var _1=Object.defineProperty;var g1=(e,r,t)=>r in e?_1(e,r,{enumerable:!0,configurable:!0,writable:!0,value:t}):e[r]=t;var le=(e,r,t)=>g1(e,typeof r!="symbol"?r+"":r,t);const Qh=new Map,yn=[],y1=(e,r,t)=>{if(r&&typeof r.init=="function"&&typeof r.createInferenceSessionHandler=="function"){const n=Qh.get(e);if(n===void 0)Qh.set(e,{backend:r,priority:t});else{if(n.priority>t)return;if(n.priority===t&&n.backend!==r)throw new Error(`cannot register backend "${e}" using priority ${t}`)}if(t>=0){const o=yn.indexOf(e);o!==-1&&yn.splice(o,1);for(let i=0;i<yn.length;i++)if(Qh.get(yn[i]).priority<=t){yn.splice(i,0,e);return}yn.push(e)}return}throw new TypeError("not a valid backend")},w1=async e=>{const r=Qh.get(e);if(!r)return"backend not found.";if(r.initialized)return r.backend;if(r.aborted)return r.error;{const t=!!r.initPromise;try{return t||(r.initPromise=r.backend.init(e)),await r.initPromise,r.initialized=!0,r.backend}catch(n){return t||(r.error=`${n}`,r.aborted=!0),r.error}finally{delete r.initPromise}}},M1=async e=>{const r=e.executionProviders||[],t=r.map(h=>typeof h=="string"?h:h.name),n=t.length===0?yn:t;let o;const i=[],u=new Set;for(const h of n){const f=await w1(h);typeof f=="string"?i.push({name:h,err:f}):(o||(o=f),o===f&&u.add(h))}if(!o)throw new Error(`no available backend found. ERR: ${i.map(h=>`[${h.name}] ${h.err}`).join(", ")}`);for(const{name:h,err:f}of i)t.includes(h)&&console.warn(`removing requested execution provider "${h}" from session options because it is not available: ${f}`);const d=r.filter(h=>u.has(typeof h=="string"?h:h.name));return[o,new Proxy(e,{get:(h,f)=>f==="executionProviders"?d:Reflect.get(h,f)})]},b1="1.23.0";let E_="warning";const Xr={wasm:{},webgl:{},webgpu:{},versions:{common:b1},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}`);E_=e}},get logLevel(){return E_}};Object.defineProperty(Xr,"logLevel",{enumerable:!0});const v1=Xr,x1=(e,r)=>{const t=typeof document<"u"?document.createElement("canvas"):new OffscreenCanvas(1,1);t.width=e.dims[3],t.height=e.dims[2];const n=t.getContext("2d");if(n!=null){let o,i;(r==null?void 0:r.tensorLayout)!==void 0&&r.tensorLayout==="NHWC"?(o=e.dims[2],i=e.dims[3]):(o=e.dims[3],i=e.dims[2]);const u=(r==null?void 0:r.format)!==void 0?r.format:"RGB",d=r==null?void 0:r.norm;let h,f;d===void 0||d.mean===void 0?h=[255,255,255,255]:typeof d.mean=="number"?h=[d.mean,d.mean,d.mean,d.mean]:(h=[d.mean[0],d.mean[1],d.mean[2],0],d.mean[3]!==void 0&&(h[3]=d.mean[3])),d===void 0||d.bias===void 0?f=[0,0,0,0]:typeof d.bias=="number"?f=[d.bias,d.bias,d.bias,d.bias]:(f=[d.bias[0],d.bias[1],d.bias[2],0],d.bias[3]!==void 0&&(f[3]=d.bias[3]));const y=i*o;let m=0,g=y,v=y*2,b=-1;u==="RGBA"?(m=0,g=y,v=y*2,b=y*3):u==="RGB"?(m=0,g=y,v=y*2):u==="RBG"&&(m=0,v=y,g=y*2);for(let k=0;k<i;k++)for(let z=0;z<o;z++){const E=(e.data[m++]-f[0])*h[0],T=(e.data[g++]-f[1])*h[1],D=(e.data[v++]-f[2])*h[2],I=b===-1?255:(e.data[b++]-f[3])*h[3];n.fillStyle="rgba("+E+","+T+","+D+","+I+")",n.fillRect(z,k,1,1)}if("toDataURL"in t)return t.toDataURL();throw new Error("toDataURL is not supported")}else throw new Error("Can not access image data")},T1=(e,r)=>{const t=typeof document<"u"?document.createElement("canvas").getContext("2d"):new OffscreenCanvas(1,1).getContext("2d");let n;if(t!=null){let o,i,u;(r==null?void 0:r.tensorLayout)!==void 0&&r.tensorLayout==="NHWC"?(o=e.dims[2],i=e.dims[1],u=e.dims[3]):(o=e.dims[3],i=e.dims[2],u=e.dims[1]);const d=r!==void 0&&r.format!==void 0?r.format:"RGB",h=r==null?void 0:r.norm;let f,y;h===void 0||h.mean===void 0?f=[255,255,255,255]:typeof h.mean=="number"?f=[h.mean,h.mean,h.mean,h.mean]:(f=[h.mean[0],h.mean[1],h.mean[2],255],h.mean[3]!==void 0&&(f[3]=h.mean[3])),h===void 0||h.bias===void 0?y=[0,0,0,0]:typeof h.bias=="number"?y=[h.bias,h.bias,h.bias,h.bias]:(y=[h.bias[0],h.bias[1],h.bias[2],0],h.bias[3]!==void 0&&(y[3]=h.bias[3]));const m=i*o;if(r!==void 0&&(r.format!==void 0&&u===4&&r.format!=="RGBA"||u===3&&r.format!=="RGB"&&r.format!=="BGR"))throw new Error("Tensor format doesn't match input tensor dims");const g=4;let v=0,b=1,k=2,z=3,E=0,T=m,D=m*2,I=-1;d==="RGBA"?(E=0,T=m,D=m*2,I=m*3):d==="RGB"?(E=0,T=m,D=m*2):d==="RBG"&&(E=0,D=m,T=m*2),n=t.createImageData(o,i);for(let $=0;$<i*o;v+=g,b+=g,k+=g,z+=g,$++)n.data[v]=(e.data[E++]-y[0])*f[0],n.data[b]=(e.data[T++]-y[1])*f[1],n.data[k]=(e.data[D++]-y[2])*f[2],n.data[z]=I===-1?255:(e.data[I++]-y[3])*f[3]}else throw new Error("Can not access image data");return n},Sf=(e,r)=>{if(e===void 0)throw new Error("Image buffer must be defined");if(r.height===void 0||r.width===void 0)throw new Error("Image height and width must be defined");if(r.tensorLayout==="NHWC")throw new Error("NHWC Tensor layout is not supported yet");const{height:t,width:n}=r,o=r.norm??{mean:255,bias:0};let i,u;typeof o.mean=="number"?i=[o.mean,o.mean,o.mean,o.mean]:i=[o.mean[0],o.mean[1],o.mean[2],o.mean[3]??255],typeof o.bias=="number"?u=[o.bias,o.bias,o.bias,o.bias]:u=[o.bias[0],o.bias[1],o.bias[2],o.bias[3]??0];const d=r.format!==void 0?r.format:"RGBA",h=r.tensorFormat!==void 0&&r.tensorFormat!==void 0?r.tensorFormat:"RGB",f=t*n,y=h==="RGBA"?new Float32Array(f*4):new Float32Array(f*3);let m=4,g=0,v=1,b=2,k=3,z=0,E=f,T=f*2,D=-1;d==="RGB"&&(m=3,g=0,v=1,b=2,k=-1),h==="RGBA"?D=f*3:h==="RBG"?(z=0,T=f,E=f*2):h==="BGR"&&(T=0,E=f,z=f*2);for(let $=0;$<f;$++,g+=m,b+=m,v+=m,k+=m)y[z++]=(e[g]+u[0])/i[0],y[E++]=(e[v]+u[1])/i[1],y[T++]=(e[b]+u[2])/i[2],D!==-1&&k!==-1&&(y[D++]=(e[k]+u[3])/i[3]);return h==="RGBA"?new $i("float32",y,[1,4,t,n]):new $i("float32",y,[1,3,t,n])},E1=async(e,r)=>{const t=typeof HTMLImageElement<"u"&&e instanceof HTMLImageElement,n=typeof ImageData<"u"&&e instanceof ImageData,o=typeof ImageBitmap<"u"&&e instanceof ImageBitmap,i=typeof e=="string";let u,d=r??{};const h=()=>{if(typeof document<"u")return document.createElement("canvas");if(typeof OffscreenCanvas<"u")return new OffscreenCanvas(1,1);throw new Error("Canvas is not supported")},f=y=>typeof HTMLCanvasElement<"u"&&y instanceof HTMLCanvasElement||y instanceof OffscreenCanvas?y.getContext("2d"):null;if(t){const y=h();y.width=e.width,y.height=e.height;const m=f(y);if(m!=null){let g=e.height,v=e.width;if(r!==void 0&&r.resizedHeight!==void 0&&r.resizedWidth!==void 0&&(g=r.resizedHeight,v=r.resizedWidth),r!==void 0){if(d=r,r.tensorFormat!==void 0)throw new Error("Image input config format must be RGBA for HTMLImageElement");d.tensorFormat="RGBA",d.height=g,d.width=v}else d.tensorFormat="RGBA",d.height=g,d.width=v;m.drawImage(e,0,0),u=m.getImageData(0,0,v,g).data}else throw new Error("Can not access image data")}else if(n){let y,m;if(r!==void 0&&r.resizedWidth!==void 0&&r.resizedHeight!==void 0?(y=r.resizedHeight,m=r.resizedWidth):(y=e.height,m=e.width),r!==void 0&&(d=r),d.format="RGBA",d.height=y,d.width=m,r!==void 0){const g=h();g.width=m,g.height=y;const v=f(g);if(v!=null)v.putImageData(e,0,0),u=v.getImageData(0,0,m,y).data;else throw new Error("Can not access image data")}else u=e.data}else if(o){if(r===void 0)throw new Error("Please provide image config with format for Imagebitmap");const y=h();y.width=e.width,y.height=e.height;const m=f(y);if(m!=null){const g=e.height,v=e.width;return m.drawImage(e,0,0,v,g),u=m.getImageData(0,0,v,g).data,d.height=g,d.width=v,Sf(u,d)}else throw new Error("Can not access image data")}else{if(i)return new Promise((y,m)=>{const g=h(),v=f(g);if(!e||!v)return m();const b=new Image;b.crossOrigin="Anonymous",b.src=e,b.onload=()=>{g.width=b.width,g.height=b.height,v.drawImage(b,0,0,g.width,g.height);const k=v.getImageData(0,0,g.width,g.height);d.height=g.height,d.width=g.width,y(Sf(k.data,d))}});throw new Error("Input data provided is not supported - aborted tensor creation")}if(u!==void 0)return Sf(u,d);throw new Error("Input data provided is not supported - aborted tensor creation")},P1=(e,r)=>{const{width:t,height:n,download:o,dispose:i}=r,u=[1,n,t,4];return new $i({location:"texture",type:"float32",texture:e,dims:u,download:o,dispose:i})},C1=(e,r)=>{const{dataType:t,dims:n,download:o,dispose:i}=r;return new $i({location:"gpu-buffer",type:t??"float32",gpuBuffer:e,dims:n,download:o,dispose:i})},S1=(e,r)=>{const{dataType:t,dims:n,download:o,dispose:i}=r;return new $i({location:"ml-tensor",type:t??"float32",mlTensor:e,dims:n,download:o,dispose:i})},$1=(e,r,t)=>new $i({location:"cpu-pinned",type:e,data:r,dims:t??[r.length]}),Ra=new Map([["float32",Float32Array],["uint8",Uint8Array],["int8",Int8Array],["uint16",Uint16Array],["int16",Int16Array],["int32",Int32Array],["bool",Uint8Array],["float64",Float64Array],["uint32",Uint32Array],["int4",Uint8Array],["uint4",Uint8Array]]),Xh=new Map([[Float32Array,"float32"],[Uint8Array,"uint8"],[Int8Array,"int8"],[Uint16Array,"uint16"],[Int16Array,"int16"],[Int32Array,"int32"],[Float64Array,"float64"],[Uint32Array,"uint32"]]);let P_=!1;const k1=()=>{if(!P_){P_=!0;const e=typeof BigInt64Array<"u"&&BigInt64Array.from,r=typeof BigUint64Array<"u"&&BigUint64Array.from,t=globalThis.Float16Array,n=typeof t<"u"&&t.from;e&&(Ra.set("int64",BigInt64Array),Xh.set(BigInt64Array,"int64")),r&&(Ra.set("uint64",BigUint64Array),Xh.set(BigUint64Array,"uint64")),n?(Ra.set("float16",t),Xh.set(t,"float16")):Ra.set("float16",Uint16Array)}},I1=e=>{let r=1;for(let t=0;t<e.length;t++){const n=e[t];if(typeof n!="number"||!Number.isSafeInteger(n))throw new TypeError(`dims[${t}] must be an integer, got: ${n}`);if(n<0)throw new RangeError(`dims[${t}] must be a non-negative integer, got: ${n}`);r*=n}return r},A1=(e,r)=>{switch(e.location){case"cpu":return new $i(e.type,e.data,r);case"cpu-pinned":return new $i({location:"cpu-pinned",data:e.data,type:e.type,dims:r});case"texture":return new $i({location:"texture",texture:e.texture,type:e.type,dims:r});case"gpu-buffer":return new $i({location:"gpu-buffer",gpuBuffer:e.gpuBuffer,type:e.type,dims:r});case"ml-tensor":return new $i({location:"ml-tensor",mlTensor:e.mlTensor,type:e.type,dims:r});default:throw new Error(`tensorReshape: tensor location ${e.location} is not supported`)}};let $i=class{constructor(r,t,n){k1();let o,i;if(typeof r=="object"&&"location"in r)switch(this.dataLocation=r.location,o=r.type,i=r.dims,r.location){case"cpu-pinned":{const d=Ra.get(o);if(!d)throw new TypeError(`unsupported type "${o}" to create tensor from pinned buffer`);if(!(r.data instanceof d))throw new TypeError(`buffer should be of type ${d.name}`);this.cpuData=r.data;break}case"texture":{if(o!=="float32")throw new TypeError(`unsupported type "${o}" to create tensor from texture`);this.gpuTextureData=r.texture,this.downloader=r.download,this.disposer=r.dispose;break}case"gpu-buffer":{if(o!=="float32"&&o!=="float16"&&o!=="int32"&&o!=="int64"&&o!=="uint32"&&o!=="uint8"&&o!=="bool"&&o!=="uint4"&&o!=="int4")throw new TypeError(`unsupported type "${o}" to create tensor from gpu buffer`);this.gpuBufferData=r.gpuBuffer,this.downloader=r.download,this.disposer=r.dispose;break}case"ml-tensor":{if(o!=="float32"&&o!=="float16"&&o!=="int32"&&o!=="int64"&&o!=="uint32"&&o!=="uint64"&&o!=="int8"&&o!=="uint8"&&o!=="bool"&&o!=="uint4"&&o!=="int4")throw new TypeError(`unsupported type "${o}" to create tensor from MLTensor`);this.mlTensorData=r.mlTensor,this.downloader=r.download,this.disposer=r.dispose;break}default:throw new Error(`Tensor constructor: unsupported location '${this.dataLocation}'`)}else{let d,h;if(typeof r=="string")if(o=r,h=n,r==="string"){if(!Array.isArray(t))throw new TypeError("A string tensor's data must be a string array.");d=t}else{const f=Ra.get(r);if(f===void 0)throw new TypeError(`Unsupported tensor type: ${r}.`);if(Array.isArray(t)){if(r==="float16"&&f===Uint16Array||r==="uint4"||r==="int4")throw new TypeError(`Creating a ${r} tensor from number array is not supported. Please use ${f.name} as data.`);r==="uint64"||r==="int64"?d=f.from(t,BigInt):d=f.from(t)}else if(t instanceof f)d=t;else if(t instanceof Uint8ClampedArray)if(r==="uint8")d=Uint8Array.from(t);else throw new TypeError("A Uint8ClampedArray tensor's data must be type of uint8");else if(r==="float16"&&t instanceof Uint16Array&&f!==Uint16Array)d=new globalThis.Float16Array(t.buffer,t.byteOffset,t.length);else throw new TypeError(`A ${o} tensor's data must be type of ${f}`)}else if(h=t,Array.isArray(r)){if(r.length===0)throw new TypeError("Tensor type cannot be inferred from an empty array.");const f=typeof r[0];if(f==="string")o="string",d=r;else if(f==="boolean")o="bool",d=Uint8Array.from(r);else throw new TypeError(`Invalid element type of data array: ${f}.`)}else if(r instanceof Uint8ClampedArray)o="uint8",d=Uint8Array.from(r);else{const f=Xh.get(r.constructor);if(f===void 0)throw new TypeError(`Unsupported type for tensor data: ${r.constructor}.`);o=f,d=r}if(h===void 0)h=[d.length];else if(!Array.isArray(h))throw new TypeError("A tensor's dims must be a number array");i=h,this.cpuData=d,this.dataLocation="cpu"}const u=I1(i);if(this.cpuData&&u!==this.cpuData.length&&!((o==="uint4"||o==="int4")&&Math.ceil(u/2)===this.cpuData.length))throw new Error(`Tensor's size(${u}) does not match data length(${this.cpuData.length}).`);this.type=o,this.dims=i,this.size=u}static async fromImage(r,t){return E1(r,t)}static fromTexture(r,t){return P1(r,t)}static fromGpuBuffer(r,t){return C1(r,t)}static fromMLTensor(r,t){return S1(r,t)}static fromPinnedBuffer(r,t,n){return $1(r,t,n)}toDataURL(r){return x1(this,r)}toImageData(r){return T1(this,r)}get data(){if(this.ensureValid(),!this.cpuData)throw new Error("The data is not on CPU. Use `getData()` to download GPU data to CPU, or use `texture` or `gpuBuffer` property to access the GPU data directly.");return this.cpuData}get location(){return this.dataLocation}get texture(){if(this.ensureValid(),!this.gpuTextureData)throw new Error("The data is not stored as a WebGL texture.");return this.gpuTextureData}get gpuBuffer(){if(this.ensureValid(),!this.gpuBufferData)throw new Error("The data is not stored as a WebGPU buffer.");return this.gpuBufferData}get mlTensor(){if(this.ensureValid(),!this.mlTensorData)throw new Error("The data is not stored as a WebNN MLTensor.");return this.mlTensorData}async getData(r){switch(this.ensureValid(),this.dataLocation){case"cpu":case"cpu-pinned":return this.data;case"texture":case"gpu-buffer":case"ml-tensor":{if(!this.downloader)throw new Error("The current tensor is not created with a specified data downloader.");if(this.isDownloading)throw new Error("The current tensor is being downloaded.");try{this.isDownloading=!0;const t=await this.downloader();return this.downloader=void 0,this.dataLocation="cpu",this.cpuData=t,r&&this.disposer&&(this.disposer(),this.disposer=void 0),t}finally{this.isDownloading=!1}}default:throw new Error(`cannot get data from location: ${this.dataLocation}`)}}dispose(){if(this.isDownloading)throw new Error("The current tensor is being downloaded.");this.disposer&&(this.disposer(),this.disposer=void 0),this.cpuData=void 0,this.gpuTextureData=void 0,this.gpuBufferData=void 0,this.mlTensorData=void 0,this.downloader=void 0,this.isDownloading=void 0,this.dataLocation="none"}ensureValid(){if(this.dataLocation==="none")throw new Error("The tensor is disposed.")}reshape(r){if(this.ensureValid(),this.downloader||this.disposer)throw new Error("Cannot reshape a tensor that owns GPU resource.");return A1(this,r)}};const Ba=$i,l0=(e,r)=>{(typeof Xr.trace>"u"?!Xr.wasm.trace:!Xr.trace)||console.timeStamp(`${e}::ORT::${r}`)},c0=(e,r)=>{var o;const t=((o=new Error().stack)==null?void 0:o.split(/\r\n|\r|\n/g))||[];let n=!1;for(let i=0;i<t.length;i++){if(n&&!t[i].includes("TRACE_FUNC")){let u=`FUNC_${e}::${t[i].trim().split(" ")[1]}`;r&&(u+=`::${r}`),l0("CPU",u);return}t[i].includes("TRACE_FUNC")&&(n=!0)}},Cm=e=>{(typeof Xr.trace>"u"?!Xr.wasm.trace:!Xr.trace)||c0("BEGIN",e)},Sm=e=>{(typeof Xr.trace>"u"?!Xr.wasm.trace:!Xr.trace)||c0("END",e)},$m=e=>{(typeof Xr.trace>"u"?!Xr.wasm.trace:!Xr.trace)||console.time(`ORT::${e}`)},km=e=>{(typeof Xr.trace>"u"?!Xr.wasm.trace:!Xr.trace)||console.timeEnd(`ORT::${e}`)};let O1=class u0{constructor(r){this.handler=r}async run(r,t,n){Cm(),$m("InferenceSession.run");const o={};let i={};if(typeof r!="object"||r===null||r instanceof Ba||Array.isArray(r))throw new TypeError("'feeds' must be an object that use input names as keys and OnnxValue as corresponding values.");let u=!0;if(typeof t=="object"){if(t===null)throw new TypeError("Unexpected argument[1]: cannot be null.");if(t instanceof Ba)throw new TypeError("'fetches' cannot be a Tensor");if(Array.isArray(t)){if(t.length===0)throw new TypeError("'fetches' cannot be an empty array.");u=!1;for(const f of t){if(typeof f!="string")throw new TypeError("'fetches' must be a string array or an object.");if(this.outputNames.indexOf(f)===-1)throw new RangeError(`'fetches' contains invalid output name: ${f}.`);o[f]=null}if(typeof n=="object"&&n!==null)i=n;else if(typeof n<"u")throw new TypeError("'options' must be an object.")}else{let f=!1;const y=Object.getOwnPropertyNames(t);for(const m of this.outputNames)if(y.indexOf(m)!==-1){const g=t[m];(g===null||g instanceof Ba)&&(f=!0,u=!1,o[m]=g)}if(f){if(typeof n=="object"&&n!==null)i=n;else if(typeof n<"u")throw new TypeError("'options' must be an object.")}else i=t}}else if(typeof t<"u")throw new TypeError("Unexpected argument[1]: must be 'fetches' or 'options'.");for(const f of this.inputNames)if(typeof r[f]>"u")throw new Error(`input '${f}' is missing in 'feeds'.`);if(u)for(const f of this.outputNames)o[f]=null;const d=await this.handler.run(r,o,i),h={};for(const f in d)if(Object.hasOwnProperty.call(d,f)){const y=d[f];y instanceof Ba?h[f]=y:h[f]=new Ba(y.type,y.data,y.dims)}return km("InferenceSession.run"),Sm(),h}async release(){return this.handler.dispose()}static async create(r,t,n,o){Cm(),$m("InferenceSession.create");let i,u={};if(typeof r=="string"){if(i=r,typeof t=="object"&&t!==null)u=t;else if(typeof t<"u")throw new TypeError("'options' must be an object.")}else if(r instanceof Uint8Array){if(i=r,typeof t=="object"&&t!==null)u=t;else if(typeof t<"u")throw new TypeError("'options' must be an object.")}else if(r instanceof ArrayBuffer||typeof SharedArrayBuffer<"u"&&r instanceof SharedArrayBuffer){const y=r;let m=0,g=r.byteLength;if(typeof t=="object"&&t!==null)u=t;else if(typeof t=="number"){if(m=t,!Number.isSafeInteger(m))throw new RangeError("'byteOffset' must be an integer.");if(m<0||m>=y.byteLength)throw new RangeError(`'byteOffset' is out of range [0, ${y.byteLength}).`);if(g=r.byteLength-m,typeof n=="number"){if(g=n,!Number.isSafeInteger(g))throw new RangeError("'byteLength' must be an integer.");if(g<=0||m+g>y.byteLength)throw new RangeError(`'byteLength' is out of range (0, ${y.byteLength-m}].`);if(typeof o=="object"&&o!==null)u=o;else if(typeof o<"u")throw new TypeError("'options' must be an object.")}else if(typeof n<"u")throw new TypeError("'byteLength' must be a number.")}else if(typeof t<"u")throw new TypeError("'options' must be an object.");i=new Uint8Array(y,m,g)}else throw new TypeError("Unexpected argument[0]: must be 'path' or 'buffer'.");const[d,h]=await M1(u),f=await d.createInferenceSessionHandler(i,h);return km("InferenceSession.create"),Sm(),new u0(f)}startProfiling(){this.handler.startProfiling()}endProfiling(){this.handler.endProfiling()}get inputNames(){return this.handler.inputNames}get outputNames(){return this.handler.outputNames}get inputMetadata(){return this.handler.inputMetadata}get outputMetadata(){return this.handler.outputMetadata}};const F1=O1;var D1=Object.freeze({__proto__:null,InferenceSession:F1,TRACE:l0,TRACE_EVENT_BEGIN:$m,TRACE_EVENT_END:km,TRACE_FUNC_BEGIN:Cm,TRACE_FUNC_END:Sm,Tensor:Ba,env:v1,registerBackend:y1});/*!
 * ONNX Runtime Web v1.23.0
 * Copyright (c) Microsoft Corporation. All rights reserved.
 * Licensed under the MIT License.
 */var Hm=Object.defineProperty,L1=Object.getOwnPropertyDescriptor,j1=Object.getOwnPropertyNames,z1=Object.prototype.hasOwnProperty,B1=(e=>typeof require<"u"?require:typeof Proxy<"u"?new Proxy(e,{get:(r,t)=>(typeof require<"u"?require:r)[t]}):e)(function(e){if(typeof require<"u")return require.apply(this,arguments);throw Error('Dynamic require of "'+e+'" is not supported')}),Je=(e,r)=>()=>(e&&(r=e(e=0)),r),Ga=(e,r)=>{for(var t in r)Hm(e,t,{get:r[t],enumerable:!0})},R1=(e,r,t,n)=>{if(r&&typeof r=="object"||typeof r=="function")for(let o of j1(r))!z1.call(e,o)&&o!==t&&Hm(e,o,{get:()=>r[o],enumerable:!(n=L1(r,o))||n.enumerable});return e},gd=e=>R1(Hm({},"__esModule",{value:!0}),e),ed,Fs,vn,C_,d0,p0=Je(()=>{ed=new Map,Fs=[],vn=(e,r,t)=>{if(r&&typeof r.init=="function"&&typeof r.createInferenceSessionHandler=="function"){let n=ed.get(e);if(n===void 0)ed.set(e,{backend:r,priority:t});else{if(n.priority>t)return;if(n.priority===t&&n.backend!==r)throw new Error(`cannot register backend "${e}" using priority ${t}`)}if(t>=0){let o=Fs.indexOf(e);o!==-1&&Fs.splice(o,1);for(let i=0;i<Fs.length;i++)if(ed.get(Fs[i]).priority<=t){Fs.splice(i,0,e);return}Fs.push(e)}return}throw new TypeError("not a valid backend")},C_=async e=>{let r=ed.get(e);if(!r)return"backend not found.";if(r.initialized)return r.backend;if(r.aborted)return r.error;{let t=!!r.initPromise;try{return t||(r.initPromise=r.backend.init(e)),await r.initPromise,r.initialized=!0,r.backend}catch(n){return t||(r.error=`${n}`,r.aborted=!0),r.error}finally{delete r.initPromise}}},d0=async e=>{let r=e.executionProviders||[],t=r.map(h=>typeof h=="string"?h:h.name),n=t.length===0?Fs:t,o,i=[],u=new Set;for(let h of n){let f=await C_(h);typeof f=="string"?i.push({name:h,err:f}):(o||(o=f),o===f&&u.add(h))}if(!o)throw new Error(`no available backend found. ERR: ${i.map(h=>`[${h.name}] ${h.err}`).join(", ")}`);for(let{name:h,err:f}of i)t.includes(h)&&console.warn(`removing requested execution provider "${h}" from session options because it is not available: ${f}`);let d=r.filter(h=>u.has(typeof h=="string"?h:h.name));return[o,new Proxy(e,{get:(h,f)=>f==="executionProviders"?d:Reflect.get(h,f)})]}}),N1=Je(()=>{p0()}),h0,V1=Je(()=>{h0="1.23.0"}),$f,Ur,f0=Je(()=>{V1(),$f="warning",Ur={wasm:{},webgl:{},webgpu:{},versions:{common:h0},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: ${e}`);$f=e}},get logLevel(){return $f}},Object.defineProperty(Ur,"logLevel",{enumerable:!0})}),dr,U1=Je(()=>{f0(),dr=Ur}),m0,_0,W1=Je(()=>{m0=(e,r)=>{let t=typeof document<"u"?document.createElement("canvas"):new OffscreenCanvas(1,1);t.width=e.dims[3],t.height=e.dims[2];let n=t.getContext("2d");if(n!=null){let o,i;(r==null?void 0:r.tensorLayout)!==void 0&&r.tensorLayout==="NHWC"?(o=e.dims[2],i=e.dims[3]):(o=e.dims[3],i=e.dims[2]);let u=(r==null?void 0:r.format)!==void 0?r.format:"RGB",d=r==null?void 0:r.norm,h,f;d===void 0||d.mean===void 0?h=[255,255,255,255]:typeof d.mean=="number"?h=[d.mean,d.mean,d.mean,d.mean]:(h=[d.mean[0],d.mean[1],d.mean[2],0],d.mean[3]!==void 0&&(h[3]=d.mean[3])),d===void 0||d.bias===void 0?f=[0,0,0,0]:typeof d.bias=="number"?f=[d.bias,d.bias,d.bias,d.bias]:(f=[d.bias[0],d.bias[1],d.bias[2],0],d.bias[3]!==void 0&&(f[3]=d.bias[3]));let y=i*o,m=0,g=y,v=y*2,b=-1;u==="RGBA"?(m=0,g=y,v=y*2,b=y*3):u==="RGB"?(m=0,g=y,v=y*2):u==="RBG"&&(m=0,v=y,g=y*2);for(let k=0;k<i;k++)for(let z=0;z<o;z++){let E=(e.data[m++]-f[0])*h[0],T=(e.data[g++]-f[1])*h[1],D=(e.data[v++]-f[2])*h[2],I=b===-1?255:(e.data[b++]-f[3])*h[3];n.fillStyle="rgba("+E+","+T+","+D+","+I+")",n.fillRect(z,k,1,1)}if("toDataURL"in t)return t.toDataURL();throw new Error("toDataURL is not supported")}else throw new Error("Can not access image data")},_0=(e,r)=>{let t=typeof document<"u"?document.createElement("canvas").getContext("2d"):new OffscreenCanvas(1,1).getContext("2d"),n;if(t!=null){let o,i,u;(r==null?void 0:r.tensorLayout)!==void 0&&r.tensorLayout==="NHWC"?(o=e.dims[2],i=e.dims[1],u=e.dims[3]):(o=e.dims[3],i=e.dims[2],u=e.dims[1]);let d=r!==void 0&&r.format!==void 0?r.format:"RGB",h=r==null?void 0:r.norm,f,y;h===void 0||h.mean===void 0?f=[255,255,255,255]:typeof h.mean=="number"?f=[h.mean,h.mean,h.mean,h.mean]:(f=[h.mean[0],h.mean[1],h.mean[2],255],h.mean[3]!==void 0&&(f[3]=h.mean[3])),h===void 0||h.bias===void 0?y=[0,0,0,0]:typeof h.bias=="number"?y=[h.bias,h.bias,h.bias,h.bias]:(y=[h.bias[0],h.bias[1],h.bias[2],0],h.bias[3]!==void 0&&(y[3]=h.bias[3]));let m=i*o;if(r!==void 0&&(r.format!==void 0&&u===4&&r.format!=="RGBA"||u===3&&r.format!=="RGB"&&r.format!=="BGR"))throw new Error("Tensor format doesn't match input tensor dims");let g=4,v=0,b=1,k=2,z=3,E=0,T=m,D=m*2,I=-1;d==="RGBA"?(E=0,T=m,D=m*2,I=m*3):d==="RGB"?(E=0,T=m,D=m*2):d==="RBG"&&(E=0,D=m,T=m*2),n=t.createImageData(o,i);for(let $=0;$<i*o;v+=g,b+=g,k+=g,z+=g,$++)n.data[v]=(e.data[E++]-y[0])*f[0],n.data[b]=(e.data[T++]-y[1])*f[1],n.data[k]=(e.data[D++]-y[2])*f[2],n.data[z]=I===-1?255:(e.data[I++]-y[3])*f[3]}else throw new Error("Can not access image data");return n}}),Lh,g0,y0,w0,M0,b0,G1=Je(()=>{qm(),Lh=(e,r)=>{if(e===void 0)throw new Error("Image buffer must be defined");if(r.height===void 0||r.width===void 0)throw new Error("Image height and width must be defined");if(r.tensorLayout==="NHWC")throw new Error("NHWC Tensor layout is not supported yet");let{height:t,width:n}=r,o=r.norm??{mean:255,bias:0},i,u;typeof o.mean=="number"?i=[o.mean,o.mean,o.mean,o.mean]:i=[o.mean[0],o.mean[1],o.mean[2],o.mean[3]??255],typeof o.bias=="number"?u=[o.bias,o.bias,o.bias,o.bias]:u=[o.bias[0],o.bias[1],o.bias[2],o.bias[3]??0];let d=r.format!==void 0?r.format:"RGBA",h=r.tensorFormat!==void 0&&r.tensorFormat!==void 0?r.tensorFormat:"RGB",f=t*n,y=h==="RGBA"?new Float32Array(f*4):new Float32Array(f*3),m=4,g=0,v=1,b=2,k=3,z=0,E=f,T=f*2,D=-1;d==="RGB"&&(m=3,g=0,v=1,b=2,k=-1),h==="RGBA"?D=f*3:h==="RBG"?(z=0,T=f,E=f*2):h==="BGR"&&(T=0,E=f,z=f*2);for(let I=0;I<f;I++,g+=m,b+=m,v+=m,k+=m)y[z++]=(e[g]+u[0])/i[0],y[E++]=(e[v]+u[1])/i[1],y[T++]=(e[b]+u[2])/i[2],D!==-1&&k!==-1&&(y[D++]=(e[k]+u[3])/i[3]);return h==="RGBA"?new vi("float32",y,[1,4,t,n]):new vi("float32",y,[1,3,t,n])},g0=async(e,r)=>{let t=typeof HTMLImageElement<"u"&&e instanceof HTMLImageElement,n=typeof ImageData<"u"&&e instanceof ImageData,o=typeof ImageBitmap<"u"&&e instanceof ImageBitmap,i=typeof e=="string",u,d=r??{},h=()=>{if(typeof document<"u")return document.createElement("canvas");if(typeof OffscreenCanvas<"u")return new OffscreenCanvas(1,1);throw new Error("Canvas is not supported")},f=y=>typeof HTMLCanvasElement<"u"&&y instanceof HTMLCanvasElement||y instanceof OffscreenCanvas?y.getContext("2d"):null;if(t){let y=h();y.width=e.width,y.height=e.height;let m=f(y);if(m!=null){let g=e.height,v=e.width;if(r!==void 0&&r.resizedHeight!==void 0&&r.resizedWidth!==void 0&&(g=r.resizedHeight,v=r.resizedWidth),r!==void 0){if(d=r,r.tensorFormat!==void 0)throw new Error("Image input config format must be RGBA for HTMLImageElement");d.tensorFormat="RGBA",d.height=g,d.width=v}else d.tensorFormat="RGBA",d.height=g,d.width=v;m.drawImage(e,0,0),u=m.getImageData(0,0,v,g).data}else throw new Error("Can not access image data")}else if(n){let y,m;if(r!==void 0&&r.resizedWidth!==void 0&&r.resizedHeight!==void 0?(y=r.resizedHeight,m=r.resizedWidth):(y=e.height,m=e.width),r!==void 0&&(d=r),d.format="RGBA",d.height=y,d.width=m,r!==void 0){let g=h();g.width=m,g.height=y;let v=f(g);if(v!=null)v.putImageData(e,0,0),u=v.getImageData(0,0,m,y).data;else throw new Error("Can not access image data")}else u=e.data}else if(o){if(r===void 0)throw new Error("Please provide image config with format for Imagebitmap");let y=h();y.width=e.width,y.height=e.height;let m=f(y);if(m!=null){let g=e.height,v=e.width;return m.drawImage(e,0,0,v,g),u=m.getImageData(0,0,v,g).data,d.height=g,d.width=v,Lh(u,d)}else throw new Error("Can not access image data")}else{if(i)return new Promise((y,m)=>{let g=h(),v=f(g);if(!e||!v)return m();let b=new Image;b.crossOrigin="Anonymous",b.src=e,b.onload=()=>{g.width=b.width,g.height=b.height,v.drawImage(b,0,0,g.width,g.height);let k=v.getImageData(0,0,g.width,g.height);d.height=g.height,d.width=g.width,y(Lh(k.data,d))}});throw new Error("Input data provided is not supported - aborted tensor creation")}if(u!==void 0)return Lh(u,d);throw new Error("Input data provided is not supported - aborted tensor creation")},y0=(e,r)=>{let{width:t,height:n,download:o,dispose:i}=r,u=[1,n,t,4];return new vi({location:"texture",type:"float32",texture:e,dims:u,download:o,dispose:i})},w0=(e,r)=>{let{dataType:t,dims:n,download:o,dispose:i}=r;return new vi({location:"gpu-buffer",type:t??"float32",gpuBuffer:e,dims:n,download:o,dispose:i})},M0=(e,r)=>{let{dataType:t,dims:n,download:o,dispose:i}=r;return new vi({location:"ml-tensor",type:t??"float32",mlTensor:e,dims:n,download:o,dispose:i})},b0=(e,r,t)=>new vi({location:"cpu-pinned",type:e,data:r,dims:t??[r.length]})}),wn,dd,kf,v0,K1=Je(()=>{wn=new Map([["float32",Float32Array],["uint8",Uint8Array],["int8",Int8Array],["uint16",Uint16Array],["int16",Int16Array],["int32",Int32Array],["bool",Uint8Array],["float64",Float64Array],["uint32",Uint32Array],["int4",Uint8Array],["uint4",Uint8Array]]),dd=new Map([[Float32Array,"float32"],[Uint8Array,"uint8"],[Int8Array,"int8"],[Uint16Array,"uint16"],[Int16Array,"int16"],[Int32Array,"int32"],[Float64Array,"float64"],[Uint32Array,"uint32"]]),kf=!1,v0=()=>{if(!kf){kf=!0;let e=typeof BigInt64Array<"u"&&BigInt64Array.from,r=typeof BigUint64Array<"u"&&BigUint64Array.from,t=globalThis.Float16Array,n=typeof t<"u"&&t.from;e&&(wn.set("int64",BigInt64Array),dd.set(BigInt64Array,"int64")),r&&(wn.set("uint64",BigUint64Array),dd.set(BigUint64Array,"uint64")),n?(wn.set("float16",t),dd.set(t,"float16")):wn.set("float16",Uint16Array)}}}),x0,T0,H1=Je(()=>{qm(),x0=e=>{let r=1;for(let t=0;t<e.length;t++){let n=e[t];if(typeof n!="number"||!Number.isSafeInteger(n))throw new TypeError(`dims[${t}] must be an integer, got: ${n}`);if(n<0)throw new RangeError(`dims[${t}] must be a non-negative integer, got: ${n}`);r*=n}return r},T0=(e,r)=>{switch(e.location){case"cpu":return new vi(e.type,e.data,r);case"cpu-pinned":return new vi({location:"cpu-pinned",data:e.data,type:e.type,dims:r});case"texture":return new vi({location:"texture",texture:e.texture,type:e.type,dims:r});case"gpu-buffer":return new vi({location:"gpu-buffer",gpuBuffer:e.gpuBuffer,type:e.type,dims:r});case"ml-tensor":return new vi({location:"ml-tensor",mlTensor:e.mlTensor,type:e.type,dims:r});default:throw new Error(`tensorReshape: tensor location ${e.location} is not supported`)}}}),vi,qm=Je(()=>{W1(),G1(),K1(),H1(),vi=class{constructor(e,r,t){v0();let n,o;if(typeof e=="object"&&"location"in e)switch(this.dataLocation=e.location,n=e.type,o=e.dims,e.location){case"cpu-pinned":{let u=wn.get(n);if(!u)throw new TypeError(`unsupported type "${n}" to create tensor from pinned buffer`);if(!(e.data instanceof u))throw new TypeError(`buffer should be of type ${u.name}`);this.cpuData=e.data;break}case"texture":{if(n!=="float32")throw new TypeError(`unsupported type "${n}" to create tensor from texture`);this.gpuTextureData=e.texture,this.downloader=e.download,this.disposer=e.dispose;break}case"gpu-buffer":{if(n!=="float32"&&n!=="float16"&&n!=="int32"&&n!=="int64"&&n!=="uint32"&&n!=="uint8"&&n!=="bool"&&n!=="uint4"&&n!=="int4")throw new TypeError(`unsupported type "${n}" to create tensor from gpu buffer`);this.gpuBufferData=e.gpuBuffer,this.downloader=e.download,this.disposer=e.dispose;break}case"ml-tensor":{if(n!=="float32"&&n!=="float16"&&n!=="int32"&&n!=="int64"&&n!=="uint32"&&n!=="uint64"&&n!=="int8"&&n!=="uint8"&&n!=="bool"&&n!=="uint4"&&n!=="int4")throw new TypeError(`unsupported type "${n}" 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 u,d;if(typeof e=="string")if(n=e,d=t,e==="string"){if(!Array.isArray(r))throw new TypeError("A string tensor's data must be a string array.");u=r}else{let h=wn.get(e);if(h===void 0)throw new TypeError(`Unsupported tensor type: ${e}.`);if(Array.isArray(r)){if(e==="float16"&&h===Uint16Array||e==="uint4"||e==="int4")throw new TypeError(`Creating a ${e} tensor from number array is not supported. Please use ${h.name} as data.`);e==="uint64"||e==="int64"?u=h.from(r,BigInt):u=h.from(r)}else if(r instanceof h)u=r;else if(r instanceof Uint8ClampedArray)if(e==="uint8")u=Uint8Array.from(r);else throw new TypeError("A Uint8ClampedArray tensor's data must be type of uint8");else if(e==="float16"&&r instanceof Uint16Array&&h!==Uint16Array)u=new globalThis.Float16Array(r.buffer,r.byteOffset,r.length);else throw new TypeError(`A ${n} tensor's data must be type of ${h}`)}else if(d=r,Array.isArray(e)){if(e.length===0)throw new TypeError("Tensor type cannot be inferred from an empty array.");let h=typeof e[0];if(h==="string")n="string",u=e;else if(h==="boolean")n="bool",u=Uint8Array.from(e);else throw new TypeError(`Invalid element type of data array: ${h}.`)}else if(e instanceof Uint8ClampedArray)n="uint8",u=Uint8Array.from(e);else{let h=dd.get(e.constructor);if(h===void 0)throw new TypeError(`Unsupported type for tensor data: ${e.constructor}.`);n=h,u=e}if(d===void 0)d=[u.length];else if(!Array.isArray(d))throw new TypeError("A tensor's dims must be a number array");o=d,this.cpuData=u,this.dataLocation="cpu"}let i=x0(o);if(this.cpuData&&i!==this.cpuData.length&&!((n==="uint4"||n==="int4")&&Math.ceil(i/2)===this.cpuData.length))throw new Error(`Tensor's size(${i}) does not match data length(${this.cpuData.length}).`);this.type=n,this.dims=o,this.size=i}static async fromImage(e,r){return g0(e,r)}static fromTexture(e,r){return y0(e,r)}static fromGpuBuffer(e,r){return w0(e,r)}static fromMLTensor(e,r){return M0(e,r)}static fromPinnedBuffer(e,r,t){return b0(e,r,t)}toDataURL(e){return m0(this,e)}toImageData(e){return _0(this,e)}get data(){if(this.ensureValid(),!this.cpuData)throw new Error("The data is not on CPU. Use `getData()` to download GPU data to CPU, or use `texture` or `gpuBuffer` property to access the GPU data directly.");return this.cpuData}get location(){return this.dataLocation}get texture(){if(this.ensureValid(),!this.gpuTextureData)throw new Error("The data is not stored as a WebGL texture.");return this.gpuTextureData}get gpuBuffer(){if(this.ensureValid(),!this.gpuBufferData)throw new Error("The data is not stored as a WebGPU buffer.");return this.gpuBufferData}get mlTensor(){if(this.ensureValid(),!this.mlTensorData)throw new Error("The data is not stored as a WebNN MLTensor.");return this.mlTensorData}async getData(e){switch(this.ensureValid(),this.dataLocation){case"cpu":case"cpu-pinned":return this.data;case"texture":case"gpu-buffer":case"ml-tensor":{if(!this.downloader)throw new Error("The current tensor is not created with a specified data downloader.");if(this.isDownloading)throw new Error("The current tensor is being downloaded.");try{this.isDownloading=!0;let r=await this.downloader();return this.downloader=void 0,this.dataLocation="cpu",this.cpuData=r,e&&this.disposer&&(this.disposer(),this.disposer=void 0),r}finally{this.isDownloading=!1}}default:throw new Error(`cannot get data from location: ${this.dataLocation}`)}}dispose(){if(this.isDownloading)throw new Error("The current tensor is being downloaded.");this.disposer&&(this.disposer(),this.disposer=void 0),this.cpuData=void 0,this.gpuTextureData=void 0,this.gpuBufferData=void 0,this.mlTensorData=void 0,this.downloader=void 0,this.isDownloading=void 0,this.dataLocation="none"}ensureValid(){if(this.dataLocation==="none")throw new Error("The tensor is disposed.")}reshape(e){if(this.ensureValid(),this.downloader||this.disposer)throw new Error("Cannot reshape a tensor that owns GPU resource.");return T0(this,e)}}}),Qi,E0=Je(()=>{qm(),Qi=vi}),yd,If,Xi,ki,zs,Bs,P0=Je(()=>{f0(),yd=(e,r)=>{(typeof Ur.trace>"u"?!Ur.wasm.trace:!Ur.trace)||console.timeStamp(`${e}::ORT::${r}`)},If=(e,r)=>{var o;let t=((o=new Error().stack)==null?void 0:o.split(/\r\n|\r|\n/g))||[],n=!1;for(let i=0;i<t.length;i++){if(n&&!t[i].includes("TRACE_FUNC")){let u=`FUNC_${e}::${t[i].trim().split(" ")[1]}`;r&&(u+=`::${r}`),yd("CPU",u);return}t[i].includes("TRACE_FUNC")&&(n=!0)}},Xi=e=>{(typeof Ur.trace>"u"?!Ur.wasm.trace:!Ur.trace)||If("BEGIN",e)},ki=e=>{(typeof Ur.trace>"u"?!Ur.wasm.trace:!Ur.trace)||If("END",e)},zs=e=>{(typeof Ur.trace>"u"?!Ur.wasm.trace:!Ur.trace)||console.time(`ORT::${e}`)},Bs=e=>{(typeof Ur.trace>"u"?!Ur.wasm.trace:!Ur.trace)||console.timeEnd(`ORT::${e}`)}}),C0,q1=Je(()=>{p0(),E0(),P0(),C0=class S0{constructor(r){this.handler=r}async run(r,t,n){Xi(),zs("InferenceSession.run");let o={},i={};if(typeof r!="object"||r===null||r instanceof Qi||Array.isArray(r))throw new TypeError("'feeds' must be an object that use input names as keys and OnnxValue as corresponding values.");let u=!0;if(typeof t=="object"){if(t===null)throw new TypeError("Unexpected argument[1]: cannot be null.");if(t instanceof Qi)throw new TypeError("'fetches' cannot be a Tensor");if(Array.isArray(t)){if(t.length===0)throw new TypeError("'fetches' cannot be an empty array.");u=!1;for(let f of t){if(typeof f!="string")throw new TypeError("'fetches' must be a string array or an object.");if(this.outputNames.indexOf(f)===-1)throw new RangeError(`'fetches' contains invalid output name: ${f}.`);o[f]=null}if(typeof n=="object"&&n!==null)i=n;else if(typeof n<"u")throw new TypeError("'options' must be an object.")}else{let f=!1,y=Object.getOwnPropertyNames(t);for(let m of this.outputNames)if(y.indexOf(m)!==-1){let g=t[m];(g===null||g instanceof Qi)&&(f=!0,u=!1,o[m]=g)}if(f){if(typeof n=="object"&&n!==null)i=n;else if(typeof n<"u")throw new TypeError("'options' must be an object.")}else i=t}}else if(typeof t<"u")throw new TypeError("Unexpected argument[1]: must be 'fetches' or 'options'.");for(let f of this.inputNames)if(typeof r[f]>"u")throw new Error(`input '${f}' is missing in 'feeds'.`);if(u)for(let f of this.outputNames)o[f]=null;let d=await this.handler.run(r,o,i),h={};for(let f in d)if(Object.hasOwnProperty.call(d,f)){let y=d[f];y instanceof Qi?h[f]=y:h[f]=new Qi(y.type,y.data,y.dims)}return Bs("InferenceSession.run"),ki(),h}async release(){return this.handler.dispose()}static async create(r,t,n,o){Xi(),zs("InferenceSession.create");let i,u={};if(typeof r=="string"){if(i=r,typeof t=="object"&&t!==null)u=t;else if(typeof t<"u")throw new TypeError("'options' must be an object.")}else if(r instanceof Uint8Array){if(i=r,typeof t=="object"&&t!==null)u=t;else if(typeof t<"u")throw new TypeError("'options' must be an object.")}else if(r instanceof ArrayBuffer||typeof SharedArrayBuffer<"u"&&r instanceof SharedArrayBuffer){let y=r,m=0,g=r.byteLength;if(typeof t=="object"&&t!==null)u=t;else if(typeof t=="number"){if(m=t,!Number.isSafeInteger(m))throw new RangeError("'byteOffset' must be an integer.");if(m<0||m>=y.byteLength)throw new RangeError(`'byteOffset' is out of range [0, ${y.byteLength}).`);if(g=r.byteLength-m,typeof n=="number"){if(g=n,!Number.isSafeInteger(g))throw new RangeError("'byteLength' must be an integer.");if(g<=0||m+g>y.byteLength)throw new RangeError(`'byteLength' is out of range (0, ${y.byteLength-m}].`);if(typeof o=="object"&&o!==null)u=o;else if(typeof o<"u")throw new TypeError("'options' must be an object.")}else if(typeof n<"u")throw new TypeError("'byteLength' must be a number.")}else if(typeof t<"u")throw new TypeError("'options' must be an object.");i=new Uint8Array(y,m,g)}else throw new TypeError("Unexpected argument[0]: must be 'path' or 'buffer'.");let[d,h]=await d0(u),f=await d.createInferenceSessionHandler(i,h);return Bs("InferenceSession.create"),ki(),new S0(f)}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}}}),Qm,Q1=Je(()=>{q1(),Qm=C0}),X1=Je(()=>{}),J1=Je(()=>{}),Y1=Je(()=>{}),Z1=Je(()=>{}),$0={};Ga($0,{InferenceSession:()=>Qm,TRACE:()=>yd,TRACE_EVENT_BEGIN:()=>zs,TRACE_EVENT_END:()=>Bs,TRACE_FUNC_BEGIN:()=>Xi,TRACE_FUNC_END:()=>ki,Tensor:()=>Qi,env:()=>dr,registerBackend:()=>vn});var Ii=Je(()=>{N1(),U1(),Q1(),E0(),X1(),J1(),P0(),Y1(),Z1()}),Xm=Je(()=>{}),k0={};Ga(k0,{default:()=>I0});var Af,Of,I0,eT=Je(()=>{var e;Lv(),Pn(),Jm(),Af="ort-wasm-proxy-worker",Of=((e=globalThis.self)==null?void 0:e.name)===Af,Of&&(self.onmessage=r=>{let{type:t,in:n}=r.data;try{switch(t){case"init-wasm":Ym(n.wasm).then(()=>{m_(n).then(()=>{postMessage({type:t})},o=>{postMessage({type:t,err:o})})},o=>{postMessage({type:t,err:o})});break;case"init-ep":{let{epName:o,env:i}=n;__(i,o).then(()=>{postMessage({type:t})},u=>{postMessage({type:t,err:u})});break}case"copy-from":{let{buffer:o}=n,i=of(o);postMessage({type:t,out:i});break}case"create":{let{model:o,options:i}=n;g_(o,i).then(u=>{postMessage({type:t,out:u})},u=>{postMessage({type:t,err:u})});break}case"release":y_(n),postMessage({type:t});break;case"run":{let{sessionId:o,inputIndices:i,inputs:u,outputIndices:d,options:h}=n;w_(o,i,u,d,new Array(d.length).fill(null),h).then(f=>{f.some(y=>y[3]!=="cpu")?postMessage({type:t,err:"Proxy does not support non-cpu tensor location."}):postMessage({type:t,out:f},b_([...u,...f]))},f=>{postMessage({type:t,err:f})});break}case"end-profiling":M_(n),postMessage({type:t});break;default:}}catch(o){postMessage({type:t,err:o})}}),I0=Of?null:r=>new Worker(r??bi,{type:"module",name:Af})}),A0={};Ga(A0,{default:()=>O0});var Ff,Df,O0,S_,tT=Je(()=>{var e,r;Df=(Ff=import.meta.url,async function(t={}){var T_;var n,o,i=t,u=new Promise((s,a)=>{n=s,o=a}),d=typeof window=="object",h=typeof WorkerGlobalScope<"u",f=h&&((T_=self.name)==null?void 0:T_.startsWith("em-pthread"));i.mountExternalData=(s,a)=>{s.startsWith("./")&&(s=s.substring(2)),(i.mg||(i.mg=new Map)).set(s,a)},i.unmountExternalData=()=>{delete i.mg};var y=globalThis.SharedArrayBuffer??new WebAssembly.Memory({initial:0,maximum:0,ul:!0}).buffer.constructor;let m=s=>async(...a)=>{var l;try{if(i.ng)throw Error("Session already started");let c=i.ng={jl:a[0],errors:[]},p=await s(...a);if(i.ng!==c)throw Error("Session mismatch");(l=i.Ok)==null||l.flush();let _=c.errors;if(0<_.length){let M=await Promise.all(_);if(M=M.filter(x=>x),0<M.length)throw Error(M.join(`
`))}return p}finally{i.ng=null}};i.jsepInit=(s,a)=>{if(s==="webgpu"){[i.Ok,i.$k,i.dl,i.Pk,i.cl,i.fg,i.el,i.gl,i.al,i.bl,i.fl]=a;let l=i.Ok;i.jsepRegisterBuffer=(c,p,_,M)=>l.registerBuffer(c,p,_,M),i.jsepGetBuffer=c=>l.getBuffer(c),i.jsepCreateDownloader=(c,p,_)=>l.createDownloader(c,p,_),i.jsepOnCreateSession=c=>{l.onCreateSession(c)},i.jsepOnReleaseSession=c=>{l.onReleaseSession(c)},i.jsepOnRunStart=c=>l.onRunStart(c),i.hl=(c,p)=>{l.upload(c,p)}}else if(s==="webnn"){let l=a[0];[i.rl,i.Tk,i.webnnEnsureTensor,i.Uk,i.webnnDownloadTensor,i.webnnEnableTraceEvent]=a.slice(1),i.webnnReleaseTensorId=i.Tk,i.webnnUploadTensor=i.Uk,i.webnnOnRunStart=c=>l.onRunStart(c),i.webnnOnRunEnd=l.onRunEnd.bind(l),i.webnnRegisterMLContext=(c,p)=>{l.registerMLContext(c,p)},i.webnnOnReleaseSession=c=>{l.onReleaseSession(c)},i.webnnCreateMLTensorDownloader=(c,p)=>l.createMLTensorDownloader(c,p),i.webnnRegisterMLTensor=(c,p,_,M)=>l.registerMLTensor(c,p,_,M),i.webnnCreateMLContext=c=>l.createMLContext(c),i.webnnRegisterMLConstant=(c,p,_,M,x,A)=>l.registerMLConstant(c,p,_,M,x,i.mg,A),i.webnnRegisterGraphInput=l.registerGraphInput.bind(l),i.webnnIsGraphInput=l.isGraphInput.bind(l),i.webnnRegisterGraphOutput=l.registerGraphOutput.bind(l),i.webnnIsGraphOutput=l.isGraphOutput.bind(l),i.webnnCreateTemporaryTensor=l.createTemporaryTensor.bind(l),i.webnnIsGraphInputOutputTypeSupported=l.isGraphInputOutputTypeSupported.bind(l)}};let g=()=>{let s=(a,l,c)=>(...p)=>{let _=li,M=l==null?void 0:l();p=a(...p);let x=l==null?void 0:l();return M!==x&&(a=x,c(M),l=c=null),li!=_?new Promise((A,L)=>{Sn={resolve:A,reject:L}}):p};(()=>{for(let a of["_OrtAppendExecutionProvider","_OrtCreateSession","_OrtRun","_OrtRunWithBinding","_OrtBindInput"])i[a]=s(i[a],()=>i[a],l=>i[a]=l)})(),m!==void 0&&(i._OrtRun=m(i._OrtRun),i._OrtRunWithBinding=m(i._OrtRunWithBinding)),g=void 0};i.asyncInit=()=>{g==null||g()};var v,b,k=Object.assign({},i),z=(s,a)=>{throw a},E="";(d||h)&&(h?E=self.location.href:typeof document<"u"&&document.currentScript&&(E=document.currentScript.src),Ff&&(E=Ff),E=E.startsWith("blob:")?"":E.slice(0,E.replace(/[?#].*/,"").lastIndexOf("/")+1),h&&(b=s=>{var a=new XMLHttpRequest;return a.open("GET",s,!1),a.responseType="arraybuffer",a.send(null),new Uint8Array(a.response)}),v=async s=>{if(q(s))return new Promise((l,c)=>{var p=new XMLHttpRequest;p.open("GET",s,!0),p.responseType="arraybuffer",p.onload=()=>{p.status==200||p.status==0&&p.response?l(p.response):c(p.status)},p.onerror=c,p.send(null)});var a=await fetch(s,{credentials:"same-origin"});if(a.ok)return a.arrayBuffer();throw Error(a.status+" : "+a.url)});var T=console.log.bind(console),D=console.error.bind(console),I=T,$=D;Object.assign(i,k),k=null;var P,S,O,R,G,ee,ie,H,ce,re,se,_e,ae,Ce=i.wasmBinary,Te=!1,q=s=>s.startsWith("file://");function B(){return P.buffer!=R.buffer&&pe(),R}function Q(){return P.buffer!=R.buffer&&pe(),G}function oe(){return P.buffer!=R.buffer&&pe(),ee}function Ie(){return P.buffer!=R.buffer&&pe(),ie}function ye(){return P.buffer!=R.buffer&&pe(),H}function be(){return P.buffer!=R.buffer&&pe(),ce}function et(){return P.buffer!=R.buffer&&pe(),re}function tt(){return P.buffer!=R.buffer&&pe(),ae}if(f){let s=function(a){try{var l=a.data,c=l.jg;if(c==="load"){let p=[];self.onmessage=_=>p.push(_),self.startWorker=()=>{postMessage({jg:"loaded"});for(let _ of p)s(_);self.onmessage=s};for(let _ of l.Xk)i[_]&&!i[_].proxy||(i[_]=(...M)=>{postMessage({jg:"callHandler",Wk:_,args:M})},_=="print"&&(I=i[_]),_=="printErr"&&($=i[_]));P=l.pl,pe(),Ee(l.ql)}else if(c==="run"){ue(l.ig),Qn(l.ig,0,0,1,0,0),U(),Ci(l.ig),Y||(Do(),Y=!0);try{$e(l.ll,l.pg)}catch(p){if(p!="unwind")throw p}}else l.target!=="setimmediate"&&(c==="checkMailbox"?Y&&sr():c&&($(`worker: received unknown command ${c}`),$(l)))}catch(p){throw Ko(),p}};var Ee,Y=!1;$=function(...a){a=a.join(" "),console.error(a)},self.alert=function(...a){postMessage({jg:"alert",text:a.join(" "),nl:Ss()})},self.onunhandledrejection=a=>{throw a.reason||a},self.onmessage=s}function pe(){var s=P.buffer;i.HEAP8=R=new Int8Array(s),i.HEAP16=ee=new Int16Array(s),i.HEAPU8=G=new Uint8Array(s),i.HEAPU16=ie=new Uint16Array(s),i.HEAP32=H=new Int32Array(s),i.HEAPU32=ce=new Uint32Array(s),i.HEAPF32=re=new Float32Array(s),i.HEAPF64=ae=new Float64Array(s),i.HEAP64=se=new BigInt64Array(s),i.HEAPU64=_e=new BigUint64Array(s)}function X(){f?startWorker(i):te.qf()}f||(P=new WebAssembly.Memory({initial:256,maximum:65536,shared:!0}),pe());var Ae,Oe=0,Ne=null;function Be(){if(--Oe==0&&Ne){var s=Ne;Ne=null,s()}}function Ve(s){throw $(s="Aborted("+s+")"),Te=!0,s=new WebAssembly.RuntimeError(s+". Build with -sASSERTIONS for more info."),o(s),s}function He(){return{a:{Ac:nt,$d:st,v:Dt,_:Et,b:nr,o:ri,A:pr,F:Jr,r:Ai,Yb:Ji,u:Oi,Bb:ns,of:Yi,h:gt,Ub:Zi,Hd:Gr,kf:es,mf:si,Id:qe,Fd:lt,yd:yt,Ed:ur,yc:gi,lf:Di,hf:ni,Gd:Li,jf:ji,Nd:Bi,qd:ts,cf:Vs,af:Me,pd:je,Ma:Ye,oa:ot,bf:it,sc:kr,df:Zr,Bd:Pi,ff:Ir,pf:yi,Ze:Ri,rd:Rr,Ad:Ci,Kd:ei,Ec:xr,be:Tr,he:vd,nb:Qa,ra:Xa,le:Ei,ae:Td,Fc:Ed,ka:Za,ob:Pd,fe:In,Y:Cd,Zb:Sd,ce:$d,ga:kd,ib:Id,ee:Ad,de:Od,Za:An,vd:eo,wd:Fd,xd:hf,td:ro,ud:Dd,_e:so,Pd:Ld,Md:zd,C:Bd,Xb:Rd,sd:no,Ld:jd,Db:Nd,Jd:Vd,gf:ao,ba:io,Yd:Ud,Cb:Wd,zc:oo,ef:Gd,Rd:Kd,Qd:co,Xd:mo,Dc:Jd,Cc:Yd,Bc:_o,_d:Zd,Wb:ep,xa:go,Wd:tp,Vd:rp,Zd:yo,Ud:ip,Td:sp,Sd:wo,Cd:Mo,Dd:bo,nf:cr,Vb:Bn,xc:vo,zd:xo,wc:Eo,Rb:Yp,Hb:fx,la:t1,S:i1,R:r1,Ba:Tf,Ha:Xp,Tc:ox,gc:hx,ke:Zx,Z:_s,y:jp,c:Uu,kd:yh,wa:Vp,me:Wx,Lc:Kx,g:mf,$a:Yx,ta:u1,d:ff,ja:F,k:Dp,ie:f1,j:Wu,Eb:Rx,s:Ap,t:sh,q:Qp,tb:hr,O:Hp,Aa:Yv,pa:uh,nc:fi,kb:Oa,cc:Ax,qb:yx,cd:N,nd:ih,$c:ze,lb:fs,ad:ke,_b:h1,sa:Mi,Sb:Kp,wb:Fh,ca:Jx,ma:C,ed:Or,N:fr,bd:me,Wa:or,H:yf,Jc:s1,yb:Th,Le:Se,za:lx,Ka:Ch,B:Up,qc:Mt,_c:Ke,hd:At,id:bf,rc:Ih,Je:at,Xa:vh,aa:sx,hb:o1,$b:c1,gb:d1,_a:n1,gd:Ah,Gc:p1,jc:Ef,fa:Jp,J:ah,E:Bp,Ja:Ph,T:Yu,Rc:wx,Zc:vt,fd:Sr,X:kt,e:Ip,bb:Bt,l:kp,ld:fh,Na:Hx,qa:Vr,Pb:zt,i:_f,md:hh,V:Oh,va:ui,Gb:Mx,ub:Ct,f:Gu,oe:zx,re:Dx,m:Fp,Wc:Er,n:Lp,pe:jx,Yc:Ot,se:Fx,Vc:ms,Mc:Gx,we:vx,p:Ku,ne:Ux,Ua:Os,Nb:gs,Ta:Da,fc:_x,D:Gp,G:Hu,K:wf,ab:Bx,qe:Lx,Ob:$r,$:dh,ha:gh,Ia:Sx,Fa:Aa,Ra:cx,Ie:ft,db:Eh,Ga:ph,Oc:$x,jb:Cx,Fb:Px,Pa:bx,Ge:Qr,ya:dx,te:Ix,Ae:ix,Ee:Pr,ic:Pf,Qa:ux,W:oh,sb:px,Uc:Jv,Jb:ys,ua:nx,Ya:wh,jd:kh,Qb:th,vb:Ze,dc:kx,P:mh,Ib:rx,na:eh,Qe:Mh,ea:ex,Ab:ch,bc:Ox,Pc:Ex,ac:Qx,Kc:qx,Nc:Vx,cb:vf,Ic:a1,zb:bh,z:Np,U:nh,Da:zp,La:Mf,Oe:xh,pc:Vt,Tb:Rp,L:Wp,rb:gx,Ea:tx,Hc:l1,Sc:mx,Se:lh,x:As,I:rh,da:Op,ze:ax,lc:Ju,Qc:Tx,Oa:Nx,fb:Zp,kc:La,eb:_h,hc:Zv,Xe:gf,Ne:$h,ec:xx,mc:Fa,pb:Xx,od:qp,oc:ar,ia:Dh,Xc:rr,M:br,xb:Sh,je:e1,Ca:ja,Q:Zu,Lb:Xu,Sa:Qu,Va:is,Kb:xf,Mb:qu,w:np,a:P,Od:xt,Re:Po,Te:ap,mb:op,vc:Co,Ue:lp,$e:cp,He:So,Ke:up,Be:dp,tc:$o,We:pp,Ce:hp,Pe:ko,De:fp,Ye:mp,uc:_p,ge:gp,Me:Io,ye:yp,ue:wp,xe:Ao,ve:Mp,dd:bp,Fe:Oo,Ve:vp}}}var Xe={1642916:(s,a,l,c,p)=>{if(i===void 0||!i.mg)return 1;if((s=er(Number(s>>>0))).startsWith("./")&&(s=s.substring(2)),!(s=i.mg.get(s)))return 2;if(a=Number(a>>>0),l=Number(l>>>0),c=Number(c>>>0),a+l>s.byteLength)return 3;try{let _=s.subarray(a,a+l);switch(p){case 0:Q().set(_,c>>>0);break;case 1:i.Sk?i.Sk(c,_):i.hl(c,_);break;default:return 4}return 0}catch{return 4}},1643740:(s,a,l)=>{s=Number(s),a=Number(a),l=er(l);let c=new Uint8Array(a);c.set(Q().subarray(s>>>0,s+a>>>0)),s=new File([c],l,{type:"application/octet-stream"}),s=URL.createObjectURL(s),window.open(s,"_blank")},1644264:(s,a,l)=>{i.Uk(s,Q().subarray(a>>>0,a+l>>>0))},1644328:()=>i.rl(),1644370:s=>{i.Tk(s)},1644407:()=>{i.al()},1644438:()=>{i.bl()},1644467:()=>{i.fl()},1644492:s=>i.$k(s),1644525:s=>i.dl(s),1644557:(s,a,l)=>{i.Pk(Number(s),Number(a),Number(l),!0)},1644620:(s,a,l)=>{i.Pk(Number(s),Number(a),Number(l))},1644677:()=>typeof wasmOffsetConverter<"u",1644734:s=>{i.fg("Abs",s,void 0)},1644785:s=>{i.fg("Neg",s,void 0)},1644836:s=>{i.fg("Floor",s,void 0)},1644889:s=>{i.fg("Ceil",s,void 0)},1644941:s=>{i.fg("Reciprocal",s,void 0)},1644999:s=>{i.fg("Sqrt",s,void 0)},1645051:s=>{i.fg("Exp",s,void 0)},1645102:s=>{i.fg("Erf",s,void 0)},1645153:s=>{i.fg("Sigmoid",s,void 0)},1645208:(s,a,l)=>{i.fg("HardSigmoid",s,{alpha:a,beta:l})},1645287:s=>{i.fg("Log",s,void 0)},1645338:s=>{i.fg("Sin",s,void 0)},1645389:s=>{i.fg("Cos",s,void 0)},1645440:s=>{i.fg("Tan",s,void 0)},1645491:s=>{i.fg("Asin",s,void 0)},1645543:s=>{i.fg("Acos",s,void 0)},1645595:s=>{i.fg("Atan",s,void 0)},1645647:s=>{i.fg("Sinh",s,void 0)},1645699:s=>{i.fg("Cosh",s,void 0)},1645751:s=>{i.fg("Asinh",s,void 0)},1645804:s=>{i.fg("Acosh",s,void 0)},1645857:s=>{i.fg("Atanh",s,void 0)},1645910:s=>{i.fg("Tanh",s,void 0)},1645962:s=>{i.fg("Not",s,void 0)},1646013:(s,a,l)=>{i.fg("Clip",s,{min:a,max:l})},1646082:s=>{i.fg("Clip",s,void 0)},1646134:(s,a)=>{i.fg("Elu",s,{alpha:a})},1646192:s=>{i.fg("Gelu",s,void 0)},1646244:s=>{i.fg("Relu",s,void 0)},1646296:(s,a)=>{i.fg("LeakyRelu",s,{alpha:a})},1646360:(s,a)=>{i.fg("ThresholdedRelu",s,{alpha:a})},1646430:(s,a)=>{i.fg("Cast",s,{to:a})},1646488:s=>{i.fg("Add",s,void 0)},1646539:s=>{i.fg("Sub",s,void 0)},1646590:s=>{i.fg("Mul",s,void 0)},1646641:s=>{i.fg("Div",s,void 0)},1646692:s=>{i.fg("Pow",s,void 0)},1646743:s=>{i.fg("Equal",s,void 0)},1646796:s=>{i.fg("Greater",s,void 0)},1646851:s=>{i.fg("GreaterOrEqual",s,void 0)},1646913:s=>{i.fg("Less",s,void 0)},1646965:s=>{i.fg("LessOrEqual",s,void 0)},1647024:(s,a,l,c,p)=>{i.fg("ReduceMean",s,{keepDims:!!a,noopWithEmptyAxes:!!l,axes:c?Array.from(ye().subarray(Number(c)>>>0,Number(p)>>>0)):[]})},1647199:(s,a,l,c,p)=>{i.fg("ReduceMax",s,{keepDims:!!a,noopWithEmptyAxes:!!l,axes:c?Array.from(ye().subarray(Number(c)>>>0,Number(p)>>>0)):[]})},1647373:(s,a,l,c,p)=>{i.fg("ReduceMin",s,{keepDims:!!a,noopWithEmptyAxes:!!l,axes:c?Array.from(ye().subarray(Number(c)>>>0,Number(p)>>>0)):[]})},1647547:(s,a,l,c,p)=>{i.fg("ReduceProd",s,{keepDims:!!a,noopWithEmptyAxes:!!l,axes:c?Array.from(ye().subarray(Number(c)>>>0,Number(p)>>>0)):[]})},1647722:(s,a,l,c,p)=>{i.fg("ReduceSum",s,{keepDims:!!a,noopWithEmptyAxes:!!l,axes:c?Array.from(ye().subarray(Number(c)>>>0,Number(p)>>>0)):[]})},1647896:(s,a,l,c,p)=>{i.fg("ReduceL1",s,{keepDims:!!a,noopWithEmptyAxes:!!l,axes:c?Array.from(ye().subarray(Number(c)>>>0,Number(p)>>>0)):[]})},1648069:(s,a,l,c,p)=>{i.fg("ReduceL2",s,{keepDims:!!a,noopWithEmptyAxes:!!l,axes:c?Array.from(ye().subarray(Number(c)>>>0,Number(p)>>>0)):[]})},1648242:(s,a,l,c,p)=>{i.fg("ReduceLogSum",s,{keepDims:!!a,noopWithEmptyAxes:!!l,axes:c?Array.from(ye().subarray(Number(c)>>>0,Number(p)>>>0)):[]})},1648419:(s,a,l,c,p)=>{i.fg("ReduceSumSquare",s,{keepDims:!!a,noopWithEmptyAxes:!!l,axes:c?Array.from(ye().subarray(Number(c)>>>0,Number(p)>>>0)):[]})},1648599:(s,a,l,c,p)=>{i.fg("ReduceLogSumExp",s,{keepDims:!!a,noopWithEmptyAxes:!!l,axes:c?Array.from(ye().subarray(Number(c)>>>0,Number(p)>>>0)):[]})},1648779:s=>{i.fg("Where",s,void 0)},1648832:(s,a,l)=>{i.fg("Transpose",s,{perm:a?Array.from(ye().subarray(Number(a)>>>0,Number(l)>>>0)):[]})},1648956:(s,a,l,c)=>{i.fg("DepthToSpace",s,{blocksize:a,mode:er(l),format:c?"NHWC":"NCHW"})},1649089:(s,a,l,c)=>{i.fg("DepthToSpace",s,{blocksize:a,mode:er(l),format:c?"NHWC":"NCHW"})},1649222:(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe)=>{i.fg("ConvTranspose",s,{format:A?"NHWC":"NCHW",autoPad:a,dilations:[l],group:c,kernelShape:[p],pads:[_,M],strides:[x],wIsConst:()=>!!B()[L>>>0],outputPadding:V?Array.from(ye().subarray(Number(V)>>>0,Number(W)>>>0)):[],outputShape:Z?Array.from(ye().subarray(Number(Z)>>>0,Number(de)>>>0)):[],activation:er(Pe)})},1649655:(s,a,l,c,p,_,M,x,A,L,V,W,Z,de)=>{i.fg("ConvTranspose",s,{format:x?"NHWC":"NCHW",autoPad:a,dilations:Array.from(ye().subarray(Number(l)>>>0,2+(Number(l)>>>0)>>>0)),group:c,kernelShape:Array.from(ye().subarray(Number(p)>>>0,2+(Number(p)>>>0)>>>0)),pads:Array.from(ye().subarray(Number(_)>>>0,4+(Number(_)>>>0)>>>0)),strides:Array.from(ye().subarray(Number(M)>>>0,2+(Number(M)>>>0)>>>0)),wIsConst:()=>!!B()[A>>>0],outputPadding:L?Array.from(ye().subarray(Number(L)>>>0,Number(V)>>>0)):[],outputShape:W?Array.from(ye().subarray(Number(W)>>>0,Number(Z)>>>0)):[],activation:er(de)})},1650316:(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe)=>{i.fg("ConvTranspose",s,{format:A?"NHWC":"NCHW",autoPad:a,dilations:[l],group:c,kernelShape:[p],pads:[_,M],strides:[x],wIsConst:()=>!!B()[L>>>0],outputPadding:V?Array.from(ye().subarray(Number(V)>>>0,Number(W)>>>0)):[],outputShape:Z?Array.from(ye().subarray(Number(Z)>>>0,Number(de)>>>0)):[],activation:er(Pe)})},1650749:(s,a,l,c,p,_,M,x,A,L,V,W,Z,de)=>{i.fg("ConvTranspose",s,{format:x?"NHWC":"NCHW",autoPad:a,dilations:Array.from(ye().subarray(Number(l)>>>0,2+(Number(l)>>>0)>>>0)),group:c,kernelShape:Array.from(ye().subarray(Number(p)>>>0,2+(Number(p)>>>0)>>>0)),pads:Array.from(ye().subarray(Number(_)>>>0,4+(Number(_)>>>0)>>>0)),strides:Array.from(ye().subarray(Number(M)>>>0,2+(Number(M)>>>0)>>>0)),wIsConst:()=>!!B()[A>>>0],outputPadding:L?Array.from(ye().subarray(Number(L)>>>0,Number(V)>>>0)):[],outputShape:W?Array.from(ye().subarray(Number(W)>>>0,Number(Z)>>>0)):[],activation:er(de)})},1651410:(s,a)=>{i.fg("GlobalAveragePool",s,{format:a?"NHWC":"NCHW"})},1651501:(s,a,l,c,p,_,M,x,A,L,V,W,Z,de)=>{i.fg("AveragePool",s,{format:de?"NHWC":"NCHW",auto_pad:a,ceil_mode:l,count_include_pad:c,storage_order:p,dilations:_?Array.from(ye().subarray(Number(_)>>>0,Number(M)>>>0)):[],kernel_shape:x?Array.from(ye().subarray(Number(x)>>>0,Number(A)>>>0)):[],pads:L?Array.from(ye().subarray(Number(L)>>>0,Number(V)>>>0)):[],strides:W?Array.from(ye().subarray(Number(W)>>>0,Number(Z)>>>0)):[]})},1651980:(s,a)=>{i.fg("GlobalAveragePool",s,{format:a?"NHWC":"NCHW"})},1652071:(s,a,l,c,p,_,M,x,A,L,V,W,Z,de)=>{i.fg("AveragePool",s,{format:de?"NHWC":"NCHW",auto_pad:a,ceil_mode:l,count_include_pad:c,storage_order:p,dilations:_?Array.from(ye().subarray(Number(_)>>>0,Number(M)>>>0)):[],kernel_shape:x?Array.from(ye().subarray(Number(x)>>>0,Number(A)>>>0)):[],pads:L?Array.from(ye().subarray(Number(L)>>>0,Number(V)>>>0)):[],strides:W?Array.from(ye().subarray(Number(W)>>>0,Number(Z)>>>0)):[]})},1652550:(s,a)=>{i.fg("GlobalMaxPool",s,{format:a?"NHWC":"NCHW"})},1652637:(s,a,l,c,p,_,M,x,A,L,V,W,Z,de)=>{i.fg("MaxPool",s,{format:de?"NHWC":"NCHW",auto_pad:a,ceil_mode:l,count_include_pad:c,storage_order:p,dilations:_?Array.from(ye().subarray(Number(_)>>>0,Number(M)>>>0)):[],kernel_shape:x?Array.from(ye().subarray(Number(x)>>>0,Number(A)>>>0)):[],pads:L?Array.from(ye().subarray(Number(L)>>>0,Number(V)>>>0)):[],strides:W?Array.from(ye().subarray(Number(W)>>>0,Number(Z)>>>0)):[]})},1653112:(s,a)=>{i.fg("GlobalMaxPool",s,{format:a?"NHWC":"NCHW"})},1653199:(s,a,l,c,p,_,M,x,A,L,V,W,Z,de)=>{i.fg("MaxPool",s,{format:de?"NHWC":"NCHW",auto_pad:a,ceil_mode:l,count_include_pad:c,storage_order:p,dilations:_?Array.from(ye().subarray(Number(_)>>>0,Number(M)>>>0)):[],kernel_shape:x?Array.from(ye().subarray(Number(x)>>>0,Number(A)>>>0)):[],pads:L?Array.from(ye().subarray(Number(L)>>>0,Number(V)>>>0)):[],strides:W?Array.from(ye().subarray(Number(W)>>>0,Number(Z)>>>0)):[]})},1653674:(s,a,l,c,p)=>{i.fg("Gemm",s,{alpha:a,beta:l,transA:c,transB:p})},1653778:s=>{i.fg("MatMul",s,void 0)},1653832:(s,a,l,c)=>{i.fg("ArgMax",s,{keepDims:!!a,selectLastIndex:!!l,axis:c})},1653940:(s,a,l,c)=>{i.fg("ArgMin",s,{keepDims:!!a,selectLastIndex:!!l,axis:c})},1654048:(s,a)=>{i.fg("Softmax",s,{axis:a})},1654111:(s,a)=>{i.fg("Concat",s,{axis:a})},1654171:(s,a,l,c,p)=>{i.fg("Split",s,{axis:a,numOutputs:l,splitSizes:c?Array.from(ye().subarray(Number(c)>>>0,Number(p)>>>0)):[]})},1654327:s=>{i.fg("Expand",s,void 0)},1654381:(s,a)=>{i.fg("Gather",s,{axis:Number(a)})},1654452:(s,a)=>{i.fg("GatherElements",s,{axis:Number(a)})},1654531:(s,a)=>{i.fg("GatherND",s,{batch_dims:Number(a)})},1654610:(s,a,l,c,p,_,M,x,A,L,V)=>{i.fg("Resize",s,{antialias:a,axes:l?Array.from(ye().subarray(Number(l)>>>0,Number(c)>>>0)):[],coordinateTransformMode:er(p),cubicCoeffA:_,excludeOutside:M,extrapolationValue:x,keepAspectRatioPolicy:er(A),mode:er(L),nearestMode:er(V)})},1654972:(s,a,l,c,p,_,M)=>{i.fg("Slice",s,{starts:a?Array.from(ye().subarray(Number(a)>>>0,Number(l)>>>0)):[],ends:c?Array.from(ye().subarray(Number(c)>>>0,Number(p)>>>0)):[],axes:_?Array.from(ye().subarray(Number(_)>>>0,Number(M)>>>0)):[]})},1655236:s=>{i.fg("Tile",s,void 0)},1655288:(s,a,l)=>{i.fg("InstanceNormalization",s,{epsilon:a,format:l?"NHWC":"NCHW"})},1655402:(s,a,l)=>{i.fg("InstanceNormalization",s,{epsilon:a,format:l?"NHWC":"NCHW"})},1655516:s=>{i.fg("Range",s,void 0)},1655569:(s,a)=>{i.fg("Einsum",s,{equation:er(a)})},1655650:(s,a,l,c,p)=>{i.fg("Pad",s,{mode:a,value:l,pads:c?Array.from(ye().subarray(Number(c)>>>0,Number(p)>>>0)):[]})},1655793:(s,a,l,c,p,_)=>{i.fg("BatchNormalization",s,{epsilon:a,momentum:l,spatial:!!p,trainingMode:!!c,format:_?"NHWC":"NCHW"})},1655962:(s,a,l,c,p,_)=>{i.fg("BatchNormalization",s,{epsilon:a,momentum:l,spatial:!!p,trainingMode:!!c,format:_?"NHWC":"NCHW"})},1656131:(s,a,l)=>{i.fg("CumSum",s,{exclusive:Number(a),reverse:Number(l)})},1656228:(s,a,l)=>{i.fg("DequantizeLinear",s,{axis:a,blockSize:l})},1656318:(s,a,l,c,p)=>{i.fg("GridSample",s,{align_corners:a,mode:er(l),padding_mode:er(c),format:p?"NHWC":"NCHW"})},1656488:(s,a,l,c,p)=>{i.fg("GridSample",s,{align_corners:a,mode:er(l),padding_mode:er(c),format:p?"NHWC":"NCHW"})},1656658:(s,a)=>{i.fg("ScatterND",s,{reduction:er(a)})},1656743:(s,a,l,c,p,_,M,x,A)=>{i.fg("Attention",s,{numHeads:a,isUnidirectional:l,maskFilterValue:c,scale:p,doRotary:_,qkvHiddenSizes:M?Array.from(ye().subarray(Number(x)>>>0,Number(x)+M>>>0)):[],pastPresentShareBuffer:!!A})},1657015:s=>{i.fg("BiasAdd",s,void 0)},1657070:s=>{i.fg("BiasSplitGelu",s,void 0)},1657131:s=>{i.fg("FastGelu",s,void 0)},1657187:(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue)=>{i.fg("Conv",s,{format:W?"NHWC":"NCHW",auto_pad:a,dilations:l?Array.from(ye().subarray(Number(l)>>>0,Number(c)>>>0)):[],group:p,kernel_shape:_?Array.from(ye().subarray(Number(_)>>>0,Number(M)>>>0)):[],pads:x?Array.from(ye().subarray(Number(x)>>>0,Number(A)>>>0)):[],strides:L?Array.from(ye().subarray(Number(L)>>>0,Number(V)>>>0)):[],w_is_const:()=>!!B()[Number(Z)>>>0],activation:er(de),activation_params:Pe?Array.from(et().subarray(Number(Pe)>>>0,Number(Ue)>>>0)):[]})},1657771:s=>{i.fg("Gelu",s,void 0)},1657823:(s,a,l,c,p,_,M,x,A)=>{i.fg("GroupQueryAttention",s,{numHeads:a,kvNumHeads:l,scale:c,softcap:p,doRotary:_,rotaryInterleaved:M,smoothSoftmax:x,localWindowSize:A})},1658040:(s,a,l,c)=>{i.fg("LayerNormalization",s,{axis:a,epsilon:l,simplified:!!c})},1658151:(s,a,l,c)=>{i.fg("LayerNormalization",s,{axis:a,epsilon:l,simplified:!!c})},1658262:(s,a,l,c,p,_)=>{i.fg("MatMulNBits",s,{k:a,n:l,accuracyLevel:c,bits:p,blockSize:_})},1658389:(s,a,l,c,p,_)=>{i.fg("MultiHeadAttention",s,{numHeads:a,isUnidirectional:l,maskFilterValue:c,scale:p,doRotary:_})},1658548:(s,a)=>{i.fg("QuickGelu",s,{alpha:a})},1658612:(s,a,l,c,p)=>{i.fg("RotaryEmbedding",s,{interleaved:!!a,numHeads:l,rotaryEmbeddingDim:c,scale:p})},1658751:(s,a,l)=>{i.fg("SkipLayerNormalization",s,{epsilon:a,simplified:!!l})},1658853:(s,a,l)=>{i.fg("SkipLayerNormalization",s,{epsilon:a,simplified:!!l})},1658955:(s,a,l,c)=>{i.fg("GatherBlockQuantized",s,{gatherAxis:a,quantizeAxis:l,blockSize:c})},1659076:s=>{i.el(s)},1659110:(s,a)=>i.gl(Number(s),Number(a),i.ng.jl,i.ng.errors)};function st(s,a,l){return $n(async()=>{await i.cl(Number(s),Number(a),Number(l))})}function nt(){return typeof wasmOffsetConverter<"u"}class We{constructor(a){le(this,"name","ExitStatus");this.message=`Program terminated with exit(${a})`,this.status=a}}var rt=s=>{s.terminate(),s.onmessage=()=>{}},pt=[],_t=s=>{Kt.length==0&&(K(),ge(Kt[0]));var a=Kt.pop();if(!a)return 6;wr.push(a),$t[s.ig]=a,a.ig=s.ig;var l={jg:"run",ll:s.kl,pg:s.pg,ig:s.ig};return a.postMessage(l,s.Rk),0},ct=0,dt=s=>-9007199254740992>s||9007199254740992<s?NaN:Number(s),Le=(s,a,...l)=>{for(var c=2*l.length,p=fe(),_=Xs(8*c),M=_>>>3,x=0;x<l.length;x++){var A=l[x];typeof A=="bigint"?(se[M+2*x]=1n,se[M+2*x+1]=A):(se[M+2*x]=0n,tt()[M+2*x+1>>>0]=A)}return s=Xn(s,0,c,_,a),he(p),s};function xt(s){if(f)return Le(0,1,s);if(O=s,!(0<ct)){for(var a of wr)rt(a);for(a of Kt)rt(a);Kt=[],wr=[],$t={},Te=!0}z(0,new We(s))}function Nt(s){if(f)return Le(1,0,s);cr(s)}var cr=s=>{if(O=s,f)throw Nt(s),"unwind";xt(s)},Kt=[],wr=[],zr=[],$t={},Ti=s=>{var a=s.ig;delete $t[a],Kt.push(s),wr.splice(wr.indexOf(s),1),s.ig=0,Ho(a)};function U(){zr.forEach(s=>s())}var ge=s=>new Promise(a=>{s.onmessage=p=>{var _=(p=p.data).jg;if(p.og&&p.og!=Ss()){var M=$t[p.og];M?M.postMessage(p,p.Rk):$(`Internal error! Worker sent a message "${_}" to target pthread ${p.og}, but that thread no longer exists!`)}else _==="checkMailbox"?sr():_==="spawnThread"?_t(p):_==="cleanupThread"?Ti($t[p.ml]):_==="loaded"?(s.loaded=!0,a(s)):_==="alert"?alert(`Thread ${p.nl}: ${p.text}`):p.target==="setimmediate"?s.postMessage(p):_==="callHandler"?i[p.Wk](...p.args):_&&$(`worker sent an unknown command ${_}`)},s.onerror=p=>{throw $(`worker sent an error! ${p.filename}:${p.lineno}: ${p.message}`),p};var l,c=[];for(l of[])i.propertyIsEnumerable(l)&&c.push(l);s.postMessage({jg:"load",Xk:c,pl:P,ql:S})});function K(){var s=new Worker((()=>{let a=URL;return import.meta.url>"file:"&&import.meta.url<"file;"?new a("ort.bundle.min.mjs",import.meta.url):new URL(import.meta.url)})(),{type:"module",workerData:"em-pthread",name:"em-pthread"});Kt.push(s)}var ue=s=>{pe();var a=be()[s+52>>>2>>>0];s=be()[s+56>>>2>>>0],Qo(a,a-s),he(a)},$e=(s,a)=>{ct=0,s=Ys(s,a),0<ct?O=s:Jn(s)},Ge=[],De=0;function Dt(s){var a=new jt(s>>>=0);if(B()[a.hg+12>>>0]==0){var l=1;B()[a.hg+12>>>0]=l,De--}return l=0,B()[a.hg+13>>>0]=l,Ge.push(a),Xo(s),Js(s)}var Ut=0,Et=()=>{we(0,0);var s=Ge.pop();ea(s.Bg),Ut=0};class jt{constructor(a){this.Bg=a,this.hg=a-24}}function gt(s){throw Ut||(Ut=s>>>0),Ut}var Ht=s=>{var a=Ut;if(!a)return $s(0),0;var l=new jt(a);be()[l.hg+16>>>2>>>0]=a;var c=be()[l.hg+4>>>2>>>0];if(!c)return $s(0),a;for(var p of s){if(p===0||p===c)break;if(Jo(p,c,l.hg+16))return $s(p),a}return $s(c),a};function nr(){return Ht([])}function ri(s){return Ht([s>>>0])}function pr(s,a){return Ht([s>>>0,a>>>0])}function Jr(s,a,l){return Ht([s>>>0,a>>>0,l>>>0])}function Ai(s,a,l,c){return Ht([s>>>0,a>>>0,l>>>0,c>>>0])}var Ji=()=>{var s=Ge.pop();s||Ve("no exception to throw");var a=s.Bg;if(B()[s.hg+13>>>0]==0){Ge.push(s);var l=1;B()[s.hg+13>>>0]=l,l=0,B()[s.hg+12>>>0]=l,De++}throw Ut=a};function Oi(s,a,l){var c=new jt(s>>>=0);throw a>>>=0,l>>>=0,be()[c.hg+16>>>2>>>0]=0,be()[c.hg+4>>>2>>>0]=a,be()[c.hg+8>>>2>>>0]=l,De++,Ut=s}var ns=()=>De;function ii(s,a,l,c){return f?Le(2,1,s,a,l,c):Yi(s,a,l,c)}function Yi(s,a,l,c){if(s>>>=0,l>>>=0,c>>>=0,y===void 0)return 6;var p=[];return f&&p.length===0?ii(s,a>>>=0,l,c):(s={kl:l,ig:s,pg:c,Rk:p},f?(s.jg="spawnThread",postMessage(s,p),0):_t(s))}var _i=typeof TextDecoder<"u"?new TextDecoder:void 0,Fi=(s,a=0,l=NaN)=>{var c=(a>>>=0)+l;for(l=a;s[l]&&!(l>=c);)++l;if(16<l-a&&s.buffer&&_i)return _i.decode(s.buffer instanceof ArrayBuffer?s.subarray(a,l):s.slice(a,l));for(c="";a<l;){var p=s[a++];if(128&p){var _=63&s[a++];if((224&p)==192)c+=String.fromCharCode((31&p)<<6|_);else{var M=63&s[a++];65536>(p=(240&p)==224?(15&p)<<12|_<<6|M:(7&p)<<18|_<<12|M<<6|63&s[a++])?c+=String.fromCharCode(p):(p-=65536,c+=String.fromCharCode(55296|p>>10,56320|1023&p))}}else c+=String.fromCharCode(p)}return c},er=(s,a)=>(s>>>=0)?Fi(Q(),s,a):"";function Zi(s,a,l){return f?Le(3,1,s,a,l):0}function Gr(s,a){if(f)return Le(4,1,s,a)}var Kr=s=>{for(var a=0,l=0;l<s.length;++l){var c=s.charCodeAt(l);127>=c?a++:2047>=c?a+=2:55296<=c&&57343>=c?(a+=4,++l):a+=3}return a},Hr=(s,a,l)=>{var c=Q();if(a>>>=0,0<l){var p=a;l=a+l-1;for(var _=0;_<s.length;++_){var M=s.charCodeAt(_);if(55296<=M&&57343>=M&&(M=65536+((1023&M)<<10)|1023&s.charCodeAt(++_)),127>=M){if(a>=l)break;c[a++>>>0]=M}else{if(2047>=M){if(a+1>=l)break;c[a++>>>0]=192|M>>6}else{if(65535>=M){if(a+2>=l)break;c[a++>>>0]=224|M>>12}else{if(a+3>=l)break;c[a++>>>0]=240|M>>18,c[a++>>>0]=128|M>>12&63}c[a++>>>0]=128|M>>6&63}c[a++>>>0]=128|63&M}}c[a>>>0]=0,s=a-p}else s=0;return s};function es(s,a){if(f)return Le(5,1,s,a)}function si(s,a,l){if(f)return Le(6,1,s,a,l)}function qe(s,a,l){return f?Le(7,1,s,a,l):0}function lt(s,a){if(f)return Le(8,1,s,a)}function yt(s,a,l){if(f)return Le(9,1,s,a,l)}function ur(s,a,l,c){if(f)return Le(10,1,s,a,l,c)}function gi(s,a,l,c){if(f)return Le(11,1,s,a,l,c)}function Di(s,a,l,c){if(f)return Le(12,1,s,a,l,c)}function ni(s){if(f)return Le(13,1,s)}function Li(s,a){if(f)return Le(14,1,s,a)}function ji(s,a,l){if(f)return Le(15,1,s,a,l)}var zi,qr,Bi=()=>Ve(""),Yr=s=>{for(var a="";Q()[s>>>0];)a+=zi[Q()[s++>>>0]];return a},as={},os={};function ai(s,a,l={}){return function(c,p,_={}){var M=p.name;if(!c)throw new qr(`type "${M}" must have a positive integer typeid pointer`);if(os.hasOwnProperty(c)){if(_.Yk)return;throw new qr(`Cannot register type '${M}' twice`)}os[c]=p,as.hasOwnProperty(c)&&(p=as[c],delete as[c],p.forEach(x=>x()))}(s,a,l)}var xs=(s,a,l)=>{switch(a){case 1:return l?c=>B()[c>>>0]:c=>Q()[c>>>0];case 2:return l?c=>oe()[c>>>1>>>0]:c=>Ie()[c>>>1>>>0];case 4:return l?c=>ye()[c>>>2>>>0]:c=>be()[c>>>2>>>0];case 8:return l?c=>se[c>>>3]:c=>_e[c>>>3];default:throw new TypeError(`invalid integer width (${a}): ${s}`)}};function ts(s,a,l){l>>>=0,ai(s>>>=0,{name:a=Yr(a>>>0),fromWireType:c=>c,toWireType:function(c,p){if(typeof p!="bigint"&&typeof p!="number")throw p=p===null?"null":(c=typeof p)=="object"||c==="array"||c==="function"?p.toString():""+p,new TypeError(`Cannot convert "${p}" to ${this.name}`);return typeof p=="number"&&(p=BigInt(p)),p},kg:pi,readValueFromPointer:xs(a,l,a.indexOf("u")==-1),lg:null})}var pi=8;function Vs(s,a,l,c){ai(s>>>=0,{name:a=Yr(a>>>0),fromWireType:function(p){return!!p},toWireType:function(p,_){return _?l:c},kg:pi,readValueFromPointer:function(p){return this.fromWireType(Q()[p>>>0])},lg:null})}var ls=[],oi=[];function Ei(s){9<(s>>>=0)&&--oi[s+1]==0&&(oi[s]=void 0,ls.push(s))}var ve=s=>{if(!s)throw new qr("Cannot use deleted val. handle = "+s);return oi[s]},j=s=>{switch(s){case void 0:return 2;case null:return 4;case!0:return 6;case!1:return 8;default:let a=ls.pop()||oi.length;return oi[a]=s,oi[a+1]=1,a}};function J(s){return this.fromWireType(be()[s>>>2>>>0])}var ne={name:"emscripten::val",fromWireType:s=>{var a=ve(s);return Ei(s),a},toWireType:(s,a)=>j(a),kg:pi,readValueFromPointer:J,lg:null};function Me(s){return ai(s>>>0,ne)}var xe=(s,a)=>{switch(a){case 4:return function(l){return this.fromWireType(et()[l>>>2>>>0])};case 8:return function(l){return this.fromWireType(tt()[l>>>3>>>0])};default:throw new TypeError(`invalid float width (${a}): ${s}`)}};function je(s,a,l){l>>>=0,ai(s>>>=0,{name:a=Yr(a>>>0),fromWireType:c=>c,toWireType:(c,p)=>p,kg:pi,readValueFromPointer:xe(a,l),lg:null})}function Ye(s,a,l,c,p){if(s>>>=0,l>>>=0,a=Yr(a>>>0),p===-1&&(p=4294967295),p=x=>x,c===0){var _=32-8*l;p=x=>x<<_>>>_}var M=a.includes("unsigned")?function(x,A){return A>>>0}:function(x,A){return A};ai(s,{name:a,fromWireType:p,toWireType:M,kg:pi,readValueFromPointer:xs(a,l,c!==0),lg:null})}function ot(s,a,l){function c(_){var M=be()[_>>>2>>>0];return _=be()[_+4>>>2>>>0],new p(B().buffer,_,M)}var p=[Int8Array,Uint8Array,Int16Array,Uint16Array,Int32Array,Uint32Array,Float32Array,Float64Array,BigInt64Array,BigUint64Array][a];ai(s>>>=0,{name:l=Yr(l>>>0),fromWireType:c,kg:pi,readValueFromPointer:c},{Yk:!0})}function it(s,a){ai(s>>>=0,{name:a=Yr(a>>>0),fromWireType:function(l){for(var c,p=be()[l>>>2>>>0],_=l+4,M=_,x=0;x<=p;++x){var A=_+x;x!=p&&Q()[A>>>0]!=0||(M=er(M,A-M),c===void 0?c=M:(c+="\0",c+=M),M=A+1)}return ci(l),c},toWireType:function(l,c){c instanceof ArrayBuffer&&(c=new Uint8Array(c));var p=typeof c=="string";if(!(p||c instanceof Uint8Array||c instanceof Uint8ClampedArray||c instanceof Int8Array))throw new qr("Cannot pass non-string to std::string");var _=p?Kr(c):c.length,M=ds(4+_+1),x=M+4;if(be()[M>>>2>>>0]=_,p)Hr(c,x,_+1);else if(p)for(p=0;p<_;++p){var A=c.charCodeAt(p);if(255<A)throw ci(M),new qr("String has UTF-16 code units that do not fit in 8 bits");Q()[x+p>>>0]=A}else for(p=0;p<_;++p)Q()[x+p>>>0]=c[p];return l!==null&&l.push(ci,M),M},kg:pi,readValueFromPointer:J,lg(l){ci(l)}})}var ut=typeof TextDecoder<"u"?new TextDecoder("utf-16le"):void 0,It=(s,a)=>{for(var l=s>>1,c=l+a/2;!(l>=c)&&Ie()[l>>>0];)++l;if(32<(l<<=1)-s&&ut)return ut.decode(Q().slice(s,l));for(l="",c=0;!(c>=a/2);++c){var p=oe()[s+2*c>>>1>>>0];if(p==0)break;l+=String.fromCharCode(p)}return l},Qt=(s,a,l)=>{if(l??(l=2147483647),2>l)return 0;var c=a;l=(l-=2)<2*s.length?l/2:s.length;for(var p=0;p<l;++p){var _=s.charCodeAt(p);oe()[a>>>1>>>0]=_,a+=2}return oe()[a>>>1>>>0]=0,a-c},qt=s=>2*s.length,_r=(s,a)=>{for(var l=0,c="";!(l>=a/4);){var p=ye()[s+4*l>>>2>>>0];if(p==0)break;++l,65536<=p?(p-=65536,c+=String.fromCharCode(55296|p>>10,56320|1023&p)):c+=String.fromCharCode(p)}return c},tr=(s,a,l)=>{if(a>>>=0,l??(l=2147483647),4>l)return 0;var c=a;l=c+l-4;for(var p=0;p<s.length;++p){var _=s.charCodeAt(p);if(55296<=_&&57343>=_&&(_=65536+((1023&_)<<10)|1023&s.charCodeAt(++p)),ye()[a>>>2>>>0]=_,(a+=4)+4>l)break}return ye()[a>>>2>>>0]=0,a-c},Cr=s=>{for(var a=0,l=0;l<s.length;++l){var c=s.charCodeAt(l);55296<=c&&57343>=c&&++l,a+=4}return a};function kr(s,a,l){if(s>>>=0,a>>>=0,l=Yr(l>>>=0),a===2)var c=It,p=Qt,_=qt,M=x=>Ie()[x>>>1>>>0];else a===4&&(c=_r,p=tr,_=Cr,M=x=>be()[x>>>2>>>0]);ai(s,{name:l,fromWireType:x=>{for(var A,L=be()[x>>>2>>>0],V=x+4,W=0;W<=L;++W){var Z=x+4+W*a;W!=L&&M(Z)!=0||(V=c(V,Z-V),A===void 0?A=V:(A+="\0",A+=V),V=Z+a)}return ci(x),A},toWireType:(x,A)=>{if(typeof A!="string")throw new qr(`Cannot pass non-string to C++ string type ${l}`);var L=_(A),V=ds(4+L+a);return be()[V>>>2>>>0]=L/a,p(A,V+4,L+a),x!==null&&x.push(ci,V),V},kg:pi,readValueFromPointer:J,lg(x){ci(x)}})}function Zr(s,a){ai(s>>>=0,{Zk:!0,name:a=Yr(a>>>0),kg:0,fromWireType:()=>{},toWireType:()=>{}})}function Pi(s){Qn(s>>>0,!h,1,!d,131072,!1),U()}var Fr=s=>{if(!Te)try{if(s(),!(0<ct))try{f?Jn(O):cr(O)}catch(a){a instanceof We||a=="unwind"||z(0,a)}}catch(a){a instanceof We||a=="unwind"||z(0,a)}};function Ci(s){s>>>=0,typeof Atomics.ol=="function"&&(Atomics.ol(ye(),s>>>2,s).value.then(sr),s+=128,Atomics.store(ye(),s>>>2,1))}var sr=()=>{var s=Ss();s&&(Ci(s),Fr(qo))};function Ir(s,a){(s>>>=0)==a>>>0?setTimeout(sr):f?postMessage({og:s,jg:"checkMailbox"}):(s=$t[s])&&s.postMessage({jg:"checkMailbox"})}var Br=[];function yi(s,a,l,c,p){for(a>>>=0,c/=2,Br.length=c,l=p>>>0>>>3,p=0;p<c;p++)Br[p]=se[l+2*p]?se[l+2*p+1]:tt()[l+2*p+1>>>0];return(a?Xe[a]:xp[s])(...Br)}var Ri=()=>{ct=0};function Rr(s){s>>>=0,f?postMessage({jg:"cleanupThread",ml:s}):Ti($t[s])}function ei(s){}var Mr=(s,a)=>{var l=os[s];if(l===void 0)throw s=Rn(s),l=Yr(s),ci(s),new qr(`${a} has unknown type ${l}`);return l},gr=(s,a,l)=>{var c=[];return s=s.toWireType(c,l),c.length&&(be()[a>>>2>>>0]=j(c)),s};function xr(s,a,l){return a>>>=0,l>>>=0,s=ve(s>>>0),a=Mr(a,"emval::as"),gr(a,l,s)}function Tr(s,a){return a>>>=0,s=ve(s>>>0),(a=Mr(a,"emval::as")).toWireType(null,s)}var Nr=s=>{try{s()}catch(a){Ve(a)}},Dr=0,li=null,Ka=0,Us=[],Ha={},qa={},Ts=0,Sn=null,bd=[];function $n(s){return function(a){if(!Te){if(Dr===0){var l=!1,c=!1;a((p=0)=>{if(!Te&&(Ka=p,l=!0,c)){Dr=2,Nr(()=>Wt(li)),typeof MainLoop<"u"&&MainLoop.Vk&&MainLoop.resume(),p=!1;try{var _=function(){var A=ye()[li+8>>>2>>>0];return A=te[qa[A]],--ct,A()}()}catch(A){_=A,p=!0}var M=!1;if(!li){var x=Sn;x&&(Sn=null,(p?x.reject:x.resolve)(_),M=!0)}if(p&&!M)throw _}}),c=!0,l||(Dr=1,li=function(){var p=ds(65548),_=p+12;be()[p>>>2>>>0]=_,be()[p+4>>>2>>>0]=_+65536,_=Us[0];var M=Ha[_];return M===void 0&&(M=Ts++,Ha[_]=M,qa[M]=_),_=M,ye()[p+8>>>2>>>0]=_,p}(),typeof MainLoop<"u"&&MainLoop.Vk&&MainLoop.pause(),Nr(()=>Nu(li)))}else Dr===2?(Dr=0,Nr($p),ci(li),li=null,bd.forEach(Fr)):Ve(`invalid state: ${Dr}`);return Ka}}(a=>{s().then(a)})}function vd(s){return s>>>=0,$n(async()=>{var a=await ve(s);return j(a)})}var Ws=[];function Qa(s,a,l,c){return l>>>=0,c>>>=0,(s=Ws[s>>>0])(null,a=ve(a>>>0),l,c)}var xd={},Gs=s=>{var a=xd[s];return a===void 0?Yr(s):a};function Xa(s,a,l,c,p){return l>>>=0,c>>>=0,p>>>=0,(s=Ws[s>>>0])(a=ve(a>>>0),a[l=Gs(l)],c,p)}function Td(s,a){return a>>>=0,(s=ve(s>>>0))==ve(a)}var kn=()=>typeof globalThis=="object"?globalThis:Function("return this")();function Ed(s){return(s>>>=0)==0?j(kn()):(s=Gs(s),j(kn()[s]))}var Ja=s=>{var a=Ws.length;return Ws.push(s),a},pf=(s,a)=>{for(var l=Array(s),c=0;c<s;++c)l[c]=Mr(be()[a+4*c>>>2>>>0],"parameter "+c);return l},Ya=(s,a)=>Object.defineProperty(a,"name",{value:s});function Za(s,a,l){var c=(a=pf(s,a>>>0)).shift();s--;var p=`return function (obj, func, destructorsRef, args) {
`,_=0,M=[];l===0&&M.push("obj");for(var x=["retType"],A=[c],L=0;L<s;++L)M.push("arg"+L),x.push("argType"+L),A.push(a[L]),p+=`  var arg${L} = argType${L}.readValueFromPointer(args${_?"+"+_:""});
`,_+=a[L].kg;return p+=`  var rv = ${l===1?"new func":"func.call"}(${M.join(", ")});
`,c.Zk||(x.push("emval_returnValue"),A.push(gr),p+=`  return emval_returnValue(retType, destructorsRef, rv);
`),x.push(p+`};
`),s=function(V){var W=Function;if(!(W instanceof Function))throw new TypeError(`new_ called with constructor type ${typeof W} which is not a function`);var Z=Ya(W.name||"unknownFunctionName",function(){});return Z.prototype=W.prototype,Z=new Z,(V=W.apply(Z,V))instanceof Object?V:Z}(x)(...A),l=`methodCaller<(${a.map(V=>V.name).join(", ")}) => ${c.name}>`,Ja(Ya(l,s))}function Pd(s){return s=Gs(s>>>0),j(i[s])}function In(s,a){return a>>>=0,s=ve(s>>>0),a=ve(a),j(s[a])}function Cd(s){9<(s>>>=0)&&(oi[s+1]+=1)}function Sd(){return j([])}function $d(s){s=ve(s>>>0);for(var a=Array(s.length),l=0;l<s.length;l++)a[l]=s[l];return j(a)}function kd(s){return j(Gs(s>>>0))}function Id(){return j({})}function Ad(s){for(var a=ve(s>>>=0);a.length;){var l=a.pop();a.pop()(l)}Ei(s)}function Od(s,a,l){a>>>=0,l>>>=0,s=ve(s>>>0),a=ve(a),l=ve(l),s[a]=l}function An(s,a){return a>>>=0,s=(s=Mr(s>>>0,"_emval_take_value")).readValueFromPointer(a),j(s)}function eo(s,a){s=dt(s),a>>>=0,s=new Date(1e3*s),ye()[a>>>2>>>0]=s.getUTCSeconds(),ye()[a+4>>>2>>>0]=s.getUTCMinutes(),ye()[a+8>>>2>>>0]=s.getUTCHours(),ye()[a+12>>>2>>>0]=s.getUTCDate(),ye()[a+16>>>2>>>0]=s.getUTCMonth(),ye()[a+20>>>2>>>0]=s.getUTCFullYear()-1900,ye()[a+24>>>2>>>0]=s.getUTCDay(),s=(s.getTime()-Date.UTC(s.getUTCFullYear(),0,1,0,0,0,0))/864e5|0,ye()[a+28>>>2>>>0]=s}var to=s=>s%4==0&&(s%100!=0||s%400==0),On=[0,31,60,91,121,152,182,213,244,274,305,335],rs=[0,31,59,90,120,151,181,212,243,273,304,334];function Fd(s,a){s=dt(s),a>>>=0,s=new Date(1e3*s),ye()[a>>>2>>>0]=s.getSeconds(),ye()[a+4>>>2>>>0]=s.getMinutes(),ye()[a+8>>>2>>>0]=s.getHours(),ye()[a+12>>>2>>>0]=s.getDate(),ye()[a+16>>>2>>>0]=s.getMonth(),ye()[a+20>>>2>>>0]=s.getFullYear()-1900,ye()[a+24>>>2>>>0]=s.getDay();var l=(to(s.getFullYear())?On:rs)[s.getMonth()]+s.getDate()-1|0;ye()[a+28>>>2>>>0]=l,ye()[a+36>>>2>>>0]=-60*s.getTimezoneOffset(),l=new Date(s.getFullYear(),6,1).getTimezoneOffset();var c=new Date(s.getFullYear(),0,1).getTimezoneOffset();s=0|(l!=c&&s.getTimezoneOffset()==Math.min(c,l)),ye()[a+32>>>2>>>0]=s}function hf(s){s>>>=0;var a=new Date(ye()[s+20>>>2>>>0]+1900,ye()[s+16>>>2>>>0],ye()[s+12>>>2>>>0],ye()[s+8>>>2>>>0],ye()[s+4>>>2>>>0],ye()[s>>>2>>>0],0),l=ye()[s+32>>>2>>>0],c=a.getTimezoneOffset(),p=new Date(a.getFullYear(),6,1).getTimezoneOffset(),_=new Date(a.getFullYear(),0,1).getTimezoneOffset(),M=Math.min(_,p);return 0>l?ye()[s+32>>>2>>>0]=+(p!=_&&M==c):0<l!=(M==c)&&(p=Math.max(_,p),a.setTime(a.getTime()+6e4*((0<l?M:p)-c))),ye()[s+24>>>2>>>0]=a.getDay(),l=(to(a.getFullYear())?On:rs)[a.getMonth()]+a.getDate()-1|0,ye()[s+28>>>2>>>0]=l,ye()[s>>>2>>>0]=a.getSeconds(),ye()[s+4>>>2>>>0]=a.getMinutes(),ye()[s+8>>>2>>>0]=a.getHours(),ye()[s+12>>>2>>>0]=a.getDate(),ye()[s+16>>>2>>>0]=a.getMonth(),ye()[s+20>>>2>>>0]=a.getYear(),s=a.getTime(),BigInt(isNaN(s)?-1:s/1e3)}function ro(s,a,l,c,p,_,M){return f?Le(16,1,s,a,l,c,p,_,M):-52}function Dd(s,a,l,c,p,_){if(f)return Le(17,1,s,a,l,c,p,_)}var Es={},io=()=>performance.timeOrigin+performance.now();function so(s,a){if(f)return Le(18,1,s,a);if(Es[s]&&(clearTimeout(Es[s].id),delete Es[s]),!a)return 0;var l=setTimeout(()=>{delete Es[s],Fr(()=>Yn(s,performance.timeOrigin+performance.now()))},a);return Es[s]={id:l,wl:a},0}function Ld(s,a,l,c){s>>>=0,a>>>=0,l>>>=0,c>>>=0;var p=new Date().getFullYear(),_=new Date(p,0,1).getTimezoneOffset();p=new Date(p,6,1).getTimezoneOffset();var M=Math.max(_,p);be()[s>>>2>>>0]=60*M,ye()[a>>>2>>>0]=+(_!=p),s=(a=x=>{var A=Math.abs(x);return`UTC${0<=x?"-":"+"}${String(Math.floor(A/60)).padStart(2,"0")}${String(A%60).padStart(2,"0")}`})(_),a=a(p),p<_?(Hr(s,l,17),Hr(a,c,17)):(Hr(s,c,17),Hr(a,l,17))}var jd=()=>Date.now();function zd(s,a,l){return 0<=s&&3>=s?(s===0?s=Date.now():s=performance.timeOrigin+performance.now(),se[l>>>0>>>3]=BigInt(Math.round(1e6*s)),0):28}var Fn=[],Ks=(s,a)=>{Fn.length=0;for(var l;l=Q()[s++>>>0];){var c=l!=105;a+=(c&=l!=112)&&a%8?4:0,Fn.push(l==112?be()[a>>>2>>>0]:l==106?se[a>>>3]:l==105?ye()[a>>>2>>>0]:tt()[a>>>3>>>0]),a+=c?8:4}return Fn};function Bd(s,a,l){return s>>>=0,a=Ks(a>>>0,l>>>0),Xe[s](...a)}function Rd(s,a,l){return s>>>=0,a=Ks(a>>>0,l>>>0),Xe[s](...a)}var cs=()=>{var s="getMappedRange size=0 no longer means WGPU_WHOLE_MAP_SIZE";cs.tg||(cs.tg={}),cs.tg[s]||(cs.tg[s]=1,$(s))},no=()=>{};function Nd(s,a){return $(er(s>>>0,a>>>0))}var Vd=()=>{throw ct+=1,"unwind"};function ao(){return 4294901760}var Ud=()=>1,Wd=()=>navigator.hardwareConcurrency;function oo(){return Ve("Cannot use emscripten_pc_get_function without -sUSE_OFFSET_CONVERTER"),0}function Gd(s){s>>>=0;var a=Q().length;if(s<=a||4294901760<s)return!1;for(var l=1;4>=l;l*=2){var c=a*(1+.2/l);c=Math.min(c,s+100663296);e:{c=(Math.min(4294901760,65536*Math.ceil(Math.max(s,c)/65536))-P.buffer.byteLength+65535)/65536|0;try{P.grow(c),pe();var p=1;break e}catch{}p=void 0}if(p)return!0}return!1}var Hs=()=>(Ve("Cannot use convertFrameToPC (needed by __builtin_return_address) without -sUSE_OFFSET_CONVERTER"),0),us={},lo=s=>{s.forEach(a=>{Hs()})};function Kd(){var s=Error().stack.toString().split(`
`);return s[0]=="Error"&&s.shift(),lo(s),us.Qk=Hs(),us.il=s,us.Qk}function co(s,a,l){if(s>>>=0,a>>>=0,us.Qk==s)var c=us.il;else(c=Error().stack.toString().split(`
`))[0]=="Error"&&c.shift(),lo(c);for(var p=3;c[p]&&Hs()!=s;)++p;for(s=0;s<l&&c[s+p];++s)ye()[a+4*s>>>2>>>0]=Hs();return s}var Si=s=>{var a=Kr(s)+1,l=Xs(a);return Hr(s,l,a),l},Ar=[],hi=[],qs=[],Ps=(s,a)=>{qs[s]=new Promise(l=>a.finally(()=>l(s)))},Tt=s=>{if(s)return Ar[s]},Qs=(s,a,l)=>{be()[s>>>2>>>0]=a,be()[s+4>>>2>>>0]=l},uo=s=>{var a=be()[s>>>2>>>0];return s=be()[s+4>>>2>>>0],er(a,s)},wi=s=>{var a=be()[s>>>2>>>0];return s=be()[s+4>>>2>>>0],a?er(a,s):s===0?"":void 0},po=(s,a)=>{function l(p,_){p=s[p],ye()[a+_>>>2>>>0]=p}function c(p,_){se[a+_>>>3]=BigInt(s[p])}l("maxTextureDimension1D",4),l("maxTextureDimension2D",8),l("maxTextureDimension3D",12),l("maxTextureArrayLayers",16),l("maxBindGroups",20),l("maxBindGroupsPlusVertexBuffers",24),l("maxBindingsPerBindGroup",28),l("maxDynamicUniformBuffersPerPipelineLayout",32),l("maxDynamicStorageBuffersPerPipelineLayout",36),l("maxSampledTexturesPerShaderStage",40),l("maxSamplersPerShaderStage",44),l("maxStorageBuffersPerShaderStage",48),l("maxStorageTexturesPerShaderStage",52),l("maxUniformBuffersPerShaderStage",56),l("minUniformBufferOffsetAlignment",80),l("minStorageBufferOffsetAlignment",84),c("maxUniformBufferBindingSize",64),c("maxStorageBufferBindingSize",72),l("maxVertexBuffers",88),c("maxBufferSize",96),l("maxVertexAttributes",104),l("maxVertexBufferArrayStride",108),l("maxInterStageShaderVariables",112),l("maxColorAttachments",116),l("maxColorAttachmentBytesPerSample",120),l("maxComputeWorkgroupStorageSize",124),l("maxComputeInvocationsPerWorkgroup",128),l("maxComputeWorkgroupSizeX",132),l("maxComputeWorkgroupSizeY",136),l("maxComputeWorkgroupSizeZ",140),l("maxComputeWorkgroupsPerDimension",144)},Hd={undefined:1,yl:1,sl:2},ho={1:"validation",2:"out-of-memory",3:"internal"},qd=[,"compatibility","core"],Dn={1:"depth-clip-control",2:"depth32float-stencil8",3:"timestamp-query",4:"texture-compression-bc",5:"texture-compression-bc-sliced-3d",6:"texture-compression-etc2",7:"texture-compression-astc",8:"texture-compression-astc-sliced-3d",9:"indirect-first-instance",10:"shader-f16",11:"rg11b10ufloat-renderable",12:"bgra8unorm-storage",13:"float32-filterable",14:"float32-blendable",15:"clip-distances",16:"dual-source-blending",17:"subgroups",18:"core-features-and-limits",327688:"subgroups-f16",327693:"chromium-experimental-unorm16-texture-formats",327694:"chromium-experimental-snorm16-texture-formats",327733:"chromium-experimental-multi-draw-indirect"},fo=[,"low-power","high-performance"],Qd={1:"occlusion",2:"timestamp"},Xd={"depth-clip-control":"1","depth32float-stencil8":"2","timestamp-query":"3","texture-compression-bc":"4","texture-compression-bc-sliced-3d":"5","texture-compression-etc2":"6","texture-compression-astc":"7","texture-compression-astc-sliced-3d":"8","indirect-first-instance":"9","shader-f16":"10","rg11b10ufloat-renderable":"11","bgra8unorm-storage":"12","float32-filterable":"13","float32-blendable":"14","clip-distances":"15","dual-source-blending":"16",vl:"17","core-features-and-limits":"18","subgroups-f16":"327688","chromium-experimental-unorm16-texture-formats":"327693","chromium-experimental-snorm16-texture-formats":"327694","chromium-experimental-multi-draw-indirect":"327733"};function mo(s,a,l,c,p,_){a=dt(a),l=dt(l),c>>>=0,p>>>=0,_>>>=0,s=Tt(s>>>0);var M={};if(_){var x=be()[_+12>>>2>>>0];if(x){var A=be()[_+16>>>2>>>0];M.requiredFeatures=Array.from(be().subarray(A>>>2>>>0,A+4*x>>>2>>>0),W=>Dn[W])}var L=be()[_+20>>>2>>>0];if(L){let W=function(de,Pe){Pe=L+Pe,(Pe=be()[Pe>>>2>>>0])!=4294967295&&(V[de]=Pe)},Z=function(de,Pe){Pe=L+Pe;var Ue=be()[Pe>>>2>>>0],Qe=be()[Pe+4>>>2>>>0];Ue==4294967295&&Qe==4294967295||(V[de]=4294967296*be()[Pe+4>>>2>>>0]+be()[Pe>>>2>>>0])};var V={};W("maxTextureDimension1D",4),W("maxTextureDimension2D",8),W("maxTextureDimension3D",12),W("maxTextureArrayLayers",16),W("maxBindGroups",20),W("maxBindGroupsPlusVertexBuffers",24),W("maxDynamicUniformBuffersPerPipelineLayout",32),W("maxDynamicStorageBuffersPerPipelineLayout",36),W("maxSampledTexturesPerShaderStage",40),W("maxSamplersPerShaderStage",44),W("maxStorageBuffersPerShaderStage",48),W("maxStorageTexturesPerShaderStage",52),W("maxUniformBuffersPerShaderStage",56),W("minUniformBufferOffsetAlignment",80),W("minStorageBufferOffsetAlignment",84),Z("maxUniformBufferBindingSize",64),Z("maxStorageBufferBindingSize",72),W("maxVertexBuffers",88),Z("maxBufferSize",96),W("maxVertexAttributes",104),W("maxVertexBufferArrayStride",108),W("maxInterStageShaderVariables",112),W("maxColorAttachments",116),W("maxColorAttachmentBytesPerSample",120),W("maxComputeWorkgroupStorageSize",124),W("maxComputeInvocationsPerWorkgroup",128),W("maxComputeWorkgroupSizeX",132),W("maxComputeWorkgroupSizeY",136),W("maxComputeWorkgroupSizeZ",140),W("maxComputeWorkgroupsPerDimension",144),M.requiredLimits=V}(x=be()[_+24>>>2>>>0])&&(x={label:wi(x+4)},M.defaultQueue=x),M.label=wi(_+4)}ct+=1,Ps(a,s.requestDevice(M).then(W=>{--ct,Ar[p]=W.queue,Ar[c]=W,l&&(ct+=1,Ps(l,W.lost.then(Z=>{--ct,W.onuncapturederror=()=>{};var de=fe(),Pe=Si(Z.message);Gn(l,Hd[Z.reason],Pe),he(de)}))),W.onuncapturederror=Z=>{var de=5;Z.error instanceof GPUValidationError?de=2:Z.error instanceof GPUOutOfMemoryError?de=3:Z.error instanceof GPUInternalError&&(de=4);var Pe=fe();Z=Si(Z.error.message),Go(c,de,Z),he(Pe)},qn(a,1,c,0)},W=>{--ct;var Z=fe();W=Si(W.message),qn(a,3,c,W),l&&Gn(l,4,W),he(Z)}))}function Jd(s){var a=Tt(s>>>=0),l=hi[s];if(l){for(var c=0;c<l.length;++c)l[c]();delete hi[s]}a.destroy()}function Yd(s,a,l){a>>>=0,l>>>=0;var c=Tt(s>>>=0);l===0&&cs(),l==4294967295&&(l=void 0);try{var p=c.getMappedRange(a,l)}catch{return 0}var _=Zn(16,p.byteLength);return Q().set(new Uint8Array(p),_>>>0),hi[s].push(()=>ci(_)),_}function _o(s,a,l){a>>>=0,l>>>=0;var c=Tt(s>>>=0);l===0&&cs(),l==4294967295&&(l=void 0);try{var p=c.getMappedRange(a,l)}catch{return 0}var _=Zn(16,p.byteLength);return Q().fill(0,_,p.byteLength),hi[s].push(()=>{new Uint8Array(p).set(Q().subarray(_>>>0,_+p.byteLength>>>0)),ci(_)}),_}function Zd(s,a,l,c,p){s>>>=0,a=dt(a),l=dt(l),p>>>=0;var _=Tt(s);hi[s]=[],p==4294967295&&(p=void 0),ct+=1,Ps(a,_.mapAsync(l,c>>>0,p).then(()=>{--ct,Kn(a,1,0)},M=>{--ct,fe();var x=Si(M.message);Kn(a,M.name==="AbortError"?4:M.name==="OperationError"?3:0,x),delete hi[s]}))}function ep(s){var a=Tt(s>>>=0),l=hi[s];if(l){for(var c=0;c<l.length;++c)l[c]();delete hi[s],a.unmap()}}function go(s){delete Ar[s>>>0]}function tp(s,a,l){s>>>=0,a>>>=0,l>>>=0;var c=!!be()[a+32>>>2>>>0];a={label:wi(a+4),usage:be()[a+16>>>2>>>0],size:4294967296*be()[a+28>>>2>>>0]+be()[a+24>>>2>>>0],mappedAtCreation:c},s=Tt(s);try{var p=s.createBuffer(a)}catch{return!1}return Ar[l]=p,c&&(hi[l]=[]),!0}function rp(s,a,l){s>>>=0,a>>>=0,l>>>=0;var c=be()[a>>>2>>>0],p=be()[c+4>>>2>>>0];a={label:wi(a+4),code:""},p===2&&(a.code=uo(c+8)),s=Tt(s).createShaderModule(a),Ar[l]=s}var yo=s=>{Tt(s).destroy()};function ip(s,a){a=dt(a),s=Tt(s>>>0),ct+=1,Ps(a,s.popErrorScope().then(l=>{--ct;var c=5;l?l instanceof GPUValidationError?c=2:l instanceof GPUOutOfMemoryError?c=3:l instanceof GPUInternalError&&(c=4):c=1;var p=fe();l=l?Si(l.message):0,Hn(a,1,c,l),he(p)},l=>{--ct;var c=fe();l=Si(l.message),Hn(a,1,5,l),he(c)}))}function sp(s,a,l,c){if(a=dt(a),c>>>=0,l>>>=0){var p=be()[l+4>>>2>>>0];p={featureLevel:qd[p],powerPreference:fo[be()[l+8>>>2>>>0]],forceFallbackAdapter:!!be()[l+12>>>2>>>0]},(l=be()[l>>>2>>>0])!==0&&(be(),p.Al=!!be()[l+8>>>2>>>0])}"gpu"in navigator?(ct+=1,Ps(a,navigator.gpu.requestAdapter(p).then(_=>{if(--ct,_)Ar[c]=_,ps(a,1,c,0);else{_=fe();var M=Si("WebGPU not available on this browser (requestAdapter returned null)");ps(a,3,c,M),he(_)}},_=>{--ct;var M=fe();_=Si(_.message),ps(a,4,c,_),he(M)}))):(l=fe(),p=Si("WebGPU not available on this browser (navigator.gpu is not available)"),ps(a,3,c,p),he(l))}function wo(s,a,l){return s>>>=0,a>>>=0,l>>>=0,$n(async()=>{var c=[];if(l){var p=(4294967296*be()[l+4>>>2>>>0]+be()[l>>>2>>>0])/1e6;c.length=a+1,c[a]=new Promise(x=>setTimeout(x,p,0))}else c.length=a;for(var _=0;_<a;++_){var M=4294967296*be()[s+8*_+4>>>2>>>0]+be()[s+8*_>>>2>>>0];if(!(M in qs))return M;c[_]=qs[M]}return c=await Promise.race(c),delete qs[c],c})}var Ln,jn={},zn=()=>{if(!Ln){var s,a={USER:"web_user",LOGNAME:"web_user",PATH:"/",PWD:"/",HOME:"/home/web_user",LANG:(typeof navigator=="object"&&navigator.languages&&navigator.languages[0]||"C").replace("-","_")+".UTF-8",_:"./this.program"};for(s in jn)jn[s]===void 0?delete a[s]:a[s]=jn[s];var l=[];for(s in a)l.push(`${s}=${a[s]}`);Ln=l}return Ln};function Mo(s,a){if(f)return Le(19,1,s,a);s>>>=0,a>>>=0;var l=0;return zn().forEach((c,p)=>{var _=a+l;for(p=be()[s+4*p>>>2>>>0]=_,_=0;_<c.length;++_)B()[p++>>>0]=c.charCodeAt(_);B()[p>>>0]=0,l+=c.length+1}),0}function bo(s,a){if(f)return Le(20,1,s,a);s>>>=0,a>>>=0;var l=zn();be()[s>>>2>>>0]=l.length;var c=0;return l.forEach(p=>c+=p.length+1),be()[a>>>2>>>0]=c,0}function Bn(s){return f?Le(21,1,s):52}function vo(s,a,l,c){return f?Le(22,1,s,a,l,c):52}function xo(s,a,l,c){return f?Le(23,1,s,a,l,c):70}var To=[null,[],[]];function Eo(s,a,l,c){if(f)return Le(24,1,s,a,l,c);a>>>=0,l>>>=0,c>>>=0;for(var p=0,_=0;_<l;_++){var M=be()[a>>>2>>>0],x=be()[a+4>>>2>>>0];a+=8;for(var A=0;A<x;A++){var L=Q()[M+A>>>0],V=To[s];L===0||L===10?((s===1?I:$)(Fi(V)),V.length=0):V.push(L)}p+=x}return be()[c>>>2>>>0]=p,0}function np(s){return s>>>0}function Po(s,a){return po(Tt(s>>>0).limits,a>>>0),1}function ap(s,a){return Tt(s>>>0).features.has(Dn[a])}function op(s){return BigInt(Tt(s>>>0).size)}function Co(s){return BigInt(Tt(s>>>0).usage)}function lp(s,a){if(s>>>=0,a>>>=0){var l=wi(a+4);l={label:l,timestampWrites:a=(a=be()[a+12>>>2>>>0])!==0?{querySet:Tt(be()[a+4>>>2>>>0]),beginningOfPassWriteIndex:be()[a+8>>>2>>>0],endOfPassWriteIndex:be()[a+12>>>2>>>0]}:void 0}}return a=Tt(s),s=Ro(0),l=a.beginComputePass(l),Ar[s]=l,s}function cp(s,a,l,c,p,_){l=dt(l),p=dt(p),_=dt(_),Tt(s>>>0).copyBufferToBuffer(Tt(a>>>0),l,Tt(c>>>0),p,_)}function So(s){var a=Tt(s>>>0);return s=Bo(0),a=a.finish(),Ar[s]=a,s}function up(s,a,l,c,p,_){_=dt(_),Tt(s>>>0).resolveQuerySet(Tt(a>>>0),l,c,Tt(p>>>0),_)}function dp(s,a,l,c){Tt(s>>>0).dispatchWorkgroups(a,l,c)}function $o(s){Tt(s>>>0).end()}function pp(s,a,l,c,p){if(c>>>=0,p>>>=0,s=Tt(s>>>0),l=Tt(l>>>0),c==0)s.setBindGroup(a,l);else{for(var _=[],M=0;M<c;M++,p+=4)_.push(be()[p>>>2>>>0]);s.setBindGroup(a,l,_)}}function hp(s,a){Tt(s>>>0).setPipeline(Tt(a>>>0))}function ko(s,a,l){Tt(s>>>0).zl(Tt(a>>>0),l)}function fp(s,a){var l=Tt(s>>>0);return s=zo(0),a=l.getBindGroupLayout(a),Ar[s]=a,s}function mp(s,a){s>>>=0;var l=wi(4+(a>>>=0)),c=Tt(be()[a+12>>>2>>>0]),p=be()[a+16>>>2>>>0];a=be()[a+20>>>2>>>0];for(var _=[],M=0;M<p;++M){var x=_,A=x.push,L=a+40*M,V=be()[L+8>>>2>>>0],W=be()[L+32>>>2>>>0],Z=be()[L+36>>>2>>>0],de=be()[L+4>>>2>>>0];V?(W=L+24,(W=be()[W>>>2>>>0]+4294967296*ye()[W+4>>>2>>>0])==-1&&(W=void 0),L={binding:de,resource:{buffer:Tt(V),offset:4294967296*be()[L+4+16>>>2>>>0]+be()[L+16>>>2>>>0],size:W}}):L=W?{binding:de,resource:Tt(W)}:{binding:de,resource:Tt(Z)},A.call(x,L)}return l={label:l,layout:c,entries:_},c=Tt(s),s=jo(0),l=c.createBindGroup(l),Ar[s]=l,s}function _p(s,a){var l;return s>>>=0,(a>>>=0)&&(l={label:wi(a+4)}),a=Tt(s),s=Un(0),l=a.createCommandEncoder(l),Ar[s]=l,s}function gp(s,a){s>>>=0;var l=a>>>0;a=wi(l+4);var c=(c=be()[l+12>>>2>>>0])?Tt(c):"auto";if(l+=16){var p=Tt(be()[l+4>>>2>>>0]),_=be()[l+16>>>2>>>0],M=be()[l+20>>>2>>>0];if(_){for(var x={},A=0;A<_;++A){var L=M+24*A;x[uo(L+4)]=tt()[L+16>>>3>>>0]}_=x}else _=void 0;l={module:p,constants:_,entryPoint:wi(l+8)}}else l=void 0;return a={label:a,layout:c,compute:l},c=Tt(s),s=No(0),a=c.createComputePipeline(a),Ar[s]=a,s}function Io(s,a){s>>>=0,a>>>=0,a={type:Qd[be()[a+12>>>2>>>0]],count:be()[a+16>>>2>>>0]};var l=Tt(s);return s=Vo(0),a=l.createQuerySet(a),Ar[s]=a,s}function yp(s,a){a>>>=0;var l=(s=Tt(s>>>0)).adapterInfo.vendor+s.adapterInfo.architecture+s.adapterInfo.device+s.adapterInfo.description,c=Kr(l)+1,p=ds(c);return p&&Hr(l,p,c),l=p,c=Kr(s.adapterInfo.vendor),Qs(a+4,l,c),l+=c,c=Kr(s.adapterInfo.architecture),Qs(a+12,l,c),l+=c,c=Kr(s.adapterInfo.device),Qs(a+20,l,c),Qs(a+28,l+c,Kr(s.adapterInfo.description)),ye()[a+36>>>2>>>0]=2,s=s.adapterInfo.isFallbackAdapter?3:4,ye()[a+40>>>2>>>0]=s,ye()[a+44>>>2>>>0]=0,ye()[a+48>>>2>>>0]=0,1}function wp(s,a){a>>>=0,s=Tt(s>>>0);var l=ds(4*s.features.size),c=0,p=0;s.features.forEach(_=>{(_=Xd[_])!==void 0&&(ye()[l+c>>>2>>>0]=_,c+=4,p++)}),be()[a+4>>>2>>>0]=l,be()[a>>>2>>>0]=p}function Ao(s,a){return po(Tt(s>>>0).limits,a>>>0),1}function Mp(s,a){return Tt(s>>>0).features.has(Dn[a])}function bp(s,a){Tt(s>>>0).pushErrorScope(ho[a])}function Oo(s,a,l){a>>>=0,l>>>=0,s=Tt(s>>>0),a=Array.from(ye().subarray(l>>>2>>>0,l+4*a>>>2>>>0),c=>Tt(c)),s.submit(a)}function vp(s,a,l,c,p){l=dt(l),c>>>=0,p>>>=0,s=Tt(s>>>0),a=Tt(a>>>0),c=Q().subarray(c>>>0,c+p>>>0),s.writeBuffer(a,l,c,0,p)}f||function(){for(var s=i.numThreads-1;s--;)K();pt.unshift(()=>{Oe++,function(a){f?a():Promise.all(Kt.map(ge)).then(a)}(()=>Be())})}();for(var Fo=Array(256),Cs=0;256>Cs;++Cs)Fo[Cs]=String.fromCharCode(Cs);zi=Fo,qr=i.BindingError=class extends Error{constructor(s){super(s),this.name="BindingError"}},i.InternalError=class extends Error{constructor(s){super(s),this.name="InternalError"}},oi.push(0,1,void 0,1,null,1,!0,1,!1,1),i.count_emval_handles=()=>oi.length/2-5-ls.length;var te,xp=[xt,Nt,ii,Zi,Gr,es,si,qe,lt,yt,ur,gi,Di,ni,Li,ji,ro,Dd,so,Mo,bo,Bn,vo,xo,Eo];(async function(){function s(c,p){return te=c.exports,te=function(){var _=te,M={};for(let[x,A]of Object.entries(_))M[x]=typeof A=="function"?(...L)=>{Us.push(x);try{return A(...L)}finally{Te||(Us.pop(),li&&Dr===1&&Us.length===0&&(Dr=0,ct+=1,Nr(Vu),typeof Fibers<"u"&&Fibers.xl()))}}:A;return M}(),te=function(){var _=te,M=A=>L=>A(L)>>>0,x=A=>()=>A()>>>0;return(_=Object.assign({},_)).rf=M(_.rf),_.Wf=x(_.Wf),_.Yf=M(_.Yf),_.Jg=(A=>(L,V)=>A(L,V)>>>0)(_.Jg),_.Og=M(_.Og),_.Pg=x(_.Pg),_.Tg=M(_.Tg),_}(),zr.push(te.Ag),S=p,Be(),te}Oe++;var a=He();if(i.instantiateWasm)return new Promise(c=>{i.instantiateWasm(a,(p,_)=>{s(p,_),c(p.exports)})});if(f)return new Promise(c=>{Ee=p=>{var _=new WebAssembly.Instance(p,He());c(s(_,p))}});Ae??(Ae=i.locateFile?i.locateFile?i.locateFile("ort-wasm-simd-threaded.jsep.wasm",E):E+"ort-wasm-simd-threaded.jsep.wasm":new URL("/assets/ort-wasm-simd-threaded.jsep-D5UMYBk4.wasm",import.meta.url).href);try{var l=await async function(c){var p=Ae;if(!Ce&&typeof WebAssembly.instantiateStreaming=="function"&&!q(p))try{var _=fetch(p,{credentials:"same-origin"});return await WebAssembly.instantiateStreaming(_,c)}catch(M){$(`wasm streaming compile failed: ${M}`),$("falling back to ArrayBuffer instantiation")}return async function(M,x){try{var A=await async function(L){if(!Ce)try{var V=await v(L);return new Uint8Array(V)}catch{}if(L==Ae&&Ce)L=new Uint8Array(Ce);else{if(!b)throw"both async and sync fetching of the wasm failed";L=b(L)}return L}(M);return await WebAssembly.instantiate(A,x)}catch(L){$(`failed to asynchronously prepare wasm: ${L}`),Ve(L)}}(p,c)}(a);return s(l.instance,l.module)}catch(c){return o(c),Promise.reject(c)}})();var Rn=s=>(Rn=te.rf)(s),Do=()=>(Do=te.sf)();i._OrtInit=(s,a)=>(i._OrtInit=te.tf)(s,a),i._OrtGetLastError=(s,a)=>(i._OrtGetLastError=te.uf)(s,a),i._OrtCreateSessionOptions=(s,a,l,c,p,_,M,x,A,L)=>(i._OrtCreateSessionOptions=te.vf)(s,a,l,c,p,_,M,x,A,L),i._OrtAppendExecutionProvider=(s,a,l,c,p)=>(i._OrtAppendExecutionProvider=te.wf)(s,a,l,c,p),i._OrtAddFreeDimensionOverride=(s,a,l)=>(i._OrtAddFreeDimensionOverride=te.xf)(s,a,l),i._OrtAddSessionConfigEntry=(s,a,l)=>(i._OrtAddSessionConfigEntry=te.yf)(s,a,l),i._OrtReleaseSessionOptions=s=>(i._OrtReleaseSessionOptions=te.zf)(s),i._OrtCreateSession=(s,a,l)=>(i._OrtCreateSession=te.Af)(s,a,l),i._OrtReleaseSession=s=>(i._OrtReleaseSession=te.Bf)(s),i._OrtGetInputOutputCount=(s,a,l)=>(i._OrtGetInputOutputCount=te.Cf)(s,a,l),i._OrtGetInputOutputMetadata=(s,a,l,c)=>(i._OrtGetInputOutputMetadata=te.Df)(s,a,l,c),i._OrtFree=s=>(i._OrtFree=te.Ef)(s),i._OrtCreateTensor=(s,a,l,c,p,_)=>(i._OrtCreateTensor=te.Ff)(s,a,l,c,p,_),i._OrtGetTensorData=(s,a,l,c,p)=>(i._OrtGetTensorData=te.Gf)(s,a,l,c,p),i._OrtReleaseTensor=s=>(i._OrtReleaseTensor=te.Hf)(s),i._OrtCreateRunOptions=(s,a,l,c)=>(i._OrtCreateRunOptions=te.If)(s,a,l,c),i._OrtAddRunConfigEntry=(s,a,l)=>(i._OrtAddRunConfigEntry=te.Jf)(s,a,l),i._OrtReleaseRunOptions=s=>(i._OrtReleaseRunOptions=te.Kf)(s),i._OrtCreateBinding=s=>(i._OrtCreateBinding=te.Lf)(s),i._OrtBindInput=(s,a,l)=>(i._OrtBindInput=te.Mf)(s,a,l),i._OrtBindOutput=(s,a,l,c)=>(i._OrtBindOutput=te.Nf)(s,a,l,c),i._OrtClearBoundOutputs=s=>(i._OrtClearBoundOutputs=te.Of)(s),i._OrtReleaseBinding=s=>(i._OrtReleaseBinding=te.Pf)(s),i._OrtRunWithBinding=(s,a,l,c,p)=>(i._OrtRunWithBinding=te.Qf)(s,a,l,c,p),i._OrtRun=(s,a,l,c,p,_,M,x)=>(i._OrtRun=te.Rf)(s,a,l,c,p,_,M,x),i._OrtEndProfiling=s=>(i._OrtEndProfiling=te.Sf)(s);var Nn=i._OrtGetWebGpuDevice=s=>(Nn=i._OrtGetWebGpuDevice=te.Tf)(s);i._JsepOutput=(s,a,l)=>(i._JsepOutput=te.Uf)(s,a,l),i._JsepGetNodeName=s=>(i._JsepGetNodeName=te.Vf)(s);var Ss=()=>(Ss=te.Wf)(),ci=i._free=s=>(ci=i._free=te.Xf)(s),ds=i._malloc=s=>(ds=i._malloc=te.Yf)(s),Vn=i._wgpuBufferRelease=s=>(Vn=i._wgpuBufferRelease=te.Zf)(s),Lo=i._wgpuCreateInstance=s=>(Lo=i._wgpuCreateInstance=te._f)(s),jo=s=>(jo=te.$f)(s),zo=s=>(zo=te.ag)(s),Bo=s=>(Bo=te.bg)(s),Un=s=>(Un=te.cg)(s),Ro=s=>(Ro=te.dg)(s),No=s=>(No=te.eg)(s),Vo=s=>(Vo=te.gg)(s),Uo=(s,a)=>(Uo=te.qg)(s,a),Wn=(s,a)=>(Wn=te.rg)(s,a),Wo=s=>(Wo=te.sg)(s),Gn=(s,a,l)=>(Gn=te.ug)(s,a,l),Kn=(s,a,l)=>(Kn=te.vg)(s,a,l),Hn=(s,a,l,c)=>(Hn=te.wg)(s,a,l,c),ps=(s,a,l,c)=>(ps=te.xg)(s,a,l,c),qn=(s,a,l,c)=>(qn=te.yg)(s,a,l,c),Go=(s,a,l)=>(Go=te.zg)(s,a,l),Qn=(s,a,l,c,p,_)=>(Qn=te.Cg)(s,a,l,c,p,_),Ko=()=>(Ko=te.Dg)(),Xn=(s,a,l,c,p)=>(Xn=te.Eg)(s,a,l,c,p),Ho=s=>(Ho=te.Fg)(s),Jn=s=>(Jn=te.Gg)(s),Yn=(s,a)=>(Yn=te.Hg)(s,a),qo=()=>(qo=te.Ig)(),Zn=(s,a)=>(Zn=te.Jg)(s,a),we=(s,a)=>(we=te.Kg)(s,a),$s=s=>($s=te.Lg)(s),Qo=(s,a)=>(Qo=te.Mg)(s,a),he=s=>(he=te.Ng)(s),Xs=s=>(Xs=te.Og)(s),fe=()=>(fe=te.Pg)(),ea=s=>(ea=te.Qg)(s),Xo=s=>(Xo=te.Rg)(s),Jo=(s,a,l)=>(Jo=te.Sg)(s,a,l),Js=s=>(Js=te.Tg)(s),Yo=i.dynCall_vii=(s,a,l)=>(Yo=i.dynCall_vii=te.Ug)(s,a,l),Zo=i.dynCall_iiii=(s,a,l,c)=>(Zo=i.dynCall_iiii=te.Vg)(s,a,l,c),ta=i.dynCall_iii=(s,a,l)=>(ta=i.dynCall_iii=te.Wg)(s,a,l),Ys=i.dynCall_ii=(s,a)=>(Ys=i.dynCall_ii=te.Xg)(s,a),el=i.dynCall_iiiiiii=(s,a,l,c,p,_,M)=>(el=i.dynCall_iiiiiii=te.Yg)(s,a,l,c,p,_,M),ra=i.dynCall_vi=(s,a)=>(ra=i.dynCall_vi=te.Zg)(s,a),tl=i.dynCall_v=s=>(tl=i.dynCall_v=te._g)(s),rl=i.dynCall_iiiiii=(s,a,l,c,p,_)=>(rl=i.dynCall_iiiiii=te.$g)(s,a,l,c,p,_),ks=i.dynCall_viii=(s,a,l,c)=>(ks=i.dynCall_viii=te.ah)(s,a,l,c),ia=i.dynCall_i=s=>(ia=i.dynCall_i=te.bh)(s),il=i.dynCall_iiiii=(s,a,l,c,p)=>(il=i.dynCall_iiiii=te.ch)(s,a,l,c,p),sl=i.dynCall_viiii=(s,a,l,c,p)=>(sl=i.dynCall_viiii=te.dh)(s,a,l,c,p),nl=i.dynCall_viiiii=(s,a,l,c,p,_)=>(nl=i.dynCall_viiiii=te.eh)(s,a,l,c,p,_),sa=i.dynCall_vijii=(s,a,l,c,p)=>(sa=i.dynCall_vijii=te.fh)(s,a,l,c,p),al=i.dynCall_vijjiii=(s,a,l,c,p,_,M)=>(al=i.dynCall_vijjiii=te.gh)(s,a,l,c,p,_,M),ol=i.dynCall_vijj=(s,a,l,c)=>(ol=i.dynCall_vijj=te.hh)(s,a,l,c),ll=i.dynCall_vijji=(s,a,l,c,p)=>(ll=i.dynCall_vijji=te.ih)(s,a,l,c,p),na=i.dynCall_diii=(s,a,l,c)=>(na=i.dynCall_diii=te.jh)(s,a,l,c),cl=i.dynCall_viiij=(s,a,l,c,p)=>(cl=i.dynCall_viiij=te.kh)(s,a,l,c,p),ul=i.dynCall_iiij=(s,a,l,c)=>(ul=i.dynCall_iiij=te.lh)(s,a,l,c),aa=i.dynCall_viijii=(s,a,l,c,p,_)=>(aa=i.dynCall_viijii=te.mh)(s,a,l,c,p,_),dl=i.dynCall_iif=(s,a,l)=>(dl=i.dynCall_iif=te.nh)(s,a,l),pl=i.dynCall_jiii=(s,a,l,c)=>(pl=i.dynCall_jiii=te.oh)(s,a,l,c),hl=i.dynCall_viijijj=(s,a,l,c,p,_,M)=>(hl=i.dynCall_viijijj=te.ph)(s,a,l,c,p,_,M),oa=i.dynCall_vij=(s,a,l)=>(oa=i.dynCall_vij=te.qh)(s,a,l),fl=i.dynCall_viij=(s,a,l,c)=>(fl=i.dynCall_viij=te.rh)(s,a,l,c),ml=i.dynCall_iiiiid=(s,a,l,c,p,_)=>(ml=i.dynCall_iiiiid=te.sh)(s,a,l,c,p,_),la=i.dynCall_viijj=(s,a,l,c,p)=>(la=i.dynCall_viijj=te.th)(s,a,l,c,p),_l=i.dynCall_iij=(s,a,l)=>(_l=i.dynCall_iij=te.uh)(s,a,l),gl=i.dynCall_iiiiij=(s,a,l,c,p,_)=>(gl=i.dynCall_iiiiij=te.vh)(s,a,l,c,p,_),ca=i.dynCall_j=s=>(ca=i.dynCall_j=te.wh)(s),yl=i.dynCall_viiiiiii=(s,a,l,c,p,_,M,x)=>(yl=i.dynCall_viiiiiii=te.xh)(s,a,l,c,p,_,M,x),wl=i.dynCall_iiiiiiiii=(s,a,l,c,p,_,M,x,A)=>(wl=i.dynCall_iiiiiiiii=te.yh)(s,a,l,c,p,_,M,x,A),Zs=i.dynCall_iiiiijiiiii=(s,a,l,c,p,_,M,x,A,L,V)=>(Zs=i.dynCall_iiiiijiiiii=te.zh)(s,a,l,c,p,_,M,x,A,L,V),Ml=i.dynCall_iiiiiiiiiii=(s,a,l,c,p,_,M,x,A,L,V)=>(Ml=i.dynCall_iiiiiiiiiii=te.Ah)(s,a,l,c,p,_,M,x,A,L,V),bl=i.dynCall_vijjjiiiiij=(s,a,l,c,p,_,M,x,A,L,V)=>(bl=i.dynCall_vijjjiiiiij=te.Bh)(s,a,l,c,p,_,M,x,A,L,V),vl=i.dynCall_viiiiii=(s,a,l,c,p,_,M)=>(vl=i.dynCall_viiiiii=te.Ch)(s,a,l,c,p,_,M),ua=i.dynCall_viji=(s,a,l,c)=>(ua=i.dynCall_viji=te.Dh)(s,a,l,c),xl=i.dynCall_viiiiiiiii=(s,a,l,c,p,_,M,x,A,L)=>(xl=i.dynCall_viiiiiiiii=te.Eh)(s,a,l,c,p,_,M,x,A,L),Tl=i.dynCall_fi=(s,a)=>(Tl=i.dynCall_fi=te.Fh)(s,a),da=i.dynCall_fii=(s,a,l)=>(da=i.dynCall_fii=te.Gh)(s,a,l),El=i.dynCall_ji=(s,a)=>(El=i.dynCall_ji=te.Hh)(s,a),Pl=i.dynCall_di=(s,a)=>(Pl=i.dynCall_di=te.Ih)(s,a),Cl=i.dynCall_jii=(s,a,l)=>(Cl=i.dynCall_jii=te.Jh)(s,a,l),Sl=i.dynCall_dii=(s,a,l)=>(Sl=i.dynCall_dii=te.Kh)(s,a,l),en=i.dynCall_viijiii=(s,a,l,c,p,_,M)=>(en=i.dynCall_viijiii=te.Lh)(s,a,l,c,p,_,M),$l=i.dynCall_viiiiiiii=(s,a,l,c,p,_,M,x,A)=>($l=i.dynCall_viiiiiiii=te.Mh)(s,a,l,c,p,_,M,x,A),kl=i.dynCall_iiiiiiiij=(s,a,l,c,p,_,M,x,A)=>(kl=i.dynCall_iiiiiiiij=te.Nh)(s,a,l,c,p,_,M,x,A),Il=i.dynCall_iiiiiiii=(s,a,l,c,p,_,M,x)=>(Il=i.dynCall_iiiiiiii=te.Oh)(s,a,l,c,p,_,M,x),Al=i.dynCall_viiji=(s,a,l,c,p)=>(Al=i.dynCall_viiji=te.Ph)(s,a,l,c,p),Ol=i.dynCall_viiiiij=(s,a,l,c,p,_,M)=>(Ol=i.dynCall_viiiiij=te.Qh)(s,a,l,c,p,_,M),Fl=i.dynCall_viijjjiiiiii=(s,a,l,c,p,_,M,x,A,L,V,W)=>(Fl=i.dynCall_viijjjiiiiii=te.Rh)(s,a,l,c,p,_,M,x,A,L,V,W),Dl=i.dynCall_viiijiiiiiii=(s,a,l,c,p,_,M,x,A,L,V,W)=>(Dl=i.dynCall_viiijiiiiiii=te.Sh)(s,a,l,c,p,_,M,x,A,L,V,W),Ll=i.dynCall_iiiiiiiiiiiii=(s,a,l,c,p,_,M,x,A,L,V,W,Z)=>(Ll=i.dynCall_iiiiiiiiiiiii=te.Th)(s,a,l,c,p,_,M,x,A,L,V,W,Z),jl=i.dynCall_viiiiiiiiii=(s,a,l,c,p,_,M,x,A,L,V)=>(jl=i.dynCall_viiiiiiiiii=te.Uh)(s,a,l,c,p,_,M,x,A,L,V),pa=i.dynCall_viiiiiiiiiiiiiiii=(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe)=>(pa=i.dynCall_viiiiiiiiiiiiiiii=te.Vh)(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe),zl=i.dynCall_viid=(s,a,l,c)=>(zl=i.dynCall_viid=te.Wh)(s,a,l,c),Bl=i.dynCall_vid=(s,a,l)=>(Bl=i.dynCall_vid=te.Xh)(s,a,l),ha=i.dynCall_viiiiiiiiiii=(s,a,l,c,p,_,M,x,A,L,V,W)=>(ha=i.dynCall_viiiiiiiiiii=te.Yh)(s,a,l,c,p,_,M,x,A,L,V,W),Rl=i.dynCall_viiijjjii=(s,a,l,c,p,_,M,x,A)=>(Rl=i.dynCall_viiijjjii=te.Zh)(s,a,l,c,p,_,M,x,A),Nl=i.dynCall_iid=(s,a,l)=>(Nl=i.dynCall_iid=te._h)(s,a,l),Vl=i.dynCall_viiiij=(s,a,l,c,p,_)=>(Vl=i.dynCall_viiiij=te.$h)(s,a,l,c,p,_),Ul=i.dynCall_viiijiiiii=(s,a,l,c,p,_,M,x,A,L)=>(Ul=i.dynCall_viiijiiiii=te.ai)(s,a,l,c,p,_,M,x,A,L),fa=i.dynCall_jj=(s,a)=>(fa=i.dynCall_jj=te.bi)(s,a),Wl=i.dynCall_iiiijii=(s,a,l,c,p,_,M)=>(Wl=i.dynCall_iiiijii=te.ci)(s,a,l,c,p,_,M),Gl=i.dynCall_iiijii=(s,a,l,c,p,_)=>(Gl=i.dynCall_iiijii=te.di)(s,a,l,c,p,_),ma=i.dynCall_viiiiiiiiiiiiiii=(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue)=>(ma=i.dynCall_viiiiiiiiiiiiiii=te.ei)(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue),Kl=i.dynCall_iiijjj=(s,a,l,c,p,_)=>(Kl=i.dynCall_iiijjj=te.fi)(s,a,l,c,p,_),Hl=i.dynCall_ij=(s,a)=>(Hl=i.dynCall_ij=te.gi)(s,a),_a=i.dynCall_viiiiji=(s,a,l,c,p,_,M)=>(_a=i.dynCall_viiiiji=te.hi)(s,a,l,c,p,_,M),ql=i.dynCall_iijjji=(s,a,l,c,p,_)=>(ql=i.dynCall_iijjji=te.ii)(s,a,l,c,p,_),Ql=i.dynCall_vjiiiiii=(s,a,l,c,p,_,M,x)=>(Ql=i.dynCall_vjiiiiii=te.ji)(s,a,l,c,p,_,M,x),ga=i.dynCall_vijjiiiii=(s,a,l,c,p,_,M,x,A)=>(ga=i.dynCall_vijjiiiii=te.ki)(s,a,l,c,p,_,M,x,A),Xl=i.dynCall_jiij=(s,a,l,c)=>(Xl=i.dynCall_jiij=te.li)(s,a,l,c),Jl=i.dynCall_iijijjijiji=(s,a,l,c,p,_,M,x,A,L,V)=>(Jl=i.dynCall_iijijjijiji=te.mi)(s,a,l,c,p,_,M,x,A,L,V),Yl=i.dynCall_iijijji=(s,a,l,c,p,_,M)=>(Yl=i.dynCall_iijijji=te.ni)(s,a,l,c,p,_,M),Zl=i.dynCall_ijijji=(s,a,l,c,p,_)=>(Zl=i.dynCall_ijijji=te.oi)(s,a,l,c,p,_),ya=i.dynCall_iiiiiiij=(s,a,l,c,p,_,M,x)=>(ya=i.dynCall_iiiiiiij=te.pi)(s,a,l,c,p,_,M,x),ec=i.dynCall_viiijjiii=(s,a,l,c,p,_,M,x,A)=>(ec=i.dynCall_viiijjiii=te.qi)(s,a,l,c,p,_,M,x,A),tc=i.dynCall_vif=(s,a,l)=>(tc=i.dynCall_vif=te.ri)(s,a,l),rc=i.dynCall_viif=(s,a,l,c)=>(rc=i.dynCall_viif=te.si)(s,a,l,c),ic=i.dynCall_iiiiijji=(s,a,l,c,p,_,M,x)=>(ic=i.dynCall_iiiiijji=te.ti)(s,a,l,c,p,_,M,x),sc=i.dynCall_iiiiji=(s,a,l,c,p,_)=>(sc=i.dynCall_iiiiji=te.ui)(s,a,l,c,p,_),nc=i.dynCall_iiiifi=(s,a,l,c,p,_)=>(nc=i.dynCall_iiiifi=te.vi)(s,a,l,c,p,_),wa=i.dynCall_iiiiiiiiijii=(s,a,l,c,p,_,M,x,A,L,V,W)=>(wa=i.dynCall_iiiiiiiiijii=te.wi)(s,a,l,c,p,_,M,x,A,L,V,W),ac=i.dynCall_iiiijjii=(s,a,l,c,p,_,M,x)=>(ac=i.dynCall_iiiijjii=te.xi)(s,a,l,c,p,_,M,x),oc=i.dynCall_iiiiiijjjii=(s,a,l,c,p,_,M,x,A,L,V)=>(oc=i.dynCall_iiiiiijjjii=te.yi)(s,a,l,c,p,_,M,x,A,L,V),Ma=i.dynCall_iiijiii=(s,a,l,c,p,_,M)=>(Ma=i.dynCall_iiijiii=te.zi)(s,a,l,c,p,_,M),lc=i.dynCall_iiiiiiiijjjfi=(s,a,l,c,p,_,M,x,A,L,V,W,Z)=>(lc=i.dynCall_iiiiiiiijjjfi=te.Ai)(s,a,l,c,p,_,M,x,A,L,V,W,Z),cc=i.dynCall_iijiiii=(s,a,l,c,p,_,M)=>(cc=i.dynCall_iijiiii=te.Bi)(s,a,l,c,p,_,M),Ni=i.dynCall_viiiijj=(s,a,l,c,p,_,M)=>(Ni=i.dynCall_viiiijj=te.Ci)(s,a,l,c,p,_,M),uc=i.dynCall_iijjjii=(s,a,l,c,p,_,M)=>(uc=i.dynCall_iijjjii=te.Di)(s,a,l,c,p,_,M),dc=i.dynCall_jij=(s,a,l)=>(dc=i.dynCall_jij=te.Ei)(s,a,l),pc=i.dynCall_jjj=(s,a,l)=>(pc=i.dynCall_jjj=te.Fi)(s,a,l),hc=i.dynCall_iiji=(s,a,l,c)=>(hc=i.dynCall_iiji=te.Gi)(s,a,l,c),ba=i.dynCall_viffiii=(s,a,l,c,p,_,M)=>(ba=i.dynCall_viffiii=te.Hi)(s,a,l,c,p,_,M),fc=i.dynCall_viifiii=(s,a,l,c,p,_,M)=>(fc=i.dynCall_viifiii=te.Ii)(s,a,l,c,p,_,M),mc=i.dynCall_viiiiidiidi=(s,a,l,c,p,_,M,x,A,L,V)=>(mc=i.dynCall_viiiiidiidi=te.Ji)(s,a,l,c,p,_,M,x,A,L,V),_c=i.dynCall_viiiiiiiiidi=(s,a,l,c,p,_,M,x,A,L,V,W)=>(_c=i.dynCall_viiiiiiiiidi=te.Ki)(s,a,l,c,p,_,M,x,A,L,V,W),gc=i.dynCall_viiiiiiiiiiiiiifi=(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe)=>(gc=i.dynCall_viiiiiiiiiiiiiifi=te.Li)(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe),tn=i.dynCall_ijii=(s,a,l,c)=>(tn=i.dynCall_ijii=te.Mi)(s,a,l,c),yc=i.dynCall_viijiiiijiii=(s,a,l,c,p,_,M,x,A,L,V,W)=>(yc=i.dynCall_viijiiiijiii=te.Ni)(s,a,l,c,p,_,M,x,A,L,V,W),wc=i.dynCall_vijjjjjjjjjjjjji=(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue)=>(wc=i.dynCall_vijjjjjjjjjjjjji=te.Oi)(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue),Mc=i.dynCall_viiijii=(s,a,l,c,p,_,M)=>(Mc=i.dynCall_viiijii=te.Pi)(s,a,l,c,p,_,M),Is=i.dynCall_vijjjiiji=(s,a,l,c,p,_,M,x,A)=>(Is=i.dynCall_vijjjiiji=te.Qi)(s,a,l,c,p,_,M,x,A),bc=i.dynCall_iiiijiiiiiiiiii=(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe)=>(bc=i.dynCall_iiiijiiiiiiiiii=te.Ri)(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe),vc=i.dynCall_iiiiiiiiii=(s,a,l,c,p,_,M,x,A,L)=>(vc=i.dynCall_iiiiiiiiii=te.Si)(s,a,l,c,p,_,M,x,A,L),xc=i.dynCall_vj=(s,a)=>(xc=i.dynCall_vj=te.Ti)(s,a),Tc=i.dynCall_vfiii=(s,a,l,c,p)=>(Tc=i.dynCall_vfiii=te.Ui)(s,a,l,c,p),rn=i.dynCall_viiiiff=(s,a,l,c,p,_,M)=>(rn=i.dynCall_viiiiff=te.Vi)(s,a,l,c,p,_,M),Ec=i.dynCall_viiiiiff=(s,a,l,c,p,_,M,x)=>(Ec=i.dynCall_viiiiiff=te.Wi)(s,a,l,c,p,_,M,x),Pc=i.dynCall_viiff=(s,a,l,c,p)=>(Pc=i.dynCall_viiff=te.Xi)(s,a,l,c,p),Cc=i.dynCall_viiiiiiiiifiiii=(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe)=>(Cc=i.dynCall_viiiiiiiiifiiii=te.Yi)(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe),Tp=i.dynCall_viiiiiiiijj=(s,a,l,c,p,_,M,x,A,L,V)=>(Tp=i.dynCall_viiiiiiiijj=te.Zi)(s,a,l,c,p,_,M,x,A,L,V),Sc=i.dynCall_iiiiiiiiiiiiiifii=(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe)=>(Sc=i.dynCall_iiiiiiiiiiiiiifii=te._i)(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe),$c=i.dynCall_viiiiiiiiiiiii=(s,a,l,c,p,_,M,x,A,L,V,W,Z,de)=>($c=i.dynCall_viiiiiiiiiiiii=te.$i)(s,a,l,c,p,_,M,x,A,L,V,W,Z,de),kc=i.dynCall_iiiiiiiiiiiiiiiiiiifii=(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe,ht,Pt,Rt,Xt,mr)=>(kc=i.dynCall_iiiiiiiiiiiiiiiiiiifii=te.aj)(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe,ht,Pt,Rt,Xt,mr),hs=i.dynCall_vijjiiiiiii=(s,a,l,c,p,_,M,x,A,L,V)=>(hs=i.dynCall_vijjiiiiiii=te.bj)(s,a,l,c,p,_,M,x,A,L,V),Ic=i.dynCall_iiiijjj=(s,a,l,c,p,_,M)=>(Ic=i.dynCall_iiiijjj=te.cj)(s,a,l,c,p,_,M),Ac=i.dynCall_fffffff=(s,a,l,c,p,_,M)=>(Ac=i.dynCall_fffffff=te.dj)(s,a,l,c,p,_,M),Oc=i.dynCall_viiiiiijiifiii=(s,a,l,c,p,_,M,x,A,L,V,W,Z,de)=>(Oc=i.dynCall_viiiiiijiifiii=te.ej)(s,a,l,c,p,_,M,x,A,L,V,W,Z,de),Fc=i.dynCall_vjjjjjjffjifiiiiii=(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe,ht)=>(Fc=i.dynCall_vjjjjjjffjifiiiiii=te.fj)(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe,ht),Dc=i.dynCall_viiiiiiffjifiiiii=(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe)=>(Dc=i.dynCall_viiiiiiffjifiiiii=te.gj)(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe),Lc=i.dynCall_viiiiiiffjfiiiii=(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue)=>(Lc=i.dynCall_viiiiiiffjfiiiii=te.hj)(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue),jc=i.dynCall_viiiiiiffjiiiii=(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe)=>(jc=i.dynCall_viiiiiiffjiiiii=te.ij)(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe),sn=i.dynCall_vjjjjjjjjfffjifiiiiii=(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe,ht,Pt,Rt,Xt)=>(sn=i.dynCall_vjjjjjjjjfffjifiiiiii=te.jj)(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe,ht,Pt,Rt,Xt),Ep=i.dynCall_vjjjjjjfffifijiiiii=(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe,ht,Pt)=>(Ep=i.dynCall_vjjjjjjfffifijiiiii=te.kj)(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe,ht,Pt),zc=i.dynCall_vjjjjjjfffifiiiiii=(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe,ht)=>(zc=i.dynCall_vjjjjjjfffifiiiiii=te.lj)(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe,ht),Bc=i.dynCall_vjjjjjjjjfffiiifiiiii=(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe,ht,Pt,Rt,Xt)=>(Bc=i.dynCall_vjjjjjjjjfffiiifiiiii=te.mj)(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe,ht,Pt,Rt,Xt),Rc=i.dynCall_vijiiiiiiiiii=(s,a,l,c,p,_,M,x,A,L,V,W,Z)=>(Rc=i.dynCall_vijiiiiiiiiii=te.nj)(s,a,l,c,p,_,M,x,A,L,V,W,Z),Nc=i.dynCall_vijjfffiii=(s,a,l,c,p,_,M,x,A,L)=>(Nc=i.dynCall_vijjfffiii=te.oj)(s,a,l,c,p,_,M,x,A,L),Vc=i.dynCall_jiijjiif=(s,a,l,c,p,_,M,x)=>(Vc=i.dynCall_jiijjiif=te.pj)(s,a,l,c,p,_,M,x),va=i.dynCall_vijjjjjjifiiiii=(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe)=>(va=i.dynCall_vijjjjjjifiiiii=te.qj)(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe),Uc=i.dynCall_vjjjjjiiii=(s,a,l,c,p,_,M,x,A,L)=>(Uc=i.dynCall_vjjjjjiiii=te.rj)(s,a,l,c,p,_,M,x,A,L),Wc=i.dynCall_vjjjjfiii=(s,a,l,c,p,_,M,x,A)=>(Wc=i.dynCall_vjjjjfiii=te.sj)(s,a,l,c,p,_,M,x,A),xa=i.dynCall_viiiiiijiiiiii=(s,a,l,c,p,_,M,x,A,L,V,W,Z,de)=>(xa=i.dynCall_viiiiiijiiiiii=te.tj)(s,a,l,c,p,_,M,x,A,L,V,W,Z,de),Gc=i.dynCall_vijjii=(s,a,l,c,p,_)=>(Gc=i.dynCall_vijjii=te.uj)(s,a,l,c,p,_),Kc=i.dynCall_viiiiijjiiiii=(s,a,l,c,p,_,M,x,A,L,V,W,Z)=>(Kc=i.dynCall_viiiiijjiiiii=te.vj)(s,a,l,c,p,_,M,x,A,L,V,W,Z),Ta=i.dynCall_iiiiiji=(s,a,l,c,p,_,M)=>(Ta=i.dynCall_iiiiiji=te.wj)(s,a,l,c,p,_,M),Hc=i.dynCall_viiiiijiiiiii=(s,a,l,c,p,_,M,x,A,L,V,W,Z)=>(Hc=i.dynCall_viiiiijiiiiii=te.xj)(s,a,l,c,p,_,M,x,A,L,V,W,Z),qc=i.dynCall_iiiiiiiiiiii=(s,a,l,c,p,_,M,x,A,L,V,W)=>(qc=i.dynCall_iiiiiiiiiiii=te.yj)(s,a,l,c,p,_,M,x,A,L,V,W),nn=i.dynCall_viiijiiiiii=(s,a,l,c,p,_,M,x,A,L,V)=>(nn=i.dynCall_viiijiiiiii=te.zj)(s,a,l,c,p,_,M,x,A,L,V),Qc=i.dynCall_viiiijii=(s,a,l,c,p,_,M,x)=>(Qc=i.dynCall_viiiijii=te.Aj)(s,a,l,c,p,_,M,x),Xc=i.dynCall_viijjiii=(s,a,l,c,p,_,M,x)=>(Xc=i.dynCall_viijjiii=te.Bj)(s,a,l,c,p,_,M,x),Jc=i.dynCall_viiiji=(s,a,l,c,p,_)=>(Jc=i.dynCall_viiiji=te.Cj)(s,a,l,c,p,_),Yc=i.dynCall_viiiiiijii=(s,a,l,c,p,_,M,x,A,L)=>(Yc=i.dynCall_viiiiiijii=te.Dj)(s,a,l,c,p,_,M,x,A,L),Ea=i.dynCall_viiiiijjji=(s,a,l,c,p,_,M,x,A,L)=>(Ea=i.dynCall_viiiiijjji=te.Ej)(s,a,l,c,p,_,M,x,A,L),an=i.dynCall_vijiii=(s,a,l,c,p,_)=>(an=i.dynCall_vijiii=te.Fj)(s,a,l,c,p,_),Pp=i.dynCall_iiijiiii=(s,a,l,c,p,_,M,x)=>(Pp=i.dynCall_iiijiiii=te.Gj)(s,a,l,c,p,_,M,x),Zc=i.dynCall_viiiiiijjiiiii=(s,a,l,c,p,_,M,x,A,L,V,W,Z,de)=>(Zc=i.dynCall_viiiiiijjiiiii=te.Hj)(s,a,l,c,p,_,M,x,A,L,V,W,Z,de),eu=i.dynCall_viiiiiiijiiiiii=(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe)=>(eu=i.dynCall_viiiiiiijiiiiii=te.Ij)(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe),Pa=i.dynCall_viiiiiji=(s,a,l,c,p,_,M,x)=>(Pa=i.dynCall_viiiiiji=te.Jj)(s,a,l,c,p,_,M,x),tu=i.dynCall_fiif=(s,a,l,c)=>(tu=i.dynCall_fiif=te.Kj)(s,a,l,c),ru=i.dynCall_viijjjiii=(s,a,l,c,p,_,M,x,A)=>(ru=i.dynCall_viijjjiii=te.Lj)(s,a,l,c,p,_,M,x,A),Ca=i.dynCall_viiiiiifiii=(s,a,l,c,p,_,M,x,A,L,V)=>(Ca=i.dynCall_viiiiiifiii=te.Mj)(s,a,l,c,p,_,M,x,A,L,V),iu=i.dynCall_viijji=(s,a,l,c,p,_)=>(iu=i.dynCall_viijji=te.Nj)(s,a,l,c,p,_),su=i.dynCall_iiiiiiiiiiijijji=(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue)=>(su=i.dynCall_iiiiiiiiiiijijji=te.Oj)(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue),Sa=i.dynCall_jiijjiii=(s,a,l,c,p,_,M,x)=>(Sa=i.dynCall_jiijjiii=te.Pj)(s,a,l,c,p,_,M,x),Cp=i.dynCall_viifiifijjjii=(s,a,l,c,p,_,M,x,A,L,V,W,Z)=>(Cp=i.dynCall_viifiifijjjii=te.Qj)(s,a,l,c,p,_,M,x,A,L,V,W,Z),Sp=i.dynCall_viiiiiiiiiiiiiiiiiiiiiii=(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe,ht,Pt,Rt,Xt,mr,Lr,Vi)=>(Sp=i.dynCall_viiiiiiiiiiiiiiiiiiiiiii=te.Rj)(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe,ht,Pt,Rt,Xt,mr,Lr,Vi),$a=i.dynCall_viiiiifiiiiii=(s,a,l,c,p,_,M,x,A,L,V,W,Z)=>($a=i.dynCall_viiiiifiiiiii=te.Sj)(s,a,l,c,p,_,M,x,A,L,V,W,Z),on=i.dynCall_vijjiiiiii=(s,a,l,c,p,_,M,x,A,L)=>(on=i.dynCall_vijjiiiiii=te.Tj)(s,a,l,c,p,_,M,x,A,L),nu=i.dynCall_vijiiiiiiijjii=(s,a,l,c,p,_,M,x,A,L,V,W,Z,de)=>(nu=i.dynCall_vijiiiiiiijjii=te.Uj)(s,a,l,c,p,_,M,x,A,L,V,W,Z,de),au=i.dynCall_viiiiiiiiiiii=(s,a,l,c,p,_,M,x,A,L,V,W,Z)=>(au=i.dynCall_viiiiiiiiiiii=te.Vj)(s,a,l,c,p,_,M,x,A,L,V,W,Z),ou=i.dynCall_viiiiiiiiiiiiiiiiii=(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe,ht,Pt)=>(ou=i.dynCall_viiiiiiiiiiiiiiiiii=te.Wj)(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe,ht,Pt),ln=i.dynCall_viiiiiiiiiiiiiiiiiii=(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe,ht,Pt,Rt)=>(ln=i.dynCall_viiiiiiiiiiiiiiiiiii=te.Xj)(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe,ht,Pt,Rt),lu=i.dynCall_viiijiiiiiiiiiiiiiii=(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe,ht,Pt,Rt)=>(lu=i.dynCall_viiijiiiiiiiiiiiiiii=te.Yj)(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe,ht,Pt,Rt),cu=i.dynCall_viiiijjj=(s,a,l,c,p,_,M,x)=>(cu=i.dynCall_viiiijjj=te.Zj)(s,a,l,c,p,_,M,x),uu=i.dynCall_viiiiiiijjj=(s,a,l,c,p,_,M,x,A,L,V)=>(uu=i.dynCall_viiiiiiijjj=te._j)(s,a,l,c,p,_,M,x,A,L,V),cn=i.dynCall_iiiiiiiiiiiiiiiiiiiifi=(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe,ht,Pt,Rt,Xt,mr)=>(cn=i.dynCall_iiiiiiiiiiiiiiiiiiiifi=te.$j)(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe,ht,Pt,Rt,Xt,mr),du=i.dynCall_viiijiiiiiiiiii=(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe)=>(du=i.dynCall_viiijiiiiiiiiii=te.ak)(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe),pu=i.dynCall_viiiiif=(s,a,l,c,p,_,M)=>(pu=i.dynCall_viiiiif=te.bk)(s,a,l,c,p,_,M),hu=i.dynCall_viiif=(s,a,l,c,p)=>(hu=i.dynCall_viiif=te.ck)(s,a,l,c,p),un=i.dynCall_viiiiiiiiifi=(s,a,l,c,p,_,M,x,A,L,V,W)=>(un=i.dynCall_viiiiiiiiifi=te.dk)(s,a,l,c,p,_,M,x,A,L,V,W),fu=i.dynCall_viiiiid=(s,a,l,c,p,_,M)=>(fu=i.dynCall_viiiiid=te.ek)(s,a,l,c,p,_,M),mu=i.dynCall_viiid=(s,a,l,c,p)=>(mu=i.dynCall_viiid=te.fk)(s,a,l,c,p),_u=i.dynCall_iiif=(s,a,l,c)=>(_u=i.dynCall_iiif=te.gk)(s,a,l,c),gu=i.dynCall_iiiij=(s,a,l,c,p)=>(gu=i.dynCall_iiiij=te.hk)(s,a,l,c,p),yu=i.dynCall_iiiiiif=(s,a,l,c,p,_,M)=>(yu=i.dynCall_iiiiiif=te.ik)(s,a,l,c,p,_,M),wu=i.dynCall_vijij=(s,a,l,c,p)=>(wu=i.dynCall_vijij=te.jk)(s,a,l,c,p),Mu=i.dynCall_viiijji=(s,a,l,c,p,_,M)=>(Mu=i.dynCall_viiijji=te.kk)(s,a,l,c,p,_,M),bu=i.dynCall_viiiiiiffii=(s,a,l,c,p,_,M,x,A,L,V)=>(bu=i.dynCall_viiiiiiffii=te.lk)(s,a,l,c,p,_,M,x,A,L,V),vu=i.dynCall_iifii=(s,a,l,c,p)=>(vu=i.dynCall_iifii=te.mk)(s,a,l,c,p),xu=i.dynCall_viiiiifi=(s,a,l,c,p,_,M,x)=>(xu=i.dynCall_viiiiifi=te.nk)(s,a,l,c,p,_,M,x),ka=i.dynCall_iifjii=(s,a,l,c,p,_)=>(ka=i.dynCall_iifjii=te.ok)(s,a,l,c,p,_),Tu=i.dynCall_vidi=(s,a,l,c)=>(Tu=i.dynCall_vidi=te.pk)(s,a,l,c),Eu=i.dynCall_viiijiji=(s,a,l,c,p,_,M,x)=>(Eu=i.dynCall_viiijiji=te.qk)(s,a,l,c,p,_,M,x),Ia=i.dynCall_viiijij=(s,a,l,c,p,_,M)=>(Ia=i.dynCall_viiijij=te.rk)(s,a,l,c,p,_,M),Pu=i.dynCall_vijjj=(s,a,l,c,p)=>(Pu=i.dynCall_vijjj=te.sk)(s,a,l,c,p),Cu=i.dynCall_vjiij=(s,a,l,c,p)=>(Cu=i.dynCall_vjiij=te.tk)(s,a,l,c,p),Su=i.dynCall_diiiii=(s,a,l,c,p,_)=>(Su=i.dynCall_diiiii=te.uk)(s,a,l,c,p,_),$u=i.dynCall_diiii=(s,a,l,c,p)=>($u=i.dynCall_diiii=te.vk)(s,a,l,c,p),dn=i.dynCall_iiiji=(s,a,l,c,p)=>(dn=i.dynCall_iiiji=te.wk)(s,a,l,c,p),ku=i.dynCall_ijiijji=(s,a,l,c,p,_,M)=>(ku=i.dynCall_ijiijji=te.xk)(s,a,l,c,p,_,M),Iu=i.dynCall_viiijjiiiiii=(s,a,l,c,p,_,M,x,A,L,V,W)=>(Iu=i.dynCall_viiijjiiiiii=te.yk)(s,a,l,c,p,_,M,x,A,L,V,W),Au=i.dynCall_viijjijjjjiii=(s,a,l,c,p,_,M,x,A,L,V,W,Z)=>(Au=i.dynCall_viijjijjjjiii=te.zk)(s,a,l,c,p,_,M,x,A,L,V,W,Z),Ou=i.dynCall_ijiii=(s,a,l,c,p)=>(Ou=i.dynCall_ijiii=te.Ak)(s,a,l,c,p),Fu=i.dynCall_ijiiiiji=(s,a,l,c,p,_,M,x)=>(Fu=i.dynCall_ijiiiiji=te.Bk)(s,a,l,c,p,_,M,x),pn=i.dynCall_iiifi=(s,a,l,c,p)=>(pn=i.dynCall_iiifi=te.Ck)(s,a,l,c,p),Du=i.dynCall_ijiij=(s,a,l,c,p)=>(Du=i.dynCall_ijiij=te.Dk)(s,a,l,c,p),Lu=i.dynCall_viijiiiiiiiiii=(s,a,l,c,p,_,M,x,A,L,V,W,Z,de)=>(Lu=i.dynCall_viijiiiiiiiiii=te.Ek)(s,a,l,c,p,_,M,x,A,L,V,W,Z,de),ju=i.dynCall_fiiii=(s,a,l,c,p)=>(ju=i.dynCall_fiiii=te.Fk)(s,a,l,c,p),zu=i.dynCall_jfi=(s,a,l)=>(zu=i.dynCall_jfi=te.Gk)(s,a,l),Bu=i.dynCall_viiiiiiiiiiiiii=(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe)=>(Bu=i.dynCall_viiiiiiiiiiiiii=te.Hk)(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe),hn=i.dynCall_viiiiiiiiiiiiiiiiiiii=(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe,ht,Pt,Rt,Xt)=>(hn=i.dynCall_viiiiiiiiiiiiiiiiiiii=te.Ik)(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe,ht,Pt,Rt,Xt),Ru=i.dynCall_fiii=(s,a,l,c)=>(Ru=i.dynCall_fiii=te.Jk)(s,a,l,c),Nu=s=>(Nu=te.Kk)(s),Vu=()=>(Vu=te.Lk)(),Wt=s=>(Wt=te.Mk)(s),$p=()=>($p=te.Nk)();function ff(s,a,l,c){var p=fe();try{return Zo(s,a,l,c)}catch(_){if(he(p),_!==_+0)throw _;we(1,0)}}function mf(s,a,l){var c=fe();try{return ta(s,a,l)}catch(p){if(he(c),p!==p+0)throw p;we(1,0)}}function _f(s,a,l){var c=fe();try{Yo(s,a,l)}catch(p){if(he(c),p!==p+0)throw p;we(1,0)}}function Uu(s,a){var l=fe();try{return Ys(s,a)}catch(c){if(he(l),c!==c+0)throw c;we(1,0)}}function kp(s,a){var l=fe();try{ra(s,a)}catch(c){if(he(l),c!==c+0)throw c;we(1,0)}}function Ip(s){var a=fe();try{tl(s)}catch(l){if(he(a),l!==l+0)throw l;we(1,0)}}function Ap(s,a,l,c,p,_,M){var x=fe();try{return el(s,a,l,c,p,_,M)}catch(A){if(he(x),A!==A+0)throw A;we(1,0)}}function Op(s,a,l,c,p){var _=fe();try{sa(s,a,l,c,p)}catch(M){if(he(_),M!==M+0)throw M;we(1,0)}}function Wu(s,a,l,c,p,_){var M=fe();try{return rl(s,a,l,c,p,_)}catch(x){if(he(M),x!==x+0)throw x;we(1,0)}}function Gu(s,a,l,c){var p=fe();try{ks(s,a,l,c)}catch(_){if(he(p),_!==_+0)throw _;we(1,0)}}function gf(s,a,l,c,p,_,M){var x=fe();try{al(s,a,l,c,p,_,M)}catch(A){if(he(x),A!==A+0)throw A;we(1,0)}}function Fp(s,a,l,c,p){var _=fe();try{sl(s,a,l,c,p)}catch(M){if(he(_),M!==M+0)throw M;we(1,0)}}function Dp(s,a,l,c,p){var _=fe();try{return il(s,a,l,c,p)}catch(M){if(he(_),M!==M+0)throw M;we(1,0)}}function Ku(s,a,l,c,p,_,M){var x=fe();try{vl(s,a,l,c,p,_,M)}catch(A){if(he(x),A!==A+0)throw A;we(1,0)}}function Lp(s,a,l,c,p,_){var M=fe();try{nl(s,a,l,c,p,_)}catch(x){if(he(M),x!==x+0)throw x;we(1,0)}}function jp(s){var a=fe();try{return ia(s)}catch(l){if(he(a),l!==l+0)throw l;we(1,0)}}function yf(s,a,l,c){var p=fe();try{return ul(s,a,l,c)}catch(_){if(he(p),_!==_+0)throw _;we(1,0)}}function zp(s,a,l,c,p,_){var M=fe();try{aa(s,a,l,c,p,_)}catch(x){if(he(M),x!==x+0)throw x;we(1,0)}}function Bp(s,a,l,c){var p=fe();try{return pl(s,a,l,c)}catch(_){if(he(p),_!==_+0)throw _;return we(1,0),0n}}function Rp(s,a,l,c,p,_,M){var x=fe();try{hl(s,a,l,c,p,_,M)}catch(A){if(he(x),A!==A+0)throw A;we(1,0)}}function As(s,a,l){var c=fe();try{oa(s,a,l)}catch(p){if(he(c),p!==p+0)throw p;we(1,0)}}function Np(s,a,l,c){var p=fe();try{fl(s,a,l,c)}catch(_){if(he(p),_!==_+0)throw _;we(1,0)}}function Vp(s,a,l){var c=fe();try{return dl(s,a,l)}catch(p){if(he(c),p!==p+0)throw p;we(1,0)}}function Up(s,a,l){var c=fe();try{return _l(s,a,l)}catch(p){if(he(c),p!==p+0)throw p;we(1,0)}}function Wp(s,a,l,c,p){var _=fe();try{la(s,a,l,c,p)}catch(M){if(he(_),M!==M+0)throw M;we(1,0)}}function Gp(s,a,l,c,p,_,M,x){var A=fe();try{yl(s,a,l,c,p,_,M,x)}catch(L){if(he(A),L!==L+0)throw L;we(1,0)}}function Kp(s,a,l,c,p,_,M,x,A,L,V){var W=fe();try{return Zs(s,a,l,c,p,_,M,x,A,L,V)}catch(Z){if(he(W),Z!==Z+0)throw Z;we(1,0)}}function Hp(s,a,l,c,p,_,M,x,A,L,V){var W=fe();try{return Ml(s,a,l,c,p,_,M,x,A,L,V)}catch(Z){if(he(W),Z!==Z+0)throw Z;we(1,0)}}function qp(s,a,l,c,p,_,M,x,A,L,V){var W=fe();try{bl(s,a,l,c,p,_,M,x,A,L,V)}catch(Z){if(he(W),Z!==Z+0)throw Z;we(1,0)}}function wf(s,a,l,c,p,_,M,x,A,L){var V=fe();try{xl(s,a,l,c,p,_,M,x,A,L)}catch(W){if(he(V),W!==W+0)throw W;we(1,0)}}function Qp(s,a,l,c,p,_,M,x,A){var L=fe();try{return wl(s,a,l,c,p,_,M,x,A)}catch(V){if(he(L),V!==V+0)throw V;we(1,0)}}function Xp(s,a){var l=fe();try{return Tl(s,a)}catch(c){if(he(l),c!==c+0)throw c;we(1,0)}}function Jp(s,a){var l=fe();try{return El(s,a)}catch(c){if(he(l),c!==c+0)throw c;return we(1,0),0n}}function Yp(s,a){var l=fe();try{return Pl(s,a)}catch(c){if(he(l),c!==c+0)throw c;we(1,0)}}function Zp(s,a,l,c){var p=fe();try{ol(s,a,l,c)}catch(_){if(he(p),_!==_+0)throw _;we(1,0)}}function eh(s,a,l,c,p,_,M){var x=fe();try{Mc(s,a,l,c,p,_,M)}catch(A){if(he(x),A!==A+0)throw A;we(1,0)}}function Mf(s,a,l,c,p,_,M){var x=fe();try{en(s,a,l,c,p,_,M)}catch(A){if(he(x),A!==A+0)throw A;we(1,0)}}function th(s,a,l,c,p,_,M,x){var A=fe();try{Qc(s,a,l,c,p,_,M,x)}catch(L){if(he(A),L!==L+0)throw L;we(1,0)}}function Hu(s,a,l,c,p,_,M,x,A){var L=fe();try{$l(s,a,l,c,p,_,M,x,A)}catch(V){if(he(L),V!==V+0)throw V;we(1,0)}}function rh(s,a,l,c){var p=fe();try{ua(s,a,l,c)}catch(_){if(he(p),_!==_+0)throw _;we(1,0)}}function ih(s,a,l,c,p,_,M,x,A){var L=fe();try{return kl(s,a,l,c,p,_,M,x,A)}catch(V){if(he(L),V!==V+0)throw V;we(1,0)}}function sh(s,a,l,c,p,_,M,x){var A=fe();try{return Il(s,a,l,c,p,_,M,x)}catch(L){if(he(A),L!==L+0)throw L;we(1,0)}}function nh(s,a,l,c,p){var _=fe();try{Al(s,a,l,c,p)}catch(M){if(he(_),M!==M+0)throw M;we(1,0)}}function ah(s,a,l){var c=fe();try{return Cl(s,a,l)}catch(p){if(he(c),p!==p+0)throw p;return we(1,0),0n}}function oh(s,a,l,c,p,_,M){var x=fe();try{Ol(s,a,l,c,p,_,M)}catch(A){if(he(x),A!==A+0)throw A;we(1,0)}}function lh(s,a,l,c,p,_,M,x,A,L,V,W){var Z=fe();try{Fl(s,a,l,c,p,_,M,x,A,L,V,W)}catch(de){if(he(Z),de!==de+0)throw de;we(1,0)}}function ch(s,a,l,c,p,_,M,x,A,L,V,W){var Z=fe();try{Dl(s,a,l,c,p,_,M,x,A,L,V,W)}catch(de){if(he(Z),de!==de+0)throw de;we(1,0)}}function uh(s,a,l,c,p,_,M,x,A,L,V,W,Z){var de=fe();try{return Ll(s,a,l,c,p,_,M,x,A,L,V,W,Z)}catch(Pe){if(he(de),Pe!==Pe+0)throw Pe;we(1,0)}}function dh(s,a,l,c,p,_,M,x,A,L,V){var W=fe();try{jl(s,a,l,c,p,_,M,x,A,L,V)}catch(Z){if(he(W),Z!==Z+0)throw Z;we(1,0)}}function ph(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe){var ht=fe();try{pa(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe)}catch(Pt){if(he(ht),Pt!==Pt+0)throw Pt;we(1,0)}}function hh(s,a,l,c){var p=fe();try{zl(s,a,l,c)}catch(_){if(he(p),_!==_+0)throw _;we(1,0)}}function fh(s,a,l){var c=fe();try{Bl(s,a,l)}catch(p){if(he(c),p!==p+0)throw p;we(1,0)}}function mh(s,a,l,c,p){var _=fe();try{cl(s,a,l,c,p)}catch(M){if(he(_),M!==M+0)throw M;we(1,0)}}function _h(s,a,l,c,p){var _=fe();try{ll(s,a,l,c,p)}catch(M){if(he(_),M!==M+0)throw M;we(1,0)}}function gh(s,a,l,c,p,_,M,x,A,L,V,W){var Z=fe();try{ha(s,a,l,c,p,_,M,x,A,L,V,W)}catch(de){if(he(Z),de!==de+0)throw de;we(1,0)}}function yh(s,a,l){var c=fe();try{return Nl(s,a,l)}catch(p){if(he(c),p!==p+0)throw p;we(1,0)}}function wh(s,a,l,c,p,_){var M=fe();try{Vl(s,a,l,c,p,_)}catch(x){if(he(M),x!==x+0)throw x;we(1,0)}}function Mh(s,a,l,c,p,_,M,x,A,L){var V=fe();try{Ul(s,a,l,c,p,_,M,x,A,L)}catch(W){if(he(V),W!==W+0)throw W;we(1,0)}}function bh(s,a,l,c,p,_,M,x,A){var L=fe();try{Rl(s,a,l,c,p,_,M,x,A)}catch(V){if(he(L),V!==V+0)throw V;we(1,0)}}function vh(s,a){var l=fe();try{return Hl(s,a)}catch(c){if(he(l),c!==c+0)throw c;we(1,0)}}function xh(s,a,l,c,p,_,M,x,A,L,V,W,Z,de){var Pe=fe();try{Lu(s,a,l,c,p,_,M,x,A,L,V,W,Z,de)}catch(Ue){if(he(Pe),Ue!==Ue+0)throw Ue;we(1,0)}}function Th(s,a,l,c,p,_){var M=fe();try{return Gl(s,a,l,c,p,_)}catch(x){if(he(M),x!==x+0)throw x;we(1,0)}}function Eh(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue){var Qe=fe();try{ma(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue)}catch(ht){if(he(Qe),ht!==ht+0)throw ht;we(1,0)}}function Ph(s,a,l,c){var p=fe();try{return Xl(s,a,l,c)}catch(_){if(he(p),_!==_+0)throw _;return we(1,0),0n}}function Ch(s,a,l,c,p,_){var M=fe();try{return Kl(s,a,l,c,p,_)}catch(x){if(he(M),x!==x+0)throw x;we(1,0)}}function Sh(s,a,l,c,p,_,M,x){var A=fe();try{Ql(s,a,l,c,p,_,M,x)}catch(L){if(he(A),L!==L+0)throw L;we(1,0)}}function $h(s,a,l,c,p,_,M,x,A){var L=fe();try{ga(s,a,l,c,p,_,M,x,A)}catch(V){if(he(L),V!==V+0)throw V;we(1,0)}}function kh(s,a,l,c,p,_,M){var x=fe();try{_a(s,a,l,c,p,_,M)}catch(A){if(he(x),A!==A+0)throw A;we(1,0)}}function Ih(s,a,l,c,p,_){var M=fe();try{return ql(s,a,l,c,p,_)}catch(x){if(he(M),x!==x+0)throw x;we(1,0)}}function bf(s,a,l,c,p,_,M,x,A,L,V){var W=fe();try{return Jl(s,a,l,c,p,_,M,x,A,L,V)}catch(Z){if(he(W),Z!==Z+0)throw Z;we(1,0)}}function At(s,a,l,c,p,_,M){var x=fe();try{return Yl(s,a,l,c,p,_,M)}catch(A){if(he(x),A!==A+0)throw A;we(1,0)}}function Ah(s,a,l,c,p,_){var M=fe();try{return Zl(s,a,l,c,p,_)}catch(x){if(he(M),x!==x+0)throw x;we(1,0)}}function Sr(s,a){var l=fe();try{return fa(s,a)}catch(c){if(he(l),c!==c+0)throw c;return we(1,0),0n}}function Or(s,a,l,c,p,_,M){var x=fe();try{return Wl(s,a,l,c,p,_,M)}catch(A){if(he(x),A!==A+0)throw A;we(1,0)}}function Vr(s,a,l){var c=fe();try{tc(s,a,l)}catch(p){if(he(c),p!==p+0)throw p;we(1,0)}}function fs(s,a,l,c,p,_,M,x){var A=fe();try{return ya(s,a,l,c,p,_,M,x)}catch(L){if(he(A),L!==L+0)throw L;we(1,0)}}function vf(s,a,l,c,p,_,M,x,A){var L=fe();try{ec(s,a,l,c,p,_,M,x,A)}catch(V){if(he(L),V!==V+0)throw V;we(1,0)}}function Oh(s,a,l,c){var p=fe();try{rc(s,a,l,c)}catch(_){if(he(p),_!==_+0)throw _;we(1,0)}}function Fh(s,a,l,c,p,_,M,x){var A=fe();try{return ic(s,a,l,c,p,_,M,x)}catch(L){if(he(A),L!==L+0)throw L;we(1,0)}}function C(s,a,l,c,p,_){var M=fe();try{return sc(s,a,l,c,p,_)}catch(x){if(he(M),x!==x+0)throw x;we(1,0)}}function F(s,a,l,c,p,_){var M=fe();try{return nc(s,a,l,c,p,_)}catch(x){if(he(M),x!==x+0)throw x;we(1,0)}}function N(s,a,l,c,p,_,M,x,A,L,V,W){var Z=fe();try{return wa(s,a,l,c,p,_,M,x,A,L,V,W)}catch(de){if(he(Z),de!==de+0)throw de;we(1,0)}}function me(s,a,l,c,p,_,M,x){var A=fe();try{return ac(s,a,l,c,p,_,M,x)}catch(L){if(he(A),L!==L+0)throw L;we(1,0)}}function ke(s,a,l,c,p,_,M,x,A,L,V){var W=fe();try{return oc(s,a,l,c,p,_,M,x,A,L,V)}catch(Z){if(he(W),Z!==Z+0)throw Z;we(1,0)}}function Se(s,a,l,c,p,_,M){var x=fe();try{return Ma(s,a,l,c,p,_,M)}catch(A){if(he(x),A!==A+0)throw A;we(1,0)}}function ze(s,a,l,c,p,_,M,x,A,L,V,W,Z){var de=fe();try{return lc(s,a,l,c,p,_,M,x,A,L,V,W,Z)}catch(Pe){if(he(de),Pe!==Pe+0)throw Pe;we(1,0)}}function Ke(s,a,l,c,p,_,M){var x=fe();try{return cc(s,a,l,c,p,_,M)}catch(A){if(he(x),A!==A+0)throw A;we(1,0)}}function Ze(s,a,l,c,p,_,M){var x=fe();try{Ni(s,a,l,c,p,_,M)}catch(A){if(he(x),A!==A+0)throw A;we(1,0)}}function at(s,a,l,c,p,_,M){var x=fe();try{return uc(s,a,l,c,p,_,M)}catch(A){if(he(x),A!==A+0)throw A;we(1,0)}}function vt(s,a,l){var c=fe();try{return dc(s,a,l)}catch(p){if(he(c),p!==p+0)throw p;return we(1,0),0n}}function kt(s,a,l){var c=fe();try{return pc(s,a,l)}catch(p){if(he(c),p!==p+0)throw p;return we(1,0),0n}}function Mt(s,a,l,c){var p=fe();try{return hc(s,a,l,c)}catch(_){if(he(p),_!==_+0)throw _;we(1,0)}}function zt(s,a,l,c,p,_,M){var x=fe();try{ba(s,a,l,c,p,_,M)}catch(A){if(he(x),A!==A+0)throw A;we(1,0)}}function Ct(s,a,l,c,p,_,M){var x=fe();try{fc(s,a,l,c,p,_,M)}catch(A){if(he(x),A!==A+0)throw A;we(1,0)}}function Ot(s,a,l,c,p,_,M,x,A,L,V){var W=fe();try{mc(s,a,l,c,p,_,M,x,A,L,V)}catch(Z){if(he(W),Z!==Z+0)throw Z;we(1,0)}}function ft(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe){var ht=fe();try{gc(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe)}catch(Pt){if(he(ht),Pt!==Pt+0)throw Pt;we(1,0)}}function Vt(s,a,l,c,p,_,M,x,A,L,V,W){var Z=fe();try{yc(s,a,l,c,p,_,M,x,A,L,V,W)}catch(de){if(he(Z),de!==de+0)throw de;we(1,0)}}function rr(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue){var Qe=fe();try{wc(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue)}catch(ht){if(he(Qe),ht!==ht+0)throw ht;we(1,0)}}function ar(s,a,l,c,p,_,M,x,A){var L=fe();try{Is(s,a,l,c,p,_,M,x,A)}catch(V){if(he(L),V!==V+0)throw V;we(1,0)}}function fr(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe){var Ue=fe();try{return bc(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe)}catch(Qe){if(he(Ue),Qe!==Qe+0)throw Qe;we(1,0)}}function hr(s,a,l,c,p,_,M,x,A,L){var V=fe();try{return vc(s,a,l,c,p,_,M,x,A,L)}catch(W){if(he(V),W!==W+0)throw W;we(1,0)}}function br(s,a){var l=fe();try{xc(s,a)}catch(c){if(he(l),c!==c+0)throw c;we(1,0)}}function $r(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe){var Ue=fe();try{Cc(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe)}catch(Qe){if(he(Ue),Qe!==Qe+0)throw Qe;we(1,0)}}function Bt(s,a,l,c,p){var _=fe();try{Tc(s,a,l,c,p)}catch(M){if(he(_),M!==M+0)throw M;we(1,0)}}function Er(s,a,l,c,p,_,M){var x=fe();try{rn(s,a,l,c,p,_,M)}catch(A){if(he(x),A!==A+0)throw A;we(1,0)}}function ui(s,a,l,c,p){var _=fe();try{Pc(s,a,l,c,p)}catch(M){if(he(_),M!==M+0)throw M;we(1,0)}}function ms(s,a,l,c,p,_,M,x){var A=fe();try{Ec(s,a,l,c,p,_,M,x)}catch(L){if(he(A),L!==L+0)throw L;we(1,0)}}function Qr(s,a,l,c,p,_,M,x,A,L,V){var W=fe();try{Tp(s,a,l,c,p,_,M,x,A,L,V)}catch(Z){if(he(W),Z!==Z+0)throw Z;we(1,0)}}function fi(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe){var ht=fe();try{return Sc(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe)}catch(Pt){if(he(ht),Pt!==Pt+0)throw Pt;we(1,0)}}function Aa(s,a,l,c,p,_,M,x,A,L,V,W,Z,de){var Pe=fe();try{$c(s,a,l,c,p,_,M,x,A,L,V,W,Z,de)}catch(Ue){if(he(Pe),Ue!==Ue+0)throw Ue;we(1,0)}}function _s(s,a,l,c,p){var _=fe();try{return ju(s,a,l,c,p)}catch(M){if(he(_),M!==M+0)throw M;we(1,0)}}function Oa(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe,ht,Pt,Rt,Xt,mr){var Lr=fe();try{return kc(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe,ht,Pt,Rt,Xt,mr)}catch(Vi){if(he(Lr),Vi!==Vi+0)throw Vi;we(1,0)}}function Fa(s,a,l,c,p,_,M,x,A,L,V){var W=fe();try{hs(s,a,l,c,p,_,M,x,A,L,V)}catch(Z){if(he(W),Z!==Z+0)throw Z;we(1,0)}}function or(s,a,l,c,p,_,M){var x=fe();try{return Ic(s,a,l,c,p,_,M)}catch(A){if(he(x),A!==A+0)throw A;we(1,0)}}function Pr(s,a,l,c,p,_,M,x,A,L,V,W,Z,de){var Pe=fe();try{Oc(s,a,l,c,p,_,M,x,A,L,V,W,Z,de)}catch(Ue){if(he(Pe),Ue!==Ue+0)throw Ue;we(1,0)}}function is(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe,ht){var Pt=fe();try{Fc(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe,ht)}catch(Rt){if(he(Pt),Rt!==Rt+0)throw Rt;we(1,0)}}function gs(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe){var ht=fe();try{Dc(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe)}catch(Pt){if(he(ht),Pt!==Pt+0)throw Pt;we(1,0)}}function Os(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue){var Qe=fe();try{Lc(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue)}catch(ht){if(he(Qe),ht!==ht+0)throw ht;we(1,0)}}function Da(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe){var Ue=fe();try{jc(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe)}catch(Qe){if(he(Ue),Qe!==Qe+0)throw Qe;we(1,0)}}function qu(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe,ht,Pt,Rt,Xt){var mr=fe();try{sn(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe,ht,Pt,Rt,Xt)}catch(Lr){if(he(mr),Lr!==Lr+0)throw Lr;we(1,0)}}function Qu(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe,ht,Pt){var Rt=fe();try{Ep(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe,ht,Pt)}catch(Xt){if(he(Rt),Xt!==Xt+0)throw Xt;we(1,0)}}function Xu(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe,ht){var Pt=fe();try{zc(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe,ht)}catch(Rt){if(he(Pt),Rt!==Rt+0)throw Rt;we(1,0)}}function xf(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe,ht,Pt,Rt,Xt){var mr=fe();try{Bc(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe,ht,Pt,Rt,Xt)}catch(Lr){if(he(mr),Lr!==Lr+0)throw Lr;we(1,0)}}function Ju(s,a,l,c,p,_,M,x,A,L,V,W,Z){var de=fe();try{Rc(s,a,l,c,p,_,M,x,A,L,V,W,Z)}catch(Pe){if(he(de),Pe!==Pe+0)throw Pe;we(1,0)}}function La(s,a,l,c,p,_,M,x,A,L){var V=fe();try{Nc(s,a,l,c,p,_,M,x,A,L)}catch(W){if(he(V),W!==W+0)throw W;we(1,0)}}function Yu(s,a,l,c,p,_,M,x){var A=fe();try{return Vc(s,a,l,c,p,_,M,x)}catch(L){if(he(A),L!==L+0)throw L;return we(1,0),0n}}function Dh(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe){var Ue=fe();try{va(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe)}catch(Qe){if(he(Ue),Qe!==Qe+0)throw Qe;we(1,0)}}function Zu(s,a,l,c,p,_,M,x,A,L){var V=fe();try{Uc(s,a,l,c,p,_,M,x,A,L)}catch(W){if(he(V),W!==W+0)throw W;we(1,0)}}function ja(s,a,l,c,p,_,M,x,A){var L=fe();try{Wc(s,a,l,c,p,_,M,x,A)}catch(V){if(he(L),V!==V+0)throw V;we(1,0)}}function Tf(s,a,l,c,p,_,M){var x=fe();try{return Ac(s,a,l,c,p,_,M)}catch(A){if(he(x),A!==A+0)throw A;we(1,0)}}function Ef(s,a,l){var c=fe();try{return zu(s,a,l)}catch(p){if(he(c),p!==p+0)throw p;return we(1,0),0n}}function Pf(s,a,l,c,p,_,M,x,A,L,V,W,Z,de){var Pe=fe();try{xa(s,a,l,c,p,_,M,x,A,L,V,W,Z,de)}catch(Ue){if(he(Pe),Ue!==Ue+0)throw Ue;we(1,0)}}function ys(s,a,l,c,p,_,M,x,A,L,V,W,Z){var de=fe();try{Kc(s,a,l,c,p,_,M,x,A,L,V,W,Z)}catch(Pe){if(he(de),Pe!==Pe+0)throw Pe;we(1,0)}}function Mi(s,a,l,c,p,_,M){var x=fe();try{return Ta(s,a,l,c,p,_,M)}catch(A){if(he(x),A!==A+0)throw A;we(1,0)}}function Jv(s,a,l,c,p,_,M,x,A,L,V,W,Z){var de=fe();try{Hc(s,a,l,c,p,_,M,x,A,L,V,W,Z)}catch(Pe){if(he(de),Pe!==Pe+0)throw Pe;we(1,0)}}function Yv(s,a,l,c,p,_,M,x,A,L,V,W){var Z=fe();try{return qc(s,a,l,c,p,_,M,x,A,L,V,W)}catch(de){if(he(Z),de!==de+0)throw de;we(1,0)}}function Zv(s,a,l,c,p,_){var M=fe();try{Gc(s,a,l,c,p,_)}catch(x){if(he(M),x!==x+0)throw x;we(1,0)}}function ex(s,a,l,c,p,_,M,x,A,L,V){var W=fe();try{nn(s,a,l,c,p,_,M,x,A,L,V)}catch(Z){if(he(W),Z!==Z+0)throw Z;we(1,0)}}function tx(s,a,l,c,p,_,M,x){var A=fe();try{Xc(s,a,l,c,p,_,M,x)}catch(L){if(he(A),L!==L+0)throw L;we(1,0)}}function rx(s,a,l,c,p,_){var M=fe();try{Jc(s,a,l,c,p,_)}catch(x){if(he(M),x!==x+0)throw x;we(1,0)}}function ix(s,a,l,c,p,_,M,x,A,L){var V=fe();try{Yc(s,a,l,c,p,_,M,x,A,L)}catch(W){if(he(V),W!==W+0)throw W;we(1,0)}}function sx(s,a,l,c){var p=fe();try{return tn(s,a,l,c)}catch(_){if(he(p),_!==_+0)throw _;we(1,0)}}function nx(s,a,l,c,p,_,M,x,A,L){var V=fe();try{Ea(s,a,l,c,p,_,M,x,A,L)}catch(W){if(he(V),W!==W+0)throw W;we(1,0)}}function ax(s,a,l,c,p,_){var M=fe();try{an(s,a,l,c,p,_)}catch(x){if(he(M),x!==x+0)throw x;we(1,0)}}function ox(s,a,l){var c=fe();try{return da(s,a,l)}catch(p){if(he(c),p!==p+0)throw p;we(1,0)}}function lx(s,a,l,c,p,_,M,x){var A=fe();try{return Pp(s,a,l,c,p,_,M,x)}catch(L){if(he(A),L!==L+0)throw L;we(1,0)}}function cx(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe){var Ue=fe();try{Bu(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe)}catch(Qe){if(he(Ue),Qe!==Qe+0)throw Qe;we(1,0)}}function ux(s,a,l,c,p,_,M,x,A,L,V,W,Z,de){var Pe=fe();try{Zc(s,a,l,c,p,_,M,x,A,L,V,W,Z,de)}catch(Ue){if(he(Pe),Ue!==Ue+0)throw Ue;we(1,0)}}function dx(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe){var Ue=fe();try{eu(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe)}catch(Qe){if(he(Ue),Qe!==Qe+0)throw Qe;we(1,0)}}function px(s,a,l,c,p,_,M,x){var A=fe();try{Pa(s,a,l,c,p,_,M,x)}catch(L){if(he(A),L!==L+0)throw L;we(1,0)}}function hx(s,a,l,c){var p=fe();try{return tu(s,a,l,c)}catch(_){if(he(p),_!==_+0)throw _;we(1,0)}}function fx(s,a,l){var c=fe();try{return Sl(s,a,l)}catch(p){if(he(c),p!==p+0)throw p;we(1,0)}}function mx(s,a,l,c,p,_,M,x,A){var L=fe();try{ru(s,a,l,c,p,_,M,x,A)}catch(V){if(he(L),V!==V+0)throw V;we(1,0)}}function _x(s,a,l,c,p,_,M,x,A,L,V){var W=fe();try{Ca(s,a,l,c,p,_,M,x,A,L,V)}catch(Z){if(he(W),Z!==Z+0)throw Z;we(1,0)}}function gx(s,a,l,c,p,_){var M=fe();try{iu(s,a,l,c,p,_)}catch(x){if(he(M),x!==x+0)throw x;we(1,0)}}function yx(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue){var Qe=fe();try{return su(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue)}catch(ht){if(he(Qe),ht!==ht+0)throw ht;we(1,0)}}function wx(s,a,l,c,p,_,M,x){var A=fe();try{return Sa(s,a,l,c,p,_,M,x)}catch(L){if(he(A),L!==L+0)throw L;return we(1,0),0n}}function Mx(s,a,l,c,p,_,M,x,A,L,V,W,Z){var de=fe();try{Cp(s,a,l,c,p,_,M,x,A,L,V,W,Z)}catch(Pe){if(he(de),Pe!==Pe+0)throw Pe;we(1,0)}}function bx(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe,ht,Pt,Rt,Xt,mr,Lr,Vi){var m1=fe();try{Sp(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe,ht,Pt,Rt,Xt,mr,Lr,Vi)}catch(Cf){if(he(m1),Cf!==Cf+0)throw Cf;we(1,0)}}function vx(s,a,l,c,p,_,M,x,A,L,V,W,Z){var de=fe();try{$a(s,a,l,c,p,_,M,x,A,L,V,W,Z)}catch(Pe){if(he(de),Pe!==Pe+0)throw Pe;we(1,0)}}function xx(s,a,l,c,p,_,M,x,A,L){var V=fe();try{on(s,a,l,c,p,_,M,x,A,L)}catch(W){if(he(V),W!==W+0)throw W;we(1,0)}}function Tx(s,a,l,c,p,_,M,x,A,L,V,W,Z,de){var Pe=fe();try{nu(s,a,l,c,p,_,M,x,A,L,V,W,Z,de)}catch(Ue){if(he(Pe),Ue!==Ue+0)throw Ue;we(1,0)}}function Ex(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe,ht,Pt,Rt){var Xt=fe();try{lu(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe,ht,Pt,Rt)}catch(mr){if(he(Xt),mr!==mr+0)throw mr;we(1,0)}}function Px(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe,ht,Pt,Rt,Xt){var mr=fe();try{hn(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe,ht,Pt,Rt,Xt)}catch(Lr){if(he(mr),Lr!==Lr+0)throw Lr;we(1,0)}}function Cx(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe,ht,Pt,Rt){var Xt=fe();try{ln(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe,ht,Pt,Rt)}catch(mr){if(he(Xt),mr!==mr+0)throw mr;we(1,0)}}function Sx(s,a,l,c,p,_,M,x,A,L,V,W,Z){var de=fe();try{au(s,a,l,c,p,_,M,x,A,L,V,W,Z)}catch(Pe){if(he(de),Pe!==Pe+0)throw Pe;we(1,0)}}function $x(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe,ht,Pt){var Rt=fe();try{ou(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe,ht,Pt)}catch(Xt){if(he(Rt),Xt!==Xt+0)throw Xt;we(1,0)}}function kx(s,a,l,c,p,_,M,x){var A=fe();try{cu(s,a,l,c,p,_,M,x)}catch(L){if(he(A),L!==L+0)throw L;we(1,0)}}function Ix(s,a,l,c,p,_,M,x,A,L,V){var W=fe();try{uu(s,a,l,c,p,_,M,x,A,L,V)}catch(Z){if(he(W),Z!==Z+0)throw Z;we(1,0)}}function Ax(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe,ht,Pt,Rt,Xt,mr){var Lr=fe();try{return cn(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe,Ue,Qe,ht,Pt,Rt,Xt,mr)}catch(Vi){if(he(Lr),Vi!==Vi+0)throw Vi;we(1,0)}}function Ox(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe){var Ue=fe();try{du(s,a,l,c,p,_,M,x,A,L,V,W,Z,de,Pe)}catch(Qe){if(he(Ue),Qe!==Qe+0)throw Qe;we(1,0)}}function Fx(s,a,l,c,p,_,M){var x=fe();try{pu(s,a,l,c,p,_,M)}catch(A){if(he(x),A!==A+0)throw A;we(1,0)}}function Dx(s,a,l,c,p){var _=fe();try{hu(s,a,l,c,p)}catch(M){if(he(_),M!==M+0)throw M;we(1,0)}}function Lx(s,a,l,c,p,_,M,x,A,L,V,W){var Z=fe();try{un(s,a,l,c,p,_,M,x,A,L,V,W)}catch(de){if(he(Z),de!==de+0)throw de;we(1,0)}}function jx(s,a,l,c,p,_,M){var x=fe();try{fu(s,a,l,c,p,_,M)}catch(A){if(he(x),A!==A+0)throw A;we(1,0)}}function zx(s,a,l,c,p){var _=fe();try{mu(s,a,l,c,p)}catch(M){if(he(_),M!==M+0)throw M;we(1,0)}}function Bx(s,a,l,c,p,_,M,x,A,L,V,W){var Z=fe();try{_c(s,a,l,c,p,_,M,x,A,L,V,W)}catch(de){if(he(Z),de!==de+0)throw de;we(1,0)}}function Rx(s,a,l,c,p,_,M){var x=fe();try{return yu(s,a,l,c,p,_,M)}catch(A){if(he(x),A!==A+0)throw A;we(1,0)}}function Nx(s,a,l,c,p){var _=fe();try{wu(s,a,l,c,p)}catch(M){if(he(_),M!==M+0)throw M;we(1,0)}}function Vx(s,a,l,c,p,_,M){var x=fe();try{Mu(s,a,l,c,p,_,M)}catch(A){if(he(x),A!==A+0)throw A;we(1,0)}}function Ux(s,a,l,c,p,_,M,x,A,L,V){var W=fe();try{bu(s,a,l,c,p,_,M,x,A,L,V)}catch(Z){if(he(W),Z!==Z+0)throw Z;we(1,0)}}function Wx(s,a,l,c,p){var _=fe();try{return vu(s,a,l,c,p)}catch(M){if(he(_),M!==M+0)throw M;we(1,0)}}function Gx(s,a,l,c,p,_,M,x){var A=fe();try{xu(s,a,l,c,p,_,M,x)}catch(L){if(he(A),L!==L+0)throw L;we(1,0)}}function Kx(s,a,l,c,p,_){var M=fe();try{return ka(s,a,l,c,p,_)}catch(x){if(he(M),x!==x+0)throw x;we(1,0)}}function Hx(s,a,l,c){var p=fe();try{Tu(s,a,l,c)}catch(_){if(he(p),_!==_+0)throw _;we(1,0)}}function qx(s,a,l,c,p,_,M,x){var A=fe();try{Eu(s,a,l,c,p,_,M,x)}catch(L){if(he(A),L!==L+0)throw L;we(1,0)}}function Qx(s,a,l,c,p,_,M){var x=fe();try{Ia(s,a,l,c,p,_,M)}catch(A){if(he(x),A!==A+0)throw A;we(1,0)}}function Xx(s,a,l,c,p){var _=fe();try{Pu(s,a,l,c,p)}catch(M){if(he(_),M!==M+0)throw M;we(1,0)}}function Jx(s,a,l,c,p){var _=fe();try{return gu(s,a,l,c,p)}catch(M){if(he(_),M!==M+0)throw M;we(1,0)}}function Yx(s,a,l,c){var p=fe();try{return _u(s,a,l,c)}catch(_){if(he(p),_!==_+0)throw _;we(1,0)}}function Zx(s,a,l,c){var p=fe();try{return Ru(s,a,l,c)}catch(_){if(he(p),_!==_+0)throw _;we(1,0)}}function e1(s,a,l,c,p){var _=fe();try{Cu(s,a,l,c,p)}catch(M){if(he(_),M!==M+0)throw M;we(1,0)}}function t1(s,a,l,c){var p=fe();try{return na(s,a,l,c)}catch(_){if(he(p),_!==_+0)throw _;we(1,0)}}function r1(s,a,l,c,p,_){var M=fe();try{return Su(s,a,l,c,p,_)}catch(x){if(he(M),x!==x+0)throw x;we(1,0)}}function i1(s,a,l,c,p){var _=fe();try{return $u(s,a,l,c,p)}catch(M){if(he(_),M!==M+0)throw M;we(1,0)}}function s1(s,a,l,c,p){var _=fe();try{return dn(s,a,l,c,p)}catch(M){if(he(_),M!==M+0)throw M;we(1,0)}}function n1(s,a,l,c,p,_,M){var x=fe();try{return ku(s,a,l,c,p,_,M)}catch(A){if(he(x),A!==A+0)throw A;we(1,0)}}function a1(s,a,l,c,p,_,M,x,A,L,V,W){var Z=fe();try{Iu(s,a,l,c,p,_,M,x,A,L,V,W)}catch(de){if(he(Z),de!==de+0)throw de;we(1,0)}}function o1(s,a,l,c,p){var _=fe();try{return Ou(s,a,l,c,p)}catch(M){if(he(_),M!==M+0)throw M;we(1,0)}}function l1(s,a,l,c,p,_,M,x,A,L,V,W,Z){var de=fe();try{Au(s,a,l,c,p,_,M,x,A,L,V,W,Z)}catch(Pe){if(he(de),Pe!==Pe+0)throw Pe;we(1,0)}}function c1(s,a,l,c,p,_,M,x){var A=fe();try{return Fu(s,a,l,c,p,_,M,x)}catch(L){if(he(A),L!==L+0)throw L;we(1,0)}}function u1(s,a,l,c,p){var _=fe();try{return pn(s,a,l,c,p)}catch(M){if(he(_),M!==M+0)throw M;we(1,0)}}function d1(s,a,l,c,p){var _=fe();try{return Du(s,a,l,c,p)}catch(M){if(he(_),M!==M+0)throw M;we(1,0)}}function p1(s){var a=fe();try{return ca(s)}catch(l){if(he(a),l!==l+0)throw l;return we(1,0),0n}}function h1(s,a,l,c,p,_){var M=fe();try{return gl(s,a,l,c,p,_)}catch(x){if(he(M),x!==x+0)throw x;we(1,0)}}function f1(s,a,l,c,p,_){var M=fe();try{return ml(s,a,l,c,p,_)}catch(x){if(he(M),x!==x+0)throw x;we(1,0)}}return i.stackSave=()=>fe(),i.stackRestore=s=>he(s),i.stackAlloc=s=>Xs(s),i.setValue=function(s,a,l="i8"){switch(l.endsWith("*")&&(l="*"),l){case"i1":case"i8":B()[s>>>0]=a;break;case"i16":oe()[s>>>1>>>0]=a;break;case"i32":ye()[s>>>2>>>0]=a;break;case"i64":se[s>>>3]=BigInt(a);break;case"float":et()[s>>>2>>>0]=a;break;case"double":tt()[s>>>3>>>0]=a;break;case"*":be()[s>>>2>>>0]=a;break;default:Ve(`invalid type for setValue: ${l}`)}},i.getValue=function(s,a="i8"){switch(a.endsWith("*")&&(a="*"),a){case"i1":case"i8":return B()[s>>>0];case"i16":return oe()[s>>>1>>>0];case"i32":return ye()[s>>>2>>>0];case"i64":return se[s>>>3];case"float":return et()[s>>>2>>>0];case"double":return tt()[s>>>3>>>0];case"*":return be()[s>>>2>>>0];default:Ve(`invalid type for getValue: ${a}`)}},i.UTF8ToString=er,i.stringToUTF8=Hr,i.lengthBytesUTF8=Kr,function s(){if(0<Oe)Ne=s;else if(f)n(i),X();else{for(;0<pt.length;)pt.shift()(i);0<Oe?Ne=s:(i.calledRun=!0,Te||(X(),n(i)))}}(),i.PTR_SIZE=4,i.webgpuInit=s=>{let a=new WeakMap,l,c,p=1;i.webgpuRegisterDevice=x=>{if(c!==void 0)throw Error("another WebGPU EP inference session is being created.");if(x){var A=a.get(x);if(!A){let L=((V,W=0)=>{var Z=Wo(W);return W=Wn(W,Z),Ar[Z]=V.queue,Ar[W]=V,W})(x,A=Lo(0));A=[p++,A,L],a.set(x,A)}return l=x,c=A[0],A}l=void 0,c=0};let _=new Map;i.webgpuOnCreateSession=x=>{if(c!==void 0){var A=c;if(c=void 0,x){let L=Nn(A);_.set(x,L),A===0&&s(l??Tt(L))}l=void 0}},i.webgpuOnReleaseSession=x=>{_.delete(x)};let M=Symbol("gpuBufferMetadata");i.webgpuRegisterBuffer=(x,A,L)=>{if(L)return x[M]=[L,NaN],L;if(L=x[M])return L[1]++,L[0];if((A=_.get(A))===void 0)throw Error("Invalid session handle passed to webgpuRegisterBuffer");return A=((V,W=0)=>(V.mapState!="pending"||Ve(),W=Uo(W,V.mapState=="mapped"?3:1),Ar[W]=V,V.mapState=="mapped"&&(hi[W]=[]),W))(x,A),x[M]=[A,1],A},i.webgpuUnregisterBuffer=x=>{let A=x[M];if(!A)throw Error("Buffer is not registered");A[1]--,A[1]===0&&(Vn(A[0]),delete x[M])},i.webgpuGetBuffer=x=>Tt(x),i.webgpuCreateDownloader=(x,A,L)=>{if((L=_.get(L))===void 0)throw Error("Invalid session handle passed to webgpuRegisterBuffer");let V=Tt(L),W=16*Math.ceil(Number(A)/16);return async()=>{let Z=V.createBuffer({size:W,usage:9});try{let de=V.createCommandEncoder();return de.copyBufferToBuffer(x,0,Z,0,W),V.queue.submit([de.finish()]),await Z.mapAsync(GPUMapMode.READ),Z.getMappedRange().slice(0,A)}finally{Z.destroy()}}},i.Sk=(x,A)=>{var L=A.buffer;let V=A.byteOffset,W=A.byteLength;if(A=16*Math.ceil(Number(W)/16),x=Tt(x),!l){var Z=Nn(c);l=Tt(Z)}let de=(Z=l.createBuffer({mappedAtCreation:!0,size:A,usage:6})).getMappedRange();new Uint8Array(de).set(new Uint8Array(L,V,W)),Z.unmap(),(L=l.createCommandEncoder()).copyBufferToBuffer(Z,0,x,0,A),l.queue.submit([L.finish()]),Z.destroy()}},u}),O0=Df,S_=(r=(e=globalThis.self)==null?void 0:e.name)==null?void 0:r.startsWith("em-pthread"),S_&&Df()}),Lf,Im,$_,bi,F0,jh,k_,I_,jf,A_,zf,D0,Bf,L0,Jm=Je(()=>{Xm(),Lf=typeof location>"u"?void 0:location.origin,Im=import.meta.url>"file:"&&import.meta.url<"file;",$_=()=>{{if(Im){let e=URL;return new URL(new e("ort.bundle.min.mjs",import.meta.url).href,Lf).href}return import.meta.url}},bi=$_(),F0=()=>{if(bi&&!bi.startsWith("blob:"))return bi.substring(0,bi.lastIndexOf("/")+1)},jh=(e,r)=>{try{let t=r??bi;return(t?new URL(e,t):new URL(e)).origin===Lf}catch{return!1}},k_=(e,r)=>{let t=r??bi;try{return(t?new URL(e,t):new URL(e)).href}catch{return}},I_=(e,r)=>`${r??"./"}${e}`,jf=async e=>{let r=await(await fetch(e,{credentials:"same-origin"})).blob();return URL.createObjectURL(r)},A_=async e=>(await import(e)).default,zf=(eT(),gd(k0)).default,D0=async()=>{if(!bi)throw new Error("Failed to load proxy worker: cannot determine the script source URL.");if(jh(bi))return[void 0,zf()];let e=await jf(bi);return[e,zf(e)]},Bf=(tT(),gd(A0)).default,L0=async(e,r,t)=>{if(!e&&!r&&Bf&&bi&&jh(bi))return[void 0,Bf];{let n="ort-wasm-simd-threaded.jsep.mjs",o=e??k_(n,r),i=t&&o&&!jh(o,r),u=i?await jf(o):o??I_(n,r);return[i?u:void 0,await A_(u)]}}}),Rf,zh,td,Nf,O_,F_,D_,Ym,lr,Pn=Je(()=>{Jm(),zh=!1,td=!1,Nf=!1,O_=()=>{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}},F_=()=>{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}},D_=()=>{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}},Ym=async e=>{if(zh)return Promise.resolve();if(td)throw new Error("multiple calls to 'initializeWebAssembly()' detected.");if(Nf)throw new Error("previous call to 'initializeWebAssembly()' failed.");td=!0;let r=e.initTimeout,t=e.numThreads;if(e.simd!==!1){if(e.simd==="relaxed"){if(!D_())throw new Error("Relaxed WebAssembly SIMD is not supported in the current environment.")}else if(!F_())throw new Error("WebAssembly SIMD is not supported in the current environment.")}let n=O_();t>1&&!n&&(typeof self<"u"&&!self.crossOriginIsolated&&console.warn("env.wasm.numThreads is set to "+t+", but this will not work unless you enable crossOriginIsolated mode. See https://web.dev/cross-origin-isolation-guide/ for more info."),console.warn("WebAssembly multi-threading is not supported in the current environment. Falling back to single-threading."),e.numThreads=t=1);let o=e.wasmPaths,i=typeof o=="string"?o:void 0,u=o==null?void 0:o.mjs,d=(u==null?void 0:u.href)??u,h=o==null?void 0:o.wasm,f=(h==null?void 0:h.href)??h,y=e.wasmBinary,[m,g]=await L0(d,i,t>1),v=!1,b=[];if(r>0&&b.push(new Promise(k=>{setTimeout(()=>{v=!0,k()},r)})),b.push(new Promise((k,z)=>{let E={numThreads:t};if(y)E.wasmBinary=y;else if(f||i)E.locateFile=T=>f??i+T;else if(d&&d.indexOf("blob:")!==0)E.locateFile=T=>new URL(T,d).href;else if(m){let T=F0();T&&(E.locateFile=D=>T+D)}g(E).then(T=>{td=!1,zh=!0,Rf=T,k(),m&&URL.revokeObjectURL(m)},T=>{td=!1,Nf=!0,z(T)})})),await Promise.race(b),v)throw new Error(`WebAssembly backend initializing failed due to timeout: ${r}ms`)},lr=()=>{if(zh&&Rf)return Rf;throw new Error("WebAssembly is not initialized yet.")}}),qi,ef,ir,Zm=Je(()=>{Pn(),qi=(e,r)=>{let t=lr(),n=t.lengthBytesUTF8(e)+1,o=t._malloc(n);return t.stringToUTF8(e,o,n),r.push(o),o},ef=(e,r,t,n)=>{if(typeof e=="object"&&e!==null){if(t.has(e))throw new Error("Circular reference in options");t.add(e)}Object.entries(e).forEach(([o,i])=>{let u=r?r+o:o;if(typeof i=="object")ef(i,u+".",t,n);else if(typeof i=="string"||typeof i=="number")n(u,i.toString());else if(typeof i=="boolean")n(u,i?"1":"0");else throw new Error(`Can't handle extra config type: ${typeof i}`)})},ir=e=>{let r=lr(),t=r.stackSave();try{let n=r.PTR_SIZE,o=r.stackAlloc(2*n);r._OrtGetLastError(o,o+n);let i=Number(r.getValue(o,n===4?"i32":"i64")),u=r.getValue(o+n,"*"),d=u?r.UTF8ToString(u):"";throw new Error(`${e} ERROR_CODE: ${i}, ERROR_MESSAGE: ${d}`)}finally{r.stackRestore(t)}}}),j0,rT=Je(()=>{Pn(),Zm(),j0=e=>{let r=lr(),t=0,n=[],o=e||{};try{if((e==null?void 0:e.logSeverityLevel)===void 0)o.logSeverityLevel=2;else if(typeof e.logSeverityLevel!="number"||!Number.isInteger(e.logSeverityLevel)||e.logSeverityLevel<0||e.logSeverityLevel>4)throw new Error(`log severity level is not valid: ${e.logSeverityLevel}`);if((e==null?void 0:e.logVerbosityLevel)===void 0)o.logVerbosityLevel=0;else if(typeof e.logVerbosityLevel!="number"||!Number.isInteger(e.logVerbosityLevel))throw new Error(`log verbosity level is not valid: ${e.logVerbosityLevel}`);(e==null?void 0:e.terminate)===void 0&&(o.terminate=!1);let i=0;return(e==null?void 0:e.tag)!==void 0&&(i=qi(e.tag,n)),t=r._OrtCreateRunOptions(o.logSeverityLevel,o.logVerbosityLevel,!!o.terminate,i),t===0&&ir("Can't create run options."),(e==null?void 0:e.extra)!==void 0&&ef(e.extra,"",new WeakSet,(u,d)=>{let h=qi(u,n),f=qi(d,n);r._OrtAddRunConfigEntry(t,h,f)!==0&&ir(`Can't set a run config entry: ${u} - ${d}.`)}),[t,n]}catch(i){throw t!==0&&r._OrtReleaseRunOptions(t),n.forEach(u=>r._free(u)),i}}}),L_,j_,z_,rd,B_,z0,iT=Je(()=>{Pn(),Zm(),L_=e=>{switch(e){case"disabled":return 0;case"basic":return 1;case"extended":return 2;case"all":return 99;default:throw new Error(`unsupported graph optimization level: ${e}`)}},j_=e=>{switch(e){case"sequential":return 0;case"parallel":return 1;default:throw new Error(`unsupported execution mode: ${e}`)}},z_=e=>{e.extra||(e.extra={}),e.extra.session||(e.extra.session={});let r=e.extra.session;r.use_ort_model_bytes_directly||(r.use_ort_model_bytes_directly="1"),e.executionProviders&&e.executionProviders.some(t=>(typeof t=="string"?t:t.name)==="webgpu")&&(e.enableMemPattern=!1)},rd=(e,r,t,n)=>{let o=qi(r,n),i=qi(t,n);lr()._OrtAddSessionConfigEntry(e,o,i)!==0&&ir(`Can't set a session config entry: ${r} - ${t}.`)},B_=async(e,r,t)=>{for(let n of r){let o=typeof n=="string"?n:n.name,i=[];switch(o){case"webnn":if(o="WEBNN",typeof n!="string"){let y=n==null?void 0:n.deviceType;y&&rd(e,"deviceType",y,t)}break;case"webgpu":if(o="JS",typeof n!="string"){let y=n;if(y!=null&&y.preferredLayout){if(y.preferredLayout!=="NCHW"&&y.preferredLayout!=="NHWC")throw new Error(`preferredLayout must be either 'NCHW' or 'NHWC': ${y.preferredLayout}`);rd(e,"preferredLayout",y.preferredLayout,t)}}break;case"wasm":case"cpu":continue;default:throw new Error(`not supported execution provider: ${o}`)}let u=qi(o,t),d=i.length,h=0,f=0;if(d>0){h=lr()._malloc(d*lr().PTR_SIZE),t.push(h),f=lr()._malloc(d*lr().PTR_SIZE),t.push(f);for(let y=0;y<d;y++)lr().setValue(h+y*lr().PTR_SIZE,i[y][0],"*"),lr().setValue(f+y*lr().PTR_SIZE,i[y][1],"*")}await lr()._OrtAppendExecutionProvider(e,u,h,f,d)!==0&&ir(`Can't append execution provider: ${o}.`)}},z0=async e=>{let r=lr(),t=0,n=[],o=e||{};z_(o);try{let i=L_(o.graphOptimizationLevel??"all"),u=j_(o.executionMode??"sequential"),d=typeof o.logId=="string"?qi(o.logId,n):0,h=o.logSeverityLevel??2;if(!Number.isInteger(h)||h<0||h>4)throw new Error(`log severity level is not valid: ${h}`);let f=o.logVerbosityLevel??0;if(!Number.isInteger(f)||f<0||f>4)throw new Error(`log verbosity level is not valid: ${f}`);let y=typeof o.optimizedModelFilePath=="string"?qi(o.optimizedModelFilePath,n):0;if(t=r._OrtCreateSessionOptions(i,!!o.enableCpuMemArena,!!o.enableMemPattern,u,!!o.enableProfiling,0,d,h,f,y),t===0&&ir("Can't create session options."),o.executionProviders&&await B_(t,o.executionProviders,n),o.enableGraphCapture!==void 0){if(typeof o.enableGraphCapture!="boolean")throw new Error(`enableGraphCapture must be a boolean value: ${o.enableGraphCapture}`);rd(t,"enableGraphCapture",o.enableGraphCapture.toString(),n)}if(o.freeDimensionOverrides)for(let[m,g]of Object.entries(o.freeDimensionOverrides)){if(typeof m!="string")throw new Error(`free dimension override name must be a string: ${m}`);if(typeof g!="number"||!Number.isInteger(g)||g<0)throw new Error(`free dimension override value must be a non-negative integer: ${g}`);let v=qi(m,n);r._OrtAddFreeDimensionOverride(t,v,g)!==0&&ir(`Can't set a free dimension override: ${m} - ${g}.`)}return o.extra!==void 0&&ef(o.extra,"",new WeakSet,(m,g)=>{rd(t,m,g,n)}),[t,n]}catch(i){throw t!==0&&r._OrtReleaseSessionOptions(t)!==0&&ir("Can't release session options."),n.forEach(u=>r._free(u)),i}}}),Mn,bs,bn,cf,tf,e_,t_,Am,St=Je(()=>{Mn=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}`)}},bs=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}`)}},bn=(e,r)=>{let t=[-1,4,1,1,2,2,4,8,-1,1,2,8,4,8,-1,-1,-1,-1,-1,-1,-1,.5,.5][e],n=typeof r=="number"?r:r.reduce((o,i)=>o*i,1);return t>0?Math.ceil(n*t):void 0},cf=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}`)}},tf=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}`)}},e_=e=>e==="float32"||e==="float16"||e==="int32"||e==="int64"||e==="uint32"||e==="uint8"||e==="bool"||e==="uint4"||e==="int4",t_=e=>e==="float32"||e==="float16"||e==="int32"||e==="int64"||e==="uint32"||e==="uint64"||e==="int8"||e==="uint8"||e==="bool"||e==="uint4"||e==="int4",Am=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}`)}}}),r_,B0=Je(()=>{Xm(),r_=async e=>{if(typeof e=="string"){let r=await fetch(e);if(!r.ok)throw new Error(`failed to load external data file: ${e}`);let t=r.headers.get("Content-Length"),n=t?parseInt(t,10):0;if(n<1073741824)return new Uint8Array(await r.arrayBuffer());{if(!r.body)throw new Error(`failed to load external data file: ${e}, no response body.`);let o=r.body.getReader(),i;try{i=new ArrayBuffer(n)}catch(d){if(d instanceof RangeError){let h=Math.ceil(n/65536);i=new WebAssembly.Memory({initial:h,maximum:h}).buffer}else throw d}let u=0;for(;;){let{done:d,value:h}=await o.read();if(d)break;let f=h.byteLength;new Uint8Array(i,u,f).set(h),u+=f}return new Uint8Array(i,0,n)}}else return e instanceof Blob?new Uint8Array(await e.arrayBuffer()):e instanceof Uint8Array?e:new Uint8Array(e)}}),R_,N_,V_,U_,i_,W_,Gt,vs=Je(()=>{St(),R_=["V","I","W","E","F"],N_=(e,r)=>{console.log(`[${R_[e]},${new Date().toISOString()}]${r}`)},i_=(e,r)=>{V_=e,U_=r},W_=(e,r)=>{let t=tf(e),n=tf(V_);t>=n&&N_(t,typeof r=="function"?r():r)},Gt=(...e)=>{U_&&W_(...e)}}),G_,Ua,Fe,rf,R0,N0,V0,Ft=Je(()=>{G_=class{static calcMatMulShape(e,r){return e[1]!==r[0]?void 0:[e[0],r[1]]}},Ua=class{static calcShape(e,r,t=!1){let n=e.length,o=r.length;if(n===0)return r;if(o===0)return e;let i=Math.max(e.length,r.length),u=new Array(i);if(t){if(n<2||o<2)return;let d=G_.calcMatMulShape([e[n-2],e[n-1]],[r[o-2],r[o-1]]);if(d===void 0)return;[u[i-2],u[i-1]]=d}for(let d=t?3:1;d<=i;d++){let h=n-d<0?1:e[n-d],f=o-d<0?1:r[o-d];if(h!==f&&h>1&&f>1)return;let y=Math.max(h,f);if(h&&f)u[i-d]=Math.max(h,f);else{if(y>1)return;u[i-d]=0}}return u}static isValidBroadcast(e,r){let t=e.length,n=r.length;if(t>n)return!1;for(let o=1;o<=t;o++)if(e[t-o]!==1&&e[t-o]!==r[n-o])return!1;return!0}},Fe=class Jh{static size(r){return Jh.getSizeFromDimensionRange(r,0,r.length)}static convertShape(r,t=4){let n=r.length;if(n===0)return[];let o=new Array(n),i=n-1;for(;i>=0;){if(r[i]%t===0){o[i]=r[i]/t;break}if(t%r[i]!==0)throw new Error("cannot convert shape");o[i]=1,t/=r[i],i--}for(i--;i>=0;i--)o[i]=r[i];return o}static sizeFromDimension(r,t){if(t<0||t>r.length)throw new Error(`invalid dimension of ${t} for sizeFromDimension as Tensor has ${r.length} dimensions.`);return Jh.getSizeFromDimensionRange(r,t,r.length)}static sizeToDimension(r,t){if(t<0||t>r.length)throw new Error(`invalid dimension of ${t} for sizeToDimension as Tensor has ${r.length} dimensions.`);return Jh.getSizeFromDimensionRange(r,0,t)}static getSizeFromDimensionRange(r,t,n){let o=1;for(let i=t;i<n;i++){if(r[i]<0)throw new Error("cannot get valid size from specified dimension range. Most likely the range contains negative values in them.");o*=Number(r[i])}return o}static computeStrides(r){let t=r.length;if(t===0)return[];if(t===1)return[1];let n=new Array(t);n[t-1]=1,n[t-2]=r[t-1];for(let o=t-3;o>=0;--o)n[o]=n[o+1]*r[o+1];return n}static normalizeAxis(r,t){if(r<-t&&r>=t)throw new Error("unsupported axis for this operation.");return r<0?r+t:r}static normalizeAxes(r,t){return r.map(n=>this.normalizeAxis(n,t??r.length))}static sortBasedOnPerm(r,t){return t?t.map(n=>r[n]):r.slice().reverse()}static padShape(r,t){let n=r.length;return r.map((o,i)=>o+t[i]+t[i+n])}static areEqual(r,t){return r.length!==t.length?!1:r.every((n,o)=>n===t[o])}},rf=class pd{static adjustPoolAttributes(r,t,n,o,i,u){if(!r&&n.length!==t.length-2)throw new Error("length of specified kernel shapes should be 2 less than length of input dimensions");if(r)for(let d=0;d<t.length-2;d++)d>=n.length?n.push(t[d+2]):n[d]=t[d+2];for(let d=0;d<n.length;d++)if(d<o.length){if(o[d]<0)throw new Error("strides should be greater than or equal to 1")}else o.push(1);for(let d=0;d<n.length;d++)if(d<i.length){if(i[d]<0)throw new Error("dilations should be greater than or equal to 1")}else i.push(1);for(let d=0;d<n.length*2;d++)if(d<u.length){if(u[d]<0)throw new Error("pad should be greater than or equal to 1")}else u.push(0);for(let d=0;d<n.length;d++){if(n[d]<=0)throw new Error("kernel shapes need to be greater than 0");if(u[d]>=n[d]||u[d+n.length]>=n[d])throw new Error("pads should be smaller than kernel")}}static adjustPadsBasedOnAutoPad(r,t,n,o,i,u,d){if(d){if(i.length!==2*(r.length-2))throw new Error("length of pads should be twice the length of data dimensions");if(t.length!==r.length-2)throw new Error("length of strides should be the length of data dimensions");if(o.length!==r.length-2)throw new Error("length of kernel shapes should be the length of data dimensions");for(let h=0;h<r.length-2;h++)pd.adjustPadAndReturnShape(r[h+(u?1:2)],t[h],n[h],o[h],i,h,h+r.length-2,d)}}static computePoolOutputShape(r,t,n,o,i,u,d){if(t.length<=0)throw new Error("input shape must be of size greater than 0");let h=[t[0],t[1]];return pd.computeShapeHelper(r,t,h,n,o,i,u,d),h}static computeConvOutputShape(r,t,n,o,i,u,d){if(r.length<=0||t.length<=0)throw new Error("invalid input tensor dims or invalid filter tensor dims");let h=[r[0],t[0]];return pd.computeShapeHelper(!1,r,h,n,o,i,u,d),h}static computeShapeHelper(r,t,n,o,i,u,d,h){if(r)for(let f=0;f<t.length-2;f++)n.push(1);else for(let f=0;f<t.length-2;f++)n.push(pd.adjustPadAndReturnShape(t[f+2],o[f],i[f],u[f],d,f,f+t.length-2,h))}static adjustPadAndReturnShape(r,t,n,o,i,u,d,h){let f=n*(o-1)+1;if(h&&h!=="NOTSET")switch(h){case"VALID":return i[u]=0,i[d]=0,Math.floor((r-f)/t+1);case"SAME_LOWER":case"SAME_UPPER":if(n!==1)throw new Error("Dilation not supported for SAME_UPPER or SAME_LOWER");{let y=((r+t-1)/t-1)*t+o-r;return i[u]=Math.floor(h==="SAME_LOWER"?(y+1)/2:y/2),i[d]=y-i[u],Math.floor((r+y-o)/t+1)}default:throw new Error("Unsupported AutoPad type")}else return Math.floor((r+i[u]+i[d]-f)/t+1)}},R0=class{static getShapeOfGemmResult(e,r,t,n,o){if(e.length!==2||t.length!==2)throw new Error("shape need to be of size 2");let i,u,d;r?(i=e[1],u=e[0]):(i=e[0],u=e[1]);let h=-1;if(n?(d=t[0],h=1):(d=t[1],h=0),t[h]!==u)throw new Error("dimension mismatch");if(i<=0||d<=0||u<=0)throw new Error("invalid shape specified");if(o&&!Ua.isValidBroadcast(o,[i,d]))throw new Error("gemm: invalid bias shape for broadcast");return[i,d,u]}},N0=-34028234663852886e22,V0=34028234663852886e22}),s_,U0=Je(()=>{St(),s_=(e,r)=>new(cf(r))(e)}),Vf,Om,Uf,K_,Wf,H_,Gf,Kf,Hf,q_,W0,sT=Je(()=>{St(),vs(),Vf=new Map([["float32",32],["float16",16],["int32",32],["uint32",32],["int64",64],["uint64",64],["int8",8],["uint8",8],["int4",4],["uint4",4]]),Om=(e,r)=>{if(r==="int32")return e;let t=Vf.get(r);if(!t)throw new Error(`WebNN backend does not support data type: ${r}`);let n=t/8;if(e.byteLength%n!==0)throw new Error(`Invalid Uint8Array length - must be a multiple of ${n}.`);let o=e.byteLength/n,i=new(cf(r))(e.buffer,e.byteOffset,o);switch(r){case"int64":case"uint64":{let u=new Int32Array(o);for(let d=0;d<o;d++){let h=i[d];if(h>2147483647n||h<-2147483648n)throw new Error("Can not convert int64 data to int32 - value out of range.");u[d]=Number(h)}return new Uint8Array(u.buffer)}case"int8":case"uint8":case"uint32":{if(r==="uint32"&&i.some(d=>d>2147483647))throw new Error("Can not convert uint32 data to int32 - value out of range.");let u=Int32Array.from(i,Number);return new Uint8Array(u.buffer)}default:throw new Error(`Unsupported data conversion from ${r} to 'int32'`)}},Uf=(e,r)=>{if(r==="int32")return e;if(e.byteLength%4!==0)throw new Error("Invalid Uint8Array length - must be a multiple of 4 (int32).");let t=e.byteLength/4,n=new Int32Array(e.buffer,e.byteOffset,t);switch(r){case"int64":{let o=BigInt64Array.from(n,BigInt);return new Uint8Array(o.buffer)}case"uint64":{if(n.some(i=>i<0))throw new Error("Can not convert int32 data to uin64 - negative value found.");let o=BigUint64Array.from(n,BigInt);return new Uint8Array(o.buffer)}case"int8":{if(n.some(i=>i<-128||i>127))throw new Error("Can not convert int32 data to int8 - value out of range.");let o=Int8Array.from(n,Number);return new Uint8Array(o.buffer)}case"uint8":{if(n.some(o=>o<0||o>255))throw new Error("Can not convert int32 data to uint8 - value out of range.");return Uint8Array.from(n,Number)}case"uint32":{if(n.some(i=>i<0))throw new Error("Can not convert int32 data to uint32 - negative value found.");let o=Uint32Array.from(n,Number);return new Uint8Array(o.buffer)}default:throw new Error(`Unsupported data conversion from 'int32' to ${r}`)}},K_=1,Wf=()=>K_++,H_=new Map([["int8","int32"],["uint8","int32"],["uint32","int32"],["int64","int32"]]),Gf=(e,r)=>{let t=Vf.get(e);if(!t)throw new Error(`WebNN backend does not support data type: ${e}`);return r.length>0?Math.ceil(r.reduce((n,o)=>n*o)*t/8):0},Kf=class{constructor(e){this.isDataConverted=!1;let{sessionId:r,context:t,tensor:n,dataType:o,shape:i,fallbackDataType:u}=e;this.sessionId=r,this.mlContext=t,this.mlTensor=n,this.dataType=o,this.tensorShape=i,this.fallbackDataType=u}get tensor(){return this.mlTensor}get type(){return this.dataType}get fallbackType(){return this.fallbackDataType}get shape(){return this.tensorShape}get byteLength(){return Gf(this.dataType,this.tensorShape)}destroy(){Gt("verbose",()=>"[WebNN] TensorWrapper.destroy"),this.mlTensor.destroy()}write(e){this.mlContext.writeTensor(this.mlTensor,e)}async read(e){if(this.fallbackDataType){let r=await this.mlContext.readTensor(this.mlTensor),t=Uf(new Uint8Array(r),this.dataType);if(e){(e instanceof ArrayBuffer?new Uint8Array(e):new Uint8Array(e.buffer,e.byteOffset,e.byteLength)).set(t);return}else return t.buffer}else return e?this.mlContext.readTensor(this.mlTensor,e):this.mlContext.readTensor(this.mlTensor)}canReuseTensor(e,r,t){return this.mlContext===e&&this.dataType===r&&this.tensorShape.length===t.length&&this.tensorShape.every((n,o)=>n===t[o])}setIsDataConverted(e){this.isDataConverted=e}},Hf=class{constructor(e,r){this.tensorManager=e,this.wrapper=r}get tensorWrapper(){return this.wrapper}releaseTensor(){this.tensorWrapper&&(this.tensorManager.releaseTensor(this.tensorWrapper),this.wrapper=void 0)}async ensureTensor(e,r,t,n){let o=this.tensorManager.getMLContext(e),i;if(!o.opSupportLimits().input.dataTypes.includes(r)){if(i=H_.get(r),!i||!o.opSupportLimits().input.dataTypes.includes(i))throw new Error(`WebNN backend does not support data type: ${r}`);Gt("verbose",()=>`[WebNN] TensorIdTracker.ensureTensor: fallback dataType from ${r} to ${i}`)}if(this.wrapper){if(this.wrapper.canReuseTensor(o,r,t))return this.wrapper.tensor;if(n){if(this.wrapper.byteLength!==Gf(r,t))throw new Error("Unable to copy data to tensor with different size.");this.activeUpload=new Uint8Array(await this.wrapper.read())}this.tensorManager.releaseTensor(this.wrapper)}let u=typeof MLTensorUsage>"u"?void 0:MLTensorUsage.READ|MLTensorUsage.WRITE;return this.wrapper=await this.tensorManager.getCachedTensor(e,r,t,u,!0,!0,i),n&&this.activeUpload&&(this.wrapper.write(this.activeUpload),this.activeUpload=void 0),this.wrapper.tensor}upload(e){let r=e;if(this.wrapper){if(this.wrapper.fallbackType)if(this.wrapper.fallbackType==="int32")r=Om(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(r);return}else Gt("verbose",()=>"Data size does not match tensor size. Releasing tensor."),this.releaseTensor()}this.activeUpload?this.activeUpload.set(r):this.activeUpload=new Uint8Array(r)}async download(e){var r,t;if(this.activeUpload){let n=(r=this.wrapper)!=null&&r.isDataConverted?Uf(this.activeUpload,(t=this.wrapper)==null?void 0:t.type):this.activeUpload;if(e){e instanceof ArrayBuffer?new Uint8Array(e).set(n):new Uint8Array(e.buffer,e.byteOffset,e.byteLength).set(n);return}else return n.buffer}if(!this.wrapper)throw new Error("Tensor has not been created.");return e?this.wrapper.read(e):this.wrapper.read()}},q_=class{constructor(e){this.backend=e,this.tensorTrackersById=new Map,this.freeTensors=[],this.externalTensors=new Set}getMLContext(e){let r=this.backend.getMLContext(e);if(!r)throw new Error("MLContext not found for session.");return r}reserveTensorId(){let e=Wf();return this.tensorTrackersById.set(e,new Hf(this)),e}releaseTensorId(e){let r=this.tensorTrackersById.get(e);r&&(this.tensorTrackersById.delete(e),r.tensorWrapper&&this.releaseTensor(r.tensorWrapper))}async ensureTensor(e,r,t,n,o){Gt("verbose",()=>`[WebNN] TensorManager.ensureTensor {tensorId: ${r}, dataType: ${t}, shape: ${n}, copyOld: ${o}}`);let i=this.tensorTrackersById.get(r);if(!i)throw new Error("Tensor not found.");return i.ensureTensor(e,t,n,o)}upload(e,r){let t=this.tensorTrackersById.get(e);if(!t)throw new Error("Tensor not found.");t.upload(r)}async download(e,r){Gt("verbose",()=>`[WebNN] TensorManager.download {tensorId: ${e}, dstBuffer: ${r==null?void 0:r.byteLength}}`);let t=this.tensorTrackersById.get(e);if(!t)throw new Error("Tensor not found.");return t.download(r)}releaseTensorsForSession(e){for(let r of this.freeTensors)r.sessionId===e&&r.destroy();this.freeTensors=this.freeTensors.filter(r=>r.sessionId!==e)}registerTensor(e,r,t,n){let o=this.getMLContext(e),i=Wf(),u=new Kf({sessionId:e,context:o,tensor:r,dataType:t,shape:n});return this.tensorTrackersById.set(i,new Hf(this,u)),this.externalTensors.add(u),i}async getCachedTensor(e,r,t,n,o,i,u){let d=this.getMLContext(e);for(let[f,y]of this.freeTensors.entries())if(y.canReuseTensor(d,r,t)){Gt("verbose",()=>`[WebNN] Reusing tensor {dataType: ${r}, ${u?`fallbackDataType: ${u},`:""} shape: ${t}`);let m=this.freeTensors.splice(f,1)[0];return m.sessionId=e,m}Gt("verbose",()=>`[WebNN] MLContext.createTensor {dataType: ${r}, ${u?`fallbackDataType: ${u},`:""} shape: ${t}}`);let h=await d.createTensor({dataType:u??r,shape:t,dimensions:t,usage:n,writable:o,readable:i});return new Kf({sessionId:e,context:d,tensor:h,dataType:r,shape:t,fallbackDataType:u})}releaseTensor(e){this.externalTensors.has(e)&&this.externalTensors.delete(e),this.freeTensors.push(e)}},W0=(...e)=>new q_(...e)}),id,Q_,G0,nT=Je(()=>{St(),Pn(),U0(),sT(),vs(),id=new Map([[1,"float32"],[10,"float16"],[6,"int32"],[12,"uint32"],[7,"int64"],[13,"uint64"],[22,"int4"],[21,"uint4"],[3,"int8"],[2,"uint8"],[9,"uint8"]]),Q_=(e,r)=>{if(e===r)return!0;if(e===void 0||r===void 0)return!1;let t=Object.keys(e).sort(),n=Object.keys(r).sort();return t.length===n.length&&t.every((o,i)=>o===n[i]&&e[o]===r[o])},G0=class{constructor(e){this.tensorManager=W0(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,i_(e.logLevel,!!e.debug)}get currentSessionId(){if(this.activeSessionId===void 0)throw new Error("No active session");return this.activeSessionId}onRunStart(e){Gt("verbose",()=>`[WebNN] onRunStart {sessionId: ${e}}`),this.activeSessionId=e}onRunEnd(e){Gt("verbose",()=>`[WebNN] onRunEnd {sessionId: ${e}}`);let r=this.temporarySessionTensorIds.get(e);if(r){for(let t of r)Gt("verbose",()=>`[WebNN] releasing temporary tensor {tensorId: ${t}}`),this.tensorManager.releaseTensorId(t);this.temporarySessionTensorIds.delete(e),this.activeSessionId=void 0}}async createMLContext(e){if(e instanceof GPUDevice){let t=this.mlContextCache.findIndex(n=>n.gpuDevice===e);if(t!==-1)return this.mlContextCache[t].mlContext;{let n=await navigator.ml.createContext(e);return this.mlContextCache.push({gpuDevice:e,mlContext:n}),n}}else if(e===void 0){let t=this.mlContextCache.findIndex(n=>n.options===void 0&&n.gpuDevice===void 0);if(t!==-1)return this.mlContextCache[t].mlContext;{let n=await navigator.ml.createContext();return this.mlContextCache.push({mlContext:n}),n}}let r=this.mlContextCache.findIndex(t=>Q_(t.options,e));if(r!==-1)return this.mlContextCache[r].mlContext;{let t=await navigator.ml.createContext(e);return this.mlContextCache.push({options:e,mlContext:t}),t}}registerMLContext(e,r){this.mlContextBySessionId.set(e,r);let t=this.sessionIdsByMLContext.get(r);t||(t=new Set,this.sessionIdsByMLContext.set(r,t)),t.add(e),this.temporaryGraphInputs.length>0&&(this.sessionGraphInputs.set(e,this.temporaryGraphInputs),this.temporaryGraphInputs=[]),this.temporaryGraphOutputs.length>0&&(this.sessionGraphOutputs.set(e,this.temporaryGraphOutputs),this.temporaryGraphOutputs=[])}onReleaseSession(e){this.sessionGraphInputs.delete(e),this.sessionGraphOutputs.delete(e);let r=this.mlContextBySessionId.get(e);if(!r)return;this.tensorManager.releaseTensorsForSession(e),this.mlContextBySessionId.delete(e);let t=this.sessionIdsByMLContext.get(r);if(t.delete(e),t.size===0){this.sessionIdsByMLContext.delete(r);let n=this.mlContextCache.findIndex(o=>o.mlContext===r);n!==-1&&this.mlContextCache.splice(n,1)}}getMLContext(e){return this.mlContextBySessionId.get(e)}reserveTensorId(){return this.tensorManager.reserveTensorId()}releaseTensorId(e){Gt("verbose",()=>`[WebNN] releaseTensorId {tensorId: ${e}}`),this.tensorManager.releaseTensorId(e)}async ensureTensor(e,r,t,n,o){let i=id.get(t);if(!i)throw new Error(`Unsupported ONNX data type: ${t}`);return this.tensorManager.ensureTensor(e??this.currentSessionId,r,i,n,o)}async createTemporaryTensor(e,r,t){Gt("verbose",()=>`[WebNN] createTemporaryTensor {onnxDataType: ${r}, shape: ${t}}`);let n=id.get(r);if(!n)throw new Error(`Unsupported ONNX data type: ${r}`);let o=this.tensorManager.reserveTensorId();await this.tensorManager.ensureTensor(e,o,n,t,!1);let i=this.temporarySessionTensorIds.get(e);return i?i.push(o):this.temporarySessionTensorIds.set(e,[o]),o}uploadTensor(e,r){if(!lr().shouldTransferToMLTensor)throw new Error("Trying to upload to a MLTensor while shouldTransferToMLTensor is false");Gt("verbose",()=>`[WebNN] uploadTensor {tensorId: ${e}, data: ${r.byteLength}}`),this.tensorManager.upload(e,r)}async downloadTensor(e,r){return this.tensorManager.download(e,r)}createMLTensorDownloader(e,r){return async()=>{let t=await this.tensorManager.download(e);return s_(t,r)}}registerMLTensor(e,r,t,n){let o=id.get(t);if(!o)throw new Error(`Unsupported ONNX data type: ${t}`);let i=this.tensorManager.registerTensor(e,r,o,n);return Gt("verbose",()=>`[WebNN] registerMLTensor {tensor: ${r}, dataType: ${o}, dimensions: ${n}} -> {tensorId: ${i}}`),i}registerMLConstant(e,r,t,n,o,i,u=!1){if(!i)throw new Error("External mounted files are not available.");let d=e;e.startsWith("./")&&(d=e.substring(2));let h=i.get(d);if(!h)throw new Error(`File with name ${d} not found in preloaded files.`);if(r+t>h.byteLength)throw new Error("Out of bounds: data offset and length exceed the external file data size.");let f=h.slice(r,r+t).buffer,y;switch(o.dataType){case"float32":y=new Float32Array(f);break;case"float16":y=typeof Float16Array<"u"&&Float16Array.from?new Float16Array(f):new Uint16Array(f);break;case"int32":y=new Int32Array(f);break;case"uint32":y=new Uint32Array(f);break;case"int64":if(u){let m=Om(new Uint8Array(f),"int64");y=new Int32Array(m.buffer),o.dataType="int32"}else y=new BigInt64Array(f);break;case"uint64":y=new BigUint64Array(f);break;case"int8":y=new Int8Array(f);break;case"int4":case"uint4":case"uint8":y=new Uint8Array(f);break;default:throw new Error(`Unsupported data type: ${o.dataType} in creating WebNN Constant from external data.`)}return Gt("verbose",()=>`[WebNN] registerMLConstant {dataType: ${o.dataType}, shape: ${o.shape}}} ${u?"(Note: it was int64 data type and registered to int32 as workaround)":""}`),n.constant(o,y)}registerGraphInput(e){this.temporaryGraphInputs.push(e)}registerGraphOutput(e){this.temporaryGraphOutputs.push(e)}isGraphInput(e,r){let t=this.sessionGraphInputs.get(e);return t?t.includes(r):!1}isGraphOutput(e,r){let t=this.sessionGraphOutputs.get(e);return t?t.includes(r):!1}isGraphInputOutputTypeSupported(e,r,t=!0){let n=this.mlContextBySessionId.get(e),o=id.get(Mn(r));return typeof o>"u"?!1:t?!!(n!=null&&n.opSupportLimits().input.dataTypes.includes(o)):!!(n!=null&&n.opSupportLimits().output.dataTypes.includes(o))}flush(){}}}),n_=Je(()=>{}),qf,Bh,Rh,X_,J_,Qf,Fm,Y_,K0,aT=Je(()=>{vs(),n_(),qf=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]]),Bh=[],Rh=e=>Math.ceil(Number(e)/16)*16,X_=e=>{for(let r=0;r<Bh.length;r++){let t=Bh[r];if(e<=t)return t}return Math.ceil(e/16)*16},J_=1,Qf=()=>J_++,Fm=async(e,r,t,n)=>{let o=Rh(t),i=e.device.createBuffer({size:o,usage:GPUBufferUsage.COPY_DST|GPUBufferUsage.MAP_READ});try{let u=e.getCommandEncoder();e.endComputePass(),u.copyBufferToBuffer(r,0,i,0,o),e.flush(),await i.mapAsync(GPUMapMode.READ);let d=i.getMappedRange();if(n){let h=n();return h.set(new Uint8Array(d,0,t)),h}else return new Uint8Array(d.slice(0,t))}finally{i.destroy()}},Y_=class{constructor(e){this.backend=e,this.storageCache=new Map,this.freeBuffers=new Map,this.freeUniformBuffers=new Map,this.buffersPending=[],this.capturedPendingBuffers=new Map;for(let[r]of qf)Bh.push(r),this.freeBuffers.set(r,[]),this.freeUniformBuffers.set(r,[]);this.sessionCount=0}upload(e,r){let t=r.buffer,n=r.byteOffset,o=r.byteLength,i=Rh(o),u=this.storageCache.get(e);if(!u)throw new Error("gpu data for uploading does not exist");if(Number(u.originalSize)!==o)throw new Error(`inconsistent data size. gpu data size=${u.originalSize}, data size=${o}`);let d=this.backend.device.createBuffer({mappedAtCreation:!0,size:i,usage:GPUBufferUsage.MAP_WRITE|GPUBufferUsage.COPY_SRC}),h=d.getMappedRange();new Uint8Array(h).set(new Uint8Array(t,n,o)),d.unmap();let f=this.backend.device.createCommandEncoder();f.copyBufferToBuffer(d,0,u.gpuData.buffer,0,i),this.backend.device.queue.submit([f.finish()]),d.destroy(),Gt("verbose",()=>`[WebGPU] GpuDataManager.upload(id=${e})`)}memcpy(e,r){let t=this.storageCache.get(e);if(!t)throw new Error("source gpu data for memcpy does not exist");let n=this.storageCache.get(r);if(!n)throw new Error("destination gpu data for memcpy does not exist");if(t.originalSize!==n.originalSize)throw new Error("inconsistent source and destination gpu data size");let o=Rh(t.originalSize),i=this.backend.getCommandEncoder();this.backend.endComputePass(),i.copyBufferToBuffer(t.gpuData.buffer,0,n.gpuData.buffer,0,o)}registerExternalBuffer(e,r,t){let n;if(t){if(n=t[0],e===t[1])return Gt("verbose",()=>`[WebGPU] GpuDataManager.registerExternalBuffer(size=${r}) => id=${n}, buffer is the same, skip.`),n;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 n=Qf();return this.storageCache.set(n,{gpuData:{id:n,type:0,buffer:e},originalSize:r}),Gt("verbose",()=>`[WebGPU] GpuDataManager.registerExternalBuffer(size=${r}) => id=${n}, registered.`),n}unregisterExternalBuffer(e){e!==void 0&&(this.storageCache.delete(e),Gt("verbose",()=>`[WebGPU] GpuDataManager.unregisterExternalBuffer() => id=${e}`))}create(e,r=GPUBufferUsage.STORAGE|GPUBufferUsage.COPY_SRC|GPUBufferUsage.COPY_DST){let t=X_(e),n,o=(r&GPUBufferUsage.STORAGE)===GPUBufferUsage.STORAGE,i=(r&GPUBufferUsage.UNIFORM)===GPUBufferUsage.UNIFORM;if(o||i){let d=(o?this.freeBuffers:this.freeUniformBuffers).get(t);d?d.length>0?n=d.pop():n=this.backend.device.createBuffer({size:t,usage:r}):n=this.backend.device.createBuffer({size:t,usage:r})}else n=this.backend.device.createBuffer({size:t,usage:r});let u={id:Qf(),type:0,buffer:n};return this.storageCache.set(u.id,{gpuData:u,originalSize:Number(e)}),Gt("verbose",()=>`[WebGPU] GpuDataManager.create(size=${e}) => id=${u.id}`),u}get(e){var r;return(r=this.storageCache.get(e))==null?void 0:r.gpuData}release(e){let r=typeof e=="bigint"?Number(e):e,t=this.storageCache.get(r);if(!t){if(this.storageCache.size===0)return 0;throw new Error("releasing data does not exist")}return Gt("verbose",()=>`[WebGPU] GpuDataManager.release(id=${r}), gpuDataId=${t.gpuData.id}`),this.storageCache.delete(r),this.buffersPending.push(t.gpuData.buffer),t.originalSize}async download(e,r){let t=this.storageCache.get(Number(e));if(!t)throw new Error("data does not exist");await Fm(this.backend,t.gpuData.buffer,t.originalSize,r)}refreshPendingBuffers(){if(this.buffersPending.length!==0)if(this.backend.sessionStatus==="default"){for(let e of this.buffersPending){let r=qf.get(e.size);if((e.usage&GPUBufferUsage.STORAGE)===GPUBufferUsage.STORAGE){let t=this.freeBuffers.get(e.size)||[];r===void 0||t.length>=r?e.destroy():t.push(e)}else if((e.usage&GPUBufferUsage.UNIFORM)===GPUBufferUsage.UNIFORM){let t=this.freeUniformBuffers.get(e.size)||[];r===void 0||t.length>=r?e.destroy():t.push(e)}else e.destroy()}this.buffersPending=[]}else{let e=this.capturedPendingBuffers.get(this.backend.currentSessionId);e||(e=[],this.capturedPendingBuffers.set(this.backend.currentSessionId,e));for(let r of this.buffersPending)e.push(r);this.buffersPending=[]}}dispose(){this.freeBuffers.forEach(e=>{e.forEach(r=>{r.destroy()})}),this.freeUniformBuffers.forEach(e=>{e.forEach(r=>{r.destroy()})}),this.storageCache.forEach(e=>{e.gpuData.buffer.destroy()}),this.capturedPendingBuffers.forEach(e=>{e.forEach(r=>{r.destroy()})}),this.storageCache=new Map,this.freeBuffers=new Map,this.freeUniformBuffers=new Map,this.capturedPendingBuffers=new Map}onCreateSession(){this.sessionCount+=1}onReleaseSession(e){let r=this.capturedPendingBuffers.get(e);r&&(r.forEach(t=>{t.destroy()}),this.capturedPendingBuffers.delete(e)),this.sessionCount-=1,this.sessionCount===0&&(Gt("warning",()=>"[WebGPU] Clearing webgpu buffer cache"),this.storageCache.forEach(t=>{t.gpuData.buffer.destroy()}),this.storageCache=new Map)}},K0=(...e)=>new Y_(...e)}),Z_,Zt,vr=Je(()=>{Z_=class{constructor(e){Object.assign(this,e)}get cacheKey(){return this.key||(this.key=Object.getOwnPropertyNames(this).sort().map(e=>`${this[e]}`).join(";")),this.key}},Zt=e=>new Z_(e)}),Wa,Nh,jr,ti,bt,yr,Dm,Na,Rs,wt,sd,Re,mt,H0,a_,eg,q0,Lt=Je(()=>{St(),Ft(),Wa=64,Nh=(e,r)=>{if(r===3)throw new Error("vec3 has same alignment as vec4, use vec4 instead");switch(Number(e)){case 10:return r>1?`vec${r}<f16>`:"f16";case 1:return r>1?`vec${r}<f32>`:"f32";case 6:return r>1?`vec${r}<i32>`:"i32";case 12:return r>1?`vec${r}<u32>`:"u32";case 7:if(r>1)throw new Error("currently not supported vecX of uint64 yet");return["vec2<u32>","i32"];case 13:if(r>1)throw new Error("currently not supported vecX of uint64 yet");return["vec2<u32>","u32"];case 9:if(r!==4)throw new Error("bool must be vec4");return["u32","vec4<bool>"];case 22:return"i32";case 21:return"u32";default:throw new Error(`Unknown data type: ${e}`)}},jr=(e,r=1)=>{let t=Nh(e,r);return typeof t=="string"?t:t[0]},ti=(e,r=1)=>{let t=Nh(e,r);return typeof t=="string"?t:t[1]},bt=(...e)=>{let r=[];return e.forEach(t=>{t.length!==0&&r.push({type:12,data:t},{type:12,data:Fe.computeStrides(t)})}),r},yr=e=>e%4===0?4:e%2===0?2:1,Dm=(e="f32",r,t="0")=>!r||r===1?`${e}(${t})`:`vec${r}<${e}>(${t})`,Na=(e,r,t)=>e==="f32"?t:r===1?`f32(${t})`:`vec${r}<f32>(${t})`,Rs=(e,r)=>r===4?`(${e}.x + ${e}.y + ${e}.z + ${e}.w)`:r===2?`(${e}.x + ${e}.y)`:r===3?`(${e}.x + ${e}.y + ${e}.z)`:e,wt=(e,r,t,n)=>e.startsWith("uniforms.")&&t>4?typeof r=="string"?n==="f16"?`${e}[(${r}) / 8][(${r}) % 8 / 4][(${r}) % 8 % 4]`:`${e}[(${r}) / 4][(${r}) % 4]`:n==="f16"?`${e}[${Math.floor(r/8)}][${Math.floor(r%8/4)}][${r%8%4}]`:`${e}[${Math.floor(r/4)}][${r%4}]`:t>1?`${e}[${r}]`:e,sd=(e,r,t,n,o)=>{let i=typeof t=="number",u=i?t:t.length,d=[...new Array(u).keys()],h=u<2?"u32":u<=4?`vec${u}<u32>`:`array<u32, ${u}>`,f=Nh(r,o),y=typeof f=="string"?f:f[1],m=typeof f=="string"?f:f[0],g={indices:h,value:y,storage:m,tensor:r},v=q=>typeof q=="string"?q:`${q}u`,b={offsetToIndices:!1,indicesToOffset:!1,broadcastedIndicesToOffset:!1,set:!1,setByIndices:!1,get:!1,getByIndices:!1},k=i?"uniforms.":"",z=`${k}${e}_shape`,E=`${k}${e}_strides`,T="";for(let q=0;q<u-1;q++)T+=`
    let dim${q} = current / ${wt(E,q,u)};
    let rest${q} = current % ${wt(E,q,u)};
    indices[${q}] = dim${q};
    current = rest${q};
    `;T+=`indices[${u-1}] = current;`;let D=u<2?"":`
  fn o2i_${e}(offset: u32) -> ${g.indices} {
    var indices: ${g.indices};
    var current = offset;
    ${T}
    return indices;
  }`,I=q=>(b.offsetToIndices=!0,u<2?q:`o2i_${e}(${q})`),$=[];if(u>=2)for(let q=u-1;q>=0;q--)$.push(`${wt(E,q,u)} * (indices[${q}])`);let P=u<2?"":`
  fn i2o_${e}(indices: ${g.indices}) -> u32 {
    return ${$.join("+")};
  }`,S=q=>(b.indicesToOffset=!0,u<2?q:`i2o_${e}(${q})`),O=(...q)=>u===0?"0u":`${g.indices}(${q.map(v).join(",")})`,R=(q,B)=>u<2?`${q}`:`${wt(q,B,u)}`,G=(q,B,Q)=>u<2?`${q}=${Q};`:`${wt(q,B,u)}=${Q};`,ee={},ie=(q,B)=>{b.broadcastedIndicesToOffset=!0;let Q=`${B.name}broadcastedIndicesTo${e}Offset`;if(Q in ee)return`${Q}(${q})`;let oe=[];for(let Ie=u-1;Ie>=0;Ie--){let ye=B.indicesGet("outputIndices",Ie+B.rank-u);oe.push(`${R(E,Ie)} * (${ye} % ${R(z,Ie)})`)}return ee[Q]=`fn ${Q}(outputIndices: ${B.type.indices}) -> u32 {
             return ${oe.length>0?oe.join("+"):"0u"};
           }`,`${Q}(${q})`},H=(q,B)=>(()=>{if(g.storage===g.value)return`${e}[${q}]=${B};`;if(g.storage==="vec2<u32>"&&g.value==="i32")return`${e}[${q}]=vec2<u32>(u32(${B}), select(0u, 0xFFFFFFFFu, ${B} < 0));`;if(g.storage==="vec2<u32>"&&g.value==="u32")return`${e}[${q}]=vec2<u32>(u32(${B}), 0u);`;if(g.storage==="u32"&&g.value==="vec4<bool>")return`${e}[${q}]=dot(vec4<u32>(0x1, 0x100, 0x10000, 0x1000000), vec4<u32>(${B}));`;throw new Error(`not supported combination of storage type ${g.storage} and value type ${g.value} yet`)})(),ce=q=>(()=>{if(g.storage===g.value)return`${e}[${q}]`;if(g.storage==="vec2<u32>"&&g.value==="i32")return`i32(${e}[${q}].x)`;if(g.storage==="vec2<u32>"&&g.value==="u32")return`u32(${e}[${q}].x)`;if(g.storage==="u32"&&g.value==="vec4<bool>")return`vec4<bool>(bool(${e}[${q}] & 0xFFu), bool(${e}[${q}] & 0xFF00u), bool(${e}[${q}] & 0xFF0000u), bool(${e}[${q}] & 0xFF000000u))`;throw new Error(`not supported combination of storage type ${g.storage} and value type ${g.value} yet`)})(),re=u<2?"":`
  fn get_${e}ByIndices(indices: ${g.indices}) -> ${y} {
    return ${ce(`i2o_${e}(indices)`)};
  }`,se=u<2?"":(()=>{let q=d.map(Q=>`d${Q}: u32`).join(", "),B=d.map(Q=>`d${Q}`).join(", ");return`
  fn get_${e}(${q}) -> ${y} {
    return get_${e}ByIndices(${O(B)});
  }`})(),_e=(...q)=>{if(q.length!==u)throw new Error(`indices length must be ${u}`);let B=q.map(v).join(",");return u===0?ce("0u"):u===1?ce(B[0]):(b.get=!0,b.getByIndices=!0,b.indicesToOffset=!0,`get_${e}(${B})`)},ae=q=>u<2?ce(q):(b.getByIndices=!0,b.indicesToOffset=!0,`get_${e}ByIndices(${q})`),Ce=u<2?"":`
  fn set_${e}ByIndices(indices: ${g.indices}, value: ${y}) {
    ${H(`i2o_${e}(indices)`,"value")}
  }`,Te=u<2?"":(()=>{let q=d.map(Q=>`d${Q}: u32`).join(", "),B=d.map(Q=>`d${Q}`).join(", ");return`
  fn set_${e}(${q}, value: ${y}) {
    set_${e}ByIndices(${O(B)}, value);
  }`})();return{impl:()=>{let q=[],B=!1;return b.offsetToIndices&&(q.push(D),B=!0),b.indicesToOffset&&(q.push(P),B=!0),b.broadcastedIndicesToOffset&&(Object.values(ee).forEach(Q=>q.push(Q)),B=!0),b.set&&(q.push(Te),B=!0),b.setByIndices&&(q.push(Ce),B=!0),b.get&&(q.push(se),B=!0),b.getByIndices&&(q.push(re),B=!0),!i&&B&&q.unshift(`const ${z} = ${g.indices}(${t.join(",")});`,`const ${E} = ${g.indices}(${Fe.computeStrides(t).join(",")});`),q.join(`
`)},type:g,offsetToIndices:I,indicesToOffset:S,broadcastedIndicesToOffset:ie,indices:O,indicesGet:R,indicesSet:G,set:(...q)=>{if(q.length!==u+1)throw new Error(`indices length must be ${u}`);let B=q[u];if(typeof B!="string")throw new Error("value must be string");let Q=q.slice(0,u).map(v).join(",");return u===0?H("0u",B):u===1?H(Q[0],B):(b.set=!0,b.setByIndices=!0,b.indicesToOffset=!0,`set_${e}(${Q}, ${B})`)},setByOffset:H,setByIndices:(q,B)=>u<2?H(q,B):(b.setByIndices=!0,b.indicesToOffset=!0,`set_${e}ByIndices(${q}, ${B});`),get:_e,getByOffset:ce,getByIndices:ae,usage:n,name:e,strides:E,shape:z,rank:u}},Re=(e,r,t,n=1)=>sd(e,r,t,"input",n),mt=(e,r,t,n=1)=>sd(e,r,t,"output",n),H0=(e,r,t)=>sd(e,r,t,"atomicOutput",1),a_=(e,r,t,n=1)=>sd(e,r,t,"internal",n),eg=class{constructor(e,r){this.normalizedDispatchGroup=e,this.limits=r,this.internalVariables=[],this.variables=[],this.uniforms=[],this.variableIndex=0}guardAgainstOutOfBoundsWorkgroupSizes(e){return`if (global_idx >= ${typeof e=="number"?`${e}u`:e}) { return; }`}mainStart(e=Wa){let r=typeof e=="number"?e:e[0],t=typeof e=="number"?1:e[1],n=typeof e=="number"?1:e[2];if(r>this.limits.maxComputeWorkgroupSizeX||t>this.limits.maxComputeWorkgroupSizeY||n>this.limits.maxComputeWorkgroupSizeZ)throw new Error(`workgroup size [${r}, ${t}, ${n}] exceeds the maximum workgroup size [${this.limits.maxComputeWorkgroupSizeX}, ${this.limits.maxComputeWorkgroupSizeY}, ${this.limits.maxComputeWorkgroupSizeZ}].`);if(r*t*n>this.limits.maxComputeInvocationsPerWorkgroup)throw new Error(`workgroup size [${r}, ${t}, ${n}] exceeds the maximum workgroup invocations ${this.limits.maxComputeInvocationsPerWorkgroup}.`);let o=this.normalizedDispatchGroup[1]===1&&this.normalizedDispatchGroup[2]===1,i=o?`@builtin(global_invocation_id) global_id : vec3<u32>,
    @builtin(workgroup_id) workgroup_id : vec3<u32>,
    @builtin(local_invocation_index) local_idx : u32,
    @builtin(local_invocation_id) local_id : vec3<u32>`:`@builtin(global_invocation_id) global_id : vec3<u32>,
                                             @builtin(local_invocation_id) local_id : vec3<u32>,
    @builtin(local_invocation_index) local_idx : u32,
    @builtin(workgroup_id) workgroup_id : vec3<u32>,
    @builtin(num_workgroups) num_workgroups : vec3<u32>`,u=o?`let global_idx = global_id.x;
         let workgroup_index = workgroup_id.x;`:`let workgroup_index = workgroup_id.z * num_workgroups[0] * num_workgroups[1] +
             workgroup_id.y * num_workgroups[0] + workgroup_id.x;
         let global_idx = workgroup_index * ${r*t*n}u + local_idx;`;return`@compute @workgroup_size(${r}, ${t}, ${n})
  fn main(${i}) {
    ${u}
  `}appendVariableUniforms(e){e.rank!==0&&(e.shape.startsWith("uniforms.")&&this.uniforms.push({name:e.shape.replace("uniforms.",""),type:"u32",length:e.rank}),e.strides.startsWith("uniforms.")&&this.uniforms.push({name:e.strides.replace("uniforms.",""),type:"u32",length:e.rank}))}declareVariable(e,r){if(e.usage==="internal")throw new Error("cannot use internal variable with declareVariable(). use registerInternalVariables() instead.");this.variables.push(e),this.appendVariableUniforms(e);let t=e.usage==="input"?"read":"read_write",n=e.usage==="atomicOutput"?"atomic<i32>":e.type.storage;return`@group(0) @binding(${r}) var<storage, ${t}> ${e.name}: array<${n}>;`}declareVariables(...e){return e.map(r=>this.declareVariable(r,this.variableIndex++)).join(`
`)}registerInternalVariable(e){if(e.usage!=="internal")throw new Error("cannot use input or output variable with registerInternalVariable(). use declareVariables() instead.");this.internalVariables.push(e),this.appendVariableUniforms(e)}registerInternalVariables(...e){return e.forEach(r=>this.registerInternalVariable(r)),this}registerUniform(e,r,t=1){return this.uniforms.push({name:e,type:r,length:t}),this}registerUniforms(e){return this.uniforms=this.uniforms.concat(e),this}uniformDeclaration(){if(this.uniforms.length===0)return"";let e=[];for(let{name:r,type:t,length:n}of this.uniforms)if(n&&n>4)t==="f16"?e.push(`@align(16) ${r}:array<mat2x4<${t}>, ${Math.ceil(n/8)}>`):e.push(`${r}:array<vec4<${t}>, ${Math.ceil(n/4)}>`);else{let o=n==null||n===1?t:`vec${n}<${t}>`;e.push(`${r}:${o}`)}return`
      struct Uniforms { ${e.join(", ")} };
      @group(0) @binding(${this.variableIndex}) var<uniform> uniforms: Uniforms;`}get additionalImplementations(){return this.uniformDeclaration()+this.variables.map(e=>e.impl()).join(`
`)+this.internalVariables.map(e=>e.impl()).join(`
`)}get variablesInfo(){if(this.uniforms.length===0)return;let e=r=>[12,10,1,6][["u32","f16","f32","i32"].indexOf(r)];return this.uniforms.map(r=>[e(r.type),r.length??1])}},q0=(e,r)=>new eg(e,r)}),tg,Xf,rg,ig,sg,ng,xi,Q0,X0,Ns=Je(()=>{St(),Ft(),vr(),Lt(),tg=(e,r)=>{if(!e||e.length!==1)throw new Error("Transpose requires 1 input.");if(r.length!==0&&r.length!==e[0].dims.length)throw new Error(`perm size ${r.length} does not match input rank ${e[0].dims.length}`)},Xf=(e,r)=>r.length!==0?r:[...new Array(e).keys()].reverse(),rg=(e,r)=>Fe.sortBasedOnPerm(e,Xf(e.length,r)),ig=(e,r,t,n)=>{let o=`fn perm(i: ${n.type.indices}) -> ${t.type.indices} {
    var a: ${t.type.indices};`;for(let i=0;i<r;++i)o+=`a[${e[i]}]=i[${i}];`;return o+="return a;}"},sg=(e,r)=>{let t=[],n=[];for(let o=0;o<e.length;++o)e[o]!==1&&t.push(e[o]),e[r[o]]!==1&&n.push(r[o]);return{newShape:t,newPerm:n}},ng=(e,r)=>{let t=0;for(let n=0;n<e.length;++n)if(r[e[n]]!==1){if(e[n]<t)return!1;t=e[n]}return!0},xi=(e,r)=>{let t=e.dataType,n=e.dims.length,o=Xf(n,r),i=rg(e.dims,o),u=e.dims,d=i,h=n<2||ng(o,e.dims),f;if(h)return f=b=>{let k=Re("input",t,u,4),z=mt("output",t,d,4);return`
  ${b.registerUniform("output_size","u32").declareVariables(k,z)}
  ${b.mainStart()}
    ${b.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.output_size")}
    output[global_idx] = input[global_idx];
  }`},{name:"TransposeCopy",shaderCache:{inputDependencies:["type"]},getRunData:()=>{let b=Fe.size(i);return{outputs:[{dims:i,dataType:e.dataType}],dispatchGroup:{x:Math.ceil(b/64/4)},programUniforms:[{type:12,data:Math.ceil(b/4)}]}},getShaderSource:f};let{newShape:y,newPerm:m}=sg(e.dims,o),g=Fe.areEqual(m,[2,3,1]),v=Fe.areEqual(m,[3,1,2]);if(y.length===2||g||v){u=g?[y[0],y[1]*y[2]]:v?[y[0]*y[1],y[2]]:y,d=[u[1],u[0]];let b=16;return f=k=>{let z=Re("a",t,u.length),E=mt("output",t,d.length);return`
  ${k.registerUniform("output_size","u32").declareVariables(z,E)}
  var<workgroup> tile : array<array<${E.type.value}, ${b+1}>, ${b}>;
  ${k.mainStart([b,b,1])}
    let stride = (uniforms.output_shape[1] - 1) / ${b} + 1;
    let workgroup_id_x = workgroup_index % stride;
    let workgroup_id_y = workgroup_index / stride;
    let input_col = workgroup_id_y * ${b}u + local_id.x;
    let input_row = workgroup_id_x * ${b}u + local_id.y;
    if (input_row < uniforms.a_shape[0] && input_col < uniforms.a_shape[1]) {
      tile[local_id.y][local_id.x] = ${z.getByIndices(`${z.type.indices}(input_row, input_col)`)};
    }
    workgroupBarrier();

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

  ${ig(o,n,k,z)}

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

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

    ${z.setByOffset("global_idx",k.getByIndices("aIndices"))}
  }`},{name:"Transpose",shaderCache:{hint:`${r}`,inputDependencies:["rank"]},getRunData:()=>{let b=Fe.size(i);return{outputs:[{dims:i,dataType:e.dataType}],dispatchGroup:{x:Math.ceil(b/64)},programUniforms:[{type:12,data:b},...bt(u,d)]}},getShaderSource:f}},Q0=(e,r)=>{tg(e.inputs,r.perm),e.compute(xi(e.inputs[0],r.perm))},X0=e=>Zt({perm:e.perm})}),ag,og,lg,cg,ug,dg,pg,hg,fg,mg,Ui,J0,Y0,Z0,eM,tM,rM,iM,sM,nM,aM,oT=Je(()=>{St(),Ft(),Lt(),o_(),Ns(),ag={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"},og={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"},lg={max:"_A[offset]",min:"_A[offset]",mean:"0",sum:"0",prod:"1",sumSquare:"0",logSumExp:"0",l1:"0",l2:"0",logSum:"0"},cg={max:"bestValue",min:"bestValue",sum:"bestValue",prod:"bestValue",sumSquare:"bestValue",logSumExp:"log(bestValue)",l1:"bestValue",l2:"sqrt(bestValue)",logSum:"log(bestValue)"},ug=(e,r)=>{let t=[];for(let n=r-e;n<r;++n)t.push(n);return t},dg=(e,r)=>{let t=[],n=e.length;for(let i=0;i<n;i++)r.indexOf(i)===-1&&t.push(e[i]);let o=r.map(i=>e[i]);return[t,o]},pg=(e,r)=>{let t=e.length+r.length,n=[],o=0;for(let i=0;i<t;i++)r.indexOf(i)===-1?n.push(e[o++]):n.push(1);return n},hg=(e,r)=>{for(let t=0;t<e.length;++t)if(e[e.length-t-1]!==r-1-t)return!1;return!0},fg=(e,r)=>{let t=[];if(!hg(e,r)){for(let n=0;n<r;++n)e.indexOf(n)===-1&&t.push(n);e.forEach(n=>t.push(n))}return t},mg=(e,r,t,n,o,i,u)=>{let d=t[0].dims,h=Fe.size(i),f=Fe.size(u),y=Re("_A",t[0].dataType,d),m=mt("output",o,i),g=64;h===1&&(g=256);let v=`
          var<workgroup> aBestValues : array<f32, ${g}>;
       `,b=k=>`
        ${k.registerUniform("reduceSize","u32").declareVariables(y,m)}
        ${v}
        fn DIV_CEIL(a : u32, b : u32) -> u32 {
          return ((a - 1u) / b + 1u);
         }
         ${k.mainStart(g)}

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

          var bestValue = f32(${lg[n]});
          let Length = uniforms.reduceSize;
          for (var k = local_idx; k < Length; k = k + ${g}) {
           let candidate = f32(${y.getByOffset("offset + k")});
           bestValue = ${ag[n]};
          }
          aBestValues[local_idx] = bestValue;
          workgroupBarrier();

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

         if (local_idx == 0u) {
          ${m.setByOffset("outputIndex",`${n==="mean"?`${m.type.storage}(bestValue / f32(uniforms.reduceSize))`:`${m.type.storage}(${cg[n]})`}`)};
         }
        }`;return{name:e,shaderCache:{hint:`${r};${g}`,inputDependencies:["type"]},getShaderSource:b,getRunData:()=>({outputs:[{dims:i,dataType:o}],dispatchGroup:{x:h},programUniforms:[{type:12,data:f}]})}},Ui=(e,r,t,n)=>{let o=e.inputs.length===1?t:Lm(e.inputs,t),i=o.axes;i.length===0&&!o.noopWithEmptyAxes&&(i=e.inputs[0].dims.map((v,b)=>b));let u=Fe.normalizeAxes(i,e.inputs[0].dims.length),d=u,h=e.inputs[0],f=fg(d,e.inputs[0].dims.length);f.length>0&&(h=e.compute(xi(e.inputs[0],f),{inputs:[0],outputs:[-1]})[0],d=ug(d.length,h.dims.length));let[y,m]=dg(h.dims,d),g=y;o.keepDims&&(g=pg(y,u)),e.compute(mg(r,o.cacheKey,[h],n,e.inputs[0].dataType,g,m),{inputs:[h]})},J0=(e,r)=>{Ui(e,"ReduceMeanShared",r,"mean")},Y0=(e,r)=>{Ui(e,"ReduceL1Shared",r,"l1")},Z0=(e,r)=>{Ui(e,"ReduceL2Shared",r,"l2")},eM=(e,r)=>{Ui(e,"ReduceLogSumExpShared",r,"logSumExp")},tM=(e,r)=>{Ui(e,"ReduceMaxShared",r,"max")},rM=(e,r)=>{Ui(e,"ReduceMinShared",r,"min")},iM=(e,r)=>{Ui(e,"ReduceProdShared",r,"prod")},sM=(e,r)=>{Ui(e,"ReduceSumShared",r,"sum")},nM=(e,r)=>{Ui(e,"ReduceSumSquareShared",r,"sumSquare")},aM=(e,r)=>{Ui(e,"ReduceLogSumShared",r,"logSum")}}),Wi,_g,sf,Lm,Gi,gg,yg,wg,Mg,bg,vg,xg,Tg,Eg,Pg,Ki,oM,lM,cM,uM,dM,pM,hM,fM,mM,_M,o_=Je(()=>{St(),Ft(),vr(),Lt(),oT(),Wi=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.")},_g=e=>["","",`var value = ${e.getByIndices("input_indices")};`,""],sf=(e,r,t,n,o,i,u=!1,d=!1)=>{let h=[],f=t[0].dims,y=f.length,m=Fe.normalizeAxes(o,y),g=!d&&m.length===0;f.forEach((k,z)=>{g||m.indexOf(z)>=0?u&&h.push(1):h.push(k)});let v=h.length,b=Fe.size(h);return{name:e,shaderCache:r,getShaderSource:k=>{let z=[],E=Re("_A",t[0].dataType,y),T=mt("output",i,v),D=n(E,T,m),I=D[2];for(let $=0,P=0;$<y;$++)g||m.indexOf($)>=0?(u&&P++,I=`for(var j${$}: u32 = 0; j${$} < ${f[$]}; j${$}++) {
                  ${D[2].includes("last_index")?`let last_index = j${$};`:""}
                  ${E.indicesSet("input_indices",$,`j${$}`)}
                  ${I}
                }`):(z.push(`${E.indicesSet("input_indices",$,T.indicesGet("output_indices",P))};`),P++);return`

        ${k.registerUniform("output_size","u32").declareVariables(E,T)}

        ${k.mainStart()}
          ${k.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.output_size")}
          var input_indices: ${E.type.indices};
          let output_indices = ${T.offsetToIndices("global_idx")};

          ${z.join(`
`)}
          ${D[0]}       // init ops for reduce max/min
          ${D[1]}
          ${I}
          ${D[3]}
          ${D.length===4?T.setByOffset("global_idx","value"):D.slice(4).join(`
`)}
        }`},getRunData:()=>({outputs:[{dims:h,dataType:i}],dispatchGroup:{x:Math.ceil(b/64)},programUniforms:[{type:12,data:b},...bt(f,h)]})}},Lm=(e,r)=>{let t=[];return e[1].dims[0]>0&&e[1].getBigInt64Array().forEach(n=>t.push(Number(n))),Zt({axes:t,keepDims:r.keepDims,noopWithEmptyAxes:r.noopWithEmptyAxes})},Gi=(e,r,t,n)=>{let o=e.inputs,i=o.length===1?t:Lm(o,t);e.compute(sf(r,{hint:i.cacheKey,inputDependencies:["rank"]},[o[0]],i.noopWithEmptyAxes&&i.axes.length===0?_g:n,i.axes,o[0].dataType,i.keepDims,i.noopWithEmptyAxes),{inputs:[0]})},gg=(e,r)=>{Wi(e.inputs),Gi(e,"ReduceLogSum",r,(t,n)=>[`var value = ${n.type.storage}(0);`,"",`value += ${t.getByIndices("input_indices")};`,"value = log(value);"])},yg=(e,r)=>{Wi(e.inputs),Gi(e,"ReduceL1",r,(t,n)=>[`var value = ${n.type.storage}(0);`,"",`value += abs(${t.getByIndices("input_indices")});`,""])},wg=(e,r)=>{Wi(e.inputs),Gi(e,"ReduceL2",r,(t,n)=>[`var t = ${n.type.value}(0); var value = ${n.type.value}(0);`,"",`t = ${t.getByIndices("input_indices")}; value += (t * t);`,"value = sqrt(value);"])},Mg=(e,r)=>{Wi(e.inputs),Gi(e,"ReduceLogSumExp",r,(t,n)=>[`var value = ${n.type.storage}(0);`,"",`value += exp(${t.getByIndices("input_indices")});`,"value = log(value);"])},bg=(e,r)=>{Wi(e.inputs),Gi(e,"ReduceMax",r,(t,n,o)=>{let i=[];for(let u=0;u<t.rank;u++)(o.indexOf(u)>=0||o.length===0)&&i.push(t.indicesSet("input_indices",u,0));return[`${i.join(`
`)}`,`var value = ${t.getByIndices("input_indices")};`,`value = max(value, ${t.getByIndices("input_indices")});`,""]})},vg=(e,r)=>{Wi(e.inputs),Gi(e,"ReduceMean",r,(t,n,o)=>{let i=1;for(let u=0;u<t.rank;u++)(o.indexOf(u)>=0||o.length===0)&&(i*=e.inputs[0].dims[u]);return["var sum = f32(0);","",`sum += f32(${t.getByIndices("input_indices")});`,`let value = ${n.type.value}(sum / ${i});`]})},xg=(e,r)=>{Wi(e.inputs),Gi(e,"ReduceMin",r,(t,n,o)=>{let i=[];for(let u=0;u<t.rank;u++)(o.indexOf(u)>=0||o.length===0)&&i.push(`input_indices[${u}] = 0;`);return[`${i.join(`
`)}`,`var value = ${t.getByIndices("input_indices")};`,`value = min(value, ${t.getByIndices("input_indices")});`,""]})},Tg=(e,r)=>{Wi(e.inputs),Gi(e,"ReduceProd",r,(t,n)=>[`var value = ${n.type.storage}(1);`,"",`value *= ${t.getByIndices("input_indices")};`,""])},Eg=(e,r)=>{Wi(e.inputs),Gi(e,"ReduceSum",r,(t,n)=>[`var value = ${n.type.storage}(0);`,"",`value += ${t.getByIndices("input_indices")};`,""])},Pg=(e,r)=>{Wi(e.inputs),Gi(e,"ReduceSumSquare",r,(t,n)=>[`var t = ${n.type.value}(0); var value = ${n.type.value}(0);`,"",`t = ${t.getByIndices("input_indices")}; value += t * t;`,""])},Ki=(e,r,t)=>{if(r.length===0)return t;let n=1,o=1;for(let i=0;i<r.length;i++)r.indexOf(i)===-1?n*=e[i]:o*=e[i];return o<32&&n>1024},oM=(e,r)=>{Ki(e.inputs[0].dims,r.axes,r.noopWithEmptyAxes)?vg(e,r):J0(e,r)},lM=(e,r)=>{Ki(e.inputs[0].dims,r.axes,r.noopWithEmptyAxes)?yg(e,r):Y0(e,r)},cM=(e,r)=>{Ki(e.inputs[0].dims,r.axes,r.noopWithEmptyAxes)?wg(e,r):Z0(e,r)},uM=(e,r)=>{Ki(e.inputs[0].dims,r.axes,r.noopWithEmptyAxes)?Mg(e,r):eM(e,r)},dM=(e,r)=>{Ki(e.inputs[0].dims,r.axes,r.noopWithEmptyAxes)?bg(e,r):tM(e,r)},pM=(e,r)=>{Ki(e.inputs[0].dims,r.axes,r.noopWithEmptyAxes)?xg(e,r):rM(e,r)},hM=(e,r)=>{Ki(e.inputs[0].dims,r.axes,r.noopWithEmptyAxes)?Tg(e,r):iM(e,r)},fM=(e,r)=>{Ki(e.inputs[0].dims,r.axes,r.noopWithEmptyAxes)?Eg(e,r):sM(e,r)},mM=(e,r)=>{Ki(e.inputs[0].dims,r.axes,r.noopWithEmptyAxes)?Pg(e,r):nM(e,r)},_M=(e,r)=>{Ki(e.inputs[0].dims,r.axes,r.noopWithEmptyAxes)?gg(e,r):aM(e,r)}}),Jf,gM,yM,jm,lT=Je(()=>{St(),vr(),o_(),Jf=e=>{if(!e||e.length===0||e.length>2)throw new Error("ArgMinMaxOp op requires 1 or 2 inputs.");if(e[0].dataType!==1)throw new Error("Invalid input type.")},gM=(e,r)=>{Jf(e.inputs);let t=(n,o,i)=>{let u=[];for(let d=0;d<n.rank;d++)(i.indexOf(d)>=0||i.length===0)&&u.push(`input_indices[${d}] = 0;`);return[`${u.join(`
`)}`,`var value = ${n.getByIndices("input_indices")};
var best_index : i32 = 0;`,`if (${n.getByIndices("input_indices")} ${r.selectLastIndex>0?"<=":"<"} value) {
         value = ${n.getByIndices("input_indices")};
         best_index = i32(last_index);
       }`,"",o.setByOffset("global_idx","best_index")]};e.compute(sf("ArgMin",{hint:r.cacheKey,inputDependencies:["rank"]},[e.inputs[0]],t,[r.axis],7,r.keepDims),{inputs:[0]})},yM=(e,r)=>{Jf(e.inputs);let t=(n,o,i)=>{let u=[];for(let d=0;d<n.rank;d++)(i.indexOf(d)>=0||i.length===0)&&u.push(`input_indices[${d}] = 0;`);return[`${u.join(`
`)}`,`var value = ${n.getByIndices("input_indices")};
var best_index : i32 = 0;`,`if (${n.getByIndices("input_indices")} ${r.selectLastIndex>0?">=":">"} value) {
         value = ${n.getByIndices("input_indices")};
         best_index = i32(last_index);
       }`,"",o.setByOffset("global_idx","best_index")]};e.compute(sf("argMax",{hint:r.cacheKey,inputDependencies:["rank"]},[e.inputs[0]],t,[r.axis],7,r.keepDims),{inputs:[0]})},jm=e=>Zt(e)}),Cg,Vh,Sg,$g,kg,wd,Ig,wM,l_=Je(()=>{St(),Ft(),n_(),Lt(),Cg=(e,r)=>{let t=e[0],n=e[1],o=e[2],i=e[3],u=e[4],d=e[5];if(u&&d)throw new Error("Attention cannot have both past and attention_bias");if(t.dims.length!==3)throw new Error('Input "input" must have 3 dimensions');let h=t.dims[0],f=t.dims[1],y=t.dims[2];if(o.dims.length!==1)throw new Error('Input "bias" is expected to have 1 dimensions');if(n.dims.length!==2)throw new Error('Input "weights" is expected to have 2 dimensions');if(n.dims[0]!==y)throw new Error("Input 1 dimension 0 should have same length as dimension 2 of input 0");if(o.dims[0]!==n.dims[1])throw new Error('Input "bias" dimension 0 should have same length as dimension 1 of input "weights"');let m=o.dims[0]/3,g=m,v=g;if(r.qkvHiddenSizes.length>0){if(r.qkvHiddenSizes.length!==3)throw new Error("qkv_hidden_sizes attribute should have 3 elements");for(let D of r.qkvHiddenSizes)if(D%r.numHeads!==0)throw new Error("qkv_hidden_sizes should be divisible by num_heads");m=r.qkvHiddenSizes[0],g=r.qkvHiddenSizes[1],v=r.qkvHiddenSizes[2]}let b=f;if(m!==g)throw new Error("qkv_hidden_sizes first element should be same as the second");if(o.dims[0]!==m+g+v)throw new Error('Input "bias" dimension 0 should have same length as sum of Q/K/V hidden sizes');let k=0;if(u){if(g!==v)throw new Error('Input "past" expect k_hidden_size == v_hidden_size');if(u.dims.length!==5)throw new Error('Input "past" must have 5 dimensions');if(u.dims[0]!==2)throw new Error('Input "past" first dimension must be 2');if(u.dims[1]!==h)throw new Error('Input "past" second dimension must be batch_size');if(u.dims[2]!==r.numHeads)throw new Error('Input "past" third dimension must be num_heads');if(u.dims[4]!==g/r.numHeads)throw new Error('Input "past" fifth dimension must be k_hidden_size / num_heads');r.pastPresentShareBuffer||(k=u.dims[3])}let z=b+k,E=-1,T=0;if(i)throw new Error("Mask not supported");if(u)throw new Error("past is not supported");if(d){if(d.dims.length!==4)throw new Error('Input "attention_bias" must have 4 dimensions');if(d.dims[0]!==h||d.dims[1]!==r.numHeads||d.dims[2]!==f||d.dims[3]!==z)throw new Error('Expect "attention_bias" shape (batch_size, num_heads, sequence_length, total_sequence_length)')}return{batchSize:h,sequenceLength:f,pastSequenceLength:k,kvSequenceLength:b,totalSequenceLength:z,maxSequenceLength:E,inputHiddenSize:y,hiddenSize:m,vHiddenSize:v,headSize:Math.floor(m/r.numHeads),vHeadSize:Math.floor(v/r.numHeads),numHeads:r.numHeads,isUnidirectional:!1,pastPresentShareBuffer:!1,maskFilterValue:r.maskFilterValue,maskType:T,scale:r.scale,broadcastResPosBias:!1,passPastInKv:!1,qkvFormat:1}},Vh=(e,r,t)=>r&&e?`
      let total_sequence_length_input = u32(${r.getByOffset("0")});
      let present_sequence_length = max(total_sequence_length_input, uniforms.past_sequence_length);
      let is_subsequent_prompt: bool = sequence_length > 1 && sequence_length != total_sequence_length_input;
      let is_first_prompt: bool = is_subsequent_prompt == false && sequence_length == total_sequence_length_input;
      total_sequence_length = u32(${e==null?void 0:e.getByOffset("batchIdx")}) + 1;
      var past_sequence_length: u32 = 0;
      if (is_first_prompt == false) {
        past_sequence_length = total_sequence_length - sequence_length;
      }
       `:`
    ${t?"let past_sequence_length = uniforms.past_sequence_length":""};
    let present_sequence_length = total_sequence_length;
    `,Sg=(e,r,t,n,o,i,u,d)=>{let h=yr(u?1:i),f=64,y=i/h;y<f&&(f=32);let m=Math.ceil(i/h/f),g=[{type:12,data:r},{type:12,data:t},{type:12,data:n},{type:12,data:o},{type:12,data:y},{type:12,data:m}],v=jr(e.dataType,h),b=ti(1,h),k=["type"];u&&k.push("type"),d&&k.push("type");let z=E=>{let T=mt("x",e.dataType,e.dims,h),D=[T],I=u?Re("seq_lens",u.dataType,u.dims):void 0;I&&D.push(I);let $=d?Re("total_sequence_length_input",d.dataType,d.dims):void 0;$&&D.push($);let P=ti(e.dataType),S=[{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, ${f}>;
  var<workgroup> thread_sum: array<f32, ${f}>;
  ${E.registerUniforms(S).declareVariables(...D)}
  ${E.mainStart([f,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;
    ${Vh(I,$,!1)}
    let local_offset = local_idx * uniforms.elements_per_thread;
    let offset = (global_idx / ${f}) * uniforms.total_sequence_length + local_offset;
    let seq_causal_length = ${u?"u32(past_sequence_length + workgroup_id.y + 1)":"total_sequence_length"};
    var thread_max_vector = ${b}(-3.402823e+38f);
    for (var i: u32 = 0; i < uniforms.elements_per_thread && i + local_offset < seq_causal_length; i++) {
      thread_max_vector = max(${b}(x[offset + i]), thread_max_vector);
    }
    thread_max[local_idx] = ${(()=>{switch(h){case 1:return"thread_max_vector";case 2:return"max(thread_max_vector.x, thread_max_vector.y)";case 4:return"max(max(thread_max_vector.x, thread_max_vector.y), max(thread_max_vector.z, thread_max_vector.w))";default:throw new Error(`Unsupported components: ${h}`)}})()};
    workgroupBarrier();

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

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

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

    if (sum == 0) {
      for (var i: u32 = 0; i < uniforms.elements_per_thread && i + local_offset < seq_causal_length; i++) {
        x[offset + i] = ${T.type.value}(${P}(1.0) / ${P}(seq_causal_length));
      }
    } else {
      for (var i: u32 = 0; i < uniforms.elements_per_thread && i + local_offset < seq_causal_length; i++) {
        var f32input = ${b}(x[offset + i]);
        x[offset + i] = ${T.type.value}(exp(f32input - max_value) / sum);
      }
    }
      ${u?`
        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] = ${T.type.value}(${P}(0));
        }`:""};
  }`};return{name:"AttentionProbsSoftmax",shaderCache:{hint:`${f};${v};${h}`,inputDependencies:k},getShaderSource:z,getRunData:()=>({outputs:[],dispatchGroup:{x:1,y:o,z:r*t},programUniforms:g})}},$g=(e,r,t,n,o,i,u,d,h)=>{let f=u+i.kvSequenceLength,y=[i.batchSize,i.numHeads,i.sequenceLength,f],m=e>1&&n,g=i.kvNumHeads?i.kvNumHeads:i.numHeads,v=m?[i.batchSize,g,f,i.headSize]:void 0,b=i.nReps?i.nReps:1,k=i.scale===0?1/Math.sqrt(i.headSize):i.scale,z=yr(i.headSize),E=i.headSize/z,T=12,D={x:Math.ceil(f/T),y:Math.ceil(i.sequenceLength/T),z:i.batchSize*i.numHeads},I=[{type:12,data:i.sequenceLength},{type:12,data:E},{type:12,data:f},{type:12,data:i.numHeads},{type:12,data:i.headSize},{type:1,data:k},{type:12,data:u},{type:12,data:i.kvSequenceLength},{type:12,data:b}],$=m&&n&&Fe.size(n.dims)>0,P=["type","type"];$&&P.push("type"),o&&P.push("type"),d&&P.push("type"),h&&P.push("type");let S=[{dims:y,dataType:r.dataType,gpuDataType:0}];m&&S.push({dims:v,dataType:r.dataType,gpuDataType:0});let O=R=>{let G=Re("q",r.dataType,r.dims,z),ee=Re("key",t.dataType,t.dims,z),ie=[G,ee];if($){let Ce=Re("past_key",n.dataType,n.dims,z);ie.push(Ce)}o&&ie.push(Re("attention_bias",o.dataType,o.dims));let H=d?Re("seq_lens",d.dataType,d.dims):void 0;H&&ie.push(H);let ce=h?Re("total_sequence_length_input",h.dataType,h.dims):void 0;ce&&ie.push(ce);let re=mt("output",r.dataType,y),se=[re];m&&se.push(mt("present_key",r.dataType,v,z));let _e=ti(1,z),ae=[{name:"M",type:"u32"},{name:"K",type:"u32"},{name:"N",type:"u32"},{name:"num_heads",type:"u32"},{name:"head_size",type:"u32"},{name:"alpha",type:"f32"},{name:"past_sequence_length",type:"u32"},{name:"kv_sequence_length",type:"u32"},{name:"n_reps",type:"u32"}];return`
  const TILE_SIZE = ${T}u;

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

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

      workgroupBarrier();
    }

    if (global_id.y < uniforms.M && global_id.x < total_sequence_length) {
      let headOffset = workgroup_id.z * uniforms.M * uniforms.N;
      let outputIdx = headOffset + global_id.y * uniforms.N + global_id.x;
      var sum: f32 = ${(()=>{switch(z){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: ${z}`)}})()};
        output[outputIdx] = ${re.type.value} (sum * uniforms.alpha) + ${o?"attention_bias[outputIdx]":"0.0"};
    }
  }`};return{name:"AttentionProbs",shaderCache:{hint:`${z};${o!==void 0};${n!==void 0};${e}`,inputDependencies:P},getRunData:()=>({outputs:S,dispatchGroup:D,programUniforms:I}),getShaderSource:O}},kg=(e,r,t,n,o,i,u=void 0,d=void 0)=>{let h=i+o.kvSequenceLength,f=o.nReps?o.nReps:1,y=o.vHiddenSize*f,m=e>1&&n,g=o.kvNumHeads?o.kvNumHeads:o.numHeads,v=m?[o.batchSize,g,h,o.headSize]:void 0,b=[o.batchSize,o.sequenceLength,y],k=12,z={x:Math.ceil(o.vHeadSize/k),y:Math.ceil(o.sequenceLength/k),z:o.batchSize*o.numHeads},E=[{type:12,data:o.sequenceLength},{type:12,data:h},{type:12,data:o.vHeadSize},{type:12,data:o.numHeads},{type:12,data:o.headSize},{type:12,data:y},{type:12,data:i},{type:12,data:o.kvSequenceLength},{type:12,data:f}],T=m&&n&&Fe.size(n.dims)>0,D=["type","type"];T&&D.push("type"),u&&D.push("type"),d&&D.push("type");let I=[{dims:b,dataType:r.dataType,gpuDataType:0}];m&&I.push({dims:v,dataType:r.dataType,gpuDataType:0});let $=P=>{let S=Re("probs",r.dataType,r.dims),O=Re("v",t.dataType,t.dims),R=[S,O];T&&R.push(Re("past_value",n.dataType,n.dims));let G=u?Re("seq_lens",u.dataType,u.dims):void 0;u&&R.push(G);let ee=d?Re("total_sequence_length_input",d.dataType,d.dims):void 0;d&&R.push(ee);let ie=[mt("output",r.dataType,b)];m&&ie.push(mt("present_value",r.dataType,v));let H=[{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 = ${k}u;
  var<workgroup> tileQ: array<${S.type.value}, ${k*k}>;
  var<workgroup> tileV: array<${S.type.value}, ${k*k}>;
  ${P.registerUniforms(H).declareVariables(...R,...ie)}
  ${P.mainStart([k,k,1])}
   let headIdx = workgroup_id.z % uniforms.num_heads;
   let batchIdx = workgroup_id.z / uniforms.num_heads;
   let kvHeadIdx = ${f===1?"headIdx":"headIdx / uniforms.n_reps"};
   let kv_num_heads = ${f===1?"uniforms.num_heads":"uniforms.num_heads / uniforms.n_reps"};
   let m = global_id.y;
   let n = global_id.x;
   let sequence_length = uniforms.M;
   var total_sequence_length = uniforms.K;
   ${Vh(G,ee,!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
   ${T&&m?"let pastValueOffset = absKvHeadIdx * uniforms.N * uniforms.past_sequence_length + n;":""};
   let vOffset = absKvHeadIdx * uniforms.N * uniforms.kv_sequence_length + n;
   ${m?"let presentValueOffset = absKvHeadIdx * uniforms.N * uniforms.K + n;":""}
   var value = ${S.type.storage}(0);
   for (var w: u32 = 0u; w < uniforms.K; w += TILE_SIZE) {
      if (m < uniforms.M && w + local_id.x < uniforms.K) {
        tileQ[TILE_SIZE * local_id.y + local_id.x] = probs[offsetA + w + local_id.x];
      }
      if (n < uniforms.N && w + local_id.y < uniforms.K) {
        var idx = TILE_SIZE * local_id.y + local_id.x;
        ${T&&m?`
        if (w + local_id.y < past_sequence_length) {
          tileV[idx] = past_value[pastValueOffset + (w + local_id.y) * uniforms.N];
        } else if (w + local_id.y - past_sequence_length < uniforms.kv_sequence_length) {
          tileV[idx] = v[vOffset + (w + local_id.y - past_sequence_length) * uniforms.N];
        }
      `:`
            if (w + local_id.y < uniforms.kv_sequence_length) {
              tileV[idx] = v[vOffset + (w + local_id.y) * uniforms.N];
            }`}
        ${m?`
            if (w + local_id.y < present_sequence_length) {
          present_value[presentValueOffset + (w + local_id.y) * uniforms.N] = tileV[idx];
        }`:""}
      }
     workgroupBarrier();
     for (var k: u32 = 0u; k < TILE_SIZE && w+k < total_sequence_length; k++) {
       value += tileQ[TILE_SIZE * local_id.y + k] * tileV[TILE_SIZE * k + local_id.x];
     }
     workgroupBarrier();
   }

   // we need to transpose output from BNSH_v to BSND_v
   if (m < uniforms.M && n < uniforms.N) {
     let outputIdx = batchIdx * uniforms.M * uniforms.v_hidden_size + m * uniforms.v_hidden_size
       + headIdx * uniforms.N + n;
     output[outputIdx] = value;
   }
  }`};return{name:"AttentionScore",shaderCache:{hint:`${n!==void 0};${e}`,inputDependencies:D},getRunData:()=>({outputs:I,dispatchGroup:z,programUniforms:E}),getShaderSource:$}},wd=(e,r,t,n,o,i,u,d,h,f,y=void 0,m=void 0)=>{let g=Math.min(e.outputCount,1+(u?1:0)+(d?1:0)),v=g>1?f.pastSequenceLength:0,b=v+f.kvSequenceLength,k=h&&Fe.size(h.dims)>0?h:void 0,z=[r,t];g>1&&u&&Fe.size(u.dims)>0&&z.push(u),k&&z.push(k),y&&z.push(y),m&&z.push(m);let E=e.compute($g(g,r,t,u,k,f,v,y,m),{inputs:z,outputs:g>1?[-1,1]:[-1]})[0];e.compute(Sg(E,f.batchSize,f.numHeads,v,f.sequenceLength,b,y,m),{inputs:y&&m?[E,y,m]:[E],outputs:[]});let T=[E,n];g>1&&d&&Fe.size(d.dims)>0&&T.push(d),y&&T.push(y),m&&T.push(m),e.compute(kg(g,E,n,d,f,v,y,m),{inputs:T,outputs:g>1?[0,2]:[0]})},Ig=(e,r)=>{let t=[r.batchSize,r.numHeads,r.sequenceLength,r.headSize],n=r.sequenceLength,o=r.inputHiddenSize,i=r.headSize,u=12,d={x:Math.ceil(r.headSize/u),y:Math.ceil(r.sequenceLength/u),z:r.batchSize*r.numHeads},h=[e.inputs[0],e.inputs[1],e.inputs[2]],f=[{type:12,data:n},{type:12,data:o},{type:12,data:i},{type:12,data:r.numHeads},{type:12,data:r.headSize},{type:12,data:r.hiddenSize},{type:12,data:r.hiddenSize+r.hiddenSize+r.vHiddenSize}],y=m=>{let g=mt("output_q",h[0].dataType,t),v=mt("output_k",h[0].dataType,t),b=mt("output_v",h[0].dataType,t),k=Re("input",h[0].dataType,h[0].dims),z=Re("weight",h[1].dataType,h[1].dims),E=Re("bias",h[2].dataType,h[2].dims),T=k.type.storage,D=[{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 = ${u}u;
  var<workgroup> tileInput: array<${T}, ${u*u}>;
  var<workgroup> tileWeightQ: array<${T}, ${u*u}>;
  var<workgroup> tileWeightK: array<${T}, ${u*u}>;
  var<workgroup> tileWeightV: array<${T}, ${u*u}>;
  ${m.registerUniforms(D).declareVariables(k,z,E,g,v,b)}
  ${m.mainStart([u,u,1])}
    let batchIndex = workgroup_id.z / uniforms.num_heads;
    let headNumber = workgroup_id.z % uniforms.num_heads;
    let m = global_id.y;
    let n = global_id.x;

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

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

      workgroupBarrier();
    }

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

    let offset = workgroup_id.z * uniforms.M * uniforms.N;
    if (m < uniforms.M && n < uniforms.N) {
      let outputIdx = offset + m * uniforms.N + n;
      output_q[outputIdx] = valueQ;
      output_k[outputIdx] = valueK;
      output_v[outputIdx] = valueV;
    }
  }`};return e.compute({name:"AttentionPrepare",shaderCache:{inputDependencies:["type","type","type"]},getRunData:()=>({outputs:[{dims:t,dataType:e.inputs[0].dataType,gpuDataType:0},{dims:t,dataType:e.inputs[0].dataType,gpuDataType:0},{dims:t,dataType:e.inputs[0].dataType,gpuDataType:0}],dispatchGroup:d,programUniforms:f}),getShaderSource:y},{inputs:h,outputs:[-1,-1,-1]})},wM=(e,r)=>{let t=Cg(e.inputs,r),[n,o,i]=Ig(e,t);return wd(e,n,o,i,e.inputs[4],void 0,void 0,void 0,e.inputs[5],t)}}),Ag,Og,Fg,MM,cT=Je(()=>{Ii(),St(),Ft(),vr(),Lt(),Ag=(e,r)=>{if(!e||e.length!==5)throw new Error("BatchNormalization requires 5 inputs");let t=(n,o,i)=>{let u=o.length;if(u!==n.length)throw new Error(`${i}: num dimensions != ${u}`);o.forEach((d,h)=>{if(d!==n[h])throw new Error(`${i}: dim[${h}] do not match`)})};if(e[0].dims.length>1){let n=r.format==="NHWC"?r.spatial?e[0].dims.slice(-1):e[0].dims.slice(-1).concat(e[0].dims.slice(1,e[0].dims.length-1)):e[0].dims.slice(1,r.spatial?2:void 0);t(e[1].dims,n,"Invalid input scale"),t(e[2].dims,n,"Invalid input B"),t(e[3].dims,n,"Invalid input mean"),t(e[4].dims,n,"Invalid input var")}else t(e[1].dims,[1],"Invalid input scale"),t(e[2].dims,[1],"Invalid input B"),t(e[3].dims,[1],"Invalid input mean"),t(e[4].dims,[1],"Invalid input var")},Og=(e,r)=>{let{epsilon:t,spatial:n,format:o}=r,i=e[0].dims,u=n?yr(i[i.length-1]):1,d=o==="NHWC"&&i.length>1?u:1,h=Fe.size(i)/u,f=n,y=f?i.length:i,m=Re("x",e[0].dataType,e[0].dims,u),g=Re("scale",e[1].dataType,e[1].dims,d),v=Re("bias",e[2].dataType,e[2].dims,d),b=Re("inputMean",e[3].dataType,e[3].dims,d),k=Re("inputVar",e[4].dataType,e[4].dims,d),z=mt("y",e[0].dataType,y,u),E=()=>{let D="";if(n)D=`let cOffset = ${i.length===1?"0u":o==="NHWC"?`outputIndices[${i.length-1}] / ${u}`:"outputIndices[1]"};`;else if(o==="NCHW")D=`
            ${z.indicesSet("outputIndices","0","0")}
            let cOffset = ${z.indicesToOffset("outputIndices")};`;else{D=`var cIndices = ${g.type.indices}(0);
                       cIndices[0] = outputIndices[${i.length-1}];`;for(let I=1;I<g.rank;I++)D+=`cIndices[${I}] = outputIndices[${I}];`;D+=`let cOffset = ${g.indicesToOffset("cIndices")};`}return D},T=D=>`
  const epsilon = ${t};
  ${D.registerUniform("outputSize","u32").declareVariables(m,g,v,b,k,z)}
  ${D.mainStart()}
  ${D.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.outputSize")}
    var outputIndices = ${z.offsetToIndices(`global_idx * ${u}`)};
    ${E()}
    let scale = ${g.getByOffset("cOffset")};
    let bias = ${v.getByOffset("cOffset")};
    let inputMean = ${b.getByOffset("cOffset")};
    let inputVar = ${k.getByOffset("cOffset")};
    let x = ${m.getByOffset("global_idx")};
    let value = (x - inputMean) * inverseSqrt(inputVar + epsilon) * scale + bias;
    ${z.setByOffset("global_idx","value")}
  }`;return{name:"BatchNormalization",shaderCache:{hint:`${r.epsilon}_${r.format}_${n}_${u}`,inputDependencies:f?["rank","type","type","type","type"]:void 0},getShaderSource:T,getRunData:()=>({outputs:[{dims:e[0].dims,dataType:e[0].dataType}],dispatchGroup:{x:Math.ceil(h/64)},programUniforms:f?[{type:12,data:h},...bt(i)]:[{type:12,data:h}]})}},Fg=e=>Zt(e),MM=(e,r)=>{let{inputs:t,outputCount:n}=e,o=Fg({...r,outputCount:n});if(dr.webgpu.validateInputContent&&Ag(t,o),r.trainingMode)throw new Error("BatchNormalization trainingMode is not supported yet.");e.compute(Og(t,o))}}),Dg,Lg,bM,uT=Je(()=>{Ft(),Lt(),Dg=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")},Lg=e=>{let r=e[0].dims,t=e[0].dims[2],n=Fe.size(r)/4,o=e[0].dataType,i=Re("input",o,r,4),u=Re("bias",o,[t],4),d=Re("residual",o,r,4),h=mt("output",o,r,4);return{name:"BiasAdd",getRunData:()=>({outputs:[{dims:r,dataType:e[0].dataType}],dispatchGroup:{x:Math.ceil(n/64)}}),getShaderSource:f=>`
  const channels = ${t}u / 4;
  ${f.declareVariables(i,u,d,h)}

  ${f.mainStart()}
    ${f.guardAgainstOutOfBoundsWorkgroupSizes(n)}
    let value = ${i.getByOffset("global_idx")}
      + ${u.getByOffset("global_idx % channels")} + ${d.getByOffset("global_idx")};
    ${h.setByOffset("global_idx","value")}
  }`}},bM=e=>{Dg(e.inputs),e.compute(Lg(e.inputs))}}),jg,Jt,vM,xM,TM,EM,PM,CM,SM,$M,kM,zg,IM,AM,OM,FM,hd,DM,Yh,LM,jM,zM,BM,RM,NM,VM,UM,WM,GM,KM,HM,qM,QM,XM,JM,Yf,YM,zm,Bm,ZM,eb,tb,Bg,Rg,rb,c_=Je(()=>{St(),Ft(),vr(),Lt(),jg=(e,r,t,n,o,i,u)=>{let d=Math.ceil(r/4),h="";typeof o=="string"?h=`${o}(a)`:h=o("a");let f=Re("inputData",t,[d],4),y=mt("outputData",n,[d],4),m=[{name:"vec_size",type:"u32"}];return u&&m.push(...u),`
      ${e.registerUniforms(m).declareVariables(f,y)}

  ${i??""}

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

    let a = ${f.getByOffset("global_idx")};
    ${y.setByOffset("global_idx",h)}
  }`},Jt=(e,r,t,n,o,i=e.dataType,u,d)=>{let h=[{type:12,data:Math.ceil(Fe.size(e.dims)/4)}];return u&&h.push(...u),{name:r,shaderCache:{hint:o,inputDependencies:["type"]},getShaderSource:f=>jg(f,Fe.size(e.dims),e.dataType,i,t,n,d),getRunData:f=>({outputs:[{dims:e.dims,dataType:i}],dispatchGroup:{x:Math.ceil(Fe.size(f[0].dims)/64/4)},programUniforms:h})}},vM=e=>{e.compute(Jt(e.inputs[0],"Abs","abs"))},xM=e=>{e.compute(Jt(e.inputs[0],"Acos","acos"))},TM=e=>{e.compute(Jt(e.inputs[0],"Acosh","acosh"))},EM=e=>{e.compute(Jt(e.inputs[0],"Asin","asin"))},PM=e=>{e.compute(Jt(e.inputs[0],"Asinh","asinh"))},CM=e=>{e.compute(Jt(e.inputs[0],"Atan","atan"))},SM=e=>{e.compute(Jt(e.inputs[0],"Atanh","atanh"))},$M=e=>Zt(e),kM=(e,r)=>{let t;switch(r.to){case 10:t="vec4<f16>";break;case 1:t="vec4<f32>";break;case 12:t="vec4<u32>";break;case 6:t="vec4<i32>";break;case 9:t="vec4<bool>";break;default:throw new RangeError(`not supported type (specified in attribute 'to' from 'Cast' operator): ${r.to}`)}e.compute(Jt(e.inputs[0],"Cast",t,void 0,r.cacheKey,r.to))},zg=e=>{let r,t,n=e.length>=2&&e[1].data!==0,o=e.length>=3&&e[2].data!==0;switch(e[0].dataType){case 1:r=n?e[1].getFloat32Array()[0]:-34028234663852886e22,t=o?e[2].getFloat32Array()[0]:34028234663852886e22;break;case 10:r=n?e[1].getUint16Array()[0]:64511,t=o?e[2].getUint16Array()[0]:31743;break;default:throw new Error("Unsupport data type")}return Zt({min:r,max:t})},IM=(e,r)=>{let t=r||zg(e.inputs),n=ti(e.inputs[0].dataType);e.compute(Jt(e.inputs[0],"Clip",o=>`clamp(${o}, vec4<${n}>(uniforms.min), vec4<${n}>(uniforms.max))`,void 0,t.cacheKey,void 0,[{type:e.inputs[0].dataType,data:t.min},{type:e.inputs[0].dataType,data:t.max}],[{name:"min",type:n},{name:"max",type:n}]),{inputs:[0]})},AM=e=>{e.compute(Jt(e.inputs[0],"Ceil","ceil"))},OM=e=>{e.compute(Jt(e.inputs[0],"Cos","cos"))},FM=e=>{e.compute(Jt(e.inputs[0],"Cosh","cosh"))},hd=e=>Zt(e),DM=(e,r)=>{let t=ti(e.inputs[0].dataType);e.compute(Jt(e.inputs[0],"Elu",n=>`elu_vf32(${n})`,`
  const elu_alpha_ = ${t}(${r.alpha});

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

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

fn erf_vf32(v: vec4<${e}>) -> vec4<${e}> {
  let absv = abs(v);
  let x = 1.0 / (1.0 + r0 * absv);
  return sign(v) * (1.0 - ((((r5 * x + r4) * x + r3) * x + r2) * x + r1) * x * exp(-absv * absv));
}`,LM=e=>{let r=ti(e.inputs[0].dataType);e.compute(Jt(e.inputs[0],"Erf",t=>`erf_vf32(${t})`,Yh(r)))},jM=e=>{e.compute(Jt(e.inputs[0],"Exp","exp"))},zM=e=>{e.compute(Jt(e.inputs[0],"Floor","floor"))},BM=e=>{let r=ti(e.inputs[0].dataType);e.compute(Jt(e.inputs[0],"Gelu",t=>`0.5 * ${t} * (1.0 + erf_vf32(${t} * 0.7071067811865475))`,Yh(r)))},RM=(e,r)=>{let t=ti(e.inputs[0].dataType);e.compute(Jt(e.inputs[0],"LeakyRelu",n=>`select(leaky_relu_alpha_ * ${n}, ${n}, ${n} >= vec4<${t}>(0.0))`,`const leaky_relu_alpha_ = ${t}(${r.alpha});`,r.cacheKey))},NM=e=>{e.compute(Jt(e.inputs[0],"Not",r=>`!${r}`))},VM=e=>{e.compute(Jt(e.inputs[0],"Neg",r=>`-${r}`))},UM=e=>{e.compute(Jt(e.inputs[0],"Reciprocal",r=>`1.0/${r}`))},WM=e=>{let r=ti(e.inputs[0].dataType);e.compute(Jt(e.inputs[0],"Relu",t=>`select(vec4<${r}>(0.0), ${t}, ${t} > vec4<${r}>(0.0))`))},GM=e=>{e.compute(Jt(e.inputs[0],"Sigmoid",r=>`(1.0 / (1.0 + exp(-${r})))`))},KM=e=>Zt(e),HM=(e,r)=>{let t=ti(e.inputs[0].dataType);e.compute(Jt(e.inputs[0],"HardSigmoid",n=>`max(vec4<${t}>(0.0), min(vec4<${t}>(1.0), ${r.alpha} * ${n} + vec4<${t}>(${r.beta})))`,void 0,r.cacheKey))},qM=e=>{e.compute(Jt(e.inputs[0],"Sin","sin"))},QM=e=>{e.compute(Jt(e.inputs[0],"Sinh","sinh"))},XM=e=>{e.compute(Jt(e.inputs[0],"Sqrt","sqrt"))},JM=e=>{e.compute(Jt(e.inputs[0],"Tan","tan"))},Yf=e=>`sign(${e}) * (1 - exp(-2 * abs(${e}))) / (1 + exp(-2 * abs(${e})))`,YM=e=>{e.compute(Jt(e.inputs[0],"Tanh",Yf))},zm=(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 ${Yf("v")};
}
`,Bm=e=>`(fast_gelu_a + fast_gelu_a * tanh_v(${e} * (fast_gelu_c * ${e} * ${e} + fast_gelu_b))) * ${e}`,ZM=e=>{let r=ti(e.inputs[0].dataType);e.compute(Jt(e.inputs[0],"FastGelu",Bm,zm(r),void 0,e.inputs[0].dataType))},eb=(e,r)=>{let t=ti(e.inputs[0].dataType);return e.compute(Jt(e.inputs[0],"ThresholdedRelu",n=>`select(vec4<${t}>(0.0), ${n}, ${n} > thresholded_relu_alpha_)`,`const thresholded_relu_alpha_ = vec4<${t}>(${r.alpha});`,r.cacheKey)),0},tb=e=>{e.compute(Jt(e.inputs[0],"Log","log"))},Bg=(e,r)=>`
const alpha = vec4<${e}>(${r});
const one = ${e}(1.0);
const zero = ${e}(0.0);

fn quick_gelu_impl(x: vec4<${e}>) -> vec4<${e}> {
  let v = x *alpha;
  var x1 : vec4<${e}>;
  for (var i = 0; i < 4; i = i + 1) {
    if (v[i] >= zero) {
      x1[i] = one / (one + exp(-v[i]));
    } else {
      x1[i] = one - one / (one + exp(v[i]));
    }
  }
  return x * x1;
}
`,Rg=e=>`quick_gelu_impl(${e})`,rb=(e,r)=>{let t=ti(e.inputs[0].dataType);e.compute(Jt(e.inputs[0],"QuickGelu",Rg,Bg(t,r.alpha),r.cacheKey,e.inputs[0].dataType))}}),Ng,Vg,ib,dT=Je(()=>{Ft(),Lt(),c_(),Ng=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")},Vg=e=>{let r=e[0].dims.slice();r[2]=r[2]/2;let t=Re("input",e[0].dataType,e[0].dims,4),n=Re("bias",e[0].dataType,[e[0].dims[2]],4),o=mt("output",e[0].dataType,r,4),i=Fe.size(r)/4,u=jr(e[0].dataType);return{name:"BiasSplitGelu",getRunData:()=>({outputs:[{dims:r,dataType:e[0].dataType}],dispatchGroup:{x:Math.ceil(i/64)}}),getShaderSource:d=>`
  const M_SQRT2 = sqrt(2.0);
  const halfChannels = ${e[0].dims[2]/4/2}u;

  ${d.declareVariables(t,n,o)}

  ${Yh(u)}

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

    ${o.setByOffset("global_idx","valueLeft * geluRight")}
  }`}},ib=e=>{Ng(e.inputs),e.compute(Vg(e.inputs))}}),Ug,Wg,Hi,sb,nb,ab,ob,lb,cb,ub,db,pb,hb,pT=Je(()=>{St(),Ft(),Lt(),Ug=(e,r,t,n,o,i,u,d,h,f,y,m)=>{let g,v;typeof d=="string"?g=v=(T,D)=>`${d}((${T}),(${D}))`:typeof d=="function"?g=v=d:(g=d.scalar,v=d.vector);let b=mt("outputData",y,n.length,4),k=Re("aData",h,r.length,4),z=Re("bData",f,t.length,4),E;if(o)if(i){let T=Fe.size(r)===1,D=Fe.size(t)===1,I=r.length>0&&r[r.length-1]%4===0,$=t.length>0&&t[t.length-1]%4===0;T||D?E=b.setByOffset("global_idx",v(T?`${k.type.value}(${k.getByOffset("0")}.x)`:k.getByOffset("global_idx"),D?`${z.type.value}(${z.getByOffset("0")}.x)`:z.getByOffset("global_idx"))):E=`
            let outputIndices = ${b.offsetToIndices("global_idx * 4u")};
            let offsetA = ${k.broadcastedIndicesToOffset("outputIndices",b)};
            let offsetB = ${z.broadcastedIndicesToOffset("outputIndices",b)};
            ${b.setByOffset("global_idx",v(u||I?k.getByOffset("offsetA / 4u"):`${k.type.value}(${k.getByOffset("offsetA / 4u")}[offsetA % 4u])`,u||$?z.getByOffset("offsetB / 4u"):`${z.type.value}(${z.getByOffset("offsetB / 4u")}[offsetB % 4u])`))}
          `}else E=b.setByOffset("global_idx",v(k.getByOffset("global_idx"),z.getByOffset("global_idx")));else{if(!i)throw new Error("no necessary to use scalar implementation for element-wise binary op implementation.");let T=(D,I,$="")=>{let P=`aData[indexA${I}][componentA${I}]`,S=`bData[indexB${I}][componentB${I}]`;return`
            let outputIndices${I} = ${b.offsetToIndices(`global_idx * 4u + ${I}u`)};
            let offsetA${I} = ${k.broadcastedIndicesToOffset(`outputIndices${I}`,b)};
            let offsetB${I} = ${z.broadcastedIndicesToOffset(`outputIndices${I}`,b)};
            let indexA${I} = offsetA${I} / 4u;
            let indexB${I} = offsetB${I} / 4u;
            let componentA${I} = offsetA${I} % 4u;
            let componentB${I} = offsetB${I} % 4u;
            ${D}[${I}] = ${$}(${g(P,S)});
          `};y===9?E=`
            var data = vec4<u32>(0);
            ${T("data",0,"u32")}
            ${T("data",1,"u32")}
            ${T("data",2,"u32")}
            ${T("data",3,"u32")}
            outputData[global_idx] = dot(vec4<u32>(0x1, 0x100, 0x10000, 0x1000000), vec4<u32>(data));`:E=`
            ${T("outputData[global_idx]",0)}
            ${T("outputData[global_idx]",1)}
            ${T("outputData[global_idx]",2)}
            ${T("outputData[global_idx]",3)}
          `}return`
        ${e.registerUniform("vec_size","u32").declareVariables(k,z,b)}

        ${m??""}

        ${e.mainStart()}
        ${e.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.vec_size")}
        ${E}
      }`},Wg=(e,r,t,n,o,i,u=t.dataType)=>{let d=t.dims.map(k=>Number(k)??1),h=n.dims.map(k=>Number(k)??1),f=!Fe.areEqual(d,h),y=d,m=Fe.size(d),g=!1,v=!1,b=[f];if(f){let k=Ua.calcShape(d,h,!1);if(!k)throw new Error("Can't perform binary op on the given tensors");y=k.slice(),m=Fe.size(y);let z=Fe.size(d)===1,E=Fe.size(h)===1,T=d.length>0&&d[d.length-1]%4===0,D=h.length>0&&h[h.length-1]%4===0;b.push(z),b.push(E),b.push(T),b.push(D);let I=1;for(let $=1;$<y.length;$++){let P=d[d.length-$],S=h[h.length-$];if(P===S)I*=P;else break}I%4===0?(v=!0,g=!0):(z||E||T||D)&&(g=!0)}else g=!0;return b.push(g),{name:e,shaderCache:{hint:r+b.map(k=>k.toString()).join("_"),inputDependencies:["rank","rank"]},getShaderSource:k=>Ug(k,d,h,y,g,f,v,o,t.dataType,n.dataType,u,i),getRunData:()=>({outputs:[{dims:y,dataType:u}],dispatchGroup:{x:Math.ceil(m/64/4)},programUniforms:[{type:12,data:Math.ceil(Fe.size(y)/4)},...bt(d,h,y)]})}},Hi=(e,r,t,n,o,i)=>{e.compute(Wg(r,o??"",e.inputs[0],e.inputs[1],t,n,i))},sb=e=>{Hi(e,"Add",(r,t)=>`${r}+${t}`)},nb=e=>{Hi(e,"Div",(r,t)=>`${r}/${t}`)},ab=e=>{Hi(e,"Equal",{scalar:(r,t)=>`u32(${r}==${t})`,vector:(r,t)=>`vec4<u32>(${r}==${t})`},void 0,void 0,9)},ob=e=>{Hi(e,"Mul",(r,t)=>`${r}*${t}`)},lb=e=>{let r=Re("input",e.inputs[0].dataType,e.inputs[0].dims).type.value;Hi(e,"Pow",{scalar:(t,n)=>`pow_custom(${t},${n})`,vector:(t,n)=>`pow_vector_custom(${t},${n})`},`
    fn pow_custom(a : ${r}, b : ${r}) -> ${r} {
      if (b == ${r}(0.0)) {
        return ${r}(1.0);
      } else if (a < ${r}(0.0) && f32(b) != floor(f32(b))) {
        return ${r}(pow(f32(a), f32(b))); // NaN
      }
      return select(sign(a), ${r}(1.0), round(f32(abs(b) % ${r}(2.0))) != 1.0) * ${r}(${r==="i32"?"round":""}(pow(f32(abs(a)), f32(b))));
    }
    fn pow_vector_custom(a : vec4<${r}>, b : vec4<${r}>) -> vec4<${r}> {
      // TODO: implement vectorized pow
      return vec4<${r}>(pow_custom(a.x, b.x), pow_custom(a.y, b.y), pow_custom(a.z, b.z), pow_custom(a.w, b.w));
    }
      `)},cb=e=>{Hi(e,"Sub",(r,t)=>`${r}-${t}`)},ub=e=>{Hi(e,"Greater",{scalar:(r,t)=>`u32(${r}>${t})`,vector:(r,t)=>`vec4<u32>(${r}>${t})`},void 0,void 0,9)},db=e=>{Hi(e,"Less",{scalar:(r,t)=>`u32(${r}<${t})`,vector:(r,t)=>`vec4<u32>(${r}<${t})`},void 0,void 0,9)},pb=e=>{Hi(e,"GreaterOrEqual",{scalar:(r,t)=>`u32(${r}>=${t})`,vector:(r,t)=>`vec4<u32>(${r}>=${t})`},void 0,void 0,9)},hb=e=>{Hi(e,"LessOrEqual",{scalar:(r,t)=>`u32(${r}<=${t})`,vector:(r,t)=>`vec4<u32>(${r}<=${t})`},void 0,void 0,9)}}),Gg,Kg,Hg,qg,fb,mb,hT=Je(()=>{St(),Ft(),vr(),Lt(),Gg=(e,r)=>{if(!e||e.length<1)throw new Error("too few inputs");let t=0,n=e[t],o=n.dataType,i=n.dims.length;e.forEach((u,d)=>{if(d!==t){if(u.dataType!==o)throw new Error("input tensors should be one type");if(u.dims.length!==i)throw new Error("input tensors should have the same shape");u.dims.forEach((h,f)=>{if(f!==r&&h!==n.dims[f])throw new Error("non concat dimensions must match")})}})},Kg=(e,r)=>`
  fn calculateInputIndex(index: u32) -> u32 {
    let sizeInConcatAxis = array<u32, ${e}u>(${r});
    for (var i: u32 = 0u; i < ${e}; i += 1u ) {
      if (index < sizeInConcatAxis[i]) {
        return i;
      }
    }
    return ${e}u;
  }`,Hg=(e,r)=>{let t=e.length,n=[];for(let o=0;o<t;++o){let i=r.setByOffset("global_idx",e[o].getByIndices("indices"));t===1?n.push(i):o===0?n.push(`if (inputIndex == ${o}u) { ${i} }`):o===t-1?n.push(`else { ${i} }`):n.push(`else if (inputIndex == ${o}) { ${i} }`)}return n.join(`
`)},qg=(e,r,t,n)=>{let o=Fe.size(t),i=new Array(e.length),u=new Array(e.length),d=0,h=[],f=[],y=[{type:12,data:o}];for(let k=0;k<e.length;++k)d+=e[k].dims[r],i[k]=d,f.push(e[k].dims.length),u[k]=Re(`input${k}`,n,f[k]),h.push("rank"),y.push({type:12,data:i[k]});for(let k=0;k<e.length;++k)y.push(...bt(e[k].dims));y.push(...bt(t));let m=mt("output",n,t.length),g=m.indicesGet("indices",r),v=Array.from(Array(i.length).keys()).map(k=>`uniforms.sizeInConcatAxis${k}`).join(","),b=k=>`

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

  ${Kg(i.length,v)}

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

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

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

    ${Hg(u,m)}
  }`;return{name:"Concat",shaderCache:{hint:`${r}`,inputDependencies:h},getRunData:()=>({outputs:[{dims:t,dataType:n}],dispatchGroup:{x:Math.ceil(o/64)},programUniforms:y}),getShaderSource:b}},fb=(e,r)=>{let t=e.inputs,n=t[0].dims,o=Fe.normalizeAxis(r.axis,n.length);Gg(t,o);let i=n.slice();i[o]=t.reduce((d,h)=>d+(h.dims.length>o?h.dims[o]:0),0);let u=t.filter(d=>Fe.size(d.dims)>0);e.compute(qg(u,o,i,t[0].dataType),{inputs:u})},mb=e=>Zt({axis:e.axis})}),xn,Tn,En,u_,Cn=Je(()=>{St(),Ft(),xn=(e,r,t="f32")=>{switch(e.activation){case"Relu":return`value = max(value, ${r}(0.0));`;case"Sigmoid":return`value = (${r}(1.0) / (${r}(1.0) + exp(-value)));`;case"Clip":return`value = clamp(value, ${r}(${t}(uniforms.clip_min)), ${r}(${t}(uniforms.clip_max)));`;case"HardSigmoid":return`value = max(${r}(0.0), min(${r}(1.0), ${t}(uniforms.alpha) * value + ${t}(uniforms.beta)));`;case"LeakyRelu":return`value = select(${t}(uniforms.alpha) * value, value, value >= ${r}(0.0));`;case"Tanh":return`let e2x = exp(-2.0 * abs(value));
              value = sign(value) * (1.0 - e2x) / (1.0 + e2x);
        `;case"":return"";default:throw new Error(`Unsupported activation ${e.activation}`)}},Tn=(e,r)=>{e.activation==="Clip"?r.push({type:1,data:e.clipMax},{type:1,data:e.clipMin}):e.activation==="HardSigmoid"?r.push({type:1,data:e.alpha},{type:1,data:e.beta}):e.activation==="LeakyRelu"&&r.push({type:1,data:e.alpha})},En=(e,r)=>{e.activation==="Clip"?r.push({name:"clip_max",type:"f32"},{name:"clip_min",type:"f32"}):e.activation==="HardSigmoid"?r.push({name:"alpha",type:"f32"},{name:"beta",type:"f32"}):e.activation==="LeakyRelu"&&r.push({name:"alpha",type:"f32"})},u_=e=>{let r=(e==null?void 0:e.activation)||"";if(r==="HardSigmoid"){let[t,n]=(e==null?void 0:e.activation_params)||[.2,.5];return{activation:r,alpha:t,beta:n}}else if(r==="Clip"){let[t,n]=(e==null?void 0:e.activation_params)||[N0,V0];return{activation:r,clipMax:n,clipMin:t}}else if(r==="LeakyRelu"){let[t]=(e==null?void 0:e.activation_params)||[.01];return{activation:r,alpha:t}}return{activation:r}}}),Wr,_b,d_=Je(()=>{Wr=(e,r)=>{switch(e){case 1:return r;case 2:return`vec2<${r}>`;case 3:return`vec3<${r}>`;case 4:return`vec4<${r}>`;default:throw new Error(`${e}-component is not supported.`)}},_b=e=>`
      ${e?"value = value + getBiasByOutputCoords(coords);":""}
      `}),gb,fT=Je(()=>{gb=e=>`
fn getIndexFromCoords4D(coords : vec4<i32>, shape : vec4<i32>) -> i32 {
  return dot(coords, vec4<i32>(
      shape.y * shape.z * shape.w, shape.z * shape.w, shape.w, 1));
}
fn getOutputIndexFromCoords(coords : vec4<i32>) -> i32 {
  return dot(coords, vec4<i32>(
    i32(${e}.x), i32(${e}.y), i32(${e}.z), 1));
}
`}),md,p_,h_=Je(()=>{St(),Ft(),Lt(),Cn(),md=(e,r,t,n,o)=>{let i=n-t;return`
      ${Array.from({length:t}).map((u,d)=>`
      if (${wt(r.shape,d,r.rank)} != 1) {
        ${r.indicesSet(e,d,wt(o,d+i,n))}
      } else {
        ${r.indicesSet(e,d,0)}
      }`).join("")}
`},p_=(e,r,t,n,o=!1,i)=>{let u=e[0].dims,d=e[1].dims,h=u[u.length-2],f=d[d.length-1],y=u[u.length-1],m=yr(f),g=yr(y),v=yr(h),b=Fe.size(t)/m/v,k=e.length>2,z=n?n.slice(0,-2):t.slice(0,-2),E=[Fe.size(z),h,f],T=[{type:12,data:b},{type:12,data:h},{type:12,data:f},{type:12,data:y}];Tn(r,T),T.push(...bt(z,u,d)),k&&T.push(...bt(e[2].dims)),T.push(...bt(E));let D=I=>{let $=a_("batch_dims",e[0].dataType,z.length),P=Re("a",e[0].dataType,u.length,g),S=Re("b",e[1].dataType,d.length,m),O=mt("output",e[0].dataType,E.length,m),R=jr(O.type.tensor),G=xn(r,O.type.value,R),ee=[P,S],ie="";if(k){let re=o?m:1;ee.push(Re("bias",e[2].dataType,e[2].dims.length,re)),ie=`${o?`value += bias[col / ${re}];`:`value += ${O.type.value}(bias[row + i]);`}`}let H=[{name:"output_size",type:"u32"},{name:"M",type:"u32"},{name:"N",type:"u32"},{name:"K",type:"u32"}];En(r,H);let ce=()=>{let re=`var a_data: ${P.type.value};`;for(let se=0;se<g;se++)re+=`
              let b_data${se} = b[(b_offset + (k + ${se}) * uniforms.N + col) / ${m}];`;for(let se=0;se<v;se++){re+=`a_data = a[(a_offset + (row + ${se}) * uniforms.K + k) / ${g}];`;for(let _e=0;_e<g;_e++)re+=`
            values[${se}] = fma(${S.type.value}(a_data${g===1?"":`[${_e}]`}), b_data${_e}, values[${se}]);
`}return re};return`
  ${I.registerUniforms(H).registerInternalVariables($).declareVariables(...ee,O)}
  ${I.mainStart()}
    ${I.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.output_size")}
    let col = (global_idx % (uniforms.N / ${m})) * ${m};
    var index1 = global_idx / (uniforms.N / ${m});
    let stride1 = uniforms.M / ${v};
    let row = (index1 % stride1) * ${v};
    let batch = index1 / stride1;

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

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

    var b_indices: ${S.type.indices};
    ${md("b_indices",S,S.rank-2,$.rank,"batch_indices")}
    ${S.indicesSet("b_indices",S.rank-2,0)}
    ${S.indicesSet("b_indices",S.rank-1,0)}
    let b_offset = ${S.indicesToOffset("b_indices")};
    var values: array<${O.type.value}, ${v}>;
    for (var k: u32 = 0u; k < uniforms.K; k = k + ${g}) {
      ${ce()}
    }
    for (var i = 0u; i < ${v}u; i++) {
      var value = values[i];
      ${ie}
      ${G}
      let cur_indices = ${O.type.indices}(batch, row + i, col);
      let offset = ${O.indicesToOffset("cur_indices")};
      ${O.setByOffset(`offset / ${m}`,"value")};
    }
  }
  `};return{name:"MatMulNaive",shaderCache:{hint:`${r.activation};${m};${g};${v};${o}`,inputDependencies:k?["rank","rank","rank"]:["rank","rank"]},getRunData:()=>({outputs:[{dims:i?i(t):t,dataType:e[0].dataType}],dispatchGroup:{x:Math.ceil(b/64)},programUniforms:T}),getShaderSource:D}}}),Qg,Xg,Rm,Zf,Jg,Nm,Yg,nf,f_=Je(()=>{St(),Ft(),Lt(),Cn(),h_(),d_(),Qg=(e,r)=>e?`
        mm_Asub[inputRow][inputCol] = mm_readA(batch,
          kStart + inputRow,
          globalRowStart / innerElementSize + inputCol${r?", batchIndices":""});
        `:`
        mm_Asub[inputRow][inputCol] = mm_readA(batch,
          globalRow + innerRow,
          kStart / innerElementSize + inputCol${r?", batchIndices":""});
        `,Xg=(e,r)=>e?`
        let ACached0 = mm_Asub[k * innerElementSize][localRow];
        let ACached1 = mm_Asub[k * innerElementSize + 1][localRow];
        let ACached2 = mm_Asub[k * innerElementSize + 2][localRow];
        ${r===3?"":"let ACached3 = mm_Asub[k * innerElementSize + 3][localRow];"}
        for (var i = 0; i < rowPerThread; i = i + 1) {
          acc[i] = BCached0 * ACached0[i] + acc[i];
          acc[i] = BCached1 * ACached1[i] + acc[i];
          acc[i] = BCached2 * ACached2[i] + acc[i];
          ${r===3?"":"acc[i] = BCached3 * ACached3[i] + acc[i];"}
        }`:`
        for (var i = 0; i < rowPerThread; i = i + 1) {
          let ACached = mm_Asub[tileRow + i][k];
          acc[i] = BCached0 * ACached.x + acc[i];
          acc[i] = BCached1 * ACached.y + acc[i];
          acc[i] = BCached2 * ACached.z + acc[i];
          ${r===3?"":"acc[i] = BCached3 * ACached.w + acc[i];"}
        }`,Rm=(e,r,t="f32",n,o=!1,i=32,u=!1,d=32)=>{let h=r[1]*e[1],f=r[0]*e[0],y=o?h:i,m=o?i:h,g=y/r[0],v=i/r[1];if(!((o&&g===4&&e[1]===4||!o&&(g===3||g===4))&&y%r[0]===0&&i%r[1]===0&&e[0]===4))throw new Error(`If transposeA ${o} is true, innerElementSize ${g} and workPerThread[1] ${e[1]} must be 4.
      Otherwise, innerElementSize ${g} must be 3 or 4.
  tileAWidth ${y} must be divisible by workgroupSize[0]${r[0]}. tileInner ${i} must be divisible by workgroupSize[1] ${r[1]}. colPerThread ${e[0]} must be 4.`);return`
var<workgroup> mm_Asub: array<array<vec${g}<${t}>, ${y/g}>, ${m}>;
var<workgroup> mm_Bsub: array<array<vec4<${t}>, ${f/e[0]}>, ${i}>;

const rowPerThread = ${e[1]};
const colPerThread = ${e[0]};
const innerElementSize = ${g};
const tileInner = ${i};

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

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

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

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

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

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

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

          ${Xg(o,g)}
      }

      workgroupBarrier();
  }

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

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

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

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

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

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

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

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

  workgroupBarrier();
}

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

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

    var acc : array<array<${t}, colPerThread>, rowPerThread>;
    ${z}
  }
`},Yg=(e,r,t,n,o=!1)=>{let[i,u,d,h]=n,f=jr(n[0].type.tensor);return`
    fn mm_readA(batch: i32, row: i32, colIn: i32, batchIndices: ${i.type.indices}) -> ${Wr(e,f)} {
      var value = ${Wr(e,f)}(0.0);
      let col = colIn * ${e};
      if(row < uniforms.dim_a_outer && col < uniforms.dim_inner)
      {
        var aIndices: ${u.type.indices};
        ${md("aIndices",u,u.rank-2,i.rank,"batchIndices")}
        ${u.indicesSet("aIndices",u.rank-2,"u32(row)")}
        ${u.indicesSet("aIndices",u.rank-1,"u32(colIn)")}
        value = ${u.getByIndices("aIndices")};
      }
      return value;
    }

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

    fn mm_write(batch: i32, row: i32, colIn: i32, valueIn: ${Wr(e,f)}) {
      let col = colIn * ${e};
      if (row < uniforms.dim_a_outer && col < uniforms.dim_b_outer) {
        var value = valueIn;
        let coords = vec3<i32>(batch, row, colIn);
        ${r?`value = value + ${o?"bias[colIn]":`${Wr(e,f)}(bias[row])`};`:""}
        ${t}
        ${h.setByIndices("vec3<u32>(coords)","value")}
      }
    }
    `},nf=(e,r,t,n,o=!1,i)=>{let u=e[0].dims,d=e[1].dims,h=u.slice(0,-2),f=d.slice(0,-2),y=n?n.slice(0,-2):t.slice(0,-2),m=Fe.size(y),g=u[u.length-2],v=u[u.length-1],b=d[d.length-1],k=v%4===0&&b%4===0,z=g<=8?[4,1,1]:[4,4,1],E=[8,8,1],T=[Math.ceil(b/E[0]/z[0]),Math.ceil(g/E[1]/z[1]),Math.ceil(m/E[2]/z[2])],D=k?4:1,I=[...h,g,v/D],$=I.length,P=[...f,v,b/D],S=P.length,O=[m,g,b/D],R=[{type:6,data:g},{type:6,data:b},{type:6,data:v}];Tn(r,R),R.push(...bt(y,I,P));let G=["rank","rank"],ee=e.length>2;ee&&(R.push(...bt(e[2].dims)),G.push("rank")),R.push(...bt(O));let ie=H=>{let ce=y.length,re=a_("batchDims",e[0].dataType,ce,1),se=jr(e[0].dataType),_e=Re("a",e[0].dataType,$,D),ae=Re("b",e[1].dataType,S,D),Ce=mt("result",e[0].dataType,O.length,D),Te=[_e,ae];if(ee){let Ie=o?D:1;Te.push(Re("bias",e[2].dataType,e[2].dims.length,Ie))}let q=[{name:"dim_a_outer",type:"i32"},{name:"dim_b_outer",type:"i32"},{name:"dim_inner",type:"i32"}];En(r,q);let B=jr(Ce.type.tensor),Q=xn(r,Ce.type.value,B),oe=Yg(D,ee,Q,[re,_e,ae,Ce],o);return`
  ${H.registerUniforms(q).registerInternalVariables(re).declareVariables(...Te,Ce)}
  ${oe}
  ${k?Rm(z,E,se,re):Nm(z,E,se,re)}
                   `};return{name:"MatMul",shaderCache:{hint:`${z};${r.activation};${k};${o}`,inputDependencies:G},getRunData:()=>({outputs:[{dims:i?i(t):t,dataType:e[0].dataType}],dispatchGroup:{x:T[0],y:T[1],z:T[2]},programUniforms:R}),getShaderSource:ie}}}),Zg,yb,mT=Je(()=>{St(),vs(),Lt(),Cn(),d_(),fT(),f_(),Zg=(e,r,t,n,o=!1,i,u=4,d=4,h=4,f="f32")=>{let y=R=>{switch(R){case 1:return"resData = x[xIndex];";case 3:return`resData = vec3<${f}>(x[xIndex], x[xIndex + 1], x[xIndex + 2]);`;case 4:return"resData = x[xIndex / 4];";default:throw new Error(`innerElementSize ${R} is not supported.`)}},m=R=>{switch(R){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 ${R} is not supported.`)}},g=e?`
    let coord = vec4<i32>(batch, xRow, xCol, xCh);
    `:`
    let coord = vec4<i32>(batch, xCh, xRow, xCol);
    `,v=e?`
    let coords = vec4<i32>(
      batch,
      row / outWidth,
      row % outWidth,
      col);
    `:`
    let coords = vec4<i32>(
      batch,
      row,
      col / outWidth,
      col % outWidth);
    `,b=e?"i32(uniforms.x_shape[1])":"i32(uniforms.x_shape[2])",k=e?"i32(uniforms.x_shape[2])":"i32(uniforms.x_shape[3])",z=e?"row":"col",E=e?"col":"row",T=`
    let inChannels = i32(uniforms.w_shape[2]);
    let outWidth = ${e?"i32(uniforms.result_shape[2])":"i32(uniforms.result_shape[3])"};
    let outRow = ${z} / outWidth;
    let outCol = ${z} % outWidth;

    let WRow = ${E} / (i32(uniforms.w_shape[1]) * inChannels);
    let WCol = ${E} / inChannels % i32(uniforms.w_shape[1]);
    let xRow = outRow * uniforms.stride[0] + uniforms.dilation[0] * WRow - uniforms.pad[0];
    let xCol = outCol * uniforms.stride[1] + uniforms.dilation[1] * WCol - uniforms.pad[1];
    let xCh = ${E} % inChannels;
    var resData = ${Wr(u,f)}(0.0);
    // The bounds checking is always needed since we use it to pad zero for
    // the 'same' padding type.
    if (xRow >= 0 && xRow < ${b} && xCol >= 0 && xCol < ${k}) {
      ${g}
      let xIndex = getIndexFromCoords4D(coord, vec4<i32>(uniforms.x_shape));
      ${y(u)}
    }
    return resData;`,D=e?r&&n?`
    let col = colIn * ${u};
    ${T}`:`
    let col = colIn * ${u};
    if (row < uniforms.dim_a_outer && col < uniforms.dim_inner) {
      ${T}
    }
    return ${Wr(u,f)}(0.0);`:n&&t?`
    let col = colIn * ${u};
    ${T}`:`
    let col = colIn * ${u};
    if (row < uniforms.dim_inner && col < uniforms.dim_b_outer) {
      ${T}
    }
    return ${Wr(u,f)}(0.0);`,I=e?n&&t?m(d):`
    let col = colIn * ${d};
    if (row < uniforms.dim_inner && col < uniforms.dim_b_outer) {
      ${m(d)}
    }
    return ${Wr(d,f)}(0.0);`:`
    let col = colIn * ${d};
    if (row < uniforms.dim_inner && col < uniforms.dim_a_outer) {
      ${m(d)}
    }
    return ${Wr(d,f)}(0.0);`,$=Wr(h,f),P=Wr(e?u:d,f),S=Wr(e?d:u,f),O=xn(i,$,f);return`
    fn mm_readA(batch: i32, row : i32, colIn : i32) -> ${P} {
      ${e?D:I}
    }

    fn mm_readB(batch: i32, row : i32, colIn : i32) -> ${S} {
      ${e?I:D}
    }

    fn mm_write(batch: i32, row : i32, colIn : i32, valueIn : ${$}) {
      let col = colIn * ${h};
      if (row < uniforms.dim_a_outer && col < uniforms.dim_b_outer)
      {
      var value = valueIn;
      let outWidth = ${e?"i32(uniforms.result_shape[2])":"i32(uniforms.result_shape[3])"};
      ${v}
      ${_b(o)}
      ${O}
      setOutputAtCoords(coords[0], coords[1], coords[2], coords[3], value);
      }
    }`},yb=(e,r,t,n,o,i,u,d,h)=>{let f=r.format==="NHWC",y=f?e[0].dims[3]:e[0].dims[1],m=t[0],g=f?t[2]:t[3],v=f?t[1]:t[2],b=f?t[3]:t[1],k=f&&(y%4===0||y%3===0)&&b%4===0,z=f?b:g*v,E=f?g*v:b,T=[8,8,1],D=n<=8?[4,1,1]:[4,4,1],I=[Math.ceil(z/T[0]/D[0]),Math.ceil(E/T[1]/D[1]),Math.ceil(m/T[2]/D[2])];Gt("verbose",()=>`[conv2d_mm_webgpu] dispatch = ${I}`);let $=k?f&&y%4!==0?3:4:1,P=T[1]*D[1],S=T[0]*D[0],O=Math.max(T[0]*$,T[1]),R=n%P===0,G=o%S===0,ee=i%O===0,ie=k?[$,4,4]:[1,1,1],H=[{type:6,data:n},{type:6,data:o},{type:6,data:i},{type:6,data:[r.pads[0],r.pads[1]]},{type:6,data:r.strides},{type:6,data:r.dilations}];Tn(r,H),H.push(...bt(e[0].dims,e[1].dims));let ce=["rank","rank"];u&&(H.push(...bt(e[2].dims)),ce.push("rank")),H.push(...bt(t));let re=se=>{let _e=[{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}];En(r,_e);let ae=k?4:1,Ce=jr(e[0].dataType),Te=`
      fn setOutputAtIndex(flatIndex : i32, value : ${k?`vec4<${Ce}>`:Ce}) {
        result[flatIndex] = ${k?`vec4<${Ce}>`:Ce}(value);
      }
      fn setOutputAtCoords(d0 : i32, d1 : i32, d2 : i32, d3 : i32, value : ${k?`vec4<${Ce}>`:Ce}) {
        let flatIndex = getOutputIndexFromCoords(vec4<i32>(d0, d1, d2, d3));
        setOutputAtIndex(flatIndex ${k?"/ 4":""}, value);
      }`,q=Re("x",e[0].dataType,e[0].dims.length,$===3?1:$),B=Re("w",e[1].dataType,e[1].dims.length,ae),Q=[q,B],oe=mt("result",e[0].dataType,t.length,ae);if(u){let Ie=Re("bias",e[2].dataType,e[2].dims.length,ae);Q.push(Ie),Te+=`
        fn getBiasByOutputCoords(coords : vec4<i32>) -> ${k?`vec4<${Ce}>`:Ce} {
          return bias[coords.${f?"w":"y"}${k?"/ 4":""}];
        }`}return`
        ${gb("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 };
        ${se.registerUniforms(_e).declareVariables(...Q,oe)}
        ${Te}
        ${Zg(f,R,G,ee,u,r,ie[0],ie[1],ie[2],Ce)}
        ${k?Rm(D,T,Ce,void 0,!f,O):Nm(D,T,Ce,void 0,!f,O,!1,void 0,d)}`};return{name:"Conv2DMatMul",shaderCache:{hint:`${r.cacheKey};${$};${k};${R};${G};${ee};${P};${S};${O}`,inputDependencies:ce},getRunData:()=>({outputs:[{dims:h?h(t):t,dataType:e[0].dataType}],dispatchGroup:{x:I[0],y:I[1],z:I[2]},programUniforms:H}),getShaderSource:re}}}),ey,em,nd,ty,tm,ry,wb,Mb,_T=Je(()=>{St(),vs(),Ft(),Lt(),Cn(),d_(),ey=e=>{let r=1;for(let t=0;t<e.length;t++)r*=e[t];return r},em=e=>typeof e=="number"?[e,e,e]:e,nd=(e,r)=>r<=1?e:e+(e-1)*(r-1),ty=(e,r,t,n=1)=>{let o=nd(r,n);return Math.floor((e[0]*(t-1)-t+o)/2)},tm=(e,r,t,n,o)=>{o==null&&(o=ty(e,r[0],n[0]));let i=[0,0,0,t];for(let u=0;u<3;u++)e[u]+2*o>=r[u]&&(i[u]=Math.trunc((e[u]-r[u]+2*o)/n[u]+1));return i},ry=(e,r,t,n,o,i,u,d,h,f)=>{let y,m,g,v;if(e==="VALID"&&(e=0),typeof e=="number"){y={top:e,bottom:e,left:e,right:e,front:e,back:e};let b=tm([r,t,n,1],[d,h,f],1,[o,i,u],e);m=b[0],g=b[1],v=b[2]}else if(Array.isArray(e)){if(!e.every((k,z,E)=>k===E[0]))throw Error(`Unsupported padding parameter: ${e}`);y={top:e[0],bottom:e[1],left:e[2],right:e[3],front:e[4],back:e[5]};let b=tm([r,t,n,1],[d,h,f],1,[o,i,u],e[0]);m=b[0],g=b[1],v=b[2]}else if(e==="SAME_UPPER"){m=Math.ceil(r/o),g=Math.ceil(t/i),v=Math.ceil(n/u);let b=(m-1)*o+d-r,k=(g-1)*i+h-t,z=(v-1)*u+f-n,E=Math.floor(b/2),T=b-E,D=Math.floor(k/2),I=k-D,$=Math.floor(z/2),P=z-$;y={top:D,bottom:I,left:$,right:P,front:E,back:T}}else throw Error(`Unknown padding parameter: ${e}`);return{padInfo:y,outDepth:m,outHeight:g,outWidth:v}},wb=(e,r,t,n,o,i=!1,u="channelsLast")=>{let d,h,f,y,m;if(u==="channelsLast")[d,h,f,y,m]=e;else if(u==="channelsFirst")[d,m,h,f,y]=e;else throw new Error(`Unknown dataFormat ${u}`);let[g,,v,b,k]=r,[z,E,T]=em(t),[D,I,$]=em(n),P=nd(v,D),S=nd(b,I),O=nd(k,$),{padInfo:R,outDepth:G,outHeight:ee,outWidth:ie}=ry(o,h,f,y,z,E,T,P,S,O),H=i?g*m:g,ce=[0,0,0,0,0];return u==="channelsFirst"?ce=[d,H,G,ee,ie]:u==="channelsLast"&&(ce=[d,G,ee,ie,H]),{batchSize:d,dataFormat:u,inDepth:h,inHeight:f,inWidth:y,inChannels:m,outDepth:G,outHeight:ee,outWidth:ie,outChannels:H,padInfo:R,strideDepth:z,strideHeight:E,strideWidth:T,filterDepth:v,filterHeight:b,filterWidth:k,effectiveFilterDepth:P,effectiveFilterHeight:S,effectiveFilterWidth:O,dilationDepth:D,dilationHeight:I,dilationWidth:$,inShape:e,outShape:ce,filterShape:r}},Mb=(e,r,t,n,o,i)=>{let u=i==="channelsLast";u?e[0].dims[3]:e[0].dims[1];let d=[64,1,1],h={x:t.map((z,E)=>E)},f=[Math.ceil(ey(h.x.map(z=>t[z]))/d[0]),1,1];Gt("verbose",()=>`[conv3d_naive_webgpu] dispatch = ${f}`);let y=1,m=Fe.size(t),g=[{type:12,data:m},{type:12,data:n},{type:12,data:o},{type:12,data:r.strides},{type:12,data:r.dilations}];Tn(r,g),g.push(...bt(e[0].dims,e[1].dims));let v=["rank","rank"],b=e.length===3;b&&(g.push(...bt(e[2].dims)),v.push("rank")),g.push(...bt(t));let k=z=>{let E=[{name:"output_size",type:"u32"},{name:"filter_dims",type:"u32",length:n.length},{name:"pads",type:"u32",length:o.length},{name:"strides",type:"u32",length:r.strides.length},{name:"dilations",type:"u32",length:r.dilations.length}];En(r,E);let T=1,D=jr(e[0].dataType),I=Re("x",e[0].dataType,e[0].dims.length,y),$=Re("W",e[1].dataType,e[1].dims.length,T),P=[I,$],S=mt("result",e[0].dataType,t.length,T),O="";if(b){let ee=Re("bias",e[2].dataType,e[2].dims.length,T);P.push(ee),O+=`
        fn getBiasByOutputCoords(coords : array<u32, 5>) -> ${D} {
          return bias[${u?wt("coords",4,5):wt("coords",1,5)}];
        }`}let R=Wr(y,D),G=xn(r,R,D);return`
            ${O}
            fn getX(d0 : u32, d1 : u32, d2 : u32, d3 : u32, d4 : u32) -> f32 {
              let aIndices = array<u32, 5>(d0, d1, d2, d3, d4);
              return ${I.getByIndices("aIndices")};
            }
            fn getW(d0 : u32, d1 : u32, d2 : u32, d3 : u32, d4 : u32) -> f32 {
              let aIndices = array<u32, 5>(d0, d1, d2, d3, d4);
              return ${$.getByIndices("aIndices")};
            }
          ${z.registerUniforms(E).declareVariables(...P,S)}
          ${z.mainStart()}
          ${z.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.output_size")}
              let coords = ${S.offsetToIndices("global_idx")};
              let batch = ${wt("coords",0,I.rank)};
              let d2 = ${u?wt("coords",I.rank-1,I.rank):wt("coords",1,I.rank)};
              let xFRCCorner = vec3<u32>(${u?wt("coords",1,I.rank):wt("coords",2,I.rank)},
              ${u?wt("coords",2,I.rank):wt("coords",3,I.rank)},
              ${u?wt("coords",3,I.rank):wt("coords",4,I.rank)}) * uniforms.strides - uniforms.pads;
              let xFCorner = xFRCCorner.x;
              let xRCorner = xFRCCorner.y;
              let xCCorner = xFRCCorner.z;
              let xShapeY = ${u?wt("uniforms.x_shape",1,I.rank):wt("uniforms.x_shape",2,I.rank)};
              let xShapeZ = ${u?wt("uniforms.x_shape",2,I.rank):wt("uniforms.x_shape",3,I.rank)};
              let xShapeW = ${u?wt("uniforms.x_shape",3,I.rank):wt("uniforms.x_shape",4,I.rank)};
              let xShapeU = ${u?wt("uniforms.x_shape",4,I.rank):wt("uniforms.x_shape",1,I.rank)};
              let inputDepthNearestVec4 = (xShapeU / 4) * 4;
              let inputDepthVec4Remainder = xShapeU % 4;

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

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

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

                    for (var d1 = 0u; d1 < inputDepthNearestVec4; d1 += 4) {
                      ${u?`let xValues = vec4<f32>(
                               getX(batch, xF, xR, xC, d1),
                               getX(batch, xF, xR, xC, d1 + 1),
                               getX(batch, xF, xR, xC, d1 + 2),
                               getX(batch, xF, xR, xC, d1 + 3));
                            `:`let xValues = vec4<f32>(
                               getX(batch, d1, xF, xR, xC),
                               getX(batch, d1 + 1, xF, xR, xC),
                               getX(batch, d1 + 2, xF, xR, xC),
                               getX(batch, d1 + 3, xF, xR, xC));
                            `}
                            let wValues = vec4<f32>(
                              getW(d2, d1, wF, wR, wC),
                              getW(d2, d1 + 1, wF, wR, wC),
                              getW(d2, d1 + 2, wF, wR, wC),
                              getW(d2, d1 + 3, wF, wR, wC));
                      value += dot(xValues, wValues);
                    }
                    if (inputDepthVec4Remainder == 1) {
                        ${u?`value += getX(batch, xF, xR, xC, inputDepthNearestVec4)
                          * getW(d2, inputDepthNearestVec4, wF, wR, wC);`:`value += getX(batch, inputDepthNearestVec4, xF, xR, xC)
                          * getW(d2, inputDepthNearestVec4, wF, wR, wC);`}
                    } else if (inputDepthVec4Remainder == 2) {
                      ${u?`let xValues = vec2<f32>(
                        getX(batch, xF, xR, xC, inputDepthNearestVec4),
                        getX(batch, xF, xR, xC, inputDepthNearestVec4 + 1));
                      `:`let xValues = vec2<f32>(
                        getX(batch, inputDepthNearestVec4, xF, xR, xC),
                        getX(batch, inputDepthNearestVec4 + 1, xF, xR, xC));
                    `}
                    let wValues = vec2<f32>(
                      getW(d2, inputDepthNearestVec4, wF, wR, wC),
                      getW(d2, inputDepthNearestVec4 + 1, wF, wR, wC));
                      value += dot(xValues, wValues);
                    } else if (inputDepthVec4Remainder == 3) {
                      ${u?`let xValues = vec3<f32>(
                        getX(batch, xF, xR, xC, inputDepthNearestVec4),
                        getX(batch, xF, xR, xC, inputDepthNearestVec4 + 1),
                        getX(batch, xF, xR, xC, inputDepthNearestVec4 + 2));
                      `:`let xValues = vec3<f32>(
                        getX(batch, inputDepthNearestVec4, xF, xR, xC),
                        getX(batch, inputDepthNearestVec4 + 1, xF, xR, xC),
                        getX(batch, inputDepthNearestVec4 + 2, xF, xR, xC));
                    `}
                    let wValues = vec3<f32>(
                      getW(d2, inputDepthNearestVec4, wF, wR, wC),
                      getW(d2, inputDepthNearestVec4 + 1, wF, wR, wC),
                      getW(d2, inputDepthNearestVec4 + 2, wF, wR, wC));
                      value += dot(xValues, wValues);
                    }
                  }
                }
              }
              ${b?"value = value + getBiasByOutputCoords(coords)":""};
              ${G}
              result[global_idx] = f32(value);
          }`};return{name:"Conv3DNaive",shaderCache:{hint:`${r.cacheKey};${u};${y};${b}`,inputDependencies:v},getRunData:()=>({outputs:[{dims:t,dataType:e[0].dataType}],dispatchGroup:{x:f[0],y:f[1],z:f[2]},programUniforms:g}),getShaderSource:k}}}),bb,vb,gT=Je(()=>{St(),Ft(),Lt(),Cn(),bb=(e,r,t,n)=>{let o=e.length>2,i=o?"value += b[output_channel];":"",u=e[0].dims,d=e[1].dims,h=r.format==="NHWC",f=h?t[3]:t[1],y=f/r.group,m=h&&y>=4?yr(f):1,g=Fe.size(t)/m,v=[{type:12,data:g},{type:12,data:r.dilations},{type:12,data:[r.strides[0],r.strides[1]]},{type:12,data:[r.pads[0],r.pads[1]]},{type:12,data:y}];Tn(r,v),v.push(...bt(u,[d[0],d[1],d[2],d[3]/m]));let b=o?["rank","rank","rank"]:["rank","rank"];v.push(...bt([t[0],t[1],t[2],t[3]/m]));let k=z=>{let E=mt("output",e[0].dataType,t.length,m),T=jr(E.type.tensor),D=xn(r,E.type.value,T),I=Re("x",e[0].dataType,u.length),$=Re("w",e[1].dataType,d.length,m),P=[I,$];o&&P.push(Re("b",e[2].dataType,e[2].dims,m));let S=[{name:"output_size",type:"u32"},{name:"dilations",type:"u32",length:r.dilations.length},{name:"strides",type:"u32",length:2},{name:"pads",type:"u32",length:2},{name:"output_channels_per_group",type:"u32"}];En(r,S);let O=h?`
      for (var wHeight: u32 = 0u; wHeight < uniforms.w_shape[0]; wHeight++) {
        let xHeight = xRCCorner.x + wHeight * uniforms.dilations[0];

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

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

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

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

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

            let xVal = ${I.get("batch","input_channel","xHeight","xWidth")};
            let wVal = ${$.get("output_channel","wInChannel","wHeight","wWidth")};
            value += xVal * wVal;
          }
        }
      }
      `;return`
  ${z.registerUniforms(S).declareVariables(...P,E)}

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

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

    var value: ${E.type.value} = ${E.type.value}(0);
    ${O}
    ${i}
    ${D}
    ${E.setByOffset("global_idx","value")}
  }`};return{name:"GroupedConv",shaderCache:{hint:`${r.cacheKey}_${m}`,inputDependencies:b},getRunData:()=>({outputs:[{dims:n?n(t):t,dataType:e[0].dataType}],dispatchGroup:{x:Math.ceil(g/64)},programUniforms:v}),getShaderSource:k}},vb=(e,r,t,n)=>{let o=e.length>2,i=yr(t[3]),u=yr(t[2]),d=Fe.size(t)/i/u,h=[e[0].dims[0],e[0].dims[1],e[0].dims[2],e[0].dims[3]/i],f=[e[1].dims[0],e[1].dims[1],e[1].dims[2],e[1].dims[3]/i],y=[t[0],t[1],t[2],t[3]/i],m=[{type:12,data:d},{type:6,data:[r.strides[0],r.strides[1]]},{type:6,data:[r.pads[0],r.pads[1]]}];Tn(r,m),m.push(...bt(h,f,y));let g=(u-1)*r.strides[1]+f[1],v=b=>{let k=mt("output",e[0].dataType,y.length,i),z=jr(k.type.tensor),E=xn(r,k.type.value,z),T=Re("x",e[0].dataType,h.length,i),D=Re("w",e[1].dataType,f.length,i),I=[T,D];o&&I.push(Re("b",e[2].dataType,e[2].dims,i));let $=o?"value += b[output_channel];":"",P=[{name:"output_size",type:"u32"},{name:"strides",type:"i32",length:2},{name:"pads",type:"i32",length:2}];return En(r,P),`
  ${b.registerUniforms(P).declareVariables(...I,k)}
  ${b.mainStart()}
    ${b.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.output_size")}
    let width0 = uniforms.output_shape[3];
    let output_channel = global_idx % width0;
    var index1 = global_idx / width0;
    let width1 = uniforms.output_shape[2] / ${u}u;
    let col = (index1 % width1) * ${u}u;
    index1 = index1 / width1;
    let row = index1 % uniforms.output_shape[1];
    let batch = index1 / uniforms.output_shape[1];

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

    var x_vals: array<${T.type.value}, ${g}>;
    var values: array<${k.type.value}, ${u}>;
    let input_channel = output_channel;
    // Use constant instead of uniform can give better performance for w's height/width.
    for (var w_height: u32 = 0u; w_height < ${f[0]}; w_height++) {
      let x_height = x_corner.x + i32(w_height);
      if (x_height >= 0 && u32(x_height) < uniforms.x_shape[1]) {
        for (var i = 0; i < ${g}; i++) {
          let x_width = x_corner.y + i;
          if (x_width >= 0 && u32(x_width) < uniforms.x_shape[2]) {
            x_vals[i] = ${T.get("batch","u32(x_height)","u32(x_width)","input_channel")};
          } else {
            x_vals[i] = ${T.type.value}(0);
          }
        }
        for (var w_width: u32 = 0u; w_width < ${f[1]}; w_width++) {
          let w_val = ${D.get("w_height","w_width","0","output_channel")};
          for (var i = 0u; i < ${u}u; i++) {
            values[i] = fma(x_vals[i * u32(uniforms.strides[1]) + w_width], w_val, values[i]);
          }
        }
      }
    }

    for (var i = 0u; i < ${u}u; i++) {
      var value = values[i];
      ${$}
      ${E}
      ${k.set("batch","row","col + i","output_channel","value")};
    }
  }`};return{name:"GroupedConv-Vectorize",shaderCache:{hint:`${r.cacheKey};${i};${u};${g};${f[0]};${f[1]}`,inputDependencies:o?["rank","rank","type"]:["rank","rank"]},getRunData:()=>({outputs:[{dims:n?n(t):t,dataType:e[0].dataType}],dispatchGroup:{x:Math.ceil(d/64)},programUniforms:m}),getShaderSource:v}}}),iy,Uh,sy,Wh,Vm,rm,ny,ay,Um,yT=Je(()=>{Ft(),mT(),_T(),f_(),gT(),Cn(),h_(),Ns(),iy=(e,r,t,n,o,i)=>{let u=e[0],d=e.slice(i?1:2,i?3:4),h=d.length,f=r[0],y=r.slice(2).map((g,v)=>g+(g-1)*(t[v]-1)),m=d.map((g,v)=>g+n[v]+n[v+h]).map((g,v)=>Math.floor((g-y[v]+o[v])/o[v]));return m.splice(0,0,u),m.splice(i?3:1,0,f),m},Uh=[2,3,1,0],sy=(e,r)=>{if(!e||e.length!==2&&e.length!==3)throw new Error("Conv requires 2 or 3 inputs");if(e[0].dims.length>5)throw new Error("greater than 5D is not supported");if(e[0].dims.length!==e[1].dims.length)throw new Error("filter does not have same dimension as input");let t=e[0].dims[r.format==="NHWC"?e[0].dims.length-1:1],n=e[1].dims[1]*r.group;if(t!==n)throw new Error("FILTER_IN_CHANNEL should be equal to DATA_CHANNEL");if(e.length===3&&(e[2].dims.length!==1||e[1].dims[0]!==e[2].dims[0]))throw new Error("invalid bias");let o=e[0].dims.length-2;if(r.dilations.length!==o)throw new Error(`dilations should be ${o}D`);if(r.strides.length!==o)throw new Error(`strides should be ${o}D`);if(r.pads.length!==o*2)throw new Error(`pads should be ${o*2}D`);if(r.kernelShape.length!==0&&r.kernelShape.length!==e[1].dims.length-2)throw new Error("invalid kernel shape")},Wh=(e,r)=>{let t=e.kernelShape.slice();t.length<r[1].dims.length-2&&t.push(...Array(r[1].dims.length-2-t.length).fill(0));for(let i=2;i<r[1].dims.length;++i)t[i-2]===0&&(t[i-2]=r[1].dims[i]);let n=e.pads.slice();rf.adjustPadsBasedOnAutoPad(r[0].dims,e.strides,e.dilations,t,n,e.format==="NHWC",e.autoPad);let o=Object.assign({},e);return Object.assign(o,{kernelShape:t,pads:n}),o},Vm=e=>{let r=u_(e),t=e.format,n=["NOTSET","VALID","SAME_UPPER","SAME_LOWER"][e.auto_pad],o=e.dilations,i=e.group,u=e.kernel_shape,d=e.pads,h=e.strides,f=e.w_is_const();return{autoPad:n,format:t,dilations:o,group:i,kernelShape:u,pads:d,strides:h,wIsConst:f,...r,cacheKey:`${e.format};${r.activation};`}},rm=(e,r,t,n)=>{let o=t.format==="NHWC",i=iy(r[0].dims,r[1].dims,t.dilations,t.pads,t.strides,o);if(t.group!==1){let P=[r[0]];if(o){let S=e.kernelCustomData.wT??e.compute(xi(r[1],Uh),{inputs:[1],outputs:[t.wIsConst?-2:-1]})[0];t.wIsConst&&!e.kernelCustomData.wT&&(e.kernelCustomData.wT=S),P.push(S)}else P.push(r[1]);r.length===3&&P.push(r[2]),!e.adapterInfo.isArchitecture("ampere")&&o&&r[1].dims[0]===t.group&&r[1].dims[1]===1&&t.dilations[0]===1&&t.dilations[1]===1?e.compute(vb(P,t,i,n),{inputs:P}):e.compute(bb(P,t,i,n),{inputs:P});return}let u=r.length===3,d=r[0].dims[o?1:2],h=r[0].dims[o?2:3],f=r[0].dims[o?3:1],y=r[1].dims[2],m=r[1].dims[3],g=i[o?1:2],v=i[o?2:3],b=i[o?3:1],k=o&&y===d&&m===h&&t.pads[0]===0&&t.pads[1]===0;if(k||y===1&&m===1&&t.dilations[0]===1&&t.dilations[1]===1&&t.strides[0]===1&&t.strides[1]===1&&t.pads[0]===0&&t.pads[1]===0){let P=i[0],S,O,R,G=[];if(o){let H=e.kernelCustomData.wT??e.compute(xi(r[1],Uh),{inputs:[1],outputs:[t.wIsConst?-2:-1]})[0];if(t.wIsConst&&!e.kernelCustomData.wT&&(e.kernelCustomData.wT=H),k){let ce=d*h*f;S=r[0].reshape([1,P,ce]),O=H.reshape([1,ce,b]),R=[1,P,b]}else S=r[0].reshape([P,d*h,f]),O=H.reshape([1,f,b]),R=[P,g*v,b];G.push(S),G.push(O)}else S=r[0].reshape([P,f,d*h]),O=r[1].reshape([1,b,f]),R=[P,b,g*v],G.push(O),G.push(S);u&&G.push(r[2]);let ee=R[2],ie=G[0].dims[G[0].dims.length-1];ee<8&&ie<8?e.compute(p_(G,t,i,R,o,n),{inputs:G}):e.compute(nf(G,t,i,R,o,n),{inputs:G});return}let z=!0,E=e.kernelCustomData.wT??e.compute(xi(r[1],Uh),{inputs:[1],outputs:[t.wIsConst?-2:-1]})[0];t.wIsConst&&!e.kernelCustomData.wT&&(e.kernelCustomData.wT=E);let T=[r[0],E];u&&T.push(r[2]);let D=o?g*v:b,I=o?b:g*v,$=y*m*f;e.compute(yb(T,t,i,D,I,$,u,z,n),{inputs:T})},ny=(e,r)=>{let t=r.format==="NHWC",n=[e.inputs[0].reshape(t?[e.inputs[0].dims[0],1,e.inputs[0].dims[1],e.inputs[0].dims[2]]:[e.inputs[0].dims[0],e.inputs[0].dims[1],1,e.inputs[0].dims[2]]),e.inputs[1].reshape([e.inputs[1].dims[0],e.inputs[1].dims[1],1,e.inputs[1].dims[2]])];e.inputs.length===3&&n.push(e.inputs[2]);let o=[0,r.pads[0],0,r.pads[1]],i=[1].concat(r.strides),u=[1].concat(r.dilations),d=[1].concat(r.kernelShape),h=Wh({...r,pads:o,strides:i,dilations:u,kernelShape:d},n);rm(e,n,h,f=>t?[f[0],f[2],f[3]]:[f[0],f[1],f[3]])},ay=(e,r,t)=>{let n=t.format==="NHWC"?"channelsLast":"channelsFirst",o=Wh(t,r),i=t.autoPad==="NOTSET"?t.pads:t.autoPad,u=wb(r[0].dims,r[1].dims,t.strides,t.dilations,i,!1,n);e.compute(Mb(r,o,u.outShape,[u.filterDepth,u.filterHeight,u.filterWidth],[u.padInfo.front,u.padInfo.top,u.padInfo.left],n))},Um=(e,r)=>{if(sy(e.inputs,r),e.inputs[0].dims.length===3)ny(e,r);else if(e.inputs[0].dims.length===5)ay(e,e.inputs,r);else{let t=Wh(r,e.inputs);rm(e,e.inputs,t)}}}),xb,wT=Je(()=>{St(),vs(),Ft(),Lt(),xb=(e,r,t)=>{let n=e.length>2,o=r.outputShape,i=r.format==="NHWC",u=r.group,d=e[1].dims,h=d[2]/u,f=d[3],y=i?yr(h):1,m=i&&f===1&&h>=4,g=m?Math.floor(h/4)*4:Math.floor(h/y)*y,v=h-g,b=i?yr(f):1,k=i?f===1?y:b:1,z=Fe.size(o)/b,E=[Math.ceil(z/64),1,1];Gt("verbose",()=>`[conv2d_backprop_webgpu] dispatch = ${E}`);let T=["rank","rank"],D=[r.strides[0],r.strides[1]],I=[r.kernelShape[i?1:2],r.kernelShape[i?2:3]],$=[r.dilations[0],r.dilations[1]],P=[I[0]+(r.dilations[0]<=1?0:(r.kernelShape[i?1:2]-1)*(r.dilations[0]-1)),I[1]+(r.dilations[1]<=1?0:(r.kernelShape[i?2:3]-1)*(r.dilations[1]-1))],S=[P[0]-1-Math.floor((r.pads[0]+r.pads[2])/2),P[1]-1-Math.floor((r.pads[1]+r.pads[3])/2)],O=[{type:12,data:z},{type:12,data:D},{type:12,data:I},{type:12,data:$},{type:12,data:P},{type:6,data:S},{type:12,data:g},{type:12,data:h},{type:12,data:f},...bt(e[0].dims,e[1].dims)];n&&(O.push(...bt(e[2].dims)),T.push("rank")),O.push(...bt(o));let R=G=>{let ee=[{name:"output_size",type:"u32"},{name:"strides",type:"u32",length:D.length},{name:"filter_dims",type:"u32",length:I.length},{name:"dilations",type:"u32",length:I.length},{name:"effective_filter_dims",type:"u32",length:P.length},{name:"pads",type:"i32",length:S.length},{name:"input_channels_per_group_int",type:"u32"},{name:"input_channels_per_group",type:"u32"},{name:"output_channels_per_group",type:"u32"}],ie=jr(e[0].dataType),H=i?1:2,ce=i?2:3,re=i?3:1,se=Re("W",e[1].dataType,e[1].dims.length,k),_e=Re("Dy",e[0].dataType,e[0].dims.length,y),ae=[_e,se];n&&ae.push(Re("bias",e[2].dataType,[o[re]].length,b));let Ce=mt("result",e[0].dataType,o.length,b),Te=()=>{let Q="";if(m)y===4?Q+=`
        let xValue = ${_e.getByOffset("x_offset")};
        let wValue = ${se.getByOffset("w_offset")};
        dotProd = dotProd + dot(xValue, wValue);
        x_offset += 1u;
        w_offset += 1u;`:y===2?Q+=`
          dotProd = dotProd + dot(vec4<${ie}>(${_e.getByOffset("x_offset")}, ${_e.getByOffset("x_offset + 1u")}), vec4<${ie}>(${se.getByOffset("w_offset")}, ${se.getByOffset("w_offset + 1u")}));
          x_offset += 2u;
          w_offset += 2u;`:y===1&&(Q+=`
          dotProd = dotProd + dot(vec4<${ie}>(${_e.getByOffset("x_offset")}, ${_e.getByOffset("x_offset + 1u")}, ${_e.getByOffset("x_offset + 2u")}, ${_e.getByOffset("x_offset + 3u")}), vec4<${ie}>(${se.getByOffset("w_offset")}, ${se.getByOffset("w_offset + 1u")}, ${se.getByOffset("w_offset + 2u")}, ${se.getByOffset("w_offset + 3u")}));
          x_offset += 4u;
          w_offset += 4u;`);else if(Q+=`
                  let xValue = ${i?_e.getByOffset(`${_e.indicesToOffset(`${_e.type.indices}(batch, idyR, idyC, inputChannel)`)} / ${y}`):_e.get("batch","inputChannel","idyR","idyC")};
        `,y===1)Q+=`
          let w_offset = ${se.indicesToOffset(`${se.type.indices}(u32(wRPerm), u32(wCPerm), inputChannel, wOutChannel)`)};
          let wValue = ${se.getByOffset(`w_offset / ${k}`)};
          dotProd = dotProd + xValue * wValue;`;else for(let oe=0;oe<y;oe++)Q+=`
            let wValue${oe} = ${se.getByOffset(`${se.indicesToOffset(`${se.type.indices}(u32(wRPerm), u32(wCPerm), inputChannel + ${oe}, wOutChannel)`)} / ${k}`)};
            dotProd = dotProd + xValue[${oe}] * wValue${oe};`;return Q},q=()=>{if(v===0)return"";if(!m)throw new Error(`packInputAs4 ${m} is not true.`);let Q="";if(y===1){Q+="dotProd = dotProd";for(let oe=0;oe<v;oe++)Q+=`
            + ${_e.getByOffset(`x_offset + ${oe}`)} * ${se.getByOffset(`w_offset + ${oe}`)}`;Q+=";"}else if(y===2){if(v!==2)throw new Error(`Invalid inputChannelsRemainder ${v}.`);Q+=`
          let xValue = ${_e.getByOffset("x_offset")};
          let wValue = ${se.getByOffset("w_offset")};
          dotProd = dotProd + dot(xValue, wValue);`}return Q},B=`
            let outputIndices = ${Ce.offsetToIndices(`global_idx * ${b}`)};
            let batch = ${Ce.indicesGet("outputIndices",0)};
            let d1 = ${Ce.indicesGet("outputIndices",re)};
            let r = ${Ce.indicesGet("outputIndices",H)};
            let c = ${Ce.indicesGet("outputIndices",ce)};
            let dyCorner = vec2<i32>(i32(r), i32(c)) - uniforms.pads;
            let dyRCorner = dyCorner.x;
            let dyCCorner = dyCorner.y;
            let groupId = d1 / uniforms.output_channels_per_group;
            let wOutChannel = d1 - groupId * uniforms.output_channels_per_group;
            // Convolve dy(?, ?, d2) with w(:, :, d1, d2) to compute dx(xR, xC, d1).
            // ? = to be determined. : = across all values in that axis.
            var dotProd = ${Ce.type.value}(0.0);
            var wR: u32 = 0;
            if (uniforms.dilations.x == 1) {
              // Minimum wR >= 0 that satisfies (dyRCorner + wR) % (uniforms.strides.x) == 0
              wR = u32(((dyRCorner + i32(uniforms.strides.x) - 1) / i32(uniforms.strides.x)) * i32(uniforms.strides.x) - dyRCorner);
            }
            for (; wR < uniforms.effective_filter_dims.x; wR = wR + 1) {
              if (wR % uniforms.dilations.x != 0) {
                continue;
              }
              let dyR = (${ie}(dyRCorner) + ${ie}(wR)) / ${ie}(uniforms.strides[0]);
              let wRPerm = uniforms.filter_dims.x - 1 - wR / uniforms.dilations.x;
              if (dyR < 0.0 || dyR >= ${ie}(uniforms.Dy_shape[${H}]) || fract(dyR) > 0.0 ||
                  wRPerm < 0) {
                continue;
              }
              let idyR: u32 = u32(dyR);
              var wC: u32 = 0;
              if (uniforms.dilations.y == 1) {
                // Minimum wC >= 0 that satisfies (dyCCorner + wC) % (uniforms.strides.y) == 0
                wC = u32(((dyCCorner + i32(uniforms.strides.y) - 1) / i32(uniforms.strides.y)) * i32(uniforms.strides.y) - dyCCorner);
              }
              for (; wC < uniforms.effective_filter_dims.y; wC = wC + 1) {
                if (wC % uniforms.dilations.y != 0) {
                  continue;
                }
                let dyC = (${ie}(dyCCorner) + ${ie}(wC)) / ${ie}(uniforms.strides.y);
                let wCPerm = uniforms.filter_dims.y - 1 - wC / uniforms.dilations.y;
                if (dyC < 0.0 || dyC >= ${ie}(uniforms.Dy_shape[${ce}]) ||
                    fract(dyC) > 0.0 || wCPerm < 0) {
                  continue;
                }
                let idyC: u32 = u32(dyC);
                var inputChannel = groupId * uniforms.input_channels_per_group;
                ${m?`
                var x_offset = ${_e.indicesToOffset(`${_e.type.indices}(batch, idyR, idyC, inputChannel)`)} / ${y};
                var w_offset = ${se.indicesToOffset(`${se.type.indices}(wRPerm, wCPerm, inputChannel, wOutChannel)`)} / ${k};
                  `:""}
                for (var d2: u32 = 0; d2 < uniforms.input_channels_per_group_int; d2 = d2 + ${m?4:y}) {
                  ${Te()}
                  inputChannel = inputChannel + ${m?4:y};
                }
                ${q()}
                wC = wC + uniforms.strides.y - 1;
              }
              wR = wR + uniforms.strides[0] - 1;
            }
            let value = dotProd${n?` + bias[d1 / ${b}]`:""};
            ${Ce.setByOffset("global_idx","value")};
          `;return`
    ${G.registerUniforms(ee).declareVariables(...ae,Ce)}
      ${G.mainStart()}
      ${G.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.output_size")};
    ${B}}`};return{name:"ConvTranspose2D",shaderCache:{hint:`${r.cacheKey};${y}${k}${b}${m}${v}`,inputDependencies:T},getRunData:()=>({dispatchGroup:{x:E[0],y:E[1],z:E[2]},outputs:[{dims:t?t(o):o,dataType:e[0].dataType}],programUniforms:O}),getShaderSource:R}}}),oy,ly,cy,im,Tb,uy,sm,dy,Eb,MT=Je(()=>{wT(),Cn(),Ns(),oy=(e,r,t,n,o,i)=>(e-1)*r+t+(n-1)*o+1-i,ly=(e,r,t,n,o)=>{let i=Math.floor(e/2);r==="SAME_UPPER"?(t[n]=i,t[o]=e-i):r==="SAME_LOWER"&&(t[n]=e-i,t[o]=i)},cy=(e,r,t,n,o,i,u,d,h,f)=>{let y=e.length-2,m=f.length===0;h.length<y&&h.push(...Array(y-h.length).fill(0));let g=e[0],v=r[d?3:1]*o;for(let b=0,k=e.length-y-(d?1:0);b<y;++b,++k){let z=e[k],E=m?z*u[b]:f[b],T=oy(z,u[b],i[b],r[k],t[b],E);ly(T,n,i,b,b+y),m&&f.push(u[b]*(z-1)+h[b]+(r[k]-1)*t[b]+1-i[b]-i[b+y])}f.splice(0,0,g),f.splice(d?3:1,0,v)},im=(e,r)=>{let t=e.kernelShape.slice();if(e.kernelShape.length===0||e.kernelShape.reduce((m,g)=>m*g,1)===0){t.length=0;for(let m=2;m<r[1].dims.length;++m)t.push(r[1].dims[m])}let n=e.format==="NHWC";t.splice(0,0,r[1].dims[0]),t.splice(n?3:1,0,r[1].dims[1]);let o=e.pads.slice(),i=e.outputShape.slice(),u=e.outputPadding.slice(),d=r[0].dims,h=e.dilations.slice();if(h.reduce((m,g)=>m+g,0)===0){let m=r[0].dims.length-2;h=new Array(m).fill(1)}let f=e.strides.slice();if(f.reduce((m,g)=>m+g,0)===0){let m=r[0].dims.length-2;f=new Array(m).fill(1)}cy(d,t,h,e.autoPad,e.group,o,f,n,u,i);let y=Object.assign({},e);return Object.assign(y,{kernelShape:t,pads:o,outputPadding:u,outputShape:i,dilations:h,strides:f}),y},Tb=e=>{let r=u_(e),t=e.format,n=["NOTSET","VALID","SAME_UPPER","SAME_LOWER"][typeof e.autoPad>"u"?0:e.autoPad],o=e.dilations,i=e.group,u=e.kernelShape,d=e.pads,h=e.strides,f=e.wIsConst(),y=e.outputPadding,m=e.outputShape;return{autoPad:n,format:t,dilations:o,group:i,kernelShape:u,outputPadding:y,outputShape:m,pads:d,strides:h,wIsConst:f,...r,cacheKey:`${e.format};${r.activation};`}},uy=(e,r)=>{if(!e||e.length!==2&&e.length!==3)throw new Error("Conv requires 2 or 3 inputs");if(e[0].dims.length!==4&&e[0].dims.length!==3)throw new Error("currently only support 2-dimensional conv");if(e[0].dims.length!==e[1].dims.length)throw new Error("filter does not have same dimension as input");let t=e[0].dims[r.format==="NHWC"?e[0].dims.length-1:1],n=e[1].dims[0];if(t!==n)throw new Error("FILTER_IN_CHANNEL should be equal to DATA_CHANNEL");let o=e[1].dims[1]*r.group;if(e.length===3&&(e[2].dims.length!==1||e[2].dims[0]!==o))throw new Error("invalid bias");let i=e[0].dims.length-2;if(r.dilations.reduce((u,d)=>u+d,0)>0&&r.dilations.length!==i)throw new Error(`dilations should be ${i}D`);if(r.strides.reduce((u,d)=>u+d,0)>0&&r.strides.length!==i)throw new Error(`strides should be ${i}D`);if(r.pads.reduce((u,d)=>u+d,0)>0&&r.pads.length!==i*2)throw new Error(`pads should be ${i*2}D`);if(r.outputPadding.length!==i&&r.outputPadding.length!==0)throw new Error(`output_padding should be ${i}D`);if(r.kernelShape.reduce((u,d)=>u+d,0)>0&&r.kernelShape.length!==0&&r.kernelShape.length!==e[1].dims.length-2)throw new Error("invalid kernel shape");if(r.outputShape.length!==0&&r.outputShape.length!==e[0].dims.length-2)throw new Error("invalid output shape")},sm=(e,r,t,n)=>{let o=e.kernelCustomData.wT??e.compute(xi(r[1],[2,3,0,1]),{inputs:[1],outputs:[t.wIsConst?-2:-1]})[0];t.wIsConst&&!e.kernelCustomData.wT&&(e.kernelCustomData.wT=o);let i=[r[0],o];r.length===3&&i.push(r[2]),e.compute(xb(i,t,n),{inputs:i})},dy=(e,r)=>{let t=r.format==="NHWC",n=[e.inputs[0].reshape(t?[e.inputs[0].dims[0],1,e.inputs[0].dims[1],e.inputs[0].dims[2]]:[e.inputs[0].dims[0],e.inputs[0].dims[1],1,e.inputs[0].dims[2]]),e.inputs[1].reshape([e.inputs[1].dims[0],e.inputs[1].dims[1],1,e.inputs[1].dims[2]])];e.inputs.length===3&&n.push(e.inputs[2]);let o=r.kernelShape;(o.length===0||o[0]===0)&&(o=[e.inputs[1].dims[2]]);let i=r.dilations;(i.length===0||i[0]===0)&&(i=[1]);let u=r.strides;(u.length===0||u[0]===0)&&(u=[1]);let d=r.pads;d.length===0&&(d=[0,0]),d=[0,d[0],0,d[1]],u=[1].concat(u),i=[1].concat(i),o=[1].concat(o);let h=r.outputPadding;h=[0].concat(h);let f=im({...r,pads:d,strides:u,dilations:i,kernelShape:o,outputPadding:h},n);sm(e,n,f,y=>t?[y[0],y[2],y[3]]:[y[0],y[1],y[3]])},Eb=(e,r)=>{if(uy(e.inputs,r),e.inputs[0].dims.length===3)dy(e,r);else{let t=im(r,e.inputs);sm(e,e.inputs,t)}}}),py,Pb,Cb,bT=Je(()=>{St(),Ft(),vr(),Lt(),py=(e,r,t,n)=>{let o=Fe.size(r),i=r.length,u=Re("input",e,i),d=mt("output",e,i),h=t.dataType===6?t.getInt32Array()[0]:Number(t.getBigInt64Array()[0]),f=Fe.normalizeAxis(h,i),y=m=>{let g=` i32(${u.indicesGet("inputIndices","uniforms.axis")}) `,v=wt("uniforms.input_shape","uniforms.axis",i),b=n.reverse?g+(n.exclusive?" + 1":""):"0",k=n.reverse?v:g+(n.exclusive?"":" + 1");return`
                ${m.registerUniform("outputSize","u32").registerUniform("axis","u32").declareVariables(u,d)}
                ${m.mainStart()}
                  ${m.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.outputSize")}
                  var inputIndices = ${d.offsetToIndices("global_idx")};
                  var sum = ${d.type.value}(0);
                  let first : i32 = ${b};
                  let last : i32 = ${k};
                  for (var i : i32 = first; i < last; i++) {
                    ${u.indicesSet("inputIndices","uniforms.axis","u32(i)")};
                    sum = sum + ${u.getByIndices("inputIndices")};
                  }
                  ${d.setByOffset("global_idx","sum")};
                }`};return{name:"CumSum",shaderCache:{hint:n.cacheKey,inputDependencies:["rank"]},getRunData:()=>({outputs:[{dims:r,dataType:e}],dispatchGroup:{x:Math.ceil(o/64)},programUniforms:[{type:12,data:o},{type:12,data:f},...bt(r,r)]}),getShaderSource:y}},Pb=(e,r)=>{let t=e.inputs[0].dims,n=e.inputs[0].dataType,o=e.inputs[1];e.compute(py(n,t,o,r),{inputs:[0]})},Cb=e=>{let r=e.exclusive===1,t=e.reverse===1;return Zt({exclusive:r,reverse:t})}}),hy,fy,my,Sb,$b,vT=Je(()=>{St(),Ft(),vr(),Lt(),hy=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.")},fy=(e,r,t,n)=>{let o=[];o.push(`fn perm(i: ${n.type.indices}) -> ${t.type.indices} {
    var a: ${t.type.indices};`);for(let i=0;i<r;++i)o.push(t.indicesSet("a",e[i],`i[${i}]`));return o.push("return a;}"),o.join(`
`)},my=(e,r)=>{let t,n,o,i,u,d,h=r.format==="NHWC",f=r.blocksize,y=r.mode==="DCR";h?([t,n,o,i]=e.dims,u=y?[t,n,o,f,f,i/f**2]:[t,n,o,i/f**2,f,f],d=y?[0,1,3,2,4,5]:[0,1,4,2,5,3]):([t,n,o,i]=[e.dims[0],e.dims[2],e.dims[3],e.dims[1]],u=y?[t,f,f,i/f**2,n,o]:[t,i/f**2,f,f,n,o],d=y?[0,3,4,1,5,2]:[0,1,4,2,5,3]);let m=e.reshape(u),g=m.dims.length,v=e.dataType,b=Re("a",v,g),k=mt("output",v,g),z=E=>`
  ${E.registerUniform("output_size","u32").declareVariables(b,k)}

  ${fy(d,g,b,k)}

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

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

    ${k.setByOffset("global_idx",b.getByIndices("aIndices"))}
  }`;return{name:"DepthToSpace",shaderCache:{hint:`${e.dims};${r.blocksize};${r.mode}`,inputDependencies:["rank"]},getRunData:E=>{let T=h?[t,n*f,o*f,i/f**2]:[t,i/f**2,n*f,o*f],D=Fe.size(T),I=m.dims,$=Fe.sortBasedOnPerm(I,d);return{outputs:[{dims:T,dataType:E[0].dataType}],dispatchGroup:{x:Math.ceil(D/64)},programUniforms:[{type:12,data:D},...bt(I,$)]}},getShaderSource:z}},Sb=(e,r)=>{hy(e.inputs),e.compute(my(e.inputs[0],r))},$b=e=>Zt({blocksize:e.blocksize,mode:e.mode,format:e.format})}),Gh,ad,nm,_y,gy,yy,wy,am,My,kb,Ib,xT=Je(()=>{St(),Ft(),vr(),Lt(),Gh="[a-zA-Z]|\\.\\.\\.",ad="("+Gh+")+",nm="^"+ad+"$",_y="("+ad+",)*"+ad,gy="^"+_y+"$",yy=class{constructor(e=-1){this.symbolToIndices=new Map,this.inputIndex=e}addSymbol(e,r){let t=this.symbolToIndices.get(e);t===void 0?t=[r]:t.push(r),this.symbolToIndices.set(e,t)}},wy=class{constructor(e,r){var o;this.equation=r,this.hasEllipsis=!1,this.symbolToInfo=new Map,this.lhs=new Array,this.outputDims=[];let[t,n]=r.includes("->")?r.split("->",2):[r,""];if(!t.match(RegExp(gy)))throw new Error("Invalid LHS term");if(t.split(",").forEach((i,u)=>{let d=e[u].dims.slice();if(!i.match(RegExp(nm)))throw new Error("Invalid LHS term");let h=this.processTerm(i,!0,d,u);this.lhs.push(h)}),n==="")n+=[...this.symbolToInfo.entries()].filter(([i,u])=>u.count===1||i==="...").map(([i])=>i).join("");else if(!n.match(RegExp(ad)))throw new Error("Invalid RHS");(o=n.match(RegExp(Gh,"g")))==null||o.forEach(i=>{if(i==="...")this.outputDims=this.outputDims.concat(this.ellipsisDims);else{let u=this.symbolToInfo.get(i);if(u===void 0)throw new Error("Invalid RHS symbol");this.outputDims.push(u.dimValue)}}),this.rhs=this.processTerm(n,!1,this.outputDims)}addSymbol(e,r,t){let n=this.symbolToInfo.get(e);if(n!==void 0){if(n.dimValue!==r&&n.count!==1)throw new Error("Dimension mismatch");n.count++,n.inputIndices.push(t)}else n={count:1,dimValue:r,inputIndices:[t]};this.symbolToInfo.set(e,n)}processTerm(e,r,t,n=-1){let o=t.length,i=!1,u=[],d=0;if(!e.match(RegExp(nm))&&!r&&e!=="")throw new Error("Invalid LHS term");let h=e.match(RegExp(Gh,"g")),f=new yy(n);return h==null||h.forEach((y,m)=>{if(y==="..."){if(i)throw new Error("Only one ellipsis is allowed per input term");i=!0;let g=o-h.length+1;if(g<0)throw new Error("Ellipsis out of bounds");if(u=t.slice(d,d+g),this.hasEllipsis){if(this.ellipsisDims.length!==u.length||this.ellipsisDims.toString()!==u.toString())throw new Error("Ellipsis dimensions mismatch")}else if(r)this.hasEllipsis=!0,this.ellipsisDims=u;else throw new Error("Ellipsis must be specified in the LHS");for(let v=0;v<u.length;v++){let b=String.fromCharCode(48+v);f.addSymbol(b,m+v),this.addSymbol(b,t[d++],n)}}else f.addSymbol(y,m+(this.hasEllipsis?this.ellipsisDims.length-1:0)),this.addSymbol(y,t[d++],n)}),f}},am=e=>e+"_max",My=(e,r,t,n)=>{let o=e.map(f=>f.length).map((f,y)=>Re(`input${y}`,r,f)),i=Fe.size(n),u=mt("output",r,n.length),d=[...t.symbolToInfo.keys()].filter(f=>!t.rhs.symbolToIndices.has(f)),h=f=>{let y=[],m="var prod = 1.0;",g="var sum = 0.0;",v="sum += prod;",b=[],k=[],z=[],E=[],T=t.symbolToInfo.size===t.rhs.symbolToIndices.size;t.symbolToInfo.forEach((I,$)=>{var P;if(t.rhs.symbolToIndices.has($)){let S=(P=t.rhs.symbolToIndices.get($))==null?void 0:P[0];S!==void 0&&t.lhs.forEach((O,R)=>{if(I.inputIndices.includes(R)){let G=O.symbolToIndices.get($);if(G===void 0)throw new Error("Invalid symbol error");G.forEach(ee=>{y.push(`${o[R].indicesSet(`input${R}Indices`,ee,u.indicesGet("outputIndices",S))}`)})}})}else t.lhs.forEach((S,O)=>{if(I.inputIndices.includes(O)){let R=S.symbolToIndices.get($);if(R===void 0)throw new Error("Invalid symbol error");R.forEach(G=>{b.push(`${o[O].indicesSet(`input${O}Indices`,G,`${$}`)}`)}),E.push(`prod *= ${o[O].getByIndices(`input${O}Indices`)};`)}}),k.push(`for(var ${$}: u32 = 0; ${$} < uniforms.${am($)}; ${$}++) {`),z.push("}")});let D=T?[...y,`let sum = ${o.map((I,$)=>I.getByIndices(`input${$}Indices`)).join(" * ")};`]:[...y,g,...k,...b,m,...E,v,...z];return`
            ${f.registerUniforms(d.map(I=>({name:`${am(I)}`,type:"u32"}))).registerUniform("outputSize","u32").declareVariables(...o,u)}

            ${f.mainStart()}
            ${f.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.outputSize")}
            var outputIndices = ${u.offsetToIndices("global_idx")};
            ${o.map((I,$)=>`var input${$}Indices: ${o[$].type.indices};`).join(`
`)}
            ${D.join(`
`)};
            ${u.setByOffset("global_idx","sum")};
          }`};return{name:"Einsum",shaderCache:{hint:t.equation,inputDependencies:e.map(()=>"rank")},getRunData:()=>{let f=d.filter(m=>t.symbolToInfo.has(m)).map(m=>{var g;return{type:12,data:((g=t.symbolToInfo.get(m))==null?void 0:g.dimValue)||0}});f.push({type:12,data:i});let y=e.map((m,g)=>[...bt(m)]).reduce((m,g)=>m.concat(g),f);return y.push(...bt(n)),{outputs:[{dims:n,dataType:r}],dispatchGroup:{x:Math.ceil(i/64)},programUniforms:y}},getShaderSource:h}},kb=(e,r)=>{let t=new wy(e.inputs,r.equation),n=t.outputDims,o=e.inputs.map((i,u)=>i.dims);e.compute(My(o,e.inputs[0].dataType,t,n))},Ib=e=>{let r=e.equation.replace(/\s+/g,"");return Zt({equation:r})}}),by,om,vy,xy,Ab,TT=Je(()=>{St(),Ft(),Lt(),by=e=>{if(!e||e.length!==2)throw new Error("Expand requires 2 input.");let r=e[0].dims,t=Array.from(e[1].getBigInt64Array(),Number),n=t.length<r.length?0:t.length-r.length,o=r.length<t.length?0:r.length-t.length;for(;n<t.length&&o<r.length;++n,++o)if(t[n]!==r[o]&&t[n]!==1&&r[o]!==1)throw new Error("Expand requires shape to be broadcastable to input")},om=(e,r)=>{let t=e.length-r.length,n=[];for(let o=0;o<t;++o)n.push(e[o]);for(let o=0;o<r.length;++o)n.push(r[o]===1?e[o+t]:r[o]);return n},vy=(e,r)=>e.length>r.length?om(e,r):om(r,e),xy=e=>{let r=e[0].dims,t=Array.from(e[1].getBigInt64Array(),Number),n=vy(r,t),o=e[0].dataType,i=o===9||Fe.size(r)===1,u=o===9||r.length>0&&r[r.length-1]%4===0?4:1,d=i||n.length>0&&n[n.length-1]%4===0?4:1,h=Math.ceil(Fe.size(n)/d),f=m=>{let g=Re("input",o,r.length,u),v=mt("output",o,n.length,d),b;if(o===9){let k=(z,E,T="")=>`
          let outputIndices${E} = ${v.offsetToIndices(`outputOffset + ${E}u`)};
          let offset${E} = ${g.broadcastedIndicesToOffset(`outputIndices${E}`,v)};
          let index${E} = offset${E} / 4u;
          let component${E} = offset${E} % 4u;
          ${z}[${E}] = ${T}(${g.getByOffset(`index${E}`)}[component${E}]);
        `;b=`
        let outputOffset = global_idx * ${d};
        var data = vec4<u32>(0);
        ${k("data",0,"u32")}
        ${k("data",1,"u32")}
        ${k("data",2,"u32")}
        ${k("data",3,"u32")}
        ${v.setByOffset("global_idx","data")}
      }`}else b=`
        let outputIndices = ${v.offsetToIndices(`global_idx * ${d}`)};
        let inputOffset = ${g.broadcastedIndicesToOffset("outputIndices",v)};
        let data = ${v.type.value}(${g.getByOffset(`inputOffset / ${u}`)});
        ${v.setByOffset("global_idx","data")}
      }`;return`
    ${m.registerUniform("vec_size","u32").declareVariables(g,v)}
    ${m.mainStart()}
    ${m.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.vec_size")}
    ${b}`},y=[{type:12,data:h},...bt(r,n)];return{name:"Expand",shaderCache:{hint:`${n.length};${u}${d}`,inputDependencies:["rank"]},getShaderSource:f,getRunData:()=>({outputs:[{dims:n,dataType:e[0].dataType}],dispatchGroup:{x:Math.ceil(h/64)},programUniforms:y})}},Ab=e=>{by(e.inputs),e.compute(xy(e.inputs),{inputs:[0]})}}),Ty,Ob,ET=Je(()=>{St(),Ft(),Lt(),c_(),Ty=e=>{let r=e[0].dataType,t=Fe.size(e[0].dims),n=Fe.size(e[1].dims),o=n%4===0,i=u=>{let d=Re("x",r,[1],4),h=Re("bias",r,[1],4),f=mt("y",r,[1],4),y=[{name:"output_vec_size",type:"u32"},{name:"bias_size",type:"u32"}],m=v=>`
      let bias${v}_offset: u32 = (global_idx * 4 + ${v}) % uniforms.bias_size;
      let bias${v} = ${h.getByOffset(`bias${v}_offset / 4`)}[bias${v}_offset % 4];`,g=o?`
      let bias = ${h.getByOffset("global_idx % (uniforms.bias_size / 4)")};`:`${m(0)}${m(1)}${m(2)}${m(3)}
      let bias = ${d.type.value}(bias0, bias1, bias2, bias3);`;return`${u.registerUniforms(y).declareVariables(d,h,f)}

    ${zm(ti(r))}

    ${u.mainStart(Wa)}
      ${u.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.output_vec_size")}

      let x = ${d.getByOffset("global_idx")};
      ${g}
      let x_in = x + bias;
      ${f.setByOffset("global_idx",Bm("x_in"))}
    }`};return{name:"FastGeluWithBias",shaderCache:{hint:`${o}`,inputDependencies:["type","type"]},getShaderSource:i,getRunData:u=>({outputs:[{dims:u[0].dims,dataType:u[0].dataType}],programUniforms:[{type:12,data:Math.ceil(t/4)},{type:12,data:n}],dispatchGroup:{x:Math.ceil(t/Wa/4)}})}},Ob=e=>{e.inputs.length<2||Fe.size(e.inputs[1].dims)===0?ZM(e):e.compute(Ty(e.inputs))}}),Ey,Py,Fb,Db,PT=Je(()=>{St(),Ft(),vr(),Lt(),Ey=e=>{if(!e||e.length!==2)throw new Error("Gather requires 2 inputs.")},Py=(e,r)=>{let t=e[0].dims,n=e[1].dims,o=t.length,i=Fe.normalizeAxis(r.axis,o),u=t.slice(0);u.splice(i,1,...n);let d=t[i],h=e[0].dataType===9?4:1,f=Math.ceil(Fe.size(u)/h),y=[{type:12,data:f},{type:6,data:d},{type:12,data:i},...bt(e[0].dims,e[1].dims,u)],m=g=>{let v=Re("data",e[0].dataType,e[0].dims.length,h),b=Re("inputIndices",e[1].dataType,e[1].dims.length),k=mt("output",e[0].dataType,u.length,h),z=T=>{let D=n.length,I=`var indicesIndices${T}  = ${b.type.indices}(0);`;for(let $=0;$<D;$++)I+=`${D>1?`indicesIndices${T}[${$}]`:`indicesIndices${T}`} = ${u.length>1?`outputIndices${T}[uniforms.axis + ${$}]`:`outputIndices${T}`};`;I+=`
          var idx${T} = ${b.getByIndices(`indicesIndices${T}`)};
          if (idx${T} < 0) {
            idx${T} = idx${T} + uniforms.axisDimLimit;
          }
          var dataIndices${T} : ${v.type.indices};
        `;for(let $=0,P=0;$<o;$++)$===i?(I+=`${o>1?`dataIndices${T}[${$}]`:`dataIndices${T}`} = u32(idx${T});`,P+=D):(I+=`${o>1?`dataIndices${T}[${$}]`:`dataIndices${T}`} = ${u.length>1?`outputIndices${T}[${P}]`:`outputIndices${T}`};`,P++);return I},E;if(e[0].dataType===9){let T=(D,I,$="")=>`
          let outputIndices${I} = ${k.offsetToIndices(`outputOffset + ${I}u`)};
          ${z(I)};
          let offset${I} = ${v.indicesToOffset(`dataIndices${I}`)};
          let index${I} = offset${I} / 4u;
          let component${I} = offset${I} % 4u;
          ${D}[${I}] = ${$}(${v.getByOffset(`index${I}`)}[component${I}]);
        `;E=`
        let outputOffset = global_idx * ${h};
        var value = vec4<u32>(0);
        ${T("value",0,"u32")}
        ${T("value",1,"u32")}
        ${T("value",2,"u32")}
        ${T("value",3,"u32")}
        ${k.setByOffset("global_idx","value")}
      `}else E=`
      let outputIndices = ${k.offsetToIndices("global_idx")};
      ${z("")};
      let value = ${v.getByIndices("dataIndices")};
      ${k.setByOffset("global_idx","value")};
      `;return`
      ${g.registerUniform("outputSize","u32").registerUniform("axisDimLimit","i32").registerUniform("axis","u32").declareVariables(v,b,k)}
      ${g.mainStart()}
        ${g.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.outputSize")}
        ${E}
      }`};return{name:"Gather",shaderCache:{hint:r.cacheKey,inputDependencies:["rank","rank"]},getRunData:()=>({outputs:[{dims:u,dataType:e[0].dataType}],dispatchGroup:{x:Math.ceil(f/64)},programUniforms:y}),getShaderSource:m}},Fb=e=>Zt({axis:e.axis}),Db=(e,r)=>{let t=e.inputs;Ey(t),e.compute(Py(e.inputs,r))}}),Cy,Lb,jb,CT=Je(()=>{St(),Ft(),Lt(),Cy=(e,r,t,n,o,i,u,d,h)=>{let f=[{type:12,data:i},{type:12,data:n},{type:12,data:o},{type:12,data:t},{type:12,data:u},{type:12,data:d},{type:12,data:h}],y=[i];f.push(...bt(r.dims,y));let m=g=>{let v=Re("indices_data",r.dataType,r.dims.length),b=mt("input_slice_offsets_data",12,1,1),k=[v,b],z=[{name:"output_size",type:"u32"},{name:"batch_dims",type:"u32"},{name:"input_dims",type:"u32",length:o.length},{name:"sizes_from_slice_dims_data",type:"u32",length:t.length},{name:"num_slices_per_batch",type:"u32"},{name:"input_batch_stride",type:"u32"},{name:"num_slice_dims",type:"u32"}];return`
  ${g.registerUniforms(z).declareVariables(...k)}
  ${g.mainStart()}
    ${g.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.output_size")}
    let batch_idx = global_idx / uniforms.num_slices_per_batch;
    let base_offset = batch_idx * uniforms.input_batch_stride;

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

    input_slice_offsets_data[global_idx] =  base_offset + u32(relative_slice_offset);
  }`};return e.compute({name:"computeSliceOffsets",shaderCache:{hint:`${o.length}_${t.length}`,inputDependencies:["rank"]},getRunData:()=>({outputs:[{dims:y,dataType:e.inputs[1].dataType}],dispatchGroup:{x:Math.ceil(i/64)},programUniforms:f}),getShaderSource:m},{inputs:[r],outputs:[-1]})[0]},Lb=(e,r)=>{let t=e.inputs,n=t[0].dims,o=t[0].dataType,i=t[1].dims,u=i[i.length-1],d=Fe.sizeToDimension(i,i.length-1),h=Fe.sizeFromDimension(n,r.batchDims+u),f=Fe.sizeToDimension(n,r.batchDims),y=Fe.sizeFromDimension(n,r.batchDims),m=d/f,g=new Array(u),v=h;for(let I=0;I<u;++I)g[u-1-I]=v,v*=n[r.batchDims+u-1-I];let b=Cy(e,t[1],g,r.batchDims,n,d,m,y,u),k=r.batchDims+u;if(k>n.length)throw new Error("last dimension of indices must not be larger than rank of input tensor");let z=i.slice(0,-1).concat(n.slice(k)),E=Fe.size(z),T=[{type:12,data:E},{type:12,data:h},...bt(t[0].dims,b.dims,z)],D=I=>{let $=Re("data",t[0].dataType,t[0].dims.length),P=Re("slice_offsets",12,b.dims.length),S=mt("output",t[0].dataType,z.length);return`
          ${I.registerUniform("output_size","u32").registerUniform("slice_size","u32").declareVariables($,P,S)}
            ${I.mainStart()}
            ${I.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.output_size")}
          let slice_offset = slice_offsets[global_idx / uniforms.slice_size];
          output[global_idx] = data[u32(slice_offset) + global_idx % uniforms.slice_size];
        }`};e.compute({name:"GatherND",shaderCache:{hint:r.cacheKey,inputDependencies:["rank","rank"]},getRunData:()=>({outputs:[{dims:z,dataType:o}],dispatchGroup:{x:Math.ceil(E/64)},programUniforms:T}),getShaderSource:D},{inputs:[t[0],b]})},jb=e=>({batchDims:e.batch_dims,cacheKey:""})}),Sy,$y,zb,Bb,ST=Je(()=>{St(),Ft(),vr(),Lt(),Sy=(e,r)=>{if(e.length<3||e.length>4)throw new Error("GatherBlockQuantized requires 3 or 4 inputs.");let t=Fe.normalizeAxis(r.quantizeAxis,e[0].dims.length),n=r.blockSize,o=e[0],i=e[2],u=e.length===4?e[3]:void 0;if(i.dims.length!==o.dims.length||!o.dims.map((d,h)=>h===t?Math.ceil(d/n)===i.dims[h]:d===i.dims[h]).reduce((d,h)=>d&&h,!0))throw new Error("Scales must have the same rank as the input tensor and the dims should match except on gatherAxis.");if(u){if(u.dataType!==o.dataType)throw new Error("Zero point must have the same data type as the input tensor.");if(u.dims.length!==i.dims.length||!u.dims.map((d,h)=>d===i.dims[h]).reduce((d,h)=>d&&h,!0))throw new Error("Zero point must have the same rank as the input tensor and the dims should match except on quantizeAxis.")}},$y=(e,r)=>{let t=e[0].dims,n=e[1].dims,o=t.length,i=Fe.normalizeAxis(r.gatherAxis,o),u=Fe.normalizeAxis(r.quantizeAxis,o),d=t.slice(0);d.splice(i,1,...n);let h=Fe.size(d),f=e[2].dataType,y=e[0].dataType===22,m=[{type:12,data:h},{type:12,data:u},{type:12,data:i},{type:12,data:r.blockSize},...bt(...e.map((v,b)=>v.dims),d)],g=v=>{let b=Re("data",e[0].dataType,e[0].dims.length),k=Re("inputIndices",e[1].dataType,e[1].dims.length),z=Re("scales",e[2].dataType,e[2].dims.length),E=e.length>3?Re("zeroPoint",e[3].dataType,e[3].dims.length):void 0,T=mt("output",f,d.length),D=[b,k,z];E&&D.push(E);let I=[{name:"output_size",type:"u32"},{name:"quantize_axis",type:"u32"},{name:"gather_axis",type:"u32"},{name:"block_size",type:"u32"}];return`
        ${v.registerUniforms(I).declareVariables(...D,T)}
        ${v.mainStart()}
        let output_indices = ${T.offsetToIndices("global_idx")};
        var indices_indices = ${k.type.indices}(0);
        ${n.length>1?`
          for (var i: u32 = 0; i < ${n.length}; i++) {
            let index = ${T.indicesGet("output_indices","uniforms.gather_axis + i")};
            ${k.indicesSet("indices_indices","i","index")};
          }`:`indices_indices = ${T.indicesGet("output_indices","uniforms.gather_axis")};`};
        var data_indices = ${b.type.indices}(0);
        for (var i: u32 = 0; i < uniforms.gather_axis; i++) {
          let index = ${T.indicesGet("output_indices","i")};
          ${b.indicesSet("data_indices","i","index")};
        }
        var index_from_indices = ${k.getByIndices("indices_indices")};
        if (index_from_indices < 0) {
          index_from_indices += ${t[i]};
        }
        ${b.indicesSet("data_indices","uniforms.gather_axis","u32(index_from_indices)")};
        for (var i = uniforms.gather_axis + 1; i < ${d.length}; i++) {
          let index = ${T.indicesGet("output_indices",`i + ${n.length} - 1`)};
          ${b.indicesSet("data_indices","i","index")};
        }
        let data_offset = ${b.indicesToOffset("data_indices")};
        let data_index = data_offset % 8;
        // Convert 4-bit packed data to 8-bit packed data.
        let packed_4bit_quantized_data = ${b.getByOffset("data_offset / 8")};
        let packed_8bit_quantized_data = (packed_4bit_quantized_data >> (4 * (data_index % 2))) & 0x0f0f0f0f;
        let quantized_data_vec = ${y?"unpack4xI8":"unpack4xU8"}(u32(packed_8bit_quantized_data));
        let quantized_data = quantized_data_vec[data_index / 2];
        var scale_indices = data_indices;
        let quantize_axis_index = ${z.indicesGet("data_indices","uniforms.quantize_axis")} / uniforms.block_size;
        ${z.indicesSet("scale_indices","uniforms.quantize_axis","quantize_axis_index")};
        var scale = ${z.getByIndices("scale_indices")};
        ${E?`
              let zero_point_indices = scale_indices;
              let zero_point_offset = ${E.indicesToOffset("zero_point_indices")};
              let zero_point_index = zero_point_offset % 8;
              let packed_4bit_zero_points = ${E.getByOffset("zero_point_offset / 8")};
              let packed_8bit_zero_points = (packed_4bit_zero_points >> (4 * (zero_point_index % 2))) & 0x0f0f0f0f;
              let zero_point_vec = ${y?"unpack4xI8":"unpack4xU8"}(u32(packed_8bit_zero_points));
              let zero_point = zero_point_vec[zero_point_index / 2];`:"var zero_point = 0"};
        let dequantized_data = ${ti(f)}(quantized_data - zero_point) * scale;
        ${T.setByOffset("global_idx","dequantized_data")};
    }`};return{name:"GatherBlockQuantized",shaderCache:{hint:`${r.cacheKey};${e.filter((v,b)=>b!==1).map(v=>v.dims.join("_")).join(";")}`,inputDependencies:Array.from({length:e.length},(v,b)=>"rank")},getRunData:()=>({outputs:[{dims:d,dataType:f}],dispatchGroup:{x:Math.ceil(h/64)},programUniforms:m}),getShaderSource:g}},zb=(e,r)=>{let t=e.inputs;Sy(t,r),e.compute($y(e.inputs,r))},Bb=e=>Zt({blockSize:e.blockSize,gatherAxis:e.gatherAxis,quantizeAxis:e.quantizeAxis})}),ky,Iy,Rb,Nb,$T=Je(()=>{St(),Ft(),vr(),Lt(),ky=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.`)},Iy=(e,r)=>{let t=e[0].dims,n=e[0].dataType,o=t.length,i=e[1].dims,u=e[1].dataType,d=Fe.normalizeAxis(r.axis,o),h=t[d],f=i.slice(0),y=Fe.size(f),m=Re("input",n,o),g=Re("indicesInput",u,i.length),v=mt("output",n,f.length),b=[{type:12,data:y},{type:6,data:h},{type:12,data:d}];return b.push(...bt(t,i,f)),{name:"GatherElements",shaderCache:{inputDependencies:["rank","rank"]},getRunData:()=>({outputs:[{dims:f,dataType:e[0].dataType}],dispatchGroup:{x:Math.ceil(y/64)},programUniforms:b}),getShaderSource:k=>`
      ${k.registerUniform("outputSize","u32").registerUniform("axisDimLimit","i32").registerUniform("axis","u32").declareVariables(m,g,v)}
      ${k.mainStart()}
      ${k.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.outputSize")}

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

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

      ${v.setByOffset("global_idx","value")};
  }`}},Rb=e=>Zt({axis:e.axis}),Nb=(e,r)=>{let t=e.inputs;ky(t),e.compute(Iy(e.inputs,r))}}),Ay,Oy,Vb,Ub,kT=Je(()=>{St(),Ft(),Lt(),Ay=e=>{if(!e)throw new Error("Input is missing");if(e.length<2||e.length>3)throw new Error("Invaid input number.");if(e.length===3&&e[2].dims.length>2)throw new Error("Invalid input shape of C");if(e[0].dataType!==e[1].dataType||e.length===3&&e[0].dataType!==e[2].dataType)throw new Error("Input types are mismatched")},Oy=(e,r)=>{let t=e[0].dims.slice(),n=e[1].dims.slice(),[o,i,u]=R0.getShapeOfGemmResult(t,r.transA,n,r.transB,e.length===3?e[2].dims:void 0),d=[o,i];if(!d)throw new Error("Can't use gemm on the given tensors");let h=16,f=Math.ceil(i/h),y=Math.ceil(o/h),m=!0,g=Fe.size(d),v=[{type:12,data:m?f:g},{type:12,data:o},{type:12,data:i},{type:12,data:u},{type:1,data:r.alpha},{type:1,data:r.beta}],b=["type","type"];e.length===3&&(v.push(...bt(e[2].dims)),b.push("rank")),v.push(...bt(d));let k=E=>{let T="";r.transA&&r.transB?T="value += a[k * uniforms.M + m] * b[n * uniforms.K + k];":r.transA&&!r.transB?T="value += a[k * uniforms.M + m] * b[k * uniforms.N + n];":!r.transA&&r.transB?T="value += a[m * uniforms.K + k] * b[n * uniforms.K + k];":!r.transA&&!r.transB&&(T="value += a[m * uniforms.K + k] * b[k * uniforms.N + n];");let D=r.alpha===1?"":"value *= uniforms.alpha;",I=Re("a",e[0].dataType,e[0].dims),$=Re("b",e[1].dataType,e[1].dims),P=I.type.value,S=null,O=[I,$];e.length===3&&(S=Re("c",e[2].dataType,e[2].dims.length),O.push(S));let R=mt("output",e[0].dataType,d.length);O.push(R);let G=[{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`
  ${E.registerUniforms(G).declareVariables(...O)}

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

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

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

    ${D}
    ${S!=null?`let cOffset = ${S.broadcastedIndicesToOffset("vec2(m, n)",R)}; value += ${P}(uniforms.beta) * ${S.getByOffset("cOffset")};`:""}
    output[global_idx] = value;
  }`},z=E=>{let T=Re("a",e[0].dataType,e[0].dims),D=Re("b",e[1].dataType,e[1].dims),I=null,$=[T,D];e.length===3&&(I=Re("c",e[2].dataType,e[2].dims.length),$.push(I));let P=mt("output",e[0].dataType,d.length);$.push(P);let S=[{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"}],O="",R="";r.transA&&r.transB?(R=`
      var col = tile_row_start + local_id.x;
      var row = k_start + local_id.y;
      if (col < uniforms.M && row < uniforms.K) {
        tile_a[local_id.y][local_id.x] = a[row * uniforms.M + col];
      } else {
        tile_a[local_id.y][local_id.x] = ${T.type.value}(0);
      }

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

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

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

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

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

    ${G}
    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)",P)}; value += ${P.type.value}(uniforms.beta) * ${I.getByOffset("cOffset")};`:""}
    if (m < uniforms.M && n < uniforms.N) {
      output[m * uniforms.N + n] = value;
    }
  }`};return m?{name:"GemmShared",shaderCache:{hint:`${r.cacheKey}`,inputDependencies:b},getRunData:()=>({outputs:[{dims:d,dataType:e[0].dataType}],dispatchGroup:{x:f*y},programUniforms:v}),getShaderSource:z}:{name:"Gemm",shaderCache:{hint:`${r.cacheKey}`,inputDependencies:b},getRunData:()=>({outputs:[{dims:d,dataType:e[0].dataType}],dispatchGroup:{x:Math.ceil(g/64)},programUniforms:v}),getShaderSource:k}},Vb=e=>{let r=e.transA,t=e.transB,n=e.alpha,o=e.beta;return{transA:r,transB:t,alpha:n,beta:o,cacheKey:`${e.transA};${e.transB};${e.alpha===1}`}},Ub=(e,r)=>{Ay(e.inputs),e.compute(Oy(e.inputs,r))}}),ss,ws,fn,mn,Fy,Dy,Ly,jy,zy,By,Ry,Ny,Wb,Gb,IT=Je(()=>{St(),Ft(),vr(),Lt(),[ss,ws,fn,mn]=[0,1,2,3],Fy=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 ${e[0].dims.length-2}`);if(e[0].dims[0]!==e[1].dims[0])throw new Error("grid batch size must match input batch size")},Dy=`
  fn gs_get_cubic_coeffs(x: f32) -> vec4<f32> {
    let cubic_alpha = -0.75f;
    let x_abs = abs(x);
    var coeffs: vec4<f32>;
    coeffs[0] = (((cubic_alpha * (x_abs + 1) - 5 * cubic_alpha) * (x_abs + 1) + 8 * cubic_alpha) * (x_abs + 1) - 4 * cubic_alpha);
    coeffs[1] = (((cubic_alpha + 2) * x_abs - (cubic_alpha + 3)) * x_abs * x_abs + 1);
    coeffs[2] = (((cubic_alpha + 2) * (1 - x_abs) - (cubic_alpha + 3)) * (1 - x_abs) * (1 - x_abs) + 1);
    coeffs[3] = (((cubic_alpha * (2 - x_abs) - 5 * cubic_alpha) * (2 - x_abs) + 8 * cubic_alpha) * (2 - x_abs) - 4 * cubic_alpha);
    return coeffs;
  }
`,Ly=e=>`
  fn gs_bicubic_interpolate(p: mat4x4<${e}>, x: f32, y: f32) -> ${e} {
    var v: vec4<f32>;
    var coeffs = gs_get_cubic_coeffs(x);
    for (var i = 0; i < 4; i++) {
      v[i] = coeffs[0] * p[i][0] + coeffs[1] * p[i][1] + coeffs[2] * p[i][2] + coeffs[3] * p[i][3];
    }
    coeffs = gs_get_cubic_coeffs(y);
    let pixel = ${e}(coeffs[0] * v[0] + coeffs[1] * v[1] + coeffs[2] * v[2] + coeffs[3] * v[3]);
    return pixel;
  }
`,jy=e=>`
  fn gs_denormalize(n: f32, length: i32) -> f32 {
    ${e.alignCorners===0?`
    // alignCorners: false => [-1, 1] to [-0.5, length - 0.5]
    return ((n + 1.0) * f32(length) - 1.0) / 2.0;
    `:`
    // alignCorners: true => [-1, 1] to [0, length - 1]
    return (n + 1.0) / 2.0 * (f32(length - 1));
    `}
  }
`,zy=e=>`
  ${e.paddingMode==="reflection"?`
      fn gs_reflect(x: i32, x_min: f32, x_max: f32) -> u32 {
        var dx = 0.0;
        var fx = f32(x);
        let range = x_max - x_min;
        if (fx < x_min) {
          dx = x_min - fx;
          let n = u32(dx / range);
          let r = dx - f32(n) * range;
          if (n % 2 == 0) {
            fx = x_min + r;
          } else {
            fx = x_max - r;
          }
        } else if (fx > x_max) {
          dx = fx - x_max;
          let n = u32(dx / range);
          let r = dx - f32(n) * range;
          if (n % 2 == 0) {
            fx = x_max - r;
          } else {
            fx = x_min + r;
          }
        }
        return u32(fx);
      }`:""}
`,By=(e,r,t)=>`
  fn pixel_at_grid(r: i32, c: i32, H: i32, W: i32, batch: u32, channel: u32, border: vec4<f32>) -> ${r} {
     var pixel = ${r}(0);
     var indices = vec4<u32>(0);
     indices[${ss}] = batch;
     indices[${ws}] = channel;`+(()=>{switch(t.paddingMode){case"zeros":return`
          if (r >= 0 && r < H && c >=0 && c < W) {
            indices[${fn}] = u32(r);
            indices[${mn}] = u32(c);
          } else {
            return ${r}(0);
          }
        `;case"border":return`
          indices[${fn}] = u32(clamp(r, 0, H - 1));
          indices[${mn}] = u32(clamp(c, 0, W - 1));
        `;case"reflection":return`
          indices[${fn}] = gs_reflect(r, border[1], border[3]);
          indices[${mn}] = gs_reflect(c, border[0], border[2]);
        `;default:throw new Error(`padding mode ${t.paddingMode} is not supported`)}})()+`
    return ${e.getByIndices("indices")};
  }
`,Ry=(e,r,t)=>(()=>{switch(t.mode){case"nearest":return`
          let result = pixel_at_grid(i32(round(y)), i32(round(x)), H_in, W_in, indices[${ss}], indices[${ws}], border);
        `;case"bilinear":return`
          let x1 = i32(floor(x));
          let y1 = i32(floor(y));
          let x2 = x1 + 1;
          let y2 = y1 + 1;

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

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

          let dx = x - f32(x0 + 1);
          let dy = y - f32(y0 + 1);
          let result = gs_bicubic_interpolate(p, dx, dy);
        `;default:throw new Error(`mode ${t.mode} is not supported`)}})()+`${e.setByOffset("global_idx","result")}`,Ny=(e,r)=>{let t=Re("x",e[0].dataType,e[0].dims.length),n=[e[1].dims[0],e[1].dims[1],e[1].dims[2]],o=Re("grid",e[1].dataType,n.length,2),i=[e[0].dims[0],e[0].dims[1],e[1].dims[1],e[1].dims[2]];r.format==="NHWC"&&(i=[e[0].dims[0],e[1].dims[1],e[1].dims[2],e[0].dims[3]],[ss,ws,fn,mn]=[0,3,1,2]);let u=mt("output",e[0].dataType,i.length),d=t.type.value,h=Fe.size(i),f=[{type:12,data:h},...bt(e[0].dims,n,i)],y=m=>`
  ${m.registerUniform("output_size","u32").declareVariables(t,o,u)}
  ${Dy}
  ${Ly(d)}
  ${jy(r)}
  ${zy(r)}
  ${By(t,d,r)}

  ${m.mainStart()}
    ${m.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.output_size")}
      let H_in = i32(uniforms.x_shape[${fn}]);
      let W_in = i32(uniforms.x_shape[${mn}]);

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

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

      ${Ry(u,d,r)}
  }`;return{name:"GridSample",shaderCache:{hint:`${r.cacheKey}`,inputDependencies:["type","type"]},getRunData:m=>{let g=Fe.size(i);return{outputs:[{dims:i,dataType:m[0].dataType}],dispatchGroup:{x:Math.ceil(g/64)},programUniforms:f}},getShaderSource:y}},Wb=(e,r)=>{Fy(e.inputs),e.compute(Ny(e.inputs,r))},Gb=e=>Zt({alignCorners:e.align_corners,mode:e.mode,paddingMode:e.padding_mode,format:e.format})}),di,Vy,Kb,lm,Uy,fd,Hb,qb=Je(()=>{St(),Ft(),vr(),n_(),l_(),Lt(),Ns(),di=(e,r)=>e.length>r&&e[r].dims.length>0?e[r]:void 0,Vy=(e,r)=>{let t=e[0],n=di(e,1),o=di(e,2),i=di(e,3),u=di(e,4),d=di(e,5),h=di(e,6),f=di(e,7);if(t.dims.length!==3&&t.dims.length!==5)throw new Error("Input query is expected to have 3 or 5 dimensions");let y=t.dims[0],m=t.dims[1],g=t.dims.length===3?t.dims[2]:r.numHeads*t.dims[4],v=m,b=0,k=0,z=Math.floor(g/r.numHeads);if(h&&f&&Fe.size(h.dims)&&Fe.size(f.dims)){if(h.dims.length!==4)throw new Error('Input "past_key" is expected to have 4 dimensions');if(h.dims[0]!==y||h.dims[1]!==r.numHeads||h.dims[3]!==z)throw new Error('Input "past_key" shape (batch_size, num_heads, past_sequence_length, head_size)');if(f.dims[0]!==y||f.dims[1]!==r.numHeads||f.dims[3]!==z)throw new Error('Input "past_value" shape (batch_size, num_heads, past_sequence_length, head_size)');if(h.dims[2]!==f.dims[2])throw new Error('Input "past_key" and "past_value" shall have same dim 2 (past_sequence_length)');if(f.dims.length!==4)throw new Error('Input "past_value" is expected to have 4 dimensions');b=h.dims[2],k=h.dims[2]}else if(h&&Fe.size(h.dims)||f&&Fe.size(f.dims))throw new Error('Input "past_key" and "past_value" shall be both present or both absent');let E;if(n&&Fe.size(n.dims)>0){if(t.dims.length!==3)throw new Error('Input "query" is expected to have 3 dimensions when key is given');if(n.dims.length<3||n.dims.length>5)throw new Error('Input "key" is expected to have 3, 4, or 5 dimensions');if(t.dims[0]!==n.dims[0])throw new Error('Input "query" and "key" shall have same dim 0 (batch size)');if(n.dims.length===3){if(n.dims[2]!==t.dims[2])throw new Error('Input "query" and "key" shall have same dim 2 (hidden_size)');E=2,v=n.dims[1]}else if(n.dims.length===5){if(n.dims[2]!==r.numHeads||n.dims[3]!==2||n.dims[4]!==z)throw new Error('Expect "key" shape (batch_size, kv_sequence_length, num_heads, 2, head_size) for packed kv');if(o)throw new Error('Expect "value" be none when "key" has packed kv format.');E=5,v=n.dims[1]}else{if(n.dims[1]!==r.numHeads||n.dims[3]!==z)throw new Error('Expect "key" shape (batch_size, num_heads, kv_sequence_length, head_size) for past_key');E=0,v=n.dims[2]}}else{if(t.dims.length!==5)throw new Error('Input "query" is expected to have 5 dimensions when key is empty');if(t.dims[2]!==r.numHeads||t.dims[3]!==3)throw new Error('Expect "query" shape (batch_size, kv_sequence_length, num_heads, 3, head_size) for packed kv');E=3}if(i&&Fe.size(i.dims)>0){if(i.dims.length!==1)throw new Error('Input "bias" is expected to have 1 dimension');if(n&&n.dims.length===5&&n.dims[3]===2)throw new Error("bias is not allowed for packed kv.")}let T=b+v,D=0;if(u&&Fe.size(u.dims)>0){D=8;let S=u.dims;throw S.length===1?S[0]===y?D=1:S[0]===3*y+2&&(D=3):S.length===2&&S[0]===y&&S[1]===T&&(D=5),D===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,$=g;if(o&&Fe.size(o.dims)>0){if(o.dims.length!==3&&o.dims.length!==4)throw new Error('Input "value" is expected to have 3 or 4 dimensions');if(t.dims[0]!==o.dims[0])throw new Error('Input "query" and "value" shall have same dim 0 (batch_size)');if(o.dims.length===3){if(v!==o.dims[1])throw new Error('Input "key" and "value" shall have the same dim 1 (kv_sequence_length)');$=o.dims[2]}else{if(v!==o.dims[2])throw new Error('Input "key" and "value" shall have the same dim 2 (kv_sequence_length)');$=o.dims[1]*o.dims[3],I=!0}}let P=!1;if(u&&Fe.size(u.dims)>0)throw new Error("Key padding mask is not supported");if(d&&Fe.size(d.dims)>0){if(d.dims.length!==4)throw new Error('Input "attention_bias" is expected to have 4 dimensions');if(d.dims[0]!==y||d.dims[1]!==r.numHeads||d.dims[2]!==m||d.dims[3]!==T)throw new Error('Expect "attention_bias" shape (batch_size, num_heads, sequence_length, total_sequence_length)')}return{batchSize:y,sequenceLength:m,pastSequenceLength:b,kvSequenceLength:v,totalSequenceLength:T,maxSequenceLength:k,inputHiddenSize:0,hiddenSize:g,vHiddenSize:$,headSize:z,vHeadSize:Math.floor($/r.numHeads),numHeads:r.numHeads,isUnidirectional:!1,pastPresentShareBuffer:!1,maskFilterValue:r.maskFilterValue,maskType:D,scale:r.scale,broadcastResPosBias:P,passPastInKv:I,qkvFormat:E}},Kb=e=>Zt({...e}),lm=Zt({perm:[0,2,1,3]}),Uy=(e,r,t,n,o,i,u)=>{let d=[n,o,i],h=Fe.size(d),f=[{type:12,data:h},{type:12,data:u},{type:12,data:i}],y=m=>{let g=mt("qkv_with_bias",r.dataType,d),v=Re("qkv",r.dataType,d),b=Re("bias",t.dataType,d),k=[{name:"output_size",type:"u32"},{name:"bias_offset",type:"u32"},{name:"hidden_size",type:"u32"}];return`
  ${m.registerUniforms(k).declareVariables(v,b,g)}
  ${m.mainStart()}
    ${m.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:d,dataType:r.dataType,gpuDataType:0}],dispatchGroup:{x:Math.ceil(h/64)},programUniforms:f}),getShaderSource:y},{inputs:[r,t],outputs:[-1]})[0]},fd=(e,r,t,n,o,i,u,d)=>{let h=i;if(u&&Fe.size(u.dims)>0){if(n===1)throw new Error("AddBiasReshape is not implemented. Please export your model with packed QKV or KV");return h=Uy(e,i,u,r,n,t*o,d),h=h.reshape([r,n,t,o]),t===1||n===1?h:e.compute(xi(h,lm.perm),{inputs:[h],outputs:[-1]})[0]}else return i.dims.length===3&&(h=i.reshape([r,n,t,o])),t===1||n===1?h:e.compute(xi(h,lm.perm),{inputs:[h],outputs:[-1]})[0]},Hb=(e,r)=>{let t=Vy(e.inputs,r),n=e.inputs[0],o=di(e.inputs,1),i=di(e.inputs,2),u=di(e.inputs,3),d=di(e.inputs,4),h=di(e.inputs,5),f=di(e.inputs,6),y=di(e.inputs,7);if(n.dims.length===5)throw new Error("Packed QKV is not implemented");if((o==null?void 0:o.dims.length)===5)throw new Error("Packed KV is not implemented");let m=o&&i&&o.dims.length===4&&i.dims.length===4,g=fd(e,t.batchSize,t.numHeads,t.sequenceLength,t.headSize,n,u,0);if(m)return wd(e,g,o,i,d,void 0,f,y,h,t);if(!o||!i)throw new Error("key and value must be provided");let v=fd(e,t.batchSize,t.numHeads,t.kvSequenceLength,t.headSize,o,u,t.hiddenSize),b=fd(e,t.batchSize,t.numHeads,t.kvSequenceLength,t.vHeadSize,i,u,2*t.hiddenSize);wd(e,g,v,b,d,void 0,f,y,h,t)}}),Wy,Gy,Ky,Hy,Wm,Qb,Xb,Jb=Je(()=>{St(),Ft(),vr(),Lt(),Wy=e=>{if(!e||e.length<1)throw new Error("too few inputs")},Gy=(e,r)=>{let t=[],n=r.numOutputs;return e[1].dims[0]>0&&(e[1].getBigInt64Array().forEach(o=>t.push(Number(o))),n=t.length),Zt({numOutputs:n,axis:r.axis,splitSizes:t})},Ky=e=>`
fn calculateOutputIndex(index: u32) -> u32 {
    for (var i: u32 = 0u; i < ${e}u; i += 1u ) {
    if (index < ${wt("uniforms.size_in_split_axis","i",e)}) {
        return i;
    }
    }
    return ${e}u;
}`,Hy=e=>{let r=e.length,t=[];for(let n=0;n<r;++n){let o=e[n].setByIndices("indices","input[global_idx]");r===1?t.push(o):n===0?t.push(`if (output_number == ${n}u) { ${o} }`):n===r-1?t.push(`else { ${o} }`):t.push(`else if (output_number == ${n}) { ${o} }`)}return`
      fn writeBufferData(output_number: u32, indices: ${e[0].type.indices}, global_idx: u32) {
        ${t.join(`
`)}
      }`},Wm=(e,r)=>{let t=e[0].dims,n=Fe.size(t),o=e[0].dataType,i=Fe.normalizeAxis(r.axis,t.length),u=new Array(r.numOutputs),d=Re("input",o,t.length),h=new Array(r.numOutputs),f=[],y=[],m=0,g=[{type:12,data:n}];for(let b=0;b<r.numOutputs;b++){m+=r.splitSizes[b],h[b]=m;let k=t.slice();k[i]=r.splitSizes[b],y.push(k),u[b]=mt(`output${b}`,o,k.length),f.push({dims:y[b],dataType:e[0].dataType})}g.push({type:12,data:h},...bt(t,...y));let v=b=>`
  ${b.registerUniform("input_size","u32").registerUniform("size_in_split_axis","u32",h.length).declareVariables(d,...u)}
  ${Ky(h.length)}
  ${Hy(u)}

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

    var indices = ${d.offsetToIndices("global_idx")};
    var index = ${d.indicesGet("indices",i)};
    let output_number = calculateOutputIndex(index);
    if (output_number != 0) {
      index -= ${wt("uniforms.size_in_split_axis","output_number - 1u",h.length)};
      ${d.indicesSet("indices",i,"index")};
    }
    writeBufferData(output_number, indices, global_idx);
  }`;return{name:"Split",shaderCache:{hint:r.cacheKey,inputDependencies:["rank"]},getShaderSource:v,getRunData:()=>({outputs:f,dispatchGroup:{x:Math.ceil(n/64)},programUniforms:g})}},Qb=(e,r)=>{Wy(e.inputs);let t=e.inputs.length===1?r:Gy(e.inputs,r);e.compute(Wm(e.inputs,t),{inputs:[0]})},Xb=e=>{let r=e.axis,t=e.splitSizes,n=e.numOutputs<0?t.length:e.numOutputs;if(n!==t.length)throw new Error("numOutputs and splitSizes length must be equal");return Zt({axis:r,numOutputs:n,splitSizes:t})}}),qy,af,Yb,Zb=Je(()=>{St(),Ft(),vr(),Lt(),qy=(e,r)=>{let[t,n,o,i]=e,{numHeads:u,rotaryEmbeddingDim:d}=r;if(t.dims.length!==3&&t.dims.length!==4)throw new Error(`Input 'x' is expected to have 3 or 4 dimensions, got ${t.dims.length}`);if(!Fe.areEqual(n.dims,[])&&!Fe.areEqual(n.dims,[1])&&n.dims.length!==2)throw new Error(`Input 'position_ids' is expected to have 0, 1, or 2 dimensions, got ${n.dims.length}`);if(o.dims.length!==2)throw new Error(`Input 'cos_cache' is expected to have 2 dimensions, got ${o.dims.length}`);if(i.dims.length!==2)throw new Error(`Input 'sin_cache' is expected to have 2 dimensions, got ${i.dims.length}`);if(!Fe.areEqual(o.dims,i.dims))throw new Error("Inputs 'cos_cache' and 'sin_cache' are expected to have the same shape");if(d>0&&u===0)throw new Error("num_heads must be provided if rotary_embedding_dim is specified");let h=t.dims[0],f=t.dims[t.dims.length-2],y=o.dims[0],m=Fe.sizeFromDimension(t.dims,1)/f,g=d===0?o.dims[1]*2:m/u;if(d>g)throw new Error("rotary_embedding_dim must be less than or equal to head_size");if(n.dims.length===2){if(h!==n.dims[0])throw new Error(`Input 'position_ids' dimension 0 should be of size batch_size, got ${n.dims[0]}`);if(f!==n.dims[1])throw new Error(`Input 'position_ids' dimension 1 should be of size sequence_length, got ${n.dims[1]}`)}if(g/2!==o.dims[1]&&d/2!==o.dims[1])throw new Error(`Input 'cos_cache' dimension 1 should be same as head_size / 2 or rotary_embedding_dim / 2, got ${o.dims[1]}`);if(f>y)throw new Error("Updating cos_cache and sin_cache in RotaryEmbedding is not currently supported")},af=(e,r)=>{let{interleaved:t,numHeads:n,rotaryEmbeddingDim:o,scale:i}=r,u=e[0].dims[0],d=Fe.sizeFromDimension(e[0].dims,1),h=e[0].dims[e[0].dims.length-2],f=d/h,y=e[2].dims[1],m=o===0?y*2:f/n,g=new Array(u,h,f/m,m-y),v=Fe.computeStrides(g),b=[{type:1,data:i},{type:12,data:g},{type:12,data:v},...e[0].dims.length===3?new Array({type:12,data:[d,f,m,1]}):[],...e[0].dims.length===4?new Array({type:12,data:[d,m,h*m,1]}):[],...bt(e[0].dims,e[1].dims,e[2].dims,e[3].dims,e[0].dims)],k=z=>{let E=Re("input",e[0].dataType,e[0].dims.length),T=Re("position_ids",e[1].dataType,e[1].dims.length),D=Re("cos_cache",e[2].dataType,e[2].dims.length),I=Re("sin_cache",e[3].dataType,e[3].dims.length),$=mt("output",e[0].dataType,e[0].dims.length);return z.registerUniforms([{name:"scale",type:"f32"},{name:"global_shape",type:"u32",length:g.length},{name:"global_strides",type:"u32",length:v.length},{name:"input_output_strides",type:"u32",length:v.length}]),`
        ${z.declareVariables(E,T,D,I,$)}

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

          if (bsnh[3] < half_rotary_emb_dim) {
            let position_ids_idx =
                ${T.broadcastedIndicesToOffset("bsnh.xy",mt("",T.type.tensor,2))};
            let position_id =
                u32(${T.getByOffset("position_ids_idx")}) + select(0, bsnh[1], position_ids_idx == 0);
            let i = dot(bsnh, uniforms.input_output_strides) + select(0, bsnh[3], ${t});
            let j = i + select(half_rotary_emb_dim, 1, ${t});
            let re = ${E.getByOffset("i")} * ${D.get("position_id","bsnh[3]")} -
                ${E.getByOffset("j")} * ${I.get("position_id","bsnh[3]")};
            ${$.setByOffset("i","re")}
            let im = ${E.getByOffset("i")} * ${I.get("position_id","bsnh[3]")} +
                ${E.getByOffset("j")} * ${D.get("position_id","bsnh[3]")};
            ${$.setByOffset("j","im")}
          } else {
            let k = dot(bsnh, uniforms.input_output_strides) + half_rotary_emb_dim;
            ${$.setByOffset("k",E.getByOffset("k"))}
          }
        }`};return{name:"RotaryEmbedding",shaderCache:{hint:Zt({interleaved:t}).cacheKey,inputDependencies:["rank","rank","rank","rank"]},getShaderSource:k,getRunData:()=>({outputs:[{dims:e[0].dims,dataType:e[0].dataType}],dispatchGroup:{x:Math.ceil(Fe.size(g)/Wa)},programUniforms:b})}},Yb=(e,r)=>{qy(e.inputs,r),e.compute(af(e.inputs,r))}}),Qy,Xy,cm,Jy,ev,AT=Je(()=>{vr(),St(),l_(),qb(),Jb(),Ns(),Zb(),Lt(),Qy=(e,r)=>{if(r.doRotary&&e.length<=7)throw new Error("cos_cache and sin_cache inputs are required if do_rotary is specified");let t=e[0],n=e[1],o=e[2],i=e[3],u=e[4];if(r.doRotary!==0&&e.length<=7)throw new Error("cos_cast and sin_cache are expected if do_rotary attribute is non-zero");if(r.localWindowSize!==-1)throw new Error("Local attention is not supported");if(r.softcap!==0)throw new Error("Softcap is not supported");if(r.rotaryInterleaved!==0)throw new Error("Rotary interleaved is not supported");if(r.smoothSoftmax)throw new Error("Smooth softmax is not supported");if(t.dims.length!==3&&t.dims.length!==5)throw new Error("Input query is expected to have 3 or 5 dimensions");let d=!1,h=t.dims[0],f=t.dims[1],y=t.dims.length===3?d?t.dims[2]/3:t.dims[2]:r.numHeads*t.dims[4],m=f,g=0,v=!n||n.dims.length===0,b=Math.floor(v?y/(r.numHeads+2*r.kvNumHeads):y/r.numHeads);v&&(y=b*r.numHeads);let k=i&&i.dims.length!==0,z=u&&u.dims.length!==0;if(k&&i.dims.length===4&&i.dims[0]===h&&i.dims[1]!==r.kvNumHeads&&i.dims[2]===r.kvNumHeads&&i.dims[3]===b)throw new Error("BSNH pastKey/pastValue is not supported");if(k&&z){if(i.dims.length!==4)throw new Error('Input "past_key" is expected to have 4 dimensions');if(u.dims.length!==4)throw new Error('Input "past_value" is expected to have 4 dimensions');g=i.dims[2]}else if(k||z)throw new Error('Input "past_key" and "past_value" shall be both present or both absent');let E=1;if(n&&n.dims.length>0){if(t.dims.length!==3)throw new Error('Input "query" is expected to have 3 dimensions when key is given');if(n.dims.length<3||n.dims.length>5)throw new Error('Input "key" is expected to have 3, 4, or 5 dimensions');if(t.dims[0]!==n.dims[0])throw new Error('Input "query" and "key" shall have same dim 0 (batch size)');if(n.dims.length===3){if(t.dims[2]%n.dims[2]!==0)throw new Error('Dimension 2 of "query" should be a multiple of "key"');m=n.dims[1]}else if(n.dims.length===5){if(n.dims[2]!==r.numHeads||n.dims[3]!==2||n.dims[4]!==b)throw new Error('Expect "key" shape (batch_size, kv_sequence_length, num_heads, 2, head_size) for packed kv');if(o)throw new Error('Expect "value" be none when "key" has packed kv format.');m=n.dims[1]}else{if(n.dims[1]!==r.numHeads||n.dims[3]!==b)throw new Error('Expect "key" shape (batch_size, num_heads, kv_sequence_length, head_size) for past_key');m=n.dims[2]}}else{if(t.dims.length!==3&&t.dims.length!==5)throw new Error('Input "query" is expected to have 3 or 5 dimensions when key is empty');if(t.dims.length===5&&(t.dims[2]!==r.numHeads||t.dims[3]!==3))throw new Error('Expect "query" shape (batch_size, kv_sequence_length, num_heads, 3, head_size) for packed kv');E=3}let T=0,D=!1,I=r.kvNumHeads?b*r.kvNumHeads:y;if(o&&o.dims.length>0){if(o.dims.length!==3&&o.dims.length!==4)throw new Error('Input "value" is expected to have 3 or 4 dimensions');if(t.dims[0]!==o.dims[0])throw new Error('Input "query" and "value" shall have same dim 0 (batch_size)');if(o.dims.length===3){if(m!==o.dims[1])throw new Error('Input "key" and "value" shall have the same dim 1 (kv_sequence_length)');I=o.dims[2]}else{if(m!==o.dims[2])throw new Error('Input "past_key" and "past_value" shall have the same dim 2 (kv_sequence_length)');I=o.dims[1]*o.dims[3],D=!0}}let $=e.length>4?e[5]:void 0;if($&&$.dims.length!==1&&$.dims[0]!==h)throw new Error('Input "seqlens" is expected to have 1 dimension and the same dim 0 as batch_size');return{batchSize:h,sequenceLength:f,pastSequenceLength:g,kvSequenceLength:m,totalSequenceLength:-1,maxSequenceLength:-1,inputHiddenSize:0,hiddenSize:y,vHiddenSize:I,headSize:b,vHeadSize:Math.floor(I/r.kvNumHeads),numHeads:r.numHeads,kvNumHeads:r.kvNumHeads,nReps:r.numHeads/r.kvNumHeads,pastPresentShareBuffer:!1,maskType:T,scale:r.scale,broadcastResPosBias:!1,passPastInKv:D,qkvFormat:E}},Xy=Zt({perm:[0,2,1,3]}),cm=(e,r,t)=>{let n=r,o=t.kvNumHeads;return r.dims.length===3&&t.kvSequenceLength!==0&&(n=r.reshape([t.batchSize,t.kvSequenceLength,o,t.headSize]),n=e.compute(xi(n,Xy.perm),{inputs:[n],outputs:[-1]})[0]),n},Jy=(e,r,t,n)=>{let o=7,i=["type","type"],u=[e*r],d=e*r,h=[{type:12,data:d},{type:12,data:r},{type:12,data:e}],f=y=>{let m=Re("seq_lens",t.dataType,t.dims),g=Re("total_seq_lens",n.dataType,n.dims),v=mt("pos_ids",o,u),b=[{name:"output_size",type:"u32"},{name:"sequence_length",type:"u32"},{name:"batch_size",type:"u32"}];return`
  ${y.registerUniforms(b).declareVariables(m,g,v)}
  ${y.mainStart()}
    ${y.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.output_size")}
    let total_sequence_length = u32(${g.getByOffset("0")});
    let is_subsequent_prompt = uniforms.sequence_length > 1 && uniforms.sequence_length != total_sequence_length;
    let is_first_prompt = !is_subsequent_prompt && uniforms.sequence_length == total_sequence_length;
    let batch_idx = global_idx / uniforms.sequence_length;
    let sequence_idx = i32(global_idx % uniforms.sequence_length);
    var pos_id: i32 = 0;
    let seqlen = ${m.getByOffset("batch_idx")};
    let total_seqlen = seqlen + 1;
    if (is_first_prompt) {
      if (sequence_idx < total_seqlen) {
        pos_id = sequence_idx;
      } else {
        pos_id = 1;
      }
      ${v.setByOffset("global_idx","pos_id")}
    } else if (is_subsequent_prompt) {
      let past_seqlen = total_seqlen - i32(uniforms.sequence_length);
      if (past_seqlen + sequence_idx < total_seqlen) {
        pos_id = past_seqlen + sequence_idx;
      } else {
        pos_id = 1;
      }
      ${v.setByOffset("global_idx","pos_id")}
    } else if (global_idx < uniforms.batch_size) {
      ${v.setByOffset("global_idx","seqlen")}
    };
  }
  `};return{name:"GeneratePositionIds",shaderCache:{hint:`${e};${r}`,inputDependencies:i},getRunData:()=>({outputs:[{dims:u,dataType:o}],dispatchGroup:{x:Math.ceil(d/64)},programUniforms:h}),getShaderSource:f}},ev=(e,r)=>{var I;let t=Qy(e.inputs,r);if(e.inputs[0].dims.length===5)throw new Error("Packed QKV is not implemented");if(((I=e.inputs[1])==null?void 0:I.dims.length)===5)throw new Error("Packed KV is not implemented");let n=e.inputs[0],o=e.inputs[1]&&e.inputs[1].dims.length>0?e.inputs[1]:void 0,i=e.inputs[2]&&e.inputs[2].dims.length>0?e.inputs[2]:void 0,u=e.inputs[3]&&e.inputs[3].dims.length!==0?e.inputs[3]:void 0,d=e.inputs[4]&&e.inputs[4].dims.length!==0?e.inputs[4]:void 0,h=e.inputs.length>4?e.inputs[5]:void 0,f=e.inputs.length>5?e.inputs[6]:void 0,y=t.kvNumHeads?t.kvNumHeads:t.numHeads,m=Zt({axis:2,numOutputs:3,splitSizes:[t.numHeads*t.headSize,y*t.headSize,y*t.headSize]}),[g,v,b]=!o&&!i?e.compute(Wm([n],m),{inputs:[n],outputs:[-1,-1,-1]}):[n,o,i],k,z;if(r.doRotary){let $=e.compute(Jy(t.batchSize,t.sequenceLength,h,f),{inputs:[h,f],outputs:[-1]})[0],P=e.inputs[7],S=e.inputs[8],O=Zt({interleaved:r.rotaryInterleaved!==0,numHeads:t.numHeads,rotaryEmbeddingDim:0,scale:r.scale}),R=[g,$,P,S],G=[-1];k=e.compute(af(R,O),{inputs:R,outputs:G})[0],R.splice(0,1,v);let ee=Zt({interleaved:r.rotaryInterleaved!==0,numHeads:t.kvNumHeads,rotaryEmbeddingDim:0,scale:r.scale});z=e.compute(af(R,ee),{inputs:R,outputs:G})[0]}let E=fd(e,t.batchSize,t.numHeads,t.sequenceLength,t.headSize,r.doRotary?k:g,void 0,0),T=cm(e,r.doRotary?z:v,t),D=cm(e,b,t);wd(e,E,T,D,void 0,void 0,u,d,void 0,t,h,f)}}),um,Yy,Zy,tv,OT=Je(()=>{St(),Ft(),Ns(),Lt(),um=(e,r,t,n,o,i,u,d)=>{let h=yr(i),f=h===1?"f32":`vec${h}f`,y=h===1?"vec2f":`mat2x${h}f`,m=o*u,g=64;m===1&&(g=256);let v=[o,u,i/h],b=[o,u,2],k=["rank","type","type"],z=[];z.push(...bt(v,b));let E=T=>{let D=Re("x",r.dataType,3,h),I=Re("scale",t.dataType,t.dims),$=Re("bias",n.dataType,n.dims),P=mt("output",1,3,2),S=[D,I,$,P];return`
  var<workgroup> workgroup_shared : array<${y}, ${g}>;
  const workgroup_size = ${g}u;
  ${T.declareVariables(...S)}
  ${T.mainStart(g)}
    let batch = workgroup_index / uniforms.x_shape[1];
    let channel = workgroup_index % uniforms.x_shape[1];
    let hight = uniforms.x_shape[2];
    // initialize workgroup memory
    var sum = ${f}(0);
    var squared_sum = ${f}(0);
    for (var h = local_idx; h < hight; h += workgroup_size) {
      let value = ${f}(${D.get("batch","channel","h")});
      sum += value;
      squared_sum += value * value;
    }
    workgroup_shared[local_idx] = ${y}(sum, squared_sum);
    workgroupBarrier();

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

      let inv_std_dev = inverseSqrt(squared_sum_final - sum_final * sum_final + f32(${d}));
      let channel_scale = inv_std_dev * f32(scale[channel]);
      let channel_shift = f32(bias[channel]) - sum_final * channel_scale;
      output[workgroup_index] = vec2f(channel_scale, channel_shift);
    }
  }`};return e.compute({name:"InstanceNormComputeChannelScaleShift",shaderCache:{hint:`${h};${d};${g}`,inputDependencies:k},getRunData:()=>({outputs:[{dims:b,dataType:1}],dispatchGroup:{x:m},programUniforms:z}),getShaderSource:E},{inputs:[r,t,n],outputs:[-1]})[0]},Yy=(e,r,t)=>{let n=r[0].dims,o=n,i=2,u=n[0],d=n[1],h=Fe.sizeFromDimension(n,i),f=yr(h),y=Fe.size(o)/f,m=um(e,r[0],r[1],r[2],u,h,d,t.epsilon),g=[u,d,h/f],v=[u,d],b=["type","none"],k=z=>{let E=Re("x",r[0].dataType,g.length,f),T=Re("scale_shift",1,v.length,2),D=mt("output",r[0].dataType,g.length,f),I=[E,T,D];return`
  ${z.registerUniform("output_size","u32").declareVariables(...I)}
  ${z.mainStart()}
  ${z.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.output_size")}
      let outputIndices = ${D.offsetToIndices("global_idx")};
      let batch = outputIndices[0];
      let channel = outputIndices[1];
      let scale_shift = ${T.getByIndices("vec2<u32>(batch, channel)")};
      let value = ${E.getByOffset("global_idx")} * ${D.type.value}(scale_shift.x) + ${D.type.value}(scale_shift.y);
      ${D.setByOffset("global_idx","value")};
  }`};e.compute({name:"InstanceNormalization",shaderCache:{hint:`${f}`,inputDependencies:b},getRunData:()=>({outputs:[{dims:o,dataType:r[0].dataType}],dispatchGroup:{x:Math.ceil(y/64)},programUniforms:[{type:12,data:y},...bt(g,v,g)]}),getShaderSource:k},{inputs:[r[0],m]})},Zy=(e,r,t)=>{let n=r[0].dims,o=n,i=n[0],u=n[n.length-1],d=Fe.sizeFromDimension(n,1)/u,h=yr(u),f=Fe.size(o)/h,y=[{type:12,data:d},{type:12,data:Math.floor(u/h)}],m=["type","type"],g=!1,v=[0,n.length-1];for(let E=0;E<n.length-2;E++)g=g||n[E+1]!==1,v.push(E+1);g=g&&n[n.length-1]!==1;let b=g?e.compute(xi(e.inputs[0],v),{inputs:[e.inputs[0]],outputs:[-1]})[0]:e.inputs[0].reshape(Array.from({length:n.length},(E,T)=>n[v[T]])),k=um(e,b,r[1],r[2],i,d,u,t.epsilon),z=E=>{let T=jr(r[0].dataType),D=h===1?"vec2f":`mat${h}x2f`,I=S=>{let O=S===0?"x":"y",R=h===1?"f32":`vec${h}f`;switch(h){case 1:return`${T}(${R}(scale.${O}))`;case 2:return`vec2<${T}>(${R}(scale[0].${O}, scale[1].${O}))`;case 4:return`vec4<${T}>(${R}(scale[0].${O}, scale[1].${O}, scale[2].${O}, scale[3].${O}))`;default:throw new Error(`Not supported compoents ${h}`)}},$=Re("input",r[0].dataType,r[0].dims,h),P=mt("output",r[0].dataType,o,h);return`
  @group(0) @binding(0) var<storage, read> input : array<${$.type.storage}>;
  @group(0) @binding(1) var<storage, read> scale_input : array<${D}>;
  @group(0) @binding(2) var<storage, read_write> output : array<${P.type.storage}>;
  struct Uniforms {H: u32, C : u32};
  @group(0) @binding(3) var<uniform> uniforms: Uniforms;

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

    let scale_offset = current_image_number * uniforms.C + current_channel_number;
    let scale = scale_input[scale_offset];
    output[global_idx] = fma(input[global_idx], ${I(0)}, ${I(1)});
  }`};e.compute({name:"InstanceNormalizationNHWC",shaderCache:{hint:`${h}`,inputDependencies:m},getRunData:()=>({outputs:[{dims:o,dataType:r[0].dataType}],dispatchGroup:{x:Math.ceil(f/64)},programUniforms:y}),getShaderSource:z},{inputs:[r[0],k]})},tv=(e,r)=>{r.format==="NHWC"?Zy(e,e.inputs,r):Yy(e,e.inputs,r)}}),ew,tw,rv,FT=Je(()=>{St(),Ft(),Lt(),ew=e=>{if(!e||e.length<2)throw new Error("layerNorm requires at least 2 inputs.")},tw=(e,r,t)=>{let n=r.simplified,o=e[0].dims,i=e[1],u=!n&&e[2],d=o,h=Fe.normalizeAxis(r.axis,o.length),f=Fe.sizeToDimension(o,h),y=Fe.sizeFromDimension(o,h),m=Fe.size(i.dims),g=u?Fe.size(u.dims):0;if(m!==y||u&&g!==y)throw new Error(`Size of X.shape()[axis:] == ${y}.
       Size of scale and bias (if provided) must match this.
       Got scale size of ${m} and bias size of ${g}`);let v=[];for(let $=0;$<o.length;++$)$<h?v.push(o[$]):v.push(1);let b=yr(y),k=["type","type"],z=[{type:12,data:f},{type:1,data:y},{type:12,data:Math.floor(y/b)},{type:1,data:r.epsilon}];u&&k.push("type");let E=t>1,T=t>2,D=$=>{let P=jr(e[0].dataType),S=[Re("x",e[0].dataType,e[0].dims,b),Re("scale",i.dataType,i.dims,b)];u&&S.push(Re("bias",u.dataType,u.dims,b)),S.push(mt("output",e[0].dataType,d,b)),E&&S.push(mt("mean_data_output",1,v)),T&&S.push(mt("inv_std_output",1,v));let O=[{name:"norm_count",type:"u32"},{name:"norm_size",type:"f32"},{name:"norm_size_vectorized",type:"u32"},{name:"epsilon",type:"f32"}];return`
  ${$.registerUniforms(O).declareVariables(...S)}
  ${$.mainStart()}
    ${$.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.norm_count")}
    let offset = global_idx * uniforms.norm_size_vectorized;
    var mean_vector = ${Dm("f32",b)};
    var mean_square_vector = ${Dm("f32",b)};

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

    for (var j: u32 = 0; j < uniforms.norm_size_vectorized; j++) {
      let f32input = ${Na(P,b,"x[j + offset]")};
      let f32scale = ${Na(P,b,"scale[j]")};
      output[j + offset] = ${S[0].type.value}((f32input ${n?"":"- mean"}) * inv_std_dev * f32scale
        ${u?`+ ${Na(P,b,"bias[j]")}`:""}
      );
    }

    ${E?"mean_data_output[global_idx] = mean":""};
    ${T?"inv_std_output[global_idx] = inv_std_dev":""};
  }`},I=[{dims:d,dataType:e[0].dataType}];return E&&I.push({dims:v,dataType:1}),T&&I.push({dims:v,dataType:1}),{name:"LayerNormalization",shaderCache:{hint:`${b};${t};${n}`,inputDependencies:k},getRunData:()=>({outputs:I,dispatchGroup:{x:Math.ceil(f/64)},programUniforms:z}),getShaderSource:D}},rv=(e,r)=>{ew(e.inputs),e.compute(tw(e.inputs,r,e.outputCount))}}),rw,iv,DT=Je(()=>{Ft(),h_(),f_(),rw=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.")},iv=e=>{rw(e.inputs);let r=Ua.calcShape(e.inputs[0].dims,e.inputs[1].dims,!0);if(!r)throw new Error("Can't use matmul on the given tensors");let t=r[r.length-1],n=e.inputs[0].dims[e.inputs[0].dims.length-1];if(t<8&&n<8)e.compute(p_(e.inputs,{activation:""},r));else{let o=r[r.length-2],i=Fe.size(e.inputs[0].dims.slice(0,-2)),u=Fe.size(e.inputs[1].dims.slice(0,-2));if(i!==1&&o===1&&u===1){let d=e.inputs[0].reshape([1,i,n]),h=e.inputs[1].reshape([1,n,t]),f=[1,i,t],y=[d,h];e.compute(nf(y,{activation:""},r,f),{inputs:y})}else e.compute(nf(e.inputs,{activation:""},r))}}}),iw,sw,nw,sv,nv,LT=Je(()=>{St(),Ft(),vr(),Lt(),iw=(e,r)=>{if(e.length<3||e.length>4)throw new Error("MatMulNBits requires 3 or 4 inputs");let t=e[0],n=t.dims.length;if(t.dims[n-1]!==r.k)throw new Error("The last dim of input shape does not match the k value");let o=Math.floor((r.k+r.blockSize-1)/r.blockSize),i=r.blockSize/8*r.bits,u=e[1];if(!Fe.areEqual(u.dims,[r.n,o,i]))throw new Error("The second inputs must be 3D tensor with shape N X nBlocksPerCol X blobSize");let d=e[2].dims;if(Fe.size(d)!==r.n*o)throw new Error("scales input size error.");if(e.length===4){let h=e[3].dims,f=r.bits>4?r.n*o:r.n*Math.floor((o+1)/2);if(Fe.size(h)!==f)throw new Error("zeroPoints input size error.")}},sw=(e,r)=>{let t=e[0].dims,n=t.length,o=t[n-2],i=r.k,u=r.n,d=t.slice(0,n-2),h=Fe.size(d),f=e[1].dims[2]/4,y=e[0].dataType,m=yr(r.k),g=yr(f),v=yr(u),b=d.concat([o,u]),k=o>1&&u/v%2===0?2:1,z=Fe.size(b)/v/k,E=64,T=[],D=[h,o,i/m],I=Fe.convertShape(e[1].dims).slice();I.splice(-1,1,f/g),T.push(...bt(D)),T.push(...bt(I)),T.push(...bt(e[2].dims)),e.length===4&&T.push(...bt(Fe.convertShape(e[3].dims)));let $=[h,o,u/v];T.push(...bt($));let P=S=>{let O=D.length,R=Re("a",e[0].dataType,O,m),G=Re("b",12,I.length,g),ee=Re("scales",e[2].dataType,e[2].dims.length),ie=[R,G,ee],H=e.length===4?Re("zero_points",12,e[3].dims.length):void 0;H&&ie.push(H);let ce=$.length,re=mt("output",e[0].dataType,ce,v),se=jr(e[0].dataType),_e=(()=>{switch(m){case 1:return`array<${se}, 8>`;case 2:return`mat4x2<${se}>`;case 4:return`mat2x4<${se}>`;default:throw new Error(`${m}-component is not supported.`)}})(),ae=()=>{let q=`
          // reuse a data
            var input_offset = ${R.indicesToOffset(`${R.type.indices}(batch, row, word_offset)`)};
            var a_data: ${_e};
            for (var j: u32 = 0; j < ${8/m}; j++) {
              a_data[j] = ${R.getByOffset("input_offset")};
              input_offset++;
            }
          `;for(let B=0;B<v*k;B++)q+=`
            b_value = ${g===1?`b${B}_data`:`b${B}_data[i]`};
            b_value_lower = unpack4xU8(b_value & b_mask);
            b_value_upper = unpack4xU8((b_value >> 4) & b_mask);
            b_quantized_values = ${_e}(${Array.from({length:4},(Q,oe)=>`${se}(b_value_lower[${oe}]), ${se}(b_value_upper[${oe}])`).join(", ")});
            b_dequantized_values = ${m===1?`${_e}(${Array.from({length:8},(Q,oe)=>`(b_quantized_values[${oe}] - ${H?`zero_point${B}`:"zero_point"}) * scale${B}`).join(", ")});`:`(b_quantized_values - ${_e}(${Array(8).fill(`${H?`zero_point${B}`:"zero_point"}`).join(",")})) * scale${B};`};
            workgroup_shared[local_id.x * ${k} + ${Math.floor(B/v)}]${v>1?`[${B%v}]`:""} += ${Array.from({length:8/m},(Q,oe)=>`${m===1?`a_data[${oe}] * b_dequantized_values[${oe}]`:`dot(a_data[${oe}], b_dequantized_values[${oe}])`}`).join(" + ")};
          `;return q},Ce=()=>{let q=`
            var col_index = col * ${v};
            ${H?`
            let zero_point_bytes_per_col = (nBlocksPerCol + 1) / 2;
            var zero_point_byte_count: u32;
            var zero_point_word_index: u32;
            var zero_point_byte_offset: u32;
            let zero_point_nibble_offset: u32 = block & 0x1u;
            var zero_point_bits_offset: u32;
            var zero_point_word: u32;`:`
            // The default zero point is 8 for unsigned 4-bit quantization.
            let zero_point = ${se}(8);`}
            `;for(let B=0;B<v*k;B++)q+=`
            let scale${B} = ${ee.getByOffset("col_index * nBlocksPerCol + block")};
            ${H?`
            zero_point_byte_count = col_index * zero_point_bytes_per_col + (block >> 0x1u);
            zero_point_word_index = zero_point_byte_count >> 0x2u;
            zero_point_byte_offset = zero_point_byte_count & 0x3u;
            zero_point_bits_offset = (zero_point_byte_offset << 3) + (zero_point_nibble_offset << 2);
            zero_point_word = ${H.getByOffset("zero_point_word_index")} >> zero_point_bits_offset;
            let zero_point${B} = ${se}((zero_point_word) & 0xFu);`:""}
            col_index += 1;`;return q},Te=()=>{let q=`col_index = col * ${v};`;for(let B=0;B<v*k;B++)q+=`
            let b${B}_data = ${G.getByIndices(`${G.type.indices}(col_index, block, word)`)};
            col_index += 1;`;return q+=`
            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: ${_e};
            var b_dequantized_values: ${_e};`,q};return`
        var<workgroup> workgroup_shared: array<${re.type.value}, ${k*E}>;
        ${S.declareVariables(...ie,re)}
        ${S.mainStart([E,1,1])}
          let output_indices = ${re.offsetToIndices(`(global_idx / ${E}) * ${k}`)};
          let col = output_indices[2];
          let row = output_indices[1];
          let batch = output_indices[0];
          let nBlocksPerCol = uniforms.b_shape[1];

          for (var block = local_id.x; block < nBlocksPerCol; block += ${E}) {
            //process one block
            var word_offset: u32 = block * ${r.blockSize/m};
            ${Ce()}
            for (var word: u32 = 0; word < ${f}; word += ${g}) {
              ${Te()}
              for (var i: u32 = 0; i < ${g}; i++) {
                ${ae()}
                word_offset += ${8/m};
              }
            }
          }
          workgroupBarrier();

          if (local_id.x < ${k}) {
            var output_value: ${re.type.value} = ${re.type.value}(0);
            var workgroup_shared_offset: u32 = local_id.x;
            for (var b: u32 = 0u; b < ${E}u; b++) {
              output_value += workgroup_shared[workgroup_shared_offset];
              workgroup_shared_offset += ${k};
            }
            ${re.setByIndices(`${re.type.indices}(batch, row, col + local_id.x)`,"output_value")};
          }
        }`};return{name:"MatMulNBits",shaderCache:{hint:`${r.blockSize};${r.bits};${m};${g};${v};${k};${E}`,inputDependencies:Array(e.length).fill("rank")},getRunData:()=>({outputs:[{dims:b,dataType:y}],dispatchGroup:{x:z},programUniforms:T}),getShaderSource:P}},nw=(e,r)=>{let t=e[0].dims,n=t.length,o=t[n-2],i=r.k,u=r.n,d=t.slice(0,n-2),h=Fe.size(d),f=e[1].dims[2]/4,y=e[0].dataType,m=yr(r.k),g=yr(f),v=d.concat([o,u]),b=128,k=u%8===0?8:u%4===0?4:1,z=b/k,E=z*g*8,T=E/m,D=E/r.blockSize,I=Fe.size(v)/k,$=[],P=[h,o,i/m],S=Fe.convertShape(e[1].dims).slice();S.splice(-1,1,f/g),$.push(...bt(P)),$.push(...bt(S)),$.push(...bt(e[2].dims)),e.length===4&&$.push(...bt(Fe.convertShape(e[3].dims)));let O=[h,o,u];$.push(...bt(O));let R=G=>{let ee=P.length,ie=Re("a",e[0].dataType,ee,m),H=Re("b",12,S.length,g),ce=Re("scales",e[2].dataType,e[2].dims.length),re=[ie,H,ce],se=e.length===4?Re("zero_points",12,e[3].dims.length):void 0;se&&re.push(se);let _e=O.length,ae=mt("output",e[0].dataType,_e),Ce=jr(e[0].dataType),Te=()=>{switch(m){case 1:return`
          let a_data0 = vec4<${Ce}>(sub_a[word_offset], sub_a[word_offset + 1], sub_a[word_offset + 2], sub_a[word_offset + 3]);
          let a_data1 = vec4<${Ce}>(sub_a[word_offset + 4], sub_a[word_offset + 5], sub_a[word_offset + 6], sub_a[word_offset + 7]);`;case 2:return`
          let a_data0 = vec4<${Ce}>(sub_a[word_offset], sub_a[word_offset + 1]);
          let a_data1 = vec4<${Ce}>(sub_a[word_offset + 2], sub_a[word_offset + 3]);`;case 4:return`
          let a_data0 = sub_a[word_offset];
          let a_data1 = sub_a[word_offset + 1];`;default:throw new Error(`${m}-component is not supported.`)}};return`
        var<workgroup> sub_a: array<${ie.type.value}, ${T}>;
        var<workgroup> inter_results: array<array<${ae.type.value}, ${z}>, ${k}>;
        ${G.declareVariables(...re,ae)}
        ${G.mainStart([z,k,1])}
          let output_indices = ${ae.offsetToIndices(`workgroup_index * ${k}`)};
          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) / ${D} + 1;

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

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

          if (local_idx < ${k}) {
            var output_value: ${ae.type.value} = ${ae.type.value}(0);
            for (var b = 0u; b < ${z}; b++) {
              output_value += inter_results[local_idx][b];
            }
            if (col + local_idx < uniforms.output_shape[2])
            {
              ${ae.setByIndices(`${ae.type.indices}(batch, row, col + local_idx)`,"output_value")}
            }
          }
        }`};return{name:"BlockwiseMatMulNBits32",shaderCache:{hint:`${r.blockSize};${m};${g};${z};${k}`,inputDependencies:Array(e.length).fill("rank")},getRunData:()=>({outputs:[{dims:v,dataType:y}],dispatchGroup:{x:I},programUniforms:$}),getShaderSource:R}},sv=(e,r)=>{iw(e.inputs,r),r.blockSize===32&&e.adapterInfo.isVendor("intel")&&e.adapterInfo.isArchitecture("gen-12lp")?e.compute(nw(e.inputs,r)):e.compute(sw(e.inputs,r))},nv=e=>Zt(e)}),aw,ow,lw,cw,uw,dw,pw,hw,av,jT=Je(()=>{St(),Ft(),Lt(),aw=e=>{if(!e||e.length<1)throw new Error("Too few inputs");if(e[0].dataType!==1&&e[0].dataType!==10)throw new Error("Input type must be float or float16.");if(e.length>=2){let r=e[0].dims.length*2===e[1].dims[0];if(e.length===4&&(r=e[3].dims[0]*2===e[1].dims[0]),!r)throw new Error("The pads should be a 1D tensor of shape [2 * input_rank] or [2 * num_axes].")}},ow=(e,r,t)=>{let n="";for(let o=r-1;o>=0;--o)n+=`
            k = i32(${e.indicesGet("indices",o)}) - ${wt("uniforms.pads",o,t)};
            if (k < 0) {
              break;
            }
            if (k >= i32(${wt("uniforms.x_shape",o,r)})) {
              break;
            }
            offset += k * i32(${wt("uniforms.x_strides",o,r)});
        `;return`
          value = ${e.type.value}(uniforms.constant_value);
          for (var i = 0; i < 1; i++) {
            var offset = 0;
            var k = 0;
            ${n}
            value = x[offset];
          }
      `},lw=(e,r,t)=>{let n="";for(let o=r-1;o>=0;--o)n+=`
                k = i32(${e.indicesGet("indices",o)}) - ${wt("uniforms.pads",o,t)};
                if (k < 0) {
                  k = -k;
                }
                {
                  let _2n_1 = 2 * (i32(${wt("uniforms.x_shape",o,r)}) - 1);
                  k = k % _2n_1;
                  if(k >= i32(${wt("uniforms.x_shape",o,r)})) {
                    k = _2n_1 - k;
                  }
                }
                offset += k * i32(${wt("uniforms.x_strides",o,r)});
            `;return`
              var offset = 0;
              var k = 0;
              ${n}
              value = x[offset];
          `},cw=(e,r,t)=>{let n="";for(let o=r-1;o>=0;--o)n+=`
                k = i32(${e.indicesGet("indices",o)}) - ${wt("uniforms.pads",o,t)};
                if (k < 0) {
                  k = 0;
                }
                if (k >= i32(${wt("uniforms.x_shape",o,r)})) {
                  k = i32(${wt("uniforms.x_shape",o,r)}) - 1;
                }
                offset += k * i32(${wt("uniforms.x_strides",o,r)});
            `;return`
              var offset = 0;
              var k = 0;
              ${n}
              value = x[offset];
          `},uw=(e,r,t)=>{let n="";for(let o=r-1;o>=0;--o)n+=`
                k = i32(${e.indicesGet("indices",o)}) - ${wt("uniforms.pads",o,t)};
                if (k < 0)  {
                  k += i32(${wt("uniforms.x_shape",o,r)}]);
                }
                if (k >= i32(${wt("uniforms.x_shape",o,r)})) {
                  k -= i32(${wt("uniforms.x_shape",o,r)});
                }
                offset += k * i32(${wt("uniforms.x_strides",o,r)});
            `;return`
              var offset = 0;
              var k = 0;
              ${n}
              value = x[offset];
          `},dw=(e,r,t)=>{switch(t.mode){case 0:return ow(e,r,t.pads.length);case 1:return lw(e,r,t.pads.length);case 2:return cw(e,r,t.pads.length);case 3:return uw(e,r,t.pads.length);default:throw new Error("Invalid mode")}},pw=(e,r)=>{let t=Fe.padShape(e[0].dims.slice(),r.pads),n=e[0].dims,o=Fe.size(t),i=[{type:12,data:o},{type:6,data:r.pads}],u=e.length>=3&&e[2].data;r.mode===0&&i.push({type:u?e[2].dataType:1,data:r.value}),i.push(...bt(e[0].dims,t));let d=["rank"],h=f=>{let y=mt("output",e[0].dataType,t.length),m=Re("x",e[0].dataType,n.length),g=m.type.value,v=dw(y,n.length,r),b=[{name:"output_size",type:"u32"},{name:"pads",type:"i32",length:r.pads.length}];return r.mode===0&&b.push({name:"constant_value",type:u?g:"f32"}),`
            ${f.registerUniforms(b).declareVariables(m,y)}
            ${f.mainStart()}
            ${f.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.output_size")}

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

            var value = ${g}(0);
            ${v}
            output[global_idx] = value;
        }`};return{name:"Pad",shaderCache:{hint:`${r.mode}${u}`,inputDependencies:d},getRunData:()=>({outputs:[{dims:t,dataType:e[0].dataType}],dispatchGroup:{x:Math.ceil(Fe.size(t)/64)},programUniforms:i}),getShaderSource:h}},hw=(e,r)=>{if(e.length>1){let t=e[1].getBigInt64Array(),n=e.length>=3&&e[2].data?e[2].dataType===10?e[2].getUint16Array()[0]:e[2].getFloat32Array()[0]:0,o=e[0].dims.length,i=new Int32Array(2*o).fill(0);if(e.length>=4){let d=e[3].getBigInt64Array();for(let h=0;h<d.length;h++)i[Number(d[h])]=Number(t[h]),i[Number(d[h])+o]=Number(t[h+d.length])}else t.forEach((d,h)=>i[Number(h)]=Number(d));let u=[];return i.forEach(d=>u.push(d)),{mode:r.mode,value:n,pads:u}}else return r},av=(e,r)=>{aw(e.inputs);let t=hw(e.inputs,r);e.compute(pw(e.inputs,t),{inputs:[0]})}}),od,dm,pm,hm,fm,fw,mw,mm,_m,ov,lv,gm,cv,uv,ym,dv,pv,hv,fv,zT=Je(()=>{Ii(),St(),Ft(),Lt(),od=e=>{if(dr.webgpu.validateInputContent&&(!e||e.length!==1))throw new Error("Pool ops requires 1 input.")},dm=(e,r,t)=>{let n=r.format==="NHWC",o=e.dims.slice();n&&o.splice(1,0,o.pop());let i=Object.hasOwnProperty.call(r,"dilations"),u=r.kernelShape.slice(),d=r.strides.slice(),h=i?r.dilations.slice():[],f=r.pads.slice();rf.adjustPoolAttributes(t,o,u,d,h,f);let y=rf.computePoolOutputShape(t,o,d,h,u,f,r.autoPad),m=Object.assign({},r);i?Object.assign(m,{kernelShape:u,strides:d,pads:f,dilations:h,cacheKey:r.cacheKey}):Object.assign(m,{kernelShape:u,strides:d,pads:f,cacheKey:r.cacheKey});let g=y.slice();return g.push(g.splice(1,1)[0]),[m,n?g:y]},pm=(e,r)=>{let t=r.format==="NHWC",n=Fe.size(e),o=Fe.size(r.kernelShape),i=[{type:12,data:n},{type:12,data:o}],u=[{name:"outputSize",type:"u32"},{name:"kernelSize",type:"u32"}];if(r.kernelShape.length<=2){let d=r.kernelShape[r.kernelShape.length-1],h=r.strides[r.strides.length-1],f=r.pads[r.pads.length/2-1],y=r.pads[r.pads.length-1],m=!!(f+y);i.push({type:12,data:d},{type:12,data:h},{type:12,data:f},{type:12,data:y}),u.push({name:"kw",type:"u32"},{name:"sw",type:"u32"},{name:"pwStart",type:"u32"},{name:"pwEnd",type:"u32"});let g=!1;if(r.kernelShape.length===2){let v=r.kernelShape[r.kernelShape.length-2],b=r.strides[r.strides.length-2],k=r.pads[r.pads.length/2-2],z=r.pads[r.pads.length-2];g=!!(k+z),i.push({type:12,data:v},{type:12,data:b},{type:12,data:k},{type:12,data:z}),u.push({name:"kh",type:"u32"},{name:"sh",type:"u32"},{name:"phStart",type:"u32"},{name:"phEnd",type:"u32"})}return[i,u,!0,m,g]}else{if(t)throw new Error("Pooling with kernelShape.length > 2 is not supported for NHWC format.");let d=Fe.computeStrides(r.kernelShape);i.push({type:12,data:d},{type:12,data:r.pads},{type:12,data:r.strides}),u.push({name:"kernelStrides",type:"u32",length:d.length},{name:"pads",type:"u32",length:r.pads.length},{name:"strides",type:"u32",length:r.strides.length});let h=r.pads.reduce((f,y)=>f+y);return[i,u,!!h,!1,!1]}},hm=(e,r,t,n,o,i,u,d,h,f,y,m)=>{let g=o.format==="NHWC",v=r.type.value,b=mt("output",r.type.tensor,n);if(o.kernelShape.length<=2){let k="",z="",E="",T=t-(g?2:1);if(y?k=`
                for (var i: u32 = 0u; i < uniforms.kw; i++) {
                  xIndices[${T}] = indices[${T}] * uniforms.sw - uniforms.pwStart + i;
                  if (xIndices[${T}] < 0 || xIndices[${T}]
                      >= uniforms.x_shape[${T}]) {
                    pad++;
                    continue;
                  }
                  let x_val = x[${r.indicesToOffset("xIndices")}];
                  ${i}
                }`:k=`
                for (var i: u32 = 0u; i < uniforms.kw; i++) {
                  xIndices[${T}] = indices[${T}] * uniforms.sw - uniforms.pwStart + i;
                  let x_val = x[${r.indicesToOffset("xIndices")}];
                  ${i}
                }`,o.kernelShape.length===2){let D=t-(g?3:2);m?z=`
                for (var j: u32 = 0u; j < uniforms.kh; j++) {
                  xIndices[${D}] = indices[${D}] * uniforms.sh - uniforms.phStart + j;
                  if (xIndices[${D}] < 0 || xIndices[${D}] >= uniforms.x_shape[${D}]) {
                    pad += i32(uniforms.kw);
                    continue;
                  }
              `:z=`
                for (var j: u32 = 0u; j < uniforms.kh; j++) {
                  xIndices[${D}] = indices[${D}] * uniforms.sh - uniforms.phStart + j;
                `,E=`
              }
            `}return`
            ${e.registerUniforms(h).declareVariables(r,b)}

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

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

              var value = ${v}(${d});
              var pad = 0;
              ${z}
              ${k}
              ${E}
              ${u}

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

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

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

              var value = ${v}(${d});
              var pad = 0;
              var isPad = false;

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

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

              output[global_idx] = value;
            }`}},fm=e=>`${e.format};${e.ceilMode};${e.autoPad};${e.kernelShape.length}`,fw=e=>`${fm(e)};${e.countIncludePad}`,mw=e=>`${fm(e)};${e.storageOrder};${e.dilations}`,mm=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}),_m=(e,r,t,n)=>{let[o,i]=dm(r,n,t),u=Re("x",r.dataType,r.dims.length),d=u.type.value,h="value += x_val;",f="";o.countIncludePad?f+=`value /= ${d}(uniforms.kernelSize);`:f+=`value /= ${d}(i32(uniforms.kernelSize) - pad);`;let[y,m,g,v,b]=pm(i,o);y.push(...bt(r.dims,i));let k=["rank"];return{name:e,shaderCache:{hint:`${n.cacheKey};${g};${v};${b}`,inputDependencies:k},getRunData:()=>({outputs:[{dims:i,dataType:r.dataType}],dispatchGroup:{x:Math.ceil(Fe.size(i)/64)},programUniforms:y}),getShaderSource:z=>hm(z,u,r.dims.length,i.length,o,h,f,0,m,g,v,b)}},ov=e=>{let r=e.count_include_pad!==0,t=mm(e);if(t.ceilMode!==0)throw new Error("using ceil() in shape computation is not yet supported for AveragePool");let n={countIncludePad:r,...t,cacheKey:""};return{...n,cacheKey:fw(n)}},lv=(e,r)=>{od(e.inputs),e.compute(_m("AveragePool",e.inputs[0],!1,r))},gm={autoPad:"",ceilMode:0,countIncludePad:!1,kernelShape:[],strides:[],pads:[],storageOrder:0,dilations:[]},cv=e=>{let r=e.format;return{format:r,...gm,cacheKey:r}},uv=(e,r)=>{od(e.inputs),e.compute(_m("GlobalAveragePool",e.inputs[0],!0,r))},ym=(e,r,t,n)=>{let[o,i]=dm(r,n,t),u=`
      value = max(x_val, value);
    `,d="",h=Re("x",r.dataType,r.dims.length),f=["rank"],[y,m,g,v,b]=pm(i,o);return y.push(...bt(r.dims,i)),{name:e,shaderCache:{hint:`${n.cacheKey};${g};${v};${b}`,inputDependencies:f},getRunData:()=>({outputs:[{dims:i,dataType:r.dataType}],dispatchGroup:{x:Math.ceil(Fe.size(i)/64)},programUniforms:y}),getShaderSource:k=>hm(k,h,r.dims.length,i.length,o,u,d,r.dataType===10?-65504:-1e5,m,g,v,b)}},dv=(e,r)=>{od(e.inputs),e.compute(ym("MaxPool",e.inputs[0],!1,r))},pv=e=>{let r=e.storage_order,t=e.dilations,n=mm(e);if(r!==0)throw new Error("column major storage order is not yet supported for MaxPool");if(n.ceilMode!==0)throw new Error("using ceil() in shape computation is not yet supported for MaxPool");let o={storageOrder:r,dilations:t,...n,cacheKey:""};return{...o,cacheKey:mw(o)}},hv=e=>{let r=e.format;return{format:r,...gm,cacheKey:r}},fv=(e,r)=>{od(e.inputs),e.compute(ym("GlobalMaxPool",e.inputs[0],!0,r))}}),_w,gw,mv,_v,BT=Je(()=>{St(),Ft(),vr(),Lt(),_w=(e,r)=>{if(e.length<2||e.length>3)throw new Error("DequantizeLinear requires 2 or 3 inputs.");if(e.length===3&&e[1].dims===e[2].dims)throw new Error("x-scale and x-zero-point must have the same shape.");if(e.length===3&&e[0].dataType!==e[2].dataType)throw new Error("x and x-zero-point must have the same data type.");if(e[0].dataType===6&&e.length>2)throw new Error("In the case of dequantizing int32 there is no zero point.");if(e[1].dims.length!==0&&e[1].dims.length!==1&&e[1].dims.length!==e[0].dims.length)throw new Error("scale input must be a scalar, a 1D tensor, or have the same rank as the input tensor.");if(e.length>2){if(e[0].dataType!==e[2].dataType)throw new Error("x and x-zero-point must have the same data type.");if(e[1].dims.length!==e[2].dims.length)throw new Error("scale and zero-point inputs must have the same rank.");if(!e[1].dims.map((t,n)=>t===e[2].dims[n]).reduce((t,n)=>t&&n,!0))throw new Error("scale and zero-point inputs must have the same shape.")}if(r.blockSize>0){if(e[1].dims.length===0||e[1].dims.length===1&&e[1].dims[0]===1)throw new Error("blockSize must be set only for block quantization.");if(!e[1].dims.map((o,i)=>i===r.axis||o===e[0].dims[i]).reduce((o,i)=>o&&i,!0))throw new Error("For block qunatization, scale input shape to match the input shape except for the axis");if(e[1].dims.length!==e[0].dims.length)throw new Error("For block qunatization the scale input rank must be the same as the x rank.");let t=e[0].dims[r.axis],n=e[1].dims[r.axis];if(r.blockSize<Math.ceil(t/n)||r.blockSize>Math.ceil(t/(n-1)-1))throw new Error("blockSize must be with in the range [ceil(dI / Si), ceil(dI / (Si - 1) - 1)].")}},gw=(e,r)=>{let t=Fe.normalizeAxis(r.axis,e[0].dims.length),n=e[0].dataType,o=n===3,i=e[0].dims,u=e[1].dataType,d=Fe.size(i),h=n===3||n===2,f=h?[Math.ceil(Fe.size(e[0].dims)/4)]:e[0].dims,y=e[1].dims,m=e.length>2?e[2]:void 0,g=m?h?[Math.ceil(Fe.size(m.dims)/4)]:m.dims:void 0,v=y.length===0||y.length===1&&y[0]===1,b=v===!1&&y.length===1,k=yr(d),z=v&&(!h||k===4),E=z?k:1,T=z&&!h?k:1,D=Re("input",h?12:n,f.length,T),I=Re("scale",u,y.length),$=m?Re("zero_point",h?12:n,g.length):void 0,P=mt("output",u,i.length,E),S=[D,I];$&&S.push($);let O=[f,y];m&&O.push(g);let R=[{type:12,data:d/E},{type:12,data:t},{type:12,data:r.blockSize},...bt(...O,i)],G=ee=>{let ie=[{name:"output_size",type:"u32"},{name:"axis",type:"u32"},{name:"block_size",type:"u32"}];return`
      ${ee.registerUniforms(ie).declareVariables(...S,P)}
      ${ee.mainStart()}
          ${ee.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.output_size")}
          let output_indices = ${P.offsetToIndices("global_idx")};

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

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

          // Set zero-point input
          ${$?v?h?`
                let zero_point_input = ${$.getByOffset("0")};
                let zero_point_vec =  ${o?"unpack4xI8(zero_point_input)":"unpack4xU8(zero_point_input)"};
                let zero_point_value= zero_point_vec[0]`:`let zero_point_value = ${$.getByOffset("0")}`:b?h?`
                let zero_point_index = ${P.indicesGet("output_indices","uniforms.axis")};
                let zero_point_input = ${$.getByOffset("zero_point_index / 4")};
                let zero_point_vec =  ${o?"unpack4xI8(zero_point_input)":"unpack4xU8(zero_point_input)"};
                let zero_point_value = zero_point_vec[zero_point_index % 4]`:`
                let zero_point_index = ${P.indicesGet("output_indices","uniforms.axis")};
                let zero_point_value = ${$.getByOffset("zero_point_index")};`:h?`
                let zero_point_offset = ${I.indicesToOffset("scale_indices")};
                let zero_point_input = ${$.getByOffset("zero_point_offset / 4")};
                let zero_point_vec = ${o?"unpack4xI8(zero_point_input)":"unpack4xU8(zero_point_input)"};
                let zero_point_value = zero_point_vec[zero_point_offset % 4];`:`let zero_point_value = ${$.getByIndices("scale_indices")};`:`let zero_point_value = ${h?o?"i32":"u32":D.type.value}(0);`};
      // Compute and write output
      ${P.setByOffset("global_idx",`${P.type.value}(x_value - zero_point_value) * scale_value`)};
      }`};return{name:"DequantizeLinear",shaderCache:{hint:r.cacheKey,inputDependencies:$?["rank","rank","rank"]:["rank","rank"]},getShaderSource:G,getRunData:()=>({outputs:[{dims:i,dataType:u}],dispatchGroup:{x:Math.ceil(d/E/64),y:1,z:1},programUniforms:R})}},mv=(e,r)=>{_w(e.inputs,r),e.compute(gw(e.inputs,r))},_v=e=>Zt({axis:e.axis,blockSize:e.blockSize})}),yw,ww,gv,RT=Je(()=>{Ii(),St(),Lt(),yw=(e,r,t)=>{let n=e===r,o=e<r&&t<0,i=e>r&&t>0;if(n||o||i)throw new Error("Range these inputs' contents are invalid.")},ww=(e,r,t,n)=>{let o=Math.abs(Math.ceil((r-e)/t)),i=[o],u=o,d=[{type:12,data:u},{type:n,data:e},{type:n,data:t},...bt(i)],h=f=>{let y=mt("output",n,i.length),m=y.type.value,g=[{name:"outputSize",type:"u32"},{name:"start",type:m},{name:"delta",type:m}];return`
        ${f.registerUniforms(g).declareVariables(y)}
        ${f.mainStart()}
        ${f.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.outputSize")}
        output[global_idx] = uniforms.start + ${m}(global_idx) * uniforms.delta;
      }`};return{name:"Range",shaderCache:{hint:`${n}`},getShaderSource:h,getRunData:()=>({outputs:[{dims:i,dataType:n}],dispatchGroup:{x:Math.ceil(u/64)},programUniforms:d})}},gv=e=>{let r=0,t=0,n=0;e.inputs[0].dataType===6?(r=e.inputs[0].getInt32Array()[0],t=e.inputs[1].getInt32Array()[0],n=e.inputs[2].getInt32Array()[0]):e.inputs[0].dataType===1&&(r=e.inputs[0].getFloat32Array()[0],t=e.inputs[1].getFloat32Array()[0],n=e.inputs[2].getFloat32Array()[0]),dr.webgpu.validateInputContent&&yw(r,t,n),e.compute(ww(r,t,n,e.inputs[0].dataType),{inputs:[]})}}),Mw,bw,yv,wv,NT=Je(()=>{St(),Ft(),vr(),Lt(),Mw=(e,r,t,n)=>{if(e!=="none"&&n!=="i32"&&n!=="u32"&&n!=="f32")throw new Error(`Input ${n} is not supported with reduction ${e}.`);let o=`{
                var oldValue = 0;
                loop {
                  let newValueF32 =`,i=`;
                  let newValue = bitcast<i32>(newValueF32);
                  let res = atomicCompareExchangeWeak(&${r}, oldValue, newValue);
                  if res.exchanged {
                    break;
                  }
                  oldValue = res.old_value;
                }
              }`;switch(e){case"none":return`${r}=${t};`;case"add":return n==="i32"||n==="u32"?`atomicAdd(&${r}, bitcast<${n}>(${t}));`:`
              ${o}bitcast<${n}>(oldValue) + (${t})${i}`;case"max":return n==="i32"||n==="u32"?`atomicMax(&${r}, bitcast<${n}>(${t}));`:`
                ${o}max(bitcast<f32>(oldValue), (${t}))${i}`;case"min":return n==="i32"||n==="u32"?`atomicMin(&${r}, bitcast<${n}>(${t}));`:`${o}min(bitcast<${n}>(oldValue), (${t}))${i}`;case"mul":return`${o}(bitcast<${n}>(oldValue) * (${t}))${i}`;default:throw new Error(`Reduction ${e} is not supported.`)}},bw=(e,r)=>{let t=e[0].dims,n=e[1].dims,o=t,i=1,u=Math.ceil(Fe.sizeToDimension(n,n.length-1)/i),d=n[n.length-1],h=Fe.sizeFromDimension(t,d),f=[{type:12,data:u},{type:12,data:d},{type:12,data:h},...bt(e[1].dims,e[2].dims,o)],y=m=>{let g=Re("indices",e[1].dataType,e[1].dims.length),v=Re("updates",e[2].dataType,e[2].dims.length,i),b=r.reduction!=="none"&&r.reduction!==""?H0("output",e[0].dataType,o.length):mt("output",e[0].dataType,o.length,i);return`
      ${m.registerUniform("output_size","u32").registerUniform("last_index_dimension","u32").registerUniform("num_updates_elements","u32").declareVariables(g,v,b)}
      ${m.mainStart()}
        ${m.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.output_size")}
  var data_offset = 0u;
  let indices_start = uniforms.last_index_dimension * global_idx;
  let indices_end = indices_start + uniforms.last_index_dimension;
  for (var i = indices_start; i < indices_end; i++) {
    var index = i32(indices[i].x);
    ${e[0].dims.length===1?`
    let element_count_dim = uniforms.output_strides;
    let dim_value = uniforms.output_shape;`:`
    let element_count_dim = uniforms.output_strides[i - indices_start];
    let dim_value = uniforms.output_shape[i - indices_start];`}
    if (index >= 0) {
      if (index >= i32(dim_value)) {
        index = i32(dim_value - 1);
      }
    } else {
      if (index < -i32(dim_value)) {
        index = 0;
      } else {
        index += i32(dim_value);
      }
    }
    data_offset += u32((u32(index) * element_count_dim));
  }

  for (var i = 0u; i < uniforms.num_updates_elements; i++) {
    let value = updates[uniforms.num_updates_elements * global_idx + i];
    ${Mw(r.reduction,"output[data_offset + i]","value",b.type.value)}
  }

      }`};return{name:"ScatterND",shaderCache:{hint:`${r.cacheKey}_${r.reduction}`,inputDependencies:["rank","rank"]},getRunData:()=>({outputs:[{dims:o,dataType:e[0].dataType}],dispatchGroup:{x:Math.ceil(u/64)},programUniforms:f}),getShaderSource:y}},yv=e=>Zt({reduction:e.reduction}),wv=(e,r)=>{e.compute(bw(e.inputs,r),{inputs:[e.inputs[1],e.inputs[2]],outputs:[]})}}),vw,xw,Tw,wm,Ew,Pw,Cw,Sw,$w,kw,Iw,Aw,Mm,Ow,Fw,Dw,Lw,jw,Mv,bv,VT=Je(()=>{St(),Ft(),vr(),Lt(),vw=(e,r)=>{if(e.every(t=>t>0||(()=>{throw new Error("Resize requires scales input values to be positive")})),e.length>0){if(r.mode==="linear"){if(!(e.length===2||e.length===3||e.length===4&&e[0]===1&&e[1]===1||e.length===4&&e[0]===1&&e[3]===1||e.length===5&&e[0]===1&&e[1]===1))throw new Error(`For linear mode, Resize requires scales to be 2D, 3D, 4D with either two outermost or one innermost and
            one outermost scale values equal to 1, or 5D with two outermost scale values equal to 1`)}else if(r.mode==="cubic"&&!(e.length===2||e.length===4&&e[0]===1&&e[1]===1||e.length===4&&e[0]===1&&e[3]===1))throw new Error("Resize requires scales input size to be 2 or 4 for cubic mode")}},xw=(e,r,t)=>{r.every(o=>o>=0&&o<t||(()=>{throw new Error("Resize requires axes input values to be positive and less than rank")}));let n=new Array(t).fill(1);return r.forEach((o,i)=>n[o]=e[i]),n},Tw=(e,r,t,n,o,i)=>{let[u,d,h]=t>10?[1,2,3]:[-1,e.length>1?1:-1,-1],f=e[0].dims.length;if(u>0&&e.length>u&&e[u].dims.length>0)e[u].getFloat32Array().forEach(y=>i.push(y));else if(r.coordinateTransformMode==="tf_crop_and_resize")throw new Error("Resize requires RoI input to be specified when coordinateTransformMode is tfCropAndResize");if(d>0&&e.length>d&&e[d].dims.length===1&&e[d].dims[0]>0){if(e[d].getFloat32Array().forEach(y=>n.push(y)),n.length!==0&&n.length!==f&&t>=18&&n.length!==r.axes.length)throw new Error("Resize requires scales input size to be same as input rank or axes size for opset 18 and up");vw(n,r),r.axes.length>0&&xw(n,r.axes,f).forEach((y,m)=>n[m]=y)}if(h>0&&e.length>h&&e[h].dims.length===1&&e[h].dims[0]>0&&(e[h].getBigInt64Array().forEach(y=>o.push(Number(y))),o.length!==0&&o.length!==f&&t>=18&&o.length!==r.axes.length))throw new Error("Resize requires sizes input size to be same as input rank or axes size for opset 18 and up");if(r.axes.length>0){if(n.length!==0&&n.length!==r.axes.length)throw new Error('Resize requires "scales" input size to be of axes rank when axes attributes is specified');if(o.length!==0&&o.length!==r.axes.length)throw new Error('Resize requires "sizes" input size to be of rank axes rank when axes attributes is specified')}if(typeof n<"u"&&typeof o<"u"&&n.length>0&&o.length>f)throw new Error("Resize requires only of scales or sizes to be specified")},wm=(e,r,t,n)=>`
  // The whole part and the fractional part are calculated separately due to inaccuracy of floating
  // point division. As an example, f32(21) / f32(7) may evaluate to 2.99... instead of 3, causing an
  // offset-by-one error later in floor().
  let big = (${e}) * (${r});
  let whole = ${n}(big / (${t}));
  let fract = ${n}(big % (${t})) / ${n}(${t});
  return whole + fract;
`,Ew=(e,r)=>`fn getOriginalCoordinateFromResizedCoordinate(xResized: u32, xScale: f32, lengthResized: u32,
     lengthOriginal: u32, roiStart: f32, roiEnd: f32) -> ${r} { `+(()=>{switch(e){case"asymmetric":return`
          if (xScale < 1.0 || floor(xScale) != xScale) {
            return ${r}(xResized) / ${r}(xScale);
          } else {
            ${wm("xResized","lengthOriginal","lengthResized",r)}
          }
        `;case"pytorch_half_pixel":return`if (lengthResized > 1) {
                    return (${r}(xResized) + 0.5) / ${r}(xScale) - 0.5;
                  } else {
                    return 0.0;
                  }`;case"tf_half_pixel_for_nn":return`return (${r}(xResized) + 0.5) / ${r}(xScale);`;case"align_corners":return`if (lengthResized == 1) {
                    return 0.0;
                  } else {
                    ${wm("xResized","lengthOriginal - 1","lengthResized - 1",r)}
                  }`;case"tf_crop_and_resize":return`if (lengthResized > 1) {
                    return ${r}(roiStart) * ${r}(lengthOriginal - 1) +
                        (${r}(xResized) * ${r}(roiEnd - roiStart) * ${r}(lengthOriginal - 1)) /
                        ${r}(lengthResized - 1);
                  } else {
                    return 0.5 * ${r}(roiStart + roiEnd) * ${r}(lengthOriginal - 1);
                  }`;case"half_pixel_symmetric":return`const outputWidth = ${r}xScale * ${r}(lengthResized);
                  const adjustment = ${r}(lengthResized) / outputWidth;
                  const center = ${r}(lengthOriginal) / 2;
                  const offset = center * (1 - adjustment);
                  return offset + ((${r}(xResized) + 0.5) / ${r}(xScale)) - 0.5;`;case"half_pixel":return`return ((${r}(xResized) + 0.5) / ${r}(xScale)) - 0.5;`;default:throw new Error(`Coordinate transform mode ${e} is not supported`)}})()+"}",Pw=(e,r,t)=>`fn getNearestPixelFromOriginal(xOriginal: ${t}, isDownSample: bool) -> ${t} {`+(()=>{switch(e){case"round_prefer_ceil":return"if (fract(xOriginal) == 0.5) {             return ceil(xOriginal);           } else {             return round(xOriginal);           }";case"floor":return"return floor(xOriginal);";case"ceil":return"return ceil(xOriginal);";case"round_prefer_floor":return"if (fract(xOriginal) == 0.5) {                     return floor(xOriginal);                   } else {                     return round(xOriginal);                   }";case"simple":default:if(r<11)return"if (isDownSample)                     {                       return ceil(xOriginal);                     } else {                       return xOriginal;                     }";throw new Error(`Nearest mode ${e} is not supported`)}})()+"}",Cw=(e,r,t)=>{let n=new Array(t).fill(0).concat(new Array(t).fill(1)),o=e.length===0?n:e.slice();return r.length>0?(r.forEach((i,u)=>{n[i]=o[u],n[u+t]=o[r.length+u]}),n):o},Sw=(e,r,t,n)=>{let o=[];if(t.length>0)if(n.length>0){if(e.forEach(i=>o.push(i)),Math.max(...n)>e.length)throw new Error("axes is out of bound");n.forEach((i,u)=>o[i]=t[u])}else t.forEach(i=>o.push(i));else{if(r.length===0)throw new Error("Resize requires either scales or sizes.");o=e.map((i,u)=>Math.round(i*r[u]))}return o},$w=(e,r,t)=>{let n=(()=>{switch(t.keepAspectRatioPolicy){case"not_larger":return t.axes.length>0?Math.min(...t.axes.map(i=>r[i]),Number.MAX_VALUE):Math.min(...r,Number.MAX_VALUE);case"not_smaller":return t.axes.length>0?Math.max(...t.axes.map(i=>r[i]),Number.MIN_VALUE):Math.max(...r,Number.MIN_VALUE);default:throw new Error(`Keep aspect ratio policy ${t.keepAspectRatioPolicy} is not supported`)}})();r.fill(1,0,r.length);let o=e.slice();return t.axes.length>0?(t.axes.forEach(i=>r[i]=n),t.axes.forEach(i=>o[i]=Math.round(e[i]*r[i]))):(r.fill(n,0,r.length),o.forEach((i,u)=>o[u]=Math.round(i*r[u]))),o},kw=(e,r,t,n,o)=>`
    fn calculateOriginalIndicesFromOutputIndices(output_indices: ${e.type.indices}) -> array<${e.type.value}, ${t.length}> {
      var original_indices: array<${e.type.value}, ${t.length}>;
      for (var i:u32 = 0; i < ${t.length}; i++) {
        var output_index = ${e.indicesGet("output_indices","i")};
        var scale = ${wt("uniforms.scales","i",n)};
        var roi_low = ${wt("uniforms.roi","i",o)};
        var roi_hi = ${wt("uniforms.roi",`i + ${r.length}`,o)};
        if (scale == 1.0) {
          original_indices[i] = ${e.type.value}(output_index);
        } else {
          var input_shape_i = ${wt("uniforms.input_shape","i",r.length)};
          var output_shape_i = ${wt("uniforms.output_shape","i",t.length)};
          original_indices[i] = getOriginalCoordinateFromResizedCoordinate(output_index, scale, output_shape_i,
                                                                           input_shape_i, roi_low, roi_hi);
        }
      }
      return original_indices;
    }`,Iw=(e,r,t,n,o,i,u)=>`
    fn calculateInputIndicesFromOutputIndices(output_indices: ${r.type.indices}) -> ${e.type.indices} {
      var input_indices: ${e.type.indices};
      for (var i:u32 = 0; i < ${n.length}; i++) {
        var output_index = ${r.indicesGet("output_indices","i")};
        var input_index: u32;
        var scale = ${wt("uniforms.scales","i",o)};
        if (scale == 1.0) {
          input_index = output_index;
        } else {
          var roi_low = ${wt("uniforms.roi","i",i)};
          var roi_hi = ${wt("uniforms.roi",`i + ${t.length}`,i)};
          var input_shape_i = ${wt("uniforms.input_shape","i",t.length)};
          var output_shape_i = ${wt("uniforms.output_shape","i",n.length)};
          var original_idx = getOriginalCoordinateFromResizedCoordinate(output_index, scale, output_shape_i,
                                                                        input_shape_i, roi_low, roi_hi);
          if (!${u} || (original_idx >= 0 && original_idx < ${r.type.value}(input_shape_i))) {
            if (original_idx < 0) {
              input_index = 0;
            } else if (original_idx > ${r.type.value}(input_shape_i - 1)) {
              input_index = input_shape_i - 1;
            } else {
              input_index = u32(getNearestPixelFromOriginal(original_idx, scale < 1));
            }
          } else {
            input_index = u32(original_idx);
          }
        }
        ${e.indicesSet("input_indices","i","input_index")}
      }
      return input_indices;
    }`,Aw=(e,r)=>`
    fn checkInputIndices(input_indices: ${e.type.indices}) -> bool {
      for (var i:u32 = 0; i < ${r.length}; i++) {
        var input_index = ${e.indicesGet("input_indices","i")};
        if (input_index < 0 || input_index >= ${wt("uniforms.input_shape","i",r.length)}) {
          return false;
        }
      }
      return true;
    }`,Mm=(e,r,t,n)=>e.rank>n?`
    ${e.indicesSet("input_indices",r,"channel")};
    ${e.indicesSet("input_indices",t,"batch")};
`:"",Ow=(e,r,t,n,o)=>{let[i,u,d,h]=t.length===2?[-1,0,1,-1]:[0,2,3,1],f=e.type.value;return`
    fn getInputValue(batch: u32, channel: u32, row: u32, col: u32) -> ${f} {
      var input_indices: ${e.type.indices};
      ${e.indicesSet("input_indices",u,`max(0, min(row, ${t[u]} - 1))`)};
      ${e.indicesSet("input_indices",d,`max(0, min(col, ${t[d]} - 1))`)};
      ${Mm(e,h,i,2)}
      return ${e.getByIndices("input_indices")};
    }

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

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

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

  fn bicubicInterpolation(output_indices: ${r.type.indices}) -> ${v} {
    var input_indices: ${e.type.indices} = output_indices;
    return colCubicInterpolation(input_indices, output_indices);
  }
    `},Dw=(e,r,t,n,o)=>{let[i,u,d,h,f]=t.length===3?[-1,0,1,2,-1]:[0,2,3,4,1],y=e.type.value;return`
    fn getInputValue(batch: u32, channel: u32, depth:u32, height: u32, width: u32) -> ${y} {
      var input_indices: ${e.type.indices};
      ${e.indicesSet("input_indices",u,`max(0, min(depth, ${t[u]} - 1))`)};
      ${e.indicesSet("input_indices",d,`max(0, min(height, ${t[d]} - 1))`)};
      ${e.indicesSet("input_indices",h,`max(0, min(width, ${t[h]} - 1))`)};
      ${Mm(e,f,i,3)}
      return ${e.getByIndices("input_indices")};
    }

    fn trilinearInterpolation(output_indices: ${r.type.indices}) -> ${y} {
      var originalIndices = calculateOriginalIndicesFromOutputIndices(output_indices);
      var depth:${y} = originalIndices[${u}];
      var height:${y} = originalIndices[${d}];
      var width:${y} = originalIndices[${h}];
      ${n?`if (depth < 0 || depth > (${t[u]} - 1) || height < 0 || height > (${t[d]} - 1) || width < 0 || (width > ${t[h]} - 1)) {
      return ${o};
        }`:""};

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

      var x111: ${y} = getInputValue(batch, channel, depth1, height1, width1);
      var x112: ${y} = getInputValue(batch, channel, depth1, height1, width2);
      var x121: ${y} = getInputValue(batch, channel, depth1, height2, width1);
      var x122: ${y} = getInputValue(batch, channel, depth1, height2, width2);
      var x211: ${y} = getInputValue(batch, channel, depth2, height1, width1);
      var x212: ${y} = getInputValue(batch, channel, depth2, height1, width2);
      var x221: ${y} = getInputValue(batch, channel, depth2, height2, width1);
      var x222: ${y} = getInputValue(batch, channel, depth2, height2, width2);
      var dx1: ${y} = abs(depth - ${y}(depth1));
      var dx2: ${y} = abs(${y}(depth2) - depth);
      var dy1: ${y} = abs(height - ${y}(height1));
      var dy2: ${y} = abs(${y}(height2) - height);
      var dz1: ${y} = abs(width - ${y}(width1));
      var dz2: ${y} = abs(${y}(width2) - width);
      if (depth1 == depth2) {
        dx1 = 0.5;
        dx2 = 0.5;
      }
      if (height1 == height2) {
        dy1 = 0.5;
        dy2 = 0.5;
      }
      if (width1 == width2) {
        dz1 = 0.5;
        dz2 = 0.5;
      }
      return (x111 * dx2 * dy2 * dz2 + x112 * dx2 * dy2 * dz1 + x121 * dx2 * dy1 *dz2 + x122 * dx2 * dy1 * dz1 +
              x211 * dx1 * dy2 * dz2 + x212 * dx1 * dy2 * dz1 + x221 * dx1 * dy1 *dz2 + x222 * dx1 * dy1 * dz1);
    }`},Lw=(e,r,t,n,o,i)=>{let u=e.dims,d=Cw(i,r.axes,u.length),h=Sw(u,n,o,r.axes),f=n.slice();n.length===0&&(f=u.map((T,D)=>T===0?1:h[D]/T),r.keepAspectRatioPolicy!=="stretch"&&(h=$w(u,f,r)));let y=mt("output",e.dataType,h.length),m=Re("input",e.dataType,u.length),g=Fe.size(h),v=u.length===h.length&&u.every((T,D)=>T===h[D]),b=r.coordinateTransformMode==="tf_crop_and_resize",k=r.extrapolationValue,z=m.type.value,E=T=>`
      ${v?"":`
      ${Ew(r.coordinateTransformMode,z)};
      ${(()=>{switch(r.mode){case"nearest":return`
              ${Aw(m,u)};
              ${Pw(r.nearestMode,t,z)};
              ${Iw(m,y,u,h,f.length,d.length,b)};
              `;case"linear":return`
              ${kw(y,u,h,f.length,d.length)};
              ${(()=>{if(u.length===2||u.length===4)return`${Ow(m,y,u,b,k)}`;if(u.length===3||u.length===5)return`${Dw(m,y,u,b,k)}`;throw Error("Linear mode only supports input dims 2, 3, 4 and 5 are supported in linear mode.")})()};
            `;case"cubic":return`
            ${(()=>{if(u.length===2||u.length===4)return`${Fw(m,y,u,h,f,d,r.cubicCoeffA,b,r.extrapolationValue,r.excludeOutside)}`;throw Error("Cubic mode only supports input dims 2 and 4 are supported in linear mode.")})()};
            `;default:throw Error("Invalid resize mode")}})()};
      `}
      ${T.registerUniform("output_size","u32").registerUniform("scales","f32",f.length).registerUniform("roi","f32",d.length).declareVariables(m,y)}
      ${T.mainStart()}
        ${T.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.output_size")}
        ${v?"output[global_idx] = input[global_idx];":`
        let output_indices = ${y.offsetToIndices("global_idx")};
        var input_indices: ${m.type.indices};
        ${(()=>{switch(r.mode){case"nearest":return`input_indices = calculateInputIndicesFromOutputIndices(output_indices);
                if (checkInputIndices(input_indices)) {
                  output[global_idx] = ${m.getByIndices("input_indices")};
                } else {
                  output[global_idx] = ${r.extrapolationValue};
                }`;case"linear":return`output[global_idx] = ${u.length===2||u.length===4?"bilinearInterpolation":"trilinearInterpolation"}(output_indices);`;case"cubic":return"output[global_idx] = bicubicInterpolation(output_indices);";default:throw Error(`Unsupported resize mode: ${r.mode}`)}})()};
`}
      }`;return{name:"Resize",shaderCache:{hint:`${r.cacheKey}|${t}|${f.length>0?r.mode==="cubic"?f:f.length:""}|${o.length>0?o:""}|${d.length>0?d:""}|${v}|${r.mode==="nearest"?u.length:u}`,inputDependencies:["rank"]},getShaderSource:E,getRunData:()=>({outputs:[{dims:h,dataType:e.dataType}],dispatchGroup:{x:Math.ceil(g/64)},programUniforms:[{type:12,data:g},{type:1,data:f},{type:1,data:d},...bt(u,h)]})}},jw=e=>{let r=e.customDataBuffer;return new Uint32Array(r,r.byteOffset,1)[0]},Mv=(e,r)=>{let t=[],n=[],o=[],i=jw(e);if(r.antialias!==0)throw Error("Only default value (0) for Antialias attribute is supported");Tw(e.inputs,r,i,t,n,o),e.compute(Lw(e.inputs[0],r,i,t,n,o),{inputs:[0]})},bv=e=>{let r=e.antialias,t=e.axes,n=e.coordinateTransformMode,o=e.cubicCoeffA,i=e.excludeOutside!==0,u=e.extrapolationValue,d=e.keepAspectRatioPolicy,h=e.mode,f=e.nearestMode===""?"simple":e.nearestMode;return Zt({antialias:r,axes:t,coordinateTransformMode:n,cubicCoeffA:o,excludeOutside:i,extrapolationValue:u,keepAspectRatioPolicy:d,mode:h,nearestMode:f})}}),zw,Bw,vv,UT=Je(()=>{St(),Ft(),Lt(),zw=e=>{if(!e||e.length<3)throw new Error("layerNorm requires at least 3 inputs.");let r=e[0],t=e[1],n=e[2];if(r.dataType!==t.dataType||r.dataType!==n.dataType)throw new Error("All inputs must have the same data type");if(r.dims.length!==3&&r.dims.length!==2)throw new Error("Input must be 2D or 3D");if(t.dims.length!==3&&t.dims.length!==2)throw new Error("Skip must be 2D or 3D");let o=r.dims[r.dims.length-1],i=r.dims[r.dims.length-2];if(t.dims[t.dims.length-1]!==o)throw new Error("Skip must have the same hidden size as input");if(t.dims[t.dims.length-2]!==i)throw new Error("Skip must have the same sequence length as input");if(n.dims.length!==1)throw new Error("Gamma must be 1D");if(n.dims[n.dims.length-1]!==o)throw new Error("Gamma must have the same hidden size as input");if(e.length>3){let u=e[3];if(u.dims.length!==1)throw new Error("Beta must be 1D");if(u.dims[u.dims.length-1]!==o)throw new Error("Beta must have the same hidden size as input")}if(e.length>4){let u=e[4];if(u.dims.length!==1)throw new Error("Bias must be 1D");if(u.dims[u.dims.length-1]!==o)throw new Error("Bias must have the same hidden size as input")}},Bw=(e,r,t,n)=>{let o=r.simplified,i=e[0].dims,u=Fe.size(i),d=i,h=u,f=i.slice(-1)[0],y=n?i.slice(0,-1).concat(1):[],m=!o&&e.length>3,g=e.length>4,v=n&&t>1,b=n&&t>2,k=t>3,z=64,E=yr(f),T=[{type:12,data:h},{type:12,data:E},{type:12,data:f},{type:1,data:r.epsilon}],D=$=>{let P=[{name:"output_size",type:"u32"},{name:"components",type:"u32"},{name:"hidden_size",type:"u32"},{name:"epsilon",type:"f32"}],S=[Re("x",e[0].dataType,e[0].dims,E),Re("skip",e[1].dataType,e[1].dims,E),Re("gamma",e[2].dataType,e[2].dims,E)];m&&S.push(Re("beta",e[3].dataType,e[3].dims,E)),g&&S.push(Re("bias",e[4].dataType,e[4].dims,E)),S.push(mt("output",e[0].dataType,d,E)),v&&S.push(mt("mean_output",1,y)),b&&S.push(mt("inv_std_output",1,y)),k&&S.push(mt("input_skip_bias_sum",e[0].dataType,d,E));let O=jr(e[0].dataType),R=jr(1,E);return`

      ${$.registerUniforms(P).declareVariables(...S)}
      var<workgroup> sum_shared : array<${R}, ${z}>;
      var<workgroup> sum_squared_shared : array<${R}, ${z}>;

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

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

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

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

        for (var i: u32 = 0; i < stride; i++) {
          output[offset + i] = (output[offset + i] ${o?"":`- ${O}(mean)`}) *
            ${O}(inv_std_dev) * gamma[offset1d + i]
            ${m?"+ beta[offset1d + i]":""};
        }
      }`},I=[{dims:d,dataType:e[0].dataType}];return t>1&&I.push({dims:y,dataType:1}),t>2&&I.push({dims:y,dataType:1}),t>3&&I.push({dims:i,dataType:e[0].dataType}),{name:"SkipLayerNormalization",shaderCache:{hint:`${E};${v};${b};${k}`,inputDependencies:e.map(($,P)=>"type")},getShaderSource:D,getRunData:()=>({outputs:I,dispatchGroup:{x:Math.ceil(h/f)},programUniforms:T})}},vv=(e,r)=>{zw(e.inputs);let t=[0];e.outputCount>1&&t.push(-3),e.outputCount>2&&t.push(-3),e.outputCount>3&&t.push(3),e.compute(Bw(e.inputs,r,e.outputCount,!1),{outputs:t})}}),Rw,ld,Nw,bm,Vw,Uw,xv,Tv,WT=Je(()=>{St(),Ft(),vr(),Lt(),Rw=(e,r)=>{if(!e||e.length<1)throw new Error("too few inputs");if(r.axes.length!==0){if(r.axes.length!==r.starts.length||r.axes.length!==r.ends.length)throw new Error("axes, starts and ends must have the same length")}else if(r.starts.length!==r.ends.length)throw new Error("starts and ends must have the same length");e.slice(1).forEach((t,n)=>{if(e[n+1].dataType!==6&&e[n+1].dataType!==7)throw new Error(`Input ${n} must be an array of int32 or int64`)})},ld=(e,r)=>{let t=[];if(e.length>r)if(e[r].dataType===7)e[r].getBigInt64Array().forEach(n=>t.push(Number(n)));else if(e[r].dataType===6)e[r].getInt32Array().forEach(n=>t.push(Number(n)));else throw new Error(`Input ${r} must be an array of int32 or int64`);return t},Nw=(e,r)=>{if(e.length>1){let t=ld(e,1),n=ld(e,2),o=ld(e,3);return o.length===0&&(o=[...Array(e[0].dims.length).keys()]),Zt({starts:t,ends:n,axes:o})}else return r},bm=(e,r,t,n,o)=>{let i=e;return e<0&&(i+=t[n[r]]),o[r]<0?Math.max(0,Math.min(i,t[n[r]]-1)):Math.max(0,Math.min(i,t[n[r]]))},Vw=(e,r,t)=>`fn calculateInputIndices(output_indices: ${r.type.indices}) -> ${e.type.indices} {
          var input_indices: ${e.type.indices};
          var carry = 0u;
          for (var i = ${t.length}; i >= 0; i--) {
            let input_shape_i = ${wt("uniforms.input_shape","i",t.length)};
            let steps_i = ${wt("uniforms.steps","i",t.length)};
            let signs_i = ${wt("uniforms.signs","i",t.length)};
            let starts_i = ${wt("uniforms.starts","i",t.length)};
            var output_index = ${r.indicesGet("output_indices","i")};
            var input_index = output_index * steps_i + starts_i + carry;
            carry = input_index / input_shape_i;
            input_index = input_index % input_shape_i;
            if (signs_i < 0) {
              input_index = input_shape_i - input_index - 1u + starts_i;
            }
            ${e.indicesSet("input_indices","i","input_index")};
          }
          return input_indices;
      }`,Uw=(e,r)=>{let t=e[0].dims,n=Fe.size(t),o=r.axes.length>0?Fe.normalizeAxes(r.axes,t.length):[...Array(t.length).keys()],i=ld(e,4);i.forEach(E=>E!==0||(()=>{throw new Error("step cannot be 0")})),i.length===0&&(i=Array(o.length).fill(1));let u=r.starts.map((E,T)=>bm(E,T,t,o,i)),d=r.ends.map((E,T)=>bm(E,T,t,o,i));if(o.length!==u.length||o.length!==d.length)throw new Error("start, ends and axes should have the same number of elements");if(o.length!==t.length)for(let E=0;E<t.length;++E)o.includes(E)||(u.splice(E,0,0),d.splice(E,0,t[E]),i.splice(E,0,1));let h=i.map(E=>Math.sign(E));i.forEach((E,T,D)=>{if(E<0){let I=(d[T]-u[T])/E,$=u[T],P=$+I*i[T];u[T]=P,d[T]=$,D[T]=-E}});let f=t.slice(0);o.forEach((E,T)=>{f[E]=Math.ceil((d[E]-u[E])/i[E])});let y={dims:f,dataType:e[0].dataType},m=mt("output",e[0].dataType,f.length),g=Re("input",e[0].dataType,e[0].dims.length),v=Fe.size(f),b=[{name:"outputSize",type:"u32"},{name:"starts",type:"u32",length:u.length},{name:"signs",type:"i32",length:h.length},{name:"steps",type:"u32",length:i.length}],k=[{type:12,data:v},{type:12,data:u},{type:6,data:h},{type:12,data:i},...bt(e[0].dims,f)],z=E=>`
      ${E.registerUniforms(b).declareVariables(g,m)}
        ${Vw(g,m,t)}
        ${E.mainStart()}
          ${E.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.outputSize")}
          let output_indices = ${m.offsetToIndices("global_idx")};
          let input_indices = calculateInputIndices(output_indices);
          ${m.setByOffset("global_idx",g.getByIndices("input_indices"))}
      }`;return{name:"Slice",shaderCache:{hint:`${h.length}_${u.length}_${i.length}`,inputDependencies:["rank"]},getShaderSource:z,getRunData:()=>({outputs:[y],dispatchGroup:{x:Math.ceil(n/64)},programUniforms:k})}},xv=(e,r)=>{Rw(e.inputs,r);let t=Nw(e.inputs,r);e.compute(Uw(e.inputs,t),{inputs:[0]})},Tv=e=>{let r=e.starts,t=e.ends,n=e.axes;return Zt({starts:r,ends:t,axes:n})}}),Ww,Gw,Ev,Pv,GT=Je(()=>{St(),Ft(),vr(),Ns(),Lt(),Ww=e=>{if(!e||e.length!==1)throw new Error("Softmax op requires 1 input.")},Gw=(e,r)=>{let t=e.inputs[0],n=t.dims,o=Fe.size(n),i=n.length,u=Fe.normalizeAxis(r.axis,i),d=u<n.length-1,h,f=[];d?(f=Array.from({length:i},(S,O)=>O),f[u]=i-1,f[i-1]=u,h=e.compute(xi(t,f),{inputs:[t],outputs:[-1]})[0]):h=t;let y=h.dims,m=y[i-1],g=o/m,v=yr(m),b=m/v,k=64;g===1&&(k=256);let z=(S,O)=>O===4?`max(max(${S}.x, ${S}.y), max(${S}.z, ${S}.w))`:O===2?`max(${S}.x, ${S}.y)`:O===3?`max(max(${S}.x, ${S}.y), ${S}.z)`:S,E=Re("x",h.dataType,h.dims,v),T=mt("result",h.dataType,h.dims,v),D=E.type.value,I=jr(h.dataType)==="f32"?`var threadMax = ${D}(-3.402823e+38f);`:`var threadMax = ${D}(-65504.0h);`,$=S=>`
      var<workgroup> rowMaxShared : ${D};
      var<workgroup> rowSumShared : ${D};
      var<workgroup> threadShared : array<${D}, ${k}>;

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

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

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

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

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

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

        // calculate final value for each element in the row
        for (var col = lindex; col < cols; col += wg) {
          let value = exp(getValue(row, col, row_stride) - rowMaxShared) / rowSumShared;
          setValue(row, col, row_stride, value);
        }
      }`,P=e.compute({name:"Softmax",shaderCache:{hint:`${v};${k}`,inputDependencies:["type"]},getRunData:()=>({outputs:[{dims:y,dataType:h.dataType}],dispatchGroup:{x:g},programUniforms:[{type:6,data:b}]}),getShaderSource:$},{inputs:[h],outputs:[d?-1:0]})[0];d&&e.compute(xi(P,f),{inputs:[P]})},Ev=(e,r)=>{Ww(e.inputs),Gw(e,r)},Pv=e=>Zt({axis:e.axis})}),vm,Kw,Hw,qw,Cv,KT=Je(()=>{St(),Ft(),Lt(),vm=e=>Array.from(e.getBigInt64Array(),Number),Kw=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(vm(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")},Hw=(e,r)=>{let t=[];for(let n=0;n<e.length;++n)t.push(e[n]*r[n]);return t},qw=(e,r)=>{let t=e[0].dims,n=r??vm(e[1]),o=Hw(t,n),i=Fe.size(o),u=e[0].dataType,d=Re("input",u,t.length),h=mt("output",u,o.length),f=y=>`
      const inputShape = ${d.indices(...t)};
      ${y.registerUniform("output_size","u32").declareVariables(d,h)}
      ${y.mainStart()}
      ${y.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.output_size")}
      let output_indices = ${h.offsetToIndices("global_idx")};
      var input_indices: ${d.type.indices};
      for (var i = 0; i < ${t.length}; i++) {
        let input_dim_i = ${d.indicesGet("uniforms.input_shape","i")};
        let input_dim_value = ${h.indicesGet("output_indices","i")}  % input_dim_i;

        ${d.indicesSet("input_indices","i","input_dim_value")}
      }
      ${h.setByOffset("global_idx",d.getByIndices("input_indices"))}
    }`;return{name:"Tile",shaderCache:{hint:`${n}`,inputDependencies:["rank"]},getRunData:()=>({outputs:[{dims:o,dataType:e[0].dataType}],dispatchGroup:{x:Math.ceil(i/64)},programUniforms:[{type:12,data:i},...bt(e[0].dims,o)]}),getShaderSource:f}},Cv=e=>{Kw(e.inputs),e.compute(qw(e.inputs),{inputs:[0]})}}),Qw,Xw,Sv,HT=Je(()=>{St(),Ft(),Lt(),Qw=(e,r,t,n,o)=>{let i=mt("output_data",o,t.length,4),u=Re("a_data",r[1].dataType,r[1].dims.length,4),d=Re("b_data",r[2].dataType,r[2].dims.length,4),h=Re("c_data",r[0].dataType,r[0].dims.length,4),f,y=(m,g,v)=>`select(${g}, ${m}, ${v})`;if(!n)f=i.setByOffset("global_idx",y(u.getByOffset("global_idx"),d.getByOffset("global_idx"),h.getByOffset("global_idx")));else{let m=(g,v,b="")=>{let k=`a_data[index_a${v}][component_a${v}]`,z=`b_data[index_b${v}][component_b${v}]`,E=`bool(c_data[index_c${v}] & (0xffu << (component_c${v} * 8)))`;return`
            let output_indices${v} = ${i.offsetToIndices(`global_idx * 4u + ${v}u`)};
            let offset_a${v} = ${u.broadcastedIndicesToOffset(`output_indices${v}`,i)};
            let offset_b${v} = ${d.broadcastedIndicesToOffset(`output_indices${v}`,i)};
            let offset_c${v} = ${h.broadcastedIndicesToOffset(`output_indices${v}`,i)};
            let index_a${v} = offset_a${v} / 4u;
            let index_b${v} = offset_b${v} / 4u;
            let index_c${v} = offset_c${v} / 4u;
            let component_a${v} = offset_a${v} % 4u;
            let component_b${v} = offset_b${v} % 4u;
            let component_c${v} = offset_c${v} % 4u;
            ${g}[${v}] = ${b}(${y(k,z,E)});
          `};o===9?f=`
            var data = vec4<u32>(0);
            ${m("data",0,"u32")}
            ${m("data",1,"u32")}
            ${m("data",2,"u32")}
            ${m("data",3,"u32")}
            output_data[global_idx] = dot(vec4<u32>(0x1, 0x100, 0x10000, 0x1000000), vec4<u32>(data));`:f=`
            ${m("output_data[global_idx]",0)}
            ${m("output_data[global_idx]",1)}
            ${m("output_data[global_idx]",2)}
            ${m("output_data[global_idx]",3)}
          `}return`
        ${e.registerUniform("vec_size","u32").declareVariables(h,u,d,i)}
        ${e.mainStart()}
        ${e.guardAgainstOutOfBoundsWorkgroupSizes("uniforms.vec_size")}
        ${f}
      }`},Xw=e=>{let r=e[1].dims,t=e[2].dims,n=e[0].dims,o=e[1].dataType,i=!(Fe.areEqual(r,t)&&Fe.areEqual(t,n)),u=r,d=Fe.size(r);if(i){let f=Ua.calcShape(Ua.calcShape(r,t,!1),n,!1);if(!f)throw new Error("Can't perform where op on the given tensors");u=f,d=Fe.size(u)}let h=Math.ceil(d/4);return{name:"Where",shaderCache:{inputDependencies:["rank","rank","rank"]},getShaderSource:f=>Qw(f,e,u,i,o),getRunData:()=>({outputs:[{dims:u,dataType:o}],dispatchGroup:{x:Math.ceil(d/64/4)},programUniforms:[{type:12,data:h},...bt(n,r,t,u)]})}},Sv=e=>{e.compute(Xw(e.inputs))}}),$v,qT=Je(()=>{lT(),l_(),cT(),uT(),dT(),pT(),hT(),yT(),MT(),bT(),vT(),xT(),TT(),ET(),PT(),CT(),ST(),$T(),kT(),IT(),AT(),OT(),FT(),DT(),LT(),qb(),jT(),zT(),BT(),RT(),NT(),o_(),VT(),Zb(),UT(),WT(),GT(),Jb(),KT(),Ns(),c_(),HT(),$v=new Map([["Abs",[vM]],["Acos",[xM]],["Acosh",[TM]],["Add",[sb]],["ArgMax",[yM,jm]],["ArgMin",[gM,jm]],["Asin",[EM]],["Asinh",[PM]],["Atan",[CM]],["Atanh",[SM]],["Attention",[wM]],["AveragePool",[lv,ov]],["BatchNormalization",[MM]],["BiasAdd",[bM]],["BiasSplitGelu",[ib]],["Cast",[kM,$M]],["Ceil",[AM]],["Clip",[IM]],["Concat",[fb,mb]],["Conv",[Um,Vm]],["ConvTranspose",[Eb,Tb]],["Cos",[OM]],["Cosh",[FM]],["CumSum",[Pb,Cb]],["DepthToSpace",[Sb,$b]],["DequantizeLinear",[mv,_v]],["Div",[nb]],["Einsum",[kb,Ib]],["Elu",[DM,hd]],["Equal",[ab]],["Erf",[LM]],["Exp",[jM]],["Expand",[Ab]],["FastGelu",[Ob]],["Floor",[zM]],["FusedConv",[Um,Vm]],["Gather",[Db,Fb]],["GatherElements",[Nb,Rb]],["GatherBlockQuantized",[zb,Bb]],["GatherND",[Lb,jb]],["Gelu",[BM]],["Gemm",[Ub,Vb]],["GlobalAveragePool",[uv,cv]],["GlobalMaxPool",[fv,hv]],["Greater",[ub]],["GreaterOrEqual",[pb]],["GridSample",[Wb,Gb]],["GroupQueryAttention",[ev]],["HardSigmoid",[HM,KM]],["InstanceNormalization",[tv]],["LayerNormalization",[rv]],["LeakyRelu",[RM,hd]],["Less",[db]],["LessOrEqual",[hb]],["Log",[tb]],["MatMul",[iv]],["MatMulNBits",[sv,nv]],["MaxPool",[dv,pv]],["Mul",[ob]],["MultiHeadAttention",[Hb,Kb]],["Neg",[VM]],["Not",[NM]],["Pad",[av]],["Pow",[lb]],["QuickGelu",[rb,hd]],["Range",[gv]],["Reciprocal",[UM]],["ReduceMin",[pM]],["ReduceMean",[oM]],["ReduceMax",[dM]],["ReduceSum",[fM]],["ReduceProd",[hM]],["ReduceL1",[lM]],["ReduceL2",[cM]],["ReduceLogSum",[_M]],["ReduceLogSumExp",[uM]],["ReduceSumSquare",[mM]],["Relu",[WM]],["Resize",[Mv,bv]],["RotaryEmbedding",[Yb]],["ScatterND",[wv,yv]],["Sigmoid",[GM]],["Sin",[qM]],["Sinh",[QM]],["Slice",[xv,Tv]],["SkipLayerNormalization",[vv]],["Split",[Qb,Xb]],["Sqrt",[XM]],["Softmax",[Ev,Pv]],["Sub",[cb]],["Tan",[JM]],["Tanh",[YM]],["ThresholdedRelu",[eb,hd]],["Tile",[Cv]],["Transpose",[Q0,X0]],["Where",[Sv]]])}),kv,QT=Je(()=>{Ii(),vs(),Lt(),kv=class{constructor(e){this.backend=e,this.repo=new Map,this.attributesBound=!1}getArtifact(e){return this.repo.get(e)}setArtifact(e,r){this.repo.set(e,r)}run(e,r,t,n,o){Xi(e.programInfo.name);let i=this.backend.device,u=this.backend.getComputePassEncoder();this.backend.writeTimestamp(this.backend.pendingDispatchNumber*2);let d=[];for(let f of r)d.push({binding:d.length,resource:{buffer:f.buffer}});for(let f of t)d.push({binding:d.length,resource:{buffer:f.buffer}});o&&d.push({binding:d.length,resource:o});let h=i.createBindGroup({layout:e.computePipeline.getBindGroupLayout(0),entries:d,label:e.programInfo.name});if(this.backend.sessionStatus==="capturing"){let f={kernelId:this.backend.currentKernelId,computePipeline:e.computePipeline,bindGroup:h,dispatchGroup:n};this.backend.capturedCommandList.get(this.backend.currentSessionId).push(f)}u.setPipeline(e.computePipeline),u.setBindGroup(0,h),u.dispatchWorkgroups(...n),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(),ki(e.programInfo.name)}dispose(){}build(e,r){Xi(e.name);let t=this.backend.device,n=[];[{feature:"shader-f16",extension:"f16"},{feature:"subgroups",extension:"subgroups"}].forEach(f=>{t.features.has(f.feature)&&n.push(`enable ${f.extension};`)});let o=q0(r,this.backend.device.limits),i=e.getShaderSource(o),u=`${n.join(`
`)}
${o.additionalImplementations}
${i}`,d=t.createShaderModule({code:u,label:e.name});Gt("verbose",()=>`[WebGPU] ${e.name} shader code: ${u}`);let h=t.createComputePipeline({compute:{module:d,entryPoint:"main"},layout:"auto",label:e.name});return ki(e.name),{programInfo:e,computePipeline:h,uniformVariablesInfo:o.variablesInfo}}normalizeDispatchGroupSize(e){let r=typeof e=="number"?e:e.x,t=typeof e=="number"?1:e.y||1,n=typeof e=="number"?1:e.z||1,o=this.backend.device.limits.maxComputeWorkgroupsPerDimension;if(r<=o&&t<=o&&n<=o)return[r,t,n];let i=r*t*n,u=Math.ceil(Math.sqrt(i));if(u>o){if(u=Math.ceil(Math.cbrt(i)),u>o)throw new Error("Total dispatch size exceeds WebGPU maximum.");return[u,u,u]}else return[u,u,1]}}}),Iv={};Ga(Iv,{WebGpuBackend:()=>Av});var Jw,Yw,Zw,Av,XT=Je(()=>{Ii(),St(),vs(),U0(),aT(),qT(),QT(),Jw=(e,r)=>{if(r.length!==e.length)throw new Error(`inputDependencies length ${r.length} is not equal to inputTensors length ${e.length}.`);let t=[];for(let n=0;n<e.length;++n){let o=e[n].dataType;switch(r[n]){case"none":{t.push("");break}case"type":{t.push(`${o}`);break}case"rank":{let i=e[n].dims.length;t.push(`${o};${i}`);break}case"dims":{let i=e[n].dims.join(",");t.push(`${o};${i}`);break}default:throw new Error(`unsupported input dependency: ${r[n]}`)}}return t.join("|")},Yw=(e,r,t)=>{var o,i;let n=e.name;return(o=e.shaderCache)!=null&&o.hint&&(n+="["+e.shaderCache.hint+"]"),n+=":"+t+`:${Jw(r,((i=e.shaderCache)==null?void 0:i.inputDependencies)??new Array(r.length).fill("dims"))}`,n},Zw=class{constructor(e){e&&(this.architecture=e.architecture,this.vendor=e.vendor)}isArchitecture(e){return this.architecture===e}isVendor(e){return this.vendor===e}},Av=class{constructor(){this.currentSessionId=null,this.currentKernelId=null,this.commandEncoder=null,this.computePassEncoder=null,this.maxDispatchNumber=16,this.pendingDispatchNumber=0,this.pendingKernels=[],this.pendingQueries=new Map,this.sessionStatus="default",this.capturedCommandList=new Map,this.capturedPendingKernels=new Map,this.sessionExternalDataMapping=new Map}get currentKernelCustomData(){if(this.currentKernelId===null)throw new Error("currentKernelCustomData(): currentKernelId is null. (should not happen)");let e=this.kernelCustomData.get(this.currentKernelId);return e||(e={},this.kernelCustomData.set(this.currentKernelId,e)),e}async initialize(e,r){this.env=e;let t=[],n={requiredLimits:{maxComputeWorkgroupStorageSize:r.limits.maxComputeWorkgroupStorageSize,maxComputeWorkgroupsPerDimension:r.limits.maxComputeWorkgroupsPerDimension,maxStorageBufferBindingSize:r.limits.maxStorageBufferBindingSize,maxBufferSize:r.limits.maxBufferSize,maxComputeInvocationsPerWorkgroup:r.limits.maxComputeInvocationsPerWorkgroup,maxComputeWorkgroupSizeX:r.limits.maxComputeWorkgroupSizeX,maxComputeWorkgroupSizeY:r.limits.maxComputeWorkgroupSizeY,maxComputeWorkgroupSizeZ:r.limits.maxComputeWorkgroupSizeZ},requiredFeatures:t},o=i=>r.features.has(i)&&t.push(i)&&!0;o("chromium-experimental-timestamp-query-inside-passes")||o("timestamp-query"),o("shader-f16"),o("subgroups"),this.device=await r.requestDevice(n),this.adapterInfo=new Zw(r.info||await r.requestAdapterInfo()),this.gpuDataManager=K0(this),this.programManager=new kv(this),this.kernels=new Map,this.kernelPersistentData=new Map,this.kernelCustomData=new Map,i_(e.logLevel,!!e.debug),this.device.onuncapturederror=i=>{i.error instanceof GPUValidationError&&console.error(`An uncaught WebGPU validation error was raised: ${i.error.message}`)},Object.defineProperty(this.env.webgpu,"device",{value:this.device,writable:!1,enumerable:!0,configurable:!1}),Object.defineProperty(this.env.webgpu,"adapter",{value:r,writable:!1,enumerable:!0,configurable:!1}),this.setQueryType()}dispose(){typeof this.querySet<"u"&&this.querySet.destroy(),this.gpuDataManager.dispose()}getCommandEncoder(){return this.commandEncoder||(this.commandEncoder=this.device.createCommandEncoder()),this.commandEncoder}getComputePassEncoder(){if(!this.computePassEncoder){let e=this.getCommandEncoder(),r={};this.queryType==="at-passes"&&(r.timestampWrites={querySet:this.querySet,beginningOfPassWriteIndex:this.pendingDispatchNumber*2,endOfPassWriteIndex:this.pendingDispatchNumber*2+1}),this.computePassEncoder=e.beginComputePass(r)}return this.computePassEncoder}endComputePass(){this.computePassEncoder&&(this.computePassEncoder.end(),this.computePassEncoder=null)}flush(){if(!this.commandEncoder)return;Xi(),this.endComputePass();let e;this.queryType!=="none"&&(this.commandEncoder.resolveQuerySet(this.querySet,0,this.pendingDispatchNumber*2,this.queryResolveBuffer,0),e=this.device.createBuffer({size:this.pendingDispatchNumber*2*8,usage:GPUBufferUsage.MAP_READ|GPUBufferUsage.COPY_DST}),this.pendingQueries.set(e,this.pendingKernels),this.pendingKernels=[],this.commandEncoder.copyBufferToBuffer(this.queryResolveBuffer,0,e,0,this.pendingDispatchNumber*2*8)),this.device.queue.submit([this.commandEncoder.finish()]),this.gpuDataManager.refreshPendingBuffers(),this.commandEncoder=null,this.pendingDispatchNumber=0,this.queryType!=="none"&&e.mapAsync(GPUMapMode.READ).then(()=>{var n;let r=new BigUint64Array(e.getMappedRange()),t=this.pendingQueries.get(e);for(let o=0;o<r.length/2;o++){let i=t[o],u=i.kernelId,d=this.kernels.get(u),h=d.kernelType,f=d.kernelName,y=i.programName,m=i.inputTensorViews,g=i.outputTensorViews,v=r[o*2],b=r[o*2+1];typeof this.queryTimeBase>"u"&&(this.queryTimeBase=v);let k=Number(v-this.queryTimeBase),z=Number(b-this.queryTimeBase);if(!Number.isSafeInteger(k)||!Number.isSafeInteger(z))throw new RangeError("incorrect timestamp range");if((n=this.env.webgpu.profiling)!=null&&n.ondata)this.env.webgpu.profiling.ondata({version:1,inputsMetadata:m.map(E=>({dims:E.dims,dataType:bs(E.dataType)})),outputsMetadata:g.map(E=>({dims:E.dims,dataType:bs(E.dataType)})),kernelId:u,kernelType:h,kernelName:f,programName:y,startTime:k,endTime:z});else{let E="";m.forEach((D,I)=>{E+=`input[${I}]: [${D.dims}] | ${bs(D.dataType)}, `});let T="";g.forEach((D,I)=>{T+=`output[${I}]: [${D.dims}] | ${bs(D.dataType)}, `}),console.log(`[profiling] kernel "${u}|${h}|${f}|${y}" ${E}${T}execution time: ${z-k} ns`)}yd("GPU",`${y}::${v}::${b}`)}e.unmap(),this.pendingQueries.delete(e)}),ki()}run(e,r,t,n,o,i){Xi(e.name);let u=[];for(let T=0;T<r.length;++T){let D=r[T].data;if(D===0)continue;let I=this.gpuDataManager.get(D);if(!I)throw new Error(`no GPU data for input: ${D}`);u.push(I)}let{outputs:d,dispatchGroup:h,programUniforms:f}=e.getRunData(r),y=t.length===0?d.map((T,D)=>D):t;if(y.length!==d.length)throw new Error(`Output size ${y.length} must be equal to ${d.length}.`);let m=[],g=[];for(let T=0;T<d.length;++T){if(!Number.isInteger(y[T])||y[T]<-3||y[T]>=i)throw new Error(`Invalid output index: ${y[T]}`);if(y[T]===-3)continue;let D=y[T]===-1,I=y[T]===-2,$=D||I?o(d[T].dataType,d[T].dims):n(y[T],d[T].dataType,d[T].dims);if(m.push($),$.data===0)continue;let P=this.gpuDataManager.get($.data);if(!P)throw new Error(`no GPU data for output: ${$.data}`);if(D&&this.temporaryData.push(P),I){let S=this.kernelPersistentData.get(this.currentKernelId);S||(S=[],this.kernelPersistentData.set(this.currentKernelId,S)),S.push(P)}g.push(P)}if(u.length!==r.length||g.length!==m.length){if(g.length===0)return ki(e.name),m;throw new Error(`Program ${e.name} has zero-sized tensor(s) in inputs or outputs. This is not supported now.`)}let v;if(f){let T=0,D=[];f.forEach(S=>{let O=typeof S.data=="number"?[S.data]:S.data;if(O.length===0)return;let R=S.type===10?2:4,G,ee;S.type===10?(ee=O.length>4?16:O.length>2?8:O.length*R,G=O.length>4?16:R*O.length):(ee=O.length<=2?O.length*R:16,G=16),T=Math.ceil(T/ee)*ee,D.push(T);let ie=S.type===10?8:4;T+=O.length>4?Math.ceil(O.length/ie)*G:O.length*R});let I=16;T=Math.ceil(T/I)*I;let $=new ArrayBuffer(T);f.forEach((S,O)=>{let R=D[O],G=typeof S.data=="number"?[S.data]:S.data;if(S.type===6)new Int32Array($,R,G.length).set(G);else if(S.type===12)new Uint32Array($,R,G.length).set(G);else if(S.type===10)new Uint16Array($,R,G.length).set(G);else if(S.type===1)new Float32Array($,R,G.length).set(G);else throw new Error(`Unsupported uniform type: ${bs(S.type)}`)});let P=this.gpuDataManager.create(T,GPUBufferUsage.COPY_DST|GPUBufferUsage.UNIFORM);this.device.queue.writeBuffer(P.buffer,0,$,0,T),this.gpuDataManager.release(P.id),v={offset:0,size:T,buffer:P.buffer}}let b=this.programManager.normalizeDispatchGroupSize(h),k=b[1]===1&&b[2]===1,z=Yw(e,r,k),E=this.programManager.getArtifact(z);if(E||(E=this.programManager.build(e,b),this.programManager.setArtifact(z,E),Gt("info",()=>`[artifact] key: ${z}, programName: ${e.name}`)),f&&E.uniformVariablesInfo){if(f.length!==E.uniformVariablesInfo.length)throw new Error(`Uniform variables count mismatch: expect ${E.uniformVariablesInfo.length}, got ${f.length} in program "${E.programInfo.name}".`);for(let T=0;T<f.length;T++){let D=f[T],I=D.type,$=typeof D.data=="number"?1:D.data.length,[P,S]=E.uniformVariablesInfo[T];if(I!==P||$!==S)throw new Error(`Uniform variable ${T} mismatch: expect type ${P} with size ${S}, got type ${I} with size ${$} in program "${E.programInfo.name}".`)}}if(Gt("info",()=>`[ProgramManager] run "${e.name}" (key=${z}) with ${b[0]}x${b[1]}x${b[2]}`),this.queryType!=="none"||this.sessionStatus==="capturing"){let T={kernelId:this.currentKernelId,programName:E.programInfo.name,inputTensorViews:r,outputTensorViews:m};this.pendingKernels.push(T),this.sessionStatus==="capturing"&&this.capturedPendingKernels.get(this.currentSessionId).push(T)}return this.programManager.run(E,u,g,b,v),ki(e.name),m}upload(e,r){this.gpuDataManager.upload(e,r)}memcpy(e,r){this.gpuDataManager.memcpy(e,r)}async download(e,r){await this.gpuDataManager.download(e,r)}alloc(e){return this.gpuDataManager.create(e).id}free(e){return this.gpuDataManager.release(e)}createKernel(e,r,t,n){let o=$v.get(e);if(!o)throw new Error(`kernel not implemented: ${e}`);let i={kernelType:e,kernelName:n,kernelEntry:o[0],attributes:[o[1],t]};this.kernels.set(r,i)}releaseKernel(e){let r=this.kernelPersistentData.get(e);if(r){for(let t of r)this.gpuDataManager.release(t.id);this.kernelPersistentData.delete(e)}this.kernelCustomData.delete(e),this.kernels.delete(e)}computeKernel(e,r,t){let n=this.kernels.get(e);if(!n)throw new Error(`kernel not created: ${e}`);let o=n.kernelType,i=n.kernelName,u=n.kernelEntry,d=n.attributes;if(this.currentKernelId!==null)throw new Error(`kernel "[${o}] ${i}" is not allowed to be called recursively`);this.currentKernelId=e,d[0]&&(d[1]=d[0](d[1]),d[0]=void 0),Gt("info",()=>`[WebGPU] Start to run kernel "[${o}] ${i}"...`);let h=this.env.debug;this.temporaryData=[];try{return h&&this.device.pushErrorScope("validation"),u(r,d[1]),0}catch(f){return t.push(Promise.resolve(`[WebGPU] Kernel "[${o}] ${i}" failed. ${f}`)),1}finally{h&&t.push(this.device.popErrorScope().then(f=>f?`GPU validation error for kernel "[${o}] ${i}": ${f.message}`:null));for(let f of this.temporaryData)this.gpuDataManager.release(f.id);this.temporaryData=[],this.currentKernelId=null}}registerBuffer(e,r,t,n){let o=this.sessionExternalDataMapping.get(e);o||(o=new Map,this.sessionExternalDataMapping.set(e,o));let i=o.get(r),u=this.gpuDataManager.registerExternalBuffer(t,n,i);return o.set(r,[u,t]),u}unregisterBuffers(e){let r=this.sessionExternalDataMapping.get(e);r&&(r.forEach(t=>this.gpuDataManager.unregisterExternalBuffer(t[0])),this.sessionExternalDataMapping.delete(e))}getBuffer(e){let r=this.gpuDataManager.get(e);if(!r)throw new Error(`no GPU data for buffer: ${e}`);return r.buffer}createDownloader(e,r,t){return async()=>{let n=await Fm(this,e,r);return s_(n.buffer,t)}}writeTimestamp(e){this.queryType==="inside-passes"&&this.computePassEncoder.writeTimestamp(this.querySet,e)}setQueryType(){var e;this.queryType="none",(((e=this.env.webgpu.profiling)==null?void 0:e.mode)==="default"||(typeof this.env.trace>"u"?this.env.wasm.trace:this.env.trace))&&(this.device.features.has("chromium-experimental-timestamp-query-inside-passes")?this.queryType="inside-passes":this.device.features.has("timestamp-query")&&(this.queryType="at-passes"),this.queryType!=="none"&&typeof this.querySet>"u"&&(this.querySet=this.device.createQuerySet({type:"timestamp",count:this.maxDispatchNumber*2}),this.queryResolveBuffer=this.device.createBuffer({size:this.maxDispatchNumber*2*8,usage:GPUBufferUsage.COPY_SRC|GPUBufferUsage.QUERY_RESOLVE})))}captureBegin(){Gt("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(){Gt("info","captureEnd"),this.flush(),this.sessionStatus="default"}replay(){Gt("info","replay"),this.sessionStatus="replaying";let e=this.capturedCommandList.get(this.currentSessionId),r=this.capturedPendingKernels.get(this.currentSessionId),t=e.length;this.pendingKernels=[];for(let n=0;n<t;n++){let o=this.getComputePassEncoder(),i=e[n];this.writeTimestamp(this.pendingDispatchNumber*2),o.setPipeline(i.computePipeline),o.setBindGroup(0,i.bindGroup),o.dispatchWorkgroups(...i.dispatchGroup),this.writeTimestamp(this.pendingDispatchNumber*2+1),this.pendingDispatchNumber++,this.queryType!=="none"&&this.pendingKernels.push(r[n]),(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()}}}),Ov={};Ga(Ov,{init:()=>Fv});var Kh,e0,Fv,JT=Je(()=>{St(),vs(),Ft(),nT(),Kh=class Dv{constructor(r,t,n,o){this.module=r,this.dataType=t,this.data=n,this.dims=o}getFloat32Array(){if(this.dataType!==1)throw new Error("Invalid data type");let r=Fe.size(this.dims);return r===0?new Float32Array:new Float32Array(this.module.HEAP8.buffer,this.data,r)}getBigInt64Array(){if(this.dataType!==7)throw new Error("Invalid data type");let r=Fe.size(this.dims);return r===0?new BigInt64Array:new BigInt64Array(this.module.HEAP8.buffer,this.data,r)}getInt32Array(){if(this.dataType!==6)throw new Error("Invalid data type");let r=Fe.size(this.dims);return r===0?new Int32Array:new Int32Array(this.module.HEAP8.buffer,this.data,r)}getUint16Array(){if(this.dataType!==10&&this.dataType!==4)throw new Error("Invalid data type");let r=Fe.size(this.dims);return r===0?new Uint16Array:new Uint16Array(this.module.HEAP8.buffer,this.data,r)}reshape(r){if(Fe.size(r)!==Fe.size(this.dims))throw new Error("Invalid new shape");return new Dv(this.module,this.dataType,this.data,r)}},e0=class{constructor(e,r,t){this.module=e,this.backend=r,this.customDataOffset=0,this.customDataSize=0,this.adapterInfo=r.adapterInfo;let n=e.PTR_SIZE,o=t/e.PTR_SIZE,i=n===4?"i32":"i64";this.opKernelContext=Number(e.getValue(n*o++,i));let u=Number(e.getValue(n*o++,i));this.outputCount=Number(e.getValue(n*o++,i)),this.customDataOffset=Number(e.getValue(n*o++,"*")),this.customDataSize=Number(e.getValue(n*o++,i));let d=[];for(let h=0;h<u;h++){let f=Number(e.getValue(n*o++,i)),y=Number(e.getValue(n*o++,"*")),m=Number(e.getValue(n*o++,i)),g=[];for(let v=0;v<m;v++)g.push(Number(e.getValue(n*o++,i)));d.push(new Kh(e,f,y,g))}this.inputs=d}get kernelCustomData(){return this.backend.currentKernelCustomData}get customDataBuffer(){return this.module.HEAPU8.subarray(this.customDataOffset,this.customDataOffset+this.customDataSize)}compute(e,r){var u;let t=((u=r==null?void 0:r.inputs)==null?void 0:u.map(d=>typeof d=="number"?this.inputs[d]:d))??this.inputs,n=(r==null?void 0:r.outputs)??[],o=(d,h,f)=>new Kh(this.module,h,this.output(d,f),f),i=(d,h)=>{let f=bn(d,h);if(!f)throw new Error(`Unsupported data type: ${d}`);let y=f>0?this.backend.gpuDataManager.create(f).id:0;return new Kh(this.module,d,y,h)};return this.backend.run(e,t,n,o,i,this.outputCount)}output(e,r){let t=this.module.stackSave();try{let n=this.module.PTR_SIZE,o=n===4?"i32":"i64",i=this.module.stackAlloc((1+r.length)*n);this.module.setValue(i,r.length,o);for(let u=0;u<r.length;u++)this.module.setValue(i+n*(u+1),r[u],o);return this.module._JsepOutput(this.opKernelContext,e,i)}catch(n){throw new Error(`Failed to generate kernel's output[${e}] with dims [${r}]. If you are running with pre-allocated output, please make sure the output type/dims are correct. Error: ${n}`)}finally{this.module.stackRestore(t)}}},Fv=async(e,r,t,n)=>{let o=r.jsepInit;if(!o)throw new Error("Failed to initialize JSEP. The WebAssembly module is not built with JSEP support.");if(e==="webgpu"){let i=(XT(),gd(Iv)).WebGpuBackend,u=new i;await u.initialize(t,n),o("webgpu",[u,d=>u.alloc(Number(d)),d=>u.free(d),(d,h,f,y=!1)=>{if(y)Gt("verbose",()=>`[WebGPU] jsepCopyGpuToGpu: src=${Number(d)}, dst=${Number(h)}, size=${Number(f)}`),u.memcpy(Number(d),Number(h));else{Gt("verbose",()=>`[WebGPU] jsepCopyCpuToGpu: dataOffset=${Number(d)}, gpuDataId=${Number(h)}, size=${Number(f)}`);let m=r.HEAPU8.subarray(Number(d>>>0),Number(d>>>0)+Number(f));u.upload(Number(h),m)}},async(d,h,f)=>{Gt("verbose",()=>`[WebGPU] jsepCopyGpuToCpu: gpuDataId=${d}, dataOffset=${h}, size=${f}`),await u.download(Number(d),()=>r.HEAPU8.subarray(Number(h)>>>0,Number(h+f)>>>0))},(d,h,f)=>u.createKernel(d,Number(h),f,r.UTF8ToString(r._JsepGetNodeName(Number(h)))),d=>u.releaseKernel(d),(d,h,f,y)=>{Gt("verbose",()=>`[WebGPU] jsepRun: sessionHandle=${f}, kernel=${d}, contextDataOffset=${h}`);let m=new e0(r,u,Number(h));return u.computeKernel(Number(d),m,y)},()=>u.captureBegin(),()=>u.captureEnd(),()=>u.replay()])}else{let i=new G0(t);o("webnn",[i,()=>i.reserveTensorId(),u=>i.releaseTensorId(u),async(u,d,h,f,y)=>i.ensureTensor(u,d,h,f,y),(u,d)=>{i.uploadTensor(u,d)},async(u,d)=>i.downloadTensor(u,d),!!t.trace])}}}),t0,m_,__,Ds,r0,xm,of,g_,y_,Tm,w_,M_,b_,Lv=Je(()=>{Ii(),rT(),iT(),St(),Pn(),Zm(),B0(),t0=(e,r)=>{lr()._OrtInit(e,r)!==0&&ir("Can't initialize onnxruntime.")},m_=async e=>{t0(e.wasm.numThreads,tf(e.logLevel))},__=async(e,r)=>{var t,n;(n=(t=lr()).asyncInit)==null||n.call(t);{let o=(JT(),gd(Ov)).init;if(r==="webgpu"){if(typeof navigator>"u"||!navigator.gpu)throw new Error("WebGPU is not supported in current environment");let i=e.webgpu.adapter;if(i){if(typeof i.limits!="object"||typeof i.features!="object"||typeof i.requestDevice!="function")throw new Error("Invalid GPU adapter set in `env.webgpu.adapter`. It must be a GPUAdapter object.")}else{let u=e.webgpu.powerPreference;if(u!==void 0&&u!=="low-power"&&u!=="high-performance")throw new Error(`Invalid powerPreference setting: "${u}"`);let d=e.webgpu.forceFallbackAdapter;if(d!==void 0&&typeof d!="boolean")throw new Error(`Invalid forceFallbackAdapter setting: "${d}"`);if(i=await navigator.gpu.requestAdapter({powerPreference:u,forceFallbackAdapter:d}),!i)throw new Error('Failed to get GPU adapter. You may need to enable flag "--enable-unsafe-webgpu" if you are using Chrome.')}await o("webgpu",lr(),e,i)}if(r==="webnn"){if(typeof navigator>"u"||!navigator.ml)throw new Error("WebNN is not supported in current environment");await o("webnn",lr(),e)}}},Ds=new Map,r0=e=>{let r=lr(),t=r.stackSave();try{let n=r.PTR_SIZE,o=r.stackAlloc(2*n);r._OrtGetInputOutputCount(e,o,o+n)!==0&&ir("Can't get session input/output count.");let i=n===4?"i32":"i64";return[Number(r.getValue(o,i)),Number(r.getValue(o+n,i))]}finally{r.stackRestore(t)}},xm=(e,r)=>{let t=lr(),n=t.stackSave(),o=0;try{let i=t.PTR_SIZE,u=t.stackAlloc(2*i);t._OrtGetInputOutputMetadata(e,r,u,u+i)!==0&&ir("Can't get session input/output metadata.");let d=Number(t.getValue(u,"*"));o=Number(t.getValue(u+i,"*"));let h=t.HEAP32[o/4];if(h===0)return[d,0];let f=t.HEAPU32[o/4+1],y=[];for(let m=0;m<f;m++){let g=Number(t.getValue(o+8+m*i,"*"));y.push(g!==0?t.UTF8ToString(g):Number(t.getValue(o+8+(m+f)*i,"*")))}return[d,h,y]}finally{t.stackRestore(n),o!==0&&t._OrtFree(o)}},of=e=>{let r=lr(),t=r._malloc(e.byteLength);if(t===0)throw new Error(`Can't create a session. failed to allocate a buffer of size ${e.byteLength}.`);return r.HEAPU8.set(e,t),[t,e.byteLength]},g_=async(e,r)=>{var m,g,v,b;let t,n,o=lr();Array.isArray(e)?[t,n]=e:e.buffer===o.HEAPU8.buffer?[t,n]=[e.byteOffset,e.byteLength]:[t,n]=of(e);let i=0,u=0,d=0,h=[],f=[],y=[];try{if([u,h]=await z0(r),(r==null?void 0:r.externalData)&&o.mountExternalData){let O=[];for(let R of r.externalData){let G=typeof R=="string"?R:R.path;O.push(r_(typeof R=="string"?R:R.data).then(ee=>{o.mountExternalData(G,ee)}))}await Promise.all(O)}for(let O of(r==null?void 0:r.executionProviders)??[])if((typeof O=="string"?O:O.name)==="webnn"){if(o.shouldTransferToMLTensor=!1,typeof O!="string"){let R=O,G=R==null?void 0:R.context,ee=R==null?void 0:R.gpuDevice,ie=R==null?void 0:R.deviceType,H=R==null?void 0:R.powerPreference;G?o.currentContext=G:ee?o.currentContext=await o.webnnCreateMLContext(ee):o.currentContext=await o.webnnCreateMLContext({deviceType:ie,powerPreference:H})}else o.currentContext=await o.webnnCreateMLContext();break}i=await o._OrtCreateSession(t,n,u),(m=o.webgpuOnCreateSession)==null||m.call(o,i),i===0&&ir("Can't create a session."),(g=o.jsepOnCreateSession)==null||g.call(o),o.currentContext&&(o.webnnRegisterMLContext(i,o.currentContext),o.currentContext=void 0,o.shouldTransferToMLTensor=!0);let[k,z]=r0(i),E=!!(r!=null&&r.enableGraphCapture),T=[],D=[],I=[],$=[],P=[];for(let O=0;O<k;O++){let[R,G,ee]=xm(i,O);R===0&&ir("Can't get an input name."),f.push(R);let ie=o.UTF8ToString(R);T.push(ie),I.push(G===0?{name:ie,isTensor:!1}:{name:ie,isTensor:!0,type:bs(G),shape:ee})}for(let O=0;O<z;O++){let[R,G,ee]=xm(i,O+k);R===0&&ir("Can't get an output name."),y.push(R);let ie=o.UTF8ToString(R);D.push(ie),$.push(G===0?{name:ie,isTensor:!1}:{name:ie,isTensor:!0,type:bs(G),shape:ee});{if(E&&(r==null?void 0:r.preferredOutputLocation)===void 0){P.push("gpu-buffer");continue}let H=typeof(r==null?void 0:r.preferredOutputLocation)=="string"?r.preferredOutputLocation:((v=r==null?void 0:r.preferredOutputLocation)==null?void 0:v[ie])??"cpu",ce=o.webnnIsGraphOutput;if(H==="cpu"&&ce&&ce(i,ie)){P.push("ml-tensor-cpu-output");continue}if(H!=="cpu"&&H!=="cpu-pinned"&&H!=="gpu-buffer"&&H!=="ml-tensor")throw new Error(`Not supported preferred output location: ${H}.`);if(E&&H!=="gpu-buffer")throw new Error(`Not supported preferred output location: ${H}. Only 'gpu-buffer' location is supported when enableGraphCapture is true.`);P.push(H)}}let S=null;return P.some(O=>O==="gpu-buffer"||O==="ml-tensor"||O==="ml-tensor-cpu-output")&&(d=o._OrtCreateBinding(i),d===0&&ir("Can't create IO binding."),S={handle:d,outputPreferredLocations:P,outputPreferredLocationsEncoded:P.map(O=>O==="ml-tensor-cpu-output"?"ml-tensor":O).map(O=>Am(O))}),Ds.set(i,[i,f,y,S,E,!1]),[i,T,D,I,$]}catch(k){throw f.forEach(z=>o._OrtFree(z)),y.forEach(z=>o._OrtFree(z)),d!==0&&o._OrtReleaseBinding(d)!==0&&ir("Can't release IO binding."),i!==0&&o._OrtReleaseSession(i)!==0&&ir("Can't release session."),k}finally{o._free(t),u!==0&&o._OrtReleaseSessionOptions(u)!==0&&ir("Can't release session options."),h.forEach(k=>o._free(k)),(b=o.unmountExternalData)==null||b.call(o)}},y_=e=>{var h,f,y;let r=lr(),t=Ds.get(e);if(!t)throw new Error(`cannot release session. invalid session id: ${e}`);let[n,o,i,u,d]=t;u&&(d&&r._OrtClearBoundOutputs(u.handle)!==0&&ir("Can't clear bound outputs."),r._OrtReleaseBinding(u.handle)!==0&&ir("Can't release IO binding.")),(h=r.jsepOnReleaseSession)==null||h.call(r,e),(f=r.webnnOnReleaseSession)==null||f.call(r,e),(y=r.webgpuOnReleaseSession)==null||y.call(r,e),o.forEach(m=>r._OrtFree(m)),i.forEach(m=>r._OrtFree(m)),r._OrtReleaseSession(n)!==0&&ir("Can't release session."),Ds.delete(e)},Tm=async(e,r,t,n,o,i,u=!1)=>{if(!e){r.push(0);return}let d=lr(),h=d.PTR_SIZE,f=e[0],y=e[1],m=e[3],g=m,v,b;if(f==="string"&&(m==="gpu-buffer"||m==="ml-tensor"))throw new Error("String tensor is not supported on GPU.");if(u&&m!=="gpu-buffer")throw new Error(`External buffer must be provided for input/output index ${i} when enableGraphCapture is true.`);if(m==="gpu-buffer"){let E=e[2].gpuBuffer;b=bn(Mn(f),y);{let T=d.jsepRegisterBuffer;if(!T)throw new Error('Tensor location "gpu-buffer" is not supported without using WebGPU.');v=T(n,i,E,b)}}else if(m==="ml-tensor"){let E=e[2].mlTensor;b=bn(Mn(f),y);let T=d.webnnRegisterMLTensor;if(!T)throw new Error('Tensor location "ml-tensor" is not supported without using WebNN.');v=T(n,E,Mn(f),y)}else{let E=e[2];if(Array.isArray(E)){b=h*E.length,v=d._malloc(b),t.push(v);for(let T=0;T<E.length;T++){if(typeof E[T]!="string")throw new TypeError(`tensor data at index ${T} is not a string`);d.setValue(v+T*h,qi(E[T],t),"*")}}else{let T=d.webnnIsGraphInput,D=d.webnnIsGraphOutput;if(f!=="string"&&T&&D){let I=d.UTF8ToString(o);if(T(n,I)||D(n,I)){let $=Mn(f);b=bn($,y),g="ml-tensor";let P=d.webnnCreateTemporaryTensor,S=d.webnnUploadTensor;if(!P||!S)throw new Error('Tensor location "ml-tensor" is not supported without using WebNN.');let O=await P(n,$,y);S(O,new Uint8Array(E.buffer,E.byteOffset,E.byteLength)),v=O}else b=E.byteLength,v=d._malloc(b),t.push(v),d.HEAPU8.set(new Uint8Array(E.buffer,E.byteOffset,b),v)}else b=E.byteLength,v=d._malloc(b),t.push(v),d.HEAPU8.set(new Uint8Array(E.buffer,E.byteOffset,b),v)}}let k=d.stackSave(),z=d.stackAlloc(4*y.length);try{y.forEach((T,D)=>d.setValue(z+D*h,T,h===4?"i32":"i64"));let E=d._OrtCreateTensor(Mn(f),v,b,z,y.length,Am(g));E===0&&ir(`Can't create tensor for input/output. session=${n}, index=${i}.`),r.push(E)}finally{d.stackRestore(k)}},w_=async(e,r,t,n,o,i)=>{var ee,ie,H,ce;let u=lr(),d=u.PTR_SIZE,h=Ds.get(e);if(!h)throw new Error(`cannot run inference. invalid session id: ${e}`);let f=h[0],y=h[1],m=h[2],g=h[3],v=h[4],b=h[5],k=r.length,z=n.length,E=0,T=[],D=[],I=[],$=[],P=u.stackSave(),S=u.stackAlloc(k*d),O=u.stackAlloc(k*d),R=u.stackAlloc(z*d),G=u.stackAlloc(z*d);try{[E,T]=j0(i),zs("wasm prepareInputOutputTensor");for(let ae=0;ae<k;ae++)await Tm(t[ae],D,$,e,y[r[ae]],r[ae],v);for(let ae=0;ae<z;ae++)await Tm(o[ae],I,$,e,m[n[ae]],k+n[ae],v);Bs("wasm prepareInputOutputTensor");for(let ae=0;ae<k;ae++)u.setValue(S+ae*d,D[ae],"*"),u.setValue(O+ae*d,y[r[ae]],"*");for(let ae=0;ae<z;ae++)u.setValue(R+ae*d,I[ae],"*"),u.setValue(G+ae*d,m[n[ae]],"*");if(g&&!b){let{handle:ae,outputPreferredLocations:Ce,outputPreferredLocationsEncoded:Te}=g;if(y.length!==k)throw new Error(`input count from feeds (${k}) is expected to be always equal to model's input count (${y.length}).`);zs("wasm bindInputsOutputs");for(let q=0;q<k;q++){let B=r[q];await u._OrtBindInput(ae,y[B],D[q])!==0&&ir(`Can't bind input[${q}] for session=${e}.`)}for(let q=0;q<z;q++){let B=n[q];(ee=o[q])!=null&&ee[3]?u._OrtBindOutput(ae,m[B],I[q],0)!==0&&ir(`Can't bind pre-allocated output[${q}] for session=${e}.`):u._OrtBindOutput(ae,m[B],0,Te[B])!==0&&ir(`Can't bind output[${q}] to ${Ce[q]} for session=${e}.`)}Bs("wasm bindInputsOutputs"),Ds.set(e,[f,y,m,g,v,!0])}(ie=u.jsepOnRunStart)==null||ie.call(u,f),(H=u.webnnOnRunStart)==null||H.call(u,f);let re;g?re=await u._OrtRunWithBinding(f,g.handle,z,R,E):re=await u._OrtRun(f,O,S,k,G,z,R,E),re!==0&&ir("failed to call OrtRun().");let se=[],_e=[];zs("wasm ProcessOutputTensor");for(let ae=0;ae<z;ae++){let Ce=Number(u.getValue(R+ae*d,"*"));if(Ce===I[ae]){se.push(o[ae]);continue}let Te=u.stackSave(),q=u.stackAlloc(4*d),B=!1,Q,oe=0;try{u._OrtGetTensorData(Ce,q,q+d,q+2*d,q+3*d)!==0&&ir(`Can't access output tensor data on index ${ae}.`);let Ie=d===4?"i32":"i64",ye=Number(u.getValue(q,Ie));oe=u.getValue(q+d,"*");let be=u.getValue(q+d*2,"*"),et=Number(u.getValue(q+d*3,Ie)),tt=[];for(let pe=0;pe<et;pe++)tt.push(Number(u.getValue(be+pe*d,Ie)));u._OrtFree(be)!==0&&ir("Can't free memory for tensor dims.");let Ee=tt.reduce((pe,X)=>pe*X,1);Q=bs(ye);let Y=g==null?void 0:g.outputPreferredLocations[n[ae]];if(Q==="string"){if(Y==="gpu-buffer"||Y==="ml-tensor")throw new Error("String tensor is not supported on GPU.");let pe=[];for(let X=0;X<Ee;X++){let Ae=u.getValue(oe+X*d,"*"),Oe=u.getValue(oe+(X+1)*d,"*"),Ne=X===Ee-1?void 0:Oe-Ae;pe.push(u.UTF8ToString(Ae,Ne))}se.push([Q,tt,pe,"cpu"])}else if(Y==="gpu-buffer"&&Ee>0){let pe=u.jsepGetBuffer;if(!pe)throw new Error('preferredLocation "gpu-buffer" is not supported without using WebGPU.');let X=pe(oe),Ae=bn(ye,Ee);if(Ae===void 0||!e_(Q))throw new Error(`Unsupported data type: ${Q}`);B=!0,se.push([Q,tt,{gpuBuffer:X,download:u.jsepCreateDownloader(X,Ae,Q),dispose:()=>{u._OrtReleaseTensor(Ce)!==0&&ir("Can't release tensor.")}},"gpu-buffer"])}else if(Y==="ml-tensor"&&Ee>0){let pe=u.webnnEnsureTensor,X=u.webnnIsGraphInputOutputTypeSupported;if(!pe||!X)throw new Error('preferredLocation "ml-tensor" is not supported without using WebNN.');if(bn(ye,Ee)===void 0||!t_(Q))throw new Error(`Unsupported data type: ${Q}`);if(!X(e,Q,!1))throw new Error(`preferredLocation "ml-tensor" for ${Q} output is not supported by current WebNN Context.`);let Ae=await pe(e,oe,ye,tt,!1);B=!0,se.push([Q,tt,{mlTensor:Ae,download:u.webnnCreateMLTensorDownloader(oe,Q),dispose:()=>{u.webnnReleaseTensorId(oe),u._OrtReleaseTensor(Ce)}},"ml-tensor"])}else if(Y==="ml-tensor-cpu-output"&&Ee>0){let pe=u.webnnCreateMLTensorDownloader(oe,Q)(),X=se.length;B=!0,_e.push((async()=>{let Ae=[X,await pe];return u.webnnReleaseTensorId(oe),u._OrtReleaseTensor(Ce),Ae})()),se.push([Q,tt,[],"cpu"])}else{let pe=cf(Q),X=new pe(Ee);new Uint8Array(X.buffer,X.byteOffset,X.byteLength).set(u.HEAPU8.subarray(oe,oe+X.byteLength)),se.push([Q,tt,X,"cpu"])}}finally{u.stackRestore(Te),Q==="string"&&oe&&u._free(oe),B||u._OrtReleaseTensor(Ce)}}g&&!v&&(u._OrtClearBoundOutputs(g.handle)!==0&&ir("Can't clear bound outputs."),Ds.set(e,[f,y,m,g,v,!1]));for(let[ae,Ce]of await Promise.all(_e))se[ae][2]=Ce;return Bs("wasm ProcessOutputTensor"),se}finally{(ce=u.webnnOnRunEnd)==null||ce.call(u,f),u.stackRestore(P),D.forEach(re=>u._OrtReleaseTensor(re)),I.forEach(re=>u._OrtReleaseTensor(re)),$.forEach(re=>u._free(re)),E!==0&&u._OrtReleaseRunOptions(E),T.forEach(re=>u._free(re))}},M_=e=>{let r=lr(),t=Ds.get(e);if(!t)throw new Error("invalid session id");let n=t[0],o=r._OrtEndProfiling(n);o===0&&ir("Can't get an profile file name."),r._OrtFree(o)},b_=e=>{let r=[];for(let t of e){let n=t[2];!Array.isArray(n)&&"buffer"in n&&r.push(n.buffer)}return r}}),Ls,mi,za,cd,ud,Hh,Em,qh,_n,gn,i0,jv,zv,Bv,Rv,Nv,Vv,Uv,Wv=Je(()=>{Ii(),Lv(),Pn(),Jm(),Ls=()=>!!dr.wasm.proxy&&typeof document<"u",za=!1,cd=!1,ud=!1,qh=new Map,_n=(e,r)=>{let t=qh.get(e);t?t.push(r):qh.set(e,[r])},gn=()=>{if(za||!cd||ud||!mi)throw new Error("worker not ready")},i0=e=>{switch(e.data.type){case"init-wasm":za=!1,e.data.err?(ud=!0,Em[1](e.data.err)):(cd=!0,Em[0]()),Hh&&(URL.revokeObjectURL(Hh),Hh=void 0);break;case"init-ep":case"copy-from":case"create":case"release":case"run":case"end-profiling":{let r=qh.get(e.data.type);e.data.err?r.shift()[1](e.data.err):r.shift()[0](e.data.out);break}}},jv=async()=>{if(!cd){if(za)throw new Error("multiple calls to 'initWasm()' detected.");if(ud)throw new Error("previous call to 'initWasm()' failed.");if(za=!0,Ls())return new Promise((e,r)=>{mi==null||mi.terminate(),D0().then(([t,n])=>{try{mi=n,mi.onerror=i=>r(i),mi.onmessage=i0,Em=[e,r];let o={type:"init-wasm",in:dr};!o.in.wasm.wasmPaths&&(t||Im)&&(o.in.wasm.wasmPaths={wasm:new URL("/assets/ort-wasm-simd-threaded.jsep-D5UMYBk4.wasm",import.meta.url).href}),mi.postMessage(o),Hh=t}catch(o){r(o)}},r)});try{await Ym(dr.wasm),await m_(dr),cd=!0}catch(e){throw ud=!0,e}finally{za=!1}}},zv=async e=>{if(Ls())return gn(),new Promise((r,t)=>{_n("init-ep",[r,t]);let n={type:"init-ep",in:{epName:e,env:dr}};mi.postMessage(n)});await __(dr,e)},Bv=async e=>Ls()?(gn(),new Promise((r,t)=>{_n("copy-from",[r,t]);let n={type:"copy-from",in:{buffer:e}};mi.postMessage(n,[e.buffer])})):of(e),Rv=async(e,r)=>{if(Ls()){if(r!=null&&r.preferredOutputLocation)throw new Error('session option "preferredOutputLocation" is not supported for proxy.');return gn(),new Promise((t,n)=>{_n("create",[t,n]);let o={type:"create",in:{model:e,options:{...r}}},i=[];e instanceof Uint8Array&&i.push(e.buffer),mi.postMessage(o,i)})}else return g_(e,r)},Nv=async e=>{if(Ls())return gn(),new Promise((r,t)=>{_n("release",[r,t]);let n={type:"release",in:e};mi.postMessage(n)});y_(e)},Vv=async(e,r,t,n,o,i)=>{if(Ls()){if(t.some(u=>u[3]!=="cpu"))throw new Error("input tensor on GPU is not supported for proxy.");if(o.some(u=>u))throw new Error("pre-allocated output tensor is not supported for proxy.");return gn(),new Promise((u,d)=>{_n("run",[u,d]);let h=t,f={type:"run",in:{sessionId:e,inputIndices:r,inputs:h,outputIndices:n,options:i}};mi.postMessage(f,b_(h))})}else return w_(e,r,t,n,o,i)},Uv=async e=>{if(Ls())return gn(),new Promise((r,t)=>{_n("end-profiling",[r,t]);let n={type:"end-profiling",in:e};mi.postMessage(n)});M_(e)}}),Pm,s0,Gv,YT=Je(()=>{Ii(),Wv(),St(),Xm(),B0(),Pm=(e,r)=>{switch(e.location){case"cpu":return[e.type,e.dims,e.data,"cpu"];case"gpu-buffer":return[e.type,e.dims,{gpuBuffer:e.gpuBuffer},"gpu-buffer"];case"ml-tensor":return[e.type,e.dims,{mlTensor:e.mlTensor},"ml-tensor"];default:throw new Error(`invalid data location: ${e.location} for ${r()}`)}},s0=e=>{switch(e[3]){case"cpu":return new Qi(e[0],e[2],e[1]);case"gpu-buffer":{let r=e[0];if(!e_(r))throw new Error(`not supported data type: ${r} for deserializing GPU tensor`);let{gpuBuffer:t,download:n,dispose:o}=e[2];return Qi.fromGpuBuffer(t,{dataType:r,dims:e[1],download:n,dispose:o})}case"ml-tensor":{let r=e[0];if(!t_(r))throw new Error(`not supported data type: ${r} for deserializing MLTensor tensor`);let{mlTensor:t,download:n,dispose:o}=e[2];return Qi.fromMLTensor(t,{dataType:r,dims:e[1],download:n,dispose:o})}default:throw new Error(`invalid data location: ${e[3]}`)}},Gv=class{async fetchModelAndCopyToWasmMemory(e){return Bv(await r_(e))}async loadModel(e,r){Xi();let t;typeof e=="string"?t=await this.fetchModelAndCopyToWasmMemory(e):t=e,[this.sessionId,this.inputNames,this.outputNames,this.inputMetadata,this.outputMetadata]=await Rv(t,r),ki()}async dispose(){return Nv(this.sessionId)}async run(e,r,t){Xi();let n=[],o=[];Object.entries(e).forEach(m=>{let g=m[0],v=m[1],b=this.inputNames.indexOf(g);if(b===-1)throw new Error(`invalid input '${g}'`);n.push(v),o.push(b)});let i=[],u=[];Object.entries(r).forEach(m=>{let g=m[0],v=m[1],b=this.outputNames.indexOf(g);if(b===-1)throw new Error(`invalid output '${g}'`);i.push(v),u.push(b)});let d=n.map((m,g)=>Pm(m,()=>`input "${this.inputNames[o[g]]}"`)),h=i.map((m,g)=>m?Pm(m,()=>`output "${this.outputNames[u[g]]}"`):null),f=await Vv(this.sessionId,o,d,u,h,t),y={};for(let m=0;m<f.length;m++)y[this.outputNames[u[m]]]=i[m]??s0(f[m]);return ki(),y}startProfiling(){}endProfiling(){Uv(this.sessionId)}}}),Kv={};Ga(Kv,{OnnxruntimeWebAssemblyBackend:()=>Km,initializeFlags:()=>Gm,wasmBackend:()=>Hv});var Gm,Km,Hv,ZT=Je(()=>{Ii(),Wv(),YT(),Gm=()=>{(typeof dr.wasm.initTimeout!="number"||dr.wasm.initTimeout<0)&&(dr.wasm.initTimeout=0);let e=dr.wasm.simd;if(typeof e!="boolean"&&e!==void 0&&e!=="fixed"&&e!=="relaxed"&&(console.warn(`Property "env.wasm.simd" is set to unknown value "${e}". Reset it to \`false\` and ignore SIMD feature checking.`),dr.wasm.simd=!1),typeof dr.wasm.proxy!="boolean"&&(dr.wasm.proxy=!1),typeof dr.wasm.trace!="boolean"&&(dr.wasm.trace=!1),typeof dr.wasm.numThreads!="number"||!Number.isInteger(dr.wasm.numThreads)||dr.wasm.numThreads<=0)if(typeof self<"u"&&!self.crossOriginIsolated)dr.wasm.numThreads=1;else{let r=typeof navigator>"u"?B1("node:os").cpus().length:navigator.hardwareConcurrency;dr.wasm.numThreads=Math.min(4,Math.ceil((r||1)/2))}},Km=class{async init(e){Gm(),await jv(),await zv(e)}async createInferenceSessionHandler(e,r){let t=new Gv;return await t.loadModel(e,r),t}},Hv=new Km});Ii();Ii();Ii();var eE="1.23.0",tE=$0;{let e=(ZT(),gd(Kv)).wasmBackend;vn("webgpu",e,5),vn("webnn",e,5),vn("cpu",e,10),vn("wasm",e,10)}Object.defineProperty(dr.versions,"web",{value:eE,enumerable:!0});/**
* @license
* Copyright 2021 Google LLC. All Rights Reserved.
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
* =============================================================================
*//**
 * @license
 * Copyright 2020 Google LLC. All Rights Reserved.
 * Licensed under the Apache License, Version 2.0 (the "License");
 * you may not use this file except in compliance with the License.
 * You may obtain a copy of the License at
 *
 * http://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS,
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 * See the License for the specific language governing permissions and
 * limitations under the License.
 * =============================================================================
 *//**
 * @license
 * Copyright 2019 Google LLC. All Rights Reserved.
 * Licensed under the Apache License, Version 2.0 (the "License");
 * you may not use this file except in compliance with the License.
 * You may obtain a copy of the License at
 *
 * http://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS,
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 * See the License for the specific language governing permissions and
 * limitations under the License.
 * =============================================================================
 */var rE=Object.freeze({__proto__:null,get InferenceSession(){return Qm},get TRACE(){return yd},get TRACE_EVENT_BEGIN(){return zs},get TRACE_EVENT_END(){return Bs},get TRACE_FUNC_BEGIN(){return Xi},get TRACE_FUNC_END(){return ki},get Tensor(){return Qi},default:tE,get env(){return dr},get registerBackend(){return vn}}),Ms={},iE={"./node_modules/@huggingface/jinja/dist/index.js":(e,r,t)=>{t.r(r),t.d(r,{Environment:()=>Xe,Interpreter:()=>st,Template:()=>Ti,parse:()=>Q,tokenize:()=>m});var n=Object.freeze({Text:"Text",NumericLiteral:"NumericLiteral",BooleanLiteral:"BooleanLiteral",NullLiteral:"NullLiteral",StringLiteral:"StringLiteral",Identifier:"Identifier",Equals:"Equals",OpenParen:"OpenParen",CloseParen:"CloseParen",OpenStatement:"OpenStatement",CloseStatement:"CloseStatement",OpenExpression:"OpenExpression",CloseExpression:"CloseExpression",OpenSquareBracket:"OpenSquareBracket",CloseSquareBracket:"CloseSquareBracket",OpenCurlyBracket:"OpenCurlyBracket",CloseCurlyBracket:"CloseCurlyBracket",Comma:"Comma",Dot:"Dot",Colon:"Colon",Pipe:"Pipe",CallOperator:"CallOperator",AdditiveBinaryOperator:"AdditiveBinaryOperator",MultiplicativeBinaryOperator:"MultiplicativeBinaryOperator",ComparisonBinaryOperator:"ComparisonBinaryOperator",UnaryOperator:"UnaryOperator",Set:"Set",If:"If",For:"For",In:"In",Is:"Is",NotIn:"NotIn",Else:"Else",EndSet:"EndSet",EndIf:"EndIf",ElseIf:"ElseIf",EndFor:"EndFor",And:"And",Or:"Or",Not:"UnaryOperator",Macro:"Macro",EndMacro:"EndMacro",Break:"Break",Continue:"Continue"}),o=Object.freeze({set:n.Set,for:n.For,in:n.In,is:n.Is,if:n.If,else:n.Else,endset:n.EndSet,endif:n.EndIf,elif:n.ElseIf,endfor:n.EndFor,and:n.And,or:n.Or,not:n.Not,"not in":n.NotIn,macro:n.Macro,endmacro:n.EndMacro,break:n.Break,continue:n.Continue,true:n.BooleanLiteral,false:n.BooleanLiteral,none:n.NullLiteral,True:n.BooleanLiteral,False:n.BooleanLiteral,None:n.NullLiteral}),i=class{constructor(U,ge){this.value=U,this.type=ge}};function u(U){return/\w/.test(U)}function d(U){return/[0-9]/.test(U)}var h=[["{%",n.OpenStatement],["%}",n.CloseStatement],["{{",n.OpenExpression],["}}",n.CloseExpression],["(",n.OpenParen],[")",n.CloseParen],["{",n.OpenCurlyBracket],["}",n.CloseCurlyBracket],["[",n.OpenSquareBracket],["]",n.CloseSquareBracket],[",",n.Comma],[".",n.Dot],[":",n.Colon],["|",n.Pipe],["<=",n.ComparisonBinaryOperator],[">=",n.ComparisonBinaryOperator],["==",n.ComparisonBinaryOperator],["!=",n.ComparisonBinaryOperator],["<",n.ComparisonBinaryOperator],[">",n.ComparisonBinaryOperator],["+",n.AdditiveBinaryOperator],["-",n.AdditiveBinaryOperator],["*",n.MultiplicativeBinaryOperator],["/",n.MultiplicativeBinaryOperator],["%",n.MultiplicativeBinaryOperator],["=",n.Equals]],f=new Map([["n",`
`],["t","	"],["r","\r"],["b","\b"],["f","\f"],["v","\v"],["'","'"],['"','"'],["\\","\\"]]);function y(U,ge={}){return U.endsWith(`
`)&&(U=U.slice(0,-1)),U=U.replace(/{#.*?#}/gs,"{##}"),ge.lstrip_blocks&&(U=U.replace(/^[ \t]*({[#%])/gm,"$1")),ge.trim_blocks&&(U=U.replace(/([#%]})\n/g,"$1")),U.replace(/{##}/g,"").replace(/-%}\s*/g,"%}").replace(/\s*{%-/g,"{%").replace(/-}}\s*/g,"}}").replace(/\s*{{-/g,"{{")}function m(U,ge={}){var De,Dt,Ut;const K=[],ue=y(U,ge);let $e=0;const Ge=Et=>{let jt="";for(;Et(ue[$e]);){if(ue[$e]==="\\"){if(++$e,$e>=ue.length)throw new SyntaxError("Unexpected end of input");const gt=ue[$e++],Ht=f.get(gt);if(Ht===void 0)throw new SyntaxError(`Unexpected escaped character: ${gt}`);jt+=Ht;continue}if(jt+=ue[$e++],$e>=ue.length)throw new SyntaxError("Unexpected end of input")}return jt};e:for(;$e<ue.length;){const Et=(De=K.at(-1))==null?void 0:De.type;if(Et===void 0||Et===n.CloseStatement||Et===n.CloseExpression){let gt="";for(;$e<ue.length&&!(ue[$e]==="{"&&(ue[$e+1]==="%"||ue[$e+1]==="{"));)gt+=ue[$e++];if(gt.length>0){K.push(new i(gt,n.Text));continue}}Ge(gt=>/\s/.test(gt));const jt=ue[$e];if(jt==="-"||jt==="+"){const gt=(Dt=K.at(-1))==null?void 0:Dt.type;if(gt===n.Text||gt===void 0)throw new SyntaxError(`Unexpected character: ${jt}`);switch(gt){case n.Identifier:case n.NumericLiteral:case n.BooleanLiteral:case n.NullLiteral:case n.StringLiteral:case n.CloseParen:case n.CloseSquareBracket:break;default:{++$e;const Ht=Ge(d);K.push(new i(`${jt}${Ht}`,Ht.length>0?n.NumericLiteral:n.UnaryOperator));continue}}}for(const[gt,Ht]of h)if(ue.slice($e,$e+gt.length)===gt){K.push(new i(gt,Ht)),$e+=gt.length;continue e}if(jt==="'"||jt==='"'){++$e;const gt=Ge(Ht=>Ht!==jt);K.push(new i(gt,n.StringLiteral)),++$e;continue}if(d(jt)){const gt=Ge(d);K.push(new i(gt,n.NumericLiteral));continue}if(u(jt)){const gt=Ge(u),Ht=Object.hasOwn(o,gt)?o[gt]:n.Identifier;Ht===n.In&&((Ut=K.at(-1))==null?void 0:Ut.type)===n.Not?(K.pop(),K.push(new i("not in",n.NotIn))):K.push(new i(gt,Ht));continue}throw new SyntaxError(`Unexpected character: ${jt}`)}return K}var g=class{constructor(){le(this,"type","Statement")}},v=class extends g{constructor(ge){super();le(this,"type","Program");this.body=ge}},b=class extends g{constructor(ge,K,ue){super();le(this,"type","If");this.test=ge,this.body=K,this.alternate=ue}},k=class extends g{constructor(ge,K,ue,$e){super();le(this,"type","For");this.loopvar=ge,this.iterable=K,this.body=ue,this.defaultBlock=$e}},z=class extends g{constructor(){super(...arguments);le(this,"type","Break")}},E=class extends g{constructor(){super(...arguments);le(this,"type","Continue")}},T=class extends g{constructor(ge,K,ue){super();le(this,"type","Set");this.assignee=ge,this.value=K,this.body=ue}},D=class extends g{constructor(ge,K,ue){super();le(this,"type","Macro");this.name=ge,this.args=K,this.body=ue}},I=class extends g{constructor(){super(...arguments);le(this,"type","Expression")}},$=class extends I{constructor(ge,K,ue){super();le(this,"type","MemberExpression");this.object=ge,this.property=K,this.computed=ue}},P=class extends I{constructor(ge,K){super();le(this,"type","CallExpression");this.callee=ge,this.args=K}},S=class extends I{constructor(ge){super();le(this,"type","Identifier");this.value=ge}},O=class extends I{constructor(ge){super();le(this,"type","Literal");this.value=ge}},R=class extends O{constructor(){super(...arguments);le(this,"type","NumericLiteral")}},G=class extends O{constructor(){super(...arguments);le(this,"type","StringLiteral")}},ee=class extends O{constructor(){super(...arguments);le(this,"type","BooleanLiteral")}},ie=class extends O{constructor(){super(...arguments);le(this,"type","NullLiteral")}},H=class extends O{constructor(){super(...arguments);le(this,"type","ArrayLiteral")}},ce=class extends O{constructor(){super(...arguments);le(this,"type","TupleLiteral")}},re=class extends O{constructor(){super(...arguments);le(this,"type","ObjectLiteral")}},se=class extends I{constructor(ge,K,ue){super();le(this,"type","BinaryExpression");this.operator=ge,this.left=K,this.right=ue}},_e=class extends I{constructor(ge,K){super();le(this,"type","FilterExpression");this.operand=ge,this.filter=K}},ae=class extends I{constructor(ge,K){super();le(this,"type","SelectExpression");this.iterable=ge,this.test=K}},Ce=class extends I{constructor(ge,K,ue){super();le(this,"type","TestExpression");this.operand=ge,this.negate=K,this.test=ue}},Te=class extends I{constructor(ge,K){super();le(this,"type","UnaryExpression");this.operator=ge,this.argument=K}},q=class extends I{constructor(ge=void 0,K=void 0,ue=void 0){super();le(this,"type","SliceExpression");this.start=ge,this.stop=K,this.step=ue}},B=class extends I{constructor(ge,K){super();le(this,"type","KeywordArgumentExpression");this.key=ge,this.value=K}};function Q(U){const ge=new v([]);let K=0;function ue(qe,lt){const yt=U[K++];if(!yt||yt.type!==qe)throw new Error(`Parser Error: ${lt}. ${yt.type} !== ${qe}.`);return yt}function $e(){switch(U[K].type){case n.Text:return Dt();case n.OpenStatement:return Ut();case n.OpenExpression:return Et();default:throw new SyntaxError(`Unexpected token type: ${U[K].type}`)}}function Ge(...qe){return K+qe.length<=U.length&&qe.some((lt,yt)=>lt!==U[K+yt].type)}function De(...qe){return K+qe.length<=U.length&&qe.every((lt,yt)=>lt===U[K+yt].type)}function Dt(){return new G(ue(n.Text,"Expected text token").value)}function Ut(){ue(n.OpenStatement,"Expected opening statement token");let qe;switch(U[K].type){case n.Set:++K,qe=jt(),ue(n.CloseStatement,"Expected closing statement token");break;case n.If:++K,qe=gt(),ue(n.OpenStatement,"Expected {% token"),ue(n.EndIf,"Expected endif token"),ue(n.CloseStatement,"Expected %} token");break;case n.Macro:++K,qe=Ht(),ue(n.OpenStatement,"Expected {% token"),ue(n.EndMacro,"Expected endmacro token"),ue(n.CloseStatement,"Expected %} token");break;case n.For:++K,qe=ri(),ue(n.OpenStatement,"Expected {% token"),ue(n.EndFor,"Expected endfor token"),ue(n.CloseStatement,"Expected %} token");break;case n.Break:++K,ue(n.CloseStatement,"Expected closing statement token"),qe=new z;break;case n.Continue:++K,ue(n.CloseStatement,"Expected closing statement token"),qe=new E;break;default:throw new SyntaxError(`Unknown statement type: ${U[K].type}`)}return qe}function Et(){ue(n.OpenExpression,"Expected opening expression token");const qe=pr();return ue(n.CloseExpression,"Expected closing expression token"),qe}function jt(){var lt,yt;const qe=pr();if(De(n.Equals)){++K;const ur=pr();return new T(qe,ur,[])}else{const ur=[];for(ue(n.CloseStatement,"Expected %} token");!(((lt=U[K])==null?void 0:lt.type)===n.OpenStatement&&((yt=U[K+1])==null?void 0:yt.type)===n.EndSet);){const gi=$e();ur.push(gi)}return ue(n.OpenStatement,"Expected {% token"),ue(n.EndSet,"Expected endset token"),new T(qe,null,ur)}}function gt(){var ur,gi,Di,ni,Li,ji,zi,qr;const qe=pr();ue(n.CloseStatement,"Expected closing statement token");const lt=[],yt=[];for(;!(((ur=U[K])==null?void 0:ur.type)===n.OpenStatement&&(((gi=U[K+1])==null?void 0:gi.type)===n.ElseIf||((Di=U[K+1])==null?void 0:Di.type)===n.Else||((ni=U[K+1])==null?void 0:ni.type)===n.EndIf));)lt.push($e());if(((Li=U[K])==null?void 0:Li.type)===n.OpenStatement&&((ji=U[K+1])==null?void 0:ji.type)!==n.EndIf)if(++K,De(n.ElseIf))ue(n.ElseIf,"Expected elseif token"),yt.push(gt());else for(ue(n.Else,"Expected else token"),ue(n.CloseStatement,"Expected closing statement token");!(((zi=U[K])==null?void 0:zi.type)===n.OpenStatement&&((qr=U[K+1])==null?void 0:qr.type)===n.EndIf);)yt.push($e());return new b(qe,lt,yt)}function Ht(){const qe=si();if(qe.type!=="Identifier")throw new SyntaxError("Expected identifier following macro statement");const lt=Fi();ue(n.CloseStatement,"Expected closing statement token");const yt=[];for(;Ge(n.OpenStatement,n.EndMacro);)yt.push($e());return new D(qe,lt,yt)}function nr(qe=!1){const lt=qe?si:pr,yt=[lt()],ur=De(n.Comma);for(;ur&&(++K,yt.push(lt()),!!De(n.Comma)););return ur?new ce(yt):yt[0]}function ri(){const qe=nr(!0);if(!(qe instanceof S||qe instanceof ce))throw new SyntaxError(`Expected identifier/tuple for the loop variable, got ${qe.type} instead`);ue(n.In,"Expected `in` keyword following loop variable");const lt=pr();ue(n.CloseStatement,"Expected closing statement token");const yt=[];for(;Ge(n.OpenStatement,n.EndFor)&&Ge(n.OpenStatement,n.Else);)yt.push($e());const ur=[];if(De(n.OpenStatement,n.Else))for(++K,++K,ue(n.CloseStatement,"Expected closing statement token");Ge(n.OpenStatement,n.EndFor);)ur.push($e());return new k(qe,lt,yt,ur)}function pr(){return Jr()}function Jr(){const qe=Ai();if(De(n.If)){++K;const lt=Ai();if(De(n.Else)){++K;const yt=Ai();return new b(lt,[qe],[yt])}else return new ae(qe,lt)}return qe}function Ai(){let qe=Ji();for(;De(n.Or);){const lt=U[K];++K;const yt=Ji();qe=new se(lt,qe,yt)}return qe}function Ji(){let qe=Oi();for(;De(n.And);){const lt=U[K];++K;const yt=Oi();qe=new se(lt,qe,yt)}return qe}function Oi(){let qe;for(;De(n.Not);){const lt=U[K];++K;const yt=Oi();qe=new Te(lt,yt)}return qe??ns()}function ns(){let qe=ii();for(;De(n.ComparisonBinaryOperator)||De(n.In)||De(n.NotIn);){const lt=U[K];++K;const yt=ii();qe=new se(lt,qe,yt)}return qe}function ii(){let qe=Kr();for(;De(n.AdditiveBinaryOperator);){const lt=U[K];++K;const yt=Kr();qe=new se(lt,qe,yt)}return qe}function Yi(){const qe=Gr(si());return De(n.OpenParen)?_i(qe):qe}function _i(qe){let lt=new P(qe,Fi());return lt=Gr(lt),De(n.OpenParen)&&(lt=_i(lt)),lt}function Fi(){ue(n.OpenParen,"Expected opening parenthesis for arguments list");const qe=er();return ue(n.CloseParen,"Expected closing parenthesis for arguments list"),qe}function er(){const qe=[];for(;!De(n.CloseParen);){let lt=pr();if(De(n.Equals)){if(++K,!(lt instanceof S))throw new SyntaxError("Expected identifier for keyword argument");const yt=pr();lt=new B(lt,yt)}qe.push(lt),De(n.Comma)&&++K}return qe}function Zi(){const qe=[];let lt=!1;for(;!De(n.CloseSquareBracket);)De(n.Colon)?(qe.push(void 0),++K,lt=!0):(qe.push(pr()),De(n.Colon)&&(++K,lt=!0));if(qe.length===0)throw new SyntaxError("Expected at least one argument for member/slice expression");if(lt){if(qe.length>3)throw new SyntaxError("Expected 0-3 arguments for slice expression");return new q(...qe)}return qe[0]}function Gr(qe){for(;De(n.Dot)||De(n.OpenSquareBracket);){const lt=U[K];++K;let yt;const ur=lt.type!==n.Dot;if(ur)yt=Zi(),ue(n.CloseSquareBracket,"Expected closing square bracket");else if(yt=si(),yt.type!=="Identifier")throw new SyntaxError("Expected identifier following dot operator");qe=new $(qe,yt,ur)}return qe}function Kr(){let qe=Hr();for(;De(n.MultiplicativeBinaryOperator);){const lt=U[K];++K;const yt=Hr();qe=new se(lt,qe,yt)}return qe}function Hr(){let qe=es();for(;De(n.Is);){++K;const lt=De(n.Not);lt&&++K;let yt=si();if(yt instanceof ee?yt=new S(yt.value.toString()):yt instanceof ie&&(yt=new S("none")),!(yt instanceof S))throw new SyntaxError("Expected identifier for the test");qe=new Ce(qe,lt,yt)}return qe}function es(){let qe=Yi();for(;De(n.Pipe);){++K;let lt=si();if(!(lt instanceof S))throw new SyntaxError("Expected identifier for the filter");De(n.OpenParen)&&(lt=_i(lt)),qe=new _e(qe,lt)}return qe}function si(){const qe=U[K];switch(qe.type){case n.NumericLiteral:return++K,new R(Number(qe.value));case n.StringLiteral:return++K,new G(qe.value);case n.BooleanLiteral:return++K,new ee(qe.value.toLowerCase()==="true");case n.NullLiteral:return++K,new ie(null);case n.Identifier:return++K,new S(qe.value);case n.OpenParen:{++K;const lt=nr();if(U[K].type!==n.CloseParen)throw new SyntaxError(`Expected closing parenthesis, got ${U[K].type} instead`);return++K,lt}case n.OpenSquareBracket:{++K;const lt=[];for(;!De(n.CloseSquareBracket);)lt.push(pr()),De(n.Comma)&&++K;return++K,new H(lt)}case n.OpenCurlyBracket:{++K;const lt=new Map;for(;!De(n.CloseCurlyBracket);){const yt=pr();ue(n.Colon,"Expected colon between key and value in object literal");const ur=pr();lt.set(yt,ur),De(n.Comma)&&++K}return++K,new re(lt)}default:throw new SyntaxError(`Unexpected token: ${qe.type}`)}}for(;K<U.length;)ge.body.push($e());return ge}function oe(U,ge,K=1){ge===void 0&&(ge=U,U=0);const ue=[];for(let $e=U;$e<ge;$e+=K)ue.push($e);return ue}function Ie(U,ge,K,ue=1){const $e=Math.sign(ue);$e>=0?(ge=(ge??(ge=0))<0?Math.max(U.length+ge,0):Math.min(ge,U.length),K=(K??(K=U.length))<0?Math.max(U.length+K,0):Math.min(K,U.length)):(ge=(ge??(ge=U.length-1))<0?Math.max(U.length+ge,-1):Math.min(ge,U.length-1),K=(K??(K=-1))<-1?Math.max(U.length+K,-1):Math.min(K,U.length-1));const Ge=[];for(let De=ge;$e*De<$e*K;De+=ue)Ge.push(U[De]);return Ge}function ye(U){return U.replace(/\b\w/g,ge=>ge.toUpperCase())}var be=class extends Error{},et=class extends Error{},tt=class{constructor(U=void 0){le(this,"type","RuntimeValue");le(this,"value");le(this,"builtins",new Map);this.value=U}__bool__(){return new pe(!!this.value)}},Ee=class extends tt{constructor(){super(...arguments);le(this,"type","NumericValue")}},Y=class extends tt{constructor(){super(...arguments);le(this,"type","StringValue");le(this,"builtins",new Map([["upper",new Be(()=>new Y(this.value.toUpperCase()))],["lower",new Be(()=>new Y(this.value.toLowerCase()))],["strip",new Be(()=>new Y(this.value.trim()))],["title",new Be(()=>new Y(ye(this.value)))],["length",new Ee(this.value.length)],["rstrip",new Be(()=>new Y(this.value.trimEnd()))],["lstrip",new Be(()=>new Y(this.value.trimStart()))],["startswith",new Be(ge=>{if(ge.length===0)throw new Error("startswith() requires at least one argument");const K=ge[0];if(!(K instanceof Y))throw new Error("startswith() argument must be a string");return new pe(this.value.startsWith(K.value))})],["endswith",new Be(ge=>{if(ge.length===0)throw new Error("endswith() requires at least one argument");const K=ge[0];if(!(K instanceof Y))throw new Error("endswith() argument must be a string");return new pe(this.value.endsWith(K.value))})],["split",new Be(ge=>{const K=ge[0]??new Ve;if(!(K instanceof Y||K instanceof Ve))throw new Error("sep argument must be a string or null");const ue=ge[1]??new Ee(-1);if(!(ue instanceof Ee))throw new Error("maxsplit argument must be a number");let $e=[];if(K instanceof Ve){const Ge=this.value.trimStart();for(const{0:De,index:Dt}of Ge.matchAll(/\S+/g)){if(ue.value!==-1&&$e.length>=ue.value&&Dt!==void 0){$e.push(De+Ge.slice(Dt+De.length));break}$e.push(De)}}else{if(K.value==="")throw new Error("empty separator");$e=this.value.split(K.value),ue.value!==-1&&$e.length>ue.value&&$e.push($e.splice(ue.value).join(K.value))}return new Oe($e.map(Ge=>new Y(Ge)))})]]))}},pe=class extends tt{constructor(){super(...arguments);le(this,"type","BooleanValue")}},X=class extends tt{constructor(){super(...arguments);le(this,"type","ObjectValue");le(this,"builtins",new Map([["get",new Be(([ge,K])=>{if(!(ge instanceof Y))throw new Error(`Object key must be a string: got ${ge.type}`);return this.value.get(ge.value)??K??new Ve})],["items",new Be(()=>new Oe(Array.from(this.value.entries()).map(([ge,K])=>new Oe([new Y(ge),K]))))]]))}__bool__(){return new pe(this.value.size>0)}},Ae=class extends X{constructor(){super(...arguments);le(this,"type","KeywordArgumentsValue")}},Oe=class extends tt{constructor(){super(...arguments);le(this,"type","ArrayValue");le(this,"builtins",new Map([["length",new Ee(this.value.length)]]))}__bool__(){return new pe(this.value.length>0)}},Ne=class extends Oe{constructor(){super(...arguments);le(this,"type","TupleValue")}},Be=class extends tt{constructor(){super(...arguments);le(this,"type","FunctionValue")}},Ve=class extends tt{constructor(){super(...arguments);le(this,"type","NullValue")}},He=class extends tt{constructor(){super(...arguments);le(this,"type","UndefinedValue")}},Xe=class{constructor(U){le(this,"variables",new Map([["namespace",new Be(U=>{if(U.length===0)return new X(new Map);if(U.length!==1||!(U[0]instanceof X))throw new Error("`namespace` expects either zero arguments or a single object argument");return U[0]})]]));le(this,"tests",new Map([["boolean",U=>U.type==="BooleanValue"],["callable",U=>U instanceof Be],["odd",U=>{if(U.type!=="NumericValue")throw new Error(`Cannot apply test "odd" to type: ${U.type}`);return U.value%2!==0}],["even",U=>{if(U.type!=="NumericValue")throw new Error(`Cannot apply test "even" to type: ${U.type}`);return U.value%2===0}],["false",U=>U.type==="BooleanValue"&&!U.value],["true",U=>U.type==="BooleanValue"&&U.value],["none",U=>U.type==="NullValue"],["string",U=>U.type==="StringValue"],["number",U=>U.type==="NumericValue"],["integer",U=>U.type==="NumericValue"&&Number.isInteger(U.value)],["iterable",U=>U.type==="ArrayValue"||U.type==="StringValue"],["mapping",U=>U.type==="ObjectValue"],["lower",U=>{const ge=U.value;return U.type==="StringValue"&&ge===ge.toLowerCase()}],["upper",U=>{const ge=U.value;return U.type==="StringValue"&&ge===ge.toUpperCase()}],["none",U=>U.type==="NullValue"],["defined",U=>U.type!=="UndefinedValue"],["undefined",U=>U.type==="UndefinedValue"],["equalto",(U,ge)=>U.value===ge.value],["eq",(U,ge)=>U.value===ge.value]]));this.parent=U}set(U,ge){return this.declareVariable(U,nt(ge))}declareVariable(U,ge){if(this.variables.has(U))throw new SyntaxError(`Variable already declared: ${U}`);return this.variables.set(U,ge),ge}setVariable(U,ge){return this.variables.set(U,ge),ge}resolve(U){if(this.variables.has(U))return this;if(this.parent)return this.parent.resolve(U);throw new Error(`Unknown variable: ${U}`)}lookupVariable(U){try{return this.resolve(U).variables.get(U)??new He}catch{return new He}}},st=class{constructor(U){le(this,"global");this.global=U??new Xe}run(U){return this.evaluate(U,this.global)}evaluateBinaryExpression(U,ge){const K=this.evaluate(U.left,ge);switch(U.operator.value){case"and":return K.__bool__().value?this.evaluate(U.right,ge):K;case"or":return K.__bool__().value?K:this.evaluate(U.right,ge)}const ue=this.evaluate(U.right,ge);switch(U.operator.value){case"==":return new pe(K.value==ue.value);case"!=":return new pe(K.value!=ue.value)}if(K instanceof He||ue instanceof He)throw new Error("Cannot perform operation on undefined values");if(K instanceof Ve||ue instanceof Ve)throw new Error("Cannot perform operation on null values");if(K instanceof Ee&&ue instanceof Ee)switch(U.operator.value){case"+":return new Ee(K.value+ue.value);case"-":return new Ee(K.value-ue.value);case"*":return new Ee(K.value*ue.value);case"/":return new Ee(K.value/ue.value);case"%":return new Ee(K.value%ue.value);case"<":return new pe(K.value<ue.value);case">":return new pe(K.value>ue.value);case">=":return new pe(K.value>=ue.value);case"<=":return new pe(K.value<=ue.value)}else if(K instanceof Oe&&ue instanceof Oe)switch(U.operator.value){case"+":return new Oe(K.value.concat(ue.value))}else if(ue instanceof Oe){const $e=ue.value.find(Ge=>Ge.value===K.value)!==void 0;switch(U.operator.value){case"in":return new pe($e);case"not in":return new pe(!$e)}}if(K instanceof Y||ue instanceof Y)switch(U.operator.value){case"+":return new Y(K.value.toString()+ue.value.toString())}if(K instanceof Y&&ue instanceof Y)switch(U.operator.value){case"in":return new pe(ue.value.includes(K.value));case"not in":return new pe(!ue.value.includes(K.value))}if(K instanceof Y&&ue instanceof X)switch(U.operator.value){case"in":return new pe(ue.value.has(K.value));case"not in":return new pe(!ue.value.has(K.value))}throw new SyntaxError(`Unknown operator "${U.operator.value}" between ${K.type} and ${ue.type}`)}evaluateArguments(U,ge){const K=[],ue=new Map;for(const $e of U)if($e.type==="KeywordArgumentExpression"){const Ge=$e;ue.set(Ge.key.value,this.evaluate(Ge.value,ge))}else{if(ue.size>0)throw new Error("Positional arguments must come before keyword arguments");K.push(this.evaluate($e,ge))}return[K,ue]}evaluateFilterExpression(U,ge){const K=this.evaluate(U.operand,ge);if(U.filter.type==="Identifier"){const ue=U.filter;if(ue.value==="tojson")return new Y(We(K));if(K instanceof Oe)switch(ue.value){case"list":return K;case"first":return K.value[0];case"last":return K.value[K.value.length-1];case"length":return new Ee(K.value.length);case"reverse":return new Oe(K.value.reverse());case"sort":return new Oe(K.value.sort(($e,Ge)=>{if($e.type!==Ge.type)throw new Error(`Cannot compare different types: ${$e.type} and ${Ge.type}`);switch($e.type){case"NumericValue":return $e.value-Ge.value;case"StringValue":return $e.value.localeCompare(Ge.value);default:throw new Error(`Cannot compare type: ${$e.type}`)}}));case"join":return new Y(K.value.map($e=>$e.value).join(""));case"string":return new Y(We(K));default:throw new Error(`Unknown ArrayValue filter: ${ue.value}`)}else if(K instanceof Y)switch(ue.value){case"length":return new Ee(K.value.length);case"upper":return new Y(K.value.toUpperCase());case"lower":return new Y(K.value.toLowerCase());case"title":return new Y(ye(K.value));case"capitalize":return new Y(K.value.charAt(0).toUpperCase()+K.value.slice(1));case"trim":return new Y(K.value.trim());case"indent":return new Y(K.value.split(`
`).map(($e,Ge)=>Ge===0||$e.length===0?$e:"    "+$e).join(`
`));case"join":case"string":return K;default:throw new Error(`Unknown StringValue filter: ${ue.value}`)}else if(K instanceof Ee)switch(ue.value){case"abs":return new Ee(Math.abs(K.value));default:throw new Error(`Unknown NumericValue filter: ${ue.value}`)}else if(K instanceof X)switch(ue.value){case"items":return new Oe(Array.from(K.value.entries()).map(([$e,Ge])=>new Oe([new Y($e),Ge])));case"length":return new Ee(K.value.size);default:throw new Error(`Unknown ObjectValue filter: ${ue.value}`)}throw new Error(`Cannot apply filter "${ue.value}" to type: ${K.type}`)}else if(U.filter.type==="CallExpression"){const ue=U.filter;if(ue.callee.type!=="Identifier")throw new Error(`Unknown filter: ${ue.callee.type}`);const $e=ue.callee.value;if($e==="tojson"){const[,Ge]=this.evaluateArguments(ue.args,ge),De=Ge.get("indent")??new Ve;if(!(De instanceof Ee||De instanceof Ve))throw new Error("If set, indent must be a number");return new Y(We(K,De.value))}else if($e==="join"){let Ge;if(K instanceof Y)Ge=Array.from(K.value);else if(K instanceof Oe)Ge=K.value.map(Et=>Et.value);else throw new Error(`Cannot apply filter "${$e}" to type: ${K.type}`);const[De,Dt]=this.evaluateArguments(ue.args,ge),Ut=De.at(0)??Dt.get("separator")??new Y("");if(!(Ut instanceof Y))throw new Error("separator must be a string");return new Y(Ge.join(Ut.value))}if(K instanceof Oe){switch($e){case"selectattr":case"rejectattr":{const Ge=$e==="selectattr";if(K.value.some(gt=>!(gt instanceof X)))throw new Error(`\`${$e}\` can only be applied to array of objects`);if(ue.args.some(gt=>gt.type!=="StringLiteral"))throw new Error(`arguments of \`${$e}\` must be strings`);const[De,Dt,Ut]=ue.args.map(gt=>this.evaluate(gt,ge));let Et;if(Dt){const gt=ge.tests.get(Dt.value);if(!gt)throw new Error(`Unknown test: ${Dt.value}`);Et=gt}else Et=(...gt)=>gt[0].__bool__().value;const jt=K.value.filter(gt=>{const Ht=gt.value.get(De.value),nr=Ht?Et(Ht,Ut):!1;return Ge?nr:!nr});return new Oe(jt)}case"map":{const[,Ge]=this.evaluateArguments(ue.args,ge);if(Ge.has("attribute")){const De=Ge.get("attribute");if(!(De instanceof Y))throw new Error("attribute must be a string");const Dt=Ge.get("default"),Ut=K.value.map(Et=>{if(!(Et instanceof X))throw new Error("items in map must be an object");return Et.value.get(De.value)??Dt??new He});return new Oe(Ut)}else throw new Error("`map` expressions without `attribute` set are not currently supported.")}}throw new Error(`Unknown ArrayValue filter: ${$e}`)}else if(K instanceof Y){switch($e){case"indent":{const[Ge,De]=this.evaluateArguments(ue.args,ge),Dt=Ge.at(0)??De.get("width")??new Ee(4);if(!(Dt instanceof Ee))throw new Error("width must be a number");const Ut=Ge.at(1)??De.get("first")??new pe(!1),Et=Ge.at(2)??De.get("blank")??new pe(!1),jt=K.value.split(`
`),gt=" ".repeat(Dt.value),Ht=jt.map((nr,ri)=>!Ut.value&&ri===0||!Et.value&&nr.length===0?nr:gt+nr);return new Y(Ht.join(`
`))}}throw new Error(`Unknown StringValue filter: ${$e}`)}else throw new Error(`Cannot apply filter "${$e}" to type: ${K.type}`)}throw new Error(`Unknown filter: ${U.filter.type}`)}evaluateTestExpression(U,ge){const K=this.evaluate(U.operand,ge),ue=ge.tests.get(U.test.value);if(!ue)throw new Error(`Unknown test: ${U.test.value}`);const $e=ue(K);return new pe(U.negate?!$e:$e)}evaluateUnaryExpression(U,ge){const K=this.evaluate(U.argument,ge);switch(U.operator.value){case"not":return new pe(!K.value);default:throw new SyntaxError(`Unknown operator: ${U.operator.value}`)}}evalProgram(U,ge){return this.evaluateBlock(U.body,ge)}evaluateBlock(U,ge){let K="";for(const ue of U){const $e=this.evaluate(ue,ge);$e.type!=="NullValue"&&$e.type!=="UndefinedValue"&&(K+=$e.value)}return new Y(K)}evaluateIdentifier(U,ge){return ge.lookupVariable(U.value)}evaluateCallExpression(U,ge){const[K,ue]=this.evaluateArguments(U.args,ge);ue.size>0&&K.push(new Ae(ue));const $e=this.evaluate(U.callee,ge);if($e.type!=="FunctionValue")throw new Error(`Cannot call something that is not a function: got ${$e.type}`);return $e.value(K,ge)}evaluateSliceExpression(U,ge,K){if(!(U instanceof Oe||U instanceof Y))throw new Error("Slice object must be an array or string");const ue=this.evaluate(ge.start,K),$e=this.evaluate(ge.stop,K),Ge=this.evaluate(ge.step,K);if(!(ue instanceof Ee||ue instanceof He))throw new Error("Slice start must be numeric or undefined");if(!($e instanceof Ee||$e instanceof He))throw new Error("Slice stop must be numeric or undefined");if(!(Ge instanceof Ee||Ge instanceof He))throw new Error("Slice step must be numeric or undefined");return U instanceof Oe?new Oe(Ie(U.value,ue.value,$e.value,Ge.value)):new Y(Ie(Array.from(U.value),ue.value,$e.value,Ge.value).join(""))}evaluateMemberExpression(U,ge){const K=this.evaluate(U.object,ge);let ue;if(U.computed){if(U.property.type==="SliceExpression")return this.evaluateSliceExpression(K,U.property,ge);ue=this.evaluate(U.property,ge)}else ue=new Y(U.property.value);let $e;if(K instanceof X){if(!(ue instanceof Y))throw new Error(`Cannot access property with non-string: got ${ue.type}`);$e=K.value.get(ue.value)??K.builtins.get(ue.value)}else if(K instanceof Oe||K instanceof Y)if(ue instanceof Ee)$e=K.value.at(ue.value),K instanceof Y&&($e=new Y(K.value.at(ue.value)));else if(ue instanceof Y)$e=K.builtins.get(ue.value);else throw new Error(`Cannot access property with non-string/non-number: got ${ue.type}`);else{if(!(ue instanceof Y))throw new Error(`Cannot access property with non-string: got ${ue.type}`);$e=K.builtins.get(ue.value)}return $e instanceof tt?$e:new He}evaluateSet(U,ge){const K=U.value?this.evaluate(U.value,ge):this.evaluateBlock(U.body,ge);if(U.assignee.type==="Identifier"){const ue=U.assignee.value;ge.setVariable(ue,K)}else if(U.assignee.type==="MemberExpression"){const ue=U.assignee,$e=this.evaluate(ue.object,ge);if(!($e instanceof X))throw new Error("Cannot assign to member of non-object");if(ue.property.type!=="Identifier")throw new Error("Cannot assign to member with non-identifier property");$e.value.set(ue.property.value,K)}else throw new Error(`Invalid LHS inside assignment expression: ${JSON.stringify(U.assignee)}`);return new Ve}evaluateIf(U,ge){const K=this.evaluate(U.test,ge);return this.evaluateBlock(K.__bool__().value?U.body:U.alternate,ge)}evaluateFor(U,ge){const K=new Xe(ge);let ue,$e;if(U.iterable.type==="SelectExpression"){const Et=U.iterable;$e=this.evaluate(Et.iterable,K),ue=Et.test}else $e=this.evaluate(U.iterable,K);if(!($e instanceof Oe))throw new Error(`Expected iterable type in for loop: got ${$e.type}`);const Ge=[],De=[];for(let Et=0;Et<$e.value.length;++Et){const jt=new Xe(K),gt=$e.value[Et];let Ht;if(U.loopvar.type==="Identifier")Ht=nr=>nr.setVariable(U.loopvar.value,gt);else if(U.loopvar.type==="TupleLiteral"){const nr=U.loopvar;if(gt.type!=="ArrayValue")throw new Error(`Cannot unpack non-iterable type: ${gt.type}`);const ri=gt;if(nr.value.length!==ri.value.length)throw new Error(`Too ${nr.value.length>ri.value.length?"few":"many"} items to unpack`);Ht=pr=>{for(let Jr=0;Jr<nr.value.length;++Jr){if(nr.value[Jr].type!=="Identifier")throw new Error(`Cannot unpack non-identifier type: ${nr.value[Jr].type}`);pr.setVariable(nr.value[Jr].value,ri.value[Jr])}}}else throw new Error(`Invalid loop variable(s): ${U.loopvar.type}`);ue&&(Ht(jt),!this.evaluate(ue,jt).__bool__().value)||(Ge.push(gt),De.push(Ht))}let Dt="",Ut=!0;for(let Et=0;Et<Ge.length;++Et){const jt=new Map([["index",new Ee(Et+1)],["index0",new Ee(Et)],["revindex",new Ee(Ge.length-Et)],["revindex0",new Ee(Ge.length-Et-1)],["first",new pe(Et===0)],["last",new pe(Et===Ge.length-1)],["length",new Ee(Ge.length)],["previtem",Et>0?Ge[Et-1]:new He],["nextitem",Et<Ge.length-1?Ge[Et+1]:new He]]);K.setVariable("loop",new X(jt)),De[Et](K);try{const gt=this.evaluateBlock(U.body,K);Dt+=gt.value}catch(gt){if(gt instanceof et)continue;if(gt instanceof be)break;throw gt}Ut=!1}if(Ut){const Et=this.evaluateBlock(U.defaultBlock,K);Dt+=Et.value}return new Y(Dt)}evaluateMacro(U,ge){return ge.setVariable(U.name.value,new Be((K,ue)=>{var De;const $e=new Xe(ue);K=K.slice();let Ge;((De=K.at(-1))==null?void 0:De.type)==="KeywordArgumentsValue"&&(Ge=K.pop());for(let Dt=0;Dt<U.args.length;++Dt){const Ut=U.args[Dt],Et=K[Dt];if(Ut.type==="Identifier"){const jt=Ut;if(!Et)throw new Error(`Missing positional argument: ${jt.value}`);$e.setVariable(jt.value,Et)}else if(Ut.type==="KeywordArgumentExpression"){const jt=Ut,gt=Et??(Ge==null?void 0:Ge.value.get(jt.key.value))??this.evaluate(jt.value,$e);$e.setVariable(jt.key.value,gt)}else throw new Error(`Unknown argument type: ${Ut.type}`)}return this.evaluateBlock(U.body,$e)})),new Ve}evaluate(U,ge){if(U===void 0)return new He;switch(U.type){case"Program":return this.evalProgram(U,ge);case"Set":return this.evaluateSet(U,ge);case"If":return this.evaluateIf(U,ge);case"For":return this.evaluateFor(U,ge);case"Macro":return this.evaluateMacro(U,ge);case"Break":throw new be;case"Continue":throw new et;case"NumericLiteral":return new Ee(Number(U.value));case"StringLiteral":return new Y(U.value);case"BooleanLiteral":return new pe(U.value);case"NullLiteral":return new Ve(U.value);case"ArrayLiteral":return new Oe(U.value.map(K=>this.evaluate(K,ge)));case"TupleLiteral":return new Ne(U.value.map(K=>this.evaluate(K,ge)));case"ObjectLiteral":{const K=new Map;for(const[ue,$e]of U.value){const Ge=this.evaluate(ue,ge);if(!(Ge instanceof Y))throw new Error(`Object keys must be strings: got ${Ge.type}`);K.set(Ge.value,this.evaluate($e,ge))}return new X(K)}case"Identifier":return this.evaluateIdentifier(U,ge);case"CallExpression":return this.evaluateCallExpression(U,ge);case"MemberExpression":return this.evaluateMemberExpression(U,ge);case"UnaryExpression":return this.evaluateUnaryExpression(U,ge);case"BinaryExpression":return this.evaluateBinaryExpression(U,ge);case"FilterExpression":return this.evaluateFilterExpression(U,ge);case"TestExpression":return this.evaluateTestExpression(U,ge);default:throw new SyntaxError(`Unknown node type: ${U.type}`)}}};function nt(U){switch(typeof U){case"number":return new Ee(U);case"string":return new Y(U);case"boolean":return new pe(U);case"undefined":return new He;case"object":return U===null?new Ve:Array.isArray(U)?new Oe(U.map(nt)):new X(new Map(Object.entries(U).map(([ge,K])=>[ge,nt(K)])));case"function":return new Be((ge,K)=>{const ue=U(...ge.map($e=>$e.value))??null;return nt(ue)});default:throw new Error(`Cannot convert to runtime value: ${U}`)}}function We(U,ge,K){const ue=K??0;switch(U.type){case"NullValue":case"UndefinedValue":return"null";case"NumericValue":case"StringValue":case"BooleanValue":return JSON.stringify(U.value);case"ArrayValue":case"ObjectValue":{const $e=ge?" ".repeat(ge):"",Ge=`
`+$e.repeat(ue),De=Ge+$e;if(U.type==="ArrayValue"){const Dt=U.value.map(Ut=>We(Ut,ge,ue+1));return ge?`[${De}${Dt.join(`,${De}`)}${Ge}]`:`[${Dt.join(", ")}]`}else{const Dt=Array.from(U.value.entries()).map(([Ut,Et])=>{const jt=`"${Ut}": ${We(Et,ge,ue+1)}`;return ge?`${De}${jt}`:jt});return ge?`{${Dt.join(",")}${Ge}}`:`{${Dt.join(", ")}}`}}default:throw new Error(`Cannot convert to JSON: ${U.type}`)}}var rt=`
`,pt="{%- ",_t=" -%}",ct={MultiplicativeBinaryOperator:2,AdditiveBinaryOperator:1,ComparisonBinaryOperator:0};function dt(U,ge="	"){const K=typeof ge=="number"?" ".repeat(ge):ge;return xt(U.body,0,K).replace(/\n$/,"")}function Le(...U){return pt+U.join(" ")+_t}function xt(U,ge,K){return U.map(ue=>Nt(ue,ge,K)).join(rt)}function Nt(U,ge,K){const ue=K.repeat(ge);switch(U.type){case"Program":return xt(U.body,ge,K);case"If":return cr(U,ge,K);case"For":return Kt(U,ge,K);case"Set":return wr(U,ge,K);case"Macro":return zr(U,ge,K);case"Break":return ue+Le("break");case"Continue":return ue+Le("continue");default:return ue+"{{- "+$t(U)+" -}}"}}function cr(U,ge,K){const ue=K.repeat(ge),$e=[];let Ge=U;for(;Ge&&($e.push({test:Ge.test,body:Ge.body}),Ge.alternate.length===1&&Ge.alternate[0].type==="If");)Ge=Ge.alternate[0];let De=ue+Le("if",$t($e[0].test))+rt+xt($e[0].body,ge+1,K);for(let Dt=1;Dt<$e.length;Dt++)De+=rt+ue+Le("elif",$t($e[Dt].test))+rt+xt($e[Dt].body,ge+1,K);return Ge&&Ge.alternate.length>0&&(De+=rt+ue+Le("else")+rt+xt(Ge.alternate,ge+1,K)),De+=rt+ue+Le("endif"),De}function Kt(U,ge,K){const ue=K.repeat(ge);let $e="";if(U.iterable.type==="SelectExpression"){const De=U.iterable;$e=`${$t(De.iterable)} if ${$t(De.test)}`}else $e=$t(U.iterable);let Ge=ue+Le("for",$t(U.loopvar),"in",$e)+rt+xt(U.body,ge+1,K);return U.defaultBlock.length>0&&(Ge+=rt+ue+Le("else")+rt+xt(U.defaultBlock,ge+1,K)),Ge+=rt+ue+Le("endfor"),Ge}function wr(U,ge,K){const ue=K.repeat(ge),$e=$t(U.assignee),Ge=U.value?$t(U.value):"",De=ue+Le("set",`${$e}${U.value?" = "+Ge:""}`);return U.body.length===0?De:De+rt+xt(U.body,ge+1,K)+rt+ue+Le("endset")}function zr(U,ge,K){const ue=K.repeat(ge),$e=U.args.map($t).join(", ");return ue+Le("macro",`${U.name.value}(${$e})`)+rt+xt(U.body,ge+1,K)+rt+ue+Le("endmacro")}function $t(U,ge=-1){switch(U.type){case"Identifier":return U.value;case"NullLiteral":return"none";case"NumericLiteral":case"BooleanLiteral":return`${U.value}`;case"StringLiteral":return JSON.stringify(U.value);case"BinaryExpression":{const K=U,ue=ct[K.operator.type]??0,$e=$t(K.left,ue),Ge=$t(K.right,ue+1),De=`${$e} ${K.operator.value} ${Ge}`;return ue<ge?`(${De})`:De}case"UnaryExpression":{const K=U;return K.operator.value+(K.operator.value==="not"?" ":"")+$t(K.argument,1/0)}case"LogicalNegationExpression":return`not ${$t(U.argument,1/0)}`;case"CallExpression":{const K=U,ue=K.args.map($e=>$t($e,-1)).join(", ");return`${$t(K.callee,-1)}(${ue})`}case"MemberExpression":{const K=U;let ue=$t(K.object,-1);K.object.type!=="Identifier"&&(ue=`(${ue})`);let $e=$t(K.property,-1);return!K.computed&&K.property.type!=="Identifier"&&($e=`(${$e})`),K.computed?`${ue}[${$e}]`:`${ue}.${$e}`}case"FilterExpression":{const K=U,ue=$t(K.operand,1/0);return K.filter.type==="CallExpression"?`${ue} | ${$t(K.filter,-1)}`:`${ue} | ${K.filter.value}`}case"SelectExpression":{const K=U;return`${$t(K.iterable,-1)} | select(${$t(K.test,-1)})`}case"TestExpression":{const K=U;return`${$t(K.operand,-1)} is${K.negate?" not":""} ${K.test.value}`}case"ArrayLiteral":case"TupleLiteral":{const K=U.value.map($e=>$t($e,-1)),ue=U.type==="ArrayLiteral"?"[]":"()";return`${ue[0]}${K.join(", ")}${ue[1]}`}case"ObjectLiteral":return`{ ${Array.from(U.value.entries()).map(([ue,$e])=>`${$t(ue,-1)}: ${$t($e,-1)}`).join(", ")} }`;case"SliceExpression":{const K=U,ue=K.start?$t(K.start,-1):"",$e=K.stop?$t(K.stop,-1):"",Ge=K.step?`:${$t(K.step,-1)}`:"";return`${ue}:${$e}${Ge}`}case"KeywordArgumentExpression":{const K=U;return`${K.key.value}=${$t(K.value,-1)}`}case"If":{const K=U,ue=$t(K.test,-1),$e=$t(K.body[0],0),Ge=$t(K.alternate[0],-1);return`${$e} if ${ue} else ${Ge}`}default:throw new Error(`Unknown expression type: ${U.type}`)}}var Ti=class{constructor(U){le(this,"parsed");const ge=m(U,{lstrip_blocks:!0,trim_blocks:!0});this.parsed=Q(ge)}render(U){const ge=new Xe;if(ge.set("false",!1),ge.set("true",!0),ge.set("raise_exception",$e=>{throw new Error($e)}),ge.set("range",oe),U)for(const[$e,Ge]of Object.entries(U))ge.set($e,Ge);return new st(ge).run(this.parsed).value}format(U){return dt(this.parsed,(U==null?void 0:U.indent)||"	")}}},"./src/backends/onnx.js":(e,r,t)=>{var n;t.r(r),t.d(r,{Tensor:()=>d.Tensor,createInferenceSession:()=>z,deviceToExecutionProviders:()=>b,isONNXProxy:()=>D,isONNXTensor:()=>E});var o=t("./src/env.js"),i=t("?2ce3"),u=t("onnxruntime-web"),d=t("onnxruntime-common");const h=Object.freeze({auto:null,gpu:null,cpu:"cpu",wasm:"wasm",webgpu:"webgpu",cuda:"cuda",dml:"dml",webnn:{name:"webnn",deviceType:"cpu"},"webnn-npu":{name:"webnn",deviceType:"npu"},"webnn-gpu":{name:"webnn",deviceType:"gpu"},"webnn-cpu":{name:"webnn",deviceType:"cpu"}}),f=[];let y,m;const g=Symbol.for("onnxruntime");if(g in globalThis)m=globalThis[g];else if(o.apis.IS_NODE_ENV){switch(m=i??(n||(n=t.t(i,2))),process.platform){case"win32":f.push("dml");break;case"linux":process.arch==="x64"&&f.push("cuda");break}f.push("cpu"),y=["cpu"]}else m=u,o.apis.IS_WEBNN_AVAILABLE&&f.push("webnn-npu","webnn-gpu","webnn-cpu","webnn"),o.apis.IS_WEBGPU_AVAILABLE&&f.push("webgpu"),f.push("wasm"),y=["wasm"];const v=m.InferenceSession;function b(I=null){if(!I)return y;switch(I){case"auto":return f;case"gpu":return f.filter($=>["webgpu","cuda","dml","webnn-gpu"].includes($))}if(f.includes(I))return[h[I]??I];throw new Error(`Unsupported device: "${I}". Should be one of: ${f.join(", ")}.`)}let k=null;async function z(I,$,P){k&&await k;const S=v.create(I,$);k??(k=S);const O=await S;return O.config=P,O}function E(I){return I instanceof m.Tensor}const T=m==null?void 0:m.env;T!=null&&T.wasm&&(!(typeof ServiceWorkerGlobalScope<"u"&&self instanceof ServiceWorkerGlobalScope)&&!T.wasm.wasmPaths&&(T.wasm.wasmPaths=`https://cdn.jsdelivr.net/npm/@huggingface/transformers@${o.env.version}/dist/`),T.wasm.proxy=!1),T!=null&&T.webgpu&&(T.webgpu.powerPreference="high-performance");function D(){var I;return(I=T==null?void 0:T.wasm)==null?void 0:I.proxy}o.env.backends.onnx=T},"./src/base/feature_extraction_utils.js":(e,r,t)=>{t.r(r),t.d(r,{FeatureExtractor:()=>u,validate_audio_inputs:()=>d});var n=t("./src/utils/constants.js"),o=t("./src/utils/generic.js"),i=t("./src/utils/hub.js");class u extends o.Callable{constructor(f){super(),this.config=f}static async from_pretrained(f,y){const m=await(0,i.getModelJSON)(f,n.FEATURE_EXTRACTOR_NAME,!0,y);return new this(m)}}function d(h,f){var y;if(!(h instanceof Float32Array||h instanceof Float64Array))throw new Error(`${f} expects input to be a Float32Array or a Float64Array, but got ${((y=h==null?void 0:h.constructor)==null?void 0:y.name)??typeof h} instead. If using the feature extractor directly, remember to use \`read_audio(url, sampling_rate)\` to obtain the raw audio data of the file/url.`)}},"./src/base/image_processors_utils.js":(e,r,t)=>{t.r(r),t.d(r,{ImageProcessor:()=>I,center_to_corners_format:()=>m,post_process_instance_segmentation:()=>D,post_process_object_detection:()=>g,post_process_panoptic_segmentation:()=>T,post_process_semantic_segmentation:()=>v});var n=t("./src/utils/generic.js"),o=t("./src/utils/tensor.js"),i=t("./src/utils/maths.js");t("./src/utils/image.js");var u=t("./src/utils/core.js"),d=t("./src/utils/hub.js"),h=t("./src/utils/constants.js");function f($,P,S=0,O=null){const R=$/P;let G=(0,i.bankers_round)(R)*P;return O!==null&&G>O&&(G=Math.floor(R)*P),G<S&&(G=Math.ceil(R)*P),G}function y([$,P],S){return[Math.max(Math.floor($/S),1)*S,Math.max(Math.floor(P/S),1)*S]}function m([$,P,S,O]){return[$-S/2,P-O/2,$+S/2,P+O/2]}function g($,P=.5,S=null,O=!1){const R=$.logits,G=$.pred_boxes,[ee,ie,H]=R.dims;if(S!==null&&S.length!==ee)throw Error("Make sure that you pass in as many target sizes as the batch dimension of the logits");let ce=[];for(let re=0;re<ee;++re){let se=S!==null?S[re]:null,_e={boxes:[],classes:[],scores:[]},ae=R[re],Ce=G[re];for(let Te=0;Te<ie;++Te){let q=ae[Te],B=[],Q;if(O){Q=q.sigmoid().data;for(let oe=0;oe<Q.length;++oe)Q[oe]>P&&B.push(oe)}else{let oe=(0,i.max)(q.data)[1];if(oe===H-1||(Q=(0,i.softmax)(q.data),Q[oe]<P))continue;B.push(oe)}for(const oe of B){let Ie=Ce[Te].data;Ie=m(Ie),se!==null&&(Ie=Ie.map((ye,be)=>ye*se[(be+1)%2])),_e.boxes.push(Ie),_e.classes.push(oe),_e.scores.push(Q[oe])}}ce.push(_e)}return ce}function v($,P=null){const S=$.logits,O=S.dims[0];if(P!==null&&P.length!==O)throw Error("Make sure that you pass in as many target sizes as the batch dimension of the logits");const R=[];for(let G=0;G<O;++G){const ee=P!==null?P[G]:null;let ie=S[G];ee!==null&&(ie=(0,o.interpolate)(ie,ee,"bilinear",!1));const[H,ce]=ee??ie.dims.slice(-2),re=new o.Tensor("int32",new Int32Array(H*ce),[H,ce]),se=ie[0].data,_e=re.data;for(let Te=1;Te<ie.dims[0];++Te){const q=ie[Te].data;for(let B=0;B<q.length;++B)q[B]>se[B]&&(se[B]=q[B],_e[B]=Te)}const ae=new Array(ie.dims[0]);for(let Te=0;Te<_e.length;++Te){const q=_e[Te];ae[q]=q}const Ce=ae.filter(Te=>Te!==void 0);R.push({segmentation:re,labels:Ce})}return R}function b($,P,S,O){const R=[],G=[],ee=[];for(let ie=0;ie<$.dims[0];++ie){const H=$[ie],ce=P[ie],re=(0,i.max)(H.data)[1];if(re===O)continue;const _e=(0,i.softmax)(H.data)[re];_e>S&&(R.push(ce),G.push(_e),ee.push(re))}return[R,G,ee]}function k($,P,S,O=.5,R=.8){const G=[];let ee=0,ie=0;const H=P[S].data;for(let re=0;re<$.length;++re)$[re]===S&&(G.push(re),++ee),H[re]>=O&&++ie;let ce=ee>0&&ie>0;return ce&&(ce=ee/ie>R),[ce,G]}function z($,P,S,O,R,G=null,ee=null){const[ie,H]=ee??$[0].dims,ce=new o.Tensor("int32",new Int32Array(ie*H),[ie,H]),re=[];if(ee!==null)for(let Te=0;Te<$.length;++Te)$[Te]=(0,o.interpolate)($[Te],ee,"bilinear",!1);const se=new Int32Array($[0].data.length),_e=new Float32Array($[0].data.length);for(let Te=0;Te<$.length;++Te){let q=P[Te];const B=$[Te].data;for(let Q=0;Q<B.length;++Q)B[Q]*=q,B[Q]>_e[Q]&&(se[Q]=Te,_e[Q]=B[Q])}let ae=0;const Ce=ce.data;for(let Te=0;Te<S.length;++Te){const q=S[Te],[B,Q]=k(se,$,Te,O,R);if(B){++ae;for(const oe of Q)Ce[oe]=ae;re.push({id:ae,label_id:q,score:P[Te]})}}return[ce,re]}function E($,P,S=28,O=56*56,R=14*14*4*1280){if($<S||P<S)throw new Error(`height:${$} or width:${P} must be larger than factor:${S}`);if(Math.max($,P)/Math.min($,P)>200)throw new Error(`absolute aspect ratio must be smaller than 200, got ${Math.max($,P)/Math.min($,P)}`);let G=Math.round($/S)*S,ee=Math.round(P/S)*S;if(G*ee>R){const ie=Math.sqrt($*P/R);G=Math.floor($/ie/S)*S,ee=Math.floor(P/ie/S)*S}else if(G*ee<O){const ie=Math.sqrt(O/($*P));G=Math.ceil($*ie/S)*S,ee=Math.ceil(P*ie/S)*S}return[G,ee]}function T($,P=.5,S=.5,O=.8,R=null,G=null){R===null&&(console.warn("`label_ids_to_fuse` unset. No instance will be fused."),R=new Set);const ee=$.class_queries_logits??$.logits,H=($.masks_queries_logits??$.pred_masks).sigmoid();let[ce,re,se]=ee.dims;if(se-=1,G!==null&&G.length!==ce)throw Error("Make sure that you pass in as many target sizes as the batch dimension of the logits");let _e=[];for(let ae=0;ae<ce;++ae){let Ce=G!==null?G[ae]:null,Te=ee[ae],q=H[ae],[B,Q,oe]=b(Te,q,P,se);if(oe.length===0){let[be,et]=Ce??q.dims.slice(-2),tt=new o.Tensor("int32",new Int32Array(be*et).fill(-1),[be,et]);_e.push({segmentation:tt,segments_info:[]});continue}let[Ie,ye]=z(B,Q,oe,S,O,R,Ce);_e.push({segmentation:Ie,segments_info:ye})}return _e}function D($,P=.5,S=null){throw new Error("`post_process_instance_segmentation` is not yet implemented.")}class I extends n.Callable{constructor(P){super(),this.image_mean=P.image_mean??P.mean,this.image_std=P.image_std??P.std,this.resample=P.resample??2,this.do_rescale=P.do_rescale??!0,this.rescale_factor=P.rescale_factor??1/255,this.do_normalize=P.do_normalize,this.do_thumbnail=P.do_thumbnail,this.size=P.size??P.image_size,this.do_resize=P.do_resize??this.size!==void 0,this.size_divisibility=P.size_divisibility??P.size_divisor,this.do_center_crop=P.do_center_crop,this.crop_size=P.crop_size,this.do_convert_rgb=P.do_convert_rgb??!0,this.do_crop_margin=P.do_crop_margin,this.pad_size=P.pad_size,this.do_pad=P.do_pad,this.min_pixels=P.min_pixels,this.max_pixels=P.max_pixels,this.do_pad&&!this.pad_size&&this.size&&this.size.width!==void 0&&this.size.height!==void 0&&(this.pad_size=this.size),this.do_flip_channel_order=P.do_flip_channel_order??!1,this.config=P}async thumbnail(P,S,O=2){const R=P.height,G=P.width,ee=S.height,ie=S.width;let H=Math.min(R,ee),ce=Math.min(G,ie);return H===R&&ce===G?P:(R>G?ce=Math.floor(G*H/R):G>R&&(H=Math.floor(R*ce/G)),await P.resize(ce,H,{resample:O}))}async crop_margin(P,S=200){const O=P.clone().grayscale(),R=(0,i.min)(O.data)[0],ee=(0,i.max)(O.data)[0]-R;if(ee===0)return P;const ie=S/255;let H=O.width,ce=O.height,re=0,se=0;const _e=O.data;for(let ae=0;ae<O.height;++ae){const Ce=ae*O.width;for(let Te=0;Te<O.width;++Te)(_e[Ce+Te]-R)/ee<ie&&(H=Math.min(H,Te),ce=Math.min(ce,ae),re=Math.max(re,Te),se=Math.max(se,ae))}return P=await P.crop([H,ce,re,se]),P}pad_image(P,S,O,{mode:R="constant",center:G=!1,constant_values:ee=0}={}){const[ie,H,ce]=S;let re,se;if(typeof O=="number"?(re=O,se=O):O==="square"?re=se=Math.max(ie,H):(re=O.width,se=O.height),re!==H||se!==ie){const _e=new Float32Array(re*se*ce);if(Array.isArray(ee))for(let Te=0;Te<_e.length;++Te)_e[Te]=ee[Te%ce];else ee!==0&&_e.fill(ee);const[ae,Ce]=G?[Math.floor((re-H)/2),Math.floor((se-ie)/2)]:[0,0];for(let Te=0;Te<ie;++Te){const q=(Te+Ce)*re,B=Te*H;for(let Q=0;Q<H;++Q){const oe=(q+Q+ae)*ce,Ie=(B+Q)*ce;for(let ye=0;ye<ce;++ye)_e[oe+ye]=P[Ie+ye]}}if(R==="symmetric"){if(G)throw new Error("`center` padding is not supported when `mode` is set to `symmetric`.");const Te=ie-1,q=H-1;for(let B=0;B<se;++B){const Q=B*re,oe=(0,u.calculateReflectOffset)(B,Te)*H;for(let Ie=0;Ie<re;++Ie){if(B<ie&&Ie<H)continue;const ye=(Q+Ie)*ce,be=(oe+(0,u.calculateReflectOffset)(Ie,q))*ce;for(let et=0;et<ce;++et)_e[ye+et]=P[be+et]}}}P=_e,S=[se,re,ce]}return[P,S]}rescale(P){for(let S=0;S<P.length;++S)P[S]=this.rescale_factor*P[S]}get_resize_output_image_size(P,S){const[O,R]=P.size;let G,ee;if(this.do_thumbnail){const{height:ie,width:H}=S;G=Math.min(ie,H)}else Number.isInteger(S)?(G=S,ee=this.config.max_size??G):S!==void 0&&(G=S.shortest_edge,ee=S.longest_edge);if(G!==void 0||ee!==void 0){const ie=G===void 0?1:Math.max(G/O,G/R),H=O*ie,ce=R*ie,re=ee===void 0?1:Math.min(ee/H,ee/ce);let se=Math.floor(Number((H*re).toFixed(2))),_e=Math.floor(Number((ce*re).toFixed(2)));return this.size_divisibility!==void 0&&([se,_e]=y([se,_e],this.size_divisibility)),[se,_e]}else if(S!==void 0&&S.width!==void 0&&S.height!==void 0){let ie=S.width,H=S.height;if(this.config.keep_aspect_ratio&&this.config.ensure_multiple_of){let ce=H/R,re=ie/O;Math.abs(1-re)<Math.abs(1-ce)?ce=re:re=ce,H=f(ce*R,this.config.ensure_multiple_of),ie=f(re*O,this.config.ensure_multiple_of)}return[ie,H]}else{if(this.size_divisibility!==void 0)return y([O,R],this.size_divisibility);if(this.min_pixels!==void 0&&this.max_pixels!==void 0){const ie=this.config.patch_size*this.config.merge_size;return E(R,O,ie,this.min_pixels,this.max_pixels)}else throw new Error(`Could not resize image due to unsupported \`this.size\` option in config: ${JSON.stringify(S)}`)}}async resize(P){const[S,O]=this.get_resize_output_image_size(P,this.size);return await P.resize(S,O,{resample:this.resample})}async preprocess(P,{do_normalize:S=null,do_pad:O=null,do_convert_rgb:R=null,do_convert_grayscale:G=null,do_flip_channel_order:ee=null}={}){this.do_crop_margin&&(P=await this.crop_margin(P));const[ie,H]=P.size;if(R??this.do_convert_rgb?P=P.rgb():G&&(P=P.grayscale()),this.do_resize&&(P=await this.resize(P)),this.do_thumbnail&&(P=await this.thumbnail(P,this.size,this.resample)),this.do_center_crop){let ae,Ce;Number.isInteger(this.crop_size)?(ae=this.crop_size,Ce=this.crop_size):(ae=this.crop_size.width,Ce=this.crop_size.height),P=await P.center_crop(ae,Ce)}const ce=[P.height,P.width];let re=Float32Array.from(P.data),se=[P.height,P.width,P.channels];if(this.do_rescale&&this.rescale(re),S??this.do_normalize){let ae=this.image_mean;Array.isArray(this.image_mean)||(ae=new Array(P.channels).fill(ae));let Ce=this.image_std;if(Array.isArray(this.image_std)||(Ce=new Array(P.channels).fill(ae)),ae.length!==P.channels||Ce.length!==P.channels)throw new Error(`When set to arrays, the length of \`image_mean\` (${ae.length}) and \`image_std\` (${Ce.length}) must match the number of channels in the image (${P.channels}).`);for(let Te=0;Te<re.length;Te+=P.channels)for(let q=0;q<P.channels;++q)re[Te+q]=(re[Te+q]-ae[q])/Ce[q]}if(O??this.do_pad){if(this.pad_size)[re,se]=this.pad_image(re,[P.height,P.width,P.channels],this.pad_size);else if(this.size_divisibility){const[ae,Ce]=y([se[1],se[0]],this.size_divisibility);[re,se]=this.pad_image(re,se,{width:ae,height:Ce})}}if(ee??this.do_flip_channel_order){if(se[2]!==3)throw new Error("Flipping channel order is only supported for RGB images.");for(let ae=0;ae<re.length;ae+=3){const Ce=re[ae];re[ae]=re[ae+2],re[ae+2]=Ce}}const _e=new o.Tensor("float32",re,se).permute(2,0,1);return{original_size:[H,ie],reshaped_input_size:ce,pixel_values:_e}}async _call(P,...S){Array.isArray(P)||(P=[P]);const O=await Promise.all(P.map(G=>this.preprocess(G)));return{pixel_values:(0,o.stack)(O.map(G=>G.pixel_values),0),original_sizes:O.map(G=>G.original_size),reshaped_input_sizes:O.map(G=>G.reshaped_input_size)}}static async from_pretrained(P,S){const O=await(0,d.getModelJSON)(P,h.IMAGE_PROCESSOR_NAME,!0,S);return new this(O)}}},"./src/base/processing_utils.js":(e,r,t)=>{t.r(r),t.d(r,{Processor:()=>u});var n=t("./src/utils/constants.js"),o=t("./src/utils/generic.js"),i=t("./src/utils/hub.js");class u extends o.Callable{constructor(h,f){super(),this.config=h,this.components=f}get image_processor(){return this.components.image_processor}get tokenizer(){return this.components.tokenizer}get feature_extractor(){return this.components.feature_extractor}apply_chat_template(h,f={}){if(!this.tokenizer)throw new Error("Unable to apply chat template without a tokenizer.");return this.tokenizer.apply_chat_template(h,{tokenize:!1,...f})}batch_decode(...h){if(!this.tokenizer)throw new Error("Unable to decode without a tokenizer.");return this.tokenizer.batch_decode(...h)}decode(...h){if(!this.tokenizer)throw new Error("Unable to decode without a tokenizer.");return this.tokenizer.decode(...h)}async _call(h,...f){for(const y of[this.image_processor,this.feature_extractor,this.tokenizer])if(y)return y(h,...f);throw new Error("No image processor, feature extractor, or tokenizer found.")}static async from_pretrained(h,f){const[y,m]=await Promise.all([this.uses_processor_config?(0,i.getModelJSON)(h,n.PROCESSOR_NAME,!0,f):{},Promise.all(this.classes.filter(g=>g in this).map(async g=>{const v=await this[g].from_pretrained(h,f);return[g.replace(/_class$/,""),v]})).then(Object.fromEntries)]);return new this(y,m)}}le(u,"classes",["image_processor_class","tokenizer_class","feature_extractor_class"]),le(u,"uses_processor_config",!1)},"./src/configs.js":(e,r,t)=>{t.r(r),t.d(r,{AutoConfig:()=>f,PretrainedConfig:()=>h,getKeyValueShapes:()=>d});var n=t("./src/utils/core.js"),o=t("./src/utils/hub.js");async function i(y,m){return await(0,o.getModelJSON)(y,"config.json",!0,m)}function u(y){const m={};let g={};switch(y.model_type){case"llava":case"paligemma":case"gemma3":case"florence2":case"llava_onevision":case"idefics3":case"ultravox":case"smolvlm":g=u(y.text_config);break;case"moondream1":g=u(y.phi_config);break;case"musicgen":g=u(y.decoder);break;case"multi_modality":g=u(y.language_config);break;case"gpt2":case"gptj":case"jais":case"codegen":case"gpt_bigcode":m.num_heads="n_head",m.num_layers="n_layer",m.hidden_size="n_embd";break;case"gpt_neox":case"stablelm":case"opt":case"falcon":m.num_heads="num_attention_heads",m.num_layers="num_hidden_layers",m.hidden_size="hidden_size";break;case"llama":case"olmo":case"olmo2":case"mobilellm":case"granite":case"cohere":case"mistral":case"starcoder2":case"qwen2":case"qwen2_vl":case"phi":case"phi3":case"phi3_v":m.num_heads="num_key_value_heads",m.num_layers="num_hidden_layers",m.hidden_size="hidden_size",m.num_attention_heads="num_attention_heads";break;case"qwen3":case"gemma":case"gemma2":case"gemma3_text":case"glm":case"helium":m.num_heads="num_key_value_heads",m.num_layers="num_hidden_layers",m.dim_kv="head_dim";break;case"openelm":m.num_heads="num_kv_heads",m.num_layers="num_transformer_layers",m.dim_kv="head_dim";break;case"gpt_neo":case"donut-swin":m.num_heads="num_heads",m.num_layers="num_layers",m.hidden_size="hidden_size";break;case"bloom":m.num_heads="n_head",m.num_layers="n_layer",m.hidden_size="hidden_size";break;case"mpt":m.num_heads="n_heads",m.num_layers="n_layers",m.hidden_size="d_model";break;case"exaone":m.num_heads="num_key_value_heads",m.num_layers="num_layers",m.dim_kv="head_dim",m.num_attention_heads="num_attention_heads";break;case"t5":case"mt5":case"longt5":m.num_decoder_layers="num_decoder_layers",m.num_decoder_heads="num_heads",m.decoder_dim_kv="d_kv",m.num_encoder_layers="num_layers",m.num_encoder_heads="num_heads",m.encoder_dim_kv="d_kv";break;case"bart":case"mbart":case"marian":case"whisper":case"lite-whisper":case"m2m_100":case"blenderbot":case"blenderbot-small":case"florence2_language":m.num_decoder_layers="decoder_layers",m.num_decoder_heads="decoder_attention_heads",m.decoder_hidden_size="d_model",m.num_encoder_layers="encoder_layers",m.num_encoder_heads="encoder_attention_heads",m.encoder_hidden_size="d_model";break;case"speecht5":m.num_decoder_layers="decoder_layers",m.num_decoder_heads="decoder_attention_heads",m.decoder_hidden_size="hidden_size",m.num_encoder_layers="encoder_layers",m.num_encoder_heads="encoder_attention_heads",m.encoder_hidden_size="hidden_size";break;case"trocr":m.num_encoder_layers=m.num_decoder_layers="decoder_layers",m.num_encoder_heads=m.num_decoder_heads="decoder_attention_heads",m.encoder_hidden_size=m.decoder_hidden_size="d_model";break;case"musicgen_decoder":m.num_encoder_layers=m.num_decoder_layers="num_hidden_layers",m.num_encoder_heads=m.num_decoder_heads="num_attention_heads",m.encoder_hidden_size=m.decoder_hidden_size="hidden_size";break;case"moonshine":m.num_decoder_layers="decoder_num_hidden_layers",m.num_decoder_heads="decoder_num_key_value_heads",m.num_encoder_layers="encoder_num_hidden_layers",m.num_encoder_heads="encoder_num_key_value_heads",m.encoder_hidden_size=m.decoder_hidden_size="hidden_size";break;case"vision-encoder-decoder":const b=u(y.decoder),k="num_decoder_layers"in b,z=(0,n.pick)(y,["model_type","is_encoder_decoder"]);return k?(z.num_decoder_layers=b.num_decoder_layers,z.num_decoder_heads=b.num_decoder_heads,z.decoder_hidden_size=b.decoder_hidden_size,z.num_encoder_layers=b.num_encoder_layers,z.num_encoder_heads=b.num_encoder_heads,z.encoder_hidden_size=b.encoder_hidden_size):(z.num_layers=b.num_layers,z.num_heads=b.num_heads,z.hidden_size=b.hidden_size),z}const v={...g,...(0,n.pick)(y,["model_type","multi_query","is_encoder_decoder"])};for(const b in m)v[b]=y[m[b]];return v}function d(y,{prefix:m="past_key_values",batch_size:g=1}={}){const v={},b=y.normalized_config;if(b.is_encoder_decoder&&"num_encoder_heads"in b&&"num_decoder_heads"in b){const k=b.encoder_dim_kv??b.encoder_hidden_size/b.num_encoder_heads,z=b.decoder_dim_kv??b.decoder_hidden_size/b.num_decoder_heads,E=[g,b.num_encoder_heads,0,k],T=[g,b.num_decoder_heads,0,z];for(let D=0;D<b.num_decoder_layers;++D)v[`${m}.${D}.encoder.key`]=E,v[`${m}.${D}.encoder.value`]=E,v[`${m}.${D}.decoder.key`]=T,v[`${m}.${D}.decoder.value`]=T}else{const k=b.num_heads,z=b.num_layers,E=b.dim_kv??b.hidden_size/(b.num_attention_heads??k);if(b.model_type==="falcon"){const T=[g*k,0,E];for(let D=0;D<z;++D)v[`${m}.${D}.key`]=T,v[`${m}.${D}.value`]=T}else if(b.multi_query){const T=[g*k,0,2*E];for(let D=0;D<z;++D)v[`${m}.${D}.key_value`]=T}else if(b.model_type==="bloom"){const T=[g*k,E,0],D=[g*k,0,E];for(let I=0;I<z;++I)v[`${m}.${I}.key`]=T,v[`${m}.${I}.value`]=D}else if(b.model_type==="openelm")for(let T=0;T<z;++T){const D=[g,k[T],0,E];v[`${m}.${T}.key`]=D,v[`${m}.${T}.value`]=D}else{const T=[g,k,0,E];for(let D=0;D<z;++D)v[`${m}.${D}.key`]=T,v[`${m}.${D}.value`]=T}}return v}class h{constructor(m){le(this,"model_type",null);le(this,"is_encoder_decoder",!1);le(this,"max_position_embeddings");le(this,"transformers.js_config");Object.assign(this,m),this.normalized_config=u(this)}static async from_pretrained(m,{progress_callback:g=null,config:v=null,cache_dir:b=null,local_files_only:k=!1,revision:z="main"}={}){v&&!(v instanceof h)&&(v=new h(v));const E=v??await i(m,{progress_callback:g,config:v,cache_dir:b,local_files_only:k,revision:z});return new this(E)}}class f{static async from_pretrained(...m){return h.from_pretrained(...m)}}},"./src/env.js":(e,r,t)=>{var O,R;t.r(r),t.d(r,{apis:()=>z,env:()=>P});var n=t("?569f"),o=t("?3f59"),i=t("?154a");const u="3.5.1",d=typeof window<"u"&&typeof window.document<"u",h=typeof self<"u"&&((O=self.constructor)==null?void 0:O.name)==="DedicatedWorkerGlobalScope",f=typeof self<"u"&&"caches"in self,y=typeof navigator<"u"&&"gpu"in navigator,m=typeof navigator<"u"&&"ml"in navigator,g=typeof process<"u",v=g&&((R=process==null?void 0:process.release)==null?void 0:R.name)==="node",b=!S(n),k=!S(o),z=Object.freeze({IS_BROWSER_ENV:d,IS_WEBWORKER_ENV:h,IS_WEB_CACHE_AVAILABLE:f,IS_WEBGPU_AVAILABLE:y,IS_WEBNN_AVAILABLE:m,IS_PROCESS_AVAILABLE:g,IS_NODE_ENV:v,IS_FS_AVAILABLE:b,IS_PATH_AVAILABLE:k}),E=b&&k;let T="./";if(E){const G=Object(import.meta).url;G?T=o.dirname(o.dirname(i.fileURLToPath(G))):typeof __dirname<"u"&&(T=o.dirname(__dirname))}const D=E?o.join(T,"/.cache/"):null,I="/models/",$=E?o.join(T,I):I,P={version:u,backends:{onnx:{}},allowRemoteModels:!0,remoteHost:"https://huggingface.co/",remotePathTemplate:"{model}/resolve/{revision}/",allowLocalModels:!(d||h),localModelPath:$,useFS:b,useBrowserCache:f,useFSCache:b,cacheDir:D,useCustomCache:!1,customCache:null};function S(G){return Object.keys(G).length===0}},"./src/generation/configuration_utils.js":(e,r,t)=>{t.r(r),t.d(r,{GenerationConfig:()=>o});var n=t("./src/utils/core.js");class o{constructor(u){le(this,"max_length",20);le(this,"max_new_tokens",null);le(this,"min_length",0);le(this,"min_new_tokens",null);le(this,"early_stopping",!1);le(this,"max_time",null);le(this,"do_sample",!1);le(this,"num_beams",1);le(this,"num_beam_groups",1);le(this,"penalty_alpha",null);le(this,"use_cache",!0);le(this,"temperature",1);le(this,"top_k",50);le(this,"top_p",1);le(this,"typical_p",1);le(this,"epsilon_cutoff",0);le(this,"eta_cutoff",0);le(this,"diversity_penalty",0);le(this,"repetition_penalty",1);le(this,"encoder_repetition_penalty",1);le(this,"length_penalty",1);le(this,"no_repeat_ngram_size",0);le(this,"bad_words_ids",null);le(this,"force_words_ids",null);le(this,"renormalize_logits",!1);le(this,"constraints",null);le(this,"forced_bos_token_id",null);le(this,"forced_eos_token_id",null);le(this,"remove_invalid_values",!1);le(this,"exponential_decay_length_penalty",null);le(this,"suppress_tokens",null);le(this,"streamer",null);le(this,"begin_suppress_tokens",null);le(this,"forced_decoder_ids",null);le(this,"guidance_scale",null);le(this,"num_return_sequences",1);le(this,"output_attentions",!1);le(this,"output_hidden_states",!1);le(this,"output_scores",!1);le(this,"return_dict_in_generate",!1);le(this,"pad_token_id",null);le(this,"bos_token_id",null);le(this,"eos_token_id",null);le(this,"encoder_no_repeat_ngram_size",0);le(this,"decoder_start_token_id",null);le(this,"generation_kwargs",{});Object.assign(this,(0,n.pick)(u,Object.getOwnPropertyNames(this)))}}},"./src/generation/logits_process.js":(e,r,t)=>{t.r(r),t.d(r,{ClassifierFreeGuidanceLogitsProcessor:()=>E,ForcedBOSTokenLogitsProcessor:()=>h,ForcedEOSTokenLogitsProcessor:()=>f,LogitsProcessor:()=>i,LogitsProcessorList:()=>d,LogitsWarper:()=>u,MinLengthLogitsProcessor:()=>b,MinNewTokensLengthLogitsProcessor:()=>k,NoBadWordsLogitsProcessor:()=>z,NoRepeatNGramLogitsProcessor:()=>g,RepetitionPenaltyLogitsProcessor:()=>v,SuppressTokensAtBeginLogitsProcessor:()=>y,TemperatureLogitsWarper:()=>T,TopKLogitsWarper:()=>I,TopPLogitsWarper:()=>D,WhisperTimeStampLogitsProcessor:()=>m});var n=t("./src/utils/generic.js");t("./src/utils/tensor.js");var o=t("./src/utils/maths.js");class i extends n.Callable{_call(P,S){throw Error("`_call` should be implemented in a subclass")}}class u extends n.Callable{_call(P,S){throw Error("`_call` should be implemented in a subclass")}}class d extends n.Callable{constructor(){super(),this.processors=[]}push(P){this.processors.push(P)}extend(P){this.processors.push(...P)}_call(P,S){let O=S;for(const R of this.processors)O=R(P,O);return O}[Symbol.iterator](){return this.processors.values()}}class h extends i{constructor(P){super(),this.bos_token_id=P}_call(P,S){for(let O=0;O<P.length;++O)if(P[O].length===1){const R=S[O].data;R.fill(-1/0),R[this.bos_token_id]=0}return S}}class f extends i{constructor(P,S){super(),this.max_length=P,this.eos_token_id=Array.isArray(S)?S:[S]}_call(P,S){for(let O=0;O<P.length;++O)if(P[O].length===this.max_length-1){const R=S[O].data;R.fill(-1/0);for(const G of this.eos_token_id)R[G]=0}return S}}class y extends i{constructor(P,S){super(),this.begin_suppress_tokens=P,this.begin_index=S}_call(P,S){for(let O=0;O<P.length;++O)if(P[O].length===this.begin_index){const R=S[O].data;for(const G of this.begin_suppress_tokens)R[G]=-1/0}return S}}class m extends i{constructor(P,S){super(),this.eos_token_id=Array.isArray(P.eos_token_id)?P.eos_token_id[0]:P.eos_token_id,this.no_timestamps_token_id=P.no_timestamps_token_id,this.timestamp_begin=this.no_timestamps_token_id+1,this.begin_index=S.length,S.at(-1)===this.no_timestamps_token_id&&(this.begin_index-=1),this.max_initial_timestamp_index=P.max_initial_timestamp_index}_call(P,S){for(let O=0;O<P.length;++O){const R=S[O].data;if(R[this.no_timestamps_token_id]=-1/0,P[O].length===this.begin_index-1){R.fill(-1/0),R[this.timestamp_begin]=0;continue}const G=P[O].slice(this.begin_index),ee=G.length>=1&&G[G.length-1]>=this.timestamp_begin,ie=G.length<2||G[G.length-2]>=this.timestamp_begin;if(ee&&(ie?R.subarray(this.timestamp_begin).fill(-1/0):R.subarray(0,this.eos_token_id).fill(-1/0)),P[O].length===this.begin_index&&this.max_initial_timestamp_index!==null){const se=this.timestamp_begin+this.max_initial_timestamp_index;R.subarray(se+1).fill(-1/0)}const H=(0,o.log_softmax)(R),ce=Math.log(H.subarray(this.timestamp_begin).map(Math.exp).reduce((se,_e)=>se+_e)),re=(0,o.max)(H.subarray(0,this.timestamp_begin))[0];ce>re&&R.subarray(0,this.timestamp_begin).fill(-1/0)}return S}}class g extends i{constructor(P){super(),this.no_repeat_ngram_size=P}getNgrams(P){const S=P.length,O=[];for(let G=0;G<S+1-this.no_repeat_ngram_size;++G){const ee=[];for(let ie=0;ie<this.no_repeat_ngram_size;++ie)ee.push(P[G+ie]);O.push(ee.map(Number))}const R=new Map;for(const G of O){const ee=G.slice(0,G.length-1),ie=JSON.stringify(ee),H=R.get(ie)??[];H.push(G[G.length-1]),R.set(ie,H)}return R}getGeneratedNgrams(P,S){const O=S.slice(S.length+1-this.no_repeat_ngram_size,S.length);return P.get(JSON.stringify(O.map(Number)))??[]}calcBannedNgramTokens(P){const S=[];if(P.length+1<this.no_repeat_ngram_size)return S;{const O=this.getNgrams(P);return this.getGeneratedNgrams(O,P)}}_call(P,S){for(let O=0;O<P.length;++O){const R=S[O].data,G=this.calcBannedNgramTokens(P[O]);for(const ee of G)R[ee]=-1/0}return S}}class v extends i{constructor(P){super(),this.penalty=P}_call(P,S){for(let O=0;O<P.length;++O){const R=S[O].data;for(const G of new Set(P[O])){const ee=Number(G);R[ee]<0?R[ee]*=this.penalty:R[ee]/=this.penalty}}return S}}class b extends i{constructor(P,S){super(),this.min_length=P,this.eos_token_id=Array.isArray(S)?S:[S]}_call(P,S){for(let O=0;O<P.length;++O)if(P[O].length<this.min_length){const R=S[O].data;for(const G of this.eos_token_id)R[G]=-1/0}return S}}class k extends i{constructor(P,S,O){super(),this.prompt_length_to_skip=P,this.min_new_tokens=S,this.eos_token_id=Array.isArray(O)?O:[O]}_call(P,S){for(let O=0;O<P.length;++O)if(P[O].length-this.prompt_length_to_skip<this.min_new_tokens){const G=S[O].data;for(const ee of this.eos_token_id)G[ee]=-1/0}return S}}class z extends i{constructor(P,S){super(),this.bad_words_ids=P,this.eos_token_id=Array.isArray(S)?S:[S]}_call(P,S){for(let O=0;O<P.length;++O){const R=S[O].data,G=P[O];for(const ee of this.bad_words_ids){if(G.length<ee.length-1)continue;let ie=!0;for(let H=1;H<=ee.length-1;++H)if(ee.at(-H-1)!=G.at(-H)){ie=!1;break}ie&&(R[ee.at(-1)]=-1/0)}}return S}}class E extends i{constructor(P){if(super(),P<=1)throw new Error(`Require guidance scale >1 to use the classifier free guidance processor, got guidance scale ${P}.`);this.guidance_scale=P}_call(P,S){if(S.dims[0]!==2*P.length)throw new Error(`Logits should have twice the batch size of the input ids, the first half of batches corresponding to the conditional inputs, and the second half of batches corresponding to the unconditional inputs. Got batch size ${S.dims[0]} for the logits and ${P.length} for the input ids.`);const O=P.length,R=S.slice([0,O],null),G=S.slice([O,S.dims[0]],null);for(let ee=0;ee<G.data.length;++ee)G.data[ee]+=(R.data[ee]-G.data[ee])*this.guidance_scale;return G}}class T extends u{constructor(P){super(),this.temperature=P}_call(P,S){const O=S.data;for(let R=0;R<O.length;++R)O[R]/=this.temperature;return S}}class D extends u{constructor(P,{filter_value:S=-1/0,min_tokens_to_keep:O=1}={}){if(super(),P<0||P>1)throw new Error(`\`top_p\` must be a float > 0 and < 1, but is ${P}`);if(!Number.isInteger(O)||O<1)throw new Error(`\`min_tokens_to_keep\` must be a positive integer, but is ${O}`);this.top_p=P,this.filter_value=S,this.min_tokens_to_keep=O}}class I extends u{constructor(P,{filter_value:S=-1/0,min_tokens_to_keep:O=1}={}){if(super(),!Number.isInteger(P)||P<0)throw new Error(`\`top_k\` must be a positive integer, but is ${P}`);this.top_k=Math.max(P,O),this.filter_value=S}}},"./src/generation/logits_sampler.js":(e,r,t)=>{t.r(r),t.d(r,{LogitsSampler:()=>u});var n=t("./src/utils/generic.js"),o=t("./src/utils/tensor.js"),i=t("./src/utils/maths.js");t("./src/generation/configuration_utils.js");class u extends n.Callable{constructor(m){super(),this.generation_config=m}async _call(m){return this.sample(m)}async sample(m){throw Error("sample should be implemented in subclasses.")}getLogits(m,g){let v=m.dims.at(-1),b=m.data;if(g===-1)b=b.slice(-v);else{let k=g*v;b=b.slice(k,k+v)}return b}randomSelect(m){let g=0;for(let b=0;b<m.length;++b)g+=m[b];let v=Math.random()*g;for(let b=0;b<m.length;++b)if(v-=m[b],v<=0)return b;return 0}static getSampler(m){if(m.do_sample)return new h(m);if(m.num_beams>1)return new f(m);if(m.num_return_sequences>1)throw Error(`num_return_sequences has to be 1 when doing greedy search, but is ${m.num_return_sequences}.`);return new d(m)}}class d extends u{async sample(m){const g=(0,i.max)(m.data)[1];return[[BigInt(g),0]]}}class h extends u{async sample(m){let g=m.dims.at(-1);this.generation_config.top_k>0&&(g=Math.min(this.generation_config.top_k,g));const[v,b]=await(0,o.topk)(m,g),k=(0,i.softmax)(v.data);return Array.from({length:this.generation_config.num_beams},()=>{const z=this.randomSelect(k);return[b.data[z],Math.log(k[z])]})}}class f extends u{async sample(m){let g=m.dims.at(-1);this.generation_config.top_k>0&&(g=Math.min(this.generation_config.top_k,g));const[v,b]=await(0,o.topk)(m,g),k=(0,i.softmax)(v.data);return Array.from({length:this.generation_config.num_beams},(z,E)=>[b.data[E],Math.log(k[E])])}}},"./src/generation/stopping_criteria.js":(e,r,t)=>{t.r(r),t.d(r,{EosTokenCriteria:()=>d,InterruptableStoppingCriteria:()=>h,MaxLengthCriteria:()=>u,StoppingCriteria:()=>o,StoppingCriteriaList:()=>i});var n=t("./src/utils/generic.js");class o extends n.Callable{_call(y,m){throw Error("StoppingCriteria needs to be subclassed")}}class i extends n.Callable{constructor(){super(),this.criteria=[]}push(y){this.criteria.push(y)}extend(y){y instanceof i?y=y.criteria:y instanceof o&&(y=[y]),this.criteria.push(...y)}_call(y,m){const g=new Array(y.length).fill(!1);for(const v of this.criteria){const b=v(y,m);for(let k=0;k<g.length;++k)g[k]||(g[k]=b[k])}return g}[Symbol.iterator](){return this.criteria.values()}}class u extends o{constructor(y,m=null){super(),this.max_length=y,this.max_position_embeddings=m}_call(y){return y.map(m=>m.length>=this.max_length)}}class d extends o{constructor(y){super(),Array.isArray(y)||(y=[y]),this.eos_token_id=y}_call(y,m){return y.map(g=>{const v=g.at(-1);return this.eos_token_id.some(b=>v==b)})}}class h extends o{constructor(){super(),this.interrupted=!1}interrupt(){this.interrupted=!0}reset(){this.interrupted=!1}_call(y,m){return new Array(y.length).fill(this.interrupted)}}},"./src/generation/streamers.js":(e,r,t)=>{t.r(r),t.d(r,{BaseStreamer:()=>u,TextStreamer:()=>h,WhisperTextStreamer:()=>f});var n=t("./src/utils/core.js"),o=t("./src/tokenizers.js"),i=t("./src/env.js");class u{put(m){throw Error("Not implemented")}end(){throw Error("Not implemented")}}const d=i.apis.IS_PROCESS_AVAILABLE?y=>process.stdout.write(y):y=>console.log(y);class h extends u{constructor(m,{skip_prompt:g=!1,callback_function:v=null,token_callback_function:b=null,skip_special_tokens:k=!0,decode_kwargs:z={},...E}={}){super(),this.tokenizer=m,this.skip_prompt=g,this.callback_function=v??d,this.token_callback_function=b,this.decode_kwargs={skip_special_tokens:k,...z,...E},this.token_cache=[],this.print_len=0,this.next_tokens_are_prompt=!0}put(m){var z;if(m.length>1)throw Error("TextStreamer only supports batch size of 1");const g=this.next_tokens_are_prompt;if(g&&(this.next_tokens_are_prompt=!1,this.skip_prompt))return;const v=m[0];(z=this.token_callback_function)==null||z.call(this,v),this.token_cache=(0,n.mergeArrays)(this.token_cache,v);const b=this.tokenizer.decode(this.token_cache,this.decode_kwargs);let k;g||b.endsWith(`
`)?(k=b.slice(this.print_len),this.token_cache=[],this.print_len=0):b.length>0&&(0,o.is_chinese_char)(b.charCodeAt(b.length-1))?(k=b.slice(this.print_len),this.print_len+=k.length):(k=b.slice(this.print_len,b.lastIndexOf(" ")+1),this.print_len+=k.length),this.on_finalized_text(k,!1)}end(){let m;this.token_cache.length>0?(m=this.tokenizer.decode(this.token_cache,this.decode_kwargs).slice(this.print_len),this.token_cache=[],this.print_len=0):m="",this.next_tokens_are_prompt=!0,this.on_finalized_text(m,!0)}on_finalized_text(m,g){var v,b;m.length>0&&((v=this.callback_function)==null||v.call(this,m)),g&&this.callback_function===d&&i.apis.IS_PROCESS_AVAILABLE&&((b=this.callback_function)==null||b.call(this,`
`))}}class f extends h{constructor(m,{skip_prompt:g=!1,callback_function:v=null,token_callback_function:b=null,on_chunk_start:k=null,on_chunk_end:z=null,on_finalize:E=null,time_precision:T=.02,skip_special_tokens:D=!0,decode_kwargs:I={}}={}){super(m,{skip_prompt:g,skip_special_tokens:D,callback_function:v,token_callback_function:b,decode_kwargs:I}),this.timestamp_begin=m.timestamp_begin,this.on_chunk_start=k,this.on_chunk_end=z,this.on_finalize=E,this.time_precision=T,this.waiting_for_timestamp=!1}put(m){var v,b;if(m.length>1)throw Error("WhisperTextStreamer only supports batch size of 1");const g=m[0];if(g.length===1){const k=Number(g[0])-this.timestamp_begin;if(k>=0){const z=k*this.time_precision;this.waiting_for_timestamp?(v=this.on_chunk_end)==null||v.call(this,z):(b=this.on_chunk_start)==null||b.call(this,z),this.waiting_for_timestamp=!this.waiting_for_timestamp,m=[[]]}}return super.put(m)}end(){var m;super.end(),(m=this.on_finalize)==null||m.call(this)}}},"./src/models.js":(e,r,t)=>{t.r(r),t.d(r,{ASTForAudioClassification:()=>Gs,ASTModel:()=>xd,ASTPreTrainedModel:()=>Qa,AlbertForMaskedLM:()=>ne,AlbertForQuestionAnswering:()=>J,AlbertForSequenceClassification:()=>j,AlbertModel:()=>ve,AlbertPreTrainedModel:()=>Ei,AutoModel:()=>Hu,AutoModelForAudioClassification:()=>Mh,AutoModelForAudioFrameClassification:()=>vh,AutoModelForAudioTextToText:()=>Ih,AutoModelForCTC:()=>wh,AutoModelForCausalLM:()=>lh,AutoModelForDepthEstimation:()=>Ph,AutoModelForDocumentQuestionAnswering:()=>xh,AutoModelForImageClassification:()=>ph,AutoModelForImageFeatureExtraction:()=>$h,AutoModelForImageMatting:()=>Th,AutoModelForImageSegmentation:()=>hh,AutoModelForImageTextToText:()=>kh,AutoModelForImageToImage:()=>Eh,AutoModelForMaskGeneration:()=>yh,AutoModelForMaskedLM:()=>ch,AutoModelForNormalEstimation:()=>Ch,AutoModelForObjectDetection:()=>_h,AutoModelForPoseEstimation:()=>Sh,AutoModelForQuestionAnswering:()=>uh,AutoModelForSemanticSegmentation:()=>fh,AutoModelForSeq2SeqLM:()=>sh,AutoModelForSequenceClassification:()=>rh,AutoModelForSpeechSeq2Seq:()=>nh,AutoModelForTextToSpectrogram:()=>ah,AutoModelForTextToWaveform:()=>oh,AutoModelForTokenClassification:()=>ih,AutoModelForUniversalSegmentation:()=>mh,AutoModelForVision2Seq:()=>dh,AutoModelForXVector:()=>bh,AutoModelForZeroShotObjectDetection:()=>gh,BartForConditionalGeneration:()=>tr,BartForSequenceClassification:()=>Cr,BartModel:()=>_r,BartPretrainedModel:()=>qt,BaseModelOutput:()=>Oe,BeitForImageClassification:()=>Jo,BeitModel:()=>Xo,BeitPreTrainedModel:()=>ea,BertForMaskedLM:()=>Ve,BertForQuestionAnswering:()=>st,BertForSequenceClassification:()=>He,BertForTokenClassification:()=>Xe,BertModel:()=>Be,BertPreTrainedModel:()=>Ne,BlenderbotForConditionalGeneration:()=>Br,BlenderbotModel:()=>Ir,BlenderbotPreTrainedModel:()=>sr,BlenderbotSmallForConditionalGeneration:()=>Rr,BlenderbotSmallModel:()=>Ri,BlenderbotSmallPreTrainedModel:()=>yi,BloomForCausalLM:()=>Fo,BloomModel:()=>vp,BloomPreTrainedModel:()=>Oo,CLIPModel:()=>Fd,CLIPPreTrainedModel:()=>rs,CLIPSegForImageSegmentation:()=>Vd,CLIPSegModel:()=>Nd,CLIPSegPreTrainedModel:()=>no,CLIPTextModel:()=>hf,CLIPTextModelWithProjection:()=>ro,CLIPVisionModel:()=>Dd,CLIPVisionModelWithProjection:()=>Es,CamembertForMaskedLM:()=>gt,CamembertForQuestionAnswering:()=>ri,CamembertForSequenceClassification:()=>Ht,CamembertForTokenClassification:()=>nr,CamembertModel:()=>jt,CamembertPreTrainedModel:()=>Et,CausalLMOutput:()=>fs,CausalLMOutputWithPast:()=>vf,ChineseCLIPModel:()=>Fn,ChineseCLIPPreTrainedModel:()=>zd,ClapAudioModelWithProjection:()=>Jc,ClapModel:()=>Qc,ClapPreTrainedModel:()=>nn,ClapTextModelWithProjection:()=>Xc,CodeGenForCausalLM:()=>Hd,CodeGenModel:()=>po,CodeGenPreTrainedModel:()=>wi,CohereForCausalLM:()=>xo,CohereModel:()=>vo,CoherePreTrainedModel:()=>Bn,ConvBertForMaskedLM:()=>Ti,ConvBertForQuestionAnswering:()=>K,ConvBertForSequenceClassification:()=>U,ConvBertForTokenClassification:()=>ge,ConvBertModel:()=>$t,ConvBertPreTrainedModel:()=>zr,ConvNextForImageClassification:()=>Gl,ConvNextModel:()=>Wl,ConvNextPreTrainedModel:()=>fa,ConvNextV2ForImageClassification:()=>Hl,ConvNextV2Model:()=>Kl,ConvNextV2PreTrainedModel:()=>ma,DFineForObjectDetection:()=>ul,DFineModel:()=>cl,DFinePreTrainedModel:()=>na,DPTForDepthEstimation:()=>Pl,DPTModel:()=>El,DPTPreTrainedModel:()=>da,DacDecoderModel:()=>Bu,DacDecoderOutput:()=>Lu,DacEncoderModel:()=>zu,DacEncoderOutput:()=>Du,DacModel:()=>ju,DacPreTrainedModel:()=>pn,DebertaForMaskedLM:()=>Ai,DebertaForQuestionAnswering:()=>ns,DebertaForSequenceClassification:()=>Ji,DebertaForTokenClassification:()=>Oi,DebertaModel:()=>Jr,DebertaPreTrainedModel:()=>pr,DebertaV2ForMaskedLM:()=>_i,DebertaV2ForQuestionAnswering:()=>Zi,DebertaV2ForSequenceClassification:()=>Fi,DebertaV2ForTokenClassification:()=>er,DebertaV2Model:()=>Yi,DebertaV2PreTrainedModel:()=>ii,DecisionTransformerModel:()=>yu,DecisionTransformerPreTrainedModel:()=>gu,DeiTForImageClassification:()=>ml,DeiTModel:()=>fl,DeiTPreTrainedModel:()=>oa,DepthAnythingForDepthEstimation:()=>Sl,DepthAnythingPreTrainedModel:()=>Cl,DepthProForDepthEstimation:()=>Ol,DepthProPreTrainedModel:()=>Al,DetrForObjectDetection:()=>Zo,DetrForSegmentation:()=>ta,DetrModel:()=>Yo,DetrObjectDetectionOutput:()=>Ys,DetrPreTrainedModel:()=>Js,DetrSegmentationOutput:()=>el,Dinov2ForImageClassification:()=>Ql,Dinov2Model:()=>ql,Dinov2PreTrainedModel:()=>_a,Dinov2WithRegistersForImageClassification:()=>Jl,Dinov2WithRegistersModel:()=>Xl,Dinov2WithRegistersPreTrainedModel:()=>ga,DistilBertForMaskedLM:()=>qe,DistilBertForQuestionAnswering:()=>si,DistilBertForSequenceClassification:()=>Hr,DistilBertForTokenClassification:()=>es,DistilBertModel:()=>Kr,DistilBertPreTrainedModel:()=>Gr,DonutSwinModel:()=>Ul,DonutSwinPreTrainedModel:()=>Vl,EfficientNetForImageClassification:()=>su,EfficientNetModel:()=>iu,EfficientNetPreTrainedModel:()=>Ca,ElectraForMaskedLM:()=>Ge,ElectraForQuestionAnswering:()=>Ut,ElectraForSequenceClassification:()=>De,ElectraForTokenClassification:()=>Dt,ElectraModel:()=>$e,ElectraPreTrainedModel:()=>ue,EsmForMaskedLM:()=>ur,EsmForSequenceClassification:()=>gi,EsmForTokenClassification:()=>Di,EsmModel:()=>yt,EsmPreTrainedModel:()=>lt,ExaoneForCausalLM:()=>ep,ExaoneModel:()=>Zd,ExaonePreTrainedModel:()=>_o,FalconForCausalLM:()=>qc,FalconModel:()=>Hc,FalconPreTrainedModel:()=>Ta,FastViTForImageClassification:()=>Go,FastViTModel:()=>qn,FastViTPreTrainedModel:()=>ps,Florence2ForConditionalGeneration:()=>kd,Florence2PreTrainedModel:()=>$d,GLPNForDepthEstimation:()=>Nl,GLPNModel:()=>Rl,GLPNPreTrainedModel:()=>ha,GPT2LMHeadModel:()=>Wd,GPT2Model:()=>Ud,GPT2PreTrainedModel:()=>ao,GPTBigCodeForCausalLM:()=>uo,GPTBigCodeModel:()=>Qs,GPTBigCodePreTrainedModel:()=>Tt,GPTJForCausalLM:()=>Ps,GPTJModel:()=>qs,GPTJPreTrainedModel:()=>hi,GPTNeoForCausalLM:()=>Kd,GPTNeoModel:()=>lo,GPTNeoPreTrainedModel:()=>us,GPTNeoXForCausalLM:()=>Ar,GPTNeoXModel:()=>Si,GPTNeoXPreTrainedModel:()=>co,Gemma2ForCausalLM:()=>op,Gemma2Model:()=>ap,Gemma2PreTrainedModel:()=>Po,Gemma3ForCausalLM:()=>cp,Gemma3Model:()=>lp,Gemma3PreTrainedModel:()=>Co,GemmaForCausalLM:()=>np,GemmaModel:()=>Eo,GemmaPreTrainedModel:()=>To,GlmForCausalLM:()=>Yd,GlmModel:()=>Jd,GlmPreTrainedModel:()=>mo,GraniteForCausalLM:()=>bo,GraniteModel:()=>Mo,GranitePreTrainedModel:()=>zn,GroundingDinoForObjectDetection:()=>Zl,GroundingDinoPreTrainedModel:()=>Yl,GroupViTModel:()=>Hn,GroupViTPreTrainedModel:()=>Kn,HeliumForCausalLM:()=>Xd,HeliumModel:()=>Qd,HeliumPreTrainedModel:()=>fo,HieraForImageClassification:()=>gl,HieraModel:()=>_l,HieraPreTrainedModel:()=>la,HubertForCTC:()=>$c,HubertForSequenceClassification:()=>kc,HubertModel:()=>Sc,HubertPreTrainedModel:()=>Tp,IJepaForImageClassification:()=>jo,IJepaModel:()=>Lo,IJepaPreTrainedModel:()=>Vn,Idefics3ForConditionalGeneration:()=>An,Idefics3PreTrainedModel:()=>Od,ImageMattingOutput:()=>Oh,JAISLMHeadModel:()=>Hs,JAISModel:()=>Gd,JAISPreTrainedModel:()=>oo,JinaCLIPModel:()=>Bd,JinaCLIPPreTrainedModel:()=>Ks,JinaCLIPTextModel:()=>Rd,JinaCLIPVisionModel:()=>cs,LiteWhisperForConditionalGeneration:()=>Ed,LlamaForCausalLM:()=>Dn,LlamaModel:()=>qd,LlamaPreTrainedModel:()=>ho,LlavaForConditionalGeneration:()=>In,LlavaOnevisionForConditionalGeneration:()=>Cd,LlavaPreTrainedModel:()=>Pd,LongT5ForConditionalGeneration:()=>it,LongT5Model:()=>ot,LongT5PreTrainedModel:()=>Ye,M2M100ForConditionalGeneration:()=>cc,M2M100Model:()=>lc,M2M100PreTrainedModel:()=>Ma,MBartForCausalLM:()=>Ci,MBartForConditionalGeneration:()=>Pi,MBartForSequenceClassification:()=>Fr,MBartModel:()=>Zr,MBartPreTrainedModel:()=>kr,MPNetForMaskedLM:()=>as,MPNetForQuestionAnswering:()=>xs,MPNetForSequenceClassification:()=>os,MPNetForTokenClassification:()=>ai,MPNetModel:()=>Yr,MPNetPreTrainedModel:()=>Bi,MT5ForConditionalGeneration:()=>Qt,MT5Model:()=>It,MT5PreTrainedModel:()=>ut,MarianMTModel:()=>oc,MarianModel:()=>ac,MarianPreTrainedModel:()=>wa,MaskFormerForInstanceSegmentation:()=>Bl,MaskFormerModel:()=>zl,MaskFormerPreTrainedModel:()=>pa,MaskedLMOutput:()=>Or,Metric3DForDepthEstimation:()=>Dl,Metric3DPreTrainedModel:()=>Fl,Metric3Dv2ForDepthEstimation:()=>jl,Metric3Dv2PreTrainedModel:()=>Ll,MgpstrForSceneTextRecognition:()=>xu,MgpstrModelOutput:()=>bu,MgpstrPreTrainedModel:()=>vu,MimiDecoderModel:()=>Fu,MimiDecoderOutput:()=>Iu,MimiEncoderModel:()=>Ou,MimiEncoderOutput:()=>ku,MimiModel:()=>Au,MimiPreTrainedModel:()=>dn,MistralForCausalLM:()=>Wc,MistralModel:()=>Uc,MistralPreTrainedModel:()=>va,MobileBertForMaskedLM:()=>ji,MobileBertForQuestionAnswering:()=>qr,MobileBertForSequenceClassification:()=>zi,MobileBertModel:()=>Li,MobileBertPreTrainedModel:()=>ni,MobileLLMForCausalLM:()=>rp,MobileLLMModel:()=>tp,MobileLLMPreTrainedModel:()=>go,MobileNetV1ForImageClassification:()=>au,MobileNetV1ForSemanticSegmentation:()=>ou,MobileNetV1Model:()=>nu,MobileNetV1PreTrainedModel:()=>on,MobileNetV2ForImageClassification:()=>cu,MobileNetV2ForSemanticSegmentation:()=>uu,MobileNetV2Model:()=>lu,MobileNetV2PreTrainedModel:()=>ln,MobileNetV3ForImageClassification:()=>pu,MobileNetV3ForSemanticSegmentation:()=>hu,MobileNetV3Model:()=>du,MobileNetV3PreTrainedModel:()=>cn,MobileNetV4ForImageClassification:()=>mu,MobileNetV4ForSemanticSegmentation:()=>_u,MobileNetV4Model:()=>fu,MobileNetV4PreTrainedModel:()=>un,MobileViTForImageClassification:()=>Jn,MobileViTModel:()=>Ho,MobileViTPreTrainedModel:()=>Xn,MobileViTV2ForImageClassification:()=>Zn,MobileViTV2Model:()=>qo,MobileViTV2PreTrainedModel:()=>Yn,ModelOutput:()=>Ae,ModernBertForMaskedLM:()=>rt,ModernBertForSequenceClassification:()=>pt,ModernBertForTokenClassification:()=>_t,ModernBertModel:()=>We,ModernBertPreTrainedModel:()=>nt,Moondream1ForConditionalGeneration:()=>Sd,MoonshineForConditionalGeneration:()=>Ya,MoonshineModel:()=>pf,MoonshinePreTrainedModel:()=>Ja,MptForCausalLM:()=>xp,MptModel:()=>te,MptPreTrainedModel:()=>Cs,MultiModalityCausalLM:()=>Mu,MultiModalityPreTrainedModel:()=>wu,MusicgenForCausalLM:()=>Sp,MusicgenForConditionalGeneration:()=>$a,MusicgenModel:()=>Cp,MusicgenPreTrainedModel:()=>Sa,NomicBertModel:()=>dt,NomicBertPreTrainedModel:()=>ct,OPTForCausalLM:()=>Nn,OPTModel:()=>Do,OPTPreTrainedModel:()=>Rn,Olmo2ForCausalLM:()=>jn,Olmo2Model:()=>Ln,Olmo2PreTrainedModel:()=>wo,OlmoForCausalLM:()=>sp,OlmoModel:()=>ip,OlmoPreTrainedModel:()=>yo,OpenELMForCausalLM:()=>dp,OpenELMModel:()=>up,OpenELMPreTrainedModel:()=>So,OwlViTForObjectDetection:()=>Qo,OwlViTModel:()=>$s,OwlViTPreTrainedModel:()=>we,Owlv2ForObjectDetection:()=>fe,Owlv2Model:()=>Xs,Owlv2PreTrainedModel:()=>he,PaliGemmaForConditionalGeneration:()=>Ad,PaliGemmaPreTrainedModel:()=>Id,PatchTSMixerForPrediction:()=>Cu,PatchTSMixerModel:()=>Pu,PatchTSMixerPreTrainedModel:()=>Ia,PatchTSTForPrediction:()=>Eu,PatchTSTModel:()=>Tu,PatchTSTPreTrainedModel:()=>ka,Phi3ForCausalLM:()=>bp,Phi3Model:()=>Mp,Phi3PreTrainedModel:()=>Ao,Phi3VForCausalLM:()=>On,Phi3VPreTrainedModel:()=>to,PhiForCausalLM:()=>wp,PhiModel:()=>yp,PhiPreTrainedModel:()=>Io,PreTrainedModel:()=>X,PretrainedMixin:()=>Wt,PvtForImageClassification:()=>No,PvtModel:()=>Ro,PvtPreTrainedModel:()=>Un,PyAnnoteForAudioFrameClassification:()=>mc,PyAnnoteModel:()=>fc,PyAnnotePreTrainedModel:()=>ba,QuestionAnsweringModelOutput:()=>Vr,Qwen2ForCausalLM:()=>hp,Qwen2Model:()=>pp,Qwen2PreTrainedModel:()=>$o,Qwen2VLForConditionalGeneration:()=>gp,Qwen2VLPreTrainedModel:()=>_p,Qwen3ForCausalLM:()=>mp,Qwen3Model:()=>fp,Qwen3PreTrainedModel:()=>ko,RFDetrForObjectDetection:()=>ol,RFDetrModel:()=>al,RFDetrObjectDetectionOutput:()=>ll,RFDetrPreTrainedModel:()=>sa,RTDetrForObjectDetection:()=>rl,RTDetrModel:()=>tl,RTDetrObjectDetectionOutput:()=>ks,RTDetrPreTrainedModel:()=>ra,RTDetrV2ForObjectDetection:()=>sl,RTDetrV2Model:()=>il,RTDetrV2ObjectDetectionOutput:()=>nl,RTDetrV2PreTrainedModel:()=>ia,ResNetForImageClassification:()=>wl,ResNetModel:()=>yl,ResNetPreTrainedModel:()=>ca,RoFormerForMaskedLM:()=>Nt,RoFormerForQuestionAnswering:()=>wr,RoFormerForSequenceClassification:()=>cr,RoFormerForTokenClassification:()=>Kt,RoFormerModel:()=>xt,RoFormerPreTrainedModel:()=>Le,RobertaForMaskedLM:()=>gr,RobertaForQuestionAnswering:()=>Nr,RobertaForSequenceClassification:()=>xr,RobertaForTokenClassification:()=>Tr,RobertaModel:()=>Mr,RobertaPreTrainedModel:()=>ei,SamImageSegmentationOutput:()=>nc,SamModel:()=>sc,SamPreTrainedModel:()=>ic,SapiensForDepthEstimation:()=>kl,SapiensForNormalEstimation:()=>Il,SapiensForSemanticSegmentation:()=>$l,SapiensPreTrainedModel:()=>en,SegformerForImageClassification:()=>Zc,SegformerForSemanticSegmentation:()=>eu,SegformerModel:()=>Pp,SegformerPreTrainedModel:()=>an,Seq2SeqLMOutput:()=>bf,SequenceClassifierOutput:()=>At,SiglipModel:()=>so,SiglipPreTrainedModel:()=>io,SiglipTextModel:()=>Ld,SiglipVisionModel:()=>jd,SmolVLMForConditionalGeneration:()=>eo,SnacDecoderModel:()=>Vu,SnacEncoderModel:()=>Nu,SnacModel:()=>Ru,SnacPreTrainedModel:()=>hn,SpeechT5ForSpeechToText:()=>zc,SpeechT5ForTextToSpeech:()=>Bc,SpeechT5HifiGan:()=>Rc,SpeechT5Model:()=>Ep,SpeechT5PreTrainedModel:()=>sn,SqueezeBertForMaskedLM:()=>Vs,SqueezeBertForQuestionAnswering:()=>oi,SqueezeBertForSequenceClassification:()=>ls,SqueezeBertModel:()=>pi,SqueezeBertPreTrainedModel:()=>ts,StableLmForCausalLM:()=>ru,StableLmModel:()=>tu,StableLmPreTrainedModel:()=>Pa,Starcoder2ForCausalLM:()=>Kc,Starcoder2Model:()=>Gc,Starcoder2PreTrainedModel:()=>xa,StyleTextToSpeech2Model:()=>jc,StyleTextToSpeech2PreTrainedModel:()=>Lc,Swin2SRForImageSuperResolution:()=>Tl,Swin2SRModel:()=>xl,Swin2SRPreTrainedModel:()=>ua,SwinForImageClassification:()=>bl,SwinForSemanticSegmentation:()=>vl,SwinModel:()=>Ml,SwinPreTrainedModel:()=>Zs,T5ForConditionalGeneration:()=>je,T5Model:()=>xe,T5PreTrainedModel:()=>Me,TableTransformerForObjectDetection:()=>pl,TableTransformerModel:()=>dl,TableTransformerObjectDetectionOutput:()=>hl,TableTransformerPreTrainedModel:()=>aa,TokenClassifierOutput:()=>Sr,TrOCRForCausalLM:()=>Vc,TrOCRPreTrainedModel:()=>Nc,UltravoxModel:()=>$u,UltravoxPreTrainedModel:()=>Su,UniSpeechForCTC:()=>wc,UniSpeechForSequenceClassification:()=>Mc,UniSpeechModel:()=>yc,UniSpeechPreTrainedModel:()=>tn,UniSpeechSatForAudioFrameClassification:()=>Tc,UniSpeechSatForCTC:()=>vc,UniSpeechSatForSequenceClassification:()=>xc,UniSpeechSatModel:()=>bc,UniSpeechSatPreTrainedModel:()=>Is,ViTForImageClassification:()=>ds,ViTMAEModel:()=>Uo,ViTMAEPreTrainedModel:()=>Vo,ViTMSNForImageClassification:()=>Gn,ViTMSNModel:()=>Wo,ViTMSNPreTrainedModel:()=>Wn,ViTModel:()=>ci,ViTPreTrainedModel:()=>Ss,VisionEncoderDecoderModel:()=>Za,VitMatteForImageMatting:()=>Ko,VitMattePreTrainedModel:()=>Qn,VitPoseForPoseEstimation:()=>Bo,VitPosePreTrainedModel:()=>zo,VitsModel:()=>Ea,VitsModelOutput:()=>Fh,VitsPreTrainedModel:()=>Yc,Wav2Vec2BertForCTC:()=>Pc,Wav2Vec2BertForSequenceClassification:()=>Cc,Wav2Vec2BertModel:()=>Ec,Wav2Vec2BertPreTrainedModel:()=>rn,Wav2Vec2ForAudioFrameClassification:()=>hc,Wav2Vec2ForCTC:()=>dc,Wav2Vec2ForSequenceClassification:()=>pc,Wav2Vec2Model:()=>uc,Wav2Vec2PreTrainedModel:()=>Ni,WavLMForAudioFrameClassification:()=>Dc,WavLMForCTC:()=>Ac,WavLMForSequenceClassification:()=>Oc,WavLMForXVector:()=>Fc,WavLMModel:()=>Ic,WavLMPreTrainedModel:()=>hs,WeSpeakerResNetModel:()=>gc,WeSpeakerResNetPreTrainedModel:()=>_c,WhisperForConditionalGeneration:()=>kn,WhisperModel:()=>Td,WhisperPreTrainedModel:()=>Xa,XLMForQuestionAnswering:()=>qa,XLMForSequenceClassification:()=>Us,XLMForTokenClassification:()=>Ha,XLMModel:()=>li,XLMPreTrainedModel:()=>Dr,XLMRobertaForMaskedLM:()=>bd,XLMRobertaForQuestionAnswering:()=>Ws,XLMRobertaForSequenceClassification:()=>$n,XLMRobertaForTokenClassification:()=>vd,XLMRobertaModel:()=>Sn,XLMRobertaPreTrainedModel:()=>Ts,XLMWithLMHeadModel:()=>Ka,XVectorOutput:()=>Ah,YolosForObjectDetection:()=>tc,YolosModel:()=>ec,YolosObjectDetectionOutput:()=>rc,YolosPreTrainedModel:()=>ya});var n=t("./src/configs.js"),o=t("./src/backends/onnx.js"),i=t("./src/utils/dtypes.js"),u=t("./src/utils/generic.js"),d=t("./src/utils/core.js"),h=t("./src/utils/hub.js"),f=t("./src/utils/constants.js"),y=t("./src/generation/logits_process.js"),m=t("./src/generation/configuration_utils.js"),g=t("./src/utils/tensor.js"),v=t("./src/utils/image.js"),b=t("./src/utils/maths.js"),k=t("./src/generation/stopping_criteria.js"),z=t("./src/generation/logits_sampler.js"),E=t("./src/env.js"),T=t("./src/models/whisper/generation_whisper.js"),D=t("./src/models/whisper/common_whisper.js");const I={EncoderOnly:0,EncoderDecoder:1,Seq2Seq:2,Vision2Seq:3,DecoderOnly:4,MaskGeneration:5,ImageTextToText:6,Musicgen:7,MultiModality:8,Phi3V:9,AudioTextToText:10,AutoEncoder:11},$=new Map,P=new Map,S=new Map;async function O(C,F,N){var $r;let me=(($r=N.config)==null?void 0:$r["transformers.js_config"])??{},ke=N.device??me.device;ke&&typeof ke!="string"&&(ke.hasOwnProperty(F)?ke=ke[F]:(console.warn(`device not specified for "${F}". Using the default device.`),ke=null));const Se=ke??(E.apis.IS_NODE_ENV?"cpu":"wasm"),ze=(0,o.deviceToExecutionProviders)(Se),Ke=me.device_config??{};Ke.hasOwnProperty(Se)&&(me={...me,...Ke[Se]});let Ze=N.dtype??me.dtype;if(typeof Ze!="string"&&(Ze&&Ze.hasOwnProperty(F)?Ze=Ze[F]:(Ze=i.DEFAULT_DEVICE_DTYPE_MAPPING[Se]??i.DATA_TYPES.fp32,console.warn(`dtype not specified for "${F}". Using the default dtype (${Ze}) for this device (${Se}).`))),Ze===i.DATA_TYPES.auto){let Bt=me.dtype;typeof Bt!="string"&&(Bt=Bt==null?void 0:Bt[F]),Bt&&Bt!==i.DATA_TYPES.auto&&i.DATA_TYPES.hasOwnProperty(Bt)?Ze=Bt:Ze=i.DEFAULT_DEVICE_DTYPE_MAPPING[Se]??i.DATA_TYPES.fp32}const at=Ze;if(i.DEFAULT_DTYPE_SUFFIX_MAPPING.hasOwnProperty(at)){if(at===i.DATA_TYPES.fp16&&Se==="webgpu"&&!await(0,i.isWebGpuFp16Supported)())throw new Error(`The device (${Se}) does not support fp16.`)}else throw new Error(`Invalid dtype: ${at}. Should be one of: ${Object.keys(i.DATA_TYPES).join(", ")}`);const vt=me.kv_cache_dtype,kt=vt?typeof vt=="string"?vt:vt[at]??"float32":void 0;if(kt&&!["float32","float16"].includes(kt))throw new Error(`Invalid kv_cache_dtype: ${kt}. Should be one of: float32, float16`);const Mt={dtype:at,kv_cache_dtype:kt,device:Se},zt=i.DEFAULT_DTYPE_SUFFIX_MAPPING[at],Ct=`${F}${zt}.onnx`,Ot=`${N.subfolder??""}/${Ct}`,ft={...N.session_options};ft.executionProviders??(ft.executionProviders=ze);const Vt=me.free_dimension_overrides;Vt?ft.freeDimensionOverrides??(ft.freeDimensionOverrides=Vt):Se.startsWith("webnn")&&!ft.freeDimensionOverrides&&console.warn(`WebNN does not currently support dynamic shapes and requires 'free_dimension_overrides' to be set in config.json, preferably as a field within config["transformers.js_config"]["device_config"]["${Se}"]. When 'free_dimension_overrides' is not set, you may experience significant performance degradation.`);const rr=E.apis.IS_NODE_ENV&&E.env.useFSCache,ar=(0,h.getModelFile)(C,Ot,!0,N,rr),fr=N.use_external_data_format??me.use_external_data_format;let hr=[];if(fr){let Bt;typeof fr=="object"?fr.hasOwnProperty(Ct)?Bt=fr[Ct]:fr.hasOwnProperty(F)?Bt=fr[F]:Bt=!1:Bt=fr;const Er=+Bt;if(Er>h.MAX_EXTERNAL_DATA_CHUNKS)throw new Error(`The number of external data chunks (${Er}) exceeds the maximum allowed value (${h.MAX_EXTERNAL_DATA_CHUNKS}).`);for(let ui=0;ui<Er;++ui){const ms=`${Ct}_data${ui===0?"":"_"+ui}`,Qr=`${N.subfolder??""}/${ms}`;hr.push(new Promise(async(fi,Aa)=>{const _s=await(0,h.getModelFile)(C,Qr,!0,N,rr);fi(_s instanceof Uint8Array?{path:ms,data:_s}:ms)}))}}else ft.externalData!==void 0&&(hr=ft.externalData.map(async Bt=>{if(typeof Bt.data=="string"){const Er=await(0,h.getModelFile)(C,Bt.data,!0,N);return{...Bt,data:Er}}return Bt}));if(hr.length>0){const Bt=await Promise.all(hr);E.apis.IS_NODE_ENV||(ft.externalData=Bt)}if(Se==="webgpu"){const Bt=(0,n.getKeyValueShapes)(N.config,{prefix:"present"});if(Object.keys(Bt).length>0&&!(0,o.isONNXProxy)()){const Er={};for(const ui in Bt)Er[ui]="gpu-buffer";ft.preferredOutputLocation=Er}}return{buffer_or_path:await ar,session_options:ft,session_config:Mt}}async function R(C,F,N){return Object.fromEntries(await Promise.all(Object.keys(F).map(async me=>{const{buffer_or_path:ke,session_options:Se,session_config:ze}=await O(C,F[me],N),Ke=await(0,o.createInferenceSession)(ke,Se,ze);return[me,Ke]})))}async function G(C,F,N){return Object.fromEntries(await Promise.all(Object.keys(F).map(async me=>{const ke=await(0,h.getModelJSON)(C,F[me],!1,N);return[me,ke]})))}function ee(C,F){const N=Object.create(null),me=[];for(const ze of C.inputNames){const Ke=F[ze];if(!(Ke instanceof g.Tensor)){me.push(ze);continue}N[ze]=(0,o.isONNXProxy)()?Ke.clone():Ke}if(me.length>0)throw new Error(`An error occurred during model execution: "Missing the following inputs: ${me.join(", ")}.`);const ke=Object.keys(F).length,Se=C.inputNames.length;if(ke>Se){let ze=Object.keys(F).filter(Ke=>!C.inputNames.includes(Ke));console.warn(`WARNING: Too many inputs were provided (${ke} > ${Se}). The following inputs will be ignored: "${ze.join(", ")}".`)}return N}let ie=Promise.resolve();async function H(C,F){const N=ee(C,F);try{const me=Object.fromEntries(Object.entries(N).map(([ze,Ke])=>[ze,Ke.ort_tensor])),ke=()=>C.run(me),Se=await(E.apis.IS_BROWSER_ENV||E.apis.IS_WEBWORKER_ENV?ie=ie.then(ke):ke());return ce(Se)}catch(me){const ke=Object.fromEntries(Object.entries(N).map(([Se,ze])=>{const Ke={type:ze.type,dims:ze.dims,location:ze.location};return Ke.location!=="gpu-buffer"&&(Ke.data=ze.data),[Se,Ke]}));throw console.error(`An error occurred during model execution: "${me}".`),console.error("Inputs given to model:",ke),me}}function ce(C){for(let F in C)(0,o.isONNXTensor)(C[F])?C[F]=new g.Tensor(C[F]):typeof C[F]=="object"&&ce(C[F]);return C}function re(C){if(C instanceof g.Tensor)return C;if(C.length===0)throw Error("items must be non-empty");if(Array.isArray(C[0])){if(C.some(F=>F.length!==C[0].length))throw Error("Unable to create tensor, you should probably activate truncation and/or padding with 'padding=True' and/or 'truncation=True' to have batched tensors with the same length.");return new g.Tensor("int64",BigInt64Array.from(C.flat().map(F=>BigInt(F))),[C.length,C[0].length])}else return new g.Tensor("int64",BigInt64Array.from(C.map(F=>BigInt(F))),[1,C.length])}function se(C){return new g.Tensor("bool",[C],[1])}async function _e(C,F){let{encoder_outputs:N,input_ids:me,decoder_input_ids:ke,...Se}=F;if(!N){const Ke=(0,d.pick)(F,C.sessions.model.inputNames);N=(await ae(C,Ke)).last_hidden_state}return Se.input_ids=ke,Se.encoder_hidden_states=N,C.sessions.decoder_model_merged.inputNames.includes("encoder_attention_mask")&&(Se.encoder_attention_mask=F.attention_mask),await Te(C,Se,!0)}async function ae(C,F){const N=C.sessions.model,me=(0,d.pick)(F,N.inputNames);if(N.inputNames.includes("inputs_embeds")&&!me.inputs_embeds){if(!F.input_ids)throw new Error("Both `input_ids` and `inputs_embeds` are missing in the model inputs.");me.inputs_embeds=await C.encode_text({input_ids:F.input_ids})}if(N.inputNames.includes("token_type_ids")&&!me.token_type_ids){if(!me.input_ids)throw new Error("Both `input_ids` and `token_type_ids` are missing in the model inputs.");me.token_type_ids=(0,g.zeros_like)(me.input_ids)}if(N.inputNames.includes("pixel_mask")&&!me.pixel_mask){if(!me.pixel_values)throw new Error("Both `pixel_values` and `pixel_mask` are missing in the model inputs.");const ke=me.pixel_values.dims;me.pixel_mask=(0,g.ones)([ke[0],ke[2],ke[3]])}return await H(N,me)}async function Ce(C,F){const N=await C.encode(F);return await C.decode(N)}async function Te(C,F,N=!1){const me=C.sessions[N?"decoder_model_merged":"model"],{past_key_values:ke,...Se}=F;if(me.inputNames.includes("use_cache_branch")&&(Se.use_cache_branch=se(!!ke)),me.inputNames.includes("position_ids")&&Se.attention_mask&&!Se.position_ids){const Ke=["paligemma","gemma3_text","gemma3"].includes(C.config.model_type)?1:0;Se.position_ids=et(Se,ke,Ke)}C.addPastKeyValues(Se,ke);const ze=(0,d.pick)(Se,me.inputNames);return await H(me,ze)}function q({modality_token_id:C,inputs_embeds:F,modality_features:N,input_ids:me,attention_mask:ke}){const Se=me.tolist().map(at=>at.reduce((vt,kt,Mt)=>(kt==C&&vt.push(Mt),vt),[])),ze=Se.reduce((at,vt)=>at+vt.length,0),Ke=N.dims[0];if(ze!==Ke)throw new Error(`Number of tokens and features do not match: tokens: ${ze}, features ${Ke}`);let Ze=0;for(let at=0;at<Se.length;++at){const vt=Se[at],kt=F[at];for(let Mt=0;Mt<vt.length;++Mt)kt[vt[Mt]].data.set(N[Ze++].data)}return{inputs_embeds:F,attention_mask:ke}}function B({image_token_id:C,inputs_embeds:F,image_features:N,input_ids:me,attention_mask:ke}){return q({modality_token_id:C,inputs_embeds:F,modality_features:N,input_ids:me,attention_mask:ke})}function Q({audio_token_id:C,inputs_embeds:F,audio_features:N,input_ids:me,attention_mask:ke}){return q({modality_token_id:C,inputs_embeds:F,modality_features:N,input_ids:me,attention_mask:ke})}async function oe(C,{encode_function:F,merge_function:N,modality_input_name:me,modality_output_name:ke,input_ids:Se=null,attention_mask:ze=null,position_ids:Ke=null,inputs_embeds:Ze=null,past_key_values:at=null,generation_config:vt=null,logits_processor:kt=null,...Mt}){const zt=Mt[me];if(!Ze){if(Ze=await C.encode_text({input_ids:Se,...Mt}),zt&&Se.dims[1]!==1){const Ot=await F({[me]:zt,...Mt});({inputs_embeds:Ze,attention_mask:ze}=N({[ke]:Ot,inputs_embeds:Ze,input_ids:Se,attention_mask:ze}))}else if(at&&zt&&Se.dims[1]===1){const Ot=Se.dims[1],ft=Object.values(at)[0].dims.at(-2);ze=(0,g.cat)([(0,g.ones)([Se.dims[0],ft]),ze.slice(null,[ze.dims[1]-Ot,ze.dims[1]])],1)}}if(!Ke&&C.config.model_type==="qwen2_vl"){const{image_grid_thw:Ot,video_grid_thw:ft}=Mt;[Ke]=C.get_rope_index(Se,Ot,ft,ze)}return await Te(C,{inputs_embeds:Ze,past_key_values:at,attention_mask:ze,position_ids:Ke,generation_config:vt,logits_processor:kt},!0)}async function Ie(C,F){return await oe(C,{...F,modality_input_name:"audio_values",modality_output_name:"audio_features",encode_function:C.encode_audio.bind(C),merge_function:C._merge_input_ids_with_audio_features.bind(C)})}async function ye(C,F){return await oe(C,{...F,modality_input_name:"pixel_values",modality_output_name:"image_features",encode_function:C.encode_image.bind(C),merge_function:C._merge_input_ids_with_image_features.bind(C)})}function be(C,F=0){const[N,me]=C.dims,ke=C.data,Se=new BigInt64Array(ke.length);for(let ze=0;ze<N;++ze){const Ke=ze*me;let Ze=BigInt(F);for(let at=0;at<me;++at){const vt=Ke+at;ke[vt]===0n?Se[vt]=BigInt(1):(Se[vt]=Ze,Ze+=ke[vt])}}return{data:Se,dims:C.dims}}function et(C,F=null,N=0){const{input_ids:me,inputs_embeds:ke,attention_mask:Se}=C,{data:ze,dims:Ke}=be(Se,N);let Ze=new g.Tensor("int64",ze,Ke);if(F){const at=-(me??ke).dims.at(1);Ze=Ze.slice(null,[at,null])}return Ze}function tt(C,F,N,me){if(N.past_key_values){const ke=Object.values(N.past_key_values)[0].dims.at(-2),{input_ids:Se,attention_mask:ze}=N;if(!(ze&&ze.dims[1]>Se.dims[1])){if(ke<Se.dims[1])N.input_ids=Se.slice(null,[ke,null]);else if(C.config.image_token_index!=null&&Se.data.some(Ke=>Ke==C.config.image_token_index)){const Ke=C.config.num_image_tokens;if(!Ke)throw new Error("`num_image_tokens` is missing in the model configuration.");const Ze=Se.dims[1]-(ke-Ke);N.input_ids=Se.slice(null,[-Ze,null]),N.attention_mask=(0,g.ones)([1,ke+Ze])}}}return N}function Ee(C,F,N,me){return N.past_key_values&&(F=F.map(ke=>[ke.at(-1)])),{...N,decoder_input_ids:re(F)}}function Y(C,...F){return C.config.is_encoder_decoder?Ee(C,...F):tt(C,...F)}function pe(C,F,N,me){const ke=!!N.past_key_values;return me.guidance_scale!==null&&me.guidance_scale>1&&(ke?N.input_ids=(0,g.cat)([N.input_ids,N.input_ids],0):(N.input_ids=(0,g.cat)([N.input_ids,(0,g.full_like)(N.input_ids,BigInt(me.pad_token_id))],0),N.attention_mask=(0,g.cat)([N.attention_mask,(0,g.full_like)(N.attention_mask,0n)],0))),(ke||!N.pixel_values)&&(N.pixel_values=(0,g.full)([0,0,3,384,384],1)),ke&&(N.images_seq_mask=new g.Tensor("bool",new Array(1).fill(!0).fill(!1,0,1),[1,1]),N.images_emb_mask=new g.Tensor("bool",new Array(0).fill(!1),[1,1,0])),N}class X extends u.Callable{constructor(N,me,ke){super();le(this,"main_input_name","input_ids");le(this,"forward_params",["input_ids","attention_mask"]);this.config=N,this.sessions=me,this.configs=ke;const Se=S.get(this.constructor),ze=$.get(Se);switch(this.can_generate=!1,this._forward=null,this._prepare_inputs_for_generation=null,ze){case I.DecoderOnly:this.can_generate=!0,this._forward=Te,this._prepare_inputs_for_generation=tt;break;case I.Seq2Seq:case I.Vision2Seq:case I.Musicgen:this.can_generate=!0,this._forward=_e,this._prepare_inputs_for_generation=Ee;break;case I.EncoderDecoder:this._forward=_e;break;case I.ImageTextToText:this.can_generate=!0,this._forward=ye,this._prepare_inputs_for_generation=Y;break;case I.AudioTextToText:this.can_generate=!0,this._forward=Ie,this._prepare_inputs_for_generation=Y;break;case I.Phi3V:this.can_generate=!0,this._prepare_inputs_for_generation=Y;break;case I.MultiModality:this.can_generate=!0,this._prepare_inputs_for_generation=pe;break;case I.AutoEncoder:this._forward=Ce;break;default:this._forward=ae;break}this.can_generate&&this.forward_params.push("past_key_values"),this.custom_config=this.config["transformers.js_config"]??{}}async dispose(){var me;const N=[];for(const ke of Object.values(this.sessions))(me=ke==null?void 0:ke.handler)!=null&&me.dispose&&N.push(ke.handler.dispose());return await Promise.all(N)}static async from_pretrained(N,{progress_callback:me=null,config:ke=null,cache_dir:Se=null,local_files_only:ze=!1,revision:Ke="main",model_file_name:Ze=null,subfolder:at="onnx",device:vt=null,dtype:kt=null,use_external_data_format:Mt=null,session_options:zt={}}={}){let Ct={progress_callback:me,config:ke,cache_dir:Se,local_files_only:ze,revision:Ke,model_file_name:Ze,subfolder:at,device:vt,dtype:kt,use_external_data_format:Mt,session_options:zt};const Ot=S.get(this),ft=$.get(Ot);ke=Ct.config=await n.AutoConfig.from_pretrained(N,Ct);let Vt;if(ft===I.DecoderOnly)Vt=await Promise.all([R(N,{model:Ct.model_file_name??"model"},Ct),G(N,{generation_config:"generation_config.json"},Ct)]);else if(ft===I.Seq2Seq||ft===I.Vision2Seq)Vt=await Promise.all([R(N,{model:"encoder_model",decoder_model_merged:"decoder_model_merged"},Ct),G(N,{generation_config:"generation_config.json"},Ct)]);else if(ft===I.MaskGeneration)Vt=await Promise.all([R(N,{model:"vision_encoder",prompt_encoder_mask_decoder:"prompt_encoder_mask_decoder"},Ct)]);else if(ft===I.EncoderDecoder)Vt=await Promise.all([R(N,{model:"encoder_model",decoder_model_merged:"decoder_model_merged"},Ct)]);else if(ft===I.ImageTextToText){const rr={embed_tokens:"embed_tokens",vision_encoder:"vision_encoder",decoder_model_merged:"decoder_model_merged"};ke.is_encoder_decoder&&(rr.model="encoder_model"),Vt=await Promise.all([R(N,rr,Ct),G(N,{generation_config:"generation_config.json"},Ct)])}else if(ft===I.AudioTextToText){const rr={embed_tokens:"embed_tokens",audio_encoder:"audio_encoder",decoder_model_merged:"decoder_model_merged"};Vt=await Promise.all([R(N,rr,Ct),G(N,{generation_config:"generation_config.json"},Ct)])}else if(ft===I.Musicgen)Vt=await Promise.all([R(N,{model:"text_encoder",decoder_model_merged:"decoder_model_merged",encodec_decode:"encodec_decode"},Ct),G(N,{generation_config:"generation_config.json"},Ct)]);else if(ft===I.MultiModality)Vt=await Promise.all([R(N,{prepare_inputs_embeds:"prepare_inputs_embeds",model:"language_model",lm_head:"lm_head",gen_head:"gen_head",gen_img_embeds:"gen_img_embeds",image_decode:"image_decode"},Ct),G(N,{generation_config:"generation_config.json"},Ct)]);else if(ft===I.Phi3V)Vt=await Promise.all([R(N,{prepare_inputs_embeds:"prepare_inputs_embeds",model:"model",vision_encoder:"vision_encoder"},Ct),G(N,{generation_config:"generation_config.json"},Ct)]);else if(ft===I.AutoEncoder)Vt=await Promise.all([R(N,{encoder_model:"encoder_model",decoder_model:"decoder_model"},Ct)]);else{if(ft!==I.EncoderOnly){const rr=Ot??(ke==null?void 0:ke.model_type);rr!=="custom"&&console.warn(`Model type for '${rr}' not found, assuming encoder-only architecture. Please report this at ${f.GITHUB_ISSUE_URL}.`)}Vt=await Promise.all([R(N,{model:Ct.model_file_name??"model"},Ct)])}return new this(ke,...Vt)}async _call(N){return await this.forward(N)}async forward(N){return await this._forward(this,N)}get generation_config(){var N;return((N=this.configs)==null?void 0:N.generation_config)??null}_get_logits_warper(N){const me=new y.LogitsProcessorList;return N.temperature!==null&&N.temperature!==1&&me.push(new y.TemperatureLogitsWarper(N.temperature)),N.top_k!==null&&N.top_k!==0&&me.push(new y.TopKLogitsWarper(N.top_k)),N.top_p!==null&&N.top_p<1&&me.push(new y.TopPLogitsWarper(N.top_p)),me}_get_logits_processor(N,me,ke=null){const Se=new y.LogitsProcessorList;if(N.repetition_penalty!==null&&N.repetition_penalty!==1&&Se.push(new y.RepetitionPenaltyLogitsProcessor(N.repetition_penalty)),N.no_repeat_ngram_size!==null&&N.no_repeat_ngram_size>0&&Se.push(new y.NoRepeatNGramLogitsProcessor(N.no_repeat_ngram_size)),N.bad_words_ids!==null&&Se.push(new y.NoBadWordsLogitsProcessor(N.bad_words_ids,N.eos_token_id)),N.min_length!==null&&N.eos_token_id!==null&&N.min_length>0&&Se.push(new y.MinLengthLogitsProcessor(N.min_length,N.eos_token_id)),N.min_new_tokens!==null&&N.eos_token_id!==null&&N.min_new_tokens>0&&Se.push(new y.MinNewTokensLengthLogitsProcessor(me,N.min_new_tokens,N.eos_token_id)),N.forced_bos_token_id!==null&&Se.push(new y.ForcedBOSTokenLogitsProcessor(N.forced_bos_token_id)),N.forced_eos_token_id!==null&&Se.push(new y.ForcedEOSTokenLogitsProcessor(N.max_length,N.forced_eos_token_id)),N.begin_suppress_tokens!==null){const ze=me>1||N.forced_bos_token_id===null?me:me+1;Se.push(new y.SuppressTokensAtBeginLogitsProcessor(N.begin_suppress_tokens,ze))}return N.guidance_scale!==null&&N.guidance_scale>1&&Se.push(new y.ClassifierFreeGuidanceLogitsProcessor(N.guidance_scale)),ke!==null&&Se.extend(ke),Se}_prepare_generation_config(N,me,ke=m.GenerationConfig){const Se={...this.config};for(const Ke of["decoder","generator","text_config"])Ke in Se&&Object.assign(Se,Se[Ke]);const ze=new ke(Se);return Object.assign(ze,this.generation_config??{}),N&&Object.assign(ze,N),me&&Object.assign(ze,(0,d.pick)(me,Object.getOwnPropertyNames(ze))),ze}_get_stopping_criteria(N,me=null){const ke=new k.StoppingCriteriaList;return N.max_length!==null&&ke.push(new k.MaxLengthCriteria(N.max_length,this.config.max_position_embeddings??null)),N.eos_token_id!==null&&ke.push(new k.EosTokenCriteria(N.eos_token_id)),me&&ke.extend(me),ke}_validate_model_class(){if(!this.can_generate){const N=[Gu,Ku,Wu,Uu],me=S.get(this.constructor),ke=new Set,Se=this.config.model_type;for(const Ke of N){const Ze=Ke.get(Se);Ze&&ke.add(Ze[0])}let ze=`The current model class (${me}) is not compatible with \`.generate()\`, as it doesn't have a language model head.`;throw ke.size>0&&(ze+=` Please use the following class instead: ${[...ke].join(", ")}`),Error(ze)}}prepare_inputs_for_generation(...N){return this._prepare_inputs_for_generation(this,...N)}_update_model_kwargs_for_generation({generated_input_ids:N,outputs:me,model_inputs:ke,is_encoder_decoder:Se}){return ke.past_key_values=this.getPastKeyValues(me,ke.past_key_values),ke.input_ids=new g.Tensor("int64",N.flat(),[N.length,1]),Se||(ke.attention_mask=(0,g.cat)([ke.attention_mask,(0,g.ones)([ke.attention_mask.dims[0],1])],1)),ke.position_ids=null,ke}_prepare_model_inputs({inputs:N,bos_token_id:me,model_kwargs:ke}){const Se=(0,d.pick)(ke,this.forward_params),ze=this.main_input_name;if(ze in Se){if(N)throw new Error("`inputs`: {inputs}` were passed alongside {input_name} which is not allowed. Make sure to either pass {inputs} or {input_name}=...")}else Se[ze]=N;return{inputs_tensor:Se[ze],model_inputs:Se,model_input_name:ze}}async _prepare_encoder_decoder_kwargs_for_generation({inputs_tensor:N,model_inputs:me,model_input_name:ke,generation_config:Se}){if(this.sessions.model.inputNames.includes("inputs_embeds")&&!me.inputs_embeds&&"_prepare_inputs_embeds"in this){const{input_ids:Ke,pixel_values:Ze,attention_mask:at,...vt}=me,kt=await this._prepare_inputs_embeds(me);me={...vt,...(0,d.pick)(kt,["inputs_embeds","attention_mask"])}}let{last_hidden_state:ze}=await ae(this,me);if(Se.guidance_scale!==null&&Se.guidance_scale>1)ze=(0,g.cat)([ze,(0,g.full_like)(ze,0)],0),"attention_mask"in me&&(me.attention_mask=(0,g.cat)([me.attention_mask,(0,g.zeros_like)(me.attention_mask)],0));else if(me.decoder_input_ids){const Ke=re(me.decoder_input_ids).dims[0];if(Ke!==ze.dims[0]){if(ze.dims[0]!==1)throw new Error(`The encoder outputs have a different batch size (${ze.dims[0]}) than the decoder inputs (${Ke}).`);ze=(0,g.cat)(Array.from({length:Ke},()=>ze),0)}}return me.encoder_outputs=ze,me}_prepare_decoder_input_ids_for_generation({batch_size:N,model_input_name:me,model_kwargs:ke,decoder_start_token_id:Se,bos_token_id:ze,generation_config:Ke}){let{decoder_input_ids:Ze,...at}=ke;if(!(Ze instanceof g.Tensor)){if(Ze)Array.isArray(Ze[0])||(Ze=Array.from({length:N},()=>Ze));else if(Se??(Se=ze),this.config.model_type==="musicgen")Ze=Array.from({length:N*this.config.decoder.num_codebooks},()=>[Se]);else if(Array.isArray(Se)){if(Se.length!==N)throw new Error(`\`decoder_start_token_id\` expcted to have length ${N} but got ${Se.length}`);Ze=Se}else Ze=Array.from({length:N},()=>[Se]);Ze=re(Ze)}return ke.decoder_attention_mask=(0,g.ones_like)(Ze),{input_ids:Ze,model_inputs:at}}async generate({inputs:N=null,generation_config:me=null,logits_processor:ke=null,stopping_criteria:Se=null,streamer:ze=null,...Ke}){this._validate_model_class(),me=this._prepare_generation_config(me,Ke);let{inputs_tensor:Ze,model_inputs:at,model_input_name:vt}=this._prepare_model_inputs({inputs:N,model_kwargs:Ke});const kt=this.config.is_encoder_decoder;kt&&("encoder_outputs"in at||(at=await this._prepare_encoder_decoder_kwargs_for_generation({inputs_tensor:Ze,model_inputs:at,model_input_name:vt,generation_config:me})));let Mt;kt?{input_ids:Mt,model_inputs:at}=this._prepare_decoder_input_ids_for_generation({batch_size:at[vt].dims.at(0),model_input_name:vt,model_kwargs:at,decoder_start_token_id:me.decoder_start_token_id,bos_token_id:me.bos_token_id,generation_config:me}):Mt=at[vt];let zt=Mt.dims.at(-1);me.max_new_tokens!==null&&(me.max_length=zt+me.max_new_tokens);const Ct=this._get_logits_processor(me,zt,ke),Ot=this._get_stopping_criteria(me,Se),ft=at[vt].dims.at(0),Vt=z.LogitsSampler.getSampler(me),rr=new Array(ft).fill(0),ar=Mt.tolist();ze&&ze.put(ar);let fr,hr={};for(;;){if(at=this.prepare_inputs_for_generation(ar,at,me),fr=await this.forward(at),me.output_attentions&&me.return_dict_in_generate){const Qr=this.getAttentions(fr);for(const fi in Qr)fi in hr||(hr[fi]=[]),hr[fi].push(Qr[fi])}const Bt=fr.logits.slice(null,-1,null),Er=Ct(ar,Bt),ui=[];for(let Qr=0;Qr<Er.dims.at(0);++Qr){const fi=Er[Qr],Aa=await Vt(fi);for(const[_s,Oa]of Aa){const Fa=BigInt(_s);rr[Qr]+=Oa,ar[Qr].push(Fa),ui.push([Fa]);break}}if(ze&&ze.put(ui),Ot(ar).every(Qr=>Qr))break;at=this._update_model_kwargs_for_generation({generated_input_ids:ui,outputs:fr,model_inputs:at,is_encoder_decoder:kt})}ze&&ze.end();const br=this.getPastKeyValues(fr,at.past_key_values,!0),$r=new g.Tensor("int64",ar.flat(),[ar.length,ar[0].length]);if(me.return_dict_in_generate)return{sequences:$r,past_key_values:br,...hr};for(const Bt of Object.values(fr))Bt.location==="gpu-buffer"&&Bt.dispose();return $r}getPastKeyValues(N,me,ke=!1){const Se=Object.create(null);for(const ze in N)if(ze.startsWith("present")){const Ke=ze.replace("present","past_key_values"),Ze=ze.includes("encoder");if(Ze&&me?Se[Ke]=me[Ke]:Se[Ke]=N[ze],me&&(!Ze||ke)){const at=me[Ke];at.location==="gpu-buffer"&&at.dispose()}}return Se}getAttentions(N){const me={};for(const ke of["cross_attentions","encoder_attentions","decoder_attentions"])for(const Se in N)Se.startsWith(ke)&&(ke in me||(me[ke]=[]),me[ke].push(N[Se]));return me}addPastKeyValues(N,me){var ke,Se,ze;if(me)Object.assign(N,me);else{const Ke=this.sessions.decoder_model_merged??this.sessions.model,Ze=((ke=Ke==null?void 0:Ke.config)==null?void 0:ke.kv_cache_dtype)??"float32",at=Ze==="float16"?new g.DataTypeMap.float16:[],vt=((ze=(Se=N[this.main_input_name]??N.attention_mask)==null?void 0:Se.dims)==null?void 0:ze[0])??1,kt=(0,n.getKeyValueShapes)(this.config,{batch_size:vt});for(const Mt in kt)N[Mt]=new g.Tensor(Ze,at,kt[Mt])}}async encode_image({pixel_values:N}){const me=(await H(this.sessions.vision_encoder,{pixel_values:N})).image_features;return this.config.num_image_tokens||(console.warn(`The number of image tokens was not set in the model configuration. Setting it to the number of features detected by the vision encoder (${me.dims[1]}).`),this.config.num_image_tokens=me.dims[1]),me}async encode_text({input_ids:N}){return(await H(this.sessions.embed_tokens,{input_ids:N})).inputs_embeds}async encode_audio({audio_values:N}){return(await H(this.sessions.audio_encoder,{audio_values:N})).audio_features}}class Ae{}class Oe extends Ae{constructor({last_hidden_state:F,hidden_states:N=null,attentions:me=null}){super(),this.last_hidden_state=F,this.hidden_states=N,this.attentions=me}}class Ne extends X{}class Be extends Ne{}class Ve extends Ne{async _call(F){return new Or(await super._call(F))}}class He extends Ne{async _call(F){return new At(await super._call(F))}}class Xe extends Ne{async _call(F){return new Sr(await super._call(F))}}class st extends Ne{async _call(F){return new Vr(await super._call(F))}}class nt extends X{}class We extends nt{}class rt extends nt{async _call(F){return new Or(await super._call(F))}}class pt extends nt{async _call(F){return new At(await super._call(F))}}class _t extends nt{async _call(F){return new Sr(await super._call(F))}}class ct extends X{}class dt extends ct{}class Le extends X{}class xt extends Le{}class Nt extends Le{async _call(F){return new Or(await super._call(F))}}class cr extends Le{async _call(F){return new At(await super._call(F))}}class Kt extends Le{async _call(F){return new Sr(await super._call(F))}}class wr extends Le{async _call(F){return new Vr(await super._call(F))}}class zr extends X{}class $t extends zr{}class Ti extends zr{async _call(F){return new Or(await super._call(F))}}class U extends zr{async _call(F){return new At(await super._call(F))}}class ge extends zr{async _call(F){return new Sr(await super._call(F))}}class K extends zr{async _call(F){return new Vr(await super._call(F))}}class ue extends X{}class $e extends ue{}class Ge extends ue{async _call(F){return new Or(await super._call(F))}}class De extends ue{async _call(F){return new At(await super._call(F))}}class Dt extends ue{async _call(F){return new Sr(await super._call(F))}}class Ut extends ue{async _call(F){return new Vr(await super._call(F))}}class Et extends X{}class jt extends Et{}class gt extends Et{async _call(F){return new Or(await super._call(F))}}class Ht extends Et{async _call(F){return new At(await super._call(F))}}class nr extends Et{async _call(F){return new Sr(await super._call(F))}}class ri extends Et{async _call(F){return new Vr(await super._call(F))}}class pr extends X{}class Jr extends pr{}class Ai extends pr{async _call(F){return new Or(await super._call(F))}}class Ji extends pr{async _call(F){return new At(await super._call(F))}}class Oi extends pr{async _call(F){return new Sr(await super._call(F))}}class ns extends pr{async _call(F){return new Vr(await super._call(F))}}class ii extends X{}class Yi extends ii{}class _i extends ii{async _call(F){return new Or(await super._call(F))}}class Fi extends ii{async _call(F){return new At(await super._call(F))}}class er extends ii{async _call(F){return new Sr(await super._call(F))}}class Zi extends ii{async _call(F){return new Vr(await super._call(F))}}class Gr extends X{}class Kr extends Gr{}class Hr extends Gr{async _call(F){return new At(await super._call(F))}}class es extends Gr{async _call(F){return new Sr(await super._call(F))}}class si extends Gr{async _call(F){return new Vr(await super._call(F))}}class qe extends Gr{async _call(F){return new Or(await super._call(F))}}class lt extends X{}class yt extends lt{}class ur extends lt{async _call(F){return new Or(await super._call(F))}}class gi extends lt{async _call(F){return new At(await super._call(F))}}class Di extends lt{async _call(F){return new Sr(await super._call(F))}}class ni extends X{}class Li extends ni{}class ji extends ni{async _call(F){return new Or(await super._call(F))}}class zi extends ni{async _call(F){return new At(await super._call(F))}}class qr extends ni{async _call(F){return new Vr(await super._call(F))}}class Bi extends X{}class Yr extends Bi{}class as extends Bi{async _call(F){return new Or(await super._call(F))}}class os extends Bi{async _call(F){return new At(await super._call(F))}}class ai extends Bi{async _call(F){return new Sr(await super._call(F))}}class xs extends Bi{async _call(F){return new Vr(await super._call(F))}}class ts extends X{}class pi extends ts{}class Vs extends ts{async _call(F){return new Or(await super._call(F))}}class ls extends ts{async _call(F){return new At(await super._call(F))}}class oi extends ts{async _call(F){return new Vr(await super._call(F))}}class Ei extends X{}class ve extends Ei{}class j extends Ei{async _call(F){return new At(await super._call(F))}}class J extends Ei{async _call(F){return new Vr(await super._call(F))}}class ne extends Ei{async _call(F){return new Or(await super._call(F))}}class Me extends X{constructor(){super(...arguments);le(this,"forward_params",["input_ids","attention_mask","encoder_outputs","decoder_input_ids","decoder_attention_mask","past_key_values"])}}class xe extends Me{}class je extends Me{}class Ye extends X{}class ot extends Ye{}class it extends Ye{}class ut extends X{}class It extends ut{}class Qt extends ut{}class qt extends X{}class _r extends qt{}class tr extends qt{}class Cr extends qt{async _call(F){return new At(await super._call(F))}}class kr extends X{}class Zr extends kr{}class Pi extends kr{}class Fr extends kr{async _call(F){return new At(await super._call(F))}}class Ci extends kr{}class sr extends X{}class Ir extends sr{}class Br extends sr{}class yi extends X{}class Ri extends yi{}class Rr extends yi{}class ei extends X{}class Mr extends ei{}class gr extends ei{async _call(F){return new Or(await super._call(F))}}class xr extends ei{async _call(F){return new At(await super._call(F))}}class Tr extends ei{async _call(F){return new Sr(await super._call(F))}}class Nr extends ei{async _call(F){return new Vr(await super._call(F))}}class Dr extends X{}class li extends Dr{}class Ka extends Dr{async _call(F){return new Or(await super._call(F))}}class Us extends Dr{async _call(F){return new At(await super._call(F))}}class Ha extends Dr{async _call(F){return new Sr(await super._call(F))}}class qa extends Dr{async _call(F){return new Vr(await super._call(F))}}class Ts extends X{}class Sn extends Ts{}class bd extends Ts{async _call(F){return new Or(await super._call(F))}}class $n extends Ts{async _call(F){return new At(await super._call(F))}}class vd extends Ts{async _call(F){return new Sr(await super._call(F))}}class Ws extends Ts{async _call(F){return new Vr(await super._call(F))}}class Qa extends X{}class xd extends Qa{}class Gs extends Qa{}class Xa extends X{constructor(){super(...arguments);le(this,"requires_attention_mask",!1);le(this,"main_input_name","input_features");le(this,"forward_params",["input_features","attention_mask","decoder_input_ids","decoder_attention_mask","past_key_values"])}}class Td extends Xa{}class kn extends Xa{_prepare_generation_config(F,N){return super._prepare_generation_config(F,N,T.WhisperGenerationConfig)}_retrieve_init_tokens(F){const N=[F.decoder_start_token_id];let me=F.language;const ke=F.task;if(F.is_multilingual){me||(console.warn("No language specified - defaulting to English (en)."),me="en");const ze=`<|${(0,D.whisper_language_to_code)(me)}|>`;N.push(F.lang_to_id[ze]),N.push(F.task_to_id[ke??"transcribe"])}else if(me||ke)throw new Error("Cannot specify `task` or `language` for an English-only model. If the model is intended to be multilingual, pass `is_multilingual=true` to generate, or update the generation config.");return!F.return_timestamps&&F.no_timestamps_token_id&&N.at(-1)!==F.no_timestamps_token_id?N.push(F.no_timestamps_token_id):F.return_timestamps&&N.at(-1)===F.no_timestamps_token_id&&(console.warn("<|notimestamps|> prompt token is removed from generation_config since `return_timestamps` is set to `true`."),N.pop()),N.filter(Se=>Se!=null)}async generate({inputs:F=null,generation_config:N=null,logits_processor:me=null,stopping_criteria:ke=null,...Se}){N=this._prepare_generation_config(N,Se);const ze=Se.decoder_input_ids??this._retrieve_init_tokens(N);if(N.return_timestamps&&(me??(me=new y.LogitsProcessorList),me.push(new y.WhisperTimeStampLogitsProcessor(N,ze))),N.begin_suppress_tokens&&(me??(me=new y.LogitsProcessorList),me.push(new y.SuppressTokensAtBeginLogitsProcessor(N.begin_suppress_tokens,ze.length))),N.return_token_timestamps){if(!N.alignment_heads)throw new Error("Model generation config has no `alignment_heads`, token-level timestamps not available. See https://gist.github.com/hollance/42e32852f24243b748ae6bc1f985b13a on how to add this property to the generation config.");N.task==="translate"&&console.warn("Token-level timestamps may not be reliable for task 'translate'."),N.output_attentions=!0,N.return_dict_in_generate=!0}const Ke=await super.generate({inputs:F,generation_config:N,logits_processor:me,decoder_input_ids:ze,...Se});return N.return_token_timestamps&&(Ke.token_timestamps=this._extract_token_timestamps(Ke,N.alignment_heads,N.num_frames)),Ke}_extract_token_timestamps(F,N,me=null,ke=.02){if(!F.cross_attentions)throw new Error("Model outputs must contain cross attentions to extract timestamps. This is most likely because the model was not exported with `output_attentions=True`.");me==null&&console.warn("`num_frames` has not been set, meaning the entire audio will be analyzed. This may lead to inaccurate token-level timestamps for short audios (< 30 seconds).");let Se=this.config.median_filter_width;Se===void 0&&(console.warn("Model config has no `median_filter_width`, using default value of 7."),Se=7);const ze=F.cross_attentions,Ke=Array.from({length:this.config.decoder_layers},(Ot,ft)=>(0,g.cat)(ze.map(Vt=>Vt[ft]),2)),Ze=(0,g.stack)(N.map(([Ot,ft])=>{if(Ot>=Ke.length)throw new Error(`Layer index ${Ot} is out of bounds for cross attentions (length ${Ke.length}).`);return me?Ke[Ot].slice(null,ft,null,[0,me]):Ke[Ot].slice(null,ft)})).transpose(1,0,2,3),[at,vt]=(0,g.std_mean)(Ze,-2,0,!0),kt=Ze.clone();for(let Ot=0;Ot<kt.dims[0];++Ot){const ft=kt[Ot];for(let Vt=0;Vt<ft.dims[0];++Vt){const rr=ft[Vt],ar=at[Ot][Vt][0].data,fr=vt[Ot][Vt][0].data;for(let hr=0;hr<rr.dims[0];++hr){let br=rr[hr].data;for(let $r=0;$r<br.length;++$r)br[$r]=(br[$r]-fr[$r])/ar[$r];br.set((0,b.medianFilter)(br,Se))}}}const Mt=[(0,g.mean)(kt,1)],zt=F.sequences.dims,Ct=new g.Tensor("float32",new Float32Array(zt[0]*zt[1]),zt);for(let Ot=0;Ot<zt[0];++Ot){const ft=Mt[Ot].neg().squeeze_(0),[Vt,rr]=(0,b.dynamic_time_warping)(ft.tolist()),ar=Array.from({length:Vt.length-1},(br,$r)=>Vt[$r+1]-Vt[$r]),fr=(0,d.mergeArrays)([1],ar).map(br=>!!br),hr=[];for(let br=0;br<fr.length;++br)fr[br]&&hr.push(rr[br]*ke);Ct[Ot].data.set(hr,1)}return Ct}}class Ed extends kn{}class Ja extends X{constructor(){super(...arguments);le(this,"requires_attention_mask",!1);le(this,"main_input_name","input_values");le(this,"forward_params",["input_values","decoder_input_ids","past_key_values"])}}class pf extends Ja{}class Ya extends Ja{}class Za extends X{constructor(){super(...arguments);le(this,"main_input_name","pixel_values");le(this,"forward_params",["pixel_values","decoder_input_ids","encoder_hidden_states","past_key_values"])}}class Pd extends X{constructor(){super(...arguments);le(this,"forward_params",["input_ids","attention_mask","pixel_values","position_ids","past_key_values"])}}class In extends Pd{_merge_input_ids_with_image_features({inputs_embeds:F,image_features:N,input_ids:me,attention_mask:ke}){const Se=this.config.image_token_index,Ke=me.tolist().map(Mt=>Mt.findIndex(zt=>zt==Se)),Ze=Ke.every(Mt=>Mt===-1),at=Ke.every(Mt=>Mt!==-1);if(!Ze&&!at)throw new Error("Every input should contain either 0 or 1 image token.");if(Ze)return{inputs_embeds:F,attention_mask:ke};const vt=[],kt=[];for(let Mt=0;Mt<Ke.length;++Mt){const zt=Ke[Mt],Ct=F[Mt],Ot=N[Mt],ft=ke[Mt];vt.push((0,g.cat)([Ct.slice([0,zt]),Ot,Ct.slice([zt+1,Ct.dims[0]])],0)),kt.push((0,g.cat)([ft.slice([0,zt]),(0,g.ones)([Ot.dims[0]]),ft.slice([zt+1,ft.dims[0]])],0))}return{inputs_embeds:(0,g.stack)(vt,0),attention_mask:(0,g.stack)(kt,0)}}}class Cd extends In{}class Sd extends In{}class $d extends X{constructor(){super(...arguments);le(this,"forward_params",["input_ids","inputs_embeds","attention_mask","pixel_values","encoder_outputs","decoder_input_ids","decoder_inputs_embeds","decoder_attention_mask","past_key_values"]);le(this,"main_input_name","inputs_embeds")}}class kd extends $d{_merge_input_ids_with_image_features({inputs_embeds:F,image_features:N,input_ids:me,attention_mask:ke}){return{inputs_embeds:(0,g.cat)([N,F],1),attention_mask:(0,g.cat)([(0,g.ones)(N.dims.slice(0,2)),ke],1)}}async _prepare_inputs_embeds({input_ids:F,pixel_values:N,inputs_embeds:me,attention_mask:ke}){if(!F&&!N)throw new Error("Either `input_ids` or `pixel_values` should be provided.");let Se,ze;return F&&(Se=await this.encode_text({input_ids:F})),N&&(ze=await this.encode_image({pixel_values:N})),Se&&ze?{inputs_embeds:me,attention_mask:ke}=this._merge_input_ids_with_image_features({inputs_embeds:Se,image_features:ze,input_ids:F,attention_mask:ke}):me=Se||ze,{inputs_embeds:me,attention_mask:ke}}async forward({input_ids:F,pixel_values:N,attention_mask:me,decoder_input_ids:ke,decoder_attention_mask:Se,encoder_outputs:ze,past_key_values:Ke,inputs_embeds:Ze,decoder_inputs_embeds:at}){if(Ze||({inputs_embeds:Ze,attention_mask:me}=await this._prepare_inputs_embeds({input_ids:F,pixel_values:N,inputs_embeds:Ze,attention_mask:me})),!ze){let{last_hidden_state:Mt}=await ae(this,{inputs_embeds:Ze,attention_mask:me});ze=Mt}if(!at){if(!ke)throw new Error("Either `decoder_input_ids` or `decoder_inputs_embeds` should be provided.");at=await this.encode_text({input_ids:ke})}return await Te(this,{inputs_embeds:at,attention_mask:Se,encoder_attention_mask:me,encoder_hidden_states:ze,past_key_values:Ke},!0)}}class Id extends X{constructor(){super(...arguments);le(this,"forward_params",["input_ids","attention_mask","pixel_values","position_ids","past_key_values"])}}class Ad extends Id{_merge_input_ids_with_image_features(F){const N=F.image_features.dims.at(-1),me=F.image_features.view(-1,N);return B({image_token_id:this.config.image_token_index,...F,image_features:me})}}class Od extends X{constructor(){super(...arguments);le(this,"forward_params",["input_ids","attention_mask","pixel_values","pixel_attention_mask","position_ids","past_key_values"])}}class An extends Od{async encode_image({pixel_values:F,pixel_attention_mask:N}){return(await H(this.sessions.vision_encoder,{pixel_values:F,pixel_attention_mask:N})).image_features}_merge_input_ids_with_image_features(F){const N=F.image_features.dims.at(-1),me=F.image_features.view(-1,N);return B({image_token_id:this.config.image_token_id,...F,image_features:me})}}class eo extends An{}class to extends X{constructor(){super(...arguments);le(this,"forward_params",["input_ids","inputs_embeds","attention_mask","position_ids","pixel_values","image_sizes","past_key_values"])}}class On extends to{async forward({input_ids:F=null,attention_mask:N=null,pixel_values:me=null,image_sizes:ke=null,position_ids:Se=null,inputs_embeds:ze=null,past_key_values:Ke=null,generation_config:Ze=null,logits_processor:at=null,...vt}){if(!ze){let Mt;if(me&&F.dims[1]!==1){if(!ke)throw new Error("`image_sizes` must be provided when `pixel_values` is provided.");({image_features:Mt}=await H(this.sessions.vision_encoder,{pixel_values:me,image_sizes:ke}))}else{const zt=this.config.normalized_config.hidden_size;Mt=new g.Tensor("float32",[],[0,zt])}({inputs_embeds:ze}=await H(this.sessions.prepare_inputs_embeds,{input_ids:F,image_features:Mt}))}return await Te(this,{inputs_embeds:ze,past_key_values:Ke,attention_mask:N,position_ids:Se,generation_config:Ze,logits_processor:at},!1)}}class rs extends X{}class Fd extends rs{}class hf extends rs{static async from_pretrained(F,N={}){return super.from_pretrained(F,{...N,model_file_name:N.model_file_name??"text_model"})}}class ro extends rs{static async from_pretrained(F,N={}){return super.from_pretrained(F,{...N,model_file_name:N.model_file_name??"text_model"})}}class Dd extends rs{static async from_pretrained(F,N={}){return super.from_pretrained(F,{...N,model_file_name:N.model_file_name??"vision_model"})}}class Es extends rs{static async from_pretrained(F,N={}){return super.from_pretrained(F,{...N,model_file_name:N.model_file_name??"vision_model"})}}class io extends X{}class so extends io{}class Ld extends io{static async from_pretrained(F,N={}){return super.from_pretrained(F,{...N,model_file_name:N.model_file_name??"text_model"})}}class jd extends rs{static async from_pretrained(F,N={}){return super.from_pretrained(F,{...N,model_file_name:N.model_file_name??"vision_model"})}}class zd extends X{}class Fn extends zd{}class Ks extends X{}class Bd extends Ks{async forward(F){const N=!F.input_ids,me=!F.pixel_values;if(N&&me)throw new Error("Either `input_ids` or `pixel_values` should be provided.");if(N&&(F.input_ids=(0,g.ones)([F.pixel_values.dims[0],1])),me){const{image_size:at}=this.config.vision_config;F.pixel_values=(0,g.full)([0,3,at,at],0)}const{text_embeddings:ke,image_embeddings:Se,l2norm_text_embeddings:ze,l2norm_image_embeddings:Ke}=await super.forward(F),Ze={};return N||(Ze.text_embeddings=ke,Ze.l2norm_text_embeddings=ze),me||(Ze.image_embeddings=Se,Ze.l2norm_image_embeddings=Ke),Ze}}class Rd extends Ks{static async from_pretrained(F,N={}){return super.from_pretrained(F,{...N,model_file_name:N.model_file_name??"text_model"})}}class cs extends Ks{static async from_pretrained(F,N={}){return super.from_pretrained(F,{...N,model_file_name:N.model_file_name??"vision_model"})}}class no extends X{}class Nd extends no{}class Vd extends no{}class ao extends X{}class Ud extends ao{}class Wd extends ao{}class oo extends X{}class Gd extends oo{}class Hs extends oo{}class us extends X{}class lo extends us{}class Kd extends us{}class co extends X{}class Si extends co{}class Ar extends co{}class hi extends X{}class qs extends hi{}class Ps extends hi{}class Tt extends X{}class Qs extends Tt{}class uo extends Tt{}class wi extends X{}class po extends wi{}class Hd extends wi{}class ho extends X{}class qd extends ho{}class Dn extends ho{}class fo extends X{}class Qd extends fo{}class Xd extends fo{}class mo extends X{}class Jd extends mo{}class Yd extends mo{}class _o extends X{}class Zd extends _o{}class ep extends _o{}class go extends X{}class tp extends go{}class rp extends go{}class yo extends X{}class ip extends yo{}class sp extends yo{}class wo extends X{}class Ln extends wo{}class jn extends wo{}class zn extends X{}class Mo extends zn{}class bo extends zn{}class Bn extends X{}class vo extends Bn{}class xo extends Bn{}class To extends X{}class Eo extends To{}class np extends To{}class Po extends X{}class ap extends Po{}class op extends Po{}class Co extends X{}class lp extends Co{}class cp extends Co{}class So extends X{}class up extends So{}class dp extends So{}class $o extends X{}class pp extends $o{}class hp extends $o{}class ko extends X{}class fp extends ko{}class mp extends ko{}class _p extends X{constructor(){super(...arguments);le(this,"forward_params",["input_ids","attention_mask","position_ids","past_key_values","pixel_values","image_grid_thw"])}}class gp extends _p{get_rope_index(F,N,me,ke){const{vision_config:Se,image_token_id:ze,video_token_id:Ke,vision_start_token_id:Ze}=this.config,at=Se.spatial_merge_size??2,vt=[];if(N||me){let kt=F.tolist();ke||(ke=(0,g.ones_like)(F));const Mt=ke.tolist(),zt=Array.from({length:3},rr=>Array.from({length:F.dims[0]},ar=>Array.from({length:F.dims[1]},fr=>1))),Ct=N?N.tolist():[],Ot=me?me.tolist():[];let ft=0,Vt=0;for(let rr=0;rr<kt.length;++rr){const ar=kt[rr].filter((or,Pr)=>Mt[rr][Pr]==1),hr=ar.reduce((or,Pr,is)=>(Pr==Ze&&or.push(is),or),[]).map(or=>ar[or+1]),br=hr.filter(or=>or==ze).length,$r=hr.filter(or=>or==Ke).length;let Bt=[],Er=0,ui=br,ms=$r;for(let or=0;or<hr.length;++or){const Pr=ar.findIndex((ys,Mi)=>Mi>Er&&ys==ze),is=ar.findIndex((ys,Mi)=>Mi>Er&&ys==Ke),gs=ui>0&&Pr!==-1?Pr:ar.length+1,Os=ms>0&&is!==-1?is:ar.length+1;let Da,qu,Qu,Xu;gs<Os?([qu,Qu,Xu]=Ct[ft],++ft,--ui,Da=gs):([qu,Qu,Xu]=Ot[Vt],++Vt,--ms,Da=Os);const[xf,Ju,La]=[Number(qu),Math.floor(Number(Qu)/at),Math.floor(Number(Xu)/at)],Yu=Da-Er,Dh=Bt.length>0?(0,b.max)(Bt.at(-1))[0]+1:0;Bt.push(Array.from({length:3*Yu},(ys,Mi)=>Dh+Mi%Yu));const Zu=Yu+Dh,ja=xf*Ju*La,Tf=Array.from({length:ja},(ys,Mi)=>Zu+Math.floor(Mi/(Ju*La))),Ef=Array.from({length:ja},(ys,Mi)=>Zu+Math.floor(Mi/La)%Ju),Pf=Array.from({length:ja},(ys,Mi)=>Zu+Mi%La);Bt.push([Tf,Ef,Pf].flat()),Er=Da+ja}if(Er<ar.length){const or=Bt.length>0?(0,b.max)(Bt.at(-1))[0]+1:0,Pr=ar.length-Er;Bt.push(Array.from({length:3*Pr},(is,gs)=>or+gs%Pr))}const Qr=Bt.reduce((or,Pr)=>or+Pr.length,0),fi=new Array(Qr);let Aa=0;for(let or=0;or<3;++or)for(let Pr=0;Pr<Bt.length;++Pr){const is=Bt[Pr],gs=is.length/3;for(let Os=or*gs;Os<(or+1)*gs;++Os)fi[Aa++]=is[Os]}let _s=0;const Oa=Mt[rr];for(let or=0;or<Oa.length;++or)if(Oa[or]==1){for(let Pr=0;Pr<3;++Pr)zt[Pr][rr][or]=fi[Pr*Qr/3+_s];++_s}const Fa=(0,b.max)(fi)[0];vt.push(Fa+1-kt[rr].length)}return[new g.Tensor("int64",zt.flat(1/0),[3,F.dims[0],F.dims[1]]),new g.Tensor("int64",vt,[vt.length,1])]}else if(ke){const{data:kt,dims:Mt}=be(ke),zt=BigInt64Array.from({length:3*kt.length},(Ot,ft)=>kt[ft%kt.length]),Ct=Array.from({length:Mt[0]},(Ot,ft)=>(0,b.max)(kt.subarray(Mt[1]*ft,Mt[1]*(ft+1)))[0]+1n+BigInt(Mt[1]));return[new g.Tensor("int64",zt,[3,...Mt]),new g.Tensor("int64",Ct,[Ct.length,1])]}else{const[kt,Mt]=F.dims,zt=BigInt64Array.from({length:3*kt*Mt},(Ct,Ot)=>BigInt(Math.floor(Ot%Mt/kt)));return[new g.Tensor("int64",zt,[3,...F.dims]),(0,g.zeros)([kt,1])]}}async encode_image({pixel_values:F,image_grid_thw:N}){return(await H(this.sessions.vision_encoder,{pixel_values:F,grid_thw:N})).image_features}_merge_input_ids_with_image_features(F){return B({image_token_id:this.config.image_token_id,...F})}prepare_inputs_for_generation(F,N,me){if(N.attention_mask&&!N.position_ids)if(!N.past_key_values)[N.position_ids,N.rope_deltas]=this.get_rope_index(N.input_ids,N.image_grid_thw,N.video_grid_thw,N.attention_mask);else{N.pixel_values=null;const ke=BigInt(Object.values(N.past_key_values)[0].dims.at(-2)),Se=N.rope_deltas.map(ze=>ke+ze);N.position_ids=(0,g.stack)([Se,Se,Se],0)}return N}}class Io extends X{}class yp extends Io{}class wp extends Io{}class Ao extends X{}class Mp extends Ao{}class bp extends Ao{}class Oo extends X{}class vp extends Oo{}class Fo extends Oo{}class Cs extends X{}class te extends Cs{}class xp extends Cs{}class Rn extends X{}class Do extends Rn{}class Nn extends Rn{}class Ss extends X{}class ci extends Ss{}class ds extends Ss{async _call(F){return new At(await super._call(F))}}class Vn extends X{}class Lo extends Vn{}class jo extends Vn{async _call(F){return new At(await super._call(F))}}class zo extends X{}class Bo extends zo{}class Un extends X{}class Ro extends Un{}class No extends Un{async _call(F){return new At(await super._call(F))}}class Vo extends X{}class Uo extends Vo{}class Wn extends X{}class Wo extends Wn{}class Gn extends Wn{async _call(F){return new At(await super._call(F))}}class Kn extends X{}class Hn extends Kn{}class ps extends X{}class qn extends ps{}class Go extends ps{async _call(F){return new At(await super._call(F))}}class Qn extends X{}class Ko extends Qn{async _call(F){return new Oh(await super._call(F))}}class Xn extends X{}class Ho extends Xn{}class Jn extends Xn{async _call(F){return new At(await super._call(F))}}class Yn extends X{}class qo extends Yn{}class Zn extends Yn{async _call(F){return new At(await super._call(F))}}class we extends X{}class $s extends we{}class Qo extends we{}class he extends X{}class Xs extends he{}class fe extends he{}class ea extends X{}class Xo extends ea{}class Jo extends ea{async _call(F){return new At(await super._call(F))}}class Js extends X{}class Yo extends Js{}class Zo extends Js{async _call(F){return new Ys(await super._call(F))}}class ta extends Js{async _call(F){return new el(await super._call(F))}}class Ys extends Ae{constructor({logits:F,pred_boxes:N}){super(),this.logits=F,this.pred_boxes=N}}class el extends Ae{constructor({logits:F,pred_boxes:N,pred_masks:me}){super(),this.logits=F,this.pred_boxes=N,this.pred_masks=me}}class ra extends X{}class tl extends ra{}class rl extends ra{async _call(F){return new ks(await super._call(F))}}class ks extends Ae{constructor({logits:F,pred_boxes:N}){super(),this.logits=F,this.pred_boxes=N}}class ia extends X{}class il extends ia{}class sl extends ia{async _call(F){return new nl(await super._call(F))}}class nl extends ks{}class sa extends X{}class al extends sa{}class ol extends sa{async _call(F){return new ll(await super._call(F))}}class ll extends ks{}class na extends X{}class cl extends na{}class ul extends na{async _call(F){return new ks(await super._call(F))}}class aa extends X{}class dl extends aa{}class pl extends aa{async _call(F){return new hl(await super._call(F))}}class hl extends Ys{}class oa extends X{}class fl extends oa{}class ml extends oa{async _call(F){return new At(await super._call(F))}}class la extends X{}class _l extends la{}class gl extends la{async _call(F){return new At(await super._call(F))}}class ca extends X{}class yl extends ca{}class wl extends ca{async _call(F){return new At(await super._call(F))}}class Zs extends X{}class Ml extends Zs{}class bl extends Zs{async _call(F){return new At(await super._call(F))}}class vl extends Zs{}class ua extends X{}class xl extends ua{}class Tl extends ua{}class da extends X{}class El extends da{}class Pl extends da{}class Cl extends X{}class Sl extends Cl{}class en extends X{}class $l extends en{}class kl extends en{}class Il extends en{}class Al extends X{}class Ol extends Al{}class Fl extends X{}class Dl extends Fl{}class Ll extends X{}class jl extends Ll{}class pa extends X{}class zl extends pa{}class Bl extends pa{}class ha extends X{}class Rl extends ha{}class Nl extends ha{}class Vl extends X{}class Ul extends Vl{}class fa extends X{}class Wl extends fa{}class Gl extends fa{async _call(F){return new At(await super._call(F))}}class ma extends X{}class Kl extends ma{}class Hl extends ma{async _call(F){return new At(await super._call(F))}}class _a extends X{}class ql extends _a{}class Ql extends _a{async _call(F){return new At(await super._call(F))}}class ga extends X{}class Xl extends ga{}class Jl extends ga{async _call(F){return new At(await super._call(F))}}class Yl extends X{}class Zl extends Yl{}class ya extends X{}class ec extends ya{}class tc extends ya{async _call(F){return new rc(await super._call(F))}}class rc extends Ae{constructor({logits:F,pred_boxes:N}){super(),this.logits=F,this.pred_boxes=N}}class ic extends X{}class sc extends ic{async get_image_embeddings({pixel_values:F}){return await ae(this,{pixel_values:F})}async forward(F){if((!F.image_embeddings||!F.image_positional_embeddings)&&(F={...F,...await this.get_image_embeddings(F)}),!F.input_labels&&F.input_points){const me=F.input_points.dims.slice(0,-1),ke=me.reduce((Se,ze)=>Se*ze,1);F.input_labels=new g.Tensor("int64",new BigInt64Array(ke).fill(1n),me)}const N={image_embeddings:F.image_embeddings,image_positional_embeddings:F.image_positional_embeddings};return F.input_points&&(N.input_points=F.input_points),F.input_labels&&(N.input_labels=F.input_labels),F.input_boxes&&(N.input_boxes=F.input_boxes),await H(this.sessions.prompt_encoder_mask_decoder,N)}async _call(F){return new nc(await super._call(F))}}class nc extends Ae{constructor({iou_scores:F,pred_masks:N}){super(),this.iou_scores=F,this.pred_masks=N}}class wa extends X{}class ac extends wa{}class oc extends wa{}class Ma extends X{}class lc extends Ma{}class cc extends Ma{}class Ni extends X{}class uc extends Ni{}class dc extends Ni{async _call(F){return new fs(await super._call(F))}}class pc extends Ni{async _call(F){return new At(await super._call(F))}}class hc extends Ni{async _call(F){return new Sr(await super._call(F))}}class ba extends X{}class fc extends ba{}class mc extends ba{async _call(F){return new Sr(await super._call(F))}}class _c extends X{}class gc extends _c{}class tn extends X{}class yc extends tn{}class wc extends tn{async _call(F){return new fs(await super._call(F))}}class Mc extends tn{async _call(F){return new At(await super._call(F))}}class Is extends X{}class bc extends Is{}class vc extends Is{async _call(F){return new fs(await super._call(F))}}class xc extends Is{async _call(F){return new At(await super._call(F))}}class Tc extends Is{async _call(F){return new Sr(await super._call(F))}}class rn extends X{}class Ec extends rn{}class Pc extends rn{async _call(F){return new fs(await super._call(F))}}class Cc extends rn{async _call(F){return new At(await super._call(F))}}class Tp extends X{}class Sc extends Ni{}class $c extends Ni{async _call(F){return new fs(await super._call(F))}}class kc extends Ni{async _call(F){return new At(await super._call(F))}}class hs extends X{}class Ic extends hs{}class Ac extends hs{async _call(F){return new fs(await super._call(F))}}class Oc extends hs{async _call(F){return new At(await super._call(F))}}class Fc extends hs{async _call(F){return new Ah(await super._call(F))}}class Dc extends hs{async _call(F){return new Sr(await super._call(F))}}class Lc extends X{}class jc extends Lc{}class sn extends X{}class Ep extends sn{}class zc extends sn{}class Bc extends sn{async generate_speech(F,N,{threshold:me=.5,minlenratio:ke=0,maxlenratio:Se=20,vocoder:ze=null}={}){const Ke={input_ids:F},{encoder_outputs:Ze,encoder_attention_mask:at}=await ae(this,Ke),vt=Ze.dims[1]/this.config.reduction_factor,kt=Math.floor(vt*Se),Mt=Math.floor(vt*ke),zt=this.config.num_mel_bins;let Ct=[],Ot=null,ft=null,Vt=0;for(;;){++Vt;const fr=se(!!ft);let hr;ft?hr=ft.output_sequence_out:hr=new g.Tensor("float32",new Float32Array(zt),[1,1,zt]);let br={use_cache_branch:fr,output_sequence:hr,encoder_attention_mask:at,speaker_embeddings:N,encoder_hidden_states:Ze};this.addPastKeyValues(br,Ot),ft=await H(this.sessions.decoder_model_merged,br),Ot=this.getPastKeyValues(ft,Ot);const{prob:$r,spectrum:Bt}=ft;if(Ct.push(Bt),Vt>=Mt&&(Array.from($r.data).filter(Er=>Er>=me).length>0||Vt>=kt))break}const rr=(0,g.cat)(Ct),{waveform:ar}=await H(ze.sessions.model,{spectrogram:rr});return{spectrogram:rr,waveform:ar}}}class Rc extends X{constructor(){super(...arguments);le(this,"main_input_name","spectrogram")}}class Nc extends X{}class Vc extends Nc{}class va extends X{}class Uc extends va{}class Wc extends va{}class xa extends X{}class Gc extends xa{}class Kc extends xa{}class Ta extends X{}class Hc extends Ta{}class qc extends Ta{}class nn extends X{}class Qc extends nn{}class Xc extends nn{static async from_pretrained(F,N={}){return super.from_pretrained(F,{...N,model_file_name:N.model_file_name??"text_model"})}}class Jc extends nn{static async from_pretrained(F,N={}){return super.from_pretrained(F,{...N,model_file_name:N.model_file_name??"audio_model"})}}class Yc extends X{}class Ea extends Yc{async _call(F){return new Fh(await super._call(F))}}class an extends X{}class Pp extends an{}class Zc extends an{}class eu extends an{}class Pa extends X{}class tu extends Pa{}class ru extends Pa{}class Ca extends X{}class iu extends Ca{}class su extends Ca{async _call(F){return new At(await super._call(F))}}class Sa extends X{}class Cp extends Sa{}class Sp extends Sa{}class $a extends X{constructor(){super(...arguments);le(this,"forward_params",["input_ids","attention_mask","encoder_outputs","decoder_input_ids","decoder_attention_mask","past_key_values"])}_apply_and_filter_by_delay_pattern_mask(N){const[me,ke]=N.dims,Se=this.config.decoder.num_codebooks,ze=ke-Se;let Ke=0;for(let vt=0;vt<N.size;++vt){if(N.data[vt]===this.config.decoder.pad_token_id)continue;const kt=vt%ke,Mt=Math.floor(vt/ke)%Se,zt=kt-Mt;zt>0&&zt<=ze&&(N.data[Ke++]=N.data[vt])}const Ze=Math.floor(me/Se),at=Ke/(Ze*Se);return new g.Tensor(N.type,N.data.slice(0,Ke),[Ze,Se,at])}prepare_inputs_for_generation(N,me,ke){let Se=structuredClone(N);for(let Ke=0;Ke<Se.length;++Ke)for(let Ze=0;Ze<Se[Ke].length;++Ze)Ke%this.config.decoder.num_codebooks>=Ze&&(Se[Ke][Ze]=BigInt(this.config.decoder.pad_token_id));return ke.guidance_scale!==null&&ke.guidance_scale>1&&(Se=Se.concat(Se)),super.prepare_inputs_for_generation(Se,me,ke)}async generate(N){const me=await super.generate(N),ke=this._apply_and_filter_by_delay_pattern_mask(me).unsqueeze_(0),{audio_values:Se}=await H(this.sessions.encodec_decode,{audio_codes:ke});return Se}}class on extends X{}class nu extends on{}class au extends on{async _call(F){return new At(await super._call(F))}}class ou extends on{}class ln extends X{}class lu extends ln{}class cu extends ln{async _call(F){return new At(await super._call(F))}}class uu extends ln{}class cn extends X{}class du extends cn{}class pu extends cn{async _call(F){return new At(await super._call(F))}}class hu extends cn{}class un extends X{}class fu extends un{}class mu extends un{async _call(F){return new At(await super._call(F))}}class _u extends un{}class gu extends X{}class yu extends gu{}class wu extends X{}class Mu extends wu{constructor(...N){super(...N);le(this,"forward_params",["input_ids","pixel_values","images_seq_mask","images_emb_mask","attention_mask","position_ids","past_key_values"]);this._generation_mode="text"}async forward(N){const me=this._generation_mode??"text";let ke;if(me==="text"||!N.past_key_values){const at=this.sessions.prepare_inputs_embeds,vt=(0,d.pick)(N,at.inputNames);ke=await H(at,vt)}else{const at=this.sessions.gen_img_embeds,vt=(0,d.pick)({image_ids:N.input_ids},at.inputNames);ke=await H(at,vt)}const Se={...N,...ke},ze=await Te(this,Se),Ke=this.sessions[me==="text"?"lm_head":"gen_head"];if(!Ke)throw new Error(`Unable to find "${Ke}" generation head`);const Ze=await H(Ke,(0,d.pick)(ze,Ke.inputNames));return{...ke,...ze,...Ze}}async generate(N){return this._generation_mode="text",super.generate(N)}async generate_images(N){this._generation_mode="image";const me=(N.inputs??N[this.main_input_name]).dims[1],Se=(await super.generate(N)).slice(null,[me,null]),ze=this.sessions.image_decode,{decoded_image:Ke}=await H(ze,{generated_tokens:Se}),Ze=Ke.add_(1).mul_(255/2).clamp_(0,255).to("uint8"),at=[];for(const vt of Ze){const kt=v.RawImage.fromTensor(vt);at.push(kt)}return at}}class bu extends Ae{constructor({char_logits:F,bpe_logits:N,wp_logits:me}){super(),this.char_logits=F,this.bpe_logits=N,this.wp_logits=me}get logits(){return[this.char_logits,this.bpe_logits,this.wp_logits]}}class vu extends X{}class xu extends vu{async _call(F){return new bu(await super._call(F))}}class ka extends X{}class Tu extends ka{}class Eu extends ka{}class Ia extends X{}class Pu extends Ia{}class Cu extends Ia{}class Su extends X{constructor(){super(...arguments);le(this,"forward_params",["input_ids","attention_mask","position_ids","audio_values","past_key_values"])}}class $u extends Su{_merge_input_ids_with_audio_features(F){const N=F.audio_features.dims.at(-1),me=F.audio_features.view(-1,N);return Q({audio_token_id:this.config.ignore_index,...F,audio_features:me})}}class dn extends X{constructor(){super(...arguments);le(this,"main_input_name","input_values");le(this,"forward_params",["input_values"])}}class ku extends Ae{constructor({audio_codes:F}){super(),this.audio_codes=F}}class Iu extends Ae{constructor({audio_values:F}){super(),this.audio_values=F}}class Au extends dn{async encode(F){return new ku(await H(this.sessions.encoder_model,F))}async decode(F){return new Iu(await H(this.sessions.decoder_model,F))}}class Ou extends dn{static async from_pretrained(F,N={}){return super.from_pretrained(F,{...N,model_file_name:N.model_file_name??"encoder_model"})}}class Fu extends dn{static async from_pretrained(F,N={}){return super.from_pretrained(F,{...N,model_file_name:N.model_file_name??"decoder_model"})}}class pn extends X{constructor(){super(...arguments);le(this,"main_input_name","input_values");le(this,"forward_params",["input_values"])}}class Du extends Ae{constructor({audio_codes:F}){super(),this.audio_codes=F}}class Lu extends Ae{constructor({audio_values:F}){super(),this.audio_values=F}}class ju extends pn{async encode(F){return new Du(await H(this.sessions.encoder_model,F))}async decode(F){return new Lu(await H(this.sessions.decoder_model,F))}}class zu extends pn{static async from_pretrained(F,N={}){return super.from_pretrained(F,{...N,model_file_name:N.model_file_name??"encoder_model"})}}class Bu extends pn{static async from_pretrained(F,N={}){return super.from_pretrained(F,{...N,model_file_name:N.model_file_name??"decoder_model"})}}class hn extends X{constructor(){super(...arguments);le(this,"main_input_name","input_values");le(this,"forward_params",["input_values"])}}class Ru extends hn{async encode(F){return await H(this.sessions.encoder_model,F)}async decode(F){return await H(this.sessions.decoder_model,F)}}class Nu extends hn{static async from_pretrained(F,N={}){return super.from_pretrained(F,{...N,model_file_name:N.model_file_name??"encoder_model"})}}class Vu extends hn{static async from_pretrained(F,N={}){return super.from_pretrained(F,{...N,model_file_name:N.model_file_name??"decoder_model"})}}class Wt{static async from_pretrained(F,{progress_callback:N=null,config:me=null,cache_dir:ke=null,local_files_only:Se=!1,revision:ze="main",model_file_name:Ke=null,subfolder:Ze="onnx",device:at=null,dtype:vt=null,use_external_data_format:kt=null,session_options:Mt={}}={}){const zt={progress_callback:N,config:me,cache_dir:ke,local_files_only:Se,revision:ze,model_file_name:Ke,subfolder:Ze,device:at,dtype:vt,use_external_data_format:kt,session_options:Mt};if(zt.config=await n.AutoConfig.from_pretrained(F,zt),!this.MODEL_CLASS_MAPPINGS)throw new Error("`MODEL_CLASS_MAPPINGS` not implemented for this type of `AutoClass`: "+this.name);const Ct=zt.config.model_type;for(const Ot of this.MODEL_CLASS_MAPPINGS){let ft=Ot.get(Ct);if(!ft){for(const Vt of Ot.values())if(Vt[0]===Ct){ft=Vt;break}if(!ft)continue}return await ft[1].from_pretrained(F,zt)}if(this.BASE_IF_FAIL)return th.has(Ct)||console.warn(`Unknown model class "${Ct}", attempting to construct from base class.`),await X.from_pretrained(F,zt);throw Error(`Unsupported model type: ${Ct}`)}}le(Wt,"MODEL_CLASS_MAPPINGS",null),le(Wt,"BASE_IF_FAIL",!1);const $p=new Map([["bert",["BertModel",Be]],["modernbert",["ModernBertModel",We]],["nomic_bert",["NomicBertModel",dt]],["roformer",["RoFormerModel",xt]],["electra",["ElectraModel",$e]],["esm",["EsmModel",yt]],["convbert",["ConvBertModel",$t]],["camembert",["CamembertModel",jt]],["deberta",["DebertaModel",Jr]],["deberta-v2",["DebertaV2Model",Yi]],["mpnet",["MPNetModel",Yr]],["albert",["AlbertModel",ve]],["distilbert",["DistilBertModel",Kr]],["roberta",["RobertaModel",Mr]],["xlm",["XLMModel",li]],["xlm-roberta",["XLMRobertaModel",Sn]],["clap",["ClapModel",Qc]],["clip",["CLIPModel",Fd]],["clipseg",["CLIPSegModel",Nd]],["chinese_clip",["ChineseCLIPModel",Fn]],["siglip",["SiglipModel",so]],["jina_clip",["JinaCLIPModel",Bd]],["mobilebert",["MobileBertModel",Li]],["squeezebert",["SqueezeBertModel",pi]],["wav2vec2",["Wav2Vec2Model",uc]],["wav2vec2-bert",["Wav2Vec2BertModel",Ec]],["unispeech",["UniSpeechModel",yc]],["unispeech-sat",["UniSpeechSatModel",bc]],["hubert",["HubertModel",Sc]],["wavlm",["WavLMModel",Ic]],["audio-spectrogram-transformer",["ASTModel",xd]],["vits",["VitsModel",Ea]],["pyannote",["PyAnnoteModel",fc]],["wespeaker-resnet",["WeSpeakerResNetModel",gc]],["detr",["DetrModel",Yo]],["rt_detr",["RTDetrModel",tl]],["rt_detr_v2",["RTDetrV2Model",il]],["rf_detr",["RFDetrModel",al]],["d_fine",["DFineModel",cl]],["table-transformer",["TableTransformerModel",dl]],["vit",["ViTModel",ci]],["ijepa",["IJepaModel",Lo]],["pvt",["PvtModel",Ro]],["vit_msn",["ViTMSNModel",Wo]],["vit_mae",["ViTMAEModel",Uo]],["groupvit",["GroupViTModel",Hn]],["fastvit",["FastViTModel",qn]],["mobilevit",["MobileViTModel",Ho]],["mobilevitv2",["MobileViTV2Model",qo]],["owlvit",["OwlViTModel",$s]],["owlv2",["Owlv2Model",Xs]],["beit",["BeitModel",Xo]],["deit",["DeiTModel",fl]],["hiera",["HieraModel",_l]],["convnext",["ConvNextModel",Wl]],["convnextv2",["ConvNextV2Model",Kl]],["dinov2",["Dinov2Model",ql]],["dinov2_with_registers",["Dinov2WithRegistersModel",Xl]],["resnet",["ResNetModel",yl]],["swin",["SwinModel",Ml]],["swin2sr",["Swin2SRModel",xl]],["donut-swin",["DonutSwinModel",Ul]],["yolos",["YolosModel",ec]],["dpt",["DPTModel",El]],["glpn",["GLPNModel",Rl]],["hifigan",["SpeechT5HifiGan",Rc]],["efficientnet",["EfficientNetModel",iu]],["decision_transformer",["DecisionTransformerModel",yu]],["patchtst",["PatchTSTForPrediction",Tu]],["patchtsmixer",["PatchTSMixerForPrediction",Pu]],["mobilenet_v1",["MobileNetV1Model",nu]],["mobilenet_v2",["MobileNetV2Model",lu]],["mobilenet_v3",["MobileNetV3Model",du]],["mobilenet_v4",["MobileNetV4Model",fu]],["maskformer",["MaskFormerModel",zl]],["mgp-str",["MgpstrForSceneTextRecognition",xu]],["style_text_to_speech_2",["StyleTextToSpeech2Model",jc]]]),ff=new Map([["t5",["T5Model",xe]],["longt5",["LongT5Model",ot]],["mt5",["MT5Model",It]],["bart",["BartModel",_r]],["mbart",["MBartModel",Zr]],["marian",["MarianModel",ac]],["whisper",["WhisperModel",Td]],["m2m_100",["M2M100Model",lc]],["blenderbot",["BlenderbotModel",Ir]],["blenderbot-small",["BlenderbotSmallModel",Ri]]]),mf=new Map([["mimi",["MimiModel",Au]],["dac",["DacModel",ju]],["snac",["SnacModel",Ru]]]),_f=new Map([["bloom",["BloomModel",vp]],["jais",["JAISModel",Gd]],["gpt2",["GPT2Model",Ud]],["gptj",["GPTJModel",qs]],["gpt_bigcode",["GPTBigCodeModel",Qs]],["gpt_neo",["GPTNeoModel",lo]],["gpt_neox",["GPTNeoXModel",Si]],["codegen",["CodeGenModel",po]],["llama",["LlamaModel",qd]],["exaone",["ExaoneModel",Zd]],["olmo",["OlmoModel",ip]],["olmo2",["Olmo2Model",Ln]],["mobilellm",["MobileLLMModel",tp]],["granite",["GraniteModel",Mo]],["cohere",["CohereModel",vo]],["gemma",["GemmaModel",Eo]],["gemma2",["Gemma2Model",ap]],["gemma3_text",["Gemma3Model",lp]],["helium",["HeliumModel",Qd]],["glm",["GlmModel",Jd]],["openelm",["OpenELMModel",up]],["qwen2",["Qwen2Model",pp]],["qwen3",["Qwen3Model",fp]],["phi",["PhiModel",yp]],["phi3",["Phi3Model",Mp]],["mpt",["MptModel",te]],["opt",["OPTModel",Do]],["mistral",["MistralModel",Uc]],["starcoder2",["Starcoder2Model",Gc]],["falcon",["FalconModel",Hc]],["stablelm",["StableLmModel",tu]]]),Uu=new Map([["speecht5",["SpeechT5ForSpeechToText",zc]],["whisper",["WhisperForConditionalGeneration",kn]],["lite-whisper",["LiteWhisperForConditionalGeneration",Ed]],["moonshine",["MoonshineForConditionalGeneration",Ya]]]),kp=new Map([["speecht5",["SpeechT5ForTextToSpeech",Bc]]]),Ip=new Map([["vits",["VitsModel",Ea]],["musicgen",["MusicgenForConditionalGeneration",$a]]]),Ap=new Map([["bert",["BertForSequenceClassification",He]],["modernbert",["ModernBertForSequenceClassification",pt]],["roformer",["RoFormerForSequenceClassification",cr]],["electra",["ElectraForSequenceClassification",De]],["esm",["EsmForSequenceClassification",gi]],["convbert",["ConvBertForSequenceClassification",U]],["camembert",["CamembertForSequenceClassification",Ht]],["deberta",["DebertaForSequenceClassification",Ji]],["deberta-v2",["DebertaV2ForSequenceClassification",Fi]],["mpnet",["MPNetForSequenceClassification",os]],["albert",["AlbertForSequenceClassification",j]],["distilbert",["DistilBertForSequenceClassification",Hr]],["roberta",["RobertaForSequenceClassification",xr]],["xlm",["XLMForSequenceClassification",Us]],["xlm-roberta",["XLMRobertaForSequenceClassification",$n]],["bart",["BartForSequenceClassification",Cr]],["mbart",["MBartForSequenceClassification",Fr]],["mobilebert",["MobileBertForSequenceClassification",zi]],["squeezebert",["SqueezeBertForSequenceClassification",ls]]]),Op=new Map([["bert",["BertForTokenClassification",Xe]],["modernbert",["ModernBertForTokenClassification",_t]],["roformer",["RoFormerForTokenClassification",Kt]],["electra",["ElectraForTokenClassification",Dt]],["esm",["EsmForTokenClassification",Di]],["convbert",["ConvBertForTokenClassification",ge]],["camembert",["CamembertForTokenClassification",nr]],["deberta",["DebertaForTokenClassification",Oi]],["deberta-v2",["DebertaV2ForTokenClassification",er]],["mpnet",["MPNetForTokenClassification",ai]],["distilbert",["DistilBertForTokenClassification",es]],["roberta",["RobertaForTokenClassification",Tr]],["xlm",["XLMForTokenClassification",Ha]],["xlm-roberta",["XLMRobertaForTokenClassification",vd]]]),Wu=new Map([["t5",["T5ForConditionalGeneration",je]],["longt5",["LongT5ForConditionalGeneration",it]],["mt5",["MT5ForConditionalGeneration",Qt]],["bart",["BartForConditionalGeneration",tr]],["mbart",["MBartForConditionalGeneration",Pi]],["marian",["MarianMTModel",oc]],["m2m_100",["M2M100ForConditionalGeneration",cc]],["blenderbot",["BlenderbotForConditionalGeneration",Br]],["blenderbot-small",["BlenderbotSmallForConditionalGeneration",Rr]]]),Gu=new Map([["bloom",["BloomForCausalLM",Fo]],["gpt2",["GPT2LMHeadModel",Wd]],["jais",["JAISLMHeadModel",Hs]],["gptj",["GPTJForCausalLM",Ps]],["gpt_bigcode",["GPTBigCodeForCausalLM",uo]],["gpt_neo",["GPTNeoForCausalLM",Kd]],["gpt_neox",["GPTNeoXForCausalLM",Ar]],["codegen",["CodeGenForCausalLM",Hd]],["llama",["LlamaForCausalLM",Dn]],["exaone",["ExaoneForCausalLM",ep]],["olmo",["OlmoForCausalLM",sp]],["olmo2",["Olmo2ForCausalLM",jn]],["mobilellm",["MobileLLMForCausalLM",rp]],["granite",["GraniteForCausalLM",bo]],["cohere",["CohereForCausalLM",xo]],["gemma",["GemmaForCausalLM",np]],["gemma2",["Gemma2ForCausalLM",op]],["gemma3_text",["Gemma3ForCausalLM",cp]],["helium",["HeliumForCausalLM",Xd]],["glm",["GlmForCausalLM",Yd]],["openelm",["OpenELMForCausalLM",dp]],["qwen2",["Qwen2ForCausalLM",hp]],["qwen3",["Qwen3ForCausalLM",mp]],["phi",["PhiForCausalLM",wp]],["phi3",["Phi3ForCausalLM",bp]],["mpt",["MptForCausalLM",xp]],["opt",["OPTForCausalLM",Nn]],["mbart",["MBartForCausalLM",Ci]],["mistral",["MistralForCausalLM",Wc]],["starcoder2",["Starcoder2ForCausalLM",Kc]],["falcon",["FalconForCausalLM",qc]],["trocr",["TrOCRForCausalLM",Vc]],["stablelm",["StableLmForCausalLM",ru]],["phi3_v",["Phi3VForCausalLM",On]]]),gf=new Map([["multi_modality",["MultiModalityCausalLM",Mu]]]),Fp=new Map([["bert",["BertForMaskedLM",Ve]],["modernbert",["ModernBertForMaskedLM",rt]],["roformer",["RoFormerForMaskedLM",Nt]],["electra",["ElectraForMaskedLM",Ge]],["esm",["EsmForMaskedLM",ur]],["convbert",["ConvBertForMaskedLM",Ti]],["camembert",["CamembertForMaskedLM",gt]],["deberta",["DebertaForMaskedLM",Ai]],["deberta-v2",["DebertaV2ForMaskedLM",_i]],["mpnet",["MPNetForMaskedLM",as]],["albert",["AlbertForMaskedLM",ne]],["distilbert",["DistilBertForMaskedLM",qe]],["roberta",["RobertaForMaskedLM",gr]],["xlm",["XLMWithLMHeadModel",Ka]],["xlm-roberta",["XLMRobertaForMaskedLM",bd]],["mobilebert",["MobileBertForMaskedLM",ji]],["squeezebert",["SqueezeBertForMaskedLM",Vs]]]),Dp=new Map([["bert",["BertForQuestionAnswering",st]],["roformer",["RoFormerForQuestionAnswering",wr]],["electra",["ElectraForQuestionAnswering",Ut]],["convbert",["ConvBertForQuestionAnswering",K]],["camembert",["CamembertForQuestionAnswering",ri]],["deberta",["DebertaForQuestionAnswering",ns]],["deberta-v2",["DebertaV2ForQuestionAnswering",Zi]],["mpnet",["MPNetForQuestionAnswering",xs]],["albert",["AlbertForQuestionAnswering",J]],["distilbert",["DistilBertForQuestionAnswering",si]],["roberta",["RobertaForQuestionAnswering",Nr]],["xlm",["XLMForQuestionAnswering",qa]],["xlm-roberta",["XLMRobertaForQuestionAnswering",Ws]],["mobilebert",["MobileBertForQuestionAnswering",qr]],["squeezebert",["SqueezeBertForQuestionAnswering",oi]]]),Ku=new Map([["vision-encoder-decoder",["VisionEncoderDecoderModel",Za]],["idefics3",["Idefics3ForConditionalGeneration",An]],["smolvlm",["SmolVLMForConditionalGeneration",eo]]]),Lp=new Map([["llava",["LlavaForConditionalGeneration",In]],["llava_onevision",["LlavaOnevisionForConditionalGeneration",Cd]],["moondream1",["Moondream1ForConditionalGeneration",Sd]],["florence2",["Florence2ForConditionalGeneration",kd]],["qwen2-vl",["Qwen2VLForConditionalGeneration",gp]],["idefics3",["Idefics3ForConditionalGeneration",An]],["smolvlm",["SmolVLMForConditionalGeneration",eo]],["paligemma",["PaliGemmaForConditionalGeneration",Ad]]]),jp=new Map([["ultravox",["UltravoxModel",$u]]]),yf=new Map([["vision-encoder-decoder",["VisionEncoderDecoderModel",Za]]]),zp=new Map([["vit",["ViTForImageClassification",ds]],["ijepa",["IJepaForImageClassification",jo]],["pvt",["PvtForImageClassification",No]],["vit_msn",["ViTMSNForImageClassification",Gn]],["fastvit",["FastViTForImageClassification",Go]],["mobilevit",["MobileViTForImageClassification",Jn]],["mobilevitv2",["MobileViTV2ForImageClassification",Zn]],["beit",["BeitForImageClassification",Jo]],["deit",["DeiTForImageClassification",ml]],["hiera",["HieraForImageClassification",gl]],["convnext",["ConvNextForImageClassification",Gl]],["convnextv2",["ConvNextV2ForImageClassification",Hl]],["dinov2",["Dinov2ForImageClassification",Ql]],["dinov2_with_registers",["Dinov2WithRegistersForImageClassification",Jl]],["resnet",["ResNetForImageClassification",wl]],["swin",["SwinForImageClassification",bl]],["segformer",["SegformerForImageClassification",Zc]],["efficientnet",["EfficientNetForImageClassification",su]],["mobilenet_v1",["MobileNetV1ForImageClassification",au]],["mobilenet_v2",["MobileNetV2ForImageClassification",cu]],["mobilenet_v3",["MobileNetV3ForImageClassification",pu]],["mobilenet_v4",["MobileNetV4ForImageClassification",mu]]]),Bp=new Map([["detr",["DetrForObjectDetection",Zo]],["rt_detr",["RTDetrForObjectDetection",rl]],["rt_detr_v2",["RTDetrV2ForObjectDetection",sl]],["rf_detr",["RFDetrForObjectDetection",ol]],["d_fine",["DFineForObjectDetection",ul]],["table-transformer",["TableTransformerForObjectDetection",pl]],["yolos",["YolosForObjectDetection",tc]]]),Rp=new Map([["owlvit",["OwlViTForObjectDetection",Qo]],["owlv2",["Owlv2ForObjectDetection",fe]],["grounding-dino",["GroundingDinoForObjectDetection",Zl]]]),As=new Map([["detr",["DetrForSegmentation",ta]],["clipseg",["CLIPSegForImageSegmentation",Vd]]]),Np=new Map([["segformer",["SegformerForSemanticSegmentation",eu]],["sapiens",["SapiensForSemanticSegmentation",$l]],["swin",["SwinForSemanticSegmentation",vl]],["mobilenet_v1",["MobileNetV1ForSemanticSegmentation",ou]],["mobilenet_v2",["MobileNetV2ForSemanticSegmentation",uu]],["mobilenet_v3",["MobileNetV3ForSemanticSegmentation",hu]],["mobilenet_v4",["MobileNetV4ForSemanticSegmentation",_u]]]),Vp=new Map([["detr",["DetrForSegmentation",ta]],["maskformer",["MaskFormerForInstanceSegmentation",Bl]]]),Up=new Map([["sam",["SamModel",sc]]]),Wp=new Map([["wav2vec2",["Wav2Vec2ForCTC",dc]],["wav2vec2-bert",["Wav2Vec2BertForCTC",Pc]],["unispeech",["UniSpeechForCTC",wc]],["unispeech-sat",["UniSpeechSatForCTC",vc]],["wavlm",["WavLMForCTC",Ac]],["hubert",["HubertForCTC",$c]]]),Gp=new Map([["wav2vec2",["Wav2Vec2ForSequenceClassification",pc]],["wav2vec2-bert",["Wav2Vec2BertForSequenceClassification",Cc]],["unispeech",["UniSpeechForSequenceClassification",Mc]],["unispeech-sat",["UniSpeechSatForSequenceClassification",xc]],["wavlm",["WavLMForSequenceClassification",Oc]],["hubert",["HubertForSequenceClassification",kc]],["audio-spectrogram-transformer",["ASTForAudioClassification",Gs]]]),Kp=new Map([["wavlm",["WavLMForXVector",Fc]]]),Hp=new Map([["unispeech-sat",["UniSpeechSatForAudioFrameClassification",Tc]],["wavlm",["WavLMForAudioFrameClassification",Dc]],["wav2vec2",["Wav2Vec2ForAudioFrameClassification",hc]],["pyannote",["PyAnnoteForAudioFrameClassification",mc]]]),qp=new Map([["vitmatte",["VitMatteForImageMatting",Ko]]]),wf=new Map([["patchtst",["PatchTSTForPrediction",Eu]],["patchtsmixer",["PatchTSMixerForPrediction",Cu]]]),Qp=new Map([["swin2sr",["Swin2SRForImageSuperResolution",Tl]]]),Xp=new Map([["dpt",["DPTForDepthEstimation",Pl]],["depth_anything",["DepthAnythingForDepthEstimation",Sl]],["glpn",["GLPNForDepthEstimation",Nl]],["sapiens",["SapiensForDepthEstimation",kl]],["depth_pro",["DepthProForDepthEstimation",Ol]],["metric3d",["Metric3DForDepthEstimation",Dl]],["metric3dv2",["Metric3Dv2ForDepthEstimation",jl]]]),Jp=new Map([["sapiens",["SapiensForNormalEstimation",Il]]]),Yp=new Map([["vitpose",["VitPoseForPoseEstimation",Bo]]]),Zp=new Map([["clip",["CLIPVisionModelWithProjection",Es]],["siglip",["SiglipVisionModel",jd]],["jina_clip",["JinaCLIPVisionModel",cs]]]),eh=[[$p,I.EncoderOnly],[ff,I.EncoderDecoder],[_f,I.DecoderOnly],[mf,I.AutoEncoder],[Ap,I.EncoderOnly],[Op,I.EncoderOnly],[Wu,I.Seq2Seq],[Uu,I.Seq2Seq],[Gu,I.DecoderOnly],[gf,I.MultiModality],[Fp,I.EncoderOnly],[Dp,I.EncoderOnly],[Ku,I.Vision2Seq],[Lp,I.ImageTextToText],[jp,I.AudioTextToText],[zp,I.EncoderOnly],[As,I.EncoderOnly],[Vp,I.EncoderOnly],[Np,I.EncoderOnly],[qp,I.EncoderOnly],[wf,I.EncoderOnly],[Qp,I.EncoderOnly],[Xp,I.EncoderOnly],[Jp,I.EncoderOnly],[Yp,I.EncoderOnly],[Bp,I.EncoderOnly],[Rp,I.EncoderOnly],[Up,I.MaskGeneration],[Wp,I.EncoderOnly],[Gp,I.EncoderOnly],[kp,I.Seq2Seq],[Ip,I.EncoderOnly],[Kp,I.EncoderOnly],[Hp,I.EncoderOnly],[Zp,I.EncoderOnly]];for(const[C,F]of eh)for(const[N,me]of C.values())$.set(N,F),S.set(me,N),P.set(N,me);const Mf=[["MusicgenForConditionalGeneration",$a,I.Musicgen],["Phi3VForCausalLM",On,I.Phi3V],["CLIPTextModelWithProjection",ro,I.EncoderOnly],["SiglipTextModel",Ld,I.EncoderOnly],["JinaCLIPTextModel",Rd,I.EncoderOnly],["ClapTextModelWithProjection",Xc,I.EncoderOnly],["ClapAudioModelWithProjection",Jc,I.EncoderOnly],["DacEncoderModel",zu,I.EncoderOnly],["DacDecoderModel",Bu,I.EncoderOnly],["MimiEncoderModel",Ou,I.EncoderOnly],["MimiDecoderModel",Fu,I.EncoderOnly],["SnacEncoderModel",Nu,I.EncoderOnly],["SnacDecoderModel",Vu,I.EncoderOnly]];for(const[C,F,N]of Mf)$.set(C,N),S.set(F,C),P.set(C,F);const th=new Map([["modnet",As],["birefnet",As],["isnet",As],["ben",As]]);for(const[C,F]of th.entries())F.set(C,["PreTrainedModel",X]),$.set(C,I.EncoderOnly),S.set(X,C),P.set(C,X);class Hu extends Wt{}le(Hu,"MODEL_CLASS_MAPPINGS",eh.map(F=>F[0])),le(Hu,"BASE_IF_FAIL",!0);class rh extends Wt{}le(rh,"MODEL_CLASS_MAPPINGS",[Ap]);class ih extends Wt{}le(ih,"MODEL_CLASS_MAPPINGS",[Op]);class sh extends Wt{}le(sh,"MODEL_CLASS_MAPPINGS",[Wu]);class nh extends Wt{}le(nh,"MODEL_CLASS_MAPPINGS",[Uu]);class ah extends Wt{}le(ah,"MODEL_CLASS_MAPPINGS",[kp]);class oh extends Wt{}le(oh,"MODEL_CLASS_MAPPINGS",[Ip]);class lh extends Wt{}le(lh,"MODEL_CLASS_MAPPINGS",[Gu]);class ch extends Wt{}le(ch,"MODEL_CLASS_MAPPINGS",[Fp]);class uh extends Wt{}le(uh,"MODEL_CLASS_MAPPINGS",[Dp]);class dh extends Wt{}le(dh,"MODEL_CLASS_MAPPINGS",[Ku]);class ph extends Wt{}le(ph,"MODEL_CLASS_MAPPINGS",[zp]);class hh extends Wt{}le(hh,"MODEL_CLASS_MAPPINGS",[As]);class fh extends Wt{}le(fh,"MODEL_CLASS_MAPPINGS",[Np]);class mh extends Wt{}le(mh,"MODEL_CLASS_MAPPINGS",[Vp]);class _h extends Wt{}le(_h,"MODEL_CLASS_MAPPINGS",[Bp]);class gh extends Wt{}le(gh,"MODEL_CLASS_MAPPINGS",[Rp]);class yh extends Wt{}le(yh,"MODEL_CLASS_MAPPINGS",[Up]);class wh extends Wt{}le(wh,"MODEL_CLASS_MAPPINGS",[Wp]);class Mh extends Wt{}le(Mh,"MODEL_CLASS_MAPPINGS",[Gp]);class bh extends Wt{}le(bh,"MODEL_CLASS_MAPPINGS",[Kp]);class vh extends Wt{}le(vh,"MODEL_CLASS_MAPPINGS",[Hp]);class xh extends Wt{}le(xh,"MODEL_CLASS_MAPPINGS",[yf]);class Th extends Wt{}le(Th,"MODEL_CLASS_MAPPINGS",[qp]);class Eh extends Wt{}le(Eh,"MODEL_CLASS_MAPPINGS",[Qp]);class Ph extends Wt{}le(Ph,"MODEL_CLASS_MAPPINGS",[Xp]);class Ch extends Wt{}le(Ch,"MODEL_CLASS_MAPPINGS",[Jp]);class Sh extends Wt{}le(Sh,"MODEL_CLASS_MAPPINGS",[Yp]);class $h extends Wt{}le($h,"MODEL_CLASS_MAPPINGS",[Zp]);class kh extends Wt{}le(kh,"MODEL_CLASS_MAPPINGS",[Lp]);class Ih extends Wt{}le(Ih,"MODEL_CLASS_MAPPINGS",[jp]);class bf extends Ae{constructor({logits:F,past_key_values:N,encoder_outputs:me,decoder_attentions:ke=null,cross_attentions:Se=null}){super(),this.logits=F,this.past_key_values=N,this.encoder_outputs=me,this.decoder_attentions=ke,this.cross_attentions=Se}}class At extends Ae{constructor({logits:F,...N}){super(),this.logits=F;const me=Object.values(N);me.length>0&&(this.attentions=me)}}class Ah extends Ae{constructor({logits:F,embeddings:N}){super(),this.logits=F,this.embeddings=N}}class Sr extends Ae{constructor({logits:F}){super(),this.logits=F}}class Or extends Ae{constructor({logits:F}){super(),this.logits=F}}class Vr extends Ae{constructor({start_logits:F,end_logits:N}){super(),this.start_logits=F,this.end_logits=N}}class fs extends Ae{constructor({logits:F}){super(),this.logits=F}}class vf extends Ae{constructor({logits:F,past_key_values:N}){super(),this.logits=F,this.past_key_values=N}}class Oh extends Ae{constructor({alphas:F}){super(),this.alphas=F}}class Fh extends Ae{constructor({waveform:F,spectrogram:N}){super(),this.waveform=F,this.spectrogram=N}}},"./src/models/audio_spectrogram_transformer/feature_extraction_audio_spectrogram_transformer.js":(e,r,t)=>{t.r(r),t.d(r,{ASTFeatureExtractor:()=>i});var n=t("./src/base/feature_extraction_utils.js");t("./src/utils/tensor.js");var o=t("./src/utils/audio.js");class i extends n.FeatureExtractor{constructor(d){super(d);const h=this.config.sampling_rate,f=(0,o.mel_filter_bank)(257,this.config.num_mel_bins,20,Math.floor(h/2),h,null,"kaldi",!0);this.mel_filters=f,this.window=(0,o.window_function)(400,"hann",{periodic:!1}),this.mean=this.config.mean,this.std=this.config.std}async _extract_fbank_features(d,h){return(0,o.spectrogram)(d,this.window,400,160,{fft_length:512,power:2,center:!1,preemphasis:.97,mel_filters:this.mel_filters,log_mel:"log",mel_floor:1192092955078125e-22,remove_dc_offset:!0,max_num_frames:h,transpose:!0})}async _call(d){(0,n.validate_audio_inputs)(d,"ASTFeatureExtractor");const h=await this._extract_fbank_features(d,this.config.max_length);if(this.config.do_normalize){const f=this.std*2,y=h.data;for(let m=0;m<y.length;++m)y[m]=(y[m]-this.mean)/f}return{input_values:h.unsqueeze_(0)}}}},"./src/models/auto/feature_extraction_auto.js":(e,r,t)=>{t.r(r),t.d(r,{AutoFeatureExtractor:()=>u});var n=t("./src/utils/constants.js"),o=t("./src/utils/hub.js");t("./src/base/feature_extraction_utils.js");var i=t("./src/models/feature_extractors.js");class u{static async from_pretrained(h,f={}){const y=await(0,o.getModelJSON)(h,n.FEATURE_EXTRACTOR_NAME,!0,f),m=y.feature_extractor_type,g=i[m];if(!g)throw new Error(`Unknown feature_extractor_type: '${m}'. Please report this at ${n.GITHUB_ISSUE_URL}.`);return new g(y)}}},"./src/models/auto/image_processing_auto.js":(e,r,t)=>{t.r(r),t.d(r,{AutoImageProcessor:()=>d});var n=t("./src/utils/constants.js"),o=t("./src/utils/hub.js"),i=t("./src/base/image_processors_utils.js"),u=t("./src/models/image_processors.js");class d{static async from_pretrained(f,y={}){const m=await(0,o.getModelJSON)(f,n.IMAGE_PROCESSOR_NAME,!0,y),g=m.image_processor_type??m.feature_extractor_type;let v=u[g];return v||(g!==void 0&&console.warn(`Image processor type '${g}' not found, assuming base ImageProcessor. Please report this at ${n.GITHUB_ISSUE_URL}.`),v=i.ImageProcessor),new v(m)}}},"./src/models/auto/processing_auto.js":(e,r,t)=>{t.r(r),t.d(r,{AutoProcessor:()=>f});var n=t("./src/utils/constants.js"),o=t("./src/utils/hub.js"),i=t("./src/base/processing_utils.js"),u=t("./src/models/processors.js"),d=t("./src/models/image_processors.js"),h=t("./src/models/feature_extractors.js");class f{static async from_pretrained(m,g={}){const v=await(0,o.getModelJSON)(m,n.IMAGE_PROCESSOR_NAME,!0,g),{image_processor_type:b,feature_extractor_type:k,processor_class:z}=v;if(z&&u[z])return u[z].from_pretrained(m,g);if(!b&&!k)throw new Error("No `image_processor_type` or `feature_extractor_type` found in the config.");const E={};if(b){const D=d[b];if(!D)throw new Error(`Unknown image_processor_type: '${b}'.`);E.image_processor=new D(v)}if(k){const D=d[k];if(D)E.image_processor=new D(v);else{const I=h[k];if(!I)throw new Error(`Unknown feature_extractor_type: '${k}'.`);E.feature_extractor=new I(v)}}const T={};return new i.Processor(T,E)}}},"./src/models/beit/image_processing_beit.js":(e,r,t)=>{t.r(r),t.d(r,{BeitFeatureExtractor:()=>o});var n=t("./src/base/image_processors_utils.js");class o extends n.ImageProcessor{}},"./src/models/bit/image_processing_bit.js":(e,r,t)=>{t.r(r),t.d(r,{BitImageProcessor:()=>o});var n=t("./src/base/image_processors_utils.js");class o extends n.ImageProcessor{}},"./src/models/chinese_clip/image_processing_chinese_clip.js":(e,r,t)=>{t.r(r),t.d(r,{ChineseCLIPFeatureExtractor:()=>o});var n=t("./src/base/image_processors_utils.js");class o extends n.ImageProcessor{}},"./src/models/clap/feature_extraction_clap.js":(e,r,t)=>{t.r(r),t.d(r,{ClapFeatureExtractor:()=>i});var n=t("./src/base/feature_extraction_utils.js");t("./src/utils/tensor.js");var o=t("./src/utils/audio.js");class i extends n.FeatureExtractor{constructor(d){super(d),this.mel_filters=(0,o.mel_filter_bank)(this.config.nb_frequency_bins,this.config.feature_size,this.config.frequency_min,this.config.frequency_max,this.config.sampling_rate,null,"htk"),this.mel_filters_slaney=(0,o.mel_filter_bank)(this.config.nb_frequency_bins,this.config.feature_size,this.config.frequency_min,this.config.frequency_max,this.config.sampling_rate,"slaney","slaney"),this.window=(0,o.window_function)(this.config.fft_window_size,"hann")}async _get_input_mel(d,h,f,y){let m;const g=d.length-h;if(g>0)if(f==="rand_trunc"){const v=Math.floor(Math.random()*(g+1));d=d.subarray(v,v+h),m=await this._extract_fbank_features(d,this.mel_filters_slaney,this.config.nb_max_samples)}else throw new Error(`Truncation strategy "${f}" not implemented`);else{if(g<0){let v=new Float64Array(h);if(v.set(d),y==="repeat")for(let b=d.length;b<h;b+=d.length)v.set(d.subarray(0,Math.min(d.length,h-b)),b);else if(y==="repeatpad")for(let b=d.length;b<-g;b+=d.length)v.set(d,b);d=v}if(f==="fusion")throw new Error(`Truncation strategy "${f}" not implemented`);m=await this._extract_fbank_features(d,this.mel_filters_slaney,this.config.nb_max_samples)}return m.unsqueeze_(0)}async _extract_fbank_features(d,h,f=null){return(0,o.spectrogram)(d,this.window,this.config.fft_window_size,this.config.hop_length,{power:2,mel_filters:h,log_mel:"dB",max_num_frames:f,do_pad:!1,transpose:!0})}async _call(d,{max_length:h=null}={}){return(0,n.validate_audio_inputs)(d,"ClapFeatureExtractor"),{input_features:(await this._get_input_mel(d,h??this.config.nb_max_samples,this.config.truncation,this.config.padding)).unsqueeze_(0)}}}},"./src/models/clip/image_processing_clip.js":(e,r,t)=>{t.r(r),t.d(r,{CLIPFeatureExtractor:()=>i,CLIPImageProcessor:()=>o});var n=t("./src/base/image_processors_utils.js");class o extends n.ImageProcessor{}class i extends o{}},"./src/models/convnext/image_processing_convnext.js":(e,r,t)=>{t.r(r),t.d(r,{ConvNextFeatureExtractor:()=>i,ConvNextImageProcessor:()=>o});var n=t("./src/base/image_processors_utils.js");class o extends n.ImageProcessor{constructor(d){super(d),this.crop_pct=this.config.crop_pct??224/256}async resize(d){var f;const h=(f=this.size)==null?void 0:f.shortest_edge;if(h===void 0)throw new Error("Size dictionary must contain 'shortest_edge' key.");if(h<384){const y=Math.floor(h/this.crop_pct),[m,g]=this.get_resize_output_image_size(d,{shortest_edge:y});d=await d.resize(m,g,{resample:this.resample}),d=await d.center_crop(h,h)}else d=await d.resize(h,h,{resample:this.resample});return d}}class i extends o{}},"./src/models/dac/feature_extraction_dac.js":(e,r,t)=>{t.r(r),t.d(r,{DacFeatureExtractor:()=>o});var n=t("./src/models/encodec/feature_extraction_encodec.js");class o extends n.EncodecFeatureExtractor{}},"./src/models/deit/image_processing_deit.js":(e,r,t)=>{t.r(r),t.d(r,{DeiTFeatureExtractor:()=>i,DeiTImageProcessor:()=>o});var n=t("./src/base/image_processors_utils.js");class o extends n.ImageProcessor{}class i extends o{}},"./src/models/detr/image_processing_detr.js":(e,r,t)=>{t.r(r),t.d(r,{DetrFeatureExtractor:()=>u,DetrImageProcessor:()=>i});var n=t("./src/base/image_processors_utils.js"),o=t("./src/utils/tensor.js");class i extends n.ImageProcessor{async _call(h){const f=await super._call(h),y=[f.pixel_values.dims[0],64,64],m=(0,o.full)(y,1n);return{...f,pixel_mask:m}}post_process_object_detection(...h){return(0,n.post_process_object_detection)(...h)}post_process_panoptic_segmentation(...h){return(0,n.post_process_panoptic_segmentation)(...h)}post_process_instance_segmentation(...h){return(0,n.post_process_instance_segmentation)(...h)}}class u extends i{}},"./src/models/donut/image_processing_donut.js":(e,r,t)=>{t.r(r),t.d(r,{DonutFeatureExtractor:()=>i,DonutImageProcessor:()=>o});var n=t("./src/base/image_processors_utils.js");class o extends n.ImageProcessor{pad_image(d,h,f,y={}){const[m,g,v]=h;let b=this.image_mean;Array.isArray(this.image_mean)||(b=new Array(v).fill(b));let k=this.image_std;Array.isArray(k)||(k=new Array(v).fill(b));const z=b.map((E,T)=>-E/k[T]);return super.pad_image(d,h,f,{center:!0,constant_values:z,...y})}}class i extends o{}},"./src/models/dpt/image_processing_dpt.js":(e,r,t)=>{t.r(r),t.d(r,{DPTFeatureExtractor:()=>i,DPTImageProcessor:()=>o});var n=t("./src/base/image_processors_utils.js");class o extends n.ImageProcessor{}class i extends o{}},"./src/models/efficientnet/image_processing_efficientnet.js":(e,r,t)=>{t.r(r),t.d(r,{EfficientNetImageProcessor:()=>o});var n=t("./src/base/image_processors_utils.js");class o extends n.ImageProcessor{constructor(u){super(u),this.include_top=this.config.include_top??!0,this.include_top&&(this.image_std=this.image_std.map(d=>d*d))}}},"./src/models/encodec/feature_extraction_encodec.js":(e,r,t)=>{t.r(r),t.d(r,{EncodecFeatureExtractor:()=>i});var n=t("./src/base/feature_extraction_utils.js"),o=t("./src/utils/tensor.js");class i extends n.FeatureExtractor{async _call(d){(0,n.validate_audio_inputs)(d,"EncodecFeatureExtractor"),d instanceof Float64Array&&(d=new Float32Array(d));const h=this.config.feature_size;if(d.length%h!==0)throw new Error(`The length of the audio data must be a multiple of the number of channels (${h}).`);const f=[1,h,d.length/h];return{input_values:new o.Tensor("float32",d,f)}}}},"./src/models/feature_extractors.js":(e,r,t)=>{t.r(r),t.d(r,{ASTFeatureExtractor:()=>n.ASTFeatureExtractor,ClapFeatureExtractor:()=>i.ClapFeatureExtractor,DacFeatureExtractor:()=>u.DacFeatureExtractor,EncodecFeatureExtractor:()=>o.EncodecFeatureExtractor,ImageFeatureExtractor:()=>k.ImageProcessor,MoonshineFeatureExtractor:()=>d.MoonshineFeatureExtractor,PyAnnoteFeatureExtractor:()=>h.PyAnnoteFeatureExtractor,SeamlessM4TFeatureExtractor:()=>f.SeamlessM4TFeatureExtractor,SnacFeatureExtractor:()=>y.SnacFeatureExtractor,SpeechT5FeatureExtractor:()=>m.SpeechT5FeatureExtractor,Wav2Vec2FeatureExtractor:()=>g.Wav2Vec2FeatureExtractor,WeSpeakerFeatureExtractor:()=>v.WeSpeakerFeatureExtractor,WhisperFeatureExtractor:()=>b.WhisperFeatureExtractor});var n=t("./src/models/audio_spectrogram_transformer/feature_extraction_audio_spectrogram_transformer.js"),o=t("./src/models/encodec/feature_extraction_encodec.js"),i=t("./src/models/clap/feature_extraction_clap.js"),u=t("./src/models/dac/feature_extraction_dac.js"),d=t("./src/models/moonshine/feature_extraction_moonshine.js"),h=t("./src/models/pyannote/feature_extraction_pyannote.js"),f=t("./src/models/seamless_m4t/feature_extraction_seamless_m4t.js"),y=t("./src/models/snac/feature_extraction_snac.js"),m=t("./src/models/speecht5/feature_extraction_speecht5.js"),g=t("./src/models/wav2vec2/feature_extraction_wav2vec2.js"),v=t("./src/models/wespeaker/feature_extraction_wespeaker.js"),b=t("./src/models/whisper/feature_extraction_whisper.js"),k=t("./src/base/image_processors_utils.js")},"./src/models/florence2/processing_florence2.js":(e,r,t)=>{t.r(r),t.d(r,{Florence2Processor:()=>u});var n=t("./src/base/processing_utils.js"),o=t("./src/models/auto/image_processing_auto.js"),i=t("./src/tokenizers.js");class u extends n.Processor{constructor(h,f){super(h,f);const{tasks_answer_post_processing_type:y,task_prompts_without_inputs:m,task_prompts_with_input:g}=this.image_processor.config;this.tasks_answer_post_processing_type=new Map(Object.entries(y??{})),this.task_prompts_without_inputs=new Map(Object.entries(m??{})),this.task_prompts_with_input=new Map(Object.entries(g??{})),this.regexes={quad_boxes:/(.+?)<loc_(\d+)><loc_(\d+)><loc_(\d+)><loc_(\d+)><loc_(\d+)><loc_(\d+)><loc_(\d+)><loc_(\d+)>/gm,bboxes:/([^<]+)?<loc_(\d+)><loc_(\d+)><loc_(\d+)><loc_(\d+)>/gm},this.size_per_bin=1e3}construct_prompts(h){typeof h=="string"&&(h=[h]);const f=[];for(const y of h)if(this.task_prompts_without_inputs.has(y))f.push(this.task_prompts_without_inputs.get(y));else{for(const[m,g]of this.task_prompts_with_input)if(y.includes(m)){f.push(g.replaceAll("{input}",y).replaceAll(m,""));break}f.length!==h.length&&f.push(y)}return f}post_process_generation(h,f,y){const m=this.tasks_answer_post_processing_type.get(f)??"pure_text";h=h.replaceAll("<s>","").replaceAll("</s>","");let g;switch(m){case"pure_text":g=h;break;case"description_with_bboxes":case"bboxes":case"phrase_grounding":case"ocr":const v=m==="ocr"?"quad_boxes":"bboxes",b=h.matchAll(this.regexes[v]),k=[],z=[];for(const[E,T,...D]of b)k.push(T?T.trim():k.at(-1)??""),z.push(D.map((I,$)=>(Number(I)+.5)/this.size_per_bin*y[$%2]));g={labels:k,[v]:z};break;default:throw new Error(`Task "${f}" (of type "${m}") not yet implemented.`)}return{[f]:g}}async _call(h,f=null,y={}){if(!h&&!f)throw new Error("Either text or images must be provided");const m=await this.image_processor(h,y),g=f?this.tokenizer(f,y):{};return{...m,...g}}}le(u,"tokenizer_class",i.AutoTokenizer),le(u,"image_processor_class",o.AutoImageProcessor)},"./src/models/glpn/image_processing_glpn.js":(e,r,t)=>{t.r(r),t.d(r,{GLPNFeatureExtractor:()=>o});var n=t("./src/base/image_processors_utils.js");class o extends n.ImageProcessor{}},"./src/models/grounding_dino/image_processing_grounding_dino.js":(e,r,t)=>{t.r(r),t.d(r,{GroundingDinoImageProcessor:()=>i});var n=t("./src/base/image_processors_utils.js"),o=t("./src/utils/tensor.js");class i extends n.ImageProcessor{async _call(d){const h=await super._call(d),f=h.pixel_values.dims,y=(0,o.ones)([f[0],f[2],f[3]]);return{...h,pixel_mask:y}}}},"./src/models/grounding_dino/processing_grounding_dino.js":(e,r,t)=>{t.r(r),t.d(r,{GroundingDinoProcessor:()=>h});var n=t("./src/base/processing_utils.js"),o=t("./src/models/auto/image_processing_auto.js"),i=t("./src/tokenizers.js"),u=t("./src/base/image_processors_utils.js");function d(f,y){const g=f.dims.at(-1)-1,v=f.tolist();v.fill(!1,0,1),v.fill(!1,g);const b=y.tolist();return v.map((k,z)=>k?z:null).filter(k=>k!==null).map(k=>b[k])}class h extends n.Processor{async _call(y,m,g={}){const v=y?await this.image_processor(y,g):{};return{...m?this.tokenizer(m,g):{},...v}}post_process_grounded_object_detection(y,m,{box_threshold:g=.25,text_threshold:v=.25,target_sizes:b=null}={}){const{logits:k,pred_boxes:z}=y,E=k.dims[0];if(b!==null&&b.length!==E)throw Error("Make sure that you pass in as many target sizes as the batch dimension of the logits");const T=k.dims.at(1),D=k.sigmoid(),I=D.max(-1).tolist(),$=z.tolist().map(S=>S.map(O=>(0,u.center_to_corners_format)(O))),P=[];for(let S=0;S<E;++S){const O=b!==null?b[S]:null;O!==null&&($[S]=$[S].map(H=>H.map((ce,re)=>ce*O[(re+1)%2])));const R=I[S],G=[],ee=[],ie=[];for(let H=0;H<T;++H){const ce=R[H];if(ce<=g)continue;const re=$[S][H],se=D[S][H];G.push(ce),ie.push(re);const _e=d(se.gt(v),m[S]);ee.push(_e)}P.push({scores:G,boxes:ie,labels:this.batch_decode(ee)})}return P}}le(h,"tokenizer_class",i.AutoTokenizer),le(h,"image_processor_class",o.AutoImageProcessor)},"./src/models/idefics3/image_processing_idefics3.js":(e,r,t)=>{t.r(r),t.d(r,{Idefics3ImageProcessor:()=>i});var n=t("./src/base/image_processors_utils.js"),o=t("./src/utils/tensor.js");class i extends n.ImageProcessor{constructor(d){super(d),this.do_image_splitting=d.do_image_splitting??!0,this.max_image_size=d.max_image_size}get_resize_for_vision_encoder(d,h){let[f,y]=d.dims.slice(-2);const m=y/f;return y>=f?(y=Math.ceil(y/h)*h,f=Math.floor(y/m),f=Math.ceil(f/h)*h):(f=Math.ceil(f/h)*h,y=Math.floor(f*m),y=Math.ceil(y/h)*h),{height:f,width:y}}async _call(d,{do_image_splitting:h=null,return_row_col_info:f=!1}={}){let y;if(!Array.isArray(d))y=[[d]];else{if(d.length===0||!d[0])throw new Error("No images provided.");Array.isArray(d[0])?y=d:y=[d]}let m=[],g=[],v=[];const b=[],k=[];for(const S of y){let O=await Promise.all(S.map(ee=>this.preprocess(ee)));b.push(...O.map(ee=>ee.original_size)),k.push(...O.map(ee=>ee.reshaped_input_size)),O.forEach(ee=>ee.pixel_values.unsqueeze_(0));const{longest_edge:R}=this.max_image_size;let G;if(h??this.do_image_splitting){let ee=new Array(O.length),ie=new Array(O.length);G=await Promise.all(O.map(async(H,ce)=>{const re=this.get_resize_for_vision_encoder(H.pixel_values,R),se=await(0,o.interpolate_4d)(H.pixel_values,{size:[re.height,re.width]}),{frames:_e,num_splits_h:ae,num_splits_w:Ce}=await this.split_image(se,this.max_image_size);return ee[ce]=ae,ie[ce]=Ce,(0,o.cat)(_e,0)})),g.push(ee),v.push(ie)}else{const ee=[R,R];G=await Promise.all(O.map(ie=>(0,o.interpolate_4d)(ie.pixel_values,{size:ee}))),g.push(new Array(O.length).fill(0)),v.push(new Array(O.length).fill(0))}m.push((0,o.cat)(G,0))}const z=m.length,[E,T,D,I]=m[0].dims;let $,P;if(z===1)$=m[0].unsqueeze_(0),P=(0,o.full)([z,E,D,I],!0);else{const S=Math.max(...m.map(G=>G.dims.at(0)));P=(0,o.full)([z,S,D,I],!0);const O=P.data,R=S*D*I;for(let G=0;G<z;++G){const ee=m[G].dims[0];if(ee<S){m[G]=(0,o.cat)([m[G],(0,o.full)([S-ee,T,D,I],0)],0);const ie=G*R+ee*D*I,H=(G+1)*R;O.fill(!1,ie,H)}}$=(0,o.stack)(m,0)}return{pixel_values:$,pixel_attention_mask:P,original_sizes:b,reshaped_input_sizes:k,...f?{rows:g,cols:v}:{}}}async split_image(d,{longest_edge:h}){const f=h,y=h,m=[],[g,v]=d.dims.slice(-2);let b=0,k=0;if(g>f||v>y){b=Math.ceil(g/f),k=Math.ceil(v/y);const z=Math.ceil(g/b),E=Math.ceil(v/k);for(let I=0;I<b;++I)for(let $=0;$<k;++$){let P,S,O,R;I===b-1?(S=g-z,R=g):(S=I*z,R=(I+1)*z),$===k-1?(P=v-E,O=v):(P=$*E,O=($+1)*E);const G=[S,P],ee=[R,O],ie=await(0,o.slice)(d,G,ee,[2,3]);m.push(ie)}const T=f,D=y;(g!==T||v!==D)&&(d=await(0,o.interpolate_4d)(d,{size:[T,D]}))}return m.push(d),{frames:m,num_splits_h:b,num_splits_w:k}}}},"./src/models/idefics3/processing_idefics3.js":(e,r,t)=>{t.r(r),t.d(r,{Idefics3Processor:()=>y});var n=t("./src/base/processing_utils.js"),o=t("./src/models/auto/image_processing_auto.js"),i=t("./src/tokenizers.js");t("./src/utils/image.js");var u=t("./src/utils/core.js");function d(m,g,v,b,k,z){let E="";for(let T=0;T<g;++T){for(let D=0;D<v;++D)E+=b+`<row_${T+1}_col_${D+1}>`+k.repeat(m);E+=`
`}return E+=`
${b}${z}`+k.repeat(m)+`${b}`,E}function h(m,g,v,b){return`${g}${b}`+v.repeat(m)+`${g}`}function f(m,g,v,b,k,z){return m===0&&g===0?h(v,b,k,z):d(v,m,g,b,k,z)}class y extends n.Processor{constructor(){super(...arguments);le(this,"fake_image_token","<fake_token_around_image>");le(this,"image_token","<image>");le(this,"global_img_token","<global-img>")}async _call(v,b=null,k={}){k.return_row_col_info??(k.return_row_col_info=!0);let z;b&&(z=await this.image_processor(b,k)),Array.isArray(v)||(v=[v]);const E=z.rows??[new Array(v.length).fill(0)],T=z.cols??[new Array(v.length).fill(0)],D=this.config.image_seq_len,I=[],$=[];for(let S=0;S<v.length;++S){const O=v[S],R=E[S],G=T[S];I.push((0,u.count)(O,this.image_token));const ee=R.map((ce,re)=>f(ce,G[re],D,this.fake_image_token,this.image_token,this.global_img_token)),ie=O.split(this.image_token);if(ie.length===0)throw new Error("The image token should be present in the text.");let H=ie[0];for(let ce=0;ce<ee.length;++ce)H+=ee[ce]+ie[ce+1];$.push(H)}return{...this.tokenizer($),...z}}}le(y,"image_processor_class",o.AutoImageProcessor),le(y,"tokenizer_class",i.AutoTokenizer),le(y,"uses_processor_config",!0)},"./src/models/image_processors.js":(e,r,t)=>{t.r(r),t.d(r,{BeitFeatureExtractor:()=>n.BeitFeatureExtractor,BitImageProcessor:()=>o.BitImageProcessor,CLIPFeatureExtractor:()=>u.CLIPFeatureExtractor,CLIPImageProcessor:()=>u.CLIPImageProcessor,ChineseCLIPFeatureExtractor:()=>i.ChineseCLIPFeatureExtractor,ConvNextFeatureExtractor:()=>d.ConvNextFeatureExtractor,ConvNextImageProcessor:()=>d.ConvNextImageProcessor,DPTFeatureExtractor:()=>m.DPTFeatureExtractor,DPTImageProcessor:()=>m.DPTImageProcessor,DeiTFeatureExtractor:()=>h.DeiTFeatureExtractor,DeiTImageProcessor:()=>h.DeiTImageProcessor,DetrFeatureExtractor:()=>f.DetrFeatureExtractor,DetrImageProcessor:()=>f.DetrImageProcessor,DonutFeatureExtractor:()=>y.DonutFeatureExtractor,DonutImageProcessor:()=>y.DonutImageProcessor,EfficientNetImageProcessor:()=>g.EfficientNetImageProcessor,GLPNFeatureExtractor:()=>v.GLPNFeatureExtractor,GroundingDinoImageProcessor:()=>b.GroundingDinoImageProcessor,Idefics3ImageProcessor:()=>k.Idefics3ImageProcessor,JinaCLIPImageProcessor:()=>E.JinaCLIPImageProcessor,LlavaOnevisionImageProcessor:()=>T.LlavaOnevisionImageProcessor,Mask2FormerImageProcessor:()=>D.Mask2FormerImageProcessor,MaskFormerFeatureExtractor:()=>I.MaskFormerFeatureExtractor,MaskFormerImageProcessor:()=>I.MaskFormerImageProcessor,MobileNetV1FeatureExtractor:()=>$.MobileNetV1FeatureExtractor,MobileNetV1ImageProcessor:()=>$.MobileNetV1ImageProcessor,MobileNetV2FeatureExtractor:()=>P.MobileNetV2FeatureExtractor,MobileNetV2ImageProcessor:()=>P.MobileNetV2ImageProcessor,MobileNetV3FeatureExtractor:()=>S.MobileNetV3FeatureExtractor,MobileNetV3ImageProcessor:()=>S.MobileNetV3ImageProcessor,MobileNetV4FeatureExtractor:()=>O.MobileNetV4FeatureExtractor,MobileNetV4ImageProcessor:()=>O.MobileNetV4ImageProcessor,MobileViTFeatureExtractor:()=>R.MobileViTFeatureExtractor,MobileViTImageProcessor:()=>R.MobileViTImageProcessor,NougatImageProcessor:()=>G.NougatImageProcessor,OwlViTFeatureExtractor:()=>ie.OwlViTFeatureExtractor,OwlViTImageProcessor:()=>ie.OwlViTImageProcessor,Owlv2ImageProcessor:()=>ee.Owlv2ImageProcessor,Phi3VImageProcessor:()=>H.Phi3VImageProcessor,PvtImageProcessor:()=>ce.PvtImageProcessor,Qwen2VLImageProcessor:()=>re.Qwen2VLImageProcessor,RTDetrImageProcessor:()=>se.RTDetrImageProcessor,SamImageProcessor:()=>_e.SamImageProcessor,SegformerFeatureExtractor:()=>ae.SegformerFeatureExtractor,SegformerImageProcessor:()=>ae.SegformerImageProcessor,SiglipImageProcessor:()=>Ce.SiglipImageProcessor,SmolVLMImageProcessor:()=>Te.SmolVLMImageProcessor,Swin2SRImageProcessor:()=>q.Swin2SRImageProcessor,VLMImageProcessor:()=>z.VLMImageProcessor,ViTFeatureExtractor:()=>B.ViTFeatureExtractor,ViTImageProcessor:()=>B.ViTImageProcessor,VitMatteImageProcessor:()=>Q.VitMatteImageProcessor,VitPoseImageProcessor:()=>oe.VitPoseImageProcessor,YolosFeatureExtractor:()=>Ie.YolosFeatureExtractor,YolosImageProcessor:()=>Ie.YolosImageProcessor});var n=t("./src/models/beit/image_processing_beit.js"),o=t("./src/models/bit/image_processing_bit.js"),i=t("./src/models/chinese_clip/image_processing_chinese_clip.js"),u=t("./src/models/clip/image_processing_clip.js"),d=t("./src/models/convnext/image_processing_convnext.js"),h=t("./src/models/deit/image_processing_deit.js"),f=t("./src/models/detr/image_processing_detr.js"),y=t("./src/models/donut/image_processing_donut.js"),m=t("./src/models/dpt/image_processing_dpt.js"),g=t("./src/models/efficientnet/image_processing_efficientnet.js"),v=t("./src/models/glpn/image_processing_glpn.js"),b=t("./src/models/grounding_dino/image_processing_grounding_dino.js"),k=t("./src/models/idefics3/image_processing_idefics3.js"),z=t("./src/models/janus/image_processing_janus.js"),E=t("./src/models/jina_clip/image_processing_jina_clip.js"),T=t("./src/models/llava_onevision/image_processing_llava_onevision.js"),D=t("./src/models/mask2former/image_processing_mask2former.js"),I=t("./src/models/maskformer/image_processing_maskformer.js"),$=t("./src/models/mobilenet_v1/image_processing_mobilenet_v1.js"),P=t("./src/models/mobilenet_v2/image_processing_mobilenet_v2.js"),S=t("./src/models/mobilenet_v3/image_processing_mobilenet_v3.js"),O=t("./src/models/mobilenet_v4/image_processing_mobilenet_v4.js"),R=t("./src/models/mobilevit/image_processing_mobilevit.js"),G=t("./src/models/nougat/image_processing_nougat.js"),ee=t("./src/models/owlv2/image_processing_owlv2.js"),ie=t("./src/models/owlvit/image_processing_owlvit.js"),H=t("./src/models/phi3_v/image_processing_phi3_v.js"),ce=t("./src/models/pvt/image_processing_pvt.js"),re=t("./src/models/qwen2_vl/image_processing_qwen2_vl.js"),se=t("./src/models/rt_detr/image_processing_rt_detr.js"),_e=t("./src/models/sam/image_processing_sam.js"),ae=t("./src/models/segformer/image_processing_segformer.js"),Ce=t("./src/models/siglip/image_processing_siglip.js"),Te=t("./src/models/smolvlm/image_processing_smolvlm.js"),q=t("./src/models/swin2sr/image_processing_swin2sr.js"),B=t("./src/models/vit/image_processing_vit.js"),Q=t("./src/models/vitmatte/image_processing_vitmatte.js"),oe=t("./src/models/vitpose/image_processing_vitpose.js"),Ie=t("./src/models/yolos/image_processing_yolos.js")},"./src/models/janus/image_processing_janus.js":(e,r,t)=>{t.r(r),t.d(r,{VLMImageProcessor:()=>o});var n=t("./src/base/image_processors_utils.js");class o extends n.ImageProcessor{constructor(u){super({do_pad:!0,pad_size:{width:u.image_size,height:u.image_size},...u}),this.constant_values=this.config.background_color.map(d=>d*this.rescale_factor)}pad_image(u,d,h,f){return super.pad_image(u,d,h,{constant_values:this.constant_values,center:!0,...f})}}},"./src/models/janus/processing_janus.js":(e,r,t)=>{t.r(r),t.d(r,{VLChatProcessor:()=>f});var n=t("./src/base/processing_utils.js"),o=t("./src/models/auto/image_processing_auto.js"),i=t("./src/tokenizers.js"),u=t("./src/utils/core.js"),d=t("./src/utils/tensor.js"),h=t("./src/utils/image.js");class f extends n.Processor{constructor(m,g){super(m,g),this.image_tag=this.config.image_tag,this.image_start_tag=this.config.image_start_tag,this.image_end_tag=this.config.image_end_tag,this.num_image_tokens=this.config.num_image_tokens}async _call(m,{images:g=null,chat_template:v="default"}={}){g?Array.isArray(g)||(g=[g]):g=await Promise.all(m.filter(G=>G.images).flatMap(G=>G.images).map(G=>h.RawImage.read(G)));const b=this.tokenizer,k=b.apply_chat_template(m,{tokenize:!1,add_generation_prompt:!0,chat_template:v}),z=G=>b.encode(G,{add_special_tokens:!1}),E=k.split(this.image_tag),T=E.length-1;if(g.length!==T)throw new Error(`Number of images provided (${g.length}) does not match number of "${this.image_tag}" image tags (${T})`);const[D,I,$]=b.model.convert_tokens_to_ids([this.image_tag,this.image_start_tag,this.image_end_tag]);let P=z(E[0]),S=new Array(P.length).fill(!1);for(let G=1;G<E.length;++G){const ee=new Array(this.num_image_tokens).fill(D),ie=z(E[G]);P=(0,u.mergeArrays)(P,[I],ee,[$],ie);const H=new Array(this.num_image_tokens).fill(!0);S=(0,u.mergeArrays)(S,[!1],H,[!1],new Array(ie.length).fill(!1))}const O=[1,P.length],R={input_ids:new d.Tensor("int64",P,O),attention_mask:new d.Tensor("int64",new Array(P.length).fill(1),O),images_seq_mask:new d.Tensor("bool",S,O),images_emb_mask:new d.Tensor("bool",new Array(T*this.num_image_tokens).fill(!0),[1,T,this.num_image_tokens])};if(g&&g.length>0){const G=await this.image_processor(g);return G.pixel_values.unsqueeze_(0),{...R,...G}}return R}}le(f,"image_processor_class",o.AutoImageProcessor),le(f,"tokenizer_class",i.AutoTokenizer),le(f,"uses_processor_config",!0)},"./src/models/jina_clip/image_processing_jina_clip.js":(e,r,t)=>{t.r(r),t.d(r,{JinaCLIPImageProcessor:()=>o});var n=t("./src/base/image_processors_utils.js");class o extends n.ImageProcessor{constructor(u){const{resize_mode:d,fill_color:h,interpolation:f,size:y,...m}=u,g=d==="squash"?{width:y,height:y}:d==="shortest"?{shortest_edge:y}:{longest_edge:y},v=f==="bicubic"?3:2;super({...m,size:g,resample:v,do_center_crop:!0,crop_size:y,do_normalize:!0})}}},"./src/models/jina_clip/processing_jina_clip.js":(e,r,t)=>{t.r(r),t.d(r,{JinaCLIPProcessor:()=>u});var n=t("./src/base/processing_utils.js"),o=t("./src/models/auto/image_processing_auto.js"),i=t("./src/tokenizers.js");class u extends n.Processor{async _call(h=null,f=null,y={}){if(!h&&!f)throw new Error("Either text or images must be provided");const m=h?this.tokenizer(h,y):{},g=f?await this.image_processor(f,y):{};return{...m,...g}}}le(u,"tokenizer_class",i.AutoTokenizer),le(u,"image_processor_class",o.AutoImageProcessor)},"./src/models/llava_onevision/image_processing_llava_onevision.js":(e,r,t)=>{t.r(r),t.d(r,{LlavaOnevisionImageProcessor:()=>o});var n=t("./src/base/image_processors_utils.js");class o extends n.ImageProcessor{}},"./src/models/mask2former/image_processing_mask2former.js":(e,r,t)=>{t.r(r),t.d(r,{Mask2FormerImageProcessor:()=>o});var n=t("./src/models/maskformer/image_processing_maskformer.js");class o extends n.MaskFormerImageProcessor{}},"./src/models/maskformer/image_processing_maskformer.js":(e,r,t)=>{t.r(r),t.d(r,{MaskFormerFeatureExtractor:()=>i,MaskFormerImageProcessor:()=>o});var n=t("./src/base/image_processors_utils.js");class o extends n.ImageProcessor{post_process_panoptic_segmentation(...d){return(0,n.post_process_panoptic_segmentation)(...d)}post_process_instance_segmentation(...d){return(0,n.post_process_instance_segmentation)(...d)}}class i extends o{}},"./src/models/mgp_str/processing_mgp_str.js":(e,r,t)=>{t.r(r),t.d(r,{MgpstrProcessor:()=>h});var n=t("./src/base/processing_utils.js"),o=t("./src/models/auto/image_processing_auto.js"),i=t("./src/tokenizers.js"),u=t("./src/utils/maths.js");const d={char:["char_decode",1],bpe:["bpe_decode",2],wp:["wp_decode",102]};class h extends n.Processor{get char_tokenizer(){return this.components.char_tokenizer}get bpe_tokenizer(){return this.components.bpe_tokenizer}get wp_tokenizer(){return this.components.wp_tokenizer}_decode_helper(y,m){if(!d.hasOwnProperty(m))throw new Error(`Format ${m} is not supported.`);const[g,v]=d[m],b=this[g].bind(this),[k,z]=y.dims,E=[],T=[],D=y.tolist();for(let $=0;$<k;++$){const P=D[$],S=[],O=[];for(let G=1;G<z;++G){const[ee,ie]=(0,u.max)((0,u.softmax)(P[G]));if(O.push(ee),ie==v)break;S.push(ie)}const R=O.length>0?O.reduce((G,ee)=>G*ee,1):0;T.push(S),E.push(R)}return[b(T),E]}char_decode(y){return this.char_tokenizer.batch_decode(y).map(m=>m.replaceAll(" ",""))}bpe_decode(y){return this.bpe_tokenizer.batch_decode(y)}wp_decode(y){return this.wp_tokenizer.batch_decode(y).map(m=>m.replaceAll(" ",""))}batch_decode([y,m,g]){const[v,b]=this._decode_helper(y,"char"),[k,z]=this._decode_helper(m,"bpe"),[E,T]=this._decode_helper(g,"wp"),D=[],I=[];for(let $=0;$<v.length;++$){const[P,S]=(0,u.max)([b[$],z[$],T[$]]);D.push([v[$],k[$],E[$]][S]),I.push(P)}return{generated_text:D,scores:I,char_preds:v,bpe_preds:k,wp_preds:E}}static async from_pretrained(...y){const m=await super.from_pretrained(...y),g=await i.AutoTokenizer.from_pretrained("Xenova/gpt2"),v=await i.AutoTokenizer.from_pretrained("Xenova/bert-base-uncased");return m.components={image_processor:m.image_processor,char_tokenizer:m.tokenizer,bpe_tokenizer:g,wp_tokenizer:v},m}async _call(y,m=null){const g=await this.image_processor(y);return m&&(g.labels=this.tokenizer(m).input_ids),g}}le(h,"tokenizer_class",i.AutoTokenizer),le(h,"image_processor_class",o.AutoImageProcessor)},"./src/models/mobilenet_v1/image_processing_mobilenet_v1.js":(e,r,t)=>{t.r(r),t.d(r,{MobileNetV1FeatureExtractor:()=>i,MobileNetV1ImageProcessor:()=>o});var n=t("./src/base/image_processors_utils.js");class o extends n.ImageProcessor{}class i extends o{}},"./src/models/mobilenet_v2/image_processing_mobilenet_v2.js":(e,r,t)=>{t.r(r),t.d(r,{MobileNetV2FeatureExtractor:()=>i,MobileNetV2ImageProcessor:()=>o});var n=t("./src/base/image_processors_utils.js");class o extends n.ImageProcessor{}class i extends o{}},"./src/models/mobilenet_v3/image_processing_mobilenet_v3.js":(e,r,t)=>{t.r(r),t.d(r,{MobileNetV3FeatureExtractor:()=>i,MobileNetV3ImageProcessor:()=>o});var n=t("./src/base/image_processors_utils.js");class o extends n.ImageProcessor{}class i extends o{}},"./src/models/mobilenet_v4/image_processing_mobilenet_v4.js":(e,r,t)=>{t.r(r),t.d(r,{MobileNetV4FeatureExtractor:()=>i,MobileNetV4ImageProcessor:()=>o});var n=t("./src/base/image_processors_utils.js");class o extends n.ImageProcessor{}class i extends o{}},"./src/models/mobilevit/image_processing_mobilevit.js":(e,r,t)=>{t.r(r),t.d(r,{MobileViTFeatureExtractor:()=>i,MobileViTImageProcessor:()=>o});var n=t("./src/base/image_processors_utils.js");class o extends n.ImageProcessor{}class i extends o{}},"./src/models/moonshine/feature_extraction_moonshine.js":(e,r,t)=>{t.r(r),t.d(r,{MoonshineFeatureExtractor:()=>i});var n=t("./src/base/feature_extraction_utils.js"),o=t("./src/utils/tensor.js");class i extends n.FeatureExtractor{async _call(d){(0,n.validate_audio_inputs)(d,"MoonshineFeatureExtractor"),d instanceof Float64Array&&(d=new Float32Array(d));const h=[1,d.length];return{input_values:new o.Tensor("float32",d,h)}}}},"./src/models/moonshine/processing_moonshine.js":(e,r,t)=>{t.r(r),t.d(r,{MoonshineProcessor:()=>u});var n=t("./src/models/auto/feature_extraction_auto.js"),o=t("./src/tokenizers.js"),i=t("./src/base/processing_utils.js");class u extends i.Processor{async _call(h){return await this.feature_extractor(h)}}le(u,"tokenizer_class",o.AutoTokenizer),le(u,"feature_extractor_class",n.AutoFeatureExtractor)},"./src/models/nougat/image_processing_nougat.js":(e,r,t)=>{t.r(r),t.d(r,{NougatImageProcessor:()=>o});var n=t("./src/models/donut/image_processing_donut.js");class o extends n.DonutImageProcessor{}},"./src/models/owlv2/image_processing_owlv2.js":(e,r,t)=>{t.r(r),t.d(r,{Owlv2ImageProcessor:()=>o});var n=t("./src/models/owlvit/image_processing_owlvit.js");class o extends n.OwlViTImageProcessor{}},"./src/models/owlvit/image_processing_owlvit.js":(e,r,t)=>{t.r(r),t.d(r,{OwlViTFeatureExtractor:()=>i,OwlViTImageProcessor:()=>o});var n=t("./src/base/image_processors_utils.js");class o extends n.ImageProcessor{post_process_object_detection(...d){return(0,n.post_process_object_detection)(...d)}}class i extends o{}},"./src/models/owlvit/processing_owlvit.js":(e,r,t)=>{t.r(r),t.d(r,{OwlViTProcessor:()=>u});var n=t("./src/base/processing_utils.js"),o=t("./src/models/auto/image_processing_auto.js"),i=t("./src/tokenizers.js");class u extends n.Processor{}le(u,"tokenizer_class",i.AutoTokenizer),le(u,"image_processor_class",o.AutoImageProcessor)},"./src/models/paligemma/processing_paligemma.js":(e,r,t)=>{t.r(r),t.d(r,{PaliGemmaProcessor:()=>h});var n=t("./src/base/processing_utils.js"),o=t("./src/models/auto/image_processing_auto.js"),i=t("./src/tokenizers.js");const u="<image>";function d(f,y,m,g,v){return`${g.repeat(m*v)}${y}${f}
`}class h extends n.Processor{async _call(y,m=null,g={}){m||(console.warn("You are using PaliGemma without a text prefix. It will perform as a picture-captioning model."),m=""),Array.isArray(y)||(y=[y]),Array.isArray(m)||(m=[m]);const v=this.tokenizer.bos_token,b=this.image_processor.config.image_seq_length;let k;m.some(T=>T.includes(u))?k=m.map(T=>{const D=T.replaceAll(u,u.repeat(b)),I=D.lastIndexOf(u),$=I===-1?0:I+u.length;return D.slice(0,$)+v+D.slice($)+`
`}):(console.warn("You are passing both `text` and `images` to `PaliGemmaProcessor`. The processor expects special image tokens in the text, as many tokens as there are images per each text. It is recommended to add `<image>` tokens in the very beginning of your text. For this call, we will infer how many images each text has and add special tokens."),k=m.map(T=>d(T,v,b,u,y.length)));const z=this.tokenizer(k,g);return{...await this.image_processor(y,g),...z}}}le(h,"tokenizer_class",i.AutoTokenizer),le(h,"image_processor_class",o.AutoImageProcessor),le(h,"uses_processor_config",!1)},"./src/models/phi3_v/image_processing_phi3_v.js":(e,r,t)=>{t.r(r),t.d(r,{Phi3VImageProcessor:()=>y});var n=t("./src/base/image_processors_utils.js"),o=t("./src/utils/tensor.js");const i=336,u=[2,3],{ceil:d,floor:h,sqrt:f}=Math;class y extends n.ImageProcessor{constructor(g){super({...g,do_normalize:!0,do_pad:!0,pad_size:"custom",do_convert_rgb:!0,do_resize:!0}),this._num_crops=g.num_crops}calc_num_image_tokens_from_image_size(g,v){const{num_img_tokens:b}=this.config;return h((h(v/i)*h(g/i)+1)*b+1+(h(v/i)+1)*f(b))}get_resize_output_image_size(g,v){const b=this._num_crops,[k,z]=g.size;let E=k/z,T=1;for(;T*Math.ceil(T/E)<=b;)T+=1;T-=1;const D=Math.floor(T*336),I=Math.floor(D/E);return[D,I]}pad_image(g,v,b,k={}){const[z,E]=v,T=i*d(z/i),D=i*d(E/i),I=[1,1,1].map(($,P)=>($-this.image_mean[P])/this.image_std[P]);return super.pad_image(g,v,{width:D,height:T},{center:!0,constant_values:I,...k})}async _call(g,{num_crops:v=null}={}){if(this._num_crops=v??(v=this.config.num_crops),v<4||f(v)%1!==0)throw new Error("num_crops must be a square number >= 4");Array.isArray(g)||(g=[g]);const b=g.length,k=await Promise.all(g.map(S=>this.preprocess(S))),z=k.map(S=>S.original_size),E=k.map(S=>S.reshaped_input_size),T=[];for(const{pixel_values:S}of k){S.unsqueeze_(0);const[O,R]=S.dims.slice(-2),G=await(0,o.interpolate_4d)(S,{size:[i,i],mode:"bicubic"});if(v>0){const ee=[],ie=f(v),H=h(R/ie),ce=h(O/ie);for(let se=0;se<ie;++se)for(let _e=0;_e<ie;++_e){let ae,Ce,Te,q;se===ie-1?(Ce=O-ce,q=O):(Ce=se*ce,q=(se+1)*ce),_e===ie-1?(ae=R-H,Te=R):(ae=_e*H,Te=(_e+1)*H);const B=[Ce,ae],Q=[q,Te],oe=await(0,o.slice)(S,B,Q,u);ee.push(oe)}const re=await(0,o.interpolate_4d)((0,o.cat)(ee,0),{size:[i,i],mode:"bicubic"});T.push((0,o.cat)([G,re],0))}else T.push(G)}const D=(0,o.stack)(T,0),I=E.map(S=>S.map(O=>i*d(O/i))),$=new o.Tensor("int64",I.flat(),[b,2]),P=I.map(([S,O])=>this.calc_num_image_tokens_from_image_size(O,S));return{pixel_values:D,original_sizes:z,reshaped_input_sizes:E,image_sizes:$,num_img_tokens:P}}}},"./src/models/phi3_v/processing_phi3_v.js":(e,r,t)=>{t.r(r),t.d(r,{Phi3VProcessor:()=>h});var n=t("./src/base/processing_utils.js"),o=t("./src/models/auto/image_processing_auto.js"),i=t("./src/tokenizers.js");t("./src/utils/image.js");const u="<|image|>",d=/<\|image_\d+\|>/g;class h extends n.Processor{async _call(y,m=null,{padding:g=!0,truncation:v=!0,num_crops:b=null}={}){Array.isArray(y)||(y=[y]);let k,z;if(m){z=await this.image_processor(m,{num_crops:b});const{num_img_tokens:E}=z,T=y.map((I,$)=>I.split(d).join(u.repeat(E[$])));k=this.tokenizer(T,{padding:g,truncation:v});const D=this.tokenizer.model.convert_tokens_to_ids([u])[0];k.input_ids.map_(I=>I==D?-I:I)}else k=this.tokenizer(y);return{...k,...z}}}le(h,"image_processor_class",o.AutoImageProcessor),le(h,"tokenizer_class",i.AutoTokenizer)},"./src/models/processors.js":(e,r,t)=>{t.r(r),t.d(r,{Florence2Processor:()=>n.Florence2Processor,GroundingDinoProcessor:()=>o.GroundingDinoProcessor,Idefics3Processor:()=>i.Idefics3Processor,JinaCLIPProcessor:()=>d.JinaCLIPProcessor,MgpstrProcessor:()=>h.MgpstrProcessor,MoonshineProcessor:()=>f.MoonshineProcessor,OwlViTProcessor:()=>y.OwlViTProcessor,PaliGemmaProcessor:()=>g.PaliGemmaProcessor,Phi3VProcessor:()=>m.Phi3VProcessor,PyAnnoteProcessor:()=>v.PyAnnoteProcessor,Qwen2VLProcessor:()=>b.Qwen2VLProcessor,SamProcessor:()=>k.SamProcessor,SmolVLMProcessor:()=>z.SmolVLMProcessor,SpeechT5Processor:()=>E.SpeechT5Processor,UltravoxProcessor:()=>T.UltravoxProcessor,VLChatProcessor:()=>u.VLChatProcessor,Wav2Vec2Processor:()=>D.Wav2Vec2Processor,Wav2Vec2ProcessorWithLM:()=>I.Wav2Vec2ProcessorWithLM,WhisperProcessor:()=>$.WhisperProcessor});var n=t("./src/models/florence2/processing_florence2.js"),o=t("./src/models/grounding_dino/processing_grounding_dino.js"),i=t("./src/models/idefics3/processing_idefics3.js"),u=t("./src/models/janus/processing_janus.js"),d=t("./src/models/jina_clip/processing_jina_clip.js"),h=t("./src/models/mgp_str/processing_mgp_str.js"),f=t("./src/models/moonshine/processing_moonshine.js"),y=t("./src/models/owlvit/processing_owlvit.js"),m=t("./src/models/phi3_v/processing_phi3_v.js"),g=t("./src/models/paligemma/processing_paligemma.js"),v=t("./src/models/pyannote/processing_pyannote.js"),b=t("./src/models/qwen2_vl/processing_qwen2_vl.js"),k=t("./src/models/sam/processing_sam.js"),z=t("./src/models/smolvlm/processing_smolvlm.js"),E=t("./src/models/speecht5/processing_speecht5.js"),T=t("./src/models/ultravox/processing_ultravox.js"),D=t("./src/models/wav2vec2/processing_wav2vec2.js"),I=t("./src/models/wav2vec2_with_lm/processing_wav2vec2_with_lm.js"),$=t("./src/models/whisper/processing_whisper.js")},"./src/models/pvt/image_processing_pvt.js":(e,r,t)=>{t.r(r),t.d(r,{PvtImageProcessor:()=>o});var n=t("./src/base/image_processors_utils.js");class o extends n.ImageProcessor{}},"./src/models/pyannote/feature_extraction_pyannote.js":(e,r,t)=>{t.r(r),t.d(r,{PyAnnoteFeatureExtractor:()=>u});var n=t("./src/base/feature_extraction_utils.js"),o=t("./src/utils/tensor.js"),i=t("./src/utils/maths.js");class u extends n.FeatureExtractor{async _call(h){(0,n.validate_audio_inputs)(h,"PyAnnoteFeatureExtractor"),h instanceof Float64Array&&(h=new Float32Array(h));const f=[1,1,h.length];return{input_values:new o.Tensor("float32",h,f)}}samples_to_frames(h){return(h-this.config.offset)/this.config.step}post_process_speaker_diarization(h,f){const y=f/this.samples_to_frames(f)/this.config.sampling_rate,m=[];for(const g of h.tolist()){const v=[];let b=-1;for(let k=0;k<g.length;++k){const z=(0,i.softmax)(g[k]),[E,T]=(0,i.max)(z),[D,I]=[k,k+1];T!==b?(b=T,v.push({id:T,start:D,end:I,score:E})):(v.at(-1).end=I,v.at(-1).score+=E)}m.push(v.map(({id:k,start:z,end:E,score:T})=>({id:k,start:z*y,end:E*y,confidence:T/(E-z)})))}return m}}},"./src/models/pyannote/processing_pyannote.js":(e,r,t)=>{t.r(r),t.d(r,{PyAnnoteProcessor:()=>i});var n=t("./src/base/processing_utils.js"),o=t("./src/models/pyannote/feature_extraction_pyannote.js");class i extends n.Processor{async _call(d){return await this.feature_extractor(d)}post_process_speaker_diarization(...d){return this.feature_extractor.post_process_speaker_diarization(...d)}get sampling_rate(){return this.feature_extractor.config.sampling_rate}}le(i,"feature_extractor_class",o.PyAnnoteFeatureExtractor)},"./src/models/qwen2_vl/image_processing_qwen2_vl.js":(e,r,t)=>{t.r(r),t.d(r,{Qwen2VLImageProcessor:()=>i});var n=t("./src/base/image_processors_utils.js"),o=t("./src/utils/tensor.js");class i extends n.ImageProcessor{async _call(d,...h){const{pixel_values:f,original_sizes:y,reshaped_input_sizes:m}=await super._call(d,...h);let g=f;const{temporal_patch_size:v,merge_size:b,patch_size:k}=this.config;g.dims[0]===1&&(g=(0,o.cat)(Array.from({length:v},()=>g),0));const z=g.dims[0]/v,E=g.dims[1],T=Math.floor(g.dims[2]/k),D=Math.floor(g.dims[3]/k),I=g.view(z,v,E,Math.floor(T/b),b,k,Math.floor(D/b),b,k).permute(0,3,6,4,7,2,1,5,8).view(z*T*D,E*v*k*k),$=new o.Tensor("int64",[z,T,D],[1,3]);return{pixel_values:I,image_grid_thw:$,original_sizes:y,reshaped_input_sizes:m}}}},"./src/models/qwen2_vl/processing_qwen2_vl.js":(e,r,t)=>{t.r(r),t.d(r,{Qwen2VLProcessor:()=>u});var n=t("./src/base/processing_utils.js"),o=t("./src/models/auto/image_processing_auto.js"),i=t("./src/tokenizers.js");t("./src/utils/image.js");class u extends n.Processor{async _call(h,f=null,...y){Array.isArray(h)||(h=[h]);let m,g;if(f&&(m=await this.image_processor(f),g=m.image_grid_thw),g){let b=this.image_processor.config.merge_size**2,k=0;const z=g.tolist();h=h.map(E=>{for(;E.includes("<|image_pad|>");){const T=Number(z[k++].reduce((D,I)=>D*I,1n));E=E.replace("<|image_pad|>","<|placeholder|>".repeat(Math.floor(T/b)))}return E.replaceAll("<|placeholder|>","<|image_pad|>")})}return{...this.tokenizer(h),...m}}}le(u,"image_processor_class",o.AutoImageProcessor),le(u,"tokenizer_class",i.AutoTokenizer)},"./src/models/rt_detr/image_processing_rt_detr.js":(e,r,t)=>{t.r(r),t.d(r,{RTDetrImageProcessor:()=>o});var n=t("./src/base/image_processors_utils.js");class o extends n.ImageProcessor{post_process_object_detection(...u){return(0,n.post_process_object_detection)(...u)}}},"./src/models/sam/image_processing_sam.js":(e,r,t)=>{t.r(r),t.d(r,{SamImageProcessor:()=>u});var n=t("./src/base/image_processors_utils.js"),o=t("./src/utils/core.js"),i=t("./src/utils/tensor.js");class u extends n.ImageProcessor{reshape_input_points(h,f,y,m=!1){h=structuredClone(h);let g=(0,o.calculateDimensions)(h);if(g.length===3)m||(g=[1,...g]),h=[h];else if(g.length!==4)throw Error("The input_points must be a 4D tensor of shape `batch_size`, `point_batch_size`, `nb_points_per_image`, `2`.");for(let v=0;v<h.length;++v){let b=f[v],k=y[v],z=[k[0]/b[0],k[1]/b[1]];for(let E=0;E<h[v].length;++E)for(let T=0;T<h[v][E].length;++T)for(let D=0;D<h[v][E][T].length;++D)h[v][E][T][D]*=z[D%2]}return new i.Tensor("float32",Float32Array.from(h.flat(1/0)),g)}add_input_labels(h,f){let y=(0,o.calculateDimensions)(h);if(y.length===2)y=[1,...y],h=[h];else if(y.length!==3)throw Error("The input_points must be a 4D tensor of shape `batch_size`, `point_batch_size`, `nb_points_per_image`, `2`.");if(y.some((m,g)=>m!==f.dims[g]))throw Error(`The first ${y.length} dimensions of 'input_points' and 'input_labels' must be the same.`);return new i.Tensor("int64",h.flat(1/0).map(BigInt),y)}async _call(h,{input_points:f=null,input_labels:y=null,input_boxes:m=null}={}){const g=await super._call(h);if(f&&(g.input_points=this.reshape_input_points(f,g.original_sizes,g.reshaped_input_sizes)),y){if(!g.input_points)throw Error("`input_points` must be provided if `input_labels` are provided.");g.input_labels=this.add_input_labels(y,g.input_points)}return m&&(g.input_boxes=this.reshape_input_points(m,g.original_sizes,g.reshaped_input_sizes,!0)),g}async post_process_masks(h,f,y,{mask_threshold:m=0,binarize:g=!0,pad_size:v=null}={}){const b=[];v=v??this.pad_size;const k=[v.height,v.width];for(let z=0;z<f.length;++z){const E=f[z],T=y[z];let D=await(0,i.interpolate_4d)(h[z],{mode:"bilinear",size:k});if(D=D.slice(null,null,[0,T[0]],[0,T[1]]),D=await(0,i.interpolate_4d)(D,{mode:"bilinear",size:E}),g){const I=D.data,$=new Uint8Array(I.length);for(let P=0;P<I.length;++P)I[P]>m&&($[P]=1);D=new i.Tensor("bool",$,D.dims)}b.push(D)}return b}generate_crop_boxes(h,f,{crop_n_layers:y=0,overlap_ratio:m=512/1500,points_per_crop:g=32,crop_n_points_downscale_factor:v=1}={}){}}},"./src/models/sam/processing_sam.js":(e,r,t)=>{t.r(r),t.d(r,{SamProcessor:()=>i});var n=t("./src/base/processing_utils.js"),o=t("./src/models/auto/image_processing_auto.js");class i extends n.Processor{async _call(...d){return await this.image_processor(...d)}post_process_masks(...d){return this.image_processor.post_process_masks(...d)}reshape_input_points(...d){return this.image_processor.reshape_input_points(...d)}}le(i,"image_processor_class",o.AutoImageProcessor)},"./src/models/seamless_m4t/feature_extraction_seamless_m4t.js":(e,r,t)=>{t.r(r),t.d(r,{SeamlessM4TFeatureExtractor:()=>u});var n=t("./src/base/feature_extraction_utils.js"),o=t("./src/utils/tensor.js"),i=t("./src/utils/audio.js");class u extends n.FeatureExtractor{constructor(h){super(h);const f=this.config.sampling_rate,y=(0,i.mel_filter_bank)(257,this.config.num_mel_bins,20,Math.floor(f/2),f,null,"kaldi",!0);this.mel_filters=y,this.window=(0,i.window_function)(400,"povey",{periodic:!1})}async _extract_fbank_features(h,f){return h=h.map(y=>y*32768),(0,i.spectrogram)(h,this.window,400,160,{fft_length:512,power:2,center:!1,preemphasis:.97,mel_filters:this.mel_filters,log_mel:"log",mel_floor:1192092955078125e-22,remove_dc_offset:!0,max_num_frames:f,transpose:!0})}async _call(h,{padding:f=!0,pad_to_multiple_of:y=2,do_normalize_per_mel_bins:m=!0,return_attention_mask:g=!0}={}){(0,n.validate_audio_inputs)(h,"SeamlessM4TFeatureExtractor");let v=await this._extract_fbank_features(h,this.config.max_length);if(m){const[$,P]=v.dims,S=v.data;for(let O=0;O<P;++O){let R=0;for(let H=0;H<$;++H)R+=S[H*P+O];const G=R/$;let ee=0;for(let H=0;H<$;++H)ee+=(S[H*P+O]-G)**2;ee/=$-1;const ie=Math.sqrt(ee+1e-7);for(let H=0;H<$;++H){const ce=H*P+O;S[ce]=(S[ce]-G)/ie}}}let b;if(f){const[$,P]=v.dims,S=v.data,O=$%y;if(O>0){const R=new Float32Array(P*($+O));R.set(S),R.fill(this.config.padding_value,S.length);const G=$+O;v=new o.Tensor(v.type,R,[G,P]),g&&(b=new o.Tensor("int64",new BigInt64Array(G),[1,G]),b.data.fill(1n,0,$))}}const[k,z]=v.dims,E=this.config.stride;if(k%E!==0)throw new Error(`The number of frames (${k}) must be a multiple of the stride (${E}).`);const D=v.view(1,Math.floor(k/E),z*E),I={input_features:D};if(g){const $=D.dims[1],P=new BigInt64Array($);if(b){const S=b.data;for(let O=1,R=0;O<k;O+=E,++R)P[R]=S[O]}else P.fill(1n);I.attention_mask=new o.Tensor("int64",P,[1,$])}return I}}},"./src/models/segformer/image_processing_segformer.js":(e,r,t)=>{t.r(r),t.d(r,{SegformerFeatureExtractor:()=>i,SegformerImageProcessor:()=>o});var n=t("./src/base/image_processors_utils.js");class o extends n.ImageProcessor{post_process_semantic_segmentation(...d){return(0,n.post_process_semantic_segmentation)(...d)}}class i extends o{}},"./src/models/siglip/image_processing_siglip.js":(e,r,t)=>{t.r(r),t.d(r,{SiglipImageProcessor:()=>o});var n=t("./src/base/image_processors_utils.js");class o extends n.ImageProcessor{}},"./src/models/smolvlm/image_processing_smolvlm.js":(e,r,t)=>{t.r(r),t.d(r,{SmolVLMImageProcessor:()=>n.Idefics3ImageProcessor});var n=t("./src/models/idefics3/image_processing_idefics3.js")},"./src/models/smolvlm/processing_smolvlm.js":(e,r,t)=>{t.r(r),t.d(r,{SmolVLMProcessor:()=>n.Idefics3Processor});var n=t("./src/models/idefics3/processing_idefics3.js")},"./src/models/snac/feature_extraction_snac.js":(e,r,t)=>{t.r(r),t.d(r,{SnacFeatureExtractor:()=>o});var n=t("./src/models/dac/feature_extraction_dac.js");class o extends n.DacFeatureExtractor{}},"./src/models/speecht5/feature_extraction_speecht5.js":(e,r,t)=>{t.r(r),t.d(r,{SpeechT5FeatureExtractor:()=>o});var n=t("./src/base/feature_extraction_utils.js");class o extends n.FeatureExtractor{}},"./src/models/speecht5/processing_speecht5.js":(e,r,t)=>{t.r(r),t.d(r,{SpeechT5Processor:()=>u});var n=t("./src/base/processing_utils.js"),o=t("./src/tokenizers.js"),i=t("./src/models/auto/feature_extraction_auto.js");class u extends n.Processor{async _call(h){return await this.feature_extractor(h)}}le(u,"tokenizer_class",o.AutoTokenizer),le(u,"feature_extractor_class",i.AutoFeatureExtractor)},"./src/models/swin2sr/image_processing_swin2sr.js":(e,r,t)=>{t.r(r),t.d(r,{Swin2SRImageProcessor:()=>o});var n=t("./src/base/image_processors_utils.js");class o extends n.ImageProcessor{pad_image(u,d,h,f={}){const[y,m,g]=d;return super.pad_image(u,d,{width:m+(h-m%h)%h,height:y+(h-y%h)%h},{mode:"symmetric",center:!1,constant_values:-1,...f})}}},"./src/models/ultravox/processing_ultravox.js":(e,r,t)=>{t.r(r),t.d(r,{UltravoxProcessor:()=>u});var n=t("./src/models/auto/feature_extraction_auto.js"),o=t("./src/tokenizers.js"),i=t("./src/base/processing_utils.js");class u extends i.Processor{async _call(h,f=null,y={}){if(Array.isArray(h))throw new Error("Batched inputs are not supported yet.");let m={};if(f){const v=f.length,{input_features:b}=await this.feature_extractor(f,{...y,max_length:v}),k=Math.round(v/this.config.encoder_ds_factor+1e-4),z=1+Math.ceil(k/this.config.stack_factor);m.audio_token_len=[z],m.audio_values=b;const E=this.config.audio_placeholder;if(!h.includes(E))throw new Error(`The input text does not contain the image token ${E}.`);h=h.replaceAll(E,E.repeat(z))}return{...this.tokenizer(h,{add_special_tokens:!1,...y}),...m}}}le(u,"tokenizer_class",o.AutoTokenizer),le(u,"feature_extractor_class",n.AutoFeatureExtractor),le(u,"uses_processor_config",!0)},"./src/models/vit/image_processing_vit.js":(e,r,t)=>{t.r(r),t.d(r,{ViTFeatureExtractor:()=>i,ViTImageProcessor:()=>o});var n=t("./src/base/image_processors_utils.js");class o extends n.ImageProcessor{}class i extends o{}},"./src/models/vitmatte/image_processing_vitmatte.js":(e,r,t)=>{t.r(r),t.d(r,{VitMatteImageProcessor:()=>i});var n=t("./src/base/image_processors_utils.js"),o=t("./src/utils/tensor.js");class i extends n.ImageProcessor{async _call(d,h){Array.isArray(d)||(d=[d]),Array.isArray(h)||(h=[h]);const f=await Promise.all(d.map(g=>this.preprocess(g))),y=await Promise.all(h.map(g=>this.preprocess(g,{do_normalize:!1,do_convert_rgb:!1,do_convert_grayscale:!0})));return{pixel_values:(0,o.stack)(f.map((g,v)=>(0,o.cat)([g.pixel_values,y[v].pixel_values],0)),0),original_sizes:f.map(g=>g.original_size),reshaped_input_sizes:f.map(g=>g.reshaped_input_size)}}}},"./src/models/vitpose/image_processing_vitpose.js":(e,r,t)=>{t.r(r),t.d(r,{VitPoseImageProcessor:()=>o});var n=t("./src/base/image_processors_utils.js");class o extends n.ImageProcessor{post_process_pose_estimation(u,d,{threshold:h=null}={}){const f=u.tolist(),[y,m,g,v]=u.dims,b=[];for(let k=0;k<y;++k){const z=f[k],E=d[k],T=[];for(let D=0;D<E.length;++D){const I=E[D],$=[],P=[],S=[],O=I.at(-2)/v,R=I.at(-1)/g;for(let G=0;G<z.length;++G){let[ee,ie]=[0,0],H=0,ce=-1/0;const re=z[G];for(let _e=0;_e<re.length;++_e){const ae=re[_e];for(let Ce=0;Ce<ae.length;++Ce){const Te=ae[Ce];H+=Te,ce=Math.max(ce,Te),ee+=(Ce+.5)*Te,ie+=_e*Te}}if(h!=null&&ce<h)continue;const se=[O*ee/H,R*ie/H];$.push(se),S.push(G),P.push(ce)}T.push({bbox:I,scores:P,labels:S,keypoints:$})}b.push(T)}return b}}},"./src/models/wav2vec2/feature_extraction_wav2vec2.js":(e,r,t)=>{t.r(r),t.d(r,{Wav2Vec2FeatureExtractor:()=>i});var n=t("./src/base/feature_extraction_utils.js"),o=t("./src/utils/tensor.js");class i extends n.FeatureExtractor{_zero_mean_unit_var_norm(d){const f=d.reduce((m,g)=>m+g,0)/d.length,y=d.reduce((m,g)=>m+(g-f)**2,0)/d.length;return d.map(m=>(m-f)/Math.sqrt(y+1e-7))}async _call(d){(0,n.validate_audio_inputs)(d,"Wav2Vec2FeatureExtractor"),d instanceof Float64Array&&(d=new Float32Array(d));let h=d;this.config.do_normalize&&(h=this._zero_mean_unit_var_norm(h));const f=[1,h.length];return{input_values:new o.Tensor("float32",h,f),attention_mask:new o.Tensor("int64",new BigInt64Array(h.length).fill(1n),f)}}}},"./src/models/wav2vec2/processing_wav2vec2.js":(e,r,t)=>{t.r(r),t.d(r,{Wav2Vec2Processor:()=>u});var n=t("./src/tokenizers.js"),o=t("./src/models/auto/feature_extraction_auto.js"),i=t("./src/base/processing_utils.js");class u extends i.Processor{async _call(h){return await this.feature_extractor(h)}}le(u,"tokenizer_class",n.AutoTokenizer),le(u,"feature_extractor_class",o.AutoFeatureExtractor)},"./src/models/wav2vec2_with_lm/processing_wav2vec2_with_lm.js":(e,r,t)=>{t.r(r),t.d(r,{Wav2Vec2ProcessorWithLM:()=>u});var n=t("./src/tokenizers.js"),o=t("./src/models/auto/feature_extraction_auto.js"),i=t("./src/base/processing_utils.js");class u extends i.Processor{async _call(h){return await this.feature_extractor(h)}}le(u,"tokenizer_class",n.AutoTokenizer),le(u,"feature_extractor_class",o.AutoFeatureExtractor)},"./src/models/wespeaker/feature_extraction_wespeaker.js":(e,r,t)=>{t.r(r),t.d(r,{WeSpeakerFeatureExtractor:()=>i});var n=t("./src/base/feature_extraction_utils.js");t("./src/utils/tensor.js");var o=t("./src/utils/audio.js");class i extends n.FeatureExtractor{constructor(d){super(d);const h=this.config.sampling_rate,f=(0,o.mel_filter_bank)(257,this.config.num_mel_bins,20,Math.floor(h/2),h,null,"kaldi",!0);this.mel_filters=f,this.window=(0,o.window_function)(400,"hamming",{periodic:!1}),this.min_num_frames=this.config.min_num_frames}async _extract_fbank_features(d){return d=d.map(h=>h*32768),(0,o.spectrogram)(d,this.window,400,160,{fft_length:512,power:2,center:!1,preemphasis:.97,mel_filters:this.mel_filters,log_mel:"log",mel_floor:1192092955078125e-22,remove_dc_offset:!0,transpose:!0,min_num_frames:this.min_num_frames})}async _call(d){(0,n.validate_audio_inputs)(d,"WeSpeakerFeatureExtractor");const h=(await this._extract_fbank_features(d)).unsqueeze_(0);if(this.config.fbank_centering_span===null){const f=h.mean(1).data,y=h.data,[m,g,v]=h.dims;for(let b=0;b<m;++b){const k=b*g*v,z=b*v;for(let E=0;E<g;++E){const T=k+E*v;for(let D=0;D<v;++D)y[T+D]-=f[z+D]}}}return{input_features:h}}}},"./src/models/whisper/common_whisper.js":(e,r,t)=>{t.r(r),t.d(r,{WHISPER_LANGUAGE_MAPPING:()=>o,WHISPER_TO_LANGUAGE_CODE_MAPPING:()=>i,whisper_language_to_code:()=>u});const n=[["en","english"],["zh","chinese"],["de","german"],["es","spanish"],["ru","russian"],["ko","korean"],["fr","french"],["ja","japanese"],["pt","portuguese"],["tr","turkish"],["pl","polish"],["ca","catalan"],["nl","dutch"],["ar","arabic"],["sv","swedish"],["it","italian"],["id","indonesian"],["hi","hindi"],["fi","finnish"],["vi","vietnamese"],["he","hebrew"],["uk","ukrainian"],["el","greek"],["ms","malay"],["cs","czech"],["ro","romanian"],["da","danish"],["hu","hungarian"],["ta","tamil"],["no","norwegian"],["th","thai"],["ur","urdu"],["hr","croatian"],["bg","bulgarian"],["lt","lithuanian"],["la","latin"],["mi","maori"],["ml","malayalam"],["cy","welsh"],["sk","slovak"],["te","telugu"],["fa","persian"],["lv","latvian"],["bn","bengali"],["sr","serbian"],["az","azerbaijani"],["sl","slovenian"],["kn","kannada"],["et","estonian"],["mk","macedonian"],["br","breton"],["eu","basque"],["is","icelandic"],["hy","armenian"],["ne","nepali"],["mn","mongolian"],["bs","bosnian"],["kk","kazakh"],["sq","albanian"],["sw","swahili"],["gl","galician"],["mr","marathi"],["pa","punjabi"],["si","sinhala"],["km","khmer"],["sn","shona"],["yo","yoruba"],["so","somali"],["af","afrikaans"],["oc","occitan"],["ka","georgian"],["be","belarusian"],["tg","tajik"],["sd","sindhi"],["gu","gujarati"],["am","amharic"],["yi","yiddish"],["lo","lao"],["uz","uzbek"],["fo","faroese"],["ht","haitian creole"],["ps","pashto"],["tk","turkmen"],["nn","nynorsk"],["mt","maltese"],["sa","sanskrit"],["lb","luxembourgish"],["my","myanmar"],["bo","tibetan"],["tl","tagalog"],["mg","malagasy"],["as","assamese"],["tt","tatar"],["haw","hawaiian"],["ln","lingala"],["ha","hausa"],["ba","bashkir"],["jw","javanese"],["su","sundanese"]],o=new Map(n),i=new Map([...n.map(([d,h])=>[h,d]),["burmese","my"],["valencian","ca"],["flemish","nl"],["haitian","ht"],["letzeburgesch","lb"],["pushto","ps"],["panjabi","pa"],["moldavian","ro"],["moldovan","ro"],["sinhalese","si"],["castilian","es"]]);function u(d){d=d.toLowerCase();let h=i.get(d);if(h===void 0){const f=d.match(/^<\|([a-z]{2})\|>$/);if(f&&(d=f[1]),o.has(d))h=d;else{const m=d.length===2?o.keys():o.values();throw new Error(`Language "${d}" is not supported. Must be one of: ${JSON.stringify(Array.from(m))}`)}}return h}},"./src/models/whisper/feature_extraction_whisper.js":(e,r,t)=>{t.r(r),t.d(r,{WhisperFeatureExtractor:()=>u});var n=t("./src/base/feature_extraction_utils.js");t("./src/utils/tensor.js");var o=t("./src/utils/audio.js"),i=t("./src/utils/maths.js");class u extends n.FeatureExtractor{constructor(h){var f;super(h),(f=this.config).mel_filters??(f.mel_filters=(0,o.mel_filter_bank)(Math.floor(1+this.config.n_fft/2),this.config.feature_size,0,8e3,this.config.sampling_rate,"slaney","slaney")),this.window=(0,o.window_function)(this.config.n_fft,"hann")}async _extract_fbank_features(h){const f=await(0,o.spectrogram)(h,this.window,this.config.n_fft,this.config.hop_length,{power:2,mel_filters:this.config.mel_filters,log_mel:"log10",max_num_frames:Math.min(Math.floor(h.length/this.config.hop_length),this.config.nb_max_frames)}),y=f.data,m=(0,i.max)(y)[0];for(let g=0;g<y.length;++g)y[g]=(Math.max(y[g],m-8)+4)/4;return f}async _call(h,{max_length:f=null}={}){(0,n.validate_audio_inputs)(h,"WhisperFeatureExtractor");let y;const m=f??this.config.n_samples;return h.length>m?(h.length>this.config.n_samples&&console.warn("Attempting to extract features for audio longer than 30 seconds. If using a pipeline to extract transcript from a long audio clip, remember to specify `chunk_length_s` and/or `stride_length_s`."),y=h.slice(0,m)):(y=new Float32Array(m),y.set(h)),{input_features:(await this._extract_fbank_features(y)).unsqueeze_(0)}}}},"./src/models/whisper/generation_whisper.js":(e,r,t)=>{t.r(r),t.d(r,{WhisperGenerationConfig:()=>o});var n=t("./src/generation/configuration_utils.js");class o extends n.GenerationConfig{constructor(){super(...arguments);le(this,"return_timestamps",null);le(this,"return_token_timestamps",null);le(this,"num_frames",null);le(this,"alignment_heads",null);le(this,"task",null);le(this,"language",null);le(this,"no_timestamps_token_id",null);le(this,"prompt_ids",null);le(this,"is_multilingual",null);le(this,"lang_to_id",null);le(this,"task_to_id",null);le(this,"max_initial_timestamp_index",1)}}},"./src/models/whisper/processing_whisper.js":(e,r,t)=>{t.r(r),t.d(r,{WhisperProcessor:()=>u});var n=t("./src/models/auto/feature_extraction_auto.js"),o=t("./src/tokenizers.js"),i=t("./src/base/processing_utils.js");class u extends i.Processor{async _call(h){return await this.feature_extractor(h)}}le(u,"tokenizer_class",o.AutoTokenizer),le(u,"feature_extractor_class",n.AutoFeatureExtractor)},"./src/models/yolos/image_processing_yolos.js":(e,r,t)=>{t.r(r),t.d(r,{YolosFeatureExtractor:()=>i,YolosImageProcessor:()=>o});var n=t("./src/base/image_processors_utils.js");class o extends n.ImageProcessor{post_process_object_detection(...d){return(0,n.post_process_object_detection)(...d)}}class i extends o{}},"./src/ops/registry.js":(e,r,t)=>{t.r(r),t.d(r,{TensorOpRegistry:()=>h});var n=t("./src/backends/onnx.js"),o=t("./src/utils/tensor.js"),i=t("./src/env.js");const u=i.apis.IS_BROWSER_ENV||i.apis.IS_WEBWORKER_ENV,d=async(f,y,m)=>{const g=await(0,n.createInferenceSession)(new Uint8Array(f),y);let v=Promise.resolve();return async b=>{const k=(0,n.isONNXProxy)(),z=Object.fromEntries(Object.entries(b).map(([T,D])=>[T,(k?D.clone():D).ort_tensor])),E=await(v=u?v.then(()=>g.run(z)):g.run(z));return Array.isArray(m)?m.map(T=>new o.Tensor(E[T])):new o.Tensor(E[m])}};class h{static get nearest_interpolate_4d(){return this._nearest_interpolate_4d||(this._nearest_interpolate_4d=d([8,10,18,0,58,129,1,10,41,10,1,120,10,0,10,0,10,1,115,18,1,121,34,6,82,101,115,105,122,101,42,18,10,4,109,111,100,101,34,7,110,101,97,114,101,115,116,160,1,3,18,1,114,90,31,10,1,120,18,26,10,24,8,1,18,20,10,3,18,1,98,10,3,18,1,99,10,3,18,1,104,10,3,18,1,119,90,15,10,1,115,18,10,10,8,8,7,18,4,10,2,8,4,98,31,10,1,121,18,26,10,24,8,1,18,20,10,3,18,1,98,10,3,18,1,99,10,3,18,1,104,10,3,18,1,119,66,2,16,21],this.session_options,"y")),this._nearest_interpolate_4d}static get bilinear_interpolate_4d(){return this._bilinear_interpolate_4d||(this._bilinear_interpolate_4d=d([8,9,18,0,58,128,1,10,40,10,1,120,10,0,10,0,10,1,115,18,1,121,34,6,82,101,115,105,122,101,42,17,10,4,109,111,100,101,34,6,108,105,110,101,97,114,160,1,3,18,1,114,90,31,10,1,120,18,26,10,24,8,1,18,20,10,3,18,1,98,10,3,18,1,99,10,3,18,1,104,10,3,18,1,119,90,15,10,1,115,18,10,10,8,8,7,18,4,10,2,8,4,98,31,10,1,121,18,26,10,24,8,1,18,20,10,3,18,1,98,10,3,18,1,99,10,3,18,1,104,10,3,18,1,119,66,2,16,20],this.session_options,"y")),this._bilinear_interpolate_4d}static get bicubic_interpolate_4d(){return this._bicubic_interpolate_4d||(this._bicubic_interpolate_4d=d([8,9,18,0,58,127,10,39,10,1,120,10,0,10,0,10,1,115,18,1,121,34,6,82,101,115,105,122,101,42,16,10,4,109,111,100,101,34,5,99,117,98,105,99,160,1,3,18,1,114,90,31,10,1,120,18,26,10,24,8,1,18,20,10,3,18,1,98,10,3,18,1,99,10,3,18,1,104,10,3,18,1,119,90,15,10,1,115,18,10,10,8,8,7,18,4,10,2,8,4,98,31,10,1,121,18,26,10,24,8,1,18,20,10,3,18,1,98,10,3,18,1,99,10,3,18,1,104,10,3,18,1,119,66,2,16,20],this.session_options,"y")),this._bicubic_interpolate_4d}static get matmul(){return this._matmul||(this._matmul=d([8,9,18,0,58,55,10,17,10,1,97,10,1,98,18,1,99,34,6,77,97,116,77,117,108,18,1,114,90,9,10,1,97,18,4,10,2,8,1,90,9,10,1,98,18,4,10,2,8,1,98,9,10,1,99,18,4,10,2,8,1,66,2,16,20],this.session_options,"c")),this._matmul}static get stft(){return this._stft||(this._stft=d([8,7,18,0,58,148,1,10,38,10,1,115,10,1,106,10,1,119,10,1,108,18,1,111,34,4,83,84,70,84,42,15,10,8,111,110,101,115,105,100,101,100,24,1,160,1,2,18,1,115,90,26,10,1,115,18,21,10,19,8,1,18,15,10,3,18,1,98,10,3,18,1,115,10,3,18,1,99,90,11,10,1,106,18,6,10,4,8,7,18,0,90,16,10,1,119,18,11,10,9,8,1,18,5,10,3,18,1,119,90,11,10,1,108,18,6,10,4,8,7,18,0,98,31,10,1,111,18,26,10,24,8,1,18,20,10,3,18,1,98,10,3,18,1,102,10,3,18,1,100,10,3,18,1,99,66,2,16,17],this.session_options,"o")),this._stft}static get rfft(){return this._rfft||(this._rfft=d([8,9,18,0,58,97,10,33,10,1,120,10,0,10,1,97,18,1,121,34,3,68,70,84,42,15,10,8,111,110,101,115,105,100,101,100,24,1,160,1,2,18,1,100,90,21,10,1,120,18,16,10,14,8,1,18,10,10,3,18,1,115,10,3,18,1,99,90,11,10,1,97,18,6,10,4,8,7,18,0,98,21,10,1,121,18,16,10,14,8,1,18,10,10,3,18,1,115,10,3,18,1,99,66,2,16,20],this.session_options,"y")),this._rfft}static get top_k(){return this._top_k||(this._top_k=d([8,10,18,0,58,73,10,18,10,1,120,10,1,107,18,1,118,18,1,105,34,4,84,111,112,75,18,1,116,90,9,10,1,120,18,4,10,2,8,1,90,15,10,1,107,18,10,10,8,8,7,18,4,10,2,8,1,98,9,10,1,118,18,4,10,2,8,1,98,9,10,1,105,18,4,10,2,8,7,66,2,16,21],this.session_options,["v","i"])),this._top_k}static get slice(){return this._slice||(this._slice=d([8,7,18,0,58,96,10,25,10,1,120,10,1,115,10,1,101,10,1,97,10,1,116,18,1,121,34,5,83,108,105,99,101,18,1,114,90,9,10,1,120,18,4,10,2,8,1,90,9,10,1,115,18,4,10,2,8,7,90,9,10,1,101,18,4,10,2,8,7,90,9,10,1,97,18,4,10,2,8,7,90,9,10,1,116,18,4,10,2,8,7,98,9,10,1,121,18,4,10,2,8,1,66,2,16,13],this.session_options,"y")),this._slice}}le(h,"session_options",{})},"./src/pipelines.js":(e,r,t)=>{t.r(r),t.d(r,{AudioClassificationPipeline:()=>ie,AutomaticSpeechRecognitionPipeline:()=>ce,BackgroundRemovalPipeline:()=>ae,DepthEstimationPipeline:()=>Ie,DocumentQuestionAnsweringPipeline:()=>B,FeatureExtractionPipeline:()=>G,FillMaskPipeline:()=>D,ImageClassificationPipeline:()=>se,ImageFeatureExtractionPipeline:()=>ee,ImageSegmentationPipeline:()=>_e,ImageToImagePipeline:()=>oe,ImageToTextPipeline:()=>re,ObjectDetectionPipeline:()=>Te,Pipeline:()=>k,QuestionAnsweringPipeline:()=>T,SummarizationPipeline:()=>$,Text2TextGenerationPipeline:()=>I,TextClassificationPipeline:()=>z,TextGenerationPipeline:()=>O,TextToAudioPipeline:()=>Q,TokenClassificationPipeline:()=>E,TranslationPipeline:()=>P,ZeroShotAudioClassificationPipeline:()=>H,ZeroShotClassificationPipeline:()=>R,ZeroShotImageClassificationPipeline:()=>Ce,ZeroShotObjectDetectionPipeline:()=>q,pipeline:()=>et});var n=t("./src/tokenizers.js"),o=t("./src/models.js"),i=t("./src/models/auto/processing_auto.js");t("./src/base/processing_utils.js");var u=t("./src/utils/generic.js"),d=t("./src/utils/core.js"),h=t("./src/utils/maths.js"),f=t("./src/utils/audio.js"),y=t("./src/utils/tensor.js"),m=t("./src/utils/image.js");async function g(Ee){return Array.isArray(Ee)||(Ee=[Ee]),await Promise.all(Ee.map(Y=>m.RawImage.read(Y)))}async function v(Ee,Y){return Array.isArray(Ee)||(Ee=[Ee]),await Promise.all(Ee.map(pe=>typeof pe=="string"||pe instanceof URL?(0,f.read_audio)(pe,Y):pe instanceof Float64Array?new Float32Array(pe):pe))}function b(Ee,Y){Y&&(Ee=Ee.map(Ne=>Ne|0));const[pe,X,Ae,Oe]=Ee;return{xmin:pe,ymin:X,xmax:Ae,ymax:Oe}}class k extends u.Callable{constructor({task:Y,model:pe,tokenizer:X=null,processor:Ae=null}){super(),this.task=Y,this.model=pe,this.tokenizer=X,this.processor=Ae}async dispose(){await this.model.dispose()}}class z extends k{constructor(Y){super(Y)}async _call(Y,{top_k:pe=1}={}){const X=this.tokenizer(Y,{padding:!0,truncation:!0}),Ae=await this.model(X),Oe=this.model.config.problem_type==="multi_label_classification"?Ve=>Ve.sigmoid():Ve=>new y.Tensor("float32",(0,h.softmax)(Ve.data),Ve.dims),Ne=this.model.config.id2label,Be=[];for(const Ve of Ae.logits){const He=Oe(Ve),Xe=await(0,y.topk)(He,pe),st=Xe[0].tolist(),We=Xe[1].tolist().map((rt,pt)=>({label:Ne?Ne[rt]:`LABEL_${rt}`,score:st[pt]}));pe===1?Be.push(...We):Be.push(We)}return Array.isArray(Y)||pe===1?Be:Be[0]}}class E extends k{constructor(Y){super(Y)}async _call(Y,{ignore_labels:pe=["O"]}={}){const X=Array.isArray(Y),Ae=this.tokenizer(X?Y:[Y],{padding:!0,truncation:!0}),Ne=(await this.model(Ae)).logits,Be=this.model.config.id2label,Ve=[];for(let He=0;He<Ne.dims[0];++He){const Xe=Ae.input_ids[He],st=Ne[He],nt=[];for(let We=0;We<st.dims[0];++We){const rt=st[We],pt=(0,h.max)(rt.data)[1],_t=Be?Be[pt]:`LABEL_${pt}`;if(pe.includes(_t))continue;const ct=this.tokenizer.decode([Xe[We].item()],{skip_special_tokens:!0});if(ct==="")continue;const dt=(0,h.softmax)(rt.data);nt.push({entity:_t,score:dt[pt],index:We,word:ct})}Ve.push(nt)}return X?Ve:Ve[0]}}class T extends k{constructor(Y){super(Y)}async _call(Y,pe,{top_k:X=1}={}){const Ae=this.tokenizer(Y,{text_pair:pe,padding:!0,truncation:!0}),{start_logits:Oe,end_logits:Ne}=await this.model(Ae),Be=Ae.input_ids.tolist(),Ve=Ae.attention_mask.tolist(),He=this.tokenizer.all_special_ids,Xe=[];for(let st=0;st<Oe.dims[0];++st){const nt=Be[st],We=nt.findIndex(Le=>Le==this.tokenizer.sep_token_id);Ve[st].map((Le,xt)=>Le==1&&(xt===0||xt>We&&He.findIndex(Nt=>Nt==nt[xt])===-1));const rt=Oe[st].tolist(),pt=Ne[st].tolist();for(let Le=1;Le<rt.length;++Le)(Ve[st]==0||Le<=We||He.findIndex(xt=>xt==nt[Le])!==-1)&&(rt[Le]=-1/0,pt[Le]=-1/0);const _t=(0,h.softmax)(rt).map((Le,xt)=>[Le,xt]),ct=(0,h.softmax)(pt).map((Le,xt)=>[Le,xt]);_t[0][0]=0,ct[0][0]=0;const dt=(0,d.product)(_t,ct).filter(Le=>Le[0][1]<=Le[1][1]).map(Le=>[Le[0][1],Le[1][1],Le[0][0]*Le[1][0]]).sort((Le,xt)=>xt[2]-Le[2]);for(let Le=0;Le<Math.min(dt.length,X);++Le){const[xt,Nt,cr]=dt[Le],Kt=nt.slice(xt,Nt+1),wr=this.tokenizer.decode(Kt,{skip_special_tokens:!0});Xe.push({answer:wr,score:cr})}}return X===1?Xe[0]:Xe}}class D extends k{constructor(Y){super(Y)}async _call(Y,{top_k:pe=5}={}){const X=this.tokenizer(Y,{padding:!0,truncation:!0}),{logits:Ae}=await this.model(X),Oe=[],Ne=X.input_ids.tolist();for(let Be=0;Be<Ne.length;++Be){const Ve=Ne[Be],He=Ve.findIndex(rt=>rt==this.tokenizer.mask_token_id);if(He===-1)throw Error(`Mask token (${this.tokenizer.mask_token}) not found in text.`);const Xe=Ae[Be][He],st=await(0,y.topk)(new y.Tensor("float32",(0,h.softmax)(Xe.data),Xe.dims),pe),nt=st[0].tolist(),We=st[1].tolist();Oe.push(We.map((rt,pt)=>{const _t=Ve.slice();return _t[He]=rt,{score:nt[pt],token:Number(rt),token_str:this.tokenizer.decode([rt]),sequence:this.tokenizer.decode(_t,{skip_special_tokens:!0})}}))}return Array.isArray(Y)?Oe:Oe[0]}}class I extends k{constructor(pe){super(pe);le(this,"_key","generated_text")}async _call(pe,X={}){Array.isArray(pe)||(pe=[pe]),this.model.config.prefix&&(pe=pe.map(He=>this.model.config.prefix+He));const Ae=this.model.config.task_specific_params;Ae&&Ae[this.task]&&Ae[this.task].prefix&&(pe=pe.map(He=>Ae[this.task].prefix+He));const Oe=this.tokenizer,Ne={padding:!0,truncation:!0};let Be;this instanceof P&&"_build_translation_inputs"in Oe?Be=Oe._build_translation_inputs(pe,Ne,X):Be=Oe(pe,Ne);const Ve=await this.model.generate({...Be,...X});return Oe.batch_decode(Ve,{skip_special_tokens:!0}).map(He=>({[this._key]:He}))}}class $ extends I{constructor(pe){super(pe);le(this,"_key","summary_text")}}class P extends I{constructor(pe){super(pe);le(this,"_key","translation_text")}}function S(Ee){return Array.isArray(Ee)&&Ee.every(Y=>"role"in Y&&"content"in Y)}class O extends k{constructor(Y){super(Y)}async _call(Y,pe={}){let X=!1,Ae=!1,Oe;if(typeof Y=="string")Oe=Y=[Y];else if(Array.isArray(Y)&&Y.every(We=>typeof We=="string"))X=!0,Oe=Y;else{if(S(Y))Y=[Y];else if(Array.isArray(Y)&&Y.every(S))X=!0;else throw new Error("Input must be a string, an array of strings, a Chat, or an array of Chats");Ae=!0,Oe=Y.map(We=>this.tokenizer.apply_chat_template(We,{tokenize:!1,add_generation_prompt:!0}))}const Ne=pe.add_special_tokens??!1,Be=Ae?!1:pe.return_full_text??!0;this.tokenizer.padding_side="left";const Ve=this.tokenizer(Oe,{add_special_tokens:Ne,padding:!0,truncation:!0}),He=await this.model.generate({...Ve,...pe}),Xe=this.tokenizer.batch_decode(He,{skip_special_tokens:!0});let st;!Be&&Ve.input_ids.dims.at(-1)>0&&(st=this.tokenizer.batch_decode(Ve.input_ids,{skip_special_tokens:!0}).map(We=>We.length));const nt=Array.from({length:Y.length},We=>[]);for(let We=0;We<Xe.length;++We){const rt=Math.floor(We/He.dims[0]*Y.length);st&&(Xe[We]=Xe[We].slice(st[rt])),nt[rt].push({generated_text:Ae?[...Y[rt],{role:"assistant",content:Xe[We]}]:Xe[We]})}return!X&&nt.length===1?nt[0]:nt}}class R extends k{constructor(Y){super(Y),this.label2id=Object.fromEntries(Object.entries(this.model.config.label2id).map(([pe,X])=>[pe.toLowerCase(),X])),this.entailment_id=this.label2id.entailment,this.entailment_id===void 0&&(console.warn("Could not find 'entailment' in label2id mapping. Using 2 as entailment_id."),this.entailment_id=2),this.contradiction_id=this.label2id.contradiction??this.label2id.not_entailment,this.contradiction_id===void 0&&(console.warn("Could not find 'contradiction' in label2id mapping. Using 0 as contradiction_id."),this.contradiction_id=0)}async _call(Y,pe,{hypothesis_template:X="This example is {}.",multi_label:Ae=!1}={}){const Oe=Array.isArray(Y);Oe||(Y=[Y]),Array.isArray(pe)||(pe=[pe]);const Ne=pe.map(He=>X.replace("{}",He)),Be=Ae||pe.length===1,Ve=[];for(const He of Y){const Xe=[];for(const We of Ne){const rt=this.tokenizer(He,{text_pair:We,padding:!0,truncation:!0}),pt=await this.model(rt);Be?Xe.push([pt.logits.data[this.contradiction_id],pt.logits.data[this.entailment_id]]):Xe.push(pt.logits.data[this.entailment_id])}const nt=(Be?Xe.map(We=>(0,h.softmax)(We)[1]):(0,h.softmax)(Xe)).map((We,rt)=>[We,rt]).sort((We,rt)=>rt[0]-We[0]);Ve.push({sequence:He,labels:nt.map(We=>pe[We[1]]),scores:nt.map(We=>We[0])})}return Oe?Ve:Ve[0]}}class G extends k{constructor(Y){super(Y)}async _call(Y,{pooling:pe="none",normalize:X=!1,quantize:Ae=!1,precision:Oe="binary"}={}){const Ne=this.tokenizer(Y,{padding:!0,truncation:!0}),Be=await this.model(Ne);let Ve=Be.last_hidden_state??Be.logits??Be.token_embeddings;if(pe!=="none")if(pe==="mean")Ve=(0,y.mean_pooling)(Ve,Ne.attention_mask);else if(pe==="cls")Ve=Ve.slice(null,0);else throw Error(`Pooling method '${pe}' not supported.`);return X&&(Ve=Ve.normalize(2,-1)),Ae&&(Ve=(0,y.quantize_embeddings)(Ve,Oe)),Ve}}class ee extends k{constructor(Y){super(Y)}async _call(Y,{pool:pe=null}={}){const X=await g(Y),{pixel_values:Ae}=await this.processor(X),Oe=await this.model({pixel_values:Ae});let Ne;if(pe){if(!("pooler_output"in Oe))throw Error("No pooled output was returned. Make sure the model has a 'pooler' layer when using the 'pool' option.");Ne=Oe.pooler_output}else Ne=Oe.last_hidden_state??Oe.logits??Oe.image_embeds;return Ne}}class ie extends k{constructor(Y){super(Y)}async _call(Y,{top_k:pe=5}={}){const X=this.processor.feature_extractor.config.sampling_rate,Ae=await v(Y,X),Oe=this.model.config.id2label,Ne=[];for(const Be of Ae){const Ve=await this.processor(Be),Xe=(await this.model(Ve)).logits[0],st=await(0,y.topk)(new y.Tensor("float32",(0,h.softmax)(Xe.data),Xe.dims),pe),nt=st[0].tolist(),rt=st[1].tolist().map((pt,_t)=>({label:Oe?Oe[pt]:`LABEL_${pt}`,score:nt[_t]}));Ne.push(rt)}return Array.isArray(Y)?Ne:Ne[0]}}class H extends k{constructor(Y){super(Y)}async _call(Y,pe,{hypothesis_template:X="This is a sound of {}."}={}){const Ae=!Array.isArray(Y);Ae&&(Y=[Y]);const Oe=pe.map(Xe=>X.replace("{}",Xe)),Ne=this.tokenizer(Oe,{padding:!0,truncation:!0}),Be=this.processor.feature_extractor.config.sampling_rate,Ve=await v(Y,Be),He=[];for(const Xe of Ve){const st=await this.processor(Xe),nt=await this.model({...Ne,...st}),We=(0,h.softmax)(nt.logits_per_audio.data);He.push([...We].map((rt,pt)=>({score:rt,label:pe[pt]})))}return Ae?He[0]:He}}class ce extends k{constructor(Y){super(Y)}async _call(Y,pe={}){switch(this.model.config.model_type){case"whisper":case"lite-whisper":return this._call_whisper(Y,pe);case"wav2vec2":case"wav2vec2-bert":case"unispeech":case"unispeech-sat":case"hubert":return this._call_wav2vec2(Y,pe);case"moonshine":return this._call_moonshine(Y,pe);default:throw new Error(`AutomaticSpeechRecognitionPipeline does not support model type '${this.model.config.model_type}'.`)}}async _call_wav2vec2(Y,pe){pe.language&&console.warn('`language` parameter is not yet supported for `wav2vec2` models, defaulting to "English".'),pe.task&&console.warn('`task` parameter is not yet supported for `wav2vec2` models, defaulting to "transcribe".');const X=!Array.isArray(Y);X&&(Y=[Y]);const Ae=this.processor.feature_extractor.config.sampling_rate,Oe=await v(Y,Ae),Ne=[];for(const Be of Oe){const Ve=await this.processor(Be),Xe=(await this.model(Ve)).logits[0],st=[];for(const We of Xe)st.push((0,h.max)(We.data)[1]);const nt=this.tokenizer.decode(st);Ne.push({text:nt})}return X?Ne[0]:Ne}async _call_whisper(Y,pe){const X=pe.return_timestamps??!1,Ae=pe.chunk_length_s??0,Oe=pe.force_full_sequences??!1;let Ne=pe.stride_length_s??null;const Be={...pe};X==="word"&&(Be.return_token_timestamps=!0,Be.return_timestamps=!1);const Ve=!Array.isArray(Y);Ve&&(Y=[Y]);const He=this.processor.feature_extractor.config.chunk_length/this.model.config.max_source_positions,Xe=this.processor.feature_extractor.config.hop_length,st=this.processor.feature_extractor.config.sampling_rate,nt=await v(Y,st),We=[];for(const rt of nt){let pt=[];if(Ae>0){if(Ne===null)Ne=Ae/6;else if(Ae<=Ne)throw Error("`chunk_length_s` must be larger than `stride_length_s`.");const dt=st*Ae,Le=st*Ne,xt=dt-2*Le;let Nt=0;for(;;){const cr=Nt+dt,Kt=rt.subarray(Nt,cr),wr=await this.processor(Kt),zr=Nt===0,$t=cr>=rt.length;if(pt.push({stride:[Kt.length,zr?0:Le,$t?0:Le],input_features:wr.input_features,is_last:$t}),$t)break;Nt+=xt}}else pt=[{stride:[rt.length,0,0],input_features:(await this.processor(rt)).input_features,is_last:!0}];for(const dt of pt){Be.num_frames=Math.floor(dt.stride[0]/Xe);const Le=await this.model.generate({inputs:dt.input_features,...Be});X==="word"?(dt.tokens=Le.sequences.tolist()[0],dt.token_timestamps=Le.token_timestamps.tolist()[0].map(xt=>(0,h.round)(xt,2))):dt.tokens=Le[0].tolist(),dt.stride=dt.stride.map(xt=>xt/st)}const[_t,ct]=this.tokenizer._decode_asr(pt,{time_precision:He,return_timestamps:X,force_full_sequences:Oe});We.push({text:_t,...ct})}return Ve?We[0]:We}async _call_moonshine(Y,pe){const X=!Array.isArray(Y);X&&(Y=[Y]);const Ae=this.processor.feature_extractor.config.sampling_rate,Oe=await v(Y,Ae),Ne=[];for(const Be of Oe){const Ve=await this.processor(Be),He=Math.floor(Be.length/Ae)*6,Xe=await this.model.generate({max_new_tokens:He,...pe,...Ve}),st=this.processor.batch_decode(Xe,{skip_special_tokens:!0})[0];Ne.push({text:st})}return X?Ne[0]:Ne}}class re extends k{constructor(Y){super(Y)}async _call(Y,pe={}){const X=Array.isArray(Y),Ae=await g(Y),{pixel_values:Oe}=await this.processor(Ae),Ne=[];for(const Be of Oe){Be.dims=[1,...Be.dims];const Ve=await this.model.generate({inputs:Be,...pe}),He=this.tokenizer.batch_decode(Ve,{skip_special_tokens:!0}).map(Xe=>({generated_text:Xe.trim()}));Ne.push(He)}return X?Ne:Ne[0]}}class se extends k{constructor(Y){super(Y)}async _call(Y,{top_k:pe=5}={}){const X=await g(Y),{pixel_values:Ae}=await this.processor(X),Oe=await this.model({pixel_values:Ae}),Ne=this.model.config.id2label,Be=[];for(const Ve of Oe.logits){const He=await(0,y.topk)(new y.Tensor("float32",(0,h.softmax)(Ve.data),Ve.dims),pe),Xe=He[0].tolist(),nt=He[1].tolist().map((We,rt)=>({label:Ne?Ne[We]:`LABEL_${We}`,score:Xe[rt]}));Be.push(nt)}return Array.isArray(Y)?Be:Be[0]}}class _e extends k{constructor(Y){super(Y),this.subtasks_mapping={panoptic:"post_process_panoptic_segmentation",instance:"post_process_instance_segmentation",semantic:"post_process_semantic_segmentation"}}async _call(Y,{threshold:pe=.5,mask_threshold:X=.5,overlap_mask_area_threshold:Ae=.8,label_ids_to_fuse:Oe=null,target_sizes:Ne=null,subtask:Be=null}={}){if(Array.isArray(Y)&&Y.length!==1)throw Error("Image segmentation pipeline currently only supports a batch size of 1.");const He=await g(Y),Xe=He.map(dt=>[dt.height,dt.width]),st=await this.processor(He),{inputNames:nt,outputNames:We}=this.model.sessions.model;if(!nt.includes("pixel_values")){if(nt.length!==1)throw Error(`Expected a single input name, but got ${nt.length} inputs: ${nt}.`);const dt=nt[0];if(dt in st)throw Error(`Input name ${dt} already exists in the inputs.`);st[dt]=st.pixel_values}const rt=await this.model(st);let pt=null;if(Be!==null)pt=this.subtasks_mapping[Be];else if(this.processor.image_processor){for(const[dt,Le]of Object.entries(this.subtasks_mapping))if(Le in this.processor.image_processor){pt=this.processor.image_processor[Le].bind(this.processor.image_processor),Be=dt;break}}const _t=this.model.config.id2label,ct=[];if(Be)if(Be==="panoptic"||Be==="instance"){const dt=pt(rt,pe,X,Ae,Oe,Ne??Xe)[0],Le=dt.segmentation;for(const xt of dt.segments_info){const Nt=new Uint8ClampedArray(Le.data.length);for(let Kt=0;Kt<Le.data.length;++Kt)Le.data[Kt]===xt.id&&(Nt[Kt]=255);const cr=new m.RawImage(Nt,Le.dims[1],Le.dims[0],1);ct.push({score:xt.score,label:_t[xt.label_id],mask:cr})}}else if(Be==="semantic"){const{segmentation:dt,labels:Le}=pt(rt,Ne??Xe)[0];for(const xt of Le){const Nt=new Uint8ClampedArray(dt.data.length);for(let Kt=0;Kt<dt.data.length;++Kt)dt.data[Kt]===xt&&(Nt[Kt]=255);const cr=new m.RawImage(Nt,dt.dims[1],dt.dims[0],1);ct.push({score:null,label:_t[xt],mask:cr})}}else throw Error(`Subtask ${Be} not supported.`);else{const Le=rt[We[0]];for(let xt=0;xt<Xe.length;++xt){const Nt=Xe[xt],cr=Le[xt];cr.data.some(wr=>wr<-1e-5||wr>1+1e-5)&&cr.sigmoid_();const Kt=await m.RawImage.fromTensor(cr.mul_(255).to("uint8")).resize(Nt[1],Nt[0]);ct.push({label:null,score:null,mask:Kt})}}return ct}}class ae extends _e{constructor(Y){super(Y)}async _call(Y,pe={}){if(Array.isArray(Y)&&Y.length!==1)throw Error("Background removal pipeline currently only supports a batch size of 1.");const Ae=await g(Y),Oe=await super._call(Y,pe);return Ae.map((Be,Ve)=>{const He=Be.clone();return He.putAlpha(Oe[Ve].mask),He})}}class Ce extends k{constructor(Y){super(Y)}async _call(Y,pe,{hypothesis_template:X="This is a photo of {}"}={}){const Ae=Array.isArray(Y),Oe=await g(Y),Ne=pe.map(nt=>X.replace("{}",nt)),Be=this.tokenizer(Ne,{padding:this.model.config.model_type==="siglip"?"max_length":!0,truncation:!0}),{pixel_values:Ve}=await this.processor(Oe),He=await this.model({...Be,pixel_values:Ve}),Xe=this.model.config.model_type==="siglip"?nt=>nt.sigmoid().data:nt=>(0,h.softmax)(nt.data),st=[];for(const nt of He.logits_per_image){const rt=[...Xe(nt)].map((pt,_t)=>({score:pt,label:pe[_t]}));rt.sort((pt,_t)=>_t.score-pt.score),st.push(rt)}return Ae?st:st[0]}}class Te extends k{constructor(Y){super(Y)}async _call(Y,{threshold:pe=.9,percentage:X=!1}={}){const Ae=Array.isArray(Y);if(Ae&&Y.length!==1)throw Error("Object detection pipeline currently only supports a batch size of 1.");const Oe=await g(Y),Ne=X?null:Oe.map(We=>[We.height,We.width]),{pixel_values:Be,pixel_mask:Ve}=await this.processor(Oe),He=await this.model({pixel_values:Be,pixel_mask:Ve}),Xe=this.processor.image_processor.post_process_object_detection(He,pe,Ne),st=this.model.config.id2label,nt=Xe.map(We=>We.boxes.map((rt,pt)=>({score:We.scores[pt],label:st[We.classes[pt]],box:b(rt,!X)})));return Ae?nt:nt[0]}}class q extends k{constructor(Y){super(Y)}async _call(Y,pe,{threshold:X=.1,top_k:Ae=null,percentage:Oe=!1}={}){const Ne=Array.isArray(Y),Be=await g(Y),Ve=this.tokenizer(pe,{padding:!0,truncation:!0}),He=await this.processor(Be),Xe=[];for(let st=0;st<Be.length;++st){const nt=Be[st],We=Oe?null:[[nt.height,nt.width]],rt=He.pixel_values[st].unsqueeze_(0),pt=await this.model({...Ve,pixel_values:rt});let _t;if("post_process_grounded_object_detection"in this.processor){const ct=this.processor.post_process_grounded_object_detection(pt,Ve.input_ids,{box_threshold:X,text_threshold:X,target_sizes:We})[0];_t=ct.boxes.map((dt,Le)=>({score:ct.scores[Le],label:ct.labels[Le],box:b(dt,!Oe)}))}else{const ct=this.processor.image_processor.post_process_object_detection(pt,X,We,!0)[0];_t=ct.boxes.map((dt,Le)=>({score:ct.scores[Le],label:pe[ct.classes[Le]],box:b(dt,!Oe)}))}_t.sort((ct,dt)=>dt.score-ct.score),Ae!==null&&(_t=_t.slice(0,Ae)),Xe.push(_t)}return Ne?Xe:Xe[0]}}class B extends k{constructor(Y){super(Y)}async _call(Y,pe,X={}){const Ae=(await g(Y))[0],{pixel_values:Oe}=await this.processor(Ae),Ne=`<s_docvqa><s_question>${pe}</s_question><s_answer>`,Be=this.tokenizer(Ne,{add_special_tokens:!1,padding:!0,truncation:!0}).input_ids,Ve=await this.model.generate({inputs:Oe,max_length:this.model.config.decoder.max_position_embeddings,decoder_input_ids:Be,...X}),Xe=this.tokenizer.batch_decode(Ve)[0].match(/<s_answer>(.*?)<\/s_answer>/);let st=null;return Xe&&Xe.length>=2&&(st=Xe[1].trim()),[{answer:st}]}}class Q extends k{constructor(pe){super(pe);le(this,"DEFAULT_VOCODER_ID","Xenova/speecht5_hifigan");this.vocoder=pe.vocoder??null}async _call(pe,{speaker_embeddings:X=null}={}){return this.processor?this._call_text_to_spectrogram(pe,{speaker_embeddings:X}):this._call_text_to_waveform(pe)}async _call_text_to_waveform(pe){const X=this.tokenizer(pe,{padding:!0,truncation:!0}),{waveform:Ae}=await this.model(X),Oe=this.model.config.sampling_rate;return new f.RawAudio(Ae.data,Oe)}async _call_text_to_spectrogram(pe,{speaker_embeddings:X}){if(this.vocoder||(console.log("No vocoder specified, using default HifiGan vocoder."),this.vocoder=await o.AutoModel.from_pretrained(this.DEFAULT_VOCODER_ID,{dtype:"fp32"})),(typeof X=="string"||X instanceof URL)&&(X=new Float32Array(await(await fetch(X)).arrayBuffer())),X instanceof Float32Array)X=new y.Tensor("float32",X,[1,X.length]);else if(!(X instanceof y.Tensor))throw new Error("Speaker embeddings must be a `Tensor`, `Float32Array`, `string`, or `URL`.");const{input_ids:Ae}=this.tokenizer(pe,{padding:!0,truncation:!0}),{waveform:Oe}=await this.model.generate_speech(Ae,X,{vocoder:this.vocoder}),Ne=this.processor.feature_extractor.config.sampling_rate;return new f.RawAudio(Oe.data,Ne)}}class oe extends k{constructor(Y){super(Y)}async _call(Y){const pe=await g(Y),X=await this.processor(pe),Ae=await this.model(X),Oe=[];for(const Ne of Ae.reconstruction){const Be=Ne.squeeze().clamp_(0,1).mul_(255).round_().to("uint8");Oe.push(m.RawImage.fromTensor(Be))}return Oe.length>1?Oe:Oe[0]}}class Ie extends k{constructor(Y){super(Y)}async _call(Y){const pe=await g(Y),X=await this.processor(pe),{predicted_depth:Ae}=await this.model(X),Oe=[];for(let Ne=0;Ne<pe.length;++Ne){const Be=Ae[Ne],[Ve,He]=Be.dims.slice(-2),[Xe,st]=pe[Ne].size,nt=(await(0,y.interpolate_4d)(Be.view(1,1,Ve,He),{size:[st,Xe],mode:"bilinear"})).view(st,Xe),We=nt.min().item(),rt=nt.max().item(),pt=nt.sub(We).div_(rt-We).mul_(255).to("uint8").unsqueeze(0),_t=m.RawImage.fromTensor(pt);Oe.push({predicted_depth:nt,depth:_t})}return Oe.length>1?Oe:Oe[0]}}const ye=Object.freeze({"text-classification":{tokenizer:n.AutoTokenizer,pipeline:z,model:o.AutoModelForSequenceClassification,default:{model:"Xenova/distilbert-base-uncased-finetuned-sst-2-english"},type:"text"},"token-classification":{tokenizer:n.AutoTokenizer,pipeline:E,model:o.AutoModelForTokenClassification,default:{model:"Xenova/bert-base-multilingual-cased-ner-hrl"},type:"text"},"question-answering":{tokenizer:n.AutoTokenizer,pipeline:T,model:o.AutoModelForQuestionAnswering,default:{model:"Xenova/distilbert-base-cased-distilled-squad"},type:"text"},"fill-mask":{tokenizer:n.AutoTokenizer,pipeline:D,model:o.AutoModelForMaskedLM,default:{model:"Xenova/bert-base-uncased"},type:"text"},summarization:{tokenizer:n.AutoTokenizer,pipeline:$,model:o.AutoModelForSeq2SeqLM,default:{model:"Xenova/distilbart-cnn-6-6"},type:"text"},translation:{tokenizer:n.AutoTokenizer,pipeline:P,model:o.AutoModelForSeq2SeqLM,default:{model:"Xenova/t5-small"},type:"text"},"text2text-generation":{tokenizer:n.AutoTokenizer,pipeline:I,model:o.AutoModelForSeq2SeqLM,default:{model:"Xenova/flan-t5-small"},type:"text"},"text-generation":{tokenizer:n.AutoTokenizer,pipeline:O,model:o.AutoModelForCausalLM,default:{model:"Xenova/gpt2"},type:"text"},"zero-shot-classification":{tokenizer:n.AutoTokenizer,pipeline:R,model:o.AutoModelForSequenceClassification,default:{model:"Xenova/distilbert-base-uncased-mnli"},type:"text"},"audio-classification":{pipeline:ie,model:o.AutoModelForAudioClassification,processor:i.AutoProcessor,default:{model:"Xenova/wav2vec2-base-superb-ks"},type:"audio"},"zero-shot-audio-classification":{tokenizer:n.AutoTokenizer,pipeline:H,model:o.AutoModel,processor:i.AutoProcessor,default:{model:"Xenova/clap-htsat-unfused"},type:"multimodal"},"automatic-speech-recognition":{tokenizer:n.AutoTokenizer,pipeline:ce,model:[o.AutoModelForSpeechSeq2Seq,o.AutoModelForCTC],processor:i.AutoProcessor,default:{model:"Xenova/whisper-tiny.en"},type:"multimodal"},"text-to-audio":{tokenizer:n.AutoTokenizer,pipeline:Q,model:[o.AutoModelForTextToWaveform,o.AutoModelForTextToSpectrogram],processor:[i.AutoProcessor,null],default:{model:"Xenova/speecht5_tts"},type:"text"},"image-to-text":{tokenizer:n.AutoTokenizer,pipeline:re,model:o.AutoModelForVision2Seq,processor:i.AutoProcessor,default:{model:"Xenova/vit-gpt2-image-captioning"},type:"multimodal"},"image-classification":{pipeline:se,model:o.AutoModelForImageClassification,processor:i.AutoProcessor,default:{model:"Xenova/vit-base-patch16-224"},type:"multimodal"},"image-segmentation":{pipeline:_e,model:[o.AutoModelForImageSegmentation,o.AutoModelForSemanticSegmentation,o.AutoModelForUniversalSegmentation],processor:i.AutoProcessor,default:{model:"Xenova/detr-resnet-50-panoptic"},type:"multimodal"},"background-removal":{pipeline:ae,model:[o.AutoModelForImageSegmentation,o.AutoModelForSemanticSegmentation,o.AutoModelForUniversalSegmentation],processor:i.AutoProcessor,default:{model:"Xenova/modnet"},type:"image"},"zero-shot-image-classification":{tokenizer:n.AutoTokenizer,pipeline:Ce,model:o.AutoModel,processor:i.AutoProcessor,default:{model:"Xenova/clip-vit-base-patch32"},type:"multimodal"},"object-detection":{pipeline:Te,model:o.AutoModelForObjectDetection,processor:i.AutoProcessor,default:{model:"Xenova/detr-resnet-50"},type:"multimodal"},"zero-shot-object-detection":{tokenizer:n.AutoTokenizer,pipeline:q,model:o.AutoModelForZeroShotObjectDetection,processor:i.AutoProcessor,default:{model:"Xenova/owlvit-base-patch32"},type:"multimodal"},"document-question-answering":{tokenizer:n.AutoTokenizer,pipeline:B,model:o.AutoModelForDocumentQuestionAnswering,processor:i.AutoProcessor,default:{model:"Xenova/donut-base-finetuned-docvqa"},type:"multimodal"},"image-to-image":{pipeline:oe,model:o.AutoModelForImageToImage,processor:i.AutoProcessor,default:{model:"Xenova/swin2SR-classical-sr-x2-64"},type:"image"},"depth-estimation":{pipeline:Ie,model:o.AutoModelForDepthEstimation,processor:i.AutoProcessor,default:{model:"Xenova/dpt-large"},type:"image"},"feature-extraction":{tokenizer:n.AutoTokenizer,pipeline:G,model:o.AutoModel,default:{model:"Xenova/all-MiniLM-L6-v2"},type:"text"},"image-feature-extraction":{processor:i.AutoProcessor,pipeline:ee,model:[o.AutoModelForImageFeatureExtraction,o.AutoModel],default:{model:"Xenova/vit-base-patch16-224-in21k"},type:"image"}}),be=Object.freeze({"sentiment-analysis":"text-classification",ner:"token-classification",asr:"automatic-speech-recognition","text-to-speech":"text-to-audio",embeddings:"feature-extraction"});async function et(Ee,Y=null,{progress_callback:pe=null,config:X=null,cache_dir:Ae=null,local_files_only:Oe=!1,revision:Ne="main",device:Be=null,dtype:Ve=null,subfolder:He="onnx",use_external_data_format:Xe=null,model_file_name:st=null,session_options:nt={}}={}){Ee=be[Ee]??Ee;const We=ye[Ee.split("_",1)[0]];if(!We)throw Error(`Unsupported pipeline: ${Ee}. Must be one of [${Object.keys(ye)}]`);Y||(Y=We.default.model,console.log(`No model specified. Using default model: "${Y}".`));const rt={progress_callback:pe,config:X,cache_dir:Ae,local_files_only:Oe,revision:Ne,device:Be,dtype:Ve,subfolder:He,use_external_data_format:Xe,model_file_name:st,session_options:nt},pt=new Map([["tokenizer",We.tokenizer],["model",We.model],["processor",We.processor]]),_t=await tt(pt,Y,rt);_t.task=Ee,(0,d.dispatchCallback)(pe,{status:"ready",task:Ee,model:Y});const ct=We.pipeline;return new ct(_t)}async function tt(Ee,Y,pe){const X=Object.create(null),Ae=[];for(const[Oe,Ne]of Ee.entries()){if(!Ne)continue;let Be;Array.isArray(Ne)?Be=new Promise(async(Ve,He)=>{var st,nt;let Xe;for(const We of Ne){if(We===null){Ve(null);return}try{Ve(await We.from_pretrained(Y,pe));return}catch(rt){if((st=rt.message)!=null&&st.includes("Unsupported model type"))Xe=rt;else if((nt=rt.message)!=null&&nt.includes("Could not locate file"))Xe=rt;else{He(rt);return}}}He(Xe)}):Be=Ne.from_pretrained(Y,pe),X[Oe]=Be,Ae.push(Be)}await Promise.all(Ae);for(const[Oe,Ne]of Object.entries(X))X[Oe]=await Ne;return X}},"./src/tokenizers.js":(e,r,t)=>{t.r(r),t.d(r,{AlbertTokenizer:()=>Ut,AutoTokenizer:()=>Ei,BartTokenizer:()=>Yi,BertTokenizer:()=>Dt,BlenderbotSmallTokenizer:()=>xs,BlenderbotTokenizer:()=>ai,BloomTokenizer:()=>Zi,CLIPTokenizer:()=>Bi,CamembertTokenizer:()=>Ai,CodeGenTokenizer:()=>qr,CodeLlamaTokenizer:()=>Hr,CohereTokenizer:()=>ls,ConvBertTokenizer:()=>ri,DebertaTokenizer:()=>gt,DebertaV2Tokenizer:()=>Ht,DistilBertTokenizer:()=>Jr,ElectraTokenizer:()=>Oi,EsmTokenizer:()=>yt,FalconTokenizer:()=>qe,GPT2Tokenizer:()=>ii,GPTNeoXTokenizer:()=>lt,GemmaTokenizer:()=>gi,Grok1Tokenizer:()=>Di,HerbertTokenizer:()=>nr,LlamaTokenizer:()=>Kr,M2M100Tokenizer:()=>ji,MBart50Tokenizer:()=>Fi,MBartTokenizer:()=>_i,MPNetTokenizer:()=>si,MarianTokenizer:()=>as,MgpstrTokenizer:()=>oi,MobileBertTokenizer:()=>Et,NllbTokenizer:()=>Li,NougatTokenizer:()=>pi,PreTrainedTokenizer:()=>De,Qwen2Tokenizer:()=>ur,RoFormerTokenizer:()=>pr,RobertaTokenizer:()=>er,SiglipTokenizer:()=>Yr,SpeechT5Tokenizer:()=>ts,SqueezeBertTokenizer:()=>jt,T5Tokenizer:()=>ns,TokenizerModel:()=>ee,VitsTokenizer:()=>Vs,Wav2Vec2CTCTokenizer:()=>os,WhisperTokenizer:()=>zi,XLMRobertaTokenizer:()=>es,XLMTokenizer:()=>Ji,is_chinese_char:()=>D});var n=t("./src/utils/generic.js"),o=t("./src/utils/core.js"),i=t("./src/utils/hub.js"),u=t("./src/utils/maths.js"),d=t("./src/utils/tensor.js"),h=t("./src/utils/data-structures.js"),f=t("./node_modules/@huggingface/jinja/dist/index.js"),y=t("./src/models/whisper/common_whisper.js");async function m(ve,j){const J=await Promise.all([(0,i.getModelJSON)(ve,"tokenizer.json",!0,j),(0,i.getModelJSON)(ve,"tokenizer_config.json",!0,j)]);return j.legacy!==null&&(J[1].legacy=j.legacy),J}function g(ve,j){const J=[];let ne=0;for(const Me of ve.matchAll(j)){const xe=Me[0];ne<Me.index&&J.push(ve.slice(ne,Me.index)),xe.length>0&&J.push(xe),ne=Me.index+xe.length}return ne<ve.length&&J.push(ve.slice(ne)),J}function v(ve,j=!0){if(ve.Regex!==void 0){let J=ve.Regex.replace(/\\([#&~])/g,"$1");for(const[ne,Me]of R)J=J.replaceAll(ne,Me);return new RegExp(J,"gu")}else if(ve.String!==void 0){const J=(0,o.escapeRegExp)(ve.String);return new RegExp(j?J:`(${J})`,"gu")}else return console.warn("Unknown pattern type:",ve),null}function b(ve){return new Map(Object.entries(ve))}function k(ve){const j=ve.dims;switch(j.length){case 1:return ve.tolist();case 2:if(j[0]!==1)throw new Error("Unable to decode tensor with `batch size !== 1`. Use `tokenizer.batch_decode(...)` for batched inputs.");return ve.tolist()[0];default:throw new Error(`Expected tensor to have 1-2 dimensions, got ${j.length}.`)}}function z(ve){return ve.replace(/ \./g,".").replace(/ \?/g,"?").replace(/ \!/g,"!").replace(/ ,/g,",").replace(/ \' /g,"'").replace(/ n\'t/g,"n't").replace(/ \'m/g,"'m").replace(/ \'s/g,"'s").replace(/ \'ve/g,"'ve").replace(/ \'re/g,"'re")}function E(ve){return ve.replace(new RegExp("\\p{M}","gu"),"")}function T(ve){return E(ve.toLowerCase())}function D(ve){return ve>=19968&&ve<=40959||ve>=13312&&ve<=19903||ve>=131072&&ve<=173791||ve>=173824&&ve<=177983||ve>=177984&&ve<=178207||ve>=178208&&ve<=183983||ve>=63744&&ve<=64255||ve>=194560&&ve<=195103}function I(ve,j,J){const ne=[];let Me=0;for(;Me<ve.length;){if(ne.push(ve[Me]),(j.get(ve[Me])??J)!==J){++Me;continue}for(;++Me<ve.length&&(j.get(ve[Me])??J)===J;)j.get(ne.at(-1))!==J&&(ne[ne.length-1]+=ve[Me])}return ne}function $(ve){return ve.match(/\S+/g)||[]}const P="\\p{P}\\u0021-\\u002F\\u003A-\\u0040\\u005B-\\u0060\\u007B-\\u007E",S=new RegExp(`^[${P}]+$`,"gu"),O=".,!?…。，、।۔،",R=new Map([["(?i:'s|'t|'re|'ve|'m|'ll|'d)","(?:'([sS]|[tT]|[rR][eE]|[vV][eE]|[mM]|[lL][lL]|[dD]))"],[` ?[^(\\s|[${O}])]+`,` ?[^\\s${O}]+`]]);class G{constructor(j){this.content=j.content,this.id=j.id,this.single_word=j.single_word??!1,this.lstrip=j.lstrip??!1,this.rstrip=j.rstrip??!1,this.special=j.special??!1,this.normalized=j.normalized??null}}class ee extends n.Callable{constructor(j){super(),this.config=j,this.vocab=[],this.tokens_to_ids=new Map,this.unk_token_id=void 0,this.unk_token=void 0,this.end_of_word_suffix=void 0,this.fuse_unk=this.config.fuse_unk??!1}static fromConfig(j,...J){switch(j.type){case"WordPiece":return new ie(j);case"Unigram":return new H(j,...J);case"BPE":return new se(j);default:if(j.vocab)return Array.isArray(j.vocab)?new H(j,...J):Object.hasOwn(j,"continuing_subword_prefix")&&Object.hasOwn(j,"unk_token")?Object.hasOwn(j,"merges")?new se(j):new ie(j):new _e(j,...J);throw new Error(`Unknown TokenizerModel type: ${j.type}`)}}_call(j){return j=this.encode(j),this.fuse_unk&&(j=I(j,this.tokens_to_ids,this.unk_token_id)),j}encode(j){throw Error("encode should be implemented in subclass.")}convert_tokens_to_ids(j){return j.map(J=>this.tokens_to_ids.get(J)??this.unk_token_id)}convert_ids_to_tokens(j){return j.map(J=>this.vocab[J]??this.unk_token)}}class ie extends ee{constructor(j){super(j),this.tokens_to_ids=b(j.vocab),this.unk_token_id=this.tokens_to_ids.get(j.unk_token),this.unk_token=j.unk_token,this.max_input_chars_per_word=j.max_input_chars_per_word??100,this.vocab=new Array(this.tokens_to_ids.size);for(const[J,ne]of this.tokens_to_ids)this.vocab[ne]=J}encode(j){const J=[];for(const ne of j){const Me=[...ne];if(Me.length>this.max_input_chars_per_word){J.push(this.unk_token);continue}let xe=!1,je=0;const Ye=[];for(;je<Me.length;){let ot=Me.length,it=null;for(;je<ot;){let ut=Me.slice(je,ot).join("");if(je>0&&(ut=this.config.continuing_subword_prefix+ut),this.tokens_to_ids.has(ut)){it=ut;break}--ot}if(it===null){xe=!0;break}Ye.push(it),je=ot}xe?J.push(this.unk_token):J.push(...Ye)}return J}}class H extends ee{constructor(j,J){super(j);const ne=j.vocab.length;this.vocab=new Array(ne),this.scores=new Array(ne);for(let Me=0;Me<ne;++Me)[this.vocab[Me],this.scores[Me]]=j.vocab[Me];this.unk_token_id=j.unk_id,this.unk_token=this.vocab[j.unk_id],this.tokens_to_ids=new Map(this.vocab.map((Me,xe)=>[Me,xe])),this.bos_token=" ",this.bos_token_id=this.tokens_to_ids.get(this.bos_token),this.eos_token=J.eos_token,this.eos_token_id=this.tokens_to_ids.get(this.eos_token),this.unk_token=this.vocab[this.unk_token_id],this.minScore=(0,u.min)(this.scores)[0],this.unk_score=this.minScore-10,this.scores[this.unk_token_id]=this.unk_score,this.trie=new h.CharTrie,this.trie.extend(this.vocab),this.fuse_unk=!0}populateNodes(j){const J=j.chars,ne=1;let Me=0;for(;Me<J.length;){let xe=!1;const je=J.slice(Me).join(""),Ye=this.trie.commonPrefixSearch(je);for(const ot of Ye){const it=this.tokens_to_ids.get(ot),ut=this.scores[it],It=(0,o.len)(ot);j.insert(Me,It,ut,it),!xe&&It===ne&&(xe=!0)}xe||j.insert(Me,ne,this.unk_score,this.unk_token_id),Me+=ne}}tokenize(j){const J=new h.TokenLattice(j,this.bos_token_id,this.eos_token_id);return this.populateNodes(J),J.tokens()}encode(j){const J=[];for(const ne of j){const Me=this.tokenize(ne);J.push(...Me)}return J}}const ce=(()=>{const ve=[...Array.from({length:94},(Me,xe)=>xe+33),...Array.from({length:12},(Me,xe)=>xe+161),...Array.from({length:82},(Me,xe)=>xe+174)],j=ve.slice();let J=0;for(let Me=0;Me<256;++Me)ve.includes(Me)||(ve.push(Me),j.push(256+J),J+=1);const ne=j.map(Me=>String.fromCharCode(Me));return Object.fromEntries(ve.map((Me,xe)=>[Me,ne[xe]]))})(),re=(0,o.reverseDictionary)(ce);class se extends ee{constructor(j){super(j),this.tokens_to_ids=b(j.vocab),this.unk_token_id=this.tokens_to_ids.get(j.unk_token),this.unk_token=j.unk_token,this.vocab=new Array(this.tokens_to_ids.size);for(const[ne,Me]of this.tokens_to_ids)this.vocab[Me]=ne;const J=Array.isArray(j.merges[0]);this.merges=J?j.merges:j.merges.map(ne=>ne.split(" ",2)),this.bpe_ranks=new Map(this.merges.map((ne,Me)=>[JSON.stringify(ne),Me])),this.end_of_word_suffix=j.end_of_word_suffix,this.continuing_subword_suffix=j.continuing_subword_suffix??null,this.byte_fallback=this.config.byte_fallback??!1,this.byte_fallback&&(this.text_encoder=new TextEncoder),this.ignore_merges=this.config.ignore_merges??!1,this.max_length_to_cache=256,this.cache_capacity=1e4,this.cache=new h.LRUCache(this.cache_capacity)}clear_cache(){this.cache.clear()}bpe(j){if(j.length===0)return[];const J=this.cache.get(j);if(J!==void 0)return J;const ne=Array.from(j);this.end_of_word_suffix&&(ne[ne.length-1]+=this.end_of_word_suffix);let Me=[];if(ne.length>1){const xe=new h.PriorityQueue((ot,it)=>ot.score<it.score);let je={token:ne[0],bias:0,prev:null,next:null},Ye=je;for(let ot=1;ot<ne.length;++ot){const it={bias:ot/ne.length,token:ne[ot],prev:Ye,next:null};Ye.next=it,this._add_node(xe,Ye),Ye=it}for(;!xe.isEmpty();){const ot=xe.pop();if(ot.deleted||!ot.next||ot.next.deleted)continue;if(ot.deleted=!0,ot.next.deleted=!0,ot.prev){const ut={...ot.prev};ot.prev.deleted=!0,ot.prev=ut,ut.prev?ut.prev.next=ut:je=ut}const it={token:ot.token+ot.next.token,bias:ot.bias,prev:ot.prev,next:ot.next.next};it.prev?(it.prev.next=it,this._add_node(xe,it.prev)):je=it,it.next&&(it.next.prev=it,this._add_node(xe,it))}for(let ot=je;ot!==null;ot=ot.next)Me.push(ot.token)}else Me=ne;if(this.continuing_subword_suffix)for(let xe=0;xe<Me.length-1;++xe)Me[xe]+=this.continuing_subword_suffix;return j.length<this.max_length_to_cache&&this.cache.put(j,Me),Me}_add_node(j,J){const ne=this.bpe_ranks.get(JSON.stringify([J.token,J.next.token]));ne!==void 0&&(J.score=ne+J.bias,j.push(J))}encode(j){const J=[];for(const ne of j){if(this.ignore_merges&&this.tokens_to_ids.has(ne)){J.push(ne);continue}const Me=this.bpe(ne);for(const xe of Me)if(this.tokens_to_ids.has(xe))J.push(xe);else if(this.byte_fallback){const je=Array.from(this.text_encoder.encode(xe)).map(Ye=>`<0x${Ye.toString(16).toUpperCase().padStart(2,"0")}>`);je.every(Ye=>this.tokens_to_ids.has(Ye))?J.push(...je):J.push(this.unk_token)}else J.push(this.unk_token)}return J}}class _e extends ee{constructor(j,J){super(j),this.tokens_to_ids=b(J.target_lang?j.vocab[J.target_lang]:j.vocab),this.bos_token=J.bos_token,this.bos_token_id=this.tokens_to_ids.get(this.bos_token),this.eos_token=J.eos_token,this.eos_token_id=this.tokens_to_ids.get(this.eos_token),this.pad_token=J.pad_token,this.pad_token_id=this.tokens_to_ids.get(this.pad_token),this.unk_token=J.unk_token,this.unk_token_id=this.tokens_to_ids.get(this.unk_token),this.vocab=new Array(this.tokens_to_ids.size);for(const[ne,Me]of this.tokens_to_ids)this.vocab[Me]=ne}encode(j){return j}}class ae extends n.Callable{constructor(j){super(),this.config=j}static fromConfig(j){if(j===null)return null;switch(j.type){case"BertNormalizer":return new Ee(j);case"Precompiled":return new $t(j);case"Sequence":return new tt(j);case"Replace":return new Ce(j);case"NFC":return new q(j);case"NFD":return new B(j);case"NFKC":return new Q(j);case"NFKD":return new oe(j);case"Strip":return new Ie(j);case"StripAccents":return new ye(j);case"Lowercase":return new be(j);case"Prepend":return new et(j);default:throw new Error(`Unknown Normalizer type: ${j.type}`)}}normalize(j){throw Error("normalize should be implemented in subclass.")}_call(j){return this.normalize(j)}}class Ce extends ae{normalize(j){const J=v(this.config.pattern);return J===null?j:j.replaceAll(J,this.config.content)}}class Te extends ae{constructor(){super(...arguments);le(this,"form")}normalize(J){return J=J.normalize(this.form),J}}class q extends Te{constructor(){super(...arguments);le(this,"form","NFC")}}class B extends Te{constructor(){super(...arguments);le(this,"form","NFD")}}class Q extends Te{constructor(){super(...arguments);le(this,"form","NFKC")}}class oe extends Te{constructor(){super(...arguments);le(this,"form","NFKD")}}class Ie extends ae{normalize(j){return this.config.strip_left&&this.config.strip_right?j=j.trim():(this.config.strip_left&&(j=j.trimStart()),this.config.strip_right&&(j=j.trimEnd())),j}}class ye extends ae{normalize(j){return j=E(j),j}}class be extends ae{normalize(j){return j=j.toLowerCase(),j}}class et extends ae{normalize(j){return j=this.config.prepend+j,j}}class tt extends ae{constructor(j){super(j),this.normalizers=j.normalizers.map(J=>ae.fromConfig(J))}normalize(j){return this.normalizers.reduce((J,ne)=>ne.normalize(J),j)}}class Ee extends ae{_tokenize_chinese_chars(j){const J=[];for(let ne=0;ne<j.length;++ne){const Me=j[ne],xe=Me.charCodeAt(0);D(xe)?(J.push(" "),J.push(Me),J.push(" ")):J.push(Me)}return J.join("")}stripAccents(j){return j.normalize("NFD").replace(new RegExp("\\p{Mn}","gu"),"")}_is_control(j){switch(j){case"	":case`
`:case"\r":return!1;default:return new RegExp("^\\p{Cc}|\\p{Cf}|\\p{Co}|\\p{Cs}$","u").test(j)}}_clean_text(j){const J=[];for(const ne of j){const Me=ne.charCodeAt(0);Me===0||Me===65533||this._is_control(ne)||(/^\s$/.test(ne)?J.push(" "):J.push(ne))}return J.join("")}normalize(j){return this.config.clean_text&&(j=this._clean_text(j)),this.config.handle_chinese_chars&&(j=this._tokenize_chinese_chars(j)),this.config.lowercase?(j=j.toLowerCase(),this.config.strip_accents!==!1&&(j=this.stripAccents(j))):this.config.strip_accents&&(j=this.stripAccents(j)),j}}class Y extends n.Callable{static fromConfig(j){if(j===null)return null;switch(j.type){case"BertPreTokenizer":return new pe(j);case"Sequence":return new Ti(j);case"Whitespace":return new U(j);case"WhitespaceSplit":return new ge(j);case"Metaspace":return new wr(j);case"ByteLevel":return new X(j);case"Split":return new Ae(j);case"Punctuation":return new Oe(j);case"Digits":return new Ne(j);case"Replace":return new K(j);default:throw new Error(`Unknown PreTokenizer type: ${j.type}`)}}pre_tokenize_text(j,J){throw Error("pre_tokenize_text should be implemented in subclass.")}pre_tokenize(j,J){return(Array.isArray(j)?j.map(ne=>this.pre_tokenize_text(ne,J)):this.pre_tokenize_text(j,J)).flat()}_call(j,J){return this.pre_tokenize(j,J)}}class pe extends Y{constructor(j){super(),this.pattern=new RegExp(`[^\\s${P}]+|[${P}]`,"gu")}pre_tokenize_text(j,J){return j.trim().match(this.pattern)||[]}}class X extends Y{constructor(j){super(),this.config=j,this.add_prefix_space=this.config.add_prefix_space,this.trim_offsets=this.config.trim_offsets,this.use_regex=this.config.use_regex??!0,this.pattern=new RegExp("'s|'t|'re|'ve|'m|'ll|'d| ?\\p{L}+| ?\\p{N}+| ?[^\\s\\p{L}\\p{N}]+|\\s+(?!\\S)|\\s+","gu"),this.byte_encoder=ce,this.text_encoder=new TextEncoder}pre_tokenize_text(j,J){return this.add_prefix_space&&!j.startsWith(" ")&&(j=" "+j),(this.use_regex?j.match(this.pattern)||[]:[j]).map(Me=>Array.from(this.text_encoder.encode(Me),xe=>this.byte_encoder[xe]).join(""))}}class Ae extends Y{constructor(j){super(),this.config=j,this.pattern=v(this.config.pattern,this.config.invert)}pre_tokenize_text(j,J){var ne;return this.pattern===null?[]:this.config.invert?j.match(this.pattern)||[]:((ne=this.config.behavior)==null?void 0:ne.toLowerCase())==="removed"?j.split(this.pattern).filter(Me=>Me):g(j,this.pattern)}}class Oe extends Y{constructor(j){super(),this.config=j,this.pattern=new RegExp(`[^${P}]+|[${P}]+`,"gu")}pre_tokenize_text(j,J){return j.match(this.pattern)||[]}}class Ne extends Y{constructor(j){super(),this.config=j;const J=`[^\\d]+|\\d${this.config.individual_digits?"":"+"}`;this.pattern=new RegExp(J,"gu")}pre_tokenize_text(j,J){return j.match(this.pattern)||[]}}class Be extends n.Callable{constructor(j){super(),this.config=j}static fromConfig(j){if(j===null)return null;switch(j.type){case"TemplateProcessing":return new Xe(j);case"ByteLevel":return new st(j);case"RobertaProcessing":return new He(j);case"BertProcessing":return new Ve(j);case"Sequence":return new nt(j);default:throw new Error(`Unknown PostProcessor type: ${j.type}`)}}post_process(j,...J){throw Error("post_process should be implemented in subclass.")}_call(j,...J){return this.post_process(j,...J)}}class Ve extends Be{constructor(j){super(j),this.cls=j.cls[0],this.sep=j.sep[0]}post_process(j,J=null,{add_special_tokens:ne=!0}={}){ne&&(j=(0,o.mergeArrays)([this.cls],j,[this.sep]));let Me=new Array(j.length).fill(0);if(J!==null){const xe=ne&&this instanceof He?[this.sep]:[],je=ne?[this.sep]:[];j=(0,o.mergeArrays)(j,xe,J,je),Me=(0,o.mergeArrays)(Me,new Array(J.length+xe.length+je.length).fill(1))}return{tokens:j,token_type_ids:Me}}}class He extends Ve{}class Xe extends Be{constructor(j){super(j),this.single=j.single,this.pair=j.pair}post_process(j,J=null,{add_special_tokens:ne=!0}={}){const Me=J===null?this.single:this.pair;let xe=[],je=[];for(const Ye of Me)"SpecialToken"in Ye?ne&&(xe.push(Ye.SpecialToken.id),je.push(Ye.SpecialToken.type_id)):"Sequence"in Ye&&(Ye.Sequence.id==="A"?(xe=(0,o.mergeArrays)(xe,j),je=(0,o.mergeArrays)(je,new Array(j.length).fill(Ye.Sequence.type_id))):Ye.Sequence.id==="B"&&(xe=(0,o.mergeArrays)(xe,J),je=(0,o.mergeArrays)(je,new Array(J.length).fill(Ye.Sequence.type_id))));return{tokens:xe,token_type_ids:je}}}class st extends Be{post_process(j,J=null){return J&&(j=(0,o.mergeArrays)(j,J)),{tokens:j}}}class nt extends Be{constructor(j){super(j),this.processors=j.processors.map(J=>Be.fromConfig(J))}post_process(j,J=null,ne={}){let Me;for(const xe of this.processors)if(xe instanceof st)j=xe.post_process(j).tokens,J&&(J=xe.post_process(J).tokens);else{const je=xe.post_process(j,J,ne);j=je.tokens,Me=je.token_type_ids}return{tokens:j,token_type_ids:Me}}}class We extends n.Callable{constructor(j){super(),this.config=j,this.added_tokens=[],this.end_of_word_suffix=null,this.trim_offsets=j.trim_offsets}static fromConfig(j){if(j===null)return null;switch(j.type){case"WordPiece":return new dt(j);case"Metaspace":return new zr(j);case"ByteLevel":return new Le(j);case"Replace":return new rt(j);case"ByteFallback":return new pt(j);case"Fuse":return new _t(j);case"Strip":return new ct(j);case"Sequence":return new Nt(j);case"CTC":return new xt(j);case"BPEDecoder":return new cr(j);default:throw new Error(`Unknown Decoder type: ${j.type}`)}}_call(j){return this.decode(j)}decode(j){return this.decode_chain(j).join("")}decode_chain(j){throw Error("`decode_chain` should be implemented in subclass.")}}class rt extends We{decode_chain(j){const J=v(this.config.pattern);return J===null?j:j.map(ne=>ne.replaceAll(J,this.config.content))}}class pt extends We{constructor(j){super(j),this.text_decoder=new TextDecoder}decode_chain(j){const J=[];let ne=[];for(const Me of j){let xe=null;if(Me.length===6&&Me.startsWith("<0x")&&Me.endsWith(">")){const je=parseInt(Me.slice(3,5),16);isNaN(je)||(xe=je)}if(xe!==null)ne.push(xe);else{if(ne.length>0){const je=this.text_decoder.decode(Uint8Array.from(ne));J.push(je),ne=[]}J.push(Me)}}if(ne.length>0){const Me=this.text_decoder.decode(Uint8Array.from(ne));J.push(Me),ne=[]}return J}}class _t extends We{decode_chain(j){return[j.join("")]}}class ct extends We{constructor(j){super(j),this.content=this.config.content,this.start=this.config.start,this.stop=this.config.stop}decode_chain(j){return j.map(J=>{let ne=0;for(let xe=0;xe<this.start&&J[xe]===this.content;++xe){ne=xe+1;continue}let Me=J.length;for(let xe=0;xe<this.stop;++xe){const je=J.length-xe-1;if(J[je]===this.content){Me=je;continue}else break}return J.slice(ne,Me)})}}class dt extends We{constructor(j){super(j),this.cleanup=j.cleanup}decode_chain(j){return j.map((J,ne)=>(ne!==0&&(J.startsWith(this.config.prefix)?J=J.replace(this.config.prefix,""):J=" "+J),this.cleanup&&(J=z(J)),J))}}class Le extends We{constructor(j){super(j),this.byte_decoder=re,this.text_decoder=new TextDecoder("utf-8",{fatal:!1,ignoreBOM:!0}),this.end_of_word_suffix=null}convert_tokens_to_string(j){const J=j.join(""),ne=new Uint8Array([...J].map(xe=>this.byte_decoder[xe]));return this.text_decoder.decode(ne)}decode_chain(j){const J=[];let ne=[];for(const Me of j)this.added_tokens.find(xe=>xe.content===Me)!==void 0?(ne.length>0&&(J.push(this.convert_tokens_to_string(ne)),ne=[]),J.push(Me)):ne.push(Me);return ne.length>0&&J.push(this.convert_tokens_to_string(ne)),J}}class xt extends We{constructor(j){super(j),this.pad_token=this.config.pad_token,this.word_delimiter_token=this.config.word_delimiter_token,this.cleanup=this.config.cleanup}convert_tokens_to_string(j){if(j.length===0)return"";const J=[j[0]];for(let xe=1;xe<j.length;++xe)j[xe]!==J.at(-1)&&J.push(j[xe]);let Me=J.filter(xe=>xe!==this.pad_token).join("");return this.cleanup&&(Me=z(Me).replaceAll(this.word_delimiter_token," ").trim()),Me}decode_chain(j){return[this.convert_tokens_to_string(j)]}}class Nt extends We{constructor(j){super(j),this.decoders=j.decoders.map(J=>We.fromConfig(J))}decode_chain(j){return this.decoders.reduce((J,ne)=>ne.decode_chain(J),j)}}class cr extends We{constructor(j){super(j),this.suffix=this.config.suffix}decode_chain(j){return j.map((J,ne)=>J.replaceAll(this.suffix,ne===j.length-1?"":" "))}}class Kt extends We{decode_chain(j){let J="";for(let ne=1;ne<j.length;ne+=2)J+=j[ne];return[J]}}class wr extends Y{constructor(j){super(),this.addPrefixSpace=j.add_prefix_space,this.replacement=j.replacement,this.strRep=j.str_rep||this.replacement,this.prepend_scheme=j.prepend_scheme??"always"}pre_tokenize_text(j,{section_index:J=void 0}={}){let ne=j.replaceAll(" ",this.strRep);return this.addPrefixSpace&&!ne.startsWith(this.replacement)&&(this.prepend_scheme==="always"||this.prepend_scheme==="first"&&J===0)&&(ne=this.strRep+ne),[ne]}}class zr extends We{constructor(j){super(j),this.addPrefixSpace=j.add_prefix_space,this.replacement=j.replacement}decode_chain(j){const J=[];for(let ne=0;ne<j.length;++ne){let Me=j[ne].replaceAll(this.replacement," ");this.addPrefixSpace&&ne==0&&Me.startsWith(" ")&&(Me=Me.substring(1)),J.push(Me)}return J}}class $t extends ae{constructor(j){super(j),this.charsmap=j.precompiled_charsmap}normalize(j){return j=j.replace(/[\u0001-\u0008\u000B\u000E-\u001F\u007F\u008F\u009F]/gm,""),j=j.replace(/[\u0009\u000A\u000C\u000D\u00A0\u1680\u2000-\u200F\u2028\u2029\u202F\u205F\u2581\u3000\uFEFF\uFFFD]/gm," "),j.includes("～")?j=j.split("～").map(ne=>ne.normalize("NFKC")).join("～"):j=j.normalize("NFKC"),j}}class Ti extends Y{constructor(j){super(),this.tokenizers=j.pretokenizers.map(J=>Y.fromConfig(J))}pre_tokenize_text(j,J){return this.tokenizers.reduce((ne,Me)=>Me.pre_tokenize(ne,J),[j])}}class U extends Y{constructor(j){super()}pre_tokenize_text(j,J){return j.match(/\w+|[^\w\s]+/g)||[]}}class ge extends Y{constructor(j){super()}pre_tokenize_text(j,J){return $(j)}}class K extends Y{constructor(j){super(),this.config=j,this.pattern=v(this.config.pattern),this.content=this.config.content}pre_tokenize_text(j,J){return this.pattern===null?[j]:[j.replaceAll(this.pattern,this.config.content)]}}const ue=["bos_token","eos_token","unk_token","sep_token","pad_token","cls_token","mask_token"];function $e(ve,j,J,ne){for(const Me of Object.keys(ve)){const xe=j-ve[Me].length,je=J(Me),Ye=new Array(xe).fill(je);ve[Me]=ne==="right"?(0,o.mergeArrays)(ve[Me],Ye):(0,o.mergeArrays)(Ye,ve[Me])}}function Ge(ve,j){for(const J of Object.keys(ve))ve[J].length=j}class De extends n.Callable{constructor(J,ne){super();le(this,"return_token_type_ids",!1);le(this,"padding_side","right");this._tokenizer_config=ne,this.normalizer=ae.fromConfig(J.normalizer),this.pre_tokenizer=Y.fromConfig(J.pre_tokenizer),this.model=ee.fromConfig(J.model,ne),this.post_processor=Be.fromConfig(J.post_processor),this.decoder=We.fromConfig(J.decoder),this.special_tokens=[],this.all_special_ids=[],this.added_tokens=[];for(const Me of J.added_tokens){const xe=new G(Me);this.added_tokens.push(xe),this.model.tokens_to_ids.set(xe.content,xe.id),this.model.vocab[xe.id]=xe.content,xe.special&&(this.special_tokens.push(xe.content),this.all_special_ids.push(xe.id))}if(this.additional_special_tokens=ne.additional_special_tokens??[],this.special_tokens.push(...this.additional_special_tokens),this.special_tokens=[...new Set(this.special_tokens)],this.decoder&&(this.decoder.added_tokens=this.added_tokens,this.decoder.end_of_word_suffix=this.model.end_of_word_suffix),this.added_tokens_splitter=new h.DictionarySplitter(this.added_tokens.map(Me=>Me.content)),this.added_tokens_map=new Map(this.added_tokens.map(Me=>[Me.content,Me])),this.mask_token=this.getToken("mask_token"),this.mask_token_id=this.model.tokens_to_ids.get(this.mask_token),this.pad_token=this.getToken("pad_token","eos_token"),this.pad_token_id=this.model.tokens_to_ids.get(this.pad_token),this.sep_token=this.getToken("sep_token"),this.sep_token_id=this.model.tokens_to_ids.get(this.sep_token),this.unk_token=this.getToken("unk_token"),this.unk_token_id=this.model.tokens_to_ids.get(this.unk_token),this.bos_token=this.getToken("bos_token"),this.bos_token_id=this.model.tokens_to_ids.get(this.bos_token),this.eos_token=this.getToken("eos_token"),this.eos_token_id=this.model.tokens_to_ids.get(this.eos_token),this.model_max_length=ne.model_max_length,this.remove_space=ne.remove_space,this.clean_up_tokenization_spaces=ne.clean_up_tokenization_spaces??!0,this.do_lowercase_and_remove_accent=ne.do_lowercase_and_remove_accent??!1,ne.padding_side&&(this.padding_side=ne.padding_side),this.legacy=!1,this.chat_template=ne.chat_template??null,Array.isArray(this.chat_template)){const Me=Object.create(null);for(const{name:xe,template:je}of this.chat_template){if(typeof xe!="string"||typeof je!="string")throw new Error('Chat template must be a list of objects with "name" and "template" properties');Me[xe]=je}this.chat_template=Me}this._compiled_template_cache=new Map}getToken(...J){for(const ne of J){const Me=this._tokenizer_config[ne];if(Me)if(typeof Me=="object"){if(Me.__type==="AddedToken")return Me.content;throw Error(`Unknown token: ${Me}`)}else return Me}return null}static async from_pretrained(J,{progress_callback:ne=null,config:Me=null,cache_dir:xe=null,local_files_only:je=!1,revision:Ye="main",legacy:ot=null}={}){const it=await m(J,{progress_callback:ne,config:Me,cache_dir:xe,local_files_only:je,revision:Ye,legacy:ot});return new this(...it)}_call(J,{text_pair:ne=null,add_special_tokens:Me=!0,padding:xe=!1,truncation:je=null,max_length:Ye=null,return_tensor:ot=!0,return_token_type_ids:it=null}={}){const ut=Array.isArray(J);let It;if(ut){if(J.length===0)throw Error("text array must be non-empty");if(ne!==null){if(Array.isArray(ne)){if(J.length!==ne.length)throw Error("text and text_pair must have the same length")}else throw Error("text_pair must also be an array");It=J.map((qt,_r)=>this._encode_plus(qt,{text_pair:ne[_r],add_special_tokens:Me,return_token_type_ids:it}))}else It=J.map(qt=>this._encode_plus(qt,{add_special_tokens:Me,return_token_type_ids:it}))}else{if(J==null)throw Error("text may not be null or undefined");if(Array.isArray(ne))throw Error("When specifying `text_pair`, since `text` is a string, `text_pair` must also be a string (i.e., not an array).");It=[this._encode_plus(J,{text_pair:ne,add_special_tokens:Me,return_token_type_ids:it})]}if(Ye===null?Ye=this.model_max_length:je===null&&(xe===!0?(console.warn("`max_length` is ignored when `padding: true` and there is no truncation strategy. To pad to max length, use `padding: 'max_length'`."),Ye=this.model_max_length):xe===!1&&(console.warn("Truncation was not explicitly activated but `max_length` is provided a specific value, please use `truncation: true` to explicitly truncate examples to max length."),je=!0)),xe===!0&&(Ye=Math.min((0,u.max)(It.map(qt=>qt.input_ids.length))[0],Ye??1/0)),Ye=Math.min(Ye,this.model_max_length??1/0),xe||je)for(let qt=0;qt<It.length;++qt)It[qt].input_ids.length!==Ye&&(It[qt].input_ids.length>Ye?je&&Ge(It[qt],Ye):xe&&$e(It[qt],Ye,_r=>_r==="input_ids"?this.pad_token_id:0,this.padding_side));const Qt={};if(ot){if(!(xe&&je)&&It.some(_r=>{var tr;for(const Cr of Object.keys(_r))if(_r[Cr].length!==((tr=It[0][Cr])==null?void 0:tr.length))return!0;return!1}))throw Error("Unable to create tensor, you should probably activate truncation and/or padding with 'padding=true' and 'truncation=true' to have batched tensors with the same length.");const qt=[It.length,It[0].input_ids.length];for(const _r of Object.keys(It[0]))Qt[_r]=new d.Tensor("int64",BigInt64Array.from(It.flatMap(tr=>tr[_r]).map(BigInt)),qt)}else{for(const qt of Object.keys(It[0]))Qt[qt]=It.map(_r=>_r[qt]);if(!ut)for(const qt of Object.keys(Qt))Qt[qt]=Qt[qt][0]}return Qt}_encode_text(J){if(J===null)return null;const ne=this.added_tokens_splitter.split(J);for(let xe=0;xe<ne.length;++xe){const je=this.added_tokens_map.get(ne[xe]);je&&(je.lstrip&&xe>0&&(ne[xe-1]=ne[xe-1].trimEnd()),je.rstrip&&xe<ne.length-1&&(ne[xe+1]=ne[xe+1].trimStart()))}return ne.flatMap((xe,je)=>{if(xe.length===0)return[];if(this.added_tokens_map.has(xe))return[xe];if(this.remove_space===!0&&(xe=xe.trim().split(/\s+/).join(" ")),this.do_lowercase_and_remove_accent&&(xe=T(xe)),this.normalizer!==null&&(xe=this.normalizer(xe)),xe.length===0)return[];const Ye=this.pre_tokenizer!==null?this.pre_tokenizer(xe,{section_index:je}):[xe];return this.model(Ye)})}_encode_plus(J,{text_pair:ne=null,add_special_tokens:Me=!0,return_token_type_ids:xe=null}={}){const{tokens:je,token_type_ids:Ye}=this._tokenize_helper(J,{pair:ne,add_special_tokens:Me}),ot=this.model.convert_tokens_to_ids(je),it={input_ids:ot,attention_mask:new Array(ot.length).fill(1)};return(xe??this.return_token_type_ids)&&Ye&&(it.token_type_ids=Ye),it}_tokenize_helper(J,{pair:ne=null,add_special_tokens:Me=!1}={}){const xe=this._encode_text(J),je=this._encode_text(ne);return this.post_processor?this.post_processor(xe,je,{add_special_tokens:Me}):{tokens:(0,o.mergeArrays)(xe??[],je??[])}}tokenize(J,{pair:ne=null,add_special_tokens:Me=!1}={}){return this._tokenize_helper(J,{pair:ne,add_special_tokens:Me}).tokens}encode(J,{text_pair:ne=null,add_special_tokens:Me=!0,return_token_type_ids:xe=null}={}){return this._encode_plus(J,{text_pair:ne,add_special_tokens:Me,return_token_type_ids:xe}).input_ids}batch_decode(J,ne={}){return J instanceof d.Tensor&&(J=J.tolist()),J.map(Me=>this.decode(Me,ne))}decode(J,ne={}){if(J instanceof d.Tensor&&(J=k(J)),!Array.isArray(J)||J.length===0||!(0,o.isIntegralNumber)(J[0]))throw Error("token_ids must be a non-empty array of integers.");return this.decode_single(J,ne)}decode_single(J,{skip_special_tokens:ne=!1,clean_up_tokenization_spaces:Me=null}){let xe=this.model.convert_ids_to_tokens(J);ne&&(xe=xe.filter(Ye=>!this.special_tokens.includes(Ye)));let je=this.decoder?this.decoder(xe):xe.join(" ");return this.decoder&&this.decoder.end_of_word_suffix&&(je=je.replaceAll(this.decoder.end_of_word_suffix," "),ne&&(je=je.trim())),(Me??this.clean_up_tokenization_spaces)&&(je=z(je)),je}get_chat_template({chat_template:J=null,tools:ne=null}={}){if(this.chat_template&&typeof this.chat_template=="object"){const Me=this.chat_template;if(J!==null&&Object.hasOwn(Me,J))J=Me[J];else if(J===null)if(ne!==null&&"tool_use"in Me)J=Me.tool_use;else if("default"in Me)J=Me.default;else throw Error(`This model has multiple chat templates with no default specified! Please either pass a chat template or the name of the template you wish to use to the 'chat_template' argument. Available template names are ${Object.keys(Me).sort()}.`)}else if(J===null)if(this.chat_template)J=this.chat_template;else throw Error("Cannot use apply_chat_template() because tokenizer.chat_template is not set and no template argument was passed! For information about writing templates and setting the tokenizer.chat_template attribute, please see the documentation at https://huggingface.co/docs/transformers/main/en/chat_templating");return J}apply_chat_template(J,{tools:ne=null,documents:Me=null,chat_template:xe=null,add_generation_prompt:je=!1,tokenize:Ye=!0,padding:ot=!1,truncation:it=!1,max_length:ut=null,return_tensor:It=!0,return_dict:Qt=!1,tokenizer_kwargs:qt={},..._r}={}){if(xe=this.get_chat_template({chat_template:xe,tools:ne}),typeof xe!="string")throw Error(`chat_template must be a string, but got ${typeof xe}`);let tr=this._compiled_template_cache.get(xe);tr===void 0&&(tr=new f.Template(xe),this._compiled_template_cache.set(xe,tr));const Cr=Object.create(null);for(const Zr of ue){const Pi=this.getToken(Zr);Pi&&(Cr[Zr]=Pi)}const kr=tr.render({messages:J,add_generation_prompt:je,tools:ne,documents:Me,...Cr,..._r});if(Ye){const Zr=this._call(kr,{add_special_tokens:!1,padding:ot,truncation:it,max_length:ut,return_tensor:It,...qt});return Qt?Zr:Zr.input_ids}return kr}}class Dt extends De{constructor(){super(...arguments);le(this,"return_token_type_ids",!0)}}class Ut extends De{constructor(){super(...arguments);le(this,"return_token_type_ids",!0)}}class Et extends De{constructor(){super(...arguments);le(this,"return_token_type_ids",!0)}}class jt extends De{constructor(){super(...arguments);le(this,"return_token_type_ids",!0)}}class gt extends De{constructor(){super(...arguments);le(this,"return_token_type_ids",!0)}}class Ht extends De{constructor(){super(...arguments);le(this,"return_token_type_ids",!0)}}class nr extends De{constructor(){super(...arguments);le(this,"return_token_type_ids",!0)}}class ri extends De{constructor(){super(...arguments);le(this,"return_token_type_ids",!0)}}class pr extends De{constructor(){super(...arguments);le(this,"return_token_type_ids",!0)}}class Jr extends De{}class Ai extends De{}class Ji extends De{constructor(J,ne){super(J,ne);le(this,"return_token_type_ids",!0);console.warn('WARNING: `XLMTokenizer` is not yet supported by Hugging Face\'s "fast" tokenizers library. Therefore, you may experience slightly inaccurate results.')}}class Oi extends De{constructor(){super(...arguments);le(this,"return_token_type_ids",!0)}}class ns extends De{}class ii extends De{}class Yi extends De{}class _i extends De{constructor(j,J){super(j,J),this.languageRegex=/^[a-z]{2}_[A-Z]{2}$/,this.language_codes=this.special_tokens.filter(ne=>this.languageRegex.test(ne)),this.lang_to_token=ne=>ne}_build_translation_inputs(j,J,ne){return ni(this,j,J,ne)}}class Fi extends _i{}class er extends De{}class Zi extends De{}const Gr="▁";class Kr extends De{constructor(J,ne){super(J,ne);le(this,"padding_side","left");this.legacy=ne.legacy??!0,this.legacy||(this.normalizer=null,this.pre_tokenizer=new wr({replacement:Gr,add_prefix_space:!0,prepend_scheme:"first"}))}_encode_text(J){if(J===null)return null;if(this.legacy||J.length===0)return super._encode_text(J);let ne=super._encode_text(Gr+J.replaceAll(Gr," "));return ne.length>1&&ne[0]===Gr&&this.special_tokens.includes(ne[1])&&(ne=ne.slice(1)),ne}}class Hr extends De{}class es extends De{}class si extends De{}class qe extends De{}class lt extends De{}class yt extends De{}class ur extends De{}class gi extends De{}class Di extends De{}function ni(ve,j,J,ne){if(!("language_codes"in ve)||!Array.isArray(ve.language_codes))throw new Error("Tokenizer must have `language_codes` attribute set and it should be an array of language ids.");if(!("languageRegex"in ve)||!(ve.languageRegex instanceof RegExp))throw new Error("Tokenizer must have `languageRegex` attribute set and it should be a regular expression.");if(!("lang_to_token"in ve)||typeof ve.lang_to_token!="function")throw new Error("Tokenizer must have `lang_to_token` attribute set and it should be a function.");const Me=ne.src_lang,xe=ne.tgt_lang;if(!ve.language_codes.includes(xe))throw new Error(`Target language code "${xe}" is not valid. Must be one of: {${ve.language_codes.join(", ")}}`);if(Me!==void 0){if(!ve.language_codes.includes(Me))throw new Error(`Source language code "${Me}" is not valid. Must be one of: {${ve.language_codes.join(", ")}}`);for(const je of ve.post_processor.config.single)if("SpecialToken"in je&&ve.languageRegex.test(je.SpecialToken.id)){je.SpecialToken.id=ve.lang_to_token(Me);break}}return ne.forced_bos_token_id=ve.model.convert_tokens_to_ids([ve.lang_to_token(xe)])[0],ve._call(j,J)}class Li extends De{constructor(j,J){super(j,J),this.languageRegex=/^[a-z]{3}_[A-Z][a-z]{3}$/,this.language_codes=this.special_tokens.filter(ne=>this.languageRegex.test(ne)),this.lang_to_token=ne=>ne}_build_translation_inputs(j,J,ne){return ni(this,j,J,ne)}}class ji extends De{constructor(j,J){super(j,J),this.languageRegex=/^__[a-z]{2,3}__$/,this.language_codes=this.special_tokens.filter(ne=>this.languageRegex.test(ne)).map(ne=>ne.slice(2,-2)),this.lang_to_token=ne=>`__${ne}__`}_build_translation_inputs(j,J,ne){return ni(this,j,J,ne)}}class zi extends De{get timestamp_begin(){return this.model.convert_tokens_to_ids(["<|notimestamps|>"])[0]+1}_decode_asr(j,{return_timestamps:J=!1,return_language:ne=!1,time_precision:Me=null,force_full_sequences:xe=!0}={}){if(Me===null)throw Error("Must specify time_precision");let je=null;const Ye=J==="word";function ot(){return{language:je,timestamp:[null,null],text:""}}const it=[];let ut=ot(),It=0;const Qt=this.timestamp_begin,_r=Qt+1500;let tr=[],Cr=[],kr=!1,Zr=null;const Pi=new Set(this.all_special_ids);for(const sr of j){const Ir=sr.tokens,Br=Ye?sr.token_timestamps:null;let yi=null,Ri=Qt;if("stride"in sr){const[Mr,gr,xr]=sr.stride;if(It-=gr,Zr=Mr-xr,gr&&(Ri=gr/Me+Qt),xr)for(let Tr=Ir.length-1;Tr>=0;--Tr){const Nr=Number(Ir[Tr]);if(Nr>=Qt){if(yi!==null&&(Nr-Qt)*Me<Zr)break;yi=Nr}}}let Rr=[],ei=[];for(let Mr=0;Mr<Ir.length;++Mr){const gr=Number(Ir[Mr]);if(Pi.has(gr)){const xr=this.decode([gr]),Tr=y.WHISPER_LANGUAGE_MAPPING.get(xr.slice(2,-2));if(Tr!==void 0){if(je!==null&&Tr!==je&&!J){tr.push(Rr);const Nr=this.findLongestCommonSequence(tr)[0],Dr=this.decode(Nr);ut.text=Dr,it.push(ut),tr=[],Rr=[],ut=ot()}je=ut.language=Tr}}else if(gr>=Qt&&gr<=_r){const xr=(gr-Qt)*Me+It,Tr=(0,u.round)(xr,2);if(yi!==null&&gr>=yi)kr=!0;else if(kr||tr.length>0&&gr<Ri)kr=!1;else if(ut.timestamp[0]===null)ut.timestamp[0]=Tr;else if(Tr!==ut.timestamp[0]){ut.timestamp[1]=Tr,tr.push(Rr),Ye&&Cr.push(ei);const[Nr,Dr]=this.findLongestCommonSequence(tr,Cr),li=this.decode(Nr);ut.text=li,Ye&&(ut.words=this.collateWordTimestamps(Nr,Dr,je)),it.push(ut),tr=[],Rr=[],Cr=[],ei=[],ut=ot()}}else if(Rr.push(gr),Ye){let xr=(0,u.round)(Br[Mr]+It,2),Tr;if(Mr+1<Br.length){Tr=(0,u.round)(Br[Mr+1]+It,2);const Nr=this.decode([gr]);S.test(Nr)&&(Tr=(0,u.round)(Math.min(xr+Me,Tr),2))}else Tr=null;ei.push([xr,Tr])}}if("stride"in sr){const[Mr,gr,xr]=sr.stride;It+=Mr-xr}Rr.length>0?(tr.push(Rr),Ye&&Cr.push(ei)):tr.every(Mr=>Mr.length===0)&&(ut=ot(),tr=[],Rr=[],Cr=[],ei=[])}if(tr.length>0){if(xe&&J)throw new Error("Whisper did not predict an ending timestamp, which can happen if audio is cut off in the middle of a word. Also make sure WhisperTimeStampLogitsProcessor was used during generation.");const[sr,Ir]=this.findLongestCommonSequence(tr,Cr),Br=this.decode(sr);ut.text=Br,Ye&&(ut.words=this.collateWordTimestamps(sr,Ir,je)),it.push(ut)}let Fr=Object.create(null);const Ci=it.map(sr=>sr.text).join("");if(J||ne){for(let sr=0;sr<it.length;++sr){const Ir=it[sr];J||delete Ir.timestamp,ne||delete Ir.language}if(Ye){const sr=[];for(const Ir of it)for(const Br of Ir.words)sr.push(Br);Fr={chunks:sr}}else Fr={chunks:it}}return[Ci,Fr]}findLongestCommonSequence(j,J=null){let ne=j[0],Me=ne.length,xe=[];const je=Array.isArray(J)&&J.length>0;let Ye=je?[]:null,ot=je?J[0]:null;for(let it=1;it<j.length;++it){const ut=j[it];let It=0,Qt=[Me,Me,0,0];const qt=ut.length;for(let Fr=1;Fr<Me+qt;++Fr){const Ci=Math.max(0,Me-Fr),sr=Math.min(Me,Me+qt-Fr),Ir=ne.slice(Ci,sr),Br=Math.max(0,Fr-Me),yi=Math.min(qt,Fr),Ri=ut.slice(Br,yi);if(Ir.length!==Ri.length)throw new Error("There is a bug within whisper `decode_asr` function, please report it. Dropping to prevent bad inference.");let Rr;je?Rr=Ir.filter((gr,xr)=>gr===Ri[xr]&&ot[Ci+xr]<=J[it][Br+xr]).length:Rr=Ir.filter((gr,xr)=>gr===Ri[xr]).length;const ei=Fr/1e4,Mr=Rr/Fr+ei;Rr>1&&Mr>It&&(It=Mr,Qt=[Ci,sr,Br,yi])}const[_r,tr,Cr,kr]=Qt,Zr=Math.floor((tr+_r)/2),Pi=Math.floor((kr+Cr)/2);xe.push(...ne.slice(0,Zr)),ne=ut.slice(Pi),Me=ne.length,je&&(Ye.push(...ot.slice(0,Zr)),ot=J[it].slice(Pi))}return xe.push(...ne),je?(Ye.push(...ot),[xe,Ye]):[xe,[]]}collateWordTimestamps(j,J,ne){const[Me,xe,je]=this.combineTokensIntoWords(j,ne),Ye=[];for(let ot=0;ot<Me.length;++ot){const it=je[ot];Ye.push({text:Me[ot],timestamp:[J[it.at(0)][0],J[it.at(-1)][1]]})}return Ye}combineTokensIntoWords(j,J,ne=`"'“¡¿([{-`,Me=`"'.。,，!！?？:：”)]}、`){J=J??"english";let xe,je,Ye;return["chinese","japanese","thai","lao","myanmar"].includes(J)?[xe,je,Ye]=this.splitTokensOnUnicode(j):[xe,je,Ye]=this.splitTokensOnSpaces(j),this.mergePunctuations(xe,je,Ye,ne,Me)}decode(j,J){let ne;return J!=null&&J.decode_with_timestamps?(j instanceof d.Tensor&&(j=k(j)),ne=this.decodeWithTimestamps(j,J)):ne=super.decode(j,J),ne}decodeWithTimestamps(j,J){const ne=(J==null?void 0:J.time_precision)??.02,Me=Array.from(this.all_special_ids).at(-1)+1;let xe=[[]];for(let je of j)if(je=Number(je),je>=Me){const Ye=((je-Me)*ne).toFixed(2);xe.push(`<|${Ye}|>`),xe.push([])}else xe[xe.length-1].push(je);return xe=xe.map(je=>typeof je=="string"?je:super.decode(je,J)),xe.join("")}splitTokensOnUnicode(j){const J=this.decode(j,{decode_with_timestamps:!0}),ne="�",Me=[],xe=[],je=[];let Ye=[],ot=[],it=0;for(let ut=0;ut<j.length;++ut){const It=j[ut];Ye.push(It),ot.push(ut);const Qt=this.decode(Ye,{decode_with_timestamps:!0});(!Qt.includes(ne)||J[it+Qt.indexOf(ne)]===ne)&&(Me.push(Qt),xe.push(Ye),je.push(ot),Ye=[],ot=[],it+=Qt.length)}return[Me,xe,je]}splitTokensOnSpaces(j){const[J,ne,Me]=this.splitTokensOnUnicode(j),xe=[],je=[],Ye=[],ot=new RegExp(`^[${P}]$`,"gu");for(let it=0;it<J.length;++it){const ut=J[it],It=ne[it],Qt=Me[it],qt=It[0]>=this.model.tokens_to_ids.get("<|endoftext|>"),_r=ut.startsWith(" "),tr=ut.trim(),Cr=ot.test(tr);if(qt||_r||Cr||xe.length===0)xe.push(ut),je.push(It),Ye.push(Qt);else{const kr=xe.length-1;xe[kr]+=ut,je[kr].push(...It),Ye[kr].push(...Qt)}}return[xe,je,Ye]}mergePunctuations(j,J,ne,Me,xe){const je=structuredClone(j),Ye=structuredClone(J),ot=structuredClone(ne);let it=je.length-2,ut=je.length-1;for(;it>=0;)je[it].startsWith(" ")&&Me.includes(je[it].trim())?(je[ut]=je[it]+je[ut],Ye[ut]=(0,o.mergeArrays)(Ye[it],Ye[ut]),ot[ut]=(0,o.mergeArrays)(ot[it],ot[ut]),je[it]="",Ye[it]=[],ot[it]=[]):ut=it,--it;for(it=0,ut=1;ut<je.length;)!je[it].endsWith(" ")&&xe.includes(je[ut])?(je[it]+=je[ut],Ye[it]=(0,o.mergeArrays)(Ye[it],Ye[ut]),ot[it]=(0,o.mergeArrays)(ot[it],ot[ut]),je[ut]="",Ye[ut]=[],ot[ut]=[]):it=ut,++ut;return[je.filter(It=>It),Ye.filter(It=>It.length>0),ot.filter(It=>It.length>0)]}}class qr extends De{}class Bi extends De{}class Yr extends De{}class as extends De{constructor(j,J){super(j,J),this.languageRegex=/^(>>\w+<<)\s*/g,this.supported_language_codes=this.model.vocab.filter(ne=>this.languageRegex.test(ne)),console.warn('WARNING: `MarianTokenizer` is not yet supported by Hugging Face\'s "fast" tokenizers library. Therefore, you may experience slightly inaccurate results.')}_encode_text(j){if(j===null)return null;const[J,...ne]=j.trim().split(this.languageRegex);if(ne.length===0)return super._encode_text(J);if(ne.length===2){const[Me,xe]=ne;return this.supported_language_codes.includes(Me)||console.warn(`Unsupported language code "${Me}" detected, which may lead to unexpected behavior. Should be one of: ${JSON.stringify(this.supported_language_codes)}`),(0,o.mergeArrays)([Me],super._encode_text(xe))}}}class os extends De{}class ai extends De{}class xs extends De{}class ts extends De{}class pi extends De{}class Vs extends De{constructor(j,J){super(j,J),this.decoder=new Kt({})}}class ls extends De{}class oi extends De{}class Ei{static async from_pretrained(j,{progress_callback:J=null,config:ne=null,cache_dir:Me=null,local_files_only:xe=!1,revision:je="main",legacy:Ye=null}={}){var Qt;const[ot,it]=await m(j,{progress_callback:J,config:ne,cache_dir:Me,local_files_only:xe,revision:je,legacy:Ye}),ut=((Qt=it.tokenizer_class)==null?void 0:Qt.replace(/Fast$/,""))??"PreTrainedTokenizer";let It=this.TOKENIZER_CLASS_MAPPING[ut];return It||(console.warn(`Unknown tokenizer class "${ut}", attempting to construct from base class.`),It=De),new It(ot,it)}}le(Ei,"TOKENIZER_CLASS_MAPPING",{T5Tokenizer:ns,DistilBertTokenizer:Jr,CamembertTokenizer:Ai,DebertaTokenizer:gt,DebertaV2Tokenizer:Ht,BertTokenizer:Dt,HerbertTokenizer:nr,ConvBertTokenizer:ri,RoFormerTokenizer:pr,XLMTokenizer:Ji,ElectraTokenizer:Oi,MobileBertTokenizer:Et,SqueezeBertTokenizer:jt,AlbertTokenizer:Ut,GPT2Tokenizer:ii,BartTokenizer:Yi,MBartTokenizer:_i,MBart50Tokenizer:Fi,RobertaTokenizer:er,WhisperTokenizer:zi,CodeGenTokenizer:qr,CLIPTokenizer:Bi,SiglipTokenizer:Yr,MarianTokenizer:as,BloomTokenizer:Zi,NllbTokenizer:Li,M2M100Tokenizer:ji,LlamaTokenizer:Kr,CodeLlamaTokenizer:Hr,XLMRobertaTokenizer:es,MPNetTokenizer:si,FalconTokenizer:qe,GPTNeoXTokenizer:lt,EsmTokenizer:yt,Wav2Vec2CTCTokenizer:os,BlenderbotTokenizer:ai,BlenderbotSmallTokenizer:xs,SpeechT5Tokenizer:ts,NougatTokenizer:pi,VitsTokenizer:Vs,Qwen2Tokenizer:ur,GemmaTokenizer:gi,Grok1Tokenizer:Di,CohereTokenizer:ls,MgpstrTokenizer:oi,PreTrainedTokenizer:De})},"./src/utils/audio.js":(e,r,t)=>{t.r(r),t.d(r,{RawAudio:()=>ie,hamming:()=>g,hanning:()=>m,mel_filter_bank:()=>D,read_audio:()=>f,spectrogram:()=>O,window_function:()=>R});var n=t("./src/utils/hub.js"),o=t("./src/utils/maths.js"),i=t("./src/utils/core.js"),u=t("./src/env.js"),d=t("?7a2c"),h=t("./src/utils/tensor.js");async function f(H,ce){if(typeof AudioContext>"u")throw Error("Unable to load audio from path/URL since `AudioContext` is not available in your environment. Instead, audio data should be passed directly to the pipeline/processor. For more information and some example code, see https://huggingface.co/docs/transformers.js/guides/node-audio-processing.");const re=await(await(0,n.getFile)(H)).arrayBuffer(),se=new AudioContext({sampleRate:ce});typeof ce>"u"&&console.warn(`No sampling rate provided, using default of ${se.sampleRate}Hz.`);const _e=await se.decodeAudioData(re);let ae;if(_e.numberOfChannels===2){const Ce=Math.sqrt(2),Te=_e.getChannelData(0),q=_e.getChannelData(1);ae=new Float32Array(Te.length);for(let B=0;B<_e.length;++B)ae[B]=Ce*(Te[B]+q[B])/2}else ae=_e.getChannelData(0);return ae}function y(H,ce){if(H<1)return new Float64Array;if(H===1)return new Float64Array([1]);const re=1-ce,se=2*Math.PI/(H-1),_e=new Float64Array(H);for(let ae=0;ae<H;++ae)_e[ae]=ce-re*Math.cos(ae*se);return _e}function m(H){return y(H,.5)}function g(H){return y(H,.54)}const v={htk:H=>2595*Math.log10(1+H/700),kaldi:H=>1127*Math.log(1+H/700),slaney:(H,ce=1e3,re=15,se=27/Math.log(6.4))=>H>=ce?re+Math.log(H/ce)*se:3*H/200};function b(H,ce="htk"){const re=v[ce];if(!re)throw new Error('mel_scale should be one of "htk", "slaney" or "kaldi".');return typeof H=="number"?re(H):H.map(se=>re(se))}const k={htk:H=>700*(10**(H/2595)-1),kaldi:H=>700*(Math.exp(H/1127)-1),slaney:(H,ce=1e3,re=15,se=Math.log(6.4)/27)=>H>=re?ce*Math.exp(se*(H-re)):200*H/3};function z(H,ce="htk"){const re=k[ce];if(!re)throw new Error('mel_scale should be one of "htk", "slaney" or "kaldi".');return typeof H=="number"?re(H):H.map(se=>re(se))}function E(H,ce){const re=Float64Array.from({length:ce.length-1},(Ce,Te)=>ce[Te+1]-ce[Te]),se=Array.from({length:H.length},()=>new Array(ce.length));for(let Ce=0;Ce<H.length;++Ce){const Te=se[Ce];for(let q=0;q<ce.length;++q)Te[q]=ce[q]-H[Ce]}const _e=ce.length-2,ae=Array.from({length:_e},()=>new Array(H.length));for(let Ce=0;Ce<H.length;++Ce){const Te=se[Ce];for(let q=0;q<_e;++q){const B=-Te[q]/re[q],Q=Te[q+2]/re[q+1];ae[q][Ce]=Math.max(0,Math.min(B,Q))}}return ae}function T(H,ce,re){const se=(ce-H)/(re-1);return Float64Array.from({length:re},(_e,ae)=>H+se*ae)}function D(H,ce,re,se,_e,ae=null,Ce="htk",Te=!1){if(ae!==null&&ae!=="slaney")throw new Error('norm must be one of null or "slaney"');if(H<2)throw new Error(`Require num_frequency_bins: ${H} >= 2`);if(re>se)throw new Error(`Require min_frequency: ${re} <= max_frequency: ${se}`);const q=b(re,Ce),B=b(se,Ce),Q=T(q,B,ce+2);let oe=z(Q,Ce),Ie;if(Te){const be=_e/((H-1)*2);Ie=b(Float64Array.from({length:H},(et,tt)=>tt*be),Ce),oe=Q}else Ie=T(0,Math.floor(_e/2),H);const ye=E(Ie,oe);if(ae!==null&&ae==="slaney")for(let be=0;be<ce;++be){const et=ye[be],tt=2/(oe[be+2]-oe[be]);for(let Ee=0;Ee<H;++Ee)et[Ee]*=tt}return ye}function I(H,ce,re){const se=new H.constructor(H.length+ce+re),_e=H.length-1;for(let ae=0;ae<H.length;++ae)se[ce+ae]=H[ae];for(let ae=1;ae<=ce;++ae)se[ce-ae]=H[(0,i.calculateReflectOffset)(ae,_e)];for(let ae=1;ae<=re;++ae)se[_e+ce+ae]=H[(0,i.calculateReflectOffset)(_e-ae,_e)];return se}function $(H,ce,re,se,_e){if(re<=0)throw new Error("reference must be greater than zero");if(se<=0)throw new Error("min_value must be greater than zero");re=Math.max(se,re);const ae=Math.log10(re);for(let Ce=0;Ce<H.length;++Ce)H[Ce]=ce*Math.log10(Math.max(se,H[Ce])-ae);if(_e!==null){if(_e<=0)throw new Error("db_range must be greater than zero");const Ce=(0,o.max)(H)[0]-_e;for(let Te=0;Te<H.length;++Te)H[Te]=Math.max(H[Te],Ce)}return H}function P(H,ce=1,re=1e-5,se=null){return $(H,20,ce,re,se)}function S(H,ce=1,re=1e-10,se=null){return $(H,10,ce,re,se)}async function O(H,ce,re,se,{fft_length:_e=null,power:ae=1,center:Ce=!0,pad_mode:Te="reflect",onesided:q=!0,preemphasis:B=null,mel_filters:Q=null,mel_floor:oe=1e-10,log_mel:Ie=null,reference:ye=1,min_value:be=1e-10,db_range:et=null,remove_dc_offset:tt=null,min_num_frames:Ee=null,max_num_frames:Y=null,do_pad:pe=!0,transpose:X=!1}={}){const Ae=ce.length;if(_e===null&&(_e=re),re>_e)throw Error(`frame_length (${re}) may not be larger than fft_length (${_e})`);if(Ae!==re)throw new Error(`Length of the window (${Ae}) must equal frame_length (${re})`);if(se<=0)throw new Error("hop_length must be greater than zero");if(ae===null&&Q!==null)throw new Error("You have provided `mel_filters` but `power` is `None`. Mel spectrogram computation is not yet supported for complex-valued spectrogram. Specify `power` to fix this issue.");if(Ce){if(Te!=="reflect")throw new Error(`pad_mode="${Te}" not implemented yet.`);const _t=Math.floor((_e-1)/2)+1;H=I(H,_t,_t)}let Oe=Math.floor(1+Math.floor((H.length-re)/se));Ee!==null&&Oe<Ee&&(Oe=Ee);const Ne=q?Math.floor(_e/2)+1:_e;let Be=Oe,Ve=Oe;Y!==null&&(Y>Oe?pe&&(Ve=Y):Ve=Be=Y);const He=new o.FFT(_e),Xe=new Float64Array(_e),st=new Float64Array(He.outputBufferSize),nt=new Float32Array(Ne*Ve);for(let _t=0;_t<Be;++_t){const ct=_t*se,dt=Math.min(H.length-ct,re);dt!==re&&Xe.fill(0,0,re);for(let Le=0;Le<dt;++Le)Xe[Le]=H[ct+Le];if(tt){let Le=0;for(let Nt=0;Nt<dt;++Nt)Le+=Xe[Nt];const xt=Le/dt;for(let Nt=0;Nt<dt;++Nt)Xe[Nt]-=xt}if(B!==null){for(let Le=dt-1;Le>=1;--Le)Xe[Le]-=B*Xe[Le-1];Xe[0]*=1-B}for(let Le=0;Le<ce.length;++Le)Xe[Le]*=ce[Le];He.realTransform(st,Xe);for(let Le=0;Le<Ne;++Le){const xt=Le<<1;nt[Le*Ve+_t]=st[xt]**2+st[xt+1]**2}}if(ae!==null&&ae!==2){const _t=2/ae;for(let ct=0;ct<nt.length;++ct)nt[ct]**=_t}const We=Q.length;let rt=await(0,h.matmul)(new h.Tensor("float32",Q.flat(),[We,Ne]),new h.Tensor("float32",nt,[Ne,Ve]));X&&(rt=rt.transpose(1,0));const pt=rt.data;for(let _t=0;_t<pt.length;++_t)pt[_t]=Math.max(oe,pt[_t]);if(ae!==null&&Ie!==null){const _t=Math.min(pt.length,Be*We);switch(Ie){case"log":for(let ct=0;ct<_t;++ct)pt[ct]=Math.log(pt[ct]);break;case"log10":for(let ct=0;ct<_t;++ct)pt[ct]=Math.log10(pt[ct]);break;case"dB":if(ae===1)P(pt,ye,be,et);else if(ae===2)S(pt,ye,be,et);else throw new Error(`Cannot use log_mel option '${Ie}' with power ${ae}`);break;default:throw new Error(`log_mel must be one of null, 'log', 'log10' or 'dB'. Got '${Ie}'`)}}return rt}function R(H,ce,{periodic:re=!0,frame_length:se=null,center:_e=!0}={}){const ae=re?H+1:H;let Ce;switch(ce){case"boxcar":Ce=new Float64Array(ae).fill(1);break;case"hann":case"hann_window":Ce=m(ae);break;case"hamming":Ce=g(ae);break;case"povey":Ce=m(ae).map(Te=>Math.pow(Te,.85));break;default:throw new Error(`Unknown window type ${ce}.`)}if(re&&(Ce=Ce.subarray(0,H)),se===null)return Ce;if(H>se)throw new Error(`Length of the window (${H}) may not be larger than frame_length (${se})`);return Ce}function G(H,ce){let re=44;const se=new ArrayBuffer(re+H.length*4),_e=new DataView(se);ee(_e,0,"RIFF"),_e.setUint32(4,36+H.length*4,!0),ee(_e,8,"WAVE"),ee(_e,12,"fmt "),_e.setUint32(16,16,!0),_e.setUint16(20,3,!0),_e.setUint16(22,1,!0),_e.setUint32(24,ce,!0),_e.setUint32(28,ce*4,!0),_e.setUint16(32,4,!0),_e.setUint16(34,32,!0),ee(_e,36,"data"),_e.setUint32(40,H.length*4,!0);for(let ae=0;ae<H.length;++ae,re+=4)_e.setFloat32(re,H[ae],!0);return se}function ee(H,ce,re){for(let se=0;se<re.length;++se)H.setUint8(ce+se,re.charCodeAt(se))}class ie{constructor(ce,re){this.audio=ce,this.sampling_rate=re}toWav(){return G(this.audio,this.sampling_rate)}toBlob(){const ce=this.toWav();return new Blob([ce],{type:"audio/wav"})}async save(ce){let re;if(u.apis.IS_BROWSER_ENV){if(u.apis.IS_WEBWORKER_ENV)throw new Error("Unable to save a file from a Web Worker.");re=i.saveBlob}else if(u.apis.IS_FS_AVAILABLE)re=async(se,_e)=>{let ae=await _e.arrayBuffer();d.writeFileSync(se,Buffer.from(ae))};else throw new Error("Unable to save because filesystem is disabled in this environment.");await re(ce,this.toBlob())}}},"./src/utils/constants.js":(e,r,t)=>{t.r(r),t.d(r,{CHAT_TEMPLATE_NAME:()=>h,CONFIG_NAME:()=>o,FEATURE_EXTRACTOR_NAME:()=>i,GENERATION_CONFIG_NAME:()=>f,GITHUB_ISSUE_URL:()=>n,IMAGE_PROCESSOR_NAME:()=>u,PROCESSOR_NAME:()=>d});const n="https://github.com/huggingface/transformers.js/issues/new/choose",o="config.json",i="preprocessor_config.json",u=i,d="processor_config.json",h="chat_template.json",f="generation_config.json"},"./src/utils/core.js":(e,r,t)=>{t.r(r),t.d(r,{calculateDimensions:()=>f,calculateReflectOffset:()=>v,count:()=>E,dispatchCallback:()=>n,escapeRegExp:()=>i,isIntegralNumber:()=>d,isNullishDimension:()=>h,isTypedArray:()=>u,len:()=>z,mergeArrays:()=>m,pick:()=>k,pop:()=>y,product:()=>g,reverseDictionary:()=>o,saveBlob:()=>b});function n(T,D){T&&T(D)}function o(T){return Object.fromEntries(Object.entries(T).map(([D,I])=>[I,D]))}function i(T){return T.replace(/[.*+?^${}()|[\]\\]/g,"\\$&")}function u(T){var D,I,$;return(($=(I=(D=T==null?void 0:T.prototype)==null?void 0:D.__proto__)==null?void 0:I.constructor)==null?void 0:$.name)==="TypedArray"}function d(T){return Number.isInteger(T)||typeof T=="bigint"}function h(T){return T==null||T===-1}function f(T){const D=[];let I=T;for(;Array.isArray(I);)D.push(I.length),I=I[0];return D}function y(T,D,I=void 0){const $=T[D];if($!==void 0)return delete T[D],$;if(I===void 0)throw Error(`Key ${D} does not exist in object.`);return I}function m(...T){return Array.prototype.concat.apply([],T)}function g(...T){return T.reduce((D,I)=>D.flatMap($=>I.map(P=>[$,P])))}function v(T,D){return Math.abs((T+D)%(2*D)-D)}function b(T,D){const I=URL.createObjectURL(D),$=document.createElement("a");$.href=I,$.download=T,$.click(),$.remove(),URL.revokeObjectURL(I)}function k(T,D){return Object.assign({},...D.map(I=>{if(T[I]!==void 0)return{[I]:T[I]}}))}function z(T){let D=0;for(const I of T)++D;return D}function E(T,D){let I=0;for(const $ of T)$===D&&++I;return I}},"./src/utils/data-structures.js":(e,r,t)=>{t.r(r),t.d(r,{CharTrie:()=>o,DictionarySplitter:()=>h,LRUCache:()=>f,PriorityQueue:()=>n,TokenLattice:()=>u});class n{constructor(m=(v,b)=>v>b,g=1/0){this._heap=[],this._comparator=m,this._maxSize=g}get size(){return this._heap.length}isEmpty(){return this.size===0}peek(){return this._heap[0]}push(...m){return this.extend(m)}extend(m){for(const g of m)if(this.size<this._maxSize)this._heap.push(g),this._siftUp();else{const v=this._smallest();this._comparator(g,this._heap[v])&&(this._heap[v]=g,this._siftUpFrom(v))}return this.size}pop(){const m=this.peek(),g=this.size-1;return g>0&&this._swap(0,g),this._heap.pop(),this._siftDown(),m}replace(m){const g=this.peek();return this._heap[0]=m,this._siftDown(),g}_parent(m){return(m+1>>>1)-1}_left(m){return(m<<1)+1}_right(m){return m+1<<1}_greater(m,g){return this._comparator(this._heap[m],this._heap[g])}_swap(m,g){const v=this._heap[m];this._heap[m]=this._heap[g],this._heap[g]=v}_siftUp(){this._siftUpFrom(this.size-1)}_siftUpFrom(m){for(;m>0&&this._greater(m,this._parent(m));)this._swap(m,this._parent(m)),m=this._parent(m)}_siftDown(){let m=0;for(;this._left(m)<this.size&&this._greater(this._left(m),m)||this._right(m)<this.size&&this._greater(this._right(m),m);){const g=this._right(m)<this.size&&this._greater(this._right(m),this._left(m))?this._right(m):this._left(m);this._swap(m,g),m=g}}_smallest(){return 2**Math.floor(Math.log2(this.size))-1}}class o{constructor(){this.root=i.default()}extend(m){for(const g of m)this.push(g)}push(m){let g=this.root;for(const v of m){let b=g.children.get(v);b===void 0&&(b=i.default(),g.children.set(v,b)),g=b}g.isLeaf=!0}*commonPrefixSearch(m){let g=this.root;if(g===void 0)return;let v="";for(const b of m){if(v+=b,g=g.children.get(b),g===void 0)return;g.isLeaf&&(yield v)}}}class i{constructor(m,g){this.isLeaf=m,this.children=g}static default(){return new i(!1,new Map)}}class u{constructor(m,g,v){this.chars=Array.from(m),this.len=this.chars.length,this.bosTokenId=g,this.eosTokenId=v,this.nodes=[],this.beginNodes=Array.from({length:this.len+1},()=>[]),this.endNodes=Array.from({length:this.len+1},()=>[]);const b=new d(this.bosTokenId,0,0,0,0),k=new d(this.eosTokenId,1,this.len,0,0);this.nodes.push(b.clone()),this.nodes.push(k.clone()),this.beginNodes[this.len].push(k),this.endNodes[0].push(b)}insert(m,g,v,b){const k=this.nodes.length,z=new d(b,k,m,g,v);this.beginNodes[m].push(z),this.endNodes[m+g].push(z),this.nodes.push(z)}viterbi(){const m=this.len;let g=0;for(;g<=m;){if(this.beginNodes[g].length==0)return[];for(let E of this.beginNodes[g]){E.prev=null;let T=0,D=null;for(let I of this.endNodes[g]){const $=I.backtraceScore+E.score;(D===null||$>T)&&(D=I.clone(),T=$)}if(D!==null)E.prev=D,E.backtraceScore=T;else return[]}++g}const v=[],k=this.beginNodes[m][0].prev;if(k===null)return[];let z=k.clone();for(;z.prev!==null;)v.push(z.clone()),z=z.clone().prev.clone();return v.reverse(),v}piece(m){return this.chars.slice(m.pos,m.pos+m.length).join("")}tokens(){return this.viterbi().map(g=>this.piece(g))}tokenIds(){return this.viterbi().map(g=>g.tokenId)}}class d{constructor(m,g,v,b,k){this.tokenId=m,this.nodeId=g,this.pos=v,this.length=b,this.score=k,this.prev=null,this.backtraceScore=0}clone(){const m=new d(this.tokenId,this.nodeId,this.pos,this.length,this.score);return m.prev=this.prev,m.backtraceScore=this.backtraceScore,m}}class h{constructor(m){this.trie=this._buildTrie(m)}_buildTrie(m){var v;const g=Object.create(null);for(const b of m){let k=g;for(let z=0;z<b.length;++z)k=k[v=b[z]]??(k[v]=Object.create(null));k.end=b}return g}split(m){const g=[],v=m.length;let b=0,k=0;for(;k<v;){let z=this.trie,E=null,T=k;for(;T<v&&(z=z[m[T]]);)z.end&&(E=z.end),++T;E?(k>b&&g.push(m.slice(b,k)),g.push(E),k+=E.length,b=k):++k}return b<v&&g.push(m.slice(b)),g}}class f{constructor(m){this.capacity=m,this.cache=new Map}get(m){if(!this.cache.has(m))return;const g=this.cache.get(m);return this.cache.delete(m),this.cache.set(m,g),g}put(m,g){this.cache.has(m)&&this.cache.delete(m),this.cache.set(m,g),this.cache.size>this.capacity&&this.cache.delete(this.cache.keys().next().value)}clear(){this.cache.clear()}}},"./src/utils/devices.js":(e,r,t)=>{t.r(r),t.d(r,{DEVICE_TYPES:()=>n});const n=Object.freeze({auto:"auto",gpu:"gpu",cpu:"cpu",wasm:"wasm",webgpu:"webgpu",cuda:"cuda",dml:"dml",webnn:"webnn","webnn-npu":"webnn-npu","webnn-gpu":"webnn-gpu","webnn-cpu":"webnn-cpu"})},"./src/utils/dtypes.js":(e,r,t)=>{t.r(r),t.d(r,{DATA_TYPES:()=>u,DEFAULT_DEVICE_DTYPE_MAPPING:()=>d,DEFAULT_DTYPE_SUFFIX_MAPPING:()=>h,isWebGpuFp16Supported:()=>i});var n=t("./src/env.js"),o=t("./src/utils/devices.js");const i=function(){let f;return async function(){if(f===void 0)if(!n.apis.IS_WEBGPU_AVAILABLE)f=!1;else try{f=(await navigator.gpu.requestAdapter()).features.has("shader-f16")}catch{f=!1}return f}}(),u=Object.freeze({auto:"auto",fp32:"fp32",fp16:"fp16",q8:"q8",int8:"int8",uint8:"uint8",q4:"q4",bnb4:"bnb4",q4f16:"q4f16"}),d=Object.freeze({[o.DEVICE_TYPES.wasm]:u.q8}),h=Object.freeze({[u.fp32]:"",[u.fp16]:"_fp16",[u.int8]:"_int8",[u.uint8]:"_uint8",[u.q8]:"_quantized",[u.q4]:"_q4",[u.q4f16]:"_q4f16",[u.bnb4]:"_bnb4"})},"./src/utils/generic.js":(e,r,t)=>{t.r(r),t.d(r,{Callable:()=>n});const n=class{constructor(){let o=function(...i){return o._call(...i)};return Object.setPrototypeOf(o,new.target.prototype)}_call(...o){throw Error("Must implement _call method in subclass")}}},"./src/utils/hub.js":(e,r,t)=>{t.r(r),t.d(r,{MAX_EXTERNAL_DATA_CHUNKS:()=>d,getFile:()=>v,getModelFile:()=>T,getModelJSON:()=>D});var n=t("?7a2c"),o=t("?a42a"),i=t("./src/env.js"),u=t("./src/utils/core.js");const d=100,h={txt:"text/plain",html:"text/html",css:"text/css",js:"text/javascript",json:"application/json",png:"image/png",jpg:"image/jpeg",jpeg:"image/jpeg",gif:"image/gif"};class f{constructor(S){if(this.filePath=S,this.headers=new Headers,this.exists=n.existsSync(S),this.exists){this.status=200,this.statusText="OK";let O=n.statSync(S);this.headers.set("content-length",O.size.toString()),this.updateContentType();const R=n.createReadStream(S);this.body=new ReadableStream({start(G){R.on("data",ee=>G.enqueue(ee)),R.on("end",()=>G.close()),R.on("error",ee=>G.error(ee))},cancel(){R.destroy()}})}else this.status=404,this.statusText="Not Found",this.body=null}updateContentType(){const S=this.filePath.toString().split(".").pop().toLowerCase();this.headers.set("content-type",h[S]??"application/octet-stream")}clone(){let S=new f(this.filePath);return S.exists=this.exists,S.status=this.status,S.statusText=this.statusText,S.headers=new Headers(this.headers),S}async arrayBuffer(){return(await n.promises.readFile(this.filePath)).buffer}async blob(){const S=await n.promises.readFile(this.filePath);return new Blob([S],{type:this.headers.get("content-type")})}async text(){return await n.promises.readFile(this.filePath,"utf8")}async json(){return JSON.parse(await this.text())}}function y(P,S=null,O=null){let R;try{R=new URL(P)}catch{return!1}return!(S&&!S.includes(R.protocol)||O&&!O.includes(R.hostname))}const m=/^(\b[\w\-.]+\b\/)?\b[\w\-.]{1,96}\b$/;function g(P){return!(!m.test(P)||P.includes("..")||P.includes("--")||P.endsWith(".git")||P.endsWith(".ipynb"))}async function v(P){var S;if(i.env.useFS&&!y(P,["http:","https:","blob:"]))return new f(P instanceof URL?P.protocol==="file:"?P.pathname:P.toString():P);if(typeof process<"u"&&((S=process==null?void 0:process.release)==null?void 0:S.name)==="node"){const O=!!(Ms!=null&&Ms.TESTING_REMOTELY),R=i.env.version,G=new Headers;if(G.set("User-Agent",`transformers.js/${R}; is_ci/${O};`),y(P,["http:","https:"],["huggingface.co","hf.co"])){const ie=(Ms==null?void 0:Ms.HF_TOKEN)??(Ms==null?void 0:Ms.HF_ACCESS_TOKEN);ie&&G.set("Authorization",`Bearer ${ie}`)}return fetch(P,{headers:G})}else return fetch(P)}const b={400:"Bad request error occurred while trying to load file",401:"Unauthorized access to file",403:"Forbidden access to file",404:"Could not locate file",408:"Request timeout error occurred while trying to load file",500:"Internal server error error occurred while trying to load file",502:"Bad gateway error occurred while trying to load file",503:"Service unavailable error occurred while trying to load file",504:"Gateway timeout error occurred while trying to load file"};function k(P,S,O){if(!O)return null;const R=b[P]??`Error (${P}) occurred while trying to load file`;throw Error(`${R}: "${S}".`)}class z{constructor(S){this.path=S}async match(S){let O=o.join(this.path,S),R=new f(O);if(R.exists)return R}async put(S,O,R=void 0){let G=o.join(this.path,S);try{const ee=O.headers.get("Content-Length"),ie=parseInt(ee??"0");let H=0;await n.promises.mkdir(o.dirname(G),{recursive:!0});const ce=n.createWriteStream(G),re=O.body.getReader();for(;;){const{done:se,value:_e}=await re.read();if(se)break;await new Promise((Ce,Te)=>{ce.write(_e,q=>{if(q){Te(q);return}Ce()})}),H+=_e.length;const ae=ie?H/ie*100:0;R==null||R({progress:ae,loaded:H,total:ie})}ce.close()}catch(ee){try{await n.promises.unlink(G)}catch{}throw ee}}}async function E(P,...S){for(let O of S)try{let R=await P.match(O);if(R)return R}catch{continue}}async function T(P,S,O=!0,R={},G=!1){if(!i.env.allowLocalModels){if(R.local_files_only)throw Error("Invalid configuration detected: local models are disabled (`env.allowLocalModels=false`) but you have requested to only use local models (`local_files_only=true`).");if(!i.env.allowRemoteModels)throw Error("Invalid configuration detected: both local and remote models are disabled. Fix by setting `env.allowLocalModels` or `env.allowRemoteModels` to `true`.")}(0,u.dispatchCallback)(R.progress_callback,{status:"initiate",name:P,file:S});let ee;if(!ee&&i.env.useCustomCache){if(!i.env.customCache)throw Error("`env.useCustomCache=true`, but `env.customCache` is not defined.");if(!i.env.customCache.match||!i.env.customCache.put)throw new Error("`env.customCache` must be an object which implements the `match` and `put` functions of the Web Cache API. For more information, see https://developer.mozilla.org/en-US/docs/Web/API/Cache");ee=i.env.customCache}if(!ee&&i.env.useBrowserCache){if(typeof caches>"u")throw Error("Browser cache is not available in this environment.");try{ee=await caches.open("transformers-cache")}catch(oe){console.warn("An error occurred while opening the browser cache:",oe)}}if(!ee&&i.env.useFSCache){if(!i.apis.IS_FS_AVAILABLE)throw Error("File System Cache is not available in this environment.");ee=new z(R.cache_dir??i.env.cacheDir)}const ie=R.revision??"main",H=$(P,S),ce=g(P),re=ce?$(i.env.localModelPath,H):H,se=$(i.env.remoteHost,i.env.remotePathTemplate.replaceAll("{model}",P).replaceAll("{revision}",encodeURIComponent(ie)),S);let _e;const ae=ee instanceof z?ie==="main"?H:$(P,ie,S):se;let Ce=!1,Te;ee&&(Te=await E(ee,re,ae));const q=Te!==void 0;if(Te===void 0){if(i.env.allowLocalModels)if(y(H,["http:","https:"])){if(R.local_files_only)throw new Error(`\`local_files_only=true\`, but attempted to load a remote file from: ${H}.`);if(!i.env.allowRemoteModels)throw new Error(`\`env.allowRemoteModels=false\`, but attempted to load a remote file from: ${H}.`)}else try{Te=await v(re),_e=re}catch(Ie){console.warn(`Unable to load from local path "${re}": "${Ie}"`)}if(Te===void 0||Te.status===404){if(R.local_files_only||!i.env.allowRemoteModels){if(O)throw Error(`\`local_files_only=true\` or \`env.allowRemoteModels=false\` and file was not found locally at "${re}".`);return null}if(!ce)throw Error(`Local file missing at "${re}" and download aborted due to invalid model ID "${P}".`);if(Te=await v(se),Te.status!==200)return k(Te.status,se,O);_e=ae}Ce=ee&&typeof Response<"u"&&Te instanceof Response&&Te.status===200}(0,u.dispatchCallback)(R.progress_callback,{status:"download",name:P,file:S});let B;if(!(i.apis.IS_NODE_ENV&&G)){let oe;R.progress_callback?q&&typeof navigator<"u"&&/firefox/i.test(navigator.userAgent)?(oe=new Uint8Array(await Te.arrayBuffer()),(0,u.dispatchCallback)(R.progress_callback,{status:"progress",name:P,file:S,progress:100,loaded:oe.length,total:oe.length})):oe=await I(Te,Ie=>{(0,u.dispatchCallback)(R.progress_callback,{status:"progress",name:P,file:S,...Ie})}):oe=new Uint8Array(await Te.arrayBuffer()),B=oe}if(Ce&&_e&&await ee.match(_e)===void 0&&(B?await ee.put(_e,new Response(B,{headers:Te.headers})).catch(oe=>{console.warn(`Unable to add response to browser cache: ${oe}.`)}):await ee.put(_e,Te,R.progress_callback)),(0,u.dispatchCallback)(R.progress_callback,{status:"done",name:P,file:S}),B){if(!i.apis.IS_NODE_ENV&&G)throw new Error("Cannot return path in a browser environment.");return B}if(Te instanceof f)return Te.filePath;const Q=await(ee==null?void 0:ee.match(_e));if(Q instanceof f)return Q.filePath;if(Q instanceof Response)return new Uint8Array(await Q.arrayBuffer());if(typeof Q=="string")return Q;throw new Error("Unable to get model file path or buffer.")}async function D(P,S,O=!0,R={}){const G=await T(P,S,O,R,!1);if(G===null)return{};const ie=new TextDecoder("utf-8").decode(G);return JSON.parse(ie)}async function I(P,S){const O=P.headers.get("Content-Length");O===null&&console.warn("Unable to determine content-length from response headers. Will expand buffer when needed.");let R=parseInt(O??"0"),G=new Uint8Array(R),ee=0;const ie=P.body.getReader();async function H(){const{done:ce,value:re}=await ie.read();if(ce)return;const se=ee+re.length;if(se>R){R=se;const ae=new Uint8Array(R);ae.set(G),G=ae}G.set(re,ee),ee=se;const _e=ee/R*100;return S({progress:_e,loaded:ee,total:R}),H()}return await H(),G}function $(...P){return P=P.map((S,O)=>(O&&(S=S.replace(new RegExp("^/"),"")),O!==P.length-1&&(S=S.replace(new RegExp("/$"),"")),S)),P.join("/")}},"./src/utils/image.js":(e,r,t)=>{t.r(r),t.d(r,{RawImage:()=>b,load_image:()=>k});var n=t("./src/utils/core.js"),o=t("./src/utils/hub.js"),i=t("./src/env.js"),u=t("./src/utils/tensor.js"),d=t("?2b25");let h,f,y;const m=i.apis.IS_BROWSER_ENV||i.apis.IS_WEBWORKER_ENV;if(m)h=(z,E)=>{if(!self.OffscreenCanvas)throw new Error("OffscreenCanvas not supported by this browser.");return new self.OffscreenCanvas(z,E)},y=self.createImageBitmap,f=self.ImageData;else if(d)y=async z=>{const T=(await z.metadata()).channels,{data:D,info:I}=await z.rotate().raw().toBuffer({resolveWithObject:!0}),$=new b(new Uint8ClampedArray(D),I.width,I.height,I.channels);return T!==void 0&&T!==I.channels&&$.convert(T),$};else throw new Error("Unable to load image processing library.");const g={0:"nearest",1:"lanczos",2:"bilinear",3:"bicubic",4:"box",5:"hamming"},v=new Map([["png","image/png"],["jpg","image/jpeg"],["jpeg","image/jpeg"],["gif","image/gif"]]);class b{constructor(E,T,D,I){this.data=E,this.width=T,this.height=D,this.channels=I}get size(){return[this.width,this.height]}static async read(E){if(E instanceof b)return E;if(typeof E=="string"||E instanceof URL)return await this.fromURL(E);if(E instanceof Blob)return await this.fromBlob(E);if(typeof HTMLCanvasElement<"u"&&E instanceof HTMLCanvasElement||typeof OffscreenCanvas<"u"&&E instanceof OffscreenCanvas)return this.fromCanvas(E);throw new Error(`Unsupported input type: ${typeof E}`)}static fromCanvas(E){if(!m)throw new Error("fromCanvas() is only supported in browser environments.");const D=E.getContext("2d").getImageData(0,0,E.width,E.height).data;return new b(D,E.width,E.height,4)}static async fromURL(E){const T=await(0,o.getFile)(E);if(T.status!==200)throw new Error(`Unable to read image from "${E}" (${T.status} ${T.statusText})`);const D=await T.blob();return this.fromBlob(D)}static async fromBlob(E){if(m){const T=await y(E),D=h(T.width,T.height).getContext("2d");return D.drawImage(T,0,0),new this(D.getImageData(0,0,T.width,T.height).data,T.width,T.height,4)}else{const T=d(await E.arrayBuffer());return await y(T)}}static fromTensor(E,T="CHW"){if(E.dims.length!==3)throw new Error(`Tensor should have 3 dimensions, but has ${E.dims.length} dimensions.`);if(T==="CHW")E=E.transpose(1,2,0);else if(T!=="HWC")throw new Error(`Unsupported channel format: ${T}`);if(!(E.data instanceof Uint8ClampedArray||E.data instanceof Uint8Array))throw new Error(`Unsupported tensor type: ${E.type}`);switch(E.dims[2]){case 1:case 2:case 3:case 4:return new b(E.data,E.dims[1],E.dims[0],E.dims[2]);default:throw new Error(`Unsupported number of channels: ${E.dims[2]}`)}}grayscale(){if(this.channels===1)return this;const E=new Uint8ClampedArray(this.width*this.height*1);switch(this.channels){case 3:case 4:for(let T=0,D=0;T<this.data.length;T+=this.channels){const I=this.data[T],$=this.data[T+1],P=this.data[T+2];E[D++]=Math.round(.2989*I+.587*$+.114*P)}break;default:throw new Error(`Conversion failed due to unsupported number of channels: ${this.channels}`)}return this._update(E,this.width,this.height,1)}rgb(){if(this.channels===3)return this;const E=new Uint8ClampedArray(this.width*this.height*3);switch(this.channels){case 1:for(let T=0,D=0;T<this.data.length;++T)E[D++]=this.data[T],E[D++]=this.data[T],E[D++]=this.data[T];break;case 4:for(let T=0,D=0;T<this.data.length;T+=4)E[D++]=this.data[T],E[D++]=this.data[T+1],E[D++]=this.data[T+2];break;default:throw new Error(`Conversion failed due to unsupported number of channels: ${this.channels}`)}return this._update(E,this.width,this.height,3)}rgba(){if(this.channels===4)return this;const E=new Uint8ClampedArray(this.width*this.height*4);switch(this.channels){case 1:for(let T=0,D=0;T<this.data.length;++T)E[D++]=this.data[T],E[D++]=this.data[T],E[D++]=this.data[T],E[D++]=255;break;case 3:for(let T=0,D=0;T<this.data.length;T+=3)E[D++]=this.data[T],E[D++]=this.data[T+1],E[D++]=this.data[T+2],E[D++]=255;break;default:throw new Error(`Conversion failed due to unsupported number of channels: ${this.channels}`)}return this._update(E,this.width,this.height,4)}putAlpha(E){if(E.width!==this.width||E.height!==this.height)throw new Error(`Expected mask size to be ${this.width}x${this.height}, but got ${E.width}x${E.height}`);if(E.channels!==1)throw new Error(`Expected mask to have 1 channel, but got ${E.channels}`);const T=this.data,D=E.data,I=this.width*this.height;if(this.channels===3){const $=new Uint8ClampedArray(I*4);for(let P=0,S=0,O=0;P<I;++P)$[O++]=T[S++],$[O++]=T[S++],$[O++]=T[S++],$[O++]=D[P];return this._update($,this.width,this.height,4)}else if(this.channels===4){for(let $=0;$<I;++$)T[4*$+3]=D[$];return this}throw new Error(`Expected image to have 3 or 4 channels, but got ${this.channels}`)}async resize(E,T,{resample:D=2}={}){if(this.width===E&&this.height===T)return this;let I=g[D]??D;const $=(0,n.isNullishDimension)(E),P=(0,n.isNullishDimension)(T);if($&&P)return this;if($?E=T/this.height*this.width:P&&(T=E/this.width*this.height),m){const S=this.channels,O=this.toCanvas(),R=h(E,T).getContext("2d");return R.drawImage(O,0,0,E,T),new b(R.getImageData(0,0,E,T).data,E,T,4).convert(S)}else{let S=this.toSharp();switch(I){case"box":case"hamming":(I==="box"||I==="hamming")&&(console.warn(`Resampling method ${I} is not yet supported. Using bilinear instead.`),I="bilinear");case"nearest":case"bilinear":case"bicubic":S=S.affine([E/this.width,0,0,T/this.height],{interpolator:I});break;case"lanczos":S=S.resize({width:E,height:T,fit:"fill",kernel:"lanczos3"});break;default:throw new Error(`Resampling method ${I} is not supported.`)}return await y(S)}}async pad([E,T,D,I]){if(E=Math.max(E,0),T=Math.max(T,0),D=Math.max(D,0),I=Math.max(I,0),E===0&&T===0&&D===0&&I===0)return this;if(m){const $=this.channels,P=this.toCanvas(),S=this.width+E+T,O=this.height+D+I,R=h(S,O).getContext("2d");return R.drawImage(P,0,0,this.width,this.height,E,D,this.width,this.height),new b(R.getImageData(0,0,S,O).data,S,O,4).convert($)}else{const $=this.toSharp().extend({left:E,right:T,top:D,bottom:I});return await y($)}}async crop([E,T,D,I]){if(E=Math.max(E,0),T=Math.max(T,0),D=Math.min(D,this.width-1),I=Math.min(I,this.height-1),E===0&&T===0&&D===this.width-1&&I===this.height-1)return this;const $=D-E+1,P=I-T+1;if(m){const S=this.channels,O=this.toCanvas(),R=h($,P).getContext("2d");return R.drawImage(O,E,T,$,P,0,0,$,P),new b(R.getImageData(0,0,$,P).data,$,P,4).convert(S)}else{const S=this.toSharp().extract({left:E,top:T,width:$,height:P});return await y(S)}}async center_crop(E,T){if(this.width===E&&this.height===T)return this;const D=(this.width-E)/2,I=(this.height-T)/2;if(m){const $=this.channels,P=this.toCanvas(),S=h(E,T).getContext("2d");let O=0,R=0,G=0,ee=0;return D>=0?O=D:G=-D,I>=0?R=I:ee=-I,S.drawImage(P,O,R,E,T,G,ee,E,T),new b(S.getImageData(0,0,E,T).data,E,T,4).convert($)}else{let $=this.toSharp();if(D>=0&&I>=0)$=$.extract({left:Math.floor(D),top:Math.floor(I),width:E,height:T});else if(D<=0&&I<=0){const P=Math.floor(-I),S=Math.floor(-D);$=$.extend({top:P,left:S,right:E-this.width-S,bottom:T-this.height-P})}else{let P=[0,0],S=0;I<0?(P[0]=Math.floor(-I),P[1]=T-this.height-P[0]):S=Math.floor(I);let O=[0,0],R=0;D<0?(O[0]=Math.floor(-D),O[1]=E-this.width-O[0]):R=Math.floor(D),$=$.extend({top:P[0],bottom:P[1],left:O[0],right:O[1]}).extract({left:R,top:S,width:E,height:T})}return await y($)}}async toBlob(E="image/png",T=1){if(!m)throw new Error("toBlob() is only supported in browser environments.");return await this.toCanvas().convertToBlob({type:E,quality:T})}toTensor(E="CHW"){let T=new u.Tensor("uint8",new Uint8Array(this.data),[this.height,this.width,this.channels]);if(E!=="HWC")if(E==="CHW")T=T.permute(2,0,1);else throw new Error(`Unsupported channel format: ${E}`);return T}toCanvas(){if(!m)throw new Error("toCanvas() is only supported in browser environments.");const E=this.clone().rgba(),T=h(E.width,E.height),D=new f(E.data,E.width,E.height);return T.getContext("2d").putImageData(D,0,0),T}split(){const{data:E,width:T,height:D,channels:I}=this,$=E.constructor,P=E.length/I,S=Array.from({length:I},()=>new $(P));for(let O=0;O<P;++O){const R=I*O;for(let G=0;G<I;++G)S[G][O]=E[R+G]}return S.map(O=>new b(O,T,D,1))}_update(E,T,D,I=null){return this.data=E,this.width=T,this.height=D,I!==null&&(this.channels=I),this}clone(){return new b(this.data.slice(),this.width,this.height,this.channels)}convert(E){if(this.channels===E)return this;switch(E){case 1:this.grayscale();break;case 3:this.rgb();break;case 4:this.rgba();break;default:throw new Error(`Conversion failed due to unsupported number of channels: ${this.channels}`)}return this}async save(E){if(m){if(i.apis.IS_WEBWORKER_ENV)throw new Error("Unable to save an image from a Web Worker.");const T=E.split(".").pop().toLowerCase(),D=v.get(T)??"image/png",I=await this.toBlob(D);(0,n.saveBlob)(E,I)}else{if(i.apis.IS_FS_AVAILABLE)return await this.toSharp().toFile(E);throw new Error("Unable to save the image because filesystem is disabled in this environment.")}}toSharp(){if(m)throw new Error("toSharp() is only supported in server-side environments.");return d(this.data,{raw:{width:this.width,height:this.height,channels:this.channels}})}}const k=b.read.bind(b)},"./src/utils/maths.js":(e,r,t)=>{t.r(r),t.d(r,{FFT:()=>k,bankers_round:()=>T,cos_sim:()=>h,dot:()=>d,dynamic_time_warping:()=>D,interpolate_data:()=>n,log_softmax:()=>u,magnitude:()=>f,max:()=>m,medianFilter:()=>z,min:()=>y,permute_data:()=>o,round:()=>E,softmax:()=>i});function n(I,[$,P,S],[O,R],G="bilinear",ee=!1){const ie=R/S,H=O/P,ce=new I.constructor(O*R*$),re=P*S,se=O*R;for(let _e=0;_e<O;++_e)for(let ae=0;ae<R;++ae){const Ce=_e*R+ae,Te=(ae+.5)/ie-.5,q=(_e+.5)/H-.5;let B=Math.floor(Te),Q=Math.floor(q);const oe=Math.min(B+1,S-1),Ie=Math.min(Q+1,P-1);B=Math.max(B,0),Q=Math.max(Q,0);const ye=Te-B,be=q-Q,et=(1-ye)*(1-be),tt=ye*(1-be),Ee=(1-ye)*be,Y=ye*be,pe=Q*S,X=Ie*S,Ae=pe+B,Oe=pe+oe,Ne=X+B,Be=X+oe;for(let Ve=0;Ve<$;++Ve){const He=Ve*re;ce[Ve*se+Ce]=et*I[He+Ae]+tt*I[He+Oe]+Ee*I[He+Ne]+Y*I[He+Be]}}return ce}function o(I,$,P){const S=new Array(P.length),O=new Array(P.length);for(let ee=P.length-1,ie=1;ee>=0;--ee)O[ee]=ie,S[ee]=$[P[ee]],ie*=S[ee];const R=P.map((ee,ie)=>O[P.indexOf(ie)]),G=new I.constructor(I.length);for(let ee=0;ee<I.length;++ee){let ie=0;for(let H=$.length-1,ce=ee;H>=0;--H)ie+=ce%$[H]*R[H],ce=Math.floor(ce/$[H]);G[ie]=I[ee]}return[G,S]}function i(I){const $=m(I)[0],P=I.map(R=>Math.exp(R-$)),S=P.reduce((R,G)=>R+G,0);return P.map(R=>R/S)}function u(I){const $=m(I)[0];let P=0;for(let R=0;R<I.length;++R)P+=Math.exp(I[R]-$);const S=Math.log(P);return I.map(R=>R-$-S)}function d(I,$){let P=0;for(let S=0;S<I.length;++S)P+=I[S]*$[S];return P}function h(I,$){const P=d(I,$),S=f(I),O=f($);return P/(S*O)}function f(I){return Math.sqrt(I.reduce(($,P)=>$+P*P,0))}function y(I){if(I.length===0)throw Error("Array must not be empty");let $=I[0],P=0;for(let S=1;S<I.length;++S)I[S]<$&&($=I[S],P=S);return[$,P]}function m(I){if(I.length===0)throw Error("Array must not be empty");let $=I[0],P=0;for(let S=1;S<I.length;++S)I[S]>$&&($=I[S],P=S);return[$,P]}function g(I){return I>0&&(I&I-1)===0}class v{constructor($){if(this.size=$|0,this.size<=1||!g(this.size))throw new Error("FFT size must be a power of two larger than 1");this._csize=$<<1,this.table=new Float64Array(this.size*2);for(let S=0;S<this.table.length;S+=2){const O=Math.PI*S/this.size;this.table[S]=Math.cos(O),this.table[S+1]=-Math.sin(O)}let P=0;for(let S=1;this.size>S;S<<=1)++P;this._width=P%2===0?P-1:P,this._bitrev=new Int32Array(1<<this._width);for(let S=0;S<this._bitrev.length;++S){this._bitrev[S]=0;for(let O=0;O<this._width;O+=2){const R=this._width-O-2;this._bitrev[S]|=(S>>>O&3)<<R}}}createComplexArray(){return new Float64Array(this._csize)}fromComplexArray($,P){const S=P||new Array($.length>>>1);for(let O=0;O<$.length;O+=2)S[O>>>1]=$[O];return S}toComplexArray($,P){const S=P||this.createComplexArray();for(let O=0;O<S.length;O+=2)S[O]=$[O>>>1],S[O+1]=0;return S}transform($,P){if($===P)throw new Error("Input and output buffers must be different");this._transform4($,P,1)}realTransform($,P){if($===P)throw new Error("Input and output buffers must be different");this._realTransform4($,P,1)}inverseTransform($,P){if($===P)throw new Error("Input and output buffers must be different");this._transform4($,P,-1);for(let S=0;S<$.length;++S)$[S]/=this.size}_transform4($,P,S){const O=this._csize;let G=1<<this._width,ee=O/G<<1,ie,H;const ce=this._bitrev;if(ee===4)for(ie=0,H=0;ie<O;ie+=ee,++H){const se=ce[H];this._singleTransform2(P,$,ie,se,G)}else for(ie=0,H=0;ie<O;ie+=ee,++H){const se=ce[H];this._singleTransform4(P,$,ie,se,G,S)}const re=this.table;for(G>>=2;G>=2;G>>=2){ee=O/G<<1;const se=ee>>>2;for(ie=0;ie<O;ie+=ee){const _e=ie+se-1;for(let ae=ie,Ce=0;ae<_e;ae+=2,Ce+=G){const Te=ae,q=Te+se,B=q+se,Q=B+se,oe=$[Te],Ie=$[Te+1],ye=$[q],be=$[q+1],et=$[B],tt=$[B+1],Ee=$[Q],Y=$[Q+1],pe=re[Ce],X=S*re[Ce+1],Ae=ye*pe-be*X,Oe=ye*X+be*pe,Ne=re[2*Ce],Be=S*re[2*Ce+1],Ve=et*Ne-tt*Be,He=et*Be+tt*Ne,Xe=re[3*Ce],st=S*re[3*Ce+1],nt=Ee*Xe-Y*st,We=Ee*st+Y*Xe,rt=oe+Ve,pt=Ie+He,_t=oe-Ve,ct=Ie-He,dt=Ae+nt,Le=Oe+We,xt=S*(Ae-nt),Nt=S*(Oe-We);$[Te]=rt+dt,$[Te+1]=pt+Le,$[q]=_t+Nt,$[q+1]=ct-xt,$[B]=rt-dt,$[B+1]=pt-Le,$[Q]=_t-Nt,$[Q+1]=ct+xt}}}}_singleTransform2($,P,S,O,R){const G=$[O],ee=$[O+1],ie=$[O+R],H=$[O+R+1];P[S]=G+ie,P[S+1]=ee+H,P[S+2]=G-ie,P[S+3]=ee-H}_singleTransform4($,P,S,O,R,G){const ee=R*2,ie=R*3,H=$[O],ce=$[O+1],re=$[O+R],se=$[O+R+1],_e=$[O+ee],ae=$[O+ee+1],Ce=$[O+ie],Te=$[O+ie+1],q=H+_e,B=ce+ae,Q=H-_e,oe=ce-ae,Ie=re+Ce,ye=se+Te,be=G*(re-Ce),et=G*(se-Te);P[S]=q+Ie,P[S+1]=B+ye,P[S+2]=Q+et,P[S+3]=oe-be,P[S+4]=q-Ie,P[S+5]=B-ye,P[S+6]=Q-et,P[S+7]=oe+be}_realTransform4($,P,S){const O=this._csize;let G=1<<this._width,ee=O/G<<1,ie,H;const ce=this._bitrev;if(ee===4)for(ie=0,H=0;ie<O;ie+=ee,++H){const _e=ce[H];this._singleRealTransform2(P,$,ie,_e>>>1,G>>>1)}else for(ie=0,H=0;ie<O;ie+=ee,++H){const _e=ce[H];this._singleRealTransform4(P,$,ie,_e>>>1,G>>>1,S)}const re=this.table;for(G>>=2;G>=2;G>>=2){ee=O/G<<1;const _e=ee>>>1,ae=_e>>>1,Ce=ae>>>1;for(ie=0;ie<O;ie+=ee)for(let Te=0,q=0;Te<=Ce;Te+=2,q+=G){const B=ie+Te,Q=B+ae,oe=Q+ae,Ie=oe+ae,ye=$[B],be=$[B+1],et=$[Q],tt=$[Q+1],Ee=$[oe],Y=$[oe+1],pe=$[Ie],X=$[Ie+1],Ae=ye,Oe=be,Ne=re[q],Be=S*re[q+1],Ve=et*Ne-tt*Be,He=et*Be+tt*Ne,Xe=re[2*q],st=S*re[2*q+1],nt=Ee*Xe-Y*st,We=Ee*st+Y*Xe,rt=re[3*q],pt=S*re[3*q+1],_t=pe*rt-X*pt,ct=pe*pt+X*rt,dt=Ae+nt,Le=Oe+We,xt=Ae-nt,Nt=Oe-We,cr=Ve+_t,Kt=He+ct,wr=S*(Ve-_t),zr=S*(He-ct);if($[B]=dt+cr,$[B+1]=Le+Kt,$[Q]=xt+zr,$[Q+1]=Nt-wr,Te===0){$[oe]=dt-cr,$[oe+1]=Le-Kt;continue}if(Te===Ce)continue;const $t=ie+ae-Te,Ti=ie+_e-Te;$[$t]=xt-S*zr,$[$t+1]=-Nt-S*wr,$[Ti]=dt-S*cr,$[Ti+1]=-Le+S*Kt}}const se=O>>>1;for(let _e=2;_e<se;_e+=2)$[O-_e]=$[_e],$[O-_e+1]=-$[_e+1]}_singleRealTransform2($,P,S,O,R){const G=$[O],ee=$[O+R];P[S]=G+ee,P[S+1]=0,P[S+2]=G-ee,P[S+3]=0}_singleRealTransform4($,P,S,O,R,G){const ee=R*2,ie=R*3,H=$[O],ce=$[O+R],re=$[O+ee],se=$[O+ie],_e=H+re,ae=H-re,Ce=ce+se,Te=G*(ce-se);P[S]=_e+Ce,P[S+1]=0,P[S+2]=ae,P[S+3]=-Te,P[S+4]=_e-Ce,P[S+5]=0,P[S+6]=ae,P[S+7]=Te}}class b{constructor($){const P=2*($-1),S=2*(2*$-1),O=2**Math.ceil(Math.log2(S));this.bufferSize=O,this._a=P;const R=new Float64Array(S),G=new Float64Array(O);this._chirpBuffer=new Float64Array(O),this._buffer1=new Float64Array(O),this._buffer2=new Float64Array(O),this._outBuffer1=new Float64Array(O),this._outBuffer2=new Float64Array(O);const ee=-2*Math.PI/$,ie=Math.cos(ee),H=Math.sin(ee);for(let ce=0;ce<S>>1;++ce){const re=(ce+1-$)**2/2,se=Math.sqrt(ie**2+H**2)**re,_e=re*Math.atan2(H,ie),ae=2*ce;R[ae]=se*Math.cos(_e),R[ae+1]=se*Math.sin(_e),G[ae]=R[ae],G[ae+1]=-R[ae+1]}this._slicedChirpBuffer=R.subarray(P,S),this._f=new v(O>>1),this._f.transform(this._chirpBuffer,G)}_transform($,P,S){const O=this._buffer1,R=this._buffer2,G=this._outBuffer1,ee=this._outBuffer2,ie=this._chirpBuffer,H=this._slicedChirpBuffer,ce=this._a;if(S)for(let re=0;re<H.length;re+=2){const se=re+1,_e=re>>1,ae=P[_e];O[re]=ae*H[re],O[se]=ae*H[se]}else for(let re=0;re<H.length;re+=2){const se=re+1;O[re]=P[re]*H[re]-P[se]*H[se],O[se]=P[re]*H[se]+P[se]*H[re]}this._f.transform(G,O);for(let re=0;re<ie.length;re+=2){const se=re+1;R[re]=G[re]*ie[re]-G[se]*ie[se],R[se]=G[re]*ie[se]+G[se]*ie[re]}this._f.inverseTransform(ee,R);for(let re=0;re<ee.length;re+=2){const se=ee[re+ce],_e=ee[re+ce+1],ae=H[re],Ce=H[re+1];$[re]=se*ae-_e*Ce,$[re+1]=se*Ce+_e*ae}}transform($,P){this._transform($,P,!1)}realTransform($,P){this._transform($,P,!0)}}class k{constructor($){this.fft_length=$,this.isPowerOfTwo=g($),this.isPowerOfTwo?(this.fft=new v($),this.outputBufferSize=2*$):(this.fft=new b($),this.outputBufferSize=this.fft.bufferSize)}realTransform($,P){this.fft.realTransform($,P)}transform($,P){this.fft.transform($,P)}}function z(I,$){if($%2===0||$<=0)throw new Error("Window size must be a positive odd number");const P=new I.constructor(I.length),S=new I.constructor($),O=Math.floor($/2);for(let R=0;R<I.length;++R){let G=0;for(let ee=-O;ee<=O;++ee){let ie=R+ee;ie<0?ie=Math.abs(ie):ie>=I.length&&(ie=2*(I.length-1)-ie),S[G++]=I[ie]}S.sort(),P[R]=S[O]}return P}function E(I,$){const P=Math.pow(10,$);return Math.round(I*P)/P}function T(I){const $=Math.round(I);return Math.abs(I)%1===.5?$%2===0?$:$-1:$}function D(I){const $=I.length,P=I[0].length,S=[$+1,P+1],O=Array.from({length:S[0]},()=>Array(S[1]).fill(1/0));O[0][0]=0;const R=Array.from({length:S[0]},()=>Array(S[1]).fill(-1));for(let ce=1;ce<S[1];++ce)for(let re=1;re<S[0];++re){const se=O[re-1][ce-1],_e=O[re-1][ce],ae=O[re][ce-1];let Ce,Te;se<_e&&se<ae?(Ce=se,Te=0):_e<se&&_e<ae?(Ce=_e,Te=1):(Ce=ae,Te=2),O[re][ce]=I[re-1][ce-1]+Ce,R[re][ce]=Te}for(let ce=0;ce<S[1];++ce)R[0][ce]=2;for(let ce=0;ce<S[0];++ce)R[ce][0]=1;let G=$,ee=P,ie=[],H=[];for(;G>0||ee>0;)switch(ie.push(G-1),H.push(ee-1),R[G][ee]){case 0:--G,--ee;break;case 1:--G;break;case 2:--ee;break;default:throw new Error(`Internal error in dynamic time warping. Unexpected trace[${G}, ${ee}]. Please file a bug report.`)}return ie.reverse(),H.reverse(),[ie,H]}},"./src/utils/tensor.js":(e,r,t)=>{t.r(r),t.d(r,{DataTypeMap:()=>u,Tensor:()=>d,cat:()=>P,full:()=>H,full_like:()=>ce,interpolate:()=>y,interpolate_4d:()=>m,layer_norm:()=>T,matmul:()=>g,mean:()=>G,mean_pooling:()=>E,ones:()=>re,ones_like:()=>se,permute:()=>f,quantize_embeddings:()=>Te,rand:()=>Ce,rfft:()=>v,slice:()=>z,stack:()=>S,std_mean:()=>R,topk:()=>b,zeros:()=>_e,zeros_like:()=>ae});var n=t("./src/utils/maths.js"),o=t("./src/backends/onnx.js"),i=t("./src/ops/registry.js");const u=Object.freeze({float32:Float32Array,float16:typeof Float16Array<"u"?Float16Array:Uint16Array,float64:Float64Array,string:Array,int8:Int8Array,uint8:Uint8Array,int16:Int16Array,uint16:Uint16Array,int32:Int32Array,uint32:Uint32Array,int64:BigInt64Array,uint64:BigUint64Array,bool:Uint8Array,uint4:Uint8Array,int4:Int8Array});class d{constructor(...B){le(this,"ort_tensor");return(0,o.isONNXTensor)(B[0])?this.ort_tensor=B[0]:this.ort_tensor=new o.Tensor(B[0],B[1],B[2]),new Proxy(this,{get:(Q,oe)=>{if(typeof oe=="string"){let Ie=Number(oe);if(Number.isInteger(Ie))return Q._getitem(Ie)}return Q[oe]},set:(Q,oe,Ie)=>Q[oe]=Ie})}get dims(){return this.ort_tensor.dims}set dims(B){this.ort_tensor.dims=B}get type(){return this.ort_tensor.type}get data(){return this.ort_tensor.data}get size(){return this.ort_tensor.size}get location(){return this.ort_tensor.location}dispose(){this.ort_tensor.dispose()}*[Symbol.iterator](){const[B,...Q]=this.dims;if(Q.length>0){const oe=Q.reduce((Ie,ye)=>Ie*ye);for(let Ie=0;Ie<B;++Ie)yield this._subarray(Ie,oe,Q)}else yield*this.data}_getitem(B){const[Q,...oe]=this.dims;if(B=$(B,Q),oe.length>0){const Ie=oe.reduce((ye,be)=>ye*be);return this._subarray(B,Ie,oe)}else return new d(this.type,[this.data[B]],oe)}indexOf(B){const Q=this.data;for(let oe=0;oe<Q.length;++oe)if(Q[oe]==B)return oe;return-1}_subarray(B,Q,oe){const Ie=B*Q,ye=(B+1)*Q,be="subarray"in this.data?this.data.subarray(Ie,ye):this.data.slice(Ie,ye);return new d(this.type,be,oe)}item(){const B=this.data;if(B.length!==1)throw new Error(`a Tensor with ${B.length} elements cannot be converted to Scalar`);return B[0]}tolist(){return h(this.data,this.dims)}sigmoid(){return this.clone().sigmoid_()}sigmoid_(){const B=this.data;for(let Q=0;Q<B.length;++Q)B[Q]=1/(1+Math.exp(-B[Q]));return this}map(B){return this.clone().map_(B)}map_(B){const Q=this.data;for(let oe=0;oe<Q.length;++oe)Q[oe]=B(Q[oe],oe,Q);return this}mul(B){return this.clone().mul_(B)}mul_(B){const Q=this.data;for(let oe=0;oe<Q.length;++oe)Q[oe]*=B;return this}div(B){return this.clone().div_(B)}div_(B){const Q=this.data;for(let oe=0;oe<Q.length;++oe)Q[oe]/=B;return this}add(B){return this.clone().add_(B)}add_(B){const Q=this.data;for(let oe=0;oe<Q.length;++oe)Q[oe]+=B;return this}sub(B){return this.clone().sub_(B)}sub_(B){const Q=this.data;for(let oe=0;oe<Q.length;++oe)Q[oe]-=B;return this}clone(){return new d(this.type,this.data.slice(),this.dims.slice())}slice(...B){const Q=[],oe=[];for(let Ee=0;Ee<this.dims.length;++Ee){let Y=B[Ee];if(Y==null)oe.push([0,this.dims[Ee]]),Q.push(this.dims[Ee]);else if(typeof Y=="number")Y=$(Y,this.dims[Ee],Ee),oe.push([Y,Y+1]);else if(Array.isArray(Y)&&Y.length===2){let[pe,X]=Y;if(pe=pe===null?0:$(pe,this.dims[Ee],Ee,!1),X=X===null?this.dims[Ee]:$(X,this.dims[Ee],Ee,!1),pe>X)throw new Error(`Invalid slice: ${Y}`);const Ae=[Math.max(pe,0),Math.min(X,this.dims[Ee])];oe.push(Ae),Q.push(Ae[1]-Ae[0])}else throw new Error(`Invalid slice: ${Y}`)}const Ie=oe.map(([Ee,Y])=>Y-Ee),ye=Ie.reduce((Ee,Y)=>Ee*Y),be=this.data,et=new be.constructor(ye),tt=this.stride();for(let Ee=0;Ee<ye;++Ee){let Y=0;for(let pe=Ie.length-1,X=Ee;pe>=0;--pe){const Ae=Ie[pe];Y+=(X%Ae+oe[pe][0])*tt[pe],X=Math.floor(X/Ae)}et[Ee]=be[Y]}return new d(this.type,et,Q)}permute(...B){return f(this,B)}transpose(...B){return this.permute(...B)}sum(B=null,Q=!1){return this.norm(1,B,Q)}norm(B="fro",Q=null,oe=!1){if(B==="fro")B=2;else if(typeof B=="string")throw Error(`Unsupported norm: ${B}`);const Ie=this.data,ye=(Ee,Y)=>Ee+Y**B;if(Q===null){const Ee=Ie.reduce(ye,0)**(1/B);return new d(this.type,[Ee],[])}const[be,et,tt]=O(ye,this,Q,oe);if(B!==1)for(let Ee=0;Ee<et.length;++Ee)et[Ee]=et[Ee]**(1/B);return new d(be,et,tt)}normalize_(B=2,Q=1){Q=$(Q,this.dims.length);const oe=this.norm(B,Q,!0),Ie=this.data,ye=oe.data;for(let be=0;be<Ie.length;++be){let et=0;for(let tt=this.dims.length-1,Ee=be,Y=1;tt>=0;--tt){const pe=this.dims[tt];if(tt!==Q){const X=Ee%pe;et+=X*Y,Y*=this.dims[tt]}Ee=Math.floor(Ee/pe)}Ie[be]/=ye[et]}return this}normalize(B=2,Q=1){return this.clone().normalize_(B,Q)}stride(){return ee(this.dims)}squeeze(B=null){return new d(this.type,this.data,D(this.dims,B))}squeeze_(B=null){return this.dims=D(this.dims,B),this}unsqueeze(B=null){return new d(this.type,this.data,I(this.dims,B))}unsqueeze_(B=null){return this.dims=I(this.dims,B),this}flatten_(B=0,Q=-1){Q=(Q+this.dims.length)%this.dims.length;let oe=this.dims.slice(0,B),Ie=this.dims.slice(B,Q+1),ye=this.dims.slice(Q+1);return this.dims=[...oe,Ie.reduce((be,et)=>be*et,1),...ye],this}flatten(B=0,Q=-1){return this.clone().flatten_(B,Q)}view(...B){let Q=-1;for(let Ie=0;Ie<B.length;++Ie)if(B[Ie]===-1){if(Q!==-1)throw new Error("Only one dimension can be inferred");Q=Ie}const oe=this.data;if(Q!==-1){const Ie=B.reduce((ye,be,et)=>et!==Q?ye*be:ye,1);B[Q]=oe.length/Ie}return new d(this.type,oe,B)}neg_(){const B=this.data;for(let Q=0;Q<B.length;++Q)B[Q]=-B[Q];return this}neg(){return this.clone().neg_()}gt(B){const Q=new Uint8Array(this.data.length),oe=this.data;for(let Ie=0;Ie<oe.length;++Ie)Q[Ie]=oe[Ie]>B?1:0;return new d("bool",Q,this.dims)}lt(B){const Q=new Uint8Array(this.data.length),oe=this.data;for(let Ie=0;Ie<oe.length;++Ie)Q[Ie]=oe[Ie]<B?1:0;return new d("bool",Q,this.dims)}clamp_(B,Q){const oe=this.data;for(let Ie=0;Ie<oe.length;++Ie)oe[Ie]=Math.min(Math.max(oe[Ie],B),Q);return this}clamp(B,Q){return this.clone().clamp_(B,Q)}round_(){const B=this.data;for(let Q=0;Q<B.length;++Q)B[Q]=Math.round(B[Q]);return this}round(){return this.clone().round_()}mean(B=null,Q=!1){return G(this,B,Q)}min(B=null,Q=!1){if(B===null){const be=(0,n.min)(this.data)[0];return new d(this.type,[be],[])}const[oe,Ie,ye]=O((be,et)=>Math.min(be,et),this,B,Q,1/0);return new d(oe,Ie,ye)}max(B=null,Q=!1){if(B===null){const be=(0,n.max)(this.data)[0];return new d(this.type,[be],[])}const[oe,Ie,ye]=O((be,et)=>Math.max(be,et),this,B,Q,-1/0);return new d(oe,Ie,ye)}argmin(B=null,Q=!1){if(B!==null)throw new Error("`dim !== null` not yet implemented.");const oe=(0,n.min)(this.data)[1];return new d("int64",[BigInt(oe)],[])}argmax(B=null,Q=!1){if(B!==null)throw new Error("`dim !== null` not yet implemented.");const oe=(0,n.max)(this.data)[1];return new d("int64",[BigInt(oe)],[])}to(B){if(this.type===B)return this;if(!u.hasOwnProperty(B))throw new Error(`Unsupported type: ${B}`);let Q;const oe=["int64","uint64"].includes(this.type),Ie=["int64","uint64"].includes(B);return oe&&!Ie?Q=Number:!oe&&Ie&&(Q=BigInt),new d(B,u[B].from(this.data,Q),this.dims)}}function h(q,B){const Q=q.length,oe=B.reduce((ye,be)=>ye*be);if(Q!==oe)throw Error(`cannot reshape array of size ${Q} into shape (${B})`);let Ie=q;for(let ye=B.length-1;ye>=0;ye--)Ie=Ie.reduce((be,et)=>{let tt=be[be.length-1];return tt.length<B[ye]?tt.push(et):be.push([et]),be},[[]]);return Ie[0]}function f(q,B){const[Q,oe]=(0,n.permute_data)(q.data,q.dims,B);return new d(q.type,Q,oe)}function y(q,[B,Q],oe="bilinear",Ie=!1){const ye=q.dims.at(-3)??1,be=q.dims.at(-2),et=q.dims.at(-1);let tt=(0,n.interpolate_data)(q.data,[ye,be,et],[B,Q],oe,Ie);return new d(q.type,tt,[ye,B,Q])}async function m(q,{size:B=null,mode:Q="bilinear"}={}){if(q.dims.length!==4)throw new Error("`interpolate_4d` currently only supports 4D input.");if(!B)throw new Error("`interpolate_4d` requires a `size` argument.");let oe;if(B.length===2)oe=[...q.dims.slice(0,2),...B];else if(B.length===3)oe=[q.dims[0],...B];else if(B.length===4)oe=B;else throw new Error("`size` must be of length 2, 3, or 4.");let Ie;if(Q==="nearest")Ie=await i.TensorOpRegistry.nearest_interpolate_4d;else if(Q==="bilinear")Ie=await i.TensorOpRegistry.bilinear_interpolate_4d;else if(Q==="bicubic")Ie=await i.TensorOpRegistry.bicubic_interpolate_4d;else throw new Error(`Unsupported mode: ${Q}`);const ye=new d("int64",new BigInt64Array(oe.map(BigInt)),[oe.length]);return await Ie({x:q,s:ye})}async function g(q,B){return await(await i.TensorOpRegistry.matmul)({a:q,b:B})}async function v(q,B){return await(await i.TensorOpRegistry.rfft)({x:q,a:B})}async function b(q,B){const Q=await i.TensorOpRegistry.top_k;return B==null?B=q.dims.at(-1):B=Math.min(B,q.dims.at(-1)),await Q({x:q,k:new d("int64",[BigInt(B)],[1])})}const k=q=>new d("int64",q,[q.length]);async function z(q,B,Q,oe,Ie){return await(await i.TensorOpRegistry.slice)({x:q,s:k(B),e:k(Q),a:k(oe),t:k(Ie??new Array(oe.length).fill(1))})}function E(q,B){const Q=q.data,oe=B.data,Ie=[q.dims[0],q.dims[2]],ye=new Q.constructor(Ie[0]*Ie[1]),[be,et,tt]=q.dims;let Ee=0;for(let Y=0;Y<be;++Y){const pe=Y*tt*et;for(let X=0;X<tt;++X){let Ae=0,Oe=0;const Ne=Y*et,Be=pe+X;for(let He=0;He<et;++He){const Xe=Number(oe[Ne+He]);Oe+=Xe,Ae+=Q[Be+He*tt]*Xe}const Ve=Ae/Oe;ye[Ee++]=Ve}}return new d(q.type,ye,Ie)}function T(q,B,{eps:Q=1e-5}={}){if(q.dims.length!==2)throw new Error("`layer_norm` currently only supports 2D input.");const[oe,Ie]=q.dims;if(B.length!==1&&B[0]!==Ie)throw new Error("`normalized_shape` must be a 1D array with shape `[input.dims[1]]`.");const[ye,be]=R(q,1,0,!0),et=ye.data,tt=be.data,Ee=q.data,Y=new Ee.constructor(Ee.length);for(let pe=0;pe<oe;++pe){const X=pe*Ie;for(let Ae=0;Ae<Ie;++Ae){const Oe=X+Ae;Y[Oe]=(Ee[Oe]-tt[pe])/(et[pe]+Q)}}return new d(q.type,Y,q.dims)}function D(q,B){return q=q.slice(),B===null?q=q.filter(Q=>Q!==1):typeof B=="number"?q[B]===1&&q.splice(B,1):Array.isArray(B)&&(q=q.filter((Q,oe)=>Q!==1||!B.includes(oe))),q}function I(q,B){return B=$(B,q.length+1),q=q.slice(),q.splice(B,0,1),q}function $(q,B,Q=null,oe=!0){if(q<-B||q>=B){if(oe)throw new Error(`IndexError: index ${q} is out of bounds for dimension${Q===null?"":" "+Q} with size ${B}`);return q<-B?0:B}return q<0&&(q=(q%B+B)%B),q}function P(q,B=0){B=$(B,q[0].dims.length);const Q=q[0].dims.slice();Q[B]=q.reduce((be,et)=>be+et.dims[B],0);const oe=Q.reduce((be,et)=>be*et,1),Ie=new q[0].data.constructor(oe),ye=q[0].type;if(B===0){let be=0;for(const et of q){const tt=et.data;Ie.set(tt,be),be+=tt.length}}else{let be=0;for(let et=0;et<q.length;++et){const{data:tt,dims:Ee}=q[et];for(let Y=0;Y<tt.length;++Y){let pe=0;for(let X=Ee.length-1,Ae=Y,Oe=1;X>=0;--X){const Ne=Ee[X];let Be=Ae%Ne;X===B&&(Be+=be),pe+=Be*Oe,Oe*=Q[X],Ae=Math.floor(Ae/Ne)}Ie[pe]=tt[Y]}be+=Ee[B]}}return new d(ye,Ie,Q)}function S(q,B=0){return P(q.map(Q=>Q.unsqueeze(B)),B)}function O(q,B,Q=null,oe=!1,Ie=null){const ye=B.data,be=B.dims;Q=$(Q,be.length);const et=be.slice();et[Q]=1;const tt=new ye.constructor(ye.length/be[Q]);Ie!==null&&tt.fill(Ie);for(let Ee=0;Ee<ye.length;++Ee){let Y=0;for(let pe=be.length-1,X=Ee,Ae=1;pe>=0;--pe){const Oe=be[pe];if(pe!==Q){const Ne=X%Oe;Y+=Ne*Ae,Ae*=et[pe]}X=Math.floor(X/Oe)}tt[Y]=q(tt[Y],ye[Ee],Ee,Y)}return oe||et.splice(Q,1),[B.type,tt,et]}function R(q,B=null,Q=1,oe=!1){const Ie=q.data,ye=q.dims;if(B===null){const Ae=Ie.reduce((Ve,He)=>Ve+He,0)/Ie.length,Oe=Math.sqrt(Ie.reduce((Ve,He)=>Ve+(He-Ae)**2,0)/(Ie.length-Q)),Ne=new d(q.type,[Ae],[]);return[new d(q.type,[Oe],[]),Ne]}B=$(B,ye.length);const be=G(q,B,oe),et=be.data,[tt,Ee,Y]=O((X,Ae,Oe,Ne)=>X+(Ae-et[Ne])**2,q,B,oe);for(let X=0;X<Ee.length;++X)Ee[X]=Math.sqrt(Ee[X]/(ye[B]-Q));return[new d(tt,Ee,Y),be]}function G(q,B=null,Q=!1){const oe=q.dims,Ie=q.data;if(B===null){const tt=Ie.reduce((Ee,Y)=>Ee+Y,0);return new d(q.type,[tt/Ie.length],[])}B=$(B,oe.length);const[ye,be,et]=O((tt,Ee)=>tt+Ee,q,B,Q);if(oe[B]!==1)for(let tt=0;tt<be.length;++tt)be[tt]/=oe[B];return new d(ye,be,et)}function ee(q){const B=new Array(q.length);for(let Q=q.length-1,oe=1;Q>=0;--Q)B[Q]=oe,oe*=q[Q];return B}function ie(q,B,Q,oe){const Ie=q.reduce((ye,be)=>ye*be,1);return new d(Q,new oe(Ie).fill(B),q)}function H(q,B){let Q,oe;if(typeof B=="number")Q="float32",oe=Float32Array;else if(typeof B=="bigint")Q="int64",oe=BigInt64Array;else if(typeof B=="boolean")Q="bool",oe=Uint8Array;else throw new Error(`Unsupported data type: ${typeof B}`);return ie(q,B,Q,oe)}function ce(q,B){return H(q.dims,B)}function re(q){return ie(q,1n,"int64",BigInt64Array)}function se(q){return re(q.dims)}function _e(q){return ie(q,0n,"int64",BigInt64Array)}function ae(q){return _e(q.dims)}function Ce(q){const B=q.reduce((Q,oe)=>Q*oe,1);return new d("float32",Float32Array.from({length:B},()=>Math.random()),q)}function Te(q,B){if(q.dims.length!==2)throw new Error("The tensor must have 2 dimensions");if(q.dims.at(-1)%8!==0)throw new Error("The last dimension of the tensor must be a multiple of 8");if(!["binary","ubinary"].includes(B))throw new Error("The precision must be either 'binary' or 'ubinary'");const Q=B==="binary",oe=Q?"int8":"uint8",Ie=Q?Int8Array:Uint8Array,ye=q.data,be=new Ie(ye.length/8);for(let et=0;et<ye.length;++et){const tt=ye[et]>0?1:0,Ee=Math.floor(et/8),Y=et%8;be[Ee]|=tt<<7-Y,Q&&Y===0&&(be[Ee]-=128)}return new d(oe,be,[q.dims[0],q.dims[1]/8])}},"./src/utils/video.js":(e,r,t)=>{t.r(r),t.d(r,{RawVideo:()=>u,RawVideoFrame:()=>i,load_video:()=>d});var n=t("./src/utils/image.js"),o=t("./src/env.js");class i{constructor(f,y){this.image=f,this.timestamp=y}}class u{constructor(f,y){f.length>0&&f[0]instanceof n.RawImage&&(f=f.map((m,g)=>new i(m,(g+1)/(f.length+1)*y))),this.frames=f,this.duration=y}get width(){return this.frames[0].image.width}get height(){return this.frames[0].image.height}get fps(){return this.frames.length/this.duration}}async function d(h,{num_frames:f=null,fps:y=null}={}){if(!o.apis.IS_BROWSER_ENV)throw new Error("`load_video` is currently only supported in browser environments.");if(f==null&&y==null)throw new Error("Either num_frames or fps must be provided.");const m=[],g=document.createElement("video");if(g.crossOrigin="anonymous",g.muted=!0,typeof h=="string")g.src=h;else if(h instanceof Blob)g.src=URL.createObjectURL(h);else if(h instanceof HTMLVideoElement)g.src=h.src;else throw new Error("Invalid URL or video element provided.");if(await new Promise(D=>g.onloadedmetadata=D),g.seekable.start(0)===g.seekable.end(0)){const I=await(await fetch(g.src)).blob();g.src=URL.createObjectURL(I),await new Promise($=>g.onloadedmetadata=$)}const v=g.duration;let b,k;f!=null?(b=f,k=f===1?0:v/(f-1)):(k=1/y,b=Math.floor(v/k));let z=[];for(let D=0;D<b;++D)z.push(f===1?v/2:D*k);const E=document.createElement("canvas");E.width=g.videoWidth,E.height=g.videoHeight;const T=E.getContext("2d",{willReadFrequently:!0});for(const D of z){g.currentTime=D,await new Promise(S=>{g.onseeked=S}),T.drawImage(g,0,0,E.width,E.height);const I=T.getImageData(0,0,E.width,E.height),$=new n.RawImage(I.data,E.width,E.height,4),P=new i($,D);m.push(P)}return g.remove(),new u(m,v)}},"?154a":()=>{},"?2b25":()=>{},"?2ce3":()=>{},"?3f59":()=>{},"?569f":()=>{},"?7a2c":()=>{},"?a42a":()=>{},"onnxruntime-common":e=>{e.exports=D1},"onnxruntime-web":e=>{e.exports=rE}},n0={};function Yt(e){var r=n0[e];if(r!==void 0)return r.exports;var t=n0[e]={exports:{}};return iE[e](t,t.exports,Yt),t.exports}(()=>{var e=Object.getPrototypeOf?t=>Object.getPrototypeOf(t):t=>t.__proto__,r;Yt.t=function(t,n){if(n&1&&(t=this(t)),n&8||typeof t=="object"&&t&&(n&4&&t.__esModule||n&16&&typeof t.then=="function"))return t;var o=Object.create(null);Yt.r(o);var i={};r=r||[null,e({}),e([]),e(e)];for(var u=n&2&&t;typeof u=="object"&&!~r.indexOf(u);u=e(u))Object.getOwnPropertyNames(u).forEach(d=>i[d]=()=>t[d]);return i.default=()=>t,Yt.d(o,i),o}})();Yt.d=(e,r)=>{for(var t in r)Yt.o(r,t)&&!Yt.o(e,t)&&Object.defineProperty(e,t,{enumerable:!0,get:r[t]})};Yt.o=(e,r)=>Object.prototype.hasOwnProperty.call(e,r);Yt.r=e=>{typeof Symbol<"u"&&Symbol.toStringTag&&Object.defineProperty(e,Symbol.toStringTag,{value:"Module"}),Object.defineProperty(e,"__esModule",{value:!0})};var w={};(()=>{/*!*****************************!*\
  !*** ./src/transformers.js ***!
  \*****************************/Yt.r(w),Yt.d(w,{ASTFeatureExtractor:()=>m.ASTFeatureExtractor,ASTForAudioClassification:()=>t.ASTForAudioClassification,ASTModel:()=>t.ASTModel,ASTPreTrainedModel:()=>t.ASTPreTrainedModel,AlbertForMaskedLM:()=>t.AlbertForMaskedLM,AlbertForQuestionAnswering:()=>t.AlbertForQuestionAnswering,AlbertForSequenceClassification:()=>t.AlbertForSequenceClassification,AlbertModel:()=>t.AlbertModel,AlbertPreTrainedModel:()=>t.AlbertPreTrainedModel,AlbertTokenizer:()=>n.AlbertTokenizer,AudioClassificationPipeline:()=>r.AudioClassificationPipeline,AutoConfig:()=>o.AutoConfig,AutoFeatureExtractor:()=>g.AutoFeatureExtractor,AutoImageProcessor:()=>k.AutoImageProcessor,AutoModel:()=>t.AutoModel,AutoModelForAudioClassification:()=>t.AutoModelForAudioClassification,AutoModelForAudioFrameClassification:()=>t.AutoModelForAudioFrameClassification,AutoModelForAudioTextToText:()=>t.AutoModelForAudioTextToText,AutoModelForCTC:()=>t.AutoModelForCTC,AutoModelForCausalLM:()=>t.AutoModelForCausalLM,AutoModelForDepthEstimation:()=>t.AutoModelForDepthEstimation,AutoModelForDocumentQuestionAnswering:()=>t.AutoModelForDocumentQuestionAnswering,AutoModelForImageClassification:()=>t.AutoModelForImageClassification,AutoModelForImageFeatureExtraction:()=>t.AutoModelForImageFeatureExtraction,AutoModelForImageMatting:()=>t.AutoModelForImageMatting,AutoModelForImageSegmentation:()=>t.AutoModelForImageSegmentation,AutoModelForImageTextToText:()=>t.AutoModelForImageTextToText,AutoModelForImageToImage:()=>t.AutoModelForImageToImage,AutoModelForMaskGeneration:()=>t.AutoModelForMaskGeneration,AutoModelForMaskedLM:()=>t.AutoModelForMaskedLM,AutoModelForNormalEstimation:()=>t.AutoModelForNormalEstimation,AutoModelForObjectDetection:()=>t.AutoModelForObjectDetection,AutoModelForPoseEstimation:()=>t.AutoModelForPoseEstimation,AutoModelForQuestionAnswering:()=>t.AutoModelForQuestionAnswering,AutoModelForSemanticSegmentation:()=>t.AutoModelForSemanticSegmentation,AutoModelForSeq2SeqLM:()=>t.AutoModelForSeq2SeqLM,AutoModelForSequenceClassification:()=>t.AutoModelForSequenceClassification,AutoModelForSpeechSeq2Seq:()=>t.AutoModelForSpeechSeq2Seq,AutoModelForTextToSpectrogram:()=>t.AutoModelForTextToSpectrogram,AutoModelForTextToWaveform:()=>t.AutoModelForTextToWaveform,AutoModelForTokenClassification:()=>t.AutoModelForTokenClassification,AutoModelForUniversalSegmentation:()=>t.AutoModelForUniversalSegmentation,AutoModelForVision2Seq:()=>t.AutoModelForVision2Seq,AutoModelForXVector:()=>t.AutoModelForXVector,AutoModelForZeroShotObjectDetection:()=>t.AutoModelForZeroShotObjectDetection,AutoProcessor:()=>T.AutoProcessor,AutoTokenizer:()=>n.AutoTokenizer,AutomaticSpeechRecognitionPipeline:()=>r.AutomaticSpeechRecognitionPipeline,BackgroundRemovalPipeline:()=>r.BackgroundRemovalPipeline,BartForConditionalGeneration:()=>t.BartForConditionalGeneration,BartForSequenceClassification:()=>t.BartForSequenceClassification,BartModel:()=>t.BartModel,BartPretrainedModel:()=>t.BartPretrainedModel,BartTokenizer:()=>n.BartTokenizer,BaseModelOutput:()=>t.BaseModelOutput,BaseStreamer:()=>D.BaseStreamer,BeitFeatureExtractor:()=>b.BeitFeatureExtractor,BeitForImageClassification:()=>t.BeitForImageClassification,BeitModel:()=>t.BeitModel,BeitPreTrainedModel:()=>t.BeitPreTrainedModel,BertForMaskedLM:()=>t.BertForMaskedLM,BertForQuestionAnswering:()=>t.BertForQuestionAnswering,BertForSequenceClassification:()=>t.BertForSequenceClassification,BertForTokenClassification:()=>t.BertForTokenClassification,BertModel:()=>t.BertModel,BertPreTrainedModel:()=>t.BertPreTrainedModel,BertTokenizer:()=>n.BertTokenizer,BitImageProcessor:()=>b.BitImageProcessor,BlenderbotForConditionalGeneration:()=>t.BlenderbotForConditionalGeneration,BlenderbotModel:()=>t.BlenderbotModel,BlenderbotPreTrainedModel:()=>t.BlenderbotPreTrainedModel,BlenderbotSmallForConditionalGeneration:()=>t.BlenderbotSmallForConditionalGeneration,BlenderbotSmallModel:()=>t.BlenderbotSmallModel,BlenderbotSmallPreTrainedModel:()=>t.BlenderbotSmallPreTrainedModel,BlenderbotSmallTokenizer:()=>n.BlenderbotSmallTokenizer,BlenderbotTokenizer:()=>n.BlenderbotTokenizer,BloomForCausalLM:()=>t.BloomForCausalLM,BloomModel:()=>t.BloomModel,BloomPreTrainedModel:()=>t.BloomPreTrainedModel,BloomTokenizer:()=>n.BloomTokenizer,CLIPFeatureExtractor:()=>b.CLIPFeatureExtractor,CLIPImageProcessor:()=>b.CLIPImageProcessor,CLIPModel:()=>t.CLIPModel,CLIPPreTrainedModel:()=>t.CLIPPreTrainedModel,CLIPSegForImageSegmentation:()=>t.CLIPSegForImageSegmentation,CLIPSegModel:()=>t.CLIPSegModel,CLIPSegPreTrainedModel:()=>t.CLIPSegPreTrainedModel,CLIPTextModel:()=>t.CLIPTextModel,CLIPTextModelWithProjection:()=>t.CLIPTextModelWithProjection,CLIPTokenizer:()=>n.CLIPTokenizer,CLIPVisionModel:()=>t.CLIPVisionModel,CLIPVisionModelWithProjection:()=>t.CLIPVisionModelWithProjection,CamembertForMaskedLM:()=>t.CamembertForMaskedLM,CamembertForQuestionAnswering:()=>t.CamembertForQuestionAnswering,CamembertForSequenceClassification:()=>t.CamembertForSequenceClassification,CamembertForTokenClassification:()=>t.CamembertForTokenClassification,CamembertModel:()=>t.CamembertModel,CamembertPreTrainedModel:()=>t.CamembertPreTrainedModel,CamembertTokenizer:()=>n.CamembertTokenizer,CausalLMOutput:()=>t.CausalLMOutput,CausalLMOutputWithPast:()=>t.CausalLMOutputWithPast,ChineseCLIPFeatureExtractor:()=>b.ChineseCLIPFeatureExtractor,ChineseCLIPModel:()=>t.ChineseCLIPModel,ChineseCLIPPreTrainedModel:()=>t.ChineseCLIPPreTrainedModel,ClapAudioModelWithProjection:()=>t.ClapAudioModelWithProjection,ClapFeatureExtractor:()=>m.ClapFeatureExtractor,ClapModel:()=>t.ClapModel,ClapPreTrainedModel:()=>t.ClapPreTrainedModel,ClapTextModelWithProjection:()=>t.ClapTextModelWithProjection,ClassifierFreeGuidanceLogitsProcessor:()=>$.ClassifierFreeGuidanceLogitsProcessor,CodeGenForCausalLM:()=>t.CodeGenForCausalLM,CodeGenModel:()=>t.CodeGenModel,CodeGenPreTrainedModel:()=>t.CodeGenPreTrainedModel,CodeGenTokenizer:()=>n.CodeGenTokenizer,CodeLlamaTokenizer:()=>n.CodeLlamaTokenizer,CohereForCausalLM:()=>t.CohereForCausalLM,CohereModel:()=>t.CohereModel,CoherePreTrainedModel:()=>t.CoherePreTrainedModel,CohereTokenizer:()=>n.CohereTokenizer,ConvBertForMaskedLM:()=>t.ConvBertForMaskedLM,ConvBertForQuestionAnswering:()=>t.ConvBertForQuestionAnswering,ConvBertForSequenceClassification:()=>t.ConvBertForSequenceClassification,ConvBertForTokenClassification:()=>t.ConvBertForTokenClassification,ConvBertModel:()=>t.ConvBertModel,ConvBertPreTrainedModel:()=>t.ConvBertPreTrainedModel,ConvBertTokenizer:()=>n.ConvBertTokenizer,ConvNextFeatureExtractor:()=>b.ConvNextFeatureExtractor,ConvNextForImageClassification:()=>t.ConvNextForImageClassification,ConvNextImageProcessor:()=>b.ConvNextImageProcessor,ConvNextModel:()=>t.ConvNextModel,ConvNextPreTrainedModel:()=>t.ConvNextPreTrainedModel,ConvNextV2ForImageClassification:()=>t.ConvNextV2ForImageClassification,ConvNextV2Model:()=>t.ConvNextV2Model,ConvNextV2PreTrainedModel:()=>t.ConvNextV2PreTrainedModel,DFineForObjectDetection:()=>t.DFineForObjectDetection,DFineModel:()=>t.DFineModel,DFinePreTrainedModel:()=>t.DFinePreTrainedModel,DPTFeatureExtractor:()=>b.DPTFeatureExtractor,DPTForDepthEstimation:()=>t.DPTForDepthEstimation,DPTImageProcessor:()=>b.DPTImageProcessor,DPTModel:()=>t.DPTModel,DPTPreTrainedModel:()=>t.DPTPreTrainedModel,DacDecoderModel:()=>t.DacDecoderModel,DacDecoderOutput:()=>t.DacDecoderOutput,DacEncoderModel:()=>t.DacEncoderModel,DacEncoderOutput:()=>t.DacEncoderOutput,DacFeatureExtractor:()=>m.DacFeatureExtractor,DacModel:()=>t.DacModel,DacPreTrainedModel:()=>t.DacPreTrainedModel,DataTypeMap:()=>h.DataTypeMap,DebertaForMaskedLM:()=>t.DebertaForMaskedLM,DebertaForQuestionAnswering:()=>t.DebertaForQuestionAnswering,DebertaForSequenceClassification:()=>t.DebertaForSequenceClassification,DebertaForTokenClassification:()=>t.DebertaForTokenClassification,DebertaModel:()=>t.DebertaModel,DebertaPreTrainedModel:()=>t.DebertaPreTrainedModel,DebertaTokenizer:()=>n.DebertaTokenizer,DebertaV2ForMaskedLM:()=>t.DebertaV2ForMaskedLM,DebertaV2ForQuestionAnswering:()=>t.DebertaV2ForQuestionAnswering,DebertaV2ForSequenceClassification:()=>t.DebertaV2ForSequenceClassification,DebertaV2ForTokenClassification:()=>t.DebertaV2ForTokenClassification,DebertaV2Model:()=>t.DebertaV2Model,DebertaV2PreTrainedModel:()=>t.DebertaV2PreTrainedModel,DebertaV2Tokenizer:()=>n.DebertaV2Tokenizer,DecisionTransformerModel:()=>t.DecisionTransformerModel,DecisionTransformerPreTrainedModel:()=>t.DecisionTransformerPreTrainedModel,DeiTFeatureExtractor:()=>b.DeiTFeatureExtractor,DeiTForImageClassification:()=>t.DeiTForImageClassification,DeiTImageProcessor:()=>b.DeiTImageProcessor,DeiTModel:()=>t.DeiTModel,DeiTPreTrainedModel:()=>t.DeiTPreTrainedModel,DepthAnythingForDepthEstimation:()=>t.DepthAnythingForDepthEstimation,DepthAnythingPreTrainedModel:()=>t.DepthAnythingPreTrainedModel,DepthEstimationPipeline:()=>r.DepthEstimationPipeline,DepthProForDepthEstimation:()=>t.DepthProForDepthEstimation,DepthProPreTrainedModel:()=>t.DepthProPreTrainedModel,DetrFeatureExtractor:()=>b.DetrFeatureExtractor,DetrForObjectDetection:()=>t.DetrForObjectDetection,DetrForSegmentation:()=>t.DetrForSegmentation,DetrImageProcessor:()=>b.DetrImageProcessor,DetrModel:()=>t.DetrModel,DetrObjectDetectionOutput:()=>t.DetrObjectDetectionOutput,DetrPreTrainedModel:()=>t.DetrPreTrainedModel,DetrSegmentationOutput:()=>t.DetrSegmentationOutput,Dinov2ForImageClassification:()=>t.Dinov2ForImageClassification,Dinov2Model:()=>t.Dinov2Model,Dinov2PreTrainedModel:()=>t.Dinov2PreTrainedModel,Dinov2WithRegistersForImageClassification:()=>t.Dinov2WithRegistersForImageClassification,Dinov2WithRegistersModel:()=>t.Dinov2WithRegistersModel,Dinov2WithRegistersPreTrainedModel:()=>t.Dinov2WithRegistersPreTrainedModel,DistilBertForMaskedLM:()=>t.DistilBertForMaskedLM,DistilBertForQuestionAnswering:()=>t.DistilBertForQuestionAnswering,DistilBertForSequenceClassification:()=>t.DistilBertForSequenceClassification,DistilBertForTokenClassification:()=>t.DistilBertForTokenClassification,DistilBertModel:()=>t.DistilBertModel,DistilBertPreTrainedModel:()=>t.DistilBertPreTrainedModel,DistilBertTokenizer:()=>n.DistilBertTokenizer,DocumentQuestionAnsweringPipeline:()=>r.DocumentQuestionAnsweringPipeline,DonutFeatureExtractor:()=>b.DonutFeatureExtractor,DonutImageProcessor:()=>b.DonutImageProcessor,DonutSwinModel:()=>t.DonutSwinModel,DonutSwinPreTrainedModel:()=>t.DonutSwinPreTrainedModel,EfficientNetForImageClassification:()=>t.EfficientNetForImageClassification,EfficientNetImageProcessor:()=>b.EfficientNetImageProcessor,EfficientNetModel:()=>t.EfficientNetModel,EfficientNetPreTrainedModel:()=>t.EfficientNetPreTrainedModel,ElectraForMaskedLM:()=>t.ElectraForMaskedLM,ElectraForQuestionAnswering:()=>t.ElectraForQuestionAnswering,ElectraForSequenceClassification:()=>t.ElectraForSequenceClassification,ElectraForTokenClassification:()=>t.ElectraForTokenClassification,ElectraModel:()=>t.ElectraModel,ElectraPreTrainedModel:()=>t.ElectraPreTrainedModel,ElectraTokenizer:()=>n.ElectraTokenizer,EncodecFeatureExtractor:()=>m.EncodecFeatureExtractor,EosTokenCriteria:()=>I.EosTokenCriteria,EsmForMaskedLM:()=>t.EsmForMaskedLM,EsmForSequenceClassification:()=>t.EsmForSequenceClassification,EsmForTokenClassification:()=>t.EsmForTokenClassification,EsmModel:()=>t.EsmModel,EsmPreTrainedModel:()=>t.EsmPreTrainedModel,EsmTokenizer:()=>n.EsmTokenizer,ExaoneForCausalLM:()=>t.ExaoneForCausalLM,ExaoneModel:()=>t.ExaoneModel,ExaonePreTrainedModel:()=>t.ExaonePreTrainedModel,FFT:()=>f.FFT,FalconForCausalLM:()=>t.FalconForCausalLM,FalconModel:()=>t.FalconModel,FalconPreTrainedModel:()=>t.FalconPreTrainedModel,FalconTokenizer:()=>n.FalconTokenizer,FastViTForImageClassification:()=>t.FastViTForImageClassification,FastViTModel:()=>t.FastViTModel,FastViTPreTrainedModel:()=>t.FastViTPreTrainedModel,FeatureExtractionPipeline:()=>r.FeatureExtractionPipeline,FeatureExtractor:()=>y.FeatureExtractor,FillMaskPipeline:()=>r.FillMaskPipeline,Florence2ForConditionalGeneration:()=>t.Florence2ForConditionalGeneration,Florence2PreTrainedModel:()=>t.Florence2PreTrainedModel,Florence2Processor:()=>E.Florence2Processor,ForcedBOSTokenLogitsProcessor:()=>$.ForcedBOSTokenLogitsProcessor,ForcedEOSTokenLogitsProcessor:()=>$.ForcedEOSTokenLogitsProcessor,GLPNFeatureExtractor:()=>b.GLPNFeatureExtractor,GLPNForDepthEstimation:()=>t.GLPNForDepthEstimation,GLPNModel:()=>t.GLPNModel,GLPNPreTrainedModel:()=>t.GLPNPreTrainedModel,GPT2LMHeadModel:()=>t.GPT2LMHeadModel,GPT2Model:()=>t.GPT2Model,GPT2PreTrainedModel:()=>t.GPT2PreTrainedModel,GPT2Tokenizer:()=>n.GPT2Tokenizer,GPTBigCodeForCausalLM:()=>t.GPTBigCodeForCausalLM,GPTBigCodeModel:()=>t.GPTBigCodeModel,GPTBigCodePreTrainedModel:()=>t.GPTBigCodePreTrainedModel,GPTJForCausalLM:()=>t.GPTJForCausalLM,GPTJModel:()=>t.GPTJModel,GPTJPreTrainedModel:()=>t.GPTJPreTrainedModel,GPTNeoForCausalLM:()=>t.GPTNeoForCausalLM,GPTNeoModel:()=>t.GPTNeoModel,GPTNeoPreTrainedModel:()=>t.GPTNeoPreTrainedModel,GPTNeoXForCausalLM:()=>t.GPTNeoXForCausalLM,GPTNeoXModel:()=>t.GPTNeoXModel,GPTNeoXPreTrainedModel:()=>t.GPTNeoXPreTrainedModel,GPTNeoXTokenizer:()=>n.GPTNeoXTokenizer,Gemma2ForCausalLM:()=>t.Gemma2ForCausalLM,Gemma2Model:()=>t.Gemma2Model,Gemma2PreTrainedModel:()=>t.Gemma2PreTrainedModel,Gemma3ForCausalLM:()=>t.Gemma3ForCausalLM,Gemma3Model:()=>t.Gemma3Model,Gemma3PreTrainedModel:()=>t.Gemma3PreTrainedModel,GemmaForCausalLM:()=>t.GemmaForCausalLM,GemmaModel:()=>t.GemmaModel,GemmaPreTrainedModel:()=>t.GemmaPreTrainedModel,GemmaTokenizer:()=>n.GemmaTokenizer,GlmForCausalLM:()=>t.GlmForCausalLM,GlmModel:()=>t.GlmModel,GlmPreTrainedModel:()=>t.GlmPreTrainedModel,GraniteForCausalLM:()=>t.GraniteForCausalLM,GraniteModel:()=>t.GraniteModel,GranitePreTrainedModel:()=>t.GranitePreTrainedModel,Grok1Tokenizer:()=>n.Grok1Tokenizer,GroundingDinoForObjectDetection:()=>t.GroundingDinoForObjectDetection,GroundingDinoImageProcessor:()=>b.GroundingDinoImageProcessor,GroundingDinoPreTrainedModel:()=>t.GroundingDinoPreTrainedModel,GroundingDinoProcessor:()=>E.GroundingDinoProcessor,GroupViTModel:()=>t.GroupViTModel,GroupViTPreTrainedModel:()=>t.GroupViTPreTrainedModel,HeliumForCausalLM:()=>t.HeliumForCausalLM,HeliumModel:()=>t.HeliumModel,HeliumPreTrainedModel:()=>t.HeliumPreTrainedModel,HerbertTokenizer:()=>n.HerbertTokenizer,HieraForImageClassification:()=>t.HieraForImageClassification,HieraModel:()=>t.HieraModel,HieraPreTrainedModel:()=>t.HieraPreTrainedModel,HubertForCTC:()=>t.HubertForCTC,HubertForSequenceClassification:()=>t.HubertForSequenceClassification,HubertModel:()=>t.HubertModel,HubertPreTrainedModel:()=>t.HubertPreTrainedModel,IJepaForImageClassification:()=>t.IJepaForImageClassification,IJepaModel:()=>t.IJepaModel,IJepaPreTrainedModel:()=>t.IJepaPreTrainedModel,Idefics3ForConditionalGeneration:()=>t.Idefics3ForConditionalGeneration,Idefics3ImageProcessor:()=>b.Idefics3ImageProcessor,Idefics3PreTrainedModel:()=>t.Idefics3PreTrainedModel,Idefics3Processor:()=>E.Idefics3Processor,ImageClassificationPipeline:()=>r.ImageClassificationPipeline,ImageFeatureExtractionPipeline:()=>r.ImageFeatureExtractionPipeline,ImageFeatureExtractor:()=>m.ImageFeatureExtractor,ImageMattingOutput:()=>t.ImageMattingOutput,ImageProcessor:()=>v.ImageProcessor,ImageSegmentationPipeline:()=>r.ImageSegmentationPipeline,ImageToImagePipeline:()=>r.ImageToImagePipeline,ImageToTextPipeline:()=>r.ImageToTextPipeline,InterruptableStoppingCriteria:()=>I.InterruptableStoppingCriteria,JAISLMHeadModel:()=>t.JAISLMHeadModel,JAISModel:()=>t.JAISModel,JAISPreTrainedModel:()=>t.JAISPreTrainedModel,JinaCLIPImageProcessor:()=>b.JinaCLIPImageProcessor,JinaCLIPModel:()=>t.JinaCLIPModel,JinaCLIPPreTrainedModel:()=>t.JinaCLIPPreTrainedModel,JinaCLIPProcessor:()=>E.JinaCLIPProcessor,JinaCLIPTextModel:()=>t.JinaCLIPTextModel,JinaCLIPVisionModel:()=>t.JinaCLIPVisionModel,LiteWhisperForConditionalGeneration:()=>t.LiteWhisperForConditionalGeneration,LlamaForCausalLM:()=>t.LlamaForCausalLM,LlamaModel:()=>t.LlamaModel,LlamaPreTrainedModel:()=>t.LlamaPreTrainedModel,LlamaTokenizer:()=>n.LlamaTokenizer,LlavaForConditionalGeneration:()=>t.LlavaForConditionalGeneration,LlavaOnevisionForConditionalGeneration:()=>t.LlavaOnevisionForConditionalGeneration,LlavaOnevisionImageProcessor:()=>b.LlavaOnevisionImageProcessor,LlavaPreTrainedModel:()=>t.LlavaPreTrainedModel,LogitsProcessor:()=>$.LogitsProcessor,LogitsProcessorList:()=>$.LogitsProcessorList,LogitsWarper:()=>$.LogitsWarper,LongT5ForConditionalGeneration:()=>t.LongT5ForConditionalGeneration,LongT5Model:()=>t.LongT5Model,LongT5PreTrainedModel:()=>t.LongT5PreTrainedModel,M2M100ForConditionalGeneration:()=>t.M2M100ForConditionalGeneration,M2M100Model:()=>t.M2M100Model,M2M100PreTrainedModel:()=>t.M2M100PreTrainedModel,M2M100Tokenizer:()=>n.M2M100Tokenizer,MBart50Tokenizer:()=>n.MBart50Tokenizer,MBartForCausalLM:()=>t.MBartForCausalLM,MBartForConditionalGeneration:()=>t.MBartForConditionalGeneration,MBartForSequenceClassification:()=>t.MBartForSequenceClassification,MBartModel:()=>t.MBartModel,MBartPreTrainedModel:()=>t.MBartPreTrainedModel,MBartTokenizer:()=>n.MBartTokenizer,MPNetForMaskedLM:()=>t.MPNetForMaskedLM,MPNetForQuestionAnswering:()=>t.MPNetForQuestionAnswering,MPNetForSequenceClassification:()=>t.MPNetForSequenceClassification,MPNetForTokenClassification:()=>t.MPNetForTokenClassification,MPNetModel:()=>t.MPNetModel,MPNetPreTrainedModel:()=>t.MPNetPreTrainedModel,MPNetTokenizer:()=>n.MPNetTokenizer,MT5ForConditionalGeneration:()=>t.MT5ForConditionalGeneration,MT5Model:()=>t.MT5Model,MT5PreTrainedModel:()=>t.MT5PreTrainedModel,MarianMTModel:()=>t.MarianMTModel,MarianModel:()=>t.MarianModel,MarianPreTrainedModel:()=>t.MarianPreTrainedModel,MarianTokenizer:()=>n.MarianTokenizer,Mask2FormerImageProcessor:()=>b.Mask2FormerImageProcessor,MaskFormerFeatureExtractor:()=>b.MaskFormerFeatureExtractor,MaskFormerForInstanceSegmentation:()=>t.MaskFormerForInstanceSegmentation,MaskFormerImageProcessor:()=>b.MaskFormerImageProcessor,MaskFormerModel:()=>t.MaskFormerModel,MaskFormerPreTrainedModel:()=>t.MaskFormerPreTrainedModel,MaskedLMOutput:()=>t.MaskedLMOutput,MaxLengthCriteria:()=>I.MaxLengthCriteria,Metric3DForDepthEstimation:()=>t.Metric3DForDepthEstimation,Metric3DPreTrainedModel:()=>t.Metric3DPreTrainedModel,Metric3Dv2ForDepthEstimation:()=>t.Metric3Dv2ForDepthEstimation,Metric3Dv2PreTrainedModel:()=>t.Metric3Dv2PreTrainedModel,MgpstrForSceneTextRecognition:()=>t.MgpstrForSceneTextRecognition,MgpstrModelOutput:()=>t.MgpstrModelOutput,MgpstrPreTrainedModel:()=>t.MgpstrPreTrainedModel,MgpstrProcessor:()=>E.MgpstrProcessor,MgpstrTokenizer:()=>n.MgpstrTokenizer,MimiDecoderModel:()=>t.MimiDecoderModel,MimiDecoderOutput:()=>t.MimiDecoderOutput,MimiEncoderModel:()=>t.MimiEncoderModel,MimiEncoderOutput:()=>t.MimiEncoderOutput,MimiModel:()=>t.MimiModel,MimiPreTrainedModel:()=>t.MimiPreTrainedModel,MinLengthLogitsProcessor:()=>$.MinLengthLogitsProcessor,MinNewTokensLengthLogitsProcessor:()=>$.MinNewTokensLengthLogitsProcessor,MistralForCausalLM:()=>t.MistralForCausalLM,MistralModel:()=>t.MistralModel,MistralPreTrainedModel:()=>t.MistralPreTrainedModel,MobileBertForMaskedLM:()=>t.MobileBertForMaskedLM,MobileBertForQuestionAnswering:()=>t.MobileBertForQuestionAnswering,MobileBertForSequenceClassification:()=>t.MobileBertForSequenceClassification,MobileBertModel:()=>t.MobileBertModel,MobileBertPreTrainedModel:()=>t.MobileBertPreTrainedModel,MobileBertTokenizer:()=>n.MobileBertTokenizer,MobileLLMForCausalLM:()=>t.MobileLLMForCausalLM,MobileLLMModel:()=>t.MobileLLMModel,MobileLLMPreTrainedModel:()=>t.MobileLLMPreTrainedModel,MobileNetV1FeatureExtractor:()=>b.MobileNetV1FeatureExtractor,MobileNetV1ForImageClassification:()=>t.MobileNetV1ForImageClassification,MobileNetV1ForSemanticSegmentation:()=>t.MobileNetV1ForSemanticSegmentation,MobileNetV1ImageProcessor:()=>b.MobileNetV1ImageProcessor,MobileNetV1Model:()=>t.MobileNetV1Model,MobileNetV1PreTrainedModel:()=>t.MobileNetV1PreTrainedModel,MobileNetV2FeatureExtractor:()=>b.MobileNetV2FeatureExtractor,MobileNetV2ForImageClassification:()=>t.MobileNetV2ForImageClassification,MobileNetV2ForSemanticSegmentation:()=>t.MobileNetV2ForSemanticSegmentation,MobileNetV2ImageProcessor:()=>b.MobileNetV2ImageProcessor,MobileNetV2Model:()=>t.MobileNetV2Model,MobileNetV2PreTrainedModel:()=>t.MobileNetV2PreTrainedModel,MobileNetV3FeatureExtractor:()=>b.MobileNetV3FeatureExtractor,MobileNetV3ForImageClassification:()=>t.MobileNetV3ForImageClassification,MobileNetV3ForSemanticSegmentation:()=>t.MobileNetV3ForSemanticSegmentation,MobileNetV3ImageProcessor:()=>b.MobileNetV3ImageProcessor,MobileNetV3Model:()=>t.MobileNetV3Model,MobileNetV3PreTrainedModel:()=>t.MobileNetV3PreTrainedModel,MobileNetV4FeatureExtractor:()=>b.MobileNetV4FeatureExtractor,MobileNetV4ForImageClassification:()=>t.MobileNetV4ForImageClassification,MobileNetV4ForSemanticSegmentation:()=>t.MobileNetV4ForSemanticSegmentation,MobileNetV4ImageProcessor:()=>b.MobileNetV4ImageProcessor,MobileNetV4Model:()=>t.MobileNetV4Model,MobileNetV4PreTrainedModel:()=>t.MobileNetV4PreTrainedModel,MobileViTFeatureExtractor:()=>b.MobileViTFeatureExtractor,MobileViTForImageClassification:()=>t.MobileViTForImageClassification,MobileViTImageProcessor:()=>b.MobileViTImageProcessor,MobileViTModel:()=>t.MobileViTModel,MobileViTPreTrainedModel:()=>t.MobileViTPreTrainedModel,MobileViTV2ForImageClassification:()=>t.MobileViTV2ForImageClassification,MobileViTV2Model:()=>t.MobileViTV2Model,MobileViTV2PreTrainedModel:()=>t.MobileViTV2PreTrainedModel,ModelOutput:()=>t.ModelOutput,ModernBertForMaskedLM:()=>t.ModernBertForMaskedLM,ModernBertForSequenceClassification:()=>t.ModernBertForSequenceClassification,ModernBertForTokenClassification:()=>t.ModernBertForTokenClassification,ModernBertModel:()=>t.ModernBertModel,ModernBertPreTrainedModel:()=>t.ModernBertPreTrainedModel,Moondream1ForConditionalGeneration:()=>t.Moondream1ForConditionalGeneration,MoonshineFeatureExtractor:()=>m.MoonshineFeatureExtractor,MoonshineForConditionalGeneration:()=>t.MoonshineForConditionalGeneration,MoonshineModel:()=>t.MoonshineModel,MoonshinePreTrainedModel:()=>t.MoonshinePreTrainedModel,MoonshineProcessor:()=>E.MoonshineProcessor,MptForCausalLM:()=>t.MptForCausalLM,MptModel:()=>t.MptModel,MptPreTrainedModel:()=>t.MptPreTrainedModel,MultiModalityCausalLM:()=>t.MultiModalityCausalLM,MultiModalityPreTrainedModel:()=>t.MultiModalityPreTrainedModel,MusicgenForCausalLM:()=>t.MusicgenForCausalLM,MusicgenForConditionalGeneration:()=>t.MusicgenForConditionalGeneration,MusicgenModel:()=>t.MusicgenModel,MusicgenPreTrainedModel:()=>t.MusicgenPreTrainedModel,NllbTokenizer:()=>n.NllbTokenizer,NoBadWordsLogitsProcessor:()=>$.NoBadWordsLogitsProcessor,NoRepeatNGramLogitsProcessor:()=>$.NoRepeatNGramLogitsProcessor,NomicBertModel:()=>t.NomicBertModel,NomicBertPreTrainedModel:()=>t.NomicBertPreTrainedModel,NougatImageProcessor:()=>b.NougatImageProcessor,NougatTokenizer:()=>n.NougatTokenizer,OPTForCausalLM:()=>t.OPTForCausalLM,OPTModel:()=>t.OPTModel,OPTPreTrainedModel:()=>t.OPTPreTrainedModel,ObjectDetectionPipeline:()=>r.ObjectDetectionPipeline,Olmo2ForCausalLM:()=>t.Olmo2ForCausalLM,Olmo2Model:()=>t.Olmo2Model,Olmo2PreTrainedModel:()=>t.Olmo2PreTrainedModel,OlmoForCausalLM:()=>t.OlmoForCausalLM,OlmoModel:()=>t.OlmoModel,OlmoPreTrainedModel:()=>t.OlmoPreTrainedModel,OpenELMForCausalLM:()=>t.OpenELMForCausalLM,OpenELMModel:()=>t.OpenELMModel,OpenELMPreTrainedModel:()=>t.OpenELMPreTrainedModel,OwlViTFeatureExtractor:()=>b.OwlViTFeatureExtractor,OwlViTForObjectDetection:()=>t.OwlViTForObjectDetection,OwlViTImageProcessor:()=>b.OwlViTImageProcessor,OwlViTModel:()=>t.OwlViTModel,OwlViTPreTrainedModel:()=>t.OwlViTPreTrainedModel,OwlViTProcessor:()=>E.OwlViTProcessor,Owlv2ForObjectDetection:()=>t.Owlv2ForObjectDetection,Owlv2ImageProcessor:()=>b.Owlv2ImageProcessor,Owlv2Model:()=>t.Owlv2Model,Owlv2PreTrainedModel:()=>t.Owlv2PreTrainedModel,PaliGemmaForConditionalGeneration:()=>t.PaliGemmaForConditionalGeneration,PaliGemmaPreTrainedModel:()=>t.PaliGemmaPreTrainedModel,PaliGemmaProcessor:()=>E.PaliGemmaProcessor,PatchTSMixerForPrediction:()=>t.PatchTSMixerForPrediction,PatchTSMixerModel:()=>t.PatchTSMixerModel,PatchTSMixerPreTrainedModel:()=>t.PatchTSMixerPreTrainedModel,PatchTSTForPrediction:()=>t.PatchTSTForPrediction,PatchTSTModel:()=>t.PatchTSTModel,PatchTSTPreTrainedModel:()=>t.PatchTSTPreTrainedModel,Phi3ForCausalLM:()=>t.Phi3ForCausalLM,Phi3Model:()=>t.Phi3Model,Phi3PreTrainedModel:()=>t.Phi3PreTrainedModel,Phi3VForCausalLM:()=>t.Phi3VForCausalLM,Phi3VImageProcessor:()=>b.Phi3VImageProcessor,Phi3VPreTrainedModel:()=>t.Phi3VPreTrainedModel,Phi3VProcessor:()=>E.Phi3VProcessor,PhiForCausalLM:()=>t.PhiForCausalLM,PhiModel:()=>t.PhiModel,PhiPreTrainedModel:()=>t.PhiPreTrainedModel,Pipeline:()=>r.Pipeline,PreTrainedModel:()=>t.PreTrainedModel,PreTrainedTokenizer:()=>n.PreTrainedTokenizer,PretrainedConfig:()=>o.PretrainedConfig,PretrainedMixin:()=>t.PretrainedMixin,Processor:()=>z.Processor,PvtForImageClassification:()=>t.PvtForImageClassification,PvtImageProcessor:()=>b.PvtImageProcessor,PvtModel:()=>t.PvtModel,PvtPreTrainedModel:()=>t.PvtPreTrainedModel,PyAnnoteFeatureExtractor:()=>m.PyAnnoteFeatureExtractor,PyAnnoteForAudioFrameClassification:()=>t.PyAnnoteForAudioFrameClassification,PyAnnoteModel:()=>t.PyAnnoteModel,PyAnnotePreTrainedModel:()=>t.PyAnnotePreTrainedModel,PyAnnoteProcessor:()=>E.PyAnnoteProcessor,QuestionAnsweringModelOutput:()=>t.QuestionAnsweringModelOutput,QuestionAnsweringPipeline:()=>r.QuestionAnsweringPipeline,Qwen2ForCausalLM:()=>t.Qwen2ForCausalLM,Qwen2Model:()=>t.Qwen2Model,Qwen2PreTrainedModel:()=>t.Qwen2PreTrainedModel,Qwen2Tokenizer:()=>n.Qwen2Tokenizer,Qwen2VLForConditionalGeneration:()=>t.Qwen2VLForConditionalGeneration,Qwen2VLImageProcessor:()=>b.Qwen2VLImageProcessor,Qwen2VLPreTrainedModel:()=>t.Qwen2VLPreTrainedModel,Qwen2VLProcessor:()=>E.Qwen2VLProcessor,Qwen3ForCausalLM:()=>t.Qwen3ForCausalLM,Qwen3Model:()=>t.Qwen3Model,Qwen3PreTrainedModel:()=>t.Qwen3PreTrainedModel,RFDetrForObjectDetection:()=>t.RFDetrForObjectDetection,RFDetrModel:()=>t.RFDetrModel,RFDetrObjectDetectionOutput:()=>t.RFDetrObjectDetectionOutput,RFDetrPreTrainedModel:()=>t.RFDetrPreTrainedModel,RTDetrForObjectDetection:()=>t.RTDetrForObjectDetection,RTDetrImageProcessor:()=>b.RTDetrImageProcessor,RTDetrModel:()=>t.RTDetrModel,RTDetrObjectDetectionOutput:()=>t.RTDetrObjectDetectionOutput,RTDetrPreTrainedModel:()=>t.RTDetrPreTrainedModel,RTDetrV2ForObjectDetection:()=>t.RTDetrV2ForObjectDetection,RTDetrV2Model:()=>t.RTDetrV2Model,RTDetrV2ObjectDetectionOutput:()=>t.RTDetrV2ObjectDetectionOutput,RTDetrV2PreTrainedModel:()=>t.RTDetrV2PreTrainedModel,RawAudio:()=>i.RawAudio,RawImage:()=>u.RawImage,RawVideo:()=>d.RawVideo,RawVideoFrame:()=>d.RawVideoFrame,RepetitionPenaltyLogitsProcessor:()=>$.RepetitionPenaltyLogitsProcessor,ResNetForImageClassification:()=>t.ResNetForImageClassification,ResNetModel:()=>t.ResNetModel,ResNetPreTrainedModel:()=>t.ResNetPreTrainedModel,RoFormerForMaskedLM:()=>t.RoFormerForMaskedLM,RoFormerForQuestionAnswering:()=>t.RoFormerForQuestionAnswering,RoFormerForSequenceClassification:()=>t.RoFormerForSequenceClassification,RoFormerForTokenClassification:()=>t.RoFormerForTokenClassification,RoFormerModel:()=>t.RoFormerModel,RoFormerPreTrainedModel:()=>t.RoFormerPreTrainedModel,RoFormerTokenizer:()=>n.RoFormerTokenizer,RobertaForMaskedLM:()=>t.RobertaForMaskedLM,RobertaForQuestionAnswering:()=>t.RobertaForQuestionAnswering,RobertaForSequenceClassification:()=>t.RobertaForSequenceClassification,RobertaForTokenClassification:()=>t.RobertaForTokenClassification,RobertaModel:()=>t.RobertaModel,RobertaPreTrainedModel:()=>t.RobertaPreTrainedModel,RobertaTokenizer:()=>n.RobertaTokenizer,SamImageProcessor:()=>b.SamImageProcessor,SamImageSegmentationOutput:()=>t.SamImageSegmentationOutput,SamModel:()=>t.SamModel,SamPreTrainedModel:()=>t.SamPreTrainedModel,SamProcessor:()=>E.SamProcessor,SapiensForDepthEstimation:()=>t.SapiensForDepthEstimation,SapiensForNormalEstimation:()=>t.SapiensForNormalEstimation,SapiensForSemanticSegmentation:()=>t.SapiensForSemanticSegmentation,SapiensPreTrainedModel:()=>t.SapiensPreTrainedModel,SeamlessM4TFeatureExtractor:()=>m.SeamlessM4TFeatureExtractor,SegformerFeatureExtractor:()=>b.SegformerFeatureExtractor,SegformerForImageClassification:()=>t.SegformerForImageClassification,SegformerForSemanticSegmentation:()=>t.SegformerForSemanticSegmentation,SegformerImageProcessor:()=>b.SegformerImageProcessor,SegformerModel:()=>t.SegformerModel,SegformerPreTrainedModel:()=>t.SegformerPreTrainedModel,Seq2SeqLMOutput:()=>t.Seq2SeqLMOutput,SequenceClassifierOutput:()=>t.SequenceClassifierOutput,SiglipImageProcessor:()=>b.SiglipImageProcessor,SiglipModel:()=>t.SiglipModel,SiglipPreTrainedModel:()=>t.SiglipPreTrainedModel,SiglipTextModel:()=>t.SiglipTextModel,SiglipTokenizer:()=>n.SiglipTokenizer,SiglipVisionModel:()=>t.SiglipVisionModel,SmolVLMForConditionalGeneration:()=>t.SmolVLMForConditionalGeneration,SmolVLMImageProcessor:()=>b.SmolVLMImageProcessor,SmolVLMProcessor:()=>E.SmolVLMProcessor,SnacDecoderModel:()=>t.SnacDecoderModel,SnacEncoderModel:()=>t.SnacEncoderModel,SnacFeatureExtractor:()=>m.SnacFeatureExtractor,SnacModel:()=>t.SnacModel,SnacPreTrainedModel:()=>t.SnacPreTrainedModel,SpeechT5FeatureExtractor:()=>m.SpeechT5FeatureExtractor,SpeechT5ForSpeechToText:()=>t.SpeechT5ForSpeechToText,SpeechT5ForTextToSpeech:()=>t.SpeechT5ForTextToSpeech,SpeechT5HifiGan:()=>t.SpeechT5HifiGan,SpeechT5Model:()=>t.SpeechT5Model,SpeechT5PreTrainedModel:()=>t.SpeechT5PreTrainedModel,SpeechT5Processor:()=>E.SpeechT5Processor,SpeechT5Tokenizer:()=>n.SpeechT5Tokenizer,SqueezeBertForMaskedLM:()=>t.SqueezeBertForMaskedLM,SqueezeBertForQuestionAnswering:()=>t.SqueezeBertForQuestionAnswering,SqueezeBertForSequenceClassification:()=>t.SqueezeBertForSequenceClassification,SqueezeBertModel:()=>t.SqueezeBertModel,SqueezeBertPreTrainedModel:()=>t.SqueezeBertPreTrainedModel,SqueezeBertTokenizer:()=>n.SqueezeBertTokenizer,StableLmForCausalLM:()=>t.StableLmForCausalLM,StableLmModel:()=>t.StableLmModel,StableLmPreTrainedModel:()=>t.StableLmPreTrainedModel,Starcoder2ForCausalLM:()=>t.Starcoder2ForCausalLM,Starcoder2Model:()=>t.Starcoder2Model,Starcoder2PreTrainedModel:()=>t.Starcoder2PreTrainedModel,StoppingCriteria:()=>I.StoppingCriteria,StoppingCriteriaList:()=>I.StoppingCriteriaList,StyleTextToSpeech2Model:()=>t.StyleTextToSpeech2Model,StyleTextToSpeech2PreTrainedModel:()=>t.StyleTextToSpeech2PreTrainedModel,SummarizationPipeline:()=>r.SummarizationPipeline,SuppressTokensAtBeginLogitsProcessor:()=>$.SuppressTokensAtBeginLogitsProcessor,Swin2SRForImageSuperResolution:()=>t.Swin2SRForImageSuperResolution,Swin2SRImageProcessor:()=>b.Swin2SRImageProcessor,Swin2SRModel:()=>t.Swin2SRModel,Swin2SRPreTrainedModel:()=>t.Swin2SRPreTrainedModel,SwinForImageClassification:()=>t.SwinForImageClassification,SwinForSemanticSegmentation:()=>t.SwinForSemanticSegmentation,SwinModel:()=>t.SwinModel,SwinPreTrainedModel:()=>t.SwinPreTrainedModel,T5ForConditionalGeneration:()=>t.T5ForConditionalGeneration,T5Model:()=>t.T5Model,T5PreTrainedModel:()=>t.T5PreTrainedModel,T5Tokenizer:()=>n.T5Tokenizer,TableTransformerForObjectDetection:()=>t.TableTransformerForObjectDetection,TableTransformerModel:()=>t.TableTransformerModel,TableTransformerObjectDetectionOutput:()=>t.TableTransformerObjectDetectionOutput,TableTransformerPreTrainedModel:()=>t.TableTransformerPreTrainedModel,TemperatureLogitsWarper:()=>$.TemperatureLogitsWarper,Tensor:()=>h.Tensor,Text2TextGenerationPipeline:()=>r.Text2TextGenerationPipeline,TextClassificationPipeline:()=>r.TextClassificationPipeline,TextGenerationPipeline:()=>r.TextGenerationPipeline,TextStreamer:()=>D.TextStreamer,TextToAudioPipeline:()=>r.TextToAudioPipeline,TokenClassificationPipeline:()=>r.TokenClassificationPipeline,TokenClassifierOutput:()=>t.TokenClassifierOutput,TokenizerModel:()=>n.TokenizerModel,TopKLogitsWarper:()=>$.TopKLogitsWarper,TopPLogitsWarper:()=>$.TopPLogitsWarper,TrOCRForCausalLM:()=>t.TrOCRForCausalLM,TrOCRPreTrainedModel:()=>t.TrOCRPreTrainedModel,TranslationPipeline:()=>r.TranslationPipeline,UltravoxModel:()=>t.UltravoxModel,UltravoxPreTrainedModel:()=>t.UltravoxPreTrainedModel,UltravoxProcessor:()=>E.UltravoxProcessor,UniSpeechForCTC:()=>t.UniSpeechForCTC,UniSpeechForSequenceClassification:()=>t.UniSpeechForSequenceClassification,UniSpeechModel:()=>t.UniSpeechModel,UniSpeechPreTrainedModel:()=>t.UniSpeechPreTrainedModel,UniSpeechSatForAudioFrameClassification:()=>t.UniSpeechSatForAudioFrameClassification,UniSpeechSatForCTC:()=>t.UniSpeechSatForCTC,UniSpeechSatForSequenceClassification:()=>t.UniSpeechSatForSequenceClassification,UniSpeechSatModel:()=>t.UniSpeechSatModel,UniSpeechSatPreTrainedModel:()=>t.UniSpeechSatPreTrainedModel,VLChatProcessor:()=>E.VLChatProcessor,VLMImageProcessor:()=>b.VLMImageProcessor,ViTFeatureExtractor:()=>b.ViTFeatureExtractor,ViTForImageClassification:()=>t.ViTForImageClassification,ViTImageProcessor:()=>b.ViTImageProcessor,ViTMAEModel:()=>t.ViTMAEModel,ViTMAEPreTrainedModel:()=>t.ViTMAEPreTrainedModel,ViTMSNForImageClassification:()=>t.ViTMSNForImageClassification,ViTMSNModel:()=>t.ViTMSNModel,ViTMSNPreTrainedModel:()=>t.ViTMSNPreTrainedModel,ViTModel:()=>t.ViTModel,ViTPreTrainedModel:()=>t.ViTPreTrainedModel,VisionEncoderDecoderModel:()=>t.VisionEncoderDecoderModel,VitMatteForImageMatting:()=>t.VitMatteForImageMatting,VitMatteImageProcessor:()=>b.VitMatteImageProcessor,VitMattePreTrainedModel:()=>t.VitMattePreTrainedModel,VitPoseForPoseEstimation:()=>t.VitPoseForPoseEstimation,VitPoseImageProcessor:()=>b.VitPoseImageProcessor,VitPosePreTrainedModel:()=>t.VitPosePreTrainedModel,VitsModel:()=>t.VitsModel,VitsModelOutput:()=>t.VitsModelOutput,VitsPreTrainedModel:()=>t.VitsPreTrainedModel,VitsTokenizer:()=>n.VitsTokenizer,Wav2Vec2BertForCTC:()=>t.Wav2Vec2BertForCTC,Wav2Vec2BertForSequenceClassification:()=>t.Wav2Vec2BertForSequenceClassification,Wav2Vec2BertModel:()=>t.Wav2Vec2BertModel,Wav2Vec2BertPreTrainedModel:()=>t.Wav2Vec2BertPreTrainedModel,Wav2Vec2CTCTokenizer:()=>n.Wav2Vec2CTCTokenizer,Wav2Vec2FeatureExtractor:()=>m.Wav2Vec2FeatureExtractor,Wav2Vec2ForAudioFrameClassification:()=>t.Wav2Vec2ForAudioFrameClassification,Wav2Vec2ForCTC:()=>t.Wav2Vec2ForCTC,Wav2Vec2ForSequenceClassification:()=>t.Wav2Vec2ForSequenceClassification,Wav2Vec2Model:()=>t.Wav2Vec2Model,Wav2Vec2PreTrainedModel:()=>t.Wav2Vec2PreTrainedModel,Wav2Vec2Processor:()=>E.Wav2Vec2Processor,Wav2Vec2ProcessorWithLM:()=>E.Wav2Vec2ProcessorWithLM,WavLMForAudioFrameClassification:()=>t.WavLMForAudioFrameClassification,WavLMForCTC:()=>t.WavLMForCTC,WavLMForSequenceClassification:()=>t.WavLMForSequenceClassification,WavLMForXVector:()=>t.WavLMForXVector,WavLMModel:()=>t.WavLMModel,WavLMPreTrainedModel:()=>t.WavLMPreTrainedModel,WeSpeakerFeatureExtractor:()=>m.WeSpeakerFeatureExtractor,WeSpeakerResNetModel:()=>t.WeSpeakerResNetModel,WeSpeakerResNetPreTrainedModel:()=>t.WeSpeakerResNetPreTrainedModel,WhisperFeatureExtractor:()=>m.WhisperFeatureExtractor,WhisperForConditionalGeneration:()=>t.WhisperForConditionalGeneration,WhisperModel:()=>t.WhisperModel,WhisperPreTrainedModel:()=>t.WhisperPreTrainedModel,WhisperProcessor:()=>E.WhisperProcessor,WhisperTextStreamer:()=>D.WhisperTextStreamer,WhisperTimeStampLogitsProcessor:()=>$.WhisperTimeStampLogitsProcessor,WhisperTokenizer:()=>n.WhisperTokenizer,XLMForQuestionAnswering:()=>t.XLMForQuestionAnswering,XLMForSequenceClassification:()=>t.XLMForSequenceClassification,XLMForTokenClassification:()=>t.XLMForTokenClassification,XLMModel:()=>t.XLMModel,XLMPreTrainedModel:()=>t.XLMPreTrainedModel,XLMRobertaForMaskedLM:()=>t.XLMRobertaForMaskedLM,XLMRobertaForQuestionAnswering:()=>t.XLMRobertaForQuestionAnswering,XLMRobertaForSequenceClassification:()=>t.XLMRobertaForSequenceClassification,XLMRobertaForTokenClassification:()=>t.XLMRobertaForTokenClassification,XLMRobertaModel:()=>t.XLMRobertaModel,XLMRobertaPreTrainedModel:()=>t.XLMRobertaPreTrainedModel,XLMRobertaTokenizer:()=>n.XLMRobertaTokenizer,XLMTokenizer:()=>n.XLMTokenizer,XLMWithLMHeadModel:()=>t.XLMWithLMHeadModel,XVectorOutput:()=>t.XVectorOutput,YolosFeatureExtractor:()=>b.YolosFeatureExtractor,YolosForObjectDetection:()=>t.YolosForObjectDetection,YolosImageProcessor:()=>b.YolosImageProcessor,YolosModel:()=>t.YolosModel,YolosObjectDetectionOutput:()=>t.YolosObjectDetectionOutput,YolosPreTrainedModel:()=>t.YolosPreTrainedModel,ZeroShotAudioClassificationPipeline:()=>r.ZeroShotAudioClassificationPipeline,ZeroShotClassificationPipeline:()=>r.ZeroShotClassificationPipeline,ZeroShotImageClassificationPipeline:()=>r.ZeroShotImageClassificationPipeline,ZeroShotObjectDetectionPipeline:()=>r.ZeroShotObjectDetectionPipeline,bankers_round:()=>f.bankers_round,cat:()=>h.cat,cos_sim:()=>f.cos_sim,dot:()=>f.dot,dynamic_time_warping:()=>f.dynamic_time_warping,env:()=>e.env,full:()=>h.full,full_like:()=>h.full_like,getKeyValueShapes:()=>o.getKeyValueShapes,hamming:()=>i.hamming,hanning:()=>i.hanning,interpolate:()=>h.interpolate,interpolate_4d:()=>h.interpolate_4d,interpolate_data:()=>f.interpolate_data,is_chinese_char:()=>n.is_chinese_char,layer_norm:()=>h.layer_norm,load_image:()=>u.load_image,load_video:()=>d.load_video,log_softmax:()=>f.log_softmax,magnitude:()=>f.magnitude,matmul:()=>h.matmul,max:()=>f.max,mean:()=>h.mean,mean_pooling:()=>h.mean_pooling,medianFilter:()=>f.medianFilter,mel_filter_bank:()=>i.mel_filter_bank,min:()=>f.min,ones:()=>h.ones,ones_like:()=>h.ones_like,permute:()=>h.permute,permute_data:()=>f.permute_data,pipeline:()=>r.pipeline,quantize_embeddings:()=>h.quantize_embeddings,rand:()=>h.rand,read_audio:()=>i.read_audio,rfft:()=>h.rfft,round:()=>f.round,slice:()=>h.slice,softmax:()=>f.softmax,spectrogram:()=>i.spectrogram,stack:()=>h.stack,std_mean:()=>h.std_mean,topk:()=>h.topk,window_function:()=>i.window_function,zeros:()=>h.zeros,zeros_like:()=>h.zeros_like});var e=Yt("./src/env.js"),r=Yt("./src/pipelines.js"),t=Yt("./src/models.js"),n=Yt("./src/tokenizers.js"),o=Yt("./src/configs.js"),i=Yt("./src/utils/audio.js"),u=Yt("./src/utils/image.js"),d=Yt("./src/utils/video.js"),h=Yt("./src/utils/tensor.js"),f=Yt("./src/utils/maths.js"),y=Yt("./src/base/feature_extraction_utils.js"),m=Yt("./src/models/feature_extractors.js"),g=Yt("./src/models/auto/feature_extraction_auto.js"),v=Yt("./src/base/image_processors_utils.js"),b=Yt("./src/models/image_processors.js"),k=Yt("./src/models/auto/image_processing_auto.js"),z=Yt("./src/base/processing_utils.js"),E=Yt("./src/models/processors.js"),T=Yt("./src/models/auto/processing_auto.js"),D=Yt("./src/generation/streamers.js"),I=Yt("./src/generation/stopping_criteria.js"),$=Yt("./src/generation/logits_process.js")})();w.ASTFeatureExtractor;w.ASTForAudioClassification;w.ASTModel;w.ASTPreTrainedModel;w.AlbertForMaskedLM;w.AlbertForQuestionAnswering;w.AlbertForSequenceClassification;w.AlbertModel;w.AlbertPreTrainedModel;w.AlbertTokenizer;w.AudioClassificationPipeline;w.AutoConfig;w.AutoFeatureExtractor;w.AutoImageProcessor;const sE=w.AutoModel;w.AutoModelForAudioClassification;w.AutoModelForAudioFrameClassification;w.AutoModelForAudioTextToText;w.AutoModelForCTC;w.AutoModelForCausalLM;w.AutoModelForDepthEstimation;w.AutoModelForDocumentQuestionAnswering;w.AutoModelForImageClassification;w.AutoModelForImageFeatureExtraction;w.AutoModelForImageMatting;w.AutoModelForImageSegmentation;w.AutoModelForImageTextToText;w.AutoModelForImageToImage;w.AutoModelForMaskGeneration;w.AutoModelForMaskedLM;w.AutoModelForNormalEstimation;w.AutoModelForObjectDetection;w.AutoModelForPoseEstimation;w.AutoModelForQuestionAnswering;w.AutoModelForSemanticSegmentation;w.AutoModelForSeq2SeqLM;w.AutoModelForSequenceClassification;w.AutoModelForSpeechSeq2Seq;w.AutoModelForTextToSpectrogram;w.AutoModelForTextToWaveform;w.AutoModelForTokenClassification;w.AutoModelForUniversalSegmentation;w.AutoModelForVision2Seq;w.AutoModelForXVector;w.AutoModelForZeroShotObjectDetection;w.AutoProcessor;w.AutoTokenizer;w.AutomaticSpeechRecognitionPipeline;w.BackgroundRemovalPipeline;w.BartForConditionalGeneration;w.BartForSequenceClassification;w.BartModel;w.BartPretrainedModel;w.BartTokenizer;w.BaseModelOutput;w.BaseStreamer;w.BeitFeatureExtractor;w.BeitForImageClassification;w.BeitModel;w.BeitPreTrainedModel;w.BertForMaskedLM;w.BertForQuestionAnswering;w.BertForSequenceClassification;w.BertForTokenClassification;w.BertModel;w.BertPreTrainedModel;w.BertTokenizer;w.BitImageProcessor;w.BlenderbotForConditionalGeneration;w.BlenderbotModel;w.BlenderbotPreTrainedModel;w.BlenderbotSmallForConditionalGeneration;w.BlenderbotSmallModel;w.BlenderbotSmallPreTrainedModel;w.BlenderbotSmallTokenizer;w.BlenderbotTokenizer;w.BloomForCausalLM;w.BloomModel;w.BloomPreTrainedModel;w.BloomTokenizer;w.CLIPFeatureExtractor;w.CLIPImageProcessor;w.CLIPModel;w.CLIPPreTrainedModel;w.CLIPSegForImageSegmentation;w.CLIPSegModel;w.CLIPSegPreTrainedModel;w.CLIPTextModel;w.CLIPTextModelWithProjection;w.CLIPTokenizer;w.CLIPVisionModel;w.CLIPVisionModelWithProjection;w.CamembertForMaskedLM;w.CamembertForQuestionAnswering;w.CamembertForSequenceClassification;w.CamembertForTokenClassification;w.CamembertModel;w.CamembertPreTrainedModel;w.CamembertTokenizer;w.CausalLMOutput;w.CausalLMOutputWithPast;w.ChineseCLIPFeatureExtractor;w.ChineseCLIPModel;w.ChineseCLIPPreTrainedModel;w.ClapAudioModelWithProjection;w.ClapFeatureExtractor;w.ClapModel;w.ClapPreTrainedModel;w.ClapTextModelWithProjection;w.ClassifierFreeGuidanceLogitsProcessor;w.CodeGenForCausalLM;w.CodeGenModel;w.CodeGenPreTrainedModel;w.CodeGenTokenizer;w.CodeLlamaTokenizer;w.CohereForCausalLM;w.CohereModel;w.CoherePreTrainedModel;w.CohereTokenizer;w.ConvBertForMaskedLM;w.ConvBertForQuestionAnswering;w.ConvBertForSequenceClassification;w.ConvBertForTokenClassification;w.ConvBertModel;w.ConvBertPreTrainedModel;w.ConvBertTokenizer;w.ConvNextFeatureExtractor;w.ConvNextForImageClassification;w.ConvNextImageProcessor;w.ConvNextModel;w.ConvNextPreTrainedModel;w.ConvNextV2ForImageClassification;w.ConvNextV2Model;w.ConvNextV2PreTrainedModel;w.DFineForObjectDetection;w.DFineModel;w.DFinePreTrainedModel;w.DPTFeatureExtractor;w.DPTForDepthEstimation;w.DPTImageProcessor;w.DPTModel;w.DPTPreTrainedModel;w.DacDecoderModel;w.DacDecoderOutput;w.DacEncoderModel;w.DacEncoderOutput;w.DacFeatureExtractor;w.DacModel;w.DacPreTrainedModel;w.DataTypeMap;w.DebertaForMaskedLM;w.DebertaForQuestionAnswering;w.DebertaForSequenceClassification;w.DebertaForTokenClassification;w.DebertaModel;w.DebertaPreTrainedModel;w.DebertaTokenizer;w.DebertaV2ForMaskedLM;w.DebertaV2ForQuestionAnswering;w.DebertaV2ForSequenceClassification;w.DebertaV2ForTokenClassification;w.DebertaV2Model;w.DebertaV2PreTrainedModel;w.DebertaV2Tokenizer;w.DecisionTransformerModel;w.DecisionTransformerPreTrainedModel;w.DeiTFeatureExtractor;w.DeiTForImageClassification;w.DeiTImageProcessor;w.DeiTModel;w.DeiTPreTrainedModel;w.DepthAnythingForDepthEstimation;w.DepthAnythingPreTrainedModel;w.DepthEstimationPipeline;w.DepthProForDepthEstimation;w.DepthProPreTrainedModel;w.DetrFeatureExtractor;w.DetrForObjectDetection;w.DetrForSegmentation;w.DetrImageProcessor;w.DetrModel;w.DetrObjectDetectionOutput;w.DetrPreTrainedModel;w.DetrSegmentationOutput;w.Dinov2ForImageClassification;w.Dinov2Model;w.Dinov2PreTrainedModel;w.Dinov2WithRegistersForImageClassification;w.Dinov2WithRegistersModel;w.Dinov2WithRegistersPreTrainedModel;w.DistilBertForMaskedLM;w.DistilBertForQuestionAnswering;w.DistilBertForSequenceClassification;w.DistilBertForTokenClassification;w.DistilBertModel;w.DistilBertPreTrainedModel;w.DistilBertTokenizer;w.DocumentQuestionAnsweringPipeline;w.DonutFeatureExtractor;w.DonutImageProcessor;w.DonutSwinModel;w.DonutSwinPreTrainedModel;w.EfficientNetForImageClassification;w.EfficientNetImageProcessor;w.EfficientNetModel;w.EfficientNetPreTrainedModel;w.ElectraForMaskedLM;w.ElectraForQuestionAnswering;w.ElectraForSequenceClassification;w.ElectraForTokenClassification;w.ElectraModel;w.ElectraPreTrainedModel;w.ElectraTokenizer;w.EncodecFeatureExtractor;w.EosTokenCriteria;w.EsmForMaskedLM;w.EsmForSequenceClassification;w.EsmForTokenClassification;w.EsmModel;w.EsmPreTrainedModel;w.EsmTokenizer;w.ExaoneForCausalLM;w.ExaoneModel;w.ExaonePreTrainedModel;w.FFT;w.FalconForCausalLM;w.FalconModel;w.FalconPreTrainedModel;w.FalconTokenizer;w.FastViTForImageClassification;w.FastViTModel;w.FastViTPreTrainedModel;w.FeatureExtractionPipeline;w.FeatureExtractor;w.FillMaskPipeline;w.Florence2ForConditionalGeneration;w.Florence2PreTrainedModel;w.Florence2Processor;w.ForcedBOSTokenLogitsProcessor;w.ForcedEOSTokenLogitsProcessor;w.GLPNFeatureExtractor;w.GLPNForDepthEstimation;w.GLPNModel;w.GLPNPreTrainedModel;w.GPT2LMHeadModel;w.GPT2Model;w.GPT2PreTrainedModel;w.GPT2Tokenizer;w.GPTBigCodeForCausalLM;w.GPTBigCodeModel;w.GPTBigCodePreTrainedModel;w.GPTJForCausalLM;w.GPTJModel;w.GPTJPreTrainedModel;w.GPTNeoForCausalLM;w.GPTNeoModel;w.GPTNeoPreTrainedModel;w.GPTNeoXForCausalLM;w.GPTNeoXModel;w.GPTNeoXPreTrainedModel;w.GPTNeoXTokenizer;w.Gemma2ForCausalLM;w.Gemma2Model;w.Gemma2PreTrainedModel;w.Gemma3ForCausalLM;w.Gemma3Model;w.Gemma3PreTrainedModel;w.GemmaForCausalLM;w.GemmaModel;w.GemmaPreTrainedModel;w.GemmaTokenizer;w.GlmForCausalLM;w.GlmModel;w.GlmPreTrainedModel;w.GraniteForCausalLM;w.GraniteModel;w.GranitePreTrainedModel;w.Grok1Tokenizer;w.GroundingDinoForObjectDetection;w.GroundingDinoImageProcessor;w.GroundingDinoPreTrainedModel;w.GroundingDinoProcessor;w.GroupViTModel;w.GroupViTPreTrainedModel;w.HeliumForCausalLM;w.HeliumModel;w.HeliumPreTrainedModel;w.HerbertTokenizer;w.HieraForImageClassification;w.HieraModel;w.HieraPreTrainedModel;w.HubertForCTC;w.HubertForSequenceClassification;w.HubertModel;w.HubertPreTrainedModel;w.IJepaForImageClassification;w.IJepaModel;w.IJepaPreTrainedModel;w.Idefics3ForConditionalGeneration;w.Idefics3ImageProcessor;w.Idefics3PreTrainedModel;w.Idefics3Processor;w.ImageClassificationPipeline;w.ImageFeatureExtractionPipeline;w.ImageFeatureExtractor;w.ImageMattingOutput;w.ImageProcessor;w.ImageSegmentationPipeline;w.ImageToImagePipeline;w.ImageToTextPipeline;w.InterruptableStoppingCriteria;w.JAISLMHeadModel;w.JAISModel;w.JAISPreTrainedModel;w.JinaCLIPImageProcessor;w.JinaCLIPModel;w.JinaCLIPPreTrainedModel;w.JinaCLIPProcessor;w.JinaCLIPTextModel;w.JinaCLIPVisionModel;w.LiteWhisperForConditionalGeneration;w.LlamaForCausalLM;w.LlamaModel;w.LlamaPreTrainedModel;w.LlamaTokenizer;w.LlavaForConditionalGeneration;w.LlavaOnevisionForConditionalGeneration;w.LlavaOnevisionImageProcessor;w.LlavaPreTrainedModel;w.LogitsProcessor;w.LogitsProcessorList;w.LogitsWarper;w.LongT5ForConditionalGeneration;w.LongT5Model;w.LongT5PreTrainedModel;w.M2M100ForConditionalGeneration;w.M2M100Model;w.M2M100PreTrainedModel;w.M2M100Tokenizer;w.MBart50Tokenizer;w.MBartForCausalLM;w.MBartForConditionalGeneration;w.MBartForSequenceClassification;w.MBartModel;w.MBartPreTrainedModel;w.MBartTokenizer;w.MPNetForMaskedLM;w.MPNetForQuestionAnswering;w.MPNetForSequenceClassification;w.MPNetForTokenClassification;w.MPNetModel;w.MPNetPreTrainedModel;w.MPNetTokenizer;w.MT5ForConditionalGeneration;w.MT5Model;w.MT5PreTrainedModel;w.MarianMTModel;w.MarianModel;w.MarianPreTrainedModel;w.MarianTokenizer;w.Mask2FormerImageProcessor;w.MaskFormerFeatureExtractor;w.MaskFormerForInstanceSegmentation;w.MaskFormerImageProcessor;w.MaskFormerModel;w.MaskFormerPreTrainedModel;w.MaskedLMOutput;w.MaxLengthCriteria;w.Metric3DForDepthEstimation;w.Metric3DPreTrainedModel;w.Metric3Dv2ForDepthEstimation;w.Metric3Dv2PreTrainedModel;w.MgpstrForSceneTextRecognition;w.MgpstrModelOutput;w.MgpstrPreTrainedModel;w.MgpstrProcessor;w.MgpstrTokenizer;w.MimiDecoderModel;w.MimiDecoderOutput;w.MimiEncoderModel;w.MimiEncoderOutput;w.MimiModel;w.MimiPreTrainedModel;w.MinLengthLogitsProcessor;w.MinNewTokensLengthLogitsProcessor;w.MistralForCausalLM;w.MistralModel;w.MistralPreTrainedModel;w.MobileBertForMaskedLM;w.MobileBertForQuestionAnswering;w.MobileBertForSequenceClassification;w.MobileBertModel;w.MobileBertPreTrainedModel;w.MobileBertTokenizer;w.MobileLLMForCausalLM;w.MobileLLMModel;w.MobileLLMPreTrainedModel;w.MobileNetV1FeatureExtractor;w.MobileNetV1ForImageClassification;w.MobileNetV1ForSemanticSegmentation;w.MobileNetV1ImageProcessor;w.MobileNetV1Model;w.MobileNetV1PreTrainedModel;w.MobileNetV2FeatureExtractor;w.MobileNetV2ForImageClassification;w.MobileNetV2ForSemanticSegmentation;w.MobileNetV2ImageProcessor;w.MobileNetV2Model;w.MobileNetV2PreTrainedModel;w.MobileNetV3FeatureExtractor;w.MobileNetV3ForImageClassification;w.MobileNetV3ForSemanticSegmentation;w.MobileNetV3ImageProcessor;w.MobileNetV3Model;w.MobileNetV3PreTrainedModel;w.MobileNetV4FeatureExtractor;w.MobileNetV4ForImageClassification;w.MobileNetV4ForSemanticSegmentation;w.MobileNetV4ImageProcessor;w.MobileNetV4Model;w.MobileNetV4PreTrainedModel;w.MobileViTFeatureExtractor;w.MobileViTForImageClassification;w.MobileViTImageProcessor;w.MobileViTModel;w.MobileViTPreTrainedModel;w.MobileViTV2ForImageClassification;w.MobileViTV2Model;w.MobileViTV2PreTrainedModel;w.ModelOutput;w.ModernBertForMaskedLM;w.ModernBertForSequenceClassification;w.ModernBertForTokenClassification;w.ModernBertModel;w.ModernBertPreTrainedModel;w.Moondream1ForConditionalGeneration;w.MoonshineFeatureExtractor;w.MoonshineForConditionalGeneration;w.MoonshineModel;w.MoonshinePreTrainedModel;w.MoonshineProcessor;w.MptForCausalLM;w.MptModel;w.MptPreTrainedModel;w.MultiModalityCausalLM;w.MultiModalityPreTrainedModel;w.MusicgenForCausalLM;w.MusicgenForConditionalGeneration;w.MusicgenModel;w.MusicgenPreTrainedModel;w.NllbTokenizer;w.NoBadWordsLogitsProcessor;w.NoRepeatNGramLogitsProcessor;w.NomicBertModel;w.NomicBertPreTrainedModel;w.NougatImageProcessor;w.NougatTokenizer;w.OPTForCausalLM;w.OPTModel;w.OPTPreTrainedModel;w.ObjectDetectionPipeline;w.Olmo2ForCausalLM;w.Olmo2Model;w.Olmo2PreTrainedModel;w.OlmoForCausalLM;w.OlmoModel;w.OlmoPreTrainedModel;w.OpenELMForCausalLM;w.OpenELMModel;w.OpenELMPreTrainedModel;w.OwlViTFeatureExtractor;w.OwlViTForObjectDetection;w.OwlViTImageProcessor;w.OwlViTModel;w.OwlViTPreTrainedModel;w.OwlViTProcessor;w.Owlv2ForObjectDetection;w.Owlv2ImageProcessor;w.Owlv2Model;w.Owlv2PreTrainedModel;w.PaliGemmaForConditionalGeneration;w.PaliGemmaPreTrainedModel;w.PaliGemmaProcessor;w.PatchTSMixerForPrediction;w.PatchTSMixerModel;w.PatchTSMixerPreTrainedModel;w.PatchTSTForPrediction;w.PatchTSTModel;w.PatchTSTPreTrainedModel;w.Phi3ForCausalLM;w.Phi3Model;w.Phi3PreTrainedModel;w.Phi3VForCausalLM;w.Phi3VImageProcessor;w.Phi3VPreTrainedModel;w.Phi3VProcessor;w.PhiForCausalLM;w.PhiModel;w.PhiPreTrainedModel;w.Pipeline;w.PreTrainedModel;w.PreTrainedTokenizer;w.PretrainedConfig;w.PretrainedMixin;w.Processor;w.PvtForImageClassification;w.PvtImageProcessor;w.PvtModel;w.PvtPreTrainedModel;w.PyAnnoteFeatureExtractor;w.PyAnnoteForAudioFrameClassification;w.PyAnnoteModel;w.PyAnnotePreTrainedModel;w.PyAnnoteProcessor;w.QuestionAnsweringModelOutput;w.QuestionAnsweringPipeline;w.Qwen2ForCausalLM;w.Qwen2Model;w.Qwen2PreTrainedModel;w.Qwen2Tokenizer;w.Qwen2VLForConditionalGeneration;w.Qwen2VLImageProcessor;w.Qwen2VLPreTrainedModel;w.Qwen2VLProcessor;w.Qwen3ForCausalLM;w.Qwen3Model;w.Qwen3PreTrainedModel;w.RFDetrForObjectDetection;w.RFDetrModel;w.RFDetrObjectDetectionOutput;w.RFDetrPreTrainedModel;w.RTDetrForObjectDetection;w.RTDetrImageProcessor;w.RTDetrModel;w.RTDetrObjectDetectionOutput;w.RTDetrPreTrainedModel;w.RTDetrV2ForObjectDetection;w.RTDetrV2Model;w.RTDetrV2ObjectDetectionOutput;w.RTDetrV2PreTrainedModel;w.RawAudio;w.RawImage;w.RawVideo;w.RawVideoFrame;w.RepetitionPenaltyLogitsProcessor;w.ResNetForImageClassification;w.ResNetModel;w.ResNetPreTrainedModel;w.RoFormerForMaskedLM;w.RoFormerForQuestionAnswering;w.RoFormerForSequenceClassification;w.RoFormerForTokenClassification;w.RoFormerModel;w.RoFormerPreTrainedModel;w.RoFormerTokenizer;w.RobertaForMaskedLM;w.RobertaForQuestionAnswering;w.RobertaForSequenceClassification;w.RobertaForTokenClassification;w.RobertaModel;w.RobertaPreTrainedModel;w.RobertaTokenizer;w.SamImageProcessor;w.SamImageSegmentationOutput;w.SamModel;w.SamPreTrainedModel;w.SamProcessor;w.SapiensForDepthEstimation;w.SapiensForNormalEstimation;w.SapiensForSemanticSegmentation;w.SapiensPreTrainedModel;w.SeamlessM4TFeatureExtractor;w.SegformerFeatureExtractor;w.SegformerForImageClassification;w.SegformerForSemanticSegmentation;w.SegformerImageProcessor;w.SegformerModel;w.SegformerPreTrainedModel;w.Seq2SeqLMOutput;w.SequenceClassifierOutput;w.SiglipImageProcessor;w.SiglipModel;w.SiglipPreTrainedModel;w.SiglipTextModel;w.SiglipTokenizer;w.SiglipVisionModel;w.SmolVLMForConditionalGeneration;w.SmolVLMImageProcessor;w.SmolVLMProcessor;w.SnacDecoderModel;w.SnacEncoderModel;w.SnacFeatureExtractor;w.SnacModel;w.SnacPreTrainedModel;w.SpeechT5FeatureExtractor;w.SpeechT5ForSpeechToText;w.SpeechT5ForTextToSpeech;w.SpeechT5HifiGan;w.SpeechT5Model;w.SpeechT5PreTrainedModel;w.SpeechT5Processor;w.SpeechT5Tokenizer;w.SqueezeBertForMaskedLM;w.SqueezeBertForQuestionAnswering;w.SqueezeBertForSequenceClassification;w.SqueezeBertModel;w.SqueezeBertPreTrainedModel;w.SqueezeBertTokenizer;w.StableLmForCausalLM;w.StableLmModel;w.StableLmPreTrainedModel;w.Starcoder2ForCausalLM;w.Starcoder2Model;w.Starcoder2PreTrainedModel;w.StoppingCriteria;w.StoppingCriteriaList;w.StyleTextToSpeech2Model;w.StyleTextToSpeech2PreTrainedModel;w.SummarizationPipeline;w.SuppressTokensAtBeginLogitsProcessor;w.Swin2SRForImageSuperResolution;w.Swin2SRImageProcessor;w.Swin2SRModel;w.Swin2SRPreTrainedModel;w.SwinForImageClassification;w.SwinForSemanticSegmentation;w.SwinModel;w.SwinPreTrainedModel;w.T5ForConditionalGeneration;w.T5Model;w.T5PreTrainedModel;w.T5Tokenizer;w.TableTransformerForObjectDetection;w.TableTransformerModel;w.TableTransformerObjectDetectionOutput;w.TableTransformerPreTrainedModel;w.TemperatureLogitsWarper;const v_=w.Tensor;w.Text2TextGenerationPipeline;w.TextClassificationPipeline;w.TextGenerationPipeline;w.TextStreamer;w.TextToAudioPipeline;w.TokenClassificationPipeline;w.TokenClassifierOutput;w.TokenizerModel;w.TopKLogitsWarper;w.TopPLogitsWarper;w.TrOCRForCausalLM;w.TrOCRPreTrainedModel;w.TranslationPipeline;w.UltravoxModel;w.UltravoxPreTrainedModel;w.UltravoxProcessor;w.UniSpeechForCTC;w.UniSpeechForSequenceClassification;w.UniSpeechModel;w.UniSpeechPreTrainedModel;w.UniSpeechSatForAudioFrameClassification;w.UniSpeechSatForCTC;w.UniSpeechSatForSequenceClassification;w.UniSpeechSatModel;w.UniSpeechSatPreTrainedModel;w.VLChatProcessor;w.VLMImageProcessor;w.ViTFeatureExtractor;w.ViTForImageClassification;w.ViTImageProcessor;w.ViTMAEModel;w.ViTMAEPreTrainedModel;w.ViTMSNForImageClassification;w.ViTMSNModel;w.ViTMSNPreTrainedModel;w.ViTModel;w.ViTPreTrainedModel;w.VisionEncoderDecoderModel;w.VitMatteForImageMatting;w.VitMatteImageProcessor;w.VitMattePreTrainedModel;w.VitPoseForPoseEstimation;w.VitPoseImageProcessor;w.VitPosePreTrainedModel;w.VitsModel;w.VitsModelOutput;w.VitsPreTrainedModel;w.VitsTokenizer;w.Wav2Vec2BertForCTC;w.Wav2Vec2BertForSequenceClassification;w.Wav2Vec2BertModel;w.Wav2Vec2BertPreTrainedModel;w.Wav2Vec2CTCTokenizer;w.Wav2Vec2FeatureExtractor;w.Wav2Vec2ForAudioFrameClassification;w.Wav2Vec2ForCTC;w.Wav2Vec2ForSequenceClassification;w.Wav2Vec2Model;w.Wav2Vec2PreTrainedModel;w.Wav2Vec2Processor;w.Wav2Vec2ProcessorWithLM;w.WavLMForAudioFrameClassification;w.WavLMForCTC;w.WavLMForSequenceClassification;w.WavLMForXVector;w.WavLMModel;w.WavLMPreTrainedModel;w.WeSpeakerFeatureExtractor;w.WeSpeakerResNetModel;w.WeSpeakerResNetPreTrainedModel;w.WhisperFeatureExtractor;w.WhisperForConditionalGeneration;w.WhisperModel;w.WhisperPreTrainedModel;w.WhisperProcessor;w.WhisperTextStreamer;w.WhisperTimeStampLogitsProcessor;w.WhisperTokenizer;w.XLMForQuestionAnswering;w.XLMForSequenceClassification;w.XLMForTokenClassification;w.XLMModel;w.XLMPreTrainedModel;w.XLMRobertaForMaskedLM;w.XLMRobertaForQuestionAnswering;w.XLMRobertaForSequenceClassification;w.XLMRobertaForTokenClassification;w.XLMRobertaModel;w.XLMRobertaPreTrainedModel;w.XLMRobertaTokenizer;w.XLMTokenizer;w.XLMWithLMHeadModel;w.XVectorOutput;w.YolosFeatureExtractor;w.YolosForObjectDetection;w.YolosImageProcessor;w.YolosModel;w.YolosObjectDetectionOutput;w.YolosPreTrainedModel;w.ZeroShotAudioClassificationPipeline;w.ZeroShotClassificationPipeline;w.ZeroShotImageClassificationPipeline;w.ZeroShotObjectDetectionPipeline;w.bankers_round;w.cat;w.cos_sim;w.dot;w.dynamic_time_warping;const uf=w.env;w.full;w.full_like;w.getKeyValueShapes;w.hamming;w.hanning;w.interpolate;w.interpolate_4d;w.interpolate_data;w.is_chinese_char;w.layer_norm;w.load_image;w.load_video;w.log_softmax;w.magnitude;w.matmul;w.max;w.mean;w.mean_pooling;w.medianFilter;w.mel_filter_bank;w.min;w.ones;w.ones_like;w.permute;w.permute_data;const nE=w.pipeline;w.quantize_embeddings;w.rand;w.read_audio;w.rfft;w.round;w.slice;w.softmax;w.spectrogram;w.stack;w.std_mean;w.topk;w.window_function;w.zeros;w.zeros_like;async function aE(){try{if(!await navigator.gpu.requestAdapter())throw new Error("WebGPU is not supported (no adapter found)");return!0}catch(e){self.postMessage({status:"error",data:e.toString()})}}const oE="/models/",lE="onnx-community/whisper-base",cE="onnx-community/silero-vad",df=16e3,x_=df/1e3,uE=.3,dE=.1,pE=1500*x_,hE=80,qv=hE*x_,fE=500*x_,mE=30,_E=512,gE=Math.ceil(qv/_E);uf.localModelPath=oE;uf.allowRemoteModels=!0;uf.allowLocalModels=!0;uf.backends.onnx.wasm.proxy=!1;const yE=new v_("int64",[df],[]);let a0=new v_("float32",new Float32Array(2*1*128),[2,1,128]);const Va=new Float32Array(mE*df);let js=0,_d=[],Zh=0,lf=!1,Qv=!1;class Md{static async Init(r=null){if(this.silero_vad||(this.silero_vad=await sE.from_pretrained(cE,{config:{model_type:"custom"},dtype:"fp32"})),!this.transcriber){const t="webgpu",n={webgpu:{encoder_model:"fp32",decoder_model_merged:"fp32"},wasm:{encoder_model:"fp32",decoder_model_merged:"q8"}};this.transcriber=await nE("automatic-speech-recognition",lE,{device:t,dtype:n[t]}),await this.transcriber(new Float32Array(df))}}}le(Md,"silero_vad",null),le(Md,"transcriber",null);async function wE(){await Md.Init(e=>{e.file=e.name+"/"+e.file,self.postMessage(e)}),self.postMessage({status:"ready"}),Qv=!0}async function ME(e){const r=new v_("float32",e,[1,e.length]),{stateN:t,output:n}=await Md.silero_vad({input:r,sr:yE,state:a0});a0=t;const o=n.data[0];return o>uE||lf&&o>=dE}async function bE(e){const r=await Md.transcriber(e).then(({text:t})=>t.trim());["","[BLANK_AUDIO]"].includes(r)||(console.log("Transcribed:",r),self.postMessage({status:"from_stt",text:r}))}function Xv(e=0){Va.fill(0,e),js=e,lf=!1,Zh=0}function o0(e){const r=(e==null?void 0:e.length)??0,t=Va.slice(0,js+qv),n=_d.reduce((u,d)=>u+d.length,0),o=new Float32Array(n+t.length);let i=0;for(const u of _d)o.set(u,i),i+=u.length;o.set(t,i),bE(o),e&&Va.set(e,0),Xv(r)}async function vE(e){if(!Qv)return;const r=lf,t=await ME(e);if(t&&!r&&console.log("Speech detected, starting recording..."),!r&&!t){_d.length>=gE&&_d.shift(),_d.push(e);return}const n=Va.length-js;if(e.length>=n){Va.set(e.subarray(0,n),js),js+=n;const o=e.subarray(n);o0(o);return}else Va.set(e,js),js+=e.length;if(t){lf=!0,Zh=0;return}if(Zh+=e.length,!(Zh<pE)){if(js<fE){Xv();return}o0()}}self.addEventListener("message",async e=>{const{type:r,data:t}=e.data;switch(r){case"check":aE();break;case"load":wE();break;case"audio":vE(t);break}});