openwakeword-js 0.1.21 → 0.1.27

package/openwakeword.mjs

@@ -1,39 +1,39 @@
-var Ag=Object.defineProperty;var St=(e=>typeof require<"u"?require:typeof Proxy<"u"?new Proxy(e,{get:(t,r)=>(typeof require<"u"?require:t)[r]}):e)(function(e){if(typeof require<"u")return require.apply(this,arguments);throw Error('Dynamic require of "'+e+'" is not supported')});var Og=(e,t)=>{for(var r in t)Ag(e,r,{get:t[r],enumerable:!0})};var on={};Og(on,{InferenceSession:()=>_t,TRACE:()=>cr,TRACE_EVENT_BEGIN:()=>mt,TRACE_EVENT_END:()=>gt,TRACE_FUNC_BEGIN:()=>Ze,TRACE_FUNC_END:()=>Ue,Tensor:()=>Ie,default:()=>Y0,env:()=>ge,registerBackend:()=>Ot});var Oa=Object.defineProperty,Bg=Object.getOwnPropertyDescriptor,Rg=Object.getOwnPropertyNames,Ng=Object.prototype.hasOwnProperty,Mg=(e=>typeof St<"u"?St:typeof Proxy<"u"?new Proxy(e,{get:(t,r)=>(typeof St<"u"?St:t)[r]}):e)(function(e){if(typeof St<"u")return St.apply(this,arguments);throw Error('Dynamic require of "'+e+'" is not supported')}),P=(e,t)=>()=>(e&&(t=e(e=0)),t),Gt=(e,t)=>{for(var r in t)Oa(e,r,{get:t[r],enumerable:!0})},Dg=(e,t,r,i)=>{if(t&&typeof t=="object"||typeof t=="function")for(let a of Rg(t))!Ng.call(e,a)&&a!==r&&Oa(e,a,{get:()=>t[a],enumerable:!(i=Bg(t,a))||i.enumerable});return e},pr=e=>Dg(Oa({},"__esModule",{value:!0}),e),Zt,ct,Ot,to,Nd,Md=P(()=>{"use strict";Zt=new Map,ct=[],Ot=(e,t,r)=>{if(t&&typeof t.init=="function"&&typeof t.createInferenceSessionHandler=="function"){let i=Zt.get(e);if(i===void 0)Zt.set(e,{backend:t,priority:r});else{if(i.priority>r)return;if(i.priority===r&&i.backend!==t)throw new Error(`cannot register backend "${e}" using priority ${r}`)}if(r>=0){let a=ct.indexOf(e);a!==-1&&ct.splice(a,1);for(let s=0;s<ct.length;s++)if(Zt.get(ct[s]).priority<=r){ct.splice(s,0,e);return}ct.push(e)}return}throw new TypeError("not a valid backend")},to=async e=>{let t=Zt.get(e);if(!t)return"backend not found.";if(t.initialized)return t.backend;if(t.aborted)return t.error;{let r=!!t.initPromise;try{return r||(t.initPromise=t.backend.init(e)),await t.initPromise,t.initialized=!0,t.backend}catch(i){return r||(t.error=`${i}`,t.aborted=!0),t.error}finally{delete t.initPromise}}},Nd=async e=>{let t=e.executionProviders||[],r=t.map(l=>typeof l=="string"?l:l.name),i=r.length===0?ct:r,a,s=[],n=new Set;for(let l of i){let p=await to(l);typeof p=="string"?s.push({name:l,err:p}):(a||(a=p),a===p&&n.add(l))}if(!a)throw new Error(`no available backend found. ERR: ${s.map(l=>`[${l.name}] ${l.err}`).join(", ")}`);for(let{name:l,err:p}of s)r.includes(l)&&console.warn(`removing requested execution provider "${l}" from session options because it is not available: ${p}`);let u=t.filter(l=>n.has(typeof l=="string"?l:l.name));return[a,new Proxy(e,{get:(l,p)=>p==="executionProviders"?u:Reflect.get(l,p)})]}}),Pg=P(()=>{"use strict";Md()}),Dd,Ug=P(()=>{"use strict";Dd="1.24.1"}),bi,ke,Pd=P(()=>{"use strict";Ug(),bi="warning",ke={wasm:{},webgl:{},webgpu:{},versions:{common:Dd},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}`);bi=e}},get logLevel(){return bi}},Object.defineProperty(ke,"logLevel",{enumerable:!0})}),ge,Wg=P(()=>{"use strict";Pd(),ge=ke}),Ud,Wd,qg=P(()=>{"use strict";Ud=(e,t)=>{let r=typeof document<"u"?document.createElement("canvas"):new OffscreenCanvas(1,1);r.width=e.dims[3],r.height=e.dims[2];let i=r.getContext("2d");if(i!=null){let a,s;t?.tensorLayout!==void 0&&t.tensorLayout==="NHWC"?(a=e.dims[2],s=e.dims[3]):(a=e.dims[3],s=e.dims[2]);let n=t?.format!==void 0?t.format:"RGB",u=t?.norm,l,p;u===void 0||u.mean===void 0?l=[255,255,255,255]:typeof u.mean=="number"?l=[u.mean,u.mean,u.mean,u.mean]:(l=[u.mean[0],u.mean[1],u.mean[2],0],u.mean[3]!==void 0&&(l[3]=u.mean[3])),u===void 0||u.bias===void 0?p=[0,0,0,0]:typeof u.bias=="number"?p=[u.bias,u.bias,u.bias,u.bias]:(p=[u.bias[0],u.bias[1],u.bias[2],0],u.bias[3]!==void 0&&(p[3]=u.bias[3]));let h=s*a,f=0,g=h,_=h*2,y=-1;n==="RGBA"?(f=0,g=h,_=h*2,y=h*3):n==="RGB"?(f=0,g=h,_=h*2):n==="RBG"&&(f=0,_=h,g=h*2);for(let $=0;$<s;$++)for(let S=0;S<a;S++){let v=(e.data[f++]-p[0])*l[0],w=(e.data[g++]-p[1])*l[1],E=(e.data[_++]-p[2])*l[2],I=y===-1?255:(e.data[y++]-p[3])*l[3];i.fillStyle="rgba("+v+","+w+","+E+","+I+")",i.fillRect(S,$,1,1)}if("toDataURL"in r)return r.toDataURL();throw new Error("toDataURL is not supported")}else throw new Error("Can not access image data")},Wd=(e,t)=>{let r=typeof document<"u"?document.createElement("canvas").getContext("2d"):new OffscreenCanvas(1,1).getContext("2d"),i;if(r!=null){let a,s,n;t?.tensorLayout!==void 0&&t.tensorLayout==="NHWC"?(a=e.dims[2],s=e.dims[1],n=e.dims[3]):(a=e.dims[3],s=e.dims[2],n=e.dims[1]);let u=t!==void 0&&t.format!==void 0?t.format:"RGB",l=t?.norm,p,h;l===void 0||l.mean===void 0?p=[255,255,255,255]:typeof l.mean=="number"?p=[l.mean,l.mean,l.mean,l.mean]:(p=[l.mean[0],l.mean[1],l.mean[2],255],l.mean[3]!==void 0&&(p[3]=l.mean[3])),l===void 0||l.bias===void 0?h=[0,0,0,0]:typeof l.bias=="number"?h=[l.bias,l.bias,l.bias,l.bias]:(h=[l.bias[0],l.bias[1],l.bias[2],0],l.bias[3]!==void 0&&(h[3]=l.bias[3]));let f=s*a;if(t!==void 0&&(t.format!==void 0&&n===4&&t.format!=="RGBA"||n===3&&t.format!=="RGB"&&t.format!=="BGR"))throw new Error("Tensor format doesn't match input tensor dims");let g=4,_=0,y=1,$=2,S=3,v=0,w=f,E=f*2,I=-1;u==="RGBA"?(v=0,w=f,E=f*2,I=f*3):u==="RGB"?(v=0,w=f,E=f*2):u==="RBG"&&(v=0,E=f,w=f*2),i=r.createImageData(a,s);for(let k=0;k<s*a;_+=g,y+=g,$+=g,S+=g,k++)i.data[_]=(e.data[v++]-h[0])*p[0],i.data[y]=(e.data[w++]-h[1])*p[1],i.data[$]=(e.data[E++]-h[2])*p[2],i.data[S]=I===-1?255:(e.data[I++]-h[3])*p[3]}else throw new Error("Can not access image data");return i}}),Er,qd,Ld,Vd,Gd,Fd,Lg=P(()=>{"use strict";Ba(),Er=(e,t)=>{if(e===void 0)throw new Error("Image buffer must be defined");if(t.height===void 0||t.width===void 0)throw new Error("Image height and width must be defined");if(t.tensorLayout==="NHWC")throw new Error("NHWC Tensor layout is not supported yet");let{height:r,width:i}=t,a=t.norm??{mean:255,bias:0},s,n;typeof a.mean=="number"?s=[a.mean,a.mean,a.mean,a.mean]:s=[a.mean[0],a.mean[1],a.mean[2],a.mean[3]??255],typeof a.bias=="number"?n=[a.bias,a.bias,a.bias,a.bias]:n=[a.bias[0],a.bias[1],a.bias[2],a.bias[3]??0];let u=t.format!==void 0?t.format:"RGBA",l=t.tensorFormat!==void 0&&t.tensorFormat!==void 0?t.tensorFormat:"RGB",p=r*i,h=l==="RGBA"?new Float32Array(p*4):new Float32Array(p*3),f=4,g=0,_=1,y=2,$=3,S=0,v=p,w=p*2,E=-1;u==="RGB"&&(f=3,g=0,_=1,y=2,$=-1),l==="RGBA"?E=p*3:l==="RBG"?(S=0,w=p,v=p*2):l==="BGR"&&(w=0,v=p,S=p*2);for(let I=0;I<p;I++,g+=f,y+=f,_+=f,$+=f)h[S++]=(e[g]+n[0])/s[0],h[v++]=(e[_]+n[1])/s[1],h[w++]=(e[y]+n[2])/s[2],E!==-1&&$!==-1&&(h[E++]=(e[$]+n[3])/s[3]);return l==="RGBA"?new Ne("float32",h,[1,4,r,i]):new Ne("float32",h,[1,3,r,i])},qd=async(e,t)=>{let r=typeof HTMLImageElement<"u"&&e instanceof HTMLImageElement,i=typeof ImageData<"u"&&e instanceof ImageData,a=typeof ImageBitmap<"u"&&e instanceof ImageBitmap,s=typeof e=="string",n,u=t??{},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")},p=h=>typeof HTMLCanvasElement<"u"&&h instanceof HTMLCanvasElement||h instanceof OffscreenCanvas?h.getContext("2d"):null;if(r){let h=l();h.width=e.width,h.height=e.height;let f=p(h);if(f!=null){let g=e.height,_=e.width;if(t!==void 0&&t.resizedHeight!==void 0&&t.resizedWidth!==void 0&&(g=t.resizedHeight,_=t.resizedWidth),t!==void 0){if(u=t,t.tensorFormat!==void 0)throw new Error("Image input config format must be RGBA for HTMLImageElement");u.tensorFormat="RGBA",u.height=g,u.width=_}else u.tensorFormat="RGBA",u.height=g,u.width=_;f.drawImage(e,0,0),n=f.getImageData(0,0,_,g).data}else throw new Error("Can not access image data")}else if(i){let h,f;if(t!==void 0&&t.resizedWidth!==void 0&&t.resizedHeight!==void 0?(h=t.resizedHeight,f=t.resizedWidth):(h=e.height,f=e.width),t!==void 0&&(u=t),u.format="RGBA",u.height=h,u.width=f,t!==void 0){let g=l();g.width=f,g.height=h;let _=p(g);if(_!=null)_.putImageData(e,0,0),n=_.getImageData(0,0,f,h).data;else throw new Error("Can not access image data")}else n=e.data}else if(a){if(t===void 0)throw new Error("Please provide image config with format for Imagebitmap");let h=l();h.width=e.width,h.height=e.height;let f=p(h);if(f!=null){let g=e.height,_=e.width;return f.drawImage(e,0,0,_,g),n=f.getImageData(0,0,_,g).data,u.height=g,u.width=_,Er(n,u)}else throw new Error("Can not access image data")}else{if(s)return new Promise((h,f)=>{let g=l(),_=p(g);if(!e||!_)return f();let y=new Image;y.crossOrigin="Anonymous",y.src=e,y.onload=()=>{g.width=y.width,g.height=y.height,_.drawImage(y,0,0,g.width,g.height);let $=_.getImageData(0,0,g.width,g.height);u.height=g.height,u.width=g.width,h(Er($.data,u))}});throw new Error("Input data provided is not supported - aborted tensor creation")}if(n!==void 0)return Er(n,u);throw new Error("Input data provided is not supported - aborted tensor creation")},Ld=(e,t)=>{let{width:r,height:i,download:a,dispose:s}=t,n=[1,i,r,4];return new Ne({location:"texture",type:"float32",texture:e,dims:n,download:a,dispose:s})},Vd=(e,t)=>{let{dataType:r,dims:i,download:a,dispose:s}=t;return new Ne({location:"gpu-buffer",type:r??"float32",gpuBuffer:e,dims:i,download:a,dispose:s})},Gd=(e,t)=>{let{dataType:r,dims:i,download:a,dispose:s}=t;return new Ne({location:"ml-tensor",type:r??"float32",mlTensor:e,dims:i,download:a,dispose:s})},Fd=(e,t,r)=>new Ne({location:"cpu-pinned",type:e,data:t,dims:r??[t.length]})}),zt,sr,$i,Hd,Vg=P(()=>{"use strict";zt=new Map([["float32",Float32Array],["uint8",Uint8Array],["int8",Int8Array],["uint16",Uint16Array],["int16",Int16Array],["int32",Int32Array],["bool",Uint8Array],["float64",Float64Array],["uint32",Uint32Array],["int4",Uint8Array],["uint4",Uint8Array]]),sr=new Map([[Float32Array,"float32"],[Uint8Array,"uint8"],[Int8Array,"int8"],[Uint16Array,"uint16"],[Int16Array,"int16"],[Int32Array,"int32"],[Float64Array,"float64"],[Uint32Array,"uint32"]]),$i=!1,Hd=()=>{if(!$i){$i=!0;let e=typeof BigInt64Array<"u"&&BigInt64Array.from,t=typeof BigUint64Array<"u"&&BigUint64Array.from,r=globalThis.Float16Array,i=typeof r<"u"&&r.from;e&&(zt.set("int64",BigInt64Array),sr.set(BigInt64Array,"int64")),t&&(zt.set("uint64",BigUint64Array),sr.set(BigUint64Array,"uint64")),i?(zt.set("float16",r),sr.set(r,"float16")):zt.set("float16",Uint16Array)}}}),jd,Kd,Gg=P(()=>{"use strict";Ba(),jd=e=>{let t=1;for(let r=0;r<e.length;r++){let i=e[r];if(typeof i!="number"||!Number.isSafeInteger(i))throw new TypeError(`dims[${r}] must be an integer, got: ${i}`);if(i<0)throw new RangeError(`dims[${r}] must be a non-negative integer, got: ${i}`);t*=i}return t},Kd=(e,t)=>{switch(e.location){case"cpu":return new Ne(e.type,e.data,t);case"cpu-pinned":return new Ne({location:"cpu-pinned",data:e.data,type:e.type,dims:t});case"texture":return new Ne({location:"texture",texture:e.texture,type:e.type,dims:t});case"gpu-buffer":return new Ne({location:"gpu-buffer",gpuBuffer:e.gpuBuffer,type:e.type,dims:t});case"ml-tensor":return new Ne({location:"ml-tensor",mlTensor:e.mlTensor,type:e.type,dims:t});default:throw new Error(`tensorReshape: tensor location ${e.location} is not supported`)}}}),Ne,Ba=P(()=>{"use strict";qg(),Lg(),Vg(),Gg(),Ne=class{constructor(e,t,r){Hd();let i,a;if(typeof e=="object"&&"location"in e)switch(this.dataLocation=e.location,i=e.type,a=e.dims,e.location){case"cpu-pinned":{let n=zt.get(i);if(!n)throw new TypeError(`unsupported type "${i}" to create tensor from pinned buffer`);if(!(e.data instanceof n))throw new TypeError(`buffer should be of type ${n.name}`);this.cpuData=e.data;break}case"texture":{if(i!=="float32")throw new TypeError(`unsupported type "${i}" to create tensor from texture`);this.gpuTextureData=e.texture,this.downloader=e.download,this.disposer=e.dispose;break}case"gpu-buffer":{if(i!=="float32"&&i!=="float16"&&i!=="int32"&&i!=="int64"&&i!=="uint32"&&i!=="uint8"&&i!=="bool"&&i!=="uint4"&&i!=="int4")throw new TypeError(`unsupported type "${i}" to create tensor from gpu buffer`);this.gpuBufferData=e.gpuBuffer,this.downloader=e.download,this.disposer=e.dispose;break}case"ml-tensor":{if(i!=="float32"&&i!=="float16"&&i!=="int32"&&i!=="int64"&&i!=="uint32"&&i!=="uint64"&&i!=="int8"&&i!=="uint8"&&i!=="bool"&&i!=="uint4"&&i!=="int4")throw new TypeError(`unsupported type "${i}" 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 n,u;if(typeof e=="string")if(i=e,u=r,e==="string"){if(!Array.isArray(t))throw new TypeError("A string tensor's data must be a string array.");n=t}else{let l=zt.get(e);if(l===void 0)throw new TypeError(`Unsupported tensor type: ${e}.`);if(Array.isArray(t)){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"?n=l.from(t,BigInt):n=l.from(t)}else if(t instanceof l)n=t;else if(t instanceof Uint8ClampedArray)if(e==="uint8")n=Uint8Array.from(t);else throw new TypeError("A Uint8ClampedArray tensor's data must be type of uint8");else if(e==="float16"&&t instanceof Uint16Array&&l!==Uint16Array)n=new globalThis.Float16Array(t.buffer,t.byteOffset,t.length);else throw new TypeError(`A ${i} tensor's data must be type of ${l}`)}else if(u=t,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")i="string",n=e;else if(l==="boolean")i="bool",n=Uint8Array.from(e);else throw new TypeError(`Invalid element type of data array: ${l}.`)}else if(e instanceof Uint8ClampedArray)i="uint8",n=Uint8Array.from(e);else{let l=sr.get(e.constructor);if(l===void 0)throw new TypeError(`Unsupported type for tensor data: ${e.constructor}.`);i=l,n=e}if(u===void 0)u=[n.length];else if(!Array.isArray(u))throw new TypeError("A tensor's dims must be a number array");a=u,this.cpuData=n,this.dataLocation="cpu"}let s=jd(a);if(this.cpuData&&s!==this.cpuData.length&&!((i==="uint4"||i==="int4")&&Math.ceil(s/2)===this.cpuData.length))throw new Error(`Tensor's size(${s}) does not match data length(${this.cpuData.length}).`);this.type=i,this.dims=a,this.size=s}static async fromImage(e,t){return qd(e,t)}static fromTexture(e,t){return Ld(e,t)}static fromGpuBuffer(e,t){return Vd(e,t)}static fromMLTensor(e,t){return Gd(e,t)}static fromPinnedBuffer(e,t,r){return Fd(e,t,r)}toDataURL(e){return Ud(this,e)}toImageData(e){return Wd(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 t=await this.downloader();return this.downloader=void 0,this.dataLocation="cpu",this.cpuData=t,e&&this.disposer&&(this.disposer(),this.disposer=void 0),t}finally{this.isDownloading=!1}}default:throw new Error(`cannot get data from location: ${this.dataLocation}`)}}dispose(){if(this.isDownloading)throw new Error("The current tensor is being downloaded.");this.disposer&&(this.disposer(),this.disposer=void 0),this.cpuData=void 0,this.gpuTextureData=void 0,this.gpuBufferData=void 0,this.mlTensorData=void 0,this.downloader=void 0,this.isDownloading=void 0,this.dataLocation="none"}ensureValid(){if(this.dataLocation==="none")throw new Error("The tensor is disposed.")}reshape(e){if(this.ensureValid(),this.downloader||this.disposer)throw new Error("Cannot reshape a tensor that owns GPU resource.");return Kd(this,e)}}}),Ie,Zd=P(()=>{"use strict";Ba(),Ie=Ne}),cr,vi,Ze,Ue,mt,gt,Qd=P(()=>{"use strict";Pd(),cr=(e,t)=>{(typeof ke.trace>"u"?!ke.wasm.trace:!ke.trace)||console.timeStamp(`${e}::ORT::${t}`)},vi=(e,t)=>{let r=new Error().stack?.split(/\r\n|\r|\n/g)||[],i=!1;for(let a=0;a<r.length;a++){if(i&&!r[a].includes("TRACE_FUNC")){let s=`FUNC_${e}::${r[a].trim().split(" ")[1]}`;t&&(s+=`::${t}`),cr("CPU",s);return}r[a].includes("TRACE_FUNC")&&(i=!0)}},Ze=e=>{(typeof ke.trace>"u"?!ke.wasm.trace:!ke.trace)||vi("BEGIN",e)},Ue=e=>{(typeof ke.trace>"u"?!ke.wasm.trace:!ke.trace)||vi("END",e)},mt=e=>{(typeof ke.trace>"u"?!ke.wasm.trace:!ke.trace)||console.time(`ORT::${e}`)},gt=e=>{(typeof ke.trace>"u"?!ke.wasm.trace:!ke.trace)||console.timeEnd(`ORT::${e}`)}}),Xd,Fg=P(()=>{"use strict";Md(),Zd(),Qd(),Xd=class Yd{constructor(t){this.handler=t}async run(t,r,i){Ze(),mt("InferenceSession.run");let a={},s={};if(typeof t!="object"||t===null||t instanceof Ie||Array.isArray(t))throw new TypeError("'feeds' must be an object that use input names as keys and OnnxValue as corresponding values.");let n=!0;if(typeof r=="object"){if(r===null)throw new TypeError("Unexpected argument[1]: cannot be null.");if(r instanceof Ie)throw new TypeError("'fetches' cannot be a Tensor");if(Array.isArray(r)){if(r.length===0)throw new TypeError("'fetches' cannot be an empty array.");n=!1;for(let p of r){if(typeof p!="string")throw new TypeError("'fetches' must be a string array or an object.");if(this.outputNames.indexOf(p)===-1)throw new RangeError(`'fetches' contains invalid output name: ${p}.`);a[p]=null}if(typeof i=="object"&&i!==null)s=i;else if(typeof i<"u")throw new TypeError("'options' must be an object.")}else{let p=!1,h=Object.getOwnPropertyNames(r);for(let f of this.outputNames)if(h.indexOf(f)!==-1){let g=r[f];(g===null||g instanceof Ie)&&(p=!0,n=!1,a[f]=g)}if(p){if(typeof i=="object"&&i!==null)s=i;else if(typeof i<"u")throw new TypeError("'options' must be an object.")}else s=r}}else if(typeof r<"u")throw new TypeError("Unexpected argument[1]: must be 'fetches' or 'options'.");for(let p of this.inputNames)if(typeof t[p]>"u")throw new Error(`input '${p}' is missing in 'feeds'.`);if(n)for(let p of this.outputNames)a[p]=null;let u=await this.handler.run(t,a,s),l={};for(let p in u)if(Object.hasOwnProperty.call(u,p)){let h=u[p];h instanceof Ie?l[p]=h:l[p]=new Ie(h.type,h.data,h.dims)}return gt("InferenceSession.run"),Ue(),l}async release(){return this.handler.dispose()}static async create(t,r,i,a){Ze(),mt("InferenceSession.create");let s,n={};if(typeof t=="string"){if(s=t,typeof r=="object"&&r!==null)n=r;else if(typeof r<"u")throw new TypeError("'options' must be an object.")}else if(t instanceof Uint8Array){if(s=t,typeof r=="object"&&r!==null)n=r;else if(typeof r<"u")throw new TypeError("'options' must be an object.")}else if(t instanceof ArrayBuffer||typeof SharedArrayBuffer<"u"&&t instanceof SharedArrayBuffer){let h=t,f=0,g=t.byteLength;if(typeof r=="object"&&r!==null)n=r;else if(typeof r=="number"){if(f=r,!Number.isSafeInteger(f))throw new RangeError("'byteOffset' must be an integer.");if(f<0||f>=h.byteLength)throw new RangeError(`'byteOffset' is out of range [0, ${h.byteLength}).`);if(g=t.byteLength-f,typeof i=="number"){if(g=i,!Number.isSafeInteger(g))throw new RangeError("'byteLength' must be an integer.");if(g<=0||f+g>h.byteLength)throw new RangeError(`'byteLength' is out of range (0, ${h.byteLength-f}].`);if(typeof a=="object"&&a!==null)n=a;else if(typeof a<"u")throw new TypeError("'options' must be an object.")}else if(typeof i<"u")throw new TypeError("'byteLength' must be a number.")}else if(typeof r<"u")throw new TypeError("'options' must be an object.");s=new Uint8Array(h,f,g)}else throw new TypeError("Unexpected argument[0]: must be 'path' or 'buffer'.");let[u,l]=await Nd(n),p=await u.createInferenceSessionHandler(s,l);return gt("InferenceSession.create"),Ue(),new Yd(p)}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}}}),_t,Hg=P(()=>{"use strict";Fg(),_t=Xd}),jg=P(()=>{"use strict"}),Kg=P(()=>{"use strict"}),Zg=P(()=>{"use strict"}),Qg=P(()=>{"use strict"}),Jd={};Gt(Jd,{InferenceSession:()=>_t,TRACE:()=>cr,TRACE_EVENT_BEGIN:()=>mt,TRACE_EVENT_END:()=>gt,TRACE_FUNC_BEGIN:()=>Ze,TRACE_FUNC_END:()=>Ue,Tensor:()=>Ie,env:()=>ge,registerBackend:()=>Ot});var We=P(()=>{"use strict";Pg(),Wg(),Hg(),Zd(),jg(),Kg(),Qd(),Zg(),Qg()}),Ra=P(()=>{"use strict"}),ep={};Gt(ep,{default:()=>tp});var xi,Si,tp,Xg=P(()=>{"use strict";of(),Mt(),Na(),xi="ort-wasm-proxy-worker",Si=globalThis.self?.name===xi,Si&&(self.onmessage=e=>{let{type:t,in:r}=e.data;try{switch(t){case"init-wasm":Ma(r.wasm).then(()=>{Ja(r).then(()=>{postMessage({type:t})},i=>{postMessage({type:t,err:i})})},i=>{postMessage({type:t,err:i})});break;case"init-ep":{let{epName:i,env:a}=r;en(a,i).then(()=>{postMessage({type:t})},s=>{postMessage({type:t,err:s})});break}case"copy-from":{let{buffer:i}=r,a=Kr(i);postMessage({type:t,out:a});break}case"create":{let{model:i,options:a}=r;tn(i,a).then(s=>{postMessage({type:t,out:s})},s=>{postMessage({type:t,err:s})});break}case"release":rn(r),postMessage({type:t});break;case"run":{let{sessionId:i,inputIndices:a,inputs:s,outputIndices:n,options:u}=r;an(i,a,s,n,new Array(n.length).fill(null),u).then(l=>{l.some(p=>p[3]!=="cpu")?postMessage({type:t,err:"Proxy does not support non-cpu tensor location."}):postMessage({type:t,out:l},sn([...s,...l]))},l=>{postMessage({type:t,err:l})});break}case"end-profiling":nn(r),postMessage({type:t});break;default:}}catch(i){postMessage({type:t,err:i})}}),tp=Si?null:e=>new Worker(e??Re,{type:"module",name:xi})}),rp={};Gt(rp,{default:()=>ip});async function ro(e={}){var t=e,r=!!globalThis.window,i=!!globalThis.WorkerGlobalScope,a=i&&self.name?.startsWith("em-pthread");t.mountExternalData=(o,d)=>{o.startsWith("./")&&(o=o.substring(2)),(t.Zc||(t.Zc=new Map)).set(o,d)},t.unmountExternalData=()=>{delete t.Zc},globalThis.SharedArrayBuffer??new WebAssembly.Memory({initial:0,maximum:0,ae:!0}).buffer.constructor;let s=o=>async(...d)=>{try{if(t.$c)throw Error("Session already started");let m=t.$c={Nd:d[0],errors:[]},c=await o(...d);if(t.$c!==m)throw Error("Session mismatch");t.gd?.flush();let b=m.errors;if(0<b.length){let T=await Promise.all(b);if(T=T.filter(z=>z),0<T.length)throw Error(T.join(`
-`))}return c}finally{t.$c=null}};t.jsepInit=(o,d)=>{if(o==="webgpu"){[t.gd,t.Dd,t.Hd,t.jd,t.Gd,t.ac,t.Id,t.Kd,t.Ed,t.Fd,t.Jd]=d;let m=t.gd;t.jsepRegisterBuffer=(c,b,T,z)=>m.registerBuffer(c,b,T,z),t.jsepGetBuffer=c=>m.getBuffer(c),t.jsepCreateDownloader=(c,b,T)=>m.createDownloader(c,b,T),t.jsepOnCreateSession=c=>{m.onCreateSession(c)},t.jsepOnReleaseSession=c=>{m.onReleaseSession(c)},t.jsepOnRunStart=c=>m.onRunStart(c),t.Ld=(c,b)=>{m.upload(c,b)}}else if(o==="webnn"){let m=d[0];[t.Zd,t.vd,t.webnnEnsureTensor,t.xd,t.webnnDownloadTensor,t.Yd,t.webnnEnableTraceEvent]=d.slice(1),t.webnnReleaseTensorId=t.vd,t.webnnUploadTensor=t.xd,t.webnnRegisterMLContext=t.Yd,t.webnnOnRunStart=c=>m.onRunStart(c),t.webnnOnRunEnd=m.onRunEnd.bind(m),t.webnnOnReleaseSession=c=>{m.onReleaseSession(c)},t.webnnCreateMLTensorDownloader=(c,b)=>m.createMLTensorDownloader(c,b),t.webnnRegisterMLTensor=(c,b,T,z)=>m.registerMLTensor(c,b,T,z),t.webnnCreateMLContext=c=>m.createMLContext(c),t.webnnRegisterMLConstant=(c,b,T,z,R,W)=>m.registerMLConstant(c,b,T,z,R,t.Zc,W),t.webnnRegisterGraphInput=m.registerGraphInput.bind(m),t.webnnIsGraphInput=m.isGraphInput.bind(m),t.webnnRegisterGraphOutput=m.registerGraphOutput.bind(m),t.webnnIsGraphOutput=m.isGraphOutput.bind(m),t.webnnCreateTemporaryTensor=m.createTemporaryTensor.bind(m),t.webnnIsGraphInputOutputTypeSupported=m.isGraphInputOutputTypeSupported.bind(m)}};let n=()=>{let o=d=>(...m)=>{let c=Xe;return m=d(...m),Xe!=c?new Promise((b,T)=>{oi={resolve:b,reject:T}}):m};(()=>{for(let d of["_OrtAppendExecutionProvider","_OrtCreateSession","_OrtRun","_OrtRunWithBinding","_OrtBindInput"])t[d]=o(t[d])})(),s!==void 0&&(t._OrtRun=s(t._OrtRun),t._OrtRunWithBinding=s(t._OrtRunWithBinding)),n=void 0};t.asyncInit=()=>{n?.()};var u,l,p=(o,d)=>{throw d},h=import.meta.url,f="";if(r||i){try{f=new URL(".",h).href}catch{}i&&(l=o=>{var d=new XMLHttpRequest;return d.open("GET",o,!1),d.responseType="arraybuffer",d.send(null),new Uint8Array(d.response)}),u=async o=>{if(O(o))return new Promise((m,c)=>{var b=new XMLHttpRequest;b.open("GET",o,!0),b.responseType="arraybuffer",b.onload=()=>{b.status==200||b.status==0&&b.response?m(b.response):c(b.status)},b.onerror=c,b.send(null)});var d=await fetch(o,{credentials:"same-origin"});if(d.ok)return d.arrayBuffer();throw Error(d.status+" : "+d.url)}}var g,_,y,$,S,v,w=console.log.bind(console),E=console.error.bind(console),I=w,k=E,C=!1,O=o=>o.startsWith("file://");function x(){ut.buffer!=Q.buffer&&J()}if(a){let o=function(d){try{var m=d.data,c=m.Uc;if(c==="load"){let b=[];self.onmessage=T=>b.push(T),v=()=>{postMessage({Uc:"loaded"});for(let T of b)o(T);self.onmessage=o};for(let T of m.Ad)t[T]&&!t[T].proxy||(t[T]=(...z)=>{postMessage({Uc:"callHandler",zd:T,args:z})},T=="print"&&(I=t[T]),T=="printErr"&&(k=t[T]));ut=m.Vd,J(),_=m.Wd,qe(),Ir()}else if(c==="run"){(function(b){var T=(x(),U)[b+52>>>2>>>0];b=(x(),U)[b+56>>>2>>>0],ds(T,T-b),ae(T)})(m.Tc),ci(m.Tc,0,0,1,0,0),pn(),ai(m.Tc),q||(as(),q=!0);try{bf(m.Pd,m.dd)}catch(b){if(b!="unwind")throw b}}else m.target!=="setimmediate"&&(c==="checkMailbox"?q&&wr():c&&(k(`worker: received unknown command ${c}`),k(m)))}catch(b){throw ns(),b}};var D=o,q=!1;self.onunhandledrejection=d=>{throw d.reason||d},self.onmessage=o}var Q,L,K,se,A,U,ee,X,H,le,M,V=!1;function J(){var o=ut.buffer;t.HEAP8=Q=new Int8Array(o),K=new Int16Array(o),t.HEAPU8=L=new Uint8Array(o),se=new Uint16Array(o),t.HEAP32=A=new Int32Array(o),t.HEAPU32=U=new Uint32Array(o),ee=new Float32Array(o),X=new Float64Array(o),H=new BigInt64Array(o),le=new BigUint64Array(o)}function de(){V=!0,a?v():rt.tb()}function Te(o){throw k(o="Aborted("+o+")"),C=!0,o=new WebAssembly.RuntimeError(o+". Build with -sASSERTIONS for more info."),S?.(o),o}function Be(){return{a:{ma:Fm,hb:Gm,g:$f,J:vf,f:xf,o:Sf,h:Tf,ha:If,b:Ef,T:kf,Ia:yn,n:zf,_:$n,Ya:vn,Ea:xn,Ga:Sn,Za:Tn,Wa:In,Pa:En,Va:kn,ka:zn,Fa:Cn,Ca:An,Xa:On,Da:Bn,cb:Cf,ea:Of,xa:Bf,va:Nf,da:Df,O:Pf,H:Uf,wa:Wf,Z:jf,ya:Kf,Sa:Zf,Aa:Xf,Ja:Yf,ta:Jf,fa:em,Ra:ai,$a:tm,R:nm,s:dm,c:ri,ib:pm,y:cm,M:hm,D:fm,m:mm,t:qn,jb:gm,I:ym,S:_m,j:wm,v:bm,r:$m,l:vm,Ma:xm,Na:Sm,Oa:Tm,Ka:Fn,La:Hn,ua:jn,eb:Em,bb:Cm,u:Am,aa:Om,ga:Bm,ab:km,V:Rm,_a:Nm,Ba:Mm,F:Im,U:Dm,la:Sr,za:Um,gb:Pm,fb:Wm,Ta:Xn,Ua:Yn,Ha:Xr,$:Jn,ja:es,Qa:ts,ia:rs,lb:kg,na:$g,mb:Eg,oa:bg,G:pg,d:Zm,q:jm,w:Hm,B:ng,pb:yg,K:ug,x:Xm,pa:_g,X:vg,ba:gg,nb:Ig,ob:Tg,ra:cg,qa:mg,qb:hg,N:lg,Y:wg,e:Qm,A:Ym,k:Km,kb:zg,p:eg,z:tg,C:Jm,E:rg,L:sg,rb:dg,Q:xg,ca:og,W:Sg,sb:ag,sa:ig,P:fg,i:Lm,a:ut,db:Qr}}}async function qe(){function o(c,b){var T=rt=c.exports;c={};for(let[z,R]of Object.entries(T))typeof R=="function"?(T=rm(R),c[z]=T):c[z]=R;return rt=c,rt=(function(){var z=rt,R=G=>ie=>G(ie)>>>0,W=G=>()=>G()>>>0;return(z=Object.assign({},z)).ub=R(z.ub),z.Yb=W(z.Yb),z._b=R(z._b),z.mc=R(z.mc),z.nc=W(z.nc),z.rc=R(z.rc),z})(),ln.push(rt.$b),is=(c=rt).ub,as=c.vb,t._OrtInit=c.wb,t._OrtGetLastError=c.xb,t._OrtCreateSessionOptions=c.yb,t._OrtAppendExecutionProvider=c.zb,t._OrtAddFreeDimensionOverride=c.Ab,t._OrtAddSessionConfigEntry=c.Bb,t._OrtReleaseSessionOptions=c.Cb,t._OrtCreateSession=c.Db,t._OrtReleaseSession=c.Eb,t._OrtGetInputOutputCount=c.Fb,t._OrtGetInputOutputMetadata=c.Gb,t._OrtFree=c.Hb,t._OrtCreateTensor=c.Ib,t._OrtGetTensorData=c.Jb,t._OrtReleaseTensor=c.Kb,t._OrtCreateRunOptions=c.Lb,t._OrtAddRunConfigEntry=c.Mb,t._OrtReleaseRunOptions=c.Nb,t._OrtCreateBinding=c.Ob,t._OrtBindInput=c.Pb,t._OrtBindOutput=c.Qb,t._OrtClearBoundOutputs=c.Rb,t._OrtReleaseBinding=c.Sb,t._OrtRunWithBinding=c.Tb,t._OrtRun=c.Ub,t._OrtEndProfiling=c.Vb,t._JsepOutput=c.Wb,t._JsepGetNodeName=c.Xb,Tr=c.Yb,Ye=t._free=c.Zb,jt=t._malloc=c._b,ci=c.bc,ns=c.cc,ss=c.dc,os=c.ec,hi=c.fc,us=c.gc,ls=c.hc,oe=c.ic,Kt=c.jc,ds=c.kc,ae=c.lc,fi=c.mc,ne=c.nc,ps=c.oc,mi=c.pc,cs=c.qc,hs=c.rc,fs=c.sc,gi=c.tc,ms=c.uc,gs=c.vc,ys=c.wc,_s=c.xc,ws=c.yc,bs=c.zc,$s=c.Ac,vs=c.Bc,xs=c.Cc,Ss=c.Dc,Ts=c.Ec,Is=c.Fc,Es=c.Gc,ks=c.Hc,zs=c.Ic,Cs=c.Jc,As=c.Kc,Os=c.Lc,Bs=c.Mc,Rs=c.Nc,Ns=c.Oc,Ms=c.Pc,Ds=c.Rc,Ps=c.Sc,Us=c.bd,Ws=c.cd,qs=c.hd,Ls=c.kd,Vs=c.ld,Gs=c.md,Fs=c.nd,Hs=c.od,js=c.pd,Ks=c.qd,Zs=c.rd,Qs=c.wd,Xs=c.Rd,Ys=c.Sd,Js=c.Td,eo=c.Ud,_=b,rt}var d,m=Be();return t.instantiateWasm?new Promise(c=>{t.instantiateWasm(m,(b,T)=>{c(o(b,T))})}):a?o(new WebAssembly.Instance(_,Be()),_):(M??=t.locateFile?t.locateFile?t.locateFile("ort-wasm-simd-threaded.jsep.wasm",f):f+"ort-wasm-simd-threaded.jsep.wasm":new URL("ort-wasm-simd-threaded.jsep.wasm",import.meta.url).href,d=await(async function(c){var b=M;if(!g&&!O(b))try{var T=fetch(b,{credentials:"same-origin"});return await WebAssembly.instantiateStreaming(T,c)}catch(z){k(`wasm streaming compile failed: ${z}`),k("falling back to ArrayBuffer instantiation")}return(async function(z,R){try{var W=await(async function(G){if(!g)try{var ie=await u(G);return new Uint8Array(ie)}catch{}if(G==M&&g)G=new Uint8Array(g);else{if(!l)throw"both async and sync fetching of the wasm failed";G=l(G)}return G})(z);return await WebAssembly.instantiate(W,R)}catch(G){k(`failed to asynchronously prepare wasm: ${G}`),Te(G)}})(b,c)})(m),o(d.instance,d.module))}class et{name="ExitStatus";constructor(d){this.message=`Program terminated with exit(${d})`,this.status=d}}var xe=o=>{o.terminate(),o.onmessage=()=>{}},be=[],Oe=0,bt=null,fr=o=>{ot.length==0&&(hn(),cn(ot[0]));var d=ot.pop();if(!d)return 6;Ft.push(d),$t[o.Tc]=d,d.Tc=o.Tc;var m={Uc:"run",Pd:o.Od,dd:o.dd,Tc:o.Tc};return d.postMessage(m,o.ud),0},st=0,we=(o,d,...m)=>{var c,b=16*m.length,T=ne(),z=fi(b),R=z>>>3;for(c of m)typeof c=="bigint"?((x(),H)[R++>>>0]=1n,(x(),H)[R++>>>0]=c):((x(),H)[R++>>>0]=0n,(x(),X)[R++>>>0]=c);return o=ss(o,0,b,z,d),ae(T),o};function Qr(o){if(a)return we(0,1,o);if(y=o,!(0<st)){for(var d of Ft)xe(d);for(d of ot)xe(d);ot=[],Ft=[],$t={},C=!0}p(0,new et(o))}function un(o){if(a)return we(1,0,o);Xr(o)}var Xr=o=>{if(y=o,a)throw un(o),"unwind";Qr(o)},ot=[],Ft=[],ln=[],$t={},dn=o=>{var d=o.Tc;delete $t[d],ot.push(o),Ft.splice(Ft.indexOf(o),1),o.Tc=0,os(d)};function pn(){ln.forEach(o=>o())}var cn=o=>new Promise(d=>{o.onmessage=b=>{var T=b.data;if(b=T.Uc,T.ad&&T.ad!=Tr()){var z=$t[T.ad];z?z.postMessage(T,T.ud):k(`Internal error! Worker sent a message "${b}" to target pthread ${T.ad}, but that thread no longer exists!`)}else b==="checkMailbox"?wr():b==="spawnThread"?fr(T):b==="cleanupThread"?_r(()=>{dn($t[T.Qd])}):b==="loaded"?(o.loaded=!0,d(o)):T.target==="setimmediate"?o.postMessage(T):b==="uncaughtException"?o.onerror(T.error):b==="callHandler"?t[T.zd](...T.args):b&&k(`worker sent an unknown command ${b}`)},o.onerror=b=>{throw k(`worker sent an error! ${b.filename}:${b.lineno}: ${b.message}`),b};var m,c=[];for(m of[])t.propertyIsEnumerable(m)&&c.push(m);o.postMessage({Uc:"load",Ad:c,Vd:ut,Wd:_})});function hn(){var o=new Worker((()=>{let d=URL;return import.meta.url>"file:"&&import.meta.url<"file;"?new d("ort.bundle.min.mjs",import.meta.url):new URL(import.meta.url)})(),{type:"module",workerData:"em-pthread",name:"em-pthread"});ot.push(o)}var ut,bf=(o,d)=>{st=0,o=gi(o,d),0<st?y=o:hi(o)},mr=[],gr=0;function $f(o){var d=new Yr(o>>>=0);return(x(),Q)[d.Vc+12>>>0]==0&&(fn(d,!0),gr--),mn(d,!1),mr.push(d),hs(o)}var Pt=0,vf=()=>{oe(0,0);var o=mr.pop();ps(o.ed),Pt=0};function fn(o,d){d=d?1:0,(x(),Q)[o.Vc+12>>>0]=d}function mn(o,d){d=d?1:0,(x(),Q)[o.Vc+13>>>0]=d}class Yr{constructor(d){this.ed=d,this.Vc=d-24}}var Jr=o=>{var d=Pt;if(!d)return Kt(0),0;var m=new Yr(d);(x(),U)[m.Vc+16>>>2>>>0]=d;var c=(x(),U)[m.Vc+4>>>2>>>0];if(!c)return Kt(0),d;for(var b of o){if(b===0||b===c)break;if(cs(b,c,m.Vc+16))return Kt(b),d}return Kt(c),d};function xf(){return Jr([])}function Sf(o){return Jr([o>>>0])}function Tf(o,d,m,c){return Jr([o>>>0,d>>>0,m>>>0,c>>>0])}var If=()=>{var o=mr.pop();o||Te("no exception to throw");var d=o.ed;throw(x(),Q)[o.Vc+13>>>0]==0&&(mr.push(o),mn(o,!0),fn(o,!1),gr++),mi(d),Pt=d};function Ef(o,d,m){var c=new Yr(o>>>=0);throw d>>>=0,m>>>=0,(x(),U)[c.Vc+16>>>2>>>0]=0,(x(),U)[c.Vc+4>>>2>>>0]=d,(x(),U)[c.Vc+8>>>2>>>0]=m,mi(o),gr++,Pt=o}var kf=()=>gr;function gn(o,d,m,c){return a?we(2,1,o,d,m,c):yn(o,d,m,c)}function yn(o,d,m,c){if(o>>>=0,d>>>=0,m>>>=0,c>>>=0,!globalThis.SharedArrayBuffer)return 6;var b=[];return a&&b.length===0?gn(o,d,m,c):(o={Od:m,Tc:o,dd:c,ud:b},a?(o.Uc="spawnThread",postMessage(o,b),0):fr(o))}function zf(o){throw Pt||=o>>>0,Pt}var _n=globalThis.TextDecoder&&new TextDecoder,wn=(o,d,m,c)=>{if(m=d+m,c)return m;for(;o[d]&&!(d>=m);)++d;return d},bn=(o,d=0,m,c)=>{if(16<(m=wn(o,d>>>=0,m,c))-d&&o.buffer&&_n)return _n.decode(o.buffer instanceof ArrayBuffer?o.subarray(d,m):o.slice(d,m));for(c="";d<m;){var b=o[d++];if(128&b){var T=63&o[d++];if((224&b)==192)c+=String.fromCharCode((31&b)<<6|T);else{var z=63&o[d++];65536>(b=(240&b)==224?(15&b)<<12|T<<6|z:(7&b)<<18|T<<12|z<<6|63&o[d++])?c+=String.fromCharCode(b):(b-=65536,c+=String.fromCharCode(55296|b>>10,56320|1023&b))}}else c+=String.fromCharCode(b)}return c},Se=(o,d,m)=>(o>>>=0)?bn((x(),L),o,d,m):"";function $n(o,d,m){return a?we(3,1,o,d,m):0}function vn(o,d){if(a)return we(4,1,o,d)}function xn(o,d){if(a)return we(5,1,o,d)}function Sn(o,d,m){if(a)return we(6,1,o,d,m)}function Tn(o,d,m){return a?we(7,1,o,d,m):0}function In(o,d){if(a)return we(8,1,o,d)}function En(o,d,m){if(a)return we(9,1,o,d,m)}function kn(o,d,m,c){if(a)return we(10,1,o,d,m,c)}function zn(o,d,m,c){if(a)return we(11,1,o,d,m,c)}function Cn(o,d,m,c){if(a)return we(12,1,o,d,m,c)}function An(o){if(a)return we(13,1,o)}function On(o,d){if(a)return we(14,1,o,d)}function Bn(o,d,m){if(a)return we(15,1,o,d,m)}var Cf=()=>Te(""),Qe=o=>{o>>>=0;for(var d="";;){var m=(x(),L)[o++>>>0];if(!m)return d;d+=String.fromCharCode(m)}},ei={},ti={},Af={},Ut=class extends Error{constructor(o){super(o),this.name="BindingError"}};function tt(o,d,m={}){return(function(c,b,T={}){var z=b.name;if(!c)throw new Ut(`type "${z}" must have a positive integer typeid pointer`);if(ti.hasOwnProperty(c)){if(T.Bd)return;throw new Ut(`Cannot register type '${z}' twice`)}ti[c]=b,delete Af[c],ei.hasOwnProperty(c)&&(b=ei[c],delete ei[c],b.forEach(R=>R()))})(o,d,m)}var Rn=(o,d,m)=>{switch(d){case 1:return m?c=>(x(),Q)[c>>>0]:c=>(x(),L)[c>>>0];case 2:return m?c=>(x(),K)[c>>>1>>>0]:c=>(x(),se)[c>>>1>>>0];case 4:return m?c=>(x(),A)[c>>>2>>>0]:c=>(x(),U)[c>>>2>>>0];case 8:return m?c=>(x(),H)[c>>>3>>>0]:c=>(x(),le)[c>>>3>>>0];default:throw new TypeError(`invalid integer width (${d}): ${o}`)}};function Of(o,d,m,c,b){o>>>=0,m>>>=0,d=Qe(d>>>0);let T=z=>z;if(c=c===0n){let z=8*m;T=R=>BigInt.asUintN(z,R),b=T(b)}tt(o,{name:d,Qc:T,Xc:(z,R)=>(typeof R=="number"&&(R=BigInt(R)),R),Wc:Rn(d,m,!c),Yc:null})}function Bf(o,d,m,c){tt(o>>>=0,{name:d=Qe(d>>>0),Qc:function(b){return!!b},Xc:function(b,T){return T?m:c},Wc:function(b){return this.Qc((x(),L)[b>>>0])},Yc:null})}var Nn=[],vt=[0,1,,1,null,1,!0,1,!1,1];function ri(o){9<(o>>>=0)&&--vt[o+1]==0&&(vt[o]=void 0,Nn.push(o))}var De=o=>{if(!o)throw new Ut(`Cannot use deleted val. handle = ${o}`);return vt[o]},Le=o=>{switch(o){case void 0:return 2;case null:return 4;case!0:return 6;case!1:return 8;default:let d=Nn.pop()||vt.length;return vt[d]=o,vt[d+1]=1,d}};function ii(o){return this.Qc((x(),U)[o>>>2>>>0])}var Rf={name:"emscripten::val",Qc:o=>{var d=De(o);return ri(o),d},Xc:(o,d)=>Le(d),Wc:ii,Yc:null};function Nf(o){return tt(o>>>0,Rf)}var Mf=(o,d)=>{switch(d){case 4:return function(m){return this.Qc((x(),ee)[m>>>2>>>0])};case 8:return function(m){return this.Qc((x(),X)[m>>>3>>>0])};default:throw new TypeError(`invalid float width (${d}): ${o}`)}};function Df(o,d,m){m>>>=0,tt(o>>>=0,{name:d=Qe(d>>>0),Qc:c=>c,Xc:(c,b)=>b,Wc:Mf(d,m),Yc:null})}function Pf(o,d,m,c,b){o>>>=0,m>>>=0,d=Qe(d>>>0);let T=R=>R;if(c===0){var z=32-8*m;T=R=>R<<z>>>z,b=T(b)}tt(o,{name:d,Qc:T,Xc:(R,W)=>W,Wc:Rn(d,m,c!==0),Yc:null})}function Uf(o,d,m){function c(T){var z=(x(),U)[T>>>2>>>0];return T=(x(),U)[T+4>>>2>>>0],new b((x(),Q).buffer,T,z)}var b=[Int8Array,Uint8Array,Int16Array,Uint16Array,Int32Array,Uint32Array,Float32Array,Float64Array,BigInt64Array,BigUint64Array][d];tt(o>>>=0,{name:m=Qe(m>>>0),Qc:c,Wc:c},{Bd:!0})}var lt=(o,d,m)=>{var c=(x(),L);if(d>>>=0,0<m){var b=d;m=d+m-1;for(var T=0;T<o.length;++T){var z=o.codePointAt(T);if(127>=z){if(d>=m)break;c[d++>>>0]=z}else if(2047>=z){if(d+1>=m)break;c[d++>>>0]=192|z>>6,c[d++>>>0]=128|63&z}else if(65535>=z){if(d+2>=m)break;c[d++>>>0]=224|z>>12,c[d++>>>0]=128|z>>6&63,c[d++>>>0]=128|63&z}else{if(d+3>=m)break;c[d++>>>0]=240|z>>18,c[d++>>>0]=128|z>>12&63,c[d++>>>0]=128|z>>6&63,c[d++>>>0]=128|63&z,T++}}c[d>>>0]=0,o=d-b}else o=0;return o},yr=o=>{for(var d=0,m=0;m<o.length;++m){var c=o.charCodeAt(m);127>=c?d++:2047>=c?d+=2:55296<=c&&57343>=c?(d+=4,++m):d+=3}return d};function Wf(o,d){tt(o>>>=0,{name:d=Qe(d>>>0),Qc(m){var c=(x(),U)[m>>>2>>>0];return c=Se(m+4,c,!0),Ye(m),c},Xc(m,c){c instanceof ArrayBuffer&&(c=new Uint8Array(c));var b=typeof c=="string";if(!(b||ArrayBuffer.isView(c)&&c.BYTES_PER_ELEMENT==1))throw new Ut("Cannot pass non-string to std::string");var T=b?yr(c):c.length,z=jt(4+T+1),R=z+4;return(x(),U)[z>>>2>>>0]=T,b?lt(c,R,T+1):(x(),L).set(c,R>>>0),m!==null&&m.push(Ye,z),z},Wc:ii,Yc(m){Ye(m)}})}var Mn=globalThis.TextDecoder?new TextDecoder("utf-16le"):void 0,qf=(o,d,m)=>{if(o>>>=1,16<(d=wn((x(),se),o,d/2,m))-o&&Mn)return Mn.decode((x(),se).slice(o,d));for(m="";o<d;++o){var c=(x(),se)[o>>>0];m+=String.fromCharCode(c)}return m},Lf=(o,d,m)=>{if(m??=2147483647,2>m)return 0;var c=d;m=(m-=2)<2*o.length?m/2:o.length;for(var b=0;b<m;++b){var T=o.charCodeAt(b);(x(),K)[d>>>1>>>0]=T,d+=2}return(x(),K)[d>>>1>>>0]=0,d-c},Vf=o=>2*o.length,Gf=(o,d,m)=>{var c="";o>>>=2;for(var b=0;!(b>=d/4);b++){var T=(x(),U)[o+b>>>0];if(!T&&!m)break;c+=String.fromCodePoint(T)}return c},Ff=(o,d,m)=>{if(d>>>=0,m??=2147483647,4>m)return 0;var c=d;m=c+m-4;for(var b=0;b<o.length;++b){var T=o.codePointAt(b);if(65535<T&&b++,(x(),A)[d>>>2>>>0]=T,(d+=4)+4>m)break}return(x(),A)[d>>>2>>>0]=0,d-c},Hf=o=>{for(var d=0,m=0;m<o.length;++m)65535<o.codePointAt(m)&&m++,d+=4;return d};function jf(o,d,m){if(o>>>=0,d>>>=0,m=Qe(m>>>=0),d===2)var c=qf,b=Lf,T=Vf;else c=Gf,b=Ff,T=Hf;tt(o,{name:m,Qc:z=>{var R=(x(),U)[z>>>2>>>0];return R=c(z+4,R*d,!0),Ye(z),R},Xc:(z,R)=>{if(typeof R!="string")throw new Ut(`Cannot pass non-string to C++ string type ${m}`);var W=T(R),G=jt(4+W+d);return(x(),U)[G>>>2>>>0]=W/d,b(R,G+4,W+d),z!==null&&z.push(Ye,G),G},Wc:ii,Yc(z){Ye(z)}})}function Kf(o,d){tt(o>>>=0,{Cd:!0,name:d=Qe(d>>>0),Qc:()=>{},Xc:()=>{}})}function Zf(o){ci(o>>>0,!i,1,!r,131072,!1),pn()}var _r=o=>{if(!C)try{if(o(),!(0<st))try{a?Tr()&&hi(y):Xr(y)}catch(d){d instanceof et||d=="unwind"||p(0,d)}}catch(d){d instanceof et||d=="unwind"||p(0,d)}},Qf=!Atomics.waitAsync||globalThis.navigator?.userAgent&&91>Number((navigator.userAgent.match(/Chrom(e|ium)\/([0-9]+)\./)||[])[2]);function ai(o){o>>>=0,Qf||(Atomics.waitAsync((x(),A),o>>>2,o).value.then(wr),o+=128,Atomics.store((x(),A),o>>>2,1))}var wr=()=>_r(()=>{var o=Tr();o&&(ai(o),ls())});function Xf(o,d){(o>>>=0)==d>>>0?setTimeout(wr):a?postMessage({ad:o,Uc:"checkMailbox"}):(o=$t[o])&&o.postMessage({Uc:"checkMailbox"})}var ni=[];function Yf(o,d,m,c,b){for(d>>>=0,b>>>=0,ni.length=0,m=b>>>3,c=b+c>>>3;m<c;){var T;T=(x(),H)[m++>>>0]?(x(),H)[m++>>>0]:(x(),X)[m++>>>0],ni.push(T)}return(d?yi[d]:Vm[o])(...ni)}var Jf=()=>{st=0};function em(o){o>>>=0,a?postMessage({Uc:"cleanupThread",Qd:o}):dn($t[o])}function tm(o){}var br=o=>{try{o()}catch(d){Te(d)}};function rm(o){var d=(...m)=>{$r.push(o);try{return o(...m)}finally{C||($r.pop(),Xe&&dt===1&&$r.length===0&&(dt=0,st+=1,br(Ys),typeof Fibers<"u"&&Fibers.ce()))}};return Un.set(o,d),d}var dt=0,Xe=null,Dn=0,$r=[],si=new Map,Pn=new Map,Un=new Map,im=0,oi=null,am=[],Wn=o=>(function(d){if(!C){if(dt===0){var m=!1,c=!1;d((b=0)=>{if(!C&&(Dn=b,m=!0,c)){dt=2,br(()=>Js(Xe)),typeof MainLoop<"u"&&MainLoop.yd&&MainLoop.resume(),b=!1;try{var T=(function(){var W=(x(),A)[Xe+8>>>2>>>0];return W=Pn.get(W),W=Un.get(W),--st,W()})()}catch(W){T=W,b=!0}var z=!1;if(!Xe){var R=oi;R&&(oi=null,(b?R.reject:R.resolve)(T),z=!0)}if(b&&!z)throw T}}),c=!0,m||(dt=1,Xe=(function(){var b=jt(65548),T=b+12;if((x(),U)[b>>>2>>>0]=T,(x(),U)[b+4>>>2>>>0]=T+65536,T=$r[0],!si.has(T)){var z=im++;si.set(T,z),Pn.set(z,T)}return T=si.get(T),(x(),A)[b+8>>>2>>>0]=T,b})(),typeof MainLoop<"u"&&MainLoop.yd&&MainLoop.pause(),br(()=>Xs(Xe)))}else dt===2?(dt=0,br(eo),Ye(Xe),Xe=null,am.forEach(_r)):Te(`invalid state: ${dt}`);return Dn}})(d=>{o().then(d)});function nm(o){return o>>>=0,Wn(async()=>{var d=await De(o);return Le(d)})}var ui=[],sm=o=>{var d=ui.length;return ui.push(o),d},om=(o,d)=>{for(var m=Array(o),c=0;c<o;++c){var b=c,T=(x(),U)[d+4*c>>>2>>>0],z=ti[T];if(z===void 0)throw o=`parameter ${c}`,T=is(T),d=Qe(T),Ye(T),new Ut(`${o} has unknown type ${d}`);m[b]=z}return m},um=(o,d,m)=>{var c=[];return o=o(c,m),c.length&&((x(),U)[d>>>2>>>0]=Le(c)),o},lm={},vr=o=>{var d=lm[o];return d===void 0?Qe(o):d};function dm(o,d,m){var[c,...b]=om(o,d>>>0);d=c.Xc.bind(c);var T=b.map(W=>W.Wc.bind(W));o--;var z={toValue:De};switch(o=T.map((W,G)=>{var ie=`argFromPtr${G}`;return z[ie]=W,`${ie}(args${G?"+"+8*G:""})`}),m){case 0:var R="toValue(handle)";break;case 2:R="new (toValue(handle))";break;case 3:R="";break;case 1:z.getStringOrSymbol=vr,R="toValue(handle)[getStringOrSymbol(methodName)]"}return R+=`(${o})`,c.Cd||(z.toReturnWire=d,z.emval_returnValue=um,R=`return emval_returnValue(toReturnWire, destructorsRef, ${R})`),R=`return function (handle, methodName, destructorsRef, args) {
+var Ag=Object.defineProperty;var St=(e=>typeof require<"u"?require:typeof Proxy<"u"?new Proxy(e,{get:(t,r)=>(typeof require<"u"?require:t)[r]}):e)(function(e){if(typeof require<"u")return require.apply(this,arguments);throw Error('Dynamic require of "'+e+'" is not supported')});var Og=(e,t)=>{for(var r in t)Ag(e,r,{get:t[r],enumerable:!0})};var on={};Og(on,{InferenceSession:()=>_t,TRACE:()=>cr,TRACE_EVENT_BEGIN:()=>mt,TRACE_EVENT_END:()=>gt,TRACE_FUNC_BEGIN:()=>Ze,TRACE_FUNC_END:()=>Ue,Tensor:()=>Te,default:()=>Y0,env:()=>ge,registerBackend:()=>Ot});var Oa=Object.defineProperty,Bg=Object.getOwnPropertyDescriptor,Rg=Object.getOwnPropertyNames,Mg=Object.prototype.hasOwnProperty,Ng=(e=>typeof St<"u"?St:typeof Proxy<"u"?new Proxy(e,{get:(t,r)=>(typeof St<"u"?St:t)[r]}):e)(function(e){if(typeof St<"u")return St.apply(this,arguments);throw Error('Dynamic require of "'+e+'" is not supported')}),P=(e,t)=>()=>(e&&(t=e(e=0)),t),Gt=(e,t)=>{for(var r in t)Oa(e,r,{get:t[r],enumerable:!0})},Dg=(e,t,r,i)=>{if(t&&typeof t=="object"||typeof t=="function")for(let a of Rg(t))!Mg.call(e,a)&&a!==r&&Oa(e,a,{get:()=>t[a],enumerable:!(i=Bg(t,a))||i.enumerable});return e},pr=e=>Dg(Oa({},"__esModule",{value:!0}),e),Zt,ct,Ot,to,Md,Nd=P(()=>{"use strict";Zt=new Map,ct=[],Ot=(e,t,r)=>{if(t&&typeof t.init=="function"&&typeof t.createInferenceSessionHandler=="function"){let i=Zt.get(e);if(i===void 0)Zt.set(e,{backend:t,priority:r});else{if(i.priority>r)return;if(i.priority===r&&i.backend!==t)throw new Error(`cannot register backend "${e}" using priority ${r}`)}if(r>=0){let a=ct.indexOf(e);a!==-1&&ct.splice(a,1);for(let s=0;s<ct.length;s++)if(Zt.get(ct[s]).priority<=r){ct.splice(s,0,e);return}ct.push(e)}return}throw new TypeError("not a valid backend")},to=async e=>{let t=Zt.get(e);if(!t)return"backend not found.";if(t.initialized)return t.backend;if(t.aborted)return t.error;{let r=!!t.initPromise;try{return r||(t.initPromise=t.backend.init(e)),await t.initPromise,t.initialized=!0,t.backend}catch(i){return r||(t.error=`${i}`,t.aborted=!0),t.error}finally{delete t.initPromise}}},Md=async e=>{let t=e.executionProviders||[],r=t.map(l=>typeof l=="string"?l:l.name),i=r.length===0?ct:r,a,s=[],n=new Set;for(let l of i){let p=await to(l);typeof p=="string"?s.push({name:l,err:p}):(a||(a=p),a===p&&n.add(l))}if(!a)throw new Error(`no available backend found. ERR: ${s.map(l=>`[${l.name}] ${l.err}`).join(", ")}`);for(let{name:l,err:p}of s)r.includes(l)&&console.warn(`removing requested execution provider "${l}" from session options because it is not available: ${p}`);let u=t.filter(l=>n.has(typeof l=="string"?l:l.name));return[a,new Proxy(e,{get:(l,p)=>p==="executionProviders"?u:Reflect.get(l,p)})]}}),Pg=P(()=>{"use strict";Nd()}),Dd,Ug=P(()=>{"use strict";Dd="1.24.1"}),bi,Ee,Pd=P(()=>{"use strict";Ug(),bi="warning",Ee={wasm:{},webgl:{},webgpu:{},versions:{common:Dd},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}`);bi=e}},get logLevel(){return bi}},Object.defineProperty(Ee,"logLevel",{enumerable:!0})}),ge,Wg=P(()=>{"use strict";Pd(),ge=Ee}),Ud,Wd,Lg=P(()=>{"use strict";Ud=(e,t)=>{let r=typeof document<"u"?document.createElement("canvas"):new OffscreenCanvas(1,1);r.width=e.dims[3],r.height=e.dims[2];let i=r.getContext("2d");if(i!=null){let a,s;t?.tensorLayout!==void 0&&t.tensorLayout==="NHWC"?(a=e.dims[2],s=e.dims[3]):(a=e.dims[3],s=e.dims[2]);let n=t?.format!==void 0?t.format:"RGB",u=t?.norm,l,p;u===void 0||u.mean===void 0?l=[255,255,255,255]:typeof u.mean=="number"?l=[u.mean,u.mean,u.mean,u.mean]:(l=[u.mean[0],u.mean[1],u.mean[2],0],u.mean[3]!==void 0&&(l[3]=u.mean[3])),u===void 0||u.bias===void 0?p=[0,0,0,0]:typeof u.bias=="number"?p=[u.bias,u.bias,u.bias,u.bias]:(p=[u.bias[0],u.bias[1],u.bias[2],0],u.bias[3]!==void 0&&(p[3]=u.bias[3]));let h=s*a,f=0,g=h,_=h*2,y=-1;n==="RGBA"?(f=0,g=h,_=h*2,y=h*3):n==="RGB"?(f=0,g=h,_=h*2):n==="RBG"&&(f=0,_=h,g=h*2);for(let $=0;$<s;$++)for(let S=0;S<a;S++){let v=(e.data[f++]-p[0])*l[0],w=(e.data[g++]-p[1])*l[1],k=(e.data[_++]-p[2])*l[2],I=y===-1?255:(e.data[y++]-p[3])*l[3];i.fillStyle="rgba("+v+","+w+","+k+","+I+")",i.fillRect(S,$,1,1)}if("toDataURL"in r)return r.toDataURL();throw new Error("toDataURL is not supported")}else throw new Error("Can not access image data")},Wd=(e,t)=>{let r=typeof document<"u"?document.createElement("canvas").getContext("2d"):new OffscreenCanvas(1,1).getContext("2d"),i;if(r!=null){let a,s,n;t?.tensorLayout!==void 0&&t.tensorLayout==="NHWC"?(a=e.dims[2],s=e.dims[1],n=e.dims[3]):(a=e.dims[3],s=e.dims[2],n=e.dims[1]);let u=t!==void 0&&t.format!==void 0?t.format:"RGB",l=t?.norm,p,h;l===void 0||l.mean===void 0?p=[255,255,255,255]:typeof l.mean=="number"?p=[l.mean,l.mean,l.mean,l.mean]:(p=[l.mean[0],l.mean[1],l.mean[2],255],l.mean[3]!==void 0&&(p[3]=l.mean[3])),l===void 0||l.bias===void 0?h=[0,0,0,0]:typeof l.bias=="number"?h=[l.bias,l.bias,l.bias,l.bias]:(h=[l.bias[0],l.bias[1],l.bias[2],0],l.bias[3]!==void 0&&(h[3]=l.bias[3]));let f=s*a;if(t!==void 0&&(t.format!==void 0&&n===4&&t.format!=="RGBA"||n===3&&t.format!=="RGB"&&t.format!=="BGR"))throw new Error("Tensor format doesn't match input tensor dims");let g=4,_=0,y=1,$=2,S=3,v=0,w=f,k=f*2,I=-1;u==="RGBA"?(v=0,w=f,k=f*2,I=f*3):u==="RGB"?(v=0,w=f,k=f*2):u==="RBG"&&(v=0,k=f,w=f*2),i=r.createImageData(a,s);for(let E=0;E<s*a;_+=g,y+=g,$+=g,S+=g,E++)i.data[_]=(e.data[v++]-h[0])*p[0],i.data[y]=(e.data[w++]-h[1])*p[1],i.data[$]=(e.data[k++]-h[2])*p[2],i.data[S]=I===-1?255:(e.data[I++]-h[3])*p[3]}else throw new Error("Can not access image data");return i}}),kr,Ld,qd,Vd,Gd,Fd,qg=P(()=>{"use strict";Ba(),kr=(e,t)=>{if(e===void 0)throw new Error("Image buffer must be defined");if(t.height===void 0||t.width===void 0)throw new Error("Image height and width must be defined");if(t.tensorLayout==="NHWC")throw new Error("NHWC Tensor layout is not supported yet");let{height:r,width:i}=t,a=t.norm??{mean:255,bias:0},s,n;typeof a.mean=="number"?s=[a.mean,a.mean,a.mean,a.mean]:s=[a.mean[0],a.mean[1],a.mean[2],a.mean[3]??255],typeof a.bias=="number"?n=[a.bias,a.bias,a.bias,a.bias]:n=[a.bias[0],a.bias[1],a.bias[2],a.bias[3]??0];let u=t.format!==void 0?t.format:"RGBA",l=t.tensorFormat!==void 0&&t.tensorFormat!==void 0?t.tensorFormat:"RGB",p=r*i,h=l==="RGBA"?new Float32Array(p*4):new Float32Array(p*3),f=4,g=0,_=1,y=2,$=3,S=0,v=p,w=p*2,k=-1;u==="RGB"&&(f=3,g=0,_=1,y=2,$=-1),l==="RGBA"?k=p*3:l==="RBG"?(S=0,w=p,v=p*2):l==="BGR"&&(w=0,v=p,S=p*2);for(let I=0;I<p;I++,g+=f,y+=f,_+=f,$+=f)h[S++]=(e[g]+n[0])/s[0],h[v++]=(e[_]+n[1])/s[1],h[w++]=(e[y]+n[2])/s[2],k!==-1&&$!==-1&&(h[k++]=(e[$]+n[3])/s[3]);return l==="RGBA"?new Me("float32",h,[1,4,r,i]):new Me("float32",h,[1,3,r,i])},Ld=async(e,t)=>{let r=typeof HTMLImageElement<"u"&&e instanceof HTMLImageElement,i=typeof ImageData<"u"&&e instanceof ImageData,a=typeof ImageBitmap<"u"&&e instanceof ImageBitmap,s=typeof e=="string",n,u=t??{},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")},p=h=>typeof HTMLCanvasElement<"u"&&h instanceof HTMLCanvasElement||h instanceof OffscreenCanvas?h.getContext("2d"):null;if(r){let h=l();h.width=e.width,h.height=e.height;let f=p(h);if(f!=null){let g=e.height,_=e.width;if(t!==void 0&&t.resizedHeight!==void 0&&t.resizedWidth!==void 0&&(g=t.resizedHeight,_=t.resizedWidth),t!==void 0){if(u=t,t.tensorFormat!==void 0)throw new Error("Image input config format must be RGBA for HTMLImageElement");u.tensorFormat="RGBA",u.height=g,u.width=_}else u.tensorFormat="RGBA",u.height=g,u.width=_;f.drawImage(e,0,0),n=f.getImageData(0,0,_,g).data}else throw new Error("Can not access image data")}else if(i){let h,f;if(t!==void 0&&t.resizedWidth!==void 0&&t.resizedHeight!==void 0?(h=t.resizedHeight,f=t.resizedWidth):(h=e.height,f=e.width),t!==void 0&&(u=t),u.format="RGBA",u.height=h,u.width=f,t!==void 0){let g=l();g.width=f,g.height=h;let _=p(g);if(_!=null)_.putImageData(e,0,0),n=_.getImageData(0,0,f,h).data;else throw new Error("Can not access image data")}else n=e.data}else if(a){if(t===void 0)throw new Error("Please provide image config with format for Imagebitmap");let h=l();h.width=e.width,h.height=e.height;let f=p(h);if(f!=null){let g=e.height,_=e.width;return f.drawImage(e,0,0,_,g),n=f.getImageData(0,0,_,g).data,u.height=g,u.width=_,kr(n,u)}else throw new Error("Can not access image data")}else{if(s)return new Promise((h,f)=>{let g=l(),_=p(g);if(!e||!_)return f();let y=new Image;y.crossOrigin="Anonymous",y.src=e,y.onload=()=>{g.width=y.width,g.height=y.height,_.drawImage(y,0,0,g.width,g.height);let $=_.getImageData(0,0,g.width,g.height);u.height=g.height,u.width=g.width,h(kr($.data,u))}});throw new Error("Input data provided is not supported - aborted tensor creation")}if(n!==void 0)return kr(n,u);throw new Error("Input data provided is not supported - aborted tensor creation")},qd=(e,t)=>{let{width:r,height:i,download:a,dispose:s}=t,n=[1,i,r,4];return new Me({location:"texture",type:"float32",texture:e,dims:n,download:a,dispose:s})},Vd=(e,t)=>{let{dataType:r,dims:i,download:a,dispose:s}=t;return new Me({location:"gpu-buffer",type:r??"float32",gpuBuffer:e,dims:i,download:a,dispose:s})},Gd=(e,t)=>{let{dataType:r,dims:i,download:a,dispose:s}=t;return new Me({location:"ml-tensor",type:r??"float32",mlTensor:e,dims:i,download:a,dispose:s})},Fd=(e,t,r)=>new Me({location:"cpu-pinned",type:e,data:t,dims:r??[t.length]})}),zt,sr,$i,Hd,Vg=P(()=>{"use strict";zt=new Map([["float32",Float32Array],["uint8",Uint8Array],["int8",Int8Array],["uint16",Uint16Array],["int16",Int16Array],["int32",Int32Array],["bool",Uint8Array],["float64",Float64Array],["uint32",Uint32Array],["int4",Uint8Array],["uint4",Uint8Array]]),sr=new Map([[Float32Array,"float32"],[Uint8Array,"uint8"],[Int8Array,"int8"],[Uint16Array,"uint16"],[Int16Array,"int16"],[Int32Array,"int32"],[Float64Array,"float64"],[Uint32Array,"uint32"]]),$i=!1,Hd=()=>{if(!$i){$i=!0;let e=typeof BigInt64Array<"u"&&BigInt64Array.from,t=typeof BigUint64Array<"u"&&BigUint64Array.from,r=globalThis.Float16Array,i=typeof r<"u"&&r.from;e&&(zt.set("int64",BigInt64Array),sr.set(BigInt64Array,"int64")),t&&(zt.set("uint64",BigUint64Array),sr.set(BigUint64Array,"uint64")),i?(zt.set("float16",r),sr.set(r,"float16")):zt.set("float16",Uint16Array)}}}),jd,Kd,Gg=P(()=>{"use strict";Ba(),jd=e=>{let t=1;for(let r=0;r<e.length;r++){let i=e[r];if(typeof i!="number"||!Number.isSafeInteger(i))throw new TypeError(`dims[${r}] must be an integer, got: ${i}`);if(i<0)throw new RangeError(`dims[${r}] must be a non-negative integer, got: ${i}`);t*=i}return t},Kd=(e,t)=>{switch(e.location){case"cpu":return new Me(e.type,e.data,t);case"cpu-pinned":return new Me({location:"cpu-pinned",data:e.data,type:e.type,dims:t});case"texture":return new Me({location:"texture",texture:e.texture,type:e.type,dims:t});case"gpu-buffer":return new Me({location:"gpu-buffer",gpuBuffer:e.gpuBuffer,type:e.type,dims:t});case"ml-tensor":return new Me({location:"ml-tensor",mlTensor:e.mlTensor,type:e.type,dims:t});default:throw new Error(`tensorReshape: tensor location ${e.location} is not supported`)}}}),Me,Ba=P(()=>{"use strict";Lg(),qg(),Vg(),Gg(),Me=class{constructor(e,t,r){Hd();let i,a;if(typeof e=="object"&&"location"in e)switch(this.dataLocation=e.location,i=e.type,a=e.dims,e.location){case"cpu-pinned":{let n=zt.get(i);if(!n)throw new TypeError(`unsupported type "${i}" to create tensor from pinned buffer`);if(!(e.data instanceof n))throw new TypeError(`buffer should be of type ${n.name}`);this.cpuData=e.data;break}case"texture":{if(i!=="float32")throw new TypeError(`unsupported type "${i}" to create tensor from texture`);this.gpuTextureData=e.texture,this.downloader=e.download,this.disposer=e.dispose;break}case"gpu-buffer":{if(i!=="float32"&&i!=="float16"&&i!=="int32"&&i!=="int64"&&i!=="uint32"&&i!=="uint8"&&i!=="bool"&&i!=="uint4"&&i!=="int4")throw new TypeError(`unsupported type "${i}" to create tensor from gpu buffer`);this.gpuBufferData=e.gpuBuffer,this.downloader=e.download,this.disposer=e.dispose;break}case"ml-tensor":{if(i!=="float32"&&i!=="float16"&&i!=="int32"&&i!=="int64"&&i!=="uint32"&&i!=="uint64"&&i!=="int8"&&i!=="uint8"&&i!=="bool"&&i!=="uint4"&&i!=="int4")throw new TypeError(`unsupported type "${i}" 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 n,u;if(typeof e=="string")if(i=e,u=r,e==="string"){if(!Array.isArray(t))throw new TypeError("A string tensor's data must be a string array.");n=t}else{let l=zt.get(e);if(l===void 0)throw new TypeError(`Unsupported tensor type: ${e}.`);if(Array.isArray(t)){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"?n=l.from(t,BigInt):n=l.from(t)}else if(t instanceof l)n=t;else if(t instanceof Uint8ClampedArray)if(e==="uint8")n=Uint8Array.from(t);else throw new TypeError("A Uint8ClampedArray tensor's data must be type of uint8");else if(e==="float16"&&t instanceof Uint16Array&&l!==Uint16Array)n=new globalThis.Float16Array(t.buffer,t.byteOffset,t.length);else throw new TypeError(`A ${i} tensor's data must be type of ${l}`)}else if(u=t,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")i="string",n=e;else if(l==="boolean")i="bool",n=Uint8Array.from(e);else throw new TypeError(`Invalid element type of data array: ${l}.`)}else if(e instanceof Uint8ClampedArray)i="uint8",n=Uint8Array.from(e);else{let l=sr.get(e.constructor);if(l===void 0)throw new TypeError(`Unsupported type for tensor data: ${e.constructor}.`);i=l,n=e}if(u===void 0)u=[n.length];else if(!Array.isArray(u))throw new TypeError("A tensor's dims must be a number array");a=u,this.cpuData=n,this.dataLocation="cpu"}let s=jd(a);if(this.cpuData&&s!==this.cpuData.length&&!((i==="uint4"||i==="int4")&&Math.ceil(s/2)===this.cpuData.length))throw new Error(`Tensor's size(${s}) does not match data length(${this.cpuData.length}).`);this.type=i,this.dims=a,this.size=s}static async fromImage(e,t){return Ld(e,t)}static fromTexture(e,t){return qd(e,t)}static fromGpuBuffer(e,t){return Vd(e,t)}static fromMLTensor(e,t){return Gd(e,t)}static fromPinnedBuffer(e,t,r){return Fd(e,t,r)}toDataURL(e){return Ud(this,e)}toImageData(e){return Wd(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 t=await this.downloader();return this.downloader=void 0,this.dataLocation="cpu",this.cpuData=t,e&&this.disposer&&(this.disposer(),this.disposer=void 0),t}finally{this.isDownloading=!1}}default:throw new Error(`cannot get data from location: ${this.dataLocation}`)}}dispose(){if(this.isDownloading)throw new Error("The current tensor is being downloaded.");this.disposer&&(this.disposer(),this.disposer=void 0),this.cpuData=void 0,this.gpuTextureData=void 0,this.gpuBufferData=void 0,this.mlTensorData=void 0,this.downloader=void 0,this.isDownloading=void 0,this.dataLocation="none"}ensureValid(){if(this.dataLocation==="none")throw new Error("The tensor is disposed.")}reshape(e){if(this.ensureValid(),this.downloader||this.disposer)throw new Error("Cannot reshape a tensor that owns GPU resource.");return Kd(this,e)}}}),Te,Zd=P(()=>{"use strict";Ba(),Te=Me}),cr,vi,Ze,Ue,mt,gt,Qd=P(()=>{"use strict";Pd(),cr=(e,t)=>{(typeof Ee.trace>"u"?!Ee.wasm.trace:!Ee.trace)||console.timeStamp(`${e}::ORT::${t}`)},vi=(e,t)=>{let r=new Error().stack?.split(/\r\n|\r|\n/g)||[],i=!1;for(let a=0;a<r.length;a++){if(i&&!r[a].includes("TRACE_FUNC")){let s=`FUNC_${e}::${r[a].trim().split(" ")[1]}`;t&&(s+=`::${t}`),cr("CPU",s);return}r[a].includes("TRACE_FUNC")&&(i=!0)}},Ze=e=>{(typeof Ee.trace>"u"?!Ee.wasm.trace:!Ee.trace)||vi("BEGIN",e)},Ue=e=>{(typeof Ee.trace>"u"?!Ee.wasm.trace:!Ee.trace)||vi("END",e)},mt=e=>{(typeof Ee.trace>"u"?!Ee.wasm.trace:!Ee.trace)||console.time(`ORT::${e}`)},gt=e=>{(typeof Ee.trace>"u"?!Ee.wasm.trace:!Ee.trace)||console.timeEnd(`ORT::${e}`)}}),Xd,Fg=P(()=>{"use strict";Nd(),Zd(),Qd(),Xd=class Yd{constructor(t){this.handler=t}async run(t,r,i){Ze(),mt("InferenceSession.run");let a={},s={};if(typeof t!="object"||t===null||t instanceof Te||Array.isArray(t))throw new TypeError("'feeds' must be an object that use input names as keys and OnnxValue as corresponding values.");let n=!0;if(typeof r=="object"){if(r===null)throw new TypeError("Unexpected argument[1]: cannot be null.");if(r instanceof Te)throw new TypeError("'fetches' cannot be a Tensor");if(Array.isArray(r)){if(r.length===0)throw new TypeError("'fetches' cannot be an empty array.");n=!1;for(let p of r){if(typeof p!="string")throw new TypeError("'fetches' must be a string array or an object.");if(this.outputNames.indexOf(p)===-1)throw new RangeError(`'fetches' contains invalid output name: ${p}.`);a[p]=null}if(typeof i=="object"&&i!==null)s=i;else if(typeof i<"u")throw new TypeError("'options' must be an object.")}else{let p=!1,h=Object.getOwnPropertyNames(r);for(let f of this.outputNames)if(h.indexOf(f)!==-1){let g=r[f];(g===null||g instanceof Te)&&(p=!0,n=!1,a[f]=g)}if(p){if(typeof i=="object"&&i!==null)s=i;else if(typeof i<"u")throw new TypeError("'options' must be an object.")}else s=r}}else if(typeof r<"u")throw new TypeError("Unexpected argument[1]: must be 'fetches' or 'options'.");for(let p of this.inputNames)if(typeof t[p]>"u")throw new Error(`input '${p}' is missing in 'feeds'.`);if(n)for(let p of this.outputNames)a[p]=null;let u=await this.handler.run(t,a,s),l={};for(let p in u)if(Object.hasOwnProperty.call(u,p)){let h=u[p];h instanceof Te?l[p]=h:l[p]=new Te(h.type,h.data,h.dims)}return gt("InferenceSession.run"),Ue(),l}async release(){return this.handler.dispose()}static async create(t,r,i,a){Ze(),mt("InferenceSession.create");let s,n={};if(typeof t=="string"){if(s=t,typeof r=="object"&&r!==null)n=r;else if(typeof r<"u")throw new TypeError("'options' must be an object.")}else if(t instanceof Uint8Array){if(s=t,typeof r=="object"&&r!==null)n=r;else if(typeof r<"u")throw new TypeError("'options' must be an object.")}else if(t instanceof ArrayBuffer||typeof SharedArrayBuffer<"u"&&t instanceof SharedArrayBuffer){let h=t,f=0,g=t.byteLength;if(typeof r=="object"&&r!==null)n=r;else if(typeof r=="number"){if(f=r,!Number.isSafeInteger(f))throw new RangeError("'byteOffset' must be an integer.");if(f<0||f>=h.byteLength)throw new RangeError(`'byteOffset' is out of range [0, ${h.byteLength}).`);if(g=t.byteLength-f,typeof i=="number"){if(g=i,!Number.isSafeInteger(g))throw new RangeError("'byteLength' must be an integer.");if(g<=0||f+g>h.byteLength)throw new RangeError(`'byteLength' is out of range (0, ${h.byteLength-f}].`);if(typeof a=="object"&&a!==null)n=a;else if(typeof a<"u")throw new TypeError("'options' must be an object.")}else if(typeof i<"u")throw new TypeError("'byteLength' must be a number.")}else if(typeof r<"u")throw new TypeError("'options' must be an object.");s=new Uint8Array(h,f,g)}else throw new TypeError("Unexpected argument[0]: must be 'path' or 'buffer'.");let[u,l]=await Md(n),p=await u.createInferenceSessionHandler(s,l);return gt("InferenceSession.create"),Ue(),new Yd(p)}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}}}),_t,Hg=P(()=>{"use strict";Fg(),_t=Xd}),jg=P(()=>{"use strict"}),Kg=P(()=>{"use strict"}),Zg=P(()=>{"use strict"}),Qg=P(()=>{"use strict"}),Jd={};Gt(Jd,{InferenceSession:()=>_t,TRACE:()=>cr,TRACE_EVENT_BEGIN:()=>mt,TRACE_EVENT_END:()=>gt,TRACE_FUNC_BEGIN:()=>Ze,TRACE_FUNC_END:()=>Ue,Tensor:()=>Te,env:()=>ge,registerBackend:()=>Ot});var We=P(()=>{"use strict";Pg(),Wg(),Hg(),Zd(),jg(),Kg(),Qd(),Zg(),Qg()}),Ra=P(()=>{"use strict"}),ep={};Gt(ep,{default:()=>tp});var xi,Si,tp,Xg=P(()=>{"use strict";of(),Nt(),Ma(),xi="ort-wasm-proxy-worker",Si=globalThis.self?.name===xi,Si&&(self.onmessage=e=>{let{type:t,in:r}=e.data;try{switch(t){case"init-wasm":Na(r.wasm).then(()=>{Ja(r).then(()=>{postMessage({type:t})},i=>{postMessage({type:t,err:i})})},i=>{postMessage({type:t,err:i})});break;case"init-ep":{let{epName:i,env:a}=r;en(a,i).then(()=>{postMessage({type:t})},s=>{postMessage({type:t,err:s})});break}case"copy-from":{let{buffer:i}=r,a=Kr(i);postMessage({type:t,out:a});break}case"create":{let{model:i,options:a}=r;tn(i,a).then(s=>{postMessage({type:t,out:s})},s=>{postMessage({type:t,err:s})});break}case"release":rn(r),postMessage({type:t});break;case"run":{let{sessionId:i,inputIndices:a,inputs:s,outputIndices:n,options:u}=r;an(i,a,s,n,new Array(n.length).fill(null),u).then(l=>{l.some(p=>p[3]!=="cpu")?postMessage({type:t,err:"Proxy does not support non-cpu tensor location."}):postMessage({type:t,out:l},sn([...s,...l]))},l=>{postMessage({type:t,err:l})});break}case"end-profiling":nn(r),postMessage({type:t});break;default:}}catch(i){postMessage({type:t,err:i})}}),tp=Si?null:e=>new Worker(e??Re,{type:"module",name:xi})}),rp={};Gt(rp,{default:()=>ip});async function ro(e={}){var t=e,r=!!globalThis.window,i=!!globalThis.WorkerGlobalScope,a=i&&self.name?.startsWith("em-pthread");t.mountExternalData=(o,d)=>{o.startsWith("./")&&(o=o.substring(2)),(t.Zc||(t.Zc=new Map)).set(o,d)},t.unmountExternalData=()=>{delete t.Zc},globalThis.SharedArrayBuffer??new WebAssembly.Memory({initial:0,maximum:0,ae:!0}).buffer.constructor;let s=o=>async(...d)=>{try{if(t.$c)throw Error("Session already started");let m=t.$c={Nd:d[0],errors:[]},c=await o(...d);if(t.$c!==m)throw Error("Session mismatch");t.gd?.flush();let b=m.errors;if(0<b.length){let T=await Promise.all(b);if(T=T.filter(z=>z),0<T.length)throw Error(T.join(`
+`))}return c}finally{t.$c=null}};t.jsepInit=(o,d)=>{if(o==="webgpu"){[t.gd,t.Dd,t.Hd,t.jd,t.Gd,t.ac,t.Id,t.Kd,t.Ed,t.Fd,t.Jd]=d;let m=t.gd;t.jsepRegisterBuffer=(c,b,T,z)=>m.registerBuffer(c,b,T,z),t.jsepGetBuffer=c=>m.getBuffer(c),t.jsepCreateDownloader=(c,b,T)=>m.createDownloader(c,b,T),t.jsepOnCreateSession=c=>{m.onCreateSession(c)},t.jsepOnReleaseSession=c=>{m.onReleaseSession(c)},t.jsepOnRunStart=c=>m.onRunStart(c),t.Ld=(c,b)=>{m.upload(c,b)}}else if(o==="webnn"){let m=d[0];[t.Zd,t.vd,t.webnnEnsureTensor,t.xd,t.webnnDownloadTensor,t.Yd,t.webnnEnableTraceEvent]=d.slice(1),t.webnnReleaseTensorId=t.vd,t.webnnUploadTensor=t.xd,t.webnnRegisterMLContext=t.Yd,t.webnnOnRunStart=c=>m.onRunStart(c),t.webnnOnRunEnd=m.onRunEnd.bind(m),t.webnnOnReleaseSession=c=>{m.onReleaseSession(c)},t.webnnCreateMLTensorDownloader=(c,b)=>m.createMLTensorDownloader(c,b),t.webnnRegisterMLTensor=(c,b,T,z)=>m.registerMLTensor(c,b,T,z),t.webnnCreateMLContext=c=>m.createMLContext(c),t.webnnRegisterMLConstant=(c,b,T,z,R,W)=>m.registerMLConstant(c,b,T,z,R,t.Zc,W),t.webnnRegisterGraphInput=m.registerGraphInput.bind(m),t.webnnIsGraphInput=m.isGraphInput.bind(m),t.webnnRegisterGraphOutput=m.registerGraphOutput.bind(m),t.webnnIsGraphOutput=m.isGraphOutput.bind(m),t.webnnCreateTemporaryTensor=m.createTemporaryTensor.bind(m),t.webnnIsGraphInputOutputTypeSupported=m.isGraphInputOutputTypeSupported.bind(m)}};let n=()=>{let o=d=>(...m)=>{let c=Xe;return m=d(...m),Xe!=c?new Promise((b,T)=>{oi={resolve:b,reject:T}}):m};(()=>{for(let d of["_OrtAppendExecutionProvider","_OrtCreateSession","_OrtRun","_OrtRunWithBinding","_OrtBindInput"])t[d]=o(t[d])})(),s!==void 0&&(t._OrtRun=s(t._OrtRun),t._OrtRunWithBinding=s(t._OrtRunWithBinding)),n=void 0};t.asyncInit=()=>{n?.()};var u,l,p=(o,d)=>{throw d},h=import.meta.url,f="";if(r||i){try{f=new URL(".",h).href}catch{}i&&(l=o=>{var d=new XMLHttpRequest;return d.open("GET",o,!1),d.responseType="arraybuffer",d.send(null),new Uint8Array(d.response)}),u=async o=>{if(O(o))return new Promise((m,c)=>{var b=new XMLHttpRequest;b.open("GET",o,!0),b.responseType="arraybuffer",b.onload=()=>{b.status==200||b.status==0&&b.response?m(b.response):c(b.status)},b.onerror=c,b.send(null)});var d=await fetch(o,{credentials:"same-origin"});if(d.ok)return d.arrayBuffer();throw Error(d.status+" : "+d.url)}}var g,_,y,$,S,v,w=console.log.bind(console),k=console.error.bind(console),I=w,E=k,C=!1,O=o=>o.startsWith("file://");function x(){ut.buffer!=Q.buffer&&J()}if(a){let o=function(d){try{var m=d.data,c=m.Uc;if(c==="load"){let b=[];self.onmessage=T=>b.push(T),v=()=>{postMessage({Uc:"loaded"});for(let T of b)o(T);self.onmessage=o};for(let T of m.Ad)t[T]&&!t[T].proxy||(t[T]=(...z)=>{postMessage({Uc:"callHandler",zd:T,args:z})},T=="print"&&(I=t[T]),T=="printErr"&&(E=t[T]));ut=m.Vd,J(),_=m.Wd,Le(),Ir()}else if(c==="run"){(function(b){var T=(x(),U)[b+52>>>2>>>0];b=(x(),U)[b+56>>>2>>>0],ds(T,T-b),ae(T)})(m.Tc),ci(m.Tc,0,0,1,0,0),pn(),ai(m.Tc),L||(as(),L=!0);try{bf(m.Pd,m.dd)}catch(b){if(b!="unwind")throw b}}else m.target!=="setimmediate"&&(c==="checkMailbox"?L&&wr():c&&(E(`worker: received unknown command ${c}`),E(m)))}catch(b){throw ns(),b}};var D=o,L=!1;self.onunhandledrejection=d=>{throw d.reason||d},self.onmessage=o}var Q,q,K,se,A,U,ee,X,H,le,N,V=!1;function J(){var o=ut.buffer;t.HEAP8=Q=new Int8Array(o),K=new Int16Array(o),t.HEAPU8=q=new Uint8Array(o),se=new Uint16Array(o),t.HEAP32=A=new Int32Array(o),t.HEAPU32=U=new Uint32Array(o),ee=new Float32Array(o),X=new Float64Array(o),H=new BigInt64Array(o),le=new BigUint64Array(o)}function de(){V=!0,a?v():rt.tb()}function Ie(o){throw E(o="Aborted("+o+")"),C=!0,o=new WebAssembly.RuntimeError(o+". Build with -sASSERTIONS for more info."),S?.(o),o}function Be(){return{a:{ma:Fm,hb:Gm,g:$f,J:vf,f:xf,o:Sf,h:Tf,ha:If,b:kf,T:Ef,Ia:yn,n:zf,_:$n,Ya:vn,Ea:xn,Ga:Sn,Za:Tn,Wa:In,Pa:kn,Va:En,ka:zn,Fa:Cn,Ca:An,Xa:On,Da:Bn,cb:Cf,ea:Of,xa:Bf,va:Mf,da:Df,O:Pf,H:Uf,wa:Wf,Z:jf,ya:Kf,Sa:Zf,Aa:Xf,Ja:Yf,ta:Jf,fa:em,Ra:ai,$a:tm,R:nm,s:dm,c:ri,ib:pm,y:cm,M:hm,D:fm,m:mm,t:Ln,jb:gm,I:ym,S:_m,j:wm,v:bm,r:$m,l:vm,Ma:xm,Na:Sm,Oa:Tm,Ka:Fn,La:Hn,ua:jn,eb:km,bb:Cm,u:Am,aa:Om,ga:Bm,ab:Em,V:Rm,_a:Mm,Ba:Nm,F:Im,U:Dm,la:Sr,za:Um,gb:Pm,fb:Wm,Ta:Xn,Ua:Yn,Ha:Xr,$:Jn,ja:es,Qa:ts,ia:rs,lb:Eg,na:$g,mb:kg,oa:bg,G:pg,d:Zm,q:jm,w:Hm,B:ng,pb:yg,K:ug,x:Xm,pa:_g,X:vg,ba:gg,nb:Ig,ob:Tg,ra:cg,qa:mg,qb:hg,N:lg,Y:wg,e:Qm,A:Ym,k:Km,kb:zg,p:eg,z:tg,C:Jm,E:rg,L:sg,rb:dg,Q:xg,ca:og,W:Sg,sb:ag,sa:ig,P:fg,i:qm,a:ut,db:Qr}}}async function Le(){function o(c,b){var T=rt=c.exports;c={};for(let[z,R]of Object.entries(T))typeof R=="function"?(T=rm(R),c[z]=T):c[z]=R;return rt=c,rt=(function(){var z=rt,R=G=>ie=>G(ie)>>>0,W=G=>()=>G()>>>0;return(z=Object.assign({},z)).ub=R(z.ub),z.Yb=W(z.Yb),z._b=R(z._b),z.mc=R(z.mc),z.nc=W(z.nc),z.rc=R(z.rc),z})(),ln.push(rt.$b),is=(c=rt).ub,as=c.vb,t._OrtInit=c.wb,t._OrtGetLastError=c.xb,t._OrtCreateSessionOptions=c.yb,t._OrtAppendExecutionProvider=c.zb,t._OrtAddFreeDimensionOverride=c.Ab,t._OrtAddSessionConfigEntry=c.Bb,t._OrtReleaseSessionOptions=c.Cb,t._OrtCreateSession=c.Db,t._OrtReleaseSession=c.Eb,t._OrtGetInputOutputCount=c.Fb,t._OrtGetInputOutputMetadata=c.Gb,t._OrtFree=c.Hb,t._OrtCreateTensor=c.Ib,t._OrtGetTensorData=c.Jb,t._OrtReleaseTensor=c.Kb,t._OrtCreateRunOptions=c.Lb,t._OrtAddRunConfigEntry=c.Mb,t._OrtReleaseRunOptions=c.Nb,t._OrtCreateBinding=c.Ob,t._OrtBindInput=c.Pb,t._OrtBindOutput=c.Qb,t._OrtClearBoundOutputs=c.Rb,t._OrtReleaseBinding=c.Sb,t._OrtRunWithBinding=c.Tb,t._OrtRun=c.Ub,t._OrtEndProfiling=c.Vb,t._JsepOutput=c.Wb,t._JsepGetNodeName=c.Xb,Tr=c.Yb,Ye=t._free=c.Zb,jt=t._malloc=c._b,ci=c.bc,ns=c.cc,ss=c.dc,os=c.ec,hi=c.fc,us=c.gc,ls=c.hc,oe=c.ic,Kt=c.jc,ds=c.kc,ae=c.lc,fi=c.mc,ne=c.nc,ps=c.oc,mi=c.pc,cs=c.qc,hs=c.rc,fs=c.sc,gi=c.tc,ms=c.uc,gs=c.vc,ys=c.wc,_s=c.xc,ws=c.yc,bs=c.zc,$s=c.Ac,vs=c.Bc,xs=c.Cc,Ss=c.Dc,Ts=c.Ec,Is=c.Fc,ks=c.Gc,Es=c.Hc,zs=c.Ic,Cs=c.Jc,As=c.Kc,Os=c.Lc,Bs=c.Mc,Rs=c.Nc,Ms=c.Oc,Ns=c.Pc,Ds=c.Rc,Ps=c.Sc,Us=c.bd,Ws=c.cd,Ls=c.hd,qs=c.kd,Vs=c.ld,Gs=c.md,Fs=c.nd,Hs=c.od,js=c.pd,Ks=c.qd,Zs=c.rd,Qs=c.wd,Xs=c.Rd,Ys=c.Sd,Js=c.Td,eo=c.Ud,_=b,rt}var d,m=Be();return t.instantiateWasm?new Promise(c=>{t.instantiateWasm(m,(b,T)=>{c(o(b,T))})}):a?o(new WebAssembly.Instance(_,Be()),_):(N??=t.locateFile?t.locateFile?t.locateFile("ort-wasm-simd-threaded.jsep.wasm",f):f+"ort-wasm-simd-threaded.jsep.wasm":new URL("ort-wasm-simd-threaded.jsep.wasm",import.meta.url).href,d=await(async function(c){var b=N;if(!g&&!O(b))try{var T=fetch(b,{credentials:"same-origin"});return await WebAssembly.instantiateStreaming(T,c)}catch(z){E(`wasm streaming compile failed: ${z}`),E("falling back to ArrayBuffer instantiation")}return(async function(z,R){try{var W=await(async function(G){if(!g)try{var ie=await u(G);return new Uint8Array(ie)}catch{}if(G==N&&g)G=new Uint8Array(g);else{if(!l)throw"both async and sync fetching of the wasm failed";G=l(G)}return G})(z);return await WebAssembly.instantiate(W,R)}catch(G){E(`failed to asynchronously prepare wasm: ${G}`),Ie(G)}})(b,c)})(m),o(d.instance,d.module))}class et{name="ExitStatus";constructor(d){this.message=`Program terminated with exit(${d})`,this.status=d}}var xe=o=>{o.terminate(),o.onmessage=()=>{}},be=[],Oe=0,bt=null,fr=o=>{ot.length==0&&(hn(),cn(ot[0]));var d=ot.pop();if(!d)return 6;Ft.push(d),$t[o.Tc]=d,d.Tc=o.Tc;var m={Uc:"run",Pd:o.Od,dd:o.dd,Tc:o.Tc};return d.postMessage(m,o.ud),0},st=0,we=(o,d,...m)=>{var c,b=16*m.length,T=ne(),z=fi(b),R=z>>>3;for(c of m)typeof c=="bigint"?((x(),H)[R++>>>0]=1n,(x(),H)[R++>>>0]=c):((x(),H)[R++>>>0]=0n,(x(),X)[R++>>>0]=c);return o=ss(o,0,b,z,d),ae(T),o};function Qr(o){if(a)return we(0,1,o);if(y=o,!(0<st)){for(var d of Ft)xe(d);for(d of ot)xe(d);ot=[],Ft=[],$t={},C=!0}p(0,new et(o))}function un(o){if(a)return we(1,0,o);Xr(o)}var Xr=o=>{if(y=o,a)throw un(o),"unwind";Qr(o)},ot=[],Ft=[],ln=[],$t={},dn=o=>{var d=o.Tc;delete $t[d],ot.push(o),Ft.splice(Ft.indexOf(o),1),o.Tc=0,os(d)};function pn(){ln.forEach(o=>o())}var cn=o=>new Promise(d=>{o.onmessage=b=>{var T=b.data;if(b=T.Uc,T.ad&&T.ad!=Tr()){var z=$t[T.ad];z?z.postMessage(T,T.ud):E(`Internal error! Worker sent a message "${b}" to target pthread ${T.ad}, but that thread no longer exists!`)}else b==="checkMailbox"?wr():b==="spawnThread"?fr(T):b==="cleanupThread"?_r(()=>{dn($t[T.Qd])}):b==="loaded"?(o.loaded=!0,d(o)):T.target==="setimmediate"?o.postMessage(T):b==="uncaughtException"?o.onerror(T.error):b==="callHandler"?t[T.zd](...T.args):b&&E(`worker sent an unknown command ${b}`)},o.onerror=b=>{throw E(`worker sent an error! ${b.filename}:${b.lineno}: ${b.message}`),b};var m,c=[];for(m of[])t.propertyIsEnumerable(m)&&c.push(m);o.postMessage({Uc:"load",Ad:c,Vd:ut,Wd:_})});function hn(){var o=new Worker((()=>{let d=URL;return import.meta.url>"file:"&&import.meta.url<"file;"?new d("ort.bundle.min.mjs",import.meta.url):new URL(import.meta.url)})(),{type:"module",workerData:"em-pthread",name:"em-pthread"});ot.push(o)}var ut,bf=(o,d)=>{st=0,o=gi(o,d),0<st?y=o:hi(o)},mr=[],gr=0;function $f(o){var d=new Yr(o>>>=0);return(x(),Q)[d.Vc+12>>>0]==0&&(fn(d,!0),gr--),mn(d,!1),mr.push(d),hs(o)}var Pt=0,vf=()=>{oe(0,0);var o=mr.pop();ps(o.ed),Pt=0};function fn(o,d){d=d?1:0,(x(),Q)[o.Vc+12>>>0]=d}function mn(o,d){d=d?1:0,(x(),Q)[o.Vc+13>>>0]=d}class Yr{constructor(d){this.ed=d,this.Vc=d-24}}var Jr=o=>{var d=Pt;if(!d)return Kt(0),0;var m=new Yr(d);(x(),U)[m.Vc+16>>>2>>>0]=d;var c=(x(),U)[m.Vc+4>>>2>>>0];if(!c)return Kt(0),d;for(var b of o){if(b===0||b===c)break;if(cs(b,c,m.Vc+16))return Kt(b),d}return Kt(c),d};function xf(){return Jr([])}function Sf(o){return Jr([o>>>0])}function Tf(o,d,m,c){return Jr([o>>>0,d>>>0,m>>>0,c>>>0])}var If=()=>{var o=mr.pop();o||Ie("no exception to throw");var d=o.ed;throw(x(),Q)[o.Vc+13>>>0]==0&&(mr.push(o),mn(o,!0),fn(o,!1),gr++),mi(d),Pt=d};function kf(o,d,m){var c=new Yr(o>>>=0);throw d>>>=0,m>>>=0,(x(),U)[c.Vc+16>>>2>>>0]=0,(x(),U)[c.Vc+4>>>2>>>0]=d,(x(),U)[c.Vc+8>>>2>>>0]=m,mi(o),gr++,Pt=o}var Ef=()=>gr;function gn(o,d,m,c){return a?we(2,1,o,d,m,c):yn(o,d,m,c)}function yn(o,d,m,c){if(o>>>=0,d>>>=0,m>>>=0,c>>>=0,!globalThis.SharedArrayBuffer)return 6;var b=[];return a&&b.length===0?gn(o,d,m,c):(o={Od:m,Tc:o,dd:c,ud:b},a?(o.Uc="spawnThread",postMessage(o,b),0):fr(o))}function zf(o){throw Pt||=o>>>0,Pt}var _n=globalThis.TextDecoder&&new TextDecoder,wn=(o,d,m,c)=>{if(m=d+m,c)return m;for(;o[d]&&!(d>=m);)++d;return d},bn=(o,d=0,m,c)=>{if(16<(m=wn(o,d>>>=0,m,c))-d&&o.buffer&&_n)return _n.decode(o.buffer instanceof ArrayBuffer?o.subarray(d,m):o.slice(d,m));for(c="";d<m;){var b=o[d++];if(128&b){var T=63&o[d++];if((224&b)==192)c+=String.fromCharCode((31&b)<<6|T);else{var z=63&o[d++];65536>(b=(240&b)==224?(15&b)<<12|T<<6|z:(7&b)<<18|T<<12|z<<6|63&o[d++])?c+=String.fromCharCode(b):(b-=65536,c+=String.fromCharCode(55296|b>>10,56320|1023&b))}}else c+=String.fromCharCode(b)}return c},Se=(o,d,m)=>(o>>>=0)?bn((x(),q),o,d,m):"";function $n(o,d,m){return a?we(3,1,o,d,m):0}function vn(o,d){if(a)return we(4,1,o,d)}function xn(o,d){if(a)return we(5,1,o,d)}function Sn(o,d,m){if(a)return we(6,1,o,d,m)}function Tn(o,d,m){return a?we(7,1,o,d,m):0}function In(o,d){if(a)return we(8,1,o,d)}function kn(o,d,m){if(a)return we(9,1,o,d,m)}function En(o,d,m,c){if(a)return we(10,1,o,d,m,c)}function zn(o,d,m,c){if(a)return we(11,1,o,d,m,c)}function Cn(o,d,m,c){if(a)return we(12,1,o,d,m,c)}function An(o){if(a)return we(13,1,o)}function On(o,d){if(a)return we(14,1,o,d)}function Bn(o,d,m){if(a)return we(15,1,o,d,m)}var Cf=()=>Ie(""),Qe=o=>{o>>>=0;for(var d="";;){var m=(x(),q)[o++>>>0];if(!m)return d;d+=String.fromCharCode(m)}},ei={},ti={},Af={},Ut=class extends Error{constructor(o){super(o),this.name="BindingError"}};function tt(o,d,m={}){return(function(c,b,T={}){var z=b.name;if(!c)throw new Ut(`type "${z}" must have a positive integer typeid pointer`);if(ti.hasOwnProperty(c)){if(T.Bd)return;throw new Ut(`Cannot register type '${z}' twice`)}ti[c]=b,delete Af[c],ei.hasOwnProperty(c)&&(b=ei[c],delete ei[c],b.forEach(R=>R()))})(o,d,m)}var Rn=(o,d,m)=>{switch(d){case 1:return m?c=>(x(),Q)[c>>>0]:c=>(x(),q)[c>>>0];case 2:return m?c=>(x(),K)[c>>>1>>>0]:c=>(x(),se)[c>>>1>>>0];case 4:return m?c=>(x(),A)[c>>>2>>>0]:c=>(x(),U)[c>>>2>>>0];case 8:return m?c=>(x(),H)[c>>>3>>>0]:c=>(x(),le)[c>>>3>>>0];default:throw new TypeError(`invalid integer width (${d}): ${o}`)}};function Of(o,d,m,c,b){o>>>=0,m>>>=0,d=Qe(d>>>0);let T=z=>z;if(c=c===0n){let z=8*m;T=R=>BigInt.asUintN(z,R),b=T(b)}tt(o,{name:d,Qc:T,Xc:(z,R)=>(typeof R=="number"&&(R=BigInt(R)),R),Wc:Rn(d,m,!c),Yc:null})}function Bf(o,d,m,c){tt(o>>>=0,{name:d=Qe(d>>>0),Qc:function(b){return!!b},Xc:function(b,T){return T?m:c},Wc:function(b){return this.Qc((x(),q)[b>>>0])},Yc:null})}var Mn=[],vt=[0,1,,1,null,1,!0,1,!1,1];function ri(o){9<(o>>>=0)&&--vt[o+1]==0&&(vt[o]=void 0,Mn.push(o))}var De=o=>{if(!o)throw new Ut(`Cannot use deleted val. handle = ${o}`);return vt[o]},qe=o=>{switch(o){case void 0:return 2;case null:return 4;case!0:return 6;case!1:return 8;default:let d=Mn.pop()||vt.length;return vt[d]=o,vt[d+1]=1,d}};function ii(o){return this.Qc((x(),U)[o>>>2>>>0])}var Rf={name:"emscripten::val",Qc:o=>{var d=De(o);return ri(o),d},Xc:(o,d)=>qe(d),Wc:ii,Yc:null};function Mf(o){return tt(o>>>0,Rf)}var Nf=(o,d)=>{switch(d){case 4:return function(m){return this.Qc((x(),ee)[m>>>2>>>0])};case 8:return function(m){return this.Qc((x(),X)[m>>>3>>>0])};default:throw new TypeError(`invalid float width (${d}): ${o}`)}};function Df(o,d,m){m>>>=0,tt(o>>>=0,{name:d=Qe(d>>>0),Qc:c=>c,Xc:(c,b)=>b,Wc:Nf(d,m),Yc:null})}function Pf(o,d,m,c,b){o>>>=0,m>>>=0,d=Qe(d>>>0);let T=R=>R;if(c===0){var z=32-8*m;T=R=>R<<z>>>z,b=T(b)}tt(o,{name:d,Qc:T,Xc:(R,W)=>W,Wc:Rn(d,m,c!==0),Yc:null})}function Uf(o,d,m){function c(T){var z=(x(),U)[T>>>2>>>0];return T=(x(),U)[T+4>>>2>>>0],new b((x(),Q).buffer,T,z)}var b=[Int8Array,Uint8Array,Int16Array,Uint16Array,Int32Array,Uint32Array,Float32Array,Float64Array,BigInt64Array,BigUint64Array][d];tt(o>>>=0,{name:m=Qe(m>>>0),Qc:c,Wc:c},{Bd:!0})}var lt=(o,d,m)=>{var c=(x(),q);if(d>>>=0,0<m){var b=d;m=d+m-1;for(var T=0;T<o.length;++T){var z=o.codePointAt(T);if(127>=z){if(d>=m)break;c[d++>>>0]=z}else if(2047>=z){if(d+1>=m)break;c[d++>>>0]=192|z>>6,c[d++>>>0]=128|63&z}else if(65535>=z){if(d+2>=m)break;c[d++>>>0]=224|z>>12,c[d++>>>0]=128|z>>6&63,c[d++>>>0]=128|63&z}else{if(d+3>=m)break;c[d++>>>0]=240|z>>18,c[d++>>>0]=128|z>>12&63,c[d++>>>0]=128|z>>6&63,c[d++>>>0]=128|63&z,T++}}c[d>>>0]=0,o=d-b}else o=0;return o},yr=o=>{for(var d=0,m=0;m<o.length;++m){var c=o.charCodeAt(m);127>=c?d++:2047>=c?d+=2:55296<=c&&57343>=c?(d+=4,++m):d+=3}return d};function Wf(o,d){tt(o>>>=0,{name:d=Qe(d>>>0),Qc(m){var c=(x(),U)[m>>>2>>>0];return c=Se(m+4,c,!0),Ye(m),c},Xc(m,c){c instanceof ArrayBuffer&&(c=new Uint8Array(c));var b=typeof c=="string";if(!(b||ArrayBuffer.isView(c)&&c.BYTES_PER_ELEMENT==1))throw new Ut("Cannot pass non-string to std::string");var T=b?yr(c):c.length,z=jt(4+T+1),R=z+4;return(x(),U)[z>>>2>>>0]=T,b?lt(c,R,T+1):(x(),q).set(c,R>>>0),m!==null&&m.push(Ye,z),z},Wc:ii,Yc(m){Ye(m)}})}var Nn=globalThis.TextDecoder?new TextDecoder("utf-16le"):void 0,Lf=(o,d,m)=>{if(o>>>=1,16<(d=wn((x(),se),o,d/2,m))-o&&Nn)return Nn.decode((x(),se).slice(o,d));for(m="";o<d;++o){var c=(x(),se)[o>>>0];m+=String.fromCharCode(c)}return m},qf=(o,d,m)=>{if(m??=2147483647,2>m)return 0;var c=d;m=(m-=2)<2*o.length?m/2:o.length;for(var b=0;b<m;++b){var T=o.charCodeAt(b);(x(),K)[d>>>1>>>0]=T,d+=2}return(x(),K)[d>>>1>>>0]=0,d-c},Vf=o=>2*o.length,Gf=(o,d,m)=>{var c="";o>>>=2;for(var b=0;!(b>=d/4);b++){var T=(x(),U)[o+b>>>0];if(!T&&!m)break;c+=String.fromCodePoint(T)}return c},Ff=(o,d,m)=>{if(d>>>=0,m??=2147483647,4>m)return 0;var c=d;m=c+m-4;for(var b=0;b<o.length;++b){var T=o.codePointAt(b);if(65535<T&&b++,(x(),A)[d>>>2>>>0]=T,(d+=4)+4>m)break}return(x(),A)[d>>>2>>>0]=0,d-c},Hf=o=>{for(var d=0,m=0;m<o.length;++m)65535<o.codePointAt(m)&&m++,d+=4;return d};function jf(o,d,m){if(o>>>=0,d>>>=0,m=Qe(m>>>=0),d===2)var c=Lf,b=qf,T=Vf;else c=Gf,b=Ff,T=Hf;tt(o,{name:m,Qc:z=>{var R=(x(),U)[z>>>2>>>0];return R=c(z+4,R*d,!0),Ye(z),R},Xc:(z,R)=>{if(typeof R!="string")throw new Ut(`Cannot pass non-string to C++ string type ${m}`);var W=T(R),G=jt(4+W+d);return(x(),U)[G>>>2>>>0]=W/d,b(R,G+4,W+d),z!==null&&z.push(Ye,G),G},Wc:ii,Yc(z){Ye(z)}})}function Kf(o,d){tt(o>>>=0,{Cd:!0,name:d=Qe(d>>>0),Qc:()=>{},Xc:()=>{}})}function Zf(o){ci(o>>>0,!i,1,!r,131072,!1),pn()}var _r=o=>{if(!C)try{if(o(),!(0<st))try{a?Tr()&&hi(y):Xr(y)}catch(d){d instanceof et||d=="unwind"||p(0,d)}}catch(d){d instanceof et||d=="unwind"||p(0,d)}},Qf=!Atomics.waitAsync||globalThis.navigator?.userAgent&&91>Number((navigator.userAgent.match(/Chrom(e|ium)\/([0-9]+)\./)||[])[2]);function ai(o){o>>>=0,Qf||(Atomics.waitAsync((x(),A),o>>>2,o).value.then(wr),o+=128,Atomics.store((x(),A),o>>>2,1))}var wr=()=>_r(()=>{var o=Tr();o&&(ai(o),ls())});function Xf(o,d){(o>>>=0)==d>>>0?setTimeout(wr):a?postMessage({ad:o,Uc:"checkMailbox"}):(o=$t[o])&&o.postMessage({Uc:"checkMailbox"})}var ni=[];function Yf(o,d,m,c,b){for(d>>>=0,b>>>=0,ni.length=0,m=b>>>3,c=b+c>>>3;m<c;){var T;T=(x(),H)[m++>>>0]?(x(),H)[m++>>>0]:(x(),X)[m++>>>0],ni.push(T)}return(d?yi[d]:Vm[o])(...ni)}var Jf=()=>{st=0};function em(o){o>>>=0,a?postMessage({Uc:"cleanupThread",Qd:o}):dn($t[o])}function tm(o){}var br=o=>{try{o()}catch(d){Ie(d)}};function rm(o){var d=(...m)=>{$r.push(o);try{return o(...m)}finally{C||($r.pop(),Xe&&dt===1&&$r.length===0&&(dt=0,st+=1,br(Ys),typeof Fibers<"u"&&Fibers.ce()))}};return Un.set(o,d),d}var dt=0,Xe=null,Dn=0,$r=[],si=new Map,Pn=new Map,Un=new Map,im=0,oi=null,am=[],Wn=o=>(function(d){if(!C){if(dt===0){var m=!1,c=!1;d((b=0)=>{if(!C&&(Dn=b,m=!0,c)){dt=2,br(()=>Js(Xe)),typeof MainLoop<"u"&&MainLoop.yd&&MainLoop.resume(),b=!1;try{var T=(function(){var W=(x(),A)[Xe+8>>>2>>>0];return W=Pn.get(W),W=Un.get(W),--st,W()})()}catch(W){T=W,b=!0}var z=!1;if(!Xe){var R=oi;R&&(oi=null,(b?R.reject:R.resolve)(T),z=!0)}if(b&&!z)throw T}}),c=!0,m||(dt=1,Xe=(function(){var b=jt(65548),T=b+12;if((x(),U)[b>>>2>>>0]=T,(x(),U)[b+4>>>2>>>0]=T+65536,T=$r[0],!si.has(T)){var z=im++;si.set(T,z),Pn.set(z,T)}return T=si.get(T),(x(),A)[b+8>>>2>>>0]=T,b})(),typeof MainLoop<"u"&&MainLoop.yd&&MainLoop.pause(),br(()=>Xs(Xe)))}else dt===2?(dt=0,br(eo),Ye(Xe),Xe=null,am.forEach(_r)):Ie(`invalid state: ${dt}`);return Dn}})(d=>{o().then(d)});function nm(o){return o>>>=0,Wn(async()=>{var d=await De(o);return qe(d)})}var ui=[],sm=o=>{var d=ui.length;return ui.push(o),d},om=(o,d)=>{for(var m=Array(o),c=0;c<o;++c){var b=c,T=(x(),U)[d+4*c>>>2>>>0],z=ti[T];if(z===void 0)throw o=`parameter ${c}`,T=is(T),d=Qe(T),Ye(T),new Ut(`${o} has unknown type ${d}`);m[b]=z}return m},um=(o,d,m)=>{var c=[];return o=o(c,m),c.length&&((x(),U)[d>>>2>>>0]=qe(c)),o},lm={},vr=o=>{var d=lm[o];return d===void 0?Qe(o):d};function dm(o,d,m){var[c,...b]=om(o,d>>>0);d=c.Xc.bind(c);var T=b.map(W=>W.Wc.bind(W));o--;var z={toValue:De};switch(o=T.map((W,G)=>{var ie=`argFromPtr${G}`;return z[ie]=W,`${ie}(args${G?"+"+8*G:""})`}),m){case 0:var R="toValue(handle)";break;case 2:R="new (toValue(handle))";break;case 3:R="";break;case 1:z.getStringOrSymbol=vr,R="toValue(handle)[getStringOrSymbol(methodName)]"}return R+=`(${o})`,c.Cd||(z.toReturnWire=d,z.emval_returnValue=um,R=`return emval_returnValue(toReturnWire, destructorsRef, ${R})`),R=`return function (handle, methodName, destructorsRef, args) {
   ${R}
-  }`,m=new Function(Object.keys(z),R)(...Object.values(z)),R=`methodCaller<(${b.map(W=>W.name)}) => ${c.name}>`,sm(Object.defineProperty(m,"name",{value:R}))}function pm(o,d){return d>>>=0,(o=De(o>>>0))==De(d)}function cm(o){return(o>>>=0)?(o=vr(o),Le(globalThis[o])):Le(globalThis)}function hm(o){return o=vr(o>>>0),Le(t[o])}function fm(o,d){return d>>>=0,o=De(o>>>0),d=De(d),Le(o[d])}function mm(o){9<(o>>>=0)&&(vt[o+1]+=1)}function qn(o,d,m,c,b){return ui[o>>>0](d>>>0,m>>>0,c>>>0,b>>>0)}function gm(o,d,m,c,b){return qn(o>>>0,d>>>0,m>>>0,c>>>0,b>>>0)}function ym(){return Le([])}function _m(o){o=De(o>>>0);for(var d=Array(o.length),m=0;m<o.length;m++)d[m]=o[m];return Le(d)}function wm(o){return Le(vr(o>>>0))}function bm(){return Le({})}function $m(o){for(var d=De(o>>>=0);d.length;){var m=d.pop();d.pop()(m)}ri(o)}function vm(o,d,m){d>>>=0,m>>>=0,o=De(o>>>0),d=De(d),m=De(m),o[d]=m}function xm(o,d){o=-9007199254740992>o||9007199254740992<o?NaN:Number(o),d>>>=0,o=new Date(1e3*o),(x(),A)[d>>>2>>>0]=o.getUTCSeconds(),(x(),A)[d+4>>>2>>>0]=o.getUTCMinutes(),(x(),A)[d+8>>>2>>>0]=o.getUTCHours(),(x(),A)[d+12>>>2>>>0]=o.getUTCDate(),(x(),A)[d+16>>>2>>>0]=o.getUTCMonth(),(x(),A)[d+20>>>2>>>0]=o.getUTCFullYear()-1900,(x(),A)[d+24>>>2>>>0]=o.getUTCDay(),o=(o.getTime()-Date.UTC(o.getUTCFullYear(),0,1,0,0,0,0))/864e5|0,(x(),A)[d+28>>>2>>>0]=o}var Ln=o=>o%4==0&&(o%100!=0||o%400==0),Vn=[0,31,60,91,121,152,182,213,244,274,305,335],Gn=[0,31,59,90,120,151,181,212,243,273,304,334];function Sm(o,d){o=-9007199254740992>o||9007199254740992<o?NaN:Number(o),d>>>=0,o=new Date(1e3*o),(x(),A)[d>>>2>>>0]=o.getSeconds(),(x(),A)[d+4>>>2>>>0]=o.getMinutes(),(x(),A)[d+8>>>2>>>0]=o.getHours(),(x(),A)[d+12>>>2>>>0]=o.getDate(),(x(),A)[d+16>>>2>>>0]=o.getMonth(),(x(),A)[d+20>>>2>>>0]=o.getFullYear()-1900,(x(),A)[d+24>>>2>>>0]=o.getDay();var m=(Ln(o.getFullYear())?Vn:Gn)[o.getMonth()]+o.getDate()-1|0;(x(),A)[d+28>>>2>>>0]=m,(x(),A)[d+36>>>2>>>0]=-60*o.getTimezoneOffset(),m=new Date(o.getFullYear(),6,1).getTimezoneOffset();var c=new Date(o.getFullYear(),0,1).getTimezoneOffset();o=0|(m!=c&&o.getTimezoneOffset()==Math.min(c,m)),(x(),A)[d+32>>>2>>>0]=o}function Tm(o){o>>>=0;var d=new Date((x(),A)[o+20>>>2>>>0]+1900,(x(),A)[o+16>>>2>>>0],(x(),A)[o+12>>>2>>>0],(x(),A)[o+8>>>2>>>0],(x(),A)[o+4>>>2>>>0],(x(),A)[o>>>2>>>0],0),m=(x(),A)[o+32>>>2>>>0],c=d.getTimezoneOffset(),b=new Date(d.getFullYear(),6,1).getTimezoneOffset(),T=new Date(d.getFullYear(),0,1).getTimezoneOffset(),z=Math.min(T,b);return 0>m?(x(),A)[o+32>>>2>>>0]=+(b!=T&&z==c):0<m!=(z==c)&&(b=Math.max(T,b),d.setTime(d.getTime()+6e4*((0<m?z:b)-c))),(x(),A)[o+24>>>2>>>0]=d.getDay(),m=(Ln(d.getFullYear())?Vn:Gn)[d.getMonth()]+d.getDate()-1|0,(x(),A)[o+28>>>2>>>0]=m,(x(),A)[o>>>2>>>0]=d.getSeconds(),(x(),A)[o+4>>>2>>>0]=d.getMinutes(),(x(),A)[o+8>>>2>>>0]=d.getHours(),(x(),A)[o+12>>>2>>>0]=d.getDate(),(x(),A)[o+16>>>2>>>0]=d.getMonth(),(x(),A)[o+20>>>2>>>0]=d.getYear(),o=d.getTime(),BigInt(isNaN(o)?-1:o/1e3)}function Fn(o,d,m,c,b,T,z){return a?we(16,1,o,d,m,c,b,T,z):-52}function Hn(o,d,m,c,b,T){if(a)return we(17,1,o,d,m,c,b,T)}var Ht={},Im=()=>performance.timeOrigin+performance.now();function jn(o,d){if(a)return we(18,1,o,d);if(Ht[o]&&(clearTimeout(Ht[o].id),delete Ht[o]),!d)return 0;var m=setTimeout(()=>{delete Ht[o],_r(()=>us(o,performance.timeOrigin+performance.now()))},d);return Ht[o]={id:m,be:d},0}function Em(o,d,m,c){o>>>=0,d>>>=0,m>>>=0,c>>>=0;var b=new Date().getFullYear(),T=new Date(b,0,1).getTimezoneOffset();b=new Date(b,6,1).getTimezoneOffset();var z=Math.max(T,b);(x(),U)[o>>>2>>>0]=60*z,(x(),A)[d>>>2>>>0]=+(T!=b),o=(d=R=>{var W=Math.abs(R);return`UTC${0<=R?"-":"+"}${String(Math.floor(W/60)).padStart(2,"0")}${String(W%60).padStart(2,"0")}`})(T),d=d(b),b<T?(lt(o,m,17),lt(d,c,17)):(lt(o,c,17),lt(d,m,17))}var km=()=>Date.now(),zm=1;function Cm(o,d,m){if(m>>>=0,!(0<=o&&3>=o))return 28;if(o===0)o=Date.now();else{if(!zm)return 52;o=performance.timeOrigin+performance.now()}return o=Math.round(1e6*o),(x(),H)[m>>>3>>>0]=BigInt(o),0}var li=[],Kn=(o,d)=>{li.length=0;for(var m;m=(x(),L)[o++>>>0];){var c=m!=105;d+=(c&=m!=112)&&d%8?4:0,li.push(m==112?(x(),U)[d>>>2>>>0]:m==106?(x(),H)[d>>>3>>>0]:m==105?(x(),A)[d>>>2>>>0]:(x(),X)[d>>>3>>>0]),d+=c?8:4}return li};function Am(o,d,m){return o>>>=0,d=Kn(d>>>0,m>>>0),yi[o](...d)}function Om(o,d,m){return o>>>=0,d=Kn(d>>>0,m>>>0),yi[o](...d)}var Bm=()=>{};function Rm(o,d){return k(Se(o>>>0,d>>>0))}var Nm=()=>{throw st+=1,"unwind"};function Mm(){return 4294901760}var Dm=()=>navigator.hardwareConcurrency,xt={},xr=o=>{var d;return(d=/\bwasm-function\[\d+\]:(0x[0-9a-f]+)/.exec(o))?+d[1]:(d=/:(\d+):\d+(?:\)|$)/.exec(o))?2147483648|+d[1]:0},Zn=o=>{for(var d of o)(o=xr(d))&&(xt[o]=d)};function Pm(){var o=Error().stack.toString().split(`
-`);return o[0]=="Error"&&o.shift(),Zn(o),xt.sd=xr(o[3]),xt.Md=o,xt.sd}function Sr(o){if(!(o=xt[o>>>0]))return 0;var d;if(d=/^\s+at .*\.wasm\.(.*) \(.*\)$/.exec(o))o=d[1];else if(d=/^\s+at (.*) \(.*\)$/.exec(o))o=d[1];else{if(!(d=/^(.+?)@/.exec(o)))return 0;o=d[1]}Ye(Sr.td??0),d=yr(o)+1;var m=jt(d);return m&&lt(o,m,d),Sr.td=m,Sr.td}function Um(o){o>>>=0;var d=(x(),L).length;if(o<=d||4294901760<o)return!1;for(var m=1;4>=m;m*=2){var c=d*(1+.2/m);c=Math.min(c,o+100663296);e:{c=(Math.min(4294901760,65536*Math.ceil(Math.max(o,c)/65536))-ut.buffer.byteLength+65535)/65536|0;try{ut.grow(c),J();var b=1;break e}catch{}b=void 0}if(b)return!0}return!1}function Wm(o,d,m){if(o>>>=0,d>>>=0,xt.sd==o)var c=xt.Md;else(c=Error().stack.toString().split(`
-`))[0]=="Error"&&c.shift(),Zn(c);for(var b=3;c[b]&&xr(c[b])!=o;)++b;for(o=0;o<m&&c[o+b];++o)(x(),A)[d+4*o>>>2>>>0]=xr(c[o+b]);return o}var di,pi={},Qn=()=>{if(!di){var o,d={USER:"web_user",LOGNAME:"web_user",PATH:"/",PWD:"/",HOME:"/home/web_user",LANG:(globalThis.navigator?.language??"C").replace("-","_")+".UTF-8",_:"./this.program"};for(o in pi)pi[o]===void 0?delete d[o]:d[o]=pi[o];var m=[];for(o in d)m.push(`${o}=${d[o]}`);di=m}return di};function Xn(o,d){if(a)return we(19,1,o,d);o>>>=0,d>>>=0;var m,c=0,b=0;for(m of Qn()){var T=d+c;(x(),U)[o+b>>>2>>>0]=T,c+=lt(m,T,1/0)+1,b+=4}return 0}function Yn(o,d){if(a)return we(20,1,o,d);o>>>=0,d>>>=0;var m=Qn();for(var c of((x(),U)[o>>>2>>>0]=m.length,o=0,m))o+=yr(c)+1;return(x(),U)[d>>>2>>>0]=o,0}function Jn(o){return a?we(21,1,o):52}function es(o,d,m,c){return a?we(22,1,o,d,m,c):52}function ts(o,d,m,c){return a?we(23,1,o,d,m,c):70}var qm=[null,[],[]];function rs(o,d,m,c){if(a)return we(24,1,o,d,m,c);d>>>=0,m>>>=0,c>>>=0;for(var b=0,T=0;T<m;T++){var z=(x(),U)[d>>>2>>>0],R=(x(),U)[d+4>>>2>>>0];d+=8;for(var W=0;W<R;W++){var G=o,ie=(x(),L)[z+W>>>0],pe=qm[G];ie===0||ie===10?((G===1?I:k)(bn(pe)),pe.length=0):pe.push(ie)}b+=R}return(x(),U)[c>>>2>>>0]=b,0}function Lm(o){return o>>>0}a||(function(){for(var o=t.numThreads-1;o--;)hn();be.push(async()=>{var d=(async function(){if(!a)return Promise.all(ot.map(cn))})();Oe++,await d,--Oe==0&&bt&&(d=bt,bt=null,d())})})(),a||(ut=new WebAssembly.Memory({initial:256,maximum:65536,shared:!0}),J()),t.wasmBinary&&(g=t.wasmBinary),t.stackSave=()=>ne(),t.stackRestore=o=>ae(o),t.stackAlloc=o=>fi(o),t.setValue=function(o,d,m="i8"){switch(m.endsWith("*")&&(m="*"),m){case"i1":case"i8":(x(),Q)[o>>>0]=d;break;case"i16":(x(),K)[o>>>1>>>0]=d;break;case"i32":(x(),A)[o>>>2>>>0]=d;break;case"i64":(x(),H)[o>>>3>>>0]=BigInt(d);break;case"float":(x(),ee)[o>>>2>>>0]=d;break;case"double":(x(),X)[o>>>3>>>0]=d;break;case"*":(x(),U)[o>>>2>>>0]=d;break;default:Te(`invalid type for setValue: ${m}`)}},t.getValue=function(o,d="i8"){switch(d.endsWith("*")&&(d="*"),d){case"i1":case"i8":return(x(),Q)[o>>>0];case"i16":return(x(),K)[o>>>1>>>0];case"i32":return(x(),A)[o>>>2>>>0];case"i64":return(x(),H)[o>>>3>>>0];case"float":return(x(),ee)[o>>>2>>>0];case"double":return(x(),X)[o>>>3>>>0];case"*":return(x(),U)[o>>>2>>>0];default:Te(`invalid type for getValue: ${d}`)}},t.UTF8ToString=Se,t.stringToUTF8=lt,t.lengthBytesUTF8=yr;var is,as,Tr,Ye,jt,ci,ns,ss,os,hi,us,ls,oe,Kt,ds,ae,fi,ne,ps,mi,cs,hs,fs,gi,ms,gs,ys,_s,ws,bs,$s,vs,xs,Ss,Ts,Is,Es,ks,zs,Cs,As,Os,Bs,Rs,Ns,Ms,Ds,Ps,Us,Ws,qs,Ls,Vs,Gs,Fs,Hs,js,Ks,Zs,Qs,Xs,Ys,Js,eo,rt,Vm=[Qr,un,gn,$n,vn,xn,Sn,Tn,In,En,kn,zn,Cn,An,On,Bn,Fn,Hn,jn,Xn,Yn,Jn,es,ts,rs],yi={927244:(o,d,m,c,b)=>{if(t===void 0||!t.Zc)return 1;if((o=Se(Number(o>>>0))).startsWith("./")&&(o=o.substring(2)),!(o=t.Zc.get(o)))return 2;if(d=Number(d>>>0),m=Number(m>>>0),c=Number(c>>>0),d+m>o.byteLength)return 3;try{let T=o.subarray(d,d+m);switch(b){case 0:(x(),L).set(T,c>>>0);break;case 1:t.Xd?t.Xd(c,T):t.Ld(c,T);break;default:return 4}return 0}catch{return 4}},928068:(o,d,m)=>{t.xd(o,(x(),L).subarray(d>>>0,d+m>>>0))},928132:()=>t.Zd(),928174:o=>{t.vd(o)},928211:()=>{t.Ed()},928242:()=>{t.Fd()},928271:()=>{t.Jd()},928296:o=>t.Dd(o),928329:o=>t.Hd(o),928361:(o,d,m)=>{t.jd(Number(o),Number(d),Number(m),!0)},928424:(o,d,m)=>{t.jd(Number(o),Number(d),Number(m))},928481:()=>typeof wasmOffsetConverter<"u",928538:o=>{t.ac("Abs",o,void 0)},928589:o=>{t.ac("Neg",o,void 0)},928640:o=>{t.ac("Floor",o,void 0)},928693:o=>{t.ac("Ceil",o,void 0)},928745:o=>{t.ac("Reciprocal",o,void 0)},928803:o=>{t.ac("Sqrt",o,void 0)},928855:o=>{t.ac("Exp",o,void 0)},928906:o=>{t.ac("Erf",o,void 0)},928957:o=>{t.ac("Sigmoid",o,void 0)},929012:(o,d,m)=>{t.ac("HardSigmoid",o,{alpha:d,beta:m})},929091:o=>{t.ac("Log",o,void 0)},929142:o=>{t.ac("Sin",o,void 0)},929193:o=>{t.ac("Cos",o,void 0)},929244:o=>{t.ac("Tan",o,void 0)},929295:o=>{t.ac("Asin",o,void 0)},929347:o=>{t.ac("Acos",o,void 0)},929399:o=>{t.ac("Atan",o,void 0)},929451:o=>{t.ac("Sinh",o,void 0)},929503:o=>{t.ac("Cosh",o,void 0)},929555:o=>{t.ac("Asinh",o,void 0)},929608:o=>{t.ac("Acosh",o,void 0)},929661:o=>{t.ac("Atanh",o,void 0)},929714:o=>{t.ac("Tanh",o,void 0)},929766:o=>{t.ac("Not",o,void 0)},929817:(o,d,m)=>{t.ac("Clip",o,{min:d,max:m})},929886:o=>{t.ac("Clip",o,void 0)},929938:(o,d)=>{t.ac("Elu",o,{alpha:d})},929996:o=>{t.ac("Gelu",o,void 0)},930048:o=>{t.ac("Relu",o,void 0)},930100:(o,d)=>{t.ac("LeakyRelu",o,{alpha:d})},930164:(o,d)=>{t.ac("ThresholdedRelu",o,{alpha:d})},930234:(o,d)=>{t.ac("Cast",o,{to:d})},930292:o=>{t.ac("Add",o,void 0)},930343:o=>{t.ac("Sub",o,void 0)},930394:o=>{t.ac("Mul",o,void 0)},930445:o=>{t.ac("Div",o,void 0)},930496:o=>{t.ac("Pow",o,void 0)},930547:o=>{t.ac("Equal",o,void 0)},930600:o=>{t.ac("Greater",o,void 0)},930655:o=>{t.ac("GreaterOrEqual",o,void 0)},930717:o=>{t.ac("Less",o,void 0)},930769:o=>{t.ac("LessOrEqual",o,void 0)},930828:(o,d,m,c,b)=>{t.ac("ReduceMean",o,{keepDims:!!d,noopWithEmptyAxes:!!m,axes:c?Array.from((x(),A).subarray(Number(c)>>>0,Number(b)>>>0)):[]})},931003:(o,d,m,c,b)=>{t.ac("ReduceMax",o,{keepDims:!!d,noopWithEmptyAxes:!!m,axes:c?Array.from((x(),A).subarray(Number(c)>>>0,Number(b)>>>0)):[]})},931177:(o,d,m,c,b)=>{t.ac("ReduceMin",o,{keepDims:!!d,noopWithEmptyAxes:!!m,axes:c?Array.from((x(),A).subarray(Number(c)>>>0,Number(b)>>>0)):[]})},931351:(o,d,m,c,b)=>{t.ac("ReduceProd",o,{keepDims:!!d,noopWithEmptyAxes:!!m,axes:c?Array.from((x(),A).subarray(Number(c)>>>0,Number(b)>>>0)):[]})},931526:(o,d,m,c,b)=>{t.ac("ReduceSum",o,{keepDims:!!d,noopWithEmptyAxes:!!m,axes:c?Array.from((x(),A).subarray(Number(c)>>>0,Number(b)>>>0)):[]})},931700:(o,d,m,c,b)=>{t.ac("ReduceL1",o,{keepDims:!!d,noopWithEmptyAxes:!!m,axes:c?Array.from((x(),A).subarray(Number(c)>>>0,Number(b)>>>0)):[]})},931873:(o,d,m,c,b)=>{t.ac("ReduceL2",o,{keepDims:!!d,noopWithEmptyAxes:!!m,axes:c?Array.from((x(),A).subarray(Number(c)>>>0,Number(b)>>>0)):[]})},932046:(o,d,m,c,b)=>{t.ac("ReduceLogSum",o,{keepDims:!!d,noopWithEmptyAxes:!!m,axes:c?Array.from((x(),A).subarray(Number(c)>>>0,Number(b)>>>0)):[]})},932223:(o,d,m,c,b)=>{t.ac("ReduceSumSquare",o,{keepDims:!!d,noopWithEmptyAxes:!!m,axes:c?Array.from((x(),A).subarray(Number(c)>>>0,Number(b)>>>0)):[]})},932403:(o,d,m,c,b)=>{t.ac("ReduceLogSumExp",o,{keepDims:!!d,noopWithEmptyAxes:!!m,axes:c?Array.from((x(),A).subarray(Number(c)>>>0,Number(b)>>>0)):[]})},932583:o=>{t.ac("Where",o,void 0)},932636:(o,d,m)=>{t.ac("Transpose",o,{perm:d?Array.from((x(),A).subarray(Number(d)>>>0,Number(m)>>>0)):[]})},932760:(o,d,m,c)=>{t.ac("DepthToSpace",o,{blocksize:d,mode:Se(m),format:c?"NHWC":"NCHW"})},932893:(o,d,m,c)=>{t.ac("DepthToSpace",o,{blocksize:d,mode:Se(m),format:c?"NHWC":"NCHW"})},933026:(o,d,m,c,b,T,z,R,W,G,ie,pe,me,_e,pt)=>{t.ac("ConvTranspose",o,{format:W?"NHWC":"NCHW",autoPad:d,dilations:[m],group:c,kernelShape:[b],pads:[T,z],strides:[R],wIsConst:()=>!!(x(),Q)[G>>>0],outputPadding:ie?Array.from((x(),A).subarray(Number(ie)>>>0,Number(pe)>>>0)):[],outputShape:me?Array.from((x(),A).subarray(Number(me)>>>0,Number(_e)>>>0)):[],activation:Se(pt)})},933459:(o,d,m,c,b,T,z,R,W,G,ie,pe,me,_e)=>{t.ac("ConvTranspose",o,{format:R?"NHWC":"NCHW",autoPad:d,dilations:Array.from((x(),A).subarray(Number(m)>>>0,2+(Number(m)>>>0)>>>0)),group:c,kernelShape:Array.from((x(),A).subarray(Number(b)>>>0,2+(Number(b)>>>0)>>>0)),pads:Array.from((x(),A).subarray(Number(T)>>>0,4+(Number(T)>>>0)>>>0)),strides:Array.from((x(),A).subarray(Number(z)>>>0,2+(Number(z)>>>0)>>>0)),wIsConst:()=>!!(x(),Q)[W>>>0],outputPadding:G?Array.from((x(),A).subarray(Number(G)>>>0,Number(ie)>>>0)):[],outputShape:pe?Array.from((x(),A).subarray(Number(pe)>>>0,Number(me)>>>0)):[],activation:Se(_e)})},934120:(o,d,m,c,b,T,z,R,W,G,ie,pe,me,_e,pt)=>{t.ac("ConvTranspose",o,{format:W?"NHWC":"NCHW",autoPad:d,dilations:[m],group:c,kernelShape:[b],pads:[T,z],strides:[R],wIsConst:()=>!!(x(),Q)[G>>>0],outputPadding:ie?Array.from((x(),A).subarray(Number(ie)>>>0,Number(pe)>>>0)):[],outputShape:me?Array.from((x(),A).subarray(Number(me)>>>0,Number(_e)>>>0)):[],activation:Se(pt)})},934553:(o,d,m,c,b,T,z,R,W,G,ie,pe,me,_e)=>{t.ac("ConvTranspose",o,{format:R?"NHWC":"NCHW",autoPad:d,dilations:Array.from((x(),A).subarray(Number(m)>>>0,2+(Number(m)>>>0)>>>0)),group:c,kernelShape:Array.from((x(),A).subarray(Number(b)>>>0,2+(Number(b)>>>0)>>>0)),pads:Array.from((x(),A).subarray(Number(T)>>>0,4+(Number(T)>>>0)>>>0)),strides:Array.from((x(),A).subarray(Number(z)>>>0,2+(Number(z)>>>0)>>>0)),wIsConst:()=>!!(x(),Q)[W>>>0],outputPadding:G?Array.from((x(),A).subarray(Number(G)>>>0,Number(ie)>>>0)):[],outputShape:pe?Array.from((x(),A).subarray(Number(pe)>>>0,Number(me)>>>0)):[],activation:Se(_e)})},935214:(o,d)=>{t.ac("GlobalAveragePool",o,{format:d?"NHWC":"NCHW"})},935305:(o,d,m,c,b,T,z,R,W,G,ie,pe,me,_e)=>{t.ac("AveragePool",o,{format:_e?"NHWC":"NCHW",auto_pad:d,ceil_mode:m,count_include_pad:c,storage_order:b,dilations:T?Array.from((x(),A).subarray(Number(T)>>>0,Number(z)>>>0)):[],kernel_shape:R?Array.from((x(),A).subarray(Number(R)>>>0,Number(W)>>>0)):[],pads:G?Array.from((x(),A).subarray(Number(G)>>>0,Number(ie)>>>0)):[],strides:pe?Array.from((x(),A).subarray(Number(pe)>>>0,Number(me)>>>0)):[]})},935784:(o,d)=>{t.ac("GlobalAveragePool",o,{format:d?"NHWC":"NCHW"})},935875:(o,d,m,c,b,T,z,R,W,G,ie,pe,me,_e)=>{t.ac("AveragePool",o,{format:_e?"NHWC":"NCHW",auto_pad:d,ceil_mode:m,count_include_pad:c,storage_order:b,dilations:T?Array.from((x(),A).subarray(Number(T)>>>0,Number(z)>>>0)):[],kernel_shape:R?Array.from((x(),A).subarray(Number(R)>>>0,Number(W)>>>0)):[],pads:G?Array.from((x(),A).subarray(Number(G)>>>0,Number(ie)>>>0)):[],strides:pe?Array.from((x(),A).subarray(Number(pe)>>>0,Number(me)>>>0)):[]})},936354:(o,d)=>{t.ac("GlobalMaxPool",o,{format:d?"NHWC":"NCHW"})},936441:(o,d,m,c,b,T,z,R,W,G,ie,pe,me,_e)=>{t.ac("MaxPool",o,{format:_e?"NHWC":"NCHW",auto_pad:d,ceil_mode:m,count_include_pad:c,storage_order:b,dilations:T?Array.from((x(),A).subarray(Number(T)>>>0,Number(z)>>>0)):[],kernel_shape:R?Array.from((x(),A).subarray(Number(R)>>>0,Number(W)>>>0)):[],pads:G?Array.from((x(),A).subarray(Number(G)>>>0,Number(ie)>>>0)):[],strides:pe?Array.from((x(),A).subarray(Number(pe)>>>0,Number(me)>>>0)):[]})},936916:(o,d)=>{t.ac("GlobalMaxPool",o,{format:d?"NHWC":"NCHW"})},937003:(o,d,m,c,b,T,z,R,W,G,ie,pe,me,_e)=>{t.ac("MaxPool",o,{format:_e?"NHWC":"NCHW",auto_pad:d,ceil_mode:m,count_include_pad:c,storage_order:b,dilations:T?Array.from((x(),A).subarray(Number(T)>>>0,Number(z)>>>0)):[],kernel_shape:R?Array.from((x(),A).subarray(Number(R)>>>0,Number(W)>>>0)):[],pads:G?Array.from((x(),A).subarray(Number(G)>>>0,Number(ie)>>>0)):[],strides:pe?Array.from((x(),A).subarray(Number(pe)>>>0,Number(me)>>>0)):[]})},937478:(o,d,m,c,b)=>{t.ac("Gemm",o,{alpha:d,beta:m,transA:c,transB:b})},937582:o=>{t.ac("MatMul",o,void 0)},937636:(o,d,m,c)=>{t.ac("ArgMax",o,{keepDims:!!d,selectLastIndex:!!m,axis:c})},937744:(o,d,m,c)=>{t.ac("ArgMin",o,{keepDims:!!d,selectLastIndex:!!m,axis:c})},937852:(o,d)=>{t.ac("Softmax",o,{axis:d})},937915:(o,d)=>{t.ac("Concat",o,{axis:d})},937975:(o,d,m,c,b)=>{t.ac("Split",o,{axis:d,numOutputs:m,splitSizes:c?Array.from((x(),A).subarray(Number(c)>>>0,Number(b)>>>0)):[]})},938131:o=>{t.ac("Expand",o,void 0)},938185:(o,d)=>{t.ac("Gather",o,{axis:Number(d)})},938256:(o,d)=>{t.ac("GatherElements",o,{axis:Number(d)})},938335:(o,d)=>{t.ac("GatherND",o,{batch_dims:Number(d)})},938414:(o,d,m,c,b,T,z,R,W,G,ie)=>{t.ac("Resize",o,{antialias:d,axes:m?Array.from((x(),A).subarray(Number(m)>>>0,Number(c)>>>0)):[],coordinateTransformMode:Se(b),cubicCoeffA:T,excludeOutside:z,extrapolationValue:R,keepAspectRatioPolicy:Se(W),mode:Se(G),nearestMode:Se(ie)})},938776:(o,d,m,c,b,T,z)=>{t.ac("Slice",o,{starts:d?Array.from((x(),A).subarray(Number(d)>>>0,Number(m)>>>0)):[],ends:c?Array.from((x(),A).subarray(Number(c)>>>0,Number(b)>>>0)):[],axes:T?Array.from((x(),A).subarray(Number(T)>>>0,Number(z)>>>0)):[]})},939040:o=>{t.ac("Tile",o,void 0)},939092:(o,d,m)=>{t.ac("InstanceNormalization",o,{epsilon:d,format:m?"NHWC":"NCHW"})},939206:(o,d,m)=>{t.ac("InstanceNormalization",o,{epsilon:d,format:m?"NHWC":"NCHW"})},939320:o=>{t.ac("Range",o,void 0)},939373:(o,d)=>{t.ac("Einsum",o,{equation:Se(d)})},939454:(o,d,m,c,b)=>{t.ac("Pad",o,{mode:d,value:m,pads:c?Array.from((x(),A).subarray(Number(c)>>>0,Number(b)>>>0)):[]})},939597:(o,d,m,c,b,T)=>{t.ac("BatchNormalization",o,{epsilon:d,momentum:m,spatial:!!b,trainingMode:!!c,format:T?"NHWC":"NCHW"})},939766:(o,d,m,c,b,T)=>{t.ac("BatchNormalization",o,{epsilon:d,momentum:m,spatial:!!b,trainingMode:!!c,format:T?"NHWC":"NCHW"})},939935:(o,d,m)=>{t.ac("CumSum",o,{exclusive:Number(d),reverse:Number(m)})},940032:(o,d,m)=>{t.ac("DequantizeLinear",o,{axis:d,blockSize:m})},940122:(o,d,m,c,b)=>{t.ac("GridSample",o,{align_corners:d,mode:Se(m),padding_mode:Se(c),format:b?"NHWC":"NCHW"})},940292:(o,d,m,c,b)=>{t.ac("GridSample",o,{align_corners:d,mode:Se(m),padding_mode:Se(c),format:b?"NHWC":"NCHW"})},940462:(o,d)=>{t.ac("ScatterND",o,{reduction:Se(d)})},940547:(o,d,m,c,b,T,z,R,W)=>{t.ac("Attention",o,{numHeads:d,isUnidirectional:m,maskFilterValue:c,scale:b,doRotary:T,qkvHiddenSizes:z?Array.from((x(),A).subarray(Number(R)>>>0,Number(R)+z>>>0)):[],pastPresentShareBuffer:!!W})},940819:o=>{t.ac("BiasAdd",o,void 0)},940874:o=>{t.ac("BiasSplitGelu",o,void 0)},940935:o=>{t.ac("FastGelu",o,void 0)},940991:(o,d,m,c,b,T,z,R,W,G,ie,pe,me,_e,pt,_i)=>{t.ac("Conv",o,{format:pe?"NHWC":"NCHW",auto_pad:d,dilations:m?Array.from((x(),A).subarray(Number(m)>>>0,Number(c)>>>0)):[],group:b,kernel_shape:T?Array.from((x(),A).subarray(Number(T)>>>0,Number(z)>>>0)):[],pads:R?Array.from((x(),A).subarray(Number(R)>>>0,Number(W)>>>0)):[],strides:G?Array.from((x(),A).subarray(Number(G)>>>0,Number(ie)>>>0)):[],w_is_const:()=>!!(x(),Q)[Number(me)>>>0],activation:Se(_e),activation_params:pt?Array.from((x(),ee).subarray(Number(pt)>>>0,Number(_i)>>>0)):[]})},941575:o=>{t.ac("Gelu",o,void 0)},941627:(o,d,m,c,b,T,z,R,W)=>{t.ac("GroupQueryAttention",o,{numHeads:d,kvNumHeads:m,scale:c,softcap:b,doRotary:T,rotaryInterleaved:z,smoothSoftmax:R,localWindowSize:W})},941844:(o,d,m,c)=>{t.ac("LayerNormalization",o,{axis:d,epsilon:m,simplified:!!c})},941955:(o,d,m,c)=>{t.ac("LayerNormalization",o,{axis:d,epsilon:m,simplified:!!c})},942066:(o,d,m,c,b,T)=>{t.ac("MatMulNBits",o,{k:d,n:m,accuracyLevel:c,bits:b,blockSize:T})},942193:(o,d,m,c,b,T)=>{t.ac("MultiHeadAttention",o,{numHeads:d,isUnidirectional:m,maskFilterValue:c,scale:b,doRotary:T})},942352:(o,d)=>{t.ac("QuickGelu",o,{alpha:d})},942416:(o,d,m,c,b)=>{t.ac("RotaryEmbedding",o,{interleaved:!!d,numHeads:m,rotaryEmbeddingDim:c,scale:b})},942555:(o,d,m)=>{t.ac("SkipLayerNormalization",o,{epsilon:d,simplified:!!m})},942657:(o,d,m)=>{t.ac("SkipLayerNormalization",o,{epsilon:d,simplified:!!m})},942759:(o,d,m,c)=>{t.ac("GatherBlockQuantized",o,{gatherAxis:d,quantizeAxis:m,blockSize:c})},942880:o=>{t.Id(o)},942914:(o,d)=>t.Kd(Number(o),Number(d),t.$c.Nd,t.$c.errors)};function Gm(o,d,m){return Wn(async()=>{await t.Gd(Number(o),Number(d),Number(m))})}function Fm(){return typeof wasmOffsetConverter<"u"}function Hm(o,d,m,c){var b=ne();try{return vs(o,d,m,c)}catch(T){if(ae(b),T!==T+0)throw T;oe(1,0)}}function jm(o,d,m){var c=ne();try{return _s(o,d,m)}catch(b){if(ae(c),b!==b+0)throw b;oe(1,0)}}function Km(o,d,m){var c=ne();try{fs(o,d,m)}catch(b){if(ae(c),b!==b+0)throw b;oe(1,0)}}function Zm(o,d){var m=ne();try{return gi(o,d)}catch(c){if(ae(m),c!==c+0)throw c;oe(1,0)}}function Qm(o){var d=ne();try{ms(o)}catch(m){if(ae(d),m!==m+0)throw m;oe(1,0)}}function Xm(o,d,m,c,b,T,z){var R=ne();try{return bs(o,d,m,c,b,T,z)}catch(W){if(ae(R),W!==W+0)throw W;oe(1,0)}}function Ym(o,d){var m=ne();try{xs(o,d)}catch(c){if(ae(m),c!==c+0)throw c;oe(1,0)}}function Jm(o,d,m,c,b,T){var z=ne();try{gs(o,d,m,c,b,T)}catch(R){if(ae(z),R!==R+0)throw R;oe(1,0)}}function eg(o,d,m,c){var b=ne();try{$s(o,d,m,c)}catch(T){if(ae(b),T!==T+0)throw T;oe(1,0)}}function tg(o,d,m,c,b){var T=ne();try{ys(o,d,m,c,b)}catch(z){if(ae(T),z!==z+0)throw z;oe(1,0)}}function rg(o,d,m,c,b,T,z){var R=ne();try{Ts(o,d,m,c,b,T,z)}catch(W){if(ae(R),W!==W+0)throw W;oe(1,0)}}function ig(o,d,m,c,b,T,z){var R=ne();try{Is(o,d,m,c,b,T,z)}catch(W){if(ae(R),W!==W+0)throw W;oe(1,0)}}function ag(o,d,m,c,b,T,z,R){var W=ne();try{Cs(o,d,m,c,b,T,z,R)}catch(G){if(ae(W),G!==G+0)throw G;oe(1,0)}}function ng(o,d,m,c,b){var T=ne();try{return Ss(o,d,m,c,b)}catch(z){if(ae(T),z!==z+0)throw z;oe(1,0)}}function sg(o,d,m,c,b,T,z,R){var W=ne();try{As(o,d,m,c,b,T,z,R)}catch(G){if(ae(W),G!==G+0)throw G;oe(1,0)}}function og(o,d,m,c,b,T,z,R,W,G,ie,pe){var me=ne();try{Es(o,d,m,c,b,T,z,R,W,G,ie,pe)}catch(_e){if(ae(me),_e!==_e+0)throw _e;oe(1,0)}}function ug(o,d,m,c,b,T){var z=ne();try{return ks(o,d,m,c,b,T)}catch(R){if(ae(z),R!==R+0)throw R;oe(1,0)}}function lg(o,d,m){var c=ne();try{return Os(o,d,m)}catch(b){if(ae(c),b!==b+0)throw b;return oe(1,0),0n}}function dg(o,d,m,c,b,T,z,R,W){var G=ne();try{ws(o,d,m,c,b,T,z,R,W)}catch(ie){if(ae(G),ie!==ie+0)throw ie;oe(1,0)}}function pg(o){var d=ne();try{return Bs(o)}catch(m){if(ae(d),m!==m+0)throw m;oe(1,0)}}function cg(o,d,m){var c=ne();try{return Rs(o,d,m)}catch(b){if(ae(c),b!==b+0)throw b;oe(1,0)}}function hg(o,d){var m=ne();try{return Qs(o,d)}catch(c){if(ae(m),c!==c+0)throw c;return oe(1,0),0n}}function fg(o,d,m,c,b){var T=ne();try{Ns(o,d,m,c,b)}catch(z){if(ae(T),z!==z+0)throw z;oe(1,0)}}function mg(o){var d=ne();try{return Ms(o)}catch(m){if(ae(d),m!==m+0)throw m;return oe(1,0),0n}}function gg(o,d,m,c,b,T){var z=ne();try{return Ls(o,d,m,c,b,T)}catch(R){if(ae(z),R!==R+0)throw R;oe(1,0)}}function yg(o,d,m,c,b,T){var z=ne();try{return Vs(o,d,m,c,b,T)}catch(R){if(ae(z),R!==R+0)throw R;oe(1,0)}}function _g(o,d,m,c,b,T,z,R){var W=ne();try{return zs(o,d,m,c,b,T,z,R)}catch(G){if(ae(W),G!==G+0)throw G;oe(1,0)}}function wg(o,d,m,c,b){var T=ne();try{return Gs(o,d,m,c,b)}catch(z){if(ae(T),z!==z+0)throw z;return oe(1,0),0n}}function bg(o,d,m,c){var b=ne();try{return Fs(o,d,m,c)}catch(T){if(ae(b),T!==T+0)throw T;oe(1,0)}}function $g(o,d,m,c){var b=ne();try{return Hs(o,d,m,c)}catch(T){if(ae(b),T!==T+0)throw T;oe(1,0)}}function vg(o,d,m,c,b,T,z,R,W,G,ie,pe){var me=ne();try{return js(o,d,m,c,b,T,z,R,W,G,ie,pe)}catch(_e){if(ae(me),_e!==_e+0)throw _e;oe(1,0)}}function xg(o,d,m,c,b,T,z,R,W,G,ie){var pe=ne();try{Ws(o,d,m,c,b,T,z,R,W,G,ie)}catch(me){if(ae(pe),me!==me+0)throw me;oe(1,0)}}function Sg(o,d,m,c,b,T,z,R,W,G,ie,pe,me,_e,pt,_i){var Cg=ne();try{qs(o,d,m,c,b,T,z,R,W,G,ie,pe,me,_e,pt,_i)}catch(wi){if(ae(Cg),wi!==wi+0)throw wi;oe(1,0)}}function Tg(o,d,m,c){var b=ne();try{return Ks(o,d,m,c)}catch(T){if(ae(b),T!==T+0)throw T;oe(1,0)}}function Ig(o,d,m,c,b){var T=ne();try{return Zs(o,d,m,c,b)}catch(z){if(ae(T),z!==z+0)throw z;oe(1,0)}}function Eg(o,d,m){var c=ne();try{return Ds(o,d,m)}catch(b){if(ae(c),b!==b+0)throw b;oe(1,0)}}function kg(o,d,m){var c=ne();try{return Ps(o,d,m)}catch(b){if(ae(c),b!==b+0)throw b;oe(1,0)}}function zg(o,d,m,c){var b=ne();try{Us(o,d,m,c)}catch(T){if(ae(b),T!==T+0)throw T;oe(1,0)}}function Ir(){if(0<Oe)bt=Ir;else if(a)$?.(t),de();else{for(var o=be;0<o.length;)o.shift()(t);0<Oe?bt=Ir:(t.calledRun=!0,C||(de(),$?.(t)))}}return a||(rt=await qe(),Ir()),t.PTR_SIZE=4,V?t:new Promise((o,d)=>{$=o,S=d})}var ip,io,Yg=P(()=>{"use strict";ip=ro,io=globalThis.self?.name?.startsWith("em-pthread"),io&&ro()}),Ti,ga,ao,Re,ap,kr,no,so,Ii,oo,Ei,np,ki,sp,Na=P(()=>{"use strict";Ra(),Ti=typeof location>"u"?void 0:location.origin,ga=import.meta.url>"file:"&&import.meta.url<"file;",ao=()=>{if(ga){let e=URL;return new URL(new e("ort.bundle.min.mjs",import.meta.url).href,Ti).href}return import.meta.url},Re=ao(),ap=()=>{if(Re&&!Re.startsWith("blob:"))return Re.substring(0,Re.lastIndexOf("/")+1)},kr=(e,t)=>{try{let r=t??Re;return(r?new URL(e,r):new URL(e)).origin===Ti}catch{return!1}},no=(e,t)=>{let r=t??Re;try{return(r?new URL(e,r):new URL(e)).href}catch{return}},so=(e,t)=>`${t??"./"}${e}`,Ii=async e=>{let t=await(await fetch(e,{credentials:"same-origin"})).blob();return URL.createObjectURL(t)},oo=async e=>(await import(e)).default,Ei=(Xg(),pr(ep)).default,np=async()=>{if(!Re)throw new Error("Failed to load proxy worker: cannot determine the script source URL.");if(kr(Re))return[void 0,Ei()];let e=await Ii(Re);return[e,Ei(e)]},ki=(Yg(),pr(rp)).default,sp=async(e,t,r,i)=>{let a=ki&&!(e||t);if(a)if(Re)a=kr(Re);else if(i&&!r)a=!0;else throw new Error("cannot determine the script source URL.");if(a)return[void 0,ki];{let s="ort-wasm-simd-threaded.jsep.mjs",n=e??no(s,t),u=r&&n&&!kr(n,t),l=u?await Ii(n):n??so(s,t);return[u?l:void 0,await oo(l)]}}}),zi,zr,Qt,Ci,uo,lo,po,Ma,ye,Mt=P(()=>{"use strict";Na(),zr=!1,Qt=!1,Ci=!1,uo=()=>{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}},lo=()=>{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}},po=()=>{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}},Ma=async e=>{if(zr)return Promise.resolve();if(Qt)throw new Error("multiple calls to 'initializeWebAssembly()' detected.");if(Ci)throw new Error("previous call to 'initializeWebAssembly()' failed.");Qt=!0;let t=e.initTimeout,r=e.numThreads;if(e.simd!==!1){if(e.simd==="relaxed"){if(!po())throw new Error("Relaxed WebAssembly SIMD is not supported in the current environment.")}else if(!lo())throw new Error("WebAssembly SIMD is not supported in the current environment.")}let i=uo();r>1&&!i&&(typeof self<"u"&&!self.crossOriginIsolated&&console.warn("env.wasm.numThreads is set to "+r+", 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=r=1);let a=e.wasmPaths,s=typeof a=="string"?a:void 0,n=a?.mjs,u=n?.href??n,l=a?.wasm,p=l?.href??l,h=e.wasmBinary,[f,g]=await sp(u,s,r>1,!!h||!!p),_=!1,y=[];if(t>0&&y.push(new Promise($=>{setTimeout(()=>{_=!0,$()},t)})),y.push(new Promise(($,S)=>{let v={numThreads:r};if(h)v.wasmBinary=h;else if(p||s)v.locateFile=w=>p??s+w;else if(u&&u.indexOf("blob:")!==0)v.locateFile=w=>new URL(w,u).href;else if(f){let w=ap();w&&(v.locateFile=E=>w+E)}g(v).then(w=>{Qt=!1,zr=!0,zi=w,$(),f&&URL.revokeObjectURL(f)},w=>{Qt=!1,Ci=!0,S(w)})})),await Promise.race(y),_)throw new Error(`WebAssembly backend initializing failed due to timeout: ${t}ms`)},ye=()=>{if(zr&&zi)return zi;throw new Error("WebAssembly is not initialized yet.")}}),Ke,Lr,fe,Da=P(()=>{"use strict";Mt(),Ke=(e,t)=>{let r=ye(),i=r.lengthBytesUTF8(e)+1,a=r._malloc(i);return r.stringToUTF8(e,a,i),t.push(a),a},Lr=(e,t,r,i)=>{if(typeof e=="object"&&e!==null){if(r.has(e))throw new Error("Circular reference in options");r.add(e)}Object.entries(e).forEach(([a,s])=>{let n=t?t+a:a;if(typeof s=="object")Lr(s,n+".",r,i);else if(typeof s=="string"||typeof s=="number")i(n,s.toString());else if(typeof s=="boolean")i(n,s?"1":"0");else throw new Error(`Can't handle extra config type: ${typeof s}`)})},fe=e=>{let t=ye(),r=t.stackSave();try{let i=t.PTR_SIZE,a=t.stackAlloc(2*i);t._OrtGetLastError(a,a+i);let s=Number(t.getValue(a,i===4?"i32":"i64")),n=t.getValue(a+i,"*"),u=n?t.UTF8ToString(n):"";throw new Error(`${e} ERROR_CODE: ${s}, ERROR_MESSAGE: ${u}`)}finally{t.stackRestore(r)}}}),op,Jg=P(()=>{"use strict";Mt(),Da(),op=e=>{let t=ye(),r=0,i=[],a=e||{};try{if(e?.logSeverityLevel===void 0)a.logSeverityLevel=2;else if(typeof e.logSeverityLevel!="number"||!Number.isInteger(e.logSeverityLevel)||e.logSeverityLevel<0||e.logSeverityLevel>4)throw new Error(`log severity level is not valid: ${e.logSeverityLevel}`);if(e?.logVerbosityLevel===void 0)a.logVerbosityLevel=0;else if(typeof e.logVerbosityLevel!="number"||!Number.isInteger(e.logVerbosityLevel))throw new Error(`log verbosity level is not valid: ${e.logVerbosityLevel}`);e?.terminate===void 0&&(a.terminate=!1);let s=0;return e?.tag!==void 0&&(s=Ke(e.tag,i)),r=t._OrtCreateRunOptions(a.logSeverityLevel,a.logVerbosityLevel,!!a.terminate,s),r===0&&fe("Can't create run options."),e?.extra!==void 0&&Lr(e.extra,"",new WeakSet,(n,u)=>{let l=Ke(n,i),p=Ke(u,i);t._OrtAddRunConfigEntry(r,l,p)!==0&&fe(`Can't set a run config entry: ${n} - ${u}.`)}),[r,i]}catch(s){throw r!==0&&t._OrtReleaseRunOptions(r),i.forEach(n=>t._free(n)),s}}}),co,ho,fo,Xt,mo,up,e0=P(()=>{"use strict";Mt(),Da(),co=e=>{switch(e){case"disabled":return 0;case"basic":return 1;case"extended":return 2;case"layout":return 3;case"all":return 99;default:throw new Error(`unsupported graph optimization level: ${e}`)}},ho=e=>{switch(e){case"sequential":return 0;case"parallel":return 1;default:throw new Error(`unsupported execution mode: ${e}`)}},fo=e=>{e.extra||(e.extra={}),e.extra.session||(e.extra.session={});let t=e.extra.session;t.use_ort_model_bytes_directly||(t.use_ort_model_bytes_directly="1"),e.executionProviders&&e.executionProviders.some(r=>(typeof r=="string"?r:r.name)==="webgpu")&&(e.enableMemPattern=!1)},Xt=(e,t,r,i)=>{let a=Ke(t,i),s=Ke(r,i);ye()._OrtAddSessionConfigEntry(e,a,s)!==0&&fe(`Can't set a session config entry: ${t} - ${r}.`)},mo=async(e,t,r)=>{let i=t.executionProviders;for(let a of i){let s=typeof a=="string"?a:a.name,n=[];switch(s){case"webnn":if(s="WEBNN",typeof a!="string"){let f=a?.deviceType;f&&Xt(e,"deviceType",f,r)}break;case"webgpu":if(s="JS",typeof a!="string"){let f=a;if(f?.preferredLayout){if(f.preferredLayout!=="NCHW"&&f.preferredLayout!=="NHWC")throw new Error(`preferredLayout must be either 'NCHW' or 'NHWC': ${f.preferredLayout}`);Xt(e,"preferredLayout",f.preferredLayout,r)}}break;case"wasm":case"cpu":continue;default:throw new Error(`not supported execution provider: ${s}`)}let u=Ke(s,r),l=n.length,p=0,h=0;if(l>0){p=ye()._malloc(l*ye().PTR_SIZE),r.push(p),h=ye()._malloc(l*ye().PTR_SIZE),r.push(h);for(let f=0;f<l;f++)ye().setValue(p+f*ye().PTR_SIZE,n[f][0],"*"),ye().setValue(h+f*ye().PTR_SIZE,n[f][1],"*")}await ye()._OrtAppendExecutionProvider(e,u,p,h,l)!==0&&fe(`Can't append execution provider: ${s}.`)}},up=async e=>{let t=ye(),r=0,i=[],a=e||{};fo(a);try{let s=co(a.graphOptimizationLevel??"all"),n=ho(a.executionMode??"sequential"),u=typeof a.logId=="string"?Ke(a.logId,i):0,l=a.logSeverityLevel??2;if(!Number.isInteger(l)||l<0||l>4)throw new Error(`log severity level is not valid: ${l}`);let p=a.logVerbosityLevel??0;if(!Number.isInteger(p)||p<0||p>4)throw new Error(`log verbosity level is not valid: ${p}`);let h=typeof a.optimizedModelFilePath=="string"?Ke(a.optimizedModelFilePath,i):0;if(r=t._OrtCreateSessionOptions(s,!!a.enableCpuMemArena,!!a.enableMemPattern,n,!!a.enableProfiling,0,u,l,p,h),r===0&&fe("Can't create session options."),a.executionProviders&&await mo(r,a,i),a.enableGraphCapture!==void 0){if(typeof a.enableGraphCapture!="boolean")throw new Error(`enableGraphCapture must be a boolean value: ${a.enableGraphCapture}`);Xt(r,"enableGraphCapture",a.enableGraphCapture.toString(),i)}if(a.freeDimensionOverrides)for(let[f,g]of Object.entries(a.freeDimensionOverrides)){if(typeof f!="string")throw new Error(`free dimension override name must be a string: ${f}`);if(typeof g!="number"||!Number.isInteger(g)||g<0)throw new Error(`free dimension override value must be a non-negative integer: ${g}`);let _=Ke(f,i);t._OrtAddFreeDimensionOverride(r,_,g)!==0&&fe(`Can't set a free dimension override: ${f} - ${g}.`)}return a.extra!==void 0&&Lr(a.extra,"",new WeakSet,(f,g)=>{Xt(r,f,g,i)}),[r,i]}catch(s){throw r!==0&&t._OrtReleaseSessionOptions(r)!==0&&fe("Can't release session options."),i.forEach(n=>t._free(n)),s}}}),Ct,at,At,Zr,Vr,Pa,Ua,ya,Y=P(()=>{"use strict";Ct=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}`)}},at=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}`)}},At=(e,t)=>{let r=[-1,4,1,1,2,2,4,8,-1,1,2,8,4,8,-1,-1,-1,-1,-1,-1,-1,.5,.5][e],i=typeof t=="number"?t:t.reduce((a,s)=>a*s,1);return r>0?Math.ceil(i*r):void 0},Zr=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}`)}},Vr=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}`)}},Pa=e=>e==="float32"||e==="float16"||e==="int32"||e==="int64"||e==="uint32"||e==="uint8"||e==="bool"||e==="uint4"||e==="int4",Ua=e=>e==="float32"||e==="float16"||e==="int32"||e==="int64"||e==="uint32"||e==="uint64"||e==="int8"||e==="uint8"||e==="bool"||e==="uint4"||e==="int4",ya=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}`)}}}),Wa,lp=P(()=>{"use strict";Ra(),Wa=async e=>{if(typeof e=="string"){let t=await fetch(e);if(!t.ok)throw new Error(`failed to load external data file: ${e}`);let r=t.headers.get("Content-Length"),i=r?parseInt(r,10):0;if(i<1073741824)return new Uint8Array(await t.arrayBuffer());{if(!t.body)throw new Error(`failed to load external data file: ${e}, no response body.`);let a=t.body.getReader(),s;try{s=new ArrayBuffer(i)}catch(u){if(u instanceof RangeError){let l=Math.ceil(i/65536);s=new WebAssembly.Memory({initial:l,maximum:l}).buffer}else throw u}let n=0;for(;;){let{done:u,value:l}=await a.read();if(u)break;let p=l.byteLength;new Uint8Array(s,n,p).set(l),n+=p}return new Uint8Array(s,0,i)}}else return e instanceof Blob?new Uint8Array(await e.arrayBuffer()):e instanceof Uint8Array?e:new Uint8Array(e)}}),go,yo,_o,wo,qa,bo,ue,nt=P(()=>{"use strict";Y(),go=["V","I","W","E","F"],yo=(e,t)=>{console.log(`[${go[e]},${new Date().toISOString()}]${t}`)},qa=(e,t)=>{_o=e,wo=t},bo=(e,t)=>{let r=Vr(e),i=Vr(_o);r>=i&&yo(r,typeof t=="function"?t():t)},ue=(...e)=>{wo&&bo(...e)}}),$o,Lt,B,Gr,dp,pp,cp,te=P(()=>{"use strict";$o=class{static calcMatMulShape(e,t){return e[1]!==t[0]?void 0:[e[0],t[1]]}},Lt=class{static calcShape(e,t,r=!1){let i=e.length,a=t.length;if(i===0)return t;if(a===0)return e;let s=Math.max(e.length,t.length),n=new Array(s);if(r){if(i<2||a<2)return;let u=$o.calcMatMulShape([e[i-2],e[i-1]],[t[a-2],t[a-1]]);if(u===void 0)return;[n[s-2],n[s-1]]=u}for(let u=r?3:1;u<=s;u++){let l=i-u<0?1:e[i-u],p=a-u<0?1:t[a-u];if(l!==p&&l>1&&p>1)return;let h=Math.max(l,p);if(l&&p)n[s-u]=Math.max(l,p);else{if(h>1)return;n[s-u]=0}}return n}static isValidBroadcast(e,t){let r=e.length,i=t.length;if(r>i)return!1;for(let a=1;a<=r;a++)if(e[r-a]!==1&&e[r-a]!==t[i-a])return!1;return!0}},B=class Wr{static size(t){return Wr.getSizeFromDimensionRange(t,0,t.length)}static convertShape(t,r=4){let i=t.length;if(i===0)return[];let a=new Array(i),s=i-1;for(;s>=0;){if(t[s]%r===0){a[s]=t[s]/r;break}if(r%t[s]!==0)throw new Error("cannot convert shape");a[s]=1,r/=t[s],s--}for(s--;s>=0;s--)a[s]=t[s];return a}static sizeFromDimension(t,r){if(r<0||r>t.length)throw new Error(`invalid dimension of ${r} for sizeFromDimension as Tensor has ${t.length} dimensions.`);return Wr.getSizeFromDimensionRange(t,r,t.length)}static sizeToDimension(t,r){if(r<0||r>t.length)throw new Error(`invalid dimension of ${r} for sizeToDimension as Tensor has ${t.length} dimensions.`);return Wr.getSizeFromDimensionRange(t,0,r)}static getSizeFromDimensionRange(t,r,i){let a=1;for(let s=r;s<i;s++){if(t[s]<0)throw new Error("cannot get valid size from specified dimension range. Most likely the range contains negative values in them.");a*=Number(t[s])}return a}static computeStrides(t){let r=t.length;if(r===0)return[];if(r===1)return[1];let i=new Array(r);i[r-1]=1,i[r-2]=t[r-1];for(let a=r-3;a>=0;--a)i[a]=i[a+1]*t[a+1];return i}static normalizeAxis(t,r){if(t<-r&&t>=r)throw new Error("unsupported axis for this operation.");return t<0?t+r:t}static normalizeAxes(t,r){return t.map(i=>this.normalizeAxis(i,r??t.length))}static sortBasedOnPerm(t,r){return r?r.map(i=>t[i]):t.slice().reverse()}static padShape(t,r){let i=t.length;return t.map((a,s)=>a+r[s]+r[s+i])}static areEqual(t,r){return t.length!==r.length?!1:t.every((i,a)=>i===r[a])}},Gr=class or{static adjustPoolAttributes(t,r,i,a,s,n){if(!t&&i.length!==r.length-2)throw new Error("length of specified kernel shapes should be 2 less than length of input dimensions");if(t)for(let u=0;u<r.length-2;u++)u>=i.length?i.push(r[u+2]):i[u]=r[u+2];for(let u=0;u<i.length;u++)if(u<a.length){if(a[u]<0)throw new Error("strides should be greater than or equal to 1")}else a.push(1);for(let u=0;u<i.length;u++)if(u<s.length){if(s[u]<0)throw new Error("dilations should be greater than or equal to 1")}else s.push(1);for(let u=0;u<i.length*2;u++)if(u<n.length){if(n[u]<0)throw new Error("pad should be greater than or equal to 1")}else n.push(0);for(let u=0;u<i.length;u++){if(i[u]<=0)throw new Error("kernel shapes need to be greater than 0");if(n[u]>=i[u]||n[u+i.length]>=i[u])throw new Error("pads should be smaller than kernel")}}static adjustPadsBasedOnAutoPad(t,r,i,a,s,n,u){if(u){if(s.length!==2*(t.length-2))throw new Error("length of pads should be twice the length of data dimensions");if(r.length!==t.length-2)throw new Error("length of strides should be the length of data dimensions");if(a.length!==t.length-2)throw new Error("length of kernel shapes should be the length of data dimensions");for(let l=0;l<t.length-2;l++)or.adjustPadAndReturnShape(t[l+(n?1:2)],r[l],i[l],a[l],s,l,l+t.length-2,u)}}static computePoolOutputShape(t,r,i,a,s,n,u){if(r.length<=0)throw new Error("input shape must be of size greater than 0");let l=[r[0],r[1]];return or.computeShapeHelper(t,r,l,i,a,s,n,u),l}static computeConvOutputShape(t,r,i,a,s,n,u){if(t.length<=0||r.length<=0)throw new Error("invalid input tensor dims or invalid filter tensor dims");let l=[t[0],r[0]];return or.computeShapeHelper(!1,t,l,i,a,s,n,u),l}static computeShapeHelper(t,r,i,a,s,n,u,l){if(t)for(let p=0;p<r.length-2;p++)i.push(1);else for(let p=0;p<r.length-2;p++)i.push(or.adjustPadAndReturnShape(r[p+2],a[p],s[p],n[p],u,p,p+r.length-2,l))}static adjustPadAndReturnShape(t,r,i,a,s,n,u,l){let p=i*(a-1)+1;if(l&&l!=="NOTSET")switch(l){case"VALID":return s[n]=0,s[u]=0,Math.floor((t-p)/r+1);case"SAME_LOWER":case"SAME_UPPER":if(i!==1)throw new Error("Dilation not supported for SAME_UPPER or SAME_LOWER");{let h=((t+r-1)/r-1)*r+a-t;return s[n]=Math.floor(l==="SAME_LOWER"?(h+1)/2:h/2),s[u]=h-s[n],Math.floor((t+h-a)/r+1)}default:throw new Error("Unsupported AutoPad type")}else return Math.floor((t+s[n]+s[u]-p)/r+1)}},dp=class{static getShapeOfGemmResult(e,t,r,i,a){if(e.length!==2||r.length!==2)throw new Error("shape need to be of size 2");let s,n,u;t?(s=e[1],n=e[0]):(s=e[0],n=e[1]);let l=-1;if(i?(u=r[0],l=1):(u=r[1],l=0),r[l]!==n)throw new Error("dimension mismatch");if(s<=0||u<=0||n<=0)throw new Error("invalid shape specified");if(a&&!Lt.isValidBroadcast(a,[s,u]))throw new Error("gemm: invalid bias shape for broadcast");return[s,u,n]}},pp=-34028234663852886e22,cp=34028234663852886e22}),La,hp=P(()=>{"use strict";Y(),La=(e,t)=>new(Zr(t))(e)}),Ai,_a,Oi,vo,Bi,xo,Ri,Ni,Mi,So,fp,t0=P(()=>{"use strict";Y(),nt(),Ai=new Map([["float32",32],["float16",16],["int32",32],["uint32",32],["int64",64],["uint64",64],["int8",8],["uint8",8],["int4",4],["uint4",4]]),_a=(e,t)=>{if(t==="int32")return e;let r=Ai.get(t);if(!r)throw new Error(`WebNN backend does not support data type: ${t}`);let i=r/8;if(e.byteLength%i!==0)throw new Error(`Invalid Uint8Array length - must be a multiple of ${i}.`);let a=e.byteLength/i,s=new(Zr(t))(e.buffer,e.byteOffset,a);switch(t){case"int64":case"uint64":{let n=new Int32Array(a);for(let u=0;u<a;u++){let l=s[u];if(l>2147483647n||l<-2147483648n)throw new Error("Can not convert int64 data to int32 - value out of range.");n[u]=Number(l)}return new Uint8Array(n.buffer)}case"int8":case"uint8":case"uint32":{if(t==="uint32"&&s.some(u=>u>2147483647))throw new Error("Can not convert uint32 data to int32 - value out of range.");let n=Int32Array.from(s,Number);return new Uint8Array(n.buffer)}default:throw new Error(`Unsupported data conversion from ${t} to 'int32'`)}},Oi=(e,t)=>{if(t==="int32")return e;if(e.byteLength%4!==0)throw new Error("Invalid Uint8Array length - must be a multiple of 4 (int32).");let r=e.byteLength/4,i=new Int32Array(e.buffer,e.byteOffset,r);switch(t){case"int64":{let a=BigInt64Array.from(i,BigInt);return new Uint8Array(a.buffer)}case"uint64":{if(i.some(s=>s<0))throw new Error("Can not convert int32 data to uin64 - negative value found.");let a=BigUint64Array.from(i,BigInt);return new Uint8Array(a.buffer)}case"int8":{if(i.some(s=>s<-128||s>127))throw new Error("Can not convert int32 data to int8 - value out of range.");let a=Int8Array.from(i,Number);return new Uint8Array(a.buffer)}case"uint8":{if(i.some(a=>a<0||a>255))throw new Error("Can not convert int32 data to uint8 - value out of range.");return Uint8Array.from(i,Number)}case"uint32":{if(i.some(s=>s<0))throw new Error("Can not convert int32 data to uint32 - negative value found.");let a=Uint32Array.from(i,Number);return new Uint8Array(a.buffer)}default:throw new Error(`Unsupported data conversion from 'int32' to ${t}`)}},vo=1,Bi=()=>vo++,xo=new Map([["int8","int32"],["uint8","int32"],["uint32","int32"],["int64","int32"]]),Ri=(e,t)=>{let r=Ai.get(e);if(!r)throw new Error(`WebNN backend does not support data type: ${e}`);return t.length>0?Math.ceil(t.reduce((i,a)=>i*a)*r/8):0},Ni=class{constructor(e){this.isDataConverted=!1;let{sessionId:t,context:r,tensor:i,dataType:a,shape:s,fallbackDataType:n}=e;this.sessionId=t,this.mlContext=r,this.mlTensor=i,this.dataType=a,this.tensorShape=s,this.fallbackDataType=n}get tensor(){return this.mlTensor}get type(){return this.dataType}get fallbackType(){return this.fallbackDataType}get shape(){return this.tensorShape}get byteLength(){return Ri(this.dataType,this.tensorShape)}destroy(){ue("verbose",()=>"[WebNN] TensorWrapper.destroy"),this.mlTensor.destroy()}write(e){this.mlContext.writeTensor(this.mlTensor,e)}async read(e){if(this.fallbackDataType){let t=await this.mlContext.readTensor(this.mlTensor),r=Oi(new Uint8Array(t),this.dataType);if(e){(e instanceof ArrayBuffer?new Uint8Array(e):new Uint8Array(e.buffer,e.byteOffset,e.byteLength)).set(r);return}else return r.buffer}else return e?this.mlContext.readTensor(this.mlTensor,e):this.mlContext.readTensor(this.mlTensor)}canReuseTensor(e,t,r){return this.mlContext===e&&this.dataType===t&&this.tensorShape.length===r.length&&this.tensorShape.every((i,a)=>i===r[a])}setIsDataConverted(e){this.isDataConverted=e}},Mi=class{constructor(e,t){this.tensorManager=e,this.wrapper=t}get tensorWrapper(){return this.wrapper}releaseTensor(){this.tensorWrapper&&(this.tensorManager.releaseTensor(this.tensorWrapper),this.wrapper=void 0)}async ensureTensor(e,t,r,i){let a=this.tensorManager.getMLContext(e),s=this.tensorManager.getMLOpSupportLimits(e),n;if(!s?.input.dataTypes.includes(t)){if(n=xo.get(t),!n||s?.input.dataTypes.includes(n))throw new Error(`WebNN backend does not support data type: ${t}`);ue("verbose",()=>`[WebNN] TensorIdTracker.ensureTensor: fallback dataType from ${t} to ${n}`)}if(this.wrapper){if(this.wrapper.canReuseTensor(a,t,r))return this.wrapper.tensor;if(i){if(this.wrapper.byteLength!==Ri(t,r))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 u=typeof MLTensorUsage>"u"?void 0:MLTensorUsage.READ|MLTensorUsage.WRITE;return this.wrapper=await this.tensorManager.getCachedTensor(e,t,r,u,!0,!0,n),i&&this.activeUpload&&(this.wrapper.write(this.activeUpload),this.activeUpload=void 0),this.wrapper.tensor}upload(e){let t=e;if(this.wrapper){if(this.wrapper.fallbackType)if(this.wrapper.fallbackType==="int32")t=_a(e,this.wrapper.type),this.wrapper.setIsDataConverted(!0);else throw new Error(`Unsupported fallback data type: ${this.wrapper.fallbackType}`);if(e.byteLength===this.wrapper.byteLength){this.wrapper.write(t);return}else ue("verbose",()=>"Data size does not match tensor size. Releasing tensor."),this.releaseTensor()}this.activeUpload?this.activeUpload.set(t):this.activeUpload=new Uint8Array(t)}async download(e){if(this.activeUpload){let t=this.wrapper?.isDataConverted?Oi(this.activeUpload,this.wrapper?.type):this.activeUpload;if(e){e instanceof ArrayBuffer?new Uint8Array(e).set(t):new Uint8Array(e.buffer,e.byteOffset,e.byteLength).set(t);return}else return t.buffer}if(!this.wrapper)throw new Error("Tensor has not been created.");return e?this.wrapper.read(e):this.wrapper.read()}},So=class{constructor(e){this.backend=e,this.tensorTrackersById=new Map,this.freeTensors=[],this.externalTensors=new Set}getMLContext(e){let t=this.backend.getMLContext(e);if(!t)throw new Error("MLContext not found for session.");return t}getMLOpSupportLimits(e){return this.backend.getMLOpSupportLimits(e)}reserveTensorId(){let e=Bi();return this.tensorTrackersById.set(e,new Mi(this)),e}releaseTensorId(e){let t=this.tensorTrackersById.get(e);t&&(this.tensorTrackersById.delete(e),t.tensorWrapper&&this.releaseTensor(t.tensorWrapper))}async ensureTensor(e,t,r,i,a){ue("verbose",()=>`[WebNN] TensorManager.ensureTensor {tensorId: ${t}, dataType: ${r}, shape: ${i}, copyOld: ${a}}`);let s=this.tensorTrackersById.get(t);if(!s)throw new Error("Tensor not found.");return s.ensureTensor(e,r,i,a)}upload(e,t){let r=this.tensorTrackersById.get(e);if(!r)throw new Error("Tensor not found.");r.upload(t)}async download(e,t){ue("verbose",()=>`[WebNN] TensorManager.download {tensorId: ${e}, dstBuffer: ${t?.byteLength}}`);let r=this.tensorTrackersById.get(e);if(!r)throw new Error("Tensor not found.");return r.download(t)}releaseTensorsForSession(e){for(let t of this.freeTensors)t.sessionId===e&&t.destroy();this.freeTensors=this.freeTensors.filter(t=>t.sessionId!==e)}registerTensor(e,t,r,i){let a=this.getMLContext(e),s=Bi(),n=new Ni({sessionId:e,context:a,tensor:t,dataType:r,shape:i});return this.tensorTrackersById.set(s,new Mi(this,n)),this.externalTensors.add(n),s}async getCachedTensor(e,t,r,i,a,s,n){let u=this.getMLContext(e);for(let[p,h]of this.freeTensors.entries())if(h.canReuseTensor(u,t,r)){ue("verbose",()=>`[WebNN] Reusing tensor {dataType: ${t}, ${n?`fallbackDataType: ${n},`:""} shape: ${r}`);let f=this.freeTensors.splice(p,1)[0];return f.sessionId=e,f}ue("verbose",()=>`[WebNN] MLContext.createTensor {dataType: ${t}, ${n?`fallbackDataType: ${n},`:""} shape: ${r}}`);let l=await u.createTensor({dataType:n??t,shape:r,dimensions:r,usage:i,writable:a,readable:s});return new Ni({sessionId:e,context:u,tensor:l,dataType:t,shape:r,fallbackDataType:n})}releaseTensor(e){this.externalTensors.has(e)&&this.externalTensors.delete(e),this.freeTensors.push(e)}},fp=(...e)=>new So(...e)}),Yt,To,mp,r0=P(()=>{"use strict";Y(),Mt(),hp(),t0(),nt(),Yt=new Map([[1,"float32"],[10,"float16"],[6,"int32"],[12,"uint32"],[7,"int64"],[13,"uint64"],[22,"int4"],[21,"uint4"],[3,"int8"],[2,"uint8"],[9,"uint8"]]),To=(e,t)=>{if(e===t)return!0;if(e===void 0||t===void 0)return!1;let r=Object.keys(e).sort(),i=Object.keys(t).sort();return r.length===i.length&&r.every((a,s)=>a===i[s]&&e[a]===t[a])},mp=class{constructor(e){this.tensorManager=fp(this),this.mlContextBySessionId=new Map,this.sessionIdsByMLContext=new Map,this.mlContextCache=[],this.sessionGraphInputs=new Map,this.sessionGraphOutputs=new Map,this.temporaryGraphInputs=[],this.temporaryGraphOutputs=[],this.temporarySessionTensorIds=new Map,this.mlOpSupportLimitsBySessionId=new Map,qa(e.logLevel,!!e.debug)}get currentSessionId(){if(this.activeSessionId===void 0)throw new Error("No active session");return this.activeSessionId}onRunStart(e){ue("verbose",()=>`[WebNN] onRunStart {sessionId: ${e}}`),this.activeSessionId=e}onRunEnd(e){ue("verbose",()=>`[WebNN] onRunEnd {sessionId: ${e}}`);let t=this.temporarySessionTensorIds.get(e);if(t){for(let r of t)ue("verbose",()=>`[WebNN] releasing temporary tensor {tensorId: ${r}}`),this.tensorManager.releaseTensorId(r);this.temporarySessionTensorIds.delete(e),this.activeSessionId=void 0}}async createMLContext(e){if(e instanceof GPUDevice){let r=this.mlContextCache.findIndex(i=>i.gpuDevice===e);if(r!==-1)return this.mlContextCache[r].mlContext;{let i=await navigator.ml.createContext(e);return this.mlContextCache.push({gpuDevice:e,mlContext:i}),i}}else if(e===void 0){let r=this.mlContextCache.findIndex(i=>i.options===void 0&&i.gpuDevice===void 0);if(r!==-1)return this.mlContextCache[r].mlContext;{let i=await navigator.ml.createContext();return this.mlContextCache.push({mlContext:i}),i}}let t=this.mlContextCache.findIndex(r=>To(r.options,e));if(t!==-1)return this.mlContextCache[t].mlContext;{let r=await navigator.ml.createContext(e);return this.mlContextCache.push({options:e,mlContext:r}),r}}registerMLContext(e,t){this.mlContextBySessionId.set(e,t);let r=this.sessionIdsByMLContext.get(t);r||(r=new Set,this.sessionIdsByMLContext.set(t,r)),r.add(e),this.mlOpSupportLimitsBySessionId.has(e)||this.mlOpSupportLimitsBySessionId.set(e,t.opSupportLimits()),this.temporaryGraphInputs.length>0&&(this.sessionGraphInputs.set(e,this.temporaryGraphInputs),this.temporaryGraphInputs=[]),this.temporaryGraphOutputs.length>0&&(this.sessionGraphOutputs.set(e,this.temporaryGraphOutputs),this.temporaryGraphOutputs=[])}onReleaseSession(e){this.sessionGraphInputs.delete(e),this.sessionGraphOutputs.delete(e);let t=this.mlContextBySessionId.get(e);if(!t)return;this.tensorManager.releaseTensorsForSession(e),this.mlContextBySessionId.delete(e),this.mlOpSupportLimitsBySessionId.delete(e);let r=this.sessionIdsByMLContext.get(t);if(r.delete(e),r.size===0){this.sessionIdsByMLContext.delete(t);let i=this.mlContextCache.findIndex(a=>a.mlContext===t);i!==-1&&this.mlContextCache.splice(i,1)}}getMLContext(e){return this.mlContextBySessionId.get(e)}getMLOpSupportLimits(e){return this.mlOpSupportLimitsBySessionId.get(e)}reserveTensorId(){return this.tensorManager.reserveTensorId()}releaseTensorId(e){ue("verbose",()=>`[WebNN] releaseTensorId {tensorId: ${e}}`),this.tensorManager.releaseTensorId(e)}async ensureTensor(e,t,r,i,a){let s=Yt.get(r);if(!s)throw new Error(`Unsupported ONNX data type: ${r}`);return this.tensorManager.ensureTensor(e??this.currentSessionId,t,s,i,a)}async createTemporaryTensor(e,t,r){ue("verbose",()=>`[WebNN] createTemporaryTensor {onnxDataType: ${t}, shape: ${r}}`);let i=Yt.get(t);if(!i)throw new Error(`Unsupported ONNX data type: ${t}`);let a=this.tensorManager.reserveTensorId();await this.tensorManager.ensureTensor(e,a,i,r,!1);let s=this.temporarySessionTensorIds.get(e);return s?s.push(a):this.temporarySessionTensorIds.set(e,[a]),a}uploadTensor(e,t){if(!ye().shouldTransferToMLTensor)throw new Error("Trying to upload to a MLTensor while shouldTransferToMLTensor is false");ue("verbose",()=>`[WebNN] uploadTensor {tensorId: ${e}, data: ${t.byteLength}}`),this.tensorManager.upload(e,t)}async downloadTensor(e,t){return this.tensorManager.download(e,t)}createMLTensorDownloader(e,t){return async()=>{let r=await this.tensorManager.download(e);return La(r,t)}}registerMLTensor(e,t,r,i){let a=Yt.get(r);if(!a)throw new Error(`Unsupported ONNX data type: ${r}`);let s=this.tensorManager.registerTensor(e,t,a,i);return ue("verbose",()=>`[WebNN] registerMLTensor {tensor: ${t}, dataType: ${a}, dimensions: ${i}} -> {tensorId: ${s}}`),s}registerMLConstant(e,t,r,i,a,s,n=!1){if(!s)throw new Error("External mounted files are not available.");let u=e;e.startsWith("./")&&(u=e.substring(2));let l=s.get(u);if(!l)throw new Error(`File with name ${u} not found in preloaded files.`);if(t+r>l.byteLength)throw new Error("Out of bounds: data offset and length exceed the external file data size.");let p=l.slice(t,t+r).buffer,h;switch(a.dataType){case"float32":h=new Float32Array(p);break;case"float16":h=typeof Float16Array<"u"&&Float16Array.from?new Float16Array(p):new Uint16Array(p);break;case"int32":h=new Int32Array(p);break;case"uint32":h=new Uint32Array(p);break;case"int64":if(n){let f=_a(new Uint8Array(p),"int64");h=new Int32Array(f.buffer),a.dataType="int32"}else h=new BigInt64Array(p);break;case"uint64":h=new BigUint64Array(p);break;case"int8":h=new Int8Array(p);break;case"int4":case"uint4":case"uint8":h=new Uint8Array(p);break;default:throw new Error(`Unsupported data type: ${a.dataType} in creating WebNN Constant from external data.`)}return ue("verbose",()=>`[WebNN] registerMLConstant {dataType: ${a.dataType}, shape: ${a.shape}}} ${n?"(Note: it was int64 data type and registered to int32 as workaround)":""}`),i.constant(a,h)}registerGraphInput(e){this.temporaryGraphInputs.push(e)}registerGraphOutput(e){this.temporaryGraphOutputs.push(e)}isGraphInput(e,t){let r=this.sessionGraphInputs.get(e);return r?r.includes(t):!1}isGraphOutput(e,t){let r=this.sessionGraphOutputs.get(e);return r?r.includes(t):!1}isGraphInputOutputTypeSupported(e,t,r=!0){let i=Yt.get(Ct(t)),a=this.mlOpSupportLimitsBySessionId.get(e);return typeof i>"u"?!1:r?!!a?.input.dataTypes.includes(i):!!a?.output.dataTypes.includes(i)}flush(){}}}),Va=P(()=>{"use strict"}),Di,Cr,Ar,Io,Eo,Pi,wa,ko,gp,i0=P(()=>{"use strict";nt(),Va(),Di=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]]),Cr=[],Ar=e=>Math.ceil(Number(e)/16)*16,Io=e=>{for(let t=0;t<Cr.length;t++){let r=Cr[t];if(e<=r)return r}return Math.ceil(e/16)*16},Eo=1,Pi=()=>Eo++,wa=async(e,t,r,i)=>{let a=Ar(r),s=e.device.createBuffer({size:a,usage:GPUBufferUsage.COPY_DST|GPUBufferUsage.MAP_READ});try{let n=e.getCommandEncoder();e.endComputePass(),n.copyBufferToBuffer(t,0,s,0,a),e.flush(),await s.mapAsync(GPUMapMode.READ);let u=s.getMappedRange();if(i){let l=i();return l.set(new Uint8Array(u,0,r)),l}else return new Uint8Array(u.slice(0,r))}finally{s.destroy()}},ko=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[t]of Di)Cr.push(t),this.freeBuffers.set(t,[]),this.freeUniformBuffers.set(t,[]);this.sessionCount=0}upload(e,t){let r=t.buffer,i=t.byteOffset,a=t.byteLength,s=Ar(a),n=this.storageCache.get(e);if(!n)throw new Error("gpu data for uploading does not exist");if(Number(n.originalSize)!==a)throw new Error(`inconsistent data size. gpu data size=${n.originalSize}, data size=${a}`);let u=this.backend.device.createBuffer({mappedAtCreation:!0,size:s,usage:GPUBufferUsage.MAP_WRITE|GPUBufferUsage.COPY_SRC}),l=u.getMappedRange();new Uint8Array(l).set(new Uint8Array(r,i,a)),u.unmap();let p=this.backend.device.createCommandEncoder();p.copyBufferToBuffer(u,0,n.gpuData.buffer,0,s),this.backend.device.queue.submit([p.finish()]),u.destroy(),ue("verbose",()=>`[WebGPU] GpuDataManager.upload(id=${e})`)}memcpy(e,t){let r=this.storageCache.get(e);if(!r)throw new Error("source gpu data for memcpy does not exist");let i=this.storageCache.get(t);if(!i)throw new Error("destination gpu data for memcpy does not exist");if(r.originalSize!==i.originalSize)throw new Error("inconsistent source and destination gpu data size");let a=Ar(r.originalSize),s=this.backend.getCommandEncoder();this.backend.endComputePass(),s.copyBufferToBuffer(r.gpuData.buffer,0,i.gpuData.buffer,0,a)}registerExternalBuffer(e,t,r){let i;if(r){if(i=r[0],e===r[1])return ue("verbose",()=>`[WebGPU] GpuDataManager.registerExternalBuffer(size=${t}) => id=${i}, buffer is the same, skip.`),i;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 i=Pi();return this.storageCache.set(i,{gpuData:{id:i,type:0,buffer:e},originalSize:t}),ue("verbose",()=>`[WebGPU] GpuDataManager.registerExternalBuffer(size=${t}) => id=${i}, registered.`),i}unregisterExternalBuffer(e){e!==void 0&&(this.storageCache.delete(e),ue("verbose",()=>`[WebGPU] GpuDataManager.unregisterExternalBuffer() => id=${e}`))}create(e,t=GPUBufferUsage.STORAGE|GPUBufferUsage.COPY_SRC|GPUBufferUsage.COPY_DST){let r=Io(e),i,a=(t&GPUBufferUsage.STORAGE)===GPUBufferUsage.STORAGE,s=(t&GPUBufferUsage.UNIFORM)===GPUBufferUsage.UNIFORM;if(a||s){let u=(a?this.freeBuffers:this.freeUniformBuffers).get(r);u?u.length>0?i=u.pop():i=this.backend.device.createBuffer({size:r,usage:t}):i=this.backend.device.createBuffer({size:r,usage:t})}else i=this.backend.device.createBuffer({size:r,usage:t});let n={id:Pi(),type:0,buffer:i};return this.storageCache.set(n.id,{gpuData:n,originalSize:Number(e)}),ue("verbose",()=>`[WebGPU] GpuDataManager.create(size=${e}) => id=${n.id}`),n}get(e){return this.storageCache.get(e)?.gpuData}release(e){let t=typeof e=="bigint"?Number(e):e,r=this.storageCache.get(t);if(!r){if(this.storageCache.size===0)return 0;throw new Error("releasing data does not exist")}return ue("verbose",()=>`[WebGPU] GpuDataManager.release(id=${t}), gpuDataId=${r.gpuData.id}`),this.storageCache.delete(t),this.buffersPending.push(r.gpuData.buffer),r.originalSize}async download(e,t){let r=this.storageCache.get(Number(e));if(!r)throw new Error("data does not exist");await wa(this.backend,r.gpuData.buffer,r.originalSize,t)}refreshPendingBuffers(){if(this.buffersPending.length!==0)if(this.backend.sessionStatus==="default"){for(let e of this.buffersPending){let t=Di.get(e.size);if((e.usage&GPUBufferUsage.STORAGE)===GPUBufferUsage.STORAGE){let r=this.freeBuffers.get(e.size)||[];t===void 0||r.length>=t?e.destroy():r.push(e)}else if((e.usage&GPUBufferUsage.UNIFORM)===GPUBufferUsage.UNIFORM){let r=this.freeUniformBuffers.get(e.size)||[];t===void 0||r.length>=t?e.destroy():r.push(e)}else e.destroy()}this.buffersPending=[]}else{let e=this.capturedPendingBuffers.get(this.backend.currentSessionId);e||(e=[],this.capturedPendingBuffers.set(this.backend.currentSessionId,e));for(let t of this.buffersPending)e.push(t);this.buffersPending=[]}}dispose(){this.freeBuffers.forEach(e=>{e.forEach(t=>{t.destroy()})}),this.freeUniformBuffers.forEach(e=>{e.forEach(t=>{t.destroy()})}),this.storageCache.forEach(e=>{e.gpuData.buffer.destroy()}),this.capturedPendingBuffers.forEach(e=>{e.forEach(t=>{t.destroy()})}),this.storageCache=new Map,this.freeBuffers=new Map,this.freeUniformBuffers=new Map,this.capturedPendingBuffers=new Map}onCreateSession(){this.sessionCount+=1}onReleaseSession(e){let t=this.capturedPendingBuffers.get(e);t&&(t.forEach(r=>{r.destroy()}),this.capturedPendingBuffers.delete(e)),this.sessionCount-=1,this.sessionCount===0&&(ue("warning",()=>"[WebGPU] Clearing webgpu buffer cache"),this.storageCache.forEach(r=>{r.gpuData.buffer.destroy()}),this.storageCache=new Map)}},gp=(...e)=>new ko(...e)}),zo,he,ve=P(()=>{"use strict";zo=class{constructor(e){Object.assign(this,e)}get cacheKey(){return this.key||(this.key=Object.getOwnPropertyNames(this).sort().map(e=>`${this[e]}`).join(";")),this.key}},he=e=>new zo(e)}),Vt,Or,Ee,Ce,Z,$e,ba,qt,yt,j,Jt,N,F,yp,Ga,Co,_p,re=P(()=>{"use strict";Y(),te(),Vt=64,Or=(e,t)=>{if(t===3)throw new Error("vec3 has same alignment as vec4, use vec4 instead");switch(Number(e)){case 10:return t>1?`vec${t}<f16>`:"f16";case 1:return t>1?`vec${t}<f32>`:"f32";case 6:return t>1?`vec${t}<i32>`:"i32";case 12:return t>1?`vec${t}<u32>`:"u32";case 7:if(t>1)throw new Error("currently not supported vecX of uint64 yet");return["vec2<u32>","i32"];case 13:if(t>1)throw new Error("currently not supported vecX of uint64 yet");return["vec2<u32>","u32"];case 9:if(t!==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: ${e}`)}},Ee=(e,t=1)=>{let r=Or(e,t);return typeof r=="string"?r:r[0]},Ce=(e,t=1)=>{let r=Or(e,t);return typeof r=="string"?r:r[1]},Z=(...e)=>{let t=[];return e.forEach(r=>{r.length!==0&&t.push({type:12,data:r},{type:12,data:B.computeStrides(r)})}),t},$e=e=>e%4===0?4:e%2===0?2:1,ba=(e="f32",t,r="0")=>!t||t===1?`${e}(${r})`:`vec${t}<${e}>(${r})`,qt=(e,t,r)=>e==="f32"?r:t===1?`f32(${r})`:`vec${t}<f32>(${r})`,yt=(e,t)=>t===4?`(${e}.x + ${e}.y + ${e}.z + ${e}.w)`:t===2?`(${e}.x + ${e}.y)`:t===3?`(${e}.x + ${e}.y + ${e}.z)`:e,j=(e,t,r,i)=>e.startsWith("uniforms.")&&r>4?typeof t=="string"?i==="f16"?`${e}[(${t}) / 8][(${t}) % 8 / 4][(${t}) % 8 % 4]`:`${e}[(${t}) / 4][(${t}) % 4]`:i==="f16"?`${e}[${Math.floor(t/8)}][${Math.floor(t%8/4)}][${t%8%4}]`:`${e}[${Math.floor(t/4)}][${t%4}]`:r>1?`${e}[${t}]`:e,Jt=(e,t,r,i,a)=>{let s=typeof r=="number",n=s?r:r.length,u=[...new Array(n).keys()],l=n<2?"u32":n<=4?`vec${n}<u32>`:`array<u32, ${n}>`,p=Or(t,a),h=typeof p=="string"?p:p[1],f=typeof p=="string"?p:p[0],g={indices:l,value:h,storage:f,tensor:t},_=M=>typeof M=="string"?M:`${M}u`,y={offsetToIndices:!1,indicesToOffset:!1,broadcastedIndicesToOffset:!1,set:!1,setByIndices:!1,get:!1,getByIndices:!1},$=s?"uniforms.":"",S=`${$}${e}_shape`,v=`${$}${e}_strides`,w="";for(let M=0;M<n-1;M++)w+=`
-    let dim${M} = current / ${j(v,M,n)};
-    let rest${M} = current % ${j(v,M,n)};
-    indices[${M}] = dim${M};
-    current = rest${M};
-    `;w+=`indices[${n-1}] = current;`;let E=n<2?"":`
+  }`,m=new Function(Object.keys(z),R)(...Object.values(z)),R=`methodCaller<(${b.map(W=>W.name)}) => ${c.name}>`,sm(Object.defineProperty(m,"name",{value:R}))}function pm(o,d){return d>>>=0,(o=De(o>>>0))==De(d)}function cm(o){return(o>>>=0)?(o=vr(o),qe(globalThis[o])):qe(globalThis)}function hm(o){return o=vr(o>>>0),qe(t[o])}function fm(o,d){return d>>>=0,o=De(o>>>0),d=De(d),qe(o[d])}function mm(o){9<(o>>>=0)&&(vt[o+1]+=1)}function Ln(o,d,m,c,b){return ui[o>>>0](d>>>0,m>>>0,c>>>0,b>>>0)}function gm(o,d,m,c,b){return Ln(o>>>0,d>>>0,m>>>0,c>>>0,b>>>0)}function ym(){return qe([])}function _m(o){o=De(o>>>0);for(var d=Array(o.length),m=0;m<o.length;m++)d[m]=o[m];return qe(d)}function wm(o){return qe(vr(o>>>0))}function bm(){return qe({})}function $m(o){for(var d=De(o>>>=0);d.length;){var m=d.pop();d.pop()(m)}ri(o)}function vm(o,d,m){d>>>=0,m>>>=0,o=De(o>>>0),d=De(d),m=De(m),o[d]=m}function xm(o,d){o=-9007199254740992>o||9007199254740992<o?NaN:Number(o),d>>>=0,o=new Date(1e3*o),(x(),A)[d>>>2>>>0]=o.getUTCSeconds(),(x(),A)[d+4>>>2>>>0]=o.getUTCMinutes(),(x(),A)[d+8>>>2>>>0]=o.getUTCHours(),(x(),A)[d+12>>>2>>>0]=o.getUTCDate(),(x(),A)[d+16>>>2>>>0]=o.getUTCMonth(),(x(),A)[d+20>>>2>>>0]=o.getUTCFullYear()-1900,(x(),A)[d+24>>>2>>>0]=o.getUTCDay(),o=(o.getTime()-Date.UTC(o.getUTCFullYear(),0,1,0,0,0,0))/864e5|0,(x(),A)[d+28>>>2>>>0]=o}var qn=o=>o%4==0&&(o%100!=0||o%400==0),Vn=[0,31,60,91,121,152,182,213,244,274,305,335],Gn=[0,31,59,90,120,151,181,212,243,273,304,334];function Sm(o,d){o=-9007199254740992>o||9007199254740992<o?NaN:Number(o),d>>>=0,o=new Date(1e3*o),(x(),A)[d>>>2>>>0]=o.getSeconds(),(x(),A)[d+4>>>2>>>0]=o.getMinutes(),(x(),A)[d+8>>>2>>>0]=o.getHours(),(x(),A)[d+12>>>2>>>0]=o.getDate(),(x(),A)[d+16>>>2>>>0]=o.getMonth(),(x(),A)[d+20>>>2>>>0]=o.getFullYear()-1900,(x(),A)[d+24>>>2>>>0]=o.getDay();var m=(qn(o.getFullYear())?Vn:Gn)[o.getMonth()]+o.getDate()-1|0;(x(),A)[d+28>>>2>>>0]=m,(x(),A)[d+36>>>2>>>0]=-60*o.getTimezoneOffset(),m=new Date(o.getFullYear(),6,1).getTimezoneOffset();var c=new Date(o.getFullYear(),0,1).getTimezoneOffset();o=0|(m!=c&&o.getTimezoneOffset()==Math.min(c,m)),(x(),A)[d+32>>>2>>>0]=o}function Tm(o){o>>>=0;var d=new Date((x(),A)[o+20>>>2>>>0]+1900,(x(),A)[o+16>>>2>>>0],(x(),A)[o+12>>>2>>>0],(x(),A)[o+8>>>2>>>0],(x(),A)[o+4>>>2>>>0],(x(),A)[o>>>2>>>0],0),m=(x(),A)[o+32>>>2>>>0],c=d.getTimezoneOffset(),b=new Date(d.getFullYear(),6,1).getTimezoneOffset(),T=new Date(d.getFullYear(),0,1).getTimezoneOffset(),z=Math.min(T,b);return 0>m?(x(),A)[o+32>>>2>>>0]=+(b!=T&&z==c):0<m!=(z==c)&&(b=Math.max(T,b),d.setTime(d.getTime()+6e4*((0<m?z:b)-c))),(x(),A)[o+24>>>2>>>0]=d.getDay(),m=(qn(d.getFullYear())?Vn:Gn)[d.getMonth()]+d.getDate()-1|0,(x(),A)[o+28>>>2>>>0]=m,(x(),A)[o>>>2>>>0]=d.getSeconds(),(x(),A)[o+4>>>2>>>0]=d.getMinutes(),(x(),A)[o+8>>>2>>>0]=d.getHours(),(x(),A)[o+12>>>2>>>0]=d.getDate(),(x(),A)[o+16>>>2>>>0]=d.getMonth(),(x(),A)[o+20>>>2>>>0]=d.getYear(),o=d.getTime(),BigInt(isNaN(o)?-1:o/1e3)}function Fn(o,d,m,c,b,T,z){return a?we(16,1,o,d,m,c,b,T,z):-52}function Hn(o,d,m,c,b,T){if(a)return we(17,1,o,d,m,c,b,T)}var Ht={},Im=()=>performance.timeOrigin+performance.now();function jn(o,d){if(a)return we(18,1,o,d);if(Ht[o]&&(clearTimeout(Ht[o].id),delete Ht[o]),!d)return 0;var m=setTimeout(()=>{delete Ht[o],_r(()=>us(o,performance.timeOrigin+performance.now()))},d);return Ht[o]={id:m,be:d},0}function km(o,d,m,c){o>>>=0,d>>>=0,m>>>=0,c>>>=0;var b=new Date().getFullYear(),T=new Date(b,0,1).getTimezoneOffset();b=new Date(b,6,1).getTimezoneOffset();var z=Math.max(T,b);(x(),U)[o>>>2>>>0]=60*z,(x(),A)[d>>>2>>>0]=+(T!=b),o=(d=R=>{var W=Math.abs(R);return`UTC${0<=R?"-":"+"}${String(Math.floor(W/60)).padStart(2,"0")}${String(W%60).padStart(2,"0")}`})(T),d=d(b),b<T?(lt(o,m,17),lt(d,c,17)):(lt(o,c,17),lt(d,m,17))}var Em=()=>Date.now(),zm=1;function Cm(o,d,m){if(m>>>=0,!(0<=o&&3>=o))return 28;if(o===0)o=Date.now();else{if(!zm)return 52;o=performance.timeOrigin+performance.now()}return o=Math.round(1e6*o),(x(),H)[m>>>3>>>0]=BigInt(o),0}var li=[],Kn=(o,d)=>{li.length=0;for(var m;m=(x(),q)[o++>>>0];){var c=m!=105;d+=(c&=m!=112)&&d%8?4:0,li.push(m==112?(x(),U)[d>>>2>>>0]:m==106?(x(),H)[d>>>3>>>0]:m==105?(x(),A)[d>>>2>>>0]:(x(),X)[d>>>3>>>0]),d+=c?8:4}return li};function Am(o,d,m){return o>>>=0,d=Kn(d>>>0,m>>>0),yi[o](...d)}function Om(o,d,m){return o>>>=0,d=Kn(d>>>0,m>>>0),yi[o](...d)}var Bm=()=>{};function Rm(o,d){return E(Se(o>>>0,d>>>0))}var Mm=()=>{throw st+=1,"unwind"};function Nm(){return 4294901760}var Dm=()=>navigator.hardwareConcurrency,xt={},xr=o=>{var d;return(d=/\bwasm-function\[\d+\]:(0x[0-9a-f]+)/.exec(o))?+d[1]:(d=/:(\d+):\d+(?:\)|$)/.exec(o))?2147483648|+d[1]:0},Zn=o=>{for(var d of o)(o=xr(d))&&(xt[o]=d)};function Pm(){var o=Error().stack.toString().split(`
+`);return o[0]=="Error"&&o.shift(),Zn(o),xt.sd=xr(o[3]),xt.Md=o,xt.sd}function Sr(o){if(!(o=xt[o>>>0]))return 0;var d;if(d=/^\s+at .*\.wasm\.(.*) \(.*\)$/.exec(o))o=d[1];else if(d=/^\s+at (.*) \(.*\)$/.exec(o))o=d[1];else{if(!(d=/^(.+?)@/.exec(o)))return 0;o=d[1]}Ye(Sr.td??0),d=yr(o)+1;var m=jt(d);return m&&lt(o,m,d),Sr.td=m,Sr.td}function Um(o){o>>>=0;var d=(x(),q).length;if(o<=d||4294901760<o)return!1;for(var m=1;4>=m;m*=2){var c=d*(1+.2/m);c=Math.min(c,o+100663296);e:{c=(Math.min(4294901760,65536*Math.ceil(Math.max(o,c)/65536))-ut.buffer.byteLength+65535)/65536|0;try{ut.grow(c),J();var b=1;break e}catch{}b=void 0}if(b)return!0}return!1}function Wm(o,d,m){if(o>>>=0,d>>>=0,xt.sd==o)var c=xt.Md;else(c=Error().stack.toString().split(`
+`))[0]=="Error"&&c.shift(),Zn(c);for(var b=3;c[b]&&xr(c[b])!=o;)++b;for(o=0;o<m&&c[o+b];++o)(x(),A)[d+4*o>>>2>>>0]=xr(c[o+b]);return o}var di,pi={},Qn=()=>{if(!di){var o,d={USER:"web_user",LOGNAME:"web_user",PATH:"/",PWD:"/",HOME:"/home/web_user",LANG:(globalThis.navigator?.language??"C").replace("-","_")+".UTF-8",_:"./this.program"};for(o in pi)pi[o]===void 0?delete d[o]:d[o]=pi[o];var m=[];for(o in d)m.push(`${o}=${d[o]}`);di=m}return di};function Xn(o,d){if(a)return we(19,1,o,d);o>>>=0,d>>>=0;var m,c=0,b=0;for(m of Qn()){var T=d+c;(x(),U)[o+b>>>2>>>0]=T,c+=lt(m,T,1/0)+1,b+=4}return 0}function Yn(o,d){if(a)return we(20,1,o,d);o>>>=0,d>>>=0;var m=Qn();for(var c of((x(),U)[o>>>2>>>0]=m.length,o=0,m))o+=yr(c)+1;return(x(),U)[d>>>2>>>0]=o,0}function Jn(o){return a?we(21,1,o):52}function es(o,d,m,c){return a?we(22,1,o,d,m,c):52}function ts(o,d,m,c){return a?we(23,1,o,d,m,c):70}var Lm=[null,[],[]];function rs(o,d,m,c){if(a)return we(24,1,o,d,m,c);d>>>=0,m>>>=0,c>>>=0;for(var b=0,T=0;T<m;T++){var z=(x(),U)[d>>>2>>>0],R=(x(),U)[d+4>>>2>>>0];d+=8;for(var W=0;W<R;W++){var G=o,ie=(x(),q)[z+W>>>0],pe=Lm[G];ie===0||ie===10?((G===1?I:E)(bn(pe)),pe.length=0):pe.push(ie)}b+=R}return(x(),U)[c>>>2>>>0]=b,0}function qm(o){return o>>>0}a||(function(){for(var o=t.numThreads-1;o--;)hn();be.push(async()=>{var d=(async function(){if(!a)return Promise.all(ot.map(cn))})();Oe++,await d,--Oe==0&&bt&&(d=bt,bt=null,d())})})(),a||(ut=new WebAssembly.Memory({initial:256,maximum:65536,shared:!0}),J()),t.wasmBinary&&(g=t.wasmBinary),t.stackSave=()=>ne(),t.stackRestore=o=>ae(o),t.stackAlloc=o=>fi(o),t.setValue=function(o,d,m="i8"){switch(m.endsWith("*")&&(m="*"),m){case"i1":case"i8":(x(),Q)[o>>>0]=d;break;case"i16":(x(),K)[o>>>1>>>0]=d;break;case"i32":(x(),A)[o>>>2>>>0]=d;break;case"i64":(x(),H)[o>>>3>>>0]=BigInt(d);break;case"float":(x(),ee)[o>>>2>>>0]=d;break;case"double":(x(),X)[o>>>3>>>0]=d;break;case"*":(x(),U)[o>>>2>>>0]=d;break;default:Ie(`invalid type for setValue: ${m}`)}},t.getValue=function(o,d="i8"){switch(d.endsWith("*")&&(d="*"),d){case"i1":case"i8":return(x(),Q)[o>>>0];case"i16":return(x(),K)[o>>>1>>>0];case"i32":return(x(),A)[o>>>2>>>0];case"i64":return(x(),H)[o>>>3>>>0];case"float":return(x(),ee)[o>>>2>>>0];case"double":return(x(),X)[o>>>3>>>0];case"*":return(x(),U)[o>>>2>>>0];default:Ie(`invalid type for getValue: ${d}`)}},t.UTF8ToString=Se,t.stringToUTF8=lt,t.lengthBytesUTF8=yr;var is,as,Tr,Ye,jt,ci,ns,ss,os,hi,us,ls,oe,Kt,ds,ae,fi,ne,ps,mi,cs,hs,fs,gi,ms,gs,ys,_s,ws,bs,$s,vs,xs,Ss,Ts,Is,ks,Es,zs,Cs,As,Os,Bs,Rs,Ms,Ns,Ds,Ps,Us,Ws,Ls,qs,Vs,Gs,Fs,Hs,js,Ks,Zs,Qs,Xs,Ys,Js,eo,rt,Vm=[Qr,un,gn,$n,vn,xn,Sn,Tn,In,kn,En,zn,Cn,An,On,Bn,Fn,Hn,jn,Xn,Yn,Jn,es,ts,rs],yi={927244:(o,d,m,c,b)=>{if(t===void 0||!t.Zc)return 1;if((o=Se(Number(o>>>0))).startsWith("./")&&(o=o.substring(2)),!(o=t.Zc.get(o)))return 2;if(d=Number(d>>>0),m=Number(m>>>0),c=Number(c>>>0),d+m>o.byteLength)return 3;try{let T=o.subarray(d,d+m);switch(b){case 0:(x(),q).set(T,c>>>0);break;case 1:t.Xd?t.Xd(c,T):t.Ld(c,T);break;default:return 4}return 0}catch{return 4}},928068:(o,d,m)=>{t.xd(o,(x(),q).subarray(d>>>0,d+m>>>0))},928132:()=>t.Zd(),928174:o=>{t.vd(o)},928211:()=>{t.Ed()},928242:()=>{t.Fd()},928271:()=>{t.Jd()},928296:o=>t.Dd(o),928329:o=>t.Hd(o),928361:(o,d,m)=>{t.jd(Number(o),Number(d),Number(m),!0)},928424:(o,d,m)=>{t.jd(Number(o),Number(d),Number(m))},928481:()=>typeof wasmOffsetConverter<"u",928538:o=>{t.ac("Abs",o,void 0)},928589:o=>{t.ac("Neg",o,void 0)},928640:o=>{t.ac("Floor",o,void 0)},928693:o=>{t.ac("Ceil",o,void 0)},928745:o=>{t.ac("Reciprocal",o,void 0)},928803:o=>{t.ac("Sqrt",o,void 0)},928855:o=>{t.ac("Exp",o,void 0)},928906:o=>{t.ac("Erf",o,void 0)},928957:o=>{t.ac("Sigmoid",o,void 0)},929012:(o,d,m)=>{t.ac("HardSigmoid",o,{alpha:d,beta:m})},929091:o=>{t.ac("Log",o,void 0)},929142:o=>{t.ac("Sin",o,void 0)},929193:o=>{t.ac("Cos",o,void 0)},929244:o=>{t.ac("Tan",o,void 0)},929295:o=>{t.ac("Asin",o,void 0)},929347:o=>{t.ac("Acos",o,void 0)},929399:o=>{t.ac("Atan",o,void 0)},929451:o=>{t.ac("Sinh",o,void 0)},929503:o=>{t.ac("Cosh",o,void 0)},929555:o=>{t.ac("Asinh",o,void 0)},929608:o=>{t.ac("Acosh",o,void 0)},929661:o=>{t.ac("Atanh",o,void 0)},929714:o=>{t.ac("Tanh",o,void 0)},929766:o=>{t.ac("Not",o,void 0)},929817:(o,d,m)=>{t.ac("Clip",o,{min:d,max:m})},929886:o=>{t.ac("Clip",o,void 0)},929938:(o,d)=>{t.ac("Elu",o,{alpha:d})},929996:o=>{t.ac("Gelu",o,void 0)},930048:o=>{t.ac("Relu",o,void 0)},930100:(o,d)=>{t.ac("LeakyRelu",o,{alpha:d})},930164:(o,d)=>{t.ac("ThresholdedRelu",o,{alpha:d})},930234:(o,d)=>{t.ac("Cast",o,{to:d})},930292:o=>{t.ac("Add",o,void 0)},930343:o=>{t.ac("Sub",o,void 0)},930394:o=>{t.ac("Mul",o,void 0)},930445:o=>{t.ac("Div",o,void 0)},930496:o=>{t.ac("Pow",o,void 0)},930547:o=>{t.ac("Equal",o,void 0)},930600:o=>{t.ac("Greater",o,void 0)},930655:o=>{t.ac("GreaterOrEqual",o,void 0)},930717:o=>{t.ac("Less",o,void 0)},930769:o=>{t.ac("LessOrEqual",o,void 0)},930828:(o,d,m,c,b)=>{t.ac("ReduceMean",o,{keepDims:!!d,noopWithEmptyAxes:!!m,axes:c?Array.from((x(),A).subarray(Number(c)>>>0,Number(b)>>>0)):[]})},931003:(o,d,m,c,b)=>{t.ac("ReduceMax",o,{keepDims:!!d,noopWithEmptyAxes:!!m,axes:c?Array.from((x(),A).subarray(Number(c)>>>0,Number(b)>>>0)):[]})},931177:(o,d,m,c,b)=>{t.ac("ReduceMin",o,{keepDims:!!d,noopWithEmptyAxes:!!m,axes:c?Array.from((x(),A).subarray(Number(c)>>>0,Number(b)>>>0)):[]})},931351:(o,d,m,c,b)=>{t.ac("ReduceProd",o,{keepDims:!!d,noopWithEmptyAxes:!!m,axes:c?Array.from((x(),A).subarray(Number(c)>>>0,Number(b)>>>0)):[]})},931526:(o,d,m,c,b)=>{t.ac("ReduceSum",o,{keepDims:!!d,noopWithEmptyAxes:!!m,axes:c?Array.from((x(),A).subarray(Number(c)>>>0,Number(b)>>>0)):[]})},931700:(o,d,m,c,b)=>{t.ac("ReduceL1",o,{keepDims:!!d,noopWithEmptyAxes:!!m,axes:c?Array.from((x(),A).subarray(Number(c)>>>0,Number(b)>>>0)):[]})},931873:(o,d,m,c,b)=>{t.ac("ReduceL2",o,{keepDims:!!d,noopWithEmptyAxes:!!m,axes:c?Array.from((x(),A).subarray(Number(c)>>>0,Number(b)>>>0)):[]})},932046:(o,d,m,c,b)=>{t.ac("ReduceLogSum",o,{keepDims:!!d,noopWithEmptyAxes:!!m,axes:c?Array.from((x(),A).subarray(Number(c)>>>0,Number(b)>>>0)):[]})},932223:(o,d,m,c,b)=>{t.ac("ReduceSumSquare",o,{keepDims:!!d,noopWithEmptyAxes:!!m,axes:c?Array.from((x(),A).subarray(Number(c)>>>0,Number(b)>>>0)):[]})},932403:(o,d,m,c,b)=>{t.ac("ReduceLogSumExp",o,{keepDims:!!d,noopWithEmptyAxes:!!m,axes:c?Array.from((x(),A).subarray(Number(c)>>>0,Number(b)>>>0)):[]})},932583:o=>{t.ac("Where",o,void 0)},932636:(o,d,m)=>{t.ac("Transpose",o,{perm:d?Array.from((x(),A).subarray(Number(d)>>>0,Number(m)>>>0)):[]})},932760:(o,d,m,c)=>{t.ac("DepthToSpace",o,{blocksize:d,mode:Se(m),format:c?"NHWC":"NCHW"})},932893:(o,d,m,c)=>{t.ac("DepthToSpace",o,{blocksize:d,mode:Se(m),format:c?"NHWC":"NCHW"})},933026:(o,d,m,c,b,T,z,R,W,G,ie,pe,me,_e,pt)=>{t.ac("ConvTranspose",o,{format:W?"NHWC":"NCHW",autoPad:d,dilations:[m],group:c,kernelShape:[b],pads:[T,z],strides:[R],wIsConst:()=>!!(x(),Q)[G>>>0],outputPadding:ie?Array.from((x(),A).subarray(Number(ie)>>>0,Number(pe)>>>0)):[],outputShape:me?Array.from((x(),A).subarray(Number(me)>>>0,Number(_e)>>>0)):[],activation:Se(pt)})},933459:(o,d,m,c,b,T,z,R,W,G,ie,pe,me,_e)=>{t.ac("ConvTranspose",o,{format:R?"NHWC":"NCHW",autoPad:d,dilations:Array.from((x(),A).subarray(Number(m)>>>0,2+(Number(m)>>>0)>>>0)),group:c,kernelShape:Array.from((x(),A).subarray(Number(b)>>>0,2+(Number(b)>>>0)>>>0)),pads:Array.from((x(),A).subarray(Number(T)>>>0,4+(Number(T)>>>0)>>>0)),strides:Array.from((x(),A).subarray(Number(z)>>>0,2+(Number(z)>>>0)>>>0)),wIsConst:()=>!!(x(),Q)[W>>>0],outputPadding:G?Array.from((x(),A).subarray(Number(G)>>>0,Number(ie)>>>0)):[],outputShape:pe?Array.from((x(),A).subarray(Number(pe)>>>0,Number(me)>>>0)):[],activation:Se(_e)})},934120:(o,d,m,c,b,T,z,R,W,G,ie,pe,me,_e,pt)=>{t.ac("ConvTranspose",o,{format:W?"NHWC":"NCHW",autoPad:d,dilations:[m],group:c,kernelShape:[b],pads:[T,z],strides:[R],wIsConst:()=>!!(x(),Q)[G>>>0],outputPadding:ie?Array.from((x(),A).subarray(Number(ie)>>>0,Number(pe)>>>0)):[],outputShape:me?Array.from((x(),A).subarray(Number(me)>>>0,Number(_e)>>>0)):[],activation:Se(pt)})},934553:(o,d,m,c,b,T,z,R,W,G,ie,pe,me,_e)=>{t.ac("ConvTranspose",o,{format:R?"NHWC":"NCHW",autoPad:d,dilations:Array.from((x(),A).subarray(Number(m)>>>0,2+(Number(m)>>>0)>>>0)),group:c,kernelShape:Array.from((x(),A).subarray(Number(b)>>>0,2+(Number(b)>>>0)>>>0)),pads:Array.from((x(),A).subarray(Number(T)>>>0,4+(Number(T)>>>0)>>>0)),strides:Array.from((x(),A).subarray(Number(z)>>>0,2+(Number(z)>>>0)>>>0)),wIsConst:()=>!!(x(),Q)[W>>>0],outputPadding:G?Array.from((x(),A).subarray(Number(G)>>>0,Number(ie)>>>0)):[],outputShape:pe?Array.from((x(),A).subarray(Number(pe)>>>0,Number(me)>>>0)):[],activation:Se(_e)})},935214:(o,d)=>{t.ac("GlobalAveragePool",o,{format:d?"NHWC":"NCHW"})},935305:(o,d,m,c,b,T,z,R,W,G,ie,pe,me,_e)=>{t.ac("AveragePool",o,{format:_e?"NHWC":"NCHW",auto_pad:d,ceil_mode:m,count_include_pad:c,storage_order:b,dilations:T?Array.from((x(),A).subarray(Number(T)>>>0,Number(z)>>>0)):[],kernel_shape:R?Array.from((x(),A).subarray(Number(R)>>>0,Number(W)>>>0)):[],pads:G?Array.from((x(),A).subarray(Number(G)>>>0,Number(ie)>>>0)):[],strides:pe?Array.from((x(),A).subarray(Number(pe)>>>0,Number(me)>>>0)):[]})},935784:(o,d)=>{t.ac("GlobalAveragePool",o,{format:d?"NHWC":"NCHW"})},935875:(o,d,m,c,b,T,z,R,W,G,ie,pe,me,_e)=>{t.ac("AveragePool",o,{format:_e?"NHWC":"NCHW",auto_pad:d,ceil_mode:m,count_include_pad:c,storage_order:b,dilations:T?Array.from((x(),A).subarray(Number(T)>>>0,Number(z)>>>0)):[],kernel_shape:R?Array.from((x(),A).subarray(Number(R)>>>0,Number(W)>>>0)):[],pads:G?Array.from((x(),A).subarray(Number(G)>>>0,Number(ie)>>>0)):[],strides:pe?Array.from((x(),A).subarray(Number(pe)>>>0,Number(me)>>>0)):[]})},936354:(o,d)=>{t.ac("GlobalMaxPool",o,{format:d?"NHWC":"NCHW"})},936441:(o,d,m,c,b,T,z,R,W,G,ie,pe,me,_e)=>{t.ac("MaxPool",o,{format:_e?"NHWC":"NCHW",auto_pad:d,ceil_mode:m,count_include_pad:c,storage_order:b,dilations:T?Array.from((x(),A).subarray(Number(T)>>>0,Number(z)>>>0)):[],kernel_shape:R?Array.from((x(),A).subarray(Number(R)>>>0,Number(W)>>>0)):[],pads:G?Array.from((x(),A).subarray(Number(G)>>>0,Number(ie)>>>0)):[],strides:pe?Array.from((x(),A).subarray(Number(pe)>>>0,Number(me)>>>0)):[]})},936916:(o,d)=>{t.ac("GlobalMaxPool",o,{format:d?"NHWC":"NCHW"})},937003:(o,d,m,c,b,T,z,R,W,G,ie,pe,me,_e)=>{t.ac("MaxPool",o,{format:_e?"NHWC":"NCHW",auto_pad:d,ceil_mode:m,count_include_pad:c,storage_order:b,dilations:T?Array.from((x(),A).subarray(Number(T)>>>0,Number(z)>>>0)):[],kernel_shape:R?Array.from((x(),A).subarray(Number(R)>>>0,Number(W)>>>0)):[],pads:G?Array.from((x(),A).subarray(Number(G)>>>0,Number(ie)>>>0)):[],strides:pe?Array.from((x(),A).subarray(Number(pe)>>>0,Number(me)>>>0)):[]})},937478:(o,d,m,c,b)=>{t.ac("Gemm",o,{alpha:d,beta:m,transA:c,transB:b})},937582:o=>{t.ac("MatMul",o,void 0)},937636:(o,d,m,c)=>{t.ac("ArgMax",o,{keepDims:!!d,selectLastIndex:!!m,axis:c})},937744:(o,d,m,c)=>{t.ac("ArgMin",o,{keepDims:!!d,selectLastIndex:!!m,axis:c})},937852:(o,d)=>{t.ac("Softmax",o,{axis:d})},937915:(o,d)=>{t.ac("Concat",o,{axis:d})},937975:(o,d,m,c,b)=>{t.ac("Split",o,{axis:d,numOutputs:m,splitSizes:c?Array.from((x(),A).subarray(Number(c)>>>0,Number(b)>>>0)):[]})},938131:o=>{t.ac("Expand",o,void 0)},938185:(o,d)=>{t.ac("Gather",o,{axis:Number(d)})},938256:(o,d)=>{t.ac("GatherElements",o,{axis:Number(d)})},938335:(o,d)=>{t.ac("GatherND",o,{batch_dims:Number(d)})},938414:(o,d,m,c,b,T,z,R,W,G,ie)=>{t.ac("Resize",o,{antialias:d,axes:m?Array.from((x(),A).subarray(Number(m)>>>0,Number(c)>>>0)):[],coordinateTransformMode:Se(b),cubicCoeffA:T,excludeOutside:z,extrapolationValue:R,keepAspectRatioPolicy:Se(W),mode:Se(G),nearestMode:Se(ie)})},938776:(o,d,m,c,b,T,z)=>{t.ac("Slice",o,{starts:d?Array.from((x(),A).subarray(Number(d)>>>0,Number(m)>>>0)):[],ends:c?Array.from((x(),A).subarray(Number(c)>>>0,Number(b)>>>0)):[],axes:T?Array.from((x(),A).subarray(Number(T)>>>0,Number(z)>>>0)):[]})},939040:o=>{t.ac("Tile",o,void 0)},939092:(o,d,m)=>{t.ac("InstanceNormalization",o,{epsilon:d,format:m?"NHWC":"NCHW"})},939206:(o,d,m)=>{t.ac("InstanceNormalization",o,{epsilon:d,format:m?"NHWC":"NCHW"})},939320:o=>{t.ac("Range",o,void 0)},939373:(o,d)=>{t.ac("Einsum",o,{equation:Se(d)})},939454:(o,d,m,c,b)=>{t.ac("Pad",o,{mode:d,value:m,pads:c?Array.from((x(),A).subarray(Number(c)>>>0,Number(b)>>>0)):[]})},939597:(o,d,m,c,b,T)=>{t.ac("BatchNormalization",o,{epsilon:d,momentum:m,spatial:!!b,trainingMode:!!c,format:T?"NHWC":"NCHW"})},939766:(o,d,m,c,b,T)=>{t.ac("BatchNormalization",o,{epsilon:d,momentum:m,spatial:!!b,trainingMode:!!c,format:T?"NHWC":"NCHW"})},939935:(o,d,m)=>{t.ac("CumSum",o,{exclusive:Number(d),reverse:Number(m)})},940032:(o,d,m)=>{t.ac("DequantizeLinear",o,{axis:d,blockSize:m})},940122:(o,d,m,c,b)=>{t.ac("GridSample",o,{align_corners:d,mode:Se(m),padding_mode:Se(c),format:b?"NHWC":"NCHW"})},940292:(o,d,m,c,b)=>{t.ac("GridSample",o,{align_corners:d,mode:Se(m),padding_mode:Se(c),format:b?"NHWC":"NCHW"})},940462:(o,d)=>{t.ac("ScatterND",o,{reduction:Se(d)})},940547:(o,d,m,c,b,T,z,R,W)=>{t.ac("Attention",o,{numHeads:d,isUnidirectional:m,maskFilterValue:c,scale:b,doRotary:T,qkvHiddenSizes:z?Array.from((x(),A).subarray(Number(R)>>>0,Number(R)+z>>>0)):[],pastPresentShareBuffer:!!W})},940819:o=>{t.ac("BiasAdd",o,void 0)},940874:o=>{t.ac("BiasSplitGelu",o,void 0)},940935:o=>{t.ac("FastGelu",o,void 0)},940991:(o,d,m,c,b,T,z,R,W,G,ie,pe,me,_e,pt,_i)=>{t.ac("Conv",o,{format:pe?"NHWC":"NCHW",auto_pad:d,dilations:m?Array.from((x(),A).subarray(Number(m)>>>0,Number(c)>>>0)):[],group:b,kernel_shape:T?Array.from((x(),A).subarray(Number(T)>>>0,Number(z)>>>0)):[],pads:R?Array.from((x(),A).subarray(Number(R)>>>0,Number(W)>>>0)):[],strides:G?Array.from((x(),A).subarray(Number(G)>>>0,Number(ie)>>>0)):[],w_is_const:()=>!!(x(),Q)[Number(me)>>>0],activation:Se(_e),activation_params:pt?Array.from((x(),ee).subarray(Number(pt)>>>0,Number(_i)>>>0)):[]})},941575:o=>{t.ac("Gelu",o,void 0)},941627:(o,d,m,c,b,T,z,R,W)=>{t.ac("GroupQueryAttention",o,{numHeads:d,kvNumHeads:m,scale:c,softcap:b,doRotary:T,rotaryInterleaved:z,smoothSoftmax:R,localWindowSize:W})},941844:(o,d,m,c)=>{t.ac("LayerNormalization",o,{axis:d,epsilon:m,simplified:!!c})},941955:(o,d,m,c)=>{t.ac("LayerNormalization",o,{axis:d,epsilon:m,simplified:!!c})},942066:(o,d,m,c,b,T)=>{t.ac("MatMulNBits",o,{k:d,n:m,accuracyLevel:c,bits:b,blockSize:T})},942193:(o,d,m,c,b,T)=>{t.ac("MultiHeadAttention",o,{numHeads:d,isUnidirectional:m,maskFilterValue:c,scale:b,doRotary:T})},942352:(o,d)=>{t.ac("QuickGelu",o,{alpha:d})},942416:(o,d,m,c,b)=>{t.ac("RotaryEmbedding",o,{interleaved:!!d,numHeads:m,rotaryEmbeddingDim:c,scale:b})},942555:(o,d,m)=>{t.ac("SkipLayerNormalization",o,{epsilon:d,simplified:!!m})},942657:(o,d,m)=>{t.ac("SkipLayerNormalization",o,{epsilon:d,simplified:!!m})},942759:(o,d,m,c)=>{t.ac("GatherBlockQuantized",o,{gatherAxis:d,quantizeAxis:m,blockSize:c})},942880:o=>{t.Id(o)},942914:(o,d)=>t.Kd(Number(o),Number(d),t.$c.Nd,t.$c.errors)};function Gm(o,d,m){return Wn(async()=>{await t.Gd(Number(o),Number(d),Number(m))})}function Fm(){return typeof wasmOffsetConverter<"u"}function Hm(o,d,m,c){var b=ne();try{return vs(o,d,m,c)}catch(T){if(ae(b),T!==T+0)throw T;oe(1,0)}}function jm(o,d,m){var c=ne();try{return _s(o,d,m)}catch(b){if(ae(c),b!==b+0)throw b;oe(1,0)}}function Km(o,d,m){var c=ne();try{fs(o,d,m)}catch(b){if(ae(c),b!==b+0)throw b;oe(1,0)}}function Zm(o,d){var m=ne();try{return gi(o,d)}catch(c){if(ae(m),c!==c+0)throw c;oe(1,0)}}function Qm(o){var d=ne();try{ms(o)}catch(m){if(ae(d),m!==m+0)throw m;oe(1,0)}}function Xm(o,d,m,c,b,T,z){var R=ne();try{return bs(o,d,m,c,b,T,z)}catch(W){if(ae(R),W!==W+0)throw W;oe(1,0)}}function Ym(o,d){var m=ne();try{xs(o,d)}catch(c){if(ae(m),c!==c+0)throw c;oe(1,0)}}function Jm(o,d,m,c,b,T){var z=ne();try{gs(o,d,m,c,b,T)}catch(R){if(ae(z),R!==R+0)throw R;oe(1,0)}}function eg(o,d,m,c){var b=ne();try{$s(o,d,m,c)}catch(T){if(ae(b),T!==T+0)throw T;oe(1,0)}}function tg(o,d,m,c,b){var T=ne();try{ys(o,d,m,c,b)}catch(z){if(ae(T),z!==z+0)throw z;oe(1,0)}}function rg(o,d,m,c,b,T,z){var R=ne();try{Ts(o,d,m,c,b,T,z)}catch(W){if(ae(R),W!==W+0)throw W;oe(1,0)}}function ig(o,d,m,c,b,T,z){var R=ne();try{Is(o,d,m,c,b,T,z)}catch(W){if(ae(R),W!==W+0)throw W;oe(1,0)}}function ag(o,d,m,c,b,T,z,R){var W=ne();try{Cs(o,d,m,c,b,T,z,R)}catch(G){if(ae(W),G!==G+0)throw G;oe(1,0)}}function ng(o,d,m,c,b){var T=ne();try{return Ss(o,d,m,c,b)}catch(z){if(ae(T),z!==z+0)throw z;oe(1,0)}}function sg(o,d,m,c,b,T,z,R){var W=ne();try{As(o,d,m,c,b,T,z,R)}catch(G){if(ae(W),G!==G+0)throw G;oe(1,0)}}function og(o,d,m,c,b,T,z,R,W,G,ie,pe){var me=ne();try{ks(o,d,m,c,b,T,z,R,W,G,ie,pe)}catch(_e){if(ae(me),_e!==_e+0)throw _e;oe(1,0)}}function ug(o,d,m,c,b,T){var z=ne();try{return Es(o,d,m,c,b,T)}catch(R){if(ae(z),R!==R+0)throw R;oe(1,0)}}function lg(o,d,m){var c=ne();try{return Os(o,d,m)}catch(b){if(ae(c),b!==b+0)throw b;return oe(1,0),0n}}function dg(o,d,m,c,b,T,z,R,W){var G=ne();try{ws(o,d,m,c,b,T,z,R,W)}catch(ie){if(ae(G),ie!==ie+0)throw ie;oe(1,0)}}function pg(o){var d=ne();try{return Bs(o)}catch(m){if(ae(d),m!==m+0)throw m;oe(1,0)}}function cg(o,d,m){var c=ne();try{return Rs(o,d,m)}catch(b){if(ae(c),b!==b+0)throw b;oe(1,0)}}function hg(o,d){var m=ne();try{return Qs(o,d)}catch(c){if(ae(m),c!==c+0)throw c;return oe(1,0),0n}}function fg(o,d,m,c,b){var T=ne();try{Ms(o,d,m,c,b)}catch(z){if(ae(T),z!==z+0)throw z;oe(1,0)}}function mg(o){var d=ne();try{return Ns(o)}catch(m){if(ae(d),m!==m+0)throw m;return oe(1,0),0n}}function gg(o,d,m,c,b,T){var z=ne();try{return qs(o,d,m,c,b,T)}catch(R){if(ae(z),R!==R+0)throw R;oe(1,0)}}function yg(o,d,m,c,b,T){var z=ne();try{return Vs(o,d,m,c,b,T)}catch(R){if(ae(z),R!==R+0)throw R;oe(1,0)}}function _g(o,d,m,c,b,T,z,R){var W=ne();try{return zs(o,d,m,c,b,T,z,R)}catch(G){if(ae(W),G!==G+0)throw G;oe(1,0)}}function wg(o,d,m,c,b){var T=ne();try{return Gs(o,d,m,c,b)}catch(z){if(ae(T),z!==z+0)throw z;return oe(1,0),0n}}function bg(o,d,m,c){var b=ne();try{return Fs(o,d,m,c)}catch(T){if(ae(b),T!==T+0)throw T;oe(1,0)}}function $g(o,d,m,c){var b=ne();try{return Hs(o,d,m,c)}catch(T){if(ae(b),T!==T+0)throw T;oe(1,0)}}function vg(o,d,m,c,b,T,z,R,W,G,ie,pe){var me=ne();try{return js(o,d,m,c,b,T,z,R,W,G,ie,pe)}catch(_e){if(ae(me),_e!==_e+0)throw _e;oe(1,0)}}function xg(o,d,m,c,b,T,z,R,W,G,ie){var pe=ne();try{Ws(o,d,m,c,b,T,z,R,W,G,ie)}catch(me){if(ae(pe),me!==me+0)throw me;oe(1,0)}}function Sg(o,d,m,c,b,T,z,R,W,G,ie,pe,me,_e,pt,_i){var Cg=ne();try{Ls(o,d,m,c,b,T,z,R,W,G,ie,pe,me,_e,pt,_i)}catch(wi){if(ae(Cg),wi!==wi+0)throw wi;oe(1,0)}}function Tg(o,d,m,c){var b=ne();try{return Ks(o,d,m,c)}catch(T){if(ae(b),T!==T+0)throw T;oe(1,0)}}function Ig(o,d,m,c,b){var T=ne();try{return Zs(o,d,m,c,b)}catch(z){if(ae(T),z!==z+0)throw z;oe(1,0)}}function kg(o,d,m){var c=ne();try{return Ds(o,d,m)}catch(b){if(ae(c),b!==b+0)throw b;oe(1,0)}}function Eg(o,d,m){var c=ne();try{return Ps(o,d,m)}catch(b){if(ae(c),b!==b+0)throw b;oe(1,0)}}function zg(o,d,m,c){var b=ne();try{Us(o,d,m,c)}catch(T){if(ae(b),T!==T+0)throw T;oe(1,0)}}function Ir(){if(0<Oe)bt=Ir;else if(a)$?.(t),de();else{for(var o=be;0<o.length;)o.shift()(t);0<Oe?bt=Ir:(t.calledRun=!0,C||(de(),$?.(t)))}}return a||(rt=await Le(),Ir()),t.PTR_SIZE=4,V?t:new Promise((o,d)=>{$=o,S=d})}var ip,io,Yg=P(()=>{"use strict";ip=ro,io=globalThis.self?.name?.startsWith("em-pthread"),io&&ro()}),Ti,ga,ao,Re,ap,Er,no,so,Ii,oo,ki,np,Ei,sp,Ma=P(()=>{"use strict";Ra(),Ti=typeof location>"u"?void 0:location.origin,ga=import.meta.url>"file:"&&import.meta.url<"file;",ao=()=>{if(ga){let e=URL;return new URL(new e("ort.bundle.min.mjs",import.meta.url).href,Ti).href}return import.meta.url},Re=ao(),ap=()=>{if(Re&&!Re.startsWith("blob:"))return Re.substring(0,Re.lastIndexOf("/")+1)},Er=(e,t)=>{try{let r=t??Re;return(r?new URL(e,r):new URL(e)).origin===Ti}catch{return!1}},no=(e,t)=>{let r=t??Re;try{return(r?new URL(e,r):new URL(e)).href}catch{return}},so=(e,t)=>`${t??"./"}${e}`,Ii=async e=>{let t=await(await fetch(e,{credentials:"same-origin"})).blob();return URL.createObjectURL(t)},oo=async e=>(await import(e)).default,ki=(Xg(),pr(ep)).default,np=async()=>{if(!Re)throw new Error("Failed to load proxy worker: cannot determine the script source URL.");if(Er(Re))return[void 0,ki()];let e=await Ii(Re);return[e,ki(e)]},Ei=(Yg(),pr(rp)).default,sp=async(e,t,r,i)=>{let a=Ei&&!(e||t);if(a)if(Re)a=Er(Re);else if(i&&!r)a=!0;else throw new Error("cannot determine the script source URL.");if(a)return[void 0,Ei];{let s="ort-wasm-simd-threaded.jsep.mjs",n=e??no(s,t),u=r&&n&&!Er(n,t),l=u?await Ii(n):n??so(s,t);return[u?l:void 0,await oo(l)]}}}),zi,zr,Qt,Ci,uo,lo,po,Na,ye,Nt=P(()=>{"use strict";Ma(),zr=!1,Qt=!1,Ci=!1,uo=()=>{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}},lo=()=>{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}},po=()=>{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}},Na=async e=>{if(zr)return Promise.resolve();if(Qt)throw new Error("multiple calls to 'initializeWebAssembly()' detected.");if(Ci)throw new Error("previous call to 'initializeWebAssembly()' failed.");Qt=!0;let t=e.initTimeout,r=e.numThreads;if(e.simd!==!1){if(e.simd==="relaxed"){if(!po())throw new Error("Relaxed WebAssembly SIMD is not supported in the current environment.")}else if(!lo())throw new Error("WebAssembly SIMD is not supported in the current environment.")}let i=uo();r>1&&!i&&(typeof self<"u"&&!self.crossOriginIsolated&&console.warn("env.wasm.numThreads is set to "+r+", 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=r=1);let a=e.wasmPaths,s=typeof a=="string"?a:void 0,n=a?.mjs,u=n?.href??n,l=a?.wasm,p=l?.href??l,h=e.wasmBinary,[f,g]=await sp(u,s,r>1,!!h||!!p),_=!1,y=[];if(t>0&&y.push(new Promise($=>{setTimeout(()=>{_=!0,$()},t)})),y.push(new Promise(($,S)=>{let v={numThreads:r};if(h)v.wasmBinary=h;else if(p||s)v.locateFile=w=>p??s+w;else if(u&&u.indexOf("blob:")!==0)v.locateFile=w=>new URL(w,u).href;else if(f){let w=ap();w&&(v.locateFile=k=>w+k)}g(v).then(w=>{Qt=!1,zr=!0,zi=w,$(),f&&URL.revokeObjectURL(f)},w=>{Qt=!1,Ci=!0,S(w)})})),await Promise.race(y),_)throw new Error(`WebAssembly backend initializing failed due to timeout: ${t}ms`)},ye=()=>{if(zr&&zi)return zi;throw new Error("WebAssembly is not initialized yet.")}}),Ke,qr,fe,Da=P(()=>{"use strict";Nt(),Ke=(e,t)=>{let r=ye(),i=r.lengthBytesUTF8(e)+1,a=r._malloc(i);return r.stringToUTF8(e,a,i),t.push(a),a},qr=(e,t,r,i)=>{if(typeof e=="object"&&e!==null){if(r.has(e))throw new Error("Circular reference in options");r.add(e)}Object.entries(e).forEach(([a,s])=>{let n=t?t+a:a;if(typeof s=="object")qr(s,n+".",r,i);else if(typeof s=="string"||typeof s=="number")i(n,s.toString());else if(typeof s=="boolean")i(n,s?"1":"0");else throw new Error(`Can't handle extra config type: ${typeof s}`)})},fe=e=>{let t=ye(),r=t.stackSave();try{let i=t.PTR_SIZE,a=t.stackAlloc(2*i);t._OrtGetLastError(a,a+i);let s=Number(t.getValue(a,i===4?"i32":"i64")),n=t.getValue(a+i,"*"),u=n?t.UTF8ToString(n):"";throw new Error(`${e} ERROR_CODE: ${s}, ERROR_MESSAGE: ${u}`)}finally{t.stackRestore(r)}}}),op,Jg=P(()=>{"use strict";Nt(),Da(),op=e=>{let t=ye(),r=0,i=[],a=e||{};try{if(e?.logSeverityLevel===void 0)a.logSeverityLevel=2;else if(typeof e.logSeverityLevel!="number"||!Number.isInteger(e.logSeverityLevel)||e.logSeverityLevel<0||e.logSeverityLevel>4)throw new Error(`log severity level is not valid: ${e.logSeverityLevel}`);if(e?.logVerbosityLevel===void 0)a.logVerbosityLevel=0;else if(typeof e.logVerbosityLevel!="number"||!Number.isInteger(e.logVerbosityLevel))throw new Error(`log verbosity level is not valid: ${e.logVerbosityLevel}`);e?.terminate===void 0&&(a.terminate=!1);let s=0;return e?.tag!==void 0&&(s=Ke(e.tag,i)),r=t._OrtCreateRunOptions(a.logSeverityLevel,a.logVerbosityLevel,!!a.terminate,s),r===0&&fe("Can't create run options."),e?.extra!==void 0&&qr(e.extra,"",new WeakSet,(n,u)=>{let l=Ke(n,i),p=Ke(u,i);t._OrtAddRunConfigEntry(r,l,p)!==0&&fe(`Can't set a run config entry: ${n} - ${u}.`)}),[r,i]}catch(s){throw r!==0&&t._OrtReleaseRunOptions(r),i.forEach(n=>t._free(n)),s}}}),co,ho,fo,Xt,mo,up,e0=P(()=>{"use strict";Nt(),Da(),co=e=>{switch(e){case"disabled":return 0;case"basic":return 1;case"extended":return 2;case"layout":return 3;case"all":return 99;default:throw new Error(`unsupported graph optimization level: ${e}`)}},ho=e=>{switch(e){case"sequential":return 0;case"parallel":return 1;default:throw new Error(`unsupported execution mode: ${e}`)}},fo=e=>{e.extra||(e.extra={}),e.extra.session||(e.extra.session={});let t=e.extra.session;t.use_ort_model_bytes_directly||(t.use_ort_model_bytes_directly="1"),e.executionProviders&&e.executionProviders.some(r=>(typeof r=="string"?r:r.name)==="webgpu")&&(e.enableMemPattern=!1)},Xt=(e,t,r,i)=>{let a=Ke(t,i),s=Ke(r,i);ye()._OrtAddSessionConfigEntry(e,a,s)!==0&&fe(`Can't set a session config entry: ${t} - ${r}.`)},mo=async(e,t,r)=>{let i=t.executionProviders;for(let a of i){let s=typeof a=="string"?a:a.name,n=[];switch(s){case"webnn":if(s="WEBNN",typeof a!="string"){let f=a?.deviceType;f&&Xt(e,"deviceType",f,r)}break;case"webgpu":if(s="JS",typeof a!="string"){let f=a;if(f?.preferredLayout){if(f.preferredLayout!=="NCHW"&&f.preferredLayout!=="NHWC")throw new Error(`preferredLayout must be either 'NCHW' or 'NHWC': ${f.preferredLayout}`);Xt(e,"preferredLayout",f.preferredLayout,r)}}break;case"wasm":case"cpu":continue;default:throw new Error(`not supported execution provider: ${s}`)}let u=Ke(s,r),l=n.length,p=0,h=0;if(l>0){p=ye()._malloc(l*ye().PTR_SIZE),r.push(p),h=ye()._malloc(l*ye().PTR_SIZE),r.push(h);for(let f=0;f<l;f++)ye().setValue(p+f*ye().PTR_SIZE,n[f][0],"*"),ye().setValue(h+f*ye().PTR_SIZE,n[f][1],"*")}await ye()._OrtAppendExecutionProvider(e,u,p,h,l)!==0&&fe(`Can't append execution provider: ${s}.`)}},up=async e=>{let t=ye(),r=0,i=[],a=e||{};fo(a);try{let s=co(a.graphOptimizationLevel??"all"),n=ho(a.executionMode??"sequential"),u=typeof a.logId=="string"?Ke(a.logId,i):0,l=a.logSeverityLevel??2;if(!Number.isInteger(l)||l<0||l>4)throw new Error(`log severity level is not valid: ${l}`);let p=a.logVerbosityLevel??0;if(!Number.isInteger(p)||p<0||p>4)throw new Error(`log verbosity level is not valid: ${p}`);let h=typeof a.optimizedModelFilePath=="string"?Ke(a.optimizedModelFilePath,i):0;if(r=t._OrtCreateSessionOptions(s,!!a.enableCpuMemArena,!!a.enableMemPattern,n,!!a.enableProfiling,0,u,l,p,h),r===0&&fe("Can't create session options."),a.executionProviders&&await mo(r,a,i),a.enableGraphCapture!==void 0){if(typeof a.enableGraphCapture!="boolean")throw new Error(`enableGraphCapture must be a boolean value: ${a.enableGraphCapture}`);Xt(r,"enableGraphCapture",a.enableGraphCapture.toString(),i)}if(a.freeDimensionOverrides)for(let[f,g]of Object.entries(a.freeDimensionOverrides)){if(typeof f!="string")throw new Error(`free dimension override name must be a string: ${f}`);if(typeof g!="number"||!Number.isInteger(g)||g<0)throw new Error(`free dimension override value must be a non-negative integer: ${g}`);let _=Ke(f,i);t._OrtAddFreeDimensionOverride(r,_,g)!==0&&fe(`Can't set a free dimension override: ${f} - ${g}.`)}return a.extra!==void 0&&qr(a.extra,"",new WeakSet,(f,g)=>{Xt(r,f,g,i)}),[r,i]}catch(s){throw r!==0&&t._OrtReleaseSessionOptions(r)!==0&&fe("Can't release session options."),i.forEach(n=>t._free(n)),s}}}),Ct,at,At,Zr,Vr,Pa,Ua,ya,Y=P(()=>{"use strict";Ct=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}`)}},at=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}`)}},At=(e,t)=>{let r=[-1,4,1,1,2,2,4,8,-1,1,2,8,4,8,-1,-1,-1,-1,-1,-1,-1,.5,.5][e],i=typeof t=="number"?t:t.reduce((a,s)=>a*s,1);return r>0?Math.ceil(i*r):void 0},Zr=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}`)}},Vr=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}`)}},Pa=e=>e==="float32"||e==="float16"||e==="int32"||e==="int64"||e==="uint32"||e==="uint8"||e==="bool"||e==="uint4"||e==="int4",Ua=e=>e==="float32"||e==="float16"||e==="int32"||e==="int64"||e==="uint32"||e==="uint64"||e==="int8"||e==="uint8"||e==="bool"||e==="uint4"||e==="int4",ya=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}`)}}}),Wa,lp=P(()=>{"use strict";Ra(),Wa=async e=>{if(typeof e=="string"){let t=await fetch(e);if(!t.ok)throw new Error(`failed to load external data file: ${e}`);let r=t.headers.get("Content-Length"),i=r?parseInt(r,10):0;if(i<1073741824)return new Uint8Array(await t.arrayBuffer());{if(!t.body)throw new Error(`failed to load external data file: ${e}, no response body.`);let a=t.body.getReader(),s;try{s=new ArrayBuffer(i)}catch(u){if(u instanceof RangeError){let l=Math.ceil(i/65536);s=new WebAssembly.Memory({initial:l,maximum:l}).buffer}else throw u}let n=0;for(;;){let{done:u,value:l}=await a.read();if(u)break;let p=l.byteLength;new Uint8Array(s,n,p).set(l),n+=p}return new Uint8Array(s,0,i)}}else return e instanceof Blob?new Uint8Array(await e.arrayBuffer()):e instanceof Uint8Array?e:new Uint8Array(e)}}),go,yo,_o,wo,La,bo,ue,nt=P(()=>{"use strict";Y(),go=["V","I","W","E","F"],yo=(e,t)=>{console.log(`[${go[e]},${new Date().toISOString()}]${t}`)},La=(e,t)=>{_o=e,wo=t},bo=(e,t)=>{let r=Vr(e),i=Vr(_o);r>=i&&yo(r,typeof t=="function"?t():t)},ue=(...e)=>{wo&&bo(...e)}}),$o,qt,B,Gr,dp,pp,cp,te=P(()=>{"use strict";$o=class{static calcMatMulShape(e,t){return e[1]!==t[0]?void 0:[e[0],t[1]]}},qt=class{static calcShape(e,t,r=!1){let i=e.length,a=t.length;if(i===0)return t;if(a===0)return e;let s=Math.max(e.length,t.length),n=new Array(s);if(r){if(i<2||a<2)return;let u=$o.calcMatMulShape([e[i-2],e[i-1]],[t[a-2],t[a-1]]);if(u===void 0)return;[n[s-2],n[s-1]]=u}for(let u=r?3:1;u<=s;u++){let l=i-u<0?1:e[i-u],p=a-u<0?1:t[a-u];if(l!==p&&l>1&&p>1)return;let h=Math.max(l,p);if(l&&p)n[s-u]=Math.max(l,p);else{if(h>1)return;n[s-u]=0}}return n}static isValidBroadcast(e,t){let r=e.length,i=t.length;if(r>i)return!1;for(let a=1;a<=r;a++)if(e[r-a]!==1&&e[r-a]!==t[i-a])return!1;return!0}},B=class Wr{static size(t){return Wr.getSizeFromDimensionRange(t,0,t.length)}static convertShape(t,r=4){let i=t.length;if(i===0)return[];let a=new Array(i),s=i-1;for(;s>=0;){if(t[s]%r===0){a[s]=t[s]/r;break}if(r%t[s]!==0)throw new Error("cannot convert shape");a[s]=1,r/=t[s],s--}for(s--;s>=0;s--)a[s]=t[s];return a}static sizeFromDimension(t,r){if(r<0||r>t.length)throw new Error(`invalid dimension of ${r} for sizeFromDimension as Tensor has ${t.length} dimensions.`);return Wr.getSizeFromDimensionRange(t,r,t.length)}static sizeToDimension(t,r){if(r<0||r>t.length)throw new Error(`invalid dimension of ${r} for sizeToDimension as Tensor has ${t.length} dimensions.`);return Wr.getSizeFromDimensionRange(t,0,r)}static getSizeFromDimensionRange(t,r,i){let a=1;for(let s=r;s<i;s++){if(t[s]<0)throw new Error("cannot get valid size from specified dimension range. Most likely the range contains negative values in them.");a*=Number(t[s])}return a}static computeStrides(t){let r=t.length;if(r===0)return[];if(r===1)return[1];let i=new Array(r);i[r-1]=1,i[r-2]=t[r-1];for(let a=r-3;a>=0;--a)i[a]=i[a+1]*t[a+1];return i}static normalizeAxis(t,r){if(t<-r&&t>=r)throw new Error("unsupported axis for this operation.");return t<0?t+r:t}static normalizeAxes(t,r){return t.map(i=>this.normalizeAxis(i,r??t.length))}static sortBasedOnPerm(t,r){return r?r.map(i=>t[i]):t.slice().reverse()}static padShape(t,r){let i=t.length;return t.map((a,s)=>a+r[s]+r[s+i])}static areEqual(t,r){return t.length!==r.length?!1:t.every((i,a)=>i===r[a])}},Gr=class or{static adjustPoolAttributes(t,r,i,a,s,n){if(!t&&i.length!==r.length-2)throw new Error("length of specified kernel shapes should be 2 less than length of input dimensions");if(t)for(let u=0;u<r.length-2;u++)u>=i.length?i.push(r[u+2]):i[u]=r[u+2];for(let u=0;u<i.length;u++)if(u<a.length){if(a[u]<0)throw new Error("strides should be greater than or equal to 1")}else a.push(1);for(let u=0;u<i.length;u++)if(u<s.length){if(s[u]<0)throw new Error("dilations should be greater than or equal to 1")}else s.push(1);for(let u=0;u<i.length*2;u++)if(u<n.length){if(n[u]<0)throw new Error("pad should be greater than or equal to 1")}else n.push(0);for(let u=0;u<i.length;u++){if(i[u]<=0)throw new Error("kernel shapes need to be greater than 0");if(n[u]>=i[u]||n[u+i.length]>=i[u])throw new Error("pads should be smaller than kernel")}}static adjustPadsBasedOnAutoPad(t,r,i,a,s,n,u){if(u){if(s.length!==2*(t.length-2))throw new Error("length of pads should be twice the length of data dimensions");if(r.length!==t.length-2)throw new Error("length of strides should be the length of data dimensions");if(a.length!==t.length-2)throw new Error("length of kernel shapes should be the length of data dimensions");for(let l=0;l<t.length-2;l++)or.adjustPadAndReturnShape(t[l+(n?1:2)],r[l],i[l],a[l],s,l,l+t.length-2,u)}}static computePoolOutputShape(t,r,i,a,s,n,u){if(r.length<=0)throw new Error("input shape must be of size greater than 0");let l=[r[0],r[1]];return or.computeShapeHelper(t,r,l,i,a,s,n,u),l}static computeConvOutputShape(t,r,i,a,s,n,u){if(t.length<=0||r.length<=0)throw new Error("invalid input tensor dims or invalid filter tensor dims");let l=[t[0],r[0]];return or.computeShapeHelper(!1,t,l,i,a,s,n,u),l}static computeShapeHelper(t,r,i,a,s,n,u,l){if(t)for(let p=0;p<r.length-2;p++)i.push(1);else for(let p=0;p<r.length-2;p++)i.push(or.adjustPadAndReturnShape(r[p+2],a[p],s[p],n[p],u,p,p+r.length-2,l))}static adjustPadAndReturnShape(t,r,i,a,s,n,u,l){let p=i*(a-1)+1;if(l&&l!=="NOTSET")switch(l){case"VALID":return s[n]=0,s[u]=0,Math.floor((t-p)/r+1);case"SAME_LOWER":case"SAME_UPPER":if(i!==1)throw new Error("Dilation not supported for SAME_UPPER or SAME_LOWER");{let h=((t+r-1)/r-1)*r+a-t;return s[n]=Math.floor(l==="SAME_LOWER"?(h+1)/2:h/2),s[u]=h-s[n],Math.floor((t+h-a)/r+1)}default:throw new Error("Unsupported AutoPad type")}else return Math.floor((t+s[n]+s[u]-p)/r+1)}},dp=class{static getShapeOfGemmResult(e,t,r,i,a){if(e.length!==2||r.length!==2)throw new Error("shape need to be of size 2");let s,n,u;t?(s=e[1],n=e[0]):(s=e[0],n=e[1]);let l=-1;if(i?(u=r[0],l=1):(u=r[1],l=0),r[l]!==n)throw new Error("dimension mismatch");if(s<=0||u<=0||n<=0)throw new Error("invalid shape specified");if(a&&!qt.isValidBroadcast(a,[s,u]))throw new Error("gemm: invalid bias shape for broadcast");return[s,u,n]}},pp=-34028234663852886e22,cp=34028234663852886e22}),qa,hp=P(()=>{"use strict";Y(),qa=(e,t)=>new(Zr(t))(e)}),Ai,_a,Oi,vo,Bi,xo,Ri,Mi,Ni,So,fp,t0=P(()=>{"use strict";Y(),nt(),Ai=new Map([["float32",32],["float16",16],["int32",32],["uint32",32],["int64",64],["uint64",64],["int8",8],["uint8",8],["int4",4],["uint4",4]]),_a=(e,t)=>{if(t==="int32")return e;let r=Ai.get(t);if(!r)throw new Error(`WebNN backend does not support data type: ${t}`);let i=r/8;if(e.byteLength%i!==0)throw new Error(`Invalid Uint8Array length - must be a multiple of ${i}.`);let a=e.byteLength/i,s=new(Zr(t))(e.buffer,e.byteOffset,a);switch(t){case"int64":case"uint64":{let n=new Int32Array(a);for(let u=0;u<a;u++){let l=s[u];if(l>2147483647n||l<-2147483648n)throw new Error("Can not convert int64 data to int32 - value out of range.");n[u]=Number(l)}return new Uint8Array(n.buffer)}case"int8":case"uint8":case"uint32":{if(t==="uint32"&&s.some(u=>u>2147483647))throw new Error("Can not convert uint32 data to int32 - value out of range.");let n=Int32Array.from(s,Number);return new Uint8Array(n.buffer)}default:throw new Error(`Unsupported data conversion from ${t} to 'int32'`)}},Oi=(e,t)=>{if(t==="int32")return e;if(e.byteLength%4!==0)throw new Error("Invalid Uint8Array length - must be a multiple of 4 (int32).");let r=e.byteLength/4,i=new Int32Array(e.buffer,e.byteOffset,r);switch(t){case"int64":{let a=BigInt64Array.from(i,BigInt);return new Uint8Array(a.buffer)}case"uint64":{if(i.some(s=>s<0))throw new Error("Can not convert int32 data to uin64 - negative value found.");let a=BigUint64Array.from(i,BigInt);return new Uint8Array(a.buffer)}case"int8":{if(i.some(s=>s<-128||s>127))throw new Error("Can not convert int32 data to int8 - value out of range.");let a=Int8Array.from(i,Number);return new Uint8Array(a.buffer)}case"uint8":{if(i.some(a=>a<0||a>255))throw new Error("Can not convert int32 data to uint8 - value out of range.");return Uint8Array.from(i,Number)}case"uint32":{if(i.some(s=>s<0))throw new Error("Can not convert int32 data to uint32 - negative value found.");let a=Uint32Array.from(i,Number);return new Uint8Array(a.buffer)}default:throw new Error(`Unsupported data conversion from 'int32' to ${t}`)}},vo=1,Bi=()=>vo++,xo=new Map([["int8","int32"],["uint8","int32"],["uint32","int32"],["int64","int32"]]),Ri=(e,t)=>{let r=Ai.get(e);if(!r)throw new Error(`WebNN backend does not support data type: ${e}`);return t.length>0?Math.ceil(t.reduce((i,a)=>i*a)*r/8):0},Mi=class{constructor(e){this.isDataConverted=!1;let{sessionId:t,context:r,tensor:i,dataType:a,shape:s,fallbackDataType:n}=e;this.sessionId=t,this.mlContext=r,this.mlTensor=i,this.dataType=a,this.tensorShape=s,this.fallbackDataType=n}get tensor(){return this.mlTensor}get type(){return this.dataType}get fallbackType(){return this.fallbackDataType}get shape(){return this.tensorShape}get byteLength(){return Ri(this.dataType,this.tensorShape)}destroy(){ue("verbose",()=>"[WebNN] TensorWrapper.destroy"),this.mlTensor.destroy()}write(e){this.mlContext.writeTensor(this.mlTensor,e)}async read(e){if(this.fallbackDataType){let t=await this.mlContext.readTensor(this.mlTensor),r=Oi(new Uint8Array(t),this.dataType);if(e){(e instanceof ArrayBuffer?new Uint8Array(e):new Uint8Array(e.buffer,e.byteOffset,e.byteLength)).set(r);return}else return r.buffer}else return e?this.mlContext.readTensor(this.mlTensor,e):this.mlContext.readTensor(this.mlTensor)}canReuseTensor(e,t,r){return this.mlContext===e&&this.dataType===t&&this.tensorShape.length===r.length&&this.tensorShape.every((i,a)=>i===r[a])}setIsDataConverted(e){this.isDataConverted=e}},Ni=class{constructor(e,t){this.tensorManager=e,this.wrapper=t}get tensorWrapper(){return this.wrapper}releaseTensor(){this.tensorWrapper&&(this.tensorManager.releaseTensor(this.tensorWrapper),this.wrapper=void 0)}async ensureTensor(e,t,r,i){let a=this.tensorManager.getMLContext(e),s=this.tensorManager.getMLOpSupportLimits(e),n;if(!s?.input.dataTypes.includes(t)){if(n=xo.get(t),!n||s?.input.dataTypes.includes(n))throw new Error(`WebNN backend does not support data type: ${t}`);ue("verbose",()=>`[WebNN] TensorIdTracker.ensureTensor: fallback dataType from ${t} to ${n}`)}if(this.wrapper){if(this.wrapper.canReuseTensor(a,t,r))return this.wrapper.tensor;if(i){if(this.wrapper.byteLength!==Ri(t,r))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 u=typeof MLTensorUsage>"u"?void 0:MLTensorUsage.READ|MLTensorUsage.WRITE;return this.wrapper=await this.tensorManager.getCachedTensor(e,t,r,u,!0,!0,n),i&&this.activeUpload&&(this.wrapper.write(this.activeUpload),this.activeUpload=void 0),this.wrapper.tensor}upload(e){let t=e;if(this.wrapper){if(this.wrapper.fallbackType)if(this.wrapper.fallbackType==="int32")t=_a(e,this.wrapper.type),this.wrapper.setIsDataConverted(!0);else throw new Error(`Unsupported fallback data type: ${this.wrapper.fallbackType}`);if(e.byteLength===this.wrapper.byteLength){this.wrapper.write(t);return}else ue("verbose",()=>"Data size does not match tensor size. Releasing tensor."),this.releaseTensor()}this.activeUpload?this.activeUpload.set(t):this.activeUpload=new Uint8Array(t)}async download(e){if(this.activeUpload){let t=this.wrapper?.isDataConverted?Oi(this.activeUpload,this.wrapper?.type):this.activeUpload;if(e){e instanceof ArrayBuffer?new Uint8Array(e).set(t):new Uint8Array(e.buffer,e.byteOffset,e.byteLength).set(t);return}else return t.buffer}if(!this.wrapper)throw new Error("Tensor has not been created.");return e?this.wrapper.read(e):this.wrapper.read()}},So=class{constructor(e){this.backend=e,this.tensorTrackersById=new Map,this.freeTensors=[],this.externalTensors=new Set}getMLContext(e){let t=this.backend.getMLContext(e);if(!t)throw new Error("MLContext not found for session.");return t}getMLOpSupportLimits(e){return this.backend.getMLOpSupportLimits(e)}reserveTensorId(){let e=Bi();return this.tensorTrackersById.set(e,new Ni(this)),e}releaseTensorId(e){let t=this.tensorTrackersById.get(e);t&&(this.tensorTrackersById.delete(e),t.tensorWrapper&&this.releaseTensor(t.tensorWrapper))}async ensureTensor(e,t,r,i,a){ue("verbose",()=>`[WebNN] TensorManager.ensureTensor {tensorId: ${t}, dataType: ${r}, shape: ${i}, copyOld: ${a}}`);let s=this.tensorTrackersById.get(t);if(!s)throw new Error("Tensor not found.");return s.ensureTensor(e,r,i,a)}upload(e,t){let r=this.tensorTrackersById.get(e);if(!r)throw new Error("Tensor not found.");r.upload(t)}async download(e,t){ue("verbose",()=>`[WebNN] TensorManager.download {tensorId: ${e}, dstBuffer: ${t?.byteLength}}`);let r=this.tensorTrackersById.get(e);if(!r)throw new Error("Tensor not found.");return r.download(t)}releaseTensorsForSession(e){for(let t of this.freeTensors)t.sessionId===e&&t.destroy();this.freeTensors=this.freeTensors.filter(t=>t.sessionId!==e)}registerTensor(e,t,r,i){let a=this.getMLContext(e),s=Bi(),n=new Mi({sessionId:e,context:a,tensor:t,dataType:r,shape:i});return this.tensorTrackersById.set(s,new Ni(this,n)),this.externalTensors.add(n),s}async getCachedTensor(e,t,r,i,a,s,n){let u=this.getMLContext(e);for(let[p,h]of this.freeTensors.entries())if(h.canReuseTensor(u,t,r)){ue("verbose",()=>`[WebNN] Reusing tensor {dataType: ${t}, ${n?`fallbackDataType: ${n},`:""} shape: ${r}`);let f=this.freeTensors.splice(p,1)[0];return f.sessionId=e,f}ue("verbose",()=>`[WebNN] MLContext.createTensor {dataType: ${t}, ${n?`fallbackDataType: ${n},`:""} shape: ${r}}`);let l=await u.createTensor({dataType:n??t,shape:r,dimensions:r,usage:i,writable:a,readable:s});return new Mi({sessionId:e,context:u,tensor:l,dataType:t,shape:r,fallbackDataType:n})}releaseTensor(e){this.externalTensors.has(e)&&this.externalTensors.delete(e),this.freeTensors.push(e)}},fp=(...e)=>new So(...e)}),Yt,To,mp,r0=P(()=>{"use strict";Y(),Nt(),hp(),t0(),nt(),Yt=new Map([[1,"float32"],[10,"float16"],[6,"int32"],[12,"uint32"],[7,"int64"],[13,"uint64"],[22,"int4"],[21,"uint4"],[3,"int8"],[2,"uint8"],[9,"uint8"]]),To=(e,t)=>{if(e===t)return!0;if(e===void 0||t===void 0)return!1;let r=Object.keys(e).sort(),i=Object.keys(t).sort();return r.length===i.length&&r.every((a,s)=>a===i[s]&&e[a]===t[a])},mp=class{constructor(e){this.tensorManager=fp(this),this.mlContextBySessionId=new Map,this.sessionIdsByMLContext=new Map,this.mlContextCache=[],this.sessionGraphInputs=new Map,this.sessionGraphOutputs=new Map,this.temporaryGraphInputs=[],this.temporaryGraphOutputs=[],this.temporarySessionTensorIds=new Map,this.mlOpSupportLimitsBySessionId=new Map,La(e.logLevel,!!e.debug)}get currentSessionId(){if(this.activeSessionId===void 0)throw new Error("No active session");return this.activeSessionId}onRunStart(e){ue("verbose",()=>`[WebNN] onRunStart {sessionId: ${e}}`),this.activeSessionId=e}onRunEnd(e){ue("verbose",()=>`[WebNN] onRunEnd {sessionId: ${e}}`);let t=this.temporarySessionTensorIds.get(e);if(t){for(let r of t)ue("verbose",()=>`[WebNN] releasing temporary tensor {tensorId: ${r}}`),this.tensorManager.releaseTensorId(r);this.temporarySessionTensorIds.delete(e),this.activeSessionId=void 0}}async createMLContext(e){if(e instanceof GPUDevice){let r=this.mlContextCache.findIndex(i=>i.gpuDevice===e);if(r!==-1)return this.mlContextCache[r].mlContext;{let i=await navigator.ml.createContext(e);return this.mlContextCache.push({gpuDevice:e,mlContext:i}),i}}else if(e===void 0){let r=this.mlContextCache.findIndex(i=>i.options===void 0&&i.gpuDevice===void 0);if(r!==-1)return this.mlContextCache[r].mlContext;{let i=await navigator.ml.createContext();return this.mlContextCache.push({mlContext:i}),i}}let t=this.mlContextCache.findIndex(r=>To(r.options,e));if(t!==-1)return this.mlContextCache[t].mlContext;{let r=await navigator.ml.createContext(e);return this.mlContextCache.push({options:e,mlContext:r}),r}}registerMLContext(e,t){this.mlContextBySessionId.set(e,t);let r=this.sessionIdsByMLContext.get(t);r||(r=new Set,this.sessionIdsByMLContext.set(t,r)),r.add(e),this.mlOpSupportLimitsBySessionId.has(e)||this.mlOpSupportLimitsBySessionId.set(e,t.opSupportLimits()),this.temporaryGraphInputs.length>0&&(this.sessionGraphInputs.set(e,this.temporaryGraphInputs),this.temporaryGraphInputs=[]),this.temporaryGraphOutputs.length>0&&(this.sessionGraphOutputs.set(e,this.temporaryGraphOutputs),this.temporaryGraphOutputs=[])}onReleaseSession(e){this.sessionGraphInputs.delete(e),this.sessionGraphOutputs.delete(e);let t=this.mlContextBySessionId.get(e);if(!t)return;this.tensorManager.releaseTensorsForSession(e),this.mlContextBySessionId.delete(e),this.mlOpSupportLimitsBySessionId.delete(e);let r=this.sessionIdsByMLContext.get(t);if(r.delete(e),r.size===0){this.sessionIdsByMLContext.delete(t);let i=this.mlContextCache.findIndex(a=>a.mlContext===t);i!==-1&&this.mlContextCache.splice(i,1)}}getMLContext(e){return this.mlContextBySessionId.get(e)}getMLOpSupportLimits(e){return this.mlOpSupportLimitsBySessionId.get(e)}reserveTensorId(){return this.tensorManager.reserveTensorId()}releaseTensorId(e){ue("verbose",()=>`[WebNN] releaseTensorId {tensorId: ${e}}`),this.tensorManager.releaseTensorId(e)}async ensureTensor(e,t,r,i,a){let s=Yt.get(r);if(!s)throw new Error(`Unsupported ONNX data type: ${r}`);return this.tensorManager.ensureTensor(e??this.currentSessionId,t,s,i,a)}async createTemporaryTensor(e,t,r){ue("verbose",()=>`[WebNN] createTemporaryTensor {onnxDataType: ${t}, shape: ${r}}`);let i=Yt.get(t);if(!i)throw new Error(`Unsupported ONNX data type: ${t}`);let a=this.tensorManager.reserveTensorId();await this.tensorManager.ensureTensor(e,a,i,r,!1);let s=this.temporarySessionTensorIds.get(e);return s?s.push(a):this.temporarySessionTensorIds.set(e,[a]),a}uploadTensor(e,t){if(!ye().shouldTransferToMLTensor)throw new Error("Trying to upload to a MLTensor while shouldTransferToMLTensor is false");ue("verbose",()=>`[WebNN] uploadTensor {tensorId: ${e}, data: ${t.byteLength}}`),this.tensorManager.upload(e,t)}async downloadTensor(e,t){return this.tensorManager.download(e,t)}createMLTensorDownloader(e,t){return async()=>{let r=await this.tensorManager.download(e);return qa(r,t)}}registerMLTensor(e,t,r,i){let a=Yt.get(r);if(!a)throw new Error(`Unsupported ONNX data type: ${r}`);let s=this.tensorManager.registerTensor(e,t,a,i);return ue("verbose",()=>`[WebNN] registerMLTensor {tensor: ${t}, dataType: ${a}, dimensions: ${i}} -> {tensorId: ${s}}`),s}registerMLConstant(e,t,r,i,a,s,n=!1){if(!s)throw new Error("External mounted files are not available.");let u=e;e.startsWith("./")&&(u=e.substring(2));let l=s.get(u);if(!l)throw new Error(`File with name ${u} not found in preloaded files.`);if(t+r>l.byteLength)throw new Error("Out of bounds: data offset and length exceed the external file data size.");let p=l.slice(t,t+r).buffer,h;switch(a.dataType){case"float32":h=new Float32Array(p);break;case"float16":h=typeof Float16Array<"u"&&Float16Array.from?new Float16Array(p):new Uint16Array(p);break;case"int32":h=new Int32Array(p);break;case"uint32":h=new Uint32Array(p);break;case"int64":if(n){let f=_a(new Uint8Array(p),"int64");h=new Int32Array(f.buffer),a.dataType="int32"}else h=new BigInt64Array(p);break;case"uint64":h=new BigUint64Array(p);break;case"int8":h=new Int8Array(p);break;case"int4":case"uint4":case"uint8":h=new Uint8Array(p);break;default:throw new Error(`Unsupported data type: ${a.dataType} in creating WebNN Constant from external data.`)}return ue("verbose",()=>`[WebNN] registerMLConstant {dataType: ${a.dataType}, shape: ${a.shape}}} ${n?"(Note: it was int64 data type and registered to int32 as workaround)":""}`),i.constant(a,h)}registerGraphInput(e){this.temporaryGraphInputs.push(e)}registerGraphOutput(e){this.temporaryGraphOutputs.push(e)}isGraphInput(e,t){let r=this.sessionGraphInputs.get(e);return r?r.includes(t):!1}isGraphOutput(e,t){let r=this.sessionGraphOutputs.get(e);return r?r.includes(t):!1}isGraphInputOutputTypeSupported(e,t,r=!0){let i=Yt.get(Ct(t)),a=this.mlOpSupportLimitsBySessionId.get(e);return typeof i>"u"?!1:r?!!a?.input.dataTypes.includes(i):!!a?.output.dataTypes.includes(i)}flush(){}}}),Va=P(()=>{"use strict"}),Di,Cr,Ar,Io,ko,Pi,wa,Eo,gp,i0=P(()=>{"use strict";nt(),Va(),Di=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]]),Cr=[],Ar=e=>Math.ceil(Number(e)/16)*16,Io=e=>{for(let t=0;t<Cr.length;t++){let r=Cr[t];if(e<=r)return r}return Math.ceil(e/16)*16},ko=1,Pi=()=>ko++,wa=async(e,t,r,i)=>{let a=Ar(r),s=e.device.createBuffer({size:a,usage:GPUBufferUsage.COPY_DST|GPUBufferUsage.MAP_READ});try{let n=e.getCommandEncoder();e.endComputePass(),n.copyBufferToBuffer(t,0,s,0,a),e.flush(),await s.mapAsync(GPUMapMode.READ);let u=s.getMappedRange();if(i){let l=i();return l.set(new Uint8Array(u,0,r)),l}else return new Uint8Array(u.slice(0,r))}finally{s.destroy()}},Eo=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[t]of Di)Cr.push(t),this.freeBuffers.set(t,[]),this.freeUniformBuffers.set(t,[]);this.sessionCount=0}upload(e,t){let r=t.buffer,i=t.byteOffset,a=t.byteLength,s=Ar(a),n=this.storageCache.get(e);if(!n)throw new Error("gpu data for uploading does not exist");if(Number(n.originalSize)!==a)throw new Error(`inconsistent data size. gpu data size=${n.originalSize}, data size=${a}`);let u=this.backend.device.createBuffer({mappedAtCreation:!0,size:s,usage:GPUBufferUsage.MAP_WRITE|GPUBufferUsage.COPY_SRC}),l=u.getMappedRange();new Uint8Array(l).set(new Uint8Array(r,i,a)),u.unmap();let p=this.backend.device.createCommandEncoder();p.copyBufferToBuffer(u,0,n.gpuData.buffer,0,s),this.backend.device.queue.submit([p.finish()]),u.destroy(),ue("verbose",()=>`[WebGPU] GpuDataManager.upload(id=${e})`)}memcpy(e,t){let r=this.storageCache.get(e);if(!r)throw new Error("source gpu data for memcpy does not exist");let i=this.storageCache.get(t);if(!i)throw new Error("destination gpu data for memcpy does not exist");if(r.originalSize!==i.originalSize)throw new Error("inconsistent source and destination gpu data size");let a=Ar(r.originalSize),s=this.backend.getCommandEncoder();this.backend.endComputePass(),s.copyBufferToBuffer(r.gpuData.buffer,0,i.gpuData.buffer,0,a)}registerExternalBuffer(e,t,r){let i;if(r){if(i=r[0],e===r[1])return ue("verbose",()=>`[WebGPU] GpuDataManager.registerExternalBuffer(size=${t}) => id=${i}, buffer is the same, skip.`),i;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 i=Pi();return this.storageCache.set(i,{gpuData:{id:i,type:0,buffer:e},originalSize:t}),ue("verbose",()=>`[WebGPU] GpuDataManager.registerExternalBuffer(size=${t}) => id=${i}, registered.`),i}unregisterExternalBuffer(e){e!==void 0&&(this.storageCache.delete(e),ue("verbose",()=>`[WebGPU] GpuDataManager.unregisterExternalBuffer() => id=${e}`))}create(e,t=GPUBufferUsage.STORAGE|GPUBufferUsage.COPY_SRC|GPUBufferUsage.COPY_DST){let r=Io(e),i,a=(t&GPUBufferUsage.STORAGE)===GPUBufferUsage.STORAGE,s=(t&GPUBufferUsage.UNIFORM)===GPUBufferUsage.UNIFORM;if(a||s){let u=(a?this.freeBuffers:this.freeUniformBuffers).get(r);u?u.length>0?i=u.pop():i=this.backend.device.createBuffer({size:r,usage:t}):i=this.backend.device.createBuffer({size:r,usage:t})}else i=this.backend.device.createBuffer({size:r,usage:t});let n={id:Pi(),type:0,buffer:i};return this.storageCache.set(n.id,{gpuData:n,originalSize:Number(e)}),ue("verbose",()=>`[WebGPU] GpuDataManager.create(size=${e}) => id=${n.id}`),n}get(e){return this.storageCache.get(e)?.gpuData}release(e){let t=typeof e=="bigint"?Number(e):e,r=this.storageCache.get(t);if(!r){if(this.storageCache.size===0)return 0;throw new Error("releasing data does not exist")}return ue("verbose",()=>`[WebGPU] GpuDataManager.release(id=${t}), gpuDataId=${r.gpuData.id}`),this.storageCache.delete(t),this.buffersPending.push(r.gpuData.buffer),r.originalSize}async download(e,t){let r=this.storageCache.get(Number(e));if(!r)throw new Error("data does not exist");await wa(this.backend,r.gpuData.buffer,r.originalSize,t)}refreshPendingBuffers(){if(this.buffersPending.length!==0)if(this.backend.sessionStatus==="default"){for(let e of this.buffersPending){let t=Di.get(e.size);if((e.usage&GPUBufferUsage.STORAGE)===GPUBufferUsage.STORAGE){let r=this.freeBuffers.get(e.size)||[];t===void 0||r.length>=t?e.destroy():r.push(e)}else if((e.usage&GPUBufferUsage.UNIFORM)===GPUBufferUsage.UNIFORM){let r=this.freeUniformBuffers.get(e.size)||[];t===void 0||r.length>=t?e.destroy():r.push(e)}else e.destroy()}this.buffersPending=[]}else{let e=this.capturedPendingBuffers.get(this.backend.currentSessionId);e||(e=[],this.capturedPendingBuffers.set(this.backend.currentSessionId,e));for(let t of this.buffersPending)e.push(t);this.buffersPending=[]}}dispose(){this.freeBuffers.forEach(e=>{e.forEach(t=>{t.destroy()})}),this.freeUniformBuffers.forEach(e=>{e.forEach(t=>{t.destroy()})}),this.storageCache.forEach(e=>{e.gpuData.buffer.destroy()}),this.capturedPendingBuffers.forEach(e=>{e.forEach(t=>{t.destroy()})}),this.storageCache=new Map,this.freeBuffers=new Map,this.freeUniformBuffers=new Map,this.capturedPendingBuffers=new Map}onCreateSession(){this.sessionCount+=1}onReleaseSession(e){let t=this.capturedPendingBuffers.get(e);t&&(t.forEach(r=>{r.destroy()}),this.capturedPendingBuffers.delete(e)),this.sessionCount-=1,this.sessionCount===0&&(ue("warning",()=>"[WebGPU] Clearing webgpu buffer cache"),this.storageCache.forEach(r=>{r.gpuData.buffer.destroy()}),this.storageCache=new Map)}},gp=(...e)=>new Eo(...e)}),zo,he,ve=P(()=>{"use strict";zo=class{constructor(e){Object.assign(this,e)}get cacheKey(){return this.key||(this.key=Object.getOwnPropertyNames(this).sort().map(e=>`${this[e]}`).join(";")),this.key}},he=e=>new zo(e)}),Vt,Or,ke,Ce,Z,$e,ba,Lt,yt,j,Jt,M,F,yp,Ga,Co,_p,re=P(()=>{"use strict";Y(),te(),Vt=64,Or=(e,t)=>{if(t===3)throw new Error("vec3 has same alignment as vec4, use vec4 instead");switch(Number(e)){case 10:return t>1?`vec${t}<f16>`:"f16";case 1:return t>1?`vec${t}<f32>`:"f32";case 6:return t>1?`vec${t}<i32>`:"i32";case 12:return t>1?`vec${t}<u32>`:"u32";case 7:if(t>1)throw new Error("currently not supported vecX of uint64 yet");return["vec2<u32>","i32"];case 13:if(t>1)throw new Error("currently not supported vecX of uint64 yet");return["vec2<u32>","u32"];case 9:if(t!==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: ${e}`)}},ke=(e,t=1)=>{let r=Or(e,t);return typeof r=="string"?r:r[0]},Ce=(e,t=1)=>{let r=Or(e,t);return typeof r=="string"?r:r[1]},Z=(...e)=>{let t=[];return e.forEach(r=>{r.length!==0&&t.push({type:12,data:r},{type:12,data:B.computeStrides(r)})}),t},$e=e=>e%4===0?4:e%2===0?2:1,ba=(e="f32",t,r="0")=>!t||t===1?`${e}(${r})`:`vec${t}<${e}>(${r})`,Lt=(e,t,r)=>e==="f32"?r:t===1?`f32(${r})`:`vec${t}<f32>(${r})`,yt=(e,t)=>t===4?`(${e}.x + ${e}.y + ${e}.z + ${e}.w)`:t===2?`(${e}.x + ${e}.y)`:t===3?`(${e}.x + ${e}.y + ${e}.z)`:e,j=(e,t,r,i)=>e.startsWith("uniforms.")&&r>4?typeof t=="string"?i==="f16"?`${e}[(${t}) / 8][(${t}) % 8 / 4][(${t}) % 8 % 4]`:`${e}[(${t}) / 4][(${t}) % 4]`:i==="f16"?`${e}[${Math.floor(t/8)}][${Math.floor(t%8/4)}][${t%8%4}]`:`${e}[${Math.floor(t/4)}][${t%4}]`:r>1?`${e}[${t}]`:e,Jt=(e,t,r,i,a)=>{let s=typeof r=="number",n=s?r:r.length,u=[...new Array(n).keys()],l=n<2?"u32":n<=4?`vec${n}<u32>`:`array<u32, ${n}>`,p=Or(t,a),h=typeof p=="string"?p:p[1],f=typeof p=="string"?p:p[0],g={indices:l,value:h,storage:f,tensor:t},_=N=>typeof N=="string"?N:`${N}u`,y={offsetToIndices:!1,indicesToOffset:!1,broadcastedIndicesToOffset:!1,set:!1,setByIndices:!1,get:!1,getByIndices:!1},$=s?"uniforms.":"",S=`${$}${e}_shape`,v=`${$}${e}_strides`,w="";for(let N=0;N<n-1;N++)w+=`
+    let dim${N} = current / ${j(v,N,n)};
+    let rest${N} = current % ${j(v,N,n)};
+    indices[${N}] = dim${N};
+    current = rest${N};
+    `;w+=`indices[${n-1}] = current;`;let k=n<2?"":`
   fn o2i_${e}(offset: u32) -> ${g.indices} {
     var indices: ${g.indices};
     var current = offset;
     ${w}
     return indices;
-  }`,I=M=>(y.offsetToIndices=!0,n<2?M:`o2i_${e}(${M})`),k=[];if(n>=2)for(let M=n-1;M>=0;M--)k.push(`${j(v,M,n)} * (indices[${M}])`);let C=n<2?"":`
+  }`,I=N=>(y.offsetToIndices=!0,n<2?N:`o2i_${e}(${N})`),E=[];if(n>=2)for(let N=n-1;N>=0;N--)E.push(`${j(v,N,n)} * (indices[${N}])`);let C=n<2?"":`
   fn i2o_${e}(indices: ${g.indices}) -> u32 {
-    return ${k.join("+")};
-  }`,O=M=>(y.indicesToOffset=!0,n<2?M:`i2o_${e}(${M})`),x=(...M)=>n===0?"0u":`${g.indices}(${M.map(_).join(",")})`,D=(M,V)=>n<2?`${M}`:`${j(M,V,n)}`,q=(M,V,J)=>n<2?`${M}=${J};`:`${j(M,V,n)}=${J};`,Q={},L=(M,V)=>{y.broadcastedIndicesToOffset=!0;let J=`${V.name}broadcastedIndicesTo${e}Offset`;if(J in Q)return`${J}(${M})`;let de=[];for(let Te=n-1;Te>=0;Te--){let Be=V.indicesGet("outputIndices",Te+V.rank-n);de.push(`${D(v,Te)} * (${Be} % ${D(S,Te)})`)}return Q[J]=`fn ${J}(outputIndices: ${V.type.indices}) -> u32 {
+    return ${E.join("+")};
+  }`,O=N=>(y.indicesToOffset=!0,n<2?N:`i2o_${e}(${N})`),x=(...N)=>n===0?"0u":`${g.indices}(${N.map(_).join(",")})`,D=(N,V)=>n<2?`${N}`:`${j(N,V,n)}`,L=(N,V,J)=>n<2?`${N}=${J};`:`${j(N,V,n)}=${J};`,Q={},q=(N,V)=>{y.broadcastedIndicesToOffset=!0;let J=`${V.name}broadcastedIndicesTo${e}Offset`;if(J in Q)return`${J}(${N})`;let de=[];for(let Ie=n-1;Ie>=0;Ie--){let Be=V.indicesGet("outputIndices",Ie+V.rank-n);de.push(`${D(v,Ie)} * (${Be} % ${D(S,Ie)})`)}return Q[J]=`fn ${J}(outputIndices: ${V.type.indices}) -> u32 {
              return ${de.length>0?de.join("+"):"0u"};
-           }`,`${J}(${M})`},K=(M,V)=>(()=>{if(g.storage===g.value)return`${e}[${M}]=${V};`;if(g.storage==="vec2<u32>"&&g.value==="i32")return`${e}[${M}]=vec2<u32>(u32(${V}), select(0u, 0xFFFFFFFFu, ${V} < 0));`;if(g.storage==="vec2<u32>"&&g.value==="u32")return`${e}[${M}]=vec2<u32>(u32(${V}), 0u);`;if(g.storage==="u32"&&g.value==="vec4<bool>")return`${e}[${M}]=dot(vec4<u32>(0x1, 0x100, 0x10000, 0x1000000), vec4<u32>(${V}));`;throw new Error(`not supported combination of storage type ${g.storage} and value type ${g.value} yet`)})(),se=M=>(()=>{if(g.storage===g.value)return`${e}[${M}]`;if(g.storage==="vec2<u32>"&&g.value==="i32")return`i32(${e}[${M}].x)`;if(g.storage==="vec2<u32>"&&g.value==="u32")return`u32(${e}[${M}].x)`;if(g.storage==="u32"&&g.value==="vec4<bool>")return`vec4<bool>(bool(${e}[${M}] & 0xFFu), bool(${e}[${M}] & 0xFF00u), bool(${e}[${M}] & 0xFF0000u), bool(${e}[${M}] & 0xFF000000u))`;throw new Error(`not supported combination of storage type ${g.storage} and value type ${g.value} yet`)})(),A=n<2?"":`
+           }`,`${J}(${N})`},K=(N,V)=>(()=>{if(g.storage===g.value)return`${e}[${N}]=${V};`;if(g.storage==="vec2<u32>"&&g.value==="i32")return`${e}[${N}]=vec2<u32>(u32(${V}), select(0u, 0xFFFFFFFFu, ${V} < 0));`;if(g.storage==="vec2<u32>"&&g.value==="u32")return`${e}[${N}]=vec2<u32>(u32(${V}), 0u);`;if(g.storage==="u32"&&g.value==="vec4<bool>")return`${e}[${N}]=dot(vec4<u32>(0x1, 0x100, 0x10000, 0x1000000), vec4<u32>(${V}));`;throw new Error(`not supported combination of storage type ${g.storage} and value type ${g.value} yet`)})(),se=N=>(()=>{if(g.storage===g.value)return`${e}[${N}]`;if(g.storage==="vec2<u32>"&&g.value==="i32")return`i32(${e}[${N}].x)`;if(g.storage==="vec2<u32>"&&g.value==="u32")return`u32(${e}[${N}].x)`;if(g.storage==="u32"&&g.value==="vec4<bool>")return`vec4<bool>(bool(${e}[${N}] & 0xFFu), bool(${e}[${N}] & 0xFF00u), bool(${e}[${N}] & 0xFF0000u), bool(${e}[${N}] & 0xFF000000u))`;throw new Error(`not supported combination of storage type ${g.storage} and value type ${g.value} yet`)})(),A=n<2?"":`
   fn get_${e}ByIndices(indices: ${g.indices}) -> ${h} {
     return ${se(`i2o_${e}(indices)`)};
-  }`,U=n<2?"":(()=>{let M=u.map(J=>`d${J}: u32`).join(", "),V=u.map(J=>`d${J}`).join(", ");return`
-  fn get_${e}(${M}) -> ${h} {
+  }`,U=n<2?"":(()=>{let N=u.map(J=>`d${J}: u32`).join(", "),V=u.map(J=>`d${J}`).join(", ");return`
+  fn get_${e}(${N}) -> ${h} {
     return get_${e}ByIndices(${x(V)});
-  }`})(),ee=(...M)=>{if(M.length!==n)throw new Error(`indices length must be ${n}`);let V=M.map(_).join(",");return n===0?se("0u"):n===1?se(V[0]):(y.get=!0,y.getByIndices=!0,y.indicesToOffset=!0,`get_${e}(${V})`)},X=M=>n<2?se(M):(y.getByIndices=!0,y.indicesToOffset=!0,`get_${e}ByIndices(${M})`),H=n<2?"":`
+  }`})(),ee=(...N)=>{if(N.length!==n)throw new Error(`indices length must be ${n}`);let V=N.map(_).join(",");return n===0?se("0u"):n===1?se(V[0]):(y.get=!0,y.getByIndices=!0,y.indicesToOffset=!0,`get_${e}(${V})`)},X=N=>n<2?se(N):(y.getByIndices=!0,y.indicesToOffset=!0,`get_${e}ByIndices(${N})`),H=n<2?"":`
   fn set_${e}ByIndices(indices: ${g.indices}, value: ${h}) {
     ${K(`i2o_${e}(indices)`,"value")}
-  }`,le=n<2?"":(()=>{let M=u.map(J=>`d${J}: u32`).join(", "),V=u.map(J=>`d${J}`).join(", ");return`
-  fn set_${e}(${M}, value: ${h}) {
+  }`,le=n<2?"":(()=>{let N=u.map(J=>`d${J}: u32`).join(", "),V=u.map(J=>`d${J}`).join(", ");return`
+  fn set_${e}(${N}, value: ${h}) {
     set_${e}ByIndices(${x(V)}, value);
-  }`})();return{impl:()=>{let M=[],V=!1;return y.offsetToIndices&&(M.push(E),V=!0),y.indicesToOffset&&(M.push(C),V=!0),y.broadcastedIndicesToOffset&&(Object.values(Q).forEach(J=>M.push(J)),V=!0),y.set&&(M.push(le),V=!0),y.setByIndices&&(M.push(H),V=!0),y.get&&(M.push(U),V=!0),y.getByIndices&&(M.push(A),V=!0),!s&&V&&M.unshift(`const ${S} = ${g.indices}(${r.join(",")});`,`const ${v} = ${g.indices}(${B.computeStrides(r).join(",")});`),M.join(`
-`)},type:g,offsetToIndices:I,indicesToOffset:O,broadcastedIndicesToOffset:L,indices:x,indicesGet:D,indicesSet:q,set:(...M)=>{if(M.length!==n+1)throw new Error(`indices length must be ${n}`);let V=M[n];if(typeof V!="string")throw new Error("value must be string");let J=M.slice(0,n).map(_).join(",");return n===0?K("0u",V):n===1?K(J[0],V):(y.set=!0,y.setByIndices=!0,y.indicesToOffset=!0,`set_${e}(${J}, ${V})`)},setByOffset:K,setByIndices:(M,V)=>n<2?K(M,V):(y.setByIndices=!0,y.indicesToOffset=!0,`set_${e}ByIndices(${M}, ${V});`),get:ee,getByOffset:se,getByIndices:X,usage:i,name:e,strides:v,shape:S,rank:n}},N=(e,t,r,i=1)=>Jt(e,t,r,"input",i),F=(e,t,r,i=1)=>Jt(e,t,r,"output",i),yp=(e,t,r)=>Jt(e,t,r,"atomicOutput",1),Ga=(e,t,r,i=1)=>Jt(e,t,r,"internal",i),Co=class{constructor(e,t){this.normalizedDispatchGroup=e,this.limits=t,this.internalVariables=[],this.variables=[],this.uniforms=[],this.variableIndex=0}guardAgainstOutOfBoundsWorkgroupSizes(e){return`if (global_idx >= ${typeof e=="number"?`${e}u`:e}) { return; }`}mainStart(e=Vt){let t=typeof e=="number"?e:e[0],r=typeof e=="number"?1:e[1],i=typeof e=="number"?1:e[2];if(t>this.limits.maxComputeWorkgroupSizeX||r>this.limits.maxComputeWorkgroupSizeY||i>this.limits.maxComputeWorkgroupSizeZ)throw new Error(`workgroup size [${t}, ${r}, ${i}] exceeds the maximum workgroup size [${this.limits.maxComputeWorkgroupSizeX}, ${this.limits.maxComputeWorkgroupSizeY}, ${this.limits.maxComputeWorkgroupSizeZ}].`);if(t*r*i>this.limits.maxComputeInvocationsPerWorkgroup)throw new Error(`workgroup size [${t}, ${r}, ${i}] exceeds the maximum workgroup invocations ${this.limits.maxComputeInvocationsPerWorkgroup}.`);let a=this.normalizedDispatchGroup[1]===1&&this.normalizedDispatchGroup[2]===1,s=a?`@builtin(global_invocation_id) global_id : vec3<u32>,
+  }`})();return{impl:()=>{let N=[],V=!1;return y.offsetToIndices&&(N.push(k),V=!0),y.indicesToOffset&&(N.push(C),V=!0),y.broadcastedIndicesToOffset&&(Object.values(Q).forEach(J=>N.push(J)),V=!0),y.set&&(N.push(le),V=!0),y.setByIndices&&(N.push(H),V=!0),y.get&&(N.push(U),V=!0),y.getByIndices&&(N.push(A),V=!0),!s&&V&&N.unshift(`const ${S} = ${g.indices}(${r.join(",")});`,`const ${v} = ${g.indices}(${B.computeStrides(r).join(",")});`),N.join(`
+`)},type:g,offsetToIndices:I,indicesToOffset:O,broadcastedIndicesToOffset:q,indices:x,indicesGet:D,indicesSet:L,set:(...N)=>{if(N.length!==n+1)throw new Error(`indices length must be ${n}`);let V=N[n];if(typeof V!="string")throw new Error("value must be string");let J=N.slice(0,n).map(_).join(",");return n===0?K("0u",V):n===1?K(J[0],V):(y.set=!0,y.setByIndices=!0,y.indicesToOffset=!0,`set_${e}(${J}, ${V})`)},setByOffset:K,setByIndices:(N,V)=>n<2?K(N,V):(y.setByIndices=!0,y.indicesToOffset=!0,`set_${e}ByIndices(${N}, ${V});`),get:ee,getByOffset:se,getByIndices:X,usage:i,name:e,strides:v,shape:S,rank:n}},M=(e,t,r,i=1)=>Jt(e,t,r,"input",i),F=(e,t,r,i=1)=>Jt(e,t,r,"output",i),yp=(e,t,r)=>Jt(e,t,r,"atomicOutput",1),Ga=(e,t,r,i=1)=>Jt(e,t,r,"internal",i),Co=class{constructor(e,t){this.normalizedDispatchGroup=e,this.limits=t,this.internalVariables=[],this.variables=[],this.uniforms=[],this.variableIndex=0}guardAgainstOutOfBoundsWorkgroupSizes(e){return`if (global_idx >= ${typeof e=="number"?`${e}u`:e}) { return; }`}mainStart(e=Vt){let t=typeof e=="number"?e:e[0],r=typeof e=="number"?1:e[1],i=typeof e=="number"?1:e[2];if(t>this.limits.maxComputeWorkgroupSizeX||r>this.limits.maxComputeWorkgroupSizeY||i>this.limits.maxComputeWorkgroupSizeZ)throw new Error(`workgroup size [${t}, ${r}, ${i}] exceeds the maximum workgroup size [${this.limits.maxComputeWorkgroupSizeX}, ${this.limits.maxComputeWorkgroupSizeY}, ${this.limits.maxComputeWorkgroupSizeZ}].`);if(t*r*i>this.limits.maxComputeInvocationsPerWorkgroup)throw new Error(`workgroup size [${t}, ${r}, ${i}] exceeds the maximum workgroup invocations ${this.limits.maxComputeInvocationsPerWorkgroup}.`);let a=this.normalizedDispatchGroup[1]===1&&this.normalizedDispatchGroup[2]===1,s=a?`@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>,
@@ -51,13 +51,13 @@ var Ag=Object.defineProperty;var St=(e=>typeof require<"u"?require:typeof Proxy<
       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=t=>[12,10,1,6][["u32","f16","f32","i32"].indexOf(t)];return this.uniforms.map(t=>[e(t.type),t.length??1])}},_p=(e,t)=>new Co(e,t)}),Ao,Ui,Oo,Bo,Ro,No,Me,wp,bp,wt=P(()=>{"use strict";Y(),te(),ve(),re(),Ao=(e,t)=>{if(!e||e.length!==1)throw new Error("Transpose requires 1 input.");if(t.length!==0&&t.length!==e[0].dims.length)throw new Error(`perm size ${t.length} does not match input rank ${e[0].dims.length}`)},Ui=(e,t)=>t.length!==0?t:[...new Array(e).keys()].reverse(),Oo=(e,t)=>B.sortBasedOnPerm(e,Ui(e.length,t)),Bo=(e,t,r,i)=>{let a=`fn perm(i: ${i.type.indices}) -> ${r.type.indices} {
-    var a: ${r.type.indices};`;for(let s=0;s<t;++s)a+=`a[${e[s]}]=i[${s}];`;return a+="return a;}"},Ro=(e,t)=>{let r=[],i=[];for(let a=0;a<e.length;++a)e[a]!==1&&r.push(e[a]),e[t[a]]!==1&&i.push(t[a]);return{newShape:r,newPerm:i}},No=(e,t)=>{let r=0;for(let i=0;i<e.length;++i)if(t[e[i]]!==1){if(e[i]<r)return!1;r=e[i]}return!0},Me=(e,t)=>{let r=e.dataType,i=e.dims.length,a=Ui(i,t),s=Oo(e.dims,a),n=e.dims,u=s,l=i<2||No(a,e.dims),p;if(l)return p=y=>{let $=N("input",r,n,4),S=F("output",r,u,4);return`
+`)}get variablesInfo(){if(this.uniforms.length===0)return;let e=t=>[12,10,1,6][["u32","f16","f32","i32"].indexOf(t)];return this.uniforms.map(t=>[e(t.type),t.length??1])}},_p=(e,t)=>new Co(e,t)}),Ao,Ui,Oo,Bo,Ro,Mo,Ne,wp,bp,wt=P(()=>{"use strict";Y(),te(),ve(),re(),Ao=(e,t)=>{if(!e||e.length!==1)throw new Error("Transpose requires 1 input.");if(t.length!==0&&t.length!==e[0].dims.length)throw new Error(`perm size ${t.length} does not match input rank ${e[0].dims.length}`)},Ui=(e,t)=>t.length!==0?t:[...new Array(e).keys()].reverse(),Oo=(e,t)=>B.sortBasedOnPerm(e,Ui(e.length,t)),Bo=(e,t,r,i)=>{let a=`fn perm(i: ${i.type.indices}) -> ${r.type.indices} {
+    var a: ${r.type.indices};`;for(let s=0;s<t;++s)a+=`a[${e[s]}]=i[${s}];`;return a+="return a;}"},Ro=(e,t)=>{let r=[],i=[];for(let a=0;a<e.length;++a)e[a]!==1&&r.push(e[a]),e[t[a]]!==1&&i.push(t[a]);return{newShape:r,newPerm:i}},Mo=(e,t)=>{let r=0;for(let i=0;i<e.length;++i)if(t[e[i]]!==1){if(e[i]<r)return!1;r=e[i]}return!0},Ne=(e,t)=>{let r=e.dataType,i=e.dims.length,a=Ui(i,t),s=Oo(e.dims,a),n=e.dims,u=s,l=i<2||Mo(a,e.dims),p;if(l)return p=y=>{let $=M("input",r,n,4),S=F("output",r,u,4);return`
   ${y.registerUniform("output_size","u32").declareVariables($,S)}
   ${y.mainStart()}
     ${y.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.output_size")}
     output[global_idx] = input[global_idx];
-  }`},{name:"TransposeCopy",shaderCache:{inputDependencies:["type"]},getRunData:()=>{let y=B.size(s);return{outputs:[{dims:s,dataType:e.dataType}],dispatchGroup:{x:Math.ceil(y/64/4)},programUniforms:[{type:12,data:Math.ceil(y/4)}]}},getShaderSource:p};let{newShape:h,newPerm:f}=Ro(e.dims,a),g=B.areEqual(f,[2,3,1]),_=B.areEqual(f,[3,1,2]);if(h.length===2||g||_){n=g?[h[0],h[1]*h[2]]:_?[h[0]*h[1],h[2]]:h,u=[n[1],n[0]];let y=16;return p=$=>{let S=N("a",r,n.length),v=F("output",r,u.length);return`
+  }`},{name:"TransposeCopy",shaderCache:{inputDependencies:["type"]},getRunData:()=>{let y=B.size(s);return{outputs:[{dims:s,dataType:e.dataType}],dispatchGroup:{x:Math.ceil(y/64/4)},programUniforms:[{type:12,data:Math.ceil(y/4)}]}},getShaderSource:p};let{newShape:h,newPerm:f}=Ro(e.dims,a),g=B.areEqual(f,[2,3,1]),_=B.areEqual(f,[3,1,2]);if(h.length===2||g||_){n=g?[h[0],h[1]*h[2]]:_?[h[0]*h[1],h[2]]:h,u=[n[1],n[0]];let y=16;return p=$=>{let S=M("a",r,n.length),v=F("output",r,u.length);return`
   ${$.registerUniform("output_size","u32").declareVariables(S,v)}
   var<workgroup> tile : array<array<${v.type.value}, ${y+1}>, ${y}>;
   ${$.mainStart([y,y,1])}
@@ -76,7 +76,7 @@ var Ag=Object.defineProperty;var St=(e=>typeof require<"u"?require:typeof Proxy<
     if (output_row < uniforms.output_shape[0] && output_col < uniforms.output_shape[1]) {
       ${v.setByIndices(`${v.type.indices}(output_row, output_col)`,"tile[local_id.x][local_id.y]")}
     }
-  }`},{name:"TransposeShared",shaderCache:{inputDependencies:["type"]},getRunData:()=>{let $=B.size(s);return{outputs:[{dims:s,dataType:e.dataType}],dispatchGroup:{x:Math.ceil(u[1]/y),y:Math.ceil(u[0]/y)},programUniforms:[{type:12,data:$},...Z(n,u)]}},getShaderSource:p}}return p=y=>{let $=N("a",r,n.length),S=F("output",r,u.length);return`
+  }`},{name:"TransposeShared",shaderCache:{inputDependencies:["type"]},getRunData:()=>{let $=B.size(s);return{outputs:[{dims:s,dataType:e.dataType}],dispatchGroup:{x:Math.ceil(u[1]/y),y:Math.ceil(u[0]/y)},programUniforms:[{type:12,data:$},...Z(n,u)]}},getShaderSource:p}}return p=y=>{let $=M("a",r,n.length),S=F("output",r,u.length);return`
   ${y.registerUniform("output_size","u32").declareVariables($,S)}
 
   ${Bo(a,i,$,S)}
@@ -88,7 +88,7 @@ var Ag=Object.defineProperty;var St=(e=>typeof require<"u"?require:typeof Proxy<
     let aIndices = perm(indices);
 
     ${S.setByOffset("global_idx",$.getByIndices("aIndices"))}
-  }`},{name:"Transpose",shaderCache:{hint:`${t}`,inputDependencies:["rank"]},getRunData:()=>{let y=B.size(s);return{outputs:[{dims:s,dataType:e.dataType}],dispatchGroup:{x:Math.ceil(y/64)},programUniforms:[{type:12,data:y},...Z(n,u)]}},getShaderSource:p}},wp=(e,t)=>{Ao(e.inputs,t.perm),e.compute(Me(e.inputs[0],t.perm))},bp=e=>he({perm:e.perm})}),Mo,Do,Po,Uo,Wo,qo,Lo,Vo,Go,Fo,Ve,$p,vp,xp,Sp,Tp,Ip,Ep,kp,zp,Cp,a0=P(()=>{"use strict";Y(),te(),re(),Fa(),wt(),Mo={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"},Do={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"},Po={max:"_A[offset]",min:"_A[offset]",mean:"0",sum:"0",prod:"1",sumSquare:"0",logSumExp:"0",l1:"0",l2:"0",logSum:"0"},Uo={max:"bestValue",min:"bestValue",sum:"bestValue",prod:"bestValue",sumSquare:"bestValue",logSumExp:"log(bestValue)",l1:"bestValue",l2:"sqrt(bestValue)",logSum:"log(bestValue)"},Wo=(e,t)=>{let r=[];for(let i=t-e;i<t;++i)r.push(i);return r},qo=(e,t)=>{let r=[],i=e.length;for(let s=0;s<i;s++)t.indexOf(s)===-1&&r.push(e[s]);let a=t.map(s=>e[s]);return[r,a]},Lo=(e,t)=>{let r=e.length+t.length,i=[],a=0;for(let s=0;s<r;s++)t.indexOf(s)===-1?i.push(e[a++]):i.push(1);return i},Vo=(e,t)=>{for(let r=0;r<e.length;++r)if(e[e.length-r-1]!==t-1-r)return!1;return!0},Go=(e,t)=>{let r=[];if(!Vo(e,t)){for(let i=0;i<t;++i)e.indexOf(i)===-1&&r.push(i);e.forEach(i=>r.push(i))}return r},Fo=(e,t,r,i,a,s,n)=>{let u=r[0].dims,l=B.size(s),p=B.size(n),h=N("_A",r[0].dataType,u),f=F("output",a,s),g=64;l===1&&(g=256);let _=`
+  }`},{name:"Transpose",shaderCache:{hint:`${t}`,inputDependencies:["rank"]},getRunData:()=>{let y=B.size(s);return{outputs:[{dims:s,dataType:e.dataType}],dispatchGroup:{x:Math.ceil(y/64)},programUniforms:[{type:12,data:y},...Z(n,u)]}},getShaderSource:p}},wp=(e,t)=>{Ao(e.inputs,t.perm),e.compute(Ne(e.inputs[0],t.perm))},bp=e=>he({perm:e.perm})}),No,Do,Po,Uo,Wo,Lo,qo,Vo,Go,Fo,Ve,$p,vp,xp,Sp,Tp,Ip,kp,Ep,zp,Cp,a0=P(()=>{"use strict";Y(),te(),re(),Fa(),wt(),No={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"},Do={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"},Po={max:"_A[offset]",min:"_A[offset]",mean:"0",sum:"0",prod:"1",sumSquare:"0",logSumExp:"0",l1:"0",l2:"0",logSum:"0"},Uo={max:"bestValue",min:"bestValue",sum:"bestValue",prod:"bestValue",sumSquare:"bestValue",logSumExp:"log(bestValue)",l1:"bestValue",l2:"sqrt(bestValue)",logSum:"log(bestValue)"},Wo=(e,t)=>{let r=[];for(let i=t-e;i<t;++i)r.push(i);return r},Lo=(e,t)=>{let r=[],i=e.length;for(let s=0;s<i;s++)t.indexOf(s)===-1&&r.push(e[s]);let a=t.map(s=>e[s]);return[r,a]},qo=(e,t)=>{let r=e.length+t.length,i=[],a=0;for(let s=0;s<r;s++)t.indexOf(s)===-1?i.push(e[a++]):i.push(1);return i},Vo=(e,t)=>{for(let r=0;r<e.length;++r)if(e[e.length-r-1]!==t-1-r)return!1;return!0},Go=(e,t)=>{let r=[];if(!Vo(e,t)){for(let i=0;i<t;++i)e.indexOf(i)===-1&&r.push(i);e.forEach(i=>r.push(i))}return r},Fo=(e,t,r,i,a,s,n)=>{let u=r[0].dims,l=B.size(s),p=B.size(n),h=M("_A",r[0].dataType,u),f=F("output",a,s),g=64;l===1&&(g=256);let _=`
           var<workgroup> aBestValues : array<f32, ${g}>;
        `,y=$=>`
         ${$.registerUniform("reduceSize","u32").declareVariables(h,f)}
@@ -105,7 +105,7 @@ var Ag=Object.defineProperty;var St=(e=>typeof require<"u"?require:typeof Proxy<
           let Length = uniforms.reduceSize;
           for (var k = local_idx; k < Length; k = k + ${g}) {
            let candidate = f32(${h.getByOffset("offset + k")});
-           bestValue = ${Mo[i]};
+           bestValue = ${No[i]};
           }
           aBestValues[local_idx] = bestValue;
           workgroupBarrier();
@@ -126,11 +126,11 @@ var Ag=Object.defineProperty;var St=(e=>typeof require<"u"?require:typeof Proxy<
          if (local_idx == 0u) {
           ${f.setByOffset("outputIndex",`${i==="mean"?`${f.type.storage}(bestValue / f32(uniforms.reduceSize))`:`${f.type.storage}(${Uo[i]})`}`)};
          }
-        }`;return{name:e,shaderCache:{hint:`${t};${g}`,inputDependencies:["type"]},getShaderSource:y,getRunData:()=>({outputs:[{dims:s,dataType:a}],dispatchGroup:{x:l},programUniforms:[{type:12,data:p}]})}},Ve=(e,t,r,i)=>{let a=e.inputs.length===1?r:$a(e.inputs,r),s=a.axes;s.length===0&&!a.noopWithEmptyAxes&&(s=e.inputs[0].dims.map((_,y)=>y));let n=B.normalizeAxes(s,e.inputs[0].dims.length),u=n,l=e.inputs[0],p=Go(u,e.inputs[0].dims.length);p.length>0&&(l=e.compute(Me(e.inputs[0],p),{inputs:[0],outputs:[-1]})[0],u=Wo(u.length,l.dims.length));let[h,f]=qo(l.dims,u),g=h;a.keepDims&&(g=Lo(h,n)),e.compute(Fo(t,a.cacheKey,[l],i,e.inputs[0].dataType,g,f),{inputs:[l]})},$p=(e,t)=>{Ve(e,"ReduceMeanShared",t,"mean")},vp=(e,t)=>{Ve(e,"ReduceL1Shared",t,"l1")},xp=(e,t)=>{Ve(e,"ReduceL2Shared",t,"l2")},Sp=(e,t)=>{Ve(e,"ReduceLogSumExpShared",t,"logSumExp")},Tp=(e,t)=>{Ve(e,"ReduceMaxShared",t,"max")},Ip=(e,t)=>{Ve(e,"ReduceMinShared",t,"min")},Ep=(e,t)=>{Ve(e,"ReduceProdShared",t,"prod")},kp=(e,t)=>{Ve(e,"ReduceSumShared",t,"sum")},zp=(e,t)=>{Ve(e,"ReduceSumSquareShared",t,"sumSquare")},Cp=(e,t)=>{Ve(e,"ReduceLogSumShared",t,"logSum")}}),Ge,Ho,Fr,$a,Fe,jo,Ko,Zo,Qo,Xo,Yo,Jo,eu,tu,ru,He,Ap,Op,Bp,Rp,Np,Mp,Dp,Pp,Up,Wp,Fa=P(()=>{"use strict";Y(),te(),ve(),re(),a0(),Ge=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.")},Ho=e=>["","",`var value = ${e.getByIndices("input_indices")};`,""],Fr=(e,t,r,i,a,s,n=!1,u=!1)=>{let l=[],p=r[0].dims,h=p.length,f=B.normalizeAxes(a,h),g=!u&&f.length===0;p.forEach(($,S)=>{g||f.indexOf(S)>=0?n&&l.push(1):l.push($)});let _=l.length,y=B.size(l);return{name:e,shaderCache:t,getShaderSource:$=>{let S=[],v=N("_A",r[0].dataType,h),w=F("output",s,_),E=i(v,w,f),I=E[2];for(let k=0,C=0;k<h;k++)g||f.indexOf(k)>=0?(n&&C++,I=`for(var j${k}: u32 = 0; j${k} < ${p[k]}; j${k}++) {
-                  ${E[2].includes("last_index")?`let last_index = j${k};`:""}
-                  ${v.indicesSet("input_indices",k,`j${k}`)}
+        }`;return{name:e,shaderCache:{hint:`${t};${g}`,inputDependencies:["type"]},getShaderSource:y,getRunData:()=>({outputs:[{dims:s,dataType:a}],dispatchGroup:{x:l},programUniforms:[{type:12,data:p}]})}},Ve=(e,t,r,i)=>{let a=e.inputs.length===1?r:$a(e.inputs,r),s=a.axes;s.length===0&&!a.noopWithEmptyAxes&&(s=e.inputs[0].dims.map((_,y)=>y));let n=B.normalizeAxes(s,e.inputs[0].dims.length),u=n,l=e.inputs[0],p=Go(u,e.inputs[0].dims.length);p.length>0&&(l=e.compute(Ne(e.inputs[0],p),{inputs:[0],outputs:[-1]})[0],u=Wo(u.length,l.dims.length));let[h,f]=Lo(l.dims,u),g=h;a.keepDims&&(g=qo(h,n)),e.compute(Fo(t,a.cacheKey,[l],i,e.inputs[0].dataType,g,f),{inputs:[l]})},$p=(e,t)=>{Ve(e,"ReduceMeanShared",t,"mean")},vp=(e,t)=>{Ve(e,"ReduceL1Shared",t,"l1")},xp=(e,t)=>{Ve(e,"ReduceL2Shared",t,"l2")},Sp=(e,t)=>{Ve(e,"ReduceLogSumExpShared",t,"logSumExp")},Tp=(e,t)=>{Ve(e,"ReduceMaxShared",t,"max")},Ip=(e,t)=>{Ve(e,"ReduceMinShared",t,"min")},kp=(e,t)=>{Ve(e,"ReduceProdShared",t,"prod")},Ep=(e,t)=>{Ve(e,"ReduceSumShared",t,"sum")},zp=(e,t)=>{Ve(e,"ReduceSumSquareShared",t,"sumSquare")},Cp=(e,t)=>{Ve(e,"ReduceLogSumShared",t,"logSum")}}),Ge,Ho,Fr,$a,Fe,jo,Ko,Zo,Qo,Xo,Yo,Jo,eu,tu,ru,He,Ap,Op,Bp,Rp,Mp,Np,Dp,Pp,Up,Wp,Fa=P(()=>{"use strict";Y(),te(),ve(),re(),a0(),Ge=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.")},Ho=e=>["","",`var value = ${e.getByIndices("input_indices")};`,""],Fr=(e,t,r,i,a,s,n=!1,u=!1)=>{let l=[],p=r[0].dims,h=p.length,f=B.normalizeAxes(a,h),g=!u&&f.length===0;p.forEach(($,S)=>{g||f.indexOf(S)>=0?n&&l.push(1):l.push($)});let _=l.length,y=B.size(l);return{name:e,shaderCache:t,getShaderSource:$=>{let S=[],v=M("_A",r[0].dataType,h),w=F("output",s,_),k=i(v,w,f),I=k[2];for(let E=0,C=0;E<h;E++)g||f.indexOf(E)>=0?(n&&C++,I=`for(var j${E}: u32 = 0; j${E} < ${p[E]}; j${E}++) {
+                  ${k[2].includes("last_index")?`let last_index = j${E};`:""}
+                  ${v.indicesSet("input_indices",E,`j${E}`)}
                   ${I}
-                }`):(S.push(`${v.indicesSet("input_indices",k,w.indicesGet("output_indices",C))};`),C++);return`
+                }`):(S.push(`${v.indicesSet("input_indices",E,w.indicesGet("output_indices",C))};`),C++);return`
 
         ${$.registerUniform("output_size","u32").declareVariables(v,w)}
 
@@ -141,25 +141,25 @@ var Ag=Object.defineProperty;var St=(e=>typeof require<"u"?require:typeof Proxy<
 
           ${S.join(`
 `)}
-          ${E[0]}       // init ops for reduce max/min
-          ${E[1]}
+          ${k[0]}       // init ops for reduce max/min
+          ${k[1]}
           ${I}
-          ${E[3]}
-          ${E.length===4?w.setByOffset("global_idx","value"):E.slice(4).join(`
+          ${k[3]}
+          ${k.length===4?w.setByOffset("global_idx","value"):k.slice(4).join(`
 `)}
         }`},getRunData:()=>({outputs:[{dims:l,dataType:s}],dispatchGroup:{x:Math.ceil(y/64)},programUniforms:[{type:12,data:y},...Z(p,l)]})}},$a=(e,t)=>{let r=[];return e[1].dims[0]>0&&e[1].getBigInt64Array().forEach(i=>r.push(Number(i))),he({axes:r,keepDims:t.keepDims,noopWithEmptyAxes:t.noopWithEmptyAxes})},Fe=(e,t,r,i)=>{let a=e.inputs,s=a.length===1?r:$a(a,r);e.compute(Fr(t,{hint:s.cacheKey,inputDependencies:["rank"]},[a[0]],s.noopWithEmptyAxes&&s.axes.length===0?Ho:i,s.axes,a[0].dataType,s.keepDims,s.noopWithEmptyAxes),{inputs:[0]})},jo=(e,t)=>{Ge(e.inputs),Fe(e,"ReduceLogSum",t,(r,i)=>[`var value = ${i.type.storage}(0);`,"",`value += ${r.getByIndices("input_indices")};`,"value = log(value);"])},Ko=(e,t)=>{Ge(e.inputs),Fe(e,"ReduceL1",t,(r,i)=>[`var value = ${i.type.storage}(0);`,"",`value += abs(${r.getByIndices("input_indices")});`,""])},Zo=(e,t)=>{Ge(e.inputs),Fe(e,"ReduceL2",t,(r,i)=>[`var t = ${i.type.value}(0); var value = ${i.type.value}(0);`,"",`t = ${r.getByIndices("input_indices")}; value += (t * t);`,"value = sqrt(value);"])},Qo=(e,t)=>{Ge(e.inputs),Fe(e,"ReduceLogSumExp",t,(r,i)=>[`var value = ${i.type.storage}(0);`,"",`value += exp(${r.getByIndices("input_indices")});`,"value = log(value);"])},Xo=(e,t)=>{Ge(e.inputs),Fe(e,"ReduceMax",t,(r,i,a)=>{let s=[];for(let n=0;n<r.rank;n++)(a.indexOf(n)>=0||a.length===0)&&s.push(r.indicesSet("input_indices",n,0));return[`${s.join(`
 `)}`,`var value = ${r.getByIndices("input_indices")};`,`value = max(value, ${r.getByIndices("input_indices")});`,""]})},Yo=(e,t)=>{Ge(e.inputs),Fe(e,"ReduceMean",t,(r,i,a)=>{let s=1;for(let n=0;n<r.rank;n++)(a.indexOf(n)>=0||a.length===0)&&(s*=e.inputs[0].dims[n]);return["var sum = f32(0);","",`sum += f32(${r.getByIndices("input_indices")});`,`let value = ${i.type.value}(sum / ${s});`]})},Jo=(e,t)=>{Ge(e.inputs),Fe(e,"ReduceMin",t,(r,i,a)=>{let s=[];for(let n=0;n<r.rank;n++)(a.indexOf(n)>=0||a.length===0)&&s.push(`input_indices[${n}] = 0;`);return[`${s.join(`
-`)}`,`var value = ${r.getByIndices("input_indices")};`,`value = min(value, ${r.getByIndices("input_indices")});`,""]})},eu=(e,t)=>{Ge(e.inputs),Fe(e,"ReduceProd",t,(r,i)=>[`var value = ${i.type.storage}(1);`,"",`value *= ${r.getByIndices("input_indices")};`,""])},tu=(e,t)=>{Ge(e.inputs),Fe(e,"ReduceSum",t,(r,i)=>[`var value = ${i.type.storage}(0);`,"",`value += ${r.getByIndices("input_indices")};`,""])},ru=(e,t)=>{Ge(e.inputs),Fe(e,"ReduceSumSquare",t,(r,i)=>[`var t = ${i.type.value}(0); var value = ${i.type.value}(0);`,"",`t = ${r.getByIndices("input_indices")}; value += t * t;`,""])},He=(e,t,r)=>{if(t.length===0)return r;let i=1,a=1;for(let s=0;s<t.length;s++)t.indexOf(s)===-1?i*=e[s]:a*=e[s];return a<32&&i>1024},Ap=(e,t)=>{He(e.inputs[0].dims,t.axes,t.noopWithEmptyAxes)?Yo(e,t):$p(e,t)},Op=(e,t)=>{He(e.inputs[0].dims,t.axes,t.noopWithEmptyAxes)?Ko(e,t):vp(e,t)},Bp=(e,t)=>{He(e.inputs[0].dims,t.axes,t.noopWithEmptyAxes)?Zo(e,t):xp(e,t)},Rp=(e,t)=>{He(e.inputs[0].dims,t.axes,t.noopWithEmptyAxes)?Qo(e,t):Sp(e,t)},Np=(e,t)=>{He(e.inputs[0].dims,t.axes,t.noopWithEmptyAxes)?Xo(e,t):Tp(e,t)},Mp=(e,t)=>{He(e.inputs[0].dims,t.axes,t.noopWithEmptyAxes)?Jo(e,t):Ip(e,t)},Dp=(e,t)=>{He(e.inputs[0].dims,t.axes,t.noopWithEmptyAxes)?eu(e,t):Ep(e,t)},Pp=(e,t)=>{He(e.inputs[0].dims,t.axes,t.noopWithEmptyAxes)?tu(e,t):kp(e,t)},Up=(e,t)=>{He(e.inputs[0].dims,t.axes,t.noopWithEmptyAxes)?ru(e,t):zp(e,t)},Wp=(e,t)=>{He(e.inputs[0].dims,t.axes,t.noopWithEmptyAxes)?jo(e,t):Cp(e,t)}}),Wi,qp,Lp,va,n0=P(()=>{"use strict";Y(),ve(),Fa(),Wi=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.")},qp=(e,t)=>{Wi(e.inputs);let r=(i,a,s)=>{let n=[];for(let u=0;u<i.rank;u++)(s.indexOf(u)>=0||s.length===0)&&n.push(`input_indices[${u}] = 0;`);return[`${n.join(`
+`)}`,`var value = ${r.getByIndices("input_indices")};`,`value = min(value, ${r.getByIndices("input_indices")});`,""]})},eu=(e,t)=>{Ge(e.inputs),Fe(e,"ReduceProd",t,(r,i)=>[`var value = ${i.type.storage}(1);`,"",`value *= ${r.getByIndices("input_indices")};`,""])},tu=(e,t)=>{Ge(e.inputs),Fe(e,"ReduceSum",t,(r,i)=>[`var value = ${i.type.storage}(0);`,"",`value += ${r.getByIndices("input_indices")};`,""])},ru=(e,t)=>{Ge(e.inputs),Fe(e,"ReduceSumSquare",t,(r,i)=>[`var t = ${i.type.value}(0); var value = ${i.type.value}(0);`,"",`t = ${r.getByIndices("input_indices")}; value += t * t;`,""])},He=(e,t,r)=>{if(t.length===0)return r;let i=1,a=1;for(let s=0;s<t.length;s++)t.indexOf(s)===-1?i*=e[s]:a*=e[s];return a<32&&i>1024},Ap=(e,t)=>{He(e.inputs[0].dims,t.axes,t.noopWithEmptyAxes)?Yo(e,t):$p(e,t)},Op=(e,t)=>{He(e.inputs[0].dims,t.axes,t.noopWithEmptyAxes)?Ko(e,t):vp(e,t)},Bp=(e,t)=>{He(e.inputs[0].dims,t.axes,t.noopWithEmptyAxes)?Zo(e,t):xp(e,t)},Rp=(e,t)=>{He(e.inputs[0].dims,t.axes,t.noopWithEmptyAxes)?Qo(e,t):Sp(e,t)},Mp=(e,t)=>{He(e.inputs[0].dims,t.axes,t.noopWithEmptyAxes)?Xo(e,t):Tp(e,t)},Np=(e,t)=>{He(e.inputs[0].dims,t.axes,t.noopWithEmptyAxes)?Jo(e,t):Ip(e,t)},Dp=(e,t)=>{He(e.inputs[0].dims,t.axes,t.noopWithEmptyAxes)?eu(e,t):kp(e,t)},Pp=(e,t)=>{He(e.inputs[0].dims,t.axes,t.noopWithEmptyAxes)?tu(e,t):Ep(e,t)},Up=(e,t)=>{He(e.inputs[0].dims,t.axes,t.noopWithEmptyAxes)?ru(e,t):zp(e,t)},Wp=(e,t)=>{He(e.inputs[0].dims,t.axes,t.noopWithEmptyAxes)?jo(e,t):Cp(e,t)}}),Wi,Lp,qp,va,n0=P(()=>{"use strict";Y(),ve(),Fa(),Wi=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.")},Lp=(e,t)=>{Wi(e.inputs);let r=(i,a,s)=>{let n=[];for(let u=0;u<i.rank;u++)(s.indexOf(u)>=0||s.length===0)&&n.push(`input_indices[${u}] = 0;`);return[`${n.join(`
 `)}`,`var value = ${i.getByIndices("input_indices")};
 var best_index : i32 = 0;`,`if (${i.getByIndices("input_indices")} ${t.selectLastIndex>0?"<=":"<"} value) {
          value = ${i.getByIndices("input_indices")};
          best_index = i32(last_index);
-       }`,"",a.setByOffset("global_idx","best_index")]};e.compute(Fr("ArgMin",{hint:t.cacheKey,inputDependencies:["rank"]},[e.inputs[0]],r,[t.axis],7,t.keepDims),{inputs:[0]})},Lp=(e,t)=>{Wi(e.inputs);let r=(i,a,s)=>{let n=[];for(let u=0;u<i.rank;u++)(s.indexOf(u)>=0||s.length===0)&&n.push(`input_indices[${u}] = 0;`);return[`${n.join(`
+       }`,"",a.setByOffset("global_idx","best_index")]};e.compute(Fr("ArgMin",{hint:t.cacheKey,inputDependencies:["rank"]},[e.inputs[0]],r,[t.axis],7,t.keepDims),{inputs:[0]})},qp=(e,t)=>{Wi(e.inputs);let r=(i,a,s)=>{let n=[];for(let u=0;u<i.rank;u++)(s.indexOf(u)>=0||s.length===0)&&n.push(`input_indices[${u}] = 0;`);return[`${n.join(`
 `)}`,`var value = ${i.getByIndices("input_indices")};
 var best_index : i32 = 0;`,`if (${i.getByIndices("input_indices")} ${t.selectLastIndex>0?">=":">"} value) {
          value = ${i.getByIndices("input_indices")};
          best_index = i32(last_index);
-       }`,"",a.setByOffset("global_idx","best_index")]};e.compute(Fr("argMax",{hint:t.cacheKey,inputDependencies:["rank"]},[e.inputs[0]],r,[t.axis],7,t.keepDims),{inputs:[0]})},va=e=>he(e)}),iu,Br,au,nu,su,hr,ou,Vp,Ha=P(()=>{"use strict";Y(),te(),Va(),re(),iu=(e,t)=>{let r=e[0],i=e[1],a=e[2],s=e[3],n=e[4],u=e[5];if(n&&u)throw new Error("Attention cannot have both past and attention_bias");if(r.dims.length!==3)throw new Error('Input "input" must have 3 dimensions');let l=r.dims[0],p=r.dims[1],h=r.dims[2];if(a.dims.length!==1)throw new Error('Input "bias" is expected to have 1 dimensions');if(i.dims.length!==2)throw new Error('Input "weights" is expected to have 2 dimensions');if(i.dims[0]!==h)throw new Error("Input 1 dimension 0 should have same length as dimension 2 of input 0");if(a.dims[0]!==i.dims[1])throw new Error('Input "bias" dimension 0 should have same length as dimension 1 of input "weights"');let f=a.dims[0]/3,g=f,_=g;if(t.qkvHiddenSizes.length>0){if(t.qkvHiddenSizes.length!==3)throw new Error("qkv_hidden_sizes attribute should have 3 elements");for(let E of t.qkvHiddenSizes)if(E%t.numHeads!==0)throw new Error("qkv_hidden_sizes should be divisible by num_heads");f=t.qkvHiddenSizes[0],g=t.qkvHiddenSizes[1],_=t.qkvHiddenSizes[2]}let y=p;if(f!==g)throw new Error("qkv_hidden_sizes first element should be same as the second");if(a.dims[0]!==f+g+_)throw new Error('Input "bias" dimension 0 should have same length as sum of Q/K/V hidden sizes');let $=0;if(n){if(g!==_)throw new Error('Input "past" expect k_hidden_size == v_hidden_size');if(n.dims.length!==5)throw new Error('Input "past" must have 5 dimensions');if(n.dims[0]!==2)throw new Error('Input "past" first dimension must be 2');if(n.dims[1]!==l)throw new Error('Input "past" second dimension must be batch_size');if(n.dims[2]!==t.numHeads)throw new Error('Input "past" third dimension must be num_heads');if(n.dims[4]!==g/t.numHeads)throw new Error('Input "past" fifth dimension must be k_hidden_size / num_heads');t.pastPresentShareBuffer||($=n.dims[3])}let S=y+$,v=-1,w=0;if(s)throw new Error("Mask not supported");if(n)throw new Error("past is not supported");if(u){if(u.dims.length!==4)throw new Error('Input "attention_bias" must have 4 dimensions');if(u.dims[0]!==l||u.dims[1]!==t.numHeads||u.dims[2]!==p||u.dims[3]!==S)throw new Error('Expect "attention_bias" shape (batch_size, num_heads, sequence_length, total_sequence_length)')}return{batchSize:l,sequenceLength:p,pastSequenceLength:$,kvSequenceLength:y,totalSequenceLength:S,maxSequenceLength:v,inputHiddenSize:h,hiddenSize:f,vHiddenSize:_,headSize:Math.floor(f/t.numHeads),vHeadSize:Math.floor(_/t.numHeads),numHeads:t.numHeads,isUnidirectional:!1,pastPresentShareBuffer:!1,maskFilterValue:t.maskFilterValue,maskType:w,scale:t.scale,broadcastResPosBias:!1,passPastInKv:!1,qkvFormat:1}},Br=(e,t,r)=>t&&e?`
+       }`,"",a.setByOffset("global_idx","best_index")]};e.compute(Fr("argMax",{hint:t.cacheKey,inputDependencies:["rank"]},[e.inputs[0]],r,[t.axis],7,t.keepDims),{inputs:[0]})},va=e=>he(e)}),iu,Br,au,nu,su,hr,ou,Vp,Ha=P(()=>{"use strict";Y(),te(),Va(),re(),iu=(e,t)=>{let r=e[0],i=e[1],a=e[2],s=e[3],n=e[4],u=e[5];if(n&&u)throw new Error("Attention cannot have both past and attention_bias");if(r.dims.length!==3)throw new Error('Input "input" must have 3 dimensions');let l=r.dims[0],p=r.dims[1],h=r.dims[2];if(a.dims.length!==1)throw new Error('Input "bias" is expected to have 1 dimensions');if(i.dims.length!==2)throw new Error('Input "weights" is expected to have 2 dimensions');if(i.dims[0]!==h)throw new Error("Input 1 dimension 0 should have same length as dimension 2 of input 0");if(a.dims[0]!==i.dims[1])throw new Error('Input "bias" dimension 0 should have same length as dimension 1 of input "weights"');let f=a.dims[0]/3,g=f,_=g;if(t.qkvHiddenSizes.length>0){if(t.qkvHiddenSizes.length!==3)throw new Error("qkv_hidden_sizes attribute should have 3 elements");for(let k of t.qkvHiddenSizes)if(k%t.numHeads!==0)throw new Error("qkv_hidden_sizes should be divisible by num_heads");f=t.qkvHiddenSizes[0],g=t.qkvHiddenSizes[1],_=t.qkvHiddenSizes[2]}let y=p;if(f!==g)throw new Error("qkv_hidden_sizes first element should be same as the second");if(a.dims[0]!==f+g+_)throw new Error('Input "bias" dimension 0 should have same length as sum of Q/K/V hidden sizes');let $=0;if(n){if(g!==_)throw new Error('Input "past" expect k_hidden_size == v_hidden_size');if(n.dims.length!==5)throw new Error('Input "past" must have 5 dimensions');if(n.dims[0]!==2)throw new Error('Input "past" first dimension must be 2');if(n.dims[1]!==l)throw new Error('Input "past" second dimension must be batch_size');if(n.dims[2]!==t.numHeads)throw new Error('Input "past" third dimension must be num_heads');if(n.dims[4]!==g/t.numHeads)throw new Error('Input "past" fifth dimension must be k_hidden_size / num_heads');t.pastPresentShareBuffer||($=n.dims[3])}let S=y+$,v=-1,w=0;if(s)throw new Error("Mask not supported");if(n)throw new Error("past is not supported");if(u){if(u.dims.length!==4)throw new Error('Input "attention_bias" must have 4 dimensions');if(u.dims[0]!==l||u.dims[1]!==t.numHeads||u.dims[2]!==p||u.dims[3]!==S)throw new Error('Expect "attention_bias" shape (batch_size, num_heads, sequence_length, total_sequence_length)')}return{batchSize:l,sequenceLength:p,pastSequenceLength:$,kvSequenceLength:y,totalSequenceLength:S,maxSequenceLength:v,inputHiddenSize:h,hiddenSize:f,vHiddenSize:_,headSize:Math.floor(f/t.numHeads),vHeadSize:Math.floor(_/t.numHeads),numHeads:t.numHeads,isUnidirectional:!1,pastPresentShareBuffer:!1,maskFilterValue:t.maskFilterValue,maskType:w,scale:t.scale,broadcastResPosBias:!1,passPastInKv:!1,qkvFormat:1}},Br=(e,t,r)=>t&&e?`
       let total_sequence_length_input = u32(${t.getByOffset("0")});
       let present_sequence_length = max(total_sequence_length_input, uniforms.past_sequence_length);
       let is_subsequent_prompt: bool = sequence_length > 1 && sequence_length != total_sequence_length_input;
@@ -172,16 +172,16 @@ var best_index : i32 = 0;`,`if (${i.getByIndices("input_indices")} ${t.selectLas
        `:`
     ${r?"let past_sequence_length = uniforms.past_sequence_length":""};
     let present_sequence_length = total_sequence_length;
-    `,au=(e,t,r,i,a,s,n,u)=>{let l=$e(n?1:s),p=64,h=s/l;h<p&&(p=32);let f=Math.ceil(s/l/p),g=[{type:12,data:t},{type:12,data:r},{type:12,data:i},{type:12,data:a},{type:12,data:h},{type:12,data:f}],_=Ee(e.dataType,l),y=Ce(1,l),$=["type"];n&&$.push("type"),u&&$.push("type");let S=v=>{let w=F("x",e.dataType,e.dims,l),E=[w],I=n?N("seq_lens",n.dataType,n.dims):void 0;I&&E.push(I);let k=u?N("total_sequence_length_input",u.dataType,u.dims):void 0;k&&E.push(k);let C=Ce(e.dataType),O=[{name:"batch_size",type:"u32"},{name:"num_heads",type:"u32"},{name:"past_sequence_length",type:"u32"},{name:"sequence_length",type:"u32"},{name:"total_sequence_length",type:"u32"},{name:"elements_per_thread",type:"u32"}];return`
+    `,au=(e,t,r,i,a,s,n,u)=>{let l=$e(n?1:s),p=64,h=s/l;h<p&&(p=32);let f=Math.ceil(s/l/p),g=[{type:12,data:t},{type:12,data:r},{type:12,data:i},{type:12,data:a},{type:12,data:h},{type:12,data:f}],_=ke(e.dataType,l),y=Ce(1,l),$=["type"];n&&$.push("type"),u&&$.push("type");let S=v=>{let w=F("x",e.dataType,e.dims,l),k=[w],I=n?M("seq_lens",n.dataType,n.dims):void 0;I&&k.push(I);let E=u?M("total_sequence_length_input",u.dataType,u.dims):void 0;E&&k.push(E);let C=Ce(e.dataType),O=[{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, ${p}>;
   var<workgroup> thread_sum: array<f32, ${p}>;
-  ${v.registerUniforms(O).declareVariables(...E)}
+  ${v.registerUniforms(O).declareVariables(...k)}
   ${v.mainStart([p,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;
-    ${Br(I,k,!1)}
+    ${Br(I,E,!1)}
     let local_offset = local_idx * uniforms.elements_per_thread;
     let offset = (global_idx / ${p}) * uniforms.total_sequence_length + local_offset;
     let seq_causal_length = ${n?"u32(past_sequence_length + workgroup_id.y + 1)":"total_sequence_length"};
@@ -223,12 +223,12 @@ var best_index : i32 = 0;`,`if (${i.getByIndices("input_indices")} ${t.selectLas
         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] = ${w.type.value}(${C}(0));
         }`:""};
-  }`};return{name:"AttentionProbsSoftmax",shaderCache:{hint:`${p};${_};${l}`,inputDependencies:$},getShaderSource:S,getRunData:()=>({outputs:[],dispatchGroup:{x:1,y:a,z:t*r},programUniforms:g})}},nu=(e,t,r,i,a,s,n,u,l)=>{let p=n+s.kvSequenceLength,h=[s.batchSize,s.numHeads,s.sequenceLength,p],f=e>1&&i,g=s.kvNumHeads?s.kvNumHeads:s.numHeads,_=f?[s.batchSize,g,p,s.headSize]:void 0,y=s.nReps?s.nReps:1,$=s.scale===0?1/Math.sqrt(s.headSize):s.scale,S=$e(s.headSize),v=s.headSize/S,w=12,E={x:Math.ceil(p/w),y:Math.ceil(s.sequenceLength/w),z:s.batchSize*s.numHeads},I=[{type:12,data:s.sequenceLength},{type:12,data:v},{type:12,data:p},{type:12,data:s.numHeads},{type:12,data:s.headSize},{type:1,data:$},{type:12,data:n},{type:12,data:s.kvSequenceLength},{type:12,data:y}],k=f&&i&&B.size(i.dims)>0,C=["type","type"];k&&C.push("type"),a&&C.push("type"),u&&C.push("type"),l&&C.push("type");let O=[{dims:h,dataType:t.dataType,gpuDataType:0}];f&&O.push({dims:_,dataType:t.dataType,gpuDataType:0});let x=D=>{let q=N("q",t.dataType,t.dims,S),Q=N("key",r.dataType,r.dims,S),L=[q,Q];if(k){let H=N("past_key",i.dataType,i.dims,S);L.push(H)}a&&L.push(N("attention_bias",a.dataType,a.dims));let K=u?N("seq_lens",u.dataType,u.dims):void 0;K&&L.push(K);let se=l?N("total_sequence_length_input",l.dataType,l.dims):void 0;se&&L.push(se);let A=F("output",t.dataType,h),U=[A];f&&U.push(F("present_key",t.dataType,_,S));let ee=Ce(1,S),X=[{name:"M",type:"u32"},{name:"K",type:"u32"},{name:"N",type:"u32"},{name:"num_heads",type:"u32"},{name:"head_size",type:"u32"},{name:"alpha",type:"f32"},{name:"past_sequence_length",type:"u32"},{name:"kv_sequence_length",type:"u32"},{name:"n_reps",type:"u32"}];return`
+  }`};return{name:"AttentionProbsSoftmax",shaderCache:{hint:`${p};${_};${l}`,inputDependencies:$},getShaderSource:S,getRunData:()=>({outputs:[],dispatchGroup:{x:1,y:a,z:t*r},programUniforms:g})}},nu=(e,t,r,i,a,s,n,u,l)=>{let p=n+s.kvSequenceLength,h=[s.batchSize,s.numHeads,s.sequenceLength,p],f=e>1&&i,g=s.kvNumHeads?s.kvNumHeads:s.numHeads,_=f?[s.batchSize,g,p,s.headSize]:void 0,y=s.nReps?s.nReps:1,$=s.scale===0?1/Math.sqrt(s.headSize):s.scale,S=$e(s.headSize),v=s.headSize/S,w=12,k={x:Math.ceil(p/w),y:Math.ceil(s.sequenceLength/w),z:s.batchSize*s.numHeads},I=[{type:12,data:s.sequenceLength},{type:12,data:v},{type:12,data:p},{type:12,data:s.numHeads},{type:12,data:s.headSize},{type:1,data:$},{type:12,data:n},{type:12,data:s.kvSequenceLength},{type:12,data:y}],E=f&&i&&B.size(i.dims)>0,C=["type","type"];E&&C.push("type"),a&&C.push("type"),u&&C.push("type"),l&&C.push("type");let O=[{dims:h,dataType:t.dataType,gpuDataType:0}];f&&O.push({dims:_,dataType:t.dataType,gpuDataType:0});let x=D=>{let L=M("q",t.dataType,t.dims,S),Q=M("key",r.dataType,r.dims,S),q=[L,Q];if(E){let H=M("past_key",i.dataType,i.dims,S);q.push(H)}a&&q.push(M("attention_bias",a.dataType,a.dims));let K=u?M("seq_lens",u.dataType,u.dims):void 0;K&&q.push(K);let se=l?M("total_sequence_length_input",l.dataType,l.dims):void 0;se&&q.push(se);let A=F("output",t.dataType,h),U=[A];f&&U.push(F("present_key",t.dataType,_,S));let ee=Ce(1,S),X=[{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 = ${w}u;
 
-  var<workgroup> tileQ: array<${q.type.storage}, ${w*w}>;
-  var<workgroup> tileK: array<${q.type.storage}, ${w*w}>;
-  ${D.registerUniforms(X).declareVariables(...L,...U)}
+  var<workgroup> tileQ: array<${L.type.storage}, ${w*w}>;
+  var<workgroup> tileK: array<${L.type.storage}, ${w*w}>;
+  ${D.registerUniforms(X).declareVariables(...q,...U)}
   ${D.mainStart([w,w,1])}
     // x holds the N and y holds the M
     let headIdx = workgroup_id.z % uniforms.num_heads;
@@ -242,7 +242,7 @@ var best_index : i32 = 0;`,`if (${i.getByIndices("input_indices")} ${t.selectLas
     ${Br(K,se,!0)}
     let absKvHeadIdx = batchIdx * kv_num_heads + kvHeadIdx;
     let qOffset = workgroup_id.z * uniforms.M * uniforms.K + m * uniforms.K;
-    ${k&&f?"let pastKeyOffset = absKvHeadIdx * uniforms.past_sequence_length * uniforms.K;":""};
+    ${E&&f?"let pastKeyOffset = absKvHeadIdx * uniforms.past_sequence_length * uniforms.K;":""};
     let kOffset = absKvHeadIdx * uniforms.kv_sequence_length * uniforms.K;
     ${f?"let presentKeyOffset = absKvHeadIdx * uniforms.N * uniforms.K;":""}
     var value = ${ee}(0);
@@ -252,7 +252,7 @@ var best_index : i32 = 0;`,`if (${i.getByIndices("input_indices")} ${t.selectLas
       }
       if (n + local_id.y < uniforms.N && w + local_id.x < uniforms.K) {
         var idx = TILE_SIZE * local_id.y + local_id.x;
-      ${k&&f?`
+      ${E&&f?`
               if (n + local_id.y < past_sequence_length) {
                 tileK[idx] = past_key[pastKeyOffset + (n + local_id.y) * uniforms.K + w + local_id.x];
               } else if (n + local_id.y - past_sequence_length < uniforms.kv_sequence_length) {
@@ -280,11 +280,11 @@ var best_index : i32 = 0;`,`if (${i.getByIndices("input_indices")} ${t.selectLas
       var sum: f32 = ${(()=>{switch(S){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: ${S}`)}})()};
         output[outputIdx] = ${A.type.value} (sum * uniforms.alpha) + ${a?"attention_bias[outputIdx]":"0.0"};
     }
-  }`};return{name:"AttentionProbs",shaderCache:{hint:`${S};${a!==void 0};${i!==void 0};${e}`,inputDependencies:C},getRunData:()=>({outputs:O,dispatchGroup:E,programUniforms:I}),getShaderSource:x}},su=(e,t,r,i,a,s,n=void 0,u=void 0)=>{let l=s+a.kvSequenceLength,p=a.nReps?a.nReps:1,h=a.vHiddenSize*p,f=e>1&&i,g=a.kvNumHeads?a.kvNumHeads:a.numHeads,_=f?[a.batchSize,g,l,a.headSize]:void 0,y=[a.batchSize,a.sequenceLength,h],$=12,S={x:Math.ceil(a.vHeadSize/$),y:Math.ceil(a.sequenceLength/$),z:a.batchSize*a.numHeads},v=[{type:12,data:a.sequenceLength},{type:12,data:l},{type:12,data:a.vHeadSize},{type:12,data:a.numHeads},{type:12,data:a.headSize},{type:12,data:h},{type:12,data:s},{type:12,data:a.kvSequenceLength},{type:12,data:p}],w=f&&i&&B.size(i.dims)>0,E=["type","type"];w&&E.push("type"),n&&E.push("type"),u&&E.push("type");let I=[{dims:y,dataType:t.dataType,gpuDataType:0}];f&&I.push({dims:_,dataType:t.dataType,gpuDataType:0});let k=C=>{let O=N("probs",t.dataType,t.dims),x=N("v",r.dataType,r.dims),D=[O,x];w&&D.push(N("past_value",i.dataType,i.dims));let q=n?N("seq_lens",n.dataType,n.dims):void 0;n&&D.push(q);let Q=u?N("total_sequence_length_input",u.dataType,u.dims):void 0;u&&D.push(Q);let L=[F("output",t.dataType,y)];f&&L.push(F("present_value",t.dataType,_));let K=[{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`
+  }`};return{name:"AttentionProbs",shaderCache:{hint:`${S};${a!==void 0};${i!==void 0};${e}`,inputDependencies:C},getRunData:()=>({outputs:O,dispatchGroup:k,programUniforms:I}),getShaderSource:x}},su=(e,t,r,i,a,s,n=void 0,u=void 0)=>{let l=s+a.kvSequenceLength,p=a.nReps?a.nReps:1,h=a.vHiddenSize*p,f=e>1&&i,g=a.kvNumHeads?a.kvNumHeads:a.numHeads,_=f?[a.batchSize,g,l,a.headSize]:void 0,y=[a.batchSize,a.sequenceLength,h],$=12,S={x:Math.ceil(a.vHeadSize/$),y:Math.ceil(a.sequenceLength/$),z:a.batchSize*a.numHeads},v=[{type:12,data:a.sequenceLength},{type:12,data:l},{type:12,data:a.vHeadSize},{type:12,data:a.numHeads},{type:12,data:a.headSize},{type:12,data:h},{type:12,data:s},{type:12,data:a.kvSequenceLength},{type:12,data:p}],w=f&&i&&B.size(i.dims)>0,k=["type","type"];w&&k.push("type"),n&&k.push("type"),u&&k.push("type");let I=[{dims:y,dataType:t.dataType,gpuDataType:0}];f&&I.push({dims:_,dataType:t.dataType,gpuDataType:0});let E=C=>{let O=M("probs",t.dataType,t.dims),x=M("v",r.dataType,r.dims),D=[O,x];w&&D.push(M("past_value",i.dataType,i.dims));let L=n?M("seq_lens",n.dataType,n.dims):void 0;n&&D.push(L);let Q=u?M("total_sequence_length_input",u.dataType,u.dims):void 0;u&&D.push(Q);let q=[F("output",t.dataType,y)];f&&q.push(F("present_value",t.dataType,_));let K=[{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 = ${$}u;
   var<workgroup> tileQ: array<${O.type.value}, ${$*$}>;
   var<workgroup> tileV: array<${O.type.value}, ${$*$}>;
-  ${C.registerUniforms(K).declareVariables(...D,...L)}
+  ${C.registerUniforms(K).declareVariables(...D,...q)}
   ${C.mainStart([$,$,1])}
    let headIdx = workgroup_id.z % uniforms.num_heads;
    let batchIdx = workgroup_id.z / uniforms.num_heads;
@@ -294,7 +294,7 @@ var best_index : i32 = 0;`,`if (${i.getByIndices("input_indices")} ${t.selectLas
    let n = global_id.x;
    let sequence_length = uniforms.M;
    var total_sequence_length = uniforms.K;
-   ${Br(q,Q,!0)}
+   ${Br(L,Q,!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
    ${w&&f?"let pastValueOffset = absKvHeadIdx * uniforms.N * uniforms.past_sequence_length + n;":""};
@@ -335,13 +335,13 @@ var best_index : i32 = 0;`,`if (${i.getByIndices("input_indices")} ${t.selectLas
        + headIdx * uniforms.N + n;
      output[outputIdx] = value;
    }
-  }`};return{name:"AttentionScore",shaderCache:{hint:`${i!==void 0};${e}`,inputDependencies:E},getRunData:()=>({outputs:I,dispatchGroup:S,programUniforms:v}),getShaderSource:k}},hr=(e,t,r,i,a,s,n,u,l,p,h=void 0,f=void 0)=>{let g=Math.min(e.outputCount,1+(n?1:0)+(u?1:0)),_=g>1?p.pastSequenceLength:0,y=_+p.kvSequenceLength,$=l&&B.size(l.dims)>0?l:void 0,S=[t,r];g>1&&n&&B.size(n.dims)>0&&S.push(n),$&&S.push($),h&&S.push(h),f&&S.push(f);let v=e.compute(nu(g,t,r,n,$,p,_,h,f),{inputs:S,outputs:g>1?[-1,1]:[-1]})[0];e.compute(au(v,p.batchSize,p.numHeads,_,p.sequenceLength,y,h,f),{inputs:h&&f?[v,h,f]:[v],outputs:[]});let w=[v,i];g>1&&u&&B.size(u.dims)>0&&w.push(u),h&&w.push(h),f&&w.push(f),e.compute(su(g,v,i,u,p,_,h,f),{inputs:w,outputs:g>1?[0,2]:[0]})},ou=(e,t)=>{let r=[t.batchSize,t.numHeads,t.sequenceLength,t.headSize],i=t.sequenceLength,a=t.inputHiddenSize,s=t.headSize,n=12,u={x:Math.ceil(t.headSize/n),y:Math.ceil(t.sequenceLength/n),z:t.batchSize*t.numHeads},l=[e.inputs[0],e.inputs[1],e.inputs[2]],p=[{type:12,data:i},{type:12,data:a},{type:12,data:s},{type:12,data:t.numHeads},{type:12,data:t.headSize},{type:12,data:t.hiddenSize},{type:12,data:t.hiddenSize+t.hiddenSize+t.vHiddenSize}],h=f=>{let g=F("output_q",l[0].dataType,r),_=F("output_k",l[0].dataType,r),y=F("output_v",l[0].dataType,r),$=N("input",l[0].dataType,l[0].dims),S=N("weight",l[1].dataType,l[1].dims),v=N("bias",l[2].dataType,l[2].dims),w=$.type.storage,E=[{name:"M",type:"u32"},{name:"K",type:"u32"},{name:"N",type:"u32"},{name:"num_heads",type:"u32"},{name:"head_size",type:"u32"},{name:"hidden_size",type:"u32"},{name:"ldb",type:"u32"}];return`
+  }`};return{name:"AttentionScore",shaderCache:{hint:`${i!==void 0};${e}`,inputDependencies:k},getRunData:()=>({outputs:I,dispatchGroup:S,programUniforms:v}),getShaderSource:E}},hr=(e,t,r,i,a,s,n,u,l,p,h=void 0,f=void 0)=>{let g=Math.min(e.outputCount,1+(n?1:0)+(u?1:0)),_=g>1?p.pastSequenceLength:0,y=_+p.kvSequenceLength,$=l&&B.size(l.dims)>0?l:void 0,S=[t,r];g>1&&n&&B.size(n.dims)>0&&S.push(n),$&&S.push($),h&&S.push(h),f&&S.push(f);let v=e.compute(nu(g,t,r,n,$,p,_,h,f),{inputs:S,outputs:g>1?[-1,1]:[-1]})[0];e.compute(au(v,p.batchSize,p.numHeads,_,p.sequenceLength,y,h,f),{inputs:h&&f?[v,h,f]:[v],outputs:[]});let w=[v,i];g>1&&u&&B.size(u.dims)>0&&w.push(u),h&&w.push(h),f&&w.push(f),e.compute(su(g,v,i,u,p,_,h,f),{inputs:w,outputs:g>1?[0,2]:[0]})},ou=(e,t)=>{let r=[t.batchSize,t.numHeads,t.sequenceLength,t.headSize],i=t.sequenceLength,a=t.inputHiddenSize,s=t.headSize,n=12,u={x:Math.ceil(t.headSize/n),y:Math.ceil(t.sequenceLength/n),z:t.batchSize*t.numHeads},l=[e.inputs[0],e.inputs[1],e.inputs[2]],p=[{type:12,data:i},{type:12,data:a},{type:12,data:s},{type:12,data:t.numHeads},{type:12,data:t.headSize},{type:12,data:t.hiddenSize},{type:12,data:t.hiddenSize+t.hiddenSize+t.vHiddenSize}],h=f=>{let g=F("output_q",l[0].dataType,r),_=F("output_k",l[0].dataType,r),y=F("output_v",l[0].dataType,r),$=M("input",l[0].dataType,l[0].dims),S=M("weight",l[1].dataType,l[1].dims),v=M("bias",l[2].dataType,l[2].dims),w=$.type.storage,k=[{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 = ${n}u;
   var<workgroup> tileInput: array<${w}, ${n*n}>;
   var<workgroup> tileWeightQ: array<${w}, ${n*n}>;
   var<workgroup> tileWeightK: array<${w}, ${n*n}>;
   var<workgroup> tileWeightV: array<${w}, ${n*n}>;
-  ${f.registerUniforms(E).declareVariables($,S,v,g,_,y)}
+  ${f.registerUniforms(k).declareVariables($,S,v,g,_,y)}
   ${f.mainStart([n,n,1])}
     let batchIndex = workgroup_id.z / uniforms.num_heads;
     let headNumber = workgroup_id.z % uniforms.num_heads;
@@ -390,14 +390,14 @@ var best_index : i32 = 0;`,`if (${i.getByIndices("input_indices")} ${t.selectLas
       output_k[outputIdx] = valueK;
       output_v[outputIdx] = valueV;
     }
-  }`};return e.compute({name:"AttentionPrepare",shaderCache:{inputDependencies:["type","type","type"]},getRunData:()=>({outputs:[{dims:r,dataType:e.inputs[0].dataType,gpuDataType:0},{dims:r,dataType:e.inputs[0].dataType,gpuDataType:0},{dims:r,dataType:e.inputs[0].dataType,gpuDataType:0}],dispatchGroup:u,programUniforms:p}),getShaderSource:h},{inputs:l,outputs:[-1,-1,-1]})},Vp=(e,t)=>{let r=iu(e.inputs,t),[i,a,s]=ou(e,r);return hr(e,i,a,s,e.inputs[4],void 0,void 0,void 0,e.inputs[5],r)}}),uu,lu,du,Gp,s0=P(()=>{"use strict";We(),Y(),te(),ve(),re(),uu=(e,t)=>{if(!e||e.length!==5)throw new Error("BatchNormalization requires 5 inputs");let r=(i,a,s)=>{let n=a.length;if(n!==i.length)throw new Error(`${s}: num dimensions != ${n}`);a.forEach((u,l)=>{if(u!==i[l])throw new Error(`${s}: dim[${l}] do not match`)})};if(e[0].dims.length>1){let i=t.format==="NHWC"?t.spatial?e[0].dims.slice(-1):e[0].dims.slice(-1).concat(e[0].dims.slice(1,e[0].dims.length-1)):e[0].dims.slice(1,t.spatial?2:void 0);r(e[1].dims,i,"Invalid input scale"),r(e[2].dims,i,"Invalid input B"),r(e[3].dims,i,"Invalid input mean"),r(e[4].dims,i,"Invalid input var")}else r(e[1].dims,[1],"Invalid input scale"),r(e[2].dims,[1],"Invalid input B"),r(e[3].dims,[1],"Invalid input mean"),r(e[4].dims,[1],"Invalid input var")},lu=(e,t)=>{let{epsilon:r,spatial:i,format:a}=t,s=e[0].dims,n=i?$e(s[s.length-1]):1,u=a==="NHWC"&&s.length>1?n:1,l=B.size(s)/n,p=i,h=p?s.length:s,f=N("x",e[0].dataType,e[0].dims,n),g=N("scale",e[1].dataType,e[1].dims,u),_=N("bias",e[2].dataType,e[2].dims,u),y=N("inputMean",e[3].dataType,e[3].dims,u),$=N("inputVar",e[4].dataType,e[4].dims,u),S=F("y",e[0].dataType,h,n),v=()=>{let E="";if(i)E=`let cOffset = ${s.length===1?"0u":a==="NHWC"?`outputIndices[${s.length-1}] / ${n}`:"outputIndices[1]"};`;else if(a==="NCHW")E=`
+  }`};return e.compute({name:"AttentionPrepare",shaderCache:{inputDependencies:["type","type","type"]},getRunData:()=>({outputs:[{dims:r,dataType:e.inputs[0].dataType,gpuDataType:0},{dims:r,dataType:e.inputs[0].dataType,gpuDataType:0},{dims:r,dataType:e.inputs[0].dataType,gpuDataType:0}],dispatchGroup:u,programUniforms:p}),getShaderSource:h},{inputs:l,outputs:[-1,-1,-1]})},Vp=(e,t)=>{let r=iu(e.inputs,t),[i,a,s]=ou(e,r);return hr(e,i,a,s,e.inputs[4],void 0,void 0,void 0,e.inputs[5],r)}}),uu,lu,du,Gp,s0=P(()=>{"use strict";We(),Y(),te(),ve(),re(),uu=(e,t)=>{if(!e||e.length!==5)throw new Error("BatchNormalization requires 5 inputs");let r=(i,a,s)=>{let n=a.length;if(n!==i.length)throw new Error(`${s}: num dimensions != ${n}`);a.forEach((u,l)=>{if(u!==i[l])throw new Error(`${s}: dim[${l}] do not match`)})};if(e[0].dims.length>1){let i=t.format==="NHWC"?t.spatial?e[0].dims.slice(-1):e[0].dims.slice(-1).concat(e[0].dims.slice(1,e[0].dims.length-1)):e[0].dims.slice(1,t.spatial?2:void 0);r(e[1].dims,i,"Invalid input scale"),r(e[2].dims,i,"Invalid input B"),r(e[3].dims,i,"Invalid input mean"),r(e[4].dims,i,"Invalid input var")}else r(e[1].dims,[1],"Invalid input scale"),r(e[2].dims,[1],"Invalid input B"),r(e[3].dims,[1],"Invalid input mean"),r(e[4].dims,[1],"Invalid input var")},lu=(e,t)=>{let{epsilon:r,spatial:i,format:a}=t,s=e[0].dims,n=i?$e(s[s.length-1]):1,u=a==="NHWC"&&s.length>1?n:1,l=B.size(s)/n,p=i,h=p?s.length:s,f=M("x",e[0].dataType,e[0].dims,n),g=M("scale",e[1].dataType,e[1].dims,u),_=M("bias",e[2].dataType,e[2].dims,u),y=M("inputMean",e[3].dataType,e[3].dims,u),$=M("inputVar",e[4].dataType,e[4].dims,u),S=F("y",e[0].dataType,h,n),v=()=>{let k="";if(i)k=`let cOffset = ${s.length===1?"0u":a==="NHWC"?`outputIndices[${s.length-1}] / ${n}`:"outputIndices[1]"};`;else if(a==="NCHW")k=`
             ${S.indicesSet("outputIndices","0","0")}
-            let cOffset = ${S.indicesToOffset("outputIndices")};`;else{E=`var cIndices = ${g.type.indices}(0);
-                       cIndices[0] = outputIndices[${s.length-1}];`;for(let I=1;I<g.rank;I++)E+=`cIndices[${I}] = outputIndices[${I}];`;E+=`let cOffset = ${g.indicesToOffset("cIndices")};`}return E},w=E=>`
+            let cOffset = ${S.indicesToOffset("outputIndices")};`;else{k=`var cIndices = ${g.type.indices}(0);
+                       cIndices[0] = outputIndices[${s.length-1}];`;for(let I=1;I<g.rank;I++)k+=`cIndices[${I}] = outputIndices[${I}];`;k+=`let cOffset = ${g.indicesToOffset("cIndices")};`}return k},w=k=>`
   const epsilon = ${r};
-  ${E.registerUniform("outputSize","u32").declareVariables(f,g,_,y,$,S)}
-  ${E.mainStart()}
-  ${E.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.outputSize")}
+  ${k.registerUniform("outputSize","u32").declareVariables(f,g,_,y,$,S)}
+  ${k.mainStart()}
+  ${k.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.outputSize")}
     var outputIndices = ${S.offsetToIndices(`global_idx * ${n}`)};
     ${v()}
     let scale = ${g.getByOffset("cOffset")};
@@ -407,7 +407,7 @@ var best_index : i32 = 0;`,`if (${i.getByIndices("input_indices")} ${t.selectLas
     let x = ${f.getByOffset("global_idx")};
     let value = (x - inputMean) * inverseSqrt(inputVar + epsilon) * scale + bias;
     ${S.setByOffset("global_idx","value")}
-  }`;return{name:"BatchNormalization",shaderCache:{hint:`${t.epsilon}_${t.format}_${i}_${n}`,inputDependencies:p?["rank","type","type","type","type"]:void 0},getShaderSource:w,getRunData:()=>({outputs:[{dims:e[0].dims,dataType:e[0].dataType}],dispatchGroup:{x:Math.ceil(l/64)},programUniforms:p?[{type:12,data:l},...Z(s)]:[{type:12,data:l}]})}},du=e=>he(e),Gp=(e,t)=>{let{inputs:r,outputCount:i}=e,a=du({...t,outputCount:i});if(ge.webgpu.validateInputContent&&uu(r,a),t.trainingMode)throw new Error("BatchNormalization trainingMode is not supported yet.");e.compute(lu(r,a))}}),pu,cu,Fp,o0=P(()=>{"use strict";te(),re(),pu=e=>{if(e[0].dims.length!==3)throw new Error("input should have 3 dimensions");if(![320,640,1280].includes(e[0].dims[2]))throw new Error("number of channels should be 320, 640 or 1280");if(e[1].dims.length!==1)throw new Error("bias is expected to have 1 dimensions");if(e[0].dims[2]!==e[1].dims[0])throw new Error("last dimension of input and bias are not the same")},cu=e=>{let t=e[0].dims,r=e[0].dims[2],i=B.size(t)/4,a=e[0].dataType,s=N("input",a,t,4),n=N("bias",a,[r],4),u=N("residual",a,t,4),l=F("output",a,t,4);return{name:"BiasAdd",getRunData:()=>({outputs:[{dims:t,dataType:e[0].dataType}],dispatchGroup:{x:Math.ceil(i/64)}}),getShaderSource:p=>`
+  }`;return{name:"BatchNormalization",shaderCache:{hint:`${t.epsilon}_${t.format}_${i}_${n}`,inputDependencies:p?["rank","type","type","type","type"]:void 0},getShaderSource:w,getRunData:()=>({outputs:[{dims:e[0].dims,dataType:e[0].dataType}],dispatchGroup:{x:Math.ceil(l/64)},programUniforms:p?[{type:12,data:l},...Z(s)]:[{type:12,data:l}]})}},du=e=>he(e),Gp=(e,t)=>{let{inputs:r,outputCount:i}=e,a=du({...t,outputCount:i});if(ge.webgpu.validateInputContent&&uu(r,a),t.trainingMode)throw new Error("BatchNormalization trainingMode is not supported yet.");e.compute(lu(r,a))}}),pu,cu,Fp,o0=P(()=>{"use strict";te(),re(),pu=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")},cu=e=>{let t=e[0].dims,r=e[0].dims[2],i=B.size(t)/4,a=e[0].dataType,s=M("input",a,t,4),n=M("bias",a,[r],4),u=M("residual",a,t,4),l=F("output",a,t,4);return{name:"BiasAdd",getRunData:()=>({outputs:[{dims:t,dataType:e[0].dataType}],dispatchGroup:{x:Math.ceil(i/64)}}),getShaderSource:p=>`
   const channels = ${r}u / 4;
   ${p.declareVariables(s,n,u,l)}
 
@@ -416,7 +416,7 @@ var best_index : i32 = 0;`,`if (${i.getByIndices("input_indices")} ${t.selectLas
     let value = ${s.getByOffset("global_idx")}
       + ${n.getByOffset("global_idx % channels")} + ${u.getByOffset("global_idx")};
     ${l.setByOffset("global_idx","value")}
-  }`}},Fp=e=>{pu(e.inputs),e.compute(cu(e.inputs))}}),hu,ce,Hp,jp,Kp,Zp,Qp,Xp,Yp,Jp,ec,fu,tc,rc,ic,ac,ur,nc,qr,sc,oc,uc,lc,dc,pc,cc,hc,fc,mc,gc,yc,_c,wc,bc,$c,qi,vc,xa,Sa,xc,Sc,Tc,mu,gu,Ic,ja=P(()=>{"use strict";Y(),te(),ve(),re(),hu=(e,t,r,i,a,s,n)=>{let u=Math.ceil(t/4),l="";typeof a=="string"?l=`${a}(a)`:l=a("a");let p=N("inputData",r,[u],4),h=F("outputData",i,[u],4),f=[{name:"vec_size",type:"u32"}];return n&&f.push(...n),`
+  }`}},Fp=e=>{pu(e.inputs),e.compute(cu(e.inputs))}}),hu,ce,Hp,jp,Kp,Zp,Qp,Xp,Yp,Jp,ec,fu,tc,rc,ic,ac,ur,nc,Lr,sc,oc,uc,lc,dc,pc,cc,hc,fc,mc,gc,yc,_c,wc,bc,$c,Li,vc,xa,Sa,xc,Sc,Tc,mu,gu,Ic,ja=P(()=>{"use strict";Y(),te(),ve(),re(),hu=(e,t,r,i,a,s,n)=>{let u=Math.ceil(t/4),l="";typeof a=="string"?l=`${a}(a)`:l=a("a");let p=M("inputData",r,[u],4),h=F("outputData",i,[u],4),f=[{name:"vec_size",type:"u32"}];return n&&f.push(...n),`
       ${e.registerUniforms(f).declareVariables(p,h)}
 
   ${s??""}
@@ -435,7 +435,7 @@ var best_index : i32 = 0;`,`if (${i.getByIndices("input_indices")} ${t.selectLas
 
   fn elu_vf32(v: vec4<${r}>) -> vec4<${r}> {
   return vec4(elu_f32(v.x), elu_f32(v.y), elu_f32(v.z), elu_f32(v.w));
-  }`,t.cacheKey))},qr=(e="f32")=>`
+  }`,t.cacheKey))},Lr=(e="f32")=>`
 const r0: ${e} = 0.3275911;
 const r1: ${e} = 0.254829592;
 const r2: ${e} = -0.284496736;
@@ -447,13 +447,13 @@ fn erf_vf32(v: vec4<${e}>) -> vec4<${e}> {
   let absv = abs(v);
   let x = 1.0 / (1.0 + r0 * absv);
   return sign(v) * (1.0 - ((((r5 * x + r4) * x + r3) * x + r2) * x + r1) * x * exp(-absv * absv));
-}`,sc=e=>{let t=Ce(e.inputs[0].dataType);e.compute(ce(e.inputs[0],"Erf",r=>`erf_vf32(${r})`,qr(t)))},oc=e=>{e.compute(ce(e.inputs[0],"Exp","exp"))},uc=e=>{e.compute(ce(e.inputs[0],"Floor","floor"))},lc=e=>{let t=Ce(e.inputs[0].dataType);e.compute(ce(e.inputs[0],"Gelu",r=>`0.5 * ${r} * (1.0 + erf_vf32(${r} * 0.7071067811865475))`,qr(t)))},dc=(e,t)=>{let r=Ce(e.inputs[0].dataType);e.compute(ce(e.inputs[0],"LeakyRelu",i=>`select(leaky_relu_alpha_ * ${i}, ${i}, ${i} >= vec4<${r}>(0.0))`,`const leaky_relu_alpha_ = ${r}(${t.alpha});`,t.cacheKey))},pc=e=>{e.compute(ce(e.inputs[0],"Not",t=>`!${t}`))},cc=e=>{e.compute(ce(e.inputs[0],"Neg",t=>`-${t}`))},hc=e=>{e.compute(ce(e.inputs[0],"Reciprocal",t=>`1.0/${t}`))},fc=e=>{let t=Ce(e.inputs[0].dataType);e.compute(ce(e.inputs[0],"Relu",r=>`select(vec4<${t}>(0.0), ${r}, ${r} > vec4<${t}>(0.0))`))},mc=e=>{e.compute(ce(e.inputs[0],"Sigmoid",t=>`(1.0 / (1.0 + exp(-${t})))`))},gc=e=>he(e),yc=(e,t)=>{let r=Ce(e.inputs[0].dataType);e.compute(ce(e.inputs[0],"HardSigmoid",i=>`max(vec4<${r}>(0.0), min(vec4<${r}>(1.0), ${t.alpha} * ${i} + vec4<${r}>(${t.beta})))`,void 0,t.cacheKey))},_c=e=>{e.compute(ce(e.inputs[0],"Sin","sin"))},wc=e=>{e.compute(ce(e.inputs[0],"Sinh","sinh"))},bc=e=>{e.compute(ce(e.inputs[0],"Sqrt","sqrt"))},$c=e=>{e.compute(ce(e.inputs[0],"Tan","tan"))},qi=e=>`sign(${e}) * (1 - exp(-2 * abs(${e}))) / (1 + exp(-2 * abs(${e})))`,vc=e=>{e.compute(ce(e.inputs[0],"Tanh",qi))},xa=(e="f32")=>`
+}`,sc=e=>{let t=Ce(e.inputs[0].dataType);e.compute(ce(e.inputs[0],"Erf",r=>`erf_vf32(${r})`,Lr(t)))},oc=e=>{e.compute(ce(e.inputs[0],"Exp","exp"))},uc=e=>{e.compute(ce(e.inputs[0],"Floor","floor"))},lc=e=>{let t=Ce(e.inputs[0].dataType);e.compute(ce(e.inputs[0],"Gelu",r=>`0.5 * ${r} * (1.0 + erf_vf32(${r} * 0.7071067811865475))`,Lr(t)))},dc=(e,t)=>{let r=Ce(e.inputs[0].dataType);e.compute(ce(e.inputs[0],"LeakyRelu",i=>`select(leaky_relu_alpha_ * ${i}, ${i}, ${i} >= vec4<${r}>(0.0))`,`const leaky_relu_alpha_ = ${r}(${t.alpha});`,t.cacheKey))},pc=e=>{e.compute(ce(e.inputs[0],"Not",t=>`!${t}`))},cc=e=>{e.compute(ce(e.inputs[0],"Neg",t=>`-${t}`))},hc=e=>{e.compute(ce(e.inputs[0],"Reciprocal",t=>`1.0/${t}`))},fc=e=>{let t=Ce(e.inputs[0].dataType);e.compute(ce(e.inputs[0],"Relu",r=>`select(vec4<${t}>(0.0), ${r}, ${r} > vec4<${t}>(0.0))`))},mc=e=>{e.compute(ce(e.inputs[0],"Sigmoid",t=>`(1.0 / (1.0 + exp(-${t})))`))},gc=e=>he(e),yc=(e,t)=>{let r=Ce(e.inputs[0].dataType);e.compute(ce(e.inputs[0],"HardSigmoid",i=>`max(vec4<${r}>(0.0), min(vec4<${r}>(1.0), ${t.alpha} * ${i} + vec4<${r}>(${t.beta})))`,void 0,t.cacheKey))},_c=e=>{e.compute(ce(e.inputs[0],"Sin","sin"))},wc=e=>{e.compute(ce(e.inputs[0],"Sinh","sinh"))},bc=e=>{e.compute(ce(e.inputs[0],"Sqrt","sqrt"))},$c=e=>{e.compute(ce(e.inputs[0],"Tan","tan"))},Li=e=>`sign(${e}) * (1 - exp(-2 * abs(${e}))) / (1 + exp(-2 * abs(${e})))`,vc=e=>{e.compute(ce(e.inputs[0],"Tanh",Li))},xa=(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 ${qi("v")};
+  return ${Li("v")};
 }
 `,Sa=e=>`(fast_gelu_a + fast_gelu_a * tanh_v(${e} * (fast_gelu_c * ${e} * ${e} + fast_gelu_b))) * ${e}`,xc=e=>{let t=Ce(e.inputs[0].dataType);e.compute(ce(e.inputs[0],"FastGelu",Sa,xa(t),void 0,e.inputs[0].dataType))},Sc=(e,t)=>{let r=Ce(e.inputs[0].dataType);return e.compute(ce(e.inputs[0],"ThresholdedRelu",i=>`select(vec4<${r}>(0.0), ${i}, ${i} > thresholded_relu_alpha_)`,`const thresholded_relu_alpha_ = vec4<${r}>(${t.alpha});`,t.cacheKey)),0},Tc=e=>{e.compute(ce(e.inputs[0],"Log","log"))},mu=(e,t)=>`
 const alpha = vec4<${e}>(${t});
@@ -472,13 +472,13 @@ fn quick_gelu_impl(x: vec4<${e}>) -> vec4<${e}> {
   }
   return x * x1;
 }
-`,gu=e=>`quick_gelu_impl(${e})`,Ic=(e,t)=>{let r=Ce(e.inputs[0].dataType);e.compute(ce(e.inputs[0],"QuickGelu",gu,mu(r,t.alpha),t.cacheKey,e.inputs[0].dataType))}}),yu,_u,Ec,u0=P(()=>{"use strict";te(),re(),ja(),yu=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")},_u=e=>{let t=e[0].dims.slice();t[2]=t[2]/2;let r=N("input",e[0].dataType,e[0].dims,4),i=N("bias",e[0].dataType,[e[0].dims[2]],4),a=F("output",e[0].dataType,t,4),s=B.size(t)/4,n=Ee(e[0].dataType);return{name:"BiasSplitGelu",getRunData:()=>({outputs:[{dims:t,dataType:e[0].dataType}],dispatchGroup:{x:Math.ceil(s/64)}}),getShaderSource:u=>`
+`,gu=e=>`quick_gelu_impl(${e})`,Ic=(e,t)=>{let r=Ce(e.inputs[0].dataType);e.compute(ce(e.inputs[0],"QuickGelu",gu,mu(r,t.alpha),t.cacheKey,e.inputs[0].dataType))}}),yu,_u,kc,u0=P(()=>{"use strict";te(),re(),ja(),yu=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")},_u=e=>{let t=e[0].dims.slice();t[2]=t[2]/2;let r=M("input",e[0].dataType,e[0].dims,4),i=M("bias",e[0].dataType,[e[0].dims[2]],4),a=F("output",e[0].dataType,t,4),s=B.size(t)/4,n=ke(e[0].dataType);return{name:"BiasSplitGelu",getRunData:()=>({outputs:[{dims:t,dataType:e[0].dataType}],dispatchGroup:{x:Math.ceil(s/64)}}),getShaderSource:u=>`
   const M_SQRT2 = sqrt(2.0);
   const halfChannels = ${e[0].dims[2]/4/2}u;
 
   ${u.declareVariables(r,i,a)}
 
-  ${qr(n)}
+  ${Lr(n)}
 
   ${u.mainStart()}
     ${u.guardAgainstOutOfBoundsWorkgroupSizes(s)}
@@ -490,12 +490,12 @@ fn quick_gelu_impl(x: vec4<${e}>) -> vec4<${e}> {
     let geluRight = valueRight * 0.5 * (erf_vf32(valueRight / M_SQRT2) + 1);
 
     ${a.setByOffset("global_idx","valueLeft * geluRight")}
-  }`}},Ec=e=>{yu(e.inputs),e.compute(_u(e.inputs))}}),wu,bu,je,kc,zc,Cc,Ac,Oc,Bc,Rc,Nc,Mc,Dc,l0=P(()=>{"use strict";Y(),te(),re(),wu=(e,t,r,i,a,s,n,u,l,p,h,f)=>{let g,_;typeof u=="string"?g=_=(w,E)=>`${u}((${w}),(${E}))`:typeof u=="function"?g=_=u:(g=u.scalar,_=u.vector);let y=F("outputData",h,i.length,4),$=N("aData",l,t.length,4),S=N("bData",p,r.length,4),v;if(a)if(s){let w=B.size(t)===1,E=B.size(r)===1,I=t.length>0&&t[t.length-1]%4===0,k=r.length>0&&r[r.length-1]%4===0;w||E?v=y.setByOffset("global_idx",_(w?`${$.type.value}(${$.getByOffset("0")}.x)`:$.getByOffset("global_idx"),E?`${S.type.value}(${S.getByOffset("0")}.x)`:S.getByOffset("global_idx"))):v=`
+  }`}},kc=e=>{yu(e.inputs),e.compute(_u(e.inputs))}}),wu,bu,je,Ec,zc,Cc,Ac,Oc,Bc,Rc,Mc,Nc,Dc,l0=P(()=>{"use strict";Y(),te(),re(),wu=(e,t,r,i,a,s,n,u,l,p,h,f)=>{let g,_;typeof u=="string"?g=_=(w,k)=>`${u}((${w}),(${k}))`:typeof u=="function"?g=_=u:(g=u.scalar,_=u.vector);let y=F("outputData",h,i.length,4),$=M("aData",l,t.length,4),S=M("bData",p,r.length,4),v;if(a)if(s){let w=B.size(t)===1,k=B.size(r)===1,I=t.length>0&&t[t.length-1]%4===0,E=r.length>0&&r[r.length-1]%4===0;w||k?v=y.setByOffset("global_idx",_(w?`${$.type.value}(${$.getByOffset("0")}.x)`:$.getByOffset("global_idx"),k?`${S.type.value}(${S.getByOffset("0")}.x)`:S.getByOffset("global_idx"))):v=`
             let outputIndices = ${y.offsetToIndices("global_idx * 4u")};
             let offsetA = ${$.broadcastedIndicesToOffset("outputIndices",y)};
             let offsetB = ${S.broadcastedIndicesToOffset("outputIndices",y)};
-            ${y.setByOffset("global_idx",_(n||I?$.getByOffset("offsetA / 4u"):`${$.type.value}(${$.getByOffset("offsetA / 4u")}[offsetA % 4u])`,n||k?S.getByOffset("offsetB / 4u"):`${S.type.value}(${S.getByOffset("offsetB / 4u")}[offsetB % 4u])`))}
-          `}else v=y.setByOffset("global_idx",_($.getByOffset("global_idx"),S.getByOffset("global_idx")));else{if(!s)throw new Error("no necessary to use scalar implementation for element-wise binary op implementation.");let w=(E,I,k="")=>{let C=`aData[indexA${I}][componentA${I}]`,O=`bData[indexB${I}][componentB${I}]`;return`
+            ${y.setByOffset("global_idx",_(n||I?$.getByOffset("offsetA / 4u"):`${$.type.value}(${$.getByOffset("offsetA / 4u")}[offsetA % 4u])`,n||E?S.getByOffset("offsetB / 4u"):`${S.type.value}(${S.getByOffset("offsetB / 4u")}[offsetB % 4u])`))}
+          `}else v=y.setByOffset("global_idx",_($.getByOffset("global_idx"),S.getByOffset("global_idx")));else{if(!s)throw new Error("no necessary to use scalar implementation for element-wise binary op implementation.");let w=(k,I,E="")=>{let C=`aData[indexA${I}][componentA${I}]`,O=`bData[indexB${I}][componentB${I}]`;return`
             let outputIndices${I} = ${y.offsetToIndices(`global_idx * 4u + ${I}u`)};
             let offsetA${I} = ${$.broadcastedIndicesToOffset(`outputIndices${I}`,y)};
             let offsetB${I} = ${S.broadcastedIndicesToOffset(`outputIndices${I}`,y)};
@@ -503,7 +503,7 @@ fn quick_gelu_impl(x: vec4<${e}>) -> vec4<${e}> {
             let indexB${I} = offsetB${I} / 4u;
             let componentA${I} = offsetA${I} % 4u;
             let componentB${I} = offsetB${I} % 4u;
-            ${E}[${I}] = ${k}(${g(C,O)});
+            ${k}[${I}] = ${E}(${g(C,O)});
           `};h===9?v=`
             var data = vec4<u32>(0);
             ${w("data",0,"u32")}
@@ -523,7 +523,7 @@ fn quick_gelu_impl(x: vec4<${e}>) -> vec4<${e}> {
         ${e.mainStart()}
         ${e.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.vec_size")}
         ${v}
-      }`},bu=(e,t,r,i,a,s,n=r.dataType)=>{let u=r.dims.map(Number),l=i.dims.map(Number),p=!B.areEqual(u,l),h=u,f=B.size(u),g=!1,_=!1,y=[p];if(p){let $=Lt.calcShape(u,l,!1);if(!$)throw new Error("Can't perform binary op on the given tensors");h=$.slice(),f=B.size(h);let S=B.size(u)===1,v=B.size(l)===1,w=u.length>0&&u[u.length-1]%4===0,E=l.length>0&&l[l.length-1]%4===0;y.push(S),y.push(v),y.push(w),y.push(E);let I=1;for(let k=1;k<h.length;k++){let C=u[u.length-k],O=l[l.length-k];if(C===O)I*=C;else break}I%4===0?(_=!0,g=!0):(S||v||w||E)&&(g=!0)}else g=!0;return y.push(g),{name:e,shaderCache:{hint:t+y.map($=>$.toString()).join("_"),inputDependencies:["rank","rank"]},getShaderSource:$=>wu($,u,l,h,g,p,_,a,r.dataType,i.dataType,n,s),getRunData:()=>({outputs:[{dims:h,dataType:n}],dispatchGroup:{x:Math.ceil(f/64/4)},programUniforms:[{type:12,data:Math.ceil(B.size(h)/4)},...Z(u,l,h)]})}},je=(e,t,r,i,a,s)=>{e.compute(bu(t,a??"",e.inputs[0],e.inputs[1],r,i,s))},kc=e=>{je(e,"Add",(t,r)=>`${t}+${r}`)},zc=e=>{je(e,"Div",(t,r)=>`${t}/${r}`)},Cc=e=>{je(e,"Equal",{scalar:(t,r)=>`u32(${t}==${r})`,vector:(t,r)=>`vec4<u32>(${t}==${r})`},void 0,void 0,9)},Ac=e=>{je(e,"Mul",(t,r)=>`${t}*${r}`)},Oc=e=>{let t=N("input",e.inputs[0].dataType,e.inputs[0].dims).type.value;je(e,"Pow",{scalar:(r,i)=>`pow_custom(${r},${i})`,vector:(r,i)=>`pow_vector_custom(${r},${i})`},`
+      }`},bu=(e,t,r,i,a,s,n=r.dataType)=>{let u=r.dims.map(Number),l=i.dims.map(Number),p=!B.areEqual(u,l),h=u,f=B.size(u),g=!1,_=!1,y=[p];if(p){let $=qt.calcShape(u,l,!1);if(!$)throw new Error("Can't perform binary op on the given tensors");h=$.slice(),f=B.size(h);let S=B.size(u)===1,v=B.size(l)===1,w=u.length>0&&u[u.length-1]%4===0,k=l.length>0&&l[l.length-1]%4===0;y.push(S),y.push(v),y.push(w),y.push(k);let I=1;for(let E=1;E<h.length;E++){let C=u[u.length-E],O=l[l.length-E];if(C===O)I*=C;else break}I%4===0?(_=!0,g=!0):(S||v||w||k)&&(g=!0)}else g=!0;return y.push(g),{name:e,shaderCache:{hint:t+y.map($=>$.toString()).join("_"),inputDependencies:["rank","rank"]},getShaderSource:$=>wu($,u,l,h,g,p,_,a,r.dataType,i.dataType,n,s),getRunData:()=>({outputs:[{dims:h,dataType:n}],dispatchGroup:{x:Math.ceil(f/64/4)},programUniforms:[{type:12,data:Math.ceil(B.size(h)/4)},...Z(u,l,h)]})}},je=(e,t,r,i,a,s)=>{e.compute(bu(t,a??"",e.inputs[0],e.inputs[1],r,i,s))},Ec=e=>{je(e,"Add",(t,r)=>`${t}+${r}`)},zc=e=>{je(e,"Div",(t,r)=>`${t}/${r}`)},Cc=e=>{je(e,"Equal",{scalar:(t,r)=>`u32(${t}==${r})`,vector:(t,r)=>`vec4<u32>(${t}==${r})`},void 0,void 0,9)},Ac=e=>{je(e,"Mul",(t,r)=>`${t}*${r}`)},Oc=e=>{let t=M("input",e.inputs[0].dataType,e.inputs[0].dims).type.value;je(e,"Pow",{scalar:(r,i)=>`pow_custom(${r},${i})`,vector:(r,i)=>`pow_vector_custom(${r},${i})`},`
     fn pow_custom(a : ${t}, b : ${t}) -> ${t} {
       if (b == ${t}(0.0)) {
         return ${t}(1.0);
@@ -536,7 +536,7 @@ fn quick_gelu_impl(x: vec4<${e}>) -> vec4<${e}> {
       // TODO: implement vectorized pow
       return vec4<${t}>(pow_custom(a.x, b.x), pow_custom(a.y, b.y), pow_custom(a.z, b.z), pow_custom(a.w, b.w));
     }
-      `)},Bc=e=>{je(e,"Sub",(t,r)=>`${t}-${r}`)},Rc=e=>{je(e,"Greater",{scalar:(t,r)=>`u32(${t}>${r})`,vector:(t,r)=>`vec4<u32>(${t}>${r})`},void 0,void 0,9)},Nc=e=>{je(e,"Less",{scalar:(t,r)=>`u32(${t}<${r})`,vector:(t,r)=>`vec4<u32>(${t}<${r})`},void 0,void 0,9)},Mc=e=>{je(e,"GreaterOrEqual",{scalar:(t,r)=>`u32(${t}>=${r})`,vector:(t,r)=>`vec4<u32>(${t}>=${r})`},void 0,void 0,9)},Dc=e=>{je(e,"LessOrEqual",{scalar:(t,r)=>`u32(${t}<=${r})`,vector:(t,r)=>`vec4<u32>(${t}<=${r})`},void 0,void 0,9)}}),$u,vu,xu,Su,Pc,Uc,d0=P(()=>{"use strict";Y(),te(),ve(),re(),$u=(e,t)=>{if(!e||e.length<1)throw new Error("too few inputs");let r=0,i=e[r],a=i.dataType,s=i.dims.length;e.forEach((n,u)=>{if(u!==r){if(n.dataType!==a)throw new Error("input tensors should be one type");if(n.dims.length!==s)throw new Error("input tensors should have the same shape");n.dims.forEach((l,p)=>{if(p!==t&&l!==i.dims[p])throw new Error("non concat dimensions must match")})}})},vu=(e,t)=>`
+      `)},Bc=e=>{je(e,"Sub",(t,r)=>`${t}-${r}`)},Rc=e=>{je(e,"Greater",{scalar:(t,r)=>`u32(${t}>${r})`,vector:(t,r)=>`vec4<u32>(${t}>${r})`},void 0,void 0,9)},Mc=e=>{je(e,"Less",{scalar:(t,r)=>`u32(${t}<${r})`,vector:(t,r)=>`vec4<u32>(${t}<${r})`},void 0,void 0,9)},Nc=e=>{je(e,"GreaterOrEqual",{scalar:(t,r)=>`u32(${t}>=${r})`,vector:(t,r)=>`vec4<u32>(${t}>=${r})`},void 0,void 0,9)},Dc=e=>{je(e,"LessOrEqual",{scalar:(t,r)=>`u32(${t}<=${r})`,vector:(t,r)=>`vec4<u32>(${t}<=${r})`},void 0,void 0,9)}}),$u,vu,xu,Su,Pc,Uc,d0=P(()=>{"use strict";Y(),te(),ve(),re(),$u=(e,t)=>{if(!e||e.length<1)throw new Error("too few inputs");let r=0,i=e[r],a=i.dataType,s=i.dims.length;e.forEach((n,u)=>{if(u!==r){if(n.dataType!==a)throw new Error("input tensors should be one type");if(n.dims.length!==s)throw new Error("input tensors should have the same shape");n.dims.forEach((l,p)=>{if(p!==t&&l!==i.dims[p])throw new Error("non concat dimensions must match")})}})},vu=(e,t)=>`
   fn calculateInputIndex(index: u32) -> u32 {
     let sizeInConcatAxis = array<u32, ${e}u>(${t});
     for (var i: u32 = 0u; i < ${e}; i += 1u ) {
@@ -546,7 +546,7 @@ fn quick_gelu_impl(x: vec4<${e}>) -> vec4<${e}> {
     }
     return ${e}u;
   }`,xu=(e,t)=>{let r=e.length,i=[];for(let a=0;a<r;++a){let s=t.setByOffset("global_idx",e[a].getByIndices("indices"));r===1?i.push(s):a===0?i.push(`if (inputIndex == ${a}u) { ${s} }`):a===r-1?i.push(`else { ${s} }`):i.push(`else if (inputIndex == ${a}) { ${s} }`)}return i.join(`
-`)},Su=(e,t,r,i)=>{let a=B.size(r),s=new Array(e.length),n=new Array(e.length),u=0,l=[],p=[],h=[{type:12,data:a}];for(let $=0;$<e.length;++$)u+=e[$].dims[t],s[$]=u,p.push(e[$].dims.length),n[$]=N(`input${$}`,i,p[$]),l.push("rank"),h.push({type:12,data:s[$]});for(let $=0;$<e.length;++$)h.push(...Z(e[$].dims));h.push(...Z(r));let f=F("output",i,r.length),g=f.indicesGet("indices",t),_=Array.from(Array(s.length).keys()).map($=>`uniforms.sizeInConcatAxis${$}`).join(","),y=$=>`
+`)},Su=(e,t,r,i)=>{let a=B.size(r),s=new Array(e.length),n=new Array(e.length),u=0,l=[],p=[],h=[{type:12,data:a}];for(let $=0;$<e.length;++$)u+=e[$].dims[t],s[$]=u,p.push(e[$].dims.length),n[$]=M(`input${$}`,i,p[$]),l.push("rank"),h.push({type:12,data:s[$]});for(let $=0;$<e.length;++$)h.push(...Z(e[$].dims));h.push(...Z(r));let f=F("output",i,r.length),g=f.indicesGet("indices",t),_=Array.from(Array(s.length).keys()).map($=>`uniforms.sizeInConcatAxis${$}`).join(","),y=$=>`
 
   ${(()=>{$.registerUniform("outputSize","u32");for(let S=0;S<e.length;S++)$.registerUniform(`sizeInConcatAxis${S}`,"u32");return $.declareVariables(...n,f)})()}
 
@@ -564,11 +564,11 @@ fn quick_gelu_impl(x: vec4<${e}>) -> vec4<${e}> {
     }
 
     ${xu(n,f)}
-  }`;return{name:"Concat",shaderCache:{hint:`${t}`,inputDependencies:l},getRunData:()=>({outputs:[{dims:r,dataType:i}],dispatchGroup:{x:Math.ceil(a/64)},programUniforms:h}),getShaderSource:y}},Pc=(e,t)=>{let r=e.inputs,i=r[0].dims,a=B.normalizeAxis(t.axis,i.length);$u(r,a);let s=i.slice();s[a]=r.reduce((u,l)=>u+(l.dims.length>a?l.dims[a]:0),0);let n=r.filter(u=>B.size(u.dims)>0);e.compute(Su(n,a,s,r[0].dataType),{inputs:n})},Uc=e=>he({axis:e.axis})}),Bt,Rt,Nt,Ka,Dt=P(()=>{"use strict";Y(),te(),Bt=(e,t,r="f32")=>{switch(e.activation){case"Relu":return`value = max(value, ${t}(0.0));`;case"Sigmoid":return`value = (${t}(1.0) / (${t}(1.0) + exp(-value)));`;case"Clip":return`value = clamp(value, ${t}(${r}(uniforms.clip_min)), ${t}(${r}(uniforms.clip_max)));`;case"HardSigmoid":return`value = max(${t}(0.0), min(${t}(1.0), ${r}(uniforms.alpha) * value + ${r}(uniforms.beta)));`;case"LeakyRelu":return`value = select(${r}(uniforms.alpha) * value, value, value >= ${t}(0.0));`;case"Tanh":return`let e2x = exp(-2.0 * abs(value));
+  }`;return{name:"Concat",shaderCache:{hint:`${t}`,inputDependencies:l},getRunData:()=>({outputs:[{dims:r,dataType:i}],dispatchGroup:{x:Math.ceil(a/64)},programUniforms:h}),getShaderSource:y}},Pc=(e,t)=>{let r=e.inputs,i=r[0].dims,a=B.normalizeAxis(t.axis,i.length);$u(r,a);let s=i.slice();s[a]=r.reduce((u,l)=>u+(l.dims.length>a?l.dims[a]:0),0);let n=r.filter(u=>B.size(u.dims)>0);e.compute(Su(n,a,s,r[0].dataType),{inputs:n})},Uc=e=>he({axis:e.axis})}),Bt,Rt,Mt,Ka,Dt=P(()=>{"use strict";Y(),te(),Bt=(e,t,r="f32")=>{switch(e.activation){case"Relu":return`value = max(value, ${t}(0.0));`;case"Sigmoid":return`value = (${t}(1.0) / (${t}(1.0) + exp(-value)));`;case"Clip":return`value = clamp(value, ${t}(${r}(uniforms.clip_min)), ${t}(${r}(uniforms.clip_max)));`;case"HardSigmoid":return`value = max(${t}(0.0), min(${t}(1.0), ${r}(uniforms.alpha) * value + ${r}(uniforms.beta)));`;case"LeakyRelu":return`value = select(${r}(uniforms.alpha) * value, value, value >= ${t}(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 ${e.activation}`)}},Rt=(e,t)=>{e.activation==="Clip"?t.push({type:1,data:e.clipMax},{type:1,data:e.clipMin}):e.activation==="HardSigmoid"?t.push({type:1,data:e.alpha},{type:1,data:e.beta}):e.activation==="LeakyRelu"&&t.push({type:1,data:e.alpha})},Nt=(e,t)=>{e.activation==="Clip"?t.push({name:"clip_max",type:"f32"},{name:"clip_min",type:"f32"}):e.activation==="HardSigmoid"?t.push({name:"alpha",type:"f32"},{name:"beta",type:"f32"}):e.activation==="LeakyRelu"&&t.push({name:"alpha",type:"f32"})},Ka=e=>{let t=e?.activation||"";if(t==="HardSigmoid"){let[r,i]=e?.activation_params||[.2,.5];return{activation:t,alpha:r,beta:i}}else if(t==="Clip"){let[r,i]=e?.activation_params||[pp,cp];return{activation:t,clipMax:i,clipMin:r}}else if(t==="LeakyRelu"){let[r]=e?.activation_params||[.01];return{activation:t,alpha:r}}return{activation:t}}}),ze,Wc,Za=P(()=>{"use strict";ze=(e,t)=>{switch(e){case 1:return t;case 2:return`vec2<${t}>`;case 3:return`vec3<${t}>`;case 4:return`vec4<${t}>`;default:throw new Error(`${e}-component is not supported.`)}},Wc=e=>`
+        `;case"":return"";default:throw new Error(`Unsupported activation ${e.activation}`)}},Rt=(e,t)=>{e.activation==="Clip"?t.push({type:1,data:e.clipMax},{type:1,data:e.clipMin}):e.activation==="HardSigmoid"?t.push({type:1,data:e.alpha},{type:1,data:e.beta}):e.activation==="LeakyRelu"&&t.push({type:1,data:e.alpha})},Mt=(e,t)=>{e.activation==="Clip"?t.push({name:"clip_max",type:"f32"},{name:"clip_min",type:"f32"}):e.activation==="HardSigmoid"?t.push({name:"alpha",type:"f32"},{name:"beta",type:"f32"}):e.activation==="LeakyRelu"&&t.push({name:"alpha",type:"f32"})},Ka=e=>{let t=e?.activation||"";if(t==="HardSigmoid"){let[r,i]=e?.activation_params||[.2,.5];return{activation:t,alpha:r,beta:i}}else if(t==="Clip"){let[r,i]=e?.activation_params||[pp,cp];return{activation:t,clipMax:i,clipMin:r}}else if(t==="LeakyRelu"){let[r]=e?.activation_params||[.01];return{activation:t,alpha:r}}return{activation:t}}}),ze,Wc,Za=P(()=>{"use strict";ze=(e,t)=>{switch(e){case 1:return t;case 2:return`vec2<${t}>`;case 3:return`vec3<${t}>`;case 4:return`vec4<${t}>`;default:throw new Error(`${e}-component is not supported.`)}},Wc=e=>`
       ${e?"value = value + getBiasByOutputCoords(coords);":""}
-      `}),qc,p0=P(()=>{"use strict";qc=e=>`
+      `}),Lc,p0=P(()=>{"use strict";Lc=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));
@@ -584,11 +584,11 @@ fn getOutputIndexFromCoords(coords : vec4<i32>) -> i32 {
       } else {
         ${t.indicesSet(e,u,0)}
       }`).join("")}
-`},Qa=(e,t,r,i,a=!1,s)=>{let n=e[0].dims,u=e[1].dims,l=n[n.length-2],p=u[u.length-1],h=n[n.length-1],f=$e(p),g=$e(h),_=$e(l),y=B.size(r)/f/_,$=e.length>2,S=i?i.slice(0,-2):r.slice(0,-2),v=[B.size(S),l,p],w=[{type:12,data:y},{type:12,data:l},{type:12,data:p},{type:12,data:h}];Rt(t,w),w.push(...Z(S,n,u)),$&&w.push(...Z(e[2].dims)),w.push(...Z(v));let E=I=>{let k=Ga("batch_dims",e[0].dataType,S.length),C=N("a",e[0].dataType,n.length,g),O=N("b",e[1].dataType,u.length,f),x=F("output",e[0].dataType,v.length,f),D=Ee(x.type.tensor),q=Bt(t,x.type.value,D),Q=[C,O],L="";if($){let A=a?f:1;Q.push(N("bias",e[2].dataType,e[2].dims.length,A)),L=`${a?`value += bias[col / ${A}];`:`value += ${x.type.value}(bias[row + i]);`}`}let K=[{name:"output_size",type:"u32"},{name:"M",type:"u32"},{name:"N",type:"u32"},{name:"K",type:"u32"}];Nt(t,K);let se=()=>{let A=`var a_data: ${C.type.value};`;for(let U=0;U<g;U++)A+=`
+`},Qa=(e,t,r,i,a=!1,s)=>{let n=e[0].dims,u=e[1].dims,l=n[n.length-2],p=u[u.length-1],h=n[n.length-1],f=$e(p),g=$e(h),_=$e(l),y=B.size(r)/f/_,$=e.length>2,S=i?i.slice(0,-2):r.slice(0,-2),v=[B.size(S),l,p],w=[{type:12,data:y},{type:12,data:l},{type:12,data:p},{type:12,data:h}];Rt(t,w),w.push(...Z(S,n,u)),$&&w.push(...Z(e[2].dims)),w.push(...Z(v));let k=I=>{let E=Ga("batch_dims",e[0].dataType,S.length),C=M("a",e[0].dataType,n.length,g),O=M("b",e[1].dataType,u.length,f),x=F("output",e[0].dataType,v.length,f),D=ke(x.type.tensor),L=Bt(t,x.type.value,D),Q=[C,O],q="";if($){let A=a?f:1;Q.push(M("bias",e[2].dataType,e[2].dims.length,A)),q=`${a?`value += bias[col / ${A}];`:`value += ${x.type.value}(bias[row + i]);`}`}let K=[{name:"output_size",type:"u32"},{name:"M",type:"u32"},{name:"N",type:"u32"},{name:"K",type:"u32"}];Mt(t,K);let se=()=>{let A=`var a_data: ${C.type.value};`;for(let U=0;U<g;U++)A+=`
               let b_data${U} = b[(b_offset + (k + ${U}) * uniforms.N + col) / ${f}];`;for(let U=0;U<_;U++){A+=`a_data = a[(a_offset + (row + ${U}) * uniforms.K + k) / ${g}];`;for(let ee=0;ee<g;ee++)A+=`
             values[${U}] = fma(${O.type.value}(a_data${g===1?"":`[${ee}]`}), b_data${ee}, values[${U}]);
 `}return A};return`
-  ${I.registerUniforms(K).registerInternalVariables(k).declareVariables(...Q,x)}
+  ${I.registerUniforms(K).registerInternalVariables(E).declareVariables(...Q,x)}
   ${I.mainStart()}
     ${I.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.output_size")}
     let col = (global_idx % (uniforms.N / ${f})) * ${f};
@@ -597,16 +597,16 @@ fn getOutputIndexFromCoords(coords : vec4<i32>) -> i32 {
     let row = (index1 % stride1) * ${_};
     let batch = index1 / stride1;
 
-    ${r.length===2?"":`let batch_indices = ${k.offsetToIndices("batch")};`}
+    ${r.length===2?"":`let batch_indices = ${E.offsetToIndices("batch")};`}
 
     var a_indices: ${C.type.indices};
-    ${dr("a_indices",C,C.rank-2,k.rank,"batch_indices")}
+    ${dr("a_indices",C,C.rank-2,E.rank,"batch_indices")}
     ${C.indicesSet("a_indices",C.rank-2,0)}
     ${C.indicesSet("a_indices",C.rank-1,0)}
     let a_offset = ${C.indicesToOffset("a_indices")};
 
     var b_indices: ${O.type.indices};
-    ${dr("b_indices",O,O.rank-2,k.rank,"batch_indices")}
+    ${dr("b_indices",O,O.rank-2,E.rank,"batch_indices")}
     ${O.indicesSet("b_indices",O.rank-2,0)}
     ${O.indicesSet("b_indices",O.rank-1,0)}
     let b_offset = ${O.indicesToOffset("b_indices")};
@@ -616,14 +616,14 @@ fn getOutputIndexFromCoords(coords : vec4<i32>) -> i32 {
     }
     for (var i = 0u; i < ${_}u; i++) {
       var value = values[i];
-      ${L}
       ${q}
+      ${L}
       let cur_indices = ${x.type.indices}(batch, row + i, col);
       let offset = ${x.indicesToOffset("cur_indices")};
       ${x.setByOffset(`offset / ${f}`,"value")};
     }
   }
-  `};return{name:"MatMulNaive",shaderCache:{hint:`${t.activation};${f};${g};${_};${a}`,inputDependencies:$?["rank","rank","rank"]:["rank","rank"]},getRunData:()=>({outputs:[{dims:s?s(r):r,dataType:e[0].dataType}],dispatchGroup:{x:Math.ceil(y/64)},programUniforms:w}),getShaderSource:E}}}),Tu,Iu,Ta,Li,Eu,Ia,ku,Hr,Ya=P(()=>{"use strict";Y(),te(),re(),Dt(),Xa(),Za(),Tu=(e,t)=>e?`
+  `};return{name:"MatMulNaive",shaderCache:{hint:`${t.activation};${f};${g};${_};${a}`,inputDependencies:$?["rank","rank","rank"]:["rank","rank"]},getRunData:()=>({outputs:[{dims:s?s(r):r,dataType:e[0].dataType}],dispatchGroup:{x:Math.ceil(y/64)},programUniforms:w}),getShaderSource:k}}}),Tu,Iu,Ta,qi,ku,Ia,Eu,Hr,Ya=P(()=>{"use strict";Y(),te(),re(),Dt(),Xa(),Za(),Tu=(e,t)=>e?`
         mm_Asub[inputRow][inputCol] = mm_readA(batch,
           kStart + inputRow,
           globalRowStart / innerElementSize + inputCol${t?", batchIndices":""});
@@ -713,7 +713,7 @@ fn main(@builtin(local_invocation_id) localId : vec3<u32>,
   for (var innerRow = 0; innerRow < rowPerThread; innerRow = innerRow + 1) {
       mm_write(batch, globalRow + innerRow, globalCol, acc[innerRow]);
   }
-}`},Li=(e,t)=>e?`
+}`},qi=(e,t)=>e?`
             mm_Asub[inputRow][inputCol] = mm_readA(batch,
               kStart + inputRow,
               globalRowStart + inputCol${t?", batchIndices":""});
@@ -721,7 +721,7 @@ fn main(@builtin(local_invocation_id) localId : vec3<u32>,
             mm_Asub[inputRow][inputCol] = mm_readA(batch,
               globalRowStart + inputRow,
               kStart + inputCol${t?", batchIndices":""});
-            `,Eu=e=>e?"let ACached = mm_Asub[k][tileRow + innerRow];":"let ACached = mm_Asub[tileRow + innerRow][k];",Ia=(e,t,r="f32",i,a=!1,s=32,n=!1,u=32,l=!1)=>{let p=e[1]*t[1],h=e[0]*t[0],f=a?p:s,g=a?s:p;if(!(g%t[1]===0&&f%t[0]===0&&s%t[1]===0))throw new Error(`tileAHight ${g} must be divisible by workgroupSize[1]${t[1]}, tileAWidth ${f} must be divisible by workgroupSize[0]${t[0]}, tileInner ${s} must be divisible by workgroupSize[1]${t[1]}`);let _=g/t[1],y=f/t[0],$=s/t[1],S=l?`
+            `,ku=e=>e?"let ACached = mm_Asub[k][tileRow + innerRow];":"let ACached = mm_Asub[tileRow + innerRow][k];",Ia=(e,t,r="f32",i,a=!1,s=32,n=!1,u=32,l=!1)=>{let p=e[1]*t[1],h=e[0]*t[0],f=a?p:s,g=a?s:p;if(!(g%t[1]===0&&f%t[0]===0&&s%t[1]===0))throw new Error(`tileAHight ${g} must be divisible by workgroupSize[1]${t[1]}, tileAWidth ${f} must be divisible by workgroupSize[0]${t[0]}, tileInner ${s} must be divisible by workgroupSize[1]${t[1]}`);let _=g/t[1],y=f/t[0],$=s/t[1],S=l?`
     let localRow = i32(localId.y);
     let localCol = i32(localId.x);
     let globalRowStart = i32(workgroupId.y) * ${p};
@@ -732,7 +732,7 @@ fn main(@builtin(local_invocation_id) localId : vec3<u32>,
       // Load one tile of A into local memory.
       for (var inputRow = localRow; inputRow < ${g}; inputRow = inputRow + ${t[1]}) {
         for (var inputCol = localCol; inputCol < ${f}; inputCol = inputCol + ${t[0]}) {
-          ${Li(a,i)}
+          ${qi(a,i)}
         }
       }
       // Load one tile of B into local memory.
@@ -787,7 +787,7 @@ for (var t = 0; t < num_tiles; t = t + 1) {
     for (var innerCol = 0; innerCol < ${y}; innerCol = innerCol + 1) {
       let inputRow = tileRowA + innerRow;
       let inputCol = tileColA + innerCol;
-      ${Li(a,i)}
+      ${qi(a,i)}
     }
   }
 
@@ -812,7 +812,7 @@ for (var t = 0; t < num_tiles; t = t + 1) {
     }
 
     for (var innerRow = 0; innerRow < rowPerThread; innerRow = innerRow + 1) {
-      ${Eu(a)}
+      ${ku(a)}
       for (var innerCol = 0; innerCol < colPerThread; innerCol = innerCol + 1) {
         acc[innerRow][innerCol] = acc[innerRow][innerCol] + ACached * BCached[innerCol];
       }
@@ -847,7 +847,7 @@ fn main(@builtin(local_invocation_id) localId : vec3<u32>,
     var acc : array<array<${r}, colPerThread>, rowPerThread>;
     ${S}
   }
-`},ku=(e,t,r,i,a=!1)=>{let[s,n,u,l]=i,p=Ee(i[0].type.tensor);return`
+`},Eu=(e,t,r,i,a=!1)=>{let[s,n,u,l]=i,p=ke(i[0].type.tensor);return`
     fn mm_readA(batch: i32, row: i32, colIn: i32, batchIndices: ${s.type.indices}) -> ${ze(e,p)} {
       var value = ${ze(e,p)}(0.0);
       let col = colIn * ${e};
@@ -886,11 +886,11 @@ fn main(@builtin(local_invocation_id) localId : vec3<u32>,
         ${l.setByIndices("vec3<u32>(coords)","value")}
       }
     }
-    `},Hr=(e,t,r,i,a=!1,s)=>{let n=e[0].dims,u=e[1].dims,l=n.slice(0,-2),p=u.slice(0,-2),h=i?i.slice(0,-2):r.slice(0,-2),f=B.size(h),g=n[n.length-2],_=n[n.length-1],y=u[u.length-1],$=_%4===0&&y%4===0,S=g<=8?[4,1,1]:[4,4,1],v=[8,8,1],w=[Math.ceil(y/v[0]/S[0]),Math.ceil(g/v[1]/S[1]),Math.ceil(f/v[2]/S[2])],E=$?4:1,I=[...l,g,_/E],k=I.length,C=[...p,_,y/E],O=C.length,x=[f,g,y/E],D=[{type:6,data:g},{type:6,data:y},{type:6,data:_}];Rt(t,D),D.push(...Z(h,I,C));let q=["rank","rank"],Q=e.length>2;Q&&(D.push(...Z(e[2].dims)),q.push("rank")),D.push(...Z(x));let L=K=>{let se=h.length,A=Ga("batchDims",e[0].dataType,se,1),U=Ee(e[0].dataType),ee=N("a",e[0].dataType,k,E),X=N("b",e[1].dataType,O,E),H=F("result",e[0].dataType,x.length,E),le=[ee,X];if(Q){let Te=a?E:1;le.push(N("bias",e[2].dataType,e[2].dims.length,Te))}let M=[{name:"dim_a_outer",type:"i32"},{name:"dim_b_outer",type:"i32"},{name:"dim_inner",type:"i32"}];Nt(t,M);let V=Ee(H.type.tensor),J=Bt(t,H.type.value,V),de=ku(E,Q,J,[A,ee,X,H],a);return`
-  ${K.registerUniforms(M).registerInternalVariables(A).declareVariables(...le,H)}
+    `},Hr=(e,t,r,i,a=!1,s)=>{let n=e[0].dims,u=e[1].dims,l=n.slice(0,-2),p=u.slice(0,-2),h=i?i.slice(0,-2):r.slice(0,-2),f=B.size(h),g=n[n.length-2],_=n[n.length-1],y=u[u.length-1],$=_%4===0&&y%4===0,S=g<=8?[4,1,1]:[4,4,1],v=[8,8,1],w=[Math.ceil(y/v[0]/S[0]),Math.ceil(g/v[1]/S[1]),Math.ceil(f/v[2]/S[2])],k=$?4:1,I=[...l,g,_/k],E=I.length,C=[...p,_,y/k],O=C.length,x=[f,g,y/k],D=[{type:6,data:g},{type:6,data:y},{type:6,data:_}];Rt(t,D),D.push(...Z(h,I,C));let L=["rank","rank"],Q=e.length>2;Q&&(D.push(...Z(e[2].dims)),L.push("rank")),D.push(...Z(x));let q=K=>{let se=h.length,A=Ga("batchDims",e[0].dataType,se,1),U=ke(e[0].dataType),ee=M("a",e[0].dataType,E,k),X=M("b",e[1].dataType,O,k),H=F("result",e[0].dataType,x.length,k),le=[ee,X];if(Q){let Ie=a?k:1;le.push(M("bias",e[2].dataType,e[2].dims.length,Ie))}let N=[{name:"dim_a_outer",type:"i32"},{name:"dim_b_outer",type:"i32"},{name:"dim_inner",type:"i32"}];Mt(t,N);let V=ke(H.type.tensor),J=Bt(t,H.type.value,V),de=Eu(k,Q,J,[A,ee,X,H],a);return`
+  ${K.registerUniforms(N).registerInternalVariables(A).declareVariables(...le,H)}
   ${de}
   ${$?Ta(S,v,U,A):Ia(S,v,U,A)}
-                   `};return{name:"MatMul",shaderCache:{hint:`${S};${t.activation};${$};${a}`,inputDependencies:q},getRunData:()=>({outputs:[{dims:s?s(r):r,dataType:e[0].dataType}],dispatchGroup:{x:w[0],y:w[1],z:w[2]},programUniforms:D}),getShaderSource:L}}}),zu,Lc,c0=P(()=>{"use strict";Y(),nt(),re(),Dt(),Za(),p0(),Ya(),zu=(e,t,r,i,a=!1,s,n=4,u=4,l=4,p="f32")=>{let h=D=>{switch(D){case 1:return"resData = x[xIndex];";case 3:return`resData = vec3<${p}>(x[xIndex], x[xIndex + 1], x[xIndex + 2]);`;case 4:return"resData = x[xIndex / 4];";default:throw new Error(`innerElementSize ${D} is not supported.`)}},f=D=>{switch(D){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 ${D} is not supported.`)}},g=e?`
+                   `};return{name:"MatMul",shaderCache:{hint:`${S};${t.activation};${$};${a}`,inputDependencies:L},getRunData:()=>({outputs:[{dims:s?s(r):r,dataType:e[0].dataType}],dispatchGroup:{x:w[0],y:w[1],z:w[2]},programUniforms:D}),getShaderSource:q}}}),zu,qc,c0=P(()=>{"use strict";Y(),nt(),re(),Dt(),Za(),p0(),Ya(),zu=(e,t,r,i,a=!1,s,n=4,u=4,l=4,p="f32")=>{let h=D=>{switch(D){case 1:return"resData = x[xIndex];";case 3:return`resData = vec3<${p}>(x[xIndex], x[xIndex + 1], x[xIndex + 2]);`;case 4:return"resData = x[xIndex / 4];";default:throw new Error(`innerElementSize ${D} is not supported.`)}},f=D=>{switch(D){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 ${D} is not supported.`)}},g=e?`
     let coord = vec4<i32>(batch, xRow, xCol, xCh);
     `:`
     let coord = vec4<i32>(batch, xCh, xRow, xCol);
@@ -925,7 +925,7 @@ fn main(@builtin(local_invocation_id) localId : vec3<u32>,
       let xIndex = getIndexFromCoords4D(coord, vec4<i32>(uniforms.x_shape));
       ${h(n)}
     }
-    return resData;`,E=e?t&&i?`
+    return resData;`,k=e?t&&i?`
     let col = colIn * ${n};
     ${w}`:`
     let col = colIn * ${n};
@@ -949,16 +949,16 @@ fn main(@builtin(local_invocation_id) localId : vec3<u32>,
     if (row < uniforms.dim_inner && col < uniforms.dim_a_outer) {
       ${f(u)}
     }
-    return ${ze(u,p)}(0.0);`,k=ze(l,p),C=ze(e?n:u,p),O=ze(e?u:n,p),x=Bt(s,k,p);return`
+    return ${ze(u,p)}(0.0);`,E=ze(l,p),C=ze(e?n:u,p),O=ze(e?u:n,p),x=Bt(s,E,p);return`
     fn mm_readA(batch: i32, row : i32, colIn : i32) -> ${C} {
-      ${e?E:I}
+      ${e?k:I}
     }
 
     fn mm_readB(batch: i32, row : i32, colIn : i32) -> ${O} {
-      ${e?I:E}
+      ${e?I:k}
     }
 
-    fn mm_write(batch: i32, row : i32, colIn : i32, valueIn : ${k}) {
+    fn mm_write(batch: i32, row : i32, colIn : i32, valueIn : ${E}) {
       let col = colIn * ${l};
       if (row < uniforms.dim_a_outer && col < uniforms.dim_b_outer)
       {
@@ -969,29 +969,29 @@ fn main(@builtin(local_invocation_id) localId : vec3<u32>,
       ${x}
       setOutputAtCoords(coords[0], coords[1], coords[2], coords[3], value);
       }
-    }`},Lc=(e,t,r,i,a,s,n,u,l)=>{let p=t.format==="NHWC",h=p?e[0].dims[3]:e[0].dims[1],f=r[0],g=p?r[2]:r[3],_=p?r[1]:r[2],y=p?r[3]:r[1],$=p&&(h%4===0||h%3===0)&&y%4===0,S=p?y:g*_,v=p?g*_:y,w=[8,8,1],E=i<=8?[4,1,1]:[4,4,1],I=[Math.ceil(S/w[0]/E[0]),Math.ceil(v/w[1]/E[1]),Math.ceil(f/w[2]/E[2])];ue("verbose",()=>`[conv2d_mm_webgpu] dispatch = ${I}`);let k=$?p&&h%4!==0?3:4:1,C=w[1]*E[1],O=w[0]*E[0],x=Math.max(w[0]*k,w[1]),D=i%C===0,q=a%O===0,Q=s%x===0,L=$?[k,4,4]:[1,1,1],K=[{type:6,data:i},{type:6,data:a},{type:6,data:s},{type:6,data:[t.pads[0],t.pads[1]]},{type:6,data:t.strides},{type:6,data:t.dilations}];Rt(t,K),K.push(...Z(e[0].dims,e[1].dims));let se=["rank","rank"];n&&(K.push(...Z(e[2].dims)),se.push("rank")),K.push(...Z(r));let A=U=>{let ee=[{name:"dim_a_outer",type:"i32"},{name:"dim_b_outer",type:"i32"},{name:"dim_inner",type:"i32"},{name:"pad",type:"i32",length:2},{name:"stride",type:"i32",length:2},{name:"dilation",type:"i32",length:2}];Nt(t,ee);let X=$?4:1,H=Ee(e[0].dataType),le=`
+    }`},qc=(e,t,r,i,a,s,n,u,l)=>{let p=t.format==="NHWC",h=p?e[0].dims[3]:e[0].dims[1],f=r[0],g=p?r[2]:r[3],_=p?r[1]:r[2],y=p?r[3]:r[1],$=p&&(h%4===0||h%3===0)&&y%4===0,S=p?y:g*_,v=p?g*_:y,w=[8,8,1],k=i<=8?[4,1,1]:[4,4,1],I=[Math.ceil(S/w[0]/k[0]),Math.ceil(v/w[1]/k[1]),Math.ceil(f/w[2]/k[2])];ue("verbose",()=>`[conv2d_mm_webgpu] dispatch = ${I}`);let E=$?p&&h%4!==0?3:4:1,C=w[1]*k[1],O=w[0]*k[0],x=Math.max(w[0]*E,w[1]),D=i%C===0,L=a%O===0,Q=s%x===0,q=$?[E,4,4]:[1,1,1],K=[{type:6,data:i},{type:6,data:a},{type:6,data:s},{type:6,data:[t.pads[0],t.pads[1]]},{type:6,data:t.strides},{type:6,data:t.dilations}];Rt(t,K),K.push(...Z(e[0].dims,e[1].dims));let se=["rank","rank"];n&&(K.push(...Z(e[2].dims)),se.push("rank")),K.push(...Z(r));let A=U=>{let ee=[{name:"dim_a_outer",type:"i32"},{name:"dim_b_outer",type:"i32"},{name:"dim_inner",type:"i32"},{name:"pad",type:"i32",length:2},{name:"stride",type:"i32",length:2},{name:"dilation",type:"i32",length:2}];Mt(t,ee);let X=$?4:1,H=ke(e[0].dataType),le=`
       fn setOutputAtIndex(flatIndex : i32, value : ${$?`vec4<${H}>`:H}) {
         result[flatIndex] = ${$?`vec4<${H}>`:H}(value);
       }
       fn setOutputAtCoords(d0 : i32, d1 : i32, d2 : i32, d3 : i32, value : ${$?`vec4<${H}>`:H}) {
         let flatIndex = getOutputIndexFromCoords(vec4<i32>(d0, d1, d2, d3));
         setOutputAtIndex(flatIndex ${$?"/ 4":""}, value);
-      }`,M=N("x",e[0].dataType,e[0].dims.length,k===3?1:k),V=N("w",e[1].dataType,e[1].dims.length,X),J=[M,V],de=F("result",e[0].dataType,r.length,X);if(n){let Te=N("bias",e[2].dataType,e[2].dims.length,X);J.push(Te),le+=`
+      }`,N=M("x",e[0].dataType,e[0].dims.length,E===3?1:E),V=M("w",e[1].dataType,e[1].dims.length,X),J=[N,V],de=F("result",e[0].dataType,r.length,X);if(n){let Ie=M("bias",e[2].dataType,e[2].dims.length,X);J.push(Ie),le+=`
         fn getBiasByOutputCoords(coords : vec4<i32>) -> ${$?`vec4<${H}>`:H} {
           return bias[coords.${p?"w":"y"}${$?"/ 4":""}];
         }`}return`
-        ${qc("uniforms.result_strides")}
+        ${Lc("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 };
         ${U.registerUniforms(ee).declareVariables(...J,de)}
         ${le}
-        ${zu(p,D,q,Q,n,t,L[0],L[1],L[2],H)}
-        ${$?Ta(E,w,H,void 0,!p,x):Ia(E,w,H,void 0,!p,x,!1,void 0,u)}`};return{name:"Conv2DMatMul",shaderCache:{hint:`${t.cacheKey};${k};${$};${D};${q};${Q};${C};${O};${x}`,inputDependencies:se},getRunData:()=>({outputs:[{dims:l?l(r):r,dataType:e[0].dataType}],dispatchGroup:{x:I[0],y:I[1],z:I[2]},programUniforms:K}),getShaderSource:A}}}),Cu,Vi,er,Au,Gi,Ou,Vc,Gc,h0=P(()=>{"use strict";Y(),nt(),te(),re(),Dt(),Za(),Cu=e=>{let t=1;for(let r=0;r<e.length;r++)t*=e[r];return t},Vi=e=>typeof e=="number"?[e,e,e]:e,er=(e,t)=>t<=1?e:e+(e-1)*(t-1),Au=(e,t,r,i=1)=>{let a=er(t,i);return Math.floor((e[0]*(r-1)-r+a)/2)},Gi=(e,t,r,i,a)=>{a==null&&(a=Au(e,t[0],i[0]));let s=[0,0,0,r];for(let n=0;n<3;n++)e[n]+2*a>=t[n]&&(s[n]=Math.trunc((e[n]-t[n]+2*a)/i[n]+1));return s},Ou=(e,t,r,i,a,s,n,u,l,p)=>{let h,f,g,_;if(e==="VALID"&&(e=0),typeof e=="number"){h={top:e,bottom:e,left:e,right:e,front:e,back:e};let y=Gi([t,r,i,1],[u,l,p],1,[a,s,n],e);f=y[0],g=y[1],_=y[2]}else if(Array.isArray(e)){if(!e.every(($,S,v)=>$===v[0]))throw Error(`Unsupported padding parameter: ${e}`);h={top:e[0],bottom:e[1],left:e[2],right:e[3],front:e[4],back:e[5]};let y=Gi([t,r,i,1],[u,l,p],1,[a,s,n],e[0]);f=y[0],g=y[1],_=y[2]}else if(e==="SAME_UPPER"){f=Math.ceil(t/a),g=Math.ceil(r/s),_=Math.ceil(i/n);let y=(f-1)*a+u-t,$=(g-1)*s+l-r,S=(_-1)*n+p-i,v=Math.floor(y/2),w=y-v,E=Math.floor($/2),I=$-E,k=Math.floor(S/2),C=S-k;h={top:E,bottom:I,left:k,right:C,front:v,back:w}}else throw Error(`Unknown padding parameter: ${e}`);return{padInfo:h,outDepth:f,outHeight:g,outWidth:_}},Vc=(e,t,r,i,a,s=!1,n="channelsLast")=>{let u,l,p,h,f;if(n==="channelsLast")[u,l,p,h,f]=e;else if(n==="channelsFirst")[u,f,l,p,h]=e;else throw new Error(`Unknown dataFormat ${n}`);let[g,,_,y,$]=t,[S,v,w]=Vi(r),[E,I,k]=Vi(i),C=er(_,E),O=er(y,I),x=er($,k),{padInfo:D,outDepth:q,outHeight:Q,outWidth:L}=Ou(a,l,p,h,S,v,w,C,O,x),K=s?g*f:g,se=[0,0,0,0,0];return n==="channelsFirst"?se=[u,K,q,Q,L]:n==="channelsLast"&&(se=[u,q,Q,L,K]),{batchSize:u,dataFormat:n,inDepth:l,inHeight:p,inWidth:h,inChannels:f,outDepth:q,outHeight:Q,outWidth:L,outChannels:K,padInfo:D,strideDepth:S,strideHeight:v,strideWidth:w,filterDepth:_,filterHeight:y,filterWidth:$,effectiveFilterDepth:C,effectiveFilterHeight:O,effectiveFilterWidth:x,dilationDepth:E,dilationHeight:I,dilationWidth:k,inShape:e,outShape:se,filterShape:t}},Gc=(e,t,r,i,a,s)=>{let n=s==="channelsLast",u=n?e[0].dims[3]:e[0].dims[1],l=!1,p=[64,1,1],h={x:r.map((w,E)=>E)},f=[Math.ceil(Cu(h.x.map(w=>r[w]))/p[0]),1,1];ue("verbose",()=>`[conv3d_naive_webgpu] dispatch = ${f}`);let g=l?n&&u%4!==0?3:4:1,_=B.size(r),y=[{type:12,data:_},{type:12,data:i},{type:12,data:a},{type:12,data:t.strides},{type:12,data:t.dilations}];Rt(t,y),y.push(...Z(e[0].dims,e[1].dims));let $=["rank","rank"],S=e.length===3;S&&(y.push(...Z(e[2].dims)),$.push("rank")),y.push(...Z(r));let v=w=>{let E=[{name:"output_size",type:"u32"},{name:"filter_dims",type:"u32",length:i.length},{name:"pads",type:"u32",length:a.length},{name:"strides",type:"u32",length:t.strides.length},{name:"dilations",type:"u32",length:t.dilations.length}];Nt(t,E);let I=l?4:1,k=Ee(e[0].dataType),C=N("x",e[0].dataType,e[0].dims.length,g===3?1:g),O=N("W",e[1].dataType,e[1].dims.length,I),x=[C,O],D=F("result",e[0].dataType,r.length,I),q="";if(S){let K=N("bias",e[2].dataType,e[2].dims.length,I);x.push(K),q+=`
-        fn getBiasByOutputCoords(coords : array<u32, 5>) -> ${l?`vec4<${k}>`:k} {
+        ${zu(p,D,L,Q,n,t,q[0],q[1],q[2],H)}
+        ${$?Ta(k,w,H,void 0,!p,x):Ia(k,w,H,void 0,!p,x,!1,void 0,u)}`};return{name:"Conv2DMatMul",shaderCache:{hint:`${t.cacheKey};${E};${$};${D};${L};${Q};${C};${O};${x}`,inputDependencies:se},getRunData:()=>({outputs:[{dims:l?l(r):r,dataType:e[0].dataType}],dispatchGroup:{x:I[0],y:I[1],z:I[2]},programUniforms:K}),getShaderSource:A}}}),Cu,Vi,er,Au,Gi,Ou,Vc,Gc,h0=P(()=>{"use strict";Y(),nt(),te(),re(),Dt(),Za(),Cu=e=>{let t=1;for(let r=0;r<e.length;r++)t*=e[r];return t},Vi=e=>typeof e=="number"?[e,e,e]:e,er=(e,t)=>t<=1?e:e+(e-1)*(t-1),Au=(e,t,r,i=1)=>{let a=er(t,i);return Math.floor((e[0]*(r-1)-r+a)/2)},Gi=(e,t,r,i,a)=>{a==null&&(a=Au(e,t[0],i[0]));let s=[0,0,0,r];for(let n=0;n<3;n++)e[n]+2*a>=t[n]&&(s[n]=Math.trunc((e[n]-t[n]+2*a)/i[n]+1));return s},Ou=(e,t,r,i,a,s,n,u,l,p)=>{let h,f,g,_;if(e==="VALID"&&(e=0),typeof e=="number"){h={top:e,bottom:e,left:e,right:e,front:e,back:e};let y=Gi([t,r,i,1],[u,l,p],1,[a,s,n],e);f=y[0],g=y[1],_=y[2]}else if(Array.isArray(e)){if(!e.every(($,S,v)=>$===v[0]))throw Error(`Unsupported padding parameter: ${e}`);h={top:e[0],bottom:e[1],left:e[2],right:e[3],front:e[4],back:e[5]};let y=Gi([t,r,i,1],[u,l,p],1,[a,s,n],e[0]);f=y[0],g=y[1],_=y[2]}else if(e==="SAME_UPPER"){f=Math.ceil(t/a),g=Math.ceil(r/s),_=Math.ceil(i/n);let y=(f-1)*a+u-t,$=(g-1)*s+l-r,S=(_-1)*n+p-i,v=Math.floor(y/2),w=y-v,k=Math.floor($/2),I=$-k,E=Math.floor(S/2),C=S-E;h={top:k,bottom:I,left:E,right:C,front:v,back:w}}else throw Error(`Unknown padding parameter: ${e}`);return{padInfo:h,outDepth:f,outHeight:g,outWidth:_}},Vc=(e,t,r,i,a,s=!1,n="channelsLast")=>{let u,l,p,h,f;if(n==="channelsLast")[u,l,p,h,f]=e;else if(n==="channelsFirst")[u,f,l,p,h]=e;else throw new Error(`Unknown dataFormat ${n}`);let[g,,_,y,$]=t,[S,v,w]=Vi(r),[k,I,E]=Vi(i),C=er(_,k),O=er(y,I),x=er($,E),{padInfo:D,outDepth:L,outHeight:Q,outWidth:q}=Ou(a,l,p,h,S,v,w,C,O,x),K=s?g*f:g,se=[0,0,0,0,0];return n==="channelsFirst"?se=[u,K,L,Q,q]:n==="channelsLast"&&(se=[u,L,Q,q,K]),{batchSize:u,dataFormat:n,inDepth:l,inHeight:p,inWidth:h,inChannels:f,outDepth:L,outHeight:Q,outWidth:q,outChannels:K,padInfo:D,strideDepth:S,strideHeight:v,strideWidth:w,filterDepth:_,filterHeight:y,filterWidth:$,effectiveFilterDepth:C,effectiveFilterHeight:O,effectiveFilterWidth:x,dilationDepth:k,dilationHeight:I,dilationWidth:E,inShape:e,outShape:se,filterShape:t}},Gc=(e,t,r,i,a,s)=>{let n=s==="channelsLast",u=n?e[0].dims[3]:e[0].dims[1],l=!1,p=[64,1,1],h={x:r.map((w,k)=>k)},f=[Math.ceil(Cu(h.x.map(w=>r[w]))/p[0]),1,1];ue("verbose",()=>`[conv3d_naive_webgpu] dispatch = ${f}`);let g=l?n&&u%4!==0?3:4:1,_=B.size(r),y=[{type:12,data:_},{type:12,data:i},{type:12,data:a},{type:12,data:t.strides},{type:12,data:t.dilations}];Rt(t,y),y.push(...Z(e[0].dims,e[1].dims));let $=["rank","rank"],S=e.length===3;S&&(y.push(...Z(e[2].dims)),$.push("rank")),y.push(...Z(r));let v=w=>{let k=[{name:"output_size",type:"u32"},{name:"filter_dims",type:"u32",length:i.length},{name:"pads",type:"u32",length:a.length},{name:"strides",type:"u32",length:t.strides.length},{name:"dilations",type:"u32",length:t.dilations.length}];Mt(t,k);let I=l?4:1,E=ke(e[0].dataType),C=M("x",e[0].dataType,e[0].dims.length,g===3?1:g),O=M("W",e[1].dataType,e[1].dims.length,I),x=[C,O],D=F("result",e[0].dataType,r.length,I),L="";if(S){let K=M("bias",e[2].dataType,e[2].dims.length,I);x.push(K),L+=`
+        fn getBiasByOutputCoords(coords : array<u32, 5>) -> ${l?`vec4<${E}>`:E} {
           return bias[${n?j("coords",4,5):j("coords",1,5)}${l?"/ 4":""}];
-        }`}let Q=ze(g,k),L=Bt(t,Q,k);return`
-            ${q}
+        }`}let Q=ze(g,E),q=Bt(t,Q,E);return`
+            ${L}
             fn getX(d0 : u32, d1 : u32, d2 : u32, d3 : u32, d4 : u32) -> f32 {
               let aIndices = array<u32, 5>(d0, d1, d2, d3, d4);
               return ${C.getByIndices("aIndices")};
@@ -1000,7 +1000,7 @@ fn main(@builtin(local_invocation_id) localId : vec3<u32>,
               let aIndices = array<u32, 5>(d0, d1, d2, d3, d4);
               return ${O.getByIndices("aIndices")};
             }
-          ${w.registerUniforms(E).declareVariables(...x,D)}
+          ${w.registerUniforms(k).declareVariables(...x,D)}
           ${w.mainStart()}
           ${w.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.output_size")}
               let coords = ${D.offsetToIndices("global_idx")};
@@ -1093,9 +1093,9 @@ fn main(@builtin(local_invocation_id) localId : vec3<u32>,
                 }
               }
               ${S?"value = value + getBiasByOutputCoords(coords)":""};
-              ${L}
+              ${q}
               result[global_idx] = f32(value);
-          }`};return{name:"Conv3DNaive",shaderCache:{hint:`${t.cacheKey};${n};${g};${S}`,inputDependencies:$},getRunData:()=>({outputs:[{dims:r,dataType:e[0].dataType}],dispatchGroup:{x:f[0],y:f[1],z:f[2]},programUniforms:y}),getShaderSource:v}}}),Fc,Hc,f0=P(()=>{"use strict";Y(),te(),re(),Dt(),Fc=(e,t,r,i)=>{let a=e.length>2,s=a?"value += b[output_channel];":"",n=e[0].dims,u=e[1].dims,l=t.format==="NHWC",p=l?r[3]:r[1],h=p/t.group,f=l&&h>=4?$e(p):1,g=B.size(r)/f,_=[{type:12,data:g},{type:12,data:t.dilations},{type:12,data:[t.strides[0],t.strides[1]]},{type:12,data:[t.pads[0],t.pads[1]]},{type:12,data:h}];Rt(t,_),_.push(...Z(n,[u[0],u[1],u[2],u[3]/f]));let y=a?["rank","rank","rank"]:["rank","rank"];_.push(...Z([r[0],r[1],r[2],r[3]/f]));let $=S=>{let v=F("output",e[0].dataType,r.length,f),w=Ee(v.type.tensor),E=Bt(t,v.type.value,w),I=N("x",e[0].dataType,n.length),k=N("w",e[1].dataType,u.length,f),C=[I,k];a&&C.push(N("b",e[2].dataType,e[2].dims,f));let O=[{name:"output_size",type:"u32"},{name:"dilations",type:"u32",length:t.dilations.length},{name:"strides",type:"u32",length:2},{name:"pads",type:"u32",length:2},{name:"output_channels_per_group",type:"u32"}];Nt(t,O);let x=l?`
+          }`};return{name:"Conv3DNaive",shaderCache:{hint:`${t.cacheKey};${n};${g};${S}`,inputDependencies:$},getRunData:()=>({outputs:[{dims:r,dataType:e[0].dataType}],dispatchGroup:{x:f[0],y:f[1],z:f[2]},programUniforms:y}),getShaderSource:v}}}),Fc,Hc,f0=P(()=>{"use strict";Y(),te(),re(),Dt(),Fc=(e,t,r,i)=>{let a=e.length>2,s=a?"value += b[output_channel];":"",n=e[0].dims,u=e[1].dims,l=t.format==="NHWC",p=l?r[3]:r[1],h=p/t.group,f=l&&h>=4?$e(p):1,g=B.size(r)/f,_=[{type:12,data:g},{type:12,data:t.dilations},{type:12,data:[t.strides[0],t.strides[1]]},{type:12,data:[t.pads[0],t.pads[1]]},{type:12,data:h}];Rt(t,_),_.push(...Z(n,[u[0],u[1],u[2],u[3]/f]));let y=a?["rank","rank","rank"]:["rank","rank"];_.push(...Z([r[0],r[1],r[2],r[3]/f]));let $=S=>{let v=F("output",e[0].dataType,r.length,f),w=ke(v.type.tensor),k=Bt(t,v.type.value,w),I=M("x",e[0].dataType,n.length),E=M("w",e[1].dataType,u.length,f),C=[I,E];a&&C.push(M("b",e[2].dataType,e[2].dims,f));let O=[{name:"output_size",type:"u32"},{name:"dilations",type:"u32",length:t.dilations.length},{name:"strides",type:"u32",length:2},{name:"pads",type:"u32",length:2},{name:"output_channels_per_group",type:"u32"}];Mt(t,O);let x=l?`
       for (var wHeight: u32 = 0u; wHeight < uniforms.w_shape[0]; wHeight++) {
         let xHeight = xRCCorner.x + wHeight * uniforms.dilations[0];
 
@@ -1112,7 +1112,7 @@ fn main(@builtin(local_invocation_id) localId : vec3<u32>,
           for (var wInChannel: u32 = 0u; wInChannel < uniforms.w_shape[2]; wInChannel++) {
             let input_channel = in_channel_offset + wInChannel;
             let xVal = ${I.get("batch","xHeight","xWidth","input_channel")};
-            let wVal = ${k.get("wHeight","wWidth","wInChannel","output_channel")};
+            let wVal = ${E.get("wHeight","wWidth","wInChannel","output_channel")};
             value += xVal * wVal;
           }
         }
@@ -1134,7 +1134,7 @@ fn main(@builtin(local_invocation_id) localId : vec3<u32>,
             }
 
             let xVal = ${I.get("batch","input_channel","xHeight","xWidth")};
-            let wVal = ${k.get("output_channel","wInChannel","wHeight","wWidth")};
+            let wVal = ${E.get("output_channel","wInChannel","wHeight","wWidth")};
             value += xVal * wVal;
           }
         }
@@ -1155,9 +1155,9 @@ fn main(@builtin(local_invocation_id) localId : vec3<u32>,
     var value: ${v.type.value} = ${v.type.value}(0);
     ${x}
     ${s}
-    ${E}
+    ${k}
     ${v.setByOffset("global_idx","value")}
-  }`};return{name:"GroupedConv",shaderCache:{hint:`${t.cacheKey}_${f}`,inputDependencies:y},getRunData:()=>({outputs:[{dims:i?i(r):r,dataType:e[0].dataType}],dispatchGroup:{x:Math.ceil(g/64)},programUniforms:_}),getShaderSource:$}},Hc=(e,t,r,i)=>{let a=e.length>2,s=$e(r[3]),n=$e(r[2]),u=B.size(r)/s/n,l=[e[0].dims[0],e[0].dims[1],e[0].dims[2],e[0].dims[3]/s],p=[e[1].dims[0],e[1].dims[1],e[1].dims[2],e[1].dims[3]/s],h=[r[0],r[1],r[2],r[3]/s],f=[{type:12,data:u},{type:6,data:[t.strides[0],t.strides[1]]},{type:6,data:[t.pads[0],t.pads[1]]}];Rt(t,f),f.push(...Z(l,p,h));let g=(n-1)*t.strides[1]+p[1],_=y=>{let $=F("output",e[0].dataType,h.length,s),S=Ee($.type.tensor),v=Bt(t,$.type.value,S),w=N("x",e[0].dataType,l.length,s),E=N("w",e[1].dataType,p.length,s),I=[w,E];a&&I.push(N("b",e[2].dataType,e[2].dims,s));let k=a?"value += b[output_channel];":"",C=[{name:"output_size",type:"u32"},{name:"strides",type:"i32",length:2},{name:"pads",type:"i32",length:2}];return Nt(t,C),`
+  }`};return{name:"GroupedConv",shaderCache:{hint:`${t.cacheKey}_${f}`,inputDependencies:y},getRunData:()=>({outputs:[{dims:i?i(r):r,dataType:e[0].dataType}],dispatchGroup:{x:Math.ceil(g/64)},programUniforms:_}),getShaderSource:$}},Hc=(e,t,r,i)=>{let a=e.length>2,s=$e(r[3]),n=$e(r[2]),u=B.size(r)/s/n,l=[e[0].dims[0],e[0].dims[1],e[0].dims[2],e[0].dims[3]/s],p=[e[1].dims[0],e[1].dims[1],e[1].dims[2],e[1].dims[3]/s],h=[r[0],r[1],r[2],r[3]/s],f=[{type:12,data:u},{type:6,data:[t.strides[0],t.strides[1]]},{type:6,data:[t.pads[0],t.pads[1]]}];Rt(t,f),f.push(...Z(l,p,h));let g=(n-1)*t.strides[1]+p[1],_=y=>{let $=F("output",e[0].dataType,h.length,s),S=ke($.type.tensor),v=Bt(t,$.type.value,S),w=M("x",e[0].dataType,l.length,s),k=M("w",e[1].dataType,p.length,s),I=[w,k];a&&I.push(M("b",e[2].dataType,e[2].dims,s));let E=a?"value += b[output_channel];":"",C=[{name:"output_size",type:"u32"},{name:"strides",type:"i32",length:2},{name:"pads",type:"i32",length:2}];return Mt(t,C),`
   ${y.registerUniforms(C).declareVariables(...I,$)}
   ${y.mainStart()}
     ${y.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.output_size")}
@@ -1188,7 +1188,7 @@ fn main(@builtin(local_invocation_id) localId : vec3<u32>,
           }
         }
         for (var w_width: u32 = 0u; w_width < ${p[1]}; w_width++) {
-          let w_val = ${E.get("w_height","w_width","0","output_channel")};
+          let w_val = ${k.get("w_height","w_width","0","output_channel")};
           for (var i = 0u; i < ${n}u; i++) {
             values[i] = fma(x_vals[i * u32(uniforms.strides[1]) + w_width], w_val, values[i]);
           }
@@ -1198,20 +1198,20 @@ fn main(@builtin(local_invocation_id) localId : vec3<u32>,
 
     for (var i = 0u; i < ${n}u; i++) {
       var value = values[i];
-      ${k}
+      ${E}
       ${v}
       ${$.set("batch","row","col + i","output_channel","value")};
     }
-  }`};return{name:"GroupedConv-Vectorize",shaderCache:{hint:`${t.cacheKey};${s};${n};${g};${p[0]};${p[1]}`,inputDependencies:a?["rank","rank","type"]:["rank","rank"]},getRunData:()=>({outputs:[{dims:i?i(r):r,dataType:e[0].dataType}],dispatchGroup:{x:Math.ceil(u/64)},programUniforms:f}),getShaderSource:_}}}),Bu,Rr,Ru,Nr,Ea,Fi,Nu,Mu,ka,m0=P(()=>{"use strict";te(),c0(),h0(),Ya(),f0(),Dt(),Xa(),wt(),Bu=(e,t,r,i,a,s)=>{let n=e[0],u=e.slice(s?1:2,s?3:4),l=u.length,p=t[0],h=t.slice(2).map((g,_)=>g+(g-1)*(r[_]-1)),f=u.map((g,_)=>g+i[_]+i[_+l]).map((g,_)=>Math.floor((g-h[_]+a[_])/a[_]));return f.splice(0,0,n),f.splice(s?3:1,0,p),f},Rr=[2,3,1,0],Ru=(e,t)=>{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 r=e[0].dims[t.format==="NHWC"?e[0].dims.length-1:1],i=e[1].dims[1]*t.group;if(r!==i)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 a=e[0].dims.length-2;if(t.dilations.length!==a)throw new Error(`dilations should be ${a}D`);if(t.strides.length!==a)throw new Error(`strides should be ${a}D`);if(t.pads.length!==a*2)throw new Error(`pads should be ${a*2}D`);if(t.kernelShape.length!==0&&t.kernelShape.length!==e[1].dims.length-2)throw new Error("invalid kernel shape")},Nr=(e,t)=>{let r=e.kernelShape.slice();r.length<t[1].dims.length-2&&r.push(...Array(t[1].dims.length-2-r.length).fill(0));for(let s=2;s<t[1].dims.length;++s)r[s-2]===0&&(r[s-2]=t[1].dims[s]);let i=e.pads.slice();Gr.adjustPadsBasedOnAutoPad(t[0].dims,e.strides,e.dilations,r,i,e.format==="NHWC",e.autoPad);let a=Object.assign({},e);return Object.assign(a,{kernelShape:r,pads:i}),a},Ea=e=>{let t=Ka(e),r=e.format,i=["NOTSET","VALID","SAME_UPPER","SAME_LOWER"][e.auto_pad],a=e.dilations,s=e.group,n=e.kernel_shape,u=e.pads,l=e.strides,p=e.w_is_const();return{autoPad:i,format:r,dilations:a,group:s,kernelShape:n,pads:u,strides:l,wIsConst:p,...t,cacheKey:`${e.format};${t.activation};`}},Fi=(e,t,r,i)=>{let a=r.format==="NHWC",s=Bu(t[0].dims,t[1].dims,r.dilations,r.pads,r.strides,a);if(r.group!==1){let C=[t[0]];if(a){let O=e.kernelCustomData.wT??e.compute(Me(t[1],Rr),{inputs:[1],outputs:[r.wIsConst?-2:-1]})[0];r.wIsConst&&!e.kernelCustomData.wT&&(e.kernelCustomData.wT=O),C.push(O)}else C.push(t[1]);t.length===3&&C.push(t[2]),!e.adapterInfo.isArchitecture("ampere")&&a&&t[1].dims[0]===r.group&&t[1].dims[1]===1&&r.dilations[0]===1&&r.dilations[1]===1?e.compute(Hc(C,r,s,i),{inputs:C}):e.compute(Fc(C,r,s,i),{inputs:C});return}let n=t.length===3,u=t[0].dims[a?1:2],l=t[0].dims[a?2:3],p=t[0].dims[a?3:1],h=t[1].dims[2],f=t[1].dims[3],g=s[a?1:2],_=s[a?2:3],y=s[a?3:1],$=a&&h===u&&f===l&&r.pads[0]===0&&r.pads[1]===0;if($||h===1&&f===1&&r.dilations[0]===1&&r.dilations[1]===1&&r.strides[0]===1&&r.strides[1]===1&&r.pads[0]===0&&r.pads[1]===0){let C=s[0],O,x,D,q=[];if(a){let K=e.kernelCustomData.wT??e.compute(Me(t[1],Rr),{inputs:[1],outputs:[r.wIsConst?-2:-1]})[0];if(r.wIsConst&&!e.kernelCustomData.wT&&(e.kernelCustomData.wT=K),$){let se=u*l*p;O=t[0].reshape([1,C,se]),x=K.reshape([1,se,y]),D=[1,C,y]}else O=t[0].reshape([C,u*l,p]),x=K.reshape([1,p,y]),D=[C,g*_,y];q.push(O),q.push(x)}else O=t[0].reshape([C,p,u*l]),x=t[1].reshape([1,y,p]),D=[C,y,g*_],q.push(x),q.push(O);n&&q.push(t[2]);let Q=D[2],L=q[0].dims[q[0].dims.length-1];Q<8&&L<8?e.compute(Qa(q,r,s,D,a,i),{inputs:q}):e.compute(Hr(q,r,s,D,a,i),{inputs:q});return}let S=!0,v=e.kernelCustomData.wT??e.compute(Me(t[1],Rr),{inputs:[1],outputs:[r.wIsConst?-2:-1]})[0];r.wIsConst&&!e.kernelCustomData.wT&&(e.kernelCustomData.wT=v);let w=[t[0],v];n&&w.push(t[2]);let E=a?g*_:y,I=a?y:g*_,k=h*f*p;e.compute(Lc(w,r,s,E,I,k,n,S,i),{inputs:w})},Nu=(e,t)=>{let r=t.format==="NHWC",i=[e.inputs[0].reshape(r?[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&&i.push(e.inputs[2]);let a=[0,t.pads[0],0,t.pads[1]],s=[1].concat(t.strides),n=[1].concat(t.dilations),u=[1].concat(t.kernelShape),l=Nr({...t,pads:a,strides:s,dilations:n,kernelShape:u},i);Fi(e,i,l,p=>r?[p[0],p[2],p[3]]:[p[0],p[1],p[3]])},Mu=(e,t,r)=>{let i=r.format==="NHWC"?"channelsLast":"channelsFirst",a=Nr(r,t),s=r.autoPad==="NOTSET"?r.pads:r.autoPad,n=Vc(t[0].dims,t[1].dims,r.strides,r.dilations,s,!1,i);e.compute(Gc(t,a,n.outShape,[n.filterDepth,n.filterHeight,n.filterWidth],[n.padInfo.front,n.padInfo.top,n.padInfo.left],i))},ka=(e,t)=>{if(Ru(e.inputs,t),e.inputs[0].dims.length===3)Nu(e,t);else if(e.inputs[0].dims.length===5)Mu(e,e.inputs,t);else{let r=Nr(t,e.inputs);Fi(e,e.inputs,r)}}}),jc,g0=P(()=>{"use strict";Y(),nt(),te(),re(),jc=(e,t,r)=>{let i=e.length>2,a=t.outputShape,s=t.format==="NHWC",n=t.group,u=e[1].dims,l=u[2]/n,p=u[3],h=s?$e(l):1,f=s&&p===1&&l>=4,g=f?Math.floor(l/4)*4:Math.floor(l/h)*h,_=l-g,y=s?$e(p):1,$=s?p===1?h:y:1,S=B.size(a)/y,v=[Math.ceil(S/64),1,1];ue("verbose",()=>`[conv2d_backprop_webgpu] dispatch = ${v}`);let w=["rank","rank"],E=[t.strides[0],t.strides[1]],I=[t.kernelShape[s?1:2],t.kernelShape[s?2:3]],k=[t.dilations[0],t.dilations[1]],C=[I[0]+(t.dilations[0]<=1?0:(t.kernelShape[s?1:2]-1)*(t.dilations[0]-1)),I[1]+(t.dilations[1]<=1?0:(t.kernelShape[s?2:3]-1)*(t.dilations[1]-1))],O=[C[0]-1-Math.floor((t.pads[0]+t.pads[2])/2),C[1]-1-Math.floor((t.pads[1]+t.pads[3])/2)],x=[{type:12,data:S},{type:12,data:E},{type:12,data:I},{type:12,data:k},{type:12,data:C},{type:6,data:O},{type:12,data:g},{type:12,data:l},{type:12,data:p},...Z(e[0].dims,e[1].dims)];i&&(x.push(...Z(e[2].dims)),w.push("rank")),x.push(...Z(a));let D=q=>{let Q=[{name:"output_size",type:"u32"},{name:"strides",type:"u32",length:E.length},{name:"filter_dims",type:"u32",length:I.length},{name:"dilations",type:"u32",length:I.length},{name:"effective_filter_dims",type:"u32",length:C.length},{name:"pads",type:"i32",length:O.length},{name:"input_channels_per_group_int",type:"u32"},{name:"input_channels_per_group",type:"u32"},{name:"output_channels_per_group",type:"u32"}],L=Ee(e[0].dataType),K=s?1:2,se=s?2:3,A=s?3:1,U=N("W",e[1].dataType,e[1].dims.length,$),ee=N("Dy",e[0].dataType,e[0].dims.length,h),X=[ee,U];i&&X.push(N("bias",e[2].dataType,[a[A]].length,y));let H=F("result",e[0].dataType,a.length,y),le=()=>{let J="";if(f)h===4?J+=`
+  }`};return{name:"GroupedConv-Vectorize",shaderCache:{hint:`${t.cacheKey};${s};${n};${g};${p[0]};${p[1]}`,inputDependencies:a?["rank","rank","type"]:["rank","rank"]},getRunData:()=>({outputs:[{dims:i?i(r):r,dataType:e[0].dataType}],dispatchGroup:{x:Math.ceil(u/64)},programUniforms:f}),getShaderSource:_}}}),Bu,Rr,Ru,Mr,ka,Fi,Mu,Nu,Ea,m0=P(()=>{"use strict";te(),c0(),h0(),Ya(),f0(),Dt(),Xa(),wt(),Bu=(e,t,r,i,a,s)=>{let n=e[0],u=e.slice(s?1:2,s?3:4),l=u.length,p=t[0],h=t.slice(2).map((g,_)=>g+(g-1)*(r[_]-1)),f=u.map((g,_)=>g+i[_]+i[_+l]).map((g,_)=>Math.floor((g-h[_]+a[_])/a[_]));return f.splice(0,0,n),f.splice(s?3:1,0,p),f},Rr=[2,3,1,0],Ru=(e,t)=>{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 r=e[0].dims[t.format==="NHWC"?e[0].dims.length-1:1],i=e[1].dims[1]*t.group;if(r!==i)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 a=e[0].dims.length-2;if(t.dilations.length!==a)throw new Error(`dilations should be ${a}D`);if(t.strides.length!==a)throw new Error(`strides should be ${a}D`);if(t.pads.length!==a*2)throw new Error(`pads should be ${a*2}D`);if(t.kernelShape.length!==0&&t.kernelShape.length!==e[1].dims.length-2)throw new Error("invalid kernel shape")},Mr=(e,t)=>{let r=e.kernelShape.slice();r.length<t[1].dims.length-2&&r.push(...Array(t[1].dims.length-2-r.length).fill(0));for(let s=2;s<t[1].dims.length;++s)r[s-2]===0&&(r[s-2]=t[1].dims[s]);let i=e.pads.slice();Gr.adjustPadsBasedOnAutoPad(t[0].dims,e.strides,e.dilations,r,i,e.format==="NHWC",e.autoPad);let a=Object.assign({},e);return Object.assign(a,{kernelShape:r,pads:i}),a},ka=e=>{let t=Ka(e),r=e.format,i=["NOTSET","VALID","SAME_UPPER","SAME_LOWER"][e.auto_pad],a=e.dilations,s=e.group,n=e.kernel_shape,u=e.pads,l=e.strides,p=e.w_is_const();return{autoPad:i,format:r,dilations:a,group:s,kernelShape:n,pads:u,strides:l,wIsConst:p,...t,cacheKey:`${e.format};${t.activation};`}},Fi=(e,t,r,i)=>{let a=r.format==="NHWC",s=Bu(t[0].dims,t[1].dims,r.dilations,r.pads,r.strides,a);if(r.group!==1){let C=[t[0]];if(a){let O=e.kernelCustomData.wT??e.compute(Ne(t[1],Rr),{inputs:[1],outputs:[r.wIsConst?-2:-1]})[0];r.wIsConst&&!e.kernelCustomData.wT&&(e.kernelCustomData.wT=O),C.push(O)}else C.push(t[1]);t.length===3&&C.push(t[2]),!e.adapterInfo.isArchitecture("ampere")&&a&&t[1].dims[0]===r.group&&t[1].dims[1]===1&&r.dilations[0]===1&&r.dilations[1]===1?e.compute(Hc(C,r,s,i),{inputs:C}):e.compute(Fc(C,r,s,i),{inputs:C});return}let n=t.length===3,u=t[0].dims[a?1:2],l=t[0].dims[a?2:3],p=t[0].dims[a?3:1],h=t[1].dims[2],f=t[1].dims[3],g=s[a?1:2],_=s[a?2:3],y=s[a?3:1],$=a&&h===u&&f===l&&r.pads[0]===0&&r.pads[1]===0;if($||h===1&&f===1&&r.dilations[0]===1&&r.dilations[1]===1&&r.strides[0]===1&&r.strides[1]===1&&r.pads[0]===0&&r.pads[1]===0){let C=s[0],O,x,D,L=[];if(a){let K=e.kernelCustomData.wT??e.compute(Ne(t[1],Rr),{inputs:[1],outputs:[r.wIsConst?-2:-1]})[0];if(r.wIsConst&&!e.kernelCustomData.wT&&(e.kernelCustomData.wT=K),$){let se=u*l*p;O=t[0].reshape([1,C,se]),x=K.reshape([1,se,y]),D=[1,C,y]}else O=t[0].reshape([C,u*l,p]),x=K.reshape([1,p,y]),D=[C,g*_,y];L.push(O),L.push(x)}else O=t[0].reshape([C,p,u*l]),x=t[1].reshape([1,y,p]),D=[C,y,g*_],L.push(x),L.push(O);n&&L.push(t[2]);let Q=D[2],q=L[0].dims[L[0].dims.length-1];Q<8&&q<8?e.compute(Qa(L,r,s,D,a,i),{inputs:L}):e.compute(Hr(L,r,s,D,a,i),{inputs:L});return}let S=!0,v=e.kernelCustomData.wT??e.compute(Ne(t[1],Rr),{inputs:[1],outputs:[r.wIsConst?-2:-1]})[0];r.wIsConst&&!e.kernelCustomData.wT&&(e.kernelCustomData.wT=v);let w=[t[0],v];n&&w.push(t[2]);let k=a?g*_:y,I=a?y:g*_,E=h*f*p;e.compute(qc(w,r,s,k,I,E,n,S,i),{inputs:w})},Mu=(e,t)=>{let r=t.format==="NHWC",i=[e.inputs[0].reshape(r?[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&&i.push(e.inputs[2]);let a=[0,t.pads[0],0,t.pads[1]],s=[1].concat(t.strides),n=[1].concat(t.dilations),u=[1].concat(t.kernelShape),l=Mr({...t,pads:a,strides:s,dilations:n,kernelShape:u},i);Fi(e,i,l,p=>r?[p[0],p[2],p[3]]:[p[0],p[1],p[3]])},Nu=(e,t,r)=>{let i=r.format==="NHWC"?"channelsLast":"channelsFirst",a=Mr(r,t),s=r.autoPad==="NOTSET"?r.pads:r.autoPad,n=Vc(t[0].dims,t[1].dims,r.strides,r.dilations,s,!1,i);e.compute(Gc(t,a,n.outShape,[n.filterDepth,n.filterHeight,n.filterWidth],[n.padInfo.front,n.padInfo.top,n.padInfo.left],i))},Ea=(e,t)=>{if(Ru(e.inputs,t),e.inputs[0].dims.length===3)Mu(e,t);else if(e.inputs[0].dims.length===5)Nu(e,e.inputs,t);else{let r=Mr(t,e.inputs);Fi(e,e.inputs,r)}}}),jc,g0=P(()=>{"use strict";Y(),nt(),te(),re(),jc=(e,t,r)=>{let i=e.length>2,a=t.outputShape,s=t.format==="NHWC",n=t.group,u=e[1].dims,l=u[2]/n,p=u[3],h=s?$e(l):1,f=s&&p===1&&l>=4,g=f?Math.floor(l/4)*4:Math.floor(l/h)*h,_=l-g,y=s?$e(p):1,$=s?p===1?h:y:1,S=B.size(a)/y,v=[Math.ceil(S/64),1,1];ue("verbose",()=>`[conv2d_backprop_webgpu] dispatch = ${v}`);let w=["rank","rank"],k=[t.strides[0],t.strides[1]],I=[t.kernelShape[s?1:2],t.kernelShape[s?2:3]],E=[t.dilations[0],t.dilations[1]],C=[I[0]+(t.dilations[0]<=1?0:(t.kernelShape[s?1:2]-1)*(t.dilations[0]-1)),I[1]+(t.dilations[1]<=1?0:(t.kernelShape[s?2:3]-1)*(t.dilations[1]-1))],O=[C[0]-1-Math.floor((t.pads[0]+t.pads[2])/2),C[1]-1-Math.floor((t.pads[1]+t.pads[3])/2)],x=[{type:12,data:S},{type:12,data:k},{type:12,data:I},{type:12,data:E},{type:12,data:C},{type:6,data:O},{type:12,data:g},{type:12,data:l},{type:12,data:p},...Z(e[0].dims,e[1].dims)];i&&(x.push(...Z(e[2].dims)),w.push("rank")),x.push(...Z(a));let D=L=>{let Q=[{name:"output_size",type:"u32"},{name:"strides",type:"u32",length:k.length},{name:"filter_dims",type:"u32",length:I.length},{name:"dilations",type:"u32",length:I.length},{name:"effective_filter_dims",type:"u32",length:C.length},{name:"pads",type:"i32",length:O.length},{name:"input_channels_per_group_int",type:"u32"},{name:"input_channels_per_group",type:"u32"},{name:"output_channels_per_group",type:"u32"}],q=ke(e[0].dataType),K=s?1:2,se=s?2:3,A=s?3:1,U=M("W",e[1].dataType,e[1].dims.length,$),ee=M("Dy",e[0].dataType,e[0].dims.length,h),X=[ee,U];i&&X.push(M("bias",e[2].dataType,[a[A]].length,y));let H=F("result",e[0].dataType,a.length,y),le=()=>{let J="";if(f)h===4?J+=`
         let xValue = ${ee.getByOffset("x_offset")};
         let wValue = ${U.getByOffset("w_offset")};
         dotProd = dotProd + dot(xValue, wValue);
         x_offset += 1u;
         w_offset += 1u;`:h===2?J+=`
-          dotProd = dotProd + dot(vec4<${L}>(${ee.getByOffset("x_offset")}, ${ee.getByOffset("x_offset + 1u")}), vec4<${L}>(${U.getByOffset("w_offset")}, ${U.getByOffset("w_offset + 1u")}));
+          dotProd = dotProd + dot(vec4<${q}>(${ee.getByOffset("x_offset")}, ${ee.getByOffset("x_offset + 1u")}), vec4<${q}>(${U.getByOffset("w_offset")}, ${U.getByOffset("w_offset + 1u")}));
           x_offset += 2u;
           w_offset += 2u;`:h===1&&(J+=`
-          dotProd = dotProd + dot(vec4<${L}>(${ee.getByOffset("x_offset")}, ${ee.getByOffset("x_offset + 1u")}, ${ee.getByOffset("x_offset + 2u")}, ${ee.getByOffset("x_offset + 3u")}), vec4<${L}>(${U.getByOffset("w_offset")}, ${U.getByOffset("w_offset + 1u")}, ${U.getByOffset("w_offset + 2u")}, ${U.getByOffset("w_offset + 3u")}));
+          dotProd = dotProd + dot(vec4<${q}>(${ee.getByOffset("x_offset")}, ${ee.getByOffset("x_offset + 1u")}, ${ee.getByOffset("x_offset + 2u")}, ${ee.getByOffset("x_offset + 3u")}), vec4<${q}>(${U.getByOffset("w_offset")}, ${U.getByOffset("w_offset + 1u")}, ${U.getByOffset("w_offset + 2u")}, ${U.getByOffset("w_offset + 3u")}));
           x_offset += 4u;
           w_offset += 4u;`);else if(J+=`
                   let xValue = ${s?ee.getByOffset(`${ee.indicesToOffset(`${ee.type.indices}(batch, idyR, idyC, inputChannel)`)} / ${h}`):ee.get("batch","inputChannel","idyR","idyC")};
@@ -1220,7 +1220,7 @@ fn main(@builtin(local_invocation_id) localId : vec3<u32>,
           let wValue = ${U.getByOffset(`w_offset / ${$}`)};
           dotProd = dotProd + xValue * wValue;`;else for(let de=0;de<h;de++)J+=`
             let wValue${de} = ${U.getByOffset(`${U.indicesToOffset(`${U.type.indices}(u32(wRPerm), u32(wCPerm), inputChannel + ${de}, wOutChannel)`)} / ${$}`)};
-            dotProd = dotProd + xValue[${de}] * wValue${de};`;return J},M=()=>{if(_===0)return"";if(!f)throw new Error(`packInputAs4 ${f} is not true.`);let J="";if(h===1){J+="dotProd = dotProd";for(let de=0;de<_;de++)J+=`
+            dotProd = dotProd + xValue[${de}] * wValue${de};`;return J},N=()=>{if(_===0)return"";if(!f)throw new Error(`packInputAs4 ${f} is not true.`);let J="";if(h===1){J+="dotProd = dotProd";for(let de=0;de<_;de++)J+=`
             + ${ee.getByOffset(`x_offset + ${de}`)} * ${U.getByOffset(`w_offset + ${de}`)}`;J+=";"}else if(h===2){if(_!==2)throw new Error(`Invalid inputChannelsRemainder ${_}.`);J+=`
           let xValue = ${ee.getByOffset("x_offset")};
           let wValue = ${U.getByOffset("w_offset")};
@@ -1247,9 +1247,9 @@ fn main(@builtin(local_invocation_id) localId : vec3<u32>,
               if (wR % uniforms.dilations.x != 0) {
                 continue;
               }
-              let dyR = (${L}(dyRCorner) + ${L}(wR)) / ${L}(uniforms.strides[0]);
+              let dyR = (${q}(dyRCorner) + ${q}(wR)) / ${q}(uniforms.strides[0]);
               let wRPerm = uniforms.filter_dims.x - 1 - wR / uniforms.dilations.x;
-              if (dyR < 0.0 || dyR >= ${L}(uniforms.Dy_shape[${K}]) || fract(dyR) > 0.0 ||
+              if (dyR < 0.0 || dyR >= ${q}(uniforms.Dy_shape[${K}]) || fract(dyR) > 0.0 ||
                   wRPerm < 0) {
                 continue;
               }
@@ -1263,9 +1263,9 @@ fn main(@builtin(local_invocation_id) localId : vec3<u32>,
                 if (wC % uniforms.dilations.y != 0) {
                   continue;
                 }
-                let dyC = (${L}(dyCCorner) + ${L}(wC)) / ${L}(uniforms.strides.y);
+                let dyC = (${q}(dyCCorner) + ${q}(wC)) / ${q}(uniforms.strides.y);
                 let wCPerm = uniforms.filter_dims.y - 1 - wC / uniforms.dilations.y;
-                if (dyC < 0.0 || dyC >= ${L}(uniforms.Dy_shape[${se}]) ||
+                if (dyC < 0.0 || dyC >= ${q}(uniforms.Dy_shape[${se}]) ||
                     fract(dyC) > 0.0 || wCPerm < 0) {
                   continue;
                 }
@@ -1279,7 +1279,7 @@ fn main(@builtin(local_invocation_id) localId : vec3<u32>,
                   ${le()}
                   inputChannel = inputChannel + ${f?4:h};
                 }
-                ${M()}
+                ${N()}
                 wC = wC + uniforms.strides.y - 1;
               }
               wR = wR + uniforms.strides[0] - 1;
@@ -1287,10 +1287,10 @@ fn main(@builtin(local_invocation_id) localId : vec3<u32>,
             let value = dotProd${i?` + bias[d1 / ${y}]`:""};
             ${H.setByOffset("global_idx","value")};
           `;return`
-    ${q.registerUniforms(Q).declareVariables(...X,H)}
-      ${q.mainStart()}
-      ${q.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.output_size")};
-    ${V}}`};return{name:"ConvTranspose2D",shaderCache:{hint:`${t.cacheKey};${h}${$}${y}${f}${_}`,inputDependencies:w},getRunData:()=>({dispatchGroup:{x:v[0],y:v[1],z:v[2]},outputs:[{dims:r?r(a):a,dataType:e[0].dataType}],programUniforms:x}),getShaderSource:D}}}),Du,Pu,Uu,Hi,Kc,Wu,ji,qu,Zc,y0=P(()=>{"use strict";g0(),Dt(),wt(),Du=(e,t,r,i,a,s)=>(e-1)*t+r+(i-1)*a+1-s,Pu=(e,t,r,i,a)=>{let s=Math.floor(e/2);t==="SAME_UPPER"?(r[i]=s,r[a]=e-s):t==="SAME_LOWER"&&(r[i]=e-s,r[a]=s)},Uu=(e,t,r,i,a,s,n,u,l,p)=>{let h=e.length-2,f=p.length===0;l.length<h&&l.push(...Array(h-l.length).fill(0));let g=e[0],_=t[u?3:1]*a;for(let y=0,$=e.length-h-(u?1:0);y<h;++y,++$){let S=e[$],v=f?S*n[y]:p[y],w=Du(S,n[y],s[y],t[$],r[y],v);Pu(w,i,s,y,y+h),f&&p.push(n[y]*(S-1)+l[y]+(t[$]-1)*r[y]+1-s[y]-s[y+h])}p.splice(0,0,g),p.splice(u?3:1,0,_)},Hi=(e,t)=>{let r=e.kernelShape.slice();if(e.kernelShape.length===0||e.kernelShape.reduce((f,g)=>f*g,1)===0){r.length=0;for(let f=2;f<t[1].dims.length;++f)r.push(t[1].dims[f])}let i=e.format==="NHWC";r.splice(0,0,t[1].dims[0]),r.splice(i?3:1,0,t[1].dims[1]);let a=e.pads.slice(),s=e.outputShape.slice(),n=e.outputPadding.slice(),u=t[0].dims,l=e.dilations.slice();if(l.reduce((f,g)=>f+g,0)===0){let f=t[0].dims.length-2;l=new Array(f).fill(1)}let p=e.strides.slice();if(p.reduce((f,g)=>f+g,0)===0){let f=t[0].dims.length-2;p=new Array(f).fill(1)}Uu(u,r,l,e.autoPad,e.group,a,p,i,n,s);let h=Object.assign({},e);return Object.assign(h,{kernelShape:r,pads:a,outputPadding:n,outputShape:s,dilations:l,strides:p}),h},Kc=e=>{let t=Ka(e),r=e.format,i=["NOTSET","VALID","SAME_UPPER","SAME_LOWER"][typeof e.autoPad>"u"?0:e.autoPad],a=e.dilations,s=e.group,n=e.kernelShape,u=e.pads,l=e.strides,p=e.wIsConst(),h=e.outputPadding,f=e.outputShape;return{autoPad:i,format:r,dilations:a,group:s,kernelShape:n,outputPadding:h,outputShape:f,pads:u,strides:l,wIsConst:p,...t,cacheKey:`${e.format};${t.activation};`}},Wu=(e,t)=>{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 r=e[0].dims[t.format==="NHWC"?e[0].dims.length-1:1],i=e[1].dims[0];if(r!==i)throw new Error("FILTER_IN_CHANNEL should be equal to DATA_CHANNEL");let a=e[1].dims[1]*t.group;if(e.length===3&&(e[2].dims.length!==1||e[2].dims[0]!==a))throw new Error("invalid bias");let s=e[0].dims.length-2;if(t.dilations.reduce((n,u)=>n+u,0)>0&&t.dilations.length!==s)throw new Error(`dilations should be ${s}D`);if(t.strides.reduce((n,u)=>n+u,0)>0&&t.strides.length!==s)throw new Error(`strides should be ${s}D`);if(t.pads.reduce((n,u)=>n+u,0)>0&&t.pads.length!==s*2)throw new Error(`pads should be ${s*2}D`);if(t.outputPadding.length!==s&&t.outputPadding.length!==0)throw new Error(`output_padding should be ${s}D`);if(t.kernelShape.reduce((n,u)=>n+u,0)>0&&t.kernelShape.length!==0&&t.kernelShape.length!==e[1].dims.length-2)throw new Error("invalid kernel shape");if(t.outputShape.length!==0&&t.outputShape.length!==e[0].dims.length-2)throw new Error("invalid output shape")},ji=(e,t,r,i)=>{let a=e.kernelCustomData.wT??e.compute(Me(t[1],[2,3,0,1]),{inputs:[1],outputs:[r.wIsConst?-2:-1]})[0];r.wIsConst&&!e.kernelCustomData.wT&&(e.kernelCustomData.wT=a);let s=[t[0],a];t.length===3&&s.push(t[2]),e.compute(jc(s,r,i),{inputs:s})},qu=(e,t)=>{let r=t.format==="NHWC",i=[e.inputs[0].reshape(r?[e.inputs[0].dims[0],1,e.inputs[0].dims[1],e.inputs[0].dims[2]]:[e.inputs[0].dims[0],e.inputs[0].dims[1],1,e.inputs[0].dims[2]]),e.inputs[1].reshape([e.inputs[1].dims[0],e.inputs[1].dims[1],1,e.inputs[1].dims[2]])];e.inputs.length===3&&i.push(e.inputs[2]);let a=t.kernelShape;(a.length===0||a[0]===0)&&(a=[e.inputs[1].dims[2]]);let s=t.dilations;(s.length===0||s[0]===0)&&(s=[1]);let n=t.strides;(n.length===0||n[0]===0)&&(n=[1]);let u=t.pads;u.length===0&&(u=[0,0]),u=[0,u[0],0,u[1]],n=[1].concat(n),s=[1].concat(s),a=[1].concat(a);let l=t.outputPadding;l=[0].concat(l);let p=Hi({...t,pads:u,strides:n,dilations:s,kernelShape:a,outputPadding:l},i);ji(e,i,p,h=>r?[h[0],h[2],h[3]]:[h[0],h[1],h[3]])},Zc=(e,t)=>{if(Wu(e.inputs,t),e.inputs[0].dims.length===3)qu(e,t);else{let r=Hi(t,e.inputs);ji(e,e.inputs,r)}}}),Lu,Qc,Xc,_0=P(()=>{"use strict";Y(),te(),ve(),re(),Lu=(e,t,r,i)=>{let a=B.size(t),s=t.length,n=N("input",e,s),u=F("output",e,s),l=r.dataType===6?r.getInt32Array()[0]:Number(r.getBigInt64Array()[0]),p=B.normalizeAxis(l,s),h=f=>{let g=` i32(${n.indicesGet("inputIndices","uniforms.axis")}) `,_=j("uniforms.input_shape","uniforms.axis",s),y=i.reverse?g+(i.exclusive?" + 1":""):"0",$=i.reverse?_:g+(i.exclusive?"":" + 1");return`
+    ${L.registerUniforms(Q).declareVariables(...X,H)}
+      ${L.mainStart()}
+      ${L.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.output_size")};
+    ${V}}`};return{name:"ConvTranspose2D",shaderCache:{hint:`${t.cacheKey};${h}${$}${y}${f}${_}`,inputDependencies:w},getRunData:()=>({dispatchGroup:{x:v[0],y:v[1],z:v[2]},outputs:[{dims:r?r(a):a,dataType:e[0].dataType}],programUniforms:x}),getShaderSource:D}}}),Du,Pu,Uu,Hi,Kc,Wu,ji,Lu,Zc,y0=P(()=>{"use strict";g0(),Dt(),wt(),Du=(e,t,r,i,a,s)=>(e-1)*t+r+(i-1)*a+1-s,Pu=(e,t,r,i,a)=>{let s=Math.floor(e/2);t==="SAME_UPPER"?(r[i]=s,r[a]=e-s):t==="SAME_LOWER"&&(r[i]=e-s,r[a]=s)},Uu=(e,t,r,i,a,s,n,u,l,p)=>{let h=e.length-2,f=p.length===0;l.length<h&&l.push(...Array(h-l.length).fill(0));let g=e[0],_=t[u?3:1]*a;for(let y=0,$=e.length-h-(u?1:0);y<h;++y,++$){let S=e[$],v=f?S*n[y]:p[y],w=Du(S,n[y],s[y],t[$],r[y],v);Pu(w,i,s,y,y+h),f&&p.push(n[y]*(S-1)+l[y]+(t[$]-1)*r[y]+1-s[y]-s[y+h])}p.splice(0,0,g),p.splice(u?3:1,0,_)},Hi=(e,t)=>{let r=e.kernelShape.slice();if(e.kernelShape.length===0||e.kernelShape.reduce((f,g)=>f*g,1)===0){r.length=0;for(let f=2;f<t[1].dims.length;++f)r.push(t[1].dims[f])}let i=e.format==="NHWC";r.splice(0,0,t[1].dims[0]),r.splice(i?3:1,0,t[1].dims[1]);let a=e.pads.slice(),s=e.outputShape.slice(),n=e.outputPadding.slice(),u=t[0].dims,l=e.dilations.slice();if(l.reduce((f,g)=>f+g,0)===0){let f=t[0].dims.length-2;l=new Array(f).fill(1)}let p=e.strides.slice();if(p.reduce((f,g)=>f+g,0)===0){let f=t[0].dims.length-2;p=new Array(f).fill(1)}Uu(u,r,l,e.autoPad,e.group,a,p,i,n,s);let h=Object.assign({},e);return Object.assign(h,{kernelShape:r,pads:a,outputPadding:n,outputShape:s,dilations:l,strides:p}),h},Kc=e=>{let t=Ka(e),r=e.format,i=["NOTSET","VALID","SAME_UPPER","SAME_LOWER"][typeof e.autoPad>"u"?0:e.autoPad],a=e.dilations,s=e.group,n=e.kernelShape,u=e.pads,l=e.strides,p=e.wIsConst(),h=e.outputPadding,f=e.outputShape;return{autoPad:i,format:r,dilations:a,group:s,kernelShape:n,outputPadding:h,outputShape:f,pads:u,strides:l,wIsConst:p,...t,cacheKey:`${e.format};${t.activation};`}},Wu=(e,t)=>{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 r=e[0].dims[t.format==="NHWC"?e[0].dims.length-1:1],i=e[1].dims[0];if(r!==i)throw new Error("FILTER_IN_CHANNEL should be equal to DATA_CHANNEL");let a=e[1].dims[1]*t.group;if(e.length===3&&(e[2].dims.length!==1||e[2].dims[0]!==a))throw new Error("invalid bias");let s=e[0].dims.length-2;if(t.dilations.reduce((n,u)=>n+u,0)>0&&t.dilations.length!==s)throw new Error(`dilations should be ${s}D`);if(t.strides.reduce((n,u)=>n+u,0)>0&&t.strides.length!==s)throw new Error(`strides should be ${s}D`);if(t.pads.reduce((n,u)=>n+u,0)>0&&t.pads.length!==s*2)throw new Error(`pads should be ${s*2}D`);if(t.outputPadding.length!==s&&t.outputPadding.length!==0)throw new Error(`output_padding should be ${s}D`);if(t.kernelShape.reduce((n,u)=>n+u,0)>0&&t.kernelShape.length!==0&&t.kernelShape.length!==e[1].dims.length-2)throw new Error("invalid kernel shape");if(t.outputShape.length!==0&&t.outputShape.length!==e[0].dims.length-2)throw new Error("invalid output shape")},ji=(e,t,r,i)=>{let a=e.kernelCustomData.wT??e.compute(Ne(t[1],[2,3,0,1]),{inputs:[1],outputs:[r.wIsConst?-2:-1]})[0];r.wIsConst&&!e.kernelCustomData.wT&&(e.kernelCustomData.wT=a);let s=[t[0],a];t.length===3&&s.push(t[2]),e.compute(jc(s,r,i),{inputs:s})},Lu=(e,t)=>{let r=t.format==="NHWC",i=[e.inputs[0].reshape(r?[e.inputs[0].dims[0],1,e.inputs[0].dims[1],e.inputs[0].dims[2]]:[e.inputs[0].dims[0],e.inputs[0].dims[1],1,e.inputs[0].dims[2]]),e.inputs[1].reshape([e.inputs[1].dims[0],e.inputs[1].dims[1],1,e.inputs[1].dims[2]])];e.inputs.length===3&&i.push(e.inputs[2]);let a=t.kernelShape;(a.length===0||a[0]===0)&&(a=[e.inputs[1].dims[2]]);let s=t.dilations;(s.length===0||s[0]===0)&&(s=[1]);let n=t.strides;(n.length===0||n[0]===0)&&(n=[1]);let u=t.pads;u.length===0&&(u=[0,0]),u=[0,u[0],0,u[1]],n=[1].concat(n),s=[1].concat(s),a=[1].concat(a);let l=t.outputPadding;l=[0].concat(l);let p=Hi({...t,pads:u,strides:n,dilations:s,kernelShape:a,outputPadding:l},i);ji(e,i,p,h=>r?[h[0],h[2],h[3]]:[h[0],h[1],h[3]])},Zc=(e,t)=>{if(Wu(e.inputs,t),e.inputs[0].dims.length===3)Lu(e,t);else{let r=Hi(t,e.inputs);ji(e,e.inputs,r)}}}),qu,Qc,Xc,_0=P(()=>{"use strict";Y(),te(),ve(),re(),qu=(e,t,r,i)=>{let a=B.size(t),s=t.length,n=M("input",e,s),u=F("output",e,s),l=r.dataType===6?r.getInt32Array()[0]:Number(r.getBigInt64Array()[0]),p=B.normalizeAxis(l,s),h=f=>{let g=` i32(${n.indicesGet("inputIndices","uniforms.axis")}) `,_=j("uniforms.input_shape","uniforms.axis",s),y=i.reverse?g+(i.exclusive?" + 1":""):"0",$=i.reverse?_:g+(i.exclusive?"":" + 1");return`
                 ${f.registerUniform("outputSize","u32").registerUniform("axis","u32").declareVariables(n,u)}
                 ${f.mainStart()}
                   ${f.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.outputSize")}
@@ -1303,9 +1303,9 @@ fn main(@builtin(local_invocation_id) localId : vec3<u32>,
                     sum = sum + ${n.getByIndices("inputIndices")};
                   }
                   ${u.setByOffset("global_idx","sum")};
-                }`};return{name:"CumSum",shaderCache:{hint:i.cacheKey,inputDependencies:["rank"]},getRunData:()=>({outputs:[{dims:t,dataType:e}],dispatchGroup:{x:Math.ceil(a/64)},programUniforms:[{type:12,data:a},{type:12,data:p},...Z(t,t)]}),getShaderSource:h}},Qc=(e,t)=>{let r=e.inputs[0].dims,i=e.inputs[0].dataType,a=e.inputs[1];e.compute(Lu(i,r,a,t),{inputs:[0]})},Xc=e=>{let t=e.exclusive===1,r=e.reverse===1;return he({exclusive:t,reverse:r})}}),Vu,Gu,Fu,Yc,Jc,w0=P(()=>{"use strict";Y(),te(),ve(),re(),Vu=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.")},Gu=(e,t,r,i)=>{let a=[];a.push(`fn perm(i: ${i.type.indices}) -> ${r.type.indices} {
+                }`};return{name:"CumSum",shaderCache:{hint:i.cacheKey,inputDependencies:["rank"]},getRunData:()=>({outputs:[{dims:t,dataType:e}],dispatchGroup:{x:Math.ceil(a/64)},programUniforms:[{type:12,data:a},{type:12,data:p},...Z(t,t)]}),getShaderSource:h}},Qc=(e,t)=>{let r=e.inputs[0].dims,i=e.inputs[0].dataType,a=e.inputs[1];e.compute(qu(i,r,a,t),{inputs:[0]})},Xc=e=>{let t=e.exclusive===1,r=e.reverse===1;return he({exclusive:t,reverse:r})}}),Vu,Gu,Fu,Yc,Jc,w0=P(()=>{"use strict";Y(),te(),ve(),re(),Vu=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.")},Gu=(e,t,r,i)=>{let a=[];a.push(`fn perm(i: ${i.type.indices}) -> ${r.type.indices} {
     var a: ${r.type.indices};`);for(let s=0;s<t;++s)a.push(r.indicesSet("a",e[s],`i[${s}]`));return a.push("return a;}"),a.join(`
-`)},Fu=(e,t)=>{let r,i,a,s,n,u,l=t.format==="NHWC",p=t.blocksize,h=t.mode==="DCR";l?([r,i,a,s]=e.dims,n=h?[r,i,a,p,p,s/p**2]:[r,i,a,s/p**2,p,p],u=h?[0,1,3,2,4,5]:[0,1,4,2,5,3]):([r,i,a,s]=[e.dims[0],e.dims[2],e.dims[3],e.dims[1]],n=h?[r,p,p,s/p**2,i,a]:[r,s/p**2,p,p,i,a],u=h?[0,3,4,1,5,2]:[0,1,4,2,5,3]);let f=e.reshape(n),g=f.dims.length,_=e.dataType,y=N("a",_,g),$=F("output",_,g),S=v=>`
+`)},Fu=(e,t)=>{let r,i,a,s,n,u,l=t.format==="NHWC",p=t.blocksize,h=t.mode==="DCR";l?([r,i,a,s]=e.dims,n=h?[r,i,a,p,p,s/p**2]:[r,i,a,s/p**2,p,p],u=h?[0,1,3,2,4,5]:[0,1,4,2,5,3]):([r,i,a,s]=[e.dims[0],e.dims[2],e.dims[3],e.dims[1]],n=h?[r,p,p,s/p**2,i,a]:[r,s/p**2,p,p,i,a],u=h?[0,3,4,1,5,2]:[0,1,4,2,5,3]);let f=e.reshape(n),g=f.dims.length,_=e.dataType,y=M("a",_,g),$=F("output",_,g),S=v=>`
   ${v.registerUniform("output_size","u32").declareVariables(y,$)}
 
   ${Gu(u,g,y,$)}
@@ -1317,18 +1317,18 @@ fn main(@builtin(local_invocation_id) localId : vec3<u32>,
     let aIndices = perm(indices);
 
     ${$.setByOffset("global_idx",y.getByIndices("aIndices"))}
-  }`;return{name:"DepthToSpace",shaderCache:{hint:`${e.dims};${t.blocksize};${t.mode}`,inputDependencies:["rank"]},getRunData:v=>{let w=l?[r,i*p,a*p,s/p**2]:[r,s/p**2,i*p,a*p],E=B.size(w),I=f.dims,k=B.sortBasedOnPerm(I,u);return{outputs:[{dims:w,dataType:v[0].dataType}],dispatchGroup:{x:Math.ceil(E/64)},programUniforms:[{type:12,data:E},...Z(I,k)]}},getShaderSource:S}},Yc=(e,t)=>{Vu(e.inputs),e.compute(Fu(e.inputs[0],t))},Jc=e=>he({blocksize:e.blocksize,mode:e.mode,format:e.format})}),Mr,tr,Ki,Hu,ju,Ku,Zu,Zi,Qu,eh,th,b0=P(()=>{"use strict";Y(),te(),ve(),re(),Mr="[a-zA-Z]|\\.\\.\\.",tr="("+Mr+")+",Ki="^"+tr+"$",Hu="("+tr+",)*"+tr,ju="^"+Hu+"$",Ku=class{constructor(e=-1){this.symbolToIndices=new Map,this.inputIndex=e}addSymbol(e,t){let r=this.symbolToIndices.get(e);r===void 0?r=[t]:r.push(t),this.symbolToIndices.set(e,r)}},Zu=class{constructor(e,t){this.equation=t,this.hasEllipsis=!1,this.symbolToInfo=new Map,this.lhs=new Array,this.outputDims=[];let[r,i]=t.includes("->")?t.split("->",2):[t,""];if(!r.match(RegExp(ju)))throw new Error("Invalid LHS term");if(r.split(",").forEach((a,s)=>{let n=e[s].dims.slice();if(!a.match(RegExp(Ki)))throw new Error("Invalid LHS term");let u=this.processTerm(a,!0,n,s);this.lhs.push(u)}),i==="")i+=[...this.symbolToInfo.entries()].filter(([a,s])=>s.count===1||a==="...").map(([a])=>a).join("");else if(!i.match(RegExp(tr)))throw new Error("Invalid RHS");i.match(RegExp(Mr,"g"))?.forEach(a=>{if(a==="...")this.outputDims=this.outputDims.concat(this.ellipsisDims);else{let s=this.symbolToInfo.get(a);if(s===void 0)throw new Error("Invalid RHS symbol");this.outputDims.push(s.dimValue)}}),this.rhs=this.processTerm(i,!1,this.outputDims)}addSymbol(e,t,r){let i=this.symbolToInfo.get(e);if(i!==void 0){if(i.dimValue!==t&&i.count!==1)throw new Error("Dimension mismatch");i.count++,i.inputIndices.push(r)}else i={count:1,dimValue:t,inputIndices:[r]};this.symbolToInfo.set(e,i)}processTerm(e,t,r,i=-1){let a=r.length,s=!1,n=[],u=0;if(!e.match(RegExp(Ki))&&!t&&e!=="")throw new Error("Invalid LHS term");let l=e.match(RegExp(Mr,"g")),p=new Ku(i);return l?.forEach((h,f)=>{if(h==="..."){if(s)throw new Error("Only one ellipsis is allowed per input term");s=!0;let g=a-l.length+1;if(g<0)throw new Error("Ellipsis out of bounds");if(n=r.slice(u,u+g),this.hasEllipsis){if(this.ellipsisDims.length!==n.length||this.ellipsisDims.toString()!==n.toString())throw new Error("Ellipsis dimensions mismatch")}else if(t)this.hasEllipsis=!0,this.ellipsisDims=n;else throw new Error("Ellipsis must be specified in the LHS");for(let _=0;_<n.length;_++){let y=String.fromCharCode(48+_);p.addSymbol(y,f+_),this.addSymbol(y,r[u++],i)}}else p.addSymbol(h,f+(this.hasEllipsis?this.ellipsisDims.length-1:0)),this.addSymbol(h,r[u++],i)}),p}},Zi=e=>e+"_max",Qu=(e,t,r,i)=>{let a=e.map(p=>p.length).map((p,h)=>N(`input${h}`,t,p)),s=B.size(i),n=F("output",t,i.length),u=[...r.symbolToInfo.keys()].filter(p=>!r.rhs.symbolToIndices.has(p)),l=p=>{let h=[],f="var prod = 1.0;",g="var sum = 0.0;",_="sum += prod;",y=[],$=[],S=[],v=[],w=r.symbolToInfo.size===r.rhs.symbolToIndices.size;r.symbolToInfo.forEach((I,k)=>{if(r.rhs.symbolToIndices.has(k)){let C=r.rhs.symbolToIndices.get(k)?.[0];C!==void 0&&r.lhs.forEach((O,x)=>{if(I.inputIndices.includes(x)){let D=O.symbolToIndices.get(k);if(D===void 0)throw new Error("Invalid symbol error");D.forEach(q=>{h.push(`${a[x].indicesSet(`input${x}Indices`,q,n.indicesGet("outputIndices",C))}`)})}})}else r.lhs.forEach((C,O)=>{if(I.inputIndices.includes(O)){let x=C.symbolToIndices.get(k);if(x===void 0)throw new Error("Invalid symbol error");x.forEach(D=>{y.push(`${a[O].indicesSet(`input${O}Indices`,D,`${k}`)}`)}),v.push(`prod *= ${a[O].getByIndices(`input${O}Indices`)};`)}}),$.push(`for(var ${k}: u32 = 0; ${k} < uniforms.${Zi(k)}; ${k}++) {`),S.push("}")});let E=w?[...h,`let sum = ${a.map((I,k)=>I.getByIndices(`input${k}Indices`)).join(" * ")};`]:[...h,g,...$,...y,f,...v,_,...S];return`
+  }`;return{name:"DepthToSpace",shaderCache:{hint:`${e.dims};${t.blocksize};${t.mode}`,inputDependencies:["rank"]},getRunData:v=>{let w=l?[r,i*p,a*p,s/p**2]:[r,s/p**2,i*p,a*p],k=B.size(w),I=f.dims,E=B.sortBasedOnPerm(I,u);return{outputs:[{dims:w,dataType:v[0].dataType}],dispatchGroup:{x:Math.ceil(k/64)},programUniforms:[{type:12,data:k},...Z(I,E)]}},getShaderSource:S}},Yc=(e,t)=>{Vu(e.inputs),e.compute(Fu(e.inputs[0],t))},Jc=e=>he({blocksize:e.blocksize,mode:e.mode,format:e.format})}),Nr,tr,Ki,Hu,ju,Ku,Zu,Zi,Qu,eh,th,b0=P(()=>{"use strict";Y(),te(),ve(),re(),Nr="[a-zA-Z]|\\.\\.\\.",tr="("+Nr+")+",Ki="^"+tr+"$",Hu="("+tr+",)*"+tr,ju="^"+Hu+"$",Ku=class{constructor(e=-1){this.symbolToIndices=new Map,this.inputIndex=e}addSymbol(e,t){let r=this.symbolToIndices.get(e);r===void 0?r=[t]:r.push(t),this.symbolToIndices.set(e,r)}},Zu=class{constructor(e,t){this.equation=t,this.hasEllipsis=!1,this.symbolToInfo=new Map,this.lhs=new Array,this.outputDims=[];let[r,i]=t.includes("->")?t.split("->",2):[t,""];if(!r.match(RegExp(ju)))throw new Error("Invalid LHS term");if(r.split(",").forEach((a,s)=>{let n=e[s].dims.slice();if(!a.match(RegExp(Ki)))throw new Error("Invalid LHS term");let u=this.processTerm(a,!0,n,s);this.lhs.push(u)}),i==="")i+=[...this.symbolToInfo.entries()].filter(([a,s])=>s.count===1||a==="...").map(([a])=>a).join("");else if(!i.match(RegExp(tr)))throw new Error("Invalid RHS");i.match(RegExp(Nr,"g"))?.forEach(a=>{if(a==="...")this.outputDims=this.outputDims.concat(this.ellipsisDims);else{let s=this.symbolToInfo.get(a);if(s===void 0)throw new Error("Invalid RHS symbol");this.outputDims.push(s.dimValue)}}),this.rhs=this.processTerm(i,!1,this.outputDims)}addSymbol(e,t,r){let i=this.symbolToInfo.get(e);if(i!==void 0){if(i.dimValue!==t&&i.count!==1)throw new Error("Dimension mismatch");i.count++,i.inputIndices.push(r)}else i={count:1,dimValue:t,inputIndices:[r]};this.symbolToInfo.set(e,i)}processTerm(e,t,r,i=-1){let a=r.length,s=!1,n=[],u=0;if(!e.match(RegExp(Ki))&&!t&&e!=="")throw new Error("Invalid LHS term");let l=e.match(RegExp(Nr,"g")),p=new Ku(i);return l?.forEach((h,f)=>{if(h==="..."){if(s)throw new Error("Only one ellipsis is allowed per input term");s=!0;let g=a-l.length+1;if(g<0)throw new Error("Ellipsis out of bounds");if(n=r.slice(u,u+g),this.hasEllipsis){if(this.ellipsisDims.length!==n.length||this.ellipsisDims.toString()!==n.toString())throw new Error("Ellipsis dimensions mismatch")}else if(t)this.hasEllipsis=!0,this.ellipsisDims=n;else throw new Error("Ellipsis must be specified in the LHS");for(let _=0;_<n.length;_++){let y=String.fromCharCode(48+_);p.addSymbol(y,f+_),this.addSymbol(y,r[u++],i)}}else p.addSymbol(h,f+(this.hasEllipsis?this.ellipsisDims.length-1:0)),this.addSymbol(h,r[u++],i)}),p}},Zi=e=>e+"_max",Qu=(e,t,r,i)=>{let a=e.map(p=>p.length).map((p,h)=>M(`input${h}`,t,p)),s=B.size(i),n=F("output",t,i.length),u=[...r.symbolToInfo.keys()].filter(p=>!r.rhs.symbolToIndices.has(p)),l=p=>{let h=[],f="var prod = 1.0;",g="var sum = 0.0;",_="sum += prod;",y=[],$=[],S=[],v=[],w=r.symbolToInfo.size===r.rhs.symbolToIndices.size;r.symbolToInfo.forEach((I,E)=>{if(r.rhs.symbolToIndices.has(E)){let C=r.rhs.symbolToIndices.get(E)?.[0];C!==void 0&&r.lhs.forEach((O,x)=>{if(I.inputIndices.includes(x)){let D=O.symbolToIndices.get(E);if(D===void 0)throw new Error("Invalid symbol error");D.forEach(L=>{h.push(`${a[x].indicesSet(`input${x}Indices`,L,n.indicesGet("outputIndices",C))}`)})}})}else r.lhs.forEach((C,O)=>{if(I.inputIndices.includes(O)){let x=C.symbolToIndices.get(E);if(x===void 0)throw new Error("Invalid symbol error");x.forEach(D=>{y.push(`${a[O].indicesSet(`input${O}Indices`,D,`${E}`)}`)}),v.push(`prod *= ${a[O].getByIndices(`input${O}Indices`)};`)}}),$.push(`for(var ${E}: u32 = 0; ${E} < uniforms.${Zi(E)}; ${E}++) {`),S.push("}")});let k=w?[...h,`let sum = ${a.map((I,E)=>I.getByIndices(`input${E}Indices`)).join(" * ")};`]:[...h,g,...$,...y,f,...v,_,...S];return`
             ${p.registerUniforms(u.map(I=>({name:`${Zi(I)}`,type:"u32"}))).registerUniform("outputSize","u32").declareVariables(...a,n)}
 
             ${p.mainStart()}
             ${p.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.outputSize")}
             var outputIndices = ${n.offsetToIndices("global_idx")};
-            ${a.map((I,k)=>`var input${k}Indices: ${a[k].type.indices};`).join(`
+            ${a.map((I,E)=>`var input${E}Indices: ${a[E].type.indices};`).join(`
 `)}
-            ${E.join(`
+            ${k.join(`
 `)};
             ${n.setByOffset("global_idx","sum")};
-          }`};return{name:"Einsum",shaderCache:{hint:r.equation,inputDependencies:e.map(()=>"rank")},getRunData:()=>{let p=u.filter(f=>r.symbolToInfo.has(f)).map(f=>({type:12,data:r.symbolToInfo.get(f)?.dimValue||0}));p.push({type:12,data:s});let h=e.map((f,g)=>[...Z(f)]).reduce((f,g)=>f.concat(g),p);return h.push(...Z(i)),{outputs:[{dims:i,dataType:t}],dispatchGroup:{x:Math.ceil(s/64)},programUniforms:h}},getShaderSource:l}},eh=(e,t)=>{let r=new Zu(e.inputs,t.equation),i=r.outputDims,a=e.inputs.map((s,n)=>s.dims);e.compute(Qu(a,e.inputs[0].dataType,r,i))},th=e=>{let t=e.equation.replace(/\s+/g,"");return he({equation:t})}}),Xu,Qi,Yu,Ju,rh,$0=P(()=>{"use strict";Y(),te(),re(),Xu=e=>{if(!e||e.length!==2)throw new Error("Expand requires 2 input.");let t=e[0].dims,r=Array.from(e[1].getBigInt64Array(),Number),i=r.length<t.length?0:r.length-t.length,a=t.length<r.length?0:t.length-r.length;for(;i<r.length&&a<t.length;++i,++a)if(r[i]!==t[a]&&r[i]!==1&&t[a]!==1)throw new Error("Expand requires shape to be broadcastable to input")},Qi=(e,t)=>{let r=e.length-t.length,i=[];for(let a=0;a<r;++a)i.push(e[a]);for(let a=0;a<t.length;++a)i.push(t[a]===1?e[a+r]:t[a]);return i},Yu=(e,t)=>e.length>t.length?Qi(e,t):Qi(t,e),Ju=e=>{let t=e[0].dims,r=Array.from(e[1].getBigInt64Array(),Number),i=Yu(t,r),a=e[0].dataType,s=a===9||B.size(t)===1,n=a===9||t.length>0&&t[t.length-1]%4===0?4:1,u=s||i.length>0&&i[i.length-1]%4===0?4:1,l=Math.ceil(B.size(i)/u),p=f=>{let g=N("input",a,t.length,n),_=F("output",a,i.length,u),y;if(a===9){let $=(S,v,w="")=>`
+          }`};return{name:"Einsum",shaderCache:{hint:r.equation,inputDependencies:e.map(()=>"rank")},getRunData:()=>{let p=u.filter(f=>r.symbolToInfo.has(f)).map(f=>({type:12,data:r.symbolToInfo.get(f)?.dimValue||0}));p.push({type:12,data:s});let h=e.map((f,g)=>[...Z(f)]).reduce((f,g)=>f.concat(g),p);return h.push(...Z(i)),{outputs:[{dims:i,dataType:t}],dispatchGroup:{x:Math.ceil(s/64)},programUniforms:h}},getShaderSource:l}},eh=(e,t)=>{let r=new Zu(e.inputs,t.equation),i=r.outputDims,a=e.inputs.map((s,n)=>s.dims);e.compute(Qu(a,e.inputs[0].dataType,r,i))},th=e=>{let t=e.equation.replace(/\s+/g,"");return he({equation:t})}}),Xu,Qi,Yu,Ju,rh,$0=P(()=>{"use strict";Y(),te(),re(),Xu=e=>{if(!e||e.length!==2)throw new Error("Expand requires 2 input.");let t=e[0].dims,r=Array.from(e[1].getBigInt64Array(),Number),i=r.length<t.length?0:r.length-t.length,a=t.length<r.length?0:t.length-r.length;for(;i<r.length&&a<t.length;++i,++a)if(r[i]!==t[a]&&r[i]!==1&&t[a]!==1)throw new Error("Expand requires shape to be broadcastable to input")},Qi=(e,t)=>{let r=e.length-t.length,i=[];for(let a=0;a<r;++a)i.push(e[a]);for(let a=0;a<t.length;++a)i.push(t[a]===1?e[a+r]:t[a]);return i},Yu=(e,t)=>e.length>t.length?Qi(e,t):Qi(t,e),Ju=e=>{let t=e[0].dims,r=Array.from(e[1].getBigInt64Array(),Number),i=Yu(t,r),a=e[0].dataType,s=a===9||B.size(t)===1,n=a===9||t.length>0&&t[t.length-1]%4===0?4:1,u=s||i.length>0&&i[i.length-1]%4===0?4:1,l=Math.ceil(B.size(i)/u),p=f=>{let g=M("input",a,t.length,n),_=F("output",a,i.length,u),y;if(a===9){let $=(S,v,w="")=>`
           let outputIndices${v} = ${_.offsetToIndices(`outputOffset + ${v}u`)};
           let offset${v} = ${g.broadcastedIndicesToOffset(`outputIndices${v}`,_)};
           let index${v} = offset${v} / 4u;
@@ -1351,7 +1351,7 @@ fn main(@builtin(local_invocation_id) localId : vec3<u32>,
     ${f.registerUniform("vec_size","u32").declareVariables(g,_)}
     ${f.mainStart()}
     ${f.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.vec_size")}
-    ${y}`},h=[{type:12,data:l},...Z(t,i)];return{name:"Expand",shaderCache:{hint:`${i.length};${n}${u}`,inputDependencies:["rank"]},getShaderSource:p,getRunData:()=>({outputs:[{dims:i,dataType:e[0].dataType}],dispatchGroup:{x:Math.ceil(l/64)},programUniforms:h})}},rh=e=>{Xu(e.inputs),e.compute(Ju(e.inputs),{inputs:[0]})}}),el,ih,v0=P(()=>{"use strict";Y(),te(),re(),ja(),el=e=>{let t=e[0].dataType,r=B.size(e[0].dims),i=B.size(e[1].dims),a=i%4===0,s=n=>{let u=N("x",t,[1],4),l=N("bias",t,[1],4),p=F("y",t,[1],4),h=[{name:"output_vec_size",type:"u32"},{name:"bias_size",type:"u32"}],f=_=>`
+    ${y}`},h=[{type:12,data:l},...Z(t,i)];return{name:"Expand",shaderCache:{hint:`${i.length};${n}${u}`,inputDependencies:["rank"]},getShaderSource:p,getRunData:()=>({outputs:[{dims:i,dataType:e[0].dataType}],dispatchGroup:{x:Math.ceil(l/64)},programUniforms:h})}},rh=e=>{Xu(e.inputs),e.compute(Ju(e.inputs),{inputs:[0]})}}),el,ih,v0=P(()=>{"use strict";Y(),te(),re(),ja(),el=e=>{let t=e[0].dataType,r=B.size(e[0].dims),i=B.size(e[1].dims),a=i%4===0,s=n=>{let u=M("x",t,[1],4),l=M("bias",t,[1],4),p=F("y",t,[1],4),h=[{name:"output_vec_size",type:"u32"},{name:"bias_size",type:"u32"}],f=_=>`
       let bias${_}_offset: u32 = (global_idx * 4 + ${_}) % uniforms.bias_size;
       let bias${_} = ${l.getByOffset(`bias${_}_offset / 4`)}[bias${_}_offset % 4];`,g=a?`
       let bias = ${l.getByOffset("global_idx % (uniforms.bias_size / 4)")};`:`${f(0)}${f(1)}${f(2)}${f(3)}
@@ -1366,19 +1366,19 @@ fn main(@builtin(local_invocation_id) localId : vec3<u32>,
       ${g}
       let x_in = x + bias;
       ${p.setByOffset("global_idx",Sa("x_in"))}
-    }`};return{name:"FastGeluWithBias",shaderCache:{hint:`${a}`,inputDependencies:["type","type"]},getShaderSource:s,getRunData:n=>({outputs:[{dims:n[0].dims,dataType:n[0].dataType}],programUniforms:[{type:12,data:Math.ceil(r/4)},{type:12,data:i}],dispatchGroup:{x:Math.ceil(r/Vt/4)}})}},ih=e=>{e.inputs.length<2||B.size(e.inputs[1].dims)===0?xc(e):e.compute(el(e.inputs))}}),tl,rl,ah,nh,x0=P(()=>{"use strict";Y(),te(),ve(),re(),tl=e=>{if(!e||e.length!==2)throw new Error("Gather requires 2 inputs.")},rl=(e,t)=>{let r=e[0].dims,i=e[1].dims,a=r.length,s=B.normalizeAxis(t.axis,a),n=r.slice(0);n.splice(s,1,...i);let u=r[s],l=e[0].dataType===9?4:1,p=Math.ceil(B.size(n)/l),h=[{type:12,data:p},{type:6,data:u},{type:12,data:s},...Z(e[0].dims,e[1].dims,n)],f=g=>{let _=N("data",e[0].dataType,e[0].dims.length,l),y=N("inputIndices",e[1].dataType,e[1].dims.length),$=F("output",e[0].dataType,n.length,l),S=w=>{let E=i.length,I=`var indicesIndices${w}  = ${y.type.indices}(0);`;for(let k=0;k<E;k++)I+=`${E>1?`indicesIndices${w}[${k}]`:`indicesIndices${w}`} = ${n.length>1?`outputIndices${w}[uniforms.axis + ${k}]`:`outputIndices${w}`};`;I+=`
+    }`};return{name:"FastGeluWithBias",shaderCache:{hint:`${a}`,inputDependencies:["type","type"]},getShaderSource:s,getRunData:n=>({outputs:[{dims:n[0].dims,dataType:n[0].dataType}],programUniforms:[{type:12,data:Math.ceil(r/4)},{type:12,data:i}],dispatchGroup:{x:Math.ceil(r/Vt/4)}})}},ih=e=>{e.inputs.length<2||B.size(e.inputs[1].dims)===0?xc(e):e.compute(el(e.inputs))}}),tl,rl,ah,nh,x0=P(()=>{"use strict";Y(),te(),ve(),re(),tl=e=>{if(!e||e.length!==2)throw new Error("Gather requires 2 inputs.")},rl=(e,t)=>{let r=e[0].dims,i=e[1].dims,a=r.length,s=B.normalizeAxis(t.axis,a),n=r.slice(0);n.splice(s,1,...i);let u=r[s],l=e[0].dataType===9?4:1,p=Math.ceil(B.size(n)/l),h=[{type:12,data:p},{type:6,data:u},{type:12,data:s},...Z(e[0].dims,e[1].dims,n)],f=g=>{let _=M("data",e[0].dataType,e[0].dims.length,l),y=M("inputIndices",e[1].dataType,e[1].dims.length),$=F("output",e[0].dataType,n.length,l),S=w=>{let k=i.length,I=`var indicesIndices${w}  = ${y.type.indices}(0);`;for(let E=0;E<k;E++)I+=`${k>1?`indicesIndices${w}[${E}]`:`indicesIndices${w}`} = ${n.length>1?`outputIndices${w}[uniforms.axis + ${E}]`:`outputIndices${w}`};`;I+=`
           var idx${w} = ${y.getByIndices(`indicesIndices${w}`)};
           if (idx${w} < 0) {
             idx${w} = idx${w} + uniforms.axisDimLimit;
           }
           var dataIndices${w} : ${_.type.indices};
-        `;for(let k=0,C=0;k<a;k++)k===s?(I+=`${a>1?`dataIndices${w}[${k}]`:`dataIndices${w}`} = u32(idx${w});`,C+=E):(I+=`${a>1?`dataIndices${w}[${k}]`:`dataIndices${w}`} = ${n.length>1?`outputIndices${w}[${C}]`:`outputIndices${w}`};`,C++);return I},v;if(e[0].dataType===9){let w=(E,I,k="")=>`
+        `;for(let E=0,C=0;E<a;E++)E===s?(I+=`${a>1?`dataIndices${w}[${E}]`:`dataIndices${w}`} = u32(idx${w});`,C+=k):(I+=`${a>1?`dataIndices${w}[${E}]`:`dataIndices${w}`} = ${n.length>1?`outputIndices${w}[${C}]`:`outputIndices${w}`};`,C++);return I},v;if(e[0].dataType===9){let w=(k,I,E="")=>`
           let outputIndices${I} = ${$.offsetToIndices(`outputOffset + ${I}u`)};
           ${S(I)};
           let offset${I} = ${_.indicesToOffset(`dataIndices${I}`)};
           let index${I} = offset${I} / 4u;
           let component${I} = offset${I} % 4u;
-          ${E}[${I}] = ${k}(${_.getByOffset(`index${I}`)}[component${I}]);
+          ${k}[${I}] = ${E}(${_.getByOffset(`index${I}`)}[component${I}]);
         `;v=`
         let outputOffset = global_idx * ${l};
         var value = vec4<u32>(0);
@@ -1397,7 +1397,7 @@ fn main(@builtin(local_invocation_id) localId : vec3<u32>,
       ${g.mainStart()}
         ${g.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.outputSize")}
         ${v}
-      }`};return{name:"Gather",shaderCache:{hint:t.cacheKey,inputDependencies:["rank","rank"]},getRunData:()=>({outputs:[{dims:n,dataType:e[0].dataType}],dispatchGroup:{x:Math.ceil(p/64)},programUniforms:h}),getShaderSource:f}},ah=e=>he({axis:e.axis}),nh=(e,t)=>{let r=e.inputs;tl(r),e.compute(rl(e.inputs,t))}}),il,sh,oh,S0=P(()=>{"use strict";Y(),te(),re(),il=(e,t,r,i,a,s,n,u,l)=>{let p=[{type:12,data:s},{type:12,data:i},{type:12,data:a},{type:12,data:r},{type:12,data:n},{type:12,data:u},{type:12,data:l}],h=[s];p.push(...Z(t.dims,h));let f=g=>{let _=N("indices_data",t.dataType,t.dims.length),y=F("input_slice_offsets_data",12,1,1),$=[_,y],S=[{name:"output_size",type:"u32"},{name:"batch_dims",type:"u32"},{name:"input_dims",type:"u32",length:a.length},{name:"sizes_from_slice_dims_data",type:"u32",length:r.length},{name:"num_slices_per_batch",type:"u32"},{name:"input_batch_stride",type:"u32"},{name:"num_slice_dims",type:"u32"}];return`
+      }`};return{name:"Gather",shaderCache:{hint:t.cacheKey,inputDependencies:["rank","rank"]},getRunData:()=>({outputs:[{dims:n,dataType:e[0].dataType}],dispatchGroup:{x:Math.ceil(p/64)},programUniforms:h}),getShaderSource:f}},ah=e=>he({axis:e.axis}),nh=(e,t)=>{let r=e.inputs;tl(r),e.compute(rl(e.inputs,t))}}),il,sh,oh,S0=P(()=>{"use strict";Y(),te(),re(),il=(e,t,r,i,a,s,n,u,l)=>{let p=[{type:12,data:s},{type:12,data:i},{type:12,data:a},{type:12,data:r},{type:12,data:n},{type:12,data:u},{type:12,data:l}],h=[s];p.push(...Z(t.dims,h));let f=g=>{let _=M("indices_data",t.dataType,t.dims.length),y=F("input_slice_offsets_data",12,1,1),$=[_,y],S=[{name:"output_size",type:"u32"},{name:"batch_dims",type:"u32"},{name:"input_dims",type:"u32",length:a.length},{name:"sizes_from_slice_dims_data",type:"u32",length:r.length},{name:"num_slices_per_batch",type:"u32"},{name:"input_batch_stride",type:"u32"},{name:"num_slice_dims",type:"u32"}];return`
   ${g.registerUniforms(S).declareVariables(...$)}
   ${g.mainStart()}
     ${g.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.output_size")}
@@ -1416,14 +1416,14 @@ fn main(@builtin(local_invocation_id) localId : vec3<u32>,
     }
 
     input_slice_offsets_data[global_idx] =  base_offset + u32(relative_slice_offset);
-  }`};return e.compute({name:"computeSliceOffsets",shaderCache:{hint:`${a.length}_${r.length}`,inputDependencies:["rank"]},getRunData:()=>({outputs:[{dims:h,dataType:e.inputs[1].dataType}],dispatchGroup:{x:Math.ceil(s/64)},programUniforms:p}),getShaderSource:f},{inputs:[t],outputs:[-1]})[0]},sh=(e,t)=>{let r=e.inputs,i=r[0].dims,a=r[0].dataType,s=r[1].dims,n=s[s.length-1],u=B.sizeToDimension(s,s.length-1),l=B.sizeFromDimension(i,t.batchDims+n),p=B.sizeToDimension(i,t.batchDims),h=B.sizeFromDimension(i,t.batchDims),f=u/p,g=new Array(n),_=l;for(let I=0;I<n;++I)g[n-1-I]=_,_*=i[t.batchDims+n-1-I];let y=il(e,r[1],g,t.batchDims,i,u,f,h,n),$=t.batchDims+n;if($>i.length)throw new Error("last dimension of indices must not be larger than rank of input tensor");let S=s.slice(0,-1).concat(i.slice($)),v=B.size(S),w=[{type:12,data:v},{type:12,data:l},...Z(r[0].dims,y.dims,S)],E=I=>{let k=N("data",r[0].dataType,r[0].dims.length),C=N("slice_offsets",12,y.dims.length),O=F("output",r[0].dataType,S.length);return`
-          ${I.registerUniform("output_size","u32").registerUniform("slice_size","u32").declareVariables(k,C,O)}
+  }`};return e.compute({name:"computeSliceOffsets",shaderCache:{hint:`${a.length}_${r.length}`,inputDependencies:["rank"]},getRunData:()=>({outputs:[{dims:h,dataType:e.inputs[1].dataType}],dispatchGroup:{x:Math.ceil(s/64)},programUniforms:p}),getShaderSource:f},{inputs:[t],outputs:[-1]})[0]},sh=(e,t)=>{let r=e.inputs,i=r[0].dims,a=r[0].dataType,s=r[1].dims,n=s[s.length-1],u=B.sizeToDimension(s,s.length-1),l=B.sizeFromDimension(i,t.batchDims+n),p=B.sizeToDimension(i,t.batchDims),h=B.sizeFromDimension(i,t.batchDims),f=u/p,g=new Array(n),_=l;for(let I=0;I<n;++I)g[n-1-I]=_,_*=i[t.batchDims+n-1-I];let y=il(e,r[1],g,t.batchDims,i,u,f,h,n),$=t.batchDims+n;if($>i.length)throw new Error("last dimension of indices must not be larger than rank of input tensor");let S=s.slice(0,-1).concat(i.slice($)),v=B.size(S),w=[{type:12,data:v},{type:12,data:l},...Z(r[0].dims,y.dims,S)],k=I=>{let E=M("data",r[0].dataType,r[0].dims.length),C=M("slice_offsets",12,y.dims.length),O=F("output",r[0].dataType,S.length);return`
+          ${I.registerUniform("output_size","u32").registerUniform("slice_size","u32").declareVariables(E,C,O)}
             ${I.mainStart()}
             ${I.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.output_size")}
           let slice_offset = slice_offsets[global_idx / uniforms.slice_size];
           output[global_idx] = data[u32(slice_offset) + global_idx % uniforms.slice_size];
-        }`};e.compute({name:"GatherND",shaderCache:{hint:t.cacheKey,inputDependencies:["rank","rank"]},getRunData:()=>({outputs:[{dims:S,dataType:a}],dispatchGroup:{x:Math.ceil(v/64)},programUniforms:w}),getShaderSource:E},{inputs:[r[0],y]})},oh=e=>({batchDims:e.batch_dims,cacheKey:""})}),al,nl,uh,lh,T0=P(()=>{"use strict";Y(),te(),ve(),re(),al=(e,t)=>{if(e.length<3||e.length>4)throw new Error("GatherBlockQuantized requires 3 or 4 inputs.");let r=B.normalizeAxis(t.quantizeAxis,e[0].dims.length),i=t.blockSize,a=e[0],s=e[2],n=e.length===4?e[3]:void 0;if(s.dims.length!==a.dims.length||!a.dims.map((u,l)=>l===r?Math.ceil(u/i)===s.dims[l]:u===s.dims[l]).reduce((u,l)=>u&&l,!0))throw new Error("Scales must have the same rank as the input tensor and the dims should match except on gatherAxis.");if(n){if(n.dataType!==a.dataType)throw new Error("Zero point must have the same data type as the input tensor.");if(n.dims.length!==s.dims.length||!n.dims.map((u,l)=>u===s.dims[l]).reduce((u,l)=>u&&l,!0))throw new Error("Zero point must have the same rank as the input tensor and the dims should match except on quantizeAxis.")}},nl=(e,t)=>{let r=e[0].dims,i=e[1].dims,a=r.length,s=B.normalizeAxis(t.gatherAxis,a),n=B.normalizeAxis(t.quantizeAxis,a),u=r.slice(0);u.splice(s,1,...i);let l=B.size(u),p=e[2].dataType,h=e[0].dataType===22,f=[{type:12,data:l},{type:12,data:n},{type:12,data:s},{type:12,data:t.blockSize},...Z(...e.map((_,y)=>_.dims),u)],g=_=>{let y=N("data",e[0].dataType,e[0].dims.length),$=N("inputIndices",e[1].dataType,e[1].dims.length),S=N("scales",e[2].dataType,e[2].dims.length),v=e.length>3?N("zeroPoint",e[3].dataType,e[3].dims.length):void 0,w=F("output",p,u.length),E=[y,$,S];v&&E.push(v);let I=[{name:"output_size",type:"u32"},{name:"quantize_axis",type:"u32"},{name:"gather_axis",type:"u32"},{name:"block_size",type:"u32"}];return`
-        ${_.registerUniforms(I).declareVariables(...E,w)}
+        }`};e.compute({name:"GatherND",shaderCache:{hint:t.cacheKey,inputDependencies:["rank","rank"]},getRunData:()=>({outputs:[{dims:S,dataType:a}],dispatchGroup:{x:Math.ceil(v/64)},programUniforms:w}),getShaderSource:k},{inputs:[r[0],y]})},oh=e=>({batchDims:e.batch_dims,cacheKey:""})}),al,nl,uh,lh,T0=P(()=>{"use strict";Y(),te(),ve(),re(),al=(e,t)=>{if(e.length<3||e.length>4)throw new Error("GatherBlockQuantized requires 3 or 4 inputs.");let r=B.normalizeAxis(t.quantizeAxis,e[0].dims.length),i=t.blockSize,a=e[0],s=e[2],n=e.length===4?e[3]:void 0;if(s.dims.length!==a.dims.length||!a.dims.map((u,l)=>l===r?Math.ceil(u/i)===s.dims[l]:u===s.dims[l]).reduce((u,l)=>u&&l,!0))throw new Error("Scales must have the same rank as the input tensor and the dims should match except on gatherAxis.");if(n){if(n.dataType!==a.dataType)throw new Error("Zero point must have the same data type as the input tensor.");if(n.dims.length!==s.dims.length||!n.dims.map((u,l)=>u===s.dims[l]).reduce((u,l)=>u&&l,!0))throw new Error("Zero point must have the same rank as the input tensor and the dims should match except on quantizeAxis.")}},nl=(e,t)=>{let r=e[0].dims,i=e[1].dims,a=r.length,s=B.normalizeAxis(t.gatherAxis,a),n=B.normalizeAxis(t.quantizeAxis,a),u=r.slice(0);u.splice(s,1,...i);let l=B.size(u),p=e[2].dataType,h=e[0].dataType===22,f=[{type:12,data:l},{type:12,data:n},{type:12,data:s},{type:12,data:t.blockSize},...Z(...e.map((_,y)=>_.dims),u)],g=_=>{let y=M("data",e[0].dataType,e[0].dims.length),$=M("inputIndices",e[1].dataType,e[1].dims.length),S=M("scales",e[2].dataType,e[2].dims.length),v=e.length>3?M("zeroPoint",e[3].dataType,e[3].dims.length):void 0,w=F("output",p,u.length),k=[y,$,S];v&&k.push(v);let I=[{name:"output_size",type:"u32"},{name:"quantize_axis",type:"u32"},{name:"gather_axis",type:"u32"},{name:"block_size",type:"u32"}];return`
+        ${_.registerUniforms(I).declareVariables(...k,w)}
         ${_.mainStart()}
         let output_indices = ${w.offsetToIndices("global_idx")};
         var indices_indices = ${$.type.indices}(0);
@@ -1468,7 +1468,7 @@ fn main(@builtin(local_invocation_id) localId : vec3<u32>,
         let dequantized_data = ${Ce(p)}(quantized_data - zero_point) * scale;
         ${w.setByOffset("global_idx","dequantized_data")};
     }`};return{name:"GatherBlockQuantized",shaderCache:{hint:`${t.cacheKey};${e.filter((_,y)=>y!==1).map(_=>_.dims.join("_")).join(";")}`,inputDependencies:Array.from({length:e.length},(_,y)=>"rank")},getRunData:()=>({outputs:[{dims:u,dataType:p}],dispatchGroup:{x:Math.ceil(l/64)},programUniforms:f}),getShaderSource:g}},uh=(e,t)=>{let r=e.inputs;al(r,t),e.compute(nl(e.inputs,t))},lh=e=>he({blockSize:e.blockSize,gatherAxis:e.gatherAxis,quantizeAxis:e.quantizeAxis})}),sl,ol,dh,ph,I0=P(()=>{"use strict";Y(),te(),ve(),re(),sl=e=>{if(!e||e.length!==2)throw new Error("GatherElements requires 2 inputs.");if(e[0].dims.length<1)throw new Error("GatherElements requires that the data input be rank >= 1.");if(e[0].dims.length!==e[1].dims.length)throw new Error(`GatherElements requires that the data input and
-                     indices input tensors be of same rank.`)},ol=(e,t)=>{let r=e[0].dims,i=e[0].dataType,a=r.length,s=e[1].dims,n=e[1].dataType,u=B.normalizeAxis(t.axis,a),l=r[u],p=s.slice(0),h=B.size(p),f=N("input",i,a),g=N("indicesInput",n,s.length),_=F("output",i,p.length),y=[{type:12,data:h},{type:6,data:l},{type:12,data:u}];return y.push(...Z(r,s,p)),{name:"GatherElements",shaderCache:{inputDependencies:["rank","rank"]},getRunData:()=>({outputs:[{dims:p,dataType:e[0].dataType}],dispatchGroup:{x:Math.ceil(h/64)},programUniforms:y}),getShaderSource:$=>`
+                     indices input tensors be of same rank.`)},ol=(e,t)=>{let r=e[0].dims,i=e[0].dataType,a=r.length,s=e[1].dims,n=e[1].dataType,u=B.normalizeAxis(t.axis,a),l=r[u],p=s.slice(0),h=B.size(p),f=M("input",i,a),g=M("indicesInput",n,s.length),_=F("output",i,p.length),y=[{type:12,data:h},{type:6,data:l},{type:12,data:u}];return y.push(...Z(r,s,p)),{name:"GatherElements",shaderCache:{inputDependencies:["rank","rank"]},getRunData:()=>({outputs:[{dims:p,dataType:e[0].dataType}],dispatchGroup:{x:Math.ceil(h/64)},programUniforms:y}),getShaderSource:$=>`
       ${$.registerUniform("outputSize","u32").registerUniform("axisDimLimit","i32").registerUniform("axis","u32").declareVariables(f,g,_)}
       ${$.mainStart()}
       ${$.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.outputSize")}
@@ -1484,8 +1484,8 @@ fn main(@builtin(local_invocation_id) localId : vec3<u32>,
       let value = ${f.getByIndices("inputIndices")};
 
       ${_.setByOffset("global_idx","value")};
-  }`}},dh=e=>he({axis:e.axis}),ph=(e,t)=>{let r=e.inputs;sl(r),e.compute(ol(e.inputs,t))}}),ul,ll,ch,hh,E0=P(()=>{"use strict";Y(),te(),re(),ul=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")},ll=(e,t)=>{let r=e[0].dims.slice(),i=e[1].dims.slice(),[a,s,n]=dp.getShapeOfGemmResult(r,t.transA,i,t.transB,e.length===3?e[2].dims:void 0),u=[a,s];if(!u)throw new Error("Can't use gemm on the given tensors");let l=16,p=Math.ceil(s/l),h=Math.ceil(a/l),f=!0,g=B.size(u),_=[{type:12,data:f?p:g},{type:12,data:a},{type:12,data:s},{type:12,data:n},{type:1,data:t.alpha},{type:1,data:t.beta}],y=["type","type"];e.length===3&&(_.push(...Z(e[2].dims)),y.push("rank")),_.push(...Z(u));let $=v=>{let w="";t.transA&&t.transB?w="value += a[k * uniforms.M + m] * b[n * uniforms.K + k];":t.transA&&!t.transB?w="value += a[k * uniforms.M + m] * b[k * uniforms.N + n];":!t.transA&&t.transB?w="value += a[m * uniforms.K + k] * b[n * uniforms.K + k];":!t.transA&&!t.transB&&(w="value += a[m * uniforms.K + k] * b[k * uniforms.N + n];");let E=t.alpha===1?"":"value *= uniforms.alpha;",I=N("a",e[0].dataType,e[0].dims),k=N("b",e[1].dataType,e[1].dims),C=I.type.value,O=null,x=[I,k];e.length===3&&(O=N("c",e[2].dataType,e[2].dims.length),x.push(O));let D=F("output",e[0].dataType,u.length);x.push(D);let q=[{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`
-  ${v.registerUniforms(q).declareVariables(...x)}
+  }`}},dh=e=>he({axis:e.axis}),ph=(e,t)=>{let r=e.inputs;sl(r),e.compute(ol(e.inputs,t))}}),ul,ll,ch,hh,k0=P(()=>{"use strict";Y(),te(),re(),ul=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")},ll=(e,t)=>{let r=e[0].dims.slice(),i=e[1].dims.slice(),[a,s,n]=dp.getShapeOfGemmResult(r,t.transA,i,t.transB,e.length===3?e[2].dims:void 0),u=[a,s];if(!u)throw new Error("Can't use gemm on the given tensors");let l=16,p=Math.ceil(s/l),h=Math.ceil(a/l),f=!0,g=B.size(u),_=[{type:12,data:f?p:g},{type:12,data:a},{type:12,data:s},{type:12,data:n},{type:1,data:t.alpha},{type:1,data:t.beta}],y=["type","type"];e.length===3&&(_.push(...Z(e[2].dims)),y.push("rank")),_.push(...Z(u));let $=v=>{let w="";t.transA&&t.transB?w="value += a[k * uniforms.M + m] * b[n * uniforms.K + k];":t.transA&&!t.transB?w="value += a[k * uniforms.M + m] * b[k * uniforms.N + n];":!t.transA&&t.transB?w="value += a[m * uniforms.K + k] * b[n * uniforms.K + k];":!t.transA&&!t.transB&&(w="value += a[m * uniforms.K + k] * b[k * uniforms.N + n];");let k=t.alpha===1?"":"value *= uniforms.alpha;",I=M("a",e[0].dataType,e[0].dims),E=M("b",e[1].dataType,e[1].dims),C=I.type.value,O=null,x=[I,E];e.length===3&&(O=M("c",e[2].dataType,e[2].dims.length),x.push(O));let D=F("output",e[0].dataType,u.length);x.push(D);let L=[{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`
+  ${v.registerUniforms(L).declareVariables(...x)}
 
   ${v.mainStart()}
     ${v.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.output_size")}
@@ -1498,10 +1498,10 @@ fn main(@builtin(local_invocation_id) localId : vec3<u32>,
       ${w}
     }
 
-    ${E}
+    ${k}
     ${O!=null?`let cOffset = ${O.broadcastedIndicesToOffset("vec2(m, n)",D)}; value += ${C}(uniforms.beta) * ${O.getByOffset("cOffset")};`:""}
     output[global_idx] = value;
-  }`},S=v=>{let w=N("a",e[0].dataType,e[0].dims),E=N("b",e[1].dataType,e[1].dims),I=null,k=[w,E];e.length===3&&(I=N("c",e[2].dataType,e[2].dims.length),k.push(I));let C=F("output",e[0].dataType,u.length);k.push(C);let O=[{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"}],x="",D="";t.transA&&t.transB?(D=`
+  }`},S=v=>{let w=M("a",e[0].dataType,e[0].dims),k=M("b",e[1].dataType,e[1].dims),I=null,E=[w,k];e.length===3&&(I=M("c",e[2].dataType,e[2].dims.length),E.push(I));let C=F("output",e[0].dataType,u.length);E.push(C);let O=[{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"}],x="",D="";t.transA&&t.transB?(D=`
       var col = tile_row_start + local_id.x;
       var row = k_start + local_id.y;
       if (col < uniforms.M && row < uniforms.K) {
@@ -1515,7 +1515,7 @@ fn main(@builtin(local_invocation_id) localId : vec3<u32>,
       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] = ${E.type.value}(0);
+        tile_b[local_id.y][local_id.x] = ${k.type.value}(0);
       }
       `,x="value += tile_a[k][local_id.y] * tile_b[local_id.x][k];"):t.transA&&!t.transB?(D=`
       var col = tile_row_start + local_id.x;
@@ -1531,7 +1531,7 @@ fn main(@builtin(local_invocation_id) localId : vec3<u32>,
       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] = ${E.type.value}(0);
+        tile_b[local_id.y][local_id.x] = ${k.type.value}(0);
       }
       `,x="value += tile_a[k][local_id.y] * tile_b[k][local_id.x];"):!t.transA&&t.transB?(D=`
       var col = k_start + local_id.x;
@@ -1547,7 +1547,7 @@ fn main(@builtin(local_invocation_id) localId : vec3<u32>,
       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] = ${E.type.value}(0);
+        tile_b[local_id.y][local_id.x] = ${k.type.value}(0);
       }
       `,x="value += tile_a[local_id.y][k] * tile_b[local_id.x][k];"):!t.transA&&!t.transB&&(D=`
       var col = k_start + local_id.x;
@@ -1563,12 +1563,12 @@ fn main(@builtin(local_invocation_id) localId : vec3<u32>,
       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] = ${E.type.value}(0);
+        tile_b[local_id.y][local_id.x] = ${k.type.value}(0);
       }
-      `,x="value += tile_a[local_id.y][k] * tile_b[k][local_id.x];");let q=t.alpha===1?"":"value *= uniforms.alpha;";return`
-  ${v.registerUniforms(O).declareVariables(...k)}
+      `,x="value += tile_a[local_id.y][k] * tile_b[k][local_id.x];");let L=t.alpha===1?"":"value *= uniforms.alpha;";return`
+  ${v.registerUniforms(O).declareVariables(...E)}
   var<workgroup> tile_a: array<array<${w.type.storage}, ${l}>, ${l}>;
-  var<workgroup> tile_b: array<array<${E.type.storage}, ${l}>, ${l}>;
+  var<workgroup> tile_b: array<array<${k.type.storage}, ${l}>, ${l}>;
   ${v.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};
@@ -1586,14 +1586,14 @@ fn main(@builtin(local_invocation_id) localId : vec3<u32>,
       workgroupBarrier();
     }
 
-    ${q}
+    ${L}
     let m = tile_row_start + local_id.y;
     let n = tile_col_start + local_id.x;
     ${I!=null?`let cOffset = ${I.broadcastedIndicesToOffset("vec2(m, n)",C)}; value += ${C.type.value}(uniforms.beta) * ${I.getByOffset("cOffset")};`:""}
     if (m < uniforms.M && n < uniforms.N) {
       output[m * uniforms.N + n] = value;
     }
-  }`};return f?{name:"GemmShared",shaderCache:{hint:`${t.cacheKey}`,inputDependencies:y},getRunData:()=>({outputs:[{dims:u,dataType:e[0].dataType}],dispatchGroup:{x:p*h},programUniforms:_}),getShaderSource:S}:{name:"Gemm",shaderCache:{hint:`${t.cacheKey}`,inputDependencies:y},getRunData:()=>({outputs:[{dims:u,dataType:e[0].dataType}],dispatchGroup:{x:Math.ceil(g/64)},programUniforms:_}),getShaderSource:$}},ch=e=>{let t=e.transA,r=e.transB,i=e.alpha,a=e.beta;return{transA:t,transB:r,alpha:i,beta:a,cacheKey:`${e.transA};${e.transB};${e.alpha===1}`}},hh=(e,t)=>{ul(e.inputs),e.compute(ll(e.inputs,t))}}),Je,it,Tt,It,dl,pl,cl,hl,fl,ml,gl,yl,fh,mh,k0=P(()=>{"use strict";Y(),te(),ve(),re(),[Je,it,Tt,It]=[0,1,2,3],dl=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 ${e[0].dims.length-2}`);if(e[0].dims[0]!==e[1].dims[0])throw new Error("grid batch size must match input batch size")},pl=`
+  }`};return f?{name:"GemmShared",shaderCache:{hint:`${t.cacheKey}`,inputDependencies:y},getRunData:()=>({outputs:[{dims:u,dataType:e[0].dataType}],dispatchGroup:{x:p*h},programUniforms:_}),getShaderSource:S}:{name:"Gemm",shaderCache:{hint:`${t.cacheKey}`,inputDependencies:y},getRunData:()=>({outputs:[{dims:u,dataType:e[0].dataType}],dispatchGroup:{x:Math.ceil(g/64)},programUniforms:_}),getShaderSource:$}},ch=e=>{let t=e.transA,r=e.transB,i=e.alpha,a=e.beta;return{transA:t,transB:r,alpha:i,beta:a,cacheKey:`${e.transA};${e.transB};${e.alpha===1}`}},hh=(e,t)=>{ul(e.inputs),e.compute(ll(e.inputs,t))}}),Je,it,Tt,It,dl,pl,cl,hl,fl,ml,gl,yl,fh,mh,E0=P(()=>{"use strict";Y(),te(),ve(),re(),[Je,it,Tt,It]=[0,1,2,3],dl=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")},pl=`
   fn gs_get_cubic_coeffs(x: f32) -> vec4<f32> {
     let cubic_alpha = -0.75f;
     let x_abs = abs(x);
@@ -1704,7 +1704,7 @@ fn main(@builtin(local_invocation_id) localId : vec3<u32>,
           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 ${r.mode} is not supported`)}})()+`${e.setByOffset("global_idx","result")}`,yl=(e,t)=>{let r=N("x",e[0].dataType,e[0].dims.length),i=[e[1].dims[0],e[1].dims[1],e[1].dims[2]],a=N("grid",e[1].dataType,i.length,2),s=[e[0].dims[0],e[0].dims[1],e[1].dims[1],e[1].dims[2]];t.format==="NHWC"&&(s=[e[0].dims[0],e[1].dims[1],e[1].dims[2],e[0].dims[3]],[Je,it,Tt,It]=[0,3,1,2]);let n=F("output",e[0].dataType,s.length),u=r.type.value,l=B.size(s),p=[{type:12,data:l},...Z(e[0].dims,i,s)],h=f=>`
+        `;default:throw new Error(`mode ${r.mode} is not supported`)}})()+`${e.setByOffset("global_idx","result")}`,yl=(e,t)=>{let r=M("x",e[0].dataType,e[0].dims.length),i=[e[1].dims[0],e[1].dims[1],e[1].dims[2]],a=M("grid",e[1].dataType,i.length,2),s=[e[0].dims[0],e[0].dims[1],e[1].dims[1],e[1].dims[2]];t.format==="NHWC"&&(s=[e[0].dims[0],e[1].dims[1],e[1].dims[2],e[0].dims[3]],[Je,it,Tt,It]=[0,3,1,2]);let n=F("output",e[0].dataType,s.length),u=r.type.value,l=B.size(s),p=[{type:12,data:l},...Z(e[0].dims,i,s)],h=f=>`
   ${f.registerUniform("output_size","u32").declareVariables(r,a,n)}
   ${pl}
   ${cl(u)}
@@ -1737,14 +1737,14 @@ fn main(@builtin(local_invocation_id) localId : vec3<u32>,
       var y = gs_denormalize(f32(nxy[1]), H_in);
 
       ${gl(n,u,t)}
-  }`;return{name:"GridSample",shaderCache:{hint:`${t.cacheKey}`,inputDependencies:["type","type"]},getRunData:f=>{let g=B.size(s);return{outputs:[{dims:s,dataType:f[0].dataType}],dispatchGroup:{x:Math.ceil(g/64)},programUniforms:p}},getShaderSource:h}},fh=(e,t)=>{dl(e.inputs),e.compute(yl(e.inputs,t))},mh=e=>he({alignCorners:e.align_corners,mode:e.mode,paddingMode:e.padding_mode,format:e.format})}),Ae,_l,gh,Xi,wl,lr,yh,_h=P(()=>{"use strict";Y(),te(),ve(),Va(),Ha(),re(),wt(),Ae=(e,t)=>e.length>t&&e[t].dims.length>0?e[t]:void 0,_l=(e,t)=>{let r=e[0],i=Ae(e,1),a=Ae(e,2),s=Ae(e,3),n=Ae(e,4),u=Ae(e,5),l=Ae(e,6),p=Ae(e,7);if(r.dims.length!==3&&r.dims.length!==5)throw new Error("Input query is expected to have 3 or 5 dimensions");let h=r.dims[0],f=r.dims[1],g=r.dims.length===3?r.dims[2]:t.numHeads*r.dims[4],_=f,y=0,$=0,S=Math.floor(g/t.numHeads);if(l&&p&&B.size(l.dims)&&B.size(p.dims)){if(l.dims.length!==4)throw new Error('Input "past_key" is expected to have 4 dimensions');if(l.dims[0]!==h||l.dims[1]!==t.numHeads||l.dims[3]!==S)throw new Error('Input "past_key" shape (batch_size, num_heads, past_sequence_length, head_size)');if(p.dims[0]!==h||p.dims[1]!==t.numHeads||p.dims[3]!==S)throw new Error('Input "past_value" shape (batch_size, num_heads, past_sequence_length, head_size)');if(l.dims[2]!==p.dims[2])throw new Error('Input "past_key" and "past_value" shall have same dim 2 (past_sequence_length)');if(p.dims.length!==4)throw new Error('Input "past_value" is expected to have 4 dimensions');y=l.dims[2],$=l.dims[2]}else if(l&&B.size(l.dims)||p&&B.size(p.dims))throw new Error('Input "past_key" and "past_value" shall be both present or both absent');let v;if(i&&B.size(i.dims)>0){if(r.dims.length!==3)throw new Error('Input "query" is expected to have 3 dimensions when key is given');if(i.dims.length<3||i.dims.length>5)throw new Error('Input "key" is expected to have 3, 4, or 5 dimensions');if(r.dims[0]!==i.dims[0])throw new Error('Input "query" and "key" shall have same dim 0 (batch size)');if(i.dims.length===3){if(i.dims[2]!==r.dims[2])throw new Error('Input "query" and "key" shall have same dim 2 (hidden_size)');v=2,_=i.dims[1]}else if(i.dims.length===5){if(i.dims[2]!==t.numHeads||i.dims[3]!==2||i.dims[4]!==S)throw new Error('Expect "key" shape (batch_size, kv_sequence_length, num_heads, 2, head_size) for packed kv');if(a)throw new Error('Expect "value" be none when "key" has packed kv format.');v=5,_=i.dims[1]}else{if(i.dims[1]!==t.numHeads||i.dims[3]!==S)throw new Error('Expect "key" shape (batch_size, num_heads, kv_sequence_length, head_size) for past_key');v=0,_=i.dims[2]}}else{if(r.dims.length!==5)throw new Error('Input "query" is expected to have 5 dimensions when key is empty');if(r.dims[2]!==t.numHeads||r.dims[3]!==3)throw new Error('Expect "query" shape (batch_size, kv_sequence_length, num_heads, 3, head_size) for packed kv');v=3}if(s&&B.size(s.dims)>0){if(s.dims.length!==1)throw new Error('Input "bias" is expected to have 1 dimension');if(i&&i.dims.length===5&&i.dims[3]===2)throw new Error("bias is not allowed for packed kv.")}let w=y+_,E=0;if(n&&B.size(n.dims)>0){E=8;let O=n.dims;throw O.length===1?O[0]===h?E=1:O[0]===3*h+2&&(E=3):O.length===2&&O[0]===h&&O[1]===w&&(E=5),E===8?new Error('Input "key_padding_mask" shape shall be (batch_size) or (batch_size, total_sequence_length)'):new Error("Mask not supported")}let I=!1,k=g;if(a&&B.size(a.dims)>0){if(a.dims.length!==3&&a.dims.length!==4)throw new Error('Input "value" is expected to have 3 or 4 dimensions');if(r.dims[0]!==a.dims[0])throw new Error('Input "query" and "value" shall have same dim 0 (batch_size)');if(a.dims.length===3){if(_!==a.dims[1])throw new Error('Input "key" and "value" shall have the same dim 1 (kv_sequence_length)');k=a.dims[2]}else{if(_!==a.dims[2])throw new Error('Input "key" and "value" shall have the same dim 2 (kv_sequence_length)');k=a.dims[1]*a.dims[3],I=!0}}let C=!1;if(n&&B.size(n.dims)>0)throw new Error("Key padding mask is not supported");if(u&&B.size(u.dims)>0){if(u.dims.length!==4)throw new Error('Input "attention_bias" is expected to have 4 dimensions');if(u.dims[0]!==h||u.dims[1]!==t.numHeads||u.dims[2]!==f||u.dims[3]!==w)throw new Error('Expect "attention_bias" shape (batch_size, num_heads, sequence_length, total_sequence_length)')}return{batchSize:h,sequenceLength:f,pastSequenceLength:y,kvSequenceLength:_,totalSequenceLength:w,maxSequenceLength:$,inputHiddenSize:0,hiddenSize:g,vHiddenSize:k,headSize:S,vHeadSize:Math.floor(k/t.numHeads),numHeads:t.numHeads,isUnidirectional:!1,pastPresentShareBuffer:!1,maskFilterValue:t.maskFilterValue,maskType:E,scale:t.scale,broadcastResPosBias:C,passPastInKv:I,qkvFormat:v}},gh=e=>he({...e}),Xi=he({perm:[0,2,1,3]}),wl=(e,t,r,i,a,s,n)=>{let u=[i,a,s],l=B.size(u),p=[{type:12,data:l},{type:12,data:n},{type:12,data:s}],h=f=>{let g=F("qkv_with_bias",t.dataType,u),_=N("qkv",t.dataType,u),y=N("bias",r.dataType,u),$=[{name:"output_size",type:"u32"},{name:"bias_offset",type:"u32"},{name:"hidden_size",type:"u32"}];return`
+  }`;return{name:"GridSample",shaderCache:{hint:`${t.cacheKey}`,inputDependencies:["type","type"]},getRunData:f=>{let g=B.size(s);return{outputs:[{dims:s,dataType:f[0].dataType}],dispatchGroup:{x:Math.ceil(g/64)},programUniforms:p}},getShaderSource:h}},fh=(e,t)=>{dl(e.inputs),e.compute(yl(e.inputs,t))},mh=e=>he({alignCorners:e.align_corners,mode:e.mode,paddingMode:e.padding_mode,format:e.format})}),Ae,_l,gh,Xi,wl,lr,yh,_h=P(()=>{"use strict";Y(),te(),ve(),Va(),Ha(),re(),wt(),Ae=(e,t)=>e.length>t&&e[t].dims.length>0?e[t]:void 0,_l=(e,t)=>{let r=e[0],i=Ae(e,1),a=Ae(e,2),s=Ae(e,3),n=Ae(e,4),u=Ae(e,5),l=Ae(e,6),p=Ae(e,7);if(r.dims.length!==3&&r.dims.length!==5)throw new Error("Input query is expected to have 3 or 5 dimensions");let h=r.dims[0],f=r.dims[1],g=r.dims.length===3?r.dims[2]:t.numHeads*r.dims[4],_=f,y=0,$=0,S=Math.floor(g/t.numHeads);if(l&&p&&B.size(l.dims)&&B.size(p.dims)){if(l.dims.length!==4)throw new Error('Input "past_key" is expected to have 4 dimensions');if(l.dims[0]!==h||l.dims[1]!==t.numHeads||l.dims[3]!==S)throw new Error('Input "past_key" shape (batch_size, num_heads, past_sequence_length, head_size)');if(p.dims[0]!==h||p.dims[1]!==t.numHeads||p.dims[3]!==S)throw new Error('Input "past_value" shape (batch_size, num_heads, past_sequence_length, head_size)');if(l.dims[2]!==p.dims[2])throw new Error('Input "past_key" and "past_value" shall have same dim 2 (past_sequence_length)');if(p.dims.length!==4)throw new Error('Input "past_value" is expected to have 4 dimensions');y=l.dims[2],$=l.dims[2]}else if(l&&B.size(l.dims)||p&&B.size(p.dims))throw new Error('Input "past_key" and "past_value" shall be both present or both absent');let v;if(i&&B.size(i.dims)>0){if(r.dims.length!==3)throw new Error('Input "query" is expected to have 3 dimensions when key is given');if(i.dims.length<3||i.dims.length>5)throw new Error('Input "key" is expected to have 3, 4, or 5 dimensions');if(r.dims[0]!==i.dims[0])throw new Error('Input "query" and "key" shall have same dim 0 (batch size)');if(i.dims.length===3){if(i.dims[2]!==r.dims[2])throw new Error('Input "query" and "key" shall have same dim 2 (hidden_size)');v=2,_=i.dims[1]}else if(i.dims.length===5){if(i.dims[2]!==t.numHeads||i.dims[3]!==2||i.dims[4]!==S)throw new Error('Expect "key" shape (batch_size, kv_sequence_length, num_heads, 2, head_size) for packed kv');if(a)throw new Error('Expect "value" be none when "key" has packed kv format.');v=5,_=i.dims[1]}else{if(i.dims[1]!==t.numHeads||i.dims[3]!==S)throw new Error('Expect "key" shape (batch_size, num_heads, kv_sequence_length, head_size) for past_key');v=0,_=i.dims[2]}}else{if(r.dims.length!==5)throw new Error('Input "query" is expected to have 5 dimensions when key is empty');if(r.dims[2]!==t.numHeads||r.dims[3]!==3)throw new Error('Expect "query" shape (batch_size, kv_sequence_length, num_heads, 3, head_size) for packed kv');v=3}if(s&&B.size(s.dims)>0){if(s.dims.length!==1)throw new Error('Input "bias" is expected to have 1 dimension');if(i&&i.dims.length===5&&i.dims[3]===2)throw new Error("bias is not allowed for packed kv.")}let w=y+_,k=0;if(n&&B.size(n.dims)>0){k=8;let O=n.dims;throw O.length===1?O[0]===h?k=1:O[0]===3*h+2&&(k=3):O.length===2&&O[0]===h&&O[1]===w&&(k=5),k===8?new Error('Input "key_padding_mask" shape shall be (batch_size) or (batch_size, total_sequence_length)'):new Error("Mask not supported")}let I=!1,E=g;if(a&&B.size(a.dims)>0){if(a.dims.length!==3&&a.dims.length!==4)throw new Error('Input "value" is expected to have 3 or 4 dimensions');if(r.dims[0]!==a.dims[0])throw new Error('Input "query" and "value" shall have same dim 0 (batch_size)');if(a.dims.length===3){if(_!==a.dims[1])throw new Error('Input "key" and "value" shall have the same dim 1 (kv_sequence_length)');E=a.dims[2]}else{if(_!==a.dims[2])throw new Error('Input "key" and "value" shall have the same dim 2 (kv_sequence_length)');E=a.dims[1]*a.dims[3],I=!0}}let C=!1;if(n&&B.size(n.dims)>0)throw new Error("Key padding mask is not supported");if(u&&B.size(u.dims)>0){if(u.dims.length!==4)throw new Error('Input "attention_bias" is expected to have 4 dimensions');if(u.dims[0]!==h||u.dims[1]!==t.numHeads||u.dims[2]!==f||u.dims[3]!==w)throw new Error('Expect "attention_bias" shape (batch_size, num_heads, sequence_length, total_sequence_length)')}return{batchSize:h,sequenceLength:f,pastSequenceLength:y,kvSequenceLength:_,totalSequenceLength:w,maxSequenceLength:$,inputHiddenSize:0,hiddenSize:g,vHiddenSize:E,headSize:S,vHeadSize:Math.floor(E/t.numHeads),numHeads:t.numHeads,isUnidirectional:!1,pastPresentShareBuffer:!1,maskFilterValue:t.maskFilterValue,maskType:k,scale:t.scale,broadcastResPosBias:C,passPastInKv:I,qkvFormat:v}},gh=e=>he({...e}),Xi=he({perm:[0,2,1,3]}),wl=(e,t,r,i,a,s,n)=>{let u=[i,a,s],l=B.size(u),p=[{type:12,data:l},{type:12,data:n},{type:12,data:s}],h=f=>{let g=F("qkv_with_bias",t.dataType,u),_=M("qkv",t.dataType,u),y=M("bias",r.dataType,u),$=[{name:"output_size",type:"u32"},{name:"bias_offset",type:"u32"},{name:"hidden_size",type:"u32"}];return`
   ${f.registerUniforms($).declareVariables(_,y,g)}
   ${f.mainStart()}
     ${f.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:u,dataType:t.dataType,gpuDataType:0}],dispatchGroup:{x:Math.ceil(l/64)},programUniforms:p}),getShaderSource:h},{inputs:[t,r],outputs:[-1]})[0]},lr=(e,t,r,i,a,s,n,u)=>{let l=s;if(n&&B.size(n.dims)>0){if(i===1)throw new Error("AddBiasReshape is not implemented. Please export your model with packed QKV or KV");return l=wl(e,s,n,t,i,r*a,u),l=l.reshape([t,i,r,a]),r===1||i===1?l:e.compute(Me(l,Xi.perm),{inputs:[l],outputs:[-1]})[0]}else return s.dims.length===3&&(l=s.reshape([t,i,r,a])),r===1||i===1?l:e.compute(Me(l,Xi.perm),{inputs:[l],outputs:[-1]})[0]},yh=(e,t)=>{let r=_l(e.inputs,t),i=e.inputs[0],a=Ae(e.inputs,1),s=Ae(e.inputs,2),n=Ae(e.inputs,3),u=Ae(e.inputs,4),l=Ae(e.inputs,5),p=Ae(e.inputs,6),h=Ae(e.inputs,7);if(i.dims.length===5)throw new Error("Packed QKV is not implemented");if(a?.dims.length===5)throw new Error("Packed KV is not implemented");let f=a&&s&&a.dims.length===4&&s.dims.length===4,g=lr(e,r.batchSize,r.numHeads,r.sequenceLength,r.headSize,i,n,0);if(f)return hr(e,g,a,s,u,void 0,p,h,l,r);if(!a||!s)throw new Error("key and value must be provided");let _=lr(e,r.batchSize,r.numHeads,r.kvSequenceLength,r.headSize,a,n,r.hiddenSize),y=lr(e,r.batchSize,r.numHeads,r.kvSequenceLength,r.vHeadSize,s,n,2*r.hiddenSize);hr(e,g,_,y,u,void 0,p,h,l,r)}}),bl,$l,vl,xl,za,wh,bh,$h=P(()=>{"use strict";Y(),te(),ve(),re(),bl=e=>{if(!e||e.length<1)throw new Error("too few inputs")},$l=(e,t)=>{let r=[],i=t.numOutputs;return e[1].dims[0]>0&&(e[1].getBigInt64Array().forEach(a=>r.push(Number(a))),i=r.length),he({numOutputs:i,axis:t.axis,splitSizes:r})},vl=e=>`
+  }`};return e.compute({name:"MultiHeadAttentionAddBias",shaderCache:{inputDependencies:["type","type"]},getRunData:()=>({outputs:[{dims:u,dataType:t.dataType,gpuDataType:0}],dispatchGroup:{x:Math.ceil(l/64)},programUniforms:p}),getShaderSource:h},{inputs:[t,r],outputs:[-1]})[0]},lr=(e,t,r,i,a,s,n,u)=>{let l=s;if(n&&B.size(n.dims)>0){if(i===1)throw new Error("AddBiasReshape is not implemented. Please export your model with packed QKV or KV");return l=wl(e,s,n,t,i,r*a,u),l=l.reshape([t,i,r,a]),r===1||i===1?l:e.compute(Ne(l,Xi.perm),{inputs:[l],outputs:[-1]})[0]}else return s.dims.length===3&&(l=s.reshape([t,i,r,a])),r===1||i===1?l:e.compute(Ne(l,Xi.perm),{inputs:[l],outputs:[-1]})[0]},yh=(e,t)=>{let r=_l(e.inputs,t),i=e.inputs[0],a=Ae(e.inputs,1),s=Ae(e.inputs,2),n=Ae(e.inputs,3),u=Ae(e.inputs,4),l=Ae(e.inputs,5),p=Ae(e.inputs,6),h=Ae(e.inputs,7);if(i.dims.length===5)throw new Error("Packed QKV is not implemented");if(a?.dims.length===5)throw new Error("Packed KV is not implemented");let f=a&&s&&a.dims.length===4&&s.dims.length===4,g=lr(e,r.batchSize,r.numHeads,r.sequenceLength,r.headSize,i,n,0);if(f)return hr(e,g,a,s,u,void 0,p,h,l,r);if(!a||!s)throw new Error("key and value must be provided");let _=lr(e,r.batchSize,r.numHeads,r.kvSequenceLength,r.headSize,a,n,r.hiddenSize),y=lr(e,r.batchSize,r.numHeads,r.kvSequenceLength,r.vHeadSize,s,n,2*r.hiddenSize);hr(e,g,_,y,u,void 0,p,h,l,r)}}),bl,$l,vl,xl,za,wh,bh,$h=P(()=>{"use strict";Y(),te(),ve(),re(),bl=e=>{if(!e||e.length<1)throw new Error("too few inputs")},$l=(e,t)=>{let r=[],i=t.numOutputs;return e[1].dims[0]>0&&(e[1].getBigInt64Array().forEach(a=>r.push(Number(a))),i=r.length),he({numOutputs:i,axis:t.axis,splitSizes:r})},vl=e=>`
 fn calculateOutputIndex(index: u32) -> u32 {
     for (var i: u32 = 0u; i < ${e}u; i += 1u ) {
     if (index < ${j("uniforms.size_in_split_axis","i",e)}) {
@@ -1756,7 +1756,7 @@ fn calculateOutputIndex(index: u32) -> u32 {
       fn writeBufferData(output_number: u32, indices: ${e[0].type.indices}, global_idx: u32) {
         ${r.join(`
 `)}
-      }`},za=(e,t)=>{let r=e[0].dims,i=B.size(r),a=e[0].dataType,s=B.normalizeAxis(t.axis,r.length),n=new Array(t.numOutputs),u=N("input",a,r.length),l=new Array(t.numOutputs),p=[],h=[],f=0,g=[{type:12,data:i}];for(let y=0;y<t.numOutputs;y++){f+=t.splitSizes[y],l[y]=f;let $=r.slice();$[s]=t.splitSizes[y],h.push($),n[y]=F(`output${y}`,a,$.length),p.push({dims:h[y],dataType:e[0].dataType})}g.push({type:12,data:l},...Z(r,...h));let _=y=>`
+      }`},za=(e,t)=>{let r=e[0].dims,i=B.size(r),a=e[0].dataType,s=B.normalizeAxis(t.axis,r.length),n=new Array(t.numOutputs),u=M("input",a,r.length),l=new Array(t.numOutputs),p=[],h=[],f=0,g=[{type:12,data:i}];for(let y=0;y<t.numOutputs;y++){f+=t.splitSizes[y],l[y]=f;let $=r.slice();$[s]=t.splitSizes[y],h.push($),n[y]=F(`output${y}`,a,$.length),p.push({dims:h[y],dataType:e[0].dataType})}g.push({type:12,data:l},...Z(r,...h));let _=y=>`
   ${y.registerUniform("input_size","u32").registerUniform("size_in_split_axis","u32",l.length).declareVariables(u,...n)}
   ${vl(l.length)}
   ${xl(n)}
@@ -1772,11 +1772,11 @@ fn calculateOutputIndex(index: u32) -> u32 {
       ${u.indicesSet("indices",s,"index")};
     }
     writeBufferData(output_number, indices, global_idx);
-  }`;return{name:"Split",shaderCache:{hint:t.cacheKey,inputDependencies:["rank"]},getShaderSource:_,getRunData:()=>({outputs:p,dispatchGroup:{x:Math.ceil(i/64)},programUniforms:g})}},wh=(e,t)=>{bl(e.inputs);let r=e.inputs.length===1?t:$l(e.inputs,t);e.compute(za(e.inputs,r),{inputs:[0]})},bh=e=>{let t=e.axis,r=e.splitSizes,i=e.numOutputs<0?r.length:e.numOutputs;if(i!==r.length)throw new Error("numOutputs and splitSizes length must be equal");return he({axis:t,numOutputs:i,splitSizes:r})}}),Sl,jr,vh,xh=P(()=>{"use strict";Y(),te(),ve(),re(),Sl=(e,t)=>{let[r,i,a,s]=e,{numHeads:n,rotaryEmbeddingDim:u}=t;if(r.dims.length!==3&&r.dims.length!==4)throw new Error(`Input 'x' is expected to have 3 or 4 dimensions, got ${r.dims.length}`);if(!B.areEqual(i.dims,[])&&!B.areEqual(i.dims,[1])&&i.dims.length!==2)throw new Error(`Input 'position_ids' is expected to have 0, 1, or 2 dimensions, got ${i.dims.length}`);if(a.dims.length!==2)throw new Error(`Input 'cos_cache' is expected to have 2 dimensions, got ${a.dims.length}`);if(s.dims.length!==2)throw new Error(`Input 'sin_cache' is expected to have 2 dimensions, got ${s.dims.length}`);if(!B.areEqual(a.dims,s.dims))throw new Error("Inputs 'cos_cache' and 'sin_cache' are expected to have the same shape");if(u>0&&n===0)throw new Error("num_heads must be provided if rotary_embedding_dim is specified");let l=r.dims[0],p=r.dims[r.dims.length-2],h=a.dims[0],f=B.sizeFromDimension(r.dims,1)/p,g=u===0?a.dims[1]*2:f/n;if(u>g)throw new Error("rotary_embedding_dim must be less than or equal to head_size");if(i.dims.length===2){if(l!==i.dims[0])throw new Error(`Input 'position_ids' dimension 0 should be of size batch_size, got ${i.dims[0]}`);if(p!==i.dims[1])throw new Error(`Input 'position_ids' dimension 1 should be of size sequence_length, got ${i.dims[1]}`)}if(g/2!==a.dims[1]&&u/2!==a.dims[1])throw new Error(`Input 'cos_cache' dimension 1 should be same as head_size / 2 or rotary_embedding_dim / 2, got ${a.dims[1]}`);if(p>h)throw new Error("Updating cos_cache and sin_cache in RotaryEmbedding is not currently supported")},jr=(e,t)=>{let{interleaved:r,numHeads:i,rotaryEmbeddingDim:a,scale:s}=t,n=e[0].dims[0],u=B.sizeFromDimension(e[0].dims,1),l=e[0].dims[e[0].dims.length-2],p=u/l,h=e[2].dims[1],f=a===0?h*2:p/i,g=new Array(n,l,p/f,f-h),_=B.computeStrides(g),y=[{type:1,data:s},{type:12,data:g},{type:12,data:_},...e[0].dims.length===3?new Array({type:12,data:[u,p,f,1]}):[],...e[0].dims.length===4?new Array({type:12,data:[u,f,l*f,1]}):[],...Z(e[0].dims,e[1].dims,e[2].dims,e[3].dims,e[0].dims)],$=S=>{let v=N("input",e[0].dataType,e[0].dims.length),w=N("position_ids",e[1].dataType,e[1].dims.length),E=N("cos_cache",e[2].dataType,e[2].dims.length),I=N("sin_cache",e[3].dataType,e[3].dims.length),k=F("output",e[0].dataType,e[0].dims.length);return S.registerUniforms([{name:"scale",type:"f32"},{name:"global_shape",type:"u32",length:g.length},{name:"global_strides",type:"u32",length:_.length},{name:"input_output_strides",type:"u32",length:_.length}]),`
-        ${S.declareVariables(v,w,E,I,k)}
+  }`;return{name:"Split",shaderCache:{hint:t.cacheKey,inputDependencies:["rank"]},getShaderSource:_,getRunData:()=>({outputs:p,dispatchGroup:{x:Math.ceil(i/64)},programUniforms:g})}},wh=(e,t)=>{bl(e.inputs);let r=e.inputs.length===1?t:$l(e.inputs,t);e.compute(za(e.inputs,r),{inputs:[0]})},bh=e=>{let t=e.axis,r=e.splitSizes,i=e.numOutputs<0?r.length:e.numOutputs;if(i!==r.length)throw new Error("numOutputs and splitSizes length must be equal");return he({axis:t,numOutputs:i,splitSizes:r})}}),Sl,jr,vh,xh=P(()=>{"use strict";Y(),te(),ve(),re(),Sl=(e,t)=>{let[r,i,a,s]=e,{numHeads:n,rotaryEmbeddingDim:u}=t;if(r.dims.length!==3&&r.dims.length!==4)throw new Error(`Input 'x' is expected to have 3 or 4 dimensions, got ${r.dims.length}`);if(!B.areEqual(i.dims,[])&&!B.areEqual(i.dims,[1])&&i.dims.length!==2)throw new Error(`Input 'position_ids' is expected to have 0, 1, or 2 dimensions, got ${i.dims.length}`);if(a.dims.length!==2)throw new Error(`Input 'cos_cache' is expected to have 2 dimensions, got ${a.dims.length}`);if(s.dims.length!==2)throw new Error(`Input 'sin_cache' is expected to have 2 dimensions, got ${s.dims.length}`);if(!B.areEqual(a.dims,s.dims))throw new Error("Inputs 'cos_cache' and 'sin_cache' are expected to have the same shape");if(u>0&&n===0)throw new Error("num_heads must be provided if rotary_embedding_dim is specified");let l=r.dims[0],p=r.dims[r.dims.length-2],h=a.dims[0],f=B.sizeFromDimension(r.dims,1)/p,g=u===0?a.dims[1]*2:f/n;if(u>g)throw new Error("rotary_embedding_dim must be less than or equal to head_size");if(i.dims.length===2){if(l!==i.dims[0])throw new Error(`Input 'position_ids' dimension 0 should be of size batch_size, got ${i.dims[0]}`);if(p!==i.dims[1])throw new Error(`Input 'position_ids' dimension 1 should be of size sequence_length, got ${i.dims[1]}`)}if(g/2!==a.dims[1]&&u/2!==a.dims[1])throw new Error(`Input 'cos_cache' dimension 1 should be same as head_size / 2 or rotary_embedding_dim / 2, got ${a.dims[1]}`);if(p>h)throw new Error("Updating cos_cache and sin_cache in RotaryEmbedding is not currently supported")},jr=(e,t)=>{let{interleaved:r,numHeads:i,rotaryEmbeddingDim:a,scale:s}=t,n=e[0].dims[0],u=B.sizeFromDimension(e[0].dims,1),l=e[0].dims[e[0].dims.length-2],p=u/l,h=e[2].dims[1],f=a===0?h*2:p/i,g=new Array(n,l,p/f,f-h),_=B.computeStrides(g),y=[{type:1,data:s},{type:12,data:g},{type:12,data:_},...e[0].dims.length===3?new Array({type:12,data:[u,p,f,1]}):[],...e[0].dims.length===4?new Array({type:12,data:[u,f,l*f,1]}):[],...Z(e[0].dims,e[1].dims,e[2].dims,e[3].dims,e[0].dims)],$=S=>{let v=M("input",e[0].dataType,e[0].dims.length),w=M("position_ids",e[1].dataType,e[1].dims.length),k=M("cos_cache",e[2].dataType,e[2].dims.length),I=M("sin_cache",e[3].dataType,e[3].dims.length),E=F("output",e[0].dataType,e[0].dims.length);return S.registerUniforms([{name:"scale",type:"f32"},{name:"global_shape",type:"u32",length:g.length},{name:"global_strides",type:"u32",length:_.length},{name:"input_output_strides",type:"u32",length:_.length}]),`
+        ${S.declareVariables(v,w,k,I,E)}
 
         ${S.mainStart(Vt)}
-          let half_rotary_emb_dim = uniforms.${E.name}_shape[1];
+          let half_rotary_emb_dim = uniforms.${k.name}_shape[1];
           let bsnh = global_idx / uniforms.global_strides % uniforms.global_shape;
           let size = uniforms.global_shape[0] * uniforms.global_strides[0];
           ${S.guardAgainstOutOfBoundsWorkgroupSizes("size")}
@@ -1788,17 +1788,17 @@ fn calculateOutputIndex(index: u32) -> u32 {
                 u32(${w.getByOffset("position_ids_idx")}) + select(0, bsnh[1], position_ids_idx == 0);
             let i = dot(bsnh, uniforms.input_output_strides) + select(0, bsnh[3], ${r});
             let j = i + select(half_rotary_emb_dim, 1, ${r});
-            let re = ${v.getByOffset("i")} * ${E.get("position_id","bsnh[3]")} -
+            let re = ${v.getByOffset("i")} * ${k.get("position_id","bsnh[3]")} -
                 ${v.getByOffset("j")} * ${I.get("position_id","bsnh[3]")};
-            ${k.setByOffset("i","re")}
+            ${E.setByOffset("i","re")}
             let im = ${v.getByOffset("i")} * ${I.get("position_id","bsnh[3]")} +
-                ${v.getByOffset("j")} * ${E.get("position_id","bsnh[3]")};
-            ${k.setByOffset("j","im")}
+                ${v.getByOffset("j")} * ${k.get("position_id","bsnh[3]")};
+            ${E.setByOffset("j","im")}
           } else {
             let k = dot(bsnh, uniforms.input_output_strides) + half_rotary_emb_dim;
-            ${k.setByOffset("k",v.getByOffset("k"))}
+            ${E.setByOffset("k",v.getByOffset("k"))}
           }
-        }`};return{name:"RotaryEmbedding",shaderCache:{hint:he({interleaved:r}).cacheKey,inputDependencies:["rank","rank","rank","rank"]},getShaderSource:$,getRunData:()=>({outputs:[{dims:e[0].dims,dataType:e[0].dataType}],dispatchGroup:{x:Math.ceil(B.size(g)/Vt)},programUniforms:y})}},vh=(e,t)=>{Sl(e.inputs,t),e.compute(jr(e.inputs,t))}}),Tl,Il,Yi,El,Sh,z0=P(()=>{"use strict";ve(),Y(),Ha(),_h(),$h(),wt(),xh(),re(),Tl=(e,t)=>{if(t.doRotary&&e.length<=7)throw new Error("cos_cache and sin_cache inputs are required if do_rotary is specified");let r=e[0],i=e[1],a=e[2],s=e[3],n=e[4];if(t.doRotary!==0&&e.length<=7)throw new Error("cos_cast and sin_cache are expected if do_rotary attribute is non-zero");if(t.localWindowSize!==-1)throw new Error("Local attention is not supported");if(t.softcap!==0)throw new Error("Softcap is not supported");if(t.rotaryInterleaved!==0)throw new Error("Rotary interleaved is not supported");if(t.smoothSoftmax)throw new Error("Smooth softmax is not supported");if(r.dims.length!==3&&r.dims.length!==5)throw new Error("Input query is expected to have 3 or 5 dimensions");let u=!1,l=r.dims[0],p=r.dims[1],h=r.dims.length===3?u?r.dims[2]/3:r.dims[2]:t.numHeads*r.dims[4],f=p,g=0,_=!i||i.dims.length===0,y=Math.floor(_?h/(t.numHeads+2*t.kvNumHeads):h/t.numHeads);_&&(h=y*t.numHeads);let $=s&&s.dims.length!==0,S=n&&n.dims.length!==0;if($&&s.dims.length===4&&s.dims[0]===l&&s.dims[1]!==t.kvNumHeads&&s.dims[2]===t.kvNumHeads&&s.dims[3]===y)throw new Error("BSNH pastKey/pastValue is not supported");if($&&S){if(s.dims.length!==4)throw new Error('Input "past_key" is expected to have 4 dimensions');if(n.dims.length!==4)throw new Error('Input "past_value" is expected to have 4 dimensions');g=s.dims[2]}else if($||S)throw new Error('Input "past_key" and "past_value" shall be both present or both absent');let v=1;if(i&&i.dims.length>0){if(r.dims.length!==3)throw new Error('Input "query" is expected to have 3 dimensions when key is given');if(i.dims.length<3||i.dims.length>5)throw new Error('Input "key" is expected to have 3, 4, or 5 dimensions');if(r.dims[0]!==i.dims[0])throw new Error('Input "query" and "key" shall have same dim 0 (batch size)');if(i.dims.length===3){if(r.dims[2]%i.dims[2]!==0)throw new Error('Dimension 2 of "query" should be a multiple of "key"');f=i.dims[1]}else if(i.dims.length===5){if(i.dims[2]!==t.numHeads||i.dims[3]!==2||i.dims[4]!==y)throw new Error('Expect "key" shape (batch_size, kv_sequence_length, num_heads, 2, head_size) for packed kv');if(a)throw new Error('Expect "value" be none when "key" has packed kv format.');f=i.dims[1]}else{if(i.dims[1]!==t.numHeads||i.dims[3]!==y)throw new Error('Expect "key" shape (batch_size, num_heads, kv_sequence_length, head_size) for past_key');f=i.dims[2]}}else{if(r.dims.length!==3&&r.dims.length!==5)throw new Error('Input "query" is expected to have 3 or 5 dimensions when key is empty');if(r.dims.length===5&&(r.dims[2]!==t.numHeads||r.dims[3]!==3))throw new Error('Expect "query" shape (batch_size, kv_sequence_length, num_heads, 3, head_size) for packed kv');v=3}let w=0,E=!1,I=t.kvNumHeads?y*t.kvNumHeads:h;if(a&&a.dims.length>0){if(a.dims.length!==3&&a.dims.length!==4)throw new Error('Input "value" is expected to have 3 or 4 dimensions');if(r.dims[0]!==a.dims[0])throw new Error('Input "query" and "value" shall have same dim 0 (batch_size)');if(a.dims.length===3){if(f!==a.dims[1])throw new Error('Input "key" and "value" shall have the same dim 1 (kv_sequence_length)');I=a.dims[2]}else{if(f!==a.dims[2])throw new Error('Input "past_key" and "past_value" shall have the same dim 2 (kv_sequence_length)');I=a.dims[1]*a.dims[3],E=!0}}let k=e.length>4?e[5]:void 0;if(k&&k.dims.length!==1&&k.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:p,pastSequenceLength:g,kvSequenceLength:f,totalSequenceLength:-1,maxSequenceLength:-1,inputHiddenSize:0,hiddenSize:h,vHiddenSize:I,headSize:y,vHeadSize:Math.floor(I/t.kvNumHeads),numHeads:t.numHeads,kvNumHeads:t.kvNumHeads,nReps:t.numHeads/t.kvNumHeads,pastPresentShareBuffer:!1,maskType:w,scale:t.scale,broadcastResPosBias:!1,passPastInKv:E,qkvFormat:v}},Il=he({perm:[0,2,1,3]}),Yi=(e,t,r)=>{let i=t,a=r.kvNumHeads;return t.dims.length===3&&r.kvSequenceLength!==0&&(i=t.reshape([r.batchSize,r.kvSequenceLength,a,r.headSize]),i=e.compute(Me(i,Il.perm),{inputs:[i],outputs:[-1]})[0]),i},El=(e,t,r,i)=>{let a=7,s=["type","type"],n=[e*t],u=e*t,l=[{type:12,data:u},{type:12,data:t},{type:12,data:e}],p=h=>{let f=N("seq_lens",r.dataType,r.dims),g=N("total_seq_lens",i.dataType,i.dims),_=F("pos_ids",a,n),y=[{name:"output_size",type:"u32"},{name:"sequence_length",type:"u32"},{name:"batch_size",type:"u32"}];return`
+        }`};return{name:"RotaryEmbedding",shaderCache:{hint:he({interleaved:r}).cacheKey,inputDependencies:["rank","rank","rank","rank"]},getShaderSource:$,getRunData:()=>({outputs:[{dims:e[0].dims,dataType:e[0].dataType}],dispatchGroup:{x:Math.ceil(B.size(g)/Vt)},programUniforms:y})}},vh=(e,t)=>{Sl(e.inputs,t),e.compute(jr(e.inputs,t))}}),Tl,Il,Yi,kl,Sh,z0=P(()=>{"use strict";ve(),Y(),Ha(),_h(),$h(),wt(),xh(),re(),Tl=(e,t)=>{if(t.doRotary&&e.length<=7)throw new Error("cos_cache and sin_cache inputs are required if do_rotary is specified");let r=e[0],i=e[1],a=e[2],s=e[3],n=e[4];if(t.doRotary!==0&&e.length<=7)throw new Error("cos_cast and sin_cache are expected if do_rotary attribute is non-zero");if(t.localWindowSize!==-1)throw new Error("Local attention is not supported");if(t.softcap!==0)throw new Error("Softcap is not supported");if(t.rotaryInterleaved!==0)throw new Error("Rotary interleaved is not supported");if(t.smoothSoftmax)throw new Error("Smooth softmax is not supported");if(r.dims.length!==3&&r.dims.length!==5)throw new Error("Input query is expected to have 3 or 5 dimensions");let u=!1,l=r.dims[0],p=r.dims[1],h=r.dims.length===3?u?r.dims[2]/3:r.dims[2]:t.numHeads*r.dims[4],f=p,g=0,_=!i||i.dims.length===0,y=Math.floor(_?h/(t.numHeads+2*t.kvNumHeads):h/t.numHeads);_&&(h=y*t.numHeads);let $=s&&s.dims.length!==0,S=n&&n.dims.length!==0;if($&&s.dims.length===4&&s.dims[0]===l&&s.dims[1]!==t.kvNumHeads&&s.dims[2]===t.kvNumHeads&&s.dims[3]===y)throw new Error("BSNH pastKey/pastValue is not supported");if($&&S){if(s.dims.length!==4)throw new Error('Input "past_key" is expected to have 4 dimensions');if(n.dims.length!==4)throw new Error('Input "past_value" is expected to have 4 dimensions');g=s.dims[2]}else if($||S)throw new Error('Input "past_key" and "past_value" shall be both present or both absent');let v=1;if(i&&i.dims.length>0){if(r.dims.length!==3)throw new Error('Input "query" is expected to have 3 dimensions when key is given');if(i.dims.length<3||i.dims.length>5)throw new Error('Input "key" is expected to have 3, 4, or 5 dimensions');if(r.dims[0]!==i.dims[0])throw new Error('Input "query" and "key" shall have same dim 0 (batch size)');if(i.dims.length===3){if(r.dims[2]%i.dims[2]!==0)throw new Error('Dimension 2 of "query" should be a multiple of "key"');f=i.dims[1]}else if(i.dims.length===5){if(i.dims[2]!==t.numHeads||i.dims[3]!==2||i.dims[4]!==y)throw new Error('Expect "key" shape (batch_size, kv_sequence_length, num_heads, 2, head_size) for packed kv');if(a)throw new Error('Expect "value" be none when "key" has packed kv format.');f=i.dims[1]}else{if(i.dims[1]!==t.numHeads||i.dims[3]!==y)throw new Error('Expect "key" shape (batch_size, num_heads, kv_sequence_length, head_size) for past_key');f=i.dims[2]}}else{if(r.dims.length!==3&&r.dims.length!==5)throw new Error('Input "query" is expected to have 3 or 5 dimensions when key is empty');if(r.dims.length===5&&(r.dims[2]!==t.numHeads||r.dims[3]!==3))throw new Error('Expect "query" shape (batch_size, kv_sequence_length, num_heads, 3, head_size) for packed kv');v=3}let w=0,k=!1,I=t.kvNumHeads?y*t.kvNumHeads:h;if(a&&a.dims.length>0){if(a.dims.length!==3&&a.dims.length!==4)throw new Error('Input "value" is expected to have 3 or 4 dimensions');if(r.dims[0]!==a.dims[0])throw new Error('Input "query" and "value" shall have same dim 0 (batch_size)');if(a.dims.length===3){if(f!==a.dims[1])throw new Error('Input "key" and "value" shall have the same dim 1 (kv_sequence_length)');I=a.dims[2]}else{if(f!==a.dims[2])throw new Error('Input "past_key" and "past_value" shall have the same dim 2 (kv_sequence_length)');I=a.dims[1]*a.dims[3],k=!0}}let E=e.length>4?e[5]:void 0;if(E&&E.dims.length!==1&&E.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:p,pastSequenceLength:g,kvSequenceLength:f,totalSequenceLength:-1,maxSequenceLength:-1,inputHiddenSize:0,hiddenSize:h,vHiddenSize:I,headSize:y,vHeadSize:Math.floor(I/t.kvNumHeads),numHeads:t.numHeads,kvNumHeads:t.kvNumHeads,nReps:t.numHeads/t.kvNumHeads,pastPresentShareBuffer:!1,maskType:w,scale:t.scale,broadcastResPosBias:!1,passPastInKv:k,qkvFormat:v}},Il=he({perm:[0,2,1,3]}),Yi=(e,t,r)=>{let i=t,a=r.kvNumHeads;return t.dims.length===3&&r.kvSequenceLength!==0&&(i=t.reshape([r.batchSize,r.kvSequenceLength,a,r.headSize]),i=e.compute(Ne(i,Il.perm),{inputs:[i],outputs:[-1]})[0]),i},kl=(e,t,r,i)=>{let a=7,s=["type","type"],n=[e*t],u=e*t,l=[{type:12,data:u},{type:12,data:t},{type:12,data:e}],p=h=>{let f=M("seq_lens",r.dataType,r.dims),g=M("total_seq_lens",i.dataType,i.dims),_=F("pos_ids",a,n),y=[{name:"output_size",type:"u32"},{name:"sequence_length",type:"u32"},{name:"batch_size",type:"u32"}];return`
   ${h.registerUniforms(y).declareVariables(f,g,_)}
   ${h.mainStart()}
     ${h.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.output_size")}
@@ -1829,7 +1829,7 @@ fn calculateOutputIndex(index: u32) -> u32 {
       ${_.setByOffset("global_idx","seqlen")}
     };
   }
-  `};return{name:"GeneratePositionIds",shaderCache:{hint:`${e};${t}`,inputDependencies:s},getRunData:()=>({outputs:[{dims:n,dataType:a}],dispatchGroup:{x:Math.ceil(u/64)},programUniforms:l}),getShaderSource:p}},Sh=(e,t)=>{let r=Tl(e.inputs,t);if(e.inputs[0].dims.length===5)throw new Error("Packed QKV is not implemented");if(e.inputs[1]?.dims.length===5)throw new Error("Packed KV is not implemented");let i=e.inputs[0],a=e.inputs[1]&&e.inputs[1].dims.length>0?e.inputs[1]:void 0,s=e.inputs[2]&&e.inputs[2].dims.length>0?e.inputs[2]:void 0,n=e.inputs[3]&&e.inputs[3].dims.length!==0?e.inputs[3]:void 0,u=e.inputs[4]&&e.inputs[4].dims.length!==0?e.inputs[4]:void 0,l=e.inputs.length>4?e.inputs[5]:void 0,p=e.inputs.length>5?e.inputs[6]:void 0,h=r.kvNumHeads?r.kvNumHeads:r.numHeads,f=he({axis:2,numOutputs:3,splitSizes:[r.numHeads*r.headSize,h*r.headSize,h*r.headSize]}),[g,_,y]=!a&&!s?e.compute(za([i],f),{inputs:[i],outputs:[-1,-1,-1]}):[i,a,s],$,S;if(t.doRotary){let I=e.compute(El(r.batchSize,r.sequenceLength,l,p),{inputs:[l,p],outputs:[-1]})[0],k=e.inputs[7],C=e.inputs[8],O=he({interleaved:t.rotaryInterleaved!==0,numHeads:r.numHeads,rotaryEmbeddingDim:0,scale:t.scale}),x=[g,I,k,C],D=[-1];$=e.compute(jr(x,O),{inputs:x,outputs:D})[0],x.splice(0,1,_);let q=he({interleaved:t.rotaryInterleaved!==0,numHeads:r.kvNumHeads,rotaryEmbeddingDim:0,scale:t.scale});S=e.compute(jr(x,q),{inputs:x,outputs:D})[0]}let v=lr(e,r.batchSize,r.numHeads,r.sequenceLength,r.headSize,t.doRotary?$:g,void 0,0),w=Yi(e,t.doRotary?S:_,r),E=Yi(e,y,r);hr(e,v,w,E,void 0,void 0,n,u,void 0,r,l,p)}}),Ji,kl,zl,Th,C0=P(()=>{"use strict";Y(),te(),wt(),re(),Ji=(e,t,r,i,a,s,n,u)=>{let l=$e(s),p=l===1?"f32":`vec${l}f`,h=l===1?"vec2f":`mat2x${l}f`,f=a*n,g=64;f===1&&(g=256);let _=[a,n,s/l],y=[a,n,2],$=["rank","type","type"],S=[];S.push(...Z(_,y));let v=w=>{let E=N("x",t.dataType,3,l),I=N("scale",r.dataType,r.dims),k=N("bias",i.dataType,i.dims),C=F("output",1,3,2),O=[E,I,k,C];return`
+  `};return{name:"GeneratePositionIds",shaderCache:{hint:`${e};${t}`,inputDependencies:s},getRunData:()=>({outputs:[{dims:n,dataType:a}],dispatchGroup:{x:Math.ceil(u/64)},programUniforms:l}),getShaderSource:p}},Sh=(e,t)=>{let r=Tl(e.inputs,t);if(e.inputs[0].dims.length===5)throw new Error("Packed QKV is not implemented");if(e.inputs[1]?.dims.length===5)throw new Error("Packed KV is not implemented");let i=e.inputs[0],a=e.inputs[1]&&e.inputs[1].dims.length>0?e.inputs[1]:void 0,s=e.inputs[2]&&e.inputs[2].dims.length>0?e.inputs[2]:void 0,n=e.inputs[3]&&e.inputs[3].dims.length!==0?e.inputs[3]:void 0,u=e.inputs[4]&&e.inputs[4].dims.length!==0?e.inputs[4]:void 0,l=e.inputs.length>4?e.inputs[5]:void 0,p=e.inputs.length>5?e.inputs[6]:void 0,h=r.kvNumHeads?r.kvNumHeads:r.numHeads,f=he({axis:2,numOutputs:3,splitSizes:[r.numHeads*r.headSize,h*r.headSize,h*r.headSize]}),[g,_,y]=!a&&!s?e.compute(za([i],f),{inputs:[i],outputs:[-1,-1,-1]}):[i,a,s],$,S;if(t.doRotary){let I=e.compute(kl(r.batchSize,r.sequenceLength,l,p),{inputs:[l,p],outputs:[-1]})[0],E=e.inputs[7],C=e.inputs[8],O=he({interleaved:t.rotaryInterleaved!==0,numHeads:r.numHeads,rotaryEmbeddingDim:0,scale:t.scale}),x=[g,I,E,C],D=[-1];$=e.compute(jr(x,O),{inputs:x,outputs:D})[0],x.splice(0,1,_);let L=he({interleaved:t.rotaryInterleaved!==0,numHeads:r.kvNumHeads,rotaryEmbeddingDim:0,scale:t.scale});S=e.compute(jr(x,L),{inputs:x,outputs:D})[0]}let v=lr(e,r.batchSize,r.numHeads,r.sequenceLength,r.headSize,t.doRotary?$:g,void 0,0),w=Yi(e,t.doRotary?S:_,r),k=Yi(e,y,r);hr(e,v,w,k,void 0,void 0,n,u,void 0,r,l,p)}}),Ji,El,zl,Th,C0=P(()=>{"use strict";Y(),te(),wt(),re(),Ji=(e,t,r,i,a,s,n,u)=>{let l=$e(s),p=l===1?"f32":`vec${l}f`,h=l===1?"vec2f":`mat2x${l}f`,f=a*n,g=64;f===1&&(g=256);let _=[a,n,s/l],y=[a,n,2],$=["rank","type","type"],S=[];S.push(...Z(_,y));let v=w=>{let k=M("x",t.dataType,3,l),I=M("scale",r.dataType,r.dims),E=M("bias",i.dataType,i.dims),C=F("output",1,3,2),O=[k,I,E,C];return`
   var<workgroup> workgroup_shared : array<${h}, ${g}>;
   const workgroup_size = ${g}u;
   ${w.declareVariables(...O)}
@@ -1841,7 +1841,7 @@ fn calculateOutputIndex(index: u32) -> u32 {
     var sum = ${p}(0);
     var squared_sum = ${p}(0);
     for (var h = local_idx; h < hight; h += workgroup_size) {
-      let value = ${p}(${E.get("batch","channel","h")});
+      let value = ${p}(${k.get("batch","channel","h")});
       sum += value;
       squared_sum += value * value;
     }
@@ -1863,19 +1863,19 @@ fn calculateOutputIndex(index: u32) -> u32 {
       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:`${l};${u};${g}`,inputDependencies:$},getRunData:()=>({outputs:[{dims:y,dataType:1}],dispatchGroup:{x:f},programUniforms:S}),getShaderSource:v},{inputs:[t,r,i],outputs:[-1]})[0]},kl=(e,t,r)=>{let i=t[0].dims,a=i,s=2,n=i[0],u=i[1],l=B.sizeFromDimension(i,s),p=$e(l),h=B.size(a)/p,f=Ji(e,t[0],t[1],t[2],n,l,u,r.epsilon),g=[n,u,l/p],_=[n,u],y=["type","none"],$=S=>{let v=N("x",t[0].dataType,g.length,p),w=N("scale_shift",1,_.length,2),E=F("output",t[0].dataType,g.length,p),I=[v,w,E];return`
+  }`};return e.compute({name:"InstanceNormComputeChannelScaleShift",shaderCache:{hint:`${l};${u};${g}`,inputDependencies:$},getRunData:()=>({outputs:[{dims:y,dataType:1}],dispatchGroup:{x:f},programUniforms:S}),getShaderSource:v},{inputs:[t,r,i],outputs:[-1]})[0]},El=(e,t,r)=>{let i=t[0].dims,a=i,s=2,n=i[0],u=i[1],l=B.sizeFromDimension(i,s),p=$e(l),h=B.size(a)/p,f=Ji(e,t[0],t[1],t[2],n,l,u,r.epsilon),g=[n,u,l/p],_=[n,u],y=["type","none"],$=S=>{let v=M("x",t[0].dataType,g.length,p),w=M("scale_shift",1,_.length,2),k=F("output",t[0].dataType,g.length,p),I=[v,w,k];return`
   ${S.registerUniform("output_size","u32").declareVariables(...I)}
   ${S.mainStart()}
   ${S.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.output_size")}
-      let outputIndices = ${E.offsetToIndices("global_idx")};
+      let outputIndices = ${k.offsetToIndices("global_idx")};
       let batch = outputIndices[0];
       let channel = outputIndices[1];
       let scale_shift = ${w.getByIndices("vec2<u32>(batch, channel)")};
-      let value = ${v.getByOffset("global_idx")} * ${E.type.value}(scale_shift.x) + ${E.type.value}(scale_shift.y);
-      ${E.setByOffset("global_idx","value")};
-  }`};e.compute({name:"InstanceNormalization",shaderCache:{hint:`${p}`,inputDependencies:y},getRunData:()=>({outputs:[{dims:a,dataType:t[0].dataType}],dispatchGroup:{x:Math.ceil(h/64)},programUniforms:[{type:12,data:h},...Z(g,_,g)]}),getShaderSource:$},{inputs:[t[0],f]})},zl=(e,t,r)=>{let i=t[0].dims,a=i,s=i[0],n=i[i.length-1],u=B.sizeFromDimension(i,1)/n,l=$e(n),p=B.size(a)/l,h=[{type:12,data:u},{type:12,data:Math.floor(n/l)}],f=["type","type"],g=!1,_=[0,i.length-1];for(let v=0;v<i.length-2;v++)g=g||i[v+1]!==1,_.push(v+1);g=g&&i[i.length-1]!==1;let y=g?e.compute(Me(e.inputs[0],_),{inputs:[e.inputs[0]],outputs:[-1]})[0]:e.inputs[0].reshape(Array.from({length:i.length},(v,w)=>i[_[w]])),$=Ji(e,y,t[1],t[2],s,u,n,r.epsilon),S=v=>{let w=Ee(t[0].dataType),E=l===1?"vec2f":`mat${l}x2f`,I=O=>{let x=O===0?"x":"y",D=l===1?"f32":`vec${l}f`;switch(l){case 1:return`${w}(${D}(scale.${x}))`;case 2:return`vec2<${w}>(${D}(scale[0].${x}, scale[1].${x}))`;case 4:return`vec4<${w}>(${D}(scale[0].${x}, scale[1].${x}, scale[2].${x}, scale[3].${x}))`;default:throw new Error(`Not supported compoents ${l}`)}},k=N("input",t[0].dataType,t[0].dims,l),C=F("output",t[0].dataType,a,l);return`
-  @group(0) @binding(0) var<storage, read> input : array<${k.type.storage}>;
-  @group(0) @binding(1) var<storage, read> scale_input : array<${E}>;
+      let value = ${v.getByOffset("global_idx")} * ${k.type.value}(scale_shift.x) + ${k.type.value}(scale_shift.y);
+      ${k.setByOffset("global_idx","value")};
+  }`};e.compute({name:"InstanceNormalization",shaderCache:{hint:`${p}`,inputDependencies:y},getRunData:()=>({outputs:[{dims:a,dataType:t[0].dataType}],dispatchGroup:{x:Math.ceil(h/64)},programUniforms:[{type:12,data:h},...Z(g,_,g)]}),getShaderSource:$},{inputs:[t[0],f]})},zl=(e,t,r)=>{let i=t[0].dims,a=i,s=i[0],n=i[i.length-1],u=B.sizeFromDimension(i,1)/n,l=$e(n),p=B.size(a)/l,h=[{type:12,data:u},{type:12,data:Math.floor(n/l)}],f=["type","type"],g=!1,_=[0,i.length-1];for(let v=0;v<i.length-2;v++)g=g||i[v+1]!==1,_.push(v+1);g=g&&i[i.length-1]!==1;let y=g?e.compute(Ne(e.inputs[0],_),{inputs:[e.inputs[0]],outputs:[-1]})[0]:e.inputs[0].reshape(Array.from({length:i.length},(v,w)=>i[_[w]])),$=Ji(e,y,t[1],t[2],s,u,n,r.epsilon),S=v=>{let w=ke(t[0].dataType),k=l===1?"vec2f":`mat${l}x2f`,I=O=>{let x=O===0?"x":"y",D=l===1?"f32":`vec${l}f`;switch(l){case 1:return`${w}(${D}(scale.${x}))`;case 2:return`vec2<${w}>(${D}(scale[0].${x}, scale[1].${x}))`;case 4:return`vec4<${w}>(${D}(scale[0].${x}, scale[1].${x}, scale[2].${x}, scale[3].${x}))`;default:throw new Error(`Not supported compoents ${l}`)}},E=M("input",t[0].dataType,t[0].dims,l),C=F("output",t[0].dataType,a,l);return`
+  @group(0) @binding(0) var<storage, read> input : array<${E.type.storage}>;
+  @group(0) @binding(1) var<storage, read> scale_input : array<${k}>;
   @group(0) @binding(2) var<storage, read_write> output : array<${C.type.storage}>;
   struct Uniforms {H: u32, C : u32};
   @group(0) @binding(3) var<uniform> uniforms: Uniforms;
@@ -1887,18 +1887,18 @@ fn calculateOutputIndex(index: u32) -> u32 {
     let scale_offset = current_image_number * uniforms.C + current_channel_number;
     let scale = scale_input[scale_offset];
     output[global_idx] = fma(input[global_idx], ${I(0)}, ${I(1)});
-  }`};e.compute({name:"InstanceNormalizationNHWC",shaderCache:{hint:`${l}`,inputDependencies:f},getRunData:()=>({outputs:[{dims:a,dataType:t[0].dataType}],dispatchGroup:{x:Math.ceil(p/64)},programUniforms:h}),getShaderSource:S},{inputs:[t[0],$]})},Th=(e,t)=>{t.format==="NHWC"?zl(e,e.inputs,t):kl(e,e.inputs,t)}}),Cl,Al,Ih,A0=P(()=>{"use strict";Y(),te(),re(),Cl=e=>{if(!e||e.length<2)throw new Error("layerNorm requires at least 2 inputs.")},Al=(e,t,r)=>{let i=t.simplified,a=e[0].dims,s=e[1],n=!i&&e[2],u=a,l=B.normalizeAxis(t.axis,a.length),p=B.sizeToDimension(a,l),h=B.sizeFromDimension(a,l),f=B.size(s.dims),g=n?B.size(n.dims):0;if(f!==h||n&&g!==h)throw new Error(`Size of X.shape()[axis:] == ${h}.
+  }`};e.compute({name:"InstanceNormalizationNHWC",shaderCache:{hint:`${l}`,inputDependencies:f},getRunData:()=>({outputs:[{dims:a,dataType:t[0].dataType}],dispatchGroup:{x:Math.ceil(p/64)},programUniforms:h}),getShaderSource:S},{inputs:[t[0],$]})},Th=(e,t)=>{t.format==="NHWC"?zl(e,e.inputs,t):El(e,e.inputs,t)}}),Cl,Al,Ih,A0=P(()=>{"use strict";Y(),te(),re(),Cl=e=>{if(!e||e.length<2)throw new Error("layerNorm requires at least 2 inputs.")},Al=(e,t,r)=>{let i=t.simplified,a=e[0].dims,s=e[1],n=!i&&e[2],u=a,l=B.normalizeAxis(t.axis,a.length),p=B.sizeToDimension(a,l),h=B.sizeFromDimension(a,l),f=B.size(s.dims),g=n?B.size(n.dims):0;if(f!==h||n&&g!==h)throw new Error(`Size of X.shape()[axis:] == ${h}.
        Size of scale and bias (if provided) must match this.
-       Got scale size of ${f} and bias size of ${g}`);let _=[];for(let k=0;k<a.length;++k)k<l?_.push(a[k]):_.push(1);let y=$e(h),$=["type","type"],S=[{type:12,data:p},{type:1,data:h},{type:12,data:Math.floor(h/y)},{type:1,data:t.epsilon}];n&&$.push("type");let v=r>1,w=r>2,E=k=>{let C=Ee(e[0].dataType),O=[N("x",e[0].dataType,e[0].dims,y),N("scale",s.dataType,s.dims,y)];n&&O.push(N("bias",n.dataType,n.dims,y)),O.push(F("output",e[0].dataType,u,y)),v&&O.push(F("mean_data_output",1,_)),w&&O.push(F("inv_std_output",1,_));let x=[{name:"norm_count",type:"u32"},{name:"norm_size",type:"f32"},{name:"norm_size_vectorized",type:"u32"},{name:"epsilon",type:"f32"}];return`
-  ${k.registerUniforms(x).declareVariables(...O)}
-  ${k.mainStart()}
-    ${k.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.norm_count")}
+       Got scale size of ${f} and bias size of ${g}`);let _=[];for(let E=0;E<a.length;++E)E<l?_.push(a[E]):_.push(1);let y=$e(h),$=["type","type"],S=[{type:12,data:p},{type:1,data:h},{type:12,data:Math.floor(h/y)},{type:1,data:t.epsilon}];n&&$.push("type");let v=r>1,w=r>2,k=E=>{let C=ke(e[0].dataType),O=[M("x",e[0].dataType,e[0].dims,y),M("scale",s.dataType,s.dims,y)];n&&O.push(M("bias",n.dataType,n.dims,y)),O.push(F("output",e[0].dataType,u,y)),v&&O.push(F("mean_data_output",1,_)),w&&O.push(F("inv_std_output",1,_));let x=[{name:"norm_count",type:"u32"},{name:"norm_size",type:"f32"},{name:"norm_size_vectorized",type:"u32"},{name:"epsilon",type:"f32"}];return`
+  ${E.registerUniforms(x).declareVariables(...O)}
+  ${E.mainStart()}
+    ${E.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.norm_count")}
     let offset = global_idx * uniforms.norm_size_vectorized;
     var mean_vector = ${ba("f32",y)};
     var mean_square_vector = ${ba("f32",y)};
 
     for (var h: u32 = 0u; h < uniforms.norm_size_vectorized; h++) {
-      let value = ${qt(C,y,"x[h + offset]")};
+      let value = ${Lt(C,y,"x[h + offset]")};
       mean_vector += value;
       mean_square_vector += value * value;
     }
@@ -1906,16 +1906,16 @@ fn calculateOutputIndex(index: u32) -> u32 {
     let inv_std_dev = inverseSqrt(${yt("mean_square_vector",y)} / uniforms.norm_size ${i?"":"- mean * mean"} + uniforms.epsilon);
 
     for (var j: u32 = 0; j < uniforms.norm_size_vectorized; j++) {
-      let f32input = ${qt(C,y,"x[j + offset]")};
-      let f32scale = ${qt(C,y,"scale[j]")};
+      let f32input = ${Lt(C,y,"x[j + offset]")};
+      let f32scale = ${Lt(C,y,"scale[j]")};
       output[j + offset] = ${O[0].type.value}((f32input ${i?"":"- mean"}) * inv_std_dev * f32scale
-        ${n?`+ ${qt(C,y,"bias[j]")}`:""}
+        ${n?`+ ${Lt(C,y,"bias[j]")}`:""}
       );
     }
 
     ${v?"mean_data_output[global_idx] = mean":""};
     ${w?"inv_std_output[global_idx] = inv_std_dev":""};
-  }`},I=[{dims:u,dataType:e[0].dataType}];return v&&I.push({dims:_,dataType:1}),w&&I.push({dims:_,dataType:1}),{name:"LayerNormalization",shaderCache:{hint:`${y};${r};${i}`,inputDependencies:$},getRunData:()=>({outputs:I,dispatchGroup:{x:Math.ceil(p/64)},programUniforms:S}),getShaderSource:E}},Ih=(e,t)=>{Cl(e.inputs),e.compute(Al(e.inputs,t,e.outputCount))}}),Ol,Eh,O0=P(()=>{"use strict";te(),Xa(),Ya(),Ol=e=>{if(!e||e.length!==2)throw new Error("MatMul requires 2 inputs.");if(e[0].dims[e[0].dims.length-1]!==e[1].dims[e[1].dims.length-2])throw new Error("shared dimension does not match.")},Eh=e=>{Ol(e.inputs);let t=Lt.calcShape(e.inputs[0].dims,e.inputs[1].dims,!0);if(!t)throw new Error("Can't use matmul on the given tensors");let r=t[t.length-1],i=e.inputs[0].dims[e.inputs[0].dims.length-1];if(r<8&&i<8)e.compute(Qa(e.inputs,{activation:""},t));else{let a=t[t.length-2],s=B.size(e.inputs[0].dims.slice(0,-2)),n=B.size(e.inputs[1].dims.slice(0,-2));if(s!==1&&a===1&&n===1){let u=e.inputs[0].reshape([1,s,i]),l=e.inputs[1].reshape([1,i,r]),p=[1,s,r],h=[u,l];e.compute(Hr(h,{activation:""},t,p),{inputs:h})}else e.compute(Hr(e.inputs,{activation:""},t))}}}),Bl,Rl,Nl,kh,zh,B0=P(()=>{"use strict";Y(),te(),ve(),re(),Bl=(e,t)=>{if(e.length<3||e.length>4)throw new Error("MatMulNBits requires 3 or 4 inputs");let r=e[0],i=r.dims.length;if(r.dims[i-1]!==t.k)throw new Error("The last dim of input shape does not match the k value");let a=Math.floor((t.k+t.blockSize-1)/t.blockSize),s=t.blockSize/8*t.bits,n=e[1];if(!B.areEqual(n.dims,[t.n,a,s]))throw new Error("The second inputs must be 3D tensor with shape N X nBlocksPerCol X blobSize");let u=e[2].dims;if(B.size(u)!==t.n*a)throw new Error("scales input size error.");if(e.length===4){let l=e[3].dims,p=t.n*(t.bits===8?a:Math.floor((a*t.bits+7)/8));if(B.size(l)!==p)throw new Error("zeroPoints input size error.")}},Rl=(e,t)=>{let r=e[0].dims,i=r.length,a=r[i-2],s=t.k,n=t.n,u=r.slice(0,i-2),l=B.size(u),p=e[1].dims[2]/4,h=e[0].dataType,f=$e(t.k),g=$e(p),_=$e(n),y=u.concat([a,n]),$=a>1&&n/_%2===0?2:1,S=B.size(y)/_/$,v=64,w=[],E=[l,a,s/f],I=B.convertShape(e[1].dims).slice();I.splice(-1,1,p/g),w.push(...Z(E)),w.push(...Z(I)),w.push(...Z(e[2].dims)),e.length===4&&w.push(...Z(B.convertShape(e[3].dims)));let k=[l,a,n/_];w.push(...Z(k));let C=O=>{let x=E.length,D=N("a",e[0].dataType,x,f),q=N("b",12,I.length,g),Q=N("scales",e[2].dataType,e[2].dims.length),L=[D,q,Q],K=e.length===4?N("zero_points",12,e[3].dims.length):void 0;K&&L.push(K);let se=k.length,A=F("output",e[0].dataType,se,_),U=Ee(e[0].dataType),ee=(()=>{switch(f){case 1:return`array<${U}, 8>`;case 2:return`mat4x2<${U}>`;case 4:return`mat2x4<${U}>`;default:throw new Error(`${f}-component is not supported.`)}})(),X=()=>{let M=`
+  }`},I=[{dims:u,dataType:e[0].dataType}];return v&&I.push({dims:_,dataType:1}),w&&I.push({dims:_,dataType:1}),{name:"LayerNormalization",shaderCache:{hint:`${y};${r};${i}`,inputDependencies:$},getRunData:()=>({outputs:I,dispatchGroup:{x:Math.ceil(p/64)},programUniforms:S}),getShaderSource:k}},Ih=(e,t)=>{Cl(e.inputs),e.compute(Al(e.inputs,t,e.outputCount))}}),Ol,kh,O0=P(()=>{"use strict";te(),Xa(),Ya(),Ol=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.")},kh=e=>{Ol(e.inputs);let t=qt.calcShape(e.inputs[0].dims,e.inputs[1].dims,!0);if(!t)throw new Error("Can't use matmul on the given tensors");let r=t[t.length-1],i=e.inputs[0].dims[e.inputs[0].dims.length-1];if(r<8&&i<8)e.compute(Qa(e.inputs,{activation:""},t));else{let a=t[t.length-2],s=B.size(e.inputs[0].dims.slice(0,-2)),n=B.size(e.inputs[1].dims.slice(0,-2));if(s!==1&&a===1&&n===1){let u=e.inputs[0].reshape([1,s,i]),l=e.inputs[1].reshape([1,i,r]),p=[1,s,r],h=[u,l];e.compute(Hr(h,{activation:""},t,p),{inputs:h})}else e.compute(Hr(e.inputs,{activation:""},t))}}}),Bl,Rl,Ml,Eh,zh,B0=P(()=>{"use strict";Y(),te(),ve(),re(),Bl=(e,t)=>{if(e.length<3||e.length>4)throw new Error("MatMulNBits requires 3 or 4 inputs");let r=e[0],i=r.dims.length;if(r.dims[i-1]!==t.k)throw new Error("The last dim of input shape does not match the k value");let a=Math.floor((t.k+t.blockSize-1)/t.blockSize),s=t.blockSize/8*t.bits,n=e[1];if(!B.areEqual(n.dims,[t.n,a,s]))throw new Error("The second inputs must be 3D tensor with shape N X nBlocksPerCol X blobSize");let u=e[2].dims;if(B.size(u)!==t.n*a)throw new Error("scales input size error.");if(e.length===4){let l=e[3].dims,p=t.n*(t.bits===8?a:Math.floor((a*t.bits+7)/8));if(B.size(l)!==p)throw new Error("zeroPoints input size error.")}},Rl=(e,t)=>{let r=e[0].dims,i=r.length,a=r[i-2],s=t.k,n=t.n,u=r.slice(0,i-2),l=B.size(u),p=e[1].dims[2]/4,h=e[0].dataType,f=$e(t.k),g=$e(p),_=$e(n),y=u.concat([a,n]),$=a>1&&n/_%2===0?2:1,S=B.size(y)/_/$,v=64,w=[],k=[l,a,s/f],I=B.convertShape(e[1].dims).slice();I.splice(-1,1,p/g),w.push(...Z(k)),w.push(...Z(I)),w.push(...Z(e[2].dims)),e.length===4&&w.push(...Z(B.convertShape(e[3].dims)));let E=[l,a,n/_];w.push(...Z(E));let C=O=>{let x=k.length,D=M("a",e[0].dataType,x,f),L=M("b",12,I.length,g),Q=M("scales",e[2].dataType,e[2].dims.length),q=[D,L,Q],K=e.length===4?M("zero_points",12,e[3].dims.length):void 0;K&&q.push(K);let se=E.length,A=F("output",e[0].dataType,se,_),U=ke(e[0].dataType),ee=(()=>{switch(f){case 1:return`array<${U}, 8>`;case 2:return`mat4x2<${U}>`;case 4:return`mat2x4<${U}>`;default:throw new Error(`${f}-component is not supported.`)}})(),X=()=>{let N=`
           // reuse a data
             var input_offset = ${D.indicesToOffset(`${D.type.indices}(batch, row, word_offset)`)};
             var a_data: ${ee};
@@ -1923,14 +1923,14 @@ fn calculateOutputIndex(index: u32) -> u32 {
               a_data[j] = ${D.getByOffset("input_offset")};
               input_offset++;
             }
-          `;for(let V=0;V<_*$;V++)M+=`
+          `;for(let V=0;V<_*$;V++)N+=`
             b_value = ${g===1?`b${V}_data`:`b${V}_data[i]`};
             b_value_lower = unpack4xU8(b_value & b_mask);
             b_value_upper = unpack4xU8((b_value >> 4) & b_mask);
             b_quantized_values = ${ee}(${Array.from({length:4},(J,de)=>`${U}(b_value_lower[${de}]), ${U}(b_value_upper[${de}])`).join(", ")});
             b_dequantized_values = ${f===1?`${ee}(${Array.from({length:8},(J,de)=>`(b_quantized_values[${de}] - ${K?`zero_point${V}`:"zero_point"}) * scale${V}`).join(", ")});`:`(b_quantized_values - ${ee}(${Array(8).fill(`${K?`zero_point${V}`:"zero_point"}`).join(",")})) * scale${V};`};
             workgroup_shared[local_id.x * ${$} + ${Math.floor(V/_)}]${_>1?`[${V%_}]`:""} += ${Array.from({length:8/f},(J,de)=>`${f===1?`a_data[${de}] * b_dequantized_values[${de}]`:`dot(a_data[${de}], b_dequantized_values[${de}])`}`).join(" + ")};
-          `;return M},H=()=>{let M=`
+          `;return N},H=()=>{let N=`
             var col_index = col * ${_};
             ${K?`
             let zero_point_bytes_per_col = (nBlocksPerCol + 1) / 2;
@@ -1942,7 +1942,7 @@ fn calculateOutputIndex(index: u32) -> u32 {
             var zero_point_word: u32;`:`
             // The default zero point is 8 for unsigned 4-bit quantization.
             let zero_point = ${U}(8);`}
-            `;for(let V=0;V<_*$;V++)M+=`
+            `;for(let V=0;V<_*$;V++)N+=`
             let scale${V} = ${Q.getByOffset("col_index * nBlocksPerCol + block")};
             ${K?`
             zero_point_byte_count = col_index * zero_point_bytes_per_col + (block >> 0x1u);
@@ -1951,17 +1951,17 @@ fn calculateOutputIndex(index: u32) -> u32 {
             zero_point_bits_offset = (zero_point_byte_offset << 3) + (zero_point_nibble_offset << 2);
             zero_point_word = ${K.getByOffset("zero_point_word_index")} >> zero_point_bits_offset;
             let zero_point${V} = ${U}((zero_point_word) & 0xFu);`:""}
-            col_index += 1;`;return M},le=()=>{let M=`col_index = col * ${_};`;for(let V=0;V<_*$;V++)M+=`
-            let b${V}_data = ${q.getByIndices(`${q.type.indices}(col_index, block, word)`)};
-            col_index += 1;`;return M+=`
+            col_index += 1;`;return N},le=()=>{let N=`col_index = col * ${_};`;for(let V=0;V<_*$;V++)N+=`
+            let b${V}_data = ${L.getByIndices(`${L.type.indices}(col_index, block, word)`)};
+            col_index += 1;`;return N+=`
             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: ${ee};
-            var b_dequantized_values: ${ee};`,M};return`
+            var b_dequantized_values: ${ee};`,N};return`
         var<workgroup> workgroup_shared: array<${A.type.value}, ${$*v}>;
-        ${O.declareVariables(...L,A)}
+        ${O.declareVariables(...q,A)}
         ${O.mainStart([v,1,1])}
           let output_indices = ${A.offsetToIndices(`(global_idx / ${v}) * ${$}`)};
           let col = output_indices[2];
@@ -1992,23 +1992,23 @@ fn calculateOutputIndex(index: u32) -> u32 {
             }
             ${A.setByIndices(`${A.type.indices}(batch, row, col + local_id.x)`,"output_value")};
           }
-        }`};return{name:"MatMulNBits",shaderCache:{hint:`${t.blockSize};${t.bits};${f};${g};${_};${$};${v}`,inputDependencies:Array(e.length).fill("rank")},getRunData:()=>({outputs:[{dims:y,dataType:h}],dispatchGroup:{x:S},programUniforms:w}),getShaderSource:C}},Nl=(e,t)=>{let r=e[0].dims,i=r.length,a=r[i-2],s=t.k,n=t.n,u=r.slice(0,i-2),l=B.size(u),p=e[1].dims[2]/4,h=e[0].dataType,f=$e(t.k),g=$e(p),_=u.concat([a,n]),y=128,$=n%8===0?8:n%4===0?4:1,S=y/$,v=S*g*8,w=v/f,E=v/t.blockSize,I=B.size(_)/$,k=[],C=[l,a,s/f],O=B.convertShape(e[1].dims).slice();O.splice(-1,1,p/g),k.push(...Z(C)),k.push(...Z(O)),k.push(...Z(e[2].dims)),e.length===4&&k.push(...Z(B.convertShape(e[3].dims)));let x=[l,a,n];k.push(...Z(x));let D=q=>{let Q=C.length,L=N("a",e[0].dataType,Q,f),K=N("b",12,O.length,g),se=N("scales",e[2].dataType,e[2].dims.length),A=[L,K,se],U=e.length===4?N("zero_points",12,e[3].dims.length):void 0;U&&A.push(U);let ee=x.length,X=F("output",e[0].dataType,ee),H=Ee(e[0].dataType),le=()=>{switch(f){case 1:return`
+        }`};return{name:"MatMulNBits",shaderCache:{hint:`${t.blockSize};${t.bits};${f};${g};${_};${$};${v}`,inputDependencies:Array(e.length).fill("rank")},getRunData:()=>({outputs:[{dims:y,dataType:h}],dispatchGroup:{x:S},programUniforms:w}),getShaderSource:C}},Ml=(e,t)=>{let r=e[0].dims,i=r.length,a=r[i-2],s=t.k,n=t.n,u=r.slice(0,i-2),l=B.size(u),p=e[1].dims[2]/4,h=e[0].dataType,f=$e(t.k),g=$e(p),_=u.concat([a,n]),y=128,$=n%8===0?8:n%4===0?4:1,S=y/$,v=S*g*8,w=v/f,k=v/t.blockSize,I=B.size(_)/$,E=[],C=[l,a,s/f],O=B.convertShape(e[1].dims).slice();O.splice(-1,1,p/g),E.push(...Z(C)),E.push(...Z(O)),E.push(...Z(e[2].dims)),e.length===4&&E.push(...Z(B.convertShape(e[3].dims)));let x=[l,a,n];E.push(...Z(x));let D=L=>{let Q=C.length,q=M("a",e[0].dataType,Q,f),K=M("b",12,O.length,g),se=M("scales",e[2].dataType,e[2].dims.length),A=[q,K,se],U=e.length===4?M("zero_points",12,e[3].dims.length):void 0;U&&A.push(U);let ee=x.length,X=F("output",e[0].dataType,ee),H=ke(e[0].dataType),le=()=>{switch(f){case 1:return`
           let a_data0 = vec4<${H}>(sub_a[word_offset], sub_a[word_offset + 1], sub_a[word_offset + 2], sub_a[word_offset + 3]);
           let a_data1 = vec4<${H}>(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<${H}>(sub_a[word_offset], sub_a[word_offset + 1]);
           let a_data1 = vec4<${H}>(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(`${f}-component is not supported.`)}};return`
-        var<workgroup> sub_a: array<${L.type.value}, ${w}>;
+        var<workgroup> sub_a: array<${q.type.value}, ${w}>;
         var<workgroup> inter_results: array<array<${X.type.value}, ${S}>, ${$}>;
-        ${q.declareVariables(...A,X)}
-        ${q.mainStart([S,$,1])}
+        ${L.declareVariables(...A,X)}
+        ${L.mainStart([S,$,1])}
           let output_indices = ${X.offsetToIndices(`workgroup_index * ${$}`)};
           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) / ${E} + 1;
+          let num_tiles =  (n_blocks_per_col - 1) / ${k} + 1;
 
           // Loop over shared dimension.
           for (var tile: u32 = 0; tile < num_tiles; tile += 1) {
@@ -2019,16 +2019,16 @@ fn calculateOutputIndex(index: u32) -> u32 {
               let a_col = a_col_start + a_offset;
               if (a_col < uniforms.a_shape[2])
               {
-                sub_a[a_offset] = ${L.getByIndices(`${L.type.indices}(batch, row, a_col)`)};
+                sub_a[a_offset] = ${q.getByIndices(`${q.type.indices}(batch, row, a_col)`)};
               } else {
-                sub_a[a_offset] = ${L.type.value}(0);
+                sub_a[a_offset] = ${q.type.value}(0);
               }
             }
             workgroupBarrier();
 
             // each thread process one block
             let b_row = col + local_id.y;
-            let block = tile * ${E} + local_id.x;
+            let block = tile * ${k} + local_id.x;
             ${U?`
             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);
@@ -2048,9 +2048,9 @@ fn calculateOutputIndex(index: u32) -> u32 {
               let b_value = ${g===1?"b_data":"b_data[i]"};
               let b_value_lower = unpack4xU8(b_value & 0x0F0F0F0Fu);
               let b_value_upper = unpack4xU8((b_value >> 4) & 0x0F0F0F0Fu);
-              let b_quantized_values = mat2x4<${H}>(${Array.from({length:4},(M,V)=>`${H}(b_value_lower[${V}]), ${H}(b_value_upper[${V}])`).join(", ")});
+              let b_quantized_values = mat2x4<${H}>(${Array.from({length:4},(N,V)=>`${H}(b_value_lower[${V}]), ${H}(b_value_upper[${V}])`).join(", ")});
               let b_dequantized_values = (b_quantized_values - mat2x4<${H}>(${Array(8).fill("zero_point").join(",")})) * scale;
-              inter_results[local_id.y][local_id.x] += ${Array.from({length:2},(M,V)=>`${`dot(a_data${V}, b_dequantized_values[${V}])`}`).join(" + ")};
+              inter_results[local_id.y][local_id.x] += ${Array.from({length:2},(N,V)=>`${`dot(a_data${V}, b_dequantized_values[${V}])`}`).join(" + ")};
               word_offset += ${8/f};
             }
             workgroupBarrier();
@@ -2066,7 +2066,7 @@ fn calculateOutputIndex(index: u32) -> u32 {
               ${X.setByIndices(`${X.type.indices}(batch, row, col + local_idx)`,"output_value")}
             }
           }
-        }`};return{name:"BlockwiseMatMulNBits32",shaderCache:{hint:`${t.blockSize};${f};${g};${S};${$}`,inputDependencies:Array(e.length).fill("rank")},getRunData:()=>({outputs:[{dims:_,dataType:h}],dispatchGroup:{x:I},programUniforms:k}),getShaderSource:D}},kh=(e,t)=>{Bl(e.inputs,t),t.blockSize===32&&e.adapterInfo.isVendor("intel")&&e.adapterInfo.isArchitecture("gen-12lp")?e.compute(Nl(e.inputs,t)):e.compute(Rl(e.inputs,t))},zh=e=>he(e)}),Ml,Dl,Pl,Ul,Wl,ql,Ll,Vl,Ch,R0=P(()=>{"use strict";Y(),te(),re(),Ml=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 t=e[0].dims.length*2===e[1].dims[0];if(e.length===4&&(t=e[3].dims[0]*2===e[1].dims[0]),!t)throw new Error("The pads should be a 1D tensor of shape [2 * input_rank] or [2 * num_axes].")}},Dl=(e,t,r)=>{let i="";for(let a=t-1;a>=0;--a)i+=`
+        }`};return{name:"BlockwiseMatMulNBits32",shaderCache:{hint:`${t.blockSize};${f};${g};${S};${$}`,inputDependencies:Array(e.length).fill("rank")},getRunData:()=>({outputs:[{dims:_,dataType:h}],dispatchGroup:{x:I},programUniforms:E}),getShaderSource:D}},Eh=(e,t)=>{Bl(e.inputs,t),t.blockSize===32&&e.adapterInfo.isVendor("intel")&&e.adapterInfo.isArchitecture("gen-12lp")?e.compute(Ml(e.inputs,t)):e.compute(Rl(e.inputs,t))},zh=e=>he(e)}),Nl,Dl,Pl,Ul,Wl,Ll,ql,Vl,Ch,R0=P(()=>{"use strict";Y(),te(),re(),Nl=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 t=e[0].dims.length*2===e[1].dims[0];if(e.length===4&&(t=e[3].dims[0]*2===e[1].dims[0]),!t)throw new Error("The pads should be a 1D tensor of shape [2 * input_rank] or [2 * num_axes].")}},Dl=(e,t,r)=>{let i="";for(let a=t-1;a>=0;--a)i+=`
             k = i32(${e.indicesGet("indices",a)}) - ${j("uniforms.pads",a,r)};
             if (k < 0) {
               break;
@@ -2129,7 +2129,7 @@ fn calculateOutputIndex(index: u32) -> u32 {
               var k = 0;
               ${i}
               value = x[offset];
-          `},ql=(e,t,r)=>{switch(r.mode){case 0:return Dl(e,t,r.pads.length);case 1:return Pl(e,t,r.pads.length);case 2:return Ul(e,t,r.pads.length);case 3:return Wl(e,t,r.pads.length);default:throw new Error("Invalid mode")}},Ll=(e,t)=>{let r=B.padShape(e[0].dims.slice(),t.pads),i=e[0].dims,a=B.size(r),s=[{type:12,data:a},{type:6,data:t.pads}],n=e.length>=3&&e[2].data;t.mode===0&&s.push({type:n?e[2].dataType:1,data:t.value}),s.push(...Z(e[0].dims,r));let u=["rank"],l=p=>{let h=F("output",e[0].dataType,r.length),f=N("x",e[0].dataType,i.length),g=f.type.value,_=ql(h,i.length,t),y=[{name:"output_size",type:"u32"},{name:"pads",type:"i32",length:t.pads.length}];return t.mode===0&&y.push({name:"constant_value",type:n?g:"f32"}),`
+          `},Ll=(e,t,r)=>{switch(r.mode){case 0:return Dl(e,t,r.pads.length);case 1:return Pl(e,t,r.pads.length);case 2:return Ul(e,t,r.pads.length);case 3:return Wl(e,t,r.pads.length);default:throw new Error("Invalid mode")}},ql=(e,t)=>{let r=B.padShape(e[0].dims.slice(),t.pads),i=e[0].dims,a=B.size(r),s=[{type:12,data:a},{type:6,data:t.pads}],n=e.length>=3&&e[2].data;t.mode===0&&s.push({type:n?e[2].dataType:1,data:t.value}),s.push(...Z(e[0].dims,r));let u=["rank"],l=p=>{let h=F("output",e[0].dataType,r.length),f=M("x",e[0].dataType,i.length),g=f.type.value,_=Ll(h,i.length,t),y=[{name:"output_size",type:"u32"},{name:"pads",type:"i32",length:t.pads.length}];return t.mode===0&&y.push({name:"constant_value",type:n?g:"f32"}),`
             ${p.registerUniforms(y).declareVariables(f,h)}
             ${p.mainStart()}
             ${p.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.output_size")}
@@ -2139,7 +2139,7 @@ fn calculateOutputIndex(index: u32) -> u32 {
             var value = ${g}(0);
             ${_}
             output[global_idx] = value;
-        }`};return{name:"Pad",shaderCache:{hint:`${t.mode}${n}`,inputDependencies:u},getRunData:()=>({outputs:[{dims:r,dataType:e[0].dataType}],dispatchGroup:{x:Math.ceil(B.size(r)/64)},programUniforms:s}),getShaderSource:l}},Vl=(e,t)=>{if(e.length>1){let r=e[1].getBigInt64Array(),i=e.length>=3&&e[2].data?e[2].dataType===10?e[2].getUint16Array()[0]:e[2].getFloat32Array()[0]:0,a=e[0].dims.length,s=new Int32Array(2*a).fill(0);if(e.length>=4){let u=e[3].getBigInt64Array();for(let l=0;l<u.length;l++)s[Number(u[l])]=Number(r[l]),s[Number(u[l])+a]=Number(r[l+u.length])}else r.forEach((u,l)=>s[Number(l)]=Number(u));let n=[];return s.forEach(u=>n.push(u)),{mode:t.mode,value:i,pads:n}}else return t},Ch=(e,t)=>{Ml(e.inputs);let r=Vl(e.inputs,t);e.compute(Ll(e.inputs,r),{inputs:[0]})}}),rr,ea,ta,ra,ia,Gl,Fl,aa,na,Ah,Oh,sa,Bh,Rh,oa,Nh,Mh,Dh,Ph,N0=P(()=>{"use strict";We(),Y(),te(),re(),rr=e=>{if(ge.webgpu.validateInputContent&&(!e||e.length!==1))throw new Error("Pool ops requires 1 input.")},ea=(e,t,r)=>{let i=t.format==="NHWC",a=e.dims.slice();i&&a.splice(1,0,a.pop());let s=Object.hasOwnProperty.call(t,"dilations"),n=t.kernelShape.slice(),u=t.strides.slice(),l=s?t.dilations.slice():[],p=t.pads.slice();Gr.adjustPoolAttributes(r,a,n,u,l,p);let h=Gr.computePoolOutputShape(r,a,u,l,n,p,t.autoPad),f=Object.assign({},t);s?Object.assign(f,{kernelShape:n,strides:u,pads:p,dilations:l,cacheKey:t.cacheKey}):Object.assign(f,{kernelShape:n,strides:u,pads:p,cacheKey:t.cacheKey});let g=h.slice();return g.push(g.splice(1,1)[0]),[f,i?g:h]},ta=(e,t)=>{let r=t.format==="NHWC",i=B.size(e),a=B.size(t.kernelShape),s=[{type:12,data:i},{type:12,data:a}],n=[{name:"outputSize",type:"u32"},{name:"kernelSize",type:"u32"}];if(t.kernelShape.length<=2){let u=t.kernelShape[t.kernelShape.length-1],l=t.strides[t.strides.length-1],p=t.pads[t.pads.length/2-1],h=t.pads[t.pads.length-1],f=!!(p+h);s.push({type:12,data:u},{type:12,data:l},{type:12,data:p},{type:12,data:h}),n.push({name:"kw",type:"u32"},{name:"sw",type:"u32"},{name:"pwStart",type:"u32"},{name:"pwEnd",type:"u32"});let g=!1;if(t.kernelShape.length===2){let _=t.kernelShape[t.kernelShape.length-2],y=t.strides[t.strides.length-2],$=t.pads[t.pads.length/2-2],S=t.pads[t.pads.length-2];g=!!($+S),s.push({type:12,data:_},{type:12,data:y},{type:12,data:$},{type:12,data:S}),n.push({name:"kh",type:"u32"},{name:"sh",type:"u32"},{name:"phStart",type:"u32"},{name:"phEnd",type:"u32"})}return[s,n,!0,f,g]}else{if(r)throw new Error("Pooling with kernelShape.length > 2 is not supported for NHWC format.");let u=B.computeStrides(t.kernelShape);s.push({type:12,data:u},{type:12,data:t.pads},{type:12,data:t.strides}),n.push({name:"kernelStrides",type:"u32",length:u.length},{name:"pads",type:"u32",length:t.pads.length},{name:"strides",type:"u32",length:t.strides.length});let l=t.pads.reduce((p,h)=>p+h);return[s,n,!!l,!1,!1]}},ra=(e,t,r,i,a,s,n,u,l,p,h,f)=>{let g=a.format==="NHWC",_=t.type.value,y=F("output",t.type.tensor,i);if(a.kernelShape.length<=2){let $="",S="",v="",w=r-(g?2:1);if(h?$=`
+        }`};return{name:"Pad",shaderCache:{hint:`${t.mode}${n}`,inputDependencies:u},getRunData:()=>({outputs:[{dims:r,dataType:e[0].dataType}],dispatchGroup:{x:Math.ceil(B.size(r)/64)},programUniforms:s}),getShaderSource:l}},Vl=(e,t)=>{if(e.length>1){let r=e[1].getBigInt64Array(),i=e.length>=3&&e[2].data?e[2].dataType===10?e[2].getUint16Array()[0]:e[2].getFloat32Array()[0]:0,a=e[0].dims.length,s=new Int32Array(2*a).fill(0);if(e.length>=4){let u=e[3].getBigInt64Array();for(let l=0;l<u.length;l++)s[Number(u[l])]=Number(r[l]),s[Number(u[l])+a]=Number(r[l+u.length])}else r.forEach((u,l)=>s[Number(l)]=Number(u));let n=[];return s.forEach(u=>n.push(u)),{mode:t.mode,value:i,pads:n}}else return t},Ch=(e,t)=>{Nl(e.inputs);let r=Vl(e.inputs,t);e.compute(ql(e.inputs,r),{inputs:[0]})}}),rr,ea,ta,ra,ia,Gl,Fl,aa,na,Ah,Oh,sa,Bh,Rh,oa,Mh,Nh,Dh,Ph,M0=P(()=>{"use strict";We(),Y(),te(),re(),rr=e=>{if(ge.webgpu.validateInputContent&&(!e||e.length!==1))throw new Error("Pool ops requires 1 input.")},ea=(e,t,r)=>{let i=t.format==="NHWC",a=e.dims.slice();i&&a.splice(1,0,a.pop());let s=Object.hasOwnProperty.call(t,"dilations"),n=t.kernelShape.slice(),u=t.strides.slice(),l=s?t.dilations.slice():[],p=t.pads.slice();Gr.adjustPoolAttributes(r,a,n,u,l,p);let h=Gr.computePoolOutputShape(r,a,u,l,n,p,t.autoPad),f=Object.assign({},t);s?Object.assign(f,{kernelShape:n,strides:u,pads:p,dilations:l,cacheKey:t.cacheKey}):Object.assign(f,{kernelShape:n,strides:u,pads:p,cacheKey:t.cacheKey});let g=h.slice();return g.push(g.splice(1,1)[0]),[f,i?g:h]},ta=(e,t)=>{let r=t.format==="NHWC",i=B.size(e),a=B.size(t.kernelShape),s=[{type:12,data:i},{type:12,data:a}],n=[{name:"outputSize",type:"u32"},{name:"kernelSize",type:"u32"}];if(t.kernelShape.length<=2){let u=t.kernelShape[t.kernelShape.length-1],l=t.strides[t.strides.length-1],p=t.pads[t.pads.length/2-1],h=t.pads[t.pads.length-1],f=!!(p+h);s.push({type:12,data:u},{type:12,data:l},{type:12,data:p},{type:12,data:h}),n.push({name:"kw",type:"u32"},{name:"sw",type:"u32"},{name:"pwStart",type:"u32"},{name:"pwEnd",type:"u32"});let g=!1;if(t.kernelShape.length===2){let _=t.kernelShape[t.kernelShape.length-2],y=t.strides[t.strides.length-2],$=t.pads[t.pads.length/2-2],S=t.pads[t.pads.length-2];g=!!($+S),s.push({type:12,data:_},{type:12,data:y},{type:12,data:$},{type:12,data:S}),n.push({name:"kh",type:"u32"},{name:"sh",type:"u32"},{name:"phStart",type:"u32"},{name:"phEnd",type:"u32"})}return[s,n,!0,f,g]}else{if(r)throw new Error("Pooling with kernelShape.length > 2 is not supported for NHWC format.");let u=B.computeStrides(t.kernelShape);s.push({type:12,data:u},{type:12,data:t.pads},{type:12,data:t.strides}),n.push({name:"kernelStrides",type:"u32",length:u.length},{name:"pads",type:"u32",length:t.pads.length},{name:"strides",type:"u32",length:t.strides.length});let l=t.pads.reduce((p,h)=>p+h);return[s,n,!!l,!1,!1]}},ra=(e,t,r,i,a,s,n,u,l,p,h,f)=>{let g=a.format==="NHWC",_=t.type.value,y=F("output",t.type.tensor,i);if(a.kernelShape.length<=2){let $="",S="",v="",w=r-(g?2:1);if(h?$=`
                 for (var i: u32 = 0u; i < uniforms.kw; i++) {
                   xIndices[${w}] = indices[${w}] * uniforms.sw - uniforms.pwStart + i;
                   if (xIndices[${w}] < 0 || xIndices[${w}]
@@ -2154,16 +2154,16 @@ fn calculateOutputIndex(index: u32) -> u32 {
                   xIndices[${w}] = indices[${w}] * uniforms.sw - uniforms.pwStart + i;
                   let x_val = x[${t.indicesToOffset("xIndices")}];
                   ${s}
-                }`,a.kernelShape.length===2){let E=r-(g?3:2);f?S=`
+                }`,a.kernelShape.length===2){let k=r-(g?3:2);f?S=`
                 for (var j: u32 = 0u; j < uniforms.kh; j++) {
-                  xIndices[${E}] = indices[${E}] * uniforms.sh - uniforms.phStart + j;
-                  if (xIndices[${E}] < 0 || xIndices[${E}] >= uniforms.x_shape[${E}]) {
+                  xIndices[${k}] = indices[${k}] * uniforms.sh - uniforms.phStart + j;
+                  if (xIndices[${k}] < 0 || xIndices[${k}] >= uniforms.x_shape[${k}]) {
                     pad += i32(uniforms.kw);
                     continue;
                   }
               `:S=`
                 for (var j: u32 = 0u; j < uniforms.kh; j++) {
-                  xIndices[${E}] = indices[${E}] * uniforms.sh - uniforms.phStart + j;
+                  xIndices[${k}] = indices[${k}] * uniforms.sh - uniforms.phStart + j;
                 `,v=`
               }
             `}return`
@@ -2228,19 +2228,19 @@ fn calculateOutputIndex(index: u32) -> u32 {
               ${n}
 
               output[global_idx] = value;
-            }`}},ia=e=>`${e.format};${e.ceilMode};${e.autoPad};${e.kernelShape.length}`,Gl=e=>`${ia(e)};${e.countIncludePad}`,Fl=e=>`${ia(e)};${e.storageOrder};${e.dilations}`,aa=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}),na=(e,t,r,i)=>{let[a,s]=ea(t,i,r),n=N("x",t.dataType,t.dims.length),u=n.type.value,l="value += x_val;",p="";a.countIncludePad?p+=`value /= ${u}(uniforms.kernelSize);`:p+=`value /= ${u}(i32(uniforms.kernelSize) - pad);`;let[h,f,g,_,y]=ta(s,a);h.push(...Z(t.dims,s));let $=["rank"];return{name:e,shaderCache:{hint:`${i.cacheKey};${g};${_};${y}`,inputDependencies:$},getRunData:()=>({outputs:[{dims:s,dataType:t.dataType}],dispatchGroup:{x:Math.ceil(B.size(s)/64)},programUniforms:h}),getShaderSource:S=>ra(S,n,t.dims.length,s.length,a,l,p,0,f,g,_,y)}},Ah=e=>{let t=e.count_include_pad!==0,r=aa(e);if(r.ceilMode!==0)throw new Error("using ceil() in shape computation is not yet supported for AveragePool");let i={countIncludePad:t,...r,cacheKey:""};return{...i,cacheKey:Gl(i)}},Oh=(e,t)=>{rr(e.inputs),e.compute(na("AveragePool",e.inputs[0],!1,t))},sa={autoPad:"",ceilMode:0,countIncludePad:!1,kernelShape:[],strides:[],pads:[],storageOrder:0,dilations:[]},Bh=e=>{let t=e.format;return{format:t,...sa,cacheKey:t}},Rh=(e,t)=>{rr(e.inputs),e.compute(na("GlobalAveragePool",e.inputs[0],!0,t))},oa=(e,t,r,i)=>{let[a,s]=ea(t,i,r),n=`
+            }`}},ia=e=>`${e.format};${e.ceilMode};${e.autoPad};${e.kernelShape.length}`,Gl=e=>`${ia(e)};${e.countIncludePad}`,Fl=e=>`${ia(e)};${e.storageOrder};${e.dilations}`,aa=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}),na=(e,t,r,i)=>{let[a,s]=ea(t,i,r),n=M("x",t.dataType,t.dims.length),u=n.type.value,l="value += x_val;",p="";a.countIncludePad?p+=`value /= ${u}(uniforms.kernelSize);`:p+=`value /= ${u}(i32(uniforms.kernelSize) - pad);`;let[h,f,g,_,y]=ta(s,a);h.push(...Z(t.dims,s));let $=["rank"];return{name:e,shaderCache:{hint:`${i.cacheKey};${g};${_};${y}`,inputDependencies:$},getRunData:()=>({outputs:[{dims:s,dataType:t.dataType}],dispatchGroup:{x:Math.ceil(B.size(s)/64)},programUniforms:h}),getShaderSource:S=>ra(S,n,t.dims.length,s.length,a,l,p,0,f,g,_,y)}},Ah=e=>{let t=e.count_include_pad!==0,r=aa(e);if(r.ceilMode!==0)throw new Error("using ceil() in shape computation is not yet supported for AveragePool");let i={countIncludePad:t,...r,cacheKey:""};return{...i,cacheKey:Gl(i)}},Oh=(e,t)=>{rr(e.inputs),e.compute(na("AveragePool",e.inputs[0],!1,t))},sa={autoPad:"",ceilMode:0,countIncludePad:!1,kernelShape:[],strides:[],pads:[],storageOrder:0,dilations:[]},Bh=e=>{let t=e.format;return{format:t,...sa,cacheKey:t}},Rh=(e,t)=>{rr(e.inputs),e.compute(na("GlobalAveragePool",e.inputs[0],!0,t))},oa=(e,t,r,i)=>{let[a,s]=ea(t,i,r),n=`
       value = max(x_val, value);
-    `,u="",l=N("x",t.dataType,t.dims.length),p=["rank"],[h,f,g,_,y]=ta(s,a);return h.push(...Z(t.dims,s)),{name:e,shaderCache:{hint:`${i.cacheKey};${g};${_};${y}`,inputDependencies:p},getRunData:()=>({outputs:[{dims:s,dataType:t.dataType}],dispatchGroup:{x:Math.ceil(B.size(s)/64)},programUniforms:h}),getShaderSource:$=>ra($,l,t.dims.length,s.length,a,n,u,t.dataType===10?-65504:-1e5,f,g,_,y)}},Nh=(e,t)=>{rr(e.inputs),e.compute(oa("MaxPool",e.inputs[0],!1,t))},Mh=e=>{let t=e.storage_order,r=e.dilations,i=aa(e);if(t!==0)throw new Error("column major storage order is not yet supported for MaxPool");if(i.ceilMode!==0)throw new Error("using ceil() in shape computation is not yet supported for MaxPool");let a={storageOrder:t,dilations:r,...i,cacheKey:""};return{...a,cacheKey:Fl(a)}},Dh=e=>{let t=e.format;return{format:t,...sa,cacheKey:t}},Ph=(e,t)=>{rr(e.inputs),e.compute(oa("GlobalMaxPool",e.inputs[0],!0,t))}}),Hl,jl,Uh,Wh,M0=P(()=>{"use strict";Y(),te(),ve(),re(),Hl=(e,t)=>{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((r,i)=>r===e[2].dims[i]).reduce((r,i)=>r&&i,!0))throw new Error("scale and zero-point inputs must have the same shape.")}if(t.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((a,s)=>s===t.axis||a===e[0].dims[s]).reduce((a,s)=>a&&s,!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 r=e[0].dims[t.axis],i=e[1].dims[t.axis];if(t.blockSize<Math.ceil(r/i)||t.blockSize>Math.ceil(r/(i-1)-1))throw new Error("blockSize must be with in the range [ceil(dI / Si), ceil(dI / (Si - 1) - 1)].")}},jl=(e,t)=>{let r=B.normalizeAxis(t.axis,e[0].dims.length),i=e[0].dataType,a=i===3,s=e[0].dims,n=e[1].dataType,u=B.size(s),l=i===3||i===2,p=l?[Math.ceil(B.size(e[0].dims)/4)]:e[0].dims,h=e[1].dims,f=e.length>2?e[2]:void 0,g=f?l?[Math.ceil(B.size(f.dims)/4)]:f.dims:void 0,_=h.length===0||h.length===1&&h[0]===1,y=_===!1&&h.length===1,$=$e(u),S=_&&(!l||$===4),v=S?$:1,w=S&&!l?$:1,E=N("input",l?12:i,p.length,w),I=N("scale",n,h.length),k=f?N("zero_point",l?12:i,g.length):void 0,C=F("output",n,s.length,v),O=[E,I];k&&O.push(k);let x=[p,h];f&&x.push(g);let D=[{type:12,data:u/v},{type:12,data:r},{type:12,data:t.blockSize},...Z(...x,s)],q=Q=>{let L=[{name:"output_size",type:"u32"},{name:"axis",type:"u32"},{name:"block_size",type:"u32"}];return`
-      ${Q.registerUniforms(L).declareVariables(...O,C)}
+    `,u="",l=M("x",t.dataType,t.dims.length),p=["rank"],[h,f,g,_,y]=ta(s,a);return h.push(...Z(t.dims,s)),{name:e,shaderCache:{hint:`${i.cacheKey};${g};${_};${y}`,inputDependencies:p},getRunData:()=>({outputs:[{dims:s,dataType:t.dataType}],dispatchGroup:{x:Math.ceil(B.size(s)/64)},programUniforms:h}),getShaderSource:$=>ra($,l,t.dims.length,s.length,a,n,u,t.dataType===10?-65504:-1e5,f,g,_,y)}},Mh=(e,t)=>{rr(e.inputs),e.compute(oa("MaxPool",e.inputs[0],!1,t))},Nh=e=>{let t=e.storage_order,r=e.dilations,i=aa(e);if(t!==0)throw new Error("column major storage order is not yet supported for MaxPool");if(i.ceilMode!==0)throw new Error("using ceil() in shape computation is not yet supported for MaxPool");let a={storageOrder:t,dilations:r,...i,cacheKey:""};return{...a,cacheKey:Fl(a)}},Dh=e=>{let t=e.format;return{format:t,...sa,cacheKey:t}},Ph=(e,t)=>{rr(e.inputs),e.compute(oa("GlobalMaxPool",e.inputs[0],!0,t))}}),Hl,jl,Uh,Wh,N0=P(()=>{"use strict";Y(),te(),ve(),re(),Hl=(e,t)=>{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((r,i)=>r===e[2].dims[i]).reduce((r,i)=>r&&i,!0))throw new Error("scale and zero-point inputs must have the same shape.")}if(t.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((a,s)=>s===t.axis||a===e[0].dims[s]).reduce((a,s)=>a&&s,!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 r=e[0].dims[t.axis],i=e[1].dims[t.axis];if(t.blockSize<Math.ceil(r/i)||t.blockSize>Math.ceil(r/(i-1)-1))throw new Error("blockSize must be with in the range [ceil(dI / Si), ceil(dI / (Si - 1) - 1)].")}},jl=(e,t)=>{let r=B.normalizeAxis(t.axis,e[0].dims.length),i=e[0].dataType,a=i===3,s=e[0].dims,n=e[1].dataType,u=B.size(s),l=i===3||i===2,p=l?[Math.ceil(B.size(e[0].dims)/4)]:e[0].dims,h=e[1].dims,f=e.length>2?e[2]:void 0,g=f?l?[Math.ceil(B.size(f.dims)/4)]:f.dims:void 0,_=h.length===0||h.length===1&&h[0]===1,y=_===!1&&h.length===1,$=$e(u),S=_&&(!l||$===4),v=S?$:1,w=S&&!l?$:1,k=M("input",l?12:i,p.length,w),I=M("scale",n,h.length),E=f?M("zero_point",l?12:i,g.length):void 0,C=F("output",n,s.length,v),O=[k,I];E&&O.push(E);let x=[p,h];f&&x.push(g);let D=[{type:12,data:u/v},{type:12,data:r},{type:12,data:t.blockSize},...Z(...x,s)],L=Q=>{let q=[{name:"output_size",type:"u32"},{name:"axis",type:"u32"},{name:"block_size",type:"u32"}];return`
+      ${Q.registerUniforms(q).declareVariables(...O,C)}
       ${Q.mainStart()}
           ${Q.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.output_size")}
           let output_indices = ${C.offsetToIndices("global_idx")};
 
           // Set input x
           ${l?`
-            let input = ${E.getByOffset("global_idx / 4")};
+            let input = ${k.getByOffset("global_idx / 4")};
             let x_vec = ${a?"unpack4xI8(input)":"unpack4xU8(input)"};
-            let x_value = ${v===1?"x_vec[global_idx % 4]":"x_vec"};`:`let x_value = ${E.getByOffset("global_idx")};`};
+            let x_value = ${v===1?"x_vec[global_idx % 4]":"x_vec"};`:`let x_value = ${k.getByOffset("global_idx")};`};
 
           // Set scale input
           ${_?`let scale_value= ${I.getByOffset("0")}`:y?`
@@ -2252,28 +2252,28 @@ fn calculateOutputIndex(index: u32) -> u32 {
             let scale_value= ${I.getByIndices("scale_indices")};`};
 
           // Set zero-point input
-          ${k?_?l?`
-                let zero_point_input = ${k.getByOffset("0")};
+          ${E?_?l?`
+                let zero_point_input = ${E.getByOffset("0")};
                 let zero_point_vec =  ${a?"unpack4xI8(zero_point_input)":"unpack4xU8(zero_point_input)"};
-                let zero_point_value= zero_point_vec[0]`:`let zero_point_value = ${k.getByOffset("0")}`:y?l?`
+                let zero_point_value= zero_point_vec[0]`:`let zero_point_value = ${E.getByOffset("0")}`:y?l?`
                 let zero_point_index = ${C.indicesGet("output_indices","uniforms.axis")};
-                let zero_point_input = ${k.getByOffset("zero_point_index / 4")};
+                let zero_point_input = ${E.getByOffset("zero_point_index / 4")};
                 let zero_point_vec =  ${a?"unpack4xI8(zero_point_input)":"unpack4xU8(zero_point_input)"};
                 let zero_point_value = zero_point_vec[zero_point_index % 4]`:`
                 let zero_point_index = ${C.indicesGet("output_indices","uniforms.axis")};
-                let zero_point_value = ${k.getByOffset("zero_point_index")};`:l?`
+                let zero_point_value = ${E.getByOffset("zero_point_index")};`:l?`
                 let zero_point_offset = ${I.indicesToOffset("scale_indices")};
-                let zero_point_input = ${k.getByOffset("zero_point_offset / 4")};
+                let zero_point_input = ${E.getByOffset("zero_point_offset / 4")};
                 let zero_point_vec = ${a?"unpack4xI8(zero_point_input)":"unpack4xU8(zero_point_input)"};
-                let zero_point_value = zero_point_vec[zero_point_offset % 4];`:`let zero_point_value = ${k.getByIndices("scale_indices")};`:`let zero_point_value = ${l?a?"i32":"u32":E.type.value}(0);`};
+                let zero_point_value = zero_point_vec[zero_point_offset % 4];`:`let zero_point_value = ${E.getByIndices("scale_indices")};`:`let zero_point_value = ${l?a?"i32":"u32":k.type.value}(0);`};
       // Compute and write output
       ${C.setByOffset("global_idx",`${C.type.value}(x_value - zero_point_value) * scale_value`)};
-      }`};return{name:"DequantizeLinear",shaderCache:{hint:t.cacheKey,inputDependencies:k?["rank","rank","rank"]:["rank","rank"]},getShaderSource:q,getRunData:()=>({outputs:[{dims:s,dataType:n}],dispatchGroup:{x:Math.ceil(u/v/64),y:1,z:1},programUniforms:D})}},Uh=(e,t)=>{Hl(e.inputs,t),e.compute(jl(e.inputs,t))},Wh=e=>he({axis:e.axis,blockSize:e.blockSize})}),Kl,Zl,qh,D0=P(()=>{"use strict";We(),Y(),re(),Kl=(e,t,r)=>{let i=e===t,a=e<t&&r<0,s=e>t&&r>0;if(i||a||s)throw new Error("Range these inputs' contents are invalid.")},Zl=(e,t,r,i)=>{let a=Math.abs(Math.ceil((t-e)/r)),s=[a],n=a,u=[{type:12,data:n},{type:i,data:e},{type:i,data:r},...Z(s)],l=p=>{let h=F("output",i,s.length),f=h.type.value,g=[{name:"outputSize",type:"u32"},{name:"start",type:f},{name:"delta",type:f}];return`
+      }`};return{name:"DequantizeLinear",shaderCache:{hint:t.cacheKey,inputDependencies:E?["rank","rank","rank"]:["rank","rank"]},getShaderSource:L,getRunData:()=>({outputs:[{dims:s,dataType:n}],dispatchGroup:{x:Math.ceil(u/v/64),y:1,z:1},programUniforms:D})}},Uh=(e,t)=>{Hl(e.inputs,t),e.compute(jl(e.inputs,t))},Wh=e=>he({axis:e.axis,blockSize:e.blockSize})}),Kl,Zl,Lh,D0=P(()=>{"use strict";We(),Y(),re(),Kl=(e,t,r)=>{let i=e===t,a=e<t&&r<0,s=e>t&&r>0;if(i||a||s)throw new Error("Range these inputs' contents are invalid.")},Zl=(e,t,r,i)=>{let a=Math.abs(Math.ceil((t-e)/r)),s=[a],n=a,u=[{type:12,data:n},{type:i,data:e},{type:i,data:r},...Z(s)],l=p=>{let h=F("output",i,s.length),f=h.type.value,g=[{name:"outputSize",type:"u32"},{name:"start",type:f},{name:"delta",type:f}];return`
         ${p.registerUniforms(g).declareVariables(h)}
         ${p.mainStart()}
         ${p.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.outputSize")}
         output[global_idx] = uniforms.start + ${f}(global_idx) * uniforms.delta;
-      }`};return{name:"Range",shaderCache:{hint:`${i}`},getShaderSource:l,getRunData:()=>({outputs:[{dims:s,dataType:i}],dispatchGroup:{x:Math.ceil(n/64)},programUniforms:u})}},qh=e=>{let t=0,r=0,i=0;e.inputs[0].dataType===6?(t=e.inputs[0].getInt32Array()[0],r=e.inputs[1].getInt32Array()[0],i=e.inputs[2].getInt32Array()[0]):e.inputs[0].dataType===1&&(t=e.inputs[0].getFloat32Array()[0],r=e.inputs[1].getFloat32Array()[0],i=e.inputs[2].getFloat32Array()[0]),ge.webgpu.validateInputContent&&Kl(t,r,i),e.compute(Zl(t,r,i,e.inputs[0].dataType),{inputs:[]})}}),Ql,Xl,Lh,Vh,P0=P(()=>{"use strict";Y(),te(),ve(),re(),Ql=(e,t,r,i)=>{if(e!=="none"&&i!=="i32"&&i!=="u32"&&i!=="f32")throw new Error(`Input ${i} is not supported with reduction ${e}.`);let a=`{
+      }`};return{name:"Range",shaderCache:{hint:`${i}`},getShaderSource:l,getRunData:()=>({outputs:[{dims:s,dataType:i}],dispatchGroup:{x:Math.ceil(n/64)},programUniforms:u})}},Lh=e=>{let t=0,r=0,i=0;e.inputs[0].dataType===6?(t=e.inputs[0].getInt32Array()[0],r=e.inputs[1].getInt32Array()[0],i=e.inputs[2].getInt32Array()[0]):e.inputs[0].dataType===1&&(t=e.inputs[0].getFloat32Array()[0],r=e.inputs[1].getFloat32Array()[0],i=e.inputs[2].getFloat32Array()[0]),ge.webgpu.validateInputContent&&Kl(t,r,i),e.compute(Zl(t,r,i,e.inputs[0].dataType),{inputs:[]})}}),Ql,Xl,qh,Vh,P0=P(()=>{"use strict";Y(),te(),ve(),re(),Ql=(e,t,r,i)=>{if(e!=="none"&&i!=="i32"&&i!=="u32"&&i!=="f32")throw new Error(`Input ${i} is not supported with reduction ${e}.`);let a=`{
                 var oldValue = 0;
                 loop {
                   let newValueF32 =`,s=`;
@@ -2286,7 +2286,7 @@ fn calculateOutputIndex(index: u32) -> u32 {
                 }
               }`;switch(e){case"none":return`${t}=${r};`;case"add":return i==="i32"||i==="u32"?`atomicAdd(&${t}, bitcast<${i}>(${r}));`:`
               ${a}bitcast<${i}>(oldValue) + (${r})${s}`;case"max":return i==="i32"||i==="u32"?`atomicMax(&${t}, bitcast<${i}>(${r}));`:`
-                ${a}max(bitcast<f32>(oldValue), (${r}))${s}`;case"min":return i==="i32"||i==="u32"?`atomicMin(&${t}, bitcast<${i}>(${r}));`:`${a}min(bitcast<${i}>(oldValue), (${r}))${s}`;case"mul":return`${a}(bitcast<${i}>(oldValue) * (${r}))${s}`;default:throw new Error(`Reduction ${e} is not supported.`)}},Xl=(e,t)=>{let r=e[0].dims,i=e[1].dims,a=r,s=1,n=Math.ceil(B.sizeToDimension(i,i.length-1)/s),u=i[i.length-1],l=B.sizeFromDimension(r,u),p=[{type:12,data:n},{type:12,data:u},{type:12,data:l},...Z(e[1].dims,e[2].dims,a)],h=f=>{let g=N("indices",e[1].dataType,e[1].dims.length),_=N("updates",e[2].dataType,e[2].dims.length,s),y=t.reduction!=="none"&&t.reduction!==""?yp("output",e[0].dataType,a.length):F("output",e[0].dataType,a.length,s);return`
+                ${a}max(bitcast<f32>(oldValue), (${r}))${s}`;case"min":return i==="i32"||i==="u32"?`atomicMin(&${t}, bitcast<${i}>(${r}));`:`${a}min(bitcast<${i}>(oldValue), (${r}))${s}`;case"mul":return`${a}(bitcast<${i}>(oldValue) * (${r}))${s}`;default:throw new Error(`Reduction ${e} is not supported.`)}},Xl=(e,t)=>{let r=e[0].dims,i=e[1].dims,a=r,s=1,n=Math.ceil(B.sizeToDimension(i,i.length-1)/s),u=i[i.length-1],l=B.sizeFromDimension(r,u),p=[{type:12,data:n},{type:12,data:u},{type:12,data:l},...Z(e[1].dims,e[2].dims,a)],h=f=>{let g=M("indices",e[1].dataType,e[1].dims.length),_=M("updates",e[2].dataType,e[2].dims.length,s),y=t.reduction!=="none"&&t.reduction!==""?yp("output",e[0].dataType,a.length):F("output",e[0].dataType,a.length,s);return`
       ${f.registerUniform("output_size","u32").registerUniform("last_index_dimension","u32").registerUniform("num_updates_elements","u32").declareVariables(g,_,y)}
       ${f.mainStart()}
         ${f.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.output_size")}
@@ -2319,7 +2319,7 @@ fn calculateOutputIndex(index: u32) -> u32 {
     ${Ql(t.reduction,"output[data_offset + i]","value",y.type.value)}
   }
 
-      }`};return{name:"ScatterND",shaderCache:{hint:`${t.cacheKey}_${t.reduction}`,inputDependencies:["rank","rank"]},getRunData:()=>({outputs:[{dims:a,dataType:e[0].dataType}],dispatchGroup:{x:Math.ceil(n/64)},programUniforms:p}),getShaderSource:h}},Lh=e=>he({reduction:e.reduction}),Vh=(e,t)=>{e.compute(Xl(e.inputs,t),{inputs:[e.inputs[1],e.inputs[2]],outputs:[]})}}),Yl,Jl,ed,ua,td,rd,id,ad,nd,sd,od,ud,la,ld,dd,pd,cd,hd,Gh,Fh,U0=P(()=>{"use strict";Y(),te(),ve(),re(),Yl=(e,t)=>{if(e.every(r=>r>0||(()=>{throw new Error("Resize requires scales input values to be positive")})),e.length>0){if(t.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
+      }`};return{name:"ScatterND",shaderCache:{hint:`${t.cacheKey}_${t.reduction}`,inputDependencies:["rank","rank"]},getRunData:()=>({outputs:[{dims:a,dataType:e[0].dataType}],dispatchGroup:{x:Math.ceil(n/64)},programUniforms:p}),getShaderSource:h}},qh=e=>he({reduction:e.reduction}),Vh=(e,t)=>{e.compute(Xl(e.inputs,t),{inputs:[e.inputs[1],e.inputs[2]],outputs:[]})}}),Yl,Jl,ed,ua,td,rd,id,ad,nd,sd,od,ud,la,ld,dd,pd,cd,hd,Gh,Fh,U0=P(()=>{"use strict";Y(),te(),ve(),re(),Yl=(e,t)=>{if(e.every(r=>r>0||(()=>{throw new Error("Resize requires scales input values to be positive")})),e.length>0){if(t.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(t.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")}},Jl=(e,t,r)=>{t.every(a=>a>=0&&a<r||(()=>{throw new Error("Resize requires axes input values to be positive and less than rank")}));let i=new Array(r).fill(1);return t.forEach((a,s)=>i[a]=e[s]),i},ed=(e,t,r,i,a,s)=>{let[n,u,l]=r>10?[1,2,3]:[-1,e.length>1?1:-1,-1],p=e[0].dims.length;if(n>0&&e.length>n&&e[n].dims.length>0)e[n].getFloat32Array().forEach(h=>s.push(h));else if(t.coordinateTransformMode==="tf_crop_and_resize")throw new Error("Resize requires RoI input to be specified when coordinateTransformMode is tfCropAndResize");if(u>0&&e.length>u&&e[u].dims.length===1&&e[u].dims[0]>0){if(e[u].getFloat32Array().forEach(h=>i.push(h)),i.length!==0&&i.length!==p&&r>=18&&i.length!==t.axes.length)throw new Error("Resize requires scales input size to be same as input rank or axes size for opset 18 and up");Yl(i,t),t.axes.length>0&&Jl(i,t.axes,p).forEach((h,f)=>i[f]=h)}if(l>0&&e.length>l&&e[l].dims.length===1&&e[l].dims[0]>0&&(e[l].getBigInt64Array().forEach(h=>a.push(Number(h))),a.length!==0&&a.length!==p&&r>=18&&a.length!==t.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(t.axes.length>0){if(i.length!==0&&i.length!==t.axes.length)throw new Error('Resize requires "scales" input size to be of axes rank when axes attributes is specified');if(a.length!==0&&a.length!==t.axes.length)throw new Error('Resize requires "sizes" input size to be of rank axes rank when axes attributes is specified')}if(typeof i<"u"&&typeof a<"u"&&i.length>0&&a.length>p)throw new Error("Resize requires only of scales or sizes to be specified")},ua=(e,t,r,i)=>`
   // 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
@@ -2562,7 +2562,7 @@ fn calculateOutputIndex(index: u32) -> u32 {
       }
       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);
-    }`},cd=(e,t,r,i,a,s)=>{let n=e.dims,u=id(s,t.axes,n.length),l=ad(n,i,a,t.axes),p=i.slice();i.length===0&&(p=n.map((w,E)=>w===0?1:l[E]/w),t.keepAspectRatioPolicy!=="stretch"&&(l=nd(n,p,t)));let h=F("output",e.dataType,l.length),f=N("input",e.dataType,n.length),g=B.size(l),_=n.length===l.length&&n.every((w,E)=>w===l[E]),y=t.coordinateTransformMode==="tf_crop_and_resize",$=t.extrapolationValue,S=f.type.value,v=w=>`
+    }`},cd=(e,t,r,i,a,s)=>{let n=e.dims,u=id(s,t.axes,n.length),l=ad(n,i,a,t.axes),p=i.slice();i.length===0&&(p=n.map((w,k)=>w===0?1:l[k]/w),t.keepAspectRatioPolicy!=="stretch"&&(l=nd(n,p,t)));let h=F("output",e.dataType,l.length),f=M("input",e.dataType,n.length),g=B.size(l),_=n.length===l.length&&n.every((w,k)=>w===l[k]),y=t.coordinateTransformMode==="tf_crop_and_resize",$=t.extrapolationValue,S=f.type.value,v=w=>`
       ${_?"":`
       ${td(t.coordinateTransformMode,S)};
       ${(()=>{switch(t.mode){case"nearest":return`
@@ -2589,13 +2589,13 @@ fn calculateOutputIndex(index: u32) -> u32 {
                   output[global_idx] = ${t.extrapolationValue};
                 }`;case"linear":return`output[global_idx] = ${n.length===2||n.length===4?"bilinearInterpolation":"trilinearInterpolation"}(output_indices);`;case"cubic":return"output[global_idx] = bicubicInterpolation(output_indices);";default:throw Error(`Unsupported resize mode: ${t.mode}`)}})()};
 `}
-      }`;return{name:"Resize",shaderCache:{hint:`${t.cacheKey}|${r}|${p.length>0?t.mode==="cubic"?p:p.length:""}|${a.length>0?a:""}|${u.length>0?u:""}|${_}|${t.mode==="nearest"?n.length:n}`,inputDependencies:["rank"]},getShaderSource:v,getRunData:()=>({outputs:[{dims:l,dataType:e.dataType}],dispatchGroup:{x:Math.ceil(g/64)},programUniforms:[{type:12,data:g},{type:1,data:p},{type:1,data:u},...Z(n,l)]})}},hd=e=>{let t=e.customDataBuffer;return new Uint32Array(t,t.byteOffset,1)[0]},Gh=(e,t)=>{let r=[],i=[],a=[],s=hd(e);if(t.antialias!==0)throw Error("Only default value (0) for Antialias attribute is supported");ed(e.inputs,t,s,r,i,a),e.compute(cd(e.inputs[0],t,s,r,i,a),{inputs:[0]})},Fh=e=>{let t=e.antialias,r=e.axes,i=e.coordinateTransformMode,a=e.cubicCoeffA,s=e.excludeOutside!==0,n=e.extrapolationValue,u=e.keepAspectRatioPolicy,l=e.mode,p=e.nearestMode===""?"simple":e.nearestMode;return he({antialias:t,axes:r,coordinateTransformMode:i,cubicCoeffA:a,excludeOutside:s,extrapolationValue:n,keepAspectRatioPolicy:u,mode:l,nearestMode:p})}}),fd,md,Hh,W0=P(()=>{"use strict";Y(),te(),re(),fd=e=>{if(!e||e.length<3)throw new Error("layerNorm requires at least 3 inputs.");let t=e[0],r=e[1],i=e[2];if(t.dataType!==r.dataType||t.dataType!==i.dataType)throw new Error("All inputs must have the same data type");if(t.dims.length!==3&&t.dims.length!==2)throw new Error("Input must be 2D or 3D");if(r.dims.length!==3&&r.dims.length!==2)throw new Error("Skip must be 2D or 3D");let a=t.dims[t.dims.length-1],s=t.dims[t.dims.length-2];if(r.dims[r.dims.length-1]!==a)throw new Error("Skip must have the same hidden size as input");if(r.dims[r.dims.length-2]!==s)throw new Error("Skip must have the same sequence length as input");if(i.dims.length!==1)throw new Error("Gamma must be 1D");if(i.dims[i.dims.length-1]!==a)throw new Error("Gamma must have the same hidden size as input");if(e.length>3){let n=e[3];if(n.dims.length!==1)throw new Error("Beta must be 1D");if(n.dims[n.dims.length-1]!==a)throw new Error("Beta must have the same hidden size as input")}if(e.length>4){let n=e[4];if(n.dims.length!==1)throw new Error("Bias must be 1D");if(n.dims[n.dims.length-1]!==a)throw new Error("Bias must have the same hidden size as input")}},md=(e,t,r,i)=>{let a=t.simplified,s=e[0].dims,n=B.size(s),u=s,l=n,p=s.slice(-1)[0],h=i?s.slice(0,-1).concat(1):[],f=!a&&e.length>3,g=e.length>4,_=i&&r>1,y=i&&r>2,$=r>3,S=64,v=$e(p),w=[{type:12,data:l},{type:12,data:v},{type:12,data:p},{type:1,data:t.epsilon}],E=k=>{let C=[{name:"output_size",type:"u32"},{name:"components",type:"u32"},{name:"hidden_size",type:"u32"},{name:"epsilon",type:"f32"}],O=[N("x",e[0].dataType,e[0].dims,v),N("skip",e[1].dataType,e[1].dims,v),N("gamma",e[2].dataType,e[2].dims,v)];f&&O.push(N("beta",e[3].dataType,e[3].dims,v)),g&&O.push(N("bias",e[4].dataType,e[4].dims,v)),O.push(F("output",e[0].dataType,u,v)),_&&O.push(F("mean_output",1,h)),y&&O.push(F("inv_std_output",1,h)),$&&O.push(F("input_skip_bias_sum",e[0].dataType,u,v));let x=Ee(e[0].dataType),D=Ee(1,v);return`
+      }`;return{name:"Resize",shaderCache:{hint:`${t.cacheKey}|${r}|${p.length>0?t.mode==="cubic"?p:p.length:""}|${a.length>0?a:""}|${u.length>0?u:""}|${_}|${t.mode==="nearest"?n.length:n}`,inputDependencies:["rank"]},getShaderSource:v,getRunData:()=>({outputs:[{dims:l,dataType:e.dataType}],dispatchGroup:{x:Math.ceil(g/64)},programUniforms:[{type:12,data:g},{type:1,data:p},{type:1,data:u},...Z(n,l)]})}},hd=e=>{let t=e.customDataBuffer;return new Uint32Array(t,t.byteOffset,1)[0]},Gh=(e,t)=>{let r=[],i=[],a=[],s=hd(e);if(t.antialias!==0)throw Error("Only default value (0) for Antialias attribute is supported");ed(e.inputs,t,s,r,i,a),e.compute(cd(e.inputs[0],t,s,r,i,a),{inputs:[0]})},Fh=e=>{let t=e.antialias,r=e.axes,i=e.coordinateTransformMode,a=e.cubicCoeffA,s=e.excludeOutside!==0,n=e.extrapolationValue,u=e.keepAspectRatioPolicy,l=e.mode,p=e.nearestMode===""?"simple":e.nearestMode;return he({antialias:t,axes:r,coordinateTransformMode:i,cubicCoeffA:a,excludeOutside:s,extrapolationValue:n,keepAspectRatioPolicy:u,mode:l,nearestMode:p})}}),fd,md,Hh,W0=P(()=>{"use strict";Y(),te(),re(),fd=e=>{if(!e||e.length<3)throw new Error("layerNorm requires at least 3 inputs.");let t=e[0],r=e[1],i=e[2];if(t.dataType!==r.dataType||t.dataType!==i.dataType)throw new Error("All inputs must have the same data type");if(t.dims.length!==3&&t.dims.length!==2)throw new Error("Input must be 2D or 3D");if(r.dims.length!==3&&r.dims.length!==2)throw new Error("Skip must be 2D or 3D");let a=t.dims[t.dims.length-1],s=t.dims[t.dims.length-2];if(r.dims[r.dims.length-1]!==a)throw new Error("Skip must have the same hidden size as input");if(r.dims[r.dims.length-2]!==s)throw new Error("Skip must have the same sequence length as input");if(i.dims.length!==1)throw new Error("Gamma must be 1D");if(i.dims[i.dims.length-1]!==a)throw new Error("Gamma must have the same hidden size as input");if(e.length>3){let n=e[3];if(n.dims.length!==1)throw new Error("Beta must be 1D");if(n.dims[n.dims.length-1]!==a)throw new Error("Beta must have the same hidden size as input")}if(e.length>4){let n=e[4];if(n.dims.length!==1)throw new Error("Bias must be 1D");if(n.dims[n.dims.length-1]!==a)throw new Error("Bias must have the same hidden size as input")}},md=(e,t,r,i)=>{let a=t.simplified,s=e[0].dims,n=B.size(s),u=s,l=n,p=s.slice(-1)[0],h=i?s.slice(0,-1).concat(1):[],f=!a&&e.length>3,g=e.length>4,_=i&&r>1,y=i&&r>2,$=r>3,S=64,v=$e(p),w=[{type:12,data:l},{type:12,data:v},{type:12,data:p},{type:1,data:t.epsilon}],k=E=>{let C=[{name:"output_size",type:"u32"},{name:"components",type:"u32"},{name:"hidden_size",type:"u32"},{name:"epsilon",type:"f32"}],O=[M("x",e[0].dataType,e[0].dims,v),M("skip",e[1].dataType,e[1].dims,v),M("gamma",e[2].dataType,e[2].dims,v)];f&&O.push(M("beta",e[3].dataType,e[3].dims,v)),g&&O.push(M("bias",e[4].dataType,e[4].dims,v)),O.push(F("output",e[0].dataType,u,v)),_&&O.push(F("mean_output",1,h)),y&&O.push(F("inv_std_output",1,h)),$&&O.push(F("input_skip_bias_sum",e[0].dataType,u,v));let x=ke(e[0].dataType),D=ke(1,v);return`
 
-      ${k.registerUniforms(C).declareVariables(...O)}
+      ${E.registerUniforms(C).declareVariables(...O)}
       var<workgroup> sum_shared : array<${D}, ${S}>;
       var<workgroup> sum_squared_shared : array<${D}, ${S}>;
 
-      ${k.mainStart([S,1,1])}
+      ${E.mainStart([S,1,1])}
         let ix = local_id.x;
         let iy = global_id.x / ${S};
 
@@ -2613,7 +2613,7 @@ fn calculateOutputIndex(index: u32) -> u32 {
           let value = input_value + skip_value + bias_value;
           ${$?"input_skip_bias_sum[offset + i] = value;":""}
           output[offset + i] = value;
-          let f32_value = ${qt(x,v,"value")};
+          let f32_value = ${Lt(x,v,"value")};
           sum_shared[ix] += f32_value;
           sum_squared_shared[ix] += f32_value * f32_value;
         }
@@ -2641,7 +2641,7 @@ fn calculateOutputIndex(index: u32) -> u32 {
             ${x}(inv_std_dev) * gamma[offset1d + i]
             ${f?"+ beta[offset1d + i]":""};
         }
-      }`},I=[{dims:u,dataType:e[0].dataType}];return r>1&&I.push({dims:h,dataType:1}),r>2&&I.push({dims:h,dataType:1}),r>3&&I.push({dims:s,dataType:e[0].dataType}),{name:"SkipLayerNormalization",shaderCache:{hint:`${v};${_};${y};${$}`,inputDependencies:e.map((k,C)=>"type")},getShaderSource:E,getRunData:()=>({outputs:I,dispatchGroup:{x:Math.ceil(l/p)},programUniforms:w})}},Hh=(e,t)=>{fd(e.inputs);let r=[0];e.outputCount>1&&r.push(-3),e.outputCount>2&&r.push(-3),e.outputCount>3&&r.push(3),e.compute(md(e.inputs,t,e.outputCount,!1),{outputs:r})}}),gd,ir,yd,da,_d,wd,jh,Kh,q0=P(()=>{"use strict";Y(),te(),ve(),re(),gd=(e,t)=>{if(!e||e.length<1)throw new Error("too few inputs");if(t.axes.length!==0){if(t.axes.length!==t.starts.length||t.axes.length!==t.ends.length)throw new Error("axes, starts and ends must have the same length")}else if(t.starts.length!==t.ends.length)throw new Error("starts and ends must have the same length");e.slice(1).forEach((r,i)=>{if(e[i+1].dataType!==6&&e[i+1].dataType!==7)throw new Error(`Input ${i} must be an array of int32 or int64`)})},ir=(e,t)=>{let r=[];if(e.length>t)if(e[t].dataType===7)e[t].getBigInt64Array().forEach(i=>r.push(Number(i)));else if(e[t].dataType===6)e[t].getInt32Array().forEach(i=>r.push(Number(i)));else throw new Error(`Input ${t} must be an array of int32 or int64`);return r},yd=(e,t)=>{if(e.length>1){let r=ir(e,1),i=ir(e,2),a=ir(e,3);return a.length===0&&(a=[...Array(e[0].dims.length).keys()]),he({starts:r,ends:i,axes:a})}else return t},da=(e,t,r,i,a)=>{let s=e;return e<0&&(s+=r[i[t]]),a[t]<0?Math.max(0,Math.min(s,r[i[t]]-1)):Math.max(0,Math.min(s,r[i[t]]))},_d=(e,t,r)=>`fn calculateInputIndices(output_indices: ${t.type.indices}) -> ${e.type.indices} {
+      }`},I=[{dims:u,dataType:e[0].dataType}];return r>1&&I.push({dims:h,dataType:1}),r>2&&I.push({dims:h,dataType:1}),r>3&&I.push({dims:s,dataType:e[0].dataType}),{name:"SkipLayerNormalization",shaderCache:{hint:`${v};${_};${y};${$}`,inputDependencies:e.map((E,C)=>"type")},getShaderSource:k,getRunData:()=>({outputs:I,dispatchGroup:{x:Math.ceil(l/p)},programUniforms:w})}},Hh=(e,t)=>{fd(e.inputs);let r=[0];e.outputCount>1&&r.push(-3),e.outputCount>2&&r.push(-3),e.outputCount>3&&r.push(3),e.compute(md(e.inputs,t,e.outputCount,!1),{outputs:r})}}),gd,ir,yd,da,_d,wd,jh,Kh,L0=P(()=>{"use strict";Y(),te(),ve(),re(),gd=(e,t)=>{if(!e||e.length<1)throw new Error("too few inputs");if(t.axes.length!==0){if(t.axes.length!==t.starts.length||t.axes.length!==t.ends.length)throw new Error("axes, starts and ends must have the same length")}else if(t.starts.length!==t.ends.length)throw new Error("starts and ends must have the same length");e.slice(1).forEach((r,i)=>{if(e[i+1].dataType!==6&&e[i+1].dataType!==7)throw new Error(`Input ${i} must be an array of int32 or int64`)})},ir=(e,t)=>{let r=[];if(e.length>t)if(e[t].dataType===7)e[t].getBigInt64Array().forEach(i=>r.push(Number(i)));else if(e[t].dataType===6)e[t].getInt32Array().forEach(i=>r.push(Number(i)));else throw new Error(`Input ${t} must be an array of int32 or int64`);return r},yd=(e,t)=>{if(e.length>1){let r=ir(e,1),i=ir(e,2),a=ir(e,3);return a.length===0&&(a=[...Array(e[0].dims.length).keys()]),he({starts:r,ends:i,axes:a})}else return t},da=(e,t,r,i,a)=>{let s=e;return e<0&&(s+=r[i[t]]),a[t]<0?Math.max(0,Math.min(s,r[i[t]]-1)):Math.max(0,Math.min(s,r[i[t]]))},_d=(e,t,r)=>`fn calculateInputIndices(output_indices: ${t.type.indices}) -> ${e.type.indices} {
           var input_indices: ${e.type.indices};
           var carry = 0u;
           for (var i = ${r.length-1}; i >= 0; i--) {
@@ -2659,7 +2659,7 @@ fn calculateOutputIndex(index: u32) -> u32 {
             ${e.indicesSet("input_indices","i","input_index")};
           }
           return input_indices;
-      }`,wd=(e,t)=>{let r=e[0].dims,i=B.size(r),a=t.axes.length>0?B.normalizeAxes(t.axes,r.length):[...Array(r.length).keys()],s=ir(e,4);s.forEach(v=>v!==0||(()=>{throw new Error("step cannot be 0")})),s.length===0&&(s=Array(a.length).fill(1));let n=t.starts.map((v,w)=>da(v,w,r,a,s)),u=t.ends.map((v,w)=>da(v,w,r,a,s));if(a.length!==n.length||a.length!==u.length)throw new Error("start, ends and axes should have the same number of elements");if(a.length!==r.length)for(let v=0;v<r.length;++v)a.includes(v)||(n.splice(v,0,0),u.splice(v,0,r[v]),s.splice(v,0,1));let l=s.map(v=>Math.sign(v));s.forEach((v,w,E)=>{if(v<0){let I=(u[w]-n[w])/v,k=n[w],C=k+I*s[w];n[w]=C,u[w]=k,E[w]=-v}});let p=r.slice(0);a.forEach((v,w)=>{p[v]=Math.ceil((u[v]-n[v])/s[v])});let h={dims:p,dataType:e[0].dataType},f=F("output",e[0].dataType,p.length),g=N("input",e[0].dataType,e[0].dims.length),_=B.size(p),y=[{name:"outputSize",type:"u32"},{name:"starts",type:"u32",length:n.length},{name:"signs",type:"i32",length:l.length},{name:"steps",type:"u32",length:s.length}],$=[{type:12,data:_},{type:12,data:n},{type:6,data:l},{type:12,data:s},...Z(e[0].dims,p)],S=v=>`
+      }`,wd=(e,t)=>{let r=e[0].dims,i=B.size(r),a=t.axes.length>0?B.normalizeAxes(t.axes,r.length):[...Array(r.length).keys()],s=ir(e,4);s.forEach(v=>v!==0||(()=>{throw new Error("step cannot be 0")})),s.length===0&&(s=Array(a.length).fill(1));let n=t.starts.map((v,w)=>da(v,w,r,a,s)),u=t.ends.map((v,w)=>da(v,w,r,a,s));if(a.length!==n.length||a.length!==u.length)throw new Error("start, ends and axes should have the same number of elements");if(a.length!==r.length)for(let v=0;v<r.length;++v)a.includes(v)||(n.splice(v,0,0),u.splice(v,0,r[v]),s.splice(v,0,1));let l=s.map(v=>Math.sign(v));s.forEach((v,w,k)=>{if(v<0){let I=(u[w]-n[w])/v,E=n[w],C=E+I*s[w];n[w]=C,u[w]=E,k[w]=-v}});let p=r.slice(0);a.forEach((v,w)=>{p[v]=Math.ceil((u[v]-n[v])/s[v])});let h={dims:p,dataType:e[0].dataType},f=F("output",e[0].dataType,p.length),g=M("input",e[0].dataType,e[0].dims.length),_=B.size(p),y=[{name:"outputSize",type:"u32"},{name:"starts",type:"u32",length:n.length},{name:"signs",type:"i32",length:l.length},{name:"steps",type:"u32",length:s.length}],$=[{type:12,data:_},{type:12,data:n},{type:6,data:l},{type:12,data:s},...Z(e[0].dims,p)],S=v=>`
       ${v.registerUniforms(y).declareVariables(g,f)}
         ${_d(g,f,r)}
         ${v.mainStart()}
@@ -2667,17 +2667,17 @@ fn calculateOutputIndex(index: u32) -> u32 {
           let output_indices = ${f.offsetToIndices("global_idx")};
           let input_indices = calculateInputIndices(output_indices);
           ${f.setByOffset("global_idx",g.getByIndices("input_indices"))}
-      }`;return{name:"Slice",shaderCache:{hint:`${l.length}_${n.length}_${s.length}`,inputDependencies:["rank"]},getShaderSource:S,getRunData:()=>({outputs:[h],dispatchGroup:{x:Math.ceil(i/64)},programUniforms:$})}},jh=(e,t)=>{gd(e.inputs,t);let r=yd(e.inputs,t);e.compute(wd(e.inputs,r),{inputs:[0]})},Kh=e=>{let t=e.starts,r=e.ends,i=e.axes;return he({starts:t,ends:r,axes:i})}}),bd,$d,Zh,Qh,L0=P(()=>{"use strict";Y(),te(),ve(),wt(),re(),bd=e=>{if(!e||e.length!==1)throw new Error("Softmax op requires 1 input.")},$d=(e,t)=>{let r=e.inputs[0],i=r.dims,a=B.size(i),s=i.length,n=B.normalizeAxis(t.axis,s),u=n<i.length-1,l,p=[];u?(p=Array.from({length:s},(O,x)=>x),p[n]=s-1,p[s-1]=n,l=e.compute(Me(r,p),{inputs:[r],outputs:[-1]})[0]):l=r;let h=l.dims,f=h[s-1],g=a/f,_=$e(f),y=f/_,$=64;g===1&&($=256);let S=(O,x)=>x===4?`max(max(${O}.x, ${O}.y), max(${O}.z, ${O}.w))`:x===2?`max(${O}.x, ${O}.y)`:x===3?`max(max(${O}.x, ${O}.y), ${O}.z)`:O,v=N("x",l.dataType,l.dims,_),w=F("result",l.dataType,l.dims,_),E=v.type.value,I=Ee(l.dataType)==="f32"?`var threadMax = ${E}(-3.4028234663852886e+38f);`:`var threadMax = ${E}(-65504.0h);`,k=O=>`
-      var<workgroup> rowMaxShared : ${E};
-      var<workgroup> rowSumShared : ${E};
-      var<workgroup> threadShared : array<${E}, ${$}>;
+      }`;return{name:"Slice",shaderCache:{hint:`${l.length}_${n.length}_${s.length}`,inputDependencies:["rank"]},getShaderSource:S,getRunData:()=>({outputs:[h],dispatchGroup:{x:Math.ceil(i/64)},programUniforms:$})}},jh=(e,t)=>{gd(e.inputs,t);let r=yd(e.inputs,t);e.compute(wd(e.inputs,r),{inputs:[0]})},Kh=e=>{let t=e.starts,r=e.ends,i=e.axes;return he({starts:t,ends:r,axes:i})}}),bd,$d,Zh,Qh,q0=P(()=>{"use strict";Y(),te(),ve(),wt(),re(),bd=e=>{if(!e||e.length!==1)throw new Error("Softmax op requires 1 input.")},$d=(e,t)=>{let r=e.inputs[0],i=r.dims,a=B.size(i),s=i.length,n=B.normalizeAxis(t.axis,s),u=n<i.length-1,l,p=[];u?(p=Array.from({length:s},(O,x)=>x),p[n]=s-1,p[s-1]=n,l=e.compute(Ne(r,p),{inputs:[r],outputs:[-1]})[0]):l=r;let h=l.dims,f=h[s-1],g=a/f,_=$e(f),y=f/_,$=64;g===1&&($=256);let S=(O,x)=>x===4?`max(max(${O}.x, ${O}.y), max(${O}.z, ${O}.w))`:x===2?`max(${O}.x, ${O}.y)`:x===3?`max(max(${O}.x, ${O}.y), ${O}.z)`:O,v=M("x",l.dataType,l.dims,_),w=F("result",l.dataType,l.dims,_),k=v.type.value,I=ke(l.dataType)==="f32"?`var threadMax = ${k}(-3.4028234663852886e+38f);`:`var threadMax = ${k}(-65504.0h);`,E=O=>`
+      var<workgroup> rowMaxShared : ${k};
+      var<workgroup> rowSumShared : ${k};
+      var<workgroup> threadShared : array<${k}, ${$}>;
 
-      fn getValue(row: i32, col: i32, row_stride: i32) -> ${E} {
+      fn getValue(row: i32, col: i32, row_stride: i32) -> ${k} {
         let index = row * row_stride + col;
         return x[index];
       }
 
-      fn setValue(row: i32, col: i32, row_stride: i32, value: ${E}) {
+      fn setValue(row: i32, col: i32, row_stride: i32, value: ${k}) {
         let index = row * row_stride + col;
         result[index] = value;
       }
@@ -2710,12 +2710,12 @@ fn calculateOutputIndex(index: u32) -> u32 {
           workgroupBarrier();
         }
         if (lindex == 0) {
-          rowMaxShared = ${E}(${S("threadShared[0]",_)});
+          rowMaxShared = ${k}(${S("threadShared[0]",_)});
         }
         workgroupBarrier();
 
         // find the rows sum
-        var threadSum = ${E}(0.0);
+        var threadSum = ${k}(0.0);
         for (var col = lindex; col < cols; col += wg) {
           let subExp = exp(getValue(row, col, row_stride) - rowMaxShared);
           threadSum += subExp;
@@ -2730,7 +2730,7 @@ fn calculateOutputIndex(index: u32) -> u32 {
           workgroupBarrier();
         }
         if (lindex == 0) {
-          rowSumShared = ${E}(${yt("threadShared[0]",_)});
+          rowSumShared = ${k}(${yt("threadShared[0]",_)});
         }
         workgroupBarrier();
 
@@ -2738,10 +2738,10 @@ fn calculateOutputIndex(index: u32) -> u32 {
         for (var col = lindex; col < cols; col += wg) {
           var value = exp(getValue(row, col, row_stride) - rowMaxShared) / rowSumShared;
           // max operation protects against NaN since all values should be >=0
-          value = max(value, ${E}(0.0));
+          value = max(value, ${k}(0.0));
           setValue(row, col, row_stride, value);
         }
-      }`,C=e.compute({name:"Softmax",shaderCache:{hint:`${_};${$}`,inputDependencies:["type"]},getRunData:()=>({outputs:[{dims:h,dataType:l.dataType}],dispatchGroup:{x:g},programUniforms:[{type:6,data:y}]}),getShaderSource:k},{inputs:[l],outputs:[u?-1:0]})[0];u&&e.compute(Me(C,p),{inputs:[C]})},Zh=(e,t)=>{bd(e.inputs),$d(e,t)},Qh=e=>he({axis:e.axis})}),pa,vd,xd,Sd,Xh,V0=P(()=>{"use strict";Y(),te(),re(),pa=e=>Array.from(e.getBigInt64Array(),Number),vd=e=>{if(!e||e.length!==2)throw new Error("Tile requires 2 inputs.");if(e[0].dataType!==1&&e[0].dataType!==10&&e[0].dataType!==6&&e[0].dataType!==12)throw new Error("Tile only support float, float16, int32, and uint32 data types");if(e[1].dataType!==7)throw new Error("Tile `repeats` input should be of int64 data type");if(e[1].dims.length!==1)throw new Error("Tile `repeats` input should be 1-D");if(pa(e[1]).length!==e[0].dims.length)throw new Error("Tile `repeats` input should have same number of elements as rank of input data tensor")},xd=(e,t)=>{let r=[];for(let i=0;i<e.length;++i)r.push(e[i]*t[i]);return r},Sd=(e,t)=>{let r=e[0].dims,i=t??pa(e[1]),a=xd(r,i),s=B.size(a),n=e[0].dataType,u=N("input",n,r.length),l=F("output",n,a.length),p=h=>`
+      }`,C=e.compute({name:"Softmax",shaderCache:{hint:`${_};${$}`,inputDependencies:["type"]},getRunData:()=>({outputs:[{dims:h,dataType:l.dataType}],dispatchGroup:{x:g},programUniforms:[{type:6,data:y}]}),getShaderSource:E},{inputs:[l],outputs:[u?-1:0]})[0];u&&e.compute(Ne(C,p),{inputs:[C]})},Zh=(e,t)=>{bd(e.inputs),$d(e,t)},Qh=e=>he({axis:e.axis})}),pa,vd,xd,Sd,Xh,V0=P(()=>{"use strict";Y(),te(),re(),pa=e=>Array.from(e.getBigInt64Array(),Number),vd=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(pa(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")},xd=(e,t)=>{let r=[];for(let i=0;i<e.length;++i)r.push(e[i]*t[i]);return r},Sd=(e,t)=>{let r=e[0].dims,i=t??pa(e[1]),a=xd(r,i),s=B.size(a),n=e[0].dataType,u=M("input",n,r.length),l=F("output",n,a.length),p=h=>`
       const inputShape = ${u.indices(...r)};
       ${h.registerUniform("output_size","u32").declareVariables(u,l)}
       ${h.mainStart()}
@@ -2755,7 +2755,7 @@ fn calculateOutputIndex(index: u32) -> u32 {
         ${u.indicesSet("input_indices","i","input_dim_value")}
       }
       ${l.setByOffset("global_idx",u.getByIndices("input_indices"))}
-    }`;return{name:"Tile",shaderCache:{hint:`${i}`,inputDependencies:["rank"]},getRunData:()=>({outputs:[{dims:a,dataType:e[0].dataType}],dispatchGroup:{x:Math.ceil(s/64)},programUniforms:[{type:12,data:s},...Z(e[0].dims,a)]}),getShaderSource:p}},Xh=e=>{vd(e.inputs),e.compute(Sd(e.inputs),{inputs:[0]})}}),Td,Id,Yh,G0=P(()=>{"use strict";Y(),te(),re(),Td=(e,t,r,i,a)=>{let s=F("output_data",a,r.length,4),n=N("a_data",t[1].dataType,t[1].dims.length,4),u=N("b_data",t[2].dataType,t[2].dims.length,4),l=N("c_data",t[0].dataType,t[0].dims.length,4),p,h=(f,g,_)=>`select(${g}, ${f}, ${_})`;if(!i)p=s.setByOffset("global_idx",h(n.getByOffset("global_idx"),u.getByOffset("global_idx"),l.getByOffset("global_idx")));else{let f=(g,_,y="")=>{let $=`a_data[index_a${_}][component_a${_}]`,S=`b_data[index_b${_}][component_b${_}]`,v=`bool(c_data[index_c${_}] & (0xffu << (component_c${_} * 8)))`;return`
+    }`;return{name:"Tile",shaderCache:{hint:`${i}`,inputDependencies:["rank"]},getRunData:()=>({outputs:[{dims:a,dataType:e[0].dataType}],dispatchGroup:{x:Math.ceil(s/64)},programUniforms:[{type:12,data:s},...Z(e[0].dims,a)]}),getShaderSource:p}},Xh=e=>{vd(e.inputs),e.compute(Sd(e.inputs),{inputs:[0]})}}),Td,Id,Yh,G0=P(()=>{"use strict";Y(),te(),re(),Td=(e,t,r,i,a)=>{let s=F("output_data",a,r.length,4),n=M("a_data",t[1].dataType,t[1].dims.length,4),u=M("b_data",t[2].dataType,t[2].dims.length,4),l=M("c_data",t[0].dataType,t[0].dims.length,4),p,h=(f,g,_)=>`select(${g}, ${f}, ${_})`;if(!i)p=s.setByOffset("global_idx",h(n.getByOffset("global_idx"),u.getByOffset("global_idx"),l.getByOffset("global_idx")));else{let f=(g,_,y="")=>{let $=`a_data[index_a${_}][component_a${_}]`,S=`b_data[index_b${_}][component_b${_}]`,v=`bool(c_data[index_c${_}] & (0xffu << (component_c${_} * 8)))`;return`
             let output_indices${_} = ${s.offsetToIndices(`global_idx * 4u + ${_}u`)};
             let offset_a${_} = ${n.broadcastedIndicesToOffset(`output_indices${_}`,s)};
             let offset_b${_} = ${u.broadcastedIndicesToOffset(`output_indices${_}`,s)};
@@ -2783,10 +2783,10 @@ fn calculateOutputIndex(index: u32) -> u32 {
         ${e.mainStart()}
         ${e.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.vec_size")}
         ${p}
-      }`},Id=e=>{let t=e[1].dims,r=e[2].dims,i=e[0].dims,a=e[1].dataType,s=!(B.areEqual(t,r)&&B.areEqual(r,i)),n=t,u=B.size(t);if(s){let p=Lt.calcShape(Lt.calcShape(t,r,!1),i,!1);if(!p)throw new Error("Can't perform where op on the given tensors");n=p,u=B.size(n)}let l=Math.ceil(u/4);return{name:"Where",shaderCache:{inputDependencies:["rank","rank","rank"]},getShaderSource:p=>Td(p,e,n,s,a),getRunData:()=>({outputs:[{dims:n,dataType:a}],dispatchGroup:{x:Math.ceil(u/64/4)},programUniforms:[{type:12,data:l},...Z(i,t,r,n)]})}},Yh=e=>{e.compute(Id(e.inputs))}}),Jh,F0=P(()=>{"use strict";n0(),Ha(),s0(),o0(),u0(),l0(),d0(),m0(),y0(),_0(),w0(),b0(),$0(),v0(),x0(),S0(),T0(),I0(),E0(),k0(),z0(),C0(),A0(),O0(),B0(),_h(),R0(),N0(),M0(),D0(),P0(),Fa(),U0(),xh(),W0(),q0(),L0(),$h(),V0(),wt(),ja(),G0(),Jh=new Map([["Abs",[Hp]],["Acos",[jp]],["Acosh",[Kp]],["Add",[kc]],["ArgMax",[Lp,va]],["ArgMin",[qp,va]],["Asin",[Zp]],["Asinh",[Qp]],["Atan",[Xp]],["Atanh",[Yp]],["Attention",[Vp]],["AveragePool",[Oh,Ah]],["BatchNormalization",[Gp]],["BiasAdd",[Fp]],["BiasSplitGelu",[Ec]],["Cast",[ec,Jp]],["Ceil",[rc]],["Clip",[tc]],["Concat",[Pc,Uc]],["Conv",[ka,Ea]],["ConvTranspose",[Zc,Kc]],["Cos",[ic]],["Cosh",[ac]],["CumSum",[Qc,Xc]],["DepthToSpace",[Yc,Jc]],["DequantizeLinear",[Uh,Wh]],["Div",[zc]],["Einsum",[eh,th]],["Elu",[nc,ur]],["Equal",[Cc]],["Erf",[sc]],["Exp",[oc]],["Expand",[rh]],["FastGelu",[ih]],["Floor",[uc]],["FusedConv",[ka,Ea]],["Gather",[nh,ah]],["GatherElements",[ph,dh]],["GatherBlockQuantized",[uh,lh]],["GatherND",[sh,oh]],["Gelu",[lc]],["Gemm",[hh,ch]],["GlobalAveragePool",[Rh,Bh]],["GlobalMaxPool",[Ph,Dh]],["Greater",[Rc]],["GreaterOrEqual",[Mc]],["GridSample",[fh,mh]],["GroupQueryAttention",[Sh]],["HardSigmoid",[yc,gc]],["InstanceNormalization",[Th]],["LayerNormalization",[Ih]],["LeakyRelu",[dc,ur]],["Less",[Nc]],["LessOrEqual",[Dc]],["Log",[Tc]],["MatMul",[Eh]],["MatMulNBits",[kh,zh]],["MaxPool",[Nh,Mh]],["Mul",[Ac]],["MultiHeadAttention",[yh,gh]],["Neg",[cc]],["Not",[pc]],["Pad",[Ch]],["Pow",[Oc]],["QuickGelu",[Ic,ur]],["Range",[qh]],["Reciprocal",[hc]],["ReduceMin",[Mp]],["ReduceMean",[Ap]],["ReduceMax",[Np]],["ReduceSum",[Pp]],["ReduceProd",[Dp]],["ReduceL1",[Op]],["ReduceL2",[Bp]],["ReduceLogSum",[Wp]],["ReduceLogSumExp",[Rp]],["ReduceSumSquare",[Up]],["Relu",[fc]],["Resize",[Gh,Fh]],["RotaryEmbedding",[vh]],["ScatterND",[Vh,Lh]],["Sigmoid",[mc]],["Sin",[_c]],["Sinh",[wc]],["Slice",[jh,Kh]],["SkipLayerNormalization",[Hh]],["Split",[wh,bh]],["Sqrt",[bc]],["Softmax",[Zh,Qh]],["Sub",[Bc]],["Tan",[$c]],["Tanh",[vc]],["ThresholdedRelu",[Sc,ur]],["Tile",[Xh]],["Transpose",[wp,bp]],["Where",[Yh]]])}),ef,H0=P(()=>{"use strict";We(),nt(),re(),ef=class{constructor(e){this.backend=e,this.repo=new Map,this.attributesBound=!1}getArtifact(e){return this.repo.get(e)}setArtifact(e,t){this.repo.set(e,t)}run(e,t,r,i,a){Ze(e.programInfo.name);let s=this.backend.device,n=this.backend.getComputePassEncoder();this.backend.writeTimestamp(this.backend.pendingDispatchNumber*2);let u=[];for(let p of t)u.push({binding:u.length,resource:{buffer:p.buffer}});for(let p of r)u.push({binding:u.length,resource:{buffer:p.buffer}});a&&u.push({binding:u.length,resource:a});let l=s.createBindGroup({layout:e.computePipeline.getBindGroupLayout(0),entries:u,label:e.programInfo.name});if(this.backend.sessionStatus==="capturing"){let p={kernelId:this.backend.currentKernelId,computePipeline:e.computePipeline,bindGroup:l,dispatchGroup:i};this.backend.capturedCommandList.get(this.backend.currentSessionId).push(p)}n.setPipeline(e.computePipeline),n.setBindGroup(0,l),n.dispatchWorkgroups(...i),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(),Ue(e.programInfo.name)}dispose(){}build(e,t){Ze(e.name);let r=this.backend.device,i=[];[{feature:"shader-f16",extension:"f16"},{feature:"subgroups",extension:"subgroups"}].forEach(p=>{r.features.has(p.feature)&&i.push(`enable ${p.extension};`)});let a=_p(t,this.backend.device.limits),s=e.getShaderSource(a),n=`${i.join(`
+      }`},Id=e=>{let t=e[1].dims,r=e[2].dims,i=e[0].dims,a=e[1].dataType,s=!(B.areEqual(t,r)&&B.areEqual(r,i)),n=t,u=B.size(t);if(s){let p=qt.calcShape(qt.calcShape(t,r,!1),i,!1);if(!p)throw new Error("Can't perform where op on the given tensors");n=p,u=B.size(n)}let l=Math.ceil(u/4);return{name:"Where",shaderCache:{inputDependencies:["rank","rank","rank"]},getShaderSource:p=>Td(p,e,n,s,a),getRunData:()=>({outputs:[{dims:n,dataType:a}],dispatchGroup:{x:Math.ceil(u/64/4)},programUniforms:[{type:12,data:l},...Z(i,t,r,n)]})}},Yh=e=>{e.compute(Id(e.inputs))}}),Jh,F0=P(()=>{"use strict";n0(),Ha(),s0(),o0(),u0(),l0(),d0(),m0(),y0(),_0(),w0(),b0(),$0(),v0(),x0(),S0(),T0(),I0(),k0(),E0(),z0(),C0(),A0(),O0(),B0(),_h(),R0(),M0(),N0(),D0(),P0(),Fa(),U0(),xh(),W0(),L0(),q0(),$h(),V0(),wt(),ja(),G0(),Jh=new Map([["Abs",[Hp]],["Acos",[jp]],["Acosh",[Kp]],["Add",[Ec]],["ArgMax",[qp,va]],["ArgMin",[Lp,va]],["Asin",[Zp]],["Asinh",[Qp]],["Atan",[Xp]],["Atanh",[Yp]],["Attention",[Vp]],["AveragePool",[Oh,Ah]],["BatchNormalization",[Gp]],["BiasAdd",[Fp]],["BiasSplitGelu",[kc]],["Cast",[ec,Jp]],["Ceil",[rc]],["Clip",[tc]],["Concat",[Pc,Uc]],["Conv",[Ea,ka]],["ConvTranspose",[Zc,Kc]],["Cos",[ic]],["Cosh",[ac]],["CumSum",[Qc,Xc]],["DepthToSpace",[Yc,Jc]],["DequantizeLinear",[Uh,Wh]],["Div",[zc]],["Einsum",[eh,th]],["Elu",[nc,ur]],["Equal",[Cc]],["Erf",[sc]],["Exp",[oc]],["Expand",[rh]],["FastGelu",[ih]],["Floor",[uc]],["FusedConv",[Ea,ka]],["Gather",[nh,ah]],["GatherElements",[ph,dh]],["GatherBlockQuantized",[uh,lh]],["GatherND",[sh,oh]],["Gelu",[lc]],["Gemm",[hh,ch]],["GlobalAveragePool",[Rh,Bh]],["GlobalMaxPool",[Ph,Dh]],["Greater",[Rc]],["GreaterOrEqual",[Nc]],["GridSample",[fh,mh]],["GroupQueryAttention",[Sh]],["HardSigmoid",[yc,gc]],["InstanceNormalization",[Th]],["LayerNormalization",[Ih]],["LeakyRelu",[dc,ur]],["Less",[Mc]],["LessOrEqual",[Dc]],["Log",[Tc]],["MatMul",[kh]],["MatMulNBits",[Eh,zh]],["MaxPool",[Mh,Nh]],["Mul",[Ac]],["MultiHeadAttention",[yh,gh]],["Neg",[cc]],["Not",[pc]],["Pad",[Ch]],["Pow",[Oc]],["QuickGelu",[Ic,ur]],["Range",[Lh]],["Reciprocal",[hc]],["ReduceMin",[Np]],["ReduceMean",[Ap]],["ReduceMax",[Mp]],["ReduceSum",[Pp]],["ReduceProd",[Dp]],["ReduceL1",[Op]],["ReduceL2",[Bp]],["ReduceLogSum",[Wp]],["ReduceLogSumExp",[Rp]],["ReduceSumSquare",[Up]],["Relu",[fc]],["Resize",[Gh,Fh]],["RotaryEmbedding",[vh]],["ScatterND",[Vh,qh]],["Sigmoid",[mc]],["Sin",[_c]],["Sinh",[wc]],["Slice",[jh,Kh]],["SkipLayerNormalization",[Hh]],["Split",[wh,bh]],["Sqrt",[bc]],["Softmax",[Zh,Qh]],["Sub",[Bc]],["Tan",[$c]],["Tanh",[vc]],["ThresholdedRelu",[Sc,ur]],["Tile",[Xh]],["Transpose",[wp,bp]],["Where",[Yh]]])}),ef,H0=P(()=>{"use strict";We(),nt(),re(),ef=class{constructor(e){this.backend=e,this.repo=new Map,this.attributesBound=!1}getArtifact(e){return this.repo.get(e)}setArtifact(e,t){this.repo.set(e,t)}run(e,t,r,i,a){Ze(e.programInfo.name);let s=this.backend.device,n=this.backend.getComputePassEncoder();this.backend.writeTimestamp(this.backend.pendingDispatchNumber*2);let u=[];for(let p of t)u.push({binding:u.length,resource:{buffer:p.buffer}});for(let p of r)u.push({binding:u.length,resource:{buffer:p.buffer}});a&&u.push({binding:u.length,resource:a});let l=s.createBindGroup({layout:e.computePipeline.getBindGroupLayout(0),entries:u,label:e.programInfo.name});if(this.backend.sessionStatus==="capturing"){let p={kernelId:this.backend.currentKernelId,computePipeline:e.computePipeline,bindGroup:l,dispatchGroup:i};this.backend.capturedCommandList.get(this.backend.currentSessionId).push(p)}n.setPipeline(e.computePipeline),n.setBindGroup(0,l),n.dispatchWorkgroups(...i),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(),Ue(e.programInfo.name)}dispose(){}build(e,t){Ze(e.name);let r=this.backend.device,i=[];[{feature:"shader-f16",extension:"f16"},{feature:"subgroups",extension:"subgroups"}].forEach(p=>{r.features.has(p.feature)&&i.push(`enable ${p.extension};`)});let a=_p(t,this.backend.device.limits),s=e.getShaderSource(a),n=`${i.join(`
 `)}
 ${a.additionalImplementations}
-${s}`,u=r.createShaderModule({code:n,label:e.name});ue("verbose",()=>`[WebGPU] ${e.name} shader code: ${n}`);let l=r.createComputePipeline({compute:{module:u,entryPoint:"main"},layout:"auto",label:e.name});return Ue(e.name),{programInfo:e,computePipeline:l,uniformVariablesInfo:a.variablesInfo}}normalizeDispatchGroupSize(e){let t=typeof e=="number"?e:e.x,r=typeof e=="number"?1:e.y||1,i=typeof e=="number"?1:e.z||1,a=this.backend.device.limits.maxComputeWorkgroupsPerDimension;if(t<=a&&r<=a&&i<=a)return[t,r,i];let s=t*r*i,n=Math.ceil(Math.sqrt(s));if(n>a){if(n=Math.ceil(Math.cbrt(s)),n>a)throw new Error("Total dispatch size exceeds WebGPU maximum.");return[n,n,n]}else return[n,n,1]}}}),tf={};Gt(tf,{WebGpuBackend:()=>rf});var Ed,kd,zd,rf,j0=P(()=>{"use strict";We(),Y(),nt(),hp(),i0(),F0(),H0(),Ed=(e,t)=>{if(t.length!==e.length)throw new Error(`inputDependencies length ${t.length} is not equal to inputTensors length ${e.length}.`);let r=[];for(let i=0;i<e.length;++i){let a=e[i].dataType;switch(t[i]){case"none":{r.push("");break}case"type":{r.push(`${a}`);break}case"rank":{let s=e[i].dims.length;r.push(`${a};${s}`);break}case"dims":{let s=e[i].dims.join(",");r.push(`${a};${s}`);break}default:throw new Error(`unsupported input dependency: ${t[i]}`)}}return r.join("|")},kd=(e,t,r)=>{let i=e.name;return e.shaderCache?.hint&&(i+="["+e.shaderCache.hint+"]"),i+=":"+r+`:${Ed(t,e.shaderCache?.inputDependencies??new Array(t.length).fill("dims"))}`,i},zd=class{constructor(e){e&&(this.architecture=e.architecture,this.vendor=e.vendor)}isArchitecture(e){return this.architecture===e}isVendor(e){return this.vendor===e}},rf=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,t){this.env=e;let r=[],i={requiredLimits:{maxComputeWorkgroupStorageSize:t.limits.maxComputeWorkgroupStorageSize,maxComputeWorkgroupsPerDimension:t.limits.maxComputeWorkgroupsPerDimension,maxStorageBufferBindingSize:t.limits.maxStorageBufferBindingSize,maxBufferSize:t.limits.maxBufferSize,maxComputeInvocationsPerWorkgroup:t.limits.maxComputeInvocationsPerWorkgroup,maxComputeWorkgroupSizeX:t.limits.maxComputeWorkgroupSizeX,maxComputeWorkgroupSizeY:t.limits.maxComputeWorkgroupSizeY,maxComputeWorkgroupSizeZ:t.limits.maxComputeWorkgroupSizeZ},requiredFeatures:r},a=s=>t.features.has(s)&&r.push(s)&&!0;a("chromium-experimental-timestamp-query-inside-passes")||a("timestamp-query"),a("shader-f16"),a("subgroups"),this.device=await t.requestDevice(i),this.adapterInfo=new zd(t.info||await t.requestAdapterInfo()),this.gpuDataManager=gp(this),this.programManager=new ef(this),this.kernels=new Map,this.kernelPersistentData=new Map,this.kernelCustomData=new Map,qa(e.logLevel,!!e.debug),this.device.onuncapturederror=s=>{s.error instanceof GPUValidationError&&console.error(`An uncaught WebGPU validation error was raised: ${s.error.message}`)},Object.defineProperty(this.env.webgpu,"device",{value:this.device,writable:!1,enumerable:!0,configurable:!1}),Object.defineProperty(this.env.webgpu,"adapter",{value:t,writable:!1,enumerable:!0,configurable:!1}),this.setQueryType()}dispose(){typeof this.querySet<"u"&&this.querySet.destroy(),this.gpuDataManager.dispose()}getCommandEncoder(){return this.commandEncoder||(this.commandEncoder=this.device.createCommandEncoder()),this.commandEncoder}getComputePassEncoder(){if(!this.computePassEncoder){let e=this.getCommandEncoder(),t={};this.queryType==="at-passes"&&(t.timestampWrites={querySet:this.querySet,beginningOfPassWriteIndex:this.pendingDispatchNumber*2,endOfPassWriteIndex:this.pendingDispatchNumber*2+1}),this.computePassEncoder=e.beginComputePass(t)}return this.computePassEncoder}endComputePass(){this.computePassEncoder&&(this.computePassEncoder.end(),this.computePassEncoder=null)}flush(){if(!this.commandEncoder)return;Ze(),this.endComputePass();let e;this.queryType!=="none"&&(this.commandEncoder.resolveQuerySet(this.querySet,0,this.pendingDispatchNumber*2,this.queryResolveBuffer,0),e=this.device.createBuffer({size:this.pendingDispatchNumber*2*8,usage:GPUBufferUsage.MAP_READ|GPUBufferUsage.COPY_DST}),this.pendingQueries.set(e,this.pendingKernels),this.pendingKernels=[],this.commandEncoder.copyBufferToBuffer(this.queryResolveBuffer,0,e,0,this.pendingDispatchNumber*2*8)),this.device.queue.submit([this.commandEncoder.finish()]),this.gpuDataManager.refreshPendingBuffers(),this.commandEncoder=null,this.pendingDispatchNumber=0,this.queryType!=="none"&&e.mapAsync(GPUMapMode.READ).then(()=>{let t=new BigUint64Array(e.getMappedRange()),r=this.pendingQueries.get(e);for(let i=0;i<t.length/2;i++){let a=r[i],s=a.kernelId,n=this.kernels.get(s),u=n.kernelType,l=n.kernelName,p=a.programName,h=a.inputTensorViews,f=a.outputTensorViews,g=t[i*2],_=t[i*2+1];typeof this.queryTimeBase>"u"&&(this.queryTimeBase=g);let y=Number(g-this.queryTimeBase),$=Number(_-this.queryTimeBase);if(!Number.isSafeInteger(y)||!Number.isSafeInteger($))throw new RangeError("incorrect timestamp range");if(this.env.webgpu.profiling?.ondata)this.env.webgpu.profiling.ondata({version:1,inputsMetadata:h.map(S=>({dims:S.dims,dataType:at(S.dataType)})),outputsMetadata:f.map(S=>({dims:S.dims,dataType:at(S.dataType)})),kernelId:s,kernelType:u,kernelName:l,programName:p,startTime:y,endTime:$});else{let S="";h.forEach((w,E)=>{S+=`input[${E}]: [${w.dims}] | ${at(w.dataType)}, `});let v="";f.forEach((w,E)=>{v+=`output[${E}]: [${w.dims}] | ${at(w.dataType)}, `}),console.log(`[profiling] kernel "${s}|${u}|${l}|${p}" ${S}${v}start time: ${y} ns, execution time: ${$-y} ns`)}cr("GPU",`${p}::${g}::${_}`)}e.unmap(),this.pendingQueries.delete(e)}),Ue()}run(e,t,r,i,a,s){Ze(e.name);let n=[];for(let w=0;w<t.length;++w){let E=t[w].data;if(E===0)continue;let I=this.gpuDataManager.get(E);if(!I)throw new Error(`no GPU data for input: ${E}`);n.push(I)}let{outputs:u,dispatchGroup:l,programUniforms:p}=e.getRunData(t),h=r.length===0?u.map((w,E)=>E):r;if(h.length!==u.length)throw new Error(`Output size ${h.length} must be equal to ${u.length}.`);let f=[],g=[];for(let w=0;w<u.length;++w){if(!Number.isInteger(h[w])||h[w]<-3||h[w]>=s)throw new Error(`Invalid output index: ${h[w]}`);if(h[w]===-3)continue;let E=h[w]===-1,I=h[w]===-2,k=E||I?a(u[w].dataType,u[w].dims):i(h[w],u[w].dataType,u[w].dims);if(f.push(k),k.data===0)continue;let C=this.gpuDataManager.get(k.data);if(!C)throw new Error(`no GPU data for output: ${k.data}`);if(E&&this.temporaryData.push(C),I){let O=this.kernelPersistentData.get(this.currentKernelId);O||(O=[],this.kernelPersistentData.set(this.currentKernelId,O)),O.push(C)}g.push(C)}if(n.length!==t.length||g.length!==f.length){if(g.length===0)return Ue(e.name),f;throw new Error(`Program ${e.name} has zero-sized tensor(s) in inputs or outputs. This is not supported now.`)}let _;if(p){let w=0,E=[];p.forEach(O=>{let x=typeof O.data=="number"?[O.data]:O.data;if(x.length===0)return;let D=O.type===10?2:4,q,Q;O.type===10?(Q=x.length>4?16:x.length>2?8:x.length*D,q=x.length>4?16:D*x.length):(Q=x.length<=2?x.length*D:16,q=16),w=Math.ceil(w/Q)*Q,E.push(w);let L=O.type===10?8:4;w+=x.length>4?Math.ceil(x.length/L)*q:x.length*D});let I=16;w=Math.ceil(w/I)*I;let k=new ArrayBuffer(w);p.forEach((O,x)=>{let D=E[x],q=typeof O.data=="number"?[O.data]:O.data;if(O.type===6)new Int32Array(k,D,q.length).set(q);else if(O.type===12)new Uint32Array(k,D,q.length).set(q);else if(O.type===10)new Uint16Array(k,D,q.length).set(q);else if(O.type===1)new Float32Array(k,D,q.length).set(q);else throw new Error(`Unsupported uniform type: ${at(O.type)}`)});let C=this.gpuDataManager.create(w,GPUBufferUsage.COPY_DST|GPUBufferUsage.UNIFORM);this.device.queue.writeBuffer(C.buffer,0,k,0,w),this.gpuDataManager.release(C.id),_={offset:0,size:w,buffer:C.buffer}}let y=this.programManager.normalizeDispatchGroupSize(l),$=y[1]===1&&y[2]===1,S=kd(e,t,$),v=this.programManager.getArtifact(S);if(v||(v=this.programManager.build(e,y),this.programManager.setArtifact(S,v),ue("info",()=>`[artifact] key: ${S}, programName: ${e.name}`)),p&&v.uniformVariablesInfo){if(p.length!==v.uniformVariablesInfo.length)throw new Error(`Uniform variables count mismatch: expect ${v.uniformVariablesInfo.length}, got ${p.length} in program "${v.programInfo.name}".`);for(let w=0;w<p.length;w++){let E=p[w],I=E.type,k=typeof E.data=="number"?1:E.data.length,[C,O]=v.uniformVariablesInfo[w];if(I!==C||k!==O)throw new Error(`Uniform variable ${w} mismatch: expect type ${C} with size ${O}, got type ${I} with size ${k} in program "${v.programInfo.name}".`)}}if(ue("info",()=>`[ProgramManager] run "${e.name}" (key=${S}) with ${y[0]}x${y[1]}x${y[2]}`),this.queryType!=="none"||this.sessionStatus==="capturing"){let w={kernelId:this.currentKernelId,programName:v.programInfo.name,inputTensorViews:t,outputTensorViews:f};this.pendingKernels.push(w),this.sessionStatus==="capturing"&&this.capturedPendingKernels.get(this.currentSessionId).push(w)}return this.programManager.run(v,n,g,y,_),Ue(e.name),f}upload(e,t){this.gpuDataManager.upload(e,t)}memcpy(e,t){this.gpuDataManager.memcpy(e,t)}async download(e,t){await this.gpuDataManager.download(e,t)}alloc(e){return this.gpuDataManager.create(e).id}free(e){return this.gpuDataManager.release(e)}createKernel(e,t,r,i){let a=Jh.get(e);if(!a)throw new Error(`kernel not implemented: ${e}`);let s={kernelType:e,kernelName:i,kernelEntry:a[0],attributes:[a[1],r]};this.kernels.set(t,s)}releaseKernel(e){let t=this.kernelPersistentData.get(e);if(t){for(let r of t)this.gpuDataManager.release(r.id);this.kernelPersistentData.delete(e)}this.kernelCustomData.delete(e),this.kernels.delete(e)}computeKernel(e,t,r){let i=this.kernels.get(e);if(!i)throw new Error(`kernel not created: ${e}`);let a=i.kernelType,s=i.kernelName,n=i.kernelEntry,u=i.attributes;if(this.currentKernelId!==null)throw new Error(`kernel "[${a}] ${s}" is not allowed to be called recursively`);this.currentKernelId=e,u[0]&&(u[1]=u[0](u[1]),u[0]=void 0),ue("info",()=>`[WebGPU] Start to run kernel "[${a}] ${s}"...`);let l=this.env.debug;this.temporaryData=[];try{return l&&this.device.pushErrorScope("validation"),n(t,u[1]),0}catch(p){return r.push(Promise.resolve(`[WebGPU] Kernel "[${a}] ${s}" failed. ${p}`)),1}finally{l&&r.push(this.device.popErrorScope().then(p=>p?`GPU validation error for kernel "[${a}] ${s}": ${p.message}`:null));for(let p of this.temporaryData)this.gpuDataManager.release(p.id);this.temporaryData=[],this.currentKernelId=null}}registerBuffer(e,t,r,i){let a=this.sessionExternalDataMapping.get(e);a||(a=new Map,this.sessionExternalDataMapping.set(e,a));let s=a.get(t),n=this.gpuDataManager.registerExternalBuffer(r,i,s);return a.set(t,[n,r]),n}unregisterBuffers(e){let t=this.sessionExternalDataMapping.get(e);t&&(t.forEach(r=>this.gpuDataManager.unregisterExternalBuffer(r[0])),this.sessionExternalDataMapping.delete(e))}getBuffer(e){let t=this.gpuDataManager.get(e);if(!t)throw new Error(`no GPU data for buffer: ${e}`);return t.buffer}createDownloader(e,t,r){return async()=>{let i=await wa(this,e,t);return La(i.buffer,r)}}writeTimestamp(e){this.queryType==="inside-passes"&&this.computePassEncoder.writeTimestamp(this.querySet,e)}setQueryType(){this.queryType="none",(this.env.webgpu.profiling?.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(){ue("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(){ue("info","captureEnd"),this.flush(),this.sessionStatus="default"}replay(){ue("info","replay"),this.sessionStatus="replaying";let e=this.capturedCommandList.get(this.currentSessionId),t=this.capturedPendingKernels.get(this.currentSessionId),r=e.length;this.pendingKernels=[];for(let i=0;i<r;i++){let a=this.getComputePassEncoder(),s=e[i];this.writeTimestamp(this.pendingDispatchNumber*2),a.setPipeline(s.computePipeline),a.setBindGroup(0,s.bindGroup),a.dispatchWorkgroups(...s.dispatchGroup),this.writeTimestamp(this.pendingDispatchNumber*2+1),this.pendingDispatchNumber++,this.queryType!=="none"&&this.pendingKernels.push(t[i]),(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()}}}),af={};Gt(af,{init:()=>nf});var Dr,Cd,nf,K0=P(()=>{"use strict";Y(),nt(),te(),r0(),Dr=class sf{constructor(t,r,i,a){this.module=t,this.dataType=r,this.data=i,this.dims=a}getFloat32Array(){if(this.dataType!==1)throw new Error("Invalid data type");let t=B.size(this.dims);return t===0?new Float32Array:new Float32Array(this.module.HEAP8.buffer,this.data,t)}getBigInt64Array(){if(this.dataType!==7)throw new Error("Invalid data type");let t=B.size(this.dims);return t===0?new BigInt64Array:new BigInt64Array(this.module.HEAP8.buffer,this.data,t)}getInt32Array(){if(this.dataType!==6)throw new Error("Invalid data type");let t=B.size(this.dims);return t===0?new Int32Array:new Int32Array(this.module.HEAP8.buffer,this.data,t)}getUint16Array(){if(this.dataType!==10&&this.dataType!==4)throw new Error("Invalid data type");let t=B.size(this.dims);return t===0?new Uint16Array:new Uint16Array(this.module.HEAP8.buffer,this.data,t)}reshape(t){if(B.size(t)!==B.size(this.dims))throw new Error("Invalid new shape");return new sf(this.module,this.dataType,this.data,t)}},Cd=class{constructor(e,t,r){this.module=e,this.backend=t,this.customDataOffset=0,this.customDataSize=0,this.adapterInfo=t.adapterInfo;let i=e.PTR_SIZE,a=r/e.PTR_SIZE,s=i===4?"i32":"i64";this.opKernelContext=Number(e.getValue(i*a++,s));let n=Number(e.getValue(i*a++,s));this.outputCount=Number(e.getValue(i*a++,s)),this.customDataOffset=Number(e.getValue(i*a++,"*")),this.customDataSize=Number(e.getValue(i*a++,s));let u=[];for(let l=0;l<n;l++){let p=Number(e.getValue(i*a++,s)),h=Number(e.getValue(i*a++,"*")),f=Number(e.getValue(i*a++,s)),g=[];for(let _=0;_<f;_++)g.push(Number(e.getValue(i*a++,s)));u.push(new Dr(e,p,h,g))}this.inputs=u}get kernelCustomData(){return this.backend.currentKernelCustomData}get customDataBuffer(){return this.module.HEAPU8.subarray(this.customDataOffset,this.customDataOffset+this.customDataSize)}compute(e,t){let r=t?.inputs?.map(n=>typeof n=="number"?this.inputs[n]:n)??this.inputs,i=t?.outputs??[],a=(n,u,l)=>new Dr(this.module,u,this.output(n,l),l),s=(n,u)=>{let l=At(n,u);if(!l)throw new Error(`Unsupported data type: ${n}`);let p=l>0?this.backend.gpuDataManager.create(l).id:0;return new Dr(this.module,n,p,u)};return this.backend.run(e,r,i,a,s,this.outputCount)}output(e,t){let r=this.module.stackSave();try{let i=this.module.PTR_SIZE,a=i===4?"i32":"i64",s=this.module.stackAlloc((1+t.length)*i);this.module.setValue(s,t.length,a);for(let n=0;n<t.length;n++)this.module.setValue(s+i*(n+1),t[n],a);return this.module._JsepOutput(this.opKernelContext,e,s)}catch(i){throw new Error(`Failed to generate kernel's output[${e}] with dims [${t}]. If you are running with pre-allocated output, please make sure the output type/dims are correct. Error: ${i}`)}finally{this.module.stackRestore(r)}}},nf=async(e,t,r,i)=>{let a=t.jsepInit;if(!a)throw new Error("Failed to initialize JSEP. The WebAssembly module is not built with JSEP support.");if(e==="webgpu"){let s=(j0(),pr(tf)).WebGpuBackend,n=new s;await n.initialize(r,i),a("webgpu",[n,u=>n.alloc(Number(u)),u=>n.free(u),(u,l,p,h=!1)=>{if(h)ue("verbose",()=>`[WebGPU] jsepCopyGpuToGpu: src=${Number(u)}, dst=${Number(l)}, size=${Number(p)}`),n.memcpy(Number(u),Number(l));else{ue("verbose",()=>`[WebGPU] jsepCopyCpuToGpu: dataOffset=${Number(u)}, gpuDataId=${Number(l)}, size=${Number(p)}`);let f=t.HEAPU8.subarray(Number(u>>>0),Number(u>>>0)+Number(p));n.upload(Number(l),f)}},async(u,l,p)=>{ue("verbose",()=>`[WebGPU] jsepCopyGpuToCpu: gpuDataId=${u}, dataOffset=${l}, size=${p}`),await n.download(Number(u),()=>t.HEAPU8.subarray(Number(l)>>>0,Number(l+p)>>>0))},(u,l,p)=>n.createKernel(u,Number(l),p,t.UTF8ToString(t._JsepGetNodeName(Number(l)))),u=>n.releaseKernel(u),(u,l,p,h)=>{ue("verbose",()=>`[WebGPU] jsepRun: sessionHandle=${p}, kernel=${u}, contextDataOffset=${l}`);let f=new Cd(t,n,Number(l));return n.computeKernel(Number(u),f,h)},()=>n.captureBegin(),()=>n.captureEnd(),()=>n.replay()])}else{let s=new mp(r);a("webnn",[s,()=>s.reserveTensorId(),n=>s.releaseTensorId(n),async(n,u,l,p,h)=>s.ensureTensor(n,u,l,p,h),(n,u)=>{s.uploadTensor(n,u)},async(n,u)=>s.downloadTensor(n,u),(n,u)=>s.registerMLContext(n,u),!!r.trace])}}}),Ad,Ja,en,ht,Od,ca,Kr,tn,rn,ha,an,nn,sn,of=P(()=>{"use strict";We(),Jg(),e0(),Y(),Mt(),Da(),lp(),Ad=(e,t)=>{ye()._OrtInit(e,t)!==0&&fe("Can't initialize onnxruntime.")},Ja=async e=>{Ad(e.wasm.numThreads,Vr(e.logLevel))},en=async(e,t)=>{ye().asyncInit?.();let r=e.webgpu.adapter;if(t==="webgpu"){if(typeof navigator>"u"||!navigator.gpu)throw new Error("WebGPU is not supported in current environment");if(r){if(typeof r.limits!="object"||typeof r.features!="object"||typeof r.requestDevice!="function")throw new Error("Invalid GPU adapter set in `env.webgpu.adapter`. It must be a GPUAdapter object.")}else{let i=e.webgpu.powerPreference;if(i!==void 0&&i!=="low-power"&&i!=="high-performance")throw new Error(`Invalid powerPreference setting: "${i}"`);let a=e.webgpu.forceFallbackAdapter;if(a!==void 0&&typeof a!="boolean")throw new Error(`Invalid forceFallbackAdapter setting: "${a}"`);if(r=await navigator.gpu.requestAdapter({powerPreference:i,forceFallbackAdapter:a}),!r)throw new Error('Failed to get GPU adapter. You may need to enable flag "--enable-unsafe-webgpu" if you are using Chrome.')}}if(t==="webnn"&&(typeof navigator>"u"||!navigator.ml))throw new Error("WebNN is not supported in current environment");{let i=(K0(),pr(af)).init;t==="webgpu"&&await i("webgpu",ye(),e,r),t==="webnn"&&await i("webnn",ye(),e)}},ht=new Map,Od=e=>{let t=ye(),r=t.stackSave();try{let i=t.PTR_SIZE,a=t.stackAlloc(2*i);t._OrtGetInputOutputCount(e,a,a+i)!==0&&fe("Can't get session input/output count.");let s=i===4?"i32":"i64";return[Number(t.getValue(a,s)),Number(t.getValue(a+i,s))]}finally{t.stackRestore(r)}},ca=(e,t)=>{let r=ye(),i=r.stackSave(),a=0;try{let s=r.PTR_SIZE,n=r.stackAlloc(2*s);r._OrtGetInputOutputMetadata(e,t,n,n+s)!==0&&fe("Can't get session input/output metadata.");let u=Number(r.getValue(n,"*"));a=Number(r.getValue(n+s,"*"));let l=r.HEAP32[a/4];if(l===0)return[u,0];let p=r.HEAPU32[a/4+1],h=[];for(let f=0;f<p;f++){let g=Number(r.getValue(a+8+f*s,"*"));h.push(g!==0?r.UTF8ToString(g):Number(r.getValue(a+8+(f+p)*s,"*")))}return[u,l,h]}finally{r.stackRestore(i),a!==0&&r._OrtFree(a)}},Kr=e=>{let t=ye(),r=t._malloc(e.byteLength);if(r===0)throw new Error(`Can't create a session. failed to allocate a buffer of size ${e.byteLength}.`);return t.HEAPU8.set(e,r),[r,e.byteLength]},tn=async(e,t)=>{let r,i,a=ye();Array.isArray(e)?[r,i]=e:e.buffer===a.HEAPU8.buffer?[r,i]=[e.byteOffset,e.byteLength]:[r,i]=Kr(e);let s=0,n=0,u=0,l=[],p=[],h=[];try{if([n,l]=await up(t),t?.externalData&&a.mountExternalData){let I=[];for(let k of t.externalData){let C=typeof k=="string"?k:k.path;I.push(Wa(typeof k=="string"?k:k.data).then(O=>{a.mountExternalData(C,O)}))}await Promise.all(I)}for(let I of t?.executionProviders??[])if((typeof I=="string"?I:I.name)==="webnn"){if(a.shouldTransferToMLTensor=!1,typeof I!="string"){let k=I,C=k?.context,O=k?.gpuDevice,x=k?.deviceType,D=k?.powerPreference;C?a.currentContext=C:O?a.currentContext=await a.webnnCreateMLContext(O):a.currentContext=await a.webnnCreateMLContext({deviceType:x,powerPreference:D})}else a.currentContext=await a.webnnCreateMLContext();break}s=await a._OrtCreateSession(r,i,n),a.webgpuOnCreateSession?.(s),s===0&&fe("Can't create a session."),a.jsepOnCreateSession?.(),a.currentContext&&(a.webnnRegisterMLContext(s,a.currentContext),a.currentContext=void 0,a.shouldTransferToMLTensor=!0);let[f,g]=Od(s),_=!!t?.enableGraphCapture,y=[],$=[],S=[],v=[],w=[];for(let I=0;I<f;I++){let[k,C,O]=ca(s,I);k===0&&fe("Can't get an input name."),p.push(k);let x=a.UTF8ToString(k);y.push(x),S.push(C===0?{name:x,isTensor:!1}:{name:x,isTensor:!0,type:at(C),shape:O})}for(let I=0;I<g;I++){let[k,C,O]=ca(s,I+f);k===0&&fe("Can't get an output name."),h.push(k);let x=a.UTF8ToString(k);$.push(x),v.push(C===0?{name:x,isTensor:!1}:{name:x,isTensor:!0,type:at(C),shape:O});{if(_&&t?.preferredOutputLocation===void 0){w.push("gpu-buffer");continue}let D=typeof t?.preferredOutputLocation=="string"?t.preferredOutputLocation:t?.preferredOutputLocation?.[x]??"cpu",q=a.webnnIsGraphOutput;if(D==="cpu"&&q&&q(s,x)){w.push("ml-tensor-cpu-output");continue}if(D!=="cpu"&&D!=="cpu-pinned"&&D!=="gpu-buffer"&&D!=="ml-tensor")throw new Error(`Not supported preferred output location: ${D}.`);if(_&&D!=="gpu-buffer")throw new Error(`Not supported preferred output location: ${D}. Only 'gpu-buffer' location is supported when enableGraphCapture is true.`);w.push(D)}}let E=null;return w.some(I=>I==="gpu-buffer"||I==="ml-tensor"||I==="ml-tensor-cpu-output")&&(u=a._OrtCreateBinding(s),u===0&&fe("Can't create IO binding."),E={handle:u,outputPreferredLocations:w,outputPreferredLocationsEncoded:w.map(I=>I==="ml-tensor-cpu-output"?"ml-tensor":I).map(I=>ya(I))}),ht.set(s,[s,p,h,E,_,!1]),[s,y,$,S,v]}catch(f){throw p.forEach(g=>a._OrtFree(g)),h.forEach(g=>a._OrtFree(g)),u!==0&&a._OrtReleaseBinding(u)!==0&&fe("Can't release IO binding."),s!==0&&a._OrtReleaseSession(s)!==0&&fe("Can't release session."),f}finally{a._free(r),n!==0&&a._OrtReleaseSessionOptions(n)!==0&&fe("Can't release session options."),l.forEach(f=>a._free(f)),a.unmountExternalData?.()}},rn=e=>{let t=ye(),r=ht.get(e);if(!r)throw new Error(`cannot release session. invalid session id: ${e}`);let[i,a,s,n,u]=r;n&&(u&&t._OrtClearBoundOutputs(n.handle)!==0&&fe("Can't clear bound outputs."),t._OrtReleaseBinding(n.handle)!==0&&fe("Can't release IO binding.")),t.jsepOnReleaseSession?.(e),t.webnnOnReleaseSession?.(e),t.webgpuOnReleaseSession?.(e),a.forEach(l=>t._OrtFree(l)),s.forEach(l=>t._OrtFree(l)),t._OrtReleaseSession(i)!==0&&fe("Can't release session."),ht.delete(e)},ha=async(e,t,r,i,a,s,n=!1)=>{if(!e){t.push(0);return}let u=ye(),l=u.PTR_SIZE,p=e[0],h=e[1],f=e[3],g=f,_,y;if(p==="string"&&(f==="gpu-buffer"||f==="ml-tensor"))throw new Error("String tensor is not supported on GPU.");if(n&&f!=="gpu-buffer")throw new Error(`External buffer must be provided for input/output index ${s} when enableGraphCapture is true.`);if(f==="gpu-buffer"){let v=e[2].gpuBuffer;y=At(Ct(p),h);{let w=u.jsepRegisterBuffer;if(!w)throw new Error('Tensor location "gpu-buffer" is not supported without using WebGPU.');_=w(i,s,v,y)}}else if(f==="ml-tensor"){let v=e[2].mlTensor;y=At(Ct(p),h);let w=u.webnnRegisterMLTensor;if(!w)throw new Error('Tensor location "ml-tensor" is not supported without using WebNN.');_=w(i,v,Ct(p),h)}else{let v=e[2];if(Array.isArray(v)){y=l*v.length,_=u._malloc(y),r.push(_);for(let w=0;w<v.length;w++){if(typeof v[w]!="string")throw new TypeError(`tensor data at index ${w} is not a string`);u.setValue(_+w*l,Ke(v[w],r),"*")}}else{let w=u.webnnIsGraphInput,E=u.webnnIsGraphOutput;if(p!=="string"&&w&&E){let I=u.UTF8ToString(a);if(w(i,I)||E(i,I)){let k=Ct(p);y=At(k,h),g="ml-tensor";let C=u.webnnCreateTemporaryTensor,O=u.webnnUploadTensor;if(!C||!O)throw new Error('Tensor location "ml-tensor" is not supported without using WebNN.');let x=await C(i,k,h);O(x,new Uint8Array(v.buffer,v.byteOffset,v.byteLength)),_=x}else y=v.byteLength,_=u._malloc(y),r.push(_),u.HEAPU8.set(new Uint8Array(v.buffer,v.byteOffset,y),_)}else y=v.byteLength,_=u._malloc(y),r.push(_),u.HEAPU8.set(new Uint8Array(v.buffer,v.byteOffset,y),_)}}let $=u.stackSave(),S=u.stackAlloc(4*h.length);try{h.forEach((w,E)=>u.setValue(S+E*l,w,l===4?"i32":"i64"));let v=u._OrtCreateTensor(Ct(p),_,y,S,h.length,ya(g));v===0&&fe(`Can't create tensor for input/output. session=${i}, index=${s}.`),t.push(v)}finally{u.stackRestore($)}},an=async(e,t,r,i,a,s)=>{let n=ye(),u=n.PTR_SIZE,l=ht.get(e);if(!l)throw new Error(`cannot run inference. invalid session id: ${e}`);let p=l[0],h=l[1],f=l[2],g=l[3],_=l[4],y=l[5],$=t.length,S=i.length,v=0,w=[],E=[],I=[],k=[],C=[],O=n.stackSave(),x=n.stackAlloc($*u),D=n.stackAlloc($*u),q=n.stackAlloc(S*u),Q=n.stackAlloc(S*u);try{[v,w]=op(s),mt("wasm prepareInputOutputTensor");for(let A=0;A<$;A++)await ha(r[A],E,k,e,h[t[A]],t[A],_);for(let A=0;A<S;A++)await ha(a[A],I,k,e,f[i[A]],$+i[A],_);gt("wasm prepareInputOutputTensor");for(let A=0;A<$;A++)n.setValue(x+A*u,E[A],"*"),n.setValue(D+A*u,h[t[A]],"*");for(let A=0;A<S;A++)n.setValue(q+A*u,I[A],"*"),n.setValue(Q+A*u,f[i[A]],"*");if(g&&!y){let{handle:A,outputPreferredLocations:U,outputPreferredLocationsEncoded:ee}=g;if(h.length!==$)throw new Error(`input count from feeds (${$}) is expected to be always equal to model's input count (${h.length}).`);mt("wasm bindInputsOutputs");for(let X=0;X<$;X++){let H=t[X];await n._OrtBindInput(A,h[H],E[X])!==0&&fe(`Can't bind input[${X}] for session=${e}.`)}for(let X=0;X<S;X++){let H=i[X];a[X]?.[3]?(C.push(I[X]),n._OrtBindOutput(A,f[H],I[X],0)!==0&&fe(`Can't bind pre-allocated output[${X}] for session=${e}.`)):n._OrtBindOutput(A,f[H],0,ee[H])!==0&&fe(`Can't bind output[${X}] to ${U[X]} for session=${e}.`)}gt("wasm bindInputsOutputs"),ht.set(e,[p,h,f,g,_,!0])}n.jsepOnRunStart?.(p),n.webnnOnRunStart?.(p);let L;g?L=await n._OrtRunWithBinding(p,g.handle,S,q,v):L=await n._OrtRun(p,D,x,$,Q,S,q,v),L!==0&&fe("failed to call OrtRun().");let K=[],se=[];mt("wasm ProcessOutputTensor");for(let A=0;A<S;A++){let U=Number(n.getValue(q+A*u,"*"));if(U===I[A]||C.includes(I[A])){K.push(a[A]),U!==I[A]&&n._OrtReleaseTensor(U)!==0&&fe("Can't release tensor.");continue}let ee=n.stackSave(),X=n.stackAlloc(4*u),H=!1,le,M=0;try{n._OrtGetTensorData(U,X,X+u,X+2*u,X+3*u)!==0&&fe(`Can't access output tensor data on index ${A}.`);let V=u===4?"i32":"i64",J=Number(n.getValue(X,V));M=n.getValue(X+u,"*");let de=n.getValue(X+u*2,"*"),Te=Number(n.getValue(X+u*3,V)),Be=[];for(let xe=0;xe<Te;xe++)Be.push(Number(n.getValue(de+xe*u,V)));n._OrtFree(de)!==0&&fe("Can't free memory for tensor dims.");let qe=Be.reduce((xe,be)=>xe*be,1);le=at(J);let et=g?.outputPreferredLocations[i[A]];if(le==="string"){if(et==="gpu-buffer"||et==="ml-tensor")throw new Error("String tensor is not supported on GPU.");let xe=[];for(let be=0;be<qe;be++){let Oe=n.getValue(M+be*u,"*"),bt=n.getValue(M+(be+1)*u,"*"),fr=be===qe-1?void 0:bt-Oe;xe.push(n.UTF8ToString(Oe,fr))}K.push([le,Be,xe,"cpu"])}else if(et==="gpu-buffer"&&qe>0){let xe=n.jsepGetBuffer;if(!xe)throw new Error('preferredLocation "gpu-buffer" is not supported without using WebGPU.');let be=xe(M),Oe=At(J,qe);if(Oe===void 0||!Pa(le))throw new Error(`Unsupported data type: ${le}`);H=!0,K.push([le,Be,{gpuBuffer:be,download:n.jsepCreateDownloader(be,Oe,le),dispose:()=>{n._OrtReleaseTensor(U)!==0&&fe("Can't release tensor.")}},"gpu-buffer"])}else if(et==="ml-tensor"&&qe>0){let xe=n.webnnEnsureTensor,be=n.webnnIsGraphInputOutputTypeSupported;if(!xe||!be)throw new Error('preferredLocation "ml-tensor" is not supported without using WebNN.');if(At(J,qe)===void 0||!Ua(le))throw new Error(`Unsupported data type: ${le}`);if(!be(e,le,!1))throw new Error(`preferredLocation "ml-tensor" for ${le} output is not supported by current WebNN Context.`);let Oe=await xe(e,M,J,Be,!1);H=!0,K.push([le,Be,{mlTensor:Oe,download:n.webnnCreateMLTensorDownloader(M,le),dispose:()=>{n.webnnReleaseTensorId(M),n._OrtReleaseTensor(U)}},"ml-tensor"])}else if(et==="ml-tensor-cpu-output"&&qe>0){let xe=n.webnnCreateMLTensorDownloader(M,le)(),be=K.length;H=!0,se.push((async()=>{let Oe=[be,await xe];return n.webnnReleaseTensorId(M),n._OrtReleaseTensor(U),Oe})()),K.push([le,Be,[],"cpu"])}else{let xe=Zr(le),be=new xe(qe);new Uint8Array(be.buffer,be.byteOffset,be.byteLength).set(n.HEAPU8.subarray(M,M+be.byteLength)),K.push([le,Be,be,"cpu"])}}finally{n.stackRestore(ee),le==="string"&&M&&n._free(M),H||n._OrtReleaseTensor(U)}}g&&!_&&(n._OrtClearBoundOutputs(g.handle)!==0&&fe("Can't clear bound outputs."),ht.set(e,[p,h,f,g,_,!1]));for(let[A,U]of await Promise.all(se))K[A][2]=U;return gt("wasm ProcessOutputTensor"),K}finally{n.webnnOnRunEnd?.(p),n.stackRestore(O),E.forEach(L=>n._OrtReleaseTensor(L)),I.forEach(L=>n._OrtReleaseTensor(L)),k.forEach(L=>n._free(L)),v!==0&&n._OrtReleaseRunOptions(v),w.forEach(L=>n._free(L))}},nn=e=>{let t=ye(),r=ht.get(e);if(!r)throw new Error("invalid session id");let i=r[0],a=t._OrtEndProfiling(i);a===0&&fe("Can't get an profile file name."),t._OrtFree(a)},sn=e=>{let t=[];for(let r of e){let i=r[2];!Array.isArray(i)&&"buffer"in i&&t.push(i.buffer)}return t}}),ft,Pe,Wt,ar,nr,Pr,fa,Ur,Et,kt,Bd,uf,lf,df,pf,cf,hf,ff,mf=P(()=>{"use strict";We(),of(),Mt(),Na(),ft=()=>!!ge.wasm.proxy&&typeof document<"u",Wt=!1,ar=!1,nr=!1,Ur=new Map,Et=(e,t)=>{let r=Ur.get(e);r?r.push(t):Ur.set(e,[t])},kt=()=>{if(Wt||!ar||nr||!Pe)throw new Error("worker not ready")},Bd=e=>{switch(e.data.type){case"init-wasm":Wt=!1,e.data.err?(nr=!0,fa[1](e.data.err)):(ar=!0,fa[0]()),Pr&&(URL.revokeObjectURL(Pr),Pr=void 0);break;case"init-ep":case"copy-from":case"create":case"release":case"run":case"end-profiling":{let t=Ur.get(e.data.type);e.data.err?t.shift()[1](e.data.err):t.shift()[0](e.data.out);break}default:}},uf=async()=>{if(!ar){if(Wt)throw new Error("multiple calls to 'initWasm()' detected.");if(nr)throw new Error("previous call to 'initWasm()' failed.");if(Wt=!0,ft())return new Promise((e,t)=>{Pe?.terminate(),np().then(([r,i])=>{try{Pe=i,Pe.onerror=s=>t(s),Pe.onmessage=Bd,fa=[e,t];let a={type:"init-wasm",in:ge};!a.in.wasm.wasmPaths&&(r||ga)&&(a.in.wasm.wasmPaths={wasm:new URL("ort-wasm-simd-threaded.jsep.wasm",import.meta.url).href}),Pe.postMessage(a),Pr=r}catch(a){t(a)}},t)});try{await Ma(ge.wasm),await Ja(ge),ar=!0}catch(e){throw nr=!0,e}finally{Wt=!1}}},lf=async e=>{if(ft())return kt(),new Promise((t,r)=>{Et("init-ep",[t,r]);let i={type:"init-ep",in:{epName:e,env:ge}};Pe.postMessage(i)});await en(ge,e)},df=async e=>ft()?(kt(),new Promise((t,r)=>{Et("copy-from",[t,r]);let i={type:"copy-from",in:{buffer:e}};Pe.postMessage(i,[e.buffer])})):Kr(e),pf=async(e,t)=>{if(ft()){if(t?.preferredOutputLocation)throw new Error('session option "preferredOutputLocation" is not supported for proxy.');return kt(),new Promise((r,i)=>{Et("create",[r,i]);let a={type:"create",in:{model:e,options:{...t}}},s=[];e instanceof Uint8Array&&s.push(e.buffer),Pe.postMessage(a,s)})}else return tn(e,t)},cf=async e=>{if(ft())return kt(),new Promise((t,r)=>{Et("release",[t,r]);let i={type:"release",in:e};Pe.postMessage(i)});rn(e)},hf=async(e,t,r,i,a,s)=>{if(ft()){if(r.some(n=>n[3]!=="cpu"))throw new Error("input tensor on GPU is not supported for proxy.");if(a.some(n=>n))throw new Error("pre-allocated output tensor is not supported for proxy.");return kt(),new Promise((n,u)=>{Et("run",[n,u]);let l=r,p={type:"run",in:{sessionId:e,inputIndices:t,inputs:l,outputIndices:i,options:s}};Pe.postMessage(p,sn(l))})}else return an(e,t,r,i,a,s)},ff=async e=>{if(ft())return kt(),new Promise((t,r)=>{Et("end-profiling",[t,r]);let i={type:"end-profiling",in:e};Pe.postMessage(i)});nn(e)}}),ma,Rd,gf,Z0=P(()=>{"use strict";We(),mf(),Y(),Ra(),lp(),ma=(e,t)=>{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 ${t()}`)}},Rd=e=>{switch(e[3]){case"cpu":return new Ie(e[0],e[2],e[1]);case"gpu-buffer":{let t=e[0];if(!Pa(t))throw new Error(`not supported data type: ${t} for deserializing GPU tensor`);let{gpuBuffer:r,download:i,dispose:a}=e[2];return Ie.fromGpuBuffer(r,{dataType:t,dims:e[1],download:i,dispose:a})}case"ml-tensor":{let t=e[0];if(!Ua(t))throw new Error(`not supported data type: ${t} for deserializing MLTensor tensor`);let{mlTensor:r,download:i,dispose:a}=e[2];return Ie.fromMLTensor(r,{dataType:t,dims:e[1],download:i,dispose:a})}default:throw new Error(`invalid data location: ${e[3]}`)}},gf=class{async fetchModelAndCopyToWasmMemory(e){return df(await Wa(e))}async loadModel(e,t){Ze();let r;typeof e=="string"?r=await this.fetchModelAndCopyToWasmMemory(e):r=e,[this.sessionId,this.inputNames,this.outputNames,this.inputMetadata,this.outputMetadata]=await pf(r,t),Ue()}async dispose(){return cf(this.sessionId)}async run(e,t,r){Ze();let i=[],a=[];Object.entries(e).forEach(f=>{let g=f[0],_=f[1],y=this.inputNames.indexOf(g);if(y===-1)throw new Error(`invalid input '${g}'`);i.push(_),a.push(y)});let s=[],n=[];Object.entries(t).forEach(f=>{let g=f[0],_=f[1],y=this.outputNames.indexOf(g);if(y===-1)throw new Error(`invalid output '${g}'`);s.push(_),n.push(y)});let u=i.map((f,g)=>ma(f,()=>`input "${this.inputNames[a[g]]}"`)),l=s.map((f,g)=>f?ma(f,()=>`output "${this.outputNames[n[g]]}"`):null),p=await hf(this.sessionId,a,u,n,l,r),h={};for(let f=0;f<p.length;f++)h[this.outputNames[n[f]]]=s[f]??Rd(p[f]);return Ue(),h}startProfiling(){}endProfiling(){ff(this.sessionId)}}}),yf={};Gt(yf,{OnnxruntimeWebAssemblyBackend:()=>Aa,initializeFlags:()=>Ca,wasmBackend:()=>_f});var Ca,Aa,_f,Q0=P(()=>{"use strict";We(),mf(),Z0(),Ca=()=>{(typeof ge.wasm.initTimeout!="number"||ge.wasm.initTimeout<0)&&(ge.wasm.initTimeout=0);let e=ge.wasm.simd;if(typeof e!="boolean"&&e!==void 0&&e!=="fixed"&&e!=="relaxed"&&(console.warn(`Property "env.wasm.simd" is set to unknown value "${e}". Reset it to \`false\` and ignore SIMD feature checking.`),ge.wasm.simd=!1),typeof ge.wasm.proxy!="boolean"&&(ge.wasm.proxy=!1),typeof ge.wasm.trace!="boolean"&&(ge.wasm.trace=!1),typeof ge.wasm.numThreads!="number"||!Number.isInteger(ge.wasm.numThreads)||ge.wasm.numThreads<=0)if(typeof self<"u"&&!self.crossOriginIsolated)ge.wasm.numThreads=1;else{let t=typeof navigator>"u"?Mg("node:os").cpus().length:navigator.hardwareConcurrency;ge.wasm.numThreads=Math.min(4,Math.ceil((t||1)/2))}},Aa=class{async init(e){Ca(),await uf(),await lf(e)}async createInferenceSessionHandler(e,t){let r=new gf;return await r.loadModel(e,t),r}},_f=new Aa});We();We();We();var X0="1.24.1",Y0=Jd;{let e=(Q0(),pr(yf)).wasmBackend;Ot("webgpu",e,5),Ot("webnn",e,5),Ot("cpu",e,10),Ot("wasm",e,10)}Object.defineProperty(ge.versions,"web",{value:X0,enumerable:!0});var wf=class{constructor(t){this.options=t;let r=ge?on:globalThis.ort;r&&t.wasmPaths&&(r.env.wasm.wasmPaths=t.wasmPaths),this.melContextBuffer=new Float32Array(this.MEL_CONTEXT).fill(0)}melSession=null;embeddingSession=null;vadSession=null;customSessions=new Map;melBuffer=[];embeddingBuffers=new Map;predictionBuffers=new Map;vadBuffer=[];rawAudioRemainder=new Float32Array(0);melContextBuffer;noiseSeededEmbeddings=new Map;CHUNK_SIZE=1280;MEL_CONTEXT=480;SAMPLE_RATE=16e3;MEL_BINS=32;FRAMES_PER_CHUNK=8;MEL_WINDOW_SIZE=76;EMBEDDING_WINDOW_SIZE=24;MAX_MEL_FRAMES=970;INITIAL_FRAMES_SUPPRESS=5;PREDICTION_BUFFER_MAX=30;vadStateH=new Float32Array(128).fill(0);vadStateC=new Float32Array(128).fill(0);isLoaded=!1;async init(){try{this.melSession=await _t.create(this.options.melspectrogramModelPath),this.embeddingSession=await _t.create(this.options.embeddingModelPath),this.options.vadModelPath&&this.options.vadThreshold&&this.options.vadThreshold>0&&(this.vadSession=await _t.create(this.options.vadModelPath)),this.melBuffer=Array(this.MEL_WINDOW_SIZE).fill(0).map(()=>new Float32Array(this.MEL_BINS).fill(1));let t=new Float32Array(this.SAMPLE_RATE*4);for(let a=0;a<t.length;a++)t[a]=Math.random()*2e3-1e3;let r=new Float32Array(this.MEL_CONTEXT).fill(0),i=[];for(let a=0;a<=t.length-this.CHUNK_SIZE;a+=this.CHUNK_SIZE){let s=t.subarray(a,a+this.CHUNK_SIZE),n=new Float32Array(this.CHUNK_SIZE+this.MEL_CONTEXT);n.set(r),n.set(s,this.MEL_CONTEXT),r.set(s.subarray(this.CHUNK_SIZE-this.MEL_CONTEXT));let u=await this.runMelSpectrogram(n);for(let p=0;p<this.FRAMES_PER_CHUNK;p++){let h=new Float32Array(this.MEL_BINS);for(let f=0;f<this.MEL_BINS;f++){let g=p*this.MEL_BINS+f;h[f]=u[g]/10+2}this.melBuffer.push(h)}for(;this.melBuffer.length>this.MAX_MEL_FRAMES;)this.melBuffer.shift();let l=await this.runEmbeddingModel();i.push(l)}for(let a of this.options.wakewordModels){let s=await _t.create(a),n=this.extractModelName(a);this.customSessions.set(n,s);let u=i.slice(-this.EMBEDDING_WINDOW_SIZE).map(l=>new Float32Array(l));this.noiseSeededEmbeddings.set(n,u),this.embeddingBuffers.set(n,u.map(l=>new Float32Array(l))),this.predictionBuffers.set(n,[])}this.isLoaded=!0,console.log("OpenWakeWord models loaded and bit-perfectly aligned")}catch(t){throw console.error("Failed to initialize OpenWakeWord models:",t),t}}async predict(t){if(!this.isLoaded)throw new Error("Model not initialized");let r;if(t instanceof Int16Array){r=new Float32Array(t.length);for(let n=0;n<t.length;n++)r[n]=t[n]}else{let n=0;for(let u=0;u<Math.min(t.length,1e3);u++){let l=Math.abs(t[u]);l>n&&(n=l)}if(n<=1){r=new Float32Array(t.length);for(let u=0;u<t.length;u++)r[u]=t[u]*32768}else r=t}let i=new Float32Array(this.rawAudioRemainder.length+r.length);i.set(this.rawAudioRemainder),i.set(r,this.rawAudioRemainder.length);let a={};for(let n of this.customSessions.keys())a[n]=0;let s=0;for(;s+this.CHUNK_SIZE<=i.length;){let n=i.subarray(s,s+this.CHUNK_SIZE);s+=this.CHUNK_SIZE;let u=new Float32Array(this.CHUNK_SIZE+this.MEL_CONTEXT);if(u.set(this.melContextBuffer),u.set(n,this.MEL_CONTEXT),this.melContextBuffer.set(n.subarray(this.CHUNK_SIZE-this.MEL_CONTEXT)),this.vadSession&&this.options.vadThreshold){let h=await this.runVAD(n);for(this.vadBuffer.push(h);this.vadBuffer.length>30;)this.vadBuffer.shift()}let l=await this.runMelSpectrogram(u);for(let h=0;h<this.FRAMES_PER_CHUNK;h++){let f=new Float32Array(this.MEL_BINS);for(let g=0;g<this.MEL_BINS;g++){let _=h*this.MEL_BINS+g;f[g]=l[_]/10+2}this.melBuffer.push(f)}for(;this.melBuffer.length>this.MAX_MEL_FRAMES;)this.melBuffer.shift();let p=await this.runEmbeddingModel();for(let[h,f]of this.customSessions.entries()){let g=this.embeddingBuffers.get(h);g.shift(),g.push(p);let _=await this.runClassifier(h,f);if(this.vadSession&&this.options.vadThreshold){let $=this.vadBuffer.slice(-7,-4);($.length>0?Math.max(...$):0)<this.options.vadThreshold&&(_=0)}let y=this.predictionBuffers.get(h);for(y.push(_);y.length>this.PREDICTION_BUFFER_MAX;)y.shift();if(y.length<this.INITIAL_FRAMES_SUPPRESS)_=0;else if(this.options.patience?.[h]||this.options.debounceTime&&this.options.debounceTime>0){let $=this.options.thresholds?.[h]??.5;if(this.options.patience?.[h]){let S=this.options.patience[h];y.slice(-S).filter(E=>E>=$).length<S&&(_=0)}else if(this.options.debounceTime){let S=Math.ceil(this.options.debounceTime/.08),w=y.slice(-S-1,-1).some(E=>E>=$);_>=$&&w&&(_=0)}}a[h]=Math.max(a[h],_)}}return this.rawAudioRemainder=i.slice(s),a}async runMelSpectrogram(t){let r=new Ie("float32",t,[1,t.length]);return(await this.melSession.run({[this.melSession.inputNames[0]]:r}))[this.melSession.outputNames[0]].data}async runEmbeddingModel(){let t=new Float32Array(this.MEL_WINDOW_SIZE*this.MEL_BINS),r=this.melBuffer.length-this.MEL_WINDOW_SIZE;for(let u=0;u<this.MEL_WINDOW_SIZE;u++)t.set(this.melBuffer[r+u],u*this.MEL_BINS);let i=new Ie("float32",t,[1,this.MEL_WINDOW_SIZE,this.MEL_BINS,1]),s=(await this.embeddingSession.run({[this.embeddingSession.inputNames[0]]:i}))[this.embeddingSession.outputNames[0]].data,n=new Float32Array(96);for(let u=0;u<96;u++){let l=s[u]??0;(isNaN(l)||!isFinite(l))&&(l=0),n[u]=l}return n}async runClassifier(t,r){let i=this.embeddingBuffers.get(t),a=new Float32Array(this.EMBEDDING_WINDOW_SIZE*96);for(let u=0;u<this.EMBEDDING_WINDOW_SIZE;u++)a.set(i[u],u*96);let s=new Ie("float32",a,[1,this.EMBEDDING_WINDOW_SIZE,96]);return(await r.run({[r.inputNames[0]]:s}))[r.outputNames[0]].data[0]}async runVAD(t){let r=new Float32Array(t.length);for(let p=0;p<t.length;p++)r[p]=t[p]/32768;let i=new Ie("int64",BigInt64Array.from([BigInt(this.SAMPLE_RATE)]),[1]),a=new Ie("float32",this.vadStateH,[2,1,64]),s=new Ie("float32",this.vadStateC,[2,1,64]),n=new Ie("float32",r,[1,t.length]),u={[this.vadSession.inputNames[0]]:n,[this.vadSession.inputNames[1]]:i,[this.vadSession.inputNames[2]]:a,[this.vadSession.inputNames[3]]:s},l=await this.vadSession.run(u);return this.vadStateH=l[this.vadSession.outputNames[1]].data,this.vadStateC=l[this.vadSession.outputNames[2]].data,l[this.vadSession.outputNames[0]].data[0]}extractModelName(t){return(t.split("/").pop()||t).replace(".onnx","").replace(".tflite","")}reset(){this.melBuffer=Array(this.MEL_WINDOW_SIZE).fill(0).map(()=>new Float32Array(this.MEL_BINS).fill(1)),this.rawAudioRemainder=new Float32Array(0),this.melContextBuffer.fill(0),this.vadBuffer=[],this.vadStateH.fill(0),this.vadStateC.fill(0);for(let t of this.embeddingBuffers.keys()){this.predictionBuffers.set(t,[]);let r=this.noiseSeededEmbeddings.get(t);r?this.embeddingBuffers.set(t,r.map(i=>new Float32Array(i))):this.embeddingBuffers.set(t,Array(this.EMBEDDING_WINDOW_SIZE).fill(0).map(()=>new Float32Array(96).fill(0)))}}};export{wf as Model};
+${s}`,u=r.createShaderModule({code:n,label:e.name});ue("verbose",()=>`[WebGPU] ${e.name} shader code: ${n}`);let l=r.createComputePipeline({compute:{module:u,entryPoint:"main"},layout:"auto",label:e.name});return Ue(e.name),{programInfo:e,computePipeline:l,uniformVariablesInfo:a.variablesInfo}}normalizeDispatchGroupSize(e){let t=typeof e=="number"?e:e.x,r=typeof e=="number"?1:e.y||1,i=typeof e=="number"?1:e.z||1,a=this.backend.device.limits.maxComputeWorkgroupsPerDimension;if(t<=a&&r<=a&&i<=a)return[t,r,i];let s=t*r*i,n=Math.ceil(Math.sqrt(s));if(n>a){if(n=Math.ceil(Math.cbrt(s)),n>a)throw new Error("Total dispatch size exceeds WebGPU maximum.");return[n,n,n]}else return[n,n,1]}}}),tf={};Gt(tf,{WebGpuBackend:()=>rf});var kd,Ed,zd,rf,j0=P(()=>{"use strict";We(),Y(),nt(),hp(),i0(),F0(),H0(),kd=(e,t)=>{if(t.length!==e.length)throw new Error(`inputDependencies length ${t.length} is not equal to inputTensors length ${e.length}.`);let r=[];for(let i=0;i<e.length;++i){let a=e[i].dataType;switch(t[i]){case"none":{r.push("");break}case"type":{r.push(`${a}`);break}case"rank":{let s=e[i].dims.length;r.push(`${a};${s}`);break}case"dims":{let s=e[i].dims.join(",");r.push(`${a};${s}`);break}default:throw new Error(`unsupported input dependency: ${t[i]}`)}}return r.join("|")},Ed=(e,t,r)=>{let i=e.name;return e.shaderCache?.hint&&(i+="["+e.shaderCache.hint+"]"),i+=":"+r+`:${kd(t,e.shaderCache?.inputDependencies??new Array(t.length).fill("dims"))}`,i},zd=class{constructor(e){e&&(this.architecture=e.architecture,this.vendor=e.vendor)}isArchitecture(e){return this.architecture===e}isVendor(e){return this.vendor===e}},rf=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,t){this.env=e;let r=[],i={requiredLimits:{maxComputeWorkgroupStorageSize:t.limits.maxComputeWorkgroupStorageSize,maxComputeWorkgroupsPerDimension:t.limits.maxComputeWorkgroupsPerDimension,maxStorageBufferBindingSize:t.limits.maxStorageBufferBindingSize,maxBufferSize:t.limits.maxBufferSize,maxComputeInvocationsPerWorkgroup:t.limits.maxComputeInvocationsPerWorkgroup,maxComputeWorkgroupSizeX:t.limits.maxComputeWorkgroupSizeX,maxComputeWorkgroupSizeY:t.limits.maxComputeWorkgroupSizeY,maxComputeWorkgroupSizeZ:t.limits.maxComputeWorkgroupSizeZ},requiredFeatures:r},a=s=>t.features.has(s)&&r.push(s)&&!0;a("chromium-experimental-timestamp-query-inside-passes")||a("timestamp-query"),a("shader-f16"),a("subgroups"),this.device=await t.requestDevice(i),this.adapterInfo=new zd(t.info||await t.requestAdapterInfo()),this.gpuDataManager=gp(this),this.programManager=new ef(this),this.kernels=new Map,this.kernelPersistentData=new Map,this.kernelCustomData=new Map,La(e.logLevel,!!e.debug),this.device.onuncapturederror=s=>{s.error instanceof GPUValidationError&&console.error(`An uncaught WebGPU validation error was raised: ${s.error.message}`)},Object.defineProperty(this.env.webgpu,"device",{value:this.device,writable:!1,enumerable:!0,configurable:!1}),Object.defineProperty(this.env.webgpu,"adapter",{value:t,writable:!1,enumerable:!0,configurable:!1}),this.setQueryType()}dispose(){typeof this.querySet<"u"&&this.querySet.destroy(),this.gpuDataManager.dispose()}getCommandEncoder(){return this.commandEncoder||(this.commandEncoder=this.device.createCommandEncoder()),this.commandEncoder}getComputePassEncoder(){if(!this.computePassEncoder){let e=this.getCommandEncoder(),t={};this.queryType==="at-passes"&&(t.timestampWrites={querySet:this.querySet,beginningOfPassWriteIndex:this.pendingDispatchNumber*2,endOfPassWriteIndex:this.pendingDispatchNumber*2+1}),this.computePassEncoder=e.beginComputePass(t)}return this.computePassEncoder}endComputePass(){this.computePassEncoder&&(this.computePassEncoder.end(),this.computePassEncoder=null)}flush(){if(!this.commandEncoder)return;Ze(),this.endComputePass();let e;this.queryType!=="none"&&(this.commandEncoder.resolveQuerySet(this.querySet,0,this.pendingDispatchNumber*2,this.queryResolveBuffer,0),e=this.device.createBuffer({size:this.pendingDispatchNumber*2*8,usage:GPUBufferUsage.MAP_READ|GPUBufferUsage.COPY_DST}),this.pendingQueries.set(e,this.pendingKernels),this.pendingKernels=[],this.commandEncoder.copyBufferToBuffer(this.queryResolveBuffer,0,e,0,this.pendingDispatchNumber*2*8)),this.device.queue.submit([this.commandEncoder.finish()]),this.gpuDataManager.refreshPendingBuffers(),this.commandEncoder=null,this.pendingDispatchNumber=0,this.queryType!=="none"&&e.mapAsync(GPUMapMode.READ).then(()=>{let t=new BigUint64Array(e.getMappedRange()),r=this.pendingQueries.get(e);for(let i=0;i<t.length/2;i++){let a=r[i],s=a.kernelId,n=this.kernels.get(s),u=n.kernelType,l=n.kernelName,p=a.programName,h=a.inputTensorViews,f=a.outputTensorViews,g=t[i*2],_=t[i*2+1];typeof this.queryTimeBase>"u"&&(this.queryTimeBase=g);let y=Number(g-this.queryTimeBase),$=Number(_-this.queryTimeBase);if(!Number.isSafeInteger(y)||!Number.isSafeInteger($))throw new RangeError("incorrect timestamp range");if(this.env.webgpu.profiling?.ondata)this.env.webgpu.profiling.ondata({version:1,inputsMetadata:h.map(S=>({dims:S.dims,dataType:at(S.dataType)})),outputsMetadata:f.map(S=>({dims:S.dims,dataType:at(S.dataType)})),kernelId:s,kernelType:u,kernelName:l,programName:p,startTime:y,endTime:$});else{let S="";h.forEach((w,k)=>{S+=`input[${k}]: [${w.dims}] | ${at(w.dataType)}, `});let v="";f.forEach((w,k)=>{v+=`output[${k}]: [${w.dims}] | ${at(w.dataType)}, `}),console.log(`[profiling] kernel "${s}|${u}|${l}|${p}" ${S}${v}start time: ${y} ns, execution time: ${$-y} ns`)}cr("GPU",`${p}::${g}::${_}`)}e.unmap(),this.pendingQueries.delete(e)}),Ue()}run(e,t,r,i,a,s){Ze(e.name);let n=[];for(let w=0;w<t.length;++w){let k=t[w].data;if(k===0)continue;let I=this.gpuDataManager.get(k);if(!I)throw new Error(`no GPU data for input: ${k}`);n.push(I)}let{outputs:u,dispatchGroup:l,programUniforms:p}=e.getRunData(t),h=r.length===0?u.map((w,k)=>k):r;if(h.length!==u.length)throw new Error(`Output size ${h.length} must be equal to ${u.length}.`);let f=[],g=[];for(let w=0;w<u.length;++w){if(!Number.isInteger(h[w])||h[w]<-3||h[w]>=s)throw new Error(`Invalid output index: ${h[w]}`);if(h[w]===-3)continue;let k=h[w]===-1,I=h[w]===-2,E=k||I?a(u[w].dataType,u[w].dims):i(h[w],u[w].dataType,u[w].dims);if(f.push(E),E.data===0)continue;let C=this.gpuDataManager.get(E.data);if(!C)throw new Error(`no GPU data for output: ${E.data}`);if(k&&this.temporaryData.push(C),I){let O=this.kernelPersistentData.get(this.currentKernelId);O||(O=[],this.kernelPersistentData.set(this.currentKernelId,O)),O.push(C)}g.push(C)}if(n.length!==t.length||g.length!==f.length){if(g.length===0)return Ue(e.name),f;throw new Error(`Program ${e.name} has zero-sized tensor(s) in inputs or outputs. This is not supported now.`)}let _;if(p){let w=0,k=[];p.forEach(O=>{let x=typeof O.data=="number"?[O.data]:O.data;if(x.length===0)return;let D=O.type===10?2:4,L,Q;O.type===10?(Q=x.length>4?16:x.length>2?8:x.length*D,L=x.length>4?16:D*x.length):(Q=x.length<=2?x.length*D:16,L=16),w=Math.ceil(w/Q)*Q,k.push(w);let q=O.type===10?8:4;w+=x.length>4?Math.ceil(x.length/q)*L:x.length*D});let I=16;w=Math.ceil(w/I)*I;let E=new ArrayBuffer(w);p.forEach((O,x)=>{let D=k[x],L=typeof O.data=="number"?[O.data]:O.data;if(O.type===6)new Int32Array(E,D,L.length).set(L);else if(O.type===12)new Uint32Array(E,D,L.length).set(L);else if(O.type===10)new Uint16Array(E,D,L.length).set(L);else if(O.type===1)new Float32Array(E,D,L.length).set(L);else throw new Error(`Unsupported uniform type: ${at(O.type)}`)});let C=this.gpuDataManager.create(w,GPUBufferUsage.COPY_DST|GPUBufferUsage.UNIFORM);this.device.queue.writeBuffer(C.buffer,0,E,0,w),this.gpuDataManager.release(C.id),_={offset:0,size:w,buffer:C.buffer}}let y=this.programManager.normalizeDispatchGroupSize(l),$=y[1]===1&&y[2]===1,S=Ed(e,t,$),v=this.programManager.getArtifact(S);if(v||(v=this.programManager.build(e,y),this.programManager.setArtifact(S,v),ue("info",()=>`[artifact] key: ${S}, programName: ${e.name}`)),p&&v.uniformVariablesInfo){if(p.length!==v.uniformVariablesInfo.length)throw new Error(`Uniform variables count mismatch: expect ${v.uniformVariablesInfo.length}, got ${p.length} in program "${v.programInfo.name}".`);for(let w=0;w<p.length;w++){let k=p[w],I=k.type,E=typeof k.data=="number"?1:k.data.length,[C,O]=v.uniformVariablesInfo[w];if(I!==C||E!==O)throw new Error(`Uniform variable ${w} mismatch: expect type ${C} with size ${O}, got type ${I} with size ${E} in program "${v.programInfo.name}".`)}}if(ue("info",()=>`[ProgramManager] run "${e.name}" (key=${S}) with ${y[0]}x${y[1]}x${y[2]}`),this.queryType!=="none"||this.sessionStatus==="capturing"){let w={kernelId:this.currentKernelId,programName:v.programInfo.name,inputTensorViews:t,outputTensorViews:f};this.pendingKernels.push(w),this.sessionStatus==="capturing"&&this.capturedPendingKernels.get(this.currentSessionId).push(w)}return this.programManager.run(v,n,g,y,_),Ue(e.name),f}upload(e,t){this.gpuDataManager.upload(e,t)}memcpy(e,t){this.gpuDataManager.memcpy(e,t)}async download(e,t){await this.gpuDataManager.download(e,t)}alloc(e){return this.gpuDataManager.create(e).id}free(e){return this.gpuDataManager.release(e)}createKernel(e,t,r,i){let a=Jh.get(e);if(!a)throw new Error(`kernel not implemented: ${e}`);let s={kernelType:e,kernelName:i,kernelEntry:a[0],attributes:[a[1],r]};this.kernels.set(t,s)}releaseKernel(e){let t=this.kernelPersistentData.get(e);if(t){for(let r of t)this.gpuDataManager.release(r.id);this.kernelPersistentData.delete(e)}this.kernelCustomData.delete(e),this.kernels.delete(e)}computeKernel(e,t,r){let i=this.kernels.get(e);if(!i)throw new Error(`kernel not created: ${e}`);let a=i.kernelType,s=i.kernelName,n=i.kernelEntry,u=i.attributes;if(this.currentKernelId!==null)throw new Error(`kernel "[${a}] ${s}" is not allowed to be called recursively`);this.currentKernelId=e,u[0]&&(u[1]=u[0](u[1]),u[0]=void 0),ue("info",()=>`[WebGPU] Start to run kernel "[${a}] ${s}"...`);let l=this.env.debug;this.temporaryData=[];try{return l&&this.device.pushErrorScope("validation"),n(t,u[1]),0}catch(p){return r.push(Promise.resolve(`[WebGPU] Kernel "[${a}] ${s}" failed. ${p}`)),1}finally{l&&r.push(this.device.popErrorScope().then(p=>p?`GPU validation error for kernel "[${a}] ${s}": ${p.message}`:null));for(let p of this.temporaryData)this.gpuDataManager.release(p.id);this.temporaryData=[],this.currentKernelId=null}}registerBuffer(e,t,r,i){let a=this.sessionExternalDataMapping.get(e);a||(a=new Map,this.sessionExternalDataMapping.set(e,a));let s=a.get(t),n=this.gpuDataManager.registerExternalBuffer(r,i,s);return a.set(t,[n,r]),n}unregisterBuffers(e){let t=this.sessionExternalDataMapping.get(e);t&&(t.forEach(r=>this.gpuDataManager.unregisterExternalBuffer(r[0])),this.sessionExternalDataMapping.delete(e))}getBuffer(e){let t=this.gpuDataManager.get(e);if(!t)throw new Error(`no GPU data for buffer: ${e}`);return t.buffer}createDownloader(e,t,r){return async()=>{let i=await wa(this,e,t);return qa(i.buffer,r)}}writeTimestamp(e){this.queryType==="inside-passes"&&this.computePassEncoder.writeTimestamp(this.querySet,e)}setQueryType(){this.queryType="none",(this.env.webgpu.profiling?.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(){ue("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(){ue("info","captureEnd"),this.flush(),this.sessionStatus="default"}replay(){ue("info","replay"),this.sessionStatus="replaying";let e=this.capturedCommandList.get(this.currentSessionId),t=this.capturedPendingKernels.get(this.currentSessionId),r=e.length;this.pendingKernels=[];for(let i=0;i<r;i++){let a=this.getComputePassEncoder(),s=e[i];this.writeTimestamp(this.pendingDispatchNumber*2),a.setPipeline(s.computePipeline),a.setBindGroup(0,s.bindGroup),a.dispatchWorkgroups(...s.dispatchGroup),this.writeTimestamp(this.pendingDispatchNumber*2+1),this.pendingDispatchNumber++,this.queryType!=="none"&&this.pendingKernels.push(t[i]),(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()}}}),af={};Gt(af,{init:()=>nf});var Dr,Cd,nf,K0=P(()=>{"use strict";Y(),nt(),te(),r0(),Dr=class sf{constructor(t,r,i,a){this.module=t,this.dataType=r,this.data=i,this.dims=a}getFloat32Array(){if(this.dataType!==1)throw new Error("Invalid data type");let t=B.size(this.dims);return t===0?new Float32Array:new Float32Array(this.module.HEAP8.buffer,this.data,t)}getBigInt64Array(){if(this.dataType!==7)throw new Error("Invalid data type");let t=B.size(this.dims);return t===0?new BigInt64Array:new BigInt64Array(this.module.HEAP8.buffer,this.data,t)}getInt32Array(){if(this.dataType!==6)throw new Error("Invalid data type");let t=B.size(this.dims);return t===0?new Int32Array:new Int32Array(this.module.HEAP8.buffer,this.data,t)}getUint16Array(){if(this.dataType!==10&&this.dataType!==4)throw new Error("Invalid data type");let t=B.size(this.dims);return t===0?new Uint16Array:new Uint16Array(this.module.HEAP8.buffer,this.data,t)}reshape(t){if(B.size(t)!==B.size(this.dims))throw new Error("Invalid new shape");return new sf(this.module,this.dataType,this.data,t)}},Cd=class{constructor(e,t,r){this.module=e,this.backend=t,this.customDataOffset=0,this.customDataSize=0,this.adapterInfo=t.adapterInfo;let i=e.PTR_SIZE,a=r/e.PTR_SIZE,s=i===4?"i32":"i64";this.opKernelContext=Number(e.getValue(i*a++,s));let n=Number(e.getValue(i*a++,s));this.outputCount=Number(e.getValue(i*a++,s)),this.customDataOffset=Number(e.getValue(i*a++,"*")),this.customDataSize=Number(e.getValue(i*a++,s));let u=[];for(let l=0;l<n;l++){let p=Number(e.getValue(i*a++,s)),h=Number(e.getValue(i*a++,"*")),f=Number(e.getValue(i*a++,s)),g=[];for(let _=0;_<f;_++)g.push(Number(e.getValue(i*a++,s)));u.push(new Dr(e,p,h,g))}this.inputs=u}get kernelCustomData(){return this.backend.currentKernelCustomData}get customDataBuffer(){return this.module.HEAPU8.subarray(this.customDataOffset,this.customDataOffset+this.customDataSize)}compute(e,t){let r=t?.inputs?.map(n=>typeof n=="number"?this.inputs[n]:n)??this.inputs,i=t?.outputs??[],a=(n,u,l)=>new Dr(this.module,u,this.output(n,l),l),s=(n,u)=>{let l=At(n,u);if(!l)throw new Error(`Unsupported data type: ${n}`);let p=l>0?this.backend.gpuDataManager.create(l).id:0;return new Dr(this.module,n,p,u)};return this.backend.run(e,r,i,a,s,this.outputCount)}output(e,t){let r=this.module.stackSave();try{let i=this.module.PTR_SIZE,a=i===4?"i32":"i64",s=this.module.stackAlloc((1+t.length)*i);this.module.setValue(s,t.length,a);for(let n=0;n<t.length;n++)this.module.setValue(s+i*(n+1),t[n],a);return this.module._JsepOutput(this.opKernelContext,e,s)}catch(i){throw new Error(`Failed to generate kernel's output[${e}] with dims [${t}]. If you are running with pre-allocated output, please make sure the output type/dims are correct. Error: ${i}`)}finally{this.module.stackRestore(r)}}},nf=async(e,t,r,i)=>{let a=t.jsepInit;if(!a)throw new Error("Failed to initialize JSEP. The WebAssembly module is not built with JSEP support.");if(e==="webgpu"){let s=(j0(),pr(tf)).WebGpuBackend,n=new s;await n.initialize(r,i),a("webgpu",[n,u=>n.alloc(Number(u)),u=>n.free(u),(u,l,p,h=!1)=>{if(h)ue("verbose",()=>`[WebGPU] jsepCopyGpuToGpu: src=${Number(u)}, dst=${Number(l)}, size=${Number(p)}`),n.memcpy(Number(u),Number(l));else{ue("verbose",()=>`[WebGPU] jsepCopyCpuToGpu: dataOffset=${Number(u)}, gpuDataId=${Number(l)}, size=${Number(p)}`);let f=t.HEAPU8.subarray(Number(u>>>0),Number(u>>>0)+Number(p));n.upload(Number(l),f)}},async(u,l,p)=>{ue("verbose",()=>`[WebGPU] jsepCopyGpuToCpu: gpuDataId=${u}, dataOffset=${l}, size=${p}`),await n.download(Number(u),()=>t.HEAPU8.subarray(Number(l)>>>0,Number(l+p)>>>0))},(u,l,p)=>n.createKernel(u,Number(l),p,t.UTF8ToString(t._JsepGetNodeName(Number(l)))),u=>n.releaseKernel(u),(u,l,p,h)=>{ue("verbose",()=>`[WebGPU] jsepRun: sessionHandle=${p}, kernel=${u}, contextDataOffset=${l}`);let f=new Cd(t,n,Number(l));return n.computeKernel(Number(u),f,h)},()=>n.captureBegin(),()=>n.captureEnd(),()=>n.replay()])}else{let s=new mp(r);a("webnn",[s,()=>s.reserveTensorId(),n=>s.releaseTensorId(n),async(n,u,l,p,h)=>s.ensureTensor(n,u,l,p,h),(n,u)=>{s.uploadTensor(n,u)},async(n,u)=>s.downloadTensor(n,u),(n,u)=>s.registerMLContext(n,u),!!r.trace])}}}),Ad,Ja,en,ht,Od,ca,Kr,tn,rn,ha,an,nn,sn,of=P(()=>{"use strict";We(),Jg(),e0(),Y(),Nt(),Da(),lp(),Ad=(e,t)=>{ye()._OrtInit(e,t)!==0&&fe("Can't initialize onnxruntime.")},Ja=async e=>{Ad(e.wasm.numThreads,Vr(e.logLevel))},en=async(e,t)=>{ye().asyncInit?.();let r=e.webgpu.adapter;if(t==="webgpu"){if(typeof navigator>"u"||!navigator.gpu)throw new Error("WebGPU is not supported in current environment");if(r){if(typeof r.limits!="object"||typeof r.features!="object"||typeof r.requestDevice!="function")throw new Error("Invalid GPU adapter set in `env.webgpu.adapter`. It must be a GPUAdapter object.")}else{let i=e.webgpu.powerPreference;if(i!==void 0&&i!=="low-power"&&i!=="high-performance")throw new Error(`Invalid powerPreference setting: "${i}"`);let a=e.webgpu.forceFallbackAdapter;if(a!==void 0&&typeof a!="boolean")throw new Error(`Invalid forceFallbackAdapter setting: "${a}"`);if(r=await navigator.gpu.requestAdapter({powerPreference:i,forceFallbackAdapter:a}),!r)throw new Error('Failed to get GPU adapter. You may need to enable flag "--enable-unsafe-webgpu" if you are using Chrome.')}}if(t==="webnn"&&(typeof navigator>"u"||!navigator.ml))throw new Error("WebNN is not supported in current environment");{let i=(K0(),pr(af)).init;t==="webgpu"&&await i("webgpu",ye(),e,r),t==="webnn"&&await i("webnn",ye(),e)}},ht=new Map,Od=e=>{let t=ye(),r=t.stackSave();try{let i=t.PTR_SIZE,a=t.stackAlloc(2*i);t._OrtGetInputOutputCount(e,a,a+i)!==0&&fe("Can't get session input/output count.");let s=i===4?"i32":"i64";return[Number(t.getValue(a,s)),Number(t.getValue(a+i,s))]}finally{t.stackRestore(r)}},ca=(e,t)=>{let r=ye(),i=r.stackSave(),a=0;try{let s=r.PTR_SIZE,n=r.stackAlloc(2*s);r._OrtGetInputOutputMetadata(e,t,n,n+s)!==0&&fe("Can't get session input/output metadata.");let u=Number(r.getValue(n,"*"));a=Number(r.getValue(n+s,"*"));let l=r.HEAP32[a/4];if(l===0)return[u,0];let p=r.HEAPU32[a/4+1],h=[];for(let f=0;f<p;f++){let g=Number(r.getValue(a+8+f*s,"*"));h.push(g!==0?r.UTF8ToString(g):Number(r.getValue(a+8+(f+p)*s,"*")))}return[u,l,h]}finally{r.stackRestore(i),a!==0&&r._OrtFree(a)}},Kr=e=>{let t=ye(),r=t._malloc(e.byteLength);if(r===0)throw new Error(`Can't create a session. failed to allocate a buffer of size ${e.byteLength}.`);return t.HEAPU8.set(e,r),[r,e.byteLength]},tn=async(e,t)=>{let r,i,a=ye();Array.isArray(e)?[r,i]=e:e.buffer===a.HEAPU8.buffer?[r,i]=[e.byteOffset,e.byteLength]:[r,i]=Kr(e);let s=0,n=0,u=0,l=[],p=[],h=[];try{if([n,l]=await up(t),t?.externalData&&a.mountExternalData){let I=[];for(let E of t.externalData){let C=typeof E=="string"?E:E.path;I.push(Wa(typeof E=="string"?E:E.data).then(O=>{a.mountExternalData(C,O)}))}await Promise.all(I)}for(let I of t?.executionProviders??[])if((typeof I=="string"?I:I.name)==="webnn"){if(a.shouldTransferToMLTensor=!1,typeof I!="string"){let E=I,C=E?.context,O=E?.gpuDevice,x=E?.deviceType,D=E?.powerPreference;C?a.currentContext=C:O?a.currentContext=await a.webnnCreateMLContext(O):a.currentContext=await a.webnnCreateMLContext({deviceType:x,powerPreference:D})}else a.currentContext=await a.webnnCreateMLContext();break}s=await a._OrtCreateSession(r,i,n),a.webgpuOnCreateSession?.(s),s===0&&fe("Can't create a session."),a.jsepOnCreateSession?.(),a.currentContext&&(a.webnnRegisterMLContext(s,a.currentContext),a.currentContext=void 0,a.shouldTransferToMLTensor=!0);let[f,g]=Od(s),_=!!t?.enableGraphCapture,y=[],$=[],S=[],v=[],w=[];for(let I=0;I<f;I++){let[E,C,O]=ca(s,I);E===0&&fe("Can't get an input name."),p.push(E);let x=a.UTF8ToString(E);y.push(x),S.push(C===0?{name:x,isTensor:!1}:{name:x,isTensor:!0,type:at(C),shape:O})}for(let I=0;I<g;I++){let[E,C,O]=ca(s,I+f);E===0&&fe("Can't get an output name."),h.push(E);let x=a.UTF8ToString(E);$.push(x),v.push(C===0?{name:x,isTensor:!1}:{name:x,isTensor:!0,type:at(C),shape:O});{if(_&&t?.preferredOutputLocation===void 0){w.push("gpu-buffer");continue}let D=typeof t?.preferredOutputLocation=="string"?t.preferredOutputLocation:t?.preferredOutputLocation?.[x]??"cpu",L=a.webnnIsGraphOutput;if(D==="cpu"&&L&&L(s,x)){w.push("ml-tensor-cpu-output");continue}if(D!=="cpu"&&D!=="cpu-pinned"&&D!=="gpu-buffer"&&D!=="ml-tensor")throw new Error(`Not supported preferred output location: ${D}.`);if(_&&D!=="gpu-buffer")throw new Error(`Not supported preferred output location: ${D}. Only 'gpu-buffer' location is supported when enableGraphCapture is true.`);w.push(D)}}let k=null;return w.some(I=>I==="gpu-buffer"||I==="ml-tensor"||I==="ml-tensor-cpu-output")&&(u=a._OrtCreateBinding(s),u===0&&fe("Can't create IO binding."),k={handle:u,outputPreferredLocations:w,outputPreferredLocationsEncoded:w.map(I=>I==="ml-tensor-cpu-output"?"ml-tensor":I).map(I=>ya(I))}),ht.set(s,[s,p,h,k,_,!1]),[s,y,$,S,v]}catch(f){throw p.forEach(g=>a._OrtFree(g)),h.forEach(g=>a._OrtFree(g)),u!==0&&a._OrtReleaseBinding(u)!==0&&fe("Can't release IO binding."),s!==0&&a._OrtReleaseSession(s)!==0&&fe("Can't release session."),f}finally{a._free(r),n!==0&&a._OrtReleaseSessionOptions(n)!==0&&fe("Can't release session options."),l.forEach(f=>a._free(f)),a.unmountExternalData?.()}},rn=e=>{let t=ye(),r=ht.get(e);if(!r)throw new Error(`cannot release session. invalid session id: ${e}`);let[i,a,s,n,u]=r;n&&(u&&t._OrtClearBoundOutputs(n.handle)!==0&&fe("Can't clear bound outputs."),t._OrtReleaseBinding(n.handle)!==0&&fe("Can't release IO binding.")),t.jsepOnReleaseSession?.(e),t.webnnOnReleaseSession?.(e),t.webgpuOnReleaseSession?.(e),a.forEach(l=>t._OrtFree(l)),s.forEach(l=>t._OrtFree(l)),t._OrtReleaseSession(i)!==0&&fe("Can't release session."),ht.delete(e)},ha=async(e,t,r,i,a,s,n=!1)=>{if(!e){t.push(0);return}let u=ye(),l=u.PTR_SIZE,p=e[0],h=e[1],f=e[3],g=f,_,y;if(p==="string"&&(f==="gpu-buffer"||f==="ml-tensor"))throw new Error("String tensor is not supported on GPU.");if(n&&f!=="gpu-buffer")throw new Error(`External buffer must be provided for input/output index ${s} when enableGraphCapture is true.`);if(f==="gpu-buffer"){let v=e[2].gpuBuffer;y=At(Ct(p),h);{let w=u.jsepRegisterBuffer;if(!w)throw new Error('Tensor location "gpu-buffer" is not supported without using WebGPU.');_=w(i,s,v,y)}}else if(f==="ml-tensor"){let v=e[2].mlTensor;y=At(Ct(p),h);let w=u.webnnRegisterMLTensor;if(!w)throw new Error('Tensor location "ml-tensor" is not supported without using WebNN.');_=w(i,v,Ct(p),h)}else{let v=e[2];if(Array.isArray(v)){y=l*v.length,_=u._malloc(y),r.push(_);for(let w=0;w<v.length;w++){if(typeof v[w]!="string")throw new TypeError(`tensor data at index ${w} is not a string`);u.setValue(_+w*l,Ke(v[w],r),"*")}}else{let w=u.webnnIsGraphInput,k=u.webnnIsGraphOutput;if(p!=="string"&&w&&k){let I=u.UTF8ToString(a);if(w(i,I)||k(i,I)){let E=Ct(p);y=At(E,h),g="ml-tensor";let C=u.webnnCreateTemporaryTensor,O=u.webnnUploadTensor;if(!C||!O)throw new Error('Tensor location "ml-tensor" is not supported without using WebNN.');let x=await C(i,E,h);O(x,new Uint8Array(v.buffer,v.byteOffset,v.byteLength)),_=x}else y=v.byteLength,_=u._malloc(y),r.push(_),u.HEAPU8.set(new Uint8Array(v.buffer,v.byteOffset,y),_)}else y=v.byteLength,_=u._malloc(y),r.push(_),u.HEAPU8.set(new Uint8Array(v.buffer,v.byteOffset,y),_)}}let $=u.stackSave(),S=u.stackAlloc(4*h.length);try{h.forEach((w,k)=>u.setValue(S+k*l,w,l===4?"i32":"i64"));let v=u._OrtCreateTensor(Ct(p),_,y,S,h.length,ya(g));v===0&&fe(`Can't create tensor for input/output. session=${i}, index=${s}.`),t.push(v)}finally{u.stackRestore($)}},an=async(e,t,r,i,a,s)=>{let n=ye(),u=n.PTR_SIZE,l=ht.get(e);if(!l)throw new Error(`cannot run inference. invalid session id: ${e}`);let p=l[0],h=l[1],f=l[2],g=l[3],_=l[4],y=l[5],$=t.length,S=i.length,v=0,w=[],k=[],I=[],E=[],C=[],O=n.stackSave(),x=n.stackAlloc($*u),D=n.stackAlloc($*u),L=n.stackAlloc(S*u),Q=n.stackAlloc(S*u);try{[v,w]=op(s),mt("wasm prepareInputOutputTensor");for(let A=0;A<$;A++)await ha(r[A],k,E,e,h[t[A]],t[A],_);for(let A=0;A<S;A++)await ha(a[A],I,E,e,f[i[A]],$+i[A],_);gt("wasm prepareInputOutputTensor");for(let A=0;A<$;A++)n.setValue(x+A*u,k[A],"*"),n.setValue(D+A*u,h[t[A]],"*");for(let A=0;A<S;A++)n.setValue(L+A*u,I[A],"*"),n.setValue(Q+A*u,f[i[A]],"*");if(g&&!y){let{handle:A,outputPreferredLocations:U,outputPreferredLocationsEncoded:ee}=g;if(h.length!==$)throw new Error(`input count from feeds (${$}) is expected to be always equal to model's input count (${h.length}).`);mt("wasm bindInputsOutputs");for(let X=0;X<$;X++){let H=t[X];await n._OrtBindInput(A,h[H],k[X])!==0&&fe(`Can't bind input[${X}] for session=${e}.`)}for(let X=0;X<S;X++){let H=i[X];a[X]?.[3]?(C.push(I[X]),n._OrtBindOutput(A,f[H],I[X],0)!==0&&fe(`Can't bind pre-allocated output[${X}] for session=${e}.`)):n._OrtBindOutput(A,f[H],0,ee[H])!==0&&fe(`Can't bind output[${X}] to ${U[X]} for session=${e}.`)}gt("wasm bindInputsOutputs"),ht.set(e,[p,h,f,g,_,!0])}n.jsepOnRunStart?.(p),n.webnnOnRunStart?.(p);let q;g?q=await n._OrtRunWithBinding(p,g.handle,S,L,v):q=await n._OrtRun(p,D,x,$,Q,S,L,v),q!==0&&fe("failed to call OrtRun().");let K=[],se=[];mt("wasm ProcessOutputTensor");for(let A=0;A<S;A++){let U=Number(n.getValue(L+A*u,"*"));if(U===I[A]||C.includes(I[A])){K.push(a[A]),U!==I[A]&&n._OrtReleaseTensor(U)!==0&&fe("Can't release tensor.");continue}let ee=n.stackSave(),X=n.stackAlloc(4*u),H=!1,le,N=0;try{n._OrtGetTensorData(U,X,X+u,X+2*u,X+3*u)!==0&&fe(`Can't access output tensor data on index ${A}.`);let V=u===4?"i32":"i64",J=Number(n.getValue(X,V));N=n.getValue(X+u,"*");let de=n.getValue(X+u*2,"*"),Ie=Number(n.getValue(X+u*3,V)),Be=[];for(let xe=0;xe<Ie;xe++)Be.push(Number(n.getValue(de+xe*u,V)));n._OrtFree(de)!==0&&fe("Can't free memory for tensor dims.");let Le=Be.reduce((xe,be)=>xe*be,1);le=at(J);let et=g?.outputPreferredLocations[i[A]];if(le==="string"){if(et==="gpu-buffer"||et==="ml-tensor")throw new Error("String tensor is not supported on GPU.");let xe=[];for(let be=0;be<Le;be++){let Oe=n.getValue(N+be*u,"*"),bt=n.getValue(N+(be+1)*u,"*"),fr=be===Le-1?void 0:bt-Oe;xe.push(n.UTF8ToString(Oe,fr))}K.push([le,Be,xe,"cpu"])}else if(et==="gpu-buffer"&&Le>0){let xe=n.jsepGetBuffer;if(!xe)throw new Error('preferredLocation "gpu-buffer" is not supported without using WebGPU.');let be=xe(N),Oe=At(J,Le);if(Oe===void 0||!Pa(le))throw new Error(`Unsupported data type: ${le}`);H=!0,K.push([le,Be,{gpuBuffer:be,download:n.jsepCreateDownloader(be,Oe,le),dispose:()=>{n._OrtReleaseTensor(U)!==0&&fe("Can't release tensor.")}},"gpu-buffer"])}else if(et==="ml-tensor"&&Le>0){let xe=n.webnnEnsureTensor,be=n.webnnIsGraphInputOutputTypeSupported;if(!xe||!be)throw new Error('preferredLocation "ml-tensor" is not supported without using WebNN.');if(At(J,Le)===void 0||!Ua(le))throw new Error(`Unsupported data type: ${le}`);if(!be(e,le,!1))throw new Error(`preferredLocation "ml-tensor" for ${le} output is not supported by current WebNN Context.`);let Oe=await xe(e,N,J,Be,!1);H=!0,K.push([le,Be,{mlTensor:Oe,download:n.webnnCreateMLTensorDownloader(N,le),dispose:()=>{n.webnnReleaseTensorId(N),n._OrtReleaseTensor(U)}},"ml-tensor"])}else if(et==="ml-tensor-cpu-output"&&Le>0){let xe=n.webnnCreateMLTensorDownloader(N,le)(),be=K.length;H=!0,se.push((async()=>{let Oe=[be,await xe];return n.webnnReleaseTensorId(N),n._OrtReleaseTensor(U),Oe})()),K.push([le,Be,[],"cpu"])}else{let xe=Zr(le),be=new xe(Le);new Uint8Array(be.buffer,be.byteOffset,be.byteLength).set(n.HEAPU8.subarray(N,N+be.byteLength)),K.push([le,Be,be,"cpu"])}}finally{n.stackRestore(ee),le==="string"&&N&&n._free(N),H||n._OrtReleaseTensor(U)}}g&&!_&&(n._OrtClearBoundOutputs(g.handle)!==0&&fe("Can't clear bound outputs."),ht.set(e,[p,h,f,g,_,!1]));for(let[A,U]of await Promise.all(se))K[A][2]=U;return gt("wasm ProcessOutputTensor"),K}finally{n.webnnOnRunEnd?.(p),n.stackRestore(O),k.forEach(q=>n._OrtReleaseTensor(q)),I.forEach(q=>n._OrtReleaseTensor(q)),E.forEach(q=>n._free(q)),v!==0&&n._OrtReleaseRunOptions(v),w.forEach(q=>n._free(q))}},nn=e=>{let t=ye(),r=ht.get(e);if(!r)throw new Error("invalid session id");let i=r[0],a=t._OrtEndProfiling(i);a===0&&fe("Can't get an profile file name."),t._OrtFree(a)},sn=e=>{let t=[];for(let r of e){let i=r[2];!Array.isArray(i)&&"buffer"in i&&t.push(i.buffer)}return t}}),ft,Pe,Wt,ar,nr,Pr,fa,Ur,kt,Et,Bd,uf,lf,df,pf,cf,hf,ff,mf=P(()=>{"use strict";We(),of(),Nt(),Ma(),ft=()=>!!ge.wasm.proxy&&typeof document<"u",Wt=!1,ar=!1,nr=!1,Ur=new Map,kt=(e,t)=>{let r=Ur.get(e);r?r.push(t):Ur.set(e,[t])},Et=()=>{if(Wt||!ar||nr||!Pe)throw new Error("worker not ready")},Bd=e=>{switch(e.data.type){case"init-wasm":Wt=!1,e.data.err?(nr=!0,fa[1](e.data.err)):(ar=!0,fa[0]()),Pr&&(URL.revokeObjectURL(Pr),Pr=void 0);break;case"init-ep":case"copy-from":case"create":case"release":case"run":case"end-profiling":{let t=Ur.get(e.data.type);e.data.err?t.shift()[1](e.data.err):t.shift()[0](e.data.out);break}default:}},uf=async()=>{if(!ar){if(Wt)throw new Error("multiple calls to 'initWasm()' detected.");if(nr)throw new Error("previous call to 'initWasm()' failed.");if(Wt=!0,ft())return new Promise((e,t)=>{Pe?.terminate(),np().then(([r,i])=>{try{Pe=i,Pe.onerror=s=>t(s),Pe.onmessage=Bd,fa=[e,t];let a={type:"init-wasm",in:ge};!a.in.wasm.wasmPaths&&(r||ga)&&(a.in.wasm.wasmPaths={wasm:new URL("ort-wasm-simd-threaded.jsep.wasm",import.meta.url).href}),Pe.postMessage(a),Pr=r}catch(a){t(a)}},t)});try{await Na(ge.wasm),await Ja(ge),ar=!0}catch(e){throw nr=!0,e}finally{Wt=!1}}},lf=async e=>{if(ft())return Et(),new Promise((t,r)=>{kt("init-ep",[t,r]);let i={type:"init-ep",in:{epName:e,env:ge}};Pe.postMessage(i)});await en(ge,e)},df=async e=>ft()?(Et(),new Promise((t,r)=>{kt("copy-from",[t,r]);let i={type:"copy-from",in:{buffer:e}};Pe.postMessage(i,[e.buffer])})):Kr(e),pf=async(e,t)=>{if(ft()){if(t?.preferredOutputLocation)throw new Error('session option "preferredOutputLocation" is not supported for proxy.');return Et(),new Promise((r,i)=>{kt("create",[r,i]);let a={type:"create",in:{model:e,options:{...t}}},s=[];e instanceof Uint8Array&&s.push(e.buffer),Pe.postMessage(a,s)})}else return tn(e,t)},cf=async e=>{if(ft())return Et(),new Promise((t,r)=>{kt("release",[t,r]);let i={type:"release",in:e};Pe.postMessage(i)});rn(e)},hf=async(e,t,r,i,a,s)=>{if(ft()){if(r.some(n=>n[3]!=="cpu"))throw new Error("input tensor on GPU is not supported for proxy.");if(a.some(n=>n))throw new Error("pre-allocated output tensor is not supported for proxy.");return Et(),new Promise((n,u)=>{kt("run",[n,u]);let l=r,p={type:"run",in:{sessionId:e,inputIndices:t,inputs:l,outputIndices:i,options:s}};Pe.postMessage(p,sn(l))})}else return an(e,t,r,i,a,s)},ff=async e=>{if(ft())return Et(),new Promise((t,r)=>{kt("end-profiling",[t,r]);let i={type:"end-profiling",in:e};Pe.postMessage(i)});nn(e)}}),ma,Rd,gf,Z0=P(()=>{"use strict";We(),mf(),Y(),Ra(),lp(),ma=(e,t)=>{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 ${t()}`)}},Rd=e=>{switch(e[3]){case"cpu":return new Te(e[0],e[2],e[1]);case"gpu-buffer":{let t=e[0];if(!Pa(t))throw new Error(`not supported data type: ${t} for deserializing GPU tensor`);let{gpuBuffer:r,download:i,dispose:a}=e[2];return Te.fromGpuBuffer(r,{dataType:t,dims:e[1],download:i,dispose:a})}case"ml-tensor":{let t=e[0];if(!Ua(t))throw new Error(`not supported data type: ${t} for deserializing MLTensor tensor`);let{mlTensor:r,download:i,dispose:a}=e[2];return Te.fromMLTensor(r,{dataType:t,dims:e[1],download:i,dispose:a})}default:throw new Error(`invalid data location: ${e[3]}`)}},gf=class{async fetchModelAndCopyToWasmMemory(e){return df(await Wa(e))}async loadModel(e,t){Ze();let r;typeof e=="string"?r=await this.fetchModelAndCopyToWasmMemory(e):r=e,[this.sessionId,this.inputNames,this.outputNames,this.inputMetadata,this.outputMetadata]=await pf(r,t),Ue()}async dispose(){return cf(this.sessionId)}async run(e,t,r){Ze();let i=[],a=[];Object.entries(e).forEach(f=>{let g=f[0],_=f[1],y=this.inputNames.indexOf(g);if(y===-1)throw new Error(`invalid input '${g}'`);i.push(_),a.push(y)});let s=[],n=[];Object.entries(t).forEach(f=>{let g=f[0],_=f[1],y=this.outputNames.indexOf(g);if(y===-1)throw new Error(`invalid output '${g}'`);s.push(_),n.push(y)});let u=i.map((f,g)=>ma(f,()=>`input "${this.inputNames[a[g]]}"`)),l=s.map((f,g)=>f?ma(f,()=>`output "${this.outputNames[n[g]]}"`):null),p=await hf(this.sessionId,a,u,n,l,r),h={};for(let f=0;f<p.length;f++)h[this.outputNames[n[f]]]=s[f]??Rd(p[f]);return Ue(),h}startProfiling(){}endProfiling(){ff(this.sessionId)}}}),yf={};Gt(yf,{OnnxruntimeWebAssemblyBackend:()=>Aa,initializeFlags:()=>Ca,wasmBackend:()=>_f});var Ca,Aa,_f,Q0=P(()=>{"use strict";We(),mf(),Z0(),Ca=()=>{(typeof ge.wasm.initTimeout!="number"||ge.wasm.initTimeout<0)&&(ge.wasm.initTimeout=0);let e=ge.wasm.simd;if(typeof e!="boolean"&&e!==void 0&&e!=="fixed"&&e!=="relaxed"&&(console.warn(`Property "env.wasm.simd" is set to unknown value "${e}". Reset it to \`false\` and ignore SIMD feature checking.`),ge.wasm.simd=!1),typeof ge.wasm.proxy!="boolean"&&(ge.wasm.proxy=!1),typeof ge.wasm.trace!="boolean"&&(ge.wasm.trace=!1),typeof ge.wasm.numThreads!="number"||!Number.isInteger(ge.wasm.numThreads)||ge.wasm.numThreads<=0)if(typeof self<"u"&&!self.crossOriginIsolated)ge.wasm.numThreads=1;else{let t=typeof navigator>"u"?Ng("node:os").cpus().length:navigator.hardwareConcurrency;ge.wasm.numThreads=Math.min(4,Math.ceil((t||1)/2))}},Aa=class{async init(e){Ca(),await uf(),await lf(e)}async createInferenceSessionHandler(e,t){let r=new gf;return await r.loadModel(e,t),r}},_f=new Aa});We();We();We();var X0="1.24.1",Y0=Jd;{let e=(Q0(),pr(yf)).wasmBackend;Ot("webgpu",e,5),Ot("webnn",e,5),Ot("cpu",e,10),Ot("wasm",e,10)}Object.defineProperty(ge.versions,"web",{value:X0,enumerable:!0});var wf=class{constructor(t){this.options=t;let r=ge?on:globalThis.ort;r&&t.wasmPaths&&(r.env.wasm.wasmPaths=t.wasmPaths),this.melContextBuffer=new Float32Array(this.MEL_CONTEXT).fill(0)}melSession=null;embeddingSession=null;vadSession=null;customSessions=new Map;embeddingWindowSizes=new Map;melBuffer=[];embeddingBuffers=[];predictionBuffers=new Map;vadBuffer=[];rawAudioRemainder=new Float32Array(0);melContextBuffer;noiseSeededEmbeddings=[];CHUNK_SIZE=1280;MEL_CONTEXT=480;SAMPLE_RATE=16e3;MEL_BINS=32;FRAMES_PER_CHUNK=8;MEL_WINDOW_SIZE=76;EMBEDDING_WINDOW_SIZE=24;MAX_MEL_FRAMES=970;INITIAL_FRAMES_SUPPRESS=5;PREDICTION_BUFFER_MAX=30;GLOBAL_MAX_EMBEDDING_WINDOW=50;vadStateH=new Float32Array(128).fill(0);vadStateC=new Float32Array(128).fill(0);isLoaded=!1;async init(){try{this.melSession=await _t.create(this.options.melspectrogramModelPath),this.embeddingSession=await _t.create(this.options.embeddingModelPath),this.options.vadModelPath&&this.options.vadThreshold&&this.options.vadThreshold>0&&(this.vadSession=await _t.create(this.options.vadModelPath)),this.melBuffer=Array(this.MEL_WINDOW_SIZE).fill(0).map(()=>new Float32Array(this.MEL_BINS).fill(1));let t=new Float32Array(this.SAMPLE_RATE*4);for(let a=0;a<t.length;a++)t[a]=Math.random()*2e3-1e3;let r=new Float32Array(this.MEL_CONTEXT).fill(0),i=[];for(let a=0;a<=t.length-this.CHUNK_SIZE;a+=this.CHUNK_SIZE){let s=t.subarray(a,a+this.CHUNK_SIZE),n=new Float32Array(this.CHUNK_SIZE+this.MEL_CONTEXT);n.set(r),n.set(s,this.MEL_CONTEXT),r.set(s.subarray(this.CHUNK_SIZE-this.MEL_CONTEXT));let u=await this.runMelSpectrogram(n);for(let p=0;p<this.FRAMES_PER_CHUNK;p++){let h=new Float32Array(this.MEL_BINS);for(let f=0;f<this.MEL_BINS;f++){let g=p*this.MEL_BINS+f;h[f]=u[g]/10+2}this.melBuffer.push(h)}for(;this.melBuffer.length>this.MAX_MEL_FRAMES;)this.melBuffer.shift();let l=await this.runEmbeddingModel();i.push(l)}this.noiseSeededEmbeddings=i.slice(-this.GLOBAL_MAX_EMBEDDING_WINDOW).map(a=>new Float32Array(a)),this.embeddingBuffers=this.noiseSeededEmbeddings.map(a=>new Float32Array(a));for(let a of this.options.wakewordModels){let s=await _t.create(a),n=this.extractModelName(a);this.customSessions.set(n,s);let u=s.inputNames[0],l=24;try{let p=new Te("float32",new Float32Array(2304),[1,24,96]);await s.run({[u]:p})}catch(p){let h=p.toString(),f=h.match(/Got: \d+ Expected: (\d+)/);f?(l=parseInt(f[1],10),console.log(`Model [${n}] dimension auto-corrected from 24 to ${l} via runtime inspection`)):h.includes("Expected")&&console.warn(`Model [${n}] dummy run failed, but couldn't parse exact dimension. Error: ${h}`)}console.log(`Model [${n}] initialized with dynamic window size: ${l}`),this.embeddingWindowSizes.set(n,l),this.predictionBuffers.set(n,[])}this.isLoaded=!0,console.log("OpenWakeWord models loaded with dynamic dimensionality support")}catch(t){throw console.error("Failed to initialize OpenWakeWord models:",t),t}}async predict(t){if(!this.isLoaded)throw new Error("Model not initialized");let r;if(t instanceof Int16Array){r=new Float32Array(t.length);for(let n=0;n<t.length;n++)r[n]=t[n]}else{let n=0;for(let u=0;u<Math.min(t.length,1e3);u++){let l=Math.abs(t[u]);l>n&&(n=l)}if(n<=1){r=new Float32Array(t.length);for(let u=0;u<t.length;u++)r[u]=t[u]*32768}else r=t}let i=new Float32Array(this.rawAudioRemainder.length+r.length);i.set(this.rawAudioRemainder),i.set(r,this.rawAudioRemainder.length);let a={};for(let n of this.customSessions.keys())a[n]=0;let s=0;for(;s+this.CHUNK_SIZE<=i.length;){let n=i.subarray(s,s+this.CHUNK_SIZE);s+=this.CHUNK_SIZE;let u=new Float32Array(this.CHUNK_SIZE+this.MEL_CONTEXT);if(u.set(this.melContextBuffer),u.set(n,this.MEL_CONTEXT),this.melContextBuffer.set(n.subarray(this.CHUNK_SIZE-this.MEL_CONTEXT)),this.vadSession&&this.options.vadThreshold){let h=await this.runVAD(n);for(this.vadBuffer.push(h);this.vadBuffer.length>30;)this.vadBuffer.shift()}let l=await this.runMelSpectrogram(u);for(let h=0;h<this.FRAMES_PER_CHUNK;h++){let f=new Float32Array(this.MEL_BINS);for(let g=0;g<this.MEL_BINS;g++){let _=h*this.MEL_BINS+g;f[g]=l[_]/10+2}this.melBuffer.push(f)}for(;this.melBuffer.length>this.MAX_MEL_FRAMES;)this.melBuffer.shift();let p=await this.runEmbeddingModel();for(this.embeddingBuffers.push(p);this.embeddingBuffers.length>this.GLOBAL_MAX_EMBEDDING_WINDOW;)this.embeddingBuffers.shift();for(let[h,f]of this.customSessions.entries()){let g=this.embeddingWindowSizes.get(h)||24,_=await this.runClassifier(h,f,g);if(this.vadSession&&this.options.vadThreshold){let $=this.vadBuffer.slice(-7,-4);($.length>0?Math.max(...$):0)<this.options.vadThreshold&&(_=0)}let y=this.predictionBuffers.get(h);for(y.push(_);y.length>this.PREDICTION_BUFFER_MAX;)y.shift();if(y.length<this.INITIAL_FRAMES_SUPPRESS)_=0;else if(this.options.patience?.[h]||this.options.debounceTime&&this.options.debounceTime>0){let $=this.options.thresholds?.[h]??.5;if(this.options.patience?.[h]){let S=this.options.patience[h];y.slice(-S).filter(k=>k>=$).length<S&&(_=0)}else if(this.options.debounceTime){let S=Math.ceil(this.options.debounceTime/.08),w=y.slice(-S-1,-1).some(k=>k>=$);_>=$&&w&&(_=0)}}a[h]=Math.max(a[h],_)}}return this.rawAudioRemainder=i.slice(s),a}async runMelSpectrogram(t){let r=new Te("float32",t,[1,t.length]);return(await this.melSession.run({[this.melSession.inputNames[0]]:r}))[this.melSession.outputNames[0]].data}async runEmbeddingModel(){let t=new Float32Array(this.MEL_WINDOW_SIZE*this.MEL_BINS),r=this.melBuffer.length-this.MEL_WINDOW_SIZE;for(let u=0;u<this.MEL_WINDOW_SIZE;u++)t.set(this.melBuffer[r+u],u*this.MEL_BINS);let i=new Te("float32",t,[1,this.MEL_WINDOW_SIZE,this.MEL_BINS,1]),s=(await this.embeddingSession.run({[this.embeddingSession.inputNames[0]]:i}))[this.embeddingSession.outputNames[0]].data,n=new Float32Array(96);for(let u=0;u<96;u++){let l=s[u]??0;(isNaN(l)||!isFinite(l))&&(l=0),n[u]=l}return n}async runClassifier(t,r,i){let a=new Float32Array(i*96),s=this.embeddingBuffers.length-i;for(let l=0;l<i;l++)a.set(this.embeddingBuffers[s+l],l*96);let n=new Te("float32",a,[1,i,96]);return(await r.run({[r.inputNames[0]]:n}))[r.outputNames[0]].data[0]}async runVAD(t){let r=new Float32Array(t.length);for(let p=0;p<t.length;p++)r[p]=t[p]/32768;let i=new Te("int64",BigInt64Array.from([BigInt(this.SAMPLE_RATE)]),[1]),a=new Te("float32",this.vadStateH,[2,1,64]),s=new Te("float32",this.vadStateC,[2,1,64]),n=new Te("float32",r,[1,t.length]),u={[this.vadSession.inputNames[0]]:n,[this.vadSession.inputNames[1]]:i,[this.vadSession.inputNames[2]]:a,[this.vadSession.inputNames[3]]:s},l=await this.vadSession.run(u);return this.vadStateH=l[this.vadSession.outputNames[1]].data,this.vadStateC=l[this.vadSession.outputNames[2]].data,l[this.vadSession.outputNames[0]].data[0]}extractModelName(t){return(t.split("/").pop()||t).replace(".onnx","").replace(".tflite","").replace(/\\/g,"/")}reset(){this.melBuffer=Array(this.MEL_WINDOW_SIZE).fill(0).map(()=>new Float32Array(this.MEL_BINS).fill(1)),this.rawAudioRemainder=new Float32Array(0),this.melContextBuffer.fill(0),this.vadBuffer=[],this.vadStateH.fill(0),this.vadStateC.fill(0),this.embeddingBuffers=this.noiseSeededEmbeddings.map(t=>new Float32Array(t));for(let t of this.customSessions.keys())this.predictionBuffers.set(t,[])}};export{wf as Model};
 /*! Bundled license information:
 
 onnxruntime-web/dist/ort.bundle.min.mjs: