@huggingface/transformers 3.0.1 → 3.0.2

package/dist/transformers.min.js

@@ -151,10 +151,10 @@ var r,a,s,i,o,l,u,d,c,p,h,m,f,g,_,w,y,b,v,x,M,T,k,$,C,S,P,E,F,A,I,z,O,B,L=Object
   \******************************/(e,t,n)=>{var r;n.r(t),n.d(t,{Tensor:()=>o.Tensor,createInferenceSession:()=>g,deviceToExecutionProviders:()=>m,isONNXProxy:()=>y,isONNXTensor:()=>_});var a=n(/*! ../env.js */"./src/env.js"),s=n(/*! onnxruntime-node */"?2ce3"),i=n(/*! #onnxruntime-webgpu */"./node_modules/onnxruntime-web/dist/ort.webgpu.bundle.min.mjs"),o=n(/*! onnxruntime-common */"./node_modules/onnxruntime-common/dist/esm/index.js");const l=Object.freeze({auto:null,gpu:null,cpu:"cpu",wasm:"wasm",webgpu:"webgpu",cuda:"cuda",dml:"dml",webnn:{name:"webnn",deviceType:"cpu"},"webnn-npu":{name:"webnn",deviceType:"npu"},"webnn-gpu":{name:"webnn",deviceType:"gpu"},"webnn-cpu":{name:"webnn",deviceType:"cpu"}}),u=[];let d,c;const p=Symbol.for("onnxruntime");if(p in globalThis)c=globalThis[p];else if(a.apis.IS_NODE_ENV){switch(c=s??(r||(r=n.t(s,2))),process.platform){case"win32":u.push("dml");break;case"linux":"x64"===process.arch&&u.push("cuda")}u.push("cpu"),d=["cpu"]}else c=i,a.apis.IS_WEBNN_AVAILABLE&&u.push("webnn-npu","webnn-gpu","webnn-cpu","webnn"),a.apis.IS_WEBGPU_AVAILABLE&&u.push("webgpu"),u.push("wasm"),d=["wasm"];const h=c.InferenceSession;function m(e=null){if(!e)return d;switch(e){case"auto":return u;case"gpu":return u.filter((e=>["webgpu","cuda","dml","webnn-gpu"].includes(e)))}if(u.includes(e))return[l[e]??e];throw new Error(`Unsupported device: "${e}". Should be one of: ${u.join(", ")}.`)}let f=null;async function g(e,t,n){f&&await f;const r=h.create(e,t);f??=r;const a=await r;return a.config=n,a}function _(e){return e instanceof c.Tensor}const w=c?.env;function y(){return w?.wasm?.proxy}w?.wasm&&(w.wasm.wasmPaths=`https://cdn.jsdelivr.net/npm/@huggingface/transformers@${a.env.version}/dist/`,w.wasm.proxy=!1,"undefined"!=typeof crossOriginIsolated&&crossOriginIsolated||(w.wasm.numThreads=1)),w?.webgpu&&(w.webgpu.powerPreference="high-performance"),a.env.backends.onnx=w},"./src/configs.js":
 /*!************************!*\
   !*** ./src/configs.js ***!
-  \************************/(e,t,n)=>{n.r(t),n.d(t,{AutoConfig:()=>l,PretrainedConfig:()=>o,getKeyValueShapes:()=>i});var r=n(/*! ./utils/core.js */"./src/utils/core.js"),a=n(/*! ./utils/hub.js */"./src/utils/hub.js");function s(e){const t={};let n={};switch(e.model_type){case"llava":case"paligemma":case"florence2":n=s(e.text_config);break;case"moondream1":n=s(e.phi_config);break;case"musicgen":n=s(e.decoder);break;case"gpt2":case"gptj":case"jais":case"codegen":case"gpt_bigcode":t.num_heads="n_head",t.num_layers="n_layer",t.hidden_size="n_embd";break;case"gpt_neox":case"stablelm":case"opt":case"phi":case"phi3":case"falcon":t.num_heads="num_attention_heads",t.num_layers="num_hidden_layers",t.hidden_size="hidden_size";break;case"llama":case"granite":case"cohere":case"mistral":case"starcoder2":case"qwen2":t.num_heads="num_key_value_heads",t.num_layers="num_hidden_layers",t.hidden_size="hidden_size",t.num_attention_heads="num_attention_heads";break;case"gemma":case"gemma2":t.num_heads="num_key_value_heads",t.num_layers="num_hidden_layers",t.dim_kv="head_dim";break;case"openelm":t.num_heads="num_kv_heads",t.num_layers="num_transformer_layers",t.dim_kv="head_dim";break;case"gpt_neo":case"donut-swin":t.num_heads="num_heads",t.num_layers="num_layers",t.hidden_size="hidden_size";break;case"bloom":t.num_heads="n_head",t.num_layers="n_layer",t.hidden_size="hidden_size";break;case"mpt":t.num_heads="n_heads",t.num_layers="n_layers",t.hidden_size="d_model";break;case"t5":case"mt5":case"longt5":t.num_decoder_layers="num_decoder_layers",t.num_decoder_heads="num_heads",t.decoder_dim_kv="d_kv",t.num_encoder_layers="num_layers",t.num_encoder_heads="num_heads",t.encoder_dim_kv="d_kv";break;case"bart":case"mbart":case"marian":case"whisper":case"m2m_100":case"blenderbot":case"blenderbot-small":case"florence2_language":t.num_decoder_layers="decoder_layers",t.num_decoder_heads="decoder_attention_heads",t.decoder_hidden_size="d_model",t.num_encoder_layers="encoder_layers",t.num_encoder_heads="encoder_attention_heads",t.encoder_hidden_size="d_model";break;case"speecht5":t.num_decoder_layers="decoder_layers",t.num_decoder_heads="decoder_attention_heads",t.decoder_hidden_size="hidden_size",t.num_encoder_layers="encoder_layers",t.num_encoder_heads="encoder_attention_heads",t.encoder_hidden_size="hidden_size";break;case"trocr":t.num_encoder_layers=t.num_decoder_layers="decoder_layers",t.num_encoder_heads=t.num_decoder_heads="decoder_attention_heads",t.encoder_hidden_size=t.decoder_hidden_size="d_model";break;case"musicgen_decoder":t.num_encoder_layers=t.num_decoder_layers="num_hidden_layers",t.num_encoder_heads=t.num_decoder_heads="num_attention_heads",t.encoder_hidden_size=t.decoder_hidden_size="hidden_size";break;case"vision-encoder-decoder":const a=s(e.decoder),i="num_decoder_layers"in a,o=(0,r.pick)(e,["model_type","is_encoder_decoder"]);return i?(o.num_decoder_layers=a.num_decoder_layers,o.num_decoder_heads=a.num_decoder_heads,o.decoder_hidden_size=a.decoder_hidden_size,o.num_encoder_layers=a.num_encoder_layers,o.num_encoder_heads=a.num_encoder_heads,o.encoder_hidden_size=a.encoder_hidden_size):(o.num_layers=a.num_layers,o.num_heads=a.num_heads,o.hidden_size=a.hidden_size),o}const a={...n,...(0,r.pick)(e,["model_type","multi_query","is_encoder_decoder"])};for(const n in t)a[n]=e[t[n]];return a}function i(e,{prefix:t="past_key_values"}={}){const n={},r=e.normalized_config;if(r.is_encoder_decoder&&"num_encoder_heads"in r&&"num_decoder_heads"in r){const e=r.encoder_dim_kv??r.encoder_hidden_size/r.num_encoder_heads,a=r.decoder_dim_kv??r.decoder_hidden_size/r.num_decoder_heads,s=[1,r.num_encoder_heads,0,e],i=[1,r.num_decoder_heads,0,a];for(let e=0;e<r.num_decoder_layers;++e)n[`${t}.${e}.encoder.key`]=s,n[`${t}.${e}.encoder.value`]=s,n[`${t}.${e}.decoder.key`]=i,n[`${t}.${e}.decoder.value`]=i}else{const e=r.num_heads,a=r.num_layers,s=r.dim_kv??r.hidden_size/(r.num_attention_heads??e);if("falcon"===r.model_type){const r=[1*e,0,s];for(let e=0;e<a;++e)n[`${t}.${e}.key`]=r,n[`${t}.${e}.value`]=r}else if(r.multi_query){const r=[1*e,0,2*s];for(let e=0;e<a;++e)n[`${t}.${e}.key_value`]=r}else if("bloom"===r.model_type){const r=[1*e,s,0],i=[1*e,0,s];for(let e=0;e<a;++e)n[`${t}.${e}.key`]=r,n[`${t}.${e}.value`]=i}else if("openelm"===r.model_type)for(let r=0;r<a;++r){const a=[1,e[r],0,s];n[`${t}.${r}.key`]=a,n[`${t}.${r}.value`]=a}else{const r=[1,e,0,s];for(let e=0;e<a;++e)n[`${t}.${e}.key`]=r,n[`${t}.${e}.value`]=r}}return n}class o{model_type=null;is_encoder_decoder=!1;max_position_embeddings;"transformers.js_config";constructor(e){Object.assign(this,e),this.normalized_config=s(this)}static async from_pretrained(e,{progress_callback:t=null,config:n=null,cache_dir:r=null,local_files_only:s=!1,revision:i="main"}={}){!n||n instanceof o||(n=new o(n));const l=n??await async function(e,t){return await(0,a.getModelJSON)(e,"config.json",!0,t)}(e,{progress_callback:t,config:n,cache_dir:r,local_files_only:s,revision:i});return new this(l)}}class l{static async from_pretrained(...e){return o.from_pretrained(...e)}}},"./src/env.js":
+  \************************/(e,t,n)=>{n.r(t),n.d(t,{AutoConfig:()=>l,PretrainedConfig:()=>o,getKeyValueShapes:()=>i});var r=n(/*! ./utils/core.js */"./src/utils/core.js"),a=n(/*! ./utils/hub.js */"./src/utils/hub.js");function s(e){const t={};let n={};switch(e.model_type){case"llava":case"paligemma":case"florence2":n=s(e.text_config);break;case"moondream1":n=s(e.phi_config);break;case"musicgen":n=s(e.decoder);break;case"gpt2":case"gptj":case"jais":case"codegen":case"gpt_bigcode":t.num_heads="n_head",t.num_layers="n_layer",t.hidden_size="n_embd";break;case"gpt_neox":case"stablelm":case"opt":case"phi":case"phi3":case"falcon":t.num_heads="num_attention_heads",t.num_layers="num_hidden_layers",t.hidden_size="hidden_size";break;case"llama":case"olmo":case"mobilellm":case"granite":case"cohere":case"mistral":case"starcoder2":case"qwen2":t.num_heads="num_key_value_heads",t.num_layers="num_hidden_layers",t.hidden_size="hidden_size",t.num_attention_heads="num_attention_heads";break;case"gemma":case"gemma2":t.num_heads="num_key_value_heads",t.num_layers="num_hidden_layers",t.dim_kv="head_dim";break;case"openelm":t.num_heads="num_kv_heads",t.num_layers="num_transformer_layers",t.dim_kv="head_dim";break;case"gpt_neo":case"donut-swin":t.num_heads="num_heads",t.num_layers="num_layers",t.hidden_size="hidden_size";break;case"bloom":t.num_heads="n_head",t.num_layers="n_layer",t.hidden_size="hidden_size";break;case"mpt":t.num_heads="n_heads",t.num_layers="n_layers",t.hidden_size="d_model";break;case"t5":case"mt5":case"longt5":t.num_decoder_layers="num_decoder_layers",t.num_decoder_heads="num_heads",t.decoder_dim_kv="d_kv",t.num_encoder_layers="num_layers",t.num_encoder_heads="num_heads",t.encoder_dim_kv="d_kv";break;case"bart":case"mbart":case"marian":case"whisper":case"m2m_100":case"blenderbot":case"blenderbot-small":case"florence2_language":t.num_decoder_layers="decoder_layers",t.num_decoder_heads="decoder_attention_heads",t.decoder_hidden_size="d_model",t.num_encoder_layers="encoder_layers",t.num_encoder_heads="encoder_attention_heads",t.encoder_hidden_size="d_model";break;case"speecht5":t.num_decoder_layers="decoder_layers",t.num_decoder_heads="decoder_attention_heads",t.decoder_hidden_size="hidden_size",t.num_encoder_layers="encoder_layers",t.num_encoder_heads="encoder_attention_heads",t.encoder_hidden_size="hidden_size";break;case"trocr":t.num_encoder_layers=t.num_decoder_layers="decoder_layers",t.num_encoder_heads=t.num_decoder_heads="decoder_attention_heads",t.encoder_hidden_size=t.decoder_hidden_size="d_model";break;case"musicgen_decoder":t.num_encoder_layers=t.num_decoder_layers="num_hidden_layers",t.num_encoder_heads=t.num_decoder_heads="num_attention_heads",t.encoder_hidden_size=t.decoder_hidden_size="hidden_size";break;case"vision-encoder-decoder":const a=s(e.decoder),i="num_decoder_layers"in a,o=(0,r.pick)(e,["model_type","is_encoder_decoder"]);return i?(o.num_decoder_layers=a.num_decoder_layers,o.num_decoder_heads=a.num_decoder_heads,o.decoder_hidden_size=a.decoder_hidden_size,o.num_encoder_layers=a.num_encoder_layers,o.num_encoder_heads=a.num_encoder_heads,o.encoder_hidden_size=a.encoder_hidden_size):(o.num_layers=a.num_layers,o.num_heads=a.num_heads,o.hidden_size=a.hidden_size),o}const a={...n,...(0,r.pick)(e,["model_type","multi_query","is_encoder_decoder"])};for(const n in t)a[n]=e[t[n]];return a}function i(e,{prefix:t="past_key_values"}={}){const n={},r=e.normalized_config;if(r.is_encoder_decoder&&"num_encoder_heads"in r&&"num_decoder_heads"in r){const e=r.encoder_dim_kv??r.encoder_hidden_size/r.num_encoder_heads,a=r.decoder_dim_kv??r.decoder_hidden_size/r.num_decoder_heads,s=[1,r.num_encoder_heads,0,e],i=[1,r.num_decoder_heads,0,a];for(let e=0;e<r.num_decoder_layers;++e)n[`${t}.${e}.encoder.key`]=s,n[`${t}.${e}.encoder.value`]=s,n[`${t}.${e}.decoder.key`]=i,n[`${t}.${e}.decoder.value`]=i}else{const e=r.num_heads,a=r.num_layers,s=r.dim_kv??r.hidden_size/(r.num_attention_heads??e);if("falcon"===r.model_type){const r=[1*e,0,s];for(let e=0;e<a;++e)n[`${t}.${e}.key`]=r,n[`${t}.${e}.value`]=r}else if(r.multi_query){const r=[1*e,0,2*s];for(let e=0;e<a;++e)n[`${t}.${e}.key_value`]=r}else if("bloom"===r.model_type){const r=[1*e,s,0],i=[1*e,0,s];for(let e=0;e<a;++e)n[`${t}.${e}.key`]=r,n[`${t}.${e}.value`]=i}else if("openelm"===r.model_type)for(let r=0;r<a;++r){const a=[1,e[r],0,s];n[`${t}.${r}.key`]=a,n[`${t}.${r}.value`]=a}else{const r=[1,e,0,s];for(let e=0;e<a;++e)n[`${t}.${e}.key`]=r,n[`${t}.${e}.value`]=r}}return n}class o{model_type=null;is_encoder_decoder=!1;max_position_embeddings;"transformers.js_config";constructor(e){Object.assign(this,e),this.normalized_config=s(this)}static async from_pretrained(e,{progress_callback:t=null,config:n=null,cache_dir:r=null,local_files_only:s=!1,revision:i="main"}={}){!n||n instanceof o||(n=new o(n));const l=n??await async function(e,t){return await(0,a.getModelJSON)(e,"config.json",!0,t)}(e,{progress_callback:t,config:n,cache_dir:r,local_files_only:s,revision:i});return new this(l)}}class l{static async from_pretrained(...e){return o.from_pretrained(...e)}}},"./src/env.js":
 /*!********************!*\
   !*** ./src/env.js ***!
-  \********************/(e,t,n)=>{n.r(t),n.d(t,{apis:()=>f,env:()=>b});var r=n(/*! fs */"?569f"),a=n(/*! path */"?3f59"),s=n(/*! url */"?154a");const i="undefined"!=typeof self,o=i&&"DedicatedWorkerGlobalScope"===self.constructor.name,l=i&&"caches"in self,u="undefined"!=typeof navigator&&"gpu"in navigator,d="undefined"!=typeof navigator&&"ml"in navigator,c="undefined"!=typeof process,p=c&&"node"===process?.release?.name,h=!v(r),m=!v(a),f=Object.freeze({IS_BROWSER_ENV:i,IS_WEBWORKER_ENV:o,IS_WEB_CACHE_AVAILABLE:l,IS_WEBGPU_AVAILABLE:u,IS_WEBNN_AVAILABLE:d,IS_PROCESS_AVAILABLE:c,IS_NODE_ENV:p,IS_FS_AVAILABLE:h,IS_PATH_AVAILABLE:m}),g=h&&m,_=g?a.dirname(a.dirname(s.fileURLToPath(import.meta.url))):"./",w=g?a.join(_,"/.cache/"):null,y="/models/",b={version:"3.0.1",backends:{onnx:{}},allowRemoteModels:!0,remoteHost:"https://huggingface.co/",remotePathTemplate:"{model}/resolve/{revision}/",allowLocalModels:!i,localModelPath:g?a.join(_,y):y,useFS:h,useBrowserCache:l,useFSCache:h,cacheDir:w,useCustomCache:!1,customCache:null};function v(e){return 0===Object.keys(e).length}},"./src/generation/configuration_utils.js":
+  \********************/(e,t,n)=>{n.r(t),n.d(t,{apis:()=>f,env:()=>b});var r=n(/*! fs */"?569f"),a=n(/*! path */"?3f59"),s=n(/*! url */"?154a");const i="undefined"!=typeof self,o=i&&"DedicatedWorkerGlobalScope"===self.constructor.name,l=i&&"caches"in self,u="undefined"!=typeof navigator&&"gpu"in navigator,d="undefined"!=typeof navigator&&"ml"in navigator,c="undefined"!=typeof process,p=c&&"node"===process?.release?.name,h=!v(r),m=!v(a),f=Object.freeze({IS_BROWSER_ENV:i,IS_WEBWORKER_ENV:o,IS_WEB_CACHE_AVAILABLE:l,IS_WEBGPU_AVAILABLE:u,IS_WEBNN_AVAILABLE:d,IS_PROCESS_AVAILABLE:c,IS_NODE_ENV:p,IS_FS_AVAILABLE:h,IS_PATH_AVAILABLE:m}),g=h&&m;let _="./";if(g){const e=Object(import.meta).url;e?_=a.dirname(a.dirname(s.fileURLToPath(e))):"undefined"!=typeof __dirname&&(_=a.dirname(__dirname))}const w=g?a.join(_,"/.cache/"):null,y="/models/",b={version:"3.0.2",backends:{onnx:{}},allowRemoteModels:!0,remoteHost:"https://huggingface.co/",remotePathTemplate:"{model}/resolve/{revision}/",allowLocalModels:!i,localModelPath:g?a.join(_,y):y,useFS:h,useBrowserCache:l,useFSCache:h,cacheDir:w,useCustomCache:!1,customCache:null};function v(e){return 0===Object.keys(e).length}},"./src/generation/configuration_utils.js":
 /*!***********************************************!*\
   !*** ./src/generation/configuration_utils.js ***!
   \***********************************************/(e,t,n)=>{n.r(t),n.d(t,{GenerationConfig:()=>a});var r=n(/*! ../utils/core.js */"./src/utils/core.js");class a{max_length=20;max_new_tokens=null;min_length=0;min_new_tokens=null;early_stopping=!1;max_time=null;do_sample=!1;num_beams=1;num_beam_groups=1;penalty_alpha=null;use_cache=!0;temperature=1;top_k=50;top_p=1;typical_p=1;epsilon_cutoff=0;eta_cutoff=0;diversity_penalty=0;repetition_penalty=1;encoder_repetition_penalty=1;length_penalty=1;no_repeat_ngram_size=0;bad_words_ids=null;force_words_ids=null;renormalize_logits=!1;constraints=null;forced_bos_token_id=null;forced_eos_token_id=null;remove_invalid_values=!1;exponential_decay_length_penalty=null;suppress_tokens=null;begin_suppress_tokens=null;forced_decoder_ids=null;guidance_scale=null;num_return_sequences=1;output_attentions=!1;output_hidden_states=!1;output_scores=!1;return_dict_in_generate=!1;pad_token_id=null;bos_token_id=null;eos_token_id=null;encoder_no_repeat_ngram_size=0;decoder_start_token_id=null;generation_kwargs={};constructor(e){Object.assign(this,(0,r.pick)(e,Object.getOwnPropertyNames(this)))}}},"./src/generation/logits_process.js":
@@ -172,7 +172,7 @@ var r,a,s,i,o,l,u,d,c,p,h,m,f,g,_,w,y,b,v,x,M,T,k,$,C,S,P,E,F,A,I,z,O,B,L=Object
   \*************************************/(e,t,n)=>{n.r(t),n.d(t,{BaseStreamer:()=>i,TextStreamer:()=>l,WhisperTextStreamer:()=>u});var r=n(/*! ../utils/core.js */"./src/utils/core.js"),a=n(/*! ../tokenizers.js */"./src/tokenizers.js"),s=n(/*! ../env.js */"./src/env.js");class i{put(e){throw Error("Not implemented")}end(){throw Error("Not implemented")}}const o=s.apis.IS_PROCESS_AVAILABLE?e=>process.stdout.write(e):e=>console.log(e);class l extends i{constructor(e,{skip_prompt:t=!1,callback_function:n=null,token_callback_function:r=null,decode_kwargs:a={},...s}={}){super(),this.tokenizer=e,this.skip_prompt=t,this.callback_function=n??o,this.token_callback_function=r,this.decode_kwargs={...a,...s},this.token_cache=[],this.print_len=0,this.next_tokens_are_prompt=!0}put(e){if(e.length>1)throw Error("TextStreamer only supports batch size of 1");if(this.skip_prompt&&this.next_tokens_are_prompt)return void(this.next_tokens_are_prompt=!1);const t=e[0];this.token_callback_function?.(t),this.token_cache=(0,r.mergeArrays)(this.token_cache,t);const n=this.tokenizer.decode(this.token_cache,this.decode_kwargs);let s;n.endsWith("\n")?(s=n.slice(this.print_len),this.token_cache=[],this.print_len=0):n.length>0&&(0,a.is_chinese_char)(n.charCodeAt(n.length-1))?(s=n.slice(this.print_len),this.print_len+=s.length):(s=n.slice(this.print_len,n.lastIndexOf(" ")+1),this.print_len+=s.length),this.on_finalized_text(s,!1)}end(){let e;if(this.token_cache.length>0){e=this.tokenizer.decode(this.token_cache,this.decode_kwargs).slice(this.print_len),this.token_cache=[],this.print_len=0}else e="";this.next_tokens_are_prompt=!0,this.on_finalized_text(e,!0)}on_finalized_text(e,t){e.length>0&&this.callback_function?.(e),t&&this.callback_function===o&&s.apis.IS_PROCESS_AVAILABLE&&this.callback_function?.("\n")}}class u extends l{constructor(e,{skip_prompt:t=!1,callback_function:n=null,token_callback_function:r=null,on_chunk_start:a=null,on_chunk_end:s=null,on_finalize:i=null,time_precision:o=.02,skip_special_tokens:l=!0,decode_kwargs:u={}}={}){super(e,{skip_prompt:t,callback_function:n,token_callback_function:r,decode_kwargs:{skip_special_tokens:l,...u}}),this.timestamp_begin=e.timestamp_begin,this.on_chunk_start=a,this.on_chunk_end=s,this.on_finalize=i,this.time_precision=o,this.waiting_for_timestamp=!1}put(e){if(e.length>1)throw Error("WhisperTextStreamer only supports batch size of 1");const t=e[0];if(1===t.length){const n=Number(t[0])-this.timestamp_begin;if(n>=0){const t=n*this.time_precision;this.waiting_for_timestamp?this.on_chunk_end?.(t):this.on_chunk_start?.(t),this.waiting_for_timestamp=!this.waiting_for_timestamp,e=[[]]}}return super.put(e)}end(){super.end(),this.on_finalize?.()}}},"./src/models.js":
 /*!***********************!*\
   !*** ./src/models.js ***!
-  \***********************/(e,t,n)=>{n.r(t),n.d(t,{ASTForAudioClassification:()=>an,ASTModel:()=>rn,ASTPreTrainedModel:()=>nn,AlbertForMaskedLM:()=>ht,AlbertForQuestionAnswering:()=>pt,AlbertForSequenceClassification:()=>ct,AlbertModel:()=>dt,AlbertPreTrainedModel:()=>ut,AutoModel:()=>Do,AutoModelForAudioClassification:()=>rl,AutoModelForAudioFrameClassification:()=>sl,AutoModelForCTC:()=>nl,AutoModelForCausalLM:()=>Uo,AutoModelForDepthEstimation:()=>ul,AutoModelForDocumentQuestionAnswering:()=>il,AutoModelForImageClassification:()=>Ko,AutoModelForImageFeatureExtraction:()=>cl,AutoModelForImageMatting:()=>ol,AutoModelForImageSegmentation:()=>Qo,AutoModelForImageToImage:()=>ll,AutoModelForMaskGeneration:()=>tl,AutoModelForMaskedLM:()=>Wo,AutoModelForNormalEstimation:()=>dl,AutoModelForObjectDetection:()=>Jo,AutoModelForQuestionAnswering:()=>Ho,AutoModelForSemanticSegmentation:()=>Yo,AutoModelForSeq2SeqLM:()=>Vo,AutoModelForSequenceClassification:()=>Ro,AutoModelForSpeechSeq2Seq:()=>jo,AutoModelForTextToSpectrogram:()=>qo,AutoModelForTextToWaveform:()=>Go,AutoModelForTokenClassification:()=>No,AutoModelForUniversalSegmentation:()=>Zo,AutoModelForVision2Seq:()=>Xo,AutoModelForXVector:()=>al,AutoModelForZeroShotObjectDetection:()=>el,BartForConditionalGeneration:()=>kt,BartForSequenceClassification:()=>$t,BartModel:()=>Tt,BartPretrainedModel:()=>Mt,BaseModelOutput:()=>U,BeitForImageClassification:()=>ca,BeitModel:()=>da,BeitPreTrainedModel:()=>ua,BertForMaskedLM:()=>X,BertForQuestionAnswering:()=>Y,BertForSequenceClassification:()=>K,BertForTokenClassification:()=>Q,BertModel:()=>H,BertPreTrainedModel:()=>W,BlenderbotForConditionalGeneration:()=>zt,BlenderbotModel:()=>It,BlenderbotPreTrainedModel:()=>At,BlenderbotSmallForConditionalGeneration:()=>Lt,BlenderbotSmallModel:()=>Bt,BlenderbotSmallPreTrainedModel:()=>Ot,BloomForCausalLM:()=>$r,BloomModel:()=>kr,BloomPreTrainedModel:()=>Tr,CLIPModel:()=>gn,CLIPPreTrainedModel:()=>fn,CLIPSegForImageSegmentation:()=>Pn,CLIPSegModel:()=>Sn,CLIPSegPreTrainedModel:()=>Cn,CLIPTextModel:()=>_n,CLIPTextModelWithProjection:()=>wn,CLIPVisionModel:()=>yn,CLIPVisionModelWithProjection:()=>bn,CamembertForMaskedLM:()=>be,CamembertForQuestionAnswering:()=>Me,CamembertForSequenceClassification:()=>ve,CamembertForTokenClassification:()=>xe,CamembertModel:()=>ye,CamembertPreTrainedModel:()=>we,CausalLMOutput:()=>wl,CausalLMOutputWithPast:()=>yl,ChineseCLIPModel:()=>$n,ChineseCLIPPreTrainedModel:()=>kn,ClapAudioModelWithProjection:()=>$i,ClapModel:()=>Ti,ClapPreTrainedModel:()=>Mi,ClapTextModelWithProjection:()=>ki,CodeGenForCausalLM:()=>Qn,CodeGenModel:()=>Kn,CodeGenPreTrainedModel:()=>Xn,CohereForCausalLM:()=>sr,CohereModel:()=>ar,CoherePreTrainedModel:()=>rr,ConvBertForMaskedLM:()=>le,ConvBertForQuestionAnswering:()=>ce,ConvBertForSequenceClassification:()=>ue,ConvBertForTokenClassification:()=>de,ConvBertModel:()=>oe,ConvBertPreTrainedModel:()=>ie,ConvNextForImageClassification:()=>ls,ConvNextModel:()=>os,ConvNextPreTrainedModel:()=>is,ConvNextV2ForImageClassification:()=>cs,ConvNextV2Model:()=>ds,ConvNextV2PreTrainedModel:()=>us,DPTForDepthEstimation:()=>qa,DPTModel:()=>ja,DPTPreTrainedModel:()=>Va,DebertaForMaskedLM:()=>$e,DebertaForQuestionAnswering:()=>Pe,DebertaForSequenceClassification:()=>Ce,DebertaForTokenClassification:()=>Se,DebertaModel:()=>ke,DebertaPreTrainedModel:()=>Te,DebertaV2ForMaskedLM:()=>Ae,DebertaV2ForQuestionAnswering:()=>Oe,DebertaV2ForSequenceClassification:()=>Ie,DebertaV2ForTokenClassification:()=>ze,DebertaV2Model:()=>Fe,DebertaV2PreTrainedModel:()=>Ee,DecisionTransformerModel:()=>no,DecisionTransformerPreTrainedModel:()=>to,DeiTForImageClassification:()=>Sa,DeiTModel:()=>Ca,DeiTPreTrainedModel:()=>$a,DepthAnythingForDepthEstimation:()=>Ua,DepthAnythingPreTrainedModel:()=>Ga,DepthProForDepthEstimation:()=>Ya,DepthProPreTrainedModel:()=>Qa,DetrForObjectDetection:()=>ma,DetrForSegmentation:()=>fa,DetrModel:()=>ha,DetrObjectDetectionOutput:()=>ga,DetrPreTrainedModel:()=>pa,DetrSegmentationOutput:()=>_a,Dinov2ForImageClassification:()=>ms,Dinov2Model:()=>hs,Dinov2PreTrainedModel:()=>ps,DistilBertForMaskedLM:()=>Ve,DistilBertForQuestionAnswering:()=>Ne,DistilBertForSequenceClassification:()=>De,DistilBertForTokenClassification:()=>Re,DistilBertModel:()=>Le,DistilBertPreTrainedModel:()=>Be,DonutSwinModel:()=>ss,DonutSwinPreTrainedModel:()=>as,EfficientNetForImageClassification:()=>Di,EfficientNetModel:()=>Li,EfficientNetPreTrainedModel:()=>Bi,ElectraForMaskedLM:()=>me,ElectraForQuestionAnswering:()=>_e,ElectraForSequenceClassification:()=>fe,ElectraForTokenClassification:()=>ge,ElectraModel:()=>he,ElectraPreTrainedModel:()=>pe,EsmForMaskedLM:()=>Ge,EsmForSequenceClassification:()=>Ue,EsmForTokenClassification:()=>We,EsmModel:()=>qe,EsmPreTrainedModel:()=>je,FalconForCausalLM:()=>xi,FalconModel:()=>vi,FalconPreTrainedModel:()=>bi,FastViTForImageClassification:()=>Xr,FastViTModel:()=>Hr,FastViTPreTrainedModel:()=>Wr,Florence2ForConditionalGeneration:()=>mn,Florence2PreTrainedModel:()=>hn,GLPNForDepthEstimation:()=>rs,GLPNModel:()=>ns,GLPNPreTrainedModel:()=>ts,GPT2LMHeadModel:()=>An,GPT2Model:()=>Fn,GPT2PreTrainedModel:()=>En,GPTBigCodeForCausalLM:()=>Hn,GPTBigCodeModel:()=>Wn,GPTBigCodePreTrainedModel:()=>Un,GPTJForCausalLM:()=>Gn,GPTJModel:()=>qn,GPTJPreTrainedModel:()=>jn,GPTNeoForCausalLM:()=>Dn,GPTNeoModel:()=>Ln,GPTNeoPreTrainedModel:()=>Bn,GPTNeoXForCausalLM:()=>Vn,GPTNeoXModel:()=>Nn,GPTNeoXPreTrainedModel:()=>Rn,Gemma2ForCausalLM:()=>cr,Gemma2Model:()=>dr,Gemma2PreTrainedModel:()=>ur,GemmaForCausalLM:()=>lr,GemmaModel:()=>or,GemmaPreTrainedModel:()=>ir,GraniteForCausalLM:()=>nr,GraniteModel:()=>tr,GranitePreTrainedModel:()=>er,GroupViTModel:()=>Ur,GroupViTPreTrainedModel:()=>Gr,HieraForImageClassification:()=>Fa,HieraModel:()=>Ea,HieraPreTrainedModel:()=>Pa,HubertForCTC:()=>Js,HubertForSequenceClassification:()=>ei,HubertModel:()=>Zs,HubertPreTrainedModel:()=>Ys,ImageMattingOutput:()=>bl,JAISLMHeadModel:()=>On,JAISModel:()=>zn,JAISPreTrainedModel:()=>In,LlamaForCausalLM:()=>Jn,LlamaModel:()=>Zn,LlamaPreTrainedModel:()=>Yn,LlavaForConditionalGeneration:()=>cn,LlavaPreTrainedModel:()=>dn,LongT5ForConditionalGeneration:()=>yt,LongT5Model:()=>wt,LongT5PreTrainedModel:()=>_t,M2M100ForConditionalGeneration:()=>Cs,M2M100Model:()=>$s,M2M100PreTrainedModel:()=>ks,MBartForCausalLM:()=>Ft,MBartForConditionalGeneration:()=>Pt,MBartForSequenceClassification:()=>Et,MBartModel:()=>St,MBartPreTrainedModel:()=>Ct,MPNetForMaskedLM:()=>et,MPNetForQuestionAnswering:()=>rt,MPNetForSequenceClassification:()=>tt,MPNetForTokenClassification:()=>nt,MPNetModel:()=>Je,MPNetPreTrainedModel:()=>Ze,MT5ForConditionalGeneration:()=>xt,MT5Model:()=>vt,MT5PreTrainedModel:()=>bt,MarianMTModel:()=>Ts,MarianModel:()=>Ms,MarianPreTrainedModel:()=>xs,MaskFormerForInstanceSegmentation:()=>es,MaskFormerModel:()=>Ja,MaskFormerPreTrainedModel:()=>Za,MaskedLMOutput:()=>gl,MistralForCausalLM:()=>gi,MistralModel:()=>fi,MistralPreTrainedModel:()=>mi,MobileBertForMaskedLM:()=>Ke,MobileBertForQuestionAnswering:()=>Ye,MobileBertForSequenceClassification:()=>Qe,MobileBertModel:()=>Xe,MobileBertPreTrainedModel:()=>He,MobileNetV1ForImageClassification:()=>Ui,MobileNetV1Model:()=>Gi,MobileNetV1PreTrainedModel:()=>qi,MobileNetV2ForImageClassification:()=>Xi,MobileNetV2Model:()=>Hi,MobileNetV2PreTrainedModel:()=>Wi,MobileNetV3ForImageClassification:()=>Yi,MobileNetV3Model:()=>Qi,MobileNetV3PreTrainedModel:()=>Ki,MobileNetV4ForImageClassification:()=>eo,MobileNetV4Model:()=>Ji,MobileNetV4PreTrainedModel:()=>Zi,MobileViTForImageClassification:()=>Jr,MobileViTModel:()=>Zr,MobileViTPreTrainedModel:()=>Yr,MobileViTV2ForImageClassification:()=>na,MobileViTV2Model:()=>ta,MobileViTV2PreTrainedModel:()=>ea,ModelOutput:()=>G,Moondream1ForConditionalGeneration:()=>pn,MptForCausalLM:()=>Pr,MptModel:()=>Sr,MptPreTrainedModel:()=>Cr,MusicgenForCausalLM:()=>Vi,MusicgenForConditionalGeneration:()=>ji,MusicgenModel:()=>Ni,MusicgenPreTrainedModel:()=>Ri,NomicBertModel:()=>J,NomicBertPreTrainedModel:()=>Z,OPTForCausalLM:()=>Ar,OPTModel:()=>Fr,OPTPreTrainedModel:()=>Er,OpenELMForCausalLM:()=>mr,OpenELMModel:()=>hr,OpenELMPreTrainedModel:()=>pr,OwlViTForObjectDetection:()=>sa,OwlViTModel:()=>aa,OwlViTPreTrainedModel:()=>ra,Owlv2ForObjectDetection:()=>la,Owlv2Model:()=>oa,Owlv2PreTrainedModel:()=>ia,Phi3ForCausalLM:()=>Mr,Phi3Model:()=>xr,Phi3PreTrainedModel:()=>vr,PhiForCausalLM:()=>br,PhiModel:()=>yr,PhiPreTrainedModel:()=>wr,PreTrainedModel:()=>q,PretrainedMixin:()=>ro,PvtForImageClassification:()=>Dr,PvtModel:()=>Lr,PvtPreTrainedModel:()=>Br,PyAnnoteForAudioFrameClassification:()=>Os,PyAnnoteModel:()=>zs,PyAnnotePreTrainedModel:()=>Is,QuestionAnsweringModelOutput:()=>_l,Qwen2ForCausalLM:()=>_r,Qwen2Model:()=>gr,Qwen2PreTrainedModel:()=>fr,RTDetrForObjectDetection:()=>ba,RTDetrModel:()=>ya,RTDetrObjectDetectionOutput:()=>va,RTDetrPreTrainedModel:()=>wa,ResNetForImageClassification:()=>za,ResNetModel:()=>Ia,ResNetPreTrainedModel:()=>Aa,RoFormerForMaskedLM:()=>ne,RoFormerForQuestionAnswering:()=>se,RoFormerForSequenceClassification:()=>re,RoFormerForTokenClassification:()=>ae,RoFormerModel:()=>te,RoFormerPreTrainedModel:()=>ee,RobertaForMaskedLM:()=>Nt,RobertaForQuestionAnswering:()=>qt,RobertaForSequenceClassification:()=>Vt,RobertaForTokenClassification:()=>jt,RobertaModel:()=>Rt,RobertaPreTrainedModel:()=>Dt,SamImageSegmentationOutput:()=>vs,SamModel:()=>bs,SamPreTrainedModel:()=>ys,SapiensForDepthEstimation:()=>Xa,SapiensForNormalEstimation:()=>Ka,SapiensForSemanticSegmentation:()=>Ha,SapiensPreTrainedModel:()=>Wa,SegformerForImageClassification:()=>Fi,SegformerForSemanticSegmentation:()=>Ai,SegformerModel:()=>Ei,SegformerPreTrainedModel:()=>Pi,Seq2SeqLMOutput:()=>pl,SequenceClassifierOutput:()=>hl,SiglipModel:()=>xn,SiglipPreTrainedModel:()=>vn,SiglipTextModel:()=>Mn,SiglipVisionModel:()=>Tn,SpeechT5ForSpeechToText:()=>ui,SpeechT5ForTextToSpeech:()=>di,SpeechT5HifiGan:()=>ci,SpeechT5Model:()=>li,SpeechT5PreTrainedModel:()=>oi,SqueezeBertForMaskedLM:()=>it,SqueezeBertForQuestionAnswering:()=>lt,SqueezeBertForSequenceClassification:()=>ot,SqueezeBertModel:()=>st,SqueezeBertPreTrainedModel:()=>at,StableLmForCausalLM:()=>Oi,StableLmModel:()=>zi,StableLmPreTrainedModel:()=>Ii,Starcoder2ForCausalLM:()=>yi,Starcoder2Model:()=>wi,Starcoder2PreTrainedModel:()=>_i,Swin2SRForImageSuperResolution:()=>Na,Swin2SRModel:()=>Ra,Swin2SRPreTrainedModel:()=>Da,SwinForImageClassification:()=>La,SwinModel:()=>Ba,SwinPreTrainedModel:()=>Oa,T5ForConditionalGeneration:()=>gt,T5Model:()=>ft,T5PreTrainedModel:()=>mt,TableTransformerForObjectDetection:()=>Ta,TableTransformerModel:()=>Ma,TableTransformerObjectDetectionOutput:()=>ka,TableTransformerPreTrainedModel:()=>xa,TokenClassifierOutput:()=>fl,TrOCRForCausalLM:()=>hi,TrOCRPreTrainedModel:()=>pi,UniSpeechForCTC:()=>Ns,UniSpeechForSequenceClassification:()=>Vs,UniSpeechModel:()=>Rs,UniSpeechPreTrainedModel:()=>Ds,UniSpeechSatForAudioFrameClassification:()=>Ws,UniSpeechSatForCTC:()=>Gs,UniSpeechSatForSequenceClassification:()=>Us,UniSpeechSatModel:()=>qs,UniSpeechSatPreTrainedModel:()=>js,ViTForImageClassification:()=>Or,ViTMAEModel:()=>Nr,ViTMAEPreTrainedModel:()=>Rr,ViTMSNForImageClassification:()=>qr,ViTMSNModel:()=>jr,ViTMSNPreTrainedModel:()=>Vr,ViTModel:()=>zr,ViTPreTrainedModel:()=>Ir,VisionEncoderDecoderModel:()=>un,VitMatteForImageMatting:()=>Qr,VitMattePreTrainedModel:()=>Kr,VitsModel:()=>Si,VitsModelOutput:()=>vl,VitsPreTrainedModel:()=>Ci,Wav2Vec2BertForCTC:()=>Ks,Wav2Vec2BertForSequenceClassification:()=>Qs,Wav2Vec2BertModel:()=>Xs,Wav2Vec2BertPreTrainedModel:()=>Hs,Wav2Vec2ForAudioFrameClassification:()=>As,Wav2Vec2ForCTC:()=>Es,Wav2Vec2ForSequenceClassification:()=>Fs,Wav2Vec2Model:()=>Ps,Wav2Vec2PreTrainedModel:()=>Ss,WavLMForAudioFrameClassification:()=>ii,WavLMForCTC:()=>ri,WavLMForSequenceClassification:()=>ai,WavLMForXVector:()=>si,WavLMModel:()=>ni,WavLMPreTrainedModel:()=>ti,WeSpeakerResNetModel:()=>Ls,WeSpeakerResNetPreTrainedModel:()=>Bs,WhisperForConditionalGeneration:()=>ln,WhisperModel:()=>on,WhisperPreTrainedModel:()=>sn,XLMForQuestionAnswering:()=>Kt,XLMForSequenceClassification:()=>Ht,XLMForTokenClassification:()=>Xt,XLMModel:()=>Ut,XLMPreTrainedModel:()=>Gt,XLMRobertaForMaskedLM:()=>Zt,XLMRobertaForQuestionAnswering:()=>tn,XLMRobertaForSequenceClassification:()=>Jt,XLMRobertaForTokenClassification:()=>en,XLMRobertaModel:()=>Yt,XLMRobertaPreTrainedModel:()=>Qt,XLMWithLMHeadModel:()=>Wt,XVectorOutput:()=>ml,YolosForObjectDetection:()=>_s,YolosModel:()=>gs,YolosObjectDetectionOutput:()=>ws,YolosPreTrainedModel:()=>fs});var r=n(/*! ./configs.js */"./src/configs.js"),a=n(/*! ./backends/onnx.js */"./src/backends/onnx.js"),s=n(/*! ./utils/dtypes.js */"./src/utils/dtypes.js"),i=n(/*! ./utils/generic.js */"./src/utils/generic.js"),o=n(/*! ./utils/core.js */"./src/utils/core.js"),l=n(/*! ./utils/hub.js */"./src/utils/hub.js"),u=n(/*! ./utils/constants.js */"./src/utils/constants.js"),d=n(/*! ./generation/logits_process.js */"./src/generation/logits_process.js"),c=n(/*! ./generation/configuration_utils.js */"./src/generation/configuration_utils.js"),p=n(/*! ./utils/tensor.js */"./src/utils/tensor.js"),h=n(/*! ./utils/maths.js */"./src/utils/maths.js"),m=n(/*! ./generation/stopping_criteria.js */"./src/generation/stopping_criteria.js"),f=n(/*! ./generation/logits_sampler.js */"./src/generation/logits_sampler.js"),g=n(/*! ./env.js */"./src/env.js"),_=n(/*! ./models/whisper/generation_whisper.js */"./src/models/whisper/generation_whisper.js"),w=n(/*! ./models/whisper/common_whisper.js */"./src/models/whisper/common_whisper.js");const y=0,b=1,v=2,x=3,M=4,T=5,k=6,$=7,C=new Map,S=new Map,P=new Map;async function E(e,t,n){return Object.fromEntries(await Promise.all(Object.keys(t).map((async i=>{const{buffer:o,session_options:u,session_config:d}=await async function(e,t,n){const i=n.config?.["transformers.js_config"]??{};let o=n.device??i.device;o&&"string"!=typeof o&&(o.hasOwnProperty(t)?o=o[t]:(console.warn(`device not specified for "${t}". Using the default device.`),o=null));const u=o??(g.apis.IS_NODE_ENV?"cpu":"wasm"),d=(0,a.deviceToExecutionProviders)(u);let c=n.dtype??i.dtype;"string"!=typeof c&&(c&&c.hasOwnProperty(t)?c=c[t]:(c=s.DEFAULT_DEVICE_DTYPE_MAPPING[u]??s.DATA_TYPES.fp32,console.warn(`dtype not specified for "${t}". Using the default dtype (${c}) for this device (${u}).`)));const p=c;if(!s.DEFAULT_DTYPE_SUFFIX_MAPPING.hasOwnProperty(p))throw new Error(`Invalid dtype: ${p}. Should be one of: ${Object.keys(s.DATA_TYPES).join(", ")}`);if(p===s.DATA_TYPES.fp16&&"webgpu"===u&&!await(0,s.isWebGpuFp16Supported)())throw new Error(`The device (${u}) does not support fp16.`);const h=i.kv_cache_dtype?"string"==typeof i.kv_cache_dtype?i.kv_cache_dtype:i.kv_cache_dtype[p]??"float32":void 0;if(h&&!["float32","float16"].includes(h))throw new Error(`Invalid kv_cache_dtype: ${h}. Should be one of: float32, float16`);const m={dtype:p,kv_cache_dtype:h},f=s.DEFAULT_DTYPE_SUFFIX_MAPPING[p],_=`${n.subfolder??""}/${t}${f}.onnx`,w={...n.session_options};w.executionProviders??=d;const y=i.free_dimension_overrides;y?w.freeDimensionOverrides??=y:u.startsWith("webnn")&&!w.freeDimensionOverrides&&console.warn('WebNN does not currently support dynamic shapes and requires `free_dimension_overrides` to be set in config.json as a field within "transformers.js_config". When `free_dimension_overrides` is not set, you may experience significant performance degradation.');const b=(0,l.getModelFile)(e,_,!0,n),v=n.use_external_data_format??i.use_external_data_format;let x=[];if(v&&(!0===v||"object"==typeof v&&v.hasOwnProperty(t)&&!0===v[t])){if(g.apis.IS_NODE_ENV)throw new Error("External data format is not yet supported in Node.js");const r=`${t}${f}.onnx_data`,a=`${n.subfolder??""}/${r}`;x.push(new Promise((async(t,s)=>{const i=await(0,l.getModelFile)(e,a,!0,n);t({path:r,data:i})})))}else void 0!==w.externalData&&(x=w.externalData.map((async t=>{if("string"==typeof t.data){const r=await(0,l.getModelFile)(e,t.data,!0,n);return{...t,data:r}}return t})));if(x.length>0&&(w.externalData=await Promise.all(x)),"webgpu"===u){const e=(0,r.getKeyValueShapes)(n.config,{prefix:"present"});if(Object.keys(e).length>0&&!(0,a.isONNXProxy)()){const t={};for(const n in e)t[n]="gpu-buffer";w.preferredOutputLocation=t}}return{buffer:await b,session_options:w,session_config:m}}(e,t[i],n);return[i,await(0,a.createInferenceSession)(o,u,d)]}))))}async function F(e,t,n){return Object.fromEntries(await Promise.all(Object.keys(t).map((async r=>[r,await(0,l.getModelJSON)(e,t[r],!1,n)]))))}async function A(e,t){const n=function(e,t){const n=Object.create(null),r=[];for(const s of e.inputNames){const e=t[s];e instanceof p.Tensor?n[s]=(0,a.isONNXProxy)()?e.clone():e:r.push(s)}if(r.length>0)throw new Error(`An error occurred during model execution: "Missing the following inputs: ${r.join(", ")}.`);const s=Object.keys(t).length,i=e.inputNames.length;if(s>i){let n=Object.keys(t).filter((t=>!e.inputNames.includes(t)));console.warn(`WARNING: Too many inputs were provided (${s} > ${i}). The following inputs will be ignored: "${n.join(", ")}".`)}return n}(e,t);try{const t=Object.fromEntries(Object.entries(n).map((([e,t])=>[e,t.ort_tensor])));let r=await e.run(t);return r=I(r),r}catch(e){throw console.error(`An error occurred during model execution: "${e}".`),console.error("Inputs given to model:",n),e}}function I(e){for(let t in e)(0,a.isONNXTensor)(e[t])?e[t]=new p.Tensor(e[t]):"object"==typeof e[t]&&I(e[t]);return e}function z(e){if(e instanceof p.Tensor)return e;if(0===e.length)throw Error("items must be non-empty");if(Array.isArray(e[0])){if(e.some((t=>t.length!==e[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 p.Tensor("int64",BigInt64Array.from(e.flat().map((e=>BigInt(e)))),[e.length,e[0].length])}return new p.Tensor("int64",BigInt64Array.from(e.map((e=>BigInt(e)))),[1,e.length])}function O(e){return new p.Tensor("bool",[e],[1])}async function B(e,t){let{encoder_outputs:n,input_ids:r,decoder_input_ids:a,...s}=t;if(!n){const r=(0,o.pick)(t,e.sessions.model.inputNames);n=(await L(e,r)).last_hidden_state}s.input_ids=a,s.encoder_hidden_states=n,e.sessions.decoder_model_merged.inputNames.includes("encoder_attention_mask")&&(s.encoder_attention_mask=t.attention_mask);return await D(e,s,!0)}async function L(e,t){const n=e.sessions.model,r=(0,o.pick)(t,n.inputNames);if(n.inputNames.includes("inputs_embeds")&&!r.inputs_embeds){if(!t.input_ids)throw new Error("Both `input_ids` and `inputs_embeds` are missing in the model inputs.");r.inputs_embeds=await e.encode_text({input_ids:t.input_ids})}return n.inputNames.includes("token_type_ids")&&!r.token_type_ids&&(r.token_type_ids=new p.Tensor("int64",new BigInt64Array(r.input_ids.data.length),r.input_ids.dims)),await A(n,r)}async function D(e,t,n=!1){const r=e.sessions[n?"decoder_model_merged":"model"],{past_key_values:a,...s}=t;r.inputNames.includes("use_cache_branch")&&(s.use_cache_branch=O(!!a)),r.inputNames.includes("position_ids")&&s.attention_mask&&!s.position_ids&&(s.position_ids=function(e,t=null){const{input_ids:n,inputs_embeds:r,attention_mask:a}=e,[s,i]=a.dims,o=new BigInt64Array(a.data.length);for(let e=0;e<s;++e){const t=e*i;let n=BigInt(0);for(let e=0;e<i;++e){const r=t+e;0n===a.data[r]?o[r]=BigInt(1):(o[r]=n,n+=a.data[r])}}let l=new p.Tensor("int64",o,a.dims);if(t){const e=-(n??r).dims.at(1);l=l.slice(null,[e,null])}return l}(s,a)),e.addPastKeyValues(s,a);const i=(0,o.pick)(s,r.inputNames);return await A(r,i)}async function R(e,{input_ids:t=null,attention_mask:n=null,pixel_values:r=null,position_ids:a=null,inputs_embeds:s=null,past_key_values:i=null,generation_config:o=null,logits_processor:l=null,...u}){if(!s)if(s=await e.encode_text({input_ids:t}),r&&1!==t.dims[1]){const a=await e.encode_image({pixel_values:r});({inputs_embeds:s,attention_mask:n}=e._merge_input_ids_with_image_features({image_features:a,inputs_embeds:s,input_ids:t,attention_mask:n}))}else if(i&&r&&1===t.dims[1]){const e=t.dims[1],r=Object.values(i)[0].dims.at(-2);n=(0,p.cat)([(0,p.ones)([t.dims[0],r]),n.slice(null,[n.dims[1]-e,n.dims[1]])],1)}return await D(e,{inputs_embeds:s,past_key_values:i,attention_mask:n,position_ids:a,generation_config:o,logits_processor:l},!0)}function N(e,t,n,r){if(n.past_key_values){const t=Object.values(n.past_key_values)[0].dims.at(-2),{input_ids:r,attention_mask:a}=n;if(a&&a.dims[1]>r.dims[1]);else if(t<r.dims[1])n.input_ids=r.slice(null,[t,null]);else if(null!=e.config.image_token_index&&r.data.some((t=>t==e.config.image_token_index))){const a=e.config.num_image_tokens;if(!a)throw new Error("`num_image_tokens` is missing in the model configuration.");const s=r.dims[1]-(t-a);n.input_ids=r.slice(null,[-s,null]),n.attention_mask=(0,p.ones)([1,t+s])}}return n}function V(e,t,n,r){return n.past_key_values&&(t=t.map((e=>[e.at(-1)]))),{...n,decoder_input_ids:z(t)}}function j(e,...t){return e.config.is_encoder_decoder?V(e,...t):N(e,...t)}class q extends i.Callable{main_input_name="input_ids";forward_params=["input_ids","attention_mask"];constructor(e,t,n){super(),this.config=e,this.sessions=t,this.configs=n;const r=P.get(this.constructor),a=C.get(r);switch(this.can_generate=!1,this._forward=null,this._prepare_inputs_for_generation=null,a){case M:this.can_generate=!0,this._forward=D,this._prepare_inputs_for_generation=N;break;case v:case x:case $:this.can_generate=!0,this._forward=B,this._prepare_inputs_for_generation=V;break;case b:this._forward=B;break;case k:this.can_generate=!0,this._forward=R,this._prepare_inputs_for_generation=j;break;default:this._forward=L}this.can_generate&&this.forward_params.push("past_key_values"),this.custom_config=this.config["transformers.js_config"]??{}}async dispose(){const e=[];for(const t of Object.values(this.sessions))t?.handler?.dispose&&e.push(t.handler.dispose());return await Promise.all(e)}static async from_pretrained(e,{progress_callback:t=null,config:n=null,cache_dir:a=null,local_files_only:s=!1,revision:i="main",model_file_name:o=null,subfolder:l="onnx",device:d=null,dtype:c=null,use_external_data_format:p=null,session_options:h={}}={}){let m={progress_callback:t,config:n,cache_dir:a,local_files_only:s,revision:i,model_file_name:o,subfolder:l,device:d,dtype:c,use_external_data_format:p,session_options:h};const f=P.get(this),g=C.get(f);let _;if(n=m.config=await r.AutoConfig.from_pretrained(e,m),g===M)_=await Promise.all([E(e,{model:m.model_file_name??"model"},m),F(e,{generation_config:"generation_config.json"},m)]);else if(g===v||g===x)_=await Promise.all([E(e,{model:"encoder_model",decoder_model_merged:"decoder_model_merged"},m),F(e,{generation_config:"generation_config.json"},m)]);else if(g===T)_=await Promise.all([E(e,{model:"vision_encoder",prompt_encoder_mask_decoder:"prompt_encoder_mask_decoder"},m)]);else if(g===b)_=await Promise.all([E(e,{model:"encoder_model",decoder_model_merged:"decoder_model_merged"},m)]);else if(g===k){const t={embed_tokens:"embed_tokens",vision_encoder:"vision_encoder",decoder_model_merged:"decoder_model_merged"};n.is_encoder_decoder&&(t.model="encoder_model"),_=await Promise.all([E(e,t,m),F(e,{generation_config:"generation_config.json"},m)])}else g===$?_=await Promise.all([E(e,{model:"text_encoder",decoder_model_merged:"decoder_model_merged",encodec_decode:"encodec_decode"},m),F(e,{generation_config:"generation_config.json"},m)]):(g!==y&&console.warn(`Model type for '${f??n?.model_type}' not found, assuming encoder-only architecture. Please report this at ${u.GITHUB_ISSUE_URL}.`),_=await Promise.all([E(e,{model:m.model_file_name??"model"},m)]));return new this(n,..._)}async _call(e){return await this.forward(e)}async forward(e){return await this._forward(this,e)}get generation_config(){return this.configs?.generation_config??null}_get_logits_warper(e){const t=new d.LogitsProcessorList;return null!==e.temperature&&1!==e.temperature&&t.push(new d.TemperatureLogitsWarper(e.temperature)),null!==e.top_k&&0!==e.top_k&&t.push(new d.TopKLogitsWarper(e.top_k)),null!==e.top_p&&e.top_p<1&&t.push(new d.TopPLogitsWarper(e.top_p)),t}_get_logits_processor(e,t,n=null){const r=new d.LogitsProcessorList;if(null!==e.repetition_penalty&&1!==e.repetition_penalty&&r.push(new d.RepetitionPenaltyLogitsProcessor(e.repetition_penalty)),null!==e.no_repeat_ngram_size&&e.no_repeat_ngram_size>0&&r.push(new d.NoRepeatNGramLogitsProcessor(e.no_repeat_ngram_size)),null!==e.bad_words_ids&&r.push(new d.NoBadWordsLogitsProcessor(e.bad_words_ids,e.eos_token_id)),null!==e.min_length&&null!==e.eos_token_id&&e.min_length>0&&r.push(new d.MinLengthLogitsProcessor(e.min_length,e.eos_token_id)),null!==e.min_new_tokens&&null!==e.eos_token_id&&e.min_new_tokens>0&&r.push(new d.MinNewTokensLengthLogitsProcessor(t,e.min_new_tokens,e.eos_token_id)),null!==e.forced_bos_token_id&&r.push(new d.ForcedBOSTokenLogitsProcessor(e.forced_bos_token_id)),null!==e.forced_eos_token_id&&r.push(new d.ForcedEOSTokenLogitsProcessor(e.max_length,e.forced_eos_token_id)),null!==e.begin_suppress_tokens){const n=t>1||null===e.forced_bos_token_id?t:t+1;r.push(new d.SuppressTokensAtBeginLogitsProcessor(e.begin_suppress_tokens,n))}return null!==e.guidance_scale&&e.guidance_scale>1&&r.push(new d.ClassifierFreeGuidanceLogitsProcessor(e.guidance_scale)),null!==n&&r.extend(n),r}_prepare_generation_config(e,t,n=c.GenerationConfig){const r={...this.config};for(const e of["decoder","generator","text_config"])e in r&&Object.assign(r,r[e]);const a=new n(r);return Object.assign(a,this.generation_config??{}),e&&Object.assign(a,e),t&&Object.assign(a,(0,o.pick)(t,Object.getOwnPropertyNames(a))),a}_get_stopping_criteria(e,t=null){const n=new m.StoppingCriteriaList;return null!==e.max_length&&n.push(new m.MaxLengthCriteria(e.max_length,this.config.max_position_embeddings??null)),null!==e.eos_token_id&&n.push(new m.EosTokenCriteria(e.eos_token_id)),t&&n.extend(t),n}_validate_model_class(){if(!this.can_generate){const e=[mo,_o,ho,oo],t=P.get(this.constructor),n=new Set,r=this.config.model_type;for(const t of e){const e=t.get(r);e&&n.add(e[0])}let a=`The current model class (${t}) is not compatible with \`.generate()\`, as it doesn't have a language model head.`;throw n.size>0&&(a+=` Please use the following class instead: ${[...n].join(", ")}`),Error(a)}}prepare_inputs_for_generation(...e){return this._prepare_inputs_for_generation(this,...e)}_update_model_kwargs_for_generation({generated_input_ids:e,outputs:t,model_inputs:n,is_encoder_decoder:r}){return n.past_key_values=this.getPastKeyValues(t,n.past_key_values),n.input_ids=new p.Tensor("int64",e.flat(),[e.length,1]),r||(n.attention_mask=(0,p.cat)([n.attention_mask,(0,p.ones)([n.attention_mask.dims[0],1])],1)),n.position_ids=null,n}_prepare_model_inputs({inputs:e,bos_token_id:t,model_kwargs:n}){const r=(0,o.pick)(n,this.forward_params),a=this.main_input_name;if(a in r){if(e)throw new Error("`inputs`: {inputs}` were passed alongside {input_name} which is not allowed. Make sure to either pass {inputs} or {input_name}=...")}else r[a]=e;return{inputs_tensor:r[a],model_inputs:r,model_input_name:a}}async _prepare_encoder_decoder_kwargs_for_generation({inputs_tensor:e,model_inputs:t,model_input_name:n,generation_config:r}){if(this.sessions.model.inputNames.includes("inputs_embeds")&&!t.inputs_embeds&&"_prepare_inputs_embeds"in this){const{input_ids:e,pixel_values:n,attention_mask:r,...a}=t,s=await this._prepare_inputs_embeds(t);t={...a,...(0,o.pick)(s,["inputs_embeds","attention_mask"])}}let{last_hidden_state:a}=await L(this,t);if(null!==r.guidance_scale&&r.guidance_scale>1)a=(0,p.cat)([a,(0,p.full_like)(a,0)],0),"attention_mask"in t&&(t.attention_mask=(0,p.cat)([t.attention_mask,(0,p.zeros_like)(t.attention_mask)],0));else if(t.decoder_input_ids){const e=z(t.decoder_input_ids).dims[0];if(e!==a.dims[0]){if(1!==a.dims[0])throw new Error(`The encoder outputs have a different batch size (${a.dims[0]}) than the decoder inputs (${e}).`);a=(0,p.cat)(Array.from({length:e},(()=>a)),0)}}return t.encoder_outputs=a,t}_prepare_decoder_input_ids_for_generation({batch_size:e,model_input_name:t,model_kwargs:n,decoder_start_token_id:r,bos_token_id:a,generation_config:s}){let{decoder_input_ids:i,...o}=n;if(!(i instanceof p.Tensor)){if(i)Array.isArray(i[0])||(i=Array.from({length:e},(()=>i)));else if(r??=a,"musicgen"===this.config.model_type)i=Array.from({length:e*this.config.decoder.num_codebooks},(()=>[r]));else if(Array.isArray(r)){if(r.length!==e)throw new Error(`\`decoder_start_token_id\` expcted to have length ${e} but got ${r.length}`);i=r}else i=Array.from({length:e},(()=>[r]));i=z(i)}return n.decoder_attention_mask=(0,p.ones_like)(i),{input_ids:i,model_inputs:o}}async generate({inputs:e=null,generation_config:t=null,logits_processor:n=null,stopping_criteria:r=null,streamer:a=null,...s}){this._validate_model_class(),t=this._prepare_generation_config(t,s);let{inputs_tensor:i,model_inputs:o,model_input_name:l}=this._prepare_model_inputs({inputs:e,model_kwargs:s});const u=this.config.is_encoder_decoder;let d;u&&("encoder_outputs"in o||(o=await this._prepare_encoder_decoder_kwargs_for_generation({inputs_tensor:i,model_inputs:o,model_input_name:l,generation_config:t}))),u?({input_ids:d,model_inputs:o}=this._prepare_decoder_input_ids_for_generation({batch_size:o[l].dims.at(0),model_input_name:l,model_kwargs:o,decoder_start_token_id:t.decoder_start_token_id,bos_token_id:t.bos_token_id,generation_config:t})):d=o[l];let c=d.dims.at(-1);null!==t.max_new_tokens&&(t.max_length=c+t.max_new_tokens);const h=this._get_logits_processor(t,c,n),m=this._get_stopping_criteria(t,r),g=o[l].dims.at(0),_=f.LogitsSampler.getSampler(t),w=new Array(g).fill(0),y=d.tolist();let b;a&&a.put(y);let v={};for(;;){if(o=this.prepare_inputs_for_generation(y,o,t),b=await this.forward(o),t.output_attentions&&t.return_dict_in_generate){const e=this.getAttentions(b);for(const t in e)t in v||(v[t]=[]),v[t].push(e[t])}const e=h(y,b.logits.slice(null,-1,null)),n=[];for(let t=0;t<e.dims.at(0);++t){const r=e[t],a=await _(r);for(const[e,r]of a){const a=BigInt(e);w[t]+=r,y[t].push(a),n.push([a]);break}}a&&a.put(n);if(m(y).every((e=>e)))break;o=this._update_model_kwargs_for_generation({generated_input_ids:n,outputs:b,model_inputs:o,is_encoder_decoder:u})}a&&a.end();const x=this.getPastKeyValues(b,o.past_key_values,!0),M=new p.Tensor("int64",y.flat(),[y.length,y[0].length]);if(t.return_dict_in_generate)return{sequences:M,past_key_values:x,...v};for(const e of Object.values(b))"gpu-buffer"===e.location&&e.dispose();return M}getPastKeyValues(e,t,n=!1){const r=Object.create(null);for(const a in e)if(a.startsWith("present")){const s=a.replace("present","past_key_values"),i=a.includes("encoder");if(r[s]=i&&t?t[s]:e[a],t&&(!i||n)){const e=t[s];"gpu-buffer"===e.location&&e.dispose()}}return r}getAttentions(e){const t={};for(const n of["cross_attentions","encoder_attentions","decoder_attentions"])for(const r in e)r.startsWith(n)&&(n in t||(t[n]=[]),t[n].push(e[r]));return t}addPastKeyValues(e,t){if(t)Object.assign(e,t);else{const t=this.sessions.decoder_model_merged??this.sessions.model,n=t?.config?.kv_cache_dtype??"float32",a="float16"===n?new Uint16Array:[],s=(0,r.getKeyValueShapes)(this.config);for(const t in s)e[t]=new p.Tensor(n,a,s[t])}}async encode_image({pixel_values:e}){const t=(await A(this.sessions.vision_encoder,{pixel_values:e})).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 (${t.dims[1]}).`),this.config.num_image_tokens=t.dims[1]),t}async encode_text({input_ids:e}){return(await A(this.sessions.embed_tokens,{input_ids:e})).inputs_embeds}}class G{}class U extends G{constructor({last_hidden_state:e,hidden_states:t=null,attentions:n=null}){super(),this.last_hidden_state=e,this.hidden_states=t,this.attentions=n}}class W extends q{}class H extends W{}class X extends W{async _call(e){return new gl(await super._call(e))}}class K extends W{async _call(e){return new hl(await super._call(e))}}class Q extends W{async _call(e){return new fl(await super._call(e))}}class Y extends W{async _call(e){return new _l(await super._call(e))}}class Z extends q{}class J extends Z{}class ee extends q{}class te extends ee{}class ne extends ee{async _call(e){return new gl(await super._call(e))}}class re extends ee{async _call(e){return new hl(await super._call(e))}}class ae extends ee{async _call(e){return new fl(await super._call(e))}}class se extends ee{async _call(e){return new _l(await super._call(e))}}class ie extends q{}class oe extends ie{}class le extends ie{async _call(e){return new gl(await super._call(e))}}class ue extends ie{async _call(e){return new hl(await super._call(e))}}class de extends ie{async _call(e){return new fl(await super._call(e))}}class ce extends ie{async _call(e){return new _l(await super._call(e))}}class pe extends q{}class he extends pe{}class me extends pe{async _call(e){return new gl(await super._call(e))}}class fe extends pe{async _call(e){return new hl(await super._call(e))}}class ge extends pe{async _call(e){return new fl(await super._call(e))}}class _e extends pe{async _call(e){return new _l(await super._call(e))}}class we extends q{}class ye extends we{}class be extends we{async _call(e){return new gl(await super._call(e))}}class ve extends we{async _call(e){return new hl(await super._call(e))}}class xe extends we{async _call(e){return new fl(await super._call(e))}}class Me extends we{async _call(e){return new _l(await super._call(e))}}class Te extends q{}class ke extends Te{}class $e extends Te{async _call(e){return new gl(await super._call(e))}}class Ce extends Te{async _call(e){return new hl(await super._call(e))}}class Se extends Te{async _call(e){return new fl(await super._call(e))}}class Pe extends Te{async _call(e){return new _l(await super._call(e))}}class Ee extends q{}class Fe extends Ee{}class Ae extends Ee{async _call(e){return new gl(await super._call(e))}}class Ie extends Ee{async _call(e){return new hl(await super._call(e))}}class ze extends Ee{async _call(e){return new fl(await super._call(e))}}class Oe extends Ee{async _call(e){return new _l(await super._call(e))}}class Be extends q{}class Le extends Be{}class De extends Be{async _call(e){return new hl(await super._call(e))}}class Re extends Be{async _call(e){return new fl(await super._call(e))}}class Ne extends Be{async _call(e){return new _l(await super._call(e))}}class Ve extends Be{async _call(e){return new gl(await super._call(e))}}class je extends q{}class qe extends je{}class Ge extends je{async _call(e){return new gl(await super._call(e))}}class Ue extends je{async _call(e){return new hl(await super._call(e))}}class We extends je{async _call(e){return new fl(await super._call(e))}}class He extends q{}class Xe extends He{}class Ke extends He{async _call(e){return new gl(await super._call(e))}}class Qe extends He{async _call(e){return new hl(await super._call(e))}}class Ye extends He{async _call(e){return new _l(await super._call(e))}}class Ze extends q{}class Je extends Ze{}class et extends Ze{async _call(e){return new gl(await super._call(e))}}class tt extends Ze{async _call(e){return new hl(await super._call(e))}}class nt extends Ze{async _call(e){return new fl(await super._call(e))}}class rt extends Ze{async _call(e){return new _l(await super._call(e))}}class at extends q{}class st extends at{}class it extends at{async _call(e){return new gl(await super._call(e))}}class ot extends at{async _call(e){return new hl(await super._call(e))}}class lt extends at{async _call(e){return new _l(await super._call(e))}}class ut extends q{}class dt extends ut{}class ct extends ut{async _call(e){return new hl(await super._call(e))}}class pt extends ut{async _call(e){return new _l(await super._call(e))}}class ht extends ut{async _call(e){return new gl(await super._call(e))}}class mt extends q{forward_params=["input_ids","attention_mask","encoder_outputs","decoder_input_ids","decoder_attention_mask","past_key_values"]}class ft extends mt{}class gt extends mt{}class _t extends q{}class wt extends _t{}class yt extends _t{}class bt extends q{}class vt extends bt{}class xt extends bt{}class Mt extends q{}class Tt extends Mt{}class kt extends Mt{}class $t extends Mt{async _call(e){return new hl(await super._call(e))}}class Ct extends q{}class St extends Ct{}class Pt extends Ct{}class Et extends Ct{async _call(e){return new hl(await super._call(e))}}class Ft extends Ct{}class At extends q{}class It extends At{}class zt extends At{}class Ot extends q{}class Bt extends Ot{}class Lt extends Ot{}class Dt extends q{}class Rt extends Dt{}class Nt extends Dt{async _call(e){return new gl(await super._call(e))}}class Vt extends Dt{async _call(e){return new hl(await super._call(e))}}class jt extends Dt{async _call(e){return new fl(await super._call(e))}}class qt extends Dt{async _call(e){return new _l(await super._call(e))}}class Gt extends q{}class Ut extends Gt{}class Wt extends Gt{async _call(e){return new gl(await super._call(e))}}class Ht extends Gt{async _call(e){return new hl(await super._call(e))}}class Xt extends Gt{async _call(e){return new fl(await super._call(e))}}class Kt extends Gt{async _call(e){return new _l(await super._call(e))}}class Qt extends q{}class Yt extends Qt{}class Zt extends Qt{async _call(e){return new gl(await super._call(e))}}class Jt extends Qt{async _call(e){return new hl(await super._call(e))}}class en extends Qt{async _call(e){return new fl(await super._call(e))}}class tn extends Qt{async _call(e){return new _l(await super._call(e))}}class nn extends q{}class rn extends nn{}class an extends nn{}class sn extends q{requires_attention_mask=!1;main_input_name="input_features";forward_params=["input_features","attention_mask","decoder_input_ids","decoder_attention_mask","past_key_values"]}class on extends sn{}class ln extends sn{_prepare_generation_config(e,t){return super._prepare_generation_config(e,t,_.WhisperGenerationConfig)}_retrieve_init_tokens(e){const t=[e.decoder_start_token_id];let n=e.language;const r=e.task;if(e.is_multilingual){n||(console.warn("No language specified - defaulting to English (en)."),n="en");const a=`<|${(0,w.whisper_language_to_code)(n)}|>`;t.push(e.lang_to_id[a]),t.push(e.task_to_id[r??"transcribe"])}else if(n||r)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!e.return_timestamps&&e.no_timestamps_token_id&&t.at(-1)!==e.no_timestamps_token_id?t.push(e.no_timestamps_token_id):e.return_timestamps&&t.at(-1)===e.no_timestamps_token_id&&(console.warn("<|notimestamps|> prompt token is removed from generation_config since `return_timestamps` is set to `true`."),t.pop()),t.filter((e=>null!=e))}async generate({inputs:e=null,generation_config:t=null,logits_processor:n=null,stopping_criteria:r=null,...a}){t=this._prepare_generation_config(t,a);const s=a.decoder_input_ids??this._retrieve_init_tokens(t);if(t.return_timestamps&&(n??=new d.LogitsProcessorList,n.push(new d.WhisperTimeStampLogitsProcessor(t,s))),t.begin_suppress_tokens&&(n??=new d.LogitsProcessorList,n.push(new d.SuppressTokensAtBeginLogitsProcessor(t.begin_suppress_tokens,s.length))),t.return_token_timestamps){if(!t.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.");"translate"===t.task&&console.warn("Token-level timestamps may not be reliable for task 'translate'."),t.output_attentions=!0,t.return_dict_in_generate=!0}const i=await super.generate({inputs:e,generation_config:t,logits_processor:n,decoder_input_ids:s,...a});return t.return_token_timestamps&&(i.token_timestamps=this._extract_token_timestamps(i,t.alignment_heads,t.num_frames)),i}_extract_token_timestamps(e,t,n=null,r=.02){if(!e.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`.");null==n&&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 a=this.config.median_filter_width;void 0===a&&(console.warn("Model config has no `median_filter_width`, using default value of 7."),a=7);const s=e.cross_attentions,i=Array.from({length:this.config.decoder_layers},((e,t)=>(0,p.cat)(s.map((e=>e[t])),2))),l=(0,p.stack)(t.map((([e,t])=>{if(e>=i.length)throw new Error(`Layer index ${e} is out of bounds for cross attentions (length ${i.length}).`);return n?i[e].slice(null,t,null,[0,n]):i[e].slice(null,t)}))).transpose(1,0,2,3),[u,d]=(0,p.std_mean)(l,-2,0,!0),c=l.clone();for(let e=0;e<c.dims[0];++e){const t=c[e];for(let n=0;n<t.dims[0];++n){const r=t[n],s=u[e][n][0].data,i=d[e][n][0].data;for(let e=0;e<r.dims[0];++e){let t=r[e].data;for(let e=0;e<t.length;++e)t[e]=(t[e]-i[e])/s[e];t.set((0,h.medianFilter)(t,a))}}}const m=[(0,p.mean)(c,1)],f=e.sequences.dims,g=new p.Tensor("float32",new Float32Array(f[0]*f[1]),f);for(let e=0;e<f[0];++e){const t=m[e].neg().squeeze_(0),[n,a]=(0,h.dynamic_time_warping)(t.tolist()),s=Array.from({length:n.length-1},((e,t)=>n[t+1]-n[t])),i=(0,o.mergeArrays)([1],s).map((e=>!!e)),l=[];for(let e=0;e<i.length;++e)i[e]&&l.push(a[e]*r);g[e].data.set(l,1)}return g}}class un extends q{main_input_name="pixel_values";forward_params=["pixel_values","decoder_input_ids","encoder_hidden_states","past_key_values"]}class dn extends q{forward_params=["input_ids","pixel_values","attention_mask","position_ids","past_key_values"]}class cn extends dn{_merge_input_ids_with_image_features({inputs_embeds:e,image_features:t,input_ids:n,attention_mask:r}){const a=this.config.image_token_index,s=n.tolist().map((e=>e.findIndex((e=>e==a)))),i=s.every((e=>-1===e)),o=s.every((e=>-1!==e));if(!i&&!o)throw new Error("Every input should contain either 0 or 1 image token.");if(i)return{inputs_embeds:e,attention_mask:r};const l=[],u=[];for(let n=0;n<s.length;++n){const a=s[n],i=e[n],o=t[n],d=r[n];l.push((0,p.cat)([i.slice([0,a]),o,i.slice([a+1,i.dims[0]])],0)),u.push((0,p.cat)([d.slice([0,a]),(0,p.ones)([o.dims[0]]),d.slice([a+1,d.dims[0]])],0))}return{inputs_embeds:(0,p.stack)(l,0),attention_mask:(0,p.stack)(u,0)}}}class pn extends cn{}class hn extends q{forward_params=["input_ids","inputs_embeds","attention_mask","pixel_values","encoder_outputs","decoder_input_ids","decoder_inputs_embeds","decoder_attention_mask","past_key_values"];main_input_name="inputs_embeds"}class mn extends hn{_merge_input_ids_with_image_features({inputs_embeds:e,image_features:t,input_ids:n,attention_mask:r}){return{inputs_embeds:(0,p.cat)([t,e],1),attention_mask:(0,p.cat)([(0,p.ones)(t.dims.slice(0,2)),r],1)}}async _prepare_inputs_embeds({input_ids:e,pixel_values:t,inputs_embeds:n,attention_mask:r}){if(!e&&!t)throw new Error("Either `input_ids` or `pixel_values` should be provided.");let a,s;return e&&(a=await this.encode_text({input_ids:e})),t&&(s=await this.encode_image({pixel_values:t})),a&&s?({inputs_embeds:n,attention_mask:r}=this._merge_input_ids_with_image_features({inputs_embeds:a,image_features:s,input_ids:e,attention_mask:r})):n=a||s,{inputs_embeds:n,attention_mask:r}}async forward({input_ids:e,pixel_values:t,attention_mask:n,decoder_input_ids:r,decoder_attention_mask:a,encoder_outputs:s,past_key_values:i,inputs_embeds:o,decoder_inputs_embeds:l}){if(o||({inputs_embeds:o,attention_mask:n}=await this._prepare_inputs_embeds({input_ids:e,pixel_values:t,inputs_embeds:o,attention_mask:n})),!s){let{last_hidden_state:e}=await L(this,{inputs_embeds:o,attention_mask:n});s=e}if(!l){if(!r)throw new Error("Either `decoder_input_ids` or `decoder_inputs_embeds` should be provided.");l=await this.encode_text({input_ids:r})}const u={inputs_embeds:l,attention_mask:a,encoder_attention_mask:n,encoder_hidden_states:s,past_key_values:i};return await D(this,u,!0)}}class fn extends q{}class gn extends fn{}class _n extends fn{static async from_pretrained(e,t={}){return t.model_file_name??="text_model",super.from_pretrained(e,t)}}class wn extends fn{static async from_pretrained(e,t={}){return t.model_file_name??="text_model",super.from_pretrained(e,t)}}class yn extends fn{static async from_pretrained(e,t={}){return t.model_file_name??="vision_model",super.from_pretrained(e,t)}}class bn extends fn{static async from_pretrained(e,t={}){return t.model_file_name??="vision_model",super.from_pretrained(e,t)}}class vn extends q{}class xn extends vn{}class Mn extends vn{static async from_pretrained(e,t={}){return t.model_file_name??="text_model",super.from_pretrained(e,t)}}class Tn extends fn{static async from_pretrained(e,t={}){return t.model_file_name??="vision_model",super.from_pretrained(e,t)}}class kn extends q{}class $n extends kn{}class Cn extends q{}class Sn extends Cn{}class Pn extends Cn{}class En extends q{}class Fn extends En{}class An extends En{}class In extends q{}class zn extends In{}class On extends In{}class Bn extends q{}class Ln extends Bn{}class Dn extends Bn{}class Rn extends q{}class Nn extends Rn{}class Vn extends Rn{}class jn extends q{}class qn extends jn{}class Gn extends jn{}class Un extends q{}class Wn extends Un{}class Hn extends Un{}class Xn extends q{}class Kn extends Xn{}class Qn extends Xn{}class Yn extends q{}class Zn extends Yn{}class Jn extends Yn{}class er extends q{}class tr extends er{}class nr extends er{}class rr extends q{}class ar extends rr{}class sr extends rr{}class ir extends q{}class or extends ir{}class lr extends ir{}class ur extends q{}class dr extends ur{}class cr extends ur{}class pr extends q{}class hr extends pr{}class mr extends pr{}class fr extends q{}class gr extends fr{}class _r extends fr{}class wr extends q{}class yr extends wr{}class br extends wr{}class vr extends q{}class xr extends vr{}class Mr extends vr{}class Tr extends q{}class kr extends Tr{}class $r extends Tr{}class Cr extends q{}class Sr extends Cr{}class Pr extends Cr{}class Er extends q{}class Fr extends Er{}class Ar extends Er{}class Ir extends q{}class zr extends Ir{}class Or extends Ir{async _call(e){return new hl(await super._call(e))}}class Br extends q{}class Lr extends Br{}class Dr extends Br{async _call(e){return new hl(await super._call(e))}}class Rr extends q{}class Nr extends Rr{}class Vr extends q{}class jr extends Vr{}class qr extends Vr{async _call(e){return new hl(await super._call(e))}}class Gr extends q{}class Ur extends Gr{}class Wr extends q{}class Hr extends Wr{}class Xr extends Wr{async _call(e){return new hl(await super._call(e))}}class Kr extends q{}class Qr extends Kr{async _call(e){return new bl(await super._call(e))}}class Yr extends q{}class Zr extends Yr{}class Jr extends Yr{async _call(e){return new hl(await super._call(e))}}class ea extends q{}class ta extends ea{}class na extends ea{async _call(e){return new hl(await super._call(e))}}class ra extends q{}class aa extends ra{}class sa extends ra{}class ia extends q{}class oa extends ia{}class la extends ia{}class ua extends q{}class da extends ua{}class ca extends ua{async _call(e){return new hl(await super._call(e))}}class pa extends q{}class ha extends pa{}class ma extends pa{async _call(e){return new ga(await super._call(e))}}class fa extends pa{async _call(e){return new _a(await super._call(e))}}class ga extends G{constructor({logits:e,pred_boxes:t}){super(),this.logits=e,this.pred_boxes=t}}class _a extends G{constructor({logits:e,pred_boxes:t,pred_masks:n}){super(),this.logits=e,this.pred_boxes=t,this.pred_masks=n}}class wa extends q{}class ya extends wa{}class ba extends wa{async _call(e){return new va(await super._call(e))}}class va extends G{constructor({logits:e,pred_boxes:t}){super(),this.logits=e,this.pred_boxes=t}}class xa extends q{}class Ma extends xa{}class Ta extends xa{async _call(e){return new ka(await super._call(e))}}class ka extends ga{}class $a extends q{}class Ca extends $a{}class Sa extends $a{async _call(e){return new hl(await super._call(e))}}class Pa extends q{}class Ea extends Pa{}class Fa extends Pa{async _call(e){return new hl(await super._call(e))}}class Aa extends q{}class Ia extends Aa{}class za extends Aa{async _call(e){return new hl(await super._call(e))}}class Oa extends q{}class Ba extends Oa{}class La extends Oa{async _call(e){return new hl(await super._call(e))}}class Da extends q{}class Ra extends Da{}class Na extends Da{}class Va extends q{}class ja extends Va{}class qa extends Va{}class Ga extends q{}class Ua extends Ga{}class Wa extends q{}class Ha extends Wa{}class Xa extends Wa{}class Ka extends Wa{}class Qa extends q{}class Ya extends Qa{}class Za extends q{}class Ja extends Za{}class es extends Za{}class ts extends q{}class ns extends ts{}class rs extends ts{}class as extends q{}class ss extends as{}class is extends q{}class os extends is{}class ls extends is{async _call(e){return new hl(await super._call(e))}}class us extends q{}class ds extends us{}class cs extends us{async _call(e){return new hl(await super._call(e))}}class ps extends q{}class hs extends ps{}class ms extends ps{async _call(e){return new hl(await super._call(e))}}class fs extends q{}class gs extends fs{}class _s extends fs{async _call(e){return new ws(await super._call(e))}}class ws extends G{constructor({logits:e,pred_boxes:t}){super(),this.logits=e,this.pred_boxes=t}}class ys extends q{}class bs extends ys{async get_image_embeddings({pixel_values:e}){return await L(this,{pixel_values:e})}async forward(e){if(e.image_embeddings&&e.image_positional_embeddings||(e={...e,...await this.get_image_embeddings(e)}),!e.input_labels&&e.input_points){const t=e.input_points.dims.slice(0,-1),n=t.reduce(((e,t)=>e*t),1);e.input_labels=new p.Tensor("int64",new BigInt64Array(n).fill(1n),t)}const t={image_embeddings:e.image_embeddings,image_positional_embeddings:e.image_positional_embeddings};return e.input_points&&(t.input_points=e.input_points),e.input_labels&&(t.input_labels=e.input_labels),e.input_boxes&&(t.input_boxes=e.input_boxes),await A(this.sessions.prompt_encoder_mask_decoder,t)}async _call(e){return new vs(await super._call(e))}}class vs extends G{constructor({iou_scores:e,pred_masks:t}){super(),this.iou_scores=e,this.pred_masks=t}}class xs extends q{}class Ms extends xs{}class Ts extends xs{}class ks extends q{}class $s extends ks{}class Cs extends ks{}class Ss extends q{}class Ps extends Ss{}class Es extends Ss{async _call(e){return new wl(await super._call(e))}}class Fs extends Ss{async _call(e){return new hl(await super._call(e))}}class As extends Ss{async _call(e){return new fl(await super._call(e))}}class Is extends q{}class zs extends Is{}class Os extends Is{async _call(e){return new fl(await super._call(e))}}class Bs extends q{}class Ls extends Bs{}class Ds extends q{}class Rs extends Ds{}class Ns extends Ds{async _call(e){return new wl(await super._call(e))}}class Vs extends Ds{async _call(e){return new hl(await super._call(e))}}class js extends q{}class qs extends js{}class Gs extends js{async _call(e){return new wl(await super._call(e))}}class Us extends js{async _call(e){return new hl(await super._call(e))}}class Ws extends js{async _call(e){return new fl(await super._call(e))}}class Hs extends q{}class Xs extends Hs{}class Ks extends Hs{async _call(e){return new wl(await super._call(e))}}class Qs extends Hs{async _call(e){return new hl(await super._call(e))}}class Ys extends q{}class Zs extends Ss{}class Js extends Ss{async _call(e){return new wl(await super._call(e))}}class ei extends Ss{async _call(e){return new hl(await super._call(e))}}class ti extends q{}class ni extends ti{}class ri extends ti{async _call(e){return new wl(await super._call(e))}}class ai extends ti{async _call(e){return new hl(await super._call(e))}}class si extends ti{async _call(e){return new ml(await super._call(e))}}class ii extends ti{async _call(e){return new fl(await super._call(e))}}class oi extends q{}class li extends oi{}class ui extends oi{}class di extends oi{async generate_speech(e,t,{threshold:n=.5,minlenratio:r=0,maxlenratio:a=20,vocoder:s=null}={}){const i={input_ids:e},{encoder_outputs:o,encoder_attention_mask:l}=await L(this,i),u=o.dims[1]/this.config.reduction_factor,d=Math.floor(u*a),c=Math.floor(u*r),h=this.config.num_mel_bins;let m=[],f=null,g=null,_=0;for(;;){++_;const e=O(!!g);let r;r=g?g.output_sequence_out:new p.Tensor("float32",new Float32Array(h),[1,1,h]);let a={use_cache_branch:e,output_sequence:r,encoder_attention_mask:l,speaker_embeddings:t,encoder_hidden_states:o};this.addPastKeyValues(a,f),g=await A(this.sessions.decoder_model_merged,a),f=this.getPastKeyValues(g,f);const{prob:s,spectrum:i}=g;if(m.push(i),_>=c&&(Array.from(s.data).filter((e=>e>=n)).length>0||_>=d))break}const w=(0,p.cat)(m),{waveform:y}=await A(s.sessions.model,{spectrogram:w});return{spectrogram:w,waveform:y}}}class ci extends q{main_input_name="spectrogram"}class pi extends q{}class hi extends pi{}class mi extends q{}class fi extends mi{}class gi extends mi{}class _i extends q{}class wi extends _i{}class yi extends _i{}class bi extends q{}class vi extends bi{}class xi extends bi{}class Mi extends q{}class Ti extends Mi{}class ki extends Mi{static async from_pretrained(e,t={}){return t.model_file_name??="text_model",super.from_pretrained(e,t)}}class $i extends Mi{static async from_pretrained(e,t={}){return t.model_file_name??="audio_model",super.from_pretrained(e,t)}}class Ci extends q{}class Si extends Ci{async _call(e){return new vl(await super._call(e))}}class Pi extends q{}class Ei extends Pi{}class Fi extends Pi{}class Ai extends Pi{}class Ii extends q{}class zi extends Ii{}class Oi extends Ii{}class Bi extends q{}class Li extends Bi{}class Di extends Bi{async _call(e){return new hl(await super._call(e))}}class Ri extends q{}class Ni extends Ri{}class Vi extends Ri{}class ji extends q{forward_params=["input_ids","attention_mask","encoder_outputs","decoder_input_ids","decoder_attention_mask","past_key_values"];_apply_and_filter_by_delay_pattern_mask(e){const[t,n]=e.dims,r=this.config.decoder.num_codebooks,a=n-r;let s=0;for(let t=0;t<e.size;++t){if(e.data[t]===this.config.decoder.pad_token_id)continue;const i=t%n-Math.floor(t/n)%r;i>0&&i<=a&&(e.data[s++]=e.data[t])}const i=Math.floor(t/r),o=s/(i*r);return new p.Tensor(e.type,e.data.slice(0,s),[i,r,o])}prepare_inputs_for_generation(e,t,n){let r=structuredClone(e);for(let e=0;e<r.length;++e)for(let t=0;t<r[e].length;++t)e%this.config.decoder.num_codebooks>=t&&(r[e][t]=BigInt(this.config.decoder.pad_token_id));null!==n.guidance_scale&&n.guidance_scale>1&&(r=r.concat(r));return super.prepare_inputs_for_generation(r,t,n)}async generate(e){const t=await super.generate(e),n=this._apply_and_filter_by_delay_pattern_mask(t).unsqueeze_(0),{audio_values:r}=await A(this.sessions.encodec_decode,{audio_codes:n});return r}}class qi extends q{}class Gi extends qi{}class Ui extends qi{async _call(e){return new hl(await super._call(e))}}class Wi extends q{}class Hi extends Wi{}class Xi extends Wi{async _call(e){return new hl(await super._call(e))}}class Ki extends q{}class Qi extends Ki{}class Yi extends Ki{async _call(e){return new hl(await super._call(e))}}class Zi extends q{}class Ji extends Zi{}class eo extends Zi{async _call(e){return new hl(await super._call(e))}}class to extends q{}class no extends to{}class ro{static MODEL_CLASS_MAPPINGS=null;static BASE_IF_FAIL=!1;static async from_pretrained(e,{progress_callback:t=null,config:n=null,cache_dir:a=null,local_files_only:s=!1,revision:i="main",model_file_name:o=null,subfolder:l="onnx",device:u=null,dtype:d=null,use_external_data_format:c=null,session_options:p={}}={}){const h={progress_callback:t,config:n,cache_dir:a,local_files_only:s,revision:i,model_file_name:o,subfolder:l,device:u,dtype:d,use_external_data_format:c,session_options:p};if(h.config=await r.AutoConfig.from_pretrained(e,h),!this.MODEL_CLASS_MAPPINGS)throw new Error("`MODEL_CLASS_MAPPINGS` not implemented for this type of `AutoClass`: "+this.name);for(const t of this.MODEL_CLASS_MAPPINGS){const n=t.get(h.config.model_type);if(n)return await n[1].from_pretrained(e,h)}if(this.BASE_IF_FAIL)return console.warn(`Unknown model class "${h.config.model_type}", attempting to construct from base class.`),await q.from_pretrained(e,h);throw Error(`Unsupported model type: ${h.config.model_type}`)}}const ao=new Map([["bert",["BertModel",H]],["nomic_bert",["NomicBertModel",J]],["roformer",["RoFormerModel",te]],["electra",["ElectraModel",he]],["esm",["EsmModel",qe]],["convbert",["ConvBertModel",oe]],["camembert",["CamembertModel",ye]],["deberta",["DebertaModel",ke]],["deberta-v2",["DebertaV2Model",Fe]],["mpnet",["MPNetModel",Je]],["albert",["AlbertModel",dt]],["distilbert",["DistilBertModel",Le]],["roberta",["RobertaModel",Rt]],["xlm",["XLMModel",Ut]],["xlm-roberta",["XLMRobertaModel",Yt]],["clap",["ClapModel",Ti]],["clip",["CLIPModel",gn]],["clipseg",["CLIPSegModel",Sn]],["chinese_clip",["ChineseCLIPModel",$n]],["siglip",["SiglipModel",xn]],["mobilebert",["MobileBertModel",Xe]],["squeezebert",["SqueezeBertModel",st]],["wav2vec2",["Wav2Vec2Model",Ps]],["wav2vec2-bert",["Wav2Vec2BertModel",Xs]],["unispeech",["UniSpeechModel",Rs]],["unispeech-sat",["UniSpeechSatModel",qs]],["hubert",["HubertModel",Zs]],["wavlm",["WavLMModel",ni]],["audio-spectrogram-transformer",["ASTModel",rn]],["vits",["VitsModel",Si]],["pyannote",["PyAnnoteModel",zs]],["wespeaker-resnet",["WeSpeakerResNetModel",Ls]],["detr",["DetrModel",ha]],["rt_detr",["RTDetrModel",ya]],["table-transformer",["TableTransformerModel",Ma]],["vit",["ViTModel",zr]],["pvt",["PvtModel",Lr]],["vit_msn",["ViTMSNModel",jr]],["vit_mae",["ViTMAEModel",Nr]],["groupvit",["GroupViTModel",Ur]],["fastvit",["FastViTModel",Hr]],["mobilevit",["MobileViTModel",Zr]],["mobilevitv2",["MobileViTV2Model",ta]],["owlvit",["OwlViTModel",aa]],["owlv2",["Owlv2Model",oa]],["beit",["BeitModel",da]],["deit",["DeiTModel",Ca]],["hiera",["HieraModel",Ea]],["convnext",["ConvNextModel",os]],["convnextv2",["ConvNextV2Model",ds]],["dinov2",["Dinov2Model",hs]],["resnet",["ResNetModel",Ia]],["swin",["SwinModel",Ba]],["swin2sr",["Swin2SRModel",Ra]],["donut-swin",["DonutSwinModel",ss]],["yolos",["YolosModel",gs]],["dpt",["DPTModel",ja]],["glpn",["GLPNModel",ns]],["hifigan",["SpeechT5HifiGan",ci]],["efficientnet",["EfficientNetModel",Li]],["decision_transformer",["DecisionTransformerModel",no]],["mobilenet_v1",["MobileNetV1Model",Gi]],["mobilenet_v2",["MobileNetV2Model",Hi]],["mobilenet_v3",["MobileNetV3Model",Qi]],["mobilenet_v4",["MobileNetV4Model",Ji]],["maskformer",["MaskFormerModel",Ja]]]),so=new Map([["t5",["T5Model",ft]],["longt5",["LongT5Model",wt]],["mt5",["MT5Model",vt]],["bart",["BartModel",Tt]],["mbart",["MBartModel",St]],["marian",["MarianModel",Ms]],["whisper",["WhisperModel",on]],["m2m_100",["M2M100Model",$s]],["blenderbot",["BlenderbotModel",It]],["blenderbot-small",["BlenderbotSmallModel",Bt]]]),io=new Map([["bloom",["BloomModel",kr]],["jais",["JAISModel",zn]],["gpt2",["GPT2Model",Fn]],["gptj",["GPTJModel",qn]],["gpt_bigcode",["GPTBigCodeModel",Wn]],["gpt_neo",["GPTNeoModel",Ln]],["gpt_neox",["GPTNeoXModel",Nn]],["codegen",["CodeGenModel",Kn]],["llama",["LlamaModel",Zn]],["granite",["GraniteModel",tr]],["cohere",["CohereModel",ar]],["gemma",["GemmaModel",or]],["gemma2",["Gemma2Model",dr]],["openelm",["OpenELMModel",hr]],["qwen2",["Qwen2Model",gr]],["phi",["PhiModel",yr]],["phi3",["Phi3Model",xr]],["mpt",["MptModel",Sr]],["opt",["OPTModel",Fr]],["mistral",["MistralModel",fi]],["starcoder2",["Starcoder2Model",wi]],["falcon",["FalconModel",vi]],["stablelm",["StableLmModel",zi]]]),oo=new Map([["speecht5",["SpeechT5ForSpeechToText",ui]],["whisper",["WhisperForConditionalGeneration",ln]]]),lo=new Map([["speecht5",["SpeechT5ForTextToSpeech",di]]]),uo=new Map([["vits",["VitsModel",Si]],["musicgen",["MusicgenForConditionalGeneration",ji]]]),co=new Map([["bert",["BertForSequenceClassification",K]],["roformer",["RoFormerForSequenceClassification",re]],["electra",["ElectraForSequenceClassification",fe]],["esm",["EsmForSequenceClassification",Ue]],["convbert",["ConvBertForSequenceClassification",ue]],["camembert",["CamembertForSequenceClassification",ve]],["deberta",["DebertaForSequenceClassification",Ce]],["deberta-v2",["DebertaV2ForSequenceClassification",Ie]],["mpnet",["MPNetForSequenceClassification",tt]],["albert",["AlbertForSequenceClassification",ct]],["distilbert",["DistilBertForSequenceClassification",De]],["roberta",["RobertaForSequenceClassification",Vt]],["xlm",["XLMForSequenceClassification",Ht]],["xlm-roberta",["XLMRobertaForSequenceClassification",Jt]],["bart",["BartForSequenceClassification",$t]],["mbart",["MBartForSequenceClassification",Et]],["mobilebert",["MobileBertForSequenceClassification",Qe]],["squeezebert",["SqueezeBertForSequenceClassification",ot]]]),po=new Map([["bert",["BertForTokenClassification",Q]],["roformer",["RoFormerForTokenClassification",ae]],["electra",["ElectraForTokenClassification",ge]],["esm",["EsmForTokenClassification",We]],["convbert",["ConvBertForTokenClassification",de]],["camembert",["CamembertForTokenClassification",xe]],["deberta",["DebertaForTokenClassification",Se]],["deberta-v2",["DebertaV2ForTokenClassification",ze]],["mpnet",["MPNetForTokenClassification",nt]],["distilbert",["DistilBertForTokenClassification",Re]],["roberta",["RobertaForTokenClassification",jt]],["xlm",["XLMForTokenClassification",Xt]],["xlm-roberta",["XLMRobertaForTokenClassification",en]]]),ho=new Map([["t5",["T5ForConditionalGeneration",gt]],["longt5",["LongT5ForConditionalGeneration",yt]],["mt5",["MT5ForConditionalGeneration",xt]],["bart",["BartForConditionalGeneration",kt]],["mbart",["MBartForConditionalGeneration",Pt]],["marian",["MarianMTModel",Ts]],["m2m_100",["M2M100ForConditionalGeneration",Cs]],["blenderbot",["BlenderbotForConditionalGeneration",zt]],["blenderbot-small",["BlenderbotSmallForConditionalGeneration",Lt]]]),mo=new Map([["bloom",["BloomForCausalLM",$r]],["gpt2",["GPT2LMHeadModel",An]],["jais",["JAISLMHeadModel",On]],["gptj",["GPTJForCausalLM",Gn]],["gpt_bigcode",["GPTBigCodeForCausalLM",Hn]],["gpt_neo",["GPTNeoForCausalLM",Dn]],["gpt_neox",["GPTNeoXForCausalLM",Vn]],["codegen",["CodeGenForCausalLM",Qn]],["llama",["LlamaForCausalLM",Jn]],["granite",["GraniteForCausalLM",nr]],["cohere",["CohereForCausalLM",sr]],["gemma",["GemmaForCausalLM",lr]],["gemma2",["Gemma2ForCausalLM",cr]],["openelm",["OpenELMForCausalLM",mr]],["qwen2",["Qwen2ForCausalLM",_r]],["phi",["PhiForCausalLM",br]],["phi3",["Phi3ForCausalLM",Mr]],["mpt",["MptForCausalLM",Pr]],["opt",["OPTForCausalLM",Ar]],["mbart",["MBartForCausalLM",Ft]],["mistral",["MistralForCausalLM",gi]],["starcoder2",["Starcoder2ForCausalLM",yi]],["falcon",["FalconForCausalLM",xi]],["trocr",["TrOCRForCausalLM",hi]],["stablelm",["StableLmForCausalLM",Oi]]]),fo=new Map([["bert",["BertForMaskedLM",X]],["roformer",["RoFormerForMaskedLM",ne]],["electra",["ElectraForMaskedLM",me]],["esm",["EsmForMaskedLM",Ge]],["convbert",["ConvBertForMaskedLM",le]],["camembert",["CamembertForMaskedLM",be]],["deberta",["DebertaForMaskedLM",$e]],["deberta-v2",["DebertaV2ForMaskedLM",Ae]],["mpnet",["MPNetForMaskedLM",et]],["albert",["AlbertForMaskedLM",ht]],["distilbert",["DistilBertForMaskedLM",Ve]],["roberta",["RobertaForMaskedLM",Nt]],["xlm",["XLMWithLMHeadModel",Wt]],["xlm-roberta",["XLMRobertaForMaskedLM",Zt]],["mobilebert",["MobileBertForMaskedLM",Ke]],["squeezebert",["SqueezeBertForMaskedLM",it]]]),go=new Map([["bert",["BertForQuestionAnswering",Y]],["roformer",["RoFormerForQuestionAnswering",se]],["electra",["ElectraForQuestionAnswering",_e]],["convbert",["ConvBertForQuestionAnswering",ce]],["camembert",["CamembertForQuestionAnswering",Me]],["deberta",["DebertaForQuestionAnswering",Pe]],["deberta-v2",["DebertaV2ForQuestionAnswering",Oe]],["mpnet",["MPNetForQuestionAnswering",rt]],["albert",["AlbertForQuestionAnswering",pt]],["distilbert",["DistilBertForQuestionAnswering",Ne]],["roberta",["RobertaForQuestionAnswering",qt]],["xlm",["XLMForQuestionAnswering",Kt]],["xlm-roberta",["XLMRobertaForQuestionAnswering",tn]],["mobilebert",["MobileBertForQuestionAnswering",Ye]],["squeezebert",["SqueezeBertForQuestionAnswering",lt]]]),_o=new Map([["vision-encoder-decoder",["VisionEncoderDecoderModel",un]]]),wo=new Map([["llava",["LlavaForConditionalGeneration",cn]],["moondream1",["Moondream1ForConditionalGeneration",pn]],["florence2",["Florence2ForConditionalGeneration",mn]]]),yo=new Map([["vision-encoder-decoder",["VisionEncoderDecoderModel",un]]]),bo=new Map([["vit",["ViTForImageClassification",Or]],["pvt",["PvtForImageClassification",Dr]],["vit_msn",["ViTMSNForImageClassification",qr]],["fastvit",["FastViTForImageClassification",Xr]],["mobilevit",["MobileViTForImageClassification",Jr]],["mobilevitv2",["MobileViTV2ForImageClassification",na]],["beit",["BeitForImageClassification",ca]],["deit",["DeiTForImageClassification",Sa]],["hiera",["HieraForImageClassification",Fa]],["convnext",["ConvNextForImageClassification",ls]],["convnextv2",["ConvNextV2ForImageClassification",cs]],["dinov2",["Dinov2ForImageClassification",ms]],["resnet",["ResNetForImageClassification",za]],["swin",["SwinForImageClassification",La]],["segformer",["SegformerForImageClassification",Fi]],["efficientnet",["EfficientNetForImageClassification",Di]],["mobilenet_v1",["MobileNetV1ForImageClassification",Ui]],["mobilenet_v2",["MobileNetV2ForImageClassification",Xi]],["mobilenet_v3",["MobileNetV3ForImageClassification",Yi]],["mobilenet_v4",["MobileNetV4ForImageClassification",eo]]]),vo=new Map([["detr",["DetrForObjectDetection",ma]],["rt_detr",["RTDetrForObjectDetection",ba]],["table-transformer",["TableTransformerForObjectDetection",Ta]],["yolos",["YolosForObjectDetection",_s]]]),xo=new Map([["owlvit",["OwlViTForObjectDetection",sa]],["owlv2",["Owlv2ForObjectDetection",la]]]),Mo=new Map([["detr",["DetrForSegmentation",fa]],["clipseg",["CLIPSegForImageSegmentation",Pn]]]),To=new Map([["segformer",["SegformerForSemanticSegmentation",Ai]],["sapiens",["SapiensForSemanticSegmentation",Ha]]]),ko=new Map([["detr",["DetrForSegmentation",fa]],["maskformer",["MaskFormerForInstanceSegmentation",es]]]),$o=new Map([["sam",["SamModel",bs]]]),Co=new Map([["wav2vec2",["Wav2Vec2ForCTC",Es]],["wav2vec2-bert",["Wav2Vec2BertForCTC",Ks]],["unispeech",["UniSpeechForCTC",Ns]],["unispeech-sat",["UniSpeechSatForCTC",Gs]],["wavlm",["WavLMForCTC",ri]],["hubert",["HubertForCTC",Js]]]),So=new Map([["wav2vec2",["Wav2Vec2ForSequenceClassification",Fs]],["wav2vec2-bert",["Wav2Vec2BertForSequenceClassification",Qs]],["unispeech",["UniSpeechForSequenceClassification",Vs]],["unispeech-sat",["UniSpeechSatForSequenceClassification",Us]],["wavlm",["WavLMForSequenceClassification",ai]],["hubert",["HubertForSequenceClassification",ei]],["audio-spectrogram-transformer",["ASTForAudioClassification",an]]]),Po=new Map([["wavlm",["WavLMForXVector",si]]]),Eo=new Map([["unispeech-sat",["UniSpeechSatForAudioFrameClassification",Ws]],["wavlm",["WavLMForAudioFrameClassification",ii]],["wav2vec2",["Wav2Vec2ForAudioFrameClassification",As]],["pyannote",["PyAnnoteForAudioFrameClassification",Os]]]),Fo=new Map([["vitmatte",["VitMatteForImageMatting",Qr]]]),Ao=new Map([["swin2sr",["Swin2SRForImageSuperResolution",Na]]]),Io=new Map([["dpt",["DPTForDepthEstimation",qa]],["depth_anything",["DepthAnythingForDepthEstimation",Ua]],["glpn",["GLPNForDepthEstimation",rs]],["sapiens",["SapiensForDepthEstimation",Xa]],["depth_pro",["DepthProForDepthEstimation",Ya]]]),zo=new Map([["sapiens",["SapiensForNormalEstimation",Ka]]]),Oo=new Map([["clip",["CLIPVisionModelWithProjection",bn]],["siglip",["SiglipVisionModel",Tn]]]),Bo=[[ao,y],[so,b],[io,M],[co,y],[po,y],[ho,v],[oo,v],[mo,M],[fo,y],[go,y],[_o,x],[wo,k],[bo,y],[Mo,y],[ko,y],[To,y],[Fo,y],[Ao,y],[Io,y],[zo,y],[vo,y],[xo,y],[$o,T],[Co,y],[So,y],[lo,v],[uo,y],[Po,y],[Eo,y],[Oo,y]];for(const[e,t]of Bo)for(const[n,r]of e.values())C.set(n,t),P.set(r,n),S.set(n,r);const Lo=[["MusicgenForConditionalGeneration",ji,$],["CLIPTextModelWithProjection",wn,y],["SiglipTextModel",Mn,y],["ClapTextModelWithProjection",ki,y],["ClapAudioModelWithProjection",$i,y]];for(const[e,t,n]of Lo)C.set(e,n),P.set(t,e),S.set(e,t);class Do extends ro{static MODEL_CLASS_MAPPINGS=Bo.map((e=>e[0]));static BASE_IF_FAIL=!0}class Ro extends ro{static MODEL_CLASS_MAPPINGS=[co]}class No extends ro{static MODEL_CLASS_MAPPINGS=[po]}class Vo extends ro{static MODEL_CLASS_MAPPINGS=[ho]}class jo extends ro{static MODEL_CLASS_MAPPINGS=[oo]}class qo extends ro{static MODEL_CLASS_MAPPINGS=[lo]}class Go extends ro{static MODEL_CLASS_MAPPINGS=[uo]}class Uo extends ro{static MODEL_CLASS_MAPPINGS=[mo]}class Wo extends ro{static MODEL_CLASS_MAPPINGS=[fo]}class Ho extends ro{static MODEL_CLASS_MAPPINGS=[go]}class Xo extends ro{static MODEL_CLASS_MAPPINGS=[_o]}class Ko extends ro{static MODEL_CLASS_MAPPINGS=[bo]}class Qo extends ro{static MODEL_CLASS_MAPPINGS=[Mo]}class Yo extends ro{static MODEL_CLASS_MAPPINGS=[To]}class Zo extends ro{static MODEL_CLASS_MAPPINGS=[ko]}class Jo extends ro{static MODEL_CLASS_MAPPINGS=[vo]}class el extends ro{static MODEL_CLASS_MAPPINGS=[xo]}class tl extends ro{static MODEL_CLASS_MAPPINGS=[$o]}class nl extends ro{static MODEL_CLASS_MAPPINGS=[Co]}class rl extends ro{static MODEL_CLASS_MAPPINGS=[So]}class al extends ro{static MODEL_CLASS_MAPPINGS=[Po]}class sl extends ro{static MODEL_CLASS_MAPPINGS=[Eo]}class il extends ro{static MODEL_CLASS_MAPPINGS=[yo]}class ol extends ro{static MODEL_CLASS_MAPPINGS=[Fo]}class ll extends ro{static MODEL_CLASS_MAPPINGS=[Ao]}class ul extends ro{static MODEL_CLASS_MAPPINGS=[Io]}class dl extends ro{static MODEL_CLASS_MAPPINGS=[zo]}class cl extends ro{static MODEL_CLASS_MAPPINGS=[Oo]}class pl extends G{constructor({logits:e,past_key_values:t,encoder_outputs:n,decoder_attentions:r=null,cross_attentions:a=null}){super(),this.logits=e,this.past_key_values=t,this.encoder_outputs=n,this.decoder_attentions=r,this.cross_attentions=a}}class hl extends G{constructor({logits:e}){super(),this.logits=e}}class ml extends G{constructor({logits:e,embeddings:t}){super(),this.logits=e,this.embeddings=t}}class fl extends G{constructor({logits:e}){super(),this.logits=e}}class gl extends G{constructor({logits:e}){super(),this.logits=e}}class _l extends G{constructor({start_logits:e,end_logits:t}){super(),this.start_logits=e,this.end_logits=t}}class wl extends G{constructor({logits:e}){super(),this.logits=e}}class yl extends G{constructor({logits:e,past_key_values:t}){super(),this.logits=e,this.past_key_values=t}}class bl extends G{constructor({alphas:e}){super(),this.alphas=e}}class vl extends G{constructor({waveform:e,spectrogram:t}){super(),this.waveform=e,this.spectrogram=t}}},"./src/models/whisper/common_whisper.js":
+  \***********************/(e,t,n)=>{n.r(t),n.d(t,{ASTForAudioClassification:()=>an,ASTModel:()=>rn,ASTPreTrainedModel:()=>nn,AlbertForMaskedLM:()=>ht,AlbertForQuestionAnswering:()=>pt,AlbertForSequenceClassification:()=>ct,AlbertModel:()=>dt,AlbertPreTrainedModel:()=>ut,AutoModel:()=>Go,AutoModelForAudioClassification:()=>ul,AutoModelForAudioFrameClassification:()=>cl,AutoModelForCTC:()=>ll,AutoModelForCausalLM:()=>Yo,AutoModelForDepthEstimation:()=>fl,AutoModelForDocumentQuestionAnswering:()=>pl,AutoModelForImageClassification:()=>tl,AutoModelForImageFeatureExtraction:()=>_l,AutoModelForImageMatting:()=>hl,AutoModelForImageSegmentation:()=>nl,AutoModelForImageToImage:()=>ml,AutoModelForMaskGeneration:()=>ol,AutoModelForMaskedLM:()=>Zo,AutoModelForNormalEstimation:()=>gl,AutoModelForObjectDetection:()=>sl,AutoModelForQuestionAnswering:()=>Jo,AutoModelForSemanticSegmentation:()=>rl,AutoModelForSeq2SeqLM:()=>Ho,AutoModelForSequenceClassification:()=>Uo,AutoModelForSpeechSeq2Seq:()=>Xo,AutoModelForTextToSpectrogram:()=>Ko,AutoModelForTextToWaveform:()=>Qo,AutoModelForTokenClassification:()=>Wo,AutoModelForUniversalSegmentation:()=>al,AutoModelForVision2Seq:()=>el,AutoModelForXVector:()=>dl,AutoModelForZeroShotObjectDetection:()=>il,BartForConditionalGeneration:()=>kt,BartForSequenceClassification:()=>$t,BartModel:()=>Tt,BartPretrainedModel:()=>Mt,BaseModelOutput:()=>U,BeitForImageClassification:()=>_a,BeitModel:()=>ga,BeitPreTrainedModel:()=>fa,BertForMaskedLM:()=>X,BertForQuestionAnswering:()=>Y,BertForSequenceClassification:()=>K,BertForTokenClassification:()=>Q,BertModel:()=>H,BertPreTrainedModel:()=>W,BlenderbotForConditionalGeneration:()=>zt,BlenderbotModel:()=>It,BlenderbotPreTrainedModel:()=>At,BlenderbotSmallForConditionalGeneration:()=>Lt,BlenderbotSmallModel:()=>Bt,BlenderbotSmallPreTrainedModel:()=>Ot,BloomForCausalLM:()=>Ar,BloomModel:()=>Fr,BloomPreTrainedModel:()=>Er,CLIPModel:()=>gn,CLIPPreTrainedModel:()=>fn,CLIPSegForImageSegmentation:()=>Pn,CLIPSegModel:()=>Sn,CLIPSegPreTrainedModel:()=>Cn,CLIPTextModel:()=>_n,CLIPTextModelWithProjection:()=>wn,CLIPVisionModel:()=>yn,CLIPVisionModelWithProjection:()=>bn,CamembertForMaskedLM:()=>be,CamembertForQuestionAnswering:()=>Me,CamembertForSequenceClassification:()=>ve,CamembertForTokenClassification:()=>xe,CamembertModel:()=>ye,CamembertPreTrainedModel:()=>we,CausalLMOutput:()=>Tl,CausalLMOutputWithPast:()=>kl,ChineseCLIPModel:()=>$n,ChineseCLIPPreTrainedModel:()=>kn,ClapAudioModelWithProjection:()=>Ai,ClapModel:()=>Ei,ClapPreTrainedModel:()=>Pi,ClapTextModelWithProjection:()=>Fi,CodeGenForCausalLM:()=>Qn,CodeGenModel:()=>Kn,CodeGenPreTrainedModel:()=>Xn,CohereForCausalLM:()=>cr,CohereModel:()=>dr,CoherePreTrainedModel:()=>ur,ConvBertForMaskedLM:()=>le,ConvBertForQuestionAnswering:()=>ce,ConvBertForSequenceClassification:()=>ue,ConvBertForTokenClassification:()=>de,ConvBertModel:()=>oe,ConvBertPreTrainedModel:()=>ie,ConvNextForImageClassification:()=>ms,ConvNextModel:()=>hs,ConvNextPreTrainedModel:()=>ps,ConvNextV2ForImageClassification:()=>_s,ConvNextV2Model:()=>gs,ConvNextV2PreTrainedModel:()=>fs,DPTForDepthEstimation:()=>Ka,DPTModel:()=>Xa,DPTPreTrainedModel:()=>Ha,DebertaForMaskedLM:()=>$e,DebertaForQuestionAnswering:()=>Pe,DebertaForSequenceClassification:()=>Ce,DebertaForTokenClassification:()=>Se,DebertaModel:()=>ke,DebertaPreTrainedModel:()=>Te,DebertaV2ForMaskedLM:()=>Ae,DebertaV2ForQuestionAnswering:()=>Oe,DebertaV2ForSequenceClassification:()=>Ie,DebertaV2ForTokenClassification:()=>ze,DebertaV2Model:()=>Fe,DebertaV2PreTrainedModel:()=>Ee,DecisionTransformerModel:()=>lo,DecisionTransformerPreTrainedModel:()=>oo,DeiTForImageClassification:()=>za,DeiTModel:()=>Ia,DeiTPreTrainedModel:()=>Aa,DepthAnythingForDepthEstimation:()=>Ya,DepthAnythingPreTrainedModel:()=>Qa,DepthProForDepthEstimation:()=>rs,DepthProPreTrainedModel:()=>ns,DetrForObjectDetection:()=>ba,DetrForSegmentation:()=>va,DetrModel:()=>ya,DetrObjectDetectionOutput:()=>xa,DetrPreTrainedModel:()=>wa,DetrSegmentationOutput:()=>Ma,Dinov2ForImageClassification:()=>bs,Dinov2Model:()=>ys,Dinov2PreTrainedModel:()=>ws,DistilBertForMaskedLM:()=>Ve,DistilBertForQuestionAnswering:()=>Ne,DistilBertForSequenceClassification:()=>De,DistilBertForTokenClassification:()=>Re,DistilBertModel:()=>Le,DistilBertPreTrainedModel:()=>Be,DonutSwinModel:()=>cs,DonutSwinPreTrainedModel:()=>ds,EfficientNetForImageClassification:()=>Gi,EfficientNetModel:()=>qi,EfficientNetPreTrainedModel:()=>ji,ElectraForMaskedLM:()=>me,ElectraForQuestionAnswering:()=>_e,ElectraForSequenceClassification:()=>fe,ElectraForTokenClassification:()=>ge,ElectraModel:()=>he,ElectraPreTrainedModel:()=>pe,EsmForMaskedLM:()=>Ge,EsmForSequenceClassification:()=>Ue,EsmForTokenClassification:()=>We,EsmModel:()=>qe,EsmPreTrainedModel:()=>je,FalconForCausalLM:()=>Si,FalconModel:()=>Ci,FalconPreTrainedModel:()=>$i,FastViTForImageClassification:()=>ea,FastViTModel:()=>Jr,FastViTPreTrainedModel:()=>Zr,Florence2ForConditionalGeneration:()=>mn,Florence2PreTrainedModel:()=>hn,GLPNForDepthEstimation:()=>us,GLPNModel:()=>ls,GLPNPreTrainedModel:()=>os,GPT2LMHeadModel:()=>An,GPT2Model:()=>Fn,GPT2PreTrainedModel:()=>En,GPTBigCodeForCausalLM:()=>Hn,GPTBigCodeModel:()=>Wn,GPTBigCodePreTrainedModel:()=>Un,GPTJForCausalLM:()=>Gn,GPTJModel:()=>qn,GPTJPreTrainedModel:()=>jn,GPTNeoForCausalLM:()=>Dn,GPTNeoModel:()=>Ln,GPTNeoPreTrainedModel:()=>Bn,GPTNeoXForCausalLM:()=>Vn,GPTNeoXModel:()=>Nn,GPTNeoXPreTrainedModel:()=>Rn,Gemma2ForCausalLM:()=>_r,Gemma2Model:()=>gr,Gemma2PreTrainedModel:()=>fr,GemmaForCausalLM:()=>mr,GemmaModel:()=>hr,GemmaPreTrainedModel:()=>pr,GraniteForCausalLM:()=>lr,GraniteModel:()=>or,GranitePreTrainedModel:()=>ir,GroupViTModel:()=>Yr,GroupViTPreTrainedModel:()=>Qr,HieraForImageClassification:()=>La,HieraModel:()=>Ba,HieraPreTrainedModel:()=>Oa,HubertForCTC:()=>si,HubertForSequenceClassification:()=>ii,HubertModel:()=>ai,HubertPreTrainedModel:()=>ri,ImageMattingOutput:()=>$l,JAISLMHeadModel:()=>On,JAISModel:()=>zn,JAISPreTrainedModel:()=>In,LlamaForCausalLM:()=>Jn,LlamaModel:()=>Zn,LlamaPreTrainedModel:()=>Yn,LlavaForConditionalGeneration:()=>cn,LlavaPreTrainedModel:()=>dn,LongT5ForConditionalGeneration:()=>yt,LongT5Model:()=>wt,LongT5PreTrainedModel:()=>_t,M2M100ForConditionalGeneration:()=>Is,M2M100Model:()=>As,M2M100PreTrainedModel:()=>Fs,MBartForCausalLM:()=>Ft,MBartForConditionalGeneration:()=>Pt,MBartForSequenceClassification:()=>Et,MBartModel:()=>St,MBartPreTrainedModel:()=>Ct,MPNetForMaskedLM:()=>et,MPNetForQuestionAnswering:()=>rt,MPNetForSequenceClassification:()=>tt,MPNetForTokenClassification:()=>nt,MPNetModel:()=>Je,MPNetPreTrainedModel:()=>Ze,MT5ForConditionalGeneration:()=>xt,MT5Model:()=>vt,MT5PreTrainedModel:()=>bt,MarianMTModel:()=>Es,MarianModel:()=>Ps,MarianPreTrainedModel:()=>Ss,MaskFormerForInstanceSegmentation:()=>is,MaskFormerModel:()=>ss,MaskFormerPreTrainedModel:()=>as,MaskedLMOutput:()=>xl,MistralForCausalLM:()=>xi,MistralModel:()=>vi,MistralPreTrainedModel:()=>bi,MobileBertForMaskedLM:()=>Ke,MobileBertForQuestionAnswering:()=>Ye,MobileBertForSequenceClassification:()=>Qe,MobileBertModel:()=>Xe,MobileBertPreTrainedModel:()=>He,MobileLLMForCausalLM:()=>nr,MobileLLMModel:()=>tr,MobileLLMPreTrainedModel:()=>er,MobileNetV1ForImageClassification:()=>Yi,MobileNetV1Model:()=>Qi,MobileNetV1PreTrainedModel:()=>Ki,MobileNetV2ForImageClassification:()=>eo,MobileNetV2Model:()=>Ji,MobileNetV2PreTrainedModel:()=>Zi,MobileNetV3ForImageClassification:()=>ro,MobileNetV3Model:()=>no,MobileNetV3PreTrainedModel:()=>to,MobileNetV4ForImageClassification:()=>io,MobileNetV4Model:()=>so,MobileNetV4PreTrainedModel:()=>ao,MobileViTForImageClassification:()=>sa,MobileViTModel:()=>aa,MobileViTPreTrainedModel:()=>ra,MobileViTV2ForImageClassification:()=>la,MobileViTV2Model:()=>oa,MobileViTV2PreTrainedModel:()=>ia,ModelOutput:()=>G,Moondream1ForConditionalGeneration:()=>pn,MptForCausalLM:()=>Or,MptModel:()=>zr,MptPreTrainedModel:()=>Ir,MusicgenForCausalLM:()=>Hi,MusicgenForConditionalGeneration:()=>Xi,MusicgenModel:()=>Wi,MusicgenPreTrainedModel:()=>Ui,NomicBertModel:()=>J,NomicBertPreTrainedModel:()=>Z,OPTForCausalLM:()=>Dr,OPTModel:()=>Lr,OPTPreTrainedModel:()=>Br,OlmoForCausalLM:()=>sr,OlmoModel:()=>ar,OlmoPreTrainedModel:()=>rr,OpenELMForCausalLM:()=>br,OpenELMModel:()=>yr,OpenELMPreTrainedModel:()=>wr,OwlViTForObjectDetection:()=>ca,OwlViTModel:()=>da,OwlViTPreTrainedModel:()=>ua,Owlv2ForObjectDetection:()=>ma,Owlv2Model:()=>ha,Owlv2PreTrainedModel:()=>pa,Phi3ForCausalLM:()=>Pr,Phi3Model:()=>Sr,Phi3PreTrainedModel:()=>Cr,PhiForCausalLM:()=>$r,PhiModel:()=>kr,PhiPreTrainedModel:()=>Tr,PreTrainedModel:()=>q,PretrainedMixin:()=>uo,PvtForImageClassification:()=>Gr,PvtModel:()=>qr,PvtPreTrainedModel:()=>jr,PyAnnoteForAudioFrameClassification:()=>Vs,PyAnnoteModel:()=>Ns,PyAnnotePreTrainedModel:()=>Rs,QuestionAnsweringModelOutput:()=>Ml,Qwen2ForCausalLM:()=>Mr,Qwen2Model:()=>xr,Qwen2PreTrainedModel:()=>vr,RTDetrForObjectDetection:()=>$a,RTDetrModel:()=>ka,RTDetrObjectDetectionOutput:()=>Ca,RTDetrPreTrainedModel:()=>Ta,ResNetForImageClassification:()=>Na,ResNetModel:()=>Ra,ResNetPreTrainedModel:()=>Da,RoFormerForMaskedLM:()=>ne,RoFormerForQuestionAnswering:()=>se,RoFormerForSequenceClassification:()=>re,RoFormerForTokenClassification:()=>ae,RoFormerModel:()=>te,RoFormerPreTrainedModel:()=>ee,RobertaForMaskedLM:()=>Nt,RobertaForQuestionAnswering:()=>qt,RobertaForSequenceClassification:()=>Vt,RobertaForTokenClassification:()=>jt,RobertaModel:()=>Rt,RobertaPreTrainedModel:()=>Dt,SamImageSegmentationOutput:()=>Cs,SamModel:()=>$s,SamPreTrainedModel:()=>ks,SapiensForDepthEstimation:()=>es,SapiensForNormalEstimation:()=>ts,SapiensForSemanticSegmentation:()=>Ja,SapiensPreTrainedModel:()=>Za,SegformerForImageClassification:()=>Li,SegformerForSemanticSegmentation:()=>Di,SegformerModel:()=>Bi,SegformerPreTrainedModel:()=>Oi,Seq2SeqLMOutput:()=>wl,SequenceClassifierOutput:()=>yl,SiglipModel:()=>xn,SiglipPreTrainedModel:()=>vn,SiglipTextModel:()=>Mn,SiglipVisionModel:()=>Tn,SpeechT5ForSpeechToText:()=>fi,SpeechT5ForTextToSpeech:()=>gi,SpeechT5HifiGan:()=>_i,SpeechT5Model:()=>mi,SpeechT5PreTrainedModel:()=>hi,SqueezeBertForMaskedLM:()=>it,SqueezeBertForQuestionAnswering:()=>lt,SqueezeBertForSequenceClassification:()=>ot,SqueezeBertModel:()=>st,SqueezeBertPreTrainedModel:()=>at,StableLmForCausalLM:()=>Vi,StableLmModel:()=>Ni,StableLmPreTrainedModel:()=>Ri,Starcoder2ForCausalLM:()=>ki,Starcoder2Model:()=>Ti,Starcoder2PreTrainedModel:()=>Mi,Swin2SRForImageSuperResolution:()=>Wa,Swin2SRModel:()=>Ua,Swin2SRPreTrainedModel:()=>Ga,SwinForImageClassification:()=>qa,SwinModel:()=>ja,SwinPreTrainedModel:()=>Va,T5ForConditionalGeneration:()=>gt,T5Model:()=>ft,T5PreTrainedModel:()=>mt,TableTransformerForObjectDetection:()=>Ea,TableTransformerModel:()=>Pa,TableTransformerObjectDetectionOutput:()=>Fa,TableTransformerPreTrainedModel:()=>Sa,TokenClassifierOutput:()=>vl,TrOCRForCausalLM:()=>yi,TrOCRPreTrainedModel:()=>wi,UniSpeechForCTC:()=>Ws,UniSpeechForSequenceClassification:()=>Hs,UniSpeechModel:()=>Us,UniSpeechPreTrainedModel:()=>Gs,UniSpeechSatForAudioFrameClassification:()=>Zs,UniSpeechSatForCTC:()=>Qs,UniSpeechSatForSequenceClassification:()=>Ys,UniSpeechSatModel:()=>Ks,UniSpeechSatPreTrainedModel:()=>Xs,ViTForImageClassification:()=>Vr,ViTMAEModel:()=>Wr,ViTMAEPreTrainedModel:()=>Ur,ViTMSNForImageClassification:()=>Kr,ViTMSNModel:()=>Xr,ViTMSNPreTrainedModel:()=>Hr,ViTModel:()=>Nr,ViTPreTrainedModel:()=>Rr,VisionEncoderDecoderModel:()=>un,VitMatteForImageMatting:()=>na,VitMattePreTrainedModel:()=>ta,VitsModel:()=>zi,VitsModelOutput:()=>Cl,VitsPreTrainedModel:()=>Ii,Wav2Vec2BertForCTC:()=>ti,Wav2Vec2BertForSequenceClassification:()=>ni,Wav2Vec2BertModel:()=>ei,Wav2Vec2BertPreTrainedModel:()=>Js,Wav2Vec2ForAudioFrameClassification:()=>Ds,Wav2Vec2ForCTC:()=>Bs,Wav2Vec2ForSequenceClassification:()=>Ls,Wav2Vec2Model:()=>Os,Wav2Vec2PreTrainedModel:()=>zs,WavLMForAudioFrameClassification:()=>pi,WavLMForCTC:()=>ui,WavLMForSequenceClassification:()=>di,WavLMForXVector:()=>ci,WavLMModel:()=>li,WavLMPreTrainedModel:()=>oi,WeSpeakerResNetModel:()=>qs,WeSpeakerResNetPreTrainedModel:()=>js,WhisperForConditionalGeneration:()=>ln,WhisperModel:()=>on,WhisperPreTrainedModel:()=>sn,XLMForQuestionAnswering:()=>Kt,XLMForSequenceClassification:()=>Ht,XLMForTokenClassification:()=>Xt,XLMModel:()=>Ut,XLMPreTrainedModel:()=>Gt,XLMRobertaForMaskedLM:()=>Zt,XLMRobertaForQuestionAnswering:()=>tn,XLMRobertaForSequenceClassification:()=>Jt,XLMRobertaForTokenClassification:()=>en,XLMRobertaModel:()=>Yt,XLMRobertaPreTrainedModel:()=>Qt,XLMWithLMHeadModel:()=>Wt,XVectorOutput:()=>bl,YolosForObjectDetection:()=>Ms,YolosModel:()=>xs,YolosObjectDetectionOutput:()=>Ts,YolosPreTrainedModel:()=>vs});var r=n(/*! ./configs.js */"./src/configs.js"),a=n(/*! ./backends/onnx.js */"./src/backends/onnx.js"),s=n(/*! ./utils/dtypes.js */"./src/utils/dtypes.js"),i=n(/*! ./utils/generic.js */"./src/utils/generic.js"),o=n(/*! ./utils/core.js */"./src/utils/core.js"),l=n(/*! ./utils/hub.js */"./src/utils/hub.js"),u=n(/*! ./utils/constants.js */"./src/utils/constants.js"),d=n(/*! ./generation/logits_process.js */"./src/generation/logits_process.js"),c=n(/*! ./generation/configuration_utils.js */"./src/generation/configuration_utils.js"),p=n(/*! ./utils/tensor.js */"./src/utils/tensor.js"),h=n(/*! ./utils/maths.js */"./src/utils/maths.js"),m=n(/*! ./generation/stopping_criteria.js */"./src/generation/stopping_criteria.js"),f=n(/*! ./generation/logits_sampler.js */"./src/generation/logits_sampler.js"),g=n(/*! ./env.js */"./src/env.js"),_=n(/*! ./models/whisper/generation_whisper.js */"./src/models/whisper/generation_whisper.js"),w=n(/*! ./models/whisper/common_whisper.js */"./src/models/whisper/common_whisper.js");const y=0,b=1,v=2,x=3,M=4,T=5,k=6,$=7,C=new Map,S=new Map,P=new Map;async function E(e,t,n){return Object.fromEntries(await Promise.all(Object.keys(t).map((async i=>{const{buffer:o,session_options:u,session_config:d}=await async function(e,t,n){const i=n.config?.["transformers.js_config"]??{};let o=n.device??i.device;o&&"string"!=typeof o&&(o.hasOwnProperty(t)?o=o[t]:(console.warn(`device not specified for "${t}". Using the default device.`),o=null));const u=o??(g.apis.IS_NODE_ENV?"cpu":"wasm"),d=(0,a.deviceToExecutionProviders)(u);let c=n.dtype??i.dtype;"string"!=typeof c&&(c&&c.hasOwnProperty(t)?c=c[t]:(c=s.DEFAULT_DEVICE_DTYPE_MAPPING[u]??s.DATA_TYPES.fp32,console.warn(`dtype not specified for "${t}". Using the default dtype (${c}) for this device (${u}).`)));const p=c;if(!s.DEFAULT_DTYPE_SUFFIX_MAPPING.hasOwnProperty(p))throw new Error(`Invalid dtype: ${p}. Should be one of: ${Object.keys(s.DATA_TYPES).join(", ")}`);if(p===s.DATA_TYPES.fp16&&"webgpu"===u&&!await(0,s.isWebGpuFp16Supported)())throw new Error(`The device (${u}) does not support fp16.`);const h=i.kv_cache_dtype?"string"==typeof i.kv_cache_dtype?i.kv_cache_dtype:i.kv_cache_dtype[p]??"float32":void 0;if(h&&!["float32","float16"].includes(h))throw new Error(`Invalid kv_cache_dtype: ${h}. Should be one of: float32, float16`);const m={dtype:p,kv_cache_dtype:h},f=s.DEFAULT_DTYPE_SUFFIX_MAPPING[p],_=`${n.subfolder??""}/${t}${f}.onnx`,w={...n.session_options};w.executionProviders??=d;const y=i.free_dimension_overrides;y?w.freeDimensionOverrides??=y:u.startsWith("webnn")&&!w.freeDimensionOverrides&&console.warn('WebNN does not currently support dynamic shapes and requires `free_dimension_overrides` to be set in config.json as a field within "transformers.js_config". When `free_dimension_overrides` is not set, you may experience significant performance degradation.');const b=(0,l.getModelFile)(e,_,!0,n),v=n.use_external_data_format??i.use_external_data_format;let x=[];if(v&&(!0===v||"object"==typeof v&&v.hasOwnProperty(t)&&!0===v[t])){if(g.apis.IS_NODE_ENV)throw new Error("External data format is not yet supported in Node.js");const r=`${t}${f}.onnx_data`,a=`${n.subfolder??""}/${r}`;x.push(new Promise((async(t,s)=>{const i=await(0,l.getModelFile)(e,a,!0,n);t({path:r,data:i})})))}else void 0!==w.externalData&&(x=w.externalData.map((async t=>{if("string"==typeof t.data){const r=await(0,l.getModelFile)(e,t.data,!0,n);return{...t,data:r}}return t})));if(x.length>0&&(w.externalData=await Promise.all(x)),"webgpu"===u){const e=(0,r.getKeyValueShapes)(n.config,{prefix:"present"});if(Object.keys(e).length>0&&!(0,a.isONNXProxy)()){const t={};for(const n in e)t[n]="gpu-buffer";w.preferredOutputLocation=t}}return{buffer:await b,session_options:w,session_config:m}}(e,t[i],n);return[i,await(0,a.createInferenceSession)(o,u,d)]}))))}async function F(e,t,n){return Object.fromEntries(await Promise.all(Object.keys(t).map((async r=>[r,await(0,l.getModelJSON)(e,t[r],!1,n)]))))}async function A(e,t){const n=function(e,t){const n=Object.create(null),r=[];for(const s of e.inputNames){const e=t[s];e instanceof p.Tensor?n[s]=(0,a.isONNXProxy)()?e.clone():e:r.push(s)}if(r.length>0)throw new Error(`An error occurred during model execution: "Missing the following inputs: ${r.join(", ")}.`);const s=Object.keys(t).length,i=e.inputNames.length;if(s>i){let n=Object.keys(t).filter((t=>!e.inputNames.includes(t)));console.warn(`WARNING: Too many inputs were provided (${s} > ${i}). The following inputs will be ignored: "${n.join(", ")}".`)}return n}(e,t);try{const t=Object.fromEntries(Object.entries(n).map((([e,t])=>[e,t.ort_tensor])));let r=await e.run(t);return r=I(r),r}catch(e){throw console.error(`An error occurred during model execution: "${e}".`),console.error("Inputs given to model:",n),e}}function I(e){for(let t in e)(0,a.isONNXTensor)(e[t])?e[t]=new p.Tensor(e[t]):"object"==typeof e[t]&&I(e[t]);return e}function z(e){if(e instanceof p.Tensor)return e;if(0===e.length)throw Error("items must be non-empty");if(Array.isArray(e[0])){if(e.some((t=>t.length!==e[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 p.Tensor("int64",BigInt64Array.from(e.flat().map((e=>BigInt(e)))),[e.length,e[0].length])}return new p.Tensor("int64",BigInt64Array.from(e.map((e=>BigInt(e)))),[1,e.length])}function O(e){return new p.Tensor("bool",[e],[1])}async function B(e,t){let{encoder_outputs:n,input_ids:r,decoder_input_ids:a,...s}=t;if(!n){const r=(0,o.pick)(t,e.sessions.model.inputNames);n=(await L(e,r)).last_hidden_state}s.input_ids=a,s.encoder_hidden_states=n,e.sessions.decoder_model_merged.inputNames.includes("encoder_attention_mask")&&(s.encoder_attention_mask=t.attention_mask);return await D(e,s,!0)}async function L(e,t){const n=e.sessions.model,r=(0,o.pick)(t,n.inputNames);if(n.inputNames.includes("inputs_embeds")&&!r.inputs_embeds){if(!t.input_ids)throw new Error("Both `input_ids` and `inputs_embeds` are missing in the model inputs.");r.inputs_embeds=await e.encode_text({input_ids:t.input_ids})}return n.inputNames.includes("token_type_ids")&&!r.token_type_ids&&(r.token_type_ids=new p.Tensor("int64",new BigInt64Array(r.input_ids.data.length),r.input_ids.dims)),await A(n,r)}async function D(e,t,n=!1){const r=e.sessions[n?"decoder_model_merged":"model"],{past_key_values:a,...s}=t;r.inputNames.includes("use_cache_branch")&&(s.use_cache_branch=O(!!a)),r.inputNames.includes("position_ids")&&s.attention_mask&&!s.position_ids&&(s.position_ids=function(e,t=null){const{input_ids:n,inputs_embeds:r,attention_mask:a}=e,[s,i]=a.dims,o=new BigInt64Array(a.data.length);for(let e=0;e<s;++e){const t=e*i;let n=BigInt(0);for(let e=0;e<i;++e){const r=t+e;0n===a.data[r]?o[r]=BigInt(1):(o[r]=n,n+=a.data[r])}}let l=new p.Tensor("int64",o,a.dims);if(t){const e=-(n??r).dims.at(1);l=l.slice(null,[e,null])}return l}(s,a)),e.addPastKeyValues(s,a);const i=(0,o.pick)(s,r.inputNames);return await A(r,i)}async function R(e,{input_ids:t=null,attention_mask:n=null,pixel_values:r=null,position_ids:a=null,inputs_embeds:s=null,past_key_values:i=null,generation_config:o=null,logits_processor:l=null,...u}){if(!s)if(s=await e.encode_text({input_ids:t}),r&&1!==t.dims[1]){const a=await e.encode_image({pixel_values:r});({inputs_embeds:s,attention_mask:n}=e._merge_input_ids_with_image_features({image_features:a,inputs_embeds:s,input_ids:t,attention_mask:n}))}else if(i&&r&&1===t.dims[1]){const e=t.dims[1],r=Object.values(i)[0].dims.at(-2);n=(0,p.cat)([(0,p.ones)([t.dims[0],r]),n.slice(null,[n.dims[1]-e,n.dims[1]])],1)}return await D(e,{inputs_embeds:s,past_key_values:i,attention_mask:n,position_ids:a,generation_config:o,logits_processor:l},!0)}function N(e,t,n,r){if(n.past_key_values){const t=Object.values(n.past_key_values)[0].dims.at(-2),{input_ids:r,attention_mask:a}=n;if(a&&a.dims[1]>r.dims[1]);else if(t<r.dims[1])n.input_ids=r.slice(null,[t,null]);else if(null!=e.config.image_token_index&&r.data.some((t=>t==e.config.image_token_index))){const a=e.config.num_image_tokens;if(!a)throw new Error("`num_image_tokens` is missing in the model configuration.");const s=r.dims[1]-(t-a);n.input_ids=r.slice(null,[-s,null]),n.attention_mask=(0,p.ones)([1,t+s])}}return n}function V(e,t,n,r){return n.past_key_values&&(t=t.map((e=>[e.at(-1)]))),{...n,decoder_input_ids:z(t)}}function j(e,...t){return e.config.is_encoder_decoder?V(e,...t):N(e,...t)}class q extends i.Callable{main_input_name="input_ids";forward_params=["input_ids","attention_mask"];constructor(e,t,n){super(),this.config=e,this.sessions=t,this.configs=n;const r=P.get(this.constructor),a=C.get(r);switch(this.can_generate=!1,this._forward=null,this._prepare_inputs_for_generation=null,a){case M:this.can_generate=!0,this._forward=D,this._prepare_inputs_for_generation=N;break;case v:case x:case $:this.can_generate=!0,this._forward=B,this._prepare_inputs_for_generation=V;break;case b:this._forward=B;break;case k:this.can_generate=!0,this._forward=R,this._prepare_inputs_for_generation=j;break;default:this._forward=L}this.can_generate&&this.forward_params.push("past_key_values"),this.custom_config=this.config["transformers.js_config"]??{}}async dispose(){const e=[];for(const t of Object.values(this.sessions))t?.handler?.dispose&&e.push(t.handler.dispose());return await Promise.all(e)}static async from_pretrained(e,{progress_callback:t=null,config:n=null,cache_dir:a=null,local_files_only:s=!1,revision:i="main",model_file_name:o=null,subfolder:l="onnx",device:d=null,dtype:c=null,use_external_data_format:p=null,session_options:h={}}={}){let m={progress_callback:t,config:n,cache_dir:a,local_files_only:s,revision:i,model_file_name:o,subfolder:l,device:d,dtype:c,use_external_data_format:p,session_options:h};const f=P.get(this),g=C.get(f);let _;if(n=m.config=await r.AutoConfig.from_pretrained(e,m),g===M)_=await Promise.all([E(e,{model:m.model_file_name??"model"},m),F(e,{generation_config:"generation_config.json"},m)]);else if(g===v||g===x)_=await Promise.all([E(e,{model:"encoder_model",decoder_model_merged:"decoder_model_merged"},m),F(e,{generation_config:"generation_config.json"},m)]);else if(g===T)_=await Promise.all([E(e,{model:"vision_encoder",prompt_encoder_mask_decoder:"prompt_encoder_mask_decoder"},m)]);else if(g===b)_=await Promise.all([E(e,{model:"encoder_model",decoder_model_merged:"decoder_model_merged"},m)]);else if(g===k){const t={embed_tokens:"embed_tokens",vision_encoder:"vision_encoder",decoder_model_merged:"decoder_model_merged"};n.is_encoder_decoder&&(t.model="encoder_model"),_=await Promise.all([E(e,t,m),F(e,{generation_config:"generation_config.json"},m)])}else g===$?_=await Promise.all([E(e,{model:"text_encoder",decoder_model_merged:"decoder_model_merged",encodec_decode:"encodec_decode"},m),F(e,{generation_config:"generation_config.json"},m)]):(g!==y&&console.warn(`Model type for '${f??n?.model_type}' not found, assuming encoder-only architecture. Please report this at ${u.GITHUB_ISSUE_URL}.`),_=await Promise.all([E(e,{model:m.model_file_name??"model"},m)]));return new this(n,..._)}async _call(e){return await this.forward(e)}async forward(e){return await this._forward(this,e)}get generation_config(){return this.configs?.generation_config??null}_get_logits_warper(e){const t=new d.LogitsProcessorList;return null!==e.temperature&&1!==e.temperature&&t.push(new d.TemperatureLogitsWarper(e.temperature)),null!==e.top_k&&0!==e.top_k&&t.push(new d.TopKLogitsWarper(e.top_k)),null!==e.top_p&&e.top_p<1&&t.push(new d.TopPLogitsWarper(e.top_p)),t}_get_logits_processor(e,t,n=null){const r=new d.LogitsProcessorList;if(null!==e.repetition_penalty&&1!==e.repetition_penalty&&r.push(new d.RepetitionPenaltyLogitsProcessor(e.repetition_penalty)),null!==e.no_repeat_ngram_size&&e.no_repeat_ngram_size>0&&r.push(new d.NoRepeatNGramLogitsProcessor(e.no_repeat_ngram_size)),null!==e.bad_words_ids&&r.push(new d.NoBadWordsLogitsProcessor(e.bad_words_ids,e.eos_token_id)),null!==e.min_length&&null!==e.eos_token_id&&e.min_length>0&&r.push(new d.MinLengthLogitsProcessor(e.min_length,e.eos_token_id)),null!==e.min_new_tokens&&null!==e.eos_token_id&&e.min_new_tokens>0&&r.push(new d.MinNewTokensLengthLogitsProcessor(t,e.min_new_tokens,e.eos_token_id)),null!==e.forced_bos_token_id&&r.push(new d.ForcedBOSTokenLogitsProcessor(e.forced_bos_token_id)),null!==e.forced_eos_token_id&&r.push(new d.ForcedEOSTokenLogitsProcessor(e.max_length,e.forced_eos_token_id)),null!==e.begin_suppress_tokens){const n=t>1||null===e.forced_bos_token_id?t:t+1;r.push(new d.SuppressTokensAtBeginLogitsProcessor(e.begin_suppress_tokens,n))}return null!==e.guidance_scale&&e.guidance_scale>1&&r.push(new d.ClassifierFreeGuidanceLogitsProcessor(e.guidance_scale)),null!==n&&r.extend(n),r}_prepare_generation_config(e,t,n=c.GenerationConfig){const r={...this.config};for(const e of["decoder","generator","text_config"])e in r&&Object.assign(r,r[e]);const a=new n(r);return Object.assign(a,this.generation_config??{}),e&&Object.assign(a,e),t&&Object.assign(a,(0,o.pick)(t,Object.getOwnPropertyNames(a))),a}_get_stopping_criteria(e,t=null){const n=new m.StoppingCriteriaList;return null!==e.max_length&&n.push(new m.MaxLengthCriteria(e.max_length,this.config.max_position_embeddings??null)),null!==e.eos_token_id&&n.push(new m.EosTokenCriteria(e.eos_token_id)),t&&n.extend(t),n}_validate_model_class(){if(!this.can_generate){const e=[bo,Mo,yo,mo],t=P.get(this.constructor),n=new Set,r=this.config.model_type;for(const t of e){const e=t.get(r);e&&n.add(e[0])}let a=`The current model class (${t}) is not compatible with \`.generate()\`, as it doesn't have a language model head.`;throw n.size>0&&(a+=` Please use the following class instead: ${[...n].join(", ")}`),Error(a)}}prepare_inputs_for_generation(...e){return this._prepare_inputs_for_generation(this,...e)}_update_model_kwargs_for_generation({generated_input_ids:e,outputs:t,model_inputs:n,is_encoder_decoder:r}){return n.past_key_values=this.getPastKeyValues(t,n.past_key_values),n.input_ids=new p.Tensor("int64",e.flat(),[e.length,1]),r||(n.attention_mask=(0,p.cat)([n.attention_mask,(0,p.ones)([n.attention_mask.dims[0],1])],1)),n.position_ids=null,n}_prepare_model_inputs({inputs:e,bos_token_id:t,model_kwargs:n}){const r=(0,o.pick)(n,this.forward_params),a=this.main_input_name;if(a in r){if(e)throw new Error("`inputs`: {inputs}` were passed alongside {input_name} which is not allowed. Make sure to either pass {inputs} or {input_name}=...")}else r[a]=e;return{inputs_tensor:r[a],model_inputs:r,model_input_name:a}}async _prepare_encoder_decoder_kwargs_for_generation({inputs_tensor:e,model_inputs:t,model_input_name:n,generation_config:r}){if(this.sessions.model.inputNames.includes("inputs_embeds")&&!t.inputs_embeds&&"_prepare_inputs_embeds"in this){const{input_ids:e,pixel_values:n,attention_mask:r,...a}=t,s=await this._prepare_inputs_embeds(t);t={...a,...(0,o.pick)(s,["inputs_embeds","attention_mask"])}}let{last_hidden_state:a}=await L(this,t);if(null!==r.guidance_scale&&r.guidance_scale>1)a=(0,p.cat)([a,(0,p.full_like)(a,0)],0),"attention_mask"in t&&(t.attention_mask=(0,p.cat)([t.attention_mask,(0,p.zeros_like)(t.attention_mask)],0));else if(t.decoder_input_ids){const e=z(t.decoder_input_ids).dims[0];if(e!==a.dims[0]){if(1!==a.dims[0])throw new Error(`The encoder outputs have a different batch size (${a.dims[0]}) than the decoder inputs (${e}).`);a=(0,p.cat)(Array.from({length:e},(()=>a)),0)}}return t.encoder_outputs=a,t}_prepare_decoder_input_ids_for_generation({batch_size:e,model_input_name:t,model_kwargs:n,decoder_start_token_id:r,bos_token_id:a,generation_config:s}){let{decoder_input_ids:i,...o}=n;if(!(i instanceof p.Tensor)){if(i)Array.isArray(i[0])||(i=Array.from({length:e},(()=>i)));else if(r??=a,"musicgen"===this.config.model_type)i=Array.from({length:e*this.config.decoder.num_codebooks},(()=>[r]));else if(Array.isArray(r)){if(r.length!==e)throw new Error(`\`decoder_start_token_id\` expcted to have length ${e} but got ${r.length}`);i=r}else i=Array.from({length:e},(()=>[r]));i=z(i)}return n.decoder_attention_mask=(0,p.ones_like)(i),{input_ids:i,model_inputs:o}}async generate({inputs:e=null,generation_config:t=null,logits_processor:n=null,stopping_criteria:r=null,streamer:a=null,...s}){this._validate_model_class(),t=this._prepare_generation_config(t,s);let{inputs_tensor:i,model_inputs:o,model_input_name:l}=this._prepare_model_inputs({inputs:e,model_kwargs:s});const u=this.config.is_encoder_decoder;let d;u&&("encoder_outputs"in o||(o=await this._prepare_encoder_decoder_kwargs_for_generation({inputs_tensor:i,model_inputs:o,model_input_name:l,generation_config:t}))),u?({input_ids:d,model_inputs:o}=this._prepare_decoder_input_ids_for_generation({batch_size:o[l].dims.at(0),model_input_name:l,model_kwargs:o,decoder_start_token_id:t.decoder_start_token_id,bos_token_id:t.bos_token_id,generation_config:t})):d=o[l];let c=d.dims.at(-1);null!==t.max_new_tokens&&(t.max_length=c+t.max_new_tokens);const h=this._get_logits_processor(t,c,n),m=this._get_stopping_criteria(t,r),g=o[l].dims.at(0),_=f.LogitsSampler.getSampler(t),w=new Array(g).fill(0),y=d.tolist();let b;a&&a.put(y);let v={};for(;;){if(o=this.prepare_inputs_for_generation(y,o,t),b=await this.forward(o),t.output_attentions&&t.return_dict_in_generate){const e=this.getAttentions(b);for(const t in e)t in v||(v[t]=[]),v[t].push(e[t])}const e=h(y,b.logits.slice(null,-1,null)),n=[];for(let t=0;t<e.dims.at(0);++t){const r=e[t],a=await _(r);for(const[e,r]of a){const a=BigInt(e);w[t]+=r,y[t].push(a),n.push([a]);break}}a&&a.put(n);if(m(y).every((e=>e)))break;o=this._update_model_kwargs_for_generation({generated_input_ids:n,outputs:b,model_inputs:o,is_encoder_decoder:u})}a&&a.end();const x=this.getPastKeyValues(b,o.past_key_values,!0),M=new p.Tensor("int64",y.flat(),[y.length,y[0].length]);if(t.return_dict_in_generate)return{sequences:M,past_key_values:x,...v};for(const e of Object.values(b))"gpu-buffer"===e.location&&e.dispose();return M}getPastKeyValues(e,t,n=!1){const r=Object.create(null);for(const a in e)if(a.startsWith("present")){const s=a.replace("present","past_key_values"),i=a.includes("encoder");if(r[s]=i&&t?t[s]:e[a],t&&(!i||n)){const e=t[s];"gpu-buffer"===e.location&&e.dispose()}}return r}getAttentions(e){const t={};for(const n of["cross_attentions","encoder_attentions","decoder_attentions"])for(const r in e)r.startsWith(n)&&(n in t||(t[n]=[]),t[n].push(e[r]));return t}addPastKeyValues(e,t){if(t)Object.assign(e,t);else{const t=this.sessions.decoder_model_merged??this.sessions.model,n=t?.config?.kv_cache_dtype??"float32",a="float16"===n?new Uint16Array:[],s=(0,r.getKeyValueShapes)(this.config);for(const t in s)e[t]=new p.Tensor(n,a,s[t])}}async encode_image({pixel_values:e}){const t=(await A(this.sessions.vision_encoder,{pixel_values:e})).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 (${t.dims[1]}).`),this.config.num_image_tokens=t.dims[1]),t}async encode_text({input_ids:e}){return(await A(this.sessions.embed_tokens,{input_ids:e})).inputs_embeds}}class G{}class U extends G{constructor({last_hidden_state:e,hidden_states:t=null,attentions:n=null}){super(),this.last_hidden_state=e,this.hidden_states=t,this.attentions=n}}class W extends q{}class H extends W{}class X extends W{async _call(e){return new xl(await super._call(e))}}class K extends W{async _call(e){return new yl(await super._call(e))}}class Q extends W{async _call(e){return new vl(await super._call(e))}}class Y extends W{async _call(e){return new Ml(await super._call(e))}}class Z extends q{}class J extends Z{}class ee extends q{}class te extends ee{}class ne extends ee{async _call(e){return new xl(await super._call(e))}}class re extends ee{async _call(e){return new yl(await super._call(e))}}class ae extends ee{async _call(e){return new vl(await super._call(e))}}class se extends ee{async _call(e){return new Ml(await super._call(e))}}class ie extends q{}class oe extends ie{}class le extends ie{async _call(e){return new xl(await super._call(e))}}class ue extends ie{async _call(e){return new yl(await super._call(e))}}class de extends ie{async _call(e){return new vl(await super._call(e))}}class ce extends ie{async _call(e){return new Ml(await super._call(e))}}class pe extends q{}class he extends pe{}class me extends pe{async _call(e){return new xl(await super._call(e))}}class fe extends pe{async _call(e){return new yl(await super._call(e))}}class ge extends pe{async _call(e){return new vl(await super._call(e))}}class _e extends pe{async _call(e){return new Ml(await super._call(e))}}class we extends q{}class ye extends we{}class be extends we{async _call(e){return new xl(await super._call(e))}}class ve extends we{async _call(e){return new yl(await super._call(e))}}class xe extends we{async _call(e){return new vl(await super._call(e))}}class Me extends we{async _call(e){return new Ml(await super._call(e))}}class Te extends q{}class ke extends Te{}class $e extends Te{async _call(e){return new xl(await super._call(e))}}class Ce extends Te{async _call(e){return new yl(await super._call(e))}}class Se extends Te{async _call(e){return new vl(await super._call(e))}}class Pe extends Te{async _call(e){return new Ml(await super._call(e))}}class Ee extends q{}class Fe extends Ee{}class Ae extends Ee{async _call(e){return new xl(await super._call(e))}}class Ie extends Ee{async _call(e){return new yl(await super._call(e))}}class ze extends Ee{async _call(e){return new vl(await super._call(e))}}class Oe extends Ee{async _call(e){return new Ml(await super._call(e))}}class Be extends q{}class Le extends Be{}class De extends Be{async _call(e){return new yl(await super._call(e))}}class Re extends Be{async _call(e){return new vl(await super._call(e))}}class Ne extends Be{async _call(e){return new Ml(await super._call(e))}}class Ve extends Be{async _call(e){return new xl(await super._call(e))}}class je extends q{}class qe extends je{}class Ge extends je{async _call(e){return new xl(await super._call(e))}}class Ue extends je{async _call(e){return new yl(await super._call(e))}}class We extends je{async _call(e){return new vl(await super._call(e))}}class He extends q{}class Xe extends He{}class Ke extends He{async _call(e){return new xl(await super._call(e))}}class Qe extends He{async _call(e){return new yl(await super._call(e))}}class Ye extends He{async _call(e){return new Ml(await super._call(e))}}class Ze extends q{}class Je extends Ze{}class et extends Ze{async _call(e){return new xl(await super._call(e))}}class tt extends Ze{async _call(e){return new yl(await super._call(e))}}class nt extends Ze{async _call(e){return new vl(await super._call(e))}}class rt extends Ze{async _call(e){return new Ml(await super._call(e))}}class at extends q{}class st extends at{}class it extends at{async _call(e){return new xl(await super._call(e))}}class ot extends at{async _call(e){return new yl(await super._call(e))}}class lt extends at{async _call(e){return new Ml(await super._call(e))}}class ut extends q{}class dt extends ut{}class ct extends ut{async _call(e){return new yl(await super._call(e))}}class pt extends ut{async _call(e){return new Ml(await super._call(e))}}class ht extends ut{async _call(e){return new xl(await super._call(e))}}class mt extends q{forward_params=["input_ids","attention_mask","encoder_outputs","decoder_input_ids","decoder_attention_mask","past_key_values"]}class ft extends mt{}class gt extends mt{}class _t extends q{}class wt extends _t{}class yt extends _t{}class bt extends q{}class vt extends bt{}class xt extends bt{}class Mt extends q{}class Tt extends Mt{}class kt extends Mt{}class $t extends Mt{async _call(e){return new yl(await super._call(e))}}class Ct extends q{}class St extends Ct{}class Pt extends Ct{}class Et extends Ct{async _call(e){return new yl(await super._call(e))}}class Ft extends Ct{}class At extends q{}class It extends At{}class zt extends At{}class Ot extends q{}class Bt extends Ot{}class Lt extends Ot{}class Dt extends q{}class Rt extends Dt{}class Nt extends Dt{async _call(e){return new xl(await super._call(e))}}class Vt extends Dt{async _call(e){return new yl(await super._call(e))}}class jt extends Dt{async _call(e){return new vl(await super._call(e))}}class qt extends Dt{async _call(e){return new Ml(await super._call(e))}}class Gt extends q{}class Ut extends Gt{}class Wt extends Gt{async _call(e){return new xl(await super._call(e))}}class Ht extends Gt{async _call(e){return new yl(await super._call(e))}}class Xt extends Gt{async _call(e){return new vl(await super._call(e))}}class Kt extends Gt{async _call(e){return new Ml(await super._call(e))}}class Qt extends q{}class Yt extends Qt{}class Zt extends Qt{async _call(e){return new xl(await super._call(e))}}class Jt extends Qt{async _call(e){return new yl(await super._call(e))}}class en extends Qt{async _call(e){return new vl(await super._call(e))}}class tn extends Qt{async _call(e){return new Ml(await super._call(e))}}class nn extends q{}class rn extends nn{}class an extends nn{}class sn extends q{requires_attention_mask=!1;main_input_name="input_features";forward_params=["input_features","attention_mask","decoder_input_ids","decoder_attention_mask","past_key_values"]}class on extends sn{}class ln extends sn{_prepare_generation_config(e,t){return super._prepare_generation_config(e,t,_.WhisperGenerationConfig)}_retrieve_init_tokens(e){const t=[e.decoder_start_token_id];let n=e.language;const r=e.task;if(e.is_multilingual){n||(console.warn("No language specified - defaulting to English (en)."),n="en");const a=`<|${(0,w.whisper_language_to_code)(n)}|>`;t.push(e.lang_to_id[a]),t.push(e.task_to_id[r??"transcribe"])}else if(n||r)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!e.return_timestamps&&e.no_timestamps_token_id&&t.at(-1)!==e.no_timestamps_token_id?t.push(e.no_timestamps_token_id):e.return_timestamps&&t.at(-1)===e.no_timestamps_token_id&&(console.warn("<|notimestamps|> prompt token is removed from generation_config since `return_timestamps` is set to `true`."),t.pop()),t.filter((e=>null!=e))}async generate({inputs:e=null,generation_config:t=null,logits_processor:n=null,stopping_criteria:r=null,...a}){t=this._prepare_generation_config(t,a);const s=a.decoder_input_ids??this._retrieve_init_tokens(t);if(t.return_timestamps&&(n??=new d.LogitsProcessorList,n.push(new d.WhisperTimeStampLogitsProcessor(t,s))),t.begin_suppress_tokens&&(n??=new d.LogitsProcessorList,n.push(new d.SuppressTokensAtBeginLogitsProcessor(t.begin_suppress_tokens,s.length))),t.return_token_timestamps){if(!t.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.");"translate"===t.task&&console.warn("Token-level timestamps may not be reliable for task 'translate'."),t.output_attentions=!0,t.return_dict_in_generate=!0}const i=await super.generate({inputs:e,generation_config:t,logits_processor:n,decoder_input_ids:s,...a});return t.return_token_timestamps&&(i.token_timestamps=this._extract_token_timestamps(i,t.alignment_heads,t.num_frames)),i}_extract_token_timestamps(e,t,n=null,r=.02){if(!e.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`.");null==n&&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 a=this.config.median_filter_width;void 0===a&&(console.warn("Model config has no `median_filter_width`, using default value of 7."),a=7);const s=e.cross_attentions,i=Array.from({length:this.config.decoder_layers},((e,t)=>(0,p.cat)(s.map((e=>e[t])),2))),l=(0,p.stack)(t.map((([e,t])=>{if(e>=i.length)throw new Error(`Layer index ${e} is out of bounds for cross attentions (length ${i.length}).`);return n?i[e].slice(null,t,null,[0,n]):i[e].slice(null,t)}))).transpose(1,0,2,3),[u,d]=(0,p.std_mean)(l,-2,0,!0),c=l.clone();for(let e=0;e<c.dims[0];++e){const t=c[e];for(let n=0;n<t.dims[0];++n){const r=t[n],s=u[e][n][0].data,i=d[e][n][0].data;for(let e=0;e<r.dims[0];++e){let t=r[e].data;for(let e=0;e<t.length;++e)t[e]=(t[e]-i[e])/s[e];t.set((0,h.medianFilter)(t,a))}}}const m=[(0,p.mean)(c,1)],f=e.sequences.dims,g=new p.Tensor("float32",new Float32Array(f[0]*f[1]),f);for(let e=0;e<f[0];++e){const t=m[e].neg().squeeze_(0),[n,a]=(0,h.dynamic_time_warping)(t.tolist()),s=Array.from({length:n.length-1},((e,t)=>n[t+1]-n[t])),i=(0,o.mergeArrays)([1],s).map((e=>!!e)),l=[];for(let e=0;e<i.length;++e)i[e]&&l.push(a[e]*r);g[e].data.set(l,1)}return g}}class un extends q{main_input_name="pixel_values";forward_params=["pixel_values","decoder_input_ids","encoder_hidden_states","past_key_values"]}class dn extends q{forward_params=["input_ids","pixel_values","attention_mask","position_ids","past_key_values"]}class cn extends dn{_merge_input_ids_with_image_features({inputs_embeds:e,image_features:t,input_ids:n,attention_mask:r}){const a=this.config.image_token_index,s=n.tolist().map((e=>e.findIndex((e=>e==a)))),i=s.every((e=>-1===e)),o=s.every((e=>-1!==e));if(!i&&!o)throw new Error("Every input should contain either 0 or 1 image token.");if(i)return{inputs_embeds:e,attention_mask:r};const l=[],u=[];for(let n=0;n<s.length;++n){const a=s[n],i=e[n],o=t[n],d=r[n];l.push((0,p.cat)([i.slice([0,a]),o,i.slice([a+1,i.dims[0]])],0)),u.push((0,p.cat)([d.slice([0,a]),(0,p.ones)([o.dims[0]]),d.slice([a+1,d.dims[0]])],0))}return{inputs_embeds:(0,p.stack)(l,0),attention_mask:(0,p.stack)(u,0)}}}class pn extends cn{}class hn extends q{forward_params=["input_ids","inputs_embeds","attention_mask","pixel_values","encoder_outputs","decoder_input_ids","decoder_inputs_embeds","decoder_attention_mask","past_key_values"];main_input_name="inputs_embeds"}class mn extends hn{_merge_input_ids_with_image_features({inputs_embeds:e,image_features:t,input_ids:n,attention_mask:r}){return{inputs_embeds:(0,p.cat)([t,e],1),attention_mask:(0,p.cat)([(0,p.ones)(t.dims.slice(0,2)),r],1)}}async _prepare_inputs_embeds({input_ids:e,pixel_values:t,inputs_embeds:n,attention_mask:r}){if(!e&&!t)throw new Error("Either `input_ids` or `pixel_values` should be provided.");let a,s;return e&&(a=await this.encode_text({input_ids:e})),t&&(s=await this.encode_image({pixel_values:t})),a&&s?({inputs_embeds:n,attention_mask:r}=this._merge_input_ids_with_image_features({inputs_embeds:a,image_features:s,input_ids:e,attention_mask:r})):n=a||s,{inputs_embeds:n,attention_mask:r}}async forward({input_ids:e,pixel_values:t,attention_mask:n,decoder_input_ids:r,decoder_attention_mask:a,encoder_outputs:s,past_key_values:i,inputs_embeds:o,decoder_inputs_embeds:l}){if(o||({inputs_embeds:o,attention_mask:n}=await this._prepare_inputs_embeds({input_ids:e,pixel_values:t,inputs_embeds:o,attention_mask:n})),!s){let{last_hidden_state:e}=await L(this,{inputs_embeds:o,attention_mask:n});s=e}if(!l){if(!r)throw new Error("Either `decoder_input_ids` or `decoder_inputs_embeds` should be provided.");l=await this.encode_text({input_ids:r})}const u={inputs_embeds:l,attention_mask:a,encoder_attention_mask:n,encoder_hidden_states:s,past_key_values:i};return await D(this,u,!0)}}class fn extends q{}class gn extends fn{}class _n extends fn{static async from_pretrained(e,t={}){return t.model_file_name??="text_model",super.from_pretrained(e,t)}}class wn extends fn{static async from_pretrained(e,t={}){return t.model_file_name??="text_model",super.from_pretrained(e,t)}}class yn extends fn{static async from_pretrained(e,t={}){return t.model_file_name??="vision_model",super.from_pretrained(e,t)}}class bn extends fn{static async from_pretrained(e,t={}){return t.model_file_name??="vision_model",super.from_pretrained(e,t)}}class vn extends q{}class xn extends vn{}class Mn extends vn{static async from_pretrained(e,t={}){return t.model_file_name??="text_model",super.from_pretrained(e,t)}}class Tn extends fn{static async from_pretrained(e,t={}){return t.model_file_name??="vision_model",super.from_pretrained(e,t)}}class kn extends q{}class $n extends kn{}class Cn extends q{}class Sn extends Cn{}class Pn extends Cn{}class En extends q{}class Fn extends En{}class An extends En{}class In extends q{}class zn extends In{}class On extends In{}class Bn extends q{}class Ln extends Bn{}class Dn extends Bn{}class Rn extends q{}class Nn extends Rn{}class Vn extends Rn{}class jn extends q{}class qn extends jn{}class Gn extends jn{}class Un extends q{}class Wn extends Un{}class Hn extends Un{}class Xn extends q{}class Kn extends Xn{}class Qn extends Xn{}class Yn extends q{}class Zn extends Yn{}class Jn extends Yn{}class er extends q{}class tr extends er{}class nr extends er{}class rr extends q{}class ar extends rr{}class sr extends rr{}class ir extends q{}class or extends ir{}class lr extends ir{}class ur extends q{}class dr extends ur{}class cr extends ur{}class pr extends q{}class hr extends pr{}class mr extends pr{}class fr extends q{}class gr extends fr{}class _r extends fr{}class wr extends q{}class yr extends wr{}class br extends wr{}class vr extends q{}class xr extends vr{}class Mr extends vr{}class Tr extends q{}class kr extends Tr{}class $r extends Tr{}class Cr extends q{}class Sr extends Cr{}class Pr extends Cr{}class Er extends q{}class Fr extends Er{}class Ar extends Er{}class Ir extends q{}class zr extends Ir{}class Or extends Ir{}class Br extends q{}class Lr extends Br{}class Dr extends Br{}class Rr extends q{}class Nr extends Rr{}class Vr extends Rr{async _call(e){return new yl(await super._call(e))}}class jr extends q{}class qr extends jr{}class Gr extends jr{async _call(e){return new yl(await super._call(e))}}class Ur extends q{}class Wr extends Ur{}class Hr extends q{}class Xr extends Hr{}class Kr extends Hr{async _call(e){return new yl(await super._call(e))}}class Qr extends q{}class Yr extends Qr{}class Zr extends q{}class Jr extends Zr{}class ea extends Zr{async _call(e){return new yl(await super._call(e))}}class ta extends q{}class na extends ta{async _call(e){return new $l(await super._call(e))}}class ra extends q{}class aa extends ra{}class sa extends ra{async _call(e){return new yl(await super._call(e))}}class ia extends q{}class oa extends ia{}class la extends ia{async _call(e){return new yl(await super._call(e))}}class ua extends q{}class da extends ua{}class ca extends ua{}class pa extends q{}class ha extends pa{}class ma extends pa{}class fa extends q{}class ga extends fa{}class _a extends fa{async _call(e){return new yl(await super._call(e))}}class wa extends q{}class ya extends wa{}class ba extends wa{async _call(e){return new xa(await super._call(e))}}class va extends wa{async _call(e){return new Ma(await super._call(e))}}class xa extends G{constructor({logits:e,pred_boxes:t}){super(),this.logits=e,this.pred_boxes=t}}class Ma extends G{constructor({logits:e,pred_boxes:t,pred_masks:n}){super(),this.logits=e,this.pred_boxes=t,this.pred_masks=n}}class Ta extends q{}class ka extends Ta{}class $a extends Ta{async _call(e){return new Ca(await super._call(e))}}class Ca extends G{constructor({logits:e,pred_boxes:t}){super(),this.logits=e,this.pred_boxes=t}}class Sa extends q{}class Pa extends Sa{}class Ea extends Sa{async _call(e){return new Fa(await super._call(e))}}class Fa extends xa{}class Aa extends q{}class Ia extends Aa{}class za extends Aa{async _call(e){return new yl(await super._call(e))}}class Oa extends q{}class Ba extends Oa{}class La extends Oa{async _call(e){return new yl(await super._call(e))}}class Da extends q{}class Ra extends Da{}class Na extends Da{async _call(e){return new yl(await super._call(e))}}class Va extends q{}class ja extends Va{}class qa extends Va{async _call(e){return new yl(await super._call(e))}}class Ga extends q{}class Ua extends Ga{}class Wa extends Ga{}class Ha extends q{}class Xa extends Ha{}class Ka extends Ha{}class Qa extends q{}class Ya extends Qa{}class Za extends q{}class Ja extends Za{}class es extends Za{}class ts extends Za{}class ns extends q{}class rs extends ns{}class as extends q{}class ss extends as{}class is extends as{}class os extends q{}class ls extends os{}class us extends os{}class ds extends q{}class cs extends ds{}class ps extends q{}class hs extends ps{}class ms extends ps{async _call(e){return new yl(await super._call(e))}}class fs extends q{}class gs extends fs{}class _s extends fs{async _call(e){return new yl(await super._call(e))}}class ws extends q{}class ys extends ws{}class bs extends ws{async _call(e){return new yl(await super._call(e))}}class vs extends q{}class xs extends vs{}class Ms extends vs{async _call(e){return new Ts(await super._call(e))}}class Ts extends G{constructor({logits:e,pred_boxes:t}){super(),this.logits=e,this.pred_boxes=t}}class ks extends q{}class $s extends ks{async get_image_embeddings({pixel_values:e}){return await L(this,{pixel_values:e})}async forward(e){if(e.image_embeddings&&e.image_positional_embeddings||(e={...e,...await this.get_image_embeddings(e)}),!e.input_labels&&e.input_points){const t=e.input_points.dims.slice(0,-1),n=t.reduce(((e,t)=>e*t),1);e.input_labels=new p.Tensor("int64",new BigInt64Array(n).fill(1n),t)}const t={image_embeddings:e.image_embeddings,image_positional_embeddings:e.image_positional_embeddings};return e.input_points&&(t.input_points=e.input_points),e.input_labels&&(t.input_labels=e.input_labels),e.input_boxes&&(t.input_boxes=e.input_boxes),await A(this.sessions.prompt_encoder_mask_decoder,t)}async _call(e){return new Cs(await super._call(e))}}class Cs extends G{constructor({iou_scores:e,pred_masks:t}){super(),this.iou_scores=e,this.pred_masks=t}}class Ss extends q{}class Ps extends Ss{}class Es extends Ss{}class Fs extends q{}class As extends Fs{}class Is extends Fs{}class zs extends q{}class Os extends zs{}class Bs extends zs{async _call(e){return new Tl(await super._call(e))}}class Ls extends zs{async _call(e){return new yl(await super._call(e))}}class Ds extends zs{async _call(e){return new vl(await super._call(e))}}class Rs extends q{}class Ns extends Rs{}class Vs extends Rs{async _call(e){return new vl(await super._call(e))}}class js extends q{}class qs extends js{}class Gs extends q{}class Us extends Gs{}class Ws extends Gs{async _call(e){return new Tl(await super._call(e))}}class Hs extends Gs{async _call(e){return new yl(await super._call(e))}}class Xs extends q{}class Ks extends Xs{}class Qs extends Xs{async _call(e){return new Tl(await super._call(e))}}class Ys extends Xs{async _call(e){return new yl(await super._call(e))}}class Zs extends Xs{async _call(e){return new vl(await super._call(e))}}class Js extends q{}class ei extends Js{}class ti extends Js{async _call(e){return new Tl(await super._call(e))}}class ni extends Js{async _call(e){return new yl(await super._call(e))}}class ri extends q{}class ai extends zs{}class si extends zs{async _call(e){return new Tl(await super._call(e))}}class ii extends zs{async _call(e){return new yl(await super._call(e))}}class oi extends q{}class li extends oi{}class ui extends oi{async _call(e){return new Tl(await super._call(e))}}class di extends oi{async _call(e){return new yl(await super._call(e))}}class ci extends oi{async _call(e){return new bl(await super._call(e))}}class pi extends oi{async _call(e){return new vl(await super._call(e))}}class hi extends q{}class mi extends hi{}class fi extends hi{}class gi extends hi{async generate_speech(e,t,{threshold:n=.5,minlenratio:r=0,maxlenratio:a=20,vocoder:s=null}={}){const i={input_ids:e},{encoder_outputs:o,encoder_attention_mask:l}=await L(this,i),u=o.dims[1]/this.config.reduction_factor,d=Math.floor(u*a),c=Math.floor(u*r),h=this.config.num_mel_bins;let m=[],f=null,g=null,_=0;for(;;){++_;const e=O(!!g);let r;r=g?g.output_sequence_out:new p.Tensor("float32",new Float32Array(h),[1,1,h]);let a={use_cache_branch:e,output_sequence:r,encoder_attention_mask:l,speaker_embeddings:t,encoder_hidden_states:o};this.addPastKeyValues(a,f),g=await A(this.sessions.decoder_model_merged,a),f=this.getPastKeyValues(g,f);const{prob:s,spectrum:i}=g;if(m.push(i),_>=c&&(Array.from(s.data).filter((e=>e>=n)).length>0||_>=d))break}const w=(0,p.cat)(m),{waveform:y}=await A(s.sessions.model,{spectrogram:w});return{spectrogram:w,waveform:y}}}class _i extends q{main_input_name="spectrogram"}class wi extends q{}class yi extends wi{}class bi extends q{}class vi extends bi{}class xi extends bi{}class Mi extends q{}class Ti extends Mi{}class ki extends Mi{}class $i extends q{}class Ci extends $i{}class Si extends $i{}class Pi extends q{}class Ei extends Pi{}class Fi extends Pi{static async from_pretrained(e,t={}){return t.model_file_name??="text_model",super.from_pretrained(e,t)}}class Ai extends Pi{static async from_pretrained(e,t={}){return t.model_file_name??="audio_model",super.from_pretrained(e,t)}}class Ii extends q{}class zi extends Ii{async _call(e){return new Cl(await super._call(e))}}class Oi extends q{}class Bi extends Oi{}class Li extends Oi{}class Di extends Oi{}class Ri extends q{}class Ni extends Ri{}class Vi extends Ri{}class ji extends q{}class qi extends ji{}class Gi extends ji{async _call(e){return new yl(await super._call(e))}}class Ui extends q{}class Wi extends Ui{}class Hi extends Ui{}class Xi extends q{forward_params=["input_ids","attention_mask","encoder_outputs","decoder_input_ids","decoder_attention_mask","past_key_values"];_apply_and_filter_by_delay_pattern_mask(e){const[t,n]=e.dims,r=this.config.decoder.num_codebooks,a=n-r;let s=0;for(let t=0;t<e.size;++t){if(e.data[t]===this.config.decoder.pad_token_id)continue;const i=t%n-Math.floor(t/n)%r;i>0&&i<=a&&(e.data[s++]=e.data[t])}const i=Math.floor(t/r),o=s/(i*r);return new p.Tensor(e.type,e.data.slice(0,s),[i,r,o])}prepare_inputs_for_generation(e,t,n){let r=structuredClone(e);for(let e=0;e<r.length;++e)for(let t=0;t<r[e].length;++t)e%this.config.decoder.num_codebooks>=t&&(r[e][t]=BigInt(this.config.decoder.pad_token_id));null!==n.guidance_scale&&n.guidance_scale>1&&(r=r.concat(r));return super.prepare_inputs_for_generation(r,t,n)}async generate(e){const t=await super.generate(e),n=this._apply_and_filter_by_delay_pattern_mask(t).unsqueeze_(0),{audio_values:r}=await A(this.sessions.encodec_decode,{audio_codes:n});return r}}class Ki extends q{}class Qi extends Ki{}class Yi extends Ki{async _call(e){return new yl(await super._call(e))}}class Zi extends q{}class Ji extends Zi{}class eo extends Zi{async _call(e){return new yl(await super._call(e))}}class to extends q{}class no extends to{}class ro extends to{async _call(e){return new yl(await super._call(e))}}class ao extends q{}class so extends ao{}class io extends ao{async _call(e){return new yl(await super._call(e))}}class oo extends q{}class lo extends oo{}class uo{static MODEL_CLASS_MAPPINGS=null;static BASE_IF_FAIL=!1;static async from_pretrained(e,{progress_callback:t=null,config:n=null,cache_dir:a=null,local_files_only:s=!1,revision:i="main",model_file_name:o=null,subfolder:l="onnx",device:u=null,dtype:d=null,use_external_data_format:c=null,session_options:p={}}={}){const h={progress_callback:t,config:n,cache_dir:a,local_files_only:s,revision:i,model_file_name:o,subfolder:l,device:u,dtype:d,use_external_data_format:c,session_options:p};if(h.config=await r.AutoConfig.from_pretrained(e,h),!this.MODEL_CLASS_MAPPINGS)throw new Error("`MODEL_CLASS_MAPPINGS` not implemented for this type of `AutoClass`: "+this.name);for(const t of this.MODEL_CLASS_MAPPINGS){const n=t.get(h.config.model_type);if(n)return await n[1].from_pretrained(e,h)}if(this.BASE_IF_FAIL)return console.warn(`Unknown model class "${h.config.model_type}", attempting to construct from base class.`),await q.from_pretrained(e,h);throw Error(`Unsupported model type: ${h.config.model_type}`)}}const co=new Map([["bert",["BertModel",H]],["nomic_bert",["NomicBertModel",J]],["roformer",["RoFormerModel",te]],["electra",["ElectraModel",he]],["esm",["EsmModel",qe]],["convbert",["ConvBertModel",oe]],["camembert",["CamembertModel",ye]],["deberta",["DebertaModel",ke]],["deberta-v2",["DebertaV2Model",Fe]],["mpnet",["MPNetModel",Je]],["albert",["AlbertModel",dt]],["distilbert",["DistilBertModel",Le]],["roberta",["RobertaModel",Rt]],["xlm",["XLMModel",Ut]],["xlm-roberta",["XLMRobertaModel",Yt]],["clap",["ClapModel",Ei]],["clip",["CLIPModel",gn]],["clipseg",["CLIPSegModel",Sn]],["chinese_clip",["ChineseCLIPModel",$n]],["siglip",["SiglipModel",xn]],["mobilebert",["MobileBertModel",Xe]],["squeezebert",["SqueezeBertModel",st]],["wav2vec2",["Wav2Vec2Model",Os]],["wav2vec2-bert",["Wav2Vec2BertModel",ei]],["unispeech",["UniSpeechModel",Us]],["unispeech-sat",["UniSpeechSatModel",Ks]],["hubert",["HubertModel",ai]],["wavlm",["WavLMModel",li]],["audio-spectrogram-transformer",["ASTModel",rn]],["vits",["VitsModel",zi]],["pyannote",["PyAnnoteModel",Ns]],["wespeaker-resnet",["WeSpeakerResNetModel",qs]],["detr",["DetrModel",ya]],["rt_detr",["RTDetrModel",ka]],["table-transformer",["TableTransformerModel",Pa]],["vit",["ViTModel",Nr]],["pvt",["PvtModel",qr]],["vit_msn",["ViTMSNModel",Xr]],["vit_mae",["ViTMAEModel",Wr]],["groupvit",["GroupViTModel",Yr]],["fastvit",["FastViTModel",Jr]],["mobilevit",["MobileViTModel",aa]],["mobilevitv2",["MobileViTV2Model",oa]],["owlvit",["OwlViTModel",da]],["owlv2",["Owlv2Model",ha]],["beit",["BeitModel",ga]],["deit",["DeiTModel",Ia]],["hiera",["HieraModel",Ba]],["convnext",["ConvNextModel",hs]],["convnextv2",["ConvNextV2Model",gs]],["dinov2",["Dinov2Model",ys]],["resnet",["ResNetModel",Ra]],["swin",["SwinModel",ja]],["swin2sr",["Swin2SRModel",Ua]],["donut-swin",["DonutSwinModel",cs]],["yolos",["YolosModel",xs]],["dpt",["DPTModel",Xa]],["glpn",["GLPNModel",ls]],["hifigan",["SpeechT5HifiGan",_i]],["efficientnet",["EfficientNetModel",qi]],["decision_transformer",["DecisionTransformerModel",lo]],["mobilenet_v1",["MobileNetV1Model",Qi]],["mobilenet_v2",["MobileNetV2Model",Ji]],["mobilenet_v3",["MobileNetV3Model",no]],["mobilenet_v4",["MobileNetV4Model",so]],["maskformer",["MaskFormerModel",ss]]]),po=new Map([["t5",["T5Model",ft]],["longt5",["LongT5Model",wt]],["mt5",["MT5Model",vt]],["bart",["BartModel",Tt]],["mbart",["MBartModel",St]],["marian",["MarianModel",Ps]],["whisper",["WhisperModel",on]],["m2m_100",["M2M100Model",As]],["blenderbot",["BlenderbotModel",It]],["blenderbot-small",["BlenderbotSmallModel",Bt]]]),ho=new Map([["bloom",["BloomModel",Fr]],["jais",["JAISModel",zn]],["gpt2",["GPT2Model",Fn]],["gptj",["GPTJModel",qn]],["gpt_bigcode",["GPTBigCodeModel",Wn]],["gpt_neo",["GPTNeoModel",Ln]],["gpt_neox",["GPTNeoXModel",Nn]],["codegen",["CodeGenModel",Kn]],["llama",["LlamaModel",Zn]],["olmo",["OlmoModel",ar]],["mobilellm",["MobileLLMModel",tr]],["granite",["GraniteModel",or]],["cohere",["CohereModel",dr]],["gemma",["GemmaModel",hr]],["gemma2",["Gemma2Model",gr]],["openelm",["OpenELMModel",yr]],["qwen2",["Qwen2Model",xr]],["phi",["PhiModel",kr]],["phi3",["Phi3Model",Sr]],["mpt",["MptModel",zr]],["opt",["OPTModel",Lr]],["mistral",["MistralModel",vi]],["starcoder2",["Starcoder2Model",Ti]],["falcon",["FalconModel",Ci]],["stablelm",["StableLmModel",Ni]]]),mo=new Map([["speecht5",["SpeechT5ForSpeechToText",fi]],["whisper",["WhisperForConditionalGeneration",ln]]]),fo=new Map([["speecht5",["SpeechT5ForTextToSpeech",gi]]]),go=new Map([["vits",["VitsModel",zi]],["musicgen",["MusicgenForConditionalGeneration",Xi]]]),_o=new Map([["bert",["BertForSequenceClassification",K]],["roformer",["RoFormerForSequenceClassification",re]],["electra",["ElectraForSequenceClassification",fe]],["esm",["EsmForSequenceClassification",Ue]],["convbert",["ConvBertForSequenceClassification",ue]],["camembert",["CamembertForSequenceClassification",ve]],["deberta",["DebertaForSequenceClassification",Ce]],["deberta-v2",["DebertaV2ForSequenceClassification",Ie]],["mpnet",["MPNetForSequenceClassification",tt]],["albert",["AlbertForSequenceClassification",ct]],["distilbert",["DistilBertForSequenceClassification",De]],["roberta",["RobertaForSequenceClassification",Vt]],["xlm",["XLMForSequenceClassification",Ht]],["xlm-roberta",["XLMRobertaForSequenceClassification",Jt]],["bart",["BartForSequenceClassification",$t]],["mbart",["MBartForSequenceClassification",Et]],["mobilebert",["MobileBertForSequenceClassification",Qe]],["squeezebert",["SqueezeBertForSequenceClassification",ot]]]),wo=new Map([["bert",["BertForTokenClassification",Q]],["roformer",["RoFormerForTokenClassification",ae]],["electra",["ElectraForTokenClassification",ge]],["esm",["EsmForTokenClassification",We]],["convbert",["ConvBertForTokenClassification",de]],["camembert",["CamembertForTokenClassification",xe]],["deberta",["DebertaForTokenClassification",Se]],["deberta-v2",["DebertaV2ForTokenClassification",ze]],["mpnet",["MPNetForTokenClassification",nt]],["distilbert",["DistilBertForTokenClassification",Re]],["roberta",["RobertaForTokenClassification",jt]],["xlm",["XLMForTokenClassification",Xt]],["xlm-roberta",["XLMRobertaForTokenClassification",en]]]),yo=new Map([["t5",["T5ForConditionalGeneration",gt]],["longt5",["LongT5ForConditionalGeneration",yt]],["mt5",["MT5ForConditionalGeneration",xt]],["bart",["BartForConditionalGeneration",kt]],["mbart",["MBartForConditionalGeneration",Pt]],["marian",["MarianMTModel",Es]],["m2m_100",["M2M100ForConditionalGeneration",Is]],["blenderbot",["BlenderbotForConditionalGeneration",zt]],["blenderbot-small",["BlenderbotSmallForConditionalGeneration",Lt]]]),bo=new Map([["bloom",["BloomForCausalLM",Ar]],["gpt2",["GPT2LMHeadModel",An]],["jais",["JAISLMHeadModel",On]],["gptj",["GPTJForCausalLM",Gn]],["gpt_bigcode",["GPTBigCodeForCausalLM",Hn]],["gpt_neo",["GPTNeoForCausalLM",Dn]],["gpt_neox",["GPTNeoXForCausalLM",Vn]],["codegen",["CodeGenForCausalLM",Qn]],["llama",["LlamaForCausalLM",Jn]],["olmo",["OlmoForCausalLM",sr]],["mobilellm",["MobileLLMForCausalLM",nr]],["granite",["GraniteForCausalLM",lr]],["cohere",["CohereForCausalLM",cr]],["gemma",["GemmaForCausalLM",mr]],["gemma2",["Gemma2ForCausalLM",_r]],["openelm",["OpenELMForCausalLM",br]],["qwen2",["Qwen2ForCausalLM",Mr]],["phi",["PhiForCausalLM",$r]],["phi3",["Phi3ForCausalLM",Pr]],["mpt",["MptForCausalLM",Or]],["opt",["OPTForCausalLM",Dr]],["mbart",["MBartForCausalLM",Ft]],["mistral",["MistralForCausalLM",xi]],["starcoder2",["Starcoder2ForCausalLM",ki]],["falcon",["FalconForCausalLM",Si]],["trocr",["TrOCRForCausalLM",yi]],["stablelm",["StableLmForCausalLM",Vi]]]),vo=new Map([["bert",["BertForMaskedLM",X]],["roformer",["RoFormerForMaskedLM",ne]],["electra",["ElectraForMaskedLM",me]],["esm",["EsmForMaskedLM",Ge]],["convbert",["ConvBertForMaskedLM",le]],["camembert",["CamembertForMaskedLM",be]],["deberta",["DebertaForMaskedLM",$e]],["deberta-v2",["DebertaV2ForMaskedLM",Ae]],["mpnet",["MPNetForMaskedLM",et]],["albert",["AlbertForMaskedLM",ht]],["distilbert",["DistilBertForMaskedLM",Ve]],["roberta",["RobertaForMaskedLM",Nt]],["xlm",["XLMWithLMHeadModel",Wt]],["xlm-roberta",["XLMRobertaForMaskedLM",Zt]],["mobilebert",["MobileBertForMaskedLM",Ke]],["squeezebert",["SqueezeBertForMaskedLM",it]]]),xo=new Map([["bert",["BertForQuestionAnswering",Y]],["roformer",["RoFormerForQuestionAnswering",se]],["electra",["ElectraForQuestionAnswering",_e]],["convbert",["ConvBertForQuestionAnswering",ce]],["camembert",["CamembertForQuestionAnswering",Me]],["deberta",["DebertaForQuestionAnswering",Pe]],["deberta-v2",["DebertaV2ForQuestionAnswering",Oe]],["mpnet",["MPNetForQuestionAnswering",rt]],["albert",["AlbertForQuestionAnswering",pt]],["distilbert",["DistilBertForQuestionAnswering",Ne]],["roberta",["RobertaForQuestionAnswering",qt]],["xlm",["XLMForQuestionAnswering",Kt]],["xlm-roberta",["XLMRobertaForQuestionAnswering",tn]],["mobilebert",["MobileBertForQuestionAnswering",Ye]],["squeezebert",["SqueezeBertForQuestionAnswering",lt]]]),Mo=new Map([["vision-encoder-decoder",["VisionEncoderDecoderModel",un]]]),To=new Map([["llava",["LlavaForConditionalGeneration",cn]],["moondream1",["Moondream1ForConditionalGeneration",pn]],["florence2",["Florence2ForConditionalGeneration",mn]]]),ko=new Map([["vision-encoder-decoder",["VisionEncoderDecoderModel",un]]]),$o=new Map([["vit",["ViTForImageClassification",Vr]],["pvt",["PvtForImageClassification",Gr]],["vit_msn",["ViTMSNForImageClassification",Kr]],["fastvit",["FastViTForImageClassification",ea]],["mobilevit",["MobileViTForImageClassification",sa]],["mobilevitv2",["MobileViTV2ForImageClassification",la]],["beit",["BeitForImageClassification",_a]],["deit",["DeiTForImageClassification",za]],["hiera",["HieraForImageClassification",La]],["convnext",["ConvNextForImageClassification",ms]],["convnextv2",["ConvNextV2ForImageClassification",_s]],["dinov2",["Dinov2ForImageClassification",bs]],["resnet",["ResNetForImageClassification",Na]],["swin",["SwinForImageClassification",qa]],["segformer",["SegformerForImageClassification",Li]],["efficientnet",["EfficientNetForImageClassification",Gi]],["mobilenet_v1",["MobileNetV1ForImageClassification",Yi]],["mobilenet_v2",["MobileNetV2ForImageClassification",eo]],["mobilenet_v3",["MobileNetV3ForImageClassification",ro]],["mobilenet_v4",["MobileNetV4ForImageClassification",io]]]),Co=new Map([["detr",["DetrForObjectDetection",ba]],["rt_detr",["RTDetrForObjectDetection",$a]],["table-transformer",["TableTransformerForObjectDetection",Ea]],["yolos",["YolosForObjectDetection",Ms]]]),So=new Map([["owlvit",["OwlViTForObjectDetection",ca]],["owlv2",["Owlv2ForObjectDetection",ma]]]),Po=new Map([["detr",["DetrForSegmentation",va]],["clipseg",["CLIPSegForImageSegmentation",Pn]]]),Eo=new Map([["segformer",["SegformerForSemanticSegmentation",Di]],["sapiens",["SapiensForSemanticSegmentation",Ja]]]),Fo=new Map([["detr",["DetrForSegmentation",va]],["maskformer",["MaskFormerForInstanceSegmentation",is]]]),Ao=new Map([["sam",["SamModel",$s]]]),Io=new Map([["wav2vec2",["Wav2Vec2ForCTC",Bs]],["wav2vec2-bert",["Wav2Vec2BertForCTC",ti]],["unispeech",["UniSpeechForCTC",Ws]],["unispeech-sat",["UniSpeechSatForCTC",Qs]],["wavlm",["WavLMForCTC",ui]],["hubert",["HubertForCTC",si]]]),zo=new Map([["wav2vec2",["Wav2Vec2ForSequenceClassification",Ls]],["wav2vec2-bert",["Wav2Vec2BertForSequenceClassification",ni]],["unispeech",["UniSpeechForSequenceClassification",Hs]],["unispeech-sat",["UniSpeechSatForSequenceClassification",Ys]],["wavlm",["WavLMForSequenceClassification",di]],["hubert",["HubertForSequenceClassification",ii]],["audio-spectrogram-transformer",["ASTForAudioClassification",an]]]),Oo=new Map([["wavlm",["WavLMForXVector",ci]]]),Bo=new Map([["unispeech-sat",["UniSpeechSatForAudioFrameClassification",Zs]],["wavlm",["WavLMForAudioFrameClassification",pi]],["wav2vec2",["Wav2Vec2ForAudioFrameClassification",Ds]],["pyannote",["PyAnnoteForAudioFrameClassification",Vs]]]),Lo=new Map([["vitmatte",["VitMatteForImageMatting",na]]]),Do=new Map([["swin2sr",["Swin2SRForImageSuperResolution",Wa]]]),Ro=new Map([["dpt",["DPTForDepthEstimation",Ka]],["depth_anything",["DepthAnythingForDepthEstimation",Ya]],["glpn",["GLPNForDepthEstimation",us]],["sapiens",["SapiensForDepthEstimation",es]],["depth_pro",["DepthProForDepthEstimation",rs]]]),No=new Map([["sapiens",["SapiensForNormalEstimation",ts]]]),Vo=new Map([["clip",["CLIPVisionModelWithProjection",bn]],["siglip",["SiglipVisionModel",Tn]]]),jo=[[co,y],[po,b],[ho,M],[_o,y],[wo,y],[yo,v],[mo,v],[bo,M],[vo,y],[xo,y],[Mo,x],[To,k],[$o,y],[Po,y],[Fo,y],[Eo,y],[Lo,y],[Do,y],[Ro,y],[No,y],[Co,y],[So,y],[Ao,T],[Io,y],[zo,y],[fo,v],[go,y],[Oo,y],[Bo,y],[Vo,y]];for(const[e,t]of jo)for(const[n,r]of e.values())C.set(n,t),P.set(r,n),S.set(n,r);const qo=[["MusicgenForConditionalGeneration",Xi,$],["CLIPTextModelWithProjection",wn,y],["SiglipTextModel",Mn,y],["ClapTextModelWithProjection",Fi,y],["ClapAudioModelWithProjection",Ai,y]];for(const[e,t,n]of qo)C.set(e,n),P.set(t,e),S.set(e,t);class Go extends uo{static MODEL_CLASS_MAPPINGS=jo.map((e=>e[0]));static BASE_IF_FAIL=!0}class Uo extends uo{static MODEL_CLASS_MAPPINGS=[_o]}class Wo extends uo{static MODEL_CLASS_MAPPINGS=[wo]}class Ho extends uo{static MODEL_CLASS_MAPPINGS=[yo]}class Xo extends uo{static MODEL_CLASS_MAPPINGS=[mo]}class Ko extends uo{static MODEL_CLASS_MAPPINGS=[fo]}class Qo extends uo{static MODEL_CLASS_MAPPINGS=[go]}class Yo extends uo{static MODEL_CLASS_MAPPINGS=[bo]}class Zo extends uo{static MODEL_CLASS_MAPPINGS=[vo]}class Jo extends uo{static MODEL_CLASS_MAPPINGS=[xo]}class el extends uo{static MODEL_CLASS_MAPPINGS=[Mo]}class tl extends uo{static MODEL_CLASS_MAPPINGS=[$o]}class nl extends uo{static MODEL_CLASS_MAPPINGS=[Po]}class rl extends uo{static MODEL_CLASS_MAPPINGS=[Eo]}class al extends uo{static MODEL_CLASS_MAPPINGS=[Fo]}class sl extends uo{static MODEL_CLASS_MAPPINGS=[Co]}class il extends uo{static MODEL_CLASS_MAPPINGS=[So]}class ol extends uo{static MODEL_CLASS_MAPPINGS=[Ao]}class ll extends uo{static MODEL_CLASS_MAPPINGS=[Io]}class ul extends uo{static MODEL_CLASS_MAPPINGS=[zo]}class dl extends uo{static MODEL_CLASS_MAPPINGS=[Oo]}class cl extends uo{static MODEL_CLASS_MAPPINGS=[Bo]}class pl extends uo{static MODEL_CLASS_MAPPINGS=[ko]}class hl extends uo{static MODEL_CLASS_MAPPINGS=[Lo]}class ml extends uo{static MODEL_CLASS_MAPPINGS=[Do]}class fl extends uo{static MODEL_CLASS_MAPPINGS=[Ro]}class gl extends uo{static MODEL_CLASS_MAPPINGS=[No]}class _l extends uo{static MODEL_CLASS_MAPPINGS=[Vo]}class wl extends G{constructor({logits:e,past_key_values:t,encoder_outputs:n,decoder_attentions:r=null,cross_attentions:a=null}){super(),this.logits=e,this.past_key_values=t,this.encoder_outputs=n,this.decoder_attentions=r,this.cross_attentions=a}}class yl extends G{constructor({logits:e}){super(),this.logits=e}}class bl extends G{constructor({logits:e,embeddings:t}){super(),this.logits=e,this.embeddings=t}}class vl extends G{constructor({logits:e}){super(),this.logits=e}}class xl extends G{constructor({logits:e}){super(),this.logits=e}}class Ml extends G{constructor({start_logits:e,end_logits:t}){super(),this.start_logits=e,this.end_logits=t}}class Tl extends G{constructor({logits:e}){super(),this.logits=e}}class kl extends G{constructor({logits:e,past_key_values:t}){super(),this.logits=e,this.past_key_values=t}}class $l extends G{constructor({alphas:e}){super(),this.alphas=e}}class Cl extends G{constructor({waveform:e,spectrogram:t}){super(),this.waveform=e,this.spectrogram=t}}},"./src/models/whisper/common_whisper.js":
 /*!**********************************************!*\
   !*** ./src/models/whisper/common_whisper.js ***!
   \**********************************************/(e,t,n)=>{n.r(t),n.d(t,{WHISPER_LANGUAGE_MAPPING:()=>a,WHISPER_TO_LANGUAGE_CODE_MAPPING:()=>s,whisper_language_to_code:()=>i});const r=[["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"]],a=new Map(r),s=new Map([...r.map((([e,t])=>[t,e])),["burmese","my"],["valencian","ca"],["flemish","nl"],["haitian","ht"],["letzeburgesch","lb"],["pushto","ps"],["panjabi","pa"],["moldavian","ro"],["moldovan","ro"],["sinhalese","si"],["castilian","es"]]);function i(e){e=e.toLowerCase();let t=s.get(e);if(void 0===t){if(!a.has(e)){const t=2===e.length?a.keys():a.values();throw new Error(`Language "${e}" is not supported. Must be one of: ${JSON.stringify(t)}`)}t=e}return t}},"./src/models/whisper/generation_whisper.js":
@@ -226,5 +226,5 @@ var r,a,s,i,o,l,u,d,c,p,h,m,f,g,_,w,y,b,v,x,M,T,k,$,C,S,P,E,F,A,I,z,O,B,L=Object
   \*****************************/(e,t,n)=>{n.r(t),n.d(t,{Tensor:()=>o,cat:()=>y,full:()=>T,full_like:()=>k,interpolate:()=>u,interpolate_4d:()=>d,layer_norm:()=>f,matmul:()=>c,mean:()=>x,mean_pooling:()=>m,ones:()=>$,ones_like:()=>C,permute:()=>l,quantize_embeddings:()=>E,rfft:()=>p,stack:()=>b,std_mean:()=>v,topk:()=>h,zeros:()=>S,zeros_like:()=>P});var r=n(/*! ./maths.js */"./src/utils/maths.js"),a=n(/*! ../backends/onnx.js */"./src/backends/onnx.js"),s=n(/*! ../ops/registry.js */"./src/ops/registry.js");const i=Object.freeze({float32:Float32Array,float16:Uint16Array,float64:Float64Array,string:Array,int8:Int8Array,uint8:Uint8Array,int16:Int16Array,uint16:Uint16Array,int32:Int32Array,uint32:Uint32Array,int64:BigInt64Array,uint64:BigUint64Array,bool:Uint8Array});class o{get dims(){return this.ort_tensor.dims}set dims(e){this.ort_tensor.dims=e}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}ort_tensor;constructor(...e){return(0,a.isONNXTensor)(e[0])?this.ort_tensor=e[0]:this.ort_tensor=new a.Tensor(e[0],e[1],e[2]),new Proxy(this,{get:(e,t)=>{if("string"==typeof t){let n=Number(t);if(Number.isInteger(n))return e._getitem(n)}return e[t]},set:(e,t,n)=>e[t]=n})}dispose(){this.ort_tensor.dispose()}*[Symbol.iterator](){const[e,...t]=this.dims;if(t.length>0){const n=t.reduce(((e,t)=>e*t));for(let r=0;r<e;++r)yield this._subarray(r,n,t)}else yield*this.data}_getitem(e){const[t,...n]=this.dims;if(e=w(e,t),n.length>0){const t=n.reduce(((e,t)=>e*t));return this._subarray(e,t,n)}return new o(this.type,[this.data[e]],n)}indexOf(e){const t=this.data;for(let n=0;n<t.length;++n)if(t[n]==e)return n;return-1}_subarray(e,t,n){const r=e*t,a=(e+1)*t,s="subarray"in this.data?this.data.subarray(r,a):this.data.slice(r,a);return new o(this.type,s,n)}item(){const e=this.data;if(1!==e.length)throw new Error(`a Tensor with ${e.length} elements cannot be converted to Scalar`);return e[0]}tolist(){return function(e,t){const n=e.length,r=t.reduce(((e,t)=>e*t));if(n!==r)throw Error(`cannot reshape array of size ${n} into shape (${t})`);let a=e;for(let e=t.length-1;e>=0;e--)a=a.reduce(((n,r)=>{let a=n[n.length-1];return a.length<t[e]?a.push(r):n.push([r]),n}),[[]]);return a[0]}(this.data,this.dims)}sigmoid(){return this.clone().sigmoid_()}sigmoid_(){const e=this.data;for(let t=0;t<e.length;++t)e[t]=1/(1+Math.exp(-e[t]));return this}map(e){return this.clone().map_(e)}map_(e){const t=this.data;for(let n=0;n<t.length;++n)t[n]=e(t[n],n,t);return this}mul(e){return this.clone().mul_(e)}mul_(e){const t=this.data;for(let n=0;n<t.length;++n)t[n]*=e;return this}div(e){return this.clone().div_(e)}div_(e){const t=this.data;for(let n=0;n<t.length;++n)t[n]/=e;return this}add(e){return this.clone().add_(e)}add_(e){const t=this.data;for(let n=0;n<t.length;++n)t[n]+=e;return this}sub(e){return this.clone().sub_(e)}sub_(e){const t=this.data;for(let n=0;n<t.length;++n)t[n]-=e;return this}clone(){return new o(this.type,this.data.slice(),this.dims.slice())}slice(...e){const t=[],n=[];for(let r=0;r<this.dims.length;++r){let a=e[r];if(null==a)n.push([0,this.dims[r]]),t.push(this.dims[r]);else if("number"==typeof a)a=w(a,this.dims[r],r),n.push([a,a+1]);else{if(!Array.isArray(a)||2!==a.length)throw new Error(`Invalid slice: ${a}`);{let[e,s]=a;if(e=null===e?0:w(e,this.dims[r],r,!1),s=null===s?this.dims[r]:w(s,this.dims[r],r,!1),e>s)throw new Error(`Invalid slice: ${a}`);const i=[Math.max(e,0),Math.min(s,this.dims[r])];n.push(i),t.push(i[1]-i[0])}}}const r=n.map((([e,t])=>t-e)),a=r.reduce(((e,t)=>e*t)),s=this.data,i=new s.constructor(a),l=this.stride();for(let e=0;e<a;++e){let t=0;for(let a=r.length-1,s=e;a>=0;--a){const e=r[a];t+=(s%e+n[a][0])*l[a],s=Math.floor(s/e)}i[e]=s[t]}return new o(this.type,i,t)}permute(...e){return l(this,e)}transpose(...e){return this.permute(...e)}sum(e=null,t=!1){return this.norm(1,e,t)}norm(e="fro",t=null,n=!1){if("fro"===e)e=2;else if("string"==typeof e)throw Error(`Unsupported norm: ${e}`);const r=this.data;if(null===t){let t=r.reduce(((t,n)=>t+n**e),0)**(1/e);return new o(this.type,[t],[])}t=w(t,this.dims.length);const a=this.dims.slice();a[t]=1;const s=new r.constructor(r.length/this.dims[t]);for(let n=0;n<r.length;++n){let i=0;for(let e=this.dims.length-1,r=n,s=1;e>=0;--e){const n=this.dims[e];if(e!==t){i+=r%n*s,s*=a[e]}r=Math.floor(r/n)}s[i]+=r[n]**e}if(1!==e)for(let t=0;t<s.length;++t)s[t]=s[t]**(1/e);return n||a.splice(t,1),new o(this.type,s,a)}normalize_(e=2,t=1){t=w(t,this.dims.length);const n=this.norm(e,t,!0),r=this.data,a=n.data;for(let e=0;e<r.length;++e){let n=0;for(let r=this.dims.length-1,a=e,s=1;r>=0;--r){const e=this.dims[r];if(r!==t){n+=a%e*s,s*=this.dims[r]}a=Math.floor(a/e)}r[e]/=a[n]}return this}normalize(e=2,t=1){return this.clone().normalize_(e,t)}stride(){return function(e){const t=new Array(e.length);for(let n=e.length-1,r=1;n>=0;--n)t[n]=r,r*=e[n];return t}(this.dims)}squeeze(e=null){return new o(this.type,this.data,g(this.dims,e))}squeeze_(e=null){return this.dims=g(this.dims,e),this}unsqueeze(e=null){return new o(this.type,this.data,_(this.dims,e))}unsqueeze_(e=null){return this.dims=_(this.dims,e),this}flatten_(e=0,t=-1){t=(t+this.dims.length)%this.dims.length;let n=this.dims.slice(0,e),r=this.dims.slice(e,t+1),a=this.dims.slice(t+1);return this.dims=[...n,r.reduce(((e,t)=>e*t),1),...a],this}flatten(e=0,t=-1){return this.clone().flatten_(e,t)}view(...e){let t=-1;for(let n=0;n<e.length;++n)if(-1===e[n]){if(-1!==t)throw new Error("Only one dimension can be inferred");t=n}const n=this.data;if(-1!==t){const r=e.reduce(((e,n,r)=>r!==t?e*n:e),1);e[t]=n.length/r}return new o(this.type,n,e)}neg_(){const e=this.data;for(let t=0;t<e.length;++t)e[t]=-e[t];return this}neg(){return this.clone().neg_()}clamp_(e,t){const n=this.data;for(let r=0;r<n.length;++r)n[r]=Math.min(Math.max(n[r],e),t);return this}clamp(e,t){return this.clone().clamp_(e,t)}round_(){const e=this.data;for(let t=0;t<e.length;++t)e[t]=Math.round(e[t]);return this}round(){return this.clone().round_()}mean(e=null,t=!1){return x(this,e,t)}to(e){if(this.type===e)return this;if(!i.hasOwnProperty(e))throw new Error(`Unsupported type: ${e}`);return new o(e,i[e].from(this.data),this.dims)}}function l(e,t){const[n,a]=(0,r.permute_data)(e.data,e.dims,t);return new o(e.type,n,a)}function u(e,[t,n],a="bilinear",s=!1){const i=e.dims.at(-3)??1,l=e.dims.at(-2),u=e.dims.at(-1);let d=(0,r.interpolate_data)(e.data,[i,l,u],[t,n],a,s);return new o(e.type,d,[i,t,n])}async function d(e,{size:t=null,mode:n="bilinear"}={}){if(4!==e.dims.length)throw new Error("`interpolate_4d` currently only supports 4D input.");if(!t)throw new Error("`interpolate_4d` requires a `size` argument.");let r,a;if(2===t.length)r=[...e.dims.slice(0,2),...t];else if(3===t.length)r=[e.dims[0],...t];else{if(4!==t.length)throw new Error("`size` must be of length 2, 3, or 4.");r=t}if("bilinear"===n)a=await s.TensorOpRegistry.bilinear_interpolate_4d;else{if("bicubic"!==n)throw new Error(`Unsupported mode: ${n}`);a=await s.TensorOpRegistry.bicubic_interpolate_4d}const i=new o("int64",new BigInt64Array(r.map(BigInt)),[r.length]);return await a({x:e,s:i})}async function c(e,t){const n=await s.TensorOpRegistry.matmul;return await n({a:e,b:t})}async function p(e,t){const n=await s.TensorOpRegistry.rfft;return await n({x:e,a:t})}async function h(e,t){const n=await s.TensorOpRegistry.top_k;return t=null===t?e.dims.at(-1):Math.min(t,e.dims.at(-1)),await n({x:e,k:new o("int64",[BigInt(t)],[1])})}function m(e,t){const n=e.data,r=t.data,a=[e.dims[0],e.dims[2]],s=new n.constructor(a[0]*a[1]),[i,l,u]=e.dims;let d=0;for(let e=0;e<i;++e){const t=e*u*l;for(let a=0;a<u;++a){let i=0,o=0;const c=e*l,p=t+a;for(let e=0;e<l;++e){const t=Number(r[c+e]);o+=t,i+=n[p+e*u]*t}const h=i/o;s[d++]=h}}return new o(e.type,s,a)}function f(e,t,{eps:n=1e-5}={}){if(2!==e.dims.length)throw new Error("`layer_norm` currently only supports 2D input.");const[r,a]=e.dims;if(1!==t.length&&t[0]!==a)throw new Error("`normalized_shape` must be a 1D array with shape `[input.dims[1]]`.");const[s,i]=v(e,1,0,!0),l=s.data,u=i.data,d=e.data,c=new d.constructor(d.length);for(let e=0;e<r;++e){const t=e*a;for(let r=0;r<a;++r){const a=t+r;c[a]=(d[a]-u[e])/(l[e]+n)}}return new o(e.type,c,e.dims)}function g(e,t){return e=e.slice(),null===t?e=e.filter((e=>1!==e)):"number"==typeof t?1===e[t]&&e.splice(t,1):Array.isArray(t)&&(e=e.filter(((e,n)=>1!==e||!t.includes(n)))),e}function _(e,t){return t=w(t,e.length+1),(e=e.slice()).splice(t,0,1),e}function w(e,t,n=null,r=!0){if(r&&(e<-t||e>=t))throw new Error(`IndexError: index ${e} is out of bounds for dimension${null===n?"":" "+n} with size ${t}`);return e<0&&(e=(e%t+t)%t),e}function y(e,t=0){t=w(t,e[0].dims.length);const n=e[0].dims.slice();n[t]=e.reduce(((e,n)=>e+n.dims[t]),0);const r=n.reduce(((e,t)=>e*t),1),a=new e[0].data.constructor(r),s=e[0].type;if(0===t){let t=0;for(const n of e){const e=n.data;a.set(e,t),t+=e.length}}else{let r=0;for(let s=0;s<e.length;++s){const{data:i,dims:o}=e[s];for(let e=0;e<i.length;++e){let s=0;for(let a=o.length-1,i=e,l=1;a>=0;--a){const e=o[a];let u=i%e;a===t&&(u+=r),s+=u*l,l*=n[a],i=Math.floor(i/e)}a[s]=i[e]}r+=o[t]}}return new o(s,a,n)}function b(e,t=0){return y(e.map((e=>e.unsqueeze(t))),t)}function v(e,t=null,n=1,r=!1){const a=e.data,s=e.dims;if(null===t){const t=a.reduce(((e,t)=>e+t),0)/a.length,r=Math.sqrt(a.reduce(((e,n)=>e+(n-t)**2),0)/(a.length-n)),s=new o(e.type,[t],[]);return[new o(e.type,[r],[]),s]}const i=x(e,t=w(t,s.length),r),l=i.data,u=s.slice();u[t]=1;const d=new a.constructor(a.length/s[t]);for(let e=0;e<a.length;++e){let n=0;for(let r=s.length-1,a=e,i=1;r>=0;--r){const e=s[r];if(r!==t){n+=a%e*i,i*=u[r]}a=Math.floor(a/e)}d[n]+=(a[e]-l[n])**2}for(let e=0;e<d.length;++e)d[e]=Math.sqrt(d[e]/(s[t]-n));r||u.splice(t,1);return[new o(e.type,d,u),i]}function x(e,t=null,n=!1){const r=e.data;if(null===t){const t=r.reduce(((e,t)=>e+t),0);return new o(e.type,[t/r.length],[])}const a=e.dims;t=w(t,a.length);const s=a.slice();s[t]=1;const i=new r.constructor(r.length/a[t]);for(let e=0;e<r.length;++e){let n=0;for(let r=a.length-1,i=e,o=1;r>=0;--r){const e=a[r];if(r!==t){n+=i%e*o,o*=s[r]}i=Math.floor(i/e)}i[n]+=r[e]}if(1!==a[t])for(let e=0;e<i.length;++e)i[e]=i[e]/a[t];return n||s.splice(t,1),new o(e.type,i,s)}function M(e,t,n,r){const a=e.reduce(((e,t)=>e*t),1);return new o(n,new r(a).fill(t),e)}function T(e,t){let n,r;if("number"==typeof t)n="float32",r=Float32Array;else{if("bigint"!=typeof t)throw new Error("Unsupported data type: "+typeof t);n="int64",r=BigInt64Array}return M(e,t,n,r)}function k(e,t){return T(e.dims,t)}function $(e){return M(e,1n,"int64",BigInt64Array)}function C(e){return $(e.dims)}function S(e){return M(e,0n,"int64",BigInt64Array)}function P(e){return S(e.dims)}function E(e,t){if(2!==e.dims.length)throw new Error("The tensor must have 2 dimensions");if(e.dims.at(-1)%8!=0)throw new Error("The last dimension of the tensor must be a multiple of 8");if(!["binary","ubinary"].includes(t))throw new Error("The precision must be either 'binary' or 'ubinary'");const n="binary"===t,r=n?"int8":"uint8",a=n?Int8Array:Uint8Array,s=e.data,i=new a(s.length/8);for(let e=0;e<s.length;++e){const t=s[e]>0?1:0,r=Math.floor(e/8),a=e%8;i[r]|=t<<7-a,n&&0===a&&(i[r]-=128)}return new o(r,i,[e.dims[0],e.dims[1]/8])}}},r={};function a(e){var t=r[e];if(void 0!==t)return t.exports;var s=r[e]={exports:{}};return n[e](s,s.exports,a),s.exports}a.m=n,t=Object.getPrototypeOf?e=>Object.getPrototypeOf(e):e=>e.__proto__,a.t=function(n,r){if(1&r&&(n=this(n)),8&r)return n;if("object"==typeof n&&n){if(4&r&&n.__esModule)return n;if(16&r&&"function"==typeof n.then)return n}var s=Object.create(null);a.r(s);var i={};e=e||[null,t({}),t([]),t(t)];for(var o=2&r&&n;"object"==typeof o&&!~e.indexOf(o);o=t(o))Object.getOwnPropertyNames(o).forEach((e=>i[e]=()=>n[e]));return i.default=()=>n,a.d(s,i),s},a.d=(e,t)=>{for(var n in t)a.o(t,n)&&!a.o(e,n)&&Object.defineProperty(e,n,{enumerable:!0,get:t[n]})},a.o=(e,t)=>Object.prototype.hasOwnProperty.call(e,t),a.r=e=>{"undefined"!=typeof Symbol&&Symbol.toStringTag&&Object.defineProperty(e,Symbol.toStringTag,{value:"Module"}),Object.defineProperty(e,"__esModule",{value:!0})},(()=>{var e;if("string"==typeof import.meta.url&&(e=import.meta.url),!e)throw new Error("Automatic publicPath is not supported in this browser");e=e.replace(/#.*$/,"").replace(/\?.*$/,"").replace(/\/[^\/]+$/,"/"),a.p=e})(),a.b=new URL("./",import.meta.url);var s={};
 /*!*****************************!*\
   !*** ./src/transformers.js ***!
-  \*****************************/a.r(s),a.d(s,{ASTFeatureExtractor:()=>d.ASTFeatureExtractor,ASTForAudioClassification:()=>l.ASTForAudioClassification,ASTModel:()=>l.ASTModel,ASTPreTrainedModel:()=>l.ASTPreTrainedModel,AlbertForMaskedLM:()=>l.AlbertForMaskedLM,AlbertForQuestionAnswering:()=>l.AlbertForQuestionAnswering,AlbertForSequenceClassification:()=>l.AlbertForSequenceClassification,AlbertModel:()=>l.AlbertModel,AlbertPreTrainedModel:()=>l.AlbertPreTrainedModel,AlbertTokenizer:()=>u.AlbertTokenizer,AudioClassificationPipeline:()=>o.AudioClassificationPipeline,AutoConfig:()=>c.AutoConfig,AutoModel:()=>l.AutoModel,AutoModelForAudioClassification:()=>l.AutoModelForAudioClassification,AutoModelForAudioFrameClassification:()=>l.AutoModelForAudioFrameClassification,AutoModelForCTC:()=>l.AutoModelForCTC,AutoModelForCausalLM:()=>l.AutoModelForCausalLM,AutoModelForDepthEstimation:()=>l.AutoModelForDepthEstimation,AutoModelForDocumentQuestionAnswering:()=>l.AutoModelForDocumentQuestionAnswering,AutoModelForImageClassification:()=>l.AutoModelForImageClassification,AutoModelForImageFeatureExtraction:()=>l.AutoModelForImageFeatureExtraction,AutoModelForImageMatting:()=>l.AutoModelForImageMatting,AutoModelForImageSegmentation:()=>l.AutoModelForImageSegmentation,AutoModelForImageToImage:()=>l.AutoModelForImageToImage,AutoModelForMaskGeneration:()=>l.AutoModelForMaskGeneration,AutoModelForMaskedLM:()=>l.AutoModelForMaskedLM,AutoModelForNormalEstimation:()=>l.AutoModelForNormalEstimation,AutoModelForObjectDetection:()=>l.AutoModelForObjectDetection,AutoModelForQuestionAnswering:()=>l.AutoModelForQuestionAnswering,AutoModelForSemanticSegmentation:()=>l.AutoModelForSemanticSegmentation,AutoModelForSeq2SeqLM:()=>l.AutoModelForSeq2SeqLM,AutoModelForSequenceClassification:()=>l.AutoModelForSequenceClassification,AutoModelForSpeechSeq2Seq:()=>l.AutoModelForSpeechSeq2Seq,AutoModelForTextToSpectrogram:()=>l.AutoModelForTextToSpectrogram,AutoModelForTextToWaveform:()=>l.AutoModelForTextToWaveform,AutoModelForTokenClassification:()=>l.AutoModelForTokenClassification,AutoModelForUniversalSegmentation:()=>l.AutoModelForUniversalSegmentation,AutoModelForVision2Seq:()=>l.AutoModelForVision2Seq,AutoModelForXVector:()=>l.AutoModelForXVector,AutoModelForZeroShotObjectDetection:()=>l.AutoModelForZeroShotObjectDetection,AutoProcessor:()=>d.AutoProcessor,AutoTokenizer:()=>u.AutoTokenizer,AutomaticSpeechRecognitionPipeline:()=>o.AutomaticSpeechRecognitionPipeline,BartForConditionalGeneration:()=>l.BartForConditionalGeneration,BartForSequenceClassification:()=>l.BartForSequenceClassification,BartModel:()=>l.BartModel,BartPretrainedModel:()=>l.BartPretrainedModel,BartTokenizer:()=>u.BartTokenizer,BaseModelOutput:()=>l.BaseModelOutput,BaseStreamer:()=>g.BaseStreamer,BeitFeatureExtractor:()=>d.BeitFeatureExtractor,BeitForImageClassification:()=>l.BeitForImageClassification,BeitModel:()=>l.BeitModel,BeitPreTrainedModel:()=>l.BeitPreTrainedModel,BertForMaskedLM:()=>l.BertForMaskedLM,BertForQuestionAnswering:()=>l.BertForQuestionAnswering,BertForSequenceClassification:()=>l.BertForSequenceClassification,BertForTokenClassification:()=>l.BertForTokenClassification,BertModel:()=>l.BertModel,BertPreTrainedModel:()=>l.BertPreTrainedModel,BertTokenizer:()=>u.BertTokenizer,BitImageProcessor:()=>d.BitImageProcessor,BlenderbotForConditionalGeneration:()=>l.BlenderbotForConditionalGeneration,BlenderbotModel:()=>l.BlenderbotModel,BlenderbotPreTrainedModel:()=>l.BlenderbotPreTrainedModel,BlenderbotSmallForConditionalGeneration:()=>l.BlenderbotSmallForConditionalGeneration,BlenderbotSmallModel:()=>l.BlenderbotSmallModel,BlenderbotSmallPreTrainedModel:()=>l.BlenderbotSmallPreTrainedModel,BlenderbotSmallTokenizer:()=>u.BlenderbotSmallTokenizer,BlenderbotTokenizer:()=>u.BlenderbotTokenizer,BloomForCausalLM:()=>l.BloomForCausalLM,BloomModel:()=>l.BloomModel,BloomPreTrainedModel:()=>l.BloomPreTrainedModel,BloomTokenizer:()=>u.BloomTokenizer,CLIPFeatureExtractor:()=>d.CLIPFeatureExtractor,CLIPImageProcessor:()=>d.CLIPImageProcessor,CLIPModel:()=>l.CLIPModel,CLIPPreTrainedModel:()=>l.CLIPPreTrainedModel,CLIPSegForImageSegmentation:()=>l.CLIPSegForImageSegmentation,CLIPSegModel:()=>l.CLIPSegModel,CLIPSegPreTrainedModel:()=>l.CLIPSegPreTrainedModel,CLIPTextModel:()=>l.CLIPTextModel,CLIPTextModelWithProjection:()=>l.CLIPTextModelWithProjection,CLIPTokenizer:()=>u.CLIPTokenizer,CLIPVisionModel:()=>l.CLIPVisionModel,CLIPVisionModelWithProjection:()=>l.CLIPVisionModelWithProjection,CamembertForMaskedLM:()=>l.CamembertForMaskedLM,CamembertForQuestionAnswering:()=>l.CamembertForQuestionAnswering,CamembertForSequenceClassification:()=>l.CamembertForSequenceClassification,CamembertForTokenClassification:()=>l.CamembertForTokenClassification,CamembertModel:()=>l.CamembertModel,CamembertPreTrainedModel:()=>l.CamembertPreTrainedModel,CamembertTokenizer:()=>u.CamembertTokenizer,CausalLMOutput:()=>l.CausalLMOutput,CausalLMOutputWithPast:()=>l.CausalLMOutputWithPast,ChineseCLIPFeatureExtractor:()=>d.ChineseCLIPFeatureExtractor,ChineseCLIPModel:()=>l.ChineseCLIPModel,ChineseCLIPPreTrainedModel:()=>l.ChineseCLIPPreTrainedModel,ClapAudioModelWithProjection:()=>l.ClapAudioModelWithProjection,ClapFeatureExtractor:()=>d.ClapFeatureExtractor,ClapModel:()=>l.ClapModel,ClapPreTrainedModel:()=>l.ClapPreTrainedModel,ClapTextModelWithProjection:()=>l.ClapTextModelWithProjection,CodeGenForCausalLM:()=>l.CodeGenForCausalLM,CodeGenModel:()=>l.CodeGenModel,CodeGenPreTrainedModel:()=>l.CodeGenPreTrainedModel,CodeGenTokenizer:()=>u.CodeGenTokenizer,CodeLlamaTokenizer:()=>u.CodeLlamaTokenizer,CohereForCausalLM:()=>l.CohereForCausalLM,CohereModel:()=>l.CohereModel,CoherePreTrainedModel:()=>l.CoherePreTrainedModel,CohereTokenizer:()=>u.CohereTokenizer,ConvBertForMaskedLM:()=>l.ConvBertForMaskedLM,ConvBertForQuestionAnswering:()=>l.ConvBertForQuestionAnswering,ConvBertForSequenceClassification:()=>l.ConvBertForSequenceClassification,ConvBertForTokenClassification:()=>l.ConvBertForTokenClassification,ConvBertModel:()=>l.ConvBertModel,ConvBertPreTrainedModel:()=>l.ConvBertPreTrainedModel,ConvBertTokenizer:()=>u.ConvBertTokenizer,ConvNextFeatureExtractor:()=>d.ConvNextFeatureExtractor,ConvNextForImageClassification:()=>l.ConvNextForImageClassification,ConvNextImageProcessor:()=>d.ConvNextImageProcessor,ConvNextModel:()=>l.ConvNextModel,ConvNextPreTrainedModel:()=>l.ConvNextPreTrainedModel,ConvNextV2ForImageClassification:()=>l.ConvNextV2ForImageClassification,ConvNextV2Model:()=>l.ConvNextV2Model,ConvNextV2PreTrainedModel:()=>l.ConvNextV2PreTrainedModel,DPTFeatureExtractor:()=>d.DPTFeatureExtractor,DPTForDepthEstimation:()=>l.DPTForDepthEstimation,DPTImageProcessor:()=>d.DPTImageProcessor,DPTModel:()=>l.DPTModel,DPTPreTrainedModel:()=>l.DPTPreTrainedModel,DebertaForMaskedLM:()=>l.DebertaForMaskedLM,DebertaForQuestionAnswering:()=>l.DebertaForQuestionAnswering,DebertaForSequenceClassification:()=>l.DebertaForSequenceClassification,DebertaForTokenClassification:()=>l.DebertaForTokenClassification,DebertaModel:()=>l.DebertaModel,DebertaPreTrainedModel:()=>l.DebertaPreTrainedModel,DebertaTokenizer:()=>u.DebertaTokenizer,DebertaV2ForMaskedLM:()=>l.DebertaV2ForMaskedLM,DebertaV2ForQuestionAnswering:()=>l.DebertaV2ForQuestionAnswering,DebertaV2ForSequenceClassification:()=>l.DebertaV2ForSequenceClassification,DebertaV2ForTokenClassification:()=>l.DebertaV2ForTokenClassification,DebertaV2Model:()=>l.DebertaV2Model,DebertaV2PreTrainedModel:()=>l.DebertaV2PreTrainedModel,DebertaV2Tokenizer:()=>u.DebertaV2Tokenizer,DecisionTransformerModel:()=>l.DecisionTransformerModel,DecisionTransformerPreTrainedModel:()=>l.DecisionTransformerPreTrainedModel,DeiTFeatureExtractor:()=>d.DeiTFeatureExtractor,DeiTForImageClassification:()=>l.DeiTForImageClassification,DeiTModel:()=>l.DeiTModel,DeiTPreTrainedModel:()=>l.DeiTPreTrainedModel,DepthAnythingForDepthEstimation:()=>l.DepthAnythingForDepthEstimation,DepthAnythingPreTrainedModel:()=>l.DepthAnythingPreTrainedModel,DepthEstimationPipeline:()=>o.DepthEstimationPipeline,DepthProForDepthEstimation:()=>l.DepthProForDepthEstimation,DepthProPreTrainedModel:()=>l.DepthProPreTrainedModel,DetrFeatureExtractor:()=>d.DetrFeatureExtractor,DetrForObjectDetection:()=>l.DetrForObjectDetection,DetrForSegmentation:()=>l.DetrForSegmentation,DetrModel:()=>l.DetrModel,DetrObjectDetectionOutput:()=>l.DetrObjectDetectionOutput,DetrPreTrainedModel:()=>l.DetrPreTrainedModel,DetrSegmentationOutput:()=>l.DetrSegmentationOutput,Dinov2ForImageClassification:()=>l.Dinov2ForImageClassification,Dinov2Model:()=>l.Dinov2Model,Dinov2PreTrainedModel:()=>l.Dinov2PreTrainedModel,DistilBertForMaskedLM:()=>l.DistilBertForMaskedLM,DistilBertForQuestionAnswering:()=>l.DistilBertForQuestionAnswering,DistilBertForSequenceClassification:()=>l.DistilBertForSequenceClassification,DistilBertForTokenClassification:()=>l.DistilBertForTokenClassification,DistilBertModel:()=>l.DistilBertModel,DistilBertPreTrainedModel:()=>l.DistilBertPreTrainedModel,DistilBertTokenizer:()=>u.DistilBertTokenizer,DocumentQuestionAnsweringPipeline:()=>o.DocumentQuestionAnsweringPipeline,DonutFeatureExtractor:()=>d.DonutFeatureExtractor,DonutImageProcessor:()=>d.DonutImageProcessor,DonutSwinModel:()=>l.DonutSwinModel,DonutSwinPreTrainedModel:()=>l.DonutSwinPreTrainedModel,EfficientNetForImageClassification:()=>l.EfficientNetForImageClassification,EfficientNetImageProcessor:()=>d.EfficientNetImageProcessor,EfficientNetModel:()=>l.EfficientNetModel,EfficientNetPreTrainedModel:()=>l.EfficientNetPreTrainedModel,ElectraForMaskedLM:()=>l.ElectraForMaskedLM,ElectraForQuestionAnswering:()=>l.ElectraForQuestionAnswering,ElectraForSequenceClassification:()=>l.ElectraForSequenceClassification,ElectraForTokenClassification:()=>l.ElectraForTokenClassification,ElectraModel:()=>l.ElectraModel,ElectraPreTrainedModel:()=>l.ElectraPreTrainedModel,ElectraTokenizer:()=>u.ElectraTokenizer,EosTokenCriteria:()=>_.EosTokenCriteria,EsmForMaskedLM:()=>l.EsmForMaskedLM,EsmForSequenceClassification:()=>l.EsmForSequenceClassification,EsmForTokenClassification:()=>l.EsmForTokenClassification,EsmModel:()=>l.EsmModel,EsmPreTrainedModel:()=>l.EsmPreTrainedModel,EsmTokenizer:()=>u.EsmTokenizer,FFT:()=>f.FFT,FalconForCausalLM:()=>l.FalconForCausalLM,FalconModel:()=>l.FalconModel,FalconPreTrainedModel:()=>l.FalconPreTrainedModel,FalconTokenizer:()=>u.FalconTokenizer,FastViTForImageClassification:()=>l.FastViTForImageClassification,FastViTModel:()=>l.FastViTModel,FastViTPreTrainedModel:()=>l.FastViTPreTrainedModel,FeatureExtractionPipeline:()=>o.FeatureExtractionPipeline,FeatureExtractor:()=>d.FeatureExtractor,FillMaskPipeline:()=>o.FillMaskPipeline,Florence2ForConditionalGeneration:()=>l.Florence2ForConditionalGeneration,Florence2PreTrainedModel:()=>l.Florence2PreTrainedModel,Florence2Processor:()=>d.Florence2Processor,GLPNFeatureExtractor:()=>d.GLPNFeatureExtractor,GLPNForDepthEstimation:()=>l.GLPNForDepthEstimation,GLPNModel:()=>l.GLPNModel,GLPNPreTrainedModel:()=>l.GLPNPreTrainedModel,GPT2LMHeadModel:()=>l.GPT2LMHeadModel,GPT2Model:()=>l.GPT2Model,GPT2PreTrainedModel:()=>l.GPT2PreTrainedModel,GPT2Tokenizer:()=>u.GPT2Tokenizer,GPTBigCodeForCausalLM:()=>l.GPTBigCodeForCausalLM,GPTBigCodeModel:()=>l.GPTBigCodeModel,GPTBigCodePreTrainedModel:()=>l.GPTBigCodePreTrainedModel,GPTJForCausalLM:()=>l.GPTJForCausalLM,GPTJModel:()=>l.GPTJModel,GPTJPreTrainedModel:()=>l.GPTJPreTrainedModel,GPTNeoForCausalLM:()=>l.GPTNeoForCausalLM,GPTNeoModel:()=>l.GPTNeoModel,GPTNeoPreTrainedModel:()=>l.GPTNeoPreTrainedModel,GPTNeoXForCausalLM:()=>l.GPTNeoXForCausalLM,GPTNeoXModel:()=>l.GPTNeoXModel,GPTNeoXPreTrainedModel:()=>l.GPTNeoXPreTrainedModel,GPTNeoXTokenizer:()=>u.GPTNeoXTokenizer,Gemma2ForCausalLM:()=>l.Gemma2ForCausalLM,Gemma2Model:()=>l.Gemma2Model,Gemma2PreTrainedModel:()=>l.Gemma2PreTrainedModel,GemmaForCausalLM:()=>l.GemmaForCausalLM,GemmaModel:()=>l.GemmaModel,GemmaPreTrainedModel:()=>l.GemmaPreTrainedModel,GemmaTokenizer:()=>u.GemmaTokenizer,GraniteForCausalLM:()=>l.GraniteForCausalLM,GraniteModel:()=>l.GraniteModel,GranitePreTrainedModel:()=>l.GranitePreTrainedModel,Grok1Tokenizer:()=>u.Grok1Tokenizer,GroupViTModel:()=>l.GroupViTModel,GroupViTPreTrainedModel:()=>l.GroupViTPreTrainedModel,HerbertTokenizer:()=>u.HerbertTokenizer,HieraForImageClassification:()=>l.HieraForImageClassification,HieraModel:()=>l.HieraModel,HieraPreTrainedModel:()=>l.HieraPreTrainedModel,HubertForCTC:()=>l.HubertForCTC,HubertForSequenceClassification:()=>l.HubertForSequenceClassification,HubertModel:()=>l.HubertModel,HubertPreTrainedModel:()=>l.HubertPreTrainedModel,ImageClassificationPipeline:()=>o.ImageClassificationPipeline,ImageFeatureExtractionPipeline:()=>o.ImageFeatureExtractionPipeline,ImageFeatureExtractor:()=>d.ImageFeatureExtractor,ImageMattingOutput:()=>l.ImageMattingOutput,ImageSegmentationPipeline:()=>o.ImageSegmentationPipeline,ImageToImagePipeline:()=>o.ImageToImagePipeline,ImageToTextPipeline:()=>o.ImageToTextPipeline,InterruptableStoppingCriteria:()=>_.InterruptableStoppingCriteria,JAISLMHeadModel:()=>l.JAISLMHeadModel,JAISModel:()=>l.JAISModel,JAISPreTrainedModel:()=>l.JAISPreTrainedModel,LlamaForCausalLM:()=>l.LlamaForCausalLM,LlamaModel:()=>l.LlamaModel,LlamaPreTrainedModel:()=>l.LlamaPreTrainedModel,LlamaTokenizer:()=>u.LlamaTokenizer,LlavaForConditionalGeneration:()=>l.LlavaForConditionalGeneration,LlavaPreTrainedModel:()=>l.LlavaPreTrainedModel,LongT5ForConditionalGeneration:()=>l.LongT5ForConditionalGeneration,LongT5Model:()=>l.LongT5Model,LongT5PreTrainedModel:()=>l.LongT5PreTrainedModel,M2M100ForConditionalGeneration:()=>l.M2M100ForConditionalGeneration,M2M100Model:()=>l.M2M100Model,M2M100PreTrainedModel:()=>l.M2M100PreTrainedModel,M2M100Tokenizer:()=>u.M2M100Tokenizer,MBart50Tokenizer:()=>u.MBart50Tokenizer,MBartForCausalLM:()=>l.MBartForCausalLM,MBartForConditionalGeneration:()=>l.MBartForConditionalGeneration,MBartForSequenceClassification:()=>l.MBartForSequenceClassification,MBartModel:()=>l.MBartModel,MBartPreTrainedModel:()=>l.MBartPreTrainedModel,MBartTokenizer:()=>u.MBartTokenizer,MPNetForMaskedLM:()=>l.MPNetForMaskedLM,MPNetForQuestionAnswering:()=>l.MPNetForQuestionAnswering,MPNetForSequenceClassification:()=>l.MPNetForSequenceClassification,MPNetForTokenClassification:()=>l.MPNetForTokenClassification,MPNetModel:()=>l.MPNetModel,MPNetPreTrainedModel:()=>l.MPNetPreTrainedModel,MPNetTokenizer:()=>u.MPNetTokenizer,MT5ForConditionalGeneration:()=>l.MT5ForConditionalGeneration,MT5Model:()=>l.MT5Model,MT5PreTrainedModel:()=>l.MT5PreTrainedModel,MarianMTModel:()=>l.MarianMTModel,MarianModel:()=>l.MarianModel,MarianPreTrainedModel:()=>l.MarianPreTrainedModel,MarianTokenizer:()=>u.MarianTokenizer,MaskFormerFeatureExtractor:()=>d.MaskFormerFeatureExtractor,MaskFormerForInstanceSegmentation:()=>l.MaskFormerForInstanceSegmentation,MaskFormerModel:()=>l.MaskFormerModel,MaskFormerPreTrainedModel:()=>l.MaskFormerPreTrainedModel,MaskedLMOutput:()=>l.MaskedLMOutput,MaxLengthCriteria:()=>_.MaxLengthCriteria,MistralForCausalLM:()=>l.MistralForCausalLM,MistralModel:()=>l.MistralModel,MistralPreTrainedModel:()=>l.MistralPreTrainedModel,MobileBertForMaskedLM:()=>l.MobileBertForMaskedLM,MobileBertForQuestionAnswering:()=>l.MobileBertForQuestionAnswering,MobileBertForSequenceClassification:()=>l.MobileBertForSequenceClassification,MobileBertModel:()=>l.MobileBertModel,MobileBertPreTrainedModel:()=>l.MobileBertPreTrainedModel,MobileBertTokenizer:()=>u.MobileBertTokenizer,MobileNetV1FeatureExtractor:()=>d.MobileNetV1FeatureExtractor,MobileNetV1ForImageClassification:()=>l.MobileNetV1ForImageClassification,MobileNetV1Model:()=>l.MobileNetV1Model,MobileNetV1PreTrainedModel:()=>l.MobileNetV1PreTrainedModel,MobileNetV2FeatureExtractor:()=>d.MobileNetV2FeatureExtractor,MobileNetV2ForImageClassification:()=>l.MobileNetV2ForImageClassification,MobileNetV2Model:()=>l.MobileNetV2Model,MobileNetV2PreTrainedModel:()=>l.MobileNetV2PreTrainedModel,MobileNetV3FeatureExtractor:()=>d.MobileNetV3FeatureExtractor,MobileNetV3ForImageClassification:()=>l.MobileNetV3ForImageClassification,MobileNetV3Model:()=>l.MobileNetV3Model,MobileNetV3PreTrainedModel:()=>l.MobileNetV3PreTrainedModel,MobileNetV4FeatureExtractor:()=>d.MobileNetV4FeatureExtractor,MobileNetV4ForImageClassification:()=>l.MobileNetV4ForImageClassification,MobileNetV4Model:()=>l.MobileNetV4Model,MobileNetV4PreTrainedModel:()=>l.MobileNetV4PreTrainedModel,MobileViTFeatureExtractor:()=>d.MobileViTFeatureExtractor,MobileViTForImageClassification:()=>l.MobileViTForImageClassification,MobileViTImageProcessor:()=>d.MobileViTImageProcessor,MobileViTModel:()=>l.MobileViTModel,MobileViTPreTrainedModel:()=>l.MobileViTPreTrainedModel,MobileViTV2ForImageClassification:()=>l.MobileViTV2ForImageClassification,MobileViTV2Model:()=>l.MobileViTV2Model,MobileViTV2PreTrainedModel:()=>l.MobileViTV2PreTrainedModel,ModelOutput:()=>l.ModelOutput,Moondream1ForConditionalGeneration:()=>l.Moondream1ForConditionalGeneration,MptForCausalLM:()=>l.MptForCausalLM,MptModel:()=>l.MptModel,MptPreTrainedModel:()=>l.MptPreTrainedModel,MusicgenForCausalLM:()=>l.MusicgenForCausalLM,MusicgenForConditionalGeneration:()=>l.MusicgenForConditionalGeneration,MusicgenModel:()=>l.MusicgenModel,MusicgenPreTrainedModel:()=>l.MusicgenPreTrainedModel,NllbTokenizer:()=>u.NllbTokenizer,NomicBertModel:()=>l.NomicBertModel,NomicBertPreTrainedModel:()=>l.NomicBertPreTrainedModel,NougatImageProcessor:()=>d.NougatImageProcessor,NougatTokenizer:()=>u.NougatTokenizer,OPTForCausalLM:()=>l.OPTForCausalLM,OPTModel:()=>l.OPTModel,OPTPreTrainedModel:()=>l.OPTPreTrainedModel,ObjectDetectionPipeline:()=>o.ObjectDetectionPipeline,OpenELMForCausalLM:()=>l.OpenELMForCausalLM,OpenELMModel:()=>l.OpenELMModel,OpenELMPreTrainedModel:()=>l.OpenELMPreTrainedModel,OwlViTFeatureExtractor:()=>d.OwlViTFeatureExtractor,OwlViTForObjectDetection:()=>l.OwlViTForObjectDetection,OwlViTModel:()=>l.OwlViTModel,OwlViTPreTrainedModel:()=>l.OwlViTPreTrainedModel,OwlViTProcessor:()=>d.OwlViTProcessor,Owlv2ForObjectDetection:()=>l.Owlv2ForObjectDetection,Owlv2ImageProcessor:()=>d.Owlv2ImageProcessor,Owlv2Model:()=>l.Owlv2Model,Owlv2PreTrainedModel:()=>l.Owlv2PreTrainedModel,Phi3ForCausalLM:()=>l.Phi3ForCausalLM,Phi3Model:()=>l.Phi3Model,Phi3PreTrainedModel:()=>l.Phi3PreTrainedModel,PhiForCausalLM:()=>l.PhiForCausalLM,PhiModel:()=>l.PhiModel,PhiPreTrainedModel:()=>l.PhiPreTrainedModel,Pipeline:()=>o.Pipeline,PreTrainedModel:()=>l.PreTrainedModel,PreTrainedTokenizer:()=>u.PreTrainedTokenizer,PretrainedConfig:()=>c.PretrainedConfig,PretrainedMixin:()=>l.PretrainedMixin,Processor:()=>d.Processor,PvtForImageClassification:()=>l.PvtForImageClassification,PvtImageProcessor:()=>d.PvtImageProcessor,PvtModel:()=>l.PvtModel,PvtPreTrainedModel:()=>l.PvtPreTrainedModel,PyAnnoteFeatureExtractor:()=>d.PyAnnoteFeatureExtractor,PyAnnoteForAudioFrameClassification:()=>l.PyAnnoteForAudioFrameClassification,PyAnnoteModel:()=>l.PyAnnoteModel,PyAnnotePreTrainedModel:()=>l.PyAnnotePreTrainedModel,PyAnnoteProcessor:()=>d.PyAnnoteProcessor,QuestionAnsweringModelOutput:()=>l.QuestionAnsweringModelOutput,QuestionAnsweringPipeline:()=>o.QuestionAnsweringPipeline,Qwen2ForCausalLM:()=>l.Qwen2ForCausalLM,Qwen2Model:()=>l.Qwen2Model,Qwen2PreTrainedModel:()=>l.Qwen2PreTrainedModel,Qwen2Tokenizer:()=>u.Qwen2Tokenizer,RTDetrForObjectDetection:()=>l.RTDetrForObjectDetection,RTDetrImageProcessor:()=>d.RTDetrImageProcessor,RTDetrModel:()=>l.RTDetrModel,RTDetrObjectDetectionOutput:()=>l.RTDetrObjectDetectionOutput,RTDetrPreTrainedModel:()=>l.RTDetrPreTrainedModel,RawImage:()=>h.RawImage,ResNetForImageClassification:()=>l.ResNetForImageClassification,ResNetModel:()=>l.ResNetModel,ResNetPreTrainedModel:()=>l.ResNetPreTrainedModel,RoFormerForMaskedLM:()=>l.RoFormerForMaskedLM,RoFormerForQuestionAnswering:()=>l.RoFormerForQuestionAnswering,RoFormerForSequenceClassification:()=>l.RoFormerForSequenceClassification,RoFormerForTokenClassification:()=>l.RoFormerForTokenClassification,RoFormerModel:()=>l.RoFormerModel,RoFormerPreTrainedModel:()=>l.RoFormerPreTrainedModel,RoFormerTokenizer:()=>u.RoFormerTokenizer,RobertaForMaskedLM:()=>l.RobertaForMaskedLM,RobertaForQuestionAnswering:()=>l.RobertaForQuestionAnswering,RobertaForSequenceClassification:()=>l.RobertaForSequenceClassification,RobertaForTokenClassification:()=>l.RobertaForTokenClassification,RobertaModel:()=>l.RobertaModel,RobertaPreTrainedModel:()=>l.RobertaPreTrainedModel,RobertaTokenizer:()=>u.RobertaTokenizer,SamImageProcessor:()=>d.SamImageProcessor,SamImageSegmentationOutput:()=>l.SamImageSegmentationOutput,SamModel:()=>l.SamModel,SamPreTrainedModel:()=>l.SamPreTrainedModel,SamProcessor:()=>d.SamProcessor,SapiensFeatureExtractor:()=>d.SapiensFeatureExtractor,SapiensForDepthEstimation:()=>l.SapiensForDepthEstimation,SapiensForNormalEstimation:()=>l.SapiensForNormalEstimation,SapiensForSemanticSegmentation:()=>l.SapiensForSemanticSegmentation,SapiensPreTrainedModel:()=>l.SapiensPreTrainedModel,SeamlessM4TFeatureExtractor:()=>d.SeamlessM4TFeatureExtractor,SegformerFeatureExtractor:()=>d.SegformerFeatureExtractor,SegformerForImageClassification:()=>l.SegformerForImageClassification,SegformerForSemanticSegmentation:()=>l.SegformerForSemanticSegmentation,SegformerModel:()=>l.SegformerModel,SegformerPreTrainedModel:()=>l.SegformerPreTrainedModel,Seq2SeqLMOutput:()=>l.Seq2SeqLMOutput,SequenceClassifierOutput:()=>l.SequenceClassifierOutput,SiglipImageProcessor:()=>d.SiglipImageProcessor,SiglipModel:()=>l.SiglipModel,SiglipPreTrainedModel:()=>l.SiglipPreTrainedModel,SiglipTextModel:()=>l.SiglipTextModel,SiglipTokenizer:()=>u.SiglipTokenizer,SiglipVisionModel:()=>l.SiglipVisionModel,SpeechT5FeatureExtractor:()=>d.SpeechT5FeatureExtractor,SpeechT5ForSpeechToText:()=>l.SpeechT5ForSpeechToText,SpeechT5ForTextToSpeech:()=>l.SpeechT5ForTextToSpeech,SpeechT5HifiGan:()=>l.SpeechT5HifiGan,SpeechT5Model:()=>l.SpeechT5Model,SpeechT5PreTrainedModel:()=>l.SpeechT5PreTrainedModel,SpeechT5Processor:()=>d.SpeechT5Processor,SpeechT5Tokenizer:()=>u.SpeechT5Tokenizer,SqueezeBertForMaskedLM:()=>l.SqueezeBertForMaskedLM,SqueezeBertForQuestionAnswering:()=>l.SqueezeBertForQuestionAnswering,SqueezeBertForSequenceClassification:()=>l.SqueezeBertForSequenceClassification,SqueezeBertModel:()=>l.SqueezeBertModel,SqueezeBertPreTrainedModel:()=>l.SqueezeBertPreTrainedModel,SqueezeBertTokenizer:()=>u.SqueezeBertTokenizer,StableLmForCausalLM:()=>l.StableLmForCausalLM,StableLmModel:()=>l.StableLmModel,StableLmPreTrainedModel:()=>l.StableLmPreTrainedModel,Starcoder2ForCausalLM:()=>l.Starcoder2ForCausalLM,Starcoder2Model:()=>l.Starcoder2Model,Starcoder2PreTrainedModel:()=>l.Starcoder2PreTrainedModel,StoppingCriteria:()=>_.StoppingCriteria,StoppingCriteriaList:()=>_.StoppingCriteriaList,SummarizationPipeline:()=>o.SummarizationPipeline,Swin2SRForImageSuperResolution:()=>l.Swin2SRForImageSuperResolution,Swin2SRImageProcessor:()=>d.Swin2SRImageProcessor,Swin2SRModel:()=>l.Swin2SRModel,Swin2SRPreTrainedModel:()=>l.Swin2SRPreTrainedModel,SwinForImageClassification:()=>l.SwinForImageClassification,SwinModel:()=>l.SwinModel,SwinPreTrainedModel:()=>l.SwinPreTrainedModel,T5ForConditionalGeneration:()=>l.T5ForConditionalGeneration,T5Model:()=>l.T5Model,T5PreTrainedModel:()=>l.T5PreTrainedModel,T5Tokenizer:()=>u.T5Tokenizer,TableTransformerForObjectDetection:()=>l.TableTransformerForObjectDetection,TableTransformerModel:()=>l.TableTransformerModel,TableTransformerObjectDetectionOutput:()=>l.TableTransformerObjectDetectionOutput,TableTransformerPreTrainedModel:()=>l.TableTransformerPreTrainedModel,Tensor:()=>m.Tensor,Text2TextGenerationPipeline:()=>o.Text2TextGenerationPipeline,TextClassificationPipeline:()=>o.TextClassificationPipeline,TextGenerationPipeline:()=>o.TextGenerationPipeline,TextStreamer:()=>g.TextStreamer,TextToAudioPipeline:()=>o.TextToAudioPipeline,TokenClassificationPipeline:()=>o.TokenClassificationPipeline,TokenClassifierOutput:()=>l.TokenClassifierOutput,TokenizerModel:()=>u.TokenizerModel,TrOCRForCausalLM:()=>l.TrOCRForCausalLM,TrOCRPreTrainedModel:()=>l.TrOCRPreTrainedModel,TranslationPipeline:()=>o.TranslationPipeline,UniSpeechForCTC:()=>l.UniSpeechForCTC,UniSpeechForSequenceClassification:()=>l.UniSpeechForSequenceClassification,UniSpeechModel:()=>l.UniSpeechModel,UniSpeechPreTrainedModel:()=>l.UniSpeechPreTrainedModel,UniSpeechSatForAudioFrameClassification:()=>l.UniSpeechSatForAudioFrameClassification,UniSpeechSatForCTC:()=>l.UniSpeechSatForCTC,UniSpeechSatForSequenceClassification:()=>l.UniSpeechSatForSequenceClassification,UniSpeechSatModel:()=>l.UniSpeechSatModel,UniSpeechSatPreTrainedModel:()=>l.UniSpeechSatPreTrainedModel,ViTFeatureExtractor:()=>d.ViTFeatureExtractor,ViTForImageClassification:()=>l.ViTForImageClassification,ViTImageProcessor:()=>d.ViTImageProcessor,ViTMAEModel:()=>l.ViTMAEModel,ViTMAEPreTrainedModel:()=>l.ViTMAEPreTrainedModel,ViTMSNForImageClassification:()=>l.ViTMSNForImageClassification,ViTMSNModel:()=>l.ViTMSNModel,ViTMSNPreTrainedModel:()=>l.ViTMSNPreTrainedModel,ViTModel:()=>l.ViTModel,ViTPreTrainedModel:()=>l.ViTPreTrainedModel,VisionEncoderDecoderModel:()=>l.VisionEncoderDecoderModel,VitMatteForImageMatting:()=>l.VitMatteForImageMatting,VitMatteImageProcessor:()=>d.VitMatteImageProcessor,VitMattePreTrainedModel:()=>l.VitMattePreTrainedModel,VitsModel:()=>l.VitsModel,VitsModelOutput:()=>l.VitsModelOutput,VitsPreTrainedModel:()=>l.VitsPreTrainedModel,VitsTokenizer:()=>u.VitsTokenizer,Wav2Vec2BertForCTC:()=>l.Wav2Vec2BertForCTC,Wav2Vec2BertForSequenceClassification:()=>l.Wav2Vec2BertForSequenceClassification,Wav2Vec2BertModel:()=>l.Wav2Vec2BertModel,Wav2Vec2BertPreTrainedModel:()=>l.Wav2Vec2BertPreTrainedModel,Wav2Vec2CTCTokenizer:()=>u.Wav2Vec2CTCTokenizer,Wav2Vec2FeatureExtractor:()=>d.Wav2Vec2FeatureExtractor,Wav2Vec2ForAudioFrameClassification:()=>l.Wav2Vec2ForAudioFrameClassification,Wav2Vec2ForCTC:()=>l.Wav2Vec2ForCTC,Wav2Vec2ForSequenceClassification:()=>l.Wav2Vec2ForSequenceClassification,Wav2Vec2Model:()=>l.Wav2Vec2Model,Wav2Vec2PreTrainedModel:()=>l.Wav2Vec2PreTrainedModel,Wav2Vec2ProcessorWithLM:()=>d.Wav2Vec2ProcessorWithLM,WavLMForAudioFrameClassification:()=>l.WavLMForAudioFrameClassification,WavLMForCTC:()=>l.WavLMForCTC,WavLMForSequenceClassification:()=>l.WavLMForSequenceClassification,WavLMForXVector:()=>l.WavLMForXVector,WavLMModel:()=>l.WavLMModel,WavLMPreTrainedModel:()=>l.WavLMPreTrainedModel,WeSpeakerFeatureExtractor:()=>d.WeSpeakerFeatureExtractor,WeSpeakerResNetModel:()=>l.WeSpeakerResNetModel,WeSpeakerResNetPreTrainedModel:()=>l.WeSpeakerResNetPreTrainedModel,WhisperFeatureExtractor:()=>d.WhisperFeatureExtractor,WhisperForConditionalGeneration:()=>l.WhisperForConditionalGeneration,WhisperModel:()=>l.WhisperModel,WhisperPreTrainedModel:()=>l.WhisperPreTrainedModel,WhisperProcessor:()=>d.WhisperProcessor,WhisperTextStreamer:()=>g.WhisperTextStreamer,WhisperTokenizer:()=>u.WhisperTokenizer,XLMForQuestionAnswering:()=>l.XLMForQuestionAnswering,XLMForSequenceClassification:()=>l.XLMForSequenceClassification,XLMForTokenClassification:()=>l.XLMForTokenClassification,XLMModel:()=>l.XLMModel,XLMPreTrainedModel:()=>l.XLMPreTrainedModel,XLMRobertaForMaskedLM:()=>l.XLMRobertaForMaskedLM,XLMRobertaForQuestionAnswering:()=>l.XLMRobertaForQuestionAnswering,XLMRobertaForSequenceClassification:()=>l.XLMRobertaForSequenceClassification,XLMRobertaForTokenClassification:()=>l.XLMRobertaForTokenClassification,XLMRobertaModel:()=>l.XLMRobertaModel,XLMRobertaPreTrainedModel:()=>l.XLMRobertaPreTrainedModel,XLMRobertaTokenizer:()=>u.XLMRobertaTokenizer,XLMTokenizer:()=>u.XLMTokenizer,XLMWithLMHeadModel:()=>l.XLMWithLMHeadModel,XVectorOutput:()=>l.XVectorOutput,YolosFeatureExtractor:()=>d.YolosFeatureExtractor,YolosForObjectDetection:()=>l.YolosForObjectDetection,YolosModel:()=>l.YolosModel,YolosObjectDetectionOutput:()=>l.YolosObjectDetectionOutput,YolosPreTrainedModel:()=>l.YolosPreTrainedModel,ZeroShotAudioClassificationPipeline:()=>o.ZeroShotAudioClassificationPipeline,ZeroShotClassificationPipeline:()=>o.ZeroShotClassificationPipeline,ZeroShotImageClassificationPipeline:()=>o.ZeroShotImageClassificationPipeline,ZeroShotObjectDetectionPipeline:()=>o.ZeroShotObjectDetectionPipeline,bankers_round:()=>f.bankers_round,cat:()=>m.cat,cos_sim:()=>f.cos_sim,dot:()=>f.dot,dynamic_time_warping:()=>f.dynamic_time_warping,env:()=>i.env,full:()=>m.full,full_like:()=>m.full_like,getKeyValueShapes:()=>c.getKeyValueShapes,hamming:()=>p.hamming,hanning:()=>p.hanning,interpolate:()=>m.interpolate,interpolate_4d:()=>m.interpolate_4d,interpolate_data:()=>f.interpolate_data,is_chinese_char:()=>u.is_chinese_char,layer_norm:()=>m.layer_norm,log_softmax:()=>f.log_softmax,magnitude:()=>f.magnitude,matmul:()=>m.matmul,max:()=>f.max,mean:()=>m.mean,mean_pooling:()=>m.mean_pooling,medianFilter:()=>f.medianFilter,mel_filter_bank:()=>p.mel_filter_bank,min:()=>f.min,ones:()=>m.ones,ones_like:()=>m.ones_like,permute:()=>m.permute,permute_data:()=>f.permute_data,pipeline:()=>o.pipeline,quantize_embeddings:()=>m.quantize_embeddings,read_audio:()=>p.read_audio,rfft:()=>m.rfft,round:()=>f.round,softmax:()=>f.softmax,spectrogram:()=>p.spectrogram,stack:()=>m.stack,std_mean:()=>m.std_mean,topk:()=>m.topk,window_function:()=>p.window_function,zeros:()=>m.zeros,zeros_like:()=>m.zeros_like});var i=a(/*! ./env.js */"./src/env.js"),o=a(/*! ./pipelines.js */"./src/pipelines.js"),l=a(/*! ./models.js */"./src/models.js"),u=a(/*! ./tokenizers.js */"./src/tokenizers.js"),d=a(/*! ./processors.js */"./src/processors.js"),c=a(/*! ./configs.js */"./src/configs.js"),p=a(/*! ./utils/audio.js */"./src/utils/audio.js"),h=a(/*! ./utils/image.js */"./src/utils/image.js"),m=a(/*! ./utils/tensor.js */"./src/utils/tensor.js"),f=a(/*! ./utils/maths.js */"./src/utils/maths.js"),g=a(/*! ./generation/streamers.js */"./src/generation/streamers.js"),_=a(/*! ./generation/stopping_criteria.js */"./src/generation/stopping_criteria.js"),w=s.ASTFeatureExtractor,y=s.ASTForAudioClassification,b=s.ASTModel,v=s.ASTPreTrainedModel,x=s.AlbertForMaskedLM,M=s.AlbertForQuestionAnswering,T=s.AlbertForSequenceClassification,k=s.AlbertModel,$=s.AlbertPreTrainedModel,C=s.AlbertTokenizer,S=s.AudioClassificationPipeline,P=s.AutoConfig,E=s.AutoModel,F=s.AutoModelForAudioClassification,A=s.AutoModelForAudioFrameClassification,I=s.AutoModelForCTC,z=s.AutoModelForCausalLM,O=s.AutoModelForDepthEstimation,B=s.AutoModelForDocumentQuestionAnswering,L=s.AutoModelForImageClassification,D=s.AutoModelForImageFeatureExtraction,R=s.AutoModelForImageMatting,N=s.AutoModelForImageSegmentation,V=s.AutoModelForImageToImage,j=s.AutoModelForMaskGeneration,q=s.AutoModelForMaskedLM,G=s.AutoModelForNormalEstimation,U=s.AutoModelForObjectDetection,W=s.AutoModelForQuestionAnswering,H=s.AutoModelForSemanticSegmentation,X=s.AutoModelForSeq2SeqLM,K=s.AutoModelForSequenceClassification,Q=s.AutoModelForSpeechSeq2Seq,Y=s.AutoModelForTextToSpectrogram,Z=s.AutoModelForTextToWaveform,J=s.AutoModelForTokenClassification,ee=s.AutoModelForUniversalSegmentation,te=s.AutoModelForVision2Seq,ne=s.AutoModelForXVector,re=s.AutoModelForZeroShotObjectDetection,ae=s.AutoProcessor,se=s.AutoTokenizer,ie=s.AutomaticSpeechRecognitionPipeline,oe=s.BartForConditionalGeneration,le=s.BartForSequenceClassification,ue=s.BartModel,de=s.BartPretrainedModel,ce=s.BartTokenizer,pe=s.BaseModelOutput,he=s.BaseStreamer,me=s.BeitFeatureExtractor,fe=s.BeitForImageClassification,ge=s.BeitModel,_e=s.BeitPreTrainedModel,we=s.BertForMaskedLM,ye=s.BertForQuestionAnswering,be=s.BertForSequenceClassification,ve=s.BertForTokenClassification,xe=s.BertModel,Me=s.BertPreTrainedModel,Te=s.BertTokenizer,ke=s.BitImageProcessor,$e=s.BlenderbotForConditionalGeneration,Ce=s.BlenderbotModel,Se=s.BlenderbotPreTrainedModel,Pe=s.BlenderbotSmallForConditionalGeneration,Ee=s.BlenderbotSmallModel,Fe=s.BlenderbotSmallPreTrainedModel,Ae=s.BlenderbotSmallTokenizer,Ie=s.BlenderbotTokenizer,ze=s.BloomForCausalLM,Oe=s.BloomModel,Be=s.BloomPreTrainedModel,Le=s.BloomTokenizer,De=s.CLIPFeatureExtractor,Re=s.CLIPImageProcessor,Ne=s.CLIPModel,Ve=s.CLIPPreTrainedModel,je=s.CLIPSegForImageSegmentation,qe=s.CLIPSegModel,Ge=s.CLIPSegPreTrainedModel,Ue=s.CLIPTextModel,We=s.CLIPTextModelWithProjection,He=s.CLIPTokenizer,Xe=s.CLIPVisionModel,Ke=s.CLIPVisionModelWithProjection,Qe=s.CamembertForMaskedLM,Ye=s.CamembertForQuestionAnswering,Ze=s.CamembertForSequenceClassification,Je=s.CamembertForTokenClassification,et=s.CamembertModel,tt=s.CamembertPreTrainedModel,nt=s.CamembertTokenizer,rt=s.CausalLMOutput,at=s.CausalLMOutputWithPast,st=s.ChineseCLIPFeatureExtractor,it=s.ChineseCLIPModel,ot=s.ChineseCLIPPreTrainedModel,lt=s.ClapAudioModelWithProjection,ut=s.ClapFeatureExtractor,dt=s.ClapModel,ct=s.ClapPreTrainedModel,pt=s.ClapTextModelWithProjection,ht=s.CodeGenForCausalLM,mt=s.CodeGenModel,ft=s.CodeGenPreTrainedModel,gt=s.CodeGenTokenizer,_t=s.CodeLlamaTokenizer,wt=s.CohereForCausalLM,yt=s.CohereModel,bt=s.CoherePreTrainedModel,vt=s.CohereTokenizer,xt=s.ConvBertForMaskedLM,Mt=s.ConvBertForQuestionAnswering,Tt=s.ConvBertForSequenceClassification,kt=s.ConvBertForTokenClassification,$t=s.ConvBertModel,Ct=s.ConvBertPreTrainedModel,St=s.ConvBertTokenizer,Pt=s.ConvNextFeatureExtractor,Et=s.ConvNextForImageClassification,Ft=s.ConvNextImageProcessor,At=s.ConvNextModel,It=s.ConvNextPreTrainedModel,zt=s.ConvNextV2ForImageClassification,Ot=s.ConvNextV2Model,Bt=s.ConvNextV2PreTrainedModel,Lt=s.DPTFeatureExtractor,Dt=s.DPTForDepthEstimation,Rt=s.DPTImageProcessor,Nt=s.DPTModel,Vt=s.DPTPreTrainedModel,jt=s.DebertaForMaskedLM,qt=s.DebertaForQuestionAnswering,Gt=s.DebertaForSequenceClassification,Ut=s.DebertaForTokenClassification,Wt=s.DebertaModel,Ht=s.DebertaPreTrainedModel,Xt=s.DebertaTokenizer,Kt=s.DebertaV2ForMaskedLM,Qt=s.DebertaV2ForQuestionAnswering,Yt=s.DebertaV2ForSequenceClassification,Zt=s.DebertaV2ForTokenClassification,Jt=s.DebertaV2Model,en=s.DebertaV2PreTrainedModel,tn=s.DebertaV2Tokenizer,nn=s.DecisionTransformerModel,rn=s.DecisionTransformerPreTrainedModel,an=s.DeiTFeatureExtractor,sn=s.DeiTForImageClassification,on=s.DeiTModel,ln=s.DeiTPreTrainedModel,un=s.DepthAnythingForDepthEstimation,dn=s.DepthAnythingPreTrainedModel,cn=s.DepthEstimationPipeline,pn=s.DepthProForDepthEstimation,hn=s.DepthProPreTrainedModel,mn=s.DetrFeatureExtractor,fn=s.DetrForObjectDetection,gn=s.DetrForSegmentation,_n=s.DetrModel,wn=s.DetrObjectDetectionOutput,yn=s.DetrPreTrainedModel,bn=s.DetrSegmentationOutput,vn=s.Dinov2ForImageClassification,xn=s.Dinov2Model,Mn=s.Dinov2PreTrainedModel,Tn=s.DistilBertForMaskedLM,kn=s.DistilBertForQuestionAnswering,$n=s.DistilBertForSequenceClassification,Cn=s.DistilBertForTokenClassification,Sn=s.DistilBertModel,Pn=s.DistilBertPreTrainedModel,En=s.DistilBertTokenizer,Fn=s.DocumentQuestionAnsweringPipeline,An=s.DonutFeatureExtractor,In=s.DonutImageProcessor,zn=s.DonutSwinModel,On=s.DonutSwinPreTrainedModel,Bn=s.EfficientNetForImageClassification,Ln=s.EfficientNetImageProcessor,Dn=s.EfficientNetModel,Rn=s.EfficientNetPreTrainedModel,Nn=s.ElectraForMaskedLM,Vn=s.ElectraForQuestionAnswering,jn=s.ElectraForSequenceClassification,qn=s.ElectraForTokenClassification,Gn=s.ElectraModel,Un=s.ElectraPreTrainedModel,Wn=s.ElectraTokenizer,Hn=s.EosTokenCriteria,Xn=s.EsmForMaskedLM,Kn=s.EsmForSequenceClassification,Qn=s.EsmForTokenClassification,Yn=s.EsmModel,Zn=s.EsmPreTrainedModel,Jn=s.EsmTokenizer,er=s.FFT,tr=s.FalconForCausalLM,nr=s.FalconModel,rr=s.FalconPreTrainedModel,ar=s.FalconTokenizer,sr=s.FastViTForImageClassification,ir=s.FastViTModel,or=s.FastViTPreTrainedModel,lr=s.FeatureExtractionPipeline,ur=s.FeatureExtractor,dr=s.FillMaskPipeline,cr=s.Florence2ForConditionalGeneration,pr=s.Florence2PreTrainedModel,hr=s.Florence2Processor,mr=s.GLPNFeatureExtractor,fr=s.GLPNForDepthEstimation,gr=s.GLPNModel,_r=s.GLPNPreTrainedModel,wr=s.GPT2LMHeadModel,yr=s.GPT2Model,br=s.GPT2PreTrainedModel,vr=s.GPT2Tokenizer,xr=s.GPTBigCodeForCausalLM,Mr=s.GPTBigCodeModel,Tr=s.GPTBigCodePreTrainedModel,kr=s.GPTJForCausalLM,$r=s.GPTJModel,Cr=s.GPTJPreTrainedModel,Sr=s.GPTNeoForCausalLM,Pr=s.GPTNeoModel,Er=s.GPTNeoPreTrainedModel,Fr=s.GPTNeoXForCausalLM,Ar=s.GPTNeoXModel,Ir=s.GPTNeoXPreTrainedModel,zr=s.GPTNeoXTokenizer,Or=s.Gemma2ForCausalLM,Br=s.Gemma2Model,Lr=s.Gemma2PreTrainedModel,Dr=s.GemmaForCausalLM,Rr=s.GemmaModel,Nr=s.GemmaPreTrainedModel,Vr=s.GemmaTokenizer,jr=s.GraniteForCausalLM,qr=s.GraniteModel,Gr=s.GranitePreTrainedModel,Ur=s.Grok1Tokenizer,Wr=s.GroupViTModel,Hr=s.GroupViTPreTrainedModel,Xr=s.HerbertTokenizer,Kr=s.HieraForImageClassification,Qr=s.HieraModel,Yr=s.HieraPreTrainedModel,Zr=s.HubertForCTC,Jr=s.HubertForSequenceClassification,ea=s.HubertModel,ta=s.HubertPreTrainedModel,na=s.ImageClassificationPipeline,ra=s.ImageFeatureExtractionPipeline,aa=s.ImageFeatureExtractor,sa=s.ImageMattingOutput,ia=s.ImageSegmentationPipeline,oa=s.ImageToImagePipeline,la=s.ImageToTextPipeline,ua=s.InterruptableStoppingCriteria,da=s.JAISLMHeadModel,ca=s.JAISModel,pa=s.JAISPreTrainedModel,ha=s.LlamaForCausalLM,ma=s.LlamaModel,fa=s.LlamaPreTrainedModel,ga=s.LlamaTokenizer,_a=s.LlavaForConditionalGeneration,wa=s.LlavaPreTrainedModel,ya=s.LongT5ForConditionalGeneration,ba=s.LongT5Model,va=s.LongT5PreTrainedModel,xa=s.M2M100ForConditionalGeneration,Ma=s.M2M100Model,Ta=s.M2M100PreTrainedModel,ka=s.M2M100Tokenizer,$a=s.MBart50Tokenizer,Ca=s.MBartForCausalLM,Sa=s.MBartForConditionalGeneration,Pa=s.MBartForSequenceClassification,Ea=s.MBartModel,Fa=s.MBartPreTrainedModel,Aa=s.MBartTokenizer,Ia=s.MPNetForMaskedLM,za=s.MPNetForQuestionAnswering,Oa=s.MPNetForSequenceClassification,Ba=s.MPNetForTokenClassification,La=s.MPNetModel,Da=s.MPNetPreTrainedModel,Ra=s.MPNetTokenizer,Na=s.MT5ForConditionalGeneration,Va=s.MT5Model,ja=s.MT5PreTrainedModel,qa=s.MarianMTModel,Ga=s.MarianModel,Ua=s.MarianPreTrainedModel,Wa=s.MarianTokenizer,Ha=s.MaskFormerFeatureExtractor,Xa=s.MaskFormerForInstanceSegmentation,Ka=s.MaskFormerModel,Qa=s.MaskFormerPreTrainedModel,Ya=s.MaskedLMOutput,Za=s.MaxLengthCriteria,Ja=s.MistralForCausalLM,es=s.MistralModel,ts=s.MistralPreTrainedModel,ns=s.MobileBertForMaskedLM,rs=s.MobileBertForQuestionAnswering,as=s.MobileBertForSequenceClassification,ss=s.MobileBertModel,is=s.MobileBertPreTrainedModel,os=s.MobileBertTokenizer,ls=s.MobileNetV1FeatureExtractor,us=s.MobileNetV1ForImageClassification,ds=s.MobileNetV1Model,cs=s.MobileNetV1PreTrainedModel,ps=s.MobileNetV2FeatureExtractor,hs=s.MobileNetV2ForImageClassification,ms=s.MobileNetV2Model,fs=s.MobileNetV2PreTrainedModel,gs=s.MobileNetV3FeatureExtractor,_s=s.MobileNetV3ForImageClassification,ws=s.MobileNetV3Model,ys=s.MobileNetV3PreTrainedModel,bs=s.MobileNetV4FeatureExtractor,vs=s.MobileNetV4ForImageClassification,xs=s.MobileNetV4Model,Ms=s.MobileNetV4PreTrainedModel,Ts=s.MobileViTFeatureExtractor,ks=s.MobileViTForImageClassification,$s=s.MobileViTImageProcessor,Cs=s.MobileViTModel,Ss=s.MobileViTPreTrainedModel,Ps=s.MobileViTV2ForImageClassification,Es=s.MobileViTV2Model,Fs=s.MobileViTV2PreTrainedModel,As=s.ModelOutput,Is=s.Moondream1ForConditionalGeneration,zs=s.MptForCausalLM,Os=s.MptModel,Bs=s.MptPreTrainedModel,Ls=s.MusicgenForCausalLM,Ds=s.MusicgenForConditionalGeneration,Rs=s.MusicgenModel,Ns=s.MusicgenPreTrainedModel,Vs=s.NllbTokenizer,js=s.NomicBertModel,qs=s.NomicBertPreTrainedModel,Gs=s.NougatImageProcessor,Us=s.NougatTokenizer,Ws=s.OPTForCausalLM,Hs=s.OPTModel,Xs=s.OPTPreTrainedModel,Ks=s.ObjectDetectionPipeline,Qs=s.OpenELMForCausalLM,Ys=s.OpenELMModel,Zs=s.OpenELMPreTrainedModel,Js=s.OwlViTFeatureExtractor,ei=s.OwlViTForObjectDetection,ti=s.OwlViTModel,ni=s.OwlViTPreTrainedModel,ri=s.OwlViTProcessor,ai=s.Owlv2ForObjectDetection,si=s.Owlv2ImageProcessor,ii=s.Owlv2Model,oi=s.Owlv2PreTrainedModel,li=s.Phi3ForCausalLM,ui=s.Phi3Model,di=s.Phi3PreTrainedModel,ci=s.PhiForCausalLM,pi=s.PhiModel,hi=s.PhiPreTrainedModel,mi=s.Pipeline,fi=s.PreTrainedModel,gi=s.PreTrainedTokenizer,_i=s.PretrainedConfig,wi=s.PretrainedMixin,yi=s.Processor,bi=s.PvtForImageClassification,vi=s.PvtImageProcessor,xi=s.PvtModel,Mi=s.PvtPreTrainedModel,Ti=s.PyAnnoteFeatureExtractor,ki=s.PyAnnoteForAudioFrameClassification,$i=s.PyAnnoteModel,Ci=s.PyAnnotePreTrainedModel,Si=s.PyAnnoteProcessor,Pi=s.QuestionAnsweringModelOutput,Ei=s.QuestionAnsweringPipeline,Fi=s.Qwen2ForCausalLM,Ai=s.Qwen2Model,Ii=s.Qwen2PreTrainedModel,zi=s.Qwen2Tokenizer,Oi=s.RTDetrForObjectDetection,Bi=s.RTDetrImageProcessor,Li=s.RTDetrModel,Di=s.RTDetrObjectDetectionOutput,Ri=s.RTDetrPreTrainedModel,Ni=s.RawImage,Vi=s.ResNetForImageClassification,ji=s.ResNetModel,qi=s.ResNetPreTrainedModel,Gi=s.RoFormerForMaskedLM,Ui=s.RoFormerForQuestionAnswering,Wi=s.RoFormerForSequenceClassification,Hi=s.RoFormerForTokenClassification,Xi=s.RoFormerModel,Ki=s.RoFormerPreTrainedModel,Qi=s.RoFormerTokenizer,Yi=s.RobertaForMaskedLM,Zi=s.RobertaForQuestionAnswering,Ji=s.RobertaForSequenceClassification,eo=s.RobertaForTokenClassification,to=s.RobertaModel,no=s.RobertaPreTrainedModel,ro=s.RobertaTokenizer,ao=s.SamImageProcessor,so=s.SamImageSegmentationOutput,io=s.SamModel,oo=s.SamPreTrainedModel,lo=s.SamProcessor,uo=s.SapiensFeatureExtractor,co=s.SapiensForDepthEstimation,po=s.SapiensForNormalEstimation,ho=s.SapiensForSemanticSegmentation,mo=s.SapiensPreTrainedModel,fo=s.SeamlessM4TFeatureExtractor,go=s.SegformerFeatureExtractor,_o=s.SegformerForImageClassification,wo=s.SegformerForSemanticSegmentation,yo=s.SegformerModel,bo=s.SegformerPreTrainedModel,vo=s.Seq2SeqLMOutput,xo=s.SequenceClassifierOutput,Mo=s.SiglipImageProcessor,To=s.SiglipModel,ko=s.SiglipPreTrainedModel,$o=s.SiglipTextModel,Co=s.SiglipTokenizer,So=s.SiglipVisionModel,Po=s.SpeechT5FeatureExtractor,Eo=s.SpeechT5ForSpeechToText,Fo=s.SpeechT5ForTextToSpeech,Ao=s.SpeechT5HifiGan,Io=s.SpeechT5Model,zo=s.SpeechT5PreTrainedModel,Oo=s.SpeechT5Processor,Bo=s.SpeechT5Tokenizer,Lo=s.SqueezeBertForMaskedLM,Do=s.SqueezeBertForQuestionAnswering,Ro=s.SqueezeBertForSequenceClassification,No=s.SqueezeBertModel,Vo=s.SqueezeBertPreTrainedModel,jo=s.SqueezeBertTokenizer,qo=s.StableLmForCausalLM,Go=s.StableLmModel,Uo=s.StableLmPreTrainedModel,Wo=s.Starcoder2ForCausalLM,Ho=s.Starcoder2Model,Xo=s.Starcoder2PreTrainedModel,Ko=s.StoppingCriteria,Qo=s.StoppingCriteriaList,Yo=s.SummarizationPipeline,Zo=s.Swin2SRForImageSuperResolution,Jo=s.Swin2SRImageProcessor,el=s.Swin2SRModel,tl=s.Swin2SRPreTrainedModel,nl=s.SwinForImageClassification,rl=s.SwinModel,al=s.SwinPreTrainedModel,sl=s.T5ForConditionalGeneration,il=s.T5Model,ol=s.T5PreTrainedModel,ll=s.T5Tokenizer,ul=s.TableTransformerForObjectDetection,dl=s.TableTransformerModel,cl=s.TableTransformerObjectDetectionOutput,pl=s.TableTransformerPreTrainedModel,hl=s.Tensor,ml=s.Text2TextGenerationPipeline,fl=s.TextClassificationPipeline,gl=s.TextGenerationPipeline,_l=s.TextStreamer,wl=s.TextToAudioPipeline,yl=s.TokenClassificationPipeline,bl=s.TokenClassifierOutput,vl=s.TokenizerModel,xl=s.TrOCRForCausalLM,Ml=s.TrOCRPreTrainedModel,Tl=s.TranslationPipeline,kl=s.UniSpeechForCTC,$l=s.UniSpeechForSequenceClassification,Cl=s.UniSpeechModel,Sl=s.UniSpeechPreTrainedModel,Pl=s.UniSpeechSatForAudioFrameClassification,El=s.UniSpeechSatForCTC,Fl=s.UniSpeechSatForSequenceClassification,Al=s.UniSpeechSatModel,Il=s.UniSpeechSatPreTrainedModel,zl=s.ViTFeatureExtractor,Ol=s.ViTForImageClassification,Bl=s.ViTImageProcessor,Ll=s.ViTMAEModel,Dl=s.ViTMAEPreTrainedModel,Rl=s.ViTMSNForImageClassification,Nl=s.ViTMSNModel,Vl=s.ViTMSNPreTrainedModel,jl=s.ViTModel,ql=s.ViTPreTrainedModel,Gl=s.VisionEncoderDecoderModel,Ul=s.VitMatteForImageMatting,Wl=s.VitMatteImageProcessor,Hl=s.VitMattePreTrainedModel,Xl=s.VitsModel,Kl=s.VitsModelOutput,Ql=s.VitsPreTrainedModel,Yl=s.VitsTokenizer,Zl=s.Wav2Vec2BertForCTC,Jl=s.Wav2Vec2BertForSequenceClassification,eu=s.Wav2Vec2BertModel,tu=s.Wav2Vec2BertPreTrainedModel,nu=s.Wav2Vec2CTCTokenizer,ru=s.Wav2Vec2FeatureExtractor,au=s.Wav2Vec2ForAudioFrameClassification,su=s.Wav2Vec2ForCTC,iu=s.Wav2Vec2ForSequenceClassification,ou=s.Wav2Vec2Model,lu=s.Wav2Vec2PreTrainedModel,uu=s.Wav2Vec2ProcessorWithLM,du=s.WavLMForAudioFrameClassification,cu=s.WavLMForCTC,pu=s.WavLMForSequenceClassification,hu=s.WavLMForXVector,mu=s.WavLMModel,fu=s.WavLMPreTrainedModel,gu=s.WeSpeakerFeatureExtractor,_u=s.WeSpeakerResNetModel,wu=s.WeSpeakerResNetPreTrainedModel,yu=s.WhisperFeatureExtractor,bu=s.WhisperForConditionalGeneration,vu=s.WhisperModel,xu=s.WhisperPreTrainedModel,Mu=s.WhisperProcessor,Tu=s.WhisperTextStreamer,ku=s.WhisperTokenizer,$u=s.XLMForQuestionAnswering,Cu=s.XLMForSequenceClassification,Su=s.XLMForTokenClassification,Pu=s.XLMModel,Eu=s.XLMPreTrainedModel,Fu=s.XLMRobertaForMaskedLM,Au=s.XLMRobertaForQuestionAnswering,Iu=s.XLMRobertaForSequenceClassification,zu=s.XLMRobertaForTokenClassification,Ou=s.XLMRobertaModel,Bu=s.XLMRobertaPreTrainedModel,Lu=s.XLMRobertaTokenizer,Du=s.XLMTokenizer,Ru=s.XLMWithLMHeadModel,Nu=s.XVectorOutput,Vu=s.YolosFeatureExtractor,ju=s.YolosForObjectDetection,qu=s.YolosModel,Gu=s.YolosObjectDetectionOutput,Uu=s.YolosPreTrainedModel,Wu=s.ZeroShotAudioClassificationPipeline,Hu=s.ZeroShotClassificationPipeline,Xu=s.ZeroShotImageClassificationPipeline,Ku=s.ZeroShotObjectDetectionPipeline,Qu=s.bankers_round,Yu=s.cat,Zu=s.cos_sim,Ju=s.dot,ed=s.dynamic_time_warping,td=s.env,nd=s.full,rd=s.full_like,ad=s.getKeyValueShapes,sd=s.hamming,id=s.hanning,od=s.interpolate,ld=s.interpolate_4d,ud=s.interpolate_data,dd=s.is_chinese_char,cd=s.layer_norm,pd=s.log_softmax,hd=s.magnitude,md=s.matmul,fd=s.max,gd=s.mean,_d=s.mean_pooling,wd=s.medianFilter,yd=s.mel_filter_bank,bd=s.min,vd=s.ones,xd=s.ones_like,Md=s.permute,Td=s.permute_data,kd=s.pipeline,$d=s.quantize_embeddings,Cd=s.read_audio,Sd=s.rfft,Pd=s.round,Ed=s.softmax,Fd=s.spectrogram,Ad=s.stack,Id=s.std_mean,zd=s.topk,Od=s.window_function,Bd=s.zeros,Ld=s.zeros_like;export{w as ASTFeatureExtractor,y as ASTForAudioClassification,b as ASTModel,v as ASTPreTrainedModel,x as AlbertForMaskedLM,M as AlbertForQuestionAnswering,T as AlbertForSequenceClassification,k as AlbertModel,$ as AlbertPreTrainedModel,C as AlbertTokenizer,S as AudioClassificationPipeline,P as AutoConfig,E as AutoModel,F as AutoModelForAudioClassification,A as AutoModelForAudioFrameClassification,I as AutoModelForCTC,z as AutoModelForCausalLM,O as AutoModelForDepthEstimation,B as AutoModelForDocumentQuestionAnswering,L as AutoModelForImageClassification,D as AutoModelForImageFeatureExtraction,R as AutoModelForImageMatting,N as AutoModelForImageSegmentation,V as AutoModelForImageToImage,j as AutoModelForMaskGeneration,q as AutoModelForMaskedLM,G as AutoModelForNormalEstimation,U as AutoModelForObjectDetection,W as AutoModelForQuestionAnswering,H as AutoModelForSemanticSegmentation,X as AutoModelForSeq2SeqLM,K as AutoModelForSequenceClassification,Q as AutoModelForSpeechSeq2Seq,Y as AutoModelForTextToSpectrogram,Z as AutoModelForTextToWaveform,J as AutoModelForTokenClassification,ee as AutoModelForUniversalSegmentation,te as AutoModelForVision2Seq,ne as AutoModelForXVector,re as AutoModelForZeroShotObjectDetection,ae as AutoProcessor,se as AutoTokenizer,ie as AutomaticSpeechRecognitionPipeline,oe as BartForConditionalGeneration,le as BartForSequenceClassification,ue as BartModel,de as BartPretrainedModel,ce as BartTokenizer,pe as BaseModelOutput,he as BaseStreamer,me as BeitFeatureExtractor,fe as BeitForImageClassification,ge as BeitModel,_e as BeitPreTrainedModel,we as BertForMaskedLM,ye as BertForQuestionAnswering,be as BertForSequenceClassification,ve as BertForTokenClassification,xe as BertModel,Me as BertPreTrainedModel,Te as BertTokenizer,ke as BitImageProcessor,$e as BlenderbotForConditionalGeneration,Ce as BlenderbotModel,Se as BlenderbotPreTrainedModel,Pe as BlenderbotSmallForConditionalGeneration,Ee as BlenderbotSmallModel,Fe as BlenderbotSmallPreTrainedModel,Ae as BlenderbotSmallTokenizer,Ie as BlenderbotTokenizer,ze as BloomForCausalLM,Oe as BloomModel,Be as BloomPreTrainedModel,Le as BloomTokenizer,De as CLIPFeatureExtractor,Re as CLIPImageProcessor,Ne as CLIPModel,Ve as CLIPPreTrainedModel,je as CLIPSegForImageSegmentation,qe as CLIPSegModel,Ge as CLIPSegPreTrainedModel,Ue as CLIPTextModel,We as CLIPTextModelWithProjection,He as CLIPTokenizer,Xe as CLIPVisionModel,Ke as CLIPVisionModelWithProjection,Qe as CamembertForMaskedLM,Ye as CamembertForQuestionAnswering,Ze as CamembertForSequenceClassification,Je as CamembertForTokenClassification,et as CamembertModel,tt as CamembertPreTrainedModel,nt as CamembertTokenizer,rt as CausalLMOutput,at as CausalLMOutputWithPast,st as ChineseCLIPFeatureExtractor,it as ChineseCLIPModel,ot as ChineseCLIPPreTrainedModel,lt as ClapAudioModelWithProjection,ut as ClapFeatureExtractor,dt as ClapModel,ct as ClapPreTrainedModel,pt as ClapTextModelWithProjection,ht as CodeGenForCausalLM,mt as CodeGenModel,ft as CodeGenPreTrainedModel,gt as CodeGenTokenizer,_t as CodeLlamaTokenizer,wt as CohereForCausalLM,yt as CohereModel,bt as CoherePreTrainedModel,vt as CohereTokenizer,xt as ConvBertForMaskedLM,Mt as ConvBertForQuestionAnswering,Tt as ConvBertForSequenceClassification,kt as ConvBertForTokenClassification,$t as ConvBertModel,Ct as ConvBertPreTrainedModel,St as ConvBertTokenizer,Pt as ConvNextFeatureExtractor,Et as ConvNextForImageClassification,Ft as ConvNextImageProcessor,At as ConvNextModel,It as ConvNextPreTrainedModel,zt as ConvNextV2ForImageClassification,Ot as ConvNextV2Model,Bt as ConvNextV2PreTrainedModel,Lt as DPTFeatureExtractor,Dt as DPTForDepthEstimation,Rt as DPTImageProcessor,Nt as DPTModel,Vt as DPTPreTrainedModel,jt as DebertaForMaskedLM,qt as DebertaForQuestionAnswering,Gt as DebertaForSequenceClassification,Ut as DebertaForTokenClassification,Wt as DebertaModel,Ht as DebertaPreTrainedModel,Xt as DebertaTokenizer,Kt as DebertaV2ForMaskedLM,Qt as DebertaV2ForQuestionAnswering,Yt as DebertaV2ForSequenceClassification,Zt as DebertaV2ForTokenClassification,Jt as DebertaV2Model,en as DebertaV2PreTrainedModel,tn as DebertaV2Tokenizer,nn as DecisionTransformerModel,rn as DecisionTransformerPreTrainedModel,an as DeiTFeatureExtractor,sn as DeiTForImageClassification,on as DeiTModel,ln as DeiTPreTrainedModel,un as DepthAnythingForDepthEstimation,dn as DepthAnythingPreTrainedModel,cn as DepthEstimationPipeline,pn as DepthProForDepthEstimation,hn as DepthProPreTrainedModel,mn as DetrFeatureExtractor,fn as DetrForObjectDetection,gn as DetrForSegmentation,_n as DetrModel,wn as DetrObjectDetectionOutput,yn as DetrPreTrainedModel,bn as DetrSegmentationOutput,vn as Dinov2ForImageClassification,xn as Dinov2Model,Mn as Dinov2PreTrainedModel,Tn as DistilBertForMaskedLM,kn as DistilBertForQuestionAnswering,$n as DistilBertForSequenceClassification,Cn as DistilBertForTokenClassification,Sn as DistilBertModel,Pn as DistilBertPreTrainedModel,En as DistilBertTokenizer,Fn as DocumentQuestionAnsweringPipeline,An as DonutFeatureExtractor,In as DonutImageProcessor,zn as DonutSwinModel,On as DonutSwinPreTrainedModel,Bn as EfficientNetForImageClassification,Ln as EfficientNetImageProcessor,Dn as EfficientNetModel,Rn as EfficientNetPreTrainedModel,Nn as ElectraForMaskedLM,Vn as ElectraForQuestionAnswering,jn as ElectraForSequenceClassification,qn as ElectraForTokenClassification,Gn as ElectraModel,Un as ElectraPreTrainedModel,Wn as ElectraTokenizer,Hn as EosTokenCriteria,Xn as EsmForMaskedLM,Kn as EsmForSequenceClassification,Qn as EsmForTokenClassification,Yn as EsmModel,Zn as EsmPreTrainedModel,Jn as EsmTokenizer,er as FFT,tr as FalconForCausalLM,nr as FalconModel,rr as FalconPreTrainedModel,ar as FalconTokenizer,sr as FastViTForImageClassification,ir as FastViTModel,or as FastViTPreTrainedModel,lr as FeatureExtractionPipeline,ur as FeatureExtractor,dr as FillMaskPipeline,cr as Florence2ForConditionalGeneration,pr as Florence2PreTrainedModel,hr as Florence2Processor,mr as GLPNFeatureExtractor,fr as GLPNForDepthEstimation,gr as GLPNModel,_r as GLPNPreTrainedModel,wr as GPT2LMHeadModel,yr as GPT2Model,br as GPT2PreTrainedModel,vr as GPT2Tokenizer,xr as GPTBigCodeForCausalLM,Mr as GPTBigCodeModel,Tr as GPTBigCodePreTrainedModel,kr as GPTJForCausalLM,$r as GPTJModel,Cr as GPTJPreTrainedModel,Sr as GPTNeoForCausalLM,Pr as GPTNeoModel,Er as GPTNeoPreTrainedModel,Fr as GPTNeoXForCausalLM,Ar as GPTNeoXModel,Ir as GPTNeoXPreTrainedModel,zr as GPTNeoXTokenizer,Or as Gemma2ForCausalLM,Br as Gemma2Model,Lr as Gemma2PreTrainedModel,Dr as GemmaForCausalLM,Rr as GemmaModel,Nr as GemmaPreTrainedModel,Vr as GemmaTokenizer,jr as GraniteForCausalLM,qr as GraniteModel,Gr as GranitePreTrainedModel,Ur as Grok1Tokenizer,Wr as GroupViTModel,Hr as GroupViTPreTrainedModel,Xr as HerbertTokenizer,Kr as HieraForImageClassification,Qr as HieraModel,Yr as HieraPreTrainedModel,Zr as HubertForCTC,Jr as HubertForSequenceClassification,ea as HubertModel,ta as HubertPreTrainedModel,na as ImageClassificationPipeline,ra as ImageFeatureExtractionPipeline,aa as ImageFeatureExtractor,sa as ImageMattingOutput,ia as ImageSegmentationPipeline,oa as ImageToImagePipeline,la as ImageToTextPipeline,ua as InterruptableStoppingCriteria,da as JAISLMHeadModel,ca as JAISModel,pa as JAISPreTrainedModel,ha as LlamaForCausalLM,ma as LlamaModel,fa as LlamaPreTrainedModel,ga as LlamaTokenizer,_a as LlavaForConditionalGeneration,wa as LlavaPreTrainedModel,ya as LongT5ForConditionalGeneration,ba as LongT5Model,va as LongT5PreTrainedModel,xa as M2M100ForConditionalGeneration,Ma as M2M100Model,Ta as M2M100PreTrainedModel,ka as M2M100Tokenizer,$a as MBart50Tokenizer,Ca as MBartForCausalLM,Sa as MBartForConditionalGeneration,Pa as MBartForSequenceClassification,Ea as MBartModel,Fa as MBartPreTrainedModel,Aa as MBartTokenizer,Ia as MPNetForMaskedLM,za as MPNetForQuestionAnswering,Oa as MPNetForSequenceClassification,Ba as MPNetForTokenClassification,La as MPNetModel,Da as MPNetPreTrainedModel,Ra as MPNetTokenizer,Na as MT5ForConditionalGeneration,Va as MT5Model,ja as MT5PreTrainedModel,qa as MarianMTModel,Ga as MarianModel,Ua as MarianPreTrainedModel,Wa as MarianTokenizer,Ha as MaskFormerFeatureExtractor,Xa as MaskFormerForInstanceSegmentation,Ka as MaskFormerModel,Qa as MaskFormerPreTrainedModel,Ya as MaskedLMOutput,Za as MaxLengthCriteria,Ja as MistralForCausalLM,es as MistralModel,ts as MistralPreTrainedModel,ns as MobileBertForMaskedLM,rs as MobileBertForQuestionAnswering,as as MobileBertForSequenceClassification,ss as MobileBertModel,is as MobileBertPreTrainedModel,os as MobileBertTokenizer,ls as MobileNetV1FeatureExtractor,us as MobileNetV1ForImageClassification,ds as MobileNetV1Model,cs as MobileNetV1PreTrainedModel,ps as MobileNetV2FeatureExtractor,hs as MobileNetV2ForImageClassification,ms as MobileNetV2Model,fs as MobileNetV2PreTrainedModel,gs as MobileNetV3FeatureExtractor,_s as MobileNetV3ForImageClassification,ws as MobileNetV3Model,ys as MobileNetV3PreTrainedModel,bs as MobileNetV4FeatureExtractor,vs as MobileNetV4ForImageClassification,xs as MobileNetV4Model,Ms as MobileNetV4PreTrainedModel,Ts as MobileViTFeatureExtractor,ks as MobileViTForImageClassification,$s as MobileViTImageProcessor,Cs as MobileViTModel,Ss as MobileViTPreTrainedModel,Ps as MobileViTV2ForImageClassification,Es as MobileViTV2Model,Fs as MobileViTV2PreTrainedModel,As as ModelOutput,Is as Moondream1ForConditionalGeneration,zs as MptForCausalLM,Os as MptModel,Bs as MptPreTrainedModel,Ls as MusicgenForCausalLM,Ds as MusicgenForConditionalGeneration,Rs as MusicgenModel,Ns as MusicgenPreTrainedModel,Vs as NllbTokenizer,js as NomicBertModel,qs as NomicBertPreTrainedModel,Gs as NougatImageProcessor,Us as NougatTokenizer,Ws as OPTForCausalLM,Hs as OPTModel,Xs as OPTPreTrainedModel,Ks as ObjectDetectionPipeline,Qs as OpenELMForCausalLM,Ys as OpenELMModel,Zs as OpenELMPreTrainedModel,Js as OwlViTFeatureExtractor,ei as OwlViTForObjectDetection,ti as OwlViTModel,ni as OwlViTPreTrainedModel,ri as OwlViTProcessor,ai as Owlv2ForObjectDetection,si as Owlv2ImageProcessor,ii as Owlv2Model,oi as Owlv2PreTrainedModel,li as Phi3ForCausalLM,ui as Phi3Model,di as Phi3PreTrainedModel,ci as PhiForCausalLM,pi as PhiModel,hi as PhiPreTrainedModel,mi as Pipeline,fi as PreTrainedModel,gi as PreTrainedTokenizer,_i as PretrainedConfig,wi as PretrainedMixin,yi as Processor,bi as PvtForImageClassification,vi as PvtImageProcessor,xi as PvtModel,Mi as PvtPreTrainedModel,Ti as PyAnnoteFeatureExtractor,ki as PyAnnoteForAudioFrameClassification,$i as PyAnnoteModel,Ci as PyAnnotePreTrainedModel,Si as PyAnnoteProcessor,Pi as QuestionAnsweringModelOutput,Ei as QuestionAnsweringPipeline,Fi as Qwen2ForCausalLM,Ai as Qwen2Model,Ii as Qwen2PreTrainedModel,zi as Qwen2Tokenizer,Oi as RTDetrForObjectDetection,Bi as RTDetrImageProcessor,Li as RTDetrModel,Di as RTDetrObjectDetectionOutput,Ri as RTDetrPreTrainedModel,Ni as RawImage,Vi as ResNetForImageClassification,ji as ResNetModel,qi as ResNetPreTrainedModel,Gi as RoFormerForMaskedLM,Ui as RoFormerForQuestionAnswering,Wi as RoFormerForSequenceClassification,Hi as RoFormerForTokenClassification,Xi as RoFormerModel,Ki as RoFormerPreTrainedModel,Qi as RoFormerTokenizer,Yi as RobertaForMaskedLM,Zi as RobertaForQuestionAnswering,Ji as RobertaForSequenceClassification,eo as RobertaForTokenClassification,to as RobertaModel,no as RobertaPreTrainedModel,ro as RobertaTokenizer,ao as SamImageProcessor,so as SamImageSegmentationOutput,io as SamModel,oo as SamPreTrainedModel,lo as SamProcessor,uo as SapiensFeatureExtractor,co as SapiensForDepthEstimation,po as SapiensForNormalEstimation,ho as SapiensForSemanticSegmentation,mo as SapiensPreTrainedModel,fo as SeamlessM4TFeatureExtractor,go as SegformerFeatureExtractor,_o as SegformerForImageClassification,wo as SegformerForSemanticSegmentation,yo as SegformerModel,bo as SegformerPreTrainedModel,vo as Seq2SeqLMOutput,xo as SequenceClassifierOutput,Mo as SiglipImageProcessor,To as SiglipModel,ko as SiglipPreTrainedModel,$o as SiglipTextModel,Co as SiglipTokenizer,So as SiglipVisionModel,Po as SpeechT5FeatureExtractor,Eo as SpeechT5ForSpeechToText,Fo as SpeechT5ForTextToSpeech,Ao as SpeechT5HifiGan,Io as SpeechT5Model,zo as SpeechT5PreTrainedModel,Oo as SpeechT5Processor,Bo as SpeechT5Tokenizer,Lo as SqueezeBertForMaskedLM,Do as SqueezeBertForQuestionAnswering,Ro as SqueezeBertForSequenceClassification,No as SqueezeBertModel,Vo as SqueezeBertPreTrainedModel,jo as SqueezeBertTokenizer,qo as StableLmForCausalLM,Go as StableLmModel,Uo as StableLmPreTrainedModel,Wo as Starcoder2ForCausalLM,Ho as Starcoder2Model,Xo as Starcoder2PreTrainedModel,Ko as StoppingCriteria,Qo as StoppingCriteriaList,Yo as SummarizationPipeline,Zo as Swin2SRForImageSuperResolution,Jo as Swin2SRImageProcessor,el as Swin2SRModel,tl as Swin2SRPreTrainedModel,nl as SwinForImageClassification,rl as SwinModel,al as SwinPreTrainedModel,sl as T5ForConditionalGeneration,il as T5Model,ol as T5PreTrainedModel,ll as T5Tokenizer,ul as TableTransformerForObjectDetection,dl as TableTransformerModel,cl as TableTransformerObjectDetectionOutput,pl as TableTransformerPreTrainedModel,hl as Tensor,ml as Text2TextGenerationPipeline,fl as TextClassificationPipeline,gl as TextGenerationPipeline,_l as TextStreamer,wl as TextToAudioPipeline,yl as TokenClassificationPipeline,bl as TokenClassifierOutput,vl as TokenizerModel,xl as TrOCRForCausalLM,Ml as TrOCRPreTrainedModel,Tl as TranslationPipeline,kl as UniSpeechForCTC,$l as UniSpeechForSequenceClassification,Cl as UniSpeechModel,Sl as UniSpeechPreTrainedModel,Pl as UniSpeechSatForAudioFrameClassification,El as UniSpeechSatForCTC,Fl as UniSpeechSatForSequenceClassification,Al as UniSpeechSatModel,Il as UniSpeechSatPreTrainedModel,zl as ViTFeatureExtractor,Ol as ViTForImageClassification,Bl as ViTImageProcessor,Ll as ViTMAEModel,Dl as ViTMAEPreTrainedModel,Rl as ViTMSNForImageClassification,Nl as ViTMSNModel,Vl as ViTMSNPreTrainedModel,jl as ViTModel,ql as ViTPreTrainedModel,Gl as VisionEncoderDecoderModel,Ul as VitMatteForImageMatting,Wl as VitMatteImageProcessor,Hl as VitMattePreTrainedModel,Xl as VitsModel,Kl as VitsModelOutput,Ql as VitsPreTrainedModel,Yl as VitsTokenizer,Zl as Wav2Vec2BertForCTC,Jl as Wav2Vec2BertForSequenceClassification,eu as Wav2Vec2BertModel,tu as Wav2Vec2BertPreTrainedModel,nu as Wav2Vec2CTCTokenizer,ru as Wav2Vec2FeatureExtractor,au as Wav2Vec2ForAudioFrameClassification,su as Wav2Vec2ForCTC,iu as Wav2Vec2ForSequenceClassification,ou as Wav2Vec2Model,lu as Wav2Vec2PreTrainedModel,uu as Wav2Vec2ProcessorWithLM,du as WavLMForAudioFrameClassification,cu as WavLMForCTC,pu as WavLMForSequenceClassification,hu as WavLMForXVector,mu as WavLMModel,fu as WavLMPreTrainedModel,gu as WeSpeakerFeatureExtractor,_u as WeSpeakerResNetModel,wu as WeSpeakerResNetPreTrainedModel,yu as WhisperFeatureExtractor,bu as WhisperForConditionalGeneration,vu as WhisperModel,xu as WhisperPreTrainedModel,Mu as WhisperProcessor,Tu as WhisperTextStreamer,ku as WhisperTokenizer,$u as XLMForQuestionAnswering,Cu as XLMForSequenceClassification,Su as XLMForTokenClassification,Pu as XLMModel,Eu as XLMPreTrainedModel,Fu as XLMRobertaForMaskedLM,Au as XLMRobertaForQuestionAnswering,Iu as XLMRobertaForSequenceClassification,zu as XLMRobertaForTokenClassification,Ou as XLMRobertaModel,Bu as XLMRobertaPreTrainedModel,Lu as XLMRobertaTokenizer,Du as XLMTokenizer,Ru as XLMWithLMHeadModel,Nu as XVectorOutput,Vu as YolosFeatureExtractor,ju as YolosForObjectDetection,qu as YolosModel,Gu as YolosObjectDetectionOutput,Uu as YolosPreTrainedModel,Wu as ZeroShotAudioClassificationPipeline,Hu as ZeroShotClassificationPipeline,Xu as ZeroShotImageClassificationPipeline,Ku as ZeroShotObjectDetectionPipeline,Qu as bankers_round,Yu as cat,Zu as cos_sim,Ju as dot,ed as dynamic_time_warping,td as env,nd as full,rd as full_like,ad as getKeyValueShapes,sd as hamming,id as hanning,od as interpolate,ld as interpolate_4d,ud as interpolate_data,dd as is_chinese_char,cd as layer_norm,pd as log_softmax,hd as magnitude,md as matmul,fd as max,gd as mean,_d as mean_pooling,wd as medianFilter,yd as mel_filter_bank,bd as min,vd as ones,xd as ones_like,Md as permute,Td as permute_data,kd as pipeline,$d as quantize_embeddings,Cd as read_audio,Sd as rfft,Pd as round,Ed as softmax,Fd as spectrogram,Ad as stack,Id as std_mean,zd as topk,Od as window_function,Bd as zeros,Ld as zeros_like};
+  \*****************************/a.r(s),a.d(s,{ASTFeatureExtractor:()=>d.ASTFeatureExtractor,ASTForAudioClassification:()=>l.ASTForAudioClassification,ASTModel:()=>l.ASTModel,ASTPreTrainedModel:()=>l.ASTPreTrainedModel,AlbertForMaskedLM:()=>l.AlbertForMaskedLM,AlbertForQuestionAnswering:()=>l.AlbertForQuestionAnswering,AlbertForSequenceClassification:()=>l.AlbertForSequenceClassification,AlbertModel:()=>l.AlbertModel,AlbertPreTrainedModel:()=>l.AlbertPreTrainedModel,AlbertTokenizer:()=>u.AlbertTokenizer,AudioClassificationPipeline:()=>o.AudioClassificationPipeline,AutoConfig:()=>c.AutoConfig,AutoModel:()=>l.AutoModel,AutoModelForAudioClassification:()=>l.AutoModelForAudioClassification,AutoModelForAudioFrameClassification:()=>l.AutoModelForAudioFrameClassification,AutoModelForCTC:()=>l.AutoModelForCTC,AutoModelForCausalLM:()=>l.AutoModelForCausalLM,AutoModelForDepthEstimation:()=>l.AutoModelForDepthEstimation,AutoModelForDocumentQuestionAnswering:()=>l.AutoModelForDocumentQuestionAnswering,AutoModelForImageClassification:()=>l.AutoModelForImageClassification,AutoModelForImageFeatureExtraction:()=>l.AutoModelForImageFeatureExtraction,AutoModelForImageMatting:()=>l.AutoModelForImageMatting,AutoModelForImageSegmentation:()=>l.AutoModelForImageSegmentation,AutoModelForImageToImage:()=>l.AutoModelForImageToImage,AutoModelForMaskGeneration:()=>l.AutoModelForMaskGeneration,AutoModelForMaskedLM:()=>l.AutoModelForMaskedLM,AutoModelForNormalEstimation:()=>l.AutoModelForNormalEstimation,AutoModelForObjectDetection:()=>l.AutoModelForObjectDetection,AutoModelForQuestionAnswering:()=>l.AutoModelForQuestionAnswering,AutoModelForSemanticSegmentation:()=>l.AutoModelForSemanticSegmentation,AutoModelForSeq2SeqLM:()=>l.AutoModelForSeq2SeqLM,AutoModelForSequenceClassification:()=>l.AutoModelForSequenceClassification,AutoModelForSpeechSeq2Seq:()=>l.AutoModelForSpeechSeq2Seq,AutoModelForTextToSpectrogram:()=>l.AutoModelForTextToSpectrogram,AutoModelForTextToWaveform:()=>l.AutoModelForTextToWaveform,AutoModelForTokenClassification:()=>l.AutoModelForTokenClassification,AutoModelForUniversalSegmentation:()=>l.AutoModelForUniversalSegmentation,AutoModelForVision2Seq:()=>l.AutoModelForVision2Seq,AutoModelForXVector:()=>l.AutoModelForXVector,AutoModelForZeroShotObjectDetection:()=>l.AutoModelForZeroShotObjectDetection,AutoProcessor:()=>d.AutoProcessor,AutoTokenizer:()=>u.AutoTokenizer,AutomaticSpeechRecognitionPipeline:()=>o.AutomaticSpeechRecognitionPipeline,BartForConditionalGeneration:()=>l.BartForConditionalGeneration,BartForSequenceClassification:()=>l.BartForSequenceClassification,BartModel:()=>l.BartModel,BartPretrainedModel:()=>l.BartPretrainedModel,BartTokenizer:()=>u.BartTokenizer,BaseModelOutput:()=>l.BaseModelOutput,BaseStreamer:()=>g.BaseStreamer,BeitFeatureExtractor:()=>d.BeitFeatureExtractor,BeitForImageClassification:()=>l.BeitForImageClassification,BeitModel:()=>l.BeitModel,BeitPreTrainedModel:()=>l.BeitPreTrainedModel,BertForMaskedLM:()=>l.BertForMaskedLM,BertForQuestionAnswering:()=>l.BertForQuestionAnswering,BertForSequenceClassification:()=>l.BertForSequenceClassification,BertForTokenClassification:()=>l.BertForTokenClassification,BertModel:()=>l.BertModel,BertPreTrainedModel:()=>l.BertPreTrainedModel,BertTokenizer:()=>u.BertTokenizer,BitImageProcessor:()=>d.BitImageProcessor,BlenderbotForConditionalGeneration:()=>l.BlenderbotForConditionalGeneration,BlenderbotModel:()=>l.BlenderbotModel,BlenderbotPreTrainedModel:()=>l.BlenderbotPreTrainedModel,BlenderbotSmallForConditionalGeneration:()=>l.BlenderbotSmallForConditionalGeneration,BlenderbotSmallModel:()=>l.BlenderbotSmallModel,BlenderbotSmallPreTrainedModel:()=>l.BlenderbotSmallPreTrainedModel,BlenderbotSmallTokenizer:()=>u.BlenderbotSmallTokenizer,BlenderbotTokenizer:()=>u.BlenderbotTokenizer,BloomForCausalLM:()=>l.BloomForCausalLM,BloomModel:()=>l.BloomModel,BloomPreTrainedModel:()=>l.BloomPreTrainedModel,BloomTokenizer:()=>u.BloomTokenizer,CLIPFeatureExtractor:()=>d.CLIPFeatureExtractor,CLIPImageProcessor:()=>d.CLIPImageProcessor,CLIPModel:()=>l.CLIPModel,CLIPPreTrainedModel:()=>l.CLIPPreTrainedModel,CLIPSegForImageSegmentation:()=>l.CLIPSegForImageSegmentation,CLIPSegModel:()=>l.CLIPSegModel,CLIPSegPreTrainedModel:()=>l.CLIPSegPreTrainedModel,CLIPTextModel:()=>l.CLIPTextModel,CLIPTextModelWithProjection:()=>l.CLIPTextModelWithProjection,CLIPTokenizer:()=>u.CLIPTokenizer,CLIPVisionModel:()=>l.CLIPVisionModel,CLIPVisionModelWithProjection:()=>l.CLIPVisionModelWithProjection,CamembertForMaskedLM:()=>l.CamembertForMaskedLM,CamembertForQuestionAnswering:()=>l.CamembertForQuestionAnswering,CamembertForSequenceClassification:()=>l.CamembertForSequenceClassification,CamembertForTokenClassification:()=>l.CamembertForTokenClassification,CamembertModel:()=>l.CamembertModel,CamembertPreTrainedModel:()=>l.CamembertPreTrainedModel,CamembertTokenizer:()=>u.CamembertTokenizer,CausalLMOutput:()=>l.CausalLMOutput,CausalLMOutputWithPast:()=>l.CausalLMOutputWithPast,ChineseCLIPFeatureExtractor:()=>d.ChineseCLIPFeatureExtractor,ChineseCLIPModel:()=>l.ChineseCLIPModel,ChineseCLIPPreTrainedModel:()=>l.ChineseCLIPPreTrainedModel,ClapAudioModelWithProjection:()=>l.ClapAudioModelWithProjection,ClapFeatureExtractor:()=>d.ClapFeatureExtractor,ClapModel:()=>l.ClapModel,ClapPreTrainedModel:()=>l.ClapPreTrainedModel,ClapTextModelWithProjection:()=>l.ClapTextModelWithProjection,CodeGenForCausalLM:()=>l.CodeGenForCausalLM,CodeGenModel:()=>l.CodeGenModel,CodeGenPreTrainedModel:()=>l.CodeGenPreTrainedModel,CodeGenTokenizer:()=>u.CodeGenTokenizer,CodeLlamaTokenizer:()=>u.CodeLlamaTokenizer,CohereForCausalLM:()=>l.CohereForCausalLM,CohereModel:()=>l.CohereModel,CoherePreTrainedModel:()=>l.CoherePreTrainedModel,CohereTokenizer:()=>u.CohereTokenizer,ConvBertForMaskedLM:()=>l.ConvBertForMaskedLM,ConvBertForQuestionAnswering:()=>l.ConvBertForQuestionAnswering,ConvBertForSequenceClassification:()=>l.ConvBertForSequenceClassification,ConvBertForTokenClassification:()=>l.ConvBertForTokenClassification,ConvBertModel:()=>l.ConvBertModel,ConvBertPreTrainedModel:()=>l.ConvBertPreTrainedModel,ConvBertTokenizer:()=>u.ConvBertTokenizer,ConvNextFeatureExtractor:()=>d.ConvNextFeatureExtractor,ConvNextForImageClassification:()=>l.ConvNextForImageClassification,ConvNextImageProcessor:()=>d.ConvNextImageProcessor,ConvNextModel:()=>l.ConvNextModel,ConvNextPreTrainedModel:()=>l.ConvNextPreTrainedModel,ConvNextV2ForImageClassification:()=>l.ConvNextV2ForImageClassification,ConvNextV2Model:()=>l.ConvNextV2Model,ConvNextV2PreTrainedModel:()=>l.ConvNextV2PreTrainedModel,DPTFeatureExtractor:()=>d.DPTFeatureExtractor,DPTForDepthEstimation:()=>l.DPTForDepthEstimation,DPTImageProcessor:()=>d.DPTImageProcessor,DPTModel:()=>l.DPTModel,DPTPreTrainedModel:()=>l.DPTPreTrainedModel,DebertaForMaskedLM:()=>l.DebertaForMaskedLM,DebertaForQuestionAnswering:()=>l.DebertaForQuestionAnswering,DebertaForSequenceClassification:()=>l.DebertaForSequenceClassification,DebertaForTokenClassification:()=>l.DebertaForTokenClassification,DebertaModel:()=>l.DebertaModel,DebertaPreTrainedModel:()=>l.DebertaPreTrainedModel,DebertaTokenizer:()=>u.DebertaTokenizer,DebertaV2ForMaskedLM:()=>l.DebertaV2ForMaskedLM,DebertaV2ForQuestionAnswering:()=>l.DebertaV2ForQuestionAnswering,DebertaV2ForSequenceClassification:()=>l.DebertaV2ForSequenceClassification,DebertaV2ForTokenClassification:()=>l.DebertaV2ForTokenClassification,DebertaV2Model:()=>l.DebertaV2Model,DebertaV2PreTrainedModel:()=>l.DebertaV2PreTrainedModel,DebertaV2Tokenizer:()=>u.DebertaV2Tokenizer,DecisionTransformerModel:()=>l.DecisionTransformerModel,DecisionTransformerPreTrainedModel:()=>l.DecisionTransformerPreTrainedModel,DeiTFeatureExtractor:()=>d.DeiTFeatureExtractor,DeiTForImageClassification:()=>l.DeiTForImageClassification,DeiTModel:()=>l.DeiTModel,DeiTPreTrainedModel:()=>l.DeiTPreTrainedModel,DepthAnythingForDepthEstimation:()=>l.DepthAnythingForDepthEstimation,DepthAnythingPreTrainedModel:()=>l.DepthAnythingPreTrainedModel,DepthEstimationPipeline:()=>o.DepthEstimationPipeline,DepthProForDepthEstimation:()=>l.DepthProForDepthEstimation,DepthProPreTrainedModel:()=>l.DepthProPreTrainedModel,DetrFeatureExtractor:()=>d.DetrFeatureExtractor,DetrForObjectDetection:()=>l.DetrForObjectDetection,DetrForSegmentation:()=>l.DetrForSegmentation,DetrModel:()=>l.DetrModel,DetrObjectDetectionOutput:()=>l.DetrObjectDetectionOutput,DetrPreTrainedModel:()=>l.DetrPreTrainedModel,DetrSegmentationOutput:()=>l.DetrSegmentationOutput,Dinov2ForImageClassification:()=>l.Dinov2ForImageClassification,Dinov2Model:()=>l.Dinov2Model,Dinov2PreTrainedModel:()=>l.Dinov2PreTrainedModel,DistilBertForMaskedLM:()=>l.DistilBertForMaskedLM,DistilBertForQuestionAnswering:()=>l.DistilBertForQuestionAnswering,DistilBertForSequenceClassification:()=>l.DistilBertForSequenceClassification,DistilBertForTokenClassification:()=>l.DistilBertForTokenClassification,DistilBertModel:()=>l.DistilBertModel,DistilBertPreTrainedModel:()=>l.DistilBertPreTrainedModel,DistilBertTokenizer:()=>u.DistilBertTokenizer,DocumentQuestionAnsweringPipeline:()=>o.DocumentQuestionAnsweringPipeline,DonutFeatureExtractor:()=>d.DonutFeatureExtractor,DonutImageProcessor:()=>d.DonutImageProcessor,DonutSwinModel:()=>l.DonutSwinModel,DonutSwinPreTrainedModel:()=>l.DonutSwinPreTrainedModel,EfficientNetForImageClassification:()=>l.EfficientNetForImageClassification,EfficientNetImageProcessor:()=>d.EfficientNetImageProcessor,EfficientNetModel:()=>l.EfficientNetModel,EfficientNetPreTrainedModel:()=>l.EfficientNetPreTrainedModel,ElectraForMaskedLM:()=>l.ElectraForMaskedLM,ElectraForQuestionAnswering:()=>l.ElectraForQuestionAnswering,ElectraForSequenceClassification:()=>l.ElectraForSequenceClassification,ElectraForTokenClassification:()=>l.ElectraForTokenClassification,ElectraModel:()=>l.ElectraModel,ElectraPreTrainedModel:()=>l.ElectraPreTrainedModel,ElectraTokenizer:()=>u.ElectraTokenizer,EosTokenCriteria:()=>_.EosTokenCriteria,EsmForMaskedLM:()=>l.EsmForMaskedLM,EsmForSequenceClassification:()=>l.EsmForSequenceClassification,EsmForTokenClassification:()=>l.EsmForTokenClassification,EsmModel:()=>l.EsmModel,EsmPreTrainedModel:()=>l.EsmPreTrainedModel,EsmTokenizer:()=>u.EsmTokenizer,FFT:()=>f.FFT,FalconForCausalLM:()=>l.FalconForCausalLM,FalconModel:()=>l.FalconModel,FalconPreTrainedModel:()=>l.FalconPreTrainedModel,FalconTokenizer:()=>u.FalconTokenizer,FastViTForImageClassification:()=>l.FastViTForImageClassification,FastViTModel:()=>l.FastViTModel,FastViTPreTrainedModel:()=>l.FastViTPreTrainedModel,FeatureExtractionPipeline:()=>o.FeatureExtractionPipeline,FeatureExtractor:()=>d.FeatureExtractor,FillMaskPipeline:()=>o.FillMaskPipeline,Florence2ForConditionalGeneration:()=>l.Florence2ForConditionalGeneration,Florence2PreTrainedModel:()=>l.Florence2PreTrainedModel,Florence2Processor:()=>d.Florence2Processor,GLPNFeatureExtractor:()=>d.GLPNFeatureExtractor,GLPNForDepthEstimation:()=>l.GLPNForDepthEstimation,GLPNModel:()=>l.GLPNModel,GLPNPreTrainedModel:()=>l.GLPNPreTrainedModel,GPT2LMHeadModel:()=>l.GPT2LMHeadModel,GPT2Model:()=>l.GPT2Model,GPT2PreTrainedModel:()=>l.GPT2PreTrainedModel,GPT2Tokenizer:()=>u.GPT2Tokenizer,GPTBigCodeForCausalLM:()=>l.GPTBigCodeForCausalLM,GPTBigCodeModel:()=>l.GPTBigCodeModel,GPTBigCodePreTrainedModel:()=>l.GPTBigCodePreTrainedModel,GPTJForCausalLM:()=>l.GPTJForCausalLM,GPTJModel:()=>l.GPTJModel,GPTJPreTrainedModel:()=>l.GPTJPreTrainedModel,GPTNeoForCausalLM:()=>l.GPTNeoForCausalLM,GPTNeoModel:()=>l.GPTNeoModel,GPTNeoPreTrainedModel:()=>l.GPTNeoPreTrainedModel,GPTNeoXForCausalLM:()=>l.GPTNeoXForCausalLM,GPTNeoXModel:()=>l.GPTNeoXModel,GPTNeoXPreTrainedModel:()=>l.GPTNeoXPreTrainedModel,GPTNeoXTokenizer:()=>u.GPTNeoXTokenizer,Gemma2ForCausalLM:()=>l.Gemma2ForCausalLM,Gemma2Model:()=>l.Gemma2Model,Gemma2PreTrainedModel:()=>l.Gemma2PreTrainedModel,GemmaForCausalLM:()=>l.GemmaForCausalLM,GemmaModel:()=>l.GemmaModel,GemmaPreTrainedModel:()=>l.GemmaPreTrainedModel,GemmaTokenizer:()=>u.GemmaTokenizer,GraniteForCausalLM:()=>l.GraniteForCausalLM,GraniteModel:()=>l.GraniteModel,GranitePreTrainedModel:()=>l.GranitePreTrainedModel,Grok1Tokenizer:()=>u.Grok1Tokenizer,GroupViTModel:()=>l.GroupViTModel,GroupViTPreTrainedModel:()=>l.GroupViTPreTrainedModel,HerbertTokenizer:()=>u.HerbertTokenizer,HieraForImageClassification:()=>l.HieraForImageClassification,HieraModel:()=>l.HieraModel,HieraPreTrainedModel:()=>l.HieraPreTrainedModel,HubertForCTC:()=>l.HubertForCTC,HubertForSequenceClassification:()=>l.HubertForSequenceClassification,HubertModel:()=>l.HubertModel,HubertPreTrainedModel:()=>l.HubertPreTrainedModel,ImageClassificationPipeline:()=>o.ImageClassificationPipeline,ImageFeatureExtractionPipeline:()=>o.ImageFeatureExtractionPipeline,ImageFeatureExtractor:()=>d.ImageFeatureExtractor,ImageMattingOutput:()=>l.ImageMattingOutput,ImageSegmentationPipeline:()=>o.ImageSegmentationPipeline,ImageToImagePipeline:()=>o.ImageToImagePipeline,ImageToTextPipeline:()=>o.ImageToTextPipeline,InterruptableStoppingCriteria:()=>_.InterruptableStoppingCriteria,JAISLMHeadModel:()=>l.JAISLMHeadModel,JAISModel:()=>l.JAISModel,JAISPreTrainedModel:()=>l.JAISPreTrainedModel,LlamaForCausalLM:()=>l.LlamaForCausalLM,LlamaModel:()=>l.LlamaModel,LlamaPreTrainedModel:()=>l.LlamaPreTrainedModel,LlamaTokenizer:()=>u.LlamaTokenizer,LlavaForConditionalGeneration:()=>l.LlavaForConditionalGeneration,LlavaPreTrainedModel:()=>l.LlavaPreTrainedModel,LongT5ForConditionalGeneration:()=>l.LongT5ForConditionalGeneration,LongT5Model:()=>l.LongT5Model,LongT5PreTrainedModel:()=>l.LongT5PreTrainedModel,M2M100ForConditionalGeneration:()=>l.M2M100ForConditionalGeneration,M2M100Model:()=>l.M2M100Model,M2M100PreTrainedModel:()=>l.M2M100PreTrainedModel,M2M100Tokenizer:()=>u.M2M100Tokenizer,MBart50Tokenizer:()=>u.MBart50Tokenizer,MBartForCausalLM:()=>l.MBartForCausalLM,MBartForConditionalGeneration:()=>l.MBartForConditionalGeneration,MBartForSequenceClassification:()=>l.MBartForSequenceClassification,MBartModel:()=>l.MBartModel,MBartPreTrainedModel:()=>l.MBartPreTrainedModel,MBartTokenizer:()=>u.MBartTokenizer,MPNetForMaskedLM:()=>l.MPNetForMaskedLM,MPNetForQuestionAnswering:()=>l.MPNetForQuestionAnswering,MPNetForSequenceClassification:()=>l.MPNetForSequenceClassification,MPNetForTokenClassification:()=>l.MPNetForTokenClassification,MPNetModel:()=>l.MPNetModel,MPNetPreTrainedModel:()=>l.MPNetPreTrainedModel,MPNetTokenizer:()=>u.MPNetTokenizer,MT5ForConditionalGeneration:()=>l.MT5ForConditionalGeneration,MT5Model:()=>l.MT5Model,MT5PreTrainedModel:()=>l.MT5PreTrainedModel,MarianMTModel:()=>l.MarianMTModel,MarianModel:()=>l.MarianModel,MarianPreTrainedModel:()=>l.MarianPreTrainedModel,MarianTokenizer:()=>u.MarianTokenizer,MaskFormerFeatureExtractor:()=>d.MaskFormerFeatureExtractor,MaskFormerForInstanceSegmentation:()=>l.MaskFormerForInstanceSegmentation,MaskFormerModel:()=>l.MaskFormerModel,MaskFormerPreTrainedModel:()=>l.MaskFormerPreTrainedModel,MaskedLMOutput:()=>l.MaskedLMOutput,MaxLengthCriteria:()=>_.MaxLengthCriteria,MistralForCausalLM:()=>l.MistralForCausalLM,MistralModel:()=>l.MistralModel,MistralPreTrainedModel:()=>l.MistralPreTrainedModel,MobileBertForMaskedLM:()=>l.MobileBertForMaskedLM,MobileBertForQuestionAnswering:()=>l.MobileBertForQuestionAnswering,MobileBertForSequenceClassification:()=>l.MobileBertForSequenceClassification,MobileBertModel:()=>l.MobileBertModel,MobileBertPreTrainedModel:()=>l.MobileBertPreTrainedModel,MobileBertTokenizer:()=>u.MobileBertTokenizer,MobileLLMForCausalLM:()=>l.MobileLLMForCausalLM,MobileLLMModel:()=>l.MobileLLMModel,MobileLLMPreTrainedModel:()=>l.MobileLLMPreTrainedModel,MobileNetV1FeatureExtractor:()=>d.MobileNetV1FeatureExtractor,MobileNetV1ForImageClassification:()=>l.MobileNetV1ForImageClassification,MobileNetV1Model:()=>l.MobileNetV1Model,MobileNetV1PreTrainedModel:()=>l.MobileNetV1PreTrainedModel,MobileNetV2FeatureExtractor:()=>d.MobileNetV2FeatureExtractor,MobileNetV2ForImageClassification:()=>l.MobileNetV2ForImageClassification,MobileNetV2Model:()=>l.MobileNetV2Model,MobileNetV2PreTrainedModel:()=>l.MobileNetV2PreTrainedModel,MobileNetV3FeatureExtractor:()=>d.MobileNetV3FeatureExtractor,MobileNetV3ForImageClassification:()=>l.MobileNetV3ForImageClassification,MobileNetV3Model:()=>l.MobileNetV3Model,MobileNetV3PreTrainedModel:()=>l.MobileNetV3PreTrainedModel,MobileNetV4FeatureExtractor:()=>d.MobileNetV4FeatureExtractor,MobileNetV4ForImageClassification:()=>l.MobileNetV4ForImageClassification,MobileNetV4Model:()=>l.MobileNetV4Model,MobileNetV4PreTrainedModel:()=>l.MobileNetV4PreTrainedModel,MobileViTFeatureExtractor:()=>d.MobileViTFeatureExtractor,MobileViTForImageClassification:()=>l.MobileViTForImageClassification,MobileViTImageProcessor:()=>d.MobileViTImageProcessor,MobileViTModel:()=>l.MobileViTModel,MobileViTPreTrainedModel:()=>l.MobileViTPreTrainedModel,MobileViTV2ForImageClassification:()=>l.MobileViTV2ForImageClassification,MobileViTV2Model:()=>l.MobileViTV2Model,MobileViTV2PreTrainedModel:()=>l.MobileViTV2PreTrainedModel,ModelOutput:()=>l.ModelOutput,Moondream1ForConditionalGeneration:()=>l.Moondream1ForConditionalGeneration,MptForCausalLM:()=>l.MptForCausalLM,MptModel:()=>l.MptModel,MptPreTrainedModel:()=>l.MptPreTrainedModel,MusicgenForCausalLM:()=>l.MusicgenForCausalLM,MusicgenForConditionalGeneration:()=>l.MusicgenForConditionalGeneration,MusicgenModel:()=>l.MusicgenModel,MusicgenPreTrainedModel:()=>l.MusicgenPreTrainedModel,NllbTokenizer:()=>u.NllbTokenizer,NomicBertModel:()=>l.NomicBertModel,NomicBertPreTrainedModel:()=>l.NomicBertPreTrainedModel,NougatImageProcessor:()=>d.NougatImageProcessor,NougatTokenizer:()=>u.NougatTokenizer,OPTForCausalLM:()=>l.OPTForCausalLM,OPTModel:()=>l.OPTModel,OPTPreTrainedModel:()=>l.OPTPreTrainedModel,ObjectDetectionPipeline:()=>o.ObjectDetectionPipeline,OlmoForCausalLM:()=>l.OlmoForCausalLM,OlmoModel:()=>l.OlmoModel,OlmoPreTrainedModel:()=>l.OlmoPreTrainedModel,OpenELMForCausalLM:()=>l.OpenELMForCausalLM,OpenELMModel:()=>l.OpenELMModel,OpenELMPreTrainedModel:()=>l.OpenELMPreTrainedModel,OwlViTFeatureExtractor:()=>d.OwlViTFeatureExtractor,OwlViTForObjectDetection:()=>l.OwlViTForObjectDetection,OwlViTModel:()=>l.OwlViTModel,OwlViTPreTrainedModel:()=>l.OwlViTPreTrainedModel,OwlViTProcessor:()=>d.OwlViTProcessor,Owlv2ForObjectDetection:()=>l.Owlv2ForObjectDetection,Owlv2ImageProcessor:()=>d.Owlv2ImageProcessor,Owlv2Model:()=>l.Owlv2Model,Owlv2PreTrainedModel:()=>l.Owlv2PreTrainedModel,Phi3ForCausalLM:()=>l.Phi3ForCausalLM,Phi3Model:()=>l.Phi3Model,Phi3PreTrainedModel:()=>l.Phi3PreTrainedModel,PhiForCausalLM:()=>l.PhiForCausalLM,PhiModel:()=>l.PhiModel,PhiPreTrainedModel:()=>l.PhiPreTrainedModel,Pipeline:()=>o.Pipeline,PreTrainedModel:()=>l.PreTrainedModel,PreTrainedTokenizer:()=>u.PreTrainedTokenizer,PretrainedConfig:()=>c.PretrainedConfig,PretrainedMixin:()=>l.PretrainedMixin,Processor:()=>d.Processor,PvtForImageClassification:()=>l.PvtForImageClassification,PvtImageProcessor:()=>d.PvtImageProcessor,PvtModel:()=>l.PvtModel,PvtPreTrainedModel:()=>l.PvtPreTrainedModel,PyAnnoteFeatureExtractor:()=>d.PyAnnoteFeatureExtractor,PyAnnoteForAudioFrameClassification:()=>l.PyAnnoteForAudioFrameClassification,PyAnnoteModel:()=>l.PyAnnoteModel,PyAnnotePreTrainedModel:()=>l.PyAnnotePreTrainedModel,PyAnnoteProcessor:()=>d.PyAnnoteProcessor,QuestionAnsweringModelOutput:()=>l.QuestionAnsweringModelOutput,QuestionAnsweringPipeline:()=>o.QuestionAnsweringPipeline,Qwen2ForCausalLM:()=>l.Qwen2ForCausalLM,Qwen2Model:()=>l.Qwen2Model,Qwen2PreTrainedModel:()=>l.Qwen2PreTrainedModel,Qwen2Tokenizer:()=>u.Qwen2Tokenizer,RTDetrForObjectDetection:()=>l.RTDetrForObjectDetection,RTDetrImageProcessor:()=>d.RTDetrImageProcessor,RTDetrModel:()=>l.RTDetrModel,RTDetrObjectDetectionOutput:()=>l.RTDetrObjectDetectionOutput,RTDetrPreTrainedModel:()=>l.RTDetrPreTrainedModel,RawImage:()=>h.RawImage,ResNetForImageClassification:()=>l.ResNetForImageClassification,ResNetModel:()=>l.ResNetModel,ResNetPreTrainedModel:()=>l.ResNetPreTrainedModel,RoFormerForMaskedLM:()=>l.RoFormerForMaskedLM,RoFormerForQuestionAnswering:()=>l.RoFormerForQuestionAnswering,RoFormerForSequenceClassification:()=>l.RoFormerForSequenceClassification,RoFormerForTokenClassification:()=>l.RoFormerForTokenClassification,RoFormerModel:()=>l.RoFormerModel,RoFormerPreTrainedModel:()=>l.RoFormerPreTrainedModel,RoFormerTokenizer:()=>u.RoFormerTokenizer,RobertaForMaskedLM:()=>l.RobertaForMaskedLM,RobertaForQuestionAnswering:()=>l.RobertaForQuestionAnswering,RobertaForSequenceClassification:()=>l.RobertaForSequenceClassification,RobertaForTokenClassification:()=>l.RobertaForTokenClassification,RobertaModel:()=>l.RobertaModel,RobertaPreTrainedModel:()=>l.RobertaPreTrainedModel,RobertaTokenizer:()=>u.RobertaTokenizer,SamImageProcessor:()=>d.SamImageProcessor,SamImageSegmentationOutput:()=>l.SamImageSegmentationOutput,SamModel:()=>l.SamModel,SamPreTrainedModel:()=>l.SamPreTrainedModel,SamProcessor:()=>d.SamProcessor,SapiensFeatureExtractor:()=>d.SapiensFeatureExtractor,SapiensForDepthEstimation:()=>l.SapiensForDepthEstimation,SapiensForNormalEstimation:()=>l.SapiensForNormalEstimation,SapiensForSemanticSegmentation:()=>l.SapiensForSemanticSegmentation,SapiensPreTrainedModel:()=>l.SapiensPreTrainedModel,SeamlessM4TFeatureExtractor:()=>d.SeamlessM4TFeatureExtractor,SegformerFeatureExtractor:()=>d.SegformerFeatureExtractor,SegformerForImageClassification:()=>l.SegformerForImageClassification,SegformerForSemanticSegmentation:()=>l.SegformerForSemanticSegmentation,SegformerModel:()=>l.SegformerModel,SegformerPreTrainedModel:()=>l.SegformerPreTrainedModel,Seq2SeqLMOutput:()=>l.Seq2SeqLMOutput,SequenceClassifierOutput:()=>l.SequenceClassifierOutput,SiglipImageProcessor:()=>d.SiglipImageProcessor,SiglipModel:()=>l.SiglipModel,SiglipPreTrainedModel:()=>l.SiglipPreTrainedModel,SiglipTextModel:()=>l.SiglipTextModel,SiglipTokenizer:()=>u.SiglipTokenizer,SiglipVisionModel:()=>l.SiglipVisionModel,SpeechT5FeatureExtractor:()=>d.SpeechT5FeatureExtractor,SpeechT5ForSpeechToText:()=>l.SpeechT5ForSpeechToText,SpeechT5ForTextToSpeech:()=>l.SpeechT5ForTextToSpeech,SpeechT5HifiGan:()=>l.SpeechT5HifiGan,SpeechT5Model:()=>l.SpeechT5Model,SpeechT5PreTrainedModel:()=>l.SpeechT5PreTrainedModel,SpeechT5Processor:()=>d.SpeechT5Processor,SpeechT5Tokenizer:()=>u.SpeechT5Tokenizer,SqueezeBertForMaskedLM:()=>l.SqueezeBertForMaskedLM,SqueezeBertForQuestionAnswering:()=>l.SqueezeBertForQuestionAnswering,SqueezeBertForSequenceClassification:()=>l.SqueezeBertForSequenceClassification,SqueezeBertModel:()=>l.SqueezeBertModel,SqueezeBertPreTrainedModel:()=>l.SqueezeBertPreTrainedModel,SqueezeBertTokenizer:()=>u.SqueezeBertTokenizer,StableLmForCausalLM:()=>l.StableLmForCausalLM,StableLmModel:()=>l.StableLmModel,StableLmPreTrainedModel:()=>l.StableLmPreTrainedModel,Starcoder2ForCausalLM:()=>l.Starcoder2ForCausalLM,Starcoder2Model:()=>l.Starcoder2Model,Starcoder2PreTrainedModel:()=>l.Starcoder2PreTrainedModel,StoppingCriteria:()=>_.StoppingCriteria,StoppingCriteriaList:()=>_.StoppingCriteriaList,SummarizationPipeline:()=>o.SummarizationPipeline,Swin2SRForImageSuperResolution:()=>l.Swin2SRForImageSuperResolution,Swin2SRImageProcessor:()=>d.Swin2SRImageProcessor,Swin2SRModel:()=>l.Swin2SRModel,Swin2SRPreTrainedModel:()=>l.Swin2SRPreTrainedModel,SwinForImageClassification:()=>l.SwinForImageClassification,SwinModel:()=>l.SwinModel,SwinPreTrainedModel:()=>l.SwinPreTrainedModel,T5ForConditionalGeneration:()=>l.T5ForConditionalGeneration,T5Model:()=>l.T5Model,T5PreTrainedModel:()=>l.T5PreTrainedModel,T5Tokenizer:()=>u.T5Tokenizer,TableTransformerForObjectDetection:()=>l.TableTransformerForObjectDetection,TableTransformerModel:()=>l.TableTransformerModel,TableTransformerObjectDetectionOutput:()=>l.TableTransformerObjectDetectionOutput,TableTransformerPreTrainedModel:()=>l.TableTransformerPreTrainedModel,Tensor:()=>m.Tensor,Text2TextGenerationPipeline:()=>o.Text2TextGenerationPipeline,TextClassificationPipeline:()=>o.TextClassificationPipeline,TextGenerationPipeline:()=>o.TextGenerationPipeline,TextStreamer:()=>g.TextStreamer,TextToAudioPipeline:()=>o.TextToAudioPipeline,TokenClassificationPipeline:()=>o.TokenClassificationPipeline,TokenClassifierOutput:()=>l.TokenClassifierOutput,TokenizerModel:()=>u.TokenizerModel,TrOCRForCausalLM:()=>l.TrOCRForCausalLM,TrOCRPreTrainedModel:()=>l.TrOCRPreTrainedModel,TranslationPipeline:()=>o.TranslationPipeline,UniSpeechForCTC:()=>l.UniSpeechForCTC,UniSpeechForSequenceClassification:()=>l.UniSpeechForSequenceClassification,UniSpeechModel:()=>l.UniSpeechModel,UniSpeechPreTrainedModel:()=>l.UniSpeechPreTrainedModel,UniSpeechSatForAudioFrameClassification:()=>l.UniSpeechSatForAudioFrameClassification,UniSpeechSatForCTC:()=>l.UniSpeechSatForCTC,UniSpeechSatForSequenceClassification:()=>l.UniSpeechSatForSequenceClassification,UniSpeechSatModel:()=>l.UniSpeechSatModel,UniSpeechSatPreTrainedModel:()=>l.UniSpeechSatPreTrainedModel,ViTFeatureExtractor:()=>d.ViTFeatureExtractor,ViTForImageClassification:()=>l.ViTForImageClassification,ViTImageProcessor:()=>d.ViTImageProcessor,ViTMAEModel:()=>l.ViTMAEModel,ViTMAEPreTrainedModel:()=>l.ViTMAEPreTrainedModel,ViTMSNForImageClassification:()=>l.ViTMSNForImageClassification,ViTMSNModel:()=>l.ViTMSNModel,ViTMSNPreTrainedModel:()=>l.ViTMSNPreTrainedModel,ViTModel:()=>l.ViTModel,ViTPreTrainedModel:()=>l.ViTPreTrainedModel,VisionEncoderDecoderModel:()=>l.VisionEncoderDecoderModel,VitMatteForImageMatting:()=>l.VitMatteForImageMatting,VitMatteImageProcessor:()=>d.VitMatteImageProcessor,VitMattePreTrainedModel:()=>l.VitMattePreTrainedModel,VitsModel:()=>l.VitsModel,VitsModelOutput:()=>l.VitsModelOutput,VitsPreTrainedModel:()=>l.VitsPreTrainedModel,VitsTokenizer:()=>u.VitsTokenizer,Wav2Vec2BertForCTC:()=>l.Wav2Vec2BertForCTC,Wav2Vec2BertForSequenceClassification:()=>l.Wav2Vec2BertForSequenceClassification,Wav2Vec2BertModel:()=>l.Wav2Vec2BertModel,Wav2Vec2BertPreTrainedModel:()=>l.Wav2Vec2BertPreTrainedModel,Wav2Vec2CTCTokenizer:()=>u.Wav2Vec2CTCTokenizer,Wav2Vec2FeatureExtractor:()=>d.Wav2Vec2FeatureExtractor,Wav2Vec2ForAudioFrameClassification:()=>l.Wav2Vec2ForAudioFrameClassification,Wav2Vec2ForCTC:()=>l.Wav2Vec2ForCTC,Wav2Vec2ForSequenceClassification:()=>l.Wav2Vec2ForSequenceClassification,Wav2Vec2Model:()=>l.Wav2Vec2Model,Wav2Vec2PreTrainedModel:()=>l.Wav2Vec2PreTrainedModel,Wav2Vec2ProcessorWithLM:()=>d.Wav2Vec2ProcessorWithLM,WavLMForAudioFrameClassification:()=>l.WavLMForAudioFrameClassification,WavLMForCTC:()=>l.WavLMForCTC,WavLMForSequenceClassification:()=>l.WavLMForSequenceClassification,WavLMForXVector:()=>l.WavLMForXVector,WavLMModel:()=>l.WavLMModel,WavLMPreTrainedModel:()=>l.WavLMPreTrainedModel,WeSpeakerFeatureExtractor:()=>d.WeSpeakerFeatureExtractor,WeSpeakerResNetModel:()=>l.WeSpeakerResNetModel,WeSpeakerResNetPreTrainedModel:()=>l.WeSpeakerResNetPreTrainedModel,WhisperFeatureExtractor:()=>d.WhisperFeatureExtractor,WhisperForConditionalGeneration:()=>l.WhisperForConditionalGeneration,WhisperModel:()=>l.WhisperModel,WhisperPreTrainedModel:()=>l.WhisperPreTrainedModel,WhisperProcessor:()=>d.WhisperProcessor,WhisperTextStreamer:()=>g.WhisperTextStreamer,WhisperTokenizer:()=>u.WhisperTokenizer,XLMForQuestionAnswering:()=>l.XLMForQuestionAnswering,XLMForSequenceClassification:()=>l.XLMForSequenceClassification,XLMForTokenClassification:()=>l.XLMForTokenClassification,XLMModel:()=>l.XLMModel,XLMPreTrainedModel:()=>l.XLMPreTrainedModel,XLMRobertaForMaskedLM:()=>l.XLMRobertaForMaskedLM,XLMRobertaForQuestionAnswering:()=>l.XLMRobertaForQuestionAnswering,XLMRobertaForSequenceClassification:()=>l.XLMRobertaForSequenceClassification,XLMRobertaForTokenClassification:()=>l.XLMRobertaForTokenClassification,XLMRobertaModel:()=>l.XLMRobertaModel,XLMRobertaPreTrainedModel:()=>l.XLMRobertaPreTrainedModel,XLMRobertaTokenizer:()=>u.XLMRobertaTokenizer,XLMTokenizer:()=>u.XLMTokenizer,XLMWithLMHeadModel:()=>l.XLMWithLMHeadModel,XVectorOutput:()=>l.XVectorOutput,YolosFeatureExtractor:()=>d.YolosFeatureExtractor,YolosForObjectDetection:()=>l.YolosForObjectDetection,YolosModel:()=>l.YolosModel,YolosObjectDetectionOutput:()=>l.YolosObjectDetectionOutput,YolosPreTrainedModel:()=>l.YolosPreTrainedModel,ZeroShotAudioClassificationPipeline:()=>o.ZeroShotAudioClassificationPipeline,ZeroShotClassificationPipeline:()=>o.ZeroShotClassificationPipeline,ZeroShotImageClassificationPipeline:()=>o.ZeroShotImageClassificationPipeline,ZeroShotObjectDetectionPipeline:()=>o.ZeroShotObjectDetectionPipeline,bankers_round:()=>f.bankers_round,cat:()=>m.cat,cos_sim:()=>f.cos_sim,dot:()=>f.dot,dynamic_time_warping:()=>f.dynamic_time_warping,env:()=>i.env,full:()=>m.full,full_like:()=>m.full_like,getKeyValueShapes:()=>c.getKeyValueShapes,hamming:()=>p.hamming,hanning:()=>p.hanning,interpolate:()=>m.interpolate,interpolate_4d:()=>m.interpolate_4d,interpolate_data:()=>f.interpolate_data,is_chinese_char:()=>u.is_chinese_char,layer_norm:()=>m.layer_norm,log_softmax:()=>f.log_softmax,magnitude:()=>f.magnitude,matmul:()=>m.matmul,max:()=>f.max,mean:()=>m.mean,mean_pooling:()=>m.mean_pooling,medianFilter:()=>f.medianFilter,mel_filter_bank:()=>p.mel_filter_bank,min:()=>f.min,ones:()=>m.ones,ones_like:()=>m.ones_like,permute:()=>m.permute,permute_data:()=>f.permute_data,pipeline:()=>o.pipeline,quantize_embeddings:()=>m.quantize_embeddings,read_audio:()=>p.read_audio,rfft:()=>m.rfft,round:()=>f.round,softmax:()=>f.softmax,spectrogram:()=>p.spectrogram,stack:()=>m.stack,std_mean:()=>m.std_mean,topk:()=>m.topk,window_function:()=>p.window_function,zeros:()=>m.zeros,zeros_like:()=>m.zeros_like});var i=a(/*! ./env.js */"./src/env.js"),o=a(/*! ./pipelines.js */"./src/pipelines.js"),l=a(/*! ./models.js */"./src/models.js"),u=a(/*! ./tokenizers.js */"./src/tokenizers.js"),d=a(/*! ./processors.js */"./src/processors.js"),c=a(/*! ./configs.js */"./src/configs.js"),p=a(/*! ./utils/audio.js */"./src/utils/audio.js"),h=a(/*! ./utils/image.js */"./src/utils/image.js"),m=a(/*! ./utils/tensor.js */"./src/utils/tensor.js"),f=a(/*! ./utils/maths.js */"./src/utils/maths.js"),g=a(/*! ./generation/streamers.js */"./src/generation/streamers.js"),_=a(/*! ./generation/stopping_criteria.js */"./src/generation/stopping_criteria.js"),w=s.ASTFeatureExtractor,y=s.ASTForAudioClassification,b=s.ASTModel,v=s.ASTPreTrainedModel,x=s.AlbertForMaskedLM,M=s.AlbertForQuestionAnswering,T=s.AlbertForSequenceClassification,k=s.AlbertModel,$=s.AlbertPreTrainedModel,C=s.AlbertTokenizer,S=s.AudioClassificationPipeline,P=s.AutoConfig,E=s.AutoModel,F=s.AutoModelForAudioClassification,A=s.AutoModelForAudioFrameClassification,I=s.AutoModelForCTC,z=s.AutoModelForCausalLM,O=s.AutoModelForDepthEstimation,B=s.AutoModelForDocumentQuestionAnswering,L=s.AutoModelForImageClassification,D=s.AutoModelForImageFeatureExtraction,R=s.AutoModelForImageMatting,N=s.AutoModelForImageSegmentation,V=s.AutoModelForImageToImage,j=s.AutoModelForMaskGeneration,q=s.AutoModelForMaskedLM,G=s.AutoModelForNormalEstimation,U=s.AutoModelForObjectDetection,W=s.AutoModelForQuestionAnswering,H=s.AutoModelForSemanticSegmentation,X=s.AutoModelForSeq2SeqLM,K=s.AutoModelForSequenceClassification,Q=s.AutoModelForSpeechSeq2Seq,Y=s.AutoModelForTextToSpectrogram,Z=s.AutoModelForTextToWaveform,J=s.AutoModelForTokenClassification,ee=s.AutoModelForUniversalSegmentation,te=s.AutoModelForVision2Seq,ne=s.AutoModelForXVector,re=s.AutoModelForZeroShotObjectDetection,ae=s.AutoProcessor,se=s.AutoTokenizer,ie=s.AutomaticSpeechRecognitionPipeline,oe=s.BartForConditionalGeneration,le=s.BartForSequenceClassification,ue=s.BartModel,de=s.BartPretrainedModel,ce=s.BartTokenizer,pe=s.BaseModelOutput,he=s.BaseStreamer,me=s.BeitFeatureExtractor,fe=s.BeitForImageClassification,ge=s.BeitModel,_e=s.BeitPreTrainedModel,we=s.BertForMaskedLM,ye=s.BertForQuestionAnswering,be=s.BertForSequenceClassification,ve=s.BertForTokenClassification,xe=s.BertModel,Me=s.BertPreTrainedModel,Te=s.BertTokenizer,ke=s.BitImageProcessor,$e=s.BlenderbotForConditionalGeneration,Ce=s.BlenderbotModel,Se=s.BlenderbotPreTrainedModel,Pe=s.BlenderbotSmallForConditionalGeneration,Ee=s.BlenderbotSmallModel,Fe=s.BlenderbotSmallPreTrainedModel,Ae=s.BlenderbotSmallTokenizer,Ie=s.BlenderbotTokenizer,ze=s.BloomForCausalLM,Oe=s.BloomModel,Be=s.BloomPreTrainedModel,Le=s.BloomTokenizer,De=s.CLIPFeatureExtractor,Re=s.CLIPImageProcessor,Ne=s.CLIPModel,Ve=s.CLIPPreTrainedModel,je=s.CLIPSegForImageSegmentation,qe=s.CLIPSegModel,Ge=s.CLIPSegPreTrainedModel,Ue=s.CLIPTextModel,We=s.CLIPTextModelWithProjection,He=s.CLIPTokenizer,Xe=s.CLIPVisionModel,Ke=s.CLIPVisionModelWithProjection,Qe=s.CamembertForMaskedLM,Ye=s.CamembertForQuestionAnswering,Ze=s.CamembertForSequenceClassification,Je=s.CamembertForTokenClassification,et=s.CamembertModel,tt=s.CamembertPreTrainedModel,nt=s.CamembertTokenizer,rt=s.CausalLMOutput,at=s.CausalLMOutputWithPast,st=s.ChineseCLIPFeatureExtractor,it=s.ChineseCLIPModel,ot=s.ChineseCLIPPreTrainedModel,lt=s.ClapAudioModelWithProjection,ut=s.ClapFeatureExtractor,dt=s.ClapModel,ct=s.ClapPreTrainedModel,pt=s.ClapTextModelWithProjection,ht=s.CodeGenForCausalLM,mt=s.CodeGenModel,ft=s.CodeGenPreTrainedModel,gt=s.CodeGenTokenizer,_t=s.CodeLlamaTokenizer,wt=s.CohereForCausalLM,yt=s.CohereModel,bt=s.CoherePreTrainedModel,vt=s.CohereTokenizer,xt=s.ConvBertForMaskedLM,Mt=s.ConvBertForQuestionAnswering,Tt=s.ConvBertForSequenceClassification,kt=s.ConvBertForTokenClassification,$t=s.ConvBertModel,Ct=s.ConvBertPreTrainedModel,St=s.ConvBertTokenizer,Pt=s.ConvNextFeatureExtractor,Et=s.ConvNextForImageClassification,Ft=s.ConvNextImageProcessor,At=s.ConvNextModel,It=s.ConvNextPreTrainedModel,zt=s.ConvNextV2ForImageClassification,Ot=s.ConvNextV2Model,Bt=s.ConvNextV2PreTrainedModel,Lt=s.DPTFeatureExtractor,Dt=s.DPTForDepthEstimation,Rt=s.DPTImageProcessor,Nt=s.DPTModel,Vt=s.DPTPreTrainedModel,jt=s.DebertaForMaskedLM,qt=s.DebertaForQuestionAnswering,Gt=s.DebertaForSequenceClassification,Ut=s.DebertaForTokenClassification,Wt=s.DebertaModel,Ht=s.DebertaPreTrainedModel,Xt=s.DebertaTokenizer,Kt=s.DebertaV2ForMaskedLM,Qt=s.DebertaV2ForQuestionAnswering,Yt=s.DebertaV2ForSequenceClassification,Zt=s.DebertaV2ForTokenClassification,Jt=s.DebertaV2Model,en=s.DebertaV2PreTrainedModel,tn=s.DebertaV2Tokenizer,nn=s.DecisionTransformerModel,rn=s.DecisionTransformerPreTrainedModel,an=s.DeiTFeatureExtractor,sn=s.DeiTForImageClassification,on=s.DeiTModel,ln=s.DeiTPreTrainedModel,un=s.DepthAnythingForDepthEstimation,dn=s.DepthAnythingPreTrainedModel,cn=s.DepthEstimationPipeline,pn=s.DepthProForDepthEstimation,hn=s.DepthProPreTrainedModel,mn=s.DetrFeatureExtractor,fn=s.DetrForObjectDetection,gn=s.DetrForSegmentation,_n=s.DetrModel,wn=s.DetrObjectDetectionOutput,yn=s.DetrPreTrainedModel,bn=s.DetrSegmentationOutput,vn=s.Dinov2ForImageClassification,xn=s.Dinov2Model,Mn=s.Dinov2PreTrainedModel,Tn=s.DistilBertForMaskedLM,kn=s.DistilBertForQuestionAnswering,$n=s.DistilBertForSequenceClassification,Cn=s.DistilBertForTokenClassification,Sn=s.DistilBertModel,Pn=s.DistilBertPreTrainedModel,En=s.DistilBertTokenizer,Fn=s.DocumentQuestionAnsweringPipeline,An=s.DonutFeatureExtractor,In=s.DonutImageProcessor,zn=s.DonutSwinModel,On=s.DonutSwinPreTrainedModel,Bn=s.EfficientNetForImageClassification,Ln=s.EfficientNetImageProcessor,Dn=s.EfficientNetModel,Rn=s.EfficientNetPreTrainedModel,Nn=s.ElectraForMaskedLM,Vn=s.ElectraForQuestionAnswering,jn=s.ElectraForSequenceClassification,qn=s.ElectraForTokenClassification,Gn=s.ElectraModel,Un=s.ElectraPreTrainedModel,Wn=s.ElectraTokenizer,Hn=s.EosTokenCriteria,Xn=s.EsmForMaskedLM,Kn=s.EsmForSequenceClassification,Qn=s.EsmForTokenClassification,Yn=s.EsmModel,Zn=s.EsmPreTrainedModel,Jn=s.EsmTokenizer,er=s.FFT,tr=s.FalconForCausalLM,nr=s.FalconModel,rr=s.FalconPreTrainedModel,ar=s.FalconTokenizer,sr=s.FastViTForImageClassification,ir=s.FastViTModel,or=s.FastViTPreTrainedModel,lr=s.FeatureExtractionPipeline,ur=s.FeatureExtractor,dr=s.FillMaskPipeline,cr=s.Florence2ForConditionalGeneration,pr=s.Florence2PreTrainedModel,hr=s.Florence2Processor,mr=s.GLPNFeatureExtractor,fr=s.GLPNForDepthEstimation,gr=s.GLPNModel,_r=s.GLPNPreTrainedModel,wr=s.GPT2LMHeadModel,yr=s.GPT2Model,br=s.GPT2PreTrainedModel,vr=s.GPT2Tokenizer,xr=s.GPTBigCodeForCausalLM,Mr=s.GPTBigCodeModel,Tr=s.GPTBigCodePreTrainedModel,kr=s.GPTJForCausalLM,$r=s.GPTJModel,Cr=s.GPTJPreTrainedModel,Sr=s.GPTNeoForCausalLM,Pr=s.GPTNeoModel,Er=s.GPTNeoPreTrainedModel,Fr=s.GPTNeoXForCausalLM,Ar=s.GPTNeoXModel,Ir=s.GPTNeoXPreTrainedModel,zr=s.GPTNeoXTokenizer,Or=s.Gemma2ForCausalLM,Br=s.Gemma2Model,Lr=s.Gemma2PreTrainedModel,Dr=s.GemmaForCausalLM,Rr=s.GemmaModel,Nr=s.GemmaPreTrainedModel,Vr=s.GemmaTokenizer,jr=s.GraniteForCausalLM,qr=s.GraniteModel,Gr=s.GranitePreTrainedModel,Ur=s.Grok1Tokenizer,Wr=s.GroupViTModel,Hr=s.GroupViTPreTrainedModel,Xr=s.HerbertTokenizer,Kr=s.HieraForImageClassification,Qr=s.HieraModel,Yr=s.HieraPreTrainedModel,Zr=s.HubertForCTC,Jr=s.HubertForSequenceClassification,ea=s.HubertModel,ta=s.HubertPreTrainedModel,na=s.ImageClassificationPipeline,ra=s.ImageFeatureExtractionPipeline,aa=s.ImageFeatureExtractor,sa=s.ImageMattingOutput,ia=s.ImageSegmentationPipeline,oa=s.ImageToImagePipeline,la=s.ImageToTextPipeline,ua=s.InterruptableStoppingCriteria,da=s.JAISLMHeadModel,ca=s.JAISModel,pa=s.JAISPreTrainedModel,ha=s.LlamaForCausalLM,ma=s.LlamaModel,fa=s.LlamaPreTrainedModel,ga=s.LlamaTokenizer,_a=s.LlavaForConditionalGeneration,wa=s.LlavaPreTrainedModel,ya=s.LongT5ForConditionalGeneration,ba=s.LongT5Model,va=s.LongT5PreTrainedModel,xa=s.M2M100ForConditionalGeneration,Ma=s.M2M100Model,Ta=s.M2M100PreTrainedModel,ka=s.M2M100Tokenizer,$a=s.MBart50Tokenizer,Ca=s.MBartForCausalLM,Sa=s.MBartForConditionalGeneration,Pa=s.MBartForSequenceClassification,Ea=s.MBartModel,Fa=s.MBartPreTrainedModel,Aa=s.MBartTokenizer,Ia=s.MPNetForMaskedLM,za=s.MPNetForQuestionAnswering,Oa=s.MPNetForSequenceClassification,Ba=s.MPNetForTokenClassification,La=s.MPNetModel,Da=s.MPNetPreTrainedModel,Ra=s.MPNetTokenizer,Na=s.MT5ForConditionalGeneration,Va=s.MT5Model,ja=s.MT5PreTrainedModel,qa=s.MarianMTModel,Ga=s.MarianModel,Ua=s.MarianPreTrainedModel,Wa=s.MarianTokenizer,Ha=s.MaskFormerFeatureExtractor,Xa=s.MaskFormerForInstanceSegmentation,Ka=s.MaskFormerModel,Qa=s.MaskFormerPreTrainedModel,Ya=s.MaskedLMOutput,Za=s.MaxLengthCriteria,Ja=s.MistralForCausalLM,es=s.MistralModel,ts=s.MistralPreTrainedModel,ns=s.MobileBertForMaskedLM,rs=s.MobileBertForQuestionAnswering,as=s.MobileBertForSequenceClassification,ss=s.MobileBertModel,is=s.MobileBertPreTrainedModel,os=s.MobileBertTokenizer,ls=s.MobileLLMForCausalLM,us=s.MobileLLMModel,ds=s.MobileLLMPreTrainedModel,cs=s.MobileNetV1FeatureExtractor,ps=s.MobileNetV1ForImageClassification,hs=s.MobileNetV1Model,ms=s.MobileNetV1PreTrainedModel,fs=s.MobileNetV2FeatureExtractor,gs=s.MobileNetV2ForImageClassification,_s=s.MobileNetV2Model,ws=s.MobileNetV2PreTrainedModel,ys=s.MobileNetV3FeatureExtractor,bs=s.MobileNetV3ForImageClassification,vs=s.MobileNetV3Model,xs=s.MobileNetV3PreTrainedModel,Ms=s.MobileNetV4FeatureExtractor,Ts=s.MobileNetV4ForImageClassification,ks=s.MobileNetV4Model,$s=s.MobileNetV4PreTrainedModel,Cs=s.MobileViTFeatureExtractor,Ss=s.MobileViTForImageClassification,Ps=s.MobileViTImageProcessor,Es=s.MobileViTModel,Fs=s.MobileViTPreTrainedModel,As=s.MobileViTV2ForImageClassification,Is=s.MobileViTV2Model,zs=s.MobileViTV2PreTrainedModel,Os=s.ModelOutput,Bs=s.Moondream1ForConditionalGeneration,Ls=s.MptForCausalLM,Ds=s.MptModel,Rs=s.MptPreTrainedModel,Ns=s.MusicgenForCausalLM,Vs=s.MusicgenForConditionalGeneration,js=s.MusicgenModel,qs=s.MusicgenPreTrainedModel,Gs=s.NllbTokenizer,Us=s.NomicBertModel,Ws=s.NomicBertPreTrainedModel,Hs=s.NougatImageProcessor,Xs=s.NougatTokenizer,Ks=s.OPTForCausalLM,Qs=s.OPTModel,Ys=s.OPTPreTrainedModel,Zs=s.ObjectDetectionPipeline,Js=s.OlmoForCausalLM,ei=s.OlmoModel,ti=s.OlmoPreTrainedModel,ni=s.OpenELMForCausalLM,ri=s.OpenELMModel,ai=s.OpenELMPreTrainedModel,si=s.OwlViTFeatureExtractor,ii=s.OwlViTForObjectDetection,oi=s.OwlViTModel,li=s.OwlViTPreTrainedModel,ui=s.OwlViTProcessor,di=s.Owlv2ForObjectDetection,ci=s.Owlv2ImageProcessor,pi=s.Owlv2Model,hi=s.Owlv2PreTrainedModel,mi=s.Phi3ForCausalLM,fi=s.Phi3Model,gi=s.Phi3PreTrainedModel,_i=s.PhiForCausalLM,wi=s.PhiModel,yi=s.PhiPreTrainedModel,bi=s.Pipeline,vi=s.PreTrainedModel,xi=s.PreTrainedTokenizer,Mi=s.PretrainedConfig,Ti=s.PretrainedMixin,ki=s.Processor,$i=s.PvtForImageClassification,Ci=s.PvtImageProcessor,Si=s.PvtModel,Pi=s.PvtPreTrainedModel,Ei=s.PyAnnoteFeatureExtractor,Fi=s.PyAnnoteForAudioFrameClassification,Ai=s.PyAnnoteModel,Ii=s.PyAnnotePreTrainedModel,zi=s.PyAnnoteProcessor,Oi=s.QuestionAnsweringModelOutput,Bi=s.QuestionAnsweringPipeline,Li=s.Qwen2ForCausalLM,Di=s.Qwen2Model,Ri=s.Qwen2PreTrainedModel,Ni=s.Qwen2Tokenizer,Vi=s.RTDetrForObjectDetection,ji=s.RTDetrImageProcessor,qi=s.RTDetrModel,Gi=s.RTDetrObjectDetectionOutput,Ui=s.RTDetrPreTrainedModel,Wi=s.RawImage,Hi=s.ResNetForImageClassification,Xi=s.ResNetModel,Ki=s.ResNetPreTrainedModel,Qi=s.RoFormerForMaskedLM,Yi=s.RoFormerForQuestionAnswering,Zi=s.RoFormerForSequenceClassification,Ji=s.RoFormerForTokenClassification,eo=s.RoFormerModel,to=s.RoFormerPreTrainedModel,no=s.RoFormerTokenizer,ro=s.RobertaForMaskedLM,ao=s.RobertaForQuestionAnswering,so=s.RobertaForSequenceClassification,io=s.RobertaForTokenClassification,oo=s.RobertaModel,lo=s.RobertaPreTrainedModel,uo=s.RobertaTokenizer,co=s.SamImageProcessor,po=s.SamImageSegmentationOutput,ho=s.SamModel,mo=s.SamPreTrainedModel,fo=s.SamProcessor,go=s.SapiensFeatureExtractor,_o=s.SapiensForDepthEstimation,wo=s.SapiensForNormalEstimation,yo=s.SapiensForSemanticSegmentation,bo=s.SapiensPreTrainedModel,vo=s.SeamlessM4TFeatureExtractor,xo=s.SegformerFeatureExtractor,Mo=s.SegformerForImageClassification,To=s.SegformerForSemanticSegmentation,ko=s.SegformerModel,$o=s.SegformerPreTrainedModel,Co=s.Seq2SeqLMOutput,So=s.SequenceClassifierOutput,Po=s.SiglipImageProcessor,Eo=s.SiglipModel,Fo=s.SiglipPreTrainedModel,Ao=s.SiglipTextModel,Io=s.SiglipTokenizer,zo=s.SiglipVisionModel,Oo=s.SpeechT5FeatureExtractor,Bo=s.SpeechT5ForSpeechToText,Lo=s.SpeechT5ForTextToSpeech,Do=s.SpeechT5HifiGan,Ro=s.SpeechT5Model,No=s.SpeechT5PreTrainedModel,Vo=s.SpeechT5Processor,jo=s.SpeechT5Tokenizer,qo=s.SqueezeBertForMaskedLM,Go=s.SqueezeBertForQuestionAnswering,Uo=s.SqueezeBertForSequenceClassification,Wo=s.SqueezeBertModel,Ho=s.SqueezeBertPreTrainedModel,Xo=s.SqueezeBertTokenizer,Ko=s.StableLmForCausalLM,Qo=s.StableLmModel,Yo=s.StableLmPreTrainedModel,Zo=s.Starcoder2ForCausalLM,Jo=s.Starcoder2Model,el=s.Starcoder2PreTrainedModel,tl=s.StoppingCriteria,nl=s.StoppingCriteriaList,rl=s.SummarizationPipeline,al=s.Swin2SRForImageSuperResolution,sl=s.Swin2SRImageProcessor,il=s.Swin2SRModel,ol=s.Swin2SRPreTrainedModel,ll=s.SwinForImageClassification,ul=s.SwinModel,dl=s.SwinPreTrainedModel,cl=s.T5ForConditionalGeneration,pl=s.T5Model,hl=s.T5PreTrainedModel,ml=s.T5Tokenizer,fl=s.TableTransformerForObjectDetection,gl=s.TableTransformerModel,_l=s.TableTransformerObjectDetectionOutput,wl=s.TableTransformerPreTrainedModel,yl=s.Tensor,bl=s.Text2TextGenerationPipeline,vl=s.TextClassificationPipeline,xl=s.TextGenerationPipeline,Ml=s.TextStreamer,Tl=s.TextToAudioPipeline,kl=s.TokenClassificationPipeline,$l=s.TokenClassifierOutput,Cl=s.TokenizerModel,Sl=s.TrOCRForCausalLM,Pl=s.TrOCRPreTrainedModel,El=s.TranslationPipeline,Fl=s.UniSpeechForCTC,Al=s.UniSpeechForSequenceClassification,Il=s.UniSpeechModel,zl=s.UniSpeechPreTrainedModel,Ol=s.UniSpeechSatForAudioFrameClassification,Bl=s.UniSpeechSatForCTC,Ll=s.UniSpeechSatForSequenceClassification,Dl=s.UniSpeechSatModel,Rl=s.UniSpeechSatPreTrainedModel,Nl=s.ViTFeatureExtractor,Vl=s.ViTForImageClassification,jl=s.ViTImageProcessor,ql=s.ViTMAEModel,Gl=s.ViTMAEPreTrainedModel,Ul=s.ViTMSNForImageClassification,Wl=s.ViTMSNModel,Hl=s.ViTMSNPreTrainedModel,Xl=s.ViTModel,Kl=s.ViTPreTrainedModel,Ql=s.VisionEncoderDecoderModel,Yl=s.VitMatteForImageMatting,Zl=s.VitMatteImageProcessor,Jl=s.VitMattePreTrainedModel,eu=s.VitsModel,tu=s.VitsModelOutput,nu=s.VitsPreTrainedModel,ru=s.VitsTokenizer,au=s.Wav2Vec2BertForCTC,su=s.Wav2Vec2BertForSequenceClassification,iu=s.Wav2Vec2BertModel,ou=s.Wav2Vec2BertPreTrainedModel,lu=s.Wav2Vec2CTCTokenizer,uu=s.Wav2Vec2FeatureExtractor,du=s.Wav2Vec2ForAudioFrameClassification,cu=s.Wav2Vec2ForCTC,pu=s.Wav2Vec2ForSequenceClassification,hu=s.Wav2Vec2Model,mu=s.Wav2Vec2PreTrainedModel,fu=s.Wav2Vec2ProcessorWithLM,gu=s.WavLMForAudioFrameClassification,_u=s.WavLMForCTC,wu=s.WavLMForSequenceClassification,yu=s.WavLMForXVector,bu=s.WavLMModel,vu=s.WavLMPreTrainedModel,xu=s.WeSpeakerFeatureExtractor,Mu=s.WeSpeakerResNetModel,Tu=s.WeSpeakerResNetPreTrainedModel,ku=s.WhisperFeatureExtractor,$u=s.WhisperForConditionalGeneration,Cu=s.WhisperModel,Su=s.WhisperPreTrainedModel,Pu=s.WhisperProcessor,Eu=s.WhisperTextStreamer,Fu=s.WhisperTokenizer,Au=s.XLMForQuestionAnswering,Iu=s.XLMForSequenceClassification,zu=s.XLMForTokenClassification,Ou=s.XLMModel,Bu=s.XLMPreTrainedModel,Lu=s.XLMRobertaForMaskedLM,Du=s.XLMRobertaForQuestionAnswering,Ru=s.XLMRobertaForSequenceClassification,Nu=s.XLMRobertaForTokenClassification,Vu=s.XLMRobertaModel,ju=s.XLMRobertaPreTrainedModel,qu=s.XLMRobertaTokenizer,Gu=s.XLMTokenizer,Uu=s.XLMWithLMHeadModel,Wu=s.XVectorOutput,Hu=s.YolosFeatureExtractor,Xu=s.YolosForObjectDetection,Ku=s.YolosModel,Qu=s.YolosObjectDetectionOutput,Yu=s.YolosPreTrainedModel,Zu=s.ZeroShotAudioClassificationPipeline,Ju=s.ZeroShotClassificationPipeline,ed=s.ZeroShotImageClassificationPipeline,td=s.ZeroShotObjectDetectionPipeline,nd=s.bankers_round,rd=s.cat,ad=s.cos_sim,sd=s.dot,id=s.dynamic_time_warping,od=s.env,ld=s.full,ud=s.full_like,dd=s.getKeyValueShapes,cd=s.hamming,pd=s.hanning,hd=s.interpolate,md=s.interpolate_4d,fd=s.interpolate_data,gd=s.is_chinese_char,_d=s.layer_norm,wd=s.log_softmax,yd=s.magnitude,bd=s.matmul,vd=s.max,xd=s.mean,Md=s.mean_pooling,Td=s.medianFilter,kd=s.mel_filter_bank,$d=s.min,Cd=s.ones,Sd=s.ones_like,Pd=s.permute,Ed=s.permute_data,Fd=s.pipeline,Ad=s.quantize_embeddings,Id=s.read_audio,zd=s.rfft,Od=s.round,Bd=s.softmax,Ld=s.spectrogram,Dd=s.stack,Rd=s.std_mean,Nd=s.topk,Vd=s.window_function,jd=s.zeros,qd=s.zeros_like;export{w as ASTFeatureExtractor,y as ASTForAudioClassification,b as ASTModel,v as ASTPreTrainedModel,x as AlbertForMaskedLM,M as AlbertForQuestionAnswering,T as AlbertForSequenceClassification,k as AlbertModel,$ as AlbertPreTrainedModel,C as AlbertTokenizer,S as AudioClassificationPipeline,P as AutoConfig,E as AutoModel,F as AutoModelForAudioClassification,A as AutoModelForAudioFrameClassification,I as AutoModelForCTC,z as AutoModelForCausalLM,O as AutoModelForDepthEstimation,B as AutoModelForDocumentQuestionAnswering,L as AutoModelForImageClassification,D as AutoModelForImageFeatureExtraction,R as AutoModelForImageMatting,N as AutoModelForImageSegmentation,V as AutoModelForImageToImage,j as AutoModelForMaskGeneration,q as AutoModelForMaskedLM,G as AutoModelForNormalEstimation,U as AutoModelForObjectDetection,W as AutoModelForQuestionAnswering,H as AutoModelForSemanticSegmentation,X as AutoModelForSeq2SeqLM,K as AutoModelForSequenceClassification,Q as AutoModelForSpeechSeq2Seq,Y as AutoModelForTextToSpectrogram,Z as AutoModelForTextToWaveform,J as AutoModelForTokenClassification,ee as AutoModelForUniversalSegmentation,te as AutoModelForVision2Seq,ne as AutoModelForXVector,re as AutoModelForZeroShotObjectDetection,ae as AutoProcessor,se as AutoTokenizer,ie as AutomaticSpeechRecognitionPipeline,oe as BartForConditionalGeneration,le as BartForSequenceClassification,ue as BartModel,de as BartPretrainedModel,ce as BartTokenizer,pe as BaseModelOutput,he as BaseStreamer,me as BeitFeatureExtractor,fe as BeitForImageClassification,ge as BeitModel,_e as BeitPreTrainedModel,we as BertForMaskedLM,ye as BertForQuestionAnswering,be as BertForSequenceClassification,ve as BertForTokenClassification,xe as BertModel,Me as BertPreTrainedModel,Te as BertTokenizer,ke as BitImageProcessor,$e as BlenderbotForConditionalGeneration,Ce as BlenderbotModel,Se as BlenderbotPreTrainedModel,Pe as BlenderbotSmallForConditionalGeneration,Ee as BlenderbotSmallModel,Fe as BlenderbotSmallPreTrainedModel,Ae as BlenderbotSmallTokenizer,Ie as BlenderbotTokenizer,ze as BloomForCausalLM,Oe as BloomModel,Be as BloomPreTrainedModel,Le as BloomTokenizer,De as CLIPFeatureExtractor,Re as CLIPImageProcessor,Ne as CLIPModel,Ve as CLIPPreTrainedModel,je as CLIPSegForImageSegmentation,qe as CLIPSegModel,Ge as CLIPSegPreTrainedModel,Ue as CLIPTextModel,We as CLIPTextModelWithProjection,He as CLIPTokenizer,Xe as CLIPVisionModel,Ke as CLIPVisionModelWithProjection,Qe as CamembertForMaskedLM,Ye as CamembertForQuestionAnswering,Ze as CamembertForSequenceClassification,Je as CamembertForTokenClassification,et as CamembertModel,tt as CamembertPreTrainedModel,nt as CamembertTokenizer,rt as CausalLMOutput,at as CausalLMOutputWithPast,st as ChineseCLIPFeatureExtractor,it as ChineseCLIPModel,ot as ChineseCLIPPreTrainedModel,lt as ClapAudioModelWithProjection,ut as ClapFeatureExtractor,dt as ClapModel,ct as ClapPreTrainedModel,pt as ClapTextModelWithProjection,ht as CodeGenForCausalLM,mt as CodeGenModel,ft as CodeGenPreTrainedModel,gt as CodeGenTokenizer,_t as CodeLlamaTokenizer,wt as CohereForCausalLM,yt as CohereModel,bt as CoherePreTrainedModel,vt as CohereTokenizer,xt as ConvBertForMaskedLM,Mt as ConvBertForQuestionAnswering,Tt as ConvBertForSequenceClassification,kt as ConvBertForTokenClassification,$t as ConvBertModel,Ct as ConvBertPreTrainedModel,St as ConvBertTokenizer,Pt as ConvNextFeatureExtractor,Et as ConvNextForImageClassification,Ft as ConvNextImageProcessor,At as ConvNextModel,It as ConvNextPreTrainedModel,zt as ConvNextV2ForImageClassification,Ot as ConvNextV2Model,Bt as ConvNextV2PreTrainedModel,Lt as DPTFeatureExtractor,Dt as DPTForDepthEstimation,Rt as DPTImageProcessor,Nt as DPTModel,Vt as DPTPreTrainedModel,jt as DebertaForMaskedLM,qt as DebertaForQuestionAnswering,Gt as DebertaForSequenceClassification,Ut as DebertaForTokenClassification,Wt as DebertaModel,Ht as DebertaPreTrainedModel,Xt as DebertaTokenizer,Kt as DebertaV2ForMaskedLM,Qt as DebertaV2ForQuestionAnswering,Yt as DebertaV2ForSequenceClassification,Zt as DebertaV2ForTokenClassification,Jt as DebertaV2Model,en as DebertaV2PreTrainedModel,tn as DebertaV2Tokenizer,nn as DecisionTransformerModel,rn as DecisionTransformerPreTrainedModel,an as DeiTFeatureExtractor,sn as DeiTForImageClassification,on as DeiTModel,ln as DeiTPreTrainedModel,un as DepthAnythingForDepthEstimation,dn as DepthAnythingPreTrainedModel,cn as DepthEstimationPipeline,pn as DepthProForDepthEstimation,hn as DepthProPreTrainedModel,mn as DetrFeatureExtractor,fn as DetrForObjectDetection,gn as DetrForSegmentation,_n as DetrModel,wn as DetrObjectDetectionOutput,yn as DetrPreTrainedModel,bn as DetrSegmentationOutput,vn as Dinov2ForImageClassification,xn as Dinov2Model,Mn as Dinov2PreTrainedModel,Tn as DistilBertForMaskedLM,kn as DistilBertForQuestionAnswering,$n as DistilBertForSequenceClassification,Cn as DistilBertForTokenClassification,Sn as DistilBertModel,Pn as DistilBertPreTrainedModel,En as DistilBertTokenizer,Fn as DocumentQuestionAnsweringPipeline,An as DonutFeatureExtractor,In as DonutImageProcessor,zn as DonutSwinModel,On as DonutSwinPreTrainedModel,Bn as EfficientNetForImageClassification,Ln as EfficientNetImageProcessor,Dn as EfficientNetModel,Rn as EfficientNetPreTrainedModel,Nn as ElectraForMaskedLM,Vn as ElectraForQuestionAnswering,jn as ElectraForSequenceClassification,qn as ElectraForTokenClassification,Gn as ElectraModel,Un as ElectraPreTrainedModel,Wn as ElectraTokenizer,Hn as EosTokenCriteria,Xn as EsmForMaskedLM,Kn as EsmForSequenceClassification,Qn as EsmForTokenClassification,Yn as EsmModel,Zn as EsmPreTrainedModel,Jn as EsmTokenizer,er as FFT,tr as FalconForCausalLM,nr as FalconModel,rr as FalconPreTrainedModel,ar as FalconTokenizer,sr as FastViTForImageClassification,ir as FastViTModel,or as FastViTPreTrainedModel,lr as FeatureExtractionPipeline,ur as FeatureExtractor,dr as FillMaskPipeline,cr as Florence2ForConditionalGeneration,pr as Florence2PreTrainedModel,hr as Florence2Processor,mr as GLPNFeatureExtractor,fr as GLPNForDepthEstimation,gr as GLPNModel,_r as GLPNPreTrainedModel,wr as GPT2LMHeadModel,yr as GPT2Model,br as GPT2PreTrainedModel,vr as GPT2Tokenizer,xr as GPTBigCodeForCausalLM,Mr as GPTBigCodeModel,Tr as GPTBigCodePreTrainedModel,kr as GPTJForCausalLM,$r as GPTJModel,Cr as GPTJPreTrainedModel,Sr as GPTNeoForCausalLM,Pr as GPTNeoModel,Er as GPTNeoPreTrainedModel,Fr as GPTNeoXForCausalLM,Ar as GPTNeoXModel,Ir as GPTNeoXPreTrainedModel,zr as GPTNeoXTokenizer,Or as Gemma2ForCausalLM,Br as Gemma2Model,Lr as Gemma2PreTrainedModel,Dr as GemmaForCausalLM,Rr as GemmaModel,Nr as GemmaPreTrainedModel,Vr as GemmaTokenizer,jr as GraniteForCausalLM,qr as GraniteModel,Gr as GranitePreTrainedModel,Ur as Grok1Tokenizer,Wr as GroupViTModel,Hr as GroupViTPreTrainedModel,Xr as HerbertTokenizer,Kr as HieraForImageClassification,Qr as HieraModel,Yr as HieraPreTrainedModel,Zr as HubertForCTC,Jr as HubertForSequenceClassification,ea as HubertModel,ta as HubertPreTrainedModel,na as ImageClassificationPipeline,ra as ImageFeatureExtractionPipeline,aa as ImageFeatureExtractor,sa as ImageMattingOutput,ia as ImageSegmentationPipeline,oa as ImageToImagePipeline,la as ImageToTextPipeline,ua as InterruptableStoppingCriteria,da as JAISLMHeadModel,ca as JAISModel,pa as JAISPreTrainedModel,ha as LlamaForCausalLM,ma as LlamaModel,fa as LlamaPreTrainedModel,ga as LlamaTokenizer,_a as LlavaForConditionalGeneration,wa as LlavaPreTrainedModel,ya as LongT5ForConditionalGeneration,ba as LongT5Model,va as LongT5PreTrainedModel,xa as M2M100ForConditionalGeneration,Ma as M2M100Model,Ta as M2M100PreTrainedModel,ka as M2M100Tokenizer,$a as MBart50Tokenizer,Ca as MBartForCausalLM,Sa as MBartForConditionalGeneration,Pa as MBartForSequenceClassification,Ea as MBartModel,Fa as MBartPreTrainedModel,Aa as MBartTokenizer,Ia as MPNetForMaskedLM,za as MPNetForQuestionAnswering,Oa as MPNetForSequenceClassification,Ba as MPNetForTokenClassification,La as MPNetModel,Da as MPNetPreTrainedModel,Ra as MPNetTokenizer,Na as MT5ForConditionalGeneration,Va as MT5Model,ja as MT5PreTrainedModel,qa as MarianMTModel,Ga as MarianModel,Ua as MarianPreTrainedModel,Wa as MarianTokenizer,Ha as MaskFormerFeatureExtractor,Xa as MaskFormerForInstanceSegmentation,Ka as MaskFormerModel,Qa as MaskFormerPreTrainedModel,Ya as MaskedLMOutput,Za as MaxLengthCriteria,Ja as MistralForCausalLM,es as MistralModel,ts as MistralPreTrainedModel,ns as MobileBertForMaskedLM,rs as MobileBertForQuestionAnswering,as as MobileBertForSequenceClassification,ss as MobileBertModel,is as MobileBertPreTrainedModel,os as MobileBertTokenizer,ls as MobileLLMForCausalLM,us as MobileLLMModel,ds as MobileLLMPreTrainedModel,cs as MobileNetV1FeatureExtractor,ps as MobileNetV1ForImageClassification,hs as MobileNetV1Model,ms as MobileNetV1PreTrainedModel,fs as MobileNetV2FeatureExtractor,gs as MobileNetV2ForImageClassification,_s as MobileNetV2Model,ws as MobileNetV2PreTrainedModel,ys as MobileNetV3FeatureExtractor,bs as MobileNetV3ForImageClassification,vs as MobileNetV3Model,xs as MobileNetV3PreTrainedModel,Ms as MobileNetV4FeatureExtractor,Ts as MobileNetV4ForImageClassification,ks as MobileNetV4Model,$s as MobileNetV4PreTrainedModel,Cs as MobileViTFeatureExtractor,Ss as MobileViTForImageClassification,Ps as MobileViTImageProcessor,Es as MobileViTModel,Fs as MobileViTPreTrainedModel,As as MobileViTV2ForImageClassification,Is as MobileViTV2Model,zs as MobileViTV2PreTrainedModel,Os as ModelOutput,Bs as Moondream1ForConditionalGeneration,Ls as MptForCausalLM,Ds as MptModel,Rs as MptPreTrainedModel,Ns as MusicgenForCausalLM,Vs as MusicgenForConditionalGeneration,js as MusicgenModel,qs as MusicgenPreTrainedModel,Gs as NllbTokenizer,Us as NomicBertModel,Ws as NomicBertPreTrainedModel,Hs as NougatImageProcessor,Xs as NougatTokenizer,Ks as OPTForCausalLM,Qs as OPTModel,Ys as OPTPreTrainedModel,Zs as ObjectDetectionPipeline,Js as OlmoForCausalLM,ei as OlmoModel,ti as OlmoPreTrainedModel,ni as OpenELMForCausalLM,ri as OpenELMModel,ai as OpenELMPreTrainedModel,si as OwlViTFeatureExtractor,ii as OwlViTForObjectDetection,oi as OwlViTModel,li as OwlViTPreTrainedModel,ui as OwlViTProcessor,di as Owlv2ForObjectDetection,ci as Owlv2ImageProcessor,pi as Owlv2Model,hi as Owlv2PreTrainedModel,mi as Phi3ForCausalLM,fi as Phi3Model,gi as Phi3PreTrainedModel,_i as PhiForCausalLM,wi as PhiModel,yi as PhiPreTrainedModel,bi as Pipeline,vi as PreTrainedModel,xi as PreTrainedTokenizer,Mi as PretrainedConfig,Ti as PretrainedMixin,ki as Processor,$i as PvtForImageClassification,Ci as PvtImageProcessor,Si as PvtModel,Pi as PvtPreTrainedModel,Ei as PyAnnoteFeatureExtractor,Fi as PyAnnoteForAudioFrameClassification,Ai as PyAnnoteModel,Ii as PyAnnotePreTrainedModel,zi as PyAnnoteProcessor,Oi as QuestionAnsweringModelOutput,Bi as QuestionAnsweringPipeline,Li as Qwen2ForCausalLM,Di as Qwen2Model,Ri as Qwen2PreTrainedModel,Ni as Qwen2Tokenizer,Vi as RTDetrForObjectDetection,ji as RTDetrImageProcessor,qi as RTDetrModel,Gi as RTDetrObjectDetectionOutput,Ui as RTDetrPreTrainedModel,Wi as RawImage,Hi as ResNetForImageClassification,Xi as ResNetModel,Ki as ResNetPreTrainedModel,Qi as RoFormerForMaskedLM,Yi as RoFormerForQuestionAnswering,Zi as RoFormerForSequenceClassification,Ji as RoFormerForTokenClassification,eo as RoFormerModel,to as RoFormerPreTrainedModel,no as RoFormerTokenizer,ro as RobertaForMaskedLM,ao as RobertaForQuestionAnswering,so as RobertaForSequenceClassification,io as RobertaForTokenClassification,oo as RobertaModel,lo as RobertaPreTrainedModel,uo as RobertaTokenizer,co as SamImageProcessor,po as SamImageSegmentationOutput,ho as SamModel,mo as SamPreTrainedModel,fo as SamProcessor,go as SapiensFeatureExtractor,_o as SapiensForDepthEstimation,wo as SapiensForNormalEstimation,yo as SapiensForSemanticSegmentation,bo as SapiensPreTrainedModel,vo as SeamlessM4TFeatureExtractor,xo as SegformerFeatureExtractor,Mo as SegformerForImageClassification,To as SegformerForSemanticSegmentation,ko as SegformerModel,$o as SegformerPreTrainedModel,Co as Seq2SeqLMOutput,So as SequenceClassifierOutput,Po as SiglipImageProcessor,Eo as SiglipModel,Fo as SiglipPreTrainedModel,Ao as SiglipTextModel,Io as SiglipTokenizer,zo as SiglipVisionModel,Oo as SpeechT5FeatureExtractor,Bo as SpeechT5ForSpeechToText,Lo as SpeechT5ForTextToSpeech,Do as SpeechT5HifiGan,Ro as SpeechT5Model,No as SpeechT5PreTrainedModel,Vo as SpeechT5Processor,jo as SpeechT5Tokenizer,qo as SqueezeBertForMaskedLM,Go as SqueezeBertForQuestionAnswering,Uo as SqueezeBertForSequenceClassification,Wo as SqueezeBertModel,Ho as SqueezeBertPreTrainedModel,Xo as SqueezeBertTokenizer,Ko as StableLmForCausalLM,Qo as StableLmModel,Yo as StableLmPreTrainedModel,Zo as Starcoder2ForCausalLM,Jo as Starcoder2Model,el as Starcoder2PreTrainedModel,tl as StoppingCriteria,nl as StoppingCriteriaList,rl as SummarizationPipeline,al as Swin2SRForImageSuperResolution,sl as Swin2SRImageProcessor,il as Swin2SRModel,ol as Swin2SRPreTrainedModel,ll as SwinForImageClassification,ul as SwinModel,dl as SwinPreTrainedModel,cl as T5ForConditionalGeneration,pl as T5Model,hl as T5PreTrainedModel,ml as T5Tokenizer,fl as TableTransformerForObjectDetection,gl as TableTransformerModel,_l as TableTransformerObjectDetectionOutput,wl as TableTransformerPreTrainedModel,yl as Tensor,bl as Text2TextGenerationPipeline,vl as TextClassificationPipeline,xl as TextGenerationPipeline,Ml as TextStreamer,Tl as TextToAudioPipeline,kl as TokenClassificationPipeline,$l as TokenClassifierOutput,Cl as TokenizerModel,Sl as TrOCRForCausalLM,Pl as TrOCRPreTrainedModel,El as TranslationPipeline,Fl as UniSpeechForCTC,Al as UniSpeechForSequenceClassification,Il as UniSpeechModel,zl as UniSpeechPreTrainedModel,Ol as UniSpeechSatForAudioFrameClassification,Bl as UniSpeechSatForCTC,Ll as UniSpeechSatForSequenceClassification,Dl as UniSpeechSatModel,Rl as UniSpeechSatPreTrainedModel,Nl as ViTFeatureExtractor,Vl as ViTForImageClassification,jl as ViTImageProcessor,ql as ViTMAEModel,Gl as ViTMAEPreTrainedModel,Ul as ViTMSNForImageClassification,Wl as ViTMSNModel,Hl as ViTMSNPreTrainedModel,Xl as ViTModel,Kl as ViTPreTrainedModel,Ql as VisionEncoderDecoderModel,Yl as VitMatteForImageMatting,Zl as VitMatteImageProcessor,Jl as VitMattePreTrainedModel,eu as VitsModel,tu as VitsModelOutput,nu as VitsPreTrainedModel,ru as VitsTokenizer,au as Wav2Vec2BertForCTC,su as Wav2Vec2BertForSequenceClassification,iu as Wav2Vec2BertModel,ou as Wav2Vec2BertPreTrainedModel,lu as Wav2Vec2CTCTokenizer,uu as Wav2Vec2FeatureExtractor,du as Wav2Vec2ForAudioFrameClassification,cu as Wav2Vec2ForCTC,pu as Wav2Vec2ForSequenceClassification,hu as Wav2Vec2Model,mu as Wav2Vec2PreTrainedModel,fu as Wav2Vec2ProcessorWithLM,gu as WavLMForAudioFrameClassification,_u as WavLMForCTC,wu as WavLMForSequenceClassification,yu as WavLMForXVector,bu as WavLMModel,vu as WavLMPreTrainedModel,xu as WeSpeakerFeatureExtractor,Mu as WeSpeakerResNetModel,Tu as WeSpeakerResNetPreTrainedModel,ku as WhisperFeatureExtractor,$u as WhisperForConditionalGeneration,Cu as WhisperModel,Su as WhisperPreTrainedModel,Pu as WhisperProcessor,Eu as WhisperTextStreamer,Fu as WhisperTokenizer,Au as XLMForQuestionAnswering,Iu as XLMForSequenceClassification,zu as XLMForTokenClassification,Ou as XLMModel,Bu as XLMPreTrainedModel,Lu as XLMRobertaForMaskedLM,Du as XLMRobertaForQuestionAnswering,Ru as XLMRobertaForSequenceClassification,Nu as XLMRobertaForTokenClassification,Vu as XLMRobertaModel,ju as XLMRobertaPreTrainedModel,qu as XLMRobertaTokenizer,Gu as XLMTokenizer,Uu as XLMWithLMHeadModel,Wu as XVectorOutput,Hu as YolosFeatureExtractor,Xu as YolosForObjectDetection,Ku as YolosModel,Qu as YolosObjectDetectionOutput,Yu as YolosPreTrainedModel,Zu as ZeroShotAudioClassificationPipeline,Ju as ZeroShotClassificationPipeline,ed as ZeroShotImageClassificationPipeline,td as ZeroShotObjectDetectionPipeline,nd as bankers_round,rd as cat,ad as cos_sim,sd as dot,id as dynamic_time_warping,od as env,ld as full,ud as full_like,dd as getKeyValueShapes,cd as hamming,pd as hanning,hd as interpolate,md as interpolate_4d,fd as interpolate_data,gd as is_chinese_char,_d as layer_norm,wd as log_softmax,yd as magnitude,bd as matmul,vd as max,xd as mean,Md as mean_pooling,Td as medianFilter,kd as mel_filter_bank,$d as min,Cd as ones,Sd as ones_like,Pd as permute,Ed as permute_data,Fd as pipeline,Ad as quantize_embeddings,Id as read_audio,zd as rfft,Od as round,Bd as softmax,Ld as spectrogram,Dd as stack,Rd as std_mean,Nd as topk,Vd as window_function,jd as zeros,qd as zeros_like};
 //# sourceMappingURL=transformers.min.js.map