectorAll(e):e,u=(e,t,n)=>{const r=e instanceof Element?[e]:e;return Array.from(r).reduce((e,t)=>{if(t instanceof HTMLImageElement)e.push(t),"gif"!==(e=>e.split(".").pop().toLowerCase().substring(0,3))(t.src)&&n.warnings&&((e,...t)=>{console.warn(i(e),...t)})("Image does not appear to be a gif",t);else{const n=t.querySelectorAll("img");n.length?e=e.concat(Array.from(n)):o("Invalid element",t)}return e},[])},h=e=>[...new Set(e)],d=e=>{const t=window.document.createElement("div");t.innerHTML=e;const n=t.childNodes;return n.length>1?n:n[0]};var p,c,f=function(){function e(e,t){for(var n=0;n<t.length;n++){var r=t[n];r.enumerable=r.enumerable||!1,r.configurable=!0,"value"in r&&(r.writable=!0),Object.defineProperty(e,r.key,r)}}return function(t,n,r){return n&&e(t.prototype,n),r&&e(t,r),t}}(),m=(p=["",""],c=["",""],Object.freeze(Object.defineProperties(p,{raw:{value:Object.freeze(c)}})));function g(e,t){if(!(e instanceof t))throw new TypeError("Cannot call a class as a function")}var y=function(){function e(){for(var t=this,n=arguments.length,r=Array(n),a=0;a<n;a++)r[a]=arguments[a];return g(this,e),this.tag=function(e){for(var n=arguments.length,r=Array(n>1?n-1:0),a=1;a<n;a++)r[a-1]=arguments[a];return"function"==typeof e?t.interimTag.bind(t,e):"string"==typeof e?t.transformEndResult(e):(e=e.map(t.transformString.bind(t)),t.transformEndResult(e.reduce(t.processSubstitutions.bind(t,r))))},r.length>0&&Array.isArray(r[0])&&(r=r[0]),this.transformers=r.map(function(e){return"function"==typeof e?e():e}),this.tag}return f(e,[{key:"interimTag",value:function(e,t){for(var n=arguments.length,r=Array(n>2?n-2:0),a=2;a<n;a++)r[a-2]=arguments[a];return this.tag(m,e.apply(void 0,[t].concat(r)))}},{key:"processSubstitutions",value:function(e,t,n){var r=this.transformSubstitution(e.shift(),t);return"".concat(t,r,n)}},{key:"transformString",value:function(e){return this.transformers.reduce(function(e,t){return t.onString?t.onString(e):e},e)}},{key:"transformSubstitution",value:function(e,t){return this.transformers.reduce(function(e,n){return n.onSubstitution?n.onSubstitution(e,t):e},e)}},{key:"transformEndResult",value:function(e){return this.transformers.reduce(function(e,t){return t.onEndResult?t.onEndResult(e):e},e)}}]),e}(),b=function(){var e=arguments.length>0&&void 0!==arguments[0]?arguments[0]:"";return{onEndResult:function(t){if(""===e)return t.trim();if("start"===(e=e.toLowerCase())||"left"===e)return t.replace(/^\s*/,"");if("end"===e||"right"===e)return t.replace(/\s*$/,"");throw new Error("Side not supported: "+e)}}};function x(e){if(Array.isArray(e)){for(var t=0,n=Array(e.length);t<e.length;t++)n[t]=e[t];return n}return Array.from(e)}var v=function(){var e=arguments.length>0&&void 0!==arguments[0]?arguments[0]:"initial";return{onEndResult:function(t){if("initial"===e){var n=t.match(/^[^\S\n]*(?=\S)/gm),r=n&&Math.min.apply(Math,x(n.map(function(e){return e.length})));if(r){var a=new RegExp("^.{"+r+"}","gm");return t.replace(a,"")}return t}if("all"===e)return t.replace(/^[^\S\n]+/gm,"");throw new Error("Unknown type: "+e)}}},w=function(e,t){return{onEndResult:function(n){if(null==e||null==t)throw new Error("replaceResultTransformer requires at least 2 arguments.");return n.replace(e,t)}}},k=function(e,t){return{onSubstitution:function(n,r){if(null==e||null==t)throw new Error("replaceSubstitutionTransformer requires at least 2 arguments.");return null==n?n:n.toString().replace(e,t)}}},I={separator:"",conjunction:"",serial:!1},S=function(){var e=arguments.length>0&&void 0!==arguments[0]?arguments[0]:I;return{onSubstitution:function(t,n){if(Array.isArray(t)){var r=t.length,a=e.separator,s=e.conjunction,i=e.serial,o=n.match(/(\n?[^\S\n]+)$/);if(t=o?t.join(a+o[1]):t.join(a+" "),s&&r>1){var l=t.lastIndexOf(a);t=t.slice(0,l)+(i?a:"")+" "+s+t.slice(l+1)}}return t}}},_=function(e){return{onSubstitution:function(t,n){if(null==e||"string"!=typeof e)throw new Error("You need to specify a string character to split by.");return"string"==typeof t&&t.includes(e)&&(t=t.split(e)),t}}},N=function(e){return null!=e&&!Number.isNaN(e)&&"boolean"!=typeof e},T=function(){return{onSubstitution:function(e){return Array.isArray(e)?e.filter(N):N(e)?e:""}}},C=(new y(S({separator:","}),v,b),new y(S({separator:",",conjunction:"and"}),v,b),new y(S({separator:",",conjunction:"or"}),v,b),new y(_("\n"),T,S,v,b));new y(_("\n"),S,v,b,k(/&/g,"&amp;"),k(/</g,"&lt;"),k(/>/g,"&gt;"),k(/"/g,"&quot;"),k(/'/g,"&#x27;"),k(/`/g,"&#x60;")),new y(w(/(?:\n(?:\s*))+/g," "),b),new y(w(/(?:\n\s*)/g,""),b),new y(S({separator:","}),w(/(?:\s+)/g," "),b),new y(S({separator:",",conjunction:"or"}),w(/(?:\s+)/g," "),b),new y(S({separator:",",conjunction:"and"}),w(/(?:\s+)/g," "),b),new y(S,v,b),new y(S,w(/(?:\s+)/g," "),b),new y(v,b),new y(v("all"),b);const E=".freezeframe",A="ff-container",$="ff-loading-icon",R="ff-image",F="ff-canvas",D="ff-ready",M="ff-inactive",O="ff-active",L="ff-canvas-ready",P="ff-responsive",z="ff-overlay",B={selector:E,responsive:!0,trigger:"hover",overlay:!1,warnings:!0};var W,V,U,G=n(1),H=n.n(G),j=function(e,t,n,r){return new(n||(n=Promise))(function(a,s){function i(e){try{l(r.next(e))}catch(e){s(e)}}function o(e){try{l(r.throw(e))}catch(e){s(e)}}function l(e){var t;e.done?a(e.value):(t=e.value,t instanceof n?t:new n(function(e){e(t)})).then(i,o)}l((r=r.apply(e,t||[])).next())})},q=function(e,t,n){if(!t.has(e))throw new TypeError("attempted to set private field on non-instance");return t.set(e,n),n},K=function(e,t){if(!t.has(e))throw new TypeError("attempted to get private field on non-instance");return t.get(e)};W=new WeakMap,V=new WeakMap,U=new WeakMap,t.default=class{constructor(e=E,t){this.items=[],this.$images=[],W.set(this,void 0),V.set(this,void 0),this._eventListeners=Object.assign({},Object.values(r).reduce((e,t)=>(e[t]=[],e),{})),U.set(this,[]),"string"==typeof e||e instanceof Element||e instanceof HTMLCollection||e instanceof NodeList?(this.options=Object.assign(Object.assign({},B),t),q(this,W,e)):"object"!=typeof e||t?o("Invalid Freezeframe.js configuration:",...[e,t].filter(e=>void 0!==e)):(this.options=Object.assign(Object.assign({},B),e),q(this,W,this.options.selector)),this._init(K(this,W))}get _stylesInjected(){return!!document.querySelector("style#ff-styles")}_init(e){this._injectStylesheet(),q(this,V,"ontouchstart"in window||"onmsgesturechange"in window),this._capture(e),this._load(this.$images)}_capture(e){this.$images=((...e)=>(...t)=>e.reduceRight((e,n)=>(...r)=>e(n(...r,...t)))())(l,u,h)(e,this.options)}_load(e){s()(e).on("progress",(e,{img:t})=>{this._setup(t)})}_setup(e){return j(this,void 0,void 0,function*(){const t=this._wrap(e);this.items.push(t),yield this._process(t),this._attach(t)})}_wrap(e){const t=d(C`
    <div class="${A} ${$}">
    </div>
  `),n=d(C`
    <canvas class="${F}" width="0" height="0">
    </canvas>
  `);var r,a;return this.options.responsive&&t.classList.add(P),this.options.overlay&&t.appendChild(d(C`
    <div class="${z}">
    </div>
  `)),e.classList.add(R),t.appendChild(n),a=t,(r=e).parentNode.insertBefore(a,r),a.appendChild(r),{$container:t,$canvas:n,$image:e}}_process(e){return new Promise(t=>{const{$canvas:n,$image:r,$container:a}=e,{clientWidth:s,clientHeight:i}=r;n.setAttribute("width",s.toString()),n.setAttribute("height",i.toString()),n.getContext("2d").drawImage(r,0,0,s,i),n.classList.add(L),n.addEventListener("transitionend",()=>{this._ready(a),t(e)},{once:!0})})}_ready(e){e.classList.add(D),e.classList.add(M),e.classList.remove($)}_attach(e){const{$image:t}=e;if(!K(this,V)&&"hover"===this.options.trigger){const n=()=>{this._toggleOn(e),this._emit(r.START,e,!0),this._emit(r.TOGGLE,e,!0)},a=()=>{this._toggleOff(e),this._emit(r.START,e,!1),this._emit(r.TOGGLE,e,!1)};this._addEventListener(t,"mouseleave",a),this._addEventListener(t,"mouseenter",n)}if(K(this,V)||"click"===this.options.trigger){const n=()=>{this._toggle(e)};this._addEventListener(t,"click",n)}}_addEventListener(e,t,n){K(this,U).push({$image:e,event:t,listener:n}),e.addEventListener(t,n)}_removeEventListener(e,t,n){e.removeEventListener(t,n)}_injectStylesheet(){this._stylesInjected||document.head.appendChild(d(C(`\n    <style id="ff-styles">\n      ${H.a.toString()}\n    </style>\n  `)))}_emit(e,t,n){this._eventListeners[e].forEach(e=>{e(Array.isArray(t)&&1===t.length?t[0]:t,n)})}_toggleOn(e){const{$container:t,$image:n}=e;t.classList.contains(D)&&(n.setAttribute("src",n.src),t.classList.remove(M),t.classList.add(O))}_toggleOff(e){const{$container:t}=e;t.classList.contains(D)&&(t.classList.add(M),t.classList.remove(O))}_toggle(e){const{$container:t}=e,n=t.classList.contains(O);return n?this._toggleOff(e):this._toggleOn(e),!n}start(){return this.items.forEach(e=>{this._toggleOn(e)}),this._emit(r.START,this.items,!0),this._emit(r.TOGGLE,this.items,!0),this}stop(){return this.items.forEach(e=>{this._toggleOff(e)}),this._emit(r.STOP,this.items,!1),this._emit(r.TOGGLE,this.items,!1),this}toggle(){return this.items.forEach(e=>{const t=this._toggle(e);t?this._emit(r.START,this.items,!1):this._emit(r.STOP,this.items,!1),this._emit(r.TOGGLE,this.items,t)}),this}on(e,t){return this._eventListeners[e].push(t),this}destroy(){K(this,U).forEach(({$image:e,event:t,listener:n})=>{this._removeEventListener(e,t,n)})}}}]).default},25938(e){const t=Math.atan2,n=Math.cos,r=Math.sin,a=Math.sqrt,s=Math.PI;function i(e){return e*s/180}function o(e){return function(e){return e*e}(r(e/2))}e.exports=function(e,r){const s=i(Array.isArray(e)?e[1]:e.latitude??e.lat),l=i(Array.isArray(r)?r[1]:r.latitude??r.lat),u=i(Array.isArray(e)?e[0]:e.longitude??e.lng??e.lon),h=i(Array.isArray(r)?r[0]:r.longitude??r.lng??r.lon),d=o(l-s)+n(s)*n(l)*o(h-u);return 12756274*t(a(d),a(1-d))}},90308(e,t,n){var r;!function(){"use strict";var a={};function s(e,t,n){for(var r=n.data,a=n.width,s=~~e.x,i=~~(e.x+e.width),o=~~e.y,l=~~(e.y+e.height),u=255*e.weight,h=o;h<l;h++)for(var d=s;d<i;d++){r[4*(h*a+d)+3]+=u}}function i(e,t,n){for(var r={detail:0,saturation:0,skin:0,boost:0,total:0},a=t.data,s=e.scoreDownSample,i=1/s,l=t.height*s,u=t.width*s,h=t.width,d=0;d<l;d+=s)for(var p=0;p<u;p+=s){var c=4*(~~(d*i)*h+~~(p*i)),f=o(e,n,p,d),m=a[c+1]/255;r.skin+=a[c]/255*(m+e.skinBias)*f,r.detail+=m*f,r.saturation+=a[c+2]/255*(m+e.saturationBias)*f,r.boost+=a[c+3]/255*f}return r.total=(r.detail*e.detailWeight+r.skin*e.skinWeight+r.saturation*e.saturationWeight+r.boost*e.boostWeight)/(n.width*n.height),r}function o(e,t,n,r){if(t.x>n||n>=t.x+t.width||t.y>r||r>=t.y+t.height)return e.outsideImportance;n=(n-t.x)/t.width,r=(r-t.y)/t.height;var a=2*m(.5-n),s=2*m(.5-r),i=Math.max(a-1+e.edgeRadius,0),o=Math.max(s-1+e.edgeRadius,0),l=(i*i+o*o)*e.edgeWeight,u=1.41-g(a*a+s*s);return e.ruleOfThirds&&(u+=1.2*Math.max(0,u+l+.5)*(b(a)+b(s))),u+l}function l(e,t,n,r){var a=g(t*t+n*n+r*r),s=t/a-e.skinColor[0],i=n/a-e.skinColor[1],o=r/a-e.skinColor[2];return 1-g(s*s+i*i+o*o)}function u(e,t,n){this.width=e,this.height=t,this.data=n?new Uint8ClampedArray(n):new Uint8ClampedArray(e*t*4)}function h(e,t){for(var n=e.data,r=e.width,a=Math.floor(e.width/t),s=Math.floor(e.height/t),i=new u(a,s),o=i.data,l=1/(t*t),h=0;h<s;h++)for(var d=0;d<a;d++){for(var p=4*(h*a+d),c=0,f=0,m=0,g=0,y=0,b=0,x=0;x<t;x++)for(var v=0;v<t;v++){var w=4*((h*t+x)*r+(d*t+v));c+=n[w],f+=n[w+1],m+=n[w+2],g+=n[w+3],y=Math.max(y,n[w]),b=Math.max(b,n[w+1])}o[p]=c*l*.5+.5*y,o[p+1]=f*l*.7+.3*b,o[p+2]=m*l,o[p+3]=g*l}return i}function d(e,t){var n=document.createElement("canvas");return n.width=e,n.height=t,n}function p(e){return{open:function(t){var n=t.naturalWidth||t.width,r=t.naturalHeight||t.height,s=e(n,r),i=s.getContext("2d");return!t.naturalWidth||t.naturalWidth==t.width&&t.naturalHeight==t.height?(s.width=t.width,s.height=t.height):(s.width=t.naturalWidth,s.height=t.naturalHeight),i.drawImage(t,0,0),a.Promise.resolve(s)},resample:function(t,n,r){return Promise.resolve(t).then(function(t){var s=e(~~n,~~r);return s.getContext("2d").drawImage(t,0,0,t.width,t.height,0,0,s.width,s.height),a.Promise.resolve(s)})},getData:function(e){return Promise.resolve(e).then(function(e){var t=e.getContext("2d").getImageData(0,0,e.width,e.height);return new u(e.width,e.height,t.data)})}}}a.Promise="undefined"!=typeof Promise?Promise:function(){throw new Error("No native promises and smartcrop.Promise not set.")},a.DEFAULTS={width:0,height:0,aspect:0,cropWidth:0,cropHeight:0,detailWeight:.2,skinColor:[.78,.57,.44],skinBias:.01,skinBrightnessMin:.2,skinBrightnessMax:1,skinThreshold:.8,skinWeight:1.8,saturationBrightnessMin:.05,saturationBrightnessMax:.9,saturationThreshold:.4,saturationBias:.2,saturationWeight:.1,scoreDownSample:8,step:8,scaleStep:.1,minScale:1,maxScale:1,edgeRadius:.4,edgeWeight:-20,outsideImportance:-.5,boostWeight:100,ruleOfThirds:!0,prescale:!0,imageOperations:null,canvasFactory:d,debug:!1},a.crop=function(e,t,n){var r=y({},a.DEFAULTS,t);r.aspect&&(r.width=r.aspect,r.height=1),null===r.imageOperations&&(r.imageOperations=p(r.canvasFactory));var o=r.imageOperations,d=1,m=1;return o.open(e,r.input).then(function(e){return r.width&&r.height&&(d=c(e.width/r.width,e.height/r.height),r.cropWidth=~~(r.width*d),r.cropHeight=~~(r.height*d),r.minScale=c(r.maxScale,f(1/d,r.minScale)),!1!==r.prescale&&((m=c(f(256/e.width,256/e.height),1))<1?(e=o.resample(e,e.width*m,e.height*m),r.cropWidth=~~(r.cropWidth*m),r.cropHeight=~~(r.cropHeight*m),r.boost&&(r.boost=r.boost.map(function(e){return{x:~~(e.x*m),y:~~(e.y*m),width:~~(e.width*m),height:~~(e.height*m),weight:e.weight}}))):m=1)),e}).then(function(e){return o.getData(e).then(function(e){for(var t=function(e,t){var n={},r=new u(t.width,t.height);(function(e,t){for(var n=e.data,r=t.data,a=e.width,s=e.height,i=0;i<s;i++)for(var o=0;o<a;o++){var l,u=4*(i*a+o);l=0===o||o>=a-1||0===i||i>=s-1?v(n,u):4*v(n,u)-v(n,u-4*a)-v(n,u-4)-v(n,u+4)-v(n,u+4*a),r[u+1]=l}})(t,r),function(e,t,n){for(var r=t.data,a=n.data,s=t.width,i=t.height,o=0;o<i;o++)for(var u=0;u<s;u++){var h=4*(o*s+u),d=x(r[h],r[h+1],r[h+2])/255,p=l(e,r[h],r[h+1],r[h+2]),c=p>e.skinThreshold,f=d>=e.skinBrightnessMin&&d<=e.skinBrightnessMax;a[h]=c&&f?(p-e.skinThreshold)*(255/(1-e.skinThreshold)):0}}(e,t,r),function(e,t,n){for(var r=t.data,a=n.data,s=t.width,i=t.height,o=0;o<i;o++)for(var l=0;l<s;l++){var u=4*(o*s+l),h=x(r[u],r[u+1],r[u+2])/255,d=w(r[u],r[u+1],r[u+2]),p=d>e.saturationThreshold,c=h>=e.saturationBrightnessMin&&h<=e.saturationBrightnessMax;a[u+2]=c&&p?(d-e.saturationThreshold)*(255/(1-e.saturationThreshold)):0}}(e,t,r),function(e,t){if(!e.boost)return;for(var n=t.data,r=0;r<t.width;r+=4)n[r+3]=0;for(r=0;r<e.boost.length;r++)s(e.boost[r],e,t)}(e,r);for(var a=h(r,e.scoreDownSample),o=-1/0,d=null,p=function(e,t,n){for(var r=[],a=c(t,n),s=e.cropWidth||a,i=e.cropHeight||a,o=e.maxScale;o>=e.minScale;o-=e.scaleStep)for(var l=0;l+i*o<=n;l+=e.step)for(var u=0;u+s*o<=t;u+=e.step)r.push({x:u,y:l,width:s*o,height:i*o});return r}(e,t.width,t.height),f=0,m=p.length;f<m;f++){var g=p[f];g.score=i(e,a,g),g.score.total>o&&(d=g,o=g.score.total)}n.topCrop=d,e.debug&&d&&(n.crops=p,n.debugOutput=r,n.debugOptions=e,n.debugTopCrop=y({},n.topCrop));return n}(r,e),a=t.crops||[t.topCrop],o=0,d=a.length;o<d;o++){var p=a[o];p.x=~~(p.x/m),p.y=~~(p.y/m),p.width=~~(p.width/m),p.height=~~(p.height/m)}return n&&n(t),t})})},a.isAvailable=function(e){if(!a.Promise)return!1;if((e?e.canvasFactory:d)===d&&!document.createElement("canvas").getContext("2d"))return!1;return!0},a.importance=o,a.ImgData=u,a._downSample=h,a._canvasImageOperations=p;var c=Math.min,f=Math.max,m=Math.abs,g=Math.sqrt;function y(e){for(var t=1,n=arguments.length;t<n;t++){var r=arguments[t];if(r)for(var a in r)e[a]=r[a]}return e}function b(e){return e=16*((e-1/3+1)%2*.5-.5),Math.max(1-e*e,0)}function x(e,t,n){return.5126*n+.7152*t+.0722*e}function v(e,t){return x(e[t],e[t+1],e[t+2])}function w(e,t,n){var r=f(e/255,t/255,n/255),a=c(e/255,t/255,n/255);if(r===a)return 0;var s=r-a;return(r+a)/2>.5?s/(2-r-a):s/(r+a)}void 0===(r=function(){return a}.call(t,n,t,e))||(e.exports=r),t.smartcrop=a,e.exports=a}()},81920(e,t,n){"use strict";n.d(t,{A:()=>d});const r="[FfƑƒ]",a="[aàáâãäåāăąαа]",s="[GgĜĝĞğĠġĢģɡɢ]",i="[\\s\\W]*",o="real|official",l=`${r}[iìíîïīĭįıɩɪіїι][nñńņňŋηи][dďđɗ]${i}${a}${i}${s}[rŕŗř]${a}[vν][eèéêëēĕėęěεе]`,u=`${r}${s}`,h=`${r}${a}${s}`;class d{static FULL_ID_REGEX=/^(cemetery|city|county|state|country|continent)_\d+$/;static NUMBER_ID_REGEX=/^\d+$/;static FG_PATTERN=new RegExp(`^${l}$|^(?=.*\\b(?:${o})\\b).*(?:^|\\s)${l}(?![\\S]).*$|^${u}$|^(?=.*\\b(?:${o})\\b).*(?:^|\\s)${u}(?![\\S]).*$|^${h}$|^(?=.*\\b(?:${o})\\b).*(?:^|\\s)${h}(?![\\S]).*$`,"iu");static validLocationId(e,t){let n=t?d.NUMBER_ID_REGEX:this.FULL_ID_REGEX;return Array.isArray(e)?void 0===e.find(e=>!n.test(e)):n.test(e)}static validLatitude(e){if(e||0===e){let t="string"==typeof e?Number.parseFloat(e):e;return t>=-90&&t<=90}return!1}static validLongitude(e){if(e||0===e){let t="string"==typeof e?Number.parseFloat(e):e;return t>=-180&&t<=180}return!1}static validEmail(e){return/^(([^<>()\[\]\\.,;:\s@"]+(\.[^<>()\[\]\\.,;:\s@"]+)*)|(".+"))@((\[[0-9]{1,3}\.[0-9]{1,3}\.[0-9]{1,3}\.[0-9]{1,3}])|(([a-zA-Z\-0-9]+\.)+[a-zA-Z]{2,}))$/.test(e)}static containsFGPattern(e){return this.FG_PATTERN.test(e)}}}}]);
//# sourceMappingURL=3919.411019e5423960c4130a.js.map