behind_glass/dist/assets/aiDepth.worker-DnfO2O4Q.js

var n1=Object.defineProperty;var o1=($s,Gr,jn)=>Gr in $s?n1($s,Gr,{enumerable:!0,configurable:!0,writable:!0,value:jn}):$s[Gr]=jn;var J=($s,Gr,jn)=>o1($s,typeof Gr!="symbol"?Gr+"":Gr,jn);(function(){"use strict";const $s=new Map,Gr=[],jn=(e,r,t)=>{if(r&&typeof r.init=="function"&&typeof r.createInferenceSessionHandler=="function"){const s=$s.get(e);if(s===void 0)$s.set(e,{backend:r,priority:t});else{if(s.priority>t)return;if(s.priority===t&&s.backend!==r)throw new Error(`cannot register backend "${e}" using priority ${t}`)}if(t>=0){const o=Gr.indexOf(e);o!==-1&&Gr.splice(o,1);for(let n=0;n<Gr.length;n++)if($s.get(Gr[n]).priority<=t){Gr.splice(n,0,e);return}Gr.push(e)}return}throw new TypeError("not a valid backend")},Xv=async e=>{const r=$s.get(e);if(!r)return"backend not found.";if(r.initialized)return r.backend;if(r.aborted)return r.error;{const t=!!r.initPromise;try{return t||(r.initPromise=r.backend.init(e)),await r.initPromise,r.initialized=!0,r.backend}catch(s){return t||(r.error=`${s}`,r.aborted=!0),r.error}finally{delete r.initPromise}}},Jv=async e=>{const r=e.executionProviders||[],t=r.map(l=>typeof l=="string"?l:l.name),s=t.length===0?Gr:t;let o;const n=[],a=new Set;for(const l of s){const c=await Xv(l);typeof c=="string"?n.push({name:l,err:c}):(o||(o=c),o===c&&a.add(l))}if(!o)throw new Error(`no available backend found. ERR: ${n.map(l=>`[${l.name}] ${l.err}`).join(", ")}`);for(const{name:l,err:c}of n)t.includes(l)&&console.warn(`removing requested execution provider "${l}" from session options because it is not available: ${c}`);const i=r.filter(l=>a.has(typeof l=="string"?l:l.name));return[o,new Proxy(e,{get:(l,c)=>c==="executionProviders"?i:Reflect.get(l,c)})]},Yv="1.21.0";let Gu="warning";const ms={wasm:{},webgl:{},webgpu:{},versions:{common:Yv},set logLevel(e){if(e!==void 0){if(typeof e!="string"||["verbose","info","warning","error","fatal"].indexOf(e)===-1)throw new Error(`Unsupported logging level: ${e}`);Gu=e}},get logLevel(){return Gu}};Object.defineProperty(ms,"logLevel",{enumerable:!0});const Zv=ms,ex=(e,r)=>{const t=typeof document<"u"?document.createElement("canvas"):new OffscreenCanvas(1,1);t.width=e.dims[3],t.height=e.dims[2];const s=t.getContext("2d");if(s!=null){let o,n;(r==null?void 0:r.tensorLayout)!==void 0&&r.tensorLayout==="NHWC"?(o=e.dims[2],n=e.dims[3]):(o=e.dims[3],n=e.dims[2]);const a=(r==null?void 0:r.format)!==void 0?r.format:"RGB",i=r==null?void 0:r.norm;let l,c;i===void 0||i.mean===void 0?l=[255,255,255,255]:typeof i.mean=="number"?l=[i.mean,i.mean,i.mean,i.mean]:(l=[i.mean[0],i.mean[1],i.mean[2],0],i.mean[3]!==void 0&&(l[3]=i.mean[3])),i===void 0||i.bias===void 0?c=[0,0,0,0]:typeof i.bias=="number"?c=[i.bias,i.bias,i.bias,i.bias]:(c=[i.bias[0],i.bias[1],i.bias[2],0],i.bias[3]!==void 0&&(c[3]=i.bias[3]));const p=n*o;let d=0,u=p,f=p*2,_=-1;a==="RGBA"?(d=0,u=p,f=p*2,_=p*3):a==="RGB"?(d=0,u=p,f=p*2):a==="RBG"&&(d=0,f=p,u=p*2);for(let b=0;b<n;b++)for(let k=0;k<o;k++){const w=(e.data[d++]-c[0])*l[0],y=(e.data[u++]-c[1])*l[1],$=(e.data[f++]-c[2])*l[2],T=_===-1?255:(e.data[_++]-c[3])*l[3];s.fillStyle="rgba("+w+","+y+","+$+","+T+")",s.fillRect(k,b,1,1)}if("toDataURL"in t)return t.toDataURL();throw new Error("toDataURL is not supported")}else throw new Error("Can not access image data")},tx=(e,r)=>{const t=typeof document<"u"?document.createElement("canvas").getContext("2d"):new OffscreenCanvas(1,1).getContext("2d");let s;if(t!=null){let o,n,a;(r==null?void 0:r.tensorLayout)!==void 0&&r.tensorLayout==="NHWC"?(o=e.dims[2],n=e.dims[1],a=e.dims[3]):(o=e.dims[3],n=e.dims[2],a=e.dims[1]);const i=r!==void 0&&r.format!==void 0?r.format:"RGB",l=r==null?void 0:r.norm;let c,p;l===void 0||l.mean===void 0?c=[255,255,255,255]:typeof l.mean=="number"?c=[l.mean,l.mean,l.mean,l.mean]:(c=[l.mean[0],l.mean[1],l.mean[2],255],l.mean[3]!==void 0&&(c[3]=l.mean[3])),l===void 0||l.bias===void 0?p=[0,0,0,0]:typeof l.bias=="number"?p=[l.bias,l.bias,l.bias,l.bias]:(p=[l.bias[0],l.bias[1],l.bias[2],0],l.bias[3]!==void 0&&(p[3]=l.bias[3]));const d=n*o;if(r!==void 0&&(r.format!==void 0&&a===4&&r.format!=="RGBA"||a===3&&r.format!=="RGB"&&r.format!=="BGR"))throw new Error("Tensor format doesn't match input tensor dims");const u=4;let f=0,_=1,b=2,k=3,w=0,y=d,$=d*2,T=-1;i==="RGBA"?(w=0,y=d,$=d*2,T=d*3):i==="RGB"?(w=0,y=d,$=d*2):i==="RBG"&&(w=0,$=d,y=d*2),s=t.createImageData(o,n);for(let M=0;M<n*o;f+=u,_+=u,b+=u,k+=u,M++)s.data[f]=(e.data[w++]-p[0])*c[0],s.data[_]=(e.data[y++]-p[1])*c[1],s.data[b]=(e.data[$++]-p[2])*c[2],s.data[k]=T===-1?255:(e.data[T++]-p[3])*c[3]}else throw new Error("Can not access image data");return s},yi=(e,r)=>{if(e===void 0)throw new Error("Image buffer must be defined");if(r.height===void 0||r.width===void 0)throw new Error("Image height and width must be defined");if(r.tensorLayout==="NHWC")throw new Error("NHWC Tensor layout is not supported yet");const{height:t,width:s}=r,o=r.norm??{mean:255,bias:0};let n,a;typeof o.mean=="number"?n=[o.mean,o.mean,o.mean,o.mean]:n=[o.mean[0],o.mean[1],o.mean[2],o.mean[3]??255],typeof o.bias=="number"?a=[o.bias,o.bias,o.bias,o.bias]:a=[o.bias[0],o.bias[1],o.bias[2],o.bias[3]??0];const i=r.format!==void 0?r.format:"RGBA",l=r.tensorFormat!==void 0&&r.tensorFormat!==void 0?r.tensorFormat:"RGB",c=t*s,p=l==="RGBA"?new Float32Array(c*4):new Float32Array(c*3);let d=4,u=0,f=1,_=2,b=3,k=0,w=c,y=c*2,$=-1;i==="RGB"&&(d=3,u=0,f=1,_=2,b=-1),l==="RGBA"?$=c*3:l==="RBG"?(k=0,y=c,w=c*2):l==="BGR"&&(y=0,w=c,k=c*2);for(let M=0;M<c;M++,u+=d,_+=d,f+=d,b+=d)p[k++]=(e[u]+a[0])/n[0],p[w++]=(e[f]+a[1])/n[1],p[y++]=(e[_]+a[2])/n[2],$!==-1&&b!==-1&&(p[$++]=(e[b]+a[3])/n[3]);return l==="RGBA"?new os("float32",p,[1,4,t,s]):new os("float32",p,[1,3,t,s])},rx=async(e,r)=>{const t=typeof HTMLImageElement<"u"&&e instanceof HTMLImageElement,s=typeof ImageData<"u"&&e instanceof ImageData,o=typeof ImageBitmap<"u"&&e instanceof ImageBitmap,n=typeof e=="string";let a,i=r??{};const l=()=>{if(typeof document<"u")return document.createElement("canvas");if(typeof OffscreenCanvas<"u")return new OffscreenCanvas(1,1);throw new Error("Canvas is not supported")},c=p=>typeof HTMLCanvasElement<"u"&&p instanceof HTMLCanvasElement||p instanceof OffscreenCanvas?p.getContext("2d"):null;if(t){const p=l();p.width=e.width,p.height=e.height;const d=c(p);if(d!=null){let u=e.height,f=e.width;if(r!==void 0&&r.resizedHeight!==void 0&&r.resizedWidth!==void 0&&(u=r.resizedHeight,f=r.resizedWidth),r!==void 0){if(i=r,r.tensorFormat!==void 0)throw new Error("Image input config format must be RGBA for HTMLImageElement");i.tensorFormat="RGBA",i.height=u,i.width=f}else i.tensorFormat="RGBA",i.height=u,i.width=f;d.drawImage(e,0,0),a=d.getImageData(0,0,f,u).data}else throw new Error("Can not access image data")}else if(s){let p,d;if(r!==void 0&&r.resizedWidth!==void 0&&r.resizedHeight!==void 0?(p=r.resizedHeight,d=r.resizedWidth):(p=e.height,d=e.width),r!==void 0&&(i=r),i.format="RGBA",i.height=p,i.width=d,r!==void 0){const u=l();u.width=d,u.height=p;const f=c(u);if(f!=null)f.putImageData(e,0,0),a=f.getImageData(0,0,d,p).data;else throw new Error("Can not access image data")}else a=e.data}else if(o){if(r===void 0)throw new Error("Please provide image config with format for Imagebitmap");const p=l();p.width=e.width,p.height=e.height;const d=c(p);if(d!=null){const u=e.height,f=e.width;return d.drawImage(e,0,0,f,u),a=d.getImageData(0,0,f,u).data,i.height=u,i.width=f,yi(a,i)}else throw new Error("Can not access image data")}else{if(n)return new Promise((p,d)=>{const u=l(),f=c(u);if(!e||!f)return d();const _=new Image;_.crossOrigin="Anonymous",_.src=e,_.onload=()=>{u.width=_.width,u.height=_.height,f.drawImage(_,0,0,u.width,u.height);const b=f.getImageData(0,0,u.width,u.height);i.height=u.height,i.width=u.width,p(yi(b.data,i))}});throw new Error("Input data provided is not supported - aborted tensor creation")}if(a!==void 0)return yi(a,i);throw new Error("Input data provided is not supported - aborted tensor creation")},sx=(e,r)=>{const{width:t,height:s,download:o,dispose:n}=r,a=[1,s,t,4];return new os({location:"texture",type:"float32",texture:e,dims:a,download:o,dispose:n})},nx=(e,r)=>{const{dataType:t,dims:s,download:o,dispose:n}=r;return new os({location:"gpu-buffer",type:t??"float32",gpuBuffer:e,dims:s,download:o,dispose:n})},ox=(e,r)=>{const{dataType:t,dims:s,download:o,dispose:n}=r;return new os({location:"ml-tensor",type:t??"float32",mlTensor:e,dims:s,download:o,dispose:n})},ax=(e,r,t)=>new os({location:"cpu-pinned",type:e,data:r,dims:t??[r.length]}),Nn=new Map([["float32",Float32Array],["uint8",Uint8Array],["int8",Int8Array],["uint16",Uint16Array],["int16",Int16Array],["int32",Int32Array],["bool",Uint8Array],["float64",Float64Array],["uint32",Uint32Array],["int4",Uint8Array],["uint4",Uint8Array]]),ea=new Map([[Float32Array,"float32"],[Uint8Array,"uint8"],[Int8Array,"int8"],[Uint16Array,"uint16"],[Int16Array,"int16"],[Int32Array,"int32"],[Float64Array,"float64"],[Uint32Array,"uint32"]]);let Ku=!1;const ix=()=>{if(!Ku){Ku=!0;const e=typeof BigInt64Array<"u"&&BigInt64Array.from,r=typeof BigUint64Array<"u"&&BigUint64Array.from,t=globalThis.Float16Array,s=typeof t<"u"&&t.from;e&&(Nn.set("int64",BigInt64Array),ea.set(BigInt64Array,"int64")),r&&(Nn.set("uint64",BigUint64Array),ea.set(BigUint64Array,"uint64")),s?(Nn.set("float16",t),ea.set(t,"float16")):Nn.set("float16",Uint16Array)}},lx=e=>{let r=1;for(let t=0;t<e.length;t++){const s=e[t];if(typeof s!="number"||!Number.isSafeInteger(s))throw new TypeError(`dims[${t}] must be an integer, got: ${s}`);if(s<0)throw new RangeError(`dims[${t}] must be a non-negative integer, got: ${s}`);r*=s}return r},cx=(e,r)=>{switch(e.location){case"cpu":return new os(e.type,e.data,r);case"cpu-pinned":return new os({location:"cpu-pinned",data:e.data,type:e.type,dims:r});case"texture":return new os({location:"texture",texture:e.texture,type:e.type,dims:r});case"gpu-buffer":return new os({location:"gpu-buffer",gpuBuffer:e.gpuBuffer,type:e.type,dims:r});case"ml-tensor":return new os({location:"ml-tensor",mlTensor:e.mlTensor,type:e.type,dims:r});default:throw new Error(`tensorReshape: tensor location ${e.location} is not supported`)}};let os=class{constructor(r,t,s){ix();let o,n;if(typeof r=="object"&&"location"in r)switch(this.dataLocation=r.location,o=r.type,n=r.dims,r.location){case"cpu-pinned":{const i=Nn.get(o);if(!i)throw new TypeError(`unsupported type "${o}" to create tensor from pinned buffer`);if(!(r.data instanceof i))throw new TypeError(`buffer should be of type ${i.name}`);this.cpuData=r.data;break}case"texture":{if(o!=="float32")throw new TypeError(`unsupported type "${o}" to create tensor from texture`);this.gpuTextureData=r.texture,this.downloader=r.download,this.disposer=r.dispose;break}case"gpu-buffer":{if(o!=="float32"&&o!=="float16"&&o!=="int32"&&o!=="int64"&&o!=="uint32"&&o!=="uint8"&&o!=="bool"&&o!=="uint4"&&o!=="int4")throw new TypeError(`unsupported type "${o}" to create tensor from gpu buffer`);this.gpuBufferData=r.gpuBuffer,this.downloader=r.download,this.disposer=r.dispose;break}case"ml-tensor":{if(o!=="float32"&&o!=="float16"&&o!=="int32"&&o!=="int64"&&o!=="uint32"&&o!=="uint64"&&o!=="int8"&&o!=="uint8"&&o!=="bool"&&o!=="uint4"&&o!=="int4")throw new TypeError(`unsupported type "${o}" to create tensor from MLTensor`);this.mlTensorData=r.mlTensor,this.downloader=r.download,this.disposer=r.dispose;break}default:throw new Error(`Tensor constructor: unsupported location '${this.dataLocation}'`)}else{let i,l;if(typeof r=="string")if(o=r,l=s,r==="string"){if(!Array.isArray(t))throw new TypeError("A string tensor's data must be a string array.");i=t}else{const c=Nn.get(r);if(c===void 0)throw new TypeError(`Unsupported tensor type: ${r}.`);if(Array.isArray(t)){if(r==="float16"&&c===Uint16Array||r==="uint4"||r==="int4")throw new TypeError(`Creating a ${r} tensor from number array is not supported. Please use ${c.name} as data.`);r==="uint64"||r==="int64"?i=c.from(t,BigInt):i=c.from(t)}else if(t instanceof c)i=t;else if(t instanceof Uint8ClampedArray)if(r==="uint8")i=Uint8Array.from(t);else throw new TypeError("A Uint8ClampedArray tensor's data must be type of uint8");else if(r==="float16"&&t instanceof Uint16Array&&c!==Uint16Array)i=new globalThis.Float16Array(t.buffer,t.byteOffset,t.length);else throw new TypeError(`A ${o} tensor's data must be type of ${c}`)}else if(l=t,Array.isArray(r)){if(r.length===0)throw new TypeError("Tensor type cannot be inferred from an empty array.");const c=typeof r[0];if(c==="string")o="string",i=r;else if(c==="boolean")o="bool",i=Uint8Array.from(r);else throw new TypeError(`Invalid element type of data array: ${c}.`)}else if(r instanceof Uint8ClampedArray)o="uint8",i=Uint8Array.from(r);else{const c=ea.get(r.constructor);if(c===void 0)throw new TypeError(`Unsupported type for tensor data: ${r.constructor}.`);o=c,i=r}if(l===void 0)l=[i.length];else if(!Array.isArray(l))throw new TypeError("A tensor's dims must be a number array");n=l,this.cpuData=i,this.dataLocation="cpu"}const a=lx(n);if(this.cpuData&&a!==this.cpuData.length&&!((o==="uint4"||o==="int4")&&Math.ceil(a/2)===this.cpuData.length))throw new Error(`Tensor's size(${a}) does not match data length(${this.cpuData.length}).`);this.type=o,this.dims=n,this.size=a}static async fromImage(r,t){return rx(r,t)}static fromTexture(r,t){return sx(r,t)}static fromGpuBuffer(r,t){return nx(r,t)}static fromMLTensor(r,t){return ox(r,t)}static fromPinnedBuffer(r,t,s){return ax(r,t,s)}toDataURL(r){return ex(this,r)}toImageData(r){return tx(this,r)}get data(){if(this.ensureValid(),!this.cpuData)throw new Error("The data is not on CPU. Use `getData()` to download GPU data to CPU, or use `texture` or `gpuBuffer` property to access the GPU data directly.");return this.cpuData}get location(){return this.dataLocation}get texture(){if(this.ensureValid(),!this.gpuTextureData)throw new Error("The data is not stored as a WebGL texture.");return this.gpuTextureData}get gpuBuffer(){if(this.ensureValid(),!this.gpuBufferData)throw new Error("The data is not stored as a WebGPU buffer.");return this.gpuBufferData}get mlTensor(){if(this.ensureValid(),!this.mlTensorData)throw new Error("The data is not stored as a WebNN MLTensor.");return this.mlTensorData}async getData(r){switch(this.ensureValid(),this.dataLocation){case"cpu":case"cpu-pinned":return this.data;case"texture":case"gpu-buffer":case"ml-tensor":{if(!this.downloader)throw new Error("The current tensor is not created with a specified data downloader.");if(this.isDownloading)throw new Error("The current tensor is being downloaded.");try{this.isDownloading=!0;const t=await this.downloader();return this.downloader=void 0,this.dataLocation="cpu",this.cpuData=t,r&&this.disposer&&(this.disposer(),this.disposer=void 0),t}finally{this.isDownloading=!1}}default:throw new Error(`cannot get data from location: ${this.dataLocation}`)}}dispose(){if(this.isDownloading)throw new Error("The current tensor is being downloaded.");this.disposer&&(this.disposer(),this.disposer=void 0),this.cpuData=void 0,this.gpuTextureData=void 0,this.gpuBufferData=void 0,this.mlTensorData=void 0,this.downloader=void 0,this.isDownloading=void 0,this.dataLocation="none"}ensureValid(){if(this.dataLocation==="none")throw new Error("The tensor is disposed.")}reshape(r){if(this.ensureValid(),this.downloader||this.disposer)throw new Error("Cannot reshape a tensor that owns GPU resource.");return cx(this,r)}};const Vn=os,Hu=(e,r)=>{(typeof ms.trace>"u"?!ms.wasm.trace:!ms.trace)||console.timeStamp(`${e}::ORT::${r}`)},qu=(e,r)=>{var o;const t=((o=new Error().stack)==null?void 0:o.split(/\r\n|\r|\n/g))||[];let s=!1;for(let n=0;n<t.length;n++){if(s&&!t[n].includes("TRACE_FUNC")){let a=`FUNC_${e}::${t[n].trim().split(" ")[1]}`;r&&(a+=`::${r}`),Hu("CPU",a);return}t[n].includes("TRACE_FUNC")&&(s=!0)}},vi=e=>{(typeof ms.trace>"u"?!ms.wasm.trace:!ms.trace)||qu("BEGIN",e)},xi=e=>{(typeof ms.trace>"u"?!ms.wasm.trace:!ms.trace)||qu("END",e)};var ux=Object.freeze({__proto__:null,InferenceSession:class Hv{constructor(r){this.handler=r}async run(r,t,s){vi();const o={};let n={};if(typeof r!="object"||r===null||r instanceof Vn||Array.isArray(r))throw new TypeError("'feeds' must be an object that use input names as keys and OnnxValue as corresponding values.");let a=!0;if(typeof t=="object"){if(t===null)throw new TypeError("Unexpected argument[1]: cannot be null.");if(t instanceof Vn)throw new TypeError("'fetches' cannot be a Tensor");if(Array.isArray(t)){if(t.length===0)throw new TypeError("'fetches' cannot be an empty array.");a=!1;for(const c of t){if(typeof c!="string")throw new TypeError("'fetches' must be a string array or an object.");if(this.outputNames.indexOf(c)===-1)throw new RangeError(`'fetches' contains invalid output name: ${c}.`);o[c]=null}if(typeof s=="object"&&s!==null)n=s;else if(typeof s<"u")throw new TypeError("'options' must be an object.")}else{let c=!1;const p=Object.getOwnPropertyNames(t);for(const d of this.outputNames)if(p.indexOf(d)!==-1){const u=t[d];(u===null||u instanceof Vn)&&(c=!0,a=!1,o[d]=u)}if(c){if(typeof s=="object"&&s!==null)n=s;else if(typeof s<"u")throw new TypeError("'options' must be an object.")}else n=t}}else if(typeof t<"u")throw new TypeError("Unexpected argument[1]: must be 'fetches' or 'options'.");for(const c of this.inputNames)if(typeof r[c]>"u")throw new Error(`input '${c}' is missing in 'feeds'.`);if(a)for(const c of this.outputNames)o[c]=null;const i=await this.handler.run(r,o,n),l={};for(const c in i)if(Object.hasOwnProperty.call(i,c)){const p=i[c];p instanceof Vn?l[c]=p:l[c]=new Vn(p.type,p.data,p.dims)}return xi(),l}async release(){return this.handler.dispose()}static async create(r,t,s,o){vi();let n,a={};if(typeof r=="string"){if(n=r,typeof t=="object"&&t!==null)a=t;else if(typeof t<"u")throw new TypeError("'options' must be an object.")}else if(r instanceof Uint8Array){if(n=r,typeof t=="object"&&t!==null)a=t;else if(typeof t<"u")throw new TypeError("'options' must be an object.")}else if(r instanceof ArrayBuffer||typeof SharedArrayBuffer<"u"&&r instanceof SharedArrayBuffer){const p=r;let d=0,u=r.byteLength;if(typeof t=="object"&&t!==null)a=t;else if(typeof t=="number"){if(d=t,!Number.isSafeInteger(d))throw new RangeError("'byteOffset' must be an integer.");if(d<0||d>=p.byteLength)throw new RangeError(`'byteOffset' is out of range [0, ${p.byteLength}).`);if(u=r.byteLength-d,typeof s=="number"){if(u=s,!Number.isSafeInteger(u))throw new RangeError("'byteLength' must be an integer.");if(u<=0||d+u>p.byteLength)throw new RangeError(`'byteLength' is out of range (0, ${p.byteLength-d}].`);if(typeof o=="object"&&o!==null)a=o;else if(typeof o<"u")throw new TypeError("'options' must be an object.")}else if(typeof s<"u")throw new TypeError("'byteLength' must be a number.")}else if(typeof t<"u")throw new TypeError("'options' must be an object.");n=new Uint8Array(p,d,u)}else throw new TypeError("Unexpected argument[0]: must be 'path' or 'buffer'.");const[i,l]=await Jv(a),c=await i.createInferenceSessionHandler(n,l);return xi(),new Hv(c)}startProfiling(){this.handler.startProfiling()}endProfiling(){this.handler.endProfiling()}get inputNames(){return this.handler.inputNames}get outputNames(){return this.handler.outputNames}},TRACE:Hu,TRACE_FUNC_BEGIN:vi,TRACE_FUNC_END:xi,Tensor:Vn,env:Zv,registerBackend:jn});/*!
 * ONNX Runtime Web v1.22.0-dev.20250409-89f8206ba4
 * Copyright (c) Microsoft Corporation. All rights reserved.
 * Licensed under the MIT License.
 */var Ti=Object.defineProperty,dx=Object.getOwnPropertyDescriptor,px=Object.getOwnPropertyNames,mx=Object.prototype.hasOwnProperty,hx=(e=>typeof require<"u"?require:typeof Proxy<"u"?new Proxy(e,{get:(r,t)=>(typeof require<"u"?require:r)[t]}):e)(function(e){if(typeof require<"u")return require.apply(this,arguments);throw Error('Dynamic require of "'+e+'" is not supported')}),Ue=(e,r)=>()=>(e&&(r=e(e=0)),r),Un=(e,r)=>{for(var t in r)Ti(e,t,{get:r[t],enumerable:!0})},_x=(e,r,t,s)=>{if(r&&typeof r=="object"||typeof r=="function")for(let o of px(r))!mx.call(e,o)&&o!==t&&Ti(e,o,{get:()=>r[o],enumerable:!(s=dx(r,o))||s.enumerable});return e},uo=e=>_x(Ti({},"__esModule",{value:!0}),e),po,Zs,mn,Qu,Xu,Ju=Ue(()=>{po=new Map,Zs=[],mn=(e,r,t)=>{if(r&&typeof r.init=="function"&&typeof r.createInferenceSessionHandler=="function"){let s=po.get(e);if(s===void 0)po.set(e,{backend:r,priority:t});else{if(s.priority>t)return;if(s.priority===t&&s.backend!==r)throw new Error(`cannot register backend "${e}" using priority ${t}`)}if(t>=0){let o=Zs.indexOf(e);o!==-1&&Zs.splice(o,1);for(let n=0;n<Zs.length;n++)if(po.get(Zs[n]).priority<=t){Zs.splice(n,0,e);return}Zs.push(e)}return}throw new TypeError("not a valid backend")},Qu=async e=>{let r=po.get(e);if(!r)return"backend not found.";if(r.initialized)return r.backend;if(r.aborted)return r.error;{let t=!!r.initPromise;try{return t||(r.initPromise=r.backend.init(e)),await r.initPromise,r.initialized=!0,r.backend}catch(s){return t||(r.error=`${s}`,r.aborted=!0),r.error}finally{delete r.initPromise}}},Xu=async e=>{let r=e.executionProviders||[],t=r.map(l=>typeof l=="string"?l:l.name),s=t.length===0?Zs:t,o,n=[],a=new Set;for(let l of s){let c=await Qu(l);typeof c=="string"?n.push({name:l,err:c}):(o||(o=c),o===c&&a.add(l))}if(!o)throw new Error(`no available backend found. ERR: ${n.map(l=>`[${l.name}] ${l.err}`).join(", ")}`);for(let{name:l,err:c}of n)t.includes(l)&&console.warn(`removing requested execution provider "${l}" from session options because it is not available: ${c}`);let i=r.filter(l=>a.has(typeof l=="string"?l:l.name));return[o,new Proxy(e,{get:(l,c)=>c==="executionProviders"?i:Reflect.get(l,c)})]}}),fx=Ue(()=>{Ju()}),Yu,gx=Ue(()=>{Yu="1.22.0-dev.20250409-89f8206ba4"}),Ei,as,Zu=Ue(()=>{gx(),Ei="warning",as={wasm:{},webgl:{},webgpu:{},versions:{common:Yu},set logLevel(e){if(e!==void 0){if(typeof e!="string"||["verbose","info","warning","error","fatal"].indexOf(e)===-1)throw new Error(`Unsupported logging level: ${e}`);Ei=e}},get logLevel(){return Ei}},Object.defineProperty(as,"logLevel",{enumerable:!0})}),tr,wx=Ue(()=>{Zu(),tr=as}),ed,td,Mx=Ue(()=>{ed=(e,r)=>{let t=typeof document<"u"?document.createElement("canvas"):new OffscreenCanvas(1,1);t.width=e.dims[3],t.height=e.dims[2];let s=t.getContext("2d");if(s!=null){let o,n;(r==null?void 0:r.tensorLayout)!==void 0&&r.tensorLayout==="NHWC"?(o=e.dims[2],n=e.dims[3]):(o=e.dims[3],n=e.dims[2]);let a=(r==null?void 0:r.format)!==void 0?r.format:"RGB",i=r==null?void 0:r.norm,l,c;i===void 0||i.mean===void 0?l=[255,255,255,255]:typeof i.mean=="number"?l=[i.mean,i.mean,i.mean,i.mean]:(l=[i.mean[0],i.mean[1],i.mean[2],0],i.mean[3]!==void 0&&(l[3]=i.mean[3])),i===void 0||i.bias===void 0?c=[0,0,0,0]:typeof i.bias=="number"?c=[i.bias,i.bias,i.bias,i.bias]:(c=[i.bias[0],i.bias[1],i.bias[2],0],i.bias[3]!==void 0&&(c[3]=i.bias[3]));let p=n*o,d=0,u=p,f=p*2,_=-1;a==="RGBA"?(d=0,u=p,f=p*2,_=p*3):a==="RGB"?(d=0,u=p,f=p*2):a==="RBG"&&(d=0,f=p,u=p*2);for(let b=0;b<n;b++)for(let k=0;k<o;k++){let w=(e.data[d++]-c[0])*l[0],y=(e.data[u++]-c[1])*l[1],$=(e.data[f++]-c[2])*l[2],T=_===-1?255:(e.data[_++]-c[3])*l[3];s.fillStyle="rgba("+w+","+y+","+$+","+T+")",s.fillRect(k,b,1,1)}if("toDataURL"in t)return t.toDataURL();throw new Error("toDataURL is not supported")}else throw new Error("Can not access image data")},td=(e,r)=>{let t=typeof document<"u"?document.createElement("canvas").getContext("2d"):new OffscreenCanvas(1,1).getContext("2d"),s;if(t!=null){let o,n,a;(r==null?void 0:r.tensorLayout)!==void 0&&r.tensorLayout==="NHWC"?(o=e.dims[2],n=e.dims[1],a=e.dims[3]):(o=e.dims[3],n=e.dims[2],a=e.dims[1]);let i=r!==void 0&&r.format!==void 0?r.format:"RGB",l=r==null?void 0:r.norm,c,p;l===void 0||l.mean===void 0?c=[255,255,255,255]:typeof l.mean=="number"?c=[l.mean,l.mean,l.mean,l.mean]:(c=[l.mean[0],l.mean[1],l.mean[2],255],l.mean[3]!==void 0&&(c[3]=l.mean[3])),l===void 0||l.bias===void 0?p=[0,0,0,0]:typeof l.bias=="number"?p=[l.bias,l.bias,l.bias,l.bias]:(p=[l.bias[0],l.bias[1],l.bias[2],0],l.bias[3]!==void 0&&(p[3]=l.bias[3]));let d=n*o;if(r!==void 0&&(r.format!==void 0&&a===4&&r.format!=="RGBA"||a===3&&r.format!=="RGB"&&r.format!=="BGR"))throw new Error("Tensor format doesn't match input tensor dims");let u=4,f=0,_=1,b=2,k=3,w=0,y=d,$=d*2,T=-1;i==="RGBA"?(w=0,y=d,$=d*2,T=d*3):i==="RGB"?(w=0,y=d,$=d*2):i==="RBG"&&(w=0,$=d,y=d*2),s=t.createImageData(o,n);for(let M=0;M<n*o;f+=u,_+=u,b+=u,k+=u,M++)s.data[f]=(e.data[w++]-p[0])*c[0],s.data[_]=(e.data[y++]-p[1])*c[1],s.data[b]=(e.data[$++]-p[2])*c[2],s.data[k]=T===-1?255:(e.data[T++]-p[3])*c[3]}else throw new Error("Can not access image data");return s}}),ta,rd,sd,nd,od,ad,bx=Ue(()=>{Ci(),ta=(e,r)=>{if(e===void 0)throw new Error("Image buffer must be defined");if(r.height===void 0||r.width===void 0)throw new Error("Image height and width must be defined");if(r.tensorLayout==="NHWC")throw new Error("NHWC Tensor layout is not supported yet");let{height:t,width:s}=r,o=r.norm??{mean:255,bias:0},n,a;typeof o.mean=="number"?n=[o.mean,o.mean,o.mean,o.mean]:n=[o.mean[0],o.mean[1],o.mean[2],o.mean[3]??255],typeof o.bias=="number"?a=[o.bias,o.bias,o.bias,o.bias]:a=[o.bias[0],o.bias[1],o.bias[2],o.bias[3]??0];let i=r.format!==void 0?r.format:"RGBA",l=r.tensorFormat!==void 0&&r.tensorFormat!==void 0?r.tensorFormat:"RGB",c=t*s,p=l==="RGBA"?new Float32Array(c*4):new Float32Array(c*3),d=4,u=0,f=1,_=2,b=3,k=0,w=c,y=c*2,$=-1;i==="RGB"&&(d=3,u=0,f=1,_=2,b=-1),l==="RGBA"?$=c*3:l==="RBG"?(k=0,y=c,w=c*2):l==="BGR"&&(y=0,w=c,k=c*2);for(let T=0;T<c;T++,u+=d,_+=d,f+=d,b+=d)p[k++]=(e[u]+a[0])/n[0],p[w++]=(e[f]+a[1])/n[1],p[y++]=(e[_]+a[2])/n[2],$!==-1&&b!==-1&&(p[$++]=(e[b]+a[3])/n[3]);return l==="RGBA"?new es("float32",p,[1,4,t,s]):new es("float32",p,[1,3,t,s])},rd=async(e,r)=>{let t=typeof HTMLImageElement<"u"&&e instanceof HTMLImageElement,s=typeof ImageData<"u"&&e instanceof ImageData,o=typeof ImageBitmap<"u"&&e instanceof ImageBitmap,n=typeof e=="string",a,i=r??{},l=()=>{if(typeof document<"u")return document.createElement("canvas");if(typeof OffscreenCanvas<"u")return new OffscreenCanvas(1,1);throw new Error("Canvas is not supported")},c=p=>typeof HTMLCanvasElement<"u"&&p instanceof HTMLCanvasElement||p instanceof OffscreenCanvas?p.getContext("2d"):null;if(t){let p=l();p.width=e.width,p.height=e.height;let d=c(p);if(d!=null){let u=e.height,f=e.width;if(r!==void 0&&r.resizedHeight!==void 0&&r.resizedWidth!==void 0&&(u=r.resizedHeight,f=r.resizedWidth),r!==void 0){if(i=r,r.tensorFormat!==void 0)throw new Error("Image input config format must be RGBA for HTMLImageElement");i.tensorFormat="RGBA",i.height=u,i.width=f}else i.tensorFormat="RGBA",i.height=u,i.width=f;d.drawImage(e,0,0),a=d.getImageData(0,0,f,u).data}else throw new Error("Can not access image data")}else if(s){let p,d;if(r!==void 0&&r.resizedWidth!==void 0&&r.resizedHeight!==void 0?(p=r.resizedHeight,d=r.resizedWidth):(p=e.height,d=e.width),r!==void 0&&(i=r),i.format="RGBA",i.height=p,i.width=d,r!==void 0){let u=l();u.width=d,u.height=p;let f=c(u);if(f!=null)f.putImageData(e,0,0),a=f.getImageData(0,0,d,p).data;else throw new Error("Can not access image data")}else a=e.data}else if(o){if(r===void 0)throw new Error("Please provide image config with format for Imagebitmap");let p=l();p.width=e.width,p.height=e.height;let d=c(p);if(d!=null){let u=e.height,f=e.width;return d.drawImage(e,0,0,f,u),a=d.getImageData(0,0,f,u).data,i.height=u,i.width=f,ta(a,i)}else throw new Error("Can not access image data")}else{if(n)return new Promise((p,d)=>{let u=l(),f=c(u);if(!e||!f)return d();let _=new Image;_.crossOrigin="Anonymous",_.src=e,_.onload=()=>{u.width=_.width,u.height=_.height,f.drawImage(_,0,0,u.width,u.height);let b=f.getImageData(0,0,u.width,u.height);i.height=u.height,i.width=u.width,p(ta(b.data,i))}});throw new Error("Input data provided is not supported - aborted tensor creation")}if(a!==void 0)return ta(a,i);throw new Error("Input data provided is not supported - aborted tensor creation")},sd=(e,r)=>{let{width:t,height:s,download:o,dispose:n}=r,a=[1,s,t,4];return new es({location:"texture",type:"float32",texture:e,dims:a,download:o,dispose:n})},nd=(e,r)=>{let{dataType:t,dims:s,download:o,dispose:n}=r;return new es({location:"gpu-buffer",type:t??"float32",gpuBuffer:e,dims:s,download:o,dispose:n})},od=(e,r)=>{let{dataType:t,dims:s,download:o,dispose:n}=r;return new es({location:"ml-tensor",type:t??"float32",mlTensor:e,dims:s,download:o,dispose:n})},ad=(e,r,t)=>new es({location:"cpu-pinned",type:e,data:r,dims:t??[r.length]})}),hn,mo,Pi,id,yx=Ue(()=>{hn=new Map([["float32",Float32Array],["uint8",Uint8Array],["int8",Int8Array],["uint16",Uint16Array],["int16",Int16Array],["int32",Int32Array],["bool",Uint8Array],["float64",Float64Array],["uint32",Uint32Array],["int4",Uint8Array],["uint4",Uint8Array]]),mo=new Map([[Float32Array,"float32"],[Uint8Array,"uint8"],[Int8Array,"int8"],[Uint16Array,"uint16"],[Int16Array,"int16"],[Int32Array,"int32"],[Float64Array,"float64"],[Uint32Array,"uint32"]]),Pi=!1,id=()=>{if(!Pi){Pi=!0;let e=typeof BigInt64Array<"u"&&BigInt64Array.from,r=typeof BigUint64Array<"u"&&BigUint64Array.from,t=globalThis.Float16Array,s=typeof t<"u"&&t.from;e&&(hn.set("int64",BigInt64Array),mo.set(BigInt64Array,"int64")),r&&(hn.set("uint64",BigUint64Array),mo.set(BigUint64Array,"uint64")),s?(hn.set("float16",t),mo.set(t,"float16")):hn.set("float16",Uint16Array)}}}),ld,cd,vx=Ue(()=>{Ci(),ld=e=>{let r=1;for(let t=0;t<e.length;t++){let s=e[t];if(typeof s!="number"||!Number.isSafeInteger(s))throw new TypeError(`dims[${t}] must be an integer, got: ${s}`);if(s<0)throw new RangeError(`dims[${t}] must be a non-negative integer, got: ${s}`);r*=s}return r},cd=(e,r)=>{switch(e.location){case"cpu":return new es(e.type,e.data,r);case"cpu-pinned":return new es({location:"cpu-pinned",data:e.data,type:e.type,dims:r});case"texture":return new es({location:"texture",texture:e.texture,type:e.type,dims:r});case"gpu-buffer":return new es({location:"gpu-buffer",gpuBuffer:e.gpuBuffer,type:e.type,dims:r});case"ml-tensor":return new es({location:"ml-tensor",mlTensor:e.mlTensor,type:e.type,dims:r});default:throw new Error(`tensorReshape: tensor location ${e.location} is not supported`)}}}),es,Ci=Ue(()=>{Mx(),bx(),yx(),vx(),es=class{constructor(e,r,t){id();let s,o;if(typeof e=="object"&&"location"in e)switch(this.dataLocation=e.location,s=e.type,o=e.dims,e.location){case"cpu-pinned":{let a=hn.get(s);if(!a)throw new TypeError(`unsupported type "${s}" to create tensor from pinned buffer`);if(!(e.data instanceof a))throw new TypeError(`buffer should be of type ${a.name}`);this.cpuData=e.data;break}case"texture":{if(s!=="float32")throw new TypeError(`unsupported type "${s}" to create tensor from texture`);this.gpuTextureData=e.texture,this.downloader=e.download,this.disposer=e.dispose;break}case"gpu-buffer":{if(s!=="float32"&&s!=="float16"&&s!=="int32"&&s!=="int64"&&s!=="uint32"&&s!=="uint8"&&s!=="bool"&&s!=="uint4"&&s!=="int4")throw new TypeError(`unsupported type "${s}" to create tensor from gpu buffer`);this.gpuBufferData=e.gpuBuffer,this.downloader=e.download,this.disposer=e.dispose;break}case"ml-tensor":{if(s!=="float32"&&s!=="float16"&&s!=="int32"&&s!=="int64"&&s!=="uint32"&&s!=="uint64"&&s!=="int8"&&s!=="uint8"&&s!=="bool"&&s!=="uint4"&&s!=="int4")throw new TypeError(`unsupported type "${s}" to create tensor from MLTensor`);this.mlTensorData=e.mlTensor,this.downloader=e.download,this.disposer=e.dispose;break}default:throw new Error(`Tensor constructor: unsupported location '${this.dataLocation}'`)}else{let a,i;if(typeof e=="string")if(s=e,i=t,e==="string"){if(!Array.isArray(r))throw new TypeError("A string tensor's data must be a string array.");a=r}else{let l=hn.get(e);if(l===void 0)throw new TypeError(`Unsupported tensor type: ${e}.`);if(Array.isArray(r)){if(e==="float16"&&l===Uint16Array||e==="uint4"||e==="int4")throw new TypeError(`Creating a ${e} tensor from number array is not supported. Please use ${l.name} as data.`);e==="uint64"||e==="int64"?a=l.from(r,BigInt):a=l.from(r)}else if(r instanceof l)a=r;else if(r instanceof Uint8ClampedArray)if(e==="uint8")a=Uint8Array.from(r);else throw new TypeError("A Uint8ClampedArray tensor's data must be type of uint8");else if(e==="float16"&&r instanceof Uint16Array&&l!==Uint16Array)a=new globalThis.Float16Array(r.buffer,r.byteOffset,r.length);else throw new TypeError(`A ${s} tensor's data must be type of ${l}`)}else if(i=r,Array.isArray(e)){if(e.length===0)throw new TypeError("Tensor type cannot be inferred from an empty array.");let l=typeof e[0];if(l==="string")s="string",a=e;else if(l==="boolean")s="bool",a=Uint8Array.from(e);else throw new TypeError(`Invalid element type of data array: ${l}.`)}else if(e instanceof Uint8ClampedArray)s="uint8",a=Uint8Array.from(e);else{let l=mo.get(e.constructor);if(l===void 0)throw new TypeError(`Unsupported type for tensor data: ${e.constructor}.`);s=l,a=e}if(i===void 0)i=[a.length];else if(!Array.isArray(i))throw new TypeError("A tensor's dims must be a number array");o=i,this.cpuData=a,this.dataLocation="cpu"}let n=ld(o);if(this.cpuData&&n!==this.cpuData.length&&!((s==="uint4"||s==="int4")&&Math.ceil(n/2)===this.cpuData.length))throw new Error(`Tensor's size(${n}) does not match data length(${this.cpuData.length}).`);this.type=s,this.dims=o,this.size=n}static async fromImage(e,r){return rd(e,r)}static fromTexture(e,r){return sd(e,r)}static fromGpuBuffer(e,r){return nd(e,r)}static fromMLTensor(e,r){return od(e,r)}static fromPinnedBuffer(e,r,t){return ad(e,r,t)}toDataURL(e){return ed(this,e)}toImageData(e){return td(this,e)}get data(){if(this.ensureValid(),!this.cpuData)throw new Error("The data is not on CPU. Use `getData()` to download GPU data to CPU, or use `texture` or `gpuBuffer` property to access the GPU data directly.");return this.cpuData}get location(){return this.dataLocation}get texture(){if(this.ensureValid(),!this.gpuTextureData)throw new Error("The data is not stored as a WebGL texture.");return this.gpuTextureData}get gpuBuffer(){if(this.ensureValid(),!this.gpuBufferData)throw new Error("The data is not stored as a WebGPU buffer.");return this.gpuBufferData}get mlTensor(){if(this.ensureValid(),!this.mlTensorData)throw new Error("The data is not stored as a WebNN MLTensor.");return this.mlTensorData}async getData(e){switch(this.ensureValid(),this.dataLocation){case"cpu":case"cpu-pinned":return this.data;case"texture":case"gpu-buffer":case"ml-tensor":{if(!this.downloader)throw new Error("The current tensor is not created with a specified data downloader.");if(this.isDownloading)throw new Error("The current tensor is being downloaded.");try{this.isDownloading=!0;let r=await this.downloader();return this.downloader=void 0,this.dataLocation="cpu",this.cpuData=r,e&&this.disposer&&(this.disposer(),this.disposer=void 0),r}finally{this.isDownloading=!1}}default:throw new Error(`cannot get data from location: ${this.dataLocation}`)}}dispose(){if(this.isDownloading)throw new Error("The current tensor is being downloaded.");this.disposer&&(this.disposer(),this.disposer=void 0),this.cpuData=void 0,this.gpuTextureData=void 0,this.gpuBufferData=void 0,this.mlTensorData=void 0,this.downloader=void 0,this.isDownloading=void 0,this.dataLocation="none"}ensureValid(){if(this.dataLocation==="none")throw new Error("The tensor is disposed.")}reshape(e){if(this.ensureValid(),this.downloader||this.disposer)throw new Error("Cannot reshape a tensor that owns GPU resource.");return cd(this,e)}}}),hs,ud=Ue(()=>{Ci(),hs=es}),ho,Si,_s,is,dd=Ue(()=>{Zu(),ho=(e,r)=>{(typeof as.trace>"u"?!as.wasm.trace:!as.trace)||console.timeStamp(`${e}::ORT::${r}`)},Si=(e,r)=>{var o;let t=((o=new Error().stack)==null?void 0:o.split(/\r\n|\r|\n/g))||[],s=!1;for(let n=0;n<t.length;n++){if(s&&!t[n].includes("TRACE_FUNC")){let a=`FUNC_${e}::${t[n].trim().split(" ")[1]}`;r&&(a+=`::${r}`),ho("CPU",a);return}t[n].includes("TRACE_FUNC")&&(s=!0)}},_s=e=>{(typeof as.trace>"u"?!as.wasm.trace:!as.trace)||Si("BEGIN",e)},is=e=>{(typeof as.trace>"u"?!as.wasm.trace:!as.trace)||Si("END",e)}}),pd,xx=Ue(()=>{Ju(),ud(),dd(),pd=class qv{constructor(r){this.handler=r}async run(r,t,s){_s();let o={},n={};if(typeof r!="object"||r===null||r instanceof hs||Array.isArray(r))throw new TypeError("'feeds' must be an object that use input names as keys and OnnxValue as corresponding values.");let a=!0;if(typeof t=="object"){if(t===null)throw new TypeError("Unexpected argument[1]: cannot be null.");if(t instanceof hs)throw new TypeError("'fetches' cannot be a Tensor");if(Array.isArray(t)){if(t.length===0)throw new TypeError("'fetches' cannot be an empty array.");a=!1;for(let c of t){if(typeof c!="string")throw new TypeError("'fetches' must be a string array or an object.");if(this.outputNames.indexOf(c)===-1)throw new RangeError(`'fetches' contains invalid output name: ${c}.`);o[c]=null}if(typeof s=="object"&&s!==null)n=s;else if(typeof s<"u")throw new TypeError("'options' must be an object.")}else{let c=!1,p=Object.getOwnPropertyNames(t);for(let d of this.outputNames)if(p.indexOf(d)!==-1){let u=t[d];(u===null||u instanceof hs)&&(c=!0,a=!1,o[d]=u)}if(c){if(typeof s=="object"&&s!==null)n=s;else if(typeof s<"u")throw new TypeError("'options' must be an object.")}else n=t}}else if(typeof t<"u")throw new TypeError("Unexpected argument[1]: must be 'fetches' or 'options'.");for(let c of this.inputNames)if(typeof r[c]>"u")throw new Error(`input '${c}' is missing in 'feeds'.`);if(a)for(let c of this.outputNames)o[c]=null;let i=await this.handler.run(r,o,n),l={};for(let c in i)if(Object.hasOwnProperty.call(i,c)){let p=i[c];p instanceof hs?l[c]=p:l[c]=new hs(p.type,p.data,p.dims)}return is(),l}async release(){return this.handler.dispose()}static async create(r,t,s,o){_s();let n,a={};if(typeof r=="string"){if(n=r,typeof t=="object"&&t!==null)a=t;else if(typeof t<"u")throw new TypeError("'options' must be an object.")}else if(r instanceof Uint8Array){if(n=r,typeof t=="object"&&t!==null)a=t;else if(typeof t<"u")throw new TypeError("'options' must be an object.")}else if(r instanceof ArrayBuffer||typeof SharedArrayBuffer<"u"&&r instanceof SharedArrayBuffer){let p=r,d=0,u=r.byteLength;if(typeof t=="object"&&t!==null)a=t;else if(typeof t=="number"){if(d=t,!Number.isSafeInteger(d))throw new RangeError("'byteOffset' must be an integer.");if(d<0||d>=p.byteLength)throw new RangeError(`'byteOffset' is out of range [0, ${p.byteLength}).`);if(u=r.byteLength-d,typeof s=="number"){if(u=s,!Number.isSafeInteger(u))throw new RangeError("'byteLength' must be an integer.");if(u<=0||d+u>p.byteLength)throw new RangeError(`'byteLength' is out of range (0, ${p.byteLength-d}].`);if(typeof o=="object"&&o!==null)a=o;else if(typeof o<"u")throw new TypeError("'options' must be an object.")}else if(typeof s<"u")throw new TypeError("'byteLength' must be a number.")}else if(typeof t<"u")throw new TypeError("'options' must be an object.");n=new Uint8Array(p,d,u)}else throw new TypeError("Unexpected argument[0]: must be 'path' or 'buffer'.");let[i,l]=await Xu(a),c=await i.createInferenceSessionHandler(n,l);return is(),new qv(c)}startProfiling(){this.handler.startProfiling()}endProfiling(){this.handler.endProfiling()}get inputNames(){return this.handler.inputNames}get outputNames(){return this.handler.outputNames}get inputMetadata(){return this.handler.inputMetadata}get outputMetadata(){return this.handler.outputMetadata}}}),$i,Tx=Ue(()=>{xx(),$i=pd}),Ex=Ue(()=>{}),Px=Ue(()=>{}),Cx=Ue(()=>{}),Sx=Ue(()=>{}),md={};Un(md,{InferenceSession:()=>$i,TRACE:()=>ho,TRACE_FUNC_BEGIN:()=>_s,TRACE_FUNC_END:()=>is,Tensor:()=>hs,env:()=>tr,registerBackend:()=>mn});var fs=Ue(()=>{fx(),wx(),Tx(),ud(),Ex(),Px(),dd(),Cx(),Sx()}),ki=Ue(()=>{}),hd={};Un(hd,{default:()=>_d});var Ii,Ai,_d,$x=Ue(()=>{var e;Qg(),_n(),ji(),Ii="ort-wasm-proxy-worker",Ai=((e=globalThis.self)==null?void 0:e.name)===Ii,Ai&&(self.onmessage=r=>{let{type:t,in:s}=r.data;try{switch(t){case"init-wasm":Ui(s.wasm).then(()=>{rc(s).then(()=>{postMessage({type:t})},o=>{postMessage({type:t,err:o})})},o=>{postMessage({type:t,err:o})});break;case"init-ep":{let{epName:o,env:n}=s;sc(n,o).then(()=>{postMessage({type:t})},a=>{postMessage({type:t,err:a})});break}case"copy-from":{let{buffer:o}=s,n=ba(o);postMessage({type:t,out:n});break}case"create":{let{model:o,options:n}=s;oc(o,n).then(a=>{postMessage({type:t,out:a})},a=>{postMessage({type:t,err:a})});break}case"release":ac(s),postMessage({type:t});break;case"run":{let{sessionId:o,inputIndices:n,inputs:a,outputIndices:i,options:l}=s;lc(o,n,a,i,new Array(i.length).fill(null),l).then(c=>{c.some(p=>p[3]!=="cpu")?postMessage({type:t,err:"Proxy does not support non-cpu tensor location."}):postMessage({type:t,out:c},uc([...a,...c]))},c=>{postMessage({type:t,err:c})});break}case"end-profiling":cc(s),postMessage({type:t});break;default:}}catch(o){postMessage({type:t,err:o})}}),_d=Ai?null:r=>new Worker(r??ts,{type:"module",name:Ii})}),fd={};Un(fd,{default:()=>gd});var Fi,Oi,gd,wd,kx=Ue(()=>{var e,r;Oi=(Fi=self.location.href,async function(t={}){var ao;var s,o,n=t,a=new Promise((h,E)=>{s=h,o=E}),i=typeof window=="object",l=typeof WorkerGlobalScope<"u",c=l&&((ao=self.name)==null?void 0:ao.startsWith("em-pthread"));n.mountExternalData=(h,E)=>{h.startsWith("./")&&(h=h.substring(2)),(n.Eb||(n.Eb=new Map)).set(h,E)},n.unmountExternalData=()=>{delete n.Eb};var p=globalThis.SharedArrayBuffer??new WebAssembly.Memory({initial:0,maximum:0,pc:!0}).buffer.constructor;let d=h=>async(...E)=>{var A;try{if(n.Fb)throw Error("Session already started");let z=n.Fb={dc:E[0],errors:[]},V=await h(...E);if(n.Fb!==z)throw Error("Session mismatch");(A=n.Jb)==null||A.flush();let ce=z.errors;if(0<ce.length){let Pe=await Promise.all(ce);if(Pe=Pe.filter(Re=>Re),0<Pe.length)throw Error(Pe.join(`
`))}return V}finally{n.Fb=null}};n.jsepInit=(h,E)=>{if(h==="webgpu"){[n.Jb,n.Ub,n.Yb,n.Kb,n.Xb,n.jb,n.Zb,n.ac,n.Vb,n.Wb,n.$b]=E;let A=n.Jb;n.jsepRegisterBuffer=(z,V,ce,Pe)=>A.registerBuffer(z,V,ce,Pe),n.jsepGetBuffer=z=>A.getBuffer(z),n.jsepCreateDownloader=(z,V,ce)=>A.createDownloader(z,V,ce),n.jsepOnCreateSession=z=>{A.onCreateSession(z)},n.jsepOnReleaseSession=z=>{A.onReleaseSession(z)},n.jsepOnRunStart=z=>A.onRunStart(z),n.bc=(z,V)=>{A.upload(z,V)}}else if(h==="webnn"){let A=E[0];[n.nc,n.Nb,n.webnnEnsureTensor,n.Ob,n.webnnDownloadTensor]=E.slice(1),n.webnnReleaseTensorId=n.Nb,n.webnnUploadTensor=n.Ob,n.webnnOnRunStart=z=>A.onRunStart(z),n.webnnOnRunEnd=A.onRunEnd.bind(A),n.webnnRegisterMLContext=(z,V)=>{A.registerMLContext(z,V)},n.webnnOnReleaseSession=z=>{A.onReleaseSession(z)},n.webnnCreateMLTensorDownloader=(z,V)=>A.createMLTensorDownloader(z,V),n.webnnRegisterMLTensor=(z,V,ce,Pe)=>A.registerMLTensor(z,V,ce,Pe),n.webnnCreateMLContext=z=>A.createMLContext(z),n.webnnRegisterMLConstant=(z,V,ce,Pe,Re,Je)=>A.registerMLConstant(z,V,ce,Pe,Re,n.Eb,Je),n.webnnRegisterGraphInput=A.registerGraphInput.bind(A),n.webnnIsGraphInput=A.isGraphInput.bind(A),n.webnnCreateTemporaryTensor=A.createTemporaryTensor.bind(A),n.webnnIsInt64Supported=A.isInt64Supported.bind(A)}};let u=()=>{let h=(E,A,z)=>(...V)=>{let ce=xr,Pe=A==null?void 0:A();V=E(...V);let Re=A==null?void 0:A();return Pe!==Re&&(E=Re,z(Pe),A=z=null),xr!=ce?new Promise((Je,it)=>{Yr={resolve:Je,reject:it}}):V};(()=>{for(let E of["_OrtAppendExecutionProvider","_OrtCreateSession","_OrtRun","_OrtRunWithBinding","_OrtBindInput"])n[E]=h(n[E],()=>n[E],A=>n[E]=A)})(),d!==void 0&&(n._OrtRun=d(n._OrtRun),n._OrtRunWithBinding=d(n._OrtRunWithBinding)),u=void 0};n.asyncInit=()=>{u==null||u()};var f,_,b=Object.assign({},n),k=(h,E)=>{throw E},w="";(i||l)&&(l?w=self.location.href:typeof document<"u"&&document.currentScript&&(w=document.currentScript.src),Fi&&(w=Fi),w=w.startsWith("blob:")?"":w.slice(0,w.replace(/[?#].*/,"").lastIndexOf("/")+1),l&&(_=h=>{var E=new XMLHttpRequest;return E.open("GET",h,!1),E.responseType="arraybuffer",E.send(null),new Uint8Array(E.response)}),f=async h=>{if(oe(h))return new Promise((A,z)=>{var V=new XMLHttpRequest;V.open("GET",h,!0),V.responseType="arraybuffer",V.onload=()=>{V.status==200||V.status==0&&V.response?A(V.response):z(V.status)},V.onerror=z,V.send(null)});var E=await fetch(h,{credentials:"same-origin"});if(E.ok)return E.arrayBuffer();throw Error(E.status+" : "+E.url)});var y=console.log.bind(console),$=console.error.bind(console),T=y,M=$;Object.assign(n,b),b=null;var x,v,C,F,D,K,U,j,se,q,re,Z,ae,_e=n.wasmBinary,pe=!1,oe=h=>h.startsWith("file://");function G(){return x.buffer!=F.buffer&&O(),F}function B(){return x.buffer!=F.buffer&&O(),D}function Q(){return x.buffer!=F.buffer&&O(),K}function le(){return x.buffer!=F.buffer&&O(),U}function ee(){return x.buffer!=F.buffer&&O(),j}function Se(){return x.buffer!=F.buffer&&O(),se}function De(){return x.buffer!=F.buffer&&O(),q}function Ne(){return x.buffer!=F.buffer&&O(),ae}if(c){let h=function(E){try{var A=E.data,z=A.Bb;if(z==="load"){let V=[];self.onmessage=ce=>V.push(ce),self.startWorker=()=>{postMessage({Bb:"loaded"});for(let ce of V)h(ce);self.onmessage=h};for(let ce of A.Rb)n[ce]&&!n[ce].proxy||(n[ce]=(...Pe)=>{postMessage({Bb:"callHandler",Qb:ce,args:Pe})},ce=="print"&&(T=n[ce]),ce=="printErr"&&(M=n[ce]));x=A.kc,O(),Ae(A.lc)}else if(z==="run"){nn(A.Ab),to(A.Ab,0,0,1,0,0),Ur(),de(A.Ab),te||(Vo(),te=!0);try{cr(A.fc,A.Hb)}catch(V){if(V!="unwind")throw V}}else A.target!=="setimmediate"&&(z==="checkMailbox"?te&&Te():z&&(M(`worker: received unknown command ${z}`),M(A)))}catch(V){throw ro(),V}};var Ae,te=!1;M=function(...E){E=E.join(" "),console.error(E)},self.alert=function(...E){postMessage({Bb:"alert",text:E.join(" "),ic:In()})},self.onunhandledrejection=E=>{throw E.reason||E},self.onmessage=h}function O(){var h=x.buffer;n.HEAP8=F=new Int8Array(h),n.HEAP16=K=new Int16Array(h),n.HEAPU8=D=new Uint8Array(h),n.HEAPU16=U=new Uint16Array(h),n.HEAP32=j=new Int32Array(h),n.HEAPU32=se=new Uint32Array(h),n.HEAPF32=q=new Float32Array(h),n.HEAPF64=ae=new Float64Array(h),n.HEAP64=re=new BigInt64Array(h),n.HEAPU64=Z=new BigUint64Array(h)}function me(){c?startWorker(n):mt.Ca()}c||(x=new WebAssembly.Memory({initial:256,maximum:65536,shared:!0}),O());var xe,be=0,Ce=null;function $e(){if(--be==0&&Ce){var h=Ce;Ce=null,h()}}function Me(h){throw M(h="Aborted("+h+")"),pe=!0,h=new WebAssembly.RuntimeError(h+". Build with -sASSERTIONS for more info."),o(h),h}function Le(){return{a:{L:He,Aa:ke,b:ls,$:gt,A:Jr,pa:$t,X:H,Z:R,qa:W,na:X,ga:ue,ma:Ee,J:et,Y:ot,V:rt,oa:St,W:lt,va:Dt,E:Os,Q:Hs,O:Ls,D:us,u:Wr,r:je,P:Ve,z:I,R:N,ja:Y,T:Ge,aa:qe,M:tt,F:bt,ia:de,sa:jt,t:Jt,Ba:yr,w:ds,o:_r,l:qr,c:Hr,n:Rs,j:Ea,v:Pa,p:on,f:Ca,s:Sa,m:$a,e:ka,k:Ia,i:an,g:Aa,d:Fa,da:Oa,ea:Da,fa:La,ba:Jn,ca:Ao,N:Yn,xa:gc,ua:Fo,h:Ra,C:ja,G:Na,ta:Ba,x:Va,ra:Ua,U:Wa,q:za,y:Oo,K:Ga,S:Ka,za:Do,ya:Ha,ka:qa,la:zo,_:ft,B:Qa,I:Bo,ha:Zn,H:Ro,a:x,wa:ze}}}var we={829644:(h,E,A,z,V)=>{if(n===void 0||!n.Eb)return 1;if((h=Yt(Number(h>>>0))).startsWith("./")&&(h=h.substring(2)),!(h=n.Eb.get(h)))return 2;if(E=Number(E>>>0),A=Number(A>>>0),z=Number(z>>>0),E+A>h.byteLength)return 3;try{let ce=h.subarray(E,E+A);switch(V){case 0:B().set(ce,z>>>0);break;case 1:n.mc?n.mc(z,ce):n.bc(z,ce);break;default:return 4}return 0}catch{return 4}},830468:(h,E,A)=>{n.Ob(h,B().subarray(E>>>0,E+A>>>0))},830532:()=>n.nc(),830574:h=>{n.Nb(h)},830611:()=>{n.Vb()},830642:()=>{n.Wb()},830671:()=>{n.$b()},830696:h=>n.Ub(h),830729:h=>n.Yb(h),830761:(h,E,A)=>{n.Kb(Number(h),Number(E),Number(A),!0)},830824:(h,E,A)=>{n.Kb(Number(h),Number(E),Number(A))},830881:()=>typeof wasmOffsetConverter<"u",830938:h=>{n.jb("Abs",h,void 0)},830989:h=>{n.jb("Neg",h,void 0)},831040:h=>{n.jb("Floor",h,void 0)},831093:h=>{n.jb("Ceil",h,void 0)},831145:h=>{n.jb("Reciprocal",h,void 0)},831203:h=>{n.jb("Sqrt",h,void 0)},831255:h=>{n.jb("Exp",h,void 0)},831306:h=>{n.jb("Erf",h,void 0)},831357:h=>{n.jb("Sigmoid",h,void 0)},831412:(h,E,A)=>{n.jb("HardSigmoid",h,{alpha:E,beta:A})},831491:h=>{n.jb("Log",h,void 0)},831542:h=>{n.jb("Sin",h,void 0)},831593:h=>{n.jb("Cos",h,void 0)},831644:h=>{n.jb("Tan",h,void 0)},831695:h=>{n.jb("Asin",h,void 0)},831747:h=>{n.jb("Acos",h,void 0)},831799:h=>{n.jb("Atan",h,void 0)},831851:h=>{n.jb("Sinh",h,void 0)},831903:h=>{n.jb("Cosh",h,void 0)},831955:h=>{n.jb("Asinh",h,void 0)},832008:h=>{n.jb("Acosh",h,void 0)},832061:h=>{n.jb("Atanh",h,void 0)},832114:h=>{n.jb("Tanh",h,void 0)},832166:h=>{n.jb("Not",h,void 0)},832217:(h,E,A)=>{n.jb("Clip",h,{min:E,max:A})},832286:h=>{n.jb("Clip",h,void 0)},832338:(h,E)=>{n.jb("Elu",h,{alpha:E})},832396:h=>{n.jb("Gelu",h,void 0)},832448:h=>{n.jb("Relu",h,void 0)},832500:(h,E)=>{n.jb("LeakyRelu",h,{alpha:E})},832564:(h,E)=>{n.jb("ThresholdedRelu",h,{alpha:E})},832634:(h,E)=>{n.jb("Cast",h,{to:E})},832692:h=>{n.jb("Add",h,void 0)},832743:h=>{n.jb("Sub",h,void 0)},832794:h=>{n.jb("Mul",h,void 0)},832845:h=>{n.jb("Div",h,void 0)},832896:h=>{n.jb("Pow",h,void 0)},832947:h=>{n.jb("Equal",h,void 0)},833e3:h=>{n.jb("Greater",h,void 0)},833055:h=>{n.jb("GreaterOrEqual",h,void 0)},833117:h=>{n.jb("Less",h,void 0)},833169:h=>{n.jb("LessOrEqual",h,void 0)},833228:(h,E,A,z,V)=>{n.jb("ReduceMean",h,{keepDims:!!E,noopWithEmptyAxes:!!A,axes:z?Array.from(ee().subarray(Number(z)>>>0,Number(V)>>>0)):[]})},833403:(h,E,A,z,V)=>{n.jb("ReduceMax",h,{keepDims:!!E,noopWithEmptyAxes:!!A,axes:z?Array.from(ee().subarray(Number(z)>>>0,Number(V)>>>0)):[]})},833577:(h,E,A,z,V)=>{n.jb("ReduceMin",h,{keepDims:!!E,noopWithEmptyAxes:!!A,axes:z?Array.from(ee().subarray(Number(z)>>>0,Number(V)>>>0)):[]})},833751:(h,E,A,z,V)=>{n.jb("ReduceProd",h,{keepDims:!!E,noopWithEmptyAxes:!!A,axes:z?Array.from(ee().subarray(Number(z)>>>0,Number(V)>>>0)):[]})},833926:(h,E,A,z,V)=>{n.jb("ReduceSum",h,{keepDims:!!E,noopWithEmptyAxes:!!A,axes:z?Array.from(ee().subarray(Number(z)>>>0,Number(V)>>>0)):[]})},834100:(h,E,A,z,V)=>{n.jb("ReduceL1",h,{keepDims:!!E,noopWithEmptyAxes:!!A,axes:z?Array.from(ee().subarray(Number(z)>>>0,Number(V)>>>0)):[]})},834273:(h,E,A,z,V)=>{n.jb("ReduceL2",h,{keepDims:!!E,noopWithEmptyAxes:!!A,axes:z?Array.from(ee().subarray(Number(z)>>>0,Number(V)>>>0)):[]})},834446:(h,E,A,z,V)=>{n.jb("ReduceLogSum",h,{keepDims:!!E,noopWithEmptyAxes:!!A,axes:z?Array.from(ee().subarray(Number(z)>>>0,Number(V)>>>0)):[]})},834623:(h,E,A,z,V)=>{n.jb("ReduceSumSquare",h,{keepDims:!!E,noopWithEmptyAxes:!!A,axes:z?Array.from(ee().subarray(Number(z)>>>0,Number(V)>>>0)):[]})},834803:(h,E,A,z,V)=>{n.jb("ReduceLogSumExp",h,{keepDims:!!E,noopWithEmptyAxes:!!A,axes:z?Array.from(ee().subarray(Number(z)>>>0,Number(V)>>>0)):[]})},834983:h=>{n.jb("Where",h,void 0)},835036:(h,E,A)=>{n.jb("Transpose",h,{perm:E?Array.from(ee().subarray(Number(E)>>>0,Number(A)>>>0)):[]})},835160:(h,E,A,z)=>{n.jb("DepthToSpace",h,{blocksize:E,mode:Yt(A),format:z?"NHWC":"NCHW"})},835293:(h,E,A,z)=>{n.jb("DepthToSpace",h,{blocksize:E,mode:Yt(A),format:z?"NHWC":"NCHW"})},835426:(h,E,A,z,V,ce,Pe,Re,Je,it,xt,Rt,Qt,Er,Ys)=>{n.jb("ConvTranspose",h,{format:Je?"NHWC":"NCHW",autoPad:E,dilations:[A],group:z,kernelShape:[V],pads:[ce,Pe],strides:[Re],wIsConst:()=>!!G()[it>>>0],outputPadding:xt?Array.from(ee().subarray(Number(xt)>>>0,Number(Rt)>>>0)):[],outputShape:Qt?Array.from(ee().subarray(Number(Qt)>>>0,Number(Er)>>>0)):[],activation:Yt(Ys)})},835859:(h,E,A,z,V,ce,Pe,Re,Je,it,xt,Rt,Qt,Er)=>{n.jb("ConvTranspose",h,{format:Re?"NHWC":"NCHW",autoPad:E,dilations:Array.from(ee().subarray(Number(A)>>>0,2+(Number(A)>>>0)>>>0)),group:z,kernelShape:Array.from(ee().subarray(Number(V)>>>0,2+(Number(V)>>>0)>>>0)),pads:Array.from(ee().subarray(Number(ce)>>>0,4+(Number(ce)>>>0)>>>0)),strides:Array.from(ee().subarray(Number(Pe)>>>0,2+(Number(Pe)>>>0)>>>0)),wIsConst:()=>!!G()[Je>>>0],outputPadding:it?Array.from(ee().subarray(Number(it)>>>0,Number(xt)>>>0)):[],outputShape:Rt?Array.from(ee().subarray(Number(Rt)>>>0,Number(Qt)>>>0)):[],activation:Yt(Er)})},836520:(h,E,A,z,V,ce,Pe,Re,Je,it,xt,Rt,Qt,Er,Ys)=>{n.jb("ConvTranspose",h,{format:Je?"NHWC":"NCHW",autoPad:E,dilations:[A],group:z,kernelShape:[V],pads:[ce,Pe],strides:[Re],wIsConst:()=>!!G()[it>>>0],outputPadding:xt?Array.from(ee().subarray(Number(xt)>>>0,Number(Rt)>>>0)):[],outputShape:Qt?Array.from(ee().subarray(Number(Qt)>>>0,Number(Er)>>>0)):[],activation:Yt(Ys)})},836953:(h,E,A,z,V,ce,Pe,Re,Je,it,xt,Rt,Qt,Er)=>{n.jb("ConvTranspose",h,{format:Re?"NHWC":"NCHW",autoPad:E,dilations:Array.from(ee().subarray(Number(A)>>>0,2+(Number(A)>>>0)>>>0)),group:z,kernelShape:Array.from(ee().subarray(Number(V)>>>0,2+(Number(V)>>>0)>>>0)),pads:Array.from(ee().subarray(Number(ce)>>>0,4+(Number(ce)>>>0)>>>0)),strides:Array.from(ee().subarray(Number(Pe)>>>0,2+(Number(Pe)>>>0)>>>0)),wIsConst:()=>!!G()[Je>>>0],outputPadding:it?Array.from(ee().subarray(Number(it)>>>0,Number(xt)>>>0)):[],outputShape:Rt?Array.from(ee().subarray(Number(Rt)>>>0,Number(Qt)>>>0)):[],activation:Yt(Er)})},837614:(h,E)=>{n.jb("GlobalAveragePool",h,{format:E?"NHWC":"NCHW"})},837705:(h,E,A,z,V,ce,Pe,Re,Je,it,xt,Rt,Qt,Er)=>{n.jb("AveragePool",h,{format:Er?"NHWC":"NCHW",auto_pad:E,ceil_mode:A,count_include_pad:z,storage_order:V,dilations:ce?Array.from(ee().subarray(Number(ce)>>>0,Number(Pe)>>>0)):[],kernel_shape:Re?Array.from(ee().subarray(Number(Re)>>>0,Number(Je)>>>0)):[],pads:it?Array.from(ee().subarray(Number(it)>>>0,Number(xt)>>>0)):[],strides:Rt?Array.from(ee().subarray(Number(Rt)>>>0,Number(Qt)>>>0)):[]})},838184:(h,E)=>{n.jb("GlobalAveragePool",h,{format:E?"NHWC":"NCHW"})},838275:(h,E,A,z,V,ce,Pe,Re,Je,it,xt,Rt,Qt,Er)=>{n.jb("AveragePool",h,{format:Er?"NHWC":"NCHW",auto_pad:E,ceil_mode:A,count_include_pad:z,storage_order:V,dilations:ce?Array.from(ee().subarray(Number(ce)>>>0,Number(Pe)>>>0)):[],kernel_shape:Re?Array.from(ee().subarray(Number(Re)>>>0,Number(Je)>>>0)):[],pads:it?Array.from(ee().subarray(Number(it)>>>0,Number(xt)>>>0)):[],strides:Rt?Array.from(ee().subarray(Number(Rt)>>>0,Number(Qt)>>>0)):[]})},838754:(h,E)=>{n.jb("GlobalMaxPool",h,{format:E?"NHWC":"NCHW"})},838841:(h,E,A,z,V,ce,Pe,Re,Je,it,xt,Rt,Qt,Er)=>{n.jb("MaxPool",h,{format:Er?"NHWC":"NCHW",auto_pad:E,ceil_mode:A,count_include_pad:z,storage_order:V,dilations:ce?Array.from(ee().subarray(Number(ce)>>>0,Number(Pe)>>>0)):[],kernel_shape:Re?Array.from(ee().subarray(Number(Re)>>>0,Number(Je)>>>0)):[],pads:it?Array.from(ee().subarray(Number(it)>>>0,Number(xt)>>>0)):[],strides:Rt?Array.from(ee().subarray(Number(Rt)>>>0,Number(Qt)>>>0)):[]})},839316:(h,E)=>{n.jb("GlobalMaxPool",h,{format:E?"NHWC":"NCHW"})},839403:(h,E,A,z,V,ce,Pe,Re,Je,it,xt,Rt,Qt,Er)=>{n.jb("MaxPool",h,{format:Er?"NHWC":"NCHW",auto_pad:E,ceil_mode:A,count_include_pad:z,storage_order:V,dilations:ce?Array.from(ee().subarray(Number(ce)>>>0,Number(Pe)>>>0)):[],kernel_shape:Re?Array.from(ee().subarray(Number(Re)>>>0,Number(Je)>>>0)):[],pads:it?Array.from(ee().subarray(Number(it)>>>0,Number(xt)>>>0)):[],strides:Rt?Array.from(ee().subarray(Number(Rt)>>>0,Number(Qt)>>>0)):[]})},839878:(h,E,A,z,V)=>{n.jb("Gemm",h,{alpha:E,beta:A,transA:z,transB:V})},839982:h=>{n.jb("MatMul",h,void 0)},840036:(h,E,A,z)=>{n.jb("ArgMax",h,{keepDims:!!E,selectLastIndex:!!A,axis:z})},840144:(h,E,A,z)=>{n.jb("ArgMin",h,{keepDims:!!E,selectLastIndex:!!A,axis:z})},840252:(h,E)=>{n.jb("Softmax",h,{axis:E})},840315:(h,E)=>{n.jb("Concat",h,{axis:E})},840375:(h,E,A,z,V)=>{n.jb("Split",h,{axis:E,numOutputs:A,splitSizes:z?Array.from(ee().subarray(Number(z)>>>0,Number(V)>>>0)):[]})},840531:h=>{n.jb("Expand",h,void 0)},840585:(h,E)=>{n.jb("Gather",h,{axis:Number(E)})},840656:(h,E)=>{n.jb("GatherElements",h,{axis:Number(E)})},840735:(h,E)=>{n.jb("GatherND",h,{batch_dims:Number(E)})},840814:(h,E,A,z,V,ce,Pe,Re,Je,it,xt)=>{n.jb("Resize",h,{antialias:E,axes:A?Array.from(ee().subarray(Number(A)>>>0,Number(z)>>>0)):[],coordinateTransformMode:Yt(V),cubicCoeffA:ce,excludeOutside:Pe,extrapolationValue:Re,keepAspectRatioPolicy:Yt(Je),mode:Yt(it),nearestMode:Yt(xt)})},841176:(h,E,A,z,V,ce,Pe)=>{n.jb("Slice",h,{starts:E?Array.from(ee().subarray(Number(E)>>>0,Number(A)>>>0)):[],ends:z?Array.from(ee().subarray(Number(z)>>>0,Number(V)>>>0)):[],axes:ce?Array.from(ee().subarray(Number(ce)>>>0,Number(Pe)>>>0)):[]})},841440:h=>{n.jb("Tile",h,void 0)},841492:(h,E,A)=>{n.jb("InstanceNormalization",h,{epsilon:E,format:A?"NHWC":"NCHW"})},841606:(h,E,A)=>{n.jb("InstanceNormalization",h,{epsilon:E,format:A?"NHWC":"NCHW"})},841720:h=>{n.jb("Range",h,void 0)},841773:(h,E)=>{n.jb("Einsum",h,{equation:Yt(E)})},841854:(h,E,A,z,V)=>{n.jb("Pad",h,{mode:E,value:A,pads:z?Array.from(ee().subarray(Number(z)>>>0,Number(V)>>>0)):[]})},841997:(h,E,A,z,V,ce)=>{n.jb("BatchNormalization",h,{epsilon:E,momentum:A,spatial:!!V,trainingMode:!!z,format:ce?"NHWC":"NCHW"})},842166:(h,E,A,z,V,ce)=>{n.jb("BatchNormalization",h,{epsilon:E,momentum:A,spatial:!!V,trainingMode:!!z,format:ce?"NHWC":"NCHW"})},842335:(h,E,A)=>{n.jb("CumSum",h,{exclusive:Number(E),reverse:Number(A)})},842432:(h,E,A)=>{n.jb("DequantizeLinear",h,{axis:E,blockSize:A})},842522:(h,E,A,z,V)=>{n.jb("GridSample",h,{align_corners:E,mode:Yt(A),padding_mode:Yt(z),format:V?"NHWC":"NCHW"})},842692:(h,E,A,z,V)=>{n.jb("GridSample",h,{align_corners:E,mode:Yt(A),padding_mode:Yt(z),format:V?"NHWC":"NCHW"})},842862:(h,E)=>{n.jb("ScatterND",h,{reduction:Yt(E)})},842947:(h,E,A,z,V,ce,Pe,Re,Je)=>{n.jb("Attention",h,{numHeads:E,isUnidirectional:A,maskFilterValue:z,scale:V,doRotary:ce,qkvHiddenSizes:Pe?Array.from(ee().subarray(Number(Re)>>>0,Number(Re)+Pe>>>0)):[],pastPresentShareBuffer:!!Je})},843219:h=>{n.jb("BiasAdd",h,void 0)},843274:h=>{n.jb("BiasSplitGelu",h,void 0)},843335:h=>{n.jb("FastGelu",h,void 0)},843391:(h,E,A,z,V,ce,Pe,Re,Je,it,xt,Rt,Qt,Er,Ys,Ja)=>{n.jb("Conv",h,{format:Rt?"NHWC":"NCHW",auto_pad:E,dilations:A?Array.from(ee().subarray(Number(A)>>>0,Number(z)>>>0)):[],group:V,kernel_shape:ce?Array.from(ee().subarray(Number(ce)>>>0,Number(Pe)>>>0)):[],pads:Re?Array.from(ee().subarray(Number(Re)>>>0,Number(Je)>>>0)):[],strides:it?Array.from(ee().subarray(Number(it)>>>0,Number(xt)>>>0)):[],w_is_const:()=>!!G()[Number(Qt)>>>0],activation:Yt(Er),activation_params:Ys?Array.from(De().subarray(Number(Ys)>>>0,Number(Ja)>>>0)):[]})},843975:h=>{n.jb("Gelu",h,void 0)},844027:(h,E,A,z,V,ce,Pe,Re,Je)=>{n.jb("GroupQueryAttention",h,{numHeads:E,kvNumHeads:A,scale:z,softcap:V,doRotary:ce,rotaryInterleaved:Pe,smoothSoftmax:Re,localWindowSize:Je})},844244:(h,E,A,z)=>{n.jb("LayerNormalization",h,{axis:E,epsilon:A,simplified:!!z})},844355:(h,E,A,z)=>{n.jb("LayerNormalization",h,{axis:E,epsilon:A,simplified:!!z})},844466:(h,E,A,z,V,ce)=>{n.jb("MatMulNBits",h,{k:E,n:A,accuracyLevel:z,bits:V,blockSize:ce})},844593:(h,E,A,z,V,ce)=>{n.jb("MultiHeadAttention",h,{numHeads:E,isUnidirectional:A,maskFilterValue:z,scale:V,doRotary:ce})},844752:(h,E)=>{n.jb("QuickGelu",h,{alpha:E})},844816:(h,E,A,z,V)=>{n.jb("RotaryEmbedding",h,{interleaved:!!E,numHeads:A,rotaryEmbeddingDim:z,scale:V})},844955:(h,E,A)=>{n.jb("SkipLayerNormalization",h,{epsilon:E,simplified:!!A})},845057:(h,E,A)=>{n.jb("SkipLayerNormalization",h,{epsilon:E,simplified:!!A})},845159:(h,E,A,z)=>{n.jb("GatherBlockQuantized",h,{gatherAxis:E,quantizeAxis:A,blockSize:z})},845280:h=>{n.Zb(h)},845314:(h,E)=>n.ac(Number(h),Number(E),n.Fb.dc,n.Fb.errors)};function ke(h,E,A){return Fr(async()=>{await n.Xb(Number(h),Number(E),Number(A))})}function He(){return typeof wasmOffsetConverter<"u"}class Fe{constructor(E){J(this,"name","ExitStatus");this.message=`Program terminated with exit(${E})`,this.status=E}}var Ze=h=>{h.terminate(),h.onmessage=()=>{}},We=[],yt=h=>{Lt.length==0&&(Gs(),Ws(Lt[0]));var E=Lt.pop();if(!E)return 6;kt.push(E),fr[h.Ab]=E,E.Ab=h.Ab;var A={Bb:"run",fc:h.ec,Hb:h.Hb,Ab:h.Ab};return E.postMessage(A,h.Mb),0},Qe=0,Oe=(h,E,...A)=>{for(var z=2*A.length,V=Dn(),ce=no(8*z),Pe=ce>>>3,Re=0;Re<A.length;Re++){var Je=A[Re];typeof Je=="bigint"?(re[Pe+2*Re]=1n,re[Pe+2*Re+1]=Je):(re[Pe+2*Re]=0n,Ne()[Pe+2*Re+1>>>0]=Je)}return h=Uo(h,0,z,ce,E),On(V),h};function ze(h){if(c)return Oe(0,1,h);if(C=h,!(0<Qe)){for(var E of kt)Ze(E);for(E of Lt)Ze(E);Lt=[],kt=[],fr={},pe=!0}k(0,new Fe(h))}function pt(h){if(c)return Oe(1,0,h);ft(h)}var ft=h=>{if(C=h,c)throw pt(h),"unwind";ze(h)},Lt=[],kt=[],Mr=[],fr={},Ut=h=>{var E=h.Ab;delete fr[E],Lt.push(h),kt.splice(kt.indexOf(h),1),h.Ab=0,Wo(E)};function Ur(){Mr.forEach(h=>h())}var Ws=h=>new Promise(E=>{h.onmessage=V=>{var ce=(V=V.data).Bb;if(V.Gb&&V.Gb!=In()){var Pe=fr[V.Gb];Pe?Pe.postMessage(V,V.Mb):M(`Internal error! Worker sent a message "${ce}" to target pthread ${V.Gb}, but that thread no longer exists!`)}else ce==="checkMailbox"?Te():ce==="spawnThread"?yt(V):ce==="cleanupThread"?Ut(fr[V.hc]):ce==="loaded"?(h.loaded=!0,E(h)):ce==="alert"?alert(`Thread ${V.ic}: ${V.text}`):V.target==="setimmediate"?h.postMessage(V):ce==="callHandler"?n[V.Qb](...V.args):ce&&M(`worker sent an unknown command ${ce}`)},h.onerror=V=>{throw M(`worker sent an error! ${V.filename}:${V.lineno}: ${V.message}`),V};var A,z=[];for(A of[])n.propertyIsEnumerable(A)&&z.push(A);h.postMessage({Bb:"load",Rb:z,kc:x,lc:v})});function Gs(){var h=new Worker((()=>{let E=URL;return self.location.href>"file:"&&self.location.href<"file;"?new E("ort.bundle.min.mjs",self.location.href):new URL(self.location.href)})(),{type:"module",workerData:"em-pthread",name:"em-pthread"});Lt.push(h)}var nn=h=>{O();var E=Se()[h+52>>>2>>>0];h=Se()[h+56>>>2>>>0],so(E,E-h),On(E)},cr=(h,E)=>{Qe=0,h=Ho(h,E),0<Qe?C=h:Fn(h)};class zr{constructor(E){this.Ib=E-24}}function ls(h,E,A){var z=new zr(h>>>=0);throw E>>>=0,A>>>=0,Se()[z.Ib+16>>>2>>>0]=0,Se()[z.Ib+4>>>2>>>0]=E,Se()[z.Ib+8>>>2>>>0]=A,h}function Ks(h,E,A,z){return c?Oe(2,1,h,E,A,z):gt(h,E,A,z)}function gt(h,E,A,z){if(h>>>=0,A>>>=0,z>>>=0,p===void 0)return 6;var V=[];return c&&V.length===0?Ks(h,E>>>=0,A,z):(h={ec:A,Ab:h,Hb:z,Mb:V},c?(h.Bb="spawnThread",postMessage(h,V),0):yt(h))}var Is=typeof TextDecoder<"u"?new TextDecoder:void 0,As=(h,E=0,A=NaN)=>{var z=(E>>>=0)+A;for(A=E;h[A]&&!(A>=z);)++A;if(16<A-E&&h.buffer&&Is)return Is.decode(h.buffer instanceof ArrayBuffer?h.subarray(E,A):h.slice(E,A));for(z="";E<A;){var V=h[E++];if(128&V){var ce=63&h[E++];if((224&V)==192)z+=String.fromCharCode((31&V)<<6|ce);else{var Pe=63&h[E++];65536>(V=(240&V)==224?(15&V)<<12|ce<<6|Pe:(7&V)<<18|ce<<12|Pe<<6|63&h[E++])?z+=String.fromCharCode(V):(V-=65536,z+=String.fromCharCode(55296|V>>10,56320|1023&V))}}else z+=String.fromCharCode(V)}return z},Yt=(h,E)=>(h>>>=0)?As(B(),h,E):"";function Jr(h,E,A){return c?Oe(3,1,h,E,A):0}function $t(h,E){if(c)return Oe(4,1,h,E)}var Fs=h=>{for(var E=0,A=0;A<h.length;++A){var z=h.charCodeAt(A);127>=z?E++:2047>=z?E+=2:55296<=z&&57343>=z?(E+=4,++A):E+=3}return E},S=(h,E,A)=>{var z=B();if(E>>>=0,0<A){var V=E;A=E+A-1;for(var ce=0;ce<h.length;++ce){var Pe=h.charCodeAt(ce);if(55296<=Pe&&57343>=Pe&&(Pe=65536+((1023&Pe)<<10)|1023&h.charCodeAt(++ce)),127>=Pe){if(E>=A)break;z[E++>>>0]=Pe}else{if(2047>=Pe){if(E+1>=A)break;z[E++>>>0]=192|Pe>>6}else{if(65535>=Pe){if(E+2>=A)break;z[E++>>>0]=224|Pe>>12}else{if(E+3>=A)break;z[E++>>>0]=240|Pe>>18,z[E++>>>0]=128|Pe>>12&63}z[E++>>>0]=128|Pe>>6&63}z[E++>>>0]=128|63&Pe}}z[E>>>0]=0,h=E-V}else h=0;return h};function H(h,E){if(c)return Oe(5,1,h,E)}function R(h,E,A){if(c)return Oe(6,1,h,E,A)}function W(h,E,A){return c?Oe(7,1,h,E,A):0}function X(h,E){if(c)return Oe(8,1,h,E)}function ue(h,E,A){if(c)return Oe(9,1,h,E,A)}function Ee(h,E,A,z){if(c)return Oe(10,1,h,E,A,z)}function et(h,E,A,z){if(c)return Oe(11,1,h,E,A,z)}function ot(h,E,A,z){if(c)return Oe(12,1,h,E,A,z)}function rt(h){if(c)return Oe(13,1,h)}function St(h,E){if(c)return Oe(14,1,h,E)}function lt(h,E,A){if(c)return Oe(15,1,h,E,A)}var Ot,at,Dt=()=>Me(""),Wt=h=>{for(var E="";B()[h>>>0];)E+=Ot[B()[h++>>>0]];return E},Br={},br={};function Rr(h,E,A={}){return function(z,V,ce={}){var Pe=V.name;if(!z)throw new at(`type "${Pe}" must have a positive integer typeid pointer`);if(br.hasOwnProperty(z)){if(ce.Sb)return;throw new at(`Cannot register type '${Pe}' twice`)}br[z]=V,Br.hasOwnProperty(z)&&(V=Br[z],delete Br[z],V.forEach(Re=>Re()))}(h,E,A)}var Ts=(h,E,A)=>{switch(E){case 1:return A?z=>G()[z>>>0]:z=>B()[z>>>0];case 2:return A?z=>Q()[z>>>1>>>0]:z=>le()[z>>>1>>>0];case 4:return A?z=>ee()[z>>>2>>>0]:z=>Se()[z>>>2>>>0];case 8:return A?z=>re[z>>>3]:z=>Z[z>>>3];default:throw new TypeError(`invalid integer width (${E}): ${h}`)}};function Os(h,E,A){A>>>=0,Rr(h>>>=0,{name:E=Wt(E>>>0),fromWireType:z=>z,toWireType:function(z,V){if(typeof V!="bigint"&&typeof V!="number")throw V=V===null?"null":(z=typeof V)=="object"||z==="array"||z==="function"?V.toString():""+V,new TypeError(`Cannot convert "${V}" to ${this.name}`);return typeof V=="number"&&(V=BigInt(V)),V},Cb:gr,readValueFromPointer:Ts(E,A,E.indexOf("u")==-1),Db:null})}var gr=8;function Hs(h,E,A,z){Rr(h>>>=0,{name:E=Wt(E>>>0),fromWireType:function(V){return!!V},toWireType:function(V,ce){return ce?A:z},Cb:gr,readValueFromPointer:function(V){return this.fromWireType(B()[V>>>0])},Db:null})}var cs=[],jr=[];function Hr(h){9<(h>>>=0)&&--jr[h+1]==0&&(jr[h]=void 0,cs.push(h))}var ur=h=>{if(!h)throw new at("Cannot use deleted val. handle = "+h);return jr[h]},mr=h=>{switch(h){case void 0:return 2;case null:return 4;case!0:return 6;case!1:return 8;default:let E=cs.pop()||jr.length;return jr[E]=h,jr[E+1]=1,E}};function Es(h){return this.fromWireType(Se()[h>>>2>>>0])}var Ds={name:"emscripten::val",fromWireType:h=>{var E=ur(h);return Hr(h),E},toWireType:(h,E)=>mr(E),Cb:gr,readValueFromPointer:Es,Db:null};function Ls(h){return Rr(h>>>0,Ds)}var zs=(h,E)=>{switch(E){case 4:return function(A){return this.fromWireType(De()[A>>>2>>>0])};case 8:return function(A){return this.fromWireType(Ne()[A>>>3>>>0])};default:throw new TypeError(`invalid float width (${E}): ${h}`)}};function us(h,E,A){A>>>=0,Rr(h>>>=0,{name:E=Wt(E>>>0),fromWireType:z=>z,toWireType:(z,V)=>V,Cb:gr,readValueFromPointer:zs(E,A),Db:null})}function Wr(h,E,A,z,V){if(h>>>=0,A>>>=0,E=Wt(E>>>0),V===-1&&(V=4294967295),V=Re=>Re,z===0){var ce=32-8*A;V=Re=>Re<<ce>>>ce}var Pe=E.includes("unsigned")?function(Re,Je){return Je>>>0}:function(Re,Je){return Je};Rr(h,{name:E,fromWireType:V,toWireType:Pe,Cb:gr,readValueFromPointer:Ts(E,A,z!==0),Db:null})}function je(h,E,A){function z(ce){var Pe=Se()[ce>>>2>>>0];return ce=Se()[ce+4>>>2>>>0],new V(G().buffer,ce,Pe)}var V=[Int8Array,Uint8Array,Int16Array,Uint16Array,Int32Array,Uint32Array,Float32Array,Float64Array,BigInt64Array,BigUint64Array][E];Rr(h>>>=0,{name:A=Wt(A>>>0),fromWireType:z,Cb:gr,readValueFromPointer:z},{Sb:!0})}function Ve(h,E){Rr(h>>>=0,{name:E=Wt(E>>>0),fromWireType:function(A){for(var z,V=Se()[A>>>2>>>0],ce=A+4,Pe=ce,Re=0;Re<=V;++Re){var Je=ce+Re;Re!=V&&B()[Je>>>0]!=0||(Pe=Yt(Pe,Je-Pe),z===void 0?z=Pe:(z+="\0",z+=Pe),Pe=Je+1)}return Zr(A),z},toWireType:function(A,z){z instanceof ArrayBuffer&&(z=new Uint8Array(z));var V=typeof z=="string";if(!(V||z instanceof Uint8Array||z instanceof Uint8ClampedArray||z instanceof Int8Array))throw new at("Cannot pass non-string to std::string");var ce=V?Fs(z):z.length,Pe=An(4+ce+1),Re=Pe+4;if(Se()[Pe>>>2>>>0]=ce,V)S(z,Re,ce+1);else if(V)for(V=0;V<ce;++V){var Je=z.charCodeAt(V);if(255<Je)throw Zr(Pe),new at("String has UTF-16 code units that do not fit in 8 bits");B()[Re+V>>>0]=Je}else for(V=0;V<ce;++V)B()[Re+V>>>0]=z[V];return A!==null&&A.push(Zr,Pe),Pe},Cb:gr,readValueFromPointer:Es,Db(A){Zr(A)}})}var st=typeof TextDecoder<"u"?new TextDecoder("utf-16le"):void 0,qt=(h,E)=>{for(var A=h>>1,z=A+E/2;!(A>=z)&&le()[A>>>0];)++A;if(32<(A<<=1)-h&&st)return st.decode(B().slice(h,A));for(A="",z=0;!(z>=E/2);++z){var V=Q()[h+2*z>>>1>>>0];if(V==0)break;A+=String.fromCharCode(V)}return A},qs=(h,E,A)=>{if(A??(A=2147483647),2>A)return 0;var z=E;A=(A-=2)<2*h.length?A/2:h.length;for(var V=0;V<A;++V){var ce=h.charCodeAt(V);Q()[E>>>1>>>0]=ce,E+=2}return Q()[E>>>1>>>0]=0,E-z},Bs=h=>2*h.length,Qs=(h,E)=>{for(var A=0,z="";!(A>=E/4);){var V=ee()[h+4*A>>>2>>>0];if(V==0)break;++A,65536<=V?(V-=65536,z+=String.fromCharCode(55296|V>>10,56320|1023&V)):z+=String.fromCharCode(V)}return z},En=(h,E,A)=>{if(E>>>=0,A??(A=2147483647),4>A)return 0;var z=E;A=z+A-4;for(var V=0;V<h.length;++V){var ce=h.charCodeAt(V);if(55296<=ce&&57343>=ce&&(ce=65536+((1023&ce)<<10)|1023&h.charCodeAt(++V)),ee()[E>>>2>>>0]=ce,(E+=4)+4>A)break}return ee()[E>>>2>>>0]=0,E-z},he=h=>{for(var E=0,A=0;A<h.length;++A){var z=h.charCodeAt(A);55296<=z&&57343>=z&&++A,E+=4}return E};function I(h,E,A){if(h>>>=0,E>>>=0,A=Wt(A>>>=0),E===2)var z=qt,V=qs,ce=Bs,Pe=Re=>le()[Re>>>1>>>0];else E===4&&(z=Qs,V=En,ce=he,Pe=Re=>Se()[Re>>>2>>>0]);Rr(h,{name:A,fromWireType:Re=>{for(var Je,it=Se()[Re>>>2>>>0],xt=Re+4,Rt=0;Rt<=it;++Rt){var Qt=Re+4+Rt*E;Rt!=it&&Pe(Qt)!=0||(xt=z(xt,Qt-xt),Je===void 0?Je=xt:(Je+="\0",Je+=xt),xt=Qt+E)}return Zr(Re),Je},toWireType:(Re,Je)=>{if(typeof Je!="string")throw new at(`Cannot pass non-string to C++ string type ${A}`);var it=ce(Je),xt=An(4+it+E);return Se()[xt>>>2>>>0]=it/E,V(Je,xt+4,it+E),Re!==null&&Re.push(Zr,xt),xt},Cb:gr,readValueFromPointer:Es,Db(Re){Zr(Re)}})}function N(h,E){Rr(h>>>=0,{Tb:!0,name:E=Wt(E>>>0),Cb:0,fromWireType:()=>{},toWireType:()=>{}})}function Y(h){to(h>>>0,!l,1,!i,131072,!1),Ur()}var ie=h=>{if(!pe)try{if(h(),!(0<Qe))try{c?Fn(C):ft(C)}catch(E){E instanceof Fe||E=="unwind"||k(0,E)}}catch(E){E instanceof Fe||E=="unwind"||k(0,E)}};function de(h){h>>>=0,typeof Atomics.jc=="function"&&(Atomics.jc(ee(),h>>>2,h).value.then(Te),h+=128,Atomics.store(ee(),h>>>2,1))}var Te=()=>{var h=In();h&&(de(h),ie(Ko))};function Ge(h,E){(h>>>=0)==E>>>0?setTimeout(Te):c?postMessage({Gb:h,Bb:"checkMailbox"}):(h=fr[h])&&h.postMessage({Bb:"checkMailbox"})}var Xe=[];function qe(h,E,A,z,V){for(E>>>=0,z/=2,Xe.length=z,A=V>>>0>>>3,V=0;V<z;V++)Xe[V]=re[A+2*V]?re[A+2*V+1]:Ne()[A+2*V+1>>>0];return(E?we[E]:eo[h])(...Xe)}var tt=()=>{Qe=0};function bt(h){h>>>=0,c?postMessage({Bb:"cleanupThread",hc:h}):Ut(fr[h])}function jt(h){}var Bt=(h,E)=>{var A=br[h];if(A===void 0)throw h=No(h),A=Wt(h),Zr(h),new at(`${E} has unknown type ${A}`);return A},nr=(h,E,A)=>{var z=[];return h=h.toWireType(z,A),z.length&&(Se()[E>>>2>>>0]=mr(z)),h};function Jt(h,E,A){return E>>>=0,A>>>=0,h=ur(h>>>0),E=Bt(E,"emval::as"),nr(E,A,h)}function yr(h,E){return E>>>=0,h=ur(h>>>0),(E=Bt(E,"emval::as")).toWireType(null,h)}var vr=h=>{try{h()}catch(E){Me(E)}},hr=0,xr=null,Cr=0,ss=[],rr={},Tr={},Ar=0,Yr=null,Ps=[];function Fr(h){return function(E){if(!pe){if(hr===0){var A=!1,z=!1;E((V=0)=>{if(!pe&&(Cr=V,A=!0,z)){hr=2,vr(()=>Qo(xr)),typeof MainLoop<"u"&&MainLoop.Pb&&MainLoop.resume(),V=!1;try{var ce=function(){var Je=ee()[xr+8>>>2>>>0];return Je=mt[Tr[Je]],--Qe,Je()}()}catch(Je){ce=Je,V=!0}var Pe=!1;if(!xr){var Re=Yr;Re&&(Yr=null,(V?Re.reject:Re.resolve)(ce),Pe=!0)}if(V&&!Pe)throw ce}}),z=!0,A||(hr=1,xr=function(){var V=An(65548),ce=V+12;Se()[V>>>2>>>0]=ce,Se()[V+4>>>2>>>0]=ce+65536,ce=ss[0];var Pe=rr[ce];return Pe===void 0&&(Pe=Ar++,rr[ce]=Pe,Tr[Pe]=ce),ce=Pe,ee()[V+8>>>2>>>0]=ce,V}(),typeof MainLoop<"u"&&MainLoop.Pb&&MainLoop.pause(),vr(()=>qo(xr)))}else hr===2?(hr=0,vr(Xo),Zr(xr),xr=null,Ps.forEach(ie)):Me(`invalid state: ${hr}`);return Cr}}(E=>{h().then(E)})}function ds(h){return h>>>=0,Fr(async()=>{var E=await ur(h);return mr(E)})}var dr=[];function _r(h,E,A,z){return A>>>=0,z>>>=0,(h=dr[h>>>0])(null,E=ur(E>>>0),A,z)}var wr={},or=h=>{var E=wr[h];return E===void 0?Wt(h):E};function qr(h,E,A,z,V){return A>>>=0,z>>>=0,V>>>=0,(h=dr[h>>>0])(E=ur(E>>>0),E[A=or(A)],z,V)}var Xs=()=>typeof globalThis=="object"?globalThis:Function("return this")();function Rs(h){return(h>>>=0)==0?mr(Xs()):(h=or(h),mr(Xs()[h]))}var xa=h=>{var E=dr.length;return dr.push(h),E},Ta=(h,E)=>{for(var A=Array(h),z=0;z<h;++z)A[z]=Bt(Se()[E+4*z>>>2>>>0],"parameter "+z);return A},$o=(h,E)=>Object.defineProperty(E,"name",{value:h});function Ea(h,E,A){var z=(E=Ta(h,E>>>0)).shift();h--;var V=`return function (obj, func, destructorsRef, args) {
`,ce=0,Pe=[];A===0&&Pe.push("obj");for(var Re=["retType"],Je=[z],it=0;it<h;++it)Pe.push("arg"+it),Re.push("argType"+it),Je.push(E[it]),V+=`  var arg${it} = argType${it}.readValueFromPointer(args${ce?"+"+ce:""});
`,ce+=E[it].Cb;return V+=`  var rv = ${A===1?"new func":"func.call"}(${Pe.join(", ")});
`,z.Tb||(Re.push("emval_returnValue"),Je.push(nr),V+=`  return emval_returnValue(retType, destructorsRef, rv);
`),Re.push(V+`};
`),h=function(xt){var Rt=Function;if(!(Rt instanceof Function))throw new TypeError(`new_ called with constructor type ${typeof Rt} which is not a function`);var Qt=$o(Rt.name||"unknownFunctionName",function(){});return Qt.prototype=Rt.prototype,Qt=new Qt,(xt=Rt.apply(Qt,xt))instanceof Object?xt:Qt}(Re)(...Je),A=`methodCaller<(${E.map(xt=>xt.name).join(", ")}) => ${z.name}>`,xa($o(A,h))}function Pa(h){return h=or(h>>>0),mr(n[h])}function on(h,E){return E>>>=0,h=ur(h>>>0),E=ur(E),mr(h[E])}function Ca(h){9<(h>>>=0)&&(jr[h+1]+=1)}function Sa(){return mr([])}function $a(h){h=ur(h>>>0);for(var E=Array(h.length),A=0;A<h.length;A++)E[A]=h[A];return mr(E)}function ka(h){return mr(or(h>>>0))}function Ia(){return mr({})}function an(h){for(var E=ur(h>>>=0);E.length;){var A=E.pop();E.pop()(A)}Hr(h)}function Aa(h,E,A){E>>>=0,A>>>=0,h=ur(h>>>0),E=ur(E),A=ur(A),h[E]=A}function Fa(h,E){return E>>>=0,h=(h=Bt(h>>>0,"_emval_take_value")).readValueFromPointer(E),mr(h)}function Oa(h,E){h=-9007199254740992>h||9007199254740992<h?NaN:Number(h),E>>>=0,h=new Date(1e3*h),ee()[E>>>2>>>0]=h.getUTCSeconds(),ee()[E+4>>>2>>>0]=h.getUTCMinutes(),ee()[E+8>>>2>>>0]=h.getUTCHours(),ee()[E+12>>>2>>>0]=h.getUTCDate(),ee()[E+16>>>2>>>0]=h.getUTCMonth(),ee()[E+20>>>2>>>0]=h.getUTCFullYear()-1900,ee()[E+24>>>2>>>0]=h.getUTCDay(),h=(h.getTime()-Date.UTC(h.getUTCFullYear(),0,1,0,0,0,0))/864e5|0,ee()[E+28>>>2>>>0]=h}var ko=h=>h%4==0&&(h%100!=0||h%400==0),Io=[0,31,60,91,121,152,182,213,244,274,305,335],Xn=[0,31,59,90,120,151,181,212,243,273,304,334];function Da(h,E){h=-9007199254740992>h||9007199254740992<h?NaN:Number(h),E>>>=0,h=new Date(1e3*h),ee()[E>>>2>>>0]=h.getSeconds(),ee()[E+4>>>2>>>0]=h.getMinutes(),ee()[E+8>>>2>>>0]=h.getHours(),ee()[E+12>>>2>>>0]=h.getDate(),ee()[E+16>>>2>>>0]=h.getMonth(),ee()[E+20>>>2>>>0]=h.getFullYear()-1900,ee()[E+24>>>2>>>0]=h.getDay();var A=(ko(h.getFullYear())?Io:Xn)[h.getMonth()]+h.getDate()-1|0;ee()[E+28>>>2>>>0]=A,ee()[E+36>>>2>>>0]=-60*h.getTimezoneOffset(),A=new Date(h.getFullYear(),6,1).getTimezoneOffset();var z=new Date(h.getFullYear(),0,1).getTimezoneOffset();h=0|(A!=z&&h.getTimezoneOffset()==Math.min(z,A)),ee()[E+32>>>2>>>0]=h}function La(h){h>>>=0;var E=new Date(ee()[h+20>>>2>>>0]+1900,ee()[h+16>>>2>>>0],ee()[h+12>>>2>>>0],ee()[h+8>>>2>>>0],ee()[h+4>>>2>>>0],ee()[h>>>2>>>0],0),A=ee()[h+32>>>2>>>0],z=E.getTimezoneOffset(),V=new Date(E.getFullYear(),6,1).getTimezoneOffset(),ce=new Date(E.getFullYear(),0,1).getTimezoneOffset(),Pe=Math.min(ce,V);return 0>A?ee()[h+32>>>2>>>0]=+(V!=ce&&Pe==z):0<A!=(Pe==z)&&(V=Math.max(ce,V),E.setTime(E.getTime()+6e4*((0<A?Pe:V)-z))),ee()[h+24>>>2>>>0]=E.getDay(),A=(ko(E.getFullYear())?Io:Xn)[E.getMonth()]+E.getDate()-1|0,ee()[h+28>>>2>>>0]=A,ee()[h>>>2>>>0]=E.getSeconds(),ee()[h+4>>>2>>>0]=E.getMinutes(),ee()[h+8>>>2>>>0]=E.getHours(),ee()[h+12>>>2>>>0]=E.getDate(),ee()[h+16>>>2>>>0]=E.getMonth(),ee()[h+20>>>2>>>0]=E.getYear(),h=E.getTime(),BigInt(isNaN(h)?-1:h/1e3)}function Jn(h,E,A,z,V,ce,Pe){return c?Oe(16,1,h,E,A,z,V,ce,Pe):-52}function Ao(h,E,A,z,V,ce){if(c)return Oe(17,1,h,E,A,z,V,ce)}var Js={},za=()=>performance.timeOrigin+performance.now();function Yn(h,E){if(c)return Oe(18,1,h,E);if(Js[h]&&(clearTimeout(Js[h].id),delete Js[h]),!E)return 0;var A=setTimeout(()=>{delete Js[h],ie(()=>Go(h,performance.timeOrigin+performance.now()))},E);return Js[h]={id:A,qc:E},0}function gc(h,E,A,z){h>>>=0,E>>>=0,A>>>=0,z>>>=0;var V=new Date().getFullYear(),ce=new Date(V,0,1).getTimezoneOffset();V=new Date(V,6,1).getTimezoneOffset();var Pe=Math.max(ce,V);Se()[h>>>2>>>0]=60*Pe,ee()[E>>>2>>>0]=+(ce!=V),h=(E=Re=>{var Je=Math.abs(Re);return`UTC${0<=Re?"-":"+"}${String(Math.floor(Je/60)).padStart(2,"0")}${String(Je%60).padStart(2,"0")}`})(ce),E=E(V),V<ce?(S(h,A,17),S(E,z,17)):(S(h,z,17),S(E,A,17))}var Ba=()=>Date.now();function Fo(h,E,A){return 0<=h&&3>=h?(h===0?h=Date.now():h=performance.timeOrigin+performance.now(),re[A>>>0>>>3]=BigInt(Math.round(1e6*h)),0):28}var Pn=[],Cn=(h,E)=>{Pn.length=0;for(var A;A=B()[h++>>>0];){var z=A!=105;E+=(z&=A!=112)&&E%8?4:0,Pn.push(A==112?Se()[E>>>2>>>0]:A==106?re[E>>>3]:A==105?ee()[E>>>2>>>0]:Ne()[E>>>3>>>0]),E+=z?8:4}return Pn};function Ra(h,E,A){return h>>>=0,E=Cn(E>>>0,A>>>0),we[h](...E)}function ja(h,E,A){return h>>>=0,E=Cn(E>>>0,A>>>0),we[h](...E)}var Na=()=>{};function Va(h,E){return M(Yt(h>>>0,E>>>0))}var Ua=()=>{throw Qe+=1,"unwind"};function Wa(){return 4294901760}var Oo=()=>navigator.hardwareConcurrency;function Ga(){return Me("Cannot use emscripten_pc_get_function without -sUSE_OFFSET_CONVERTER"),0}function Ka(h){h>>>=0;var E=B().length;if(h<=E||4294901760<h)return!1;for(var A=1;4>=A;A*=2){var z=E*(1+.2/A);z=Math.min(z,h+100663296);e:{z=(Math.min(4294901760,65536*Math.ceil(Math.max(h,z)/65536))-x.buffer.byteLength+65535)/65536|0;try{x.grow(z),O();var V=1;break e}catch{}V=void 0}if(V)return!0}return!1}var ln=()=>(Me("Cannot use convertFrameToPC (needed by __builtin_return_address) without -sUSE_OFFSET_CONVERTER"),0),cn={},Sn=h=>{h.forEach(E=>{ln()})};function Do(){var h=Error().stack.toString().split(`
`);return h[0]=="Error"&&h.shift(),Sn(h),cn.Lb=ln(),cn.cc=h,cn.Lb}function Ha(h,E,A){if(h>>>=0,E>>>=0,cn.Lb==h)var z=cn.cc;else(z=Error().stack.toString().split(`
`))[0]=="Error"&&z.shift(),Sn(z);for(var V=3;z[V]&&ln()!=h;)++V;for(h=0;h<A&&z[h+V];++h)ee()[E+4*h>>>2>>>0]=ln();return h}var $n,Cs={},Lo=()=>{if(!$n){var h,E={USER:"web_user",LOGNAME:"web_user",PATH:"/",PWD:"/",HOME:"/home/web_user",LANG:(typeof navigator=="object"&&navigator.languages&&navigator.languages[0]||"C").replace("-","_")+".UTF-8",_:"./this.program"};for(h in Cs)Cs[h]===void 0?delete E[h]:E[h]=Cs[h];var A=[];for(h in E)A.push(`${h}=${E[h]}`);$n=A}return $n};function qa(h,E){if(c)return Oe(19,1,h,E);h>>>=0,E>>>=0;var A=0;return Lo().forEach((z,V)=>{var ce=E+A;for(V=Se()[h+4*V>>>2>>>0]=ce,ce=0;ce<z.length;++ce)G()[V++>>>0]=z.charCodeAt(ce);G()[V>>>0]=0,A+=z.length+1}),0}function zo(h,E){if(c)return Oe(20,1,h,E);h>>>=0,E>>>=0;var A=Lo();Se()[h>>>2>>>0]=A.length;var z=0;return A.forEach(V=>z+=V.length+1),Se()[E>>>2>>>0]=z,0}function Qa(h){return c?Oe(21,1,h):52}function Bo(h,E,A,z){return c?Oe(22,1,h,E,A,z):52}function Zn(h,E,A,z){return c?Oe(23,1,h,E,A,z):70}var Xa=[null,[],[]];function Ro(h,E,A,z){if(c)return Oe(24,1,h,E,A,z);E>>>=0,A>>>=0,z>>>=0;for(var V=0,ce=0;ce<A;ce++){var Pe=Se()[E>>>2>>>0],Re=Se()[E+4>>>2>>>0];E+=8;for(var Je=0;Je<Re;Je++){var it=B()[Pe+Je>>>0],xt=Xa[h];it===0||it===10?((h===1?T:M)(As(xt)),xt.length=0):xt.push(it)}V+=Re}return Se()[z>>>2>>>0]=V,0}c||function(){for(var h=n.numThreads-1;h--;)Gs();We.unshift(()=>{be++,function(E){c?E():Promise.all(Lt.map(Ws)).then(E)}(()=>$e())})}();for(var jo=Array(256),kn=0;256>kn;++kn)jo[kn]=String.fromCharCode(kn);Ot=jo,at=n.BindingError=class extends Error{constructor(h){super(h),this.name="BindingError"}},n.InternalError=class extends Error{constructor(h){super(h),this.name="InternalError"}},jr.push(0,1,void 0,1,null,1,!0,1,!1,1),n.count_emval_handles=()=>jr.length/2-5-cs.length;var mt,eo=[ze,pt,Ks,Jr,$t,H,R,W,X,ue,Ee,et,ot,rt,St,lt,Jn,Ao,Yn,qa,zo,Qa,Bo,Zn,Ro];(async function(){function h(z,V){return mt=z.exports,mt=function(){var ce=mt,Pe={};for(let[Re,Je]of Object.entries(ce))Pe[Re]=typeof Je=="function"?(...it)=>{ss.push(Re);try{return Je(...it)}finally{pe||(ss.pop(),xr&&hr===1&&ss.length===0&&(hr=0,Qe+=1,vr(oo),typeof Fibers<"u"&&Fibers.rc()))}}:Je;return Pe}(),mt=function(){var ce=mt,Pe=Je=>it=>Je(it)>>>0,Re=Je=>()=>Je()>>>0;return(ce=Object.assign({},ce)).Da=Pe(ce.Da),ce.fb=Re(ce.fb),ce.hb=Pe(ce.hb),ce.tb=Pe(ce.tb),ce.ub=Re(ce.ub),ce.__cxa_get_exception_ptr=Pe(ce.__cxa_get_exception_ptr),ce}(),Mr.push(mt.ib),v=V,$e(),mt}be++;var E=Le();if(n.instantiateWasm)return new Promise(z=>{n.instantiateWasm(E,(V,ce)=>{h(V,ce),z(V.exports)})});if(c)return new Promise(z=>{Ae=V=>{var ce=new WebAssembly.Instance(V,Le());z(h(ce,V))}});xe??(xe=n.locateFile?n.locateFile?n.locateFile("ort-wasm-simd-threaded.jsep.wasm",w):w+"ort-wasm-simd-threaded.jsep.wasm":new URL("/assets/ort-wasm-simd-threaded.jsep-B0T3yYHD.wasm",self.location.href).href);try{var A=await async function(z){var V=xe;if(!_e&&typeof WebAssembly.instantiateStreaming=="function"&&!oe(V))try{var ce=fetch(V,{credentials:"same-origin"});return await WebAssembly.instantiateStreaming(ce,z)}catch(Pe){M(`wasm streaming compile failed: ${Pe}`),M("falling back to ArrayBuffer instantiation")}return async function(Pe,Re){try{var Je=await async function(it){if(!_e)try{var xt=await f(it);return new Uint8Array(xt)}catch{}if(it==xe&&_e)it=new Uint8Array(_e);else{if(!_)throw"both async and sync fetching of the wasm failed";it=_(it)}return it}(Pe);return await WebAssembly.instantiate(Je,Re)}catch(it){M(`failed to asynchronously prepare wasm: ${it}`),Me(it)}}(V,z)}(E);return h(A.instance,A.module)}catch(z){return o(z),Promise.reject(z)}})();var No=h=>(No=mt.Da)(h),Vo=()=>(Vo=mt.Ea)();n._OrtInit=(h,E)=>(n._OrtInit=mt.Fa)(h,E),n._OrtGetLastError=(h,E)=>(n._OrtGetLastError=mt.Ga)(h,E),n._OrtCreateSessionOptions=(h,E,A,z,V,ce,Pe,Re,Je,it)=>(n._OrtCreateSessionOptions=mt.Ha)(h,E,A,z,V,ce,Pe,Re,Je,it),n._OrtAppendExecutionProvider=(h,E,A,z,V)=>(n._OrtAppendExecutionProvider=mt.Ia)(h,E,A,z,V),n._OrtAddFreeDimensionOverride=(h,E,A)=>(n._OrtAddFreeDimensionOverride=mt.Ja)(h,E,A),n._OrtAddSessionConfigEntry=(h,E,A)=>(n._OrtAddSessionConfigEntry=mt.Ka)(h,E,A),n._OrtReleaseSessionOptions=h=>(n._OrtReleaseSessionOptions=mt.La)(h),n._OrtCreateSession=(h,E,A)=>(n._OrtCreateSession=mt.Ma)(h,E,A),n._OrtReleaseSession=h=>(n._OrtReleaseSession=mt.Na)(h),n._OrtGetInputOutputCount=(h,E,A)=>(n._OrtGetInputOutputCount=mt.Oa)(h,E,A),n._OrtGetInputOutputMetadata=(h,E,A,z)=>(n._OrtGetInputOutputMetadata=mt.Pa)(h,E,A,z),n._OrtFree=h=>(n._OrtFree=mt.Qa)(h),n._OrtCreateTensor=(h,E,A,z,V,ce)=>(n._OrtCreateTensor=mt.Ra)(h,E,A,z,V,ce),n._OrtGetTensorData=(h,E,A,z,V)=>(n._OrtGetTensorData=mt.Sa)(h,E,A,z,V),n._OrtReleaseTensor=h=>(n._OrtReleaseTensor=mt.Ta)(h),n._OrtCreateRunOptions=(h,E,A,z)=>(n._OrtCreateRunOptions=mt.Ua)(h,E,A,z),n._OrtAddRunConfigEntry=(h,E,A)=>(n._OrtAddRunConfigEntry=mt.Va)(h,E,A),n._OrtReleaseRunOptions=h=>(n._OrtReleaseRunOptions=mt.Wa)(h),n._OrtCreateBinding=h=>(n._OrtCreateBinding=mt.Xa)(h),n._OrtBindInput=(h,E,A)=>(n._OrtBindInput=mt.Ya)(h,E,A),n._OrtBindOutput=(h,E,A,z)=>(n._OrtBindOutput=mt.Za)(h,E,A,z),n._OrtClearBoundOutputs=h=>(n._OrtClearBoundOutputs=mt._a)(h),n._OrtReleaseBinding=h=>(n._OrtReleaseBinding=mt.$a)(h),n._OrtRunWithBinding=(h,E,A,z,V)=>(n._OrtRunWithBinding=mt.ab)(h,E,A,z,V),n._OrtRun=(h,E,A,z,V,ce,Pe,Re)=>(n._OrtRun=mt.bb)(h,E,A,z,V,ce,Pe,Re),n._OrtEndProfiling=h=>(n._OrtEndProfiling=mt.cb)(h),n._JsepOutput=(h,E,A)=>(n._JsepOutput=mt.db)(h,E,A),n._JsepGetNodeName=h=>(n._JsepGetNodeName=mt.eb)(h);var In=()=>(In=mt.fb)(),Zr=n._free=h=>(Zr=n._free=mt.gb)(h),An=n._malloc=h=>(An=n._malloc=mt.hb)(h),to=(h,E,A,z,V,ce)=>(to=mt.kb)(h,E,A,z,V,ce),ro=()=>(ro=mt.lb)(),Uo=(h,E,A,z,V)=>(Uo=mt.mb)(h,E,A,z,V),Wo=h=>(Wo=mt.nb)(h),Fn=h=>(Fn=mt.ob)(h),Go=(h,E)=>(Go=mt.pb)(h,E),Ko=()=>(Ko=mt.qb)(),so=(h,E)=>(so=mt.rb)(h,E),On=h=>(On=mt.sb)(h),no=h=>(no=mt.tb)(h),Dn=()=>(Dn=mt.ub)(),Ho=n.dynCall_ii=(h,E)=>(Ho=n.dynCall_ii=mt.vb)(h,E),qo=h=>(qo=mt.wb)(h),oo=()=>(oo=mt.xb)(),Qo=h=>(Qo=mt.yb)(h),Xo=()=>(Xo=mt.zb)();return n.stackSave=()=>Dn(),n.stackRestore=h=>On(h),n.stackAlloc=h=>no(h),n.setValue=function(h,E,A="i8"){switch(A.endsWith("*")&&(A="*"),A){case"i1":case"i8":G()[h>>>0]=E;break;case"i16":Q()[h>>>1>>>0]=E;break;case"i32":ee()[h>>>2>>>0]=E;break;case"i64":re[h>>>3]=BigInt(E);break;case"float":De()[h>>>2>>>0]=E;break;case"double":Ne()[h>>>3>>>0]=E;break;case"*":Se()[h>>>2>>>0]=E;break;default:Me(`invalid type for setValue: ${A}`)}},n.getValue=function(h,E="i8"){switch(E.endsWith("*")&&(E="*"),E){case"i1":case"i8":return G()[h>>>0];case"i16":return Q()[h>>>1>>>0];case"i32":return ee()[h>>>2>>>0];case"i64":return re[h>>>3];case"float":return De()[h>>>2>>>0];case"double":return Ne()[h>>>3>>>0];case"*":return Se()[h>>>2>>>0];default:Me(`invalid type for getValue: ${E}`)}},n.UTF8ToString=Yt,n.stringToUTF8=S,n.lengthBytesUTF8=Fs,function h(){if(0<be)Ce=h;else if(c)s(n),me();else{for(;0<We.length;)We.shift()(n);0<be?Ce=h:(n.calledRun=!0,pe||(me(),s(n)))}}(),n.PTR_SIZE=4,a}),gd=Oi,wd=(r=(e=globalThis.self)==null?void 0:e.name)==null?void 0:r.startsWith("em-pthread"),wd&&Oi()}),Di,Li,Md,ts,bd,ra,yd,vd,zi,xd,Bi,Td,Ri,Ed,ji=Ue(()=>{ki(),Di=typeof location>"u"?void 0:location.origin,Li=self.location.href>"file:"&&self.location.href<"file;",Md=()=>{{if(Li){let e=URL;return new URL(new e("ort.bundle.min.mjs",self.location.href).href,Di).href}return self.location.href}},ts=Md(),bd=()=>{if(ts&&!ts.startsWith("blob:"))return ts.substring(0,ts.lastIndexOf("/")+1)},ra=(e,r)=>{try{let t=r??ts;return(t?new URL(e,t):new URL(e)).origin===Di}catch{return!1}},yd=(e,r)=>{let t=r??ts;try{return(t?new URL(e,t):new URL(e)).href}catch{return}},vd=(e,r)=>`${r??"./"}${e}`,zi=async e=>{let r=await(await fetch(e,{credentials:"same-origin"})).blob();return URL.createObjectURL(r)},xd=async e=>(await import(e)).default,Bi=($x(),uo(hd)).default,Td=async()=>{if(!ts)throw new Error("Failed to load proxy worker: cannot determine the script source URL.");if(ra(ts))return[void 0,Bi()];let e=await zi(ts);return[e,Bi(e)]},Ri=(kx(),uo(fd)).default,Ed=async(e,r,t)=>{if(!e&&!r&&Ri&&ts&&ra(ts))return[void 0,Ri];{let s="ort-wasm-simd-threaded.jsep.mjs",o=e??yd(s,r),n=t&&o&&!ra(o,r),a=n?await zi(o):o??vd(s,r);return[n?a:void 0,await xd(a)]}}}),Ni,sa,_o,Vi,Pd,Cd,Sd,Ui,Zt,_n=Ue(()=>{ji(),sa=!1,_o=!1,Vi=!1,Pd=()=>{if(typeof SharedArrayBuffer>"u")return!1;try{return typeof MessageChannel<"u"&&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{return!1}},Cd=()=>{try{return WebAssembly.validate(new Uint8Array([0,97,115,109,1,0,0,0,1,4,1,96,0,0,3,2,1,0,10,30,1,28,0,65,0,253,15,253,12,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,253,186,1,26,11]))}catch{return!1}},Sd=()=>{try{return WebAssembly.validate(new Uint8Array([0,97,115,109,1,0,0,0,1,5,1,96,0,1,123,3,2,1,0,10,19,1,17,0,65,1,253,15,65,2,253,15,65,3,253,15,253,147,2,11]))}catch{return!1}},Ui=async e=>{if(sa)return Promise.resolve();if(_o)throw new Error("multiple calls to 'initializeWebAssembly()' detected.");if(Vi)throw new Error("previous call to 'initializeWebAssembly()' failed.");_o=!0;let r=e.initTimeout,t=e.numThreads;if(e.simd!==!1){if(e.simd==="relaxed"){if(!Sd())throw new Error("Relaxed WebAssembly SIMD is not supported in the current environment.")}else if(!Cd())throw new Error("WebAssembly SIMD is not supported in the current environment.")}let s=Pd();t>1&&!s&&(typeof self<"u"&&!self.crossOriginIsolated&&console.warn("env.wasm.numThreads is set to "+t+", but this will not work unless you enable crossOriginIsolated mode. See https://web.dev/cross-origin-isolation-guide/ for more info."),console.warn("WebAssembly multi-threading is not supported in the current environment. Falling back to single-threading."),e.numThreads=t=1);let o=e.wasmPaths,n=typeof o=="string"?o:void 0,a=o==null?void 0:o.mjs,i=(a==null?void 0:a.href)??a,l=o==null?void 0:o.wasm,c=(l==null?void 0:l.href)??l,p=e.wasmBinary,[d,u]=await Ed(i,n,t>1),f=!1,_=[];if(r>0&&_.push(new Promise(b=>{setTimeout(()=>{f=!0,b()},r)})),_.push(new Promise((b,k)=>{let w={numThreads:t};if(p)w.wasmBinary=p;else if(c||n)w.locateFile=y=>c??n+y;else if(i&&i.indexOf("blob:")!==0)w.locateFile=y=>new URL(y,i).href;else if(d){let y=bd();y&&(w.locateFile=$=>y+$)}u(w).then(y=>{_o=!1,sa=!0,Ni=y,b(),d&&URL.revokeObjectURL(d)},y=>{_o=!1,Vi=!0,k(y)})})),await Promise.race(_),f)throw new Error(`WebAssembly backend initializing failed due to timeout: ${r}ms`)},Zt=()=>{if(sa&&Ni)return Ni;throw new Error("WebAssembly is not initialized yet.")}}),gs,na,Xt,Wi=Ue(()=>{_n(),gs=(e,r)=>{let t=Zt(),s=t.lengthBytesUTF8(e)+1,o=t._malloc(s);return t.stringToUTF8(e,o,s),r.push(o),o},na=(e,r,t,s)=>{if(typeof e=="object"&&e!==null){if(t.has(e))throw new Error("Circular reference in options");t.add(e)}Object.entries(e).forEach(([o,n])=>{let a=r?r+o:o;if(typeof n=="object")na(n,a+".",t,s);else if(typeof n=="string"||typeof n=="number")s(a,n.toString());else if(typeof n=="boolean")s(a,n?"1":"0");else throw new Error(`Can't handle extra config type: ${typeof n}`)})},Xt=e=>{let r=Zt(),t=r.stackSave();try{let s=r.PTR_SIZE,o=r.stackAlloc(2*s);r._OrtGetLastError(o,o+s);let n=Number(r.getValue(o,s===4?"i32":"i64")),a=r.getValue(o+s,"*"),i=a?r.UTF8ToString(a):"";throw new Error(`${e} ERROR_CODE: ${n}, ERROR_MESSAGE: ${i}`)}finally{r.stackRestore(t)}}}),$d,Ix=Ue(()=>{_n(),Wi(),$d=e=>{let r=Zt(),t=0,s=[],o=e||{};try{if((e==null?void 0:e.logSeverityLevel)===void 0)o.logSeverityLevel=2;else if(typeof e.logSeverityLevel!="number"||!Number.isInteger(e.logSeverityLevel)||e.logSeverityLevel<0||e.logSeverityLevel>4)throw new Error(`log serverity level is not valid: ${e.logSeverityLevel}`);if((e==null?void 0:e.logVerbosityLevel)===void 0)o.logVerbosityLevel=0;else if(typeof e.logVerbosityLevel!="number"||!Number.isInteger(e.logVerbosityLevel))throw new Error(`log verbosity level is not valid: ${e.logVerbosityLevel}`);(e==null?void 0:e.terminate)===void 0&&(o.terminate=!1);let n=0;return(e==null?void 0:e.tag)!==void 0&&(n=gs(e.tag,s)),t=r._OrtCreateRunOptions(o.logSeverityLevel,o.logVerbosityLevel,!!o.terminate,n),t===0&&Xt("Can't create run options."),(e==null?void 0:e.extra)!==void 0&&na(e.extra,"",new WeakSet,(a,i)=>{let l=gs(a,s),c=gs(i,s);r._OrtAddRunConfigEntry(t,l,c)!==0&&Xt(`Can't set a run config entry: ${a} - ${i}.`)}),[t,s]}catch(n){throw t!==0&&r._OrtReleaseRunOptions(t),s.forEach(a=>r._free(a)),n}}}),kd,Id,Ad,fo,Fd,Od,Ax=Ue(()=>{_n(),Wi(),kd=e=>{switch(e){case"disabled":return 0;case"basic":return 1;case"extended":return 2;case"all":return 99;default:throw new Error(`unsupported graph optimization level: ${e}`)}},Id=e=>{switch(e){case"sequential":return 0;case"parallel":return 1;default:throw new Error(`unsupported execution mode: ${e}`)}},Ad=e=>{e.extra||(e.extra={}),e.extra.session||(e.extra.session={});let r=e.extra.session;r.use_ort_model_bytes_directly||(r.use_ort_model_bytes_directly="1"),e.executionProviders&&e.executionProviders.some(t=>(typeof t=="string"?t:t.name)==="webgpu")&&(e.enableMemPattern=!1)},fo=(e,r,t,s)=>{let o=gs(r,s),n=gs(t,s);Zt()._OrtAddSessionConfigEntry(e,o,n)!==0&&Xt(`Can't set a session config entry: ${r} - ${t}.`)},Fd=async(e,r,t)=>{for(let s of r){let o=typeof s=="string"?s:s.name,n=[];switch(o){case"webnn":if(o="WEBNN",typeof s!="string"){let p=s==null?void 0:s.deviceType;p&&fo(e,"deviceType",p,t)}break;case"webgpu":if(o="JS",typeof s!="string"){let p=s;if(p!=null&&p.preferredLayout){if(p.preferredLayout!=="NCHW"&&p.preferredLayout!=="NHWC")throw new Error(`preferredLayout must be either 'NCHW' or 'NHWC': ${p.preferredLayout}`);fo(e,"preferredLayout",p.preferredLayout,t)}}break;case"wasm":case"cpu":continue;default:throw new Error(`not supported execution provider: ${o}`)}let a=gs(o,t),i=n.length,l=0,c=0;if(i>0){l=Zt()._malloc(i*Zt().PTR_SIZE),t.push(l),c=Zt()._malloc(i*Zt().PTR_SIZE),t.push(c);for(let p=0;p<i;p++)Zt().setValue(l+p*Zt().PTR_SIZE,n[p][0],"*"),Zt().setValue(c+p*Zt().PTR_SIZE,n[p][1],"*")}await Zt()._OrtAppendExecutionProvider(e,a,l,c,i)!==0&&Xt(`Can't append execution provider: ${o}.`)}},Od=async e=>{let r=Zt(),t=0,s=[],o=e||{};Ad(o);try{let n=kd(o.graphOptimizationLevel??"all"),a=Id(o.executionMode??"sequential"),i=typeof o.logId=="string"?gs(o.logId,s):0,l=o.logSeverityLevel??2;if(!Number.isInteger(l)||l<0||l>4)throw new Error(`log serverity level is not valid: ${l}`);let c=o.logVerbosityLevel??0;if(!Number.isInteger(c)||c<0||c>4)throw new Error(`log verbosity level is not valid: ${c}`);let p=typeof o.optimizedModelFilePath=="string"?gs(o.optimizedModelFilePath,s):0;if(t=r._OrtCreateSessionOptions(n,!!o.enableCpuMemArena,!!o.enableMemPattern,a,!!o.enableProfiling,0,i,l,c,p),t===0&&Xt("Can't create session options."),o.executionProviders&&await Fd(t,o.executionProviders,s),o.enableGraphCapture!==void 0){if(typeof o.enableGraphCapture!="boolean")throw new Error(`enableGraphCapture must be a boolean value: ${o.enableGraphCapture}`);fo(t,"enableGraphCapture",o.enableGraphCapture.toString(),s)}if(o.freeDimensionOverrides)for(let[d,u]of Object.entries(o.freeDimensionOverrides)){if(typeof d!="string")throw new Error(`free dimension override name must be a string: ${d}`);if(typeof u!="number"||!Number.isInteger(u)||u<0)throw new Error(`free dimension override value must be a non-negative integer: ${u}`);let f=gs(d,s);r._OrtAddFreeDimensionOverride(t,f,u)!==0&&Xt(`Can't set a free dimension override: ${d} - ${u}.`)}return o.extra!==void 0&&na(o.extra,"",new WeakSet,(d,u)=>{fo(t,d,u,s)}),[t,s]}catch(n){throw t!==0&&r._OrtReleaseSessionOptions(t)!==0&&Xt("Can't release session options."),s.forEach(a=>r._free(a)),n}}}),Wn,js,fn,Gi,oa,Ki,Hi,qi,Mt=Ue(()=>{Wn=e=>{switch(e){case"int8":return 3;case"uint8":return 2;case"bool":return 9;case"int16":return 5;case"uint16":return 4;case"int32":return 6;case"uint32":return 12;case"float16":return 10;case"float32":return 1;case"float64":return 11;case"string":return 8;case"int64":return 7;case"uint64":return 13;case"int4":return 22;case"uint4":return 21;default:throw new Error(`unsupported data type: ${e}`)}},js=e=>{switch(e){case 3:return"int8";case 2:return"uint8";case 9:return"bool";case 5:return"int16";case 4:return"uint16";case 6:return"int32";case 12:return"uint32";case 10:return"float16";case 1:return"float32";case 11:return"float64";case 8:return"string";case 7:return"int64";case 13:return"uint64";case 22:return"int4";case 21:return"uint4";default:throw new Error(`unsupported data type: ${e}`)}},fn=(e,r)=>{let t=[-1,4,1,1,2,2,4,8,-1,1,2,8,4,8,-1,-1,-1,-1,-1,-1,-1,.5,.5][e],s=typeof r=="number"?r:r.reduce((o,n)=>o*n,1);return t>0?Math.ceil(s*t):void 0},Gi=e=>{switch(e){case"float16":return typeof Float16Array<"u"&&Float16Array.from?Float16Array:Uint16Array;case"float32":return Float32Array;case"uint8":return Uint8Array;case"int8":return Int8Array;case"uint16":return Uint16Array;case"int16":return Int16Array;case"int32":return Int32Array;case"bool":return Uint8Array;case"float64":return Float64Array;case"uint32":return Uint32Array;case"int64":return BigInt64Array;case"uint64":return BigUint64Array;default:throw new Error(`unsupported type: ${e}`)}},oa=e=>{switch(e){case"verbose":return 0;case"info":return 1;case"warning":return 2;case"error":return 3;case"fatal":return 4;default:throw new Error(`unsupported logging level: ${e}`)}},Ki=e=>e==="float32"||e==="float16"||e==="int32"||e==="int64"||e==="uint32"||e==="uint8"||e==="bool"||e==="uint4"||e==="int4",Hi=e=>e==="float32"||e==="float16"||e==="int32"||e==="int64"||e==="uint32"||e==="uint64"||e==="int8"||e==="uint8"||e==="bool"||e==="uint4"||e==="int4",qi=e=>{switch(e){case"none":return 0;case"cpu":return 1;case"cpu-pinned":return 2;case"texture":return 3;case"gpu-buffer":return 4;case"ml-tensor":return 5;default:throw new Error(`unsupported data location: ${e}`)}}}),Qi,Dd=Ue(()=>{ki(),Qi=async e=>{if(typeof e=="string"){let r=await fetch(e);if(!r.ok)throw new Error(`failed to load external data file: ${e}`);let t=r.headers.get("Content-Length"),s=t?parseInt(t,10):0;if(s<1073741824)return new Uint8Array(await r.arrayBuffer());{if(!r.body)throw new Error(`failed to load external data file: ${e}, no response body.`);let o=r.body.getReader(),n;try{n=new ArrayBuffer(s)}catch(i){if(i instanceof RangeError){let l=Math.ceil(s/65536);n=new WebAssembly.Memory({initial:l,maximum:l}).buffer}else throw i}let a=0;for(;;){let{done:i,value:l}=await o.read();if(i)break;let c=l.byteLength;new Uint8Array(n,a,c).set(l),a+=c}return new Uint8Array(n,0,s)}}else return e instanceof Blob?new Uint8Array(await e.arrayBuffer()):e instanceof Uint8Array?e:new Uint8Array(e)}}),Ld,zd,Bd,Rd,Xi,jd,zt,Ns=Ue(()=>{Mt(),Ld=["V","I","W","E","F"],zd=(e,r)=>{console.log(`[${Ld[e]},${new Date().toISOString()}]${r}`)},Xi=(e,r)=>{Bd=e,Rd=r},jd=(e,r)=>{let t=oa(e),s=oa(Bd);t>=s&&zd(t,typeof r=="function"?r():r)},zt=(...e)=>{Rd&&jd(...e)}}),Nd,Gn,ye,aa,Vd,Ud,Wd,Pt=Ue(()=>{Nd=class{static calcMatMulShape(e,r){return e[1]!==r[0]?void 0:[e[0],r[1]]}},Gn=class{static calcShape(e,r,t=!1){let s=e.length,o=r.length;if(s===0)return r;if(o===0)return e;let n=Math.max(e.length,r.length),a=new Array(n);if(t){if(s<2||o<2)return;let i=Nd.calcMatMulShape([e[s-2],e[s-1]],[r[o-2],r[o-1]]);if(i===void 0)return;[a[n-2],a[n-1]]=i}for(let i=t?3:1;i<=n;i++){let l=s-i<0?1:e[s-i],c=o-i<0?1:r[o-i];if(l!==c&&l>1&&c>1)return;let p=Math.max(l,c);if(l&&c)a[n-i]=Math.max(l,c);else{if(p>1)return;a[n-i]=0}}return a}static isValidBroadcast(e,r){let t=e.length,s=r.length;if(t>s)return!1;for(let o=1;o<=t;o++)if(e[t-o]!==1&&e[t-o]!==r[s-o])return!1;return!0}},ye=class bi{static size(r){return bi.getSizeFromDimensionRange(r,0,r.length)}static convertShape(r,t=4){let s=r.length;if(s===0)return[];let o=new Array(s),n=s-1;for(;n>=0;){if(r[n]%t===0){o[n]=r[n]/t;break}if(t%r[n]!==0)throw new Error("cannot convert shape");o[n]=1,t/=r[n],n--}for(n--;n>=0;n--)o[n]=r[n];return o}static sizeFromDimension(r,t){if(t<0||t>r.length)throw new Error(`invalid dimension of ${t} for sizeFromDimension as Tensor has ${r.length} dimensions.`);return bi.getSizeFromDimensionRange(r,t,r.length)}static sizeToDimension(r,t){if(t<0||t>r.length)throw new Error(`invalid dimension of ${t} for sizeToDimension as Tensor has ${r.length} dimensions.`);return bi.getSizeFromDimensionRange(r,0,t)}static getSizeFromDimensionRange(r,t,s){let o=1;for(let n=t;n<s;n++){if(r[n]<0)throw new Error("cannot get valid size from specified dimension range. Most likely the range contains negative values in them.");o*=Number(r[n])}return o}static computeStrides(r){let t=r.length;if(t===0)return[];if(t===1)return[1];let s=new Array(t);s[t-1]=1,s[t-2]=r[t-1];for(let o=t-3;o>=0;--o)s[o]=s[o+1]*r[o+1];return s}static normalizeAxis(r,t){if(r<-t&&r>=t)throw new Error("unsupported axis for this operation.");return r<0?r+t:r}static normalizeAxes(r,t){return r.map(s=>this.normalizeAxis(s,t??r.length))}static sortBasedOnPerm(r,t){return t?t.map(s=>r[s]):r.slice().reverse()}static padShape(r,t){let s=r.length;return r.map((o,n)=>o+t[n]+t[n+s])}static areEqual(r,t){return r.length!==t.length?!1:r.every((s,o)=>s===t[o])}},aa=class Zo{static adjustPoolAttributes(r,t,s,o,n,a){if(!r&&s.length!==t.length-2)throw new Error("length of specified kernel shapes should be 2 less than length of input dimensions");if(r)for(let i=0;i<t.length-2;i++)i>=s.length?s.push(t[i+2]):s[i]=t[i+2];for(let i=0;i<s.length;i++)if(i<o.length){if(o[i]<0)throw new Error("strides should be greater than or equal to 1")}else o.push(1);for(let i=0;i<s.length;i++)if(i<n.length){if(n[i]<0)throw new Error("dilations should be greater than or equal to 1")}else n.push(1);for(let i=0;i<s.length*2;i++)if(i<a.length){if(a[i]<0)throw new Error("pad should be greater than or equal to 1")}else a.push(0);for(let i=0;i<s.length;i++){if(s[i]<=0)throw new Error("kernel shapes need to be greater than 0");if(a[i]>=s[i]||a[i+s.length]>=s[i])throw new Error("pads should be smaller than kernel")}}static adjustPadsBasedOnAutoPad(r,t,s,o,n,a,i){if(i){if(n.length!==2*(r.length-2))throw new Error("length of pads should be twice the length of data dimensions");if(t.length!==r.length-2)throw new Error("length of strides should be the length of data dimensions");if(o.length!==r.length-2)throw new Error("length of kernel shapes should be the length of data dimensions");for(let l=0;l<r.length-2;l++)Zo.adjustPadAndReturnShape(r[l+(a?1:2)],t[l],s[l],o[l],n,l,l+r.length-2,i)}}static computePoolOutputShape(r,t,s,o,n,a,i){if(t.length<=0)throw new Error("input shape must be of size greater than 0");let l=[t[0],t[1]];return Zo.computeShapeHelper(r,t,l,s,o,n,a,i),l}static computeConvOutputShape(r,t,s,o,n,a,i){if(r.length<=0||t.length<=0)throw new Error("invalid input tensor dims or invalid filter tensor dims");let l=[r[0],t[0]];return Zo.computeShapeHelper(!1,r,l,s,o,n,a,i),l}static computeShapeHelper(r,t,s,o,n,a,i,l){if(r)for(let c=0;c<t.length-2;c++)s.push(1);else for(let c=0;c<t.length-2;c++)s.push(Zo.adjustPadAndReturnShape(t[c+2],o[c],n[c],a[c],i,c,c+t.length-2,l))}static adjustPadAndReturnShape(r,t,s,o,n,a,i,l){let c=s*(o-1)+1;if(l&&l!=="NOTSET")switch(l){case"VALID":return n[a]=0,n[i]=0,Math.floor((r-c)/t+1);case"SAME_LOWER":case"SAME_UPPER":if(s!==1)throw new Error("Dilation not supported for SAME_UPPER or SAME_LOWER");{let p=((r+t-1)/t-1)*t+o-r;return n[a]=Math.floor(l==="SAME_LOWER"?(p+1)/2:p/2),n[i]=p-n[a],Math.floor((r+p-o)/t+1)}default:throw new Error("Unsupported AutoPad type")}else return Math.floor((r+n[a]+n[i]-c)/t+1)}},Vd=class{static getShapeOfGemmResult(e,r,t,s,o){if(e.length!==2||t.length!==2)throw new Error("shape need to be of size 2");let n,a,i;r?(n=e[1],a=e[0]):(n=e[0],a=e[1]);let l=-1;if(s?(i=t[0],l=1):(i=t[1],l=0),t[l]!==a)throw new Error("dimension mismatch");if(n<=0||i<=0||a<=0)throw new Error("invalid shape specified");if(o&&!Gn.isValidBroadcast(o,[n,i]))throw new Error("gemm: invalid bias shape for broadcast");return[n,i,a]}},Ud=-34028234663852886e22,Wd=34028234663852886e22}),Ji,Gd=Ue(()=>{Mt(),Ji=(e,r)=>new(Gi(r))(e)}),Yi,Zi,Kd,el,Hd,tl,rl,sl,qd,Qd,Fx=Ue(()=>{Ns(),Yi=(e,r=!0)=>{if(e.byteLength%8!==0)throw new Error("Invalid Uint8Array length - must be a multiple of 8 (BigInt).");let t=e.byteLength/8,s=new BigInt64Array(e.buffer,e.byteOffset,t),o=new Int32Array(t);for(let n=0;n<t;n++){let a=s[n];if(a>2147483647n||a<-2147483648n)throw new Error(`Overflow occurred when converting BigInt to Int32 at index ${n}: ${a}`);o[n]=Number(a)}return r?new Uint8Array(o.buffer):o},Zi=(e,r=!0)=>{if(e.byteLength%4!==0)throw new Error("Invalid Uint8Array length - must be a multiple of 4 (Int32).");let t=e.byteLength/4,s=new Int32Array(e.buffer,e.byteOffset,t),o=BigInt64Array.from(s,BigInt);return r?new Uint8Array(o.buffer):o},Kd=1,el=()=>Kd++,Hd=new Map([["float32",32],["float16",16],["int32",32],["uint32",32],["int64",64],["uint64",64],["int8",8],["uint8",8],["int4",4],["uint4",4]]),tl=(e,r)=>{let t=Hd.get(e);if(!t)throw new Error("Unsupported data type.");return r.length>0?Math.ceil(r.reduce((s,o)=>s*o)*t/8):0},rl=class{constructor(e){this.shouldConvertInt64toInt32=!1,this.isInt64ToInt32Converted=!1;let{sessionId:r,context:t,tensor:s,dataType:o,shape:n,shouldConvertInt64toInt32:a=!1}=e;this.sessionId=r,this.mlContext=t,this.mlTensor=s,this.dataType=o,this.tensorShape=n,this.shouldConvertInt64toInt32=a}get tensor(){return this.mlTensor}get type(){return this.dataType}get shape(){return this.tensorShape}get byteLength(){return tl(this.dataType,this.tensorShape)}destroy(){zt("verbose",()=>"[WebNN] TensorWrapper.destroy"),this.mlTensor.destroy()}write(e){this.mlContext.writeTensor(this.mlTensor,e)}async read(e,r){if(e){let t=await this.mlContext.readTensor(this.mlTensor),s=Zi(new Uint8Array(t));if(r){(r instanceof ArrayBuffer?new Uint8Array(r):new Uint8Array(r.buffer,r.byteOffset,r.byteLength)).set(s);return}else return s.buffer}else return r?this.mlContext.readTensor(this.mlTensor,r):this.mlContext.readTensor(this.mlTensor)}canReuseTensor(e,r,t){return this.mlContext===e&&this.dataType===r&&this.tensorShape.length===t.length&&this.tensorShape.every((s,o)=>s===t[o])}setIsInt64ToInt32Converted(e){this.isInt64ToInt32Converted=e}},sl=class{constructor(e,r){this.tensorManager=e,this.wrapper=r}get tensorWrapper(){return this.wrapper}releaseTensor(){this.tensorWrapper&&(this.tensorManager.releaseTensor(this.tensorWrapper),this.wrapper=void 0)}async ensureTensor(e,r,t,s){let o=r,n=this.tensorManager.getMLContext(e),a=o==="int64"&&!n.opSupportLimits().input.dataTypes.includes("int64");if(a&&(o="int32",zt("verbose",()=>"[WebNN] TensorIdTracker.ensureTensor: convert dataType from int64 to int32")),this.wrapper){if(this.wrapper.canReuseTensor(n,o,t))return this.wrapper.tensor;if(s){if(this.wrapper.byteLength!==tl(o,t))throw new Error("Unable to copy data to tensor with different size.");this.activeUpload=new Uint8Array(await this.wrapper.read())}this.tensorManager.releaseTensor(this.wrapper)}let i=typeof MLTensorUsage>"u"?void 0:MLTensorUsage.READ|MLTensorUsage.WRITE;return this.wrapper=await this.tensorManager.getCachedTensor(e,o,t,i,!0,!0,a),s&&this.activeUpload&&(this.wrapper.write(this.activeUpload),this.activeUpload=void 0),this.wrapper.tensor}upload(e){let r=e;if(this.wrapper)if(this.wrapper.shouldConvertInt64toInt32&&(r=Yi(e,!0),this.wrapper.setIsInt64ToInt32Converted(!0)),r.byteLength===this.wrapper.byteLength){this.wrapper.write(r);return}else zt("verbose",()=>"Data size does not match tensor size. Releasing tensor."),this.releaseTensor();this.activeUpload?this.activeUpload.set(r):this.activeUpload=new Uint8Array(r)}async download(e){var r,t,s;if(this.activeUpload){let o=(r=this.wrapper)!=null&&r.isInt64ToInt32Converted?Zi(this.activeUpload):this.activeUpload;if(e){e instanceof ArrayBuffer?new Uint8Array(e).set(o):new Uint8Array(e.buffer,e.byteOffset,e.byteLength).set(o);return}else return o.buffer}if(!this.wrapper)throw new Error("Tensor has not been created.");return e?this.wrapper.read((t=this.wrapper)==null?void 0:t.shouldConvertInt64toInt32,e):this.wrapper.read((s=this.wrapper)==null?void 0:s.shouldConvertInt64toInt32)}},qd=class{constructor(e){this.backend=e,this.tensorTrackersById=new Map,this.freeTensors=[],this.externalTensors=new Set}getMLContext(e){let r=this.backend.getMLContext(e);if(!r)throw new Error("MLContext not found for session.");return r}reserveTensorId(){let e=el();return this.tensorTrackersById.set(e,new sl(this)),e}releaseTensorId(e){let r=this.tensorTrackersById.get(e);r&&(this.tensorTrackersById.delete(e),r.tensorWrapper&&this.releaseTensor(r.tensorWrapper))}async ensureTensor(e,r,t,s,o){zt("verbose",()=>`[WebNN] TensorManager.ensureTensor {tensorId: ${r}, dataType: ${t}, shape: ${s}, copyOld: ${o}}`);let n=this.tensorTrackersById.get(r);if(!n)throw new Error("Tensor not found.");return n.ensureTensor(e,t,s,o)}upload(e,r){let t=this.tensorTrackersById.get(e);if(!t)throw new Error("Tensor not found.");t.upload(r)}async download(e,r){zt("verbose",()=>`[WebNN] TensorManager.download {tensorId: ${e}, dstBuffer: ${r==null?void 0:r.byteLength}}`);let t=this.tensorTrackersById.get(e);if(!t)throw new Error("Tensor not found.");return t.download(r)}releaseTensorsForSession(e){for(let r of this.freeTensors)r.sessionId===e&&r.destroy();this.freeTensors=this.freeTensors.filter(r=>r.sessionId!==e)}registerTensor(e,r,t,s){let o=this.getMLContext(e),n=el(),a=new rl({sessionId:e,context:o,tensor:r,dataType:t,shape:s});return this.tensorTrackersById.set(n,new sl(this,a)),this.externalTensors.add(a),n}async getCachedTensor(e,r,t,s,o,n,a=!1){let i=this.getMLContext(e);for(let[c,p]of this.freeTensors.entries())if(p.canReuseTensor(i,r,t)){zt("verbose",()=>`[WebNN] Reusing tensor {dataType: ${r}, shape: ${t}}`);let d=this.freeTensors.splice(c,1)[0];return d.sessionId=e,d}zt("verbose",()=>`[WebNN] MLContext.createTensor {dataType: ${r}, shape: ${t}}`);let l=await i.createTensor({dataType:r,shape:t,dimensions:t,usage:s,writable:o,readable:n});return new rl({sessionId:e,context:i,tensor:l,dataType:r,shape:t,shouldConvertInt64toInt32:a})}releaseTensor(e){this.externalTensors.has(e)&&this.externalTensors.delete(e),this.freeTensors.push(e)}},Qd=(...e)=>new qd(...e)}),ia,Xd,Jd,Ox=Ue(()=>{Mt(),_n(),Gd(),Fx(),Ns(),ia=new Map([[1,"float32"],[10,"float16"],[6,"int32"],[12,"uint32"],[7,"int64"],[13,"uint64"],[22,"int4"],[21,"uint4"],[3,"int8"],[2,"uint8"],[9,"uint8"]]),Xd=(e,r)=>{if(e===r)return!0;if(e===void 0||r===void 0)return!1;let t=Object.keys(e).sort(),s=Object.keys(r).sort();return t.length===s.length&&t.every((o,n)=>o===s[n]&&e[o]===r[o])},Jd=class{constructor(e){this.tensorManager=Qd(this),this.mlContextBySessionId=new Map,this.sessionIdsByMLContext=new Map,this.mlContextCache=[],this.sessionGraphInputs=new Map,this.temporaryGraphInputs=[],this.temporarySessionTensorIds=new Map,Xi(e.logLevel,!!e.debug)}get currentSessionId(){if(this.activeSessionId===void 0)throw new Error("No active session");return this.activeSessionId}onRunStart(e){zt("verbose",()=>`[WebNN] onRunStart {sessionId: ${e}}`),this.activeSessionId=e}onRunEnd(e){zt("verbose",()=>`[WebNN] onRunEnd {sessionId: ${e}}`);let r=this.temporarySessionTensorIds.get(e);if(r){for(let t of r)zt("verbose",()=>`[WebNN] releasing temporary tensor {tensorId: ${t}}`),this.tensorManager.releaseTensorId(t);this.temporarySessionTensorIds.delete(e),this.activeSessionId=void 0}}async createMLContext(e){if(e instanceof GPUDevice){let t=this.mlContextCache.findIndex(s=>s.gpuDevice===e);if(t!==-1)return this.mlContextCache[t].mlContext;{let s=await navigator.ml.createContext(e);return this.mlContextCache.push({gpuDevice:e,mlContext:s}),s}}else if(e===void 0){let t=this.mlContextCache.findIndex(s=>s.options===void 0&&s.gpuDevice===void 0);if(t!==-1)return this.mlContextCache[t].mlContext;{let s=await navigator.ml.createContext();return this.mlContextCache.push({mlContext:s}),s}}let r=this.mlContextCache.findIndex(t=>Xd(t.options,e));if(r!==-1)return this.mlContextCache[r].mlContext;{let t=await navigator.ml.createContext(e);return this.mlContextCache.push({options:e,mlContext:t}),t}}registerMLContext(e,r){this.mlContextBySessionId.set(e,r);let t=this.sessionIdsByMLContext.get(r);t||(t=new Set,this.sessionIdsByMLContext.set(r,t)),t.add(e),this.temporaryGraphInputs.length>0&&(this.sessionGraphInputs.set(e,this.temporaryGraphInputs),this.temporaryGraphInputs=[])}onReleaseSession(e){this.sessionGraphInputs.delete(e);let r=this.mlContextBySessionId.get(e);if(!r)return;this.tensorManager.releaseTensorsForSession(e),this.mlContextBySessionId.delete(e);let t=this.sessionIdsByMLContext.get(r);if(t.delete(e),t.size===0){this.sessionIdsByMLContext.delete(r);let s=this.mlContextCache.findIndex(o=>o.mlContext===r);s!==-1&&this.mlContextCache.splice(s,1)}}getMLContext(e){return this.mlContextBySessionId.get(e)}reserveTensorId(){return this.tensorManager.reserveTensorId()}releaseTensorId(e){zt("verbose",()=>`[WebNN] releaseTensorId {tensorId: ${e}}`),this.tensorManager.releaseTensorId(e)}async ensureTensor(e,r,t,s,o){let n=ia.get(t);if(!n)throw new Error(`Unsupported ONNX data type: ${t}`);return this.tensorManager.ensureTensor(e??this.currentSessionId,r,n,s,o)}async createTemporaryTensor(e,r,t){zt("verbose",()=>`[WebNN] createTemporaryTensor {onnxDataType: ${r}, shape: ${t}}`);let s=ia.get(r);if(!s)throw new Error(`Unsupported ONNX data type: ${r}`);let o=this.tensorManager.reserveTensorId();await this.tensorManager.ensureTensor(e,o,s,t,!1);let n=this.temporarySessionTensorIds.get(e);return n?n.push(o):this.temporarySessionTensorIds.set(e,[o]),o}uploadTensor(e,r){if(!Zt().shouldTransferToMLTensor)throw new Error("Trying to upload to a MLTensor while shouldTransferToMLTensor is false");zt("verbose",()=>`[WebNN] uploadTensor {tensorId: ${e}, data: ${r.byteLength}}`),this.tensorManager.upload(e,r)}async downloadTensor(e,r){return this.tensorManager.download(e,r)}createMLTensorDownloader(e,r){return async()=>{let t=await this.tensorManager.download(e);return Ji(t,r)}}registerMLTensor(e,r,t,s){let o=ia.get(t);if(!o)throw new Error(`Unsupported ONNX data type: ${t}`);let n=this.tensorManager.registerTensor(e,r,o,s);return zt("verbose",()=>`[WebNN] registerMLTensor {tensor: ${r}, dataType: ${o}, dimensions: ${s}} -> {tensorId: ${n}}`),n}registerMLConstant(e,r,t,s,o,n,a=!1){if(!n)throw new Error("External mounted files are not available.");let i=e;e.startsWith("./")&&(i=e.substring(2));let l=n.get(i);if(!l)throw new Error(`File with name ${i} not found in preloaded files.`);if(r+t>l.byteLength)throw new Error("Out of bounds: data offset and length exceed the external file data size.");let c=l.slice(r,r+t).buffer,p;switch(o.dataType){case"float32":p=new Float32Array(c);break;case"float16":p=typeof Float16Array<"u"&&Float16Array.from?new Float16Array(c):new Uint16Array(c);break;case"int32":p=new Int32Array(c);break;case"uint32":p=new Uint32Array(c);break;case"int64":a?(p=Yi(new Uint8Array(c),!1),o.dataType="int32"):p=new BigInt64Array(c);break;case"uint64":p=new BigUint64Array(c);break;case"int8":p=new Int8Array(c);break;case"int4":case"uint4":case"uint8":p=new Uint8Array(c);break;default:throw new Error(`Unsupported data type: ${o.dataType} in creating WebNN Constant from external data.`)}return zt("verbose",()=>`[WebNN] registerMLConstant {dataType: ${o.dataType}, shape: ${o.shape}}} ${a?"(Note: it was int64 data type and registered to int32 as workaround)":""}`),s.constant(o,p)}registerGraphInput(e){this.temporaryGraphInputs.push(e)}isGraphInput(e,r){let t=this.sessionGraphInputs.get(e);return t?t.includes(r):!1}isInt64Supported(e){var r;return!!((r=this.mlContextBySessionId.get(e))!=null&&r.opSupportLimits().input.dataTypes.includes("int64"))}flush(){}}}),nl=Ue(()=>{}),ol,la,ca,Yd,Zd,al,il,ep,tp,Dx=Ue(()=>{Ns(),nl(),ol=new Map([[64,250],[128,200],[256,200],[512,200],[2048,230],[4096,200],[8192,50],[16384,50],[32768,50],[65536,50],[131072,50],[262144,50],[524288,50],[1048576,50],[2097152,30],[4194304,20],[8388608,10],[12582912,10],[16777216,10],[26214400,15],[33554432,22],[44236800,2],[58982400,6],[67108864,6],[134217728,6],[167772160,6]]),la=[],ca=e=>Math.ceil(Number(e)/16)*16,Yd=e=>{for(let r=0;r<la.length;r++){let t=la[r];if(e<=t)return t}return Math.ceil(e/16)*16},Zd=1,al=()=>Zd++,il=async(e,r,t,s)=>{let o=ca(t),n=e.device.createBuffer({size:o,usage:GPUBufferUsage.COPY_DST|GPUBufferUsage.MAP_READ});try{let a=e.getCommandEncoder();e.endComputePass(),a.copyBufferToBuffer(r,0,n,0,o),e.flush(),await n.mapAsync(GPUMapMode.READ);let i=n.getMappedRange();if(s){let l=s();return l.set(new Uint8Array(i,0,t)),l}else return new Uint8Array(i.slice(0,t))}finally{n.destroy()}},ep=class{constructor(e){this.backend=e,this.storageCache=new Map,this.freeBuffers=new Map,this.freeUniformBuffers=new Map,this.buffersPending=[],this.capturedPendingBuffers=new Map;for(let[r]of ol)la.push(r),this.freeBuffers.set(r,[]),this.freeUniformBuffers.set(r,[]);this.sessionCount=0}upload(e,r){let t=r.buffer,s=r.byteOffset,o=r.byteLength,n=ca(o),a=this.storageCache.get(e);if(!a)throw new Error("gpu data for uploading does not exist");if(Number(a.originalSize)!==o)throw new Error(`inconsistent data size. gpu data size=${a.originalSize}, data size=${o}`);let i=this.backend.device.createBuffer({mappedAtCreation:!0,size:n,usage:GPUBufferUsage.MAP_WRITE|GPUBufferUsage.COPY_SRC}),l=i.getMappedRange();new Uint8Array(l).set(new Uint8Array(t,s,o)),i.unmap();let c=this.backend.device.createCommandEncoder();c.copyBufferToBuffer(i,0,a.gpuData.buffer,0,n),this.backend.device.queue.submit([c.finish()]),i.destroy(),zt("verbose",()=>`[WebGPU] GpuDataManager.upload(id=${e})`)}memcpy(e,r){let t=this.storageCache.get(e);if(!t)throw new Error("source gpu data for memcpy does not exist");let s=this.storageCache.get(r);if(!s)throw new Error("destination gpu data for memcpy does not exist");if(t.originalSize!==s.originalSize)throw new Error("inconsistent source and destination gpu data size");let o=ca(t.originalSize),n=this.backend.getCommandEncoder();this.backend.endComputePass(),n.copyBufferToBuffer(t.gpuData.buffer,0,s.gpuData.buffer,0,o)}registerExternalBuffer(e,r,t){let s;if(t){if(s=t[0],e===t[1])return zt("verbose",()=>`[WebGPU] GpuDataManager.registerExternalBuffer(size=${r}) => id=${s}, buffer is the same, skip.`),s;if(this.backend.capturedCommandList.has(this.backend.currentSessionId))throw new Error(`Registering a different external buffer under graph capture mode is not supported yet.
             Please use the previous external buffer!`)}else s=al();return this.storageCache.set(s,{gpuData:{id:s,type:0,buffer:e},originalSize:r}),zt("verbose",()=>`[WebGPU] GpuDataManager.registerExternalBuffer(size=${r}) => id=${s}, registered.`),s}unregisterExternalBuffer(e){e!==void 0&&(this.storageCache.delete(e),zt("verbose",()=>`[WebGPU] GpuDataManager.unregisterExternalBuffer() => id=${e}`))}create(e,r=GPUBufferUsage.STORAGE|GPUBufferUsage.COPY_SRC|GPUBufferUsage.COPY_DST){let t=Yd(e),s,o=(r&GPUBufferUsage.STORAGE)===GPUBufferUsage.STORAGE,n=(r&GPUBufferUsage.UNIFORM)===GPUBufferUsage.UNIFORM;if(o||n){let i=(o?this.freeBuffers:this.freeUniformBuffers).get(t);i?i.length>0?s=i.pop():s=this.backend.device.createBuffer({size:t,usage:r}):s=this.backend.device.createBuffer({size:t,usage:r})}else s=this.backend.device.createBuffer({size:t,usage:r});let a={id:al(),type:0,buffer:s};return this.storageCache.set(a.id,{gpuData:a,originalSize:Number(e)}),zt("verbose",()=>`[WebGPU] GpuDataManager.create(size=${e}) => id=${a.id}`),a}get(e){var r;return(r=this.storageCache.get(e))==null?void 0:r.gpuData}release(e){let r=typeof e=="bigint"?Number(e):e,t=this.storageCache.get(r);if(!t){if(this.storageCache.size===0)return 0;throw new Error("releasing data does not exist")}return zt("verbose",()=>`[WebGPU] GpuDataManager.release(id=${r}), gpuDataId=${t.gpuData.id}`),this.storageCache.delete(r),this.buffersPending.push(t.gpuData.buffer),t.originalSize}async download(e,r){let t=this.storageCache.get(Number(e));if(!t)throw new Error("data does not exist");await il(this.backend,t.gpuData.buffer,t.originalSize,r)}refreshPendingBuffers(){if(this.buffersPending.length!==0)if(this.backend.sessionStatus==="default"){for(let e of this.buffersPending){let r=ol.get(e.size);if((e.usage&GPUBufferUsage.STORAGE)===GPUBufferUsage.STORAGE){let t=this.freeBuffers.get(e.size)||[];r===void 0||t.length>=r?e.destroy():t.push(e)}else if((e.usage&GPUBufferUsage.UNIFORM)===GPUBufferUsage.UNIFORM){let t=this.freeUniformBuffers.get(e.size)||[];r===void 0||t.length>=r?e.destroy():t.push(e)}else e.destroy()}this.buffersPending=[]}else{let e=this.capturedPendingBuffers.get(this.backend.currentSessionId);e||(e=[],this.capturedPendingBuffers.set(this.backend.currentSessionId,e));for(let r of this.buffersPending)e.push(r);this.buffersPending=[]}}dispose(){this.freeBuffers.forEach(e=>{e.forEach(r=>{r.destroy()})}),this.freeUniformBuffers.forEach(e=>{e.forEach(r=>{r.destroy()})}),this.storageCache.forEach(e=>{e.gpuData.buffer.destroy()}),this.capturedPendingBuffers.forEach(e=>{e.forEach(r=>{r.destroy()})}),this.storageCache=new Map,this.freeBuffers=new Map,this.freeUniformBuffers=new Map,this.capturedPendingBuffers=new Map}onCreateSession(){this.sessionCount+=1}onReleaseSession(e){let r=this.capturedPendingBuffers.get(e);r&&(r.forEach(t=>{t.destroy()}),this.capturedPendingBuffers.delete(e)),this.sessionCount-=1,this.sessionCount===0&&(zt("warning",()=>"[WebGPU] Clearing webgpu buffer cache"),this.storageCache.forEach(t=>{t.gpuData.buffer.destroy()}),this.storageCache=new Map)}},tp=(...e)=>new ep(...e)}),rp,Kt,pr=Ue(()=>{rp=class{constructor(e){Object.assign(this,e)}get cacheKey(){return this.key||(this.key=Object.getOwnPropertyNames(this).sort().map(e=>`${this[e]}`).join(";")),this.key}},Kt=e=>new rp(e)}),Kn,ua,Ir,Vr,dt,lr,ll,Hn,en,ct,go,Ie,nt,sp,cl,np,op,Ct=Ue(()=>{Mt(),Pt(),Kn=64,ua=(e,r)=>{if(r===3)throw new Error("vec3 has same alignment as vec4, use vec4 instead");switch(Number(e)){case 10:return r>1?`vec${r}<f16>`:"f16";case 1:return r>1?`vec${r}<f32>`:"f32";case 6:return r>1?`vec${r}<i32>`:"i32";case 12:return r>1?`vec${r}<u32>`:"u32";case 7:if(r>1)throw new Error("currently not supported vecX of uint64 yet");return["vec2<u32>","i32"];case 13:if(r>1)throw new Error("currently not supported vecX of uint64 yet");return["vec2<u32>","u32"];case 9:if(r!==4)throw new Error("bool must be vec4");return["u32","vec4<bool>"];case 22:return"i32";case 21:return"u32";default:throw new Error(`Unknown data type: ${e}`)}},Ir=(e,r=1)=>{let t=ua(e,r);return typeof t=="string"?t:t[0]},Vr=(e,r=1)=>{let t=ua(e,r);return typeof t=="string"?t:t[1]},dt=(...e)=>{let r=[];return e.forEach(t=>{t.length!==0&&r.push({type:12,data:t},{type:12,data:ye.computeStrides(t)})}),r},lr=e=>e%4===0?4:e%2===0?2:1,ll=(e="f32",r,t="0")=>!r||r===1?`${e}(${t})`:`vec${r}<${e}>(${t})`,Hn=(e,r,t)=>e==="f32"?t:r===1?`f32(${t})`:`vec${r}<f32>(${t})`,en=(e,r)=>r===4?`(${e}.x + ${e}.y + ${e}.z + ${e}.w)`:r===2?`(${e}.x + ${e}.y)`:r===3?`(${e}.x + ${e}.y + ${e}.z)`:e,ct=(e,r,t,s)=>e.startsWith("uniforms.")&&t>4?typeof r=="string"?s==="f16"?`${e}[(${r}) / 8][(${r}) % 8 / 4][(${r}) % 8 % 4]`:`${e}[(${r}) / 4][(${r}) % 4]`:s==="f16"?`${e}[${Math.floor(r/8)}][${Math.floor(r%8/4)}][${r%8%4}]`:`${e}[${Math.floor(r/4)}][${r%4}]`:t>1?`${e}[${r}]`:e,go=(e,r,t,s,o)=>{let n=typeof t=="number",a=n?t:t.length,i=[...new Array(a).keys()],l=a<2?"u32":a<=4?`vec${a}<u32>`:`array<u32, ${a}>`,c=ua(r,o),p=typeof c=="string"?c:c[1],d=typeof c=="string"?c:c[0],u={indices:l,value:p,storage:d,tensor:r},f=oe=>typeof oe=="string"?oe:`${oe}u`,_={offsetToIndices:!1,indicesToOffset:!1,broadcastedIndicesToOffset:!1,set:!1,setByIndices:!1,get:!1,getByIndices:!1},b=n?"uniforms.":"",k=`${b}${e}_shape`,w=`${b}${e}_strides`,y="";for(let oe=0;oe<a-1;oe++)y+=`
    let dim${oe} = current / ${ct(w,oe,a)};
    let rest${oe} = current % ${ct(w,oe,a)};
    indices[${oe}] = dim${oe};
    current = rest${oe};
    `;y+=`indices[${a-1}] = current;`;let $=a<2?"":`
  fn o2i_${e}(offset: u32) -> ${u.indices} {
    var indices: ${u.indices};
    var current = offset;
    ${y}
    return indices;
  }`,T=oe=>(_.offsetToIndices=!0,a<2?oe:`o2i_${e}(${oe})`),M=[];if(a>=2)for(let oe=a-1;oe>=0;oe--)M.push(`${ct(w,oe,a)} * (indices[${oe}])`);let x=a<2?"":`
  fn i2o_${e}(indices: ${u.indices}) -> u32 {
    return ${M.join("+")};
  }`,v=oe=>(_.indicesToOffset=!0,a<2?oe:`i2o_${e}(${oe})`),C=(...oe)=>a===0?"0u":`${u.indices}(${oe.map(f).join(",")})`,F=(oe,G)=>a<2?`${oe}`:`${ct(oe,G,a)}`,D=(oe,G,B)=>a<2?`${oe}=${B};`:`${ct(oe,G,a)}=${B};`,K={},U=(oe,G)=>{_.broadcastedIndicesToOffset=!0;let B=`${G.name}broadcastedIndicesTo${e}Offset`;if(B in K)return`${B}(${oe})`;let Q=[];for(let le=a-1;le>=0;le--){let ee=G.indicesGet("outputIndices",le+G.rank-a);Q.push(`${F(w,le)} * (${ee} % ${F(k,le)})`)}return K[B]=`fn ${B}(outputIndices: ${G.type.indices}) -> u32 {
             return ${Q.length>0?Q.join("+"):"0u"};
           }`,`${B}(${oe})`},j=(oe,G)=>(()=>{if(u.storage===u.value)return`${e}[${oe}]=${G};`;if(u.storage==="vec2<u32>"&&u.value==="i32")return`${e}[${oe}]=vec2<u32>(u32(${G}), select(0u, 0xFFFFFFFFu, ${G} < 0));`;if(u.storage==="vec2<u32>"&&u.value==="u32")return`${e}[${oe}]=vec2<u32>(u32(${G}), 0u);`;if(u.storage==="u32"&&u.value==="vec4<bool>")return`${e}[${oe}]=dot(vec4<u32>(0x1, 0x100, 0x10000, 0x1000000), vec4<u32>(${G}));`;throw new Error(`not supported combination of storage type ${u.storage} and value type ${u.value} yet`)})(),se=oe=>(()=>{if(u.storage===u.value)return`${e}[${oe}]`;if(u.storage==="vec2<u32>"&&u.value==="i32")return`i32(${e}[${oe}].x)`;if(u.storage==="vec2<u32>"&&u.value==="u32")return`u32(${e}[${oe}].x)`;if(u.storage==="u32"&&u.value==="vec4<bool>")return`vec4<bool>(bool(${e}[${oe}] & 0xFFu), bool(${e}[${oe}] & 0xFF00u), bool(${e}[${oe}] & 0xFF0000u), bool(${e}[${oe}] & 0xFF000000u))`;throw new Error(`not supported combination of storage type ${u.storage} and value type ${u.value} yet`)})(),q=a<2?"":`
  fn get_${e}ByIndices(indices: ${u.indices}) -> ${p} {
    return ${se(`i2o_${e}(indices)`)};
  }`,re=a<2?"":(()=>{let oe=i.map(B=>`d${B}: u32`).join(", "),G=i.map(B=>`d${B}`).join(", ");return`
  fn get_${e}(${oe}) -> ${p} {
    return get_${e}ByIndices(${C(G)});
  }`})(),Z=(...oe)=>{if(oe.length!==a)throw new Error(`indices length must be ${a}`);let G=oe.map(f).join(",");return a===0?se("0u"):a===1?se(G[0]):(_.get=!0,_.getByIndices=!0,_.indicesToOffset=!0,`get_${e}(${G})`)},ae=oe=>a<2?se(oe):(_.getByIndices=!0,_.indicesToOffset=!0,`get_${e}ByIndices(${oe})`),_e=a<2?"":`
  fn set_${e}ByIndices(indices: ${u.indices}, value: ${p}) {
    ${j(`i2o_${e}(indices)`,"value")}
  }`,pe=a<2?"":(()=>{let oe=i.map(B=>`d${B}: u32`).join(", "),G=i.map(B=>`d${B}`).join(", ");return`
  fn set_${e}(${oe}, value: ${p}) {
    set_${e}ByIndices(${C(G)}, value);
  }`})();return{impl:()=>{let oe=[],G=!1;return _.offsetToIndices&&(oe.push($),G=!0),_.indicesToOffset&&(oe.push(x),G=!0),_.broadcastedIndicesToOffset&&(Object.values(K).forEach(B=>oe.push(B)),G=!0),_.set&&(oe.push(pe),G=!0),_.setByIndices&&(oe.push(_e),G=!0),_.get&&(oe.push(re),G=!0),_.getByIndices&&(oe.push(q),G=!0),!n&&G&&oe.unshift(`const ${k} = ${u.indices}(${t.join(",")});`,`const ${w} = ${u.indices}(${ye.computeStrides(t).join(",")});`),oe.join(`
`)},type:u,offsetToIndices:T,indicesToOffset:v,broadcastedIndicesToOffset:U,indices:C,indicesGet:F,indicesSet:D,set:(...oe)=>{if(oe.length!==a+1)throw new Error(`indices length must be ${a}`);let G=oe[a];if(typeof G!="string")throw new Error("value must be string");let B=oe.slice(0,a).map(f).join(",");return a===0?j("0u",G):a===1?j(B[0],G):(_.set=!0,_.setByIndices=!0,_.indicesToOffset=!0,`set_${e}(${B}, ${G})`)},setByOffset:j,setByIndices:(oe,G)=>a<2?j(oe,G):(_.setByIndices=!0,_.indicesToOffset=!0,`set_${e}ByIndices(${oe}, ${G});`),get:Z,getByOffset:se,getByIndices:ae,usage:s,name:e,strides:w,shape:k,rank:a}},Ie=(e,r,t,s=1)=>go(e,r,t,"input",s),nt=(e,r,t,s=1)=>go(e,r,t,"output",s),sp=(e,r,t)=>go(e,r,t,"atomicOutput",1),cl=(e,r,t,s=1)=>go(e,r,t,"internal",s),np=class{constructor(e,r){this.normalizedDispatchGroup=e,this.limits=r,this.internalVariables=[],this.variables=[],this.uniforms=[],this.variableIndex=0}guardAgainstOutOfBoundsWorkgroupSizes(e){return`if (global_idx >= ${typeof e=="number"?`${e}u`:e}) { return; }`}mainStart(e=Kn){let r=typeof e=="number"?e:e[0],t=typeof e=="number"?1:e[1],s=typeof e=="number"?1:e[2];if(r>this.limits.maxComputeWorkgroupSizeX||t>this.limits.maxComputeWorkgroupSizeY||s>this.limits.maxComputeWorkgroupSizeZ)throw new Error(`workgroup size [${r}, ${t}, ${s}] exceeds the maximum workgroup size [${this.limits.maxComputeWorkgroupSizeX}, ${this.limits.maxComputeWorkgroupSizeY}, ${this.limits.maxComputeWorkgroupSizeZ}].`);if(r*t*s>this.limits.maxComputeInvocationsPerWorkgroup)throw new Error(`workgroup size [${r}, ${t}, ${s}] exceeds the maximum workgroup invocations ${this.limits.maxComputeInvocationsPerWorkgroup}.`);let o=this.normalizedDispatchGroup[1]===1&&this.normalizedDispatchGroup[2]===1,n=o?`@builtin(global_invocation_id) global_id : vec3<u32>,
    @builtin(workgroup_id) workgroup_id : vec3<u32>,
    @builtin(local_invocation_index) local_idx : u32,
    @builtin(local_invocation_id) local_id : vec3<u32>`:`@builtin(global_invocation_id) global_id : vec3<u32>,
                                             @builtin(local_invocation_id) local_id : vec3<u32>,
    @builtin(local_invocation_index) local_idx : u32,
    @builtin(workgroup_id) workgroup_id : vec3<u32>,
    @builtin(num_workgroups) num_workgroups : vec3<u32>`,a=o?`let global_idx = global_id.x;
         let workgroup_index = workgroup_id.x;`:`let workgroup_index = workgroup_id.z * num_workgroups[0] * num_workgroups[1] +
             workgroup_id.y * num_workgroups[0] + workgroup_id.x;
         let global_idx = workgroup_index * ${r*t*s}u + local_idx;`;return`@compute @workgroup_size(${r}, ${t}, ${s})
  fn main(${n}) {
    ${a}
  `}appendVariableUniforms(e){e.rank!==0&&(e.shape.startsWith("uniforms.")&&this.uniforms.push({name:e.shape.replace("uniforms.",""),type:"u32",length:e.rank}),e.strides.startsWith("uniforms.")&&this.uniforms.push({name:e.strides.replace("uniforms.",""),type:"u32",length:e.rank}))}declareVariable(e,r){if(e.usage==="internal")throw new Error("cannot use internal variable with declareVariable(). use registerInternalVariables() instead.");this.variables.push(e),this.appendVariableUniforms(e);let t=e.usage==="input"?"read":"read_write",s=e.usage==="atomicOutput"?"atomic<i32>":e.type.storage;return`@group(0) @binding(${r}) var<storage, ${t}> ${e.name}: array<${s}>;`}declareVariables(...e){return e.map(r=>this.declareVariable(r,this.variableIndex++)).join(`
`)}registerInternalVariable(e){if(e.usage!=="internal")throw new Error("cannot use input or output variable with registerInternalVariable(). use declareVariables() instead.");this.internalVariables.push(e),this.appendVariableUniforms(e)}registerInternalVariables(...e){return e.forEach(r=>this.registerInternalVariable(r)),this}registerUniform(e,r,t=1){return this.uniforms.push({name:e,type:r,length:t}),this}registerUniforms(e){return this.uniforms=this.uniforms.concat(e),this}uniformDeclaration(){if(this.uniforms.length===0)return"";let e=[];for(let{name:r,type:t,length:s}of this.uniforms)if(s&&s>4)t==="f16"?e.push(`@align(16) ${r}:array<mat2x4<${t}>, ${Math.ceil(s/8)}>`):e.push(`${r}:array<vec4<${t}>, ${Math.ceil(s/4)}>`);else{let o=s==null||s===1?t:`vec${s}<${t}>`;e.push(`${r}:${o}`)}return`
      struct Uniforms { ${e.join(", ")} };
      @group(0) @binding(${this.variableIndex}) var<uniform> uniforms: Uniforms;`}get additionalImplementations(){return this.uniformDeclaration()+this.variables.map(e=>e.impl()).join(`
`)+this.internalVariables.map(e=>e.impl()).join(`
`)}get variablesInfo(){if(this.uniforms.length===0)return;let e=r=>[12,10,1,6][["u32","f16","f32","i32"].indexOf(r)];return this.uniforms.map(r=>[e(r.type),r.length??1])}},op=(e,r)=>new np(e,r)}),ap,ul,ip,lp,cp,up,rs,dp,pp,tn=Ue(()=>{Mt(),Pt(),pr(),Ct(),ap=(e,r)=>{if(!e||e.length!==1)throw new Error("Transpose requires 1 input.");if(r.length!==0&&r.length!==e[0].dims.length)throw new Error(`perm size ${r.length} does not match input rank ${e[0].dims.length}`)},ul=(e,r)=>r.length!==0?r:[...new Array(e).keys()].reverse(),ip=(e,r)=>ye.sortBasedOnPerm(e,ul(e.length,r)),lp=(e,r,t,s)=>{let o=`fn perm(i: ${s.type.indices}) -> ${t.type.indices} {
    var a: ${t.type.indices};`;for(let n=0;n<r;++n)o+=`a[${e[n]}]=i[${n}];`;return o+="return a;}"},cp=(e,r)=>{let t=[],s=[];for(let o=0;o<e.length;++o)e[o]!==1&&t.push(e[o]),e[r[o]]!==1&&s.push(r[o]);return{newShape:t,newPerm:s}},up=(e,r)=>{let t=0;for(let s=0;s<e.length;++s)if(r[e[s]]!==1){if(e[s]<t)return!1;t=e[s]}return!0},rs=(e,r)=>{let t=e.dataType,s=e.dims.length,o=ul(s,r),n=ip(e.dims,o),a=e.dims,i=n,l=s<2||up(o,e.dims),c;if(l)return c=_=>{let b=Ie("input",t,a,4),k=nt("output",t,i,4);return`
  ${_.registerUniform("output_size","u32").declareVariables(b,k)}
  ${_.mainStart()}
    ${_.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.output_size")}
    output[global_idx] = input[global_idx];
  }`},{name:"TransposeCopy",shaderCache:{inputDependencies:["type"]},getRunData:()=>{let _=ye.size(n);return{outputs:[{dims:n,dataType:e.dataType}],dispatchGroup:{x:Math.ceil(_/64/4)},programUniforms:[{type:12,data:Math.ceil(_/4)}]}},getShaderSource:c};let{newShape:p,newPerm:d}=cp(e.dims,o),u=ye.areEqual(d,[2,3,1]),f=ye.areEqual(d,[3,1,2]);if(p.length===2||u||f){a=u?[p[0],p[1]*p[2]]:f?[p[0]*p[1],p[2]]:p,i=[a[1],a[0]];let _=16;return c=b=>{let k=Ie("a",t,a.length),w=nt("output",t,i.length);return`
  ${b.registerUniform("output_size","u32").declareVariables(k,w)}
  var<workgroup> tile : array<array<${w.type.value}, ${_+1}>, ${_}>;
  ${b.mainStart([_,_,1])}
    let stride = (uniforms.output_shape[1] - 1) / ${_} + 1;
    let workgroup_id_x = workgroup_index % stride;
    let workgroup_id_y = workgroup_index / stride;
    let input_col = workgroup_id_y * ${_}u + local_id.x;
    let input_row = workgroup_id_x * ${_}u + local_id.y;
    if (input_row < uniforms.a_shape[0] && input_col < uniforms.a_shape[1]) {
      tile[local_id.y][local_id.x] = ${k.getByIndices(`${k.type.indices}(input_row, input_col)`)};
    }
    workgroupBarrier();

    let output_col = workgroup_id_x * ${_}u + local_id.x;
    let output_row = workgroup_id_y * ${_}u + local_id.y;
    if (output_row < uniforms.output_shape[0] && output_col < uniforms.output_shape[1]) {
      ${w.setByIndices(`${w.type.indices}(output_row, output_col)`,"tile[local_id.x][local_id.y]")}
    }
  }`},{name:"TransposeShared",shaderCache:{inputDependencies:["type"]},getRunData:()=>{let b=ye.size(n);return{outputs:[{dims:n,dataType:e.dataType}],dispatchGroup:{x:Math.ceil(i[1]/_),y:Math.ceil(i[0]/_)},programUniforms:[{type:12,data:b},...dt(a,i)]}},getShaderSource:c}}return c=_=>{let b=Ie("a",t,a.length),k=nt("output",t,i.length);return`
  ${_.registerUniform("output_size","u32").declareVariables(b,k)}

  ${lp(o,s,b,k)}

  ${_.mainStart()}
    ${_.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.output_size")}

    let indices = ${k.offsetToIndices("global_idx")};
    let aIndices = perm(indices);

    ${k.setByOffset("global_idx",b.getByIndices("aIndices"))}
  }`},{name:"Transpose",shaderCache:{hint:`${r}`,inputDependencies:["rank"]},getRunData:()=>{let _=ye.size(n);return{outputs:[{dims:n,dataType:e.dataType}],dispatchGroup:{x:Math.ceil(_/64)},programUniforms:[{type:12,data:_},...dt(a,i)]}},getShaderSource:c}},dp=(e,r)=>{ap(e.inputs,r.perm),e.compute(rs(e.inputs[0],r.perm))},pp=e=>Kt({perm:e.perm})}),mp,hp,_p,fp,gp,wp,Mp,bp,yp,vp,ws,xp,Tp,Ep,Pp,Cp,Sp,$p,kp,Ip,Ap,Lx=Ue(()=>{Mt(),Pt(),Ct(),pl(),tn(),mp={max:"select(bestValue, candidate, candidate > bestValue)",min:"select(bestValue, candidate, candidate < bestValue)",mean:"bestValue + candidate",sum:"bestValue + candidate",prod:"bestValue * candidate",sumSquare:"bestValue + candidate * candidate",logSumExp:"bestValue + exp(candidate)",l1:"bestValue + abs(candidate)",l2:"bestValue + candidate * candidate",logSum:"bestValue + candidate"},hp={max:"select(bestValue, candidate, candidate > bestValue)",min:"select(bestValue, candidate, candidate < bestValue)",mean:"bestValue + candidate",sum:"bestValue + candidate",prod:"bestValue * candidate",sumSquare:"bestValue + candidate",logSumExp:"bestValue + candidate",l1:"bestValue + candidate",l2:"bestValue + candidate",logSum:"bestValue + candidate"},_p={max:"_A[offset]",min:"_A[offset]",mean:"0",sum:"0",prod:"1",sumSquare:"0",logSumExp:"0",l1:"0",l2:"0",logSum:"0"},fp={max:"bestValue",min:"bestValue",sum:"bestValue",prod:"bestValue",sumSquare:"bestValue",logSumExp:"log(bestValue)",l1:"bestValue",l2:"sqrt(bestValue)",logSum:"log(bestValue)"},gp=(e,r)=>{let t=[];for(let s=r-e;s<r;++s)t.push(s);return t},wp=(e,r)=>{let t=[],s=e.length;for(let n=0;n<s;n++)r.indexOf(n)===-1&&t.push(e[n]);let o=r.map(n=>e[n]);return[t,o]},Mp=(e,r)=>{let t=e.length+r.length,s=[],o=0;for(let n=0;n<t;n++)r.indexOf(n)===-1?s.push(e[o++]):s.push(1);return s},bp=(e,r)=>{for(let t=0;t<e.length;++t)if(e[e.length-t-1]!==r-1-t)return!1;return!0},yp=(e,r)=>{let t=[];if(!bp(e,r)){for(let s=0;s<r;++s)e.indexOf(s)===-1&&t.push(s);e.forEach(s=>t.push(s))}return t},vp=(e,r,t,s,o,n,a)=>{let i=t[0].dims,l=ye.size(n),c=ye.size(a),p=Ie("_A",t[0].dataType,i),d=nt("output",o,n),u=64;l===1&&(u=256);let f=`
          var<workgroup> aBestValues : array<f32, ${u}>;
       `,_=b=>`
        ${b.registerUniform("reduceSize","u32").declareVariables(p,d)}
        ${f}
        fn DIV_CEIL(a : u32, b : u32) -> u32 {
          return ((a - 1u) / b + 1u);
         }
         ${b.mainStart(u)}

          let outputIndex = global_idx / ${u};
          let offset = outputIndex * uniforms.reduceSize;

          var bestValue = f32(${_p[s]});
          let Length = uniforms.reduceSize;
          for (var k = local_idx; k < Length; k = k + ${u}) {
           let candidate = f32(${p.getByOffset("offset + k")});
           bestValue = ${mp[s]};
          }
          aBestValues[local_idx] = bestValue;
          workgroupBarrier();

         var reduceSize = min(Length, ${u}u);
         for (var currentSize = reduceSize / 2u; reduceSize > 1u;
             currentSize = reduceSize / 2u) {
           let interval = DIV_CEIL(reduceSize, 2u);
           if (local_idx < currentSize) {
            let candidate = aBestValues[local_idx + interval];
            bestValue = ${hp[s]};
            aBestValues[local_idx] = bestValue;
           }
           reduceSize = interval;
           workgroupBarrier();
         }

         if (local_idx == 0u) {
          ${d.setByOffset("outputIndex",`${s==="mean"?`${d.type.storage}(bestValue / f32(uniforms.reduceSize))`:`${d.type.storage}(${fp[s]})`}`)};
         }
        }`;return{name:e,shaderCache:{hint:`${r};${u}`,inputDependencies:["type"]},getShaderSource:_,getRunData:()=>({outputs:[{dims:n,dataType:o}],dispatchGroup:{x:l},programUniforms:[{type:12,data:c}]})}},ws=(e,r,t,s)=>{let o=e.inputs.length===1?t:dl(e.inputs,t),n=o.axes;n.length===0&&!o.noopWithEmptyAxes&&(n=e.inputs[0].dims.map((f,_)=>_));let a=ye.normalizeAxes(n,e.inputs[0].dims.length),i=a,l=e.inputs[0],c=yp(i,e.inputs[0].dims.length);c.length>0&&(l=e.compute(rs(e.inputs[0],c),{inputs:[0],outputs:[-1]})[0],i=gp(i.length,l.dims.length));let[p,d]=wp(l.dims,i),u=p;o.keepDims&&(u=Mp(p,a)),e.compute(vp(r,o.cacheKey,[l],s,e.inputs[0].dataType,u,d),{inputs:[l]})},xp=(e,r)=>{ws(e,"ReduceMeanShared",r,"mean")},Tp=(e,r)=>{ws(e,"ReduceL1Shared",r,"l1")},Ep=(e,r)=>{ws(e,"ReduceL2Shared",r,"l2")},Pp=(e,r)=>{ws(e,"ReduceLogSumExpShared",r,"logSumExp")},Cp=(e,r)=>{ws(e,"ReduceMaxShared",r,"max")},Sp=(e,r)=>{ws(e,"ReduceMinShared",r,"min")},$p=(e,r)=>{ws(e,"ReduceProdShared",r,"prod")},kp=(e,r)=>{ws(e,"ReduceSumShared",r,"sum")},Ip=(e,r)=>{ws(e,"ReduceSumSquareShared",r,"sumSquare")},Ap=(e,r)=>{ws(e,"ReduceLogSumShared",r,"logSum")}}),Ms,Fp,da,dl,bs,Op,Dp,Lp,zp,Bp,Rp,jp,Np,Vp,Up,ys,Wp,Gp,Kp,Hp,qp,Qp,Xp,Jp,Yp,Zp,pl=Ue(()=>{Mt(),Pt(),pr(),Ct(),Lx(),Ms=e=>{if(!e||e.length===0||e.length>2)throw new Error("Reduce op requires 1 or 2 inputs.");if(e.length===2&&e[1].dims.length!==1)throw new Error("Invalid axes input dims.")},Fp=e=>["","",`var value = ${e.getByIndices("input_indices")};`,""],da=(e,r,t,s,o,n,a=!1,i=!1)=>{let l=[],c=t[0].dims,p=c.length,d=ye.normalizeAxes(o,p),u=!i&&d.length===0;c.forEach((b,k)=>{u||d.indexOf(k)>=0?a&&l.push(1):l.push(b)});let f=l.length,_=ye.size(l);return{name:e,shaderCache:r,getShaderSource:b=>{let k=[],w=Ie("_A",t[0].dataType,p),y=nt("output",n,f),$=s(w,y,d),T=$[2];for(let M=0,x=0;M<p;M++)u||d.indexOf(M)>=0?(a&&x++,T=`for(var j${M}: u32 = 0; j${M} < ${c[M]}; j${M}++) {
                  ${$[2].includes("last_index")?`let last_index = j${M};`:""}
                  ${w.indicesSet("input_indices",M,`j${M}`)}
                  ${T}
                }`):(k.push(`${w.indicesSet("input_indices",M,y.indicesGet("output_indices",x))};`),x++);return`

        ${b.registerUniform("output_size","u32").declareVariables(w,y)}

        ${b.mainStart()}
          ${b.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.output_size")}
          var input_indices: ${w.type.indices};
          let output_indices = ${y.offsetToIndices("global_idx")};

          ${k.join(`
`)}
          ${$[0]}       // init ops for reduce max/min
          ${$[1]}
          ${T}
          ${$[3]}
          ${$.length===4?y.setByOffset("global_idx","value"):$.slice(4).join(`
`)}
        }`},getRunData:()=>({outputs:[{dims:l,dataType:n}],dispatchGroup:{x:Math.ceil(_/64)},programUniforms:[{type:12,data:_},...dt(c,l)]})}},dl=(e,r)=>{let t=[];return e[1].dims[0]>0&&e[1].getBigInt64Array().forEach(s=>t.push(Number(s))),Kt({axes:t,keepDims:r.keepDims,noopWithEmptyAxes:r.noopWithEmptyAxes})},bs=(e,r,t,s)=>{let o=e.inputs,n=o.length===1?t:dl(o,t);e.compute(da(r,{hint:n.cacheKey,inputDependencies:["rank"]},[o[0]],n.noopWithEmptyAxes&&n.axes.length===0?Fp:s,n.axes,o[0].dataType,n.keepDims,n.noopWithEmptyAxes),{inputs:[0]})},Op=(e,r)=>{Ms(e.inputs),bs(e,"ReduceLogSum",r,(t,s)=>[`var value = ${s.type.storage}(0);`,"",`value += ${t.getByIndices("input_indices")};`,"value = log(value);"])},Dp=(e,r)=>{Ms(e.inputs),bs(e,"ReduceL1",r,(t,s)=>[`var value = ${s.type.storage}(0);`,"",`value += abs(${t.getByIndices("input_indices")});`,""])},Lp=(e,r)=>{Ms(e.inputs),bs(e,"ReduceL2",r,(t,s)=>[`var t = ${s.type.value}(0); var value = ${s.type.value}(0);`,"",`t = ${t.getByIndices("input_indices")}; value += (t * t);`,"value = sqrt(value);"])},zp=(e,r)=>{Ms(e.inputs),bs(e,"ReduceLogSumExp",r,(t,s)=>[`var value = ${s.type.storage}(0);`,"",`value += exp(${t.getByIndices("input_indices")});`,"value = log(value);"])},Bp=(e,r)=>{Ms(e.inputs),bs(e,"ReduceMax",r,(t,s,o)=>{let n=[];for(let a=0;a<t.rank;a++)(o.indexOf(a)>=0||o.length===0)&&n.push(t.indicesSet("input_indices",a,0));return[`${n.join(`
`)}`,`var value = ${t.getByIndices("input_indices")};`,`value = max(value, ${t.getByIndices("input_indices")});`,""]})},Rp=(e,r)=>{Ms(e.inputs),bs(e,"ReduceMean",r,(t,s,o)=>{let n=1;for(let a=0;a<t.rank;a++)(o.indexOf(a)>=0||o.length===0)&&(n*=e.inputs[0].dims[a]);return["var sum = f32(0);","",`sum += f32(${t.getByIndices("input_indices")});`,`let value = ${s.type.value}(sum / ${n});`]})},jp=(e,r)=>{Ms(e.inputs),bs(e,"ReduceMin",r,(t,s,o)=>{let n=[];for(let a=0;a<t.rank;a++)(o.indexOf(a)>=0||o.length===0)&&n.push(`input_indices[${a}] = 0;`);return[`${n.join(`
`)}`,`var value = ${t.getByIndices("input_indices")};`,`value = min(value, ${t.getByIndices("input_indices")});`,""]})},Np=(e,r)=>{Ms(e.inputs),bs(e,"ReduceProd",r,(t,s)=>[`var value = ${s.type.storage}(1);`,"",`value *= ${t.getByIndices("input_indices")};`,""])},Vp=(e,r)=>{Ms(e.inputs),bs(e,"ReduceSum",r,(t,s)=>[`var value = ${s.type.storage}(0);`,"",`value += ${t.getByIndices("input_indices")};`,""])},Up=(e,r)=>{Ms(e.inputs),bs(e,"ReduceSumSquare",r,(t,s)=>[`var t = ${s.type.value}(0); var value = ${s.type.value}(0);`,"",`t = ${t.getByIndices("input_indices")}; value += t * t;`,""])},ys=(e,r,t)=>{if(r.length===0)return t;let s=1,o=1;for(let n=0;n<r.length;n++)r.indexOf(n)===-1?s*=e[n]:o*=e[n];return o<32&&s>1024},Wp=(e,r)=>{ys(e.inputs[0].dims,r.axes,r.noopWithEmptyAxes)?Rp(e,r):xp(e,r)},Gp=(e,r)=>{ys(e.inputs[0].dims,r.axes,r.noopWithEmptyAxes)?Dp(e,r):Tp(e,r)},Kp=(e,r)=>{ys(e.inputs[0].dims,r.axes,r.noopWithEmptyAxes)?Lp(e,r):Ep(e,r)},Hp=(e,r)=>{ys(e.inputs[0].dims,r.axes,r.noopWithEmptyAxes)?zp(e,r):Pp(e,r)},qp=(e,r)=>{ys(e.inputs[0].dims,r.axes,r.noopWithEmptyAxes)?Bp(e,r):Cp(e,r)},Qp=(e,r)=>{ys(e.inputs[0].dims,r.axes,r.noopWithEmptyAxes)?jp(e,r):Sp(e,r)},Xp=(e,r)=>{ys(e.inputs[0].dims,r.axes,r.noopWithEmptyAxes)?Np(e,r):$p(e,r)},Jp=(e,r)=>{ys(e.inputs[0].dims,r.axes,r.noopWithEmptyAxes)?Vp(e,r):kp(e,r)},Yp=(e,r)=>{ys(e.inputs[0].dims,r.axes,r.noopWithEmptyAxes)?Up(e,r):Ip(e,r)},Zp=(e,r)=>{ys(e.inputs[0].dims,r.axes,r.noopWithEmptyAxes)?Op(e,r):Ap(e,r)}}),ml,em,tm,hl,zx=Ue(()=>{Mt(),pr(),pl(),ml=e=>{if(!e||e.length===0||e.length>2)throw new Error("ArgMinMaxOp op requires 1 or 2 inputs.");if(e[0].dataType!==1)throw new Error("Invalid input type.")},em=(e,r)=>{ml(e.inputs);let t=(s,o,n)=>{let a=[];for(let i=0;i<s.rank;i++)(n.indexOf(i)>=0||n.length===0)&&a.push(`input_indices[${i}] = 0;`);return[`${a.join(`
`)}`,`var value = ${s.getByIndices("input_indices")};
var best_index : i32 = 0;`,`if (${s.getByIndices("input_indices")} ${r.selectLastIndex>0?"<=":"<"} value) {
         value = ${s.getByIndices("input_indices")};
         best_index = i32(last_index);
       }`,"",o.setByOffset("global_idx","best_index")]};e.compute(da("ArgMin",{hint:r.cacheKey,inputDependencies:["rank"]},[e.inputs[0]],t,[r.axis],7,r.keepDims),{inputs:[0]})},tm=(e,r)=>{ml(e.inputs);let t=(s,o,n)=>{let a=[];for(let i=0;i<s.rank;i++)(n.indexOf(i)>=0||n.length===0)&&a.push(`input_indices[${i}] = 0;`);return[`${a.join(`
`)}`,`var value = ${s.getByIndices("input_indices")};
var best_index : i32 = 0;`,`if (${s.getByIndices("input_indices")} ${r.selectLastIndex>0?">=":">"} value) {
         value = ${s.getByIndices("input_indices")};
         best_index = i32(last_index);
       }`,"",o.setByOffset("global_idx","best_index")]};e.compute(da("argMax",{hint:r.cacheKey,inputDependencies:["rank"]},[e.inputs[0]],t,[r.axis],7,r.keepDims),{inputs:[0]})},hl=e=>Kt(e)}),rm,pa,sm,nm,om,wo,am,im,_l=Ue(()=>{Mt(),Pt(),nl(),Ct(),rm=(e,r)=>{let t=e[0],s=e[1],o=e[2],n=e[3],a=e[4],i=e[5];if(a&&i)throw new Error("Attention cannot have both past and attention_bias");if(t.dims.length!==3)throw new Error('Input "input" must have 3 dimensions');let l=t.dims[0],c=t.dims[1],p=t.dims[2];if(o.dims.length!==1)throw new Error('Input "bias" is expected to have 1 dimensions');if(s.dims.length!==2)throw new Error('Input "weights" is expected to have 2 dimensions');if(s.dims[0]!==p)throw new Error("Input 1 dimension 0 should have same length as dimension 2 of input 0");if(o.dims[0]!==s.dims[1])throw new Error('Input "bias" dimension 0 should have same length as dimension 1 of input "weights"');let d=o.dims[0]/3,u=d,f=u;if(r.qkvHiddenSizes.length>0){if(r.qkvHiddenSizes.length!==3)throw new Error("qkv_hidden_sizes attribute should have 3 elements");for(let $ of r.qkvHiddenSizes)if($%r.numHeads!==0)throw new Error("qkv_hidden_sizes should be divisible by num_heads");d=r.qkvHiddenSizes[0],u=r.qkvHiddenSizes[1],f=r.qkvHiddenSizes[2]}let _=c;if(d!==u)throw new Error("qkv_hidden_sizes first element should be same as the second");if(o.dims[0]!==d+u+f)throw new Error('Input "bias" dimension 0 should have same length as sum of Q/K/V hidden sizes');let b=0;if(a){if(u!==f)throw new Error('Input "past" expect k_hidden_size == v_hidden_size');if(a.dims.length!==5)throw new Error('Input "past" must have 5 dimensions');if(a.dims[0]!==2)throw new Error('Input "past" first dimension must be 2');if(a.dims[1]!==l)throw new Error('Input "past" second dimension must be batch_size');if(a.dims[2]!==r.numHeads)throw new Error('Input "past" third dimension must be num_heads');if(a.dims[4]!==u/r.numHeads)throw new Error('Input "past" fifth dimension must be k_hidden_size / num_heads');r.pastPresentShareBuffer||(b=a.dims[3])}let k=_+b,w=-1,y=0;if(n)throw new Error("Mask not supported");if(a)throw new Error("past is not supported");if(i){if(i.dims.length!==4)throw new Error('Input "attention_bias" must have 4 dimensions');if(i.dims[0]!==l||i.dims[1]!==r.numHeads||i.dims[2]!==c||i.dims[3]!==k)throw new Error('Expect "attention_bias" shape (batch_size, num_heads, sequence_length, total_sequence_length)')}return{batchSize:l,sequenceLength:c,pastSequenceLength:b,kvSequenceLength:_,totalSequenceLength:k,maxSequenceLength:w,inputHiddenSize:p,hiddenSize:d,vHiddenSize:f,headSize:Math.floor(d/r.numHeads),vHeadSize:Math.floor(f/r.numHeads),numHeads:r.numHeads,isUnidirectional:!1,pastPresentShareBuffer:!1,maskFilterValue:r.maskFilterValue,maskType:y,scale:r.scale,broadcastResPosBias:!1,passPastInKv:!1,qkvFormat:1}},pa=(e,r,t)=>r&&e?`
      let total_sequence_length_input = u32(${r.getByOffset("0")});
      let present_sequence_length = max(total_sequence_length_input, uniforms.past_sequence_length);
      let is_subsequent_prompt: bool = sequence_length > 1 && sequence_length != total_sequence_length_input;
      let is_first_prompt: bool = is_subsequent_prompt == false && sequence_length == total_sequence_length_input;
      total_sequence_length = u32(${e==null?void 0:e.getByOffset("batchIdx")}) + 1;
      var past_sequence_length: u32 = 0;
      if (is_first_prompt == false) {
        past_sequence_length = total_sequence_length - sequence_length;
      }
       `:`
    ${t?"let past_sequence_length = uniforms.past_sequence_length":""};
    let present_sequence_length = total_sequence_length;
    `,sm=(e,r,t,s,o,n,a,i)=>{let l=lr(a?1:n),c=64,p=n/l;p<c&&(c=32);let d=Math.ceil(n/l/c),u=[{type:12,data:r},{type:12,data:t},{type:12,data:s},{type:12,data:o},{type:12,data:p},{type:12,data:d}],f=Ir(e.dataType,l),_=Vr(1,l),b=["type"];a&&b.push("type"),i&&b.push("type");let k=w=>{let y=nt("x",e.dataType,e.dims,l),$=[y],T=a?Ie("seq_lens",a.dataType,a.dims):void 0;T&&$.push(T);let M=i?Ie("total_sequence_length_input",i.dataType,i.dims):void 0;M&&$.push(M);let x=Vr(e.dataType),v=[{name:"batch_size",type:"u32"},{name:"num_heads",type:"u32"},{name:"past_sequence_length",type:"u32"},{name:"sequence_length",type:"u32"},{name:"total_sequence_length",type:"u32"},{name:"elements_per_thread",type:"u32"}];return`
  var<workgroup> thread_max: array<f32, ${c}>;
  var<workgroup> thread_sum: array<f32, ${c}>;
  ${w.registerUniforms(v).declareVariables(...$)}
  ${w.mainStart([c,1,1])}
    let batchIdx = workgroup_id.z / uniforms.num_heads;
    let headIdx = workgroup_id.z % uniforms.num_heads;
    let sequence_length = uniforms.sequence_length;
    var total_sequence_length = uniforms.total_sequence_length;
    ${pa(T,M,!1)}
    let local_offset = local_idx * uniforms.elements_per_thread;
    let offset = (global_idx / ${c}) * uniforms.total_sequence_length + local_offset;
    let seq_causal_length = ${a?"u32(past_sequence_length + workgroup_id.y + 1)":"total_sequence_length"};
    var thread_max_vector = ${_}(-3.402823e+38f);
    for (var i: u32 = 0; i < uniforms.elements_per_thread && i + local_offset < seq_causal_length; i++) {
      thread_max_vector = max(${_}(x[offset + i]), thread_max_vector);
    }
    thread_max[local_idx] = ${(()=>{switch(l){case 1:return"thread_max_vector";case 2:return"max(thread_max_vector.x, thread_max_vector.y)";case 4:return"max(max(thread_max_vector.x, thread_max_vector.y), max(thread_max_vector.z, thread_max_vector.w))";default:throw new Error(`Unsupported components: ${l}`)}})()};
    workgroupBarrier();

    var max_value =  f32(-3.402823e+38f);
    for (var i = 0u; i < ${c}; i++) {
      max_value = max(thread_max[i], max_value);
    }

    var sum_vector = ${_}(0);
    for (var i: u32 = 0; i < uniforms.elements_per_thread && i + local_offset < seq_causal_length; i++) {
      sum_vector += exp(${_}(x[offset + i]) - max_value);
    }
    thread_sum[local_idx] = ${(()=>{switch(l){case 1:return"sum_vector";case 2:return"sum_vector.x + sum_vector.y";case 4:return"sum_vector.x + sum_vector.y + sum_vector.z + sum_vector.w";default:throw new Error(`Unsupported components: ${l}`)}})()};
    workgroupBarrier();

    var sum: f32 = 0;
    for (var i = 0u; i < ${c}; i++) {
      sum += thread_sum[i];
    }

    if (sum == 0) {
      for (var i: u32 = 0; i < uniforms.elements_per_thread && i + local_offset < seq_causal_length; i++) {
        x[offset + i] = ${y.type.value}(${x}(1.0) / ${x}(seq_causal_length));
      }
    } else {
      for (var i: u32 = 0; i < uniforms.elements_per_thread && i + local_offset < seq_causal_length; i++) {
        var f32input = ${_}(x[offset + i]);
        x[offset + i] = ${y.type.value}(exp(f32input - max_value) / sum);
      }
    }
      ${a?`
        for (var total_seq_id: u32 = seq_causal_length; total_seq_id + local_offset < uniforms.total_sequence_length; total_seq_id++) {
          x[offset + total_seq_id] = ${y.type.value}(${x}(0));
        }`:""};
  }`};return{name:"AttentionProbsSoftmax",shaderCache:{hint:`${c};${f};${l}`,inputDependencies:b},getShaderSource:k,getRunData:()=>({outputs:[],dispatchGroup:{x:1,y:o,z:r*t},programUniforms:u})}},nm=(e,r,t,s,o,n,a,i,l)=>{let c=a+n.kvSequenceLength,p=[n.batchSize,n.numHeads,n.sequenceLength,c],d=e>1&&s,u=n.kvNumHeads?n.kvNumHeads:n.numHeads,f=d?[n.batchSize,u,c,n.headSize]:void 0,_=n.nReps?n.nReps:1,b=n.scale===0?1/Math.sqrt(n.headSize):n.scale,k=lr(n.headSize),w=n.headSize/k,y=12,$={x:Math.ceil(c/y),y:Math.ceil(n.sequenceLength/y),z:n.batchSize*n.numHeads},T=[{type:12,data:n.sequenceLength},{type:12,data:w},{type:12,data:c},{type:12,data:n.numHeads},{type:12,data:n.headSize},{type:1,data:b},{type:12,data:a},{type:12,data:n.kvSequenceLength},{type:12,data:_}],M=d&&s&&ye.size(s.dims)>0,x=["type","type"];M&&x.push("type"),o&&x.push("type"),i&&x.push("type"),l&&x.push("type");let v=[{dims:p,dataType:r.dataType,gpuDataType:0}];d&&v.push({dims:f,dataType:r.dataType,gpuDataType:0});let C=F=>{let D=Ie("q",r.dataType,r.dims,k),K=Ie("key",t.dataType,t.dims,k),U=[D,K];if(M){let _e=Ie("past_key",s.dataType,s.dims,k);U.push(_e)}o&&U.push(Ie("attention_bias",o.dataType,o.dims));let j=i?Ie("seq_lens",i.dataType,i.dims):void 0;j&&U.push(j);let se=l?Ie("total_sequence_length_input",l.dataType,l.dims):void 0;se&&U.push(se);let q=nt("output",r.dataType,p),re=[q];d&&re.push(nt("present_key",r.dataType,f,k));let Z=Vr(1,k),ae=[{name:"M",type:"u32"},{name:"K",type:"u32"},{name:"N",type:"u32"},{name:"num_heads",type:"u32"},{name:"head_size",type:"u32"},{name:"alpha",type:"f32"},{name:"past_sequence_length",type:"u32"},{name:"kv_sequence_length",type:"u32"},{name:"n_reps",type:"u32"}];return`
  const TILE_SIZE = ${y}u;

  var<workgroup> tileQ: array<${D.type.storage}, ${y*y}>;
  var<workgroup> tileK: array<${D.type.storage}, ${y*y}>;
  ${F.registerUniforms(ae).declareVariables(...U,...re)}
  ${F.mainStart([y,y,1])}
    // x holds the N and y holds the M
    let headIdx = workgroup_id.z % uniforms.num_heads;
    let kvHeadIdx = ${_===1?"headIdx":"headIdx / uniforms.n_reps"};
    let kv_num_heads = ${_===1?"uniforms.num_heads":"uniforms.num_heads / uniforms.n_reps"};
    let batchIdx = workgroup_id.z / uniforms.num_heads;
    let m = workgroup_id.y * TILE_SIZE;
    let n = workgroup_id.x * TILE_SIZE;
    let sequence_length = uniforms.M;
    var total_sequence_length = uniforms.N;
    ${pa(j,se,!0)}
    let absKvHeadIdx = batchIdx * kv_num_heads + kvHeadIdx;
    let qOffset = workgroup_id.z * uniforms.M * uniforms.K + m * uniforms.K;
    ${M&&d?"let pastKeyOffset = absKvHeadIdx * uniforms.past_sequence_length * uniforms.K;":""};
    let kOffset = absKvHeadIdx * uniforms.kv_sequence_length * uniforms.K;
    ${d?"let presentKeyOffset = absKvHeadIdx * uniforms.N * uniforms.K;":""}
    var value = ${Z}(0);
    for (var w: u32 = 0u; w < uniforms.K; w += TILE_SIZE) {
      if (global_id.y < uniforms.M && w + local_id.x < uniforms.K) {
        tileQ[TILE_SIZE * local_id.y + local_id.x] = q[qOffset + local_id.y * uniforms.K + w + local_id.x];
      }
      if (n + local_id.y < uniforms.N && w + local_id.x < uniforms.K) {
        var idx = TILE_SIZE * local_id.y + local_id.x;
      ${M&&d?`
              if (n + local_id.y < past_sequence_length) {
                tileK[idx] = past_key[pastKeyOffset + (n + local_id.y) * uniforms.K + w + local_id.x];
              } else if (n + local_id.y - past_sequence_length < uniforms.kv_sequence_length) {
                tileK[idx] = key[kOffset + (n + local_id.y - past_sequence_length) * uniforms.K + w + local_id.x];
              }`:`
          if (n + local_id.y < uniforms.kv_sequence_length) {
            tileK[idx] = key[kOffset + (n + local_id.y) * uniforms.K + w + local_id.x];
          }`}
      ${d?`if (n + local_id.y < present_sequence_length) {
        present_key[presentKeyOffset + (n + local_id.y) * uniforms.K + w + local_id.x] = tileK[idx];
      }`:""}
      }
      workgroupBarrier();

      for (var k: u32 = 0u; k < TILE_SIZE && w+k < uniforms.K; k++) {
          value += ${Z}(tileQ[TILE_SIZE * local_id.y + k] * tileK[TILE_SIZE * local_id.x + k]);
      }

      workgroupBarrier();
    }

    if (global_id.y < uniforms.M && global_id.x < total_sequence_length) {
      let headOffset = workgroup_id.z * uniforms.M * uniforms.N;
      let outputIdx = headOffset + global_id.y * uniforms.N + global_id.x;
      var sum: f32 = ${(()=>{switch(k){case 1:return"value";case 2:return"value.x + value.y";case 4:return"value.x + value.y + value.z + value.w";default:throw new Error(`Unsupported components: ${k}`)}})()};
        output[outputIdx] = ${q.type.value} (sum * uniforms.alpha) + ${o?"attention_bias[outputIdx]":"0.0"};
    }
  }`};return{name:"AttentionProbs",shaderCache:{hint:`${k};${o!==void 0};${s!==void 0};${e}`,inputDependencies:x},getRunData:()=>({outputs:v,dispatchGroup:$,programUniforms:T}),getShaderSource:C}},om=(e,r,t,s,o,n,a=void 0,i=void 0)=>{let l=n+o.kvSequenceLength,c=o.nReps?o.nReps:1,p=o.vHiddenSize*c,d=e>1&&s,u=o.kvNumHeads?o.kvNumHeads:o.numHeads,f=d?[o.batchSize,u,l,o.headSize]:void 0,_=[o.batchSize,o.sequenceLength,p],b=12,k={x:Math.ceil(o.vHeadSize/b),y:Math.ceil(o.sequenceLength/b),z:o.batchSize*o.numHeads},w=[{type:12,data:o.sequenceLength},{type:12,data:l},{type:12,data:o.vHeadSize},{type:12,data:o.numHeads},{type:12,data:o.headSize},{type:12,data:p},{type:12,data:n},{type:12,data:o.kvSequenceLength},{type:12,data:c}],y=d&&s&&ye.size(s.dims)>0,$=["type","type"];y&&$.push("type"),a&&$.push("type"),i&&$.push("type");let T=[{dims:_,dataType:r.dataType,gpuDataType:0}];d&&T.push({dims:f,dataType:r.dataType,gpuDataType:0});let M=x=>{let v=Ie("probs",r.dataType,r.dims),C=Ie("v",t.dataType,t.dims),F=[v,C];y&&F.push(Ie("past_value",s.dataType,s.dims));let D=a?Ie("seq_lens",a.dataType,a.dims):void 0;a&&F.push(D);let K=i?Ie("total_sequence_length_input",i.dataType,i.dims):void 0;i&&F.push(K);let U=[nt("output",r.dataType,_)];d&&U.push(nt("present_value",r.dataType,f));let j=[{name:"M",type:"u32"},{name:"K",type:"u32"},{name:"N",type:"u32"},{name:"num_heads",type:"u32"},{name:"head_size",type:"u32"},{name:"v_hidden_size",type:"u32"},{name:"past_sequence_length",type:"u32"},{name:"kv_sequence_length",type:"u32"},{name:"n_reps",type:"u32"}];return`
  const TILE_SIZE = ${b}u;
  var<workgroup> tileQ: array<${v.type.value}, ${b*b}>;
  var<workgroup> tileV: array<${v.type.value}, ${b*b}>;
  ${x.registerUniforms(j).declareVariables(...F,...U)}
  ${x.mainStart([b,b,1])}
   let headIdx = workgroup_id.z % uniforms.num_heads;
   let batchIdx = workgroup_id.z / uniforms.num_heads;
   let kvHeadIdx = ${c===1?"headIdx":"headIdx / uniforms.n_reps"};
   let kv_num_heads = ${c===1?"uniforms.num_heads":"uniforms.num_heads / uniforms.n_reps"};
   let m = global_id.y;
   let n = global_id.x;
   let sequence_length = uniforms.M;
   var total_sequence_length = uniforms.K;
   ${pa(D,K,!0)}
   let offsetA = workgroup_id.z * uniforms.M * uniforms.K + m * uniforms.K;
   let absKvHeadIdx = batchIdx * kv_num_heads + kvHeadIdx; // kvHeadIdx is relative to the batch
   ${y&&d?"let pastValueOffset = absKvHeadIdx * uniforms.N * uniforms.past_sequence_length + n;":""};
   let vOffset = absKvHeadIdx * uniforms.N * uniforms.kv_sequence_length + n;
   ${d?"let presentValueOffset = absKvHeadIdx * uniforms.N * uniforms.K + n;":""}
   var value = ${v.type.storage}(0);
   for (var w: u32 = 0u; w < uniforms.K; w += TILE_SIZE) {
      if (m < uniforms.M && w + local_id.x < uniforms.K) {
        tileQ[TILE_SIZE * local_id.y + local_id.x] = probs[offsetA + w + local_id.x];
      }
      if (n < uniforms.N && w + local_id.y < uniforms.K) {
        var idx = TILE_SIZE * local_id.y + local_id.x;
        ${y&&d?`
        if (w + local_id.y < past_sequence_length) {
          tileV[idx] = past_value[pastValueOffset + (w + local_id.y) * uniforms.N];
        } else if (w + local_id.y - past_sequence_length < uniforms.kv_sequence_length) {
          tileV[idx] = v[vOffset + (w + local_id.y - past_sequence_length) * uniforms.N];
        }
      `:`
            if (w + local_id.y < uniforms.kv_sequence_length) {
              tileV[idx] = v[vOffset + (w + local_id.y) * uniforms.N];
            }`}
        ${d?`
            if (w + local_id.y < present_sequence_length) {
          present_value[presentValueOffset + (w + local_id.y) * uniforms.N] = tileV[idx];
        }`:""}
      }
     workgroupBarrier();
     for (var k: u32 = 0u; k < TILE_SIZE && w+k < total_sequence_length; k++) {
       value += tileQ[TILE_SIZE * local_id.y + k] * tileV[TILE_SIZE * k + local_id.x];
     }
     workgroupBarrier();
   }

   // we need to transpose output from BNSH_v to BSND_v
   if (m < uniforms.M && n < uniforms.N) {
     let outputIdx = batchIdx * uniforms.M * uniforms.v_hidden_size + m * uniforms.v_hidden_size
       + headIdx * uniforms.N + n;
     output[outputIdx] = value;
   }
  }`};return{name:"AttentionScore",shaderCache:{hint:`${s!==void 0};${e}`,inputDependencies:$},getRunData:()=>({outputs:T,dispatchGroup:k,programUniforms:w}),getShaderSource:M}},wo=(e,r,t,s,o,n,a,i,l,c,p=void 0,d=void 0)=>{let u=Math.min(e.outputCount,1+(a?1:0)+(i?1:0)),f=u>1?c.pastSequenceLength:0,_=f+c.kvSequenceLength,b=l&&ye.size(l.dims)>0?l:void 0,k=[r,t];u>1&&a&&ye.size(a.dims)>0&&k.push(a),b&&k.push(b),p&&k.push(p),d&&k.push(d);let w=e.compute(nm(u,r,t,a,b,c,f,p,d),{inputs:k,outputs:u>1?[-1,1]:[-1]})[0];e.compute(sm(w,c.batchSize,c.numHeads,f,c.sequenceLength,_,p,d),{inputs:p&&d?[w,p,d]:[w],outputs:[]});let y=[w,s];u>1&&i&&ye.size(i.dims)>0&&y.push(i),p&&y.push(p),d&&y.push(d),e.compute(om(u,w,s,i,c,f,p,d),{inputs:y,outputs:u>1?[0,2]:[0]})},am=(e,r)=>{let t=[r.batchSize,r.numHeads,r.sequenceLength,r.headSize],s=r.sequenceLength,o=r.inputHiddenSize,n=r.headSize,a=12,i={x:Math.ceil(r.headSize/a),y:Math.ceil(r.sequenceLength/a),z:r.batchSize*r.numHeads},l=[e.inputs[0],e.inputs[1],e.inputs[2]],c=[{type:12,data:s},{type:12,data:o},{type:12,data:n},{type:12,data:r.numHeads},{type:12,data:r.headSize},{type:12,data:r.hiddenSize},{type:12,data:r.hiddenSize+r.hiddenSize+r.vHiddenSize}],p=d=>{let u=nt("output_q",l[0].dataType,t),f=nt("output_k",l[0].dataType,t),_=nt("output_v",l[0].dataType,t),b=Ie("input",l[0].dataType,l[0].dims),k=Ie("weight",l[1].dataType,l[1].dims),w=Ie("bias",l[2].dataType,l[2].dims),y=b.type.storage,$=[{name:"M",type:"u32"},{name:"K",type:"u32"},{name:"N",type:"u32"},{name:"num_heads",type:"u32"},{name:"head_size",type:"u32"},{name:"hidden_size",type:"u32"},{name:"ldb",type:"u32"}];return`
  const TILE_SIZE = ${a}u;
  var<workgroup> tileInput: array<${y}, ${a*a}>;
  var<workgroup> tileWeightQ: array<${y}, ${a*a}>;
  var<workgroup> tileWeightK: array<${y}, ${a*a}>;
  var<workgroup> tileWeightV: array<${y}, ${a*a}>;
  ${d.registerUniforms($).declareVariables(b,k,w,u,f,_)}
  ${d.mainStart([a,a,1])}
    let batchIndex = workgroup_id.z / uniforms.num_heads;
    let headNumber = workgroup_id.z % uniforms.num_heads;
    let m = global_id.y;
    let n = global_id.x;

    let inputOffset = batchIndex * (uniforms.M * uniforms.K) + m * uniforms.K;
    let biasOffsetQ = headNumber * uniforms.head_size;
    let biasOffsetK = uniforms.hidden_size + biasOffsetQ;
    let biasOffsetV = uniforms.hidden_size + biasOffsetK;

    var valueQ = ${y}(0);
    var valueK = ${y}(0);
    var valueV = ${y}(0);
    for (var w: u32 = 0u; w < uniforms.K; w += TILE_SIZE) {
      if (m < uniforms.M && w + local_id.x < uniforms.K) {
        tileInput[TILE_SIZE * local_id.y + local_id.x] = input[inputOffset + w + local_id.x];
      }
      if (n < uniforms.N && w + local_id.y < uniforms.K) {
        let offset = n + (w + local_id.y) * uniforms.ldb;
        tileWeightQ[TILE_SIZE * local_id.y + local_id.x] = weight[biasOffsetQ + offset];
        tileWeightK[TILE_SIZE * local_id.y + local_id.x] = weight[biasOffsetK + offset];
        tileWeightV[TILE_SIZE * local_id.y + local_id.x] = weight[biasOffsetV + offset];
      }
      workgroupBarrier();
      for (var k: u32 = 0u; k<TILE_SIZE && w+k < uniforms.K; k++) {
        let inputTileOffset = TILE_SIZE * local_id.y + k;
        let weightTileOffset = TILE_SIZE * k + local_id.x;
        valueQ += tileInput[inputTileOffset] * tileWeightQ[weightTileOffset];
        valueK += tileInput[inputTileOffset] * tileWeightK[weightTileOffset];
        valueV += tileInput[inputTileOffset] * tileWeightV[weightTileOffset];
      }

      workgroupBarrier();
    }

    let headOffset = (m * uniforms.N + n) % uniforms.head_size;
    valueQ += bias[headOffset + biasOffsetQ];
    valueK += bias[headOffset + biasOffsetK];
    valueV += bias[headOffset + biasOffsetV];

    let offset = workgroup_id.z * uniforms.M * uniforms.N;
    if (m < uniforms.M && n < uniforms.N) {
      let outputIdx = offset + m * uniforms.N + n;
      output_q[outputIdx] = valueQ;
      output_k[outputIdx] = valueK;
      output_v[outputIdx] = valueV;
    }
  }`};return e.compute({name:"AttentionPrepare",shaderCache:{inputDependencies:["type","type","type"]},getRunData:()=>({outputs:[{dims:t,dataType:e.inputs[0].dataType,gpuDataType:0},{dims:t,dataType:e.inputs[0].dataType,gpuDataType:0},{dims:t,dataType:e.inputs[0].dataType,gpuDataType:0}],dispatchGroup:i,programUniforms:c}),getShaderSource:p},{inputs:l,outputs:[-1,-1,-1]})},im=(e,r)=>{let t=rm(e.inputs,r),[s,o,n]=am(e,t);return wo(e,s,o,n,e.inputs[4],void 0,void 0,void 0,e.inputs[5],t)}}),lm,cm,um,dm,Bx=Ue(()=>{fs(),Mt(),Pt(),pr(),Ct(),lm=(e,r)=>{if(!e||e.length!==5)throw new Error("BatchNormalization requires 5 inputs");let t=(s,o,n)=>{let a=o.length;if(a!==s.length)throw new Error(`${n}: num dimensions != ${a}`);o.forEach((i,l)=>{if(i!==s[l])throw new Error(`${n}: dim[${l}] do not match`)})};if(e[0].dims.length>1){let s=r.format==="NHWC"?r.spatial?e[0].dims.slice(-1):e[0].dims.slice(-1).concat(e[0].dims.slice(1,e[0].dims.length-1)):e[0].dims.slice(1,r.spatial?2:void 0);t(e[1].dims,s,"Invalid input scale"),t(e[2].dims,s,"Invalid input B"),t(e[3].dims,s,"Invalid input mean"),t(e[4].dims,s,"Invalid input var")}else t(e[1].dims,[1],"Invalid input scale"),t(e[2].dims,[1],"Invalid input B"),t(e[3].dims,[1],"Invalid input mean"),t(e[4].dims,[1],"Invalid input var")},cm=(e,r)=>{let{epsilon:t,spatial:s,format:o}=r,n=e[0].dims,a=s?lr(n[n.length-1]):1,i=o==="NHWC"&&n.length>1?a:1,l=ye.size(n)/a,c=s,p=c?n.length:n,d=Ie("x",e[0].dataType,e[0].dims,a),u=Ie("scale",e[1].dataType,e[1].dims,i),f=Ie("bias",e[2].dataType,e[2].dims,i),_=Ie("inputMean",e[3].dataType,e[3].dims,i),b=Ie("inputVar",e[4].dataType,e[4].dims,i),k=nt("y",e[0].dataType,p,a),w=()=>{let $="";if(s)$=`let cOffset = ${n.length===1?"0u":o==="NHWC"?`outputIndices[${n.length-1}] / ${a}`:"outputIndices[1]"};`;else if(o==="NCHW")$=`
            ${k.indicesSet("outputIndices","0","0")}
            let cOffset = ${k.indicesToOffset("outputIndices")};`;else{$=`var cIndices = ${u.type.indices}(0);
                       cIndices[0] = outputIndices[${n.length-1}];`;for(let T=1;T<u.rank;T++)$+=`cIndices[${T}] = outputIndices[${T}];`;$+=`let cOffset = ${u.indicesToOffset("cIndices")};`}return $},y=$=>`
  const epsilon = ${t};
  ${$.registerUniform("outputSize","u32").declareVariables(d,u,f,_,b,k)}
  ${$.mainStart()}
  ${$.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.outputSize")}
    var outputIndices = ${k.offsetToIndices(`global_idx * ${a}`)};
    ${w()}
    let scale = ${u.getByOffset("cOffset")};
    let bias = ${f.getByOffset("cOffset")};
    let inputMean = ${_.getByOffset("cOffset")};
    let inputVar = ${b.getByOffset("cOffset")};
    let x = ${d.getByOffset("global_idx")};
    let value = (x - inputMean) * inverseSqrt(inputVar + epsilon) * scale + bias;
    ${k.setByOffset("global_idx","value")}
  }`;return{name:"BatchNormalization",shaderCache:{hint:`${r.epsilon}_${r.format}_${s}_${a}`,inputDependencies:c?["rank","type","type","type","type"]:void 0},getShaderSource:y,getRunData:()=>({outputs:[{dims:e[0].dims,dataType:e[0].dataType}],dispatchGroup:{x:Math.ceil(l/64)},programUniforms:c?[{type:12,data:l},...dt(n)]:[{type:12,data:l}]})}},um=e=>Kt(e),dm=(e,r)=>{let{inputs:t,outputCount:s}=e,o=um({...r,outputCount:s});if(tr.webgpu.validateInputContent&&lm(t,o),r.trainingMode)throw new Error("BatchNormalization trainingMode is not supported yet.");e.compute(cm(t,o))}}),pm,mm,hm,Rx=Ue(()=>{Pt(),Ct(),pm=e=>{if(e[0].dims.length!==3)throw new Error("input should have 3 dimensions");if(![320,640,1280].includes(e[0].dims[2]))throw new Error("number of channels should be 320, 640 or 1280");if(e[1].dims.length!==1)throw new Error("bias is expected to have 1 dimensions");if(e[0].dims[2]!==e[1].dims[0])throw new Error("last dimension of input and bias are not the same")},mm=e=>{let r=e[0].dims,t=e[0].dims[2],s=ye.size(r)/4,o=e[0].dataType,n=Ie("input",o,r,4),a=Ie("bias",o,[t],4),i=Ie("residual",o,r,4),l=nt("output",o,r,4);return{name:"BiasAdd",getRunData:()=>({outputs:[{dims:r,dataType:e[0].dataType}],dispatchGroup:{x:Math.ceil(s/64)}}),getShaderSource:c=>`
  const channels = ${t}u / 4;
  ${c.declareVariables(n,a,i,l)}

  ${c.mainStart()}
    ${c.guardAgainstOutOfBoundsWorkgroupSizes(s)}
    let value = ${n.getByOffset("global_idx")}
      + ${a.getByOffset("global_idx % channels")} + ${i.getByOffset("global_idx")};
    ${l.setByOffset("global_idx","value")}
  }`}},hm=e=>{pm(e.inputs),e.compute(mm(e.inputs))}}),_m,Vt,fm,gm,wm,Mm,bm,ym,vm,xm,Tm,Em,Pm,Cm,Sm,$m,Mo,km,ma,Im,Am,Fm,Om,Dm,Lm,zm,Bm,Rm,jm,Nm,Vm,Um,Wm,Gm,Km,fl,Hm,gl,wl,qm,Qm,Xm,Jm,Ym,Zm,Ml=Ue(()=>{Mt(),Pt(),pr(),Ct(),_m=(e,r,t,s,o,n,a)=>{let i=Math.ceil(r/4),l="";typeof o=="string"?l=`${o}(a)`:l=o("a");let c=Ie("inputData",t,[i],4),p=nt("outputData",s,[i],4),d=[{name:"vec_size",type:"u32"}];return a&&d.push(...a),`
      ${e.registerUniforms(d).declareVariables(c,p)}

  ${n??""}

  ${e.mainStart()}
    ${e.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.vec_size")}

    let a = ${c.getByOffset("global_idx")};
    ${p.setByOffset("global_idx",l)}
  }`},Vt=(e,r,t,s,o,n=e.dataType,a,i)=>{let l=[{type:12,data:Math.ceil(ye.size(e.dims)/4)}];return a&&l.push(...a),{name:r,shaderCache:{hint:o,inputDependencies:["type"]},getShaderSource:c=>_m(c,ye.size(e.dims),e.dataType,n,t,s,i),getRunData:c=>({outputs:[{dims:e.dims,dataType:n}],dispatchGroup:{x:Math.ceil(ye.size(c[0].dims)/64/4)},programUniforms:l})}},fm=e=>{e.compute(Vt(e.inputs[0],"Abs","abs"))},gm=e=>{e.compute(Vt(e.inputs[0],"Acos","acos"))},wm=e=>{e.compute(Vt(e.inputs[0],"Acosh","acosh"))},Mm=e=>{e.compute(Vt(e.inputs[0],"Asin","asin"))},bm=e=>{e.compute(Vt(e.inputs[0],"Asinh","asinh"))},ym=e=>{e.compute(Vt(e.inputs[0],"Atan","atan"))},vm=e=>{e.compute(Vt(e.inputs[0],"Atanh","atanh"))},xm=e=>Kt(e),Tm=(e,r)=>{let t;switch(r.to){case 10:t="vec4<f16>";break;case 1:t="vec4<f32>";break;case 12:t="vec4<u32>";break;case 6:t="vec4<i32>";break;case 9:t="vec4<bool>";break;default:throw new RangeError(`not supported type (specified in attribute 'to' from 'Cast' operator): ${r.to}`)}e.compute(Vt(e.inputs[0],"Cast",t,void 0,r.cacheKey,r.to))},Em=e=>{let r,t,s=e.length>=2&&e[1].data!==0,o=e.length>=3&&e[2].data!==0;switch(e[0].dataType){case 1:r=s?e[1].getFloat32Array()[0]:-34028234663852886e22,t=o?e[2].getFloat32Array()[0]:34028234663852886e22;break;case 10:r=s?e[1].getUint16Array()[0]:64511,t=o?e[2].getUint16Array()[0]:31743;break;default:throw new Error("Unsupport data type")}return Kt({min:r,max:t})},Pm=(e,r)=>{let t=r||Em(e.inputs),s=Vr(e.inputs[0].dataType);e.compute(Vt(e.inputs[0],"Clip",o=>`clamp(${o}, vec4<${s}>(uniforms.min), vec4<${s}>(uniforms.max))`,void 0,t.cacheKey,void 0,[{type:e.inputs[0].dataType,data:t.min},{type:e.inputs[0].dataType,data:t.max}],[{name:"min",type:s},{name:"max",type:s}]),{inputs:[0]})},Cm=e=>{e.compute(Vt(e.inputs[0],"Ceil","ceil"))},Sm=e=>{e.compute(Vt(e.inputs[0],"Cos","cos"))},$m=e=>{e.compute(Vt(e.inputs[0],"Cosh","cosh"))},Mo=e=>Kt(e),km=(e,r)=>{let t=Vr(e.inputs[0].dataType);e.compute(Vt(e.inputs[0],"Elu",s=>`elu_vf32(${s})`,`
  const elu_alpha_ = ${t}(${r.alpha});

  fn elu_f32(a: ${t}) -> ${t} {
  return select((exp(a) - 1.0) * elu_alpha_, a, a >= 0.0);
  }

  fn elu_vf32(v: vec4<${t}>) -> vec4<${t}> {
  return vec4(elu_f32(v.x), elu_f32(v.y), elu_f32(v.z), elu_f32(v.w));
  }`,r.cacheKey))},ma=(e="f32")=>`
const r0: ${e} = 0.3275911;
const r1: ${e} = 0.254829592;
const r2: ${e} = -0.284496736;
const r3: ${e} = 1.421413741;
const r4: ${e} = -1.453152027;
const r5: ${e} = 1.061405429;

fn erf_vf32(v: vec4<${e}>) -> vec4<${e}> {
  let absv = abs(v);
  let x = 1.0 / (1.0 + r0 * absv);
  return sign(v) * (1.0 - ((((r5 * x + r4) * x + r3) * x + r2) * x + r1) * x * exp(-absv * absv));
}`,Im=e=>{let r=Vr(e.inputs[0].dataType);e.compute(Vt(e.inputs[0],"Erf",t=>`erf_vf32(${t})`,ma(r)))},Am=e=>{e.compute(Vt(e.inputs[0],"Exp","exp"))},Fm=e=>{e.compute(Vt(e.inputs[0],"Floor","floor"))},Om=e=>{let r=Vr(e.inputs[0].dataType);e.compute(Vt(e.inputs[0],"Gelu",t=>`0.5 * ${t} * (1.0 + erf_vf32(${t} * 0.7071067811865475))`,ma(r)))},Dm=(e,r)=>{let t=Vr(e.inputs[0].dataType);e.compute(Vt(e.inputs[0],"LeakyRelu",s=>`select(leaky_relu_alpha_ * ${s}, ${s}, ${s} >= vec4<${t}>(0.0))`,`const leaky_relu_alpha_ = ${t}(${r.alpha});`,r.cacheKey))},Lm=e=>{e.compute(Vt(e.inputs[0],"Not",r=>`!${r}`))},zm=e=>{e.compute(Vt(e.inputs[0],"Neg",r=>`-${r}`))},Bm=e=>{e.compute(Vt(e.inputs[0],"Reciprocal",r=>`1.0/${r}`))},Rm=e=>{let r=Vr(e.inputs[0].dataType);e.compute(Vt(e.inputs[0],"Relu",t=>`select(vec4<${r}>(0.0), ${t}, ${t} > vec4<${r}>(0.0))`))},jm=e=>{e.compute(Vt(e.inputs[0],"Sigmoid",r=>`(1.0 / (1.0 + exp(-${r})))`))},Nm=e=>Kt(e),Vm=(e,r)=>{let t=Vr(e.inputs[0].dataType);e.compute(Vt(e.inputs[0],"HardSigmoid",s=>`max(vec4<${t}>(0.0), min(vec4<${t}>(1.0), ${r.alpha} * ${s} + vec4<${t}>(${r.beta})))`,void 0,r.cacheKey))},Um=e=>{e.compute(Vt(e.inputs[0],"Sin","sin"))},Wm=e=>{e.compute(Vt(e.inputs[0],"Sinh","sinh"))},Gm=e=>{e.compute(Vt(e.inputs[0],"Sqrt","sqrt"))},Km=e=>{e.compute(Vt(e.inputs[0],"Tan","tan"))},fl=e=>`sign(${e}) * (1 - exp(-2 * abs(${e}))) / (1 + exp(-2 * abs(${e})))`,Hm=e=>{e.compute(Vt(e.inputs[0],"Tanh",fl))},gl=(e="f32")=>`
const fast_gelu_a: ${e} = 0.5;
const fast_gelu_b: ${e} = 0.7978845608028654;
const fast_gelu_c: ${e} = 0.035677408136300125;

fn tanh_v(v: vec4<${e}>) -> vec4<${e}> {
  return ${fl("v")};
}
`,wl=e=>`(fast_gelu_a + fast_gelu_a * tanh_v(${e} * (fast_gelu_c * ${e} * ${e} + fast_gelu_b))) * ${e}`,qm=e=>{let r=Vr(e.inputs[0].dataType);e.compute(Vt(e.inputs[0],"FastGelu",wl,gl(r),void 0,e.inputs[0].dataType))},Qm=(e,r)=>{let t=Vr(e.inputs[0].dataType);return e.compute(Vt(e.inputs[0],"ThresholdedRelu",s=>`select(vec4<${t}>(0.0), ${s}, ${s} > thresholded_relu_alpha_)`,`const thresholded_relu_alpha_ = vec4<${t}>(${r.alpha});`,r.cacheKey)),0},Xm=e=>{e.compute(Vt(e.inputs[0],"Log","log"))},Jm=(e,r)=>`
const alpha = vec4<${e}>(${r});
const one = ${e}(1.0);
const zero = ${e}(0.0);

fn quick_gelu_impl(x: vec4<${e}>) -> vec4<${e}> {
  let v = x *alpha;
  var x1 : vec4<${e}>;
  for (var i = 0; i < 4; i = i + 1) {
    if (v[i] >= zero) {
      x1[i] = one / (one + exp(-v[i]));
    } else {
      x1[i] = one - one / (one + exp(v[i]));
    }
  }
  return x * x1;
}
`,Ym=e=>`quick_gelu_impl(${e})`,Zm=(e,r)=>{let t=Vr(e.inputs[0].dataType);e.compute(Vt(e.inputs[0],"QuickGelu",Ym,Jm(t,r.alpha),r.cacheKey,e.inputs[0].dataType))}}),eh,th,rh,jx=Ue(()=>{Pt(),Ct(),Ml(),eh=e=>{if(e[0].dims.length!==3)throw new Error("input should have 3 dimensions");if(![2560,5120,10240].includes(e[0].dims[2]))throw new Error("hidden state should be 2560, 5120 or 10240");if(e[1].dims.length!==1)throw new Error("bias is expected to have 1 dimensions");if(e[0].dims[2]!==e[1].dims[0])throw new Error("last dimension of input and bias are not the same")},th=e=>{let r=e[0].dims.slice();r[2]=r[2]/2;let t=Ie("input",e[0].dataType,e[0].dims,4),s=Ie("bias",e[0].dataType,[e[0].dims[2]],4),o=nt("output",e[0].dataType,r,4),n=ye.size(r)/4,a=Ir(e[0].dataType);return{name:"BiasSplitGelu",getRunData:()=>({outputs:[{dims:r,dataType:e[0].dataType}],dispatchGroup:{x:Math.ceil(n/64)}}),getShaderSource:i=>`
  const M_SQRT2 = sqrt(2.0);
  const halfChannels = ${e[0].dims[2]/4/2}u;

  ${i.declareVariables(t,s,o)}

  ${ma(a)}

  ${i.mainStart()}
    ${i.guardAgainstOutOfBoundsWorkgroupSizes(n)}
    let biasIdx = global_idx % halfChannels;
    let batchIndex = global_idx / halfChannels;
    let inputOffset = biasIdx + batchIndex * halfChannels * 2;
    let valueLeft = input[inputOffset] + bias[biasIdx];
    let valueRight = input[inputOffset + halfChannels] + bias[biasIdx + halfChannels];
    let geluRight = valueRight * 0.5 * (erf_vf32(valueRight / M_SQRT2) + 1);

    ${o.setByOffset("global_idx","valueLeft * geluRight")}
  }`}},rh=e=>{eh(e.inputs),e.compute(th(e.inputs))}}),sh,nh,vs,oh,ah,ih,lh,ch,uh,dh,ph,mh,hh,Nx=Ue(()=>{Mt(),Pt(),Ct(),sh=(e,r,t,s,o,n,a,i,l,c,p,d)=>{let u,f;typeof i=="string"?u=f=(y,$)=>`${i}((${y}),(${$}))`:typeof i=="function"?u=f=i:(u=i.scalar,f=i.vector);let _=nt("outputData",p,s.length,4),b=Ie("aData",l,r.length,4),k=Ie("bData",c,t.length,4),w;if(o)if(n){let y=ye.size(r)===1,$=ye.size(t)===1,T=r.length>0&&r[r.length-1]%4===0,M=t.length>0&&t[t.length-1]%4===0;y||$?w=_.setByOffset("global_idx",f(y?`${b.type.value}(${b.getByOffset("0")}.x)`:b.getByOffset("global_idx"),$?`${k.type.value}(${k.getByOffset("0")}.x)`:k.getByOffset("global_idx"))):w=`
            let outputIndices = ${_.offsetToIndices("global_idx * 4u")};
            let offsetA = ${b.broadcastedIndicesToOffset("outputIndices",_)};
            let offsetB = ${k.broadcastedIndicesToOffset("outputIndices",_)};
            ${_.setByOffset("global_idx",f(a||T?b.getByOffset("offsetA / 4u"):`${b.type.value}(${b.getByOffset("offsetA / 4u")}[offsetA % 4u])`,a||M?k.getByOffset("offsetB / 4u"):`${k.type.value}(${k.getByOffset("offsetB / 4u")}[offsetB % 4u])`))}
          `}else w=_.setByOffset("global_idx",f(b.getByOffset("global_idx"),k.getByOffset("global_idx")));else{if(!n)throw new Error("no necessary to use scalar implementation for element-wise binary op implementation.");let y=($,T,M="")=>{let x=`aData[indexA${T}][componentA${T}]`,v=`bData[indexB${T}][componentB${T}]`;return`
            let outputIndices${T} = ${_.offsetToIndices(`global_idx * 4u + ${T}u`)};
            let offsetA${T} = ${b.broadcastedIndicesToOffset(`outputIndices${T}`,_)};
            let offsetB${T} = ${k.broadcastedIndicesToOffset(`outputIndices${T}`,_)};
            let indexA${T} = offsetA${T} / 4u;
            let indexB${T} = offsetB${T} / 4u;
            let componentA${T} = offsetA${T} % 4u;
            let componentB${T} = offsetB${T} % 4u;
            ${$}[${T}] = ${M}(${u(x,v)});
          `};p===9?w=`
            var data = vec4<u32>(0);
            ${y("data",0,"u32")}
            ${y("data",1,"u32")}
            ${y("data",2,"u32")}
            ${y("data",3,"u32")}
            outputData[global_idx] = dot(vec4<u32>(0x1, 0x100, 0x10000, 0x1000000), vec4<u32>(data));`:w=`
            ${y("outputData[global_idx]",0)}
            ${y("outputData[global_idx]",1)}
            ${y("outputData[global_idx]",2)}
            ${y("outputData[global_idx]",3)}
          `}return`
        ${e.registerUniform("vec_size","u32").declareVariables(b,k,_)}

        ${d??""}

        ${e.mainStart()}
        ${e.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.vec_size")}
        ${w}
      }`},nh=(e,r,t,s,o,n,a=t.dataType)=>{let i=t.dims.map(b=>Number(b)??1),l=s.dims.map(b=>Number(b)??1),c=!ye.areEqual(i,l),p=i,d=ye.size(i),u=!1,f=!1,_=[c];if(c){let b=Gn.calcShape(i,l,!1);if(!b)throw new Error("Can't perform binary op on the given tensors");p=b.slice(),d=ye.size(p);let k=ye.size(i)===1,w=ye.size(l)===1,y=i.length>0&&i[i.length-1]%4===0,$=l.length>0&&l[l.length-1]%4===0;_.push(k),_.push(w),_.push(y),_.push($);let T=1;for(let M=1;M<p.length;M++){let x=i[i.length-M],v=l[l.length-M];if(x===v)T*=x;else break}T%4===0?(f=!0,u=!0):(k||w||y||$)&&(u=!0)}else u=!0;return _.push(u),{name:e,shaderCache:{hint:r+_.map(b=>b.toString()).join("_"),inputDependencies:["rank","rank"]},getShaderSource:b=>sh(b,i,l,p,u,c,f,o,t.dataType,s.dataType,a,n),getRunData:()=>({outputs:[{dims:p,dataType:a}],dispatchGroup:{x:Math.ceil(d/64/4)},programUniforms:[{type:12,data:Math.ceil(ye.size(p)/4)},...dt(i,l,p)]})}},vs=(e,r,t,s,o,n)=>{e.compute(nh(r,o??"",e.inputs[0],e.inputs[1],t,s,n))},oh=e=>{vs(e,"Add",(r,t)=>`${r}+${t}`)},ah=e=>{vs(e,"Div",(r,t)=>`${r}/${t}`)},ih=e=>{vs(e,"Equal",{scalar:(r,t)=>`u32(${r}==${t})`,vector:(r,t)=>`vec4<u32>(${r}==${t})`},void 0,void 0,9)},lh=e=>{vs(e,"Mul",(r,t)=>`${r}*${t}`)},ch=e=>{let r=Ie("input",e.inputs[0].dataType,e.inputs[0].dims).type.value;vs(e,"Pow",{scalar:(t,s)=>`pow_custom(${t},${s})`,vector:(t,s)=>`pow_vector_custom(${t},${s})`},`
    fn pow_custom(a : ${r}, b : ${r}) -> ${r} {
      if (b == ${r}(0.0)) {
        return ${r}(1.0);
      } else if (a < ${r}(0.0) && f32(b) != floor(f32(b))) {
        return ${r}(pow(f32(a), f32(b))); // NaN
      }
      return select(sign(a), ${r}(1.0), round(f32(abs(b) % ${r}(2.0))) != 1.0) * ${r}(${r==="i32"?"round":""}(pow(f32(abs(a)), f32(b))));
    }
    fn pow_vector_custom(a : vec4<${r}>, b : vec4<${r}>) -> vec4<${r}> {
      // TODO: implement vectorized pow
      return vec4<${r}>(pow_custom(a.x, b.x), pow_custom(a.y, b.y), pow_custom(a.z, b.z), pow_custom(a.w, b.w));
    }
      `)},uh=e=>{vs(e,"Sub",(r,t)=>`${r}-${t}`)},dh=e=>{vs(e,"Greater",{scalar:(r,t)=>`u32(${r}>${t})`,vector:(r,t)=>`vec4<u32>(${r}>${t})`},void 0,void 0,9)},ph=e=>{vs(e,"Less",{scalar:(r,t)=>`u32(${r}<${t})`,vector:(r,t)=>`vec4<u32>(${r}<${t})`},void 0,void 0,9)},mh=e=>{vs(e,"GreaterOrEqual",{scalar:(r,t)=>`u32(${r}>=${t})`,vector:(r,t)=>`vec4<u32>(${r}>=${t})`},void 0,void 0,9)},hh=e=>{vs(e,"LessOrEqual",{scalar:(r,t)=>`u32(${r}<=${t})`,vector:(r,t)=>`vec4<u32>(${r}<=${t})`},void 0,void 0,9)}}),_h,fh,gh,wh,Mh,bh,Vx=Ue(()=>{Mt(),Pt(),pr(),Ct(),_h=(e,r)=>{if(!e||e.length<1)throw new Error("too few inputs");let t=0,s=e[t],o=s.dataType,n=s.dims.length;e.forEach((a,i)=>{if(i!==t){if(a.dataType!==o)throw new Error("input tensors should be one type");if(a.dims.length!==n)throw new Error("input tensors should have the same shape");a.dims.forEach((l,c)=>{if(c!==r&&l!==s.dims[c])throw new Error("non concat dimensions must match")})}})},fh=(e,r)=>`
  fn calculateInputIndex(index: u32) -> u32 {
    let sizeInConcatAxis = array<u32, ${e}u>(${r});
    for (var i: u32 = 0u; i < ${e}; i += 1u ) {
      if (index < sizeInConcatAxis[i]) {
        return i;
      }
    }
    return ${e}u;
  }`,gh=(e,r)=>{let t=e.length,s=[];for(let o=0;o<t;++o){let n=r.setByOffset("global_idx",e[o].getByIndices("indices"));t===1?s.push(n):o===0?s.push(`if (inputIndex == ${o}u) { ${n} }`):o===t-1?s.push(`else { ${n} }`):s.push(`else if (inputIndex == ${o}) { ${n} }`)}return s.join(`
`)},wh=(e,r,t,s)=>{let o=ye.size(t),n=new Array(e.length),a=new Array(e.length),i=0,l=[],c=[],p=[{type:12,data:o}];for(let b=0;b<e.length;++b)i+=e[b].dims[r],n[b]=i,c.push(e[b].dims.length),a[b]=Ie(`input${b}`,s,c[b]),l.push("rank"),p.push({type:12,data:n[b]});for(let b=0;b<e.length;++b)p.push(...dt(e[b].dims));p.push(...dt(t));let d=nt("output",s,t.length),u=d.indicesGet("indices",r),f=Array.from(Array(n.length).keys()).map(b=>`uniforms.sizeInConcatAxis${b}`).join(","),_=b=>`

  ${(()=>{b.registerUniform("outputSize","u32");for(let k=0;k<e.length;k++)b.registerUniform(`sizeInConcatAxis${k}`,"u32");return b.declareVariables(...a,d)})()}

  ${fh(n.length,f)}

  ${b.mainStart()}
    ${b.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.outputSize")}

    var indices = ${d.offsetToIndices("global_idx")};

    let inputIndex = calculateInputIndex(${u});
    if (inputIndex != 0u) {
      let sizeInConcatAxis = array<u32, ${n.length}u>(${f});
      ${u} -= sizeInConcatAxis[inputIndex - 1u];
    }

    ${gh(a,d)}
  }`;return{name:"Concat",shaderCache:{hint:`${r}`,inputDependencies:l},getRunData:()=>({outputs:[{dims:t,dataType:s}],dispatchGroup:{x:Math.ceil(o/64)},programUniforms:p}),getShaderSource:_}},Mh=(e,r)=>{let t=e.inputs,s=t[0].dims,o=ye.normalizeAxis(r.axis,s.length);_h(t,o);let n=s.slice();n[o]=t.reduce((i,l)=>i+(l.dims.length>o?l.dims[o]:0),0);let a=t.filter(i=>ye.size(i.dims)>0);e.compute(wh(a,o,n,t[0].dataType),{inputs:a})},bh=e=>Kt({axis:e.axis})}),gn,wn,Mn,bl,bn=Ue(()=>{Mt(),Pt(),gn=(e,r,t="f32")=>{switch(e.activation){case"Relu":return`value = max(value, ${r}(0.0));`;case"Sigmoid":return`value = (${r}(1.0) / (${r}(1.0) + exp(-value)));`;case"Clip":return`value = clamp(value, ${r}(${t}(uniforms.clip_min)), ${r}(${t}(uniforms.clip_max)));`;case"HardSigmoid":return`value = max(${r}(0.0), min(${r}(1.0), ${t}(uniforms.alpha) * value + ${t}(uniforms.beta)));`;case"LeakyRelu":return`value = select(${t}(uniforms.alpha) * value, value, value >= ${r}(0.0));`;case"Tanh":return`let e2x = exp(-2.0 * abs(value));
              value = sign(value) * (1.0 - e2x) / (1.0 + e2x);
        `;case"":return"";default:throw new Error(`Unsupported activation ${e.activation}`)}},wn=(e,r)=>{e.activation==="Clip"?r.push({type:1,data:e.clipMax},{type:1,data:e.clipMin}):e.activation==="HardSigmoid"?r.push({type:1,data:e.alpha},{type:1,data:e.beta}):e.activation==="LeakyRelu"&&r.push({type:1,data:e.alpha})},Mn=(e,r)=>{e.activation==="Clip"?r.push({name:"clip_max",type:"f32"},{name:"clip_min",type:"f32"}):e.activation==="HardSigmoid"?r.push({name:"alpha",type:"f32"},{name:"beta",type:"f32"}):e.activation==="LeakyRelu"&&r.push({name:"alpha",type:"f32"})},bl=e=>{let r=(e==null?void 0:e.activation)||"";if(r==="HardSigmoid"){let[t,s]=(e==null?void 0:e.activation_params)||[.2,.5];return{activation:r,alpha:t,beta:s}}else if(r==="Clip"){let[t,s]=(e==null?void 0:e.activation_params)||[Ud,Wd];return{activation:r,clipMax:s,clipMin:t}}else if(r==="LeakyRelu"){let[t]=(e==null?void 0:e.activation_params)||[.01];return{activation:r,alpha:t}}return{activation:r}}}),Lr,yh,yl=Ue(()=>{Lr=(e,r)=>{switch(e){case 1:return r;case 2:return`vec2<${r}>`;case 3:return`vec3<${r}>`;case 4:return`vec4<${r}>`;default:throw new Error(`${e}-component is not supported.`)}},yh=e=>`
      ${e?"value = value + getBiasByOutputCoords(coords);":""}
      `}),vh,Ux=Ue(()=>{vh=e=>`
fn getIndexFromCoords4D(coords : vec4<i32>, shape : vec4<i32>) -> i32 {
  return dot(coords, vec4<i32>(
      shape.y * shape.z * shape.w, shape.z * shape.w, shape.w, 1));
}
fn getOutputIndexFromCoords(coords : vec4<i32>) -> i32 {
  return dot(coords, vec4<i32>(
    i32(${e}.x), i32(${e}.y), i32(${e}.z), 1));
}
`}),bo,vl,xl=Ue(()=>{Mt(),Pt(),Ct(),bn(),bo=(e,r,t,s,o)=>{let n=s-t;return`
      ${Array.from({length:t}).map((a,i)=>`
      if (${ct(r.shape,i,r.rank)} != 1) {
        ${r.indicesSet(e,i,ct(o,i+n,s))}
      } else {
        ${r.indicesSet(e,i,0)}
      }`).join("")}
`},vl=(e,r,t,s,o=!1,n)=>{let a=e[0].dims,i=e[1].dims,l=a[a.length-2],c=i[i.length-1],p=a[a.length-1],d=lr(c),u=lr(p),f=lr(l),_=ye.size(t)/d/f,b=e.length>2,k=s?s.slice(0,-2):t.slice(0,-2),w=[ye.size(k),l,c],y=[{type:12,data:_},{type:12,data:l},{type:12,data:c},{type:12,data:p}];wn(r,y),y.push(...dt(k,a,i)),b&&y.push(...dt(e[2].dims)),y.push(...dt(w));let $=T=>{let M=cl("batch_dims",e[0].dataType,k.length),x=Ie("a",e[0].dataType,a.length,u),v=Ie("b",e[1].dataType,i.length,d),C=nt("output",e[0].dataType,w.length,d),F=Ir(C.type.tensor),D=gn(r,C.type.value,F),K=[x,v],U="";if(b){let q=o?d:1;K.push(Ie("bias",e[2].dataType,e[2].dims.length,q)),U=`${o?`value += bias[col / ${q}];`:`value += ${C.type.value}(bias[row + i]);`}`}let j=[{name:"output_size",type:"u32"},{name:"M",type:"u32"},{name:"N",type:"u32"},{name:"K",type:"u32"}];Mn(r,j);let se=()=>{let q=`var a_data: ${x.type.value};`;for(let re=0;re<u;re++)q+=`
              let b_data${re} = b[(b_offset + (k + ${re}) * uniforms.N + col) / ${d}];`;for(let re=0;re<f;re++){q+=`a_data = a[(a_offset + (row + ${re}) * uniforms.K + k) / ${u}];`;for(let Z=0;Z<u;Z++)q+=`
            values[${re}] = fma(${v.type.value}(a_data${u===1?"":`[${Z}]`}), b_data${Z}, values[${re}]);
`}return q};return`
  ${T.registerUniforms(j).registerInternalVariables(M).declareVariables(...K,C)}
  ${T.mainStart()}
    ${T.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.output_size")}
    let col = (global_idx % (uniforms.N / ${d})) * ${d};
    var index1 = global_idx / (uniforms.N / ${d});
    let stride1 = uniforms.M / ${f};
    let row = (index1 % stride1) * ${f};
    let batch = index1 / stride1;

    ${t.length===2?"":`let batch_indices = ${M.offsetToIndices("batch")};`}

    var a_indices: ${x.type.indices};
    ${bo("a_indices",x,x.rank-2,M.rank,"batch_indices")}
    ${x.indicesSet("a_indices",x.rank-2,0)}
    ${x.indicesSet("a_indices",x.rank-1,0)}
    let a_offset = ${x.indicesToOffset("a_indices")};

    var b_indices: ${v.type.indices};
    ${bo("b_indices",v,v.rank-2,M.rank,"batch_indices")}
    ${v.indicesSet("b_indices",v.rank-2,0)}
    ${v.indicesSet("b_indices",v.rank-1,0)}
    let b_offset = ${v.indicesToOffset("b_indices")};
    var values: array<${C.type.value}, ${f}>;
    for (var k: u32 = 0u; k < uniforms.K; k = k + ${u}) {
      ${se()}
    }
    for (var i = 0u; i < ${f}u; i++) {
      var value = values[i];
      ${U}
      ${D}
      let cur_indices = ${C.type.indices}(batch, row + i, col);
      let offset = ${C.indicesToOffset("cur_indices")};
      ${C.setByOffset(`offset / ${d}`,"value")};
    }
  }
  `};return{name:"MatMulNaive",shaderCache:{hint:`${r.activation};${d};${u};${f};${o}`,inputDependencies:b?["rank","rank","rank"]:["rank","rank"]},getRunData:()=>({outputs:[{dims:n?n(t):t,dataType:e[0].dataType}],dispatchGroup:{x:Math.ceil(_/64)},programUniforms:y}),getShaderSource:$}}}),xh,Th,Tl,El,Eh,Pl,Ph,ha,Cl=Ue(()=>{Mt(),Pt(),Ct(),bn(),xl(),yl(),xh=(e,r)=>e?`
        mm_Asub[inputRow][inputCol] = mm_readA(batch,
          kStart + inputRow,
          globalRowStart / innerElementSize + inputCol${r?", batchIndices":""});
        `:`
        mm_Asub[inputRow][inputCol] = mm_readA(batch,
          globalRow + innerRow,
          kStart / innerElementSize + inputCol${r?", batchIndices":""});
        `,Th=(e,r)=>e?`
        let ACached0 = mm_Asub[k * innerElementSize][localRow];
        let ACached1 = mm_Asub[k * innerElementSize + 1][localRow];
        let ACached2 = mm_Asub[k * innerElementSize + 2][localRow];
        ${r===3?"":"let ACached3 = mm_Asub[k * innerElementSize + 3][localRow];"}
        for (var i = 0; i < rowPerThread; i = i + 1) {
          acc[i] = BCached0 * ACached0[i] + acc[i];
          acc[i] = BCached1 * ACached1[i] + acc[i];
          acc[i] = BCached2 * ACached2[i] + acc[i];
          ${r===3?"":"acc[i] = BCached3 * ACached3[i] + acc[i];"}
        }`:`
        for (var i = 0; i < rowPerThread; i = i + 1) {
          let ACached = mm_Asub[tileRow + i][k];
          acc[i] = BCached0 * ACached.x + acc[i];
          acc[i] = BCached1 * ACached.y + acc[i];
          acc[i] = BCached2 * ACached.z + acc[i];
          ${r===3?"":"acc[i] = BCached3 * ACached.w + acc[i];"}
        }`,Tl=(e,r,t="f32",s,o=!1,n=32,a=!1,i=32)=>{let l=r[1]*e[1],c=r[0]*e[0],p=o?l:n,d=o?n:l,u=p/r[0],f=n/r[1];if(!((o&&u===4&&e[1]===4||!o&&(u===3||u===4))&&p%r[0]===0&&n%r[1]===0&&e[0]===4))throw new Error(`If transposeA ${o} is true, innerElementSize ${u} and workPerThread[1] ${e[1]} must be 4.
      Otherwise, innerElementSize ${u} must be 3 or 4.
  tileAWidth ${p} must be divisible by workgroupSize[0]${r[0]}. tileInner ${n} must be divisible by workgroupSize[1] ${r[1]}. colPerThread ${e[0]} must be 4.`);return`
var<workgroup> mm_Asub: array<array<vec${u}<${t}>, ${p/u}>, ${d}>;
var<workgroup> mm_Bsub: array<array<vec4<${t}>, ${c/e[0]}>, ${n}>;

const rowPerThread = ${e[1]};
const colPerThread = ${e[0]};
const innerElementSize = ${u};
const tileInner = ${n};

@compute @workgroup_size(${r[0]}, ${r[1]}, ${r[2]})
fn main(@builtin(local_invocation_id) localId : vec3<u32>,
        @builtin(global_invocation_id) globalId : vec3<u32>,
        @builtin(workgroup_id) workgroupId : vec3<u32>) {
  let localRow = i32(localId.y);
  let tileRow = localRow * rowPerThread;
  let tileCol = i32(localId.x);

  let globalRow =i32(globalId.y) * rowPerThread;
  let globalCol = i32(globalId.x);
  let batch = ${a?"0":"i32(globalId.z)"};
  ${s?`let batchIndices = ${s.offsetToIndices("u32(batch)")};`:""}
  let globalRowStart = i32(workgroupId.y) * ${l};

  let num_tiles = ${a?`${Math.ceil(i/n)}`:"(uniforms.dim_inner - 1) / tileInner + 1"};
  var kStart = ${a?`i32(globalId.z) * ${i}`:"0"};

  var acc: array<vec4<${t}>, rowPerThread>;

  // Loop over shared dimension.
  let tileRowB = localRow * ${f};
  for (var t = 0; t < num_tiles; t = t + 1) {
      // Load one tile of A into local memory.
      for (var innerRow = 0; innerRow < rowPerThread; innerRow = innerRow + 1) {
          let inputRow = tileRow + innerRow;
          let inputCol = tileCol;
          ${xh(o,s)}
      }

      // Load one tile of B into local memory.
      for (var innerRow = 0; innerRow < ${f}; innerRow = innerRow + 1) {
          let inputRow = tileRowB + innerRow;
          let inputCol = tileCol;
          mm_Bsub[inputRow][inputCol] = mm_readB(batch, kStart + inputRow, globalCol${s?", batchIndices":""});
      }
      kStart = kStart + tileInner;
      workgroupBarrier();

      // Compute acc values for a single thread.
      for (var k = 0; k < tileInner / innerElementSize; k = k + 1) {
          let BCached0 = mm_Bsub[k * innerElementSize][tileCol];
          let BCached1 = mm_Bsub[k * innerElementSize + 1][tileCol];
          let BCached2 = mm_Bsub[k * innerElementSize + 2][tileCol];
          ${u===3?"":"let BCached3 = mm_Bsub[k * innerElementSize + 3][tileCol];"}

          ${Th(o,u)}
      }

      workgroupBarrier();
  }

  for (var innerRow = 0; innerRow < rowPerThread; innerRow = innerRow + 1) {
      mm_write(batch, globalRow + innerRow, globalCol, acc[innerRow]);
  }
}`},El=(e,r)=>e?`
            mm_Asub[inputRow][inputCol] = mm_readA(batch,
              kStart + inputRow,
              globalRowStart + inputCol${r?", batchIndices":""});
            `:`
            mm_Asub[inputRow][inputCol] = mm_readA(batch,
              globalRowStart + inputRow,
              kStart + inputCol${r?", batchIndices":""});
            `,Eh=e=>e?"let ACached = mm_Asub[k][tileRow + innerRow];":"let ACached = mm_Asub[tileRow + innerRow][k];",Pl=(e,r,t="f32",s,o=!1,n=32,a=!1,i=32,l=!1)=>{let c=e[1]*r[1],p=e[0]*r[0],d=o?c:n,u=o?n:c;if(!(u%r[1]===0&&d%r[0]===0&&n%r[1]===0))throw new Error(`tileAHight ${u} must be divisible by workgroupSize[1]${r[1]}, tileAWidth ${d} must be divisible by workgroupSize[0]${r[0]}, tileInner ${n} must be divisible by workgroupSize[1]${r[1]}`);let f=u/r[1],_=d/r[0],b=n/r[1],k=l?`
    let localRow = i32(localId.y);
    let localCol = i32(localId.x);
    let globalRowStart = i32(workgroupId.y) * ${c};
    let globalColStart = i32(workgroupId.x) * ${p};

    // Loop over shared dimension.
    for (var t = 0; t < num_tiles; t = t + 1) {
      // Load one tile of A into local memory.
      for (var inputRow = localRow; inputRow < ${u}; inputRow = inputRow + ${r[1]}) {
        for (var inputCol = localCol; inputCol < ${d}; inputCol = inputCol + ${r[0]}) {
          ${El(o,s)}
        }
      }
      // Load one tile of B into local memory.
      for (var inputRow = localRow; inputRow < ${n}; inputRow = inputRow + ${r[1]}) {
            for (var inputCol = localCol; inputCol < ${p}; inputCol = inputCol + ${r[0]}) {
          mm_Bsub[inputRow][inputCol] = mm_readB(batch,
            kStart + inputRow,
            globalColStart + inputCol${s?", batchIndices":""});
        }
      }
      kStart = kStart + tileInner;
      workgroupBarrier();

      // Compute acc values for a single thread.
      var BCached : array<${t}, colPerThread>;
      for (var k = 0; k < tileInner; k = k + 1) {
        for (var inner = 0; inner < colPerThread; inner = inner + 1) {
          BCached[inner] = mm_Bsub[k][localCol + inner * ${r[0]}];
        }
        for (var innerRow = 0; innerRow < rowPerThread; innerRow = innerRow + 1) {
          let ACached = ${o?`mm_Asub[k][localRow + innerRow * ${r[1]}];`:`mm_Asub[localRow + innerRow * ${r[1]}][k];`}
          for (var innerCol = 0; innerCol < colPerThread; innerCol = innerCol + 1) {
            acc[innerRow][innerCol] = acc[innerRow][innerCol] +
                ACached * BCached[innerCol];
          }
        }
      }
      workgroupBarrier();
    }
    for (var innerRow = 0; innerRow < rowPerThread; innerRow = innerRow + 1) {
      let gRow = globalRowStart + localRow + innerRow * ${r[1]};
      for (var innerCol = 0; innerCol < colPerThread; innerCol = innerCol + 1) {
        let gCol = globalColStart + localCol + innerCol * ${r[0]};
        mm_write(batch, gRow, gCol, acc[innerRow][innerCol]);
      }
    }
    `:`
let tileRow = i32(localId.y) * rowPerThread;
let tileCol = i32(localId.x) * colPerThread;

let globalRow = i32(globalId.y) * rowPerThread;
let globalCol = i32(globalId.x) * colPerThread;
let globalRowStart = i32(workgroupId.y) * ${c};

let tileRowA = i32(localId.y) * ${f};
let tileColA = i32(localId.x) * ${_};
let tileRowB = i32(localId.y) * ${b};
// Loop over shared dimension.
for (var t = 0; t < num_tiles; t = t + 1) {
  // Load one tile of A into local memory.
  for (var innerRow = 0; innerRow < ${f}; innerRow = innerRow + 1) {
    for (var innerCol = 0; innerCol < ${_}; innerCol = innerCol + 1) {
      let inputRow = tileRowA + innerRow;
      let inputCol = tileColA + innerCol;
      ${El(o,s)}
    }
  }

  // Load one tile of B into local memory.
  for (var innerRow = 0; innerRow < ${b}; innerRow = innerRow + 1) {
    for (var innerCol = 0; innerCol < colPerThread; innerCol = innerCol + 1) {
      let inputRow = tileRowB + innerRow;
      let inputCol = tileCol + innerCol;
      mm_Bsub[inputRow][inputCol] = mm_readB(batch,
        kStart + inputRow,
        globalCol + innerCol${s?", batchIndices":""});
    }
  }
  kStart = kStart + tileInner;
  workgroupBarrier();

  // Compute acc values for a single thread.
  var BCached : array<${t}, colPerThread>;
  for (var k = 0; k < tileInner; k = k + 1) {
    for (var inner = 0; inner < colPerThread; inner = inner + 1) {
      BCached[inner] = mm_Bsub[k][tileCol + inner];
    }

    for (var innerRow = 0; innerRow < rowPerThread; innerRow = innerRow + 1) {
      ${Eh(o)}
      for (var innerCol = 0; innerCol < colPerThread; innerCol = innerCol + 1) {
        acc[innerRow][innerCol] = acc[innerRow][innerCol] + ACached * BCached[innerCol];
      }
    }
  }

  workgroupBarrier();
}

for (var innerRow = 0; innerRow < rowPerThread; innerRow = innerRow + 1) {
  for (var innerCol = 0; innerCol < colPerThread; innerCol = innerCol + 1) {
    mm_write(batch, globalRow + innerRow, globalCol + innerCol,
        acc[innerRow][innerCol]);
  }
}
`;return`
  var<workgroup> mm_Asub : array<array<${t}, ${d}>, ${u}>;
  var<workgroup> mm_Bsub : array<array<${t}, ${p}>, ${n}>;
  const rowPerThread = ${e[1]};
  const colPerThread = ${e[0]};
  const tileInner = ${n};

@compute @workgroup_size(${r[0]}, ${r[1]}, ${r[2]})
fn main(@builtin(local_invocation_id) localId : vec3<u32>,
        @builtin(global_invocation_id) globalId : vec3<u32>,
        @builtin(workgroup_id) workgroupId : vec3<u32>) {
    let batch = ${a?"0":"i32(globalId.z)"};
    ${s?`let batchIndices = ${s.offsetToIndices("u32(batch)")};`:""}
    let num_tiles = ${a?`${Math.ceil(i/n)}`:"(uniforms.dim_inner - 1) / tileInner + 1"};
    var kStart = ${a?`i32(globalId.z) * ${i}`:"0"};

    var acc : array<array<${t}, colPerThread>, rowPerThread>;
    ${k}
  }
`},Ph=(e,r,t,s,o=!1)=>{let[n,a,i,l]=s,c=Ir(s[0].type.tensor);return`
    fn mm_readA(batch: i32, row: i32, colIn: i32, batchIndices: ${n.type.indices}) -> ${Lr(e,c)} {
      var value = ${Lr(e,c)}(0.0);
      let col = colIn * ${e};
      if(row < uniforms.dim_a_outer && col < uniforms.dim_inner)
      {
        var aIndices: ${a.type.indices};
        ${bo("aIndices",a,a.rank-2,n.rank,"batchIndices")}
        ${a.indicesSet("aIndices",a.rank-2,"u32(row)")}
        ${a.indicesSet("aIndices",a.rank-1,"u32(colIn)")}
        value = ${a.getByIndices("aIndices")};
      }
      return value;
    }

    fn mm_readB(batch: i32, row: i32, colIn: i32, batchIndices: ${n.type.indices}) -> ${Lr(e,c)} {
      var value = ${Lr(e,c)}(0.0);
      let col = colIn * ${e};
      if(row < uniforms.dim_inner && col < uniforms.dim_b_outer)
      {
        var bIndices: ${i.type.indices};
        ${bo("bIndices",i,i.rank-2,n.rank,"batchIndices")}
        ${i.indicesSet("bIndices",i.rank-2,"u32(row)")}
        ${i.indicesSet("bIndices",i.rank-1,"u32(colIn)")}
        value = ${i.getByIndices("bIndices")};
      }
      return value;
    }

    fn mm_write(batch: i32, row: i32, colIn: i32, valueIn: ${Lr(e,c)}) {
      let col = colIn * ${e};
      if (row < uniforms.dim_a_outer && col < uniforms.dim_b_outer) {
        var value = valueIn;
        let coords = vec3<i32>(batch, row, colIn);
        ${r?`value = value + ${o?"bias[colIn]":`${Lr(e,c)}(bias[row])`};`:""}
        ${t}
        ${l.setByIndices("vec3<u32>(coords)","value")}
      }
    }
    `},ha=(e,r,t,s,o=!1,n)=>{let a=e[0].dims,i=e[1].dims,l=a.slice(0,-2),c=i.slice(0,-2),p=s?s.slice(0,-2):t.slice(0,-2),d=ye.size(p),u=a[a.length-2],f=a[a.length-1],_=i[i.length-1],b=f%4===0&&_%4===0,k=u<=8?[4,1,1]:[4,4,1],w=[8,8,1],y=[Math.ceil(_/w[0]/k[0]),Math.ceil(u/w[1]/k[1]),Math.ceil(d/w[2]/k[2])],$=b?4:1,T=[...l,u,f/$],M=T.length,x=[...c,f,_/$],v=x.length,C=[d,u,_/$],F=[{type:6,data:u},{type:6,data:_},{type:6,data:f}];wn(r,F),F.push(...dt(p,T,x));let D=["rank","rank"],K=e.length>2;K&&(F.push(...dt(e[2].dims)),D.push("rank")),F.push(...dt(C));let U=j=>{let se=p.length,q=cl("batchDims",e[0].dataType,se,1),re=Ir(e[0].dataType),Z=Ie("a",e[0].dataType,M,$),ae=Ie("b",e[1].dataType,v,$),_e=nt("result",e[0].dataType,C.length,$),pe=[Z,ae];if(K){let le=o?$:1;pe.push(Ie("bias",e[2].dataType,e[2].dims.length,le))}let oe=[{name:"dim_a_outer",type:"i32"},{name:"dim_b_outer",type:"i32"},{name:"dim_inner",type:"i32"}];Mn(r,oe);let G=Ir(_e.type.tensor),B=gn(r,_e.type.value,G),Q=Ph($,K,B,[q,Z,ae,_e],o);return`
  ${j.registerUniforms(oe).registerInternalVariables(q).declareVariables(...pe,_e)}
  ${Q}
  ${b?Tl(k,w,re,q):Pl(k,w,re,q)}
                   `};return{name:"MatMul",shaderCache:{hint:`${k};${r.activation};${b};${o}`,inputDependencies:D},getRunData:()=>({outputs:[{dims:n?n(t):t,dataType:e[0].dataType}],dispatchGroup:{x:y[0],y:y[1],z:y[2]},programUniforms:F}),getShaderSource:U}}}),Ch,Sh,Wx=Ue(()=>{Mt(),Ns(),Ct(),bn(),yl(),Ux(),Cl(),Ch=(e,r,t,s,o=!1,n,a=4,i=4,l=4,c="f32")=>{let p=F=>{switch(F){case 1:return"resData = x[xIndex];";case 3:return`resData = vec3<${c}>(x[xIndex], x[xIndex + 1], x[xIndex + 2]);`;case 4:return"resData = x[xIndex / 4];";default:throw new Error(`innerElementSize ${F} is not supported.`)}},d=F=>{switch(F){case 1:return"return w[row * i32(uniforms.w_shape[3]) + colIn];";case 4:return"return w[row * i32(uniforms.w_shape[3]) / 4 + colIn];";default:throw new Error(`innerElementSize ${F} is not supported.`)}},u=e?`
    let coord = vec4<i32>(batch, xRow, xCol, xCh);
    `:`
    let coord = vec4<i32>(batch, xCh, xRow, xCol);
    `,f=e?`
    let coords = vec4<i32>(
      batch,
      row / outWidth,
      row % outWidth,
      col);
    `:`
    let coords = vec4<i32>(
      batch,
      row,
      col / outWidth,
      col % outWidth);
    `,_=e?"i32(uniforms.x_shape[1])":"i32(uniforms.x_shape[2])",b=e?"i32(uniforms.x_shape[2])":"i32(uniforms.x_shape[3])",k=e?"row":"col",w=e?"col":"row",y=`
    let inChannels = i32(uniforms.w_shape[2]);
    let outWidth = ${e?"i32(uniforms.result_shape[2])":"i32(uniforms.result_shape[3])"};
    let outRow = ${k} / outWidth;
    let outCol = ${k} % outWidth;

    let WRow = ${w} / (i32(uniforms.w_shape[1]) * inChannels);
    let WCol = ${w} / inChannels % i32(uniforms.w_shape[1]);
    let xRow = outRow * uniforms.stride[0] + uniforms.dilation[0] * WRow - uniforms.pad[0];
    let xCol = outCol * uniforms.stride[1] + uniforms.dilation[1] * WCol - uniforms.pad[1];
    let xCh = ${w} % inChannels;
    var resData = ${Lr(a,c)}(0.0);
    // The bounds checking is always needed since we use it to pad zero for
    // the 'same' padding type.
    if (xRow >= 0 && xRow < ${_} && xCol >= 0 && xCol < ${b}) {
      ${u}
      let xIndex = getIndexFromCoords4D(coord, vec4<i32>(uniforms.x_shape));
      ${p(a)}
    }
    return resData;`,$=e?r&&s?`
    let col = colIn * ${a};
    ${y}`:`
    let col = colIn * ${a};
    if (row < uniforms.dim_a_outer && col < uniforms.dim_inner) {
      ${y}
    }
    return ${Lr(a,c)}(0.0);`:s&&t?`
    let col = colIn * ${a};
    ${y}`:`
    let col = colIn * ${a};
    if (row < uniforms.dim_inner && col < uniforms.dim_b_outer) {
      ${y}
    }
    return ${Lr(a,c)}(0.0);`,T=e?s&&t?d(i):`
    let col = colIn * ${i};
    if (row < uniforms.dim_inner && col < uniforms.dim_b_outer) {
      ${d(i)}
    }
    return ${Lr(i,c)}(0.0);`:`
    let col = colIn * ${i};
    if (row < uniforms.dim_inner && col < uniforms.dim_a_outer) {
      ${d(i)}
    }
    return ${Lr(i,c)}(0.0);`,M=Lr(l,c),x=Lr(e?a:i,c),v=Lr(e?i:a,c),C=gn(n,M,c);return`
    fn mm_readA(batch: i32, row : i32, colIn : i32) -> ${x} {
      ${e?$:T}
    }

    fn mm_readB(batch: i32, row : i32, colIn : i32) -> ${v} {
      ${e?T:$}
    }

    fn mm_write(batch: i32, row : i32, colIn : i32, valueIn : ${M}) {
      let col = colIn * ${l};
      if (row < uniforms.dim_a_outer && col < uniforms.dim_b_outer)
      {
      var value = valueIn;
      let outWidth = ${e?"i32(uniforms.result_shape[2])":"i32(uniforms.result_shape[3])"};
      ${f}
      ${yh(o)}
      ${C}
      setOutputAtCoords(coords[0], coords[1], coords[2], coords[3], value);
      }
    }`},Sh=(e,r,t,s,o,n,a,i,l)=>{let c=r.format==="NHWC",p=c?e[0].dims[3]:e[0].dims[1],d=t[0],u=c?t[2]:t[3],f=c?t[1]:t[2],_=c?t[3]:t[1],b=c&&(p%4===0||p%3===0)&&_%4===0,k=c?_:u*f,w=c?u*f:_,y=[8,8,1],$=s<=8?[4,1,1]:[4,4,1],T=[Math.ceil(k/y[0]/$[0]),Math.ceil(w/y[1]/$[1]),Math.ceil(d/y[2]/$[2])];zt("verbose",()=>`[conv2d_mm_webgpu] dispatch = ${T}`);let M=b?c&&p%4!==0?3:4:1,x=y[1]*$[1],v=y[0]*$[0],C=Math.max(y[0]*M,y[1]),F=s%x===0,D=o%v===0,K=n%C===0,U=b?[M,4,4]:[1,1,1],j=[{type:6,data:s},{type:6,data:o},{type:6,data:n},{type:6,data:[r.pads[0],r.pads[1]]},{type:6,data:r.strides},{type:6,data:r.dilations}];wn(r,j),j.push(...dt(e[0].dims,e[1].dims));let se=["rank","rank"];a&&(j.push(...dt(e[2].dims)),se.push("rank")),j.push(...dt(t));let q=re=>{let Z=[{name:"dim_a_outer",type:"i32"},{name:"dim_b_outer",type:"i32"},{name:"dim_inner",type:"i32"},{name:"pad",type:"i32",length:2},{name:"stride",type:"i32",length:2},{name:"dilation",type:"i32",length:2}];Mn(r,Z);let ae=b?4:1,_e=Ir(e[0].dataType),pe=`
      fn setOutputAtIndex(flatIndex : i32, value : ${b?`vec4<${_e}>`:_e}) {
        result[flatIndex] = ${b?`vec4<${_e}>`:_e}(value);
      }
      fn setOutputAtCoords(d0 : i32, d1 : i32, d2 : i32, d3 : i32, value : ${b?`vec4<${_e}>`:_e}) {
        let flatIndex = getOutputIndexFromCoords(vec4<i32>(d0, d1, d2, d3));
        setOutputAtIndex(flatIndex ${b?"/ 4":""}, value);
      }`,oe=Ie("x",e[0].dataType,e[0].dims.length,M===3?1:M),G=Ie("w",e[1].dataType,e[1].dims.length,ae),B=[oe,G],Q=nt("result",e[0].dataType,t.length,ae);if(a){let le=Ie("bias",e[2].dataType,e[2].dims.length,ae);B.push(le),pe+=`
        fn getBiasByOutputCoords(coords : vec4<i32>) -> ${b?`vec4<${_e}>`:_e} {
          return bias[coords.${c?"w":"y"}${b?"/ 4":""}];
        }`}return`
        ${vh("uniforms.result_strides")}
        //struct Uniforms { xShape : vec4<i32>, wShape : vec4<i32>, outShape : vec4<i32>,
        //  outShapeStrides: vec3<i32>, filterDims : vec2<i32>, pad : vec2<i32>, stride : vec2<i32>,
        //  dilation : vec2<i32>, dimAOuter : i32, dimBOuter : i32, dimInner : i32 };
        ${re.registerUniforms(Z).declareVariables(...B,Q)}
        ${pe}
        ${Ch(c,F,D,K,a,r,U[0],U[1],U[2],_e)}
        ${b?Tl($,y,_e,void 0,!c,C):Pl($,y,_e,void 0,!c,C,!1,void 0,i)}`};return{name:"Conv2DMatMul",shaderCache:{hint:`${r.cacheKey};${M};${b};${F};${D};${K};${x};${v};${C}`,inputDependencies:se},getRunData:()=>({outputs:[{dims:l?l(t):t,dataType:e[0].dataType}],dispatchGroup:{x:T[0],y:T[1],z:T[2]},programUniforms:j}),getShaderSource:q}}}),$h,Sl,yo,kh,$l,Ih,Ah,Fh,Gx=Ue(()=>{Mt(),Ns(),Pt(),Ct(),bn(),yl(),$h=e=>{let r=1;for(let t=0;t<e.length;t++)r*=e[t];return r},Sl=e=>typeof e=="number"?[e,e,e]:e,yo=(e,r)=>r<=1?e:e+(e-1)*(r-1),kh=(e,r,t,s=1)=>{let o=yo(r,s);return Math.floor((e[0]*(t-1)-t+o)/2)},$l=(e,r,t,s,o)=>{o==null&&(o=kh(e,r[0],s[0]));let n=[0,0,0,t];for(let a=0;a<3;a++)e[a]+2*o>=r[a]&&(n[a]=Math.trunc((e[a]-r[a]+2*o)/s[a]+1));return n},Ih=(e,r,t,s,o,n,a,i,l,c)=>{let p,d,u,f;if(e==="VALID"&&(e=0),typeof e=="number"){p={top:e,bottom:e,left:e,right:e,front:e,back:e};let _=$l([r,t,s,1],[i,l,c],1,[o,n,a],e);d=_[0],u=_[1],f=_[2]}else if(Array.isArray(e)){if(!e.every((b,k,w)=>b===w[0]))throw Error(`Unsupported padding parameter: ${e}`);p={top:e[0],bottom:e[1],left:e[2],right:e[3],front:e[4],back:e[5]};let _=$l([r,t,s,1],[i,l,c],1,[o,n,a],e[0]);d=_[0],u=_[1],f=_[2]}else if(e==="SAME_UPPER"){d=Math.ceil(r/o),u=Math.ceil(t/n),f=Math.ceil(s/a);let _=(d-1)*o+i-r,b=(u-1)*n+l-t,k=(f-1)*a+c-s,w=Math.floor(_/2),y=_-w,$=Math.floor(b/2),T=b-$,M=Math.floor(k/2),x=k-M;p={top:$,bottom:T,left:M,right:x,front:w,back:y}}else throw Error(`Unknown padding parameter: ${e}`);return{padInfo:p,outDepth:d,outHeight:u,outWidth:f}},Ah=(e,r,t,s,o,n=!1,a="channelsLast")=>{let i,l,c,p,d;if(a==="channelsLast")[i,l,c,p,d]=e;else if(a==="channelsFirst")[i,d,l,c,p]=e;else throw new Error(`Unknown dataFormat ${a}`);let[u,,f,_,b]=r,[k,w,y]=Sl(t),[$,T,M]=Sl(s),x=yo(f,$),v=yo(_,T),C=yo(b,M),{padInfo:F,outDepth:D,outHeight:K,outWidth:U}=Ih(o,l,c,p,k,w,y,x,v,C),j=n?u*d:u,se=[0,0,0,0,0];return a==="channelsFirst"?se=[i,j,D,K,U]:a==="channelsLast"&&(se=[i,D,K,U,j]),{batchSize:i,dataFormat:a,inDepth:l,inHeight:c,inWidth:p,inChannels:d,outDepth:D,outHeight:K,outWidth:U,outChannels:j,padInfo:F,strideDepth:k,strideHeight:w,strideWidth:y,filterDepth:f,filterHeight:_,filterWidth:b,effectiveFilterDepth:x,effectiveFilterHeight:v,effectiveFilterWidth:C,dilationDepth:$,dilationHeight:T,dilationWidth:M,inShape:e,outShape:se,filterShape:r}},Fh=(e,r,t,s,o,n)=>{let a=n==="channelsLast";a?e[0].dims[3]:e[0].dims[1];let i=[64,1,1],l={x:t.map((k,w)=>w)},c=[Math.ceil($h(l.x.map(k=>t[k]))/i[0]),1,1];zt("verbose",()=>`[conv3d_naive_webgpu] dispatch = ${c}`);let p=1,d=ye.size(t),u=[{type:12,data:d},{type:12,data:s},{type:12,data:o},{type:12,data:r.strides},{type:12,data:r.dilations}];wn(r,u),u.push(...dt(e[0].dims,e[1].dims));let f=["rank","rank"],_=e.length===3;_&&(u.push(...dt(e[2].dims)),f.push("rank")),u.push(...dt(t));let b=k=>{let w=[{name:"output_size",type:"u32"},{name:"filter_dims",type:"u32",length:s.length},{name:"pads",type:"u32",length:o.length},{name:"strides",type:"u32",length:r.strides.length},{name:"dilations",type:"u32",length:r.dilations.length}];Mn(r,w);let y=1,$=Ir(e[0].dataType),T=Ie("x",e[0].dataType,e[0].dims.length,p),M=Ie("W",e[1].dataType,e[1].dims.length,y),x=[T,M],v=nt("result",e[0].dataType,t.length,y),C="";if(_){let K=Ie("bias",e[2].dataType,e[2].dims.length,y);x.push(K),C+=`
        fn getBiasByOutputCoords(coords : array<u32, 5>) -> ${$} {
          return bias[${a?ct("coords",4,5):ct("coords",1,5)}];
        }`}let F=Lr(p,$),D=gn(r,F,$);return`
            ${C}
            fn getX(d0 : u32, d1 : u32, d2 : u32, d3 : u32, d4 : u32) -> f32 {
              let aIndices = array<u32, 5>(d0, d1, d2, d3, d4);
              return ${T.getByIndices("aIndices")};
            }
            fn getW(d0 : u32, d1 : u32, d2 : u32, d3 : u32, d4 : u32) -> f32 {
              let aIndices = array<u32, 5>(d0, d1, d2, d3, d4);
              return ${M.getByIndices("aIndices")};
            }
          ${k.registerUniforms(w).declareVariables(...x,v)}
          ${k.mainStart()}
          ${k.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.output_size")}
              let coords = ${v.offsetToIndices("global_idx")};
              let batch = ${ct("coords",0,T.rank)};
              let d2 = ${a?ct("coords",T.rank-1,T.rank):ct("coords",1,T.rank)};
              let xFRCCorner = vec3<u32>(${a?ct("coords",1,T.rank):ct("coords",2,T.rank)},
              ${a?ct("coords",2,T.rank):ct("coords",3,T.rank)},
              ${a?ct("coords",3,T.rank):ct("coords",4,T.rank)}) * uniforms.strides - uniforms.pads;
              let xFCorner = xFRCCorner.x;
              let xRCorner = xFRCCorner.y;
              let xCCorner = xFRCCorner.z;
              let xShapeY = ${a?ct("uniforms.x_shape",1,T.rank):ct("uniforms.x_shape",2,T.rank)};
              let xShapeZ = ${a?ct("uniforms.x_shape",2,T.rank):ct("uniforms.x_shape",3,T.rank)};
              let xShapeW = ${a?ct("uniforms.x_shape",3,T.rank):ct("uniforms.x_shape",4,T.rank)};
              let xShapeU = ${a?ct("uniforms.x_shape",4,T.rank):ct("uniforms.x_shape",1,T.rank)};
              let inputDepthNearestVec4 = (xShapeU / 4) * 4;
              let inputDepthVec4Remainder = xShapeU % 4;

              var value = 0.0;
              for (var wF = 0u; wF < uniforms.filter_dims[0]; wF++) {
                let xF = xFCorner + wF * uniforms.dilations[0];
                if (xF < 0 || xF >= xShapeY) {
                  continue;
                }

                for (var wR = 0u; wR < uniforms.filter_dims[1]; wR++) {
                  let xR = xRCorner + wR * uniforms.dilations[1];
                  if (xR < 0 || xR >= xShapeZ) {
                    continue;
                  }

                  for (var wC = 0u; wC < uniforms.filter_dims[2]; wC++) {
                    let xC = xCCorner + wC * uniforms.dilations[2];
                    if (xC < 0 || xC >= xShapeW) {
                      continue;
                    }

                    for (var d1 = 0u; d1 < inputDepthNearestVec4; d1 += 4) {
                      ${a?`let xValues = vec4<f32>(
                               getX(batch, xF, xR, xC, d1),
                               getX(batch, xF, xR, xC, d1 + 1),
                               getX(batch, xF, xR, xC, d1 + 2),
                               getX(batch, xF, xR, xC, d1 + 3));
                            `:`let xValues = vec4<f32>(
                               getX(batch, d1, xF, xR, xC),
                               getX(batch, d1 + 1, xF, xR, xC),
                               getX(batch, d1 + 2, xF, xR, xC),
                               getX(batch, d1 + 3, xF, xR, xC));
                            `}
                            let wValues = vec4<f32>(
                              getW(d2, d1, wF, wR, wC),
                              getW(d2, d1 + 1, wF, wR, wC),
                              getW(d2, d1 + 2, wF, wR, wC),
                              getW(d2, d1 + 3, wF, wR, wC));
                      value += dot(xValues, wValues);
                    }
                    if (inputDepthVec4Remainder == 1) {
                        ${a?`value += getX(batch, xF, xR, xC, inputDepthNearestVec4)
                          * getW(d2, inputDepthNearestVec4, wF, wR, wC);`:`value += getX(batch, inputDepthNearestVec4, xF, xR, xC)
                          * getW(d2, inputDepthNearestVec4, wF, wR, wC);`}
                    } else if (inputDepthVec4Remainder == 2) {
                      ${a?`let xValues = vec2<f32>(
                        getX(batch, xF, xR, xC, inputDepthNearestVec4),
                        getX(batch, xF, xR, xC, inputDepthNearestVec4 + 1));
                      `:`let xValues = vec2<f32>(
                        getX(batch, inputDepthNearestVec4, xF, xR, xC),
                        getX(batch, inputDepthNearestVec4 + 1, xF, xR, xC));
                    `}
                    let wValues = vec2<f32>(
                      getW(d2, inputDepthNearestVec4, wF, wR, wC),
                      getW(d2, inputDepthNearestVec4 + 1, wF, wR, wC));
                      value += dot(xValues, wValues);
                    } else if (inputDepthVec4Remainder == 3) {
                      ${a?`let xValues = vec3<f32>(
                        getX(batch, xF, xR, xC, inputDepthNearestVec4),
                        getX(batch, xF, xR, xC, inputDepthNearestVec4 + 1),
                        getX(batch, xF, xR, xC, inputDepthNearestVec4 + 2));
                      `:`let xValues = vec3<f32>(
                        getX(batch, inputDepthNearestVec4, xF, xR, xC),
                        getX(batch, inputDepthNearestVec4 + 1, xF, xR, xC),
                        getX(batch, inputDepthNearestVec4 + 2, xF, xR, xC));
                    `}
                    let wValues = vec3<f32>(
                      getW(d2, inputDepthNearestVec4, wF, wR, wC),
                      getW(d2, inputDepthNearestVec4 + 1, wF, wR, wC),
                      getW(d2, inputDepthNearestVec4 + 2, wF, wR, wC));
                      value += dot(xValues, wValues);
                    }
                  }
                }
              }
              ${_?"value = value + getBiasByOutputCoords(coords)":""};
              ${D}
              result[global_idx] = f32(value);
          }`};return{name:"Conv3DNaive",shaderCache:{hint:`${r.cacheKey};${a};${p};${_}`,inputDependencies:f},getRunData:()=>({outputs:[{dims:t,dataType:e[0].dataType}],dispatchGroup:{x:c[0],y:c[1],z:c[2]},programUniforms:u}),getShaderSource:b}}}),Oh,Dh,Kx=Ue(()=>{Mt(),Pt(),Ct(),bn(),Oh=(e,r,t,s)=>{let o=e.length>2,n=o?"value += b[output_channel];":"",a=e[0].dims,i=e[1].dims,l=r.format==="NHWC",c=l?t[3]:t[1],p=c/r.group,d=l&&p>=4?lr(c):1,u=ye.size(t)/d,f=[{type:12,data:u},{type:12,data:r.dilations},{type:12,data:[r.strides[0],r.strides[1]]},{type:12,data:[r.pads[0],r.pads[1]]},{type:12,data:p}];wn(r,f),f.push(...dt(a,[i[0],i[1],i[2],i[3]/d]));let _=o?["rank","rank","rank"]:["rank","rank"];f.push(...dt([t[0],t[1],t[2],t[3]/d]));let b=k=>{let w=nt("output",e[0].dataType,t.length,d),y=Ir(w.type.tensor),$=gn(r,w.type.value,y),T=Ie("x",e[0].dataType,a.length),M=Ie("w",e[1].dataType,i.length,d),x=[T,M];o&&x.push(Ie("b",e[2].dataType,e[2].dims,d));let v=[{name:"output_size",type:"u32"},{name:"dilations",type:"u32",length:r.dilations.length},{name:"strides",type:"u32",length:2},{name:"pads",type:"u32",length:2},{name:"output_channels_per_group",type:"u32"}];Mn(r,v);let C=l?`
      for (var wHeight: u32 = 0u; wHeight < uniforms.w_shape[0]; wHeight++) {
        let xHeight = xRCCorner.x + wHeight * uniforms.dilations[0];

        if (xHeight < 0u || xHeight >= uniforms.x_shape[1]) {
          continue;
        }

        for (var wWidth: u32 = 0u; wWidth < uniforms.w_shape[1]; wWidth++) {
          let xWidth = xRCCorner.y + wWidth * uniforms.dilations[1];
          if (xWidth < 0u || xWidth >= uniforms.x_shape[2]) {
            continue;
          }

          for (var wInChannel: u32 = 0u; wInChannel < uniforms.w_shape[2]; wInChannel++) {
            let input_channel = in_channel_offset + wInChannel;
            let xVal = ${T.get("batch","xHeight","xWidth","input_channel")};
            let wVal = ${M.get("wHeight","wWidth","wInChannel","output_channel")};
            value += xVal * wVal;
          }
        }
      }
      `:`
      for (var wInChannel: u32 = 0u; wInChannel < uniforms.w_shape[1]; wInChannel++) {
        let input_channel = in_channel_offset + wInChannel;
        for (var wHeight: u32 = 0u; wHeight < uniforms.w_shape[2]; wHeight++) {
          let xHeight = xRCCorner.x + wHeight * uniforms.dilations[0];

          if (xHeight < 0u || xHeight >= uniforms.x_shape[2]) {
            continue;
          }

          for (var wWidth: u32 = 0u; wWidth < uniforms.w_shape[3]; wWidth++) {
            let xWidth = xRCCorner.y + wWidth * uniforms.dilations[1];
            if (xWidth < 0u || xWidth >= uniforms.x_shape[3]) {
              continue;
            }

            let xVal = ${T.get("batch","input_channel","xHeight","xWidth")};
            let wVal = ${M.get("output_channel","wInChannel","wHeight","wWidth")};
            value += xVal * wVal;
          }
        }
      }
      `;return`
  ${k.registerUniforms(v).declareVariables(...x,w)}

  ${k.mainStart()}
    ${k.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.output_size")}

    let outputIndices = ${w.offsetToIndices("global_idx")};
    let batch: u32 = outputIndices[0];
    let output_channel: u32 = outputIndices[${l?3:1}];
    let xRCCorner: vec2<u32> = vec2<u32>(outputIndices[${l?1:2}], outputIndices[${l?2:3}]) * uniforms.strides - uniforms.pads;
    let group_id: u32 = output_channel * ${d} / uniforms.output_channels_per_group;
    var in_channel_offset = group_id * uniforms.w_shape[${l?2:1}];

    var value: ${w.type.value} = ${w.type.value}(0);
    ${C}
    ${n}
    ${$}
    ${w.setByOffset("global_idx","value")}
  }`};return{name:"GroupedConv",shaderCache:{hint:`${r.cacheKey}_${d}`,inputDependencies:_},getRunData:()=>({outputs:[{dims:s?s(t):t,dataType:e[0].dataType}],dispatchGroup:{x:Math.ceil(u/64)},programUniforms:f}),getShaderSource:b}},Dh=(e,r,t,s)=>{let o=e.length>2,n=lr(t[3]),a=lr(t[2]),i=ye.size(t)/n/a,l=[e[0].dims[0],e[0].dims[1],e[0].dims[2],e[0].dims[3]/n],c=[e[1].dims[0],e[1].dims[1],e[1].dims[2],e[1].dims[3]/n],p=[t[0],t[1],t[2],t[3]/n],d=[{type:12,data:i},{type:6,data:[r.strides[0],r.strides[1]]},{type:6,data:[r.pads[0],r.pads[1]]}];wn(r,d),d.push(...dt(l,c,p));let u=(a-1)*r.strides[1]+c[1],f=_=>{let b=nt("output",e[0].dataType,p.length,n),k=Ir(b.type.tensor),w=gn(r,b.type.value,k),y=Ie("x",e[0].dataType,l.length,n),$=Ie("w",e[1].dataType,c.length,n),T=[y,$];o&&T.push(Ie("b",e[2].dataType,e[2].dims,n));let M=o?"value += b[output_channel];":"",x=[{name:"output_size",type:"u32"},{name:"strides",type:"i32",length:2},{name:"pads",type:"i32",length:2}];return Mn(r,x),`
  ${_.registerUniforms(x).declareVariables(...T,b)}
  ${_.mainStart()}
    ${_.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.output_size")}
    let width0 = uniforms.output_shape[3];
    let output_channel = global_idx % width0;
    var index1 = global_idx / width0;
    let width1 = uniforms.output_shape[2] / ${a}u;
    let col = (index1 % width1) * ${a}u;
    index1 = index1 / width1;
    let row = index1 % uniforms.output_shape[1];
    let batch = index1 / uniforms.output_shape[1];

    let x_corner = vec2<i32>(i32(row), i32(col)) * uniforms.strides - uniforms.pads;

    var x_vals: array<${y.type.value}, ${u}>;
    var values: array<${b.type.value}, ${a}>;
    let input_channel = output_channel;
    // Use constant instead of uniform can give better performance for w's height/width.
    for (var w_height: u32 = 0u; w_height < ${c[0]}; w_height++) {
      let x_height = x_corner.x + i32(w_height);
      if (x_height >= 0 && u32(x_height) < uniforms.x_shape[1]) {
        for (var i = 0; i < ${u}; i++) {
          let x_width = x_corner.y + i;
          if (x_width >= 0 && u32(x_width) < uniforms.x_shape[2]) {
            x_vals[i] = ${y.get("batch","u32(x_height)","u32(x_width)","input_channel")};
          } else {
            x_vals[i] = ${y.type.value}(0);
          }
        }
        for (var w_width: u32 = 0u; w_width < ${c[1]}; w_width++) {
          let w_val = ${$.get("w_height","w_width","0","output_channel")};
          for (var i = 0u; i < ${a}u; i++) {
            values[i] = fma(x_vals[i * u32(uniforms.strides[1]) + w_width], w_val, values[i]);
          }
        }
      }
    }

    for (var i = 0u; i < ${a}u; i++) {
      var value = values[i];
      ${M}
      ${w}
      ${b.set("batch","row","col + i","output_channel","value")};
    }
  }`};return{name:"GroupedConv-Vectorize",shaderCache:{hint:`${r.cacheKey};${n};${a};${u};${c[0]};${c[1]}`,inputDependencies:o?["rank","rank","type"]:["rank","rank"]},getRunData:()=>({outputs:[{dims:s?s(t):t,dataType:e[0].dataType}],dispatchGroup:{x:Math.ceil(i/64)},programUniforms:d}),getShaderSource:f}}}),Lh,_a,zh,fa,kl,Il,Bh,Rh,Al,Hx=Ue(()=>{Pt(),Wx(),Gx(),Cl(),Kx(),bn(),xl(),tn(),Lh=(e,r,t,s,o,n)=>{let a=e[0],i=e.slice(n?1:2,n?3:4),l=i.length,c=r[0],p=r.slice(2).map((u,f)=>u+(u-1)*(t[f]-1)),d=i.map((u,f)=>u+s[f]+s[f+l]).map((u,f)=>Math.floor((u-p[f]+o[f])/o[f]));return d.splice(0,0,a),d.splice(n?3:1,0,c),d},_a=[2,3,1,0],zh=(e,r)=>{if(!e||e.length!==2&&e.length!==3)throw new Error("Conv requires 2 or 3 inputs");if(e[0].dims.length>5)throw new Error("greater than 5D is not supported");if(e[0].dims.length!==e[1].dims.length)throw new Error("filter does not have same dimension as input");let t=e[0].dims[r.format==="NHWC"?e[0].dims.length-1:1],s=e[1].dims[1]*r.group;if(t!==s)throw new Error("FILTER_IN_CHANNEL should be equal to DATA_CHANNEL");if(e.length===3&&(e[2].dims.length!==1||e[1].dims[0]!==e[2].dims[0]))throw new Error("invalid bias");let o=e[0].dims.length-2;if(r.dilations.length!==o)throw new Error(`dilations should be ${o}D`);if(r.strides.length!==o)throw new Error(`strides should be ${o}D`);if(r.pads.length!==o*2)throw new Error(`pads should be ${o*2}D`);if(r.kernelShape.length!==0&&r.kernelShape.length!==e[1].dims.length-2)throw new Error("invalid kernel shape")},fa=(e,r)=>{let t=e.kernelShape.slice();t.length<r[1].dims.length-2&&t.push(...Array(r[1].dims.length-2-t.length).fill(0));for(let n=2;n<r[1].dims.length;++n)t[n-2]===0&&(t[n-2]=r[1].dims[n]);let s=e.pads.slice();aa.adjustPadsBasedOnAutoPad(r[0].dims,e.strides,e.dilations,t,s,e.format==="NHWC",e.autoPad);let o=Object.assign({},e);return Object.assign(o,{kernelShape:t,pads:s}),o},kl=e=>{let r=bl(e),t=e.format,s=["NOTSET","VALID","SAME_UPPER","SAME_LOWER"][e.auto_pad],o=e.dilations,n=e.group,a=e.kernel_shape,i=e.pads,l=e.strides,c=e.w_is_const();return{autoPad:s,format:t,dilations:o,group:n,kernelShape:a,pads:i,strides:l,wIsConst:c,...r,cacheKey:`${e.format};${r.activation};`}},Il=(e,r,t,s)=>{let o=t.format==="NHWC",n=Lh(r[0].dims,r[1].dims,t.dilations,t.pads,t.strides,o);if(t.group!==1){let x=[r[0]];if(o){let v=e.kernelCustomData.wT??e.compute(rs(r[1],_a),{inputs:[1],outputs:[t.wIsConst?-2:-1]})[0];t.wIsConst&&!e.kernelCustomData.wT&&(e.kernelCustomData.wT=v),x.push(v)}else x.push(r[1]);r.length===3&&x.push(r[2]),!e.adapterInfo.isArchitecture("ampere")&&o&&r[1].dims[0]===t.group&&r[1].dims[1]===1&&t.dilations[0]===1&&t.dilations[1]===1?e.compute(Dh(x,t,n,s),{inputs:x}):e.compute(Oh(x,t,n,s),{inputs:x});return}let a=r.length===3,i=r[0].dims[o?1:2],l=r[0].dims[o?2:3],c=r[0].dims[o?3:1],p=r[1].dims[2],d=r[1].dims[3],u=n[o?1:2],f=n[o?2:3],_=n[o?3:1],b=o&&p===i&&d===l&&t.pads[0]===0&&t.pads[1]===0;if(b||p===1&&d===1&&t.dilations[0]===1&&t.dilations[1]===1&&t.strides[0]===1&&t.strides[1]===1&&t.pads[0]===0&&t.pads[1]===0){let x=n[0],v,C,F,D=[];if(o){let j=e.kernelCustomData.wT??e.compute(rs(r[1],_a),{inputs:[1],outputs:[t.wIsConst?-2:-1]})[0];if(t.wIsConst&&!e.kernelCustomData.wT&&(e.kernelCustomData.wT=j),b){let se=i*l*c;v=r[0].reshape([1,x,se]),C=j.reshape([1,se,_]),F=[1,x,_]}else v=r[0].reshape([x,i*l,c]),C=j.reshape([1,c,_]),F=[x,u*f,_];D.push(v),D.push(C)}else v=r[0].reshape([x,c,i*l]),C=r[1].reshape([1,_,c]),F=[x,_,u*f],D.push(C),D.push(v);a&&D.push(r[2]);let K=F[2],U=D[0].dims[D[0].dims.length-1];K<8&&U<8?e.compute(vl(D,t,n,F,o,s),{inputs:D}):e.compute(ha(D,t,n,F,o,s),{inputs:D});return}let k=!0,w=e.kernelCustomData.wT??e.compute(rs(r[1],_a),{inputs:[1],outputs:[t.wIsConst?-2:-1]})[0];t.wIsConst&&!e.kernelCustomData.wT&&(e.kernelCustomData.wT=w);let y=[r[0],w];a&&y.push(r[2]);let $=o?u*f:_,T=o?_:u*f,M=p*d*c;e.compute(Sh(y,t,n,$,T,M,a,k,s),{inputs:y})},Bh=(e,r)=>{let t=r.format==="NHWC",s=[e.inputs[0].reshape(t?[e.inputs[0].dims[0],1,e.inputs[0].dims[1],e.inputs[0].dims[2]]:[e.inputs[0].dims[0],e.inputs[0].dims[1],1,e.inputs[0].dims[2]]),e.inputs[1].reshape([e.inputs[1].dims[0],e.inputs[1].dims[1],1,e.inputs[1].dims[2]])];e.inputs.length===3&&s.push(e.inputs[2]);let o=[0,r.pads[0],0,r.pads[1]],n=[1].concat(r.strides),a=[1].concat(r.dilations),i=[1].concat(r.kernelShape),l=fa({...r,pads:o,strides:n,dilations:a,kernelShape:i},s);Il(e,s,l,c=>t?[c[0],c[2],c[3]]:[c[0],c[1],c[3]])},Rh=(e,r,t)=>{let s=t.format==="NHWC"?"channelsLast":"channelsFirst",o=fa(t,r),n=t.autoPad==="NOTSET"?t.pads:t.autoPad,a=Ah(r[0].dims,r[1].dims,t.strides,t.dilations,n,!1,s);e.compute(Fh(r,o,a.outShape,[a.filterDepth,a.filterHeight,a.filterWidth],[a.padInfo.front,a.padInfo.top,a.padInfo.left],s))},Al=(e,r)=>{if(zh(e.inputs,r),e.inputs[0].dims.length===3)Bh(e,r);else if(e.inputs[0].dims.length===5)Rh(e,e.inputs,r);else{let t=fa(r,e.inputs);Il(e,e.inputs,t)}}}),jh,qx=Ue(()=>{Mt(),Ns(),Pt(),Ct(),jh=(e,r,t)=>{let s=e.length>2,o=r.outputShape,n=r.format==="NHWC",a=r.group,i=e[1].dims,l=i[2]/a,c=i[3],p=n?lr(l):1,d=n&&c===1&&l>=4,u=d?Math.floor(l/4)*4:Math.floor(l/p)*p,f=l-u,_=n?lr(c):1,b=n?c===1?p:_:1,k=ye.size(o)/_,w=[Math.ceil(k/64),1,1];zt("verbose",()=>`[conv2d_backprop_webgpu] dispatch = ${w}`);let y=["rank","rank"],$=[r.strides[0],r.strides[1]],T=[r.kernelShape[n?1:2],r.kernelShape[n?2:3]],M=[r.dilations[0],r.dilations[1]],x=[T[0]+(r.dilations[0]<=1?0:(r.kernelShape[n?1:2]-1)*(r.dilations[0]-1)),T[1]+(r.dilations[1]<=1?0:(r.kernelShape[n?2:3]-1)*(r.dilations[1]-1))],v=[x[0]-1-Math.floor((r.pads[0]+r.pads[2])/2),x[1]-1-Math.floor((r.pads[1]+r.pads[3])/2)],C=[{type:12,data:k},{type:12,data:$},{type:12,data:T},{type:12,data:M},{type:12,data:x},{type:6,data:v},{type:12,data:u},{type:12,data:l},{type:12,data:c},...dt(e[0].dims,e[1].dims)];s&&(C.push(...dt(e[2].dims)),y.push("rank")),C.push(...dt(o));let F=D=>{let K=[{name:"output_size",type:"u32"},{name:"strides",type:"u32",length:$.length},{name:"filter_dims",type:"u32",length:T.length},{name:"dilations",type:"u32",length:T.length},{name:"effective_filter_dims",type:"u32",length:x.length},{name:"pads",type:"i32",length:v.length},{name:"input_channels_per_group_int",type:"u32"},{name:"input_channels_per_group",type:"u32"},{name:"output_channels_per_group",type:"u32"}],U=Ir(e[0].dataType),j=n?1:2,se=n?2:3,q=n?3:1,re=Ie("W",e[1].dataType,e[1].dims.length,b),Z=Ie("Dy",e[0].dataType,e[0].dims.length,p),ae=[Z,re];s&&ae.push(Ie("bias",e[2].dataType,[o[q]].length,_));let _e=nt("result",e[0].dataType,o.length,_),pe=()=>{let B="";if(d)p===4?B+=`
        let xValue = ${Z.getByOffset("x_offset")};
        let wValue = ${re.getByOffset("w_offset")};
        dotProd = dotProd + dot(xValue, wValue);
        x_offset += 1u;
        w_offset += 1u;`:p===2?B+=`
          dotProd = dotProd + dot(vec4<${U}>(${Z.getByOffset("x_offset")}, ${Z.getByOffset("x_offset + 1u")}), vec4<${U}>(${re.getByOffset("w_offset")}, ${re.getByOffset("w_offset + 1u")}));
          x_offset += 2u;
          w_offset += 2u;`:p===1&&(B+=`
          dotProd = dotProd + dot(vec4<${U}>(${Z.getByOffset("x_offset")}, ${Z.getByOffset("x_offset + 1u")}, ${Z.getByOffset("x_offset + 2u")}, ${Z.getByOffset("x_offset + 3u")}), vec4<${U}>(${re.getByOffset("w_offset")}, ${re.getByOffset("w_offset + 1u")}, ${re.getByOffset("w_offset + 2u")}, ${re.getByOffset("w_offset + 3u")}));
          x_offset += 4u;
          w_offset += 4u;`);else if(B+=`
                  let xValue = ${n?Z.getByOffset(`${Z.indicesToOffset(`${Z.type.indices}(batch, idyR, idyC, inputChannel)`)} / ${p}`):Z.get("batch","inputChannel","idyR","idyC")};
        `,p===1)B+=`
          let w_offset = ${re.indicesToOffset(`${re.type.indices}(u32(wRPerm), u32(wCPerm), inputChannel, wOutChannel)`)};
          let wValue = ${re.getByOffset(`w_offset / ${b}`)};
          dotProd = dotProd + xValue * wValue;`;else for(let Q=0;Q<p;Q++)B+=`
            let wValue${Q} = ${re.getByOffset(`${re.indicesToOffset(`${re.type.indices}(u32(wRPerm), u32(wCPerm), inputChannel + ${Q}, wOutChannel)`)} / ${b}`)};
            dotProd = dotProd + xValue[${Q}] * wValue${Q};`;return B},oe=()=>{if(f===0)return"";if(!d)throw new Error(`packInputAs4 ${d} is not true.`);let B="";if(p===1){B+="dotProd = dotProd";for(let Q=0;Q<f;Q++)B+=`
            + ${Z.getByOffset(`x_offset + ${Q}`)} * ${re.getByOffset(`w_offset + ${Q}`)}`;B+=";"}else if(p===2){if(f!==2)throw new Error(`Invalid inputChannelsRemainder ${f}.`);B+=`
          let xValue = ${Z.getByOffset("x_offset")};
          let wValue = ${re.getByOffset("w_offset")};
          dotProd = dotProd + dot(xValue, wValue);`}return B},G=`
            let outputIndices = ${_e.offsetToIndices(`global_idx * ${_}`)};
            let batch = ${_e.indicesGet("outputIndices",0)};
            let d1 = ${_e.indicesGet("outputIndices",q)};
            let r = ${_e.indicesGet("outputIndices",j)};
            let c = ${_e.indicesGet("outputIndices",se)};
            let dyCorner = vec2<i32>(i32(r), i32(c)) - uniforms.pads;
            let dyRCorner = dyCorner.x;
            let dyCCorner = dyCorner.y;
            let groupId = d1 / uniforms.output_channels_per_group;
            let wOutChannel = d1 - groupId * uniforms.output_channels_per_group;
            // Convolve dy(?, ?, d2) with w(:, :, d1, d2) to compute dx(xR, xC, d1).
            // ? = to be determined. : = across all values in that axis.
            var dotProd = ${_e.type.value}(0.0);
            var wR: u32 = 0;
            if (uniforms.dilations.x == 1) {
              // Minimum wR >= 0 that satisfies (dyRCorner + wR) % (uniforms.strides.x) == 0
              wR = u32(((dyRCorner + i32(uniforms.strides.x) - 1) / i32(uniforms.strides.x)) * i32(uniforms.strides.x) - dyRCorner);
            }
            for (; wR < uniforms.effective_filter_dims.x; wR = wR + 1) {
              if (wR % uniforms.dilations.x != 0) {
                continue;
              }
              let dyR = (${U}(dyRCorner) + ${U}(wR)) / ${U}(uniforms.strides[0]);
              let wRPerm = uniforms.filter_dims.x - 1 - wR / uniforms.dilations.x;
              if (dyR < 0.0 || dyR >= ${U}(uniforms.Dy_shape[${j}]) || fract(dyR) > 0.0 ||
                  wRPerm < 0) {
                continue;
              }
              let idyR: u32 = u32(dyR);
              var wC: u32 = 0;
              if (uniforms.dilations.y == 1) {
                // Minimum wC >= 0 that satisfies (dyCCorner + wC) % (uniforms.strides.y) == 0
                wC = u32(((dyCCorner + i32(uniforms.strides.y) - 1) / i32(uniforms.strides.y)) * i32(uniforms.strides.y) - dyCCorner);
              }
              for (; wC < uniforms.effective_filter_dims.y; wC = wC + 1) {
                if (wC % uniforms.dilations.y != 0) {
                  continue;
                }
                let dyC = (${U}(dyCCorner) + ${U}(wC)) / ${U}(uniforms.strides.y);
                let wCPerm = uniforms.filter_dims.y - 1 - wC / uniforms.dilations.y;
                if (dyC < 0.0 || dyC >= ${U}(uniforms.Dy_shape[${se}]) ||
                    fract(dyC) > 0.0 || wCPerm < 0) {
                  continue;
                }
                let idyC: u32 = u32(dyC);
                var inputChannel = groupId * uniforms.input_channels_per_group;
                ${d?`
                var x_offset = ${Z.indicesToOffset(`${Z.type.indices}(batch, idyR, idyC, inputChannel)`)} / ${p};
                var w_offset = ${re.indicesToOffset(`${re.type.indices}(wRPerm, wCPerm, inputChannel, wOutChannel)`)} / ${b};
                  `:""}
                for (var d2: u32 = 0; d2 < uniforms.input_channels_per_group_int; d2 = d2 + ${d?4:p}) {
                  ${pe()}
                  inputChannel = inputChannel + ${d?4:p};
                }
                ${oe()}
                wC = wC + uniforms.strides.y - 1;
              }
              wR = wR + uniforms.strides[0] - 1;
            }
            let value = dotProd${s?` + bias[d1 / ${_}]`:""};
            ${_e.setByOffset("global_idx","value")};
          `;return`
    ${D.registerUniforms(K).declareVariables(...ae,_e)}
      ${D.mainStart()}
      ${D.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.output_size")};
    ${G}}`};return{name:"ConvTranspose2D",shaderCache:{hint:`${r.cacheKey};${p}${b}${_}${d}${f}`,inputDependencies:y},getRunData:()=>({dispatchGroup:{x:w[0],y:w[1],z:w[2]},outputs:[{dims:t?t(o):o,dataType:e[0].dataType}],programUniforms:C}),getShaderSource:F}}}),Nh,Vh,Uh,Fl,Wh,Gh,Ol,Kh,Hh,Qx=Ue(()=>{qx(),bn(),tn(),Nh=(e,r,t,s,o,n)=>(e-1)*r+t+(s-1)*o+1-n,Vh=(e,r,t,s,o)=>{let n=Math.floor(e/2);r==="SAME_UPPER"?(t[s]=n,t[o]=e-n):r==="SAME_LOWER"&&(t[s]=e-n,t[o]=n)},Uh=(e,r,t,s,o,n,a,i,l,c)=>{let p=e.length-2,d=c.length===0;l.length<p&&l.push(...Array(p-l.length).fill(0));let u=e[0],f=r[i?3:1]*o;for(let _=0,b=e.length-p-(i?1:0);_<p;++_,++b){let k=e[b],w=d?k*a[_]:c[_],y=Nh(k,a[_],n[_],r[b],t[_],w);Vh(y,s,n,_,_+p),d&&c.push(a[_]*(k-1)+l[_]+(r[b]-1)*t[_]+1-n[_]-n[_+p])}c.splice(0,0,u),c.splice(i?3:1,0,f)},Fl=(e,r)=>{let t=e.kernelShape.slice();if(e.kernelShape.length===0||e.kernelShape.reduce((d,u)=>d*u,1)===0){t.length=0;for(let d=2;d<r[1].dims.length;++d)t.push(r[1].dims[d])}let s=e.format==="NHWC";t.splice(0,0,r[1].dims[0]),t.splice(s?3:1,0,r[1].dims[1]);let o=e.pads.slice(),n=e.outputShape.slice(),a=e.outputPadding.slice(),i=r[0].dims,l=e.dilations.slice();if(l.reduce((d,u)=>d+u,0)===0){let d=r[0].dims.length-2;l=new Array(d).fill(1)}let c=e.strides.slice();if(c.reduce((d,u)=>d+u,0)===0){let d=r[0].dims.length-2;c=new Array(d).fill(1)}Uh(i,t,l,e.autoPad,e.group,o,c,s,a,n);let p=Object.assign({},e);return Object.assign(p,{kernelShape:t,pads:o,outputPadding:a,outputShape:n,dilations:l,strides:c}),p},Wh=e=>{let r=bl(e),t=e.format,s=["NOTSET","VALID","SAME_UPPER","SAME_LOWER"][typeof e.autoPad>"u"?0:e.autoPad],o=e.dilations,n=e.group,a=e.kernelShape,i=e.pads,l=e.strides,c=e.wIsConst(),p=e.outputPadding,d=e.outputShape;return{autoPad:s,format:t,dilations:o,group:n,kernelShape:a,outputPadding:p,outputShape:d,pads:i,strides:l,wIsConst:c,...r,cacheKey:`${e.format};${r.activation};`}},Gh=(e,r)=>{if(!e||e.length!==2&&e.length!==3)throw new Error("Conv requires 2 or 3 inputs");if(e[0].dims.length!==4&&e[0].dims.length!==3)throw new Error("currently only support 2-dimensional conv");if(e[0].dims.length!==e[1].dims.length)throw new Error("filter does not have same dimension as input");let t=e[0].dims[r.format==="NHWC"?e[0].dims.length-1:1],s=e[1].dims[0];if(t!==s)throw new Error("FILTER_IN_CHANNEL should be equal to DATA_CHANNEL");let o=e[1].dims[1]*r.group;if(e.length===3&&(e[2].dims.length!==1||e[2].dims[0]!==o))throw new Error("invalid bias");let n=e[0].dims.length-2;if(r.dilations.reduce((a,i)=>a+i,0)>0&&r.dilations.length!==n)throw new Error(`dilations should be ${n}D`);if(r.strides.reduce((a,i)=>a+i,0)>0&&r.strides.length!==n)throw new Error(`strides should be ${n}D`);if(r.pads.reduce((a,i)=>a+i,0)>0&&r.pads.length!==n*2)throw new Error(`pads should be ${n*2}D`);if(r.outputPadding.length!==n&&r.outputPadding.length!==0)throw new Error(`output_padding should be ${n}D`);if(r.kernelShape.reduce((a,i)=>a+i,0)>0&&r.kernelShape.length!==0&&r.kernelShape.length!==e[1].dims.length-2)throw new Error("invalid kernel shape");if(r.outputShape.length!==0&&r.outputShape.length!==e[0].dims.length-2)throw new Error("invalid output shape")},Ol=(e,r,t,s)=>{let o=e.kernelCustomData.wT??e.compute(rs(r[1],[2,3,0,1]),{inputs:[1],outputs:[t.wIsConst?-2:-1]})[0];t.wIsConst&&!e.kernelCustomData.wT&&(e.kernelCustomData.wT=o);let n=[r[0],o];r.length===3&&n.push(r[2]),e.compute(jh(n,t,s),{inputs:n})},Kh=(e,r)=>{let t=r.format==="NHWC",s=[e.inputs[0].reshape(t?[e.inputs[0].dims[0],1,e.inputs[0].dims[1],e.inputs[0].dims[2]]:[e.inputs[0].dims[0],e.inputs[0].dims[1],1,e.inputs[0].dims[2]]),e.inputs[1].reshape([e.inputs[1].dims[0],e.inputs[1].dims[1],1,e.inputs[1].dims[2]])];e.inputs.length===3&&s.push(e.inputs[2]);let o=r.kernelShape;(o.length===0||o[0]===0)&&(o=[e.inputs[1].dims[2]]);let n=r.dilations;(n.length===0||n[0]===0)&&(n=[1]);let a=r.strides;(a.length===0||a[0]===0)&&(a=[1]);let i=r.pads;i.length===0&&(i=[0,0]),i=[0,i[0],0,i[1]],a=[1].concat(a),n=[1].concat(n),o=[1].concat(o);let l=r.outputPadding;l=[0].concat(l);let c=Fl({...r,pads:i,strides:a,dilations:n,kernelShape:o,outputPadding:l},s);Ol(e,s,c,p=>t?[p[0],p[2],p[3]]:[p[0],p[1],p[3]])},Hh=(e,r)=>{if(Gh(e.inputs,r),e.inputs[0].dims.length===3)Kh(e,r);else{let t=Fl(r,e.inputs);Ol(e,e.inputs,t)}}}),qh,Qh,Xh,Xx=Ue(()=>{Mt(),Pt(),pr(),Ct(),qh=(e,r,t,s)=>{let o=ye.size(r),n=r.length,a=Ie("input",e,n),i=nt("output",e,n),l=t.dataType===6?t.getInt32Array()[0]:Number(t.getBigInt64Array()[0]),c=ye.normalizeAxis(l,n),p=d=>{let u=` i32(${a.indicesGet("inputIndices","uniforms.axis")}) `,f=ct("uniforms.input_shape","uniforms.axis",n),_=s.reverse?u+(s.exclusive?" + 1":""):"0",b=s.reverse?f:u+(s.exclusive?"":" + 1");return`
                ${d.registerUniform("outputSize","u32").registerUniform("axis","u32").declareVariables(a,i)}
                ${d.mainStart()}
                  ${d.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.outputSize")}
                  var inputIndices = ${i.offsetToIndices("global_idx")};
                  var sum = ${i.type.value}(0);
                  let first : i32 = ${_};
                  let last : i32 = ${b};
                  for (var i : i32 = first; i < last; i++) {
                    ${a.indicesSet("inputIndices","uniforms.axis","u32(i)")};
                    sum = sum + ${a.getByIndices("inputIndices")};
                  }
                  ${i.setByOffset("global_idx","sum")};
                }`};return{name:"CumSum",shaderCache:{hint:s.cacheKey,inputDependencies:["rank"]},getRunData:()=>({outputs:[{dims:r,dataType:e}],dispatchGroup:{x:Math.ceil(o/64)},programUniforms:[{type:12,data:o},{type:12,data:c},...dt(r,r)]}),getShaderSource:p}},Qh=(e,r)=>{let t=e.inputs[0].dims,s=e.inputs[0].dataType,o=e.inputs[1];e.compute(qh(s,t,o,r),{inputs:[0]})},Xh=e=>{let r=e.exclusive===1,t=e.reverse===1;return Kt({exclusive:r,reverse:t})}}),Jh,Yh,Zh,e_,t_,Jx=Ue(()=>{Mt(),Pt(),pr(),Ct(),Jh=e=>{if(!e||e.length!==1)throw new Error("DepthToSpace requires 1 input.");if(e[0].dims.length!==4)throw new Error("DepthToSpace requires 4D input.")},Yh=(e,r,t,s)=>{let o=[];o.push(`fn perm(i: ${s.type.indices}) -> ${t.type.indices} {
    var a: ${t.type.indices};`);for(let n=0;n<r;++n)o.push(t.indicesSet("a",e[n],`i[${n}]`));return o.push("return a;}"),o.join(`
`)},Zh=(e,r)=>{let t,s,o,n,a,i,l=r.format==="NHWC",c=r.blocksize,p=r.mode==="DCR";l?([t,s,o,n]=e.dims,a=p?[t,s,o,c,c,n/c**2]:[t,s,o,n/c**2,c,c],i=p?[0,1,3,2,4,5]:[0,1,4,2,5,3]):([t,s,o,n]=[e.dims[0],e.dims[2],e.dims[3],e.dims[1]],a=p?[t,c,c,n/c**2,s,o]:[t,n/c**2,c,c,s,o],i=p?[0,3,4,1,5,2]:[0,1,4,2,5,3]);let d=e.reshape(a),u=d.dims.length,f=e.dataType,_=Ie("a",f,u),b=nt("output",f,u),k=w=>`
  ${w.registerUniform("output_size","u32").declareVariables(_,b)}

  ${Yh(i,u,_,b)}

  ${w.mainStart()}
    ${w.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.output_size")}

    let indices = ${b.offsetToIndices("global_idx")};
    let aIndices = perm(indices);

    ${b.setByOffset("global_idx",_.getByIndices("aIndices"))}
  }`;return{name:"DepthToSpace",shaderCache:{hint:`${e.dims};${r.blocksize};${r.mode}`,inputDependencies:["rank"]},getRunData:w=>{let y=l?[t,s*c,o*c,n/c**2]:[t,n/c**2,s*c,o*c],$=ye.size(y),T=d.dims,M=ye.sortBasedOnPerm(T,i);return{outputs:[{dims:y,dataType:w[0].dataType}],dispatchGroup:{x:Math.ceil($/64)},programUniforms:[{type:12,data:$},...dt(T,M)]}},getShaderSource:k}},e_=(e,r)=>{Jh(e.inputs),e.compute(Zh(e.inputs[0],r))},t_=e=>Kt({blocksize:e.blocksize,mode:e.mode,format:e.format})}),ga,vo,Dl,r_,s_,n_,o_,Ll,a_,i_,l_,Yx=Ue(()=>{Mt(),Pt(),pr(),Ct(),ga="[a-zA-Z]|\\.\\.\\.",vo="("+ga+")+",Dl="^"+vo+"$",r_="("+vo+",)*"+vo,s_="^"+r_+"$",n_=class{constructor(e=-1){this.symbolToIndices=new Map,this.inputIndex=e}addSymbol(e,r){let t=this.symbolToIndices.get(e);t===void 0?t=[r]:t.push(r),this.symbolToIndices.set(e,t)}},o_=class{constructor(e,r){var o;this.equation=r,this.hasEllipsis=!1,this.symbolToInfo=new Map,this.lhs=new Array,this.outputDims=[];let[t,s]=r.includes("->")?r.split("->",2):[r,""];if(!t.match(RegExp(s_)))throw new Error("Invalid LHS term");if(t.split(",").forEach((n,a)=>{let i=e[a].dims.slice();if(!n.match(RegExp(Dl)))throw new Error("Invalid LHS term");let l=this.processTerm(n,!0,i,a);this.lhs.push(l)}),s==="")s+=[...this.symbolToInfo.entries()].filter(([n,a])=>a.count===1||n==="...").map(([n])=>n).join("");else if(!s.match(RegExp(vo)))throw new Error("Invalid RHS");(o=s.match(RegExp(ga,"g")))==null||o.forEach(n=>{if(n==="...")this.outputDims=this.outputDims.concat(this.ellipsisDims);else{let a=this.symbolToInfo.get(n);if(a===void 0)throw new Error("Invalid RHS symbol");this.outputDims.push(a.dimValue)}}),this.rhs=this.processTerm(s,!1,this.outputDims)}addSymbol(e,r,t){let s=this.symbolToInfo.get(e);if(s!==void 0){if(s.dimValue!==r&&s.count!==1)throw new Error("Dimension mismatch");s.count++,s.inputIndices.push(t)}else s={count:1,dimValue:r,inputIndices:[t]};this.symbolToInfo.set(e,s)}processTerm(e,r,t,s=-1){let o=t.length,n=!1,a=[],i=0;if(!e.match(RegExp(Dl))&&!r&&e!=="")throw new Error("Invalid LHS term");let l=e.match(RegExp(ga,"g")),c=new n_(s);return l==null||l.forEach((p,d)=>{if(p==="..."){if(n)throw new Error("Only one ellipsis is allowed per input term");n=!0;let u=o-l.length+1;if(u<0)throw new Error("Ellipsis out of bounds");if(a=t.slice(i,i+u),this.hasEllipsis){if(this.ellipsisDims.length!==a.length||this.ellipsisDims.toString()!==a.toString())throw new Error("Ellipsis dimensions mismatch")}else if(r)this.hasEllipsis=!0,this.ellipsisDims=a;else throw new Error("Ellipsis must be specified in the LHS");for(let f=0;f<a.length;f++){let _=String.fromCharCode(48+f);c.addSymbol(_,d+f),this.addSymbol(_,t[i++],s)}}else c.addSymbol(p,d+(this.hasEllipsis?this.ellipsisDims.length-1:0)),this.addSymbol(p,t[i++],s)}),c}},Ll=e=>e+"_max",a_=(e,r,t,s)=>{let o=e.map(c=>c.length).map((c,p)=>Ie(`input${p}`,r,c)),n=ye.size(s),a=nt("output",r,s.length),i=[...t.symbolToInfo.keys()].filter(c=>!t.rhs.symbolToIndices.has(c)),l=c=>{let p=[],d="var prod = 1.0;",u="var sum = 0.0;",f="sum += prod;",_=[],b=[],k=[],w=[],y=t.symbolToInfo.size===t.rhs.symbolToIndices.size;t.symbolToInfo.forEach((T,M)=>{var x;if(t.rhs.symbolToIndices.has(M)){let v=(x=t.rhs.symbolToIndices.get(M))==null?void 0:x[0];v!==void 0&&t.lhs.forEach((C,F)=>{if(T.inputIndices.includes(F)){let D=C.symbolToIndices.get(M);if(D===void 0)throw new Error("Invalid symbol error");D.forEach(K=>{p.push(`${o[F].indicesSet(`input${F}Indices`,K,a.indicesGet("outputIndices",v))}`)})}})}else t.lhs.forEach((v,C)=>{if(T.inputIndices.includes(C)){let F=v.symbolToIndices.get(M);if(F===void 0)throw new Error("Invalid symbol error");F.forEach(D=>{_.push(`${o[C].indicesSet(`input${C}Indices`,D,`${M}`)}`)}),w.push(`prod *= ${o[C].getByIndices(`input${C}Indices`)};`)}}),b.push(`for(var ${M}: u32 = 0; ${M} < uniforms.${Ll(M)}; ${M}++) {`),k.push("}")});let $=y?[...p,`let sum = ${o.map((T,M)=>T.getByIndices(`input${M}Indices`)).join(" * ")};`]:[...p,u,...b,..._,d,...w,f,...k];return`
            ${c.registerUniforms(i.map(T=>({name:`${Ll(T)}`,type:"u32"}))).registerUniform("outputSize","u32").declareVariables(...o,a)}

            ${c.mainStart()}
            ${c.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.outputSize")}
            var outputIndices = ${a.offsetToIndices("global_idx")};
            ${o.map((T,M)=>`var input${M}Indices: ${o[M].type.indices};`).join(`
`)}
            ${$.join(`
`)};
            ${a.setByOffset("global_idx","sum")};
          }`};return{name:"Einsum",shaderCache:{hint:t.equation,inputDependencies:e.map(()=>"rank")},getRunData:()=>{let c=i.filter(d=>t.symbolToInfo.has(d)).map(d=>{var u;return{type:12,data:((u=t.symbolToInfo.get(d))==null?void 0:u.dimValue)||0}});c.push({type:12,data:n});let p=e.map((d,u)=>[...dt(d)]).reduce((d,u)=>d.concat(u),c);return p.push(...dt(s)),{outputs:[{dims:s,dataType:r}],dispatchGroup:{x:Math.ceil(n/64)},programUniforms:p}},getShaderSource:l}},i_=(e,r)=>{let t=new o_(e.inputs,r.equation),s=t.outputDims,o=e.inputs.map((n,a)=>n.dims);e.compute(a_(o,e.inputs[0].dataType,t,s))},l_=e=>{let r=e.equation.replace(/\s+/g,"");return Kt({equation:r})}}),c_,zl,u_,d_,p_,Zx=Ue(()=>{Mt(),Pt(),Ct(),c_=e=>{if(!e||e.length!==2)throw new Error("Expand requires 2 input.");let r=e[0].dims,t=Array.from(e[1].getBigInt64Array(),Number),s=t.length<r.length?0:t.length-r.length,o=r.length<t.length?0:r.length-t.length;for(;s<t.length&&o<r.length;++s,++o)if(t[s]!==r[o]&&t[s]!==1&&r[o]!==1)throw new Error("Expand requires shape to be broadcastable to input")},zl=(e,r)=>{let t=e.length-r.length,s=[];for(let o=0;o<t;++o)s.push(e[o]);for(let o=0;o<r.length;++o)s.push(r[o]===1?e[o+t]:r[o]);return s},u_=(e,r)=>e.length>r.length?zl(e,r):zl(r,e),d_=e=>{let r=e[0].dims,t=Array.from(e[1].getBigInt64Array(),Number),s=u_(r,t),o=e[0].dataType,n=o===9||ye.size(r)===1,a=o===9||r.length>0&&r[r.length-1]%4===0?4:1,i=n||s.length>0&&s[s.length-1]%4===0?4:1,l=Math.ceil(ye.size(s)/i),c=d=>{let u=Ie("input",o,r.length,a),f=nt("output",o,s.length,i),_;if(o===9){let b=(k,w,y="")=>`
          let outputIndices${w} = ${f.offsetToIndices(`outputOffset + ${w}u`)};
          let offset${w} = ${u.broadcastedIndicesToOffset(`outputIndices${w}`,f)};
          let index${w} = offset${w} / 4u;
          let component${w} = offset${w} % 4u;
          ${k}[${w}] = ${y}(${u.getByOffset(`index${w}`)}[component${w}]);
        `;_=`
        let outputOffset = global_idx * ${i};
        var data = vec4<u32>(0);
        ${b("data",0,"u32")}
        ${b("data",1,"u32")}
        ${b("data",2,"u32")}
        ${b("data",3,"u32")}
        ${f.setByOffset("global_idx","data")}
      }`}else _=`
        let outputIndices = ${f.offsetToIndices(`global_idx * ${i}`)};
        let inputOffset = ${u.broadcastedIndicesToOffset("outputIndices",f)};
        let data = ${f.type.value}(${u.getByOffset(`inputOffset / ${a}`)});
        ${f.setByOffset("global_idx","data")}
      }`;return`
    ${d.registerUniform("vec_size","u32").declareVariables(u,f)}
    ${d.mainStart()}
    ${d.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.vec_size")}
    ${_}`},p=[{type:12,data:l},...dt(r,s)];return{name:"Expand",shaderCache:{hint:`${s.length};${a}${i}`,inputDependencies:["rank"]},getShaderSource:c,getRunData:()=>({outputs:[{dims:s,dataType:e[0].dataType}],dispatchGroup:{x:Math.ceil(l/64)},programUniforms:p})}},p_=e=>{c_(e.inputs),e.compute(d_(e.inputs),{inputs:[0]})}}),m_,h_,eT=Ue(()=>{Mt(),Pt(),Ct(),Ml(),m_=e=>{let r=e[0].dataType,t=ye.size(e[0].dims),s=ye.size(e[1].dims),o=s%4===0,n=a=>{let i=Ie("x",r,[1],4),l=Ie("bias",r,[1],4),c=nt("y",r,[1],4),p=[{name:"output_vec_size",type:"u32"},{name:"bias_size",type:"u32"}],d=f=>`
      let bias${f}_offset: u32 = (global_idx * 4 + ${f}) % uniforms.bias_size;
      let bias${f} = ${l.getByOffset(`bias${f}_offset / 4`)}[bias${f}_offset % 4];`,u=o?`
      let bias = ${l.getByOffset("global_idx % (uniforms.bias_size / 4)")};`:`${d(0)}${d(1)}${d(2)}${d(3)}
      let bias = ${i.type.value}(bias0, bias1, bias2, bias3);`;return`${a.registerUniforms(p).declareVariables(i,l,c)}

    ${gl(Vr(r))}

    ${a.mainStart(Kn)}
      ${a.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.output_vec_size")}

      let x = ${i.getByOffset("global_idx")};
      ${u}
      let x_in = x + bias;
      ${c.setByOffset("global_idx",wl("x_in"))}
    }`};return{name:"FastGeluWithBias",shaderCache:{hint:`${o}`,inputDependencies:["type","type"]},getShaderSource:n,getRunData:a=>({outputs:[{dims:a[0].dims,dataType:a[0].dataType}],programUniforms:[{type:12,data:Math.ceil(t/4)},{type:12,data:s}],dispatchGroup:{x:Math.ceil(t/Kn/4)}})}},h_=e=>{e.inputs.length<2||ye.size(e.inputs[1].dims)===0?qm(e):e.compute(m_(e.inputs))}}),__,f_,g_,w_,tT=Ue(()=>{Mt(),Pt(),pr(),Ct(),__=e=>{if(!e||e.length!==2)throw new Error("Gather requires 2 inputs.")},f_=(e,r)=>{let t=e[0].dims,s=e[1].dims,o=t.length,n=ye.normalizeAxis(r.axis,o),a=t.slice(0);a.splice(n,1,...s);let i=t[n],l=e[0].dataType===9?4:1,c=Math.ceil(ye.size(a)/l),p=[{type:12,data:c},{type:6,data:i},{type:12,data:n},...dt(e[0].dims,e[1].dims,a)],d=u=>{let f=Ie("data",e[0].dataType,e[0].dims.length,l),_=Ie("inputIndices",e[1].dataType,e[1].dims.length),b=nt("output",e[0].dataType,a.length,l),k=y=>{let $=s.length,T=`var indicesIndices${y}  = ${_.type.indices}(0);`;for(let M=0;M<$;M++)T+=`${$>1?`indicesIndices${y}[${M}]`:`indicesIndices${y}`} = ${a.length>1?`outputIndices${y}[uniforms.axis + ${M}]`:`outputIndices${y}`};`;T+=`
          var idx${y} = ${_.getByIndices(`indicesIndices${y}`)};
          if (idx${y} < 0) {
            idx${y} = idx${y} + uniforms.axisDimLimit;
          }
          var dataIndices${y} : ${f.type.indices};
        `;for(let M=0,x=0;M<o;M++)M===n?(T+=`${o>1?`dataIndices${y}[${M}]`:`dataIndices${y}`} = u32(idx${y});`,x+=$):(T+=`${o>1?`dataIndices${y}[${M}]`:`dataIndices${y}`} = ${a.length>1?`outputIndices${y}[${x}]`:`outputIndices${y}`};`,x++);return T},w;if(e[0].dataType===9){let y=($,T,M="")=>`
          let outputIndices${T} = ${b.offsetToIndices(`outputOffset + ${T}u`)};
          ${k(T)};
          let offset${T} = ${f.indicesToOffset(`dataIndices${T}`)};
          let index${T} = offset${T} / 4u;
          let component${T} = offset${T} % 4u;
          ${$}[${T}] = ${M}(${f.getByOffset(`index${T}`)}[component${T}]);
        `;w=`
        let outputOffset = global_idx * ${l};
        var value = vec4<u32>(0);
        ${y("value",0,"u32")}
        ${y("value",1,"u32")}
        ${y("value",2,"u32")}
        ${y("value",3,"u32")}
        ${b.setByOffset("global_idx","value")}
      `}else w=`
      let outputIndices = ${b.offsetToIndices("global_idx")};
      ${k("")};
      let value = ${f.getByIndices("dataIndices")};
      ${b.setByOffset("global_idx","value")};
      `;return`
      ${u.registerUniform("outputSize","u32").registerUniform("axisDimLimit","i32").registerUniform("axis","u32").declareVariables(f,_,b)}
      ${u.mainStart()}
        ${u.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.outputSize")}
        ${w}
      }`};return{name:"Gather",shaderCache:{hint:r.cacheKey,inputDependencies:["rank","rank"]},getRunData:()=>({outputs:[{dims:a,dataType:e[0].dataType}],dispatchGroup:{x:Math.ceil(c/64)},programUniforms:p}),getShaderSource:d}},g_=e=>Kt({axis:e.axis}),w_=(e,r)=>{let t=e.inputs;__(t),e.compute(f_(e.inputs,r))}}),M_,b_,y_,rT=Ue(()=>{Mt(),Pt(),Ct(),M_=(e,r,t,s,o,n,a,i,l)=>{let c=[{type:12,data:n},{type:12,data:s},{type:12,data:o},{type:12,data:t},{type:12,data:a},{type:12,data:i},{type:12,data:l}],p=[n];c.push(...dt(r.dims,p));let d=u=>{let f=Ie("indices_data",r.dataType,r.dims.length),_=nt("input_slice_offsets_data",12,1,1),b=[f,_],k=[{name:"output_size",type:"u32"},{name:"batch_dims",type:"u32"},{name:"input_dims",type:"u32",length:o.length},{name:"sizes_from_slice_dims_data",type:"u32",length:t.length},{name:"num_slices_per_batch",type:"u32"},{name:"input_batch_stride",type:"u32"},{name:"num_slice_dims",type:"u32"}];return`
  ${u.registerUniforms(k).declareVariables(...b)}
  ${u.mainStart()}
    ${u.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.output_size")}
    let batch_idx = global_idx / uniforms.num_slices_per_batch;
    let base_offset = batch_idx * uniforms.input_batch_stride;

    let slice_indices_base_offset = global_idx * uniforms.num_slice_dims;
    var relative_slice_offset = 0;
    for (var dim_idx = 0u; dim_idx < uniforms.num_slice_dims; dim_idx ++) {
      var index = i32(indices_data[dim_idx + slice_indices_base_offset].x);
      let input_dim_idx = uniforms.batch_dims + dim_idx;
      if (index < 0) {
        ${o.length===1?"index += i32(uniforms.input_dims);":"index += i32(uniforms.input_dims[input_dim_idx]);"}
      }
      ${t.length===1?"relative_slice_offset += index * i32(uniforms.sizes_from_slice_dims_data);":"relative_slice_offset += index * i32(uniforms.sizes_from_slice_dims_data[dim_idx]);"}
    }

    input_slice_offsets_data[global_idx] =  base_offset + u32(relative_slice_offset);
  }`};return e.compute({name:"computeSliceOffsets",shaderCache:{hint:`${o.length}_${t.length}`,inputDependencies:["rank"]},getRunData:()=>({outputs:[{dims:p,dataType:e.inputs[1].dataType}],dispatchGroup:{x:Math.ceil(n/64)},programUniforms:c}),getShaderSource:d},{inputs:[r],outputs:[-1]})[0]},b_=(e,r)=>{let t=e.inputs,s=t[0].dims,o=t[0].dataType,n=t[1].dims,a=n[n.length-1],i=ye.sizeToDimension(n,n.length-1),l=ye.sizeFromDimension(s,r.batchDims+a),c=ye.sizeToDimension(s,r.batchDims),p=ye.sizeFromDimension(s,r.batchDims),d=i/c,u=new Array(a),f=l;for(let T=0;T<a;++T)u[a-1-T]=f,f*=s[r.batchDims+a-1-T];let _=M_(e,t[1],u,r.batchDims,s,i,d,p,a),b=r.batchDims+a;if(b>s.length)throw new Error("last dimension of indices must not be larger than rank of input tensor");let k=n.slice(0,-1).concat(s.slice(b)),w=ye.size(k),y=[{type:12,data:w},{type:12,data:l},...dt(t[0].dims,_.dims,k)],$=T=>{let M=Ie("data",t[0].dataType,t[0].dims.length),x=Ie("slice_offsets",12,_.dims.length),v=nt("output",t[0].dataType,k.length);return`
          ${T.registerUniform("output_size","u32").registerUniform("slice_size","u32").declareVariables(M,x,v)}
            ${T.mainStart()}
            ${T.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.output_size")}
          let slice_offset = slice_offsets[global_idx / uniforms.slice_size];
          output[global_idx] = data[u32(slice_offset) + global_idx % uniforms.slice_size];
        }`};e.compute({name:"GatherND",shaderCache:{hint:r.cacheKey,inputDependencies:["rank","rank"]},getRunData:()=>({outputs:[{dims:k,dataType:o}],dispatchGroup:{x:Math.ceil(w/64)},programUniforms:y}),getShaderSource:$},{inputs:[t[0],_]})},y_=e=>({batchDims:e.batch_dims,cacheKey:""})}),v_,x_,T_,E_,sT=Ue(()=>{Mt(),Pt(),pr(),Ct(),v_=(e,r)=>{if(e.length<3||e.length>4)throw new Error("GatherBlockQuantized requires 3 or 4 inputs.");let t=ye.normalizeAxis(r.quantizeAxis,e[0].dims.length),s=r.blockSize,o=e[0],n=e[2],a=e.length===4?e[3]:void 0;if(n.dims.length!==o.dims.length||!o.dims.map((i,l)=>l===t?Math.ceil(i/s)===n.dims[l]:i===n.dims[l]).reduce((i,l)=>i&&l,!0))throw new Error("Scales must have the same rank as the input tensor and the dims should match except on gatherAxis.");if(a){if(a.dataType!==o.dataType)throw new Error("Zero point must have the same data type as the input tensor.");if(a.dims.length!==n.dims.length||!a.dims.map((i,l)=>i===n.dims[l]).reduce((i,l)=>i&&l,!0))throw new Error("Zero point must have the same rank as the input tensor and the dims should match except on quantizeAxis.")}},x_=(e,r)=>{let t=e[0].dims,s=e[1].dims,o=t.length,n=ye.normalizeAxis(r.gatherAxis,o),a=ye.normalizeAxis(r.quantizeAxis,o),i=t.slice(0);i.splice(n,1,...s);let l=ye.size(i),c=e[2].dataType,p=e[0].dataType===22,d=[{type:12,data:l},{type:12,data:a},{type:12,data:n},{type:12,data:r.blockSize},...dt(...e.map((f,_)=>f.dims),i)],u=f=>{let _=Ie("data",e[0].dataType,e[0].dims.length),b=Ie("inputIndices",e[1].dataType,e[1].dims.length),k=Ie("scales",e[2].dataType,e[2].dims.length),w=e.length>3?Ie("zeroPoint",e[3].dataType,e[3].dims.length):void 0,y=nt("output",c,i.length),$=[_,b,k];w&&$.push(w);let T=[{name:"output_size",type:"u32"},{name:"quantize_axis",type:"u32"},{name:"gather_axis",type:"u32"},{name:"block_size",type:"u32"}];return`
        ${f.registerUniforms(T).declareVariables(...$,y)}
        ${f.mainStart()}
        let output_indices = ${y.offsetToIndices("global_idx")};
        var indices_indices = ${b.type.indices}(0);
        ${s.length>1?`
          for (var i: u32 = 0; i < ${s.length}; i++) {
            let index = ${y.indicesGet("output_indices","uniforms.gather_axis + i")};
            ${b.indicesSet("indices_indices","i","index")};
          }`:`indices_indices = ${y.indicesGet("output_indices","uniforms.gather_axis")};`};
        var data_indices = ${_.type.indices}(0);
        for (var i: u32 = 0; i < uniforms.gather_axis; i++) {
          let index = ${y.indicesGet("output_indices","i")};
          ${_.indicesSet("data_indices","i","index")};
        }
        var index_from_indices = ${b.getByIndices("indices_indices")};
        if (index_from_indices < 0) {
          index_from_indices += ${t[n]};
        }
        ${_.indicesSet("data_indices","uniforms.gather_axis","u32(index_from_indices)")};
        for (var i = uniforms.gather_axis + 1; i < ${i.length}; i++) {
          let index = ${y.indicesGet("output_indices",`i + ${s.length} - 1`)};
          ${_.indicesSet("data_indices","i","index")};
        }
        let data_offset = ${_.indicesToOffset("data_indices")};
        let data_index = data_offset % 8;
        // Convert 4-bit packed data to 8-bit packed data.
        let packed_4bit_quantized_data = ${_.getByOffset("data_offset / 8")};
        let packed_8bit_quantized_data = (packed_4bit_quantized_data >> (4 * (data_index % 2))) & 0x0f0f0f0f;
        let quantized_data_vec = ${p?"unpack4xI8":"unpack4xU8"}(u32(packed_8bit_quantized_data));
        let quantized_data = quantized_data_vec[data_index / 2];
        var scale_indices = data_indices;
        let quantize_axis_index = ${k.indicesGet("data_indices","uniforms.quantize_axis")} / uniforms.block_size;
        ${k.indicesSet("scale_indices","uniforms.quantize_axis","quantize_axis_index")};
        var scale = ${k.getByIndices("scale_indices")};
        ${w?`
              let zero_point_indices = scale_indices;
              let zero_point_offset = ${w.indicesToOffset("zero_point_indices")};
              let zero_point_index = zero_point_offset % 8;
              let packed_4bit_zero_points = ${w.getByOffset("zero_point_offset / 8")};
              let packed_8bit_zero_points = (packed_4bit_zero_points >> (4 * (zero_point_index % 2))) & 0x0f0f0f0f;
              let zero_point_vec = ${p?"unpack4xI8":"unpack4xU8"}(u32(packed_8bit_zero_points));
              let zero_point = zero_point_vec[zero_point_index / 2];`:"var zero_point = 0"};
        let dequantized_data = ${Vr(c)}(quantized_data - zero_point) * scale;
        ${y.setByOffset("global_idx","dequantized_data")};
    }`};return{name:"GatherBlockQuantized",shaderCache:{hint:`${r.cacheKey};${e.filter((f,_)=>_!==1).map(f=>f.dims.join("_")).join(";")}`,inputDependencies:Array.from({length:e.length},(f,_)=>"rank")},getRunData:()=>({outputs:[{dims:i,dataType:c}],dispatchGroup:{x:Math.ceil(l/64)},programUniforms:d}),getShaderSource:u}},T_=(e,r)=>{let t=e.inputs;v_(t,r),e.compute(x_(e.inputs,r))},E_=e=>Kt({blockSize:e.blockSize,gatherAxis:e.gatherAxis,quantizeAxis:e.quantizeAxis})}),P_,C_,S_,$_,nT=Ue(()=>{Mt(),Pt(),pr(),Ct(),P_=e=>{if(!e||e.length!==2)throw new Error("GatherElements requires 2 inputs.");if(e[0].dims.length<1)throw new Error("GatherElements requires that the data input be rank >= 1.");if(e[0].dims.length!==e[1].dims.length)throw new Error(`GatherElements requires that the data input and
                     indices input tensors be of same rank.`)},C_=(e,r)=>{let t=e[0].dims,s=e[0].dataType,o=t.length,n=e[1].dims,a=e[1].dataType,i=ye.normalizeAxis(r.axis,o),l=t[i],c=n.slice(0),p=ye.size(c),d=Ie("input",s,o),u=Ie("indicesInput",a,n.length),f=nt("output",s,c.length),_=[{type:12,data:p},{type:6,data:l},{type:12,data:i}];return _.push(...dt(t,n,c)),{name:"GatherElements",shaderCache:{inputDependencies:["rank","rank"]},getRunData:()=>({outputs:[{dims:c,dataType:e[0].dataType}],dispatchGroup:{x:Math.ceil(p/64)},programUniforms:_}),getShaderSource:b=>`
      ${b.registerUniform("outputSize","u32").registerUniform("axisDimLimit","i32").registerUniform("axis","u32").declareVariables(d,u,f)}
      ${b.mainStart()}
      ${b.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.outputSize")}

      let outputIndices = ${f.offsetToIndices("global_idx")};

      var idx = ${u.getByOffset("global_idx")};
      if (idx < 0) {
        idx = idx + uniforms.axisDimLimit;
      }
      var inputIndices = ${d.type.indices}(outputIndices);
      ${d.indicesSet("inputIndices","uniforms.axis","u32(idx)")};
      let value = ${d.getByIndices("inputIndices")};

      ${f.setByOffset("global_idx","value")};
  }`}},S_=e=>Kt({axis:e.axis}),$_=(e,r)=>{let t=e.inputs;P_(t),e.compute(C_(e.inputs,r))}}),k_,I_,A_,F_,oT=Ue(()=>{Mt(),Pt(),Ct(),k_=e=>{if(!e)throw new Error("Input is missing");if(e.length<2||e.length>3)throw new Error("Invaid input number.");if(e.length===3&&e[2].dims.length>2)throw new Error("Invalid input shape of C");if(e[0].dataType!==e[1].dataType||e.length===3&&e[0].dataType!==e[2].dataType)throw new Error("Input types are mismatched")},I_=(e,r)=>{let t=e[0].dims.slice(),s=e[1].dims.slice(),[o,n,a]=Vd.getShapeOfGemmResult(t,r.transA,s,r.transB,e.length===3?e[2].dims:void 0),i=[o,n];if(!i)throw new Error("Can't use gemm on the given tensors");let l=16,c=Math.ceil(n/l),p=Math.ceil(o/l),d=!0,u=ye.size(i),f=[{type:12,data:d?c:u},{type:12,data:o},{type:12,data:n},{type:12,data:a},{type:1,data:r.alpha},{type:1,data:r.beta}],_=["type","type"];e.length===3&&(f.push(...dt(e[2].dims)),_.push("rank")),f.push(...dt(i));let b=w=>{let y="";r.transA&&r.transB?y="value += a[k * uniforms.M + m] * b[n * uniforms.K + k];":r.transA&&!r.transB?y="value += a[k * uniforms.M + m] * b[k * uniforms.N + n];":!r.transA&&r.transB?y="value += a[m * uniforms.K + k] * b[n * uniforms.K + k];":!r.transA&&!r.transB&&(y="value += a[m * uniforms.K + k] * b[k * uniforms.N + n];");let $=r.alpha===1?"":"value *= uniforms.alpha;",T=Ie("a",e[0].dataType,e[0].dims),M=Ie("b",e[1].dataType,e[1].dims),x=T.type.value,v=null,C=[T,M];e.length===3&&(v=Ie("c",e[2].dataType,e[2].dims.length),C.push(v));let F=nt("output",e[0].dataType,i.length);C.push(F);let D=[{name:"output_size",type:"u32"},{name:"M",type:"u32"},{name:"N",type:"u32"},{name:"K",type:"u32"},{name:"alpha",type:"f32"},{name:"beta",type:"f32"}];return`
  ${w.registerUniforms(D).declareVariables(...C)}

  ${w.mainStart()}
    ${w.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.output_size")}

    let m = global_idx / uniforms.N;
    let n = global_idx % uniforms.N;

    var value = ${x}(0);
    for (var k: u32 = 0u; k < uniforms.K; k++) {
      ${y}
    }

    ${$}
    ${v!=null?`let cOffset = ${v.broadcastedIndicesToOffset("vec2(m, n)",F)}; value += ${x}(uniforms.beta) * ${v.getByOffset("cOffset")};`:""}
    output[global_idx] = value;
  }`},k=w=>{let y=Ie("a",e[0].dataType,e[0].dims),$=Ie("b",e[1].dataType,e[1].dims),T=null,M=[y,$];e.length===3&&(T=Ie("c",e[2].dataType,e[2].dims.length),M.push(T));let x=nt("output",e[0].dataType,i.length);M.push(x);let v=[{name:"num_tile_n",type:"u32"},{name:"M",type:"u32"},{name:"N",type:"u32"},{name:"K",type:"u32"},{name:"alpha",type:"f32"},{name:"beta",type:"f32"}],C="",F="";r.transA&&r.transB?(F=`
      var col = tile_row_start + local_id.x;
      var row = k_start + local_id.y;
      if (col < uniforms.M && row < uniforms.K) {
        tile_a[local_id.y][local_id.x] = a[row * uniforms.M + col];
      } else {
        tile_a[local_id.y][local_id.x] = ${y.type.value}(0);
      }

      col = k_start + local_id.x;
      row = tile_col_start + local_id.y;
      if (col < uniforms.K && row < uniforms.N) {
        tile_b[local_id.y][local_id.x] = b[row * uniforms.K + col];
      } else {
        tile_b[local_id.y][local_id.x] = ${$.type.value}(0);
      }
      `,C="value += tile_a[k][local_id.y] * tile_b[local_id.x][k];"):r.transA&&!r.transB?(F=`
      var col = tile_row_start + local_id.x;
      var row = k_start + local_id.y;
      if (col < uniforms.M && row < uniforms.K) {
        tile_a[local_id.y][local_id.x] = a[row * uniforms.M + col];
      } else {
        tile_a[local_id.y][local_id.x] = ${y.type.value}(0);
      }

      col = tile_col_start + local_id.x;
      row = k_start + local_id.y;
      if (col < uniforms.N && row < uniforms.K) {
        tile_b[local_id.y][local_id.x] = b[row * uniforms.N + col];
      } else {
        tile_b[local_id.y][local_id.x] = ${$.type.value}(0);
      }
      `,C="value += tile_a[k][local_id.y] * tile_b[k][local_id.x];"):!r.transA&&r.transB?(F=`
      var col = k_start + local_id.x;
      var row = tile_row_start + local_id.y;
      if (col < uniforms.K && row < uniforms.M) {
        tile_a[local_id.y][local_id.x] = a[row * uniforms.K + col];
      } else {
        tile_a[local_id.y][local_id.x] = ${y.type.value}(0);
      }

      col = k_start + local_id.x;
      row = tile_col_start + local_id.y;
      if (col < uniforms.K && row < uniforms.N) {
        tile_b[local_id.y][local_id.x] = b[row * uniforms.K + col];
      } else {
        tile_b[local_id.y][local_id.x] = ${$.type.value}(0);
      }
      `,C="value += tile_a[local_id.y][k] * tile_b[local_id.x][k];"):!r.transA&&!r.transB&&(F=`
      var col = k_start + local_id.x;
      var row = tile_row_start + local_id.y;
      if (col < uniforms.K && row < uniforms.M) {
        tile_a[local_id.y][local_id.x] = a[row * uniforms.K + col];
      } else {
        tile_a[local_id.y][local_id.x] = ${y.type.value}(0);
      }

      col = tile_col_start + local_id.x;
      row = k_start + local_id.y;
      if (col < uniforms.N && row < uniforms.K) {
        tile_b[local_id.y][local_id.x] = b[row * uniforms.N + col];
      } else {
        tile_b[local_id.y][local_id.x] = ${$.type.value}(0);
      }
      `,C="value += tile_a[local_id.y][k] * tile_b[k][local_id.x];");let D=r.alpha===1?"":"value *= uniforms.alpha;";return`
  ${w.registerUniforms(v).declareVariables(...M)}
  var<workgroup> tile_a: array<array<${y.type.storage}, ${l}>, ${l}>;
  var<workgroup> tile_b: array<array<${$.type.storage}, ${l}>, ${l}>;
  ${w.mainStart([l,l,1])}
    let tile_col_start = (workgroup_index % uniforms.num_tile_n) * ${l};
    let tile_row_start = (workgroup_index / uniforms.num_tile_n) * ${l};
    let num_tiles = (uniforms.K - 1) / ${l} + 1;
    var k_start = 0u;
    var value = ${x.type.value}(0);
    for (var t: u32 = 0u; t < num_tiles; t++) {
      ${F}
      k_start = k_start + ${l};
      workgroupBarrier();

      for (var k: u32 = 0u; k < ${l}; k++) {
        ${C}
      }
      workgroupBarrier();
    }

    ${D}
    let m = tile_row_start + local_id.y;
    let n = tile_col_start + local_id.x;
    ${T!=null?`let cOffset = ${T.broadcastedIndicesToOffset("vec2(m, n)",x)}; value += ${x.type.value}(uniforms.beta) * ${T.getByOffset("cOffset")};`:""}
    if (m < uniforms.M && n < uniforms.N) {
      output[m * uniforms.N + n] = value;
    }
  }`};return d?{name:"GemmShared",shaderCache:{hint:`${r.cacheKey}`,inputDependencies:_},getRunData:()=>({outputs:[{dims:i,dataType:e[0].dataType}],dispatchGroup:{x:c*p},programUniforms:f}),getShaderSource:k}:{name:"Gemm",shaderCache:{hint:`${r.cacheKey}`,inputDependencies:_},getRunData:()=>({outputs:[{dims:i,dataType:e[0].dataType}],dispatchGroup:{x:Math.ceil(u/64)},programUniforms:f}),getShaderSource:b}},A_=e=>{let r=e.transA,t=e.transB,s=e.alpha,o=e.beta;return{transA:r,transB:t,alpha:s,beta:o,cacheKey:`${e.transA};${e.transB};${e.alpha===1}`}},F_=(e,r)=>{k_(e.inputs),e.compute(I_(e.inputs,r))}}),ks,Vs,yn,vn,O_,D_,L_,z_,B_,R_,j_,N_,V_,U_,aT=Ue(()=>{Mt(),Pt(),pr(),Ct(),[ks,Vs,yn,vn]=[0,1,2,3],O_=e=>{if(e[0].dims.length!==4)throw new Error("only 4-D tensor is supported.");if(e[0].dims.length!==e[1].dims.length)throw new Error("input dimensions must be equal to grid dimensions");if(e[0].dims.length-2!==e[1].dims[e[1].dims.length-1])throw new Error(`last dimension of grid must be equal to ${e[0].dims.length-2}`);if(e[0].dims[0]!==e[1].dims[0])throw new Error("grid batch size must match input batch size")},D_=`
  fn gs_get_cubic_coeffs(x: f32) -> vec4<f32> {
    let cubic_alpha = -0.75f;
    let x_abs = abs(x);
    var coeffs: vec4<f32>;
    coeffs[0] = (((cubic_alpha * (x_abs + 1) - 5 * cubic_alpha) * (x_abs + 1) + 8 * cubic_alpha) * (x_abs + 1) - 4 * cubic_alpha);
    coeffs[1] = (((cubic_alpha + 2) * x_abs - (cubic_alpha + 3)) * x_abs * x_abs + 1);
    coeffs[2] = (((cubic_alpha + 2) * (1 - x_abs) - (cubic_alpha + 3)) * (1 - x_abs) * (1 - x_abs) + 1);
    coeffs[3] = (((cubic_alpha * (2 - x_abs) - 5 * cubic_alpha) * (2 - x_abs) + 8 * cubic_alpha) * (2 - x_abs) - 4 * cubic_alpha);
    return coeffs;
  }
`,L_=e=>`
  fn gs_bicubic_interpolate(p: mat4x4<${e}>, x: f32, y: f32) -> ${e} {
    var v: vec4<f32>;
    var coeffs = gs_get_cubic_coeffs(x);
    for (var i = 0; i < 4; i++) {
      v[i] = coeffs[0] * p[i][0] + coeffs[1] * p[i][1] + coeffs[2] * p[i][2] + coeffs[3] * p[i][3];
    }
    coeffs = gs_get_cubic_coeffs(y);
    let pixel = ${e}(coeffs[0] * v[0] + coeffs[1] * v[1] + coeffs[2] * v[2] + coeffs[3] * v[3]);
    return pixel;
  }
`,z_=e=>`
  fn gs_denormalize(n: f32, length: i32) -> f32 {
    ${e.alignCorners===0?`
    // alignCorners: false => [-1, 1] to [-0.5, length - 0.5]
    return ((n + 1.0) * f32(length) - 1.0) / 2.0;
    `:`
    // alignCorners: true => [-1, 1] to [0, length - 1]
    return (n + 1.0) / 2.0 * (f32(length - 1));
    `}
  }
`,B_=e=>`
  ${e.paddingMode==="reflection"?`
      fn gs_reflect(x: i32, x_min: f32, x_max: f32) -> u32 {
        var dx = 0.0;
        var fx = f32(x);
        let range = x_max - x_min;
        if (fx < x_min) {
          dx = x_min - fx;
          let n = u32(dx / range);
          let r = dx - f32(n) * range;
          if (n % 2 == 0) {
            fx = x_min + r;
          } else {
            fx = x_max - r;
          }
        } else if (fx > x_max) {
          dx = fx - x_max;
          let n = u32(dx / range);
          let r = dx - f32(n) * range;
          if (n % 2 == 0) {
            fx = x_max - r;
          } else {
            fx = x_min + r;
          }
        }
        return u32(fx);
      }`:""}
`,R_=(e,r,t)=>`
  fn pixel_at_grid(r: i32, c: i32, H: i32, W: i32, batch: u32, channel: u32, border: vec4<f32>) -> ${r} {
     var pixel = ${r}(0);
     var indices = vec4<u32>(0);
     indices[${ks}] = batch;
     indices[${Vs}] = channel;`+(()=>{switch(t.paddingMode){case"zeros":return`
          if (r >= 0 && r < H && c >=0 && c < W) {
            indices[${yn}] = u32(r);
            indices[${vn}] = u32(c);
          } else {
            return ${r}(0);
          }
        `;case"border":return`
          indices[${yn}] = u32(clamp(r, 0, H - 1));
          indices[${vn}] = u32(clamp(c, 0, W - 1));
        `;case"reflection":return`
          indices[${yn}] = gs_reflect(r, border[1], border[3]);
          indices[${vn}] = gs_reflect(c, border[0], border[2]);
        `;default:throw new Error(`padding mode ${t.paddingMode} is not supported`)}})()+`
    return ${e.getByIndices("indices")};
  }
`,j_=(e,r,t)=>(()=>{switch(t.mode){case"nearest":return`
          let result = pixel_at_grid(i32(round(y)), i32(round(x)), H_in, W_in, indices[${ks}], indices[${Vs}], border);
        `;case"bilinear":return`
          let x1 = i32(floor(x));
          let y1 = i32(floor(y));
          let x2 = x1 + 1;
          let y2 = y1 + 1;

          let p11 = pixel_at_grid(y1, x1, H_in, W_in, indices[${ks}], indices[${Vs}], border);
          let p12 = pixel_at_grid(y1, x2, H_in, W_in, indices[${ks}], indices[${Vs}], border);
          let p21 = pixel_at_grid(y2, x1, H_in, W_in, indices[${ks}], indices[${Vs}], border);
          let p22 = pixel_at_grid(y2, x2, H_in, W_in, indices[${ks}], indices[${Vs}], border);

          let dx2 = ${r}(f32(x2) - x);
          let dx1 = ${r}(x - f32(x1));
          let dy2 = ${r}(f32(y2) - y);
          let dy1 = ${r}(y - f32(y1));
          let result = dy2 * (dx2 * p11 + dx1 * p12) + dy1 * (dx2 * p21 + dx1 * p22);
        `;case"bicubic":return`
          let x0 = i32(floor(x)) - 1;
          let y0 = i32(floor(y)) - 1;
          var p: mat4x4<${r}>;
          for (var h = 0; h < 4; h++) {
            for (var w = 0; w < 4; w++) {
              p[h][w] = pixel_at_grid(h + y0, w + x0, H_in, W_in, indices[${ks}], indices[${Vs}], border);
            }
          }

          let dx = x - f32(x0 + 1);
          let dy = y - f32(y0 + 1);
          let result = gs_bicubic_interpolate(p, dx, dy);
        `;default:throw new Error(`mode ${t.mode} is not supported`)}})()+`${e.setByOffset("global_idx","result")}`,N_=(e,r)=>{let t=Ie("x",e[0].dataType,e[0].dims.length),s=[e[1].dims[0],e[1].dims[1],e[1].dims[2]],o=Ie("grid",e[1].dataType,s.length,2),n=[e[0].dims[0],e[0].dims[1],e[1].dims[1],e[1].dims[2]];r.format==="NHWC"&&(n=[e[0].dims[0],e[1].dims[1],e[1].dims[2],e[0].dims[3]],[ks,Vs,yn,vn]=[0,3,1,2]);let a=nt("output",e[0].dataType,n.length),i=t.type.value,l=ye.size(n),c=[{type:12,data:l},...dt(e[0].dims,s,n)],p=d=>`
  ${d.registerUniform("output_size","u32").declareVariables(t,o,a)}
  ${D_}
  ${L_(i)}
  ${z_(r)}
  ${B_(r)}
  ${R_(t,i,r)}

  ${d.mainStart()}
    ${d.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.output_size")}
      let H_in = i32(uniforms.x_shape[${yn}]);
      let W_in = i32(uniforms.x_shape[${vn}]);

      ${r.alignCorners===0?`
      let x_min = -0.5;
      let x_max = f32(W_in) - 0.5;
      let y_min = -0.5;
      let y_max = f32(H_in) - 0.5;
      `:`
      let x_min = 0.0;
      let x_max = f32(W_in) - 1.0;
      let y_min = 0.0;
      let y_max = f32(H_in) - 1.0;
      `};
      let border = vec4<f32>(x_min, y_min, x_max, y_max);

      let indices = ${a.offsetToIndices("global_idx")};
      var grid_indices = vec3<u32>(indices[${ks}], indices[${yn}], indices[${vn}]);
      let nxy = ${o.getByIndices("grid_indices")};
      var x = gs_denormalize(f32(nxy[0]), W_in);
      var y = gs_denormalize(f32(nxy[1]), H_in);

      ${j_(a,i,r)}
  }`;return{name:"GridSample",shaderCache:{hint:`${r.cacheKey}`,inputDependencies:["type","type"]},getRunData:d=>{let u=ye.size(n);return{outputs:[{dims:n,dataType:d[0].dataType}],dispatchGroup:{x:Math.ceil(u/64)},programUniforms:c}},getShaderSource:p}},V_=(e,r)=>{O_(e.inputs),e.compute(N_(e.inputs,r))},U_=e=>Kt({alignCorners:e.align_corners,mode:e.mode,paddingMode:e.padding_mode,format:e.format})}),Kr,W_,G_,Bl,K_,xo,H_,q_=Ue(()=>{Mt(),Pt(),pr(),nl(),_l(),Ct(),tn(),Kr=(e,r)=>e.length>r&&e[r].dims.length>0?e[r]:void 0,W_=(e,r)=>{let t=e[0],s=Kr(e,1),o=Kr(e,2),n=Kr(e,3),a=Kr(e,4),i=Kr(e,5),l=Kr(e,6),c=Kr(e,7);if(t.dims.length!==3&&t.dims.length!==5)throw new Error("Input query is expected to have 3 or 5 dimensions");let p=t.dims[0],d=t.dims[1],u=t.dims.length===3?t.dims[2]:r.numHeads*t.dims[4],f=d,_=0,b=0,k=Math.floor(u/r.numHeads);if(l&&c&&ye.size(l.dims)&&ye.size(c.dims)){if(l.dims.length!==4)throw new Error('Input "past_key" is expected to have 4 dimensions');if(l.dims[0]!==p||l.dims[1]!==r.numHeads||l.dims[3]!==k)throw new Error('Input "past_key" shape (batch_size, num_heads, past_sequence_length, head_size)');if(c.dims[0]!==p||c.dims[1]!==r.numHeads||c.dims[3]!==k)throw new Error('Input "past_value" shape (batch_size, num_heads, past_sequence_length, head_size)');if(l.dims[2]!==c.dims[2])throw new Error('Input "past_key" and "past_value" shall have same dim 2 (past_sequence_length)');if(c.dims.length!==4)throw new Error('Input "past_value" is expected to have 4 dimensions');_=l.dims[2],b=l.dims[2]}else if(l&&ye.size(l.dims)||c&&ye.size(c.dims))throw new Error('Input "past_key" and "past_value" shall be both present or both absent');let w;if(s&&ye.size(s.dims)>0){if(t.dims.length!==3)throw new Error('Input "query" is expected to have 3 dimensions when key is given');if(s.dims.length<3||s.dims.length>5)throw new Error('Input "key" is expected to have 3, 4, or 5 dimensions');if(t.dims[0]!==s.dims[0])throw new Error('Input "query" and "key" shall have same dim 0 (batch size)');if(s.dims.length===3){if(s.dims[2]!==t.dims[2])throw new Error('Input "query" and "key" shall have same dim 2 (hidden_size)');w=2,f=s.dims[1]}else if(s.dims.length===5){if(s.dims[2]!==r.numHeads||s.dims[3]!==2||s.dims[4]!==k)throw new Error('Expect "key" shape (batch_size, kv_sequence_length, num_heads, 2, head_size) for packed kv');if(o)throw new Error('Expect "value" be none when "key" has packed kv format.');w=5,f=s.dims[1]}else{if(s.dims[1]!==r.numHeads||s.dims[3]!==k)throw new Error('Expect "key" shape (batch_size, num_heads, kv_sequence_length, head_size) for past_key');w=0,f=s.dims[2]}}else{if(t.dims.length!==5)throw new Error('Input "query" is expected to have 5 dimensions when key is empty');if(t.dims[2]!==r.numHeads||t.dims[3]!==3)throw new Error('Expect "query" shape (batch_size, kv_sequence_length, num_heads, 3, head_size) for packed kv');w=3}if(n&&ye.size(n.dims)>0){if(n.dims.length!==1)throw new Error('Input "bias" is expected to have 1 dimension');if(s&&s.dims.length===5&&s.dims[3]===2)throw new Error("bias is not allowed for packed kv.")}let y=_+f,$=0;if(a&&ye.size(a.dims)>0){$=8;let v=a.dims;throw v.length===1?v[0]===p?$=1:v[0]===3*p+2&&($=3):v.length===2&&v[0]===p&&v[1]===y&&($=5),$===8?new Error('Input "key_padding_mask" shape shall be (batch_size) or (batch_size, total_sequence_length)'):new Error("Mask not supported")}let T=!1,M=u;if(o&&ye.size(o.dims)>0){if(o.dims.length!==3&&o.dims.length!==4)throw new Error('Input "value" is expected to have 3 or 4 dimensions');if(t.dims[0]!==o.dims[0])throw new Error('Input "query" and "value" shall have same dim 0 (batch_size)');if(o.dims.length===3){if(f!==o.dims[1])throw new Error('Input "key" and "value" shall have the same dim 1 (kv_sequence_length)');M=o.dims[2]}else{if(f!==o.dims[2])throw new Error('Input "key" and "value" shall have the same dim 2 (kv_sequence_length)');M=o.dims[1]*o.dims[3],T=!0}}let x=!1;if(a&&ye.size(a.dims)>0)throw new Error("Key padding mask is not supported");if(i&&ye.size(i.dims)>0){if(i.dims.length!==4)throw new Error('Input "attention_bias" is expected to have 4 dimensions');if(i.dims[0]!==p||i.dims[1]!==r.numHeads||i.dims[2]!==d||i.dims[3]!==y)throw new Error('Expect "attention_bias" shape (batch_size, num_heads, sequence_length, total_sequence_length)')}return{batchSize:p,sequenceLength:d,pastSequenceLength:_,kvSequenceLength:f,totalSequenceLength:y,maxSequenceLength:b,inputHiddenSize:0,hiddenSize:u,vHiddenSize:M,headSize:k,vHeadSize:Math.floor(M/r.numHeads),numHeads:r.numHeads,isUnidirectional:!1,pastPresentShareBuffer:!1,maskFilterValue:r.maskFilterValue,maskType:$,scale:r.scale,broadcastResPosBias:x,passPastInKv:T,qkvFormat:w}},G_=e=>Kt({...e}),Bl=Kt({perm:[0,2,1,3]}),K_=(e,r,t,s,o,n,a)=>{let i=[s,o,n],l=ye.size(i),c=[{type:12,data:l},{type:12,data:a},{type:12,data:n}],p=d=>{let u=nt("qkv_with_bias",r.dataType,i),f=Ie("qkv",r.dataType,i),_=Ie("bias",t.dataType,i),b=[{name:"output_size",type:"u32"},{name:"bias_offset",type:"u32"},{name:"hidden_size",type:"u32"}];return`
  ${d.registerUniforms(b).declareVariables(f,_,u)}
  ${d.mainStart()}
    ${d.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.output_size")}
    let bias_offset_idx = (global_idx % uniforms.hidden_size) + uniforms.bias_offset;

    qkv_with_bias[global_idx] = qkv[global_idx] + bias[bias_offset_idx];
  }`};return e.compute({name:"MultiHeadAttentionAddBias",shaderCache:{inputDependencies:["type","type"]},getRunData:()=>({outputs:[{dims:i,dataType:r.dataType,gpuDataType:0}],dispatchGroup:{x:Math.ceil(l/64)},programUniforms:c}),getShaderSource:p},{inputs:[r,t],outputs:[-1]})[0]},xo=(e,r,t,s,o,n,a,i)=>{let l=n;if(a&&ye.size(a.dims)>0){if(s===1)throw new Error("AddBiasReshape is not implemented. Please export your model with packed QKV or KV");return l=K_(e,n,a,r,s,t*o,i),l=l.reshape([r,s,t,o]),t===1||s===1?l:e.compute(rs(l,Bl.perm),{inputs:[l],outputs:[-1]})[0]}else return n.dims.length===3&&(l=n.reshape([r,s,t,o])),t===1||s===1?l:e.compute(rs(l,Bl.perm),{inputs:[l],outputs:[-1]})[0]},H_=(e,r)=>{let t=W_(e.inputs,r),s=e.inputs[0],o=Kr(e.inputs,1),n=Kr(e.inputs,2),a=Kr(e.inputs,3),i=Kr(e.inputs,4),l=Kr(e.inputs,5),c=Kr(e.inputs,6),p=Kr(e.inputs,7);if(s.dims.length===5)throw new Error("Packed QKV is not implemented");if((o==null?void 0:o.dims.length)===5)throw new Error("Packed KV is not implemented");let d=o&&n&&o.dims.length===4&&n.dims.length===4,u=xo(e,t.batchSize,t.numHeads,t.sequenceLength,t.headSize,s,a,0);if(d)return wo(e,u,o,n,i,void 0,c,p,l,t);if(!o||!n)throw new Error("key and value must be provided");let f=xo(e,t.batchSize,t.numHeads,t.kvSequenceLength,t.headSize,o,a,t.hiddenSize),_=xo(e,t.batchSize,t.numHeads,t.kvSequenceLength,t.vHeadSize,n,a,2*t.hiddenSize);wo(e,u,f,_,i,void 0,c,p,l,t)}}),Q_,X_,J_,Y_,Rl,Z_,ef,tf=Ue(()=>{Mt(),Pt(),pr(),Ct(),Q_=e=>{if(!e||e.length<1)throw new Error("too few inputs")},X_=(e,r)=>{let t=[],s=r.numOutputs;return e[1].dims[0]>0&&(e[1].getBigInt64Array().forEach(o=>t.push(Number(o))),s=t.length),Kt({numOutputs:s,axis:r.axis,splitSizes:t})},J_=e=>`
fn calculateOutputIndex(index: u32) -> u32 {
    for (var i: u32 = 0u; i < ${e}u; i += 1u ) {
    if (index < ${ct("uniforms.size_in_split_axis","i",e)}) {
        return i;
    }
    }
    return ${e}u;
}`,Y_=e=>{let r=e.length,t=[];for(let s=0;s<r;++s){let o=e[s].setByIndices("indices","input[global_idx]");r===1?t.push(o):s===0?t.push(`if (output_number == ${s}u) { ${o} }`):s===r-1?t.push(`else { ${o} }`):t.push(`else if (output_number == ${s}) { ${o} }`)}return`
      fn writeBufferData(output_number: u32, indices: ${e[0].type.indices}, global_idx: u32) {
        ${t.join(`
`)}
      }`},Rl=(e,r)=>{let t=e[0].dims,s=ye.size(t),o=e[0].dataType,n=ye.normalizeAxis(r.axis,t.length),a=new Array(r.numOutputs),i=Ie("input",o,t.length),l=new Array(r.numOutputs),c=[],p=[],d=0,u=[{type:12,data:s}];for(let _=0;_<r.numOutputs;_++){d+=r.splitSizes[_],l[_]=d;let b=t.slice();b[n]=r.splitSizes[_],p.push(b),a[_]=nt(`output${_}`,o,b.length),c.push({dims:p[_],dataType:e[0].dataType})}u.push({type:12,data:l},...dt(t,...p));let f=_=>`
  ${_.registerUniform("input_size","u32").registerUniform("size_in_split_axis","u32",l.length).declareVariables(i,...a)}
  ${J_(l.length)}
  ${Y_(a)}

  ${_.mainStart()}
    ${_.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.input_size")}

    var indices = ${i.offsetToIndices("global_idx")};
    var index = ${i.indicesGet("indices",n)};
    let output_number = calculateOutputIndex(index);
    if (output_number != 0) {
      index -= ${ct("uniforms.size_in_split_axis","output_number - 1u",l.length)};
      ${i.indicesSet("indices",n,"index")};
    }
    writeBufferData(output_number, indices, global_idx);
  }`;return{name:"Split",shaderCache:{hint:r.cacheKey,inputDependencies:["rank"]},getShaderSource:f,getRunData:()=>({outputs:c,dispatchGroup:{x:Math.ceil(s/64)},programUniforms:u})}},Z_=(e,r)=>{Q_(e.inputs);let t=e.inputs.length===1?r:X_(e.inputs,r);e.compute(Rl(e.inputs,t),{inputs:[0]})},ef=e=>{let r=e.axis,t=e.splitSizes,s=e.numOutputs<0?t.length:e.numOutputs;if(s!==t.length)throw new Error("numOutputs and splitSizes lengh must be equal");return Kt({axis:r,numOutputs:s,splitSizes:t})}}),rf,wa,sf,nf=Ue(()=>{Mt(),Pt(),pr(),Ct(),rf=(e,r)=>{let[t,s,o,n]=e,{numHeads:a,rotaryEmbeddingDim:i}=r;if(t.dims.length!==3&&t.dims.length!==4)throw new Error(`Input 'x' is expected to have 3 or 4 dimensions, got ${t.dims.length}`);if(!ye.areEqual(s.dims,[])&&!ye.areEqual(s.dims,[1])&&s.dims.length!==2)throw new Error(`Input 'position_ids' is expected to have 0, 1, or 2 dimensions, got ${s.dims.length}`);if(o.dims.length!==2)throw new Error(`Input 'cos_cache' is expected to have 2 dimensions, got ${o.dims.length}`);if(n.dims.length!==2)throw new Error(`Input 'sin_cache' is expected to have 2 dimensions, got ${n.dims.length}`);if(!ye.areEqual(o.dims,n.dims))throw new Error("Inputs 'cos_cache' and 'sin_cache' are expected to have the same shape");if(i>0&&a===0)throw new Error("num_heads must be provided if rotary_embedding_dim is specified");let l=t.dims[0],c=t.dims[t.dims.length-2],p=o.dims[0],d=ye.sizeFromDimension(t.dims,1)/c,u=i===0?o.dims[1]*2:d/a;if(i>u)throw new Error("rotary_embedding_dim must be less than or equal to head_size");if(s.dims.length===2){if(l!==s.dims[0])throw new Error(`Input 'position_ids' dimension 0 should be of size batch_size, got ${s.dims[0]}`);if(c!==s.dims[1])throw new Error(`Input 'position_ids' dimension 1 should be of size sequence_length, got ${s.dims[1]}`)}if(u/2!==o.dims[1]&&i/2!==o.dims[1])throw new Error(`Input 'cos_cache' dimension 1 should be same as head_size / 2 or rotary_embedding_dim / 2, got ${o.dims[1]}`);if(c>p)throw new Error("Updating cos_cache and sin_cache in RotaryEmbedding is not currently supported")},wa=(e,r)=>{let{interleaved:t,numHeads:s,rotaryEmbeddingDim:o,scale:n}=r,a=e[0].dims[0],i=ye.sizeFromDimension(e[0].dims,1),l=e[0].dims[e[0].dims.length-2],c=i/l,p=e[2].dims[1],d=o===0?p*2:c/s,u=new Array(a,l,c/d,d-p),f=ye.computeStrides(u),_=[{type:1,data:n},{type:12,data:u},{type:12,data:f},...e[0].dims.length===3?new Array({type:12,data:[i,c,d,1]}):[],...e[0].dims.length===4?new Array({type:12,data:[i,d,l*d,1]}):[],...dt(e[0].dims,e[1].dims,e[2].dims,e[3].dims,e[0].dims)],b=k=>{let w=Ie("input",e[0].dataType,e[0].dims.length),y=Ie("position_ids",e[1].dataType,e[1].dims.length),$=Ie("cos_cache",e[2].dataType,e[2].dims.length),T=Ie("sin_cache",e[3].dataType,e[3].dims.length),M=nt("output",e[0].dataType,e[0].dims.length);return k.registerUniforms([{name:"scale",type:"f32"},{name:"global_shape",type:"u32",length:u.length},{name:"global_strides",type:"u32",length:f.length},{name:"input_output_strides",type:"u32",length:f.length}]),`
        ${k.declareVariables(w,y,$,T,M)}

        ${k.mainStart(Kn)}
          let half_rotary_emb_dim = uniforms.${$.name}_shape[1];
          let bsnh = global_idx / uniforms.global_strides % uniforms.global_shape;
          let size = uniforms.global_shape[0] * uniforms.global_strides[0];
          ${k.guardAgainstOutOfBoundsWorkgroupSizes("size")}

          if (bsnh[3] < half_rotary_emb_dim) {
            let position_ids_idx =
                ${y.broadcastedIndicesToOffset("bsnh.xy",nt("",y.type.tensor,2))};
            let position_id =
                u32(${y.getByOffset("position_ids_idx")}) + select(0, bsnh[1], position_ids_idx == 0);
            let i = dot(bsnh, uniforms.input_output_strides) + select(0, bsnh[3], ${t});
            let j = i + select(half_rotary_emb_dim, 1, ${t});
            let re = ${w.getByOffset("i")} * ${$.get("position_id","bsnh[3]")} -
                ${w.getByOffset("j")} * ${T.get("position_id","bsnh[3]")};
            ${M.setByOffset("i","re")}
            let im = ${w.getByOffset("i")} * ${T.get("position_id","bsnh[3]")} +
                ${w.getByOffset("j")} * ${$.get("position_id","bsnh[3]")};
            ${M.setByOffset("j","im")}
          } else {
            let k = dot(bsnh, uniforms.input_output_strides) + half_rotary_emb_dim;
            ${M.setByOffset("k",w.getByOffset("k"))}
          }
        }`};return{name:"RotaryEmbedding",shaderCache:{hint:Kt({interleaved:t}).cacheKey,inputDependencies:["rank","rank","rank","rank"]},getShaderSource:b,getRunData:()=>({outputs:[{dims:e[0].dims,dataType:e[0].dataType}],dispatchGroup:{x:Math.ceil(ye.size(u)/Kn)},programUniforms:_})}},sf=(e,r)=>{rf(e.inputs,r),e.compute(wa(e.inputs,r))}}),of,af,jl,lf,cf,iT=Ue(()=>{pr(),Mt(),_l(),q_(),tf(),tn(),nf(),Ct(),of=(e,r)=>{if(r.doRotary&&e.length<=7)throw new Error("cos_cache and sin_cache inputs are required if do_rotary is specified");let t=e[0],s=e[1],o=e[2],n=e[3],a=e[4];if(r.doRotary!==0&&e.length<=7)throw new Error("cos_cast and sin_cache are expected if do_rotary attribute is non-zero");if(r.localWindowSize!==-1)throw new Error("Local attention is not supported");if(r.softcap!==0)throw new Error("Softcap is not supported");if(r.rotaryInterleaved!==0)throw new Error("Rotary interleaved is not supported");if(r.smoothSoftmax)throw new Error("Smooth softmax is not supported");if(t.dims.length!==3&&t.dims.length!==5)throw new Error("Input query is expected to have 3 or 5 dimensions");let i=!1,l=t.dims[0],c=t.dims[1],p=t.dims.length===3?i?t.dims[2]/3:t.dims[2]:r.numHeads*t.dims[4],d=c,u=0,f=!s||s.dims.length===0,_=Math.floor(f?p/(r.numHeads+2*r.kvNumHeads):p/r.numHeads);f&&(p=_*r.numHeads);let b=n&&n.dims.length!==0,k=a&&a.dims.length!==0;if(b&&n.dims.length===4&&n.dims[0]===l&&n.dims[1]!==r.kvNumHeads&&n.dims[2]===r.kvNumHeads&&n.dims[3]===_)throw new Error("BSNH pastKey/pastValue is not supported");if(b&&k){if(n.dims.length!==4)throw new Error('Input "past_key" is expected to have 4 dimensions');if(a.dims.length!==4)throw new Error('Input "past_value" is expected to have 4 dimensions');u=n.dims[2]}else if(b||k)throw new Error('Input "past_key" and "past_value" shall be both present or both absent');let w=1;if(s&&s.dims.length>0){if(t.dims.length!==3)throw new Error('Input "query" is expected to have 3 dimensions when key is given');if(s.dims.length<3||s.dims.length>5)throw new Error('Input "key" is expected to have 3, 4, or 5 dimensions');if(t.dims[0]!==s.dims[0])throw new Error('Input "query" and "key" shall have same dim 0 (batch size)');if(s.dims.length===3){if(t.dims[2]%s.dims[2]!==0)throw new Error('Dimension 2 of "query" should be a multiple of "key"');d=s.dims[1]}else if(s.dims.length===5){if(s.dims[2]!==r.numHeads||s.dims[3]!==2||s.dims[4]!==_)throw new Error('Expect "key" shape (batch_size, kv_sequence_length, num_heads, 2, head_size) for packed kv');if(o)throw new Error('Expect "value" be none when "key" has packed kv format.');d=s.dims[1]}else{if(s.dims[1]!==r.numHeads||s.dims[3]!==_)throw new Error('Expect "key" shape (batch_size, num_heads, kv_sequence_length, head_size) for past_key');d=s.dims[2]}}else{if(t.dims.length!==3&&t.dims.length!==5)throw new Error('Input "query" is expected to have 3 or 5 dimensions when key is empty');if(t.dims.length===5&&(t.dims[2]!==r.numHeads||t.dims[3]!==3))throw new Error('Expect "query" shape (batch_size, kv_sequence_length, num_heads, 3, head_size) for packed kv');w=3}let y=0,$=!1,T=r.kvNumHeads?_*r.kvNumHeads:p;if(o&&o.dims.length>0){if(o.dims.length!==3&&o.dims.length!==4)throw new Error('Input "value" is expected to have 3 or 4 dimensions');if(t.dims[0]!==o.dims[0])throw new Error('Input "query" and "value" shall have same dim 0 (batch_size)');if(o.dims.length===3){if(d!==o.dims[1])throw new Error('Input "key" and "value" shall have the same dim 1 (kv_sequence_length)');T=o.dims[2]}else{if(d!==o.dims[2])throw new Error('Input "past_key" and "past_value" shall have the same dim 2 (kv_sequence_length)');T=o.dims[1]*o.dims[3],$=!0}}let M=e.length>4?e[5]:void 0;if(M&&M.dims.length!==1&&M.dims[0]!==l)throw new Error('Input "seqlens" is expected to have 1 dimension and the same dim 0 as batch_size');return{batchSize:l,sequenceLength:c,pastSequenceLength:u,kvSequenceLength:d,totalSequenceLength:-1,maxSequenceLength:-1,inputHiddenSize:0,hiddenSize:p,vHiddenSize:T,headSize:_,vHeadSize:Math.floor(T/r.kvNumHeads),numHeads:r.numHeads,kvNumHeads:r.kvNumHeads,nReps:r.numHeads/r.kvNumHeads,pastPresentShareBuffer:!1,maskType:y,scale:r.scale,broadcastResPosBias:!1,passPastInKv:$,qkvFormat:w}},af=Kt({perm:[0,2,1,3]}),jl=(e,r,t)=>{let s=r,o=t.kvNumHeads;return r.dims.length===3&&t.kvSequenceLength!==0&&(s=r.reshape([t.batchSize,t.kvSequenceLength,o,t.headSize]),s=e.compute(rs(s,af.perm),{inputs:[s],outputs:[-1]})[0]),s},lf=(e,r,t,s)=>{let o=7,n=["type","type"],a=[e*r],i=e*r,l=[{type:12,data:i},{type:12,data:r},{type:12,data:e}],c=p=>{let d=Ie("seq_lens",t.dataType,t.dims),u=Ie("total_seq_lens",s.dataType,s.dims),f=nt("pos_ids",o,a),_=[{name:"output_size",type:"u32"},{name:"sequence_length",type:"u32"},{name:"batch_size",type:"u32"}];return`
  ${p.registerUniforms(_).declareVariables(d,u,f)}
  ${p.mainStart()}
    ${p.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.output_size")}
    let total_sequence_length = u32(${u.getByOffset("0")});
    let is_subsequent_prompt = uniforms.sequence_length > 1 && uniforms.sequence_length != total_sequence_length;
    let is_first_prompt = !is_subsequent_prompt && uniforms.sequence_length == total_sequence_length;
    let batch_idx = global_idx / uniforms.sequence_length;
    let sequence_idx = i32(global_idx % uniforms.sequence_length);
    var pos_id: i32 = 0;
    let seqlen = ${d.getByOffset("batch_idx")};
    let total_seqlen = seqlen + 1;
    if (is_first_prompt) {
      if (sequence_idx < total_seqlen) {
        pos_id = sequence_idx;
      } else {
        pos_id = 1;
      }
      ${f.setByOffset("global_idx","pos_id")}
    } else if (is_subsequent_prompt) {
      let past_seqlen = total_seqlen - i32(uniforms.sequence_length);
      if (past_seqlen + sequence_idx < total_seqlen) {
        pos_id = past_seqlen + sequence_idx;
      } else {
        pos_id = 1;
      }
      ${f.setByOffset("global_idx","pos_id")}
    } else if (global_idx < uniforms.batch_size) {
      ${f.setByOffset("global_idx","seqlen")}
    };
  }
  `};return{name:"GeneratePositionIds",shaderCache:{hint:`${e};${r}`,inputDependencies:n},getRunData:()=>({outputs:[{dims:a,dataType:o}],dispatchGroup:{x:Math.ceil(i/64)},programUniforms:l}),getShaderSource:c}},cf=(e,r)=>{var T;let t=of(e.inputs,r);if(e.inputs[0].dims.length===5)throw new Error("Packed QKV is not implemented");if(((T=e.inputs[1])==null?void 0:T.dims.length)===5)throw new Error("Packed KV is not implemented");let s=e.inputs[0],o=e.inputs[1]&&e.inputs[1].dims.length>0?e.inputs[1]:void 0,n=e.inputs[2]&&e.inputs[2].dims.length>0?e.inputs[2]:void 0,a=e.inputs[3]&&e.inputs[3].dims.length!==0?e.inputs[3]:void 0,i=e.inputs[4]&&e.inputs[4].dims.length!==0?e.inputs[4]:void 0,l=e.inputs.length>4?e.inputs[5]:void 0,c=e.inputs.length>5?e.inputs[6]:void 0,p=t.kvNumHeads?t.kvNumHeads:t.numHeads,d=Kt({axis:2,numOutputs:3,splitSizes:[t.numHeads*t.headSize,p*t.headSize,p*t.headSize]}),[u,f,_]=!o&&!n?e.compute(Rl([s],d),{inputs:[s],outputs:[-1,-1,-1]}):[s,o,n],b,k;if(r.doRotary){let M=e.compute(lf(t.batchSize,t.sequenceLength,l,c),{inputs:[l,c],outputs:[-1]})[0],x=e.inputs[7],v=e.inputs[8],C=Kt({interleaved:r.rotaryInterleaved!==0,numHeads:t.numHeads,rotaryEmbeddingDim:0,scale:r.scale}),F=[u,M,x,v],D=[-1];b=e.compute(wa(F,C),{inputs:F,outputs:D})[0],F.splice(0,1,f);let K=Kt({interleaved:r.rotaryInterleaved!==0,numHeads:t.kvNumHeads,rotaryEmbeddingDim:0,scale:r.scale});k=e.compute(wa(F,K),{inputs:F,outputs:D})[0]}let w=xo(e,t.batchSize,t.numHeads,t.sequenceLength,t.headSize,r.doRotary?b:u,void 0,0),y=jl(e,r.doRotary?k:f,t),$=jl(e,_,t);wo(e,w,y,$,void 0,void 0,a,i,void 0,t,l,c)}}),Nl,uf,df,pf,lT=Ue(()=>{Mt(),Pt(),tn(),Ct(),Nl=(e,r,t,s,o,n,a,i)=>{let l=lr(n),c=l===1?"f32":`vec${l}f`,p=l===1?"vec2f":`mat2x${l}f`,d=o*a,u=64;d===1&&(u=256);let f=[o,a,n/l],_=[o,a,2],b=["rank","type","type"],k=[];k.push(...dt(f,_));let w=y=>{let $=Ie("x",r.dataType,3,l),T=Ie("scale",t.dataType,t.dims),M=Ie("bias",s.dataType,s.dims),x=nt("output",1,3,2),v=[$,T,M,x];return`
  var<workgroup> workgroup_shared : array<${p}, ${u}>;
  const workgroup_size = ${u}u;
  ${y.declareVariables(...v)}
  ${y.mainStart(u)}
    let batch = workgroup_index / uniforms.x_shape[1];
    let channel = workgroup_index % uniforms.x_shape[1];
    let hight = uniforms.x_shape[2];
    // initialize workgroup memory
    var sum = ${c}(0);
    var squared_sum = ${c}(0);
    for (var h = local_idx; h < hight; h += workgroup_size) {
      let value = ${c}(${$.get("batch","channel","h")});
      sum += value;
      squared_sum += value * value;
    }
    workgroup_shared[local_idx] = ${p}(sum, squared_sum);
    workgroupBarrier();

    for (var currSize = workgroup_size >> 1;  currSize > 0; currSize = currSize >> 1) {
      if (local_idx < currSize) {
        workgroup_shared[local_idx] = workgroup_shared[local_idx] + workgroup_shared[local_idx + currSize];
      }
      workgroupBarrier();
    }
    if (local_idx == 0) {
      let sum_final = ${en("workgroup_shared[0][0]",l)} / f32(hight * ${l});
      let squared_sum_final = ${en("workgroup_shared[0][1]",l)} / f32(hight * ${l});

      let inv_std_dev = inverseSqrt(squared_sum_final - sum_final * sum_final + f32(${i}));
      let channel_scale = inv_std_dev * f32(scale[channel]);
      let channel_shift = f32(bias[channel]) - sum_final * channel_scale;
      output[workgroup_index] = vec2f(channel_scale, channel_shift);
    }
  }`};return e.compute({name:"InstanceNormComputeChannelScaleShift",shaderCache:{hint:`${l};${i};${u}`,inputDependencies:b},getRunData:()=>({outputs:[{dims:_,dataType:1}],dispatchGroup:{x:d},programUniforms:k}),getShaderSource:w},{inputs:[r,t,s],outputs:[-1]})[0]},uf=(e,r,t)=>{let s=r[0].dims,o=s,n=2,a=s[0],i=s[1],l=ye.sizeFromDimension(s,n),c=lr(l),p=ye.size(o)/c,d=Nl(e,r[0],r[1],r[2],a,l,i,t.epsilon),u=[a,i,l/c],f=[a,i],_=["type","none"],b=k=>{let w=Ie("x",r[0].dataType,u.length,c),y=Ie("scale_shift",1,f.length,2),$=nt("output",r[0].dataType,u.length,c),T=[w,y,$];return`
  ${k.registerUniform("output_size","u32").declareVariables(...T)}
  ${k.mainStart()}
  ${k.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.output_size")}
      let outputIndices = ${$.offsetToIndices("global_idx")};
      let batch = outputIndices[0];
      let channel = outputIndices[1];
      let scale_shift = ${y.getByIndices("vec2<u32>(batch, channel)")};
      let value = ${w.getByOffset("global_idx")} * ${$.type.value}(scale_shift.x) + ${$.type.value}(scale_shift.y);
      ${$.setByOffset("global_idx","value")};
  }`};e.compute({name:"InstanceNormalization",shaderCache:{hint:`${c}`,inputDependencies:_},getRunData:()=>({outputs:[{dims:o,dataType:r[0].dataType}],dispatchGroup:{x:Math.ceil(p/64)},programUniforms:[{type:12,data:p},...dt(u,f,u)]}),getShaderSource:b},{inputs:[r[0],d]})},df=(e,r,t)=>{let s=r[0].dims,o=s,n=s[0],a=s[s.length-1],i=ye.sizeFromDimension(s,1)/a,l=lr(a),c=ye.size(o)/l,p=[{type:12,data:i},{type:12,data:Math.floor(a/l)}],d=["type","type"],u=!1,f=[0,s.length-1];for(let w=0;w<s.length-2;w++)u=u||s[w+1]!==1,f.push(w+1);u=u&&s[s.length-1]!==1;let _=u?e.compute(rs(e.inputs[0],f),{inputs:[e.inputs[0]],outputs:[-1]})[0]:e.inputs[0].reshape(Array.from({length:s.length},(w,y)=>s[f[y]])),b=Nl(e,_,r[1],r[2],n,i,a,t.epsilon),k=w=>{let y=Ir(r[0].dataType),$=l===1?"vec2f":`mat${l}x2f`,T=v=>{let C=v===0?"x":"y",F=l===1?"f32":`vec${l}f`;switch(l){case 1:return`${y}(${F}(scale.${C}))`;case 2:return`vec2<${y}>(${F}(scale[0].${C}, scale[1].${C}))`;case 4:return`vec4<${y}>(${F}(scale[0].${C}, scale[1].${C}, scale[2].${C}, scale[3].${C}))`;default:throw new Error(`Not supported compoents ${l}`)}},M=Ie("input",r[0].dataType,r[0].dims,l),x=nt("output",r[0].dataType,o,l);return`
  @group(0) @binding(0) var<storage, read> input : array<${M.type.storage}>;
  @group(0) @binding(1) var<storage, read> scale_input : array<${$}>;
  @group(0) @binding(2) var<storage, read_write> output : array<${x.type.storage}>;
  struct Uniforms {H: u32, C : u32};
  @group(0) @binding(3) var<uniform> uniforms: Uniforms;

  ${w.mainStart()}
    let current_image_number = global_idx / (uniforms.C * uniforms.H);
    let current_channel_number = global_idx % uniforms.C;

    let scale_offset = current_image_number * uniforms.C + current_channel_number;
    let scale = scale_input[scale_offset];
    output[global_idx] = fma(input[global_idx], ${T(0)}, ${T(1)});
  }`};e.compute({name:"InstanceNormalizationNHWC",shaderCache:{hint:`${l}`,inputDependencies:d},getRunData:()=>({outputs:[{dims:o,dataType:r[0].dataType}],dispatchGroup:{x:Math.ceil(c/64)},programUniforms:p}),getShaderSource:k},{inputs:[r[0],b]})},pf=(e,r)=>{r.format==="NHWC"?df(e,e.inputs,r):uf(e,e.inputs,r)}}),mf,hf,_f,cT=Ue(()=>{Mt(),Pt(),Ct(),mf=e=>{if(!e||e.length<2)throw new Error("layerNorm requires at least 2 inputs.")},hf=(e,r,t)=>{let s=r.simplified,o=e[0].dims,n=e[1],a=!s&&e[2],i=o,l=ye.normalizeAxis(r.axis,o.length),c=ye.sizeToDimension(o,l),p=ye.sizeFromDimension(o,l),d=ye.size(n.dims),u=a?ye.size(a.dims):0;if(d!==p||a&&u!==p)throw new Error(`Size of X.shape()[axis:] == ${p}.
       Size of scale and bias (if provided) must match this.
       Got scale size of ${d} and bias size of ${u}`);let f=[];for(let M=0;M<o.length;++M)M<l?f.push(o[M]):f.push(1);let _=lr(p),b=["type","type"],k=[{type:12,data:c},{type:1,data:p},{type:12,data:Math.floor(p/_)},{type:1,data:r.epsilon}];a&&b.push("type");let w=t>1,y=t>2,$=M=>{let x=Ir(e[0].dataType),v=[Ie("x",e[0].dataType,e[0].dims,_),Ie("scale",n.dataType,n.dims,_)];a&&v.push(Ie("bias",a.dataType,a.dims,_)),v.push(nt("output",e[0].dataType,i,_)),w&&v.push(nt("mean_data_output",1,f)),y&&v.push(nt("inv_std_output",1,f));let C=[{name:"norm_count",type:"u32"},{name:"norm_size",type:"f32"},{name:"norm_size_vectorized",type:"u32"},{name:"epsilon",type:"f32"}];return`
  ${M.registerUniforms(C).declareVariables(...v)}
  ${M.mainStart()}
    ${M.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.norm_count")}
    let offset = global_idx * uniforms.norm_size_vectorized;
    var mean_vector = ${ll("f32",_)};
    var mean_square_vector = ${ll("f32",_)};

    for (var h: u32 = 0u; h < uniforms.norm_size_vectorized; h++) {
      let value = ${Hn(x,_,"x[h + offset]")};
      mean_vector += value;
      mean_square_vector += value * value;
    }
    let mean = ${en("mean_vector",_)} / uniforms.norm_size;
    let inv_std_dev = inverseSqrt(${en("mean_square_vector",_)} / uniforms.norm_size ${s?"":"- mean * mean"} + uniforms.epsilon);

    for (var j: u32 = 0; j < uniforms.norm_size_vectorized; j++) {
      let f32input = ${Hn(x,_,"x[j + offset]")};
      let f32scale = ${Hn(x,_,"scale[j]")};
      output[j + offset] = ${v[0].type.value}((f32input ${s?"":"- mean"}) * inv_std_dev * f32scale
        ${a?`+ ${Hn(x,_,"bias[j]")}`:""}
      );
    }

    ${w?"mean_data_output[global_idx] = mean":""};
    ${y?"inv_std_output[global_idx] = inv_std_dev":""};
  }`},T=[{dims:i,dataType:e[0].dataType}];return w&&T.push({dims:f,dataType:1}),y&&T.push({dims:f,dataType:1}),{name:"LayerNormalization",shaderCache:{hint:`${_};${t};${s}`,inputDependencies:b},getRunData:()=>({outputs:T,dispatchGroup:{x:Math.ceil(c/64)},programUniforms:k}),getShaderSource:$}},_f=(e,r)=>{mf(e.inputs),e.compute(hf(e.inputs,r,e.outputCount))}}),ff,gf,uT=Ue(()=>{Pt(),xl(),Cl(),ff=e=>{if(!e||e.length!==2)throw new Error("MatMul requires 2 inputs.");if(e[0].dims[e[0].dims.length-1]!==e[1].dims[e[1].dims.length-2])throw new Error("shared dimension does not match.")},gf=e=>{ff(e.inputs);let r=Gn.calcShape(e.inputs[0].dims,e.inputs[1].dims,!0);if(!r)throw new Error("Can't use matmul on the given tensors");let t=r[r.length-1],s=e.inputs[0].dims[e.inputs[0].dims.length-1];if(t<8&&s<8)e.compute(vl(e.inputs,{activation:""},r));else{let o=r[r.length-2],n=ye.size(e.inputs[0].dims.slice(0,-2)),a=ye.size(e.inputs[1].dims.slice(0,-2));if(n!==1&&o===1&&a===1){let i=e.inputs[0].reshape([1,n,s]),l=e.inputs[1].reshape([1,s,t]),c=[1,n,t],p=[i,l];e.compute(ha(p,{activation:""},r,c),{inputs:p})}else e.compute(ha(e.inputs,{activation:""},r))}}}),wf,Mf,bf,yf,vf,dT=Ue(()=>{Mt(),Pt(),pr(),Ct(),wf=(e,r)=>{if(e.length<3||e.length>4)throw new Error("MatMulNBits requires 3 or 4 inputs");let t=e[0],s=t.dims.length;if(t.dims[s-1]!==r.k)throw new Error("The last dim of input shape does not match the k value");let o=Math.floor((r.k+r.blockSize-1)/r.blockSize),n=r.blockSize/8*r.bits,a=e[1];if(!ye.areEqual(a.dims,[r.n,o,n]))throw new Error("The second inputs must be 3D tensor with shape N X nBlocksPerCol X blobSize");let i=e[2].dims;if(ye.size(i)!==r.n*o)throw new Error("scales input size error.");if(e.length===4){let l=e[3].dims,c=r.bits>4?r.n*o:r.n*Math.floor((o+1)/2);if(ye.size(l)!==c)throw new Error("zeroPoints input size error.")}},Mf=(e,r)=>{let t=e[0].dims,s=t.length,o=t[s-2],n=r.k,a=r.n,i=t.slice(0,s-2),l=ye.size(i),c=e[1].dims[2]/4,p=e[0].dataType,d=lr(r.k),u=lr(c),f=lr(a),_=i.concat([o,a]),b=o>1&&a/f%2===0?2:1,k=ye.size(_)/f/b,w=64,y=[],$=[l,o,n/d],T=ye.convertShape(e[1].dims).slice();T.splice(-1,1,c/u),y.push(...dt($)),y.push(...dt(T)),y.push(...dt(e[2].dims)),e.length===4&&y.push(...dt(ye.convertShape(e[3].dims)));let M=[l,o,a/f];y.push(...dt(M));let x=v=>{let C=$.length,F=Ie("a",e[0].dataType,C,d),D=Ie("b",12,T.length,u),K=Ie("scales",e[2].dataType,e[2].dims.length),U=[F,D,K],j=e.length===4?Ie("zero_points",12,e[3].dims.length):void 0;j&&U.push(j);let se=M.length,q=nt("output",e[0].dataType,se,f),re=Ir(e[0].dataType),Z=(()=>{switch(d){case 1:return`array<${re}, 8>`;case 2:return`mat4x2<${re}>`;case 4:return`mat2x4<${re}>`;default:throw new Error(`${d}-component is not supported.`)}})(),ae=()=>{let oe=`
          // reuse a data
            var input_offset = ${F.indicesToOffset(`${F.type.indices}(batch, row, word_offset)`)};
            var a_data: ${Z};
            for (var j: u32 = 0; j < ${8/d}; j++) {
              a_data[j] = ${F.getByOffset("input_offset")};
              input_offset++;
            }
          `;for(let G=0;G<f*b;G++)oe+=`
            b_value = ${u===1?`b${G}_data`:`b${G}_data[i]`};
            b_value_lower = unpack4xU8(b_value & b_mask);
            b_value_upper = unpack4xU8((b_value >> 4) & b_mask);
            b_quantized_values = ${Z}(${Array.from({length:4},(B,Q)=>`${re}(b_value_lower[${Q}]), ${re}(b_value_upper[${Q}])`).join(", ")});
            b_dequantized_values = ${d===1?`${Z}(${Array.from({length:8},(B,Q)=>`(b_quantized_values[${Q}] - ${j?`zero_point${G}`:"zero_point"}) * scale${G}`).join(", ")});`:`(b_quantized_values - ${Z}(${Array(8).fill(`${j?`zero_point${G}`:"zero_point"}`).join(",")})) * scale${G};`};
            workgroup_shared[local_id.x * ${b} + ${Math.floor(G/f)}]${f>1?`[${G%f}]`:""} += ${Array.from({length:8/d},(B,Q)=>`${d===1?`a_data[${Q}] * b_dequantized_values[${Q}]`:`dot(a_data[${Q}], b_dequantized_values[${Q}])`}`).join(" + ")};
          `;return oe},_e=()=>{let oe=`
            var col_index = col * ${f};
            ${j?`
            let zero_point_bytes_per_col = (nBlocksPerCol + 1) / 2;
            var zero_point_byte_count: u32;
            var zero_point_word_index: u32;
            var zero_point_byte_offset: u32;
            let zero_point_nibble_offset: u32 = block & 0x1u;
            var zero_point_bits_offset: u32;
            var zero_point_word: u32;`:`
            // The default zero point is 8 for unsigned 4-bit quantization.
            let zero_point = ${re}(8);`}
            `;for(let G=0;G<f*b;G++)oe+=`
            let scale${G} = ${K.getByOffset("col_index * nBlocksPerCol + block")};
            ${j?`
            zero_point_byte_count = col_index * zero_point_bytes_per_col + (block >> 0x1u);
            zero_point_word_index = zero_point_byte_count >> 0x2u;
            zero_point_byte_offset = zero_point_byte_count & 0x3u;
            zero_point_bits_offset = (zero_point_byte_offset << 3) + (zero_point_nibble_offset << 2);
            zero_point_word = ${j.getByOffset("zero_point_word_index")} >> zero_point_bits_offset;
            let zero_point${G} = ${re}((zero_point_word) & 0xFu);`:""}
            col_index += 1;`;return oe},pe=()=>{let oe=`col_index = col * ${f};`;for(let G=0;G<f*b;G++)oe+=`
            let b${G}_data = ${D.getByIndices(`${D.type.indices}(col_index, block, word)`)};
            col_index += 1;`;return oe+=`
            var b_value: u32;
            let b_mask: u32 = 0x0F0F0F0Fu;
            var b_value_lower: vec4<u32>;
            var b_value_upper: vec4<u32>;
            var b_quantized_values: ${Z};
            var b_dequantized_values: ${Z};`,oe};return`
        var<workgroup> workgroup_shared: array<${q.type.value}, ${b*w}>;
        ${v.declareVariables(...U,q)}
        ${v.mainStart([w,1,1])}
          let output_indices = ${q.offsetToIndices(`(global_idx / ${w}) * ${b}`)};
          let col = output_indices[2];
          let row = output_indices[1];
          let batch = output_indices[0];
          let nBlocksPerCol = uniforms.b_shape[1];

          for (var block = local_id.x; block < nBlocksPerCol; block += ${w}) {
            //process one block
            var word_offset: u32 = block * ${r.blockSize/d};
            ${_e()}
            for (var word: u32 = 0; word < ${c}; word += ${u}) {
              ${pe()}
              for (var i: u32 = 0; i < ${u}; i++) {
                ${ae()}
                word_offset += ${8/d};
              }
            }
          }
          workgroupBarrier();

          if (local_id.x < ${b}) {
            var output_value: ${q.type.value} = ${q.type.value}(0);
            var workgroup_shared_offset: u32 = local_id.x;
            for (var b: u32 = 0u; b < ${w}u; b++) {
              output_value += workgroup_shared[workgroup_shared_offset];
              workgroup_shared_offset += ${b};
            }
            ${q.setByIndices(`${q.type.indices}(batch, row, col + local_id.x)`,"output_value")};
          }
        }`};return{name:"MatMulNBits",shaderCache:{hint:`${r.blockSize};${r.bits};${d};${u};${f};${b};${w}`,inputDependencies:Array(e.length).fill("rank")},getRunData:()=>({outputs:[{dims:_,dataType:p}],dispatchGroup:{x:k},programUniforms:y}),getShaderSource:x}},bf=(e,r)=>{let t=e[0].dims,s=t.length,o=t[s-2],n=r.k,a=r.n,i=t.slice(0,s-2),l=ye.size(i),c=e[1].dims[2]/4,p=e[0].dataType,d=lr(r.k),u=lr(c),f=i.concat([o,a]),_=128,b=a%8===0?8:a%4===0?4:1,k=_/b,w=k*u*8,y=w/d,$=w/r.blockSize,T=ye.size(f)/b,M=[],x=[l,o,n/d],v=ye.convertShape(e[1].dims).slice();v.splice(-1,1,c/u),M.push(...dt(x)),M.push(...dt(v)),M.push(...dt(e[2].dims)),e.length===4&&M.push(...dt(ye.convertShape(e[3].dims)));let C=[l,o,a];M.push(...dt(C));let F=D=>{let K=x.length,U=Ie("a",e[0].dataType,K,d),j=Ie("b",12,v.length,u),se=Ie("scales",e[2].dataType,e[2].dims.length),q=[U,j,se],re=e.length===4?Ie("zero_points",12,e[3].dims.length):void 0;re&&q.push(re);let Z=C.length,ae=nt("output",e[0].dataType,Z),_e=Ir(e[0].dataType),pe=()=>{switch(d){case 1:return`
          let a_data0 = vec4<${_e}>(sub_a[word_offset], sub_a[word_offset + 1], sub_a[word_offset + 2], sub_a[word_offset + 3]);
          let a_data1 = vec4<${_e}>(sub_a[word_offset + 4], sub_a[word_offset + 5], sub_a[word_offset + 6], sub_a[word_offset + 7]);`;case 2:return`
          let a_data0 = vec4<${_e}>(sub_a[word_offset], sub_a[word_offset + 1]);
          let a_data1 = vec4<${_e}>(sub_a[word_offset + 2], sub_a[word_offset + 3]);`;case 4:return`
          let a_data0 = sub_a[word_offset];
          let a_data1 = sub_a[word_offset + 1];`;default:throw new Error(`${d}-component is not supported.`)}};return`
        var<workgroup> sub_a: array<${U.type.value}, ${y}>;
        var<workgroup> inter_results: array<array<${ae.type.value}, ${k}>, ${b}>;
        ${D.declareVariables(...q,ae)}
        ${D.mainStart([k,b,1])}
          let output_indices = ${ae.offsetToIndices(`workgroup_index * ${b}`)};
          let col = output_indices[2];
          let row = output_indices[1];
          let batch = output_indices[0];
          let n_blocks_per_col = uniforms.b_shape[1];
          let num_tiles =  (n_blocks_per_col - 1) / ${$} + 1;

          // Loop over shared dimension.
          for (var tile: u32 = 0; tile < num_tiles; tile += 1) {
            let a_col_start = tile * ${y};
            // load one tile A data into shared memory.
            for (var a_offset = local_idx; a_offset < ${y}; a_offset += ${_})
            {
              let a_col = a_col_start + a_offset;
              if (a_col < uniforms.a_shape[2])
              {
                sub_a[a_offset] = ${U.getByIndices(`${U.type.indices}(batch, row, a_col)`)};
              } else {
                sub_a[a_offset] = ${U.type.value}(0);
              }
            }
            workgroupBarrier();

            // each thread process one block
            let b_row = col + local_id.y;
            let block = tile * ${$} + local_id.x;
            ${re?`
            let zero_point_bytes_per_col = (n_blocks_per_col + 1) / 2;
            let zero_point_byte_count = b_row * zero_point_bytes_per_col + (block >> 0x1u);
            let zero_point_word_index = zero_point_byte_count >> 0x2u;
            let zero_point_byte_offset = zero_point_byte_count & 0x3u;
            let zero_point_nibble_offset: u32 = block & 0x1u;
            let zero_point_bits_offset = (zero_point_byte_offset << 3) + (zero_point_nibble_offset << 2);
            let zero_point_word = ${re.getByOffset("zero_point_word_index")} >> zero_point_bits_offset;
            let zero_point = ${_e}((zero_point_word) & 0xFu);`:`
            // The default zero point is 8 for unsigned 4-bit quantization.
            let zero_point = ${_e}(8);`}
            let scale = ${se.getByOffset("b_row * n_blocks_per_col + block")};
            let b_data = ${j.getByIndices(`${j.type.indices}(b_row, block, 0)`)};
            var word_offset = local_id.x * ${r.blockSize/d};
            for (var i: u32 = 0; i < ${u}; i++) {
              ${pe()}
              let b_value = ${u===1?"b_data":"b_data[i]"};
              let b_value_lower = unpack4xU8(b_value & 0x0F0F0F0Fu);
              let b_value_upper = unpack4xU8((b_value >> 4) & 0x0F0F0F0Fu);
              let b_quantized_values = mat2x4<${_e}>(${Array.from({length:4},(oe,G)=>`${_e}(b_value_lower[${G}]), ${_e}(b_value_upper[${G}])`).join(", ")});
              let b_dequantized_values = (b_quantized_values - mat2x4<${_e}>(${Array(8).fill("zero_point").join(",")})) * scale;
              inter_results[local_id.y][local_id.x] += ${Array.from({length:2},(oe,G)=>`${`dot(a_data${G}, b_dequantized_values[${G}])`}`).join(" + ")};
              word_offset += ${8/d};
            }
            workgroupBarrier();
          }

          if (local_idx < ${b}) {
            var output_value: ${ae.type.value} = ${ae.type.value}(0);
            for (var b = 0u; b < ${k}; b++) {
              output_value += inter_results[local_idx][b];
            }
            if (col + local_idx < uniforms.output_shape[2])
            {
              ${ae.setByIndices(`${ae.type.indices}(batch, row, col + local_idx)`,"output_value")}
            }
          }
        }`};return{name:"BlockwiseMatMulNBits32",shaderCache:{hint:`${r.blockSize};${d};${u};${k};${b}`,inputDependencies:Array(e.length).fill("rank")},getRunData:()=>({outputs:[{dims:f,dataType:p}],dispatchGroup:{x:T},programUniforms:M}),getShaderSource:F}},yf=(e,r)=>{wf(e.inputs,r),r.blockSize===32&&e.adapterInfo.isVendor("intel")&&e.adapterInfo.isArchitecture("gen-12lp")?e.compute(bf(e.inputs,r)):e.compute(Mf(e.inputs,r))},vf=e=>Kt(e)}),xf,Tf,Ef,Pf,Cf,Sf,$f,kf,If,pT=Ue(()=>{Mt(),Pt(),Ct(),xf=e=>{if(!e||e.length<1)throw new Error("Too few inputs");if(e[0].dataType!==1&&e[0].dataType!==10)throw new Error("Input type must be float or float16.");if(e.length>=2){let r=e[0].dims.length*2===e[1].dims[0];if(e.length===4&&(r=e[3].dims[0]*2===e[1].dims[0]),!r)throw new Error("The pads should be a 1D tensor of shape [2 * input_rank] or [2 * num_axes].")}},Tf=(e,r,t)=>{let s="";for(let o=r-1;o>=0;--o)s+=`
            k = i32(${e.indicesGet("indices",o)}) - ${ct("uniforms.pads",o,t)};
            if (k < 0) {
              break;
            }
            if (k >= i32(${ct("uniforms.x_shape",o,r)})) {
              break;
            }
            offset += k * i32(${ct("uniforms.x_strides",o,r)});
        `;return`
          value = ${e.type.value}(uniforms.constant_value);
          for (var i = 0; i < 1; i++) {
            var offset = 0;
            var k = 0;
            ${s}
            value = x[offset];
          }
      `},Ef=(e,r,t)=>{let s="";for(let o=r-1;o>=0;--o)s+=`
                k = i32(${e.indicesGet("indices",o)}) - ${ct("uniforms.pads",o,t)};
                if (k < 0) {
                  k = -k;
                }
                {
                  let _2n_1 = 2 * (i32(${ct("uniforms.x_shape",o,r)}) - 1);
                  k = k % _2n_1;
                  if(k >= i32(${ct("uniforms.x_shape",o,r)})) {
                    k = _2n_1 - k;
                  }
                }
                offset += k * i32(${ct("uniforms.x_strides",o,r)});
            `;return`
              var offset = 0;
              var k = 0;
              ${s}
              value = x[offset];
          `},Pf=(e,r,t)=>{let s="";for(let o=r-1;o>=0;--o)s+=`
                k = i32(${e.indicesGet("indices",o)}) - ${ct("uniforms.pads",o,t)};
                if (k < 0) {
                  k = 0;
                }
                if (k >= i32(${ct("uniforms.x_shape",o,r)})) {
                  k = i32(${ct("uniforms.x_shape",o,r)}) - 1;
                }
                offset += k * i32(${ct("uniforms.x_strides",o,r)});
            `;return`
              var offset = 0;
              var k = 0;
              ${s}
              value = x[offset];
          `},Cf=(e,r,t)=>{let s="";for(let o=r-1;o>=0;--o)s+=`
                k = i32(${e.indicesGet("indices",o)}) - ${ct("uniforms.pads",o,t)};
                if (k < 0)  {
                  k += i32(${ct("uniforms.x_shape",o,r)}]);
                }
                if (k >= i32(${ct("uniforms.x_shape",o,r)})) {
                  k -= i32(${ct("uniforms.x_shape",o,r)});
                }
                offset += k * i32(${ct("uniforms.x_strides",o,r)});
            `;return`
              var offset = 0;
              var k = 0;
              ${s}
              value = x[offset];
          `},Sf=(e,r,t)=>{switch(t.mode){case 0:return Tf(e,r,t.pads.length);case 1:return Ef(e,r,t.pads.length);case 2:return Pf(e,r,t.pads.length);case 3:return Cf(e,r,t.pads.length);default:throw new Error("Invalid mode")}},$f=(e,r)=>{let t=ye.padShape(e[0].dims.slice(),r.pads),s=e[0].dims,o=ye.size(t),n=[{type:12,data:o},{type:6,data:r.pads}],a=e.length>=3&&e[2].data;r.mode===0&&n.push({type:a?e[2].dataType:1,data:r.value}),n.push(...dt(e[0].dims,t));let i=["rank"],l=c=>{let p=nt("output",e[0].dataType,t.length),d=Ie("x",e[0].dataType,s.length),u=d.type.value,f=Sf(p,s.length,r),_=[{name:"output_size",type:"u32"},{name:"pads",type:"i32",length:r.pads.length}];return r.mode===0&&_.push({name:"constant_value",type:a?u:"f32"}),`
            ${c.registerUniforms(_).declareVariables(d,p)}
            ${c.mainStart()}
            ${c.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.output_size")}

            let indices = ${p.offsetToIndices("global_idx")};

            var value = ${u}(0);
            ${f}
            output[global_idx] = value;
        }`};return{name:"Pad",shaderCache:{hint:`${r.mode}${a}`,inputDependencies:i},getRunData:()=>({outputs:[{dims:t,dataType:e[0].dataType}],dispatchGroup:{x:Math.ceil(ye.size(t)/64)},programUniforms:n}),getShaderSource:l}},kf=(e,r)=>{if(e.length>1){let t=e[1].getBigInt64Array(),s=e.length>=3&&e[2].data?e[2].dataType===10?e[2].getUint16Array()[0]:e[2].getFloat32Array()[0]:0,o=e[0].dims.length,n=new Int32Array(2*o).fill(0);if(e.length>=4){let i=e[3].getBigInt64Array();for(let l=0;l<i.length;l++)n[Number(i[l])]=Number(t[l]),n[Number(i[l])+o]=Number(t[l+i.length])}else t.forEach((i,l)=>n[Number(l)]=Number(i));let a=[];return n.forEach(i=>a.push(i)),{mode:r.mode,value:s,pads:a}}else return r},If=(e,r)=>{xf(e.inputs);let t=kf(e.inputs,r);e.compute($f(e.inputs,t),{inputs:[0]})}}),To,Vl,Ul,Wl,Gl,Af,Ff,Kl,Hl,Of,Df,ql,Lf,zf,Ql,Bf,Rf,jf,Nf,mT=Ue(()=>{fs(),Mt(),Pt(),Ct(),To=e=>{if(tr.webgpu.validateInputContent&&(!e||e.length!==1))throw new Error("Pool ops requires 1 input.")},Vl=(e,r,t)=>{let s=r.format==="NHWC",o=e.dims.slice();s&&o.splice(1,0,o.pop());let n=Object.hasOwnProperty.call(r,"dilations"),a=r.kernelShape.slice(),i=r.strides.slice(),l=n?r.dilations.slice():[],c=r.pads.slice();aa.adjustPoolAttributes(t,o,a,i,l,c);let p=aa.computePoolOutputShape(t,o,i,l,a,c,r.autoPad),d=Object.assign({},r);n?Object.assign(d,{kernelShape:a,strides:i,pads:c,dilations:l,cacheKey:r.cacheKey}):Object.assign(d,{kernelShape:a,strides:i,pads:c,cacheKey:r.cacheKey});let u=p.slice();return u.push(u.splice(1,1)[0]),[d,s?u:p]},Ul=(e,r)=>{let t=r.format==="NHWC",s=ye.size(e),o=ye.size(r.kernelShape),n=[{type:12,data:s},{type:12,data:o}],a=[{name:"outputSize",type:"u32"},{name:"kernelSize",type:"u32"}];if(r.kernelShape.length<=2){let i=r.kernelShape[r.kernelShape.length-1],l=r.strides[r.strides.length-1],c=r.pads[r.pads.length/2-1],p=r.pads[r.pads.length-1],d=!!(c+p);n.push({type:12,data:i},{type:12,data:l},{type:12,data:c},{type:12,data:p}),a.push({name:"kw",type:"u32"},{name:"sw",type:"u32"},{name:"pwStart",type:"u32"},{name:"pwEnd",type:"u32"});let u=!1;if(r.kernelShape.length===2){let f=r.kernelShape[r.kernelShape.length-2],_=r.strides[r.strides.length-2],b=r.pads[r.pads.length/2-2],k=r.pads[r.pads.length-2];u=!!(b+k),n.push({type:12,data:f},{type:12,data:_},{type:12,data:b},{type:12,data:k}),a.push({name:"kh",type:"u32"},{name:"sh",type:"u32"},{name:"phStart",type:"u32"},{name:"phEnd",type:"u32"})}return[n,a,!0,d,u]}else{if(t)throw new Error("Pooling with kernelShape.length > 2 is not supported for NHWC format.");let i=ye.computeStrides(r.kernelShape);n.push({type:12,data:i},{type:12,data:r.pads},{type:12,data:r.strides}),a.push({name:"kernelStrides",type:"u32",length:i.length},{name:"pads",type:"u32",length:r.pads.length},{name:"strides",type:"u32",length:r.strides.length});let l=r.pads.reduce((c,p)=>c+p);return[n,a,!!l,!1,!1]}},Wl=(e,r,t,s,o,n,a,i,l,c,p,d)=>{let u=o.format==="NHWC",f=r.type.value,_=nt("output",r.type.tensor,s);if(o.kernelShape.length<=2){let b="",k="",w="",y=t-(u?2:1);if(p?b=`
                for (var i: u32 = 0u; i < uniforms.kw; i++) {
                  xIndices[${y}] = indices[${y}] * uniforms.sw - uniforms.pwStart + i;
                  if (xIndices[${y}] < 0 || xIndices[${y}]
                      >= uniforms.x_shape[${y}]) {
                    pad++;
                    continue;
                  }
                  let x_val = x[${r.indicesToOffset("xIndices")}];
                  ${n}
                }`:b=`
                for (var i: u32 = 0u; i < uniforms.kw; i++) {
                  xIndices[${y}] = indices[${y}] * uniforms.sw - uniforms.pwStart + i;
                  let x_val = x[${r.indicesToOffset("xIndices")}];
                  ${n}
                }`,o.kernelShape.length===2){let $=t-(u?3:2);d?k=`
                for (var j: u32 = 0u; j < uniforms.kh; j++) {
                  xIndices[${$}] = indices[${$}] * uniforms.sh - uniforms.phStart + j;
                  if (xIndices[${$}] < 0 || xIndices[${$}] >= uniforms.x_shape[${$}]) {
                    pad += i32(uniforms.kw);
                    continue;
                  }
              `:k=`
                for (var j: u32 = 0u; j < uniforms.kh; j++) {
                  xIndices[${$}] = indices[${$}] * uniforms.sh - uniforms.phStart + j;
                `,w=`
              }
            `}return`
            ${e.registerUniforms(l).declareVariables(r,_)}

            ${e.mainStart()}
              ${e.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.outputSize")}

              let indices = ${_.offsetToIndices("global_idx")};
              var xIndices = ${_.offsetToIndices("global_idx")};

              var value = ${f}(${i});
              var pad = 0;
              ${k}
              ${b}
              ${w}
              ${a}

              output[global_idx] = value;
            }`}else{if(u)throw new Error("Pooling with kernelShape.length > 2 is not supported for NHWC format.");let b=o.kernelShape.length,k=o.pads.length,w="";return c?w=`
                if (xIndices[j] >= uniforms.x_shape[j]) {
                  pad++;
                  isPad = true;
                  break;
                }
              }
              if (!isPad) {
                let x_val = x[${r.indicesToOffset("xIndices")}];
                ${n}
              }`:w=`
              }
              let x_val = x[${r.indicesToOffset("xIndices")}];
              ${n}
            `,`
            ${e.registerUniforms(l).declareVariables(r,_)}

            ${e.mainStart()}
              ${e.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.outputSize")}
              let indices = ${_.offsetToIndices("global_idx")};
              var xIndices = ${_.offsetToIndices("global_idx")};

              var offsets: array<u32, ${b}>;

              var value = ${f}(${i});
              var pad = 0;
              var isPad = false;

              for (var i: u32 = 0u; i < uniforms.kernelSize; i++) {
                var offset = i;
                for (var j = 0u; j < ${b-1}u; j++) {
                  offsets[j] = offset / ${ct("uniforms.kernelStrides","j",b)};
                  offset -= offsets[j] * ${ct("uniforms.kernelStrides","j",b)};
                }
                offsets[${b-1}] = offset;

                isPad = false;
                for (var j = ${t-b}u; j < ${t}u; j++) {
                  xIndices[j] = indices[j] * ${ct("uniforms.strides",`j - ${t-b}u`,b)}
                    + offsets[j - ${t-b}u] - ${ct("uniforms.pads","j - 2u",k)};
                  ${w}
              }
              ${a}

              output[global_idx] = value;
            }`}},Gl=e=>`${e.format};${e.ceilMode};${e.autoPad};${e.kernelShape.length}`,Af=e=>`${Gl(e)};${e.countIncludePad}`,Ff=e=>`${Gl(e)};${e.storageOrder};${e.dilations}`,Kl=e=>({format:e.format,autoPad:["NOTSET","VALID","SAME_UPPER","SAME_LOWER"][e.auto_pad],ceilMode:e.ceil_mode,kernelShape:e.kernel_shape,strides:e.strides,pads:e.pads}),Hl=(e,r,t,s)=>{let[o,n]=Vl(r,s,t),a=Ie("x",r.dataType,r.dims.length),i=a.type.value,l="value += x_val;",c="";o.countIncludePad?c+=`value /= ${i}(uniforms.kernelSize);`:c+=`value /= ${i}(i32(uniforms.kernelSize) - pad);`;let[p,d,u,f,_]=Ul(n,o);p.push(...dt(r.dims,n));let b=["rank"];return{name:e,shaderCache:{hint:`${s.cacheKey};${u};${f};${_}`,inputDependencies:b},getRunData:()=>({outputs:[{dims:n,dataType:r.dataType}],dispatchGroup:{x:Math.ceil(ye.size(n)/64)},programUniforms:p}),getShaderSource:k=>Wl(k,a,r.dims.length,n.length,o,l,c,0,d,u,f,_)}},Of=e=>{let r=e.count_include_pad!==0,t=Kl(e);if(t.ceilMode!==0)throw new Error("using ceil() in shape computation is not yet supported for AveragePool");let s={countIncludePad:r,...t,cacheKey:""};return{...s,cacheKey:Af(s)}},Df=(e,r)=>{To(e.inputs),e.compute(Hl("AveragePool",e.inputs[0],!1,r))},ql={autoPad:"",ceilMode:0,countIncludePad:!1,kernelShape:[],strides:[],pads:[],storageOrder:0,dilations:[]},Lf=e=>{let r=e.format;return{format:r,...ql,cacheKey:r}},zf=(e,r)=>{To(e.inputs),e.compute(Hl("GlobalAveragePool",e.inputs[0],!0,r))},Ql=(e,r,t,s)=>{let[o,n]=Vl(r,s,t),a=`
      value = max(x_val, value);
    `,i="",l=Ie("x",r.dataType,r.dims.length),c=["rank"],[p,d,u,f,_]=Ul(n,o);return p.push(...dt(r.dims,n)),{name:e,shaderCache:{hint:`${s.cacheKey};${u};${f};${_}`,inputDependencies:c},getRunData:()=>({outputs:[{dims:n,dataType:r.dataType}],dispatchGroup:{x:Math.ceil(ye.size(n)/64)},programUniforms:p}),getShaderSource:b=>Wl(b,l,r.dims.length,n.length,o,a,i,r.dataType===10?-65504:-1e5,d,u,f,_)}},Bf=(e,r)=>{To(e.inputs),e.compute(Ql("MaxPool",e.inputs[0],!1,r))},Rf=e=>{let r=e.storage_order,t=e.dilations,s=Kl(e);if(r!==0)throw new Error("column major storage order is not yet supported for MaxPool");if(s.ceilMode!==0)throw new Error("using ceil() in shape computation is not yet supported for MaxPool");let o={storageOrder:r,dilations:t,...s,cacheKey:""};return{...o,cacheKey:Ff(o)}},jf=e=>{let r=e.format;return{format:r,...ql,cacheKey:r}},Nf=(e,r)=>{To(e.inputs),e.compute(Ql("GlobalMaxPool",e.inputs[0],!0,r))}}),Vf,Uf,Wf,Gf,hT=Ue(()=>{Mt(),Pt(),pr(),Ct(),Vf=(e,r)=>{if(e.length<2||e.length>3)throw new Error("DequantizeLinear requires 2 or 3 inputs.");if(e.length===3&&e[1].dims===e[2].dims)throw new Error("x-scale and x-zero-point must have the same shape.");if(e.length===3&&e[0].dataType!==e[2].dataType)throw new Error("x and x-zero-point must have the same data type.");if(e[0].dataType===6&&e.length>2)throw new Error("In the case of dequantizing int32 there is no zero point.");if(e[1].dims.length!==0&&e[1].dims.length!==1&&e[1].dims.length!==e[0].dims.length)throw new Error("scale input must be a scalar, a 1D tensor, or have the same rank as the input tensor.");if(e.length>2){if(e[0].dataType!==e[2].dataType)throw new Error("x and x-zero-point must have the same data type.");if(e[1].dims.length!==e[2].dims.length)throw new Error("scale and zero-point inputs must have the same rank.");if(!e[1].dims.map((t,s)=>t===e[2].dims[s]).reduce((t,s)=>t&&s,!0))throw new Error("scale and zero-point inputs must have the same shape.")}if(r.blockSize>0){if(e[1].dims.length===0||e[1].dims.length===1&&e[1].dims[0]===1)throw new Error("blockSize must be set only for block quantization.");if(!e[1].dims.map((o,n)=>n===r.axis||o===e[0].dims[n]).reduce((o,n)=>o&&n,!0))throw new Error("For block qunatization, scale input shape to match the input shape except for the axis");if(e[1].dims.length!==e[0].dims.length)throw new Error("For block qunatization the scale input rank must be the same as the x rank.");let t=e[0].dims[r.axis],s=e[1].dims[r.axis];if(r.blockSize<Math.ceil(t/s)||r.blockSize>Math.ceil(t/(s-1)-1))throw new Error("blockSize must be with in the range [ceil(dI / Si), ceil(dI / (Si - 1) - 1)].")}},Uf=(e,r)=>{let t=ye.normalizeAxis(r.axis,e[0].dims.length),s=e[0].dataType,o=s===3,n=e[0].dims,a=e[1].dataType,i=ye.size(n),l=s===3||s===2,c=l?[Math.ceil(ye.size(e[0].dims)/4)]:e[0].dims,p=e[1].dims,d=e.length>2?e[2]:void 0,u=d?l?[Math.ceil(ye.size(d.dims)/4)]:d.dims:void 0,f=p.length===0||p.length===1&&p[0]===1,_=f===!1&&p.length===1,b=lr(i),k=f&&(!l||b===4),w=k?b:1,y=k&&!l?b:1,$=Ie("input",l?12:s,c.length,y),T=Ie("scale",a,p.length),M=d?Ie("zero_point",l?12:s,u.length):void 0,x=nt("output",a,n.length,w),v=[$,T];M&&v.push(M);let C=[c,p];d&&C.push(u);let F=[{type:12,data:i/w},{type:12,data:t},{type:12,data:r.blockSize},...dt(...C,n)],D=K=>{let U=[{name:"output_size",type:"u32"},{name:"axis",type:"u32"},{name:"block_size",type:"u32"}];return`
      ${K.registerUniforms(U).declareVariables(...v,x)}
      ${K.mainStart()}
          ${K.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.output_size")}
          let output_indices = ${x.offsetToIndices("global_idx")};

          // Set input x
          ${l?`
            let input = ${$.getByOffset("global_idx / 4")};
            let x_vec = ${o?"unpack4xI8(input)":"unpack4xU8(input)"};
            let x_value = ${w===1?"x_vec[global_idx % 4]":"x_vec"};`:`let x_value = ${$.getByOffset("global_idx")};`};

          // Set scale input
          ${f?`let scale_value= ${T.getByOffset("0")}`:_?`
            let scale_index = ${x.indicesGet("output_indices","uniforms.axis")};
            let scale_value= ${T.getByOffset("scale_index")};`:`
            var scale_indices: ${T.type.indices} = output_indices;
            let index = ${T.indicesGet("scale_indices","uniforms.axis")} / uniforms.block_size;
            ${T.indicesSet("scale_indices","uniforms.axis","index")};
            let scale_value= ${T.getByIndices("scale_indices")};`};

          // Set zero-point input
          ${M?f?l?`
                let zero_point_input = ${M.getByOffset("0")};
                let zero_point_vec =  ${o?"unpack4xI8(zero_point_input)":"unpack4xU8(zero_point_input)"};
                let zero_point_value= zero_point_vec[0]`:`let zero_point_value = ${M.getByOffset("0")}`:_?l?`
                let zero_point_index = ${x.indicesGet("output_indices","uniforms.axis")};
                let zero_point_input = ${M.getByOffset("zero_point_index / 4")};
                let zero_point_vec =  ${o?"unpack4xI8(zero_point_input)":"unpack4xU8(zero_point_input)"};
                let zero_point_value = zero_point_vec[zero_point_index % 4]`:`
                let zero_point_index = ${x.indicesGet("output_indices","uniforms.axis")};
                let zero_point_value = ${M.getByOffset("zero_point_index")};`:l?`
                let zero_point_offset = ${T.indicesToOffset("scale_indices")};
                let zero_point_input = ${M.getByOffset("zero_point_offset / 4")};
                let zero_point_vec = ${o?"unpack4xI8(zero_point_input)":"unpack4xU8(zero_point_input)"};
                let zero_point_value = zero_point_vec[zero_point_offset % 4];`:`let zero_point_value = ${M.getByIndices("scale_indices")};`:`let zero_point_value = ${l?o?"i32":"u32":$.type.value}(0);`};
      // Compute and write output
      ${x.setByOffset("global_idx",`${x.type.value}(x_value - zero_point_value) * scale_value`)};
      }`};return{name:"DequantizeLinear",shaderCache:{hint:r.cacheKey,inputDependencies:M?["rank","rank","rank"]:["rank","rank"]},getShaderSource:D,getRunData:()=>({outputs:[{dims:n,dataType:a}],dispatchGroup:{x:Math.ceil(i/w/64),y:1,z:1},programUniforms:F})}},Wf=(e,r)=>{Vf(e.inputs,r),e.compute(Uf(e.inputs,r))},Gf=e=>Kt({axis:e.axis,blockSize:e.blockSize})}),Kf,Hf,qf,_T=Ue(()=>{fs(),Mt(),Ct(),Kf=(e,r,t)=>{let s=e===r,o=e<r&&t<0,n=e>r&&t>0;if(s||o||n)throw new Error("Range these inputs' contents are invalid.")},Hf=(e,r,t,s)=>{let o=Math.abs(Math.ceil((r-e)/t)),n=[o],a=o,i=[{type:12,data:a},{type:s,data:e},{type:s,data:t},...dt(n)],l=c=>{let p=nt("output",s,n.length),d=p.type.value,u=[{name:"outputSize",type:"u32"},{name:"start",type:d},{name:"delta",type:d}];return`
        ${c.registerUniforms(u).declareVariables(p)}
        ${c.mainStart()}
        ${c.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.outputSize")}
        output[global_idx] = uniforms.start + ${d}(global_idx) * uniforms.delta;
      }`};return{name:"Range",shaderCache:{hint:`${s}`},getShaderSource:l,getRunData:()=>({outputs:[{dims:n,dataType:s}],dispatchGroup:{x:Math.ceil(a/64)},programUniforms:i})}},qf=e=>{let r=0,t=0,s=0;e.inputs[0].dataType===6?(r=e.inputs[0].getInt32Array()[0],t=e.inputs[1].getInt32Array()[0],s=e.inputs[2].getInt32Array()[0]):e.inputs[0].dataType===1&&(r=e.inputs[0].getFloat32Array()[0],t=e.inputs[1].getFloat32Array()[0],s=e.inputs[2].getFloat32Array()[0]),tr.webgpu.validateInputContent&&Kf(r,t,s),e.compute(Hf(r,t,s,e.inputs[0].dataType),{inputs:[]})}}),Qf,Xl,Jl,Xf,Jf,Yf,fT=Ue(()=>{Mt(),Pt(),pr(),Ct(),Qf=(e,r,t,s)=>{if(e!=="none"&&s!=="i32"&&s!=="u32"&&s!=="f32")throw new Error(`Input ${s} is not supported with reduction ${e}.`);let o=`{
                var oldValue = 0;
                loop {
                  let newValueF32 =`,n=`;
                  let newValue = bitcast<i32>(newValueF32);
                  let res = atomicCompareExchangeWeak(&${r}, oldValue, newValue);
                  if res.exchanged {
                    break;
                  }
                  oldValue = res.old_value;
                }
              }`;switch(e){case"none":return`${r}=${t};`;case"add":return s==="i32"||s==="u32"?`atomicAdd(&${r}, bitcast<${s}>(${t}));`:`
              ${o}bitcast<${s}>(oldValue) + (${t})${n}`;case"max":return s==="i32"||s==="u32"?`atomicMax(&${r}, bitcast<${s}>(${t}));`:`
                ${o}max(bitcast<f32>(oldValue), (${t}))${n}`;case"min":return s==="i32"||s==="u32"?`atomicMin(&${r}, bitcast<${s}>(${t}));`:`${o}min(bitcast<${s}>(oldValue), (${t}))${n}`;case"mul":return`${o}(bitcast<${s}>(oldValue) * (${t}))${n}`;default:throw new Error(`Reduction ${e} is not supported.`)}},Xl=(e,r)=>`${e===1?`
    let element_count_dim = uniforms.output_strides;
    let dim_value = uniforms.output_shape;`:`
    let element_count_dim = uniforms.output_strides[${r?"i - indices_start":"i"}];
    let dim_value = uniforms.output_shape[${r?"i - indices_start":"i"} + uniforms.last_index_dimension];`}
    
    if (index >= 0) {
      if (index >= i32(dim_value)) {
        index = i32(dim_value - 1);
      }
    } else {
      if (index < -i32(dim_value)) {
        index = 0;
      } else {
        index += i32(dim_value);
      }
    }
    data_offset += u32((u32(index) * element_count_dim));`,Jl=(e,r,t)=>`for (var i = 0u; i < uniforms.num_updates_elements; i++) {
        let value = updates[uniforms.num_updates_elements * ${t?"global_idx":"idx"} + i];
        ${Qf(e.reduction,"output[data_offset + i]","value",r)}
      }`,Xf=(e,r)=>{let t=e[0].dims,s=e[1].dims,o=t,n=1,a=Math.ceil(ye.size(s)/n),i=s[s.length-1],l=ye.sizeFromDimension(t,i),c=ye.sizeFromDimension(s,0)/i,p=[{type:12,data:a},{type:12,data:i},{type:12,data:l},...dt(e[1].dims,e[2].dims,o)],d=u=>{let f=Ie("indices",e[1].dataType,e[1].dims.length),_=Ie("updates",e[2].dataType,e[2].dims.length,n),b=r.reduction!=="none"&&r.reduction!==""?sp("output",e[0].dataType,o.length):nt("output",e[0].dataType,o.length,n);return`
      ${u.registerUniform("output_size","u32").registerUniform("last_index_dimension","u32").registerUniform("num_updates_elements","u32").declareVariables(f,_,b)}
      ${u.mainStart()}
        ${u.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.output_size")}
  var hasDuplicates = false;
  if (${r.reduction==="none"}) {
    for (var i = 0; i < ${c}; i = i + 1) {
      for (var j = i + 1; j < ${c}; j = j + 1) {
        var index_i = i32(indices[i].x);
        var index_j = i32(indices[j].x);
        if (index_i == index_j) {
          hasDuplicates = true;
          break;
        }
      }
      if (hasDuplicates) {
        break;
      }
    }
  }

  if (${r.reduction==="none"} && hasDuplicates) {
    if (global_idx != 0u) {
      return;
    }
    // Process each index-update pair individually when duplicates exist
    for (var idx = 0u; idx < ${c}u; idx++) {
      var data_offset = 0u;
      for (var i = 0u; i < uniforms.last_index_dimension; i++) {
        var index = i32(indices[idx * uniforms.last_index_dimension + i].x);
        ${Xl(t.length,!1)}
      }
      ${Jl(r,b.type.value,!1)}
    }
    return;
  }

  var data_offset = 0u;
  var indices_start = uniforms.last_index_dimension * global_idx;
  var indices_end = indices_start + uniforms.last_index_dimension;
  for (var i = indices_start; i < indices_end; i++) {
    var index = i32(indices[i].x);
    ${Xl(t.length,!0)}
  }
  ${Jl(r,b.type.value,!0)}
  }`};return{name:"ScatterND",shaderCache:{hint:`${r.cacheKey}_${r.reduction}`,inputDependencies:["rank","rank"]},getRunData:()=>({outputs:[{dims:o,dataType:e[0].dataType}],dispatchGroup:{x:Math.ceil(a/64)},programUniforms:p}),getShaderSource:d}},Jf=e=>Kt({reduction:e.reduction}),Yf=(e,r)=>{e.compute(Xf(e.inputs,r),{inputs:[e.inputs[1],e.inputs[2]],outputs:[]})}}),Zf,eg,tg,Yl,rg,sg,ng,og,ag,ig,lg,cg,Zl,ug,dg,pg,mg,hg,_g,fg,gT=Ue(()=>{Mt(),Pt(),pr(),Ct(),Zf=(e,r)=>{if(e.every(t=>t>0||(()=>{throw new Error("Resize requires scales input values to be positive")})),e.length>0){if(r.mode==="linear"){if(!(e.length===2||e.length===3||e.length===4&&e[0]===1&&e[1]===1||e.length===4&&e[0]===1&&e[3]===1||e.length===5&&e[0]===1&&e[1]===1))throw new Error(`For linear mode, Resize requires scales to be 2D, 3D, 4D with either two outermost or one innermost and
            one outermost scale values equal to 1, or 5D with two outermost scale values equal to 1`)}else if(r.mode==="cubic"&&!(e.length===2||e.length===4&&e[0]===1&&e[1]===1||e.length===4&&e[0]===1&&e[3]===1))throw new Error("Resize requires scales input size to be 2 or 4 for cubic mode")}},eg=(e,r,t)=>{r.every(o=>o>=0&&o<t||(()=>{throw new Error("Resize requires axes input values to be positive and less than rank")}));let s=new Array(t).fill(1);return r.forEach((o,n)=>s[o]=e[n]),s},tg=(e,r,t,s,o,n)=>{let[a,i,l]=t>10?[1,2,3]:[-1,e.length>1?1:-1,-1],c=e[0].dims.length;if(a>0&&e.length>a&&e[a].dims.length>0)e[a].getFloat32Array().forEach(p=>n.push(p));else if(r.coordinateTransformMode==="tf_crop_and_resize")throw new Error("Resize requires RoI input to be specified when coordinateTransformMode is tfCropAndResize");if(i>0&&e.length>i&&e[i].dims.length===1&&e[i].dims[0]>0){if(e[i].getFloat32Array().forEach(p=>s.push(p)),s.length!==0&&s.length!==c&&t>=18&&s.length!==r.axes.length)throw new Error("Resize requires scales input size to be same as input rank or axes size for opset 18 and up");Zf(s,r),r.axes.length>0&&eg(s,r.axes,c).forEach((p,d)=>s[d]=p)}if(l>0&&e.length>l&&e[l].dims.length===1&&e[l].dims[0]>0&&(e[l].getBigInt64Array().forEach(p=>o.push(Number(p))),o.length!==0&&o.length!==c&&t>=18&&o.length!==r.axes.length))throw new Error("Resize requires sizes input size to be same as input rank or axes size for opset 18 and up");if(r.axes.length>0){if(s.length!==0&&s.length!==r.axes.length)throw new Error('Resize requires "scales" input size to be of axes rank when axes attributes is specified');if(o.length!==0&&o.length!==r.axes.length)throw new Error('Resize requires "sizes" input size to be of rank axes rank when axes attributes is specified')}if(typeof s<"u"&&typeof o<"u"&&s.length>0&&o.length>c)throw new Error("Resize requires only of scales or sizes to be specified")},Yl=(e,r,t,s)=>`
  // The whole part and the fractional part are calculated separately due to inaccuracy of floating
  // point division. As an example, f32(21) / f32(7) may evaluate to 2.99... instead of 3, causing an
  // offset-by-one error later in floor().
  let big = (${e}) * (${r});
  let whole = ${s}(big / (${t}));
  let fract = ${s}(big % (${t})) / ${s}(${t});
  return whole + fract;
`,rg=(e,r)=>`fn getOriginalCoordinateFromResizedCoordinate(xResized: u32, xScale: f32, lengthResized: u32,
     lengthOriginal: u32, roiStart: f32, roiEnd: f32) -> ${r} { `+(()=>{switch(e){case"asymmetric":return`
          if (xScale < 1.0 || floor(xScale) != xScale) {
            return ${r}(xResized) / ${r}(xScale);
          } else {
            ${Yl("xResized","lengthOriginal","lengthResized",r)}
          }
        `;case"pytorch_half_pixel":return`if (lengthResized > 1) {
                    return (${r}(xResized) + 0.5) / ${r}(xScale) - 0.5;
                  } else {
                    return 0.0;
                  }`;case"tf_half_pixel_for_nn":return`return (${r}(xResized) + 0.5) / ${r}(xScale);`;case"align_corners":return`if (lengthResized == 1) {
                    return 0.0;
                  } else {
                    ${Yl("xResized","lengthOriginal - 1","lengthResized - 1",r)}
                  }`;case"tf_crop_and_resize":return`if (lengthResized > 1) {
                    return ${r}(roiStart) * ${r}(lengthOriginal - 1) +
                        (${r}(xResized) * ${r}(roiEnd - roiStart) * ${r}(lengthOriginal - 1)) /
                        ${r}(lengthResized - 1);
                  } else {
                    return 0.5 * ${r}(roiStart + roiEnd) * ${r}(lengthOriginal - 1);
                  }`;case"half_pixel_symmetric":return`const outputWidth = ${r}xScale * ${r}(lengthResized);
                  const adjustment = ${r}(lengthResized) / outputWidth;
                  const center = ${r}(lengthOriginal) / 2;
                  const offset = center * (1 - adjustment);
                  return offset + ((${r}(xResized) + 0.5) / ${r}(xScale)) - 0.5;`;case"half_pixel":return`return ((${r}(xResized) + 0.5) / ${r}(xScale)) - 0.5;`;default:throw new Error(`Coordinate transform mode ${e} is not supported`)}})()+"}",sg=(e,r,t)=>`fn getNearestPixelFromOriginal(xOriginal: ${t}, isDownSample: bool) -> ${t} {`+(()=>{switch(e){case"round_prefer_ceil":return"if (fract(xOriginal) == 0.5) {             return ceil(xOriginal);           } else {             return round(xOriginal);           }";case"floor":return"return floor(xOriginal);";case"ceil":return"return ceil(xOriginal);";case"round_prefer_floor":return"if (fract(xOriginal) == 0.5) {                     return floor(xOriginal);                   } else {                     return round(xOriginal);                   }";case"simple":default:if(r<11)return"if (isDownSample)                     {                       return ceil(xOriginal);                     } else {                       return xOriginal;                     }";throw new Error(`Nearest mode ${e} is not supported`)}})()+"}",ng=(e,r,t)=>{let s=new Array(t).fill(0).concat(new Array(t).fill(1)),o=e.length===0?s:e.slice();return r.length>0?(r.forEach((n,a)=>{s[n]=o[a],s[a+t]=o[r.length+a]}),s):o},og=(e,r,t,s)=>{let o=[];if(t.length>0)if(s.length>0){if(e.forEach(n=>o.push(n)),Math.max(...s)>e.length)throw new Error("axes is out of bound");s.forEach((n,a)=>o[n]=t[a])}else t.forEach(n=>o.push(n));else{if(r.length===0)throw new Error("Resize requires either scales or sizes.");o=e.map((n,a)=>Math.round(n*r[a]))}return o},ag=(e,r,t)=>{let s=(()=>{switch(t.keepAspectRatioPolicy){case"not_larger":return t.axes.length>0?Math.min(...t.axes.map(n=>r[n]),Number.MAX_VALUE):Math.min(...r,Number.MAX_VALUE);case"not_smaller":return t.axes.length>0?Math.max(...t.axes.map(n=>r[n]),Number.MIN_VALUE):Math.max(...r,Number.MIN_VALUE);default:throw new Error(`Keep aspect ratio policy ${t.keepAspectRatioPolicy} is not supported`)}})();r.fill(1,0,r.length);let o=e.slice();return t.axes.length>0?(t.axes.forEach(n=>r[n]=s),t.axes.forEach(n=>o[n]=Math.round(e[n]*r[n]))):(r.fill(s,0,r.length),o.forEach((n,a)=>o[a]=Math.round(n*r[a]))),o},ig=(e,r,t,s,o)=>`
    fn calculateOriginalIndicesFromOutputIndices(output_indices: ${e.type.indices}) -> array<${e.type.value}, ${t.length}> {
      var original_indices: array<${e.type.value}, ${t.length}>;
      for (var i:u32 = 0; i < ${t.length}; i++) {
        var output_index = ${e.indicesGet("output_indices","i")};
        var scale = ${ct("uniforms.scales","i",s)};
        var roi_low = ${ct("uniforms.roi","i",o)};
        var roi_hi = ${ct("uniforms.roi",`i + ${r.length}`,o)};
        if (scale == 1.0) {
          original_indices[i] = ${e.type.value}(output_index);
        } else {
          var input_shape_i = ${ct("uniforms.input_shape","i",r.length)};
          var output_shape_i = ${ct("uniforms.output_shape","i",t.length)};
          original_indices[i] = getOriginalCoordinateFromResizedCoordinate(output_index, scale, output_shape_i,
                                                                           input_shape_i, roi_low, roi_hi);
        }
      }
      return original_indices;
    }`,lg=(e,r,t,s,o,n,a)=>`
    fn calculateInputIndicesFromOutputIndices(output_indices: ${r.type.indices}) -> ${e.type.indices} {
      var input_indices: ${e.type.indices};
      for (var i:u32 = 0; i < ${s.length}; i++) {
        var output_index = ${r.indicesGet("output_indices","i")};
        var input_index: u32;
        var scale = ${ct("uniforms.scales","i",o)};
        if (scale == 1.0) {
          input_index = output_index;
        } else {
          var roi_low = ${ct("uniforms.roi","i",n)};
          var roi_hi = ${ct("uniforms.roi",`i + ${t.length}`,n)};
          var input_shape_i = ${ct("uniforms.input_shape","i",t.length)};
          var output_shape_i = ${ct("uniforms.output_shape","i",s.length)};
          var original_idx = getOriginalCoordinateFromResizedCoordinate(output_index, scale, output_shape_i,
                                                                        input_shape_i, roi_low, roi_hi);
          if (!${a} || (original_idx >= 0 && original_idx < ${r.type.value}(input_shape_i))) {
            if (original_idx < 0) {
              input_index = 0;
            } else if (original_idx > ${r.type.value}(input_shape_i - 1)) {
              input_index = input_shape_i - 1;
            } else {
              input_index = u32(getNearestPixelFromOriginal(original_idx, scale < 1));
            }
          } else {
            input_index = u32(original_idx);
          }
        }
        ${e.indicesSet("input_indices","i","input_index")}
      }
      return input_indices;
    }`,cg=(e,r)=>`
    fn checkInputIndices(input_indices: ${e.type.indices}) -> bool {
      for (var i:u32 = 0; i < ${r.length}; i++) {
        var input_index = ${e.indicesGet("input_indices","i")};
        if (input_index < 0 || input_index >= ${ct("uniforms.input_shape","i",r.length)}) {
          return false;
        }
      }
      return true;
    }`,Zl=(e,r,t,s)=>e.rank>s?`
    ${e.indicesSet("input_indices",r,"channel")};
    ${e.indicesSet("input_indices",t,"batch")};
`:"",ug=(e,r,t,s,o)=>{let[n,a,i,l]=t.length===2?[-1,0,1,-1]:[0,2,3,1],c=e.type.value;return`
    fn getInputValue(batch: u32, channel: u32, row: u32, col: u32) -> ${c} {
      var input_indices: ${e.type.indices};
      ${e.indicesSet("input_indices",a,`max(0, min(row, ${t[a]} - 1))`)};
      ${e.indicesSet("input_indices",i,`max(0, min(col, ${t[i]} - 1))`)};
      ${Zl(e,l,n,2)}
      return ${e.getByIndices("input_indices")};
    }

    fn bilinearInterpolation(output_indices: ${r.type.indices}) -> ${c} {
      var originalIndices = calculateOriginalIndicesFromOutputIndices(output_indices);
      var row:${c} = originalIndices[${a}];
      var col:${c} = originalIndices[${i}];
      ${s?`if (row < 0 || row > (${t[a]} - 1) || col < 0 || col > (${t[i]} - 1)) {
        return ${o};
      }`:""};
      row = max(0, min(row, ${t[a]} - 1));
      col = max(0, min(col, ${t[i]} - 1));
      var row1: u32 = u32(row);
      var col1: u32 = u32(col);
      var row2: u32 = u32(row + 1);
      var col2: u32 = u32(col + 1);
      var channel: u32 = ${t.length>2?`u32(originalIndices[${l}])`:"0"};
      var batch: u32 =  ${t.length>2?`u32(originalIndices[${n}])`:"0"};
      var x11: ${c} = getInputValue(batch, channel, row1, col1);
      var x12: ${c} = getInputValue(batch, channel, row1, col2);
      var x21: ${c} = getInputValue(batch, channel, row2, col1);
      var x22: ${c} = getInputValue(batch, channel, row2, col2);
      var dx1: ${c} = abs(row - ${c}(row1));
      var dx2: ${c} = abs(${c}(row2) - row);
      var dy1: ${c} = abs(col - ${c}(col1));
      var dy2: ${c} = abs(${c}(col2) - col);
      if (row1 == row2) {
        dx1 = 0.5;
        dx2 = 0.5;
      }
      if (col1 == col2) {
        dy1 = 0.5;
        dy2 = 0.5;
      }
      return (x11 * dx2 * dy2 + x12 * dx2 * dy1 + x21 * dx1 * dy2 + x22 * dx1 * dy1);
    }`},dg=(e,r,t,s,o,n,a,i,l,c)=>{let p=t.length===2,[d,u]=p?[0,1]:[2,3],f=e.type.value,_=b=>{let k=b===d?"row":"col";return`
      fn ${k}CubicInterpolation(input_indices: ${e.type.indices}, output_indices: ${r.type.indices}) -> ${f} {
        var output_index = ${r.indicesGet("output_indices",b)};
        var originalIdx: ${f} = getOriginalCoordinateFromResizedCoordinate(output_index, ${o[b]},
        ${s[b]}, ${t[b]}, ${n[b]}, ${n[b]} + ${t.length});
        var fractOriginalIdx: ${f} = originalIdx - floor(originalIdx);
        var coefs = getCubicInterpolationCoefs(fractOriginalIdx);

        if (${i} && (originalIdx < 0 || originalIdx > (${t[b]} - 1))) {
          return ${l};
        }
        var data: array<${f}, 4> = array<${f}, 4>(0.0, 0.0, 0.0, 0.0);
        for (var i: i32 = -1; i < 3; i++) {
          var ${k}: ${f} = originalIdx + ${f}(i);
          if (${k} < 0 || ${k} >= ${t[b]}) {
            ${c?`coefs[i + 1] = 0.0;
                        continue;`:i?`return ${l};`:`${k} = max(0, min(${k}, ${t[b]} - 1));`};
          }
        var input_indices_copy: ${e.type.indices} = input_indices;
          ${e.indicesSet("input_indices_copy",b,`u32(${k})`)};
          data[i + 1] = ${b===d?e.getByIndices("input_indices_copy"):"rowCubicInterpolation(input_indices_copy, output_indices)"};
        }
        return cubicInterpolation1D(data, coefs);
      }`};return`
    ${_(d)};
    ${_(u)};
  fn getCubicInterpolationCoefs(s: ${f}) -> array<${f}, 4> {
    var absS = abs(s);
    var coeffs: array<${f}, 4> = array<${f}, 4>(0.0, 0.0, 0.0, 0.0);
    var oneMinusAbsS: ${f} = 1.0 - absS;
    var twoMinusAbsS: ${f} = 2.0 - absS;
    var onePlusAbsS: ${f} = 1.0 + absS;
    coeffs[0] = ((${a} * onePlusAbsS - 5 * ${a}) * onePlusAbsS + 8 * ${a}) * onePlusAbsS - 4 * ${a};
    coeffs[1] = ((${a} + 2) * absS - (${a} + 3)) * absS * absS + 1;
    coeffs[2] = ((${a} + 2) * oneMinusAbsS - (${a} + 3)) * oneMinusAbsS * oneMinusAbsS + 1;
    coeffs[3] = ((${a} * twoMinusAbsS - 5 * ${a}) * twoMinusAbsS + 8 * ${a}) * twoMinusAbsS - 4 * ${a};
    return coeffs;
  }

  fn cubicInterpolation1D(x: array<${f}, 4>, coefs: array<${f}, 4>) -> ${f} {
    var coefsSum: ${f} = coefs[0] + coefs[1] + coefs[2] + coefs[3];
    return (x[0] * coefs[0] + x[1] * coefs[1]+ x[2] * coefs[2]+ x[3] * coefs[3]) / coefsSum;
  }

  fn bicubicInterpolation(output_indices: ${r.type.indices}) -> ${f} {
    var input_indices: ${e.type.indices} = output_indices;
    return colCubicInterpolation(input_indices, output_indices);
  }
    `},pg=(e,r,t,s,o)=>{let[n,a,i,l,c]=t.length===3?[-1,0,1,2,-1]:[0,2,3,4,1],p=e.type.value;return`
    fn getInputValue(batch: u32, channel: u32, depth:u32, height: u32, width: u32) -> ${p} {
      var input_indices: ${e.type.indices};
      ${e.indicesSet("input_indices",a,`max(0, min(depth, ${t[a]} - 1))`)};
      ${e.indicesSet("input_indices",i,`max(0, min(height, ${t[i]} - 1))`)};
      ${e.indicesSet("input_indices",l,`max(0, min(width, ${t[l]} - 1))`)};
      ${Zl(e,c,n,3)}
      return ${e.getByIndices("input_indices")};
    }

    fn trilinearInterpolation(output_indices: ${r.type.indices}) -> ${p} {
      var originalIndices = calculateOriginalIndicesFromOutputIndices(output_indices);
      var depth:${p} = originalIndices[${a}];
      var height:${p} = originalIndices[${i}];
      var width:${p} = originalIndices[${l}];
      ${s?`if (depth < 0 || depth > (${t[a]} - 1) || height < 0 || height > (${t[i]} - 1) || width < 0 || (width > ${t[l]} - 1)) {
      return ${o};
        }`:""};

    depth = max(0, min(depth, ${t[a]} - 1));
      height = max(0, min(height, ${t[i]} - 1));
      width = max(0, min(width, ${t[l]} - 1));
      var depth1: u32 = u32(depth);
      var height1: u32 = u32(height);
      var width1: u32 = u32(width);
      var depth2: u32 = u32(depth + 1);
      var height2: u32 = u32(height + 1);
      var width2: u32 = u32(width + 1);
      var channel: u32 = ${t.length>3?`u32(originalIndices[${c}])`:"0"};
      var batch: u32 =  ${t.length>3?`u32(originalIndices[${n}])`:"0"};

      var x111: ${p} = getInputValue(batch, channel, depth1, height1, width1);
      var x112: ${p} = getInputValue(batch, channel, depth1, height1, width2);
      var x121: ${p} = getInputValue(batch, channel, depth1, height2, width1);
      var x122: ${p} = getInputValue(batch, channel, depth1, height2, width2);
      var x211: ${p} = getInputValue(batch, channel, depth2, height1, width1);
      var x212: ${p} = getInputValue(batch, channel, depth2, height1, width2);
      var x221: ${p} = getInputValue(batch, channel, depth2, height2, width1);
      var x222: ${p} = getInputValue(batch, channel, depth2, height2, width2);
      var dx1: ${p} = abs(depth - ${p}(depth1));
      var dx2: ${p} = abs(${p}(depth2) - depth);
      var dy1: ${p} = abs(height - ${p}(height1));
      var dy2: ${p} = abs(${p}(height2) - height);
      var dz1: ${p} = abs(width - ${p}(width1));
      var dz2: ${p} = abs(${p}(width2) - width);
      if (depth1 == depth2) {
        dx1 = 0.5;
        dx2 = 0.5;
      }
      if (height1 == height2) {
        dy1 = 0.5;
        dy2 = 0.5;
      }
      if (width1 == width2) {
        dz1 = 0.5;
        dz2 = 0.5;
      }
      return (x111 * dx2 * dy2 * dz2 + x112 * dx2 * dy2 * dz1 + x121 * dx2 * dy1 *dz2 + x122 * dx2 * dy1 * dz1 +
              x211 * dx1 * dy2 * dz2 + x212 * dx1 * dy2 * dz1 + x221 * dx1 * dy1 *dz2 + x222 * dx1 * dy1 * dz1);
    }`},mg=(e,r,t,s,o,n)=>{let a=e.dims,i=ng(n,r.axes,a.length),l=og(a,s,o,r.axes),c=s.slice();s.length===0&&(c=a.map((y,$)=>y===0?1:l[$]/y),r.keepAspectRatioPolicy!=="stretch"&&(l=ag(a,c,r)));let p=nt("output",e.dataType,l.length),d=Ie("input",e.dataType,a.length),u=ye.size(l),f=a.length===l.length&&a.every((y,$)=>y===l[$]),_=r.coordinateTransformMode==="tf_crop_and_resize",b=r.extrapolationValue,k=d.type.value,w=y=>`
      ${f?"":`
      ${rg(r.coordinateTransformMode,k)};
      ${(()=>{switch(r.mode){case"nearest":return`
              ${cg(d,a)};
              ${sg(r.nearestMode,t,k)};
              ${lg(d,p,a,l,c.length,i.length,_)};
              `;case"linear":return`
              ${ig(p,a,l,c.length,i.length)};
              ${(()=>{if(a.length===2||a.length===4)return`${ug(d,p,a,_,b)}`;if(a.length===3||a.length===5)return`${pg(d,p,a,_,b)}`;throw Error("Linear mode only supports input dims 2, 3, 4 and 5 are supported in linear mode.")})()};
            `;case"cubic":return`
            ${(()=>{if(a.length===2||a.length===4)return`${dg(d,p,a,l,c,i,r.cubicCoeffA,_,r.extrapolationValue,r.excludeOutside)}`;throw Error("Cubic mode only supports input dims 2 and 4 are supported in linear mode.")})()};
            `;default:throw Error("Invalid resize mode")}})()};
      `}
      ${y.registerUniform("output_size","u32").registerUniform("scales","f32",c.length).registerUniform("roi","f32",i.length).declareVariables(d,p)}
      ${y.mainStart()}
        ${y.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.output_size")}
        ${f?"output[global_idx] = input[global_idx];":`
        let output_indices = ${p.offsetToIndices("global_idx")};
        var input_indices: ${d.type.indices};
        ${(()=>{switch(r.mode){case"nearest":return`input_indices = calculateInputIndicesFromOutputIndices(output_indices);
                if (checkInputIndices(input_indices)) {
                  output[global_idx] = ${d.getByIndices("input_indices")};
                } else {
                  output[global_idx] = ${r.extrapolationValue};
                }`;case"linear":return`output[global_idx] = ${a.length===2||a.length===4?"bilinearInterpolation":"trilinearInterpolation"}(output_indices);`;case"cubic":return"output[global_idx] = bicubicInterpolation(output_indices);";default:throw Error(`Unsupported resize mode: ${r.mode}`)}})()};
`}
      }`;return{name:"Resize",shaderCache:{hint:`${r.cacheKey}|${t}|${c.length>0?r.mode==="cubic"?c:c.length:""}|${o.length>0?o:""}|${i.length>0?i:""}|${f}|${r.mode==="nearest"?a.length:a}`,inputDependencies:["rank"]},getShaderSource:w,getRunData:()=>({outputs:[{dims:l,dataType:e.dataType}],dispatchGroup:{x:Math.ceil(u/64)},programUniforms:[{type:12,data:u},{type:1,data:c},{type:1,data:i},...dt(a,l)]})}},hg=e=>{let r=e.customDataBuffer;return new Uint32Array(r,r.byteOffset,1)[0]},_g=(e,r)=>{let t=[],s=[],o=[],n=hg(e);if(r.antialias!==0)throw Error("Only default value (0) for Antialias attribute is supported");tg(e.inputs,r,n,t,s,o),e.compute(mg(e.inputs[0],r,n,t,s,o),{inputs:[0]})},fg=e=>{let r=e.antialias,t=e.axes,s=e.coordinateTransformMode,o=e.cubicCoeffA,n=e.excludeOutside!==0,a=e.extrapolationValue,i=e.keepAspectRatioPolicy,l=e.mode,c=e.nearestMode===""?"simple":e.nearestMode;return Kt({antialias:r,axes:t,coordinateTransformMode:s,cubicCoeffA:o,excludeOutside:n,extrapolationValue:a,keepAspectRatioPolicy:i,mode:l,nearestMode:c})}}),gg,wg,Mg,wT=Ue(()=>{Mt(),Pt(),Ct(),gg=e=>{if(!e||e.length<3)throw new Error("layerNorm requires at least 3 inputs.");let r=e[0],t=e[1],s=e[2];if(r.dataType!==t.dataType||r.dataType!==s.dataType)throw new Error("All inputs must have the same data type");if(r.dims.length!==3&&r.dims.length!==2)throw new Error("Input must be 2D or 3D");if(t.dims.length!==3&&t.dims.length!==2)throw new Error("Skip must be 2D or 3D");let o=r.dims[r.dims.length-1],n=r.dims[r.dims.length-2];if(t.dims[t.dims.length-1]!==o)throw new Error("Skip must have the same hidden size as input");if(t.dims[t.dims.length-2]!==n)throw new Error("Skip must have the same sequence length as input");if(s.dims.length!==1)throw new Error("Gamma must be 1D");if(s.dims[s.dims.length-1]!==o)throw new Error("Gamma must have the same hidden size as input");if(e.length>3){let a=e[3];if(a.dims.length!==1)throw new Error("Beta must be 1D");if(a.dims[a.dims.length-1]!==o)throw new Error("Beta must have the same hidden size as input")}if(e.length>4){let a=e[4];if(a.dims.length!==1)throw new Error("Bias must be 1D");if(a.dims[a.dims.length-1]!==o)throw new Error("Bias must have the same hidden size as input")}},wg=(e,r,t,s)=>{let o=r.simplified,n=e[0].dims,a=ye.size(n),i=n,l=a,c=n.slice(-1)[0],p=s?n.slice(0,-1).concat(1):[],d=!o&&e.length>3,u=e.length>4,f=s&&t>1,_=s&&t>2,b=t>3,k=64,w=lr(c),y=[{type:12,data:l},{type:12,data:w},{type:12,data:c},{type:1,data:r.epsilon}],$=M=>{let x=[{name:"output_size",type:"u32"},{name:"components",type:"u32"},{name:"hidden_size",type:"u32"},{name:"epsilon",type:"f32"}],v=[Ie("x",e[0].dataType,e[0].dims,w),Ie("skip",e[1].dataType,e[1].dims,w),Ie("gamma",e[2].dataType,e[2].dims,w)];d&&v.push(Ie("beta",e[3].dataType,e[3].dims,w)),u&&v.push(Ie("bias",e[4].dataType,e[4].dims,w)),v.push(nt("output",e[0].dataType,i,w)),f&&v.push(nt("mean_output",1,p)),_&&v.push(nt("inv_std_output",1,p)),b&&v.push(nt("input_skip_bias_sum",e[0].dataType,i,w));let C=Ir(e[0].dataType),F=Ir(1,w);return`

      ${M.registerUniforms(x).declareVariables(...v)}
      var<workgroup> sum_shared : array<${F}, ${k}>;
      var<workgroup> sum_squared_shared : array<${F}, ${k}>;

      ${M.mainStart([k,1,1])}
        let ix = local_id.x;
        let iy = global_id.x / ${k};

        let hidden_size_vectorized: u32 = uniforms.hidden_size / uniforms.components;
        var stride = hidden_size_vectorized / ${k};
        let offset = ix * stride + iy * hidden_size_vectorized;
        let offset1d = stride * ix;
        if (ix == ${k-1}) {
          stride = hidden_size_vectorized - stride * ix;
        }
        for (var i: u32 = 0; i < stride; i++) {
          let skip_value = skip[offset + i];
          let bias_value = ${u?"bias[offset1d + i]":C+"(0.0)"};
          let input_value = x[offset + i];
          let value = input_value + skip_value + bias_value;
          ${b?"input_skip_bias_sum[offset + i] = value;":""}
          output[offset + i] = value;
          let f32_value = ${Hn(C,w,"value")};
          sum_shared[ix] += f32_value;
          sum_squared_shared[ix] += f32_value * f32_value;
        }
        workgroupBarrier();

        var reduce_size : u32 = ${k};
        for (var curr_size = reduce_size >> 1;  curr_size > 0; curr_size = reduce_size >> 1) {
          reduce_size = curr_size + (reduce_size & 1);
          if (ix < curr_size) {
            sum_shared[ix] += sum_shared[ix + reduce_size];
            sum_squared_shared[ix] += sum_squared_shared[ix + reduce_size];
          }
          workgroupBarrier();
        }

        let sum = sum_shared[0];
        let square_sum = sum_squared_shared[0];
        let mean = ${en("sum",w)} / f32(uniforms.hidden_size);
        let inv_std_dev = inverseSqrt(${en("square_sum",w)} / f32(uniforms.hidden_size) ${o?"":"- mean * mean"} + uniforms.epsilon);
        ${f?"mean_output[global_idx] = mean;":""}
        ${_?"inv_std_output[global_idx] = inv_std_dev;":""}

        for (var i: u32 = 0; i < stride; i++) {
          output[offset + i] = (output[offset + i] ${o?"":`- ${C}(mean)`}) *
            ${C}(inv_std_dev) * gamma[offset1d + i]
            ${d?"+ beta[offset1d + i]":""};
        }
      }`},T=[{dims:i,dataType:e[0].dataType}];return t>1&&T.push({dims:p,dataType:1}),t>2&&T.push({dims:p,dataType:1}),t>3&&T.push({dims:n,dataType:e[0].dataType}),{name:"SkipLayerNormalization",shaderCache:{hint:`${w};${f};${_};${b}`,inputDependencies:e.map((M,x)=>"type")},getShaderSource:$,getRunData:()=>({outputs:T,dispatchGroup:{x:Math.ceil(l/c)},programUniforms:y})}},Mg=(e,r)=>{gg(e.inputs);let t=[0];e.outputCount>1&&t.push(-3),e.outputCount>2&&t.push(-3),e.outputCount>3&&t.push(3),e.compute(wg(e.inputs,r,e.outputCount,!1),{outputs:t})}}),bg,Eo,yg,ec,vg,xg,Tg,Eg,MT=Ue(()=>{Mt(),Pt(),pr(),Ct(),bg=(e,r)=>{if(!e||e.length<1)throw new Error("too few inputs");if(r.axes.length!==0){if(r.axes.length!==r.starts.length||r.axes.length!==r.ends.length)throw new Error("axes, starts and ends must have the same length")}else if(r.starts.length!==r.ends.length)throw new Error("starts and ends must have the same length");e.slice(1).forEach((t,s)=>{if(e[s+1].dataType!==6&&e[s+1].dataType!==7)throw new Error(`Input ${s} must be an array of int32 or int64`)})},Eo=(e,r)=>{let t=[];if(e.length>r)if(e[r].dataType===7)e[r].getBigInt64Array().forEach(s=>t.push(Number(s)));else if(e[r].dataType===6)e[r].getInt32Array().forEach(s=>t.push(Number(s)));else throw new Error(`Input ${r} must be an array of int32 or int64`);return t},yg=(e,r)=>{if(e.length>1){let t=Eo(e,1),s=Eo(e,2),o=Eo(e,3);return o.length===0&&(o=[...Array(e[0].dims.length).keys()]),Kt({starts:t,ends:s,axes:o})}else return r},ec=(e,r,t,s,o)=>{let n=e;return e<0&&(n+=t[s[r]]),o[r]<0?Math.max(0,Math.min(n,t[s[r]]-1)):Math.max(0,Math.min(n,t[s[r]]))},vg=(e,r,t)=>`fn calculateInputIndices(output_indices: ${r.type.indices}) -> ${e.type.indices} {
          var input_indices: ${e.type.indices};
          var carry = 0u;
          for (var i = ${t.length}; i >= 0; i--) {
            let input_shape_i = ${ct("uniforms.input_shape","i",t.length)};
            let steps_i = ${ct("uniforms.steps","i",t.length)};
            let signs_i = ${ct("uniforms.signs","i",t.length)};
            let starts_i = ${ct("uniforms.starts","i",t.length)};
            var output_index = ${r.indicesGet("output_indices","i")};
            var input_index = output_index * steps_i + starts_i + carry;
            carry = input_index / input_shape_i;
            input_index = input_index % input_shape_i;
            if (signs_i < 0) {
              input_index = input_shape_i - input_index - 1u + starts_i;
            }
            ${e.indicesSet("input_indices","i","input_index")};
          }
          return input_indices;
      }`,xg=(e,r)=>{let t=e[0].dims,s=ye.size(t),o=r.axes.length>0?ye.normalizeAxes(r.axes,t.length):[...Array(t.length).keys()],n=Eo(e,4);n.forEach(w=>w!==0||(()=>{throw new Error("step cannot be 0")})),n.length===0&&(n=Array(o.length).fill(1));let a=r.starts.map((w,y)=>ec(w,y,t,o,n)),i=r.ends.map((w,y)=>ec(w,y,t,o,n));if(o.length!==a.length||o.length!==i.length)throw new Error("start, ends and axes should have the same number of elements");if(o.length!==t.length)for(let w=0;w<t.length;++w)o.includes(w)||(a.splice(w,0,0),i.splice(w,0,t[w]),n.splice(w,0,1));let l=n.map(w=>Math.sign(w));n.forEach((w,y,$)=>{if(w<0){let T=(i[y]-a[y])/w,M=a[y],x=M+T*n[y];a[y]=x,i[y]=M,$[y]=-w}});let c=t.slice(0);o.forEach((w,y)=>{c[w]=Math.ceil((i[w]-a[w])/n[w])});let p={dims:c,dataType:e[0].dataType},d=nt("output",e[0].dataType,c.length),u=Ie("input",e[0].dataType,e[0].dims.length),f=ye.size(c),_=[{name:"outputSize",type:"u32"},{name:"starts",type:"u32",length:a.length},{name:"signs",type:"i32",length:l.length},{name:"steps",type:"u32",length:n.length}],b=[{type:12,data:f},{type:12,data:a},{type:6,data:l},{type:12,data:n},...dt(e[0].dims,c)],k=w=>`
      ${w.registerUniforms(_).declareVariables(u,d)}
        ${vg(u,d,t)}
        ${w.mainStart()}
          ${w.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.outputSize")}
          let output_indices = ${d.offsetToIndices("global_idx")};
          let input_indices = calculateInputIndices(output_indices);
          ${d.setByOffset("global_idx",u.getByIndices("input_indices"))}
      }`;return{name:"Slice",shaderCache:{hint:`${l.length}_${a.length}_${n.length}`,inputDependencies:["rank"]},getShaderSource:k,getRunData:()=>({outputs:[p],dispatchGroup:{x:Math.ceil(s/64)},programUniforms:b})}},Tg=(e,r)=>{bg(e.inputs,r);let t=yg(e.inputs,r);e.compute(xg(e.inputs,t),{inputs:[0]})},Eg=e=>{let r=e.starts,t=e.ends,s=e.axes;return Kt({starts:r,ends:t,axes:s})}}),Pg,Cg,Sg,$g,bT=Ue(()=>{Mt(),Pt(),pr(),tn(),Ct(),Pg=e=>{if(!e||e.length!==1)throw new Error("Softmax op requires 1 input.")},Cg=(e,r)=>{let t=e.inputs[0],s=t.dims,o=ye.size(s),n=s.length,a=ye.normalizeAxis(r.axis,n),i=a<s.length-1,l,c=[];i?(c=Array.from({length:n},(v,C)=>C),c[a]=n-1,c[n-1]=a,l=e.compute(rs(t,c),{inputs:[t],outputs:[-1]})[0]):l=t;let p=l.dims,d=p[n-1],u=o/d,f=lr(d),_=d/f,b=64;u===1&&(b=256);let k=(v,C)=>C===4?`max(max(${v}.x, ${v}.y), max(${v}.z, ${v}.w))`:C===2?`max(${v}.x, ${v}.y)`:C===3?`max(max(${v}.x, ${v}.y), ${v}.z)`:v,w=Ie("x",l.dataType,l.dims,f),y=nt("result",l.dataType,l.dims,f),$=w.type.value,T=Ir(l.dataType)==="f32"?`var threadMax = ${$}(-3.402823e+38f);`:`var threadMax = ${$}(-65504.0h);`,M=v=>`
      var<workgroup> rowMaxShared : ${$};
      var<workgroup> rowSumShared : ${$};
      var<workgroup> threadShared : array<${$}, ${b}>;

      fn getValue(row: i32, col: i32, row_stride: i32) -> ${$} {
        let index = row * row_stride + col;
        return x[index];
      }

      fn setValue(row: i32, col: i32, row_stride: i32, value: ${$}) {
        let index = row * row_stride + col;
        result[index] = value;
      }
      ${v.registerUniform("packedCols","i32").declareVariables(w,y)}
      ${v.mainStart(b)}
        let gindex = i32(global_idx);
        let lindex = i32(local_idx);
        const wg = ${b};
        let row = gindex / wg;
        let cols = uniforms.packedCols;
        let row_stride : i32 = uniforms.packedCols;

        // find the rows max
        ${T}
        for (var col = lindex; col < cols; col += wg) {
          let value = getValue(row, col, row_stride);
          threadMax = max(threadMax, value);
        }
        if (lindex < cols) {
          threadShared[lindex] = threadMax;
        }
        workgroupBarrier();

        var reduceSize = min(cols, wg);
        for (var currSize = reduceSize >> 1;  currSize > 0; currSize = reduceSize >> 1) {
          reduceSize = currSize + (reduceSize & 1);
          if (lindex < currSize) {
            threadShared[lindex] = max(threadShared[lindex], threadShared[lindex + reduceSize]);
          }
          workgroupBarrier();
        }
        if (lindex == 0) {
          rowMaxShared = ${$}(${k("threadShared[0]",f)});
        }
        workgroupBarrier();

        // find the rows sum
        var threadSum = ${$}(0.0);
        for (var col = lindex; col < cols; col += wg) {
          let subExp = exp(getValue(row, col, row_stride) - rowMaxShared);
          threadSum += subExp;
        }
        threadShared[lindex] = threadSum;
        workgroupBarrier();

        for (var currSize = wg >> 1;  currSize > 0; currSize = currSize >> 1) {
          if (lindex < currSize) {
            threadShared[lindex] = threadShared[lindex] + threadShared[lindex + currSize];
          }
          workgroupBarrier();
        }
        if (lindex == 0) {
          rowSumShared = ${$}(${en("threadShared[0]",f)});
        }
        workgroupBarrier();

        // calculate final value for each element in the row
        for (var col = lindex; col < cols; col += wg) {
          let value = exp(getValue(row, col, row_stride) - rowMaxShared) / rowSumShared;
          setValue(row, col, row_stride, value);
        }
      }`,x=e.compute({name:"Softmax",shaderCache:{hint:`${f};${b}`,inputDependencies:["type"]},getRunData:()=>({outputs:[{dims:p,dataType:l.dataType}],dispatchGroup:{x:u},programUniforms:[{type:6,data:_}]}),getShaderSource:M},{inputs:[l],outputs:[i?-1:0]})[0];i&&e.compute(rs(x,c),{inputs:[x]})},Sg=(e,r)=>{Pg(e.inputs),Cg(e,r)},$g=e=>Kt({axis:e.axis})}),tc,kg,Ig,Ag,Fg,yT=Ue(()=>{Mt(),Pt(),Ct(),tc=e=>Array.from(e.getBigInt64Array(),Number),kg=e=>{if(!e||e.length!==2)throw new Error("Tile requires 2 inputs.");if(e[0].dataType!==1&&e[0].dataType!==10&&e[0].dataType!==6&&e[0].dataType!==12)throw new Error("Tile only support float, float16, int32, and uint32 data types");if(e[1].dataType!==7)throw new Error("Tile `repeats` input should be of int64 data type");if(e[1].dims.length!==1)throw new Error("Tile `repeats` input should be 1-D");if(tc(e[1]).length!==e[0].dims.length)throw new Error("Tile `repeats` input should have same number of elements as rank of input data tensor")},Ig=(e,r)=>{let t=[];for(let s=0;s<e.length;++s)t.push(e[s]*r[s]);return t},Ag=(e,r)=>{let t=e[0].dims,s=r??tc(e[1]),o=Ig(t,s),n=ye.size(o),a=e[0].dataType,i=Ie("input",a,t.length),l=nt("output",a,o.length),c=p=>`
      const inputShape = ${i.indices(...t)};
      ${p.registerUniform("output_size","u32").declareVariables(i,l)}
      ${p.mainStart()}
      ${p.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.output_size")}
      let output_indices = ${l.offsetToIndices("global_idx")};
      var input_indices: ${i.type.indices};
      for (var i = 0; i < ${t.length}; i++) {
        let input_dim_i = ${i.indicesGet("uniforms.input_shape","i")};
        let input_dim_value = ${l.indicesGet("output_indices","i")}  % input_dim_i;

        ${i.indicesSet("input_indices","i","input_dim_value")}
      }
      ${l.setByOffset("global_idx",i.getByIndices("input_indices"))}
    }`;return{name:"Tile",shaderCache:{hint:`${s}`,inputDependencies:["rank"]},getRunData:()=>({outputs:[{dims:o,dataType:e[0].dataType}],dispatchGroup:{x:Math.ceil(n/64)},programUniforms:[{type:12,data:n},...dt(e[0].dims,o)]}),getShaderSource:c}},Fg=e=>{kg(e.inputs),e.compute(Ag(e.inputs),{inputs:[0]})}}),Og,Dg,Lg,vT=Ue(()=>{Mt(),Pt(),Ct(),Og=(e,r,t,s,o)=>{let n=nt("output_data",o,t.length,4),a=Ie("a_data",r[1].dataType,r[1].dims.length,4),i=Ie("b_data",r[2].dataType,r[2].dims.length,4),l=Ie("c_data",r[0].dataType,r[0].dims.length,4),c,p=(d,u,f)=>`select(${u}, ${d}, ${f})`;if(!s)c=n.setByOffset("global_idx",p(a.getByOffset("global_idx"),i.getByOffset("global_idx"),l.getByOffset("global_idx")));else{let d=(u,f,_="")=>{let b=`a_data[index_a${f}][component_a${f}]`,k=`b_data[index_b${f}][component_b${f}]`,w=`bool(c_data[index_c${f}] & (0xffu << (component_c${f} * 8)))`;return`
            let output_indices${f} = ${n.offsetToIndices(`global_idx * 4u + ${f}u`)};
            let offset_a${f} = ${a.broadcastedIndicesToOffset(`output_indices${f}`,n)};
            let offset_b${f} = ${i.broadcastedIndicesToOffset(`output_indices${f}`,n)};
            let offset_c${f} = ${l.broadcastedIndicesToOffset(`output_indices${f}`,n)};
            let index_a${f} = offset_a${f} / 4u;
            let index_b${f} = offset_b${f} / 4u;
            let index_c${f} = offset_c${f} / 4u;
            let component_a${f} = offset_a${f} % 4u;
            let component_b${f} = offset_b${f} % 4u;
            let component_c${f} = offset_c${f} % 4u;
            ${u}[${f}] = ${_}(${p(b,k,w)});
          `};o===9?c=`
            var data = vec4<u32>(0);
            ${d("data",0,"u32")}
            ${d("data",1,"u32")}
            ${d("data",2,"u32")}
            ${d("data",3,"u32")}
            output_data[global_idx] = dot(vec4<u32>(0x1, 0x100, 0x10000, 0x1000000), vec4<u32>(data));`:c=`
            ${d("output_data[global_idx]",0)}
            ${d("output_data[global_idx]",1)}
            ${d("output_data[global_idx]",2)}
            ${d("output_data[global_idx]",3)}
          `}return`
        ${e.registerUniform("vec_size","u32").declareVariables(l,a,i,n)}
        ${e.mainStart()}
        ${e.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.vec_size")}
        ${c}
      }`},Dg=e=>{let r=e[1].dims,t=e[2].dims,s=e[0].dims,o=e[1].dataType,n=!(ye.areEqual(r,t)&&ye.areEqual(t,s)),a=r,i=ye.size(r);if(n){let c=Gn.calcShape(Gn.calcShape(r,t,!1),s,!1);if(!c)throw new Error("Can't perform where op on the given tensors");a=c,i=ye.size(a)}let l=Math.ceil(i/4);return{name:"Where",shaderCache:{inputDependencies:["rank","rank","rank"]},getShaderSource:c=>Og(c,e,a,n,o),getRunData:()=>({outputs:[{dims:a,dataType:o}],dispatchGroup:{x:Math.ceil(i/64/4)},programUniforms:[{type:12,data:l},...dt(s,r,t,a)]})}},Lg=e=>{e.compute(Dg(e.inputs))}}),zg,xT=Ue(()=>{zx(),_l(),Bx(),Rx(),jx(),Nx(),Vx(),Hx(),Qx(),Xx(),Jx(),Yx(),Zx(),eT(),tT(),rT(),sT(),nT(),oT(),aT(),iT(),lT(),cT(),uT(),dT(),q_(),pT(),mT(),hT(),_T(),fT(),pl(),gT(),nf(),wT(),MT(),bT(),tf(),yT(),tn(),Ml(),vT(),zg=new Map([["Abs",[fm]],["Acos",[gm]],["Acosh",[wm]],["Add",[oh]],["ArgMax",[tm,hl]],["ArgMin",[em,hl]],["Asin",[Mm]],["Asinh",[bm]],["Atan",[ym]],["Atanh",[vm]],["Attention",[im]],["AveragePool",[Df,Of]],["BatchNormalization",[dm]],["BiasAdd",[hm]],["BiasSplitGelu",[rh]],["Cast",[Tm,xm]],["Ceil",[Cm]],["Clip",[Pm]],["Concat",[Mh,bh]],["Conv",[Al,kl]],["ConvTranspose",[Hh,Wh]],["Cos",[Sm]],["Cosh",[$m]],["CumSum",[Qh,Xh]],["DepthToSpace",[e_,t_]],["DequantizeLinear",[Wf,Gf]],["Div",[ah]],["Einsum",[i_,l_]],["Elu",[km,Mo]],["Equal",[ih]],["Erf",[Im]],["Exp",[Am]],["Expand",[p_]],["FastGelu",[h_]],["Floor",[Fm]],["FusedConv",[Al,kl]],["Gather",[w_,g_]],["GatherElements",[$_,S_]],["GatherBlockQuantized",[T_,E_]],["GatherND",[b_,y_]],["Gelu",[Om]],["Gemm",[F_,A_]],["GlobalAveragePool",[zf,Lf]],["GlobalMaxPool",[Nf,jf]],["Greater",[dh]],["GreaterOrEqual",[mh]],["GridSample",[V_,U_]],["GroupQueryAttention",[cf]],["HardSigmoid",[Vm,Nm]],["InstanceNormalization",[pf]],["LayerNormalization",[_f]],["LeakyRelu",[Dm,Mo]],["Less",[ph]],["LessOrEqual",[hh]],["Log",[Xm]],["MatMul",[gf]],["MatMulNBits",[yf,vf]],["MaxPool",[Bf,Rf]],["Mul",[lh]],["MultiHeadAttention",[H_,G_]],["Neg",[zm]],["Not",[Lm]],["Pad",[If]],["Pow",[ch]],["QuickGelu",[Zm,Mo]],["Range",[qf]],["Reciprocal",[Bm]],["ReduceMin",[Qp]],["ReduceMean",[Wp]],["ReduceMax",[qp]],["ReduceSum",[Jp]],["ReduceProd",[Xp]],["ReduceL1",[Gp]],["ReduceL2",[Kp]],["ReduceLogSum",[Zp]],["ReduceLogSumExp",[Hp]],["ReduceSumSquare",[Yp]],["Relu",[Rm]],["Resize",[_g,fg]],["RotaryEmbedding",[sf]],["ScatterND",[Yf,Jf]],["Sigmoid",[jm]],["Sin",[Um]],["Sinh",[Wm]],["Slice",[Tg,Eg]],["SkipLayerNormalization",[Mg]],["Split",[Z_,ef]],["Sqrt",[Gm]],["Softmax",[Sg,$g]],["Sub",[uh]],["Tan",[Km]],["Tanh",[Hm]],["ThresholdedRelu",[Qm,Mo]],["Tile",[Fg]],["Transpose",[dp,pp]],["Where",[Lg]]])}),Bg,TT=Ue(()=>{fs(),Ns(),Ct(),Bg=class{constructor(e){this.backend=e,this.repo=new Map,this.attributesBound=!1}getArtifact(e){return this.repo.get(e)}setArtifact(e,r){this.repo.set(e,r)}run(e,r,t,s,o){_s(e.programInfo.name);let n=this.backend.device,a=this.backend.getComputePassEncoder();this.backend.writeTimestamp(this.backend.pendingDispatchNumber*2);let i=[];for(let c of r)i.push({binding:i.length,resource:{buffer:c.buffer}});for(let c of t)i.push({binding:i.length,resource:{buffer:c.buffer}});o&&i.push({binding:i.length,resource:o});let l=n.createBindGroup({layout:e.computePipeline.getBindGroupLayout(0),entries:i,label:e.programInfo.name});if(this.backend.sessionStatus==="capturing"){let c={kernelId:this.backend.currentKernelId,computePipeline:e.computePipeline,bindGroup:l,dispatchGroup:s};this.backend.capturedCommandList.get(this.backend.currentSessionId).push(c)}a.setPipeline(e.computePipeline),a.setBindGroup(0,l),a.dispatchWorkgroups(...s),this.backend.writeTimestamp(this.backend.pendingDispatchNumber*2+1),this.backend.pendingDispatchNumber++,(this.backend.pendingDispatchNumber>=this.backend.maxDispatchNumber||this.backend.queryType==="at-passes")&&this.backend.endComputePass(),this.backend.pendingDispatchNumber>=this.backend.maxDispatchNumber&&this.backend.flush(),is(e.programInfo.name)}dispose(){}build(e,r){_s(e.name);let t=this.backend.device,s=[];[{feature:"shader-f16",extension:"f16"},{feature:"subgroups",extension:"subgroups"}].forEach(c=>{t.features.has(c.feature)&&s.push(`enable ${c.extension};`)});let o=op(r,this.backend.device.limits),n=e.getShaderSource(o),a=`${s.join(`
`)}
${o.additionalImplementations}
${n}`,i=t.createShaderModule({code:a,label:e.name});zt("verbose",()=>`[WebGPU] ${e.name} shader code: ${a}`);let l=t.createComputePipeline({compute:{module:i,entryPoint:"main"},layout:"auto",label:e.name});return is(e.name),{programInfo:e,computePipeline:l,uniformVariablesInfo:o.variablesInfo}}normalizeDispatchGroupSize(e){let r=typeof e=="number"?e:e.x,t=typeof e=="number"?1:e.y||1,s=typeof e=="number"?1:e.z||1,o=this.backend.device.limits.maxComputeWorkgroupsPerDimension;if(r<=o&&t<=o&&s<=o)return[r,t,s];let n=r*t*s,a=Math.ceil(Math.sqrt(n));if(a>o){if(a=Math.ceil(Math.cbrt(n)),a>o)throw new Error("Total dispatch size exceeds WebGPU maximum.");return[a,a,a]}else return[a,a,1]}}}),Rg={};Un(Rg,{WebGpuBackend:()=>Ug});var jg,Ng,Vg,Ug,ET=Ue(()=>{fs(),Mt(),Ns(),Gd(),Dx(),xT(),TT(),jg=(e,r)=>{if(r.length!==e.length)throw new Error(`inputDependencies length ${r.length} is not equal to inputTensors length ${e.length}.`);let t=[];for(let s=0;s<e.length;++s){let o=e[s].dataType;switch(r[s]){case"none":{t.push("");break}case"type":{t.push(`${o}`);break}case"rank":{let n=e[s].dims.length;t.push(`${o};${n}`);break}case"dims":{let n=e[s].dims.join(",");t.push(`${o};${n}`);break}default:throw new Error(`unsupported input dependency: ${r[s]}`)}}return t.join("|")},Ng=(e,r,t)=>{var o,n;let s=e.name;return(o=e.shaderCache)!=null&&o.hint&&(s+="["+e.shaderCache.hint+"]"),s+=":"+t+`:${jg(r,((n=e.shaderCache)==null?void 0:n.inputDependencies)??new Array(r.length).fill("dims"))}`,s},Vg=class{constructor(e){e&&(this.architecture=e.architecture,this.vendor=e.vendor)}isArchitecture(e){return this.architecture===e}isVendor(e){return this.vendor===e}},Ug=class{constructor(){this.currentSessionId=null,this.currentKernelId=null,this.commandEncoder=null,this.computePassEncoder=null,this.maxDispatchNumber=16,this.pendingDispatchNumber=0,this.pendingKernels=[],this.pendingQueries=new Map,this.sessionStatus="default",this.capturedCommandList=new Map,this.capturedPendingKernels=new Map,this.sessionExternalDataMapping=new Map}get currentKernelCustomData(){if(this.currentKernelId===null)throw new Error("currentKernelCustomData(): currentKernelId is null. (should not happen)");let e=this.kernelCustomData.get(this.currentKernelId);return e||(e={},this.kernelCustomData.set(this.currentKernelId,e)),e}async initialize(e,r){this.env=e;let t=[],s={requiredLimits:{maxComputeWorkgroupStorageSize:r.limits.maxComputeWorkgroupStorageSize,maxComputeWorkgroupsPerDimension:r.limits.maxComputeWorkgroupsPerDimension,maxStorageBufferBindingSize:r.limits.maxStorageBufferBindingSize,maxBufferSize:r.limits.maxBufferSize,maxComputeInvocationsPerWorkgroup:r.limits.maxComputeInvocationsPerWorkgroup,maxComputeWorkgroupSizeX:r.limits.maxComputeWorkgroupSizeX,maxComputeWorkgroupSizeY:r.limits.maxComputeWorkgroupSizeY,maxComputeWorkgroupSizeZ:r.limits.maxComputeWorkgroupSizeZ},requiredFeatures:t},o=n=>r.features.has(n)&&t.push(n)&&!0;o("chromium-experimental-timestamp-query-inside-passes")||o("timestamp-query"),o("shader-f16"),o("subgroups"),this.device=await r.requestDevice(s),this.adapterInfo=new Vg(r.info||await r.requestAdapterInfo()),this.gpuDataManager=tp(this),this.programManager=new Bg(this),this.kernels=new Map,this.kernelPersistentData=new Map,this.kernelCustomData=new Map,Xi(e.logLevel,!!e.debug),this.device.onuncapturederror=n=>{n.error instanceof GPUValidationError&&console.error(`An uncaught WebGPU validation error was raised: ${n.error.message}`)},Object.defineProperty(this.env.webgpu,"device",{value:this.device,writable:!1,enumerable:!0,configurable:!1}),Object.defineProperty(this.env.webgpu,"adapter",{value:r,writable:!1,enumerable:!0,configurable:!1}),this.setQueryType()}dispose(){typeof this.querySet<"u"&&this.querySet.destroy(),this.gpuDataManager.dispose()}getCommandEncoder(){return this.commandEncoder||(this.commandEncoder=this.device.createCommandEncoder()),this.commandEncoder}getComputePassEncoder(){if(!this.computePassEncoder){let e=this.getCommandEncoder(),r={};this.queryType==="at-passes"&&(r.timestampWrites={querySet:this.querySet,beginningOfPassWriteIndex:this.pendingDispatchNumber*2,endOfPassWriteIndex:this.pendingDispatchNumber*2+1}),this.computePassEncoder=e.beginComputePass(r)}return this.computePassEncoder}endComputePass(){this.computePassEncoder&&(this.computePassEncoder.end(),this.computePassEncoder=null)}flush(){if(!this.commandEncoder)return;_s(),this.endComputePass();let e;this.queryType!=="none"&&(this.commandEncoder.resolveQuerySet(this.querySet,0,this.pendingDispatchNumber*2,this.queryResolveBuffer,0),e=this.device.createBuffer({size:this.pendingDispatchNumber*2*8,usage:GPUBufferUsage.MAP_READ|GPUBufferUsage.COPY_DST}),this.pendingQueries.set(e,this.pendingKernels),this.pendingKernels=[],this.commandEncoder.copyBufferToBuffer(this.queryResolveBuffer,0,e,0,this.pendingDispatchNumber*2*8)),this.device.queue.submit([this.commandEncoder.finish()]),this.gpuDataManager.refreshPendingBuffers(),this.commandEncoder=null,this.pendingDispatchNumber=0,this.queryType!=="none"&&e.mapAsync(GPUMapMode.READ).then(()=>{var s;let r=new BigUint64Array(e.getMappedRange()),t=this.pendingQueries.get(e);for(let o=0;o<r.length/2;o++){let n=t[o],a=n.kernelId,i=this.kernels.get(a),l=i.kernelType,c=i.kernelName,p=n.programName,d=n.inputTensorViews,u=n.outputTensorViews,f=r[o*2],_=r[o*2+1];typeof this.queryTimeBase>"u"&&(this.queryTimeBase=f);let b=Number(f-this.queryTimeBase),k=Number(_-this.queryTimeBase);if(!Number.isSafeInteger(b)||!Number.isSafeInteger(k))throw new RangeError("incorrect timestamp range");if((s=this.env.webgpu.profiling)!=null&&s.ondata)this.env.webgpu.profiling.ondata({version:1,inputsMetadata:d.map(w=>({dims:w.dims,dataType:js(w.dataType)})),outputsMetadata:u.map(w=>({dims:w.dims,dataType:js(w.dataType)})),kernelId:a,kernelType:l,kernelName:c,programName:p,startTime:b,endTime:k});else{let w="";d.forEach(($,T)=>{w+=`input[${T}]: [${$.dims}] | ${js($.dataType)}, `});let y="";u.forEach(($,T)=>{y+=`output[${T}]: [${$.dims}] | ${js($.dataType)}, `}),console.log(`[profiling] kernel "${a}|${l}|${c}|${p}" ${w}${y}execution time: ${k-b} ns`)}ho("GPU",`${p}::${f}::${_}`)}e.unmap(),this.pendingQueries.delete(e)}),is()}run(e,r,t,s,o,n){_s(e.name);let a=[];for(let y=0;y<r.length;++y){let $=r[y].data;if($===0)continue;let T=this.gpuDataManager.get($);if(!T)throw new Error(`no GPU data for input: ${$}`);a.push(T)}let{outputs:i,dispatchGroup:l,programUniforms:c}=e.getRunData(r),p=t.length===0?i.map((y,$)=>$):t;if(p.length!==i.length)throw new Error(`Output size ${p.length} must be equal to ${i.length}.`);let d=[],u=[];for(let y=0;y<i.length;++y){if(!Number.isInteger(p[y])||p[y]<-3||p[y]>=n)throw new Error(`Invalid output index: ${p[y]}`);if(p[y]===-3)continue;let $=p[y]===-1,T=p[y]===-2,M=$||T?o(i[y].dataType,i[y].dims):s(p[y],i[y].dataType,i[y].dims);if(d.push(M),M.data===0)continue;let x=this.gpuDataManager.get(M.data);if(!x)throw new Error(`no GPU data for output: ${M.data}`);if($&&this.temporaryData.push(x),T){let v=this.kernelPersistentData.get(this.currentKernelId);v||(v=[],this.kernelPersistentData.set(this.currentKernelId,v)),v.push(x)}u.push(x)}if(a.length!==r.length||u.length!==d.length){if(u.length===0)return is(e.name),d;throw new Error(`Program ${e.name} has zero-sized tensor(s) in inputs or outputs. This is not supported now.`)}let f;if(c){let y=0,$=[];c.forEach(v=>{let C=typeof v.data=="number"?[v.data]:v.data;if(C.length===0)return;let F=v.type===10?2:4,D,K;v.type===10?(K=C.length>4?16:C.length>2?8:C.length*F,D=C.length>4?16:F*C.length):(K=C.length<=2?C.length*F:16,D=16),y=Math.ceil(y/K)*K,$.push(y);let U=v.type===10?8:4;y+=C.length>4?Math.ceil(C.length/U)*D:C.length*F});let T=16;y=Math.ceil(y/T)*T;let M=new ArrayBuffer(y);c.forEach((v,C)=>{let F=$[C],D=typeof v.data=="number"?[v.data]:v.data;if(v.type===6)new Int32Array(M,F,D.length).set(D);else if(v.type===12)new Uint32Array(M,F,D.length).set(D);else if(v.type===10)new Uint16Array(M,F,D.length).set(D);else if(v.type===1)new Float32Array(M,F,D.length).set(D);else throw new Error(`Unsupported uniform type: ${js(v.type)}`)});let x=this.gpuDataManager.create(y,GPUBufferUsage.COPY_DST|GPUBufferUsage.UNIFORM);this.device.queue.writeBuffer(x.buffer,0,M,0,y),this.gpuDataManager.release(x.id),f={offset:0,size:y,buffer:x.buffer}}let _=this.programManager.normalizeDispatchGroupSize(l),b=_[1]===1&&_[2]===1,k=Ng(e,r,b),w=this.programManager.getArtifact(k);if(w||(w=this.programManager.build(e,_),this.programManager.setArtifact(k,w),zt("info",()=>`[artifact] key: ${k}, programName: ${e.name}`)),c&&w.uniformVariablesInfo){if(c.length!==w.uniformVariablesInfo.length)throw new Error(`Uniform variables count mismatch: expect ${w.uniformVariablesInfo.length}, got ${c.length} in program "${w.programInfo.name}".`);for(let y=0;y<c.length;y++){let $=c[y],T=$.type,M=typeof $.data=="number"?1:$.data.length,[x,v]=w.uniformVariablesInfo[y];if(T!==x||M!==v)throw new Error(`Uniform variable ${y} mismatch: expect type ${x} with size ${v}, got type ${T} with size ${M} in program "${w.programInfo.name}".`)}}if(zt("info",()=>`[ProgramManager] run "${e.name}" (key=${k}) with ${_[0]}x${_[1]}x${_[2]}`),this.queryType!=="none"||this.sessionStatus==="capturing"){let y={kernelId:this.currentKernelId,programName:w.programInfo.name,inputTensorViews:r,outputTensorViews:d};this.pendingKernels.push(y),this.sessionStatus==="capturing"&&this.capturedPendingKernels.get(this.currentSessionId).push(y)}return this.programManager.run(w,a,u,_,f),is(e.name),d}upload(e,r){this.gpuDataManager.upload(e,r)}memcpy(e,r){this.gpuDataManager.memcpy(e,r)}async download(e,r){await this.gpuDataManager.download(e,r)}alloc(e){return this.gpuDataManager.create(e).id}free(e){return this.gpuDataManager.release(e)}createKernel(e,r,t,s){let o=zg.get(e);if(!o)throw new Error(`kernel not implemented: ${e}`);let n={kernelType:e,kernelName:s,kernelEntry:o[0],attributes:[o[1],t]};this.kernels.set(r,n)}releaseKernel(e){let r=this.kernelPersistentData.get(e);if(r){for(let t of r)this.gpuDataManager.release(t.id);this.kernelPersistentData.delete(e)}this.kernelCustomData.delete(e),this.kernels.delete(e)}computeKernel(e,r,t){let s=this.kernels.get(e);if(!s)throw new Error(`kernel not created: ${e}`);let o=s.kernelType,n=s.kernelName,a=s.kernelEntry,i=s.attributes;if(this.currentKernelId!==null)throw new Error(`kernel "[${o}] ${n}" is not allowed to be called recursively`);this.currentKernelId=e,i[0]&&(i[1]=i[0](i[1]),i[0]=void 0),zt("info",()=>`[WebGPU] Start to run kernel "[${o}] ${n}"...`);let l=this.env.debug;this.temporaryData=[];try{return l&&this.device.pushErrorScope("validation"),a(r,i[1]),0}catch(c){return t.push(Promise.resolve(`[WebGPU] Kernel "[${o}] ${n}" failed. ${c}`)),1}finally{l&&t.push(this.device.popErrorScope().then(c=>c?`GPU validation error for kernel "[${o}] ${n}": ${c.message}`:null));for(let c of this.temporaryData)this.gpuDataManager.release(c.id);this.temporaryData=[],this.currentKernelId=null}}registerBuffer(e,r,t,s){let o=this.sessionExternalDataMapping.get(e);o||(o=new Map,this.sessionExternalDataMapping.set(e,o));let n=o.get(r),a=this.gpuDataManager.registerExternalBuffer(t,s,n);return o.set(r,[a,t]),a}unregisterBuffers(e){let r=this.sessionExternalDataMapping.get(e);r&&(r.forEach(t=>this.gpuDataManager.unregisterExternalBuffer(t[0])),this.sessionExternalDataMapping.delete(e))}getBuffer(e){let r=this.gpuDataManager.get(e);if(!r)throw new Error(`no GPU data for buffer: ${e}`);return r.buffer}createDownloader(e,r,t){return async()=>{let s=await il(this,e,r);return Ji(s.buffer,t)}}writeTimestamp(e){this.queryType==="inside-passes"&&this.computePassEncoder.writeTimestamp(this.querySet,e)}setQueryType(){var e;this.queryType="none",(((e=this.env.webgpu.profiling)==null?void 0:e.mode)==="default"||(typeof this.env.trace>"u"?this.env.wasm.trace:this.env.trace))&&(this.device.features.has("chromium-experimental-timestamp-query-inside-passes")?this.queryType="inside-passes":this.device.features.has("timestamp-query")&&(this.queryType="at-passes"),this.queryType!=="none"&&typeof this.querySet>"u"&&(this.querySet=this.device.createQuerySet({type:"timestamp",count:this.maxDispatchNumber*2}),this.queryResolveBuffer=this.device.createBuffer({size:this.maxDispatchNumber*2*8,usage:GPUBufferUsage.COPY_SRC|GPUBufferUsage.QUERY_RESOLVE})))}captureBegin(){zt("info","captureBegin"),this.capturedCommandList.get(this.currentSessionId)||this.capturedCommandList.set(this.currentSessionId,[]),this.capturedPendingKernels.get(this.currentSessionId)||this.capturedPendingKernels.set(this.currentSessionId,[]),this.flush(),this.sessionStatus="capturing"}captureEnd(){zt("info","captureEnd"),this.flush(),this.sessionStatus="default"}replay(){zt("info","replay"),this.sessionStatus="replaying";let e=this.capturedCommandList.get(this.currentSessionId),r=this.capturedPendingKernels.get(this.currentSessionId),t=e.length;this.pendingKernels=[];for(let s=0;s<t;s++){let o=this.getComputePassEncoder(),n=e[s];this.writeTimestamp(this.pendingDispatchNumber*2),o.setPipeline(n.computePipeline),o.setBindGroup(0,n.bindGroup),o.dispatchWorkgroups(...n.dispatchGroup),this.writeTimestamp(this.pendingDispatchNumber*2+1),this.pendingDispatchNumber++,this.queryType!=="none"&&this.pendingKernels.push(r[s]),(this.pendingDispatchNumber>=this.maxDispatchNumber||this.queryType==="at-passes")&&this.endComputePass(),this.pendingDispatchNumber>=this.maxDispatchNumber&&this.flush()}this.flush(),this.sessionStatus="default"}onCreateSession(){this.gpuDataManager.onCreateSession()}onReleaseSession(e){this.unregisterBuffers(e),this.capturedCommandList.has(e)&&this.capturedCommandList.delete(e),this.capturedPendingKernels.has(e)&&this.capturedPendingKernels.delete(e),this.gpuDataManager.onReleaseSession(e)}onRunStart(e){this.currentSessionId=e,this.setQueryType()}}}),Wg={};Un(Wg,{init:()=>Kg});var Ma,Gg,Kg,PT=Ue(()=>{Mt(),Ns(),Pt(),Ox(),Ma=class Qv{constructor(r,t,s,o){this.module=r,this.dataType=t,this.data=s,this.dims=o}getFloat32Array(){if(this.dataType!==1)throw new Error("Invalid data type");let r=ye.size(this.dims);return r===0?new Float32Array:new Float32Array(this.module.HEAP8.buffer,this.data,r)}getBigInt64Array(){if(this.dataType!==7)throw new Error("Invalid data type");let r=ye.size(this.dims);return r===0?new BigInt64Array:new BigInt64Array(this.module.HEAP8.buffer,this.data,r)}getInt32Array(){if(this.dataType!==6)throw new Error("Invalid data type");let r=ye.size(this.dims);return r===0?new Int32Array:new Int32Array(this.module.HEAP8.buffer,this.data,r)}getUint16Array(){if(this.dataType!==10&&this.dataType!==4)throw new Error("Invalid data type");let r=ye.size(this.dims);return r===0?new Uint16Array:new Uint16Array(this.module.HEAP8.buffer,this.data,r)}reshape(r){if(ye.size(r)!==ye.size(this.dims))throw new Error("Invalid new shape");return new Qv(this.module,this.dataType,this.data,r)}},Gg=class{constructor(e,r,t){this.module=e,this.backend=r,this.customDataOffset=0,this.customDataSize=0,this.adapterInfo=r.adapterInfo;let s=e.PTR_SIZE,o=t/e.PTR_SIZE,n=s===4?"i32":"i64";this.opKernelContext=Number(e.getValue(s*o++,n));let a=Number(e.getValue(s*o++,n));this.outputCount=Number(e.getValue(s*o++,n)),this.customDataOffset=Number(e.getValue(s*o++,"*")),this.customDataSize=Number(e.getValue(s*o++,n));let i=[];for(let l=0;l<a;l++){let c=Number(e.getValue(s*o++,n)),p=Number(e.getValue(s*o++,"*")),d=Number(e.getValue(s*o++,n)),u=[];for(let f=0;f<d;f++)u.push(Number(e.getValue(s*o++,n)));i.push(new Ma(e,c,p,u))}this.inputs=i}get kernelCustomData(){return this.backend.currentKernelCustomData}get customDataBuffer(){return this.module.HEAPU8.subarray(this.customDataOffset,this.customDataOffset+this.customDataSize)}compute(e,r){var a;let t=((a=r==null?void 0:r.inputs)==null?void 0:a.map(i=>typeof i=="number"?this.inputs[i]:i))??this.inputs,s=(r==null?void 0:r.outputs)??[],o=(i,l,c)=>new Ma(this.module,l,this.output(i,c),c),n=(i,l)=>{let c=fn(i,l);if(!c)throw new Error(`Unsupported data type: ${i}`);let p=c>0?this.backend.gpuDataManager.create(c).id:0;return new Ma(this.module,i,p,l)};return this.backend.run(e,t,s,o,n,this.outputCount)}output(e,r){let t=this.module.stackSave();try{let s=this.module.PTR_SIZE,o=s===4?"i32":"i64",n=this.module.stackAlloc((1+r.length)*s);this.module.setValue(n,r.length,o);for(let a=0;a<r.length;a++)this.module.setValue(n+s*(a+1),r[a],o);return this.module._JsepOutput(this.opKernelContext,e,n)}catch(s){throw new Error(`Failed to generate kernel's output[${e}] with dims [${r}]. If you are running with pre-allocated output, please make sure the output type/dims are correct. Error: ${s}`)}finally{this.module.stackRestore(t)}}},Kg=async(e,r,t,s)=>{let o=r.jsepInit;if(!o)throw new Error("Failed to initialize JSEP. The WebAssembly module is not built with JSEP support.");if(e==="webgpu"){let n=(ET(),uo(Rg)).WebGpuBackend,a=new n;await a.initialize(t,s),o("webgpu",[a,i=>a.alloc(Number(i)),i=>a.free(i),(i,l,c,p=!1)=>{if(p)zt("verbose",()=>`[WebGPU] jsepCopyGpuToGpu: src=${Number(i)}, dst=${Number(l)}, size=${Number(c)}`),a.memcpy(Number(i),Number(l));else{zt("verbose",()=>`[WebGPU] jsepCopyCpuToGpu: dataOffset=${Number(i)}, gpuDataId=${Number(l)}, size=${Number(c)}`);let d=r.HEAPU8.subarray(Number(i>>>0),Number(i>>>0)+Number(c));a.upload(Number(l),d)}},async(i,l,c)=>{zt("verbose",()=>`[WebGPU] jsepCopyGpuToCpu: gpuDataId=${i}, dataOffset=${l}, size=${c}`),await a.download(Number(i),()=>r.HEAPU8.subarray(Number(l)>>>0,Number(l+c)>>>0))},(i,l,c)=>a.createKernel(i,Number(l),c,r.UTF8ToString(r._JsepGetNodeName(Number(l)))),i=>a.releaseKernel(i),(i,l,c,p)=>{zt("verbose",()=>`[WebGPU] jsepRun: sessionHandle=${c}, kernel=${i}, contextDataOffset=${l}`);let d=new Gg(r,a,Number(l));return a.computeKernel(Number(i),d,p)},()=>a.captureBegin(),()=>a.captureEnd(),()=>a.replay()])}else{let n=new Jd(t);o("webnn",[n,()=>n.reserveTensorId(),a=>n.releaseTensorId(a),async(a,i,l,c,p)=>n.ensureTensor(a,i,l,c,p),(a,i)=>{n.uploadTensor(a,i)},async(a,i)=>n.downloadTensor(a,i)])}}}),Hg,rc,sc,rn,qg,nc,ba,oc,ac,ic,lc,cc,uc,Qg=Ue(()=>{Ix(),Ax(),Mt(),_n(),Wi(),Dd(),Hg=(e,r)=>{Zt()._OrtInit(e,r)!==0&&Xt("Can't initialize onnxruntime.")},rc=async e=>{Hg(e.wasm.numThreads,oa(e.logLevel))},sc=async(e,r)=>{var t,s;(s=(t=Zt()).asyncInit)==null||s.call(t);{let o=(PT(),uo(Wg)).init;if(r==="webgpu"){if(typeof navigator>"u"||!navigator.gpu)throw new Error("WebGPU is not supported in current environment");let n=e.webgpu.adapter;if(n){if(typeof n.limits!="object"||typeof n.features!="object"||typeof n.requestDevice!="function")throw new Error("Invalid GPU adapter set in `env.webgpu.adapter`. It must be a GPUAdapter object.")}else{let a=e.webgpu.powerPreference;if(a!==void 0&&a!=="low-power"&&a!=="high-performance")throw new Error(`Invalid powerPreference setting: "${a}"`);let i=e.webgpu.forceFallbackAdapter;if(i!==void 0&&typeof i!="boolean")throw new Error(`Invalid forceFallbackAdapter setting: "${i}"`);if(n=await navigator.gpu.requestAdapter({powerPreference:a,forceFallbackAdapter:i}),!n)throw new Error('Failed to get GPU adapter. You may need to enable flag "--enable-unsafe-webgpu" if you are using Chrome.')}await o("webgpu",Zt(),e,n)}if(r==="webnn"){if(typeof navigator>"u"||!navigator.ml)throw new Error("WebNN is not supported in current environment");await o("webnn",Zt(),e)}}},rn=new Map,qg=e=>{let r=Zt(),t=r.stackSave();try{let s=r.PTR_SIZE,o=r.stackAlloc(2*s);r._OrtGetInputOutputCount(e,o,o+s)!==0&&Xt("Can't get session input/output count.");let n=s===4?"i32":"i64";return[Number(r.getValue(o,n)),Number(r.getValue(o+s,n))]}finally{r.stackRestore(t)}},nc=(e,r)=>{let t=Zt(),s=t.stackSave(),o=0;try{let n=t.PTR_SIZE,a=t.stackAlloc(2*n);t._OrtGetInputOutputMetadata(e,r,a,a+n)!==0&&Xt("Can't get session input/output metadata.");let i=Number(t.getValue(a,"*"));o=Number(t.getValue(a+n,"*"));let l=t.HEAP32[o/4];if(l===0)return[i,0];let c=t.HEAPU32[o/4+1],p=[];for(let d=0;d<c;d++){let u=Number(t.getValue(o+8+d*n,"*"));p.push(u!==0?t.UTF8ToString(u):Number(t.getValue(o+8+(d+c)*n,"*")))}return[i,l,p]}finally{t.stackRestore(s),o!==0&&t._OrtFree(o)}},ba=e=>{let r=Zt(),t=r._malloc(e.byteLength);if(t===0)throw new Error(`Can't create a session. failed to allocate a buffer of size ${e.byteLength}.`);return r.HEAPU8.set(e,t),[t,e.byteLength]},oc=async(e,r)=>{var d,u,f,_;let t,s,o=Zt();Array.isArray(e)?[t,s]=e:e.buffer===o.HEAPU8.buffer?[t,s]=[e.byteOffset,e.byteLength]:[t,s]=ba(e);let n=0,a=0,i=0,l=[],c=[],p=[];try{if([a,l]=await Od(r),(r==null?void 0:r.externalData)&&o.mountExternalData){let C=[];for(let F of r.externalData){let D=typeof F=="string"?F:F.path;C.push(Qi(typeof F=="string"?F:F.data).then(K=>{o.mountExternalData(D,K)}))}await Promise.all(C)}for(let C of(r==null?void 0:r.executionProviders)??[])if((typeof C=="string"?C:C.name)==="webnn"){if(o.shouldTransferToMLTensor=!1,typeof C!="string"){let F=C,D=F==null?void 0:F.context,K=F==null?void 0:F.gpuDevice,U=F==null?void 0:F.deviceType,j=F==null?void 0:F.powerPreference;D?o.currentContext=D:K?o.currentContext=await o.webnnCreateMLContext(K):o.currentContext=await o.webnnCreateMLContext({deviceType:U,powerPreference:j})}else o.currentContext=await o.webnnCreateMLContext();break}n=await o._OrtCreateSession(t,s,a),(d=o.webgpuOnCreateSession)==null||d.call(o,n),n===0&&Xt("Can't create a session."),(u=o.jsepOnCreateSession)==null||u.call(o),o.currentContext&&(o.webnnRegisterMLContext(n,o.currentContext),o.currentContext=void 0,o.shouldTransferToMLTensor=!0);let[b,k]=qg(n),w=!!(r!=null&&r.enableGraphCapture),y=[],$=[],T=[],M=[],x=[];for(let C=0;C<b;C++){let[F,D,K]=nc(n,C);F===0&&Xt("Can't get an input name."),c.push(F);let U=o.UTF8ToString(F);y.push(U),T.push(D===0?{name:U,isTensor:!1}:{name:U,isTensor:!0,type:js(D),shape:K})}for(let C=0;C<k;C++){let[F,D,K]=nc(n,C+b);F===0&&Xt("Can't get an output name."),p.push(F);let U=o.UTF8ToString(F);$.push(U),M.push(D===0?{name:U,isTensor:!1}:{name:U,isTensor:!0,type:js(D),shape:K});{if(w&&(r==null?void 0:r.preferredOutputLocation)===void 0){x.push("gpu-buffer");continue}let j=typeof(r==null?void 0:r.preferredOutputLocation)=="string"?r.preferredOutputLocation:((f=r==null?void 0:r.preferredOutputLocation)==null?void 0:f[U])??"cpu";if(j!=="cpu"&&j!=="cpu-pinned"&&j!=="gpu-buffer"&&j!=="ml-tensor")throw new Error(`Not supported preferred output location: ${j}.`);if(w&&j!=="gpu-buffer")throw new Error(`Not supported preferred output location: ${j}. Only 'gpu-buffer' location is supported when enableGraphCapture is true.`);x.push(j)}}let v=null;return x.some(C=>C==="gpu-buffer"||C==="ml-tensor")&&(i=o._OrtCreateBinding(n),i===0&&Xt("Can't create IO binding."),v={handle:i,outputPreferredLocations:x,outputPreferredLocationsEncoded:x.map(C=>qi(C))}),rn.set(n,[n,c,p,v,w,!1]),[n,y,$,T,M]}catch(b){throw c.forEach(k=>o._OrtFree(k)),p.forEach(k=>o._OrtFree(k)),i!==0&&o._OrtReleaseBinding(i)!==0&&Xt("Can't release IO binding."),n!==0&&o._OrtReleaseSession(n)!==0&&Xt("Can't release session."),b}finally{o._free(t),a!==0&&o._OrtReleaseSessionOptions(a)!==0&&Xt("Can't release session options."),l.forEach(b=>o._free(b)),(_=o.unmountExternalData)==null||_.call(o)}},ac=e=>{var l,c,p;let r=Zt(),t=rn.get(e);if(!t)throw new Error(`cannot release session. invalid session id: ${e}`);let[s,o,n,a,i]=t;a&&(i&&r._OrtClearBoundOutputs(a.handle)!==0&&Xt("Can't clear bound outputs."),r._OrtReleaseBinding(a.handle)!==0&&Xt("Can't release IO binding.")),(l=r.jsepOnReleaseSession)==null||l.call(r,e),(c=r.webnnOnReleaseSession)==null||c.call(r,e),(p=r.webgpuOnReleaseSession)==null||p.call(r,e),o.forEach(d=>r._OrtFree(d)),n.forEach(d=>r._OrtFree(d)),r._OrtReleaseSession(s)!==0&&Xt("Can't release session."),rn.delete(e)},ic=async(e,r,t,s,o,n,a=!1)=>{if(!e){r.push(0);return}let i=Zt(),l=i.PTR_SIZE,c=e[0],p=e[1],d=e[3],u=d,f,_;if(c==="string"&&(d==="gpu-buffer"||d==="ml-tensor"))throw new Error("String tensor is not supported on GPU.");if(a&&d!=="gpu-buffer")throw new Error(`External buffer must be provided for input/output index ${n} when enableGraphCapture is true.`);if(d==="gpu-buffer"){let w=e[2].gpuBuffer;_=fn(Wn(c),p);{let y=i.jsepRegisterBuffer;if(!y)throw new Error('Tensor location "gpu-buffer" is not supported without using WebGPU.');f=y(s,n,w,_)}}else if(d==="ml-tensor"){let w=e[2].mlTensor;_=fn(Wn(c),p);let y=i.webnnRegisterMLTensor;if(!y)throw new Error('Tensor location "ml-tensor" is not supported without using WebNN.');f=y(s,w,Wn(c),p)}else{let w=e[2];if(Array.isArray(w)){_=l*w.length,f=i._malloc(_),t.push(f);for(let y=0;y<w.length;y++){if(typeof w[y]!="string")throw new TypeError(`tensor data at index ${y} is not a string`);i.setValue(f+y*l,gs(w[y],t),"*")}}else{let y=i.webnnIsGraphInput;if(c!=="string"&&y){let $=i.UTF8ToString(o);if(y(s,$)){let T=Wn(c);_=fn(T,p),u="ml-tensor";let M=i.webnnCreateTemporaryTensor,x=i.webnnUploadTensor;if(!M||!x)throw new Error('Tensor location "ml-tensor" is not supported without using WebNN.');let v=await M(s,T,p);x(v,new Uint8Array(w.buffer,w.byteOffset,w.byteLength)),f=v}else _=w.byteLength,f=i._malloc(_),t.push(f),i.HEAPU8.set(new Uint8Array(w.buffer,w.byteOffset,_),f)}else _=w.byteLength,f=i._malloc(_),t.push(f),i.HEAPU8.set(new Uint8Array(w.buffer,w.byteOffset,_),f)}}let b=i.stackSave(),k=i.stackAlloc(4*p.length);try{p.forEach((y,$)=>i.setValue(k+$*l,y,l===4?"i32":"i64"));let w=i._OrtCreateTensor(Wn(c),f,_,k,p.length,qi(u));w===0&&Xt(`Can't create tensor for input/output. session=${s}, index=${n}.`),r.push(w)}finally{i.stackRestore(b)}},lc=async(e,r,t,s,o,n)=>{var K,U,j,se;let a=Zt(),i=a.PTR_SIZE,l=rn.get(e);if(!l)throw new Error(`cannot run inference. invalid session id: ${e}`);let c=l[0],p=l[1],d=l[2],u=l[3],f=l[4],_=l[5],b=r.length,k=s.length,w=0,y=[],$=[],T=[],M=[],x=a.stackSave(),v=a.stackAlloc(b*i),C=a.stackAlloc(b*i),F=a.stackAlloc(k*i),D=a.stackAlloc(k*i);try{[w,y]=$d(n);for(let Z=0;Z<b;Z++)await ic(t[Z],$,M,e,p[r[Z]],r[Z],f);for(let Z=0;Z<k;Z++)await ic(o[Z],T,M,e,d[s[Z]],b+s[Z],f);for(let Z=0;Z<b;Z++)a.setValue(v+Z*i,$[Z],"*"),a.setValue(C+Z*i,p[r[Z]],"*");for(let Z=0;Z<k;Z++)a.setValue(F+Z*i,T[Z],"*"),a.setValue(D+Z*i,d[s[Z]],"*");if(u&&!_){let{handle:Z,outputPreferredLocations:ae,outputPreferredLocationsEncoded:_e}=u;if(p.length!==b)throw new Error(`input count from feeds (${b}) is expected to be always equal to model's input count (${p.length}).`);for(let pe=0;pe<b;pe++){let oe=r[pe];await a._OrtBindInput(Z,p[oe],$[pe])!==0&&Xt(`Can't bind input[${pe}] for session=${e}.`)}for(let pe=0;pe<k;pe++){let oe=s[pe];(K=o[pe])!=null&&K[3]?a._OrtBindOutput(Z,d[oe],T[pe],0)!==0&&Xt(`Can't bind pre-allocated output[${pe}] for session=${e}.`):a._OrtBindOutput(Z,d[oe],0,_e[oe])!==0&&Xt(`Can't bind output[${pe}] to ${ae[pe]} for session=${e}.`)}rn.set(e,[c,p,d,u,f,!0])}(U=a.jsepOnRunStart)==null||U.call(a,c),(j=a.webnnOnRunStart)==null||j.call(a,c);let q;u?q=await a._OrtRunWithBinding(c,u.handle,k,F,w):q=await a._OrtRun(c,C,v,b,D,k,F,w),q!==0&&Xt("failed to call OrtRun().");let re=[];for(let Z=0;Z<k;Z++){let ae=Number(a.getValue(F+Z*i,"*"));if(ae===T[Z]){re.push(o[Z]);continue}let _e=a.stackSave(),pe=a.stackAlloc(4*i),oe=!1,G,B=0;try{a._OrtGetTensorData(ae,pe,pe+i,pe+2*i,pe+3*i)!==0&&Xt(`Can't access output tensor data on index ${Z}.`);let Q=i===4?"i32":"i64",le=Number(a.getValue(pe,Q));B=a.getValue(pe+i,"*");let ee=a.getValue(pe+i*2,"*"),Se=Number(a.getValue(pe+i*3,Q)),De=[];for(let te=0;te<Se;te++)De.push(Number(a.getValue(ee+te*i,Q)));a._OrtFree(ee)!==0&&Xt("Can't free memory for tensor dims.");let Ne=De.reduce((te,O)=>te*O,1);G=js(le);let Ae=u==null?void 0:u.outputPreferredLocations[s[Z]];if(G==="string"){if(Ae==="gpu-buffer"||Ae==="ml-tensor")throw new Error("String tensor is not supported on GPU.");let te=[];for(let O=0;O<Ne;O++){let me=a.getValue(B+O*i,"*"),xe=a.getValue(B+(O+1)*i,"*"),be=O===Ne-1?void 0:xe-me;te.push(a.UTF8ToString(me,be))}re.push([G,De,te,"cpu"])}else if(Ae==="gpu-buffer"&&Ne>0){let te=a.jsepGetBuffer;if(!te)throw new Error('preferredLocation "gpu-buffer" is not supported without using WebGPU.');let O=te(B),me=fn(le,Ne);if(me===void 0||!Ki(G))throw new Error(`Unsupported data type: ${G}`);oe=!0,re.push([G,De,{gpuBuffer:O,download:a.jsepCreateDownloader(O,me,G),dispose:()=>{a._OrtReleaseTensor(ae)!==0&&Xt("Can't release tensor.")}},"gpu-buffer"])}else if(Ae==="ml-tensor"&&Ne>0){let te=a.webnnEnsureTensor,O=a.webnnIsInt64Supported;if(!te||!O)throw new Error('preferredLocation "ml-tensor" is not supported without using WebNN.');if(fn(le,Ne)===void 0||!Hi(G))throw new Error(`Unsupported data type: ${G}`);if(G==="int64"&&!O(e))throw new Error('preferredLocation "ml-tensor" for int64 output is not supported by current WebNN Context.');let me=await te(e,B,le,De,!1);oe=!0,re.push([G,De,{mlTensor:me,download:a.webnnCreateMLTensorDownloader(B,G),dispose:()=>{a.webnnReleaseTensorId(B),a._OrtReleaseTensor(ae)}},"ml-tensor"])}else{let te=Gi(G),O=new te(Ne);new Uint8Array(O.buffer,O.byteOffset,O.byteLength).set(a.HEAPU8.subarray(B,B+O.byteLength)),re.push([G,De,O,"cpu"])}}finally{a.stackRestore(_e),G==="string"&&B&&a._free(B),oe||a._OrtReleaseTensor(ae),(se=a.webnnOnRunEnd)==null||se.call(a,c)}}return u&&!f&&(a._OrtClearBoundOutputs(u.handle)!==0&&Xt("Can't clear bound outputs."),rn.set(e,[c,p,d,u,f,!1])),re}finally{a.stackRestore(x),$.forEach(q=>a._OrtReleaseTensor(q)),T.forEach(q=>a._OrtReleaseTensor(q)),M.forEach(q=>a._free(q)),w!==0&&a._OrtReleaseRunOptions(w),y.forEach(q=>a._free(q))}},cc=e=>{let r=Zt(),t=rn.get(e);if(!t)throw new Error("invalid session id");let s=t[0],o=r._OrtEndProfiling(s);o===0&&Xt("Can't get an profile file name."),r._OrtFree(o)},uc=e=>{let r=[];for(let t of e){let s=t[2];!Array.isArray(s)&&"buffer"in s&&r.push(s.buffer)}return r}}),sn,Xr,qn,Po,Co,ya,dc,va,xn,Tn,Xg,Jg,Yg,Zg,ew,tw,rw,sw,nw=Ue(()=>{fs(),Qg(),_n(),ji(),sn=()=>!!tr.wasm.proxy&&typeof document<"u",qn=!1,Po=!1,Co=!1,va=new Map,xn=(e,r)=>{let t=va.get(e);t?t.push(r):va.set(e,[r])},Tn=()=>{if(qn||!Po||Co||!Xr)throw new Error("worker not ready")},Xg=e=>{switch(e.data.type){case"init-wasm":qn=!1,e.data.err?(Co=!0,dc[1](e.data.err)):(Po=!0,dc[0]()),ya&&(URL.revokeObjectURL(ya),ya=void 0);break;case"init-ep":case"copy-from":case"create":case"release":case"run":case"end-profiling":{let r=va.get(e.data.type);e.data.err?r.shift()[1](e.data.err):r.shift()[0](e.data.out);break}}},Jg=async()=>{if(!Po){if(qn)throw new Error("multiple calls to 'initWasm()' detected.");if(Co)throw new Error("previous call to 'initWasm()' failed.");if(qn=!0,sn())return new Promise((e,r)=>{Xr==null||Xr.terminate(),Td().then(([t,s])=>{try{Xr=s,Xr.onerror=n=>r(n),Xr.onmessage=Xg,dc=[e,r];let o={type:"init-wasm",in:tr};!o.in.wasm.wasmPaths&&(t||Li)&&(o.in.wasm.wasmPaths={wasm:new URL("/assets/ort-wasm-simd-threaded.jsep-B0T3yYHD.wasm",self.location.href).href}),Xr.postMessage(o),ya=t}catch(o){r(o)}},r)});try{await Ui(tr.wasm),await rc(tr),Po=!0}catch(e){throw Co=!0,e}finally{qn=!1}}},Yg=async e=>{if(sn())return Tn(),new Promise((r,t)=>{xn("init-ep",[r,t]);let s={type:"init-ep",in:{epName:e,env:tr}};Xr.postMessage(s)});await sc(tr,e)},Zg=async e=>sn()?(Tn(),new Promise((r,t)=>{xn("copy-from",[r,t]);let s={type:"copy-from",in:{buffer:e}};Xr.postMessage(s,[e.buffer])})):ba(e),ew=async(e,r)=>{if(sn()){if(r!=null&&r.preferredOutputLocation)throw new Error('session option "preferredOutputLocation" is not supported for proxy.');return Tn(),new Promise((t,s)=>{xn("create",[t,s]);let o={type:"create",in:{model:e,options:{...r}}},n=[];e instanceof Uint8Array&&n.push(e.buffer),Xr.postMessage(o,n)})}else return oc(e,r)},tw=async e=>{if(sn())return Tn(),new Promise((r,t)=>{xn("release",[r,t]);let s={type:"release",in:e};Xr.postMessage(s)});ac(e)},rw=async(e,r,t,s,o,n)=>{if(sn()){if(t.some(a=>a[3]!=="cpu"))throw new Error("input tensor on GPU is not supported for proxy.");if(o.some(a=>a))throw new Error("pre-allocated output tensor is not supported for proxy.");return Tn(),new Promise((a,i)=>{xn("run",[a,i]);let l=t,c={type:"run",in:{sessionId:e,inputIndices:r,inputs:l,outputIndices:s,options:n}};Xr.postMessage(c,uc(l))})}else return lc(e,r,t,s,o,n)},sw=async e=>{if(sn())return Tn(),new Promise((r,t)=>{xn("end-profiling",[r,t]);let s={type:"end-profiling",in:e};Xr.postMessage(s)});cc(e)}}),pc,ow,aw,CT=Ue(()=>{fs(),nw(),Mt(),ki(),Dd(),pc=(e,r)=>{switch(e.location){case"cpu":return[e.type,e.dims,e.data,"cpu"];case"gpu-buffer":return[e.type,e.dims,{gpuBuffer:e.gpuBuffer},"gpu-buffer"];case"ml-tensor":return[e.type,e.dims,{mlTensor:e.mlTensor},"ml-tensor"];default:throw new Error(`invalid data location: ${e.location} for ${r()}`)}},ow=e=>{switch(e[3]){case"cpu":return new hs(e[0],e[2],e[1]);case"gpu-buffer":{let r=e[0];if(!Ki(r))throw new Error(`not supported data type: ${r} for deserializing GPU tensor`);let{gpuBuffer:t,download:s,dispose:o}=e[2];return hs.fromGpuBuffer(t,{dataType:r,dims:e[1],download:s,dispose:o})}case"ml-tensor":{let r=e[0];if(!Hi(r))throw new Error(`not supported data type: ${r} for deserializing MLTensor tensor`);let{mlTensor:t,download:s,dispose:o}=e[2];return hs.fromMLTensor(t,{dataType:r,dims:e[1],download:s,dispose:o})}default:throw new Error(`invalid data location: ${e[3]}`)}},aw=class{async fetchModelAndCopyToWasmMemory(e){return Zg(await Qi(e))}async loadModel(e,r){_s();let t;typeof e=="string"?t=await this.fetchModelAndCopyToWasmMemory(e):t=e,[this.sessionId,this.inputNames,this.outputNames,this.inputMetadata,this.outputMetadata]=await ew(t,r),is()}async dispose(){return tw(this.sessionId)}async run(e,r,t){_s();let s=[],o=[];Object.entries(e).forEach(d=>{let u=d[0],f=d[1],_=this.inputNames.indexOf(u);if(_===-1)throw new Error(`invalid input '${u}'`);s.push(f),o.push(_)});let n=[],a=[];Object.entries(r).forEach(d=>{let u=d[0],f=d[1],_=this.outputNames.indexOf(u);if(_===-1)throw new Error(`invalid output '${u}'`);n.push(f),a.push(_)});let i=s.map((d,u)=>pc(d,()=>`input "${this.inputNames[o[u]]}"`)),l=n.map((d,u)=>d?pc(d,()=>`output "${this.outputNames[a[u]]}"`):null),c=await rw(this.sessionId,o,i,a,l,t),p={};for(let d=0;d<c.length;d++)p[this.outputNames[a[d]]]=n[d]??ow(c[d]);return is(),p}startProfiling(){}endProfiling(){sw(this.sessionId)}}}),iw={};Un(iw,{OnnxruntimeWebAssemblyBackend:()=>hc,initializeFlags:()=>mc,wasmBackend:()=>lw});var mc,hc,lw,ST=Ue(()=>{fs(),nw(),CT(),mc=()=>{(typeof tr.wasm.initTimeout!="number"||tr.wasm.initTimeout<0)&&(tr.wasm.initTimeout=0);let e=tr.wasm.simd;if(typeof e!="boolean"&&e!==void 0&&e!=="fixed"&&e!=="relaxed"&&(console.warn(`Property "env.wasm.simd" is set to unknown value "${e}". Reset it to \`false\` and ignore SIMD feature checking.`),tr.wasm.simd=!1),typeof tr.wasm.proxy!="boolean"&&(tr.wasm.proxy=!1),typeof tr.wasm.trace!="boolean"&&(tr.wasm.trace=!1),typeof tr.wasm.numThreads!="number"||!Number.isInteger(tr.wasm.numThreads)||tr.wasm.numThreads<=0)if(typeof self<"u"&&!self.crossOriginIsolated)tr.wasm.numThreads=1;else{let r=typeof navigator>"u"?hx("node:os").cpus().length:navigator.hardwareConcurrency;tr.wasm.numThreads=Math.min(4,Math.ceil((r||1)/2))}},hc=class{async init(e){mc(),await Jg(),await Yg(e)}async createInferenceSessionHandler(e,r){let t=new aw;return await t.loadModel(e,r),t}},lw=new hc});fs(),fs(),fs();var $T="1.22.0-dev.20250409-89f8206ba4",kT=md;{let e=(ST(),uo(iw)).wasmBackend;mn("webgpu",e,5),mn("webnn",e,5),mn("cpu",e,10),mn("wasm",e,10)}Object.defineProperty(tr.versions,"web",{value:$T,enumerable:!0});/**
* @license
* Copyright 2021 Google LLC. All Rights Reserved.
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
* =============================================================================
*//**
 * @license
 * Copyright 2020 Google LLC. All Rights Reserved.
 * Licensed under the Apache License, Version 2.0 (the "License");
 * you may not use this file except in compliance with the License.
 * You may obtain a copy of the License at
 *
 * http://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS,
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 * See the License for the specific language governing permissions and
 * limitations under the License.
 * =============================================================================
 *//**
 * @license
 * Copyright 2019 Google LLC. All Rights Reserved.
 * Licensed under the Apache License, Version 2.0 (the "License");
 * you may not use this file except in compliance with the License.
 * You may obtain a copy of the License at
 *
 * http://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS,
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 * See the License for the specific language governing permissions and
 * limitations under the License.
 * =============================================================================
 */var IT=Object.freeze({__proto__:null,get InferenceSession(){return $i},get TRACE(){return ho},get TRACE_FUNC_BEGIN(){return _s},get TRACE_FUNC_END(){return is},get Tensor(){return hs},default:kT,get env(){return tr},get registerBackend(){return mn}}),Us={},AT={"onnxruntime-common":e=>{e.exports=ux},"onnxruntime-web":e=>{e.exports=IT},"?2ce3":()=>{},"?7992":()=>{},"?5af5":()=>{},"?2b25":()=>{},"?db59":()=>{},"?383f":()=>{},"?fa4b":()=>{},"./node_modules/@huggingface/jinja/dist/index.js":(e,r,t)=>{t.r(r),t.d(r,{Environment:()=>pt,Interpreter:()=>Mr,Template:()=>Fs,parse:()=>Se,tokenize:()=>d});var s=Object.freeze({Text:"Text",NumericLiteral:"NumericLiteral",StringLiteral:"StringLiteral",Identifier:"Identifier",Equals:"Equals",OpenParen:"OpenParen",CloseParen:"CloseParen",OpenStatement:"OpenStatement",CloseStatement:"CloseStatement",OpenExpression:"OpenExpression",CloseExpression:"CloseExpression",OpenSquareBracket:"OpenSquareBracket",CloseSquareBracket:"CloseSquareBracket",OpenCurlyBracket:"OpenCurlyBracket",CloseCurlyBracket:"CloseCurlyBracket",Comma:"Comma",Dot:"Dot",Colon:"Colon",Pipe:"Pipe",CallOperator:"CallOperator",AdditiveBinaryOperator:"AdditiveBinaryOperator",MultiplicativeBinaryOperator:"MultiplicativeBinaryOperator",ComparisonBinaryOperator:"ComparisonBinaryOperator",UnaryOperator:"UnaryOperator",Comment:"Comment"}),o=class{constructor(S,H){this.value=S,this.type=H}};function n(S){return/\w/.test(S)}function a(S){return/[0-9]/.test(S)}function i(S){return/\s/.test(S)}var l=[["{%",s.OpenStatement],["%}",s.CloseStatement],["{{",s.OpenExpression],["}}",s.CloseExpression],["(",s.OpenParen],[")",s.CloseParen],["{",s.OpenCurlyBracket],["}",s.CloseCurlyBracket],["[",s.OpenSquareBracket],["]",s.CloseSquareBracket],[",",s.Comma],[".",s.Dot],[":",s.Colon],["|",s.Pipe],["<=",s.ComparisonBinaryOperator],[">=",s.ComparisonBinaryOperator],["==",s.ComparisonBinaryOperator],["!=",s.ComparisonBinaryOperator],["<",s.ComparisonBinaryOperator],[">",s.ComparisonBinaryOperator],["+",s.AdditiveBinaryOperator],["-",s.AdditiveBinaryOperator],["~",s.AdditiveBinaryOperator],["*",s.MultiplicativeBinaryOperator],["/",s.MultiplicativeBinaryOperator],["%",s.MultiplicativeBinaryOperator],["=",s.Equals]],c=new Map([["n",`
`],["t","	"],["r","\r"],["b","\b"],["f","\f"],["v","\v"],["'","'"],['"','"'],["\\","\\"]]);function p(S,H={}){return S.endsWith(`
`)&&(S=S.slice(0,-1)),H.lstrip_blocks&&(S=S.replace(/^[ \t]*({[#%-])/gm,"$1")),H.trim_blocks&&(S=S.replace(/([#%-]})\n/g,"$1")),S.replace(/{%\s*(end)?generation\s*%}/gs,"")}function d(S,H={}){var rt,St;const R=[],W=p(S,H);let X=0,ue=0;const Ee=lt=>{let Ot="";for(;lt(W[X]);){if(W[X]==="\\"){if(++X,X>=W.length)throw new SyntaxError("Unexpected end of input");const at=W[X++],Dt=c.get(at);if(Dt===void 0)throw new SyntaxError(`Unexpected escaped character: ${at}`);Ot+=Dt;continue}if(Ot+=W[X++],X>=W.length)throw new SyntaxError("Unexpected end of input")}return Ot},et=()=>{const lt=R.at(-1);lt&&lt.type===s.Text&&(lt.value=lt.value.trimEnd(),lt.value===""&&R.pop())},ot=()=>{for(;X<W.length&&i(W[X]);)++X};e:for(;X<W.length;){const lt=(rt=R.at(-1))==null?void 0:rt.type;if(lt===void 0||lt===s.CloseStatement||lt===s.CloseExpression||lt===s.Comment){let at="";for(;X<W.length&&!(W[X]==="{"&&(W[X+1]==="%"||W[X+1]==="{"||W[X+1]==="#"));)at+=W[X++];if(at.length>0){R.push(new o(at,s.Text));continue}}if(W[X]==="{"&&W[X+1]==="#"){X+=2;const at=W[X]==="-";at&&++X;let Dt="";for(;W[X]!=="#"||W[X+1]!=="}";){if(X+2>=W.length)throw new SyntaxError("Missing end of comment tag");Dt+=W[X++]}const Wt=Dt.endsWith("-");Wt&&(Dt=Dt.slice(0,-1)),at&&et(),R.push(new o(Dt,s.Comment)),X+=2,Wt&&ot();continue}if(W.slice(X,X+3)==="{%-"){et(),R.push(new o("{%",s.OpenStatement)),X+=3;continue}if(W.slice(X,X+3)==="{{-"){et(),R.push(new o("{{",s.OpenExpression)),ue=0,X+=3;continue}if(Ee(i),W.slice(X,X+3)==="-%}"){R.push(new o("%}",s.CloseStatement)),X+=3,ot();continue}if(W.slice(X,X+3)==="-}}"){R.push(new o("}}",s.CloseExpression)),X+=3,ot();continue}const Ot=W[X];if(Ot==="-"||Ot==="+"){const at=(St=R.at(-1))==null?void 0:St.type;if(at===s.Text||at===void 0)throw new SyntaxError(`Unexpected character: ${Ot}`);switch(at){case s.Identifier:case s.NumericLiteral:case s.StringLiteral:case s.CloseParen:case s.CloseSquareBracket:break;default:{++X;const Dt=Ee(a);R.push(new o(`${Ot}${Dt}`,Dt.length>0?s.NumericLiteral:s.UnaryOperator));continue}}}for(const[at,Dt]of l){if(at==="}}"&&ue>0)continue;if(W.slice(X,X+at.length)===at){R.push(new o(at,Dt)),Dt===s.OpenExpression?ue=0:Dt===s.OpenCurlyBracket?++ue:Dt===s.CloseCurlyBracket&&--ue,X+=at.length;continue e}}if(Ot==="'"||Ot==='"'){++X;const at=Ee(Dt=>Dt!==Ot);R.push(new o(at,s.StringLiteral)),++X;continue}if(a(Ot)){let at=Ee(a);if(W[X]==="."&&a(W[X+1])){++X;const Dt=Ee(a);at=`${at}.${Dt}`}R.push(new o(at,s.NumericLiteral));continue}if(n(Ot)){const at=Ee(n);R.push(new o(at,s.Identifier));continue}throw new SyntaxError(`Unexpected character: ${Ot}`)}return R}var u=class{constructor(){J(this,"type","Statement")}},f=class extends u{constructor(H){super();J(this,"type","Program");this.body=H}},_=class extends u{constructor(H,R,W){super();J(this,"type","If");this.test=H,this.body=R,this.alternate=W}},b=class extends u{constructor(H,R,W,X){super();J(this,"type","For");this.loopvar=H,this.iterable=R,this.body=W,this.defaultBlock=X}},k=class extends u{constructor(){super(...arguments);J(this,"type","Break")}},w=class extends u{constructor(){super(...arguments);J(this,"type","Continue")}},y=class extends u{constructor(H,R,W){super();J(this,"type","Set");this.assignee=H,this.value=R,this.body=W}},$=class extends u{constructor(H,R,W){super();J(this,"type","Macro");this.name=H,this.args=R,this.body=W}},T=class extends u{constructor(H){super();J(this,"type","Comment");this.value=H}},M=class extends u{constructor(){super(...arguments);J(this,"type","Expression")}},x=class extends M{constructor(H,R,W){super();J(this,"type","MemberExpression");this.object=H,this.property=R,this.computed=W}},v=class extends M{constructor(H,R){super();J(this,"type","CallExpression");this.callee=H,this.args=R}},C=class extends M{constructor(H){super();J(this,"type","Identifier");this.value=H}},F=class extends M{constructor(H){super();J(this,"type","Literal");this.value=H}},D=class extends F{constructor(){super(...arguments);J(this,"type","IntegerLiteral")}},K=class extends F{constructor(){super(...arguments);J(this,"type","FloatLiteral")}},U=class extends F{constructor(){super(...arguments);J(this,"type","StringLiteral")}},j=class extends F{constructor(){super(...arguments);J(this,"type","ArrayLiteral")}},se=class extends F{constructor(){super(...arguments);J(this,"type","TupleLiteral")}},q=class extends F{constructor(){super(...arguments);J(this,"type","ObjectLiteral")}},re=class extends M{constructor(H,R,W){super();J(this,"type","BinaryExpression");this.operator=H,this.left=R,this.right=W}},Z=class extends M{constructor(H,R){super();J(this,"type","FilterExpression");this.operand=H,this.filter=R}},ae=class extends u{constructor(H,R){super();J(this,"type","FilterStatement");this.filter=H,this.body=R}},_e=class extends M{constructor(H,R){super();J(this,"type","SelectExpression");this.lhs=H,this.test=R}},pe=class extends M{constructor(H,R,W){super();J(this,"type","TestExpression");this.operand=H,this.negate=R,this.test=W}},oe=class extends M{constructor(H,R){super();J(this,"type","UnaryExpression");this.operator=H,this.argument=R}},G=class extends M{constructor(H=void 0,R=void 0,W=void 0){super();J(this,"type","SliceExpression");this.start=H,this.stop=R,this.step=W}},B=class extends M{constructor(H,R){super();J(this,"type","KeywordArgumentExpression");this.key=H,this.value=R}},Q=class extends M{constructor(H){super();J(this,"type","SpreadExpression");this.argument=H}},le=class extends u{constructor(H,R,W){super();J(this,"type","CallStatement");this.call=H,this.callerArgs=R,this.body=W}},ee=class extends M{constructor(H,R,W){super();J(this,"type","Ternary");this.condition=H,this.trueExpr=R,this.falseExpr=W}};function Se(S){const H=new f([]);let R=0;function W(je,Ve){const st=S[R++];if(!st||st.type!==je)throw new Error(`Parser Error: ${Ve}. ${st.type} !== ${je}.`);return st}function X(je){if(!ot(je))throw new SyntaxError(`Expected ${je}`);++R}function ue(){switch(S[R].type){case s.Comment:return new T(S[R++].value);case s.Text:return rt();case s.OpenStatement:return St();case s.OpenExpression:return lt();default:throw new SyntaxError(`Unexpected token type: ${S[R].type}`)}}function Ee(...je){return R+je.length<=S.length&&je.every((Ve,st)=>Ve===S[R+st].type)}function et(...je){var Ve,st,qt;return((Ve=S[R])==null?void 0:Ve.type)===s.OpenStatement&&((st=S[R+1])==null?void 0:st.type)===s.Identifier&&je.includes((qt=S[R+1])==null?void 0:qt.value)}function ot(...je){return R+je.length<=S.length&&je.every((Ve,st)=>S[R+st].type==="Identifier"&&Ve===S[R+st].value)}function rt(){return new U(W(s.Text,"Expected text token").value)}function St(){if(W(s.OpenStatement,"Expected opening statement token"),S[R].type!==s.Identifier)throw new SyntaxError(`Unknown statement, got ${S[R].type}`);const je=S[R].value;let Ve;switch(je){case"set":++R,Ve=Ot();break;case"if":++R,Ve=at(),W(s.OpenStatement,"Expected {% token"),X("endif"),W(s.CloseStatement,"Expected %} token");break;case"macro":++R,Ve=Dt(),W(s.OpenStatement,"Expected {% token"),X("endmacro"),W(s.CloseStatement,"Expected %} token");break;case"for":++R,Ve=Br(),W(s.OpenStatement,"Expected {% token"),X("endfor"),W(s.CloseStatement,"Expected %} token");break;case"call":{++R;let st=null;Ee(s.OpenParen)&&(st=ur());const qt=Wr();if(qt.type!=="Identifier")throw new SyntaxError("Expected identifier following call statement");const qs=ur();W(s.CloseStatement,"Expected closing statement token");const Bs=[];for(;!et("endcall");)Bs.push(ue());W(s.OpenStatement,"Expected '{%'"),X("endcall"),W(s.CloseStatement,"Expected closing statement token");const Qs=new v(qt,qs);Ve=new le(Qs,st,Bs);break}case"break":++R,W(s.CloseStatement,"Expected closing statement token"),Ve=new k;break;case"continue":++R,W(s.CloseStatement,"Expected closing statement token"),Ve=new w;break;case"filter":{++R;let st=Wr();st instanceof C&&Ee(s.OpenParen)&&(st=Hr(st)),W(s.CloseStatement,"Expected closing statement token");const qt=[];for(;!et("endfilter");)qt.push(ue());W(s.OpenStatement,"Expected '{%'"),X("endfilter"),W(s.CloseStatement,"Expected '%}'"),Ve=new ae(st,qt);break}default:throw new SyntaxError(`Unknown statement type: ${je}`)}return Ve}function lt(){W(s.OpenExpression,"Expected opening expression token");const je=br();return W(s.CloseExpression,"Expected closing expression token"),je}function Ot(){const je=Wt();let Ve=null;const st=[];if(Ee(s.Equals))++R,Ve=Wt();else{for(W(s.CloseStatement,"Expected %} token");!et("endset");)st.push(ue());W(s.OpenStatement,"Expected {% token"),X("endset")}return W(s.CloseStatement,"Expected closing statement token"),new y(je,Ve,st)}function at(){const je=br();W(s.CloseStatement,"Expected closing statement token");const Ve=[],st=[];for(;!et("elif","else","endif");)Ve.push(ue());if(et("elif")){++R,++R;const qt=at();st.push(qt)}else if(et("else"))for(++R,++R,W(s.CloseStatement,"Expected closing statement token");!et("endif");)st.push(ue());return new _(je,Ve,st)}function Dt(){const je=Wr();if(je.type!=="Identifier")throw new SyntaxError("Expected identifier following macro statement");const Ve=ur();W(s.CloseStatement,"Expected closing statement token");const st=[];for(;!et("endmacro");)st.push(ue());return new $(je,Ve,st)}function Wt(je=!1){const Ve=je?Wr:br,st=[Ve()],qt=Ee(s.Comma);for(;qt&&(++R,st.push(Ve()),!!Ee(s.Comma)););return qt?new se(st):st[0]}function Br(){const je=Wt(!0);if(!(je instanceof C||je instanceof se))throw new SyntaxError(`Expected identifier/tuple for the loop variable, got ${je.type} instead`);if(!ot("in"))throw new SyntaxError("Expected `in` keyword following loop variable");++R;const Ve=br();W(s.CloseStatement,"Expected closing statement token");const st=[];for(;!et("endfor","else");)st.push(ue());const qt=[];if(et("else"))for(++R,++R,W(s.CloseStatement,"Expected closing statement token");!et("endfor");)qt.push(ue());return new b(je,Ve,st,qt)}function br(){return Rr()}function Rr(){const je=Ts();if(ot("if")){++R;const Ve=Ts();if(ot("else")){++R;const st=Rr();return new ee(Ve,je,st)}else return new _e(je,Ve)}return je}function Ts(){let je=Os();for(;ot("or");){const Ve=S[R];++R;const st=Os();je=new re(Ve,je,st)}return je}function Os(){let je=gr();for(;ot("and");){const Ve=S[R];++R;const st=gr();je=new re(Ve,je,st)}return je}function gr(){let je;for(;ot("not");){const Ve=S[R];++R;const st=gr();je=new oe(Ve,st)}return je??Hs()}function Hs(){let je=cs();for(;;){let Ve;if(ot("not","in"))Ve=new o("not in",s.Identifier),R+=2;else if(ot("in"))Ve=S[R++];else if(Ee(s.ComparisonBinaryOperator))Ve=S[R++];else break;const st=cs();je=new re(Ve,je,st)}return je}function cs(){let je=Ls();for(;Ee(s.AdditiveBinaryOperator);){const Ve=S[R];++R;const st=Ls();je=new re(Ve,je,st)}return je}function jr(){const je=Ds(Wr());return Ee(s.OpenParen)?Hr(je):je}function Hr(je){let Ve=new v(je,ur());return Ve=Ds(Ve),Ee(s.OpenParen)&&(Ve=Hr(Ve)),Ve}function ur(){W(s.OpenParen,"Expected opening parenthesis for arguments list");const je=mr();return W(s.CloseParen,"Expected closing parenthesis for arguments list"),je}function mr(){const je=[];for(;!Ee(s.CloseParen);){let Ve;if(S[R].type===s.MultiplicativeBinaryOperator&&S[R].value==="*"){++R;const st=br();Ve=new Q(st)}else if(Ve=br(),Ee(s.Equals)){if(++R,!(Ve instanceof C))throw new SyntaxError("Expected identifier for keyword argument");const st=br();Ve=new B(Ve,st)}je.push(Ve),Ee(s.Comma)&&++R}return je}function Es(){const je=[];let Ve=!1;for(;!Ee(s.CloseSquareBracket);)Ee(s.Colon)?(je.push(void 0),++R,Ve=!0):(je.push(br()),Ee(s.Colon)&&(++R,Ve=!0));if(je.length===0)throw new SyntaxError("Expected at least one argument for member/slice expression");if(Ve){if(je.length>3)throw new SyntaxError("Expected 0-3 arguments for slice expression");return new G(...je)}return je[0]}function Ds(je){for(;Ee(s.Dot)||Ee(s.OpenSquareBracket);){const Ve=S[R];++R;let st;const qt=Ve.type===s.OpenSquareBracket;if(qt)st=Es(),W(s.CloseSquareBracket,"Expected closing square bracket");else if(st=Wr(),st.type!=="Identifier")throw new SyntaxError("Expected identifier following dot operator");je=new x(je,st,qt)}return je}function Ls(){let je=zs();for(;Ee(s.MultiplicativeBinaryOperator);){const Ve=S[R++],st=zs();je=new re(Ve,je,st)}return je}function zs(){let je=us();for(;ot("is");){++R;const Ve=ot("not");Ve&&++R;const st=Wr();if(!(st instanceof C))throw new SyntaxError("Expected identifier for the test");je=new pe(je,Ve,st)}return je}function us(){let je=jr();for(;Ee(s.Pipe);){++R;let Ve=Wr();if(!(Ve instanceof C))throw new SyntaxError("Expected identifier for the filter");Ee(s.OpenParen)&&(Ve=Hr(Ve)),je=new Z(je,Ve)}return je}function Wr(){const je=S[R++];switch(je.type){case s.NumericLiteral:{const Ve=je.value;return Ve.includes(".")?new K(Number(Ve)):new D(Number(Ve))}case s.StringLiteral:{let Ve=je.value;for(;Ee(s.StringLiteral);)Ve+=S[R++].value;return new U(Ve)}case s.Identifier:return new C(je.value);case s.OpenParen:{const Ve=Wt();return W(s.CloseParen,"Expected closing parenthesis, got ${tokens[current].type} instead."),Ve}case s.OpenSquareBracket:{const Ve=[];for(;!Ee(s.CloseSquareBracket);)Ve.push(br()),Ee(s.Comma)&&++R;return++R,new j(Ve)}case s.OpenCurlyBracket:{const Ve=new Map;for(;!Ee(s.CloseCurlyBracket);){const st=br();W(s.Colon,"Expected colon between key and value in object literal");const qt=br();Ve.set(st,qt),Ee(s.Comma)&&++R}return++R,new q(Ve)}default:throw new SyntaxError(`Unexpected token: ${je.type}`)}}for(;R<S.length;)H.body.push(ue());return H}function De(S,H,R=1){H===void 0&&(H=S,S=0);const W=[];for(let X=S;X<H;X+=R)W.push(X);return W}function Ne(S,H,R,W=1){const X=Math.sign(W);X>=0?(H=(H??(H=0))<0?Math.max(S.length+H,0):Math.min(H,S.length),R=(R??(R=S.length))<0?Math.max(S.length+R,0):Math.min(R,S.length)):(H=(H??(H=S.length-1))<0?Math.max(S.length+H,-1):Math.min(H,S.length-1),R=(R??(R=-1))<-1?Math.max(S.length+R,-1):Math.min(R,S.length-1));const ue=[];for(let Ee=H;X*Ee<X*R;Ee+=W)ue.push(S[Ee]);return ue}function Ae(S){return S.replace(/\b\w/g,H=>H.toUpperCase())}function te(S){return O(new Date,S)}function O(S,H){const R=new Intl.DateTimeFormat(void 0,{month:"long"}),W=new Intl.DateTimeFormat(void 0,{month:"short"}),X=ue=>ue<10?"0"+ue:ue.toString();return H.replace(/%[YmdbBHM%]/g,ue=>{switch(ue){case"%Y":return S.getFullYear().toString();case"%m":return X(S.getMonth()+1);case"%d":return X(S.getDate());case"%b":return W.format(S);case"%B":return R.format(S);case"%H":return X(S.getHours());case"%M":return X(S.getMinutes());case"%%":return"%";default:return ue}})}function me(S){return S.replace(/[.*+?^${}()|[\]\\]/g,"\\$&")}function xe(S,H,R,W){if(W===0)return S;let X=W==null||W<0?1/0:W;const ue=H.length===0?new RegExp("(?=)","gu"):new RegExp(me(H),"gu");return S.replaceAll(ue,Ee=>X>0?(--X,R):Ee)}var be=class extends Error{},Ce=class extends Error{},$e=class{constructor(S=void 0){J(this,"type","RuntimeValue");J(this,"value");J(this,"builtins",new Map);this.value=S}__bool__(){return new ke(!!this.value)}toString(){return String(this.value)}},Me=class extends $e{constructor(){super(...arguments);J(this,"type","IntegerValue")}},Le=class extends $e{constructor(){super(...arguments);J(this,"type","FloatValue")}toString(){return this.value%1===0?this.value.toFixed(1):this.value.toString()}},we=class extends $e{constructor(){super(...arguments);J(this,"type","StringValue");J(this,"builtins",new Map([["upper",new Qe(()=>new we(this.value.toUpperCase()))],["lower",new Qe(()=>new we(this.value.toLowerCase()))],["strip",new Qe(()=>new we(this.value.trim()))],["title",new Qe(()=>new we(Ae(this.value)))],["capitalize",new Qe(()=>new we(this.value.charAt(0).toUpperCase()+this.value.slice(1)))],["length",new Me(this.value.length)],["rstrip",new Qe(()=>new we(this.value.trimEnd()))],["lstrip",new Qe(()=>new we(this.value.trimStart()))],["startswith",new Qe(H=>{if(H.length===0)throw new Error("startswith() requires at least one argument");const R=H[0];if(R instanceof we)return new ke(this.value.startsWith(R.value));if(R instanceof We){for(const W of R.value){if(!(W instanceof we))throw new Error("startswith() tuple elements must be strings");if(this.value.startsWith(W.value))return new ke(!0)}return new ke(!1)}throw new Error("startswith() argument must be a string or tuple of strings")})],["endswith",new Qe(H=>{if(H.length===0)throw new Error("endswith() requires at least one argument");const R=H[0];if(R instanceof we)return new ke(this.value.endsWith(R.value));if(R instanceof We){for(const W of R.value){if(!(W instanceof we))throw new Error("endswith() tuple elements must be strings");if(this.value.endsWith(W.value))return new ke(!0)}return new ke(!1)}throw new Error("endswith() argument must be a string or tuple of strings")})],["split",new Qe(H=>{const R=H[0]??new Oe;if(!(R instanceof we||R instanceof Oe))throw new Error("sep argument must be a string or null");const W=H[1]??new Me(-1);if(!(W instanceof Me))throw new Error("maxsplit argument must be a number");let X=[];if(R instanceof Oe){const ue=this.value.trimStart();for(const{0:Ee,index:et}of ue.matchAll(/\S+/g)){if(W.value!==-1&&X.length>=W.value&&et!==void 0){X.push(Ee+ue.slice(et+Ee.length));break}X.push(Ee)}}else{if(R.value==="")throw new Error("empty separator");X=this.value.split(R.value),W.value!==-1&&X.length>W.value&&X.push(X.splice(W.value).join(R.value))}return new We(X.map(ue=>new we(ue)))})],["replace",new Qe(H=>{if(H.length<2)throw new Error("replace() requires at least two arguments");const R=H[0],W=H[1];if(!(R instanceof we&&W instanceof we))throw new Error("replace() arguments must be strings");let X;if(H.length>2?H[2].type==="KeywordArgumentsValue"?X=H[2].value.get("count")??new Oe:X=H[2]:X=new Oe,!(X instanceof Me||X instanceof Oe))throw new Error("replace() count argument must be a number or null");return new we(xe(this.value,R.value,W.value,X.value))})]]))}},ke=class extends $e{constructor(){super(...arguments);J(this,"type","BooleanValue")}};function He(S,H,R,W=!0){const X=R??0;switch(S.type){case"NullValue":return"null";case"UndefinedValue":return W?"null":"undefined";case"IntegerValue":case"FloatValue":case"StringValue":case"BooleanValue":return JSON.stringify(S.value);case"ArrayValue":case"ObjectValue":{const ue=H?" ".repeat(H):"",Ee=`
`+ue.repeat(X),et=Ee+ue;if(S.type==="ArrayValue"){const ot=S.value.map(rt=>He(rt,H,X+1,W));return H?`[${et}${ot.join(`,${et}`)}${Ee}]`:`[${ot.join(", ")}]`}else{const ot=Array.from(S.value.entries()).map(([rt,St])=>{const lt=`"${rt}": ${He(St,H,X+1,W)}`;return H?`${et}${lt}`:lt});return H?`{${ot.join(",")}${Ee}}`:`{${ot.join(", ")}}`}}default:throw new Error(`Cannot convert to JSON: ${S.type}`)}}var Fe=class extends $e{constructor(){super(...arguments);J(this,"type","ObjectValue");J(this,"builtins",new Map([["get",new Qe(([H,R])=>{if(!(H instanceof we))throw new Error(`Object key must be a string: got ${H.type}`);return this.value.get(H.value)??R??new Oe})],["items",new Qe(()=>this.items())],["keys",new Qe(()=>this.keys())],["values",new Qe(()=>this.values())],["dictsort",new Qe(H=>{let R=new Map;const W=H.filter(ot=>ot instanceof Ze?(R=ot.value,!1):!0),X=W.at(0)??R.get("case_sensitive")??new ke(!1);if(!(X instanceof ke))throw new Error("case_sensitive must be a boolean");const ue=W.at(1)??R.get("by")??new we("key");if(!(ue instanceof we))throw new Error("by must be a string");if(!["key","value"].includes(ue.value))throw new Error("by must be either 'key' or 'value'");const Ee=W.at(2)??R.get("reverse")??new ke(!1);if(!(Ee instanceof ke))throw new Error("reverse must be a boolean");const et=Array.from(this.value.entries()).map(([ot,rt])=>new We([new we(ot),rt])).sort((ot,rt)=>{const St=ue.value==="key"?0:1,lt=ot.value[St],Ot=rt.value[St],at=kt(lt,Ot,X.value);return Ee.value?-at:at});return new We(et)})]]))}__bool__(){return new ke(this.value.size>0)}items(){return new We(Array.from(this.value.entries()).map(([H,R])=>new We([new we(H),R])))}keys(){return new We(Array.from(this.value.keys()).map(H=>new we(H)))}values(){return new We(Array.from(this.value.values()))}toString(){return He(this,null,0,!1)}},Ze=class extends Fe{constructor(){super(...arguments);J(this,"type","KeywordArgumentsValue")}},We=class extends $e{constructor(){super(...arguments);J(this,"type","ArrayValue");J(this,"builtins",new Map([["length",new Me(this.value.length)]]))}__bool__(){return new ke(this.value.length>0)}toString(){return He(this,null,0,!1)}},yt=class extends We{constructor(){super(...arguments);J(this,"type","TupleValue")}},Qe=class extends $e{constructor(){super(...arguments);J(this,"type","FunctionValue")}},Oe=class extends $e{constructor(){super(...arguments);J(this,"type","NullValue")}},ze=class extends $e{constructor(){super(...arguments);J(this,"type","UndefinedValue")}},pt=class{constructor(S){J(this,"variables",new Map([["namespace",new Qe(S=>{if(S.length===0)return new Fe(new Map);if(S.length!==1||!(S[0]instanceof Fe))throw new Error("`namespace` expects either zero arguments or a single object argument");return S[0]})]]));J(this,"tests",new Map([["boolean",S=>S.type==="BooleanValue"],["callable",S=>S instanceof Qe],["odd",S=>{if(!(S instanceof Me))throw new Error(`cannot odd on ${S.type}`);return S.value%2!==0}],["even",S=>{if(!(S instanceof Me))throw new Error(`cannot even on ${S.type}`);return S.value%2===0}],["false",S=>S.type==="BooleanValue"&&!S.value],["true",S=>S.type==="BooleanValue"&&S.value],["none",S=>S.type==="NullValue"],["string",S=>S.type==="StringValue"],["number",S=>S instanceof Me||S instanceof Le],["integer",S=>S instanceof Me],["iterable",S=>S.type==="ArrayValue"||S.type==="StringValue"],["mapping",S=>S.type==="ObjectValue"],["lower",S=>{const H=S.value;return S.type==="StringValue"&&H===H.toLowerCase()}],["upper",S=>{const H=S.value;return S.type==="StringValue"&&H===H.toUpperCase()}],["none",S=>S.type==="NullValue"],["defined",S=>S.type!=="UndefinedValue"],["undefined",S=>S.type==="UndefinedValue"],["equalto",(S,H)=>S.value===H.value],["eq",(S,H)=>S.value===H.value]]));this.parent=S}set(S,H){return this.declareVariable(S,fr(H))}declareVariable(S,H){if(this.variables.has(S))throw new SyntaxError(`Variable already declared: ${S}`);return this.variables.set(S,H),H}setVariable(S,H){return this.variables.set(S,H),H}resolve(S){if(this.variables.has(S))return this;if(this.parent)return this.parent.resolve(S);throw new Error(`Unknown variable: ${S}`)}lookupVariable(S){try{return this.resolve(S).variables.get(S)??new ze}catch{return new ze}}};function ft(S){S.set("false",!1),S.set("true",!0),S.set("none",null),S.set("raise_exception",H=>{throw new Error(H)}),S.set("range",De),S.set("strftime_now",te),S.set("True",!0),S.set("False",!1),S.set("None",null)}function Lt(S,H){const R=H.split(".");let W=S;for(const X of R)if(W instanceof Fe)W=W.value.get(X)??new ze;else if(W instanceof We){const ue=parseInt(X,10);if(!isNaN(ue)&&ue>=0&&ue<W.value.length)W=W.value[ue];else return new ze}else return new ze;return W}function kt(S,H,R=!1){if(S instanceof Oe&&H instanceof Oe)return 0;if(S instanceof Oe||H instanceof Oe)throw new Error(`Cannot compare ${S.type} with ${H.type}`);if(S instanceof ze&&H instanceof ze)return 0;if(S instanceof ze||H instanceof ze)throw new Error(`Cannot compare ${S.type} with ${H.type}`);const W=ue=>ue instanceof Me||ue instanceof Le||ue instanceof ke,X=ue=>ue instanceof ke?ue.value?1:0:ue.value;if(W(S)&&W(H)){const ue=X(S),Ee=X(H);return ue<Ee?-1:ue>Ee?1:0}if(S.type!==H.type)throw new Error(`Cannot compare different types: ${S.type} and ${H.type}`);switch(S.type){case"StringValue":{let ue=S.value,Ee=H.value;return R||(ue=ue.toLowerCase(),Ee=Ee.toLowerCase()),ue<Ee?-1:ue>Ee?1:0}default:throw new Error(`Cannot compare type: ${S.type}`)}}var Mr=class{constructor(S){J(this,"global");this.global=S??new pt}run(S){return this.evaluate(S,this.global)}evaluateBinaryExpression(S,H){const R=this.evaluate(S.left,H);switch(S.operator.value){case"and":return R.__bool__().value?this.evaluate(S.right,H):R;case"or":return R.__bool__().value?R:this.evaluate(S.right,H)}const W=this.evaluate(S.right,H);switch(S.operator.value){case"==":return new ke(R.value==W.value);case"!=":return new ke(R.value!=W.value)}if(R instanceof ze||W instanceof ze){if(W instanceof ze&&["in","not in"].includes(S.operator.value))return new ke(S.operator.value==="not in");throw new Error(`Cannot perform operation ${S.operator.value} on undefined values`)}else{if(R instanceof Oe||W instanceof Oe)throw new Error("Cannot perform operation on null values");if(S.operator.value==="~")return new we(R.value.toString()+W.value.toString());if((R instanceof Me||R instanceof Le)&&(W instanceof Me||W instanceof Le)){const X=R.value,ue=W.value;switch(S.operator.value){case"+":case"-":case"*":{const Ee=S.operator.value==="+"?X+ue:S.operator.value==="-"?X-ue:X*ue;return R instanceof Le||W instanceof Le?new Le(Ee):new Me(Ee)}case"/":return new Le(X/ue);case"%":{const Ee=X%ue;return R instanceof Le||W instanceof Le?new Le(Ee):new Me(Ee)}case"<":return new ke(X<ue);case">":return new ke(X>ue);case">=":return new ke(X>=ue);case"<=":return new ke(X<=ue)}}else if(R instanceof We&&W instanceof We)switch(S.operator.value){case"+":return new We(R.value.concat(W.value))}else if(W instanceof We){const X=W.value.find(ue=>ue.value===R.value)!==void 0;switch(S.operator.value){case"in":return new ke(X);case"not in":return new ke(!X)}}}if(R instanceof we||W instanceof we)switch(S.operator.value){case"+":return new we(R.value.toString()+W.value.toString())}if(R instanceof we&&W instanceof we)switch(S.operator.value){case"in":return new ke(W.value.includes(R.value));case"not in":return new ke(!W.value.includes(R.value))}if(R instanceof we&&W instanceof Fe)switch(S.operator.value){case"in":return new ke(W.value.has(R.value));case"not in":return new ke(!W.value.has(R.value))}throw new SyntaxError(`Unknown operator "${S.operator.value}" between ${R.type} and ${W.type}`)}evaluateArguments(S,H){const R=[],W=new Map;for(const X of S)if(X.type==="SpreadExpression"){const ue=X,Ee=this.evaluate(ue.argument,H);if(!(Ee instanceof We))throw new Error(`Cannot unpack non-iterable type: ${Ee.type}`);for(const et of Ee.value)R.push(et)}else if(X.type==="KeywordArgumentExpression"){const ue=X;W.set(ue.key.value,this.evaluate(ue.value,H))}else{if(W.size>0)throw new Error("Positional arguments must come before keyword arguments");R.push(this.evaluate(X,H))}return[R,W]}applyFilter(S,H,R){if(H.type==="Identifier"){const W=H;if(W.value==="tojson")return new we(He(S));if(S instanceof We)switch(W.value){case"list":return S;case"first":return S.value[0];case"last":return S.value[S.value.length-1];case"length":return new Me(S.value.length);case"reverse":return new We(S.value.slice().reverse());case"sort":return new We(S.value.slice().sort((X,ue)=>kt(X,ue,!1)));case"join":return new we(S.value.map(X=>X.value).join(""));case"string":return new we(He(S,null,0,!1));case"unique":{const X=new Set,ue=[];for(const Ee of S.value)X.has(Ee.value)||(X.add(Ee.value),ue.push(Ee));return new We(ue)}default:throw new Error(`Unknown ArrayValue filter: ${W.value}`)}else if(S instanceof we)switch(W.value){case"length":case"upper":case"lower":case"title":case"capitalize":{const X=S.builtins.get(W.value);if(X instanceof Qe)return X.value([],R);if(X instanceof Me)return X;throw new Error(`Unknown StringValue filter: ${W.value}`)}case"trim":return new we(S.value.trim());case"indent":return new we(S.value.split(`
`).map((X,ue)=>ue===0||X.length===0?X:"    "+X).join(`
`));case"join":case"string":return S;case"int":{const X=parseInt(S.value,10);return new Me(isNaN(X)?0:X)}case"float":{const X=parseFloat(S.value);return new Le(isNaN(X)?0:X)}default:throw new Error(`Unknown StringValue filter: ${W.value}`)}else if(S instanceof Me||S instanceof Le)switch(W.value){case"abs":return S instanceof Me?new Me(Math.abs(S.value)):new Le(Math.abs(S.value));case"int":return new Me(Math.floor(S.value));case"float":return new Le(S.value);default:throw new Error(`Unknown NumericValue filter: ${W.value}`)}else if(S instanceof Fe)switch(W.value){case"items":return new We(Array.from(S.value.entries()).map(([X,ue])=>new We([new we(X),ue])));case"length":return new Me(S.value.size);default:{const X=S.builtins.get(W.value);if(X)return X instanceof Qe?X.value([],R):X;throw new Error(`Unknown ObjectValue filter: ${W.value}`)}}else if(S instanceof ke)switch(W.value){case"bool":return new ke(S.value);case"int":return new Me(S.value?1:0);case"float":return new Le(S.value?1:0);case"string":return new we(S.value?"true":"false");default:throw new Error(`Unknown BooleanValue filter: ${W.value}`)}throw new Error(`Cannot apply filter "${W.value}" to type: ${S.type}`)}else if(H.type==="CallExpression"){const W=H;if(W.callee.type!=="Identifier")throw new Error(`Unknown filter: ${W.callee.type}`);const X=W.callee.value;if(X==="tojson"){const[,ue]=this.evaluateArguments(W.args,R),Ee=ue.get("indent")??new Oe;if(!(Ee instanceof Me||Ee instanceof Oe))throw new Error("If set, indent must be a number");return new we(He(S,Ee.value))}else if(X==="join"){let ue;if(S instanceof we)ue=Array.from(S.value);else if(S instanceof We)ue=S.value.map(rt=>rt.value);else throw new Error(`Cannot apply filter "${X}" to type: ${S.type}`);const[Ee,et]=this.evaluateArguments(W.args,R),ot=Ee.at(0)??et.get("separator")??new we("");if(!(ot instanceof we))throw new Error("separator must be a string");return new we(ue.join(ot.value))}else if(X==="int"||X==="float"){const[ue,Ee]=this.evaluateArguments(W.args,R),et=ue.at(0)??Ee.get("default")??(X==="int"?new Me(0):new Le(0));if(S instanceof we){const ot=X==="int"?parseInt(S.value,10):parseFloat(S.value);return isNaN(ot)?et:X==="int"?new Me(ot):new Le(ot)}else{if(S instanceof Me||S instanceof Le)return S;if(S instanceof ke)return X==="int"?new Me(S.value?1:0):new Le(S.value?1:0);throw new Error(`Cannot apply filter "${X}" to type: ${S.type}`)}}else if(X==="default"){const[ue,Ee]=this.evaluateArguments(W.args,R),et=ue[0]??new we(""),ot=ue[1]??Ee.get("boolean")??new ke(!1);if(!(ot instanceof ke))throw new Error("`default` filter flag must be a boolean");return S instanceof ze||ot.value&&!S.__bool__().value?et:S}if(S instanceof We){switch(X){case"sort":{const[ue,Ee]=this.evaluateArguments(W.args,R),et=ue.at(0)??Ee.get("reverse")??new ke(!1);if(!(et instanceof ke))throw new Error("reverse must be a boolean");const ot=ue.at(1)??Ee.get("case_sensitive")??new ke(!1);if(!(ot instanceof ke))throw new Error("case_sensitive must be a boolean");const rt=ue.at(2)??Ee.get("attribute")??new Oe;if(!(rt instanceof we||rt instanceof Me||rt instanceof Oe))throw new Error("attribute must be a string, integer, or null");const St=lt=>{if(rt instanceof Oe)return lt;const Ot=rt instanceof Me?String(rt.value):rt.value;return Lt(lt,Ot)};return new We(S.value.slice().sort((lt,Ot)=>{const at=St(lt),Dt=St(Ot),Wt=kt(at,Dt,ot.value);return et.value?-Wt:Wt}))}case"selectattr":case"rejectattr":{const ue=X==="selectattr";if(S.value.some(lt=>!(lt instanceof Fe)))throw new Error(`\`${X}\` can only be applied to array of objects`);if(W.args.some(lt=>lt.type!=="StringLiteral"))throw new Error(`arguments of \`${X}\` must be strings`);const[Ee,et,ot]=W.args.map(lt=>this.evaluate(lt,R));let rt;if(et){const lt=R.tests.get(et.value);if(!lt)throw new Error(`Unknown test: ${et.value}`);rt=lt}else rt=(...lt)=>lt[0].__bool__().value;const St=S.value.filter(lt=>{const Ot=lt.value.get(Ee.value),at=Ot?rt(Ot,ot):!1;return ue?at:!at});return new We(St)}case"map":{const[,ue]=this.evaluateArguments(W.args,R);if(ue.has("attribute")){const Ee=ue.get("attribute");if(!(Ee instanceof we))throw new Error("attribute must be a string");const et=ue.get("default"),ot=S.value.map(rt=>{if(!(rt instanceof Fe))throw new Error("items in map must be an object");const St=Lt(rt,Ee.value);return St instanceof ze?et??new ze:St});return new We(ot)}else throw new Error("`map` expressions without `attribute` set are not currently supported.")}}throw new Error(`Unknown ArrayValue filter: ${X}`)}else if(S instanceof we){switch(X){case"indent":{const[ue,Ee]=this.evaluateArguments(W.args,R),et=ue.at(0)??Ee.get("width")??new Me(4);if(!(et instanceof Me))throw new Error("width must be a number");const ot=ue.at(1)??Ee.get("first")??new ke(!1),rt=ue.at(2)??Ee.get("blank")??new ke(!1),St=S.value.split(`
`),lt=" ".repeat(et.value),Ot=St.map((at,Dt)=>!ot.value&&Dt===0||!rt.value&&at.length===0?at:lt+at);return new we(Ot.join(`
`))}case"replace":{const ue=S.builtins.get("replace");if(!(ue instanceof Qe))throw new Error("replace filter not available");const[Ee,et]=this.evaluateArguments(W.args,R);return ue.value([...Ee,new Ze(et)],R)}}throw new Error(`Unknown StringValue filter: ${X}`)}else if(S instanceof Fe){const ue=S.builtins.get(X);if(ue&&ue instanceof Qe){const[Ee,et]=this.evaluateArguments(W.args,R);return et.size>0&&Ee.push(new Ze(et)),ue.value(Ee,R)}throw new Error(`Unknown ObjectValue filter: ${X}`)}else throw new Error(`Cannot apply filter "${X}" to type: ${S.type}`)}throw new Error(`Unknown filter: ${H.type}`)}evaluateFilterExpression(S,H){const R=this.evaluate(S.operand,H);return this.applyFilter(R,S.filter,H)}evaluateTestExpression(S,H){const R=this.evaluate(S.operand,H),W=H.tests.get(S.test.value);if(!W)throw new Error(`Unknown test: ${S.test.value}`);const X=W(R);return new ke(S.negate?!X:X)}evaluateSelectExpression(S,H){return this.evaluate(S.test,H).__bool__().value?this.evaluate(S.lhs,H):new ze}evaluateUnaryExpression(S,H){const R=this.evaluate(S.argument,H);switch(S.operator.value){case"not":return new ke(!R.value);default:throw new SyntaxError(`Unknown operator: ${S.operator.value}`)}}evaluateTernaryExpression(S,H){return this.evaluate(S.condition,H).__bool__().value?this.evaluate(S.trueExpr,H):this.evaluate(S.falseExpr,H)}evalProgram(S,H){return this.evaluateBlock(S.body,H)}evaluateBlock(S,H){let R="";for(const W of S){const X=this.evaluate(W,H);X.type!=="NullValue"&&X.type!=="UndefinedValue"&&(R+=X.toString())}return new we(R)}evaluateIdentifier(S,H){return H.lookupVariable(S.value)}evaluateCallExpression(S,H){const[R,W]=this.evaluateArguments(S.args,H);W.size>0&&R.push(new Ze(W));const X=this.evaluate(S.callee,H);if(X.type!=="FunctionValue")throw new Error(`Cannot call something that is not a function: got ${X.type}`);return X.value(R,H)}evaluateSliceExpression(S,H,R){if(!(S instanceof We||S instanceof we))throw new Error("Slice object must be an array or string");const W=this.evaluate(H.start,R),X=this.evaluate(H.stop,R),ue=this.evaluate(H.step,R);if(!(W instanceof Me||W instanceof ze))throw new Error("Slice start must be numeric or undefined");if(!(X instanceof Me||X instanceof ze))throw new Error("Slice stop must be numeric or undefined");if(!(ue instanceof Me||ue instanceof ze))throw new Error("Slice step must be numeric or undefined");return S instanceof We?new We(Ne(S.value,W.value,X.value,ue.value)):new we(Ne(Array.from(S.value),W.value,X.value,ue.value).join(""))}evaluateMemberExpression(S,H){const R=this.evaluate(S.object,H);let W;if(S.computed){if(S.property.type==="SliceExpression")return this.evaluateSliceExpression(R,S.property,H);W=this.evaluate(S.property,H)}else W=new we(S.property.value);let X;if(R instanceof Fe){if(!(W instanceof we))throw new Error(`Cannot access property with non-string: got ${W.type}`);X=R.value.get(W.value)??R.builtins.get(W.value)}else if(R instanceof We||R instanceof we)if(W instanceof Me)X=R.value.at(W.value),R instanceof we&&(X=new we(R.value.at(W.value)));else if(W instanceof we)X=R.builtins.get(W.value);else throw new Error(`Cannot access property with non-string/non-number: got ${W.type}`);else{if(!(W instanceof we))throw new Error(`Cannot access property with non-string: got ${W.type}`);X=R.builtins.get(W.value)}return X instanceof $e?X:new ze}evaluateSet(S,H){const R=S.value?this.evaluate(S.value,H):this.evaluateBlock(S.body,H);if(S.assignee.type==="Identifier"){const W=S.assignee.value;H.setVariable(W,R)}else if(S.assignee.type==="TupleLiteral"){const W=S.assignee;if(!(R instanceof We))throw new Error(`Cannot unpack non-iterable type in set: ${R.type}`);const X=R.value;if(X.length!==W.value.length)throw new Error(`Too ${W.value.length>X.length?"few":"many"} items to unpack in set`);for(let ue=0;ue<W.value.length;++ue){const Ee=W.value[ue];if(Ee.type!=="Identifier")throw new Error(`Cannot unpack to non-identifier in set: ${Ee.type}`);H.setVariable(Ee.value,X[ue])}}else if(S.assignee.type==="MemberExpression"){const W=S.assignee,X=this.evaluate(W.object,H);if(!(X instanceof Fe))throw new Error("Cannot assign to member of non-object");if(W.property.type!=="Identifier")throw new Error("Cannot assign to member with non-identifier property");X.value.set(W.property.value,R)}else throw new Error(`Invalid LHS inside assignment expression: ${JSON.stringify(S.assignee)}`);return new Oe}evaluateIf(S,H){const R=this.evaluate(S.test,H);return this.evaluateBlock(R.__bool__().value?S.body:S.alternate,H)}evaluateFor(S,H){const R=new pt(H);let W,X;if(S.iterable.type==="SelectExpression"){const rt=S.iterable;X=this.evaluate(rt.lhs,R),W=rt.test}else X=this.evaluate(S.iterable,R);if(!(X instanceof We||X instanceof Fe))throw new Error(`Expected iterable or object type in for loop: got ${X.type}`);X instanceof Fe&&(X=X.keys());const ue=[],Ee=[];for(let rt=0;rt<X.value.length;++rt){const St=new pt(R),lt=X.value[rt];let Ot;if(S.loopvar.type==="Identifier")Ot=at=>at.setVariable(S.loopvar.value,lt);else if(S.loopvar.type==="TupleLiteral"){const at=S.loopvar;if(lt.type!=="ArrayValue")throw new Error(`Cannot unpack non-iterable type: ${lt.type}`);const Dt=lt;if(at.value.length!==Dt.value.length)throw new Error(`Too ${at.value.length>Dt.value.length?"few":"many"} items to unpack`);Ot=Wt=>{for(let Br=0;Br<at.value.length;++Br){if(at.value[Br].type!=="Identifier")throw new Error(`Cannot unpack non-identifier type: ${at.value[Br].type}`);Wt.setVariable(at.value[Br].value,Dt.value[Br])}}}else throw new Error(`Invalid loop variable(s): ${S.loopvar.type}`);W&&(Ot(St),!this.evaluate(W,St).__bool__().value)||(ue.push(lt),Ee.push(Ot))}let et="",ot=!0;for(let rt=0;rt<ue.length;++rt){const St=new Map([["index",new Me(rt+1)],["index0",new Me(rt)],["revindex",new Me(ue.length-rt)],["revindex0",new Me(ue.length-rt-1)],["first",new ke(rt===0)],["last",new ke(rt===ue.length-1)],["length",new Me(ue.length)],["previtem",rt>0?ue[rt-1]:new ze],["nextitem",rt<ue.length-1?ue[rt+1]:new ze]]);R.setVariable("loop",new Fe(St)),Ee[rt](R);try{const lt=this.evaluateBlock(S.body,R);et+=lt.value}catch(lt){if(lt instanceof Ce)continue;if(lt instanceof be)break;throw lt}ot=!1}if(ot){const rt=this.evaluateBlock(S.defaultBlock,R);et+=rt.value}return new we(et)}evaluateMacro(S,H){return H.setVariable(S.name.value,new Qe((R,W)=>{var Ee;const X=new pt(W);R=R.slice();let ue;((Ee=R.at(-1))==null?void 0:Ee.type)==="KeywordArgumentsValue"&&(ue=R.pop());for(let et=0;et<S.args.length;++et){const ot=S.args[et],rt=R[et];if(ot.type==="Identifier"){const St=ot;if(!rt)throw new Error(`Missing positional argument: ${St.value}`);X.setVariable(St.value,rt)}else if(ot.type==="KeywordArgumentExpression"){const St=ot,lt=rt??(ue==null?void 0:ue.value.get(St.key.value))??this.evaluate(St.value,X);X.setVariable(St.key.value,lt)}else throw new Error(`Unknown argument type: ${ot.type}`)}return this.evaluateBlock(S.body,X)})),new Oe}evaluateCallStatement(S,H){const R=new Qe((et,ot)=>{const rt=new pt(ot);if(S.callerArgs)for(let St=0;St<S.callerArgs.length;++St){const lt=S.callerArgs[St];if(lt.type!=="Identifier")throw new Error(`Caller parameter must be an identifier, got ${lt.type}`);rt.setVariable(lt.value,et[St]??new ze)}return this.evaluateBlock(S.body,rt)}),[W,X]=this.evaluateArguments(S.call.args,H);W.push(new Ze(X));const ue=this.evaluate(S.call.callee,H);if(ue.type!=="FunctionValue")throw new Error(`Cannot call something that is not a function: got ${ue.type}`);const Ee=new pt(H);return Ee.setVariable("caller",R),ue.value(W,Ee)}evaluateFilterStatement(S,H){const R=this.evaluateBlock(S.body,H);return this.applyFilter(R,S.filter,H)}evaluate(S,H){if(!S)return new ze;switch(S.type){case"Program":return this.evalProgram(S,H);case"Set":return this.evaluateSet(S,H);case"If":return this.evaluateIf(S,H);case"For":return this.evaluateFor(S,H);case"Macro":return this.evaluateMacro(S,H);case"CallStatement":return this.evaluateCallStatement(S,H);case"Break":throw new be;case"Continue":throw new Ce;case"IntegerLiteral":return new Me(S.value);case"FloatLiteral":return new Le(S.value);case"StringLiteral":return new we(S.value);case"ArrayLiteral":return new We(S.value.map(R=>this.evaluate(R,H)));case"TupleLiteral":return new yt(S.value.map(R=>this.evaluate(R,H)));case"ObjectLiteral":{const R=new Map;for(const[W,X]of S.value){const ue=this.evaluate(W,H);if(!(ue instanceof we))throw new Error(`Object keys must be strings: got ${ue.type}`);R.set(ue.value,this.evaluate(X,H))}return new Fe(R)}case"Identifier":return this.evaluateIdentifier(S,H);case"CallExpression":return this.evaluateCallExpression(S,H);case"MemberExpression":return this.evaluateMemberExpression(S,H);case"UnaryExpression":return this.evaluateUnaryExpression(S,H);case"BinaryExpression":return this.evaluateBinaryExpression(S,H);case"FilterExpression":return this.evaluateFilterExpression(S,H);case"FilterStatement":return this.evaluateFilterStatement(S,H);case"TestExpression":return this.evaluateTestExpression(S,H);case"SelectExpression":return this.evaluateSelectExpression(S,H);case"Ternary":return this.evaluateTernaryExpression(S,H);case"Comment":return new Oe;default:throw new SyntaxError(`Unknown node type: ${S.type}`)}}};function fr(S){switch(typeof S){case"number":return Number.isInteger(S)?new Me(S):new Le(S);case"string":return new we(S);case"boolean":return new ke(S);case"undefined":return new ze;case"object":return S===null?new Oe:Array.isArray(S)?new We(S.map(fr)):new Fe(new Map(Object.entries(S).map(([H,R])=>[H,fr(R)])));case"function":return new Qe((H,R)=>{const W=S(...H.map(X=>X.value))??null;return fr(W)});default:throw new Error(`Cannot convert to runtime value: ${S}`)}}var Ut=`
`,Ur="{%- ",Ws=" -%}";function Gs(S){switch(S.operator.type){case"MultiplicativeBinaryOperator":return 4;case"AdditiveBinaryOperator":return 3;case"ComparisonBinaryOperator":return 2;case"Identifier":return S.operator.value==="and"?1:S.operator.value==="in"||S.operator.value==="not in"?2:0}return 0}function nn(S,H="	"){const R=typeof H=="number"?" ".repeat(H):H;return zr(S.body,0,R).replace(/\n$/,"")}function cr(...S){return Ur+S.join(" ")+Ws}function zr(S,H,R){return S.map(W=>ls(W,H,R)).join(Ut)}function ls(S,H,R){const W=R.repeat(H);switch(S.type){case"Program":return zr(S.body,H,R);case"If":return Ks(S,H,R);case"For":return gt(S,H,R);case"Set":return Is(S,H,R);case"Macro":return As(S,H,R);case"Break":return W+cr("break");case"Continue":return W+cr("continue");case"CallStatement":return Yt(S,H,R);case"FilterStatement":return Jr(S,H,R);case"Comment":return W+"{# "+S.value+" #}";default:return W+"{{- "+$t(S)+" -}}"}}function Ks(S,H,R){const W=R.repeat(H),X=[];let ue=S;for(;ue&&(X.push({test:ue.test,body:ue.body}),ue.alternate.length===1&&ue.alternate[0].type==="If");)ue=ue.alternate[0];let Ee=W+cr("if",$t(X[0].test))+Ut+zr(X[0].body,H+1,R);for(let et=1;et<X.length;++et)Ee+=Ut+W+cr("elif",$t(X[et].test))+Ut+zr(X[et].body,H+1,R);return ue&&ue.alternate.length>0&&(Ee+=Ut+W+cr("else")+Ut+zr(ue.alternate,H+1,R)),Ee+=Ut+W+cr("endif"),Ee}function gt(S,H,R){const W=R.repeat(H);let X="";if(S.iterable.type==="SelectExpression"){const Ee=S.iterable;X=`${$t(Ee.lhs)} if ${$t(Ee.test)}`}else X=$t(S.iterable);let ue=W+cr("for",$t(S.loopvar),"in",X)+Ut+zr(S.body,H+1,R);return S.defaultBlock.length>0&&(ue+=Ut+W+cr("else")+Ut+zr(S.defaultBlock,H+1,R)),ue+=Ut+W+cr("endfor"),ue}function Is(S,H,R){const W=R.repeat(H),X=$t(S.assignee),ue=S.value?$t(S.value):"",Ee=W+cr("set",`${X}${S.value?" = "+ue:""}`);return S.body.length===0?Ee:Ee+Ut+zr(S.body,H+1,R)+Ut+W+cr("endset")}function As(S,H,R){const W=R.repeat(H),X=S.args.map($t).join(", ");return W+cr("macro",`${S.name.value}(${X})`)+Ut+zr(S.body,H+1,R)+Ut+W+cr("endmacro")}function Yt(S,H,R){const W=R.repeat(H),X=S.callerArgs&&S.callerArgs.length>0?`(${S.callerArgs.map($t).join(", ")})`:"",ue=$t(S.call);let Ee=W+cr(`call${X}`,ue)+Ut;return Ee+=zr(S.body,H+1,R)+Ut,Ee+=W+cr("endcall"),Ee}function Jr(S,H,R){const W=R.repeat(H),X=S.filter.type==="Identifier"?S.filter.value:$t(S.filter);let ue=W+cr("filter",X)+Ut;return ue+=zr(S.body,H+1,R)+Ut,ue+=W+cr("endfilter"),ue}function $t(S,H=-1){switch(S.type){case"SpreadExpression":return`*${$t(S.argument)}`;case"Identifier":return S.value;case"IntegerLiteral":return`${S.value}`;case"FloatLiteral":return`${S.value}`;case"StringLiteral":return JSON.stringify(S.value);case"BinaryExpression":{const R=S,W=Gs(R),X=$t(R.left,W),ue=$t(R.right,W+1),Ee=`${X} ${R.operator.value} ${ue}`;return W<H?`(${Ee})`:Ee}case"UnaryExpression":{const R=S;return R.operator.value+(R.operator.value==="not"?" ":"")+$t(R.argument,1/0)}case"CallExpression":{const R=S,W=R.args.map($t).join(", ");return`${$t(R.callee)}(${W})`}case"MemberExpression":{const R=S;let W=$t(R.object);["Identifier","MemberExpression","CallExpression","StringLiteral","IntegerLiteral","FloatLiteral","ArrayLiteral","TupleLiteral","ObjectLiteral"].includes(R.object.type)||(W=`(${W})`);let X=$t(R.property);return!R.computed&&R.property.type!=="Identifier"&&(X=`(${X})`),R.computed?`${W}[${X}]`:`${W}.${X}`}case"FilterExpression":{const R=S,W=$t(R.operand,1/0);return R.filter.type==="CallExpression"?`${W} | ${$t(R.filter)}`:`${W} | ${R.filter.value}`}case"SelectExpression":{const R=S;return`${$t(R.lhs)} if ${$t(R.test)}`}case"TestExpression":{const R=S;return`${$t(R.operand)} is${R.negate?" not":""} ${R.test.value}`}case"ArrayLiteral":case"TupleLiteral":{const R=S.value.map($t),W=S.type==="ArrayLiteral"?"[]":"()";return`${W[0]}${R.join(", ")}${W[1]}`}case"ObjectLiteral":return`{${Array.from(S.value.entries()).map(([W,X])=>`${$t(W)}: ${$t(X)}`).join(", ")}}`;case"SliceExpression":{const R=S,W=R.start?$t(R.start):"",X=R.stop?$t(R.stop):"",ue=R.step?`:${$t(R.step)}`:"";return`${W}:${X}${ue}`}case"KeywordArgumentExpression":{const R=S;return`${R.key.value}=${$t(R.value)}`}case"Ternary":{const R=S,W=`${$t(R.trueExpr)} if ${$t(R.condition,0)} else ${$t(R.falseExpr)}`;return H>-1?`(${W})`:W}default:throw new Error(`Unknown expression type: ${S.type}`)}}var Fs=class{constructor(S){J(this,"parsed");const H=d(S,{lstrip_blocks:!0,trim_blocks:!0});this.parsed=Se(H)}render(S){const H=new pt;if(ft(H),S)for(const[X,ue]of Object.entries(S))H.set(X,ue);return new Mr(H).run(this.parsed).value}format(S){return nn(this.parsed,(S==null?void 0:S.indent)||"	")}}},"./src/backends/onnx.js":(e,r,t)=>{var s;t.r(r),t.d(r,{Tensor:()=>i.Tensor,createInferenceSession:()=>k,deviceToExecutionProviders:()=>_,isONNXProxy:()=>x,isONNXTensor:()=>T,runInferenceSession:()=>$});var o=t("./src/env.js"),n=t("?2ce3"),a=t("onnxruntime-web"),i=t("onnxruntime-common");const l=Object.freeze({auto:null,gpu:null,cpu:"cpu",wasm:"wasm",webgpu:"webgpu",cuda:"cuda",dml:"dml",webnn:{name:"webnn",deviceType:"cpu"},"webnn-npu":{name:"webnn",deviceType:"npu"},"webnn-gpu":{name:"webnn",deviceType:"gpu"},"webnn-cpu":{name:"webnn",deviceType:"cpu"}}),c=[];let p,d;const u=Symbol.for("onnxruntime");if(u in globalThis)d=globalThis[u];else if(o.apis.IS_NODE_ENV){switch(d=n??(s||(s=t.t(n,2))),process.platform){case"win32":c.push("dml");break;case"linux":process.arch==="x64"&&c.push("cuda");break}c.push("cpu"),p=["cpu"]}else d=a,o.apis.IS_WEBNN_AVAILABLE&&c.push("webnn-npu","webnn-gpu","webnn-cpu","webnn"),o.apis.IS_WEBGPU_AVAILABLE&&c.push("webgpu"),c.push("wasm"),p=["wasm"];const f=d.InferenceSession;function _(v=null){if(!v)return p;switch(v){case"auto":return c;case"gpu":return c.filter(C=>["webgpu","cuda","dml","webnn-gpu"].includes(C))}if(c.includes(v))return[l[v]??v];throw new Error(`Unsupported device: "${v}". Should be one of: ${c.join(", ")}.`)}let b=null;async function k(v,C,F){b&&await b;const D=f.create(v,C);b??(b=D);const K=await D;return K.config=F,K}let w=Promise.resolve();const y=o.apis.IS_BROWSER_ENV||o.apis.IS_WEBWORKER_ENV;async function $(v,C){const F=()=>v.run(C);return await(y?w=w.then(F):F())}function T(v){return v instanceof d.Tensor}const M=d==null?void 0:d.env;M!=null&&M.wasm&&(!(typeof ServiceWorkerGlobalScope<"u"&&self instanceof ServiceWorkerGlobalScope)&&!M.wasm.wasmPaths&&(M.wasm.wasmPaths=`https://cdn.jsdelivr.net/npm/@huggingface/transformers@${o.env.version}/dist/`),M.wasm.proxy=!1),M!=null&&M.webgpu&&(M.webgpu.powerPreference="high-performance");function x(){var v;return(v=M==null?void 0:M.wasm)==null?void 0:v.proxy}o.env.backends.onnx=M},"./src/base/feature_extraction_utils.js":(e,r,t)=>{t.r(r),t.d(r,{FeatureExtractor:()=>a,validate_audio_inputs:()=>i});var s=t("./src/utils/constants.js"),o=t("./src/utils/generic.js"),n=t("./src/utils/hub.js");class a extends o.Callable{constructor(c){super(),this.config=c}static async from_pretrained(c,p={}){const d=await(0,n.getModelJSON)(c,s.FEATURE_EXTRACTOR_NAME,!0,p);return new this(d)}}function i(l,c){var p;if(!(l instanceof Float32Array||l instanceof Float64Array))throw new Error(`${c} expects input to be a Float32Array or a Float64Array, but got ${((p=l==null?void 0:l.constructor)==null?void 0:p.name)??typeof l} instead. If using the feature extractor directly, remember to use \`read_audio(url, sampling_rate)\` to obtain the raw audio data of the file/url.`)}},"./src/base/image_processors_utils.js":(e,r,t)=>{t.r(r),t.d(r,{ImageProcessor:()=>T,center_to_corners_format:()=>d,post_process_instance_segmentation:()=>$,post_process_object_detection:()=>u,post_process_panoptic_segmentation:()=>y,post_process_semantic_segmentation:()=>f});var s=t("./src/utils/generic.js"),o=t("./src/utils/tensor.js"),n=t("./src/utils/maths.js");t("./src/utils/image.js");var a=t("./src/utils/core.js"),i=t("./src/utils/hub.js"),l=t("./src/utils/constants.js");function c(M,x,v=0,C=null){const F=M/x;let D=(0,n.bankers_round)(F)*x;return C!==null&&D>C&&(D=Math.floor(F)*x),D<v&&(D=Math.ceil(F)*x),D}function p([M,x],v){return[Math.max(Math.floor(M/v),1)*v,Math.max(Math.floor(x/v),1)*v]}function d([M,x,v,C]){return[M-v/2,x-C/2,M+v/2,x+C/2]}function u(M,x=.5,v=null,C=!1){const F=M.logits,D=M.pred_boxes,[K,U,j]=F.dims;if(v!==null&&v.length!==K)throw Error("Make sure that you pass in as many target sizes as the batch dimension of the logits");let se=[];for(let q=0;q<K;++q){let re=v!==null?v[q]:null,Z={boxes:[],classes:[],scores:[]},ae=F[q],_e=D[q];for(let pe=0;pe<U;++pe){let oe=ae[pe],G=[],B;if(C){B=oe.sigmoid().data;for(let Q=0;Q<B.length;++Q)B[Q]>x&&G.push(Q)}else{let Q=(0,n.max)(oe.data)[1];if(Q===j-1||(B=(0,n.softmax)(oe.data),B[Q]<x))continue;G.push(Q)}for(const Q of G){let le=_e[pe].data;le=d(le),re!==null&&(le=le.map((ee,Se)=>ee*re[(Se+1)%2])),Z.boxes.push(le),Z.classes.push(Q),Z.scores.push(B[Q])}}se.push(Z)}return se}function f(M,x=null){const v=M.logits,C=v.dims[0];if(x!==null&&x.length!==C)throw Error("Make sure that you pass in as many target sizes as the batch dimension of the logits");const F=[];for(let D=0;D<C;++D){const K=x!==null?x[D]:null;let U=v[D];K!==null&&(U=(0,o.interpolate)(U,K,"bilinear",!1));const[j,se]=K??U.dims.slice(-2),q=new o.Tensor("int32",new Int32Array(j*se),[j,se]),re=U[0].data,Z=q.data;for(let pe=1;pe<U.dims[0];++pe){const oe=U[pe].data;for(let G=0;G<oe.length;++G)oe[G]>re[G]&&(re[G]=oe[G],Z[G]=pe)}const ae=new Array(U.dims[0]);for(let pe=0;pe<Z.length;++pe){const oe=Z[pe];ae[oe]=oe}const _e=ae.filter(pe=>pe!==void 0);F.push({segmentation:q,labels:_e})}return F}function _(M,x,v,C){const F=[],D=[],K=[];for(let U=0;U<M.dims[0];++U){const j=M[U],se=x[U],q=(0,n.max)(j.data)[1];if(q===C)continue;const Z=(0,n.softmax)(j.data)[q];Z>v&&(F.push(se),D.push(Z),K.push(q))}return[F,D,K]}function b(M,x,v,C=.5,F=.8){const D=[];let K=0,U=0;const j=x[v].data;for(let q=0;q<M.length;++q)M[q]===v&&(D.push(q),++K),j[q]>=C&&++U;let se=K>0&&U>0;return se&&(se=K/U>F),[se,D]}function k(M,x,v,C,F,D=null,K=null){const[U,j]=K??M[0].dims,se=new o.Tensor("int32",new Int32Array(U*j),[U,j]),q=[];if(K!==null)for(let pe=0;pe<M.length;++pe)M[pe]=(0,o.interpolate)(M[pe],K,"bilinear",!1);const re=new Int32Array(M[0].data.length),Z=new Float32Array(M[0].data.length);for(let pe=0;pe<M.length;++pe){let oe=x[pe];const G=M[pe].data;for(let B=0;B<G.length;++B)G[B]*=oe,G[B]>Z[B]&&(re[B]=pe,Z[B]=G[B])}let ae=0;const _e=se.data;for(let pe=0;pe<v.length;++pe){const oe=v[pe],[G,B]=b(re,M,pe,C,F);if(G){++ae;for(const Q of B)_e[Q]=ae;q.push({id:ae,label_id:oe,score:x[pe]})}}return[se,q]}function w(M,x,v=28,C=56*56,F=14*14*4*1280){if(M<v||x<v)throw new Error(`height:${M} or width:${x} must be larger than factor:${v}`);if(Math.max(M,x)/Math.min(M,x)>200)throw new Error(`absolute aspect ratio must be smaller than 200, got ${Math.max(M,x)/Math.min(M,x)}`);let D=Math.round(M/v)*v,K=Math.round(x/v)*v;if(D*K>F){const U=Math.sqrt(M*x/F);D=Math.floor(M/U/v)*v,K=Math.floor(x/U/v)*v}else if(D*K<C){const U=Math.sqrt(C/(M*x));D=Math.ceil(M*U/v)*v,K=Math.ceil(x*U/v)*v}return[D,K]}function y(M,x=.5,v=.5,C=.8,F=null,D=null){F===null&&(console.warn("`label_ids_to_fuse` unset. No instance will be fused."),F=new Set);const K=M.class_queries_logits??M.logits,j=(M.masks_queries_logits??M.pred_masks).sigmoid();let[se,q,re]=K.dims;if(re-=1,D!==null&&D.length!==se)throw Error("Make sure that you pass in as many target sizes as the batch dimension of the logits");let Z=[];for(let ae=0;ae<se;++ae){let _e=D!==null?D[ae]:null,pe=K[ae],oe=j[ae],[G,B,Q]=_(pe,oe,x,re);if(Q.length===0){let[Se,De]=_e??oe.dims.slice(-2),Ne=new o.Tensor("int32",new Int32Array(Se*De).fill(-1),[Se,De]);Z.push({segmentation:Ne,segments_info:[]});continue}let[le,ee]=k(G,B,Q,v,C,F,_e);Z.push({segmentation:le,segments_info:ee})}return Z}function $(M,x=.5,v=null){throw new Error("`post_process_instance_segmentation` is not yet implemented.")}class T extends s.Callable{constructor(x){super(),this.image_mean=x.image_mean??x.mean,this.image_std=x.image_std??x.std,this.resample=x.resample??2,this.do_rescale=x.do_rescale??!0,this.rescale_factor=x.rescale_factor??1/255,this.do_normalize=x.do_normalize,this.do_thumbnail=x.do_thumbnail,this.size=x.size??x.image_size,this.do_resize=x.do_resize??this.size!==void 0,this.size_divisibility=x.size_divisibility??x.size_divisor,this.do_center_crop=x.do_center_crop,this.crop_size=x.crop_size,this.do_convert_rgb=x.do_convert_rgb??!0,this.do_crop_margin=x.do_crop_margin,this.pad_size=x.pad_size,this.do_pad=x.do_pad,this.min_pixels=x.min_pixels,this.max_pixels=x.max_pixels,this.do_pad&&!this.pad_size&&this.size&&this.size.width!==void 0&&this.size.height!==void 0&&(this.pad_size=this.size),this.do_flip_channel_order=x.do_flip_channel_order??!1,this.config=x}async thumbnail(x,v,C=2){const F=x.height,D=x.width,K=v.height,U=v.width;let j=Math.min(F,K),se=Math.min(D,U);return j===F&&se===D?x:(F>D?se=Math.floor(D*j/F):D>F&&(j=Math.floor(F*se/D)),await x.resize(se,j,{resample:C}))}async crop_margin(x,v=200){const C=x.clone().grayscale(),F=(0,n.min)(C.data)[0],K=(0,n.max)(C.data)[0]-F;if(K===0)return x;const U=v/255;let j=C.width,se=C.height,q=0,re=0;const Z=C.data;for(let ae=0;ae<C.height;++ae){const _e=ae*C.width;for(let pe=0;pe<C.width;++pe)(Z[_e+pe]-F)/K<U&&(j=Math.min(j,pe),se=Math.min(se,ae),q=Math.max(q,pe),re=Math.max(re,ae))}return x=await x.crop([j,se,q,re]),x}pad_image(x,v,C,{mode:F="constant",center:D=!1,constant_values:K=0}={}){const[U,j,se]=v;let q,re;if(typeof C=="number"?(q=C,re=C):C==="square"?q=re=Math.max(U,j):(q=C.width,re=C.height),q!==j||re!==U){const Z=new Float32Array(q*re*se);if(Array.isArray(K))for(let pe=0;pe<Z.length;++pe)Z[pe]=K[pe%se];else K!==0&&Z.fill(K);const[ae,_e]=D?[Math.floor((q-j)/2),Math.floor((re-U)/2)]:[0,0];for(let pe=0;pe<U;++pe){const oe=(pe+_e)*q,G=pe*j;for(let B=0;B<j;++B){const Q=(oe+B+ae)*se,le=(G+B)*se;for(let ee=0;ee<se;++ee)Z[Q+ee]=x[le+ee]}}if(F==="symmetric"){if(D)throw new Error("`center` padding is not supported when `mode` is set to `symmetric`.");const pe=U-1,oe=j-1;for(let G=0;G<re;++G){const B=G*q,Q=(0,a.calculateReflectOffset)(G,pe)*j;for(let le=0;le<q;++le){if(G<U&&le<j)continue;const ee=(B+le)*se,Se=(Q+(0,a.calculateReflectOffset)(le,oe))*se;for(let De=0;De<se;++De)Z[ee+De]=x[Se+De]}}}x=Z,v=[re,q,se]}return[x,v]}rescale(x){for(let v=0;v<x.length;++v)x[v]=this.rescale_factor*x[v]}get_resize_output_image_size(x,v){const[C,F]=x.size;let D,K;if(this.do_thumbnail){const{height:U,width:j}=v;D=Math.min(U,j)}else Number.isInteger(v)?(D=v,K=this.config.max_size??D):v!==void 0&&(D=v.shortest_edge,K=v.longest_edge);if(D!==void 0||K!==void 0){const U=D===void 0?1:Math.max(D/C,D/F),j=C*U,se=F*U,q=K===void 0?1:Math.min(K/j,K/se);let re=Math.floor(Number((j*q).toFixed(2))),Z=Math.floor(Number((se*q).toFixed(2)));return this.size_divisibility!==void 0&&([re,Z]=p([re,Z],this.size_divisibility)),[re,Z]}else if(v!==void 0&&v.width!==void 0&&v.height!==void 0){let U=v.width,j=v.height;if(this.config.keep_aspect_ratio&&this.config.ensure_multiple_of){let se=j/F,q=U/C;Math.abs(1-q)<Math.abs(1-se)?se=q:q=se,j=c(se*F,this.config.ensure_multiple_of),U=c(q*C,this.config.ensure_multiple_of)}return[U,j]}else{if(this.size_divisibility!==void 0)return p([C,F],this.size_divisibility);if(this.min_pixels!==void 0&&this.max_pixels!==void 0){const U=this.config.patch_size*this.config.merge_size;return w(F,C,U,this.min_pixels,this.max_pixels)}else throw new Error(`Could not resize image due to unsupported \`this.size\` option in config: ${JSON.stringify(v)}`)}}async resize(x){const[v,C]=this.get_resize_output_image_size(x,this.size);return await x.resize(v,C,{resample:this.resample})}async preprocess(x,{do_normalize:v=null,do_pad:C=null,do_convert_rgb:F=null,do_convert_grayscale:D=null,do_flip_channel_order:K=null}={}){this.do_crop_margin&&(x=await this.crop_margin(x));const[U,j]=x.size;if(F??this.do_convert_rgb?x=x.rgb():D&&(x=x.grayscale()),this.do_resize&&(x=await this.resize(x)),this.do_thumbnail&&(x=await this.thumbnail(x,this.size,this.resample)),this.do_center_crop){let ae,_e;Number.isInteger(this.crop_size)?(ae=this.crop_size,_e=this.crop_size):(ae=this.crop_size.width,_e=this.crop_size.height),x=await x.center_crop(ae,_e)}const se=[x.height,x.width];let q=Float32Array.from(x.data),re=[x.height,x.width,x.channels];if(this.do_rescale&&this.rescale(q),v??this.do_normalize){let ae=this.image_mean;Array.isArray(this.image_mean)||(ae=new Array(x.channels).fill(ae));let _e=this.image_std;if(Array.isArray(this.image_std)||(_e=new Array(x.channels).fill(_e)),ae.length!==x.channels||_e.length!==x.channels)throw new Error(`When set to arrays, the length of \`image_mean\` (${ae.length}) and \`image_std\` (${_e.length}) must match the number of channels in the image (${x.channels}).`);for(let pe=0;pe<q.length;pe+=x.channels)for(let oe=0;oe<x.channels;++oe)q[pe+oe]=(q[pe+oe]-ae[oe])/_e[oe]}if(C??this.do_pad){if(this.pad_size)[q,re]=this.pad_image(q,[x.height,x.width,x.channels],this.pad_size);else if(this.size_divisibility){const[ae,_e]=p([re[1],re[0]],this.size_divisibility);[q,re]=this.pad_image(q,re,{width:ae,height:_e})}}if(K??this.do_flip_channel_order){if(re[2]!==3)throw new Error("Flipping channel order is only supported for RGB images.");for(let ae=0;ae<q.length;ae+=3){const _e=q[ae];q[ae]=q[ae+2],q[ae+2]=_e}}const Z=new o.Tensor("float32",q,re).permute(2,0,1);return{original_size:[j,U],reshaped_input_size:se,pixel_values:Z}}async _call(x,...v){Array.isArray(x)||(x=[x]);const C=await Promise.all(x.map(D=>this.preprocess(D)));return{pixel_values:(0,o.stack)(C.map(D=>D.pixel_values),0),original_sizes:C.map(D=>D.original_size),reshaped_input_sizes:C.map(D=>D.reshaped_input_size)}}static async from_pretrained(x,v={}){const C=await(0,i.getModelJSON)(x,l.IMAGE_PROCESSOR_NAME,!0,v);return new this(C)}}},"./src/base/processing_utils.js":(e,r,t)=>{t.r(r),t.d(r,{Processor:()=>a});var s=t("./src/utils/constants.js"),o=t("./src/utils/generic.js"),n=t("./src/utils/hub.js");class a extends o.Callable{constructor(l,c,p){super(),this.config=l,this.components=c,this.chat_template=p}get image_processor(){return this.components.image_processor}get tokenizer(){return this.components.tokenizer}get feature_extractor(){return this.components.feature_extractor}apply_chat_template(l,c={}){if(!this.tokenizer)throw new Error("Unable to apply chat template without a tokenizer.");return this.tokenizer.apply_chat_template(l,{tokenize:!1,chat_template:this.chat_template??void 0,...c})}batch_decode(...l){if(!this.tokenizer)throw new Error("Unable to decode without a tokenizer.");return this.tokenizer.batch_decode(...l)}decode(...l){if(!this.tokenizer)throw new Error("Unable to decode without a tokenizer.");return this.tokenizer.decode(...l)}async _call(l,...c){for(const p of[this.image_processor,this.feature_extractor,this.tokenizer])if(p)return p(l,...c);throw new Error("No image processor, feature extractor, or tokenizer found.")}static async from_pretrained(l,c={}){const[p,d,u]=await Promise.all([this.uses_processor_config?(0,n.getModelJSON)(l,s.PROCESSOR_NAME,!0,c):{},Promise.all(this.classes.filter(f=>f in this).map(async f=>{const _=await this[f].from_pretrained(l,c);return[f.replace(/_class$/,""),_]})).then(Object.fromEntries),this.uses_chat_template_file?(0,n.getModelText)(l,s.CHAT_TEMPLATE_NAME,!0,c):null]);return new this(p,d,u)}}J(a,"classes",["image_processor_class","tokenizer_class","feature_extractor_class"]),J(a,"uses_processor_config",!1),J(a,"uses_chat_template_file",!1)},"./src/configs.js":(e,r,t)=>{t.r(r),t.d(r,{AutoConfig:()=>p,PretrainedConfig:()=>c,getCacheShapes:()=>i});var s=t("./src/utils/core.js"),o=t("./src/utils/hub.js");async function n(d,u){return await(0,o.getModelJSON)(d,"config.json",!0,u)}function a(d){const u={};let f={};switch(d.model_type){case"llava":case"paligemma":case"gemma3":case"florence2":case"llava_onevision":case"idefics3":case"ultravox":case"voxtral":case"smolvlm":case"gemma3n":case"mistral3":f=a(d.text_config);break;case"moondream1":f=a(d.phi_config);break;case"musicgen":f=a(d.decoder);break;case"multi_modality":f=a(d.language_config);break;case"gpt2":case"gptj":case"jais":case"codegen":case"gpt_bigcode":u.num_heads="n_head",u.num_layers="n_layer",u.hidden_size="n_embd";break;case"gpt_neox":case"stablelm":case"opt":case"falcon":case"modernbert-decoder":u.num_heads="num_attention_heads",u.num_layers="num_hidden_layers",u.hidden_size="hidden_size";break;case"llama":case"llama4_text":case"nanochat":case"arcee":case"lfm2":case"smollm3":case"olmo":case"olmo2":case"mobilellm":case"granite":case"granitemoehybrid":case"cohere":case"mistral":case"starcoder2":case"qwen2":case"qwen2_vl":case"phi":case"phi3":case"phi3_v":case"llava_qwen2":u.num_heads="num_key_value_heads",u.num_layers="num_hidden_layers",u.hidden_size="hidden_size",u.num_attention_heads="num_attention_heads",u.dim_kv="head_dim";break;case"qwen3":case"gemma":case"gemma2":case"vaultgemma":case"gemma3_text":case"gemma3n_text":case"glm":case"helium":case"ernie4_5":case"ministral":case"ministral3":u.num_heads="num_key_value_heads",u.num_layers="num_hidden_layers",u.dim_kv="head_dim";break;case"openelm":u.num_heads="num_kv_heads",u.num_layers="num_transformer_layers",u.dim_kv="head_dim";break;case"gpt_neo":case"donut-swin":u.num_heads="num_heads",u.num_layers="num_layers",u.hidden_size="hidden_size";break;case"bloom":u.num_heads="n_head",u.num_layers="n_layer",u.hidden_size="hidden_size";break;case"mpt":u.num_heads="n_heads",u.num_layers="n_layers",u.hidden_size="d_model";break;case"exaone":u.num_heads="num_key_value_heads",u.num_layers="num_layers",u.dim_kv="head_dim",u.num_attention_heads="num_attention_heads";break;case"t5":case"mt5":case"longt5":u.num_decoder_layers="num_decoder_layers",u.num_decoder_heads="num_heads",u.decoder_dim_kv="d_kv",u.num_encoder_layers="num_layers",u.num_encoder_heads="num_heads",u.encoder_dim_kv="d_kv";break;case"bart":case"mbart":case"marian":case"whisper":case"lite-whisper":case"m2m_100":case"blenderbot":case"blenderbot-small":case"florence2_language":u.num_decoder_layers="decoder_layers",u.num_decoder_heads="decoder_attention_heads",u.decoder_hidden_size="d_model",u.num_encoder_layers="encoder_layers",u.num_encoder_heads="encoder_attention_heads",u.encoder_hidden_size="d_model";break;case"speecht5":u.num_decoder_layers="decoder_layers",u.num_decoder_heads="decoder_attention_heads",u.decoder_hidden_size="hidden_size",u.num_encoder_layers="encoder_layers",u.num_encoder_heads="encoder_attention_heads",u.encoder_hidden_size="hidden_size";break;case"trocr":u.num_encoder_layers=u.num_decoder_layers="decoder_layers",u.num_encoder_heads=u.num_decoder_heads="decoder_attention_heads",u.encoder_hidden_size=u.decoder_hidden_size="d_model";break;case"musicgen_decoder":u.num_encoder_layers=u.num_decoder_layers="num_hidden_layers",u.num_encoder_heads=u.num_decoder_heads="num_attention_heads",u.encoder_hidden_size=u.decoder_hidden_size="hidden_size";break;case"moonshine":u.num_decoder_layers="decoder_num_hidden_layers",u.num_decoder_heads="decoder_num_key_value_heads",u.num_encoder_layers="encoder_num_hidden_layers",u.num_encoder_heads="encoder_num_key_value_heads",u.encoder_hidden_size=u.decoder_hidden_size="hidden_size";break;case"vision-encoder-decoder":const b=a(d.decoder),k="num_decoder_layers"in b,w=(0,s.pick)(d,["model_type","is_encoder_decoder"]);return k?(w.num_decoder_layers=b.num_decoder_layers,w.num_decoder_heads=b.num_decoder_heads,w.decoder_hidden_size=b.decoder_hidden_size,w.num_encoder_layers=b.num_encoder_layers,w.num_encoder_heads=b.num_encoder_heads,w.encoder_hidden_size=b.encoder_hidden_size):(w.num_layers=b.num_layers,w.num_heads=b.num_heads,w.hidden_size=b.hidden_size),w}const _={...f,...(0,s.pick)(d,["model_type","multi_query","is_encoder_decoder"])};for(const b in u)_[b]=d[u[b]];return _}function i(d,u){if(d.model_type==="lfm2"){const f=(u==null?void 0:u.prefix)??"past_key_values",_=f==="present"?"present":"past",b={},{layer_types:k,num_attention_heads:w,num_key_value_heads:y,hidden_size:$,conv_L_cache:T}=d,M=$/w,x=(u==null?void 0:u.batch_size)??1;for(let v=0;v<k.length;++v)if(k[v]==="full_attention")for(const C of["key","value"])b[`${f}.${v}.${C}`]=[x,y,0,M];else if(k[v]==="conv")b[`${_}_conv.${v}`]=[x,$,T];else throw new Error(`Unsupported layer type: ${k[v]}`);return b}return l(d,u)}function l(d,{prefix:u="past_key_values",batch_size:f=1}={}){const _={},b=d.normalized_config;if(b.is_encoder_decoder&&"num_encoder_heads"in b&&"num_decoder_heads"in b){const k=b.encoder_dim_kv??b.encoder_hidden_size/b.num_encoder_heads,w=b.decoder_dim_kv??b.decoder_hidden_size/b.num_decoder_heads,y=[f,b.num_encoder_heads,0,k],$=[f,b.num_decoder_heads,0,w];for(let T=0;T<b.num_decoder_layers;++T)_[`${u}.${T}.encoder.key`]=y,_[`${u}.${T}.encoder.value`]=y,_[`${u}.${T}.decoder.key`]=$,_[`${u}.${T}.decoder.value`]=$}else{const k=b.num_heads,w=b.num_layers,y=b.dim_kv??b.hidden_size/(b.num_attention_heads??k);if(b.model_type==="falcon"){const $=[f*k,0,y];for(let T=0;T<w;++T)_[`${u}.${T}.key`]=$,_[`${u}.${T}.value`]=$}else if(b.multi_query){const $=[f*k,0,2*y];for(let T=0;T<w;++T)_[`${u}.${T}.key_value`]=$}else if(b.model_type==="bloom"){const $=[f*k,y,0],T=[f*k,0,y];for(let M=0;M<w;++M)_[`${u}.${M}.key`]=$,_[`${u}.${M}.value`]=T}else if(b.model_type==="openelm")for(let $=0;$<w;++$){const T=[f,k[$],0,y];_[`${u}.${$}.key`]=T,_[`${u}.${$}.value`]=T}else{const $=[f,k,0,y];for(let T=0;T<w;++T)_[`${u}.${T}.key`]=$,_[`${u}.${T}.value`]=$}}return _}class c{constructor(u){J(this,"model_type",null);J(this,"is_encoder_decoder",!1);J(this,"max_position_embeddings");J(this,"transformers.js_config");Object.assign(this,u),this.normalized_config=a(this)}static async from_pretrained(u,{progress_callback:f=null,config:_=null,cache_dir:b=null,local_files_only:k=!1,revision:w="main"}={}){_&&!(_ instanceof c)&&(_=new c(_));const y=_??await n(u,{progress_callback:f,config:_,cache_dir:b,local_files_only:k,revision:w});return new this(y)}}class p{static async from_pretrained(...u){return c.from_pretrained(...u)}}},"./src/env.js":(e,r,t)=>{var F,D;t.r(r),t.d(r,{apis:()=>w,env:()=>v});var s=t("?db59"),o=t("?383f"),n=t("?fa4b");const a="3.8.1",i=typeof window<"u"&&typeof window.document<"u",l=typeof self<"u"&&["DedicatedWorkerGlobalScope","ServiceWorkerGlobalScope","SharedWorkerGlobalScope"].includes((F=self.constructor)==null?void 0:F.name),c=typeof self<"u"&&"caches"in self,p=typeof navigator<"u"&&"gpu"in navigator,d=typeof navigator<"u"&&"ml"in navigator,u=typeof process<"u",f=u&&((D=process==null?void 0:process.release)==null?void 0:D.name)==="node",_=!C(s),b=!C(o),k=typeof globalThis.Deno<"u",w=Object.freeze({IS_BROWSER_ENV:i,IS_WEBWORKER_ENV:l,IS_WEB_CACHE_AVAILABLE:c,IS_WEBGPU_AVAILABLE:p,IS_WEBNN_AVAILABLE:d,IS_PROCESS_AVAILABLE:u,IS_NODE_ENV:f,IS_FS_AVAILABLE:_,IS_PATH_AVAILABLE:b}),y=_&&b;let $="./";if(y){const K=Object({url:self.location.href}).url;K?$=o.dirname(o.dirname(n.fileURLToPath(K))):typeof __dirname<"u"&&($=o.dirname(__dirname))}const T=y?o.join($,"/.cache/"):null,M="/models/",x=y?o.join($,M):M,v={version:a,backends:{onnx:{}},allowRemoteModels:!0,remoteHost:"https://huggingface.co/",remotePathTemplate:"{model}/resolve/{revision}/",allowLocalModels:!(i||l),localModelPath:x,useFS:_,useBrowserCache:c&&!k,useFSCache:_,cacheDir:T,useCustomCache:!1,customCache:null};function C(K){return Object.keys(K).length===0}},"./src/generation/configuration_utils.js":(e,r,t)=>{t.r(r),t.d(r,{GenerationConfig:()=>o});var s=t("./src/utils/core.js");class o{constructor(a){J(this,"max_length",20);J(this,"max_new_tokens",null);J(this,"min_length",0);J(this,"min_new_tokens",null);J(this,"early_stopping",!1);J(this,"max_time",null);J(this,"do_sample",!1);J(this,"num_beams",1);J(this,"num_beam_groups",1);J(this,"penalty_alpha",null);J(this,"use_cache",!0);J(this,"temperature",1);J(this,"top_k",50);J(this,"top_p",1);J(this,"typical_p",1);J(this,"epsilon_cutoff",0);J(this,"eta_cutoff",0);J(this,"diversity_penalty",0);J(this,"repetition_penalty",1);J(this,"encoder_repetition_penalty",1);J(this,"length_penalty",1);J(this,"no_repeat_ngram_size",0);J(this,"bad_words_ids",null);J(this,"force_words_ids",null);J(this,"renormalize_logits",!1);J(this,"constraints",null);J(this,"forced_bos_token_id",null);J(this,"forced_eos_token_id",null);J(this,"remove_invalid_values",!1);J(this,"exponential_decay_length_penalty",null);J(this,"suppress_tokens",null);J(this,"streamer",null);J(this,"begin_suppress_tokens",null);J(this,"forced_decoder_ids",null);J(this,"guidance_scale",null);J(this,"num_return_sequences",1);J(this,"output_attentions",!1);J(this,"output_hidden_states",!1);J(this,"output_scores",!1);J(this,"return_dict_in_generate",!1);J(this,"pad_token_id",null);J(this,"bos_token_id",null);J(this,"eos_token_id",null);J(this,"encoder_no_repeat_ngram_size",0);J(this,"decoder_start_token_id",null);J(this,"generation_kwargs",{});Object.assign(this,(0,s.pick)(a,Object.getOwnPropertyNames(this)))}}},"./src/generation/logits_process.js":(e,r,t)=>{t.r(r),t.d(r,{ClassifierFreeGuidanceLogitsProcessor:()=>w,ForcedBOSTokenLogitsProcessor:()=>l,ForcedEOSTokenLogitsProcessor:()=>c,LogitsProcessor:()=>n,LogitsProcessorList:()=>i,LogitsWarper:()=>a,MinLengthLogitsProcessor:()=>_,MinNewTokensLengthLogitsProcessor:()=>b,NoBadWordsLogitsProcessor:()=>k,NoRepeatNGramLogitsProcessor:()=>u,RepetitionPenaltyLogitsProcessor:()=>f,SuppressTokensAtBeginLogitsProcessor:()=>p,TemperatureLogitsWarper:()=>y,TopKLogitsWarper:()=>T,TopPLogitsWarper:()=>$,WhisperTimeStampLogitsProcessor:()=>d});var s=t("./src/utils/generic.js");t("./src/utils/tensor.js");var o=t("./src/utils/maths.js");class n extends s.Callable{_call(x,v){throw Error("`_call` should be implemented in a subclass")}}class a extends s.Callable{_call(x,v){throw Error("`_call` should be implemented in a subclass")}}class i extends s.Callable{constructor(){super(),this.processors=[]}push(x){this.processors.push(x)}extend(x){this.processors.push(...x)}_call(x,v){let C=v;for(const F of this.processors)C=F(x,C);return C}[Symbol.iterator](){return this.processors.values()}}class l extends n{constructor(x){super(),this.bos_token_id=x}_call(x,v){for(let C=0;C<x.length;++C)if(x[C].length===1){const F=v[C].data;F.fill(-1/0),F[this.bos_token_id]=0}return v}}class c extends n{constructor(x,v){super(),this.max_length=x,this.eos_token_id=Array.isArray(v)?v:[v]}_call(x,v){for(let C=0;C<x.length;++C)if(x[C].length===this.max_length-1){const F=v[C].data;F.fill(-1/0);for(const D of this.eos_token_id)F[D]=0}return v}}class p extends n{constructor(x,v){super(),this.begin_suppress_tokens=x,this.begin_index=v}_call(x,v){for(let C=0;C<x.length;++C)if(x[C].length===this.begin_index){const F=v[C].data;for(const D of this.begin_suppress_tokens)F[D]=-1/0}return v}}class d extends n{constructor(x,v){super(),this.eos_token_id=Array.isArray(x.eos_token_id)?x.eos_token_id[0]:x.eos_token_id,this.no_timestamps_token_id=x.no_timestamps_token_id,this.timestamp_begin=this.no_timestamps_token_id+1,this.begin_index=v.length,v.at(-1)===this.no_timestamps_token_id&&(this.begin_index-=1),this.max_initial_timestamp_index=x.max_initial_timestamp_index}_call(x,v){for(let C=0;C<x.length;++C){const F=v[C].data;if(F[this.no_timestamps_token_id]=-1/0,x[C].length===this.begin_index-1){F.fill(-1/0),F[this.timestamp_begin]=0;continue}const D=x[C].slice(this.begin_index),K=D.length>=1&&D[D.length-1]>=this.timestamp_begin,U=D.length<2||D[D.length-2]>=this.timestamp_begin;if(K&&(U?F.subarray(this.timestamp_begin).fill(-1/0):F.subarray(0,this.eos_token_id).fill(-1/0)),x[C].length===this.begin_index&&this.max_initial_timestamp_index!==null){const re=this.timestamp_begin+this.max_initial_timestamp_index;F.subarray(re+1).fill(-1/0)}const j=(0,o.log_softmax)(F),se=Math.log(j.subarray(this.timestamp_begin).map(Math.exp).reduce((re,Z)=>re+Z)),q=(0,o.max)(j.subarray(0,this.timestamp_begin))[0];se>q&&F.subarray(0,this.timestamp_begin).fill(-1/0)}return v}}class u extends n{constructor(x){super(),this.no_repeat_ngram_size=x}getNgrams(x){const v=x.length,C=[];for(let D=0;D<v+1-this.no_repeat_ngram_size;++D){const K=[];for(let U=0;U<this.no_repeat_ngram_size;++U)K.push(x[D+U]);C.push(K.map(Number))}const F=new Map;for(const D of C){const K=D.slice(0,D.length-1),U=JSON.stringify(K),j=F.get(U)??[];j.push(D[D.length-1]),F.set(U,j)}return F}getGeneratedNgrams(x,v){const C=v.slice(v.length+1-this.no_repeat_ngram_size,v.length);return x.get(JSON.stringify(C.map(Number)))??[]}calcBannedNgramTokens(x){const v=[];if(x.length+1<this.no_repeat_ngram_size)return v;{const C=this.getNgrams(x);return this.getGeneratedNgrams(C,x)}}_call(x,v){for(let C=0;C<x.length;++C){const F=v[C].data,D=this.calcBannedNgramTokens(x[C]);for(const K of D)F[K]=-1/0}return v}}class f extends n{constructor(x){super(),this.penalty=x}_call(x,v){for(let C=0;C<x.length;++C){const F=v[C].data;for(const D of new Set(x[C])){const K=Number(D);F[K]<0?F[K]*=this.penalty:F[K]/=this.penalty}}return v}}class _ extends n{constructor(x,v){super(),this.min_length=x,this.eos_token_id=Array.isArray(v)?v:[v]}_call(x,v){for(let C=0;C<x.length;++C)if(x[C].length<this.min_length){const F=v[C].data;for(const D of this.eos_token_id)F[D]=-1/0}return v}}class b extends n{constructor(x,v,C){super(),this.prompt_length_to_skip=x,this.min_new_tokens=v,this.eos_token_id=Array.isArray(C)?C:[C]}_call(x,v){for(let C=0;C<x.length;++C)if(x[C].length-this.prompt_length_to_skip<this.min_new_tokens){const D=v[C].data;for(const K of this.eos_token_id)D[K]=-1/0}return v}}class k extends n{constructor(x,v){super(),this.bad_words_ids=x,this.eos_token_id=Array.isArray(v)?v:[v]}_call(x,v){for(let C=0;C<x.length;++C){const F=v[C].data,D=x[C];for(const K of this.bad_words_ids){if(D.length<K.length-1)continue;let U=!0;for(let j=1;j<=K.length-1;++j)if(K.at(-j-1)!=D.at(-j)){U=!1;break}U&&(F[K.at(-1)]=-1/0)}}return v}}class w extends n{constructor(x){if(super(),x<=1)throw new Error(`Require guidance scale >1 to use the classifier free guidance processor, got guidance scale ${x}.`);this.guidance_scale=x}_call(x,v){if(v.dims[0]!==2*x.length)throw new Error(`Logits should have twice the batch size of the input ids, the first half of batches corresponding to the conditional inputs, and the second half of batches corresponding to the unconditional inputs. Got batch size ${v.dims[0]} for the logits and ${x.length} for the input ids.`);const C=x.length,F=v.slice([0,C],null),D=v.slice([C,v.dims[0]],null);for(let K=0;K<D.data.length;++K)D.data[K]+=(F.data[K]-D.data[K])*this.guidance_scale;return D}}class y extends a{constructor(x){super(),this.temperature=x}_call(x,v){const C=v.data;for(let F=0;F<C.length;++F)C[F]/=this.temperature;return v}}class $ extends a{constructor(x,{filter_value:v=-1/0,min_tokens_to_keep:C=1}={}){if(super(),x<0||x>1)throw new Error(`\`top_p\` must be a float > 0 and < 1, but is ${x}`);if(!Number.isInteger(C)||C<1)throw new Error(`\`min_tokens_to_keep\` must be a positive integer, but is ${C}`);this.top_p=x,this.filter_value=v,this.min_tokens_to_keep=C}}class T extends a{constructor(x,{filter_value:v=-1/0,min_tokens_to_keep:C=1}={}){if(super(),!Number.isInteger(x)||x<0)throw new Error(`\`top_k\` must be a positive integer, but is ${x}`);this.top_k=Math.max(x,C),this.filter_value=v}}},"./src/generation/logits_sampler.js":(e,r,t)=>{t.r(r),t.d(r,{LogitsSampler:()=>a});var s=t("./src/utils/generic.js"),o=t("./src/utils/tensor.js"),n=t("./src/utils/maths.js");t("./src/generation/configuration_utils.js");class a extends s.Callable{constructor(d){super(),this.generation_config=d}async _call(d){return this.sample(d)}async sample(d){throw Error("sample should be implemented in subclasses.")}getLogits(d,u){let f=d.dims.at(-1),_=d.data;if(u===-1)_=_.slice(-f);else{let b=u*f;_=_.slice(b,b+f)}return _}randomSelect(d){let u=0;for(let _=0;_<d.length;++_)u+=d[_];let f=Math.random()*u;for(let _=0;_<d.length;++_)if(f-=d[_],f<=0)return _;return 0}static getSampler(d){if(d.do_sample)return new l(d);if(d.num_beams>1)return new c(d);if(d.num_return_sequences>1)throw Error(`num_return_sequences has to be 1 when doing greedy search, but is ${d.num_return_sequences}.`);return new i(d)}}class i extends a{async sample(d){const u=(0,n.max)(d.data)[1];return[[BigInt(u),0]]}}class l extends a{async sample(d){let u=d.dims.at(-1);this.generation_config.top_k>0&&(u=Math.min(this.generation_config.top_k,u));const[f,_]=await(0,o.topk)(d,u),b=(0,n.softmax)(f.data);return Array.from({length:this.generation_config.num_beams},()=>{const k=this.randomSelect(b);return[_.data[k],Math.log(b[k])]})}}class c extends a{async sample(d){let u=d.dims.at(-1);this.generation_config.top_k>0&&(u=Math.min(this.generation_config.top_k,u));const[f,_]=await(0,o.topk)(d,u),b=(0,n.softmax)(f.data);return Array.from({length:this.generation_config.num_beams},(k,w)=>[_.data[w],Math.log(b[w])])}}},"./src/generation/stopping_criteria.js":(e,r,t)=>{t.r(r),t.d(r,{EosTokenCriteria:()=>i,InterruptableStoppingCriteria:()=>l,MaxLengthCriteria:()=>a,StoppingCriteria:()=>o,StoppingCriteriaList:()=>n});var s=t("./src/utils/generic.js");class o extends s.Callable{_call(p,d){throw Error("StoppingCriteria needs to be subclassed")}}class n extends s.Callable{constructor(){super(),this.criteria=[]}push(p){this.criteria.push(p)}extend(p){p instanceof n?p=p.criteria:p instanceof o&&(p=[p]),this.criteria.push(...p)}_call(p,d){const u=new Array(p.length).fill(!1);for(const f of this.criteria){const _=f(p,d);for(let b=0;b<u.length;++b)u[b]||(u[b]=_[b])}return u}[Symbol.iterator](){return this.criteria.values()}}class a extends o{constructor(p,d=null){super(),this.max_length=p,this.max_position_embeddings=d}_call(p){return p.map(d=>d.length>=this.max_length)}}class i extends o{constructor(p){super(),Array.isArray(p)||(p=[p]),this.eos_token_id=p}_call(p,d){return p.map(u=>{const f=u.at(-1);return this.eos_token_id.some(_=>f==_)})}}class l extends o{constructor(){super(),this.interrupted=!1}interrupt(){this.interrupted=!0}reset(){this.interrupted=!1}_call(p,d){return new Array(p.length).fill(this.interrupted)}}},"./src/generation/streamers.js":(e,r,t)=>{t.r(r),t.d(r,{BaseStreamer:()=>a,TextStreamer:()=>l,WhisperTextStreamer:()=>c});var s=t("./src/utils/core.js"),o=t("./src/tokenizers.js"),n=t("./src/env.js");class a{put(d){throw Error("Not implemented")}end(){throw Error("Not implemented")}}const i=n.apis.IS_PROCESS_AVAILABLE?p=>process.stdout.write(p):p=>console.log(p);class l extends a{constructor(d,{skip_prompt:u=!1,callback_function:f=null,token_callback_function:_=null,skip_special_tokens:b=!0,decode_kwargs:k={},...w}={}){super(),this.tokenizer=d,this.skip_prompt=u,this.callback_function=f??i,this.token_callback_function=_,this.decode_kwargs={skip_special_tokens:b,...k,...w},this.token_cache=[],this.print_len=0,this.next_tokens_are_prompt=!0}put(d){var k;if(d.length>1)throw Error("TextStreamer only supports batch size of 1");const u=this.next_tokens_are_prompt;if(u&&(this.next_tokens_are_prompt=!1,this.skip_prompt))return;const f=d[0];(k=this.token_callback_function)==null||k.call(this,f),this.token_cache=(0,s.mergeArrays)(this.token_cache,f);const _=this.tokenizer.decode(this.token_cache,this.decode_kwargs);let b;u||_.endsWith(`
`)?(b=_.slice(this.print_len),this.token_cache=[],this.print_len=0):_.length>0&&(0,o.is_chinese_char)(_.charCodeAt(_.length-1))?(b=_.slice(this.print_len),this.print_len+=b.length):(b=_.slice(this.print_len,_.lastIndexOf(" ")+1),this.print_len+=b.length),this.on_finalized_text(b,!1)}end(){let d;this.token_cache.length>0?(d=this.tokenizer.decode(this.token_cache,this.decode_kwargs).slice(this.print_len),this.token_cache=[],this.print_len=0):d="",this.next_tokens_are_prompt=!0,this.on_finalized_text(d,!0)}on_finalized_text(d,u){var f,_;d.length>0&&((f=this.callback_function)==null||f.call(this,d)),u&&this.callback_function===i&&n.apis.IS_PROCESS_AVAILABLE&&((_=this.callback_function)==null||_.call(this,`
`))}}class c extends l{constructor(d,{skip_prompt:u=!1,callback_function:f=null,token_callback_function:_=null,on_chunk_start:b=null,on_chunk_end:k=null,on_finalize:w=null,time_precision:y=.02,skip_special_tokens:$=!0,decode_kwargs:T={}}={}){super(d,{skip_prompt:u,skip_special_tokens:$,callback_function:f,token_callback_function:_,decode_kwargs:T}),this.timestamp_begin=d.timestamp_begin,this.on_chunk_start=b,this.on_chunk_end=k,this.on_finalize=w,this.time_precision=y,this.waiting_for_timestamp=!1}put(d){var f,_,b;if(d.length>1)throw Error("WhisperTextStreamer only supports batch size of 1");const u=d[0];if(u.length===1){const k=Number(u[0])-this.timestamp_begin;if(k>=0){const w=k*this.time_precision;this.waiting_for_timestamp?(f=this.on_chunk_end)==null||f.call(this,w):(_=this.on_chunk_start)==null||_.call(this,w),this.waiting_for_timestamp=!this.waiting_for_timestamp,(b=this.token_callback_function)==null||b.call(this,u);return}}return super.put(d)}end(){var d;super.end(),(d=this.on_finalize)==null||d.call(this)}}},"./src/models.js":(e,r,t)=>{t.r(r),t.d(r,{ASTForAudioClassification:()=>La,ASTModel:()=>Da,ASTPreTrainedModel:()=>Xn,AlbertForMaskedLM:()=>tt,AlbertForQuestionAnswering:()=>qe,AlbertForSequenceClassification:()=>Xe,AlbertModel:()=>Ge,AlbertPreTrainedModel:()=>Te,ArceeForCausalLM:()=>Ja,ArceeModel:()=>Ys,ArceePreTrainedModel:()=>Er,AutoModel:()=>Bu,AutoModelForAudioClassification:()=>Iv,AutoModelForAudioFrameClassification:()=>Fv,AutoModelForAudioTextToText:()=>Vv,AutoModelForCTC:()=>kv,AutoModelForCausalLM:()=>Mv,AutoModelForDepthEstimation:()=>zv,AutoModelForDocumentQuestionAnswering:()=>Ov,AutoModelForImageClassification:()=>xv,AutoModelForImageFeatureExtraction:()=>jv,AutoModelForImageMatting:()=>Dv,AutoModelForImageSegmentation:()=>Tv,AutoModelForImageTextToText:()=>Nv,AutoModelForImageToImage:()=>Lv,AutoModelForMaskGeneration:()=>$v,AutoModelForMaskedLM:()=>bv,AutoModelForNormalEstimation:()=>Bv,AutoModelForObjectDetection:()=>Cv,AutoModelForPoseEstimation:()=>Rv,AutoModelForQuestionAnswering:()=>yv,AutoModelForSemanticSegmentation:()=>Ev,AutoModelForSeq2SeqLM:()=>_v,AutoModelForSequenceClassification:()=>mv,AutoModelForSpeechSeq2Seq:()=>fv,AutoModelForTextToSpectrogram:()=>gv,AutoModelForTextToWaveform:()=>wv,AutoModelForTokenClassification:()=>hv,AutoModelForUniversalSegmentation:()=>Pv,AutoModelForVision2Seq:()=>vv,AutoModelForXVector:()=>Av,AutoModelForZeroShotObjectDetection:()=>Sv,BartForConditionalGeneration:()=>rr,BartForSequenceClassification:()=>Tr,BartModel:()=>ss,BartPretrainedModel:()=>Cr,BaseModelOutput:()=>xe,BeitForImageClassification:()=>AM,BeitModel:()=>IM,BeitPreTrainedModel:()=>Xc,BertForMaskedLM:()=>$e,BertForQuestionAnswering:()=>we,BertForSequenceClassification:()=>Me,BertForTokenClassification:()=>Le,BertModel:()=>Ce,BertPreTrainedModel:()=>be,BlenderbotForConditionalGeneration:()=>wr,BlenderbotModel:()=>_r,BlenderbotPreTrainedModel:()=>dr,BlenderbotSmallForConditionalGeneration:()=>Xs,BlenderbotSmallModel:()=>qr,BlenderbotSmallPreTrainedModel:()=>or,BloomForCausalLM:()=>eM,BloomModel:()=>Zw,BloomPreTrainedModel:()=>Bc,CLIPModel:()=>Lo,CLIPPreTrainedModel:()=>Cs,CLIPSegForImageSegmentation:()=>to,CLIPSegModel:()=>An,CLIPSegPreTrainedModel:()=>Zr,CLIPTextModel:()=>qa,CLIPTextModelWithProjection:()=>zo,CLIPVisionModel:()=>Qa,CLIPVisionModelWithProjection:()=>Bo,CamembertForMaskedLM:()=>ue,CamembertForQuestionAnswering:()=>ot,CamembertForSequenceClassification:()=>Ee,CamembertForTokenClassification:()=>et,CamembertModel:()=>X,CamembertPreTrainedModel:()=>W,CausalLMOutput:()=>dn,CausalLMOutputWithPast:()=>ZT,ChineseCLIPModel:()=>mt,ChineseCLIPPreTrainedModel:()=>kn,ClapAudioModelWithProjection:()=>Qy,ClapModel:()=>Hy,ClapPreTrainedModel:()=>oi,ClapTextModelWithProjection:()=>qy,CodeGenForCausalLM:()=>V,CodeGenModel:()=>z,CodeGenPreTrainedModel:()=>A,CohereForCausalLM:()=>Aw,CohereModel:()=>Iw,CoherePreTrainedModel:()=>Sc,ConvBertForMaskedLM:()=>gt,ConvBertForQuestionAnswering:()=>Yt,ConvBertForSequenceClassification:()=>Is,ConvBertForTokenClassification:()=>As,ConvBertModel:()=>Ks,ConvBertPreTrainedModel:()=>ls,ConvNextForImageClassification:()=>Pb,ConvNextModel:()=>Eb,ConvNextPreTrainedModel:()=>du,ConvNextV2ForImageClassification:()=>Sb,ConvNextV2Model:()=>Cb,ConvNextV2PreTrainedModel:()=>pu,DFineForObjectDetection:()=>GM,DFineModel:()=>WM,DFinePreTrainedModel:()=>ru,DINOv3ConvNextModel:()=>Lb,DINOv3ConvNextPreTrainedModel:()=>Db,DINOv3ViTModel:()=>Ob,DINOv3ViTPreTrainedModel:()=>Fb,DPTForDepthEstimation:()=>ib,DPTModel:()=>ab,DPTPreTrainedModel:()=>lu,DacDecoderModel:()=>z0,DacDecoderOutput:()=>O0,DacEncoderModel:()=>L0,DacEncoderOutput:()=>F0,DacModel:()=>D0,DacPreTrainedModel:()=>pi,DebertaForMaskedLM:()=>lt,DebertaForQuestionAnswering:()=>Dt,DebertaForSequenceClassification:()=>Ot,DebertaForTokenClassification:()=>at,DebertaModel:()=>St,DebertaPreTrainedModel:()=>rt,DebertaV2ForMaskedLM:()=>br,DebertaV2ForQuestionAnswering:()=>Os,DebertaV2ForSequenceClassification:()=>Rr,DebertaV2ForTokenClassification:()=>Ts,DebertaV2Model:()=>Br,DebertaV2PreTrainedModel:()=>Wt,DecisionTransformerModel:()=>f0,DecisionTransformerPreTrainedModel:()=>_0,DeiTForImageClassification:()=>XM,DeiTModel:()=>QM,DeiTPreTrainedModel:()=>nu,DepthAnythingForDepthEstimation:()=>cb,DepthAnythingPreTrainedModel:()=>lb,DepthProForDepthEstimation:()=>hb,DepthProPreTrainedModel:()=>mb,DetrForObjectDetection:()=>OM,DetrForSegmentation:()=>Jc,DetrModel:()=>FM,DetrObjectDetectionOutput:()=>Yc,DetrPreTrainedModel:()=>Ya,DetrSegmentationOutput:()=>DM,Dinov2ForImageClassification:()=>kb,Dinov2Model:()=>$b,Dinov2PreTrainedModel:()=>mu,Dinov2WithRegistersForImageClassification:()=>Ab,Dinov2WithRegistersModel:()=>Ib,Dinov2WithRegistersPreTrainedModel:()=>hu,DistilBertForMaskedLM:()=>ur,DistilBertForQuestionAnswering:()=>Hr,DistilBertForSequenceClassification:()=>cs,DistilBertForTokenClassification:()=>jr,DistilBertModel:()=>Hs,DistilBertPreTrainedModel:()=>gr,DonutSwinModel:()=>Tb,DonutSwinPreTrainedModel:()=>xb,EdgeTamModel:()=>Hb,EfficientNetForImageClassification:()=>r0,EfficientNetModel:()=>t0,EfficientNetPreTrainedModel:()=>Su,ElectraForMaskedLM:()=>Fs,ElectraForQuestionAnswering:()=>R,ElectraForSequenceClassification:()=>S,ElectraForTokenClassification:()=>H,ElectraModel:()=>$t,ElectraPreTrainedModel:()=>Jr,Ernie4_5ForCausalLM:()=>Vy,Ernie4_5Model:()=>Ny,Ernie4_5PreTrainedModel:()=>xu,EsmForMaskedLM:()=>Ds,EsmForSequenceClassification:()=>Ls,EsmForTokenClassification:()=>zs,EsmModel:()=>Es,EsmPreTrainedModel:()=>mr,ExaoneForCausalLM:()=>bw,ExaoneModel:()=>Mw,ExaonePreTrainedModel:()=>vc,FalconForCausalLM:()=>Ky,FalconModel:()=>Gy,FalconPreTrainedModel:()=>Eu,FastViTForImageClassification:()=>bM,FastViTModel:()=>MM,FastViTPreTrainedModel:()=>Gc,Florence2ForConditionalGeneration:()=>Va,Florence2PreTrainedModel:()=>Na,GLPNForDepthEstimation:()=>vb,GLPNModel:()=>yb,GLPNPreTrainedModel:()=>uu,GPT2LMHeadModel:()=>Wo,GPT2Model:()=>Uo,GPT2PreTrainedModel:()=>ro,GPTBigCodeForCausalLM:()=>E,GPTBigCodeModel:()=>h,GPTBigCodePreTrainedModel:()=>ao,GPTJForCausalLM:()=>Xo,GPTJModel:()=>Qo,GPTJPreTrainedModel:()=>oo,GPTNeoForCausalLM:()=>no,GPTNeoModel:()=>On,GPTNeoPreTrainedModel:()=>so,GPTNeoXForCausalLM:()=>qo,GPTNeoXModel:()=>Ho,GPTNeoXPreTrainedModel:()=>Dn,Gemma2ForCausalLM:()=>Lw,Gemma2Model:()=>Dw,Gemma2PreTrainedModel:()=>kc,Gemma3ForCausalLM:()=>jw,Gemma3Model:()=>Rw,Gemma3PreTrainedModel:()=>Ac,Gemma3nForConditionalGeneration:()=>ln,Gemma3nPreTrainedModel:()=>Ka,GemmaForCausalLM:()=>Ow,GemmaModel:()=>Fw,GemmaPreTrainedModel:()=>$c,GlmForCausalLM:()=>ww,GlmModel:()=>gw,GlmPreTrainedModel:()=>yc,GraniteForCausalLM:()=>Sw,GraniteModel:()=>Cw,GraniteMoeHybridForCausalLM:()=>kw,GraniteMoeHybridModel:()=>$w,GraniteMoeHybridPreTrainedModel:()=>Cc,GranitePreTrainedModel:()=>Pc,GroundingDinoForObjectDetection:()=>Bb,GroundingDinoPreTrainedModel:()=>zb,GroupViTModel:()=>wM,GroupViTPreTrainedModel:()=>gM,HeliumForCausalLM:()=>fw,HeliumModel:()=>_w,HeliumPreTrainedModel:()=>bc,HieraForImageClassification:()=>YM,HieraModel:()=>JM,HieraPreTrainedModel:()=>ou,HubertForCTC:()=>by,HubertForSequenceClassification:()=>yy,HubertModel:()=>My,HubertPreTrainedModel:()=>RT,IJepaForImageClassification:()=>lM,IJepaModel:()=>iM,IJepaPreTrainedModel:()=>Vc,Idefics3ForConditionalGeneration:()=>Sn,Idefics3PreTrainedModel:()=>cn,ImageMattingOutput:()=>Wv,JAISLMHeadModel:()=>Ko,JAISModel:()=>Go,JAISPreTrainedModel:()=>Fn,JinaCLIPModel:()=>No,JinaCLIPPreTrainedModel:()=>eo,JinaCLIPTextModel:()=>Vo,JinaCLIPVisionModel:()=>In,Lfm2ForCausalLM:()=>pw,Lfm2Model:()=>dw,Lfm2PreTrainedModel:()=>wc,LiteWhisperForConditionalGeneration:()=>za,Llama4ForCausalLM:()=>it,Llama4PreTrainedModel:()=>Je,LlamaForCausalLM:()=>Re,LlamaModel:()=>Pe,LlamaPreTrainedModel:()=>ce,LlavaForConditionalGeneration:()=>Cn,LlavaOnevisionForConditionalGeneration:()=>Ra,LlavaPreTrainedModel:()=>Pn,LlavaQwen2ForCausalLM:()=>Oo,LongT5ForConditionalGeneration:()=>yr,LongT5Model:()=>Jt,LongT5PreTrainedModel:()=>nr,M2M100ForConditionalGeneration:()=>Yb,M2M100Model:()=>Jb,M2M100PreTrainedModel:()=>gu,MBartForCausalLM:()=>ds,MBartForConditionalGeneration:()=>Ps,MBartForSequenceClassification:()=>Fr,MBartModel:()=>Yr,MBartPreTrainedModel:()=>Ar,MPNetForMaskedLM:()=>Bs,MPNetForQuestionAnswering:()=>he,MPNetForSequenceClassification:()=>Qs,MPNetForTokenClassification:()=>En,MPNetModel:()=>qs,MPNetPreTrainedModel:()=>qt,MT5ForConditionalGeneration:()=>xr,MT5Model:()=>hr,MT5PreTrainedModel:()=>vr,MarianMTModel:()=>Xb,MarianModel:()=>Qb,MarianPreTrainedModel:()=>fu,MaskFormerForInstanceSegmentation:()=>bb,MaskFormerModel:()=>Mb,MaskFormerPreTrainedModel:()=>cu,MaskedLMOutput:()=>Or,Metric3DForDepthEstimation:()=>fb,Metric3DPreTrainedModel:()=>_b,Metric3Dv2ForDepthEstimation:()=>wb,Metric3Dv2PreTrainedModel:()=>gb,MgpstrForSceneTextRecognition:()=>y0,MgpstrModelOutput:()=>M0,MgpstrPreTrainedModel:()=>b0,MimiDecoderModel:()=>A0,MimiDecoderOutput:()=>$0,MimiEncoderModel:()=>I0,MimiEncoderOutput:()=>S0,MimiModel:()=>k0,MimiPreTrainedModel:()=>di,Ministral3ForCausalLM:()=>jy,Ministral3Model:()=>Ry,Ministral3PreTrainedModel:()=>vu,MinistralForCausalLM:()=>By,MinistralModel:()=>zy,MinistralPreTrainedModel:()=>yu,Mistral3ForConditionalGeneration:()=>Ga,MistralForCausalLM:()=>Ly,MistralModel:()=>Dy,MistralPreTrainedModel:()=>bu,MobileBertForMaskedLM:()=>je,MobileBertForQuestionAnswering:()=>st,MobileBertForSequenceClassification:()=>Ve,MobileBertModel:()=>Wr,MobileBertPreTrainedModel:()=>us,MobileLLMForCausalLM:()=>vw,MobileLLMModel:()=>yw,MobileLLMPreTrainedModel:()=>xc,MobileNetV1ForImageClassification:()=>n0,MobileNetV1ForSemanticSegmentation:()=>o0,MobileNetV1Model:()=>s0,MobileNetV1PreTrainedModel:()=>ii,MobileNetV2ForImageClassification:()=>i0,MobileNetV2ForSemanticSegmentation:()=>l0,MobileNetV2Model:()=>a0,MobileNetV2PreTrainedModel:()=>li,MobileNetV3ForImageClassification:()=>u0,MobileNetV3ForSemanticSegmentation:()=>d0,MobileNetV3Model:()=>c0,MobileNetV3PreTrainedModel:()=>ci,MobileNetV4ForImageClassification:()=>m0,MobileNetV4ForSemanticSegmentation:()=>h0,MobileNetV4Model:()=>p0,MobileNetV4PreTrainedModel:()=>ui,MobileViTForImageClassification:()=>TM,MobileViTModel:()=>xM,MobileViTPreTrainedModel:()=>Kc,MobileViTV2ForImageClassification:()=>PM,MobileViTV2Model:()=>EM,MobileViTV2PreTrainedModel:()=>Hc,ModelOutput:()=>me,ModernBertDecoderForCausalLM:()=>Mr,ModernBertDecoderModel:()=>kt,ModernBertDecoderPreTrainedModel:()=>Lt,ModernBertForMaskedLM:()=>ze,ModernBertForSequenceClassification:()=>pt,ModernBertForTokenClassification:()=>ft,ModernBertModel:()=>Oe,ModernBertPreTrainedModel:()=>Qe,Moondream1ForConditionalGeneration:()=>ja,MoonshineForConditionalGeneration:()=>Ba,MoonshineModel:()=>gc,MoonshinePreTrainedModel:()=>Yn,MptForCausalLM:()=>rM,MptModel:()=>tM,MptPreTrainedModel:()=>Rc,MultiModalityCausalLM:()=>w0,MultiModalityPreTrainedModel:()=>g0,MusicgenForCausalLM:()=>UT,MusicgenForConditionalGeneration:()=>ku,MusicgenModel:()=>VT,MusicgenPreTrainedModel:()=>$u,NanoChatForCausalLM:()=>Qt,NanoChatModel:()=>Rt,NanoChatPreTrainedModel:()=>xt,NeoBertForMaskedLM:()=>Fe,NeoBertForQuestionAnswering:()=>yt,NeoBertForSequenceClassification:()=>Ze,NeoBertForTokenClassification:()=>We,NeoBertModel:()=>He,NeoBertPreTrainedModel:()=>ke,NomicBertModel:()=>Ut,NomicBertPreTrainedModel:()=>fr,OPTForCausalLM:()=>nM,OPTModel:()=>sM,OPTPreTrainedModel:()=>jc,Olmo2ForCausalLM:()=>Pw,Olmo2Model:()=>Ew,Olmo2PreTrainedModel:()=>Ec,OlmoForCausalLM:()=>Tw,OlmoModel:()=>xw,OlmoPreTrainedModel:()=>Tc,OpenELMForCausalLM:()=>Vw,OpenELMModel:()=>Nw,OpenELMPreTrainedModel:()=>Fc,OwlViTForObjectDetection:()=>SM,OwlViTModel:()=>CM,OwlViTPreTrainedModel:()=>qc,Owlv2ForObjectDetection:()=>kM,Owlv2Model:()=>$M,Owlv2PreTrainedModel:()=>Qc,PaliGemmaForConditionalGeneration:()=>Wa,PaliGemmaPreTrainedModel:()=>Ua,ParakeetForCTC:()=>ny,ParakeetPreTrainedModel:()=>sy,PatchTSMixerForPrediction:()=>E0,PatchTSMixerModel:()=>T0,PatchTSMixerPreTrainedModel:()=>Au,PatchTSTForPrediction:()=>x0,PatchTSTModel:()=>v0,PatchTSTPreTrainedModel:()=>Iu,Phi3ForCausalLM:()=>Yw,Phi3Model:()=>Jw,Phi3PreTrainedModel:()=>zc,Phi3VForCausalLM:()=>$n,Phi3VPreTrainedModel:()=>Ha,PhiForCausalLM:()=>Xw,PhiModel:()=>Qw,PhiPreTrainedModel:()=>Lc,PreTrainedModel:()=>O,PretrainedMixin:()=>Gt,PvtForImageClassification:()=>pM,PvtModel:()=>dM,PvtPreTrainedModel:()=>Uc,PyAnnoteForAudioFrameClassification:()=>ay,PyAnnoteModel:()=>oy,PyAnnotePreTrainedModel:()=>wu,QuestionAnsweringModelOutput:()=>Nr,Qwen2ForCausalLM:()=>Ww,Qwen2Model:()=>Uw,Qwen2PreTrainedModel:()=>Oc,Qwen2VLForConditionalGeneration:()=>qw,Qwen2VLPreTrainedModel:()=>Hw,Qwen3ForCausalLM:()=>Kw,Qwen3Model:()=>Gw,Qwen3PreTrainedModel:()=>Dc,RFDetrForObjectDetection:()=>VM,RFDetrModel:()=>NM,RFDetrObjectDetectionOutput:()=>UM,RFDetrPreTrainedModel:()=>tu,RTDetrForObjectDetection:()=>zM,RTDetrModel:()=>LM,RTDetrObjectDetectionOutput:()=>Jo,RTDetrPreTrainedModel:()=>Zc,RTDetrV2ForObjectDetection:()=>RM,RTDetrV2Model:()=>BM,RTDetrV2ObjectDetectionOutput:()=>jM,RTDetrV2PreTrainedModel:()=>eu,ResNetForImageClassification:()=>eb,ResNetModel:()=>ZM,ResNetPreTrainedModel:()=>au,RoFormerForMaskedLM:()=>Gs,RoFormerForQuestionAnswering:()=>zr,RoFormerForSequenceClassification:()=>nn,RoFormerForTokenClassification:()=>cr,RoFormerModel:()=>Ws,RoFormerPreTrainedModel:()=>Ur,RobertaForMaskedLM:()=>Ta,RobertaForQuestionAnswering:()=>Pa,RobertaForSequenceClassification:()=>$o,RobertaForTokenClassification:()=>Ea,RobertaModel:()=>xa,RobertaPreTrainedModel:()=>Rs,Sam2ImageSegmentationOutput:()=>Gb,Sam2Model:()=>ti,Sam2PreTrainedModel:()=>Kb,Sam3TrackerModel:()=>qb,SamImageSegmentationOutput:()=>Wb,SamModel:()=>Ub,SamPreTrainedModel:()=>Vb,SapiensForDepthEstimation:()=>db,SapiensForNormalEstimation:()=>pb,SapiensForSemanticSegmentation:()=>ub,SapiensPreTrainedModel:()=>ei,SegformerForImageClassification:()=>Jy,SegformerForSemanticSegmentation:()=>Yy,SegformerModel:()=>NT,SegformerPreTrainedModel:()=>ai,Seq2SeqLMOutput:()=>YT,SequenceClassifierOutput:()=>Et,SiglipModel:()=>Xa,SiglipPreTrainedModel:()=>Zn,SiglipTextModel:()=>Ro,SiglipVisionModel:()=>jo,SmolLM3ForCausalLM:()=>hw,SmolLM3Model:()=>mw,SmolLM3PreTrainedModel:()=>Mc,SmolVLMForConditionalGeneration:()=>Do,SnacDecoderModel:()=>j0,SnacEncoderModel:()=>R0,SnacModel:()=>B0,SnacPreTrainedModel:()=>mi,SpeechT5ForSpeechToText:()=>$y,SpeechT5ForTextToSpeech:()=>ky,SpeechT5HifiGan:()=>Iy,SpeechT5Model:()=>jT,SpeechT5PreTrainedModel:()=>ni,SqueezeBertForMaskedLM:()=>Y,SqueezeBertForQuestionAnswering:()=>de,SqueezeBertForSequenceClassification:()=>ie,SqueezeBertModel:()=>N,SqueezeBertPreTrainedModel:()=>I,StableLmForCausalLM:()=>e0,StableLmModel:()=>Zy,StableLmPreTrainedModel:()=>Cu,Starcoder2ForCausalLM:()=>Wy,Starcoder2Model:()=>Uy,Starcoder2PreTrainedModel:()=>Tu,StyleTextToSpeech2Model:()=>Sy,StyleTextToSpeech2PreTrainedModel:()=>Cy,SupertonicForConditionalGeneration:()=>Mu,SupertonicPreTrainedModel:()=>Ay,Swin2SRForImageSuperResolution:()=>ob,Swin2SRModel:()=>nb,Swin2SRPreTrainedModel:()=>iu,SwinForImageClassification:()=>rb,SwinForSemanticSegmentation:()=>sb,SwinModel:()=>tb,SwinPreTrainedModel:()=>Za,T5ForConditionalGeneration:()=>Bt,T5Model:()=>jt,T5PreTrainedModel:()=>bt,TableTransformerForObjectDetection:()=>HM,TableTransformerModel:()=>KM,TableTransformerObjectDetectionOutput:()=>qM,TableTransformerPreTrainedModel:()=>su,TokenClassifierOutput:()=>Sr,TrOCRForCausalLM:()=>Oy,TrOCRPreTrainedModel:()=>Fy,UltravoxModel:()=>Fu,UltravoxPreTrainedModel:()=>P0,UniSpeechForCTC:()=>uy,UniSpeechForSequenceClassification:()=>dy,UniSpeechModel:()=>cy,UniSpeechPreTrainedModel:()=>ri,UniSpeechSatForAudioFrameClassification:()=>_y,UniSpeechSatForCTC:()=>my,UniSpeechSatForSequenceClassification:()=>hy,UniSpeechSatModel:()=>py,UniSpeechSatPreTrainedModel:()=>Yo,VaultGemmaForCausalLM:()=>Bw,VaultGemmaModel:()=>zw,VaultGemmaPreTrainedModel:()=>Ic,ViTForImageClassification:()=>aM,ViTMAEModel:()=>hM,ViTMAEPreTrainedModel:()=>mM,ViTMSNForImageClassification:()=>fM,ViTMSNModel:()=>_M,ViTMSNPreTrainedModel:()=>Wc,ViTModel:()=>oM,ViTPreTrainedModel:()=>Nc,VisionEncoderDecoderModel:()=>Fo,VitMatteForImageMatting:()=>vM,VitMattePreTrainedModel:()=>yM,VitPoseForPoseEstimation:()=>uM,VitPosePreTrainedModel:()=>cM,VitsModel:()=>Pu,VitsModelOutput:()=>Gv,VitsPreTrainedModel:()=>Xy,VoxtralForConditionalGeneration:()=>C0,Wav2Vec2BertForCTC:()=>gy,Wav2Vec2BertForSequenceClassification:()=>wy,Wav2Vec2BertModel:()=>fy,Wav2Vec2BertPreTrainedModel:()=>si,Wav2Vec2ForAudioFrameClassification:()=>ry,Wav2Vec2ForCTC:()=>ey,Wav2Vec2ForSequenceClassification:()=>ty,Wav2Vec2Model:()=>Zb,Wav2Vec2PreTrainedModel:()=>un,WavLMForAudioFrameClassification:()=>Py,WavLMForCTC:()=>xy,WavLMForSequenceClassification:()=>Ty,WavLMForXVector:()=>Ey,WavLMModel:()=>vy,WavLMPreTrainedModel:()=>io,WeSpeakerResNetModel:()=>ly,WeSpeakerResNetPreTrainedModel:()=>iy,WhisperForConditionalGeneration:()=>Js,WhisperModel:()=>Ao,WhisperPreTrainedModel:()=>Jn,XLMForQuestionAnswering:()=>Ia,XLMForSequenceClassification:()=>$a,XLMForTokenClassification:()=>ka,XLMModel:()=>Ca,XLMPreTrainedModel:()=>on,XLMRobertaForMaskedLM:()=>Fa,XLMRobertaForQuestionAnswering:()=>Io,XLMRobertaForSequenceClassification:()=>Oa,XLMRobertaForTokenClassification:()=>ko,XLMRobertaModel:()=>Aa,XLMRobertaPreTrainedModel:()=>an,XLMWithLMHeadModel:()=>Sa,XVectorOutput:()=>Uv,YolosForObjectDetection:()=>jb,YolosModel:()=>Rb,YolosObjectDetectionOutput:()=>Nb,YolosPreTrainedModel:()=>_u});var s=t("./src/configs.js"),o=t("./src/backends/onnx.js"),n=t("./src/utils/dtypes.js"),a=t("./src/utils/generic.js"),i=t("./src/utils/core.js"),l=t("./src/utils/hub.js"),c=t("./src/utils/constants.js"),p=t("./src/generation/logits_process.js"),d=t("./src/generation/configuration_utils.js"),u=t("./src/utils/tensor.js"),f=t("./src/utils/image.js"),_=t("./src/utils/maths.js"),b=t("./src/generation/stopping_criteria.js"),k=t("./src/generation/logits_sampler.js"),w=t("./src/env.js"),y=t("./src/models/whisper/generation_whisper.js"),$=t("./src/models/whisper/common_whisper.js");const T={EncoderOnly:0,EncoderDecoder:1,Seq2Seq:2,Vision2Seq:3,DecoderOnly:4,MaskGeneration:5,ImageTextToText:6,Musicgen:7,MultiModality:8,Phi3V:9,AudioTextToText:10,AutoEncoder:11,ImageAudioTextToText:12,Supertonic:13},M=new Map,x=new Map,v=new Map;async function C(g,P,L){var Dr;let ne=((Dr=L.config)==null?void 0:Dr["transformers.js_config"])??{},ge=L.device??ne.device;ge&&typeof ge!="string"&&(ge.hasOwnProperty(P)?ge=ge[P]:(console.warn(`device not specified for "${P}". Using the default device.`),ge=null));const fe=ge??(w.apis.IS_NODE_ENV?"cpu":"wasm"),ve=(0,o.deviceToExecutionProviders)(fe),Be=ne.device_config??{};Be.hasOwnProperty(fe)&&(ne={...ne,...Be[fe]});let Ke=L.dtype??ne.dtype;if(typeof Ke!="string"&&(Ke&&Ke.hasOwnProperty(P)?Ke=Ke[P]:(Ke=n.DEFAULT_DEVICE_DTYPE_MAPPING[fe]??n.DATA_TYPES.fp32,console.warn(`dtype not specified for "${P}". Using the default dtype (${Ke}) for this device (${fe}).`))),Ke===n.DATA_TYPES.auto){let At=ne.dtype;typeof At!="string"&&(At=At==null?void 0:At[P]),At&&At!==n.DATA_TYPES.auto&&n.DATA_TYPES.hasOwnProperty(At)?Ke=At:Ke=n.DEFAULT_DEVICE_DTYPE_MAPPING[fe]??n.DATA_TYPES.fp32}const Ye=Ke;if(n.DEFAULT_DTYPE_SUFFIX_MAPPING.hasOwnProperty(Ye)){if(Ye===n.DATA_TYPES.fp16&&fe==="webgpu"&&!await(0,n.isWebGpuFp16Supported)())throw new Error(`The device (${fe}) does not support fp16.`)}else throw new Error(`Invalid dtype: ${Ye}. Should be one of: ${Object.keys(n.DATA_TYPES).join(", ")}`);const ht=ne.kv_cache_dtype,wt=ht?typeof ht=="string"?ht:ht[Ye]??"float32":void 0;if(wt&&!["float32","float16"].includes(wt))throw new Error(`Invalid kv_cache_dtype: ${wt}. Should be one of: float32, float16`);const vt={dtype:Ye,kv_cache_dtype:wt,device:fe},Ft=n.DEFAULT_DTYPE_SUFFIX_MAPPING[Ye],_t=`${P}${Ft}.onnx`,Tt=`${L.subfolder??""}/${_t}`,ut={...L.session_options};ut.executionProviders??(ut.executionProviders=ve);const It=ne.free_dimension_overrides;It?ut.freeDimensionOverrides??(ut.freeDimensionOverrides=It):fe.startsWith("webnn")&&!ut.freeDimensionOverrides&&console.warn(`WebNN does not currently support dynamic shapes and requires 'free_dimension_overrides' to be set in config.json, preferably as a field within config["transformers.js_config"]["device_config"]["${fe}"]. When 'free_dimension_overrides' is not set, you may experience significant performance degradation.`);const Nt=w.apis.IS_NODE_ENV&&w.env.useFSCache,er=(0,l.getModelFile)(g,Tt,!0,L,Nt),sr=L.use_external_data_format??ne.use_external_data_format;let ar=[];if(sr){let At;typeof sr=="object"?sr.hasOwnProperty(_t)?At=sr[_t]:sr.hasOwnProperty(P)?At=sr[P]:At=!1:At=sr;const $r=+At;if($r>l.MAX_EXTERNAL_DATA_CHUNKS)throw new Error(`The number of external data chunks (${$r}) exceeds the maximum allowed value (${l.MAX_EXTERNAL_DATA_CHUNKS}).`);for(let ns=0;ns<$r;++ns){const Ln=`${_t}_data${ns===0?"":"_"+ns}`,Qr=`${L.subfolder??""}/${Ln}`;ar.push(new Promise(async(ps,hi)=>{const zn=await(0,l.getModelFile)(g,Qr,!0,L,Nt);ps(zn instanceof Uint8Array?{path:Ln,data:zn}:Ln)}))}}else ut.externalData!==void 0&&(ar=ut.externalData.map(async At=>{if(typeof At.data=="string"){const $r=await(0,l.getModelFile)(g,At.data,!0,L);return{...At,data:$r}}return At}));if(ar.length>0){const At=await Promise.all(ar);w.apis.IS_NODE_ENV||(ut.externalData=At)}if(fe==="webgpu"){const At=(0,s.getCacheShapes)(L.config,{prefix:"present"});if(Object.keys(At).length>0&&!(0,o.isONNXProxy)()){const $r={};for(const ns in At)$r[ns]="gpu-buffer";ut.preferredOutputLocation=$r}}return{buffer_or_path:await er,session_options:ut,session_config:vt}}async function F(g,P,L){return Object.fromEntries(await Promise.all(Object.keys(P).map(async ne=>{const{buffer_or_path:ge,session_options:fe,session_config:ve}=await C(g,P[ne],L),Be=await(0,o.createInferenceSession)(ge,fe,ve);return[ne,Be]})))}async function D(g,P,L){return Object.fromEntries(await Promise.all(Object.keys(P).map(async ne=>{const ge=await(0,l.getModelJSON)(g,P[ne],!1,L);return[ne,ge]})))}function K(g,P){const L=Object.create(null),ne=[];for(const ve of g.inputNames){const Be=P[ve];if(!(Be instanceof u.Tensor)){ne.push(ve);continue}L[ve]=(0,o.isONNXProxy)()?Be.clone():Be}if(ne.length>0)throw new Error(`An error occurred during model execution: "Missing the following inputs: ${ne.join(", ")}.`);const ge=Object.keys(P).length,fe=g.inputNames.length;if(ge>fe){let ve=Object.keys(P).filter(Be=>!g.inputNames.includes(Be));console.warn(`WARNING: Too many inputs were provided (${ge} > ${fe}). The following inputs will be ignored: "${ve.join(", ")}".`)}return L}async function U(g,P){const L=K(g,P);try{const ne=Object.fromEntries(Object.entries(L).map(([fe,ve])=>[fe,ve.ort_tensor])),ge=await(0,o.runInferenceSession)(g,ne);return j(ge)}catch(ne){const ge=Object.fromEntries(Object.entries(L).map(([fe,ve])=>{const Be={type:ve.type,dims:ve.dims,location:ve.location};return Be.location!=="gpu-buffer"&&(Be.data=ve.data),[fe,Be]}));throw console.error(`An error occurred during model execution: "${ne}".`),console.error("Inputs given to model:",ge),ne}}function j(g){for(let P in g)(0,o.isONNXTensor)(g[P])?g[P]=new u.Tensor(g[P]):typeof g[P]=="object"&&j(g[P]);return g}function se(g){if(g instanceof u.Tensor)return g;if(g.length===0)throw Error("items must be non-empty");if(Array.isArray(g[0])){if(g.some(P=>P.length!==g[0].length))throw Error("Unable to create tensor, you should probably activate truncation and/or padding with 'padding=True' and/or 'truncation=True' to have batched tensors with the same length.");return new u.Tensor("int64",BigInt64Array.from(g.flat().map(P=>BigInt(P))),[g.length,g[0].length])}else return new u.Tensor("int64",BigInt64Array.from(g.map(P=>BigInt(P))),[1,g.length])}function q(g){return new u.Tensor("bool",[g],[1])}async function re(g,P){let{encoder_outputs:L,input_ids:ne,decoder_input_ids:ge,...fe}=P;if(!L){const Be=(0,i.pick)(P,g.sessions.model.inputNames);L=(await Z(g,Be)).last_hidden_state}return fe.input_ids=ge,fe.encoder_hidden_states=L,g.sessions.decoder_model_merged.inputNames.includes("encoder_attention_mask")&&(fe.encoder_attention_mask=P.attention_mask),await _e(g,fe,!0)}async function Z(g,P){const L=g.sessions.model,ne=(0,i.pick)(P,L.inputNames);if(L.inputNames.includes("inputs_embeds")&&!ne.inputs_embeds){if(!P.input_ids)throw new Error("Both `input_ids` and `inputs_embeds` are missing in the model inputs.");ne.inputs_embeds=await g.encode_text({input_ids:P.input_ids})}if(L.inputNames.includes("token_type_ids")&&!ne.token_type_ids){if(!ne.input_ids)throw new Error("Both `input_ids` and `token_type_ids` are missing in the model inputs.");ne.token_type_ids=(0,u.zeros_like)(ne.input_ids)}if(L.inputNames.includes("pixel_mask")&&!ne.pixel_mask){if(!ne.pixel_values)throw new Error("Both `pixel_values` and `pixel_mask` are missing in the model inputs.");const ge=ne.pixel_values.dims;ne.pixel_mask=(0,u.ones)([ge[0],ge[2],ge[3]])}return await U(L,ne)}async function ae(g,P){const L=await g.encode(P);return await g.decode(L)}async function _e(g,P,L=!1){const ne=g.sessions[L?"decoder_model_merged":"model"],{past_key_values:ge,...fe}=P;if(ne.inputNames.includes("use_cache_branch")&&(fe.use_cache_branch=q(!!ge)),ne.inputNames.includes("position_ids")&&fe.attention_mask&&!fe.position_ids){const Be=["paligemma","gemma3_text","gemma3"].includes(g.config.model_type)?1:0;fe.position_ids=Se(fe,ge,Be)}g.addPastKeyValues(fe,ge);const ve=(0,i.pick)(fe,ne.inputNames);return await U(ne,ve)}function pe({modality_token_id:g,inputs_embeds:P,modality_features:L,input_ids:ne,attention_mask:ge}){const fe=ne.tolist().map(Ye=>Ye.reduce((ht,wt,vt)=>(wt==g&&ht.push(vt),ht),[])),ve=fe.reduce((Ye,ht)=>Ye+ht.length,0),Be=L.dims[0];if(ve!==Be)throw new Error(`Number of tokens and features do not match: tokens: ${ve}, features ${Be}`);let Ke=0;for(let Ye=0;Ye<fe.length;++Ye){const ht=fe[Ye],wt=P[Ye];for(let vt=0;vt<ht.length;++vt)wt[ht[vt]].data.set(L[Ke++].data)}return{inputs_embeds:P,attention_mask:ge}}function oe({image_token_id:g,inputs_embeds:P,image_features:L,input_ids:ne,attention_mask:ge}){return pe({modality_token_id:g,inputs_embeds:P,modality_features:L,input_ids:ne,attention_mask:ge})}function G({audio_token_id:g,inputs_embeds:P,audio_features:L,input_ids:ne,attention_mask:ge}){return pe({modality_token_id:g,inputs_embeds:P,modality_features:L,input_ids:ne,attention_mask:ge})}async function B(g,{encode_function:P,merge_function:L,modality_input_name:ne,modality_output_name:ge,input_ids:fe=null,attention_mask:ve=null,position_ids:Be=null,inputs_embeds:Ke=null,past_key_values:Ye=null,generation_config:ht=null,logits_processor:wt=null,...vt}){const Ft=vt[ne];if(!Ke){if(Ke=await g.encode_text({input_ids:fe,...vt}),Ft&&fe.dims[1]!==1){const Tt=await P({[ne]:Ft,...vt});({inputs_embeds:Ke,attention_mask:ve}=L({[ge]:Tt,inputs_embeds:Ke,input_ids:fe,attention_mask:ve}))}else if(Ye&&Ft&&fe.dims[1]===1){const Tt=fe.dims[1],ut=Object.values(Ye)[0].dims.at(-2);ve=(0,u.cat)([(0,u.ones)([fe.dims[0],ut]),ve.slice(null,[ve.dims[1]-Tt,ve.dims[1]])],1)}}if(!Be&&g.config.model_type==="qwen2_vl"){const{image_grid_thw:Tt,video_grid_thw:ut}=vt;[Be]=g.get_rope_index(fe,Tt,ut,ve)}return await _e(g,{inputs_embeds:Ke,past_key_values:Ye,attention_mask:ve,position_ids:Be,generation_config:ht,logits_processor:wt},!0)}async function Q(g,P){return await B(g,{...P,modality_input_name:"audio_values",modality_output_name:"audio_features",encode_function:g.encode_audio.bind(g),merge_function:g._merge_input_ids_with_audio_features.bind(g)})}async function le(g,P){return await B(g,{...P,modality_input_name:"pixel_values",modality_output_name:"image_features",encode_function:g.encode_image.bind(g),merge_function:g._merge_input_ids_with_image_features.bind(g)})}function ee(g,P=0){const[L,ne]=g.dims,ge=g.data,fe=new BigInt64Array(ge.length);for(let ve=0;ve<L;++ve){const Be=ve*ne;let Ke=BigInt(P);for(let Ye=0;Ye<ne;++Ye){const ht=Be+Ye;ge[ht]===0n?fe[ht]=BigInt(1):(fe[ht]=Ke,Ke+=ge[ht])}}return{data:fe,dims:g.dims}}function Se(g,P=null,L=0){const{input_ids:ne,inputs_embeds:ge,attention_mask:fe}=g,{data:ve,dims:Be}=ee(fe,L);let Ke=new u.Tensor("int64",ve,Be);if(P){const Ye=-(ne??ge).dims.at(1);Ke=Ke.slice(null,[Ye,null])}return Ke}function De(g,P,L,ne){const ge=L.past_key_values?Object.values(L.past_key_values)[0].dims.at(-2):0;if(!L.attention_mask){let fe;for(const ve of["input_ids","inputs_embeds","position_ids"])if(L[ve]){fe=L[ve].dims;break}if(!fe)throw new Error("attention_mask is not provided, and unable to infer its shape from model inputs.");L.attention_mask=(0,u.ones)([fe[0],ge+fe[1]])}if(L.past_key_values){const{input_ids:fe,attention_mask:ve}=L;ve&&ve.dims[1]>fe.dims[1]||ge<fe.dims[1]&&(L.input_ids=fe.slice(null,[ge,null]))}return L}function Ne(g,P,L,ne){return L.past_key_values&&(P=P.map(ge=>[ge.at(-1)])),{...L,decoder_input_ids:se(P)}}function Ae(g,...P){return g.config.is_encoder_decoder?Ne(g,...P):De(g,...P)}function te(g,P,L,ne){const ge=!!L.past_key_values;return ne.guidance_scale!==null&&ne.guidance_scale>1&&(ge?L.input_ids=(0,u.cat)([L.input_ids,L.input_ids],0):(L.input_ids=(0,u.cat)([L.input_ids,(0,u.full_like)(L.input_ids,BigInt(ne.pad_token_id))],0),L.attention_mask=(0,u.cat)([L.attention_mask,(0,u.full_like)(L.attention_mask,0n)],0))),(ge||!L.pixel_values)&&(L.pixel_values=(0,u.full)([0,0,3,384,384],1)),ge&&(L.images_seq_mask=new u.Tensor("bool",new Array(1).fill(!0).fill(!1,0,1),[1,1]),L.images_emb_mask=new u.Tensor("bool",new Array(0).fill(!1),[1,1,0])),L}class O extends a.Callable{constructor(L,ne,ge){super();J(this,"main_input_name","input_ids");J(this,"forward_params",["input_ids","attention_mask"]);this.config=L,this.sessions=ne,this.configs=ge;const fe=v.get(this.constructor),ve=M.get(fe);switch(this.can_generate=!1,this._forward=null,this._prepare_inputs_for_generation=null,ve){case T.DecoderOnly:this.can_generate=!0,this._forward=_e,this._prepare_inputs_for_generation=De;break;case T.Seq2Seq:case T.Vision2Seq:case T.Musicgen:this.can_generate=!0,this._forward=re,this._prepare_inputs_for_generation=Ne;break;case T.EncoderDecoder:this._forward=re;break;case T.ImageTextToText:this.can_generate=!0,this._forward=le,this._prepare_inputs_for_generation=Ae;break;case T.AudioTextToText:this.can_generate=!0,this._forward=Q,this._prepare_inputs_for_generation=Ae;break;case T.Phi3V:case T.ImageAudioTextToText:this.can_generate=!0,this._prepare_inputs_for_generation=Ae;break;case T.MultiModality:this.can_generate=!0,this._prepare_inputs_for_generation=te;break;case T.AutoEncoder:this._forward=ae;break;default:this._forward=Z;break}this.can_generate&&this.forward_params.push("past_key_values"),this.custom_config=this.config["transformers.js_config"]??{}}async dispose(){var ne;const L=[];for(const ge of Object.values(this.sessions))(ne=ge==null?void 0:ge.handler)!=null&&ne.dispose&&L.push(ge.handler.dispose());return await Promise.all(L)}static async from_pretrained(L,{progress_callback:ne=null,config:ge=null,cache_dir:fe=null,local_files_only:ve=!1,revision:Be="main",model_file_name:Ke=null,subfolder:Ye="onnx",device:ht=null,dtype:wt=null,use_external_data_format:vt=null,session_options:Ft={}}={}){let _t={progress_callback:ne,config:ge,cache_dir:fe,local_files_only:ve,revision:Be,model_file_name:Ke,subfolder:Ye,device:ht,dtype:wt,use_external_data_format:vt,session_options:Ft};const Tt=v.get(this),ut=M.get(Tt);ge=_t.config=await s.AutoConfig.from_pretrained(L,_t);let It;if(ut===T.DecoderOnly)It=await Promise.all([F(L,{model:_t.model_file_name??"model"},_t),D(L,{generation_config:"generation_config.json"},_t)]);else if(ut===T.Seq2Seq||ut===T.Vision2Seq)It=await Promise.all([F(L,{model:"encoder_model",decoder_model_merged:"decoder_model_merged"},_t),D(L,{generation_config:"generation_config.json"},_t)]);else if(ut===T.MaskGeneration)It=await Promise.all([F(L,{model:"vision_encoder",prompt_encoder_mask_decoder:"prompt_encoder_mask_decoder"},_t)]);else if(ut===T.EncoderDecoder)It=await Promise.all([F(L,{model:"encoder_model",decoder_model_merged:"decoder_model_merged"},_t)]);else if(ut===T.ImageTextToText){const Nt={embed_tokens:"embed_tokens",vision_encoder:"vision_encoder",decoder_model_merged:"decoder_model_merged"};ge.is_encoder_decoder&&(Nt.model="encoder_model"),It=await Promise.all([F(L,Nt,_t),D(L,{generation_config:"generation_config.json"},_t)])}else if(ut===T.AudioTextToText){const Nt={embed_tokens:"embed_tokens",audio_encoder:"audio_encoder",decoder_model_merged:"decoder_model_merged"};It=await Promise.all([F(L,Nt,_t),D(L,{generation_config:"generation_config.json"},_t)])}else if(ut===T.ImageAudioTextToText){const Nt={embed_tokens:"embed_tokens",audio_encoder:"audio_encoder",vision_encoder:"vision_encoder",decoder_model_merged:"decoder_model_merged"};It=await Promise.all([F(L,Nt,_t),D(L,{generation_config:"generation_config.json"},_t)])}else if(ut===T.Musicgen)It=await Promise.all([F(L,{model:"text_encoder",decoder_model_merged:"decoder_model_merged",encodec_decode:"encodec_decode"},_t),D(L,{generation_config:"generation_config.json"},_t)]);else if(ut===T.MultiModality)It=await Promise.all([F(L,{prepare_inputs_embeds:"prepare_inputs_embeds",model:"language_model",lm_head:"lm_head",gen_head:"gen_head",gen_img_embeds:"gen_img_embeds",image_decode:"image_decode"},_t),D(L,{generation_config:"generation_config.json"},_t)]);else if(ut===T.Phi3V)It=await Promise.all([F(L,{prepare_inputs_embeds:"prepare_inputs_embeds",model:"model",vision_encoder:"vision_encoder"},_t),D(L,{generation_config:"generation_config.json"},_t)]);else if(ut===T.AutoEncoder)It=await Promise.all([F(L,{encoder_model:"encoder_model",decoder_model:"decoder_model"},_t)]);else if(ut===T.Supertonic)It=await Promise.all([F(L,{text_encoder:"text_encoder",latent_denoiser:"latent_denoiser",voice_decoder:"voice_decoder"},_t)]);else{if(ut!==T.EncoderOnly){const Nt=Tt??(ge==null?void 0:ge.model_type);Nt!=="custom"&&console.warn(`Model type for '${Nt}' not found, assuming encoder-only architecture. Please report this at ${c.GITHUB_ISSUE_URL}.`)}It=await Promise.all([F(L,{model:_t.model_file_name??"model"},_t)])}return new this(ge,...It)}async _call(L){return await this.forward(L)}async forward(L){return await this._forward(this,L)}get generation_config(){var L;return((L=this.configs)==null?void 0:L.generation_config)??null}_get_logits_processor(L,ne,ge=null){const fe=new p.LogitsProcessorList;if(L.repetition_penalty!==null&&L.repetition_penalty!==1&&fe.push(new p.RepetitionPenaltyLogitsProcessor(L.repetition_penalty)),L.no_repeat_ngram_size!==null&&L.no_repeat_ngram_size>0&&fe.push(new p.NoRepeatNGramLogitsProcessor(L.no_repeat_ngram_size)),L.bad_words_ids!==null&&fe.push(new p.NoBadWordsLogitsProcessor(L.bad_words_ids,L.eos_token_id)),L.min_length!==null&&L.eos_token_id!==null&&L.min_length>0&&fe.push(new p.MinLengthLogitsProcessor(L.min_length,L.eos_token_id)),L.min_new_tokens!==null&&L.eos_token_id!==null&&L.min_new_tokens>0&&fe.push(new p.MinNewTokensLengthLogitsProcessor(ne,L.min_new_tokens,L.eos_token_id)),L.forced_bos_token_id!==null&&fe.push(new p.ForcedBOSTokenLogitsProcessor(L.forced_bos_token_id)),L.forced_eos_token_id!==null&&fe.push(new p.ForcedEOSTokenLogitsProcessor(L.max_length,L.forced_eos_token_id)),L.begin_suppress_tokens!==null){const ve=ne>1||L.forced_bos_token_id===null?ne:ne+1;fe.push(new p.SuppressTokensAtBeginLogitsProcessor(L.begin_suppress_tokens,ve))}return L.guidance_scale!==null&&L.guidance_scale>1&&fe.push(new p.ClassifierFreeGuidanceLogitsProcessor(L.guidance_scale)),L.temperature===0&&L.do_sample&&(console.warn("`do_sample` changed to false because `temperature: 0` implies greedy sampling (always selecting the most likely token), which is incompatible with `do_sample: true`."),L.do_sample=!1),L.do_sample&&L.temperature!==null&&L.temperature!==1&&fe.push(new p.TemperatureLogitsWarper(L.temperature)),ge!==null&&fe.extend(ge),fe}_prepare_generation_config(L,ne,ge=d.GenerationConfig){const fe={...this.config};for(const Be of["decoder","generator","text_config"])Be in fe&&Object.assign(fe,fe[Be]);const ve=new ge(fe);return Object.assign(ve,this.generation_config??{}),L&&Object.assign(ve,L),ne&&Object.assign(ve,(0,i.pick)(ne,Object.getOwnPropertyNames(ve))),ve}_get_stopping_criteria(L,ne=null){const ge=new b.StoppingCriteriaList;return L.max_length!==null&&ge.push(new b.MaxLengthCriteria(L.max_length,this.config.max_position_embeddings??null)),L.eos_token_id!==null&&ge.push(new b.EosTokenCriteria(L.eos_token_id)),ne&&ge.extend(ne),ge}_validate_model_class(){if(!this.can_generate){const L=[Lu,zu,Du,Ou],ne=v.get(this.constructor),ge=new Set,fe=this.config.model_type;for(const Be of L){const Ke=Be.get(fe);Ke&&ge.add(Ke[0])}let ve=`The current model class (${ne}) is not compatible with \`.generate()\`, as it doesn't have a language model head.`;throw ge.size>0&&(ve+=` Please use the following class instead: ${[...ge].join(", ")}`),Error(ve)}}prepare_inputs_for_generation(...L){return this._prepare_inputs_for_generation(this,...L)}_update_model_kwargs_for_generation({generated_input_ids:L,outputs:ne,model_inputs:ge,is_encoder_decoder:fe}){return ge.past_key_values=this.getPastKeyValues(ne,ge.past_key_values),ge.input_ids=new u.Tensor("int64",L.flat(),[L.length,1]),fe||(ge.attention_mask=(0,u.cat)([ge.attention_mask,(0,u.ones)([ge.attention_mask.dims[0],1])],1)),ge.position_ids=null,ge}_prepare_model_inputs({inputs:L,bos_token_id:ne,model_kwargs:ge}){const fe=(0,i.pick)(ge,this.forward_params),ve=this.main_input_name;if(ve in fe){if(L)throw new Error("`inputs`: {inputs}` were passed alongside {input_name} which is not allowed. Make sure to either pass {inputs} or {input_name}=...")}else fe[ve]=L;return{inputs_tensor:fe[ve],model_inputs:fe,model_input_name:ve}}async _prepare_encoder_decoder_kwargs_for_generation({inputs_tensor:L,model_inputs:ne,model_input_name:ge,generation_config:fe}){if(this.sessions.model.inputNames.includes("inputs_embeds")&&!ne.inputs_embeds&&"_prepare_inputs_embeds"in this){const{input_ids:Be,pixel_values:Ke,attention_mask:Ye,...ht}=ne,wt=await this._prepare_inputs_embeds(ne);ne={...ht,...(0,i.pick)(wt,["inputs_embeds","attention_mask"])}}let{last_hidden_state:ve}=await Z(this,ne);if(fe.guidance_scale!==null&&fe.guidance_scale>1)ve=(0,u.cat)([ve,(0,u.full_like)(ve,0)],0),"attention_mask"in ne&&(ne.attention_mask=(0,u.cat)([ne.attention_mask,(0,u.zeros_like)(ne.attention_mask)],0));else if(ne.decoder_input_ids){const Be=se(ne.decoder_input_ids).dims[0];if(Be!==ve.dims[0]){if(ve.dims[0]!==1)throw new Error(`The encoder outputs have a different batch size (${ve.dims[0]}) than the decoder inputs (${Be}).`);ve=(0,u.cat)(Array.from({length:Be},()=>ve),0)}}return ne.encoder_outputs=ve,ne}_prepare_decoder_input_ids_for_generation({batch_size:L,model_input_name:ne,model_kwargs:ge,decoder_start_token_id:fe,bos_token_id:ve,generation_config:Be}){let{decoder_input_ids:Ke,...Ye}=ge;if(!(Ke instanceof u.Tensor)){if(Ke)Array.isArray(Ke[0])||(Ke=Array.from({length:L},()=>Ke));else if(fe??(fe=ve),this.config.model_type==="musicgen")Ke=Array.from({length:L*this.config.decoder.num_codebooks},()=>[fe]);else if(Array.isArray(fe)){if(fe.length!==L)throw new Error(`\`decoder_start_token_id\` expcted to have length ${L} but got ${fe.length}`);Ke=fe}else Ke=Array.from({length:L},()=>[fe]);Ke=se(Ke)}return ge.decoder_attention_mask=(0,u.ones_like)(Ke),{input_ids:Ke,model_inputs:Ye}}async generate({inputs:L=null,generation_config:ne=null,logits_processor:ge=null,stopping_criteria:fe=null,streamer:ve=null,...Be}){this._validate_model_class(),ne=this._prepare_generation_config(ne,Be);let{inputs_tensor:Ke,model_inputs:Ye,model_input_name:ht}=this._prepare_model_inputs({inputs:L,model_kwargs:Be});const wt=this.config.is_encoder_decoder;wt&&("encoder_outputs"in Ye||(Ye=await this._prepare_encoder_decoder_kwargs_for_generation({inputs_tensor:Ke,model_inputs:Ye,model_input_name:ht,generation_config:ne})));let vt;wt?{input_ids:vt,model_inputs:Ye}=this._prepare_decoder_input_ids_for_generation({batch_size:Ye[ht].dims.at(0),model_input_name:ht,model_kwargs:Ye,decoder_start_token_id:ne.decoder_start_token_id,bos_token_id:ne.bos_token_id,generation_config:ne}):vt=Ye[ht];let Ft=vt.dims.at(-1);ne.max_new_tokens!==null&&(ne.max_length=Ft+ne.max_new_tokens);const _t=this._get_logits_processor(ne,Ft,ge),Tt=this._get_stopping_criteria(ne,fe),ut=Ye[ht].dims.at(0),It=k.LogitsSampler.getSampler(ne),Nt=new Array(ut).fill(0),er=vt.tolist();ve&&ve.put(er);let sr,ar={};for(;;){if(Ye=this.prepare_inputs_for_generation(er,Ye,ne),sr=await this.forward(Ye),ne.output_attentions&&ne.return_dict_in_generate){const Qr=this.getAttentions(sr);for(const ps in Qr)ps in ar||(ar[ps]=[]),ar[ps].push(Qr[ps])}const At=sr.logits.slice(null,-1,null),$r=_t(er,At),ns=[];for(let Qr=0;Qr<$r.dims.at(0);++Qr){const ps=$r[Qr],hi=await It(ps);for(const[zn,_i]of hi){const fi=BigInt(zn);Nt[Qr]+=_i,er[Qr].push(fi),ns.push([fi]);break}}if(ve&&ve.put(ns),Tt(er).every(Qr=>Qr))break;Ye=this._update_model_kwargs_for_generation({generated_input_ids:ns,outputs:sr,model_inputs:Ye,is_encoder_decoder:wt})}ve&&ve.end();const Pr=this.getPastKeyValues(sr,Ye.past_key_values,!0),Dr=new u.Tensor("int64",er.flat(),[er.length,er[0].length]);if(ne.return_dict_in_generate)return{sequences:Dr,past_key_values:Pr,...ar};for(const At of Object.values(sr))At.location==="gpu-buffer"&&At.dispose();return Dr}getPastKeyValues(L,ne,ge=!1){const fe=Object.create(null);for(const ve in L)if(ve.startsWith("present")){const Be=ve.replace("present_conv","past_conv").replace("present","past_key_values"),Ke=ve.includes("encoder");if(Ke&&ne?fe[Be]=ne[Be]:fe[Be]=L[ve],ne&&(!Ke||ge)){const Ye=ne[Be];Ye.location==="gpu-buffer"&&Ye.dispose()}}return fe}getAttentions(L){const ne={};for(const ge of["cross_attentions","encoder_attentions","decoder_attentions"])for(const fe in L)fe.startsWith(ge)&&(ge in ne||(ne[ge]=[]),ne[ge].push(L[fe]));return ne}addPastKeyValues(L,ne){var ge,fe,ve;if(ne)Object.assign(L,ne);else{const Be=this.sessions.decoder_model_merged??this.sessions.model,Ke=((fe=(ge=L[this.main_input_name]??L.attention_mask)==null?void 0:ge.dims)==null?void 0:fe[0])??1,Ye=((ve=Be==null?void 0:Be.config)==null?void 0:ve.kv_cache_dtype)??"float32",ht=Ye==="float16"?u.DataTypeMap.float16:u.DataTypeMap.float32,wt=(0,s.getCacheShapes)(this.config,{batch_size:Ke});for(const vt in wt){const Ft=wt[vt].reduce((_t,Tt)=>_t*Tt,1);L[vt]=new u.Tensor(Ye,new ht(Ft),wt[vt])}}}async encode_image({pixel_values:L}){return(await U(this.sessions.vision_encoder,{pixel_values:L})).image_features}async encode_text({input_ids:L}){return(await U(this.sessions.embed_tokens,{input_ids:L})).inputs_embeds}async encode_audio({audio_values:L}){return(await U(this.sessions.audio_encoder,{audio_values:L})).audio_features}}class me{}class xe extends me{constructor({last_hidden_state:P,hidden_states:L=null,attentions:ne=null}){super(),this.last_hidden_state=P,this.hidden_states=L,this.attentions=ne}}class be extends O{}class Ce extends be{}class $e extends be{async _call(P){return new Or(await super._call(P))}}class Me extends be{async _call(P){return new Et(await super._call(P))}}class Le extends be{async _call(P){return new Sr(await super._call(P))}}class we extends be{async _call(P){return new Nr(await super._call(P))}}class ke extends O{}class He extends ke{}class Fe extends ke{async _call(P){return new Or(await super._call(P))}}class Ze extends ke{async _call(P){return new Et(await super._call(P))}}class We extends ke{async _call(P){return new Sr(await super._call(P))}}class yt extends ke{async _call(P){return new Nr(await super._call(P))}}class Qe extends O{}class Oe extends Qe{}class ze extends Qe{async _call(P){return new Or(await super._call(P))}}class pt extends Qe{async _call(P){return new Et(await super._call(P))}}class ft extends Qe{async _call(P){return new Sr(await super._call(P))}}class Lt extends O{}class kt extends Lt{}class Mr extends Lt{}class fr extends O{}class Ut extends fr{}class Ur extends O{}class Ws extends Ur{}class Gs extends Ur{async _call(P){return new Or(await super._call(P))}}class nn extends Ur{async _call(P){return new Et(await super._call(P))}}class cr extends Ur{async _call(P){return new Sr(await super._call(P))}}class zr extends Ur{async _call(P){return new Nr(await super._call(P))}}class ls extends O{}class Ks extends ls{}class gt extends ls{async _call(P){return new Or(await super._call(P))}}class Is extends ls{async _call(P){return new Et(await super._call(P))}}class As extends ls{async _call(P){return new Sr(await super._call(P))}}class Yt extends ls{async _call(P){return new Nr(await super._call(P))}}class Jr extends O{}class $t extends Jr{}class Fs extends Jr{async _call(P){return new Or(await super._call(P))}}class S extends Jr{async _call(P){return new Et(await super._call(P))}}class H extends Jr{async _call(P){return new Sr(await super._call(P))}}class R extends Jr{async _call(P){return new Nr(await super._call(P))}}class W extends O{}class X extends W{}class ue extends W{async _call(P){return new Or(await super._call(P))}}class Ee extends W{async _call(P){return new Et(await super._call(P))}}class et extends W{async _call(P){return new Sr(await super._call(P))}}class ot extends W{async _call(P){return new Nr(await super._call(P))}}class rt extends O{}class St extends rt{}class lt extends rt{async _call(P){return new Or(await super._call(P))}}class Ot extends rt{async _call(P){return new Et(await super._call(P))}}class at extends rt{async _call(P){return new Sr(await super._call(P))}}class Dt extends rt{async _call(P){return new Nr(await super._call(P))}}class Wt extends O{}class Br extends Wt{}class br extends Wt{async _call(P){return new Or(await super._call(P))}}class Rr extends Wt{async _call(P){return new Et(await super._call(P))}}class Ts extends Wt{async _call(P){return new Sr(await super._call(P))}}class Os extends Wt{async _call(P){return new Nr(await super._call(P))}}class gr extends O{}class Hs extends gr{}class cs extends gr{async _call(P){return new Et(await super._call(P))}}class jr extends gr{async _call(P){return new Sr(await super._call(P))}}class Hr extends gr{async _call(P){return new Nr(await super._call(P))}}class ur extends gr{async _call(P){return new Or(await super._call(P))}}class mr extends O{}class Es extends mr{}class Ds extends mr{async _call(P){return new Or(await super._call(P))}}class Ls extends mr{async _call(P){return new Et(await super._call(P))}}class zs extends mr{async _call(P){return new Sr(await super._call(P))}}class us extends O{}class Wr extends us{}class je extends us{async _call(P){return new Or(await super._call(P))}}class Ve extends us{async _call(P){return new Et(await super._call(P))}}class st extends us{async _call(P){return new Nr(await super._call(P))}}class qt extends O{}class qs extends qt{}class Bs extends qt{async _call(P){return new Or(await super._call(P))}}class Qs extends qt{async _call(P){return new Et(await super._call(P))}}class En extends qt{async _call(P){return new Sr(await super._call(P))}}class he extends qt{async _call(P){return new Nr(await super._call(P))}}class I extends O{}class N extends I{}class Y extends I{async _call(P){return new Or(await super._call(P))}}class ie extends I{async _call(P){return new Et(await super._call(P))}}class de extends I{async _call(P){return new Nr(await super._call(P))}}class Te extends O{}class Ge extends Te{}class Xe extends Te{async _call(P){return new Et(await super._call(P))}}class qe extends Te{async _call(P){return new Nr(await super._call(P))}}class tt extends Te{async _call(P){return new Or(await super._call(P))}}class bt extends O{constructor(){super(...arguments);J(this,"forward_params",["input_ids","attention_mask","encoder_outputs","decoder_input_ids","decoder_attention_mask","past_key_values"])}}class jt extends bt{}class Bt extends bt{}class nr extends O{}class Jt extends nr{}class yr extends nr{}class vr extends O{}class hr extends vr{}class xr extends vr{}class Cr extends O{}class ss extends Cr{}class rr extends Cr{}class Tr extends Cr{async _call(P){return new Et(await super._call(P))}}class Ar extends O{}class Yr extends Ar{}class Ps extends Ar{}class Fr extends Ar{async _call(P){return new Et(await super._call(P))}}class ds extends Ar{}class dr extends O{}class _r extends dr{}class wr extends dr{}class or extends O{}class qr extends or{}class Xs extends or{}class Rs extends O{}class xa extends Rs{}class Ta extends Rs{async _call(P){return new Or(await super._call(P))}}class $o extends Rs{async _call(P){return new Et(await super._call(P))}}class Ea extends Rs{async _call(P){return new Sr(await super._call(P))}}class Pa extends Rs{async _call(P){return new Nr(await super._call(P))}}class on extends O{}class Ca extends on{}class Sa extends on{async _call(P){return new Or(await super._call(P))}}class $a extends on{async _call(P){return new Et(await super._call(P))}}class ka extends on{async _call(P){return new Sr(await super._call(P))}}class Ia extends on{async _call(P){return new Nr(await super._call(P))}}class an extends O{}class Aa extends an{}class Fa extends an{async _call(P){return new Or(await super._call(P))}}class Oa extends an{async _call(P){return new Et(await super._call(P))}}class ko extends an{async _call(P){return new Sr(await super._call(P))}}class Io extends an{async _call(P){return new Nr(await super._call(P))}}class Xn extends O{}class Da extends Xn{}class La extends Xn{}class Jn extends O{constructor(){super(...arguments);J(this,"requires_attention_mask",!1);J(this,"main_input_name","input_features");J(this,"forward_params",["input_features","attention_mask","decoder_input_ids","decoder_attention_mask","past_key_values"])}}class Ao extends Jn{}class Js extends Jn{_prepare_generation_config(P,L){return super._prepare_generation_config(P,L,y.WhisperGenerationConfig)}_retrieve_init_tokens(P){const L=[P.decoder_start_token_id];let ne=P.language;const ge=P.task;if(P.is_multilingual){ne||(console.warn("No language specified - defaulting to English (en)."),ne="en");const ve=`<|${(0,$.whisper_language_to_code)(ne)}|>`;L.push(P.lang_to_id[ve]),L.push(P.task_to_id[ge??"transcribe"])}else if(ne||ge)throw new Error("Cannot specify `task` or `language` for an English-only model. If the model is intended to be multilingual, pass `is_multilingual=true` to generate, or update the generation config.");return!P.return_timestamps&&P.no_timestamps_token_id&&L.at(-1)!==P.no_timestamps_token_id?L.push(P.no_timestamps_token_id):P.return_timestamps&&L.at(-1)===P.no_timestamps_token_id&&(console.warn("<|notimestamps|> prompt token is removed from generation_config since `return_timestamps` is set to `true`."),L.pop()),L.filter(fe=>fe!=null)}async generate({inputs:P=null,generation_config:L=null,logits_processor:ne=null,stopping_criteria:ge=null,...fe}){L=this._prepare_generation_config(L,fe);const ve=fe.decoder_input_ids??this._retrieve_init_tokens(L);if(L.return_timestamps&&(ne??(ne=new p.LogitsProcessorList),ne.push(new p.WhisperTimeStampLogitsProcessor(L,ve))),L.begin_suppress_tokens&&(ne??(ne=new p.LogitsProcessorList),ne.push(new p.SuppressTokensAtBeginLogitsProcessor(L.begin_suppress_tokens,ve.length))),L.return_token_timestamps){if(!L.alignment_heads)throw new Error("Model generation config has no `alignment_heads`, token-level timestamps not available. See https://gist.github.com/hollance/42e32852f24243b748ae6bc1f985b13a on how to add this property to the generation config.");L.task==="translate"&&console.warn("Token-level timestamps may not be reliable for task 'translate'."),L.output_attentions=!0,L.return_dict_in_generate=!0}const Be=await super.generate({inputs:P,generation_config:L,logits_processor:ne,decoder_input_ids:ve,...fe});return L.return_token_timestamps&&(Be.token_timestamps=this._extract_token_timestamps(Be,L.alignment_heads,L.num_frames)),Be}_extract_token_timestamps(P,L,ne=null,ge=.02){if(!P.cross_attentions)throw new Error("Model outputs must contain cross attentions to extract timestamps. This is most likely because the model was not exported with `output_attentions=True`.");ne==null&&console.warn("`num_frames` has not been set, meaning the entire audio will be analyzed. This may lead to inaccurate token-level timestamps for short audios (< 30 seconds).");let fe=this.config.median_filter_width;fe===void 0&&(console.warn("Model config has no `median_filter_width`, using default value of 7."),fe=7);const ve=P.cross_attentions,Be=Array.from({length:this.config.decoder_layers},(Tt,ut)=>(0,u.cat)(ve.map(It=>It[ut]),2)),Ke=(0,u.stack)(L.map(([Tt,ut])=>{if(Tt>=Be.length)throw new Error(`Layer index ${Tt} is out of bounds for cross attentions (length ${Be.length}).`);return ne?Be[Tt].slice(null,ut,null,[0,ne]):Be[Tt].slice(null,ut)})).transpose(1,0,2,3),[Ye,ht]=(0,u.std_mean)(Ke,-2,0,!0),wt=Ke.clone();for(let Tt=0;Tt<wt.dims[0];++Tt){const ut=wt[Tt];for(let It=0;It<ut.dims[0];++It){const Nt=ut[It],er=Ye[Tt][It][0].data,sr=ht[Tt][It][0].data;for(let ar=0;ar<Nt.dims[0];++ar){let Pr=Nt[ar].data;for(let Dr=0;Dr<Pr.length;++Dr)Pr[Dr]=(Pr[Dr]-sr[Dr])/er[Dr];Pr.set((0,_.medianFilter)(Pr,fe))}}}const vt=[(0,u.mean)(wt,1)],Ft=P.sequences.dims,_t=new u.Tensor("float32",new Float32Array(Ft[0]*Ft[1]),Ft);for(let Tt=0;Tt<Ft[0];++Tt){const ut=vt[Tt].neg().squeeze_(0),[It,Nt]=(0,_.dynamic_time_warping)(ut.tolist()),er=Array.from({length:It.length-1},(Pr,Dr)=>It[Dr+1]-It[Dr]),sr=(0,i.mergeArrays)([1],er).map(Pr=>!!Pr),ar=[];for(let Pr=0;Pr<sr.length;++Pr)sr[Pr]&&ar.push(Nt[Pr]*ge);_t[Tt].data.set(ar,1)}return _t}}class za extends Js{}class Yn extends O{constructor(){super(...arguments);J(this,"requires_attention_mask",!1);J(this,"main_input_name","input_values");J(this,"forward_params",["input_values","decoder_input_ids","past_key_values"])}}class gc extends Yn{}class Ba extends Yn{}class Fo extends O{constructor(){super(...arguments);J(this,"main_input_name","pixel_values");J(this,"forward_params",["pixel_values","decoder_input_ids","encoder_hidden_states","past_key_values"])}}class Pn extends O{constructor(){super(...arguments);J(this,"forward_params",["input_ids","attention_mask","pixel_values","position_ids","past_key_values"])}}class Cn extends Pn{_merge_input_ids_with_image_features(P){const L=P.image_features.dims.at(-1),ne=P.image_features.view(-1,L);return oe({image_token_id:this.config.image_token_index,...P,image_features:ne})}}class Ra extends Cn{}class ja extends Cn{}class Na extends O{constructor(){super(...arguments);J(this,"forward_params",["input_ids","inputs_embeds","attention_mask","pixel_values","encoder_outputs","decoder_input_ids","decoder_inputs_embeds","decoder_attention_mask","past_key_values"]);J(this,"main_input_name","inputs_embeds")}}class Va extends Na{_merge_input_ids_with_image_features({inputs_embeds:P,image_features:L,input_ids:ne,attention_mask:ge}){return{inputs_embeds:(0,u.cat)([L,P],1),attention_mask:(0,u.cat)([(0,u.ones)(L.dims.slice(0,2)),ge],1)}}async _prepare_inputs_embeds({input_ids:P,pixel_values:L,inputs_embeds:ne,attention_mask:ge}){if(!P&&!L)throw new Error("Either `input_ids` or `pixel_values` should be provided.");let fe,ve;return P&&(fe=await this.encode_text({input_ids:P})),L&&(ve=await this.encode_image({pixel_values:L})),fe&&ve?{inputs_embeds:ne,attention_mask:ge}=this._merge_input_ids_with_image_features({inputs_embeds:fe,image_features:ve,input_ids:P,attention_mask:ge}):ne=fe||ve,{inputs_embeds:ne,attention_mask:ge}}async forward({input_ids:P,pixel_values:L,attention_mask:ne,decoder_input_ids:ge,decoder_attention_mask:fe,encoder_outputs:ve,past_key_values:Be,inputs_embeds:Ke,decoder_inputs_embeds:Ye}){if(Ke||({inputs_embeds:Ke,attention_mask:ne}=await this._prepare_inputs_embeds({input_ids:P,pixel_values:L,inputs_embeds:Ke,attention_mask:ne})),!ve){let{last_hidden_state:vt}=await Z(this,{inputs_embeds:Ke,attention_mask:ne});ve=vt}if(!Ye){if(!ge)throw new Error("Either `decoder_input_ids` or `decoder_inputs_embeds` should be provided.");Ye=await this.encode_text({input_ids:ge})}return await _e(this,{inputs_embeds:Ye,attention_mask:fe,encoder_attention_mask:ne,encoder_hidden_states:ve,past_key_values:Be},!0)}}class Ua extends O{constructor(){super(...arguments);J(this,"forward_params",["input_ids","attention_mask","pixel_values","position_ids","past_key_values"])}}class Wa extends Ua{_merge_input_ids_with_image_features(P){const L=P.image_features.dims.at(-1),ne=P.image_features.view(-1,L);return oe({image_token_id:this.config.image_token_index,...P,image_features:ne})}}class Oo extends Pn{_merge_input_ids_with_image_features(P){const L=P.image_features.dims.at(-1),ne=P.image_features.view(-1,L);return oe({image_token_id:this.config.image_token_index,...P,image_features:ne})}}class Ga extends Oo{}class Ka extends O{constructor(){super(...arguments);J(this,"forward_params",["input_ids","attention_mask","inputs_embeds","per_layer_inputs","position_ids","pixel_values","input_features","input_features_mask","past_key_values"])}}class ln extends Ka{async forward({input_ids:P=null,attention_mask:L=null,pixel_values:ne=null,input_features:ge=null,input_features_mask:fe=null,position_ids:ve=null,inputs_embeds:Be=null,per_layer_inputs:Ke=null,past_key_values:Ye=null,generation_config:ht=null,logits_processor:wt=null,...vt}){if((!Be||!Ke)&&({inputs_embeds:Be,per_layer_inputs:Ke}=await U(this.sessions.embed_tokens,{input_ids:P}),P.dims[1]!==1)){if(ne){const{image_features:_t}=await U(this.sessions.vision_encoder,{pixel_values:ne});({inputs_embeds:Be,attention_mask:L}=this._merge_input_ids_with_image_features({image_features:_t,inputs_embeds:Be,input_ids:P,attention_mask:L}))}if(ge){const{audio_features:_t}=await U(this.sessions.audio_encoder,{input_features:ge,input_features_mask:fe});({inputs_embeds:Be,attention_mask:L}=this._merge_input_ids_with_audio_features({audio_features:_t,inputs_embeds:Be,input_ids:P,attention_mask:L}))}}return await _e(this,{inputs_embeds:Be,per_layer_inputs:Ke,past_key_values:Ye,attention_mask:L,position_ids:ve,generation_config:ht,logits_processor:wt},!0)}_merge_input_ids_with_image_features(P){const L=P.image_features.dims.at(-1),ne=P.image_features.view(-1,L);return oe({image_token_id:this.config.image_token_id,...P,image_features:ne})}_merge_input_ids_with_audio_features(P){const L=P.audio_features.dims.at(-1),ne=P.audio_features.view(-1,L);return G({audio_token_id:this.config.audio_token_id,...P,audio_features:ne})}}class cn extends O{constructor(){super(...arguments);J(this,"forward_params",["input_ids","attention_mask","pixel_values","pixel_attention_mask","position_ids","past_key_values"])}}class Sn extends cn{async encode_image({pixel_values:P,pixel_attention_mask:L}){return(await U(this.sessions.vision_encoder,{pixel_values:P,pixel_attention_mask:L})).image_features}_merge_input_ids_with_image_features(P){const L=P.image_features.dims.at(-1),ne=P.image_features.view(-1,L);return oe({image_token_id:this.config.image_token_id,...P,image_features:ne})}}class Do extends Sn{}class Ha extends O{constructor(){super(...arguments);J(this,"forward_params",["input_ids","inputs_embeds","attention_mask","position_ids","pixel_values","image_sizes","past_key_values"])}}class $n extends Ha{async forward({input_ids:P=null,attention_mask:L=null,pixel_values:ne=null,image_sizes:ge=null,position_ids:fe=null,inputs_embeds:ve=null,past_key_values:Be=null,generation_config:Ke=null,logits_processor:Ye=null,...ht}){if(!ve){let vt;if(ne&&P.dims[1]!==1){if(!ge)throw new Error("`image_sizes` must be provided when `pixel_values` is provided.");({image_features:vt}=await U(this.sessions.vision_encoder,{pixel_values:ne,image_sizes:ge}))}else{const Ft=this.config.normalized_config.hidden_size;vt=new u.Tensor("float32",[],[0,Ft])}({inputs_embeds:ve}=await U(this.sessions.prepare_inputs_embeds,{input_ids:P,image_features:vt}))}return await _e(this,{inputs_embeds:ve,past_key_values:Be,attention_mask:L,position_ids:fe,generation_config:Ke,logits_processor:Ye},!1)}}class Cs extends O{}class Lo extends Cs{}class qa extends Cs{static async from_pretrained(P,L={}){return super.from_pretrained(P,{...L,model_file_name:L.model_file_name??"text_model"})}}class zo extends Cs{static async from_pretrained(P,L={}){return super.from_pretrained(P,{...L,model_file_name:L.model_file_name??"text_model"})}}class Qa extends Cs{static async from_pretrained(P,L={}){return super.from_pretrained(P,{...L,model_file_name:L.model_file_name??"vision_model"})}}class Bo extends Cs{static async from_pretrained(P,L={}){return super.from_pretrained(P,{...L,model_file_name:L.model_file_name??"vision_model"})}}class Zn extends O{}class Xa extends Zn{}class Ro extends Zn{static async from_pretrained(P,L={}){return super.from_pretrained(P,{...L,model_file_name:L.model_file_name??"text_model"})}}class jo extends Cs{static async from_pretrained(P,L={}){return super.from_pretrained(P,{...L,model_file_name:L.model_file_name??"vision_model"})}}class kn extends O{}class mt extends kn{}class eo extends O{}class No extends eo{async forward(P){const L=!P.input_ids,ne=!P.pixel_values;if(L&&ne)throw new Error("Either `input_ids` or `pixel_values` should be provided.");if(L&&(P.input_ids=(0,u.ones)([P.pixel_values.dims[0],1])),ne){const{image_size:Ye}=this.config.vision_config;P.pixel_values=(0,u.full)([0,3,Ye,Ye],0)}const{text_embeddings:ge,image_embeddings:fe,l2norm_text_embeddings:ve,l2norm_image_embeddings:Be}=await super.forward(P),Ke={};return L||(Ke.text_embeddings=ge,Ke.l2norm_text_embeddings=ve),ne||(Ke.image_embeddings=fe,Ke.l2norm_image_embeddings=Be),Ke}}class Vo extends eo{static async from_pretrained(P,L={}){return super.from_pretrained(P,{...L,model_file_name:L.model_file_name??"text_model"})}}class In extends eo{static async from_pretrained(P,L={}){return super.from_pretrained(P,{...L,model_file_name:L.model_file_name??"vision_model"})}}class Zr extends O{}class An extends Zr{}class to extends Zr{}class ro extends O{}class Uo extends ro{}class Wo extends ro{}class Fn extends O{}class Go extends Fn{}class Ko extends Fn{}class so extends O{}class On extends so{}class no extends so{}class Dn extends O{}class Ho extends Dn{}class qo extends Dn{}class oo extends O{}class Qo extends oo{}class Xo extends oo{}class ao extends O{}class h extends ao{}class E extends ao{}class A extends O{}class z extends A{}class V extends A{}class ce extends O{}class Pe extends ce{}class Re extends ce{}class Je extends O{}class it extends Je{}class xt extends O{}class Rt extends xt{}class Qt extends xt{}class Er extends O{}class Ys extends Er{}class Ja extends Er{}class wc extends O{}class dw extends wc{}class pw extends wc{}class Mc extends O{}class mw extends Mc{}class hw extends Mc{}class bc extends O{}class _w extends bc{}class fw extends bc{}class yc extends O{}class gw extends yc{}class ww extends yc{}class vc extends O{}class Mw extends vc{}class bw extends vc{}class xc extends O{}class yw extends xc{}class vw extends xc{}class Tc extends O{}class xw extends Tc{}class Tw extends Tc{}class Ec extends O{}class Ew extends Ec{}class Pw extends Ec{}class Pc extends O{}class Cw extends Pc{}class Sw extends Pc{}class Cc extends O{}class $w extends Cc{}class kw extends Cc{}class Sc extends O{}class Iw extends Sc{}class Aw extends Sc{}class $c extends O{}class Fw extends $c{}class Ow extends $c{}class kc extends O{}class Dw extends kc{}class Lw extends kc{}class Ic extends O{}class zw extends Ic{}class Bw extends Ic{}class Ac extends O{}class Rw extends Ac{}class jw extends Ac{}class Fc extends O{}class Nw extends Fc{}class Vw extends Fc{}class Oc extends O{}class Uw extends Oc{}class Ww extends Oc{}class Dc extends O{}class Gw extends Dc{}class Kw extends Dc{}class Hw extends O{constructor(){super(...arguments);J(this,"forward_params",["input_ids","attention_mask","position_ids","past_key_values","pixel_values","image_grid_thw"])}}class qw extends Hw{get_rope_index(P,L,ne,ge){const{vision_config:fe,image_token_id:ve,video_token_id:Be,vision_start_token_id:Ke}=this.config,Ye=fe.spatial_merge_size??2,ht=[];if(L||ne){let wt=P.tolist();ge||(ge=(0,u.ones_like)(P));const vt=ge.tolist(),Ft=Array.from({length:3},Nt=>Array.from({length:P.dims[0]},er=>Array.from({length:P.dims[1]},sr=>1))),_t=L?L.tolist():[],Tt=ne?ne.tolist():[];let ut=0,It=0;for(let Nt=0;Nt<wt.length;++Nt){const er=wt[Nt].filter((ir,kr)=>vt[Nt][kr]==1),ar=er.reduce((ir,kr,pn)=>(kr==Ke&&ir.push(pn),ir),[]).map(ir=>er[ir+1]),Pr=ar.filter(ir=>ir==ve).length,Dr=ar.filter(ir=>ir==Be).length;let At=[],$r=0,ns=Pr,Ln=Dr;for(let ir=0;ir<ar.length;++ir){const kr=er.findIndex((Rn,Ss)=>Ss>$r&&Rn==ve),pn=er.findIndex((Rn,Ss)=>Ss>$r&&Rn==Be),Bn=ns>0&&kr!==-1?kr:er.length+1,co=Ln>0&&pn!==-1?pn:er.length+1;let gi,Ru,ju,Nu;Bn<co?([Ru,ju,Nu]=_t[ut],++ut,--ns,gi=Bn):([Ru,ju,Nu]=Tt[It],++It,--Ln,gi=co);const[e1,Vu,wi]=[Number(Ru),Math.floor(Number(ju)/Ye),Math.floor(Number(Nu)/Ye)],Uu=gi-$r,Kv=At.length>0?(0,_.max)(At.at(-1))[0]+1:0;At.push(Array.from({length:3*Uu},(Rn,Ss)=>Kv+Ss%Uu));const Wu=Uu+Kv,Mi=e1*Vu*wi,t1=Array.from({length:Mi},(Rn,Ss)=>Wu+Math.floor(Ss/(Vu*wi))),r1=Array.from({length:Mi},(Rn,Ss)=>Wu+Math.floor(Ss/wi)%Vu),s1=Array.from({length:Mi},(Rn,Ss)=>Wu+Ss%wi);At.push([t1,r1,s1].flat()),$r=gi+Mi}if($r<er.length){const ir=At.length>0?(0,_.max)(At.at(-1))[0]+1:0,kr=er.length-$r;At.push(Array.from({length:3*kr},(pn,Bn)=>ir+Bn%kr))}const Qr=At.reduce((ir,kr)=>ir+kr.length,0),ps=new Array(Qr);let hi=0;for(let ir=0;ir<3;++ir)for(let kr=0;kr<At.length;++kr){const pn=At[kr],Bn=pn.length/3;for(let co=ir*Bn;co<(ir+1)*Bn;++co)ps[hi++]=pn[co]}let zn=0;const _i=vt[Nt];for(let ir=0;ir<_i.length;++ir)if(_i[ir]==1){for(let kr=0;kr<3;++kr)Ft[kr][Nt][ir]=ps[kr*Qr/3+zn];++zn}const fi=(0,_.max)(ps)[0];ht.push(fi+1-wt[Nt].length)}return[new u.Tensor("int64",Ft.flat(1/0),[3,P.dims[0],P.dims[1]]),new u.Tensor("int64",ht,[ht.length,1])]}else if(ge){const{data:wt,dims:vt}=ee(ge),Ft=BigInt64Array.from({length:3*wt.length},(Tt,ut)=>wt[ut%wt.length]),_t=Array.from({length:vt[0]},(Tt,ut)=>(0,_.max)(wt.subarray(vt[1]*ut,vt[1]*(ut+1)))[0]+1n+BigInt(vt[1]));return[new u.Tensor("int64",Ft,[3,...vt]),new u.Tensor("int64",_t,[_t.length,1])]}else{const[wt,vt]=P.dims,Ft=BigInt64Array.from({length:3*wt*vt},(_t,Tt)=>BigInt(Math.floor(Tt%vt/wt)));return[new u.Tensor("int64",Ft,[3,...P.dims]),(0,u.zeros)([wt,1])]}}async encode_image({pixel_values:P,image_grid_thw:L}){return(await U(this.sessions.vision_encoder,{pixel_values:P,grid_thw:L})).image_features}_merge_input_ids_with_image_features(P){return oe({image_token_id:this.config.image_token_id,...P})}prepare_inputs_for_generation(P,L,ne){if(L.attention_mask&&!L.position_ids)if(!L.past_key_values)[L.position_ids,L.rope_deltas]=this.get_rope_index(L.input_ids,L.image_grid_thw,L.video_grid_thw,L.attention_mask);else{L.pixel_values=null;const ge=BigInt(Object.values(L.past_key_values)[0].dims.at(-2)),fe=L.rope_deltas.map(ve=>ge+ve);L.position_ids=(0,u.stack)([fe,fe,fe],0)}return L}}class Lc extends O{}class Qw extends Lc{}class Xw extends Lc{}class zc extends O{}class Jw extends zc{}class Yw extends zc{}class Bc extends O{}class Zw extends Bc{}class eM extends Bc{}class Rc extends O{}class tM extends Rc{}class rM extends Rc{}class jc extends O{}class sM extends jc{}class nM extends jc{}class Nc extends O{}class oM extends Nc{}class aM extends Nc{async _call(P){return new Et(await super._call(P))}}class Vc extends O{}class iM extends Vc{}class lM extends Vc{async _call(P){return new Et(await super._call(P))}}class cM extends O{}class uM extends cM{}class Uc extends O{}class dM extends Uc{}class pM extends Uc{async _call(P){return new Et(await super._call(P))}}class mM extends O{}class hM extends mM{}class Wc extends O{}class _M extends Wc{}class fM extends Wc{async _call(P){return new Et(await super._call(P))}}class gM extends O{}class wM extends gM{}class Gc extends O{}class MM extends Gc{}class bM extends Gc{async _call(P){return new Et(await super._call(P))}}class yM extends O{}class vM extends yM{async _call(P){return new Wv(await super._call(P))}}class Kc extends O{}class xM extends Kc{}class TM extends Kc{async _call(P){return new Et(await super._call(P))}}class Hc extends O{}class EM extends Hc{}class PM extends Hc{async _call(P){return new Et(await super._call(P))}}class qc extends O{}class CM extends qc{}class SM extends qc{}class Qc extends O{}class $M extends Qc{}class kM extends Qc{}class Xc extends O{}class IM extends Xc{}class AM extends Xc{async _call(P){return new Et(await super._call(P))}}class Ya extends O{}class FM extends Ya{}class OM extends Ya{async _call(P){return new Yc(await super._call(P))}}class Jc extends Ya{async _call(P){return new DM(await super._call(P))}}class Yc extends me{constructor({logits:P,pred_boxes:L}){super(),this.logits=P,this.pred_boxes=L}}class DM extends me{constructor({logits:P,pred_boxes:L,pred_masks:ne}){super(),this.logits=P,this.pred_boxes=L,this.pred_masks=ne}}class Zc extends O{}class LM extends Zc{}class zM extends Zc{async _call(P){return new Jo(await super._call(P))}}class Jo extends me{constructor({logits:P,pred_boxes:L}){super(),this.logits=P,this.pred_boxes=L}}class eu extends O{}class BM extends eu{}class RM extends eu{async _call(P){return new jM(await super._call(P))}}class jM extends Jo{}class tu extends O{}class NM extends tu{}class VM extends tu{async _call(P){return new UM(await super._call(P))}}class UM extends Jo{}class ru extends O{}class WM extends ru{}class GM extends ru{async _call(P){return new Jo(await super._call(P))}}class su extends O{}class KM extends su{}class HM extends su{async _call(P){return new qM(await super._call(P))}}class qM extends Yc{}class nu extends O{}class QM extends nu{}class XM extends nu{async _call(P){return new Et(await super._call(P))}}class ou extends O{}class JM extends ou{}class YM extends ou{async _call(P){return new Et(await super._call(P))}}class au extends O{}class ZM extends au{}class eb extends au{async _call(P){return new Et(await super._call(P))}}class Za extends O{}class tb extends Za{}class rb extends Za{async _call(P){return new Et(await super._call(P))}}class sb extends Za{}class iu extends O{}class nb extends iu{}class ob extends iu{}class lu extends O{}class ab extends lu{}class ib extends lu{}class lb extends O{}class cb extends lb{}class ei extends O{}class ub extends ei{}class db extends ei{}class pb extends ei{}class mb extends O{}class hb extends mb{}class _b extends O{}class fb extends _b{}class gb extends O{}class wb extends gb{}class cu extends O{}class Mb extends cu{}class bb extends cu{}class uu extends O{}class yb extends uu{}class vb extends uu{}class xb extends O{}class Tb extends xb{}class du extends O{}class Eb extends du{}class Pb extends du{async _call(P){return new Et(await super._call(P))}}class pu extends O{}class Cb extends pu{}class Sb extends pu{async _call(P){return new Et(await super._call(P))}}class mu extends O{}class $b extends mu{}class kb extends mu{async _call(P){return new Et(await super._call(P))}}class hu extends O{}class Ib extends hu{}class Ab extends hu{async _call(P){return new Et(await super._call(P))}}class Fb extends O{}class Ob extends Fb{}class Db extends O{}class Lb extends Db{}class zb extends O{}class Bb extends zb{}class _u extends O{}class Rb extends _u{}class jb extends _u{async _call(P){return new Nb(await super._call(P))}}class Nb extends me{constructor({logits:P,pred_boxes:L}){super(),this.logits=P,this.pred_boxes=L}}class Vb extends O{}class Ub extends Vb{async get_image_embeddings({pixel_values:P}){return await Z(this,{pixel_values:P})}async forward(P){!P.image_embeddings||!P.image_positional_embeddings?P={...P,...await this.get_image_embeddings(P)}:P={...P},P.input_labels??(P.input_labels=(0,u.ones)(P.input_points.dims.slice(0,-1)));const L={image_embeddings:P.image_embeddings,image_positional_embeddings:P.image_positional_embeddings};return P.input_points&&(L.input_points=P.input_points),P.input_labels&&(L.input_labels=P.input_labels),P.input_boxes&&(L.input_boxes=P.input_boxes),await U(this.sessions.prompt_encoder_mask_decoder,L)}async _call(P){return new Wb(await super._call(P))}}class Wb extends me{constructor({iou_scores:P,pred_masks:L}){super(),this.iou_scores=P,this.pred_masks=L}}class Gb extends me{constructor({iou_scores:P,pred_masks:L,object_score_logits:ne}){super(),this.iou_scores=P,this.pred_masks=L,this.object_score_logits=ne}}class Kb extends O{}class ti extends Kb{async get_image_embeddings({pixel_values:P}){return await Z(this,{pixel_values:P})}async forward(P){const{num_feature_levels:L}=this.config.vision_config;if(Array.from({length:L},(ve,Be)=>`image_embeddings.${Be}`).some(ve=>!P[ve])?P={...P,...await this.get_image_embeddings(P)}:P={...P},P.input_points){if(P.input_boxes&&P.input_boxes.dims[1]!==1)throw new Error("When both `input_points` and `input_boxes` are provided, the number of boxes per image must be 1.");const ve=P.input_points.dims;P.input_labels??(P.input_labels=(0,u.ones)(ve.slice(0,-1))),P.input_boxes??(P.input_boxes=(0,u.full)([ve[0],0,4],0))}else if(P.input_boxes){const ve=P.input_boxes.dims;P.input_labels=(0,u.full)([ve[0],ve[1],0],-1n),P.input_points=(0,u.full)([ve[0],1,0,2],0)}else throw new Error("At least one of `input_points` or `input_boxes` must be provided.");const ge=this.sessions.prompt_encoder_mask_decoder,fe=(0,i.pick)(P,ge.inputNames);return await U(ge,fe)}async _call(P){return new Gb(await super._call(P))}}class Hb extends ti{}class qb extends ti{}class fu extends O{}class Qb extends fu{}class Xb extends fu{}class gu extends O{}class Jb extends gu{}class Yb extends gu{}class un extends O{}class Zb extends un{}class ey extends un{async _call(P){return new dn(await super._call(P))}}class ty extends un{async _call(P){return new Et(await super._call(P))}}class ry extends un{async _call(P){return new Sr(await super._call(P))}}class sy extends O{}class ny extends sy{async _call(P){return new dn(await super._call(P))}}class wu extends O{}class oy extends wu{}class ay extends wu{async _call(P){return new Sr(await super._call(P))}}class iy extends O{}class ly extends iy{}class ri extends O{}class cy extends ri{}class uy extends ri{async _call(P){return new dn(await super._call(P))}}class dy extends ri{async _call(P){return new Et(await super._call(P))}}class Yo extends O{}class py extends Yo{}class my extends Yo{async _call(P){return new dn(await super._call(P))}}class hy extends Yo{async _call(P){return new Et(await super._call(P))}}class _y extends Yo{async _call(P){return new Sr(await super._call(P))}}class si extends O{}class fy extends si{}class gy extends si{async _call(P){return new dn(await super._call(P))}}class wy extends si{async _call(P){return new Et(await super._call(P))}}class RT extends O{}class My extends un{}class by extends un{async _call(P){return new dn(await super._call(P))}}class yy extends un{async _call(P){return new Et(await super._call(P))}}class io extends O{}class vy extends io{}class xy extends io{async _call(P){return new dn(await super._call(P))}}class Ty extends io{async _call(P){return new Et(await super._call(P))}}class Ey extends io{async _call(P){return new Uv(await super._call(P))}}class Py extends io{async _call(P){return new Sr(await super._call(P))}}class Cy extends O{}class Sy extends Cy{}class ni extends O{}class jT extends ni{}class $y extends ni{}class ky extends ni{async generate_speech(P,L,{threshold:ne=.5,minlenratio:ge=0,maxlenratio:fe=20,vocoder:ve=null}={}){const Be={input_ids:P},{encoder_outputs:Ke,encoder_attention_mask:Ye}=await Z(this,Be),ht=Ke.dims[1]/this.config.reduction_factor,wt=Math.floor(ht*fe),vt=Math.floor(ht*ge),Ft=this.config.num_mel_bins;let _t=[],Tt=null,ut=null,It=0;for(;;){++It;const sr=q(!!ut);let ar;ut?ar=ut.output_sequence_out:ar=new u.Tensor("float32",new Float32Array(Ft),[1,1,Ft]);let Pr={use_cache_branch:sr,output_sequence:ar,encoder_attention_mask:Ye,speaker_embeddings:L,encoder_hidden_states:Ke};this.addPastKeyValues(Pr,Tt),ut=await U(this.sessions.decoder_model_merged,Pr),Tt=this.getPastKeyValues(ut,Tt);const{prob:Dr,spectrum:At}=ut;if(_t.push(At),It>=vt&&(Array.from(Dr.data).filter($r=>$r>=ne).length>0||It>=wt))break}const Nt=(0,u.cat)(_t),{waveform:er}=await U(ve.sessions.model,{spectrogram:Nt});return{spectrogram:Nt,waveform:er}}}class Iy extends O{constructor(){super(...arguments);J(this,"main_input_name","spectrogram")}}class Ay extends O{}class Mu extends Ay{async generate_speech({input_ids:P,attention_mask:L,style:ne,num_inference_steps:ge=5,speed:fe=1.05}){const{sampling_rate:ve,chunk_compress_factor:Be,base_chunk_size:Ke,latent_dim:Ye}=this.config,{last_hidden_state:ht,durations:wt}=await U(this.sessions.text_encoder,{input_ids:P,attention_mask:L,style:ne});wt.div_(fe);const vt=wt.max().item()*ve,Ft=Ke*Be,_t=Math.floor((vt+Ft-1)/Ft),Tt=P.dims[0],ut=(0,u.ones)([Tt,_t]),It=(0,u.full)([Tt],ge);let Nt=(0,u.randn)([Tt,Ye*Be,_t]);for(let sr=0;sr<ge;++sr){const ar=(0,u.full)([Tt],sr);({denoised_latents:Nt}=await U(this.sessions.latent_denoiser,{style:ne,noisy_latents:Nt,latent_mask:ut,encoder_outputs:ht,attention_mask:L,timestep:ar,num_inference_steps:It}))}const{waveform:er}=await U(this.sessions.voice_decoder,{latents:Nt});return{waveform:er,durations:wt}}}class Fy extends O{}class Oy extends Fy{}class bu extends O{}class Dy extends bu{}class Ly extends bu{}class yu extends O{}class zy extends yu{}class By extends yu{}class vu extends O{}class Ry extends vu{}class jy extends vu{}class xu extends O{}class Ny extends xu{}class Vy extends xu{}class Tu extends O{}class Uy extends Tu{}class Wy extends Tu{}class Eu extends O{}class Gy extends Eu{}class Ky extends Eu{}class oi extends O{}class Hy extends oi{}class qy extends oi{static async from_pretrained(P,L={}){return super.from_pretrained(P,{...L,model_file_name:L.model_file_name??"text_model"})}}class Qy extends oi{static async from_pretrained(P,L={}){return super.from_pretrained(P,{...L,model_file_name:L.model_file_name??"audio_model"})}}class Xy extends O{}class Pu extends Xy{async _call(P){return new Gv(await super._call(P))}}class ai extends O{}class NT extends ai{}class Jy extends ai{}class Yy extends ai{}class Cu extends O{}class Zy extends Cu{}class e0 extends Cu{}class Su extends O{}class t0 extends Su{}class r0 extends Su{async _call(P){return new Et(await super._call(P))}}class $u extends O{}class VT extends $u{}class UT extends $u{}class ku extends O{constructor(){super(...arguments);J(this,"forward_params",["input_ids","attention_mask","encoder_outputs","decoder_input_ids","decoder_attention_mask","past_key_values"])}_apply_and_filter_by_delay_pattern_mask(L){const[ne,ge]=L.dims,fe=this.config.decoder.num_codebooks,ve=ge-fe;let Be=0;for(let ht=0;ht<L.size;++ht){if(L.data[ht]===this.config.decoder.pad_token_id)continue;const wt=ht%ge,vt=Math.floor(ht/ge)%fe,Ft=wt-vt;Ft>0&&Ft<=ve&&(L.data[Be++]=L.data[ht])}const Ke=Math.floor(ne/fe),Ye=Be/(Ke*fe);return new u.Tensor(L.type,L.data.slice(0,Be),[Ke,fe,Ye])}prepare_inputs_for_generation(L,ne,ge){let fe=structuredClone(L);for(let Be=0;Be<fe.length;++Be)for(let Ke=0;Ke<fe[Be].length;++Ke)Be%this.config.decoder.num_codebooks>=Ke&&(fe[Be][Ke]=BigInt(this.config.decoder.pad_token_id));return ge.guidance_scale!==null&&ge.guidance_scale>1&&(fe=fe.concat(fe)),super.prepare_inputs_for_generation(fe,ne,ge)}async generate(L){const ne=await super.generate(L),ge=this._apply_and_filter_by_delay_pattern_mask(ne).unsqueeze_(0),{audio_values:fe}=await U(this.sessions.encodec_decode,{audio_codes:ge});return fe}}class ii extends O{}class s0 extends ii{}class n0 extends ii{async _call(P){return new Et(await super._call(P))}}class o0 extends ii{}class li extends O{}class a0 extends li{}class i0 extends li{async _call(P){return new Et(await super._call(P))}}class l0 extends li{}class ci extends O{}class c0 extends ci{}class u0 extends ci{async _call(P){return new Et(await super._call(P))}}class d0 extends ci{}class ui extends O{}class p0 extends ui{}class m0 extends ui{async _call(P){return new Et(await super._call(P))}}class h0 extends ui{}class _0 extends O{}class f0 extends _0{}class g0 extends O{}class w0 extends g0{constructor(...L){super(...L);J(this,"forward_params",["input_ids","pixel_values","images_seq_mask","images_emb_mask","attention_mask","position_ids","past_key_values"]);this._generation_mode="text"}async forward(L){const ne=this._generation_mode??"text";let ge;if(ne==="text"||!L.past_key_values){const Ye=this.sessions.prepare_inputs_embeds,ht=(0,i.pick)(L,Ye.inputNames);ge=await U(Ye,ht)}else{const Ye=this.sessions.gen_img_embeds,ht=(0,i.pick)({image_ids:L.input_ids},Ye.inputNames);ge=await U(Ye,ht)}const fe={...L,...ge},ve=await _e(this,fe),Be=this.sessions[ne==="text"?"lm_head":"gen_head"];if(!Be)throw new Error(`Unable to find "${Be}" generation head`);const Ke=await U(Be,(0,i.pick)(ve,Be.inputNames));return{...ge,...ve,...Ke}}async generate(L){return this._generation_mode="text",super.generate(L)}async generate_images(L){this._generation_mode="image";const ne=(L.inputs??L[this.main_input_name]).dims[1],fe=(await super.generate(L)).slice(null,[ne,null]),ve=this.sessions.image_decode,{decoded_image:Be}=await U(ve,{generated_tokens:fe}),Ke=Be.add_(1).mul_(255/2).clamp_(0,255).to("uint8"),Ye=[];for(const ht of Ke){const wt=f.RawImage.fromTensor(ht);Ye.push(wt)}return Ye}}class M0 extends me{constructor({char_logits:P,bpe_logits:L,wp_logits:ne}){super(),this.char_logits=P,this.bpe_logits=L,this.wp_logits=ne}get logits(){return[this.char_logits,this.bpe_logits,this.wp_logits]}}class b0 extends O{}class y0 extends b0{async _call(P){return new M0(await super._call(P))}}class Iu extends O{}class v0 extends Iu{}class x0 extends Iu{}class Au extends O{}class T0 extends Au{}class E0 extends Au{}class P0 extends O{constructor(){super(...arguments);J(this,"forward_params",["input_ids","attention_mask","position_ids","audio_values","past_key_values"])}}class Fu extends P0{_merge_input_ids_with_audio_features(P){const L=P.audio_features.dims.at(-1),ne=P.audio_features.view(-1,L);return G({audio_token_id:this.config.ignore_index??this.config.audio_token_id,...P,audio_features:ne})}}class C0 extends Fu{}class di extends O{constructor(){super(...arguments);J(this,"main_input_name","input_values");J(this,"forward_params",["input_values"])}}class S0 extends me{constructor({audio_codes:P}){super(),this.audio_codes=P}}class $0 extends me{constructor({audio_values:P}){super(),this.audio_values=P}}class k0 extends di{async encode(P){return new S0(await U(this.sessions.encoder_model,P))}async decode(P){return new $0(await U(this.sessions.decoder_model,P))}}class I0 extends di{static async from_pretrained(P,L={}){return super.from_pretrained(P,{...L,model_file_name:L.model_file_name??"encoder_model"})}}class A0 extends di{static async from_pretrained(P,L={}){return super.from_pretrained(P,{...L,model_file_name:L.model_file_name??"decoder_model"})}}class pi extends O{constructor(){super(...arguments);J(this,"main_input_name","input_values");J(this,"forward_params",["input_values"])}}class F0 extends me{constructor({audio_codes:P}){super(),this.audio_codes=P}}class O0 extends me{constructor({audio_values:P}){super(),this.audio_values=P}}class D0 extends pi{async encode(P){return new F0(await U(this.sessions.encoder_model,P))}async decode(P){return new O0(await U(this.sessions.decoder_model,P))}}class L0 extends pi{static async from_pretrained(P,L={}){return super.from_pretrained(P,{...L,model_file_name:L.model_file_name??"encoder_model"})}}class z0 extends pi{static async from_pretrained(P,L={}){return super.from_pretrained(P,{...L,model_file_name:L.model_file_name??"decoder_model"})}}class mi extends O{constructor(){super(...arguments);J(this,"main_input_name","input_values");J(this,"forward_params",["input_values"])}}class B0 extends mi{async encode(P){return await U(this.sessions.encoder_model,P)}async decode(P){return await U(this.sessions.decoder_model,P)}}class R0 extends mi{static async from_pretrained(P,L={}){return super.from_pretrained(P,{...L,model_file_name:L.model_file_name??"encoder_model"})}}class j0 extends mi{static async from_pretrained(P,L={}){return super.from_pretrained(P,{...L,model_file_name:L.model_file_name??"decoder_model"})}}class Gt{static async from_pretrained(P,{progress_callback:L=null,config:ne=null,cache_dir:ge=null,local_files_only:fe=!1,revision:ve="main",model_file_name:Be=null,subfolder:Ke="onnx",device:Ye=null,dtype:ht=null,use_external_data_format:wt=null,session_options:vt={}}={}){const Ft={progress_callback:L,config:ne,cache_dir:ge,local_files_only:fe,revision:ve,model_file_name:Be,subfolder:Ke,device:Ye,dtype:ht,use_external_data_format:wt,session_options:vt};if(Ft.config=await s.AutoConfig.from_pretrained(P,Ft),!this.MODEL_CLASS_MAPPINGS)throw new Error("`MODEL_CLASS_MAPPINGS` not implemented for this type of `AutoClass`: "+this.name);const _t=Ft.config.model_type;for(const Tt of this.MODEL_CLASS_MAPPINGS){let ut=Tt.get(_t);if(!ut){for(const It of Tt.values())if(It[0]===_t){ut=It;break}if(!ut)continue}return await ut[1].from_pretrained(P,Ft)}if(this.BASE_IF_FAIL)return pv.has(_t)||console.warn(`Unknown model class "${_t}", attempting to construct from base class.`),await O.from_pretrained(P,Ft);throw Error(`Unsupported model type: ${_t}`)}}J(Gt,"MODEL_CLASS_MAPPINGS",null),J(Gt,"BASE_IF_FAIL",!1);const WT=new Map([["bert",["BertModel",Ce]],["neobert",["NeoBertModel",He]],["modernbert",["ModernBertModel",Oe]],["nomic_bert",["NomicBertModel",Ut]],["roformer",["RoFormerModel",Ws]],["electra",["ElectraModel",$t]],["esm",["EsmModel",Es]],["convbert",["ConvBertModel",Ks]],["camembert",["CamembertModel",X]],["deberta",["DebertaModel",St]],["deberta-v2",["DebertaV2Model",Br]],["mpnet",["MPNetModel",qs]],["albert",["AlbertModel",Ge]],["distilbert",["DistilBertModel",Hs]],["roberta",["RobertaModel",xa]],["xlm",["XLMModel",Ca]],["xlm-roberta",["XLMRobertaModel",Aa]],["clap",["ClapModel",Hy]],["clip",["CLIPModel",Lo]],["clipseg",["CLIPSegModel",An]],["chinese_clip",["ChineseCLIPModel",mt]],["siglip",["SiglipModel",Xa]],["jina_clip",["JinaCLIPModel",No]],["mobilebert",["MobileBertModel",Wr]],["squeezebert",["SqueezeBertModel",N]],["wav2vec2",["Wav2Vec2Model",Zb]],["wav2vec2-bert",["Wav2Vec2BertModel",fy]],["unispeech",["UniSpeechModel",cy]],["unispeech-sat",["UniSpeechSatModel",py]],["hubert",["HubertModel",My]],["wavlm",["WavLMModel",vy]],["audio-spectrogram-transformer",["ASTModel",Da]],["vits",["VitsModel",Pu]],["pyannote",["PyAnnoteModel",oy]],["wespeaker-resnet",["WeSpeakerResNetModel",ly]],["detr",["DetrModel",FM]],["rt_detr",["RTDetrModel",LM]],["rt_detr_v2",["RTDetrV2Model",BM]],["rf_detr",["RFDetrModel",NM]],["d_fine",["DFineModel",WM]],["table-transformer",["TableTransformerModel",KM]],["vit",["ViTModel",oM]],["ijepa",["IJepaModel",iM]],["pvt",["PvtModel",dM]],["vit_msn",["ViTMSNModel",_M]],["vit_mae",["ViTMAEModel",hM]],["groupvit",["GroupViTModel",wM]],["fastvit",["FastViTModel",MM]],["mobilevit",["MobileViTModel",xM]],["mobilevitv2",["MobileViTV2Model",EM]],["owlvit",["OwlViTModel",CM]],["owlv2",["Owlv2Model",$M]],["beit",["BeitModel",IM]],["deit",["DeiTModel",QM]],["hiera",["HieraModel",JM]],["convnext",["ConvNextModel",Eb]],["convnextv2",["ConvNextV2Model",Cb]],["dinov2",["Dinov2Model",$b]],["dinov2_with_registers",["Dinov2WithRegistersModel",Ib]],["dinov3_vit",["DINOv3ViTModel",Ob]],["dinov3_convnext",["DINOv3ConvNextModel",Lb]],["resnet",["ResNetModel",ZM]],["swin",["SwinModel",tb]],["swin2sr",["Swin2SRModel",nb]],["donut-swin",["DonutSwinModel",Tb]],["yolos",["YolosModel",Rb]],["dpt",["DPTModel",ab]],["glpn",["GLPNModel",yb]],["hifigan",["SpeechT5HifiGan",Iy]],["efficientnet",["EfficientNetModel",t0]],["decision_transformer",["DecisionTransformerModel",f0]],["patchtst",["PatchTSTForPrediction",v0]],["patchtsmixer",["PatchTSMixerForPrediction",T0]],["mobilenet_v1",["MobileNetV1Model",s0]],["mobilenet_v2",["MobileNetV2Model",a0]],["mobilenet_v3",["MobileNetV3Model",c0]],["mobilenet_v4",["MobileNetV4Model",p0]],["maskformer",["MaskFormerModel",Mb]],["mgp-str",["MgpstrForSceneTextRecognition",y0]],["style_text_to_speech_2",["StyleTextToSpeech2Model",Sy]]]),GT=new Map([["t5",["T5Model",jt]],["longt5",["LongT5Model",Jt]],["mt5",["MT5Model",hr]],["bart",["BartModel",ss]],["mbart",["MBartModel",Yr]],["marian",["MarianModel",Qb]],["whisper",["WhisperModel",Ao]],["m2m_100",["M2M100Model",Jb]],["blenderbot",["BlenderbotModel",_r]],["blenderbot-small",["BlenderbotSmallModel",qr]]]),KT=new Map([["mimi",["MimiModel",k0]],["dac",["DacModel",D0]],["snac",["SnacModel",B0]]]),HT=new Map([["bloom",["BloomModel",Zw]],["jais",["JAISModel",Go]],["gpt2",["GPT2Model",Uo]],["gptj",["GPTJModel",Qo]],["gpt_bigcode",["GPTBigCodeModel",h]],["gpt_neo",["GPTNeoModel",On]],["gpt_neox",["GPTNeoXModel",Ho]],["codegen",["CodeGenModel",z]],["llama",["LlamaModel",Pe]],["nanochat",["NanoChatModel",Rt]],["arcee",["ArceeModel",Ys]],["lfm2",["Lfm2Model",dw]],["smollm3",["SmolLM3Model",mw]],["exaone",["ExaoneModel",Mw]],["olmo",["OlmoModel",xw]],["olmo2",["Olmo2Model",Ew]],["mobilellm",["MobileLLMModel",yw]],["granite",["GraniteModel",Cw]],["granitemoehybrid",["GraniteMoeHybridModel",$w]],["cohere",["CohereModel",Iw]],["gemma",["GemmaModel",Fw]],["gemma2",["Gemma2Model",Dw]],["vaultgemma",["VaultGemmaModel",zw]],["gemma3_text",["Gemma3Model",Rw]],["helium",["HeliumModel",_w]],["glm",["GlmModel",gw]],["openelm",["OpenELMModel",Nw]],["qwen2",["Qwen2Model",Uw]],["qwen3",["Qwen3Model",Gw]],["phi",["PhiModel",Qw]],["phi3",["Phi3Model",Jw]],["mpt",["MptModel",tM]],["opt",["OPTModel",sM]],["mistral",["MistralModel",Dy]],["ministral",["MinistralModel",zy]],["ministral3",["Ministral3Model",Ry]],["ernie4_5",["Ernie4_5Model",Ny]],["starcoder2",["Starcoder2Model",Uy]],["falcon",["FalconModel",Gy]],["stablelm",["StableLmModel",Zy]],["modernbert-decoder",["ModernBertDecoderModel",kt]]]),Ou=new Map([["speecht5",["SpeechT5ForSpeechToText",$y]],["whisper",["WhisperForConditionalGeneration",Js]],["lite-whisper",["LiteWhisperForConditionalGeneration",za]],["moonshine",["MoonshineForConditionalGeneration",Ba]]]),N0=new Map([["speecht5",["SpeechT5ForTextToSpeech",ky]]]),V0=new Map([["vits",["VitsModel",Pu]],["musicgen",["MusicgenForConditionalGeneration",ku]],["supertonic",["SupertonicForConditionalGeneration",Mu]]]),U0=new Map([["bert",["BertForSequenceClassification",Me]],["neobert",["NeoBertForSequenceClassification",Ze]],["modernbert",["ModernBertForSequenceClassification",pt]],["roformer",["RoFormerForSequenceClassification",nn]],["electra",["ElectraForSequenceClassification",S]],["esm",["EsmForSequenceClassification",Ls]],["convbert",["ConvBertForSequenceClassification",Is]],["camembert",["CamembertForSequenceClassification",Ee]],["deberta",["DebertaForSequenceClassification",Ot]],["deberta-v2",["DebertaV2ForSequenceClassification",Rr]],["mpnet",["MPNetForSequenceClassification",Qs]],["albert",["AlbertForSequenceClassification",Xe]],["distilbert",["DistilBertForSequenceClassification",cs]],["roberta",["RobertaForSequenceClassification",$o]],["xlm",["XLMForSequenceClassification",$a]],["xlm-roberta",["XLMRobertaForSequenceClassification",Oa]],["bart",["BartForSequenceClassification",Tr]],["mbart",["MBartForSequenceClassification",Fr]],["mobilebert",["MobileBertForSequenceClassification",Ve]],["squeezebert",["SqueezeBertForSequenceClassification",ie]]]),W0=new Map([["bert",["BertForTokenClassification",Le]],["neobert",["NeoBertForTokenClassification",We]],["modernbert",["ModernBertForTokenClassification",ft]],["roformer",["RoFormerForTokenClassification",cr]],["electra",["ElectraForTokenClassification",H]],["esm",["EsmForTokenClassification",zs]],["convbert",["ConvBertForTokenClassification",As]],["camembert",["CamembertForTokenClassification",et]],["deberta",["DebertaForTokenClassification",at]],["deberta-v2",["DebertaV2ForTokenClassification",Ts]],["mpnet",["MPNetForTokenClassification",En]],["distilbert",["DistilBertForTokenClassification",jr]],["roberta",["RobertaForTokenClassification",Ea]],["xlm",["XLMForTokenClassification",ka]],["xlm-roberta",["XLMRobertaForTokenClassification",ko]]]),Du=new Map([["t5",["T5ForConditionalGeneration",Bt]],["longt5",["LongT5ForConditionalGeneration",yr]],["mt5",["MT5ForConditionalGeneration",xr]],["bart",["BartForConditionalGeneration",rr]],["mbart",["MBartForConditionalGeneration",Ps]],["marian",["MarianMTModel",Xb]],["m2m_100",["M2M100ForConditionalGeneration",Yb]],["blenderbot",["BlenderbotForConditionalGeneration",wr]],["blenderbot-small",["BlenderbotSmallForConditionalGeneration",Xs]]]),Lu=new Map([["bloom",["BloomForCausalLM",eM]],["gpt2",["GPT2LMHeadModel",Wo]],["jais",["JAISLMHeadModel",Ko]],["gptj",["GPTJForCausalLM",Xo]],["gpt_bigcode",["GPTBigCodeForCausalLM",E]],["gpt_neo",["GPTNeoForCausalLM",no]],["gpt_neox",["GPTNeoXForCausalLM",qo]],["codegen",["CodeGenForCausalLM",V]],["llama",["LlamaForCausalLM",Re]],["nanochat",["NanoChatForCausalLM",Qt]],["llama4_text",["Llama4ForCausalLM",it]],["arcee",["ArceeForCausalLM",Ja]],["lfm2",["Lfm2ForCausalLM",pw]],["smollm3",["SmolLM3ForCausalLM",hw]],["exaone",["ExaoneForCausalLM",bw]],["olmo",["OlmoForCausalLM",Tw]],["olmo2",["Olmo2ForCausalLM",Pw]],["mobilellm",["MobileLLMForCausalLM",vw]],["granite",["GraniteForCausalLM",Sw]],["granitemoehybrid",["GraniteMoeHybridForCausalLM",kw]],["cohere",["CohereForCausalLM",Aw]],["gemma",["GemmaForCausalLM",Ow]],["gemma2",["Gemma2ForCausalLM",Lw]],["vaultgemma",["VaultGemmaForCausalLM",Bw]],["gemma3_text",["Gemma3ForCausalLM",jw]],["helium",["HeliumForCausalLM",fw]],["glm",["GlmForCausalLM",ww]],["openelm",["OpenELMForCausalLM",Vw]],["qwen2",["Qwen2ForCausalLM",Ww]],["qwen3",["Qwen3ForCausalLM",Kw]],["phi",["PhiForCausalLM",Xw]],["phi3",["Phi3ForCausalLM",Yw]],["mpt",["MptForCausalLM",rM]],["opt",["OPTForCausalLM",nM]],["mbart",["MBartForCausalLM",ds]],["mistral",["MistralForCausalLM",Ly]],["ministral",["MinistralForCausalLM",By]],["ministral3",["Ministral3ForCausalLM",jy]],["ernie4_5",["Ernie4_5ForCausalLM",Vy]],["starcoder2",["Starcoder2ForCausalLM",Wy]],["falcon",["FalconForCausalLM",Ky]],["trocr",["TrOCRForCausalLM",Oy]],["stablelm",["StableLmForCausalLM",e0]],["modernbert-decoder",["ModernBertDecoderForCausalLM",Mr]],["phi3_v",["Phi3VForCausalLM",$n]]]),qT=new Map([["multi_modality",["MultiModalityCausalLM",w0]]]),G0=new Map([["bert",["BertForMaskedLM",$e]],["neobert",["NeoBertForMaskedLM",Fe]],["modernbert",["ModernBertForMaskedLM",ze]],["roformer",["RoFormerForMaskedLM",Gs]],["electra",["ElectraForMaskedLM",Fs]],["esm",["EsmForMaskedLM",Ds]],["convbert",["ConvBertForMaskedLM",gt]],["camembert",["CamembertForMaskedLM",ue]],["deberta",["DebertaForMaskedLM",lt]],["deberta-v2",["DebertaV2ForMaskedLM",br]],["mpnet",["MPNetForMaskedLM",Bs]],["albert",["AlbertForMaskedLM",tt]],["distilbert",["DistilBertForMaskedLM",ur]],["roberta",["RobertaForMaskedLM",Ta]],["xlm",["XLMWithLMHeadModel",Sa]],["xlm-roberta",["XLMRobertaForMaskedLM",Fa]],["mobilebert",["MobileBertForMaskedLM",je]],["squeezebert",["SqueezeBertForMaskedLM",Y]]]),K0=new Map([["bert",["BertForQuestionAnswering",we]],["neobert",["NeoBertForQuestionAnswering",yt]],["roformer",["RoFormerForQuestionAnswering",zr]],["electra",["ElectraForQuestionAnswering",R]],["convbert",["ConvBertForQuestionAnswering",Yt]],["camembert",["CamembertForQuestionAnswering",ot]],["deberta",["DebertaForQuestionAnswering",Dt]],["deberta-v2",["DebertaV2ForQuestionAnswering",Os]],["mpnet",["MPNetForQuestionAnswering",he]],["albert",["AlbertForQuestionAnswering",qe]],["distilbert",["DistilBertForQuestionAnswering",Hr]],["roberta",["RobertaForQuestionAnswering",Pa]],["xlm",["XLMForQuestionAnswering",Ia]],["xlm-roberta",["XLMRobertaForQuestionAnswering",Io]],["mobilebert",["MobileBertForQuestionAnswering",st]],["squeezebert",["SqueezeBertForQuestionAnswering",de]]]),zu=new Map([["vision-encoder-decoder",["VisionEncoderDecoderModel",Fo]],["idefics3",["Idefics3ForConditionalGeneration",Sn]],["smolvlm",["SmolVLMForConditionalGeneration",Do]]]),H0=new Map([["llava",["LlavaForConditionalGeneration",Cn]],["llava_onevision",["LlavaOnevisionForConditionalGeneration",Ra]],["moondream1",["Moondream1ForConditionalGeneration",ja]],["florence2",["Florence2ForConditionalGeneration",Va]],["qwen2-vl",["Qwen2VLForConditionalGeneration",qw]],["idefics3",["Idefics3ForConditionalGeneration",Sn]],["smolvlm",["SmolVLMForConditionalGeneration",Do]],["paligemma",["PaliGemmaForConditionalGeneration",Wa]],["llava_qwen2",["LlavaQwen2ForCausalLM",Oo]],["gemma3n",["Gemma3nForConditionalGeneration",ln]],["mistral3",["Mistral3ForConditionalGeneration",Ga]]]),q0=new Map([["ultravox",["UltravoxModel",Fu]],["voxtral",["VoxtralForConditionalGeneration",C0]]]),QT=new Map([["vision-encoder-decoder",["VisionEncoderDecoderModel",Fo]]]),Q0=new Map([["vit",["ViTForImageClassification",aM]],["ijepa",["IJepaForImageClassification",lM]],["pvt",["PvtForImageClassification",pM]],["vit_msn",["ViTMSNForImageClassification",fM]],["fastvit",["FastViTForImageClassification",bM]],["mobilevit",["MobileViTForImageClassification",TM]],["mobilevitv2",["MobileViTV2ForImageClassification",PM]],["beit",["BeitForImageClassification",AM]],["deit",["DeiTForImageClassification",XM]],["hiera",["HieraForImageClassification",YM]],["convnext",["ConvNextForImageClassification",Pb]],["convnextv2",["ConvNextV2ForImageClassification",Sb]],["dinov2",["Dinov2ForImageClassification",kb]],["dinov2_with_registers",["Dinov2WithRegistersForImageClassification",Ab]],["resnet",["ResNetForImageClassification",eb]],["swin",["SwinForImageClassification",rb]],["segformer",["SegformerForImageClassification",Jy]],["efficientnet",["EfficientNetForImageClassification",r0]],["mobilenet_v1",["MobileNetV1ForImageClassification",n0]],["mobilenet_v2",["MobileNetV2ForImageClassification",i0]],["mobilenet_v3",["MobileNetV3ForImageClassification",u0]],["mobilenet_v4",["MobileNetV4ForImageClassification",m0]]]),X0=new Map([["detr",["DetrForObjectDetection",OM]],["rt_detr",["RTDetrForObjectDetection",zM]],["rt_detr_v2",["RTDetrV2ForObjectDetection",RM]],["rf_detr",["RFDetrForObjectDetection",VM]],["d_fine",["DFineForObjectDetection",GM]],["table-transformer",["TableTransformerForObjectDetection",HM]],["yolos",["YolosForObjectDetection",jb]]]),J0=new Map([["owlvit",["OwlViTForObjectDetection",SM]],["owlv2",["Owlv2ForObjectDetection",kM]],["grounding-dino",["GroundingDinoForObjectDetection",Bb]]]),lo=new Map([["detr",["DetrForSegmentation",Jc]],["clipseg",["CLIPSegForImageSegmentation",to]]]),Y0=new Map([["segformer",["SegformerForSemanticSegmentation",Yy]],["sapiens",["SapiensForSemanticSegmentation",ub]],["swin",["SwinForSemanticSegmentation",sb]],["mobilenet_v1",["MobileNetV1ForSemanticSegmentation",o0]],["mobilenet_v2",["MobileNetV2ForSemanticSegmentation",l0]],["mobilenet_v3",["MobileNetV3ForSemanticSegmentation",d0]],["mobilenet_v4",["MobileNetV4ForSemanticSegmentation",h0]]]),Z0=new Map([["detr",["DetrForSegmentation",Jc]],["maskformer",["MaskFormerForInstanceSegmentation",bb]]]),ev=new Map([["sam",["SamModel",Ub]],["sam2",["Sam2Model",ti]],["edgetam",["EdgeTamModel",Hb]],["sam3_tracker",["Sam3TrackerModel",qb]]]),tv=new Map([["wav2vec2",["Wav2Vec2ForCTC",ey]],["wav2vec2-bert",["Wav2Vec2BertForCTC",gy]],["unispeech",["UniSpeechForCTC",uy]],["unispeech-sat",["UniSpeechSatForCTC",my]],["wavlm",["WavLMForCTC",xy]],["hubert",["HubertForCTC",by]],["parakeet_ctc",["ParakeetForCTC",ny]]]),rv=new Map([["wav2vec2",["Wav2Vec2ForSequenceClassification",ty]],["wav2vec2-bert",["Wav2Vec2BertForSequenceClassification",wy]],["unispeech",["UniSpeechForSequenceClassification",dy]],["unispeech-sat",["UniSpeechSatForSequenceClassification",hy]],["wavlm",["WavLMForSequenceClassification",Ty]],["hubert",["HubertForSequenceClassification",yy]],["audio-spectrogram-transformer",["ASTForAudioClassification",La]]]),sv=new Map([["wavlm",["WavLMForXVector",Ey]]]),nv=new Map([["unispeech-sat",["UniSpeechSatForAudioFrameClassification",_y]],["wavlm",["WavLMForAudioFrameClassification",Py]],["wav2vec2",["Wav2Vec2ForAudioFrameClassification",ry]],["pyannote",["PyAnnoteForAudioFrameClassification",ay]]]),ov=new Map([["vitmatte",["VitMatteForImageMatting",vM]]]),XT=new Map([["patchtst",["PatchTSTForPrediction",x0]],["patchtsmixer",["PatchTSMixerForPrediction",E0]]]),av=new Map([["swin2sr",["Swin2SRForImageSuperResolution",ob]]]),iv=new Map([["dpt",["DPTForDepthEstimation",ib]],["depth_anything",["DepthAnythingForDepthEstimation",cb]],["glpn",["GLPNForDepthEstimation",vb]],["sapiens",["SapiensForDepthEstimation",db]],["depth_pro",["DepthProForDepthEstimation",hb]],["metric3d",["Metric3DForDepthEstimation",fb]],["metric3dv2",["Metric3Dv2ForDepthEstimation",wb]]]),lv=new Map([["sapiens",["SapiensForNormalEstimation",pb]]]),cv=new Map([["vitpose",["VitPoseForPoseEstimation",uM]]]),uv=new Map([["clip",["CLIPVisionModelWithProjection",Bo]],["siglip",["SiglipVisionModel",jo]],["jina_clip",["JinaCLIPVisionModel",In]]]),dv=[[WT,T.EncoderOnly],[GT,T.EncoderDecoder],[HT,T.DecoderOnly],[KT,T.AutoEncoder],[U0,T.EncoderOnly],[W0,T.EncoderOnly],[Du,T.Seq2Seq],[Ou,T.Seq2Seq],[Lu,T.DecoderOnly],[qT,T.MultiModality],[G0,T.EncoderOnly],[K0,T.EncoderOnly],[zu,T.Vision2Seq],[H0,T.ImageTextToText],[q0,T.AudioTextToText],[Q0,T.EncoderOnly],[lo,T.EncoderOnly],[Z0,T.EncoderOnly],[Y0,T.EncoderOnly],[ov,T.EncoderOnly],[XT,T.EncoderOnly],[av,T.EncoderOnly],[iv,T.EncoderOnly],[lv,T.EncoderOnly],[cv,T.EncoderOnly],[X0,T.EncoderOnly],[J0,T.EncoderOnly],[ev,T.MaskGeneration],[tv,T.EncoderOnly],[rv,T.EncoderOnly],[N0,T.Seq2Seq],[V0,T.EncoderOnly],[sv,T.EncoderOnly],[nv,T.EncoderOnly],[uv,T.EncoderOnly]];for(const[g,P]of dv)for(const[L,ne]of g.values())M.set(L,P),v.set(ne,L),x.set(L,ne);const JT=[["MusicgenForConditionalGeneration",ku,T.Musicgen],["Phi3VForCausalLM",$n,T.Phi3V],["CLIPTextModelWithProjection",zo,T.EncoderOnly],["SiglipTextModel",Ro,T.EncoderOnly],["JinaCLIPTextModel",Vo,T.EncoderOnly],["ClapTextModelWithProjection",qy,T.EncoderOnly],["ClapAudioModelWithProjection",Qy,T.EncoderOnly],["DacEncoderModel",L0,T.EncoderOnly],["DacDecoderModel",z0,T.EncoderOnly],["MimiEncoderModel",I0,T.EncoderOnly],["MimiDecoderModel",A0,T.EncoderOnly],["SnacEncoderModel",R0,T.EncoderOnly],["SnacDecoderModel",j0,T.EncoderOnly],["Gemma3nForConditionalGeneration",ln,T.ImageAudioTextToText],["SupertonicForConditionalGeneration",Mu,T.Supertonic]];for(const[g,P,L]of JT)M.set(g,L),v.set(P,g),x.set(g,P);const pv=new Map([["modnet",lo],["birefnet",lo],["isnet",lo],["ben",lo]]);for(const[g,P]of pv.entries())P.set(g,["PreTrainedModel",O]),M.set(g,T.EncoderOnly),v.set(O,g),x.set(g,O);class Bu extends Gt{}J(Bu,"MODEL_CLASS_MAPPINGS",dv.map(P=>P[0])),J(Bu,"BASE_IF_FAIL",!0);class mv extends Gt{}J(mv,"MODEL_CLASS_MAPPINGS",[U0]);class hv extends Gt{}J(hv,"MODEL_CLASS_MAPPINGS",[W0]);class _v extends Gt{}J(_v,"MODEL_CLASS_MAPPINGS",[Du]);class fv extends Gt{}J(fv,"MODEL_CLASS_MAPPINGS",[Ou]);class gv extends Gt{}J(gv,"MODEL_CLASS_MAPPINGS",[N0]);class wv extends Gt{}J(wv,"MODEL_CLASS_MAPPINGS",[V0]);class Mv extends Gt{}J(Mv,"MODEL_CLASS_MAPPINGS",[Lu]);class bv extends Gt{}J(bv,"MODEL_CLASS_MAPPINGS",[G0]);class yv extends Gt{}J(yv,"MODEL_CLASS_MAPPINGS",[K0]);class vv extends Gt{}J(vv,"MODEL_CLASS_MAPPINGS",[zu]);class xv extends Gt{}J(xv,"MODEL_CLASS_MAPPINGS",[Q0]);class Tv extends Gt{}J(Tv,"MODEL_CLASS_MAPPINGS",[lo]);class Ev extends Gt{}J(Ev,"MODEL_CLASS_MAPPINGS",[Y0]);class Pv extends Gt{}J(Pv,"MODEL_CLASS_MAPPINGS",[Z0]);class Cv extends Gt{}J(Cv,"MODEL_CLASS_MAPPINGS",[X0]);class Sv extends Gt{}J(Sv,"MODEL_CLASS_MAPPINGS",[J0]);class $v extends Gt{}J($v,"MODEL_CLASS_MAPPINGS",[ev]);class kv extends Gt{}J(kv,"MODEL_CLASS_MAPPINGS",[tv]);class Iv extends Gt{}J(Iv,"MODEL_CLASS_MAPPINGS",[rv]);class Av extends Gt{}J(Av,"MODEL_CLASS_MAPPINGS",[sv]);class Fv extends Gt{}J(Fv,"MODEL_CLASS_MAPPINGS",[nv]);class Ov extends Gt{}J(Ov,"MODEL_CLASS_MAPPINGS",[QT]);class Dv extends Gt{}J(Dv,"MODEL_CLASS_MAPPINGS",[ov]);class Lv extends Gt{}J(Lv,"MODEL_CLASS_MAPPINGS",[av]);class zv extends Gt{}J(zv,"MODEL_CLASS_MAPPINGS",[iv]);class Bv extends Gt{}J(Bv,"MODEL_CLASS_MAPPINGS",[lv]);class Rv extends Gt{}J(Rv,"MODEL_CLASS_MAPPINGS",[cv]);class jv extends Gt{}J(jv,"MODEL_CLASS_MAPPINGS",[uv]);class Nv extends Gt{}J(Nv,"MODEL_CLASS_MAPPINGS",[H0]);class Vv extends Gt{}J(Vv,"MODEL_CLASS_MAPPINGS",[q0]);class YT extends me{constructor({logits:P,past_key_values:L,encoder_outputs:ne,decoder_attentions:ge=null,cross_attentions:fe=null}){super(),this.logits=P,this.past_key_values=L,this.encoder_outputs=ne,this.decoder_attentions=ge,this.cross_attentions=fe}}class Et extends me{constructor({logits:P,...L}){super(),this.logits=P;const ne=Object.values(L);ne.length>0&&(this.attentions=ne)}}class Uv extends me{constructor({logits:P,embeddings:L}){super(),this.logits=P,this.embeddings=L}}class Sr extends me{constructor({logits:P}){super(),this.logits=P}}class Or extends me{constructor({logits:P}){super(),this.logits=P}}class Nr extends me{constructor({start_logits:P,end_logits:L}){super(),this.start_logits=P,this.end_logits=L}}class dn extends me{constructor({logits:P}){super(),this.logits=P}}class ZT extends me{constructor({logits:P,past_key_values:L}){super(),this.logits=P,this.past_key_values=L}}class Wv extends me{constructor({alphas:P}){super(),this.alphas=P}}class Gv extends me{constructor({waveform:P,spectrogram:L}){super(),this.waveform=P,this.spectrogram=L}}},"./src/models/audio_spectrogram_transformer/feature_extraction_audio_spectrogram_transformer.js":(e,r,t)=>{t.r(r),t.d(r,{ASTFeatureExtractor:()=>n});var s=t("./src/base/feature_extraction_utils.js");t("./src/utils/tensor.js");var o=t("./src/utils/audio.js");class n extends s.FeatureExtractor{constructor(i){super(i);const l=this.config.sampling_rate,c=(0,o.mel_filter_bank)(257,this.config.num_mel_bins,20,Math.floor(l/2),l,null,"kaldi",!0);this.mel_filters=c,this.window=(0,o.window_function)(400,"hann",{periodic:!1}),this.mean=this.config.mean,this.std=this.config.std}async _extract_fbank_features(i,l){return(0,o.spectrogram)(i,this.window,400,160,{fft_length:512,power:2,center:!1,preemphasis:.97,mel_filters:this.mel_filters,log_mel:"log",mel_floor:1192092955078125e-22,remove_dc_offset:!0,max_num_frames:l,transpose:!0})}async _call(i){(0,s.validate_audio_inputs)(i,"ASTFeatureExtractor");const l=await this._extract_fbank_features(i,this.config.max_length);if(this.config.do_normalize){const c=this.std*2,p=l.data;for(let d=0;d<p.length;++d)p[d]=(p[d]-this.mean)/c}return{input_values:l.unsqueeze_(0)}}}},"./src/models/auto/feature_extraction_auto.js":(e,r,t)=>{t.r(r),t.d(r,{AutoFeatureExtractor:()=>a});var s=t("./src/utils/constants.js"),o=t("./src/utils/hub.js");t("./src/base/feature_extraction_utils.js");var n=t("./src/models/feature_extractors.js");class a{static async from_pretrained(l,c={}){const p=await(0,o.getModelJSON)(l,s.FEATURE_EXTRACTOR_NAME,!0,c),d=p.feature_extractor_type,u=n[d];if(!u)throw new Error(`Unknown feature_extractor_type: '${d}'. Please report this at ${s.GITHUB_ISSUE_URL}.`);return new u(p)}}},"./src/models/auto/image_processing_auto.js":(e,r,t)=>{t.r(r),t.d(r,{AutoImageProcessor:()=>i});var s=t("./src/utils/constants.js"),o=t("./src/utils/hub.js"),n=t("./src/base/image_processors_utils.js"),a=t("./src/models/image_processors.js");class i{static async from_pretrained(c,p={}){const d=await(0,o.getModelJSON)(c,s.IMAGE_PROCESSOR_NAME,!0,p),u=d.image_processor_type??d.feature_extractor_type;let f=a[u==null?void 0:u.replace(/Fast$/,"")];return f||(u!==void 0&&console.warn(`Image processor type '${u}' not found, assuming base ImageProcessor. Please report this at ${s.GITHUB_ISSUE_URL}.`),f=n.ImageProcessor),new f(d)}}},"./src/models/auto/processing_auto.js":(e,r,t)=>{t.r(r),t.d(r,{AutoProcessor:()=>c});var s=t("./src/utils/constants.js"),o=t("./src/utils/hub.js"),n=t("./src/base/processing_utils.js"),a=t("./src/models/processors.js"),i=t("./src/models/image_processors.js"),l=t("./src/models/feature_extractors.js");class c{static async from_pretrained(d,u={}){const f=await(0,o.getModelJSON)(d,s.IMAGE_PROCESSOR_NAME,!0,u),{image_processor_type:_,feature_extractor_type:b,processor_class:k}=f;if(k&&a[k])return a[k].from_pretrained(d,u);if(!_&&!b)throw new Error("No `image_processor_type` or `feature_extractor_type` found in the config.");const w={};if(_){const $=i[_.replace(/Fast$/,"")];if(!$)throw new Error(`Unknown image_processor_type: '${_}'.`);w.image_processor=new $(f)}if(b){const $=i[b];if($)w.image_processor=new $(f);else{const T=l[b];if(!T)throw new Error(`Unknown feature_extractor_type: '${b}'.`);w.feature_extractor=new T(f)}}const y={};return new n.Processor(y,w,null)}}},"./src/models/beit/image_processing_beit.js":(e,r,t)=>{t.r(r),t.d(r,{BeitFeatureExtractor:()=>o});var s=t("./src/base/image_processors_utils.js");class o extends s.ImageProcessor{}},"./src/models/bit/image_processing_bit.js":(e,r,t)=>{t.r(r),t.d(r,{BitImageProcessor:()=>o});var s=t("./src/base/image_processors_utils.js");class o extends s.ImageProcessor{}},"./src/models/chinese_clip/image_processing_chinese_clip.js":(e,r,t)=>{t.r(r),t.d(r,{ChineseCLIPFeatureExtractor:()=>o});var s=t("./src/base/image_processors_utils.js");class o extends s.ImageProcessor{}},"./src/models/clap/feature_extraction_clap.js":(e,r,t)=>{t.r(r),t.d(r,{ClapFeatureExtractor:()=>n});var s=t("./src/base/feature_extraction_utils.js");t("./src/utils/tensor.js");var o=t("./src/utils/audio.js");class n extends s.FeatureExtractor{constructor(i){super(i),this.mel_filters=(0,o.mel_filter_bank)(this.config.nb_frequency_bins,this.config.feature_size,this.config.frequency_min,this.config.frequency_max,this.config.sampling_rate,null,"htk"),this.mel_filters_slaney=(0,o.mel_filter_bank)(this.config.nb_frequency_bins,this.config.feature_size,this.config.frequency_min,this.config.frequency_max,this.config.sampling_rate,"slaney","slaney"),this.window=(0,o.window_function)(this.config.fft_window_size,"hann")}async _get_input_mel(i,l,c,p){let d;const u=i.length-l;if(u>0)if(c==="rand_trunc"){const f=Math.floor(Math.random()*(u+1));i=i.subarray(f,f+l),d=await this._extract_fbank_features(i,this.mel_filters_slaney,this.config.nb_max_samples)}else throw new Error(`Truncation strategy "${c}" not implemented`);else{if(u<0){let f=new Float64Array(l);if(f.set(i),p==="repeat")for(let _=i.length;_<l;_+=i.length)f.set(i.subarray(0,Math.min(i.length,l-_)),_);else if(p==="repeatpad")for(let _=i.length;_<-u;_+=i.length)f.set(i,_);i=f}if(c==="fusion")throw new Error(`Truncation strategy "${c}" not implemented`);d=await this._extract_fbank_features(i,this.mel_filters_slaney,this.config.nb_max_samples)}return d.unsqueeze_(0)}async _extract_fbank_features(i,l,c=null){return(0,o.spectrogram)(i,this.window,this.config.fft_window_size,this.config.hop_length,{power:2,mel_filters:l,log_mel:"dB",max_num_frames:c,do_pad:!1,transpose:!0})}async _call(i,{max_length:l=null}={}){return(0,s.validate_audio_inputs)(i,"ClapFeatureExtractor"),{input_features:(await this._get_input_mel(i,l??this.config.nb_max_samples,this.config.truncation,this.config.padding)).unsqueeze_(0)}}}},"./src/models/clip/image_processing_clip.js":(e,r,t)=>{t.r(r),t.d(r,{CLIPFeatureExtractor:()=>n,CLIPImageProcessor:()=>o});var s=t("./src/base/image_processors_utils.js");class o extends s.ImageProcessor{}class n extends o{}},"./src/models/convnext/image_processing_convnext.js":(e,r,t)=>{t.r(r),t.d(r,{ConvNextFeatureExtractor:()=>n,ConvNextImageProcessor:()=>o});var s=t("./src/base/image_processors_utils.js");class o extends s.ImageProcessor{constructor(i){super(i),this.crop_pct=this.config.crop_pct??224/256}async resize(i){var c;const l=(c=this.size)==null?void 0:c.shortest_edge;if(l===void 0)throw new Error("Size dictionary must contain 'shortest_edge' key.");if(l<384){const p=Math.floor(l/this.crop_pct),[d,u]=this.get_resize_output_image_size(i,{shortest_edge:p});i=await i.resize(d,u,{resample:this.resample}),i=await i.center_crop(l,l)}else i=await i.resize(l,l,{resample:this.resample});return i}}class n extends o{}},"./src/models/dac/feature_extraction_dac.js":(e,r,t)=>{t.r(r),t.d(r,{DacFeatureExtractor:()=>o});var s=t("./src/models/encodec/feature_extraction_encodec.js");class o extends s.EncodecFeatureExtractor{}},"./src/models/deit/image_processing_deit.js":(e,r,t)=>{t.r(r),t.d(r,{DeiTFeatureExtractor:()=>n,DeiTImageProcessor:()=>o});var s=t("./src/base/image_processors_utils.js");class o extends s.ImageProcessor{}class n extends o{}},"./src/models/detr/image_processing_detr.js":(e,r,t)=>{t.r(r),t.d(r,{DetrFeatureExtractor:()=>a,DetrImageProcessor:()=>n});var s=t("./src/base/image_processors_utils.js"),o=t("./src/utils/tensor.js");class n extends s.ImageProcessor{async _call(l){const c=await super._call(l),p=[c.pixel_values.dims[0],64,64],d=(0,o.full)(p,1n);return{...c,pixel_mask:d}}post_process_object_detection(...l){return(0,s.post_process_object_detection)(...l)}post_process_panoptic_segmentation(...l){return(0,s.post_process_panoptic_segmentation)(...l)}post_process_instance_segmentation(...l){return(0,s.post_process_instance_segmentation)(...l)}}class a extends n{}},"./src/models/dinov3_vit/image_processing_dinov3_vit.js":(e,r,t)=>{t.r(r),t.d(r,{DINOv3ViTImageProcessor:()=>o});var s=t("./src/base/image_processors_utils.js");class o extends s.ImageProcessor{}},"./src/models/donut/image_processing_donut.js":(e,r,t)=>{t.r(r),t.d(r,{DonutFeatureExtractor:()=>n,DonutImageProcessor:()=>o});var s=t("./src/base/image_processors_utils.js");class o extends s.ImageProcessor{pad_image(i,l,c,p={}){const[d,u,f]=l;let _=this.image_mean;Array.isArray(this.image_mean)||(_=new Array(f).fill(_));let b=this.image_std;Array.isArray(b)||(b=new Array(f).fill(_));const k=_.map((w,y)=>-w/b[y]);return super.pad_image(i,l,c,{center:!0,constant_values:k,...p})}}class n extends o{}},"./src/models/dpt/image_processing_dpt.js":(e,r,t)=>{t.r(r),t.d(r,{DPTFeatureExtractor:()=>n,DPTImageProcessor:()=>o});var s=t("./src/base/image_processors_utils.js");class o extends s.ImageProcessor{}class n extends o{}},"./src/models/efficientnet/image_processing_efficientnet.js":(e,r,t)=>{t.r(r),t.d(r,{EfficientNetImageProcessor:()=>o});var s=t("./src/base/image_processors_utils.js");class o extends s.ImageProcessor{constructor(a){super(a),this.include_top=this.config.include_top??!0,this.include_top&&(this.image_std=this.image_std.map(i=>i*i))}}},"./src/models/encodec/feature_extraction_encodec.js":(e,r,t)=>{t.r(r),t.d(r,{EncodecFeatureExtractor:()=>n});var s=t("./src/base/feature_extraction_utils.js"),o=t("./src/utils/tensor.js");class n extends s.FeatureExtractor{async _call(i){(0,s.validate_audio_inputs)(i,"EncodecFeatureExtractor"),i instanceof Float64Array&&(i=new Float32Array(i));const l=this.config.feature_size;if(i.length%l!==0)throw new Error(`The length of the audio data must be a multiple of the number of channels (${l}).`);const c=[1,l,i.length/l];return{input_values:new o.Tensor("float32",i,c)}}}},"./src/models/feature_extractors.js":(e,r,t)=>{t.r(r),t.d(r,{ASTFeatureExtractor:()=>s.ASTFeatureExtractor,ClapFeatureExtractor:()=>n.ClapFeatureExtractor,DacFeatureExtractor:()=>a.DacFeatureExtractor,EncodecFeatureExtractor:()=>o.EncodecFeatureExtractor,Gemma3nAudioFeatureExtractor:()=>i.Gemma3nAudioFeatureExtractor,ImageFeatureExtractor:()=>w.ImageProcessor,MoonshineFeatureExtractor:()=>l.MoonshineFeatureExtractor,ParakeetFeatureExtractor:()=>c.ParakeetFeatureExtractor,PyAnnoteFeatureExtractor:()=>p.PyAnnoteFeatureExtractor,SeamlessM4TFeatureExtractor:()=>d.SeamlessM4TFeatureExtractor,SnacFeatureExtractor:()=>u.SnacFeatureExtractor,SpeechT5FeatureExtractor:()=>f.SpeechT5FeatureExtractor,Wav2Vec2FeatureExtractor:()=>_.Wav2Vec2FeatureExtractor,WeSpeakerFeatureExtractor:()=>b.WeSpeakerFeatureExtractor,WhisperFeatureExtractor:()=>k.WhisperFeatureExtractor});var s=t("./src/models/audio_spectrogram_transformer/feature_extraction_audio_spectrogram_transformer.js"),o=t("./src/models/encodec/feature_extraction_encodec.js"),n=t("./src/models/clap/feature_extraction_clap.js"),a=t("./src/models/dac/feature_extraction_dac.js"),i=t("./src/models/gemma3n/feature_extraction_gemma3n.js"),l=t("./src/models/moonshine/feature_extraction_moonshine.js"),c=t("./src/models/parakeet/feature_extraction_parakeet.js"),p=t("./src/models/pyannote/feature_extraction_pyannote.js"),d=t("./src/models/seamless_m4t/feature_extraction_seamless_m4t.js"),u=t("./src/models/snac/feature_extraction_snac.js"),f=t("./src/models/speecht5/feature_extraction_speecht5.js"),_=t("./src/models/wav2vec2/feature_extraction_wav2vec2.js"),b=t("./src/models/wespeaker/feature_extraction_wespeaker.js"),k=t("./src/models/whisper/feature_extraction_whisper.js"),w=t("./src/base/image_processors_utils.js")},"./src/models/florence2/processing_florence2.js":(e,r,t)=>{t.r(r),t.d(r,{Florence2Processor:()=>a});var s=t("./src/base/processing_utils.js"),o=t("./src/models/auto/image_processing_auto.js"),n=t("./src/tokenizers.js");class a extends s.Processor{constructor(l,c,p){super(l,c,p);const{tasks_answer_post_processing_type:d,task_prompts_without_inputs:u,task_prompts_with_input:f}=this.image_processor.config;this.tasks_answer_post_processing_type=new Map(Object.entries(d??{})),this.task_prompts_without_inputs=new Map(Object.entries(u??{})),this.task_prompts_with_input=new Map(Object.entries(f??{})),this.regexes={quad_boxes:/(.+?)<loc_(\d+)><loc_(\d+)><loc_(\d+)><loc_(\d+)><loc_(\d+)><loc_(\d+)><loc_(\d+)><loc_(\d+)>/gm,bboxes:/([^<]+)?<loc_(\d+)><loc_(\d+)><loc_(\d+)><loc_(\d+)>/gm},this.size_per_bin=1e3}construct_prompts(l){typeof l=="string"&&(l=[l]);const c=[];for(const p of l)if(this.task_prompts_without_inputs.has(p))c.push(this.task_prompts_without_inputs.get(p));else{for(const[d,u]of this.task_prompts_with_input)if(p.includes(d)){c.push(u.replaceAll("{input}",p).replaceAll(d,""));break}c.length!==l.length&&c.push(p)}return c}post_process_generation(l,c,p){const d=this.tasks_answer_post_processing_type.get(c)??"pure_text";l=l.replaceAll("<s>","").replaceAll("</s>","");let u;switch(d){case"pure_text":u=l;break;case"description_with_bboxes":case"bboxes":case"phrase_grounding":case"ocr":const f=d==="ocr"?"quad_boxes":"bboxes",_=l.matchAll(this.regexes[f]),b=[],k=[];for(const[w,y,...$]of _)b.push(y?y.trim():b.at(-1)??""),k.push($.map((T,M)=>(Number(T)+.5)/this.size_per_bin*p[M%2]));u={labels:b,[f]:k};break;default:throw new Error(`Task "${c}" (of type "${d}") not yet implemented.`)}return{[c]:u}}async _call(l,c=null,p={}){if(!l&&!c)throw new Error("Either text or images must be provided");const d=await this.image_processor(l,p),u=c?this.tokenizer(this.construct_prompts(c),p):{};return{...d,...u}}}J(a,"tokenizer_class",n.AutoTokenizer),J(a,"image_processor_class",o.AutoImageProcessor)},"./src/models/gemma3n/feature_extraction_gemma3n.js":(e,r,t)=>{t.r(r),t.d(r,{Gemma3nAudioFeatureExtractor:()=>a});var s=t("./src/base/feature_extraction_utils.js"),o=t("./src/utils/tensor.js"),n=t("./src/utils/audio.js");class a extends s.FeatureExtractor{constructor(l){super(l);const{fft_length:c,feature_size:p,min_frequency:d,max_frequency:u,sampling_rate:f,frame_length:_}=this.config,b=(0,n.mel_filter_bank)(Math.floor(1+c/2),p,d,u,f,null,"htk",!1);this.mel_filters=b,this.window=(0,n.window_function)(_,"hann")}async _extract_fbank_features(l,c){return(0,n.spectrogram)(l,this.window,this.config.frame_length,this.config.hop_length,{fft_length:this.config.fft_length,center:!1,onesided:!0,preemphasis:this.config.preemphasis,preemphasis_htk_flavor:this.config.preemphasis_htk_flavor,mel_filters:this.mel_filters,log_mel:"log",mel_floor:this.config.mel_floor,remove_dc_offset:!1,transpose:!0})}async _call(l,{max_length:c=48e4,truncation:p=!0,padding:d=!0,pad_to_multiple_of:u=128}={}){if((0,s.validate_audio_inputs)(l,"Gemma3nAudioFeatureExtractor"),p&&l.length>c&&(l=l.slice(0,c)),d&&l.length%u!==0){const b=u-l.length%u,k=new Float64Array(l.length+b);k.set(l),this.config.padding_value!==0&&k.fill(this.config.padding_value,l.length),l=k}const f=await this._extract_fbank_features(l,this.config.max_length),_=(0,o.full)([1,f.dims[0]],!0);return{input_features:f.unsqueeze_(0),input_features_mask:_}}}},"./src/models/gemma3n/processing_gemma3n.js":(e,r,t)=>{t.r(r),t.d(r,{Gemma3nProcessor:()=>i});var s=t("./src/base/processing_utils.js"),o=t("./src/models/auto/image_processing_auto.js"),n=t("./src/models/auto/feature_extraction_auto.js"),a=t("./src/tokenizers.js");t("./src/utils/image.js"),t("./src/utils/audio.js");class i extends s.Processor{constructor(c,p,d){super(c,p,d),this.audio_seq_length=this.config.audio_seq_length,this.image_seq_length=this.config.image_seq_length;const{audio_token_id:u,boa_token:f,audio_token:_,eoa_token:b,image_token_id:k,boi_token:w,image_token:y,eoi_token:$}=this.tokenizer.config;this.audio_token_id=u,this.boa_token=f,this.audio_token=_;const T=_.repeat(this.audio_seq_length);this.full_audio_sequence=`

${f}${T}${b}

`,this.image_token_id=k,this.boi_token=w,this.image_token=y;const M=y.repeat(this.image_seq_length);this.full_image_sequence=`

${w}${M}${$}

`}async _call(c,p=null,d=null,u={}){typeof c=="string"&&(c=[c]);let f;d&&(f=await this.feature_extractor(d,u),c=c.map(k=>k.replaceAll(this.audio_token,this.full_audio_sequence)));let _;return p&&(_=await this.image_processor(p,u),c=c.map(k=>k.replaceAll(this.image_token,this.full_image_sequence))),{...this.tokenizer(c,u),..._,...f}}}J(i,"image_processor_class",o.AutoImageProcessor),J(i,"feature_extractor_class",n.AutoFeatureExtractor),J(i,"tokenizer_class",a.AutoTokenizer),J(i,"uses_processor_config",!0),J(i,"uses_chat_template_file",!0)},"./src/models/glpn/image_processing_glpn.js":(e,r,t)=>{t.r(r),t.d(r,{GLPNFeatureExtractor:()=>o});var s=t("./src/base/image_processors_utils.js");class o extends s.ImageProcessor{}},"./src/models/grounding_dino/image_processing_grounding_dino.js":(e,r,t)=>{t.r(r),t.d(r,{GroundingDinoImageProcessor:()=>n});var s=t("./src/base/image_processors_utils.js"),o=t("./src/utils/tensor.js");class n extends s.ImageProcessor{async _call(i){const l=await super._call(i),c=l.pixel_values.dims,p=(0,o.ones)([c[0],c[2],c[3]]);return{...l,pixel_mask:p}}}},"./src/models/grounding_dino/processing_grounding_dino.js":(e,r,t)=>{t.r(r),t.d(r,{GroundingDinoProcessor:()=>l});var s=t("./src/base/processing_utils.js"),o=t("./src/models/auto/image_processing_auto.js"),n=t("./src/tokenizers.js"),a=t("./src/base/image_processors_utils.js");function i(c,p){const u=c.dims.at(-1)-1,f=c.tolist();f.fill(!1,0,1),f.fill(!1,u);const _=p.tolist();return f.map((b,k)=>b?k:null).filter(b=>b!==null).map(b=>_[b])}class l extends s.Processor{async _call(p,d,u={}){const f=p?await this.image_processor(p,u):{};return{...d?this.tokenizer(d,u):{},...f}}post_process_grounded_object_detection(p,d,{box_threshold:u=.25,text_threshold:f=.25,target_sizes:_=null}={}){const{logits:b,pred_boxes:k}=p,w=b.dims[0];if(_!==null&&_.length!==w)throw Error("Make sure that you pass in as many target sizes as the batch dimension of the logits");const y=b.dims.at(1),$=b.sigmoid(),T=$.max(-1).tolist(),M=k.tolist().map(v=>v.map(C=>(0,a.center_to_corners_format)(C))),x=[];for(let v=0;v<w;++v){const C=_!==null?_[v]:null;C!==null&&(M[v]=M[v].map(j=>j.map((se,q)=>se*C[(q+1)%2])));const F=T[v],D=[],K=[],U=[];for(let j=0;j<y;++j){const se=F[j];if(se<=u)continue;const q=M[v][j],re=$[v][j];D.push(se),U.push(q);const Z=i(re.gt(f),d[v]);K.push(Z)}x.push({scores:D,boxes:U,labels:this.batch_decode(K)})}return x}}J(l,"tokenizer_class",n.AutoTokenizer),J(l,"image_processor_class",o.AutoImageProcessor)},"./src/models/idefics3/image_processing_idefics3.js":(e,r,t)=>{t.r(r),t.d(r,{Idefics3ImageProcessor:()=>n});var s=t("./src/base/image_processors_utils.js"),o=t("./src/utils/tensor.js");class n extends s.ImageProcessor{constructor(i){super(i),this.do_image_splitting=i.do_image_splitting??!0,this.max_image_size=i.max_image_size}get_resize_for_vision_encoder(i,l){let[c,p]=i.dims.slice(-2);const d=p/c;return p>=c?(p=Math.ceil(p/l)*l,c=Math.floor(p/d),c=Math.ceil(c/l)*l):(c=Math.ceil(c/l)*l,p=Math.floor(c*d),p=Math.ceil(p/l)*l),{height:c,width:p}}async _call(i,{do_image_splitting:l=null,return_row_col_info:c=!1}={}){let p;if(!Array.isArray(i))p=[[i]];else{if(i.length===0||!i[0])throw new Error("No images provided.");Array.isArray(i[0])?p=i:p=[i]}let d=[],u=[],f=[];const _=[],b=[];for(const v of p){let C=await Promise.all(v.map(K=>this.preprocess(K)));_.push(...C.map(K=>K.original_size)),b.push(...C.map(K=>K.reshaped_input_size)),C.forEach(K=>K.pixel_values.unsqueeze_(0));const{longest_edge:F}=this.max_image_size;let D;if(l??this.do_image_splitting){let K=new Array(C.length),U=new Array(C.length);D=await Promise.all(C.map(async(j,se)=>{const q=this.get_resize_for_vision_encoder(j.pixel_values,F),re=await(0,o.interpolate_4d)(j.pixel_values,{size:[q.height,q.width]}),{frames:Z,num_splits_h:ae,num_splits_w:_e}=await this.split_image(re,this.max_image_size);return K[se]=ae,U[se]=_e,(0,o.cat)(Z,0)})),u.push(K),f.push(U)}else{const K=[F,F];D=await Promise.all(C.map(U=>(0,o.interpolate_4d)(U.pixel_values,{size:K}))),u.push(new Array(C.length).fill(0)),f.push(new Array(C.length).fill(0))}d.push((0,o.cat)(D,0))}const k=d.length,[w,y,$,T]=d[0].dims;let M,x;if(k===1)M=d[0].unsqueeze_(0),x=(0,o.full)([k,w,$,T],!0);else{const v=Math.max(...d.map(D=>D.dims.at(0)));x=(0,o.full)([k,v,$,T],!0);const C=x.data,F=v*$*T;for(let D=0;D<k;++D){const K=d[D].dims[0];if(K<v){d[D]=(0,o.cat)([d[D],(0,o.full)([v-K,y,$,T],0)],0);const U=D*F+K*$*T,j=(D+1)*F;C.fill(!1,U,j)}}M=(0,o.stack)(d,0)}return{pixel_values:M,pixel_attention_mask:x,original_sizes:_,reshaped_input_sizes:b,...c?{rows:u,cols:f}:{}}}async split_image(i,{longest_edge:l}){const c=l,p=l,d=[],[u,f]=i.dims.slice(-2);let _=0,b=0;if(u>c||f>p){_=Math.ceil(u/c),b=Math.ceil(f/p);const k=Math.ceil(u/_),w=Math.ceil(f/b);for(let T=0;T<_;++T)for(let M=0;M<b;++M){let x,v,C,F;T===_-1?(v=u-k,F=u):(v=T*k,F=(T+1)*k),M===b-1?(x=f-w,C=f):(x=M*w,C=(M+1)*w);const D=[v,x],K=[F,C],U=await(0,o.slice)(i,D,K,[2,3]);d.push(U)}const y=c,$=p;(u!==y||f!==$)&&(i=await(0,o.interpolate_4d)(i,{size:[y,$]}))}return d.push(i),{frames:d,num_splits_h:_,num_splits_w:b}}}},"./src/models/idefics3/processing_idefics3.js":(e,r,t)=>{t.r(r),t.d(r,{Idefics3Processor:()=>p});var s=t("./src/base/processing_utils.js"),o=t("./src/models/auto/image_processing_auto.js"),n=t("./src/tokenizers.js");t("./src/utils/image.js");var a=t("./src/utils/core.js");function i(d,u,f,_,b,k){let w="";for(let y=0;y<u;++y){for(let $=0;$<f;++$)w+=_+`<row_${y+1}_col_${$+1}>`+b.repeat(d);w+=`
`}return w+=`
${_}${k}`+b.repeat(d)+`${_}`,w}function l(d,u,f,_){return`${u}${_}`+f.repeat(d)+`${u}`}function c(d,u,f,_,b,k){return d===0&&u===0?l(f,_,b,k):i(f,d,u,_,b,k)}class p extends s.Processor{constructor(){super(...arguments);J(this,"fake_image_token","<fake_token_around_image>");J(this,"image_token","<image>");J(this,"global_img_token","<global-img>")}async _call(f,_=null,b={}){b.return_row_col_info??(b.return_row_col_info=!0);let k;_&&(k=await this.image_processor(_,b)),Array.isArray(f)||(f=[f]);const w=k.rows??[new Array(f.length).fill(0)],y=k.cols??[new Array(f.length).fill(0)],$=this.config.image_seq_len,T=[],M=[];for(let v=0;v<f.length;++v){const C=f[v],F=w[v],D=y[v];T.push((0,a.count)(C,this.image_token));const K=F.map((se,q)=>c(se,D[q],$,this.fake_image_token,this.image_token,this.global_img_token)),U=C.split(this.image_token);if(U.length===0)throw new Error("The image token should be present in the text.");let j=U[0];for(let se=0;se<K.length;++se)j+=K[se]+U[se+1];M.push(j)}return{...this.tokenizer(M),...k}}}J(p,"image_processor_class",o.AutoImageProcessor),J(p,"tokenizer_class",n.AutoTokenizer),J(p,"uses_processor_config",!0)},"./src/models/image_processors.js":(e,r,t)=>{t.r(r),t.d(r,{BeitFeatureExtractor:()=>s.BeitFeatureExtractor,BitImageProcessor:()=>o.BitImageProcessor,CLIPFeatureExtractor:()=>a.CLIPFeatureExtractor,CLIPImageProcessor:()=>a.CLIPImageProcessor,ChineseCLIPFeatureExtractor:()=>n.ChineseCLIPFeatureExtractor,ConvNextFeatureExtractor:()=>i.ConvNextFeatureExtractor,ConvNextImageProcessor:()=>i.ConvNextImageProcessor,DINOv3ViTImageProcessor:()=>p.DINOv3ViTImageProcessor,DPTFeatureExtractor:()=>u.DPTFeatureExtractor,DPTImageProcessor:()=>u.DPTImageProcessor,DeiTFeatureExtractor:()=>l.DeiTFeatureExtractor,DeiTImageProcessor:()=>l.DeiTImageProcessor,DetrFeatureExtractor:()=>c.DetrFeatureExtractor,DetrImageProcessor:()=>c.DetrImageProcessor,DonutFeatureExtractor:()=>d.DonutFeatureExtractor,DonutImageProcessor:()=>d.DonutImageProcessor,EfficientNetImageProcessor:()=>f.EfficientNetImageProcessor,GLPNFeatureExtractor:()=>_.GLPNFeatureExtractor,GroundingDinoImageProcessor:()=>b.GroundingDinoImageProcessor,Idefics3ImageProcessor:()=>k.Idefics3ImageProcessor,JinaCLIPImageProcessor:()=>y.JinaCLIPImageProcessor,LlavaOnevisionImageProcessor:()=>$.LlavaOnevisionImageProcessor,Mask2FormerImageProcessor:()=>T.Mask2FormerImageProcessor,MaskFormerFeatureExtractor:()=>M.MaskFormerFeatureExtractor,MaskFormerImageProcessor:()=>M.MaskFormerImageProcessor,MobileNetV1FeatureExtractor:()=>x.MobileNetV1FeatureExtractor,MobileNetV1ImageProcessor:()=>x.MobileNetV1ImageProcessor,MobileNetV2FeatureExtractor:()=>v.MobileNetV2FeatureExtractor,MobileNetV2ImageProcessor:()=>v.MobileNetV2ImageProcessor,MobileNetV3FeatureExtractor:()=>C.MobileNetV3FeatureExtractor,MobileNetV3ImageProcessor:()=>C.MobileNetV3ImageProcessor,MobileNetV4FeatureExtractor:()=>F.MobileNetV4FeatureExtractor,MobileNetV4ImageProcessor:()=>F.MobileNetV4ImageProcessor,MobileViTFeatureExtractor:()=>D.MobileViTFeatureExtractor,MobileViTImageProcessor:()=>D.MobileViTImageProcessor,NougatImageProcessor:()=>K.NougatImageProcessor,OwlViTFeatureExtractor:()=>j.OwlViTFeatureExtractor,OwlViTImageProcessor:()=>j.OwlViTImageProcessor,Owlv2ImageProcessor:()=>U.Owlv2ImageProcessor,Phi3VImageProcessor:()=>se.Phi3VImageProcessor,PixtralImageProcessor:()=>q.PixtralImageProcessor,PvtImageProcessor:()=>re.PvtImageProcessor,Qwen2VLImageProcessor:()=>Z.Qwen2VLImageProcessor,RTDetrImageProcessor:()=>ae.RTDetrImageProcessor,Sam2ImageProcessor:()=>pe.Sam2ImageProcessor,Sam3ImageProcessor:()=>oe.Sam3ImageProcessor,SamImageProcessor:()=>_e.SamImageProcessor,SegformerFeatureExtractor:()=>G.SegformerFeatureExtractor,SegformerImageProcessor:()=>G.SegformerImageProcessor,SiglipImageProcessor:()=>B.SiglipImageProcessor,SmolVLMImageProcessor:()=>Q.SmolVLMImageProcessor,Swin2SRImageProcessor:()=>le.Swin2SRImageProcessor,VLMImageProcessor:()=>w.VLMImageProcessor,ViTFeatureExtractor:()=>ee.ViTFeatureExtractor,ViTImageProcessor:()=>ee.ViTImageProcessor,VitMatteImageProcessor:()=>Se.VitMatteImageProcessor,VitPoseImageProcessor:()=>De.VitPoseImageProcessor,YolosFeatureExtractor:()=>Ne.YolosFeatureExtractor,YolosImageProcessor:()=>Ne.YolosImageProcessor});var s=t("./src/models/beit/image_processing_beit.js"),o=t("./src/models/bit/image_processing_bit.js"),n=t("./src/models/chinese_clip/image_processing_chinese_clip.js"),a=t("./src/models/clip/image_processing_clip.js"),i=t("./src/models/convnext/image_processing_convnext.js"),l=t("./src/models/deit/image_processing_deit.js"),c=t("./src/models/detr/image_processing_detr.js"),p=t("./src/models/dinov3_vit/image_processing_dinov3_vit.js"),d=t("./src/models/donut/image_processing_donut.js"),u=t("./src/models/dpt/image_processing_dpt.js"),f=t("./src/models/efficientnet/image_processing_efficientnet.js"),_=t("./src/models/glpn/image_processing_glpn.js"),b=t("./src/models/grounding_dino/image_processing_grounding_dino.js"),k=t("./src/models/idefics3/image_processing_idefics3.js"),w=t("./src/models/janus/image_processing_janus.js"),y=t("./src/models/jina_clip/image_processing_jina_clip.js"),$=t("./src/models/llava_onevision/image_processing_llava_onevision.js"),T=t("./src/models/mask2former/image_processing_mask2former.js"),M=t("./src/models/maskformer/image_processing_maskformer.js"),x=t("./src/models/mobilenet_v1/image_processing_mobilenet_v1.js"),v=t("./src/models/mobilenet_v2/image_processing_mobilenet_v2.js"),C=t("./src/models/mobilenet_v3/image_processing_mobilenet_v3.js"),F=t("./src/models/mobilenet_v4/image_processing_mobilenet_v4.js"),D=t("./src/models/mobilevit/image_processing_mobilevit.js"),K=t("./src/models/nougat/image_processing_nougat.js"),U=t("./src/models/owlv2/image_processing_owlv2.js"),j=t("./src/models/owlvit/image_processing_owlvit.js"),se=t("./src/models/phi3_v/image_processing_phi3_v.js"),q=t("./src/models/pixtral/image_processing_pixtral.js"),re=t("./src/models/pvt/image_processing_pvt.js"),Z=t("./src/models/qwen2_vl/image_processing_qwen2_vl.js"),ae=t("./src/models/rt_detr/image_processing_rt_detr.js"),_e=t("./src/models/sam/image_processing_sam.js"),pe=t("./src/models/sam2/image_processing_sam2.js"),oe=t("./src/models/sam3/image_processing_sam3.js"),G=t("./src/models/segformer/image_processing_segformer.js"),B=t("./src/models/siglip/image_processing_siglip.js"),Q=t("./src/models/smolvlm/image_processing_smolvlm.js"),le=t("./src/models/swin2sr/image_processing_swin2sr.js"),ee=t("./src/models/vit/image_processing_vit.js"),Se=t("./src/models/vitmatte/image_processing_vitmatte.js"),De=t("./src/models/vitpose/image_processing_vitpose.js"),Ne=t("./src/models/yolos/image_processing_yolos.js")},"./src/models/janus/image_processing_janus.js":(e,r,t)=>{t.r(r),t.d(r,{VLMImageProcessor:()=>o});var s=t("./src/base/image_processors_utils.js");class o extends s.ImageProcessor{constructor(a){super({do_pad:!0,pad_size:{width:a.image_size,height:a.image_size},...a}),this.constant_values=this.config.background_color.map(i=>i*this.rescale_factor)}pad_image(a,i,l,c){return super.pad_image(a,i,l,{constant_values:this.constant_values,center:!0,...c})}}},"./src/models/janus/processing_janus.js":(e,r,t)=>{t.r(r),t.d(r,{VLChatProcessor:()=>c});var s=t("./src/base/processing_utils.js"),o=t("./src/models/auto/image_processing_auto.js"),n=t("./src/tokenizers.js"),a=t("./src/utils/core.js"),i=t("./src/utils/tensor.js"),l=t("./src/utils/image.js");class c extends s.Processor{constructor(d,u,f){super(d,u,f),this.image_tag=this.config.image_tag,this.image_start_tag=this.config.image_start_tag,this.image_end_tag=this.config.image_end_tag,this.num_image_tokens=this.config.num_image_tokens}async _call(d,{images:u=null,chat_template:f="default"}={}){u?Array.isArray(u)||(u=[u]):u=await Promise.all(d.filter(D=>D.images).flatMap(D=>D.images).map(D=>l.RawImage.read(D)));const _=this.tokenizer,b=_.apply_chat_template(d,{tokenize:!1,add_generation_prompt:!0,chat_template:f}),k=D=>_.encode(D,{add_special_tokens:!1}),w=b.split(this.image_tag),y=w.length-1;if(u.length!==y)throw new Error(`Number of images provided (${u.length}) does not match number of "${this.image_tag}" image tags (${y})`);const[$,T,M]=_.model.convert_tokens_to_ids([this.image_tag,this.image_start_tag,this.image_end_tag]);let x=k(w[0]),v=new Array(x.length).fill(!1);for(let D=1;D<w.length;++D){const K=new Array(this.num_image_tokens).fill($),U=k(w[D]);x=(0,a.mergeArrays)(x,[T],K,[M],U);const j=new Array(this.num_image_tokens).fill(!0);v=(0,a.mergeArrays)(v,[!1],j,[!1],new Array(U.length).fill(!1))}const C=[1,x.length],F={input_ids:new i.Tensor("int64",x,C),attention_mask:new i.Tensor("int64",new Array(x.length).fill(1),C),images_seq_mask:new i.Tensor("bool",v,C),images_emb_mask:new i.Tensor("bool",new Array(y*this.num_image_tokens).fill(!0),[1,y,this.num_image_tokens])};if(u&&u.length>0){const D=await this.image_processor(u);return D.pixel_values.unsqueeze_(0),{...F,...D}}return F}}J(c,"image_processor_class",o.AutoImageProcessor),J(c,"tokenizer_class",n.AutoTokenizer),J(c,"uses_processor_config",!0)},"./src/models/jina_clip/image_processing_jina_clip.js":(e,r,t)=>{t.r(r),t.d(r,{JinaCLIPImageProcessor:()=>o});var s=t("./src/base/image_processors_utils.js");class o extends s.ImageProcessor{constructor(a){const{resize_mode:i,fill_color:l,interpolation:c,size:p,...d}=a,u=i==="squash"?{width:p,height:p}:i==="shortest"?{shortest_edge:p}:{longest_edge:p},f=c==="bicubic"?3:2;super({...d,size:u,resample:f,do_center_crop:!0,crop_size:p,do_normalize:!0})}}},"./src/models/jina_clip/processing_jina_clip.js":(e,r,t)=>{t.r(r),t.d(r,{JinaCLIPProcessor:()=>a});var s=t("./src/base/processing_utils.js"),o=t("./src/models/auto/image_processing_auto.js"),n=t("./src/tokenizers.js");class a extends s.Processor{async _call(l=null,c=null,p={}){if(!l&&!c)throw new Error("Either text or images must be provided");const d=l?this.tokenizer(l,p):{},u=c?await this.image_processor(c,p):{};return{...d,...u}}}J(a,"tokenizer_class",n.AutoTokenizer),J(a,"image_processor_class",o.AutoImageProcessor)},"./src/models/llava/processing_llava.js":(e,r,t)=>{t.r(r),t.d(r,{LlavaProcessor:()=>a});var s=t("./src/base/processing_utils.js"),o=t("./src/models/auto/image_processing_auto.js"),n=t("./src/tokenizers.js");class a extends s.Processor{async _call(l,c=null,p={}){const d=await this.image_processor(l,p);if(c){const[f,_]=d.pixel_values.dims.slice(-2),{image_token:b,patch_size:k,num_additional_image_tokens:w}=this.config,y=Math.floor(f/k)*Math.floor(_/k)+w;c=structuredClone(c),Array.isArray(c)||(c=[c]);for(let $=0;$<c.length;++$)c[$]=c[$].replace(b,b.repeat(y))}const u=c?this.tokenizer(c,p):{};return{...d,...u}}}J(a,"tokenizer_class",n.AutoTokenizer),J(a,"image_processor_class",o.AutoImageProcessor),J(a,"uses_processor_config",!0)},"./src/models/llava_onevision/image_processing_llava_onevision.js":(e,r,t)=>{t.r(r),t.d(r,{LlavaOnevisionImageProcessor:()=>o});var s=t("./src/base/image_processors_utils.js");class o extends s.ImageProcessor{}},"./src/models/mask2former/image_processing_mask2former.js":(e,r,t)=>{t.r(r),t.d(r,{Mask2FormerImageProcessor:()=>o});var s=t("./src/models/maskformer/image_processing_maskformer.js");class o extends s.MaskFormerImageProcessor{}},"./src/models/maskformer/image_processing_maskformer.js":(e,r,t)=>{t.r(r),t.d(r,{MaskFormerFeatureExtractor:()=>n,MaskFormerImageProcessor:()=>o});var s=t("./src/base/image_processors_utils.js");class o extends s.ImageProcessor{post_process_panoptic_segmentation(...i){return(0,s.post_process_panoptic_segmentation)(...i)}post_process_instance_segmentation(...i){return(0,s.post_process_instance_segmentation)(...i)}}class n extends o{}},"./src/models/mgp_str/processing_mgp_str.js":(e,r,t)=>{t.r(r),t.d(r,{MgpstrProcessor:()=>l});var s=t("./src/base/processing_utils.js"),o=t("./src/models/auto/image_processing_auto.js"),n=t("./src/tokenizers.js"),a=t("./src/utils/maths.js");const i={char:["char_decode",1],bpe:["bpe_decode",2],wp:["wp_decode",102]};class l extends s.Processor{get char_tokenizer(){return this.components.char_tokenizer}get bpe_tokenizer(){return this.components.bpe_tokenizer}get wp_tokenizer(){return this.components.wp_tokenizer}_decode_helper(p,d){if(!i.hasOwnProperty(d))throw new Error(`Format ${d} is not supported.`);const[u,f]=i[d],_=this[u].bind(this),[b,k]=p.dims,w=[],y=[],$=p.tolist();for(let M=0;M<b;++M){const x=$[M],v=[],C=[];for(let D=1;D<k;++D){const[K,U]=(0,a.max)((0,a.softmax)(x[D]));if(C.push(K),U==f)break;v.push(U)}const F=C.length>0?C.reduce((D,K)=>D*K,1):0;y.push(v),w.push(F)}return[_(y),w]}char_decode(p){return this.char_tokenizer.batch_decode(p).map(d=>d.replaceAll(" ",""))}bpe_decode(p){return this.bpe_tokenizer.batch_decode(p)}wp_decode(p){return this.wp_tokenizer.batch_decode(p).map(d=>d.replaceAll(" ",""))}batch_decode([p,d,u]){const[f,_]=this._decode_helper(p,"char"),[b,k]=this._decode_helper(d,"bpe"),[w,y]=this._decode_helper(u,"wp"),$=[],T=[];for(let M=0;M<f.length;++M){const[x,v]=(0,a.max)([_[M],k[M],y[M]]);$.push([f[M],b[M],w[M]][v]),T.push(x)}return{generated_text:$,scores:T,char_preds:f,bpe_preds:b,wp_preds:w}}static async from_pretrained(...p){const d=await super.from_pretrained(...p),u=await n.AutoTokenizer.from_pretrained("Xenova/gpt2"),f=await n.AutoTokenizer.from_pretrained("Xenova/bert-base-uncased");return d.components={image_processor:d.image_processor,char_tokenizer:d.tokenizer,bpe_tokenizer:u,wp_tokenizer:f},d}async _call(p,d=null){const u=await this.image_processor(p);return d&&(u.labels=this.tokenizer(d).input_ids),u}}J(l,"tokenizer_class",n.AutoTokenizer),J(l,"image_processor_class",o.AutoImageProcessor)},"./src/models/mobilenet_v1/image_processing_mobilenet_v1.js":(e,r,t)=>{t.r(r),t.d(r,{MobileNetV1FeatureExtractor:()=>n,MobileNetV1ImageProcessor:()=>o});var s=t("./src/base/image_processors_utils.js");class o extends s.ImageProcessor{}class n extends o{}},"./src/models/mobilenet_v2/image_processing_mobilenet_v2.js":(e,r,t)=>{t.r(r),t.d(r,{MobileNetV2FeatureExtractor:()=>n,MobileNetV2ImageProcessor:()=>o});var s=t("./src/base/image_processors_utils.js");class o extends s.ImageProcessor{}class n extends o{}},"./src/models/mobilenet_v3/image_processing_mobilenet_v3.js":(e,r,t)=>{t.r(r),t.d(r,{MobileNetV3FeatureExtractor:()=>n,MobileNetV3ImageProcessor:()=>o});var s=t("./src/base/image_processors_utils.js");class o extends s.ImageProcessor{}class n extends o{}},"./src/models/mobilenet_v4/image_processing_mobilenet_v4.js":(e,r,t)=>{t.r(r),t.d(r,{MobileNetV4FeatureExtractor:()=>n,MobileNetV4ImageProcessor:()=>o});var s=t("./src/base/image_processors_utils.js");class o extends s.ImageProcessor{}class n extends o{}},"./src/models/mobilevit/image_processing_mobilevit.js":(e,r,t)=>{t.r(r),t.d(r,{MobileViTFeatureExtractor:()=>n,MobileViTImageProcessor:()=>o});var s=t("./src/base/image_processors_utils.js");class o extends s.ImageProcessor{}class n extends o{}},"./src/models/moonshine/feature_extraction_moonshine.js":(e,r,t)=>{t.r(r),t.d(r,{MoonshineFeatureExtractor:()=>n});var s=t("./src/base/feature_extraction_utils.js"),o=t("./src/utils/tensor.js");class n extends s.FeatureExtractor{async _call(i){(0,s.validate_audio_inputs)(i,"MoonshineFeatureExtractor"),i instanceof Float64Array&&(i=new Float32Array(i));const l=[1,i.length];return{input_values:new o.Tensor("float32",i,l)}}}},"./src/models/moonshine/processing_moonshine.js":(e,r,t)=>{t.r(r),t.d(r,{MoonshineProcessor:()=>a});var s=t("./src/models/auto/feature_extraction_auto.js"),o=t("./src/tokenizers.js"),n=t("./src/base/processing_utils.js");class a extends n.Processor{async _call(l){return await this.feature_extractor(l)}}J(a,"tokenizer_class",o.AutoTokenizer),J(a,"feature_extractor_class",s.AutoFeatureExtractor)},"./src/models/nougat/image_processing_nougat.js":(e,r,t)=>{t.r(r),t.d(r,{NougatImageProcessor:()=>o});var s=t("./src/models/donut/image_processing_donut.js");class o extends s.DonutImageProcessor{}},"./src/models/owlv2/image_processing_owlv2.js":(e,r,t)=>{t.r(r),t.d(r,{Owlv2ImageProcessor:()=>o});var s=t("./src/models/owlvit/image_processing_owlvit.js");class o extends s.OwlViTImageProcessor{}},"./src/models/owlvit/image_processing_owlvit.js":(e,r,t)=>{t.r(r),t.d(r,{OwlViTFeatureExtractor:()=>n,OwlViTImageProcessor:()=>o});var s=t("./src/base/image_processors_utils.js");class o extends s.ImageProcessor{post_process_object_detection(...i){return(0,s.post_process_object_detection)(...i)}}class n extends o{}},"./src/models/owlvit/processing_owlvit.js":(e,r,t)=>{t.r(r),t.d(r,{OwlViTProcessor:()=>a});var s=t("./src/base/processing_utils.js"),o=t("./src/models/auto/image_processing_auto.js"),n=t("./src/tokenizers.js");class a extends s.Processor{}J(a,"tokenizer_class",n.AutoTokenizer),J(a,"image_processor_class",o.AutoImageProcessor)},"./src/models/paligemma/processing_paligemma.js":(e,r,t)=>{t.r(r),t.d(r,{PaliGemmaProcessor:()=>l});var s=t("./src/base/processing_utils.js"),o=t("./src/models/auto/image_processing_auto.js"),n=t("./src/tokenizers.js");const a="<image>";function i(c,p,d,u,f){return`${u.repeat(d*f)}${p}${c}
`}class l extends s.Processor{async _call(p,d=null,u={}){d||(console.warn("You are using PaliGemma without a text prefix. It will perform as a picture-captioning model."),d=""),Array.isArray(p)||(p=[p]),Array.isArray(d)||(d=[d]);const f=this.tokenizer.bos_token,_=this.image_processor.config.image_seq_length;let b;d.some(y=>y.includes(a))?b=d.map(y=>{const $=y.replaceAll(a,a.repeat(_)),T=$.lastIndexOf(a),M=T===-1?0:T+a.length;return $.slice(0,M)+f+$.slice(M)+`
`}):(console.warn("You are passing both `text` and `images` to `PaliGemmaProcessor`. The processor expects special image tokens in the text, as many tokens as there are images per each text. It is recommended to add `<image>` tokens in the very beginning of your text. For this call, we will infer how many images each text has and add special tokens."),b=d.map(y=>i(y,f,_,a,p.length)));const k=this.tokenizer(b,u);return{...await this.image_processor(p,u),...k}}}J(l,"tokenizer_class",n.AutoTokenizer),J(l,"image_processor_class",o.AutoImageProcessor),J(l,"uses_processor_config",!1)},"./src/models/parakeet/feature_extraction_parakeet.js":(e,r,t)=>{t.r(r),t.d(r,{ParakeetFeatureExtractor:()=>i});var s=t("./src/base/feature_extraction_utils.js"),o=t("./src/utils/tensor.js"),n=t("./src/utils/audio.js");const a=1e-5;class i extends s.FeatureExtractor{constructor(c){var u;super(c),(u=this.config).mel_filters??(u.mel_filters=(0,n.mel_filter_bank)(Math.floor(1+this.config.n_fft/2),this.config.feature_size,0,this.config.sampling_rate/2,this.config.sampling_rate,"slaney","slaney"));const p=(0,n.window_function)(this.config.win_length,"hann",{periodic:!1});this.window=new Float64Array(this.config.n_fft);const d=Math.floor((this.config.n_fft-this.config.win_length)/2);this.window.set(p,d)}async _extract_fbank_features(c){const p=this.config.preemphasis;c=new Float64Array(c);for(let u=c.length-1;u>=1;--u)c[u]-=p*c[u-1];return await(0,n.spectrogram)(c,this.window,this.window.length,this.config.hop_length,{fft_length:this.config.n_fft,power:2,mel_filters:this.config.mel_filters,log_mel:"log",mel_floor:-1/0,pad_mode:"constant",center:!0,transpose:!0,mel_offset:2**-24})}async _call(c){(0,s.validate_audio_inputs)(c,"ParakeetFeatureExtractor");const p=await this._extract_fbank_features(c),d=Math.floor((c.length+Math.floor(this.config.n_fft/2)*2-this.config.n_fft)/this.config.hop_length),u=p.data;u.fill(0,d*p.dims[1]);const[f,_]=p.dims,b=new Float64Array(_),k=new Float64Array(_);for(let $=0;$<d;++$){const T=$*_;for(let M=0;M<_;++M){const x=u[T+M];b[M]+=x,k[M]+=x*x}}const w=d>1?d-1:1;for(let $=0;$<_;++$){const T=b[$]/d,M=(k[$]-d*T*T)/w,v=1/(Math.sqrt(M)+a);for(let C=0;C<d;++C){const F=C*_+$;u[F]=(u[F]-T)*v}}const y=new BigInt64Array(f);return y.fill(1n,0,d),{input_features:p.unsqueeze_(0),attention_mask:new o.Tensor("int64",y,[1,f])}}}},"./src/models/phi3_v/image_processing_phi3_v.js":(e,r,t)=>{t.r(r),t.d(r,{Phi3VImageProcessor:()=>p});var s=t("./src/base/image_processors_utils.js"),o=t("./src/utils/tensor.js");const n=336,a=[2,3],{ceil:i,floor:l,sqrt:c}=Math;class p extends s.ImageProcessor{constructor(u){super({...u,do_normalize:!0,do_pad:!0,pad_size:"custom",do_convert_rgb:!0,do_resize:!0}),this._num_crops=u.num_crops}calc_num_image_tokens_from_image_size(u,f){const{num_img_tokens:_}=this.config;return l((l(f/n)*l(u/n)+1)*_+1+(l(f/n)+1)*c(_))}get_resize_output_image_size(u,f){const _=this._num_crops,[b,k]=u.size;let w=b/k,y=1;for(;y*Math.ceil(y/w)<=_;)y+=1;y-=1;const $=Math.floor(y*336),T=Math.floor($/w);return[$,T]}pad_image(u,f,_,b={}){const[k,w]=f,y=n*i(k/n),$=n*i(w/n),T=[1,1,1].map((M,x)=>(M-this.image_mean[x])/this.image_std[x]);return super.pad_image(u,f,{width:$,height:y},{center:!0,constant_values:T,...b})}async _call(u,{num_crops:f=null}={}){if(this._num_crops=f??(f=this.config.num_crops),f<4||c(f)%1!==0)throw new Error("num_crops must be a square number >= 4");Array.isArray(u)||(u=[u]);const _=u.length,b=await Promise.all(u.map(v=>this.preprocess(v))),k=b.map(v=>v.original_size),w=b.map(v=>v.reshaped_input_size),y=[];for(const{pixel_values:v}of b){v.unsqueeze_(0);const[C,F]=v.dims.slice(-2),D=await(0,o.interpolate_4d)(v,{size:[n,n],mode:"bicubic"});if(f>0){const K=[],U=c(f),j=l(F/U),se=l(C/U);for(let re=0;re<U;++re)for(let Z=0;Z<U;++Z){let ae,_e,pe,oe;re===U-1?(_e=C-se,oe=C):(_e=re*se,oe=(re+1)*se),Z===U-1?(ae=F-j,pe=F):(ae=Z*j,pe=(Z+1)*j);const G=[_e,ae],B=[oe,pe],Q=await(0,o.slice)(v,G,B,a);K.push(Q)}const q=await(0,o.interpolate_4d)((0,o.cat)(K,0),{size:[n,n],mode:"bicubic"});y.push((0,o.cat)([D,q],0))}else y.push(D)}const $=(0,o.stack)(y,0),T=w.map(v=>v.map(C=>n*i(C/n))),M=new o.Tensor("int64",T.flat(),[_,2]),x=T.map(([v,C])=>this.calc_num_image_tokens_from_image_size(C,v));return{pixel_values:$,original_sizes:k,reshaped_input_sizes:w,image_sizes:M,num_img_tokens:x}}}},"./src/models/phi3_v/processing_phi3_v.js":(e,r,t)=>{t.r(r),t.d(r,{Phi3VProcessor:()=>l});var s=t("./src/base/processing_utils.js"),o=t("./src/models/auto/image_processing_auto.js"),n=t("./src/tokenizers.js");t("./src/utils/image.js");const a="<|image|>",i=/<\|image_\d+\|>/g;class l extends s.Processor{async _call(p,d=null,{padding:u=!0,truncation:f=!0,num_crops:_=null}={}){Array.isArray(p)||(p=[p]);let b,k;if(d){k=await this.image_processor(d,{num_crops:_});const{num_img_tokens:w}=k,y=p.map((T,M)=>T.split(i).join(a.repeat(w[M])));b=this.tokenizer(y,{padding:u,truncation:f});const $=this.tokenizer.model.convert_tokens_to_ids([a])[0];b.input_ids.map_(T=>T==$?-T:T)}else b=this.tokenizer(p);return{...b,...k}}}J(l,"image_processor_class",o.AutoImageProcessor),J(l,"tokenizer_class",n.AutoTokenizer)},"./src/models/pixtral/image_processing_pixtral.js":(e,r,t)=>{t.r(r),t.d(r,{PixtralImageProcessor:()=>o});var s=t("./src/base/image_processors_utils.js");class o extends s.ImageProcessor{get_resize_output_image_size(a,i){const{longest_edge:l}=i;if(l===void 0)throw new Error("size must contain 'longest_edge'");const[c,p]=a.size,d=Math.max(c,p)/l;let u=c,f=p;d>1&&(u=Math.floor(c/d),f=Math.floor(p/d));const{patch_size:_,spatial_merge_size:b}=this.config;if(!b)throw new Error("config must contain 'spatial_merge_size'");const k=_*b,w=Math.floor((u-1)/k)+1,y=Math.floor((f-1)/k)+1;return[w*k,y*k]}}},"./src/models/pixtral/processing_pixtral.js":(e,r,t)=>{t.r(r),t.d(r,{PixtralProcessor:()=>a});var s=t("./src/base/processing_utils.js"),o=t("./src/models/auto/image_processing_auto.js"),n=t("./src/tokenizers.js");class a extends s.Processor{async _call(l,c=null,p={}){const d=await this.image_processor(l,p);if(c){const[f,_]=d.pixel_values.dims.slice(-2),{image_token:b,image_break_token:k,image_end_token:w,patch_size:y,spatial_merge_size:$}=this.config,T=y*$,M=Math.floor(f/T),x=Math.floor(_/T);c=structuredClone(c),Array.isArray(c)||(c=[c]);for(let v=0;v<c.length;++v){const C=b.repeat(x),F=C+k,D=C+w,K=F.repeat(M-1)+D;c[v]=c[v].replace(b,K)}}const u=c?this.tokenizer(c,p):{};return{...d,...u}}}J(a,"tokenizer_class",n.AutoTokenizer),J(a,"image_processor_class",o.AutoImageProcessor),J(a,"uses_processor_config",!0)},"./src/models/processors.js":(e,r,t)=>{t.r(r),t.d(r,{Florence2Processor:()=>s.Florence2Processor,Gemma3nProcessor:()=>o.Gemma3nProcessor,GroundingDinoProcessor:()=>n.GroundingDinoProcessor,Idefics3Processor:()=>a.Idefics3Processor,JinaCLIPProcessor:()=>l.JinaCLIPProcessor,LlavaProcessor:()=>c.LlavaProcessor,MgpstrProcessor:()=>p.MgpstrProcessor,MoonshineProcessor:()=>d.MoonshineProcessor,OwlViTProcessor:()=>u.OwlViTProcessor,PaliGemmaProcessor:()=>f.PaliGemmaProcessor,Phi3VProcessor:()=>_.Phi3VProcessor,PixtralProcessor:()=>b.PixtralProcessor,PyAnnoteProcessor:()=>k.PyAnnoteProcessor,Qwen2VLProcessor:()=>w.Qwen2VLProcessor,Sam2Processor:()=>$.Sam2Processor,Sam2VideoProcessor:()=>$.Sam2VideoProcessor,SamProcessor:()=>y.SamProcessor,SmolVLMProcessor:()=>T.SmolVLMProcessor,SpeechT5Processor:()=>M.SpeechT5Processor,UltravoxProcessor:()=>x.UltravoxProcessor,VLChatProcessor:()=>i.VLChatProcessor,VoxtralProcessor:()=>v.VoxtralProcessor,Wav2Vec2Processor:()=>C.Wav2Vec2Processor,Wav2Vec2ProcessorWithLM:()=>F.Wav2Vec2ProcessorWithLM,WhisperProcessor:()=>D.WhisperProcessor});var s=t("./src/models/florence2/processing_florence2.js"),o=t("./src/models/gemma3n/processing_gemma3n.js"),n=t("./src/models/grounding_dino/processing_grounding_dino.js"),a=t("./src/models/idefics3/processing_idefics3.js"),i=t("./src/models/janus/processing_janus.js"),l=t("./src/models/jina_clip/processing_jina_clip.js"),c=t("./src/models/llava/processing_llava.js"),p=t("./src/models/mgp_str/processing_mgp_str.js"),d=t("./src/models/moonshine/processing_moonshine.js"),u=t("./src/models/owlvit/processing_owlvit.js"),f=t("./src/models/paligemma/processing_paligemma.js"),_=t("./src/models/phi3_v/processing_phi3_v.js"),b=t("./src/models/pixtral/processing_pixtral.js"),k=t("./src/models/pyannote/processing_pyannote.js"),w=t("./src/models/qwen2_vl/processing_qwen2_vl.js"),y=t("./src/models/sam/processing_sam.js"),$=t("./src/models/sam2/processing_sam2.js"),T=t("./src/models/smolvlm/processing_smolvlm.js"),M=t("./src/models/speecht5/processing_speecht5.js"),x=t("./src/models/ultravox/processing_ultravox.js"),v=t("./src/models/voxtral/processing_voxtral.js"),C=t("./src/models/wav2vec2/processing_wav2vec2.js"),F=t("./src/models/wav2vec2_with_lm/processing_wav2vec2_with_lm.js"),D=t("./src/models/whisper/processing_whisper.js")},"./src/models/pvt/image_processing_pvt.js":(e,r,t)=>{t.r(r),t.d(r,{PvtImageProcessor:()=>o});var s=t("./src/base/image_processors_utils.js");class o extends s.ImageProcessor{}},"./src/models/pyannote/feature_extraction_pyannote.js":(e,r,t)=>{t.r(r),t.d(r,{PyAnnoteFeatureExtractor:()=>a});var s=t("./src/base/feature_extraction_utils.js"),o=t("./src/utils/tensor.js"),n=t("./src/utils/maths.js");class a extends s.FeatureExtractor{async _call(l){(0,s.validate_audio_inputs)(l,"PyAnnoteFeatureExtractor"),l instanceof Float64Array&&(l=new Float32Array(l));const c=[1,1,l.length];return{input_values:new o.Tensor("float32",l,c)}}samples_to_frames(l){return(l-this.config.offset)/this.config.step}post_process_speaker_diarization(l,c){const p=c/this.samples_to_frames(c)/this.config.sampling_rate,d=[];for(const u of l.tolist()){const f=[];let _=-1;for(let b=0;b<u.length;++b){const k=(0,n.softmax)(u[b]),[w,y]=(0,n.max)(k),[$,T]=[b,b+1];y!==_?(_=y,f.push({id:y,start:$,end:T,score:w})):(f.at(-1).end=T,f.at(-1).score+=w)}d.push(f.map(({id:b,start:k,end:w,score:y})=>({id:b,start:k*p,end:w*p,confidence:y/(w-k)})))}return d}}},"./src/models/pyannote/processing_pyannote.js":(e,r,t)=>{t.r(r),t.d(r,{PyAnnoteProcessor:()=>n});var s=t("./src/base/processing_utils.js"),o=t("./src/models/pyannote/feature_extraction_pyannote.js");class n extends s.Processor{async _call(i){return await this.feature_extractor(i)}post_process_speaker_diarization(...i){return this.feature_extractor.post_process_speaker_diarization(...i)}get sampling_rate(){return this.feature_extractor.config.sampling_rate}}J(n,"feature_extractor_class",o.PyAnnoteFeatureExtractor)},"./src/models/qwen2_vl/image_processing_qwen2_vl.js":(e,r,t)=>{t.r(r),t.d(r,{Qwen2VLImageProcessor:()=>n});var s=t("./src/base/image_processors_utils.js"),o=t("./src/utils/tensor.js");class n extends s.ImageProcessor{async _call(i,...l){const{pixel_values:c,original_sizes:p,reshaped_input_sizes:d}=await super._call(i,...l);let u=c;const{temporal_patch_size:f,merge_size:_,patch_size:b}=this.config;u.dims[0]===1&&(u=(0,o.cat)(Array.from({length:f},()=>u),0));const k=u.dims[0]/f,w=u.dims[1],y=Math.floor(u.dims[2]/b),$=Math.floor(u.dims[3]/b),T=u.view(k,f,w,Math.floor(y/_),_,b,Math.floor($/_),_,b).permute(0,3,6,4,7,2,1,5,8).view(k*y*$,w*f*b*b),M=new o.Tensor("int64",[k,y,$],[1,3]);return{pixel_values:T,image_grid_thw:M,original_sizes:p,reshaped_input_sizes:d}}}},"./src/models/qwen2_vl/processing_qwen2_vl.js":(e,r,t)=>{t.r(r),t.d(r,{Qwen2VLProcessor:()=>a});var s=t("./src/base/processing_utils.js"),o=t("./src/models/auto/image_processing_auto.js"),n=t("./src/tokenizers.js");t("./src/utils/image.js");class a extends s.Processor{async _call(l,c=null,...p){Array.isArray(l)||(l=[l]);let d,u;if(c&&(d=await this.image_processor(c),u=d.image_grid_thw),u){let _=this.image_processor.config.merge_size**2,b=0;const k=u.tolist();l=l.map(w=>{for(;w.includes("<|image_pad|>");){const y=Number(k[b++].reduce(($,T)=>$*T,1n));w=w.replace("<|image_pad|>","<|placeholder|>".repeat(Math.floor(y/_)))}return w.replaceAll("<|placeholder|>","<|image_pad|>")})}return{...this.tokenizer(l),...d}}}J(a,"image_processor_class",o.AutoImageProcessor),J(a,"tokenizer_class",n.AutoTokenizer)},"./src/models/rt_detr/image_processing_rt_detr.js":(e,r,t)=>{t.r(r),t.d(r,{RTDetrImageProcessor:()=>o});var s=t("./src/base/image_processors_utils.js");class o extends s.ImageProcessor{post_process_object_detection(...a){return(0,s.post_process_object_detection)(...a)}}},"./src/models/sam/image_processing_sam.js":(e,r,t)=>{t.r(r),t.d(r,{SamImageProcessor:()=>a});var s=t("./src/base/image_processors_utils.js"),o=t("./src/utils/core.js"),n=t("./src/utils/tensor.js");class a extends s.ImageProcessor{reshape_input_points(l,c,p,d=!1){l=structuredClone(l);let u=(0,o.calculateDimensions)(l);if(u.length===3)d||(u=[1,...u]),l=[l];else if(u.length!==4)throw Error("The input_points must be a 4D tensor of shape `batch_size`, `point_batch_size`, `nb_points_per_image`, `2`.");for(let f=0;f<l.length;++f){const[_,b]=c[f],[k,w]=p[f],y=[w/b,k/_];for(let $=0;$<l[f].length;++$)for(let T=0;T<l[f][$].length;++T)for(let M=0;M<l[f][$][T].length;++M)l[f][$][T][M]*=y[M%2]}return new n.Tensor("float32",Float32Array.from(l.flat(1/0)),u)}add_input_labels(l,c){let p=(0,o.calculateDimensions)(l);if(p.length===2)p=[1,...p],l=[l];else if(p.length!==3)throw Error("The input_points must be a 4D tensor of shape `batch_size`, `point_batch_size`, `nb_points_per_image`, `2`.");if(p.some((d,u)=>d!==c.dims[u]))throw Error(`The first ${p.length} dimensions of 'input_points' and 'input_labels' must be the same.`);return new n.Tensor("int64",l.flat(1/0).map(BigInt),p)}async _call(l,{input_points:c=null,input_labels:p=null,input_boxes:d=null}={}){const u=await super._call(l);if(c&&(u.input_points=this.reshape_input_points(c,u.original_sizes,u.reshaped_input_sizes)),p){if(!u.input_points)throw Error("`input_points` must be provided if `input_labels` are provided.");u.input_labels=this.add_input_labels(p,u.input_points)}return d&&(u.input_boxes=this.reshape_input_points(d,u.original_sizes,u.reshaped_input_sizes,!0)),u}async post_process_masks(l,c,p,{mask_threshold:d=0,binarize:u=!0,pad_size:f=null}={}){const _=[];f=f??this.pad_size??this.size;const b=[f.height,f.width];for(let k=0;k<c.length;++k){const w=c[k],y=p[k];let $=await(0,n.interpolate_4d)(l[k],{mode:"bilinear",size:b});if($=$.slice(null,null,[0,y[0]],[0,y[1]]),$=await(0,n.interpolate_4d)($,{mode:"bilinear",size:w}),u){const T=$.data,M=new Uint8Array(T.length);for(let x=0;x<T.length;++x)T[x]>d&&(M[x]=1);$=new n.Tensor("bool",M,$.dims)}_.push($)}return _}generate_crop_boxes(l,c,{crop_n_layers:p=0,overlap_ratio:d=512/1500,points_per_crop:u=32,crop_n_points_downscale_factor:f=1}={}){}}},"./src/models/sam/processing_sam.js":(e,r,t)=>{t.r(r),t.d(r,{SamProcessor:()=>n});var s=t("./src/base/processing_utils.js"),o=t("./src/models/auto/image_processing_auto.js");class n extends s.Processor{async _call(...i){return await this.image_processor(...i)}post_process_masks(...i){return this.image_processor.post_process_masks(...i)}reshape_input_points(...i){return this.image_processor.reshape_input_points(...i)}}J(n,"image_processor_class",o.AutoImageProcessor)},"./src/models/sam2/image_processing_sam2.js":(e,r,t)=>{t.r(r),t.d(r,{Sam2ImageProcessor:()=>s.SamImageProcessor});var s=t("./src/models/sam/image_processing_sam.js")},"./src/models/sam2/processing_sam2.js":(e,r,t)=>{t.r(r),t.d(r,{Sam2Processor:()=>o,Sam2VideoProcessor:()=>n});var s=t("./src/models/sam/processing_sam.js");class o extends s.SamProcessor{}class n extends o{}},"./src/models/sam3/image_processing_sam3.js":(e,r,t)=>{t.r(r),t.d(r,{Sam3ImageProcessor:()=>s.Sam2ImageProcessor});var s=t("./src/models/sam2/image_processing_sam2.js")},"./src/models/seamless_m4t/feature_extraction_seamless_m4t.js":(e,r,t)=>{t.r(r),t.d(r,{SeamlessM4TFeatureExtractor:()=>a});var s=t("./src/base/feature_extraction_utils.js"),o=t("./src/utils/tensor.js"),n=t("./src/utils/audio.js");class a extends s.FeatureExtractor{constructor(l){super(l);const c=this.config.sampling_rate,p=(0,n.mel_filter_bank)(257,this.config.num_mel_bins,20,Math.floor(c/2),c,null,"kaldi",!0);this.mel_filters=p,this.window=(0,n.window_function)(400,"povey",{periodic:!1})}async _extract_fbank_features(l,c){return l=l.map(p=>p*32768),(0,n.spectrogram)(l,this.window,400,160,{fft_length:512,power:2,center:!1,preemphasis:.97,mel_filters:this.mel_filters,log_mel:"log",mel_floor:1192092955078125e-22,remove_dc_offset:!0,max_num_frames:c,transpose:!0})}async _call(l,{padding:c=!0,pad_to_multiple_of:p=2,do_normalize_per_mel_bins:d=!0,return_attention_mask:u=!0}={}){(0,s.validate_audio_inputs)(l,"SeamlessM4TFeatureExtractor");let f=await this._extract_fbank_features(l,this.config.max_length);if(d){const[M,x]=f.dims,v=f.data;for(let C=0;C<x;++C){let F=0;for(let j=0;j<M;++j)F+=v[j*x+C];const D=F/M;let K=0;for(let j=0;j<M;++j)K+=(v[j*x+C]-D)**2;K/=M-1;const U=Math.sqrt(K+1e-7);for(let j=0;j<M;++j){const se=j*x+C;v[se]=(v[se]-D)/U}}}let _;if(c){const[M,x]=f.dims,v=f.data,C=M%p;if(C>0){const F=new Float32Array(x*(M+C));F.set(v),F.fill(this.config.padding_value,v.length);const D=M+C;f=new o.Tensor(f.type,F,[D,x]),u&&(_=new o.Tensor("int64",new BigInt64Array(D),[1,D]),_.data.fill(1n,0,M))}}const[b,k]=f.dims,w=this.config.stride;if(b%w!==0)throw new Error(`The number of frames (${b}) must be a multiple of the stride (${w}).`);const $=f.view(1,Math.floor(b/w),k*w),T={input_features:$};if(u){const M=$.dims[1],x=new BigInt64Array(M);if(_){const v=_.data;for(let C=1,F=0;C<b;C+=w,++F)x[F]=v[C]}else x.fill(1n);T.attention_mask=new o.Tensor("int64",x,[1,M])}return T}}},"./src/models/segformer/image_processing_segformer.js":(e,r,t)=>{t.r(r),t.d(r,{SegformerFeatureExtractor:()=>n,SegformerImageProcessor:()=>o});var s=t("./src/base/image_processors_utils.js");class o extends s.ImageProcessor{post_process_semantic_segmentation(...i){return(0,s.post_process_semantic_segmentation)(...i)}}class n extends o{}},"./src/models/siglip/image_processing_siglip.js":(e,r,t)=>{t.r(r),t.d(r,{SiglipImageProcessor:()=>o});var s=t("./src/base/image_processors_utils.js");class o extends s.ImageProcessor{}},"./src/models/smolvlm/image_processing_smolvlm.js":(e,r,t)=>{t.r(r),t.d(r,{SmolVLMImageProcessor:()=>s.Idefics3ImageProcessor});var s=t("./src/models/idefics3/image_processing_idefics3.js")},"./src/models/smolvlm/processing_smolvlm.js":(e,r,t)=>{t.r(r),t.d(r,{SmolVLMProcessor:()=>s.Idefics3Processor});var s=t("./src/models/idefics3/processing_idefics3.js")},"./src/models/snac/feature_extraction_snac.js":(e,r,t)=>{t.r(r),t.d(r,{SnacFeatureExtractor:()=>o});var s=t("./src/models/dac/feature_extraction_dac.js");class o extends s.DacFeatureExtractor{}},"./src/models/speecht5/feature_extraction_speecht5.js":(e,r,t)=>{t.r(r),t.d(r,{SpeechT5FeatureExtractor:()=>o});var s=t("./src/base/feature_extraction_utils.js");class o extends s.FeatureExtractor{}},"./src/models/speecht5/processing_speecht5.js":(e,r,t)=>{t.r(r),t.d(r,{SpeechT5Processor:()=>a});var s=t("./src/base/processing_utils.js"),o=t("./src/tokenizers.js"),n=t("./src/models/auto/feature_extraction_auto.js");class a extends s.Processor{async _call(l){return await this.feature_extractor(l)}}J(a,"tokenizer_class",o.AutoTokenizer),J(a,"feature_extractor_class",n.AutoFeatureExtractor)},"./src/models/swin2sr/image_processing_swin2sr.js":(e,r,t)=>{t.r(r),t.d(r,{Swin2SRImageProcessor:()=>o});var s=t("./src/base/image_processors_utils.js");class o extends s.ImageProcessor{pad_image(a,i,l,c={}){const[p,d,u]=i;return super.pad_image(a,i,{width:d+(l-d%l)%l,height:p+(l-p%l)%l},{mode:"symmetric",center:!1,constant_values:-1,...c})}}},"./src/models/ultravox/processing_ultravox.js":(e,r,t)=>{t.r(r),t.d(r,{UltravoxProcessor:()=>a});var s=t("./src/models/auto/feature_extraction_auto.js"),o=t("./src/tokenizers.js"),n=t("./src/base/processing_utils.js");class a extends n.Processor{async _call(l,c=null,p={}){if(Array.isArray(l))throw new Error("Batched inputs are not supported yet.");let d={};if(c){const f=c.length,{input_features:_}=await this.feature_extractor(c,{...p,max_length:f}),b=Math.round(f/this.config.encoder_ds_factor+1e-4),k=1+Math.ceil(b/this.config.stack_factor);d.audio_token_len=[k],d.audio_values=_;const w=this.config.audio_placeholder;if(!l.includes(w))throw new Error(`The input text does not contain the image token ${w}.`);l=l.replaceAll(w,w.repeat(k))}return{...this.tokenizer(l,{add_special_tokens:!1,...p}),...d}}}J(a,"tokenizer_class",o.AutoTokenizer),J(a,"feature_extractor_class",s.AutoFeatureExtractor),J(a,"uses_processor_config",!0)},"./src/models/vit/image_processing_vit.js":(e,r,t)=>{t.r(r),t.d(r,{ViTFeatureExtractor:()=>n,ViTImageProcessor:()=>o});var s=t("./src/base/image_processors_utils.js");class o extends s.ImageProcessor{}class n extends o{}},"./src/models/vitmatte/image_processing_vitmatte.js":(e,r,t)=>{t.r(r),t.d(r,{VitMatteImageProcessor:()=>n});var s=t("./src/base/image_processors_utils.js"),o=t("./src/utils/tensor.js");class n extends s.ImageProcessor{async _call(i,l){Array.isArray(i)||(i=[i]),Array.isArray(l)||(l=[l]);const c=await Promise.all(i.map(u=>this.preprocess(u))),p=await Promise.all(l.map(u=>this.preprocess(u,{do_normalize:!1,do_convert_rgb:!1,do_convert_grayscale:!0})));return{pixel_values:(0,o.stack)(c.map((u,f)=>(0,o.cat)([u.pixel_values,p[f].pixel_values],0)),0),original_sizes:c.map(u=>u.original_size),reshaped_input_sizes:c.map(u=>u.reshaped_input_size)}}}},"./src/models/vitpose/image_processing_vitpose.js":(e,r,t)=>{t.r(r),t.d(r,{VitPoseImageProcessor:()=>o});var s=t("./src/base/image_processors_utils.js");class o extends s.ImageProcessor{post_process_pose_estimation(a,i,{threshold:l=null}={}){const c=a.tolist(),[p,d,u,f]=a.dims,_=[];for(let b=0;b<p;++b){const k=c[b],w=i[b],y=[];for(let $=0;$<w.length;++$){const T=w[$],M=[],x=[],v=[],C=T.at(-2)/f,F=T.at(-1)/u;for(let D=0;D<k.length;++D){let[K,U]=[0,0],j=0,se=-1/0;const q=k[D];for(let Z=0;Z<q.length;++Z){const ae=q[Z];for(let _e=0;_e<ae.length;++_e){const pe=ae[_e];j+=pe,se=Math.max(se,pe),K+=(_e+.5)*pe,U+=Z*pe}}if(l!=null&&se<l)continue;const re=[C*K/j,F*U/j];M.push(re),v.push(D),x.push(se)}y.push({bbox:T,scores:x,labels:v,keypoints:M})}_.push(y)}return _}}},"./src/models/voxtral/processing_voxtral.js":(e,r,t)=>{t.r(r),t.d(r,{VoxtralProcessor:()=>d});var s=t("./src/models/auto/feature_extraction_auto.js"),o=t("./src/tokenizers.js"),n=t("./src/base/processing_utils.js"),a=t("./src/utils/tensor.js");const i="[AUDIO]",l="[BEGIN_AUDIO]",c=375;function p(u,f){const _=[];for(let b=0;b<u.length;b+=f)_.push(u.subarray(b,Math.min(b+f,u.length)));return _}class d extends n.Processor{async _call(f,_=null,b={}){if(Array.isArray(f))throw new Error("Batched inputs are not supported yet.");const k={};if(_){if(!f.includes(i))throw new Error(`The input text does not contain the audio token ${i}.`);Array.isArray(_)||(_=[_]);const y=f.split(i),$=y.length-1;if($!==_.length)throw new Error(`The number of audio inputs (${_.length}) does not match the number of audio tokens in the text (${$}).`);const T=this.feature_extractor.config.n_samples,M=_.map(D=>p(D,T)),x=M.map(D=>D.length),v=M.flat(),C=(await Promise.all(v.map(D=>this.feature_extractor(D,b)))).map(D=>D.input_features);k.audio_values=C.length>1?(0,a.cat)(C,0):C[0];let F=y[0];for(let D=0;D<x.length;++D){F+=l;for(let K=0;K<x[D];++K)F+=i.repeat(c);F+=y[D+1]}f=F}return{...this.tokenizer(f,{add_special_tokens:!1,...b}),...k}}}J(d,"tokenizer_class",o.AutoTokenizer),J(d,"feature_extractor_class",s.AutoFeatureExtractor),J(d,"uses_processor_config",!1)},"./src/models/wav2vec2/feature_extraction_wav2vec2.js":(e,r,t)=>{t.r(r),t.d(r,{Wav2Vec2FeatureExtractor:()=>n});var s=t("./src/base/feature_extraction_utils.js"),o=t("./src/utils/tensor.js");class n extends s.FeatureExtractor{_zero_mean_unit_var_norm(i){const c=i.reduce((d,u)=>d+u,0)/i.length,p=i.reduce((d,u)=>d+(u-c)**2,0)/i.length;return i.map(d=>(d-c)/Math.sqrt(p+1e-7))}async _call(i){(0,s.validate_audio_inputs)(i,"Wav2Vec2FeatureExtractor"),i instanceof Float64Array&&(i=new Float32Array(i));let l=i;this.config.do_normalize&&(l=this._zero_mean_unit_var_norm(l));const c=[1,l.length];return{input_values:new o.Tensor("float32",l,c),attention_mask:new o.Tensor("int64",new BigInt64Array(l.length).fill(1n),c)}}}},"./src/models/wav2vec2/processing_wav2vec2.js":(e,r,t)=>{t.r(r),t.d(r,{Wav2Vec2Processor:()=>a});var s=t("./src/tokenizers.js"),o=t("./src/models/auto/feature_extraction_auto.js"),n=t("./src/base/processing_utils.js");class a extends n.Processor{async _call(l){return await this.feature_extractor(l)}}J(a,"tokenizer_class",s.AutoTokenizer),J(a,"feature_extractor_class",o.AutoFeatureExtractor)},"./src/models/wav2vec2_with_lm/processing_wav2vec2_with_lm.js":(e,r,t)=>{t.r(r),t.d(r,{Wav2Vec2ProcessorWithLM:()=>a});var s=t("./src/tokenizers.js"),o=t("./src/models/auto/feature_extraction_auto.js"),n=t("./src/base/processing_utils.js");class a extends n.Processor{async _call(l){return await this.feature_extractor(l)}}J(a,"tokenizer_class",s.AutoTokenizer),J(a,"feature_extractor_class",o.AutoFeatureExtractor)},"./src/models/wespeaker/feature_extraction_wespeaker.js":(e,r,t)=>{t.r(r),t.d(r,{WeSpeakerFeatureExtractor:()=>n});var s=t("./src/base/feature_extraction_utils.js");t("./src/utils/tensor.js");var o=t("./src/utils/audio.js");class n extends s.FeatureExtractor{constructor(i){super(i);const l=this.config.sampling_rate,c=(0,o.mel_filter_bank)(257,this.config.num_mel_bins,20,Math.floor(l/2),l,null,"kaldi",!0);this.mel_filters=c,this.window=(0,o.window_function)(400,"hamming",{periodic:!1}),this.min_num_frames=this.config.min_num_frames}async _extract_fbank_features(i){return i=i.map(l=>l*32768),(0,o.spectrogram)(i,this.window,400,160,{fft_length:512,power:2,center:!1,preemphasis:.97,mel_filters:this.mel_filters,log_mel:"log",mel_floor:1192092955078125e-22,remove_dc_offset:!0,transpose:!0,min_num_frames:this.min_num_frames})}async _call(i){(0,s.validate_audio_inputs)(i,"WeSpeakerFeatureExtractor");const l=(await this._extract_fbank_features(i)).unsqueeze_(0);if(this.config.fbank_centering_span===null){const c=l.mean(1).data,p=l.data,[d,u,f]=l.dims;for(let _=0;_<d;++_){const b=_*u*f,k=_*f;for(let w=0;w<u;++w){const y=b+w*f;for(let $=0;$<f;++$)p[y+$]-=c[k+$]}}}return{input_features:l}}}},"./src/models/whisper/common_whisper.js":(e,r,t)=>{t.r(r),t.d(r,{WHISPER_LANGUAGE_MAPPING:()=>o,WHISPER_TO_LANGUAGE_CODE_MAPPING:()=>n,whisper_language_to_code:()=>a});const s=[["en","english"],["zh","chinese"],["de","german"],["es","spanish"],["ru","russian"],["ko","korean"],["fr","french"],["ja","japanese"],["pt","portuguese"],["tr","turkish"],["pl","polish"],["ca","catalan"],["nl","dutch"],["ar","arabic"],["sv","swedish"],["it","italian"],["id","indonesian"],["hi","hindi"],["fi","finnish"],["vi","vietnamese"],["he","hebrew"],["uk","ukrainian"],["el","greek"],["ms","malay"],["cs","czech"],["ro","romanian"],["da","danish"],["hu","hungarian"],["ta","tamil"],["no","norwegian"],["th","thai"],["ur","urdu"],["hr","croatian"],["bg","bulgarian"],["lt","lithuanian"],["la","latin"],["mi","maori"],["ml","malayalam"],["cy","welsh"],["sk","slovak"],["te","telugu"],["fa","persian"],["lv","latvian"],["bn","bengali"],["sr","serbian"],["az","azerbaijani"],["sl","slovenian"],["kn","kannada"],["et","estonian"],["mk","macedonian"],["br","breton"],["eu","basque"],["is","icelandic"],["hy","armenian"],["ne","nepali"],["mn","mongolian"],["bs","bosnian"],["kk","kazakh"],["sq","albanian"],["sw","swahili"],["gl","galician"],["mr","marathi"],["pa","punjabi"],["si","sinhala"],["km","khmer"],["sn","shona"],["yo","yoruba"],["so","somali"],["af","afrikaans"],["oc","occitan"],["ka","georgian"],["be","belarusian"],["tg","tajik"],["sd","sindhi"],["gu","gujarati"],["am","amharic"],["yi","yiddish"],["lo","lao"],["uz","uzbek"],["fo","faroese"],["ht","haitian creole"],["ps","pashto"],["tk","turkmen"],["nn","nynorsk"],["mt","maltese"],["sa","sanskrit"],["lb","luxembourgish"],["my","myanmar"],["bo","tibetan"],["tl","tagalog"],["mg","malagasy"],["as","assamese"],["tt","tatar"],["haw","hawaiian"],["ln","lingala"],["ha","hausa"],["ba","bashkir"],["jw","javanese"],["su","sundanese"]],o=new Map(s),n=new Map([...s.map(([i,l])=>[l,i]),["burmese","my"],["valencian","ca"],["flemish","nl"],["haitian","ht"],["letzeburgesch","lb"],["pushto","ps"],["panjabi","pa"],["moldavian","ro"],["moldovan","ro"],["sinhalese","si"],["castilian","es"]]);function a(i){i=i.toLowerCase();let l=n.get(i);if(l===void 0){const c=i.match(/^<\|([a-z]{2})\|>$/);if(c&&(i=c[1]),o.has(i))l=i;else{const d=i.length===2?o.keys():o.values();throw new Error(`Language "${i}" is not supported. Must be one of: ${JSON.stringify(Array.from(d))}`)}}return l}},"./src/models/whisper/feature_extraction_whisper.js":(e,r,t)=>{t.r(r),t.d(r,{WhisperFeatureExtractor:()=>a});var s=t("./src/base/feature_extraction_utils.js");t("./src/utils/tensor.js");var o=t("./src/utils/audio.js"),n=t("./src/utils/maths.js");class a extends s.FeatureExtractor{constructor(l){var c;super(l),(c=this.config).mel_filters??(c.mel_filters=(0,o.mel_filter_bank)(Math.floor(1+this.config.n_fft/2),this.config.feature_size,0,8e3,this.config.sampling_rate,"slaney","slaney")),this.window=(0,o.window_function)(this.config.n_fft,"hann")}async _extract_fbank_features(l){const c=await(0,o.spectrogram)(l,this.window,this.config.n_fft,this.config.hop_length,{power:2,mel_filters:this.config.mel_filters,log_mel:"log10",max_num_frames:Math.min(Math.floor(l.length/this.config.hop_length),this.config.nb_max_frames)}),p=c.data,d=(0,n.max)(p)[0];for(let u=0;u<p.length;++u)p[u]=(Math.max(p[u],d-8)+4)/4;return c}async _call(l,{max_length:c=null}={}){(0,s.validate_audio_inputs)(l,"WhisperFeatureExtractor");let p;const d=c??this.config.n_samples;return l.length>d?(l.length>this.config.n_samples&&console.warn("Attempting to extract features for audio longer than 30 seconds. If using a pipeline to extract transcript from a long audio clip, remember to specify `chunk_length_s` and/or `stride_length_s`."),p=l.slice(0,d)):(p=new Float32Array(d),p.set(l)),{input_features:(await this._extract_fbank_features(p)).unsqueeze_(0)}}}},"./src/models/whisper/generation_whisper.js":(e,r,t)=>{t.r(r),t.d(r,{WhisperGenerationConfig:()=>o});var s=t("./src/generation/configuration_utils.js");class o extends s.GenerationConfig{constructor(){super(...arguments);J(this,"return_timestamps",null);J(this,"return_token_timestamps",null);J(this,"num_frames",null);J(this,"alignment_heads",null);J(this,"task",null);J(this,"language",null);J(this,"no_timestamps_token_id",null);J(this,"prompt_ids",null);J(this,"is_multilingual",null);J(this,"lang_to_id",null);J(this,"task_to_id",null);J(this,"max_initial_timestamp_index",1)}}},"./src/models/whisper/processing_whisper.js":(e,r,t)=>{t.r(r),t.d(r,{WhisperProcessor:()=>a});var s=t("./src/models/auto/feature_extraction_auto.js"),o=t("./src/tokenizers.js"),n=t("./src/base/processing_utils.js");class a extends n.Processor{async _call(l){return await this.feature_extractor(l)}}J(a,"tokenizer_class",o.AutoTokenizer),J(a,"feature_extractor_class",s.AutoFeatureExtractor)},"./src/models/yolos/image_processing_yolos.js":(e,r,t)=>{t.r(r),t.d(r,{YolosFeatureExtractor:()=>n,YolosImageProcessor:()=>o});var s=t("./src/base/image_processors_utils.js");class o extends s.ImageProcessor{post_process_object_detection(...i){return(0,s.post_process_object_detection)(...i)}}class n extends o{}},"./src/ops/registry.js":(e,r,t)=>{t.r(r),t.d(r,{TensorOpRegistry:()=>a});var s=t("./src/backends/onnx.js"),o=t("./src/utils/tensor.js");const n=async(i,l,c)=>{const p=await(0,s.createInferenceSession)(new Uint8Array(i),l);return async d=>{const u=(0,s.isONNXProxy)(),f=Object.fromEntries(Object.entries(d).map(([b,k])=>[b,(u?k.clone():k).ort_tensor])),_=await(0,s.runInferenceSession)(p,f);return Array.isArray(c)?c.map(b=>new o.Tensor(_[b])):new o.Tensor(_[c])}};class a{static get nearest_interpolate_4d(){return this._nearest_interpolate_4d||(this._nearest_interpolate_4d=n([8,10,18,0,58,129,1,10,41,10,1,120,10,0,10,0,10,1,115,18,1,121,34,6,82,101,115,105,122,101,42,18,10,4,109,111,100,101,34,7,110,101,97,114,101,115,116,160,1,3,18,1,114,90,31,10,1,120,18,26,10,24,8,1,18,20,10,3,18,1,98,10,3,18,1,99,10,3,18,1,104,10,3,18,1,119,90,15,10,1,115,18,10,10,8,8,7,18,4,10,2,8,4,98,31,10,1,121,18,26,10,24,8,1,18,20,10,3,18,1,98,10,3,18,1,99,10,3,18,1,104,10,3,18,1,119,66,2,16,21],this.session_options,"y")),this._nearest_interpolate_4d}static get bilinear_interpolate_4d(){return this._bilinear_interpolate_4d||(this._bilinear_interpolate_4d=n([8,9,18,0,58,128,1,10,40,10,1,120,10,0,10,0,10,1,115,18,1,121,34,6,82,101,115,105,122,101,42,17,10,4,109,111,100,101,34,6,108,105,110,101,97,114,160,1,3,18,1,114,90,31,10,1,120,18,26,10,24,8,1,18,20,10,3,18,1,98,10,3,18,1,99,10,3,18,1,104,10,3,18,1,119,90,15,10,1,115,18,10,10,8,8,7,18,4,10,2,8,4,98,31,10,1,121,18,26,10,24,8,1,18,20,10,3,18,1,98,10,3,18,1,99,10,3,18,1,104,10,3,18,1,119,66,2,16,20],this.session_options,"y")),this._bilinear_interpolate_4d}static get bicubic_interpolate_4d(){return this._bicubic_interpolate_4d||(this._bicubic_interpolate_4d=n([8,9,18,0,58,127,10,39,10,1,120,10,0,10,0,10,1,115,18,1,121,34,6,82,101,115,105,122,101,42,16,10,4,109,111,100,101,34,5,99,117,98,105,99,160,1,3,18,1,114,90,31,10,1,120,18,26,10,24,8,1,18,20,10,3,18,1,98,10,3,18,1,99,10,3,18,1,104,10,3,18,1,119,90,15,10,1,115,18,10,10,8,8,7,18,4,10,2,8,4,98,31,10,1,121,18,26,10,24,8,1,18,20,10,3,18,1,98,10,3,18,1,99,10,3,18,1,104,10,3,18,1,119,66,2,16,20],this.session_options,"y")),this._bicubic_interpolate_4d}static get matmul(){return this._matmul||(this._matmul=n([8,9,18,0,58,55,10,17,10,1,97,10,1,98,18,1,99,34,6,77,97,116,77,117,108,18,1,114,90,9,10,1,97,18,4,10,2,8,1,90,9,10,1,98,18,4,10,2,8,1,98,9,10,1,99,18,4,10,2,8,1,66,2,16,20],this.session_options,"c")),this._matmul}static get stft(){return this._stft||(this._stft=n([8,7,18,0,58,148,1,10,38,10,1,115,10,1,106,10,1,119,10,1,108,18,1,111,34,4,83,84,70,84,42,15,10,8,111,110,101,115,105,100,101,100,24,1,160,1,2,18,1,115,90,26,10,1,115,18,21,10,19,8,1,18,15,10,3,18,1,98,10,3,18,1,115,10,3,18,1,99,90,11,10,1,106,18,6,10,4,8,7,18,0,90,16,10,1,119,18,11,10,9,8,1,18,5,10,3,18,1,119,90,11,10,1,108,18,6,10,4,8,7,18,0,98,31,10,1,111,18,26,10,24,8,1,18,20,10,3,18,1,98,10,3,18,1,102,10,3,18,1,100,10,3,18,1,99,66,2,16,17],this.session_options,"o")),this._stft}static get rfft(){return this._rfft||(this._rfft=n([8,9,18,0,58,97,10,33,10,1,120,10,0,10,1,97,18,1,121,34,3,68,70,84,42,15,10,8,111,110,101,115,105,100,101,100,24,1,160,1,2,18,1,100,90,21,10,1,120,18,16,10,14,8,1,18,10,10,3,18,1,115,10,3,18,1,99,90,11,10,1,97,18,6,10,4,8,7,18,0,98,21,10,1,121,18,16,10,14,8,1,18,10,10,3,18,1,115,10,3,18,1,99,66,2,16,20],this.session_options,"y")),this._rfft}static get top_k(){return this._top_k||(this._top_k=n([8,10,18,0,58,73,10,18,10,1,120,10,1,107,18,1,118,18,1,105,34,4,84,111,112,75,18,1,116,90,9,10,1,120,18,4,10,2,8,1,90,15,10,1,107,18,10,10,8,8,7,18,4,10,2,8,1,98,9,10,1,118,18,4,10,2,8,1,98,9,10,1,105,18,4,10,2,8,7,66,2,16,21],this.session_options,["v","i"])),this._top_k}static get slice(){return this._slice||(this._slice=n([8,7,18,0,58,96,10,25,10,1,120,10,1,115,10,1,101,10,1,97,10,1,116,18,1,121,34,5,83,108,105,99,101,18,1,114,90,9,10,1,120,18,4,10,2,8,1,90,9,10,1,115,18,4,10,2,8,7,90,9,10,1,101,18,4,10,2,8,7,90,9,10,1,97,18,4,10,2,8,7,90,9,10,1,116,18,4,10,2,8,7,98,9,10,1,121,18,4,10,2,8,1,66,2,16,13],this.session_options,"y")),this._slice}}J(a,"session_options",{})},"./src/pipelines.js":(e,r,t)=>{t.r(r),t.d(r,{AudioClassificationPipeline:()=>U,AutomaticSpeechRecognitionPipeline:()=>se,BackgroundRemovalPipeline:()=>ae,DepthEstimationPipeline:()=>le,DocumentQuestionAnsweringPipeline:()=>G,FeatureExtractionPipeline:()=>D,FillMaskPipeline:()=>$,ImageClassificationPipeline:()=>re,ImageFeatureExtractionPipeline:()=>K,ImageSegmentationPipeline:()=>Z,ImageToImagePipeline:()=>Q,ImageToTextPipeline:()=>q,ObjectDetectionPipeline:()=>pe,Pipeline:()=>b,QuestionAnsweringPipeline:()=>y,SummarizationPipeline:()=>M,Text2TextGenerationPipeline:()=>T,TextClassificationPipeline:()=>k,TextGenerationPipeline:()=>C,TextToAudioPipeline:()=>B,TokenClassificationPipeline:()=>w,TranslationPipeline:()=>x,ZeroShotAudioClassificationPipeline:()=>j,ZeroShotClassificationPipeline:()=>F,ZeroShotImageClassificationPipeline:()=>_e,ZeroShotObjectDetectionPipeline:()=>oe,pipeline:()=>De});var s=t("./src/tokenizers.js"),o=t("./src/models.js"),n=t("./src/models/auto/processing_auto.js");t("./src/base/processing_utils.js");var a=t("./src/utils/generic.js"),i=t("./src/utils/core.js"),l=t("./src/utils/maths.js"),c=t("./src/utils/audio.js"),p=t("./src/utils/tensor.js"),d=t("./src/utils/image.js");async function u(Ae){return Array.isArray(Ae)||(Ae=[Ae]),await Promise.all(Ae.map(te=>d.RawImage.read(te)))}async function f(Ae,te){return Array.isArray(Ae)||(Ae=[Ae]),await Promise.all(Ae.map(O=>typeof O=="string"||O instanceof URL?(0,c.read_audio)(O,te):O instanceof Float64Array?new Float32Array(O):O))}function _(Ae,te){te&&(Ae=Ae.map(Ce=>Ce|0));const[O,me,xe,be]=Ae;return{xmin:O,ymin:me,xmax:xe,ymax:be}}class b extends a.Callable{constructor({task:te,model:O,tokenizer:me=null,processor:xe=null}){super(),this.task=te,this.model=O,this.tokenizer=me,this.processor=xe}async dispose(){await this.model.dispose()}}class k extends b{constructor(te){super(te)}async _call(te,{top_k:O=1}={}){const me=this.tokenizer(te,{padding:!0,truncation:!0}),xe=await this.model(me),be=this.model.config.problem_type==="multi_label_classification"?Me=>Me.sigmoid():Me=>new p.Tensor("float32",(0,l.softmax)(Me.data),Me.dims),Ce=this.model.config.id2label,$e=[];for(const Me of xe.logits){const Le=be(Me),we=await(0,p.topk)(Le,O),ke=we[0].tolist(),Fe=we[1].tolist().map((Ze,We)=>({label:Ce?Ce[Ze]:`LABEL_${Ze}`,score:ke[We]}));O===1?$e.push(...Fe):$e.push(Fe)}return Array.isArray(te)||O===1?$e:$e[0]}}class w extends b{constructor(te){super(te)}async _call(te,{ignore_labels:O=["O"]}={}){const me=Array.isArray(te),xe=this.tokenizer(me?te:[te],{padding:!0,truncation:!0}),Ce=(await this.model(xe)).logits,$e=this.model.config.id2label,Me=[];for(let Le=0;Le<Ce.dims[0];++Le){const we=xe.input_ids[Le],ke=Ce[Le],He=[];for(let Fe=0;Fe<ke.dims[0];++Fe){const Ze=ke[Fe],We=(0,l.max)(Ze.data)[1],yt=$e?$e[We]:`LABEL_${We}`;if(O.includes(yt))continue;const Qe=this.tokenizer.decode([we[Fe].item()],{skip_special_tokens:!0});if(Qe==="")continue;const Oe=(0,l.softmax)(Ze.data);He.push({entity:yt,score:Oe[We],index:Fe,word:Qe})}Me.push(He)}return me?Me:Me[0]}}class y extends b{constructor(te){super(te)}async _call(te,O,{top_k:me=1}={}){const xe=this.tokenizer(te,{text_pair:O,padding:!0,truncation:!0}),{start_logits:be,end_logits:Ce}=await this.model(xe),$e=xe.input_ids.tolist(),Me=xe.attention_mask.tolist(),Le=this.tokenizer.all_special_ids,we=[];for(let ke=0;ke<be.dims[0];++ke){const He=$e[ke],Fe=He.findIndex(ze=>ze==this.tokenizer.sep_token_id);Me[ke].map((ze,pt)=>ze==1&&(pt===0||pt>Fe&&Le.findIndex(ft=>ft==He[pt])===-1));const Ze=be[ke].tolist(),We=Ce[ke].tolist();for(let ze=1;ze<Ze.length;++ze)(Me[ke]==0||ze<=Fe||Le.findIndex(pt=>pt==He[ze])!==-1)&&(Ze[ze]=-1/0,We[ze]=-1/0);const yt=(0,l.softmax)(Ze).map((ze,pt)=>[ze,pt]),Qe=(0,l.softmax)(We).map((ze,pt)=>[ze,pt]);yt[0][0]=0,Qe[0][0]=0;const Oe=(0,i.product)(yt,Qe).filter(ze=>ze[0][1]<=ze[1][1]).map(ze=>[ze[0][1],ze[1][1],ze[0][0]*ze[1][0]]).sort((ze,pt)=>pt[2]-ze[2]);for(let ze=0;ze<Math.min(Oe.length,me);++ze){const[pt,ft,Lt]=Oe[ze],kt=He.slice(pt,ft+1),Mr=this.tokenizer.decode(kt,{skip_special_tokens:!0});we.push({answer:Mr,score:Lt})}}return me===1?we[0]:we}}class $ extends b{constructor(te){super(te)}async _call(te,{top_k:O=5}={}){const me=this.tokenizer(te,{padding:!0,truncation:!0}),{logits:xe}=await this.model(me),be=[],Ce=me.input_ids.tolist();for(let $e=0;$e<Ce.length;++$e){const Me=Ce[$e],Le=Me.findIndex(Ze=>Ze==this.tokenizer.mask_token_id);if(Le===-1)throw Error(`Mask token (${this.tokenizer.mask_token}) not found in text.`);const we=xe[$e][Le],ke=await(0,p.topk)(new p.Tensor("float32",(0,l.softmax)(we.data),we.dims),O),He=ke[0].tolist(),Fe=ke[1].tolist();be.push(Fe.map((Ze,We)=>{const yt=Me.slice();return yt[Le]=Ze,{score:He[We],token:Number(Ze),token_str:this.tokenizer.decode([Ze]),sequence:this.tokenizer.decode(yt,{skip_special_tokens:!0})}}))}return Array.isArray(te)?be:be[0]}}class T extends b{constructor(O){super(O);J(this,"_key","generated_text")}async _call(O,me={}){Array.isArray(O)||(O=[O]),this.model.config.prefix&&(O=O.map(Le=>this.model.config.prefix+Le));const xe=this.model.config.task_specific_params;xe&&xe[this.task]&&xe[this.task].prefix&&(O=O.map(Le=>xe[this.task].prefix+Le));const be=this.tokenizer,Ce={padding:!0,truncation:!0};let $e;this instanceof x&&"_build_translation_inputs"in be?$e=be._build_translation_inputs(O,Ce,me):$e=be(O,Ce);const Me=await this.model.generate({...$e,...me});return be.batch_decode(Me,{skip_special_tokens:!0}).map(Le=>({[this._key]:Le}))}}class M extends T{constructor(O){super(O);J(this,"_key","summary_text")}}class x extends T{constructor(O){super(O);J(this,"_key","translation_text")}}function v(Ae){return Array.isArray(Ae)&&Ae.every(te=>"role"in te&&"content"in te)}class C extends b{constructor(te){super(te)}async _call(te,O={}){let me=!1,xe=!1,be=O.add_special_tokens??(this.tokenizer.add_bos_token||this.tokenizer.add_eos_token)??!1,Ce;if(typeof te=="string")Ce=te=[te];else if(Array.isArray(te)&&te.every(Fe=>typeof Fe=="string"))me=!0,Ce=te;else{if(v(te))te=[te];else if(Array.isArray(te)&&te.every(v))me=!0;else throw new Error("Input must be a string, an array of strings, a Chat, or an array of Chats");xe=!0,Ce=te.map(Fe=>this.tokenizer.apply_chat_template(Fe,{tokenize:!1,add_generation_prompt:!0})),be=!1}const $e=xe?!1:O.return_full_text??!0;this.tokenizer.padding_side="left";const Me=this.tokenizer(Ce,{add_special_tokens:be,padding:!0,truncation:!0}),Le=await this.model.generate({...Me,...O}),we=this.tokenizer.batch_decode(Le,{skip_special_tokens:!0});let ke;!$e&&Me.input_ids.dims.at(-1)>0&&(ke=this.tokenizer.batch_decode(Me.input_ids,{skip_special_tokens:!0}).map(Fe=>Fe.length));const He=Array.from({length:te.length},Fe=>[]);for(let Fe=0;Fe<we.length;++Fe){const Ze=Math.floor(Fe/Le.dims[0]*te.length);ke&&(we[Fe]=we[Fe].slice(ke[Ze])),He[Ze].push({generated_text:xe?[...te[Ze],{role:"assistant",content:we[Fe]}]:we[Fe]})}return!me&&He.length===1?He[0]:He}}class F extends b{constructor(te){super(te),this.label2id=Object.fromEntries(Object.entries(this.model.config.label2id).map(([O,me])=>[O.toLowerCase(),me])),this.entailment_id=this.label2id.entailment,this.entailment_id===void 0&&(console.warn("Could not find 'entailment' in label2id mapping. Using 2 as entailment_id."),this.entailment_id=2),this.contradiction_id=this.label2id.contradiction??this.label2id.not_entailment,this.contradiction_id===void 0&&(console.warn("Could not find 'contradiction' in label2id mapping. Using 0 as contradiction_id."),this.contradiction_id=0)}async _call(te,O,{hypothesis_template:me="This example is {}.",multi_label:xe=!1}={}){const be=Array.isArray(te);be||(te=[te]),Array.isArray(O)||(O=[O]);const Ce=O.map(Le=>me.replace("{}",Le)),$e=xe||O.length===1,Me=[];for(const Le of te){const we=[];for(const Fe of Ce){const Ze=this.tokenizer(Le,{text_pair:Fe,padding:!0,truncation:!0}),We=await this.model(Ze);$e?we.push([We.logits.data[this.contradiction_id],We.logits.data[this.entailment_id]]):we.push(We.logits.data[this.entailment_id])}const He=($e?we.map(Fe=>(0,l.softmax)(Fe)[1]):(0,l.softmax)(we)).map((Fe,Ze)=>[Fe,Ze]).sort((Fe,Ze)=>Ze[0]-Fe[0]);Me.push({sequence:Le,labels:He.map(Fe=>O[Fe[1]]),scores:He.map(Fe=>Fe[0])})}return be?Me:Me[0]}}class D extends b{constructor(te){super(te)}async _call(te,{pooling:O="none",normalize:me=!1,quantize:xe=!1,precision:be="binary"}={}){const Ce=this.tokenizer(te,{padding:!0,truncation:!0}),$e=await this.model(Ce);let Me=$e.last_hidden_state??$e.logits??$e.token_embeddings;switch(O){case"none":break;case"mean":Me=(0,p.mean_pooling)(Me,Ce.attention_mask);break;case"first_token":case"cls":Me=Me.slice(null,0);break;case"last_token":case"eos":Me=Me.slice(null,-1);break;default:throw Error(`Pooling method '${O}' not supported.`)}return me&&(Me=Me.normalize(2,-1)),xe&&(Me=(0,p.quantize_embeddings)(Me,be)),Me}}class K extends b{constructor(te){super(te)}async _call(te,{pool:O=null}={}){const me=await u(te),{pixel_values:xe}=await this.processor(me),be=await this.model({pixel_values:xe});let Ce;if(O){if(!("pooler_output"in be))throw Error("No pooled output was returned. Make sure the model has a 'pooler' layer when using the 'pool' option.");Ce=be.pooler_output}else Ce=be.last_hidden_state??be.logits??be.image_embeds;return Ce}}class U extends b{constructor(te){super(te)}async _call(te,{top_k:O=5}={}){const me=this.processor.feature_extractor.config.sampling_rate,xe=await f(te,me),be=this.model.config.id2label,Ce=[];for(const $e of xe){const Me=await this.processor($e),we=(await this.model(Me)).logits[0],ke=await(0,p.topk)(new p.Tensor("float32",(0,l.softmax)(we.data),we.dims),O),He=ke[0].tolist(),Ze=ke[1].tolist().map((We,yt)=>({label:be?be[We]:`LABEL_${We}`,score:He[yt]}));Ce.push(Ze)}return Array.isArray(te)?Ce:Ce[0]}}class j extends b{constructor(te){super(te)}async _call(te,O,{hypothesis_template:me="This is a sound of {}."}={}){const xe=!Array.isArray(te);xe&&(te=[te]);const be=O.map(we=>me.replace("{}",we)),Ce=this.tokenizer(be,{padding:!0,truncation:!0}),$e=this.processor.feature_extractor.config.sampling_rate,Me=await f(te,$e),Le=[];for(const we of Me){const ke=await this.processor(we),He=await this.model({...Ce,...ke}),Fe=(0,l.softmax)(He.logits_per_audio.data);Le.push([...Fe].map((Ze,We)=>({score:Ze,label:O[We]})))}return xe?Le[0]:Le}}class se extends b{constructor(te){super(te)}async _call(te,O={}){switch(this.model.config.model_type){case"whisper":case"lite-whisper":return this._call_whisper(te,O);case"wav2vec2":case"wav2vec2-bert":case"unispeech":case"unispeech-sat":case"hubert":case"parakeet_ctc":return this._call_wav2vec2(te,O);case"moonshine":return this._call_moonshine(te,O);default:throw new Error(`AutomaticSpeechRecognitionPipeline does not support model type '${this.model.config.model_type}'.`)}}async _call_wav2vec2(te,O){O.language&&console.warn('`language` parameter is not yet supported for `wav2vec2` models, defaulting to "English".'),O.task&&console.warn('`task` parameter is not yet supported for `wav2vec2` models, defaulting to "transcribe".');const me=!Array.isArray(te);me&&(te=[te]);const xe=this.processor.feature_extractor.config.sampling_rate,be=await f(te,xe),Ce=[];for(const $e of be){const Me=await this.processor($e),we=(await this.model(Me)).logits[0],ke=[];for(const Fe of we)ke.push((0,l.max)(Fe.data)[1]);const He=this.tokenizer.decode(ke,{skip_special_tokens:!0}).trim();Ce.push({text:He})}return me?Ce[0]:Ce}async _call_whisper(te,O){const me=O.return_timestamps??!1,xe=O.chunk_length_s??0,be=O.force_full_sequences??!1;let Ce=O.stride_length_s??null;const $e={...O};me==="word"&&($e.return_token_timestamps=!0,$e.return_timestamps=!1);const Me=!Array.isArray(te);Me&&(te=[te]);const Le=this.processor.feature_extractor.config.chunk_length/this.model.config.max_source_positions,we=this.processor.feature_extractor.config.hop_length,ke=this.processor.feature_extractor.config.sampling_rate,He=await f(te,ke),Fe=[];for(const Ze of He){let We=[];if(xe>0){if(Ce===null)Ce=xe/6;else if(xe<=Ce)throw Error("`chunk_length_s` must be larger than `stride_length_s`.");const Oe=ke*xe,ze=ke*Ce,pt=Oe-2*ze;let ft=0;for(;;){const Lt=ft+Oe,kt=Ze.subarray(ft,Lt),Mr=await this.processor(kt),fr=ft===0,Ut=Lt>=Ze.length;if(We.push({stride:[kt.length,fr?0:ze,Ut?0:ze],input_features:Mr.input_features,is_last:Ut}),Ut)break;ft+=pt}}else We=[{stride:[Ze.length,0,0],input_features:(await this.processor(Ze)).input_features,is_last:!0}];for(const Oe of We){$e.num_frames=Math.floor(Oe.stride[0]/we);const ze=await this.model.generate({inputs:Oe.input_features,...$e});me==="word"?(Oe.tokens=ze.sequences.tolist()[0],Oe.token_timestamps=ze.token_timestamps.tolist()[0].map(pt=>(0,l.round)(pt,2))):Oe.tokens=ze[0].tolist(),Oe.stride=Oe.stride.map(pt=>pt/ke)}const[yt,Qe]=this.tokenizer._decode_asr(We,{time_precision:Le,return_timestamps:me,force_full_sequences:be});Fe.push({text:yt,...Qe})}return Me?Fe[0]:Fe}async _call_moonshine(te,O){const me=!Array.isArray(te);me&&(te=[te]);const xe=this.processor.feature_extractor.config.sampling_rate,be=await f(te,xe),Ce=[];for(const $e of be){const Me=await this.processor($e),Le=Math.floor($e.length/xe)*6,we=await this.model.generate({max_new_tokens:Le,...O,...Me}),ke=this.processor.batch_decode(we,{skip_special_tokens:!0})[0];Ce.push({text:ke})}return me?Ce[0]:Ce}}class q extends b{constructor(te){super(te)}async _call(te,O={}){const me=Array.isArray(te),xe=await u(te),{pixel_values:be}=await this.processor(xe),Ce=[];for(const $e of be){$e.dims=[1,...$e.dims];const Me=await this.model.generate({inputs:$e,...O}),Le=this.tokenizer.batch_decode(Me,{skip_special_tokens:!0}).map(we=>({generated_text:we.trim()}));Ce.push(Le)}return me?Ce:Ce[0]}}class re extends b{constructor(te){super(te)}async _call(te,{top_k:O=5}={}){const me=await u(te),{pixel_values:xe}=await this.processor(me),be=await this.model({pixel_values:xe}),Ce=this.model.config.id2label,$e=[];for(const Me of be.logits){const Le=await(0,p.topk)(new p.Tensor("float32",(0,l.softmax)(Me.data),Me.dims),O),we=Le[0].tolist(),He=Le[1].tolist().map((Fe,Ze)=>({label:Ce?Ce[Fe]:`LABEL_${Fe}`,score:we[Ze]}));$e.push(He)}return Array.isArray(te)?$e:$e[0]}}class Z extends b{constructor(te){super(te),this.subtasks_mapping={panoptic:"post_process_panoptic_segmentation",instance:"post_process_instance_segmentation",semantic:"post_process_semantic_segmentation"}}async _call(te,{threshold:O=.5,mask_threshold:me=.5,overlap_mask_area_threshold:xe=.8,label_ids_to_fuse:be=null,target_sizes:Ce=null,subtask:$e=null}={}){if(Array.isArray(te)&&te.length!==1)throw Error("Image segmentation pipeline currently only supports a batch size of 1.");const Le=await u(te),we=Le.map(Oe=>[Oe.height,Oe.width]),ke=await this.processor(Le),{inputNames:He,outputNames:Fe}=this.model.sessions.model;if(!He.includes("pixel_values")){if(He.length!==1)throw Error(`Expected a single input name, but got ${He.length} inputs: ${He}.`);const Oe=He[0];if(Oe in ke)throw Error(`Input name ${Oe} already exists in the inputs.`);ke[Oe]=ke.pixel_values}const Ze=await this.model(ke);let We=null;if($e!==null)We=this.subtasks_mapping[$e];else if(this.processor.image_processor){for(const[Oe,ze]of Object.entries(this.subtasks_mapping))if(ze in this.processor.image_processor){We=this.processor.image_processor[ze].bind(this.processor.image_processor),$e=Oe;break}}const yt=this.model.config.id2label,Qe=[];if($e)if($e==="panoptic"||$e==="instance"){const Oe=We(Ze,O,me,xe,be,Ce??we)[0],ze=Oe.segmentation;for(const pt of Oe.segments_info){const ft=new Uint8ClampedArray(ze.data.length);for(let kt=0;kt<ze.data.length;++kt)ze.data[kt]===pt.id&&(ft[kt]=255);const Lt=new d.RawImage(ft,ze.dims[1],ze.dims[0],1);Qe.push({score:pt.score,label:yt[pt.label_id],mask:Lt})}}else if($e==="semantic"){const{segmentation:Oe,labels:ze}=We(Ze,Ce??we)[0];for(const pt of ze){const ft=new Uint8ClampedArray(Oe.data.length);for(let kt=0;kt<Oe.data.length;++kt)Oe.data[kt]===pt&&(ft[kt]=255);const Lt=new d.RawImage(ft,Oe.dims[1],Oe.dims[0],1);Qe.push({score:null,label:yt[pt],mask:Lt})}}else throw Error(`Subtask ${$e} not supported.`);else{const ze=Ze[Fe[0]];for(let pt=0;pt<we.length;++pt){const ft=we[pt],Lt=ze[pt];Lt.data.some(Mr=>Mr<-1e-5||Mr>1+1e-5)&&Lt.sigmoid_();const kt=await d.RawImage.fromTensor(Lt.mul_(255).to("uint8")).resize(ft[1],ft[0]);Qe.push({label:null,score:null,mask:kt})}}return Qe}}class ae extends Z{constructor(te){super(te)}async _call(te,O={}){if(Array.isArray(te)&&te.length!==1)throw Error("Background removal pipeline currently only supports a batch size of 1.");const xe=await u(te),be=await super._call(te,O);return xe.map(($e,Me)=>{const Le=$e.clone();return Le.putAlpha(be[Me].mask),Le})}}class _e extends b{constructor(te){super(te)}async _call(te,O,{hypothesis_template:me="This is a photo of {}"}={}){const xe=Array.isArray(te),be=await u(te),Ce=O.map(He=>me.replace("{}",He)),$e=this.tokenizer(Ce,{padding:this.model.config.model_type==="siglip"?"max_length":!0,truncation:!0}),{pixel_values:Me}=await this.processor(be),Le=await this.model({...$e,pixel_values:Me}),we=this.model.config.model_type==="siglip"?He=>He.sigmoid().data:He=>(0,l.softmax)(He.data),ke=[];for(const He of Le.logits_per_image){const Ze=[...we(He)].map((We,yt)=>({score:We,label:O[yt]}));Ze.sort((We,yt)=>yt.score-We.score),ke.push(Ze)}return xe?ke:ke[0]}}class pe extends b{constructor(te){super(te)}async _call(te,{threshold:O=.9,percentage:me=!1}={}){const xe=Array.isArray(te);if(xe&&te.length!==1)throw Error("Object detection pipeline currently only supports a batch size of 1.");const be=await u(te),Ce=me?null:be.map(Fe=>[Fe.height,Fe.width]),{pixel_values:$e,pixel_mask:Me}=await this.processor(be),Le=await this.model({pixel_values:$e,pixel_mask:Me}),we=this.processor.image_processor.post_process_object_detection(Le,O,Ce),ke=this.model.config.id2label,He=we.map(Fe=>Fe.boxes.map((Ze,We)=>({score:Fe.scores[We],label:ke[Fe.classes[We]],box:_(Ze,!me)})));return xe?He:He[0]}}class oe extends b{constructor(te){super(te)}async _call(te,O,{threshold:me=.1,top_k:xe=null,percentage:be=!1}={}){const Ce=Array.isArray(te),$e=await u(te),Me=this.tokenizer(O,{padding:!0,truncation:!0}),Le=await this.processor($e),we=[];for(let ke=0;ke<$e.length;++ke){const He=$e[ke],Fe=be?null:[[He.height,He.width]],Ze=Le.pixel_values[ke].unsqueeze_(0),We=await this.model({...Me,pixel_values:Ze});let yt;if("post_process_grounded_object_detection"in this.processor){const Qe=this.processor.post_process_grounded_object_detection(We,Me.input_ids,{box_threshold:me,text_threshold:me,target_sizes:Fe})[0];yt=Qe.boxes.map((Oe,ze)=>({score:Qe.scores[ze],label:Qe.labels[ze],box:_(Oe,!be)}))}else{const Qe=this.processor.image_processor.post_process_object_detection(We,me,Fe,!0)[0];yt=Qe.boxes.map((Oe,ze)=>({score:Qe.scores[ze],label:O[Qe.classes[ze]],box:_(Oe,!be)}))}yt.sort((Qe,Oe)=>Oe.score-Qe.score),xe!==null&&(yt=yt.slice(0,xe)),we.push(yt)}return Ce?we:we[0]}}class G extends b{constructor(te){super(te)}async _call(te,O,me={}){const xe=(await u(te))[0],{pixel_values:be}=await this.processor(xe),Ce=`<s_docvqa><s_question>${O}</s_question><s_answer>`,$e=this.tokenizer(Ce,{add_special_tokens:!1,padding:!0,truncation:!0}).input_ids,Me=await this.model.generate({inputs:be,max_length:this.model.config.decoder.max_position_embeddings,decoder_input_ids:$e,...me}),we=this.tokenizer.batch_decode(Me)[0].match(/<s_answer>(.*?)<\/s_answer>/);let ke=null;return we&&we.length>=2&&(ke=we[1].trim()),[{answer:ke}]}}class B extends b{constructor(O){super(O);J(this,"DEFAULT_VOCODER_ID","Xenova/speecht5_hifigan");this.vocoder=O.vocoder??null}async _prepare_speaker_embeddings(O){if((typeof O=="string"||O instanceof URL)&&(O=new Float32Array(await(await fetch(O)).arrayBuffer())),O instanceof Float32Array)O=new p.Tensor("float32",O,[O.length]);else if(!(O instanceof p.Tensor))throw new Error("Speaker embeddings must be a `Tensor`, `Float32Array`, `string`, or `URL`.");return O}async _call(O,{speaker_embeddings:me=null,num_inference_steps:xe,speed:be}={}){return this.processor?this._call_text_to_spectrogram(O,{speaker_embeddings:me}):this.model.config.model_type==="supertonic"?this._call_supertonic(O,{speaker_embeddings:me,num_inference_steps:xe,speed:be}):this._call_text_to_waveform(O)}async _call_supertonic(O,{speaker_embeddings:me,num_inference_steps:xe,speed:be}){if(!me)throw new Error("Speaker embeddings must be provided for Supertonic models.");me=await this._prepare_speaker_embeddings(me);const{sampling_rate:Ce,style_dim:$e}=this.model.config;me=me.view(1,-1,$e);const Me=this.tokenizer(O,{padding:!0,truncation:!0}),{waveform:Le}=await this.model.generate_speech({...Me,style:me,num_inference_steps:xe,speed:be});return new c.RawAudio(Le.data,Ce)}async _call_text_to_waveform(O){const me=this.tokenizer(O,{padding:!0,truncation:!0}),{waveform:xe}=await this.model(me),be=this.model.config.sampling_rate;return new c.RawAudio(xe.data,be)}async _call_text_to_spectrogram(O,{speaker_embeddings:me}){this.vocoder||(console.log("No vocoder specified, using default HifiGan vocoder."),this.vocoder=await o.AutoModel.from_pretrained(this.DEFAULT_VOCODER_ID,{dtype:"fp32"}));const{input_ids:xe}=this.tokenizer(O,{padding:!0,truncation:!0});me=await this._prepare_speaker_embeddings(me),me=me.view(1,-1);const{waveform:be}=await this.model.generate_speech(xe,me,{vocoder:this.vocoder}),Ce=this.processor.feature_extractor.config.sampling_rate;return new c.RawAudio(be.data,Ce)}}class Q extends b{constructor(te){super(te)}async _call(te){const O=await u(te),me=await this.processor(O),xe=await this.model(me),be=[];for(const Ce of xe.reconstruction){const $e=Ce.squeeze().clamp_(0,1).mul_(255).round_().to("uint8");be.push(d.RawImage.fromTensor($e))}return be.length>1?be:be[0]}}class le extends b{constructor(te){super(te)}async _call(te){const O=await u(te),me=await this.processor(O),{predicted_depth:xe}=await this.model(me),be=[];for(let Ce=0;Ce<O.length;++Ce){const $e=xe[Ce],[Me,Le]=$e.dims.slice(-2),[we,ke]=O[Ce].size,He=(await(0,p.interpolate_4d)($e.view(1,1,Me,Le),{size:[ke,we],mode:"bilinear"})).view(ke,we),Fe=He.min().item(),Ze=He.max().item(),We=He.sub(Fe).div_(Ze-Fe).mul_(255).to("uint8").unsqueeze(0),yt=d.RawImage.fromTensor(We);be.push({predicted_depth:He,depth:yt})}return be.length>1?be:be[0]}}const ee=Object.freeze({"text-classification":{tokenizer:s.AutoTokenizer,pipeline:k,model:o.AutoModelForSequenceClassification,default:{model:"Xenova/distilbert-base-uncased-finetuned-sst-2-english"},type:"text"},"token-classification":{tokenizer:s.AutoTokenizer,pipeline:w,model:o.AutoModelForTokenClassification,default:{model:"Xenova/bert-base-multilingual-cased-ner-hrl"},type:"text"},"question-answering":{tokenizer:s.AutoTokenizer,pipeline:y,model:o.AutoModelForQuestionAnswering,default:{model:"Xenova/distilbert-base-cased-distilled-squad"},type:"text"},"fill-mask":{tokenizer:s.AutoTokenizer,pipeline:$,model:o.AutoModelForMaskedLM,default:{model:"Xenova/bert-base-uncased"},type:"text"},summarization:{tokenizer:s.AutoTokenizer,pipeline:M,model:o.AutoModelForSeq2SeqLM,default:{model:"Xenova/distilbart-cnn-6-6"},type:"text"},translation:{tokenizer:s.AutoTokenizer,pipeline:x,model:o.AutoModelForSeq2SeqLM,default:{model:"Xenova/t5-small"},type:"text"},"text2text-generation":{tokenizer:s.AutoTokenizer,pipeline:T,model:o.AutoModelForSeq2SeqLM,default:{model:"Xenova/flan-t5-small"},type:"text"},"text-generation":{tokenizer:s.AutoTokenizer,pipeline:C,model:o.AutoModelForCausalLM,default:{model:"Xenova/gpt2"},type:"text"},"zero-shot-classification":{tokenizer:s.AutoTokenizer,pipeline:F,model:o.AutoModelForSequenceClassification,default:{model:"Xenova/distilbert-base-uncased-mnli"},type:"text"},"audio-classification":{pipeline:U,model:o.AutoModelForAudioClassification,processor:n.AutoProcessor,default:{model:"Xenova/wav2vec2-base-superb-ks"},type:"audio"},"zero-shot-audio-classification":{tokenizer:s.AutoTokenizer,pipeline:j,model:o.AutoModel,processor:n.AutoProcessor,default:{model:"Xenova/clap-htsat-unfused"},type:"multimodal"},"automatic-speech-recognition":{tokenizer:s.AutoTokenizer,pipeline:se,model:[o.AutoModelForSpeechSeq2Seq,o.AutoModelForCTC],processor:n.AutoProcessor,default:{model:"Xenova/whisper-tiny.en"},type:"multimodal"},"text-to-audio":{tokenizer:s.AutoTokenizer,pipeline:B,model:[o.AutoModelForTextToWaveform,o.AutoModelForTextToSpectrogram],processor:[n.AutoProcessor,null],default:{model:"Xenova/speecht5_tts"},type:"text"},"image-to-text":{tokenizer:s.AutoTokenizer,pipeline:q,model:o.AutoModelForVision2Seq,processor:n.AutoProcessor,default:{model:"Xenova/vit-gpt2-image-captioning"},type:"multimodal"},"image-classification":{pipeline:re,model:o.AutoModelForImageClassification,processor:n.AutoProcessor,default:{model:"Xenova/vit-base-patch16-224"},type:"multimodal"},"image-segmentation":{pipeline:Z,model:[o.AutoModelForImageSegmentation,o.AutoModelForSemanticSegmentation,o.AutoModelForUniversalSegmentation],processor:n.AutoProcessor,default:{model:"Xenova/detr-resnet-50-panoptic"},type:"multimodal"},"background-removal":{pipeline:ae,model:[o.AutoModelForImageSegmentation,o.AutoModelForSemanticSegmentation,o.AutoModelForUniversalSegmentation],processor:n.AutoProcessor,default:{model:"Xenova/modnet"},type:"image"},"zero-shot-image-classification":{tokenizer:s.AutoTokenizer,pipeline:_e,model:o.AutoModel,processor:n.AutoProcessor,default:{model:"Xenova/clip-vit-base-patch32"},type:"multimodal"},"object-detection":{pipeline:pe,model:o.AutoModelForObjectDetection,processor:n.AutoProcessor,default:{model:"Xenova/detr-resnet-50"},type:"multimodal"},"zero-shot-object-detection":{tokenizer:s.AutoTokenizer,pipeline:oe,model:o.AutoModelForZeroShotObjectDetection,processor:n.AutoProcessor,default:{model:"Xenova/owlvit-base-patch32"},type:"multimodal"},"document-question-answering":{tokenizer:s.AutoTokenizer,pipeline:G,model:o.AutoModelForDocumentQuestionAnswering,processor:n.AutoProcessor,default:{model:"Xenova/donut-base-finetuned-docvqa"},type:"multimodal"},"image-to-image":{pipeline:Q,model:o.AutoModelForImageToImage,processor:n.AutoProcessor,default:{model:"Xenova/swin2SR-classical-sr-x2-64"},type:"image"},"depth-estimation":{pipeline:le,model:o.AutoModelForDepthEstimation,processor:n.AutoProcessor,default:{model:"Xenova/dpt-large"},type:"image"},"feature-extraction":{tokenizer:s.AutoTokenizer,pipeline:D,model:o.AutoModel,default:{model:"Xenova/all-MiniLM-L6-v2"},type:"text"},"image-feature-extraction":{processor:n.AutoProcessor,pipeline:K,model:[o.AutoModelForImageFeatureExtraction,o.AutoModel],default:{model:"Xenova/vit-base-patch16-224-in21k"},type:"image"}}),Se=Object.freeze({"sentiment-analysis":"text-classification",ner:"token-classification",asr:"automatic-speech-recognition","text-to-speech":"text-to-audio",embeddings:"feature-extraction"});async function De(Ae,te=null,{progress_callback:O=null,config:me=null,cache_dir:xe=null,local_files_only:be=!1,revision:Ce="main",device:$e=null,dtype:Me=null,subfolder:Le="onnx",use_external_data_format:we=null,model_file_name:ke=null,session_options:He={}}={}){Ae=Se[Ae]??Ae;const Fe=ee[Ae.split("_",1)[0]];if(!Fe)throw Error(`Unsupported pipeline: ${Ae}. Must be one of [${Object.keys(ee)}]`);te||(te=Fe.default.model,console.log(`No model specified. Using default model: "${te}".`));const Ze={progress_callback:O,config:me,cache_dir:xe,local_files_only:be,revision:Ce,device:$e,dtype:Me,subfolder:Le,use_external_data_format:we,model_file_name:ke,session_options:He},We=new Map([["tokenizer",Fe.tokenizer],["model",Fe.model],["processor",Fe.processor]]),yt=await Ne(We,te,Ze);yt.task=Ae,(0,i.dispatchCallback)(O,{status:"ready",task:Ae,model:te});const Qe=Fe.pipeline;return new Qe(yt)}async function Ne(Ae,te,O){const me=Object.create(null),xe=[];for(const[be,Ce]of Ae.entries()){if(!Ce)continue;let $e;Array.isArray(Ce)?$e=new Promise(async(Me,Le)=>{var ke,He;let we;for(const Fe of Ce){if(Fe===null){Me(null);return}try{Me(await Fe.from_pretrained(te,O));return}catch(Ze){if((ke=Ze.message)!=null&&ke.includes("Unsupported model type"))we=Ze;else if((He=Ze.message)!=null&&He.includes("Could not locate file"))we=Ze;else{Le(Ze);return}}}Le(we)}):$e=Ce.from_pretrained(te,O),me[be]=$e,xe.push($e)}await Promise.all(xe);for(const[be,Ce]of Object.entries(me))me[be]=await Ce;return me}},"./src/tokenizers.js":(e,r,t)=>{t.r(r),t.d(r,{AlbertTokenizer:()=>As,AutoTokenizer:()=>En,BartTokenizer:()=>rt,BertTokenizer:()=>Is,BlenderbotSmallTokenizer:()=>Ve,BlenderbotTokenizer:()=>je,BloomTokenizer:()=>at,CLIPTokenizer:()=>Ls,CamembertTokenizer:()=>X,CodeGenTokenizer:()=>Ds,CodeLlamaTokenizer:()=>Br,CohereTokenizer:()=>Bs,ConvBertTokenizer:()=>H,DebertaTokenizer:()=>$t,DebertaV2Tokenizer:()=>Fs,DistilBertTokenizer:()=>W,ElectraTokenizer:()=>Ee,EsmTokenizer:()=>gr,FalconTokenizer:()=>Ts,GPT2Tokenizer:()=>ot,GPTNeoXTokenizer:()=>Os,GemmaTokenizer:()=>cs,Grok1Tokenizer:()=>jr,HerbertTokenizer:()=>S,LlamaTokenizer:()=>Wt,M2M100Tokenizer:()=>mr,MBart50Tokenizer:()=>lt,MBartTokenizer:()=>St,MPNetTokenizer:()=>Rr,MarianTokenizer:()=>us,MgpstrTokenizer:()=>Qs,MobileBertTokenizer:()=>Yt,NllbTokenizer:()=>ur,NougatTokenizer:()=>qt,PreTrainedTokenizer:()=>gt,Qwen2Tokenizer:()=>Hs,RoFormerTokenizer:()=>R,RobertaTokenizer:()=>Ot,SiglipTokenizer:()=>zs,SpeechT5Tokenizer:()=>st,SqueezeBertTokenizer:()=>Jr,T5Tokenizer:()=>et,TokenizerModel:()=>K,VitsTokenizer:()=>qs,Wav2Vec2CTCTokenizer:()=>Wr,WhisperTokenizer:()=>Es,XLMRobertaTokenizer:()=>br,XLMTokenizer:()=>ue,is_chinese_char:()=>$});var s=t("./src/utils/generic.js"),o=t("./src/utils/core.js"),n=t("./src/utils/hub.js"),a=t("./src/utils/maths.js"),i=t("./src/utils/tensor.js"),l=t("./src/utils/data-structures.js"),c=t("./node_modules/@huggingface/jinja/dist/index.js"),p=t("./src/models/whisper/common_whisper.js");async function d(he,I){const N=await Promise.all([(0,n.getModelJSON)(he,"tokenizer.json",!0,I),(0,n.getModelJSON)(he,"tokenizer_config.json",!0,I)]);return I.legacy!==null&&(N[1].legacy=I.legacy),N}function u(he,I){const N=[];let Y=0;for(const ie of he.matchAll(I)){const de=ie[0];Y<ie.index&&N.push(he.slice(Y,ie.index)),de.length>0&&N.push(de),Y=ie.index+de.length}return Y<he.length&&N.push(he.slice(Y)),N}function f(he,I=!0){if(he.Regex!==void 0){let N=he.Regex.replace(/\\([#&~])/g,"$1");for(const[Y,ie]of F)N=N.replaceAll(Y,ie);return new RegExp(N,"gu")}else if(he.String!==void 0){const N=(0,o.escapeRegExp)(he.String);return new RegExp(I?N:`(${N})`,"gu")}else return console.warn("Unknown pattern type:",he),null}function _(he){return new Map(Object.entries(he))}function b(he){const I=he.dims;switch(I.length){case 1:return he.tolist();case 2:if(I[0]!==1)throw new Error("Unable to decode tensor with `batch size !== 1`. Use `tokenizer.batch_decode(...)` for batched inputs.");return he.tolist()[0];default:throw new Error(`Expected tensor to have 1-2 dimensions, got ${I.length}.`)}}function k(he){return he.replace(/ \./g,".").replace(/ \?/g,"?").replace(/ \!/g,"!").replace(/ ,/g,",").replace(/ \' /g,"'").replace(/ n\'t/g,"n't").replace(/ \'m/g,"'m").replace(/ \'s/g,"'s").replace(/ \'ve/g,"'ve").replace(/ \'re/g,"'re")}function w(he){return he.replace(new RegExp("\\p{M}","gu"),"")}function y(he){return w(he.toLowerCase())}function $(he){return he>=19968&&he<=40959||he>=13312&&he<=19903||he>=131072&&he<=173791||he>=173824&&he<=177983||he>=177984&&he<=178207||he>=178208&&he<=183983||he>=63744&&he<=64255||he>=194560&&he<=195103}function T(he,I,N){const Y=[];let ie=0;for(;ie<he.length;){if(Y.push(he[ie]),(I.get(he[ie])??N)!==N){++ie;continue}for(;++ie<he.length&&(I.get(he[ie])??N)===N;)I.get(Y.at(-1))!==N&&(Y[Y.length-1]+=he[ie])}return Y}function M(he){return he.match(/\S+/g)||[]}const x="\\p{P}\\u0021-\\u002F\\u003A-\\u0040\\u005B-\\u0060\\u007B-\\u007E",v=new RegExp(`^[${x}]+$`,"gu"),C=".,!?…。，、।۔،",F=new Map([["(?i:'s|'t|'re|'ve|'m|'ll|'d)","(?:'([sS]|[tT]|[rR][eE]|[vV][eE]|[mM]|[lL][lL]|[dD]))"],["(?i:[sdmt]|ll|ve|re)","(?:[sS]|[dD]|[mM]|[tT]|[lL][lL]|[vV][eE]|[rR][eE])"],["[^\\r\\n\\p{L}\\p{N}]?+","[^\\r\\n\\p{L}\\p{N}]?"],["[^\\s\\p{L}\\p{N}]++","[^\\s\\p{L}\\p{N}]+"],[` ?[^(\\s|[${C}])]+`,` ?[^\\s${C}]+`]]);class D{constructor(I){this.content=I.content,this.id=I.id,this.single_word=I.single_word??!1,this.lstrip=I.lstrip??!1,this.rstrip=I.rstrip??!1,this.special=I.special??!1,this.normalized=I.normalized??null}}class K extends s.Callable{constructor(I){super(),this.config=I,this.vocab=[],this.tokens_to_ids=new Map,this.unk_token_id=void 0,this.unk_token=void 0,this.end_of_word_suffix=void 0,this.fuse_unk=this.config.fuse_unk??!1}static fromConfig(I,...N){switch(I.type){case"WordPiece":return new U(I);case"Unigram":return new j(I,...N);case"BPE":return new re(I);default:if(I.vocab)return Array.isArray(I.vocab)?new j(I,...N):Object.hasOwn(I,"continuing_subword_prefix")&&Object.hasOwn(I,"unk_token")?Object.hasOwn(I,"merges")?new re(I):new U(I):new Z(I,...N);throw new Error(`Unknown TokenizerModel type: ${I.type}`)}}_call(I){return I=this.encode(I),this.fuse_unk&&(I=T(I,this.tokens_to_ids,this.unk_token_id)),I}encode(I){throw Error("encode should be implemented in subclass.")}convert_tokens_to_ids(I){return I.map(N=>this.tokens_to_ids.get(N)??this.unk_token_id)}convert_ids_to_tokens(I){return I.map(N=>this.vocab[N]??this.unk_token)}}class U extends K{constructor(I){super(I),this.tokens_to_ids=_(I.vocab),this.unk_token_id=this.tokens_to_ids.get(I.unk_token),this.unk_token=I.unk_token,this.max_input_chars_per_word=I.max_input_chars_per_word??100,this.vocab=new Array(this.tokens_to_ids.size);for(const[N,Y]of this.tokens_to_ids)this.vocab[Y]=N}encode(I){const N=[];for(const Y of I){const ie=[...Y];if(ie.length>this.max_input_chars_per_word){N.push(this.unk_token);continue}let de=!1,Te=0;const Ge=[];for(;Te<ie.length;){let Xe=ie.length,qe=null;for(;Te<Xe;){let tt=ie.slice(Te,Xe).join("");if(Te>0&&(tt=this.config.continuing_subword_prefix+tt),this.tokens_to_ids.has(tt)){qe=tt;break}--Xe}if(qe===null){de=!0;break}Ge.push(qe),Te=Xe}de?N.push(this.unk_token):N.push(...Ge)}return N}}class j extends K{constructor(I,N){super(I);const Y=I.vocab.length;this.vocab=new Array(Y),this.scores=new Array(Y);for(let ie=0;ie<Y;++ie)[this.vocab[ie],this.scores[ie]]=I.vocab[ie];this.unk_token_id=I.unk_id,this.unk_token=this.vocab[I.unk_id],this.tokens_to_ids=new Map(this.vocab.map((ie,de)=>[ie,de])),this.bos_token=" ",this.bos_token_id=this.tokens_to_ids.get(this.bos_token),this.eos_token=N.eos_token,this.eos_token_id=this.tokens_to_ids.get(this.eos_token),this.unk_token=this.vocab[this.unk_token_id],this.minScore=(0,a.min)(this.scores)[0],this.unk_score=this.minScore-10,this.scores[this.unk_token_id]=this.unk_score,this.trie=new l.CharTrie,this.trie.extend(this.vocab),this.fuse_unk=!0}populateNodes(I){const N=I.chars,Y=1;let ie=0;for(;ie<N.length;){let de=!1;const Te=N.slice(ie).join(""),Ge=this.trie.commonPrefixSearch(Te);for(const Xe of Ge){const qe=this.tokens_to_ids.get(Xe),tt=this.scores[qe],bt=(0,o.len)(Xe);I.insert(ie,bt,tt,qe),!de&&bt===Y&&(de=!0)}de||I.insert(ie,Y,this.unk_score,this.unk_token_id),ie+=Y}}tokenize(I){const N=new l.TokenLattice(I,this.bos_token_id,this.eos_token_id);return this.populateNodes(N),N.tokens()}encode(I){const N=[];for(const Y of I){const ie=this.tokenize(Y);N.push(...ie)}return N}}const se=(()=>{const he=[...Array.from({length:94},(ie,de)=>de+33),...Array.from({length:12},(ie,de)=>de+161),...Array.from({length:82},(ie,de)=>de+174)],I=he.slice();let N=0;for(let ie=0;ie<256;++ie)he.includes(ie)||(he.push(ie),I.push(256+N),N+=1);const Y=I.map(ie=>String.fromCharCode(ie));return Object.fromEntries(he.map((ie,de)=>[ie,Y[de]]))})(),q=(0,o.reverseDictionary)(se);class re extends K{constructor(I){super(I),this.tokens_to_ids=_(I.vocab),this.unk_token_id=this.tokens_to_ids.get(I.unk_token),this.unk_token=I.unk_token,this.vocab=new Array(this.tokens_to_ids.size);for(const[Y,ie]of this.tokens_to_ids)this.vocab[ie]=Y;const N=Array.isArray(I.merges[0]);this.merges=N?I.merges:I.merges.map(Y=>Y.split(" ",2)),this.bpe_ranks=new Map(this.merges.map((Y,ie)=>[JSON.stringify(Y),ie])),this.end_of_word_suffix=I.end_of_word_suffix,this.continuing_subword_suffix=I.continuing_subword_suffix??null,this.byte_fallback=this.config.byte_fallback??!1,this.byte_fallback&&(this.text_encoder=new TextEncoder),this.ignore_merges=this.config.ignore_merges??!1,this.max_length_to_cache=256,this.cache_capacity=1e4,this.cache=new l.LRUCache(this.cache_capacity)}clear_cache(){this.cache.clear()}bpe(I){if(I.length===0)return[];const N=this.cache.get(I);if(N!==void 0)return N;const Y=Array.from(I);this.end_of_word_suffix&&(Y[Y.length-1]+=this.end_of_word_suffix);let ie=[];if(Y.length>1){const de=new l.PriorityQueue((Xe,qe)=>Xe.score<qe.score);let Te={token:Y[0],bias:0,prev:null,next:null},Ge=Te;for(let Xe=1;Xe<Y.length;++Xe){const qe={bias:Xe/Y.length,token:Y[Xe],prev:Ge,next:null};Ge.next=qe,this._add_node(de,Ge),Ge=qe}for(;!de.isEmpty();){const Xe=de.pop();if(Xe.deleted||!Xe.next||Xe.next.deleted)continue;if(Xe.deleted=!0,Xe.next.deleted=!0,Xe.prev){const tt={...Xe.prev};Xe.prev.deleted=!0,Xe.prev=tt,tt.prev?tt.prev.next=tt:Te=tt}const qe={token:Xe.token+Xe.next.token,bias:Xe.bias,prev:Xe.prev,next:Xe.next.next};qe.prev?(qe.prev.next=qe,this._add_node(de,qe.prev)):Te=qe,qe.next&&(qe.next.prev=qe,this._add_node(de,qe))}for(let Xe=Te;Xe!==null;Xe=Xe.next)ie.push(Xe.token)}else ie=Y;if(this.continuing_subword_suffix)for(let de=0;de<ie.length-1;++de)ie[de]+=this.continuing_subword_suffix;return I.length<this.max_length_to_cache&&this.cache.put(I,ie),ie}_add_node(I,N){const Y=this.bpe_ranks.get(JSON.stringify([N.token,N.next.token]));Y!==void 0&&(N.score=Y+N.bias,I.push(N))}encode(I){const N=[];for(const Y of I){if(this.ignore_merges&&this.tokens_to_ids.has(Y)){N.push(Y);continue}const ie=this.bpe(Y);for(const de of ie)if(this.tokens_to_ids.has(de))N.push(de);else if(this.byte_fallback){const Te=Array.from(this.text_encoder.encode(de)).map(Ge=>`<0x${Ge.toString(16).toUpperCase().padStart(2,"0")}>`);Te.every(Ge=>this.tokens_to_ids.has(Ge))?N.push(...Te):N.push(this.unk_token)}else N.push(this.unk_token)}return N}}class Z extends K{constructor(I,N){super(I),this.tokens_to_ids=_(N.target_lang?I.vocab[N.target_lang]:I.vocab),this.bos_token=N.bos_token,this.bos_token_id=this.tokens_to_ids.get(this.bos_token),this.eos_token=N.eos_token,this.eos_token_id=this.tokens_to_ids.get(this.eos_token),this.pad_token=N.pad_token,this.pad_token_id=this.tokens_to_ids.get(this.pad_token),this.unk_token=N.unk_token,this.unk_token_id=this.tokens_to_ids.get(this.unk_token),this.vocab=new Array(this.tokens_to_ids.size);for(const[Y,ie]of this.tokens_to_ids)this.vocab[ie]=Y}encode(I){return I}}class ae extends s.Callable{constructor(I){super(),this.config=I}static fromConfig(I){if(I===null)return null;switch(I.type){case"BertNormalizer":return new Ae(I);case"Precompiled":return new Ut(I);case"Sequence":return new Ne(I);case"Replace":return new _e(I);case"NFC":return new oe(I);case"NFD":return new G(I);case"NFKC":return new B(I);case"NFKD":return new Q(I);case"Strip":return new le(I);case"StripAccents":return new ee(I);case"Lowercase":return new Se(I);case"Prepend":return new De(I);default:throw new Error(`Unknown Normalizer type: ${I.type}`)}}normalize(I){throw Error("normalize should be implemented in subclass.")}_call(I){return this.normalize(I)}}class _e extends ae{normalize(I){const N=f(this.config.pattern);return N===null?I:I.replaceAll(N,this.config.content)}}class pe extends ae{constructor(){super(...arguments);J(this,"form")}normalize(N){return N=N.normalize(this.form),N}}class oe extends pe{constructor(){super(...arguments);J(this,"form","NFC")}}class G extends pe{constructor(){super(...arguments);J(this,"form","NFD")}}class B extends pe{constructor(){super(...arguments);J(this,"form","NFKC")}}class Q extends pe{constructor(){super(...arguments);J(this,"form","NFKD")}}class le extends ae{normalize(I){return this.config.strip_left&&this.config.strip_right?I=I.trim():(this.config.strip_left&&(I=I.trimStart()),this.config.strip_right&&(I=I.trimEnd())),I}}class ee extends ae{normalize(I){return I=w(I),I}}class Se extends ae{normalize(I){return I=I.toLowerCase(),I}}class De extends ae{normalize(I){return I=this.config.prepend+I,I}}class Ne extends ae{constructor(I){super(I),this.normalizers=I.normalizers.map(N=>ae.fromConfig(N))}normalize(I){return this.normalizers.reduce((N,Y)=>Y.normalize(N),I)}}class Ae extends ae{_tokenize_chinese_chars(I){const N=[];for(let Y=0;Y<I.length;++Y){const ie=I[Y],de=ie.charCodeAt(0);$(de)?(N.push(" "),N.push(ie),N.push(" ")):N.push(ie)}return N.join("")}stripAccents(I){return I.normalize("NFD").replace(new RegExp("\\p{Mn}","gu"),"")}_is_control(I){switch(I){case"	":case`
`:case"\r":return!1;default:return new RegExp("^\\p{Cc}|\\p{Cf}|\\p{Co}|\\p{Cs}$","u").test(I)}}_clean_text(I){const N=[];for(const Y of I){const ie=Y.charCodeAt(0);ie===0||ie===65533||this._is_control(Y)||(/^\s$/.test(Y)?N.push(" "):N.push(Y))}return N.join("")}normalize(I){return this.config.clean_text&&(I=this._clean_text(I)),this.config.handle_chinese_chars&&(I=this._tokenize_chinese_chars(I)),this.config.lowercase?(I=I.toLowerCase(),this.config.strip_accents!==!1&&(I=this.stripAccents(I))):this.config.strip_accents&&(I=this.stripAccents(I)),I}}class te extends s.Callable{static fromConfig(I){if(I===null)return null;switch(I.type){case"BertPreTokenizer":return new O(I);case"Sequence":return new Ur(I);case"Whitespace":return new Ws(I);case"WhitespaceSplit":return new Gs(I);case"Metaspace":return new Mr(I);case"ByteLevel":return new me(I);case"Split":return new xe(I);case"Punctuation":return new be(I);case"Digits":return new Ce(I);case"Replace":return new nn(I);case"FixedLength":return new cr(I);default:throw new Error(`Unknown PreTokenizer type: ${I.type}`)}}pre_tokenize_text(I,N){throw Error("pre_tokenize_text should be implemented in subclass.")}pre_tokenize(I,N){return(Array.isArray(I)?I.map(Y=>this.pre_tokenize_text(Y,N)):this.pre_tokenize_text(I,N)).flat()}_call(I,N){return this.pre_tokenize(I,N)}}class O extends te{constructor(I){super(),this.pattern=new RegExp(`[^\\s${x}]+|[${x}]`,"gu")}pre_tokenize_text(I,N){return I.trim().match(this.pattern)||[]}}class me extends te{constructor(I){super(),this.config=I,this.add_prefix_space=this.config.add_prefix_space,this.trim_offsets=this.config.trim_offsets,this.use_regex=this.config.use_regex??!0,this.pattern=new RegExp("'s|'t|'re|'ve|'m|'ll|'d| ?\\p{L}+| ?\\p{N}+| ?[^\\s\\p{L}\\p{N}]+|\\s+(?!\\S)|\\s+","gu"),this.byte_encoder=se,this.text_encoder=new TextEncoder}pre_tokenize_text(I,N){return this.add_prefix_space&&!I.startsWith(" ")&&(I=" "+I),(this.use_regex?I.match(this.pattern)||[]:[I]).map(ie=>Array.from(this.text_encoder.encode(ie),de=>this.byte_encoder[de]).join(""))}}class xe extends te{constructor(I){super(),this.config=I,this.pattern=f(this.config.pattern,this.config.invert)}pre_tokenize_text(I,N){var Y;return this.pattern===null?[]:this.config.invert?I.match(this.pattern)||[]:((Y=this.config.behavior)==null?void 0:Y.toLowerCase())==="removed"?I.split(this.pattern).filter(ie=>ie):u(I,this.pattern)}}class be extends te{constructor(I){super(),this.config=I,this.pattern=new RegExp(`[^${x}]+|[${x}]+`,"gu")}pre_tokenize_text(I,N){return I.match(this.pattern)||[]}}class Ce extends te{constructor(I){super(),this.config=I;const N=`[^\\d]+|\\d${this.config.individual_digits?"":"+"}`;this.pattern=new RegExp(N,"gu")}pre_tokenize_text(I,N){return I.match(this.pattern)||[]}}class $e extends s.Callable{constructor(I){super(),this.config=I}static fromConfig(I){if(I===null)return null;switch(I.type){case"TemplateProcessing":return new we(I);case"ByteLevel":return new ke(I);case"RobertaProcessing":return new Le(I);case"BertProcessing":return new Me(I);case"Sequence":return new He(I);default:throw new Error(`Unknown PostProcessor type: ${I.type}`)}}post_process(I,...N){throw Error("post_process should be implemented in subclass.")}_call(I,...N){return this.post_process(I,...N)}}class Me extends $e{constructor(I){super(I),this.cls=I.cls[0],this.sep=I.sep[0]}post_process(I,N=null,{add_special_tokens:Y=!0}={}){Y&&(I=(0,o.mergeArrays)([this.cls],I,[this.sep]));let ie=new Array(I.length).fill(0);if(N!==null){const de=Y&&this instanceof Le?[this.sep]:[],Te=Y?[this.sep]:[];I=(0,o.mergeArrays)(I,de,N,Te),ie=(0,o.mergeArrays)(ie,new Array(N.length+de.length+Te.length).fill(1))}return{tokens:I,token_type_ids:ie}}}class Le extends Me{}class we extends $e{constructor(I){super(I),this.single=I.single,this.pair=I.pair}post_process(I,N=null,{add_special_tokens:Y=!0}={}){const ie=N===null?this.single:this.pair;let de=[],Te=[];for(const Ge of ie)"SpecialToken"in Ge?Y&&(de.push(Ge.SpecialToken.id),Te.push(Ge.SpecialToken.type_id)):"Sequence"in Ge&&(Ge.Sequence.id==="A"?(de=(0,o.mergeArrays)(de,I),Te=(0,o.mergeArrays)(Te,new Array(I.length).fill(Ge.Sequence.type_id))):Ge.Sequence.id==="B"&&(de=(0,o.mergeArrays)(de,N),Te=(0,o.mergeArrays)(Te,new Array(N.length).fill(Ge.Sequence.type_id))));return{tokens:de,token_type_ids:Te}}}class ke extends $e{post_process(I,N=null){return N&&(I=(0,o.mergeArrays)(I,N)),{tokens:I}}}class He extends $e{constructor(I){super(I),this.processors=I.processors.map(N=>$e.fromConfig(N))}post_process(I,N=null,Y={}){let ie;for(const de of this.processors)if(de instanceof ke)I=de.post_process(I).tokens,N&&(N=de.post_process(N).tokens);else{const Te=de.post_process(I,N,Y);I=Te.tokens,ie=Te.token_type_ids}return{tokens:I,token_type_ids:ie}}}class Fe extends s.Callable{constructor(I){super(),this.config=I,this.added_tokens=[],this.end_of_word_suffix=null,this.trim_offsets=I.trim_offsets}static fromConfig(I){if(I===null)return null;switch(I.type){case"WordPiece":return new Oe(I);case"Metaspace":return new fr(I);case"ByteLevel":return new ze(I);case"Replace":return new Ze(I);case"ByteFallback":return new We(I);case"Fuse":return new yt(I);case"Strip":return new Qe(I);case"Sequence":return new ft(I);case"CTC":return new pt(I);case"BPEDecoder":return new Lt(I);default:throw new Error(`Unknown Decoder type: ${I.type}`)}}_call(I){return this.decode(I)}decode(I){return this.decode_chain(I).join("")}decode_chain(I){throw Error("`decode_chain` should be implemented in subclass.")}}class Ze extends Fe{decode_chain(I){const N=f(this.config.pattern);return N===null?I:I.map(Y=>Y.replaceAll(N,this.config.content))}}class We extends Fe{constructor(I){super(I),this.text_decoder=new TextDecoder}decode_chain(I){const N=[];let Y=[];for(const ie of I){let de=null;if(ie.length===6&&ie.startsWith("<0x")&&ie.endsWith(">")){const Te=parseInt(ie.slice(3,5),16);isNaN(Te)||(de=Te)}if(de!==null)Y.push(de);else{if(Y.length>0){const Te=this.text_decoder.decode(Uint8Array.from(Y));N.push(Te),Y=[]}N.push(ie)}}if(Y.length>0){const ie=this.text_decoder.decode(Uint8Array.from(Y));N.push(ie),Y=[]}return N}}class yt extends Fe{decode_chain(I){return[I.join("")]}}class Qe extends Fe{constructor(I){super(I),this.content=this.config.content,this.start=this.config.start,this.stop=this.config.stop}decode_chain(I){return I.map(N=>{let Y=0;for(let de=0;de<this.start&&N[de]===this.content;++de){Y=de+1;continue}let ie=N.length;for(let de=0;de<this.stop;++de){const Te=N.length-de-1;if(N[Te]===this.content){ie=Te;continue}else break}return N.slice(Y,ie)})}}class Oe extends Fe{constructor(I){super(I),this.cleanup=I.cleanup}decode_chain(I){return I.map((N,Y)=>(Y!==0&&(N.startsWith(this.config.prefix)?N=N.replace(this.config.prefix,""):N=" "+N),this.cleanup&&(N=k(N)),N))}}class ze extends Fe{constructor(I){super(I),this.byte_decoder=q,this.text_decoder=new TextDecoder("utf-8",{fatal:!1,ignoreBOM:!0}),this.end_of_word_suffix=null}convert_tokens_to_string(I){const N=I.join(""),Y=new Uint8Array([...N].map(de=>this.byte_decoder[de]));return this.text_decoder.decode(Y)}decode_chain(I){const N=[];let Y=[];for(const ie of I)this.added_tokens.find(de=>de.content===ie)!==void 0?(Y.length>0&&(N.push(this.convert_tokens_to_string(Y)),Y=[]),N.push(ie)):Y.push(ie);return Y.length>0&&N.push(this.convert_tokens_to_string(Y)),N}}class pt extends Fe{constructor(I){super(I),this.pad_token=this.config.pad_token,this.word_delimiter_token=this.config.word_delimiter_token,this.cleanup=this.config.cleanup}convert_tokens_to_string(I){if(I.length===0)return"";const N=[I[0]];for(let de=1;de<I.length;++de)I[de]!==N.at(-1)&&N.push(I[de]);let ie=N.filter(de=>de!==this.pad_token).join("");return this.cleanup&&(ie=k(ie).replaceAll(this.word_delimiter_token," ").trim()),ie}decode_chain(I){return[this.convert_tokens_to_string(I)]}}class ft extends Fe{constructor(I){super(I),this.decoders=I.decoders.map(N=>Fe.fromConfig(N))}decode_chain(I){return this.decoders.reduce((N,Y)=>Y.decode_chain(N),I)}}class Lt extends Fe{constructor(I){super(I),this.suffix=this.config.suffix}decode_chain(I){return I.map((N,Y)=>N.replaceAll(this.suffix,Y===I.length-1?"":" "))}}class kt extends Fe{decode_chain(I){let N="";for(let Y=1;Y<I.length;Y+=2)N+=I[Y];return[N]}}class Mr extends te{constructor(I){super(),this.replacement=I.replacement,this.strRep=I.str_rep||this.replacement,this.prepend_scheme=I.prepend_scheme??"always"}pre_tokenize_text(I,{section_index:N=void 0}={}){let Y=I.replaceAll(" ",this.strRep);return!Y.startsWith(this.replacement)&&(this.prepend_scheme==="always"||this.prepend_scheme==="first"&&N===0)&&(Y=this.strRep+Y),[Y]}}class fr extends Fe{constructor(I){super(I),this.replacement=I.replacement}decode_chain(I){const N=[];for(let Y=0;Y<I.length;++Y){let ie=I[Y].replaceAll(this.replacement," ");Y==0&&ie.startsWith(" ")&&(ie=ie.substring(1)),N.push(ie)}return N}}class Ut extends ae{constructor(I){super(I),this.charsmap=I.precompiled_charsmap}normalize(I){return I=I.replace(/[\u0001-\u0008\u000B\u000E-\u001F\u007F\u008F\u009F]/gm,""),I=I.replace(/[\u0009\u000A\u000C\u000D\u00A0\u1680\u2000-\u200F\u2028\u2029\u202F\u205F\u2581\u3000\uFEFF\uFFFD]/gm," "),I.includes("～")?I=I.split("～").map(Y=>Y.normalize("NFKC")).join("～"):I=I.normalize("NFKC"),I}}class Ur extends te{constructor(I){super(),this.tokenizers=I.pretokenizers.map(N=>te.fromConfig(N))}pre_tokenize_text(I,N){return this.tokenizers.reduce((Y,ie)=>ie.pre_tokenize(Y,N),[I])}}class Ws extends te{constructor(I){super()}pre_tokenize_text(I,N){return I.match(/\w+|[^\w\s]+/g)||[]}}class Gs extends te{constructor(I){super()}pre_tokenize_text(I,N){return M(I)}}class nn extends te{constructor(I){super(),this.config=I,this.pattern=f(this.config.pattern),this.content=this.config.content}pre_tokenize_text(I,N){return this.pattern===null?[I]:[I.replaceAll(this.pattern,this.config.content)]}}class cr extends te{constructor(I){super(),this._length=I.length}pre_tokenize_text(I,N){const Y=[];for(let ie=0;ie<I.length;ie+=this._length)Y.push(I.slice(ie,ie+this._length));return Y}}const zr=["bos_token","eos_token","unk_token","sep_token","pad_token","cls_token","mask_token"];function ls(he,I,N,Y){for(const ie of Object.keys(he)){const de=I-he[ie].length,Te=N(ie),Ge=new Array(de).fill(Te);he[ie]=Y==="right"?(0,o.mergeArrays)(he[ie],Ge):(0,o.mergeArrays)(Ge,he[ie])}}function Ks(he,I){for(const N of Object.keys(he))he[N].length=I}class gt extends s.Callable{constructor(N,Y){super();J(this,"return_token_type_ids",!1);J(this,"padding_side","right");this.config=Y,this.normalizer=ae.fromConfig(N.normalizer),this.pre_tokenizer=te.fromConfig(N.pre_tokenizer),this.model=K.fromConfig(N.model,Y),this.post_processor=$e.fromConfig(N.post_processor),this.decoder=Fe.fromConfig(N.decoder),this.special_tokens=[],this.all_special_ids=[],this.added_tokens=[];for(const ie of N.added_tokens){const de=new D(ie);this.added_tokens.push(de),this.model.tokens_to_ids.set(de.content,de.id),this.model.vocab[de.id]=de.content,de.special&&(this.special_tokens.push(de.content),this.all_special_ids.push(de.id))}if(this.additional_special_tokens=Y.additional_special_tokens??[],this.special_tokens.push(...this.additional_special_tokens),this.special_tokens=[...new Set(this.special_tokens)],this.decoder&&(this.decoder.added_tokens=this.added_tokens,this.decoder.end_of_word_suffix=this.model.end_of_word_suffix),this.added_tokens_splitter=new l.DictionarySplitter(this.added_tokens.map(ie=>ie.content)),this.added_tokens_map=new Map(this.added_tokens.map(ie=>[ie.content,ie])),this.mask_token=this.getToken("mask_token"),this.mask_token_id=this.model.tokens_to_ids.get(this.mask_token),this.pad_token=this.getToken("pad_token","eos_token"),this.pad_token_id=this.model.tokens_to_ids.get(this.pad_token),this.sep_token=this.getToken("sep_token"),this.sep_token_id=this.model.tokens_to_ids.get(this.sep_token),this.unk_token=this.getToken("unk_token"),this.unk_token_id=this.model.tokens_to_ids.get(this.unk_token),this.bos_token=this.getToken("bos_token"),this.bos_token_id=this.model.tokens_to_ids.get(this.bos_token),this.eos_token=this.getToken("eos_token"),this.eos_token_id=this.model.tokens_to_ids.get(this.eos_token),this.model_max_length=Y.model_max_length,this.remove_space=Y.remove_space,this.clean_up_tokenization_spaces=Y.clean_up_tokenization_spaces??!0,this.do_lowercase_and_remove_accent=Y.do_lowercase_and_remove_accent??!1,Y.padding_side&&(this.padding_side=Y.padding_side),this.add_bos_token=Y.add_bos_token,this.add_eos_token=Y.add_eos_token,this.legacy=!1,this.chat_template=Y.chat_template??null,Array.isArray(this.chat_template)){const ie=Object.create(null);for(const{name:de,template:Te}of this.chat_template){if(typeof de!="string"||typeof Te!="string")throw new Error('Chat template must be a list of objects with "name" and "template" properties');ie[de]=Te}this.chat_template=ie}this._compiled_template_cache=new Map}getToken(...N){for(const Y of N){const ie=this.config[Y];if(ie)if(typeof ie=="object"){if(ie.__type==="AddedToken")return ie.content;throw Error(`Unknown token: ${ie}`)}else return ie}return null}static async from_pretrained(N,{progress_callback:Y=null,config:ie=null,cache_dir:de=null,local_files_only:Te=!1,revision:Ge="main",legacy:Xe=null}={}){const qe=await d(N,{progress_callback:Y,config:ie,cache_dir:de,local_files_only:Te,revision:Ge,legacy:Xe});return new this(...qe)}_call(N,{text_pair:Y=null,add_special_tokens:ie=!0,padding:de=!1,truncation:Te=null,max_length:Ge=null,return_tensor:Xe=!0,return_token_type_ids:qe=null}={}){const tt=Array.isArray(N);let bt;if(tt){if(N.length===0)throw Error("text array must be non-empty");if(Y!==null){if(Array.isArray(Y)){if(N.length!==Y.length)throw Error("text and text_pair must have the same length")}else throw Error("text_pair must also be an array");bt=N.map((Bt,nr)=>this._encode_plus(Bt,{text_pair:Y[nr],add_special_tokens:ie,return_token_type_ids:qe}))}else bt=N.map(Bt=>this._encode_plus(Bt,{add_special_tokens:ie,return_token_type_ids:qe}))}else{if(N==null)throw Error("text may not be null or undefined");if(Array.isArray(Y))throw Error("When specifying `text_pair`, since `text` is a string, `text_pair` must also be a string (i.e., not an array).");bt=[this._encode_plus(N,{text_pair:Y,add_special_tokens:ie,return_token_type_ids:qe})]}if(Ge===null?Ge=this.model_max_length:Te===null&&(de===!0?(console.warn("`max_length` is ignored when `padding: true` and there is no truncation strategy. To pad to max length, use `padding: 'max_length'`."),Ge=this.model_max_length):de===!1&&(console.warn("Truncation was not explicitly activated but `max_length` is provided a specific value, please use `truncation: true` to explicitly truncate examples to max length."),Te=!0)),de===!0&&(Ge=Math.min((0,a.max)(bt.map(Bt=>Bt.input_ids.length))[0],Ge??1/0)),Ge=Math.min(Ge,this.model_max_length??1/0),de||Te)for(let Bt=0;Bt<bt.length;++Bt)bt[Bt].input_ids.length!==Ge&&(bt[Bt].input_ids.length>Ge?Te&&Ks(bt[Bt],Ge):de&&ls(bt[Bt],Ge,nr=>nr==="input_ids"?this.pad_token_id:0,this.padding_side));const jt={};if(Xe){if(!(de&&Te)&&bt.some(nr=>{var Jt;for(const yr of Object.keys(nr))if(nr[yr].length!==((Jt=bt[0][yr])==null?void 0:Jt.length))return!0;return!1}))throw Error("Unable to create tensor, you should probably activate truncation and/or padding with 'padding=true' and 'truncation=true' to have batched tensors with the same length.");const Bt=[bt.length,bt[0].input_ids.length];for(const nr of Object.keys(bt[0]))jt[nr]=new i.Tensor("int64",BigInt64Array.from(bt.flatMap(Jt=>Jt[nr]).map(BigInt)),Bt)}else{for(const Bt of Object.keys(bt[0]))jt[Bt]=bt.map(nr=>nr[Bt]);if(!tt)for(const Bt of Object.keys(jt))jt[Bt]=jt[Bt][0]}return jt}_encode_text(N){if(N===null)return null;const Y=this.added_tokens_splitter.split(N);for(let de=0;de<Y.length;++de){const Te=this.added_tokens_map.get(Y[de]);Te&&(Te.lstrip&&de>0&&(Y[de-1]=Y[de-1].trimEnd()),Te.rstrip&&de<Y.length-1&&(Y[de+1]=Y[de+1].trimStart()))}return Y.flatMap((de,Te)=>{if(de.length===0)return[];if(this.added_tokens_map.has(de))return[de];if(this.remove_space===!0&&(de=de.trim().split(/\s+/).join(" ")),this.do_lowercase_and_remove_accent&&(de=y(de)),this.normalizer!==null&&(de=this.normalizer(de)),de.length===0)return[];const Ge=this.pre_tokenizer!==null?this.pre_tokenizer(de,{section_index:Te}):[de];return this.model(Ge)})}_encode_plus(N,{text_pair:Y=null,add_special_tokens:ie=!0,return_token_type_ids:de=null}={}){const{tokens:Te,token_type_ids:Ge}=this._tokenize_helper(N,{pair:Y,add_special_tokens:ie}),Xe=this.model.convert_tokens_to_ids(Te),qe={input_ids:Xe,attention_mask:new Array(Xe.length).fill(1)};return(de??this.return_token_type_ids)&&Ge&&(qe.token_type_ids=Ge),qe}_tokenize_helper(N,{pair:Y=null,add_special_tokens:ie=!1}={}){const de=this._encode_text(N),Te=this._encode_text(Y);return this.post_processor?this.post_processor(de,Te,{add_special_tokens:ie}):{tokens:(0,o.mergeArrays)(de??[],Te??[])}}tokenize(N,{pair:Y=null,add_special_tokens:ie=!1}={}){return this._tokenize_helper(N,{pair:Y,add_special_tokens:ie}).tokens}encode(N,{text_pair:Y=null,add_special_tokens:ie=!0,return_token_type_ids:de=null}={}){return this._encode_plus(N,{text_pair:Y,add_special_tokens:ie,return_token_type_ids:de}).input_ids}batch_decode(N,Y={}){return N instanceof i.Tensor&&(N=N.tolist()),N.map(ie=>this.decode(ie,Y))}decode(N,Y={}){if(N instanceof i.Tensor&&(N=b(N)),!Array.isArray(N)||N.length===0||!(0,o.isIntegralNumber)(N[0]))throw Error("token_ids must be a non-empty array of integers.");return this.decode_single(N,Y)}decode_single(N,{skip_special_tokens:Y=!1,clean_up_tokenization_spaces:ie=null}){let de=this.model.convert_ids_to_tokens(N);Y&&(de=de.filter(Ge=>!this.special_tokens.includes(Ge)));let Te=this.decoder?this.decoder(de):de.join(" ");return this.decoder&&this.decoder.end_of_word_suffix&&(Te=Te.replaceAll(this.decoder.end_of_word_suffix," "),Y&&(Te=Te.trim())),(ie??this.clean_up_tokenization_spaces)&&(Te=k(Te)),Te}get_chat_template({chat_template:N=null,tools:Y=null}={}){if(this.chat_template&&typeof this.chat_template=="object"){const ie=this.chat_template;if(N!==null&&Object.hasOwn(ie,N))N=ie[N];else if(N===null)if(Y!==null&&"tool_use"in ie)N=ie.tool_use;else if("default"in ie)N=ie.default;else throw Error(`This model has multiple chat templates with no default specified! Please either pass a chat template or the name of the template you wish to use to the 'chat_template' argument. Available template names are ${Object.keys(ie).sort()}.`)}else if(N===null)if(this.chat_template)N=this.chat_template;else throw Error("Cannot use apply_chat_template() because tokenizer.chat_template is not set and no template argument was passed! For information about writing templates and setting the tokenizer.chat_template attribute, please see the documentation at https://huggingface.co/docs/transformers/main/en/chat_templating");return N}apply_chat_template(N,{tools:Y=null,documents:ie=null,chat_template:de=null,add_generation_prompt:Te=!1,tokenize:Ge=!0,padding:Xe=!1,truncation:qe=!1,max_length:tt=null,return_tensor:bt=!0,return_dict:jt=!1,tokenizer_kwargs:Bt={},...nr}={}){if(de=this.get_chat_template({chat_template:de,tools:Y}),typeof de!="string")throw Error(`chat_template must be a string, but got ${typeof de}`);let Jt=this._compiled_template_cache.get(de);Jt===void 0&&(Jt=new c.Template(de),this._compiled_template_cache.set(de,Jt));const yr=Object.create(null);for(const hr of zr){const xr=this.getToken(hr);xr&&(yr[hr]=xr)}const vr=Jt.render({messages:N,add_generation_prompt:Te,tools:Y,documents:ie,...yr,...nr});if(Ge){const hr=this._call(vr,{add_special_tokens:!1,padding:Xe,truncation:qe,max_length:tt,return_tensor:bt,...Bt});return jt?hr:hr.input_ids}return vr}}class Is extends gt{constructor(){super(...arguments);J(this,"return_token_type_ids",!0)}}class As extends gt{constructor(){super(...arguments);J(this,"return_token_type_ids",!0)}}class Yt extends gt{constructor(){super(...arguments);J(this,"return_token_type_ids",!0)}}class Jr extends gt{constructor(){super(...arguments);J(this,"return_token_type_ids",!0)}}class $t extends gt{constructor(){super(...arguments);J(this,"return_token_type_ids",!0)}}class Fs extends gt{constructor(){super(...arguments);J(this,"return_token_type_ids",!0)}}class S extends gt{constructor(){super(...arguments);J(this,"return_token_type_ids",!0)}}class H extends gt{constructor(){super(...arguments);J(this,"return_token_type_ids",!0)}}class R extends gt{constructor(){super(...arguments);J(this,"return_token_type_ids",!0)}}class W extends gt{}class X extends gt{}class ue extends gt{constructor(N,Y){super(N,Y);J(this,"return_token_type_ids",!0);console.warn('WARNING: `XLMTokenizer` is not yet supported by Hugging Face\'s "fast" tokenizers library. Therefore, you may experience slightly inaccurate results.')}}class Ee extends gt{constructor(){super(...arguments);J(this,"return_token_type_ids",!0)}}class et extends gt{}class ot extends gt{}class rt extends gt{}class St extends gt{constructor(I,N){super(I,N),this.languageRegex=/^[a-z]{2}_[A-Z]{2}$/,this.language_codes=this.special_tokens.filter(Y=>this.languageRegex.test(Y)),this.lang_to_token=Y=>Y}_build_translation_inputs(I,N,Y){return Hr(this,I,N,Y)}}class lt extends St{}class Ot extends gt{}class at extends gt{}const Dt="▁";class Wt extends gt{constructor(N,Y){super(N,Y);J(this,"padding_side","left");this.legacy=Y.legacy??!0,this.legacy||(this.normalizer=null,this.pre_tokenizer=new Mr({replacement:Dt,prepend_scheme:"first"}))}_encode_text(N){if(N===null)return null;if(this.legacy||N.length===0)return super._encode_text(N);let Y=super._encode_text(Dt+N.replaceAll(Dt," "));return Y.length>1&&Y[0]===Dt&&this.special_tokens.includes(Y[1])&&(Y=Y.slice(1)),Y}}class Br extends gt{}class br extends gt{}class Rr extends gt{}class Ts extends gt{}class Os extends gt{}class gr extends gt{}class Hs extends gt{}class cs extends gt{}class jr extends gt{}function Hr(he,I,N,Y){if(!("language_codes"in he)||!Array.isArray(he.language_codes))throw new Error("Tokenizer must have `language_codes` attribute set and it should be an array of language ids.");if(!("languageRegex"in he)||!(he.languageRegex instanceof RegExp))throw new Error("Tokenizer must have `languageRegex` attribute set and it should be a regular expression.");if(!("lang_to_token"in he)||typeof he.lang_to_token!="function")throw new Error("Tokenizer must have `lang_to_token` attribute set and it should be a function.");const ie=Y.src_lang,de=Y.tgt_lang;if(!he.language_codes.includes(de))throw new Error(`Target language code "${de}" is not valid. Must be one of: {${he.language_codes.join(", ")}}`);if(ie!==void 0){if(!he.language_codes.includes(ie))throw new Error(`Source language code "${ie}" is not valid. Must be one of: {${he.language_codes.join(", ")}}`);for(const Te of he.post_processor.config.single)if("SpecialToken"in Te&&he.languageRegex.test(Te.SpecialToken.id)){Te.SpecialToken.id=he.lang_to_token(ie);break}}return Y.forced_bos_token_id=he.model.convert_tokens_to_ids([he.lang_to_token(de)])[0],he._call(I,N)}class ur extends gt{constructor(I,N){super(I,N),this.languageRegex=/^[a-z]{3}_[A-Z][a-z]{3}$/,this.language_codes=this.special_tokens.filter(Y=>this.languageRegex.test(Y)),this.lang_to_token=Y=>Y}_build_translation_inputs(I,N,Y){return Hr(this,I,N,Y)}}class mr extends gt{constructor(I,N){super(I,N),this.languageRegex=/^__[a-z]{2,3}__$/,this.language_codes=this.special_tokens.filter(Y=>this.languageRegex.test(Y)).map(Y=>Y.slice(2,-2)),this.lang_to_token=Y=>`__${Y}__`}_build_translation_inputs(I,N,Y){return Hr(this,I,N,Y)}}class Es extends gt{get timestamp_begin(){return this.model.convert_tokens_to_ids(["<|notimestamps|>"])[0]+1}_decode_asr(I,{return_timestamps:N=!1,return_language:Y=!1,time_precision:ie=null,force_full_sequences:de=!0}={}){if(ie===null)throw Error("Must specify time_precision");let Te=null;const Ge=N==="word";function Xe(){return{language:Te,timestamp:[null,null],text:""}}const qe=[];let tt=Xe(),bt=0;const jt=this.timestamp_begin,nr=jt+1500;let Jt=[],yr=[],vr=!1,hr=null;const xr=new Set(this.all_special_ids);for(const rr of I){const Tr=rr.tokens,Ar=Ge?rr.token_timestamps:null;let Yr=null,Ps=jt;if("stride"in rr){const[dr,_r,wr]=rr.stride;if(bt-=_r,hr=dr-wr,_r&&(Ps=_r/ie+jt),wr)for(let or=Tr.length-1;or>=0;--or){const qr=Number(Tr[or]);if(qr>=jt){if(Yr!==null&&(qr-jt)*ie<hr)break;Yr=qr}}}let Fr=[],ds=[];for(let dr=0;dr<Tr.length;++dr){const _r=Number(Tr[dr]);if(xr.has(_r)){const wr=this.decode([_r]),or=p.WHISPER_LANGUAGE_MAPPING.get(wr.slice(2,-2));if(or!==void 0){if(Te!==null&&or!==Te&&!N){Jt.push(Fr);const qr=this.findLongestCommonSequence(Jt)[0],Xs=this.decode(qr);tt.text=Xs,qe.push(tt),Jt=[],Fr=[],tt=Xe()}Te=tt.language=or}}else if(_r>=jt&&_r<=nr){const wr=(_r-jt)*ie+bt,or=(0,a.round)(wr,2);if(Yr!==null&&_r>=Yr)vr=!0;else if(vr||Jt.length>0&&_r<Ps)vr=!1;else if(tt.timestamp[0]===null)tt.timestamp[0]=or;else if(or!==tt.timestamp[0]){tt.timestamp[1]=or,Jt.push(Fr),Ge&&yr.push(ds);const[qr,Xs]=this.findLongestCommonSequence(Jt,yr),Rs=this.decode(qr);tt.text=Rs,Ge&&(tt.words=this.collateWordTimestamps(qr,Xs,Te)),qe.push(tt),Jt=[],Fr=[],yr=[],ds=[],tt=Xe()}}else if(Fr.push(_r),Ge){let wr=(0,a.round)(Ar[dr]+bt,2),or;if(dr+1<Ar.length){or=(0,a.round)(Ar[dr+1]+bt,2);const qr=this.decode([_r]);v.test(qr)&&(or=(0,a.round)(Math.min(wr+ie,or),2))}else or=null;ds.push([wr,or])}}if("stride"in rr){const[dr,_r,wr]=rr.stride;bt+=dr-wr}Fr.length>0?(Jt.push(Fr),Ge&&yr.push(ds)):Jt.every(dr=>dr.length===0)&&(tt=Xe(),Jt=[],Fr=[],yr=[],ds=[])}if(Jt.length>0){if(de&&N)throw new Error("Whisper did not predict an ending timestamp, which can happen if audio is cut off in the middle of a word. Also make sure WhisperTimeStampLogitsProcessor was used during generation.");const[rr,Tr]=this.findLongestCommonSequence(Jt,yr),Ar=this.decode(rr);tt.text=Ar,Ge&&(tt.words=this.collateWordTimestamps(rr,Tr,Te)),qe.push(tt)}let Cr=Object.create(null);const ss=qe.map(rr=>rr.text).join("");if(N||Y){for(let rr=0;rr<qe.length;++rr){const Tr=qe[rr];N||delete Tr.timestamp,Y||delete Tr.language}if(Ge){const rr=[];for(const Tr of qe)for(const Ar of Tr.words)rr.push(Ar);Cr={chunks:rr}}else Cr={chunks:qe}}return[ss,Cr]}findLongestCommonSequence(I,N=null){let Y=I[0],ie=Y.length,de=[];const Te=Array.isArray(N)&&N.length>0;let Ge=Te?[]:null,Xe=Te?N[0]:null;for(let qe=1;qe<I.length;++qe){const tt=I[qe];let bt=0,jt=[ie,ie,0,0];const Bt=tt.length;for(let Cr=1;Cr<ie+Bt;++Cr){const ss=Math.max(0,ie-Cr),rr=Math.min(ie,ie+Bt-Cr),Tr=Y.slice(ss,rr),Ar=Math.max(0,Cr-ie),Yr=Math.min(Bt,Cr),Ps=tt.slice(Ar,Yr);if(Tr.length!==Ps.length)throw new Error("There is a bug within whisper `decode_asr` function, please report it. Dropping to prevent bad inference.");let Fr;Te?Fr=Tr.filter((_r,wr)=>_r===Ps[wr]&&Xe[ss+wr]<=N[qe][Ar+wr]).length:Fr=Tr.filter((_r,wr)=>_r===Ps[wr]).length;const ds=Cr/1e4,dr=Fr/Cr+ds;Fr>1&&dr>bt&&(bt=dr,jt=[ss,rr,Ar,Yr])}const[nr,Jt,yr,vr]=jt,hr=Math.floor((Jt+nr)/2),xr=Math.floor((vr+yr)/2);de.push(...Y.slice(0,hr)),Y=tt.slice(xr),ie=Y.length,Te&&(Ge.push(...Xe.slice(0,hr)),Xe=N[qe].slice(xr))}return de.push(...Y),Te?(Ge.push(...Xe),[de,Ge]):[de,[]]}collateWordTimestamps(I,N,Y){const[ie,de,Te]=this.combineTokensIntoWords(I,Y),Ge=[];for(let Xe=0;Xe<ie.length;++Xe){const qe=Te[Xe];Ge.push({text:ie[Xe],timestamp:[N[qe.at(0)][0],N[qe.at(-1)][1]]})}return Ge}combineTokensIntoWords(I,N,Y=`"'“¡¿([{-`,ie=`"'.。,，!！?？:：”)]}、`){N=N??"english";let de,Te,Ge;return["chinese","japanese","thai","lao","myanmar"].includes(N)?[de,Te,Ge]=this.splitTokensOnUnicode(I):[de,Te,Ge]=this.splitTokensOnSpaces(I),this.mergePunctuations(de,Te,Ge,Y,ie)}decode(I,N){let Y;return N!=null&&N.decode_with_timestamps?(I instanceof i.Tensor&&(I=b(I)),Y=this.decodeWithTimestamps(I,N)):Y=super.decode(I,N),Y}decodeWithTimestamps(I,N){const Y=(N==null?void 0:N.time_precision)??.02,ie=Array.from(this.all_special_ids).at(-1)+1;let de=[[]];for(let Te of I)if(Te=Number(Te),Te>=ie){const Ge=((Te-ie)*Y).toFixed(2);de.push(`<|${Ge}|>`),de.push([])}else de[de.length-1].push(Te);return de=de.map(Te=>typeof Te=="string"?Te:super.decode(Te,N)),de.join("")}splitTokensOnUnicode(I){const N=this.decode(I,{decode_with_timestamps:!0}),Y="�",ie=[],de=[],Te=[];let Ge=[],Xe=[],qe=0;for(let tt=0;tt<I.length;++tt){const bt=I[tt];Ge.push(bt),Xe.push(tt);const jt=this.decode(Ge,{decode_with_timestamps:!0});(!jt.includes(Y)||N[qe+jt.indexOf(Y)]===Y)&&(ie.push(jt),de.push(Ge),Te.push(Xe),Ge=[],Xe=[],qe+=jt.length)}return[ie,de,Te]}splitTokensOnSpaces(I){const[N,Y,ie]=this.splitTokensOnUnicode(I),de=[],Te=[],Ge=[],Xe=new RegExp(`^[${x}]$`,"gu");for(let qe=0;qe<N.length;++qe){const tt=N[qe],bt=Y[qe],jt=ie[qe],Bt=bt[0]>=this.model.tokens_to_ids.get("<|endoftext|>"),nr=tt.startsWith(" "),Jt=tt.trim(),yr=Xe.test(Jt);if(Bt||nr||yr||de.length===0)de.push(tt),Te.push(bt),Ge.push(jt);else{const vr=de.length-1;de[vr]+=tt,Te[vr].push(...bt),Ge[vr].push(...jt)}}return[de,Te,Ge]}mergePunctuations(I,N,Y,ie,de){const Te=structuredClone(I),Ge=structuredClone(N),Xe=structuredClone(Y);let qe=Te.length-2,tt=Te.length-1;for(;qe>=0;)Te[qe].startsWith(" ")&&ie.includes(Te[qe].trim())?(Te[tt]=Te[qe]+Te[tt],Ge[tt]=(0,o.mergeArrays)(Ge[qe],Ge[tt]),Xe[tt]=(0,o.mergeArrays)(Xe[qe],Xe[tt]),Te[qe]="",Ge[qe]=[],Xe[qe]=[]):tt=qe,--qe;for(qe=0,tt=1;tt<Te.length;)!Te[qe].endsWith(" ")&&de.includes(Te[tt])?(Te[qe]+=Te[tt],Ge[qe]=(0,o.mergeArrays)(Ge[qe],Ge[tt]),Xe[qe]=(0,o.mergeArrays)(Xe[qe],Xe[tt]),Te[tt]="",Ge[tt]=[],Xe[tt]=[]):qe=tt,++tt;return[Te.filter(bt=>bt),Ge.filter(bt=>bt.length>0),Xe.filter(bt=>bt.length>0)]}}class Ds extends gt{}class Ls extends gt{}class zs extends gt{}class us extends gt{constructor(I,N){super(I,N),this.languageRegex=/^(>>\w+<<)\s*/g,this.supported_language_codes=this.model.vocab.filter(Y=>this.languageRegex.test(Y)),console.warn('WARNING: `MarianTokenizer` is not yet supported by Hugging Face\'s "fast" tokenizers library. Therefore, you may experience slightly inaccurate results.')}_encode_text(I){if(I===null)return null;const[N,...Y]=I.trim().split(this.languageRegex);if(Y.length===0)return super._encode_text(N);if(Y.length===2){const[ie,de]=Y;return this.supported_language_codes.includes(ie)||console.warn(`Unsupported language code "${ie}" detected, which may lead to unexpected behavior. Should be one of: ${JSON.stringify(this.supported_language_codes)}`),(0,o.mergeArrays)([ie],super._encode_text(de))}}}class Wr extends gt{}class je extends gt{}class Ve extends gt{}class st extends gt{}class qt extends gt{}class qs extends gt{constructor(I,N){super(I,N),this.decoder=new kt({})}}class Bs extends gt{}class Qs extends gt{}class En{static async from_pretrained(I,{progress_callback:N=null,config:Y=null,cache_dir:ie=null,local_files_only:de=!1,revision:Te="main",legacy:Ge=null}={}){var jt;const[Xe,qe]=await d(I,{progress_callback:N,config:Y,cache_dir:ie,local_files_only:de,revision:Te,legacy:Ge}),tt=((jt=qe.tokenizer_class)==null?void 0:jt.replace(/Fast$/,""))??"PreTrainedTokenizer";let bt=this.TOKENIZER_CLASS_MAPPING[tt];return bt||(console.warn(`Unknown tokenizer class "${tt}", attempting to construct from base class.`),bt=gt),new bt(Xe,qe)}}J(En,"TOKENIZER_CLASS_MAPPING",{T5Tokenizer:et,DistilBertTokenizer:W,CamembertTokenizer:X,DebertaTokenizer:$t,DebertaV2Tokenizer:Fs,BertTokenizer:Is,HerbertTokenizer:S,ConvBertTokenizer:H,RoFormerTokenizer:R,XLMTokenizer:ue,ElectraTokenizer:Ee,MobileBertTokenizer:Yt,SqueezeBertTokenizer:Jr,AlbertTokenizer:As,GPT2Tokenizer:ot,BartTokenizer:rt,MBartTokenizer:St,MBart50Tokenizer:lt,RobertaTokenizer:Ot,WhisperTokenizer:Es,CodeGenTokenizer:Ds,CLIPTokenizer:Ls,SiglipTokenizer:zs,MarianTokenizer:us,BloomTokenizer:at,NllbTokenizer:ur,M2M100Tokenizer:mr,LlamaTokenizer:Wt,CodeLlamaTokenizer:Br,XLMRobertaTokenizer:br,MPNetTokenizer:Rr,FalconTokenizer:Ts,GPTNeoXTokenizer:Os,EsmTokenizer:gr,Wav2Vec2CTCTokenizer:Wr,BlenderbotTokenizer:je,BlenderbotSmallTokenizer:Ve,SpeechT5Tokenizer:st,NougatTokenizer:qt,VitsTokenizer:qs,Qwen2Tokenizer:Hs,GemmaTokenizer:cs,Grok1Tokenizer:jr,CohereTokenizer:Bs,MgpstrTokenizer:Qs,PreTrainedTokenizer:gt})},"./src/utils/audio.js":(e,r,t)=>{t.r(r),t.d(r,{RawAudio:()=>U,hamming:()=>u,hanning:()=>d,mel_filter_bank:()=>$,read_audio:()=>c,spectrogram:()=>C,window_function:()=>F});var s=t("./src/utils/hub.js"),o=t("./src/utils/maths.js"),n=t("./src/utils/core.js"),a=t("./src/env.js"),i=t("./src/utils/tensor.js"),l=t("?7992");async function c(j,se){if(typeof AudioContext>"u")throw Error("Unable to load audio from path/URL since `AudioContext` is not available in your environment. Instead, audio data should be passed directly to the pipeline/processor. For more information and some example code, see https://huggingface.co/docs/transformers.js/guides/node-audio-processing.");const q=await(await(0,s.getFile)(j)).arrayBuffer(),re=new AudioContext({sampleRate:se});typeof se>"u"&&console.warn(`No sampling rate provided, using default of ${re.sampleRate}Hz.`);const Z=await re.decodeAudioData(q);let ae;if(Z.numberOfChannels===2){const _e=Math.sqrt(2),pe=Z.getChannelData(0),oe=Z.getChannelData(1);ae=new Float32Array(pe.length);for(let G=0;G<Z.length;++G)ae[G]=_e*(pe[G]+oe[G])/2}else ae=Z.getChannelData(0);return ae}function p(j,se){if(j<1)return new Float64Array;if(j===1)return new Float64Array([1]);const q=1-se,re=2*Math.PI/(j-1),Z=new Float64Array(j);for(let ae=0;ae<j;++ae)Z[ae]=se-q*Math.cos(ae*re);return Z}function d(j){return p(j,.5)}function u(j){return p(j,.54)}const f={htk:j=>2595*Math.log10(1+j/700),kaldi:j=>1127*Math.log(1+j/700),slaney:(j,se=1e3,q=15,re=27/Math.log(6.4))=>j>=se?q+Math.log(j/se)*re:3*j/200};function _(j,se="htk"){const q=f[se];if(!q)throw new Error('mel_scale should be one of "htk", "slaney" or "kaldi".');return typeof j=="number"?q(j):j.map(re=>q(re))}const b={htk:j=>700*(10**(j/2595)-1),kaldi:j=>700*(Math.exp(j/1127)-1),slaney:(j,se=1e3,q=15,re=Math.log(6.4)/27)=>j>=q?se*Math.exp(re*(j-q)):200*j/3};function k(j,se="htk"){const q=b[se];if(!q)throw new Error('mel_scale should be one of "htk", "slaney" or "kaldi".');return typeof j=="number"?q(j):j.map(re=>q(re))}function w(j,se){const q=Float64Array.from({length:se.length-1},(_e,pe)=>se[pe+1]-se[pe]),re=Array.from({length:j.length},()=>new Array(se.length));for(let _e=0;_e<j.length;++_e){const pe=re[_e];for(let oe=0;oe<se.length;++oe)pe[oe]=se[oe]-j[_e]}const Z=se.length-2,ae=Array.from({length:Z},()=>new Array(j.length));for(let _e=0;_e<j.length;++_e){const pe=re[_e];for(let oe=0;oe<Z;++oe){const G=-pe[oe]/q[oe],B=pe[oe+2]/q[oe+1];ae[oe][_e]=Math.max(0,Math.min(G,B))}}return ae}function y(j,se,q){const re=(se-j)/(q-1);return Float64Array.from({length:q},(Z,ae)=>j+re*ae)}function $(j,se,q,re,Z,ae=null,_e="htk",pe=!1){if(ae!==null&&ae!=="slaney")throw new Error('norm must be one of null or "slaney"');if(j<2)throw new Error(`Require num_frequency_bins: ${j} >= 2`);if(q>re)throw new Error(`Require min_frequency: ${q} <= max_frequency: ${re}`);const oe=_(q,_e),G=_(re,_e),B=y(oe,G,se+2);let Q=k(B,_e),le;if(pe){const Se=Z/((j-1)*2);le=_(Float64Array.from({length:j},(De,Ne)=>Ne*Se),_e),Q=B}else le=y(0,Math.floor(Z/2),j);const ee=w(le,Q);if(ae!==null&&ae==="slaney")for(let Se=0;Se<se;++Se){const De=ee[Se],Ne=2/(Q[Se+2]-Q[Se]);for(let Ae=0;Ae<j;++Ae)De[Ae]*=Ne}return ee}function T(j,se,q){const re=new j.constructor(j.length+se+q),Z=j.length-1;for(let ae=0;ae<j.length;++ae)re[se+ae]=j[ae];for(let ae=1;ae<=se;++ae)re[se-ae]=j[(0,n.calculateReflectOffset)(ae,Z)];for(let ae=1;ae<=q;++ae)re[Z+se+ae]=j[(0,n.calculateReflectOffset)(Z-ae,Z)];return re}function M(j,se,q,re,Z){if(q<=0)throw new Error("reference must be greater than zero");if(re<=0)throw new Error("min_value must be greater than zero");q=Math.max(re,q);const ae=Math.log10(q);for(let _e=0;_e<j.length;++_e)j[_e]=se*Math.log10(Math.max(re,j[_e])-ae);if(Z!==null){if(Z<=0)throw new Error("db_range must be greater than zero");const _e=(0,o.max)(j)[0]-Z;for(let pe=0;pe<j.length;++pe)j[pe]=Math.max(j[pe],_e)}return j}function x(j,se=1,q=1e-5,re=null){return M(j,20,se,q,re)}function v(j,se=1,q=1e-10,re=null){return M(j,10,se,q,re)}async function C(j,se,q,re,{fft_length:Z=null,power:ae=1,center:_e=!0,pad_mode:pe="reflect",onesided:oe=!0,preemphasis:G=null,preemphasis_htk_flavor:B=!0,mel_filters:Q=null,mel_floor:le=1e-10,log_mel:ee=null,reference:Se=1,min_value:De=1e-10,db_range:Ne=null,remove_dc_offset:Ae=null,min_num_frames:te=null,max_num_frames:O=null,do_pad:me=!0,transpose:xe=!1,mel_offset:be=0}={}){const Ce=se.length;if(Z===null&&(Z=q),q>Z)throw Error(`frame_length (${q}) may not be larger than fft_length (${Z})`);if(Ce!==q)throw new Error(`Length of the window (${Ce}) must equal frame_length (${q})`);if(re<=0)throw new Error("hop_length must be greater than zero");if(ae===null&&Q!==null)throw new Error("You have provided `mel_filters` but `power` is `None`. Mel spectrogram computation is not yet supported for complex-valued spectrogram. Specify `power` to fix this issue.");if(!B)throw new Error("`preemphasis_htk_flavor=false` is not currently supported.");if(_e)switch(pe){case"reflect":{const Oe=Math.floor((Z-1)/2)+1;j=T(j,Oe,Oe);break}case"constant":{const Oe=Math.floor(Z/2),ze=new j.constructor(j.length+2*Oe);ze.set(j,Oe),j=ze;break}default:throw new Error(`pad_mode="${pe}" not implemented yet.`)}let $e=Math.floor(1+Math.floor((j.length-q)/re));te!==null&&$e<te&&($e=te);const Me=oe?Math.floor(Z/2)+1:Z;let Le=$e,we=$e;O!==null&&(O>$e?me&&(we=O):we=Le=O);const ke=new o.FFT(Z),He=new Float64Array(Z),Fe=new Float64Array(ke.outputBufferSize),Ze=new Float32Array(Me*we);for(let Oe=0;Oe<Le;++Oe){const ze=Oe*re,pt=Math.min(j.length-ze,q);pt!==q&&He.fill(0,0,q);for(let ft=0;ft<pt;++ft)He[ft]=j[ze+ft];if(Ae){let ft=0;for(let kt=0;kt<pt;++kt)ft+=He[kt];const Lt=ft/pt;for(let kt=0;kt<pt;++kt)He[kt]-=Lt}if(G!==null){for(let ft=pt-1;ft>=1;--ft)He[ft]-=G*He[ft-1];He[0]*=1-G}for(let ft=0;ft<se.length;++ft)He[ft]*=se[ft];ke.realTransform(Fe,He);for(let ft=0;ft<Me;++ft){const Lt=ft<<1;Ze[ft*we+Oe]=Fe[Lt]**2+Fe[Lt+1]**2}}if(ae!==null&&ae!==2){const Oe=ae/2;for(let ze=0;ze<Ze.length;++ze)Ze[ze]**=Oe}const We=Q.length;let yt=await(0,i.matmul)(new i.Tensor("float32",Q.flat(),[We,Me]),new i.Tensor("float32",Ze,[Me,we]));xe&&(yt=yt.transpose(1,0));const Qe=yt.data;for(let Oe=0;Oe<Qe.length;++Oe)Qe[Oe]=be+Math.max(le,Qe[Oe]);if(ae!==null&&ee!==null){const Oe=Math.min(Qe.length,Le*We);switch(ee){case"log":for(let ze=0;ze<Oe;++ze)Qe[ze]=Math.log(Qe[ze]);break;case"log10":for(let ze=0;ze<Oe;++ze)Qe[ze]=Math.log10(Qe[ze]);break;case"dB":if(ae===1)x(Qe,Se,De,Ne);else if(ae===2)v(Qe,Se,De,Ne);else throw new Error(`Cannot use log_mel option '${ee}' with power ${ae}`);break;default:throw new Error(`log_mel must be one of null, 'log', 'log10' or 'dB'. Got '${ee}'`)}}return yt}function F(j,se,{periodic:q=!0,frame_length:re=null,center:Z=!0}={}){const ae=q?j+1:j;let _e;switch(se){case"boxcar":_e=new Float64Array(ae).fill(1);break;case"hann":case"hann_window":_e=d(ae);break;case"hamming":_e=u(ae);break;case"povey":_e=d(ae).map(pe=>Math.pow(pe,.85));break;default:throw new Error(`Unknown window type ${se}.`)}if(q&&(_e=_e.subarray(0,j)),re===null)return _e;if(j>re)throw new Error(`Length of the window (${j}) may not be larger than frame_length (${re})`);return _e}function D(j,se){let q=44;const re=new ArrayBuffer(q+j.length*4),Z=new DataView(re);K(Z,0,"RIFF"),Z.setUint32(4,36+j.length*4,!0),K(Z,8,"WAVE"),K(Z,12,"fmt "),Z.setUint32(16,16,!0),Z.setUint16(20,3,!0),Z.setUint16(22,1,!0),Z.setUint32(24,se,!0),Z.setUint32(28,se*4,!0),Z.setUint16(32,4,!0),Z.setUint16(34,32,!0),K(Z,36,"data"),Z.setUint32(40,j.length*4,!0);for(let ae=0;ae<j.length;++ae,q+=4)Z.setFloat32(q,j[ae],!0);return re}function K(j,se,q){for(let re=0;re<q.length;++re)j.setUint8(se+re,q.charCodeAt(re))}class U{constructor(se,q){this.audio=se,this.sampling_rate=q}toWav(){return D(this.audio,this.sampling_rate)}toBlob(){const se=this.toWav();return new Blob([se],{type:"audio/wav"})}async save(se){let q;if(a.apis.IS_BROWSER_ENV){if(a.apis.IS_WEBWORKER_ENV)throw new Error("Unable to save a file from a Web Worker.");q=n.saveBlob}else if(a.apis.IS_FS_AVAILABLE)q=async(re,Z)=>{let ae=await Z.arrayBuffer();l.writeFileSync(re,Buffer.from(ae))};else throw new Error("Unable to save because filesystem is disabled in this environment.");await q(se,this.toBlob())}}},"./src/utils/constants.js":(e,r,t)=>{t.r(r),t.d(r,{CHAT_TEMPLATE_NAME:()=>l,CONFIG_NAME:()=>o,FEATURE_EXTRACTOR_NAME:()=>n,GENERATION_CONFIG_NAME:()=>c,GITHUB_ISSUE_URL:()=>s,IMAGE_PROCESSOR_NAME:()=>a,PROCESSOR_NAME:()=>i});const s="https://github.com/huggingface/transformers.js/issues/new/choose",o="config.json",n="preprocessor_config.json",a=n,i="processor_config.json",l="chat_template.jinja",c="generation_config.json"},"./src/utils/core.js":(e,r,t)=>{t.r(r),t.d(r,{calculateDimensions:()=>c,calculateReflectOffset:()=>f,count:()=>w,dispatchCallback:()=>s,escapeRegExp:()=>n,isIntegralNumber:()=>i,isNullishDimension:()=>l,isTypedArray:()=>a,len:()=>k,mergeArrays:()=>d,pick:()=>b,pop:()=>p,product:()=>u,reverseDictionary:()=>o,saveBlob:()=>_});function s(y,$){y&&y($)}function o(y){return Object.fromEntries(Object.entries(y).map(([$,T])=>[T,$]))}function n(y){return y.replace(/[.*+?^${}()|[\]\\]/g,"\\$&")}function a(y){var $,T,M;return((M=(T=($=y==null?void 0:y.prototype)==null?void 0:$.__proto__)==null?void 0:T.constructor)==null?void 0:M.name)==="TypedArray"}function i(y){return Number.isInteger(y)||typeof y=="bigint"}function l(y){return y==null||y===-1}function c(y){const $=[];let T=y;for(;Array.isArray(T);)$.push(T.length),T=T[0];return $}function p(y,$,T=void 0){const M=y[$];if(M!==void 0)return delete y[$],M;if(T===void 0)throw Error(`Key ${$} does not exist in object.`);return T}function d(...y){return Array.prototype.concat.apply([],y)}function u(...y){return y.reduce(($,T)=>$.flatMap(M=>T.map(x=>[M,x])))}function f(y,$){return Math.abs((y+$)%(2*$)-$)}function _(y,$){const T=URL.createObjectURL($),M=document.createElement("a");M.href=T,M.download=y,M.click(),M.remove(),URL.revokeObjectURL(T)}function b(y,$){return Object.assign({},...$.map(T=>{if(y[T]!==void 0)return{[T]:y[T]}}))}function k(y){let $=0;for(const T of y)++$;return $}function w(y,$){let T=0;for(const M of y)M===$&&++T;return T}},"./src/utils/data-structures.js":(e,r,t)=>{t.r(r),t.d(r,{CharTrie:()=>o,DictionarySplitter:()=>l,LRUCache:()=>c,PriorityQueue:()=>s,TokenLattice:()=>a});class s{constructor(d=(f,_)=>f>_,u=1/0){this._heap=[],this._comparator=d,this._maxSize=u}get size(){return this._heap.length}isEmpty(){return this.size===0}peek(){return this._heap[0]}push(...d){return this.extend(d)}extend(d){for(const u of d)if(this.size<this._maxSize)this._heap.push(u),this._siftUp();else{const f=this._smallest();this._comparator(u,this._heap[f])&&(this._heap[f]=u,this._siftUpFrom(f))}return this.size}pop(){const d=this.peek(),u=this.size-1;return u>0&&this._swap(0,u),this._heap.pop(),this._siftDown(),d}replace(d){const u=this.peek();return this._heap[0]=d,this._siftDown(),u}_parent(d){return(d+1>>>1)-1}_left(d){return(d<<1)+1}_right(d){return d+1<<1}_greater(d,u){return this._comparator(this._heap[d],this._heap[u])}_swap(d,u){const f=this._heap[d];this._heap[d]=this._heap[u],this._heap[u]=f}_siftUp(){this._siftUpFrom(this.size-1)}_siftUpFrom(d){for(;d>0&&this._greater(d,this._parent(d));)this._swap(d,this._parent(d)),d=this._parent(d)}_siftDown(){let d=0;for(;this._left(d)<this.size&&this._greater(this._left(d),d)||this._right(d)<this.size&&this._greater(this._right(d),d);){const u=this._right(d)<this.size&&this._greater(this._right(d),this._left(d))?this._right(d):this._left(d);this._swap(d,u),d=u}}_smallest(){return 2**Math.floor(Math.log2(this.size))-1}}class o{constructor(){this.root=n.default()}extend(d){for(const u of d)this.push(u)}push(d){let u=this.root;for(const f of d){let _=u.children.get(f);_===void 0&&(_=n.default(),u.children.set(f,_)),u=_}u.isLeaf=!0}*commonPrefixSearch(d){let u=this.root;if(u===void 0)return;let f="";for(const _ of d){if(f+=_,u=u.children.get(_),u===void 0)return;u.isLeaf&&(yield f)}}}class n{constructor(d,u){this.isLeaf=d,this.children=u}static default(){return new n(!1,new Map)}}class a{constructor(d,u,f){this.chars=Array.from(d),this.len=this.chars.length,this.bosTokenId=u,this.eosTokenId=f,this.nodes=[],this.beginNodes=Array.from({length:this.len+1},()=>[]),this.endNodes=Array.from({length:this.len+1},()=>[]);const _=new i(this.bosTokenId,0,0,0,0),b=new i(this.eosTokenId,1,this.len,0,0);this.nodes.push(_.clone()),this.nodes.push(b.clone()),this.beginNodes[this.len].push(b),this.endNodes[0].push(_)}insert(d,u,f,_){const b=this.nodes.length,k=new i(_,b,d,u,f);this.beginNodes[d].push(k),this.endNodes[d+u].push(k),this.nodes.push(k)}viterbi(){const d=this.len;let u=0;for(;u<=d;){if(this.beginNodes[u].length==0)return[];for(let w of this.beginNodes[u]){w.prev=null;let y=0,$=null;for(let T of this.endNodes[u]){const M=T.backtraceScore+w.score;($===null||M>y)&&($=T.clone(),y=M)}if($!==null)w.prev=$,w.backtraceScore=y;else return[]}++u}const f=[],b=this.beginNodes[d][0].prev;if(b===null)return[];let k=b.clone();for(;k.prev!==null;)f.push(k.clone()),k=k.clone().prev.clone();return f.reverse(),f}piece(d){return this.chars.slice(d.pos,d.pos+d.length).join("")}tokens(){return this.viterbi().map(u=>this.piece(u))}tokenIds(){return this.viterbi().map(u=>u.tokenId)}}class i{constructor(d,u,f,_,b){this.tokenId=d,this.nodeId=u,this.pos=f,this.length=_,this.score=b,this.prev=null,this.backtraceScore=0}clone(){const d=new i(this.tokenId,this.nodeId,this.pos,this.length,this.score);return d.prev=this.prev,d.backtraceScore=this.backtraceScore,d}}class l{constructor(d){this.trie=this._buildTrie(d)}_buildTrie(d){var f;const u=Object.create(null);for(const _ of d){let b=u;for(let k=0;k<_.length;++k)b=b[f=_[k]]??(b[f]=Object.create(null));b.end=_}return u}split(d){const u=[],f=d.length;let _=0,b=0;for(;b<f;){let k=this.trie,w=null,y=b;for(;y<f&&(k=k[d[y]]);)k.end&&(w=k.end),++y;w?(b>_&&u.push(d.slice(_,b)),u.push(w),b+=w.length,_=b):++b}return _<f&&u.push(d.slice(_)),u}}class c{constructor(d){this.capacity=d,this.cache=new Map}get(d){if(!this.cache.has(d))return;const u=this.cache.get(d);return this.cache.delete(d),this.cache.set(d,u),u}put(d,u){this.cache.has(d)&&this.cache.delete(d),this.cache.set(d,u),this.cache.size>this.capacity&&this.cache.delete(this.cache.keys().next().value)}clear(){this.cache.clear()}}},"./src/utils/devices.js":(e,r,t)=>{t.r(r),t.d(r,{DEVICE_TYPES:()=>s});const s=Object.freeze({auto:"auto",gpu:"gpu",cpu:"cpu",wasm:"wasm",webgpu:"webgpu",cuda:"cuda",dml:"dml",webnn:"webnn","webnn-npu":"webnn-npu","webnn-gpu":"webnn-gpu","webnn-cpu":"webnn-cpu"})},"./src/utils/dtypes.js":(e,r,t)=>{t.r(r),t.d(r,{DATA_TYPES:()=>a,DEFAULT_DEVICE_DTYPE_MAPPING:()=>i,DEFAULT_DTYPE_SUFFIX_MAPPING:()=>l,isWebGpuFp16Supported:()=>n});var s=t("./src/env.js"),o=t("./src/utils/devices.js");const n=function(){let c;return async function(){if(c===void 0)if(!s.apis.IS_WEBGPU_AVAILABLE)c=!1;else try{c=(await navigator.gpu.requestAdapter()).features.has("shader-f16")}catch{c=!1}return c}}(),a=Object.freeze({auto:"auto",fp32:"fp32",fp16:"fp16",q8:"q8",int8:"int8",uint8:"uint8",q4:"q4",bnb4:"bnb4",q4f16:"q4f16"}),i=Object.freeze({[o.DEVICE_TYPES.wasm]:a.q8}),l=Object.freeze({[a.fp32]:"",[a.fp16]:"_fp16",[a.int8]:"_int8",[a.uint8]:"_uint8",[a.q8]:"_quantized",[a.q4]:"_q4",[a.q4f16]:"_q4f16",[a.bnb4]:"_bnb4"})},"./src/utils/generic.js":(e,r,t)=>{t.r(r),t.d(r,{Callable:()=>s});const s=class{constructor(){let o=function(...n){return o._call(...n)};return Object.setPrototypeOf(o,new.target.prototype)}_call(...o){throw Error("Must implement _call method in subclass")}}},"./src/utils/hub.js":(e,r,t)=>{t.r(r),t.d(r,{MAX_EXTERNAL_DATA_CHUNKS:()=>i,getFile:()=>f,getModelFile:()=>y,getModelJSON:()=>T,getModelText:()=>$});var s=t("?7992"),o=t("?5af5"),n=t("./src/env.js"),a=t("./src/utils/core.js");const i=100,l={txt:"text/plain",html:"text/html",css:"text/css",js:"text/javascript",json:"application/json",png:"image/png",jpg:"image/jpeg",jpeg:"image/jpeg",gif:"image/gif"};class c{constructor(C){if(this.filePath=C,this.headers=new Headers,this.exists=s.existsSync(C),this.exists){this.status=200,this.statusText="OK";let F=s.statSync(C);this.headers.set("content-length",F.size.toString()),this.updateContentType();const D=s.createReadStream(C);this.body=new ReadableStream({start(K){D.on("data",U=>K.enqueue(U)),D.on("end",()=>K.close()),D.on("error",U=>K.error(U))},cancel(){D.destroy()}})}else this.status=404,this.statusText="Not Found",this.body=null}updateContentType(){const C=this.filePath.toString().split(".").pop().toLowerCase();this.headers.set("content-type",l[C]??"application/octet-stream")}clone(){let C=new c(this.filePath);return C.exists=this.exists,C.status=this.status,C.statusText=this.statusText,C.headers=new Headers(this.headers),C}async arrayBuffer(){return(await s.promises.readFile(this.filePath)).buffer}async blob(){const C=await s.promises.readFile(this.filePath);return new Blob([C],{type:this.headers.get("content-type")})}async text(){return await s.promises.readFile(this.filePath,"utf8")}async json(){return JSON.parse(await this.text())}}function p(v,C=null,F=null){let D;try{D=new URL(v)}catch{return!1}return!(C&&!C.includes(D.protocol)||F&&!F.includes(D.hostname))}const d=/^(\b[\w\-.]+\b\/)?\b[\w\-.]{1,96}\b$/;function u(v){return!(!d.test(v)||v.includes("..")||v.includes("--")||v.endsWith(".git")||v.endsWith(".ipynb"))}async function f(v){var C;if(n.env.useFS&&!p(v,["http:","https:","blob:"]))return new c(v instanceof URL?v.protocol==="file:"?v.pathname:v.toString():v);if(typeof process<"u"&&((C=process==null?void 0:process.release)==null?void 0:C.name)==="node"){const F=!!(Us!=null&&Us.TESTING_REMOTELY),D=n.env.version,K=new Headers;if(K.set("User-Agent",`transformers.js/${D}; is_ci/${F};`),p(v,["http:","https:"],["huggingface.co","hf.co"])){const j=(Us==null?void 0:Us.HF_TOKEN)??(Us==null?void 0:Us.HF_ACCESS_TOKEN);j&&K.set("Authorization",`Bearer ${j}`)}return fetch(v,{headers:K})}else return fetch(v)}const _={400:"Bad request error occurred while trying to load file",401:"Unauthorized access to file",403:"Forbidden access to file",404:"Could not locate file",408:"Request timeout error occurred while trying to load file",500:"Internal server error error occurred while trying to load file",502:"Bad gateway error occurred while trying to load file",503:"Service unavailable error occurred while trying to load file",504:"Gateway timeout error occurred while trying to load file"};function b(v,C,F){if(!F)return null;const D=_[v]??`Error (${v}) occurred while trying to load file`;throw Error(`${D}: "${C}".`)}class k{constructor(C){this.path=C}async match(C){let F=o.join(this.path,C),D=new c(F);if(D.exists)return D}async put(C,F,D=void 0){let K=o.join(this.path,C);try{const U=F.headers.get("Content-Length"),j=parseInt(U??"0");let se=0;await s.promises.mkdir(o.dirname(K),{recursive:!0});const q=s.createWriteStream(K),re=F.body.getReader();for(;;){const{done:Z,value:ae}=await re.read();if(Z)break;await new Promise((pe,oe)=>{q.write(ae,G=>{if(G){oe(G);return}pe()})}),se+=ae.length;const _e=j?se/j*100:0;D==null||D({progress:_e,loaded:se,total:j})}q.close()}catch(U){try{await s.promises.unlink(K)}catch{}throw U}}}async function w(v,...C){for(let F of C)try{let D=await v.match(F);if(D)return D}catch{continue}}async function y(v,C,F=!0,D={},K=!1){if(!n.env.allowLocalModels){if(D.local_files_only)throw Error("Invalid configuration detected: local models are disabled (`env.allowLocalModels=false`) but you have requested to only use local models (`local_files_only=true`).");if(!n.env.allowRemoteModels)throw Error("Invalid configuration detected: both local and remote models are disabled. Fix by setting `env.allowLocalModels` or `env.allowRemoteModels` to `true`.")}(0,a.dispatchCallback)(D.progress_callback,{status:"initiate",name:v,file:C});let U;if(!U&&n.env.useCustomCache){if(!n.env.customCache)throw Error("`env.useCustomCache=true`, but `env.customCache` is not defined.");if(!n.env.customCache.match||!n.env.customCache.put)throw new Error("`env.customCache` must be an object which implements the `match` and `put` functions of the Web Cache API. For more information, see https://developer.mozilla.org/en-US/docs/Web/API/Cache");U=n.env.customCache}if(!U&&n.env.useBrowserCache){if(typeof caches>"u")throw Error("Browser cache is not available in this environment.");try{U=await caches.open("transformers-cache")}catch(le){console.warn("An error occurred while opening the browser cache:",le)}}if(!U&&n.env.useFSCache){if(!n.apis.IS_FS_AVAILABLE)throw Error("File System Cache is not available in this environment.");U=new k(D.cache_dir??n.env.cacheDir)}const j=D.revision??"main",se=x(v,C),q=u(v),re=q?x(n.env.localModelPath,se):se,Z=x(n.env.remoteHost,n.env.remotePathTemplate.replaceAll("{model}",v).replaceAll("{revision}",encodeURIComponent(j)),C);let ae;const _e=U instanceof k?j==="main"?se:x(v,j,C):Z;let pe=!1,oe;U&&(oe=await w(U,re,_e));const G=oe!==void 0;if(oe===void 0){if(n.env.allowLocalModels)if(p(se,["http:","https:"])){if(D.local_files_only)throw new Error(`\`local_files_only=true\`, but attempted to load a remote file from: ${se}.`);if(!n.env.allowRemoteModels)throw new Error(`\`env.allowRemoteModels=false\`, but attempted to load a remote file from: ${se}.`)}else try{oe=await f(re),ae=re}catch(ee){console.warn(`Unable to load from local path "${re}": "${ee}"`)}if(oe===void 0||oe.status===404){if(D.local_files_only||!n.env.allowRemoteModels){if(F)throw Error(`\`local_files_only=true\` or \`env.allowRemoteModels=false\` and file was not found locally at "${re}".`);return null}if(!q)throw Error(`Local file missing at "${re}" and download aborted due to invalid model ID "${v}".`);if(oe=await f(Z),oe.status!==200)return b(oe.status,Z,F);ae=_e}pe=U&&typeof Response<"u"&&oe instanceof Response&&oe.status===200}(0,a.dispatchCallback)(D.progress_callback,{status:"download",name:v,file:C});let B;if(!(n.apis.IS_NODE_ENV&&K)){let le;D.progress_callback?G&&typeof navigator<"u"&&/firefox/i.test(navigator.userAgent)?(le=new Uint8Array(await oe.arrayBuffer()),(0,a.dispatchCallback)(D.progress_callback,{status:"progress",name:v,file:C,progress:100,loaded:le.length,total:le.length})):le=await M(oe,ee=>{(0,a.dispatchCallback)(D.progress_callback,{status:"progress",name:v,file:C,...ee})}):le=new Uint8Array(await oe.arrayBuffer()),B=le}if(pe&&ae&&await U.match(ae)===void 0)if(B)await U.put(ae,new Response(B,{headers:oe.headers})).catch(le=>{console.warn(`Unable to add response to browser cache: ${le}.`)});else{const le=D.progress_callback?ee=>(0,a.dispatchCallback)(D.progress_callback,{status:"progress",name:v,file:C,...ee}):void 0;await U.put(ae,oe,le)}if((0,a.dispatchCallback)(D.progress_callback,{status:"done",name:v,file:C}),B){if(!n.apis.IS_NODE_ENV&&K)throw new Error("Cannot return path in a browser environment.");return B}if(oe instanceof c)return oe.filePath;const Q=await(U==null?void 0:U.match(ae));if(Q instanceof c)return Q.filePath;if(Q instanceof Response)return new Uint8Array(await Q.arrayBuffer());if(typeof Q=="string")return Q;throw new Error("Unable to get model file path or buffer.")}async function $(v,C,F=!0,D={}){const K=await y(v,C,F,D,!1);return K===null?null:new TextDecoder("utf-8").decode(K)}async function T(v,C,F=!0,D={}){const K=await $(v,C,F,D);return K===null?{}:JSON.parse(K)}async function M(v,C){const F=v.headers.get("Content-Length");F===null&&console.warn("Unable to determine content-length from response headers. Will expand buffer when needed.");let D=parseInt(F??"0"),K=new Uint8Array(D),U=0;const j=v.body.getReader();async function se(){const{done:q,value:re}=await j.read();if(q)return;const Z=U+re.length;if(Z>D){D=Z;const _e=new Uint8Array(D);_e.set(K),K=_e}K.set(re,U),U=Z;const ae=U/D*100;return C({progress:ae,loaded:U,total:D}),se()}return await se(),K}function x(...v){return v=v.map((C,F)=>(F&&(C=C.replace(new RegExp("^/"),"")),F!==v.length-1&&(C=C.replace(new RegExp("/$"),"")),C)),v.join("/")}},"./src/utils/image.js":(e,r,t)=>{t.r(r),t.d(r,{RawImage:()=>_,load_image:()=>b});var s=t("./src/utils/core.js"),o=t("./src/utils/hub.js"),n=t("./src/env.js"),a=t("./src/utils/tensor.js"),i=t("?2b25");let l,c,p;const d=n.apis.IS_BROWSER_ENV||n.apis.IS_WEBWORKER_ENV;if(d)l=(k,w)=>{if(!self.OffscreenCanvas)throw new Error("OffscreenCanvas not supported by this browser.");return new self.OffscreenCanvas(k,w)},p=self.createImageBitmap,c=self.ImageData;else if(i)p=async k=>{const y=(await k.metadata()).channels,{data:$,info:T}=await k.rotate().raw().toBuffer({resolveWithObject:!0}),M=new _(new Uint8ClampedArray($),T.width,T.height,T.channels);return y!==void 0&&y!==T.channels&&M.convert(y),M};else throw new Error("Unable to load image processing library.");const u={0:"nearest",1:"lanczos",2:"bilinear",3:"bicubic",4:"box",5:"hamming"},f=new Map([["png","image/png"],["jpg","image/jpeg"],["jpeg","image/jpeg"],["gif","image/gif"]]);class _{constructor(w,y,$,T){this.data=w,this.width=y,this.height=$,this.channels=T}get size(){return[this.width,this.height]}static async read(w){if(w instanceof _)return w;if(typeof w=="string"||w instanceof URL)return await this.fromURL(w);if(w instanceof Blob)return await this.fromBlob(w);if(typeof HTMLCanvasElement<"u"&&w instanceof HTMLCanvasElement||typeof OffscreenCanvas<"u"&&w instanceof OffscreenCanvas)return this.fromCanvas(w);throw new Error(`Unsupported input type: ${typeof w}`)}static fromCanvas(w){if(!d)throw new Error("fromCanvas() is only supported in browser environments.");const $=w.getContext("2d").getImageData(0,0,w.width,w.height).data;return new _($,w.width,w.height,4)}static async fromURL(w){const y=await(0,o.getFile)(w);if(y.status!==200)throw new Error(`Unable to read image from "${w}" (${y.status} ${y.statusText})`);const $=await y.blob();return this.fromBlob($)}static async fromBlob(w){if(d){const y=await p(w),$=l(y.width,y.height).getContext("2d");return $.drawImage(y,0,0),new this($.getImageData(0,0,y.width,y.height).data,y.width,y.height,4)}else{const y=i(await w.arrayBuffer());return await p(y)}}static fromTensor(w,y="CHW"){if(w.dims.length!==3)throw new Error(`Tensor should have 3 dimensions, but has ${w.dims.length} dimensions.`);if(y==="CHW")w=w.transpose(1,2,0);else if(y!=="HWC")throw new Error(`Unsupported channel format: ${y}`);if(!(w.data instanceof Uint8ClampedArray||w.data instanceof Uint8Array))throw new Error(`Unsupported tensor type: ${w.type}`);switch(w.dims[2]){case 1:case 2:case 3:case 4:return new _(w.data,w.dims[1],w.dims[0],w.dims[2]);default:throw new Error(`Unsupported number of channels: ${w.dims[2]}`)}}grayscale(){if(this.channels===1)return this;const w=new Uint8ClampedArray(this.width*this.height*1);switch(this.channels){case 3:case 4:for(let y=0,$=0;y<this.data.length;y+=this.channels){const T=this.data[y],M=this.data[y+1],x=this.data[y+2];w[$++]=Math.round(.2989*T+.587*M+.114*x)}break;default:throw new Error(`Conversion failed due to unsupported number of channels: ${this.channels}`)}return this._update(w,this.width,this.height,1)}rgb(){if(this.channels===3)return this;const w=new Uint8ClampedArray(this.width*this.height*3);switch(this.channels){case 1:for(let y=0,$=0;y<this.data.length;++y)w[$++]=this.data[y],w[$++]=this.data[y],w[$++]=this.data[y];break;case 4:for(let y=0,$=0;y<this.data.length;y+=4)w[$++]=this.data[y],w[$++]=this.data[y+1],w[$++]=this.data[y+2];break;default:throw new Error(`Conversion failed due to unsupported number of channels: ${this.channels}`)}return this._update(w,this.width,this.height,3)}rgba(){if(this.channels===4)return this;const w=new Uint8ClampedArray(this.width*this.height*4);switch(this.channels){case 1:for(let y=0,$=0;y<this.data.length;++y)w[$++]=this.data[y],w[$++]=this.data[y],w[$++]=this.data[y],w[$++]=255;break;case 3:for(let y=0,$=0;y<this.data.length;y+=3)w[$++]=this.data[y],w[$++]=this.data[y+1],w[$++]=this.data[y+2],w[$++]=255;break;default:throw new Error(`Conversion failed due to unsupported number of channels: ${this.channels}`)}return this._update(w,this.width,this.height,4)}putAlpha(w){if(w.width!==this.width||w.height!==this.height)throw new Error(`Expected mask size to be ${this.width}x${this.height}, but got ${w.width}x${w.height}`);if(w.channels!==1)throw new Error(`Expected mask to have 1 channel, but got ${w.channels}`);const y=this.data,$=w.data,T=this.width*this.height;if(this.channels===3){const M=new Uint8ClampedArray(T*4);for(let x=0,v=0,C=0;x<T;++x)M[C++]=y[v++],M[C++]=y[v++],M[C++]=y[v++],M[C++]=$[x];return this._update(M,this.width,this.height,4)}else if(this.channels===4){for(let M=0;M<T;++M)y[4*M+3]=$[M];return this}throw new Error(`Expected image to have 3 or 4 channels, but got ${this.channels}`)}async resize(w,y,{resample:$=2}={}){if(this.width===w&&this.height===y)return this;let T=u[$]??$;const M=(0,s.isNullishDimension)(w),x=(0,s.isNullishDimension)(y);if(M&&x)return this;if(M?w=y/this.height*this.width:x&&(y=w/this.width*this.height),d){const v=this.channels,C=this.toCanvas(),F=l(w,y).getContext("2d");return F.drawImage(C,0,0,w,y),new _(F.getImageData(0,0,w,y).data,w,y,4).convert(v)}else{let v=this.toSharp();switch(T){case"box":case"hamming":(T==="box"||T==="hamming")&&(console.warn(`Resampling method ${T} is not yet supported. Using bilinear instead.`),T="bilinear");case"nearest":case"bilinear":case"bicubic":v=v.affine([w/this.width,0,0,y/this.height],{interpolator:T});break;case"lanczos":v=v.resize({width:w,height:y,fit:"fill",kernel:"lanczos3"});break;default:throw new Error(`Resampling method ${T} is not supported.`)}return await p(v)}}async pad([w,y,$,T]){if(w=Math.max(w,0),y=Math.max(y,0),$=Math.max($,0),T=Math.max(T,0),w===0&&y===0&&$===0&&T===0)return this;if(d){const M=this.channels,x=this.toCanvas(),v=this.width+w+y,C=this.height+$+T,F=l(v,C).getContext("2d");return F.drawImage(x,0,0,this.width,this.height,w,$,this.width,this.height),new _(F.getImageData(0,0,v,C).data,v,C,4).convert(M)}else{const M=this.toSharp().extend({left:w,right:y,top:$,bottom:T});return await p(M)}}async crop([w,y,$,T]){if(w=Math.max(w,0),y=Math.max(y,0),$=Math.min($,this.width-1),T=Math.min(T,this.height-1),w===0&&y===0&&$===this.width-1&&T===this.height-1)return this;const M=$-w+1,x=T-y+1;if(d){const v=this.channels,C=this.toCanvas(),F=l(M,x).getContext("2d");return F.drawImage(C,w,y,M,x,0,0,M,x),new _(F.getImageData(0,0,M,x).data,M,x,4).convert(v)}else{const v=this.toSharp().extract({left:w,top:y,width:M,height:x});return await p(v)}}async center_crop(w,y){if(this.width===w&&this.height===y)return this;const $=(this.width-w)/2,T=(this.height-y)/2;if(d){const M=this.channels,x=this.toCanvas(),v=l(w,y).getContext("2d");let C=0,F=0,D=0,K=0;return $>=0?C=$:D=-$,T>=0?F=T:K=-T,v.drawImage(x,C,F,w,y,D,K,w,y),new _(v.getImageData(0,0,w,y).data,w,y,4).convert(M)}else{let M=this.toSharp();if($>=0&&T>=0)M=M.extract({left:Math.floor($),top:Math.floor(T),width:w,height:y});else if($<=0&&T<=0){const x=Math.floor(-T),v=Math.floor(-$);M=M.extend({top:x,left:v,right:w-this.width-v,bottom:y-this.height-x})}else{let x=[0,0],v=0;T<0?(x[0]=Math.floor(-T),x[1]=y-this.height-x[0]):v=Math.floor(T);let C=[0,0],F=0;$<0?(C[0]=Math.floor(-$),C[1]=w-this.width-C[0]):F=Math.floor($),M=M.extend({top:x[0],bottom:x[1],left:C[0],right:C[1]}).extract({left:F,top:v,width:w,height:y})}return await p(M)}}async toBlob(w="image/png",y=1){if(!d)throw new Error("toBlob() is only supported in browser environments.");return await this.toCanvas().convertToBlob({type:w,quality:y})}toTensor(w="CHW"){let y=new a.Tensor("uint8",new Uint8Array(this.data),[this.height,this.width,this.channels]);if(w!=="HWC")if(w==="CHW")y=y.permute(2,0,1);else throw new Error(`Unsupported channel format: ${w}`);return y}toCanvas(){if(!d)throw new Error("toCanvas() is only supported in browser environments.");const w=this.clone().rgba(),y=l(w.width,w.height),$=new c(w.data,w.width,w.height);return y.getContext("2d").putImageData($,0,0),y}split(){const{data:w,width:y,height:$,channels:T}=this,M=w.constructor,x=w.length/T,v=Array.from({length:T},()=>new M(x));for(let C=0;C<x;++C){const F=T*C;for(let D=0;D<T;++D)v[D][C]=w[F+D]}return v.map(C=>new _(C,y,$,1))}_update(w,y,$,T=null){return this.data=w,this.width=y,this.height=$,T!==null&&(this.channels=T),this}clone(){return new _(this.data.slice(),this.width,this.height,this.channels)}convert(w){if(this.channels===w)return this;switch(w){case 1:this.grayscale();break;case 3:this.rgb();break;case 4:this.rgba();break;default:throw new Error(`Conversion failed due to unsupported number of channels: ${this.channels}`)}return this}async save(w){if(d){if(n.apis.IS_WEBWORKER_ENV)throw new Error("Unable to save an image from a Web Worker.");const y=w.split(".").pop().toLowerCase(),$=f.get(y)??"image/png",T=await this.toBlob($);(0,s.saveBlob)(w,T)}else{if(n.apis.IS_FS_AVAILABLE)return await this.toSharp().toFile(w);throw new Error("Unable to save the image because filesystem is disabled in this environment.")}}toSharp(){if(d)throw new Error("toSharp() is only supported in server-side environments.");return i(this.data,{raw:{width:this.width,height:this.height,channels:this.channels}})}}const b=_.read.bind(_)},"./src/utils/maths.js":(e,r,t)=>{t.r(r),t.d(r,{FFT:()=>b,bankers_round:()=>y,cos_sim:()=>l,dot:()=>i,dynamic_time_warping:()=>$,interpolate_data:()=>s,log_softmax:()=>a,magnitude:()=>c,max:()=>d,medianFilter:()=>k,min:()=>p,permute_data:()=>o,round:()=>w,softmax:()=>n});function s(T,[M,x,v],[C,F],D="bilinear",K=!1){const U=F/v,j=C/x,se=new T.constructor(C*F*M),q=x*v,re=C*F;for(let Z=0;Z<C;++Z)for(let ae=0;ae<F;++ae){const _e=Z*F+ae,pe=(ae+.5)/U-.5,oe=(Z+.5)/j-.5;let G=Math.floor(pe),B=Math.floor(oe);const Q=Math.min(G+1,v-1),le=Math.min(B+1,x-1);G=Math.max(G,0),B=Math.max(B,0);const ee=pe-G,Se=oe-B,De=(1-ee)*(1-Se),Ne=ee*(1-Se),Ae=(1-ee)*Se,te=ee*Se,O=B*v,me=le*v,xe=O+G,be=O+Q,Ce=me+G,$e=me+Q;for(let Me=0;Me<M;++Me){const Le=Me*q;se[Me*re+_e]=De*T[Le+xe]+Ne*T[Le+be]+Ae*T[Le+Ce]+te*T[Le+$e]}}return se}function o(T,M,x){const v=new Array(x.length),C=new Array(x.length);for(let K=x.length-1,U=1;K>=0;--K)C[K]=U,v[K]=M[x[K]],U*=v[K];const F=x.map((K,U)=>C[x.indexOf(U)]),D=new T.constructor(T.length);for(let K=0;K<T.length;++K){let U=0;for(let j=M.length-1,se=K;j>=0;--j)U+=se%M[j]*F[j],se=Math.floor(se/M[j]);D[U]=T[K]}return[D,v]}function n(T){const M=d(T)[0],x=T.map(F=>Math.exp(F-M)),v=x.reduce((F,D)=>F+D,0);return x.map(F=>F/v)}function a(T){const M=d(T)[0];let x=0;for(let F=0;F<T.length;++F)x+=Math.exp(T[F]-M);const v=Math.log(x);return T.map(F=>F-M-v)}function i(T,M){let x=0;for(let v=0;v<T.length;++v)x+=T[v]*M[v];return x}function l(T,M){const x=i(T,M),v=c(T),C=c(M);return x/(v*C)}function c(T){return Math.sqrt(T.reduce((M,x)=>M+x*x,0))}function p(T){if(T.length===0)throw Error("Array must not be empty");let M=T[0],x=0;for(let v=1;v<T.length;++v)T[v]<M&&(M=T[v],x=v);return[M,x]}function d(T){if(T.length===0)throw Error("Array must not be empty");let M=T[0],x=0;for(let v=1;v<T.length;++v)T[v]>M&&(M=T[v],x=v);return[M,x]}function u(T){return T>0&&(T&T-1)===0}class f{constructor(M){if(this.size=M|0,this.size<=1||!u(this.size))throw new Error("FFT size must be a power of two larger than 1");this._csize=M<<1,this.table=new Float64Array(this.size*2);for(let v=0;v<this.table.length;v+=2){const C=Math.PI*v/this.size;this.table[v]=Math.cos(C),this.table[v+1]=-Math.sin(C)}let x=0;for(let v=1;this.size>v;v<<=1)++x;this._width=x%2===0?x-1:x,this._bitrev=new Int32Array(1<<this._width);for(let v=0;v<this._bitrev.length;++v){this._bitrev[v]=0;for(let C=0;C<this._width;C+=2){const F=this._width-C-2;this._bitrev[v]|=(v>>>C&3)<<F}}}createComplexArray(){return new Float64Array(this._csize)}fromComplexArray(M,x){const v=x||new Array(M.length>>>1);for(let C=0;C<M.length;C+=2)v[C>>>1]=M[C];return v}toComplexArray(M,x){const v=x||this.createComplexArray();for(let C=0;C<v.length;C+=2)v[C]=M[C>>>1],v[C+1]=0;return v}transform(M,x){if(M===x)throw new Error("Input and output buffers must be different");this._transform4(M,x,1)}realTransform(M,x){if(M===x)throw new Error("Input and output buffers must be different");this._realTransform4(M,x,1)}inverseTransform(M,x){if(M===x)throw new Error("Input and output buffers must be different");this._transform4(M,x,-1);for(let v=0;v<M.length;++v)M[v]/=this.size}_transform4(M,x,v){const C=this._csize;let D=1<<this._width,K=C/D<<1,U,j;const se=this._bitrev;if(K===4)for(U=0,j=0;U<C;U+=K,++j){const re=se[j];this._singleTransform2(x,M,U,re,D)}else for(U=0,j=0;U<C;U+=K,++j){const re=se[j];this._singleTransform4(x,M,U,re,D,v)}const q=this.table;for(D>>=2;D>=2;D>>=2){K=C/D<<1;const re=K>>>2;for(U=0;U<C;U+=K){const Z=U+re-1;for(let ae=U,_e=0;ae<Z;ae+=2,_e+=D){const pe=ae,oe=pe+re,G=oe+re,B=G+re,Q=M[pe],le=M[pe+1],ee=M[oe],Se=M[oe+1],De=M[G],Ne=M[G+1],Ae=M[B],te=M[B+1],O=q[_e],me=v*q[_e+1],xe=ee*O-Se*me,be=ee*me+Se*O,Ce=q[2*_e],$e=v*q[2*_e+1],Me=De*Ce-Ne*$e,Le=De*$e+Ne*Ce,we=q[3*_e],ke=v*q[3*_e+1],He=Ae*we-te*ke,Fe=Ae*ke+te*we,Ze=Q+Me,We=le+Le,yt=Q-Me,Qe=le-Le,Oe=xe+He,ze=be+Fe,pt=v*(xe-He),ft=v*(be-Fe);M[pe]=Ze+Oe,M[pe+1]=We+ze,M[oe]=yt+ft,M[oe+1]=Qe-pt,M[G]=Ze-Oe,M[G+1]=We-ze,M[B]=yt-ft,M[B+1]=Qe+pt}}}}_singleTransform2(M,x,v,C,F){const D=M[C],K=M[C+1],U=M[C+F],j=M[C+F+1];x[v]=D+U,x[v+1]=K+j,x[v+2]=D-U,x[v+3]=K-j}_singleTransform4(M,x,v,C,F,D){const K=F*2,U=F*3,j=M[C],se=M[C+1],q=M[C+F],re=M[C+F+1],Z=M[C+K],ae=M[C+K+1],_e=M[C+U],pe=M[C+U+1],oe=j+Z,G=se+ae,B=j-Z,Q=se-ae,le=q+_e,ee=re+pe,Se=D*(q-_e),De=D*(re-pe);x[v]=oe+le,x[v+1]=G+ee,x[v+2]=B+De,x[v+3]=Q-Se,x[v+4]=oe-le,x[v+5]=G-ee,x[v+6]=B-De,x[v+7]=Q+Se}_realTransform4(M,x,v){const C=this._csize;let D=1<<this._width,K=C/D<<1,U,j;const se=this._bitrev;if(K===4)for(U=0,j=0;U<C;U+=K,++j){const Z=se[j];this._singleRealTransform2(x,M,U,Z>>>1,D>>>1)}else for(U=0,j=0;U<C;U+=K,++j){const Z=se[j];this._singleRealTransform4(x,M,U,Z>>>1,D>>>1,v)}const q=this.table;for(D>>=2;D>=2;D>>=2){K=C/D<<1;const Z=K>>>1,ae=Z>>>1,_e=ae>>>1;for(U=0;U<C;U+=K)for(let pe=0,oe=0;pe<=_e;pe+=2,oe+=D){const G=U+pe,B=G+ae,Q=B+ae,le=Q+ae,ee=M[G],Se=M[G+1],De=M[B],Ne=M[B+1],Ae=M[Q],te=M[Q+1],O=M[le],me=M[le+1],xe=ee,be=Se,Ce=q[oe],$e=v*q[oe+1],Me=De*Ce-Ne*$e,Le=De*$e+Ne*Ce,we=q[2*oe],ke=v*q[2*oe+1],He=Ae*we-te*ke,Fe=Ae*ke+te*we,Ze=q[3*oe],We=v*q[3*oe+1],yt=O*Ze-me*We,Qe=O*We+me*Ze,Oe=xe+He,ze=be+Fe,pt=xe-He,ft=be-Fe,Lt=Me+yt,kt=Le+Qe,Mr=v*(Me-yt),fr=v*(Le-Qe);if(M[G]=Oe+Lt,M[G+1]=ze+kt,M[B]=pt+fr,M[B+1]=ft-Mr,pe===0){M[Q]=Oe-Lt,M[Q+1]=ze-kt;continue}if(pe===_e)continue;const Ut=U+ae-pe,Ur=U+Z-pe;M[Ut]=pt-v*fr,M[Ut+1]=-ft-v*Mr,M[Ur]=Oe-v*Lt,M[Ur+1]=-ze+v*kt}}const re=C>>>1;for(let Z=2;Z<re;Z+=2)M[C-Z]=M[Z],M[C-Z+1]=-M[Z+1]}_singleRealTransform2(M,x,v,C,F){const D=M[C],K=M[C+F];x[v]=D+K,x[v+1]=0,x[v+2]=D-K,x[v+3]=0}_singleRealTransform4(M,x,v,C,F,D){const K=F*2,U=F*3,j=M[C],se=M[C+F],q=M[C+K],re=M[C+U],Z=j+q,ae=j-q,_e=se+re,pe=D*(se-re);x[v]=Z+_e,x[v+1]=0,x[v+2]=ae,x[v+3]=-pe,x[v+4]=Z-_e,x[v+5]=0,x[v+6]=ae,x[v+7]=pe}}class _{constructor(M){const x=2*(M-1),v=2*(2*M-1),C=2**Math.ceil(Math.log2(v));this.bufferSize=C,this._a=x;const F=new Float64Array(v),D=new Float64Array(C);this._chirpBuffer=new Float64Array(C),this._buffer1=new Float64Array(C),this._buffer2=new Float64Array(C),this._outBuffer1=new Float64Array(C),this._outBuffer2=new Float64Array(C);const K=-2*Math.PI/M,U=Math.cos(K),j=Math.sin(K);for(let se=0;se<v>>1;++se){const q=(se+1-M)**2/2,re=Math.sqrt(U**2+j**2)**q,Z=q*Math.atan2(j,U),ae=2*se;F[ae]=re*Math.cos(Z),F[ae+1]=re*Math.sin(Z),D[ae]=F[ae],D[ae+1]=-F[ae+1]}this._slicedChirpBuffer=F.subarray(x,v),this._f=new f(C>>1),this._f.transform(this._chirpBuffer,D)}_transform(M,x,v){const C=this._buffer1,F=this._buffer2,D=this._outBuffer1,K=this._outBuffer2,U=this._chirpBuffer,j=this._slicedChirpBuffer,se=this._a;if(v)for(let q=0;q<j.length;q+=2){const re=q+1,Z=q>>1,ae=x[Z];C[q]=ae*j[q],C[re]=ae*j[re]}else for(let q=0;q<j.length;q+=2){const re=q+1;C[q]=x[q]*j[q]-x[re]*j[re],C[re]=x[q]*j[re]+x[re]*j[q]}this._f.transform(D,C);for(let q=0;q<U.length;q+=2){const re=q+1;F[q]=D[q]*U[q]-D[re]*U[re],F[re]=D[q]*U[re]+D[re]*U[q]}this._f.inverseTransform(K,F);for(let q=0;q<K.length;q+=2){const re=K[q+se],Z=K[q+se+1],ae=j[q],_e=j[q+1];M[q]=re*ae-Z*_e,M[q+1]=re*_e+Z*ae}}transform(M,x){this._transform(M,x,!1)}realTransform(M,x){this._transform(M,x,!0)}}class b{constructor(M){this.fft_length=M,this.isPowerOfTwo=u(M),this.isPowerOfTwo?(this.fft=new f(M),this.outputBufferSize=2*M):(this.fft=new _(M),this.outputBufferSize=this.fft.bufferSize)}realTransform(M,x){this.fft.realTransform(M,x)}transform(M,x){this.fft.transform(M,x)}}function k(T,M){if(M%2===0||M<=0)throw new Error("Window size must be a positive odd number");const x=new T.constructor(T.length),v=new T.constructor(M),C=Math.floor(M/2);for(let F=0;F<T.length;++F){let D=0;for(let K=-C;K<=C;++K){let U=F+K;U<0?U=Math.abs(U):U>=T.length&&(U=2*(T.length-1)-U),v[D++]=T[U]}v.sort(),x[F]=v[C]}return x}function w(T,M){const x=Math.pow(10,M);return Math.round(T*x)/x}function y(T){const M=Math.round(T);return Math.abs(T)%1===.5?M%2===0?M:M-1:M}function $(T){const M=T.length,x=T[0].length,v=[M+1,x+1],C=Array.from({length:v[0]},()=>Array(v[1]).fill(1/0));C[0][0]=0;const F=Array.from({length:v[0]},()=>Array(v[1]).fill(-1));for(let se=1;se<v[1];++se)for(let q=1;q<v[0];++q){const re=C[q-1][se-1],Z=C[q-1][se],ae=C[q][se-1];let _e,pe;re<Z&&re<ae?(_e=re,pe=0):Z<re&&Z<ae?(_e=Z,pe=1):(_e=ae,pe=2),C[q][se]=T[q-1][se-1]+_e,F[q][se]=pe}for(let se=0;se<v[1];++se)F[0][se]=2;for(let se=0;se<v[0];++se)F[se][0]=1;let D=M,K=x,U=[],j=[];for(;D>0||K>0;)switch(U.push(D-1),j.push(K-1),F[D][K]){case 0:--D,--K;break;case 1:--D;break;case 2:--K;break;default:throw new Error(`Internal error in dynamic time warping. Unexpected trace[${D}, ${K}]. Please file a bug report.`)}return U.reverse(),j.reverse(),[U,j]}},"./src/utils/tensor.js":(e,r,t)=>{t.r(r),t.d(r,{DataTypeMap:()=>a,Tensor:()=>i,cat:()=>x,full:()=>j,full_like:()=>se,interpolate:()=>p,interpolate_4d:()=>d,layer_norm:()=>y,matmul:()=>u,mean:()=>D,mean_pooling:()=>w,ones:()=>q,ones_like:()=>re,permute:()=>c,quantize_embeddings:()=>oe,rand:()=>_e,randn:()=>pe,rfft:()=>f,slice:()=>k,stack:()=>v,std_mean:()=>F,topk:()=>_,zeros:()=>Z,zeros_like:()=>ae});var s=t("./src/utils/maths.js"),o=t("./src/backends/onnx.js"),n=t("./src/ops/registry.js");const a=Object.freeze({float32:Float32Array,float16:typeof Float16Array<"u"?Float16Array:Uint16Array,float64:Float64Array,string:Array,int8:Int8Array,uint8:Uint8Array,int16:Int16Array,uint16:Uint16Array,int32:Int32Array,uint32:Uint32Array,int64:BigInt64Array,uint64:BigUint64Array,bool:Uint8Array,uint4:Uint8Array,int4:Int8Array});class i{constructor(...B){J(this,"ort_tensor");return(0,o.isONNXTensor)(B[0])?this.ort_tensor=B[0]:this.ort_tensor=new o.Tensor(B[0],B[1],B[2]),new Proxy(this,{get:(Q,le)=>{if(typeof le=="string"){let ee=Number(le);if(Number.isInteger(ee))return Q._getitem(ee)}return Q[le]},set:(Q,le,ee)=>Q[le]=ee})}get dims(){return this.ort_tensor.dims}set dims(B){this.ort_tensor.dims=B}get type(){return this.ort_tensor.type}get data(){return this.ort_tensor.data}get size(){return this.ort_tensor.size}get location(){return this.ort_tensor.location}dispose(){this.ort_tensor.dispose()}*[Symbol.iterator](){const[B,...Q]=this.dims;if(Q.length>0){const le=Q.reduce((ee,Se)=>ee*Se);for(let ee=0;ee<B;++ee)yield this._subarray(ee,le,Q)}else yield*this.data}_getitem(B){const[Q,...le]=this.dims;if(B=M(B,Q),le.length>0){const ee=le.reduce((Se,De)=>Se*De);return this._subarray(B,ee,le)}else return new i(this.type,[this.data[B]],le)}indexOf(B){const Q=this.data;for(let le=0;le<Q.length;++le)if(Q[le]==B)return le;return-1}_subarray(B,Q,le){const ee=B*Q,Se=(B+1)*Q,De="subarray"in this.data?this.data.subarray(ee,Se):this.data.slice(ee,Se);return new i(this.type,De,le)}item(){const B=this.data;if(B.length!==1)throw new Error(`a Tensor with ${B.length} elements cannot be converted to Scalar`);return B[0]}tolist(){return l(this.data,this.dims)}sigmoid(){return this.clone().sigmoid_()}sigmoid_(){const B=this.data;for(let Q=0;Q<B.length;++Q)B[Q]=1/(1+Math.exp(-B[Q]));return this}map(B){return this.clone().map_(B)}map_(B){const Q=this.data;for(let le=0;le<Q.length;++le)Q[le]=B(Q[le],le,Q);return this}mul(B){return this.clone().mul_(B)}mul_(B){const Q=this.data;for(let le=0;le<Q.length;++le)Q[le]*=B;return this}div(B){return this.clone().div_(B)}div_(B){const Q=this.data;for(let le=0;le<Q.length;++le)Q[le]/=B;return this}add(B){return this.clone().add_(B)}add_(B){const Q=this.data;for(let le=0;le<Q.length;++le)Q[le]+=B;return this}sub(B){return this.clone().sub_(B)}sub_(B){const Q=this.data;for(let le=0;le<Q.length;++le)Q[le]-=B;return this}clone(){return new i(this.type,this.data.slice(),this.dims.slice())}slice(...B){const Q=[],le=[];for(let O=0;O<this.dims.length;++O){let me=B[O];if(me==null)le.push([0,this.dims[O]]),Q.push(this.dims[O]);else if(typeof me=="number")me=M(me,this.dims[O],O),le.push([me,me+1]);else if(Array.isArray(me)&&me.length===2){let[xe,be]=me;if(xe=xe===null?0:M(xe,this.dims[O],O,!1),be=be===null?this.dims[O]:M(be,this.dims[O],O,!1),xe>be)throw new Error(`Invalid slice: ${me}`);const Ce=[Math.max(xe,0),Math.min(be,this.dims[O])];le.push(Ce),Q.push(Ce[1]-Ce[0])}else throw new Error(`Invalid slice: ${me}`)}const ee=le.map(([O,me])=>me-O),Se=ee.reduce((O,me)=>O*me),De=this.data,Ne=new De.constructor(Se),Ae=this.stride();let te=!0;for(let O=1;O<ee.length;++O)if(le[O][0]!==0||le[O][1]!==this.dims[O]){te=!1;break}if(te){const O=le[0][0]*Ae[0],me=le[0][1]*Ae[0];if(ArrayBuffer.isView(De))Ne.set(De.subarray(O,me));else if(Array.isArray(De)){const xe=De.slice(O,me);for(let be=0;be<xe.length;++be)Ne[be]=xe[be]}else throw new Error("Unsupported data type for slicing")}else for(let O=0;O<Se;++O){let me=0;for(let xe=ee.length-1,be=O;xe>=0;--xe){const Ce=ee[xe];me+=(be%Ce+le[xe][0])*Ae[xe],be=Math.floor(be/Ce)}Ne[O]=De[me]}return new i(this.type,Ne,Q)}permute(...B){return c(this,B)}transpose(...B){return this.permute(...B)}sum(B=null,Q=!1){return this.norm(1,B,Q)}norm(B="fro",Q=null,le=!1){if(B==="fro")B=2;else if(typeof B=="string")throw Error(`Unsupported norm: ${B}`);const ee=this.data,Se=(te,O)=>te+O**B;if(Q===null){const te=ee.reduce(Se,0)**(1/B);return new i(this.type,[te],[])}const[De,Ne,Ae]=C(Se,this,Q,le);if(B!==1)for(let te=0;te<Ne.length;++te)Ne[te]=Ne[te]**(1/B);return new i(De,Ne,Ae)}normalize_(B=2,Q=1){Q=M(Q,this.dims.length);const le=this.norm(B,Q,!0),ee=this.data,Se=le.data;for(let De=0;De<ee.length;++De){let Ne=0;for(let Ae=this.dims.length-1,te=De,O=1;Ae>=0;--Ae){const me=this.dims[Ae];if(Ae!==Q){const xe=te%me;Ne+=xe*O,O*=this.dims[Ae]}te=Math.floor(te/me)}ee[De]/=Se[Ne]}return this}normalize(B=2,Q=1){return this.clone().normalize_(B,Q)}stride(){return K(this.dims)}squeeze(B=null){return new i(this.type,this.data,$(this.dims,B))}squeeze_(B=null){return this.dims=$(this.dims,B),this}unsqueeze(B=null){return new i(this.type,this.data,T(this.dims,B))}unsqueeze_(B=null){return this.dims=T(this.dims,B),this}flatten_(B=0,Q=-1){Q=(Q+this.dims.length)%this.dims.length;let le=this.dims.slice(0,B),ee=this.dims.slice(B,Q+1),Se=this.dims.slice(Q+1);return this.dims=[...le,ee.reduce((De,Ne)=>De*Ne,1),...Se],this}flatten(B=0,Q=-1){return this.clone().flatten_(B,Q)}view(...B){let Q=-1;for(let ee=0;ee<B.length;++ee)if(B[ee]===-1){if(Q!==-1)throw new Error("Only one dimension can be inferred");Q=ee}const le=this.data;if(Q!==-1){const ee=B.reduce((Se,De,Ne)=>Ne!==Q?Se*De:Se,1);B[Q]=le.length/ee}return new i(this.type,le,B)}neg_(){const B=this.data;for(let Q=0;Q<B.length;++Q)B[Q]=-B[Q];return this}neg(){return this.clone().neg_()}gt(B){const Q=new Uint8Array(this.data.length),le=this.data;for(let ee=0;ee<le.length;++ee)Q[ee]=le[ee]>B?1:0;return new i("bool",Q,this.dims)}lt(B){const Q=new Uint8Array(this.data.length),le=this.data;for(let ee=0;ee<le.length;++ee)Q[ee]=le[ee]<B?1:0;return new i("bool",Q,this.dims)}clamp_(B,Q){const le=this.data;for(let ee=0;ee<le.length;++ee)le[ee]=Math.min(Math.max(le[ee],B),Q);return this}clamp(B,Q){return this.clone().clamp_(B,Q)}round_(){const B=this.data;for(let Q=0;Q<B.length;++Q)B[Q]=Math.round(B[Q]);return this}round(){return this.clone().round_()}mean(B=null,Q=!1){return D(this,B,Q)}min(B=null,Q=!1){if(B===null){const De=(0,s.min)(this.data)[0];return new i(this.type,[De],[])}const[le,ee,Se]=C((De,Ne)=>Math.min(De,Ne),this,B,Q,1/0);return new i(le,ee,Se)}max(B=null,Q=!1){if(B===null){const De=(0,s.max)(this.data)[0];return new i(this.type,[De],[])}const[le,ee,Se]=C((De,Ne)=>Math.max(De,Ne),this,B,Q,-1/0);return new i(le,ee,Se)}argmin(B=null,Q=!1){if(B!==null)throw new Error("`dim !== null` not yet implemented.");const le=(0,s.min)(this.data)[1];return new i("int64",[BigInt(le)],[])}argmax(B=null,Q=!1){if(B!==null)throw new Error("`dim !== null` not yet implemented.");const le=(0,s.max)(this.data)[1];return new i("int64",[BigInt(le)],[])}to(B){if(this.type===B)return this;if(!a.hasOwnProperty(B))throw new Error(`Unsupported type: ${B}`);let Q;const le=["int64","uint64"].includes(this.type),ee=["int64","uint64"].includes(B);return le&&!ee?Q=Number:!le&&ee&&(["float16","float32","float64"].includes(this.type)?Q=Se=>BigInt(Math.floor(Se)):Q=BigInt),new i(B,a[B].from(this.data,Q),this.dims)}}function l(G,B){const Q=G.length,le=B.reduce((Se,De)=>Se*De);if(Q!==le)throw Error(`cannot reshape array of size ${Q} into shape (${B})`);let ee=G;for(let Se=B.length-1;Se>=0;Se--)ee=ee.reduce((De,Ne)=>{let Ae=De[De.length-1];return Ae.length<B[Se]?Ae.push(Ne):De.push([Ne]),De},[[]]);return ee[0]}function c(G,B){const[Q,le]=(0,s.permute_data)(G.data,G.dims,B);return new i(G.type,Q,le)}function p(G,[B,Q],le="bilinear",ee=!1){const Se=G.dims.at(-3)??1,De=G.dims.at(-2),Ne=G.dims.at(-1);let Ae=(0,s.interpolate_data)(G.data,[Se,De,Ne],[B,Q],le,ee);return new i(G.type,Ae,[Se,B,Q])}async function d(G,{size:B=null,mode:Q="bilinear"}={}){if(G.dims.length!==4)throw new Error("`interpolate_4d` currently only supports 4D input.");if(!B)throw new Error("`interpolate_4d` requires a `size` argument.");let le;if(B.length===2)le=[...G.dims.slice(0,2),...B];else if(B.length===3)le=[G.dims[0],...B];else if(B.length===4)le=B;else throw new Error("`size` must be of length 2, 3, or 4.");let ee;if(Q==="nearest")ee=await n.TensorOpRegistry.nearest_interpolate_4d;else if(Q==="bilinear")ee=await n.TensorOpRegistry.bilinear_interpolate_4d;else if(Q==="bicubic")ee=await n.TensorOpRegistry.bicubic_interpolate_4d;else throw new Error(`Unsupported mode: ${Q}`);const Se=new i("int64",new BigInt64Array(le.map(BigInt)),[le.length]);return await ee({x:G,s:Se})}async function u(G,B){return await(await n.TensorOpRegistry.matmul)({a:G,b:B})}async function f(G,B){return await(await n.TensorOpRegistry.rfft)({x:G,a:B})}async function _(G,B){const Q=await n.TensorOpRegistry.top_k;return B==null?B=G.dims.at(-1):B=Math.min(B,G.dims.at(-1)),await Q({x:G,k:new i("int64",[BigInt(B)],[1])})}const b=G=>new i("int64",G,[G.length]);async function k(G,B,Q,le,ee){return await(await n.TensorOpRegistry.slice)({x:G,s:b(B),e:b(Q),a:b(le),t:b(ee??new Array(le.length).fill(1))})}function w(G,B){const Q=G.data,le=B.data,ee=[G.dims[0],G.dims[2]],Se=new Q.constructor(ee[0]*ee[1]),[De,Ne,Ae]=G.dims;let te=0;for(let O=0;O<De;++O){const me=O*Ae*Ne;for(let xe=0;xe<Ae;++xe){let be=0,Ce=0;const $e=O*Ne,Me=me+xe;for(let we=0;we<Ne;++we){const ke=Number(le[$e+we]);Ce+=ke,be+=Q[Me+we*Ae]*ke}const Le=be/Ce;Se[te++]=Le}}return new i(G.type,Se,ee)}function y(G,B,{eps:Q=1e-5}={}){if(G.dims.length!==2)throw new Error("`layer_norm` currently only supports 2D input.");const[le,ee]=G.dims;if(B.length!==1&&B[0]!==ee)throw new Error("`normalized_shape` must be a 1D array with shape `[input.dims[1]]`.");const[Se,De]=F(G,1,0,!0),Ne=Se.data,Ae=De.data,te=G.data,O=new te.constructor(te.length);for(let me=0;me<le;++me){const xe=me*ee;for(let be=0;be<ee;++be){const Ce=xe+be;O[Ce]=(te[Ce]-Ae[me])/(Ne[me]+Q)}}return new i(G.type,O,G.dims)}function $(G,B){return G=G.slice(),B===null?G=G.filter(Q=>Q!==1):typeof B=="number"?G[B]===1&&G.splice(B,1):Array.isArray(B)&&(G=G.filter((Q,le)=>Q!==1||!B.includes(le))),G}function T(G,B){return B=M(B,G.length+1),G=G.slice(),G.splice(B,0,1),G}function M(G,B,Q=null,le=!0){if(G<-B||G>=B){if(le)throw new Error(`IndexError: index ${G} is out of bounds for dimension${Q===null?"":" "+Q} with size ${B}`);return G<-B?0:B}return G<0&&(G=(G%B+B)%B),G}function x(G,B=0){B=M(B,G[0].dims.length);const Q=G[0].dims.slice();Q[B]=G.reduce((De,Ne)=>De+Ne.dims[B],0);const le=Q.reduce((De,Ne)=>De*Ne,1),ee=new G[0].data.constructor(le),Se=G[0].type;if(B===0){let De=0;for(const Ne of G){const Ae=Ne.data;ee.set(Ae,De),De+=Ae.length}}else{let De=0;for(let Ne=0;Ne<G.length;++Ne){const{data:Ae,dims:te}=G[Ne];for(let O=0;O<Ae.length;++O){let me=0;for(let xe=te.length-1,be=O,Ce=1;xe>=0;--xe){const $e=te[xe];let Me=be%$e;xe===B&&(Me+=De),me+=Me*Ce,Ce*=Q[xe],be=Math.floor(be/$e)}ee[me]=Ae[O]}De+=te[B]}}return new i(Se,ee,Q)}function v(G,B=0){return x(G.map(Q=>Q.unsqueeze(B)),B)}function C(G,B,Q=null,le=!1,ee=null){const Se=B.data,De=B.dims;Q=M(Q,De.length);const Ne=De.slice();Ne[Q]=1;const Ae=new Se.constructor(Se.length/De[Q]);ee!==null&&Ae.fill(ee);for(let te=0;te<Se.length;++te){let O=0;for(let me=De.length-1,xe=te,be=1;me>=0;--me){const Ce=De[me];if(me!==Q){const $e=xe%Ce;O+=$e*be,be*=Ne[me]}xe=Math.floor(xe/Ce)}Ae[O]=G(Ae[O],Se[te],te,O)}return le||Ne.splice(Q,1),[B.type,Ae,Ne]}function F(G,B=null,Q=1,le=!1){const ee=G.data,Se=G.dims;if(B===null){const be=ee.reduce((Le,we)=>Le+we,0)/ee.length,Ce=Math.sqrt(ee.reduce((Le,we)=>Le+(we-be)**2,0)/(ee.length-Q)),$e=new i(G.type,[be],[]);return[new i(G.type,[Ce],[]),$e]}B=M(B,Se.length);const De=D(G,B,le),Ne=De.data,[Ae,te,O]=C((xe,be,Ce,$e)=>xe+(be-Ne[$e])**2,G,B,le);for(let xe=0;xe<te.length;++xe)te[xe]=Math.sqrt(te[xe]/(Se[B]-Q));return[new i(Ae,te,O),De]}function D(G,B=null,Q=!1){const le=G.dims,ee=G.data;if(B===null){const Ae=ee.reduce((te,O)=>te+O,0);return new i(G.type,[Ae/ee.length],[])}B=M(B,le.length);const[Se,De,Ne]=C((Ae,te)=>Ae+te,G,B,Q);if(le[B]!==1)for(let Ae=0;Ae<De.length;++Ae)De[Ae]/=le[B];return new i(Se,De,Ne)}function K(G){const B=new Array(G.length);for(let Q=G.length-1,le=1;Q>=0;--Q)B[Q]=le,le*=G[Q];return B}function U(G,B,Q,le){const ee=G.reduce((Se,De)=>Se*De,1);return new i(Q,new le(ee).fill(B),G)}function j(G,B){let Q,le;if(typeof B=="number")Q="float32",le=Float32Array;else if(typeof B=="bigint")Q="int64",le=BigInt64Array;else if(typeof B=="boolean")Q="bool",le=Uint8Array;else throw new Error(`Unsupported data type: ${typeof B}`);return U(G,B,Q,le)}function se(G,B){return j(G.dims,B)}function q(G){return U(G,1n,"int64",BigInt64Array)}function re(G){return q(G.dims)}function Z(G){return U(G,0n,"int64",BigInt64Array)}function ae(G){return Z(G.dims)}function _e(G){const B=G.reduce((Q,le)=>Q*le,1);return new i("float32",Float32Array.from({length:B},()=>Math.random()),G)}function pe(G){const B=G.reduce((le,ee)=>le*ee,1);function Q(){const le=1-Math.random(),ee=1-Math.random();return Math.sqrt(-2*Math.log(le))*Math.cos(2*Math.PI*ee)}return new i("float32",Float32Array.from({length:B},()=>Q()),G)}function oe(G,B){if(G.dims.length!==2)throw new Error("The tensor must have 2 dimensions");if(G.dims.at(-1)%8!==0)throw new Error("The last dimension of the tensor must be a multiple of 8");if(!["binary","ubinary"].includes(B))throw new Error("The precision must be either 'binary' or 'ubinary'");const Q=B==="binary",le=Q?"int8":"uint8",ee=Q?Int8Array:Uint8Array,Se=G.data,De=new ee(Se.length/8);for(let Ne=0;Ne<Se.length;++Ne){const Ae=Se[Ne]>0?1:0,te=Math.floor(Ne/8),O=Ne%8;De[te]|=Ae<<7-O,Q&&O===0&&(De[te]-=128)}return new i(le,De,[G.dims[0],G.dims[1]/8])}},"./src/utils/video.js":(e,r,t)=>{t.r(r),t.d(r,{RawVideo:()=>a,RawVideoFrame:()=>n,load_video:()=>i});var s=t("./src/utils/image.js"),o=t("./src/env.js");class n{constructor(c,p){this.image=c,this.timestamp=p}}class a{constructor(c,p){c.length>0&&c[0]instanceof s.RawImage&&(c=c.map((d,u)=>new n(d,(u+1)/(c.length+1)*p))),this.frames=c,this.duration=p}get width(){return this.frames[0].image.width}get height(){return this.frames[0].image.height}get fps(){return this.frames.length/this.duration}}async function i(l,{num_frames:c=null,fps:p=null}={}){if(!o.apis.IS_BROWSER_ENV)throw new Error("`load_video` is currently only supported in browser environments.");if(c==null&&p==null)throw new Error("Either num_frames or fps must be provided.");const d=[],u=document.createElement("video");if(u.crossOrigin="anonymous",u.muted=!0,typeof l=="string")u.src=l;else if(l instanceof Blob)u.src=URL.createObjectURL(l);else if(l instanceof HTMLVideoElement)u.src=l.src;else throw new Error("Invalid URL or video element provided.");if(await new Promise($=>u.onloadedmetadata=$),u.seekable.start(0)===u.seekable.end(0)){const T=await(await fetch(u.src)).blob();u.src=URL.createObjectURL(T),await new Promise(M=>u.onloadedmetadata=M)}const f=u.duration;let _,b;c!=null?(_=c,b=c===1?0:f/(c-1)):(b=1/p,_=Math.floor(f/b));let k=[];for(let $=0;$<_;++$)k.push(c===1?f/2:$*b);const w=document.createElement("canvas");w.width=u.videoWidth,w.height=u.videoHeight;const y=w.getContext("2d",{willReadFrequently:!0});for(const $ of k){u.currentTime=$,await new Promise(v=>{u.onseeked=v}),y.drawImage(u,0,0,w.width,w.height);const T=y.getImageData(0,0,w.width,w.height),M=new s.RawImage(T.data,w.width,w.height,4),x=new n(M,$);d.push(x)}return u.remove(),new a(d,f)}}},cw={};function Ht(e){var r=cw[e];if(r!==void 0)return r.exports;var t=cw[e]={exports:{}};return AT[e](t,t.exports,Ht),t.exports}(()=>{var e=Object.getPrototypeOf?t=>Object.getPrototypeOf(t):t=>t.__proto__,r;Ht.t=function(t,s){if(s&1&&(t=this(t)),s&8||typeof t=="object"&&t&&(s&4&&t.__esModule||s&16&&typeof t.then=="function"))return t;var o=Object.create(null);Ht.r(o);var n={};r=r||[null,e({}),e([]),e(e)];for(var a=s&2&&t;typeof a=="object"&&!~r.indexOf(a);a=e(a))Object.getOwnPropertyNames(a).forEach(i=>n[i]=()=>t[i]);return n.default=()=>t,Ht.d(o,n),o}})(),Ht.d=(e,r)=>{for(var t in r)Ht.o(r,t)&&!Ht.o(e,t)&&Object.defineProperty(e,t,{enumerable:!0,get:r[t]})},Ht.o=(e,r)=>Object.prototype.hasOwnProperty.call(e,r),Ht.r=e=>{typeof Symbol<"u"&&Symbol.toStringTag&&Object.defineProperty(e,Symbol.toStringTag,{value:"Module"}),Object.defineProperty(e,"__esModule",{value:!0})};var m={};(()=>{/*!*****************************!*\
  !*** ./src/transformers.js ***!
  \*****************************/Ht.r(m),Ht.d(m,{ASTFeatureExtractor:()=>d.ASTFeatureExtractor,ASTForAudioClassification:()=>t.ASTForAudioClassification,ASTModel:()=>t.ASTModel,ASTPreTrainedModel:()=>t.ASTPreTrainedModel,AlbertForMaskedLM:()=>t.AlbertForMaskedLM,AlbertForQuestionAnswering:()=>t.AlbertForQuestionAnswering,AlbertForSequenceClassification:()=>t.AlbertForSequenceClassification,AlbertModel:()=>t.AlbertModel,AlbertPreTrainedModel:()=>t.AlbertPreTrainedModel,AlbertTokenizer:()=>s.AlbertTokenizer,ArceeForCausalLM:()=>t.ArceeForCausalLM,ArceeModel:()=>t.ArceeModel,ArceePreTrainedModel:()=>t.ArceePreTrainedModel,AudioClassificationPipeline:()=>r.AudioClassificationPipeline,AutoConfig:()=>o.AutoConfig,AutoFeatureExtractor:()=>u.AutoFeatureExtractor,AutoImageProcessor:()=>b.AutoImageProcessor,AutoModel:()=>t.AutoModel,AutoModelForAudioClassification:()=>t.AutoModelForAudioClassification,AutoModelForAudioFrameClassification:()=>t.AutoModelForAudioFrameClassification,AutoModelForAudioTextToText:()=>t.AutoModelForAudioTextToText,AutoModelForCTC:()=>t.AutoModelForCTC,AutoModelForCausalLM:()=>t.AutoModelForCausalLM,AutoModelForDepthEstimation:()=>t.AutoModelForDepthEstimation,AutoModelForDocumentQuestionAnswering:()=>t.AutoModelForDocumentQuestionAnswering,AutoModelForImageClassification:()=>t.AutoModelForImageClassification,AutoModelForImageFeatureExtraction:()=>t.AutoModelForImageFeatureExtraction,AutoModelForImageMatting:()=>t.AutoModelForImageMatting,AutoModelForImageSegmentation:()=>t.AutoModelForImageSegmentation,AutoModelForImageTextToText:()=>t.AutoModelForImageTextToText,AutoModelForImageToImage:()=>t.AutoModelForImageToImage,AutoModelForMaskGeneration:()=>t.AutoModelForMaskGeneration,AutoModelForMaskedLM:()=>t.AutoModelForMaskedLM,AutoModelForNormalEstimation:()=>t.AutoModelForNormalEstimation,AutoModelForObjectDetection:()=>t.AutoModelForObjectDetection,AutoModelForPoseEstimation:()=>t.AutoModelForPoseEstimation,AutoModelForQuestionAnswering:()=>t.AutoModelForQuestionAnswering,AutoModelForSemanticSegmentation:()=>t.AutoModelForSemanticSegmentation,AutoModelForSeq2SeqLM:()=>t.AutoModelForSeq2SeqLM,AutoModelForSequenceClassification:()=>t.AutoModelForSequenceClassification,AutoModelForSpeechSeq2Seq:()=>t.AutoModelForSpeechSeq2Seq,AutoModelForTextToSpectrogram:()=>t.AutoModelForTextToSpectrogram,AutoModelForTextToWaveform:()=>t.AutoModelForTextToWaveform,AutoModelForTokenClassification:()=>t.AutoModelForTokenClassification,AutoModelForUniversalSegmentation:()=>t.AutoModelForUniversalSegmentation,AutoModelForVision2Seq:()=>t.AutoModelForVision2Seq,AutoModelForXVector:()=>t.AutoModelForXVector,AutoModelForZeroShotObjectDetection:()=>t.AutoModelForZeroShotObjectDetection,AutoProcessor:()=>y.AutoProcessor,AutoTokenizer:()=>s.AutoTokenizer,AutomaticSpeechRecognitionPipeline:()=>r.AutomaticSpeechRecognitionPipeline,BackgroundRemovalPipeline:()=>r.BackgroundRemovalPipeline,BartForConditionalGeneration:()=>t.BartForConditionalGeneration,BartForSequenceClassification:()=>t.BartForSequenceClassification,BartModel:()=>t.BartModel,BartPretrainedModel:()=>t.BartPretrainedModel,BartTokenizer:()=>s.BartTokenizer,BaseModelOutput:()=>t.BaseModelOutput,BaseStreamer:()=>$.BaseStreamer,BeitFeatureExtractor:()=>_.BeitFeatureExtractor,BeitForImageClassification:()=>t.BeitForImageClassification,BeitModel:()=>t.BeitModel,BeitPreTrainedModel:()=>t.BeitPreTrainedModel,BertForMaskedLM:()=>t.BertForMaskedLM,BertForQuestionAnswering:()=>t.BertForQuestionAnswering,BertForSequenceClassification:()=>t.BertForSequenceClassification,BertForTokenClassification:()=>t.BertForTokenClassification,BertModel:()=>t.BertModel,BertPreTrainedModel:()=>t.BertPreTrainedModel,BertTokenizer:()=>s.BertTokenizer,BitImageProcessor:()=>_.BitImageProcessor,BlenderbotForConditionalGeneration:()=>t.BlenderbotForConditionalGeneration,BlenderbotModel:()=>t.BlenderbotModel,BlenderbotPreTrainedModel:()=>t.BlenderbotPreTrainedModel,BlenderbotSmallForConditionalGeneration:()=>t.BlenderbotSmallForConditionalGeneration,BlenderbotSmallModel:()=>t.BlenderbotSmallModel,BlenderbotSmallPreTrainedModel:()=>t.BlenderbotSmallPreTrainedModel,BlenderbotSmallTokenizer:()=>s.BlenderbotSmallTokenizer,BlenderbotTokenizer:()=>s.BlenderbotTokenizer,BloomForCausalLM:()=>t.BloomForCausalLM,BloomModel:()=>t.BloomModel,BloomPreTrainedModel:()=>t.BloomPreTrainedModel,BloomTokenizer:()=>s.BloomTokenizer,CLIPFeatureExtractor:()=>_.CLIPFeatureExtractor,CLIPImageProcessor:()=>_.CLIPImageProcessor,CLIPModel:()=>t.CLIPModel,CLIPPreTrainedModel:()=>t.CLIPPreTrainedModel,CLIPSegForImageSegmentation:()=>t.CLIPSegForImageSegmentation,CLIPSegModel:()=>t.CLIPSegModel,CLIPSegPreTrainedModel:()=>t.CLIPSegPreTrainedModel,CLIPTextModel:()=>t.CLIPTextModel,CLIPTextModelWithProjection:()=>t.CLIPTextModelWithProjection,CLIPTokenizer:()=>s.CLIPTokenizer,CLIPVisionModel:()=>t.CLIPVisionModel,CLIPVisionModelWithProjection:()=>t.CLIPVisionModelWithProjection,CamembertForMaskedLM:()=>t.CamembertForMaskedLM,CamembertForQuestionAnswering:()=>t.CamembertForQuestionAnswering,CamembertForSequenceClassification:()=>t.CamembertForSequenceClassification,CamembertForTokenClassification:()=>t.CamembertForTokenClassification,CamembertModel:()=>t.CamembertModel,CamembertPreTrainedModel:()=>t.CamembertPreTrainedModel,CamembertTokenizer:()=>s.CamembertTokenizer,CausalLMOutput:()=>t.CausalLMOutput,CausalLMOutputWithPast:()=>t.CausalLMOutputWithPast,ChineseCLIPFeatureExtractor:()=>_.ChineseCLIPFeatureExtractor,ChineseCLIPModel:()=>t.ChineseCLIPModel,ChineseCLIPPreTrainedModel:()=>t.ChineseCLIPPreTrainedModel,ClapAudioModelWithProjection:()=>t.ClapAudioModelWithProjection,ClapFeatureExtractor:()=>d.ClapFeatureExtractor,ClapModel:()=>t.ClapModel,ClapPreTrainedModel:()=>t.ClapPreTrainedModel,ClapTextModelWithProjection:()=>t.ClapTextModelWithProjection,ClassifierFreeGuidanceLogitsProcessor:()=>M.ClassifierFreeGuidanceLogitsProcessor,CodeGenForCausalLM:()=>t.CodeGenForCausalLM,CodeGenModel:()=>t.CodeGenModel,CodeGenPreTrainedModel:()=>t.CodeGenPreTrainedModel,CodeGenTokenizer:()=>s.CodeGenTokenizer,CodeLlamaTokenizer:()=>s.CodeLlamaTokenizer,CohereForCausalLM:()=>t.CohereForCausalLM,CohereModel:()=>t.CohereModel,CoherePreTrainedModel:()=>t.CoherePreTrainedModel,CohereTokenizer:()=>s.CohereTokenizer,ConvBertForMaskedLM:()=>t.ConvBertForMaskedLM,ConvBertForQuestionAnswering:()=>t.ConvBertForQuestionAnswering,ConvBertForSequenceClassification:()=>t.ConvBertForSequenceClassification,ConvBertForTokenClassification:()=>t.ConvBertForTokenClassification,ConvBertModel:()=>t.ConvBertModel,ConvBertPreTrainedModel:()=>t.ConvBertPreTrainedModel,ConvBertTokenizer:()=>s.ConvBertTokenizer,ConvNextFeatureExtractor:()=>_.ConvNextFeatureExtractor,ConvNextForImageClassification:()=>t.ConvNextForImageClassification,ConvNextImageProcessor:()=>_.ConvNextImageProcessor,ConvNextModel:()=>t.ConvNextModel,ConvNextPreTrainedModel:()=>t.ConvNextPreTrainedModel,ConvNextV2ForImageClassification:()=>t.ConvNextV2ForImageClassification,ConvNextV2Model:()=>t.ConvNextV2Model,ConvNextV2PreTrainedModel:()=>t.ConvNextV2PreTrainedModel,DFineForObjectDetection:()=>t.DFineForObjectDetection,DFineModel:()=>t.DFineModel,DFinePreTrainedModel:()=>t.DFinePreTrainedModel,DINOv3ConvNextModel:()=>t.DINOv3ConvNextModel,DINOv3ConvNextPreTrainedModel:()=>t.DINOv3ConvNextPreTrainedModel,DINOv3ViTImageProcessor:()=>_.DINOv3ViTImageProcessor,DINOv3ViTModel:()=>t.DINOv3ViTModel,DINOv3ViTPreTrainedModel:()=>t.DINOv3ViTPreTrainedModel,DPTFeatureExtractor:()=>_.DPTFeatureExtractor,DPTForDepthEstimation:()=>t.DPTForDepthEstimation,DPTImageProcessor:()=>_.DPTImageProcessor,DPTModel:()=>t.DPTModel,DPTPreTrainedModel:()=>t.DPTPreTrainedModel,DacDecoderModel:()=>t.DacDecoderModel,DacDecoderOutput:()=>t.DacDecoderOutput,DacEncoderModel:()=>t.DacEncoderModel,DacEncoderOutput:()=>t.DacEncoderOutput,DacFeatureExtractor:()=>d.DacFeatureExtractor,DacModel:()=>t.DacModel,DacPreTrainedModel:()=>t.DacPreTrainedModel,DataTypeMap:()=>l.DataTypeMap,DebertaForMaskedLM:()=>t.DebertaForMaskedLM,DebertaForQuestionAnswering:()=>t.DebertaForQuestionAnswering,DebertaForSequenceClassification:()=>t.DebertaForSequenceClassification,DebertaForTokenClassification:()=>t.DebertaForTokenClassification,DebertaModel:()=>t.DebertaModel,DebertaPreTrainedModel:()=>t.DebertaPreTrainedModel,DebertaTokenizer:()=>s.DebertaTokenizer,DebertaV2ForMaskedLM:()=>t.DebertaV2ForMaskedLM,DebertaV2ForQuestionAnswering:()=>t.DebertaV2ForQuestionAnswering,DebertaV2ForSequenceClassification:()=>t.DebertaV2ForSequenceClassification,DebertaV2ForTokenClassification:()=>t.DebertaV2ForTokenClassification,DebertaV2Model:()=>t.DebertaV2Model,DebertaV2PreTrainedModel:()=>t.DebertaV2PreTrainedModel,DebertaV2Tokenizer:()=>s.DebertaV2Tokenizer,DecisionTransformerModel:()=>t.DecisionTransformerModel,DecisionTransformerPreTrainedModel:()=>t.DecisionTransformerPreTrainedModel,DeiTFeatureExtractor:()=>_.DeiTFeatureExtractor,DeiTForImageClassification:()=>t.DeiTForImageClassification,DeiTImageProcessor:()=>_.DeiTImageProcessor,DeiTModel:()=>t.DeiTModel,DeiTPreTrainedModel:()=>t.DeiTPreTrainedModel,DepthAnythingForDepthEstimation:()=>t.DepthAnythingForDepthEstimation,DepthAnythingPreTrainedModel:()=>t.DepthAnythingPreTrainedModel,DepthEstimationPipeline:()=>r.DepthEstimationPipeline,DepthProForDepthEstimation:()=>t.DepthProForDepthEstimation,DepthProPreTrainedModel:()=>t.DepthProPreTrainedModel,DetrFeatureExtractor:()=>_.DetrFeatureExtractor,DetrForObjectDetection:()=>t.DetrForObjectDetection,DetrForSegmentation:()=>t.DetrForSegmentation,DetrImageProcessor:()=>_.DetrImageProcessor,DetrModel:()=>t.DetrModel,DetrObjectDetectionOutput:()=>t.DetrObjectDetectionOutput,DetrPreTrainedModel:()=>t.DetrPreTrainedModel,DetrSegmentationOutput:()=>t.DetrSegmentationOutput,Dinov2ForImageClassification:()=>t.Dinov2ForImageClassification,Dinov2Model:()=>t.Dinov2Model,Dinov2PreTrainedModel:()=>t.Dinov2PreTrainedModel,Dinov2WithRegistersForImageClassification:()=>t.Dinov2WithRegistersForImageClassification,Dinov2WithRegistersModel:()=>t.Dinov2WithRegistersModel,Dinov2WithRegistersPreTrainedModel:()=>t.Dinov2WithRegistersPreTrainedModel,DistilBertForMaskedLM:()=>t.DistilBertForMaskedLM,DistilBertForQuestionAnswering:()=>t.DistilBertForQuestionAnswering,DistilBertForSequenceClassification:()=>t.DistilBertForSequenceClassification,DistilBertForTokenClassification:()=>t.DistilBertForTokenClassification,DistilBertModel:()=>t.DistilBertModel,DistilBertPreTrainedModel:()=>t.DistilBertPreTrainedModel,DistilBertTokenizer:()=>s.DistilBertTokenizer,DocumentQuestionAnsweringPipeline:()=>r.DocumentQuestionAnsweringPipeline,DonutFeatureExtractor:()=>_.DonutFeatureExtractor,DonutImageProcessor:()=>_.DonutImageProcessor,DonutSwinModel:()=>t.DonutSwinModel,DonutSwinPreTrainedModel:()=>t.DonutSwinPreTrainedModel,EdgeTamModel:()=>t.EdgeTamModel,EfficientNetForImageClassification:()=>t.EfficientNetForImageClassification,EfficientNetImageProcessor:()=>_.EfficientNetImageProcessor,EfficientNetModel:()=>t.EfficientNetModel,EfficientNetPreTrainedModel:()=>t.EfficientNetPreTrainedModel,ElectraForMaskedLM:()=>t.ElectraForMaskedLM,ElectraForQuestionAnswering:()=>t.ElectraForQuestionAnswering,ElectraForSequenceClassification:()=>t.ElectraForSequenceClassification,ElectraForTokenClassification:()=>t.ElectraForTokenClassification,ElectraModel:()=>t.ElectraModel,ElectraPreTrainedModel:()=>t.ElectraPreTrainedModel,ElectraTokenizer:()=>s.ElectraTokenizer,EncodecFeatureExtractor:()=>d.EncodecFeatureExtractor,EosTokenCriteria:()=>T.EosTokenCriteria,Ernie4_5ForCausalLM:()=>t.Ernie4_5ForCausalLM,Ernie4_5Model:()=>t.Ernie4_5Model,Ernie4_5PreTrainedModel:()=>t.Ernie4_5PreTrainedModel,EsmForMaskedLM:()=>t.EsmForMaskedLM,EsmForSequenceClassification:()=>t.EsmForSequenceClassification,EsmForTokenClassification:()=>t.EsmForTokenClassification,EsmModel:()=>t.EsmModel,EsmPreTrainedModel:()=>t.EsmPreTrainedModel,EsmTokenizer:()=>s.EsmTokenizer,ExaoneForCausalLM:()=>t.ExaoneForCausalLM,ExaoneModel:()=>t.ExaoneModel,ExaonePreTrainedModel:()=>t.ExaonePreTrainedModel,FFT:()=>c.FFT,FalconForCausalLM:()=>t.FalconForCausalLM,FalconModel:()=>t.FalconModel,FalconPreTrainedModel:()=>t.FalconPreTrainedModel,FalconTokenizer:()=>s.FalconTokenizer,FastViTForImageClassification:()=>t.FastViTForImageClassification,FastViTModel:()=>t.FastViTModel,FastViTPreTrainedModel:()=>t.FastViTPreTrainedModel,FeatureExtractionPipeline:()=>r.FeatureExtractionPipeline,FeatureExtractor:()=>p.FeatureExtractor,FillMaskPipeline:()=>r.FillMaskPipeline,Florence2ForConditionalGeneration:()=>t.Florence2ForConditionalGeneration,Florence2PreTrainedModel:()=>t.Florence2PreTrainedModel,Florence2Processor:()=>w.Florence2Processor,ForcedBOSTokenLogitsProcessor:()=>M.ForcedBOSTokenLogitsProcessor,ForcedEOSTokenLogitsProcessor:()=>M.ForcedEOSTokenLogitsProcessor,GLPNFeatureExtractor:()=>_.GLPNFeatureExtractor,GLPNForDepthEstimation:()=>t.GLPNForDepthEstimation,GLPNModel:()=>t.GLPNModel,GLPNPreTrainedModel:()=>t.GLPNPreTrainedModel,GPT2LMHeadModel:()=>t.GPT2LMHeadModel,GPT2Model:()=>t.GPT2Model,GPT2PreTrainedModel:()=>t.GPT2PreTrainedModel,GPT2Tokenizer:()=>s.GPT2Tokenizer,GPTBigCodeForCausalLM:()=>t.GPTBigCodeForCausalLM,GPTBigCodeModel:()=>t.GPTBigCodeModel,GPTBigCodePreTrainedModel:()=>t.GPTBigCodePreTrainedModel,GPTJForCausalLM:()=>t.GPTJForCausalLM,GPTJModel:()=>t.GPTJModel,GPTJPreTrainedModel:()=>t.GPTJPreTrainedModel,GPTNeoForCausalLM:()=>t.GPTNeoForCausalLM,GPTNeoModel:()=>t.GPTNeoModel,GPTNeoPreTrainedModel:()=>t.GPTNeoPreTrainedModel,GPTNeoXForCausalLM:()=>t.GPTNeoXForCausalLM,GPTNeoXModel:()=>t.GPTNeoXModel,GPTNeoXPreTrainedModel:()=>t.GPTNeoXPreTrainedModel,GPTNeoXTokenizer:()=>s.GPTNeoXTokenizer,Gemma2ForCausalLM:()=>t.Gemma2ForCausalLM,Gemma2Model:()=>t.Gemma2Model,Gemma2PreTrainedModel:()=>t.Gemma2PreTrainedModel,Gemma3ForCausalLM:()=>t.Gemma3ForCausalLM,Gemma3Model:()=>t.Gemma3Model,Gemma3PreTrainedModel:()=>t.Gemma3PreTrainedModel,Gemma3nAudioFeatureExtractor:()=>d.Gemma3nAudioFeatureExtractor,Gemma3nForConditionalGeneration:()=>t.Gemma3nForConditionalGeneration,Gemma3nPreTrainedModel:()=>t.Gemma3nPreTrainedModel,Gemma3nProcessor:()=>w.Gemma3nProcessor,GemmaForCausalLM:()=>t.GemmaForCausalLM,GemmaModel:()=>t.GemmaModel,GemmaPreTrainedModel:()=>t.GemmaPreTrainedModel,GemmaTokenizer:()=>s.GemmaTokenizer,GlmForCausalLM:()=>t.GlmForCausalLM,GlmModel:()=>t.GlmModel,GlmPreTrainedModel:()=>t.GlmPreTrainedModel,GraniteForCausalLM:()=>t.GraniteForCausalLM,GraniteModel:()=>t.GraniteModel,GraniteMoeHybridForCausalLM:()=>t.GraniteMoeHybridForCausalLM,GraniteMoeHybridModel:()=>t.GraniteMoeHybridModel,GraniteMoeHybridPreTrainedModel:()=>t.GraniteMoeHybridPreTrainedModel,GranitePreTrainedModel:()=>t.GranitePreTrainedModel,Grok1Tokenizer:()=>s.Grok1Tokenizer,GroundingDinoForObjectDetection:()=>t.GroundingDinoForObjectDetection,GroundingDinoImageProcessor:()=>_.GroundingDinoImageProcessor,GroundingDinoPreTrainedModel:()=>t.GroundingDinoPreTrainedModel,GroundingDinoProcessor:()=>w.GroundingDinoProcessor,GroupViTModel:()=>t.GroupViTModel,GroupViTPreTrainedModel:()=>t.GroupViTPreTrainedModel,HeliumForCausalLM:()=>t.HeliumForCausalLM,HeliumModel:()=>t.HeliumModel,HeliumPreTrainedModel:()=>t.HeliumPreTrainedModel,HerbertTokenizer:()=>s.HerbertTokenizer,HieraForImageClassification:()=>t.HieraForImageClassification,HieraModel:()=>t.HieraModel,HieraPreTrainedModel:()=>t.HieraPreTrainedModel,HubertForCTC:()=>t.HubertForCTC,HubertForSequenceClassification:()=>t.HubertForSequenceClassification,HubertModel:()=>t.HubertModel,HubertPreTrainedModel:()=>t.HubertPreTrainedModel,IJepaForImageClassification:()=>t.IJepaForImageClassification,IJepaModel:()=>t.IJepaModel,IJepaPreTrainedModel:()=>t.IJepaPreTrainedModel,Idefics3ForConditionalGeneration:()=>t.Idefics3ForConditionalGeneration,Idefics3ImageProcessor:()=>_.Idefics3ImageProcessor,Idefics3PreTrainedModel:()=>t.Idefics3PreTrainedModel,Idefics3Processor:()=>w.Idefics3Processor,ImageClassificationPipeline:()=>r.ImageClassificationPipeline,ImageFeatureExtractionPipeline:()=>r.ImageFeatureExtractionPipeline,ImageFeatureExtractor:()=>d.ImageFeatureExtractor,ImageMattingOutput:()=>t.ImageMattingOutput,ImageProcessor:()=>f.ImageProcessor,ImageSegmentationPipeline:()=>r.ImageSegmentationPipeline,ImageToImagePipeline:()=>r.ImageToImagePipeline,ImageToTextPipeline:()=>r.ImageToTextPipeline,InterruptableStoppingCriteria:()=>T.InterruptableStoppingCriteria,JAISLMHeadModel:()=>t.JAISLMHeadModel,JAISModel:()=>t.JAISModel,JAISPreTrainedModel:()=>t.JAISPreTrainedModel,JinaCLIPImageProcessor:()=>_.JinaCLIPImageProcessor,JinaCLIPModel:()=>t.JinaCLIPModel,JinaCLIPPreTrainedModel:()=>t.JinaCLIPPreTrainedModel,JinaCLIPProcessor:()=>w.JinaCLIPProcessor,JinaCLIPTextModel:()=>t.JinaCLIPTextModel,JinaCLIPVisionModel:()=>t.JinaCLIPVisionModel,Lfm2ForCausalLM:()=>t.Lfm2ForCausalLM,Lfm2Model:()=>t.Lfm2Model,Lfm2PreTrainedModel:()=>t.Lfm2PreTrainedModel,LiteWhisperForConditionalGeneration:()=>t.LiteWhisperForConditionalGeneration,Llama4ForCausalLM:()=>t.Llama4ForCausalLM,Llama4PreTrainedModel:()=>t.Llama4PreTrainedModel,LlamaForCausalLM:()=>t.LlamaForCausalLM,LlamaModel:()=>t.LlamaModel,LlamaPreTrainedModel:()=>t.LlamaPreTrainedModel,LlamaTokenizer:()=>s.LlamaTokenizer,LlavaForConditionalGeneration:()=>t.LlavaForConditionalGeneration,LlavaOnevisionForConditionalGeneration:()=>t.LlavaOnevisionForConditionalGeneration,LlavaOnevisionImageProcessor:()=>_.LlavaOnevisionImageProcessor,LlavaPreTrainedModel:()=>t.LlavaPreTrainedModel,LlavaProcessor:()=>w.LlavaProcessor,LlavaQwen2ForCausalLM:()=>t.LlavaQwen2ForCausalLM,LogitsProcessor:()=>M.LogitsProcessor,LogitsProcessorList:()=>M.LogitsProcessorList,LogitsWarper:()=>M.LogitsWarper,LongT5ForConditionalGeneration:()=>t.LongT5ForConditionalGeneration,LongT5Model:()=>t.LongT5Model,LongT5PreTrainedModel:()=>t.LongT5PreTrainedModel,M2M100ForConditionalGeneration:()=>t.M2M100ForConditionalGeneration,M2M100Model:()=>t.M2M100Model,M2M100PreTrainedModel:()=>t.M2M100PreTrainedModel,M2M100Tokenizer:()=>s.M2M100Tokenizer,MBart50Tokenizer:()=>s.MBart50Tokenizer,MBartForCausalLM:()=>t.MBartForCausalLM,MBartForConditionalGeneration:()=>t.MBartForConditionalGeneration,MBartForSequenceClassification:()=>t.MBartForSequenceClassification,MBartModel:()=>t.MBartModel,MBartPreTrainedModel:()=>t.MBartPreTrainedModel,MBartTokenizer:()=>s.MBartTokenizer,MPNetForMaskedLM:()=>t.MPNetForMaskedLM,MPNetForQuestionAnswering:()=>t.MPNetForQuestionAnswering,MPNetForSequenceClassification:()=>t.MPNetForSequenceClassification,MPNetForTokenClassification:()=>t.MPNetForTokenClassification,MPNetModel:()=>t.MPNetModel,MPNetPreTrainedModel:()=>t.MPNetPreTrainedModel,MPNetTokenizer:()=>s.MPNetTokenizer,MT5ForConditionalGeneration:()=>t.MT5ForConditionalGeneration,MT5Model:()=>t.MT5Model,MT5PreTrainedModel:()=>t.MT5PreTrainedModel,MarianMTModel:()=>t.MarianMTModel,MarianModel:()=>t.MarianModel,MarianPreTrainedModel:()=>t.MarianPreTrainedModel,MarianTokenizer:()=>s.MarianTokenizer,Mask2FormerImageProcessor:()=>_.Mask2FormerImageProcessor,MaskFormerFeatureExtractor:()=>_.MaskFormerFeatureExtractor,MaskFormerForInstanceSegmentation:()=>t.MaskFormerForInstanceSegmentation,MaskFormerImageProcessor:()=>_.MaskFormerImageProcessor,MaskFormerModel:()=>t.MaskFormerModel,MaskFormerPreTrainedModel:()=>t.MaskFormerPreTrainedModel,MaskedLMOutput:()=>t.MaskedLMOutput,MaxLengthCriteria:()=>T.MaxLengthCriteria,Metric3DForDepthEstimation:()=>t.Metric3DForDepthEstimation,Metric3DPreTrainedModel:()=>t.Metric3DPreTrainedModel,Metric3Dv2ForDepthEstimation:()=>t.Metric3Dv2ForDepthEstimation,Metric3Dv2PreTrainedModel:()=>t.Metric3Dv2PreTrainedModel,MgpstrForSceneTextRecognition:()=>t.MgpstrForSceneTextRecognition,MgpstrModelOutput:()=>t.MgpstrModelOutput,MgpstrPreTrainedModel:()=>t.MgpstrPreTrainedModel,MgpstrProcessor:()=>w.MgpstrProcessor,MgpstrTokenizer:()=>s.MgpstrTokenizer,MimiDecoderModel:()=>t.MimiDecoderModel,MimiDecoderOutput:()=>t.MimiDecoderOutput,MimiEncoderModel:()=>t.MimiEncoderModel,MimiEncoderOutput:()=>t.MimiEncoderOutput,MimiModel:()=>t.MimiModel,MimiPreTrainedModel:()=>t.MimiPreTrainedModel,MinLengthLogitsProcessor:()=>M.MinLengthLogitsProcessor,MinNewTokensLengthLogitsProcessor:()=>M.MinNewTokensLengthLogitsProcessor,Ministral3ForCausalLM:()=>t.Ministral3ForCausalLM,Ministral3Model:()=>t.Ministral3Model,Ministral3PreTrainedModel:()=>t.Ministral3PreTrainedModel,MinistralForCausalLM:()=>t.MinistralForCausalLM,MinistralModel:()=>t.MinistralModel,MinistralPreTrainedModel:()=>t.MinistralPreTrainedModel,Mistral3ForConditionalGeneration:()=>t.Mistral3ForConditionalGeneration,MistralForCausalLM:()=>t.MistralForCausalLM,MistralModel:()=>t.MistralModel,MistralPreTrainedModel:()=>t.MistralPreTrainedModel,MobileBertForMaskedLM:()=>t.MobileBertForMaskedLM,MobileBertForQuestionAnswering:()=>t.MobileBertForQuestionAnswering,MobileBertForSequenceClassification:()=>t.MobileBertForSequenceClassification,MobileBertModel:()=>t.MobileBertModel,MobileBertPreTrainedModel:()=>t.MobileBertPreTrainedModel,MobileBertTokenizer:()=>s.MobileBertTokenizer,MobileLLMForCausalLM:()=>t.MobileLLMForCausalLM,MobileLLMModel:()=>t.MobileLLMModel,MobileLLMPreTrainedModel:()=>t.MobileLLMPreTrainedModel,MobileNetV1FeatureExtractor:()=>_.MobileNetV1FeatureExtractor,MobileNetV1ForImageClassification:()=>t.MobileNetV1ForImageClassification,MobileNetV1ForSemanticSegmentation:()=>t.MobileNetV1ForSemanticSegmentation,MobileNetV1ImageProcessor:()=>_.MobileNetV1ImageProcessor,MobileNetV1Model:()=>t.MobileNetV1Model,MobileNetV1PreTrainedModel:()=>t.MobileNetV1PreTrainedModel,MobileNetV2FeatureExtractor:()=>_.MobileNetV2FeatureExtractor,MobileNetV2ForImageClassification:()=>t.MobileNetV2ForImageClassification,MobileNetV2ForSemanticSegmentation:()=>t.MobileNetV2ForSemanticSegmentation,MobileNetV2ImageProcessor:()=>_.MobileNetV2ImageProcessor,MobileNetV2Model:()=>t.MobileNetV2Model,MobileNetV2PreTrainedModel:()=>t.MobileNetV2PreTrainedModel,MobileNetV3FeatureExtractor:()=>_.MobileNetV3FeatureExtractor,MobileNetV3ForImageClassification:()=>t.MobileNetV3ForImageClassification,MobileNetV3ForSemanticSegmentation:()=>t.MobileNetV3ForSemanticSegmentation,MobileNetV3ImageProcessor:()=>_.MobileNetV3ImageProcessor,MobileNetV3Model:()=>t.MobileNetV3Model,MobileNetV3PreTrainedModel:()=>t.MobileNetV3PreTrainedModel,MobileNetV4FeatureExtractor:()=>_.MobileNetV4FeatureExtractor,MobileNetV4ForImageClassification:()=>t.MobileNetV4ForImageClassification,MobileNetV4ForSemanticSegmentation:()=>t.MobileNetV4ForSemanticSegmentation,MobileNetV4ImageProcessor:()=>_.MobileNetV4ImageProcessor,MobileNetV4Model:()=>t.MobileNetV4Model,MobileNetV4PreTrainedModel:()=>t.MobileNetV4PreTrainedModel,MobileViTFeatureExtractor:()=>_.MobileViTFeatureExtractor,MobileViTForImageClassification:()=>t.MobileViTForImageClassification,MobileViTImageProcessor:()=>_.MobileViTImageProcessor,MobileViTModel:()=>t.MobileViTModel,MobileViTPreTrainedModel:()=>t.MobileViTPreTrainedModel,MobileViTV2ForImageClassification:()=>t.MobileViTV2ForImageClassification,MobileViTV2Model:()=>t.MobileViTV2Model,MobileViTV2PreTrainedModel:()=>t.MobileViTV2PreTrainedModel,ModelOutput:()=>t.ModelOutput,ModernBertDecoderForCausalLM:()=>t.ModernBertDecoderForCausalLM,ModernBertDecoderModel:()=>t.ModernBertDecoderModel,ModernBertDecoderPreTrainedModel:()=>t.ModernBertDecoderPreTrainedModel,ModernBertForMaskedLM:()=>t.ModernBertForMaskedLM,ModernBertForSequenceClassification:()=>t.ModernBertForSequenceClassification,ModernBertForTokenClassification:()=>t.ModernBertForTokenClassification,ModernBertModel:()=>t.ModernBertModel,ModernBertPreTrainedModel:()=>t.ModernBertPreTrainedModel,Moondream1ForConditionalGeneration:()=>t.Moondream1ForConditionalGeneration,MoonshineFeatureExtractor:()=>d.MoonshineFeatureExtractor,MoonshineForConditionalGeneration:()=>t.MoonshineForConditionalGeneration,MoonshineModel:()=>t.MoonshineModel,MoonshinePreTrainedModel:()=>t.MoonshinePreTrainedModel,MoonshineProcessor:()=>w.MoonshineProcessor,MptForCausalLM:()=>t.MptForCausalLM,MptModel:()=>t.MptModel,MptPreTrainedModel:()=>t.MptPreTrainedModel,MultiModalityCausalLM:()=>t.MultiModalityCausalLM,MultiModalityPreTrainedModel:()=>t.MultiModalityPreTrainedModel,MusicgenForCausalLM:()=>t.MusicgenForCausalLM,MusicgenForConditionalGeneration:()=>t.MusicgenForConditionalGeneration,MusicgenModel:()=>t.MusicgenModel,MusicgenPreTrainedModel:()=>t.MusicgenPreTrainedModel,NanoChatForCausalLM:()=>t.NanoChatForCausalLM,NanoChatModel:()=>t.NanoChatModel,NanoChatPreTrainedModel:()=>t.NanoChatPreTrainedModel,NeoBertForMaskedLM:()=>t.NeoBertForMaskedLM,NeoBertForQuestionAnswering:()=>t.NeoBertForQuestionAnswering,NeoBertForSequenceClassification:()=>t.NeoBertForSequenceClassification,NeoBertForTokenClassification:()=>t.NeoBertForTokenClassification,NeoBertModel:()=>t.NeoBertModel,NeoBertPreTrainedModel:()=>t.NeoBertPreTrainedModel,NllbTokenizer:()=>s.NllbTokenizer,NoBadWordsLogitsProcessor:()=>M.NoBadWordsLogitsProcessor,NoRepeatNGramLogitsProcessor:()=>M.NoRepeatNGramLogitsProcessor,NomicBertModel:()=>t.NomicBertModel,NomicBertPreTrainedModel:()=>t.NomicBertPreTrainedModel,NougatImageProcessor:()=>_.NougatImageProcessor,NougatTokenizer:()=>s.NougatTokenizer,OPTForCausalLM:()=>t.OPTForCausalLM,OPTModel:()=>t.OPTModel,OPTPreTrainedModel:()=>t.OPTPreTrainedModel,ObjectDetectionPipeline:()=>r.ObjectDetectionPipeline,Olmo2ForCausalLM:()=>t.Olmo2ForCausalLM,Olmo2Model:()=>t.Olmo2Model,Olmo2PreTrainedModel:()=>t.Olmo2PreTrainedModel,OlmoForCausalLM:()=>t.OlmoForCausalLM,OlmoModel:()=>t.OlmoModel,OlmoPreTrainedModel:()=>t.OlmoPreTrainedModel,OpenELMForCausalLM:()=>t.OpenELMForCausalLM,OpenELMModel:()=>t.OpenELMModel,OpenELMPreTrainedModel:()=>t.OpenELMPreTrainedModel,OwlViTFeatureExtractor:()=>_.OwlViTFeatureExtractor,OwlViTForObjectDetection:()=>t.OwlViTForObjectDetection,OwlViTImageProcessor:()=>_.OwlViTImageProcessor,OwlViTModel:()=>t.OwlViTModel,OwlViTPreTrainedModel:()=>t.OwlViTPreTrainedModel,OwlViTProcessor:()=>w.OwlViTProcessor,Owlv2ForObjectDetection:()=>t.Owlv2ForObjectDetection,Owlv2ImageProcessor:()=>_.Owlv2ImageProcessor,Owlv2Model:()=>t.Owlv2Model,Owlv2PreTrainedModel:()=>t.Owlv2PreTrainedModel,PaliGemmaForConditionalGeneration:()=>t.PaliGemmaForConditionalGeneration,PaliGemmaPreTrainedModel:()=>t.PaliGemmaPreTrainedModel,PaliGemmaProcessor:()=>w.PaliGemmaProcessor,ParakeetFeatureExtractor:()=>d.ParakeetFeatureExtractor,ParakeetForCTC:()=>t.ParakeetForCTC,ParakeetPreTrainedModel:()=>t.ParakeetPreTrainedModel,PatchTSMixerForPrediction:()=>t.PatchTSMixerForPrediction,PatchTSMixerModel:()=>t.PatchTSMixerModel,PatchTSMixerPreTrainedModel:()=>t.PatchTSMixerPreTrainedModel,PatchTSTForPrediction:()=>t.PatchTSTForPrediction,PatchTSTModel:()=>t.PatchTSTModel,PatchTSTPreTrainedModel:()=>t.PatchTSTPreTrainedModel,Phi3ForCausalLM:()=>t.Phi3ForCausalLM,Phi3Model:()=>t.Phi3Model,Phi3PreTrainedModel:()=>t.Phi3PreTrainedModel,Phi3VForCausalLM:()=>t.Phi3VForCausalLM,Phi3VImageProcessor:()=>_.Phi3VImageProcessor,Phi3VPreTrainedModel:()=>t.Phi3VPreTrainedModel,Phi3VProcessor:()=>w.Phi3VProcessor,PhiForCausalLM:()=>t.PhiForCausalLM,PhiModel:()=>t.PhiModel,PhiPreTrainedModel:()=>t.PhiPreTrainedModel,Pipeline:()=>r.Pipeline,PixtralImageProcessor:()=>_.PixtralImageProcessor,PixtralProcessor:()=>w.PixtralProcessor,PreTrainedModel:()=>t.PreTrainedModel,PreTrainedTokenizer:()=>s.PreTrainedTokenizer,PretrainedConfig:()=>o.PretrainedConfig,PretrainedMixin:()=>t.PretrainedMixin,Processor:()=>k.Processor,PvtForImageClassification:()=>t.PvtForImageClassification,PvtImageProcessor:()=>_.PvtImageProcessor,PvtModel:()=>t.PvtModel,PvtPreTrainedModel:()=>t.PvtPreTrainedModel,PyAnnoteFeatureExtractor:()=>d.PyAnnoteFeatureExtractor,PyAnnoteForAudioFrameClassification:()=>t.PyAnnoteForAudioFrameClassification,PyAnnoteModel:()=>t.PyAnnoteModel,PyAnnotePreTrainedModel:()=>t.PyAnnotePreTrainedModel,PyAnnoteProcessor:()=>w.PyAnnoteProcessor,QuestionAnsweringModelOutput:()=>t.QuestionAnsweringModelOutput,QuestionAnsweringPipeline:()=>r.QuestionAnsweringPipeline,Qwen2ForCausalLM:()=>t.Qwen2ForCausalLM,Qwen2Model:()=>t.Qwen2Model,Qwen2PreTrainedModel:()=>t.Qwen2PreTrainedModel,Qwen2Tokenizer:()=>s.Qwen2Tokenizer,Qwen2VLForConditionalGeneration:()=>t.Qwen2VLForConditionalGeneration,Qwen2VLImageProcessor:()=>_.Qwen2VLImageProcessor,Qwen2VLPreTrainedModel:()=>t.Qwen2VLPreTrainedModel,Qwen2VLProcessor:()=>w.Qwen2VLProcessor,Qwen3ForCausalLM:()=>t.Qwen3ForCausalLM,Qwen3Model:()=>t.Qwen3Model,Qwen3PreTrainedModel:()=>t.Qwen3PreTrainedModel,RFDetrForObjectDetection:()=>t.RFDetrForObjectDetection,RFDetrModel:()=>t.RFDetrModel,RFDetrObjectDetectionOutput:()=>t.RFDetrObjectDetectionOutput,RFDetrPreTrainedModel:()=>t.RFDetrPreTrainedModel,RTDetrForObjectDetection:()=>t.RTDetrForObjectDetection,RTDetrImageProcessor:()=>_.RTDetrImageProcessor,RTDetrModel:()=>t.RTDetrModel,RTDetrObjectDetectionOutput:()=>t.RTDetrObjectDetectionOutput,RTDetrPreTrainedModel:()=>t.RTDetrPreTrainedModel,RTDetrV2ForObjectDetection:()=>t.RTDetrV2ForObjectDetection,RTDetrV2Model:()=>t.RTDetrV2Model,RTDetrV2ObjectDetectionOutput:()=>t.RTDetrV2ObjectDetectionOutput,RTDetrV2PreTrainedModel:()=>t.RTDetrV2PreTrainedModel,RawAudio:()=>n.RawAudio,RawImage:()=>a.RawImage,RawVideo:()=>i.RawVideo,RawVideoFrame:()=>i.RawVideoFrame,RepetitionPenaltyLogitsProcessor:()=>M.RepetitionPenaltyLogitsProcessor,ResNetForImageClassification:()=>t.ResNetForImageClassification,ResNetModel:()=>t.ResNetModel,ResNetPreTrainedModel:()=>t.ResNetPreTrainedModel,RoFormerForMaskedLM:()=>t.RoFormerForMaskedLM,RoFormerForQuestionAnswering:()=>t.RoFormerForQuestionAnswering,RoFormerForSequenceClassification:()=>t.RoFormerForSequenceClassification,RoFormerForTokenClassification:()=>t.RoFormerForTokenClassification,RoFormerModel:()=>t.RoFormerModel,RoFormerPreTrainedModel:()=>t.RoFormerPreTrainedModel,RoFormerTokenizer:()=>s.RoFormerTokenizer,RobertaForMaskedLM:()=>t.RobertaForMaskedLM,RobertaForQuestionAnswering:()=>t.RobertaForQuestionAnswering,RobertaForSequenceClassification:()=>t.RobertaForSequenceClassification,RobertaForTokenClassification:()=>t.RobertaForTokenClassification,RobertaModel:()=>t.RobertaModel,RobertaPreTrainedModel:()=>t.RobertaPreTrainedModel,RobertaTokenizer:()=>s.RobertaTokenizer,Sam2ImageProcessor:()=>_.Sam2ImageProcessor,Sam2ImageSegmentationOutput:()=>t.Sam2ImageSegmentationOutput,Sam2Model:()=>t.Sam2Model,Sam2PreTrainedModel:()=>t.Sam2PreTrainedModel,Sam2Processor:()=>w.Sam2Processor,Sam2VideoProcessor:()=>w.Sam2VideoProcessor,Sam3ImageProcessor:()=>_.Sam3ImageProcessor,Sam3TrackerModel:()=>t.Sam3TrackerModel,SamImageProcessor:()=>_.SamImageProcessor,SamImageSegmentationOutput:()=>t.SamImageSegmentationOutput,SamModel:()=>t.SamModel,SamPreTrainedModel:()=>t.SamPreTrainedModel,SamProcessor:()=>w.SamProcessor,SapiensForDepthEstimation:()=>t.SapiensForDepthEstimation,SapiensForNormalEstimation:()=>t.SapiensForNormalEstimation,SapiensForSemanticSegmentation:()=>t.SapiensForSemanticSegmentation,SapiensPreTrainedModel:()=>t.SapiensPreTrainedModel,SeamlessM4TFeatureExtractor:()=>d.SeamlessM4TFeatureExtractor,SegformerFeatureExtractor:()=>_.SegformerFeatureExtractor,SegformerForImageClassification:()=>t.SegformerForImageClassification,SegformerForSemanticSegmentation:()=>t.SegformerForSemanticSegmentation,SegformerImageProcessor:()=>_.SegformerImageProcessor,SegformerModel:()=>t.SegformerModel,SegformerPreTrainedModel:()=>t.SegformerPreTrainedModel,Seq2SeqLMOutput:()=>t.Seq2SeqLMOutput,SequenceClassifierOutput:()=>t.SequenceClassifierOutput,SiglipImageProcessor:()=>_.SiglipImageProcessor,SiglipModel:()=>t.SiglipModel,SiglipPreTrainedModel:()=>t.SiglipPreTrainedModel,SiglipTextModel:()=>t.SiglipTextModel,SiglipTokenizer:()=>s.SiglipTokenizer,SiglipVisionModel:()=>t.SiglipVisionModel,SmolLM3ForCausalLM:()=>t.SmolLM3ForCausalLM,SmolLM3Model:()=>t.SmolLM3Model,SmolLM3PreTrainedModel:()=>t.SmolLM3PreTrainedModel,SmolVLMForConditionalGeneration:()=>t.SmolVLMForConditionalGeneration,SmolVLMImageProcessor:()=>_.SmolVLMImageProcessor,SmolVLMProcessor:()=>w.SmolVLMProcessor,SnacDecoderModel:()=>t.SnacDecoderModel,SnacEncoderModel:()=>t.SnacEncoderModel,SnacFeatureExtractor:()=>d.SnacFeatureExtractor,SnacModel:()=>t.SnacModel,SnacPreTrainedModel:()=>t.SnacPreTrainedModel,SpeechT5FeatureExtractor:()=>d.SpeechT5FeatureExtractor,SpeechT5ForSpeechToText:()=>t.SpeechT5ForSpeechToText,SpeechT5ForTextToSpeech:()=>t.SpeechT5ForTextToSpeech,SpeechT5HifiGan:()=>t.SpeechT5HifiGan,SpeechT5Model:()=>t.SpeechT5Model,SpeechT5PreTrainedModel:()=>t.SpeechT5PreTrainedModel,SpeechT5Processor:()=>w.SpeechT5Processor,SpeechT5Tokenizer:()=>s.SpeechT5Tokenizer,SqueezeBertForMaskedLM:()=>t.SqueezeBertForMaskedLM,SqueezeBertForQuestionAnswering:()=>t.SqueezeBertForQuestionAnswering,SqueezeBertForSequenceClassification:()=>t.SqueezeBertForSequenceClassification,SqueezeBertModel:()=>t.SqueezeBertModel,SqueezeBertPreTrainedModel:()=>t.SqueezeBertPreTrainedModel,SqueezeBertTokenizer:()=>s.SqueezeBertTokenizer,StableLmForCausalLM:()=>t.StableLmForCausalLM,StableLmModel:()=>t.StableLmModel,StableLmPreTrainedModel:()=>t.StableLmPreTrainedModel,Starcoder2ForCausalLM:()=>t.Starcoder2ForCausalLM,Starcoder2Model:()=>t.Starcoder2Model,Starcoder2PreTrainedModel:()=>t.Starcoder2PreTrainedModel,StoppingCriteria:()=>T.StoppingCriteria,StoppingCriteriaList:()=>T.StoppingCriteriaList,StyleTextToSpeech2Model:()=>t.StyleTextToSpeech2Model,StyleTextToSpeech2PreTrainedModel:()=>t.StyleTextToSpeech2PreTrainedModel,SummarizationPipeline:()=>r.SummarizationPipeline,SupertonicForConditionalGeneration:()=>t.SupertonicForConditionalGeneration,SupertonicPreTrainedModel:()=>t.SupertonicPreTrainedModel,SuppressTokensAtBeginLogitsProcessor:()=>M.SuppressTokensAtBeginLogitsProcessor,Swin2SRForImageSuperResolution:()=>t.Swin2SRForImageSuperResolution,Swin2SRImageProcessor:()=>_.Swin2SRImageProcessor,Swin2SRModel:()=>t.Swin2SRModel,Swin2SRPreTrainedModel:()=>t.Swin2SRPreTrainedModel,SwinForImageClassification:()=>t.SwinForImageClassification,SwinForSemanticSegmentation:()=>t.SwinForSemanticSegmentation,SwinModel:()=>t.SwinModel,SwinPreTrainedModel:()=>t.SwinPreTrainedModel,T5ForConditionalGeneration:()=>t.T5ForConditionalGeneration,T5Model:()=>t.T5Model,T5PreTrainedModel:()=>t.T5PreTrainedModel,T5Tokenizer:()=>s.T5Tokenizer,TableTransformerForObjectDetection:()=>t.TableTransformerForObjectDetection,TableTransformerModel:()=>t.TableTransformerModel,TableTransformerObjectDetectionOutput:()=>t.TableTransformerObjectDetectionOutput,TableTransformerPreTrainedModel:()=>t.TableTransformerPreTrainedModel,TemperatureLogitsWarper:()=>M.TemperatureLogitsWarper,Tensor:()=>l.Tensor,Text2TextGenerationPipeline:()=>r.Text2TextGenerationPipeline,TextClassificationPipeline:()=>r.TextClassificationPipeline,TextGenerationPipeline:()=>r.TextGenerationPipeline,TextStreamer:()=>$.TextStreamer,TextToAudioPipeline:()=>r.TextToAudioPipeline,TokenClassificationPipeline:()=>r.TokenClassificationPipeline,TokenClassifierOutput:()=>t.TokenClassifierOutput,TokenizerModel:()=>s.TokenizerModel,TopKLogitsWarper:()=>M.TopKLogitsWarper,TopPLogitsWarper:()=>M.TopPLogitsWarper,TrOCRForCausalLM:()=>t.TrOCRForCausalLM,TrOCRPreTrainedModel:()=>t.TrOCRPreTrainedModel,TranslationPipeline:()=>r.TranslationPipeline,UltravoxModel:()=>t.UltravoxModel,UltravoxPreTrainedModel:()=>t.UltravoxPreTrainedModel,UltravoxProcessor:()=>w.UltravoxProcessor,UniSpeechForCTC:()=>t.UniSpeechForCTC,UniSpeechForSequenceClassification:()=>t.UniSpeechForSequenceClassification,UniSpeechModel:()=>t.UniSpeechModel,UniSpeechPreTrainedModel:()=>t.UniSpeechPreTrainedModel,UniSpeechSatForAudioFrameClassification:()=>t.UniSpeechSatForAudioFrameClassification,UniSpeechSatForCTC:()=>t.UniSpeechSatForCTC,UniSpeechSatForSequenceClassification:()=>t.UniSpeechSatForSequenceClassification,UniSpeechSatModel:()=>t.UniSpeechSatModel,UniSpeechSatPreTrainedModel:()=>t.UniSpeechSatPreTrainedModel,VLChatProcessor:()=>w.VLChatProcessor,VLMImageProcessor:()=>_.VLMImageProcessor,VaultGemmaForCausalLM:()=>t.VaultGemmaForCausalLM,VaultGemmaModel:()=>t.VaultGemmaModel,VaultGemmaPreTrainedModel:()=>t.VaultGemmaPreTrainedModel,ViTFeatureExtractor:()=>_.ViTFeatureExtractor,ViTForImageClassification:()=>t.ViTForImageClassification,ViTImageProcessor:()=>_.ViTImageProcessor,ViTMAEModel:()=>t.ViTMAEModel,ViTMAEPreTrainedModel:()=>t.ViTMAEPreTrainedModel,ViTMSNForImageClassification:()=>t.ViTMSNForImageClassification,ViTMSNModel:()=>t.ViTMSNModel,ViTMSNPreTrainedModel:()=>t.ViTMSNPreTrainedModel,ViTModel:()=>t.ViTModel,ViTPreTrainedModel:()=>t.ViTPreTrainedModel,VisionEncoderDecoderModel:()=>t.VisionEncoderDecoderModel,VitMatteForImageMatting:()=>t.VitMatteForImageMatting,VitMatteImageProcessor:()=>_.VitMatteImageProcessor,VitMattePreTrainedModel:()=>t.VitMattePreTrainedModel,VitPoseForPoseEstimation:()=>t.VitPoseForPoseEstimation,VitPoseImageProcessor:()=>_.VitPoseImageProcessor,VitPosePreTrainedModel:()=>t.VitPosePreTrainedModel,VitsModel:()=>t.VitsModel,VitsModelOutput:()=>t.VitsModelOutput,VitsPreTrainedModel:()=>t.VitsPreTrainedModel,VitsTokenizer:()=>s.VitsTokenizer,VoxtralForConditionalGeneration:()=>t.VoxtralForConditionalGeneration,VoxtralProcessor:()=>w.VoxtralProcessor,Wav2Vec2BertForCTC:()=>t.Wav2Vec2BertForCTC,Wav2Vec2BertForSequenceClassification:()=>t.Wav2Vec2BertForSequenceClassification,Wav2Vec2BertModel:()=>t.Wav2Vec2BertModel,Wav2Vec2BertPreTrainedModel:()=>t.Wav2Vec2BertPreTrainedModel,Wav2Vec2CTCTokenizer:()=>s.Wav2Vec2CTCTokenizer,Wav2Vec2FeatureExtractor:()=>d.Wav2Vec2FeatureExtractor,Wav2Vec2ForAudioFrameClassification:()=>t.Wav2Vec2ForAudioFrameClassification,Wav2Vec2ForCTC:()=>t.Wav2Vec2ForCTC,Wav2Vec2ForSequenceClassification:()=>t.Wav2Vec2ForSequenceClassification,Wav2Vec2Model:()=>t.Wav2Vec2Model,Wav2Vec2PreTrainedModel:()=>t.Wav2Vec2PreTrainedModel,Wav2Vec2Processor:()=>w.Wav2Vec2Processor,Wav2Vec2ProcessorWithLM:()=>w.Wav2Vec2ProcessorWithLM,WavLMForAudioFrameClassification:()=>t.WavLMForAudioFrameClassification,WavLMForCTC:()=>t.WavLMForCTC,WavLMForSequenceClassification:()=>t.WavLMForSequenceClassification,WavLMForXVector:()=>t.WavLMForXVector,WavLMModel:()=>t.WavLMModel,WavLMPreTrainedModel:()=>t.WavLMPreTrainedModel,WeSpeakerFeatureExtractor:()=>d.WeSpeakerFeatureExtractor,WeSpeakerResNetModel:()=>t.WeSpeakerResNetModel,WeSpeakerResNetPreTrainedModel:()=>t.WeSpeakerResNetPreTrainedModel,WhisperFeatureExtractor:()=>d.WhisperFeatureExtractor,WhisperForConditionalGeneration:()=>t.WhisperForConditionalGeneration,WhisperModel:()=>t.WhisperModel,WhisperPreTrainedModel:()=>t.WhisperPreTrainedModel,WhisperProcessor:()=>w.WhisperProcessor,WhisperTextStreamer:()=>$.WhisperTextStreamer,WhisperTimeStampLogitsProcessor:()=>M.WhisperTimeStampLogitsProcessor,WhisperTokenizer:()=>s.WhisperTokenizer,XLMForQuestionAnswering:()=>t.XLMForQuestionAnswering,XLMForSequenceClassification:()=>t.XLMForSequenceClassification,XLMForTokenClassification:()=>t.XLMForTokenClassification,XLMModel:()=>t.XLMModel,XLMPreTrainedModel:()=>t.XLMPreTrainedModel,XLMRobertaForMaskedLM:()=>t.XLMRobertaForMaskedLM,XLMRobertaForQuestionAnswering:()=>t.XLMRobertaForQuestionAnswering,XLMRobertaForSequenceClassification:()=>t.XLMRobertaForSequenceClassification,XLMRobertaForTokenClassification:()=>t.XLMRobertaForTokenClassification,XLMRobertaModel:()=>t.XLMRobertaModel,XLMRobertaPreTrainedModel:()=>t.XLMRobertaPreTrainedModel,XLMRobertaTokenizer:()=>s.XLMRobertaTokenizer,XLMTokenizer:()=>s.XLMTokenizer,XLMWithLMHeadModel:()=>t.XLMWithLMHeadModel,XVectorOutput:()=>t.XVectorOutput,YolosFeatureExtractor:()=>_.YolosFeatureExtractor,YolosForObjectDetection:()=>t.YolosForObjectDetection,YolosImageProcessor:()=>_.YolosImageProcessor,YolosModel:()=>t.YolosModel,YolosObjectDetectionOutput:()=>t.YolosObjectDetectionOutput,YolosPreTrainedModel:()=>t.YolosPreTrainedModel,ZeroShotAudioClassificationPipeline:()=>r.ZeroShotAudioClassificationPipeline,ZeroShotClassificationPipeline:()=>r.ZeroShotClassificationPipeline,ZeroShotImageClassificationPipeline:()=>r.ZeroShotImageClassificationPipeline,ZeroShotObjectDetectionPipeline:()=>r.ZeroShotObjectDetectionPipeline,bankers_round:()=>c.bankers_round,cat:()=>l.cat,cos_sim:()=>c.cos_sim,dot:()=>c.dot,dynamic_time_warping:()=>c.dynamic_time_warping,env:()=>e.env,full:()=>l.full,full_like:()=>l.full_like,getCacheShapes:()=>o.getCacheShapes,hamming:()=>n.hamming,hanning:()=>n.hanning,interpolate:()=>l.interpolate,interpolate_4d:()=>l.interpolate_4d,interpolate_data:()=>c.interpolate_data,is_chinese_char:()=>s.is_chinese_char,layer_norm:()=>l.layer_norm,load_image:()=>a.load_image,load_video:()=>i.load_video,log_softmax:()=>c.log_softmax,magnitude:()=>c.magnitude,matmul:()=>l.matmul,max:()=>c.max,mean:()=>l.mean,mean_pooling:()=>l.mean_pooling,medianFilter:()=>c.medianFilter,mel_filter_bank:()=>n.mel_filter_bank,min:()=>c.min,ones:()=>l.ones,ones_like:()=>l.ones_like,permute:()=>l.permute,permute_data:()=>c.permute_data,pipeline:()=>r.pipeline,quantize_embeddings:()=>l.quantize_embeddings,rand:()=>l.rand,randn:()=>l.randn,read_audio:()=>n.read_audio,rfft:()=>l.rfft,round:()=>c.round,slice:()=>l.slice,softmax:()=>c.softmax,spectrogram:()=>n.spectrogram,stack:()=>l.stack,std_mean:()=>l.std_mean,topk:()=>l.topk,window_function:()=>n.window_function,zeros:()=>l.zeros,zeros_like:()=>l.zeros_like});var e=Ht("./src/env.js"),r=Ht("./src/pipelines.js"),t=Ht("./src/models.js"),s=Ht("./src/tokenizers.js"),o=Ht("./src/configs.js"),n=Ht("./src/utils/audio.js"),a=Ht("./src/utils/image.js"),i=Ht("./src/utils/video.js"),l=Ht("./src/utils/tensor.js"),c=Ht("./src/utils/maths.js"),p=Ht("./src/base/feature_extraction_utils.js"),d=Ht("./src/models/feature_extractors.js"),u=Ht("./src/models/auto/feature_extraction_auto.js"),f=Ht("./src/base/image_processors_utils.js"),_=Ht("./src/models/image_processors.js"),b=Ht("./src/models/auto/image_processing_auto.js"),k=Ht("./src/base/processing_utils.js"),w=Ht("./src/models/processors.js"),y=Ht("./src/models/auto/processing_auto.js"),$=Ht("./src/generation/streamers.js"),T=Ht("./src/generation/stopping_criteria.js"),M=Ht("./src/generation/logits_process.js")})(),m.ASTFeatureExtractor,m.ASTForAudioClassification,m.ASTModel,m.ASTPreTrainedModel,m.AlbertForMaskedLM,m.AlbertForQuestionAnswering,m.AlbertForSequenceClassification,m.AlbertModel,m.AlbertPreTrainedModel,m.AlbertTokenizer,m.ArceeForCausalLM,m.ArceeModel,m.ArceePreTrainedModel,m.AudioClassificationPipeline,m.AutoConfig,m.AutoFeatureExtractor,m.AutoImageProcessor,m.AutoModel,m.AutoModelForAudioClassification,m.AutoModelForAudioFrameClassification,m.AutoModelForAudioTextToText,m.AutoModelForCTC,m.AutoModelForCausalLM,m.AutoModelForDepthEstimation,m.AutoModelForDocumentQuestionAnswering,m.AutoModelForImageClassification,m.AutoModelForImageFeatureExtraction,m.AutoModelForImageMatting,m.AutoModelForImageSegmentation,m.AutoModelForImageTextToText,m.AutoModelForImageToImage,m.AutoModelForMaskGeneration,m.AutoModelForMaskedLM,m.AutoModelForNormalEstimation,m.AutoModelForObjectDetection,m.AutoModelForPoseEstimation,m.AutoModelForQuestionAnswering,m.AutoModelForSemanticSegmentation,m.AutoModelForSeq2SeqLM,m.AutoModelForSequenceClassification,m.AutoModelForSpeechSeq2Seq,m.AutoModelForTextToSpectrogram,m.AutoModelForTextToWaveform,m.AutoModelForTokenClassification,m.AutoModelForUniversalSegmentation,m.AutoModelForVision2Seq,m.AutoModelForXVector,m.AutoModelForZeroShotObjectDetection,m.AutoProcessor,m.AutoTokenizer,m.AutomaticSpeechRecognitionPipeline,m.BackgroundRemovalPipeline,m.BartForConditionalGeneration,m.BartForSequenceClassification,m.BartModel,m.BartPretrainedModel,m.BartTokenizer,m.BaseModelOutput,m.BaseStreamer,m.BeitFeatureExtractor,m.BeitForImageClassification,m.BeitModel,m.BeitPreTrainedModel,m.BertForMaskedLM,m.BertForQuestionAnswering,m.BertForSequenceClassification,m.BertForTokenClassification,m.BertModel,m.BertPreTrainedModel,m.BertTokenizer,m.BitImageProcessor,m.BlenderbotForConditionalGeneration,m.BlenderbotModel,m.BlenderbotPreTrainedModel,m.BlenderbotSmallForConditionalGeneration,m.BlenderbotSmallModel,m.BlenderbotSmallPreTrainedModel,m.BlenderbotSmallTokenizer,m.BlenderbotTokenizer,m.BloomForCausalLM,m.BloomModel,m.BloomPreTrainedModel,m.BloomTokenizer,m.CLIPFeatureExtractor,m.CLIPImageProcessor,m.CLIPModel,m.CLIPPreTrainedModel,m.CLIPSegForImageSegmentation,m.CLIPSegModel,m.CLIPSegPreTrainedModel,m.CLIPTextModel,m.CLIPTextModelWithProjection,m.CLIPTokenizer,m.CLIPVisionModel,m.CLIPVisionModelWithProjection,m.CamembertForMaskedLM,m.CamembertForQuestionAnswering,m.CamembertForSequenceClassification,m.CamembertForTokenClassification,m.CamembertModel,m.CamembertPreTrainedModel,m.CamembertTokenizer,m.CausalLMOutput,m.CausalLMOutputWithPast,m.ChineseCLIPFeatureExtractor,m.ChineseCLIPModel,m.ChineseCLIPPreTrainedModel,m.ClapAudioModelWithProjection,m.ClapFeatureExtractor,m.ClapModel,m.ClapPreTrainedModel,m.ClapTextModelWithProjection,m.ClassifierFreeGuidanceLogitsProcessor,m.CodeGenForCausalLM,m.CodeGenModel,m.CodeGenPreTrainedModel,m.CodeGenTokenizer,m.CodeLlamaTokenizer,m.CohereForCausalLM,m.CohereModel,m.CoherePreTrainedModel,m.CohereTokenizer,m.ConvBertForMaskedLM,m.ConvBertForQuestionAnswering,m.ConvBertForSequenceClassification,m.ConvBertForTokenClassification,m.ConvBertModel,m.ConvBertPreTrainedModel,m.ConvBertTokenizer,m.ConvNextFeatureExtractor,m.ConvNextForImageClassification,m.ConvNextImageProcessor,m.ConvNextModel,m.ConvNextPreTrainedModel,m.ConvNextV2ForImageClassification,m.ConvNextV2Model,m.ConvNextV2PreTrainedModel,m.DFineForObjectDetection,m.DFineModel,m.DFinePreTrainedModel,m.DINOv3ConvNextModel,m.DINOv3ConvNextPreTrainedModel,m.DINOv3ViTImageProcessor,m.DINOv3ViTModel,m.DINOv3ViTPreTrainedModel,m.DPTFeatureExtractor,m.DPTForDepthEstimation,m.DPTImageProcessor,m.DPTModel,m.DPTPreTrainedModel,m.DacDecoderModel,m.DacDecoderOutput,m.DacEncoderModel,m.DacEncoderOutput,m.DacFeatureExtractor,m.DacModel,m.DacPreTrainedModel,m.DataTypeMap,m.DebertaForMaskedLM,m.DebertaForQuestionAnswering,m.DebertaForSequenceClassification,m.DebertaForTokenClassification,m.DebertaModel,m.DebertaPreTrainedModel,m.DebertaTokenizer,m.DebertaV2ForMaskedLM,m.DebertaV2ForQuestionAnswering,m.DebertaV2ForSequenceClassification,m.DebertaV2ForTokenClassification,m.DebertaV2Model,m.DebertaV2PreTrainedModel,m.DebertaV2Tokenizer,m.DecisionTransformerModel,m.DecisionTransformerPreTrainedModel,m.DeiTFeatureExtractor,m.DeiTForImageClassification,m.DeiTImageProcessor,m.DeiTModel,m.DeiTPreTrainedModel,m.DepthAnythingForDepthEstimation,m.DepthAnythingPreTrainedModel,m.DepthEstimationPipeline,m.DepthProForDepthEstimation,m.DepthProPreTrainedModel,m.DetrFeatureExtractor,m.DetrForObjectDetection,m.DetrForSegmentation,m.DetrImageProcessor,m.DetrModel,m.DetrObjectDetectionOutput,m.DetrPreTrainedModel,m.DetrSegmentationOutput,m.Dinov2ForImageClassification,m.Dinov2Model,m.Dinov2PreTrainedModel,m.Dinov2WithRegistersForImageClassification,m.Dinov2WithRegistersModel,m.Dinov2WithRegistersPreTrainedModel,m.DistilBertForMaskedLM,m.DistilBertForQuestionAnswering,m.DistilBertForSequenceClassification,m.DistilBertForTokenClassification,m.DistilBertModel,m.DistilBertPreTrainedModel,m.DistilBertTokenizer,m.DocumentQuestionAnsweringPipeline,m.DonutFeatureExtractor,m.DonutImageProcessor,m.DonutSwinModel,m.DonutSwinPreTrainedModel,m.EdgeTamModel,m.EfficientNetForImageClassification,m.EfficientNetImageProcessor,m.EfficientNetModel,m.EfficientNetPreTrainedModel,m.ElectraForMaskedLM,m.ElectraForQuestionAnswering,m.ElectraForSequenceClassification,m.ElectraForTokenClassification,m.ElectraModel,m.ElectraPreTrainedModel,m.ElectraTokenizer,m.EncodecFeatureExtractor,m.EosTokenCriteria,m.Ernie4_5ForCausalLM,m.Ernie4_5Model,m.Ernie4_5PreTrainedModel,m.EsmForMaskedLM,m.EsmForSequenceClassification,m.EsmForTokenClassification,m.EsmModel,m.EsmPreTrainedModel,m.EsmTokenizer,m.ExaoneForCausalLM,m.ExaoneModel,m.ExaonePreTrainedModel,m.FFT,m.FalconForCausalLM,m.FalconModel,m.FalconPreTrainedModel,m.FalconTokenizer,m.FastViTForImageClassification,m.FastViTModel,m.FastViTPreTrainedModel,m.FeatureExtractionPipeline,m.FeatureExtractor,m.FillMaskPipeline,m.Florence2ForConditionalGeneration,m.Florence2PreTrainedModel,m.Florence2Processor,m.ForcedBOSTokenLogitsProcessor,m.ForcedEOSTokenLogitsProcessor,m.GLPNFeatureExtractor,m.GLPNForDepthEstimation,m.GLPNModel,m.GLPNPreTrainedModel,m.GPT2LMHeadModel,m.GPT2Model,m.GPT2PreTrainedModel,m.GPT2Tokenizer,m.GPTBigCodeForCausalLM,m.GPTBigCodeModel,m.GPTBigCodePreTrainedModel,m.GPTJForCausalLM,m.GPTJModel,m.GPTJPreTrainedModel,m.GPTNeoForCausalLM,m.GPTNeoModel,m.GPTNeoPreTrainedModel,m.GPTNeoXForCausalLM,m.GPTNeoXModel,m.GPTNeoXPreTrainedModel,m.GPTNeoXTokenizer,m.Gemma2ForCausalLM,m.Gemma2Model,m.Gemma2PreTrainedModel,m.Gemma3ForCausalLM,m.Gemma3Model,m.Gemma3PreTrainedModel,m.Gemma3nAudioFeatureExtractor,m.Gemma3nForConditionalGeneration,m.Gemma3nPreTrainedModel,m.Gemma3nProcessor,m.GemmaForCausalLM,m.GemmaModel,m.GemmaPreTrainedModel,m.GemmaTokenizer,m.GlmForCausalLM,m.GlmModel,m.GlmPreTrainedModel,m.GraniteForCausalLM,m.GraniteModel,m.GraniteMoeHybridForCausalLM,m.GraniteMoeHybridModel,m.GraniteMoeHybridPreTrainedModel,m.GranitePreTrainedModel,m.Grok1Tokenizer,m.GroundingDinoForObjectDetection,m.GroundingDinoImageProcessor,m.GroundingDinoPreTrainedModel,m.GroundingDinoProcessor,m.GroupViTModel,m.GroupViTPreTrainedModel,m.HeliumForCausalLM,m.HeliumModel,m.HeliumPreTrainedModel,m.HerbertTokenizer,m.HieraForImageClassification,m.HieraModel,m.HieraPreTrainedModel,m.HubertForCTC,m.HubertForSequenceClassification,m.HubertModel,m.HubertPreTrainedModel,m.IJepaForImageClassification,m.IJepaModel,m.IJepaPreTrainedModel,m.Idefics3ForConditionalGeneration,m.Idefics3ImageProcessor,m.Idefics3PreTrainedModel,m.Idefics3Processor,m.ImageClassificationPipeline,m.ImageFeatureExtractionPipeline,m.ImageFeatureExtractor,m.ImageMattingOutput,m.ImageProcessor,m.ImageSegmentationPipeline,m.ImageToImagePipeline,m.ImageToTextPipeline,m.InterruptableStoppingCriteria,m.JAISLMHeadModel,m.JAISModel,m.JAISPreTrainedModel,m.JinaCLIPImageProcessor,m.JinaCLIPModel,m.JinaCLIPPreTrainedModel,m.JinaCLIPProcessor,m.JinaCLIPTextModel,m.JinaCLIPVisionModel,m.Lfm2ForCausalLM,m.Lfm2Model,m.Lfm2PreTrainedModel,m.LiteWhisperForConditionalGeneration,m.Llama4ForCausalLM,m.Llama4PreTrainedModel,m.LlamaForCausalLM,m.LlamaModel,m.LlamaPreTrainedModel,m.LlamaTokenizer,m.LlavaForConditionalGeneration,m.LlavaOnevisionForConditionalGeneration,m.LlavaOnevisionImageProcessor,m.LlavaPreTrainedModel,m.LlavaProcessor,m.LlavaQwen2ForCausalLM,m.LogitsProcessor,m.LogitsProcessorList,m.LogitsWarper,m.LongT5ForConditionalGeneration,m.LongT5Model,m.LongT5PreTrainedModel,m.M2M100ForConditionalGeneration,m.M2M100Model,m.M2M100PreTrainedModel,m.M2M100Tokenizer,m.MBart50Tokenizer,m.MBartForCausalLM,m.MBartForConditionalGeneration,m.MBartForSequenceClassification,m.MBartModel,m.MBartPreTrainedModel,m.MBartTokenizer,m.MPNetForMaskedLM,m.MPNetForQuestionAnswering,m.MPNetForSequenceClassification,m.MPNetForTokenClassification,m.MPNetModel,m.MPNetPreTrainedModel,m.MPNetTokenizer,m.MT5ForConditionalGeneration,m.MT5Model,m.MT5PreTrainedModel,m.MarianMTModel,m.MarianModel,m.MarianPreTrainedModel,m.MarianTokenizer,m.Mask2FormerImageProcessor,m.MaskFormerFeatureExtractor,m.MaskFormerForInstanceSegmentation,m.MaskFormerImageProcessor,m.MaskFormerModel,m.MaskFormerPreTrainedModel,m.MaskedLMOutput,m.MaxLengthCriteria,m.Metric3DForDepthEstimation,m.Metric3DPreTrainedModel,m.Metric3Dv2ForDepthEstimation,m.Metric3Dv2PreTrainedModel,m.MgpstrForSceneTextRecognition,m.MgpstrModelOutput,m.MgpstrPreTrainedModel,m.MgpstrProcessor,m.MgpstrTokenizer,m.MimiDecoderModel,m.MimiDecoderOutput,m.MimiEncoderModel,m.MimiEncoderOutput,m.MimiModel,m.MimiPreTrainedModel,m.MinLengthLogitsProcessor,m.MinNewTokensLengthLogitsProcessor,m.Ministral3ForCausalLM,m.Ministral3Model,m.Ministral3PreTrainedModel,m.MinistralForCausalLM,m.MinistralModel,m.MinistralPreTrainedModel,m.Mistral3ForConditionalGeneration,m.MistralForCausalLM,m.MistralModel,m.MistralPreTrainedModel,m.MobileBertForMaskedLM,m.MobileBertForQuestionAnswering,m.MobileBertForSequenceClassification,m.MobileBertModel,m.MobileBertPreTrainedModel,m.MobileBertTokenizer,m.MobileLLMForCausalLM,m.MobileLLMModel,m.MobileLLMPreTrainedModel,m.MobileNetV1FeatureExtractor,m.MobileNetV1ForImageClassification,m.MobileNetV1ForSemanticSegmentation,m.MobileNetV1ImageProcessor,m.MobileNetV1Model,m.MobileNetV1PreTrainedModel,m.MobileNetV2FeatureExtractor,m.MobileNetV2ForImageClassification,m.MobileNetV2ForSemanticSegmentation,m.MobileNetV2ImageProcessor,m.MobileNetV2Model,m.MobileNetV2PreTrainedModel,m.MobileNetV3FeatureExtractor,m.MobileNetV3ForImageClassification,m.MobileNetV3ForSemanticSegmentation,m.MobileNetV3ImageProcessor,m.MobileNetV3Model,m.MobileNetV3PreTrainedModel,m.MobileNetV4FeatureExtractor,m.MobileNetV4ForImageClassification,m.MobileNetV4ForSemanticSegmentation,m.MobileNetV4ImageProcessor,m.MobileNetV4Model,m.MobileNetV4PreTrainedModel,m.MobileViTFeatureExtractor,m.MobileViTForImageClassification,m.MobileViTImageProcessor,m.MobileViTModel,m.MobileViTPreTrainedModel,m.MobileViTV2ForImageClassification,m.MobileViTV2Model,m.MobileViTV2PreTrainedModel,m.ModelOutput,m.ModernBertDecoderForCausalLM,m.ModernBertDecoderModel,m.ModernBertDecoderPreTrainedModel,m.ModernBertForMaskedLM,m.ModernBertForSequenceClassification,m.ModernBertForTokenClassification,m.ModernBertModel,m.ModernBertPreTrainedModel,m.Moondream1ForConditionalGeneration,m.MoonshineFeatureExtractor,m.MoonshineForConditionalGeneration,m.MoonshineModel,m.MoonshinePreTrainedModel,m.MoonshineProcessor,m.MptForCausalLM,m.MptModel,m.MptPreTrainedModel,m.MultiModalityCausalLM,m.MultiModalityPreTrainedModel,m.MusicgenForCausalLM,m.MusicgenForConditionalGeneration,m.MusicgenModel,m.MusicgenPreTrainedModel,m.NanoChatForCausalLM,m.NanoChatModel,m.NanoChatPreTrainedModel,m.NeoBertForMaskedLM,m.NeoBertForQuestionAnswering,m.NeoBertForSequenceClassification,m.NeoBertForTokenClassification,m.NeoBertModel,m.NeoBertPreTrainedModel,m.NllbTokenizer,m.NoBadWordsLogitsProcessor,m.NoRepeatNGramLogitsProcessor,m.NomicBertModel,m.NomicBertPreTrainedModel,m.NougatImageProcessor,m.NougatTokenizer,m.OPTForCausalLM,m.OPTModel,m.OPTPreTrainedModel,m.ObjectDetectionPipeline,m.Olmo2ForCausalLM,m.Olmo2Model,m.Olmo2PreTrainedModel,m.OlmoForCausalLM,m.OlmoModel,m.OlmoPreTrainedModel,m.OpenELMForCausalLM,m.OpenELMModel,m.OpenELMPreTrainedModel,m.OwlViTFeatureExtractor,m.OwlViTForObjectDetection,m.OwlViTImageProcessor,m.OwlViTModel,m.OwlViTPreTrainedModel,m.OwlViTProcessor,m.Owlv2ForObjectDetection,m.Owlv2ImageProcessor,m.Owlv2Model,m.Owlv2PreTrainedModel,m.PaliGemmaForConditionalGeneration,m.PaliGemmaPreTrainedModel,m.PaliGemmaProcessor,m.ParakeetFeatureExtractor,m.ParakeetForCTC,m.ParakeetPreTrainedModel,m.PatchTSMixerForPrediction,m.PatchTSMixerModel,m.PatchTSMixerPreTrainedModel,m.PatchTSTForPrediction,m.PatchTSTModel,m.PatchTSTPreTrainedModel,m.Phi3ForCausalLM,m.Phi3Model,m.Phi3PreTrainedModel,m.Phi3VForCausalLM,m.Phi3VImageProcessor,m.Phi3VPreTrainedModel,m.Phi3VProcessor,m.PhiForCausalLM,m.PhiModel,m.PhiPreTrainedModel,m.Pipeline,m.PixtralImageProcessor,m.PixtralProcessor,m.PreTrainedModel,m.PreTrainedTokenizer,m.PretrainedConfig,m.PretrainedMixin,m.Processor,m.PvtForImageClassification,m.PvtImageProcessor,m.PvtModel,m.PvtPreTrainedModel,m.PyAnnoteFeatureExtractor,m.PyAnnoteForAudioFrameClassification,m.PyAnnoteModel,m.PyAnnotePreTrainedModel,m.PyAnnoteProcessor,m.QuestionAnsweringModelOutput,m.QuestionAnsweringPipeline,m.Qwen2ForCausalLM,m.Qwen2Model,m.Qwen2PreTrainedModel,m.Qwen2Tokenizer,m.Qwen2VLForConditionalGeneration,m.Qwen2VLImageProcessor,m.Qwen2VLPreTrainedModel,m.Qwen2VLProcessor,m.Qwen3ForCausalLM,m.Qwen3Model,m.Qwen3PreTrainedModel,m.RFDetrForObjectDetection,m.RFDetrModel,m.RFDetrObjectDetectionOutput,m.RFDetrPreTrainedModel,m.RTDetrForObjectDetection,m.RTDetrImageProcessor,m.RTDetrModel,m.RTDetrObjectDetectionOutput,m.RTDetrPreTrainedModel,m.RTDetrV2ForObjectDetection,m.RTDetrV2Model,m.RTDetrV2ObjectDetectionOutput,m.RTDetrV2PreTrainedModel,m.RawAudio,m.RawImage,m.RawVideo,m.RawVideoFrame,m.RepetitionPenaltyLogitsProcessor,m.ResNetForImageClassification,m.ResNetModel,m.ResNetPreTrainedModel,m.RoFormerForMaskedLM,m.RoFormerForQuestionAnswering,m.RoFormerForSequenceClassification,m.RoFormerForTokenClassification,m.RoFormerModel,m.RoFormerPreTrainedModel,m.RoFormerTokenizer,m.RobertaForMaskedLM,m.RobertaForQuestionAnswering,m.RobertaForSequenceClassification,m.RobertaForTokenClassification,m.RobertaModel,m.RobertaPreTrainedModel,m.RobertaTokenizer,m.Sam2ImageProcessor,m.Sam2ImageSegmentationOutput,m.Sam2Model,m.Sam2PreTrainedModel,m.Sam2Processor,m.Sam2VideoProcessor,m.Sam3ImageProcessor,m.Sam3TrackerModel,m.SamImageProcessor,m.SamImageSegmentationOutput,m.SamModel,m.SamPreTrainedModel,m.SamProcessor,m.SapiensForDepthEstimation,m.SapiensForNormalEstimation,m.SapiensForSemanticSegmentation,m.SapiensPreTrainedModel,m.SeamlessM4TFeatureExtractor,m.SegformerFeatureExtractor,m.SegformerForImageClassification,m.SegformerForSemanticSegmentation,m.SegformerImageProcessor,m.SegformerModel,m.SegformerPreTrainedModel,m.Seq2SeqLMOutput,m.SequenceClassifierOutput,m.SiglipImageProcessor,m.SiglipModel,m.SiglipPreTrainedModel,m.SiglipTextModel,m.SiglipTokenizer,m.SiglipVisionModel,m.SmolLM3ForCausalLM,m.SmolLM3Model,m.SmolLM3PreTrainedModel,m.SmolVLMForConditionalGeneration,m.SmolVLMImageProcessor,m.SmolVLMProcessor,m.SnacDecoderModel,m.SnacEncoderModel,m.SnacFeatureExtractor,m.SnacModel,m.SnacPreTrainedModel,m.SpeechT5FeatureExtractor,m.SpeechT5ForSpeechToText,m.SpeechT5ForTextToSpeech,m.SpeechT5HifiGan,m.SpeechT5Model,m.SpeechT5PreTrainedModel,m.SpeechT5Processor,m.SpeechT5Tokenizer,m.SqueezeBertForMaskedLM,m.SqueezeBertForQuestionAnswering,m.SqueezeBertForSequenceClassification,m.SqueezeBertModel,m.SqueezeBertPreTrainedModel,m.SqueezeBertTokenizer,m.StableLmForCausalLM,m.StableLmModel,m.StableLmPreTrainedModel,m.Starcoder2ForCausalLM,m.Starcoder2Model,m.Starcoder2PreTrainedModel,m.StoppingCriteria,m.StoppingCriteriaList,m.StyleTextToSpeech2Model,m.StyleTextToSpeech2PreTrainedModel,m.SummarizationPipeline,m.SupertonicForConditionalGeneration,m.SupertonicPreTrainedModel,m.SuppressTokensAtBeginLogitsProcessor,m.Swin2SRForImageSuperResolution,m.Swin2SRImageProcessor,m.Swin2SRModel,m.Swin2SRPreTrainedModel,m.SwinForImageClassification,m.SwinForSemanticSegmentation,m.SwinModel,m.SwinPreTrainedModel,m.T5ForConditionalGeneration,m.T5Model,m.T5PreTrainedModel,m.T5Tokenizer,m.TableTransformerForObjectDetection,m.TableTransformerModel,m.TableTransformerObjectDetectionOutput,m.TableTransformerPreTrainedModel,m.TemperatureLogitsWarper,m.Tensor,m.Text2TextGenerationPipeline,m.TextClassificationPipeline,m.TextGenerationPipeline,m.TextStreamer,m.TextToAudioPipeline,m.TokenClassificationPipeline,m.TokenClassifierOutput,m.TokenizerModel,m.TopKLogitsWarper,m.TopPLogitsWarper,m.TrOCRForCausalLM,m.TrOCRPreTrainedModel,m.TranslationPipeline,m.UltravoxModel,m.UltravoxPreTrainedModel,m.UltravoxProcessor,m.UniSpeechForCTC,m.UniSpeechForSequenceClassification,m.UniSpeechModel,m.UniSpeechPreTrainedModel,m.UniSpeechSatForAudioFrameClassification,m.UniSpeechSatForCTC,m.UniSpeechSatForSequenceClassification,m.UniSpeechSatModel,m.UniSpeechSatPreTrainedModel,m.VLChatProcessor,m.VLMImageProcessor,m.VaultGemmaForCausalLM,m.VaultGemmaModel,m.VaultGemmaPreTrainedModel,m.ViTFeatureExtractor,m.ViTForImageClassification,m.ViTImageProcessor,m.ViTMAEModel,m.ViTMAEPreTrainedModel,m.ViTMSNForImageClassification,m.ViTMSNModel,m.ViTMSNPreTrainedModel,m.ViTModel,m.ViTPreTrainedModel,m.VisionEncoderDecoderModel,m.VitMatteForImageMatting,m.VitMatteImageProcessor,m.VitMattePreTrainedModel,m.VitPoseForPoseEstimation,m.VitPoseImageProcessor,m.VitPosePreTrainedModel,m.VitsModel,m.VitsModelOutput,m.VitsPreTrainedModel,m.VitsTokenizer,m.VoxtralForConditionalGeneration,m.VoxtralProcessor,m.Wav2Vec2BertForCTC,m.Wav2Vec2BertForSequenceClassification,m.Wav2Vec2BertModel,m.Wav2Vec2BertPreTrainedModel,m.Wav2Vec2CTCTokenizer,m.Wav2Vec2FeatureExtractor,m.Wav2Vec2ForAudioFrameClassification,m.Wav2Vec2ForCTC,m.Wav2Vec2ForSequenceClassification,m.Wav2Vec2Model,m.Wav2Vec2PreTrainedModel,m.Wav2Vec2Processor,m.Wav2Vec2ProcessorWithLM,m.WavLMForAudioFrameClassification,m.WavLMForCTC,m.WavLMForSequenceClassification,m.WavLMForXVector,m.WavLMModel,m.WavLMPreTrainedModel,m.WeSpeakerFeatureExtractor,m.WeSpeakerResNetModel,m.WeSpeakerResNetPreTrainedModel,m.WhisperFeatureExtractor,m.WhisperForConditionalGeneration,m.WhisperModel,m.WhisperPreTrainedModel,m.WhisperProcessor,m.WhisperTextStreamer,m.WhisperTimeStampLogitsProcessor,m.WhisperTokenizer,m.XLMForQuestionAnswering,m.XLMForSequenceClassification,m.XLMForTokenClassification,m.XLMModel,m.XLMPreTrainedModel,m.XLMRobertaForMaskedLM,m.XLMRobertaForQuestionAnswering,m.XLMRobertaForSequenceClassification,m.XLMRobertaForTokenClassification,m.XLMRobertaModel,m.XLMRobertaPreTrainedModel,m.XLMRobertaTokenizer,m.XLMTokenizer,m.XLMWithLMHeadModel,m.XVectorOutput,m.YolosFeatureExtractor,m.YolosForObjectDetection,m.YolosImageProcessor,m.YolosModel,m.YolosObjectDetectionOutput,m.YolosPreTrainedModel,m.ZeroShotAudioClassificationPipeline,m.ZeroShotClassificationPipeline,m.ZeroShotImageClassificationPipeline,m.ZeroShotObjectDetectionPipeline,m.bankers_round,m.cat,m.cos_sim,m.dot,m.dynamic_time_warping,m.env,m.full,m.full_like,m.getCacheShapes,m.hamming,m.hanning,m.interpolate,m.interpolate_4d,m.interpolate_data,m.is_chinese_char,m.layer_norm,m.load_image,m.load_video,m.log_softmax,m.magnitude,m.matmul,m.max,m.mean,m.mean_pooling,m.medianFilter,m.mel_filter_bank,m.min,m.ones,m.ones_like,m.permute,m.permute_data;var FT=m.pipeline;m.quantize_embeddings,m.rand,m.randn,m.read_audio,m.rfft,m.round,m.slice,m.softmax,m.spectrogram,m.stack,m.std_mean,m.topk,m.window_function,m.zeros,m.zeros_like;const OT="onnx-community/depth-anything-v2-small-ONNX",DT=FT,uw=self;let xs=null,So=null;function _c(e,r=[]){uw.postMessage(e,r)}function Qn(e,r){_c({type:"status",id:e,message:r})}async function fc(e,r){return Qn(r,"AI 모델 준비 중…"),DT("depth-estimation",OT,{device:e,progress_callback:t=>{t.status==="progress"?Qn(r,`AI 모델 다운로드 중 · ${Math.round(t.progress??0)}%`):(t.status==="initiate"||t.status==="download")&&Qn(r,"AI 모델 다운로드 중…")}})}async function LT(e){if(xs)return xs;if("gpu"in navigator)try{return xs=await fc("webgpu",e),So="webgpu",xs}catch{xs&&await xs.dispose(),xs=null,Qn(e,"AI 모델을 준비하는 중입니다…")}return xs=await fc("wasm",e),So="wasm",xs}function zT(e){const{predicted_depth:r}=e,[t,s]=r.dims.slice(-2),o=r.data;if(!s||!t||o.length!==s*t)throw new Error("모델이 유효한 depth 크기를 반환하지 않았습니다.");let n=1/0,a=-1/0;for(let c=0;c<o.length;c+=1){const p=Number(o[c]);Number.isFinite(p)&&(n=Math.min(n,p),a=Math.max(a,p))}if(!Number.isFinite(n)||!Number.isFinite(a))throw new Error("모델의 depth 결과가 비어 있습니다.");const i=a-n,l=new Uint8Array(s*t);if(i<=1e-8)l.fill(128);else for(let c=0;c<l.length;c+=1){const p=Number(o[c]);l[c]=Number.isFinite(p)?Math.round((p-n)/i*255):128}return{values:l,width:s,height:t}}async function BT({id:e,image:r}){try{const t=new OffscreenCanvas(r.width,r.height),s=t.getContext("2d");if(!s)throw new Error("사진을 처리할 수 없습니다.");s.drawImage(r,0,0),r.close();let o=await LT(e);Qn(e,"AI 깊이를 계산하는 중입니다…");let n;try{const a=await o(t);if(Array.isArray(a))throw new Error("단일 사진 결과를 읽을 수 없습니다.");n=a}catch(a){if(So!=="webgpu")throw a;await o.dispose(),xs=null,So=null,Qn(e,"AI 추론을 다시 시도하는 중입니다…"),o=await fc("wasm",e),xs=o,So="wasm";const i=await o(t);if(Array.isArray(i))throw new Error("단일 사진 결과를 읽을 수 없습니다.");n=i}try{const a=zT(n),i=a.values.buffer;_c({type:"result",id:e,width:a.width,height:a.height,values:i},[i])}finally{n.predicted_depth.dispose()}}catch(t){_c({type:"error",id:e,message:t instanceof Error?t.message:"AI depth 생성에 실패했습니다."})}}uw.addEventListener("message",e=>{e.data.type==="infer"&&BT(e.data)})})();