@huggingface/transformers 3.0.0-alpha.12 → 3.0.0-alpha.13

package/dist/transformers.min.mjs

@@ -91,7 +91,7 @@ import*as e from"fs";import*as t from"onnxruntime-node";import*as n from"path";i
   \************************/(e,t,n)=>{n.r(t),n.d(t,{AutoConfig:()=>l,PretrainedConfig:()=>i,getKeyValueShapes:()=>a});var r=n(/*! ./utils/core.js */"./src/utils/core.js"),o=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"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 o=s(e.decoder),a="num_decoder_layers"in o,i=(0,r.pick)(e,["model_type","is_encoder_decoder"]);return a?(i.num_decoder_layers=o.num_decoder_layers,i.num_decoder_heads=o.num_decoder_heads,i.decoder_hidden_size=o.decoder_hidden_size,i.num_encoder_layers=o.num_encoder_layers,i.num_encoder_heads=o.num_encoder_heads,i.encoder_hidden_size=o.encoder_hidden_size):(i.num_layers=o.num_layers,i.num_heads=o.num_heads,i.hidden_size=o.hidden_size),i}const o={...n,...(0,r.pick)(e,["model_type","multi_query","is_encoder_decoder"])};for(const n in t)o[n]=e[t[n]];return o}function a(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,o=r.decoder_dim_kv??r.decoder_hidden_size/r.num_decoder_heads,s=[1,r.num_encoder_heads,0,e],a=[1,r.num_decoder_heads,0,o];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`]=a,n[`${t}.${e}.decoder.value`]=a}else{const e=r.num_heads,o=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<o;++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<o;++e)n[`${t}.${e}.key_value`]=r}else if("bloom"===r.model_type){const r=[1*e,s,0],a=[1*e,0,s];for(let e=0;e<o;++e)n[`${t}.${e}.key`]=r,n[`${t}.${e}.value`]=a}else if("openelm"===r.model_type)for(let r=0;r<o;++r){const o=[1,e[r],0,s];n[`${t}.${r}.key`]=o,n[`${t}.${r}.value`]=o}else{const r=[1,e,0,s];for(let e=0;e<o;++e)n[`${t}.${e}.key`]=r,n[`${t}.${e}.value`]=r}}return n}class i{max_position_embeddings;constructor(e){this.model_type=null,this.is_encoder_decoder=!1,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:a="main"}={}){!n||n instanceof i||(n=new i(n));const l=n??await async function(e,t){return await(0,o.getModelJSON)(e,"config.json",!0,t)}(e,{progress_callback:t,config:n,cache_dir:r,local_files_only:s,revision:a});return new this(l)}}class l{static async from_pretrained(...e){return i.from_pretrained(...e)}}},"./src/env.js":
 /*!********************!*\
   !*** ./src/env.js ***!
-  \********************/(e,t,n)=>{n.r(t),n.d(t,{apis:()=>_,env:()=>b});var r=n(/*! fs */"fs"),o=n(/*! path */"path"),s=n(/*! url */"url");const a="undefined"!=typeof self,i=a&&"DedicatedWorkerGlobalScope"===self.constructor.name,l=a&&"caches"in self,c="undefined"!=typeof navigator&&"gpu"in navigator,d="undefined"!=typeof navigator&&"ml"in navigator,u="undefined"!=typeof process,p=u&&"node"===process?.release?.name,h=!T(r.default),m=!T(o.default),_=Object.freeze({IS_BROWSER_ENV:a,IS_WEBWORKER_ENV:i,IS_WEB_CACHE_AVAILABLE:l,IS_WEBGPU_AVAILABLE:c,IS_WEBNN_AVAILABLE:d,IS_PROCESS_AVAILABLE:u,IS_NODE_ENV:p,IS_FS_AVAILABLE:h,IS_PATH_AVAILABLE:m}),f=h&&m,g=f?o.default.dirname(o.default.dirname(s.default.fileURLToPath(import.meta.url))):"./",M=f?o.default.join(g,"/.cache/"):null,w="/models/",b={version:"3.0.0-alpha.12",backends:{onnx:{}},allowRemoteModels:!0,remoteHost:"https://huggingface.co/",remotePathTemplate:"{model}/resolve/{revision}/",allowLocalModels:!a,localModelPath:f?o.default.join(g,w):w,useFS:h,useBrowserCache:l,useFSCache:h,cacheDir:M,useCustomCache:!1,customCache:null};function T(e){return 0===Object.keys(e).length}},"./src/generation/configuration_utils.js":
+  \********************/(e,t,n)=>{n.r(t),n.d(t,{apis:()=>_,env:()=>b});var r=n(/*! fs */"fs"),o=n(/*! path */"path"),s=n(/*! url */"url");const a="undefined"!=typeof self,i=a&&"DedicatedWorkerGlobalScope"===self.constructor.name,l=a&&"caches"in self,c="undefined"!=typeof navigator&&"gpu"in navigator,d="undefined"!=typeof navigator&&"ml"in navigator,u="undefined"!=typeof process,p=u&&"node"===process?.release?.name,h=!T(r.default),m=!T(o.default),_=Object.freeze({IS_BROWSER_ENV:a,IS_WEBWORKER_ENV:i,IS_WEB_CACHE_AVAILABLE:l,IS_WEBGPU_AVAILABLE:c,IS_WEBNN_AVAILABLE:d,IS_PROCESS_AVAILABLE:u,IS_NODE_ENV:p,IS_FS_AVAILABLE:h,IS_PATH_AVAILABLE:m}),f=h&&m,g=f?o.default.dirname(o.default.dirname(s.default.fileURLToPath(import.meta.url))):"./",M=f?o.default.join(g,"/.cache/"):null,w="/models/",b={version:"3.0.0-alpha.13",backends:{onnx:{}},allowRemoteModels:!0,remoteHost:"https://huggingface.co/",remotePathTemplate:"{model}/resolve/{revision}/",allowLocalModels:!a,localModelPath:f?o.default.join(g,w):w,useFS:h,useBrowserCache:l,useFSCache:h,cacheDir:M,useCustomCache:!1,customCache:null};function T(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:()=>o});var r=n(/*! ../utils/core.js */"./src/utils/core.js");class o{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":
@@ -109,7 +109,7 @@ import*as e from"fs";import*as t from"onnxruntime-node";import*as n from"path";i
   \*************************************/(e,t,n)=>{n.r(t),n.d(t,{BaseStreamer:()=>a,TextStreamer:()=>l,WhisperTextStreamer:()=>c});var r=n(/*! ../utils/core.js */"./src/utils/core.js"),o=n(/*! ../tokenizers.js */"./src/tokenizers.js"),s=n(/*! ../env.js */"./src/env.js");class a{put(e){throw Error("Not implemented")}end(){throw Error("Not implemented")}}const i=s.apis.IS_PROCESS_AVAILABLE?e=>process.stdout.write(e):e=>console.log(e);class l extends a{constructor(e,{skip_prompt:t=!1,callback_function:n=null,token_callback_function:r=null,decode_kwargs:o={},...s}={}){super(),this.tokenizer=e,this.skip_prompt=t,this.callback_function=n??i,this.token_callback_function=r,this.decode_kwargs={...o,...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,o.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===i&&s.apis.IS_PROCESS_AVAILABLE&&this.callback_function?.("\n")}}class c extends l{constructor(e,{skip_prompt:t=!1,callback_function:n=null,token_callback_function:r=null,on_chunk_start:o=null,on_chunk_end:s=null,on_finalize:a=null,time_precision:i=.02,skip_special_tokens:l=!0,decode_kwargs:c={}}={}){super(e,{skip_prompt:t,callback_function:n,token_callback_function:r,decode_kwargs:{skip_special_tokens:l,...c}}),this.timestamp_begin=e.timestamp_begin,this.on_chunk_start=o,this.on_chunk_end=s,this.on_finalize=a,this.time_precision=i,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:()=>nn,ASTModel:()=>tn,ASTPreTrainedModel:()=>en,AlbertForMaskedLM:()=>ut,AlbertForQuestionAnswering:()=>dt,AlbertForSequenceClassification:()=>ct,AlbertModel:()=>lt,AlbertPreTrainedModel:()=>it,AutoModel:()=>ci,AutoModelForAudioClassification:()=>Pi,AutoModelForAudioFrameClassification:()=>Si,AutoModelForCTC:()=>Ci,AutoModelForCausalLM:()=>fi,AutoModelForDepthEstimation:()=>zi,AutoModelForDocumentQuestionAnswering:()=>Ai,AutoModelForImageClassification:()=>bi,AutoModelForImageFeatureExtraction:()=>Ii,AutoModelForImageMatting:()=>Ei,AutoModelForImageSegmentation:()=>Ti,AutoModelForImageToImage:()=>Li,AutoModelForMaskGeneration:()=>Fi,AutoModelForMaskedLM:()=>gi,AutoModelForObjectDetection:()=>yi,AutoModelForQuestionAnswering:()=>Mi,AutoModelForSemanticSegmentation:()=>xi,AutoModelForSeq2SeqLM:()=>pi,AutoModelForSequenceClassification:()=>di,AutoModelForSpeechSeq2Seq:()=>hi,AutoModelForTextToSpectrogram:()=>mi,AutoModelForTextToWaveform:()=>_i,AutoModelForTokenClassification:()=>ui,AutoModelForVision2Seq:()=>wi,AutoModelForXVector:()=>vi,AutoModelForZeroShotObjectDetection:()=>ki,BartForConditionalGeneration:()=>yt,BartForSequenceClassification:()=>kt,BartModel:()=>xt,BartPretrainedModel:()=>Tt,BaseModelOutput:()=>q,BeitForImageClassification:()=>Qr,BeitModel:()=>Xr,BeitPreTrainedModel:()=>Ur,BertForMaskedLM:()=>U,BertForQuestionAnswering:()=>H,BertForSequenceClassification:()=>X,BertForTokenClassification:()=>Q,BertModel:()=>W,BertPreTrainedModel:()=>$,BlenderbotForConditionalGeneration:()=>Lt,BlenderbotModel:()=>Et,BlenderbotPreTrainedModel:()=>At,BlenderbotSmallForConditionalGeneration:()=>Bt,BlenderbotSmallModel:()=>It,BlenderbotSmallPreTrainedModel:()=>zt,BloomForCausalLM:()=>wr,BloomModel:()=>Mr,BloomPreTrainedModel:()=>gr,CLIPModel:()=>mn,CLIPPreTrainedModel:()=>hn,CLIPSegForImageSegmentation:()=>Fn,CLIPSegModel:()=>kn,CLIPSegPreTrainedModel:()=>yn,CLIPTextModelWithProjection:()=>_n,CLIPVisionModelWithProjection:()=>fn,CamembertForMaskedLM:()=>Me,CamembertForQuestionAnswering:()=>Te,CamembertForSequenceClassification:()=>we,CamembertForTokenClassification:()=>be,CamembertModel:()=>ge,CamembertPreTrainedModel:()=>fe,CausalLMOutput:()=>Ri,CausalLMOutputWithPast:()=>Gi,ChineseCLIPModel:()=>xn,ChineseCLIPPreTrainedModel:()=>Tn,ClapAudioModelWithProjection:()=>na,ClapModel:()=>ea,ClapPreTrainedModel:()=>Zs,ClapTextModelWithProjection:()=>ta,CodeGenForCausalLM:()=>Un,CodeGenModel:()=>Wn,CodeGenPreTrainedModel:()=>$n,CohereForCausalLM:()=>Kn,CohereModel:()=>Jn,CoherePreTrainedModel:()=>Yn,ConvBertForMaskedLM:()=>ae,ConvBertForQuestionAnswering:()=>ce,ConvBertForSequenceClassification:()=>ie,ConvBertForTokenClassification:()=>le,ConvBertModel:()=>se,ConvBertPreTrainedModel:()=>oe,ConvNextForImageClassification:()=>Do,ConvNextModel:()=>Oo,ConvNextPreTrainedModel:()=>No,ConvNextV2ForImageClassification:()=>Ro,ConvNextV2Model:()=>jo,ConvNextV2PreTrainedModel:()=>Vo,DPTForDepthEstimation:()=>ko,DPTModel:()=>yo,DPTPreTrainedModel:()=>xo,DebertaForMaskedLM:()=>ke,DebertaForQuestionAnswering:()=>Pe,DebertaForSequenceClassification:()=>Fe,DebertaForTokenClassification:()=>Ce,DebertaModel:()=>ye,DebertaPreTrainedModel:()=>xe,DebertaV2ForMaskedLM:()=>Ae,DebertaV2ForQuestionAnswering:()=>ze,DebertaV2ForSequenceClassification:()=>Ee,DebertaV2ForTokenClassification:()=>Le,DebertaV2Model:()=>Se,DebertaV2PreTrainedModel:()=>ve,DeiTForImageClassification:()=>po,DeiTModel:()=>uo,DeiTPreTrainedModel:()=>co,DepthAnythingForDepthEstimation:()=>Co,DepthAnythingPreTrainedModel:()=>Fo,DetrForObjectDetection:()=>Jr,DetrForSegmentation:()=>Kr,DetrModel:()=>Yr,DetrObjectDetectionOutput:()=>Zr,DetrPreTrainedModel:()=>Hr,DetrSegmentationOutput:()=>eo,Dinov2ForImageClassification:()=>$o,Dinov2Model:()=>qo,Dinov2PreTrainedModel:()=>Go,DistilBertForMaskedLM:()=>Ve,DistilBertForQuestionAnswering:()=>De,DistilBertForSequenceClassification:()=>Ne,DistilBertForTokenClassification:()=>Oe,DistilBertModel:()=>Be,DistilBertPreTrainedModel:()=>Ie,DonutSwinModel:()=>Bo,DonutSwinPreTrainedModel:()=>Io,EfficientNetForImageClassification:()=>ma,EfficientNetModel:()=>ha,EfficientNetPreTrainedModel:()=>pa,ElectraForMaskedLM:()=>pe,ElectraForQuestionAnswering:()=>_e,ElectraForSequenceClassification:()=>he,ElectraForTokenClassification:()=>me,ElectraModel:()=>ue,ElectraPreTrainedModel:()=>de,EsmForMaskedLM:()=>Ge,EsmForSequenceClassification:()=>qe,EsmForTokenClassification:()=>$e,EsmModel:()=>Re,EsmPreTrainedModel:()=>je,FalconForCausalLM:()=>Ks,FalconModel:()=>Js,FalconPreTrainedModel:()=>Ys,FastViTForImageClassification:()=>Er,FastViTModel:()=>Ar,FastViTPreTrainedModel:()=>Sr,Florence2ForConditionalGeneration:()=>pn,Florence2PreTrainedModel:()=>un,GLPNForDepthEstimation:()=>zo,GLPNModel:()=>Lo,GLPNPreTrainedModel:()=>Eo,GPT2LMHeadModel:()=>vn,GPT2Model:()=>Pn,GPT2PreTrainedModel:()=>Cn,GPTBigCodeForCausalLM:()=>qn,GPTBigCodeModel:()=>Gn,GPTBigCodePreTrainedModel:()=>Rn,GPTJForCausalLM:()=>jn,GPTJModel:()=>Vn,GPTJPreTrainedModel:()=>Dn,GPTNeoForCausalLM:()=>In,GPTNeoModel:()=>zn,GPTNeoPreTrainedModel:()=>Ln,GPTNeoXForCausalLM:()=>On,GPTNeoXModel:()=>Nn,GPTNeoXPreTrainedModel:()=>Bn,Gemma2ForCausalLM:()=>or,Gemma2Model:()=>rr,Gemma2PreTrainedModel:()=>nr,GemmaForCausalLM:()=>tr,GemmaModel:()=>er,GemmaPreTrainedModel:()=>Zn,HubertForCTC:()=>Ss,HubertForSequenceClassification:()=>As,HubertModel:()=>vs,HubertPreTrainedModel:()=>Ps,ImageMattingOutput:()=>qi,JAISLMHeadModel:()=>En,JAISModel:()=>An,JAISPreTrainedModel:()=>Sn,LlamaForCausalLM:()=>Hn,LlamaModel:()=>Qn,LlamaPreTrainedModel:()=>Xn,LlavaForConditionalGeneration:()=>cn,LlavaPreTrainedModel:()=>ln,LongT5ForConditionalGeneration:()=>gt,LongT5Model:()=>ft,LongT5PreTrainedModel:()=>_t,M2M100ForConditionalGeneration:()=>rs,M2M100Model:()=>ns,M2M100PreTrainedModel:()=>ts,MBartForCausalLM:()=>St,MBartForConditionalGeneration:()=>Pt,MBartForSequenceClassification:()=>vt,MBartModel:()=>Ct,MBartPreTrainedModel:()=>Ft,MPNetForMaskedLM:()=>Ke,MPNetForQuestionAnswering:()=>tt,MPNetForSequenceClassification:()=>Ze,MPNetForTokenClassification:()=>et,MPNetModel:()=>Je,MPNetPreTrainedModel:()=>Ye,MT5ForConditionalGeneration:()=>bt,MT5Model:()=>wt,MT5PreTrainedModel:()=>Mt,MarianMTModel:()=>es,MarianModel:()=>Zo,MarianPreTrainedModel:()=>Ko,MaskedLMOutput:()=>Vi,MistralForCausalLM:()=>Us,MistralModel:()=>Ws,MistralPreTrainedModel:()=>$s,MobileBertForMaskedLM:()=>Xe,MobileBertForQuestionAnswering:()=>He,MobileBertForSequenceClassification:()=>Qe,MobileBertModel:()=>Ue,MobileBertPreTrainedModel:()=>We,MobileNetV1ForImageClassification:()=>Ta,MobileNetV1Model:()=>ba,MobileNetV1PreTrainedModel:()=>wa,MobileNetV2ForImageClassification:()=>ka,MobileNetV2Model:()=>ya,MobileNetV2PreTrainedModel:()=>xa,MobileNetV3ForImageClassification:()=>Pa,MobileNetV3Model:()=>Ca,MobileNetV3PreTrainedModel:()=>Fa,MobileNetV4ForImageClassification:()=>Aa,MobileNetV4Model:()=>Sa,MobileNetV4PreTrainedModel:()=>va,MobileViTForImageClassification:()=>Nr,MobileViTModel:()=>Br,MobileViTPreTrainedModel:()=>Ir,MobileViTV2ForImageClassification:()=>Vr,MobileViTV2Model:()=>Dr,MobileViTV2PreTrainedModel:()=>Or,ModelOutput:()=>G,Moondream1ForConditionalGeneration:()=>dn,MptForCausalLM:()=>xr,MptModel:()=>Tr,MptPreTrainedModel:()=>br,MusicgenForCausalLM:()=>ga,MusicgenForConditionalGeneration:()=>Ma,MusicgenModel:()=>fa,MusicgenPreTrainedModel:()=>_a,NomicBertModel:()=>J,NomicBertPreTrainedModel:()=>Y,OPTForCausalLM:()=>Fr,OPTModel:()=>kr,OPTPreTrainedModel:()=>yr,OpenELMForCausalLM:()=>ir,OpenELMModel:()=>ar,OpenELMPreTrainedModel:()=>sr,OwlViTForObjectDetection:()=>Gr,OwlViTModel:()=>Rr,OwlViTPreTrainedModel:()=>jr,Owlv2ForObjectDetection:()=>Wr,Owlv2Model:()=>$r,Owlv2PreTrainedModel:()=>qr,Phi3ForCausalLM:()=>fr,Phi3Model:()=>_r,Phi3PreTrainedModel:()=>mr,PhiForCausalLM:()=>hr,PhiModel:()=>pr,PhiPreTrainedModel:()=>ur,PreTrainedModel:()=>R,PretrainedMixin:()=>Ea,PyAnnoteForAudioFrameClassification:()=>us,PyAnnoteModel:()=>ds,PyAnnotePreTrainedModel:()=>cs,QuestionAnsweringModelOutput:()=>ji,Qwen2ForCausalLM:()=>dr,Qwen2Model:()=>cr,Qwen2PreTrainedModel:()=>lr,RTDetrForObjectDetection:()=>ro,RTDetrModel:()=>no,RTDetrObjectDetectionOutput:()=>oo,RTDetrPreTrainedModel:()=>to,ResNetForImageClassification:()=>_o,ResNetModel:()=>mo,ResNetPreTrainedModel:()=>ho,RoFormerForMaskedLM:()=>ee,RoFormerForQuestionAnswering:()=>re,RoFormerForSequenceClassification:()=>te,RoFormerForTokenClassification:()=>ne,RoFormerModel:()=>Z,RoFormerPreTrainedModel:()=>K,RobertaForMaskedLM:()=>Dt,RobertaForQuestionAnswering:()=>Rt,RobertaForSequenceClassification:()=>Vt,RobertaForTokenClassification:()=>jt,RobertaModel:()=>Ot,RobertaPreTrainedModel:()=>Nt,SamImageSegmentationOutput:()=>Jo,SamModel:()=>Yo,SamPreTrainedModel:()=>Ho,SapiensForDepthEstimation:()=>So,SapiensForNormalEstimation:()=>Ao,SapiensForSemanticSegmentation:()=>vo,SapiensPreTrainedModel:()=>Po,SegformerForImageClassification:()=>ia,SegformerForSemanticSegmentation:()=>la,SegformerModel:()=>aa,SegformerPreTrainedModel:()=>sa,Seq2SeqLMOutput:()=>Bi,SequenceClassifierOutput:()=>Ni,SiglipModel:()=>Mn,SiglipPreTrainedModel:()=>gn,SiglipTextModel:()=>wn,SiglipVisionModel:()=>bn,SpeechT5ForSpeechToText:()=>Vs,SpeechT5ForTextToSpeech:()=>js,SpeechT5HifiGan:()=>Rs,SpeechT5Model:()=>Ds,SpeechT5PreTrainedModel:()=>Os,SqueezeBertForMaskedLM:()=>ot,SqueezeBertForQuestionAnswering:()=>at,SqueezeBertForSequenceClassification:()=>st,SqueezeBertModel:()=>rt,SqueezeBertPreTrainedModel:()=>nt,StableLmForCausalLM:()=>ua,StableLmModel:()=>da,StableLmPreTrainedModel:()=>ca,Starcoder2ForCausalLM:()=>Hs,Starcoder2Model:()=>Qs,Starcoder2PreTrainedModel:()=>Xs,Swin2SRForImageSuperResolution:()=>To,Swin2SRModel:()=>bo,Swin2SRPreTrainedModel:()=>wo,SwinForImageClassification:()=>Mo,SwinModel:()=>go,SwinPreTrainedModel:()=>fo,T5ForConditionalGeneration:()=>mt,T5Model:()=>ht,T5PreTrainedModel:()=>pt,TableTransformerForObjectDetection:()=>io,TableTransformerModel:()=>ao,TableTransformerObjectDetectionOutput:()=>lo,TableTransformerPreTrainedModel:()=>so,TokenClassifierOutput:()=>Di,TrOCRForCausalLM:()=>qs,TrOCRPreTrainedModel:()=>Gs,UniSpeechForCTC:()=>fs,UniSpeechForSequenceClassification:()=>gs,UniSpeechModel:()=>_s,UniSpeechPreTrainedModel:()=>ms,UniSpeechSatForAudioFrameClassification:()=>xs,UniSpeechSatForCTC:()=>bs,UniSpeechSatForSequenceClassification:()=>Ts,UniSpeechSatModel:()=>ws,UniSpeechSatPreTrainedModel:()=>Ms,ViTForImageClassification:()=>vr,ViTModel:()=>Pr,ViTPreTrainedModel:()=>Cr,VisionEncoderDecoderModel:()=>an,VitMatteForImageMatting:()=>zr,VitMattePreTrainedModel:()=>Lr,VitsModel:()=>oa,VitsModelOutput:()=>$i,VitsPreTrainedModel:()=>ra,Wav2Vec2BertForCTC:()=>Fs,Wav2Vec2BertForSequenceClassification:()=>Cs,Wav2Vec2BertModel:()=>ks,Wav2Vec2BertPreTrainedModel:()=>ys,Wav2Vec2ForAudioFrameClassification:()=>ls,Wav2Vec2ForCTC:()=>as,Wav2Vec2ForSequenceClassification:()=>is,Wav2Vec2Model:()=>ss,Wav2Vec2PreTrainedModel:()=>os,WavLMForAudioFrameClassification:()=>Ns,WavLMForCTC:()=>zs,WavLMForSequenceClassification:()=>Is,WavLMForXVector:()=>Bs,WavLMModel:()=>Ls,WavLMPreTrainedModel:()=>Es,WeSpeakerResNetModel:()=>hs,WeSpeakerResNetPreTrainedModel:()=>ps,WhisperForConditionalGeneration:()=>sn,WhisperModel:()=>on,WhisperPreTrainedModel:()=>rn,XLMForQuestionAnswering:()=>Xt,XLMForSequenceClassification:()=>Wt,XLMForTokenClassification:()=>Ut,XLMModel:()=>qt,XLMPreTrainedModel:()=>Gt,XLMRobertaForMaskedLM:()=>Yt,XLMRobertaForQuestionAnswering:()=>Zt,XLMRobertaForSequenceClassification:()=>Jt,XLMRobertaForTokenClassification:()=>Kt,XLMRobertaModel:()=>Ht,XLMRobertaPreTrainedModel:()=>Qt,XLMWithLMHeadModel:()=>$t,XVectorOutput:()=>Oi,YolosForObjectDetection:()=>Xo,YolosModel:()=>Uo,YolosObjectDetectionOutput:()=>Qo,YolosPreTrainedModel:()=>Wo});var r=n(/*! ./configs.js */"./src/configs.js"),o=n(/*! ./backends/onnx.js */"./src/backends/onnx.js"),s=n(/*! ./utils/dtypes.js */"./src/utils/dtypes.js"),a=n(/*! ./utils/generic.js */"./src/utils/generic.js"),i=n(/*! ./utils/core.js */"./src/utils/core.js"),l=n(/*! ./utils/hub.js */"./src/utils/hub.js"),c=n(/*! ./generation/logits_process.js */"./src/generation/logits_process.js"),d=n(/*! ./generation/configuration_utils.js */"./src/generation/configuration_utils.js"),u=n(/*! ./utils/tensor.js */"./src/utils/tensor.js"),p=n(/*! ./utils/maths.js */"./src/utils/maths.js"),h=n(/*! ./generation/stopping_criteria.js */"./src/generation/stopping_criteria.js"),m=n(/*! ./generation/logits_sampler.js */"./src/generation/logits_sampler.js"),_=n(/*! ./env.js */"./src/env.js"),f=n(/*! ./models/whisper/generation_whisper.js */"./src/models/whisper/generation_whisper.js"),g=n(/*! ./models/whisper/common_whisper.js */"./src/models/whisper/common_whisper.js");const M=0,w=1,b=2,T=3,x=4,y=5,k=6,F=7,C=new Map,P=new Map,v=new Map;async function S(e,t,n){return Object.fromEntries(await Promise.all(Object.keys(t).map((async a=>{const{buffer:i,session_options:c}=await async function(e,t,n){let a=n.device;a&&"string"!=typeof a&&(a.hasOwnProperty(t)?a=a[t]:(console.warn(`device not specified for "${t}". Using the default device.`),a=null));const i=a??(_.apis.IS_NODE_ENV?"cpu":"wasm"),c=(0,o.deviceToExecutionProviders)(i);let d=n.dtype;"string"!=typeof d&&(d&&d.hasOwnProperty(t)?d=d[t]:(d=s.DEFAULT_DEVICE_DTYPE_MAPPING[i]??s.DATA_TYPES.fp32,console.warn(`dtype not specified for "${t}". Using the default dtype (${d}) for this device (${i}).`)));const u=d;if(!s.DEFAULT_DTYPE_SUFFIX_MAPPING.hasOwnProperty(u))throw new Error(`Invalid dtype: ${u}. Should be one of: ${Object.keys(s.DATA_TYPES).join(", ")}`);if(u===s.DATA_TYPES.fp16&&"webgpu"===i&&!await(0,s.isWebGpuFp16Supported)())throw new Error(`The device (${i}) does not support fp16.`);const p=s.DEFAULT_DTYPE_SUFFIX_MAPPING[u],h=`${n.subfolder??""}/${t}${p}.onnx`,m={...n.session_options}??{};m.executionProviders??=c;const f=(0,l.getModelFile)(e,h,!0,n);let g=[];if(n.use_external_data_format&&(!0===n.use_external_data_format||"object"==typeof n.use_external_data_format&&n.use_external_data_format.hasOwnProperty(t)&&!0===n.use_external_data_format[t])){if(_.apis.IS_NODE_ENV)throw new Error("External data format is not yet supported in Node.js");const r=`${t}${p}.onnx_data`,o=`${n.subfolder??""}/${r}`;g.push(new Promise((async(t,s)=>{const a=await(0,l.getModelFile)(e,o,!0,n);t({path:r,data:a})})))}else void 0!==m.externalData&&(g=m.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(g.length>0&&(m.externalData=await Promise.all(g)),"webgpu"===i){const e=(0,r.getKeyValueShapes)(n.config,{prefix:"present"});if(Object.keys(e).length>0&&!(0,o.isONNXProxy)()){const t={};for(const n in e)t[n]="gpu-buffer";m.preferredOutputLocation=t}}return{buffer:await f,session_options:m}}(e,t[a],n);return[a,await(0,o.createInferenceSession)(i,c)]}))))}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 u.Tensor?n[s]=(0,o.isONNXProxy)()?e.clone():e:r.push(s)}if(r.length>0)throw new Error(`An error occurred during model execution: "Missing the following inputs: ${r.join(", ")}.`);const s=Object.keys(t).length,a=e.inputNames.length;if(s>a){let n=Object.keys(t).filter((t=>!e.inputNames.includes(t)));console.warn(`WARNING: Too many inputs were provided (${s} > ${a}). The following inputs will be ignored: "${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=E(r),r}catch(e){throw console.error(`An error occurred during model execution: "${e}".`),console.error("Inputs given to model:",n),e}}function E(e){for(let t in e)(0,o.isONNXTensor)(e[t])?e[t]=new u.Tensor(e[t]):"object"==typeof e[t]&&E(e[t]);return e}function L(e){if(e instanceof u.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 u.Tensor("int64",BigInt64Array.from(e.flat().map((e=>BigInt(e)))),[e.length,e[0].length])}return new u.Tensor("int64",BigInt64Array.from(e.map((e=>BigInt(e)))),[1,e.length])}function z(e){return new u.Tensor("bool",[e],[1])}async function I(e,t){let{encoder_outputs:n,input_ids:r,decoder_input_ids:o,...s}=t;if(!n){const r=(0,i.pick)(t,e.sessions.model.inputNames);n=(await B(e,r)).last_hidden_state}s.input_ids=o,s.encoder_hidden_states=n,e.sessions.decoder_model_merged.inputNames.includes("encoder_attention_mask")&&(s.encoder_attention_mask=t.attention_mask);return await N(e,s,!0)}async function B(e,t){const n=e.sessions.model,r=(0,i.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 u.Tensor("int64",new BigInt64Array(r.input_ids.data.length),r.input_ids.dims)),await A(n,r)}async function N(e,t,n=!1){const r=e.sessions[n?"decoder_model_merged":"model"],{past_key_values:o,...s}=t;r.inputNames.includes("use_cache_branch")&&(s.use_cache_branch=z(!!o)),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:o}=e,[s,a]=o.dims,i=new BigInt64Array(o.data.length);for(let e=0;e<s;++e){const t=e*a;let n=BigInt(0);for(let e=0;e<a;++e){const r=t+e;0n===o.data[r]?i[r]=BigInt(1):(i[r]=n,n+=o.data[r])}}let l=new u.Tensor("int64",i,o.dims);if(t){const e=-(n??r).dims.at(1);l=l.slice(null,[e,null])}return l}(s,o)),e.addPastKeyValues(s,o);const a=(0,i.pick)(s,r.inputNames);return await A(r,a)}async function O(e,{input_ids:t=null,attention_mask:n=null,pixel_values:r=null,position_ids:o=null,inputs_embeds:s=null,past_key_values:a=null,generation_config:i=null,logits_processor:l=null,...c}){if(!s)if(s=await e.encode_text({input_ids:t}),r&&1!==t.dims[1]){const o=await e.encode_image({pixel_values:r});({inputs_embeds:s,attention_mask:n}=e._merge_input_ids_with_image_features({image_features:o,inputs_embeds:s,input_ids:t,attention_mask:n}))}else if(a&&r&&1===t.dims[1]){const e=t.dims[1],r=Object.values(a)[0].dims.at(-2);n=(0,u.cat)([(0,u.ones)([t.dims[0],r]),n.slice(null,[n.dims[1]-e,n.dims[1]])],1)}return await N(e,{inputs_embeds:s,past_key_values:a,attention_mask:n,position_ids:o,generation_config:i,logits_processor:l},!0)}function D(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:o}=n;if(o&&o.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 o=e.config.num_image_tokens;if(!o)throw new Error("`num_image_tokens` is missing in the model configuration.");const s=r.dims[1]-(t-o);n.input_ids=r.slice(null,[-s,null]),n.attention_mask=(0,u.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:L(t)}}function j(e,...t){return e.config.is_encoder_decoder?V(e,...t):D(e,...t)}class R extends a.Callable{main_input_name="input_ids";forward_params=["input_ids","attention_mask"];constructor(e,t){super(),this.config=e,this.sessions=t;const n=v.get(this.constructor),r=C.get(n);switch(this.can_generate=!1,this._forward=null,this._prepare_inputs_for_generation=null,r){case x:this.can_generate=!0,this._forward=N,this._prepare_inputs_for_generation=D;break;case b:case T:case F:this.can_generate=!0,this._forward=I,this._prepare_inputs_for_generation=V;break;case w:this._forward=I;break;case k:this.can_generate=!0,this._forward=O,this._prepare_inputs_for_generation=j;break;default:this._forward=B}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:o=null,local_files_only:s=!1,revision:a="main",model_file_name:i=null,subfolder:c="onnx",device:d=null,dtype:u=null,use_external_data_format:p=null,session_options:h={}}={}){let m={progress_callback:t,config:n,cache_dir:o,local_files_only:s,revision:a,model_file_name:i,subfolder:c,device:d,dtype:u,use_external_data_format:p,session_options:h};const _=v.get(this),f=C.get(_);let g;if(n=m.config=await r.AutoConfig.from_pretrained(e,m),f===x)g=await Promise.all([S(e,{model:m.model_file_name??"model"},m),(0,l.getModelJSON)(e,"generation_config.json",!1,m)]);else if(f===b||f===T)g=await Promise.all([S(e,{model:"encoder_model",decoder_model_merged:"decoder_model_merged"},m),(0,l.getModelJSON)(e,"generation_config.json",!1,m)]);else if(f===y)g=await Promise.all([S(e,{model:"vision_encoder",prompt_encoder_mask_decoder:"prompt_encoder_mask_decoder"},m)]);else if(f===w)g=await Promise.all([S(e,{model:"encoder_model",decoder_model_merged:"decoder_model_merged"},m)]);else if(f===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"),g=await Promise.all([S(e,t,m),(0,l.getModelJSON)(e,"generation_config.json",!1,m)])}else f===F?g=await Promise.all([S(e,{model:"text_encoder",decoder_model_merged:"decoder_model_merged",encodec_decode:"encodec_decode"},m),(0,l.getModelJSON)(e,"generation_config.json",!1,m)]):(f!==M&&console.warn(`Model type for '${_??n?.model_type}' not found, assuming encoder-only architecture. Please report this at https://github.com/xenova/transformers.js/issues/new/choose.`),g=await Promise.all([S(e,{model:m.model_file_name??"model"},m)]));return new this(n,...g)}async _call(e){return await this.forward(e)}async forward(e){return await this._forward(this,e)}_get_logits_warper(e){const t=new c.LogitsProcessorList;return null!==e.temperature&&1!==e.temperature&&t.push(new c.TemperatureLogitsWarper(e.temperature)),null!==e.top_k&&0!==e.top_k&&t.push(new c.TopKLogitsWarper(e.top_k)),null!==e.top_p&&e.top_p<1&&t.push(new c.TopPLogitsWarper(e.top_p)),t}_get_logits_processor(e,t,n=null){const r=new c.LogitsProcessorList;if(null!==e.repetition_penalty&&1!==e.repetition_penalty&&r.push(new c.RepetitionPenaltyLogitsProcessor(e.repetition_penalty)),null!==e.no_repeat_ngram_size&&e.no_repeat_ngram_size>0&&r.push(new c.NoRepeatNGramLogitsProcessor(e.no_repeat_ngram_size)),null!==e.bad_words_ids&&r.push(new c.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 c.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 c.MinNewTokensLengthLogitsProcessor(t,e.min_new_tokens,e.eos_token_id)),null!==e.forced_bos_token_id&&r.push(new c.ForcedBOSTokenLogitsProcessor(e.forced_bos_token_id)),null!==e.forced_eos_token_id&&r.push(new c.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 c.SuppressTokensAtBeginLogitsProcessor(e.begin_suppress_tokens,n))}return null!==e.guidance_scale&&e.guidance_scale>1&&r.push(new c.ClassifierFreeGuidanceLogitsProcessor(e.guidance_scale)),null!==n&&r.extend(n),r}_prepare_generation_config(e,t,n=d.GenerationConfig){const r={...this.config};for(const e of["decoder","generator","text_config"])e in r&&Object.assign(r,r[e]);const o=new n(r);return"generation_config"in this&&Object.assign(o,this.generation_config),e&&Object.assign(o,e),t&&Object.assign(o,(0,i.pick)(t,Object.getOwnPropertyNames(o))),o}_get_stopping_criteria(e,t=null){const n=new h.StoppingCriteriaList;return null!==e.max_length&&n.push(new h.MaxLengthCriteria(e.max_length,this.config.max_position_embeddings??null)),null!==e.eos_token_id&&n.push(new h.EosTokenCriteria(e.eos_token_id)),t&&n.extend(t),n}_validate_model_class(){if(!this.can_generate){const e=[Ra,$a,ja,Ba],t=v.get(this.constructor),n=new Set,r=this.config.model_type;for(const t of e){const e=t.get(r);e&&n.add(e[0])}let o=`The current model class (${t}) is not compatible with \`.generate()\`, as it doesn't have a language model head.`;throw n.size>0&&(o+=` Please use the following class instead: ${[...n].join(", ")}`),Error(o)}}prepare_inputs_for_generation(...e){return this._prepare_inputs_for_generation(this,...e)}_update_model_kwargs_for_generation({generated_input_ids:e,outputs:t,model_inputs:n,is_encoder_decoder:r}){return n.past_key_values=this.getPastKeyValues(t,n.past_key_values),n.input_ids=new u.Tensor("int64",e.flat(),[e.length,1]),r||(n.attention_mask=(0,u.cat)([n.attention_mask,(0,u.ones)([n.attention_mask.dims[0],1])],1)),n.position_ids=null,n}_prepare_model_inputs({inputs:e,bos_token_id:t,model_kwargs:n}){const r=(0,i.pick)(n,this.forward_params),o=this.main_input_name;if(o in r){if(e)throw new Error("`inputs`: {inputs}` were passed alongside {input_name} which is not allowed. Make sure to either pass {inputs} or {input_name}=...")}else r[o]=e;return{inputs_tensor:r[o],model_inputs:r,model_input_name:o}}async _prepare_encoder_decoder_kwargs_for_generation({inputs_tensor:e,model_inputs:t,model_input_name:n,generation_config:r}){if(this.sessions.model.inputNames.includes("inputs_embeds")&&!t.inputs_embeds&&"_prepare_inputs_embeds"in this){const{input_ids:e,pixel_values:n,attention_mask:r,...o}=t,s=await this._prepare_inputs_embeds(t);t={...o,...(0,i.pick)(s,["inputs_embeds","attention_mask"])}}let{last_hidden_state:o}=await B(this,t);if(null!==r.guidance_scale&&r.guidance_scale>1)o=(0,u.cat)([o,(0,u.full_like)(o,0)],0),"attention_mask"in t&&(t.attention_mask=(0,u.cat)([t.attention_mask,(0,u.zeros_like)(t.attention_mask)],0));else if(t.decoder_input_ids){const e=L(t.decoder_input_ids).dims[0];if(e!==o.dims[0]){if(1!==o.dims[0])throw new Error(`The encoder outputs have a different batch size (${o.dims[0]}) than the decoder inputs (${e}).`);o=(0,u.cat)(Array.from({length:e},(()=>o)),0)}}return t.encoder_outputs=o,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:o,generation_config:s}){let{decoder_input_ids:a,...i}=n;if(a)Array.isArray(a[0])||(a=Array.from({length:e},(()=>a)));else if(r??=o,"musicgen"===this.config.model_type)a=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}`);a=r}else a=Array.from({length:e},(()=>[r]));return a=L(a),n.decoder_attention_mask=(0,u.ones_like)(a),{input_ids:a,model_inputs:i}}async generate({inputs:e=null,generation_config:t=null,logits_processor:n=null,stopping_criteria:r=null,streamer:o=null,...s}){this._validate_model_class(),t=this._prepare_generation_config(t,s);let{inputs_tensor:a,model_inputs:i,model_input_name:l}=this._prepare_model_inputs({inputs:e,model_kwargs:s});const c=this.config.is_encoder_decoder;let d;c&&("encoder_outputs"in i||(i=await this._prepare_encoder_decoder_kwargs_for_generation({inputs_tensor:a,model_inputs:i,model_input_name:l,generation_config:t}))),c?({input_ids:d,model_inputs:i}=this._prepare_decoder_input_ids_for_generation({batch_size:i[l].dims.at(0),model_input_name:l,model_kwargs:i,decoder_start_token_id:t.decoder_start_token_id,bos_token_id:t.bos_token_id,generation_config:t})):d=i[l];let p=d.dims.at(-1);null!==t.max_new_tokens&&(t.max_length=p+t.max_new_tokens);const h=this._get_logits_processor(t,p,n),_=this._get_stopping_criteria(t,r),f=i[l].dims.at(0),g=m.LogitsSampler.getSampler(t),M=new Array(f).fill(0),w=d.tolist();o&&o.put(w);let b=null,T={};for(;;){i=this.prepare_inputs_for_generation(w,i,t);const e=await this.forward(i);if(t.output_attentions&&t.return_dict_in_generate){const t=this.getAttentions(e);for(const e in t)e in T||(T[e]=[]),T[e].push(t[e])}const n=h(w,e.logits.slice(null,-1,null)),r=[];for(let e=0;e<n.dims.at(0);++e){const t=n[e],o=await g(t);for(const[t,n]of o){const o=BigInt(t);M[e]+=n,w[e].push(o),r.push([o]);break}}o&&o.put(r);if(_(w).every((e=>e))){t.return_dict_in_generate&&(b=this.getPastKeyValues(e,i.past_key_values,!1));break}i=this._update_model_kwargs_for_generation({generated_input_ids:r,outputs:e,model_inputs:i,is_encoder_decoder:c})}o&&o.end();const x=new u.Tensor("int64",w.flat(),[w.length,w[0].length]);return t.return_dict_in_generate?{sequences:x,past_key_values:b,...T}:x}getPastKeyValues(e,t,n=!0){const r=Object.create(null);for(const o in e)if(o.startsWith("present")){const s=o.replace("present","past_key_values");if(t&&o.includes("encoder"))r[s]=t[s];else{if(n&&t){const e=t[s];"gpu-buffer"===e.location&&e.dispose()}r[s]=e[o]}}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.custom_config.kv_cache_dtype??"float32",n="float16"===t?new Uint16Array:[],o=(0,r.getKeyValueShapes)(this.config);for(const r in o)e[r]=new u.Tensor(t,n,o[r])}}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 q 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 $ extends R{}class W extends ${}class U extends ${async _call(e){return new Vi(await super._call(e))}}class X extends ${async _call(e){return new Ni(await super._call(e))}}class Q extends ${async _call(e){return new Di(await super._call(e))}}class H extends ${async _call(e){return new ji(await super._call(e))}}class Y extends R{}class J extends Y{}class K extends R{}class Z extends K{}class ee extends K{async _call(e){return new Vi(await super._call(e))}}class te extends K{async _call(e){return new Ni(await super._call(e))}}class ne extends K{async _call(e){return new Di(await super._call(e))}}class re extends K{async _call(e){return new ji(await super._call(e))}}class oe extends R{}class se extends oe{}class ae extends oe{async _call(e){return new Vi(await super._call(e))}}class ie extends oe{async _call(e){return new Ni(await super._call(e))}}class le extends oe{async _call(e){return new Di(await super._call(e))}}class ce extends oe{async _call(e){return new ji(await super._call(e))}}class de extends R{}class ue extends de{}class pe extends de{async _call(e){return new Vi(await super._call(e))}}class he extends de{async _call(e){return new Ni(await super._call(e))}}class me extends de{async _call(e){return new Di(await super._call(e))}}class _e extends de{async _call(e){return new ji(await super._call(e))}}class fe extends R{}class ge extends fe{}class Me extends fe{async _call(e){return new Vi(await super._call(e))}}class we extends fe{async _call(e){return new Ni(await super._call(e))}}class be extends fe{async _call(e){return new Di(await super._call(e))}}class Te extends fe{async _call(e){return new ji(await super._call(e))}}class xe extends R{}class ye extends xe{}class ke extends xe{async _call(e){return new Vi(await super._call(e))}}class Fe extends xe{async _call(e){return new Ni(await super._call(e))}}class Ce extends xe{async _call(e){return new Di(await super._call(e))}}class Pe extends xe{async _call(e){return new ji(await super._call(e))}}class ve extends R{}class Se extends ve{}class Ae extends ve{async _call(e){return new Vi(await super._call(e))}}class Ee extends ve{async _call(e){return new Ni(await super._call(e))}}class Le extends ve{async _call(e){return new Di(await super._call(e))}}class ze extends ve{async _call(e){return new ji(await super._call(e))}}class Ie extends R{}class Be extends Ie{}class Ne extends Ie{async _call(e){return new Ni(await super._call(e))}}class Oe extends Ie{async _call(e){return new Di(await super._call(e))}}class De extends Ie{async _call(e){return new ji(await super._call(e))}}class Ve extends Ie{async _call(e){return new Vi(await super._call(e))}}class je extends R{}class Re extends je{}class Ge extends je{async _call(e){return new Vi(await super._call(e))}}class qe extends je{async _call(e){return new Ni(await super._call(e))}}class $e extends je{async _call(e){return new Di(await super._call(e))}}class We extends R{}class Ue extends We{}class Xe extends We{async _call(e){return new Vi(await super._call(e))}}class Qe extends We{async _call(e){return new Ni(await super._call(e))}}class He extends We{async _call(e){return new ji(await super._call(e))}}class Ye extends R{}class Je extends Ye{}class Ke extends Ye{async _call(e){return new Vi(await super._call(e))}}class Ze extends Ye{async _call(e){return new Ni(await super._call(e))}}class et extends Ye{async _call(e){return new Di(await super._call(e))}}class tt extends Ye{async _call(e){return new ji(await super._call(e))}}class nt extends R{}class rt extends nt{}class ot extends nt{async _call(e){return new Vi(await super._call(e))}}class st extends nt{async _call(e){return new Ni(await super._call(e))}}class at extends nt{async _call(e){return new ji(await super._call(e))}}class it extends R{}class lt extends it{}class ct extends it{async _call(e){return new Ni(await super._call(e))}}class dt extends it{async _call(e){return new ji(await super._call(e))}}class ut extends it{async _call(e){return new Vi(await super._call(e))}}class pt extends R{forward_params=["input_ids","attention_mask","encoder_outputs","decoder_input_ids","decoder_attention_mask","past_key_values"];constructor(e,t,n){super(e,t),this.generation_config=n}}class ht extends pt{}class mt extends pt{}class _t extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class ft extends _t{}class gt extends _t{}class Mt extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class wt extends Mt{}class bt extends Mt{}class Tt extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class xt extends Tt{}class yt extends Tt{}class kt extends Tt{async _call(e){return new Ni(await super._call(e))}}class Ft extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class Ct extends Ft{}class Pt extends Ft{}class vt extends Ft{async _call(e){return new Ni(await super._call(e))}}class St extends Ft{}class At extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class Et extends At{}class Lt extends At{}class zt extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class It extends zt{}class Bt extends zt{}class Nt extends R{}class Ot extends Nt{}class Dt extends Nt{async _call(e){return new Vi(await super._call(e))}}class Vt extends Nt{async _call(e){return new Ni(await super._call(e))}}class jt extends Nt{async _call(e){return new Di(await super._call(e))}}class Rt extends Nt{async _call(e){return new ji(await super._call(e))}}class Gt extends R{}class qt extends Gt{}class $t extends Gt{async _call(e){return new Vi(await super._call(e))}}class Wt extends Gt{async _call(e){return new Ni(await super._call(e))}}class Ut extends Gt{async _call(e){return new Di(await super._call(e))}}class Xt extends Gt{async _call(e){return new ji(await super._call(e))}}class Qt extends R{}class Ht extends Qt{}class Yt extends Qt{async _call(e){return new Vi(await super._call(e))}}class Jt extends Qt{async _call(e){return new Ni(await super._call(e))}}class Kt extends Qt{async _call(e){return new Di(await super._call(e))}}class Zt extends Qt{async _call(e){return new ji(await super._call(e))}}class en extends R{}class tn extends en{}class nn extends en{}class rn extends R{requires_attention_mask=!1;main_input_name="input_features";forward_params=["input_features","attention_mask","decoder_input_ids","decoder_attention_mask","past_key_values"];constructor(e,t,n){super(e,t),this.generation_config=n}}class on extends rn{}class sn extends rn{_prepare_generation_config(e,t){return super._prepare_generation_config(e,t,f.WhisperGenerationConfig)}_retrieve_init_tokens(e){const t=[e.decoder_start_token_id];let n=e.language;const r=e.task;if(e.is_multilingual){n||(console.warn("No language specified - defaulting to English (en)."),n="en");const o=`<|${(0,g.whisper_language_to_code)(n)}|>`;t.push(e.lang_to_id[o]),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,...o}){t=this._prepare_generation_config(t,o);const s=o.decoder_input_ids??this._retrieve_init_tokens(t);if(t.return_timestamps&&(n??=new c.LogitsProcessorList,n.push(new c.WhisperTimeStampLogitsProcessor(t,s))),t.begin_suppress_tokens&&(n??=new c.LogitsProcessorList,n.push(new c.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 a=await super.generate({inputs:e,generation_config:t,logits_processor:n,decoder_input_ids:s,...o});return t.return_token_timestamps&&(a.token_timestamps=this._extract_token_timestamps(a,t.alignment_heads,t.num_frames)),a}_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 o=this.config.median_filter_width;void 0===o&&(console.warn("Model config has no `median_filter_width`, using default value of 7."),o=7);const s=e.cross_attentions,a=Array.from({length:this.config.decoder_layers},((e,t)=>(0,u.cat)(s.map((e=>e[t])),2))),l=(0,u.stack)(t.map((([e,t])=>{if(e>=a.length)throw new Error(`Layer index ${e} is out of bounds for cross attentions (length ${a.length}).`);return n?a[e].slice(null,t,null,[0,n]):a[e].slice(null,t)}))).transpose(1,0,2,3),[c,d]=(0,u.std_mean)(l,-2,0,!0),h=l.clone();for(let e=0;e<h.dims[0];++e){const t=h[e];for(let n=0;n<t.dims[0];++n){const r=t[n],s=c[e][n][0].data,a=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]-a[e])/s[e];t.set((0,p.medianFilter)(t,o))}}}const m=[(0,u.mean)(h,1)],_=e.sequences.dims,f=new u.Tensor("float32",new Float32Array(_[0]*_[1]),_);for(let e=0;e<_[0];++e){const t=m[e].neg().squeeze_(0),[n,o]=(0,p.dynamic_time_warping)(t.tolist()),s=Array.from({length:n.length-1},((e,t)=>n[t+1]-n[t])),a=(0,i.mergeArrays)([1],s).map((e=>!!e)),l=[];for(let e=0;e<a.length;++e)a[e]&&l.push(o[e]*r);f[e].data.set(l,1)}return f}}class an extends R{main_input_name="pixel_values";forward_params=["pixel_values","input_ids","encoder_hidden_states","past_key_values"];constructor(e,t,n){super(e,t),this.generation_config=n}}class ln extends R{forward_params=["input_ids","pixel_values","attention_mask","position_ids","past_key_values"];constructor(e,t,n){super(e,t),this.generation_config=n}}class cn extends ln{_merge_input_ids_with_image_features({inputs_embeds:e,image_features:t,input_ids:n,attention_mask:r}){const o=this.config.image_token_index,s=n.tolist().map((e=>e.findIndex((e=>e==o)))),a=s.every((e=>-1===e)),i=s.every((e=>-1!==e));if(!a&&!i)throw new Error("Every input should contain either 0 or 1 image token.");if(a)return{inputs_embeds:e,attention_mask:r};const l=[],c=[];for(let n=0;n<s.length;++n){const o=s[n],a=e[n],i=t[n],d=r[n];l.push((0,u.cat)([a.slice([0,o]),i,a.slice([o+1,a.dims[0]])],0)),c.push((0,u.cat)([d.slice([0,o]),(0,u.ones)([i.dims[0]]),d.slice([o+1,d.dims[0]])],0))}return{inputs_embeds:(0,u.stack)(l,0),attention_mask:(0,u.stack)(c,0)}}}class dn extends cn{}class un extends R{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";constructor(e,t,n){super(e,t),this.generation_config=n}}class pn extends un{_merge_input_ids_with_image_features({inputs_embeds:e,image_features:t,input_ids:n,attention_mask:r}){return{inputs_embeds:(0,u.cat)([t,e],1),attention_mask:(0,u.cat)([(0,u.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 o,s;return e&&(o=await this.encode_text({input_ids:e})),t&&(s=await this.encode_image({pixel_values:t})),o&&s?({inputs_embeds:n,attention_mask:r}=this._merge_input_ids_with_image_features({inputs_embeds:o,image_features:s,input_ids:e,attention_mask:r})):n=o||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:o,encoder_outputs:s,past_key_values:a,inputs_embeds:i,decoder_inputs_embeds:l}){if(i||({inputs_embeds:i,attention_mask:n}=await this._prepare_inputs_embeds({input_ids:e,pixel_values:t,inputs_embeds:i,attention_mask:n})),!s){let{last_hidden_state:e}=await B(this,{inputs_embeds:i,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 c={inputs_embeds:l,attention_mask:o,encoder_attention_mask:n,encoder_hidden_states:s,past_key_values:a};return await N(this,c,!0)}}class hn extends R{}class mn extends hn{}class _n extends hn{static async from_pretrained(e,t={}){return t.model_file_name??="text_model",super.from_pretrained(e,t)}}class fn extends hn{static async from_pretrained(e,t={}){return t.model_file_name??="vision_model",super.from_pretrained(e,t)}}class gn extends R{}class Mn extends gn{}class wn extends gn{static async from_pretrained(e,t={}){return t.model_file_name??="text_model",super.from_pretrained(e,t)}}class bn extends hn{static async from_pretrained(e,t={}){return t.model_file_name??="vision_model",super.from_pretrained(e,t)}}class Tn extends R{}class xn extends Tn{}class yn extends R{}class kn extends yn{}class Fn extends yn{}class Cn extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class Pn extends Cn{}class vn extends Cn{}class Sn extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class An extends Sn{}class En extends Sn{}class Ln extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class zn extends Ln{}class In extends Ln{}class Bn extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class Nn extends Bn{}class On extends Bn{}class Dn extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class Vn extends Dn{}class jn extends Dn{}class Rn extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class Gn extends Rn{}class qn extends Rn{}class $n extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class Wn extends $n{}class Un extends $n{}class Xn extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class Qn extends Xn{}class Hn extends Xn{}class Yn extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class Jn extends Yn{}class Kn extends Yn{}class Zn extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class er extends Zn{}class tr extends Zn{}class nr extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class rr extends nr{}class or extends nr{}class sr extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class ar extends sr{}class ir extends sr{}class lr extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class cr extends lr{}class dr extends lr{}class ur extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class pr extends ur{}class hr extends ur{}class mr extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class _r extends mr{}class fr extends mr{}class gr extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class Mr extends gr{}class wr extends gr{}class br extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class Tr extends br{}class xr extends br{}class yr extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class kr extends yr{}class Fr extends yr{}class Cr extends R{}class Pr extends Cr{}class vr extends Cr{async _call(e){return new Ni(await super._call(e))}}class Sr extends R{}class Ar extends Sr{}class Er extends Sr{async _call(e){return new Ni(await super._call(e))}}class Lr extends R{}class zr extends Lr{async _call(e){return new qi(await super._call(e))}}class Ir extends R{}class Br extends Ir{}class Nr extends Ir{async _call(e){return new Ni(await super._call(e))}}class Or extends R{}class Dr extends Or{}class Vr extends Or{async _call(e){return new Ni(await super._call(e))}}class jr extends R{}class Rr extends jr{}class Gr extends jr{}class qr extends R{}class $r extends qr{}class Wr extends qr{}class Ur extends R{}class Xr extends Ur{}class Qr extends Ur{async _call(e){return new Ni(await super._call(e))}}class Hr extends R{}class Yr extends Hr{}class Jr extends Hr{async _call(e){return new Zr(await super._call(e))}}class Kr extends Hr{async _call(e){return new eo(await super._call(e))}}class Zr extends G{constructor({logits:e,pred_boxes:t}){super(),this.logits=e,this.pred_boxes=t}}class eo extends G{constructor({logits:e,pred_boxes:t,pred_masks:n}){super(),this.logits=e,this.pred_boxes=t,this.pred_masks=n}}class to extends R{}class no extends to{}class ro extends to{async _call(e){return new oo(await super._call(e))}}class oo extends G{constructor({logits:e,pred_boxes:t}){super(),this.logits=e,this.pred_boxes=t}}class so extends R{}class ao extends so{}class io extends so{async _call(e){return new lo(await super._call(e))}}class lo extends Zr{}class co extends R{}class uo extends co{}class po extends co{async _call(e){return new Ni(await super._call(e))}}class ho extends R{}class mo extends ho{}class _o extends ho{async _call(e){return new Ni(await super._call(e))}}class fo extends R{}class go extends fo{}class Mo extends fo{async _call(e){return new Ni(await super._call(e))}}class wo extends R{}class bo extends wo{}class To extends wo{}class xo extends R{}class yo extends xo{}class ko extends xo{}class Fo extends R{}class Co extends Fo{}class Po extends R{}class vo extends Po{}class So extends Po{}class Ao extends Po{}class Eo extends R{}class Lo extends Eo{}class zo extends Eo{}class Io extends R{}class Bo extends Io{}class No extends R{}class Oo extends No{}class Do extends No{async _call(e){return new Ni(await super._call(e))}}class Vo extends R{}class jo extends Vo{}class Ro extends Vo{async _call(e){return new Ni(await super._call(e))}}class Go extends R{}class qo extends Go{}class $o extends Go{async _call(e){return new Ni(await super._call(e))}}class Wo extends R{}class Uo extends Wo{}class Xo extends Wo{async _call(e){return new Qo(await super._call(e))}}class Qo extends G{constructor({logits:e,pred_boxes:t}){super(),this.logits=e,this.pred_boxes=t}}class Ho extends R{}class Yo extends Ho{async get_image_embeddings({pixel_values:e}){return await B(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 u.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 Jo(await super._call(e))}}class Jo extends G{constructor({iou_scores:e,pred_masks:t}){super(),this.iou_scores=e,this.pred_masks=t}}class Ko extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class Zo extends Ko{}class es extends Ko{}class ts extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class ns extends ts{}class rs extends ts{}class os extends R{}class ss extends os{}class as extends os{async _call(e){return new Ri(await super._call(e))}}class is extends os{async _call(e){return new Ni(await super._call(e))}}class ls extends os{async _call(e){return new Di(await super._call(e))}}class cs extends R{}class ds extends cs{}class us extends cs{async _call(e){return new Di(await super._call(e))}}class ps extends R{}class hs extends ps{}class ms extends R{}class _s extends ms{}class fs extends ms{async _call(e){return new Ri(await super._call(e))}}class gs extends ms{async _call(e){return new Ni(await super._call(e))}}class Ms extends R{}class ws extends Ms{}class bs extends Ms{async _call(e){return new Ri(await super._call(e))}}class Ts extends Ms{async _call(e){return new Ni(await super._call(e))}}class xs extends Ms{async _call(e){return new Di(await super._call(e))}}class ys extends R{}class ks extends ys{}class Fs extends ys{async _call(e){return new Ri(await super._call(e))}}class Cs extends ys{async _call(e){return new Ni(await super._call(e))}}class Ps extends R{}class vs extends os{}class Ss extends os{async _call(e){return new Ri(await super._call(e))}}class As extends os{async _call(e){return new Ni(await super._call(e))}}class Es extends R{}class Ls extends Es{}class zs extends Es{async _call(e){return new Ri(await super._call(e))}}class Is extends Es{async _call(e){return new Ni(await super._call(e))}}class Bs extends Es{async _call(e){return new Oi(await super._call(e))}}class Ns extends Es{async _call(e){return new Di(await super._call(e))}}class Os extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class Ds extends Os{}class Vs extends Os{}class js extends Os{async generate_speech(e,t,{threshold:n=.5,minlenratio:r=0,maxlenratio:o=20,vocoder:s=null}={}){const a={input_ids:e},{encoder_outputs:i,encoder_attention_mask:l}=await B(this,a),c=i.dims[1]/this.config.reduction_factor,d=Math.floor(c*o),p=Math.floor(c*r),h=this.config.num_mel_bins;let m=[],_=null,f=null,g=0;for(;;){++g;const e=z(!!f);let r;r=f?f.output_sequence_out:new u.Tensor("float32",new Float32Array(h),[1,1,h]);let o={use_cache_branch:e,output_sequence:r,encoder_attention_mask:l,speaker_embeddings:t,encoder_hidden_states:i};this.addPastKeyValues(o,_),f=await A(this.sessions.decoder_model_merged,o),_=this.getPastKeyValues(f,_);const{prob:s,spectrum:a}=f;if(m.push(a),g>=p&&(Array.from(s.data).filter((e=>e>=n)).length>0||g>=d))break}const M=(0,u.cat)(m),{waveform:w}=await A(s.sessions.model,{spectrogram:M});return{spectrogram:M,waveform:w}}}class Rs extends R{main_input_name="spectrogram"}class Gs extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class qs extends Gs{}class $s extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class Ws extends $s{}class Us extends $s{}class Xs extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class Qs extends Xs{}class Hs extends Xs{}class Ys extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class Js extends Ys{}class Ks extends Ys{}class Zs extends R{}class ea extends Zs{}class ta extends Zs{static async from_pretrained(e,t={}){return t.model_file_name??="text_model",super.from_pretrained(e,t)}}class na extends Zs{static async from_pretrained(e,t={}){return t.model_file_name??="audio_model",super.from_pretrained(e,t)}}class ra extends R{}class oa extends ra{async _call(e){return new $i(await super._call(e))}}class sa extends R{}class aa extends sa{}class ia extends sa{}class la extends sa{}class ca extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class da extends ca{}class ua extends ca{}class pa extends R{}class ha extends pa{}class ma extends pa{async _call(e){return new Ni(await super._call(e))}}class _a extends R{}class fa extends _a{}class ga extends _a{}class Ma extends R{forward_params=["input_ids","attention_mask","encoder_outputs","decoder_input_ids","decoder_attention_mask","past_key_values"];constructor(e,t,n){super(e,t),this.generation_config=n}_apply_and_filter_by_delay_pattern_mask(e){const[t,n]=e.dims,r=this.config.decoder.num_codebooks,o=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 a=t%n-Math.floor(t/n)%r;a>0&&a<=o&&(e.data[s++]=e.data[t])}const a=Math.floor(t/r),i=s/(a*r);return new u.Tensor(e.type,e.data.slice(0,s),[a,r,i])}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 wa extends R{}class ba extends wa{}class Ta extends wa{async _call(e){return new Ni(await super._call(e))}}class xa extends R{}class ya extends xa{}class ka extends xa{async _call(e){return new Ni(await super._call(e))}}class Fa extends R{}class Ca extends Fa{}class Pa extends Fa{async _call(e){return new Ni(await super._call(e))}}class va extends R{}class Sa extends va{}class Aa extends va{async _call(e){return new Ni(await super._call(e))}}class Ea{static MODEL_CLASS_MAPPINGS=null;static BASE_IF_FAIL=!1;static async from_pretrained(e,{progress_callback:t=null,config:n=null,cache_dir:o=null,local_files_only:s=!1,revision:a="main",model_file_name:i=null,subfolder:l="onnx",device:c=null,dtype:d=null,use_external_data_format:u=null,session_options:p={}}={}){let h={progress_callback:t,config:n,cache_dir:o,local_files_only:s,revision:a,model_file_name:i,subfolder:l,device:c,dtype:d,use_external_data_format:u,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(let 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 R.from_pretrained(e,h);throw Error(`Unsupported model type: ${h.config.model_type}`)}}const La=new Map([["bert",["BertModel",W]],["nomic_bert",["NomicBertModel",J]],["roformer",["RoFormerModel",Z]],["electra",["ElectraModel",ue]],["esm",["EsmModel",Re]],["convbert",["ConvBertModel",se]],["camembert",["CamembertModel",ge]],["deberta",["DebertaModel",ye]],["deberta-v2",["DebertaV2Model",Se]],["mpnet",["MPNetModel",Je]],["albert",["AlbertModel",lt]],["distilbert",["DistilBertModel",Be]],["roberta",["RobertaModel",Ot]],["xlm",["XLMModel",qt]],["xlm-roberta",["XLMRobertaModel",Ht]],["clap",["ClapModel",ea]],["clip",["CLIPModel",mn]],["clipseg",["CLIPSegModel",kn]],["chinese_clip",["ChineseCLIPModel",xn]],["siglip",["SiglipModel",Mn]],["mobilebert",["MobileBertModel",Ue]],["squeezebert",["SqueezeBertModel",rt]],["wav2vec2",["Wav2Vec2Model",ss]],["wav2vec2-bert",["Wav2Vec2BertModel",ks]],["unispeech",["UniSpeechModel",_s]],["unispeech-sat",["UniSpeechSatModel",ws]],["hubert",["HubertModel",vs]],["wavlm",["WavLMModel",Ls]],["audio-spectrogram-transformer",["ASTModel",tn]],["vits",["VitsModel",oa]],["pyannote",["PyAnnoteModel",ds]],["wespeaker-resnet",["WeSpeakerResNetModel",hs]],["detr",["DetrModel",Yr]],["rt_detr",["RTDetrModel",no]],["table-transformer",["TableTransformerModel",ao]],["vit",["ViTModel",Pr]],["fastvit",["FastViTModel",Ar]],["mobilevit",["MobileViTModel",Br]],["mobilevitv2",["MobileViTV2Model",Dr]],["owlvit",["OwlViTModel",Rr]],["owlv2",["Owlv2Model",$r]],["beit",["BeitModel",Xr]],["deit",["DeiTModel",uo]],["convnext",["ConvNextModel",Oo]],["convnextv2",["ConvNextV2Model",jo]],["dinov2",["Dinov2Model",qo]],["resnet",["ResNetModel",mo]],["swin",["SwinModel",go]],["swin2sr",["Swin2SRModel",bo]],["donut-swin",["DonutSwinModel",Bo]],["yolos",["YolosModel",Uo]],["dpt",["DPTModel",yo]],["glpn",["GLPNModel",Lo]],["hifigan",["SpeechT5HifiGan",Rs]],["efficientnet",["EfficientNetModel",ha]],["mobilenet_v1",["MobileNetV1Model",ba]],["mobilenet_v2",["MobileNetV2Model",ya]],["mobilenet_v3",["MobileNetV3Model",Ca]],["mobilenet_v4",["MobileNetV4Model",Sa]]]),za=new Map([["t5",["T5Model",ht]],["longt5",["LongT5Model",ft]],["mt5",["MT5Model",wt]],["bart",["BartModel",xt]],["mbart",["MBartModel",Ct]],["marian",["MarianModel",Zo]],["whisper",["WhisperModel",on]],["m2m_100",["M2M100Model",ns]],["blenderbot",["BlenderbotModel",Et]],["blenderbot-small",["BlenderbotSmallModel",It]]]),Ia=new Map([["bloom",["BloomModel",Mr]],["jais",["JAISModel",An]],["gpt2",["GPT2Model",Pn]],["gptj",["GPTJModel",Vn]],["gpt_bigcode",["GPTBigCodeModel",Gn]],["gpt_neo",["GPTNeoModel",zn]],["gpt_neox",["GPTNeoXModel",Nn]],["codegen",["CodeGenModel",Wn]],["llama",["LlamaModel",Qn]],["cohere",["CohereModel",Jn]],["gemma",["GemmaModel",er]],["gemma2",["Gemma2Model",rr]],["openelm",["OpenELMModel",ar]],["qwen2",["Qwen2Model",cr]],["phi",["PhiModel",pr]],["phi3",["Phi3Model",_r]],["mpt",["MptModel",Tr]],["opt",["OPTModel",kr]],["mistral",["MistralModel",Ws]],["starcoder2",["Starcoder2Model",Qs]],["falcon",["FalconModel",Js]],["stablelm",["StableLmModel",da]]]),Ba=new Map([["speecht5",["SpeechT5ForSpeechToText",Vs]],["whisper",["WhisperForConditionalGeneration",sn]]]),Na=new Map([["speecht5",["SpeechT5ForTextToSpeech",js]]]),Oa=new Map([["vits",["VitsModel",oa]],["musicgen",["MusicgenForConditionalGeneration",Ma]]]),Da=new Map([["bert",["BertForSequenceClassification",X]],["roformer",["RoFormerForSequenceClassification",te]],["electra",["ElectraForSequenceClassification",he]],["esm",["EsmForSequenceClassification",qe]],["convbert",["ConvBertForSequenceClassification",ie]],["camembert",["CamembertForSequenceClassification",we]],["deberta",["DebertaForSequenceClassification",Fe]],["deberta-v2",["DebertaV2ForSequenceClassification",Ee]],["mpnet",["MPNetForSequenceClassification",Ze]],["albert",["AlbertForSequenceClassification",ct]],["distilbert",["DistilBertForSequenceClassification",Ne]],["roberta",["RobertaForSequenceClassification",Vt]],["xlm",["XLMForSequenceClassification",Wt]],["xlm-roberta",["XLMRobertaForSequenceClassification",Jt]],["bart",["BartForSequenceClassification",kt]],["mbart",["MBartForSequenceClassification",vt]],["mobilebert",["MobileBertForSequenceClassification",Qe]],["squeezebert",["SqueezeBertForSequenceClassification",st]]]),Va=new Map([["bert",["BertForTokenClassification",Q]],["roformer",["RoFormerForTokenClassification",ne]],["electra",["ElectraForTokenClassification",me]],["esm",["EsmForTokenClassification",$e]],["convbert",["ConvBertForTokenClassification",le]],["camembert",["CamembertForTokenClassification",be]],["deberta",["DebertaForTokenClassification",Ce]],["deberta-v2",["DebertaV2ForTokenClassification",Le]],["mpnet",["MPNetForTokenClassification",et]],["distilbert",["DistilBertForTokenClassification",Oe]],["roberta",["RobertaForTokenClassification",jt]],["xlm",["XLMForTokenClassification",Ut]],["xlm-roberta",["XLMRobertaForTokenClassification",Kt]]]),ja=new Map([["t5",["T5ForConditionalGeneration",mt]],["longt5",["LongT5ForConditionalGeneration",gt]],["mt5",["MT5ForConditionalGeneration",bt]],["bart",["BartForConditionalGeneration",yt]],["mbart",["MBartForConditionalGeneration",Pt]],["marian",["MarianMTModel",es]],["m2m_100",["M2M100ForConditionalGeneration",rs]],["blenderbot",["BlenderbotForConditionalGeneration",Lt]],["blenderbot-small",["BlenderbotSmallForConditionalGeneration",Bt]]]),Ra=new Map([["bloom",["BloomForCausalLM",wr]],["gpt2",["GPT2LMHeadModel",vn]],["jais",["JAISLMHeadModel",En]],["gptj",["GPTJForCausalLM",jn]],["gpt_bigcode",["GPTBigCodeForCausalLM",qn]],["gpt_neo",["GPTNeoForCausalLM",In]],["gpt_neox",["GPTNeoXForCausalLM",On]],["codegen",["CodeGenForCausalLM",Un]],["llama",["LlamaForCausalLM",Hn]],["cohere",["CohereForCausalLM",Kn]],["gemma",["GemmaForCausalLM",tr]],["gemma2",["Gemma2ForCausalLM",or]],["openelm",["OpenELMForCausalLM",ir]],["qwen2",["Qwen2ForCausalLM",dr]],["phi",["PhiForCausalLM",hr]],["phi3",["Phi3ForCausalLM",fr]],["mpt",["MptForCausalLM",xr]],["opt",["OPTForCausalLM",Fr]],["mbart",["MBartForCausalLM",St]],["mistral",["MistralForCausalLM",Us]],["starcoder2",["Starcoder2ForCausalLM",Hs]],["falcon",["FalconForCausalLM",Ks]],["trocr",["TrOCRForCausalLM",qs]],["stablelm",["StableLmForCausalLM",ua]]]),Ga=new Map([["bert",["BertForMaskedLM",U]],["roformer",["RoFormerForMaskedLM",ee]],["electra",["ElectraForMaskedLM",pe]],["esm",["EsmForMaskedLM",Ge]],["convbert",["ConvBertForMaskedLM",ae]],["camembert",["CamembertForMaskedLM",Me]],["deberta",["DebertaForMaskedLM",ke]],["deberta-v2",["DebertaV2ForMaskedLM",Ae]],["mpnet",["MPNetForMaskedLM",Ke]],["albert",["AlbertForMaskedLM",ut]],["distilbert",["DistilBertForMaskedLM",Ve]],["roberta",["RobertaForMaskedLM",Dt]],["xlm",["XLMWithLMHeadModel",$t]],["xlm-roberta",["XLMRobertaForMaskedLM",Yt]],["mobilebert",["MobileBertForMaskedLM",Xe]],["squeezebert",["SqueezeBertForMaskedLM",ot]]]),qa=new Map([["bert",["BertForQuestionAnswering",H]],["roformer",["RoFormerForQuestionAnswering",re]],["electra",["ElectraForQuestionAnswering",_e]],["convbert",["ConvBertForQuestionAnswering",ce]],["camembert",["CamembertForQuestionAnswering",Te]],["deberta",["DebertaForQuestionAnswering",Pe]],["deberta-v2",["DebertaV2ForQuestionAnswering",ze]],["mpnet",["MPNetForQuestionAnswering",tt]],["albert",["AlbertForQuestionAnswering",dt]],["distilbert",["DistilBertForQuestionAnswering",De]],["roberta",["RobertaForQuestionAnswering",Rt]],["xlm",["XLMForQuestionAnswering",Xt]],["xlm-roberta",["XLMRobertaForQuestionAnswering",Zt]],["mobilebert",["MobileBertForQuestionAnswering",He]],["squeezebert",["SqueezeBertForQuestionAnswering",at]]]),$a=new Map([["vision-encoder-decoder",["VisionEncoderDecoderModel",an]]]),Wa=new Map([["llava",["LlavaForConditionalGeneration",cn]],["moondream1",["Moondream1ForConditionalGeneration",dn]],["florence2",["Florence2ForConditionalGeneration",pn]]]),Ua=new Map([["vision-encoder-decoder",["VisionEncoderDecoderModel",an]]]),Xa=new Map([["vit",["ViTForImageClassification",vr]],["fastvit",["FastViTForImageClassification",Er]],["mobilevit",["MobileViTForImageClassification",Nr]],["mobilevitv2",["MobileViTV2ForImageClassification",Vr]],["beit",["BeitForImageClassification",Qr]],["deit",["DeiTForImageClassification",po]],["convnext",["ConvNextForImageClassification",Do]],["convnextv2",["ConvNextV2ForImageClassification",Ro]],["dinov2",["Dinov2ForImageClassification",$o]],["resnet",["ResNetForImageClassification",_o]],["swin",["SwinForImageClassification",Mo]],["segformer",["SegformerForImageClassification",ia]],["efficientnet",["EfficientNetForImageClassification",ma]],["mobilenet_v1",["MobileNetV1ForImageClassification",Ta]],["mobilenet_v2",["MobileNetV2ForImageClassification",ka]],["mobilenet_v3",["MobileNetV3ForImageClassification",Pa]],["mobilenet_v4",["MobileNetV4ForImageClassification",Aa]]]),Qa=new Map([["detr",["DetrForObjectDetection",Jr]],["rt_detr",["RTDetrForObjectDetection",ro]],["table-transformer",["TableTransformerForObjectDetection",io]],["yolos",["YolosForObjectDetection",Xo]]]),Ha=new Map([["owlvit",["OwlViTForObjectDetection",Gr]],["owlv2",["Owlv2ForObjectDetection",Wr]]]),Ya=new Map([["detr",["DetrForSegmentation",Kr]],["clipseg",["CLIPSegForImageSegmentation",Fn]]]),Ja=new Map([["segformer",["SegformerForSemanticSegmentation",la]],["sapiens",["SapiensForSemanticSegmentation",vo]]]),Ka=new Map([["sam",["SamModel",Yo]]]),Za=new Map([["wav2vec2",["Wav2Vec2ForCTC",as]],["wav2vec2-bert",["Wav2Vec2BertForCTC",Fs]],["unispeech",["UniSpeechForCTC",fs]],["unispeech-sat",["UniSpeechSatForCTC",bs]],["wavlm",["WavLMForCTC",zs]],["hubert",["HubertForCTC",Ss]]]),ei=new Map([["wav2vec2",["Wav2Vec2ForSequenceClassification",is]],["wav2vec2-bert",["Wav2Vec2BertForSequenceClassification",Cs]],["unispeech",["UniSpeechForSequenceClassification",gs]],["unispeech-sat",["UniSpeechSatForSequenceClassification",Ts]],["wavlm",["WavLMForSequenceClassification",Is]],["hubert",["HubertForSequenceClassification",As]],["audio-spectrogram-transformer",["ASTForAudioClassification",nn]]]),ti=new Map([["wavlm",["WavLMForXVector",Bs]]]),ni=new Map([["unispeech-sat",["UniSpeechSatForAudioFrameClassification",xs]],["wavlm",["WavLMForAudioFrameClassification",Ns]],["wav2vec2",["Wav2Vec2ForAudioFrameClassification",ls]],["pyannote",["PyAnnoteForAudioFrameClassification",us]]]),ri=new Map([["vitmatte",["VitMatteForImageMatting",zr]]]),oi=new Map([["swin2sr",["Swin2SRForImageSuperResolution",To]]]),si=new Map([["dpt",["DPTForDepthEstimation",ko]],["depth_anything",["DepthAnythingForDepthEstimation",Co]],["glpn",["GLPNForDepthEstimation",zo]],["sapiens",["SapiensForDepthEstimation",So]]]),ai=new Map([["clip",["CLIPVisionModelWithProjection",fn]],["siglip",["SiglipVisionModel",bn]]]),ii=[[La,M],[za,w],[Ia,x],[Da,M],[Va,M],[ja,b],[Ba,b],[Ra,x],[Ga,M],[qa,M],[$a,T],[Wa,k],[Xa,M],[Ya,M],[Ja,M],[ri,M],[oi,M],[si,M],[Qa,M],[Ha,M],[Ka,y],[Za,M],[ei,M],[Na,b],[Oa,M],[ti,M],[ni,M],[ai,M]];for(const[e,t]of ii)for(const[n,r]of e.values())C.set(n,t),v.set(r,n),P.set(n,r);const li=[["MusicgenForConditionalGeneration",Ma,F],["CLIPTextModelWithProjection",_n,M],["SiglipTextModel",wn,M],["ClapTextModelWithProjection",ta,M],["ClapAudioModelWithProjection",na,M]];for(const[e,t,n]of li)C.set(e,n),v.set(t,e),P.set(e,t);class ci extends Ea{static MODEL_CLASS_MAPPINGS=ii.map((e=>e[0]));static BASE_IF_FAIL=!0}class di extends Ea{static MODEL_CLASS_MAPPINGS=[Da]}class ui extends Ea{static MODEL_CLASS_MAPPINGS=[Va]}class pi extends Ea{static MODEL_CLASS_MAPPINGS=[ja]}class hi extends Ea{static MODEL_CLASS_MAPPINGS=[Ba]}class mi extends Ea{static MODEL_CLASS_MAPPINGS=[Na]}class _i extends Ea{static MODEL_CLASS_MAPPINGS=[Oa]}class fi extends Ea{static MODEL_CLASS_MAPPINGS=[Ra]}class gi extends Ea{static MODEL_CLASS_MAPPINGS=[Ga]}class Mi extends Ea{static MODEL_CLASS_MAPPINGS=[qa]}class wi extends Ea{static MODEL_CLASS_MAPPINGS=[$a]}class bi extends Ea{static MODEL_CLASS_MAPPINGS=[Xa]}class Ti extends Ea{static MODEL_CLASS_MAPPINGS=[Ya]}class xi extends Ea{static MODEL_CLASS_MAPPINGS=[Ja]}class yi extends Ea{static MODEL_CLASS_MAPPINGS=[Qa]}class ki extends Ea{static MODEL_CLASS_MAPPINGS=[Ha]}class Fi extends Ea{static MODEL_CLASS_MAPPINGS=[Ka]}class Ci extends Ea{static MODEL_CLASS_MAPPINGS=[Za]}class Pi extends Ea{static MODEL_CLASS_MAPPINGS=[ei]}class vi extends Ea{static MODEL_CLASS_MAPPINGS=[ti]}class Si extends Ea{static MODEL_CLASS_MAPPINGS=[ni]}class Ai extends Ea{static MODEL_CLASS_MAPPINGS=[Ua]}class Ei extends Ea{static MODEL_CLASS_MAPPINGS=[ri]}class Li extends Ea{static MODEL_CLASS_MAPPINGS=[oi]}class zi extends Ea{static MODEL_CLASS_MAPPINGS=[si]}class Ii extends Ea{static MODEL_CLASS_MAPPINGS=[ai]}class Bi extends G{constructor({logits:e,past_key_values:t,encoder_outputs:n,decoder_attentions:r=null,cross_attentions:o=null}){super(),this.logits=e,this.past_key_values=t,this.encoder_outputs=n,this.decoder_attentions=r,this.cross_attentions=o}}class Ni extends G{constructor({logits:e}){super(),this.logits=e}}class Oi extends G{constructor({logits:e,embeddings:t}){super(),this.logits=e,this.embeddings=t}}class Di extends G{constructor({logits:e}){super(),this.logits=e}}class Vi extends G{constructor({logits:e}){super(),this.logits=e}}class ji extends G{constructor({start_logits:e,end_logits:t}){super(),this.start_logits=e,this.end_logits=t}}class Ri extends G{constructor({logits:e}){super(),this.logits=e}}class Gi extends G{constructor({logits:e,past_key_values:t}){super(),this.logits=e,this.past_key_values=t}}class qi extends G{constructor({alphas:e}){super(),this.alphas=e}}class $i extends G{constructor({waveform:e,spectrogram:t}){super(),this.waveform=e,this.spectrogram=t}}},"./src/models/whisper/common_whisper.js":
+  \***********************/(e,t,n)=>{n.r(t),n.d(t,{ASTForAudioClassification:()=>nn,ASTModel:()=>tn,ASTPreTrainedModel:()=>en,AlbertForMaskedLM:()=>ut,AlbertForQuestionAnswering:()=>dt,AlbertForSequenceClassification:()=>ct,AlbertModel:()=>lt,AlbertPreTrainedModel:()=>it,AutoModel:()=>di,AutoModelForAudioClassification:()=>vi,AutoModelForAudioFrameClassification:()=>Ai,AutoModelForCTC:()=>Pi,AutoModelForCausalLM:()=>gi,AutoModelForDepthEstimation:()=>Ii,AutoModelForDocumentQuestionAnswering:()=>Ei,AutoModelForImageClassification:()=>Ti,AutoModelForImageFeatureExtraction:()=>Ni,AutoModelForImageMatting:()=>Li,AutoModelForImageSegmentation:()=>xi,AutoModelForImageToImage:()=>zi,AutoModelForMaskGeneration:()=>Ci,AutoModelForMaskedLM:()=>Mi,AutoModelForNormalEstimation:()=>Bi,AutoModelForObjectDetection:()=>ki,AutoModelForQuestionAnswering:()=>wi,AutoModelForSemanticSegmentation:()=>yi,AutoModelForSeq2SeqLM:()=>hi,AutoModelForSequenceClassification:()=>ui,AutoModelForSpeechSeq2Seq:()=>mi,AutoModelForTextToSpectrogram:()=>_i,AutoModelForTextToWaveform:()=>fi,AutoModelForTokenClassification:()=>pi,AutoModelForVision2Seq:()=>bi,AutoModelForXVector:()=>Si,AutoModelForZeroShotObjectDetection:()=>Fi,BartForConditionalGeneration:()=>yt,BartForSequenceClassification:()=>kt,BartModel:()=>xt,BartPretrainedModel:()=>Tt,BaseModelOutput:()=>q,BeitForImageClassification:()=>Qr,BeitModel:()=>Xr,BeitPreTrainedModel:()=>Ur,BertForMaskedLM:()=>U,BertForQuestionAnswering:()=>H,BertForSequenceClassification:()=>X,BertForTokenClassification:()=>Q,BertModel:()=>W,BertPreTrainedModel:()=>$,BlenderbotForConditionalGeneration:()=>Lt,BlenderbotModel:()=>Et,BlenderbotPreTrainedModel:()=>At,BlenderbotSmallForConditionalGeneration:()=>Bt,BlenderbotSmallModel:()=>It,BlenderbotSmallPreTrainedModel:()=>zt,BloomForCausalLM:()=>wr,BloomModel:()=>Mr,BloomPreTrainedModel:()=>gr,CLIPModel:()=>mn,CLIPPreTrainedModel:()=>hn,CLIPSegForImageSegmentation:()=>Fn,CLIPSegModel:()=>kn,CLIPSegPreTrainedModel:()=>yn,CLIPTextModelWithProjection:()=>_n,CLIPVisionModelWithProjection:()=>fn,CamembertForMaskedLM:()=>Me,CamembertForQuestionAnswering:()=>Te,CamembertForSequenceClassification:()=>we,CamembertForTokenClassification:()=>be,CamembertModel:()=>ge,CamembertPreTrainedModel:()=>fe,CausalLMOutput:()=>qi,CausalLMOutputWithPast:()=>$i,ChineseCLIPModel:()=>xn,ChineseCLIPPreTrainedModel:()=>Tn,ClapAudioModelWithProjection:()=>na,ClapModel:()=>ea,ClapPreTrainedModel:()=>Zs,ClapTextModelWithProjection:()=>ta,CodeGenForCausalLM:()=>Un,CodeGenModel:()=>Wn,CodeGenPreTrainedModel:()=>$n,CohereForCausalLM:()=>Kn,CohereModel:()=>Jn,CoherePreTrainedModel:()=>Yn,ConvBertForMaskedLM:()=>ae,ConvBertForQuestionAnswering:()=>ce,ConvBertForSequenceClassification:()=>ie,ConvBertForTokenClassification:()=>le,ConvBertModel:()=>se,ConvBertPreTrainedModel:()=>oe,ConvNextForImageClassification:()=>Do,ConvNextModel:()=>Oo,ConvNextPreTrainedModel:()=>No,ConvNextV2ForImageClassification:()=>Ro,ConvNextV2Model:()=>jo,ConvNextV2PreTrainedModel:()=>Vo,DPTForDepthEstimation:()=>ko,DPTModel:()=>yo,DPTPreTrainedModel:()=>xo,DebertaForMaskedLM:()=>ke,DebertaForQuestionAnswering:()=>Pe,DebertaForSequenceClassification:()=>Fe,DebertaForTokenClassification:()=>Ce,DebertaModel:()=>ye,DebertaPreTrainedModel:()=>xe,DebertaV2ForMaskedLM:()=>Ae,DebertaV2ForQuestionAnswering:()=>ze,DebertaV2ForSequenceClassification:()=>Ee,DebertaV2ForTokenClassification:()=>Le,DebertaV2Model:()=>Se,DebertaV2PreTrainedModel:()=>ve,DeiTForImageClassification:()=>po,DeiTModel:()=>uo,DeiTPreTrainedModel:()=>co,DepthAnythingForDepthEstimation:()=>Co,DepthAnythingPreTrainedModel:()=>Fo,DetrForObjectDetection:()=>Jr,DetrForSegmentation:()=>Kr,DetrModel:()=>Yr,DetrObjectDetectionOutput:()=>Zr,DetrPreTrainedModel:()=>Hr,DetrSegmentationOutput:()=>eo,Dinov2ForImageClassification:()=>$o,Dinov2Model:()=>qo,Dinov2PreTrainedModel:()=>Go,DistilBertForMaskedLM:()=>Ve,DistilBertForQuestionAnswering:()=>De,DistilBertForSequenceClassification:()=>Ne,DistilBertForTokenClassification:()=>Oe,DistilBertModel:()=>Be,DistilBertPreTrainedModel:()=>Ie,DonutSwinModel:()=>Bo,DonutSwinPreTrainedModel:()=>Io,EfficientNetForImageClassification:()=>ma,EfficientNetModel:()=>ha,EfficientNetPreTrainedModel:()=>pa,ElectraForMaskedLM:()=>pe,ElectraForQuestionAnswering:()=>_e,ElectraForSequenceClassification:()=>he,ElectraForTokenClassification:()=>me,ElectraModel:()=>ue,ElectraPreTrainedModel:()=>de,EsmForMaskedLM:()=>Ge,EsmForSequenceClassification:()=>qe,EsmForTokenClassification:()=>$e,EsmModel:()=>Re,EsmPreTrainedModel:()=>je,FalconForCausalLM:()=>Ks,FalconModel:()=>Js,FalconPreTrainedModel:()=>Ys,FastViTForImageClassification:()=>Er,FastViTModel:()=>Ar,FastViTPreTrainedModel:()=>Sr,Florence2ForConditionalGeneration:()=>pn,Florence2PreTrainedModel:()=>un,GLPNForDepthEstimation:()=>zo,GLPNModel:()=>Lo,GLPNPreTrainedModel:()=>Eo,GPT2LMHeadModel:()=>vn,GPT2Model:()=>Pn,GPT2PreTrainedModel:()=>Cn,GPTBigCodeForCausalLM:()=>qn,GPTBigCodeModel:()=>Gn,GPTBigCodePreTrainedModel:()=>Rn,GPTJForCausalLM:()=>jn,GPTJModel:()=>Vn,GPTJPreTrainedModel:()=>Dn,GPTNeoForCausalLM:()=>In,GPTNeoModel:()=>zn,GPTNeoPreTrainedModel:()=>Ln,GPTNeoXForCausalLM:()=>On,GPTNeoXModel:()=>Nn,GPTNeoXPreTrainedModel:()=>Bn,Gemma2ForCausalLM:()=>or,Gemma2Model:()=>rr,Gemma2PreTrainedModel:()=>nr,GemmaForCausalLM:()=>tr,GemmaModel:()=>er,GemmaPreTrainedModel:()=>Zn,HubertForCTC:()=>Ss,HubertForSequenceClassification:()=>As,HubertModel:()=>vs,HubertPreTrainedModel:()=>Ps,ImageMattingOutput:()=>Wi,JAISLMHeadModel:()=>En,JAISModel:()=>An,JAISPreTrainedModel:()=>Sn,LlamaForCausalLM:()=>Hn,LlamaModel:()=>Qn,LlamaPreTrainedModel:()=>Xn,LlavaForConditionalGeneration:()=>cn,LlavaPreTrainedModel:()=>ln,LongT5ForConditionalGeneration:()=>gt,LongT5Model:()=>ft,LongT5PreTrainedModel:()=>_t,M2M100ForConditionalGeneration:()=>rs,M2M100Model:()=>ns,M2M100PreTrainedModel:()=>ts,MBartForCausalLM:()=>St,MBartForConditionalGeneration:()=>Pt,MBartForSequenceClassification:()=>vt,MBartModel:()=>Ct,MBartPreTrainedModel:()=>Ft,MPNetForMaskedLM:()=>Ke,MPNetForQuestionAnswering:()=>tt,MPNetForSequenceClassification:()=>Ze,MPNetForTokenClassification:()=>et,MPNetModel:()=>Je,MPNetPreTrainedModel:()=>Ye,MT5ForConditionalGeneration:()=>bt,MT5Model:()=>wt,MT5PreTrainedModel:()=>Mt,MarianMTModel:()=>es,MarianModel:()=>Zo,MarianPreTrainedModel:()=>Ko,MaskedLMOutput:()=>Ri,MistralForCausalLM:()=>Us,MistralModel:()=>Ws,MistralPreTrainedModel:()=>$s,MobileBertForMaskedLM:()=>Xe,MobileBertForQuestionAnswering:()=>He,MobileBertForSequenceClassification:()=>Qe,MobileBertModel:()=>Ue,MobileBertPreTrainedModel:()=>We,MobileNetV1ForImageClassification:()=>Ta,MobileNetV1Model:()=>ba,MobileNetV1PreTrainedModel:()=>wa,MobileNetV2ForImageClassification:()=>ka,MobileNetV2Model:()=>ya,MobileNetV2PreTrainedModel:()=>xa,MobileNetV3ForImageClassification:()=>Pa,MobileNetV3Model:()=>Ca,MobileNetV3PreTrainedModel:()=>Fa,MobileNetV4ForImageClassification:()=>Aa,MobileNetV4Model:()=>Sa,MobileNetV4PreTrainedModel:()=>va,MobileViTForImageClassification:()=>Nr,MobileViTModel:()=>Br,MobileViTPreTrainedModel:()=>Ir,MobileViTV2ForImageClassification:()=>Vr,MobileViTV2Model:()=>Dr,MobileViTV2PreTrainedModel:()=>Or,ModelOutput:()=>G,Moondream1ForConditionalGeneration:()=>dn,MptForCausalLM:()=>xr,MptModel:()=>Tr,MptPreTrainedModel:()=>br,MusicgenForCausalLM:()=>ga,MusicgenForConditionalGeneration:()=>Ma,MusicgenModel:()=>fa,MusicgenPreTrainedModel:()=>_a,NomicBertModel:()=>J,NomicBertPreTrainedModel:()=>Y,OPTForCausalLM:()=>Fr,OPTModel:()=>kr,OPTPreTrainedModel:()=>yr,OpenELMForCausalLM:()=>ir,OpenELMModel:()=>ar,OpenELMPreTrainedModel:()=>sr,OwlViTForObjectDetection:()=>Gr,OwlViTModel:()=>Rr,OwlViTPreTrainedModel:()=>jr,Owlv2ForObjectDetection:()=>Wr,Owlv2Model:()=>$r,Owlv2PreTrainedModel:()=>qr,Phi3ForCausalLM:()=>fr,Phi3Model:()=>_r,Phi3PreTrainedModel:()=>mr,PhiForCausalLM:()=>hr,PhiModel:()=>pr,PhiPreTrainedModel:()=>ur,PreTrainedModel:()=>R,PretrainedMixin:()=>Ea,PyAnnoteForAudioFrameClassification:()=>us,PyAnnoteModel:()=>ds,PyAnnotePreTrainedModel:()=>cs,QuestionAnsweringModelOutput:()=>Gi,Qwen2ForCausalLM:()=>dr,Qwen2Model:()=>cr,Qwen2PreTrainedModel:()=>lr,RTDetrForObjectDetection:()=>ro,RTDetrModel:()=>no,RTDetrObjectDetectionOutput:()=>oo,RTDetrPreTrainedModel:()=>to,ResNetForImageClassification:()=>_o,ResNetModel:()=>mo,ResNetPreTrainedModel:()=>ho,RoFormerForMaskedLM:()=>ee,RoFormerForQuestionAnswering:()=>re,RoFormerForSequenceClassification:()=>te,RoFormerForTokenClassification:()=>ne,RoFormerModel:()=>Z,RoFormerPreTrainedModel:()=>K,RobertaForMaskedLM:()=>Dt,RobertaForQuestionAnswering:()=>Rt,RobertaForSequenceClassification:()=>Vt,RobertaForTokenClassification:()=>jt,RobertaModel:()=>Ot,RobertaPreTrainedModel:()=>Nt,SamImageSegmentationOutput:()=>Jo,SamModel:()=>Yo,SamPreTrainedModel:()=>Ho,SapiensForDepthEstimation:()=>So,SapiensForNormalEstimation:()=>Ao,SapiensForSemanticSegmentation:()=>vo,SapiensPreTrainedModel:()=>Po,SegformerForImageClassification:()=>ia,SegformerForSemanticSegmentation:()=>la,SegformerModel:()=>aa,SegformerPreTrainedModel:()=>sa,Seq2SeqLMOutput:()=>Oi,SequenceClassifierOutput:()=>Di,SiglipModel:()=>Mn,SiglipPreTrainedModel:()=>gn,SiglipTextModel:()=>wn,SiglipVisionModel:()=>bn,SpeechT5ForSpeechToText:()=>Vs,SpeechT5ForTextToSpeech:()=>js,SpeechT5HifiGan:()=>Rs,SpeechT5Model:()=>Ds,SpeechT5PreTrainedModel:()=>Os,SqueezeBertForMaskedLM:()=>ot,SqueezeBertForQuestionAnswering:()=>at,SqueezeBertForSequenceClassification:()=>st,SqueezeBertModel:()=>rt,SqueezeBertPreTrainedModel:()=>nt,StableLmForCausalLM:()=>ua,StableLmModel:()=>da,StableLmPreTrainedModel:()=>ca,Starcoder2ForCausalLM:()=>Hs,Starcoder2Model:()=>Qs,Starcoder2PreTrainedModel:()=>Xs,Swin2SRForImageSuperResolution:()=>To,Swin2SRModel:()=>bo,Swin2SRPreTrainedModel:()=>wo,SwinForImageClassification:()=>Mo,SwinModel:()=>go,SwinPreTrainedModel:()=>fo,T5ForConditionalGeneration:()=>mt,T5Model:()=>ht,T5PreTrainedModel:()=>pt,TableTransformerForObjectDetection:()=>io,TableTransformerModel:()=>ao,TableTransformerObjectDetectionOutput:()=>lo,TableTransformerPreTrainedModel:()=>so,TokenClassifierOutput:()=>ji,TrOCRForCausalLM:()=>qs,TrOCRPreTrainedModel:()=>Gs,UniSpeechForCTC:()=>fs,UniSpeechForSequenceClassification:()=>gs,UniSpeechModel:()=>_s,UniSpeechPreTrainedModel:()=>ms,UniSpeechSatForAudioFrameClassification:()=>xs,UniSpeechSatForCTC:()=>bs,UniSpeechSatForSequenceClassification:()=>Ts,UniSpeechSatModel:()=>ws,UniSpeechSatPreTrainedModel:()=>Ms,ViTForImageClassification:()=>vr,ViTModel:()=>Pr,ViTPreTrainedModel:()=>Cr,VisionEncoderDecoderModel:()=>an,VitMatteForImageMatting:()=>zr,VitMattePreTrainedModel:()=>Lr,VitsModel:()=>oa,VitsModelOutput:()=>Ui,VitsPreTrainedModel:()=>ra,Wav2Vec2BertForCTC:()=>Fs,Wav2Vec2BertForSequenceClassification:()=>Cs,Wav2Vec2BertModel:()=>ks,Wav2Vec2BertPreTrainedModel:()=>ys,Wav2Vec2ForAudioFrameClassification:()=>ls,Wav2Vec2ForCTC:()=>as,Wav2Vec2ForSequenceClassification:()=>is,Wav2Vec2Model:()=>ss,Wav2Vec2PreTrainedModel:()=>os,WavLMForAudioFrameClassification:()=>Ns,WavLMForCTC:()=>zs,WavLMForSequenceClassification:()=>Is,WavLMForXVector:()=>Bs,WavLMModel:()=>Ls,WavLMPreTrainedModel:()=>Es,WeSpeakerResNetModel:()=>hs,WeSpeakerResNetPreTrainedModel:()=>ps,WhisperForConditionalGeneration:()=>sn,WhisperModel:()=>on,WhisperPreTrainedModel:()=>rn,XLMForQuestionAnswering:()=>Xt,XLMForSequenceClassification:()=>Wt,XLMForTokenClassification:()=>Ut,XLMModel:()=>qt,XLMPreTrainedModel:()=>Gt,XLMRobertaForMaskedLM:()=>Yt,XLMRobertaForQuestionAnswering:()=>Zt,XLMRobertaForSequenceClassification:()=>Jt,XLMRobertaForTokenClassification:()=>Kt,XLMRobertaModel:()=>Ht,XLMRobertaPreTrainedModel:()=>Qt,XLMWithLMHeadModel:()=>$t,XVectorOutput:()=>Vi,YolosForObjectDetection:()=>Xo,YolosModel:()=>Uo,YolosObjectDetectionOutput:()=>Qo,YolosPreTrainedModel:()=>Wo});var r=n(/*! ./configs.js */"./src/configs.js"),o=n(/*! ./backends/onnx.js */"./src/backends/onnx.js"),s=n(/*! ./utils/dtypes.js */"./src/utils/dtypes.js"),a=n(/*! ./utils/generic.js */"./src/utils/generic.js"),i=n(/*! ./utils/core.js */"./src/utils/core.js"),l=n(/*! ./utils/hub.js */"./src/utils/hub.js"),c=n(/*! ./generation/logits_process.js */"./src/generation/logits_process.js"),d=n(/*! ./generation/configuration_utils.js */"./src/generation/configuration_utils.js"),u=n(/*! ./utils/tensor.js */"./src/utils/tensor.js"),p=n(/*! ./utils/maths.js */"./src/utils/maths.js"),h=n(/*! ./generation/stopping_criteria.js */"./src/generation/stopping_criteria.js"),m=n(/*! ./generation/logits_sampler.js */"./src/generation/logits_sampler.js"),_=n(/*! ./env.js */"./src/env.js"),f=n(/*! ./models/whisper/generation_whisper.js */"./src/models/whisper/generation_whisper.js"),g=n(/*! ./models/whisper/common_whisper.js */"./src/models/whisper/common_whisper.js");const M=0,w=1,b=2,T=3,x=4,y=5,k=6,F=7,C=new Map,P=new Map,v=new Map;async function S(e,t,n){return Object.fromEntries(await Promise.all(Object.keys(t).map((async a=>{const{buffer:i,session_options:c}=await async function(e,t,n){let a=n.device;a&&"string"!=typeof a&&(a.hasOwnProperty(t)?a=a[t]:(console.warn(`device not specified for "${t}". Using the default device.`),a=null));const i=a??(_.apis.IS_NODE_ENV?"cpu":"wasm"),c=(0,o.deviceToExecutionProviders)(i);let d=n.dtype;"string"!=typeof d&&(d&&d.hasOwnProperty(t)?d=d[t]:(d=s.DEFAULT_DEVICE_DTYPE_MAPPING[i]??s.DATA_TYPES.fp32,console.warn(`dtype not specified for "${t}". Using the default dtype (${d}) for this device (${i}).`)));const u=d;if(!s.DEFAULT_DTYPE_SUFFIX_MAPPING.hasOwnProperty(u))throw new Error(`Invalid dtype: ${u}. Should be one of: ${Object.keys(s.DATA_TYPES).join(", ")}`);if(u===s.DATA_TYPES.fp16&&"webgpu"===i&&!await(0,s.isWebGpuFp16Supported)())throw new Error(`The device (${i}) does not support fp16.`);const p=s.DEFAULT_DTYPE_SUFFIX_MAPPING[u],h=`${n.subfolder??""}/${t}${p}.onnx`,m={...n.session_options}??{};m.executionProviders??=c;const f=(0,l.getModelFile)(e,h,!0,n);let g=[];if(n.use_external_data_format&&(!0===n.use_external_data_format||"object"==typeof n.use_external_data_format&&n.use_external_data_format.hasOwnProperty(t)&&!0===n.use_external_data_format[t])){if(_.apis.IS_NODE_ENV)throw new Error("External data format is not yet supported in Node.js");const r=`${t}${p}.onnx_data`,o=`${n.subfolder??""}/${r}`;g.push(new Promise((async(t,s)=>{const a=await(0,l.getModelFile)(e,o,!0,n);t({path:r,data:a})})))}else void 0!==m.externalData&&(g=m.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(g.length>0&&(m.externalData=await Promise.all(g)),"webgpu"===i){const e=(0,r.getKeyValueShapes)(n.config,{prefix:"present"});if(Object.keys(e).length>0&&!(0,o.isONNXProxy)()){const t={};for(const n in e)t[n]="gpu-buffer";m.preferredOutputLocation=t}}return{buffer:await f,session_options:m}}(e,t[a],n);return[a,await(0,o.createInferenceSession)(i,c)]}))))}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 u.Tensor?n[s]=(0,o.isONNXProxy)()?e.clone():e:r.push(s)}if(r.length>0)throw new Error(`An error occurred during model execution: "Missing the following inputs: ${r.join(", ")}.`);const s=Object.keys(t).length,a=e.inputNames.length;if(s>a){let n=Object.keys(t).filter((t=>!e.inputNames.includes(t)));console.warn(`WARNING: Too many inputs were provided (${s} > ${a}). The following inputs will be ignored: "${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=E(r),r}catch(e){throw console.error(`An error occurred during model execution: "${e}".`),console.error("Inputs given to model:",n),e}}function E(e){for(let t in e)(0,o.isONNXTensor)(e[t])?e[t]=new u.Tensor(e[t]):"object"==typeof e[t]&&E(e[t]);return e}function L(e){if(e instanceof u.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 u.Tensor("int64",BigInt64Array.from(e.flat().map((e=>BigInt(e)))),[e.length,e[0].length])}return new u.Tensor("int64",BigInt64Array.from(e.map((e=>BigInt(e)))),[1,e.length])}function z(e){return new u.Tensor("bool",[e],[1])}async function I(e,t){let{encoder_outputs:n,input_ids:r,decoder_input_ids:o,...s}=t;if(!n){const r=(0,i.pick)(t,e.sessions.model.inputNames);n=(await B(e,r)).last_hidden_state}s.input_ids=o,s.encoder_hidden_states=n,e.sessions.decoder_model_merged.inputNames.includes("encoder_attention_mask")&&(s.encoder_attention_mask=t.attention_mask);return await N(e,s,!0)}async function B(e,t){const n=e.sessions.model,r=(0,i.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 u.Tensor("int64",new BigInt64Array(r.input_ids.data.length),r.input_ids.dims)),await A(n,r)}async function N(e,t,n=!1){const r=e.sessions[n?"decoder_model_merged":"model"],{past_key_values:o,...s}=t;r.inputNames.includes("use_cache_branch")&&(s.use_cache_branch=z(!!o)),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:o}=e,[s,a]=o.dims,i=new BigInt64Array(o.data.length);for(let e=0;e<s;++e){const t=e*a;let n=BigInt(0);for(let e=0;e<a;++e){const r=t+e;0n===o.data[r]?i[r]=BigInt(1):(i[r]=n,n+=o.data[r])}}let l=new u.Tensor("int64",i,o.dims);if(t){const e=-(n??r).dims.at(1);l=l.slice(null,[e,null])}return l}(s,o)),e.addPastKeyValues(s,o);const a=(0,i.pick)(s,r.inputNames);return await A(r,a)}async function O(e,{input_ids:t=null,attention_mask:n=null,pixel_values:r=null,position_ids:o=null,inputs_embeds:s=null,past_key_values:a=null,generation_config:i=null,logits_processor:l=null,...c}){if(!s)if(s=await e.encode_text({input_ids:t}),r&&1!==t.dims[1]){const o=await e.encode_image({pixel_values:r});({inputs_embeds:s,attention_mask:n}=e._merge_input_ids_with_image_features({image_features:o,inputs_embeds:s,input_ids:t,attention_mask:n}))}else if(a&&r&&1===t.dims[1]){const e=t.dims[1],r=Object.values(a)[0].dims.at(-2);n=(0,u.cat)([(0,u.ones)([t.dims[0],r]),n.slice(null,[n.dims[1]-e,n.dims[1]])],1)}return await N(e,{inputs_embeds:s,past_key_values:a,attention_mask:n,position_ids:o,generation_config:i,logits_processor:l},!0)}function D(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:o}=n;if(o&&o.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 o=e.config.num_image_tokens;if(!o)throw new Error("`num_image_tokens` is missing in the model configuration.");const s=r.dims[1]-(t-o);n.input_ids=r.slice(null,[-s,null]),n.attention_mask=(0,u.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:L(t)}}function j(e,...t){return e.config.is_encoder_decoder?V(e,...t):D(e,...t)}class R extends a.Callable{main_input_name="input_ids";forward_params=["input_ids","attention_mask"];constructor(e,t){super(),this.config=e,this.sessions=t;const n=v.get(this.constructor),r=C.get(n);switch(this.can_generate=!1,this._forward=null,this._prepare_inputs_for_generation=null,r){case x:this.can_generate=!0,this._forward=N,this._prepare_inputs_for_generation=D;break;case b:case T:case F:this.can_generate=!0,this._forward=I,this._prepare_inputs_for_generation=V;break;case w:this._forward=I;break;case k:this.can_generate=!0,this._forward=O,this._prepare_inputs_for_generation=j;break;default:this._forward=B}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:o=null,local_files_only:s=!1,revision:a="main",model_file_name:i=null,subfolder:c="onnx",device:d=null,dtype:u=null,use_external_data_format:p=null,session_options:h={}}={}){let m={progress_callback:t,config:n,cache_dir:o,local_files_only:s,revision:a,model_file_name:i,subfolder:c,device:d,dtype:u,use_external_data_format:p,session_options:h};const _=v.get(this),f=C.get(_);let g;if(n=m.config=await r.AutoConfig.from_pretrained(e,m),f===x)g=await Promise.all([S(e,{model:m.model_file_name??"model"},m),(0,l.getModelJSON)(e,"generation_config.json",!1,m)]);else if(f===b||f===T)g=await Promise.all([S(e,{model:"encoder_model",decoder_model_merged:"decoder_model_merged"},m),(0,l.getModelJSON)(e,"generation_config.json",!1,m)]);else if(f===y)g=await Promise.all([S(e,{model:"vision_encoder",prompt_encoder_mask_decoder:"prompt_encoder_mask_decoder"},m)]);else if(f===w)g=await Promise.all([S(e,{model:"encoder_model",decoder_model_merged:"decoder_model_merged"},m)]);else if(f===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"),g=await Promise.all([S(e,t,m),(0,l.getModelJSON)(e,"generation_config.json",!1,m)])}else f===F?g=await Promise.all([S(e,{model:"text_encoder",decoder_model_merged:"decoder_model_merged",encodec_decode:"encodec_decode"},m),(0,l.getModelJSON)(e,"generation_config.json",!1,m)]):(f!==M&&console.warn(`Model type for '${_??n?.model_type}' not found, assuming encoder-only architecture. Please report this at https://github.com/xenova/transformers.js/issues/new/choose.`),g=await Promise.all([S(e,{model:m.model_file_name??"model"},m)]));return new this(n,...g)}async _call(e){return await this.forward(e)}async forward(e){return await this._forward(this,e)}_get_logits_warper(e){const t=new c.LogitsProcessorList;return null!==e.temperature&&1!==e.temperature&&t.push(new c.TemperatureLogitsWarper(e.temperature)),null!==e.top_k&&0!==e.top_k&&t.push(new c.TopKLogitsWarper(e.top_k)),null!==e.top_p&&e.top_p<1&&t.push(new c.TopPLogitsWarper(e.top_p)),t}_get_logits_processor(e,t,n=null){const r=new c.LogitsProcessorList;if(null!==e.repetition_penalty&&1!==e.repetition_penalty&&r.push(new c.RepetitionPenaltyLogitsProcessor(e.repetition_penalty)),null!==e.no_repeat_ngram_size&&e.no_repeat_ngram_size>0&&r.push(new c.NoRepeatNGramLogitsProcessor(e.no_repeat_ngram_size)),null!==e.bad_words_ids&&r.push(new c.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 c.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 c.MinNewTokensLengthLogitsProcessor(t,e.min_new_tokens,e.eos_token_id)),null!==e.forced_bos_token_id&&r.push(new c.ForcedBOSTokenLogitsProcessor(e.forced_bos_token_id)),null!==e.forced_eos_token_id&&r.push(new c.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 c.SuppressTokensAtBeginLogitsProcessor(e.begin_suppress_tokens,n))}return null!==e.guidance_scale&&e.guidance_scale>1&&r.push(new c.ClassifierFreeGuidanceLogitsProcessor(e.guidance_scale)),null!==n&&r.extend(n),r}_prepare_generation_config(e,t,n=d.GenerationConfig){const r={...this.config};for(const e of["decoder","generator","text_config"])e in r&&Object.assign(r,r[e]);const o=new n(r);return"generation_config"in this&&Object.assign(o,this.generation_config),e&&Object.assign(o,e),t&&Object.assign(o,(0,i.pick)(t,Object.getOwnPropertyNames(o))),o}_get_stopping_criteria(e,t=null){const n=new h.StoppingCriteriaList;return null!==e.max_length&&n.push(new h.MaxLengthCriteria(e.max_length,this.config.max_position_embeddings??null)),null!==e.eos_token_id&&n.push(new h.EosTokenCriteria(e.eos_token_id)),t&&n.extend(t),n}_validate_model_class(){if(!this.can_generate){const e=[Ra,$a,ja,Ba],t=v.get(this.constructor),n=new Set,r=this.config.model_type;for(const t of e){const e=t.get(r);e&&n.add(e[0])}let o=`The current model class (${t}) is not compatible with \`.generate()\`, as it doesn't have a language model head.`;throw n.size>0&&(o+=` Please use the following class instead: ${[...n].join(", ")}`),Error(o)}}prepare_inputs_for_generation(...e){return this._prepare_inputs_for_generation(this,...e)}_update_model_kwargs_for_generation({generated_input_ids:e,outputs:t,model_inputs:n,is_encoder_decoder:r}){return n.past_key_values=this.getPastKeyValues(t,n.past_key_values),n.input_ids=new u.Tensor("int64",e.flat(),[e.length,1]),r||(n.attention_mask=(0,u.cat)([n.attention_mask,(0,u.ones)([n.attention_mask.dims[0],1])],1)),n.position_ids=null,n}_prepare_model_inputs({inputs:e,bos_token_id:t,model_kwargs:n}){const r=(0,i.pick)(n,this.forward_params),o=this.main_input_name;if(o in r){if(e)throw new Error("`inputs`: {inputs}` were passed alongside {input_name} which is not allowed. Make sure to either pass {inputs} or {input_name}=...")}else r[o]=e;return{inputs_tensor:r[o],model_inputs:r,model_input_name:o}}async _prepare_encoder_decoder_kwargs_for_generation({inputs_tensor:e,model_inputs:t,model_input_name:n,generation_config:r}){if(this.sessions.model.inputNames.includes("inputs_embeds")&&!t.inputs_embeds&&"_prepare_inputs_embeds"in this){const{input_ids:e,pixel_values:n,attention_mask:r,...o}=t,s=await this._prepare_inputs_embeds(t);t={...o,...(0,i.pick)(s,["inputs_embeds","attention_mask"])}}let{last_hidden_state:o}=await B(this,t);if(null!==r.guidance_scale&&r.guidance_scale>1)o=(0,u.cat)([o,(0,u.full_like)(o,0)],0),"attention_mask"in t&&(t.attention_mask=(0,u.cat)([t.attention_mask,(0,u.zeros_like)(t.attention_mask)],0));else if(t.decoder_input_ids){const e=L(t.decoder_input_ids).dims[0];if(e!==o.dims[0]){if(1!==o.dims[0])throw new Error(`The encoder outputs have a different batch size (${o.dims[0]}) than the decoder inputs (${e}).`);o=(0,u.cat)(Array.from({length:e},(()=>o)),0)}}return t.encoder_outputs=o,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:o,generation_config:s}){let{decoder_input_ids:a,...i}=n;if(a)Array.isArray(a[0])||(a=Array.from({length:e},(()=>a)));else if(r??=o,"musicgen"===this.config.model_type)a=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}`);a=r}else a=Array.from({length:e},(()=>[r]));return a=L(a),n.decoder_attention_mask=(0,u.ones_like)(a),{input_ids:a,model_inputs:i}}async generate({inputs:e=null,generation_config:t=null,logits_processor:n=null,stopping_criteria:r=null,streamer:o=null,...s}){this._validate_model_class(),t=this._prepare_generation_config(t,s);let{inputs_tensor:a,model_inputs:i,model_input_name:l}=this._prepare_model_inputs({inputs:e,model_kwargs:s});const c=this.config.is_encoder_decoder;let d;c&&("encoder_outputs"in i||(i=await this._prepare_encoder_decoder_kwargs_for_generation({inputs_tensor:a,model_inputs:i,model_input_name:l,generation_config:t}))),c?({input_ids:d,model_inputs:i}=this._prepare_decoder_input_ids_for_generation({batch_size:i[l].dims.at(0),model_input_name:l,model_kwargs:i,decoder_start_token_id:t.decoder_start_token_id,bos_token_id:t.bos_token_id,generation_config:t})):d=i[l];let p=d.dims.at(-1);null!==t.max_new_tokens&&(t.max_length=p+t.max_new_tokens);const h=this._get_logits_processor(t,p,n),_=this._get_stopping_criteria(t,r),f=i[l].dims.at(0),g=m.LogitsSampler.getSampler(t),M=new Array(f).fill(0),w=d.tolist();o&&o.put(w);let b=null,T={};for(;;){i=this.prepare_inputs_for_generation(w,i,t);const e=await this.forward(i);if(t.output_attentions&&t.return_dict_in_generate){const t=this.getAttentions(e);for(const e in t)e in T||(T[e]=[]),T[e].push(t[e])}const n=h(w,e.logits.slice(null,-1,null)),r=[];for(let e=0;e<n.dims.at(0);++e){const t=n[e],o=await g(t);for(const[t,n]of o){const o=BigInt(t);M[e]+=n,w[e].push(o),r.push([o]);break}}o&&o.put(r);if(_(w).every((e=>e))){t.return_dict_in_generate&&(b=this.getPastKeyValues(e,i.past_key_values,!1));break}i=this._update_model_kwargs_for_generation({generated_input_ids:r,outputs:e,model_inputs:i,is_encoder_decoder:c})}o&&o.end();const x=new u.Tensor("int64",w.flat(),[w.length,w[0].length]);return t.return_dict_in_generate?{sequences:x,past_key_values:b,...T}:x}getPastKeyValues(e,t,n=!0){const r=Object.create(null);for(const o in e)if(o.startsWith("present")){const s=o.replace("present","past_key_values");if(t&&o.includes("encoder"))r[s]=t[s];else{if(n&&t){const e=t[s];"gpu-buffer"===e.location&&e.dispose()}r[s]=e[o]}}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.custom_config.kv_cache_dtype??"float32",n="float16"===t?new Uint16Array:[],o=(0,r.getKeyValueShapes)(this.config);for(const r in o)e[r]=new u.Tensor(t,n,o[r])}}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 q 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 $ extends R{}class W extends ${}class U extends ${async _call(e){return new Ri(await super._call(e))}}class X extends ${async _call(e){return new Di(await super._call(e))}}class Q extends ${async _call(e){return new ji(await super._call(e))}}class H extends ${async _call(e){return new Gi(await super._call(e))}}class Y extends R{}class J extends Y{}class K extends R{}class Z extends K{}class ee extends K{async _call(e){return new Ri(await super._call(e))}}class te extends K{async _call(e){return new Di(await super._call(e))}}class ne extends K{async _call(e){return new ji(await super._call(e))}}class re extends K{async _call(e){return new Gi(await super._call(e))}}class oe extends R{}class se extends oe{}class ae extends oe{async _call(e){return new Ri(await super._call(e))}}class ie extends oe{async _call(e){return new Di(await super._call(e))}}class le extends oe{async _call(e){return new ji(await super._call(e))}}class ce extends oe{async _call(e){return new Gi(await super._call(e))}}class de extends R{}class ue extends de{}class pe extends de{async _call(e){return new Ri(await super._call(e))}}class he extends de{async _call(e){return new Di(await super._call(e))}}class me extends de{async _call(e){return new ji(await super._call(e))}}class _e extends de{async _call(e){return new Gi(await super._call(e))}}class fe extends R{}class ge extends fe{}class Me extends fe{async _call(e){return new Ri(await super._call(e))}}class we extends fe{async _call(e){return new Di(await super._call(e))}}class be extends fe{async _call(e){return new ji(await super._call(e))}}class Te extends fe{async _call(e){return new Gi(await super._call(e))}}class xe extends R{}class ye extends xe{}class ke extends xe{async _call(e){return new Ri(await super._call(e))}}class Fe extends xe{async _call(e){return new Di(await super._call(e))}}class Ce extends xe{async _call(e){return new ji(await super._call(e))}}class Pe extends xe{async _call(e){return new Gi(await super._call(e))}}class ve extends R{}class Se extends ve{}class Ae extends ve{async _call(e){return new Ri(await super._call(e))}}class Ee extends ve{async _call(e){return new Di(await super._call(e))}}class Le extends ve{async _call(e){return new ji(await super._call(e))}}class ze extends ve{async _call(e){return new Gi(await super._call(e))}}class Ie extends R{}class Be extends Ie{}class Ne extends Ie{async _call(e){return new Di(await super._call(e))}}class Oe extends Ie{async _call(e){return new ji(await super._call(e))}}class De extends Ie{async _call(e){return new Gi(await super._call(e))}}class Ve extends Ie{async _call(e){return new Ri(await super._call(e))}}class je extends R{}class Re extends je{}class Ge extends je{async _call(e){return new Ri(await super._call(e))}}class qe extends je{async _call(e){return new Di(await super._call(e))}}class $e extends je{async _call(e){return new ji(await super._call(e))}}class We extends R{}class Ue extends We{}class Xe extends We{async _call(e){return new Ri(await super._call(e))}}class Qe extends We{async _call(e){return new Di(await super._call(e))}}class He extends We{async _call(e){return new Gi(await super._call(e))}}class Ye extends R{}class Je extends Ye{}class Ke extends Ye{async _call(e){return new Ri(await super._call(e))}}class Ze extends Ye{async _call(e){return new Di(await super._call(e))}}class et extends Ye{async _call(e){return new ji(await super._call(e))}}class tt extends Ye{async _call(e){return new Gi(await super._call(e))}}class nt extends R{}class rt extends nt{}class ot extends nt{async _call(e){return new Ri(await super._call(e))}}class st extends nt{async _call(e){return new Di(await super._call(e))}}class at extends nt{async _call(e){return new Gi(await super._call(e))}}class it extends R{}class lt extends it{}class ct extends it{async _call(e){return new Di(await super._call(e))}}class dt extends it{async _call(e){return new Gi(await super._call(e))}}class ut extends it{async _call(e){return new Ri(await super._call(e))}}class pt extends R{forward_params=["input_ids","attention_mask","encoder_outputs","decoder_input_ids","decoder_attention_mask","past_key_values"];constructor(e,t,n){super(e,t),this.generation_config=n}}class ht extends pt{}class mt extends pt{}class _t extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class ft extends _t{}class gt extends _t{}class Mt extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class wt extends Mt{}class bt extends Mt{}class Tt extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class xt extends Tt{}class yt extends Tt{}class kt extends Tt{async _call(e){return new Di(await super._call(e))}}class Ft extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class Ct extends Ft{}class Pt extends Ft{}class vt extends Ft{async _call(e){return new Di(await super._call(e))}}class St extends Ft{}class At extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class Et extends At{}class Lt extends At{}class zt extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class It extends zt{}class Bt extends zt{}class Nt extends R{}class Ot extends Nt{}class Dt extends Nt{async _call(e){return new Ri(await super._call(e))}}class Vt extends Nt{async _call(e){return new Di(await super._call(e))}}class jt extends Nt{async _call(e){return new ji(await super._call(e))}}class Rt extends Nt{async _call(e){return new Gi(await super._call(e))}}class Gt extends R{}class qt extends Gt{}class $t extends Gt{async _call(e){return new Ri(await super._call(e))}}class Wt extends Gt{async _call(e){return new Di(await super._call(e))}}class Ut extends Gt{async _call(e){return new ji(await super._call(e))}}class Xt extends Gt{async _call(e){return new Gi(await super._call(e))}}class Qt extends R{}class Ht extends Qt{}class Yt extends Qt{async _call(e){return new Ri(await super._call(e))}}class Jt extends Qt{async _call(e){return new Di(await super._call(e))}}class Kt extends Qt{async _call(e){return new ji(await super._call(e))}}class Zt extends Qt{async _call(e){return new Gi(await super._call(e))}}class en extends R{}class tn extends en{}class nn extends en{}class rn extends R{requires_attention_mask=!1;main_input_name="input_features";forward_params=["input_features","attention_mask","decoder_input_ids","decoder_attention_mask","past_key_values"];constructor(e,t,n){super(e,t),this.generation_config=n}}class on extends rn{}class sn extends rn{_prepare_generation_config(e,t){return super._prepare_generation_config(e,t,f.WhisperGenerationConfig)}_retrieve_init_tokens(e){const t=[e.decoder_start_token_id];let n=e.language;const r=e.task;if(e.is_multilingual){n||(console.warn("No language specified - defaulting to English (en)."),n="en");const o=`<|${(0,g.whisper_language_to_code)(n)}|>`;t.push(e.lang_to_id[o]),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,...o}){t=this._prepare_generation_config(t,o);const s=o.decoder_input_ids??this._retrieve_init_tokens(t);if(t.return_timestamps&&(n??=new c.LogitsProcessorList,n.push(new c.WhisperTimeStampLogitsProcessor(t,s))),t.begin_suppress_tokens&&(n??=new c.LogitsProcessorList,n.push(new c.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 a=await super.generate({inputs:e,generation_config:t,logits_processor:n,decoder_input_ids:s,...o});return t.return_token_timestamps&&(a.token_timestamps=this._extract_token_timestamps(a,t.alignment_heads,t.num_frames)),a}_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 o=this.config.median_filter_width;void 0===o&&(console.warn("Model config has no `median_filter_width`, using default value of 7."),o=7);const s=e.cross_attentions,a=Array.from({length:this.config.decoder_layers},((e,t)=>(0,u.cat)(s.map((e=>e[t])),2))),l=(0,u.stack)(t.map((([e,t])=>{if(e>=a.length)throw new Error(`Layer index ${e} is out of bounds for cross attentions (length ${a.length}).`);return n?a[e].slice(null,t,null,[0,n]):a[e].slice(null,t)}))).transpose(1,0,2,3),[c,d]=(0,u.std_mean)(l,-2,0,!0),h=l.clone();for(let e=0;e<h.dims[0];++e){const t=h[e];for(let n=0;n<t.dims[0];++n){const r=t[n],s=c[e][n][0].data,a=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]-a[e])/s[e];t.set((0,p.medianFilter)(t,o))}}}const m=[(0,u.mean)(h,1)],_=e.sequences.dims,f=new u.Tensor("float32",new Float32Array(_[0]*_[1]),_);for(let e=0;e<_[0];++e){const t=m[e].neg().squeeze_(0),[n,o]=(0,p.dynamic_time_warping)(t.tolist()),s=Array.from({length:n.length-1},((e,t)=>n[t+1]-n[t])),a=(0,i.mergeArrays)([1],s).map((e=>!!e)),l=[];for(let e=0;e<a.length;++e)a[e]&&l.push(o[e]*r);f[e].data.set(l,1)}return f}}class an extends R{main_input_name="pixel_values";forward_params=["pixel_values","input_ids","encoder_hidden_states","past_key_values"];constructor(e,t,n){super(e,t),this.generation_config=n}}class ln extends R{forward_params=["input_ids","pixel_values","attention_mask","position_ids","past_key_values"];constructor(e,t,n){super(e,t),this.generation_config=n}}class cn extends ln{_merge_input_ids_with_image_features({inputs_embeds:e,image_features:t,input_ids:n,attention_mask:r}){const o=this.config.image_token_index,s=n.tolist().map((e=>e.findIndex((e=>e==o)))),a=s.every((e=>-1===e)),i=s.every((e=>-1!==e));if(!a&&!i)throw new Error("Every input should contain either 0 or 1 image token.");if(a)return{inputs_embeds:e,attention_mask:r};const l=[],c=[];for(let n=0;n<s.length;++n){const o=s[n],a=e[n],i=t[n],d=r[n];l.push((0,u.cat)([a.slice([0,o]),i,a.slice([o+1,a.dims[0]])],0)),c.push((0,u.cat)([d.slice([0,o]),(0,u.ones)([i.dims[0]]),d.slice([o+1,d.dims[0]])],0))}return{inputs_embeds:(0,u.stack)(l,0),attention_mask:(0,u.stack)(c,0)}}}class dn extends cn{}class un extends R{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";constructor(e,t,n){super(e,t),this.generation_config=n}}class pn extends un{_merge_input_ids_with_image_features({inputs_embeds:e,image_features:t,input_ids:n,attention_mask:r}){return{inputs_embeds:(0,u.cat)([t,e],1),attention_mask:(0,u.cat)([(0,u.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 o,s;return e&&(o=await this.encode_text({input_ids:e})),t&&(s=await this.encode_image({pixel_values:t})),o&&s?({inputs_embeds:n,attention_mask:r}=this._merge_input_ids_with_image_features({inputs_embeds:o,image_features:s,input_ids:e,attention_mask:r})):n=o||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:o,encoder_outputs:s,past_key_values:a,inputs_embeds:i,decoder_inputs_embeds:l}){if(i||({inputs_embeds:i,attention_mask:n}=await this._prepare_inputs_embeds({input_ids:e,pixel_values:t,inputs_embeds:i,attention_mask:n})),!s){let{last_hidden_state:e}=await B(this,{inputs_embeds:i,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 c={inputs_embeds:l,attention_mask:o,encoder_attention_mask:n,encoder_hidden_states:s,past_key_values:a};return await N(this,c,!0)}}class hn extends R{}class mn extends hn{}class _n extends hn{static async from_pretrained(e,t={}){return t.model_file_name??="text_model",super.from_pretrained(e,t)}}class fn extends hn{static async from_pretrained(e,t={}){return t.model_file_name??="vision_model",super.from_pretrained(e,t)}}class gn extends R{}class Mn extends gn{}class wn extends gn{static async from_pretrained(e,t={}){return t.model_file_name??="text_model",super.from_pretrained(e,t)}}class bn extends hn{static async from_pretrained(e,t={}){return t.model_file_name??="vision_model",super.from_pretrained(e,t)}}class Tn extends R{}class xn extends Tn{}class yn extends R{}class kn extends yn{}class Fn extends yn{}class Cn extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class Pn extends Cn{}class vn extends Cn{}class Sn extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class An extends Sn{}class En extends Sn{}class Ln extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class zn extends Ln{}class In extends Ln{}class Bn extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class Nn extends Bn{}class On extends Bn{}class Dn extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class Vn extends Dn{}class jn extends Dn{}class Rn extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class Gn extends Rn{}class qn extends Rn{}class $n extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class Wn extends $n{}class Un extends $n{}class Xn extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class Qn extends Xn{}class Hn extends Xn{}class Yn extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class Jn extends Yn{}class Kn extends Yn{}class Zn extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class er extends Zn{}class tr extends Zn{}class nr extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class rr extends nr{}class or extends nr{}class sr extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class ar extends sr{}class ir extends sr{}class lr extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class cr extends lr{}class dr extends lr{}class ur extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class pr extends ur{}class hr extends ur{}class mr extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class _r extends mr{}class fr extends mr{}class gr extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class Mr extends gr{}class wr extends gr{}class br extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class Tr extends br{}class xr extends br{}class yr extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class kr extends yr{}class Fr extends yr{}class Cr extends R{}class Pr extends Cr{}class vr extends Cr{async _call(e){return new Di(await super._call(e))}}class Sr extends R{}class Ar extends Sr{}class Er extends Sr{async _call(e){return new Di(await super._call(e))}}class Lr extends R{}class zr extends Lr{async _call(e){return new Wi(await super._call(e))}}class Ir extends R{}class Br extends Ir{}class Nr extends Ir{async _call(e){return new Di(await super._call(e))}}class Or extends R{}class Dr extends Or{}class Vr extends Or{async _call(e){return new Di(await super._call(e))}}class jr extends R{}class Rr extends jr{}class Gr extends jr{}class qr extends R{}class $r extends qr{}class Wr extends qr{}class Ur extends R{}class Xr extends Ur{}class Qr extends Ur{async _call(e){return new Di(await super._call(e))}}class Hr extends R{}class Yr extends Hr{}class Jr extends Hr{async _call(e){return new Zr(await super._call(e))}}class Kr extends Hr{async _call(e){return new eo(await super._call(e))}}class Zr extends G{constructor({logits:e,pred_boxes:t}){super(),this.logits=e,this.pred_boxes=t}}class eo extends G{constructor({logits:e,pred_boxes:t,pred_masks:n}){super(),this.logits=e,this.pred_boxes=t,this.pred_masks=n}}class to extends R{}class no extends to{}class ro extends to{async _call(e){return new oo(await super._call(e))}}class oo extends G{constructor({logits:e,pred_boxes:t}){super(),this.logits=e,this.pred_boxes=t}}class so extends R{}class ao extends so{}class io extends so{async _call(e){return new lo(await super._call(e))}}class lo extends Zr{}class co extends R{}class uo extends co{}class po extends co{async _call(e){return new Di(await super._call(e))}}class ho extends R{}class mo extends ho{}class _o extends ho{async _call(e){return new Di(await super._call(e))}}class fo extends R{}class go extends fo{}class Mo extends fo{async _call(e){return new Di(await super._call(e))}}class wo extends R{}class bo extends wo{}class To extends wo{}class xo extends R{}class yo extends xo{}class ko extends xo{}class Fo extends R{}class Co extends Fo{}class Po extends R{}class vo extends Po{}class So extends Po{}class Ao extends Po{}class Eo extends R{}class Lo extends Eo{}class zo extends Eo{}class Io extends R{}class Bo extends Io{}class No extends R{}class Oo extends No{}class Do extends No{async _call(e){return new Di(await super._call(e))}}class Vo extends R{}class jo extends Vo{}class Ro extends Vo{async _call(e){return new Di(await super._call(e))}}class Go extends R{}class qo extends Go{}class $o extends Go{async _call(e){return new Di(await super._call(e))}}class Wo extends R{}class Uo extends Wo{}class Xo extends Wo{async _call(e){return new Qo(await super._call(e))}}class Qo extends G{constructor({logits:e,pred_boxes:t}){super(),this.logits=e,this.pred_boxes=t}}class Ho extends R{}class Yo extends Ho{async get_image_embeddings({pixel_values:e}){return await B(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 u.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 Jo(await super._call(e))}}class Jo extends G{constructor({iou_scores:e,pred_masks:t}){super(),this.iou_scores=e,this.pred_masks=t}}class Ko extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class Zo extends Ko{}class es extends Ko{}class ts extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class ns extends ts{}class rs extends ts{}class os extends R{}class ss extends os{}class as extends os{async _call(e){return new qi(await super._call(e))}}class is extends os{async _call(e){return new Di(await super._call(e))}}class ls extends os{async _call(e){return new ji(await super._call(e))}}class cs extends R{}class ds extends cs{}class us extends cs{async _call(e){return new ji(await super._call(e))}}class ps extends R{}class hs extends ps{}class ms extends R{}class _s extends ms{}class fs extends ms{async _call(e){return new qi(await super._call(e))}}class gs extends ms{async _call(e){return new Di(await super._call(e))}}class Ms extends R{}class ws extends Ms{}class bs extends Ms{async _call(e){return new qi(await super._call(e))}}class Ts extends Ms{async _call(e){return new Di(await super._call(e))}}class xs extends Ms{async _call(e){return new ji(await super._call(e))}}class ys extends R{}class ks extends ys{}class Fs extends ys{async _call(e){return new qi(await super._call(e))}}class Cs extends ys{async _call(e){return new Di(await super._call(e))}}class Ps extends R{}class vs extends os{}class Ss extends os{async _call(e){return new qi(await super._call(e))}}class As extends os{async _call(e){return new Di(await super._call(e))}}class Es extends R{}class Ls extends Es{}class zs extends Es{async _call(e){return new qi(await super._call(e))}}class Is extends Es{async _call(e){return new Di(await super._call(e))}}class Bs extends Es{async _call(e){return new Vi(await super._call(e))}}class Ns extends Es{async _call(e){return new ji(await super._call(e))}}class Os extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class Ds extends Os{}class Vs extends Os{}class js extends Os{async generate_speech(e,t,{threshold:n=.5,minlenratio:r=0,maxlenratio:o=20,vocoder:s=null}={}){const a={input_ids:e},{encoder_outputs:i,encoder_attention_mask:l}=await B(this,a),c=i.dims[1]/this.config.reduction_factor,d=Math.floor(c*o),p=Math.floor(c*r),h=this.config.num_mel_bins;let m=[],_=null,f=null,g=0;for(;;){++g;const e=z(!!f);let r;r=f?f.output_sequence_out:new u.Tensor("float32",new Float32Array(h),[1,1,h]);let o={use_cache_branch:e,output_sequence:r,encoder_attention_mask:l,speaker_embeddings:t,encoder_hidden_states:i};this.addPastKeyValues(o,_),f=await A(this.sessions.decoder_model_merged,o),_=this.getPastKeyValues(f,_);const{prob:s,spectrum:a}=f;if(m.push(a),g>=p&&(Array.from(s.data).filter((e=>e>=n)).length>0||g>=d))break}const M=(0,u.cat)(m),{waveform:w}=await A(s.sessions.model,{spectrogram:M});return{spectrogram:M,waveform:w}}}class Rs extends R{main_input_name="spectrogram"}class Gs extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class qs extends Gs{}class $s extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class Ws extends $s{}class Us extends $s{}class Xs extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class Qs extends Xs{}class Hs extends Xs{}class Ys extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class Js extends Ys{}class Ks extends Ys{}class Zs extends R{}class ea extends Zs{}class ta extends Zs{static async from_pretrained(e,t={}){return t.model_file_name??="text_model",super.from_pretrained(e,t)}}class na extends Zs{static async from_pretrained(e,t={}){return t.model_file_name??="audio_model",super.from_pretrained(e,t)}}class ra extends R{}class oa extends ra{async _call(e){return new Ui(await super._call(e))}}class sa extends R{}class aa extends sa{}class ia extends sa{}class la extends sa{}class ca extends R{constructor(e,t,n){super(e,t),this.generation_config=n}}class da extends ca{}class ua extends ca{}class pa extends R{}class ha extends pa{}class ma extends pa{async _call(e){return new Di(await super._call(e))}}class _a extends R{}class fa extends _a{}class ga extends _a{}class Ma extends R{forward_params=["input_ids","attention_mask","encoder_outputs","decoder_input_ids","decoder_attention_mask","past_key_values"];constructor(e,t,n){super(e,t),this.generation_config=n}_apply_and_filter_by_delay_pattern_mask(e){const[t,n]=e.dims,r=this.config.decoder.num_codebooks,o=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 a=t%n-Math.floor(t/n)%r;a>0&&a<=o&&(e.data[s++]=e.data[t])}const a=Math.floor(t/r),i=s/(a*r);return new u.Tensor(e.type,e.data.slice(0,s),[a,r,i])}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 wa extends R{}class ba extends wa{}class Ta extends wa{async _call(e){return new Di(await super._call(e))}}class xa extends R{}class ya extends xa{}class ka extends xa{async _call(e){return new Di(await super._call(e))}}class Fa extends R{}class Ca extends Fa{}class Pa extends Fa{async _call(e){return new Di(await super._call(e))}}class va extends R{}class Sa extends va{}class Aa extends va{async _call(e){return new Di(await super._call(e))}}class Ea{static MODEL_CLASS_MAPPINGS=null;static BASE_IF_FAIL=!1;static async from_pretrained(e,{progress_callback:t=null,config:n=null,cache_dir:o=null,local_files_only:s=!1,revision:a="main",model_file_name:i=null,subfolder:l="onnx",device:c=null,dtype:d=null,use_external_data_format:u=null,session_options:p={}}={}){let h={progress_callback:t,config:n,cache_dir:o,local_files_only:s,revision:a,model_file_name:i,subfolder:l,device:c,dtype:d,use_external_data_format:u,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(let 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 R.from_pretrained(e,h);throw Error(`Unsupported model type: ${h.config.model_type}`)}}const La=new Map([["bert",["BertModel",W]],["nomic_bert",["NomicBertModel",J]],["roformer",["RoFormerModel",Z]],["electra",["ElectraModel",ue]],["esm",["EsmModel",Re]],["convbert",["ConvBertModel",se]],["camembert",["CamembertModel",ge]],["deberta",["DebertaModel",ye]],["deberta-v2",["DebertaV2Model",Se]],["mpnet",["MPNetModel",Je]],["albert",["AlbertModel",lt]],["distilbert",["DistilBertModel",Be]],["roberta",["RobertaModel",Ot]],["xlm",["XLMModel",qt]],["xlm-roberta",["XLMRobertaModel",Ht]],["clap",["ClapModel",ea]],["clip",["CLIPModel",mn]],["clipseg",["CLIPSegModel",kn]],["chinese_clip",["ChineseCLIPModel",xn]],["siglip",["SiglipModel",Mn]],["mobilebert",["MobileBertModel",Ue]],["squeezebert",["SqueezeBertModel",rt]],["wav2vec2",["Wav2Vec2Model",ss]],["wav2vec2-bert",["Wav2Vec2BertModel",ks]],["unispeech",["UniSpeechModel",_s]],["unispeech-sat",["UniSpeechSatModel",ws]],["hubert",["HubertModel",vs]],["wavlm",["WavLMModel",Ls]],["audio-spectrogram-transformer",["ASTModel",tn]],["vits",["VitsModel",oa]],["pyannote",["PyAnnoteModel",ds]],["wespeaker-resnet",["WeSpeakerResNetModel",hs]],["detr",["DetrModel",Yr]],["rt_detr",["RTDetrModel",no]],["table-transformer",["TableTransformerModel",ao]],["vit",["ViTModel",Pr]],["fastvit",["FastViTModel",Ar]],["mobilevit",["MobileViTModel",Br]],["mobilevitv2",["MobileViTV2Model",Dr]],["owlvit",["OwlViTModel",Rr]],["owlv2",["Owlv2Model",$r]],["beit",["BeitModel",Xr]],["deit",["DeiTModel",uo]],["convnext",["ConvNextModel",Oo]],["convnextv2",["ConvNextV2Model",jo]],["dinov2",["Dinov2Model",qo]],["resnet",["ResNetModel",mo]],["swin",["SwinModel",go]],["swin2sr",["Swin2SRModel",bo]],["donut-swin",["DonutSwinModel",Bo]],["yolos",["YolosModel",Uo]],["dpt",["DPTModel",yo]],["glpn",["GLPNModel",Lo]],["hifigan",["SpeechT5HifiGan",Rs]],["efficientnet",["EfficientNetModel",ha]],["mobilenet_v1",["MobileNetV1Model",ba]],["mobilenet_v2",["MobileNetV2Model",ya]],["mobilenet_v3",["MobileNetV3Model",Ca]],["mobilenet_v4",["MobileNetV4Model",Sa]]]),za=new Map([["t5",["T5Model",ht]],["longt5",["LongT5Model",ft]],["mt5",["MT5Model",wt]],["bart",["BartModel",xt]],["mbart",["MBartModel",Ct]],["marian",["MarianModel",Zo]],["whisper",["WhisperModel",on]],["m2m_100",["M2M100Model",ns]],["blenderbot",["BlenderbotModel",Et]],["blenderbot-small",["BlenderbotSmallModel",It]]]),Ia=new Map([["bloom",["BloomModel",Mr]],["jais",["JAISModel",An]],["gpt2",["GPT2Model",Pn]],["gptj",["GPTJModel",Vn]],["gpt_bigcode",["GPTBigCodeModel",Gn]],["gpt_neo",["GPTNeoModel",zn]],["gpt_neox",["GPTNeoXModel",Nn]],["codegen",["CodeGenModel",Wn]],["llama",["LlamaModel",Qn]],["cohere",["CohereModel",Jn]],["gemma",["GemmaModel",er]],["gemma2",["Gemma2Model",rr]],["openelm",["OpenELMModel",ar]],["qwen2",["Qwen2Model",cr]],["phi",["PhiModel",pr]],["phi3",["Phi3Model",_r]],["mpt",["MptModel",Tr]],["opt",["OPTModel",kr]],["mistral",["MistralModel",Ws]],["starcoder2",["Starcoder2Model",Qs]],["falcon",["FalconModel",Js]],["stablelm",["StableLmModel",da]]]),Ba=new Map([["speecht5",["SpeechT5ForSpeechToText",Vs]],["whisper",["WhisperForConditionalGeneration",sn]]]),Na=new Map([["speecht5",["SpeechT5ForTextToSpeech",js]]]),Oa=new Map([["vits",["VitsModel",oa]],["musicgen",["MusicgenForConditionalGeneration",Ma]]]),Da=new Map([["bert",["BertForSequenceClassification",X]],["roformer",["RoFormerForSequenceClassification",te]],["electra",["ElectraForSequenceClassification",he]],["esm",["EsmForSequenceClassification",qe]],["convbert",["ConvBertForSequenceClassification",ie]],["camembert",["CamembertForSequenceClassification",we]],["deberta",["DebertaForSequenceClassification",Fe]],["deberta-v2",["DebertaV2ForSequenceClassification",Ee]],["mpnet",["MPNetForSequenceClassification",Ze]],["albert",["AlbertForSequenceClassification",ct]],["distilbert",["DistilBertForSequenceClassification",Ne]],["roberta",["RobertaForSequenceClassification",Vt]],["xlm",["XLMForSequenceClassification",Wt]],["xlm-roberta",["XLMRobertaForSequenceClassification",Jt]],["bart",["BartForSequenceClassification",kt]],["mbart",["MBartForSequenceClassification",vt]],["mobilebert",["MobileBertForSequenceClassification",Qe]],["squeezebert",["SqueezeBertForSequenceClassification",st]]]),Va=new Map([["bert",["BertForTokenClassification",Q]],["roformer",["RoFormerForTokenClassification",ne]],["electra",["ElectraForTokenClassification",me]],["esm",["EsmForTokenClassification",$e]],["convbert",["ConvBertForTokenClassification",le]],["camembert",["CamembertForTokenClassification",be]],["deberta",["DebertaForTokenClassification",Ce]],["deberta-v2",["DebertaV2ForTokenClassification",Le]],["mpnet",["MPNetForTokenClassification",et]],["distilbert",["DistilBertForTokenClassification",Oe]],["roberta",["RobertaForTokenClassification",jt]],["xlm",["XLMForTokenClassification",Ut]],["xlm-roberta",["XLMRobertaForTokenClassification",Kt]]]),ja=new Map([["t5",["T5ForConditionalGeneration",mt]],["longt5",["LongT5ForConditionalGeneration",gt]],["mt5",["MT5ForConditionalGeneration",bt]],["bart",["BartForConditionalGeneration",yt]],["mbart",["MBartForConditionalGeneration",Pt]],["marian",["MarianMTModel",es]],["m2m_100",["M2M100ForConditionalGeneration",rs]],["blenderbot",["BlenderbotForConditionalGeneration",Lt]],["blenderbot-small",["BlenderbotSmallForConditionalGeneration",Bt]]]),Ra=new Map([["bloom",["BloomForCausalLM",wr]],["gpt2",["GPT2LMHeadModel",vn]],["jais",["JAISLMHeadModel",En]],["gptj",["GPTJForCausalLM",jn]],["gpt_bigcode",["GPTBigCodeForCausalLM",qn]],["gpt_neo",["GPTNeoForCausalLM",In]],["gpt_neox",["GPTNeoXForCausalLM",On]],["codegen",["CodeGenForCausalLM",Un]],["llama",["LlamaForCausalLM",Hn]],["cohere",["CohereForCausalLM",Kn]],["gemma",["GemmaForCausalLM",tr]],["gemma2",["Gemma2ForCausalLM",or]],["openelm",["OpenELMForCausalLM",ir]],["qwen2",["Qwen2ForCausalLM",dr]],["phi",["PhiForCausalLM",hr]],["phi3",["Phi3ForCausalLM",fr]],["mpt",["MptForCausalLM",xr]],["opt",["OPTForCausalLM",Fr]],["mbart",["MBartForCausalLM",St]],["mistral",["MistralForCausalLM",Us]],["starcoder2",["Starcoder2ForCausalLM",Hs]],["falcon",["FalconForCausalLM",Ks]],["trocr",["TrOCRForCausalLM",qs]],["stablelm",["StableLmForCausalLM",ua]]]),Ga=new Map([["bert",["BertForMaskedLM",U]],["roformer",["RoFormerForMaskedLM",ee]],["electra",["ElectraForMaskedLM",pe]],["esm",["EsmForMaskedLM",Ge]],["convbert",["ConvBertForMaskedLM",ae]],["camembert",["CamembertForMaskedLM",Me]],["deberta",["DebertaForMaskedLM",ke]],["deberta-v2",["DebertaV2ForMaskedLM",Ae]],["mpnet",["MPNetForMaskedLM",Ke]],["albert",["AlbertForMaskedLM",ut]],["distilbert",["DistilBertForMaskedLM",Ve]],["roberta",["RobertaForMaskedLM",Dt]],["xlm",["XLMWithLMHeadModel",$t]],["xlm-roberta",["XLMRobertaForMaskedLM",Yt]],["mobilebert",["MobileBertForMaskedLM",Xe]],["squeezebert",["SqueezeBertForMaskedLM",ot]]]),qa=new Map([["bert",["BertForQuestionAnswering",H]],["roformer",["RoFormerForQuestionAnswering",re]],["electra",["ElectraForQuestionAnswering",_e]],["convbert",["ConvBertForQuestionAnswering",ce]],["camembert",["CamembertForQuestionAnswering",Te]],["deberta",["DebertaForQuestionAnswering",Pe]],["deberta-v2",["DebertaV2ForQuestionAnswering",ze]],["mpnet",["MPNetForQuestionAnswering",tt]],["albert",["AlbertForQuestionAnswering",dt]],["distilbert",["DistilBertForQuestionAnswering",De]],["roberta",["RobertaForQuestionAnswering",Rt]],["xlm",["XLMForQuestionAnswering",Xt]],["xlm-roberta",["XLMRobertaForQuestionAnswering",Zt]],["mobilebert",["MobileBertForQuestionAnswering",He]],["squeezebert",["SqueezeBertForQuestionAnswering",at]]]),$a=new Map([["vision-encoder-decoder",["VisionEncoderDecoderModel",an]]]),Wa=new Map([["llava",["LlavaForConditionalGeneration",cn]],["moondream1",["Moondream1ForConditionalGeneration",dn]],["florence2",["Florence2ForConditionalGeneration",pn]]]),Ua=new Map([["vision-encoder-decoder",["VisionEncoderDecoderModel",an]]]),Xa=new Map([["vit",["ViTForImageClassification",vr]],["fastvit",["FastViTForImageClassification",Er]],["mobilevit",["MobileViTForImageClassification",Nr]],["mobilevitv2",["MobileViTV2ForImageClassification",Vr]],["beit",["BeitForImageClassification",Qr]],["deit",["DeiTForImageClassification",po]],["convnext",["ConvNextForImageClassification",Do]],["convnextv2",["ConvNextV2ForImageClassification",Ro]],["dinov2",["Dinov2ForImageClassification",$o]],["resnet",["ResNetForImageClassification",_o]],["swin",["SwinForImageClassification",Mo]],["segformer",["SegformerForImageClassification",ia]],["efficientnet",["EfficientNetForImageClassification",ma]],["mobilenet_v1",["MobileNetV1ForImageClassification",Ta]],["mobilenet_v2",["MobileNetV2ForImageClassification",ka]],["mobilenet_v3",["MobileNetV3ForImageClassification",Pa]],["mobilenet_v4",["MobileNetV4ForImageClassification",Aa]]]),Qa=new Map([["detr",["DetrForObjectDetection",Jr]],["rt_detr",["RTDetrForObjectDetection",ro]],["table-transformer",["TableTransformerForObjectDetection",io]],["yolos",["YolosForObjectDetection",Xo]]]),Ha=new Map([["owlvit",["OwlViTForObjectDetection",Gr]],["owlv2",["Owlv2ForObjectDetection",Wr]]]),Ya=new Map([["detr",["DetrForSegmentation",Kr]],["clipseg",["CLIPSegForImageSegmentation",Fn]]]),Ja=new Map([["segformer",["SegformerForSemanticSegmentation",la]],["sapiens",["SapiensForSemanticSegmentation",vo]]]),Ka=new Map([["sam",["SamModel",Yo]]]),Za=new Map([["wav2vec2",["Wav2Vec2ForCTC",as]],["wav2vec2-bert",["Wav2Vec2BertForCTC",Fs]],["unispeech",["UniSpeechForCTC",fs]],["unispeech-sat",["UniSpeechSatForCTC",bs]],["wavlm",["WavLMForCTC",zs]],["hubert",["HubertForCTC",Ss]]]),ei=new Map([["wav2vec2",["Wav2Vec2ForSequenceClassification",is]],["wav2vec2-bert",["Wav2Vec2BertForSequenceClassification",Cs]],["unispeech",["UniSpeechForSequenceClassification",gs]],["unispeech-sat",["UniSpeechSatForSequenceClassification",Ts]],["wavlm",["WavLMForSequenceClassification",Is]],["hubert",["HubertForSequenceClassification",As]],["audio-spectrogram-transformer",["ASTForAudioClassification",nn]]]),ti=new Map([["wavlm",["WavLMForXVector",Bs]]]),ni=new Map([["unispeech-sat",["UniSpeechSatForAudioFrameClassification",xs]],["wavlm",["WavLMForAudioFrameClassification",Ns]],["wav2vec2",["Wav2Vec2ForAudioFrameClassification",ls]],["pyannote",["PyAnnoteForAudioFrameClassification",us]]]),ri=new Map([["vitmatte",["VitMatteForImageMatting",zr]]]),oi=new Map([["swin2sr",["Swin2SRForImageSuperResolution",To]]]),si=new Map([["dpt",["DPTForDepthEstimation",ko]],["depth_anything",["DepthAnythingForDepthEstimation",Co]],["glpn",["GLPNForDepthEstimation",zo]],["sapiens",["SapiensForDepthEstimation",So]]]),ai=new Map([["sapiens",["SapiensForNormalEstimation",Ao]]]),ii=new Map([["clip",["CLIPVisionModelWithProjection",fn]],["siglip",["SiglipVisionModel",bn]]]),li=[[La,M],[za,w],[Ia,x],[Da,M],[Va,M],[ja,b],[Ba,b],[Ra,x],[Ga,M],[qa,M],[$a,T],[Wa,k],[Xa,M],[Ya,M],[Ja,M],[ri,M],[oi,M],[si,M],[ai,M],[Qa,M],[Ha,M],[Ka,y],[Za,M],[ei,M],[Na,b],[Oa,M],[ti,M],[ni,M],[ii,M]];for(const[e,t]of li)for(const[n,r]of e.values())C.set(n,t),v.set(r,n),P.set(n,r);const ci=[["MusicgenForConditionalGeneration",Ma,F],["CLIPTextModelWithProjection",_n,M],["SiglipTextModel",wn,M],["ClapTextModelWithProjection",ta,M],["ClapAudioModelWithProjection",na,M]];for(const[e,t,n]of ci)C.set(e,n),v.set(t,e),P.set(e,t);class di extends Ea{static MODEL_CLASS_MAPPINGS=li.map((e=>e[0]));static BASE_IF_FAIL=!0}class ui extends Ea{static MODEL_CLASS_MAPPINGS=[Da]}class pi extends Ea{static MODEL_CLASS_MAPPINGS=[Va]}class hi extends Ea{static MODEL_CLASS_MAPPINGS=[ja]}class mi extends Ea{static MODEL_CLASS_MAPPINGS=[Ba]}class _i extends Ea{static MODEL_CLASS_MAPPINGS=[Na]}class fi extends Ea{static MODEL_CLASS_MAPPINGS=[Oa]}class gi extends Ea{static MODEL_CLASS_MAPPINGS=[Ra]}class Mi extends Ea{static MODEL_CLASS_MAPPINGS=[Ga]}class wi extends Ea{static MODEL_CLASS_MAPPINGS=[qa]}class bi extends Ea{static MODEL_CLASS_MAPPINGS=[$a]}class Ti extends Ea{static MODEL_CLASS_MAPPINGS=[Xa]}class xi extends Ea{static MODEL_CLASS_MAPPINGS=[Ya]}class yi extends Ea{static MODEL_CLASS_MAPPINGS=[Ja]}class ki extends Ea{static MODEL_CLASS_MAPPINGS=[Qa]}class Fi extends Ea{static MODEL_CLASS_MAPPINGS=[Ha]}class Ci extends Ea{static MODEL_CLASS_MAPPINGS=[Ka]}class Pi extends Ea{static MODEL_CLASS_MAPPINGS=[Za]}class vi extends Ea{static MODEL_CLASS_MAPPINGS=[ei]}class Si extends Ea{static MODEL_CLASS_MAPPINGS=[ti]}class Ai extends Ea{static MODEL_CLASS_MAPPINGS=[ni]}class Ei extends Ea{static MODEL_CLASS_MAPPINGS=[Ua]}class Li extends Ea{static MODEL_CLASS_MAPPINGS=[ri]}class zi extends Ea{static MODEL_CLASS_MAPPINGS=[oi]}class Ii extends Ea{static MODEL_CLASS_MAPPINGS=[si]}class Bi extends Ea{static MODEL_CLASS_MAPPINGS=[ai]}class Ni extends Ea{static MODEL_CLASS_MAPPINGS=[ii]}class Oi extends G{constructor({logits:e,past_key_values:t,encoder_outputs:n,decoder_attentions:r=null,cross_attentions:o=null}){super(),this.logits=e,this.past_key_values=t,this.encoder_outputs=n,this.decoder_attentions=r,this.cross_attentions=o}}class Di extends G{constructor({logits:e}){super(),this.logits=e}}class Vi extends G{constructor({logits:e,embeddings:t}){super(),this.logits=e,this.embeddings=t}}class ji extends G{constructor({logits:e}){super(),this.logits=e}}class Ri extends G{constructor({logits:e}){super(),this.logits=e}}class Gi extends G{constructor({start_logits:e,end_logits:t}){super(),this.start_logits=e,this.end_logits=t}}class qi extends G{constructor({logits:e}){super(),this.logits=e}}class $i extends G{constructor({logits:e,past_key_values:t}){super(),this.logits=e,this.past_key_values=t}}class Wi extends G{constructor({alphas:e}){super(),this.alphas=e}}class Ui 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:()=>o,WHISPER_TO_LANGUAGE_CODE_MAPPING:()=>s,whisper_language_to_code:()=>a});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"]],o=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 a(e){e=e.toLowerCase();let t=s.get(e);if(void 0===t){if(!o.has(e)){const t=2===e.length?o.keys():o.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":
@@ -160,8 +160,8 @@ import*as e from"fs";import*as t from"onnxruntime-node";import*as n from"path";i
   \****************************/(e,t,n)=>{function r(e,[t,n,r],[o,s],a="bilinear",i=!1){const l=s/r,c=o/n,d=new e.constructor(o*s*t),u=n*r,p=o*s;for(let a=0;a<o;++a)for(let o=0;o<s;++o){const i=a*s+o,h=(o+.5)/l-.5,m=(a+.5)/c-.5;let _=Math.floor(h),f=Math.floor(m);const g=Math.min(_+1,r-1),M=Math.min(f+1,n-1);_=Math.max(_,0),f=Math.max(f,0);const w=h-_,b=m-f,T=(1-w)*(1-b),x=w*(1-b),y=(1-w)*b,k=w*b,F=f*r,C=M*r,P=F+_,v=F+g,S=C+_,A=C+g;for(let n=0;n<t;++n){const t=n*u;d[n*p+i]=T*e[t+P]+x*e[t+v]+y*e[t+S]+k*e[t+A]}}return d}function o(e,t,n){const r=new Array(n.length),o=new Array(n.length);for(let e=n.length-1,s=1;e>=0;--e)o[e]=s,r[e]=t[n[e]],s*=r[e];const s=n.map(((e,t)=>o[n.indexOf(t)])),a=new e.constructor(e.length);for(let n=0;n<e.length;++n){let r=0;for(let e=t.length-1,o=n;e>=0;--e)r+=o%t[e]*s[e],o=Math.floor(o/t[e]);a[r]=e[n]}return[a,r]}function s(e){const t=u(e)[0],n=e.map((e=>Math.exp(e-t))),r=n.reduce(((e,t)=>e+t),0);return n.map((e=>e/r))}function a(e){return s(e).map((e=>Math.log(e)))}function i(e,t){let n=0;for(let r=0;r<e.length;++r)n+=e[r]*t[r];return n}function l(e,t){return i(e,t)/(c(e)*c(t))}function c(e){return Math.sqrt(e.reduce(((e,t)=>e+t*t),0))}function d(e){if(0===e.length)throw Error("Array must not be empty");let t=e[0],n=0;for(let r=1;r<e.length;++r)e[r]<t&&(t=e[r],n=r);return[t,n]}function u(e){if(0===e.length)throw Error("Array must not be empty");let t=e[0],n=0;for(let r=1;r<e.length;++r)e[r]>t&&(t=e[r],n=r);return[Number(t),n]}function p(e){return e>0&&!(e&e-1)}n.r(t),n.d(t,{FFT:()=>_,bankers_round:()=>M,cos_sim:()=>l,dot:()=>i,dynamic_time_warping:()=>w,interpolate_data:()=>r,log_softmax:()=>a,magnitude:()=>c,max:()=>u,medianFilter:()=>f,min:()=>d,permute_data:()=>o,round:()=>g,softmax:()=>s});class h{constructor(e){if(this.size=0|e,this.size<=1||!p(this.size))throw new Error("FFT size must be a power of two larger than 1");this._csize=e<<1,this.table=new Float64Array(2*this.size);for(let e=0;e<this.table.length;e+=2){const t=Math.PI*e/this.size;this.table[e]=Math.cos(t),this.table[e+1]=-Math.sin(t)}let t=0;for(let e=1;this.size>e;e<<=1)++t;this._width=t%2==0?t-1:t,this._bitrev=new Int32Array(1<<this._width);for(let e=0;e<this._bitrev.length;++e){this._bitrev[e]=0;for(let t=0;t<this._width;t+=2){const n=this._width-t-2;this._bitrev[e]|=(e>>>t&3)<<n}}}createComplexArray(){return new Float64Array(this._csize)}fromComplexArray(e,t){const n=t||new Array(e.length>>>1);for(let t=0;t<e.length;t+=2)n[t>>>1]=e[t];return n}toComplexArray(e,t){const n=t||this.createComplexArray();for(let t=0;t<n.length;t+=2)n[t]=e[t>>>1],n[t+1]=0;return n}transform(e,t){if(e===t)throw new Error("Input and output buffers must be different");this._transform4(e,t,1)}realTransform(e,t){if(e===t)throw new Error("Input and output buffers must be different");this._realTransform4(e,t,1)}inverseTransform(e,t){if(e===t)throw new Error("Input and output buffers must be different");this._transform4(e,t,-1);for(let t=0;t<e.length;++t)e[t]/=this.size}_transform4(e,t,n){const r=this._csize;let o,s,a=1<<this._width,i=r/a<<1;const l=this._bitrev;if(4===i)for(o=0,s=0;o<r;o+=i,++s){const n=l[s];this._singleTransform2(t,e,o,n,a)}else for(o=0,s=0;o<r;o+=i,++s){const r=l[s];this._singleTransform4(t,e,o,r,a,n)}const c=this.table;for(a>>=2;a>=2;a>>=2){i=r/a<<1;const t=i>>>2;for(o=0;o<r;o+=i){const r=o+t-1;for(let s=o,i=0;s<r;s+=2,i+=a){const r=s,o=r+t,a=o+t,l=a+t,d=e[r],u=e[r+1],p=e[o],h=e[o+1],m=e[a],_=e[a+1],f=e[l],g=e[l+1],M=c[i],w=n*c[i+1],b=p*M-h*w,T=p*w+h*M,x=c[2*i],y=n*c[2*i+1],k=m*x-_*y,F=m*y+_*x,C=c[3*i],P=n*c[3*i+1],v=f*C-g*P,S=f*P+g*C,A=d+k,E=u+F,L=d-k,z=u-F,I=b+v,B=T+S,N=n*(b-v),O=n*(T-S);e[r]=A+I,e[r+1]=E+B,e[o]=L+O,e[o+1]=z-N,e[a]=A-I,e[a+1]=E-B,e[l]=L-O,e[l+1]=z+N}}}}_singleTransform2(e,t,n,r,o){const s=e[r],a=e[r+1],i=e[r+o],l=e[r+o+1];t[n]=s+i,t[n+1]=a+l,t[n+2]=s-i,t[n+3]=a-l}_singleTransform4(e,t,n,r,o,s){const a=2*o,i=3*o,l=e[r],c=e[r+1],d=e[r+o],u=e[r+o+1],p=e[r+a],h=e[r+a+1],m=e[r+i],_=e[r+i+1],f=l+p,g=c+h,M=l-p,w=c-h,b=d+m,T=u+_,x=s*(d-m),y=s*(u-_);t[n]=f+b,t[n+1]=g+T,t[n+2]=M+y,t[n+3]=w-x,t[n+4]=f-b,t[n+5]=g-T,t[n+6]=M-y,t[n+7]=w+x}_realTransform4(e,t,n){const r=this._csize;let o,s,a=1<<this._width,i=r/a<<1;const l=this._bitrev;if(4===i)for(o=0,s=0;o<r;o+=i,++s){const n=l[s];this._singleRealTransform2(t,e,o,n>>>1,a>>>1)}else for(o=0,s=0;o<r;o+=i,++s){const r=l[s];this._singleRealTransform4(t,e,o,r>>>1,a>>>1,n)}const c=this.table;for(a>>=2;a>=2;a>>=2){i=r/a<<1;const t=i>>>1,s=t>>>1,l=s>>>1;for(o=0;o<r;o+=i)for(let r=0,i=0;r<=l;r+=2,i+=a){const a=o+r,d=a+s,u=d+s,p=u+s,h=e[a],m=e[a+1],_=e[d],f=e[d+1],g=e[u],M=e[u+1],w=e[p],b=e[p+1],T=h,x=m,y=c[i],k=n*c[i+1],F=_*y-f*k,C=_*k+f*y,P=c[2*i],v=n*c[2*i+1],S=g*P-M*v,A=g*v+M*P,E=c[3*i],L=n*c[3*i+1],z=w*E-b*L,I=w*L+b*E,B=T+S,N=x+A,O=T-S,D=x-A,V=F+z,j=C+I,R=n*(F-z),G=n*(C-I);if(e[a]=B+V,e[a+1]=N+j,e[d]=O+G,e[d+1]=D-R,0===r){e[u]=B-V,e[u+1]=N-j;continue}if(r===l)continue;const q=o+s-r,$=o+t-r;e[q]=O-n*G,e[q+1]=-D-n*R,e[$]=B-n*V,e[$+1]=n*j-N}}const d=r>>>1;for(let t=2;t<d;t+=2)e[r-t]=e[t],e[r-t+1]=-e[t+1]}_singleRealTransform2(e,t,n,r,o){const s=e[r],a=e[r+o];t[n]=s+a,t[n+1]=0,t[n+2]=s-a,t[n+3]=0}_singleRealTransform4(e,t,n,r,o,s){const a=2*o,i=3*o,l=e[r],c=e[r+o],d=e[r+a],u=e[r+i],p=l+d,h=l-d,m=c+u,_=s*(c-u);t[n]=p+m,t[n+1]=0,t[n+2]=h,t[n+3]=-_,t[n+4]=p-m,t[n+5]=0,t[n+6]=h,t[n+7]=_}}class m{constructor(e){const t=2*(e-1),n=2*(2*e-1),r=2**Math.ceil(Math.log2(n));this.bufferSize=r,this._a=t;const o=new Float64Array(n),s=new Float64Array(r);this._chirpBuffer=new Float64Array(r),this._buffer1=new Float64Array(r),this._buffer2=new Float64Array(r),this._outBuffer1=new Float64Array(r),this._outBuffer2=new Float64Array(r);const a=-2*Math.PI/e,i=Math.cos(a),l=Math.sin(a);for(let t=0;t<n>>1;++t){const n=(t+1-e)**2/2,r=Math.sqrt(i**2+l**2)**n,a=n*Math.atan2(l,i),c=2*t;o[c]=r*Math.cos(a),o[c+1]=r*Math.sin(a),s[c]=o[c],s[c+1]=-o[c+1]}this._slicedChirpBuffer=o.subarray(t,n),this._f=new h(r>>1),this._f.transform(this._chirpBuffer,s)}_transform(e,t,n){const r=this._buffer1,o=this._buffer2,s=this._outBuffer1,a=this._outBuffer2,i=this._chirpBuffer,l=this._slicedChirpBuffer,c=this._a;if(n)for(let e=0;e<l.length;e+=2){const n=e+1,o=t[e>>1];r[e]=o*l[e],r[n]=o*l[n]}else for(let e=0;e<l.length;e+=2){const n=e+1;r[e]=t[e]*l[e]-t[n]*l[n],r[n]=t[e]*l[n]+t[n]*l[e]}this._f.transform(s,r);for(let e=0;e<i.length;e+=2){const t=e+1;o[e]=s[e]*i[e]-s[t]*i[t],o[t]=s[e]*i[t]+s[t]*i[e]}this._f.inverseTransform(a,o);for(let t=0;t<a.length;t+=2){const n=a[t+c],r=a[t+c+1],o=l[t],s=l[t+1];e[t]=n*o-r*s,e[t+1]=n*s+r*o}}transform(e,t){this._transform(e,t,!1)}realTransform(e,t){this._transform(e,t,!0)}}class _{constructor(e){this.fft_length=e,this.isPowerOfTwo=p(e),this.isPowerOfTwo?(this.fft=new h(e),this.outputBufferSize=2*e):(this.fft=new m(e),this.outputBufferSize=this.fft.bufferSize)}realTransform(e,t){this.fft.realTransform(e,t)}transform(e,t){this.fft.transform(e,t)}}function f(e,t){if(t%2==0||t<=0)throw new Error("Window size must be a positive odd number");const n=new e.constructor(e.length),r=new e.constructor(t),o=Math.floor(t/2);for(let t=0;t<e.length;++t){let s=0;for(let n=-o;n<=o;++n){let o=t+n;o<0?o=Math.abs(o):o>=e.length&&(o=2*(e.length-1)-o),r[s++]=e[o]}r.sort(),n[t]=r[o]}return n}function g(e,t){const n=Math.pow(10,t);return Math.round(e*n)/n}function M(e){const t=Math.round(e);return Math.abs(e)%1==.5?t%2==0?t:t-1:t}function w(e){const t=e.length,n=e[0].length,r=[t+1,n+1],o=Array.from({length:r[0]},(()=>Array(r[1]).fill(1/0)));o[0][0]=0;const s=Array.from({length:r[0]},(()=>Array(r[1]).fill(-1)));for(let t=1;t<r[1];++t)for(let n=1;n<r[0];++n){const r=o[n-1][t-1],a=o[n-1][t],i=o[n][t-1];let l,c;r<a&&r<i?(l=r,c=0):a<r&&a<i?(l=a,c=1):(l=i,c=2),o[n][t]=e[n-1][t-1]+l,s[n][t]=c}for(let e=0;e<r[1];++e)s[0][e]=2;for(let e=0;e<r[0];++e)s[e][0]=1;let a=t,i=n,l=[],c=[];for(;a>0||i>0;)switch(l.push(a-1),c.push(i-1),s[a][i]){case 0:--a,--i;break;case 1:--a;break;case 2:--i;break;default:throw new Error(`Internal error in dynamic time warping. Unexpected trace[${a}, ${i}]. Please file a bug report.`)}return l.reverse(),c.reverse(),[l,c]}},"./src/utils/tensor.js":
 /*!*****************************!*\
   !*** ./src/utils/tensor.js ***!
-  \*****************************/(e,t,n)=>{n.r(t),n.d(t,{Tensor:()=>i,cat:()=>w,full:()=>k,full_like:()=>F,interpolate:()=>c,interpolate_4d:()=>d,layer_norm:()=>_,matmul:()=>u,mean:()=>x,mean_pooling:()=>m,ones:()=>C,ones_like:()=>P,permute:()=>l,quantize_embeddings:()=>A,rfft:()=>p,stack:()=>b,std_mean:()=>T,topk:()=>h,zeros:()=>v,zeros_like:()=>S});var r=n(/*! ./maths.js */"./src/utils/maths.js"),o=n(/*! ../backends/onnx.js */"./src/backends/onnx.js"),s=n(/*! ../ops/registry.js */"./src/ops/registry.js");const a=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 i{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,o.isONNXTensor)(e[0])?this.ort_tensor=e[0]:this.ort_tensor=new o.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=M(e,t),n.length>0){const t=n.reduce(((e,t)=>e*t));return this._subarray(e,t,n)}return new i(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,o=(e+1)*t,s="subarray"in this.data?this.data.subarray(r,o):this.data.slice(r,o);return new i(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 o=e;for(let e=t.length-1;e>=0;e--)o=o.reduce(((n,r)=>{let o=n[n.length-1];return o.length<t[e]?o.push(r):n.push([r]),n}),[[]]);return o[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}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}clone(){return new i(this.type,this.data.slice(),this.dims.slice())}slice(...e){const t=[],n=[];for(let r=0;r<this.dims.length;++r){let o=e[r];if(null==o)n.push([0,this.dims[r]]),t.push(this.dims[r]);else if("number"==typeof o)o=M(o,this.dims[r],r),n.push([o,o+1]);else{if(!Array.isArray(o)||2!==o.length)throw new Error(`Invalid slice: ${o}`);{let[e,s]=o;if(e=null===e?0:M(e,this.dims[r],r,!1),s=null===s?this.dims[r]:M(s,this.dims[r],r,!1),e>s)throw new Error(`Invalid slice: ${o}`);const a=[Math.max(e,0),Math.min(s,this.dims[r])];n.push(a),t.push(a[1]-a[0])}}}const r=n.map((([e,t])=>t-e)),o=r.reduce(((e,t)=>e*t)),s=this.data,a=new s.constructor(o),l=this.stride();for(let e=0;e<o;++e){let t=0;for(let o=r.length-1,s=e;o>=0;--o){const e=r[o];t+=(s%e+n[o][0])*l[o],s=Math.floor(s/e)}a[e]=s[t]}return new i(this.type,a,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 i(this.type,[t],[])}t=M(t,this.dims.length);const o=this.dims.slice();o[t]=1;const s=new r.constructor(r.length/this.dims[t]);for(let n=0;n<r.length;++n){let a=0;for(let e=this.dims.length-1,r=n,s=1;e>=0;--e){const n=this.dims[e];if(e!==t){a+=r%n*s,s*=o[e]}r=Math.floor(r/n)}s[a]+=r[n]**e}if(1!==e)for(let t=0;t<s.length;++t)s[t]=s[t]**(1/e);return n||o.splice(t,1),new i(this.type,s,o)}normalize_(e=2,t=1){t=M(t,this.dims.length);const n=this.norm(e,t,!0),r=this.data,o=n.data;for(let e=0;e<r.length;++e){let n=0;for(let r=this.dims.length-1,o=e,s=1;r>=0;--r){const e=this.dims[r];if(r!==t){n+=o%e*s,s*=this.dims[r]}o=Math.floor(o/e)}r[e]/=o[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 i(this.type,this.data,f(this.dims,e))}squeeze_(e=null){return this.dims=f(this.dims,e),this}unsqueeze(e=null){return new i(this.type,this.data,g(this.dims,e))}unsqueeze_(e=null){return this.dims=g(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),o=this.dims.slice(t+1);return this.dims=[...n,r.reduce(((e,t)=>e*t),1),...o],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 i(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(!a.hasOwnProperty(e))throw new Error(`Unsupported type: ${e}`);return new i(e,a[e].from(this.data),this.dims)}}function l(e,t){const[n,o]=(0,r.permute_data)(e.data,e.dims,t);return new i(e.type,n,o)}function c(e,[t,n],o="bilinear",s=!1){const a=e.dims.at(-3)??1,l=e.dims.at(-2),c=e.dims.at(-1);let d=(0,r.interpolate_data)(e.data,[a,l,c],[t,n],o,s);return new i(e.type,d,[a,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,o;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)o=await s.TensorOpRegistry.bilinear_interpolate_4d;else{if("bicubic"!==n)throw new Error(`Unsupported mode: ${n}`);o=await s.TensorOpRegistry.bicubic_interpolate_4d}const a=new i("int64",new BigInt64Array(r.map(BigInt)),[r.length]);return await o({x:e,s:a})}async function u(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 i("int64",[BigInt(t)],[1])})}function m(e,t){const n=e.data,r=t.data,o=[e.dims[0],e.dims[2]],s=new n.constructor(o[0]*o[1]),[a,l,c]=e.dims;let d=0;for(let e=0;e<a;++e){const t=e*c*l;for(let o=0;o<c;++o){let a=0,i=0;const u=e*l,p=t+o;for(let e=0;e<l;++e){const t=Number(r[u+e]);i+=t,a+=n[p+e*c]*t}const h=a/i;s[d++]=h}}return new i(e.type,s,o)}function _(e,t,{eps:n=1e-5}={}){if(2!==e.dims.length)throw new Error("`layer_norm` currently only supports 2D input.");const[r,o]=e.dims;if(1!==t.length&&t[0]!==o)throw new Error("`normalized_shape` must be a 1D array with shape `[input.dims[1]]`.");const[s,a]=T(e,1,0,!0),l=s.data,c=a.data,d=e.data,u=new d.constructor(d.length);for(let e=0;e<r;++e){const t=e*o;for(let r=0;r<o;++r){const o=t+r;u[o]=(d[o]-c[e])/(l[e]+n)}}return new i(e.type,u,e.dims)}function f(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 g(e,t){return t=M(t,e.length+1),(e=e.slice()).splice(t,0,1),e}function M(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 w(e,t=0){t=M(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),o=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;o.set(e,t),t+=e.length}}else{let r=0;for(let s=0;s<e.length;++s){const{data:a,dims:i}=e[s];for(let e=0;e<a.length;++e){let s=0;for(let o=i.length-1,a=e,l=1;o>=0;--o){const e=i[o];let c=a%e;o===t&&(c+=r),s+=c*l,l*=n[o],a=Math.floor(a/e)}o[s]=a[e]}r+=i[t]}}return new i(s,o,n)}function b(e,t=0){return w(e.map((e=>e.unsqueeze(t))),t)}function T(e,t=null,n=1,r=!1){const o=e.data,s=e.dims;if(null===t){const t=o.reduce(((e,t)=>e+t),0)/o.length,r=Math.sqrt(o.reduce(((e,n)=>e+(n-t)**2),0)/(o.length-n)),s=new i(e.type,[t],[]);return[new i(e.type,[r],[]),s]}const a=x(e,t=M(t,s.length),r),l=a.data,c=s.slice();c[t]=1;const d=new o.constructor(o.length/s[t]);for(let e=0;e<o.length;++e){let n=0;for(let r=s.length-1,o=e,a=1;r>=0;--r){const e=s[r];if(r!==t){n+=o%e*a,a*=c[r]}o=Math.floor(o/e)}d[n]+=(o[e]-l[n])**2}for(let e=0;e<d.length;++e)d[e]=Math.sqrt(d[e]/(s[t]-n));r||c.splice(t,1);return[new i(e.type,d,c),a]}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 i(e.type,[t/r.length],[])}const o=e.dims;t=M(t,o.length);const s=o.slice();s[t]=1;const a=new r.constructor(r.length/o[t]);for(let e=0;e<r.length;++e){let n=0;for(let r=o.length-1,a=e,i=1;r>=0;--r){const e=o[r];if(r!==t){n+=a%e*i,i*=s[r]}a=Math.floor(a/e)}a[n]+=r[e]}if(1!==o[t])for(let e=0;e<a.length;++e)a[e]=a[e]/o[t];return n||s.splice(t,1),new i(e.type,a,s)}function y(e,t,n,r){const o=e.reduce(((e,t)=>e*t),1);return new i(n,new r(o).fill(t),e)}function k(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 y(e,t,n,r)}function F(e,t){return k(e.dims,t)}function C(e){return y(e,1n,"int64",BigInt64Array)}function P(e){return C(e.dims)}function v(e){return y(e,0n,"int64",BigInt64Array)}function S(e){return v(e.dims)}function A(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",o=n?Int8Array:Uint8Array,s=e.data,a=new o(s.length/8);for(let e=0;e<s.length;++e){const t=s[e]>0?1:0,r=Math.floor(e/8),o=e%8;a[r]|=t<<7-o,n&&0===o&&(a[r]-=128)}return new i(r,a,[e.dims[0],e.dims[1]/8])}}},l={};function c(e){var t=l[e];if(void 0!==t)return t.exports;var n=l[e]={exports:{}};return i[e](n,n.exports,c),n.exports}a=Object.getPrototypeOf?e=>Object.getPrototypeOf(e):e=>e.__proto__,c.t=function(e,t){if(1&t&&(e=this(e)),8&t)return e;if("object"==typeof e&&e){if(4&t&&e.__esModule)return e;if(16&t&&"function"==typeof e.then)return e}var n=Object.create(null);c.r(n);var r={};s=s||[null,a({}),a([]),a(a)];for(var o=2&t&&e;"object"==typeof o&&!~s.indexOf(o);o=a(o))Object.getOwnPropertyNames(o).forEach((t=>r[t]=()=>e[t]));return r.default=()=>e,c.d(n,r),n},c.d=(e,t)=>{for(var n in t)c.o(t,n)&&!c.o(e,n)&&Object.defineProperty(e,n,{enumerable:!0,get:t[n]})},c.o=(e,t)=>Object.prototype.hasOwnProperty.call(e,t),c.r=e=>{"undefined"!=typeof Symbol&&Symbol.toStringTag&&Object.defineProperty(e,Symbol.toStringTag,{value:"Module"}),Object.defineProperty(e,"__esModule",{value:!0})};var d={};
+  \*****************************/(e,t,n)=>{n.r(t),n.d(t,{Tensor:()=>i,cat:()=>w,full:()=>k,full_like:()=>F,interpolate:()=>c,interpolate_4d:()=>d,layer_norm:()=>_,matmul:()=>u,mean:()=>x,mean_pooling:()=>m,ones:()=>C,ones_like:()=>P,permute:()=>l,quantize_embeddings:()=>A,rfft:()=>p,stack:()=>b,std_mean:()=>T,topk:()=>h,zeros:()=>v,zeros_like:()=>S});var r=n(/*! ./maths.js */"./src/utils/maths.js"),o=n(/*! ../backends/onnx.js */"./src/backends/onnx.js"),s=n(/*! ../ops/registry.js */"./src/ops/registry.js");const a=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 i{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,o.isONNXTensor)(e[0])?this.ort_tensor=e[0]:this.ort_tensor=new o.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=M(e,t),n.length>0){const t=n.reduce(((e,t)=>e*t));return this._subarray(e,t,n)}return new i(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,o=(e+1)*t,s="subarray"in this.data?this.data.subarray(r,o):this.data.slice(r,o);return new i(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 o=e;for(let e=t.length-1;e>=0;e--)o=o.reduce(((n,r)=>{let o=n[n.length-1];return o.length<t[e]?o.push(r):n.push([r]),n}),[[]]);return o[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}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 i(this.type,this.data.slice(),this.dims.slice())}slice(...e){const t=[],n=[];for(let r=0;r<this.dims.length;++r){let o=e[r];if(null==o)n.push([0,this.dims[r]]),t.push(this.dims[r]);else if("number"==typeof o)o=M(o,this.dims[r],r),n.push([o,o+1]);else{if(!Array.isArray(o)||2!==o.length)throw new Error(`Invalid slice: ${o}`);{let[e,s]=o;if(e=null===e?0:M(e,this.dims[r],r,!1),s=null===s?this.dims[r]:M(s,this.dims[r],r,!1),e>s)throw new Error(`Invalid slice: ${o}`);const a=[Math.max(e,0),Math.min(s,this.dims[r])];n.push(a),t.push(a[1]-a[0])}}}const r=n.map((([e,t])=>t-e)),o=r.reduce(((e,t)=>e*t)),s=this.data,a=new s.constructor(o),l=this.stride();for(let e=0;e<o;++e){let t=0;for(let o=r.length-1,s=e;o>=0;--o){const e=r[o];t+=(s%e+n[o][0])*l[o],s=Math.floor(s/e)}a[e]=s[t]}return new i(this.type,a,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 i(this.type,[t],[])}t=M(t,this.dims.length);const o=this.dims.slice();o[t]=1;const s=new r.constructor(r.length/this.dims[t]);for(let n=0;n<r.length;++n){let a=0;for(let e=this.dims.length-1,r=n,s=1;e>=0;--e){const n=this.dims[e];if(e!==t){a+=r%n*s,s*=o[e]}r=Math.floor(r/n)}s[a]+=r[n]**e}if(1!==e)for(let t=0;t<s.length;++t)s[t]=s[t]**(1/e);return n||o.splice(t,1),new i(this.type,s,o)}normalize_(e=2,t=1){t=M(t,this.dims.length);const n=this.norm(e,t,!0),r=this.data,o=n.data;for(let e=0;e<r.length;++e){let n=0;for(let r=this.dims.length-1,o=e,s=1;r>=0;--r){const e=this.dims[r];if(r!==t){n+=o%e*s,s*=this.dims[r]}o=Math.floor(o/e)}r[e]/=o[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 i(this.type,this.data,f(this.dims,e))}squeeze_(e=null){return this.dims=f(this.dims,e),this}unsqueeze(e=null){return new i(this.type,this.data,g(this.dims,e))}unsqueeze_(e=null){return this.dims=g(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),o=this.dims.slice(t+1);return this.dims=[...n,r.reduce(((e,t)=>e*t),1),...o],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 i(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(!a.hasOwnProperty(e))throw new Error(`Unsupported type: ${e}`);return new i(e,a[e].from(this.data),this.dims)}}function l(e,t){const[n,o]=(0,r.permute_data)(e.data,e.dims,t);return new i(e.type,n,o)}function c(e,[t,n],o="bilinear",s=!1){const a=e.dims.at(-3)??1,l=e.dims.at(-2),c=e.dims.at(-1);let d=(0,r.interpolate_data)(e.data,[a,l,c],[t,n],o,s);return new i(e.type,d,[a,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,o;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)o=await s.TensorOpRegistry.bilinear_interpolate_4d;else{if("bicubic"!==n)throw new Error(`Unsupported mode: ${n}`);o=await s.TensorOpRegistry.bicubic_interpolate_4d}const a=new i("int64",new BigInt64Array(r.map(BigInt)),[r.length]);return await o({x:e,s:a})}async function u(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 i("int64",[BigInt(t)],[1])})}function m(e,t){const n=e.data,r=t.data,o=[e.dims[0],e.dims[2]],s=new n.constructor(o[0]*o[1]),[a,l,c]=e.dims;let d=0;for(let e=0;e<a;++e){const t=e*c*l;for(let o=0;o<c;++o){let a=0,i=0;const u=e*l,p=t+o;for(let e=0;e<l;++e){const t=Number(r[u+e]);i+=t,a+=n[p+e*c]*t}const h=a/i;s[d++]=h}}return new i(e.type,s,o)}function _(e,t,{eps:n=1e-5}={}){if(2!==e.dims.length)throw new Error("`layer_norm` currently only supports 2D input.");const[r,o]=e.dims;if(1!==t.length&&t[0]!==o)throw new Error("`normalized_shape` must be a 1D array with shape `[input.dims[1]]`.");const[s,a]=T(e,1,0,!0),l=s.data,c=a.data,d=e.data,u=new d.constructor(d.length);for(let e=0;e<r;++e){const t=e*o;for(let r=0;r<o;++r){const o=t+r;u[o]=(d[o]-c[e])/(l[e]+n)}}return new i(e.type,u,e.dims)}function f(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 g(e,t){return t=M(t,e.length+1),(e=e.slice()).splice(t,0,1),e}function M(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 w(e,t=0){t=M(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),o=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;o.set(e,t),t+=e.length}}else{let r=0;for(let s=0;s<e.length;++s){const{data:a,dims:i}=e[s];for(let e=0;e<a.length;++e){let s=0;for(let o=i.length-1,a=e,l=1;o>=0;--o){const e=i[o];let c=a%e;o===t&&(c+=r),s+=c*l,l*=n[o],a=Math.floor(a/e)}o[s]=a[e]}r+=i[t]}}return new i(s,o,n)}function b(e,t=0){return w(e.map((e=>e.unsqueeze(t))),t)}function T(e,t=null,n=1,r=!1){const o=e.data,s=e.dims;if(null===t){const t=o.reduce(((e,t)=>e+t),0)/o.length,r=Math.sqrt(o.reduce(((e,n)=>e+(n-t)**2),0)/(o.length-n)),s=new i(e.type,[t],[]);return[new i(e.type,[r],[]),s]}const a=x(e,t=M(t,s.length),r),l=a.data,c=s.slice();c[t]=1;const d=new o.constructor(o.length/s[t]);for(let e=0;e<o.length;++e){let n=0;for(let r=s.length-1,o=e,a=1;r>=0;--r){const e=s[r];if(r!==t){n+=o%e*a,a*=c[r]}o=Math.floor(o/e)}d[n]+=(o[e]-l[n])**2}for(let e=0;e<d.length;++e)d[e]=Math.sqrt(d[e]/(s[t]-n));r||c.splice(t,1);return[new i(e.type,d,c),a]}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 i(e.type,[t/r.length],[])}const o=e.dims;t=M(t,o.length);const s=o.slice();s[t]=1;const a=new r.constructor(r.length/o[t]);for(let e=0;e<r.length;++e){let n=0;for(let r=o.length-1,a=e,i=1;r>=0;--r){const e=o[r];if(r!==t){n+=a%e*i,i*=s[r]}a=Math.floor(a/e)}a[n]+=r[e]}if(1!==o[t])for(let e=0;e<a.length;++e)a[e]=a[e]/o[t];return n||s.splice(t,1),new i(e.type,a,s)}function y(e,t,n,r){const o=e.reduce(((e,t)=>e*t),1);return new i(n,new r(o).fill(t),e)}function k(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 y(e,t,n,r)}function F(e,t){return k(e.dims,t)}function C(e){return y(e,1n,"int64",BigInt64Array)}function P(e){return C(e.dims)}function v(e){return y(e,0n,"int64",BigInt64Array)}function S(e){return v(e.dims)}function A(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",o=n?Int8Array:Uint8Array,s=e.data,a=new o(s.length/8);for(let e=0;e<s.length;++e){const t=s[e]>0?1:0,r=Math.floor(e/8),o=e%8;a[r]|=t<<7-o,n&&0===o&&(a[r]-=128)}return new i(r,a,[e.dims[0],e.dims[1]/8])}}},l={};function c(e){var t=l[e];if(void 0!==t)return t.exports;var n=l[e]={exports:{}};return i[e](n,n.exports,c),n.exports}a=Object.getPrototypeOf?e=>Object.getPrototypeOf(e):e=>e.__proto__,c.t=function(e,t){if(1&t&&(e=this(e)),8&t)return e;if("object"==typeof e&&e){if(4&t&&e.__esModule)return e;if(16&t&&"function"==typeof e.then)return e}var n=Object.create(null);c.r(n);var r={};s=s||[null,a({}),a([]),a(a)];for(var o=2&t&&e;"object"==typeof o&&!~s.indexOf(o);o=a(o))Object.getOwnPropertyNames(o).forEach((t=>r[t]=()=>e[t]));return r.default=()=>e,c.d(n,r),n},c.d=(e,t)=>{for(var n in t)c.o(t,n)&&!c.o(e,n)&&Object.defineProperty(e,n,{enumerable:!0,get:t[n]})},c.o=(e,t)=>Object.prototype.hasOwnProperty.call(e,t),c.r=e=>{"undefined"!=typeof Symbol&&Symbol.toStringTag&&Object.defineProperty(e,Symbol.toStringTag,{value:"Module"}),Object.defineProperty(e,"__esModule",{value:!0})};var d={};
 /*!*****************************!*\
   !*** ./src/transformers.js ***!
-  \*****************************/c.r(d),c.d(d,{ASTFeatureExtractor:()=>_.ASTFeatureExtractor,ASTForAudioClassification:()=>h.ASTForAudioClassification,ASTModel:()=>h.ASTModel,ASTPreTrainedModel:()=>h.ASTPreTrainedModel,AlbertForMaskedLM:()=>h.AlbertForMaskedLM,AlbertForQuestionAnswering:()=>h.AlbertForQuestionAnswering,AlbertForSequenceClassification:()=>h.AlbertForSequenceClassification,AlbertModel:()=>h.AlbertModel,AlbertPreTrainedModel:()=>h.AlbertPreTrainedModel,AlbertTokenizer:()=>m.AlbertTokenizer,AudioClassificationPipeline:()=>p.AudioClassificationPipeline,AutoConfig:()=>f.AutoConfig,AutoModel:()=>h.AutoModel,AutoModelForAudioClassification:()=>h.AutoModelForAudioClassification,AutoModelForAudioFrameClassification:()=>h.AutoModelForAudioFrameClassification,AutoModelForCTC:()=>h.AutoModelForCTC,AutoModelForCausalLM:()=>h.AutoModelForCausalLM,AutoModelForDepthEstimation:()=>h.AutoModelForDepthEstimation,AutoModelForDocumentQuestionAnswering:()=>h.AutoModelForDocumentQuestionAnswering,AutoModelForImageClassification:()=>h.AutoModelForImageClassification,AutoModelForImageFeatureExtraction:()=>h.AutoModelForImageFeatureExtraction,AutoModelForImageMatting:()=>h.AutoModelForImageMatting,AutoModelForImageSegmentation:()=>h.AutoModelForImageSegmentation,AutoModelForImageToImage:()=>h.AutoModelForImageToImage,AutoModelForMaskGeneration:()=>h.AutoModelForMaskGeneration,AutoModelForMaskedLM:()=>h.AutoModelForMaskedLM,AutoModelForObjectDetection:()=>h.AutoModelForObjectDetection,AutoModelForQuestionAnswering:()=>h.AutoModelForQuestionAnswering,AutoModelForSemanticSegmentation:()=>h.AutoModelForSemanticSegmentation,AutoModelForSeq2SeqLM:()=>h.AutoModelForSeq2SeqLM,AutoModelForSequenceClassification:()=>h.AutoModelForSequenceClassification,AutoModelForSpeechSeq2Seq:()=>h.AutoModelForSpeechSeq2Seq,AutoModelForTextToSpectrogram:()=>h.AutoModelForTextToSpectrogram,AutoModelForTextToWaveform:()=>h.AutoModelForTextToWaveform,AutoModelForTokenClassification:()=>h.AutoModelForTokenClassification,AutoModelForVision2Seq:()=>h.AutoModelForVision2Seq,AutoModelForXVector:()=>h.AutoModelForXVector,AutoModelForZeroShotObjectDetection:()=>h.AutoModelForZeroShotObjectDetection,AutoProcessor:()=>_.AutoProcessor,AutoTokenizer:()=>m.AutoTokenizer,AutomaticSpeechRecognitionPipeline:()=>p.AutomaticSpeechRecognitionPipeline,BartForConditionalGeneration:()=>h.BartForConditionalGeneration,BartForSequenceClassification:()=>h.BartForSequenceClassification,BartModel:()=>h.BartModel,BartPretrainedModel:()=>h.BartPretrainedModel,BartTokenizer:()=>m.BartTokenizer,BaseModelOutput:()=>h.BaseModelOutput,BaseStreamer:()=>T.BaseStreamer,BeitFeatureExtractor:()=>_.BeitFeatureExtractor,BeitForImageClassification:()=>h.BeitForImageClassification,BeitModel:()=>h.BeitModel,BeitPreTrainedModel:()=>h.BeitPreTrainedModel,BertForMaskedLM:()=>h.BertForMaskedLM,BertForQuestionAnswering:()=>h.BertForQuestionAnswering,BertForSequenceClassification:()=>h.BertForSequenceClassification,BertForTokenClassification:()=>h.BertForTokenClassification,BertModel:()=>h.BertModel,BertPreTrainedModel:()=>h.BertPreTrainedModel,BertTokenizer:()=>m.BertTokenizer,BitImageProcessor:()=>_.BitImageProcessor,BlenderbotForConditionalGeneration:()=>h.BlenderbotForConditionalGeneration,BlenderbotModel:()=>h.BlenderbotModel,BlenderbotPreTrainedModel:()=>h.BlenderbotPreTrainedModel,BlenderbotSmallForConditionalGeneration:()=>h.BlenderbotSmallForConditionalGeneration,BlenderbotSmallModel:()=>h.BlenderbotSmallModel,BlenderbotSmallPreTrainedModel:()=>h.BlenderbotSmallPreTrainedModel,BlenderbotSmallTokenizer:()=>m.BlenderbotSmallTokenizer,BlenderbotTokenizer:()=>m.BlenderbotTokenizer,BloomForCausalLM:()=>h.BloomForCausalLM,BloomModel:()=>h.BloomModel,BloomPreTrainedModel:()=>h.BloomPreTrainedModel,BloomTokenizer:()=>m.BloomTokenizer,CLIPFeatureExtractor:()=>_.CLIPFeatureExtractor,CLIPImageProcessor:()=>_.CLIPImageProcessor,CLIPModel:()=>h.CLIPModel,CLIPPreTrainedModel:()=>h.CLIPPreTrainedModel,CLIPSegForImageSegmentation:()=>h.CLIPSegForImageSegmentation,CLIPSegModel:()=>h.CLIPSegModel,CLIPSegPreTrainedModel:()=>h.CLIPSegPreTrainedModel,CLIPTextModelWithProjection:()=>h.CLIPTextModelWithProjection,CLIPTokenizer:()=>m.CLIPTokenizer,CLIPVisionModelWithProjection:()=>h.CLIPVisionModelWithProjection,CamembertForMaskedLM:()=>h.CamembertForMaskedLM,CamembertForQuestionAnswering:()=>h.CamembertForQuestionAnswering,CamembertForSequenceClassification:()=>h.CamembertForSequenceClassification,CamembertForTokenClassification:()=>h.CamembertForTokenClassification,CamembertModel:()=>h.CamembertModel,CamembertPreTrainedModel:()=>h.CamembertPreTrainedModel,CamembertTokenizer:()=>m.CamembertTokenizer,CausalLMOutput:()=>h.CausalLMOutput,CausalLMOutputWithPast:()=>h.CausalLMOutputWithPast,ChineseCLIPFeatureExtractor:()=>_.ChineseCLIPFeatureExtractor,ChineseCLIPModel:()=>h.ChineseCLIPModel,ChineseCLIPPreTrainedModel:()=>h.ChineseCLIPPreTrainedModel,ClapAudioModelWithProjection:()=>h.ClapAudioModelWithProjection,ClapFeatureExtractor:()=>_.ClapFeatureExtractor,ClapModel:()=>h.ClapModel,ClapPreTrainedModel:()=>h.ClapPreTrainedModel,ClapTextModelWithProjection:()=>h.ClapTextModelWithProjection,CodeGenForCausalLM:()=>h.CodeGenForCausalLM,CodeGenModel:()=>h.CodeGenModel,CodeGenPreTrainedModel:()=>h.CodeGenPreTrainedModel,CodeGenTokenizer:()=>m.CodeGenTokenizer,CodeLlamaTokenizer:()=>m.CodeLlamaTokenizer,CohereForCausalLM:()=>h.CohereForCausalLM,CohereModel:()=>h.CohereModel,CoherePreTrainedModel:()=>h.CoherePreTrainedModel,CohereTokenizer:()=>m.CohereTokenizer,ConvBertForMaskedLM:()=>h.ConvBertForMaskedLM,ConvBertForQuestionAnswering:()=>h.ConvBertForQuestionAnswering,ConvBertForSequenceClassification:()=>h.ConvBertForSequenceClassification,ConvBertForTokenClassification:()=>h.ConvBertForTokenClassification,ConvBertModel:()=>h.ConvBertModel,ConvBertPreTrainedModel:()=>h.ConvBertPreTrainedModel,ConvBertTokenizer:()=>m.ConvBertTokenizer,ConvNextFeatureExtractor:()=>_.ConvNextFeatureExtractor,ConvNextForImageClassification:()=>h.ConvNextForImageClassification,ConvNextImageProcessor:()=>_.ConvNextImageProcessor,ConvNextModel:()=>h.ConvNextModel,ConvNextPreTrainedModel:()=>h.ConvNextPreTrainedModel,ConvNextV2ForImageClassification:()=>h.ConvNextV2ForImageClassification,ConvNextV2Model:()=>h.ConvNextV2Model,ConvNextV2PreTrainedModel:()=>h.ConvNextV2PreTrainedModel,DPTFeatureExtractor:()=>_.DPTFeatureExtractor,DPTForDepthEstimation:()=>h.DPTForDepthEstimation,DPTImageProcessor:()=>_.DPTImageProcessor,DPTModel:()=>h.DPTModel,DPTPreTrainedModel:()=>h.DPTPreTrainedModel,DebertaForMaskedLM:()=>h.DebertaForMaskedLM,DebertaForQuestionAnswering:()=>h.DebertaForQuestionAnswering,DebertaForSequenceClassification:()=>h.DebertaForSequenceClassification,DebertaForTokenClassification:()=>h.DebertaForTokenClassification,DebertaModel:()=>h.DebertaModel,DebertaPreTrainedModel:()=>h.DebertaPreTrainedModel,DebertaTokenizer:()=>m.DebertaTokenizer,DebertaV2ForMaskedLM:()=>h.DebertaV2ForMaskedLM,DebertaV2ForQuestionAnswering:()=>h.DebertaV2ForQuestionAnswering,DebertaV2ForSequenceClassification:()=>h.DebertaV2ForSequenceClassification,DebertaV2ForTokenClassification:()=>h.DebertaV2ForTokenClassification,DebertaV2Model:()=>h.DebertaV2Model,DebertaV2PreTrainedModel:()=>h.DebertaV2PreTrainedModel,DebertaV2Tokenizer:()=>m.DebertaV2Tokenizer,DeiTFeatureExtractor:()=>_.DeiTFeatureExtractor,DeiTForImageClassification:()=>h.DeiTForImageClassification,DeiTModel:()=>h.DeiTModel,DeiTPreTrainedModel:()=>h.DeiTPreTrainedModel,DepthAnythingForDepthEstimation:()=>h.DepthAnythingForDepthEstimation,DepthAnythingPreTrainedModel:()=>h.DepthAnythingPreTrainedModel,DepthEstimationPipeline:()=>p.DepthEstimationPipeline,DetrFeatureExtractor:()=>_.DetrFeatureExtractor,DetrForObjectDetection:()=>h.DetrForObjectDetection,DetrForSegmentation:()=>h.DetrForSegmentation,DetrModel:()=>h.DetrModel,DetrObjectDetectionOutput:()=>h.DetrObjectDetectionOutput,DetrPreTrainedModel:()=>h.DetrPreTrainedModel,DetrSegmentationOutput:()=>h.DetrSegmentationOutput,Dinov2ForImageClassification:()=>h.Dinov2ForImageClassification,Dinov2Model:()=>h.Dinov2Model,Dinov2PreTrainedModel:()=>h.Dinov2PreTrainedModel,DistilBertForMaskedLM:()=>h.DistilBertForMaskedLM,DistilBertForQuestionAnswering:()=>h.DistilBertForQuestionAnswering,DistilBertForSequenceClassification:()=>h.DistilBertForSequenceClassification,DistilBertForTokenClassification:()=>h.DistilBertForTokenClassification,DistilBertModel:()=>h.DistilBertModel,DistilBertPreTrainedModel:()=>h.DistilBertPreTrainedModel,DistilBertTokenizer:()=>m.DistilBertTokenizer,DocumentQuestionAnsweringPipeline:()=>p.DocumentQuestionAnsweringPipeline,DonutFeatureExtractor:()=>_.DonutFeatureExtractor,DonutSwinModel:()=>h.DonutSwinModel,DonutSwinPreTrainedModel:()=>h.DonutSwinPreTrainedModel,EfficientNetForImageClassification:()=>h.EfficientNetForImageClassification,EfficientNetImageProcessor:()=>_.EfficientNetImageProcessor,EfficientNetModel:()=>h.EfficientNetModel,EfficientNetPreTrainedModel:()=>h.EfficientNetPreTrainedModel,ElectraForMaskedLM:()=>h.ElectraForMaskedLM,ElectraForQuestionAnswering:()=>h.ElectraForQuestionAnswering,ElectraForSequenceClassification:()=>h.ElectraForSequenceClassification,ElectraForTokenClassification:()=>h.ElectraForTokenClassification,ElectraModel:()=>h.ElectraModel,ElectraPreTrainedModel:()=>h.ElectraPreTrainedModel,ElectraTokenizer:()=>m.ElectraTokenizer,EosTokenCriteria:()=>x.EosTokenCriteria,EsmForMaskedLM:()=>h.EsmForMaskedLM,EsmForSequenceClassification:()=>h.EsmForSequenceClassification,EsmForTokenClassification:()=>h.EsmForTokenClassification,EsmModel:()=>h.EsmModel,EsmPreTrainedModel:()=>h.EsmPreTrainedModel,EsmTokenizer:()=>m.EsmTokenizer,FFT:()=>b.FFT,FalconForCausalLM:()=>h.FalconForCausalLM,FalconModel:()=>h.FalconModel,FalconPreTrainedModel:()=>h.FalconPreTrainedModel,FalconTokenizer:()=>m.FalconTokenizer,FastViTForImageClassification:()=>h.FastViTForImageClassification,FastViTModel:()=>h.FastViTModel,FastViTPreTrainedModel:()=>h.FastViTPreTrainedModel,FeatureExtractionPipeline:()=>p.FeatureExtractionPipeline,FeatureExtractor:()=>_.FeatureExtractor,FillMaskPipeline:()=>p.FillMaskPipeline,Florence2ForConditionalGeneration:()=>h.Florence2ForConditionalGeneration,Florence2PreTrainedModel:()=>h.Florence2PreTrainedModel,Florence2Processor:()=>_.Florence2Processor,GLPNFeatureExtractor:()=>_.GLPNFeatureExtractor,GLPNForDepthEstimation:()=>h.GLPNForDepthEstimation,GLPNModel:()=>h.GLPNModel,GLPNPreTrainedModel:()=>h.GLPNPreTrainedModel,GPT2LMHeadModel:()=>h.GPT2LMHeadModel,GPT2Model:()=>h.GPT2Model,GPT2PreTrainedModel:()=>h.GPT2PreTrainedModel,GPT2Tokenizer:()=>m.GPT2Tokenizer,GPTBigCodeForCausalLM:()=>h.GPTBigCodeForCausalLM,GPTBigCodeModel:()=>h.GPTBigCodeModel,GPTBigCodePreTrainedModel:()=>h.GPTBigCodePreTrainedModel,GPTJForCausalLM:()=>h.GPTJForCausalLM,GPTJModel:()=>h.GPTJModel,GPTJPreTrainedModel:()=>h.GPTJPreTrainedModel,GPTNeoForCausalLM:()=>h.GPTNeoForCausalLM,GPTNeoModel:()=>h.GPTNeoModel,GPTNeoPreTrainedModel:()=>h.GPTNeoPreTrainedModel,GPTNeoXForCausalLM:()=>h.GPTNeoXForCausalLM,GPTNeoXModel:()=>h.GPTNeoXModel,GPTNeoXPreTrainedModel:()=>h.GPTNeoXPreTrainedModel,GPTNeoXTokenizer:()=>m.GPTNeoXTokenizer,Gemma2ForCausalLM:()=>h.Gemma2ForCausalLM,Gemma2Model:()=>h.Gemma2Model,Gemma2PreTrainedModel:()=>h.Gemma2PreTrainedModel,GemmaForCausalLM:()=>h.GemmaForCausalLM,GemmaModel:()=>h.GemmaModel,GemmaPreTrainedModel:()=>h.GemmaPreTrainedModel,GemmaTokenizer:()=>m.GemmaTokenizer,Grok1Tokenizer:()=>m.Grok1Tokenizer,HerbertTokenizer:()=>m.HerbertTokenizer,HubertForCTC:()=>h.HubertForCTC,HubertForSequenceClassification:()=>h.HubertForSequenceClassification,HubertModel:()=>h.HubertModel,HubertPreTrainedModel:()=>h.HubertPreTrainedModel,ImageClassificationPipeline:()=>p.ImageClassificationPipeline,ImageFeatureExtractionPipeline:()=>p.ImageFeatureExtractionPipeline,ImageFeatureExtractor:()=>_.ImageFeatureExtractor,ImageMattingOutput:()=>h.ImageMattingOutput,ImageSegmentationPipeline:()=>p.ImageSegmentationPipeline,ImageToImagePipeline:()=>p.ImageToImagePipeline,ImageToTextPipeline:()=>p.ImageToTextPipeline,InterruptableStoppingCriteria:()=>x.InterruptableStoppingCriteria,JAISLMHeadModel:()=>h.JAISLMHeadModel,JAISModel:()=>h.JAISModel,JAISPreTrainedModel:()=>h.JAISPreTrainedModel,LlamaForCausalLM:()=>h.LlamaForCausalLM,LlamaModel:()=>h.LlamaModel,LlamaPreTrainedModel:()=>h.LlamaPreTrainedModel,LlamaTokenizer:()=>m.LlamaTokenizer,LlavaForConditionalGeneration:()=>h.LlavaForConditionalGeneration,LlavaPreTrainedModel:()=>h.LlavaPreTrainedModel,LongT5ForConditionalGeneration:()=>h.LongT5ForConditionalGeneration,LongT5Model:()=>h.LongT5Model,LongT5PreTrainedModel:()=>h.LongT5PreTrainedModel,M2M100ForConditionalGeneration:()=>h.M2M100ForConditionalGeneration,M2M100Model:()=>h.M2M100Model,M2M100PreTrainedModel:()=>h.M2M100PreTrainedModel,M2M100Tokenizer:()=>m.M2M100Tokenizer,MBart50Tokenizer:()=>m.MBart50Tokenizer,MBartForCausalLM:()=>h.MBartForCausalLM,MBartForConditionalGeneration:()=>h.MBartForConditionalGeneration,MBartForSequenceClassification:()=>h.MBartForSequenceClassification,MBartModel:()=>h.MBartModel,MBartPreTrainedModel:()=>h.MBartPreTrainedModel,MBartTokenizer:()=>m.MBartTokenizer,MPNetForMaskedLM:()=>h.MPNetForMaskedLM,MPNetForQuestionAnswering:()=>h.MPNetForQuestionAnswering,MPNetForSequenceClassification:()=>h.MPNetForSequenceClassification,MPNetForTokenClassification:()=>h.MPNetForTokenClassification,MPNetModel:()=>h.MPNetModel,MPNetPreTrainedModel:()=>h.MPNetPreTrainedModel,MPNetTokenizer:()=>m.MPNetTokenizer,MT5ForConditionalGeneration:()=>h.MT5ForConditionalGeneration,MT5Model:()=>h.MT5Model,MT5PreTrainedModel:()=>h.MT5PreTrainedModel,MarianMTModel:()=>h.MarianMTModel,MarianModel:()=>h.MarianModel,MarianPreTrainedModel:()=>h.MarianPreTrainedModel,MarianTokenizer:()=>m.MarianTokenizer,MaskedLMOutput:()=>h.MaskedLMOutput,MaxLengthCriteria:()=>x.MaxLengthCriteria,MistralForCausalLM:()=>h.MistralForCausalLM,MistralModel:()=>h.MistralModel,MistralPreTrainedModel:()=>h.MistralPreTrainedModel,MobileBertForMaskedLM:()=>h.MobileBertForMaskedLM,MobileBertForQuestionAnswering:()=>h.MobileBertForQuestionAnswering,MobileBertForSequenceClassification:()=>h.MobileBertForSequenceClassification,MobileBertModel:()=>h.MobileBertModel,MobileBertPreTrainedModel:()=>h.MobileBertPreTrainedModel,MobileBertTokenizer:()=>m.MobileBertTokenizer,MobileNetV1FeatureExtractor:()=>_.MobileNetV1FeatureExtractor,MobileNetV1ForImageClassification:()=>h.MobileNetV1ForImageClassification,MobileNetV1Model:()=>h.MobileNetV1Model,MobileNetV1PreTrainedModel:()=>h.MobileNetV1PreTrainedModel,MobileNetV2FeatureExtractor:()=>_.MobileNetV2FeatureExtractor,MobileNetV2ForImageClassification:()=>h.MobileNetV2ForImageClassification,MobileNetV2Model:()=>h.MobileNetV2Model,MobileNetV2PreTrainedModel:()=>h.MobileNetV2PreTrainedModel,MobileNetV3FeatureExtractor:()=>_.MobileNetV3FeatureExtractor,MobileNetV3ForImageClassification:()=>h.MobileNetV3ForImageClassification,MobileNetV3Model:()=>h.MobileNetV3Model,MobileNetV3PreTrainedModel:()=>h.MobileNetV3PreTrainedModel,MobileNetV4FeatureExtractor:()=>_.MobileNetV4FeatureExtractor,MobileNetV4ForImageClassification:()=>h.MobileNetV4ForImageClassification,MobileNetV4Model:()=>h.MobileNetV4Model,MobileNetV4PreTrainedModel:()=>h.MobileNetV4PreTrainedModel,MobileViTFeatureExtractor:()=>_.MobileViTFeatureExtractor,MobileViTForImageClassification:()=>h.MobileViTForImageClassification,MobileViTImageProcessor:()=>_.MobileViTImageProcessor,MobileViTModel:()=>h.MobileViTModel,MobileViTPreTrainedModel:()=>h.MobileViTPreTrainedModel,MobileViTV2ForImageClassification:()=>h.MobileViTV2ForImageClassification,MobileViTV2Model:()=>h.MobileViTV2Model,MobileViTV2PreTrainedModel:()=>h.MobileViTV2PreTrainedModel,ModelOutput:()=>h.ModelOutput,Moondream1ForConditionalGeneration:()=>h.Moondream1ForConditionalGeneration,MptForCausalLM:()=>h.MptForCausalLM,MptModel:()=>h.MptModel,MptPreTrainedModel:()=>h.MptPreTrainedModel,MusicgenForCausalLM:()=>h.MusicgenForCausalLM,MusicgenForConditionalGeneration:()=>h.MusicgenForConditionalGeneration,MusicgenModel:()=>h.MusicgenModel,MusicgenPreTrainedModel:()=>h.MusicgenPreTrainedModel,NllbTokenizer:()=>m.NllbTokenizer,NomicBertModel:()=>h.NomicBertModel,NomicBertPreTrainedModel:()=>h.NomicBertPreTrainedModel,NougatImageProcessor:()=>_.NougatImageProcessor,NougatTokenizer:()=>m.NougatTokenizer,OPTForCausalLM:()=>h.OPTForCausalLM,OPTModel:()=>h.OPTModel,OPTPreTrainedModel:()=>h.OPTPreTrainedModel,ObjectDetectionPipeline:()=>p.ObjectDetectionPipeline,OpenELMForCausalLM:()=>h.OpenELMForCausalLM,OpenELMModel:()=>h.OpenELMModel,OpenELMPreTrainedModel:()=>h.OpenELMPreTrainedModel,OwlViTFeatureExtractor:()=>_.OwlViTFeatureExtractor,OwlViTForObjectDetection:()=>h.OwlViTForObjectDetection,OwlViTModel:()=>h.OwlViTModel,OwlViTPreTrainedModel:()=>h.OwlViTPreTrainedModel,OwlViTProcessor:()=>_.OwlViTProcessor,Owlv2ForObjectDetection:()=>h.Owlv2ForObjectDetection,Owlv2ImageProcessor:()=>_.Owlv2ImageProcessor,Owlv2Model:()=>h.Owlv2Model,Owlv2PreTrainedModel:()=>h.Owlv2PreTrainedModel,Phi3ForCausalLM:()=>h.Phi3ForCausalLM,Phi3Model:()=>h.Phi3Model,Phi3PreTrainedModel:()=>h.Phi3PreTrainedModel,PhiForCausalLM:()=>h.PhiForCausalLM,PhiModel:()=>h.PhiModel,PhiPreTrainedModel:()=>h.PhiPreTrainedModel,Pipeline:()=>p.Pipeline,PreTrainedModel:()=>h.PreTrainedModel,PreTrainedTokenizer:()=>m.PreTrainedTokenizer,PretrainedConfig:()=>f.PretrainedConfig,PretrainedMixin:()=>h.PretrainedMixin,Processor:()=>_.Processor,PyAnnoteFeatureExtractor:()=>_.PyAnnoteFeatureExtractor,PyAnnoteForAudioFrameClassification:()=>h.PyAnnoteForAudioFrameClassification,PyAnnoteModel:()=>h.PyAnnoteModel,PyAnnotePreTrainedModel:()=>h.PyAnnotePreTrainedModel,PyAnnoteProcessor:()=>_.PyAnnoteProcessor,QuestionAnsweringModelOutput:()=>h.QuestionAnsweringModelOutput,QuestionAnsweringPipeline:()=>p.QuestionAnsweringPipeline,Qwen2ForCausalLM:()=>h.Qwen2ForCausalLM,Qwen2Model:()=>h.Qwen2Model,Qwen2PreTrainedModel:()=>h.Qwen2PreTrainedModel,Qwen2Tokenizer:()=>m.Qwen2Tokenizer,RTDetrForObjectDetection:()=>h.RTDetrForObjectDetection,RTDetrImageProcessor:()=>_.RTDetrImageProcessor,RTDetrModel:()=>h.RTDetrModel,RTDetrObjectDetectionOutput:()=>h.RTDetrObjectDetectionOutput,RTDetrPreTrainedModel:()=>h.RTDetrPreTrainedModel,RawImage:()=>M.RawImage,ResNetForImageClassification:()=>h.ResNetForImageClassification,ResNetModel:()=>h.ResNetModel,ResNetPreTrainedModel:()=>h.ResNetPreTrainedModel,RoFormerForMaskedLM:()=>h.RoFormerForMaskedLM,RoFormerForQuestionAnswering:()=>h.RoFormerForQuestionAnswering,RoFormerForSequenceClassification:()=>h.RoFormerForSequenceClassification,RoFormerForTokenClassification:()=>h.RoFormerForTokenClassification,RoFormerModel:()=>h.RoFormerModel,RoFormerPreTrainedModel:()=>h.RoFormerPreTrainedModel,RoFormerTokenizer:()=>m.RoFormerTokenizer,RobertaForMaskedLM:()=>h.RobertaForMaskedLM,RobertaForQuestionAnswering:()=>h.RobertaForQuestionAnswering,RobertaForSequenceClassification:()=>h.RobertaForSequenceClassification,RobertaForTokenClassification:()=>h.RobertaForTokenClassification,RobertaModel:()=>h.RobertaModel,RobertaPreTrainedModel:()=>h.RobertaPreTrainedModel,RobertaTokenizer:()=>m.RobertaTokenizer,SamImageProcessor:()=>_.SamImageProcessor,SamImageSegmentationOutput:()=>h.SamImageSegmentationOutput,SamModel:()=>h.SamModel,SamPreTrainedModel:()=>h.SamPreTrainedModel,SamProcessor:()=>_.SamProcessor,SapiensFeatureExtractor:()=>_.SapiensFeatureExtractor,SapiensForDepthEstimation:()=>h.SapiensForDepthEstimation,SapiensForNormalEstimation:()=>h.SapiensForNormalEstimation,SapiensForSemanticSegmentation:()=>h.SapiensForSemanticSegmentation,SapiensPreTrainedModel:()=>h.SapiensPreTrainedModel,SeamlessM4TFeatureExtractor:()=>_.SeamlessM4TFeatureExtractor,SegformerFeatureExtractor:()=>_.SegformerFeatureExtractor,SegformerForImageClassification:()=>h.SegformerForImageClassification,SegformerForSemanticSegmentation:()=>h.SegformerForSemanticSegmentation,SegformerModel:()=>h.SegformerModel,SegformerPreTrainedModel:()=>h.SegformerPreTrainedModel,Seq2SeqLMOutput:()=>h.Seq2SeqLMOutput,SequenceClassifierOutput:()=>h.SequenceClassifierOutput,SiglipImageProcessor:()=>_.SiglipImageProcessor,SiglipModel:()=>h.SiglipModel,SiglipPreTrainedModel:()=>h.SiglipPreTrainedModel,SiglipTextModel:()=>h.SiglipTextModel,SiglipTokenizer:()=>m.SiglipTokenizer,SiglipVisionModel:()=>h.SiglipVisionModel,SpeechT5FeatureExtractor:()=>_.SpeechT5FeatureExtractor,SpeechT5ForSpeechToText:()=>h.SpeechT5ForSpeechToText,SpeechT5ForTextToSpeech:()=>h.SpeechT5ForTextToSpeech,SpeechT5HifiGan:()=>h.SpeechT5HifiGan,SpeechT5Model:()=>h.SpeechT5Model,SpeechT5PreTrainedModel:()=>h.SpeechT5PreTrainedModel,SpeechT5Processor:()=>_.SpeechT5Processor,SpeechT5Tokenizer:()=>m.SpeechT5Tokenizer,SqueezeBertForMaskedLM:()=>h.SqueezeBertForMaskedLM,SqueezeBertForQuestionAnswering:()=>h.SqueezeBertForQuestionAnswering,SqueezeBertForSequenceClassification:()=>h.SqueezeBertForSequenceClassification,SqueezeBertModel:()=>h.SqueezeBertModel,SqueezeBertPreTrainedModel:()=>h.SqueezeBertPreTrainedModel,SqueezeBertTokenizer:()=>m.SqueezeBertTokenizer,StableLmForCausalLM:()=>h.StableLmForCausalLM,StableLmModel:()=>h.StableLmModel,StableLmPreTrainedModel:()=>h.StableLmPreTrainedModel,Starcoder2ForCausalLM:()=>h.Starcoder2ForCausalLM,Starcoder2Model:()=>h.Starcoder2Model,Starcoder2PreTrainedModel:()=>h.Starcoder2PreTrainedModel,StoppingCriteria:()=>x.StoppingCriteria,StoppingCriteriaList:()=>x.StoppingCriteriaList,SummarizationPipeline:()=>p.SummarizationPipeline,Swin2SRForImageSuperResolution:()=>h.Swin2SRForImageSuperResolution,Swin2SRImageProcessor:()=>_.Swin2SRImageProcessor,Swin2SRModel:()=>h.Swin2SRModel,Swin2SRPreTrainedModel:()=>h.Swin2SRPreTrainedModel,SwinForImageClassification:()=>h.SwinForImageClassification,SwinModel:()=>h.SwinModel,SwinPreTrainedModel:()=>h.SwinPreTrainedModel,T5ForConditionalGeneration:()=>h.T5ForConditionalGeneration,T5Model:()=>h.T5Model,T5PreTrainedModel:()=>h.T5PreTrainedModel,T5Tokenizer:()=>m.T5Tokenizer,TableTransformerForObjectDetection:()=>h.TableTransformerForObjectDetection,TableTransformerModel:()=>h.TableTransformerModel,TableTransformerObjectDetectionOutput:()=>h.TableTransformerObjectDetectionOutput,TableTransformerPreTrainedModel:()=>h.TableTransformerPreTrainedModel,Tensor:()=>w.Tensor,Text2TextGenerationPipeline:()=>p.Text2TextGenerationPipeline,TextClassificationPipeline:()=>p.TextClassificationPipeline,TextGenerationPipeline:()=>p.TextGenerationPipeline,TextStreamer:()=>T.TextStreamer,TextToAudioPipeline:()=>p.TextToAudioPipeline,TokenClassificationPipeline:()=>p.TokenClassificationPipeline,TokenClassifierOutput:()=>h.TokenClassifierOutput,TokenizerModel:()=>m.TokenizerModel,TrOCRForCausalLM:()=>h.TrOCRForCausalLM,TrOCRPreTrainedModel:()=>h.TrOCRPreTrainedModel,TranslationPipeline:()=>p.TranslationPipeline,UniSpeechForCTC:()=>h.UniSpeechForCTC,UniSpeechForSequenceClassification:()=>h.UniSpeechForSequenceClassification,UniSpeechModel:()=>h.UniSpeechModel,UniSpeechPreTrainedModel:()=>h.UniSpeechPreTrainedModel,UniSpeechSatForAudioFrameClassification:()=>h.UniSpeechSatForAudioFrameClassification,UniSpeechSatForCTC:()=>h.UniSpeechSatForCTC,UniSpeechSatForSequenceClassification:()=>h.UniSpeechSatForSequenceClassification,UniSpeechSatModel:()=>h.UniSpeechSatModel,UniSpeechSatPreTrainedModel:()=>h.UniSpeechSatPreTrainedModel,ViTFeatureExtractor:()=>_.ViTFeatureExtractor,ViTForImageClassification:()=>h.ViTForImageClassification,ViTImageProcessor:()=>_.ViTImageProcessor,ViTModel:()=>h.ViTModel,ViTPreTrainedModel:()=>h.ViTPreTrainedModel,VisionEncoderDecoderModel:()=>h.VisionEncoderDecoderModel,VitMatteForImageMatting:()=>h.VitMatteForImageMatting,VitMatteImageProcessor:()=>_.VitMatteImageProcessor,VitMattePreTrainedModel:()=>h.VitMattePreTrainedModel,VitsModel:()=>h.VitsModel,VitsModelOutput:()=>h.VitsModelOutput,VitsPreTrainedModel:()=>h.VitsPreTrainedModel,VitsTokenizer:()=>m.VitsTokenizer,Wav2Vec2BertForCTC:()=>h.Wav2Vec2BertForCTC,Wav2Vec2BertForSequenceClassification:()=>h.Wav2Vec2BertForSequenceClassification,Wav2Vec2BertModel:()=>h.Wav2Vec2BertModel,Wav2Vec2BertPreTrainedModel:()=>h.Wav2Vec2BertPreTrainedModel,Wav2Vec2CTCTokenizer:()=>m.Wav2Vec2CTCTokenizer,Wav2Vec2FeatureExtractor:()=>_.Wav2Vec2FeatureExtractor,Wav2Vec2ForAudioFrameClassification:()=>h.Wav2Vec2ForAudioFrameClassification,Wav2Vec2ForCTC:()=>h.Wav2Vec2ForCTC,Wav2Vec2ForSequenceClassification:()=>h.Wav2Vec2ForSequenceClassification,Wav2Vec2Model:()=>h.Wav2Vec2Model,Wav2Vec2PreTrainedModel:()=>h.Wav2Vec2PreTrainedModel,Wav2Vec2ProcessorWithLM:()=>_.Wav2Vec2ProcessorWithLM,WavLMForAudioFrameClassification:()=>h.WavLMForAudioFrameClassification,WavLMForCTC:()=>h.WavLMForCTC,WavLMForSequenceClassification:()=>h.WavLMForSequenceClassification,WavLMForXVector:()=>h.WavLMForXVector,WavLMModel:()=>h.WavLMModel,WavLMPreTrainedModel:()=>h.WavLMPreTrainedModel,WeSpeakerFeatureExtractor:()=>_.WeSpeakerFeatureExtractor,WeSpeakerResNetModel:()=>h.WeSpeakerResNetModel,WeSpeakerResNetPreTrainedModel:()=>h.WeSpeakerResNetPreTrainedModel,WhisperFeatureExtractor:()=>_.WhisperFeatureExtractor,WhisperForConditionalGeneration:()=>h.WhisperForConditionalGeneration,WhisperModel:()=>h.WhisperModel,WhisperPreTrainedModel:()=>h.WhisperPreTrainedModel,WhisperProcessor:()=>_.WhisperProcessor,WhisperTextStreamer:()=>T.WhisperTextStreamer,WhisperTokenizer:()=>m.WhisperTokenizer,XLMForQuestionAnswering:()=>h.XLMForQuestionAnswering,XLMForSequenceClassification:()=>h.XLMForSequenceClassification,XLMForTokenClassification:()=>h.XLMForTokenClassification,XLMModel:()=>h.XLMModel,XLMPreTrainedModel:()=>h.XLMPreTrainedModel,XLMRobertaForMaskedLM:()=>h.XLMRobertaForMaskedLM,XLMRobertaForQuestionAnswering:()=>h.XLMRobertaForQuestionAnswering,XLMRobertaForSequenceClassification:()=>h.XLMRobertaForSequenceClassification,XLMRobertaForTokenClassification:()=>h.XLMRobertaForTokenClassification,XLMRobertaModel:()=>h.XLMRobertaModel,XLMRobertaPreTrainedModel:()=>h.XLMRobertaPreTrainedModel,XLMRobertaTokenizer:()=>m.XLMRobertaTokenizer,XLMTokenizer:()=>m.XLMTokenizer,XLMWithLMHeadModel:()=>h.XLMWithLMHeadModel,XVectorOutput:()=>h.XVectorOutput,YolosFeatureExtractor:()=>_.YolosFeatureExtractor,YolosForObjectDetection:()=>h.YolosForObjectDetection,YolosModel:()=>h.YolosModel,YolosObjectDetectionOutput:()=>h.YolosObjectDetectionOutput,YolosPreTrainedModel:()=>h.YolosPreTrainedModel,ZeroShotAudioClassificationPipeline:()=>p.ZeroShotAudioClassificationPipeline,ZeroShotClassificationPipeline:()=>p.ZeroShotClassificationPipeline,ZeroShotImageClassificationPipeline:()=>p.ZeroShotImageClassificationPipeline,ZeroShotObjectDetectionPipeline:()=>p.ZeroShotObjectDetectionPipeline,bankers_round:()=>b.bankers_round,cat:()=>w.cat,cos_sim:()=>b.cos_sim,dot:()=>b.dot,dynamic_time_warping:()=>b.dynamic_time_warping,env:()=>u.env,full:()=>w.full,full_like:()=>w.full_like,getKeyValueShapes:()=>f.getKeyValueShapes,hamming:()=>g.hamming,hanning:()=>g.hanning,interpolate:()=>w.interpolate,interpolate_4d:()=>w.interpolate_4d,interpolate_data:()=>b.interpolate_data,is_chinese_char:()=>m.is_chinese_char,layer_norm:()=>w.layer_norm,log_softmax:()=>b.log_softmax,magnitude:()=>b.magnitude,matmul:()=>w.matmul,max:()=>b.max,mean:()=>w.mean,mean_pooling:()=>w.mean_pooling,medianFilter:()=>b.medianFilter,mel_filter_bank:()=>g.mel_filter_bank,min:()=>b.min,ones:()=>w.ones,ones_like:()=>w.ones_like,permute:()=>w.permute,permute_data:()=>b.permute_data,pipeline:()=>p.pipeline,quantize_embeddings:()=>w.quantize_embeddings,read_audio:()=>g.read_audio,rfft:()=>w.rfft,round:()=>b.round,softmax:()=>b.softmax,spectrogram:()=>g.spectrogram,stack:()=>w.stack,std_mean:()=>w.std_mean,topk:()=>w.topk,window_function:()=>g.window_function,zeros:()=>w.zeros,zeros_like:()=>w.zeros_like});var u=c(/*! ./env.js */"./src/env.js"),p=c(/*! ./pipelines.js */"./src/pipelines.js"),h=c(/*! ./models.js */"./src/models.js"),m=c(/*! ./tokenizers.js */"./src/tokenizers.js"),_=c(/*! ./processors.js */"./src/processors.js"),f=c(/*! ./configs.js */"./src/configs.js"),g=c(/*! ./utils/audio.js */"./src/utils/audio.js"),M=c(/*! ./utils/image.js */"./src/utils/image.js"),w=c(/*! ./utils/tensor.js */"./src/utils/tensor.js"),b=c(/*! ./utils/maths.js */"./src/utils/maths.js"),T=c(/*! ./generation/streamers.js */"./src/generation/streamers.js"),x=c(/*! ./generation/stopping_criteria.js */"./src/generation/stopping_criteria.js"),y=d.ASTFeatureExtractor,k=d.ASTForAudioClassification,F=d.ASTModel,C=d.ASTPreTrainedModel,P=d.AlbertForMaskedLM,v=d.AlbertForQuestionAnswering,S=d.AlbertForSequenceClassification,A=d.AlbertModel,E=d.AlbertPreTrainedModel,L=d.AlbertTokenizer,z=d.AudioClassificationPipeline,I=d.AutoConfig,B=d.AutoModel,N=d.AutoModelForAudioClassification,O=d.AutoModelForAudioFrameClassification,D=d.AutoModelForCTC,V=d.AutoModelForCausalLM,j=d.AutoModelForDepthEstimation,R=d.AutoModelForDocumentQuestionAnswering,G=d.AutoModelForImageClassification,q=d.AutoModelForImageFeatureExtraction,$=d.AutoModelForImageMatting,W=d.AutoModelForImageSegmentation,U=d.AutoModelForImageToImage,X=d.AutoModelForMaskGeneration,Q=d.AutoModelForMaskedLM,H=d.AutoModelForObjectDetection,Y=d.AutoModelForQuestionAnswering,J=d.AutoModelForSemanticSegmentation,K=d.AutoModelForSeq2SeqLM,Z=d.AutoModelForSequenceClassification,ee=d.AutoModelForSpeechSeq2Seq,te=d.AutoModelForTextToSpectrogram,ne=d.AutoModelForTextToWaveform,re=d.AutoModelForTokenClassification,oe=d.AutoModelForVision2Seq,se=d.AutoModelForXVector,ae=d.AutoModelForZeroShotObjectDetection,ie=d.AutoProcessor,le=d.AutoTokenizer,ce=d.AutomaticSpeechRecognitionPipeline,de=d.BartForConditionalGeneration,ue=d.BartForSequenceClassification,pe=d.BartModel,he=d.BartPretrainedModel,me=d.BartTokenizer,_e=d.BaseModelOutput,fe=d.BaseStreamer,ge=d.BeitFeatureExtractor,Me=d.BeitForImageClassification,we=d.BeitModel,be=d.BeitPreTrainedModel,Te=d.BertForMaskedLM,xe=d.BertForQuestionAnswering,ye=d.BertForSequenceClassification,ke=d.BertForTokenClassification,Fe=d.BertModel,Ce=d.BertPreTrainedModel,Pe=d.BertTokenizer,ve=d.BitImageProcessor,Se=d.BlenderbotForConditionalGeneration,Ae=d.BlenderbotModel,Ee=d.BlenderbotPreTrainedModel,Le=d.BlenderbotSmallForConditionalGeneration,ze=d.BlenderbotSmallModel,Ie=d.BlenderbotSmallPreTrainedModel,Be=d.BlenderbotSmallTokenizer,Ne=d.BlenderbotTokenizer,Oe=d.BloomForCausalLM,De=d.BloomModel,Ve=d.BloomPreTrainedModel,je=d.BloomTokenizer,Re=d.CLIPFeatureExtractor,Ge=d.CLIPImageProcessor,qe=d.CLIPModel,$e=d.CLIPPreTrainedModel,We=d.CLIPSegForImageSegmentation,Ue=d.CLIPSegModel,Xe=d.CLIPSegPreTrainedModel,Qe=d.CLIPTextModelWithProjection,He=d.CLIPTokenizer,Ye=d.CLIPVisionModelWithProjection,Je=d.CamembertForMaskedLM,Ke=d.CamembertForQuestionAnswering,Ze=d.CamembertForSequenceClassification,et=d.CamembertForTokenClassification,tt=d.CamembertModel,nt=d.CamembertPreTrainedModel,rt=d.CamembertTokenizer,ot=d.CausalLMOutput,st=d.CausalLMOutputWithPast,at=d.ChineseCLIPFeatureExtractor,it=d.ChineseCLIPModel,lt=d.ChineseCLIPPreTrainedModel,ct=d.ClapAudioModelWithProjection,dt=d.ClapFeatureExtractor,ut=d.ClapModel,pt=d.ClapPreTrainedModel,ht=d.ClapTextModelWithProjection,mt=d.CodeGenForCausalLM,_t=d.CodeGenModel,ft=d.CodeGenPreTrainedModel,gt=d.CodeGenTokenizer,Mt=d.CodeLlamaTokenizer,wt=d.CohereForCausalLM,bt=d.CohereModel,Tt=d.CoherePreTrainedModel,xt=d.CohereTokenizer,yt=d.ConvBertForMaskedLM,kt=d.ConvBertForQuestionAnswering,Ft=d.ConvBertForSequenceClassification,Ct=d.ConvBertForTokenClassification,Pt=d.ConvBertModel,vt=d.ConvBertPreTrainedModel,St=d.ConvBertTokenizer,At=d.ConvNextFeatureExtractor,Et=d.ConvNextForImageClassification,Lt=d.ConvNextImageProcessor,zt=d.ConvNextModel,It=d.ConvNextPreTrainedModel,Bt=d.ConvNextV2ForImageClassification,Nt=d.ConvNextV2Model,Ot=d.ConvNextV2PreTrainedModel,Dt=d.DPTFeatureExtractor,Vt=d.DPTForDepthEstimation,jt=d.DPTImageProcessor,Rt=d.DPTModel,Gt=d.DPTPreTrainedModel,qt=d.DebertaForMaskedLM,$t=d.DebertaForQuestionAnswering,Wt=d.DebertaForSequenceClassification,Ut=d.DebertaForTokenClassification,Xt=d.DebertaModel,Qt=d.DebertaPreTrainedModel,Ht=d.DebertaTokenizer,Yt=d.DebertaV2ForMaskedLM,Jt=d.DebertaV2ForQuestionAnswering,Kt=d.DebertaV2ForSequenceClassification,Zt=d.DebertaV2ForTokenClassification,en=d.DebertaV2Model,tn=d.DebertaV2PreTrainedModel,nn=d.DebertaV2Tokenizer,rn=d.DeiTFeatureExtractor,on=d.DeiTForImageClassification,sn=d.DeiTModel,an=d.DeiTPreTrainedModel,ln=d.DepthAnythingForDepthEstimation,cn=d.DepthAnythingPreTrainedModel,dn=d.DepthEstimationPipeline,un=d.DetrFeatureExtractor,pn=d.DetrForObjectDetection,hn=d.DetrForSegmentation,mn=d.DetrModel,_n=d.DetrObjectDetectionOutput,fn=d.DetrPreTrainedModel,gn=d.DetrSegmentationOutput,Mn=d.Dinov2ForImageClassification,wn=d.Dinov2Model,bn=d.Dinov2PreTrainedModel,Tn=d.DistilBertForMaskedLM,xn=d.DistilBertForQuestionAnswering,yn=d.DistilBertForSequenceClassification,kn=d.DistilBertForTokenClassification,Fn=d.DistilBertModel,Cn=d.DistilBertPreTrainedModel,Pn=d.DistilBertTokenizer,vn=d.DocumentQuestionAnsweringPipeline,Sn=d.DonutFeatureExtractor,An=d.DonutSwinModel,En=d.DonutSwinPreTrainedModel,Ln=d.EfficientNetForImageClassification,zn=d.EfficientNetImageProcessor,In=d.EfficientNetModel,Bn=d.EfficientNetPreTrainedModel,Nn=d.ElectraForMaskedLM,On=d.ElectraForQuestionAnswering,Dn=d.ElectraForSequenceClassification,Vn=d.ElectraForTokenClassification,jn=d.ElectraModel,Rn=d.ElectraPreTrainedModel,Gn=d.ElectraTokenizer,qn=d.EosTokenCriteria,$n=d.EsmForMaskedLM,Wn=d.EsmForSequenceClassification,Un=d.EsmForTokenClassification,Xn=d.EsmModel,Qn=d.EsmPreTrainedModel,Hn=d.EsmTokenizer,Yn=d.FFT,Jn=d.FalconForCausalLM,Kn=d.FalconModel,Zn=d.FalconPreTrainedModel,er=d.FalconTokenizer,tr=d.FastViTForImageClassification,nr=d.FastViTModel,rr=d.FastViTPreTrainedModel,or=d.FeatureExtractionPipeline,sr=d.FeatureExtractor,ar=d.FillMaskPipeline,ir=d.Florence2ForConditionalGeneration,lr=d.Florence2PreTrainedModel,cr=d.Florence2Processor,dr=d.GLPNFeatureExtractor,ur=d.GLPNForDepthEstimation,pr=d.GLPNModel,hr=d.GLPNPreTrainedModel,mr=d.GPT2LMHeadModel,_r=d.GPT2Model,fr=d.GPT2PreTrainedModel,gr=d.GPT2Tokenizer,Mr=d.GPTBigCodeForCausalLM,wr=d.GPTBigCodeModel,br=d.GPTBigCodePreTrainedModel,Tr=d.GPTJForCausalLM,xr=d.GPTJModel,yr=d.GPTJPreTrainedModel,kr=d.GPTNeoForCausalLM,Fr=d.GPTNeoModel,Cr=d.GPTNeoPreTrainedModel,Pr=d.GPTNeoXForCausalLM,vr=d.GPTNeoXModel,Sr=d.GPTNeoXPreTrainedModel,Ar=d.GPTNeoXTokenizer,Er=d.Gemma2ForCausalLM,Lr=d.Gemma2Model,zr=d.Gemma2PreTrainedModel,Ir=d.GemmaForCausalLM,Br=d.GemmaModel,Nr=d.GemmaPreTrainedModel,Or=d.GemmaTokenizer,Dr=d.Grok1Tokenizer,Vr=d.HerbertTokenizer,jr=d.HubertForCTC,Rr=d.HubertForSequenceClassification,Gr=d.HubertModel,qr=d.HubertPreTrainedModel,$r=d.ImageClassificationPipeline,Wr=d.ImageFeatureExtractionPipeline,Ur=d.ImageFeatureExtractor,Xr=d.ImageMattingOutput,Qr=d.ImageSegmentationPipeline,Hr=d.ImageToImagePipeline,Yr=d.ImageToTextPipeline,Jr=d.InterruptableStoppingCriteria,Kr=d.JAISLMHeadModel,Zr=d.JAISModel,eo=d.JAISPreTrainedModel,to=d.LlamaForCausalLM,no=d.LlamaModel,ro=d.LlamaPreTrainedModel,oo=d.LlamaTokenizer,so=d.LlavaForConditionalGeneration,ao=d.LlavaPreTrainedModel,io=d.LongT5ForConditionalGeneration,lo=d.LongT5Model,co=d.LongT5PreTrainedModel,uo=d.M2M100ForConditionalGeneration,po=d.M2M100Model,ho=d.M2M100PreTrainedModel,mo=d.M2M100Tokenizer,_o=d.MBart50Tokenizer,fo=d.MBartForCausalLM,go=d.MBartForConditionalGeneration,Mo=d.MBartForSequenceClassification,wo=d.MBartModel,bo=d.MBartPreTrainedModel,To=d.MBartTokenizer,xo=d.MPNetForMaskedLM,yo=d.MPNetForQuestionAnswering,ko=d.MPNetForSequenceClassification,Fo=d.MPNetForTokenClassification,Co=d.MPNetModel,Po=d.MPNetPreTrainedModel,vo=d.MPNetTokenizer,So=d.MT5ForConditionalGeneration,Ao=d.MT5Model,Eo=d.MT5PreTrainedModel,Lo=d.MarianMTModel,zo=d.MarianModel,Io=d.MarianPreTrainedModel,Bo=d.MarianTokenizer,No=d.MaskedLMOutput,Oo=d.MaxLengthCriteria,Do=d.MistralForCausalLM,Vo=d.MistralModel,jo=d.MistralPreTrainedModel,Ro=d.MobileBertForMaskedLM,Go=d.MobileBertForQuestionAnswering,qo=d.MobileBertForSequenceClassification,$o=d.MobileBertModel,Wo=d.MobileBertPreTrainedModel,Uo=d.MobileBertTokenizer,Xo=d.MobileNetV1FeatureExtractor,Qo=d.MobileNetV1ForImageClassification,Ho=d.MobileNetV1Model,Yo=d.MobileNetV1PreTrainedModel,Jo=d.MobileNetV2FeatureExtractor,Ko=d.MobileNetV2ForImageClassification,Zo=d.MobileNetV2Model,es=d.MobileNetV2PreTrainedModel,ts=d.MobileNetV3FeatureExtractor,ns=d.MobileNetV3ForImageClassification,rs=d.MobileNetV3Model,os=d.MobileNetV3PreTrainedModel,ss=d.MobileNetV4FeatureExtractor,as=d.MobileNetV4ForImageClassification,is=d.MobileNetV4Model,ls=d.MobileNetV4PreTrainedModel,cs=d.MobileViTFeatureExtractor,ds=d.MobileViTForImageClassification,us=d.MobileViTImageProcessor,ps=d.MobileViTModel,hs=d.MobileViTPreTrainedModel,ms=d.MobileViTV2ForImageClassification,_s=d.MobileViTV2Model,fs=d.MobileViTV2PreTrainedModel,gs=d.ModelOutput,Ms=d.Moondream1ForConditionalGeneration,ws=d.MptForCausalLM,bs=d.MptModel,Ts=d.MptPreTrainedModel,xs=d.MusicgenForCausalLM,ys=d.MusicgenForConditionalGeneration,ks=d.MusicgenModel,Fs=d.MusicgenPreTrainedModel,Cs=d.NllbTokenizer,Ps=d.NomicBertModel,vs=d.NomicBertPreTrainedModel,Ss=d.NougatImageProcessor,As=d.NougatTokenizer,Es=d.OPTForCausalLM,Ls=d.OPTModel,zs=d.OPTPreTrainedModel,Is=d.ObjectDetectionPipeline,Bs=d.OpenELMForCausalLM,Ns=d.OpenELMModel,Os=d.OpenELMPreTrainedModel,Ds=d.OwlViTFeatureExtractor,Vs=d.OwlViTForObjectDetection,js=d.OwlViTModel,Rs=d.OwlViTPreTrainedModel,Gs=d.OwlViTProcessor,qs=d.Owlv2ForObjectDetection,$s=d.Owlv2ImageProcessor,Ws=d.Owlv2Model,Us=d.Owlv2PreTrainedModel,Xs=d.Phi3ForCausalLM,Qs=d.Phi3Model,Hs=d.Phi3PreTrainedModel,Ys=d.PhiForCausalLM,Js=d.PhiModel,Ks=d.PhiPreTrainedModel,Zs=d.Pipeline,ea=d.PreTrainedModel,ta=d.PreTrainedTokenizer,na=d.PretrainedConfig,ra=d.PretrainedMixin,oa=d.Processor,sa=d.PyAnnoteFeatureExtractor,aa=d.PyAnnoteForAudioFrameClassification,ia=d.PyAnnoteModel,la=d.PyAnnotePreTrainedModel,ca=d.PyAnnoteProcessor,da=d.QuestionAnsweringModelOutput,ua=d.QuestionAnsweringPipeline,pa=d.Qwen2ForCausalLM,ha=d.Qwen2Model,ma=d.Qwen2PreTrainedModel,_a=d.Qwen2Tokenizer,fa=d.RTDetrForObjectDetection,ga=d.RTDetrImageProcessor,Ma=d.RTDetrModel,wa=d.RTDetrObjectDetectionOutput,ba=d.RTDetrPreTrainedModel,Ta=d.RawImage,xa=d.ResNetForImageClassification,ya=d.ResNetModel,ka=d.ResNetPreTrainedModel,Fa=d.RoFormerForMaskedLM,Ca=d.RoFormerForQuestionAnswering,Pa=d.RoFormerForSequenceClassification,va=d.RoFormerForTokenClassification,Sa=d.RoFormerModel,Aa=d.RoFormerPreTrainedModel,Ea=d.RoFormerTokenizer,La=d.RobertaForMaskedLM,za=d.RobertaForQuestionAnswering,Ia=d.RobertaForSequenceClassification,Ba=d.RobertaForTokenClassification,Na=d.RobertaModel,Oa=d.RobertaPreTrainedModel,Da=d.RobertaTokenizer,Va=d.SamImageProcessor,ja=d.SamImageSegmentationOutput,Ra=d.SamModel,Ga=d.SamPreTrainedModel,qa=d.SamProcessor,$a=d.SapiensFeatureExtractor,Wa=d.SapiensForDepthEstimation,Ua=d.SapiensForNormalEstimation,Xa=d.SapiensForSemanticSegmentation,Qa=d.SapiensPreTrainedModel,Ha=d.SeamlessM4TFeatureExtractor,Ya=d.SegformerFeatureExtractor,Ja=d.SegformerForImageClassification,Ka=d.SegformerForSemanticSegmentation,Za=d.SegformerModel,ei=d.SegformerPreTrainedModel,ti=d.Seq2SeqLMOutput,ni=d.SequenceClassifierOutput,ri=d.SiglipImageProcessor,oi=d.SiglipModel,si=d.SiglipPreTrainedModel,ai=d.SiglipTextModel,ii=d.SiglipTokenizer,li=d.SiglipVisionModel,ci=d.SpeechT5FeatureExtractor,di=d.SpeechT5ForSpeechToText,ui=d.SpeechT5ForTextToSpeech,pi=d.SpeechT5HifiGan,hi=d.SpeechT5Model,mi=d.SpeechT5PreTrainedModel,_i=d.SpeechT5Processor,fi=d.SpeechT5Tokenizer,gi=d.SqueezeBertForMaskedLM,Mi=d.SqueezeBertForQuestionAnswering,wi=d.SqueezeBertForSequenceClassification,bi=d.SqueezeBertModel,Ti=d.SqueezeBertPreTrainedModel,xi=d.SqueezeBertTokenizer,yi=d.StableLmForCausalLM,ki=d.StableLmModel,Fi=d.StableLmPreTrainedModel,Ci=d.Starcoder2ForCausalLM,Pi=d.Starcoder2Model,vi=d.Starcoder2PreTrainedModel,Si=d.StoppingCriteria,Ai=d.StoppingCriteriaList,Ei=d.SummarizationPipeline,Li=d.Swin2SRForImageSuperResolution,zi=d.Swin2SRImageProcessor,Ii=d.Swin2SRModel,Bi=d.Swin2SRPreTrainedModel,Ni=d.SwinForImageClassification,Oi=d.SwinModel,Di=d.SwinPreTrainedModel,Vi=d.T5ForConditionalGeneration,ji=d.T5Model,Ri=d.T5PreTrainedModel,Gi=d.T5Tokenizer,qi=d.TableTransformerForObjectDetection,$i=d.TableTransformerModel,Wi=d.TableTransformerObjectDetectionOutput,Ui=d.TableTransformerPreTrainedModel,Xi=d.Tensor,Qi=d.Text2TextGenerationPipeline,Hi=d.TextClassificationPipeline,Yi=d.TextGenerationPipeline,Ji=d.TextStreamer,Ki=d.TextToAudioPipeline,Zi=d.TokenClassificationPipeline,el=d.TokenClassifierOutput,tl=d.TokenizerModel,nl=d.TrOCRForCausalLM,rl=d.TrOCRPreTrainedModel,ol=d.TranslationPipeline,sl=d.UniSpeechForCTC,al=d.UniSpeechForSequenceClassification,il=d.UniSpeechModel,ll=d.UniSpeechPreTrainedModel,cl=d.UniSpeechSatForAudioFrameClassification,dl=d.UniSpeechSatForCTC,ul=d.UniSpeechSatForSequenceClassification,pl=d.UniSpeechSatModel,hl=d.UniSpeechSatPreTrainedModel,ml=d.ViTFeatureExtractor,_l=d.ViTForImageClassification,fl=d.ViTImageProcessor,gl=d.ViTModel,Ml=d.ViTPreTrainedModel,wl=d.VisionEncoderDecoderModel,bl=d.VitMatteForImageMatting,Tl=d.VitMatteImageProcessor,xl=d.VitMattePreTrainedModel,yl=d.VitsModel,kl=d.VitsModelOutput,Fl=d.VitsPreTrainedModel,Cl=d.VitsTokenizer,Pl=d.Wav2Vec2BertForCTC,vl=d.Wav2Vec2BertForSequenceClassification,Sl=d.Wav2Vec2BertModel,Al=d.Wav2Vec2BertPreTrainedModel,El=d.Wav2Vec2CTCTokenizer,Ll=d.Wav2Vec2FeatureExtractor,zl=d.Wav2Vec2ForAudioFrameClassification,Il=d.Wav2Vec2ForCTC,Bl=d.Wav2Vec2ForSequenceClassification,Nl=d.Wav2Vec2Model,Ol=d.Wav2Vec2PreTrainedModel,Dl=d.Wav2Vec2ProcessorWithLM,Vl=d.WavLMForAudioFrameClassification,jl=d.WavLMForCTC,Rl=d.WavLMForSequenceClassification,Gl=d.WavLMForXVector,ql=d.WavLMModel,$l=d.WavLMPreTrainedModel,Wl=d.WeSpeakerFeatureExtractor,Ul=d.WeSpeakerResNetModel,Xl=d.WeSpeakerResNetPreTrainedModel,Ql=d.WhisperFeatureExtractor,Hl=d.WhisperForConditionalGeneration,Yl=d.WhisperModel,Jl=d.WhisperPreTrainedModel,Kl=d.WhisperProcessor,Zl=d.WhisperTextStreamer,ec=d.WhisperTokenizer,tc=d.XLMForQuestionAnswering,nc=d.XLMForSequenceClassification,rc=d.XLMForTokenClassification,oc=d.XLMModel,sc=d.XLMPreTrainedModel,ac=d.XLMRobertaForMaskedLM,ic=d.XLMRobertaForQuestionAnswering,lc=d.XLMRobertaForSequenceClassification,cc=d.XLMRobertaForTokenClassification,dc=d.XLMRobertaModel,uc=d.XLMRobertaPreTrainedModel,pc=d.XLMRobertaTokenizer,hc=d.XLMTokenizer,mc=d.XLMWithLMHeadModel,_c=d.XVectorOutput,fc=d.YolosFeatureExtractor,gc=d.YolosForObjectDetection,Mc=d.YolosModel,wc=d.YolosObjectDetectionOutput,bc=d.YolosPreTrainedModel,Tc=d.ZeroShotAudioClassificationPipeline,xc=d.ZeroShotClassificationPipeline,yc=d.ZeroShotImageClassificationPipeline,kc=d.ZeroShotObjectDetectionPipeline,Fc=d.bankers_round,Cc=d.cat,Pc=d.cos_sim,vc=d.dot,Sc=d.dynamic_time_warping,Ac=d.env,Ec=d.full,Lc=d.full_like,zc=d.getKeyValueShapes,Ic=d.hamming,Bc=d.hanning,Nc=d.interpolate,Oc=d.interpolate_4d,Dc=d.interpolate_data,Vc=d.is_chinese_char,jc=d.layer_norm,Rc=d.log_softmax,Gc=d.magnitude,qc=d.matmul,$c=d.max,Wc=d.mean,Uc=d.mean_pooling,Xc=d.medianFilter,Qc=d.mel_filter_bank,Hc=d.min,Yc=d.ones,Jc=d.ones_like,Kc=d.permute,Zc=d.permute_data,ed=d.pipeline,td=d.quantize_embeddings,nd=d.read_audio,rd=d.rfft,od=d.round,sd=d.softmax,ad=d.spectrogram,id=d.stack,ld=d.std_mean,cd=d.topk,dd=d.window_function,ud=d.zeros,pd=d.zeros_like;export{y as ASTFeatureExtractor,k as ASTForAudioClassification,F as ASTModel,C as ASTPreTrainedModel,P as AlbertForMaskedLM,v as AlbertForQuestionAnswering,S as AlbertForSequenceClassification,A as AlbertModel,E as AlbertPreTrainedModel,L as AlbertTokenizer,z as AudioClassificationPipeline,I as AutoConfig,B as AutoModel,N as AutoModelForAudioClassification,O as AutoModelForAudioFrameClassification,D as AutoModelForCTC,V as AutoModelForCausalLM,j as AutoModelForDepthEstimation,R as AutoModelForDocumentQuestionAnswering,G as AutoModelForImageClassification,q as AutoModelForImageFeatureExtraction,$ as AutoModelForImageMatting,W as AutoModelForImageSegmentation,U as AutoModelForImageToImage,X as AutoModelForMaskGeneration,Q as AutoModelForMaskedLM,H as AutoModelForObjectDetection,Y as AutoModelForQuestionAnswering,J as AutoModelForSemanticSegmentation,K as AutoModelForSeq2SeqLM,Z as AutoModelForSequenceClassification,ee as AutoModelForSpeechSeq2Seq,te as AutoModelForTextToSpectrogram,ne as AutoModelForTextToWaveform,re as AutoModelForTokenClassification,oe as AutoModelForVision2Seq,se as AutoModelForXVector,ae as AutoModelForZeroShotObjectDetection,ie as AutoProcessor,le as AutoTokenizer,ce as AutomaticSpeechRecognitionPipeline,de as BartForConditionalGeneration,ue as BartForSequenceClassification,pe as BartModel,he as BartPretrainedModel,me as BartTokenizer,_e as BaseModelOutput,fe as BaseStreamer,ge as BeitFeatureExtractor,Me as BeitForImageClassification,we as BeitModel,be as BeitPreTrainedModel,Te as BertForMaskedLM,xe as BertForQuestionAnswering,ye as BertForSequenceClassification,ke as BertForTokenClassification,Fe as BertModel,Ce as BertPreTrainedModel,Pe as BertTokenizer,ve as BitImageProcessor,Se as BlenderbotForConditionalGeneration,Ae as BlenderbotModel,Ee as BlenderbotPreTrainedModel,Le as BlenderbotSmallForConditionalGeneration,ze as BlenderbotSmallModel,Ie as BlenderbotSmallPreTrainedModel,Be as BlenderbotSmallTokenizer,Ne as BlenderbotTokenizer,Oe as BloomForCausalLM,De as BloomModel,Ve as BloomPreTrainedModel,je as BloomTokenizer,Re as CLIPFeatureExtractor,Ge as CLIPImageProcessor,qe as CLIPModel,$e as CLIPPreTrainedModel,We as CLIPSegForImageSegmentation,Ue as CLIPSegModel,Xe as CLIPSegPreTrainedModel,Qe as CLIPTextModelWithProjection,He as CLIPTokenizer,Ye as CLIPVisionModelWithProjection,Je as CamembertForMaskedLM,Ke as CamembertForQuestionAnswering,Ze as CamembertForSequenceClassification,et as CamembertForTokenClassification,tt as CamembertModel,nt as CamembertPreTrainedModel,rt as CamembertTokenizer,ot as CausalLMOutput,st as CausalLMOutputWithPast,at as ChineseCLIPFeatureExtractor,it as ChineseCLIPModel,lt as ChineseCLIPPreTrainedModel,ct as ClapAudioModelWithProjection,dt as ClapFeatureExtractor,ut as ClapModel,pt as ClapPreTrainedModel,ht as ClapTextModelWithProjection,mt as CodeGenForCausalLM,_t as CodeGenModel,ft as CodeGenPreTrainedModel,gt as CodeGenTokenizer,Mt as CodeLlamaTokenizer,wt as CohereForCausalLM,bt as CohereModel,Tt as CoherePreTrainedModel,xt as CohereTokenizer,yt as ConvBertForMaskedLM,kt as ConvBertForQuestionAnswering,Ft as ConvBertForSequenceClassification,Ct as ConvBertForTokenClassification,Pt as ConvBertModel,vt as ConvBertPreTrainedModel,St as ConvBertTokenizer,At as ConvNextFeatureExtractor,Et as ConvNextForImageClassification,Lt as ConvNextImageProcessor,zt as ConvNextModel,It as ConvNextPreTrainedModel,Bt as ConvNextV2ForImageClassification,Nt as ConvNextV2Model,Ot as ConvNextV2PreTrainedModel,Dt as DPTFeatureExtractor,Vt as DPTForDepthEstimation,jt as DPTImageProcessor,Rt as DPTModel,Gt as DPTPreTrainedModel,qt as DebertaForMaskedLM,$t as DebertaForQuestionAnswering,Wt as DebertaForSequenceClassification,Ut as DebertaForTokenClassification,Xt as DebertaModel,Qt as DebertaPreTrainedModel,Ht as DebertaTokenizer,Yt as DebertaV2ForMaskedLM,Jt as DebertaV2ForQuestionAnswering,Kt as DebertaV2ForSequenceClassification,Zt as DebertaV2ForTokenClassification,en as DebertaV2Model,tn as DebertaV2PreTrainedModel,nn as DebertaV2Tokenizer,rn as DeiTFeatureExtractor,on as DeiTForImageClassification,sn as DeiTModel,an as DeiTPreTrainedModel,ln as DepthAnythingForDepthEstimation,cn as DepthAnythingPreTrainedModel,dn as DepthEstimationPipeline,un as DetrFeatureExtractor,pn as DetrForObjectDetection,hn as DetrForSegmentation,mn as DetrModel,_n as DetrObjectDetectionOutput,fn as DetrPreTrainedModel,gn as DetrSegmentationOutput,Mn as Dinov2ForImageClassification,wn as Dinov2Model,bn as Dinov2PreTrainedModel,Tn as DistilBertForMaskedLM,xn as DistilBertForQuestionAnswering,yn as DistilBertForSequenceClassification,kn as DistilBertForTokenClassification,Fn as DistilBertModel,Cn as DistilBertPreTrainedModel,Pn as DistilBertTokenizer,vn as DocumentQuestionAnsweringPipeline,Sn as DonutFeatureExtractor,An as DonutSwinModel,En as DonutSwinPreTrainedModel,Ln as EfficientNetForImageClassification,zn as EfficientNetImageProcessor,In as EfficientNetModel,Bn as EfficientNetPreTrainedModel,Nn as ElectraForMaskedLM,On as ElectraForQuestionAnswering,Dn as ElectraForSequenceClassification,Vn as ElectraForTokenClassification,jn as ElectraModel,Rn as ElectraPreTrainedModel,Gn as ElectraTokenizer,qn as EosTokenCriteria,$n as EsmForMaskedLM,Wn as EsmForSequenceClassification,Un as EsmForTokenClassification,Xn as EsmModel,Qn as EsmPreTrainedModel,Hn as EsmTokenizer,Yn as FFT,Jn as FalconForCausalLM,Kn as FalconModel,Zn as FalconPreTrainedModel,er as FalconTokenizer,tr as FastViTForImageClassification,nr as FastViTModel,rr as FastViTPreTrainedModel,or as FeatureExtractionPipeline,sr as FeatureExtractor,ar as FillMaskPipeline,ir as Florence2ForConditionalGeneration,lr as Florence2PreTrainedModel,cr as Florence2Processor,dr as GLPNFeatureExtractor,ur as GLPNForDepthEstimation,pr as GLPNModel,hr as GLPNPreTrainedModel,mr as GPT2LMHeadModel,_r as GPT2Model,fr as GPT2PreTrainedModel,gr as GPT2Tokenizer,Mr as GPTBigCodeForCausalLM,wr as GPTBigCodeModel,br as GPTBigCodePreTrainedModel,Tr as GPTJForCausalLM,xr as GPTJModel,yr as GPTJPreTrainedModel,kr as GPTNeoForCausalLM,Fr as GPTNeoModel,Cr as GPTNeoPreTrainedModel,Pr as GPTNeoXForCausalLM,vr as GPTNeoXModel,Sr as GPTNeoXPreTrainedModel,Ar as GPTNeoXTokenizer,Er as Gemma2ForCausalLM,Lr as Gemma2Model,zr as Gemma2PreTrainedModel,Ir as GemmaForCausalLM,Br as GemmaModel,Nr as GemmaPreTrainedModel,Or as GemmaTokenizer,Dr as Grok1Tokenizer,Vr as HerbertTokenizer,jr as HubertForCTC,Rr as HubertForSequenceClassification,Gr as HubertModel,qr as HubertPreTrainedModel,$r as ImageClassificationPipeline,Wr as ImageFeatureExtractionPipeline,Ur as ImageFeatureExtractor,Xr as ImageMattingOutput,Qr as ImageSegmentationPipeline,Hr as ImageToImagePipeline,Yr as ImageToTextPipeline,Jr as InterruptableStoppingCriteria,Kr as JAISLMHeadModel,Zr as JAISModel,eo as JAISPreTrainedModel,to as LlamaForCausalLM,no as LlamaModel,ro as LlamaPreTrainedModel,oo as LlamaTokenizer,so as LlavaForConditionalGeneration,ao as LlavaPreTrainedModel,io as LongT5ForConditionalGeneration,lo as LongT5Model,co as LongT5PreTrainedModel,uo as M2M100ForConditionalGeneration,po as M2M100Model,ho as M2M100PreTrainedModel,mo as M2M100Tokenizer,_o as MBart50Tokenizer,fo as MBartForCausalLM,go as MBartForConditionalGeneration,Mo as MBartForSequenceClassification,wo as MBartModel,bo as MBartPreTrainedModel,To as MBartTokenizer,xo as MPNetForMaskedLM,yo as MPNetForQuestionAnswering,ko as MPNetForSequenceClassification,Fo as MPNetForTokenClassification,Co as MPNetModel,Po as MPNetPreTrainedModel,vo as MPNetTokenizer,So as MT5ForConditionalGeneration,Ao as MT5Model,Eo as MT5PreTrainedModel,Lo as MarianMTModel,zo as MarianModel,Io as MarianPreTrainedModel,Bo as MarianTokenizer,No as MaskedLMOutput,Oo as MaxLengthCriteria,Do as MistralForCausalLM,Vo as MistralModel,jo as MistralPreTrainedModel,Ro as MobileBertForMaskedLM,Go as MobileBertForQuestionAnswering,qo as MobileBertForSequenceClassification,$o as MobileBertModel,Wo as MobileBertPreTrainedModel,Uo as MobileBertTokenizer,Xo as MobileNetV1FeatureExtractor,Qo as MobileNetV1ForImageClassification,Ho as MobileNetV1Model,Yo as MobileNetV1PreTrainedModel,Jo as MobileNetV2FeatureExtractor,Ko as MobileNetV2ForImageClassification,Zo as MobileNetV2Model,es as MobileNetV2PreTrainedModel,ts as MobileNetV3FeatureExtractor,ns as MobileNetV3ForImageClassification,rs as MobileNetV3Model,os as MobileNetV3PreTrainedModel,ss as MobileNetV4FeatureExtractor,as as MobileNetV4ForImageClassification,is as MobileNetV4Model,ls as MobileNetV4PreTrainedModel,cs as MobileViTFeatureExtractor,ds as MobileViTForImageClassification,us as MobileViTImageProcessor,ps as MobileViTModel,hs as MobileViTPreTrainedModel,ms as MobileViTV2ForImageClassification,_s as MobileViTV2Model,fs as MobileViTV2PreTrainedModel,gs as ModelOutput,Ms as Moondream1ForConditionalGeneration,ws as MptForCausalLM,bs as MptModel,Ts as MptPreTrainedModel,xs as MusicgenForCausalLM,ys as MusicgenForConditionalGeneration,ks as MusicgenModel,Fs as MusicgenPreTrainedModel,Cs as NllbTokenizer,Ps as NomicBertModel,vs as NomicBertPreTrainedModel,Ss as NougatImageProcessor,As as NougatTokenizer,Es as OPTForCausalLM,Ls as OPTModel,zs as OPTPreTrainedModel,Is as ObjectDetectionPipeline,Bs as OpenELMForCausalLM,Ns as OpenELMModel,Os as OpenELMPreTrainedModel,Ds as OwlViTFeatureExtractor,Vs as OwlViTForObjectDetection,js as OwlViTModel,Rs as OwlViTPreTrainedModel,Gs as OwlViTProcessor,qs as Owlv2ForObjectDetection,$s as Owlv2ImageProcessor,Ws as Owlv2Model,Us as Owlv2PreTrainedModel,Xs as Phi3ForCausalLM,Qs as Phi3Model,Hs as Phi3PreTrainedModel,Ys as PhiForCausalLM,Js as PhiModel,Ks as PhiPreTrainedModel,Zs as Pipeline,ea as PreTrainedModel,ta as PreTrainedTokenizer,na as PretrainedConfig,ra as PretrainedMixin,oa as Processor,sa as PyAnnoteFeatureExtractor,aa as PyAnnoteForAudioFrameClassification,ia as PyAnnoteModel,la as PyAnnotePreTrainedModel,ca as PyAnnoteProcessor,da as QuestionAnsweringModelOutput,ua as QuestionAnsweringPipeline,pa as Qwen2ForCausalLM,ha as Qwen2Model,ma as Qwen2PreTrainedModel,_a as Qwen2Tokenizer,fa as RTDetrForObjectDetection,ga as RTDetrImageProcessor,Ma as RTDetrModel,wa as RTDetrObjectDetectionOutput,ba as RTDetrPreTrainedModel,Ta as RawImage,xa as ResNetForImageClassification,ya as ResNetModel,ka as ResNetPreTrainedModel,Fa as RoFormerForMaskedLM,Ca as RoFormerForQuestionAnswering,Pa as RoFormerForSequenceClassification,va as RoFormerForTokenClassification,Sa as RoFormerModel,Aa as RoFormerPreTrainedModel,Ea as RoFormerTokenizer,La as RobertaForMaskedLM,za as RobertaForQuestionAnswering,Ia as RobertaForSequenceClassification,Ba as RobertaForTokenClassification,Na as RobertaModel,Oa as RobertaPreTrainedModel,Da as RobertaTokenizer,Va as SamImageProcessor,ja as SamImageSegmentationOutput,Ra as SamModel,Ga as SamPreTrainedModel,qa as SamProcessor,$a as SapiensFeatureExtractor,Wa as SapiensForDepthEstimation,Ua as SapiensForNormalEstimation,Xa as SapiensForSemanticSegmentation,Qa as SapiensPreTrainedModel,Ha as SeamlessM4TFeatureExtractor,Ya as SegformerFeatureExtractor,Ja as SegformerForImageClassification,Ka as SegformerForSemanticSegmentation,Za as SegformerModel,ei as SegformerPreTrainedModel,ti as Seq2SeqLMOutput,ni as SequenceClassifierOutput,ri as SiglipImageProcessor,oi as SiglipModel,si as SiglipPreTrainedModel,ai as SiglipTextModel,ii as SiglipTokenizer,li as SiglipVisionModel,ci as SpeechT5FeatureExtractor,di as SpeechT5ForSpeechToText,ui as SpeechT5ForTextToSpeech,pi as SpeechT5HifiGan,hi as SpeechT5Model,mi as SpeechT5PreTrainedModel,_i as SpeechT5Processor,fi as SpeechT5Tokenizer,gi as SqueezeBertForMaskedLM,Mi as SqueezeBertForQuestionAnswering,wi as SqueezeBertForSequenceClassification,bi as SqueezeBertModel,Ti as SqueezeBertPreTrainedModel,xi as SqueezeBertTokenizer,yi as StableLmForCausalLM,ki as StableLmModel,Fi as StableLmPreTrainedModel,Ci as Starcoder2ForCausalLM,Pi as Starcoder2Model,vi as Starcoder2PreTrainedModel,Si as StoppingCriteria,Ai as StoppingCriteriaList,Ei as SummarizationPipeline,Li as Swin2SRForImageSuperResolution,zi as Swin2SRImageProcessor,Ii as Swin2SRModel,Bi as Swin2SRPreTrainedModel,Ni as SwinForImageClassification,Oi as SwinModel,Di as SwinPreTrainedModel,Vi as T5ForConditionalGeneration,ji as T5Model,Ri as T5PreTrainedModel,Gi as T5Tokenizer,qi as TableTransformerForObjectDetection,$i as TableTransformerModel,Wi as TableTransformerObjectDetectionOutput,Ui as TableTransformerPreTrainedModel,Xi as Tensor,Qi as Text2TextGenerationPipeline,Hi as TextClassificationPipeline,Yi as TextGenerationPipeline,Ji as TextStreamer,Ki as TextToAudioPipeline,Zi as TokenClassificationPipeline,el as TokenClassifierOutput,tl as TokenizerModel,nl as TrOCRForCausalLM,rl as TrOCRPreTrainedModel,ol as TranslationPipeline,sl as UniSpeechForCTC,al as UniSpeechForSequenceClassification,il as UniSpeechModel,ll as UniSpeechPreTrainedModel,cl as UniSpeechSatForAudioFrameClassification,dl as UniSpeechSatForCTC,ul as UniSpeechSatForSequenceClassification,pl as UniSpeechSatModel,hl as UniSpeechSatPreTrainedModel,ml as ViTFeatureExtractor,_l as ViTForImageClassification,fl as ViTImageProcessor,gl as ViTModel,Ml as ViTPreTrainedModel,wl as VisionEncoderDecoderModel,bl as VitMatteForImageMatting,Tl as VitMatteImageProcessor,xl as VitMattePreTrainedModel,yl as VitsModel,kl as VitsModelOutput,Fl as VitsPreTrainedModel,Cl as VitsTokenizer,Pl as Wav2Vec2BertForCTC,vl as Wav2Vec2BertForSequenceClassification,Sl as Wav2Vec2BertModel,Al as Wav2Vec2BertPreTrainedModel,El as Wav2Vec2CTCTokenizer,Ll as Wav2Vec2FeatureExtractor,zl as Wav2Vec2ForAudioFrameClassification,Il as Wav2Vec2ForCTC,Bl as Wav2Vec2ForSequenceClassification,Nl as Wav2Vec2Model,Ol as Wav2Vec2PreTrainedModel,Dl as Wav2Vec2ProcessorWithLM,Vl as WavLMForAudioFrameClassification,jl as WavLMForCTC,Rl as WavLMForSequenceClassification,Gl as WavLMForXVector,ql as WavLMModel,$l as WavLMPreTrainedModel,Wl as WeSpeakerFeatureExtractor,Ul as WeSpeakerResNetModel,Xl as WeSpeakerResNetPreTrainedModel,Ql as WhisperFeatureExtractor,Hl as WhisperForConditionalGeneration,Yl as WhisperModel,Jl as WhisperPreTrainedModel,Kl as WhisperProcessor,Zl as WhisperTextStreamer,ec as WhisperTokenizer,tc as XLMForQuestionAnswering,nc as XLMForSequenceClassification,rc as XLMForTokenClassification,oc as XLMModel,sc as XLMPreTrainedModel,ac as XLMRobertaForMaskedLM,ic as XLMRobertaForQuestionAnswering,lc as XLMRobertaForSequenceClassification,cc as XLMRobertaForTokenClassification,dc as XLMRobertaModel,uc as XLMRobertaPreTrainedModel,pc as XLMRobertaTokenizer,hc as XLMTokenizer,mc as XLMWithLMHeadModel,_c as XVectorOutput,fc as YolosFeatureExtractor,gc as YolosForObjectDetection,Mc as YolosModel,wc as YolosObjectDetectionOutput,bc as YolosPreTrainedModel,Tc as ZeroShotAudioClassificationPipeline,xc as ZeroShotClassificationPipeline,yc as ZeroShotImageClassificationPipeline,kc as ZeroShotObjectDetectionPipeline,Fc as bankers_round,Cc as cat,Pc as cos_sim,vc as dot,Sc as dynamic_time_warping,Ac as env,Ec as full,Lc as full_like,zc as getKeyValueShapes,Ic as hamming,Bc as hanning,Nc as interpolate,Oc as interpolate_4d,Dc as interpolate_data,Vc as is_chinese_char,jc as layer_norm,Rc as log_softmax,Gc as magnitude,qc as matmul,$c as max,Wc as mean,Uc as mean_pooling,Xc as medianFilter,Qc as mel_filter_bank,Hc as min,Yc as ones,Jc as ones_like,Kc as permute,Zc as permute_data,ed as pipeline,td as quantize_embeddings,nd as read_audio,rd as rfft,od as round,sd as softmax,ad as spectrogram,id as stack,ld as std_mean,cd as topk,dd as window_function,ud as zeros,pd as zeros_like};
+  \*****************************/c.r(d),c.d(d,{ASTFeatureExtractor:()=>_.ASTFeatureExtractor,ASTForAudioClassification:()=>h.ASTForAudioClassification,ASTModel:()=>h.ASTModel,ASTPreTrainedModel:()=>h.ASTPreTrainedModel,AlbertForMaskedLM:()=>h.AlbertForMaskedLM,AlbertForQuestionAnswering:()=>h.AlbertForQuestionAnswering,AlbertForSequenceClassification:()=>h.AlbertForSequenceClassification,AlbertModel:()=>h.AlbertModel,AlbertPreTrainedModel:()=>h.AlbertPreTrainedModel,AlbertTokenizer:()=>m.AlbertTokenizer,AudioClassificationPipeline:()=>p.AudioClassificationPipeline,AutoConfig:()=>f.AutoConfig,AutoModel:()=>h.AutoModel,AutoModelForAudioClassification:()=>h.AutoModelForAudioClassification,AutoModelForAudioFrameClassification:()=>h.AutoModelForAudioFrameClassification,AutoModelForCTC:()=>h.AutoModelForCTC,AutoModelForCausalLM:()=>h.AutoModelForCausalLM,AutoModelForDepthEstimation:()=>h.AutoModelForDepthEstimation,AutoModelForDocumentQuestionAnswering:()=>h.AutoModelForDocumentQuestionAnswering,AutoModelForImageClassification:()=>h.AutoModelForImageClassification,AutoModelForImageFeatureExtraction:()=>h.AutoModelForImageFeatureExtraction,AutoModelForImageMatting:()=>h.AutoModelForImageMatting,AutoModelForImageSegmentation:()=>h.AutoModelForImageSegmentation,AutoModelForImageToImage:()=>h.AutoModelForImageToImage,AutoModelForMaskGeneration:()=>h.AutoModelForMaskGeneration,AutoModelForMaskedLM:()=>h.AutoModelForMaskedLM,AutoModelForNormalEstimation:()=>h.AutoModelForNormalEstimation,AutoModelForObjectDetection:()=>h.AutoModelForObjectDetection,AutoModelForQuestionAnswering:()=>h.AutoModelForQuestionAnswering,AutoModelForSemanticSegmentation:()=>h.AutoModelForSemanticSegmentation,AutoModelForSeq2SeqLM:()=>h.AutoModelForSeq2SeqLM,AutoModelForSequenceClassification:()=>h.AutoModelForSequenceClassification,AutoModelForSpeechSeq2Seq:()=>h.AutoModelForSpeechSeq2Seq,AutoModelForTextToSpectrogram:()=>h.AutoModelForTextToSpectrogram,AutoModelForTextToWaveform:()=>h.AutoModelForTextToWaveform,AutoModelForTokenClassification:()=>h.AutoModelForTokenClassification,AutoModelForVision2Seq:()=>h.AutoModelForVision2Seq,AutoModelForXVector:()=>h.AutoModelForXVector,AutoModelForZeroShotObjectDetection:()=>h.AutoModelForZeroShotObjectDetection,AutoProcessor:()=>_.AutoProcessor,AutoTokenizer:()=>m.AutoTokenizer,AutomaticSpeechRecognitionPipeline:()=>p.AutomaticSpeechRecognitionPipeline,BartForConditionalGeneration:()=>h.BartForConditionalGeneration,BartForSequenceClassification:()=>h.BartForSequenceClassification,BartModel:()=>h.BartModel,BartPretrainedModel:()=>h.BartPretrainedModel,BartTokenizer:()=>m.BartTokenizer,BaseModelOutput:()=>h.BaseModelOutput,BaseStreamer:()=>T.BaseStreamer,BeitFeatureExtractor:()=>_.BeitFeatureExtractor,BeitForImageClassification:()=>h.BeitForImageClassification,BeitModel:()=>h.BeitModel,BeitPreTrainedModel:()=>h.BeitPreTrainedModel,BertForMaskedLM:()=>h.BertForMaskedLM,BertForQuestionAnswering:()=>h.BertForQuestionAnswering,BertForSequenceClassification:()=>h.BertForSequenceClassification,BertForTokenClassification:()=>h.BertForTokenClassification,BertModel:()=>h.BertModel,BertPreTrainedModel:()=>h.BertPreTrainedModel,BertTokenizer:()=>m.BertTokenizer,BitImageProcessor:()=>_.BitImageProcessor,BlenderbotForConditionalGeneration:()=>h.BlenderbotForConditionalGeneration,BlenderbotModel:()=>h.BlenderbotModel,BlenderbotPreTrainedModel:()=>h.BlenderbotPreTrainedModel,BlenderbotSmallForConditionalGeneration:()=>h.BlenderbotSmallForConditionalGeneration,BlenderbotSmallModel:()=>h.BlenderbotSmallModel,BlenderbotSmallPreTrainedModel:()=>h.BlenderbotSmallPreTrainedModel,BlenderbotSmallTokenizer:()=>m.BlenderbotSmallTokenizer,BlenderbotTokenizer:()=>m.BlenderbotTokenizer,BloomForCausalLM:()=>h.BloomForCausalLM,BloomModel:()=>h.BloomModel,BloomPreTrainedModel:()=>h.BloomPreTrainedModel,BloomTokenizer:()=>m.BloomTokenizer,CLIPFeatureExtractor:()=>_.CLIPFeatureExtractor,CLIPImageProcessor:()=>_.CLIPImageProcessor,CLIPModel:()=>h.CLIPModel,CLIPPreTrainedModel:()=>h.CLIPPreTrainedModel,CLIPSegForImageSegmentation:()=>h.CLIPSegForImageSegmentation,CLIPSegModel:()=>h.CLIPSegModel,CLIPSegPreTrainedModel:()=>h.CLIPSegPreTrainedModel,CLIPTextModelWithProjection:()=>h.CLIPTextModelWithProjection,CLIPTokenizer:()=>m.CLIPTokenizer,CLIPVisionModelWithProjection:()=>h.CLIPVisionModelWithProjection,CamembertForMaskedLM:()=>h.CamembertForMaskedLM,CamembertForQuestionAnswering:()=>h.CamembertForQuestionAnswering,CamembertForSequenceClassification:()=>h.CamembertForSequenceClassification,CamembertForTokenClassification:()=>h.CamembertForTokenClassification,CamembertModel:()=>h.CamembertModel,CamembertPreTrainedModel:()=>h.CamembertPreTrainedModel,CamembertTokenizer:()=>m.CamembertTokenizer,CausalLMOutput:()=>h.CausalLMOutput,CausalLMOutputWithPast:()=>h.CausalLMOutputWithPast,ChineseCLIPFeatureExtractor:()=>_.ChineseCLIPFeatureExtractor,ChineseCLIPModel:()=>h.ChineseCLIPModel,ChineseCLIPPreTrainedModel:()=>h.ChineseCLIPPreTrainedModel,ClapAudioModelWithProjection:()=>h.ClapAudioModelWithProjection,ClapFeatureExtractor:()=>_.ClapFeatureExtractor,ClapModel:()=>h.ClapModel,ClapPreTrainedModel:()=>h.ClapPreTrainedModel,ClapTextModelWithProjection:()=>h.ClapTextModelWithProjection,CodeGenForCausalLM:()=>h.CodeGenForCausalLM,CodeGenModel:()=>h.CodeGenModel,CodeGenPreTrainedModel:()=>h.CodeGenPreTrainedModel,CodeGenTokenizer:()=>m.CodeGenTokenizer,CodeLlamaTokenizer:()=>m.CodeLlamaTokenizer,CohereForCausalLM:()=>h.CohereForCausalLM,CohereModel:()=>h.CohereModel,CoherePreTrainedModel:()=>h.CoherePreTrainedModel,CohereTokenizer:()=>m.CohereTokenizer,ConvBertForMaskedLM:()=>h.ConvBertForMaskedLM,ConvBertForQuestionAnswering:()=>h.ConvBertForQuestionAnswering,ConvBertForSequenceClassification:()=>h.ConvBertForSequenceClassification,ConvBertForTokenClassification:()=>h.ConvBertForTokenClassification,ConvBertModel:()=>h.ConvBertModel,ConvBertPreTrainedModel:()=>h.ConvBertPreTrainedModel,ConvBertTokenizer:()=>m.ConvBertTokenizer,ConvNextFeatureExtractor:()=>_.ConvNextFeatureExtractor,ConvNextForImageClassification:()=>h.ConvNextForImageClassification,ConvNextImageProcessor:()=>_.ConvNextImageProcessor,ConvNextModel:()=>h.ConvNextModel,ConvNextPreTrainedModel:()=>h.ConvNextPreTrainedModel,ConvNextV2ForImageClassification:()=>h.ConvNextV2ForImageClassification,ConvNextV2Model:()=>h.ConvNextV2Model,ConvNextV2PreTrainedModel:()=>h.ConvNextV2PreTrainedModel,DPTFeatureExtractor:()=>_.DPTFeatureExtractor,DPTForDepthEstimation:()=>h.DPTForDepthEstimation,DPTImageProcessor:()=>_.DPTImageProcessor,DPTModel:()=>h.DPTModel,DPTPreTrainedModel:()=>h.DPTPreTrainedModel,DebertaForMaskedLM:()=>h.DebertaForMaskedLM,DebertaForQuestionAnswering:()=>h.DebertaForQuestionAnswering,DebertaForSequenceClassification:()=>h.DebertaForSequenceClassification,DebertaForTokenClassification:()=>h.DebertaForTokenClassification,DebertaModel:()=>h.DebertaModel,DebertaPreTrainedModel:()=>h.DebertaPreTrainedModel,DebertaTokenizer:()=>m.DebertaTokenizer,DebertaV2ForMaskedLM:()=>h.DebertaV2ForMaskedLM,DebertaV2ForQuestionAnswering:()=>h.DebertaV2ForQuestionAnswering,DebertaV2ForSequenceClassification:()=>h.DebertaV2ForSequenceClassification,DebertaV2ForTokenClassification:()=>h.DebertaV2ForTokenClassification,DebertaV2Model:()=>h.DebertaV2Model,DebertaV2PreTrainedModel:()=>h.DebertaV2PreTrainedModel,DebertaV2Tokenizer:()=>m.DebertaV2Tokenizer,DeiTFeatureExtractor:()=>_.DeiTFeatureExtractor,DeiTForImageClassification:()=>h.DeiTForImageClassification,DeiTModel:()=>h.DeiTModel,DeiTPreTrainedModel:()=>h.DeiTPreTrainedModel,DepthAnythingForDepthEstimation:()=>h.DepthAnythingForDepthEstimation,DepthAnythingPreTrainedModel:()=>h.DepthAnythingPreTrainedModel,DepthEstimationPipeline:()=>p.DepthEstimationPipeline,DetrFeatureExtractor:()=>_.DetrFeatureExtractor,DetrForObjectDetection:()=>h.DetrForObjectDetection,DetrForSegmentation:()=>h.DetrForSegmentation,DetrModel:()=>h.DetrModel,DetrObjectDetectionOutput:()=>h.DetrObjectDetectionOutput,DetrPreTrainedModel:()=>h.DetrPreTrainedModel,DetrSegmentationOutput:()=>h.DetrSegmentationOutput,Dinov2ForImageClassification:()=>h.Dinov2ForImageClassification,Dinov2Model:()=>h.Dinov2Model,Dinov2PreTrainedModel:()=>h.Dinov2PreTrainedModel,DistilBertForMaskedLM:()=>h.DistilBertForMaskedLM,DistilBertForQuestionAnswering:()=>h.DistilBertForQuestionAnswering,DistilBertForSequenceClassification:()=>h.DistilBertForSequenceClassification,DistilBertForTokenClassification:()=>h.DistilBertForTokenClassification,DistilBertModel:()=>h.DistilBertModel,DistilBertPreTrainedModel:()=>h.DistilBertPreTrainedModel,DistilBertTokenizer:()=>m.DistilBertTokenizer,DocumentQuestionAnsweringPipeline:()=>p.DocumentQuestionAnsweringPipeline,DonutFeatureExtractor:()=>_.DonutFeatureExtractor,DonutSwinModel:()=>h.DonutSwinModel,DonutSwinPreTrainedModel:()=>h.DonutSwinPreTrainedModel,EfficientNetForImageClassification:()=>h.EfficientNetForImageClassification,EfficientNetImageProcessor:()=>_.EfficientNetImageProcessor,EfficientNetModel:()=>h.EfficientNetModel,EfficientNetPreTrainedModel:()=>h.EfficientNetPreTrainedModel,ElectraForMaskedLM:()=>h.ElectraForMaskedLM,ElectraForQuestionAnswering:()=>h.ElectraForQuestionAnswering,ElectraForSequenceClassification:()=>h.ElectraForSequenceClassification,ElectraForTokenClassification:()=>h.ElectraForTokenClassification,ElectraModel:()=>h.ElectraModel,ElectraPreTrainedModel:()=>h.ElectraPreTrainedModel,ElectraTokenizer:()=>m.ElectraTokenizer,EosTokenCriteria:()=>x.EosTokenCriteria,EsmForMaskedLM:()=>h.EsmForMaskedLM,EsmForSequenceClassification:()=>h.EsmForSequenceClassification,EsmForTokenClassification:()=>h.EsmForTokenClassification,EsmModel:()=>h.EsmModel,EsmPreTrainedModel:()=>h.EsmPreTrainedModel,EsmTokenizer:()=>m.EsmTokenizer,FFT:()=>b.FFT,FalconForCausalLM:()=>h.FalconForCausalLM,FalconModel:()=>h.FalconModel,FalconPreTrainedModel:()=>h.FalconPreTrainedModel,FalconTokenizer:()=>m.FalconTokenizer,FastViTForImageClassification:()=>h.FastViTForImageClassification,FastViTModel:()=>h.FastViTModel,FastViTPreTrainedModel:()=>h.FastViTPreTrainedModel,FeatureExtractionPipeline:()=>p.FeatureExtractionPipeline,FeatureExtractor:()=>_.FeatureExtractor,FillMaskPipeline:()=>p.FillMaskPipeline,Florence2ForConditionalGeneration:()=>h.Florence2ForConditionalGeneration,Florence2PreTrainedModel:()=>h.Florence2PreTrainedModel,Florence2Processor:()=>_.Florence2Processor,GLPNFeatureExtractor:()=>_.GLPNFeatureExtractor,GLPNForDepthEstimation:()=>h.GLPNForDepthEstimation,GLPNModel:()=>h.GLPNModel,GLPNPreTrainedModel:()=>h.GLPNPreTrainedModel,GPT2LMHeadModel:()=>h.GPT2LMHeadModel,GPT2Model:()=>h.GPT2Model,GPT2PreTrainedModel:()=>h.GPT2PreTrainedModel,GPT2Tokenizer:()=>m.GPT2Tokenizer,GPTBigCodeForCausalLM:()=>h.GPTBigCodeForCausalLM,GPTBigCodeModel:()=>h.GPTBigCodeModel,GPTBigCodePreTrainedModel:()=>h.GPTBigCodePreTrainedModel,GPTJForCausalLM:()=>h.GPTJForCausalLM,GPTJModel:()=>h.GPTJModel,GPTJPreTrainedModel:()=>h.GPTJPreTrainedModel,GPTNeoForCausalLM:()=>h.GPTNeoForCausalLM,GPTNeoModel:()=>h.GPTNeoModel,GPTNeoPreTrainedModel:()=>h.GPTNeoPreTrainedModel,GPTNeoXForCausalLM:()=>h.GPTNeoXForCausalLM,GPTNeoXModel:()=>h.GPTNeoXModel,GPTNeoXPreTrainedModel:()=>h.GPTNeoXPreTrainedModel,GPTNeoXTokenizer:()=>m.GPTNeoXTokenizer,Gemma2ForCausalLM:()=>h.Gemma2ForCausalLM,Gemma2Model:()=>h.Gemma2Model,Gemma2PreTrainedModel:()=>h.Gemma2PreTrainedModel,GemmaForCausalLM:()=>h.GemmaForCausalLM,GemmaModel:()=>h.GemmaModel,GemmaPreTrainedModel:()=>h.GemmaPreTrainedModel,GemmaTokenizer:()=>m.GemmaTokenizer,Grok1Tokenizer:()=>m.Grok1Tokenizer,HerbertTokenizer:()=>m.HerbertTokenizer,HubertForCTC:()=>h.HubertForCTC,HubertForSequenceClassification:()=>h.HubertForSequenceClassification,HubertModel:()=>h.HubertModel,HubertPreTrainedModel:()=>h.HubertPreTrainedModel,ImageClassificationPipeline:()=>p.ImageClassificationPipeline,ImageFeatureExtractionPipeline:()=>p.ImageFeatureExtractionPipeline,ImageFeatureExtractor:()=>_.ImageFeatureExtractor,ImageMattingOutput:()=>h.ImageMattingOutput,ImageSegmentationPipeline:()=>p.ImageSegmentationPipeline,ImageToImagePipeline:()=>p.ImageToImagePipeline,ImageToTextPipeline:()=>p.ImageToTextPipeline,InterruptableStoppingCriteria:()=>x.InterruptableStoppingCriteria,JAISLMHeadModel:()=>h.JAISLMHeadModel,JAISModel:()=>h.JAISModel,JAISPreTrainedModel:()=>h.JAISPreTrainedModel,LlamaForCausalLM:()=>h.LlamaForCausalLM,LlamaModel:()=>h.LlamaModel,LlamaPreTrainedModel:()=>h.LlamaPreTrainedModel,LlamaTokenizer:()=>m.LlamaTokenizer,LlavaForConditionalGeneration:()=>h.LlavaForConditionalGeneration,LlavaPreTrainedModel:()=>h.LlavaPreTrainedModel,LongT5ForConditionalGeneration:()=>h.LongT5ForConditionalGeneration,LongT5Model:()=>h.LongT5Model,LongT5PreTrainedModel:()=>h.LongT5PreTrainedModel,M2M100ForConditionalGeneration:()=>h.M2M100ForConditionalGeneration,M2M100Model:()=>h.M2M100Model,M2M100PreTrainedModel:()=>h.M2M100PreTrainedModel,M2M100Tokenizer:()=>m.M2M100Tokenizer,MBart50Tokenizer:()=>m.MBart50Tokenizer,MBartForCausalLM:()=>h.MBartForCausalLM,MBartForConditionalGeneration:()=>h.MBartForConditionalGeneration,MBartForSequenceClassification:()=>h.MBartForSequenceClassification,MBartModel:()=>h.MBartModel,MBartPreTrainedModel:()=>h.MBartPreTrainedModel,MBartTokenizer:()=>m.MBartTokenizer,MPNetForMaskedLM:()=>h.MPNetForMaskedLM,MPNetForQuestionAnswering:()=>h.MPNetForQuestionAnswering,MPNetForSequenceClassification:()=>h.MPNetForSequenceClassification,MPNetForTokenClassification:()=>h.MPNetForTokenClassification,MPNetModel:()=>h.MPNetModel,MPNetPreTrainedModel:()=>h.MPNetPreTrainedModel,MPNetTokenizer:()=>m.MPNetTokenizer,MT5ForConditionalGeneration:()=>h.MT5ForConditionalGeneration,MT5Model:()=>h.MT5Model,MT5PreTrainedModel:()=>h.MT5PreTrainedModel,MarianMTModel:()=>h.MarianMTModel,MarianModel:()=>h.MarianModel,MarianPreTrainedModel:()=>h.MarianPreTrainedModel,MarianTokenizer:()=>m.MarianTokenizer,MaskedLMOutput:()=>h.MaskedLMOutput,MaxLengthCriteria:()=>x.MaxLengthCriteria,MistralForCausalLM:()=>h.MistralForCausalLM,MistralModel:()=>h.MistralModel,MistralPreTrainedModel:()=>h.MistralPreTrainedModel,MobileBertForMaskedLM:()=>h.MobileBertForMaskedLM,MobileBertForQuestionAnswering:()=>h.MobileBertForQuestionAnswering,MobileBertForSequenceClassification:()=>h.MobileBertForSequenceClassification,MobileBertModel:()=>h.MobileBertModel,MobileBertPreTrainedModel:()=>h.MobileBertPreTrainedModel,MobileBertTokenizer:()=>m.MobileBertTokenizer,MobileNetV1FeatureExtractor:()=>_.MobileNetV1FeatureExtractor,MobileNetV1ForImageClassification:()=>h.MobileNetV1ForImageClassification,MobileNetV1Model:()=>h.MobileNetV1Model,MobileNetV1PreTrainedModel:()=>h.MobileNetV1PreTrainedModel,MobileNetV2FeatureExtractor:()=>_.MobileNetV2FeatureExtractor,MobileNetV2ForImageClassification:()=>h.MobileNetV2ForImageClassification,MobileNetV2Model:()=>h.MobileNetV2Model,MobileNetV2PreTrainedModel:()=>h.MobileNetV2PreTrainedModel,MobileNetV3FeatureExtractor:()=>_.MobileNetV3FeatureExtractor,MobileNetV3ForImageClassification:()=>h.MobileNetV3ForImageClassification,MobileNetV3Model:()=>h.MobileNetV3Model,MobileNetV3PreTrainedModel:()=>h.MobileNetV3PreTrainedModel,MobileNetV4FeatureExtractor:()=>_.MobileNetV4FeatureExtractor,MobileNetV4ForImageClassification:()=>h.MobileNetV4ForImageClassification,MobileNetV4Model:()=>h.MobileNetV4Model,MobileNetV4PreTrainedModel:()=>h.MobileNetV4PreTrainedModel,MobileViTFeatureExtractor:()=>_.MobileViTFeatureExtractor,MobileViTForImageClassification:()=>h.MobileViTForImageClassification,MobileViTImageProcessor:()=>_.MobileViTImageProcessor,MobileViTModel:()=>h.MobileViTModel,MobileViTPreTrainedModel:()=>h.MobileViTPreTrainedModel,MobileViTV2ForImageClassification:()=>h.MobileViTV2ForImageClassification,MobileViTV2Model:()=>h.MobileViTV2Model,MobileViTV2PreTrainedModel:()=>h.MobileViTV2PreTrainedModel,ModelOutput:()=>h.ModelOutput,Moondream1ForConditionalGeneration:()=>h.Moondream1ForConditionalGeneration,MptForCausalLM:()=>h.MptForCausalLM,MptModel:()=>h.MptModel,MptPreTrainedModel:()=>h.MptPreTrainedModel,MusicgenForCausalLM:()=>h.MusicgenForCausalLM,MusicgenForConditionalGeneration:()=>h.MusicgenForConditionalGeneration,MusicgenModel:()=>h.MusicgenModel,MusicgenPreTrainedModel:()=>h.MusicgenPreTrainedModel,NllbTokenizer:()=>m.NllbTokenizer,NomicBertModel:()=>h.NomicBertModel,NomicBertPreTrainedModel:()=>h.NomicBertPreTrainedModel,NougatImageProcessor:()=>_.NougatImageProcessor,NougatTokenizer:()=>m.NougatTokenizer,OPTForCausalLM:()=>h.OPTForCausalLM,OPTModel:()=>h.OPTModel,OPTPreTrainedModel:()=>h.OPTPreTrainedModel,ObjectDetectionPipeline:()=>p.ObjectDetectionPipeline,OpenELMForCausalLM:()=>h.OpenELMForCausalLM,OpenELMModel:()=>h.OpenELMModel,OpenELMPreTrainedModel:()=>h.OpenELMPreTrainedModel,OwlViTFeatureExtractor:()=>_.OwlViTFeatureExtractor,OwlViTForObjectDetection:()=>h.OwlViTForObjectDetection,OwlViTModel:()=>h.OwlViTModel,OwlViTPreTrainedModel:()=>h.OwlViTPreTrainedModel,OwlViTProcessor:()=>_.OwlViTProcessor,Owlv2ForObjectDetection:()=>h.Owlv2ForObjectDetection,Owlv2ImageProcessor:()=>_.Owlv2ImageProcessor,Owlv2Model:()=>h.Owlv2Model,Owlv2PreTrainedModel:()=>h.Owlv2PreTrainedModel,Phi3ForCausalLM:()=>h.Phi3ForCausalLM,Phi3Model:()=>h.Phi3Model,Phi3PreTrainedModel:()=>h.Phi3PreTrainedModel,PhiForCausalLM:()=>h.PhiForCausalLM,PhiModel:()=>h.PhiModel,PhiPreTrainedModel:()=>h.PhiPreTrainedModel,Pipeline:()=>p.Pipeline,PreTrainedModel:()=>h.PreTrainedModel,PreTrainedTokenizer:()=>m.PreTrainedTokenizer,PretrainedConfig:()=>f.PretrainedConfig,PretrainedMixin:()=>h.PretrainedMixin,Processor:()=>_.Processor,PyAnnoteFeatureExtractor:()=>_.PyAnnoteFeatureExtractor,PyAnnoteForAudioFrameClassification:()=>h.PyAnnoteForAudioFrameClassification,PyAnnoteModel:()=>h.PyAnnoteModel,PyAnnotePreTrainedModel:()=>h.PyAnnotePreTrainedModel,PyAnnoteProcessor:()=>_.PyAnnoteProcessor,QuestionAnsweringModelOutput:()=>h.QuestionAnsweringModelOutput,QuestionAnsweringPipeline:()=>p.QuestionAnsweringPipeline,Qwen2ForCausalLM:()=>h.Qwen2ForCausalLM,Qwen2Model:()=>h.Qwen2Model,Qwen2PreTrainedModel:()=>h.Qwen2PreTrainedModel,Qwen2Tokenizer:()=>m.Qwen2Tokenizer,RTDetrForObjectDetection:()=>h.RTDetrForObjectDetection,RTDetrImageProcessor:()=>_.RTDetrImageProcessor,RTDetrModel:()=>h.RTDetrModel,RTDetrObjectDetectionOutput:()=>h.RTDetrObjectDetectionOutput,RTDetrPreTrainedModel:()=>h.RTDetrPreTrainedModel,RawImage:()=>M.RawImage,ResNetForImageClassification:()=>h.ResNetForImageClassification,ResNetModel:()=>h.ResNetModel,ResNetPreTrainedModel:()=>h.ResNetPreTrainedModel,RoFormerForMaskedLM:()=>h.RoFormerForMaskedLM,RoFormerForQuestionAnswering:()=>h.RoFormerForQuestionAnswering,RoFormerForSequenceClassification:()=>h.RoFormerForSequenceClassification,RoFormerForTokenClassification:()=>h.RoFormerForTokenClassification,RoFormerModel:()=>h.RoFormerModel,RoFormerPreTrainedModel:()=>h.RoFormerPreTrainedModel,RoFormerTokenizer:()=>m.RoFormerTokenizer,RobertaForMaskedLM:()=>h.RobertaForMaskedLM,RobertaForQuestionAnswering:()=>h.RobertaForQuestionAnswering,RobertaForSequenceClassification:()=>h.RobertaForSequenceClassification,RobertaForTokenClassification:()=>h.RobertaForTokenClassification,RobertaModel:()=>h.RobertaModel,RobertaPreTrainedModel:()=>h.RobertaPreTrainedModel,RobertaTokenizer:()=>m.RobertaTokenizer,SamImageProcessor:()=>_.SamImageProcessor,SamImageSegmentationOutput:()=>h.SamImageSegmentationOutput,SamModel:()=>h.SamModel,SamPreTrainedModel:()=>h.SamPreTrainedModel,SamProcessor:()=>_.SamProcessor,SapiensFeatureExtractor:()=>_.SapiensFeatureExtractor,SapiensForDepthEstimation:()=>h.SapiensForDepthEstimation,SapiensForNormalEstimation:()=>h.SapiensForNormalEstimation,SapiensForSemanticSegmentation:()=>h.SapiensForSemanticSegmentation,SapiensPreTrainedModel:()=>h.SapiensPreTrainedModel,SeamlessM4TFeatureExtractor:()=>_.SeamlessM4TFeatureExtractor,SegformerFeatureExtractor:()=>_.SegformerFeatureExtractor,SegformerForImageClassification:()=>h.SegformerForImageClassification,SegformerForSemanticSegmentation:()=>h.SegformerForSemanticSegmentation,SegformerModel:()=>h.SegformerModel,SegformerPreTrainedModel:()=>h.SegformerPreTrainedModel,Seq2SeqLMOutput:()=>h.Seq2SeqLMOutput,SequenceClassifierOutput:()=>h.SequenceClassifierOutput,SiglipImageProcessor:()=>_.SiglipImageProcessor,SiglipModel:()=>h.SiglipModel,SiglipPreTrainedModel:()=>h.SiglipPreTrainedModel,SiglipTextModel:()=>h.SiglipTextModel,SiglipTokenizer:()=>m.SiglipTokenizer,SiglipVisionModel:()=>h.SiglipVisionModel,SpeechT5FeatureExtractor:()=>_.SpeechT5FeatureExtractor,SpeechT5ForSpeechToText:()=>h.SpeechT5ForSpeechToText,SpeechT5ForTextToSpeech:()=>h.SpeechT5ForTextToSpeech,SpeechT5HifiGan:()=>h.SpeechT5HifiGan,SpeechT5Model:()=>h.SpeechT5Model,SpeechT5PreTrainedModel:()=>h.SpeechT5PreTrainedModel,SpeechT5Processor:()=>_.SpeechT5Processor,SpeechT5Tokenizer:()=>m.SpeechT5Tokenizer,SqueezeBertForMaskedLM:()=>h.SqueezeBertForMaskedLM,SqueezeBertForQuestionAnswering:()=>h.SqueezeBertForQuestionAnswering,SqueezeBertForSequenceClassification:()=>h.SqueezeBertForSequenceClassification,SqueezeBertModel:()=>h.SqueezeBertModel,SqueezeBertPreTrainedModel:()=>h.SqueezeBertPreTrainedModel,SqueezeBertTokenizer:()=>m.SqueezeBertTokenizer,StableLmForCausalLM:()=>h.StableLmForCausalLM,StableLmModel:()=>h.StableLmModel,StableLmPreTrainedModel:()=>h.StableLmPreTrainedModel,Starcoder2ForCausalLM:()=>h.Starcoder2ForCausalLM,Starcoder2Model:()=>h.Starcoder2Model,Starcoder2PreTrainedModel:()=>h.Starcoder2PreTrainedModel,StoppingCriteria:()=>x.StoppingCriteria,StoppingCriteriaList:()=>x.StoppingCriteriaList,SummarizationPipeline:()=>p.SummarizationPipeline,Swin2SRForImageSuperResolution:()=>h.Swin2SRForImageSuperResolution,Swin2SRImageProcessor:()=>_.Swin2SRImageProcessor,Swin2SRModel:()=>h.Swin2SRModel,Swin2SRPreTrainedModel:()=>h.Swin2SRPreTrainedModel,SwinForImageClassification:()=>h.SwinForImageClassification,SwinModel:()=>h.SwinModel,SwinPreTrainedModel:()=>h.SwinPreTrainedModel,T5ForConditionalGeneration:()=>h.T5ForConditionalGeneration,T5Model:()=>h.T5Model,T5PreTrainedModel:()=>h.T5PreTrainedModel,T5Tokenizer:()=>m.T5Tokenizer,TableTransformerForObjectDetection:()=>h.TableTransformerForObjectDetection,TableTransformerModel:()=>h.TableTransformerModel,TableTransformerObjectDetectionOutput:()=>h.TableTransformerObjectDetectionOutput,TableTransformerPreTrainedModel:()=>h.TableTransformerPreTrainedModel,Tensor:()=>w.Tensor,Text2TextGenerationPipeline:()=>p.Text2TextGenerationPipeline,TextClassificationPipeline:()=>p.TextClassificationPipeline,TextGenerationPipeline:()=>p.TextGenerationPipeline,TextStreamer:()=>T.TextStreamer,TextToAudioPipeline:()=>p.TextToAudioPipeline,TokenClassificationPipeline:()=>p.TokenClassificationPipeline,TokenClassifierOutput:()=>h.TokenClassifierOutput,TokenizerModel:()=>m.TokenizerModel,TrOCRForCausalLM:()=>h.TrOCRForCausalLM,TrOCRPreTrainedModel:()=>h.TrOCRPreTrainedModel,TranslationPipeline:()=>p.TranslationPipeline,UniSpeechForCTC:()=>h.UniSpeechForCTC,UniSpeechForSequenceClassification:()=>h.UniSpeechForSequenceClassification,UniSpeechModel:()=>h.UniSpeechModel,UniSpeechPreTrainedModel:()=>h.UniSpeechPreTrainedModel,UniSpeechSatForAudioFrameClassification:()=>h.UniSpeechSatForAudioFrameClassification,UniSpeechSatForCTC:()=>h.UniSpeechSatForCTC,UniSpeechSatForSequenceClassification:()=>h.UniSpeechSatForSequenceClassification,UniSpeechSatModel:()=>h.UniSpeechSatModel,UniSpeechSatPreTrainedModel:()=>h.UniSpeechSatPreTrainedModel,ViTFeatureExtractor:()=>_.ViTFeatureExtractor,ViTForImageClassification:()=>h.ViTForImageClassification,ViTImageProcessor:()=>_.ViTImageProcessor,ViTModel:()=>h.ViTModel,ViTPreTrainedModel:()=>h.ViTPreTrainedModel,VisionEncoderDecoderModel:()=>h.VisionEncoderDecoderModel,VitMatteForImageMatting:()=>h.VitMatteForImageMatting,VitMatteImageProcessor:()=>_.VitMatteImageProcessor,VitMattePreTrainedModel:()=>h.VitMattePreTrainedModel,VitsModel:()=>h.VitsModel,VitsModelOutput:()=>h.VitsModelOutput,VitsPreTrainedModel:()=>h.VitsPreTrainedModel,VitsTokenizer:()=>m.VitsTokenizer,Wav2Vec2BertForCTC:()=>h.Wav2Vec2BertForCTC,Wav2Vec2BertForSequenceClassification:()=>h.Wav2Vec2BertForSequenceClassification,Wav2Vec2BertModel:()=>h.Wav2Vec2BertModel,Wav2Vec2BertPreTrainedModel:()=>h.Wav2Vec2BertPreTrainedModel,Wav2Vec2CTCTokenizer:()=>m.Wav2Vec2CTCTokenizer,Wav2Vec2FeatureExtractor:()=>_.Wav2Vec2FeatureExtractor,Wav2Vec2ForAudioFrameClassification:()=>h.Wav2Vec2ForAudioFrameClassification,Wav2Vec2ForCTC:()=>h.Wav2Vec2ForCTC,Wav2Vec2ForSequenceClassification:()=>h.Wav2Vec2ForSequenceClassification,Wav2Vec2Model:()=>h.Wav2Vec2Model,Wav2Vec2PreTrainedModel:()=>h.Wav2Vec2PreTrainedModel,Wav2Vec2ProcessorWithLM:()=>_.Wav2Vec2ProcessorWithLM,WavLMForAudioFrameClassification:()=>h.WavLMForAudioFrameClassification,WavLMForCTC:()=>h.WavLMForCTC,WavLMForSequenceClassification:()=>h.WavLMForSequenceClassification,WavLMForXVector:()=>h.WavLMForXVector,WavLMModel:()=>h.WavLMModel,WavLMPreTrainedModel:()=>h.WavLMPreTrainedModel,WeSpeakerFeatureExtractor:()=>_.WeSpeakerFeatureExtractor,WeSpeakerResNetModel:()=>h.WeSpeakerResNetModel,WeSpeakerResNetPreTrainedModel:()=>h.WeSpeakerResNetPreTrainedModel,WhisperFeatureExtractor:()=>_.WhisperFeatureExtractor,WhisperForConditionalGeneration:()=>h.WhisperForConditionalGeneration,WhisperModel:()=>h.WhisperModel,WhisperPreTrainedModel:()=>h.WhisperPreTrainedModel,WhisperProcessor:()=>_.WhisperProcessor,WhisperTextStreamer:()=>T.WhisperTextStreamer,WhisperTokenizer:()=>m.WhisperTokenizer,XLMForQuestionAnswering:()=>h.XLMForQuestionAnswering,XLMForSequenceClassification:()=>h.XLMForSequenceClassification,XLMForTokenClassification:()=>h.XLMForTokenClassification,XLMModel:()=>h.XLMModel,XLMPreTrainedModel:()=>h.XLMPreTrainedModel,XLMRobertaForMaskedLM:()=>h.XLMRobertaForMaskedLM,XLMRobertaForQuestionAnswering:()=>h.XLMRobertaForQuestionAnswering,XLMRobertaForSequenceClassification:()=>h.XLMRobertaForSequenceClassification,XLMRobertaForTokenClassification:()=>h.XLMRobertaForTokenClassification,XLMRobertaModel:()=>h.XLMRobertaModel,XLMRobertaPreTrainedModel:()=>h.XLMRobertaPreTrainedModel,XLMRobertaTokenizer:()=>m.XLMRobertaTokenizer,XLMTokenizer:()=>m.XLMTokenizer,XLMWithLMHeadModel:()=>h.XLMWithLMHeadModel,XVectorOutput:()=>h.XVectorOutput,YolosFeatureExtractor:()=>_.YolosFeatureExtractor,YolosForObjectDetection:()=>h.YolosForObjectDetection,YolosModel:()=>h.YolosModel,YolosObjectDetectionOutput:()=>h.YolosObjectDetectionOutput,YolosPreTrainedModel:()=>h.YolosPreTrainedModel,ZeroShotAudioClassificationPipeline:()=>p.ZeroShotAudioClassificationPipeline,ZeroShotClassificationPipeline:()=>p.ZeroShotClassificationPipeline,ZeroShotImageClassificationPipeline:()=>p.ZeroShotImageClassificationPipeline,ZeroShotObjectDetectionPipeline:()=>p.ZeroShotObjectDetectionPipeline,bankers_round:()=>b.bankers_round,cat:()=>w.cat,cos_sim:()=>b.cos_sim,dot:()=>b.dot,dynamic_time_warping:()=>b.dynamic_time_warping,env:()=>u.env,full:()=>w.full,full_like:()=>w.full_like,getKeyValueShapes:()=>f.getKeyValueShapes,hamming:()=>g.hamming,hanning:()=>g.hanning,interpolate:()=>w.interpolate,interpolate_4d:()=>w.interpolate_4d,interpolate_data:()=>b.interpolate_data,is_chinese_char:()=>m.is_chinese_char,layer_norm:()=>w.layer_norm,log_softmax:()=>b.log_softmax,magnitude:()=>b.magnitude,matmul:()=>w.matmul,max:()=>b.max,mean:()=>w.mean,mean_pooling:()=>w.mean_pooling,medianFilter:()=>b.medianFilter,mel_filter_bank:()=>g.mel_filter_bank,min:()=>b.min,ones:()=>w.ones,ones_like:()=>w.ones_like,permute:()=>w.permute,permute_data:()=>b.permute_data,pipeline:()=>p.pipeline,quantize_embeddings:()=>w.quantize_embeddings,read_audio:()=>g.read_audio,rfft:()=>w.rfft,round:()=>b.round,softmax:()=>b.softmax,spectrogram:()=>g.spectrogram,stack:()=>w.stack,std_mean:()=>w.std_mean,topk:()=>w.topk,window_function:()=>g.window_function,zeros:()=>w.zeros,zeros_like:()=>w.zeros_like});var u=c(/*! ./env.js */"./src/env.js"),p=c(/*! ./pipelines.js */"./src/pipelines.js"),h=c(/*! ./models.js */"./src/models.js"),m=c(/*! ./tokenizers.js */"./src/tokenizers.js"),_=c(/*! ./processors.js */"./src/processors.js"),f=c(/*! ./configs.js */"./src/configs.js"),g=c(/*! ./utils/audio.js */"./src/utils/audio.js"),M=c(/*! ./utils/image.js */"./src/utils/image.js"),w=c(/*! ./utils/tensor.js */"./src/utils/tensor.js"),b=c(/*! ./utils/maths.js */"./src/utils/maths.js"),T=c(/*! ./generation/streamers.js */"./src/generation/streamers.js"),x=c(/*! ./generation/stopping_criteria.js */"./src/generation/stopping_criteria.js"),y=d.ASTFeatureExtractor,k=d.ASTForAudioClassification,F=d.ASTModel,C=d.ASTPreTrainedModel,P=d.AlbertForMaskedLM,v=d.AlbertForQuestionAnswering,S=d.AlbertForSequenceClassification,A=d.AlbertModel,E=d.AlbertPreTrainedModel,L=d.AlbertTokenizer,z=d.AudioClassificationPipeline,I=d.AutoConfig,B=d.AutoModel,N=d.AutoModelForAudioClassification,O=d.AutoModelForAudioFrameClassification,D=d.AutoModelForCTC,V=d.AutoModelForCausalLM,j=d.AutoModelForDepthEstimation,R=d.AutoModelForDocumentQuestionAnswering,G=d.AutoModelForImageClassification,q=d.AutoModelForImageFeatureExtraction,$=d.AutoModelForImageMatting,W=d.AutoModelForImageSegmentation,U=d.AutoModelForImageToImage,X=d.AutoModelForMaskGeneration,Q=d.AutoModelForMaskedLM,H=d.AutoModelForNormalEstimation,Y=d.AutoModelForObjectDetection,J=d.AutoModelForQuestionAnswering,K=d.AutoModelForSemanticSegmentation,Z=d.AutoModelForSeq2SeqLM,ee=d.AutoModelForSequenceClassification,te=d.AutoModelForSpeechSeq2Seq,ne=d.AutoModelForTextToSpectrogram,re=d.AutoModelForTextToWaveform,oe=d.AutoModelForTokenClassification,se=d.AutoModelForVision2Seq,ae=d.AutoModelForXVector,ie=d.AutoModelForZeroShotObjectDetection,le=d.AutoProcessor,ce=d.AutoTokenizer,de=d.AutomaticSpeechRecognitionPipeline,ue=d.BartForConditionalGeneration,pe=d.BartForSequenceClassification,he=d.BartModel,me=d.BartPretrainedModel,_e=d.BartTokenizer,fe=d.BaseModelOutput,ge=d.BaseStreamer,Me=d.BeitFeatureExtractor,we=d.BeitForImageClassification,be=d.BeitModel,Te=d.BeitPreTrainedModel,xe=d.BertForMaskedLM,ye=d.BertForQuestionAnswering,ke=d.BertForSequenceClassification,Fe=d.BertForTokenClassification,Ce=d.BertModel,Pe=d.BertPreTrainedModel,ve=d.BertTokenizer,Se=d.BitImageProcessor,Ae=d.BlenderbotForConditionalGeneration,Ee=d.BlenderbotModel,Le=d.BlenderbotPreTrainedModel,ze=d.BlenderbotSmallForConditionalGeneration,Ie=d.BlenderbotSmallModel,Be=d.BlenderbotSmallPreTrainedModel,Ne=d.BlenderbotSmallTokenizer,Oe=d.BlenderbotTokenizer,De=d.BloomForCausalLM,Ve=d.BloomModel,je=d.BloomPreTrainedModel,Re=d.BloomTokenizer,Ge=d.CLIPFeatureExtractor,qe=d.CLIPImageProcessor,$e=d.CLIPModel,We=d.CLIPPreTrainedModel,Ue=d.CLIPSegForImageSegmentation,Xe=d.CLIPSegModel,Qe=d.CLIPSegPreTrainedModel,He=d.CLIPTextModelWithProjection,Ye=d.CLIPTokenizer,Je=d.CLIPVisionModelWithProjection,Ke=d.CamembertForMaskedLM,Ze=d.CamembertForQuestionAnswering,et=d.CamembertForSequenceClassification,tt=d.CamembertForTokenClassification,nt=d.CamembertModel,rt=d.CamembertPreTrainedModel,ot=d.CamembertTokenizer,st=d.CausalLMOutput,at=d.CausalLMOutputWithPast,it=d.ChineseCLIPFeatureExtractor,lt=d.ChineseCLIPModel,ct=d.ChineseCLIPPreTrainedModel,dt=d.ClapAudioModelWithProjection,ut=d.ClapFeatureExtractor,pt=d.ClapModel,ht=d.ClapPreTrainedModel,mt=d.ClapTextModelWithProjection,_t=d.CodeGenForCausalLM,ft=d.CodeGenModel,gt=d.CodeGenPreTrainedModel,Mt=d.CodeGenTokenizer,wt=d.CodeLlamaTokenizer,bt=d.CohereForCausalLM,Tt=d.CohereModel,xt=d.CoherePreTrainedModel,yt=d.CohereTokenizer,kt=d.ConvBertForMaskedLM,Ft=d.ConvBertForQuestionAnswering,Ct=d.ConvBertForSequenceClassification,Pt=d.ConvBertForTokenClassification,vt=d.ConvBertModel,St=d.ConvBertPreTrainedModel,At=d.ConvBertTokenizer,Et=d.ConvNextFeatureExtractor,Lt=d.ConvNextForImageClassification,zt=d.ConvNextImageProcessor,It=d.ConvNextModel,Bt=d.ConvNextPreTrainedModel,Nt=d.ConvNextV2ForImageClassification,Ot=d.ConvNextV2Model,Dt=d.ConvNextV2PreTrainedModel,Vt=d.DPTFeatureExtractor,jt=d.DPTForDepthEstimation,Rt=d.DPTImageProcessor,Gt=d.DPTModel,qt=d.DPTPreTrainedModel,$t=d.DebertaForMaskedLM,Wt=d.DebertaForQuestionAnswering,Ut=d.DebertaForSequenceClassification,Xt=d.DebertaForTokenClassification,Qt=d.DebertaModel,Ht=d.DebertaPreTrainedModel,Yt=d.DebertaTokenizer,Jt=d.DebertaV2ForMaskedLM,Kt=d.DebertaV2ForQuestionAnswering,Zt=d.DebertaV2ForSequenceClassification,en=d.DebertaV2ForTokenClassification,tn=d.DebertaV2Model,nn=d.DebertaV2PreTrainedModel,rn=d.DebertaV2Tokenizer,on=d.DeiTFeatureExtractor,sn=d.DeiTForImageClassification,an=d.DeiTModel,ln=d.DeiTPreTrainedModel,cn=d.DepthAnythingForDepthEstimation,dn=d.DepthAnythingPreTrainedModel,un=d.DepthEstimationPipeline,pn=d.DetrFeatureExtractor,hn=d.DetrForObjectDetection,mn=d.DetrForSegmentation,_n=d.DetrModel,fn=d.DetrObjectDetectionOutput,gn=d.DetrPreTrainedModel,Mn=d.DetrSegmentationOutput,wn=d.Dinov2ForImageClassification,bn=d.Dinov2Model,Tn=d.Dinov2PreTrainedModel,xn=d.DistilBertForMaskedLM,yn=d.DistilBertForQuestionAnswering,kn=d.DistilBertForSequenceClassification,Fn=d.DistilBertForTokenClassification,Cn=d.DistilBertModel,Pn=d.DistilBertPreTrainedModel,vn=d.DistilBertTokenizer,Sn=d.DocumentQuestionAnsweringPipeline,An=d.DonutFeatureExtractor,En=d.DonutSwinModel,Ln=d.DonutSwinPreTrainedModel,zn=d.EfficientNetForImageClassification,In=d.EfficientNetImageProcessor,Bn=d.EfficientNetModel,Nn=d.EfficientNetPreTrainedModel,On=d.ElectraForMaskedLM,Dn=d.ElectraForQuestionAnswering,Vn=d.ElectraForSequenceClassification,jn=d.ElectraForTokenClassification,Rn=d.ElectraModel,Gn=d.ElectraPreTrainedModel,qn=d.ElectraTokenizer,$n=d.EosTokenCriteria,Wn=d.EsmForMaskedLM,Un=d.EsmForSequenceClassification,Xn=d.EsmForTokenClassification,Qn=d.EsmModel,Hn=d.EsmPreTrainedModel,Yn=d.EsmTokenizer,Jn=d.FFT,Kn=d.FalconForCausalLM,Zn=d.FalconModel,er=d.FalconPreTrainedModel,tr=d.FalconTokenizer,nr=d.FastViTForImageClassification,rr=d.FastViTModel,or=d.FastViTPreTrainedModel,sr=d.FeatureExtractionPipeline,ar=d.FeatureExtractor,ir=d.FillMaskPipeline,lr=d.Florence2ForConditionalGeneration,cr=d.Florence2PreTrainedModel,dr=d.Florence2Processor,ur=d.GLPNFeatureExtractor,pr=d.GLPNForDepthEstimation,hr=d.GLPNModel,mr=d.GLPNPreTrainedModel,_r=d.GPT2LMHeadModel,fr=d.GPT2Model,gr=d.GPT2PreTrainedModel,Mr=d.GPT2Tokenizer,wr=d.GPTBigCodeForCausalLM,br=d.GPTBigCodeModel,Tr=d.GPTBigCodePreTrainedModel,xr=d.GPTJForCausalLM,yr=d.GPTJModel,kr=d.GPTJPreTrainedModel,Fr=d.GPTNeoForCausalLM,Cr=d.GPTNeoModel,Pr=d.GPTNeoPreTrainedModel,vr=d.GPTNeoXForCausalLM,Sr=d.GPTNeoXModel,Ar=d.GPTNeoXPreTrainedModel,Er=d.GPTNeoXTokenizer,Lr=d.Gemma2ForCausalLM,zr=d.Gemma2Model,Ir=d.Gemma2PreTrainedModel,Br=d.GemmaForCausalLM,Nr=d.GemmaModel,Or=d.GemmaPreTrainedModel,Dr=d.GemmaTokenizer,Vr=d.Grok1Tokenizer,jr=d.HerbertTokenizer,Rr=d.HubertForCTC,Gr=d.HubertForSequenceClassification,qr=d.HubertModel,$r=d.HubertPreTrainedModel,Wr=d.ImageClassificationPipeline,Ur=d.ImageFeatureExtractionPipeline,Xr=d.ImageFeatureExtractor,Qr=d.ImageMattingOutput,Hr=d.ImageSegmentationPipeline,Yr=d.ImageToImagePipeline,Jr=d.ImageToTextPipeline,Kr=d.InterruptableStoppingCriteria,Zr=d.JAISLMHeadModel,eo=d.JAISModel,to=d.JAISPreTrainedModel,no=d.LlamaForCausalLM,ro=d.LlamaModel,oo=d.LlamaPreTrainedModel,so=d.LlamaTokenizer,ao=d.LlavaForConditionalGeneration,io=d.LlavaPreTrainedModel,lo=d.LongT5ForConditionalGeneration,co=d.LongT5Model,uo=d.LongT5PreTrainedModel,po=d.M2M100ForConditionalGeneration,ho=d.M2M100Model,mo=d.M2M100PreTrainedModel,_o=d.M2M100Tokenizer,fo=d.MBart50Tokenizer,go=d.MBartForCausalLM,Mo=d.MBartForConditionalGeneration,wo=d.MBartForSequenceClassification,bo=d.MBartModel,To=d.MBartPreTrainedModel,xo=d.MBartTokenizer,yo=d.MPNetForMaskedLM,ko=d.MPNetForQuestionAnswering,Fo=d.MPNetForSequenceClassification,Co=d.MPNetForTokenClassification,Po=d.MPNetModel,vo=d.MPNetPreTrainedModel,So=d.MPNetTokenizer,Ao=d.MT5ForConditionalGeneration,Eo=d.MT5Model,Lo=d.MT5PreTrainedModel,zo=d.MarianMTModel,Io=d.MarianModel,Bo=d.MarianPreTrainedModel,No=d.MarianTokenizer,Oo=d.MaskedLMOutput,Do=d.MaxLengthCriteria,Vo=d.MistralForCausalLM,jo=d.MistralModel,Ro=d.MistralPreTrainedModel,Go=d.MobileBertForMaskedLM,qo=d.MobileBertForQuestionAnswering,$o=d.MobileBertForSequenceClassification,Wo=d.MobileBertModel,Uo=d.MobileBertPreTrainedModel,Xo=d.MobileBertTokenizer,Qo=d.MobileNetV1FeatureExtractor,Ho=d.MobileNetV1ForImageClassification,Yo=d.MobileNetV1Model,Jo=d.MobileNetV1PreTrainedModel,Ko=d.MobileNetV2FeatureExtractor,Zo=d.MobileNetV2ForImageClassification,es=d.MobileNetV2Model,ts=d.MobileNetV2PreTrainedModel,ns=d.MobileNetV3FeatureExtractor,rs=d.MobileNetV3ForImageClassification,os=d.MobileNetV3Model,ss=d.MobileNetV3PreTrainedModel,as=d.MobileNetV4FeatureExtractor,is=d.MobileNetV4ForImageClassification,ls=d.MobileNetV4Model,cs=d.MobileNetV4PreTrainedModel,ds=d.MobileViTFeatureExtractor,us=d.MobileViTForImageClassification,ps=d.MobileViTImageProcessor,hs=d.MobileViTModel,ms=d.MobileViTPreTrainedModel,_s=d.MobileViTV2ForImageClassification,fs=d.MobileViTV2Model,gs=d.MobileViTV2PreTrainedModel,Ms=d.ModelOutput,ws=d.Moondream1ForConditionalGeneration,bs=d.MptForCausalLM,Ts=d.MptModel,xs=d.MptPreTrainedModel,ys=d.MusicgenForCausalLM,ks=d.MusicgenForConditionalGeneration,Fs=d.MusicgenModel,Cs=d.MusicgenPreTrainedModel,Ps=d.NllbTokenizer,vs=d.NomicBertModel,Ss=d.NomicBertPreTrainedModel,As=d.NougatImageProcessor,Es=d.NougatTokenizer,Ls=d.OPTForCausalLM,zs=d.OPTModel,Is=d.OPTPreTrainedModel,Bs=d.ObjectDetectionPipeline,Ns=d.OpenELMForCausalLM,Os=d.OpenELMModel,Ds=d.OpenELMPreTrainedModel,Vs=d.OwlViTFeatureExtractor,js=d.OwlViTForObjectDetection,Rs=d.OwlViTModel,Gs=d.OwlViTPreTrainedModel,qs=d.OwlViTProcessor,$s=d.Owlv2ForObjectDetection,Ws=d.Owlv2ImageProcessor,Us=d.Owlv2Model,Xs=d.Owlv2PreTrainedModel,Qs=d.Phi3ForCausalLM,Hs=d.Phi3Model,Ys=d.Phi3PreTrainedModel,Js=d.PhiForCausalLM,Ks=d.PhiModel,Zs=d.PhiPreTrainedModel,ea=d.Pipeline,ta=d.PreTrainedModel,na=d.PreTrainedTokenizer,ra=d.PretrainedConfig,oa=d.PretrainedMixin,sa=d.Processor,aa=d.PyAnnoteFeatureExtractor,ia=d.PyAnnoteForAudioFrameClassification,la=d.PyAnnoteModel,ca=d.PyAnnotePreTrainedModel,da=d.PyAnnoteProcessor,ua=d.QuestionAnsweringModelOutput,pa=d.QuestionAnsweringPipeline,ha=d.Qwen2ForCausalLM,ma=d.Qwen2Model,_a=d.Qwen2PreTrainedModel,fa=d.Qwen2Tokenizer,ga=d.RTDetrForObjectDetection,Ma=d.RTDetrImageProcessor,wa=d.RTDetrModel,ba=d.RTDetrObjectDetectionOutput,Ta=d.RTDetrPreTrainedModel,xa=d.RawImage,ya=d.ResNetForImageClassification,ka=d.ResNetModel,Fa=d.ResNetPreTrainedModel,Ca=d.RoFormerForMaskedLM,Pa=d.RoFormerForQuestionAnswering,va=d.RoFormerForSequenceClassification,Sa=d.RoFormerForTokenClassification,Aa=d.RoFormerModel,Ea=d.RoFormerPreTrainedModel,La=d.RoFormerTokenizer,za=d.RobertaForMaskedLM,Ia=d.RobertaForQuestionAnswering,Ba=d.RobertaForSequenceClassification,Na=d.RobertaForTokenClassification,Oa=d.RobertaModel,Da=d.RobertaPreTrainedModel,Va=d.RobertaTokenizer,ja=d.SamImageProcessor,Ra=d.SamImageSegmentationOutput,Ga=d.SamModel,qa=d.SamPreTrainedModel,$a=d.SamProcessor,Wa=d.SapiensFeatureExtractor,Ua=d.SapiensForDepthEstimation,Xa=d.SapiensForNormalEstimation,Qa=d.SapiensForSemanticSegmentation,Ha=d.SapiensPreTrainedModel,Ya=d.SeamlessM4TFeatureExtractor,Ja=d.SegformerFeatureExtractor,Ka=d.SegformerForImageClassification,Za=d.SegformerForSemanticSegmentation,ei=d.SegformerModel,ti=d.SegformerPreTrainedModel,ni=d.Seq2SeqLMOutput,ri=d.SequenceClassifierOutput,oi=d.SiglipImageProcessor,si=d.SiglipModel,ai=d.SiglipPreTrainedModel,ii=d.SiglipTextModel,li=d.SiglipTokenizer,ci=d.SiglipVisionModel,di=d.SpeechT5FeatureExtractor,ui=d.SpeechT5ForSpeechToText,pi=d.SpeechT5ForTextToSpeech,hi=d.SpeechT5HifiGan,mi=d.SpeechT5Model,_i=d.SpeechT5PreTrainedModel,fi=d.SpeechT5Processor,gi=d.SpeechT5Tokenizer,Mi=d.SqueezeBertForMaskedLM,wi=d.SqueezeBertForQuestionAnswering,bi=d.SqueezeBertForSequenceClassification,Ti=d.SqueezeBertModel,xi=d.SqueezeBertPreTrainedModel,yi=d.SqueezeBertTokenizer,ki=d.StableLmForCausalLM,Fi=d.StableLmModel,Ci=d.StableLmPreTrainedModel,Pi=d.Starcoder2ForCausalLM,vi=d.Starcoder2Model,Si=d.Starcoder2PreTrainedModel,Ai=d.StoppingCriteria,Ei=d.StoppingCriteriaList,Li=d.SummarizationPipeline,zi=d.Swin2SRForImageSuperResolution,Ii=d.Swin2SRImageProcessor,Bi=d.Swin2SRModel,Ni=d.Swin2SRPreTrainedModel,Oi=d.SwinForImageClassification,Di=d.SwinModel,Vi=d.SwinPreTrainedModel,ji=d.T5ForConditionalGeneration,Ri=d.T5Model,Gi=d.T5PreTrainedModel,qi=d.T5Tokenizer,$i=d.TableTransformerForObjectDetection,Wi=d.TableTransformerModel,Ui=d.TableTransformerObjectDetectionOutput,Xi=d.TableTransformerPreTrainedModel,Qi=d.Tensor,Hi=d.Text2TextGenerationPipeline,Yi=d.TextClassificationPipeline,Ji=d.TextGenerationPipeline,Ki=d.TextStreamer,Zi=d.TextToAudioPipeline,el=d.TokenClassificationPipeline,tl=d.TokenClassifierOutput,nl=d.TokenizerModel,rl=d.TrOCRForCausalLM,ol=d.TrOCRPreTrainedModel,sl=d.TranslationPipeline,al=d.UniSpeechForCTC,il=d.UniSpeechForSequenceClassification,ll=d.UniSpeechModel,cl=d.UniSpeechPreTrainedModel,dl=d.UniSpeechSatForAudioFrameClassification,ul=d.UniSpeechSatForCTC,pl=d.UniSpeechSatForSequenceClassification,hl=d.UniSpeechSatModel,ml=d.UniSpeechSatPreTrainedModel,_l=d.ViTFeatureExtractor,fl=d.ViTForImageClassification,gl=d.ViTImageProcessor,Ml=d.ViTModel,wl=d.ViTPreTrainedModel,bl=d.VisionEncoderDecoderModel,Tl=d.VitMatteForImageMatting,xl=d.VitMatteImageProcessor,yl=d.VitMattePreTrainedModel,kl=d.VitsModel,Fl=d.VitsModelOutput,Cl=d.VitsPreTrainedModel,Pl=d.VitsTokenizer,vl=d.Wav2Vec2BertForCTC,Sl=d.Wav2Vec2BertForSequenceClassification,Al=d.Wav2Vec2BertModel,El=d.Wav2Vec2BertPreTrainedModel,Ll=d.Wav2Vec2CTCTokenizer,zl=d.Wav2Vec2FeatureExtractor,Il=d.Wav2Vec2ForAudioFrameClassification,Bl=d.Wav2Vec2ForCTC,Nl=d.Wav2Vec2ForSequenceClassification,Ol=d.Wav2Vec2Model,Dl=d.Wav2Vec2PreTrainedModel,Vl=d.Wav2Vec2ProcessorWithLM,jl=d.WavLMForAudioFrameClassification,Rl=d.WavLMForCTC,Gl=d.WavLMForSequenceClassification,ql=d.WavLMForXVector,$l=d.WavLMModel,Wl=d.WavLMPreTrainedModel,Ul=d.WeSpeakerFeatureExtractor,Xl=d.WeSpeakerResNetModel,Ql=d.WeSpeakerResNetPreTrainedModel,Hl=d.WhisperFeatureExtractor,Yl=d.WhisperForConditionalGeneration,Jl=d.WhisperModel,Kl=d.WhisperPreTrainedModel,Zl=d.WhisperProcessor,ec=d.WhisperTextStreamer,tc=d.WhisperTokenizer,nc=d.XLMForQuestionAnswering,rc=d.XLMForSequenceClassification,oc=d.XLMForTokenClassification,sc=d.XLMModel,ac=d.XLMPreTrainedModel,ic=d.XLMRobertaForMaskedLM,lc=d.XLMRobertaForQuestionAnswering,cc=d.XLMRobertaForSequenceClassification,dc=d.XLMRobertaForTokenClassification,uc=d.XLMRobertaModel,pc=d.XLMRobertaPreTrainedModel,hc=d.XLMRobertaTokenizer,mc=d.XLMTokenizer,_c=d.XLMWithLMHeadModel,fc=d.XVectorOutput,gc=d.YolosFeatureExtractor,Mc=d.YolosForObjectDetection,wc=d.YolosModel,bc=d.YolosObjectDetectionOutput,Tc=d.YolosPreTrainedModel,xc=d.ZeroShotAudioClassificationPipeline,yc=d.ZeroShotClassificationPipeline,kc=d.ZeroShotImageClassificationPipeline,Fc=d.ZeroShotObjectDetectionPipeline,Cc=d.bankers_round,Pc=d.cat,vc=d.cos_sim,Sc=d.dot,Ac=d.dynamic_time_warping,Ec=d.env,Lc=d.full,zc=d.full_like,Ic=d.getKeyValueShapes,Bc=d.hamming,Nc=d.hanning,Oc=d.interpolate,Dc=d.interpolate_4d,Vc=d.interpolate_data,jc=d.is_chinese_char,Rc=d.layer_norm,Gc=d.log_softmax,qc=d.magnitude,$c=d.matmul,Wc=d.max,Uc=d.mean,Xc=d.mean_pooling,Qc=d.medianFilter,Hc=d.mel_filter_bank,Yc=d.min,Jc=d.ones,Kc=d.ones_like,Zc=d.permute,ed=d.permute_data,td=d.pipeline,nd=d.quantize_embeddings,rd=d.read_audio,od=d.rfft,sd=d.round,ad=d.softmax,id=d.spectrogram,ld=d.stack,cd=d.std_mean,dd=d.topk,ud=d.window_function,pd=d.zeros,hd=d.zeros_like;export{y as ASTFeatureExtractor,k as ASTForAudioClassification,F as ASTModel,C as ASTPreTrainedModel,P as AlbertForMaskedLM,v as AlbertForQuestionAnswering,S as AlbertForSequenceClassification,A as AlbertModel,E as AlbertPreTrainedModel,L as AlbertTokenizer,z as AudioClassificationPipeline,I as AutoConfig,B as AutoModel,N as AutoModelForAudioClassification,O as AutoModelForAudioFrameClassification,D as AutoModelForCTC,V as AutoModelForCausalLM,j as AutoModelForDepthEstimation,R as AutoModelForDocumentQuestionAnswering,G as AutoModelForImageClassification,q as AutoModelForImageFeatureExtraction,$ as AutoModelForImageMatting,W as AutoModelForImageSegmentation,U as AutoModelForImageToImage,X as AutoModelForMaskGeneration,Q as AutoModelForMaskedLM,H as AutoModelForNormalEstimation,Y as AutoModelForObjectDetection,J as AutoModelForQuestionAnswering,K as AutoModelForSemanticSegmentation,Z as AutoModelForSeq2SeqLM,ee as AutoModelForSequenceClassification,te as AutoModelForSpeechSeq2Seq,ne as AutoModelForTextToSpectrogram,re as AutoModelForTextToWaveform,oe as AutoModelForTokenClassification,se as AutoModelForVision2Seq,ae as AutoModelForXVector,ie as AutoModelForZeroShotObjectDetection,le as AutoProcessor,ce as AutoTokenizer,de as AutomaticSpeechRecognitionPipeline,ue as BartForConditionalGeneration,pe as BartForSequenceClassification,he as BartModel,me as BartPretrainedModel,_e as BartTokenizer,fe as BaseModelOutput,ge as BaseStreamer,Me as BeitFeatureExtractor,we as BeitForImageClassification,be as BeitModel,Te as BeitPreTrainedModel,xe as BertForMaskedLM,ye as BertForQuestionAnswering,ke as BertForSequenceClassification,Fe as BertForTokenClassification,Ce as BertModel,Pe as BertPreTrainedModel,ve as BertTokenizer,Se as BitImageProcessor,Ae as BlenderbotForConditionalGeneration,Ee as BlenderbotModel,Le as BlenderbotPreTrainedModel,ze as BlenderbotSmallForConditionalGeneration,Ie as BlenderbotSmallModel,Be as BlenderbotSmallPreTrainedModel,Ne as BlenderbotSmallTokenizer,Oe as BlenderbotTokenizer,De as BloomForCausalLM,Ve as BloomModel,je as BloomPreTrainedModel,Re as BloomTokenizer,Ge as CLIPFeatureExtractor,qe as CLIPImageProcessor,$e as CLIPModel,We as CLIPPreTrainedModel,Ue as CLIPSegForImageSegmentation,Xe as CLIPSegModel,Qe as CLIPSegPreTrainedModel,He as CLIPTextModelWithProjection,Ye as CLIPTokenizer,Je as CLIPVisionModelWithProjection,Ke as CamembertForMaskedLM,Ze as CamembertForQuestionAnswering,et as CamembertForSequenceClassification,tt as CamembertForTokenClassification,nt as CamembertModel,rt as CamembertPreTrainedModel,ot as CamembertTokenizer,st as CausalLMOutput,at as CausalLMOutputWithPast,it as ChineseCLIPFeatureExtractor,lt as ChineseCLIPModel,ct as ChineseCLIPPreTrainedModel,dt as ClapAudioModelWithProjection,ut as ClapFeatureExtractor,pt as ClapModel,ht as ClapPreTrainedModel,mt as ClapTextModelWithProjection,_t as CodeGenForCausalLM,ft as CodeGenModel,gt as CodeGenPreTrainedModel,Mt as CodeGenTokenizer,wt as CodeLlamaTokenizer,bt as CohereForCausalLM,Tt as CohereModel,xt as CoherePreTrainedModel,yt as CohereTokenizer,kt as ConvBertForMaskedLM,Ft as ConvBertForQuestionAnswering,Ct as ConvBertForSequenceClassification,Pt as ConvBertForTokenClassification,vt as ConvBertModel,St as ConvBertPreTrainedModel,At as ConvBertTokenizer,Et as ConvNextFeatureExtractor,Lt as ConvNextForImageClassification,zt as ConvNextImageProcessor,It as ConvNextModel,Bt as ConvNextPreTrainedModel,Nt as ConvNextV2ForImageClassification,Ot as ConvNextV2Model,Dt as ConvNextV2PreTrainedModel,Vt as DPTFeatureExtractor,jt as DPTForDepthEstimation,Rt as DPTImageProcessor,Gt as DPTModel,qt as DPTPreTrainedModel,$t as DebertaForMaskedLM,Wt as DebertaForQuestionAnswering,Ut as DebertaForSequenceClassification,Xt as DebertaForTokenClassification,Qt as DebertaModel,Ht as DebertaPreTrainedModel,Yt as DebertaTokenizer,Jt as DebertaV2ForMaskedLM,Kt as DebertaV2ForQuestionAnswering,Zt as DebertaV2ForSequenceClassification,en as DebertaV2ForTokenClassification,tn as DebertaV2Model,nn as DebertaV2PreTrainedModel,rn as DebertaV2Tokenizer,on as DeiTFeatureExtractor,sn as DeiTForImageClassification,an as DeiTModel,ln as DeiTPreTrainedModel,cn as DepthAnythingForDepthEstimation,dn as DepthAnythingPreTrainedModel,un as DepthEstimationPipeline,pn as DetrFeatureExtractor,hn as DetrForObjectDetection,mn as DetrForSegmentation,_n as DetrModel,fn as DetrObjectDetectionOutput,gn as DetrPreTrainedModel,Mn as DetrSegmentationOutput,wn as Dinov2ForImageClassification,bn as Dinov2Model,Tn as Dinov2PreTrainedModel,xn as DistilBertForMaskedLM,yn as DistilBertForQuestionAnswering,kn as DistilBertForSequenceClassification,Fn as DistilBertForTokenClassification,Cn as DistilBertModel,Pn as DistilBertPreTrainedModel,vn as DistilBertTokenizer,Sn as DocumentQuestionAnsweringPipeline,An as DonutFeatureExtractor,En as DonutSwinModel,Ln as DonutSwinPreTrainedModel,zn as EfficientNetForImageClassification,In as EfficientNetImageProcessor,Bn as EfficientNetModel,Nn as EfficientNetPreTrainedModel,On as ElectraForMaskedLM,Dn as ElectraForQuestionAnswering,Vn as ElectraForSequenceClassification,jn as ElectraForTokenClassification,Rn as ElectraModel,Gn as ElectraPreTrainedModel,qn as ElectraTokenizer,$n as EosTokenCriteria,Wn as EsmForMaskedLM,Un as EsmForSequenceClassification,Xn as EsmForTokenClassification,Qn as EsmModel,Hn as EsmPreTrainedModel,Yn as EsmTokenizer,Jn as FFT,Kn as FalconForCausalLM,Zn as FalconModel,er as FalconPreTrainedModel,tr as FalconTokenizer,nr as FastViTForImageClassification,rr as FastViTModel,or as FastViTPreTrainedModel,sr as FeatureExtractionPipeline,ar as FeatureExtractor,ir as FillMaskPipeline,lr as Florence2ForConditionalGeneration,cr as Florence2PreTrainedModel,dr as Florence2Processor,ur as GLPNFeatureExtractor,pr as GLPNForDepthEstimation,hr as GLPNModel,mr as GLPNPreTrainedModel,_r as GPT2LMHeadModel,fr as GPT2Model,gr as GPT2PreTrainedModel,Mr as GPT2Tokenizer,wr as GPTBigCodeForCausalLM,br as GPTBigCodeModel,Tr as GPTBigCodePreTrainedModel,xr as GPTJForCausalLM,yr as GPTJModel,kr as GPTJPreTrainedModel,Fr as GPTNeoForCausalLM,Cr as GPTNeoModel,Pr as GPTNeoPreTrainedModel,vr as GPTNeoXForCausalLM,Sr as GPTNeoXModel,Ar as GPTNeoXPreTrainedModel,Er as GPTNeoXTokenizer,Lr as Gemma2ForCausalLM,zr as Gemma2Model,Ir as Gemma2PreTrainedModel,Br as GemmaForCausalLM,Nr as GemmaModel,Or as GemmaPreTrainedModel,Dr as GemmaTokenizer,Vr as Grok1Tokenizer,jr as HerbertTokenizer,Rr as HubertForCTC,Gr as HubertForSequenceClassification,qr as HubertModel,$r as HubertPreTrainedModel,Wr as ImageClassificationPipeline,Ur as ImageFeatureExtractionPipeline,Xr as ImageFeatureExtractor,Qr as ImageMattingOutput,Hr as ImageSegmentationPipeline,Yr as ImageToImagePipeline,Jr as ImageToTextPipeline,Kr as InterruptableStoppingCriteria,Zr as JAISLMHeadModel,eo as JAISModel,to as JAISPreTrainedModel,no as LlamaForCausalLM,ro as LlamaModel,oo as LlamaPreTrainedModel,so as LlamaTokenizer,ao as LlavaForConditionalGeneration,io as LlavaPreTrainedModel,lo as LongT5ForConditionalGeneration,co as LongT5Model,uo as LongT5PreTrainedModel,po as M2M100ForConditionalGeneration,ho as M2M100Model,mo as M2M100PreTrainedModel,_o as M2M100Tokenizer,fo as MBart50Tokenizer,go as MBartForCausalLM,Mo as MBartForConditionalGeneration,wo as MBartForSequenceClassification,bo as MBartModel,To as MBartPreTrainedModel,xo as MBartTokenizer,yo as MPNetForMaskedLM,ko as MPNetForQuestionAnswering,Fo as MPNetForSequenceClassification,Co as MPNetForTokenClassification,Po as MPNetModel,vo as MPNetPreTrainedModel,So as MPNetTokenizer,Ao as MT5ForConditionalGeneration,Eo as MT5Model,Lo as MT5PreTrainedModel,zo as MarianMTModel,Io as MarianModel,Bo as MarianPreTrainedModel,No as MarianTokenizer,Oo as MaskedLMOutput,Do as MaxLengthCriteria,Vo as MistralForCausalLM,jo as MistralModel,Ro as MistralPreTrainedModel,Go as MobileBertForMaskedLM,qo as MobileBertForQuestionAnswering,$o as MobileBertForSequenceClassification,Wo as MobileBertModel,Uo as MobileBertPreTrainedModel,Xo as MobileBertTokenizer,Qo as MobileNetV1FeatureExtractor,Ho as MobileNetV1ForImageClassification,Yo as MobileNetV1Model,Jo as MobileNetV1PreTrainedModel,Ko as MobileNetV2FeatureExtractor,Zo as MobileNetV2ForImageClassification,es as MobileNetV2Model,ts as MobileNetV2PreTrainedModel,ns as MobileNetV3FeatureExtractor,rs as MobileNetV3ForImageClassification,os as MobileNetV3Model,ss as MobileNetV3PreTrainedModel,as as MobileNetV4FeatureExtractor,is as MobileNetV4ForImageClassification,ls as MobileNetV4Model,cs as MobileNetV4PreTrainedModel,ds as MobileViTFeatureExtractor,us as MobileViTForImageClassification,ps as MobileViTImageProcessor,hs as MobileViTModel,ms as MobileViTPreTrainedModel,_s as MobileViTV2ForImageClassification,fs as MobileViTV2Model,gs as MobileViTV2PreTrainedModel,Ms as ModelOutput,ws as Moondream1ForConditionalGeneration,bs as MptForCausalLM,Ts as MptModel,xs as MptPreTrainedModel,ys as MusicgenForCausalLM,ks as MusicgenForConditionalGeneration,Fs as MusicgenModel,Cs as MusicgenPreTrainedModel,Ps as NllbTokenizer,vs as NomicBertModel,Ss as NomicBertPreTrainedModel,As as NougatImageProcessor,Es as NougatTokenizer,Ls as OPTForCausalLM,zs as OPTModel,Is as OPTPreTrainedModel,Bs as ObjectDetectionPipeline,Ns as OpenELMForCausalLM,Os as OpenELMModel,Ds as OpenELMPreTrainedModel,Vs as OwlViTFeatureExtractor,js as OwlViTForObjectDetection,Rs as OwlViTModel,Gs as OwlViTPreTrainedModel,qs as OwlViTProcessor,$s as Owlv2ForObjectDetection,Ws as Owlv2ImageProcessor,Us as Owlv2Model,Xs as Owlv2PreTrainedModel,Qs as Phi3ForCausalLM,Hs as Phi3Model,Ys as Phi3PreTrainedModel,Js as PhiForCausalLM,Ks as PhiModel,Zs as PhiPreTrainedModel,ea as Pipeline,ta as PreTrainedModel,na as PreTrainedTokenizer,ra as PretrainedConfig,oa as PretrainedMixin,sa as Processor,aa as PyAnnoteFeatureExtractor,ia as PyAnnoteForAudioFrameClassification,la as PyAnnoteModel,ca as PyAnnotePreTrainedModel,da as PyAnnoteProcessor,ua as QuestionAnsweringModelOutput,pa as QuestionAnsweringPipeline,ha as Qwen2ForCausalLM,ma as Qwen2Model,_a as Qwen2PreTrainedModel,fa as Qwen2Tokenizer,ga as RTDetrForObjectDetection,Ma as RTDetrImageProcessor,wa as RTDetrModel,ba as RTDetrObjectDetectionOutput,Ta as RTDetrPreTrainedModel,xa as RawImage,ya as ResNetForImageClassification,ka as ResNetModel,Fa as ResNetPreTrainedModel,Ca as RoFormerForMaskedLM,Pa as RoFormerForQuestionAnswering,va as RoFormerForSequenceClassification,Sa as RoFormerForTokenClassification,Aa as RoFormerModel,Ea as RoFormerPreTrainedModel,La as RoFormerTokenizer,za as RobertaForMaskedLM,Ia as RobertaForQuestionAnswering,Ba as RobertaForSequenceClassification,Na as RobertaForTokenClassification,Oa as RobertaModel,Da as RobertaPreTrainedModel,Va as RobertaTokenizer,ja as SamImageProcessor,Ra as SamImageSegmentationOutput,Ga as SamModel,qa as SamPreTrainedModel,$a as SamProcessor,Wa as SapiensFeatureExtractor,Ua as SapiensForDepthEstimation,Xa as SapiensForNormalEstimation,Qa as SapiensForSemanticSegmentation,Ha as SapiensPreTrainedModel,Ya as SeamlessM4TFeatureExtractor,Ja as SegformerFeatureExtractor,Ka as SegformerForImageClassification,Za as SegformerForSemanticSegmentation,ei as SegformerModel,ti as SegformerPreTrainedModel,ni as Seq2SeqLMOutput,ri as SequenceClassifierOutput,oi as SiglipImageProcessor,si as SiglipModel,ai as SiglipPreTrainedModel,ii as SiglipTextModel,li as SiglipTokenizer,ci as SiglipVisionModel,di as SpeechT5FeatureExtractor,ui as SpeechT5ForSpeechToText,pi as SpeechT5ForTextToSpeech,hi as SpeechT5HifiGan,mi as SpeechT5Model,_i as SpeechT5PreTrainedModel,fi as SpeechT5Processor,gi as SpeechT5Tokenizer,Mi as SqueezeBertForMaskedLM,wi as SqueezeBertForQuestionAnswering,bi as SqueezeBertForSequenceClassification,Ti as SqueezeBertModel,xi as SqueezeBertPreTrainedModel,yi as SqueezeBertTokenizer,ki as StableLmForCausalLM,Fi as StableLmModel,Ci as StableLmPreTrainedModel,Pi as Starcoder2ForCausalLM,vi as Starcoder2Model,Si as Starcoder2PreTrainedModel,Ai as StoppingCriteria,Ei as StoppingCriteriaList,Li as SummarizationPipeline,zi as Swin2SRForImageSuperResolution,Ii as Swin2SRImageProcessor,Bi as Swin2SRModel,Ni as Swin2SRPreTrainedModel,Oi as SwinForImageClassification,Di as SwinModel,Vi as SwinPreTrainedModel,ji as T5ForConditionalGeneration,Ri as T5Model,Gi as T5PreTrainedModel,qi as T5Tokenizer,$i as TableTransformerForObjectDetection,Wi as TableTransformerModel,Ui as TableTransformerObjectDetectionOutput,Xi as TableTransformerPreTrainedModel,Qi as Tensor,Hi as Text2TextGenerationPipeline,Yi as TextClassificationPipeline,Ji as TextGenerationPipeline,Ki as TextStreamer,Zi as TextToAudioPipeline,el as TokenClassificationPipeline,tl as TokenClassifierOutput,nl as TokenizerModel,rl as TrOCRForCausalLM,ol as TrOCRPreTrainedModel,sl as TranslationPipeline,al as UniSpeechForCTC,il as UniSpeechForSequenceClassification,ll as UniSpeechModel,cl as UniSpeechPreTrainedModel,dl as UniSpeechSatForAudioFrameClassification,ul as UniSpeechSatForCTC,pl as UniSpeechSatForSequenceClassification,hl as UniSpeechSatModel,ml as UniSpeechSatPreTrainedModel,_l as ViTFeatureExtractor,fl as ViTForImageClassification,gl as ViTImageProcessor,Ml as ViTModel,wl as ViTPreTrainedModel,bl as VisionEncoderDecoderModel,Tl as VitMatteForImageMatting,xl as VitMatteImageProcessor,yl as VitMattePreTrainedModel,kl as VitsModel,Fl as VitsModelOutput,Cl as VitsPreTrainedModel,Pl as VitsTokenizer,vl as Wav2Vec2BertForCTC,Sl as Wav2Vec2BertForSequenceClassification,Al as Wav2Vec2BertModel,El as Wav2Vec2BertPreTrainedModel,Ll as Wav2Vec2CTCTokenizer,zl as Wav2Vec2FeatureExtractor,Il as Wav2Vec2ForAudioFrameClassification,Bl as Wav2Vec2ForCTC,Nl as Wav2Vec2ForSequenceClassification,Ol as Wav2Vec2Model,Dl as Wav2Vec2PreTrainedModel,Vl as Wav2Vec2ProcessorWithLM,jl as WavLMForAudioFrameClassification,Rl as WavLMForCTC,Gl as WavLMForSequenceClassification,ql as WavLMForXVector,$l as WavLMModel,Wl as WavLMPreTrainedModel,Ul as WeSpeakerFeatureExtractor,Xl as WeSpeakerResNetModel,Ql as WeSpeakerResNetPreTrainedModel,Hl as WhisperFeatureExtractor,Yl as WhisperForConditionalGeneration,Jl as WhisperModel,Kl as WhisperPreTrainedModel,Zl as WhisperProcessor,ec as WhisperTextStreamer,tc as WhisperTokenizer,nc as XLMForQuestionAnswering,rc as XLMForSequenceClassification,oc as XLMForTokenClassification,sc as XLMModel,ac as XLMPreTrainedModel,ic as XLMRobertaForMaskedLM,lc as XLMRobertaForQuestionAnswering,cc as XLMRobertaForSequenceClassification,dc as XLMRobertaForTokenClassification,uc as XLMRobertaModel,pc as XLMRobertaPreTrainedModel,hc as XLMRobertaTokenizer,mc as XLMTokenizer,_c as XLMWithLMHeadModel,fc as XVectorOutput,gc as YolosFeatureExtractor,Mc as YolosForObjectDetection,wc as YolosModel,bc as YolosObjectDetectionOutput,Tc as YolosPreTrainedModel,xc as ZeroShotAudioClassificationPipeline,yc as ZeroShotClassificationPipeline,kc as ZeroShotImageClassificationPipeline,Fc as ZeroShotObjectDetectionPipeline,Cc as bankers_round,Pc as cat,vc as cos_sim,Sc as dot,Ac as dynamic_time_warping,Ec as env,Lc as full,zc as full_like,Ic as getKeyValueShapes,Bc as hamming,Nc as hanning,Oc as interpolate,Dc as interpolate_4d,Vc as interpolate_data,jc as is_chinese_char,Rc as layer_norm,Gc as log_softmax,qc as magnitude,$c as matmul,Wc as max,Uc as mean,Xc as mean_pooling,Qc as medianFilter,Hc as mel_filter_bank,Yc as min,Jc as ones,Kc as ones_like,Zc as permute,ed as permute_data,td as pipeline,nd as quantize_embeddings,rd as read_audio,od as rfft,sd as round,ad as softmax,id as spectrogram,ld as stack,cd as std_mean,dd as topk,ud as window_function,pd as zeros,hd as zeros_like};
 //# sourceMappingURL=transformers.min.mjs.map