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

package/dist/transformers.min.js

@@ -154,7 +154,7 @@ var r,i,a,s,o,l,u,d,c,p,h,m,f,g,_,w,y,b,v,x,M,T,k,$,S,C,E,P,A,F,z,I,O,B=Object.d
   \************************/(e,t,n)=>{n.r(t),n.d(t,{AutoConfig:()=>l,PretrainedConfig:()=>o,getKeyValueShapes:()=>s});var r=n(/*! ./utils/core.js */"./src/utils/core.js"),i=n(/*! ./utils/hub.js */"./src/utils/hub.js");function a(e){const t={};let n={};switch(e.model_type){case"llava":case"paligemma":case"florence2":n=a(e.text_config);break;case"moondream1":n=a(e.phi_config);break;case"musicgen":n=a(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 i=a(e.decoder),s="num_decoder_layers"in i,o=(0,r.pick)(e,["model_type","is_encoder_decoder"]);return s?(o.num_decoder_layers=i.num_decoder_layers,o.num_decoder_heads=i.num_decoder_heads,o.decoder_hidden_size=i.decoder_hidden_size,o.num_encoder_layers=i.num_encoder_layers,o.num_encoder_heads=i.num_encoder_heads,o.encoder_hidden_size=i.encoder_hidden_size):(o.num_layers=i.num_layers,o.num_heads=i.num_heads,o.hidden_size=i.hidden_size),o}const i={...n,...(0,r.pick)(e,["model_type","multi_query","is_encoder_decoder"])};for(const n in t)i[n]=e[t[n]];return i}function s(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,i=r.decoder_dim_kv??r.decoder_hidden_size/r.num_decoder_heads,a=[1,r.num_encoder_heads,0,e],s=[1,r.num_decoder_heads,0,i];for(let e=0;e<r.num_decoder_layers;++e)n[`${t}.${e}.encoder.key`]=a,n[`${t}.${e}.encoder.value`]=a,n[`${t}.${e}.decoder.key`]=s,n[`${t}.${e}.decoder.value`]=s}else{const e=r.num_heads,i=r.num_layers,a=r.dim_kv??r.hidden_size/(r.num_attention_heads??e);if("falcon"===r.model_type){const r=[1*e,0,a];for(let e=0;e<i;++e)n[`${t}.${e}.key`]=r,n[`${t}.${e}.value`]=r}else if(r.multi_query){const r=[1*e,0,2*a];for(let e=0;e<i;++e)n[`${t}.${e}.key_value`]=r}else if("bloom"===r.model_type){const r=[1*e,a,0],s=[1*e,0,a];for(let e=0;e<i;++e)n[`${t}.${e}.key`]=r,n[`${t}.${e}.value`]=s}else if("openelm"===r.model_type)for(let r=0;r<i;++r){const i=[1,e[r],0,a];n[`${t}.${r}.key`]=i,n[`${t}.${r}.value`]=i}else{const r=[1,e,0,a];for(let e=0;e<i;++e)n[`${t}.${e}.key`]=r,n[`${t}.${e}.value`]=r}}return n}class o{max_position_embeddings;constructor(e){this.model_type=null,this.is_encoder_decoder=!1,Object.assign(this,e),this.normalized_config=a(this)}static async from_pretrained(e,{progress_callback:t=null,config:n=null,cache_dir:r=null,local_files_only:a=!1,revision:s="main"}={}){!n||n instanceof o||(n=new o(n));const l=n??await async function(e,t){return await(0,i.getModelJSON)(e,"config.json",!0,t)}(e,{progress_callback:t,config:n,cache_dir:r,local_files_only:a,revision:s});return new this(l)}}class l{static async from_pretrained(...e){return o.from_pretrained(...e)}}},"./src/env.js":
 /*!********************!*\
   !*** ./src/env.js ***!
-  \********************/(e,t,n)=>{n.r(t),n.d(t,{apis:()=>f,env:()=>b});var r=n(/*! fs */"?569f"),i=n(/*! path */"?3f59"),a=n(/*! url */"?154a");const s="undefined"!=typeof self,o=s&&"DedicatedWorkerGlobalScope"===self.constructor.name,l=s&&"caches"in self,u="undefined"!=typeof navigator&&"gpu"in navigator,d="undefined"!=typeof navigator&&"ml"in navigator,c="undefined"!=typeof process,p=c&&"node"===process?.release?.name,h=!v(r),m=!v(i),f=Object.freeze({IS_BROWSER_ENV:s,IS_WEBWORKER_ENV:o,IS_WEB_CACHE_AVAILABLE:l,IS_WEBGPU_AVAILABLE:u,IS_WEBNN_AVAILABLE:d,IS_PROCESS_AVAILABLE:c,IS_NODE_ENV:p,IS_FS_AVAILABLE:h,IS_PATH_AVAILABLE:m}),g=h&&m,_=g?i.dirname(i.dirname(a.fileURLToPath(import.meta.url))):"./",w=g?i.join(_,"/.cache/"):null,y="/models/",b={version:"3.0.0-alpha.12",backends:{onnx:{}},allowRemoteModels:!0,remoteHost:"https://huggingface.co/",remotePathTemplate:"{model}/resolve/{revision}/",allowLocalModels:!s,localModelPath:g?i.join(_,y):y,useFS:h,useBrowserCache:l,useFSCache:h,cacheDir:w,useCustomCache:!1,customCache:null};function v(e){return 0===Object.keys(e).length}},"./src/generation/configuration_utils.js":
+  \********************/(e,t,n)=>{n.r(t),n.d(t,{apis:()=>f,env:()=>b});var r=n(/*! fs */"?569f"),i=n(/*! path */"?3f59"),a=n(/*! url */"?154a");const s="undefined"!=typeof self,o=s&&"DedicatedWorkerGlobalScope"===self.constructor.name,l=s&&"caches"in self,u="undefined"!=typeof navigator&&"gpu"in navigator,d="undefined"!=typeof navigator&&"ml"in navigator,c="undefined"!=typeof process,p=c&&"node"===process?.release?.name,h=!v(r),m=!v(i),f=Object.freeze({IS_BROWSER_ENV:s,IS_WEBWORKER_ENV:o,IS_WEB_CACHE_AVAILABLE:l,IS_WEBGPU_AVAILABLE:u,IS_WEBNN_AVAILABLE:d,IS_PROCESS_AVAILABLE:c,IS_NODE_ENV:p,IS_FS_AVAILABLE:h,IS_PATH_AVAILABLE:m}),g=h&&m,_=g?i.dirname(i.dirname(a.fileURLToPath(import.meta.url))):"./",w=g?i.join(_,"/.cache/"):null,y="/models/",b={version:"3.0.0-alpha.13",backends:{onnx:{}},allowRemoteModels:!0,remoteHost:"https://huggingface.co/",remotePathTemplate:"{model}/resolve/{revision}/",allowLocalModels:!s,localModelPath:g?i.join(_,y):y,useFS:h,useBrowserCache:l,useFSCache:h,cacheDir:w,useCustomCache:!1,customCache:null};function v(e){return 0===Object.keys(e).length}},"./src/generation/configuration_utils.js":
 /*!***********************************************!*\
   !*** ./src/generation/configuration_utils.js ***!
   \***********************************************/(e,t,n)=>{n.r(t),n.d(t,{GenerationConfig:()=>i});var r=n(/*! ../utils/core.js */"./src/utils/core.js");class i{max_length=20;max_new_tokens=null;min_length=0;min_new_tokens=null;early_stopping=!1;max_time=null;do_sample=!1;num_beams=1;num_beam_groups=1;penalty_alpha=null;use_cache=!0;temperature=1;top_k=50;top_p=1;typical_p=1;epsilon_cutoff=0;eta_cutoff=0;diversity_penalty=0;repetition_penalty=1;encoder_repetition_penalty=1;length_penalty=1;no_repeat_ngram_size=0;bad_words_ids=null;force_words_ids=null;renormalize_logits=!1;constraints=null;forced_bos_token_id=null;forced_eos_token_id=null;remove_invalid_values=!1;exponential_decay_length_penalty=null;suppress_tokens=null;begin_suppress_tokens=null;forced_decoder_ids=null;guidance_scale=null;num_return_sequences=1;output_attentions=!1;output_hidden_states=!1;output_scores=!1;return_dict_in_generate=!1;pad_token_id=null;bos_token_id=null;eos_token_id=null;encoder_no_repeat_ngram_size=0;decoder_start_token_id=null;generation_kwargs={};constructor(e){Object.assign(this,(0,r.pick)(e,Object.getOwnPropertyNames(this)))}}},"./src/generation/logits_process.js":
@@ -172,7 +172,7 @@ var r,i,a,s,o,l,u,d,c,p,h,m,f,g,_,w,y,b,v,x,M,T,k,$,S,C,E,P,A,F,z,I,O,B=Object.d
   \*************************************/(e,t,n)=>{n.r(t),n.d(t,{BaseStreamer:()=>s,TextStreamer:()=>l,WhisperTextStreamer:()=>u});var r=n(/*! ../utils/core.js */"./src/utils/core.js"),i=n(/*! ../tokenizers.js */"./src/tokenizers.js"),a=n(/*! ../env.js */"./src/env.js");class s{put(e){throw Error("Not implemented")}end(){throw Error("Not implemented")}}const o=a.apis.IS_PROCESS_AVAILABLE?e=>process.stdout.write(e):e=>console.log(e);class l extends s{constructor(e,{skip_prompt:t=!1,callback_function:n=null,token_callback_function:r=null,decode_kwargs:i={},...a}={}){super(),this.tokenizer=e,this.skip_prompt=t,this.callback_function=n??o,this.token_callback_function=r,this.decode_kwargs={...i,...a},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 a;n.endsWith("\n")?(a=n.slice(this.print_len),this.token_cache=[],this.print_len=0):n.length>0&&(0,i.is_chinese_char)(n.charCodeAt(n.length-1))?(a=n.slice(this.print_len),this.print_len+=a.length):(a=n.slice(this.print_len,n.lastIndexOf(" ")+1),this.print_len+=a.length),this.on_finalized_text(a,!1)}end(){let e;if(this.token_cache.length>0){e=this.tokenizer.decode(this.token_cache,this.decode_kwargs).slice(this.print_len),this.token_cache=[],this.print_len=0}else e="";this.next_tokens_are_prompt=!0,this.on_finalized_text(e,!0)}on_finalized_text(e,t){e.length>0&&this.callback_function?.(e),t&&this.callback_function===o&&a.apis.IS_PROCESS_AVAILABLE&&this.callback_function?.("\n")}}class u extends l{constructor(e,{skip_prompt:t=!1,callback_function:n=null,token_callback_function:r=null,on_chunk_start:i=null,on_chunk_end:a=null,on_finalize:s=null,time_precision:o=.02,skip_special_tokens:l=!0,decode_kwargs:u={}}={}){super(e,{skip_prompt:t,callback_function:n,token_callback_function:r,decode_kwargs:{skip_special_tokens:l,...u}}),this.timestamp_begin=e.timestamp_begin,this.on_chunk_start=i,this.on_chunk_end=a,this.on_finalize=s,this.time_precision=o,this.waiting_for_timestamp=!1}put(e){if(e.length>1)throw Error("WhisperTextStreamer only supports batch size of 1");const t=e[0];if(1===t.length){const n=Number(t[0])-this.timestamp_begin;if(n>=0){const t=n*this.time_precision;this.waiting_for_timestamp?this.on_chunk_end?.(t):this.on_chunk_start?.(t),this.waiting_for_timestamp=!this.waiting_for_timestamp,e=[[]]}}return super.put(e)}end(){super.end(),this.on_finalize?.()}}},"./src/models.js":
 /*!***********************!*\
   !*** ./src/models.js ***!
-  \***********************/(e,t,n)=>{n.r(t),n.d(t,{ASTForAudioClassification:()=>nn,ASTModel:()=>tn,ASTPreTrainedModel:()=>en,AlbertForMaskedLM:()=>ct,AlbertForQuestionAnswering:()=>dt,AlbertForSequenceClassification:()=>ut,AlbertModel:()=>lt,AlbertPreTrainedModel:()=>ot,AutoModel:()=>lo,AutoModelForAudioClassification:()=>So,AutoModelForAudioFrameClassification:()=>Eo,AutoModelForCTC:()=>$o,AutoModelForCausalLM:()=>go,AutoModelForDepthEstimation:()=>zo,AutoModelForDocumentQuestionAnswering:()=>Po,AutoModelForImageClassification:()=>bo,AutoModelForImageFeatureExtraction:()=>Io,AutoModelForImageMatting:()=>Ao,AutoModelForImageSegmentation:()=>vo,AutoModelForImageToImage:()=>Fo,AutoModelForMaskGeneration:()=>ko,AutoModelForMaskedLM:()=>_o,AutoModelForObjectDetection:()=>Mo,AutoModelForQuestionAnswering:()=>wo,AutoModelForSemanticSegmentation:()=>xo,AutoModelForSeq2SeqLM:()=>po,AutoModelForSequenceClassification:()=>uo,AutoModelForSpeechSeq2Seq:()=>ho,AutoModelForTextToSpectrogram:()=>mo,AutoModelForTextToWaveform:()=>fo,AutoModelForTokenClassification:()=>co,AutoModelForVision2Seq:()=>yo,AutoModelForXVector:()=>Co,AutoModelForZeroShotObjectDetection:()=>To,BartForConditionalGeneration:()=>Mt,BartForSequenceClassification:()=>Tt,BartModel:()=>xt,BartPretrainedModel:()=>vt,BaseModelOutput:()=>G,BeitForImageClassification:()=>Xr,BeitModel:()=>Hr,BeitPreTrainedModel:()=>Wr,BertForMaskedLM:()=>W,BertForQuestionAnswering:()=>K,BertForSequenceClassification:()=>H,BertForTokenClassification:()=>X,BertModel:()=>U,BertPreTrainedModel:()=>q,BlenderbotForConditionalGeneration:()=>Ft,BlenderbotModel:()=>At,BlenderbotPreTrainedModel:()=>Pt,BlenderbotSmallForConditionalGeneration:()=>Ot,BlenderbotSmallModel:()=>It,BlenderbotSmallPreTrainedModel:()=>zt,BloomForCausalLM:()=>yr,BloomModel:()=>wr,BloomPreTrainedModel:()=>_r,CLIPModel:()=>mn,CLIPPreTrainedModel:()=>hn,CLIPSegForImageSegmentation:()=>kn,CLIPSegModel:()=>Tn,CLIPSegPreTrainedModel:()=>Mn,CLIPTextModelWithProjection:()=>fn,CLIPVisionModelWithProjection:()=>gn,CamembertForMaskedLM:()=>we,CamembertForQuestionAnswering:()=>ve,CamembertForSequenceClassification:()=>ye,CamembertForTokenClassification:()=>be,CamembertModel:()=>_e,CamembertPreTrainedModel:()=>ge,CausalLMOutput:()=>Vo,CausalLMOutputWithPast:()=>jo,ChineseCLIPModel:()=>xn,ChineseCLIPPreTrainedModel:()=>vn,ClapAudioModelWithProjection:()=>ts,ClapModel:()=>Za,ClapPreTrainedModel:()=>Ja,ClapTextModelWithProjection:()=>es,CodeGenForCausalLM:()=>Wn,CodeGenModel:()=>Un,CodeGenPreTrainedModel:()=>qn,CohereForCausalLM:()=>Jn,CohereModel:()=>Yn,CoherePreTrainedModel:()=>Qn,ConvBertForMaskedLM:()=>se,ConvBertForQuestionAnswering:()=>ue,ConvBertForSequenceClassification:()=>oe,ConvBertForTokenClassification:()=>le,ConvBertModel:()=>ae,ConvBertPreTrainedModel:()=>ie,ConvNextForImageClassification:()=>Li,ConvNextModel:()=>Bi,ConvNextPreTrainedModel:()=>Oi,ConvNextV2ForImageClassification:()=>Ni,ConvNextV2Model:()=>Ri,ConvNextV2PreTrainedModel:()=>Di,DPTForDepthEstimation:()=>Mi,DPTModel:()=>xi,DPTPreTrainedModel:()=>vi,DebertaForMaskedLM:()=>Te,DebertaForQuestionAnswering:()=>Se,DebertaForSequenceClassification:()=>ke,DebertaForTokenClassification:()=>$e,DebertaModel:()=>Me,DebertaPreTrainedModel:()=>xe,DebertaV2ForMaskedLM:()=>Pe,DebertaV2ForQuestionAnswering:()=>ze,DebertaV2ForSequenceClassification:()=>Ae,DebertaV2ForTokenClassification:()=>Fe,DebertaV2Model:()=>Ee,DebertaV2PreTrainedModel:()=>Ce,DeiTForImageClassification:()=>ci,DeiTModel:()=>di,DeiTPreTrainedModel:()=>ui,DepthAnythingForDepthEstimation:()=>ki,DepthAnythingPreTrainedModel:()=>Ti,DetrForObjectDetection:()=>Yr,DetrForSegmentation:()=>Jr,DetrModel:()=>Qr,DetrObjectDetectionOutput:()=>Zr,DetrPreTrainedModel:()=>Kr,DetrSegmentationOutput:()=>ei,Dinov2ForImageClassification:()=>Gi,Dinov2Model:()=>ji,Dinov2PreTrainedModel:()=>Vi,DistilBertForMaskedLM:()=>Re,DistilBertForQuestionAnswering:()=>De,DistilBertForSequenceClassification:()=>Be,DistilBertForTokenClassification:()=>Le,DistilBertModel:()=>Oe,DistilBertPreTrainedModel:()=>Ie,DonutSwinModel:()=>Ii,DonutSwinPreTrainedModel:()=>zi,EfficientNetForImageClassification:()=>hs,EfficientNetModel:()=>ps,EfficientNetPreTrainedModel:()=>cs,ElectraForMaskedLM:()=>pe,ElectraForQuestionAnswering:()=>fe,ElectraForSequenceClassification:()=>he,ElectraForTokenClassification:()=>me,ElectraModel:()=>ce,ElectraPreTrainedModel:()=>de,EsmForMaskedLM:()=>je,EsmForSequenceClassification:()=>Ge,EsmForTokenClassification:()=>qe,EsmModel:()=>Ve,EsmPreTrainedModel:()=>Ne,FalconForCausalLM:()=>Ya,FalconModel:()=>Qa,FalconPreTrainedModel:()=>Ka,FastViTForImageClassification:()=>Ar,FastViTModel:()=>Pr,FastViTPreTrainedModel:()=>Er,Florence2ForConditionalGeneration:()=>pn,Florence2PreTrainedModel:()=>cn,GLPNForDepthEstimation:()=>Fi,GLPNModel:()=>Ai,GLPNPreTrainedModel:()=>Pi,GPT2LMHeadModel:()=>Cn,GPT2Model:()=>Sn,GPT2PreTrainedModel:()=>$n,GPTBigCodeForCausalLM:()=>Gn,GPTBigCodeModel:()=>jn,GPTBigCodePreTrainedModel:()=>Vn,GPTJForCausalLM:()=>Nn,GPTJModel:()=>Rn,GPTJPreTrainedModel:()=>Dn,GPTNeoForCausalLM:()=>In,GPTNeoModel:()=>zn,GPTNeoPreTrainedModel:()=>Fn,GPTNeoXForCausalLM:()=>Ln,GPTNeoXModel:()=>Bn,GPTNeoXPreTrainedModel:()=>On,Gemma2ForCausalLM:()=>ir,Gemma2Model:()=>rr,Gemma2PreTrainedModel:()=>nr,GemmaForCausalLM:()=>tr,GemmaModel:()=>er,GemmaPreTrainedModel:()=>Zn,HubertForCTC:()=>Ca,HubertForSequenceClassification:()=>Ea,HubertModel:()=>Sa,HubertPreTrainedModel:()=>$a,ImageMattingOutput:()=>Go,JAISLMHeadModel:()=>An,JAISModel:()=>Pn,JAISPreTrainedModel:()=>En,LlamaForCausalLM:()=>Kn,LlamaModel:()=>Xn,LlamaPreTrainedModel:()=>Hn,LlavaForConditionalGeneration:()=>un,LlavaPreTrainedModel:()=>ln,LongT5ForConditionalGeneration:()=>_t,LongT5Model:()=>gt,LongT5PreTrainedModel:()=>ft,M2M100ForConditionalGeneration:()=>na,M2M100Model:()=>ta,M2M100PreTrainedModel:()=>ea,MBartForCausalLM:()=>Et,MBartForConditionalGeneration:()=>St,MBartForSequenceClassification:()=>Ct,MBartModel:()=>$t,MBartPreTrainedModel:()=>kt,MPNetForMaskedLM:()=>Je,MPNetForQuestionAnswering:()=>tt,MPNetForSequenceClassification:()=>Ze,MPNetForTokenClassification:()=>et,MPNetModel:()=>Ye,MPNetPreTrainedModel:()=>Qe,MT5ForConditionalGeneration:()=>bt,MT5Model:()=>yt,MT5PreTrainedModel:()=>wt,MarianMTModel:()=>Zi,MarianModel:()=>Ji,MarianPreTrainedModel:()=>Yi,MaskedLMOutput:()=>Ro,MistralForCausalLM:()=>Ua,MistralModel:()=>qa,MistralPreTrainedModel:()=>Ga,MobileBertForMaskedLM:()=>He,MobileBertForQuestionAnswering:()=>Ke,MobileBertForSequenceClassification:()=>Xe,MobileBertModel:()=>We,MobileBertPreTrainedModel:()=>Ue,MobileNetV1ForImageClassification:()=>bs,MobileNetV1Model:()=>ys,MobileNetV1PreTrainedModel:()=>ws,MobileNetV2ForImageClassification:()=>Ms,MobileNetV2Model:()=>xs,MobileNetV2PreTrainedModel:()=>vs,MobileNetV3ForImageClassification:()=>$s,MobileNetV3Model:()=>ks,MobileNetV3PreTrainedModel:()=>Ts,MobileNetV4ForImageClassification:()=>Es,MobileNetV4Model:()=>Cs,MobileNetV4PreTrainedModel:()=>Ss,MobileViTForImageClassification:()=>Br,MobileViTModel:()=>Or,MobileViTPreTrainedModel:()=>Ir,MobileViTV2ForImageClassification:()=>Rr,MobileViTV2Model:()=>Dr,MobileViTV2PreTrainedModel:()=>Lr,ModelOutput:()=>j,Moondream1ForConditionalGeneration:()=>dn,MptForCausalLM:()=>xr,MptModel:()=>vr,MptPreTrainedModel:()=>br,MusicgenForCausalLM:()=>gs,MusicgenForConditionalGeneration:()=>_s,MusicgenModel:()=>fs,MusicgenPreTrainedModel:()=>ms,NomicBertModel:()=>Y,NomicBertPreTrainedModel:()=>Q,OPTForCausalLM:()=>kr,OPTModel:()=>Tr,OPTPreTrainedModel:()=>Mr,OpenELMForCausalLM:()=>or,OpenELMModel:()=>sr,OpenELMPreTrainedModel:()=>ar,OwlViTForObjectDetection:()=>jr,OwlViTModel:()=>Vr,OwlViTPreTrainedModel:()=>Nr,Owlv2ForObjectDetection:()=>Ur,Owlv2Model:()=>qr,Owlv2PreTrainedModel:()=>Gr,Phi3ForCausalLM:()=>gr,Phi3Model:()=>fr,Phi3PreTrainedModel:()=>mr,PhiForCausalLM:()=>hr,PhiModel:()=>pr,PhiPreTrainedModel:()=>cr,PreTrainedModel:()=>V,PretrainedMixin:()=>Ps,PyAnnoteForAudioFrameClassification:()=>da,PyAnnoteModel:()=>ua,PyAnnotePreTrainedModel:()=>la,QuestionAnsweringModelOutput:()=>No,Qwen2ForCausalLM:()=>dr,Qwen2Model:()=>ur,Qwen2PreTrainedModel:()=>lr,RTDetrForObjectDetection:()=>ri,RTDetrModel:()=>ni,RTDetrObjectDetectionOutput:()=>ii,RTDetrPreTrainedModel:()=>ti,ResNetForImageClassification:()=>mi,ResNetModel:()=>hi,ResNetPreTrainedModel:()=>pi,RoFormerForMaskedLM:()=>ee,RoFormerForQuestionAnswering:()=>re,RoFormerForSequenceClassification:()=>te,RoFormerForTokenClassification:()=>ne,RoFormerModel:()=>Z,RoFormerPreTrainedModel:()=>J,RobertaForMaskedLM:()=>Dt,RobertaForQuestionAnswering:()=>Vt,RobertaForSequenceClassification:()=>Rt,RobertaForTokenClassification:()=>Nt,RobertaModel:()=>Lt,RobertaPreTrainedModel:()=>Bt,SamImageSegmentationOutput:()=>Qi,SamModel:()=>Ki,SamPreTrainedModel:()=>Xi,SapiensForDepthEstimation:()=>Ci,SapiensForNormalEstimation:()=>Ei,SapiensForSemanticSegmentation:()=>Si,SapiensPreTrainedModel:()=>$i,SegformerForImageClassification:()=>ss,SegformerForSemanticSegmentation:()=>os,SegformerModel:()=>as,SegformerPreTrainedModel:()=>is,Seq2SeqLMOutput:()=>Oo,SequenceClassifierOutput:()=>Bo,SiglipModel:()=>wn,SiglipPreTrainedModel:()=>_n,SiglipTextModel:()=>yn,SiglipVisionModel:()=>bn,SpeechT5ForSpeechToText:()=>Da,SpeechT5ForTextToSpeech:()=>Ra,SpeechT5HifiGan:()=>Na,SpeechT5Model:()=>La,SpeechT5PreTrainedModel:()=>Ba,SqueezeBertForMaskedLM:()=>it,SqueezeBertForQuestionAnswering:()=>st,SqueezeBertForSequenceClassification:()=>at,SqueezeBertModel:()=>rt,SqueezeBertPreTrainedModel:()=>nt,StableLmForCausalLM:()=>ds,StableLmModel:()=>us,StableLmPreTrainedModel:()=>ls,Starcoder2ForCausalLM:()=>Xa,Starcoder2Model:()=>Ha,Starcoder2PreTrainedModel:()=>Wa,Swin2SRForImageSuperResolution:()=>bi,Swin2SRModel:()=>yi,Swin2SRPreTrainedModel:()=>wi,SwinForImageClassification:()=>_i,SwinModel:()=>gi,SwinPreTrainedModel:()=>fi,T5ForConditionalGeneration:()=>mt,T5Model:()=>ht,T5PreTrainedModel:()=>pt,TableTransformerForObjectDetection:()=>oi,TableTransformerModel:()=>si,TableTransformerObjectDetectionOutput:()=>li,TableTransformerPreTrainedModel:()=>ai,TokenClassifierOutput:()=>Do,TrOCRForCausalLM:()=>ja,TrOCRPreTrainedModel:()=>Va,UniSpeechForCTC:()=>fa,UniSpeechForSequenceClassification:()=>ga,UniSpeechModel:()=>ma,UniSpeechPreTrainedModel:()=>ha,UniSpeechSatForAudioFrameClassification:()=>va,UniSpeechSatForCTC:()=>ya,UniSpeechSatForSequenceClassification:()=>ba,UniSpeechSatModel:()=>wa,UniSpeechSatPreTrainedModel:()=>_a,ViTForImageClassification:()=>Cr,ViTModel:()=>Sr,ViTPreTrainedModel:()=>$r,VisionEncoderDecoderModel:()=>on,VitMatteForImageMatting:()=>zr,VitMattePreTrainedModel:()=>Fr,VitsModel:()=>rs,VitsModelOutput:()=>qo,VitsPreTrainedModel:()=>ns,Wav2Vec2BertForCTC:()=>Ta,Wav2Vec2BertForSequenceClassification:()=>ka,Wav2Vec2BertModel:()=>Ma,Wav2Vec2BertPreTrainedModel:()=>xa,Wav2Vec2ForAudioFrameClassification:()=>oa,Wav2Vec2ForCTC:()=>aa,Wav2Vec2ForSequenceClassification:()=>sa,Wav2Vec2Model:()=>ia,Wav2Vec2PreTrainedModel:()=>ra,WavLMForAudioFrameClassification:()=>Oa,WavLMForCTC:()=>Fa,WavLMForSequenceClassification:()=>za,WavLMForXVector:()=>Ia,WavLMModel:()=>Aa,WavLMPreTrainedModel:()=>Pa,WeSpeakerResNetModel:()=>pa,WeSpeakerResNetPreTrainedModel:()=>ca,WhisperForConditionalGeneration:()=>sn,WhisperModel:()=>an,WhisperPreTrainedModel:()=>rn,XLMForQuestionAnswering:()=>Ht,XLMForSequenceClassification:()=>Ut,XLMForTokenClassification:()=>Wt,XLMModel:()=>Gt,XLMPreTrainedModel:()=>jt,XLMRobertaForMaskedLM:()=>Qt,XLMRobertaForQuestionAnswering:()=>Zt,XLMRobertaForSequenceClassification:()=>Yt,XLMRobertaForTokenClassification:()=>Jt,XLMRobertaModel:()=>Kt,XLMRobertaPreTrainedModel:()=>Xt,XLMWithLMHeadModel:()=>qt,XVectorOutput:()=>Lo,YolosForObjectDetection:()=>Wi,YolosModel:()=>Ui,YolosObjectDetectionOutput:()=>Hi,YolosPreTrainedModel:()=>qi});var r=n(/*! ./configs.js */"./src/configs.js"),i=n(/*! ./backends/onnx.js */"./src/backends/onnx.js"),a=n(/*! ./utils/dtypes.js */"./src/utils/dtypes.js"),s=n(/*! ./utils/generic.js */"./src/utils/generic.js"),o=n(/*! ./utils/core.js */"./src/utils/core.js"),l=n(/*! ./utils/hub.js */"./src/utils/hub.js"),u=n(/*! ./generation/logits_process.js */"./src/generation/logits_process.js"),d=n(/*! ./generation/configuration_utils.js */"./src/generation/configuration_utils.js"),c=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"),f=n(/*! ./env.js */"./src/env.js"),g=n(/*! ./models/whisper/generation_whisper.js */"./src/models/whisper/generation_whisper.js"),_=n(/*! ./models/whisper/common_whisper.js */"./src/models/whisper/common_whisper.js");const w=0,y=1,b=2,v=3,x=4,M=5,T=6,k=7,$=new Map,S=new Map,C=new Map;async function E(e,t,n){return Object.fromEntries(await Promise.all(Object.keys(t).map((async s=>{const{buffer:o,session_options:u}=await async function(e,t,n){let s=n.device;s&&"string"!=typeof s&&(s.hasOwnProperty(t)?s=s[t]:(console.warn(`device not specified for "${t}". Using the default device.`),s=null));const o=s??(f.apis.IS_NODE_ENV?"cpu":"wasm"),u=(0,i.deviceToExecutionProviders)(o);let d=n.dtype;"string"!=typeof d&&(d&&d.hasOwnProperty(t)?d=d[t]:(d=a.DEFAULT_DEVICE_DTYPE_MAPPING[o]??a.DATA_TYPES.fp32,console.warn(`dtype not specified for "${t}". Using the default dtype (${d}) for this device (${o}).`)));const c=d;if(!a.DEFAULT_DTYPE_SUFFIX_MAPPING.hasOwnProperty(c))throw new Error(`Invalid dtype: ${c}. Should be one of: ${Object.keys(a.DATA_TYPES).join(", ")}`);if(c===a.DATA_TYPES.fp16&&"webgpu"===o&&!await(0,a.isWebGpuFp16Supported)())throw new Error(`The device (${o}) does not support fp16.`);const p=a.DEFAULT_DTYPE_SUFFIX_MAPPING[c],h=`${n.subfolder??""}/${t}${p}.onnx`,m={...n.session_options}??{};m.executionProviders??=u;const g=(0,l.getModelFile)(e,h,!0,n);let _=[];if(n.use_external_data_format&&(!0===n.use_external_data_format||"object"==typeof n.use_external_data_format&&n.use_external_data_format.hasOwnProperty(t)&&!0===n.use_external_data_format[t])){if(f.apis.IS_NODE_ENV)throw new Error("External data format is not yet supported in Node.js");const r=`${t}${p}.onnx_data`,i=`${n.subfolder??""}/${r}`;_.push(new Promise((async(t,a)=>{const s=await(0,l.getModelFile)(e,i,!0,n);t({path:r,data:s})})))}else void 0!==m.externalData&&(_=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(_.length>0&&(m.externalData=await Promise.all(_)),"webgpu"===o){const e=(0,r.getKeyValueShapes)(n.config,{prefix:"present"});if(Object.keys(e).length>0&&!(0,i.isONNXProxy)()){const t={};for(const n in e)t[n]="gpu-buffer";m.preferredOutputLocation=t}}return{buffer:await g,session_options:m}}(e,t[s],n);return[s,await(0,i.createInferenceSession)(o,u)]}))))}async function P(e,t){const n=function(e,t){const n=Object.create(null),r=[];for(const a of e.inputNames){const e=t[a];e instanceof c.Tensor?n[a]=(0,i.isONNXProxy)()?e.clone():e:r.push(a)}if(r.length>0)throw new Error(`An error occurred during model execution: "Missing the following inputs: ${r.join(", ")}.`);const a=Object.keys(t).length,s=e.inputNames.length;if(a>s){let n=Object.keys(t).filter((t=>!e.inputNames.includes(t)));console.warn(`WARNING: Too many inputs were provided (${a} > ${s}). The following inputs will be ignored: "${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=A(r),r}catch(e){throw console.error(`An error occurred during model execution: "${e}".`),console.error("Inputs given to model:",n),e}}function A(e){for(let t in e)(0,i.isONNXTensor)(e[t])?e[t]=new c.Tensor(e[t]):"object"==typeof e[t]&&A(e[t]);return e}function F(e){if(e instanceof c.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 c.Tensor("int64",BigInt64Array.from(e.flat().map((e=>BigInt(e)))),[e.length,e[0].length])}return new c.Tensor("int64",BigInt64Array.from(e.map((e=>BigInt(e)))),[1,e.length])}function z(e){return new c.Tensor("bool",[e],[1])}async function I(e,t){let{encoder_outputs:n,input_ids:r,decoder_input_ids:i,...a}=t;if(!n){const r=(0,o.pick)(t,e.sessions.model.inputNames);n=(await O(e,r)).last_hidden_state}a.input_ids=i,a.encoder_hidden_states=n,e.sessions.decoder_model_merged.inputNames.includes("encoder_attention_mask")&&(a.encoder_attention_mask=t.attention_mask);return await B(e,a,!0)}async function O(e,t){const n=e.sessions.model,r=(0,o.pick)(t,n.inputNames);if(n.inputNames.includes("inputs_embeds")&&!r.inputs_embeds){if(!t.input_ids)throw new Error("Both `input_ids` and `inputs_embeds` are missing in the model inputs.");r.inputs_embeds=await e.encode_text({input_ids:t.input_ids})}return n.inputNames.includes("token_type_ids")&&!r.token_type_ids&&(r.token_type_ids=new c.Tensor("int64",new BigInt64Array(r.input_ids.data.length),r.input_ids.dims)),await P(n,r)}async function B(e,t,n=!1){const r=e.sessions[n?"decoder_model_merged":"model"],{past_key_values:i,...a}=t;r.inputNames.includes("use_cache_branch")&&(a.use_cache_branch=z(!!i)),r.inputNames.includes("position_ids")&&a.attention_mask&&!a.position_ids&&(a.position_ids=function(e,t=null){const{input_ids:n,inputs_embeds:r,attention_mask:i}=e,[a,s]=i.dims,o=new BigInt64Array(i.data.length);for(let e=0;e<a;++e){const t=e*s;let n=BigInt(0);for(let e=0;e<s;++e){const r=t+e;0n===i.data[r]?o[r]=BigInt(1):(o[r]=n,n+=i.data[r])}}let l=new c.Tensor("int64",o,i.dims);if(t){const e=-(n??r).dims.at(1);l=l.slice(null,[e,null])}return l}(a,i)),e.addPastKeyValues(a,i);const s=(0,o.pick)(a,r.inputNames);return await P(r,s)}async function L(e,{input_ids:t=null,attention_mask:n=null,pixel_values:r=null,position_ids:i=null,inputs_embeds:a=null,past_key_values:s=null,generation_config:o=null,logits_processor:l=null,...u}){if(!a)if(a=await e.encode_text({input_ids:t}),r&&1!==t.dims[1]){const i=await e.encode_image({pixel_values:r});({inputs_embeds:a,attention_mask:n}=e._merge_input_ids_with_image_features({image_features:i,inputs_embeds:a,input_ids:t,attention_mask:n}))}else if(s&&r&&1===t.dims[1]){const e=t.dims[1],r=Object.values(s)[0].dims.at(-2);n=(0,c.cat)([(0,c.ones)([t.dims[0],r]),n.slice(null,[n.dims[1]-e,n.dims[1]])],1)}return await B(e,{inputs_embeds:a,past_key_values:s,attention_mask:n,position_ids:i,generation_config:o,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:i}=n;if(i&&i.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 i=e.config.num_image_tokens;if(!i)throw new Error("`num_image_tokens` is missing in the model configuration.");const a=r.dims[1]-(t-i);n.input_ids=r.slice(null,[-a,null]),n.attention_mask=(0,c.ones)([1,t+a])}}return n}function R(e,t,n,r){return n.past_key_values&&(t=t.map((e=>[e.at(-1)]))),{...n,decoder_input_ids:F(t)}}function N(e,...t){return e.config.is_encoder_decoder?R(e,...t):D(e,...t)}class V extends s.Callable{main_input_name="input_ids";forward_params=["input_ids","attention_mask"];constructor(e,t){super(),this.config=e,this.sessions=t;const n=C.get(this.constructor),r=$.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=B,this._prepare_inputs_for_generation=D;break;case b:case v:case k:this.can_generate=!0,this._forward=I,this._prepare_inputs_for_generation=R;break;case y:this._forward=I;break;case T:this.can_generate=!0,this._forward=L,this._prepare_inputs_for_generation=N;break;default:this._forward=O}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:i=null,local_files_only:a=!1,revision:s="main",model_file_name:o=null,subfolder:u="onnx",device:d=null,dtype:c=null,use_external_data_format:p=null,session_options:h={}}={}){let m={progress_callback:t,config:n,cache_dir:i,local_files_only:a,revision:s,model_file_name:o,subfolder:u,device:d,dtype:c,use_external_data_format:p,session_options:h};const f=C.get(this),g=$.get(f);let _;if(n=m.config=await r.AutoConfig.from_pretrained(e,m),g===x)_=await Promise.all([E(e,{model:m.model_file_name??"model"},m),(0,l.getModelJSON)(e,"generation_config.json",!1,m)]);else if(g===b||g===v)_=await Promise.all([E(e,{model:"encoder_model",decoder_model_merged:"decoder_model_merged"},m),(0,l.getModelJSON)(e,"generation_config.json",!1,m)]);else if(g===M)_=await Promise.all([E(e,{model:"vision_encoder",prompt_encoder_mask_decoder:"prompt_encoder_mask_decoder"},m)]);else if(g===y)_=await Promise.all([E(e,{model:"encoder_model",decoder_model_merged:"decoder_model_merged"},m)]);else if(g===T){const t={embed_tokens:"embed_tokens",vision_encoder:"vision_encoder",decoder_model_merged:"decoder_model_merged"};n.is_encoder_decoder&&(t.model="encoder_model"),_=await Promise.all([E(e,t,m),(0,l.getModelJSON)(e,"generation_config.json",!1,m)])}else g===k?_=await Promise.all([E(e,{model:"text_encoder",decoder_model_merged:"decoder_model_merged",encodec_decode:"encodec_decode"},m),(0,l.getModelJSON)(e,"generation_config.json",!1,m)]):(g!==w&&console.warn(`Model type for '${f??n?.model_type}' not found, assuming encoder-only architecture. Please report this at https://github.com/xenova/transformers.js/issues/new/choose.`),_=await Promise.all([E(e,{model:m.model_file_name??"model"},m)]));return new this(n,..._)}async _call(e){return await this.forward(e)}async forward(e){return await this._forward(this,e)}_get_logits_warper(e){const t=new u.LogitsProcessorList;return null!==e.temperature&&1!==e.temperature&&t.push(new u.TemperatureLogitsWarper(e.temperature)),null!==e.top_k&&0!==e.top_k&&t.push(new u.TopKLogitsWarper(e.top_k)),null!==e.top_p&&e.top_p<1&&t.push(new u.TopPLogitsWarper(e.top_p)),t}_get_logits_processor(e,t,n=null){const r=new u.LogitsProcessorList;if(null!==e.repetition_penalty&&1!==e.repetition_penalty&&r.push(new u.RepetitionPenaltyLogitsProcessor(e.repetition_penalty)),null!==e.no_repeat_ngram_size&&e.no_repeat_ngram_size>0&&r.push(new u.NoRepeatNGramLogitsProcessor(e.no_repeat_ngram_size)),null!==e.bad_words_ids&&r.push(new u.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 u.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 u.MinNewTokensLengthLogitsProcessor(t,e.min_new_tokens,e.eos_token_id)),null!==e.forced_bos_token_id&&r.push(new u.ForcedBOSTokenLogitsProcessor(e.forced_bos_token_id)),null!==e.forced_eos_token_id&&r.push(new u.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 u.SuppressTokensAtBeginLogitsProcessor(e.begin_suppress_tokens,n))}return null!==e.guidance_scale&&e.guidance_scale>1&&r.push(new u.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 i=new n(r);return"generation_config"in this&&Object.assign(i,this.generation_config),e&&Object.assign(i,e),t&&Object.assign(i,(0,o.pick)(t,Object.getOwnPropertyNames(i))),i}_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=[Ns,Gs,Rs,Is],t=C.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 i=`The current model class (${t}) is not compatible with \`.generate()\`, as it doesn't have a language model head.`;throw n.size>0&&(i+=` Please use the following class instead: ${[...n].join(", ")}`),Error(i)}}prepare_inputs_for_generation(...e){return this._prepare_inputs_for_generation(this,...e)}_update_model_kwargs_for_generation({generated_input_ids:e,outputs:t,model_inputs:n,is_encoder_decoder:r}){return n.past_key_values=this.getPastKeyValues(t,n.past_key_values),n.input_ids=new c.Tensor("int64",e.flat(),[e.length,1]),r||(n.attention_mask=(0,c.cat)([n.attention_mask,(0,c.ones)([n.attention_mask.dims[0],1])],1)),n.position_ids=null,n}_prepare_model_inputs({inputs:e,bos_token_id:t,model_kwargs:n}){const r=(0,o.pick)(n,this.forward_params),i=this.main_input_name;if(i in r){if(e)throw new Error("`inputs`: {inputs}` were passed alongside {input_name} which is not allowed. Make sure to either pass {inputs} or {input_name}=...")}else r[i]=e;return{inputs_tensor:r[i],model_inputs:r,model_input_name:i}}async _prepare_encoder_decoder_kwargs_for_generation({inputs_tensor:e,model_inputs:t,model_input_name:n,generation_config:r}){if(this.sessions.model.inputNames.includes("inputs_embeds")&&!t.inputs_embeds&&"_prepare_inputs_embeds"in this){const{input_ids:e,pixel_values:n,attention_mask:r,...i}=t,a=await this._prepare_inputs_embeds(t);t={...i,...(0,o.pick)(a,["inputs_embeds","attention_mask"])}}let{last_hidden_state:i}=await O(this,t);if(null!==r.guidance_scale&&r.guidance_scale>1)i=(0,c.cat)([i,(0,c.full_like)(i,0)],0),"attention_mask"in t&&(t.attention_mask=(0,c.cat)([t.attention_mask,(0,c.zeros_like)(t.attention_mask)],0));else if(t.decoder_input_ids){const e=F(t.decoder_input_ids).dims[0];if(e!==i.dims[0]){if(1!==i.dims[0])throw new Error(`The encoder outputs have a different batch size (${i.dims[0]}) than the decoder inputs (${e}).`);i=(0,c.cat)(Array.from({length:e},(()=>i)),0)}}return t.encoder_outputs=i,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:i,generation_config:a}){let{decoder_input_ids:s,...o}=n;if(s)Array.isArray(s[0])||(s=Array.from({length:e},(()=>s)));else if(r??=i,"musicgen"===this.config.model_type)s=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}`);s=r}else s=Array.from({length:e},(()=>[r]));return s=F(s),n.decoder_attention_mask=(0,c.ones_like)(s),{input_ids:s,model_inputs:o}}async generate({inputs:e=null,generation_config:t=null,logits_processor:n=null,stopping_criteria:r=null,streamer:i=null,...a}){this._validate_model_class(),t=this._prepare_generation_config(t,a);let{inputs_tensor:s,model_inputs:o,model_input_name:l}=this._prepare_model_inputs({inputs:e,model_kwargs:a});const u=this.config.is_encoder_decoder;let d;u&&("encoder_outputs"in o||(o=await this._prepare_encoder_decoder_kwargs_for_generation({inputs_tensor:s,model_inputs:o,model_input_name:l,generation_config:t}))),u?({input_ids:d,model_inputs:o}=this._prepare_decoder_input_ids_for_generation({batch_size:o[l].dims.at(0),model_input_name:l,model_kwargs:o,decoder_start_token_id:t.decoder_start_token_id,bos_token_id:t.bos_token_id,generation_config:t})):d=o[l];let 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),f=this._get_stopping_criteria(t,r),g=o[l].dims.at(0),_=m.LogitsSampler.getSampler(t),w=new Array(g).fill(0),y=d.tolist();i&&i.put(y);let b=null,v={};for(;;){o=this.prepare_inputs_for_generation(y,o,t);const e=await this.forward(o);if(t.output_attentions&&t.return_dict_in_generate){const t=this.getAttentions(e);for(const e in t)e in v||(v[e]=[]),v[e].push(t[e])}const n=h(y,e.logits.slice(null,-1,null)),r=[];for(let e=0;e<n.dims.at(0);++e){const t=n[e],i=await _(t);for(const[t,n]of i){const i=BigInt(t);w[e]+=n,y[e].push(i),r.push([i]);break}}i&&i.put(r);if(f(y).every((e=>e))){t.return_dict_in_generate&&(b=this.getPastKeyValues(e,o.past_key_values,!1));break}o=this._update_model_kwargs_for_generation({generated_input_ids:r,outputs:e,model_inputs:o,is_encoder_decoder:u})}i&&i.end();const x=new c.Tensor("int64",y.flat(),[y.length,y[0].length]);return t.return_dict_in_generate?{sequences:x,past_key_values:b,...v}:x}getPastKeyValues(e,t,n=!0){const r=Object.create(null);for(const i in e)if(i.startsWith("present")){const a=i.replace("present","past_key_values");if(t&&i.includes("encoder"))r[a]=t[a];else{if(n&&t){const e=t[a];"gpu-buffer"===e.location&&e.dispose()}r[a]=e[i]}}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:[],i=(0,r.getKeyValueShapes)(this.config);for(const r in i)e[r]=new c.Tensor(t,n,i[r])}}async encode_image({pixel_values:e}){const t=(await P(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 P(this.sessions.embed_tokens,{input_ids:e})).inputs_embeds}}class j{}class G extends j{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 q extends V{}class U extends q{}class W extends q{async _call(e){return new Ro(await super._call(e))}}class H extends q{async _call(e){return new Bo(await super._call(e))}}class X extends q{async _call(e){return new Do(await super._call(e))}}class K extends q{async _call(e){return new No(await super._call(e))}}class Q extends V{}class Y extends Q{}class J extends V{}class Z extends J{}class ee extends J{async _call(e){return new Ro(await super._call(e))}}class te extends J{async _call(e){return new Bo(await super._call(e))}}class ne extends J{async _call(e){return new Do(await super._call(e))}}class re extends J{async _call(e){return new No(await super._call(e))}}class ie extends V{}class ae extends ie{}class se extends ie{async _call(e){return new Ro(await super._call(e))}}class oe extends ie{async _call(e){return new Bo(await super._call(e))}}class le extends ie{async _call(e){return new Do(await super._call(e))}}class ue extends ie{async _call(e){return new No(await super._call(e))}}class de extends V{}class ce extends de{}class pe extends de{async _call(e){return new Ro(await super._call(e))}}class he extends de{async _call(e){return new Bo(await super._call(e))}}class me extends de{async _call(e){return new Do(await super._call(e))}}class fe extends de{async _call(e){return new No(await super._call(e))}}class ge extends V{}class _e extends ge{}class we extends ge{async _call(e){return new Ro(await super._call(e))}}class ye extends ge{async _call(e){return new Bo(await super._call(e))}}class be extends ge{async _call(e){return new Do(await super._call(e))}}class ve extends ge{async _call(e){return new No(await super._call(e))}}class xe extends V{}class Me extends xe{}class Te extends xe{async _call(e){return new Ro(await super._call(e))}}class ke extends xe{async _call(e){return new Bo(await super._call(e))}}class $e extends xe{async _call(e){return new Do(await super._call(e))}}class Se extends xe{async _call(e){return new No(await super._call(e))}}class Ce extends V{}class Ee extends Ce{}class Pe extends Ce{async _call(e){return new Ro(await super._call(e))}}class Ae extends Ce{async _call(e){return new Bo(await super._call(e))}}class Fe extends Ce{async _call(e){return new Do(await super._call(e))}}class ze extends Ce{async _call(e){return new No(await super._call(e))}}class Ie extends V{}class Oe extends Ie{}class Be extends Ie{async _call(e){return new Bo(await super._call(e))}}class Le extends Ie{async _call(e){return new Do(await super._call(e))}}class De extends Ie{async _call(e){return new No(await super._call(e))}}class Re extends Ie{async _call(e){return new Ro(await super._call(e))}}class Ne extends V{}class Ve extends Ne{}class je extends Ne{async _call(e){return new Ro(await super._call(e))}}class Ge extends Ne{async _call(e){return new Bo(await super._call(e))}}class qe extends Ne{async _call(e){return new Do(await super._call(e))}}class Ue extends V{}class We extends Ue{}class He extends Ue{async _call(e){return new Ro(await super._call(e))}}class Xe extends Ue{async _call(e){return new Bo(await super._call(e))}}class Ke extends Ue{async _call(e){return new No(await super._call(e))}}class Qe extends V{}class Ye extends Qe{}class Je extends Qe{async _call(e){return new Ro(await super._call(e))}}class Ze extends Qe{async _call(e){return new Bo(await super._call(e))}}class et extends Qe{async _call(e){return new Do(await super._call(e))}}class tt extends Qe{async _call(e){return new No(await super._call(e))}}class nt extends V{}class rt extends nt{}class it extends nt{async _call(e){return new Ro(await super._call(e))}}class at extends nt{async _call(e){return new Bo(await super._call(e))}}class st extends nt{async _call(e){return new No(await super._call(e))}}class ot extends V{}class lt extends ot{}class ut extends ot{async _call(e){return new Bo(await super._call(e))}}class dt extends ot{async _call(e){return new No(await super._call(e))}}class ct extends ot{async _call(e){return new Ro(await super._call(e))}}class pt extends V{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 ft extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class gt extends ft{}class _t extends ft{}class wt extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class yt extends wt{}class bt extends wt{}class vt extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class xt extends vt{}class Mt extends vt{}class Tt extends vt{async _call(e){return new Bo(await super._call(e))}}class kt extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class $t extends kt{}class St extends kt{}class Ct extends kt{async _call(e){return new Bo(await super._call(e))}}class Et extends kt{}class Pt extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class At extends Pt{}class Ft extends Pt{}class zt extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class It extends zt{}class Ot extends zt{}class Bt extends V{}class Lt extends Bt{}class Dt extends Bt{async _call(e){return new Ro(await super._call(e))}}class Rt extends Bt{async _call(e){return new Bo(await super._call(e))}}class Nt extends Bt{async _call(e){return new Do(await super._call(e))}}class Vt extends Bt{async _call(e){return new No(await super._call(e))}}class jt extends V{}class Gt extends jt{}class qt extends jt{async _call(e){return new Ro(await super._call(e))}}class Ut extends jt{async _call(e){return new Bo(await super._call(e))}}class Wt extends jt{async _call(e){return new Do(await super._call(e))}}class Ht extends jt{async _call(e){return new No(await super._call(e))}}class Xt extends V{}class Kt extends Xt{}class Qt extends Xt{async _call(e){return new Ro(await super._call(e))}}class Yt extends Xt{async _call(e){return new Bo(await super._call(e))}}class Jt extends Xt{async _call(e){return new Do(await super._call(e))}}class Zt extends Xt{async _call(e){return new No(await super._call(e))}}class en extends V{}class tn extends en{}class nn extends en{}class rn extends V{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 an extends rn{}class sn extends rn{_prepare_generation_config(e,t){return super._prepare_generation_config(e,t,g.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 i=`<|${(0,_.whisper_language_to_code)(n)}|>`;t.push(e.lang_to_id[i]),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,...i}){t=this._prepare_generation_config(t,i);const a=i.decoder_input_ids??this._retrieve_init_tokens(t);if(t.return_timestamps&&(n??=new u.LogitsProcessorList,n.push(new u.WhisperTimeStampLogitsProcessor(t,a))),t.begin_suppress_tokens&&(n??=new u.LogitsProcessorList,n.push(new u.SuppressTokensAtBeginLogitsProcessor(t.begin_suppress_tokens,a.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 s=await super.generate({inputs:e,generation_config:t,logits_processor:n,decoder_input_ids:a,...i});return t.return_token_timestamps&&(s.token_timestamps=this._extract_token_timestamps(s,t.alignment_heads,t.num_frames)),s}_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 i=this.config.median_filter_width;void 0===i&&(console.warn("Model config has no `median_filter_width`, using default value of 7."),i=7);const a=e.cross_attentions,s=Array.from({length:this.config.decoder_layers},((e,t)=>(0,c.cat)(a.map((e=>e[t])),2))),l=(0,c.stack)(t.map((([e,t])=>{if(e>=s.length)throw new Error(`Layer index ${e} is out of bounds for cross attentions (length ${s.length}).`);return n?s[e].slice(null,t,null,[0,n]):s[e].slice(null,t)}))).transpose(1,0,2,3),[u,d]=(0,c.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],a=u[e][n][0].data,s=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]-s[e])/a[e];t.set((0,p.medianFilter)(t,i))}}}const m=[(0,c.mean)(h,1)],f=e.sequences.dims,g=new c.Tensor("float32",new Float32Array(f[0]*f[1]),f);for(let e=0;e<f[0];++e){const t=m[e].neg().squeeze_(0),[n,i]=(0,p.dynamic_time_warping)(t.tolist()),a=Array.from({length:n.length-1},((e,t)=>n[t+1]-n[t])),s=(0,o.mergeArrays)([1],a).map((e=>!!e)),l=[];for(let e=0;e<s.length;++e)s[e]&&l.push(i[e]*r);g[e].data.set(l,1)}return g}}class on extends V{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 V{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 un extends ln{_merge_input_ids_with_image_features({inputs_embeds:e,image_features:t,input_ids:n,attention_mask:r}){const i=this.config.image_token_index,a=n.tolist().map((e=>e.findIndex((e=>e==i)))),s=a.every((e=>-1===e)),o=a.every((e=>-1!==e));if(!s&&!o)throw new Error("Every input should contain either 0 or 1 image token.");if(s)return{inputs_embeds:e,attention_mask:r};const l=[],u=[];for(let n=0;n<a.length;++n){const i=a[n],s=e[n],o=t[n],d=r[n];l.push((0,c.cat)([s.slice([0,i]),o,s.slice([i+1,s.dims[0]])],0)),u.push((0,c.cat)([d.slice([0,i]),(0,c.ones)([o.dims[0]]),d.slice([i+1,d.dims[0]])],0))}return{inputs_embeds:(0,c.stack)(l,0),attention_mask:(0,c.stack)(u,0)}}}class dn extends un{}class cn extends V{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 cn{_merge_input_ids_with_image_features({inputs_embeds:e,image_features:t,input_ids:n,attention_mask:r}){return{inputs_embeds:(0,c.cat)([t,e],1),attention_mask:(0,c.cat)([(0,c.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 i,a;return e&&(i=await this.encode_text({input_ids:e})),t&&(a=await this.encode_image({pixel_values:t})),i&&a?({inputs_embeds:n,attention_mask:r}=this._merge_input_ids_with_image_features({inputs_embeds:i,image_features:a,input_ids:e,attention_mask:r})):n=i||a,{inputs_embeds:n,attention_mask:r}}async forward({input_ids:e,pixel_values:t,attention_mask:n,decoder_input_ids:r,decoder_attention_mask:i,encoder_outputs:a,past_key_values:s,inputs_embeds:o,decoder_inputs_embeds:l}){if(o||({inputs_embeds:o,attention_mask:n}=await this._prepare_inputs_embeds({input_ids:e,pixel_values:t,inputs_embeds:o,attention_mask:n})),!a){let{last_hidden_state:e}=await O(this,{inputs_embeds:o,attention_mask:n});a=e}if(!l){if(!r)throw new Error("Either `decoder_input_ids` or `decoder_inputs_embeds` should be provided.");l=await this.encode_text({input_ids:r})}const u={inputs_embeds:l,attention_mask:i,encoder_attention_mask:n,encoder_hidden_states:a,past_key_values:s};return await B(this,u,!0)}}class hn extends V{}class mn extends hn{}class fn extends hn{static async from_pretrained(e,t={}){return t.model_file_name??="text_model",super.from_pretrained(e,t)}}class gn extends hn{static async from_pretrained(e,t={}){return t.model_file_name??="vision_model",super.from_pretrained(e,t)}}class _n extends V{}class wn extends _n{}class yn extends _n{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 vn extends V{}class xn extends vn{}class Mn extends V{}class Tn extends Mn{}class kn extends Mn{}class $n extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class Sn extends $n{}class Cn extends $n{}class En extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class Pn extends En{}class An extends En{}class Fn extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class zn extends Fn{}class In extends Fn{}class On extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class Bn extends On{}class Ln extends On{}class Dn extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class Rn extends Dn{}class Nn extends Dn{}class Vn extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class jn extends Vn{}class Gn extends Vn{}class qn extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class Un extends qn{}class Wn extends qn{}class Hn extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class Xn extends Hn{}class Kn extends Hn{}class Qn extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class Yn extends Qn{}class Jn extends Qn{}class Zn extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class er extends Zn{}class tr extends Zn{}class nr extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class rr extends nr{}class ir extends nr{}class ar extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class sr extends ar{}class or extends ar{}class lr extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class ur extends lr{}class dr extends lr{}class cr extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class pr extends cr{}class hr extends cr{}class mr extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class fr extends mr{}class gr extends mr{}class _r extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class wr extends _r{}class yr extends _r{}class br extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class vr extends br{}class xr extends br{}class Mr extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class Tr extends Mr{}class kr extends Mr{}class $r extends V{}class Sr extends $r{}class Cr extends $r{async _call(e){return new Bo(await super._call(e))}}class Er extends V{}class Pr extends Er{}class Ar extends Er{async _call(e){return new Bo(await super._call(e))}}class Fr extends V{}class zr extends Fr{async _call(e){return new Go(await super._call(e))}}class Ir extends V{}class Or extends Ir{}class Br extends Ir{async _call(e){return new Bo(await super._call(e))}}class Lr extends V{}class Dr extends Lr{}class Rr extends Lr{async _call(e){return new Bo(await super._call(e))}}class Nr extends V{}class Vr extends Nr{}class jr extends Nr{}class Gr extends V{}class qr extends Gr{}class Ur extends Gr{}class Wr extends V{}class Hr extends Wr{}class Xr extends Wr{async _call(e){return new Bo(await super._call(e))}}class Kr extends V{}class Qr extends Kr{}class Yr extends Kr{async _call(e){return new Zr(await super._call(e))}}class Jr extends Kr{async _call(e){return new ei(await super._call(e))}}class Zr extends j{constructor({logits:e,pred_boxes:t}){super(),this.logits=e,this.pred_boxes=t}}class ei extends j{constructor({logits:e,pred_boxes:t,pred_masks:n}){super(),this.logits=e,this.pred_boxes=t,this.pred_masks=n}}class ti extends V{}class ni extends ti{}class ri extends ti{async _call(e){return new ii(await super._call(e))}}class ii extends j{constructor({logits:e,pred_boxes:t}){super(),this.logits=e,this.pred_boxes=t}}class ai extends V{}class si extends ai{}class oi extends ai{async _call(e){return new li(await super._call(e))}}class li extends Zr{}class ui extends V{}class di extends ui{}class ci extends ui{async _call(e){return new Bo(await super._call(e))}}class pi extends V{}class hi extends pi{}class mi extends pi{async _call(e){return new Bo(await super._call(e))}}class fi extends V{}class gi extends fi{}class _i extends fi{async _call(e){return new Bo(await super._call(e))}}class wi extends V{}class yi extends wi{}class bi extends wi{}class vi extends V{}class xi extends vi{}class Mi extends vi{}class Ti extends V{}class ki extends Ti{}class $i extends V{}class Si extends $i{}class Ci extends $i{}class Ei extends $i{}class Pi extends V{}class Ai extends Pi{}class Fi extends Pi{}class zi extends V{}class Ii extends zi{}class Oi extends V{}class Bi extends Oi{}class Li extends Oi{async _call(e){return new Bo(await super._call(e))}}class Di extends V{}class Ri extends Di{}class Ni extends Di{async _call(e){return new Bo(await super._call(e))}}class Vi extends V{}class ji extends Vi{}class Gi extends Vi{async _call(e){return new Bo(await super._call(e))}}class qi extends V{}class Ui extends qi{}class Wi extends qi{async _call(e){return new Hi(await super._call(e))}}class Hi extends j{constructor({logits:e,pred_boxes:t}){super(),this.logits=e,this.pred_boxes=t}}class Xi extends V{}class Ki extends Xi{async get_image_embeddings({pixel_values:e}){return await O(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 c.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 P(this.sessions.prompt_encoder_mask_decoder,t)}async _call(e){return new Qi(await super._call(e))}}class Qi extends j{constructor({iou_scores:e,pred_masks:t}){super(),this.iou_scores=e,this.pred_masks=t}}class Yi extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class Ji extends Yi{}class Zi extends Yi{}class ea extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class ta extends ea{}class na extends ea{}class ra extends V{}class ia extends ra{}class aa extends ra{async _call(e){return new Vo(await super._call(e))}}class sa extends ra{async _call(e){return new Bo(await super._call(e))}}class oa extends ra{async _call(e){return new Do(await super._call(e))}}class la extends V{}class ua extends la{}class da extends la{async _call(e){return new Do(await super._call(e))}}class ca extends V{}class pa extends ca{}class ha extends V{}class ma extends ha{}class fa extends ha{async _call(e){return new Vo(await super._call(e))}}class ga extends ha{async _call(e){return new Bo(await super._call(e))}}class _a extends V{}class wa extends _a{}class ya extends _a{async _call(e){return new Vo(await super._call(e))}}class ba extends _a{async _call(e){return new Bo(await super._call(e))}}class va extends _a{async _call(e){return new Do(await super._call(e))}}class xa extends V{}class Ma extends xa{}class Ta extends xa{async _call(e){return new Vo(await super._call(e))}}class ka extends xa{async _call(e){return new Bo(await super._call(e))}}class $a extends V{}class Sa extends ra{}class Ca extends ra{async _call(e){return new Vo(await super._call(e))}}class Ea extends ra{async _call(e){return new Bo(await super._call(e))}}class Pa extends V{}class Aa extends Pa{}class Fa extends Pa{async _call(e){return new Vo(await super._call(e))}}class za extends Pa{async _call(e){return new Bo(await super._call(e))}}class Ia extends Pa{async _call(e){return new Lo(await super._call(e))}}class Oa extends Pa{async _call(e){return new Do(await super._call(e))}}class Ba extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class La extends Ba{}class Da extends Ba{}class Ra extends Ba{async generate_speech(e,t,{threshold:n=.5,minlenratio:r=0,maxlenratio:i=20,vocoder:a=null}={}){const s={input_ids:e},{encoder_outputs:o,encoder_attention_mask:l}=await O(this,s),u=o.dims[1]/this.config.reduction_factor,d=Math.floor(u*i),p=Math.floor(u*r),h=this.config.num_mel_bins;let m=[],f=null,g=null,_=0;for(;;){++_;const e=z(!!g);let r;r=g?g.output_sequence_out:new c.Tensor("float32",new Float32Array(h),[1,1,h]);let i={use_cache_branch:e,output_sequence:r,encoder_attention_mask:l,speaker_embeddings:t,encoder_hidden_states:o};this.addPastKeyValues(i,f),g=await P(this.sessions.decoder_model_merged,i),f=this.getPastKeyValues(g,f);const{prob:a,spectrum:s}=g;if(m.push(s),_>=p&&(Array.from(a.data).filter((e=>e>=n)).length>0||_>=d))break}const w=(0,c.cat)(m),{waveform:y}=await P(a.sessions.model,{spectrogram:w});return{spectrogram:w,waveform:y}}}class Na extends V{main_input_name="spectrogram"}class Va extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class ja extends Va{}class Ga extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class qa extends Ga{}class Ua extends Ga{}class Wa extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class Ha extends Wa{}class Xa extends Wa{}class Ka extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class Qa extends Ka{}class Ya extends Ka{}class Ja extends V{}class Za extends Ja{}class es extends Ja{static async from_pretrained(e,t={}){return t.model_file_name??="text_model",super.from_pretrained(e,t)}}class ts extends Ja{static async from_pretrained(e,t={}){return t.model_file_name??="audio_model",super.from_pretrained(e,t)}}class ns extends V{}class rs extends ns{async _call(e){return new qo(await super._call(e))}}class is extends V{}class as extends is{}class ss extends is{}class os extends is{}class ls extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class us extends ls{}class ds extends ls{}class cs extends V{}class ps extends cs{}class hs extends cs{async _call(e){return new Bo(await super._call(e))}}class ms extends V{}class fs extends ms{}class gs extends ms{}class _s extends V{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,i=n-r;let a=0;for(let t=0;t<e.size;++t){if(e.data[t]===this.config.decoder.pad_token_id)continue;const s=t%n-Math.floor(t/n)%r;s>0&&s<=i&&(e.data[a++]=e.data[t])}const s=Math.floor(t/r),o=a/(s*r);return new c.Tensor(e.type,e.data.slice(0,a),[s,r,o])}prepare_inputs_for_generation(e,t,n){let r=structuredClone(e);for(let e=0;e<r.length;++e)for(let t=0;t<r[e].length;++t)e%this.config.decoder.num_codebooks>=t&&(r[e][t]=BigInt(this.config.decoder.pad_token_id));null!==n.guidance_scale&&n.guidance_scale>1&&(r=r.concat(r));return super.prepare_inputs_for_generation(r,t,n)}async generate(e){const t=await super.generate(e),n=this._apply_and_filter_by_delay_pattern_mask(t).unsqueeze_(0),{audio_values:r}=await P(this.sessions.encodec_decode,{audio_codes:n});return r}}class ws extends V{}class ys extends ws{}class bs extends ws{async _call(e){return new Bo(await super._call(e))}}class vs extends V{}class xs extends vs{}class Ms extends vs{async _call(e){return new Bo(await super._call(e))}}class Ts extends V{}class ks extends Ts{}class $s extends Ts{async _call(e){return new Bo(await super._call(e))}}class Ss extends V{}class Cs extends Ss{}class Es extends Ss{async _call(e){return new Bo(await super._call(e))}}class Ps{static MODEL_CLASS_MAPPINGS=null;static BASE_IF_FAIL=!1;static async from_pretrained(e,{progress_callback:t=null,config:n=null,cache_dir:i=null,local_files_only:a=!1,revision:s="main",model_file_name:o=null,subfolder:l="onnx",device:u=null,dtype:d=null,use_external_data_format:c=null,session_options:p={}}={}){let h={progress_callback:t,config:n,cache_dir:i,local_files_only:a,revision:s,model_file_name:o,subfolder:l,device:u,dtype:d,use_external_data_format:c,session_options:p};if(h.config=await r.AutoConfig.from_pretrained(e,h),!this.MODEL_CLASS_MAPPINGS)throw new Error("`MODEL_CLASS_MAPPINGS` not implemented for this type of `AutoClass`: "+this.name);for(let t of this.MODEL_CLASS_MAPPINGS){const n=t.get(h.config.model_type);if(n)return await n[1].from_pretrained(e,h)}if(this.BASE_IF_FAIL)return console.warn(`Unknown model class "${h.config.model_type}", attempting to construct from base class.`),await V.from_pretrained(e,h);throw Error(`Unsupported model type: ${h.config.model_type}`)}}const As=new Map([["bert",["BertModel",U]],["nomic_bert",["NomicBertModel",Y]],["roformer",["RoFormerModel",Z]],["electra",["ElectraModel",ce]],["esm",["EsmModel",Ve]],["convbert",["ConvBertModel",ae]],["camembert",["CamembertModel",_e]],["deberta",["DebertaModel",Me]],["deberta-v2",["DebertaV2Model",Ee]],["mpnet",["MPNetModel",Ye]],["albert",["AlbertModel",lt]],["distilbert",["DistilBertModel",Oe]],["roberta",["RobertaModel",Lt]],["xlm",["XLMModel",Gt]],["xlm-roberta",["XLMRobertaModel",Kt]],["clap",["ClapModel",Za]],["clip",["CLIPModel",mn]],["clipseg",["CLIPSegModel",Tn]],["chinese_clip",["ChineseCLIPModel",xn]],["siglip",["SiglipModel",wn]],["mobilebert",["MobileBertModel",We]],["squeezebert",["SqueezeBertModel",rt]],["wav2vec2",["Wav2Vec2Model",ia]],["wav2vec2-bert",["Wav2Vec2BertModel",Ma]],["unispeech",["UniSpeechModel",ma]],["unispeech-sat",["UniSpeechSatModel",wa]],["hubert",["HubertModel",Sa]],["wavlm",["WavLMModel",Aa]],["audio-spectrogram-transformer",["ASTModel",tn]],["vits",["VitsModel",rs]],["pyannote",["PyAnnoteModel",ua]],["wespeaker-resnet",["WeSpeakerResNetModel",pa]],["detr",["DetrModel",Qr]],["rt_detr",["RTDetrModel",ni]],["table-transformer",["TableTransformerModel",si]],["vit",["ViTModel",Sr]],["fastvit",["FastViTModel",Pr]],["mobilevit",["MobileViTModel",Or]],["mobilevitv2",["MobileViTV2Model",Dr]],["owlvit",["OwlViTModel",Vr]],["owlv2",["Owlv2Model",qr]],["beit",["BeitModel",Hr]],["deit",["DeiTModel",di]],["convnext",["ConvNextModel",Bi]],["convnextv2",["ConvNextV2Model",Ri]],["dinov2",["Dinov2Model",ji]],["resnet",["ResNetModel",hi]],["swin",["SwinModel",gi]],["swin2sr",["Swin2SRModel",yi]],["donut-swin",["DonutSwinModel",Ii]],["yolos",["YolosModel",Ui]],["dpt",["DPTModel",xi]],["glpn",["GLPNModel",Ai]],["hifigan",["SpeechT5HifiGan",Na]],["efficientnet",["EfficientNetModel",ps]],["mobilenet_v1",["MobileNetV1Model",ys]],["mobilenet_v2",["MobileNetV2Model",xs]],["mobilenet_v3",["MobileNetV3Model",ks]],["mobilenet_v4",["MobileNetV4Model",Cs]]]),Fs=new Map([["t5",["T5Model",ht]],["longt5",["LongT5Model",gt]],["mt5",["MT5Model",yt]],["bart",["BartModel",xt]],["mbart",["MBartModel",$t]],["marian",["MarianModel",Ji]],["whisper",["WhisperModel",an]],["m2m_100",["M2M100Model",ta]],["blenderbot",["BlenderbotModel",At]],["blenderbot-small",["BlenderbotSmallModel",It]]]),zs=new Map([["bloom",["BloomModel",wr]],["jais",["JAISModel",Pn]],["gpt2",["GPT2Model",Sn]],["gptj",["GPTJModel",Rn]],["gpt_bigcode",["GPTBigCodeModel",jn]],["gpt_neo",["GPTNeoModel",zn]],["gpt_neox",["GPTNeoXModel",Bn]],["codegen",["CodeGenModel",Un]],["llama",["LlamaModel",Xn]],["cohere",["CohereModel",Yn]],["gemma",["GemmaModel",er]],["gemma2",["Gemma2Model",rr]],["openelm",["OpenELMModel",sr]],["qwen2",["Qwen2Model",ur]],["phi",["PhiModel",pr]],["phi3",["Phi3Model",fr]],["mpt",["MptModel",vr]],["opt",["OPTModel",Tr]],["mistral",["MistralModel",qa]],["starcoder2",["Starcoder2Model",Ha]],["falcon",["FalconModel",Qa]],["stablelm",["StableLmModel",us]]]),Is=new Map([["speecht5",["SpeechT5ForSpeechToText",Da]],["whisper",["WhisperForConditionalGeneration",sn]]]),Os=new Map([["speecht5",["SpeechT5ForTextToSpeech",Ra]]]),Bs=new Map([["vits",["VitsModel",rs]],["musicgen",["MusicgenForConditionalGeneration",_s]]]),Ls=new Map([["bert",["BertForSequenceClassification",H]],["roformer",["RoFormerForSequenceClassification",te]],["electra",["ElectraForSequenceClassification",he]],["esm",["EsmForSequenceClassification",Ge]],["convbert",["ConvBertForSequenceClassification",oe]],["camembert",["CamembertForSequenceClassification",ye]],["deberta",["DebertaForSequenceClassification",ke]],["deberta-v2",["DebertaV2ForSequenceClassification",Ae]],["mpnet",["MPNetForSequenceClassification",Ze]],["albert",["AlbertForSequenceClassification",ut]],["distilbert",["DistilBertForSequenceClassification",Be]],["roberta",["RobertaForSequenceClassification",Rt]],["xlm",["XLMForSequenceClassification",Ut]],["xlm-roberta",["XLMRobertaForSequenceClassification",Yt]],["bart",["BartForSequenceClassification",Tt]],["mbart",["MBartForSequenceClassification",Ct]],["mobilebert",["MobileBertForSequenceClassification",Xe]],["squeezebert",["SqueezeBertForSequenceClassification",at]]]),Ds=new Map([["bert",["BertForTokenClassification",X]],["roformer",["RoFormerForTokenClassification",ne]],["electra",["ElectraForTokenClassification",me]],["esm",["EsmForTokenClassification",qe]],["convbert",["ConvBertForTokenClassification",le]],["camembert",["CamembertForTokenClassification",be]],["deberta",["DebertaForTokenClassification",$e]],["deberta-v2",["DebertaV2ForTokenClassification",Fe]],["mpnet",["MPNetForTokenClassification",et]],["distilbert",["DistilBertForTokenClassification",Le]],["roberta",["RobertaForTokenClassification",Nt]],["xlm",["XLMForTokenClassification",Wt]],["xlm-roberta",["XLMRobertaForTokenClassification",Jt]]]),Rs=new Map([["t5",["T5ForConditionalGeneration",mt]],["longt5",["LongT5ForConditionalGeneration",_t]],["mt5",["MT5ForConditionalGeneration",bt]],["bart",["BartForConditionalGeneration",Mt]],["mbart",["MBartForConditionalGeneration",St]],["marian",["MarianMTModel",Zi]],["m2m_100",["M2M100ForConditionalGeneration",na]],["blenderbot",["BlenderbotForConditionalGeneration",Ft]],["blenderbot-small",["BlenderbotSmallForConditionalGeneration",Ot]]]),Ns=new Map([["bloom",["BloomForCausalLM",yr]],["gpt2",["GPT2LMHeadModel",Cn]],["jais",["JAISLMHeadModel",An]],["gptj",["GPTJForCausalLM",Nn]],["gpt_bigcode",["GPTBigCodeForCausalLM",Gn]],["gpt_neo",["GPTNeoForCausalLM",In]],["gpt_neox",["GPTNeoXForCausalLM",Ln]],["codegen",["CodeGenForCausalLM",Wn]],["llama",["LlamaForCausalLM",Kn]],["cohere",["CohereForCausalLM",Jn]],["gemma",["GemmaForCausalLM",tr]],["gemma2",["Gemma2ForCausalLM",ir]],["openelm",["OpenELMForCausalLM",or]],["qwen2",["Qwen2ForCausalLM",dr]],["phi",["PhiForCausalLM",hr]],["phi3",["Phi3ForCausalLM",gr]],["mpt",["MptForCausalLM",xr]],["opt",["OPTForCausalLM",kr]],["mbart",["MBartForCausalLM",Et]],["mistral",["MistralForCausalLM",Ua]],["starcoder2",["Starcoder2ForCausalLM",Xa]],["falcon",["FalconForCausalLM",Ya]],["trocr",["TrOCRForCausalLM",ja]],["stablelm",["StableLmForCausalLM",ds]]]),Vs=new Map([["bert",["BertForMaskedLM",W]],["roformer",["RoFormerForMaskedLM",ee]],["electra",["ElectraForMaskedLM",pe]],["esm",["EsmForMaskedLM",je]],["convbert",["ConvBertForMaskedLM",se]],["camembert",["CamembertForMaskedLM",we]],["deberta",["DebertaForMaskedLM",Te]],["deberta-v2",["DebertaV2ForMaskedLM",Pe]],["mpnet",["MPNetForMaskedLM",Je]],["albert",["AlbertForMaskedLM",ct]],["distilbert",["DistilBertForMaskedLM",Re]],["roberta",["RobertaForMaskedLM",Dt]],["xlm",["XLMWithLMHeadModel",qt]],["xlm-roberta",["XLMRobertaForMaskedLM",Qt]],["mobilebert",["MobileBertForMaskedLM",He]],["squeezebert",["SqueezeBertForMaskedLM",it]]]),js=new Map([["bert",["BertForQuestionAnswering",K]],["roformer",["RoFormerForQuestionAnswering",re]],["electra",["ElectraForQuestionAnswering",fe]],["convbert",["ConvBertForQuestionAnswering",ue]],["camembert",["CamembertForQuestionAnswering",ve]],["deberta",["DebertaForQuestionAnswering",Se]],["deberta-v2",["DebertaV2ForQuestionAnswering",ze]],["mpnet",["MPNetForQuestionAnswering",tt]],["albert",["AlbertForQuestionAnswering",dt]],["distilbert",["DistilBertForQuestionAnswering",De]],["roberta",["RobertaForQuestionAnswering",Vt]],["xlm",["XLMForQuestionAnswering",Ht]],["xlm-roberta",["XLMRobertaForQuestionAnswering",Zt]],["mobilebert",["MobileBertForQuestionAnswering",Ke]],["squeezebert",["SqueezeBertForQuestionAnswering",st]]]),Gs=new Map([["vision-encoder-decoder",["VisionEncoderDecoderModel",on]]]),qs=new Map([["llava",["LlavaForConditionalGeneration",un]],["moondream1",["Moondream1ForConditionalGeneration",dn]],["florence2",["Florence2ForConditionalGeneration",pn]]]),Us=new Map([["vision-encoder-decoder",["VisionEncoderDecoderModel",on]]]),Ws=new Map([["vit",["ViTForImageClassification",Cr]],["fastvit",["FastViTForImageClassification",Ar]],["mobilevit",["MobileViTForImageClassification",Br]],["mobilevitv2",["MobileViTV2ForImageClassification",Rr]],["beit",["BeitForImageClassification",Xr]],["deit",["DeiTForImageClassification",ci]],["convnext",["ConvNextForImageClassification",Li]],["convnextv2",["ConvNextV2ForImageClassification",Ni]],["dinov2",["Dinov2ForImageClassification",Gi]],["resnet",["ResNetForImageClassification",mi]],["swin",["SwinForImageClassification",_i]],["segformer",["SegformerForImageClassification",ss]],["efficientnet",["EfficientNetForImageClassification",hs]],["mobilenet_v1",["MobileNetV1ForImageClassification",bs]],["mobilenet_v2",["MobileNetV2ForImageClassification",Ms]],["mobilenet_v3",["MobileNetV3ForImageClassification",$s]],["mobilenet_v4",["MobileNetV4ForImageClassification",Es]]]),Hs=new Map([["detr",["DetrForObjectDetection",Yr]],["rt_detr",["RTDetrForObjectDetection",ri]],["table-transformer",["TableTransformerForObjectDetection",oi]],["yolos",["YolosForObjectDetection",Wi]]]),Xs=new Map([["owlvit",["OwlViTForObjectDetection",jr]],["owlv2",["Owlv2ForObjectDetection",Ur]]]),Ks=new Map([["detr",["DetrForSegmentation",Jr]],["clipseg",["CLIPSegForImageSegmentation",kn]]]),Qs=new Map([["segformer",["SegformerForSemanticSegmentation",os]],["sapiens",["SapiensForSemanticSegmentation",Si]]]),Ys=new Map([["sam",["SamModel",Ki]]]),Js=new Map([["wav2vec2",["Wav2Vec2ForCTC",aa]],["wav2vec2-bert",["Wav2Vec2BertForCTC",Ta]],["unispeech",["UniSpeechForCTC",fa]],["unispeech-sat",["UniSpeechSatForCTC",ya]],["wavlm",["WavLMForCTC",Fa]],["hubert",["HubertForCTC",Ca]]]),Zs=new Map([["wav2vec2",["Wav2Vec2ForSequenceClassification",sa]],["wav2vec2-bert",["Wav2Vec2BertForSequenceClassification",ka]],["unispeech",["UniSpeechForSequenceClassification",ga]],["unispeech-sat",["UniSpeechSatForSequenceClassification",ba]],["wavlm",["WavLMForSequenceClassification",za]],["hubert",["HubertForSequenceClassification",Ea]],["audio-spectrogram-transformer",["ASTForAudioClassification",nn]]]),eo=new Map([["wavlm",["WavLMForXVector",Ia]]]),to=new Map([["unispeech-sat",["UniSpeechSatForAudioFrameClassification",va]],["wavlm",["WavLMForAudioFrameClassification",Oa]],["wav2vec2",["Wav2Vec2ForAudioFrameClassification",oa]],["pyannote",["PyAnnoteForAudioFrameClassification",da]]]),no=new Map([["vitmatte",["VitMatteForImageMatting",zr]]]),ro=new Map([["swin2sr",["Swin2SRForImageSuperResolution",bi]]]),io=new Map([["dpt",["DPTForDepthEstimation",Mi]],["depth_anything",["DepthAnythingForDepthEstimation",ki]],["glpn",["GLPNForDepthEstimation",Fi]],["sapiens",["SapiensForDepthEstimation",Ci]]]),ao=new Map([["clip",["CLIPVisionModelWithProjection",gn]],["siglip",["SiglipVisionModel",bn]]]),so=[[As,w],[Fs,y],[zs,x],[Ls,w],[Ds,w],[Rs,b],[Is,b],[Ns,x],[Vs,w],[js,w],[Gs,v],[qs,T],[Ws,w],[Ks,w],[Qs,w],[no,w],[ro,w],[io,w],[Hs,w],[Xs,w],[Ys,M],[Js,w],[Zs,w],[Os,b],[Bs,w],[eo,w],[to,w],[ao,w]];for(const[e,t]of so)for(const[n,r]of e.values())$.set(n,t),C.set(r,n),S.set(n,r);const oo=[["MusicgenForConditionalGeneration",_s,k],["CLIPTextModelWithProjection",fn,w],["SiglipTextModel",yn,w],["ClapTextModelWithProjection",es,w],["ClapAudioModelWithProjection",ts,w]];for(const[e,t,n]of oo)$.set(e,n),C.set(t,e),S.set(e,t);class lo extends Ps{static MODEL_CLASS_MAPPINGS=so.map((e=>e[0]));static BASE_IF_FAIL=!0}class uo extends Ps{static MODEL_CLASS_MAPPINGS=[Ls]}class co extends Ps{static MODEL_CLASS_MAPPINGS=[Ds]}class po extends Ps{static MODEL_CLASS_MAPPINGS=[Rs]}class ho extends Ps{static MODEL_CLASS_MAPPINGS=[Is]}class mo extends Ps{static MODEL_CLASS_MAPPINGS=[Os]}class fo extends Ps{static MODEL_CLASS_MAPPINGS=[Bs]}class go extends Ps{static MODEL_CLASS_MAPPINGS=[Ns]}class _o extends Ps{static MODEL_CLASS_MAPPINGS=[Vs]}class wo extends Ps{static MODEL_CLASS_MAPPINGS=[js]}class yo extends Ps{static MODEL_CLASS_MAPPINGS=[Gs]}class bo extends Ps{static MODEL_CLASS_MAPPINGS=[Ws]}class vo extends Ps{static MODEL_CLASS_MAPPINGS=[Ks]}class xo extends Ps{static MODEL_CLASS_MAPPINGS=[Qs]}class Mo extends Ps{static MODEL_CLASS_MAPPINGS=[Hs]}class To extends Ps{static MODEL_CLASS_MAPPINGS=[Xs]}class ko extends Ps{static MODEL_CLASS_MAPPINGS=[Ys]}class $o extends Ps{static MODEL_CLASS_MAPPINGS=[Js]}class So extends Ps{static MODEL_CLASS_MAPPINGS=[Zs]}class Co extends Ps{static MODEL_CLASS_MAPPINGS=[eo]}class Eo extends Ps{static MODEL_CLASS_MAPPINGS=[to]}class Po extends Ps{static MODEL_CLASS_MAPPINGS=[Us]}class Ao extends Ps{static MODEL_CLASS_MAPPINGS=[no]}class Fo extends Ps{static MODEL_CLASS_MAPPINGS=[ro]}class zo extends Ps{static MODEL_CLASS_MAPPINGS=[io]}class Io extends Ps{static MODEL_CLASS_MAPPINGS=[ao]}class Oo extends j{constructor({logits:e,past_key_values:t,encoder_outputs:n,decoder_attentions:r=null,cross_attentions:i=null}){super(),this.logits=e,this.past_key_values=t,this.encoder_outputs=n,this.decoder_attentions=r,this.cross_attentions=i}}class Bo extends j{constructor({logits:e}){super(),this.logits=e}}class Lo extends j{constructor({logits:e,embeddings:t}){super(),this.logits=e,this.embeddings=t}}class Do extends j{constructor({logits:e}){super(),this.logits=e}}class Ro extends j{constructor({logits:e}){super(),this.logits=e}}class No extends j{constructor({start_logits:e,end_logits:t}){super(),this.start_logits=e,this.end_logits=t}}class Vo extends j{constructor({logits:e}){super(),this.logits=e}}class jo extends j{constructor({logits:e,past_key_values:t}){super(),this.logits=e,this.past_key_values=t}}class Go extends j{constructor({alphas:e}){super(),this.alphas=e}}class qo extends j{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:()=>ct,AlbertForQuestionAnswering:()=>dt,AlbertForSequenceClassification:()=>ut,AlbertModel:()=>lt,AlbertPreTrainedModel:()=>ot,AutoModel:()=>uo,AutoModelForAudioClassification:()=>Co,AutoModelForAudioFrameClassification:()=>Po,AutoModelForCTC:()=>So,AutoModelForCausalLM:()=>_o,AutoModelForDepthEstimation:()=>Io,AutoModelForDocumentQuestionAnswering:()=>Ao,AutoModelForImageClassification:()=>vo,AutoModelForImageFeatureExtraction:()=>Bo,AutoModelForImageMatting:()=>Fo,AutoModelForImageSegmentation:()=>xo,AutoModelForImageToImage:()=>zo,AutoModelForMaskGeneration:()=>$o,AutoModelForMaskedLM:()=>wo,AutoModelForNormalEstimation:()=>Oo,AutoModelForObjectDetection:()=>To,AutoModelForQuestionAnswering:()=>yo,AutoModelForSemanticSegmentation:()=>Mo,AutoModelForSeq2SeqLM:()=>ho,AutoModelForSequenceClassification:()=>co,AutoModelForSpeechSeq2Seq:()=>mo,AutoModelForTextToSpectrogram:()=>fo,AutoModelForTextToWaveform:()=>go,AutoModelForTokenClassification:()=>po,AutoModelForVision2Seq:()=>bo,AutoModelForXVector:()=>Eo,AutoModelForZeroShotObjectDetection:()=>ko,BartForConditionalGeneration:()=>Mt,BartForSequenceClassification:()=>Tt,BartModel:()=>xt,BartPretrainedModel:()=>vt,BaseModelOutput:()=>G,BeitForImageClassification:()=>Xr,BeitModel:()=>Hr,BeitPreTrainedModel:()=>Wr,BertForMaskedLM:()=>W,BertForQuestionAnswering:()=>K,BertForSequenceClassification:()=>H,BertForTokenClassification:()=>X,BertModel:()=>U,BertPreTrainedModel:()=>q,BlenderbotForConditionalGeneration:()=>Ft,BlenderbotModel:()=>At,BlenderbotPreTrainedModel:()=>Pt,BlenderbotSmallForConditionalGeneration:()=>Ot,BlenderbotSmallModel:()=>It,BlenderbotSmallPreTrainedModel:()=>zt,BloomForCausalLM:()=>yr,BloomModel:()=>wr,BloomPreTrainedModel:()=>_r,CLIPModel:()=>mn,CLIPPreTrainedModel:()=>hn,CLIPSegForImageSegmentation:()=>kn,CLIPSegModel:()=>Tn,CLIPSegPreTrainedModel:()=>Mn,CLIPTextModelWithProjection:()=>fn,CLIPVisionModelWithProjection:()=>gn,CamembertForMaskedLM:()=>we,CamembertForQuestionAnswering:()=>ve,CamembertForSequenceClassification:()=>ye,CamembertForTokenClassification:()=>be,CamembertModel:()=>_e,CamembertPreTrainedModel:()=>ge,CausalLMOutput:()=>Go,CausalLMOutputWithPast:()=>qo,ChineseCLIPModel:()=>xn,ChineseCLIPPreTrainedModel:()=>vn,ClapAudioModelWithProjection:()=>ts,ClapModel:()=>Za,ClapPreTrainedModel:()=>Ja,ClapTextModelWithProjection:()=>es,CodeGenForCausalLM:()=>Wn,CodeGenModel:()=>Un,CodeGenPreTrainedModel:()=>qn,CohereForCausalLM:()=>Jn,CohereModel:()=>Yn,CoherePreTrainedModel:()=>Qn,ConvBertForMaskedLM:()=>se,ConvBertForQuestionAnswering:()=>ue,ConvBertForSequenceClassification:()=>oe,ConvBertForTokenClassification:()=>le,ConvBertModel:()=>ae,ConvBertPreTrainedModel:()=>ie,ConvNextForImageClassification:()=>Li,ConvNextModel:()=>Bi,ConvNextPreTrainedModel:()=>Oi,ConvNextV2ForImageClassification:()=>Ni,ConvNextV2Model:()=>Ri,ConvNextV2PreTrainedModel:()=>Di,DPTForDepthEstimation:()=>Mi,DPTModel:()=>xi,DPTPreTrainedModel:()=>vi,DebertaForMaskedLM:()=>Te,DebertaForQuestionAnswering:()=>Se,DebertaForSequenceClassification:()=>ke,DebertaForTokenClassification:()=>$e,DebertaModel:()=>Me,DebertaPreTrainedModel:()=>xe,DebertaV2ForMaskedLM:()=>Pe,DebertaV2ForQuestionAnswering:()=>ze,DebertaV2ForSequenceClassification:()=>Ae,DebertaV2ForTokenClassification:()=>Fe,DebertaV2Model:()=>Ee,DebertaV2PreTrainedModel:()=>Ce,DeiTForImageClassification:()=>ci,DeiTModel:()=>di,DeiTPreTrainedModel:()=>ui,DepthAnythingForDepthEstimation:()=>ki,DepthAnythingPreTrainedModel:()=>Ti,DetrForObjectDetection:()=>Yr,DetrForSegmentation:()=>Jr,DetrModel:()=>Qr,DetrObjectDetectionOutput:()=>Zr,DetrPreTrainedModel:()=>Kr,DetrSegmentationOutput:()=>ei,Dinov2ForImageClassification:()=>Gi,Dinov2Model:()=>ji,Dinov2PreTrainedModel:()=>Vi,DistilBertForMaskedLM:()=>Re,DistilBertForQuestionAnswering:()=>De,DistilBertForSequenceClassification:()=>Be,DistilBertForTokenClassification:()=>Le,DistilBertModel:()=>Oe,DistilBertPreTrainedModel:()=>Ie,DonutSwinModel:()=>Ii,DonutSwinPreTrainedModel:()=>zi,EfficientNetForImageClassification:()=>hs,EfficientNetModel:()=>ps,EfficientNetPreTrainedModel:()=>cs,ElectraForMaskedLM:()=>pe,ElectraForQuestionAnswering:()=>fe,ElectraForSequenceClassification:()=>he,ElectraForTokenClassification:()=>me,ElectraModel:()=>ce,ElectraPreTrainedModel:()=>de,EsmForMaskedLM:()=>je,EsmForSequenceClassification:()=>Ge,EsmForTokenClassification:()=>qe,EsmModel:()=>Ve,EsmPreTrainedModel:()=>Ne,FalconForCausalLM:()=>Ya,FalconModel:()=>Qa,FalconPreTrainedModel:()=>Ka,FastViTForImageClassification:()=>Ar,FastViTModel:()=>Pr,FastViTPreTrainedModel:()=>Er,Florence2ForConditionalGeneration:()=>pn,Florence2PreTrainedModel:()=>cn,GLPNForDepthEstimation:()=>Fi,GLPNModel:()=>Ai,GLPNPreTrainedModel:()=>Pi,GPT2LMHeadModel:()=>Cn,GPT2Model:()=>Sn,GPT2PreTrainedModel:()=>$n,GPTBigCodeForCausalLM:()=>Gn,GPTBigCodeModel:()=>jn,GPTBigCodePreTrainedModel:()=>Vn,GPTJForCausalLM:()=>Nn,GPTJModel:()=>Rn,GPTJPreTrainedModel:()=>Dn,GPTNeoForCausalLM:()=>In,GPTNeoModel:()=>zn,GPTNeoPreTrainedModel:()=>Fn,GPTNeoXForCausalLM:()=>Ln,GPTNeoXModel:()=>Bn,GPTNeoXPreTrainedModel:()=>On,Gemma2ForCausalLM:()=>ir,Gemma2Model:()=>rr,Gemma2PreTrainedModel:()=>nr,GemmaForCausalLM:()=>tr,GemmaModel:()=>er,GemmaPreTrainedModel:()=>Zn,HubertForCTC:()=>Ca,HubertForSequenceClassification:()=>Ea,HubertModel:()=>Sa,HubertPreTrainedModel:()=>$a,ImageMattingOutput:()=>Uo,JAISLMHeadModel:()=>An,JAISModel:()=>Pn,JAISPreTrainedModel:()=>En,LlamaForCausalLM:()=>Kn,LlamaModel:()=>Xn,LlamaPreTrainedModel:()=>Hn,LlavaForConditionalGeneration:()=>un,LlavaPreTrainedModel:()=>ln,LongT5ForConditionalGeneration:()=>_t,LongT5Model:()=>gt,LongT5PreTrainedModel:()=>ft,M2M100ForConditionalGeneration:()=>na,M2M100Model:()=>ta,M2M100PreTrainedModel:()=>ea,MBartForCausalLM:()=>Et,MBartForConditionalGeneration:()=>St,MBartForSequenceClassification:()=>Ct,MBartModel:()=>$t,MBartPreTrainedModel:()=>kt,MPNetForMaskedLM:()=>Je,MPNetForQuestionAnswering:()=>tt,MPNetForSequenceClassification:()=>Ze,MPNetForTokenClassification:()=>et,MPNetModel:()=>Ye,MPNetPreTrainedModel:()=>Qe,MT5ForConditionalGeneration:()=>bt,MT5Model:()=>yt,MT5PreTrainedModel:()=>wt,MarianMTModel:()=>Zi,MarianModel:()=>Ji,MarianPreTrainedModel:()=>Yi,MaskedLMOutput:()=>Vo,MistralForCausalLM:()=>Ua,MistralModel:()=>qa,MistralPreTrainedModel:()=>Ga,MobileBertForMaskedLM:()=>He,MobileBertForQuestionAnswering:()=>Ke,MobileBertForSequenceClassification:()=>Xe,MobileBertModel:()=>We,MobileBertPreTrainedModel:()=>Ue,MobileNetV1ForImageClassification:()=>bs,MobileNetV1Model:()=>ys,MobileNetV1PreTrainedModel:()=>ws,MobileNetV2ForImageClassification:()=>Ms,MobileNetV2Model:()=>xs,MobileNetV2PreTrainedModel:()=>vs,MobileNetV3ForImageClassification:()=>$s,MobileNetV3Model:()=>ks,MobileNetV3PreTrainedModel:()=>Ts,MobileNetV4ForImageClassification:()=>Es,MobileNetV4Model:()=>Cs,MobileNetV4PreTrainedModel:()=>Ss,MobileViTForImageClassification:()=>Br,MobileViTModel:()=>Or,MobileViTPreTrainedModel:()=>Ir,MobileViTV2ForImageClassification:()=>Rr,MobileViTV2Model:()=>Dr,MobileViTV2PreTrainedModel:()=>Lr,ModelOutput:()=>j,Moondream1ForConditionalGeneration:()=>dn,MptForCausalLM:()=>xr,MptModel:()=>vr,MptPreTrainedModel:()=>br,MusicgenForCausalLM:()=>gs,MusicgenForConditionalGeneration:()=>_s,MusicgenModel:()=>fs,MusicgenPreTrainedModel:()=>ms,NomicBertModel:()=>Y,NomicBertPreTrainedModel:()=>Q,OPTForCausalLM:()=>kr,OPTModel:()=>Tr,OPTPreTrainedModel:()=>Mr,OpenELMForCausalLM:()=>or,OpenELMModel:()=>sr,OpenELMPreTrainedModel:()=>ar,OwlViTForObjectDetection:()=>jr,OwlViTModel:()=>Vr,OwlViTPreTrainedModel:()=>Nr,Owlv2ForObjectDetection:()=>Ur,Owlv2Model:()=>qr,Owlv2PreTrainedModel:()=>Gr,Phi3ForCausalLM:()=>gr,Phi3Model:()=>fr,Phi3PreTrainedModel:()=>mr,PhiForCausalLM:()=>hr,PhiModel:()=>pr,PhiPreTrainedModel:()=>cr,PreTrainedModel:()=>V,PretrainedMixin:()=>Ps,PyAnnoteForAudioFrameClassification:()=>da,PyAnnoteModel:()=>ua,PyAnnotePreTrainedModel:()=>la,QuestionAnsweringModelOutput:()=>jo,Qwen2ForCausalLM:()=>dr,Qwen2Model:()=>ur,Qwen2PreTrainedModel:()=>lr,RTDetrForObjectDetection:()=>ri,RTDetrModel:()=>ni,RTDetrObjectDetectionOutput:()=>ii,RTDetrPreTrainedModel:()=>ti,ResNetForImageClassification:()=>mi,ResNetModel:()=>hi,ResNetPreTrainedModel:()=>pi,RoFormerForMaskedLM:()=>ee,RoFormerForQuestionAnswering:()=>re,RoFormerForSequenceClassification:()=>te,RoFormerForTokenClassification:()=>ne,RoFormerModel:()=>Z,RoFormerPreTrainedModel:()=>J,RobertaForMaskedLM:()=>Dt,RobertaForQuestionAnswering:()=>Vt,RobertaForSequenceClassification:()=>Rt,RobertaForTokenClassification:()=>Nt,RobertaModel:()=>Lt,RobertaPreTrainedModel:()=>Bt,SamImageSegmentationOutput:()=>Qi,SamModel:()=>Ki,SamPreTrainedModel:()=>Xi,SapiensForDepthEstimation:()=>Ci,SapiensForNormalEstimation:()=>Ei,SapiensForSemanticSegmentation:()=>Si,SapiensPreTrainedModel:()=>$i,SegformerForImageClassification:()=>ss,SegformerForSemanticSegmentation:()=>os,SegformerModel:()=>as,SegformerPreTrainedModel:()=>is,Seq2SeqLMOutput:()=>Lo,SequenceClassifierOutput:()=>Do,SiglipModel:()=>wn,SiglipPreTrainedModel:()=>_n,SiglipTextModel:()=>yn,SiglipVisionModel:()=>bn,SpeechT5ForSpeechToText:()=>Da,SpeechT5ForTextToSpeech:()=>Ra,SpeechT5HifiGan:()=>Na,SpeechT5Model:()=>La,SpeechT5PreTrainedModel:()=>Ba,SqueezeBertForMaskedLM:()=>it,SqueezeBertForQuestionAnswering:()=>st,SqueezeBertForSequenceClassification:()=>at,SqueezeBertModel:()=>rt,SqueezeBertPreTrainedModel:()=>nt,StableLmForCausalLM:()=>ds,StableLmModel:()=>us,StableLmPreTrainedModel:()=>ls,Starcoder2ForCausalLM:()=>Xa,Starcoder2Model:()=>Ha,Starcoder2PreTrainedModel:()=>Wa,Swin2SRForImageSuperResolution:()=>bi,Swin2SRModel:()=>yi,Swin2SRPreTrainedModel:()=>wi,SwinForImageClassification:()=>_i,SwinModel:()=>gi,SwinPreTrainedModel:()=>fi,T5ForConditionalGeneration:()=>mt,T5Model:()=>ht,T5PreTrainedModel:()=>pt,TableTransformerForObjectDetection:()=>oi,TableTransformerModel:()=>si,TableTransformerObjectDetectionOutput:()=>li,TableTransformerPreTrainedModel:()=>ai,TokenClassifierOutput:()=>No,TrOCRForCausalLM:()=>ja,TrOCRPreTrainedModel:()=>Va,UniSpeechForCTC:()=>fa,UniSpeechForSequenceClassification:()=>ga,UniSpeechModel:()=>ma,UniSpeechPreTrainedModel:()=>ha,UniSpeechSatForAudioFrameClassification:()=>va,UniSpeechSatForCTC:()=>ya,UniSpeechSatForSequenceClassification:()=>ba,UniSpeechSatModel:()=>wa,UniSpeechSatPreTrainedModel:()=>_a,ViTForImageClassification:()=>Cr,ViTModel:()=>Sr,ViTPreTrainedModel:()=>$r,VisionEncoderDecoderModel:()=>on,VitMatteForImageMatting:()=>zr,VitMattePreTrainedModel:()=>Fr,VitsModel:()=>rs,VitsModelOutput:()=>Wo,VitsPreTrainedModel:()=>ns,Wav2Vec2BertForCTC:()=>Ta,Wav2Vec2BertForSequenceClassification:()=>ka,Wav2Vec2BertModel:()=>Ma,Wav2Vec2BertPreTrainedModel:()=>xa,Wav2Vec2ForAudioFrameClassification:()=>oa,Wav2Vec2ForCTC:()=>aa,Wav2Vec2ForSequenceClassification:()=>sa,Wav2Vec2Model:()=>ia,Wav2Vec2PreTrainedModel:()=>ra,WavLMForAudioFrameClassification:()=>Oa,WavLMForCTC:()=>Fa,WavLMForSequenceClassification:()=>za,WavLMForXVector:()=>Ia,WavLMModel:()=>Aa,WavLMPreTrainedModel:()=>Pa,WeSpeakerResNetModel:()=>pa,WeSpeakerResNetPreTrainedModel:()=>ca,WhisperForConditionalGeneration:()=>sn,WhisperModel:()=>an,WhisperPreTrainedModel:()=>rn,XLMForQuestionAnswering:()=>Ht,XLMForSequenceClassification:()=>Ut,XLMForTokenClassification:()=>Wt,XLMModel:()=>Gt,XLMPreTrainedModel:()=>jt,XLMRobertaForMaskedLM:()=>Qt,XLMRobertaForQuestionAnswering:()=>Zt,XLMRobertaForSequenceClassification:()=>Yt,XLMRobertaForTokenClassification:()=>Jt,XLMRobertaModel:()=>Kt,XLMRobertaPreTrainedModel:()=>Xt,XLMWithLMHeadModel:()=>qt,XVectorOutput:()=>Ro,YolosForObjectDetection:()=>Wi,YolosModel:()=>Ui,YolosObjectDetectionOutput:()=>Hi,YolosPreTrainedModel:()=>qi});var r=n(/*! ./configs.js */"./src/configs.js"),i=n(/*! ./backends/onnx.js */"./src/backends/onnx.js"),a=n(/*! ./utils/dtypes.js */"./src/utils/dtypes.js"),s=n(/*! ./utils/generic.js */"./src/utils/generic.js"),o=n(/*! ./utils/core.js */"./src/utils/core.js"),l=n(/*! ./utils/hub.js */"./src/utils/hub.js"),u=n(/*! ./generation/logits_process.js */"./src/generation/logits_process.js"),d=n(/*! ./generation/configuration_utils.js */"./src/generation/configuration_utils.js"),c=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"),f=n(/*! ./env.js */"./src/env.js"),g=n(/*! ./models/whisper/generation_whisper.js */"./src/models/whisper/generation_whisper.js"),_=n(/*! ./models/whisper/common_whisper.js */"./src/models/whisper/common_whisper.js");const w=0,y=1,b=2,v=3,x=4,M=5,T=6,k=7,$=new Map,S=new Map,C=new Map;async function E(e,t,n){return Object.fromEntries(await Promise.all(Object.keys(t).map((async s=>{const{buffer:o,session_options:u}=await async function(e,t,n){let s=n.device;s&&"string"!=typeof s&&(s.hasOwnProperty(t)?s=s[t]:(console.warn(`device not specified for "${t}". Using the default device.`),s=null));const o=s??(f.apis.IS_NODE_ENV?"cpu":"wasm"),u=(0,i.deviceToExecutionProviders)(o);let d=n.dtype;"string"!=typeof d&&(d&&d.hasOwnProperty(t)?d=d[t]:(d=a.DEFAULT_DEVICE_DTYPE_MAPPING[o]??a.DATA_TYPES.fp32,console.warn(`dtype not specified for "${t}". Using the default dtype (${d}) for this device (${o}).`)));const c=d;if(!a.DEFAULT_DTYPE_SUFFIX_MAPPING.hasOwnProperty(c))throw new Error(`Invalid dtype: ${c}. Should be one of: ${Object.keys(a.DATA_TYPES).join(", ")}`);if(c===a.DATA_TYPES.fp16&&"webgpu"===o&&!await(0,a.isWebGpuFp16Supported)())throw new Error(`The device (${o}) does not support fp16.`);const p=a.DEFAULT_DTYPE_SUFFIX_MAPPING[c],h=`${n.subfolder??""}/${t}${p}.onnx`,m={...n.session_options}??{};m.executionProviders??=u;const g=(0,l.getModelFile)(e,h,!0,n);let _=[];if(n.use_external_data_format&&(!0===n.use_external_data_format||"object"==typeof n.use_external_data_format&&n.use_external_data_format.hasOwnProperty(t)&&!0===n.use_external_data_format[t])){if(f.apis.IS_NODE_ENV)throw new Error("External data format is not yet supported in Node.js");const r=`${t}${p}.onnx_data`,i=`${n.subfolder??""}/${r}`;_.push(new Promise((async(t,a)=>{const s=await(0,l.getModelFile)(e,i,!0,n);t({path:r,data:s})})))}else void 0!==m.externalData&&(_=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(_.length>0&&(m.externalData=await Promise.all(_)),"webgpu"===o){const e=(0,r.getKeyValueShapes)(n.config,{prefix:"present"});if(Object.keys(e).length>0&&!(0,i.isONNXProxy)()){const t={};for(const n in e)t[n]="gpu-buffer";m.preferredOutputLocation=t}}return{buffer:await g,session_options:m}}(e,t[s],n);return[s,await(0,i.createInferenceSession)(o,u)]}))))}async function P(e,t){const n=function(e,t){const n=Object.create(null),r=[];for(const a of e.inputNames){const e=t[a];e instanceof c.Tensor?n[a]=(0,i.isONNXProxy)()?e.clone():e:r.push(a)}if(r.length>0)throw new Error(`An error occurred during model execution: "Missing the following inputs: ${r.join(", ")}.`);const a=Object.keys(t).length,s=e.inputNames.length;if(a>s){let n=Object.keys(t).filter((t=>!e.inputNames.includes(t)));console.warn(`WARNING: Too many inputs were provided (${a} > ${s}). The following inputs will be ignored: "${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=A(r),r}catch(e){throw console.error(`An error occurred during model execution: "${e}".`),console.error("Inputs given to model:",n),e}}function A(e){for(let t in e)(0,i.isONNXTensor)(e[t])?e[t]=new c.Tensor(e[t]):"object"==typeof e[t]&&A(e[t]);return e}function F(e){if(e instanceof c.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 c.Tensor("int64",BigInt64Array.from(e.flat().map((e=>BigInt(e)))),[e.length,e[0].length])}return new c.Tensor("int64",BigInt64Array.from(e.map((e=>BigInt(e)))),[1,e.length])}function z(e){return new c.Tensor("bool",[e],[1])}async function I(e,t){let{encoder_outputs:n,input_ids:r,decoder_input_ids:i,...a}=t;if(!n){const r=(0,o.pick)(t,e.sessions.model.inputNames);n=(await O(e,r)).last_hidden_state}a.input_ids=i,a.encoder_hidden_states=n,e.sessions.decoder_model_merged.inputNames.includes("encoder_attention_mask")&&(a.encoder_attention_mask=t.attention_mask);return await B(e,a,!0)}async function O(e,t){const n=e.sessions.model,r=(0,o.pick)(t,n.inputNames);if(n.inputNames.includes("inputs_embeds")&&!r.inputs_embeds){if(!t.input_ids)throw new Error("Both `input_ids` and `inputs_embeds` are missing in the model inputs.");r.inputs_embeds=await e.encode_text({input_ids:t.input_ids})}return n.inputNames.includes("token_type_ids")&&!r.token_type_ids&&(r.token_type_ids=new c.Tensor("int64",new BigInt64Array(r.input_ids.data.length),r.input_ids.dims)),await P(n,r)}async function B(e,t,n=!1){const r=e.sessions[n?"decoder_model_merged":"model"],{past_key_values:i,...a}=t;r.inputNames.includes("use_cache_branch")&&(a.use_cache_branch=z(!!i)),r.inputNames.includes("position_ids")&&a.attention_mask&&!a.position_ids&&(a.position_ids=function(e,t=null){const{input_ids:n,inputs_embeds:r,attention_mask:i}=e,[a,s]=i.dims,o=new BigInt64Array(i.data.length);for(let e=0;e<a;++e){const t=e*s;let n=BigInt(0);for(let e=0;e<s;++e){const r=t+e;0n===i.data[r]?o[r]=BigInt(1):(o[r]=n,n+=i.data[r])}}let l=new c.Tensor("int64",o,i.dims);if(t){const e=-(n??r).dims.at(1);l=l.slice(null,[e,null])}return l}(a,i)),e.addPastKeyValues(a,i);const s=(0,o.pick)(a,r.inputNames);return await P(r,s)}async function L(e,{input_ids:t=null,attention_mask:n=null,pixel_values:r=null,position_ids:i=null,inputs_embeds:a=null,past_key_values:s=null,generation_config:o=null,logits_processor:l=null,...u}){if(!a)if(a=await e.encode_text({input_ids:t}),r&&1!==t.dims[1]){const i=await e.encode_image({pixel_values:r});({inputs_embeds:a,attention_mask:n}=e._merge_input_ids_with_image_features({image_features:i,inputs_embeds:a,input_ids:t,attention_mask:n}))}else if(s&&r&&1===t.dims[1]){const e=t.dims[1],r=Object.values(s)[0].dims.at(-2);n=(0,c.cat)([(0,c.ones)([t.dims[0],r]),n.slice(null,[n.dims[1]-e,n.dims[1]])],1)}return await B(e,{inputs_embeds:a,past_key_values:s,attention_mask:n,position_ids:i,generation_config:o,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:i}=n;if(i&&i.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 i=e.config.num_image_tokens;if(!i)throw new Error("`num_image_tokens` is missing in the model configuration.");const a=r.dims[1]-(t-i);n.input_ids=r.slice(null,[-a,null]),n.attention_mask=(0,c.ones)([1,t+a])}}return n}function R(e,t,n,r){return n.past_key_values&&(t=t.map((e=>[e.at(-1)]))),{...n,decoder_input_ids:F(t)}}function N(e,...t){return e.config.is_encoder_decoder?R(e,...t):D(e,...t)}class V extends s.Callable{main_input_name="input_ids";forward_params=["input_ids","attention_mask"];constructor(e,t){super(),this.config=e,this.sessions=t;const n=C.get(this.constructor),r=$.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=B,this._prepare_inputs_for_generation=D;break;case b:case v:case k:this.can_generate=!0,this._forward=I,this._prepare_inputs_for_generation=R;break;case y:this._forward=I;break;case T:this.can_generate=!0,this._forward=L,this._prepare_inputs_for_generation=N;break;default:this._forward=O}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:i=null,local_files_only:a=!1,revision:s="main",model_file_name:o=null,subfolder:u="onnx",device:d=null,dtype:c=null,use_external_data_format:p=null,session_options:h={}}={}){let m={progress_callback:t,config:n,cache_dir:i,local_files_only:a,revision:s,model_file_name:o,subfolder:u,device:d,dtype:c,use_external_data_format:p,session_options:h};const f=C.get(this),g=$.get(f);let _;if(n=m.config=await r.AutoConfig.from_pretrained(e,m),g===x)_=await Promise.all([E(e,{model:m.model_file_name??"model"},m),(0,l.getModelJSON)(e,"generation_config.json",!1,m)]);else if(g===b||g===v)_=await Promise.all([E(e,{model:"encoder_model",decoder_model_merged:"decoder_model_merged"},m),(0,l.getModelJSON)(e,"generation_config.json",!1,m)]);else if(g===M)_=await Promise.all([E(e,{model:"vision_encoder",prompt_encoder_mask_decoder:"prompt_encoder_mask_decoder"},m)]);else if(g===y)_=await Promise.all([E(e,{model:"encoder_model",decoder_model_merged:"decoder_model_merged"},m)]);else if(g===T){const t={embed_tokens:"embed_tokens",vision_encoder:"vision_encoder",decoder_model_merged:"decoder_model_merged"};n.is_encoder_decoder&&(t.model="encoder_model"),_=await Promise.all([E(e,t,m),(0,l.getModelJSON)(e,"generation_config.json",!1,m)])}else g===k?_=await Promise.all([E(e,{model:"text_encoder",decoder_model_merged:"decoder_model_merged",encodec_decode:"encodec_decode"},m),(0,l.getModelJSON)(e,"generation_config.json",!1,m)]):(g!==w&&console.warn(`Model type for '${f??n?.model_type}' not found, assuming encoder-only architecture. Please report this at https://github.com/xenova/transformers.js/issues/new/choose.`),_=await Promise.all([E(e,{model:m.model_file_name??"model"},m)]));return new this(n,..._)}async _call(e){return await this.forward(e)}async forward(e){return await this._forward(this,e)}_get_logits_warper(e){const t=new u.LogitsProcessorList;return null!==e.temperature&&1!==e.temperature&&t.push(new u.TemperatureLogitsWarper(e.temperature)),null!==e.top_k&&0!==e.top_k&&t.push(new u.TopKLogitsWarper(e.top_k)),null!==e.top_p&&e.top_p<1&&t.push(new u.TopPLogitsWarper(e.top_p)),t}_get_logits_processor(e,t,n=null){const r=new u.LogitsProcessorList;if(null!==e.repetition_penalty&&1!==e.repetition_penalty&&r.push(new u.RepetitionPenaltyLogitsProcessor(e.repetition_penalty)),null!==e.no_repeat_ngram_size&&e.no_repeat_ngram_size>0&&r.push(new u.NoRepeatNGramLogitsProcessor(e.no_repeat_ngram_size)),null!==e.bad_words_ids&&r.push(new u.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 u.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 u.MinNewTokensLengthLogitsProcessor(t,e.min_new_tokens,e.eos_token_id)),null!==e.forced_bos_token_id&&r.push(new u.ForcedBOSTokenLogitsProcessor(e.forced_bos_token_id)),null!==e.forced_eos_token_id&&r.push(new u.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 u.SuppressTokensAtBeginLogitsProcessor(e.begin_suppress_tokens,n))}return null!==e.guidance_scale&&e.guidance_scale>1&&r.push(new u.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 i=new n(r);return"generation_config"in this&&Object.assign(i,this.generation_config),e&&Object.assign(i,e),t&&Object.assign(i,(0,o.pick)(t,Object.getOwnPropertyNames(i))),i}_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=[Ns,Gs,Rs,Is],t=C.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 i=`The current model class (${t}) is not compatible with \`.generate()\`, as it doesn't have a language model head.`;throw n.size>0&&(i+=` Please use the following class instead: ${[...n].join(", ")}`),Error(i)}}prepare_inputs_for_generation(...e){return this._prepare_inputs_for_generation(this,...e)}_update_model_kwargs_for_generation({generated_input_ids:e,outputs:t,model_inputs:n,is_encoder_decoder:r}){return n.past_key_values=this.getPastKeyValues(t,n.past_key_values),n.input_ids=new c.Tensor("int64",e.flat(),[e.length,1]),r||(n.attention_mask=(0,c.cat)([n.attention_mask,(0,c.ones)([n.attention_mask.dims[0],1])],1)),n.position_ids=null,n}_prepare_model_inputs({inputs:e,bos_token_id:t,model_kwargs:n}){const r=(0,o.pick)(n,this.forward_params),i=this.main_input_name;if(i in r){if(e)throw new Error("`inputs`: {inputs}` were passed alongside {input_name} which is not allowed. Make sure to either pass {inputs} or {input_name}=...")}else r[i]=e;return{inputs_tensor:r[i],model_inputs:r,model_input_name:i}}async _prepare_encoder_decoder_kwargs_for_generation({inputs_tensor:e,model_inputs:t,model_input_name:n,generation_config:r}){if(this.sessions.model.inputNames.includes("inputs_embeds")&&!t.inputs_embeds&&"_prepare_inputs_embeds"in this){const{input_ids:e,pixel_values:n,attention_mask:r,...i}=t,a=await this._prepare_inputs_embeds(t);t={...i,...(0,o.pick)(a,["inputs_embeds","attention_mask"])}}let{last_hidden_state:i}=await O(this,t);if(null!==r.guidance_scale&&r.guidance_scale>1)i=(0,c.cat)([i,(0,c.full_like)(i,0)],0),"attention_mask"in t&&(t.attention_mask=(0,c.cat)([t.attention_mask,(0,c.zeros_like)(t.attention_mask)],0));else if(t.decoder_input_ids){const e=F(t.decoder_input_ids).dims[0];if(e!==i.dims[0]){if(1!==i.dims[0])throw new Error(`The encoder outputs have a different batch size (${i.dims[0]}) than the decoder inputs (${e}).`);i=(0,c.cat)(Array.from({length:e},(()=>i)),0)}}return t.encoder_outputs=i,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:i,generation_config:a}){let{decoder_input_ids:s,...o}=n;if(s)Array.isArray(s[0])||(s=Array.from({length:e},(()=>s)));else if(r??=i,"musicgen"===this.config.model_type)s=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}`);s=r}else s=Array.from({length:e},(()=>[r]));return s=F(s),n.decoder_attention_mask=(0,c.ones_like)(s),{input_ids:s,model_inputs:o}}async generate({inputs:e=null,generation_config:t=null,logits_processor:n=null,stopping_criteria:r=null,streamer:i=null,...a}){this._validate_model_class(),t=this._prepare_generation_config(t,a);let{inputs_tensor:s,model_inputs:o,model_input_name:l}=this._prepare_model_inputs({inputs:e,model_kwargs:a});const u=this.config.is_encoder_decoder;let d;u&&("encoder_outputs"in o||(o=await this._prepare_encoder_decoder_kwargs_for_generation({inputs_tensor:s,model_inputs:o,model_input_name:l,generation_config:t}))),u?({input_ids:d,model_inputs:o}=this._prepare_decoder_input_ids_for_generation({batch_size:o[l].dims.at(0),model_input_name:l,model_kwargs:o,decoder_start_token_id:t.decoder_start_token_id,bos_token_id:t.bos_token_id,generation_config:t})):d=o[l];let 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),f=this._get_stopping_criteria(t,r),g=o[l].dims.at(0),_=m.LogitsSampler.getSampler(t),w=new Array(g).fill(0),y=d.tolist();i&&i.put(y);let b=null,v={};for(;;){o=this.prepare_inputs_for_generation(y,o,t);const e=await this.forward(o);if(t.output_attentions&&t.return_dict_in_generate){const t=this.getAttentions(e);for(const e in t)e in v||(v[e]=[]),v[e].push(t[e])}const n=h(y,e.logits.slice(null,-1,null)),r=[];for(let e=0;e<n.dims.at(0);++e){const t=n[e],i=await _(t);for(const[t,n]of i){const i=BigInt(t);w[e]+=n,y[e].push(i),r.push([i]);break}}i&&i.put(r);if(f(y).every((e=>e))){t.return_dict_in_generate&&(b=this.getPastKeyValues(e,o.past_key_values,!1));break}o=this._update_model_kwargs_for_generation({generated_input_ids:r,outputs:e,model_inputs:o,is_encoder_decoder:u})}i&&i.end();const x=new c.Tensor("int64",y.flat(),[y.length,y[0].length]);return t.return_dict_in_generate?{sequences:x,past_key_values:b,...v}:x}getPastKeyValues(e,t,n=!0){const r=Object.create(null);for(const i in e)if(i.startsWith("present")){const a=i.replace("present","past_key_values");if(t&&i.includes("encoder"))r[a]=t[a];else{if(n&&t){const e=t[a];"gpu-buffer"===e.location&&e.dispose()}r[a]=e[i]}}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:[],i=(0,r.getKeyValueShapes)(this.config);for(const r in i)e[r]=new c.Tensor(t,n,i[r])}}async encode_image({pixel_values:e}){const t=(await P(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 P(this.sessions.embed_tokens,{input_ids:e})).inputs_embeds}}class j{}class G extends j{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 q extends V{}class U extends q{}class W extends q{async _call(e){return new Vo(await super._call(e))}}class H extends q{async _call(e){return new Do(await super._call(e))}}class X extends q{async _call(e){return new No(await super._call(e))}}class K extends q{async _call(e){return new jo(await super._call(e))}}class Q extends V{}class Y extends Q{}class J extends V{}class Z extends J{}class ee extends J{async _call(e){return new Vo(await super._call(e))}}class te extends J{async _call(e){return new Do(await super._call(e))}}class ne extends J{async _call(e){return new No(await super._call(e))}}class re extends J{async _call(e){return new jo(await super._call(e))}}class ie extends V{}class ae extends ie{}class se extends ie{async _call(e){return new Vo(await super._call(e))}}class oe extends ie{async _call(e){return new Do(await super._call(e))}}class le extends ie{async _call(e){return new No(await super._call(e))}}class ue extends ie{async _call(e){return new jo(await super._call(e))}}class de extends V{}class ce extends de{}class pe extends de{async _call(e){return new Vo(await super._call(e))}}class he extends de{async _call(e){return new Do(await super._call(e))}}class me extends de{async _call(e){return new No(await super._call(e))}}class fe extends de{async _call(e){return new jo(await super._call(e))}}class ge extends V{}class _e extends ge{}class we extends ge{async _call(e){return new Vo(await super._call(e))}}class ye extends ge{async _call(e){return new Do(await super._call(e))}}class be extends ge{async _call(e){return new No(await super._call(e))}}class ve extends ge{async _call(e){return new jo(await super._call(e))}}class xe extends V{}class Me extends xe{}class Te extends xe{async _call(e){return new Vo(await super._call(e))}}class ke extends xe{async _call(e){return new Do(await super._call(e))}}class $e extends xe{async _call(e){return new No(await super._call(e))}}class Se extends xe{async _call(e){return new jo(await super._call(e))}}class Ce extends V{}class Ee extends Ce{}class Pe extends Ce{async _call(e){return new Vo(await super._call(e))}}class Ae extends Ce{async _call(e){return new Do(await super._call(e))}}class Fe extends Ce{async _call(e){return new No(await super._call(e))}}class ze extends Ce{async _call(e){return new jo(await super._call(e))}}class Ie extends V{}class Oe extends Ie{}class Be extends Ie{async _call(e){return new Do(await super._call(e))}}class Le extends Ie{async _call(e){return new No(await super._call(e))}}class De extends Ie{async _call(e){return new jo(await super._call(e))}}class Re extends Ie{async _call(e){return new Vo(await super._call(e))}}class Ne extends V{}class Ve extends Ne{}class je extends Ne{async _call(e){return new Vo(await super._call(e))}}class Ge extends Ne{async _call(e){return new Do(await super._call(e))}}class qe extends Ne{async _call(e){return new No(await super._call(e))}}class Ue extends V{}class We extends Ue{}class He extends Ue{async _call(e){return new Vo(await super._call(e))}}class Xe extends Ue{async _call(e){return new Do(await super._call(e))}}class Ke extends Ue{async _call(e){return new jo(await super._call(e))}}class Qe extends V{}class Ye extends Qe{}class Je extends Qe{async _call(e){return new Vo(await super._call(e))}}class Ze extends Qe{async _call(e){return new Do(await super._call(e))}}class et extends Qe{async _call(e){return new No(await super._call(e))}}class tt extends Qe{async _call(e){return new jo(await super._call(e))}}class nt extends V{}class rt extends nt{}class it extends nt{async _call(e){return new Vo(await super._call(e))}}class at extends nt{async _call(e){return new Do(await super._call(e))}}class st extends nt{async _call(e){return new jo(await super._call(e))}}class ot extends V{}class lt extends ot{}class ut extends ot{async _call(e){return new Do(await super._call(e))}}class dt extends ot{async _call(e){return new jo(await super._call(e))}}class ct extends ot{async _call(e){return new Vo(await super._call(e))}}class pt extends V{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 ft extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class gt extends ft{}class _t extends ft{}class wt extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class yt extends wt{}class bt extends wt{}class vt extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class xt extends vt{}class Mt extends vt{}class Tt extends vt{async _call(e){return new Do(await super._call(e))}}class kt extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class $t extends kt{}class St extends kt{}class Ct extends kt{async _call(e){return new Do(await super._call(e))}}class Et extends kt{}class Pt extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class At extends Pt{}class Ft extends Pt{}class zt extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class It extends zt{}class Ot extends zt{}class Bt extends V{}class Lt extends Bt{}class Dt extends Bt{async _call(e){return new Vo(await super._call(e))}}class Rt extends Bt{async _call(e){return new Do(await super._call(e))}}class Nt extends Bt{async _call(e){return new No(await super._call(e))}}class Vt extends Bt{async _call(e){return new jo(await super._call(e))}}class jt extends V{}class Gt extends jt{}class qt extends jt{async _call(e){return new Vo(await super._call(e))}}class Ut extends jt{async _call(e){return new Do(await super._call(e))}}class Wt extends jt{async _call(e){return new No(await super._call(e))}}class Ht extends jt{async _call(e){return new jo(await super._call(e))}}class Xt extends V{}class Kt extends Xt{}class Qt extends Xt{async _call(e){return new Vo(await super._call(e))}}class Yt extends Xt{async _call(e){return new Do(await super._call(e))}}class Jt extends Xt{async _call(e){return new No(await super._call(e))}}class Zt extends Xt{async _call(e){return new jo(await super._call(e))}}class en extends V{}class tn extends en{}class nn extends en{}class rn extends V{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 an extends rn{}class sn extends rn{_prepare_generation_config(e,t){return super._prepare_generation_config(e,t,g.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 i=`<|${(0,_.whisper_language_to_code)(n)}|>`;t.push(e.lang_to_id[i]),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,...i}){t=this._prepare_generation_config(t,i);const a=i.decoder_input_ids??this._retrieve_init_tokens(t);if(t.return_timestamps&&(n??=new u.LogitsProcessorList,n.push(new u.WhisperTimeStampLogitsProcessor(t,a))),t.begin_suppress_tokens&&(n??=new u.LogitsProcessorList,n.push(new u.SuppressTokensAtBeginLogitsProcessor(t.begin_suppress_tokens,a.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 s=await super.generate({inputs:e,generation_config:t,logits_processor:n,decoder_input_ids:a,...i});return t.return_token_timestamps&&(s.token_timestamps=this._extract_token_timestamps(s,t.alignment_heads,t.num_frames)),s}_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 i=this.config.median_filter_width;void 0===i&&(console.warn("Model config has no `median_filter_width`, using default value of 7."),i=7);const a=e.cross_attentions,s=Array.from({length:this.config.decoder_layers},((e,t)=>(0,c.cat)(a.map((e=>e[t])),2))),l=(0,c.stack)(t.map((([e,t])=>{if(e>=s.length)throw new Error(`Layer index ${e} is out of bounds for cross attentions (length ${s.length}).`);return n?s[e].slice(null,t,null,[0,n]):s[e].slice(null,t)}))).transpose(1,0,2,3),[u,d]=(0,c.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],a=u[e][n][0].data,s=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]-s[e])/a[e];t.set((0,p.medianFilter)(t,i))}}}const m=[(0,c.mean)(h,1)],f=e.sequences.dims,g=new c.Tensor("float32",new Float32Array(f[0]*f[1]),f);for(let e=0;e<f[0];++e){const t=m[e].neg().squeeze_(0),[n,i]=(0,p.dynamic_time_warping)(t.tolist()),a=Array.from({length:n.length-1},((e,t)=>n[t+1]-n[t])),s=(0,o.mergeArrays)([1],a).map((e=>!!e)),l=[];for(let e=0;e<s.length;++e)s[e]&&l.push(i[e]*r);g[e].data.set(l,1)}return g}}class on extends V{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 V{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 un extends ln{_merge_input_ids_with_image_features({inputs_embeds:e,image_features:t,input_ids:n,attention_mask:r}){const i=this.config.image_token_index,a=n.tolist().map((e=>e.findIndex((e=>e==i)))),s=a.every((e=>-1===e)),o=a.every((e=>-1!==e));if(!s&&!o)throw new Error("Every input should contain either 0 or 1 image token.");if(s)return{inputs_embeds:e,attention_mask:r};const l=[],u=[];for(let n=0;n<a.length;++n){const i=a[n],s=e[n],o=t[n],d=r[n];l.push((0,c.cat)([s.slice([0,i]),o,s.slice([i+1,s.dims[0]])],0)),u.push((0,c.cat)([d.slice([0,i]),(0,c.ones)([o.dims[0]]),d.slice([i+1,d.dims[0]])],0))}return{inputs_embeds:(0,c.stack)(l,0),attention_mask:(0,c.stack)(u,0)}}}class dn extends un{}class cn extends V{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 cn{_merge_input_ids_with_image_features({inputs_embeds:e,image_features:t,input_ids:n,attention_mask:r}){return{inputs_embeds:(0,c.cat)([t,e],1),attention_mask:(0,c.cat)([(0,c.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 i,a;return e&&(i=await this.encode_text({input_ids:e})),t&&(a=await this.encode_image({pixel_values:t})),i&&a?({inputs_embeds:n,attention_mask:r}=this._merge_input_ids_with_image_features({inputs_embeds:i,image_features:a,input_ids:e,attention_mask:r})):n=i||a,{inputs_embeds:n,attention_mask:r}}async forward({input_ids:e,pixel_values:t,attention_mask:n,decoder_input_ids:r,decoder_attention_mask:i,encoder_outputs:a,past_key_values:s,inputs_embeds:o,decoder_inputs_embeds:l}){if(o||({inputs_embeds:o,attention_mask:n}=await this._prepare_inputs_embeds({input_ids:e,pixel_values:t,inputs_embeds:o,attention_mask:n})),!a){let{last_hidden_state:e}=await O(this,{inputs_embeds:o,attention_mask:n});a=e}if(!l){if(!r)throw new Error("Either `decoder_input_ids` or `decoder_inputs_embeds` should be provided.");l=await this.encode_text({input_ids:r})}const u={inputs_embeds:l,attention_mask:i,encoder_attention_mask:n,encoder_hidden_states:a,past_key_values:s};return await B(this,u,!0)}}class hn extends V{}class mn extends hn{}class fn extends hn{static async from_pretrained(e,t={}){return t.model_file_name??="text_model",super.from_pretrained(e,t)}}class gn extends hn{static async from_pretrained(e,t={}){return t.model_file_name??="vision_model",super.from_pretrained(e,t)}}class _n extends V{}class wn extends _n{}class yn extends _n{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 vn extends V{}class xn extends vn{}class Mn extends V{}class Tn extends Mn{}class kn extends Mn{}class $n extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class Sn extends $n{}class Cn extends $n{}class En extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class Pn extends En{}class An extends En{}class Fn extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class zn extends Fn{}class In extends Fn{}class On extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class Bn extends On{}class Ln extends On{}class Dn extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class Rn extends Dn{}class Nn extends Dn{}class Vn extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class jn extends Vn{}class Gn extends Vn{}class qn extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class Un extends qn{}class Wn extends qn{}class Hn extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class Xn extends Hn{}class Kn extends Hn{}class Qn extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class Yn extends Qn{}class Jn extends Qn{}class Zn extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class er extends Zn{}class tr extends Zn{}class nr extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class rr extends nr{}class ir extends nr{}class ar extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class sr extends ar{}class or extends ar{}class lr extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class ur extends lr{}class dr extends lr{}class cr extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class pr extends cr{}class hr extends cr{}class mr extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class fr extends mr{}class gr extends mr{}class _r extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class wr extends _r{}class yr extends _r{}class br extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class vr extends br{}class xr extends br{}class Mr extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class Tr extends Mr{}class kr extends Mr{}class $r extends V{}class Sr extends $r{}class Cr extends $r{async _call(e){return new Do(await super._call(e))}}class Er extends V{}class Pr extends Er{}class Ar extends Er{async _call(e){return new Do(await super._call(e))}}class Fr extends V{}class zr extends Fr{async _call(e){return new Uo(await super._call(e))}}class Ir extends V{}class Or extends Ir{}class Br extends Ir{async _call(e){return new Do(await super._call(e))}}class Lr extends V{}class Dr extends Lr{}class Rr extends Lr{async _call(e){return new Do(await super._call(e))}}class Nr extends V{}class Vr extends Nr{}class jr extends Nr{}class Gr extends V{}class qr extends Gr{}class Ur extends Gr{}class Wr extends V{}class Hr extends Wr{}class Xr extends Wr{async _call(e){return new Do(await super._call(e))}}class Kr extends V{}class Qr extends Kr{}class Yr extends Kr{async _call(e){return new Zr(await super._call(e))}}class Jr extends Kr{async _call(e){return new ei(await super._call(e))}}class Zr extends j{constructor({logits:e,pred_boxes:t}){super(),this.logits=e,this.pred_boxes=t}}class ei extends j{constructor({logits:e,pred_boxes:t,pred_masks:n}){super(),this.logits=e,this.pred_boxes=t,this.pred_masks=n}}class ti extends V{}class ni extends ti{}class ri extends ti{async _call(e){return new ii(await super._call(e))}}class ii extends j{constructor({logits:e,pred_boxes:t}){super(),this.logits=e,this.pred_boxes=t}}class ai extends V{}class si extends ai{}class oi extends ai{async _call(e){return new li(await super._call(e))}}class li extends Zr{}class ui extends V{}class di extends ui{}class ci extends ui{async _call(e){return new Do(await super._call(e))}}class pi extends V{}class hi extends pi{}class mi extends pi{async _call(e){return new Do(await super._call(e))}}class fi extends V{}class gi extends fi{}class _i extends fi{async _call(e){return new Do(await super._call(e))}}class wi extends V{}class yi extends wi{}class bi extends wi{}class vi extends V{}class xi extends vi{}class Mi extends vi{}class Ti extends V{}class ki extends Ti{}class $i extends V{}class Si extends $i{}class Ci extends $i{}class Ei extends $i{}class Pi extends V{}class Ai extends Pi{}class Fi extends Pi{}class zi extends V{}class Ii extends zi{}class Oi extends V{}class Bi extends Oi{}class Li extends Oi{async _call(e){return new Do(await super._call(e))}}class Di extends V{}class Ri extends Di{}class Ni extends Di{async _call(e){return new Do(await super._call(e))}}class Vi extends V{}class ji extends Vi{}class Gi extends Vi{async _call(e){return new Do(await super._call(e))}}class qi extends V{}class Ui extends qi{}class Wi extends qi{async _call(e){return new Hi(await super._call(e))}}class Hi extends j{constructor({logits:e,pred_boxes:t}){super(),this.logits=e,this.pred_boxes=t}}class Xi extends V{}class Ki extends Xi{async get_image_embeddings({pixel_values:e}){return await O(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 c.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 P(this.sessions.prompt_encoder_mask_decoder,t)}async _call(e){return new Qi(await super._call(e))}}class Qi extends j{constructor({iou_scores:e,pred_masks:t}){super(),this.iou_scores=e,this.pred_masks=t}}class Yi extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class Ji extends Yi{}class Zi extends Yi{}class ea extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class ta extends ea{}class na extends ea{}class ra extends V{}class ia extends ra{}class aa extends ra{async _call(e){return new Go(await super._call(e))}}class sa extends ra{async _call(e){return new Do(await super._call(e))}}class oa extends ra{async _call(e){return new No(await super._call(e))}}class la extends V{}class ua extends la{}class da extends la{async _call(e){return new No(await super._call(e))}}class ca extends V{}class pa extends ca{}class ha extends V{}class ma extends ha{}class fa extends ha{async _call(e){return new Go(await super._call(e))}}class ga extends ha{async _call(e){return new Do(await super._call(e))}}class _a extends V{}class wa extends _a{}class ya extends _a{async _call(e){return new Go(await super._call(e))}}class ba extends _a{async _call(e){return new Do(await super._call(e))}}class va extends _a{async _call(e){return new No(await super._call(e))}}class xa extends V{}class Ma extends xa{}class Ta extends xa{async _call(e){return new Go(await super._call(e))}}class ka extends xa{async _call(e){return new Do(await super._call(e))}}class $a extends V{}class Sa extends ra{}class Ca extends ra{async _call(e){return new Go(await super._call(e))}}class Ea extends ra{async _call(e){return new Do(await super._call(e))}}class Pa extends V{}class Aa extends Pa{}class Fa extends Pa{async _call(e){return new Go(await super._call(e))}}class za extends Pa{async _call(e){return new Do(await super._call(e))}}class Ia extends Pa{async _call(e){return new Ro(await super._call(e))}}class Oa extends Pa{async _call(e){return new No(await super._call(e))}}class Ba extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class La extends Ba{}class Da extends Ba{}class Ra extends Ba{async generate_speech(e,t,{threshold:n=.5,minlenratio:r=0,maxlenratio:i=20,vocoder:a=null}={}){const s={input_ids:e},{encoder_outputs:o,encoder_attention_mask:l}=await O(this,s),u=o.dims[1]/this.config.reduction_factor,d=Math.floor(u*i),p=Math.floor(u*r),h=this.config.num_mel_bins;let m=[],f=null,g=null,_=0;for(;;){++_;const e=z(!!g);let r;r=g?g.output_sequence_out:new c.Tensor("float32",new Float32Array(h),[1,1,h]);let i={use_cache_branch:e,output_sequence:r,encoder_attention_mask:l,speaker_embeddings:t,encoder_hidden_states:o};this.addPastKeyValues(i,f),g=await P(this.sessions.decoder_model_merged,i),f=this.getPastKeyValues(g,f);const{prob:a,spectrum:s}=g;if(m.push(s),_>=p&&(Array.from(a.data).filter((e=>e>=n)).length>0||_>=d))break}const w=(0,c.cat)(m),{waveform:y}=await P(a.sessions.model,{spectrogram:w});return{spectrogram:w,waveform:y}}}class Na extends V{main_input_name="spectrogram"}class Va extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class ja extends Va{}class Ga extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class qa extends Ga{}class Ua extends Ga{}class Wa extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class Ha extends Wa{}class Xa extends Wa{}class Ka extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class Qa extends Ka{}class Ya extends Ka{}class Ja extends V{}class Za extends Ja{}class es extends Ja{static async from_pretrained(e,t={}){return t.model_file_name??="text_model",super.from_pretrained(e,t)}}class ts extends Ja{static async from_pretrained(e,t={}){return t.model_file_name??="audio_model",super.from_pretrained(e,t)}}class ns extends V{}class rs extends ns{async _call(e){return new Wo(await super._call(e))}}class is extends V{}class as extends is{}class ss extends is{}class os extends is{}class ls extends V{constructor(e,t,n){super(e,t),this.generation_config=n}}class us extends ls{}class ds extends ls{}class cs extends V{}class ps extends cs{}class hs extends cs{async _call(e){return new Do(await super._call(e))}}class ms extends V{}class fs extends ms{}class gs extends ms{}class _s extends V{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,i=n-r;let a=0;for(let t=0;t<e.size;++t){if(e.data[t]===this.config.decoder.pad_token_id)continue;const s=t%n-Math.floor(t/n)%r;s>0&&s<=i&&(e.data[a++]=e.data[t])}const s=Math.floor(t/r),o=a/(s*r);return new c.Tensor(e.type,e.data.slice(0,a),[s,r,o])}prepare_inputs_for_generation(e,t,n){let r=structuredClone(e);for(let e=0;e<r.length;++e)for(let t=0;t<r[e].length;++t)e%this.config.decoder.num_codebooks>=t&&(r[e][t]=BigInt(this.config.decoder.pad_token_id));null!==n.guidance_scale&&n.guidance_scale>1&&(r=r.concat(r));return super.prepare_inputs_for_generation(r,t,n)}async generate(e){const t=await super.generate(e),n=this._apply_and_filter_by_delay_pattern_mask(t).unsqueeze_(0),{audio_values:r}=await P(this.sessions.encodec_decode,{audio_codes:n});return r}}class ws extends V{}class ys extends ws{}class bs extends ws{async _call(e){return new Do(await super._call(e))}}class vs extends V{}class xs extends vs{}class Ms extends vs{async _call(e){return new Do(await super._call(e))}}class Ts extends V{}class ks extends Ts{}class $s extends Ts{async _call(e){return new Do(await super._call(e))}}class Ss extends V{}class Cs extends Ss{}class Es extends Ss{async _call(e){return new Do(await super._call(e))}}class Ps{static MODEL_CLASS_MAPPINGS=null;static BASE_IF_FAIL=!1;static async from_pretrained(e,{progress_callback:t=null,config:n=null,cache_dir:i=null,local_files_only:a=!1,revision:s="main",model_file_name:o=null,subfolder:l="onnx",device:u=null,dtype:d=null,use_external_data_format:c=null,session_options:p={}}={}){let h={progress_callback:t,config:n,cache_dir:i,local_files_only:a,revision:s,model_file_name:o,subfolder:l,device:u,dtype:d,use_external_data_format:c,session_options:p};if(h.config=await r.AutoConfig.from_pretrained(e,h),!this.MODEL_CLASS_MAPPINGS)throw new Error("`MODEL_CLASS_MAPPINGS` not implemented for this type of `AutoClass`: "+this.name);for(let t of this.MODEL_CLASS_MAPPINGS){const n=t.get(h.config.model_type);if(n)return await n[1].from_pretrained(e,h)}if(this.BASE_IF_FAIL)return console.warn(`Unknown model class "${h.config.model_type}", attempting to construct from base class.`),await V.from_pretrained(e,h);throw Error(`Unsupported model type: ${h.config.model_type}`)}}const As=new Map([["bert",["BertModel",U]],["nomic_bert",["NomicBertModel",Y]],["roformer",["RoFormerModel",Z]],["electra",["ElectraModel",ce]],["esm",["EsmModel",Ve]],["convbert",["ConvBertModel",ae]],["camembert",["CamembertModel",_e]],["deberta",["DebertaModel",Me]],["deberta-v2",["DebertaV2Model",Ee]],["mpnet",["MPNetModel",Ye]],["albert",["AlbertModel",lt]],["distilbert",["DistilBertModel",Oe]],["roberta",["RobertaModel",Lt]],["xlm",["XLMModel",Gt]],["xlm-roberta",["XLMRobertaModel",Kt]],["clap",["ClapModel",Za]],["clip",["CLIPModel",mn]],["clipseg",["CLIPSegModel",Tn]],["chinese_clip",["ChineseCLIPModel",xn]],["siglip",["SiglipModel",wn]],["mobilebert",["MobileBertModel",We]],["squeezebert",["SqueezeBertModel",rt]],["wav2vec2",["Wav2Vec2Model",ia]],["wav2vec2-bert",["Wav2Vec2BertModel",Ma]],["unispeech",["UniSpeechModel",ma]],["unispeech-sat",["UniSpeechSatModel",wa]],["hubert",["HubertModel",Sa]],["wavlm",["WavLMModel",Aa]],["audio-spectrogram-transformer",["ASTModel",tn]],["vits",["VitsModel",rs]],["pyannote",["PyAnnoteModel",ua]],["wespeaker-resnet",["WeSpeakerResNetModel",pa]],["detr",["DetrModel",Qr]],["rt_detr",["RTDetrModel",ni]],["table-transformer",["TableTransformerModel",si]],["vit",["ViTModel",Sr]],["fastvit",["FastViTModel",Pr]],["mobilevit",["MobileViTModel",Or]],["mobilevitv2",["MobileViTV2Model",Dr]],["owlvit",["OwlViTModel",Vr]],["owlv2",["Owlv2Model",qr]],["beit",["BeitModel",Hr]],["deit",["DeiTModel",di]],["convnext",["ConvNextModel",Bi]],["convnextv2",["ConvNextV2Model",Ri]],["dinov2",["Dinov2Model",ji]],["resnet",["ResNetModel",hi]],["swin",["SwinModel",gi]],["swin2sr",["Swin2SRModel",yi]],["donut-swin",["DonutSwinModel",Ii]],["yolos",["YolosModel",Ui]],["dpt",["DPTModel",xi]],["glpn",["GLPNModel",Ai]],["hifigan",["SpeechT5HifiGan",Na]],["efficientnet",["EfficientNetModel",ps]],["mobilenet_v1",["MobileNetV1Model",ys]],["mobilenet_v2",["MobileNetV2Model",xs]],["mobilenet_v3",["MobileNetV3Model",ks]],["mobilenet_v4",["MobileNetV4Model",Cs]]]),Fs=new Map([["t5",["T5Model",ht]],["longt5",["LongT5Model",gt]],["mt5",["MT5Model",yt]],["bart",["BartModel",xt]],["mbart",["MBartModel",$t]],["marian",["MarianModel",Ji]],["whisper",["WhisperModel",an]],["m2m_100",["M2M100Model",ta]],["blenderbot",["BlenderbotModel",At]],["blenderbot-small",["BlenderbotSmallModel",It]]]),zs=new Map([["bloom",["BloomModel",wr]],["jais",["JAISModel",Pn]],["gpt2",["GPT2Model",Sn]],["gptj",["GPTJModel",Rn]],["gpt_bigcode",["GPTBigCodeModel",jn]],["gpt_neo",["GPTNeoModel",zn]],["gpt_neox",["GPTNeoXModel",Bn]],["codegen",["CodeGenModel",Un]],["llama",["LlamaModel",Xn]],["cohere",["CohereModel",Yn]],["gemma",["GemmaModel",er]],["gemma2",["Gemma2Model",rr]],["openelm",["OpenELMModel",sr]],["qwen2",["Qwen2Model",ur]],["phi",["PhiModel",pr]],["phi3",["Phi3Model",fr]],["mpt",["MptModel",vr]],["opt",["OPTModel",Tr]],["mistral",["MistralModel",qa]],["starcoder2",["Starcoder2Model",Ha]],["falcon",["FalconModel",Qa]],["stablelm",["StableLmModel",us]]]),Is=new Map([["speecht5",["SpeechT5ForSpeechToText",Da]],["whisper",["WhisperForConditionalGeneration",sn]]]),Os=new Map([["speecht5",["SpeechT5ForTextToSpeech",Ra]]]),Bs=new Map([["vits",["VitsModel",rs]],["musicgen",["MusicgenForConditionalGeneration",_s]]]),Ls=new Map([["bert",["BertForSequenceClassification",H]],["roformer",["RoFormerForSequenceClassification",te]],["electra",["ElectraForSequenceClassification",he]],["esm",["EsmForSequenceClassification",Ge]],["convbert",["ConvBertForSequenceClassification",oe]],["camembert",["CamembertForSequenceClassification",ye]],["deberta",["DebertaForSequenceClassification",ke]],["deberta-v2",["DebertaV2ForSequenceClassification",Ae]],["mpnet",["MPNetForSequenceClassification",Ze]],["albert",["AlbertForSequenceClassification",ut]],["distilbert",["DistilBertForSequenceClassification",Be]],["roberta",["RobertaForSequenceClassification",Rt]],["xlm",["XLMForSequenceClassification",Ut]],["xlm-roberta",["XLMRobertaForSequenceClassification",Yt]],["bart",["BartForSequenceClassification",Tt]],["mbart",["MBartForSequenceClassification",Ct]],["mobilebert",["MobileBertForSequenceClassification",Xe]],["squeezebert",["SqueezeBertForSequenceClassification",at]]]),Ds=new Map([["bert",["BertForTokenClassification",X]],["roformer",["RoFormerForTokenClassification",ne]],["electra",["ElectraForTokenClassification",me]],["esm",["EsmForTokenClassification",qe]],["convbert",["ConvBertForTokenClassification",le]],["camembert",["CamembertForTokenClassification",be]],["deberta",["DebertaForTokenClassification",$e]],["deberta-v2",["DebertaV2ForTokenClassification",Fe]],["mpnet",["MPNetForTokenClassification",et]],["distilbert",["DistilBertForTokenClassification",Le]],["roberta",["RobertaForTokenClassification",Nt]],["xlm",["XLMForTokenClassification",Wt]],["xlm-roberta",["XLMRobertaForTokenClassification",Jt]]]),Rs=new Map([["t5",["T5ForConditionalGeneration",mt]],["longt5",["LongT5ForConditionalGeneration",_t]],["mt5",["MT5ForConditionalGeneration",bt]],["bart",["BartForConditionalGeneration",Mt]],["mbart",["MBartForConditionalGeneration",St]],["marian",["MarianMTModel",Zi]],["m2m_100",["M2M100ForConditionalGeneration",na]],["blenderbot",["BlenderbotForConditionalGeneration",Ft]],["blenderbot-small",["BlenderbotSmallForConditionalGeneration",Ot]]]),Ns=new Map([["bloom",["BloomForCausalLM",yr]],["gpt2",["GPT2LMHeadModel",Cn]],["jais",["JAISLMHeadModel",An]],["gptj",["GPTJForCausalLM",Nn]],["gpt_bigcode",["GPTBigCodeForCausalLM",Gn]],["gpt_neo",["GPTNeoForCausalLM",In]],["gpt_neox",["GPTNeoXForCausalLM",Ln]],["codegen",["CodeGenForCausalLM",Wn]],["llama",["LlamaForCausalLM",Kn]],["cohere",["CohereForCausalLM",Jn]],["gemma",["GemmaForCausalLM",tr]],["gemma2",["Gemma2ForCausalLM",ir]],["openelm",["OpenELMForCausalLM",or]],["qwen2",["Qwen2ForCausalLM",dr]],["phi",["PhiForCausalLM",hr]],["phi3",["Phi3ForCausalLM",gr]],["mpt",["MptForCausalLM",xr]],["opt",["OPTForCausalLM",kr]],["mbart",["MBartForCausalLM",Et]],["mistral",["MistralForCausalLM",Ua]],["starcoder2",["Starcoder2ForCausalLM",Xa]],["falcon",["FalconForCausalLM",Ya]],["trocr",["TrOCRForCausalLM",ja]],["stablelm",["StableLmForCausalLM",ds]]]),Vs=new Map([["bert",["BertForMaskedLM",W]],["roformer",["RoFormerForMaskedLM",ee]],["electra",["ElectraForMaskedLM",pe]],["esm",["EsmForMaskedLM",je]],["convbert",["ConvBertForMaskedLM",se]],["camembert",["CamembertForMaskedLM",we]],["deberta",["DebertaForMaskedLM",Te]],["deberta-v2",["DebertaV2ForMaskedLM",Pe]],["mpnet",["MPNetForMaskedLM",Je]],["albert",["AlbertForMaskedLM",ct]],["distilbert",["DistilBertForMaskedLM",Re]],["roberta",["RobertaForMaskedLM",Dt]],["xlm",["XLMWithLMHeadModel",qt]],["xlm-roberta",["XLMRobertaForMaskedLM",Qt]],["mobilebert",["MobileBertForMaskedLM",He]],["squeezebert",["SqueezeBertForMaskedLM",it]]]),js=new Map([["bert",["BertForQuestionAnswering",K]],["roformer",["RoFormerForQuestionAnswering",re]],["electra",["ElectraForQuestionAnswering",fe]],["convbert",["ConvBertForQuestionAnswering",ue]],["camembert",["CamembertForQuestionAnswering",ve]],["deberta",["DebertaForQuestionAnswering",Se]],["deberta-v2",["DebertaV2ForQuestionAnswering",ze]],["mpnet",["MPNetForQuestionAnswering",tt]],["albert",["AlbertForQuestionAnswering",dt]],["distilbert",["DistilBertForQuestionAnswering",De]],["roberta",["RobertaForQuestionAnswering",Vt]],["xlm",["XLMForQuestionAnswering",Ht]],["xlm-roberta",["XLMRobertaForQuestionAnswering",Zt]],["mobilebert",["MobileBertForQuestionAnswering",Ke]],["squeezebert",["SqueezeBertForQuestionAnswering",st]]]),Gs=new Map([["vision-encoder-decoder",["VisionEncoderDecoderModel",on]]]),qs=new Map([["llava",["LlavaForConditionalGeneration",un]],["moondream1",["Moondream1ForConditionalGeneration",dn]],["florence2",["Florence2ForConditionalGeneration",pn]]]),Us=new Map([["vision-encoder-decoder",["VisionEncoderDecoderModel",on]]]),Ws=new Map([["vit",["ViTForImageClassification",Cr]],["fastvit",["FastViTForImageClassification",Ar]],["mobilevit",["MobileViTForImageClassification",Br]],["mobilevitv2",["MobileViTV2ForImageClassification",Rr]],["beit",["BeitForImageClassification",Xr]],["deit",["DeiTForImageClassification",ci]],["convnext",["ConvNextForImageClassification",Li]],["convnextv2",["ConvNextV2ForImageClassification",Ni]],["dinov2",["Dinov2ForImageClassification",Gi]],["resnet",["ResNetForImageClassification",mi]],["swin",["SwinForImageClassification",_i]],["segformer",["SegformerForImageClassification",ss]],["efficientnet",["EfficientNetForImageClassification",hs]],["mobilenet_v1",["MobileNetV1ForImageClassification",bs]],["mobilenet_v2",["MobileNetV2ForImageClassification",Ms]],["mobilenet_v3",["MobileNetV3ForImageClassification",$s]],["mobilenet_v4",["MobileNetV4ForImageClassification",Es]]]),Hs=new Map([["detr",["DetrForObjectDetection",Yr]],["rt_detr",["RTDetrForObjectDetection",ri]],["table-transformer",["TableTransformerForObjectDetection",oi]],["yolos",["YolosForObjectDetection",Wi]]]),Xs=new Map([["owlvit",["OwlViTForObjectDetection",jr]],["owlv2",["Owlv2ForObjectDetection",Ur]]]),Ks=new Map([["detr",["DetrForSegmentation",Jr]],["clipseg",["CLIPSegForImageSegmentation",kn]]]),Qs=new Map([["segformer",["SegformerForSemanticSegmentation",os]],["sapiens",["SapiensForSemanticSegmentation",Si]]]),Ys=new Map([["sam",["SamModel",Ki]]]),Js=new Map([["wav2vec2",["Wav2Vec2ForCTC",aa]],["wav2vec2-bert",["Wav2Vec2BertForCTC",Ta]],["unispeech",["UniSpeechForCTC",fa]],["unispeech-sat",["UniSpeechSatForCTC",ya]],["wavlm",["WavLMForCTC",Fa]],["hubert",["HubertForCTC",Ca]]]),Zs=new Map([["wav2vec2",["Wav2Vec2ForSequenceClassification",sa]],["wav2vec2-bert",["Wav2Vec2BertForSequenceClassification",ka]],["unispeech",["UniSpeechForSequenceClassification",ga]],["unispeech-sat",["UniSpeechSatForSequenceClassification",ba]],["wavlm",["WavLMForSequenceClassification",za]],["hubert",["HubertForSequenceClassification",Ea]],["audio-spectrogram-transformer",["ASTForAudioClassification",nn]]]),eo=new Map([["wavlm",["WavLMForXVector",Ia]]]),to=new Map([["unispeech-sat",["UniSpeechSatForAudioFrameClassification",va]],["wavlm",["WavLMForAudioFrameClassification",Oa]],["wav2vec2",["Wav2Vec2ForAudioFrameClassification",oa]],["pyannote",["PyAnnoteForAudioFrameClassification",da]]]),no=new Map([["vitmatte",["VitMatteForImageMatting",zr]]]),ro=new Map([["swin2sr",["Swin2SRForImageSuperResolution",bi]]]),io=new Map([["dpt",["DPTForDepthEstimation",Mi]],["depth_anything",["DepthAnythingForDepthEstimation",ki]],["glpn",["GLPNForDepthEstimation",Fi]],["sapiens",["SapiensForDepthEstimation",Ci]]]),ao=new Map([["sapiens",["SapiensForNormalEstimation",Ei]]]),so=new Map([["clip",["CLIPVisionModelWithProjection",gn]],["siglip",["SiglipVisionModel",bn]]]),oo=[[As,w],[Fs,y],[zs,x],[Ls,w],[Ds,w],[Rs,b],[Is,b],[Ns,x],[Vs,w],[js,w],[Gs,v],[qs,T],[Ws,w],[Ks,w],[Qs,w],[no,w],[ro,w],[io,w],[ao,w],[Hs,w],[Xs,w],[Ys,M],[Js,w],[Zs,w],[Os,b],[Bs,w],[eo,w],[to,w],[so,w]];for(const[e,t]of oo)for(const[n,r]of e.values())$.set(n,t),C.set(r,n),S.set(n,r);const lo=[["MusicgenForConditionalGeneration",_s,k],["CLIPTextModelWithProjection",fn,w],["SiglipTextModel",yn,w],["ClapTextModelWithProjection",es,w],["ClapAudioModelWithProjection",ts,w]];for(const[e,t,n]of lo)$.set(e,n),C.set(t,e),S.set(e,t);class uo extends Ps{static MODEL_CLASS_MAPPINGS=oo.map((e=>e[0]));static BASE_IF_FAIL=!0}class co extends Ps{static MODEL_CLASS_MAPPINGS=[Ls]}class po extends Ps{static MODEL_CLASS_MAPPINGS=[Ds]}class ho extends Ps{static MODEL_CLASS_MAPPINGS=[Rs]}class mo extends Ps{static MODEL_CLASS_MAPPINGS=[Is]}class fo extends Ps{static MODEL_CLASS_MAPPINGS=[Os]}class go extends Ps{static MODEL_CLASS_MAPPINGS=[Bs]}class _o extends Ps{static MODEL_CLASS_MAPPINGS=[Ns]}class wo extends Ps{static MODEL_CLASS_MAPPINGS=[Vs]}class yo extends Ps{static MODEL_CLASS_MAPPINGS=[js]}class bo extends Ps{static MODEL_CLASS_MAPPINGS=[Gs]}class vo extends Ps{static MODEL_CLASS_MAPPINGS=[Ws]}class xo extends Ps{static MODEL_CLASS_MAPPINGS=[Ks]}class Mo extends Ps{static MODEL_CLASS_MAPPINGS=[Qs]}class To extends Ps{static MODEL_CLASS_MAPPINGS=[Hs]}class ko extends Ps{static MODEL_CLASS_MAPPINGS=[Xs]}class $o extends Ps{static MODEL_CLASS_MAPPINGS=[Ys]}class So extends Ps{static MODEL_CLASS_MAPPINGS=[Js]}class Co extends Ps{static MODEL_CLASS_MAPPINGS=[Zs]}class Eo extends Ps{static MODEL_CLASS_MAPPINGS=[eo]}class Po extends Ps{static MODEL_CLASS_MAPPINGS=[to]}class Ao extends Ps{static MODEL_CLASS_MAPPINGS=[Us]}class Fo extends Ps{static MODEL_CLASS_MAPPINGS=[no]}class zo extends Ps{static MODEL_CLASS_MAPPINGS=[ro]}class Io extends Ps{static MODEL_CLASS_MAPPINGS=[io]}class Oo extends Ps{static MODEL_CLASS_MAPPINGS=[ao]}class Bo extends Ps{static MODEL_CLASS_MAPPINGS=[so]}class Lo extends j{constructor({logits:e,past_key_values:t,encoder_outputs:n,decoder_attentions:r=null,cross_attentions:i=null}){super(),this.logits=e,this.past_key_values=t,this.encoder_outputs=n,this.decoder_attentions=r,this.cross_attentions=i}}class Do extends j{constructor({logits:e}){super(),this.logits=e}}class Ro extends j{constructor({logits:e,embeddings:t}){super(),this.logits=e,this.embeddings=t}}class No extends j{constructor({logits:e}){super(),this.logits=e}}class Vo extends j{constructor({logits:e}){super(),this.logits=e}}class jo extends j{constructor({start_logits:e,end_logits:t}){super(),this.start_logits=e,this.end_logits=t}}class Go extends j{constructor({logits:e}){super(),this.logits=e}}class qo extends j{constructor({logits:e,past_key_values:t}){super(),this.logits=e,this.past_key_values=t}}class Uo extends j{constructor({alphas:e}){super(),this.alphas=e}}class Wo extends j{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:()=>i,WHISPER_TO_LANGUAGE_CODE_MAPPING:()=>a,whisper_language_to_code:()=>s});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"]],i=new Map(r),a=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 s(e){e=e.toLowerCase();let t=a.get(e);if(void 0===t){if(!i.has(e)){const t=2===e.length?i.keys():i.values();throw new Error(`Language "${e}" is not supported. Must be one of: ${JSON.stringify(t)}`)}t=e}return t}},"./src/models/whisper/generation_whisper.js":
@@ -223,8 +223,8 @@ var r,i,a,s,o,l,u,d,c,p,h,m,f,g,_,w,y,b,v,x,M,T,k,$,S,C,E,P,A,F,z,I,O,B=Object.d
   \****************************/(e,t,n)=>{function r(e,[t,n,r],[i,a],s="bilinear",o=!1){const l=a/r,u=i/n,d=new e.constructor(i*a*t),c=n*r,p=i*a;for(let s=0;s<i;++s)for(let i=0;i<a;++i){const o=s*a+i,h=(i+.5)/l-.5,m=(s+.5)/u-.5;let f=Math.floor(h),g=Math.floor(m);const _=Math.min(f+1,r-1),w=Math.min(g+1,n-1);f=Math.max(f,0),g=Math.max(g,0);const y=h-f,b=m-g,v=(1-y)*(1-b),x=y*(1-b),M=(1-y)*b,T=y*b,k=g*r,$=w*r,S=k+f,C=k+_,E=$+f,P=$+_;for(let n=0;n<t;++n){const t=n*c;d[n*p+o]=v*e[t+S]+x*e[t+C]+M*e[t+E]+T*e[t+P]}}return d}function i(e,t,n){const r=new Array(n.length),i=new Array(n.length);for(let e=n.length-1,a=1;e>=0;--e)i[e]=a,r[e]=t[n[e]],a*=r[e];const a=n.map(((e,t)=>i[n.indexOf(t)])),s=new e.constructor(e.length);for(let n=0;n<e.length;++n){let r=0;for(let e=t.length-1,i=n;e>=0;--e)r+=i%t[e]*a[e],i=Math.floor(i/t[e]);s[r]=e[n]}return[s,r]}function a(e){const t=c(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 s(e){return a(e).map((e=>Math.log(e)))}function o(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 o(e,t)/(u(e)*u(t))}function u(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 c(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:()=>f,bankers_round:()=>w,cos_sim:()=>l,dot:()=>o,dynamic_time_warping:()=>y,interpolate_data:()=>r,log_softmax:()=>s,magnitude:()=>u,max:()=>c,medianFilter:()=>g,min:()=>d,permute_data:()=>i,round:()=>_,softmax:()=>a});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 i,a,s=1<<this._width,o=r/s<<1;const l=this._bitrev;if(4===o)for(i=0,a=0;i<r;i+=o,++a){const n=l[a];this._singleTransform2(t,e,i,n,s)}else for(i=0,a=0;i<r;i+=o,++a){const r=l[a];this._singleTransform4(t,e,i,r,s,n)}const u=this.table;for(s>>=2;s>=2;s>>=2){o=r/s<<1;const t=o>>>2;for(i=0;i<r;i+=o){const r=i+t-1;for(let a=i,o=0;a<r;a+=2,o+=s){const r=a,i=r+t,s=i+t,l=s+t,d=e[r],c=e[r+1],p=e[i],h=e[i+1],m=e[s],f=e[s+1],g=e[l],_=e[l+1],w=u[o],y=n*u[o+1],b=p*w-h*y,v=p*y+h*w,x=u[2*o],M=n*u[2*o+1],T=m*x-f*M,k=m*M+f*x,$=u[3*o],S=n*u[3*o+1],C=g*$-_*S,E=g*S+_*$,P=d+T,A=c+k,F=d-T,z=c-k,I=b+C,O=v+E,B=n*(b-C),L=n*(v-E);e[r]=P+I,e[r+1]=A+O,e[i]=F+L,e[i+1]=z-B,e[s]=P-I,e[s+1]=A-O,e[l]=F-L,e[l+1]=z+B}}}}_singleTransform2(e,t,n,r,i){const a=e[r],s=e[r+1],o=e[r+i],l=e[r+i+1];t[n]=a+o,t[n+1]=s+l,t[n+2]=a-o,t[n+3]=s-l}_singleTransform4(e,t,n,r,i,a){const s=2*i,o=3*i,l=e[r],u=e[r+1],d=e[r+i],c=e[r+i+1],p=e[r+s],h=e[r+s+1],m=e[r+o],f=e[r+o+1],g=l+p,_=u+h,w=l-p,y=u-h,b=d+m,v=c+f,x=a*(d-m),M=a*(c-f);t[n]=g+b,t[n+1]=_+v,t[n+2]=w+M,t[n+3]=y-x,t[n+4]=g-b,t[n+5]=_-v,t[n+6]=w-M,t[n+7]=y+x}_realTransform4(e,t,n){const r=this._csize;let i,a,s=1<<this._width,o=r/s<<1;const l=this._bitrev;if(4===o)for(i=0,a=0;i<r;i+=o,++a){const n=l[a];this._singleRealTransform2(t,e,i,n>>>1,s>>>1)}else for(i=0,a=0;i<r;i+=o,++a){const r=l[a];this._singleRealTransform4(t,e,i,r>>>1,s>>>1,n)}const u=this.table;for(s>>=2;s>=2;s>>=2){o=r/s<<1;const t=o>>>1,a=t>>>1,l=a>>>1;for(i=0;i<r;i+=o)for(let r=0,o=0;r<=l;r+=2,o+=s){const s=i+r,d=s+a,c=d+a,p=c+a,h=e[s],m=e[s+1],f=e[d],g=e[d+1],_=e[c],w=e[c+1],y=e[p],b=e[p+1],v=h,x=m,M=u[o],T=n*u[o+1],k=f*M-g*T,$=f*T+g*M,S=u[2*o],C=n*u[2*o+1],E=_*S-w*C,P=_*C+w*S,A=u[3*o],F=n*u[3*o+1],z=y*A-b*F,I=y*F+b*A,O=v+E,B=x+P,L=v-E,D=x-P,R=k+z,N=$+I,V=n*(k-z),j=n*($-I);if(e[s]=O+R,e[s+1]=B+N,e[d]=L+j,e[d+1]=D-V,0===r){e[c]=O-R,e[c+1]=B-N;continue}if(r===l)continue;const G=i+a-r,q=i+t-r;e[G]=L-n*j,e[G+1]=-D-n*V,e[q]=O-n*R,e[q+1]=n*N-B}}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,i){const a=e[r],s=e[r+i];t[n]=a+s,t[n+1]=0,t[n+2]=a-s,t[n+3]=0}_singleRealTransform4(e,t,n,r,i,a){const s=2*i,o=3*i,l=e[r],u=e[r+i],d=e[r+s],c=e[r+o],p=l+d,h=l-d,m=u+c,f=a*(u-c);t[n]=p+m,t[n+1]=0,t[n+2]=h,t[n+3]=-f,t[n+4]=p-m,t[n+5]=0,t[n+6]=h,t[n+7]=f}}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 i=new Float64Array(n),a=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 s=-2*Math.PI/e,o=Math.cos(s),l=Math.sin(s);for(let t=0;t<n>>1;++t){const n=(t+1-e)**2/2,r=Math.sqrt(o**2+l**2)**n,s=n*Math.atan2(l,o),u=2*t;i[u]=r*Math.cos(s),i[u+1]=r*Math.sin(s),a[u]=i[u],a[u+1]=-i[u+1]}this._slicedChirpBuffer=i.subarray(t,n),this._f=new h(r>>1),this._f.transform(this._chirpBuffer,a)}_transform(e,t,n){const r=this._buffer1,i=this._buffer2,a=this._outBuffer1,s=this._outBuffer2,o=this._chirpBuffer,l=this._slicedChirpBuffer,u=this._a;if(n)for(let e=0;e<l.length;e+=2){const n=e+1,i=t[e>>1];r[e]=i*l[e],r[n]=i*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(a,r);for(let e=0;e<o.length;e+=2){const t=e+1;i[e]=a[e]*o[e]-a[t]*o[t],i[t]=a[e]*o[t]+a[t]*o[e]}this._f.inverseTransform(s,i);for(let t=0;t<s.length;t+=2){const n=s[t+u],r=s[t+u+1],i=l[t],a=l[t+1];e[t]=n*i-r*a,e[t+1]=n*a+r*i}}transform(e,t){this._transform(e,t,!1)}realTransform(e,t){this._transform(e,t,!0)}}class f{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 g(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),i=Math.floor(t/2);for(let t=0;t<e.length;++t){let a=0;for(let n=-i;n<=i;++n){let i=t+n;i<0?i=Math.abs(i):i>=e.length&&(i=2*(e.length-1)-i),r[a++]=e[i]}r.sort(),n[t]=r[i]}return n}function _(e,t){const n=Math.pow(10,t);return Math.round(e*n)/n}function w(e){const t=Math.round(e);return Math.abs(e)%1==.5?t%2==0?t:t-1:t}function y(e){const t=e.length,n=e[0].length,r=[t+1,n+1],i=Array.from({length:r[0]},(()=>Array(r[1]).fill(1/0)));i[0][0]=0;const a=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=i[n-1][t-1],s=i[n-1][t],o=i[n][t-1];let l,u;r<s&&r<o?(l=r,u=0):s<r&&s<o?(l=s,u=1):(l=o,u=2),i[n][t]=e[n-1][t-1]+l,a[n][t]=u}for(let e=0;e<r[1];++e)a[0][e]=2;for(let e=0;e<r[0];++e)a[e][0]=1;let s=t,o=n,l=[],u=[];for(;s>0||o>0;)switch(l.push(s-1),u.push(o-1),a[s][o]){case 0:--s,--o;break;case 1:--s;break;case 2:--o;break;default:throw new Error(`Internal error in dynamic time warping. Unexpected trace[${s}, ${o}]. Please file a bug report.`)}return l.reverse(),u.reverse(),[l,u]}},"./src/utils/tensor.js":
 /*!*****************************!*\
   !*** ./src/utils/tensor.js ***!
-  \*****************************/(e,t,n)=>{n.r(t),n.d(t,{Tensor:()=>o,cat:()=>y,full:()=>T,full_like:()=>k,interpolate:()=>u,interpolate_4d:()=>d,layer_norm:()=>f,matmul:()=>c,mean:()=>x,mean_pooling:()=>m,ones:()=>$,ones_like:()=>S,permute:()=>l,quantize_embeddings:()=>P,rfft:()=>p,stack:()=>b,std_mean:()=>v,topk:()=>h,zeros:()=>C,zeros_like:()=>E});var r=n(/*! ./maths.js */"./src/utils/maths.js"),i=n(/*! ../backends/onnx.js */"./src/backends/onnx.js"),a=n(/*! ../ops/registry.js */"./src/ops/registry.js");const s=Object.freeze({float32:Float32Array,float16:Uint16Array,float64:Float64Array,string:Array,int8:Int8Array,uint8:Uint8Array,int16:Int16Array,uint16:Uint16Array,int32:Int32Array,uint32:Uint32Array,int64:BigInt64Array,uint64:BigUint64Array,bool:Uint8Array});class o{get dims(){return this.ort_tensor.dims}set dims(e){this.ort_tensor.dims=e}get type(){return this.ort_tensor.type}get data(){return this.ort_tensor.data}get size(){return this.ort_tensor.size}get location(){return this.ort_tensor.location}ort_tensor;constructor(...e){return(0,i.isONNXTensor)(e[0])?this.ort_tensor=e[0]:this.ort_tensor=new i.Tensor(e[0],e[1],e[2]),new Proxy(this,{get:(e,t)=>{if("string"==typeof t){let n=Number(t);if(Number.isInteger(n))return e._getitem(n)}return e[t]},set:(e,t,n)=>e[t]=n})}dispose(){this.ort_tensor.dispose()}*[Symbol.iterator](){const[e,...t]=this.dims;if(t.length>0){const n=t.reduce(((e,t)=>e*t));for(let r=0;r<e;++r)yield this._subarray(r,n,t)}else yield*this.data}_getitem(e){const[t,...n]=this.dims;if(e=w(e,t),n.length>0){const t=n.reduce(((e,t)=>e*t));return this._subarray(e,t,n)}return new o(this.type,[this.data[e]],n)}indexOf(e){const t=this.data;for(let n=0;n<t.length;++n)if(t[n]==e)return n;return-1}_subarray(e,t,n){const r=e*t,i=(e+1)*t,a="subarray"in this.data?this.data.subarray(r,i):this.data.slice(r,i);return new o(this.type,a,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 i=e;for(let e=t.length-1;e>=0;e--)i=i.reduce(((n,r)=>{let i=n[n.length-1];return i.length<t[e]?i.push(r):n.push([r]),n}),[[]]);return i[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 o(this.type,this.data.slice(),this.dims.slice())}slice(...e){const t=[],n=[];for(let r=0;r<this.dims.length;++r){let i=e[r];if(null==i)n.push([0,this.dims[r]]),t.push(this.dims[r]);else if("number"==typeof i)i=w(i,this.dims[r],r),n.push([i,i+1]);else{if(!Array.isArray(i)||2!==i.length)throw new Error(`Invalid slice: ${i}`);{let[e,a]=i;if(e=null===e?0:w(e,this.dims[r],r,!1),a=null===a?this.dims[r]:w(a,this.dims[r],r,!1),e>a)throw new Error(`Invalid slice: ${i}`);const s=[Math.max(e,0),Math.min(a,this.dims[r])];n.push(s),t.push(s[1]-s[0])}}}const r=n.map((([e,t])=>t-e)),i=r.reduce(((e,t)=>e*t)),a=this.data,s=new a.constructor(i),l=this.stride();for(let e=0;e<i;++e){let t=0;for(let i=r.length-1,a=e;i>=0;--i){const e=r[i];t+=(a%e+n[i][0])*l[i],a=Math.floor(a/e)}s[e]=a[t]}return new o(this.type,s,t)}permute(...e){return l(this,e)}transpose(...e){return this.permute(...e)}sum(e=null,t=!1){return this.norm(1,e,t)}norm(e="fro",t=null,n=!1){if("fro"===e)e=2;else if("string"==typeof e)throw Error(`Unsupported norm: ${e}`);const r=this.data;if(null===t){let t=r.reduce(((t,n)=>t+n**e),0)**(1/e);return new o(this.type,[t],[])}t=w(t,this.dims.length);const i=this.dims.slice();i[t]=1;const a=new r.constructor(r.length/this.dims[t]);for(let n=0;n<r.length;++n){let s=0;for(let e=this.dims.length-1,r=n,a=1;e>=0;--e){const n=this.dims[e];if(e!==t){s+=r%n*a,a*=i[e]}r=Math.floor(r/n)}a[s]+=r[n]**e}if(1!==e)for(let t=0;t<a.length;++t)a[t]=a[t]**(1/e);return n||i.splice(t,1),new o(this.type,a,i)}normalize_(e=2,t=1){t=w(t,this.dims.length);const n=this.norm(e,t,!0),r=this.data,i=n.data;for(let e=0;e<r.length;++e){let n=0;for(let r=this.dims.length-1,i=e,a=1;r>=0;--r){const e=this.dims[r];if(r!==t){n+=i%e*a,a*=this.dims[r]}i=Math.floor(i/e)}r[e]/=i[n]}return this}normalize(e=2,t=1){return this.clone().normalize_(e,t)}stride(){return function(e){const t=new Array(e.length);for(let n=e.length-1,r=1;n>=0;--n)t[n]=r,r*=e[n];return t}(this.dims)}squeeze(e=null){return new o(this.type,this.data,g(this.dims,e))}squeeze_(e=null){return this.dims=g(this.dims,e),this}unsqueeze(e=null){return new o(this.type,this.data,_(this.dims,e))}unsqueeze_(e=null){return this.dims=_(this.dims,e),this}flatten_(e=0,t=-1){t=(t+this.dims.length)%this.dims.length;let n=this.dims.slice(0,e),r=this.dims.slice(e,t+1),i=this.dims.slice(t+1);return this.dims=[...n,r.reduce(((e,t)=>e*t),1),...i],this}flatten(e=0,t=-1){return this.clone().flatten_(e,t)}view(...e){let t=-1;for(let n=0;n<e.length;++n)if(-1===e[n]){if(-1!==t)throw new Error("Only one dimension can be inferred");t=n}const n=this.data;if(-1!==t){const r=e.reduce(((e,n,r)=>r!==t?e*n:e),1);e[t]=n.length/r}return new o(this.type,n,e)}neg_(){const e=this.data;for(let t=0;t<e.length;++t)e[t]=-e[t];return this}neg(){return this.clone().neg_()}clamp_(e,t){const n=this.data;for(let r=0;r<n.length;++r)n[r]=Math.min(Math.max(n[r],e),t);return this}clamp(e,t){return this.clone().clamp_(e,t)}round_(){const e=this.data;for(let t=0;t<e.length;++t)e[t]=Math.round(e[t]);return this}round(){return this.clone().round_()}mean(e=null,t=!1){return x(this,e,t)}to(e){if(this.type===e)return this;if(!s.hasOwnProperty(e))throw new Error(`Unsupported type: ${e}`);return new o(e,s[e].from(this.data),this.dims)}}function l(e,t){const[n,i]=(0,r.permute_data)(e.data,e.dims,t);return new o(e.type,n,i)}function u(e,[t,n],i="bilinear",a=!1){const s=e.dims.at(-3)??1,l=e.dims.at(-2),u=e.dims.at(-1);let d=(0,r.interpolate_data)(e.data,[s,l,u],[t,n],i,a);return new o(e.type,d,[s,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,i;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)i=await a.TensorOpRegistry.bilinear_interpolate_4d;else{if("bicubic"!==n)throw new Error(`Unsupported mode: ${n}`);i=await a.TensorOpRegistry.bicubic_interpolate_4d}const s=new o("int64",new BigInt64Array(r.map(BigInt)),[r.length]);return await i({x:e,s})}async function c(e,t){const n=await a.TensorOpRegistry.matmul;return await n({a:e,b:t})}async function p(e,t){const n=await a.TensorOpRegistry.rfft;return await n({x:e,a:t})}async function h(e,t){const n=await a.TensorOpRegistry.top_k;return t=null===t?e.dims.at(-1):Math.min(t,e.dims.at(-1)),await n({x:e,k:new o("int64",[BigInt(t)],[1])})}function m(e,t){const n=e.data,r=t.data,i=[e.dims[0],e.dims[2]],a=new n.constructor(i[0]*i[1]),[s,l,u]=e.dims;let d=0;for(let e=0;e<s;++e){const t=e*u*l;for(let i=0;i<u;++i){let s=0,o=0;const c=e*l,p=t+i;for(let e=0;e<l;++e){const t=Number(r[c+e]);o+=t,s+=n[p+e*u]*t}const h=s/o;a[d++]=h}}return new o(e.type,a,i)}function f(e,t,{eps:n=1e-5}={}){if(2!==e.dims.length)throw new Error("`layer_norm` currently only supports 2D input.");const[r,i]=e.dims;if(1!==t.length&&t[0]!==i)throw new Error("`normalized_shape` must be a 1D array with shape `[input.dims[1]]`.");const[a,s]=v(e,1,0,!0),l=a.data,u=s.data,d=e.data,c=new d.constructor(d.length);for(let e=0;e<r;++e){const t=e*i;for(let r=0;r<i;++r){const i=t+r;c[i]=(d[i]-u[e])/(l[e]+n)}}return new o(e.type,c,e.dims)}function g(e,t){return e=e.slice(),null===t?e=e.filter((e=>1!==e)):"number"==typeof t?1===e[t]&&e.splice(t,1):Array.isArray(t)&&(e=e.filter(((e,n)=>1!==e||!t.includes(n)))),e}function _(e,t){return t=w(t,e.length+1),(e=e.slice()).splice(t,0,1),e}function w(e,t,n=null,r=!0){if(r&&(e<-t||e>=t))throw new Error(`IndexError: index ${e} is out of bounds for dimension${null===n?"":" "+n} with size ${t}`);return e<0&&(e=(e%t+t)%t),e}function y(e,t=0){t=w(t,e[0].dims.length);const n=e[0].dims.slice();n[t]=e.reduce(((e,n)=>e+n.dims[t]),0);const r=n.reduce(((e,t)=>e*t),1),i=new e[0].data.constructor(r),a=e[0].type;if(0===t){let t=0;for(const n of e){const e=n.data;i.set(e,t),t+=e.length}}else{let r=0;for(let a=0;a<e.length;++a){const{data:s,dims:o}=e[a];for(let e=0;e<s.length;++e){let a=0;for(let i=o.length-1,s=e,l=1;i>=0;--i){const e=o[i];let u=s%e;i===t&&(u+=r),a+=u*l,l*=n[i],s=Math.floor(s/e)}i[a]=s[e]}r+=o[t]}}return new o(a,i,n)}function b(e,t=0){return y(e.map((e=>e.unsqueeze(t))),t)}function v(e,t=null,n=1,r=!1){const i=e.data,a=e.dims;if(null===t){const t=i.reduce(((e,t)=>e+t),0)/i.length,r=Math.sqrt(i.reduce(((e,n)=>e+(n-t)**2),0)/(i.length-n)),a=new o(e.type,[t],[]);return[new o(e.type,[r],[]),a]}const s=x(e,t=w(t,a.length),r),l=s.data,u=a.slice();u[t]=1;const d=new i.constructor(i.length/a[t]);for(let e=0;e<i.length;++e){let n=0;for(let r=a.length-1,i=e,s=1;r>=0;--r){const e=a[r];if(r!==t){n+=i%e*s,s*=u[r]}i=Math.floor(i/e)}d[n]+=(i[e]-l[n])**2}for(let e=0;e<d.length;++e)d[e]=Math.sqrt(d[e]/(a[t]-n));r||u.splice(t,1);return[new o(e.type,d,u),s]}function x(e,t=null,n=!1){const r=e.data;if(null===t){const t=r.reduce(((e,t)=>e+t),0);return new o(e.type,[t/r.length],[])}const i=e.dims;t=w(t,i.length);const a=i.slice();a[t]=1;const s=new r.constructor(r.length/i[t]);for(let e=0;e<r.length;++e){let n=0;for(let r=i.length-1,s=e,o=1;r>=0;--r){const e=i[r];if(r!==t){n+=s%e*o,o*=a[r]}s=Math.floor(s/e)}s[n]+=r[e]}if(1!==i[t])for(let e=0;e<s.length;++e)s[e]=s[e]/i[t];return n||a.splice(t,1),new o(e.type,s,a)}function M(e,t,n,r){const i=e.reduce(((e,t)=>e*t),1);return new o(n,new r(i).fill(t),e)}function T(e,t){let n,r;if("number"==typeof t)n="float32",r=Float32Array;else{if("bigint"!=typeof t)throw new Error("Unsupported data type: "+typeof t);n="int64",r=BigInt64Array}return M(e,t,n,r)}function k(e,t){return T(e.dims,t)}function $(e){return M(e,1n,"int64",BigInt64Array)}function S(e){return $(e.dims)}function C(e){return M(e,0n,"int64",BigInt64Array)}function E(e){return C(e.dims)}function P(e,t){if(2!==e.dims.length)throw new Error("The tensor must have 2 dimensions");if(e.dims.at(-1)%8!=0)throw new Error("The last dimension of the tensor must be a multiple of 8");if(!["binary","ubinary"].includes(t))throw new Error("The precision must be either 'binary' or 'ubinary'");const n="binary"===t,r=n?"int8":"uint8",i=n?Int8Array:Uint8Array,a=e.data,s=new i(a.length/8);for(let e=0;e<a.length;++e){const t=a[e]>0?1:0,r=Math.floor(e/8),i=e%8;s[r]|=t<<7-i,n&&0===i&&(s[r]-=128)}return new o(r,s,[e.dims[0],e.dims[1]/8])}}},r={};function i(e){var t=r[e];if(void 0!==t)return t.exports;var a=r[e]={exports:{}};return n[e](a,a.exports,i),a.exports}i.m=n,t=Object.getPrototypeOf?e=>Object.getPrototypeOf(e):e=>e.__proto__,i.t=function(n,r){if(1&r&&(n=this(n)),8&r)return n;if("object"==typeof n&&n){if(4&r&&n.__esModule)return n;if(16&r&&"function"==typeof n.then)return n}var a=Object.create(null);i.r(a);var s={};e=e||[null,t({}),t([]),t(t)];for(var o=2&r&&n;"object"==typeof o&&!~e.indexOf(o);o=t(o))Object.getOwnPropertyNames(o).forEach((e=>s[e]=()=>n[e]));return s.default=()=>n,i.d(a,s),a},i.d=(e,t)=>{for(var n in t)i.o(t,n)&&!i.o(e,n)&&Object.defineProperty(e,n,{enumerable:!0,get:t[n]})},i.o=(e,t)=>Object.prototype.hasOwnProperty.call(e,t),i.r=e=>{"undefined"!=typeof Symbol&&Symbol.toStringTag&&Object.defineProperty(e,Symbol.toStringTag,{value:"Module"}),Object.defineProperty(e,"__esModule",{value:!0})},(()=>{var e;if("string"==typeof import.meta.url&&(e=import.meta.url),!e)throw new Error("Automatic publicPath is not supported in this browser");e=e.replace(/#.*$/,"").replace(/\?.*$/,"").replace(/\/[^\/]+$/,"/"),i.p=e})(),i.b=new URL("./",import.meta.url);var a={};
+  \*****************************/(e,t,n)=>{n.r(t),n.d(t,{Tensor:()=>o,cat:()=>y,full:()=>T,full_like:()=>k,interpolate:()=>u,interpolate_4d:()=>d,layer_norm:()=>f,matmul:()=>c,mean:()=>x,mean_pooling:()=>m,ones:()=>$,ones_like:()=>S,permute:()=>l,quantize_embeddings:()=>P,rfft:()=>p,stack:()=>b,std_mean:()=>v,topk:()=>h,zeros:()=>C,zeros_like:()=>E});var r=n(/*! ./maths.js */"./src/utils/maths.js"),i=n(/*! ../backends/onnx.js */"./src/backends/onnx.js"),a=n(/*! ../ops/registry.js */"./src/ops/registry.js");const s=Object.freeze({float32:Float32Array,float16:Uint16Array,float64:Float64Array,string:Array,int8:Int8Array,uint8:Uint8Array,int16:Int16Array,uint16:Uint16Array,int32:Int32Array,uint32:Uint32Array,int64:BigInt64Array,uint64:BigUint64Array,bool:Uint8Array});class o{get dims(){return this.ort_tensor.dims}set dims(e){this.ort_tensor.dims=e}get type(){return this.ort_tensor.type}get data(){return this.ort_tensor.data}get size(){return this.ort_tensor.size}get location(){return this.ort_tensor.location}ort_tensor;constructor(...e){return(0,i.isONNXTensor)(e[0])?this.ort_tensor=e[0]:this.ort_tensor=new i.Tensor(e[0],e[1],e[2]),new Proxy(this,{get:(e,t)=>{if("string"==typeof t){let n=Number(t);if(Number.isInteger(n))return e._getitem(n)}return e[t]},set:(e,t,n)=>e[t]=n})}dispose(){this.ort_tensor.dispose()}*[Symbol.iterator](){const[e,...t]=this.dims;if(t.length>0){const n=t.reduce(((e,t)=>e*t));for(let r=0;r<e;++r)yield this._subarray(r,n,t)}else yield*this.data}_getitem(e){const[t,...n]=this.dims;if(e=w(e,t),n.length>0){const t=n.reduce(((e,t)=>e*t));return this._subarray(e,t,n)}return new o(this.type,[this.data[e]],n)}indexOf(e){const t=this.data;for(let n=0;n<t.length;++n)if(t[n]==e)return n;return-1}_subarray(e,t,n){const r=e*t,i=(e+1)*t,a="subarray"in this.data?this.data.subarray(r,i):this.data.slice(r,i);return new o(this.type,a,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 i=e;for(let e=t.length-1;e>=0;e--)i=i.reduce(((n,r)=>{let i=n[n.length-1];return i.length<t[e]?i.push(r):n.push([r]),n}),[[]]);return i[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 o(this.type,this.data.slice(),this.dims.slice())}slice(...e){const t=[],n=[];for(let r=0;r<this.dims.length;++r){let i=e[r];if(null==i)n.push([0,this.dims[r]]),t.push(this.dims[r]);else if("number"==typeof i)i=w(i,this.dims[r],r),n.push([i,i+1]);else{if(!Array.isArray(i)||2!==i.length)throw new Error(`Invalid slice: ${i}`);{let[e,a]=i;if(e=null===e?0:w(e,this.dims[r],r,!1),a=null===a?this.dims[r]:w(a,this.dims[r],r,!1),e>a)throw new Error(`Invalid slice: ${i}`);const s=[Math.max(e,0),Math.min(a,this.dims[r])];n.push(s),t.push(s[1]-s[0])}}}const r=n.map((([e,t])=>t-e)),i=r.reduce(((e,t)=>e*t)),a=this.data,s=new a.constructor(i),l=this.stride();for(let e=0;e<i;++e){let t=0;for(let i=r.length-1,a=e;i>=0;--i){const e=r[i];t+=(a%e+n[i][0])*l[i],a=Math.floor(a/e)}s[e]=a[t]}return new o(this.type,s,t)}permute(...e){return l(this,e)}transpose(...e){return this.permute(...e)}sum(e=null,t=!1){return this.norm(1,e,t)}norm(e="fro",t=null,n=!1){if("fro"===e)e=2;else if("string"==typeof e)throw Error(`Unsupported norm: ${e}`);const r=this.data;if(null===t){let t=r.reduce(((t,n)=>t+n**e),0)**(1/e);return new o(this.type,[t],[])}t=w(t,this.dims.length);const i=this.dims.slice();i[t]=1;const a=new r.constructor(r.length/this.dims[t]);for(let n=0;n<r.length;++n){let s=0;for(let e=this.dims.length-1,r=n,a=1;e>=0;--e){const n=this.dims[e];if(e!==t){s+=r%n*a,a*=i[e]}r=Math.floor(r/n)}a[s]+=r[n]**e}if(1!==e)for(let t=0;t<a.length;++t)a[t]=a[t]**(1/e);return n||i.splice(t,1),new o(this.type,a,i)}normalize_(e=2,t=1){t=w(t,this.dims.length);const n=this.norm(e,t,!0),r=this.data,i=n.data;for(let e=0;e<r.length;++e){let n=0;for(let r=this.dims.length-1,i=e,a=1;r>=0;--r){const e=this.dims[r];if(r!==t){n+=i%e*a,a*=this.dims[r]}i=Math.floor(i/e)}r[e]/=i[n]}return this}normalize(e=2,t=1){return this.clone().normalize_(e,t)}stride(){return function(e){const t=new Array(e.length);for(let n=e.length-1,r=1;n>=0;--n)t[n]=r,r*=e[n];return t}(this.dims)}squeeze(e=null){return new o(this.type,this.data,g(this.dims,e))}squeeze_(e=null){return this.dims=g(this.dims,e),this}unsqueeze(e=null){return new o(this.type,this.data,_(this.dims,e))}unsqueeze_(e=null){return this.dims=_(this.dims,e),this}flatten_(e=0,t=-1){t=(t+this.dims.length)%this.dims.length;let n=this.dims.slice(0,e),r=this.dims.slice(e,t+1),i=this.dims.slice(t+1);return this.dims=[...n,r.reduce(((e,t)=>e*t),1),...i],this}flatten(e=0,t=-1){return this.clone().flatten_(e,t)}view(...e){let t=-1;for(let n=0;n<e.length;++n)if(-1===e[n]){if(-1!==t)throw new Error("Only one dimension can be inferred");t=n}const n=this.data;if(-1!==t){const r=e.reduce(((e,n,r)=>r!==t?e*n:e),1);e[t]=n.length/r}return new o(this.type,n,e)}neg_(){const e=this.data;for(let t=0;t<e.length;++t)e[t]=-e[t];return this}neg(){return this.clone().neg_()}clamp_(e,t){const n=this.data;for(let r=0;r<n.length;++r)n[r]=Math.min(Math.max(n[r],e),t);return this}clamp(e,t){return this.clone().clamp_(e,t)}round_(){const e=this.data;for(let t=0;t<e.length;++t)e[t]=Math.round(e[t]);return this}round(){return this.clone().round_()}mean(e=null,t=!1){return x(this,e,t)}to(e){if(this.type===e)return this;if(!s.hasOwnProperty(e))throw new Error(`Unsupported type: ${e}`);return new o(e,s[e].from(this.data),this.dims)}}function l(e,t){const[n,i]=(0,r.permute_data)(e.data,e.dims,t);return new o(e.type,n,i)}function u(e,[t,n],i="bilinear",a=!1){const s=e.dims.at(-3)??1,l=e.dims.at(-2),u=e.dims.at(-1);let d=(0,r.interpolate_data)(e.data,[s,l,u],[t,n],i,a);return new o(e.type,d,[s,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,i;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)i=await a.TensorOpRegistry.bilinear_interpolate_4d;else{if("bicubic"!==n)throw new Error(`Unsupported mode: ${n}`);i=await a.TensorOpRegistry.bicubic_interpolate_4d}const s=new o("int64",new BigInt64Array(r.map(BigInt)),[r.length]);return await i({x:e,s})}async function c(e,t){const n=await a.TensorOpRegistry.matmul;return await n({a:e,b:t})}async function p(e,t){const n=await a.TensorOpRegistry.rfft;return await n({x:e,a:t})}async function h(e,t){const n=await a.TensorOpRegistry.top_k;return t=null===t?e.dims.at(-1):Math.min(t,e.dims.at(-1)),await n({x:e,k:new o("int64",[BigInt(t)],[1])})}function m(e,t){const n=e.data,r=t.data,i=[e.dims[0],e.dims[2]],a=new n.constructor(i[0]*i[1]),[s,l,u]=e.dims;let d=0;for(let e=0;e<s;++e){const t=e*u*l;for(let i=0;i<u;++i){let s=0,o=0;const c=e*l,p=t+i;for(let e=0;e<l;++e){const t=Number(r[c+e]);o+=t,s+=n[p+e*u]*t}const h=s/o;a[d++]=h}}return new o(e.type,a,i)}function f(e,t,{eps:n=1e-5}={}){if(2!==e.dims.length)throw new Error("`layer_norm` currently only supports 2D input.");const[r,i]=e.dims;if(1!==t.length&&t[0]!==i)throw new Error("`normalized_shape` must be a 1D array with shape `[input.dims[1]]`.");const[a,s]=v(e,1,0,!0),l=a.data,u=s.data,d=e.data,c=new d.constructor(d.length);for(let e=0;e<r;++e){const t=e*i;for(let r=0;r<i;++r){const i=t+r;c[i]=(d[i]-u[e])/(l[e]+n)}}return new o(e.type,c,e.dims)}function g(e,t){return e=e.slice(),null===t?e=e.filter((e=>1!==e)):"number"==typeof t?1===e[t]&&e.splice(t,1):Array.isArray(t)&&(e=e.filter(((e,n)=>1!==e||!t.includes(n)))),e}function _(e,t){return t=w(t,e.length+1),(e=e.slice()).splice(t,0,1),e}function w(e,t,n=null,r=!0){if(r&&(e<-t||e>=t))throw new Error(`IndexError: index ${e} is out of bounds for dimension${null===n?"":" "+n} with size ${t}`);return e<0&&(e=(e%t+t)%t),e}function y(e,t=0){t=w(t,e[0].dims.length);const n=e[0].dims.slice();n[t]=e.reduce(((e,n)=>e+n.dims[t]),0);const r=n.reduce(((e,t)=>e*t),1),i=new e[0].data.constructor(r),a=e[0].type;if(0===t){let t=0;for(const n of e){const e=n.data;i.set(e,t),t+=e.length}}else{let r=0;for(let a=0;a<e.length;++a){const{data:s,dims:o}=e[a];for(let e=0;e<s.length;++e){let a=0;for(let i=o.length-1,s=e,l=1;i>=0;--i){const e=o[i];let u=s%e;i===t&&(u+=r),a+=u*l,l*=n[i],s=Math.floor(s/e)}i[a]=s[e]}r+=o[t]}}return new o(a,i,n)}function b(e,t=0){return y(e.map((e=>e.unsqueeze(t))),t)}function v(e,t=null,n=1,r=!1){const i=e.data,a=e.dims;if(null===t){const t=i.reduce(((e,t)=>e+t),0)/i.length,r=Math.sqrt(i.reduce(((e,n)=>e+(n-t)**2),0)/(i.length-n)),a=new o(e.type,[t],[]);return[new o(e.type,[r],[]),a]}const s=x(e,t=w(t,a.length),r),l=s.data,u=a.slice();u[t]=1;const d=new i.constructor(i.length/a[t]);for(let e=0;e<i.length;++e){let n=0;for(let r=a.length-1,i=e,s=1;r>=0;--r){const e=a[r];if(r!==t){n+=i%e*s,s*=u[r]}i=Math.floor(i/e)}d[n]+=(i[e]-l[n])**2}for(let e=0;e<d.length;++e)d[e]=Math.sqrt(d[e]/(a[t]-n));r||u.splice(t,1);return[new o(e.type,d,u),s]}function x(e,t=null,n=!1){const r=e.data;if(null===t){const t=r.reduce(((e,t)=>e+t),0);return new o(e.type,[t/r.length],[])}const i=e.dims;t=w(t,i.length);const a=i.slice();a[t]=1;const s=new r.constructor(r.length/i[t]);for(let e=0;e<r.length;++e){let n=0;for(let r=i.length-1,s=e,o=1;r>=0;--r){const e=i[r];if(r!==t){n+=s%e*o,o*=a[r]}s=Math.floor(s/e)}s[n]+=r[e]}if(1!==i[t])for(let e=0;e<s.length;++e)s[e]=s[e]/i[t];return n||a.splice(t,1),new o(e.type,s,a)}function M(e,t,n,r){const i=e.reduce(((e,t)=>e*t),1);return new o(n,new r(i).fill(t),e)}function T(e,t){let n,r;if("number"==typeof t)n="float32",r=Float32Array;else{if("bigint"!=typeof t)throw new Error("Unsupported data type: "+typeof t);n="int64",r=BigInt64Array}return M(e,t,n,r)}function k(e,t){return T(e.dims,t)}function $(e){return M(e,1n,"int64",BigInt64Array)}function S(e){return $(e.dims)}function C(e){return M(e,0n,"int64",BigInt64Array)}function E(e){return C(e.dims)}function P(e,t){if(2!==e.dims.length)throw new Error("The tensor must have 2 dimensions");if(e.dims.at(-1)%8!=0)throw new Error("The last dimension of the tensor must be a multiple of 8");if(!["binary","ubinary"].includes(t))throw new Error("The precision must be either 'binary' or 'ubinary'");const n="binary"===t,r=n?"int8":"uint8",i=n?Int8Array:Uint8Array,a=e.data,s=new i(a.length/8);for(let e=0;e<a.length;++e){const t=a[e]>0?1:0,r=Math.floor(e/8),i=e%8;s[r]|=t<<7-i,n&&0===i&&(s[r]-=128)}return new o(r,s,[e.dims[0],e.dims[1]/8])}}},r={};function i(e){var t=r[e];if(void 0!==t)return t.exports;var a=r[e]={exports:{}};return n[e](a,a.exports,i),a.exports}i.m=n,t=Object.getPrototypeOf?e=>Object.getPrototypeOf(e):e=>e.__proto__,i.t=function(n,r){if(1&r&&(n=this(n)),8&r)return n;if("object"==typeof n&&n){if(4&r&&n.__esModule)return n;if(16&r&&"function"==typeof n.then)return n}var a=Object.create(null);i.r(a);var s={};e=e||[null,t({}),t([]),t(t)];for(var o=2&r&&n;"object"==typeof o&&!~e.indexOf(o);o=t(o))Object.getOwnPropertyNames(o).forEach((e=>s[e]=()=>n[e]));return s.default=()=>n,i.d(a,s),a},i.d=(e,t)=>{for(var n in t)i.o(t,n)&&!i.o(e,n)&&Object.defineProperty(e,n,{enumerable:!0,get:t[n]})},i.o=(e,t)=>Object.prototype.hasOwnProperty.call(e,t),i.r=e=>{"undefined"!=typeof Symbol&&Symbol.toStringTag&&Object.defineProperty(e,Symbol.toStringTag,{value:"Module"}),Object.defineProperty(e,"__esModule",{value:!0})},(()=>{var e;if("string"==typeof import.meta.url&&(e=import.meta.url),!e)throw new Error("Automatic publicPath is not supported in this browser");e=e.replace(/#.*$/,"").replace(/\?.*$/,"").replace(/\/[^\/]+$/,"/"),i.p=e})(),i.b=new URL("./",import.meta.url);var a={};
 /*!*****************************!*\
   !*** ./src/transformers.js ***!
-  \*****************************/i.r(a),i.d(a,{ASTFeatureExtractor:()=>d.ASTFeatureExtractor,ASTForAudioClassification:()=>l.ASTForAudioClassification,ASTModel:()=>l.ASTModel,ASTPreTrainedModel:()=>l.ASTPreTrainedModel,AlbertForMaskedLM:()=>l.AlbertForMaskedLM,AlbertForQuestionAnswering:()=>l.AlbertForQuestionAnswering,AlbertForSequenceClassification:()=>l.AlbertForSequenceClassification,AlbertModel:()=>l.AlbertModel,AlbertPreTrainedModel:()=>l.AlbertPreTrainedModel,AlbertTokenizer:()=>u.AlbertTokenizer,AudioClassificationPipeline:()=>o.AudioClassificationPipeline,AutoConfig:()=>c.AutoConfig,AutoModel:()=>l.AutoModel,AutoModelForAudioClassification:()=>l.AutoModelForAudioClassification,AutoModelForAudioFrameClassification:()=>l.AutoModelForAudioFrameClassification,AutoModelForCTC:()=>l.AutoModelForCTC,AutoModelForCausalLM:()=>l.AutoModelForCausalLM,AutoModelForDepthEstimation:()=>l.AutoModelForDepthEstimation,AutoModelForDocumentQuestionAnswering:()=>l.AutoModelForDocumentQuestionAnswering,AutoModelForImageClassification:()=>l.AutoModelForImageClassification,AutoModelForImageFeatureExtraction:()=>l.AutoModelForImageFeatureExtraction,AutoModelForImageMatting:()=>l.AutoModelForImageMatting,AutoModelForImageSegmentation:()=>l.AutoModelForImageSegmentation,AutoModelForImageToImage:()=>l.AutoModelForImageToImage,AutoModelForMaskGeneration:()=>l.AutoModelForMaskGeneration,AutoModelForMaskedLM:()=>l.AutoModelForMaskedLM,AutoModelForObjectDetection:()=>l.AutoModelForObjectDetection,AutoModelForQuestionAnswering:()=>l.AutoModelForQuestionAnswering,AutoModelForSemanticSegmentation:()=>l.AutoModelForSemanticSegmentation,AutoModelForSeq2SeqLM:()=>l.AutoModelForSeq2SeqLM,AutoModelForSequenceClassification:()=>l.AutoModelForSequenceClassification,AutoModelForSpeechSeq2Seq:()=>l.AutoModelForSpeechSeq2Seq,AutoModelForTextToSpectrogram:()=>l.AutoModelForTextToSpectrogram,AutoModelForTextToWaveform:()=>l.AutoModelForTextToWaveform,AutoModelForTokenClassification:()=>l.AutoModelForTokenClassification,AutoModelForVision2Seq:()=>l.AutoModelForVision2Seq,AutoModelForXVector:()=>l.AutoModelForXVector,AutoModelForZeroShotObjectDetection:()=>l.AutoModelForZeroShotObjectDetection,AutoProcessor:()=>d.AutoProcessor,AutoTokenizer:()=>u.AutoTokenizer,AutomaticSpeechRecognitionPipeline:()=>o.AutomaticSpeechRecognitionPipeline,BartForConditionalGeneration:()=>l.BartForConditionalGeneration,BartForSequenceClassification:()=>l.BartForSequenceClassification,BartModel:()=>l.BartModel,BartPretrainedModel:()=>l.BartPretrainedModel,BartTokenizer:()=>u.BartTokenizer,BaseModelOutput:()=>l.BaseModelOutput,BaseStreamer:()=>g.BaseStreamer,BeitFeatureExtractor:()=>d.BeitFeatureExtractor,BeitForImageClassification:()=>l.BeitForImageClassification,BeitModel:()=>l.BeitModel,BeitPreTrainedModel:()=>l.BeitPreTrainedModel,BertForMaskedLM:()=>l.BertForMaskedLM,BertForQuestionAnswering:()=>l.BertForQuestionAnswering,BertForSequenceClassification:()=>l.BertForSequenceClassification,BertForTokenClassification:()=>l.BertForTokenClassification,BertModel:()=>l.BertModel,BertPreTrainedModel:()=>l.BertPreTrainedModel,BertTokenizer:()=>u.BertTokenizer,BitImageProcessor:()=>d.BitImageProcessor,BlenderbotForConditionalGeneration:()=>l.BlenderbotForConditionalGeneration,BlenderbotModel:()=>l.BlenderbotModel,BlenderbotPreTrainedModel:()=>l.BlenderbotPreTrainedModel,BlenderbotSmallForConditionalGeneration:()=>l.BlenderbotSmallForConditionalGeneration,BlenderbotSmallModel:()=>l.BlenderbotSmallModel,BlenderbotSmallPreTrainedModel:()=>l.BlenderbotSmallPreTrainedModel,BlenderbotSmallTokenizer:()=>u.BlenderbotSmallTokenizer,BlenderbotTokenizer:()=>u.BlenderbotTokenizer,BloomForCausalLM:()=>l.BloomForCausalLM,BloomModel:()=>l.BloomModel,BloomPreTrainedModel:()=>l.BloomPreTrainedModel,BloomTokenizer:()=>u.BloomTokenizer,CLIPFeatureExtractor:()=>d.CLIPFeatureExtractor,CLIPImageProcessor:()=>d.CLIPImageProcessor,CLIPModel:()=>l.CLIPModel,CLIPPreTrainedModel:()=>l.CLIPPreTrainedModel,CLIPSegForImageSegmentation:()=>l.CLIPSegForImageSegmentation,CLIPSegModel:()=>l.CLIPSegModel,CLIPSegPreTrainedModel:()=>l.CLIPSegPreTrainedModel,CLIPTextModelWithProjection:()=>l.CLIPTextModelWithProjection,CLIPTokenizer:()=>u.CLIPTokenizer,CLIPVisionModelWithProjection:()=>l.CLIPVisionModelWithProjection,CamembertForMaskedLM:()=>l.CamembertForMaskedLM,CamembertForQuestionAnswering:()=>l.CamembertForQuestionAnswering,CamembertForSequenceClassification:()=>l.CamembertForSequenceClassification,CamembertForTokenClassification:()=>l.CamembertForTokenClassification,CamembertModel:()=>l.CamembertModel,CamembertPreTrainedModel:()=>l.CamembertPreTrainedModel,CamembertTokenizer:()=>u.CamembertTokenizer,CausalLMOutput:()=>l.CausalLMOutput,CausalLMOutputWithPast:()=>l.CausalLMOutputWithPast,ChineseCLIPFeatureExtractor:()=>d.ChineseCLIPFeatureExtractor,ChineseCLIPModel:()=>l.ChineseCLIPModel,ChineseCLIPPreTrainedModel:()=>l.ChineseCLIPPreTrainedModel,ClapAudioModelWithProjection:()=>l.ClapAudioModelWithProjection,ClapFeatureExtractor:()=>d.ClapFeatureExtractor,ClapModel:()=>l.ClapModel,ClapPreTrainedModel:()=>l.ClapPreTrainedModel,ClapTextModelWithProjection:()=>l.ClapTextModelWithProjection,CodeGenForCausalLM:()=>l.CodeGenForCausalLM,CodeGenModel:()=>l.CodeGenModel,CodeGenPreTrainedModel:()=>l.CodeGenPreTrainedModel,CodeGenTokenizer:()=>u.CodeGenTokenizer,CodeLlamaTokenizer:()=>u.CodeLlamaTokenizer,CohereForCausalLM:()=>l.CohereForCausalLM,CohereModel:()=>l.CohereModel,CoherePreTrainedModel:()=>l.CoherePreTrainedModel,CohereTokenizer:()=>u.CohereTokenizer,ConvBertForMaskedLM:()=>l.ConvBertForMaskedLM,ConvBertForQuestionAnswering:()=>l.ConvBertForQuestionAnswering,ConvBertForSequenceClassification:()=>l.ConvBertForSequenceClassification,ConvBertForTokenClassification:()=>l.ConvBertForTokenClassification,ConvBertModel:()=>l.ConvBertModel,ConvBertPreTrainedModel:()=>l.ConvBertPreTrainedModel,ConvBertTokenizer:()=>u.ConvBertTokenizer,ConvNextFeatureExtractor:()=>d.ConvNextFeatureExtractor,ConvNextForImageClassification:()=>l.ConvNextForImageClassification,ConvNextImageProcessor:()=>d.ConvNextImageProcessor,ConvNextModel:()=>l.ConvNextModel,ConvNextPreTrainedModel:()=>l.ConvNextPreTrainedModel,ConvNextV2ForImageClassification:()=>l.ConvNextV2ForImageClassification,ConvNextV2Model:()=>l.ConvNextV2Model,ConvNextV2PreTrainedModel:()=>l.ConvNextV2PreTrainedModel,DPTFeatureExtractor:()=>d.DPTFeatureExtractor,DPTForDepthEstimation:()=>l.DPTForDepthEstimation,DPTImageProcessor:()=>d.DPTImageProcessor,DPTModel:()=>l.DPTModel,DPTPreTrainedModel:()=>l.DPTPreTrainedModel,DebertaForMaskedLM:()=>l.DebertaForMaskedLM,DebertaForQuestionAnswering:()=>l.DebertaForQuestionAnswering,DebertaForSequenceClassification:()=>l.DebertaForSequenceClassification,DebertaForTokenClassification:()=>l.DebertaForTokenClassification,DebertaModel:()=>l.DebertaModel,DebertaPreTrainedModel:()=>l.DebertaPreTrainedModel,DebertaTokenizer:()=>u.DebertaTokenizer,DebertaV2ForMaskedLM:()=>l.DebertaV2ForMaskedLM,DebertaV2ForQuestionAnswering:()=>l.DebertaV2ForQuestionAnswering,DebertaV2ForSequenceClassification:()=>l.DebertaV2ForSequenceClassification,DebertaV2ForTokenClassification:()=>l.DebertaV2ForTokenClassification,DebertaV2Model:()=>l.DebertaV2Model,DebertaV2PreTrainedModel:()=>l.DebertaV2PreTrainedModel,DebertaV2Tokenizer:()=>u.DebertaV2Tokenizer,DeiTFeatureExtractor:()=>d.DeiTFeatureExtractor,DeiTForImageClassification:()=>l.DeiTForImageClassification,DeiTModel:()=>l.DeiTModel,DeiTPreTrainedModel:()=>l.DeiTPreTrainedModel,DepthAnythingForDepthEstimation:()=>l.DepthAnythingForDepthEstimation,DepthAnythingPreTrainedModel:()=>l.DepthAnythingPreTrainedModel,DepthEstimationPipeline:()=>o.DepthEstimationPipeline,DetrFeatureExtractor:()=>d.DetrFeatureExtractor,DetrForObjectDetection:()=>l.DetrForObjectDetection,DetrForSegmentation:()=>l.DetrForSegmentation,DetrModel:()=>l.DetrModel,DetrObjectDetectionOutput:()=>l.DetrObjectDetectionOutput,DetrPreTrainedModel:()=>l.DetrPreTrainedModel,DetrSegmentationOutput:()=>l.DetrSegmentationOutput,Dinov2ForImageClassification:()=>l.Dinov2ForImageClassification,Dinov2Model:()=>l.Dinov2Model,Dinov2PreTrainedModel:()=>l.Dinov2PreTrainedModel,DistilBertForMaskedLM:()=>l.DistilBertForMaskedLM,DistilBertForQuestionAnswering:()=>l.DistilBertForQuestionAnswering,DistilBertForSequenceClassification:()=>l.DistilBertForSequenceClassification,DistilBertForTokenClassification:()=>l.DistilBertForTokenClassification,DistilBertModel:()=>l.DistilBertModel,DistilBertPreTrainedModel:()=>l.DistilBertPreTrainedModel,DistilBertTokenizer:()=>u.DistilBertTokenizer,DocumentQuestionAnsweringPipeline:()=>o.DocumentQuestionAnsweringPipeline,DonutFeatureExtractor:()=>d.DonutFeatureExtractor,DonutSwinModel:()=>l.DonutSwinModel,DonutSwinPreTrainedModel:()=>l.DonutSwinPreTrainedModel,EfficientNetForImageClassification:()=>l.EfficientNetForImageClassification,EfficientNetImageProcessor:()=>d.EfficientNetImageProcessor,EfficientNetModel:()=>l.EfficientNetModel,EfficientNetPreTrainedModel:()=>l.EfficientNetPreTrainedModel,ElectraForMaskedLM:()=>l.ElectraForMaskedLM,ElectraForQuestionAnswering:()=>l.ElectraForQuestionAnswering,ElectraForSequenceClassification:()=>l.ElectraForSequenceClassification,ElectraForTokenClassification:()=>l.ElectraForTokenClassification,ElectraModel:()=>l.ElectraModel,ElectraPreTrainedModel:()=>l.ElectraPreTrainedModel,ElectraTokenizer:()=>u.ElectraTokenizer,EosTokenCriteria:()=>_.EosTokenCriteria,EsmForMaskedLM:()=>l.EsmForMaskedLM,EsmForSequenceClassification:()=>l.EsmForSequenceClassification,EsmForTokenClassification:()=>l.EsmForTokenClassification,EsmModel:()=>l.EsmModel,EsmPreTrainedModel:()=>l.EsmPreTrainedModel,EsmTokenizer:()=>u.EsmTokenizer,FFT:()=>f.FFT,FalconForCausalLM:()=>l.FalconForCausalLM,FalconModel:()=>l.FalconModel,FalconPreTrainedModel:()=>l.FalconPreTrainedModel,FalconTokenizer:()=>u.FalconTokenizer,FastViTForImageClassification:()=>l.FastViTForImageClassification,FastViTModel:()=>l.FastViTModel,FastViTPreTrainedModel:()=>l.FastViTPreTrainedModel,FeatureExtractionPipeline:()=>o.FeatureExtractionPipeline,FeatureExtractor:()=>d.FeatureExtractor,FillMaskPipeline:()=>o.FillMaskPipeline,Florence2ForConditionalGeneration:()=>l.Florence2ForConditionalGeneration,Florence2PreTrainedModel:()=>l.Florence2PreTrainedModel,Florence2Processor:()=>d.Florence2Processor,GLPNFeatureExtractor:()=>d.GLPNFeatureExtractor,GLPNForDepthEstimation:()=>l.GLPNForDepthEstimation,GLPNModel:()=>l.GLPNModel,GLPNPreTrainedModel:()=>l.GLPNPreTrainedModel,GPT2LMHeadModel:()=>l.GPT2LMHeadModel,GPT2Model:()=>l.GPT2Model,GPT2PreTrainedModel:()=>l.GPT2PreTrainedModel,GPT2Tokenizer:()=>u.GPT2Tokenizer,GPTBigCodeForCausalLM:()=>l.GPTBigCodeForCausalLM,GPTBigCodeModel:()=>l.GPTBigCodeModel,GPTBigCodePreTrainedModel:()=>l.GPTBigCodePreTrainedModel,GPTJForCausalLM:()=>l.GPTJForCausalLM,GPTJModel:()=>l.GPTJModel,GPTJPreTrainedModel:()=>l.GPTJPreTrainedModel,GPTNeoForCausalLM:()=>l.GPTNeoForCausalLM,GPTNeoModel:()=>l.GPTNeoModel,GPTNeoPreTrainedModel:()=>l.GPTNeoPreTrainedModel,GPTNeoXForCausalLM:()=>l.GPTNeoXForCausalLM,GPTNeoXModel:()=>l.GPTNeoXModel,GPTNeoXPreTrainedModel:()=>l.GPTNeoXPreTrainedModel,GPTNeoXTokenizer:()=>u.GPTNeoXTokenizer,Gemma2ForCausalLM:()=>l.Gemma2ForCausalLM,Gemma2Model:()=>l.Gemma2Model,Gemma2PreTrainedModel:()=>l.Gemma2PreTrainedModel,GemmaForCausalLM:()=>l.GemmaForCausalLM,GemmaModel:()=>l.GemmaModel,GemmaPreTrainedModel:()=>l.GemmaPreTrainedModel,GemmaTokenizer:()=>u.GemmaTokenizer,Grok1Tokenizer:()=>u.Grok1Tokenizer,HerbertTokenizer:()=>u.HerbertTokenizer,HubertForCTC:()=>l.HubertForCTC,HubertForSequenceClassification:()=>l.HubertForSequenceClassification,HubertModel:()=>l.HubertModel,HubertPreTrainedModel:()=>l.HubertPreTrainedModel,ImageClassificationPipeline:()=>o.ImageClassificationPipeline,ImageFeatureExtractionPipeline:()=>o.ImageFeatureExtractionPipeline,ImageFeatureExtractor:()=>d.ImageFeatureExtractor,ImageMattingOutput:()=>l.ImageMattingOutput,ImageSegmentationPipeline:()=>o.ImageSegmentationPipeline,ImageToImagePipeline:()=>o.ImageToImagePipeline,ImageToTextPipeline:()=>o.ImageToTextPipeline,InterruptableStoppingCriteria:()=>_.InterruptableStoppingCriteria,JAISLMHeadModel:()=>l.JAISLMHeadModel,JAISModel:()=>l.JAISModel,JAISPreTrainedModel:()=>l.JAISPreTrainedModel,LlamaForCausalLM:()=>l.LlamaForCausalLM,LlamaModel:()=>l.LlamaModel,LlamaPreTrainedModel:()=>l.LlamaPreTrainedModel,LlamaTokenizer:()=>u.LlamaTokenizer,LlavaForConditionalGeneration:()=>l.LlavaForConditionalGeneration,LlavaPreTrainedModel:()=>l.LlavaPreTrainedModel,LongT5ForConditionalGeneration:()=>l.LongT5ForConditionalGeneration,LongT5Model:()=>l.LongT5Model,LongT5PreTrainedModel:()=>l.LongT5PreTrainedModel,M2M100ForConditionalGeneration:()=>l.M2M100ForConditionalGeneration,M2M100Model:()=>l.M2M100Model,M2M100PreTrainedModel:()=>l.M2M100PreTrainedModel,M2M100Tokenizer:()=>u.M2M100Tokenizer,MBart50Tokenizer:()=>u.MBart50Tokenizer,MBartForCausalLM:()=>l.MBartForCausalLM,MBartForConditionalGeneration:()=>l.MBartForConditionalGeneration,MBartForSequenceClassification:()=>l.MBartForSequenceClassification,MBartModel:()=>l.MBartModel,MBartPreTrainedModel:()=>l.MBartPreTrainedModel,MBartTokenizer:()=>u.MBartTokenizer,MPNetForMaskedLM:()=>l.MPNetForMaskedLM,MPNetForQuestionAnswering:()=>l.MPNetForQuestionAnswering,MPNetForSequenceClassification:()=>l.MPNetForSequenceClassification,MPNetForTokenClassification:()=>l.MPNetForTokenClassification,MPNetModel:()=>l.MPNetModel,MPNetPreTrainedModel:()=>l.MPNetPreTrainedModel,MPNetTokenizer:()=>u.MPNetTokenizer,MT5ForConditionalGeneration:()=>l.MT5ForConditionalGeneration,MT5Model:()=>l.MT5Model,MT5PreTrainedModel:()=>l.MT5PreTrainedModel,MarianMTModel:()=>l.MarianMTModel,MarianModel:()=>l.MarianModel,MarianPreTrainedModel:()=>l.MarianPreTrainedModel,MarianTokenizer:()=>u.MarianTokenizer,MaskedLMOutput:()=>l.MaskedLMOutput,MaxLengthCriteria:()=>_.MaxLengthCriteria,MistralForCausalLM:()=>l.MistralForCausalLM,MistralModel:()=>l.MistralModel,MistralPreTrainedModel:()=>l.MistralPreTrainedModel,MobileBertForMaskedLM:()=>l.MobileBertForMaskedLM,MobileBertForQuestionAnswering:()=>l.MobileBertForQuestionAnswering,MobileBertForSequenceClassification:()=>l.MobileBertForSequenceClassification,MobileBertModel:()=>l.MobileBertModel,MobileBertPreTrainedModel:()=>l.MobileBertPreTrainedModel,MobileBertTokenizer:()=>u.MobileBertTokenizer,MobileNetV1FeatureExtractor:()=>d.MobileNetV1FeatureExtractor,MobileNetV1ForImageClassification:()=>l.MobileNetV1ForImageClassification,MobileNetV1Model:()=>l.MobileNetV1Model,MobileNetV1PreTrainedModel:()=>l.MobileNetV1PreTrainedModel,MobileNetV2FeatureExtractor:()=>d.MobileNetV2FeatureExtractor,MobileNetV2ForImageClassification:()=>l.MobileNetV2ForImageClassification,MobileNetV2Model:()=>l.MobileNetV2Model,MobileNetV2PreTrainedModel:()=>l.MobileNetV2PreTrainedModel,MobileNetV3FeatureExtractor:()=>d.MobileNetV3FeatureExtractor,MobileNetV3ForImageClassification:()=>l.MobileNetV3ForImageClassification,MobileNetV3Model:()=>l.MobileNetV3Model,MobileNetV3PreTrainedModel:()=>l.MobileNetV3PreTrainedModel,MobileNetV4FeatureExtractor:()=>d.MobileNetV4FeatureExtractor,MobileNetV4ForImageClassification:()=>l.MobileNetV4ForImageClassification,MobileNetV4Model:()=>l.MobileNetV4Model,MobileNetV4PreTrainedModel:()=>l.MobileNetV4PreTrainedModel,MobileViTFeatureExtractor:()=>d.MobileViTFeatureExtractor,MobileViTForImageClassification:()=>l.MobileViTForImageClassification,MobileViTImageProcessor:()=>d.MobileViTImageProcessor,MobileViTModel:()=>l.MobileViTModel,MobileViTPreTrainedModel:()=>l.MobileViTPreTrainedModel,MobileViTV2ForImageClassification:()=>l.MobileViTV2ForImageClassification,MobileViTV2Model:()=>l.MobileViTV2Model,MobileViTV2PreTrainedModel:()=>l.MobileViTV2PreTrainedModel,ModelOutput:()=>l.ModelOutput,Moondream1ForConditionalGeneration:()=>l.Moondream1ForConditionalGeneration,MptForCausalLM:()=>l.MptForCausalLM,MptModel:()=>l.MptModel,MptPreTrainedModel:()=>l.MptPreTrainedModel,MusicgenForCausalLM:()=>l.MusicgenForCausalLM,MusicgenForConditionalGeneration:()=>l.MusicgenForConditionalGeneration,MusicgenModel:()=>l.MusicgenModel,MusicgenPreTrainedModel:()=>l.MusicgenPreTrainedModel,NllbTokenizer:()=>u.NllbTokenizer,NomicBertModel:()=>l.NomicBertModel,NomicBertPreTrainedModel:()=>l.NomicBertPreTrainedModel,NougatImageProcessor:()=>d.NougatImageProcessor,NougatTokenizer:()=>u.NougatTokenizer,OPTForCausalLM:()=>l.OPTForCausalLM,OPTModel:()=>l.OPTModel,OPTPreTrainedModel:()=>l.OPTPreTrainedModel,ObjectDetectionPipeline:()=>o.ObjectDetectionPipeline,OpenELMForCausalLM:()=>l.OpenELMForCausalLM,OpenELMModel:()=>l.OpenELMModel,OpenELMPreTrainedModel:()=>l.OpenELMPreTrainedModel,OwlViTFeatureExtractor:()=>d.OwlViTFeatureExtractor,OwlViTForObjectDetection:()=>l.OwlViTForObjectDetection,OwlViTModel:()=>l.OwlViTModel,OwlViTPreTrainedModel:()=>l.OwlViTPreTrainedModel,OwlViTProcessor:()=>d.OwlViTProcessor,Owlv2ForObjectDetection:()=>l.Owlv2ForObjectDetection,Owlv2ImageProcessor:()=>d.Owlv2ImageProcessor,Owlv2Model:()=>l.Owlv2Model,Owlv2PreTrainedModel:()=>l.Owlv2PreTrainedModel,Phi3ForCausalLM:()=>l.Phi3ForCausalLM,Phi3Model:()=>l.Phi3Model,Phi3PreTrainedModel:()=>l.Phi3PreTrainedModel,PhiForCausalLM:()=>l.PhiForCausalLM,PhiModel:()=>l.PhiModel,PhiPreTrainedModel:()=>l.PhiPreTrainedModel,Pipeline:()=>o.Pipeline,PreTrainedModel:()=>l.PreTrainedModel,PreTrainedTokenizer:()=>u.PreTrainedTokenizer,PretrainedConfig:()=>c.PretrainedConfig,PretrainedMixin:()=>l.PretrainedMixin,Processor:()=>d.Processor,PyAnnoteFeatureExtractor:()=>d.PyAnnoteFeatureExtractor,PyAnnoteForAudioFrameClassification:()=>l.PyAnnoteForAudioFrameClassification,PyAnnoteModel:()=>l.PyAnnoteModel,PyAnnotePreTrainedModel:()=>l.PyAnnotePreTrainedModel,PyAnnoteProcessor:()=>d.PyAnnoteProcessor,QuestionAnsweringModelOutput:()=>l.QuestionAnsweringModelOutput,QuestionAnsweringPipeline:()=>o.QuestionAnsweringPipeline,Qwen2ForCausalLM:()=>l.Qwen2ForCausalLM,Qwen2Model:()=>l.Qwen2Model,Qwen2PreTrainedModel:()=>l.Qwen2PreTrainedModel,Qwen2Tokenizer:()=>u.Qwen2Tokenizer,RTDetrForObjectDetection:()=>l.RTDetrForObjectDetection,RTDetrImageProcessor:()=>d.RTDetrImageProcessor,RTDetrModel:()=>l.RTDetrModel,RTDetrObjectDetectionOutput:()=>l.RTDetrObjectDetectionOutput,RTDetrPreTrainedModel:()=>l.RTDetrPreTrainedModel,RawImage:()=>h.RawImage,ResNetForImageClassification:()=>l.ResNetForImageClassification,ResNetModel:()=>l.ResNetModel,ResNetPreTrainedModel:()=>l.ResNetPreTrainedModel,RoFormerForMaskedLM:()=>l.RoFormerForMaskedLM,RoFormerForQuestionAnswering:()=>l.RoFormerForQuestionAnswering,RoFormerForSequenceClassification:()=>l.RoFormerForSequenceClassification,RoFormerForTokenClassification:()=>l.RoFormerForTokenClassification,RoFormerModel:()=>l.RoFormerModel,RoFormerPreTrainedModel:()=>l.RoFormerPreTrainedModel,RoFormerTokenizer:()=>u.RoFormerTokenizer,RobertaForMaskedLM:()=>l.RobertaForMaskedLM,RobertaForQuestionAnswering:()=>l.RobertaForQuestionAnswering,RobertaForSequenceClassification:()=>l.RobertaForSequenceClassification,RobertaForTokenClassification:()=>l.RobertaForTokenClassification,RobertaModel:()=>l.RobertaModel,RobertaPreTrainedModel:()=>l.RobertaPreTrainedModel,RobertaTokenizer:()=>u.RobertaTokenizer,SamImageProcessor:()=>d.SamImageProcessor,SamImageSegmentationOutput:()=>l.SamImageSegmentationOutput,SamModel:()=>l.SamModel,SamPreTrainedModel:()=>l.SamPreTrainedModel,SamProcessor:()=>d.SamProcessor,SapiensFeatureExtractor:()=>d.SapiensFeatureExtractor,SapiensForDepthEstimation:()=>l.SapiensForDepthEstimation,SapiensForNormalEstimation:()=>l.SapiensForNormalEstimation,SapiensForSemanticSegmentation:()=>l.SapiensForSemanticSegmentation,SapiensPreTrainedModel:()=>l.SapiensPreTrainedModel,SeamlessM4TFeatureExtractor:()=>d.SeamlessM4TFeatureExtractor,SegformerFeatureExtractor:()=>d.SegformerFeatureExtractor,SegformerForImageClassification:()=>l.SegformerForImageClassification,SegformerForSemanticSegmentation:()=>l.SegformerForSemanticSegmentation,SegformerModel:()=>l.SegformerModel,SegformerPreTrainedModel:()=>l.SegformerPreTrainedModel,Seq2SeqLMOutput:()=>l.Seq2SeqLMOutput,SequenceClassifierOutput:()=>l.SequenceClassifierOutput,SiglipImageProcessor:()=>d.SiglipImageProcessor,SiglipModel:()=>l.SiglipModel,SiglipPreTrainedModel:()=>l.SiglipPreTrainedModel,SiglipTextModel:()=>l.SiglipTextModel,SiglipTokenizer:()=>u.SiglipTokenizer,SiglipVisionModel:()=>l.SiglipVisionModel,SpeechT5FeatureExtractor:()=>d.SpeechT5FeatureExtractor,SpeechT5ForSpeechToText:()=>l.SpeechT5ForSpeechToText,SpeechT5ForTextToSpeech:()=>l.SpeechT5ForTextToSpeech,SpeechT5HifiGan:()=>l.SpeechT5HifiGan,SpeechT5Model:()=>l.SpeechT5Model,SpeechT5PreTrainedModel:()=>l.SpeechT5PreTrainedModel,SpeechT5Processor:()=>d.SpeechT5Processor,SpeechT5Tokenizer:()=>u.SpeechT5Tokenizer,SqueezeBertForMaskedLM:()=>l.SqueezeBertForMaskedLM,SqueezeBertForQuestionAnswering:()=>l.SqueezeBertForQuestionAnswering,SqueezeBertForSequenceClassification:()=>l.SqueezeBertForSequenceClassification,SqueezeBertModel:()=>l.SqueezeBertModel,SqueezeBertPreTrainedModel:()=>l.SqueezeBertPreTrainedModel,SqueezeBertTokenizer:()=>u.SqueezeBertTokenizer,StableLmForCausalLM:()=>l.StableLmForCausalLM,StableLmModel:()=>l.StableLmModel,StableLmPreTrainedModel:()=>l.StableLmPreTrainedModel,Starcoder2ForCausalLM:()=>l.Starcoder2ForCausalLM,Starcoder2Model:()=>l.Starcoder2Model,Starcoder2PreTrainedModel:()=>l.Starcoder2PreTrainedModel,StoppingCriteria:()=>_.StoppingCriteria,StoppingCriteriaList:()=>_.StoppingCriteriaList,SummarizationPipeline:()=>o.SummarizationPipeline,Swin2SRForImageSuperResolution:()=>l.Swin2SRForImageSuperResolution,Swin2SRImageProcessor:()=>d.Swin2SRImageProcessor,Swin2SRModel:()=>l.Swin2SRModel,Swin2SRPreTrainedModel:()=>l.Swin2SRPreTrainedModel,SwinForImageClassification:()=>l.SwinForImageClassification,SwinModel:()=>l.SwinModel,SwinPreTrainedModel:()=>l.SwinPreTrainedModel,T5ForConditionalGeneration:()=>l.T5ForConditionalGeneration,T5Model:()=>l.T5Model,T5PreTrainedModel:()=>l.T5PreTrainedModel,T5Tokenizer:()=>u.T5Tokenizer,TableTransformerForObjectDetection:()=>l.TableTransformerForObjectDetection,TableTransformerModel:()=>l.TableTransformerModel,TableTransformerObjectDetectionOutput:()=>l.TableTransformerObjectDetectionOutput,TableTransformerPreTrainedModel:()=>l.TableTransformerPreTrainedModel,Tensor:()=>m.Tensor,Text2TextGenerationPipeline:()=>o.Text2TextGenerationPipeline,TextClassificationPipeline:()=>o.TextClassificationPipeline,TextGenerationPipeline:()=>o.TextGenerationPipeline,TextStreamer:()=>g.TextStreamer,TextToAudioPipeline:()=>o.TextToAudioPipeline,TokenClassificationPipeline:()=>o.TokenClassificationPipeline,TokenClassifierOutput:()=>l.TokenClassifierOutput,TokenizerModel:()=>u.TokenizerModel,TrOCRForCausalLM:()=>l.TrOCRForCausalLM,TrOCRPreTrainedModel:()=>l.TrOCRPreTrainedModel,TranslationPipeline:()=>o.TranslationPipeline,UniSpeechForCTC:()=>l.UniSpeechForCTC,UniSpeechForSequenceClassification:()=>l.UniSpeechForSequenceClassification,UniSpeechModel:()=>l.UniSpeechModel,UniSpeechPreTrainedModel:()=>l.UniSpeechPreTrainedModel,UniSpeechSatForAudioFrameClassification:()=>l.UniSpeechSatForAudioFrameClassification,UniSpeechSatForCTC:()=>l.UniSpeechSatForCTC,UniSpeechSatForSequenceClassification:()=>l.UniSpeechSatForSequenceClassification,UniSpeechSatModel:()=>l.UniSpeechSatModel,UniSpeechSatPreTrainedModel:()=>l.UniSpeechSatPreTrainedModel,ViTFeatureExtractor:()=>d.ViTFeatureExtractor,ViTForImageClassification:()=>l.ViTForImageClassification,ViTImageProcessor:()=>d.ViTImageProcessor,ViTModel:()=>l.ViTModel,ViTPreTrainedModel:()=>l.ViTPreTrainedModel,VisionEncoderDecoderModel:()=>l.VisionEncoderDecoderModel,VitMatteForImageMatting:()=>l.VitMatteForImageMatting,VitMatteImageProcessor:()=>d.VitMatteImageProcessor,VitMattePreTrainedModel:()=>l.VitMattePreTrainedModel,VitsModel:()=>l.VitsModel,VitsModelOutput:()=>l.VitsModelOutput,VitsPreTrainedModel:()=>l.VitsPreTrainedModel,VitsTokenizer:()=>u.VitsTokenizer,Wav2Vec2BertForCTC:()=>l.Wav2Vec2BertForCTC,Wav2Vec2BertForSequenceClassification:()=>l.Wav2Vec2BertForSequenceClassification,Wav2Vec2BertModel:()=>l.Wav2Vec2BertModel,Wav2Vec2BertPreTrainedModel:()=>l.Wav2Vec2BertPreTrainedModel,Wav2Vec2CTCTokenizer:()=>u.Wav2Vec2CTCTokenizer,Wav2Vec2FeatureExtractor:()=>d.Wav2Vec2FeatureExtractor,Wav2Vec2ForAudioFrameClassification:()=>l.Wav2Vec2ForAudioFrameClassification,Wav2Vec2ForCTC:()=>l.Wav2Vec2ForCTC,Wav2Vec2ForSequenceClassification:()=>l.Wav2Vec2ForSequenceClassification,Wav2Vec2Model:()=>l.Wav2Vec2Model,Wav2Vec2PreTrainedModel:()=>l.Wav2Vec2PreTrainedModel,Wav2Vec2ProcessorWithLM:()=>d.Wav2Vec2ProcessorWithLM,WavLMForAudioFrameClassification:()=>l.WavLMForAudioFrameClassification,WavLMForCTC:()=>l.WavLMForCTC,WavLMForSequenceClassification:()=>l.WavLMForSequenceClassification,WavLMForXVector:()=>l.WavLMForXVector,WavLMModel:()=>l.WavLMModel,WavLMPreTrainedModel:()=>l.WavLMPreTrainedModel,WeSpeakerFeatureExtractor:()=>d.WeSpeakerFeatureExtractor,WeSpeakerResNetModel:()=>l.WeSpeakerResNetModel,WeSpeakerResNetPreTrainedModel:()=>l.WeSpeakerResNetPreTrainedModel,WhisperFeatureExtractor:()=>d.WhisperFeatureExtractor,WhisperForConditionalGeneration:()=>l.WhisperForConditionalGeneration,WhisperModel:()=>l.WhisperModel,WhisperPreTrainedModel:()=>l.WhisperPreTrainedModel,WhisperProcessor:()=>d.WhisperProcessor,WhisperTextStreamer:()=>g.WhisperTextStreamer,WhisperTokenizer:()=>u.WhisperTokenizer,XLMForQuestionAnswering:()=>l.XLMForQuestionAnswering,XLMForSequenceClassification:()=>l.XLMForSequenceClassification,XLMForTokenClassification:()=>l.XLMForTokenClassification,XLMModel:()=>l.XLMModel,XLMPreTrainedModel:()=>l.XLMPreTrainedModel,XLMRobertaForMaskedLM:()=>l.XLMRobertaForMaskedLM,XLMRobertaForQuestionAnswering:()=>l.XLMRobertaForQuestionAnswering,XLMRobertaForSequenceClassification:()=>l.XLMRobertaForSequenceClassification,XLMRobertaForTokenClassification:()=>l.XLMRobertaForTokenClassification,XLMRobertaModel:()=>l.XLMRobertaModel,XLMRobertaPreTrainedModel:()=>l.XLMRobertaPreTrainedModel,XLMRobertaTokenizer:()=>u.XLMRobertaTokenizer,XLMTokenizer:()=>u.XLMTokenizer,XLMWithLMHeadModel:()=>l.XLMWithLMHeadModel,XVectorOutput:()=>l.XVectorOutput,YolosFeatureExtractor:()=>d.YolosFeatureExtractor,YolosForObjectDetection:()=>l.YolosForObjectDetection,YolosModel:()=>l.YolosModel,YolosObjectDetectionOutput:()=>l.YolosObjectDetectionOutput,YolosPreTrainedModel:()=>l.YolosPreTrainedModel,ZeroShotAudioClassificationPipeline:()=>o.ZeroShotAudioClassificationPipeline,ZeroShotClassificationPipeline:()=>o.ZeroShotClassificationPipeline,ZeroShotImageClassificationPipeline:()=>o.ZeroShotImageClassificationPipeline,ZeroShotObjectDetectionPipeline:()=>o.ZeroShotObjectDetectionPipeline,bankers_round:()=>f.bankers_round,cat:()=>m.cat,cos_sim:()=>f.cos_sim,dot:()=>f.dot,dynamic_time_warping:()=>f.dynamic_time_warping,env:()=>s.env,full:()=>m.full,full_like:()=>m.full_like,getKeyValueShapes:()=>c.getKeyValueShapes,hamming:()=>p.hamming,hanning:()=>p.hanning,interpolate:()=>m.interpolate,interpolate_4d:()=>m.interpolate_4d,interpolate_data:()=>f.interpolate_data,is_chinese_char:()=>u.is_chinese_char,layer_norm:()=>m.layer_norm,log_softmax:()=>f.log_softmax,magnitude:()=>f.magnitude,matmul:()=>m.matmul,max:()=>f.max,mean:()=>m.mean,mean_pooling:()=>m.mean_pooling,medianFilter:()=>f.medianFilter,mel_filter_bank:()=>p.mel_filter_bank,min:()=>f.min,ones:()=>m.ones,ones_like:()=>m.ones_like,permute:()=>m.permute,permute_data:()=>f.permute_data,pipeline:()=>o.pipeline,quantize_embeddings:()=>m.quantize_embeddings,read_audio:()=>p.read_audio,rfft:()=>m.rfft,round:()=>f.round,softmax:()=>f.softmax,spectrogram:()=>p.spectrogram,stack:()=>m.stack,std_mean:()=>m.std_mean,topk:()=>m.topk,window_function:()=>p.window_function,zeros:()=>m.zeros,zeros_like:()=>m.zeros_like});var s=i(/*! ./env.js */"./src/env.js"),o=i(/*! ./pipelines.js */"./src/pipelines.js"),l=i(/*! ./models.js */"./src/models.js"),u=i(/*! ./tokenizers.js */"./src/tokenizers.js"),d=i(/*! ./processors.js */"./src/processors.js"),c=i(/*! ./configs.js */"./src/configs.js"),p=i(/*! ./utils/audio.js */"./src/utils/audio.js"),h=i(/*! ./utils/image.js */"./src/utils/image.js"),m=i(/*! ./utils/tensor.js */"./src/utils/tensor.js"),f=i(/*! ./utils/maths.js */"./src/utils/maths.js"),g=i(/*! ./generation/streamers.js */"./src/generation/streamers.js"),_=i(/*! ./generation/stopping_criteria.js */"./src/generation/stopping_criteria.js"),w=a.ASTFeatureExtractor,y=a.ASTForAudioClassification,b=a.ASTModel,v=a.ASTPreTrainedModel,x=a.AlbertForMaskedLM,M=a.AlbertForQuestionAnswering,T=a.AlbertForSequenceClassification,k=a.AlbertModel,$=a.AlbertPreTrainedModel,S=a.AlbertTokenizer,C=a.AudioClassificationPipeline,E=a.AutoConfig,P=a.AutoModel,A=a.AutoModelForAudioClassification,F=a.AutoModelForAudioFrameClassification,z=a.AutoModelForCTC,I=a.AutoModelForCausalLM,O=a.AutoModelForDepthEstimation,B=a.AutoModelForDocumentQuestionAnswering,L=a.AutoModelForImageClassification,D=a.AutoModelForImageFeatureExtraction,R=a.AutoModelForImageMatting,N=a.AutoModelForImageSegmentation,V=a.AutoModelForImageToImage,j=a.AutoModelForMaskGeneration,G=a.AutoModelForMaskedLM,q=a.AutoModelForObjectDetection,U=a.AutoModelForQuestionAnswering,W=a.AutoModelForSemanticSegmentation,H=a.AutoModelForSeq2SeqLM,X=a.AutoModelForSequenceClassification,K=a.AutoModelForSpeechSeq2Seq,Q=a.AutoModelForTextToSpectrogram,Y=a.AutoModelForTextToWaveform,J=a.AutoModelForTokenClassification,Z=a.AutoModelForVision2Seq,ee=a.AutoModelForXVector,te=a.AutoModelForZeroShotObjectDetection,ne=a.AutoProcessor,re=a.AutoTokenizer,ie=a.AutomaticSpeechRecognitionPipeline,ae=a.BartForConditionalGeneration,se=a.BartForSequenceClassification,oe=a.BartModel,le=a.BartPretrainedModel,ue=a.BartTokenizer,de=a.BaseModelOutput,ce=a.BaseStreamer,pe=a.BeitFeatureExtractor,he=a.BeitForImageClassification,me=a.BeitModel,fe=a.BeitPreTrainedModel,ge=a.BertForMaskedLM,_e=a.BertForQuestionAnswering,we=a.BertForSequenceClassification,ye=a.BertForTokenClassification,be=a.BertModel,ve=a.BertPreTrainedModel,xe=a.BertTokenizer,Me=a.BitImageProcessor,Te=a.BlenderbotForConditionalGeneration,ke=a.BlenderbotModel,$e=a.BlenderbotPreTrainedModel,Se=a.BlenderbotSmallForConditionalGeneration,Ce=a.BlenderbotSmallModel,Ee=a.BlenderbotSmallPreTrainedModel,Pe=a.BlenderbotSmallTokenizer,Ae=a.BlenderbotTokenizer,Fe=a.BloomForCausalLM,ze=a.BloomModel,Ie=a.BloomPreTrainedModel,Oe=a.BloomTokenizer,Be=a.CLIPFeatureExtractor,Le=a.CLIPImageProcessor,De=a.CLIPModel,Re=a.CLIPPreTrainedModel,Ne=a.CLIPSegForImageSegmentation,Ve=a.CLIPSegModel,je=a.CLIPSegPreTrainedModel,Ge=a.CLIPTextModelWithProjection,qe=a.CLIPTokenizer,Ue=a.CLIPVisionModelWithProjection,We=a.CamembertForMaskedLM,He=a.CamembertForQuestionAnswering,Xe=a.CamembertForSequenceClassification,Ke=a.CamembertForTokenClassification,Qe=a.CamembertModel,Ye=a.CamembertPreTrainedModel,Je=a.CamembertTokenizer,Ze=a.CausalLMOutput,et=a.CausalLMOutputWithPast,tt=a.ChineseCLIPFeatureExtractor,nt=a.ChineseCLIPModel,rt=a.ChineseCLIPPreTrainedModel,it=a.ClapAudioModelWithProjection,at=a.ClapFeatureExtractor,st=a.ClapModel,ot=a.ClapPreTrainedModel,lt=a.ClapTextModelWithProjection,ut=a.CodeGenForCausalLM,dt=a.CodeGenModel,ct=a.CodeGenPreTrainedModel,pt=a.CodeGenTokenizer,ht=a.CodeLlamaTokenizer,mt=a.CohereForCausalLM,ft=a.CohereModel,gt=a.CoherePreTrainedModel,_t=a.CohereTokenizer,wt=a.ConvBertForMaskedLM,yt=a.ConvBertForQuestionAnswering,bt=a.ConvBertForSequenceClassification,vt=a.ConvBertForTokenClassification,xt=a.ConvBertModel,Mt=a.ConvBertPreTrainedModel,Tt=a.ConvBertTokenizer,kt=a.ConvNextFeatureExtractor,$t=a.ConvNextForImageClassification,St=a.ConvNextImageProcessor,Ct=a.ConvNextModel,Et=a.ConvNextPreTrainedModel,Pt=a.ConvNextV2ForImageClassification,At=a.ConvNextV2Model,Ft=a.ConvNextV2PreTrainedModel,zt=a.DPTFeatureExtractor,It=a.DPTForDepthEstimation,Ot=a.DPTImageProcessor,Bt=a.DPTModel,Lt=a.DPTPreTrainedModel,Dt=a.DebertaForMaskedLM,Rt=a.DebertaForQuestionAnswering,Nt=a.DebertaForSequenceClassification,Vt=a.DebertaForTokenClassification,jt=a.DebertaModel,Gt=a.DebertaPreTrainedModel,qt=a.DebertaTokenizer,Ut=a.DebertaV2ForMaskedLM,Wt=a.DebertaV2ForQuestionAnswering,Ht=a.DebertaV2ForSequenceClassification,Xt=a.DebertaV2ForTokenClassification,Kt=a.DebertaV2Model,Qt=a.DebertaV2PreTrainedModel,Yt=a.DebertaV2Tokenizer,Jt=a.DeiTFeatureExtractor,Zt=a.DeiTForImageClassification,en=a.DeiTModel,tn=a.DeiTPreTrainedModel,nn=a.DepthAnythingForDepthEstimation,rn=a.DepthAnythingPreTrainedModel,an=a.DepthEstimationPipeline,sn=a.DetrFeatureExtractor,on=a.DetrForObjectDetection,ln=a.DetrForSegmentation,un=a.DetrModel,dn=a.DetrObjectDetectionOutput,cn=a.DetrPreTrainedModel,pn=a.DetrSegmentationOutput,hn=a.Dinov2ForImageClassification,mn=a.Dinov2Model,fn=a.Dinov2PreTrainedModel,gn=a.DistilBertForMaskedLM,_n=a.DistilBertForQuestionAnswering,wn=a.DistilBertForSequenceClassification,yn=a.DistilBertForTokenClassification,bn=a.DistilBertModel,vn=a.DistilBertPreTrainedModel,xn=a.DistilBertTokenizer,Mn=a.DocumentQuestionAnsweringPipeline,Tn=a.DonutFeatureExtractor,kn=a.DonutSwinModel,$n=a.DonutSwinPreTrainedModel,Sn=a.EfficientNetForImageClassification,Cn=a.EfficientNetImageProcessor,En=a.EfficientNetModel,Pn=a.EfficientNetPreTrainedModel,An=a.ElectraForMaskedLM,Fn=a.ElectraForQuestionAnswering,zn=a.ElectraForSequenceClassification,In=a.ElectraForTokenClassification,On=a.ElectraModel,Bn=a.ElectraPreTrainedModel,Ln=a.ElectraTokenizer,Dn=a.EosTokenCriteria,Rn=a.EsmForMaskedLM,Nn=a.EsmForSequenceClassification,Vn=a.EsmForTokenClassification,jn=a.EsmModel,Gn=a.EsmPreTrainedModel,qn=a.EsmTokenizer,Un=a.FFT,Wn=a.FalconForCausalLM,Hn=a.FalconModel,Xn=a.FalconPreTrainedModel,Kn=a.FalconTokenizer,Qn=a.FastViTForImageClassification,Yn=a.FastViTModel,Jn=a.FastViTPreTrainedModel,Zn=a.FeatureExtractionPipeline,er=a.FeatureExtractor,tr=a.FillMaskPipeline,nr=a.Florence2ForConditionalGeneration,rr=a.Florence2PreTrainedModel,ir=a.Florence2Processor,ar=a.GLPNFeatureExtractor,sr=a.GLPNForDepthEstimation,or=a.GLPNModel,lr=a.GLPNPreTrainedModel,ur=a.GPT2LMHeadModel,dr=a.GPT2Model,cr=a.GPT2PreTrainedModel,pr=a.GPT2Tokenizer,hr=a.GPTBigCodeForCausalLM,mr=a.GPTBigCodeModel,fr=a.GPTBigCodePreTrainedModel,gr=a.GPTJForCausalLM,_r=a.GPTJModel,wr=a.GPTJPreTrainedModel,yr=a.GPTNeoForCausalLM,br=a.GPTNeoModel,vr=a.GPTNeoPreTrainedModel,xr=a.GPTNeoXForCausalLM,Mr=a.GPTNeoXModel,Tr=a.GPTNeoXPreTrainedModel,kr=a.GPTNeoXTokenizer,$r=a.Gemma2ForCausalLM,Sr=a.Gemma2Model,Cr=a.Gemma2PreTrainedModel,Er=a.GemmaForCausalLM,Pr=a.GemmaModel,Ar=a.GemmaPreTrainedModel,Fr=a.GemmaTokenizer,zr=a.Grok1Tokenizer,Ir=a.HerbertTokenizer,Or=a.HubertForCTC,Br=a.HubertForSequenceClassification,Lr=a.HubertModel,Dr=a.HubertPreTrainedModel,Rr=a.ImageClassificationPipeline,Nr=a.ImageFeatureExtractionPipeline,Vr=a.ImageFeatureExtractor,jr=a.ImageMattingOutput,Gr=a.ImageSegmentationPipeline,qr=a.ImageToImagePipeline,Ur=a.ImageToTextPipeline,Wr=a.InterruptableStoppingCriteria,Hr=a.JAISLMHeadModel,Xr=a.JAISModel,Kr=a.JAISPreTrainedModel,Qr=a.LlamaForCausalLM,Yr=a.LlamaModel,Jr=a.LlamaPreTrainedModel,Zr=a.LlamaTokenizer,ei=a.LlavaForConditionalGeneration,ti=a.LlavaPreTrainedModel,ni=a.LongT5ForConditionalGeneration,ri=a.LongT5Model,ii=a.LongT5PreTrainedModel,ai=a.M2M100ForConditionalGeneration,si=a.M2M100Model,oi=a.M2M100PreTrainedModel,li=a.M2M100Tokenizer,ui=a.MBart50Tokenizer,di=a.MBartForCausalLM,ci=a.MBartForConditionalGeneration,pi=a.MBartForSequenceClassification,hi=a.MBartModel,mi=a.MBartPreTrainedModel,fi=a.MBartTokenizer,gi=a.MPNetForMaskedLM,_i=a.MPNetForQuestionAnswering,wi=a.MPNetForSequenceClassification,yi=a.MPNetForTokenClassification,bi=a.MPNetModel,vi=a.MPNetPreTrainedModel,xi=a.MPNetTokenizer,Mi=a.MT5ForConditionalGeneration,Ti=a.MT5Model,ki=a.MT5PreTrainedModel,$i=a.MarianMTModel,Si=a.MarianModel,Ci=a.MarianPreTrainedModel,Ei=a.MarianTokenizer,Pi=a.MaskedLMOutput,Ai=a.MaxLengthCriteria,Fi=a.MistralForCausalLM,zi=a.MistralModel,Ii=a.MistralPreTrainedModel,Oi=a.MobileBertForMaskedLM,Bi=a.MobileBertForQuestionAnswering,Li=a.MobileBertForSequenceClassification,Di=a.MobileBertModel,Ri=a.MobileBertPreTrainedModel,Ni=a.MobileBertTokenizer,Vi=a.MobileNetV1FeatureExtractor,ji=a.MobileNetV1ForImageClassification,Gi=a.MobileNetV1Model,qi=a.MobileNetV1PreTrainedModel,Ui=a.MobileNetV2FeatureExtractor,Wi=a.MobileNetV2ForImageClassification,Hi=a.MobileNetV2Model,Xi=a.MobileNetV2PreTrainedModel,Ki=a.MobileNetV3FeatureExtractor,Qi=a.MobileNetV3ForImageClassification,Yi=a.MobileNetV3Model,Ji=a.MobileNetV3PreTrainedModel,Zi=a.MobileNetV4FeatureExtractor,ea=a.MobileNetV4ForImageClassification,ta=a.MobileNetV4Model,na=a.MobileNetV4PreTrainedModel,ra=a.MobileViTFeatureExtractor,ia=a.MobileViTForImageClassification,aa=a.MobileViTImageProcessor,sa=a.MobileViTModel,oa=a.MobileViTPreTrainedModel,la=a.MobileViTV2ForImageClassification,ua=a.MobileViTV2Model,da=a.MobileViTV2PreTrainedModel,ca=a.ModelOutput,pa=a.Moondream1ForConditionalGeneration,ha=a.MptForCausalLM,ma=a.MptModel,fa=a.MptPreTrainedModel,ga=a.MusicgenForCausalLM,_a=a.MusicgenForConditionalGeneration,wa=a.MusicgenModel,ya=a.MusicgenPreTrainedModel,ba=a.NllbTokenizer,va=a.NomicBertModel,xa=a.NomicBertPreTrainedModel,Ma=a.NougatImageProcessor,Ta=a.NougatTokenizer,ka=a.OPTForCausalLM,$a=a.OPTModel,Sa=a.OPTPreTrainedModel,Ca=a.ObjectDetectionPipeline,Ea=a.OpenELMForCausalLM,Pa=a.OpenELMModel,Aa=a.OpenELMPreTrainedModel,Fa=a.OwlViTFeatureExtractor,za=a.OwlViTForObjectDetection,Ia=a.OwlViTModel,Oa=a.OwlViTPreTrainedModel,Ba=a.OwlViTProcessor,La=a.Owlv2ForObjectDetection,Da=a.Owlv2ImageProcessor,Ra=a.Owlv2Model,Na=a.Owlv2PreTrainedModel,Va=a.Phi3ForCausalLM,ja=a.Phi3Model,Ga=a.Phi3PreTrainedModel,qa=a.PhiForCausalLM,Ua=a.PhiModel,Wa=a.PhiPreTrainedModel,Ha=a.Pipeline,Xa=a.PreTrainedModel,Ka=a.PreTrainedTokenizer,Qa=a.PretrainedConfig,Ya=a.PretrainedMixin,Ja=a.Processor,Za=a.PyAnnoteFeatureExtractor,es=a.PyAnnoteForAudioFrameClassification,ts=a.PyAnnoteModel,ns=a.PyAnnotePreTrainedModel,rs=a.PyAnnoteProcessor,is=a.QuestionAnsweringModelOutput,as=a.QuestionAnsweringPipeline,ss=a.Qwen2ForCausalLM,os=a.Qwen2Model,ls=a.Qwen2PreTrainedModel,us=a.Qwen2Tokenizer,ds=a.RTDetrForObjectDetection,cs=a.RTDetrImageProcessor,ps=a.RTDetrModel,hs=a.RTDetrObjectDetectionOutput,ms=a.RTDetrPreTrainedModel,fs=a.RawImage,gs=a.ResNetForImageClassification,_s=a.ResNetModel,ws=a.ResNetPreTrainedModel,ys=a.RoFormerForMaskedLM,bs=a.RoFormerForQuestionAnswering,vs=a.RoFormerForSequenceClassification,xs=a.RoFormerForTokenClassification,Ms=a.RoFormerModel,Ts=a.RoFormerPreTrainedModel,ks=a.RoFormerTokenizer,$s=a.RobertaForMaskedLM,Ss=a.RobertaForQuestionAnswering,Cs=a.RobertaForSequenceClassification,Es=a.RobertaForTokenClassification,Ps=a.RobertaModel,As=a.RobertaPreTrainedModel,Fs=a.RobertaTokenizer,zs=a.SamImageProcessor,Is=a.SamImageSegmentationOutput,Os=a.SamModel,Bs=a.SamPreTrainedModel,Ls=a.SamProcessor,Ds=a.SapiensFeatureExtractor,Rs=a.SapiensForDepthEstimation,Ns=a.SapiensForNormalEstimation,Vs=a.SapiensForSemanticSegmentation,js=a.SapiensPreTrainedModel,Gs=a.SeamlessM4TFeatureExtractor,qs=a.SegformerFeatureExtractor,Us=a.SegformerForImageClassification,Ws=a.SegformerForSemanticSegmentation,Hs=a.SegformerModel,Xs=a.SegformerPreTrainedModel,Ks=a.Seq2SeqLMOutput,Qs=a.SequenceClassifierOutput,Ys=a.SiglipImageProcessor,Js=a.SiglipModel,Zs=a.SiglipPreTrainedModel,eo=a.SiglipTextModel,to=a.SiglipTokenizer,no=a.SiglipVisionModel,ro=a.SpeechT5FeatureExtractor,io=a.SpeechT5ForSpeechToText,ao=a.SpeechT5ForTextToSpeech,so=a.SpeechT5HifiGan,oo=a.SpeechT5Model,lo=a.SpeechT5PreTrainedModel,uo=a.SpeechT5Processor,co=a.SpeechT5Tokenizer,po=a.SqueezeBertForMaskedLM,ho=a.SqueezeBertForQuestionAnswering,mo=a.SqueezeBertForSequenceClassification,fo=a.SqueezeBertModel,go=a.SqueezeBertPreTrainedModel,_o=a.SqueezeBertTokenizer,wo=a.StableLmForCausalLM,yo=a.StableLmModel,bo=a.StableLmPreTrainedModel,vo=a.Starcoder2ForCausalLM,xo=a.Starcoder2Model,Mo=a.Starcoder2PreTrainedModel,To=a.StoppingCriteria,ko=a.StoppingCriteriaList,$o=a.SummarizationPipeline,So=a.Swin2SRForImageSuperResolution,Co=a.Swin2SRImageProcessor,Eo=a.Swin2SRModel,Po=a.Swin2SRPreTrainedModel,Ao=a.SwinForImageClassification,Fo=a.SwinModel,zo=a.SwinPreTrainedModel,Io=a.T5ForConditionalGeneration,Oo=a.T5Model,Bo=a.T5PreTrainedModel,Lo=a.T5Tokenizer,Do=a.TableTransformerForObjectDetection,Ro=a.TableTransformerModel,No=a.TableTransformerObjectDetectionOutput,Vo=a.TableTransformerPreTrainedModel,jo=a.Tensor,Go=a.Text2TextGenerationPipeline,qo=a.TextClassificationPipeline,Uo=a.TextGenerationPipeline,Wo=a.TextStreamer,Ho=a.TextToAudioPipeline,Xo=a.TokenClassificationPipeline,Ko=a.TokenClassifierOutput,Qo=a.TokenizerModel,Yo=a.TrOCRForCausalLM,Jo=a.TrOCRPreTrainedModel,Zo=a.TranslationPipeline,el=a.UniSpeechForCTC,tl=a.UniSpeechForSequenceClassification,nl=a.UniSpeechModel,rl=a.UniSpeechPreTrainedModel,il=a.UniSpeechSatForAudioFrameClassification,al=a.UniSpeechSatForCTC,sl=a.UniSpeechSatForSequenceClassification,ol=a.UniSpeechSatModel,ll=a.UniSpeechSatPreTrainedModel,ul=a.ViTFeatureExtractor,dl=a.ViTForImageClassification,cl=a.ViTImageProcessor,pl=a.ViTModel,hl=a.ViTPreTrainedModel,ml=a.VisionEncoderDecoderModel,fl=a.VitMatteForImageMatting,gl=a.VitMatteImageProcessor,_l=a.VitMattePreTrainedModel,wl=a.VitsModel,yl=a.VitsModelOutput,bl=a.VitsPreTrainedModel,vl=a.VitsTokenizer,xl=a.Wav2Vec2BertForCTC,Ml=a.Wav2Vec2BertForSequenceClassification,Tl=a.Wav2Vec2BertModel,kl=a.Wav2Vec2BertPreTrainedModel,$l=a.Wav2Vec2CTCTokenizer,Sl=a.Wav2Vec2FeatureExtractor,Cl=a.Wav2Vec2ForAudioFrameClassification,El=a.Wav2Vec2ForCTC,Pl=a.Wav2Vec2ForSequenceClassification,Al=a.Wav2Vec2Model,Fl=a.Wav2Vec2PreTrainedModel,zl=a.Wav2Vec2ProcessorWithLM,Il=a.WavLMForAudioFrameClassification,Ol=a.WavLMForCTC,Bl=a.WavLMForSequenceClassification,Ll=a.WavLMForXVector,Dl=a.WavLMModel,Rl=a.WavLMPreTrainedModel,Nl=a.WeSpeakerFeatureExtractor,Vl=a.WeSpeakerResNetModel,jl=a.WeSpeakerResNetPreTrainedModel,Gl=a.WhisperFeatureExtractor,ql=a.WhisperForConditionalGeneration,Ul=a.WhisperModel,Wl=a.WhisperPreTrainedModel,Hl=a.WhisperProcessor,Xl=a.WhisperTextStreamer,Kl=a.WhisperTokenizer,Ql=a.XLMForQuestionAnswering,Yl=a.XLMForSequenceClassification,Jl=a.XLMForTokenClassification,Zl=a.XLMModel,eu=a.XLMPreTrainedModel,tu=a.XLMRobertaForMaskedLM,nu=a.XLMRobertaForQuestionAnswering,ru=a.XLMRobertaForSequenceClassification,iu=a.XLMRobertaForTokenClassification,au=a.XLMRobertaModel,su=a.XLMRobertaPreTrainedModel,ou=a.XLMRobertaTokenizer,lu=a.XLMTokenizer,uu=a.XLMWithLMHeadModel,du=a.XVectorOutput,cu=a.YolosFeatureExtractor,pu=a.YolosForObjectDetection,hu=a.YolosModel,mu=a.YolosObjectDetectionOutput,fu=a.YolosPreTrainedModel,gu=a.ZeroShotAudioClassificationPipeline,_u=a.ZeroShotClassificationPipeline,wu=a.ZeroShotImageClassificationPipeline,yu=a.ZeroShotObjectDetectionPipeline,bu=a.bankers_round,vu=a.cat,xu=a.cos_sim,Mu=a.dot,Tu=a.dynamic_time_warping,ku=a.env,$u=a.full,Su=a.full_like,Cu=a.getKeyValueShapes,Eu=a.hamming,Pu=a.hanning,Au=a.interpolate,Fu=a.interpolate_4d,zu=a.interpolate_data,Iu=a.is_chinese_char,Ou=a.layer_norm,Bu=a.log_softmax,Lu=a.magnitude,Du=a.matmul,Ru=a.max,Nu=a.mean,Vu=a.mean_pooling,ju=a.medianFilter,Gu=a.mel_filter_bank,qu=a.min,Uu=a.ones,Wu=a.ones_like,Hu=a.permute,Xu=a.permute_data,Ku=a.pipeline,Qu=a.quantize_embeddings,Yu=a.read_audio,Ju=a.rfft,Zu=a.round,ed=a.softmax,td=a.spectrogram,nd=a.stack,rd=a.std_mean,id=a.topk,ad=a.window_function,sd=a.zeros,od=a.zeros_like;export{w as ASTFeatureExtractor,y as ASTForAudioClassification,b as ASTModel,v as ASTPreTrainedModel,x as AlbertForMaskedLM,M as AlbertForQuestionAnswering,T as AlbertForSequenceClassification,k as AlbertModel,$ as AlbertPreTrainedModel,S as AlbertTokenizer,C as AudioClassificationPipeline,E as AutoConfig,P as AutoModel,A as AutoModelForAudioClassification,F as AutoModelForAudioFrameClassification,z as AutoModelForCTC,I as AutoModelForCausalLM,O as AutoModelForDepthEstimation,B as AutoModelForDocumentQuestionAnswering,L as AutoModelForImageClassification,D as AutoModelForImageFeatureExtraction,R as AutoModelForImageMatting,N as AutoModelForImageSegmentation,V as AutoModelForImageToImage,j as AutoModelForMaskGeneration,G as AutoModelForMaskedLM,q as AutoModelForObjectDetection,U as AutoModelForQuestionAnswering,W as AutoModelForSemanticSegmentation,H as AutoModelForSeq2SeqLM,X as AutoModelForSequenceClassification,K as AutoModelForSpeechSeq2Seq,Q as AutoModelForTextToSpectrogram,Y as AutoModelForTextToWaveform,J as AutoModelForTokenClassification,Z as AutoModelForVision2Seq,ee as AutoModelForXVector,te as AutoModelForZeroShotObjectDetection,ne as AutoProcessor,re as AutoTokenizer,ie as AutomaticSpeechRecognitionPipeline,ae as BartForConditionalGeneration,se as BartForSequenceClassification,oe as BartModel,le as BartPretrainedModel,ue as BartTokenizer,de as BaseModelOutput,ce as BaseStreamer,pe as BeitFeatureExtractor,he as BeitForImageClassification,me as BeitModel,fe as BeitPreTrainedModel,ge as BertForMaskedLM,_e as BertForQuestionAnswering,we as BertForSequenceClassification,ye as BertForTokenClassification,be as BertModel,ve as BertPreTrainedModel,xe as BertTokenizer,Me as BitImageProcessor,Te as BlenderbotForConditionalGeneration,ke as BlenderbotModel,$e as BlenderbotPreTrainedModel,Se as BlenderbotSmallForConditionalGeneration,Ce as BlenderbotSmallModel,Ee as BlenderbotSmallPreTrainedModel,Pe as BlenderbotSmallTokenizer,Ae as BlenderbotTokenizer,Fe as BloomForCausalLM,ze as BloomModel,Ie as BloomPreTrainedModel,Oe as BloomTokenizer,Be as CLIPFeatureExtractor,Le as CLIPImageProcessor,De as CLIPModel,Re as CLIPPreTrainedModel,Ne as CLIPSegForImageSegmentation,Ve as CLIPSegModel,je as CLIPSegPreTrainedModel,Ge as CLIPTextModelWithProjection,qe as CLIPTokenizer,Ue as CLIPVisionModelWithProjection,We as CamembertForMaskedLM,He as CamembertForQuestionAnswering,Xe as CamembertForSequenceClassification,Ke as CamembertForTokenClassification,Qe as CamembertModel,Ye as CamembertPreTrainedModel,Je as CamembertTokenizer,Ze as CausalLMOutput,et as CausalLMOutputWithPast,tt as ChineseCLIPFeatureExtractor,nt as ChineseCLIPModel,rt as ChineseCLIPPreTrainedModel,it as ClapAudioModelWithProjection,at as ClapFeatureExtractor,st as ClapModel,ot as ClapPreTrainedModel,lt as ClapTextModelWithProjection,ut as CodeGenForCausalLM,dt as CodeGenModel,ct as CodeGenPreTrainedModel,pt as CodeGenTokenizer,ht as CodeLlamaTokenizer,mt as CohereForCausalLM,ft as CohereModel,gt as CoherePreTrainedModel,_t as CohereTokenizer,wt as ConvBertForMaskedLM,yt as ConvBertForQuestionAnswering,bt as ConvBertForSequenceClassification,vt as ConvBertForTokenClassification,xt as ConvBertModel,Mt as ConvBertPreTrainedModel,Tt as ConvBertTokenizer,kt as ConvNextFeatureExtractor,$t as ConvNextForImageClassification,St as ConvNextImageProcessor,Ct as ConvNextModel,Et as ConvNextPreTrainedModel,Pt as ConvNextV2ForImageClassification,At as ConvNextV2Model,Ft as ConvNextV2PreTrainedModel,zt as DPTFeatureExtractor,It as DPTForDepthEstimation,Ot as DPTImageProcessor,Bt as DPTModel,Lt as DPTPreTrainedModel,Dt as DebertaForMaskedLM,Rt as DebertaForQuestionAnswering,Nt as DebertaForSequenceClassification,Vt as DebertaForTokenClassification,jt as DebertaModel,Gt as DebertaPreTrainedModel,qt as DebertaTokenizer,Ut as DebertaV2ForMaskedLM,Wt as DebertaV2ForQuestionAnswering,Ht as DebertaV2ForSequenceClassification,Xt as DebertaV2ForTokenClassification,Kt as DebertaV2Model,Qt as DebertaV2PreTrainedModel,Yt as DebertaV2Tokenizer,Jt as DeiTFeatureExtractor,Zt as DeiTForImageClassification,en as DeiTModel,tn as DeiTPreTrainedModel,nn as DepthAnythingForDepthEstimation,rn as DepthAnythingPreTrainedModel,an as DepthEstimationPipeline,sn as DetrFeatureExtractor,on as DetrForObjectDetection,ln as DetrForSegmentation,un as DetrModel,dn as DetrObjectDetectionOutput,cn as DetrPreTrainedModel,pn as DetrSegmentationOutput,hn as Dinov2ForImageClassification,mn as Dinov2Model,fn as Dinov2PreTrainedModel,gn as DistilBertForMaskedLM,_n as DistilBertForQuestionAnswering,wn as DistilBertForSequenceClassification,yn as DistilBertForTokenClassification,bn as DistilBertModel,vn as DistilBertPreTrainedModel,xn as DistilBertTokenizer,Mn as DocumentQuestionAnsweringPipeline,Tn as DonutFeatureExtractor,kn as DonutSwinModel,$n as DonutSwinPreTrainedModel,Sn as EfficientNetForImageClassification,Cn as EfficientNetImageProcessor,En as EfficientNetModel,Pn as EfficientNetPreTrainedModel,An as ElectraForMaskedLM,Fn as ElectraForQuestionAnswering,zn as ElectraForSequenceClassification,In as ElectraForTokenClassification,On as ElectraModel,Bn as ElectraPreTrainedModel,Ln as ElectraTokenizer,Dn as EosTokenCriteria,Rn as EsmForMaskedLM,Nn as EsmForSequenceClassification,Vn as EsmForTokenClassification,jn as EsmModel,Gn as EsmPreTrainedModel,qn as EsmTokenizer,Un as FFT,Wn as FalconForCausalLM,Hn as FalconModel,Xn as FalconPreTrainedModel,Kn as FalconTokenizer,Qn as FastViTForImageClassification,Yn as FastViTModel,Jn as FastViTPreTrainedModel,Zn as FeatureExtractionPipeline,er as FeatureExtractor,tr as FillMaskPipeline,nr as Florence2ForConditionalGeneration,rr as Florence2PreTrainedModel,ir as Florence2Processor,ar as GLPNFeatureExtractor,sr as GLPNForDepthEstimation,or as GLPNModel,lr as GLPNPreTrainedModel,ur as GPT2LMHeadModel,dr as GPT2Model,cr as GPT2PreTrainedModel,pr as GPT2Tokenizer,hr as GPTBigCodeForCausalLM,mr as GPTBigCodeModel,fr as GPTBigCodePreTrainedModel,gr as GPTJForCausalLM,_r as GPTJModel,wr as GPTJPreTrainedModel,yr as GPTNeoForCausalLM,br as GPTNeoModel,vr as GPTNeoPreTrainedModel,xr as GPTNeoXForCausalLM,Mr as GPTNeoXModel,Tr as GPTNeoXPreTrainedModel,kr as GPTNeoXTokenizer,$r as Gemma2ForCausalLM,Sr as Gemma2Model,Cr as Gemma2PreTrainedModel,Er as GemmaForCausalLM,Pr as GemmaModel,Ar as GemmaPreTrainedModel,Fr as GemmaTokenizer,zr as Grok1Tokenizer,Ir as HerbertTokenizer,Or as HubertForCTC,Br as HubertForSequenceClassification,Lr as HubertModel,Dr as HubertPreTrainedModel,Rr as ImageClassificationPipeline,Nr as ImageFeatureExtractionPipeline,Vr as ImageFeatureExtractor,jr as ImageMattingOutput,Gr as ImageSegmentationPipeline,qr as ImageToImagePipeline,Ur as ImageToTextPipeline,Wr as InterruptableStoppingCriteria,Hr as JAISLMHeadModel,Xr as JAISModel,Kr as JAISPreTrainedModel,Qr as LlamaForCausalLM,Yr as LlamaModel,Jr as LlamaPreTrainedModel,Zr as LlamaTokenizer,ei as LlavaForConditionalGeneration,ti as LlavaPreTrainedModel,ni as LongT5ForConditionalGeneration,ri as LongT5Model,ii as LongT5PreTrainedModel,ai as M2M100ForConditionalGeneration,si as M2M100Model,oi as M2M100PreTrainedModel,li as M2M100Tokenizer,ui as MBart50Tokenizer,di as MBartForCausalLM,ci as MBartForConditionalGeneration,pi as MBartForSequenceClassification,hi as MBartModel,mi as MBartPreTrainedModel,fi as MBartTokenizer,gi as MPNetForMaskedLM,_i as MPNetForQuestionAnswering,wi as MPNetForSequenceClassification,yi as MPNetForTokenClassification,bi as MPNetModel,vi as MPNetPreTrainedModel,xi as MPNetTokenizer,Mi as MT5ForConditionalGeneration,Ti as MT5Model,ki as MT5PreTrainedModel,$i as MarianMTModel,Si as MarianModel,Ci as MarianPreTrainedModel,Ei as MarianTokenizer,Pi as MaskedLMOutput,Ai as MaxLengthCriteria,Fi as MistralForCausalLM,zi as MistralModel,Ii as MistralPreTrainedModel,Oi as MobileBertForMaskedLM,Bi as MobileBertForQuestionAnswering,Li as MobileBertForSequenceClassification,Di as MobileBertModel,Ri as MobileBertPreTrainedModel,Ni as MobileBertTokenizer,Vi as MobileNetV1FeatureExtractor,ji as MobileNetV1ForImageClassification,Gi as MobileNetV1Model,qi as MobileNetV1PreTrainedModel,Ui as MobileNetV2FeatureExtractor,Wi as MobileNetV2ForImageClassification,Hi as MobileNetV2Model,Xi as MobileNetV2PreTrainedModel,Ki as MobileNetV3FeatureExtractor,Qi as MobileNetV3ForImageClassification,Yi as MobileNetV3Model,Ji as MobileNetV3PreTrainedModel,Zi as MobileNetV4FeatureExtractor,ea as MobileNetV4ForImageClassification,ta as MobileNetV4Model,na as MobileNetV4PreTrainedModel,ra as MobileViTFeatureExtractor,ia as MobileViTForImageClassification,aa as MobileViTImageProcessor,sa as MobileViTModel,oa as MobileViTPreTrainedModel,la as MobileViTV2ForImageClassification,ua as MobileViTV2Model,da as MobileViTV2PreTrainedModel,ca as ModelOutput,pa as Moondream1ForConditionalGeneration,ha as MptForCausalLM,ma as MptModel,fa as MptPreTrainedModel,ga as MusicgenForCausalLM,_a as MusicgenForConditionalGeneration,wa as MusicgenModel,ya as MusicgenPreTrainedModel,ba as NllbTokenizer,va as NomicBertModel,xa as NomicBertPreTrainedModel,Ma as NougatImageProcessor,Ta as NougatTokenizer,ka as OPTForCausalLM,$a as OPTModel,Sa as OPTPreTrainedModel,Ca as ObjectDetectionPipeline,Ea as OpenELMForCausalLM,Pa as OpenELMModel,Aa as OpenELMPreTrainedModel,Fa as OwlViTFeatureExtractor,za as OwlViTForObjectDetection,Ia as OwlViTModel,Oa as OwlViTPreTrainedModel,Ba as OwlViTProcessor,La as Owlv2ForObjectDetection,Da as Owlv2ImageProcessor,Ra as Owlv2Model,Na as Owlv2PreTrainedModel,Va as Phi3ForCausalLM,ja as Phi3Model,Ga as Phi3PreTrainedModel,qa as PhiForCausalLM,Ua as PhiModel,Wa as PhiPreTrainedModel,Ha as Pipeline,Xa as PreTrainedModel,Ka as PreTrainedTokenizer,Qa as PretrainedConfig,Ya as PretrainedMixin,Ja as Processor,Za as PyAnnoteFeatureExtractor,es as PyAnnoteForAudioFrameClassification,ts as PyAnnoteModel,ns as PyAnnotePreTrainedModel,rs as PyAnnoteProcessor,is as QuestionAnsweringModelOutput,as as QuestionAnsweringPipeline,ss as Qwen2ForCausalLM,os as Qwen2Model,ls as Qwen2PreTrainedModel,us as Qwen2Tokenizer,ds as RTDetrForObjectDetection,cs as RTDetrImageProcessor,ps as RTDetrModel,hs as RTDetrObjectDetectionOutput,ms as RTDetrPreTrainedModel,fs as RawImage,gs as ResNetForImageClassification,_s as ResNetModel,ws as ResNetPreTrainedModel,ys as RoFormerForMaskedLM,bs as RoFormerForQuestionAnswering,vs as RoFormerForSequenceClassification,xs as RoFormerForTokenClassification,Ms as RoFormerModel,Ts as RoFormerPreTrainedModel,ks as RoFormerTokenizer,$s as RobertaForMaskedLM,Ss as RobertaForQuestionAnswering,Cs as RobertaForSequenceClassification,Es as RobertaForTokenClassification,Ps as RobertaModel,As as RobertaPreTrainedModel,Fs as RobertaTokenizer,zs as SamImageProcessor,Is as SamImageSegmentationOutput,Os as SamModel,Bs as SamPreTrainedModel,Ls as SamProcessor,Ds as SapiensFeatureExtractor,Rs as SapiensForDepthEstimation,Ns as SapiensForNormalEstimation,Vs as SapiensForSemanticSegmentation,js as SapiensPreTrainedModel,Gs as SeamlessM4TFeatureExtractor,qs as SegformerFeatureExtractor,Us as SegformerForImageClassification,Ws as SegformerForSemanticSegmentation,Hs as SegformerModel,Xs as SegformerPreTrainedModel,Ks as Seq2SeqLMOutput,Qs as SequenceClassifierOutput,Ys as SiglipImageProcessor,Js as SiglipModel,Zs as SiglipPreTrainedModel,eo as SiglipTextModel,to as SiglipTokenizer,no as SiglipVisionModel,ro as SpeechT5FeatureExtractor,io as SpeechT5ForSpeechToText,ao as SpeechT5ForTextToSpeech,so as SpeechT5HifiGan,oo as SpeechT5Model,lo as SpeechT5PreTrainedModel,uo as SpeechT5Processor,co as SpeechT5Tokenizer,po as SqueezeBertForMaskedLM,ho as SqueezeBertForQuestionAnswering,mo as SqueezeBertForSequenceClassification,fo as SqueezeBertModel,go as SqueezeBertPreTrainedModel,_o as SqueezeBertTokenizer,wo as StableLmForCausalLM,yo as StableLmModel,bo as StableLmPreTrainedModel,vo as Starcoder2ForCausalLM,xo as Starcoder2Model,Mo as Starcoder2PreTrainedModel,To as StoppingCriteria,ko as StoppingCriteriaList,$o as SummarizationPipeline,So as Swin2SRForImageSuperResolution,Co as Swin2SRImageProcessor,Eo as Swin2SRModel,Po as Swin2SRPreTrainedModel,Ao as SwinForImageClassification,Fo as SwinModel,zo as SwinPreTrainedModel,Io as T5ForConditionalGeneration,Oo as T5Model,Bo as T5PreTrainedModel,Lo as T5Tokenizer,Do as TableTransformerForObjectDetection,Ro as TableTransformerModel,No as TableTransformerObjectDetectionOutput,Vo as TableTransformerPreTrainedModel,jo as Tensor,Go as Text2TextGenerationPipeline,qo as TextClassificationPipeline,Uo as TextGenerationPipeline,Wo as TextStreamer,Ho as TextToAudioPipeline,Xo as TokenClassificationPipeline,Ko as TokenClassifierOutput,Qo as TokenizerModel,Yo as TrOCRForCausalLM,Jo as TrOCRPreTrainedModel,Zo as TranslationPipeline,el as UniSpeechForCTC,tl as UniSpeechForSequenceClassification,nl as UniSpeechModel,rl as UniSpeechPreTrainedModel,il as UniSpeechSatForAudioFrameClassification,al as UniSpeechSatForCTC,sl as UniSpeechSatForSequenceClassification,ol as UniSpeechSatModel,ll as UniSpeechSatPreTrainedModel,ul as ViTFeatureExtractor,dl as ViTForImageClassification,cl as ViTImageProcessor,pl as ViTModel,hl as ViTPreTrainedModel,ml as VisionEncoderDecoderModel,fl as VitMatteForImageMatting,gl as VitMatteImageProcessor,_l as VitMattePreTrainedModel,wl as VitsModel,yl as VitsModelOutput,bl as VitsPreTrainedModel,vl as VitsTokenizer,xl as Wav2Vec2BertForCTC,Ml as Wav2Vec2BertForSequenceClassification,Tl as Wav2Vec2BertModel,kl as Wav2Vec2BertPreTrainedModel,$l as Wav2Vec2CTCTokenizer,Sl as Wav2Vec2FeatureExtractor,Cl as Wav2Vec2ForAudioFrameClassification,El as Wav2Vec2ForCTC,Pl as Wav2Vec2ForSequenceClassification,Al as Wav2Vec2Model,Fl as Wav2Vec2PreTrainedModel,zl as Wav2Vec2ProcessorWithLM,Il as WavLMForAudioFrameClassification,Ol as WavLMForCTC,Bl as WavLMForSequenceClassification,Ll as WavLMForXVector,Dl as WavLMModel,Rl as WavLMPreTrainedModel,Nl as WeSpeakerFeatureExtractor,Vl as WeSpeakerResNetModel,jl as WeSpeakerResNetPreTrainedModel,Gl as WhisperFeatureExtractor,ql as WhisperForConditionalGeneration,Ul as WhisperModel,Wl as WhisperPreTrainedModel,Hl as WhisperProcessor,Xl as WhisperTextStreamer,Kl as WhisperTokenizer,Ql as XLMForQuestionAnswering,Yl as XLMForSequenceClassification,Jl as XLMForTokenClassification,Zl as XLMModel,eu as XLMPreTrainedModel,tu as XLMRobertaForMaskedLM,nu as XLMRobertaForQuestionAnswering,ru as XLMRobertaForSequenceClassification,iu as XLMRobertaForTokenClassification,au as XLMRobertaModel,su as XLMRobertaPreTrainedModel,ou as XLMRobertaTokenizer,lu as XLMTokenizer,uu as XLMWithLMHeadModel,du as XVectorOutput,cu as YolosFeatureExtractor,pu as YolosForObjectDetection,hu as YolosModel,mu as YolosObjectDetectionOutput,fu as YolosPreTrainedModel,gu as ZeroShotAudioClassificationPipeline,_u as ZeroShotClassificationPipeline,wu as ZeroShotImageClassificationPipeline,yu as ZeroShotObjectDetectionPipeline,bu as bankers_round,vu as cat,xu as cos_sim,Mu as dot,Tu as dynamic_time_warping,ku as env,$u as full,Su as full_like,Cu as getKeyValueShapes,Eu as hamming,Pu as hanning,Au as interpolate,Fu as interpolate_4d,zu as interpolate_data,Iu as is_chinese_char,Ou as layer_norm,Bu as log_softmax,Lu as magnitude,Du as matmul,Ru as max,Nu as mean,Vu as mean_pooling,ju as medianFilter,Gu as mel_filter_bank,qu as min,Uu as ones,Wu as ones_like,Hu as permute,Xu as permute_data,Ku as pipeline,Qu as quantize_embeddings,Yu as read_audio,Ju as rfft,Zu as round,ed as softmax,td as spectrogram,nd as stack,rd as std_mean,id as topk,ad as window_function,sd as zeros,od as zeros_like};
+  \*****************************/i.r(a),i.d(a,{ASTFeatureExtractor:()=>d.ASTFeatureExtractor,ASTForAudioClassification:()=>l.ASTForAudioClassification,ASTModel:()=>l.ASTModel,ASTPreTrainedModel:()=>l.ASTPreTrainedModel,AlbertForMaskedLM:()=>l.AlbertForMaskedLM,AlbertForQuestionAnswering:()=>l.AlbertForQuestionAnswering,AlbertForSequenceClassification:()=>l.AlbertForSequenceClassification,AlbertModel:()=>l.AlbertModel,AlbertPreTrainedModel:()=>l.AlbertPreTrainedModel,AlbertTokenizer:()=>u.AlbertTokenizer,AudioClassificationPipeline:()=>o.AudioClassificationPipeline,AutoConfig:()=>c.AutoConfig,AutoModel:()=>l.AutoModel,AutoModelForAudioClassification:()=>l.AutoModelForAudioClassification,AutoModelForAudioFrameClassification:()=>l.AutoModelForAudioFrameClassification,AutoModelForCTC:()=>l.AutoModelForCTC,AutoModelForCausalLM:()=>l.AutoModelForCausalLM,AutoModelForDepthEstimation:()=>l.AutoModelForDepthEstimation,AutoModelForDocumentQuestionAnswering:()=>l.AutoModelForDocumentQuestionAnswering,AutoModelForImageClassification:()=>l.AutoModelForImageClassification,AutoModelForImageFeatureExtraction:()=>l.AutoModelForImageFeatureExtraction,AutoModelForImageMatting:()=>l.AutoModelForImageMatting,AutoModelForImageSegmentation:()=>l.AutoModelForImageSegmentation,AutoModelForImageToImage:()=>l.AutoModelForImageToImage,AutoModelForMaskGeneration:()=>l.AutoModelForMaskGeneration,AutoModelForMaskedLM:()=>l.AutoModelForMaskedLM,AutoModelForNormalEstimation:()=>l.AutoModelForNormalEstimation,AutoModelForObjectDetection:()=>l.AutoModelForObjectDetection,AutoModelForQuestionAnswering:()=>l.AutoModelForQuestionAnswering,AutoModelForSemanticSegmentation:()=>l.AutoModelForSemanticSegmentation,AutoModelForSeq2SeqLM:()=>l.AutoModelForSeq2SeqLM,AutoModelForSequenceClassification:()=>l.AutoModelForSequenceClassification,AutoModelForSpeechSeq2Seq:()=>l.AutoModelForSpeechSeq2Seq,AutoModelForTextToSpectrogram:()=>l.AutoModelForTextToSpectrogram,AutoModelForTextToWaveform:()=>l.AutoModelForTextToWaveform,AutoModelForTokenClassification:()=>l.AutoModelForTokenClassification,AutoModelForVision2Seq:()=>l.AutoModelForVision2Seq,AutoModelForXVector:()=>l.AutoModelForXVector,AutoModelForZeroShotObjectDetection:()=>l.AutoModelForZeroShotObjectDetection,AutoProcessor:()=>d.AutoProcessor,AutoTokenizer:()=>u.AutoTokenizer,AutomaticSpeechRecognitionPipeline:()=>o.AutomaticSpeechRecognitionPipeline,BartForConditionalGeneration:()=>l.BartForConditionalGeneration,BartForSequenceClassification:()=>l.BartForSequenceClassification,BartModel:()=>l.BartModel,BartPretrainedModel:()=>l.BartPretrainedModel,BartTokenizer:()=>u.BartTokenizer,BaseModelOutput:()=>l.BaseModelOutput,BaseStreamer:()=>g.BaseStreamer,BeitFeatureExtractor:()=>d.BeitFeatureExtractor,BeitForImageClassification:()=>l.BeitForImageClassification,BeitModel:()=>l.BeitModel,BeitPreTrainedModel:()=>l.BeitPreTrainedModel,BertForMaskedLM:()=>l.BertForMaskedLM,BertForQuestionAnswering:()=>l.BertForQuestionAnswering,BertForSequenceClassification:()=>l.BertForSequenceClassification,BertForTokenClassification:()=>l.BertForTokenClassification,BertModel:()=>l.BertModel,BertPreTrainedModel:()=>l.BertPreTrainedModel,BertTokenizer:()=>u.BertTokenizer,BitImageProcessor:()=>d.BitImageProcessor,BlenderbotForConditionalGeneration:()=>l.BlenderbotForConditionalGeneration,BlenderbotModel:()=>l.BlenderbotModel,BlenderbotPreTrainedModel:()=>l.BlenderbotPreTrainedModel,BlenderbotSmallForConditionalGeneration:()=>l.BlenderbotSmallForConditionalGeneration,BlenderbotSmallModel:()=>l.BlenderbotSmallModel,BlenderbotSmallPreTrainedModel:()=>l.BlenderbotSmallPreTrainedModel,BlenderbotSmallTokenizer:()=>u.BlenderbotSmallTokenizer,BlenderbotTokenizer:()=>u.BlenderbotTokenizer,BloomForCausalLM:()=>l.BloomForCausalLM,BloomModel:()=>l.BloomModel,BloomPreTrainedModel:()=>l.BloomPreTrainedModel,BloomTokenizer:()=>u.BloomTokenizer,CLIPFeatureExtractor:()=>d.CLIPFeatureExtractor,CLIPImageProcessor:()=>d.CLIPImageProcessor,CLIPModel:()=>l.CLIPModel,CLIPPreTrainedModel:()=>l.CLIPPreTrainedModel,CLIPSegForImageSegmentation:()=>l.CLIPSegForImageSegmentation,CLIPSegModel:()=>l.CLIPSegModel,CLIPSegPreTrainedModel:()=>l.CLIPSegPreTrainedModel,CLIPTextModelWithProjection:()=>l.CLIPTextModelWithProjection,CLIPTokenizer:()=>u.CLIPTokenizer,CLIPVisionModelWithProjection:()=>l.CLIPVisionModelWithProjection,CamembertForMaskedLM:()=>l.CamembertForMaskedLM,CamembertForQuestionAnswering:()=>l.CamembertForQuestionAnswering,CamembertForSequenceClassification:()=>l.CamembertForSequenceClassification,CamembertForTokenClassification:()=>l.CamembertForTokenClassification,CamembertModel:()=>l.CamembertModel,CamembertPreTrainedModel:()=>l.CamembertPreTrainedModel,CamembertTokenizer:()=>u.CamembertTokenizer,CausalLMOutput:()=>l.CausalLMOutput,CausalLMOutputWithPast:()=>l.CausalLMOutputWithPast,ChineseCLIPFeatureExtractor:()=>d.ChineseCLIPFeatureExtractor,ChineseCLIPModel:()=>l.ChineseCLIPModel,ChineseCLIPPreTrainedModel:()=>l.ChineseCLIPPreTrainedModel,ClapAudioModelWithProjection:()=>l.ClapAudioModelWithProjection,ClapFeatureExtractor:()=>d.ClapFeatureExtractor,ClapModel:()=>l.ClapModel,ClapPreTrainedModel:()=>l.ClapPreTrainedModel,ClapTextModelWithProjection:()=>l.ClapTextModelWithProjection,CodeGenForCausalLM:()=>l.CodeGenForCausalLM,CodeGenModel:()=>l.CodeGenModel,CodeGenPreTrainedModel:()=>l.CodeGenPreTrainedModel,CodeGenTokenizer:()=>u.CodeGenTokenizer,CodeLlamaTokenizer:()=>u.CodeLlamaTokenizer,CohereForCausalLM:()=>l.CohereForCausalLM,CohereModel:()=>l.CohereModel,CoherePreTrainedModel:()=>l.CoherePreTrainedModel,CohereTokenizer:()=>u.CohereTokenizer,ConvBertForMaskedLM:()=>l.ConvBertForMaskedLM,ConvBertForQuestionAnswering:()=>l.ConvBertForQuestionAnswering,ConvBertForSequenceClassification:()=>l.ConvBertForSequenceClassification,ConvBertForTokenClassification:()=>l.ConvBertForTokenClassification,ConvBertModel:()=>l.ConvBertModel,ConvBertPreTrainedModel:()=>l.ConvBertPreTrainedModel,ConvBertTokenizer:()=>u.ConvBertTokenizer,ConvNextFeatureExtractor:()=>d.ConvNextFeatureExtractor,ConvNextForImageClassification:()=>l.ConvNextForImageClassification,ConvNextImageProcessor:()=>d.ConvNextImageProcessor,ConvNextModel:()=>l.ConvNextModel,ConvNextPreTrainedModel:()=>l.ConvNextPreTrainedModel,ConvNextV2ForImageClassification:()=>l.ConvNextV2ForImageClassification,ConvNextV2Model:()=>l.ConvNextV2Model,ConvNextV2PreTrainedModel:()=>l.ConvNextV2PreTrainedModel,DPTFeatureExtractor:()=>d.DPTFeatureExtractor,DPTForDepthEstimation:()=>l.DPTForDepthEstimation,DPTImageProcessor:()=>d.DPTImageProcessor,DPTModel:()=>l.DPTModel,DPTPreTrainedModel:()=>l.DPTPreTrainedModel,DebertaForMaskedLM:()=>l.DebertaForMaskedLM,DebertaForQuestionAnswering:()=>l.DebertaForQuestionAnswering,DebertaForSequenceClassification:()=>l.DebertaForSequenceClassification,DebertaForTokenClassification:()=>l.DebertaForTokenClassification,DebertaModel:()=>l.DebertaModel,DebertaPreTrainedModel:()=>l.DebertaPreTrainedModel,DebertaTokenizer:()=>u.DebertaTokenizer,DebertaV2ForMaskedLM:()=>l.DebertaV2ForMaskedLM,DebertaV2ForQuestionAnswering:()=>l.DebertaV2ForQuestionAnswering,DebertaV2ForSequenceClassification:()=>l.DebertaV2ForSequenceClassification,DebertaV2ForTokenClassification:()=>l.DebertaV2ForTokenClassification,DebertaV2Model:()=>l.DebertaV2Model,DebertaV2PreTrainedModel:()=>l.DebertaV2PreTrainedModel,DebertaV2Tokenizer:()=>u.DebertaV2Tokenizer,DeiTFeatureExtractor:()=>d.DeiTFeatureExtractor,DeiTForImageClassification:()=>l.DeiTForImageClassification,DeiTModel:()=>l.DeiTModel,DeiTPreTrainedModel:()=>l.DeiTPreTrainedModel,DepthAnythingForDepthEstimation:()=>l.DepthAnythingForDepthEstimation,DepthAnythingPreTrainedModel:()=>l.DepthAnythingPreTrainedModel,DepthEstimationPipeline:()=>o.DepthEstimationPipeline,DetrFeatureExtractor:()=>d.DetrFeatureExtractor,DetrForObjectDetection:()=>l.DetrForObjectDetection,DetrForSegmentation:()=>l.DetrForSegmentation,DetrModel:()=>l.DetrModel,DetrObjectDetectionOutput:()=>l.DetrObjectDetectionOutput,DetrPreTrainedModel:()=>l.DetrPreTrainedModel,DetrSegmentationOutput:()=>l.DetrSegmentationOutput,Dinov2ForImageClassification:()=>l.Dinov2ForImageClassification,Dinov2Model:()=>l.Dinov2Model,Dinov2PreTrainedModel:()=>l.Dinov2PreTrainedModel,DistilBertForMaskedLM:()=>l.DistilBertForMaskedLM,DistilBertForQuestionAnswering:()=>l.DistilBertForQuestionAnswering,DistilBertForSequenceClassification:()=>l.DistilBertForSequenceClassification,DistilBertForTokenClassification:()=>l.DistilBertForTokenClassification,DistilBertModel:()=>l.DistilBertModel,DistilBertPreTrainedModel:()=>l.DistilBertPreTrainedModel,DistilBertTokenizer:()=>u.DistilBertTokenizer,DocumentQuestionAnsweringPipeline:()=>o.DocumentQuestionAnsweringPipeline,DonutFeatureExtractor:()=>d.DonutFeatureExtractor,DonutSwinModel:()=>l.DonutSwinModel,DonutSwinPreTrainedModel:()=>l.DonutSwinPreTrainedModel,EfficientNetForImageClassification:()=>l.EfficientNetForImageClassification,EfficientNetImageProcessor:()=>d.EfficientNetImageProcessor,EfficientNetModel:()=>l.EfficientNetModel,EfficientNetPreTrainedModel:()=>l.EfficientNetPreTrainedModel,ElectraForMaskedLM:()=>l.ElectraForMaskedLM,ElectraForQuestionAnswering:()=>l.ElectraForQuestionAnswering,ElectraForSequenceClassification:()=>l.ElectraForSequenceClassification,ElectraForTokenClassification:()=>l.ElectraForTokenClassification,ElectraModel:()=>l.ElectraModel,ElectraPreTrainedModel:()=>l.ElectraPreTrainedModel,ElectraTokenizer:()=>u.ElectraTokenizer,EosTokenCriteria:()=>_.EosTokenCriteria,EsmForMaskedLM:()=>l.EsmForMaskedLM,EsmForSequenceClassification:()=>l.EsmForSequenceClassification,EsmForTokenClassification:()=>l.EsmForTokenClassification,EsmModel:()=>l.EsmModel,EsmPreTrainedModel:()=>l.EsmPreTrainedModel,EsmTokenizer:()=>u.EsmTokenizer,FFT:()=>f.FFT,FalconForCausalLM:()=>l.FalconForCausalLM,FalconModel:()=>l.FalconModel,FalconPreTrainedModel:()=>l.FalconPreTrainedModel,FalconTokenizer:()=>u.FalconTokenizer,FastViTForImageClassification:()=>l.FastViTForImageClassification,FastViTModel:()=>l.FastViTModel,FastViTPreTrainedModel:()=>l.FastViTPreTrainedModel,FeatureExtractionPipeline:()=>o.FeatureExtractionPipeline,FeatureExtractor:()=>d.FeatureExtractor,FillMaskPipeline:()=>o.FillMaskPipeline,Florence2ForConditionalGeneration:()=>l.Florence2ForConditionalGeneration,Florence2PreTrainedModel:()=>l.Florence2PreTrainedModel,Florence2Processor:()=>d.Florence2Processor,GLPNFeatureExtractor:()=>d.GLPNFeatureExtractor,GLPNForDepthEstimation:()=>l.GLPNForDepthEstimation,GLPNModel:()=>l.GLPNModel,GLPNPreTrainedModel:()=>l.GLPNPreTrainedModel,GPT2LMHeadModel:()=>l.GPT2LMHeadModel,GPT2Model:()=>l.GPT2Model,GPT2PreTrainedModel:()=>l.GPT2PreTrainedModel,GPT2Tokenizer:()=>u.GPT2Tokenizer,GPTBigCodeForCausalLM:()=>l.GPTBigCodeForCausalLM,GPTBigCodeModel:()=>l.GPTBigCodeModel,GPTBigCodePreTrainedModel:()=>l.GPTBigCodePreTrainedModel,GPTJForCausalLM:()=>l.GPTJForCausalLM,GPTJModel:()=>l.GPTJModel,GPTJPreTrainedModel:()=>l.GPTJPreTrainedModel,GPTNeoForCausalLM:()=>l.GPTNeoForCausalLM,GPTNeoModel:()=>l.GPTNeoModel,GPTNeoPreTrainedModel:()=>l.GPTNeoPreTrainedModel,GPTNeoXForCausalLM:()=>l.GPTNeoXForCausalLM,GPTNeoXModel:()=>l.GPTNeoXModel,GPTNeoXPreTrainedModel:()=>l.GPTNeoXPreTrainedModel,GPTNeoXTokenizer:()=>u.GPTNeoXTokenizer,Gemma2ForCausalLM:()=>l.Gemma2ForCausalLM,Gemma2Model:()=>l.Gemma2Model,Gemma2PreTrainedModel:()=>l.Gemma2PreTrainedModel,GemmaForCausalLM:()=>l.GemmaForCausalLM,GemmaModel:()=>l.GemmaModel,GemmaPreTrainedModel:()=>l.GemmaPreTrainedModel,GemmaTokenizer:()=>u.GemmaTokenizer,Grok1Tokenizer:()=>u.Grok1Tokenizer,HerbertTokenizer:()=>u.HerbertTokenizer,HubertForCTC:()=>l.HubertForCTC,HubertForSequenceClassification:()=>l.HubertForSequenceClassification,HubertModel:()=>l.HubertModel,HubertPreTrainedModel:()=>l.HubertPreTrainedModel,ImageClassificationPipeline:()=>o.ImageClassificationPipeline,ImageFeatureExtractionPipeline:()=>o.ImageFeatureExtractionPipeline,ImageFeatureExtractor:()=>d.ImageFeatureExtractor,ImageMattingOutput:()=>l.ImageMattingOutput,ImageSegmentationPipeline:()=>o.ImageSegmentationPipeline,ImageToImagePipeline:()=>o.ImageToImagePipeline,ImageToTextPipeline:()=>o.ImageToTextPipeline,InterruptableStoppingCriteria:()=>_.InterruptableStoppingCriteria,JAISLMHeadModel:()=>l.JAISLMHeadModel,JAISModel:()=>l.JAISModel,JAISPreTrainedModel:()=>l.JAISPreTrainedModel,LlamaForCausalLM:()=>l.LlamaForCausalLM,LlamaModel:()=>l.LlamaModel,LlamaPreTrainedModel:()=>l.LlamaPreTrainedModel,LlamaTokenizer:()=>u.LlamaTokenizer,LlavaForConditionalGeneration:()=>l.LlavaForConditionalGeneration,LlavaPreTrainedModel:()=>l.LlavaPreTrainedModel,LongT5ForConditionalGeneration:()=>l.LongT5ForConditionalGeneration,LongT5Model:()=>l.LongT5Model,LongT5PreTrainedModel:()=>l.LongT5PreTrainedModel,M2M100ForConditionalGeneration:()=>l.M2M100ForConditionalGeneration,M2M100Model:()=>l.M2M100Model,M2M100PreTrainedModel:()=>l.M2M100PreTrainedModel,M2M100Tokenizer:()=>u.M2M100Tokenizer,MBart50Tokenizer:()=>u.MBart50Tokenizer,MBartForCausalLM:()=>l.MBartForCausalLM,MBartForConditionalGeneration:()=>l.MBartForConditionalGeneration,MBartForSequenceClassification:()=>l.MBartForSequenceClassification,MBartModel:()=>l.MBartModel,MBartPreTrainedModel:()=>l.MBartPreTrainedModel,MBartTokenizer:()=>u.MBartTokenizer,MPNetForMaskedLM:()=>l.MPNetForMaskedLM,MPNetForQuestionAnswering:()=>l.MPNetForQuestionAnswering,MPNetForSequenceClassification:()=>l.MPNetForSequenceClassification,MPNetForTokenClassification:()=>l.MPNetForTokenClassification,MPNetModel:()=>l.MPNetModel,MPNetPreTrainedModel:()=>l.MPNetPreTrainedModel,MPNetTokenizer:()=>u.MPNetTokenizer,MT5ForConditionalGeneration:()=>l.MT5ForConditionalGeneration,MT5Model:()=>l.MT5Model,MT5PreTrainedModel:()=>l.MT5PreTrainedModel,MarianMTModel:()=>l.MarianMTModel,MarianModel:()=>l.MarianModel,MarianPreTrainedModel:()=>l.MarianPreTrainedModel,MarianTokenizer:()=>u.MarianTokenizer,MaskedLMOutput:()=>l.MaskedLMOutput,MaxLengthCriteria:()=>_.MaxLengthCriteria,MistralForCausalLM:()=>l.MistralForCausalLM,MistralModel:()=>l.MistralModel,MistralPreTrainedModel:()=>l.MistralPreTrainedModel,MobileBertForMaskedLM:()=>l.MobileBertForMaskedLM,MobileBertForQuestionAnswering:()=>l.MobileBertForQuestionAnswering,MobileBertForSequenceClassification:()=>l.MobileBertForSequenceClassification,MobileBertModel:()=>l.MobileBertModel,MobileBertPreTrainedModel:()=>l.MobileBertPreTrainedModel,MobileBertTokenizer:()=>u.MobileBertTokenizer,MobileNetV1FeatureExtractor:()=>d.MobileNetV1FeatureExtractor,MobileNetV1ForImageClassification:()=>l.MobileNetV1ForImageClassification,MobileNetV1Model:()=>l.MobileNetV1Model,MobileNetV1PreTrainedModel:()=>l.MobileNetV1PreTrainedModel,MobileNetV2FeatureExtractor:()=>d.MobileNetV2FeatureExtractor,MobileNetV2ForImageClassification:()=>l.MobileNetV2ForImageClassification,MobileNetV2Model:()=>l.MobileNetV2Model,MobileNetV2PreTrainedModel:()=>l.MobileNetV2PreTrainedModel,MobileNetV3FeatureExtractor:()=>d.MobileNetV3FeatureExtractor,MobileNetV3ForImageClassification:()=>l.MobileNetV3ForImageClassification,MobileNetV3Model:()=>l.MobileNetV3Model,MobileNetV3PreTrainedModel:()=>l.MobileNetV3PreTrainedModel,MobileNetV4FeatureExtractor:()=>d.MobileNetV4FeatureExtractor,MobileNetV4ForImageClassification:()=>l.MobileNetV4ForImageClassification,MobileNetV4Model:()=>l.MobileNetV4Model,MobileNetV4PreTrainedModel:()=>l.MobileNetV4PreTrainedModel,MobileViTFeatureExtractor:()=>d.MobileViTFeatureExtractor,MobileViTForImageClassification:()=>l.MobileViTForImageClassification,MobileViTImageProcessor:()=>d.MobileViTImageProcessor,MobileViTModel:()=>l.MobileViTModel,MobileViTPreTrainedModel:()=>l.MobileViTPreTrainedModel,MobileViTV2ForImageClassification:()=>l.MobileViTV2ForImageClassification,MobileViTV2Model:()=>l.MobileViTV2Model,MobileViTV2PreTrainedModel:()=>l.MobileViTV2PreTrainedModel,ModelOutput:()=>l.ModelOutput,Moondream1ForConditionalGeneration:()=>l.Moondream1ForConditionalGeneration,MptForCausalLM:()=>l.MptForCausalLM,MptModel:()=>l.MptModel,MptPreTrainedModel:()=>l.MptPreTrainedModel,MusicgenForCausalLM:()=>l.MusicgenForCausalLM,MusicgenForConditionalGeneration:()=>l.MusicgenForConditionalGeneration,MusicgenModel:()=>l.MusicgenModel,MusicgenPreTrainedModel:()=>l.MusicgenPreTrainedModel,NllbTokenizer:()=>u.NllbTokenizer,NomicBertModel:()=>l.NomicBertModel,NomicBertPreTrainedModel:()=>l.NomicBertPreTrainedModel,NougatImageProcessor:()=>d.NougatImageProcessor,NougatTokenizer:()=>u.NougatTokenizer,OPTForCausalLM:()=>l.OPTForCausalLM,OPTModel:()=>l.OPTModel,OPTPreTrainedModel:()=>l.OPTPreTrainedModel,ObjectDetectionPipeline:()=>o.ObjectDetectionPipeline,OpenELMForCausalLM:()=>l.OpenELMForCausalLM,OpenELMModel:()=>l.OpenELMModel,OpenELMPreTrainedModel:()=>l.OpenELMPreTrainedModel,OwlViTFeatureExtractor:()=>d.OwlViTFeatureExtractor,OwlViTForObjectDetection:()=>l.OwlViTForObjectDetection,OwlViTModel:()=>l.OwlViTModel,OwlViTPreTrainedModel:()=>l.OwlViTPreTrainedModel,OwlViTProcessor:()=>d.OwlViTProcessor,Owlv2ForObjectDetection:()=>l.Owlv2ForObjectDetection,Owlv2ImageProcessor:()=>d.Owlv2ImageProcessor,Owlv2Model:()=>l.Owlv2Model,Owlv2PreTrainedModel:()=>l.Owlv2PreTrainedModel,Phi3ForCausalLM:()=>l.Phi3ForCausalLM,Phi3Model:()=>l.Phi3Model,Phi3PreTrainedModel:()=>l.Phi3PreTrainedModel,PhiForCausalLM:()=>l.PhiForCausalLM,PhiModel:()=>l.PhiModel,PhiPreTrainedModel:()=>l.PhiPreTrainedModel,Pipeline:()=>o.Pipeline,PreTrainedModel:()=>l.PreTrainedModel,PreTrainedTokenizer:()=>u.PreTrainedTokenizer,PretrainedConfig:()=>c.PretrainedConfig,PretrainedMixin:()=>l.PretrainedMixin,Processor:()=>d.Processor,PyAnnoteFeatureExtractor:()=>d.PyAnnoteFeatureExtractor,PyAnnoteForAudioFrameClassification:()=>l.PyAnnoteForAudioFrameClassification,PyAnnoteModel:()=>l.PyAnnoteModel,PyAnnotePreTrainedModel:()=>l.PyAnnotePreTrainedModel,PyAnnoteProcessor:()=>d.PyAnnoteProcessor,QuestionAnsweringModelOutput:()=>l.QuestionAnsweringModelOutput,QuestionAnsweringPipeline:()=>o.QuestionAnsweringPipeline,Qwen2ForCausalLM:()=>l.Qwen2ForCausalLM,Qwen2Model:()=>l.Qwen2Model,Qwen2PreTrainedModel:()=>l.Qwen2PreTrainedModel,Qwen2Tokenizer:()=>u.Qwen2Tokenizer,RTDetrForObjectDetection:()=>l.RTDetrForObjectDetection,RTDetrImageProcessor:()=>d.RTDetrImageProcessor,RTDetrModel:()=>l.RTDetrModel,RTDetrObjectDetectionOutput:()=>l.RTDetrObjectDetectionOutput,RTDetrPreTrainedModel:()=>l.RTDetrPreTrainedModel,RawImage:()=>h.RawImage,ResNetForImageClassification:()=>l.ResNetForImageClassification,ResNetModel:()=>l.ResNetModel,ResNetPreTrainedModel:()=>l.ResNetPreTrainedModel,RoFormerForMaskedLM:()=>l.RoFormerForMaskedLM,RoFormerForQuestionAnswering:()=>l.RoFormerForQuestionAnswering,RoFormerForSequenceClassification:()=>l.RoFormerForSequenceClassification,RoFormerForTokenClassification:()=>l.RoFormerForTokenClassification,RoFormerModel:()=>l.RoFormerModel,RoFormerPreTrainedModel:()=>l.RoFormerPreTrainedModel,RoFormerTokenizer:()=>u.RoFormerTokenizer,RobertaForMaskedLM:()=>l.RobertaForMaskedLM,RobertaForQuestionAnswering:()=>l.RobertaForQuestionAnswering,RobertaForSequenceClassification:()=>l.RobertaForSequenceClassification,RobertaForTokenClassification:()=>l.RobertaForTokenClassification,RobertaModel:()=>l.RobertaModel,RobertaPreTrainedModel:()=>l.RobertaPreTrainedModel,RobertaTokenizer:()=>u.RobertaTokenizer,SamImageProcessor:()=>d.SamImageProcessor,SamImageSegmentationOutput:()=>l.SamImageSegmentationOutput,SamModel:()=>l.SamModel,SamPreTrainedModel:()=>l.SamPreTrainedModel,SamProcessor:()=>d.SamProcessor,SapiensFeatureExtractor:()=>d.SapiensFeatureExtractor,SapiensForDepthEstimation:()=>l.SapiensForDepthEstimation,SapiensForNormalEstimation:()=>l.SapiensForNormalEstimation,SapiensForSemanticSegmentation:()=>l.SapiensForSemanticSegmentation,SapiensPreTrainedModel:()=>l.SapiensPreTrainedModel,SeamlessM4TFeatureExtractor:()=>d.SeamlessM4TFeatureExtractor,SegformerFeatureExtractor:()=>d.SegformerFeatureExtractor,SegformerForImageClassification:()=>l.SegformerForImageClassification,SegformerForSemanticSegmentation:()=>l.SegformerForSemanticSegmentation,SegformerModel:()=>l.SegformerModel,SegformerPreTrainedModel:()=>l.SegformerPreTrainedModel,Seq2SeqLMOutput:()=>l.Seq2SeqLMOutput,SequenceClassifierOutput:()=>l.SequenceClassifierOutput,SiglipImageProcessor:()=>d.SiglipImageProcessor,SiglipModel:()=>l.SiglipModel,SiglipPreTrainedModel:()=>l.SiglipPreTrainedModel,SiglipTextModel:()=>l.SiglipTextModel,SiglipTokenizer:()=>u.SiglipTokenizer,SiglipVisionModel:()=>l.SiglipVisionModel,SpeechT5FeatureExtractor:()=>d.SpeechT5FeatureExtractor,SpeechT5ForSpeechToText:()=>l.SpeechT5ForSpeechToText,SpeechT5ForTextToSpeech:()=>l.SpeechT5ForTextToSpeech,SpeechT5HifiGan:()=>l.SpeechT5HifiGan,SpeechT5Model:()=>l.SpeechT5Model,SpeechT5PreTrainedModel:()=>l.SpeechT5PreTrainedModel,SpeechT5Processor:()=>d.SpeechT5Processor,SpeechT5Tokenizer:()=>u.SpeechT5Tokenizer,SqueezeBertForMaskedLM:()=>l.SqueezeBertForMaskedLM,SqueezeBertForQuestionAnswering:()=>l.SqueezeBertForQuestionAnswering,SqueezeBertForSequenceClassification:()=>l.SqueezeBertForSequenceClassification,SqueezeBertModel:()=>l.SqueezeBertModel,SqueezeBertPreTrainedModel:()=>l.SqueezeBertPreTrainedModel,SqueezeBertTokenizer:()=>u.SqueezeBertTokenizer,StableLmForCausalLM:()=>l.StableLmForCausalLM,StableLmModel:()=>l.StableLmModel,StableLmPreTrainedModel:()=>l.StableLmPreTrainedModel,Starcoder2ForCausalLM:()=>l.Starcoder2ForCausalLM,Starcoder2Model:()=>l.Starcoder2Model,Starcoder2PreTrainedModel:()=>l.Starcoder2PreTrainedModel,StoppingCriteria:()=>_.StoppingCriteria,StoppingCriteriaList:()=>_.StoppingCriteriaList,SummarizationPipeline:()=>o.SummarizationPipeline,Swin2SRForImageSuperResolution:()=>l.Swin2SRForImageSuperResolution,Swin2SRImageProcessor:()=>d.Swin2SRImageProcessor,Swin2SRModel:()=>l.Swin2SRModel,Swin2SRPreTrainedModel:()=>l.Swin2SRPreTrainedModel,SwinForImageClassification:()=>l.SwinForImageClassification,SwinModel:()=>l.SwinModel,SwinPreTrainedModel:()=>l.SwinPreTrainedModel,T5ForConditionalGeneration:()=>l.T5ForConditionalGeneration,T5Model:()=>l.T5Model,T5PreTrainedModel:()=>l.T5PreTrainedModel,T5Tokenizer:()=>u.T5Tokenizer,TableTransformerForObjectDetection:()=>l.TableTransformerForObjectDetection,TableTransformerModel:()=>l.TableTransformerModel,TableTransformerObjectDetectionOutput:()=>l.TableTransformerObjectDetectionOutput,TableTransformerPreTrainedModel:()=>l.TableTransformerPreTrainedModel,Tensor:()=>m.Tensor,Text2TextGenerationPipeline:()=>o.Text2TextGenerationPipeline,TextClassificationPipeline:()=>o.TextClassificationPipeline,TextGenerationPipeline:()=>o.TextGenerationPipeline,TextStreamer:()=>g.TextStreamer,TextToAudioPipeline:()=>o.TextToAudioPipeline,TokenClassificationPipeline:()=>o.TokenClassificationPipeline,TokenClassifierOutput:()=>l.TokenClassifierOutput,TokenizerModel:()=>u.TokenizerModel,TrOCRForCausalLM:()=>l.TrOCRForCausalLM,TrOCRPreTrainedModel:()=>l.TrOCRPreTrainedModel,TranslationPipeline:()=>o.TranslationPipeline,UniSpeechForCTC:()=>l.UniSpeechForCTC,UniSpeechForSequenceClassification:()=>l.UniSpeechForSequenceClassification,UniSpeechModel:()=>l.UniSpeechModel,UniSpeechPreTrainedModel:()=>l.UniSpeechPreTrainedModel,UniSpeechSatForAudioFrameClassification:()=>l.UniSpeechSatForAudioFrameClassification,UniSpeechSatForCTC:()=>l.UniSpeechSatForCTC,UniSpeechSatForSequenceClassification:()=>l.UniSpeechSatForSequenceClassification,UniSpeechSatModel:()=>l.UniSpeechSatModel,UniSpeechSatPreTrainedModel:()=>l.UniSpeechSatPreTrainedModel,ViTFeatureExtractor:()=>d.ViTFeatureExtractor,ViTForImageClassification:()=>l.ViTForImageClassification,ViTImageProcessor:()=>d.ViTImageProcessor,ViTModel:()=>l.ViTModel,ViTPreTrainedModel:()=>l.ViTPreTrainedModel,VisionEncoderDecoderModel:()=>l.VisionEncoderDecoderModel,VitMatteForImageMatting:()=>l.VitMatteForImageMatting,VitMatteImageProcessor:()=>d.VitMatteImageProcessor,VitMattePreTrainedModel:()=>l.VitMattePreTrainedModel,VitsModel:()=>l.VitsModel,VitsModelOutput:()=>l.VitsModelOutput,VitsPreTrainedModel:()=>l.VitsPreTrainedModel,VitsTokenizer:()=>u.VitsTokenizer,Wav2Vec2BertForCTC:()=>l.Wav2Vec2BertForCTC,Wav2Vec2BertForSequenceClassification:()=>l.Wav2Vec2BertForSequenceClassification,Wav2Vec2BertModel:()=>l.Wav2Vec2BertModel,Wav2Vec2BertPreTrainedModel:()=>l.Wav2Vec2BertPreTrainedModel,Wav2Vec2CTCTokenizer:()=>u.Wav2Vec2CTCTokenizer,Wav2Vec2FeatureExtractor:()=>d.Wav2Vec2FeatureExtractor,Wav2Vec2ForAudioFrameClassification:()=>l.Wav2Vec2ForAudioFrameClassification,Wav2Vec2ForCTC:()=>l.Wav2Vec2ForCTC,Wav2Vec2ForSequenceClassification:()=>l.Wav2Vec2ForSequenceClassification,Wav2Vec2Model:()=>l.Wav2Vec2Model,Wav2Vec2PreTrainedModel:()=>l.Wav2Vec2PreTrainedModel,Wav2Vec2ProcessorWithLM:()=>d.Wav2Vec2ProcessorWithLM,WavLMForAudioFrameClassification:()=>l.WavLMForAudioFrameClassification,WavLMForCTC:()=>l.WavLMForCTC,WavLMForSequenceClassification:()=>l.WavLMForSequenceClassification,WavLMForXVector:()=>l.WavLMForXVector,WavLMModel:()=>l.WavLMModel,WavLMPreTrainedModel:()=>l.WavLMPreTrainedModel,WeSpeakerFeatureExtractor:()=>d.WeSpeakerFeatureExtractor,WeSpeakerResNetModel:()=>l.WeSpeakerResNetModel,WeSpeakerResNetPreTrainedModel:()=>l.WeSpeakerResNetPreTrainedModel,WhisperFeatureExtractor:()=>d.WhisperFeatureExtractor,WhisperForConditionalGeneration:()=>l.WhisperForConditionalGeneration,WhisperModel:()=>l.WhisperModel,WhisperPreTrainedModel:()=>l.WhisperPreTrainedModel,WhisperProcessor:()=>d.WhisperProcessor,WhisperTextStreamer:()=>g.WhisperTextStreamer,WhisperTokenizer:()=>u.WhisperTokenizer,XLMForQuestionAnswering:()=>l.XLMForQuestionAnswering,XLMForSequenceClassification:()=>l.XLMForSequenceClassification,XLMForTokenClassification:()=>l.XLMForTokenClassification,XLMModel:()=>l.XLMModel,XLMPreTrainedModel:()=>l.XLMPreTrainedModel,XLMRobertaForMaskedLM:()=>l.XLMRobertaForMaskedLM,XLMRobertaForQuestionAnswering:()=>l.XLMRobertaForQuestionAnswering,XLMRobertaForSequenceClassification:()=>l.XLMRobertaForSequenceClassification,XLMRobertaForTokenClassification:()=>l.XLMRobertaForTokenClassification,XLMRobertaModel:()=>l.XLMRobertaModel,XLMRobertaPreTrainedModel:()=>l.XLMRobertaPreTrainedModel,XLMRobertaTokenizer:()=>u.XLMRobertaTokenizer,XLMTokenizer:()=>u.XLMTokenizer,XLMWithLMHeadModel:()=>l.XLMWithLMHeadModel,XVectorOutput:()=>l.XVectorOutput,YolosFeatureExtractor:()=>d.YolosFeatureExtractor,YolosForObjectDetection:()=>l.YolosForObjectDetection,YolosModel:()=>l.YolosModel,YolosObjectDetectionOutput:()=>l.YolosObjectDetectionOutput,YolosPreTrainedModel:()=>l.YolosPreTrainedModel,ZeroShotAudioClassificationPipeline:()=>o.ZeroShotAudioClassificationPipeline,ZeroShotClassificationPipeline:()=>o.ZeroShotClassificationPipeline,ZeroShotImageClassificationPipeline:()=>o.ZeroShotImageClassificationPipeline,ZeroShotObjectDetectionPipeline:()=>o.ZeroShotObjectDetectionPipeline,bankers_round:()=>f.bankers_round,cat:()=>m.cat,cos_sim:()=>f.cos_sim,dot:()=>f.dot,dynamic_time_warping:()=>f.dynamic_time_warping,env:()=>s.env,full:()=>m.full,full_like:()=>m.full_like,getKeyValueShapes:()=>c.getKeyValueShapes,hamming:()=>p.hamming,hanning:()=>p.hanning,interpolate:()=>m.interpolate,interpolate_4d:()=>m.interpolate_4d,interpolate_data:()=>f.interpolate_data,is_chinese_char:()=>u.is_chinese_char,layer_norm:()=>m.layer_norm,log_softmax:()=>f.log_softmax,magnitude:()=>f.magnitude,matmul:()=>m.matmul,max:()=>f.max,mean:()=>m.mean,mean_pooling:()=>m.mean_pooling,medianFilter:()=>f.medianFilter,mel_filter_bank:()=>p.mel_filter_bank,min:()=>f.min,ones:()=>m.ones,ones_like:()=>m.ones_like,permute:()=>m.permute,permute_data:()=>f.permute_data,pipeline:()=>o.pipeline,quantize_embeddings:()=>m.quantize_embeddings,read_audio:()=>p.read_audio,rfft:()=>m.rfft,round:()=>f.round,softmax:()=>f.softmax,spectrogram:()=>p.spectrogram,stack:()=>m.stack,std_mean:()=>m.std_mean,topk:()=>m.topk,window_function:()=>p.window_function,zeros:()=>m.zeros,zeros_like:()=>m.zeros_like});var s=i(/*! ./env.js */"./src/env.js"),o=i(/*! ./pipelines.js */"./src/pipelines.js"),l=i(/*! ./models.js */"./src/models.js"),u=i(/*! ./tokenizers.js */"./src/tokenizers.js"),d=i(/*! ./processors.js */"./src/processors.js"),c=i(/*! ./configs.js */"./src/configs.js"),p=i(/*! ./utils/audio.js */"./src/utils/audio.js"),h=i(/*! ./utils/image.js */"./src/utils/image.js"),m=i(/*! ./utils/tensor.js */"./src/utils/tensor.js"),f=i(/*! ./utils/maths.js */"./src/utils/maths.js"),g=i(/*! ./generation/streamers.js */"./src/generation/streamers.js"),_=i(/*! ./generation/stopping_criteria.js */"./src/generation/stopping_criteria.js"),w=a.ASTFeatureExtractor,y=a.ASTForAudioClassification,b=a.ASTModel,v=a.ASTPreTrainedModel,x=a.AlbertForMaskedLM,M=a.AlbertForQuestionAnswering,T=a.AlbertForSequenceClassification,k=a.AlbertModel,$=a.AlbertPreTrainedModel,S=a.AlbertTokenizer,C=a.AudioClassificationPipeline,E=a.AutoConfig,P=a.AutoModel,A=a.AutoModelForAudioClassification,F=a.AutoModelForAudioFrameClassification,z=a.AutoModelForCTC,I=a.AutoModelForCausalLM,O=a.AutoModelForDepthEstimation,B=a.AutoModelForDocumentQuestionAnswering,L=a.AutoModelForImageClassification,D=a.AutoModelForImageFeatureExtraction,R=a.AutoModelForImageMatting,N=a.AutoModelForImageSegmentation,V=a.AutoModelForImageToImage,j=a.AutoModelForMaskGeneration,G=a.AutoModelForMaskedLM,q=a.AutoModelForNormalEstimation,U=a.AutoModelForObjectDetection,W=a.AutoModelForQuestionAnswering,H=a.AutoModelForSemanticSegmentation,X=a.AutoModelForSeq2SeqLM,K=a.AutoModelForSequenceClassification,Q=a.AutoModelForSpeechSeq2Seq,Y=a.AutoModelForTextToSpectrogram,J=a.AutoModelForTextToWaveform,Z=a.AutoModelForTokenClassification,ee=a.AutoModelForVision2Seq,te=a.AutoModelForXVector,ne=a.AutoModelForZeroShotObjectDetection,re=a.AutoProcessor,ie=a.AutoTokenizer,ae=a.AutomaticSpeechRecognitionPipeline,se=a.BartForConditionalGeneration,oe=a.BartForSequenceClassification,le=a.BartModel,ue=a.BartPretrainedModel,de=a.BartTokenizer,ce=a.BaseModelOutput,pe=a.BaseStreamer,he=a.BeitFeatureExtractor,me=a.BeitForImageClassification,fe=a.BeitModel,ge=a.BeitPreTrainedModel,_e=a.BertForMaskedLM,we=a.BertForQuestionAnswering,ye=a.BertForSequenceClassification,be=a.BertForTokenClassification,ve=a.BertModel,xe=a.BertPreTrainedModel,Me=a.BertTokenizer,Te=a.BitImageProcessor,ke=a.BlenderbotForConditionalGeneration,$e=a.BlenderbotModel,Se=a.BlenderbotPreTrainedModel,Ce=a.BlenderbotSmallForConditionalGeneration,Ee=a.BlenderbotSmallModel,Pe=a.BlenderbotSmallPreTrainedModel,Ae=a.BlenderbotSmallTokenizer,Fe=a.BlenderbotTokenizer,ze=a.BloomForCausalLM,Ie=a.BloomModel,Oe=a.BloomPreTrainedModel,Be=a.BloomTokenizer,Le=a.CLIPFeatureExtractor,De=a.CLIPImageProcessor,Re=a.CLIPModel,Ne=a.CLIPPreTrainedModel,Ve=a.CLIPSegForImageSegmentation,je=a.CLIPSegModel,Ge=a.CLIPSegPreTrainedModel,qe=a.CLIPTextModelWithProjection,Ue=a.CLIPTokenizer,We=a.CLIPVisionModelWithProjection,He=a.CamembertForMaskedLM,Xe=a.CamembertForQuestionAnswering,Ke=a.CamembertForSequenceClassification,Qe=a.CamembertForTokenClassification,Ye=a.CamembertModel,Je=a.CamembertPreTrainedModel,Ze=a.CamembertTokenizer,et=a.CausalLMOutput,tt=a.CausalLMOutputWithPast,nt=a.ChineseCLIPFeatureExtractor,rt=a.ChineseCLIPModel,it=a.ChineseCLIPPreTrainedModel,at=a.ClapAudioModelWithProjection,st=a.ClapFeatureExtractor,ot=a.ClapModel,lt=a.ClapPreTrainedModel,ut=a.ClapTextModelWithProjection,dt=a.CodeGenForCausalLM,ct=a.CodeGenModel,pt=a.CodeGenPreTrainedModel,ht=a.CodeGenTokenizer,mt=a.CodeLlamaTokenizer,ft=a.CohereForCausalLM,gt=a.CohereModel,_t=a.CoherePreTrainedModel,wt=a.CohereTokenizer,yt=a.ConvBertForMaskedLM,bt=a.ConvBertForQuestionAnswering,vt=a.ConvBertForSequenceClassification,xt=a.ConvBertForTokenClassification,Mt=a.ConvBertModel,Tt=a.ConvBertPreTrainedModel,kt=a.ConvBertTokenizer,$t=a.ConvNextFeatureExtractor,St=a.ConvNextForImageClassification,Ct=a.ConvNextImageProcessor,Et=a.ConvNextModel,Pt=a.ConvNextPreTrainedModel,At=a.ConvNextV2ForImageClassification,Ft=a.ConvNextV2Model,zt=a.ConvNextV2PreTrainedModel,It=a.DPTFeatureExtractor,Ot=a.DPTForDepthEstimation,Bt=a.DPTImageProcessor,Lt=a.DPTModel,Dt=a.DPTPreTrainedModel,Rt=a.DebertaForMaskedLM,Nt=a.DebertaForQuestionAnswering,Vt=a.DebertaForSequenceClassification,jt=a.DebertaForTokenClassification,Gt=a.DebertaModel,qt=a.DebertaPreTrainedModel,Ut=a.DebertaTokenizer,Wt=a.DebertaV2ForMaskedLM,Ht=a.DebertaV2ForQuestionAnswering,Xt=a.DebertaV2ForSequenceClassification,Kt=a.DebertaV2ForTokenClassification,Qt=a.DebertaV2Model,Yt=a.DebertaV2PreTrainedModel,Jt=a.DebertaV2Tokenizer,Zt=a.DeiTFeatureExtractor,en=a.DeiTForImageClassification,tn=a.DeiTModel,nn=a.DeiTPreTrainedModel,rn=a.DepthAnythingForDepthEstimation,an=a.DepthAnythingPreTrainedModel,sn=a.DepthEstimationPipeline,on=a.DetrFeatureExtractor,ln=a.DetrForObjectDetection,un=a.DetrForSegmentation,dn=a.DetrModel,cn=a.DetrObjectDetectionOutput,pn=a.DetrPreTrainedModel,hn=a.DetrSegmentationOutput,mn=a.Dinov2ForImageClassification,fn=a.Dinov2Model,gn=a.Dinov2PreTrainedModel,_n=a.DistilBertForMaskedLM,wn=a.DistilBertForQuestionAnswering,yn=a.DistilBertForSequenceClassification,bn=a.DistilBertForTokenClassification,vn=a.DistilBertModel,xn=a.DistilBertPreTrainedModel,Mn=a.DistilBertTokenizer,Tn=a.DocumentQuestionAnsweringPipeline,kn=a.DonutFeatureExtractor,$n=a.DonutSwinModel,Sn=a.DonutSwinPreTrainedModel,Cn=a.EfficientNetForImageClassification,En=a.EfficientNetImageProcessor,Pn=a.EfficientNetModel,An=a.EfficientNetPreTrainedModel,Fn=a.ElectraForMaskedLM,zn=a.ElectraForQuestionAnswering,In=a.ElectraForSequenceClassification,On=a.ElectraForTokenClassification,Bn=a.ElectraModel,Ln=a.ElectraPreTrainedModel,Dn=a.ElectraTokenizer,Rn=a.EosTokenCriteria,Nn=a.EsmForMaskedLM,Vn=a.EsmForSequenceClassification,jn=a.EsmForTokenClassification,Gn=a.EsmModel,qn=a.EsmPreTrainedModel,Un=a.EsmTokenizer,Wn=a.FFT,Hn=a.FalconForCausalLM,Xn=a.FalconModel,Kn=a.FalconPreTrainedModel,Qn=a.FalconTokenizer,Yn=a.FastViTForImageClassification,Jn=a.FastViTModel,Zn=a.FastViTPreTrainedModel,er=a.FeatureExtractionPipeline,tr=a.FeatureExtractor,nr=a.FillMaskPipeline,rr=a.Florence2ForConditionalGeneration,ir=a.Florence2PreTrainedModel,ar=a.Florence2Processor,sr=a.GLPNFeatureExtractor,or=a.GLPNForDepthEstimation,lr=a.GLPNModel,ur=a.GLPNPreTrainedModel,dr=a.GPT2LMHeadModel,cr=a.GPT2Model,pr=a.GPT2PreTrainedModel,hr=a.GPT2Tokenizer,mr=a.GPTBigCodeForCausalLM,fr=a.GPTBigCodeModel,gr=a.GPTBigCodePreTrainedModel,_r=a.GPTJForCausalLM,wr=a.GPTJModel,yr=a.GPTJPreTrainedModel,br=a.GPTNeoForCausalLM,vr=a.GPTNeoModel,xr=a.GPTNeoPreTrainedModel,Mr=a.GPTNeoXForCausalLM,Tr=a.GPTNeoXModel,kr=a.GPTNeoXPreTrainedModel,$r=a.GPTNeoXTokenizer,Sr=a.Gemma2ForCausalLM,Cr=a.Gemma2Model,Er=a.Gemma2PreTrainedModel,Pr=a.GemmaForCausalLM,Ar=a.GemmaModel,Fr=a.GemmaPreTrainedModel,zr=a.GemmaTokenizer,Ir=a.Grok1Tokenizer,Or=a.HerbertTokenizer,Br=a.HubertForCTC,Lr=a.HubertForSequenceClassification,Dr=a.HubertModel,Rr=a.HubertPreTrainedModel,Nr=a.ImageClassificationPipeline,Vr=a.ImageFeatureExtractionPipeline,jr=a.ImageFeatureExtractor,Gr=a.ImageMattingOutput,qr=a.ImageSegmentationPipeline,Ur=a.ImageToImagePipeline,Wr=a.ImageToTextPipeline,Hr=a.InterruptableStoppingCriteria,Xr=a.JAISLMHeadModel,Kr=a.JAISModel,Qr=a.JAISPreTrainedModel,Yr=a.LlamaForCausalLM,Jr=a.LlamaModel,Zr=a.LlamaPreTrainedModel,ei=a.LlamaTokenizer,ti=a.LlavaForConditionalGeneration,ni=a.LlavaPreTrainedModel,ri=a.LongT5ForConditionalGeneration,ii=a.LongT5Model,ai=a.LongT5PreTrainedModel,si=a.M2M100ForConditionalGeneration,oi=a.M2M100Model,li=a.M2M100PreTrainedModel,ui=a.M2M100Tokenizer,di=a.MBart50Tokenizer,ci=a.MBartForCausalLM,pi=a.MBartForConditionalGeneration,hi=a.MBartForSequenceClassification,mi=a.MBartModel,fi=a.MBartPreTrainedModel,gi=a.MBartTokenizer,_i=a.MPNetForMaskedLM,wi=a.MPNetForQuestionAnswering,yi=a.MPNetForSequenceClassification,bi=a.MPNetForTokenClassification,vi=a.MPNetModel,xi=a.MPNetPreTrainedModel,Mi=a.MPNetTokenizer,Ti=a.MT5ForConditionalGeneration,ki=a.MT5Model,$i=a.MT5PreTrainedModel,Si=a.MarianMTModel,Ci=a.MarianModel,Ei=a.MarianPreTrainedModel,Pi=a.MarianTokenizer,Ai=a.MaskedLMOutput,Fi=a.MaxLengthCriteria,zi=a.MistralForCausalLM,Ii=a.MistralModel,Oi=a.MistralPreTrainedModel,Bi=a.MobileBertForMaskedLM,Li=a.MobileBertForQuestionAnswering,Di=a.MobileBertForSequenceClassification,Ri=a.MobileBertModel,Ni=a.MobileBertPreTrainedModel,Vi=a.MobileBertTokenizer,ji=a.MobileNetV1FeatureExtractor,Gi=a.MobileNetV1ForImageClassification,qi=a.MobileNetV1Model,Ui=a.MobileNetV1PreTrainedModel,Wi=a.MobileNetV2FeatureExtractor,Hi=a.MobileNetV2ForImageClassification,Xi=a.MobileNetV2Model,Ki=a.MobileNetV2PreTrainedModel,Qi=a.MobileNetV3FeatureExtractor,Yi=a.MobileNetV3ForImageClassification,Ji=a.MobileNetV3Model,Zi=a.MobileNetV3PreTrainedModel,ea=a.MobileNetV4FeatureExtractor,ta=a.MobileNetV4ForImageClassification,na=a.MobileNetV4Model,ra=a.MobileNetV4PreTrainedModel,ia=a.MobileViTFeatureExtractor,aa=a.MobileViTForImageClassification,sa=a.MobileViTImageProcessor,oa=a.MobileViTModel,la=a.MobileViTPreTrainedModel,ua=a.MobileViTV2ForImageClassification,da=a.MobileViTV2Model,ca=a.MobileViTV2PreTrainedModel,pa=a.ModelOutput,ha=a.Moondream1ForConditionalGeneration,ma=a.MptForCausalLM,fa=a.MptModel,ga=a.MptPreTrainedModel,_a=a.MusicgenForCausalLM,wa=a.MusicgenForConditionalGeneration,ya=a.MusicgenModel,ba=a.MusicgenPreTrainedModel,va=a.NllbTokenizer,xa=a.NomicBertModel,Ma=a.NomicBertPreTrainedModel,Ta=a.NougatImageProcessor,ka=a.NougatTokenizer,$a=a.OPTForCausalLM,Sa=a.OPTModel,Ca=a.OPTPreTrainedModel,Ea=a.ObjectDetectionPipeline,Pa=a.OpenELMForCausalLM,Aa=a.OpenELMModel,Fa=a.OpenELMPreTrainedModel,za=a.OwlViTFeatureExtractor,Ia=a.OwlViTForObjectDetection,Oa=a.OwlViTModel,Ba=a.OwlViTPreTrainedModel,La=a.OwlViTProcessor,Da=a.Owlv2ForObjectDetection,Ra=a.Owlv2ImageProcessor,Na=a.Owlv2Model,Va=a.Owlv2PreTrainedModel,ja=a.Phi3ForCausalLM,Ga=a.Phi3Model,qa=a.Phi3PreTrainedModel,Ua=a.PhiForCausalLM,Wa=a.PhiModel,Ha=a.PhiPreTrainedModel,Xa=a.Pipeline,Ka=a.PreTrainedModel,Qa=a.PreTrainedTokenizer,Ya=a.PretrainedConfig,Ja=a.PretrainedMixin,Za=a.Processor,es=a.PyAnnoteFeatureExtractor,ts=a.PyAnnoteForAudioFrameClassification,ns=a.PyAnnoteModel,rs=a.PyAnnotePreTrainedModel,is=a.PyAnnoteProcessor,as=a.QuestionAnsweringModelOutput,ss=a.QuestionAnsweringPipeline,os=a.Qwen2ForCausalLM,ls=a.Qwen2Model,us=a.Qwen2PreTrainedModel,ds=a.Qwen2Tokenizer,cs=a.RTDetrForObjectDetection,ps=a.RTDetrImageProcessor,hs=a.RTDetrModel,ms=a.RTDetrObjectDetectionOutput,fs=a.RTDetrPreTrainedModel,gs=a.RawImage,_s=a.ResNetForImageClassification,ws=a.ResNetModel,ys=a.ResNetPreTrainedModel,bs=a.RoFormerForMaskedLM,vs=a.RoFormerForQuestionAnswering,xs=a.RoFormerForSequenceClassification,Ms=a.RoFormerForTokenClassification,Ts=a.RoFormerModel,ks=a.RoFormerPreTrainedModel,$s=a.RoFormerTokenizer,Ss=a.RobertaForMaskedLM,Cs=a.RobertaForQuestionAnswering,Es=a.RobertaForSequenceClassification,Ps=a.RobertaForTokenClassification,As=a.RobertaModel,Fs=a.RobertaPreTrainedModel,zs=a.RobertaTokenizer,Is=a.SamImageProcessor,Os=a.SamImageSegmentationOutput,Bs=a.SamModel,Ls=a.SamPreTrainedModel,Ds=a.SamProcessor,Rs=a.SapiensFeatureExtractor,Ns=a.SapiensForDepthEstimation,Vs=a.SapiensForNormalEstimation,js=a.SapiensForSemanticSegmentation,Gs=a.SapiensPreTrainedModel,qs=a.SeamlessM4TFeatureExtractor,Us=a.SegformerFeatureExtractor,Ws=a.SegformerForImageClassification,Hs=a.SegformerForSemanticSegmentation,Xs=a.SegformerModel,Ks=a.SegformerPreTrainedModel,Qs=a.Seq2SeqLMOutput,Ys=a.SequenceClassifierOutput,Js=a.SiglipImageProcessor,Zs=a.SiglipModel,eo=a.SiglipPreTrainedModel,to=a.SiglipTextModel,no=a.SiglipTokenizer,ro=a.SiglipVisionModel,io=a.SpeechT5FeatureExtractor,ao=a.SpeechT5ForSpeechToText,so=a.SpeechT5ForTextToSpeech,oo=a.SpeechT5HifiGan,lo=a.SpeechT5Model,uo=a.SpeechT5PreTrainedModel,co=a.SpeechT5Processor,po=a.SpeechT5Tokenizer,ho=a.SqueezeBertForMaskedLM,mo=a.SqueezeBertForQuestionAnswering,fo=a.SqueezeBertForSequenceClassification,go=a.SqueezeBertModel,_o=a.SqueezeBertPreTrainedModel,wo=a.SqueezeBertTokenizer,yo=a.StableLmForCausalLM,bo=a.StableLmModel,vo=a.StableLmPreTrainedModel,xo=a.Starcoder2ForCausalLM,Mo=a.Starcoder2Model,To=a.Starcoder2PreTrainedModel,ko=a.StoppingCriteria,$o=a.StoppingCriteriaList,So=a.SummarizationPipeline,Co=a.Swin2SRForImageSuperResolution,Eo=a.Swin2SRImageProcessor,Po=a.Swin2SRModel,Ao=a.Swin2SRPreTrainedModel,Fo=a.SwinForImageClassification,zo=a.SwinModel,Io=a.SwinPreTrainedModel,Oo=a.T5ForConditionalGeneration,Bo=a.T5Model,Lo=a.T5PreTrainedModel,Do=a.T5Tokenizer,Ro=a.TableTransformerForObjectDetection,No=a.TableTransformerModel,Vo=a.TableTransformerObjectDetectionOutput,jo=a.TableTransformerPreTrainedModel,Go=a.Tensor,qo=a.Text2TextGenerationPipeline,Uo=a.TextClassificationPipeline,Wo=a.TextGenerationPipeline,Ho=a.TextStreamer,Xo=a.TextToAudioPipeline,Ko=a.TokenClassificationPipeline,Qo=a.TokenClassifierOutput,Yo=a.TokenizerModel,Jo=a.TrOCRForCausalLM,Zo=a.TrOCRPreTrainedModel,el=a.TranslationPipeline,tl=a.UniSpeechForCTC,nl=a.UniSpeechForSequenceClassification,rl=a.UniSpeechModel,il=a.UniSpeechPreTrainedModel,al=a.UniSpeechSatForAudioFrameClassification,sl=a.UniSpeechSatForCTC,ol=a.UniSpeechSatForSequenceClassification,ll=a.UniSpeechSatModel,ul=a.UniSpeechSatPreTrainedModel,dl=a.ViTFeatureExtractor,cl=a.ViTForImageClassification,pl=a.ViTImageProcessor,hl=a.ViTModel,ml=a.ViTPreTrainedModel,fl=a.VisionEncoderDecoderModel,gl=a.VitMatteForImageMatting,_l=a.VitMatteImageProcessor,wl=a.VitMattePreTrainedModel,yl=a.VitsModel,bl=a.VitsModelOutput,vl=a.VitsPreTrainedModel,xl=a.VitsTokenizer,Ml=a.Wav2Vec2BertForCTC,Tl=a.Wav2Vec2BertForSequenceClassification,kl=a.Wav2Vec2BertModel,$l=a.Wav2Vec2BertPreTrainedModel,Sl=a.Wav2Vec2CTCTokenizer,Cl=a.Wav2Vec2FeatureExtractor,El=a.Wav2Vec2ForAudioFrameClassification,Pl=a.Wav2Vec2ForCTC,Al=a.Wav2Vec2ForSequenceClassification,Fl=a.Wav2Vec2Model,zl=a.Wav2Vec2PreTrainedModel,Il=a.Wav2Vec2ProcessorWithLM,Ol=a.WavLMForAudioFrameClassification,Bl=a.WavLMForCTC,Ll=a.WavLMForSequenceClassification,Dl=a.WavLMForXVector,Rl=a.WavLMModel,Nl=a.WavLMPreTrainedModel,Vl=a.WeSpeakerFeatureExtractor,jl=a.WeSpeakerResNetModel,Gl=a.WeSpeakerResNetPreTrainedModel,ql=a.WhisperFeatureExtractor,Ul=a.WhisperForConditionalGeneration,Wl=a.WhisperModel,Hl=a.WhisperPreTrainedModel,Xl=a.WhisperProcessor,Kl=a.WhisperTextStreamer,Ql=a.WhisperTokenizer,Yl=a.XLMForQuestionAnswering,Jl=a.XLMForSequenceClassification,Zl=a.XLMForTokenClassification,eu=a.XLMModel,tu=a.XLMPreTrainedModel,nu=a.XLMRobertaForMaskedLM,ru=a.XLMRobertaForQuestionAnswering,iu=a.XLMRobertaForSequenceClassification,au=a.XLMRobertaForTokenClassification,su=a.XLMRobertaModel,ou=a.XLMRobertaPreTrainedModel,lu=a.XLMRobertaTokenizer,uu=a.XLMTokenizer,du=a.XLMWithLMHeadModel,cu=a.XVectorOutput,pu=a.YolosFeatureExtractor,hu=a.YolosForObjectDetection,mu=a.YolosModel,fu=a.YolosObjectDetectionOutput,gu=a.YolosPreTrainedModel,_u=a.ZeroShotAudioClassificationPipeline,wu=a.ZeroShotClassificationPipeline,yu=a.ZeroShotImageClassificationPipeline,bu=a.ZeroShotObjectDetectionPipeline,vu=a.bankers_round,xu=a.cat,Mu=a.cos_sim,Tu=a.dot,ku=a.dynamic_time_warping,$u=a.env,Su=a.full,Cu=a.full_like,Eu=a.getKeyValueShapes,Pu=a.hamming,Au=a.hanning,Fu=a.interpolate,zu=a.interpolate_4d,Iu=a.interpolate_data,Ou=a.is_chinese_char,Bu=a.layer_norm,Lu=a.log_softmax,Du=a.magnitude,Ru=a.matmul,Nu=a.max,Vu=a.mean,ju=a.mean_pooling,Gu=a.medianFilter,qu=a.mel_filter_bank,Uu=a.min,Wu=a.ones,Hu=a.ones_like,Xu=a.permute,Ku=a.permute_data,Qu=a.pipeline,Yu=a.quantize_embeddings,Ju=a.read_audio,Zu=a.rfft,ed=a.round,td=a.softmax,nd=a.spectrogram,rd=a.stack,id=a.std_mean,ad=a.topk,sd=a.window_function,od=a.zeros,ld=a.zeros_like;export{w as ASTFeatureExtractor,y as ASTForAudioClassification,b as ASTModel,v as ASTPreTrainedModel,x as AlbertForMaskedLM,M as AlbertForQuestionAnswering,T as AlbertForSequenceClassification,k as AlbertModel,$ as AlbertPreTrainedModel,S as AlbertTokenizer,C as AudioClassificationPipeline,E as AutoConfig,P as AutoModel,A as AutoModelForAudioClassification,F as AutoModelForAudioFrameClassification,z as AutoModelForCTC,I as AutoModelForCausalLM,O as AutoModelForDepthEstimation,B as AutoModelForDocumentQuestionAnswering,L as AutoModelForImageClassification,D as AutoModelForImageFeatureExtraction,R as AutoModelForImageMatting,N as AutoModelForImageSegmentation,V as AutoModelForImageToImage,j as AutoModelForMaskGeneration,G as AutoModelForMaskedLM,q as AutoModelForNormalEstimation,U as AutoModelForObjectDetection,W as AutoModelForQuestionAnswering,H as AutoModelForSemanticSegmentation,X as AutoModelForSeq2SeqLM,K as AutoModelForSequenceClassification,Q as AutoModelForSpeechSeq2Seq,Y as AutoModelForTextToSpectrogram,J as AutoModelForTextToWaveform,Z as AutoModelForTokenClassification,ee as AutoModelForVision2Seq,te as AutoModelForXVector,ne as AutoModelForZeroShotObjectDetection,re as AutoProcessor,ie as AutoTokenizer,ae as AutomaticSpeechRecognitionPipeline,se as BartForConditionalGeneration,oe as BartForSequenceClassification,le as BartModel,ue as BartPretrainedModel,de as BartTokenizer,ce as BaseModelOutput,pe as BaseStreamer,he as BeitFeatureExtractor,me as BeitForImageClassification,fe as BeitModel,ge as BeitPreTrainedModel,_e as BertForMaskedLM,we as BertForQuestionAnswering,ye as BertForSequenceClassification,be as BertForTokenClassification,ve as BertModel,xe as BertPreTrainedModel,Me as BertTokenizer,Te as BitImageProcessor,ke as BlenderbotForConditionalGeneration,$e as BlenderbotModel,Se as BlenderbotPreTrainedModel,Ce as BlenderbotSmallForConditionalGeneration,Ee as BlenderbotSmallModel,Pe as BlenderbotSmallPreTrainedModel,Ae as BlenderbotSmallTokenizer,Fe as BlenderbotTokenizer,ze as BloomForCausalLM,Ie as BloomModel,Oe as BloomPreTrainedModel,Be as BloomTokenizer,Le as CLIPFeatureExtractor,De as CLIPImageProcessor,Re as CLIPModel,Ne as CLIPPreTrainedModel,Ve as CLIPSegForImageSegmentation,je as CLIPSegModel,Ge as CLIPSegPreTrainedModel,qe as CLIPTextModelWithProjection,Ue as CLIPTokenizer,We as CLIPVisionModelWithProjection,He as CamembertForMaskedLM,Xe as CamembertForQuestionAnswering,Ke as CamembertForSequenceClassification,Qe as CamembertForTokenClassification,Ye as CamembertModel,Je as CamembertPreTrainedModel,Ze as CamembertTokenizer,et as CausalLMOutput,tt as CausalLMOutputWithPast,nt as ChineseCLIPFeatureExtractor,rt as ChineseCLIPModel,it as ChineseCLIPPreTrainedModel,at as ClapAudioModelWithProjection,st as ClapFeatureExtractor,ot as ClapModel,lt as ClapPreTrainedModel,ut as ClapTextModelWithProjection,dt as CodeGenForCausalLM,ct as CodeGenModel,pt as CodeGenPreTrainedModel,ht as CodeGenTokenizer,mt as CodeLlamaTokenizer,ft as CohereForCausalLM,gt as CohereModel,_t as CoherePreTrainedModel,wt as CohereTokenizer,yt as ConvBertForMaskedLM,bt as ConvBertForQuestionAnswering,vt as ConvBertForSequenceClassification,xt as ConvBertForTokenClassification,Mt as ConvBertModel,Tt as ConvBertPreTrainedModel,kt as ConvBertTokenizer,$t as ConvNextFeatureExtractor,St as ConvNextForImageClassification,Ct as ConvNextImageProcessor,Et as ConvNextModel,Pt as ConvNextPreTrainedModel,At as ConvNextV2ForImageClassification,Ft as ConvNextV2Model,zt as ConvNextV2PreTrainedModel,It as DPTFeatureExtractor,Ot as DPTForDepthEstimation,Bt as DPTImageProcessor,Lt as DPTModel,Dt as DPTPreTrainedModel,Rt as DebertaForMaskedLM,Nt as DebertaForQuestionAnswering,Vt as DebertaForSequenceClassification,jt as DebertaForTokenClassification,Gt as DebertaModel,qt as DebertaPreTrainedModel,Ut as DebertaTokenizer,Wt as DebertaV2ForMaskedLM,Ht as DebertaV2ForQuestionAnswering,Xt as DebertaV2ForSequenceClassification,Kt as DebertaV2ForTokenClassification,Qt as DebertaV2Model,Yt as DebertaV2PreTrainedModel,Jt as DebertaV2Tokenizer,Zt as DeiTFeatureExtractor,en as DeiTForImageClassification,tn as DeiTModel,nn as DeiTPreTrainedModel,rn as DepthAnythingForDepthEstimation,an as DepthAnythingPreTrainedModel,sn as DepthEstimationPipeline,on as DetrFeatureExtractor,ln as DetrForObjectDetection,un as DetrForSegmentation,dn as DetrModel,cn as DetrObjectDetectionOutput,pn as DetrPreTrainedModel,hn as DetrSegmentationOutput,mn as Dinov2ForImageClassification,fn as Dinov2Model,gn as Dinov2PreTrainedModel,_n as DistilBertForMaskedLM,wn as DistilBertForQuestionAnswering,yn as DistilBertForSequenceClassification,bn as DistilBertForTokenClassification,vn as DistilBertModel,xn as DistilBertPreTrainedModel,Mn as DistilBertTokenizer,Tn as DocumentQuestionAnsweringPipeline,kn as DonutFeatureExtractor,$n as DonutSwinModel,Sn as DonutSwinPreTrainedModel,Cn as EfficientNetForImageClassification,En as EfficientNetImageProcessor,Pn as EfficientNetModel,An as EfficientNetPreTrainedModel,Fn as ElectraForMaskedLM,zn as ElectraForQuestionAnswering,In as ElectraForSequenceClassification,On as ElectraForTokenClassification,Bn as ElectraModel,Ln as ElectraPreTrainedModel,Dn as ElectraTokenizer,Rn as EosTokenCriteria,Nn as EsmForMaskedLM,Vn as EsmForSequenceClassification,jn as EsmForTokenClassification,Gn as EsmModel,qn as EsmPreTrainedModel,Un as EsmTokenizer,Wn as FFT,Hn as FalconForCausalLM,Xn as FalconModel,Kn as FalconPreTrainedModel,Qn as FalconTokenizer,Yn as FastViTForImageClassification,Jn as FastViTModel,Zn as FastViTPreTrainedModel,er as FeatureExtractionPipeline,tr as FeatureExtractor,nr as FillMaskPipeline,rr as Florence2ForConditionalGeneration,ir as Florence2PreTrainedModel,ar as Florence2Processor,sr as GLPNFeatureExtractor,or as GLPNForDepthEstimation,lr as GLPNModel,ur as GLPNPreTrainedModel,dr as GPT2LMHeadModel,cr as GPT2Model,pr as GPT2PreTrainedModel,hr as GPT2Tokenizer,mr as GPTBigCodeForCausalLM,fr as GPTBigCodeModel,gr as GPTBigCodePreTrainedModel,_r as GPTJForCausalLM,wr as GPTJModel,yr as GPTJPreTrainedModel,br as GPTNeoForCausalLM,vr as GPTNeoModel,xr as GPTNeoPreTrainedModel,Mr as GPTNeoXForCausalLM,Tr as GPTNeoXModel,kr as GPTNeoXPreTrainedModel,$r as GPTNeoXTokenizer,Sr as Gemma2ForCausalLM,Cr as Gemma2Model,Er as Gemma2PreTrainedModel,Pr as GemmaForCausalLM,Ar as GemmaModel,Fr as GemmaPreTrainedModel,zr as GemmaTokenizer,Ir as Grok1Tokenizer,Or as HerbertTokenizer,Br as HubertForCTC,Lr as HubertForSequenceClassification,Dr as HubertModel,Rr as HubertPreTrainedModel,Nr as ImageClassificationPipeline,Vr as ImageFeatureExtractionPipeline,jr as ImageFeatureExtractor,Gr as ImageMattingOutput,qr as ImageSegmentationPipeline,Ur as ImageToImagePipeline,Wr as ImageToTextPipeline,Hr as InterruptableStoppingCriteria,Xr as JAISLMHeadModel,Kr as JAISModel,Qr as JAISPreTrainedModel,Yr as LlamaForCausalLM,Jr as LlamaModel,Zr as LlamaPreTrainedModel,ei as LlamaTokenizer,ti as LlavaForConditionalGeneration,ni as LlavaPreTrainedModel,ri as LongT5ForConditionalGeneration,ii as LongT5Model,ai as LongT5PreTrainedModel,si as M2M100ForConditionalGeneration,oi as M2M100Model,li as M2M100PreTrainedModel,ui as M2M100Tokenizer,di as MBart50Tokenizer,ci as MBartForCausalLM,pi as MBartForConditionalGeneration,hi as MBartForSequenceClassification,mi as MBartModel,fi as MBartPreTrainedModel,gi as MBartTokenizer,_i as MPNetForMaskedLM,wi as MPNetForQuestionAnswering,yi as MPNetForSequenceClassification,bi as MPNetForTokenClassification,vi as MPNetModel,xi as MPNetPreTrainedModel,Mi as MPNetTokenizer,Ti as MT5ForConditionalGeneration,ki as MT5Model,$i as MT5PreTrainedModel,Si as MarianMTModel,Ci as MarianModel,Ei as MarianPreTrainedModel,Pi as MarianTokenizer,Ai as MaskedLMOutput,Fi as MaxLengthCriteria,zi as MistralForCausalLM,Ii as MistralModel,Oi as MistralPreTrainedModel,Bi as MobileBertForMaskedLM,Li as MobileBertForQuestionAnswering,Di as MobileBertForSequenceClassification,Ri as MobileBertModel,Ni as MobileBertPreTrainedModel,Vi as MobileBertTokenizer,ji as MobileNetV1FeatureExtractor,Gi as MobileNetV1ForImageClassification,qi as MobileNetV1Model,Ui as MobileNetV1PreTrainedModel,Wi as MobileNetV2FeatureExtractor,Hi as MobileNetV2ForImageClassification,Xi as MobileNetV2Model,Ki as MobileNetV2PreTrainedModel,Qi as MobileNetV3FeatureExtractor,Yi as MobileNetV3ForImageClassification,Ji as MobileNetV3Model,Zi as MobileNetV3PreTrainedModel,ea as MobileNetV4FeatureExtractor,ta as MobileNetV4ForImageClassification,na as MobileNetV4Model,ra as MobileNetV4PreTrainedModel,ia as MobileViTFeatureExtractor,aa as MobileViTForImageClassification,sa as MobileViTImageProcessor,oa as MobileViTModel,la as MobileViTPreTrainedModel,ua as MobileViTV2ForImageClassification,da as MobileViTV2Model,ca as MobileViTV2PreTrainedModel,pa as ModelOutput,ha as Moondream1ForConditionalGeneration,ma as MptForCausalLM,fa as MptModel,ga as MptPreTrainedModel,_a as MusicgenForCausalLM,wa as MusicgenForConditionalGeneration,ya as MusicgenModel,ba as MusicgenPreTrainedModel,va as NllbTokenizer,xa as NomicBertModel,Ma as NomicBertPreTrainedModel,Ta as NougatImageProcessor,ka as NougatTokenizer,$a as OPTForCausalLM,Sa as OPTModel,Ca as OPTPreTrainedModel,Ea as ObjectDetectionPipeline,Pa as OpenELMForCausalLM,Aa as OpenELMModel,Fa as OpenELMPreTrainedModel,za as OwlViTFeatureExtractor,Ia as OwlViTForObjectDetection,Oa as OwlViTModel,Ba as OwlViTPreTrainedModel,La as OwlViTProcessor,Da as Owlv2ForObjectDetection,Ra as Owlv2ImageProcessor,Na as Owlv2Model,Va as Owlv2PreTrainedModel,ja as Phi3ForCausalLM,Ga as Phi3Model,qa as Phi3PreTrainedModel,Ua as PhiForCausalLM,Wa as PhiModel,Ha as PhiPreTrainedModel,Xa as Pipeline,Ka as PreTrainedModel,Qa as PreTrainedTokenizer,Ya as PretrainedConfig,Ja as PretrainedMixin,Za as Processor,es as PyAnnoteFeatureExtractor,ts as PyAnnoteForAudioFrameClassification,ns as PyAnnoteModel,rs as PyAnnotePreTrainedModel,is as PyAnnoteProcessor,as as QuestionAnsweringModelOutput,ss as QuestionAnsweringPipeline,os as Qwen2ForCausalLM,ls as Qwen2Model,us as Qwen2PreTrainedModel,ds as Qwen2Tokenizer,cs as RTDetrForObjectDetection,ps as RTDetrImageProcessor,hs as RTDetrModel,ms as RTDetrObjectDetectionOutput,fs as RTDetrPreTrainedModel,gs as RawImage,_s as ResNetForImageClassification,ws as ResNetModel,ys as ResNetPreTrainedModel,bs as RoFormerForMaskedLM,vs as RoFormerForQuestionAnswering,xs as RoFormerForSequenceClassification,Ms as RoFormerForTokenClassification,Ts as RoFormerModel,ks as RoFormerPreTrainedModel,$s as RoFormerTokenizer,Ss as RobertaForMaskedLM,Cs as RobertaForQuestionAnswering,Es as RobertaForSequenceClassification,Ps as RobertaForTokenClassification,As as RobertaModel,Fs as RobertaPreTrainedModel,zs as RobertaTokenizer,Is as SamImageProcessor,Os as SamImageSegmentationOutput,Bs as SamModel,Ls as SamPreTrainedModel,Ds as SamProcessor,Rs as SapiensFeatureExtractor,Ns as SapiensForDepthEstimation,Vs as SapiensForNormalEstimation,js as SapiensForSemanticSegmentation,Gs as SapiensPreTrainedModel,qs as SeamlessM4TFeatureExtractor,Us as SegformerFeatureExtractor,Ws as SegformerForImageClassification,Hs as SegformerForSemanticSegmentation,Xs as SegformerModel,Ks as SegformerPreTrainedModel,Qs as Seq2SeqLMOutput,Ys as SequenceClassifierOutput,Js as SiglipImageProcessor,Zs as SiglipModel,eo as SiglipPreTrainedModel,to as SiglipTextModel,no as SiglipTokenizer,ro as SiglipVisionModel,io as SpeechT5FeatureExtractor,ao as SpeechT5ForSpeechToText,so as SpeechT5ForTextToSpeech,oo as SpeechT5HifiGan,lo as SpeechT5Model,uo as SpeechT5PreTrainedModel,co as SpeechT5Processor,po as SpeechT5Tokenizer,ho as SqueezeBertForMaskedLM,mo as SqueezeBertForQuestionAnswering,fo as SqueezeBertForSequenceClassification,go as SqueezeBertModel,_o as SqueezeBertPreTrainedModel,wo as SqueezeBertTokenizer,yo as StableLmForCausalLM,bo as StableLmModel,vo as StableLmPreTrainedModel,xo as Starcoder2ForCausalLM,Mo as Starcoder2Model,To as Starcoder2PreTrainedModel,ko as StoppingCriteria,$o as StoppingCriteriaList,So as SummarizationPipeline,Co as Swin2SRForImageSuperResolution,Eo as Swin2SRImageProcessor,Po as Swin2SRModel,Ao as Swin2SRPreTrainedModel,Fo as SwinForImageClassification,zo as SwinModel,Io as SwinPreTrainedModel,Oo as T5ForConditionalGeneration,Bo as T5Model,Lo as T5PreTrainedModel,Do as T5Tokenizer,Ro as TableTransformerForObjectDetection,No as TableTransformerModel,Vo as TableTransformerObjectDetectionOutput,jo as TableTransformerPreTrainedModel,Go as Tensor,qo as Text2TextGenerationPipeline,Uo as TextClassificationPipeline,Wo as TextGenerationPipeline,Ho as TextStreamer,Xo as TextToAudioPipeline,Ko as TokenClassificationPipeline,Qo as TokenClassifierOutput,Yo as TokenizerModel,Jo as TrOCRForCausalLM,Zo as TrOCRPreTrainedModel,el as TranslationPipeline,tl as UniSpeechForCTC,nl as UniSpeechForSequenceClassification,rl as UniSpeechModel,il as UniSpeechPreTrainedModel,al as UniSpeechSatForAudioFrameClassification,sl as UniSpeechSatForCTC,ol as UniSpeechSatForSequenceClassification,ll as UniSpeechSatModel,ul as UniSpeechSatPreTrainedModel,dl as ViTFeatureExtractor,cl as ViTForImageClassification,pl as ViTImageProcessor,hl as ViTModel,ml as ViTPreTrainedModel,fl as VisionEncoderDecoderModel,gl as VitMatteForImageMatting,_l as VitMatteImageProcessor,wl as VitMattePreTrainedModel,yl as VitsModel,bl as VitsModelOutput,vl as VitsPreTrainedModel,xl as VitsTokenizer,Ml as Wav2Vec2BertForCTC,Tl as Wav2Vec2BertForSequenceClassification,kl as Wav2Vec2BertModel,$l as Wav2Vec2BertPreTrainedModel,Sl as Wav2Vec2CTCTokenizer,Cl as Wav2Vec2FeatureExtractor,El as Wav2Vec2ForAudioFrameClassification,Pl as Wav2Vec2ForCTC,Al as Wav2Vec2ForSequenceClassification,Fl as Wav2Vec2Model,zl as Wav2Vec2PreTrainedModel,Il as Wav2Vec2ProcessorWithLM,Ol as WavLMForAudioFrameClassification,Bl as WavLMForCTC,Ll as WavLMForSequenceClassification,Dl as WavLMForXVector,Rl as WavLMModel,Nl as WavLMPreTrainedModel,Vl as WeSpeakerFeatureExtractor,jl as WeSpeakerResNetModel,Gl as WeSpeakerResNetPreTrainedModel,ql as WhisperFeatureExtractor,Ul as WhisperForConditionalGeneration,Wl as WhisperModel,Hl as WhisperPreTrainedModel,Xl as WhisperProcessor,Kl as WhisperTextStreamer,Ql as WhisperTokenizer,Yl as XLMForQuestionAnswering,Jl as XLMForSequenceClassification,Zl as XLMForTokenClassification,eu as XLMModel,tu as XLMPreTrainedModel,nu as XLMRobertaForMaskedLM,ru as XLMRobertaForQuestionAnswering,iu as XLMRobertaForSequenceClassification,au as XLMRobertaForTokenClassification,su as XLMRobertaModel,ou as XLMRobertaPreTrainedModel,lu as XLMRobertaTokenizer,uu as XLMTokenizer,du as XLMWithLMHeadModel,cu as XVectorOutput,pu as YolosFeatureExtractor,hu as YolosForObjectDetection,mu as YolosModel,fu as YolosObjectDetectionOutput,gu as YolosPreTrainedModel,_u as ZeroShotAudioClassificationPipeline,wu as ZeroShotClassificationPipeline,yu as ZeroShotImageClassificationPipeline,bu as ZeroShotObjectDetectionPipeline,vu as bankers_round,xu as cat,Mu as cos_sim,Tu as dot,ku as dynamic_time_warping,$u as env,Su as full,Cu as full_like,Eu as getKeyValueShapes,Pu as hamming,Au as hanning,Fu as interpolate,zu as interpolate_4d,Iu as interpolate_data,Ou as is_chinese_char,Bu as layer_norm,Lu as log_softmax,Du as magnitude,Ru as matmul,Nu as max,Vu as mean,ju as mean_pooling,Gu as medianFilter,qu as mel_filter_bank,Uu as min,Wu as ones,Hu as ones_like,Xu as permute,Ku as permute_data,Qu as pipeline,Yu as quantize_embeddings,Ju as read_audio,Zu as rfft,ed as round,td as softmax,nd as spectrogram,rd as stack,id as std_mean,ad as topk,sd as window_function,od as zeros,ld as zeros_like};
 //# sourceMappingURL=transformers.min.js.map