@huggingface/transformers 3.0.0-alpha.20 → 3.0.0-alpha.21

package/dist/transformers.min.mjs

@@ -88,10 +88,10 @@ import*as e from"fs";import*as t from"onnxruntime-node";import*as r from"path";i
   \******************************/(e,t,r)=>{var n;r.r(t),r.d(t,{Tensor:()=>i.Tensor,createInferenceSession:()=>f,deviceToExecutionProviders:()=>m,isONNXProxy:()=>w,isONNXTensor:()=>g});var o=r(/*! ../env.js */"./src/env.js"),s=r(/*! onnxruntime-node */"onnxruntime-node"),a=r(/*! #onnxruntime-webgpu */"?cb4d"),i=r(/*! onnxruntime-common */"./node_modules/onnxruntime-common/dist/esm/index.js");const l=Object.freeze({auto:null,gpu:null,cpu:"cpu",wasm:"wasm",webgpu:"webgpu",cuda:"cuda",dml:"dml",webnn:{name:"webnn",deviceType:"cpu"},"webnn-npu":{name:"webnn",deviceType:"npu"},"webnn-gpu":{name:"webnn",deviceType:"gpu"},"webnn-cpu":{name:"webnn",deviceType:"cpu"}}),c=[];let d,u;const p=Symbol.for("onnxruntime");if(p in globalThis)u=globalThis[p];else if(o.apis.IS_NODE_ENV){switch(u=s.default??s,process.platform){case"win32":c.push("dml");break;case"linux":"x64"===process.arch&&c.push("cuda")}c.push("cpu"),d=["cpu"]}else u=n||(n=r.t(a,2)),o.apis.IS_WEBNN_AVAILABLE&&c.push("webnn-npu","webnn-gpu","webnn-cpu","webnn"),o.apis.IS_WEBGPU_AVAILABLE&&c.push("webgpu"),c.push("wasm"),d=["wasm"];const h=u.InferenceSession;function m(e=null){if(!e)return d;switch(e){case"auto":return c;case"gpu":return c.filter((e=>["webgpu","cuda","dml","webnn-gpu"].includes(e)))}if(c.includes(e))return[l[e]??e];throw new Error(`Unsupported device: "${e}". Should be one of: ${c.join(", ")}.`)}let _=null;async function f(e,t){_&&await _;const r=h.create(e,t);return _??=r,await r}function g(e){return e instanceof u.Tensor}const M=u?.env;function w(){return M?.wasm?.proxy}M?.wasm&&(M.wasm.wasmPaths=`https://cdn.jsdelivr.net/npm/@huggingface/transformers@${o.env.version}/dist/`,M.wasm.proxy=!1,"undefined"!=typeof crossOriginIsolated&&crossOriginIsolated||(M.wasm.numThreads=1)),M?.webgpu&&(M.webgpu.powerPreference="high-performance"),o.env.backends.onnx=M},"./src/configs.js":
 /*!************************!*\
   !*** ./src/configs.js ***!
-  \************************/(e,t,r)=>{r.r(t),r.d(t,{AutoConfig:()=>l,PretrainedConfig:()=>i,getKeyValueShapes:()=>a});var n=r(/*! ./utils/core.js */"./src/utils/core.js"),o=r(/*! ./utils/hub.js */"./src/utils/hub.js");function s(e){const t={};let r={};switch(e.model_type){case"llava":case"paligemma":case"florence2":r=s(e.text_config);break;case"moondream1":r=s(e.phi_config);break;case"musicgen":r=s(e.decoder);break;case"gpt2":case"gptj":case"jais":case"codegen":case"gpt_bigcode":t.num_heads="n_head",t.num_layers="n_layer",t.hidden_size="n_embd";break;case"gpt_neox":case"stablelm":case"opt":case"phi":case"phi3":case"falcon":t.num_heads="num_attention_heads",t.num_layers="num_hidden_layers",t.hidden_size="hidden_size";break;case"llama":case"cohere":case"mistral":case"starcoder2":case"qwen2":t.num_heads="num_key_value_heads",t.num_layers="num_hidden_layers",t.hidden_size="hidden_size",t.num_attention_heads="num_attention_heads";break;case"gemma":case"gemma2":t.num_heads="num_key_value_heads",t.num_layers="num_hidden_layers",t.dim_kv="head_dim";break;case"openelm":t.num_heads="num_kv_heads",t.num_layers="num_transformer_layers",t.dim_kv="head_dim";break;case"gpt_neo":case"donut-swin":t.num_heads="num_heads",t.num_layers="num_layers",t.hidden_size="hidden_size";break;case"bloom":t.num_heads="n_head",t.num_layers="n_layer",t.hidden_size="hidden_size";break;case"mpt":t.num_heads="n_heads",t.num_layers="n_layers",t.hidden_size="d_model";break;case"t5":case"mt5":case"longt5":t.num_decoder_layers="num_decoder_layers",t.num_decoder_heads="num_heads",t.decoder_dim_kv="d_kv",t.num_encoder_layers="num_layers",t.num_encoder_heads="num_heads",t.encoder_dim_kv="d_kv";break;case"bart":case"mbart":case"marian":case"whisper":case"m2m_100":case"blenderbot":case"blenderbot-small":case"florence2_language":t.num_decoder_layers="decoder_layers",t.num_decoder_heads="decoder_attention_heads",t.decoder_hidden_size="d_model",t.num_encoder_layers="encoder_layers",t.num_encoder_heads="encoder_attention_heads",t.encoder_hidden_size="d_model";break;case"speecht5":t.num_decoder_layers="decoder_layers",t.num_decoder_heads="decoder_attention_heads",t.decoder_hidden_size="hidden_size",t.num_encoder_layers="encoder_layers",t.num_encoder_heads="encoder_attention_heads",t.encoder_hidden_size="hidden_size";break;case"trocr":t.num_encoder_layers=t.num_decoder_layers="decoder_layers",t.num_encoder_heads=t.num_decoder_heads="decoder_attention_heads",t.encoder_hidden_size=t.decoder_hidden_size="d_model";break;case"musicgen_decoder":t.num_encoder_layers=t.num_decoder_layers="num_hidden_layers",t.num_encoder_heads=t.num_decoder_heads="num_attention_heads",t.encoder_hidden_size=t.decoder_hidden_size="hidden_size";break;case"vision-encoder-decoder":const o=s(e.decoder),a="num_decoder_layers"in o,i=(0,n.pick)(e,["model_type","is_encoder_decoder"]);return a?(i.num_decoder_layers=o.num_decoder_layers,i.num_decoder_heads=o.num_decoder_heads,i.decoder_hidden_size=o.decoder_hidden_size,i.num_encoder_layers=o.num_encoder_layers,i.num_encoder_heads=o.num_encoder_heads,i.encoder_hidden_size=o.encoder_hidden_size):(i.num_layers=o.num_layers,i.num_heads=o.num_heads,i.hidden_size=o.hidden_size),i}const o={...r,...(0,n.pick)(e,["model_type","multi_query","is_encoder_decoder"])};for(const r in t)o[r]=e[t[r]];return o}function a(e,{prefix:t="past_key_values"}={}){const r={},n=e.normalized_config;if(n.is_encoder_decoder&&"num_encoder_heads"in n&&"num_decoder_heads"in n){const e=n.encoder_dim_kv??n.encoder_hidden_size/n.num_encoder_heads,o=n.decoder_dim_kv??n.decoder_hidden_size/n.num_decoder_heads,s=[1,n.num_encoder_heads,0,e],a=[1,n.num_decoder_heads,0,o];for(let e=0;e<n.num_decoder_layers;++e)r[`${t}.${e}.encoder.key`]=s,r[`${t}.${e}.encoder.value`]=s,r[`${t}.${e}.decoder.key`]=a,r[`${t}.${e}.decoder.value`]=a}else{const e=n.num_heads,o=n.num_layers,s=n.dim_kv??n.hidden_size/(n.num_attention_heads??e);if("falcon"===n.model_type){const n=[1*e,0,s];for(let e=0;e<o;++e)r[`${t}.${e}.key`]=n,r[`${t}.${e}.value`]=n}else if(n.multi_query){const n=[1*e,0,2*s];for(let e=0;e<o;++e)r[`${t}.${e}.key_value`]=n}else if("bloom"===n.model_type){const n=[1*e,s,0],a=[1*e,0,s];for(let e=0;e<o;++e)r[`${t}.${e}.key`]=n,r[`${t}.${e}.value`]=a}else if("openelm"===n.model_type)for(let n=0;n<o;++n){const o=[1,e[n],0,s];r[`${t}.${n}.key`]=o,r[`${t}.${n}.value`]=o}else{const n=[1,e,0,s];for(let e=0;e<o;++e)r[`${t}.${e}.key`]=n,r[`${t}.${e}.value`]=n}}return r}class i{model_type=null;is_encoder_decoder=!1;max_position_embeddings;"transformers.js_config";constructor(e){Object.assign(this,e),this.normalized_config=s(this)}static async from_pretrained(e,{progress_callback:t=null,config:r=null,cache_dir:n=null,local_files_only:s=!1,revision:a="main"}={}){!r||r instanceof i||(r=new i(r));const l=r??await async function(e,t){return await(0,o.getModelJSON)(e,"config.json",!0,t)}(e,{progress_callback:t,config:r,cache_dir:n,local_files_only:s,revision:a});return new this(l)}}class l{static async from_pretrained(...e){return i.from_pretrained(...e)}}},"./src/env.js":
+  \************************/(e,t,r)=>{r.r(t),r.d(t,{AutoConfig:()=>l,PretrainedConfig:()=>i,getKeyValueShapes:()=>a});var n=r(/*! ./utils/core.js */"./src/utils/core.js"),o=r(/*! ./utils/hub.js */"./src/utils/hub.js");function s(e){const t={};let r={};switch(e.model_type){case"llava":case"paligemma":case"florence2":r=s(e.text_config);break;case"moondream1":r=s(e.phi_config);break;case"musicgen":r=s(e.decoder);break;case"gpt2":case"gptj":case"jais":case"codegen":case"gpt_bigcode":t.num_heads="n_head",t.num_layers="n_layer",t.hidden_size="n_embd";break;case"gpt_neox":case"stablelm":case"opt":case"phi":case"phi3":case"falcon":t.num_heads="num_attention_heads",t.num_layers="num_hidden_layers",t.hidden_size="hidden_size";break;case"llama":case"granite":case"cohere":case"mistral":case"starcoder2":case"qwen2":t.num_heads="num_key_value_heads",t.num_layers="num_hidden_layers",t.hidden_size="hidden_size",t.num_attention_heads="num_attention_heads";break;case"gemma":case"gemma2":t.num_heads="num_key_value_heads",t.num_layers="num_hidden_layers",t.dim_kv="head_dim";break;case"openelm":t.num_heads="num_kv_heads",t.num_layers="num_transformer_layers",t.dim_kv="head_dim";break;case"gpt_neo":case"donut-swin":t.num_heads="num_heads",t.num_layers="num_layers",t.hidden_size="hidden_size";break;case"bloom":t.num_heads="n_head",t.num_layers="n_layer",t.hidden_size="hidden_size";break;case"mpt":t.num_heads="n_heads",t.num_layers="n_layers",t.hidden_size="d_model";break;case"t5":case"mt5":case"longt5":t.num_decoder_layers="num_decoder_layers",t.num_decoder_heads="num_heads",t.decoder_dim_kv="d_kv",t.num_encoder_layers="num_layers",t.num_encoder_heads="num_heads",t.encoder_dim_kv="d_kv";break;case"bart":case"mbart":case"marian":case"whisper":case"m2m_100":case"blenderbot":case"blenderbot-small":case"florence2_language":t.num_decoder_layers="decoder_layers",t.num_decoder_heads="decoder_attention_heads",t.decoder_hidden_size="d_model",t.num_encoder_layers="encoder_layers",t.num_encoder_heads="encoder_attention_heads",t.encoder_hidden_size="d_model";break;case"speecht5":t.num_decoder_layers="decoder_layers",t.num_decoder_heads="decoder_attention_heads",t.decoder_hidden_size="hidden_size",t.num_encoder_layers="encoder_layers",t.num_encoder_heads="encoder_attention_heads",t.encoder_hidden_size="hidden_size";break;case"trocr":t.num_encoder_layers=t.num_decoder_layers="decoder_layers",t.num_encoder_heads=t.num_decoder_heads="decoder_attention_heads",t.encoder_hidden_size=t.decoder_hidden_size="d_model";break;case"musicgen_decoder":t.num_encoder_layers=t.num_decoder_layers="num_hidden_layers",t.num_encoder_heads=t.num_decoder_heads="num_attention_heads",t.encoder_hidden_size=t.decoder_hidden_size="hidden_size";break;case"vision-encoder-decoder":const o=s(e.decoder),a="num_decoder_layers"in o,i=(0,n.pick)(e,["model_type","is_encoder_decoder"]);return a?(i.num_decoder_layers=o.num_decoder_layers,i.num_decoder_heads=o.num_decoder_heads,i.decoder_hidden_size=o.decoder_hidden_size,i.num_encoder_layers=o.num_encoder_layers,i.num_encoder_heads=o.num_encoder_heads,i.encoder_hidden_size=o.encoder_hidden_size):(i.num_layers=o.num_layers,i.num_heads=o.num_heads,i.hidden_size=o.hidden_size),i}const o={...r,...(0,n.pick)(e,["model_type","multi_query","is_encoder_decoder"])};for(const r in t)o[r]=e[t[r]];return o}function a(e,{prefix:t="past_key_values"}={}){const r={},n=e.normalized_config;if(n.is_encoder_decoder&&"num_encoder_heads"in n&&"num_decoder_heads"in n){const e=n.encoder_dim_kv??n.encoder_hidden_size/n.num_encoder_heads,o=n.decoder_dim_kv??n.decoder_hidden_size/n.num_decoder_heads,s=[1,n.num_encoder_heads,0,e],a=[1,n.num_decoder_heads,0,o];for(let e=0;e<n.num_decoder_layers;++e)r[`${t}.${e}.encoder.key`]=s,r[`${t}.${e}.encoder.value`]=s,r[`${t}.${e}.decoder.key`]=a,r[`${t}.${e}.decoder.value`]=a}else{const e=n.num_heads,o=n.num_layers,s=n.dim_kv??n.hidden_size/(n.num_attention_heads??e);if("falcon"===n.model_type){const n=[1*e,0,s];for(let e=0;e<o;++e)r[`${t}.${e}.key`]=n,r[`${t}.${e}.value`]=n}else if(n.multi_query){const n=[1*e,0,2*s];for(let e=0;e<o;++e)r[`${t}.${e}.key_value`]=n}else if("bloom"===n.model_type){const n=[1*e,s,0],a=[1*e,0,s];for(let e=0;e<o;++e)r[`${t}.${e}.key`]=n,r[`${t}.${e}.value`]=a}else if("openelm"===n.model_type)for(let n=0;n<o;++n){const o=[1,e[n],0,s];r[`${t}.${n}.key`]=o,r[`${t}.${n}.value`]=o}else{const n=[1,e,0,s];for(let e=0;e<o;++e)r[`${t}.${e}.key`]=n,r[`${t}.${e}.value`]=n}}return r}class i{model_type=null;is_encoder_decoder=!1;max_position_embeddings;"transformers.js_config";constructor(e){Object.assign(this,e),this.normalized_config=s(this)}static async from_pretrained(e,{progress_callback:t=null,config:r=null,cache_dir:n=null,local_files_only:s=!1,revision:a="main"}={}){!r||r instanceof i||(r=new i(r));const l=r??await async function(e,t){return await(0,o.getModelJSON)(e,"config.json",!0,t)}(e,{progress_callback:t,config:r,cache_dir:n,local_files_only:s,revision:a});return new this(l)}}class l{static async from_pretrained(...e){return i.from_pretrained(...e)}}},"./src/env.js":
 /*!********************!*\
   !*** ./src/env.js ***!
-  \********************/(e,t,r)=>{r.r(t),r.d(t,{apis:()=>_,env:()=>T});var n=r(/*! fs */"fs"),o=r(/*! path */"path"),s=r(/*! url */"url");const a="undefined"!=typeof self,i=a&&"DedicatedWorkerGlobalScope"===self.constructor.name,l=a&&"caches"in self,c="undefined"!=typeof navigator&&"gpu"in navigator,d="undefined"!=typeof navigator&&"ml"in navigator,u="undefined"!=typeof process,p=u&&"node"===process?.release?.name,h=!b(n.default),m=!b(o.default),_=Object.freeze({IS_BROWSER_ENV:a,IS_WEBWORKER_ENV:i,IS_WEB_CACHE_AVAILABLE:l,IS_WEBGPU_AVAILABLE:c,IS_WEBNN_AVAILABLE:d,IS_PROCESS_AVAILABLE:u,IS_NODE_ENV:p,IS_FS_AVAILABLE:h,IS_PATH_AVAILABLE:m}),f=h&&m,g=f?o.default.dirname(o.default.dirname(s.default.fileURLToPath(import.meta.url))):"./",M=f?o.default.join(g,"/.cache/"):null,w="/models/",T={version:"3.0.0-alpha.20",backends:{onnx:{}},allowRemoteModels:!0,remoteHost:"https://huggingface.co/",remotePathTemplate:"{model}/resolve/{revision}/",allowLocalModels:!a,localModelPath:f?o.default.join(g,w):w,useFS:h,useBrowserCache:l,useFSCache:h,cacheDir:M,useCustomCache:!1,customCache:null};function b(e){return 0===Object.keys(e).length}},"./src/generation/configuration_utils.js":
+  \********************/(e,t,r)=>{r.r(t),r.d(t,{apis:()=>_,env:()=>T});var n=r(/*! fs */"fs"),o=r(/*! path */"path"),s=r(/*! url */"url");const a="undefined"!=typeof self,i=a&&"DedicatedWorkerGlobalScope"===self.constructor.name,l=a&&"caches"in self,c="undefined"!=typeof navigator&&"gpu"in navigator,d="undefined"!=typeof navigator&&"ml"in navigator,u="undefined"!=typeof process,p=u&&"node"===process?.release?.name,h=!b(n.default),m=!b(o.default),_=Object.freeze({IS_BROWSER_ENV:a,IS_WEBWORKER_ENV:i,IS_WEB_CACHE_AVAILABLE:l,IS_WEBGPU_AVAILABLE:c,IS_WEBNN_AVAILABLE:d,IS_PROCESS_AVAILABLE:u,IS_NODE_ENV:p,IS_FS_AVAILABLE:h,IS_PATH_AVAILABLE:m}),f=h&&m,g=f?o.default.dirname(o.default.dirname(s.default.fileURLToPath(import.meta.url))):"./",M=f?o.default.join(g,"/.cache/"):null,w="/models/",T={version:"3.0.0-alpha.21",backends:{onnx:{}},allowRemoteModels:!0,remoteHost:"https://huggingface.co/",remotePathTemplate:"{model}/resolve/{revision}/",allowLocalModels:!a,localModelPath:f?o.default.join(g,w):w,useFS:h,useBrowserCache:l,useFSCache:h,cacheDir:M,useCustomCache:!1,customCache:null};function b(e){return 0===Object.keys(e).length}},"./src/generation/configuration_utils.js":
 /*!***********************************************!*\
   !*** ./src/generation/configuration_utils.js ***!
   \***********************************************/(e,t,r)=>{r.r(t),r.d(t,{GenerationConfig:()=>o});var n=r(/*! ../utils/core.js */"./src/utils/core.js");class o{max_length=20;max_new_tokens=null;min_length=0;min_new_tokens=null;early_stopping=!1;max_time=null;do_sample=!1;num_beams=1;num_beam_groups=1;penalty_alpha=null;use_cache=!0;temperature=1;top_k=50;top_p=1;typical_p=1;epsilon_cutoff=0;eta_cutoff=0;diversity_penalty=0;repetition_penalty=1;encoder_repetition_penalty=1;length_penalty=1;no_repeat_ngram_size=0;bad_words_ids=null;force_words_ids=null;renormalize_logits=!1;constraints=null;forced_bos_token_id=null;forced_eos_token_id=null;remove_invalid_values=!1;exponential_decay_length_penalty=null;suppress_tokens=null;begin_suppress_tokens=null;forced_decoder_ids=null;guidance_scale=null;num_return_sequences=1;output_attentions=!1;output_hidden_states=!1;output_scores=!1;return_dict_in_generate=!1;pad_token_id=null;bos_token_id=null;eos_token_id=null;encoder_no_repeat_ngram_size=0;decoder_start_token_id=null;generation_kwargs={};constructor(e){Object.assign(this,(0,n.pick)(e,Object.getOwnPropertyNames(this)))}}},"./src/generation/logits_process.js":
@@ -109,7 +109,7 @@ import*as e from"fs";import*as t from"onnxruntime-node";import*as r from"path";i
   \*************************************/(e,t,r)=>{r.r(t),r.d(t,{BaseStreamer:()=>a,TextStreamer:()=>l,WhisperTextStreamer:()=>c});var n=r(/*! ../utils/core.js */"./src/utils/core.js"),o=r(/*! ../tokenizers.js */"./src/tokenizers.js"),s=r(/*! ../env.js */"./src/env.js");class a{put(e){throw Error("Not implemented")}end(){throw Error("Not implemented")}}const i=s.apis.IS_PROCESS_AVAILABLE?e=>process.stdout.write(e):e=>console.log(e);class l extends a{constructor(e,{skip_prompt:t=!1,callback_function:r=null,token_callback_function:n=null,decode_kwargs:o={},...s}={}){super(),this.tokenizer=e,this.skip_prompt=t,this.callback_function=r??i,this.token_callback_function=n,this.decode_kwargs={...o,...s},this.token_cache=[],this.print_len=0,this.next_tokens_are_prompt=!0}put(e){if(e.length>1)throw Error("TextStreamer only supports batch size of 1");if(this.skip_prompt&&this.next_tokens_are_prompt)return void(this.next_tokens_are_prompt=!1);const t=e[0];this.token_callback_function?.(t),this.token_cache=(0,n.mergeArrays)(this.token_cache,t);const r=this.tokenizer.decode(this.token_cache,this.decode_kwargs);let s;r.endsWith("\n")?(s=r.slice(this.print_len),this.token_cache=[],this.print_len=0):r.length>0&&(0,o.is_chinese_char)(r.charCodeAt(r.length-1))?(s=r.slice(this.print_len),this.print_len+=s.length):(s=r.slice(this.print_len,r.lastIndexOf(" ")+1),this.print_len+=s.length),this.on_finalized_text(s,!1)}end(){let e;if(this.token_cache.length>0){e=this.tokenizer.decode(this.token_cache,this.decode_kwargs).slice(this.print_len),this.token_cache=[],this.print_len=0}else e="";this.next_tokens_are_prompt=!0,this.on_finalized_text(e,!0)}on_finalized_text(e,t){e.length>0&&this.callback_function?.(e),t&&this.callback_function===i&&s.apis.IS_PROCESS_AVAILABLE&&this.callback_function?.("\n")}}class c extends l{constructor(e,{skip_prompt:t=!1,callback_function:r=null,token_callback_function:n=null,on_chunk_start:o=null,on_chunk_end:s=null,on_finalize:a=null,time_precision:i=.02,skip_special_tokens:l=!0,decode_kwargs:c={}}={}){super(e,{skip_prompt:t,callback_function:r,token_callback_function:n,decode_kwargs:{skip_special_tokens:l,...c}}),this.timestamp_begin=e.timestamp_begin,this.on_chunk_start=o,this.on_chunk_end=s,this.on_finalize=a,this.time_precision=i,this.waiting_for_timestamp=!1}put(e){if(e.length>1)throw Error("WhisperTextStreamer only supports batch size of 1");const t=e[0];if(1===t.length){const r=Number(t[0])-this.timestamp_begin;if(r>=0){const t=r*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,r)=>{r.r(t),r.d(t,{ASTForAudioClassification:()=>rr,ASTModel:()=>tr,ASTPreTrainedModel:()=>er,AlbertForMaskedLM:()=>ut,AlbertForQuestionAnswering:()=>dt,AlbertForSequenceClassification:()=>ct,AlbertModel:()=>lt,AlbertPreTrainedModel:()=>it,AutoModel:()=>zi,AutoModelForAudioClassification:()=>Ki,AutoModelForAudioFrameClassification:()=>el,AutoModelForCTC:()=>Ji,AutoModelForCausalLM:()=>ji,AutoModelForDepthEstimation:()=>ol,AutoModelForDocumentQuestionAnswering:()=>tl,AutoModelForImageClassification:()=>Wi,AutoModelForImageFeatureExtraction:()=>al,AutoModelForImageMatting:()=>rl,AutoModelForImageSegmentation:()=>$i,AutoModelForImageToImage:()=>nl,AutoModelForMaskGeneration:()=>Yi,AutoModelForMaskedLM:()=>Ri,AutoModelForNormalEstimation:()=>sl,AutoModelForObjectDetection:()=>Qi,AutoModelForQuestionAnswering:()=>Gi,AutoModelForSemanticSegmentation:()=>Ui,AutoModelForSeq2SeqLM:()=>Ni,AutoModelForSequenceClassification:()=>Ii,AutoModelForSpeechSeq2Seq:()=>Di,AutoModelForTextToSpectrogram:()=>Oi,AutoModelForTextToWaveform:()=>Vi,AutoModelForTokenClassification:()=>Bi,AutoModelForUniversalSegmentation:()=>Xi,AutoModelForVision2Seq:()=>qi,AutoModelForXVector:()=>Zi,AutoModelForZeroShotObjectDetection:()=>Hi,BartForConditionalGeneration:()=>yt,BartForSequenceClassification:()=>kt,BartModel:()=>xt,BartPretrainedModel:()=>bt,BaseModelOutput:()=>q,BeitForImageClassification:()=>ao,BeitModel:()=>so,BeitPreTrainedModel:()=>oo,BertForMaskedLM:()=>U,BertForQuestionAnswering:()=>H,BertForSequenceClassification:()=>X,BertForTokenClassification:()=>Q,BertModel:()=>$,BertPreTrainedModel:()=>W,BlenderbotForConditionalGeneration:()=>Lt,BlenderbotModel:()=>Et,BlenderbotPreTrainedModel:()=>At,BlenderbotSmallForConditionalGeneration:()=>Bt,BlenderbotSmallModel:()=>It,BlenderbotSmallPreTrainedModel:()=>zt,BloomForCausalLM:()=>bn,BloomModel:()=>Tn,BloomPreTrainedModel:()=>wn,CLIPModel:()=>hr,CLIPPreTrainedModel:()=>pr,CLIPSegForImageSegmentation:()=>Pr,CLIPSegModel:()=>Fr,CLIPSegPreTrainedModel:()=>kr,CLIPTextModel:()=>mr,CLIPTextModelWithProjection:()=>_r,CLIPVisionModel:()=>fr,CLIPVisionModelWithProjection:()=>gr,CamembertForMaskedLM:()=>Me,CamembertForQuestionAnswering:()=>be,CamembertForSequenceClassification:()=>we,CamembertForTokenClassification:()=>Te,CamembertModel:()=>ge,CamembertPreTrainedModel:()=>fe,CausalLMOutput:()=>hl,CausalLMOutputWithPast:()=>ml,ChineseCLIPModel:()=>yr,ChineseCLIPPreTrainedModel:()=>xr,ClapAudioModelWithProjection:()=>xa,ClapModel:()=>Ta,ClapPreTrainedModel:()=>wa,ClapTextModelWithProjection:()=>ba,CodeGenForCausalLM:()=>Xr,CodeGenModel:()=>Ur,CodeGenPreTrainedModel:()=>$r,CohereForCausalLM:()=>Zr,CohereModel:()=>Kr,CoherePreTrainedModel:()=>Jr,ConvBertForMaskedLM:()=>ae,ConvBertForQuestionAnswering:()=>ce,ConvBertForSequenceClassification:()=>ie,ConvBertForTokenClassification:()=>le,ConvBertModel:()=>se,ConvBertPreTrainedModel:()=>oe,ConvNextForImageClassification:()=>os,ConvNextModel:()=>ns,ConvNextPreTrainedModel:()=>rs,ConvNextV2ForImageClassification:()=>is,ConvNextV2Model:()=>as,ConvNextV2PreTrainedModel:()=>ss,DPTForDepthEstimation:()=>Vo,DPTModel:()=>Oo,DPTPreTrainedModel:()=>Do,DebertaForMaskedLM:()=>ke,DebertaForQuestionAnswering:()=>Ce,DebertaForSequenceClassification:()=>Fe,DebertaForTokenClassification:()=>Pe,DebertaModel:()=>ye,DebertaPreTrainedModel:()=>xe,DebertaV2ForMaskedLM:()=>Ae,DebertaV2ForQuestionAnswering:()=>ze,DebertaV2ForSequenceClassification:()=>Ee,DebertaV2ForTokenClassification:()=>Le,DebertaV2Model:()=>Se,DebertaV2PreTrainedModel:()=>ve,DecisionTransformerModel:()=>Ka,DecisionTransformerPreTrainedModel:()=>Ja,DeiTForImageClassification:()=>ko,DeiTModel:()=>yo,DeiTPreTrainedModel:()=>xo,DepthAnythingForDepthEstimation:()=>Ro,DepthAnythingPreTrainedModel:()=>jo,DepthProForDepthEstimation:()=>Xo,DepthProPreTrainedModel:()=>Uo,DetrForObjectDetection:()=>co,DetrForSegmentation:()=>uo,DetrModel:()=>lo,DetrObjectDetectionOutput:()=>po,DetrPreTrainedModel:()=>io,DetrSegmentationOutput:()=>ho,Dinov2ForImageClassification:()=>ds,Dinov2Model:()=>cs,Dinov2PreTrainedModel:()=>ls,DistilBertForMaskedLM:()=>Ve,DistilBertForQuestionAnswering:()=>Oe,DistilBertForSequenceClassification:()=>Ne,DistilBertForTokenClassification:()=>De,DistilBertModel:()=>Be,DistilBertPreTrainedModel:()=>Ie,DonutSwinModel:()=>ts,DonutSwinPreTrainedModel:()=>es,EfficientNetForImageClassification:()=>Ia,EfficientNetModel:()=>za,EfficientNetPreTrainedModel:()=>La,ElectraForMaskedLM:()=>pe,ElectraForQuestionAnswering:()=>_e,ElectraForSequenceClassification:()=>he,ElectraForTokenClassification:()=>me,ElectraModel:()=>ue,ElectraPreTrainedModel:()=>de,EsmForMaskedLM:()=>Ge,EsmForSequenceClassification:()=>qe,EsmForTokenClassification:()=>We,EsmModel:()=>Re,EsmPreTrainedModel:()=>je,FalconForCausalLM:()=>Ma,FalconModel:()=>ga,FalconPreTrainedModel:()=>fa,FastViTForImageClassification:()=>qn,FastViTModel:()=>Gn,FastViTPreTrainedModel:()=>Rn,Florence2ForConditionalGeneration:()=>ur,Florence2PreTrainedModel:()=>dr,GLPNForDepthEstimation:()=>Zo,GLPNModel:()=>Ko,GLPNPreTrainedModel:()=>Jo,GPT2LMHeadModel:()=>Sr,GPT2Model:()=>vr,GPT2PreTrainedModel:()=>Cr,GPTBigCodeForCausalLM:()=>Wr,GPTBigCodeModel:()=>qr,GPTBigCodePreTrainedModel:()=>Gr,GPTJForCausalLM:()=>Rr,GPTJModel:()=>jr,GPTJPreTrainedModel:()=>Vr,GPTNeoForCausalLM:()=>Br,GPTNeoModel:()=>Ir,GPTNeoPreTrainedModel:()=>zr,GPTNeoXForCausalLM:()=>Or,GPTNeoXModel:()=>Dr,GPTNeoXPreTrainedModel:()=>Nr,Gemma2ForCausalLM:()=>sn,Gemma2Model:()=>on,Gemma2PreTrainedModel:()=>nn,GemmaForCausalLM:()=>rn,GemmaModel:()=>tn,GemmaPreTrainedModel:()=>en,GroupViTModel:()=>jn,GroupViTPreTrainedModel:()=>Vn,HieraForImageClassification:()=>Co,HieraModel:()=>Po,HieraPreTrainedModel:()=>Fo,HubertForCTC:()=>Hs,HubertForSequenceClassification:()=>Ys,HubertModel:()=>Qs,HubertPreTrainedModel:()=>Xs,ImageMattingOutput:()=>_l,JAISLMHeadModel:()=>Lr,JAISModel:()=>Er,JAISPreTrainedModel:()=>Ar,LlamaForCausalLM:()=>Yr,LlamaModel:()=>Hr,LlamaPreTrainedModel:()=>Qr,LlavaForConditionalGeneration:()=>lr,LlavaPreTrainedModel:()=>ir,LongT5ForConditionalGeneration:()=>gt,LongT5Model:()=>ft,LongT5PreTrainedModel:()=>_t,M2M100ForConditionalGeneration:()=>ys,M2M100Model:()=>xs,M2M100PreTrainedModel:()=>bs,MBartForCausalLM:()=>St,MBartForConditionalGeneration:()=>Ct,MBartForSequenceClassification:()=>vt,MBartModel:()=>Pt,MBartPreTrainedModel:()=>Ft,MPNetForMaskedLM:()=>Ke,MPNetForQuestionAnswering:()=>tt,MPNetForSequenceClassification:()=>Ze,MPNetForTokenClassification:()=>et,MPNetModel:()=>Je,MPNetPreTrainedModel:()=>Ye,MT5ForConditionalGeneration:()=>Tt,MT5Model:()=>wt,MT5PreTrainedModel:()=>Mt,MarianMTModel:()=>Ts,MarianModel:()=>ws,MarianPreTrainedModel:()=>Ms,MaskFormerForInstanceSegmentation:()=>Yo,MaskFormerModel:()=>Ho,MaskFormerPreTrainedModel:()=>Qo,MaskedLMOutput:()=>ul,MistralForCausalLM:()=>pa,MistralModel:()=>ua,MistralPreTrainedModel:()=>da,MobileBertForMaskedLM:()=>Xe,MobileBertForQuestionAnswering:()=>He,MobileBertForSequenceClassification:()=>Qe,MobileBertModel:()=>Ue,MobileBertPreTrainedModel:()=>$e,MobileNetV1ForImageClassification:()=>Ra,MobileNetV1Model:()=>ja,MobileNetV1PreTrainedModel:()=>Va,MobileNetV2ForImageClassification:()=>Wa,MobileNetV2Model:()=>qa,MobileNetV2PreTrainedModel:()=>Ga,MobileNetV3ForImageClassification:()=>Xa,MobileNetV3Model:()=>Ua,MobileNetV3PreTrainedModel:()=>$a,MobileNetV4ForImageClassification:()=>Ya,MobileNetV4Model:()=>Ha,MobileNetV4PreTrainedModel:()=>Qa,MobileViTForImageClassification:()=>Qn,MobileViTModel:()=>Xn,MobileViTPreTrainedModel:()=>Un,MobileViTV2ForImageClassification:()=>Jn,MobileViTV2Model:()=>Yn,MobileViTV2PreTrainedModel:()=>Hn,ModelOutput:()=>G,Moondream1ForConditionalGeneration:()=>cr,MptForCausalLM:()=>kn,MptModel:()=>yn,MptPreTrainedModel:()=>xn,MusicgenForCausalLM:()=>Da,MusicgenForConditionalGeneration:()=>Oa,MusicgenModel:()=>Na,MusicgenPreTrainedModel:()=>Ba,NomicBertModel:()=>J,NomicBertPreTrainedModel:()=>Y,OPTForCausalLM:()=>Cn,OPTModel:()=>Pn,OPTPreTrainedModel:()=>Fn,OpenELMForCausalLM:()=>cn,OpenELMModel:()=>ln,OpenELMPreTrainedModel:()=>an,OwlViTForObjectDetection:()=>eo,OwlViTModel:()=>Zn,OwlViTPreTrainedModel:()=>Kn,Owlv2ForObjectDetection:()=>no,Owlv2Model:()=>ro,Owlv2PreTrainedModel:()=>to,Phi3ForCausalLM:()=>Mn,Phi3Model:()=>gn,Phi3PreTrainedModel:()=>fn,PhiForCausalLM:()=>_n,PhiModel:()=>mn,PhiPreTrainedModel:()=>hn,PreTrainedModel:()=>R,PretrainedMixin:()=>Za,PvtForImageClassification:()=>zn,PvtModel:()=>Ln,PvtPreTrainedModel:()=>En,PyAnnoteForAudioFrameClassification:()=>Es,PyAnnoteModel:()=>As,PyAnnotePreTrainedModel:()=>Ss,QuestionAnsweringModelOutput:()=>pl,Qwen2ForCausalLM:()=>pn,Qwen2Model:()=>un,Qwen2PreTrainedModel:()=>dn,RTDetrForObjectDetection:()=>fo,RTDetrModel:()=>_o,RTDetrObjectDetectionOutput:()=>go,RTDetrPreTrainedModel:()=>mo,ResNetForImageClassification:()=>Ao,ResNetModel:()=>So,ResNetPreTrainedModel:()=>vo,RoFormerForMaskedLM:()=>ee,RoFormerForQuestionAnswering:()=>ne,RoFormerForSequenceClassification:()=>te,RoFormerForTokenClassification:()=>re,RoFormerModel:()=>Z,RoFormerPreTrainedModel:()=>K,RobertaForMaskedLM:()=>Ot,RobertaForQuestionAnswering:()=>Rt,RobertaForSequenceClassification:()=>Vt,RobertaForTokenClassification:()=>jt,RobertaModel:()=>Dt,RobertaPreTrainedModel:()=>Nt,SamImageSegmentationOutput:()=>gs,SamModel:()=>fs,SamPreTrainedModel:()=>_s,SapiensForDepthEstimation:()=>Wo,SapiensForNormalEstimation:()=>$o,SapiensForSemanticSegmentation:()=>qo,SapiensPreTrainedModel:()=>Go,SegformerForImageClassification:()=>Ca,SegformerForSemanticSegmentation:()=>va,SegformerModel:()=>Pa,SegformerPreTrainedModel:()=>Fa,Seq2SeqLMOutput:()=>il,SequenceClassifierOutput:()=>ll,SiglipModel:()=>wr,SiglipPreTrainedModel:()=>Mr,SiglipTextModel:()=>Tr,SiglipVisionModel:()=>br,SpeechT5ForSpeechToText:()=>sa,SpeechT5ForTextToSpeech:()=>aa,SpeechT5HifiGan:()=>ia,SpeechT5Model:()=>oa,SpeechT5PreTrainedModel:()=>na,SqueezeBertForMaskedLM:()=>ot,SqueezeBertForQuestionAnswering:()=>at,SqueezeBertForSequenceClassification:()=>st,SqueezeBertModel:()=>nt,SqueezeBertPreTrainedModel:()=>rt,StableLmForCausalLM:()=>Ea,StableLmModel:()=>Aa,StableLmPreTrainedModel:()=>Sa,Starcoder2ForCausalLM:()=>_a,Starcoder2Model:()=>ma,Starcoder2PreTrainedModel:()=>ha,Swin2SRForImageSuperResolution:()=>No,Swin2SRModel:()=>Bo,Swin2SRPreTrainedModel:()=>Io,SwinForImageClassification:()=>zo,SwinModel:()=>Lo,SwinPreTrainedModel:()=>Eo,T5ForConditionalGeneration:()=>mt,T5Model:()=>ht,T5PreTrainedModel:()=>pt,TableTransformerForObjectDetection:()=>To,TableTransformerModel:()=>wo,TableTransformerObjectDetectionOutput:()=>bo,TableTransformerPreTrainedModel:()=>Mo,TokenClassifierOutput:()=>dl,TrOCRForCausalLM:()=>ca,TrOCRPreTrainedModel:()=>la,UniSpeechForCTC:()=>Ns,UniSpeechForSequenceClassification:()=>Ds,UniSpeechModel:()=>Bs,UniSpeechPreTrainedModel:()=>Is,UniSpeechSatForAudioFrameClassification:()=>Gs,UniSpeechSatForCTC:()=>js,UniSpeechSatForSequenceClassification:()=>Rs,UniSpeechSatModel:()=>Vs,UniSpeechSatPreTrainedModel:()=>Os,ViTForImageClassification:()=>An,ViTMAEModel:()=>Bn,ViTMAEPreTrainedModel:()=>In,ViTMSNForImageClassification:()=>On,ViTMSNModel:()=>Dn,ViTMSNPreTrainedModel:()=>Nn,ViTModel:()=>Sn,ViTPreTrainedModel:()=>vn,VisionEncoderDecoderModel:()=>ar,VitMatteForImageMatting:()=>$n,VitMattePreTrainedModel:()=>Wn,VitsModel:()=>ka,VitsModelOutput:()=>fl,VitsPreTrainedModel:()=>ya,Wav2Vec2BertForCTC:()=>$s,Wav2Vec2BertForSequenceClassification:()=>Us,Wav2Vec2BertModel:()=>Ws,Wav2Vec2BertPreTrainedModel:()=>qs,Wav2Vec2ForAudioFrameClassification:()=>vs,Wav2Vec2ForCTC:()=>Ps,Wav2Vec2ForSequenceClassification:()=>Cs,Wav2Vec2Model:()=>Fs,Wav2Vec2PreTrainedModel:()=>ks,WavLMForAudioFrameClassification:()=>ra,WavLMForCTC:()=>Zs,WavLMForSequenceClassification:()=>ea,WavLMForXVector:()=>ta,WavLMModel:()=>Ks,WavLMPreTrainedModel:()=>Js,WeSpeakerResNetModel:()=>zs,WeSpeakerResNetPreTrainedModel:()=>Ls,WhisperForConditionalGeneration:()=>sr,WhisperModel:()=>or,WhisperPreTrainedModel:()=>nr,XLMForQuestionAnswering:()=>Xt,XLMForSequenceClassification:()=>$t,XLMForTokenClassification:()=>Ut,XLMModel:()=>qt,XLMPreTrainedModel:()=>Gt,XLMRobertaForMaskedLM:()=>Yt,XLMRobertaForQuestionAnswering:()=>Zt,XLMRobertaForSequenceClassification:()=>Jt,XLMRobertaForTokenClassification:()=>Kt,XLMRobertaModel:()=>Ht,XLMRobertaPreTrainedModel:()=>Qt,XLMWithLMHeadModel:()=>Wt,XVectorOutput:()=>cl,YolosForObjectDetection:()=>hs,YolosModel:()=>ps,YolosObjectDetectionOutput:()=>ms,YolosPreTrainedModel:()=>us});var n=r(/*! ./configs.js */"./src/configs.js"),o=r(/*! ./backends/onnx.js */"./src/backends/onnx.js"),s=r(/*! ./utils/dtypes.js */"./src/utils/dtypes.js"),a=r(/*! ./utils/generic.js */"./src/utils/generic.js"),i=r(/*! ./utils/core.js */"./src/utils/core.js"),l=r(/*! ./utils/hub.js */"./src/utils/hub.js"),c=r(/*! ./generation/logits_process.js */"./src/generation/logits_process.js"),d=r(/*! ./generation/configuration_utils.js */"./src/generation/configuration_utils.js"),u=r(/*! ./utils/tensor.js */"./src/utils/tensor.js"),p=r(/*! ./utils/maths.js */"./src/utils/maths.js"),h=r(/*! ./generation/stopping_criteria.js */"./src/generation/stopping_criteria.js"),m=r(/*! ./generation/logits_sampler.js */"./src/generation/logits_sampler.js"),_=r(/*! ./env.js */"./src/env.js"),f=r(/*! ./models/whisper/generation_whisper.js */"./src/models/whisper/generation_whisper.js"),g=r(/*! ./models/whisper/common_whisper.js */"./src/models/whisper/common_whisper.js");const M=0,w=1,T=2,b=3,x=4,y=5,k=6,F=7,P=new Map,C=new Map,v=new Map;async function S(e,t,r){return Object.fromEntries(await Promise.all(Object.keys(t).map((async a=>{const{buffer:i,session_options:c}=await async function(e,t,r){const a=r.config?.["transformers.js_config"]??{};let i=r.device??a.device;i&&"string"!=typeof i&&(i.hasOwnProperty(t)?i=i[t]:(console.warn(`device not specified for "${t}". Using the default device.`),i=null));const c=i??(_.apis.IS_NODE_ENV?"cpu":"wasm"),d=(0,o.deviceToExecutionProviders)(c);let u=r.dtype??a.dtype;"string"!=typeof u&&(u&&u.hasOwnProperty(t)?u=u[t]:(u=s.DEFAULT_DEVICE_DTYPE_MAPPING[c]??s.DATA_TYPES.fp32,console.warn(`dtype not specified for "${t}". Using the default dtype (${u}) for this device (${c}).`)));const p=u;if(!s.DEFAULT_DTYPE_SUFFIX_MAPPING.hasOwnProperty(p))throw new Error(`Invalid dtype: ${p}. Should be one of: ${Object.keys(s.DATA_TYPES).join(", ")}`);if(p===s.DATA_TYPES.fp16&&"webgpu"===c&&!await(0,s.isWebGpuFp16Supported)())throw new Error(`The device (${c}) does not support fp16.`);const h=s.DEFAULT_DTYPE_SUFFIX_MAPPING[p],m=`${r.subfolder??""}/${t}${h}.onnx`,f={...r.session_options};f.executionProviders??=d;const g=a.free_dimension_overrides;g?f.freeDimensionOverrides??=g:c.startsWith("webnn")&&!f.freeDimensionOverrides&&console.warn('WebNN does not currently support dynamic shapes and requires `free_dimension_overrides` to be set in config.json as a field within "transformers.js_config". When `free_dimension_overrides` is not set, you may experience significant performance degradation.');const M=(0,l.getModelFile)(e,m,!0,r),w=r.use_external_data_format??a.use_external_data_format;let T=[];if(w&&(!0===w||"object"==typeof w&&w.hasOwnProperty(t)&&!0===w[t])){if(_.apis.IS_NODE_ENV)throw new Error("External data format is not yet supported in Node.js");const n=`${t}${h}.onnx_data`,o=`${r.subfolder??""}/${n}`;T.push(new Promise((async(t,s)=>{const a=await(0,l.getModelFile)(e,o,!0,r);t({path:n,data:a})})))}else void 0!==f.externalData&&(T=f.externalData.map((async t=>{if("string"==typeof t.data){const n=await(0,l.getModelFile)(e,t.data,!0,r);return{...t,data:n}}return t})));if(T.length>0&&(f.externalData=await Promise.all(T)),"webgpu"===c){const e=(0,n.getKeyValueShapes)(r.config,{prefix:"present"});if(Object.keys(e).length>0&&!(0,o.isONNXProxy)()){const t={};for(const r in e)t[r]="gpu-buffer";f.preferredOutputLocation=t}}return{buffer:await M,session_options:f}}(e,t[a],r);return[a,await(0,o.createInferenceSession)(i,c)]}))))}async function A(e,t){const r=function(e,t){const r=Object.create(null),n=[];for(const s of e.inputNames){const e=t[s];e instanceof u.Tensor?r[s]=(0,o.isONNXProxy)()?e.clone():e:n.push(s)}if(n.length>0)throw new Error(`An error occurred during model execution: "Missing the following inputs: ${n.join(", ")}.`);const s=Object.keys(t).length,a=e.inputNames.length;if(s>a){let r=Object.keys(t).filter((t=>!e.inputNames.includes(t)));console.warn(`WARNING: Too many inputs were provided (${s} > ${a}). The following inputs will be ignored: "${r.join(", ")}".`)}return r}(e,t);try{const t=Object.fromEntries(Object.entries(r).map((([e,t])=>[e,t.ort_tensor])));let n=await e.run(t);return n=E(n),n}catch(e){throw console.error(`An error occurred during model execution: "${e}".`),console.error("Inputs given to model:",r),e}}function E(e){for(let t in e)(0,o.isONNXTensor)(e[t])?e[t]=new u.Tensor(e[t]):"object"==typeof e[t]&&E(e[t]);return e}function L(e){if(e instanceof u.Tensor)return e;if(0===e.length)throw Error("items must be non-empty");if(Array.isArray(e[0])){if(e.some((t=>t.length!==e[0].length)))throw Error("Unable to create tensor, you should probably activate truncation and/or padding with 'padding=True' and/or 'truncation=True' to have batched tensors with the same length.");return new u.Tensor("int64",BigInt64Array.from(e.flat().map((e=>BigInt(e)))),[e.length,e[0].length])}return new u.Tensor("int64",BigInt64Array.from(e.map((e=>BigInt(e)))),[1,e.length])}function z(e){return new u.Tensor("bool",[e],[1])}async function I(e,t){let{encoder_outputs:r,input_ids:n,decoder_input_ids:o,...s}=t;if(!r){const n=(0,i.pick)(t,e.sessions.model.inputNames);r=(await B(e,n)).last_hidden_state}s.input_ids=o,s.encoder_hidden_states=r,e.sessions.decoder_model_merged.inputNames.includes("encoder_attention_mask")&&(s.encoder_attention_mask=t.attention_mask);return await N(e,s,!0)}async function B(e,t){const r=e.sessions.model,n=(0,i.pick)(t,r.inputNames);if(r.inputNames.includes("inputs_embeds")&&!n.inputs_embeds){if(!t.input_ids)throw new Error("Both `input_ids` and `inputs_embeds` are missing in the model inputs.");n.inputs_embeds=await e.encode_text({input_ids:t.input_ids})}return r.inputNames.includes("token_type_ids")&&!n.token_type_ids&&(n.token_type_ids=new u.Tensor("int64",new BigInt64Array(n.input_ids.data.length),n.input_ids.dims)),await A(r,n)}async function N(e,t,r=!1){const n=e.sessions[r?"decoder_model_merged":"model"],{past_key_values:o,...s}=t;n.inputNames.includes("use_cache_branch")&&(s.use_cache_branch=z(!!o)),n.inputNames.includes("position_ids")&&s.attention_mask&&!s.position_ids&&(s.position_ids=function(e,t=null){const{input_ids:r,inputs_embeds:n,attention_mask:o}=e,[s,a]=o.dims,i=new BigInt64Array(o.data.length);for(let e=0;e<s;++e){const t=e*a;let r=BigInt(0);for(let e=0;e<a;++e){const n=t+e;0n===o.data[n]?i[n]=BigInt(1):(i[n]=r,r+=o.data[n])}}let l=new u.Tensor("int64",i,o.dims);if(t){const e=-(r??n).dims.at(1);l=l.slice(null,[e,null])}return l}(s,o)),e.addPastKeyValues(s,o);const a=(0,i.pick)(s,n.inputNames);return await A(n,a)}async function D(e,{input_ids:t=null,attention_mask:r=null,pixel_values:n=null,position_ids:o=null,inputs_embeds:s=null,past_key_values:a=null,generation_config:i=null,logits_processor:l=null,...c}){if(!s)if(s=await e.encode_text({input_ids:t}),n&&1!==t.dims[1]){const o=await e.encode_image({pixel_values:n});({inputs_embeds:s,attention_mask:r}=e._merge_input_ids_with_image_features({image_features:o,inputs_embeds:s,input_ids:t,attention_mask:r}))}else if(a&&n&&1===t.dims[1]){const e=t.dims[1],n=Object.values(a)[0].dims.at(-2);r=(0,u.cat)([(0,u.ones)([t.dims[0],n]),r.slice(null,[r.dims[1]-e,r.dims[1]])],1)}return await N(e,{inputs_embeds:s,past_key_values:a,attention_mask:r,position_ids:o,generation_config:i,logits_processor:l},!0)}function O(e,t,r,n){if(r.past_key_values){const t=Object.values(r.past_key_values)[0].dims.at(-2),{input_ids:n,attention_mask:o}=r;if(o&&o.dims[1]>n.dims[1]);else if(t<n.dims[1])r.input_ids=n.slice(null,[t,null]);else if(null!=e.config.image_token_index&&n.data.some((t=>t==e.config.image_token_index))){const o=e.config.num_image_tokens;if(!o)throw new Error("`num_image_tokens` is missing in the model configuration.");const s=n.dims[1]-(t-o);r.input_ids=n.slice(null,[-s,null]),r.attention_mask=(0,u.ones)([1,t+s])}}return r}function V(e,t,r,n){return r.past_key_values&&(t=t.map((e=>[e.at(-1)]))),{...r,decoder_input_ids:L(t)}}function j(e,...t){return e.config.is_encoder_decoder?V(e,...t):O(e,...t)}class R extends a.Callable{main_input_name="input_ids";forward_params=["input_ids","attention_mask"];constructor(e,t){super(),this.config=e,this.sessions=t;const r=v.get(this.constructor),n=P.get(r);switch(this.can_generate=!1,this._forward=null,this._prepare_inputs_for_generation=null,n){case x:this.can_generate=!0,this._forward=N,this._prepare_inputs_for_generation=O;break;case T:case b:case F:this.can_generate=!0,this._forward=I,this._prepare_inputs_for_generation=V;break;case w:this._forward=I;break;case k:this.can_generate=!0,this._forward=D,this._prepare_inputs_for_generation=j;break;default:this._forward=B}this.can_generate&&this.forward_params.push("past_key_values"),this.custom_config=this.config["transformers.js_config"]??{}}async dispose(){const e=[];for(const t of Object.values(this.sessions))t?.handler?.dispose&&e.push(t.handler.dispose());return await Promise.all(e)}static async from_pretrained(e,{progress_callback:t=null,config:r=null,cache_dir:o=null,local_files_only:s=!1,revision:a="main",model_file_name:i=null,subfolder:c="onnx",device:d=null,dtype:u=null,use_external_data_format:p=null,session_options:h={}}={}){let m={progress_callback:t,config:r,cache_dir:o,local_files_only:s,revision:a,model_file_name:i,subfolder:c,device:d,dtype:u,use_external_data_format:p,session_options:h};const _=v.get(this),f=P.get(_);let g;if(r=m.config=await n.AutoConfig.from_pretrained(e,m),f===x)g=await Promise.all([S(e,{model:m.model_file_name??"model"},m),(0,l.getModelJSON)(e,"generation_config.json",!1,m)]);else if(f===T||f===b)g=await Promise.all([S(e,{model:"encoder_model",decoder_model_merged:"decoder_model_merged"},m),(0,l.getModelJSON)(e,"generation_config.json",!1,m)]);else if(f===y)g=await Promise.all([S(e,{model:"vision_encoder",prompt_encoder_mask_decoder:"prompt_encoder_mask_decoder"},m)]);else if(f===w)g=await Promise.all([S(e,{model:"encoder_model",decoder_model_merged:"decoder_model_merged"},m)]);else if(f===k){const t={embed_tokens:"embed_tokens",vision_encoder:"vision_encoder",decoder_model_merged:"decoder_model_merged"};r.is_encoder_decoder&&(t.model="encoder_model"),g=await Promise.all([S(e,t,m),(0,l.getModelJSON)(e,"generation_config.json",!1,m)])}else f===F?g=await Promise.all([S(e,{model:"text_encoder",decoder_model_merged:"decoder_model_merged",encodec_decode:"encodec_decode"},m),(0,l.getModelJSON)(e,"generation_config.json",!1,m)]):(f!==M&&console.warn(`Model type for '${_??r?.model_type}' not found, assuming encoder-only architecture. Please report this at https://github.com/xenova/transformers.js/issues/new/choose.`),g=await Promise.all([S(e,{model:m.model_file_name??"model"},m)]));return new this(r,...g)}async _call(e){return await this.forward(e)}async forward(e){return await this._forward(this,e)}_get_logits_warper(e){const t=new c.LogitsProcessorList;return null!==e.temperature&&1!==e.temperature&&t.push(new c.TemperatureLogitsWarper(e.temperature)),null!==e.top_k&&0!==e.top_k&&t.push(new c.TopKLogitsWarper(e.top_k)),null!==e.top_p&&e.top_p<1&&t.push(new c.TopPLogitsWarper(e.top_p)),t}_get_logits_processor(e,t,r=null){const n=new c.LogitsProcessorList;if(null!==e.repetition_penalty&&1!==e.repetition_penalty&&n.push(new c.RepetitionPenaltyLogitsProcessor(e.repetition_penalty)),null!==e.no_repeat_ngram_size&&e.no_repeat_ngram_size>0&&n.push(new c.NoRepeatNGramLogitsProcessor(e.no_repeat_ngram_size)),null!==e.bad_words_ids&&n.push(new c.NoBadWordsLogitsProcessor(e.bad_words_ids,e.eos_token_id)),null!==e.min_length&&null!==e.eos_token_id&&e.min_length>0&&n.push(new c.MinLengthLogitsProcessor(e.min_length,e.eos_token_id)),null!==e.min_new_tokens&&null!==e.eos_token_id&&e.min_new_tokens>0&&n.push(new c.MinNewTokensLengthLogitsProcessor(t,e.min_new_tokens,e.eos_token_id)),null!==e.forced_bos_token_id&&n.push(new c.ForcedBOSTokenLogitsProcessor(e.forced_bos_token_id)),null!==e.forced_eos_token_id&&n.push(new c.ForcedEOSTokenLogitsProcessor(e.max_length,e.forced_eos_token_id)),null!==e.begin_suppress_tokens){const r=t>1||null===e.forced_bos_token_id?t:t+1;n.push(new c.SuppressTokensAtBeginLogitsProcessor(e.begin_suppress_tokens,r))}return null!==e.guidance_scale&&e.guidance_scale>1&&n.push(new c.ClassifierFreeGuidanceLogitsProcessor(e.guidance_scale)),null!==r&&n.extend(r),n}_prepare_generation_config(e,t,r=d.GenerationConfig){const n={...this.config};for(const e of["decoder","generator","text_config"])e in n&&Object.assign(n,n[e]);const o=new r(n);return"generation_config"in this&&Object.assign(o,this.generation_config),e&&Object.assign(o,e),t&&Object.assign(o,(0,i.pick)(t,Object.getOwnPropertyNames(o))),o}_get_stopping_criteria(e,t=null){const r=new h.StoppingCriteriaList;return null!==e.max_length&&r.push(new h.MaxLengthCriteria(e.max_length,this.config.max_position_embeddings??null)),null!==e.eos_token_id&&r.push(new h.EosTokenCriteria(e.eos_token_id)),t&&r.extend(t),r}_validate_model_class(){if(!this.can_generate){const e=[ci,pi,li,ni],t=v.get(this.constructor),r=new Set,n=this.config.model_type;for(const t of e){const e=t.get(n);e&&r.add(e[0])}let o=`The current model class (${t}) is not compatible with \`.generate()\`, as it doesn't have a language model head.`;throw r.size>0&&(o+=` Please use the following class instead: ${[...r].join(", ")}`),Error(o)}}prepare_inputs_for_generation(...e){return this._prepare_inputs_for_generation(this,...e)}_update_model_kwargs_for_generation({generated_input_ids:e,outputs:t,model_inputs:r,is_encoder_decoder:n}){return r.past_key_values=this.getPastKeyValues(t,r.past_key_values),r.input_ids=new u.Tensor("int64",e.flat(),[e.length,1]),n||(r.attention_mask=(0,u.cat)([r.attention_mask,(0,u.ones)([r.attention_mask.dims[0],1])],1)),r.position_ids=null,r}_prepare_model_inputs({inputs:e,bos_token_id:t,model_kwargs:r}){const n=(0,i.pick)(r,this.forward_params),o=this.main_input_name;if(o in n){if(e)throw new Error("`inputs`: {inputs}` were passed alongside {input_name} which is not allowed. Make sure to either pass {inputs} or {input_name}=...")}else n[o]=e;return{inputs_tensor:n[o],model_inputs:n,model_input_name:o}}async _prepare_encoder_decoder_kwargs_for_generation({inputs_tensor:e,model_inputs:t,model_input_name:r,generation_config:n}){if(this.sessions.model.inputNames.includes("inputs_embeds")&&!t.inputs_embeds&&"_prepare_inputs_embeds"in this){const{input_ids:e,pixel_values:r,attention_mask:n,...o}=t,s=await this._prepare_inputs_embeds(t);t={...o,...(0,i.pick)(s,["inputs_embeds","attention_mask"])}}let{last_hidden_state:o}=await B(this,t);if(null!==n.guidance_scale&&n.guidance_scale>1)o=(0,u.cat)([o,(0,u.full_like)(o,0)],0),"attention_mask"in t&&(t.attention_mask=(0,u.cat)([t.attention_mask,(0,u.zeros_like)(t.attention_mask)],0));else if(t.decoder_input_ids){const e=L(t.decoder_input_ids).dims[0];if(e!==o.dims[0]){if(1!==o.dims[0])throw new Error(`The encoder outputs have a different batch size (${o.dims[0]}) than the decoder inputs (${e}).`);o=(0,u.cat)(Array.from({length:e},(()=>o)),0)}}return t.encoder_outputs=o,t}_prepare_decoder_input_ids_for_generation({batch_size:e,model_input_name:t,model_kwargs:r,decoder_start_token_id:n,bos_token_id:o,generation_config:s}){let{decoder_input_ids:a,...i}=r;if(a)Array.isArray(a[0])||(a=Array.from({length:e},(()=>a)));else if(n??=o,"musicgen"===this.config.model_type)a=Array.from({length:e*this.config.decoder.num_codebooks},(()=>[n]));else if(Array.isArray(n)){if(n.length!==e)throw new Error(`\`decoder_start_token_id\` expcted to have length ${e} but got ${n.length}`);a=n}else a=Array.from({length:e},(()=>[n]));return a=L(a),r.decoder_attention_mask=(0,u.ones_like)(a),{input_ids:a,model_inputs:i}}async generate({inputs:e=null,generation_config:t=null,logits_processor:r=null,stopping_criteria:n=null,streamer:o=null,...s}){this._validate_model_class(),t=this._prepare_generation_config(t,s);let{inputs_tensor:a,model_inputs:i,model_input_name:l}=this._prepare_model_inputs({inputs:e,model_kwargs:s});const c=this.config.is_encoder_decoder;let d;c&&("encoder_outputs"in i||(i=await this._prepare_encoder_decoder_kwargs_for_generation({inputs_tensor:a,model_inputs:i,model_input_name:l,generation_config:t}))),c?({input_ids:d,model_inputs:i}=this._prepare_decoder_input_ids_for_generation({batch_size:i[l].dims.at(0),model_input_name:l,model_kwargs:i,decoder_start_token_id:t.decoder_start_token_id,bos_token_id:t.bos_token_id,generation_config:t})):d=i[l];let p=d.dims.at(-1);null!==t.max_new_tokens&&(t.max_length=p+t.max_new_tokens);const h=this._get_logits_processor(t,p,r),_=this._get_stopping_criteria(t,n),f=i[l].dims.at(0),g=m.LogitsSampler.getSampler(t),M=new Array(f).fill(0),w=d.tolist();let T;o&&o.put(w);let b={};for(;;){if(i=this.prepare_inputs_for_generation(w,i,t),T=await this.forward(i),t.output_attentions&&t.return_dict_in_generate){const e=this.getAttentions(T);for(const t in e)t in b||(b[t]=[]),b[t].push(e[t])}const e=h(w,T.logits.slice(null,-1,null)),r=[];for(let t=0;t<e.dims.at(0);++t){const n=e[t],o=await g(n);for(const[e,n]of o){const o=BigInt(e);M[t]+=n,w[t].push(o),r.push([o]);break}}o&&o.put(r);if(_(w).every((e=>e)))break;i=this._update_model_kwargs_for_generation({generated_input_ids:r,outputs:T,model_inputs:i,is_encoder_decoder:c})}o&&o.end();const x=this.getPastKeyValues(T,i.past_key_values,!0),y=new u.Tensor("int64",w.flat(),[w.length,w[0].length]);if(t.return_dict_in_generate)return{sequences:y,past_key_values:x,...b};for(const e of Object.values(T))"gpu-buffer"===e.location&&e.dispose();return y}getPastKeyValues(e,t,r=!1){const n=Object.create(null);for(const o in e)if(o.startsWith("present")){const s=o.replace("present","past_key_values"),a=o.includes("encoder");if(n[s]=a&&t?t[s]:e[o],t&&(!a||r)){const e=t[s];"gpu-buffer"===e.location&&e.dispose()}}return n}getAttentions(e){const t={};for(const r of["cross_attentions","encoder_attentions","decoder_attentions"])for(const n in e)n.startsWith(r)&&(r in t||(t[r]=[]),t[r].push(e[n]));return t}addPastKeyValues(e,t){if(t)Object.assign(e,t);else{const t=this.custom_config.kv_cache_dtype??"float32",r="float16"===t?new Uint16Array:[],o=(0,n.getKeyValueShapes)(this.config);for(const n in o)e[n]=new u.Tensor(t,r,o[n])}}async encode_image({pixel_values:e}){const t=(await A(this.sessions.vision_encoder,{pixel_values:e})).image_features;return this.config.num_image_tokens||(console.warn(`The number of image tokens was not set in the model configuration. Setting it to the number of features detected by the vision encoder (${t.dims[1]}).`),this.config.num_image_tokens=t.dims[1]),t}async encode_text({input_ids:e}){return(await A(this.sessions.embed_tokens,{input_ids:e})).inputs_embeds}}class G{}class q extends G{constructor({last_hidden_state:e,hidden_states:t=null,attentions:r=null}){super(),this.last_hidden_state=e,this.hidden_states=t,this.attentions=r}}class W extends R{}class $ extends W{}class U extends W{async _call(e){return new ul(await super._call(e))}}class X extends W{async _call(e){return new ll(await super._call(e))}}class Q extends W{async _call(e){return new dl(await super._call(e))}}class H extends W{async _call(e){return new pl(await super._call(e))}}class Y extends R{}class J extends Y{}class K extends R{}class Z extends K{}class ee extends K{async _call(e){return new ul(await super._call(e))}}class te extends K{async _call(e){return new ll(await super._call(e))}}class re extends K{async _call(e){return new dl(await super._call(e))}}class ne extends K{async _call(e){return new pl(await super._call(e))}}class oe extends R{}class se extends oe{}class ae extends oe{async _call(e){return new ul(await super._call(e))}}class ie extends oe{async _call(e){return new ll(await super._call(e))}}class le extends oe{async _call(e){return new dl(await super._call(e))}}class ce extends oe{async _call(e){return new pl(await super._call(e))}}class de extends R{}class ue extends de{}class pe extends de{async _call(e){return new ul(await super._call(e))}}class he extends de{async _call(e){return new ll(await super._call(e))}}class me extends de{async _call(e){return new dl(await super._call(e))}}class _e extends de{async _call(e){return new pl(await super._call(e))}}class fe extends R{}class ge extends fe{}class Me extends fe{async _call(e){return new ul(await super._call(e))}}class we extends fe{async _call(e){return new ll(await super._call(e))}}class Te extends fe{async _call(e){return new dl(await super._call(e))}}class be extends fe{async _call(e){return new pl(await super._call(e))}}class xe extends R{}class ye extends xe{}class ke extends xe{async _call(e){return new ul(await super._call(e))}}class Fe extends xe{async _call(e){return new ll(await super._call(e))}}class Pe extends xe{async _call(e){return new dl(await super._call(e))}}class Ce extends xe{async _call(e){return new pl(await super._call(e))}}class ve extends R{}class Se extends ve{}class Ae extends ve{async _call(e){return new ul(await super._call(e))}}class Ee extends ve{async _call(e){return new ll(await super._call(e))}}class Le extends ve{async _call(e){return new dl(await super._call(e))}}class ze extends ve{async _call(e){return new pl(await super._call(e))}}class Ie extends R{}class Be extends Ie{}class Ne extends Ie{async _call(e){return new ll(await super._call(e))}}class De extends Ie{async _call(e){return new dl(await super._call(e))}}class Oe extends Ie{async _call(e){return new pl(await super._call(e))}}class Ve extends Ie{async _call(e){return new ul(await super._call(e))}}class je extends R{}class Re extends je{}class Ge extends je{async _call(e){return new ul(await super._call(e))}}class qe extends je{async _call(e){return new ll(await super._call(e))}}class We extends je{async _call(e){return new dl(await super._call(e))}}class $e extends R{}class Ue extends $e{}class Xe extends $e{async _call(e){return new ul(await super._call(e))}}class Qe extends $e{async _call(e){return new ll(await super._call(e))}}class He extends $e{async _call(e){return new pl(await super._call(e))}}class Ye extends R{}class Je extends Ye{}class Ke extends Ye{async _call(e){return new ul(await super._call(e))}}class Ze extends Ye{async _call(e){return new ll(await super._call(e))}}class et extends Ye{async _call(e){return new dl(await super._call(e))}}class tt extends Ye{async _call(e){return new pl(await super._call(e))}}class rt extends R{}class nt extends rt{}class ot extends rt{async _call(e){return new ul(await super._call(e))}}class st extends rt{async _call(e){return new ll(await super._call(e))}}class at extends rt{async _call(e){return new pl(await super._call(e))}}class it extends R{}class lt extends it{}class ct extends it{async _call(e){return new ll(await super._call(e))}}class dt extends it{async _call(e){return new pl(await super._call(e))}}class ut extends it{async _call(e){return new ul(await super._call(e))}}class pt extends R{forward_params=["input_ids","attention_mask","encoder_outputs","decoder_input_ids","decoder_attention_mask","past_key_values"];constructor(e,t,r){super(e,t),this.generation_config=r}}class ht extends pt{}class mt extends pt{}class _t extends R{constructor(e,t,r){super(e,t),this.generation_config=r}}class ft extends _t{}class gt extends _t{}class Mt extends R{constructor(e,t,r){super(e,t),this.generation_config=r}}class wt extends Mt{}class Tt extends Mt{}class bt extends R{constructor(e,t,r){super(e,t),this.generation_config=r}}class xt extends bt{}class yt extends bt{}class kt extends bt{async _call(e){return new ll(await super._call(e))}}class Ft extends R{constructor(e,t,r){super(e,t),this.generation_config=r}}class Pt extends Ft{}class Ct extends Ft{}class vt extends Ft{async _call(e){return new ll(await super._call(e))}}class St extends Ft{}class At extends R{constructor(e,t,r){super(e,t),this.generation_config=r}}class Et extends At{}class Lt extends At{}class zt extends R{constructor(e,t,r){super(e,t),this.generation_config=r}}class It extends zt{}class Bt extends zt{}class Nt extends R{}class Dt extends Nt{}class Ot extends Nt{async _call(e){return new ul(await super._call(e))}}class Vt extends Nt{async _call(e){return new ll(await super._call(e))}}class jt extends Nt{async _call(e){return new dl(await super._call(e))}}class Rt extends Nt{async _call(e){return new pl(await super._call(e))}}class Gt extends R{}class qt extends Gt{}class Wt extends Gt{async _call(e){return new ul(await super._call(e))}}class $t extends Gt{async _call(e){return new ll(await super._call(e))}}class Ut extends Gt{async _call(e){return new dl(await super._call(e))}}class Xt extends Gt{async _call(e){return new pl(await super._call(e))}}class Qt extends R{}class Ht extends Qt{}class Yt extends Qt{async _call(e){return new ul(await super._call(e))}}class Jt extends Qt{async _call(e){return new ll(await super._call(e))}}class Kt extends Qt{async _call(e){return new dl(await super._call(e))}}class Zt extends Qt{async _call(e){return new pl(await super._call(e))}}class er extends R{}class tr extends er{}class rr extends er{}class nr extends R{requires_attention_mask=!1;main_input_name="input_features";forward_params=["input_features","attention_mask","decoder_input_ids","decoder_attention_mask","past_key_values"];constructor(e,t,r){super(e,t),this.generation_config=r}}class or extends nr{}class sr extends nr{_prepare_generation_config(e,t){return super._prepare_generation_config(e,t,f.WhisperGenerationConfig)}_retrieve_init_tokens(e){const t=[e.decoder_start_token_id];let r=e.language;const n=e.task;if(e.is_multilingual){r||(console.warn("No language specified - defaulting to English (en)."),r="en");const o=`<|${(0,g.whisper_language_to_code)(r)}|>`;t.push(e.lang_to_id[o]),t.push(e.task_to_id[n??"transcribe"])}else if(r||n)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:r=null,stopping_criteria:n=null,...o}){t=this._prepare_generation_config(t,o);const s=o.decoder_input_ids??this._retrieve_init_tokens(t);if(t.return_timestamps&&(r??=new c.LogitsProcessorList,r.push(new c.WhisperTimeStampLogitsProcessor(t,s))),t.begin_suppress_tokens&&(r??=new c.LogitsProcessorList,r.push(new c.SuppressTokensAtBeginLogitsProcessor(t.begin_suppress_tokens,s.length))),t.return_token_timestamps){if(!t.alignment_heads)throw new Error("Model generation config has no `alignment_heads`, token-level timestamps not available. See https://gist.github.com/hollance/42e32852f24243b748ae6bc1f985b13a on how to add this property to the generation config.");"translate"===t.task&&console.warn("Token-level timestamps may not be reliable for task 'translate'."),t.output_attentions=!0,t.return_dict_in_generate=!0}const a=await super.generate({inputs:e,generation_config:t,logits_processor:r,decoder_input_ids:s,...o});return t.return_token_timestamps&&(a.token_timestamps=this._extract_token_timestamps(a,t.alignment_heads,t.num_frames)),a}_extract_token_timestamps(e,t,r=null,n=.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==r&&console.warn("`num_frames` has not been set, meaning the entire audio will be analyzed. This may lead to inaccurate token-level timestamps for short audios (< 30 seconds).");let o=this.config.median_filter_width;void 0===o&&(console.warn("Model config has no `median_filter_width`, using default value of 7."),o=7);const s=e.cross_attentions,a=Array.from({length:this.config.decoder_layers},((e,t)=>(0,u.cat)(s.map((e=>e[t])),2))),l=(0,u.stack)(t.map((([e,t])=>{if(e>=a.length)throw new Error(`Layer index ${e} is out of bounds for cross attentions (length ${a.length}).`);return r?a[e].slice(null,t,null,[0,r]):a[e].slice(null,t)}))).transpose(1,0,2,3),[c,d]=(0,u.std_mean)(l,-2,0,!0),h=l.clone();for(let e=0;e<h.dims[0];++e){const t=h[e];for(let r=0;r<t.dims[0];++r){const n=t[r],s=c[e][r][0].data,a=d[e][r][0].data;for(let e=0;e<n.dims[0];++e){let t=n[e].data;for(let e=0;e<t.length;++e)t[e]=(t[e]-a[e])/s[e];t.set((0,p.medianFilter)(t,o))}}}const m=[(0,u.mean)(h,1)],_=e.sequences.dims,f=new u.Tensor("float32",new Float32Array(_[0]*_[1]),_);for(let e=0;e<_[0];++e){const t=m[e].neg().squeeze_(0),[r,o]=(0,p.dynamic_time_warping)(t.tolist()),s=Array.from({length:r.length-1},((e,t)=>r[t+1]-r[t])),a=(0,i.mergeArrays)([1],s).map((e=>!!e)),l=[];for(let e=0;e<a.length;++e)a[e]&&l.push(o[e]*n);f[e].data.set(l,1)}return f}}class ar extends R{main_input_name="pixel_values";forward_params=["pixel_values","input_ids","encoder_hidden_states","past_key_values"];constructor(e,t,r){super(e,t),this.generation_config=r}}class ir extends R{forward_params=["input_ids","pixel_values","attention_mask","position_ids","past_key_values"];constructor(e,t,r){super(e,t),this.generation_config=r}}class lr extends ir{_merge_input_ids_with_image_features({inputs_embeds:e,image_features:t,input_ids:r,attention_mask:n}){const o=this.config.image_token_index,s=r.tolist().map((e=>e.findIndex((e=>e==o)))),a=s.every((e=>-1===e)),i=s.every((e=>-1!==e));if(!a&&!i)throw new Error("Every input should contain either 0 or 1 image token.");if(a)return{inputs_embeds:e,attention_mask:n};const l=[],c=[];for(let r=0;r<s.length;++r){const o=s[r],a=e[r],i=t[r],d=n[r];l.push((0,u.cat)([a.slice([0,o]),i,a.slice([o+1,a.dims[0]])],0)),c.push((0,u.cat)([d.slice([0,o]),(0,u.ones)([i.dims[0]]),d.slice([o+1,d.dims[0]])],0))}return{inputs_embeds:(0,u.stack)(l,0),attention_mask:(0,u.stack)(c,0)}}}class cr extends lr{}class dr extends R{forward_params=["input_ids","inputs_embeds","attention_mask","pixel_values","encoder_outputs","decoder_input_ids","decoder_inputs_embeds","decoder_attention_mask","past_key_values"];main_input_name="inputs_embeds";constructor(e,t,r){super(e,t),this.generation_config=r}}class ur extends dr{_merge_input_ids_with_image_features({inputs_embeds:e,image_features:t,input_ids:r,attention_mask:n}){return{inputs_embeds:(0,u.cat)([t,e],1),attention_mask:(0,u.cat)([(0,u.ones)(t.dims.slice(0,2)),n],1)}}async _prepare_inputs_embeds({input_ids:e,pixel_values:t,inputs_embeds:r,attention_mask:n}){if(!e&&!t)throw new Error("Either `input_ids` or `pixel_values` should be provided.");let o,s;return e&&(o=await this.encode_text({input_ids:e})),t&&(s=await this.encode_image({pixel_values:t})),o&&s?({inputs_embeds:r,attention_mask:n}=this._merge_input_ids_with_image_features({inputs_embeds:o,image_features:s,input_ids:e,attention_mask:n})):r=o||s,{inputs_embeds:r,attention_mask:n}}async forward({input_ids:e,pixel_values:t,attention_mask:r,decoder_input_ids:n,decoder_attention_mask:o,encoder_outputs:s,past_key_values:a,inputs_embeds:i,decoder_inputs_embeds:l}){if(i||({inputs_embeds:i,attention_mask:r}=await this._prepare_inputs_embeds({input_ids:e,pixel_values:t,inputs_embeds:i,attention_mask:r})),!s){let{last_hidden_state:e}=await B(this,{inputs_embeds:i,attention_mask:r});s=e}if(!l){if(!n)throw new Error("Either `decoder_input_ids` or `decoder_inputs_embeds` should be provided.");l=await this.encode_text({input_ids:n})}const c={inputs_embeds:l,attention_mask:o,encoder_attention_mask:r,encoder_hidden_states:s,past_key_values:a};return await N(this,c,!0)}}class pr extends R{}class hr extends pr{}class mr extends pr{static async from_pretrained(e,t={}){return t.model_file_name??="text_model",super.from_pretrained(e,t)}}class _r extends pr{static async from_pretrained(e,t={}){return t.model_file_name??="text_model",super.from_pretrained(e,t)}}class fr extends pr{static async from_pretrained(e,t={}){return t.model_file_name??="vision_model",super.from_pretrained(e,t)}}class gr extends pr{static async from_pretrained(e,t={}){return t.model_file_name??="vision_model",super.from_pretrained(e,t)}}class Mr extends R{}class wr extends Mr{}class Tr extends Mr{static async from_pretrained(e,t={}){return t.model_file_name??="text_model",super.from_pretrained(e,t)}}class br extends pr{static async from_pretrained(e,t={}){return t.model_file_name??="vision_model",super.from_pretrained(e,t)}}class xr extends R{}class yr extends xr{}class kr extends R{}class Fr extends kr{}class Pr extends kr{}class Cr extends R{constructor(e,t,r){super(e,t),this.generation_config=r}}class vr extends Cr{}class Sr extends Cr{}class Ar extends R{constructor(e,t,r){super(e,t),this.generation_config=r}}class Er extends Ar{}class Lr extends Ar{}class zr extends R{constructor(e,t,r){super(e,t),this.generation_config=r}}class Ir extends zr{}class Br extends zr{}class Nr extends R{constructor(e,t,r){super(e,t),this.generation_config=r}}class Dr extends Nr{}class Or extends Nr{}class Vr extends R{constructor(e,t,r){super(e,t),this.generation_config=r}}class jr extends Vr{}class Rr extends Vr{}class Gr extends R{constructor(e,t,r){super(e,t),this.generation_config=r}}class qr extends Gr{}class Wr extends Gr{}class $r extends R{constructor(e,t,r){super(e,t),this.generation_config=r}}class Ur extends $r{}class Xr extends $r{}class Qr extends R{constructor(e,t,r){super(e,t),this.generation_config=r}}class Hr extends Qr{}class Yr extends Qr{}class Jr extends R{constructor(e,t,r){super(e,t),this.generation_config=r}}class Kr extends Jr{}class Zr extends Jr{}class en extends R{constructor(e,t,r){super(e,t),this.generation_config=r}}class tn extends en{}class rn extends en{}class nn extends R{constructor(e,t,r){super(e,t),this.generation_config=r}}class on extends nn{}class sn extends nn{}class an extends R{constructor(e,t,r){super(e,t),this.generation_config=r}}class ln extends an{}class cn extends an{}class dn extends R{constructor(e,t,r){super(e,t),this.generation_config=r}}class un extends dn{}class pn extends dn{}class hn extends R{constructor(e,t,r){super(e,t),this.generation_config=r}}class mn extends hn{}class _n extends hn{}class fn extends R{constructor(e,t,r){super(e,t),this.generation_config=r}}class gn extends fn{}class Mn extends fn{}class wn extends R{constructor(e,t,r){super(e,t),this.generation_config=r}}class Tn extends wn{}class bn extends wn{}class xn extends R{constructor(e,t,r){super(e,t),this.generation_config=r}}class yn extends xn{}class kn extends xn{}class Fn extends R{constructor(e,t,r){super(e,t),this.generation_config=r}}class Pn extends Fn{}class Cn extends Fn{}class vn extends R{}class Sn extends vn{}class An extends vn{async _call(e){return new ll(await super._call(e))}}class En extends R{}class Ln extends En{}class zn extends En{async _call(e){return new ll(await super._call(e))}}class In extends R{}class Bn extends In{}class Nn extends R{}class Dn extends Nn{}class On extends Nn{async _call(e){return new ll(await super._call(e))}}class Vn extends R{}class jn extends Vn{}class Rn extends R{}class Gn extends Rn{}class qn extends Rn{async _call(e){return new ll(await super._call(e))}}class Wn extends R{}class $n extends Wn{async _call(e){return new _l(await super._call(e))}}class Un extends R{}class Xn extends Un{}class Qn extends Un{async _call(e){return new ll(await super._call(e))}}class Hn extends R{}class Yn extends Hn{}class Jn extends Hn{async _call(e){return new ll(await super._call(e))}}class Kn extends R{}class Zn extends Kn{}class eo extends Kn{}class to extends R{}class ro extends to{}class no extends to{}class oo extends R{}class so extends oo{}class ao extends oo{async _call(e){return new ll(await super._call(e))}}class io extends R{}class lo extends io{}class co extends io{async _call(e){return new po(await super._call(e))}}class uo extends io{async _call(e){return new ho(await super._call(e))}}class po extends G{constructor({logits:e,pred_boxes:t}){super(),this.logits=e,this.pred_boxes=t}}class ho extends G{constructor({logits:e,pred_boxes:t,pred_masks:r}){super(),this.logits=e,this.pred_boxes=t,this.pred_masks=r}}class mo extends R{}class _o extends mo{}class fo extends mo{async _call(e){return new go(await super._call(e))}}class go extends G{constructor({logits:e,pred_boxes:t}){super(),this.logits=e,this.pred_boxes=t}}class Mo extends R{}class wo extends Mo{}class To extends Mo{async _call(e){return new bo(await super._call(e))}}class bo extends po{}class xo extends R{}class yo extends xo{}class ko extends xo{async _call(e){return new ll(await super._call(e))}}class Fo extends R{}class Po extends Fo{}class Co extends Fo{async _call(e){return new ll(await super._call(e))}}class vo extends R{}class So extends vo{}class Ao extends vo{async _call(e){return new ll(await super._call(e))}}class Eo extends R{}class Lo extends Eo{}class zo extends Eo{async _call(e){return new ll(await super._call(e))}}class Io extends R{}class Bo extends Io{}class No extends Io{}class Do extends R{}class Oo extends Do{}class Vo extends Do{}class jo extends R{}class Ro extends jo{}class Go extends R{}class qo extends Go{}class Wo extends Go{}class $o extends Go{}class Uo extends R{}class Xo extends Uo{}class Qo extends R{}class Ho extends Qo{}class Yo extends Qo{}class Jo extends R{}class Ko extends Jo{}class Zo extends Jo{}class es extends R{}class ts extends es{}class rs extends R{}class ns extends rs{}class os extends rs{async _call(e){return new ll(await super._call(e))}}class ss extends R{}class as extends ss{}class is extends ss{async _call(e){return new ll(await super._call(e))}}class ls extends R{}class cs extends ls{}class ds extends ls{async _call(e){return new ll(await super._call(e))}}class us extends R{}class ps extends us{}class hs extends us{async _call(e){return new ms(await super._call(e))}}class ms extends G{constructor({logits:e,pred_boxes:t}){super(),this.logits=e,this.pred_boxes=t}}class _s extends R{}class fs extends _s{async get_image_embeddings({pixel_values:e}){return await B(this,{pixel_values:e})}async forward(e){if(e.image_embeddings&&e.image_positional_embeddings||(e={...e,...await this.get_image_embeddings(e)}),!e.input_labels&&e.input_points){const t=e.input_points.dims.slice(0,-1),r=t.reduce(((e,t)=>e*t),1);e.input_labels=new u.Tensor("int64",new BigInt64Array(r).fill(1n),t)}const t={image_embeddings:e.image_embeddings,image_positional_embeddings:e.image_positional_embeddings};return e.input_points&&(t.input_points=e.input_points),e.input_labels&&(t.input_labels=e.input_labels),e.input_boxes&&(t.input_boxes=e.input_boxes),await A(this.sessions.prompt_encoder_mask_decoder,t)}async _call(e){return new gs(await super._call(e))}}class gs extends G{constructor({iou_scores:e,pred_masks:t}){super(),this.iou_scores=e,this.pred_masks=t}}class Ms extends R{constructor(e,t,r){super(e,t),this.generation_config=r}}class ws extends Ms{}class Ts extends Ms{}class bs extends R{constructor(e,t,r){super(e,t),this.generation_config=r}}class xs extends bs{}class ys extends bs{}class ks extends R{}class Fs extends ks{}class Ps extends ks{async _call(e){return new hl(await super._call(e))}}class Cs extends ks{async _call(e){return new ll(await super._call(e))}}class vs extends ks{async _call(e){return new dl(await super._call(e))}}class Ss extends R{}class As extends Ss{}class Es extends Ss{async _call(e){return new dl(await super._call(e))}}class Ls extends R{}class zs extends Ls{}class Is extends R{}class Bs extends Is{}class Ns extends Is{async _call(e){return new hl(await super._call(e))}}class Ds extends Is{async _call(e){return new ll(await super._call(e))}}class Os extends R{}class Vs extends Os{}class js extends Os{async _call(e){return new hl(await super._call(e))}}class Rs extends Os{async _call(e){return new ll(await super._call(e))}}class Gs extends Os{async _call(e){return new dl(await super._call(e))}}class qs extends R{}class Ws extends qs{}class $s extends qs{async _call(e){return new hl(await super._call(e))}}class Us extends qs{async _call(e){return new ll(await super._call(e))}}class Xs extends R{}class Qs extends ks{}class Hs extends ks{async _call(e){return new hl(await super._call(e))}}class Ys extends ks{async _call(e){return new ll(await super._call(e))}}class Js extends R{}class Ks extends Js{}class Zs extends Js{async _call(e){return new hl(await super._call(e))}}class ea extends Js{async _call(e){return new ll(await super._call(e))}}class ta extends Js{async _call(e){return new cl(await super._call(e))}}class ra extends Js{async _call(e){return new dl(await super._call(e))}}class na extends R{constructor(e,t,r){super(e,t),this.generation_config=r}}class oa extends na{}class sa extends na{}class aa extends na{async generate_speech(e,t,{threshold:r=.5,minlenratio:n=0,maxlenratio:o=20,vocoder:s=null}={}){const a={input_ids:e},{encoder_outputs:i,encoder_attention_mask:l}=await B(this,a),c=i.dims[1]/this.config.reduction_factor,d=Math.floor(c*o),p=Math.floor(c*n),h=this.config.num_mel_bins;let m=[],_=null,f=null,g=0;for(;;){++g;const e=z(!!f);let n;n=f?f.output_sequence_out:new u.Tensor("float32",new Float32Array(h),[1,1,h]);let o={use_cache_branch:e,output_sequence:n,encoder_attention_mask:l,speaker_embeddings:t,encoder_hidden_states:i};this.addPastKeyValues(o,_),f=await A(this.sessions.decoder_model_merged,o),_=this.getPastKeyValues(f,_);const{prob:s,spectrum:a}=f;if(m.push(a),g>=p&&(Array.from(s.data).filter((e=>e>=r)).length>0||g>=d))break}const M=(0,u.cat)(m),{waveform:w}=await A(s.sessions.model,{spectrogram:M});return{spectrogram:M,waveform:w}}}class ia extends R{main_input_name="spectrogram"}class la extends R{constructor(e,t,r){super(e,t),this.generation_config=r}}class ca extends la{}class da extends R{constructor(e,t,r){super(e,t),this.generation_config=r}}class ua extends da{}class pa extends da{}class ha extends R{constructor(e,t,r){super(e,t),this.generation_config=r}}class ma extends ha{}class _a extends ha{}class fa extends R{constructor(e,t,r){super(e,t),this.generation_config=r}}class ga extends fa{}class Ma extends fa{}class wa extends R{}class Ta extends wa{}class ba extends wa{static async from_pretrained(e,t={}){return t.model_file_name??="text_model",super.from_pretrained(e,t)}}class xa extends wa{static async from_pretrained(e,t={}){return t.model_file_name??="audio_model",super.from_pretrained(e,t)}}class ya extends R{}class ka extends ya{async _call(e){return new fl(await super._call(e))}}class Fa extends R{}class Pa extends Fa{}class Ca extends Fa{}class va extends Fa{}class Sa extends R{constructor(e,t,r){super(e,t),this.generation_config=r}}class Aa extends Sa{}class Ea extends Sa{}class La extends R{}class za extends La{}class Ia extends La{async _call(e){return new ll(await super._call(e))}}class Ba extends R{}class Na extends Ba{}class Da extends Ba{}class Oa extends R{forward_params=["input_ids","attention_mask","encoder_outputs","decoder_input_ids","decoder_attention_mask","past_key_values"];constructor(e,t,r){super(e,t),this.generation_config=r}_apply_and_filter_by_delay_pattern_mask(e){const[t,r]=e.dims,n=this.config.decoder.num_codebooks,o=r-n;let s=0;for(let t=0;t<e.size;++t){if(e.data[t]===this.config.decoder.pad_token_id)continue;const a=t%r-Math.floor(t/r)%n;a>0&&a<=o&&(e.data[s++]=e.data[t])}const a=Math.floor(t/n),i=s/(a*n);return new u.Tensor(e.type,e.data.slice(0,s),[a,n,i])}prepare_inputs_for_generation(e,t,r){let n=structuredClone(e);for(let e=0;e<n.length;++e)for(let t=0;t<n[e].length;++t)e%this.config.decoder.num_codebooks>=t&&(n[e][t]=BigInt(this.config.decoder.pad_token_id));null!==r.guidance_scale&&r.guidance_scale>1&&(n=n.concat(n));return super.prepare_inputs_for_generation(n,t,r)}async generate(e){const t=await super.generate(e),r=this._apply_and_filter_by_delay_pattern_mask(t).unsqueeze_(0),{audio_values:n}=await A(this.sessions.encodec_decode,{audio_codes:r});return n}}class Va extends R{}class ja extends Va{}class Ra extends Va{async _call(e){return new ll(await super._call(e))}}class Ga extends R{}class qa extends Ga{}class Wa extends Ga{async _call(e){return new ll(await super._call(e))}}class $a extends R{}class Ua extends $a{}class Xa extends $a{async _call(e){return new ll(await super._call(e))}}class Qa extends R{}class Ha extends Qa{}class Ya extends Qa{async _call(e){return new ll(await super._call(e))}}class Ja extends R{}class Ka extends Ja{}class Za{static MODEL_CLASS_MAPPINGS=null;static BASE_IF_FAIL=!1;static async from_pretrained(e,{progress_callback:t=null,config:r=null,cache_dir:o=null,local_files_only:s=!1,revision:a="main",model_file_name:i=null,subfolder:l="onnx",device:c=null,dtype:d=null,use_external_data_format:u=null,session_options:p={}}={}){const h={progress_callback:t,config:r,cache_dir:o,local_files_only:s,revision:a,model_file_name:i,subfolder:l,device:c,dtype:d,use_external_data_format:u,session_options:p};if(h.config=await n.AutoConfig.from_pretrained(e,h),!this.MODEL_CLASS_MAPPINGS)throw new Error("`MODEL_CLASS_MAPPINGS` not implemented for this type of `AutoClass`: "+this.name);for(const t of this.MODEL_CLASS_MAPPINGS){const r=t.get(h.config.model_type);if(r)return await r[1].from_pretrained(e,h)}if(this.BASE_IF_FAIL)return console.warn(`Unknown model class "${h.config.model_type}", attempting to construct from base class.`),await R.from_pretrained(e,h);throw Error(`Unsupported model type: ${h.config.model_type}`)}}const ei=new Map([["bert",["BertModel",$]],["nomic_bert",["NomicBertModel",J]],["roformer",["RoFormerModel",Z]],["electra",["ElectraModel",ue]],["esm",["EsmModel",Re]],["convbert",["ConvBertModel",se]],["camembert",["CamembertModel",ge]],["deberta",["DebertaModel",ye]],["deberta-v2",["DebertaV2Model",Se]],["mpnet",["MPNetModel",Je]],["albert",["AlbertModel",lt]],["distilbert",["DistilBertModel",Be]],["roberta",["RobertaModel",Dt]],["xlm",["XLMModel",qt]],["xlm-roberta",["XLMRobertaModel",Ht]],["clap",["ClapModel",Ta]],["clip",["CLIPModel",hr]],["clipseg",["CLIPSegModel",Fr]],["chinese_clip",["ChineseCLIPModel",yr]],["siglip",["SiglipModel",wr]],["mobilebert",["MobileBertModel",Ue]],["squeezebert",["SqueezeBertModel",nt]],["wav2vec2",["Wav2Vec2Model",Fs]],["wav2vec2-bert",["Wav2Vec2BertModel",Ws]],["unispeech",["UniSpeechModel",Bs]],["unispeech-sat",["UniSpeechSatModel",Vs]],["hubert",["HubertModel",Qs]],["wavlm",["WavLMModel",Ks]],["audio-spectrogram-transformer",["ASTModel",tr]],["vits",["VitsModel",ka]],["pyannote",["PyAnnoteModel",As]],["wespeaker-resnet",["WeSpeakerResNetModel",zs]],["detr",["DetrModel",lo]],["rt_detr",["RTDetrModel",_o]],["table-transformer",["TableTransformerModel",wo]],["vit",["ViTModel",Sn]],["pvt",["PvtModel",Ln]],["vit_msn",["ViTMSNModel",Dn]],["vit_mae",["ViTMAEModel",Bn]],["groupvit",["GroupViTModel",jn]],["fastvit",["FastViTModel",Gn]],["mobilevit",["MobileViTModel",Xn]],["mobilevitv2",["MobileViTV2Model",Yn]],["owlvit",["OwlViTModel",Zn]],["owlv2",["Owlv2Model",ro]],["beit",["BeitModel",so]],["deit",["DeiTModel",yo]],["hiera",["HieraModel",Po]],["convnext",["ConvNextModel",ns]],["convnextv2",["ConvNextV2Model",as]],["dinov2",["Dinov2Model",cs]],["resnet",["ResNetModel",So]],["swin",["SwinModel",Lo]],["swin2sr",["Swin2SRModel",Bo]],["donut-swin",["DonutSwinModel",ts]],["yolos",["YolosModel",ps]],["dpt",["DPTModel",Oo]],["glpn",["GLPNModel",Ko]],["hifigan",["SpeechT5HifiGan",ia]],["efficientnet",["EfficientNetModel",za]],["decision_transformer",["DecisionTransformerModel",Ka]],["mobilenet_v1",["MobileNetV1Model",ja]],["mobilenet_v2",["MobileNetV2Model",qa]],["mobilenet_v3",["MobileNetV3Model",Ua]],["mobilenet_v4",["MobileNetV4Model",Ha]],["maskformer",["MaskFormerModel",Ho]]]),ti=new Map([["t5",["T5Model",ht]],["longt5",["LongT5Model",ft]],["mt5",["MT5Model",wt]],["bart",["BartModel",xt]],["mbart",["MBartModel",Pt]],["marian",["MarianModel",ws]],["whisper",["WhisperModel",or]],["m2m_100",["M2M100Model",xs]],["blenderbot",["BlenderbotModel",Et]],["blenderbot-small",["BlenderbotSmallModel",It]]]),ri=new Map([["bloom",["BloomModel",Tn]],["jais",["JAISModel",Er]],["gpt2",["GPT2Model",vr]],["gptj",["GPTJModel",jr]],["gpt_bigcode",["GPTBigCodeModel",qr]],["gpt_neo",["GPTNeoModel",Ir]],["gpt_neox",["GPTNeoXModel",Dr]],["codegen",["CodeGenModel",Ur]],["llama",["LlamaModel",Hr]],["cohere",["CohereModel",Kr]],["gemma",["GemmaModel",tn]],["gemma2",["Gemma2Model",on]],["openelm",["OpenELMModel",ln]],["qwen2",["Qwen2Model",un]],["phi",["PhiModel",mn]],["phi3",["Phi3Model",gn]],["mpt",["MptModel",yn]],["opt",["OPTModel",Pn]],["mistral",["MistralModel",ua]],["starcoder2",["Starcoder2Model",ma]],["falcon",["FalconModel",ga]],["stablelm",["StableLmModel",Aa]]]),ni=new Map([["speecht5",["SpeechT5ForSpeechToText",sa]],["whisper",["WhisperForConditionalGeneration",sr]]]),oi=new Map([["speecht5",["SpeechT5ForTextToSpeech",aa]]]),si=new Map([["vits",["VitsModel",ka]],["musicgen",["MusicgenForConditionalGeneration",Oa]]]),ai=new Map([["bert",["BertForSequenceClassification",X]],["roformer",["RoFormerForSequenceClassification",te]],["electra",["ElectraForSequenceClassification",he]],["esm",["EsmForSequenceClassification",qe]],["convbert",["ConvBertForSequenceClassification",ie]],["camembert",["CamembertForSequenceClassification",we]],["deberta",["DebertaForSequenceClassification",Fe]],["deberta-v2",["DebertaV2ForSequenceClassification",Ee]],["mpnet",["MPNetForSequenceClassification",Ze]],["albert",["AlbertForSequenceClassification",ct]],["distilbert",["DistilBertForSequenceClassification",Ne]],["roberta",["RobertaForSequenceClassification",Vt]],["xlm",["XLMForSequenceClassification",$t]],["xlm-roberta",["XLMRobertaForSequenceClassification",Jt]],["bart",["BartForSequenceClassification",kt]],["mbart",["MBartForSequenceClassification",vt]],["mobilebert",["MobileBertForSequenceClassification",Qe]],["squeezebert",["SqueezeBertForSequenceClassification",st]]]),ii=new Map([["bert",["BertForTokenClassification",Q]],["roformer",["RoFormerForTokenClassification",re]],["electra",["ElectraForTokenClassification",me]],["esm",["EsmForTokenClassification",We]],["convbert",["ConvBertForTokenClassification",le]],["camembert",["CamembertForTokenClassification",Te]],["deberta",["DebertaForTokenClassification",Pe]],["deberta-v2",["DebertaV2ForTokenClassification",Le]],["mpnet",["MPNetForTokenClassification",et]],["distilbert",["DistilBertForTokenClassification",De]],["roberta",["RobertaForTokenClassification",jt]],["xlm",["XLMForTokenClassification",Ut]],["xlm-roberta",["XLMRobertaForTokenClassification",Kt]]]),li=new Map([["t5",["T5ForConditionalGeneration",mt]],["longt5",["LongT5ForConditionalGeneration",gt]],["mt5",["MT5ForConditionalGeneration",Tt]],["bart",["BartForConditionalGeneration",yt]],["mbart",["MBartForConditionalGeneration",Ct]],["marian",["MarianMTModel",Ts]],["m2m_100",["M2M100ForConditionalGeneration",ys]],["blenderbot",["BlenderbotForConditionalGeneration",Lt]],["blenderbot-small",["BlenderbotSmallForConditionalGeneration",Bt]]]),ci=new Map([["bloom",["BloomForCausalLM",bn]],["gpt2",["GPT2LMHeadModel",Sr]],["jais",["JAISLMHeadModel",Lr]],["gptj",["GPTJForCausalLM",Rr]],["gpt_bigcode",["GPTBigCodeForCausalLM",Wr]],["gpt_neo",["GPTNeoForCausalLM",Br]],["gpt_neox",["GPTNeoXForCausalLM",Or]],["codegen",["CodeGenForCausalLM",Xr]],["llama",["LlamaForCausalLM",Yr]],["cohere",["CohereForCausalLM",Zr]],["gemma",["GemmaForCausalLM",rn]],["gemma2",["Gemma2ForCausalLM",sn]],["openelm",["OpenELMForCausalLM",cn]],["qwen2",["Qwen2ForCausalLM",pn]],["phi",["PhiForCausalLM",_n]],["phi3",["Phi3ForCausalLM",Mn]],["mpt",["MptForCausalLM",kn]],["opt",["OPTForCausalLM",Cn]],["mbart",["MBartForCausalLM",St]],["mistral",["MistralForCausalLM",pa]],["starcoder2",["Starcoder2ForCausalLM",_a]],["falcon",["FalconForCausalLM",Ma]],["trocr",["TrOCRForCausalLM",ca]],["stablelm",["StableLmForCausalLM",Ea]]]),di=new Map([["bert",["BertForMaskedLM",U]],["roformer",["RoFormerForMaskedLM",ee]],["electra",["ElectraForMaskedLM",pe]],["esm",["EsmForMaskedLM",Ge]],["convbert",["ConvBertForMaskedLM",ae]],["camembert",["CamembertForMaskedLM",Me]],["deberta",["DebertaForMaskedLM",ke]],["deberta-v2",["DebertaV2ForMaskedLM",Ae]],["mpnet",["MPNetForMaskedLM",Ke]],["albert",["AlbertForMaskedLM",ut]],["distilbert",["DistilBertForMaskedLM",Ve]],["roberta",["RobertaForMaskedLM",Ot]],["xlm",["XLMWithLMHeadModel",Wt]],["xlm-roberta",["XLMRobertaForMaskedLM",Yt]],["mobilebert",["MobileBertForMaskedLM",Xe]],["squeezebert",["SqueezeBertForMaskedLM",ot]]]),ui=new Map([["bert",["BertForQuestionAnswering",H]],["roformer",["RoFormerForQuestionAnswering",ne]],["electra",["ElectraForQuestionAnswering",_e]],["convbert",["ConvBertForQuestionAnswering",ce]],["camembert",["CamembertForQuestionAnswering",be]],["deberta",["DebertaForQuestionAnswering",Ce]],["deberta-v2",["DebertaV2ForQuestionAnswering",ze]],["mpnet",["MPNetForQuestionAnswering",tt]],["albert",["AlbertForQuestionAnswering",dt]],["distilbert",["DistilBertForQuestionAnswering",Oe]],["roberta",["RobertaForQuestionAnswering",Rt]],["xlm",["XLMForQuestionAnswering",Xt]],["xlm-roberta",["XLMRobertaForQuestionAnswering",Zt]],["mobilebert",["MobileBertForQuestionAnswering",He]],["squeezebert",["SqueezeBertForQuestionAnswering",at]]]),pi=new Map([["vision-encoder-decoder",["VisionEncoderDecoderModel",ar]]]),hi=new Map([["llava",["LlavaForConditionalGeneration",lr]],["moondream1",["Moondream1ForConditionalGeneration",cr]],["florence2",["Florence2ForConditionalGeneration",ur]]]),mi=new Map([["vision-encoder-decoder",["VisionEncoderDecoderModel",ar]]]),_i=new Map([["vit",["ViTForImageClassification",An]],["pvt",["PvtForImageClassification",zn]],["vit_msn",["ViTMSNForImageClassification",On]],["fastvit",["FastViTForImageClassification",qn]],["mobilevit",["MobileViTForImageClassification",Qn]],["mobilevitv2",["MobileViTV2ForImageClassification",Jn]],["beit",["BeitForImageClassification",ao]],["deit",["DeiTForImageClassification",ko]],["hiera",["HieraForImageClassification",Co]],["convnext",["ConvNextForImageClassification",os]],["convnextv2",["ConvNextV2ForImageClassification",is]],["dinov2",["Dinov2ForImageClassification",ds]],["resnet",["ResNetForImageClassification",Ao]],["swin",["SwinForImageClassification",zo]],["segformer",["SegformerForImageClassification",Ca]],["efficientnet",["EfficientNetForImageClassification",Ia]],["mobilenet_v1",["MobileNetV1ForImageClassification",Ra]],["mobilenet_v2",["MobileNetV2ForImageClassification",Wa]],["mobilenet_v3",["MobileNetV3ForImageClassification",Xa]],["mobilenet_v4",["MobileNetV4ForImageClassification",Ya]]]),fi=new Map([["detr",["DetrForObjectDetection",co]],["rt_detr",["RTDetrForObjectDetection",fo]],["table-transformer",["TableTransformerForObjectDetection",To]],["yolos",["YolosForObjectDetection",hs]]]),gi=new Map([["owlvit",["OwlViTForObjectDetection",eo]],["owlv2",["Owlv2ForObjectDetection",no]]]),Mi=new Map([["detr",["DetrForSegmentation",uo]],["clipseg",["CLIPSegForImageSegmentation",Pr]]]),wi=new Map([["segformer",["SegformerForSemanticSegmentation",va]],["sapiens",["SapiensForSemanticSegmentation",qo]]]),Ti=new Map([["detr",["DetrForSegmentation",uo]],["maskformer",["MaskFormerForInstanceSegmentation",Yo]]]),bi=new Map([["sam",["SamModel",fs]]]),xi=new Map([["wav2vec2",["Wav2Vec2ForCTC",Ps]],["wav2vec2-bert",["Wav2Vec2BertForCTC",$s]],["unispeech",["UniSpeechForCTC",Ns]],["unispeech-sat",["UniSpeechSatForCTC",js]],["wavlm",["WavLMForCTC",Zs]],["hubert",["HubertForCTC",Hs]]]),yi=new Map([["wav2vec2",["Wav2Vec2ForSequenceClassification",Cs]],["wav2vec2-bert",["Wav2Vec2BertForSequenceClassification",Us]],["unispeech",["UniSpeechForSequenceClassification",Ds]],["unispeech-sat",["UniSpeechSatForSequenceClassification",Rs]],["wavlm",["WavLMForSequenceClassification",ea]],["hubert",["HubertForSequenceClassification",Ys]],["audio-spectrogram-transformer",["ASTForAudioClassification",rr]]]),ki=new Map([["wavlm",["WavLMForXVector",ta]]]),Fi=new Map([["unispeech-sat",["UniSpeechSatForAudioFrameClassification",Gs]],["wavlm",["WavLMForAudioFrameClassification",ra]],["wav2vec2",["Wav2Vec2ForAudioFrameClassification",vs]],["pyannote",["PyAnnoteForAudioFrameClassification",Es]]]),Pi=new Map([["vitmatte",["VitMatteForImageMatting",$n]]]),Ci=new Map([["swin2sr",["Swin2SRForImageSuperResolution",No]]]),vi=new Map([["dpt",["DPTForDepthEstimation",Vo]],["depth_anything",["DepthAnythingForDepthEstimation",Ro]],["glpn",["GLPNForDepthEstimation",Zo]],["sapiens",["SapiensForDepthEstimation",Wo]],["depth_pro",["DepthProForDepthEstimation",Xo]]]),Si=new Map([["sapiens",["SapiensForNormalEstimation",$o]]]),Ai=new Map([["clip",["CLIPVisionModelWithProjection",gr]],["siglip",["SiglipVisionModel",br]]]),Ei=[[ei,M],[ti,w],[ri,x],[ai,M],[ii,M],[li,T],[ni,T],[ci,x],[di,M],[ui,M],[pi,b],[hi,k],[_i,M],[Mi,M],[Ti,M],[wi,M],[Pi,M],[Ci,M],[vi,M],[Si,M],[fi,M],[gi,M],[bi,y],[xi,M],[yi,M],[oi,T],[si,M],[ki,M],[Fi,M],[Ai,M]];for(const[e,t]of Ei)for(const[r,n]of e.values())P.set(r,t),v.set(n,r),C.set(r,n);const Li=[["MusicgenForConditionalGeneration",Oa,F],["CLIPTextModelWithProjection",_r,M],["SiglipTextModel",Tr,M],["ClapTextModelWithProjection",ba,M],["ClapAudioModelWithProjection",xa,M]];for(const[e,t,r]of Li)P.set(e,r),v.set(t,e),C.set(e,t);class zi extends Za{static MODEL_CLASS_MAPPINGS=Ei.map((e=>e[0]));static BASE_IF_FAIL=!0}class Ii extends Za{static MODEL_CLASS_MAPPINGS=[ai]}class Bi extends Za{static MODEL_CLASS_MAPPINGS=[ii]}class Ni extends Za{static MODEL_CLASS_MAPPINGS=[li]}class Di extends Za{static MODEL_CLASS_MAPPINGS=[ni]}class Oi extends Za{static MODEL_CLASS_MAPPINGS=[oi]}class Vi extends Za{static MODEL_CLASS_MAPPINGS=[si]}class ji extends Za{static MODEL_CLASS_MAPPINGS=[ci]}class Ri extends Za{static MODEL_CLASS_MAPPINGS=[di]}class Gi extends Za{static MODEL_CLASS_MAPPINGS=[ui]}class qi extends Za{static MODEL_CLASS_MAPPINGS=[pi]}class Wi extends Za{static MODEL_CLASS_MAPPINGS=[_i]}class $i extends Za{static MODEL_CLASS_MAPPINGS=[Mi]}class Ui extends Za{static MODEL_CLASS_MAPPINGS=[wi]}class Xi extends Za{static MODEL_CLASS_MAPPINGS=[Ti]}class Qi extends Za{static MODEL_CLASS_MAPPINGS=[fi]}class Hi extends Za{static MODEL_CLASS_MAPPINGS=[gi]}class Yi extends Za{static MODEL_CLASS_MAPPINGS=[bi]}class Ji extends Za{static MODEL_CLASS_MAPPINGS=[xi]}class Ki extends Za{static MODEL_CLASS_MAPPINGS=[yi]}class Zi extends Za{static MODEL_CLASS_MAPPINGS=[ki]}class el extends Za{static MODEL_CLASS_MAPPINGS=[Fi]}class tl extends Za{static MODEL_CLASS_MAPPINGS=[mi]}class rl extends Za{static MODEL_CLASS_MAPPINGS=[Pi]}class nl extends Za{static MODEL_CLASS_MAPPINGS=[Ci]}class ol extends Za{static MODEL_CLASS_MAPPINGS=[vi]}class sl extends Za{static MODEL_CLASS_MAPPINGS=[Si]}class al extends Za{static MODEL_CLASS_MAPPINGS=[Ai]}class il extends G{constructor({logits:e,past_key_values:t,encoder_outputs:r,decoder_attentions:n=null,cross_attentions:o=null}){super(),this.logits=e,this.past_key_values=t,this.encoder_outputs=r,this.decoder_attentions=n,this.cross_attentions=o}}class ll extends G{constructor({logits:e}){super(),this.logits=e}}class cl extends G{constructor({logits:e,embeddings:t}){super(),this.logits=e,this.embeddings=t}}class dl extends G{constructor({logits:e}){super(),this.logits=e}}class ul extends G{constructor({logits:e}){super(),this.logits=e}}class pl extends G{constructor({start_logits:e,end_logits:t}){super(),this.start_logits=e,this.end_logits=t}}class hl extends G{constructor({logits:e}){super(),this.logits=e}}class ml extends G{constructor({logits:e,past_key_values:t}){super(),this.logits=e,this.past_key_values=t}}class _l extends G{constructor({alphas:e}){super(),this.alphas=e}}class fl extends G{constructor({waveform:e,spectrogram:t}){super(),this.waveform=e,this.spectrogram=t}}},"./src/models/whisper/common_whisper.js":
+  \***********************/(e,t,r)=>{r.r(t),r.d(t,{ASTForAudioClassification:()=>nr,ASTModel:()=>rr,ASTPreTrainedModel:()=>tr,AlbertForMaskedLM:()=>pt,AlbertForQuestionAnswering:()=>ut,AlbertForSequenceClassification:()=>dt,AlbertModel:()=>ct,AlbertPreTrainedModel:()=>lt,AutoModel:()=>Di,AutoModelForAudioClassification:()=>rl,AutoModelForAudioFrameClassification:()=>ol,AutoModelForCTC:()=>tl,AutoModelForCausalLM:()=>Wi,AutoModelForDepthEstimation:()=>ll,AutoModelForDocumentQuestionAnswering:()=>sl,AutoModelForImageClassification:()=>Qi,AutoModelForImageFeatureExtraction:()=>dl,AutoModelForImageMatting:()=>al,AutoModelForImageSegmentation:()=>Hi,AutoModelForImageToImage:()=>il,AutoModelForMaskGeneration:()=>el,AutoModelForMaskedLM:()=>$i,AutoModelForNormalEstimation:()=>cl,AutoModelForObjectDetection:()=>Ki,AutoModelForQuestionAnswering:()=>Ui,AutoModelForSemanticSegmentation:()=>Yi,AutoModelForSeq2SeqLM:()=>ji,AutoModelForSequenceClassification:()=>Oi,AutoModelForSpeechSeq2Seq:()=>Ri,AutoModelForTextToSpectrogram:()=>Gi,AutoModelForTextToWaveform:()=>qi,AutoModelForTokenClassification:()=>Vi,AutoModelForUniversalSegmentation:()=>Ji,AutoModelForVision2Seq:()=>Xi,AutoModelForXVector:()=>nl,AutoModelForZeroShotObjectDetection:()=>Zi,BartForConditionalGeneration:()=>kt,BartForSequenceClassification:()=>Ft,BartModel:()=>yt,BartPretrainedModel:()=>xt,BaseModelOutput:()=>W,BeitForImageClassification:()=>uo,BeitModel:()=>co,BeitPreTrainedModel:()=>lo,BertForMaskedLM:()=>X,BertForQuestionAnswering:()=>Y,BertForSequenceClassification:()=>Q,BertForTokenClassification:()=>H,BertModel:()=>U,BertPreTrainedModel:()=>$,BlenderbotForConditionalGeneration:()=>zt,BlenderbotModel:()=>Lt,BlenderbotPreTrainedModel:()=>Et,BlenderbotSmallForConditionalGeneration:()=>Nt,BlenderbotSmallModel:()=>Bt,BlenderbotSmallPreTrainedModel:()=>It,BloomForCausalLM:()=>Fn,BloomModel:()=>kn,BloomPreTrainedModel:()=>yn,CLIPModel:()=>mr,CLIPPreTrainedModel:()=>hr,CLIPSegForImageSegmentation:()=>Cr,CLIPSegModel:()=>Pr,CLIPSegPreTrainedModel:()=>Fr,CLIPTextModel:()=>_r,CLIPTextModelWithProjection:()=>fr,CLIPVisionModel:()=>gr,CLIPVisionModelWithProjection:()=>Mr,CamembertForMaskedLM:()=>we,CamembertForQuestionAnswering:()=>xe,CamembertForSequenceClassification:()=>Te,CamembertForTokenClassification:()=>be,CamembertModel:()=>Me,CamembertPreTrainedModel:()=>ge,CausalLMOutput:()=>gl,CausalLMOutputWithPast:()=>Ml,ChineseCLIPModel:()=>kr,ChineseCLIPPreTrainedModel:()=>yr,ClapAudioModelWithProjection:()=>Pa,ClapModel:()=>ka,ClapPreTrainedModel:()=>ya,ClapTextModelWithProjection:()=>Fa,CodeGenForCausalLM:()=>Qr,CodeGenModel:()=>Xr,CodeGenPreTrainedModel:()=>Ur,CohereForCausalLM:()=>nn,CohereModel:()=>rn,CoherePreTrainedModel:()=>tn,ConvBertForMaskedLM:()=>ie,ConvBertForQuestionAnswering:()=>de,ConvBertForSequenceClassification:()=>le,ConvBertForTokenClassification:()=>ce,ConvBertModel:()=>ae,ConvBertPreTrainedModel:()=>se,ConvNextForImageClassification:()=>ls,ConvNextModel:()=>is,ConvNextPreTrainedModel:()=>as,ConvNextV2ForImageClassification:()=>us,ConvNextV2Model:()=>ds,ConvNextV2PreTrainedModel:()=>cs,DPTForDepthEstimation:()=>qo,DPTModel:()=>Go,DPTPreTrainedModel:()=>Ro,DebertaForMaskedLM:()=>Fe,DebertaForQuestionAnswering:()=>ve,DebertaForSequenceClassification:()=>Pe,DebertaForTokenClassification:()=>Ce,DebertaModel:()=>ke,DebertaPreTrainedModel:()=>ye,DebertaV2ForMaskedLM:()=>Ee,DebertaV2ForQuestionAnswering:()=>Ie,DebertaV2ForSequenceClassification:()=>Le,DebertaV2ForTokenClassification:()=>ze,DebertaV2Model:()=>Ae,DebertaV2PreTrainedModel:()=>Se,DecisionTransformerModel:()=>ri,DecisionTransformerPreTrainedModel:()=>ti,DeiTForImageClassification:()=>vo,DeiTModel:()=>Co,DeiTPreTrainedModel:()=>Po,DepthAnythingForDepthEstimation:()=>$o,DepthAnythingPreTrainedModel:()=>Wo,DepthProForDepthEstimation:()=>Jo,DepthProPreTrainedModel:()=>Yo,DetrForObjectDetection:()=>mo,DetrForSegmentation:()=>_o,DetrModel:()=>ho,DetrObjectDetectionOutput:()=>fo,DetrPreTrainedModel:()=>po,DetrSegmentationOutput:()=>go,Dinov2ForImageClassification:()=>ms,Dinov2Model:()=>hs,Dinov2PreTrainedModel:()=>ps,DistilBertForMaskedLM:()=>je,DistilBertForQuestionAnswering:()=>Ve,DistilBertForSequenceClassification:()=>De,DistilBertForTokenClassification:()=>Oe,DistilBertModel:()=>Ne,DistilBertPreTrainedModel:()=>Be,DonutSwinModel:()=>ss,DonutSwinPreTrainedModel:()=>os,EfficientNetForImageClassification:()=>Oa,EfficientNetModel:()=>Da,EfficientNetPreTrainedModel:()=>Na,ElectraForMaskedLM:()=>he,ElectraForQuestionAnswering:()=>fe,ElectraForSequenceClassification:()=>me,ElectraForTokenClassification:()=>_e,ElectraModel:()=>pe,ElectraPreTrainedModel:()=>ue,EsmForMaskedLM:()=>qe,EsmForSequenceClassification:()=>We,EsmForTokenClassification:()=>$e,EsmModel:()=>Ge,EsmPreTrainedModel:()=>Re,FalconForCausalLM:()=>xa,FalconModel:()=>ba,FalconPreTrainedModel:()=>Ta,FastViTForImageClassification:()=>Xn,FastViTModel:()=>Un,FastViTPreTrainedModel:()=>$n,Florence2ForConditionalGeneration:()=>pr,Florence2PreTrainedModel:()=>ur,GLPNForDepthEstimation:()=>ns,GLPNModel:()=>rs,GLPNPreTrainedModel:()=>ts,GPT2LMHeadModel:()=>Ar,GPT2Model:()=>Sr,GPT2PreTrainedModel:()=>vr,GPTBigCodeForCausalLM:()=>$r,GPTBigCodeModel:()=>Wr,GPTBigCodePreTrainedModel:()=>qr,GPTJForCausalLM:()=>Gr,GPTJModel:()=>Rr,GPTJPreTrainedModel:()=>jr,GPTNeoForCausalLM:()=>Nr,GPTNeoModel:()=>Br,GPTNeoPreTrainedModel:()=>Ir,GPTNeoXForCausalLM:()=>Vr,GPTNeoXModel:()=>Or,GPTNeoXPreTrainedModel:()=>Dr,Gemma2ForCausalLM:()=>dn,Gemma2Model:()=>cn,Gemma2PreTrainedModel:()=>ln,GemmaForCausalLM:()=>an,GemmaModel:()=>sn,GemmaPreTrainedModel:()=>on,GraniteForCausalLM:()=>en,GraniteModel:()=>Zr,GranitePreTrainedModel:()=>Kr,GroupViTModel:()=>Wn,GroupViTPreTrainedModel:()=>qn,HieraForImageClassification:()=>Eo,HieraModel:()=>Ao,HieraPreTrainedModel:()=>So,HubertForCTC:()=>Zs,HubertForSequenceClassification:()=>ea,HubertModel:()=>Ks,HubertPreTrainedModel:()=>Js,ImageMattingOutput:()=>wl,JAISLMHeadModel:()=>zr,JAISModel:()=>Lr,JAISPreTrainedModel:()=>Er,LlamaForCausalLM:()=>Jr,LlamaModel:()=>Yr,LlamaPreTrainedModel:()=>Hr,LlavaForConditionalGeneration:()=>cr,LlavaPreTrainedModel:()=>lr,LongT5ForConditionalGeneration:()=>Mt,LongT5Model:()=>gt,LongT5PreTrainedModel:()=>ft,M2M100ForConditionalGeneration:()=>Cs,M2M100Model:()=>Ps,M2M100PreTrainedModel:()=>Fs,MBartForCausalLM:()=>At,MBartForConditionalGeneration:()=>vt,MBartForSequenceClassification:()=>St,MBartModel:()=>Ct,MBartPreTrainedModel:()=>Pt,MPNetForMaskedLM:()=>Ze,MPNetForQuestionAnswering:()=>rt,MPNetForSequenceClassification:()=>et,MPNetForTokenClassification:()=>tt,MPNetModel:()=>Ke,MPNetPreTrainedModel:()=>Je,MT5ForConditionalGeneration:()=>bt,MT5Model:()=>Tt,MT5PreTrainedModel:()=>wt,MarianMTModel:()=>ks,MarianModel:()=>ys,MarianPreTrainedModel:()=>xs,MaskFormerForInstanceSegmentation:()=>es,MaskFormerModel:()=>Zo,MaskFormerPreTrainedModel:()=>Ko,MaskedLMOutput:()=>_l,MistralForCausalLM:()=>fa,MistralModel:()=>_a,MistralPreTrainedModel:()=>ma,MobileBertForMaskedLM:()=>Qe,MobileBertForQuestionAnswering:()=>Ye,MobileBertForSequenceClassification:()=>He,MobileBertModel:()=>Xe,MobileBertPreTrainedModel:()=>Ue,MobileNetV1ForImageClassification:()=>$a,MobileNetV1Model:()=>Wa,MobileNetV1PreTrainedModel:()=>qa,MobileNetV2ForImageClassification:()=>Qa,MobileNetV2Model:()=>Xa,MobileNetV2PreTrainedModel:()=>Ua,MobileNetV3ForImageClassification:()=>Ja,MobileNetV3Model:()=>Ya,MobileNetV3PreTrainedModel:()=>Ha,MobileNetV4ForImageClassification:()=>ei,MobileNetV4Model:()=>Za,MobileNetV4PreTrainedModel:()=>Ka,MobileViTForImageClassification:()=>Kn,MobileViTModel:()=>Jn,MobileViTPreTrainedModel:()=>Yn,MobileViTV2ForImageClassification:()=>to,MobileViTV2Model:()=>eo,MobileViTV2PreTrainedModel:()=>Zn,ModelOutput:()=>q,Moondream1ForConditionalGeneration:()=>dr,MptForCausalLM:()=>vn,MptModel:()=>Cn,MptPreTrainedModel:()=>Pn,MusicgenForCausalLM:()=>Ra,MusicgenForConditionalGeneration:()=>Ga,MusicgenModel:()=>ja,MusicgenPreTrainedModel:()=>Va,NomicBertModel:()=>K,NomicBertPreTrainedModel:()=>J,OPTForCausalLM:()=>En,OPTModel:()=>An,OPTPreTrainedModel:()=>Sn,OpenELMForCausalLM:()=>hn,OpenELMModel:()=>pn,OpenELMPreTrainedModel:()=>un,OwlViTForObjectDetection:()=>oo,OwlViTModel:()=>no,OwlViTPreTrainedModel:()=>ro,Owlv2ForObjectDetection:()=>io,Owlv2Model:()=>ao,Owlv2PreTrainedModel:()=>so,Phi3ForCausalLM:()=>xn,Phi3Model:()=>bn,Phi3PreTrainedModel:()=>Tn,PhiForCausalLM:()=>wn,PhiModel:()=>Mn,PhiPreTrainedModel:()=>gn,PreTrainedModel:()=>G,PretrainedMixin:()=>ni,PvtForImageClassification:()=>Dn,PvtModel:()=>Nn,PvtPreTrainedModel:()=>Bn,PyAnnoteForAudioFrameClassification:()=>Bs,PyAnnoteModel:()=>Is,PyAnnotePreTrainedModel:()=>zs,QuestionAnsweringModelOutput:()=>fl,Qwen2ForCausalLM:()=>fn,Qwen2Model:()=>_n,Qwen2PreTrainedModel:()=>mn,RTDetrForObjectDetection:()=>To,RTDetrModel:()=>wo,RTDetrObjectDetectionOutput:()=>bo,RTDetrPreTrainedModel:()=>Mo,ResNetForImageClassification:()=>Io,ResNetModel:()=>zo,ResNetPreTrainedModel:()=>Lo,RoFormerForMaskedLM:()=>te,RoFormerForQuestionAnswering:()=>oe,RoFormerForSequenceClassification:()=>re,RoFormerForTokenClassification:()=>ne,RoFormerModel:()=>ee,RoFormerPreTrainedModel:()=>Z,RobertaForMaskedLM:()=>Vt,RobertaForQuestionAnswering:()=>Gt,RobertaForSequenceClassification:()=>jt,RobertaForTokenClassification:()=>Rt,RobertaModel:()=>Ot,RobertaPreTrainedModel:()=>Dt,SamImageSegmentationOutput:()=>bs,SamModel:()=>Ts,SamPreTrainedModel:()=>ws,SapiensForDepthEstimation:()=>Qo,SapiensForNormalEstimation:()=>Ho,SapiensForSemanticSegmentation:()=>Xo,SapiensPreTrainedModel:()=>Uo,SegformerForImageClassification:()=>Ea,SegformerForSemanticSegmentation:()=>La,SegformerModel:()=>Aa,SegformerPreTrainedModel:()=>Sa,Seq2SeqLMOutput:()=>ul,SequenceClassifierOutput:()=>pl,SiglipModel:()=>Tr,SiglipPreTrainedModel:()=>wr,SiglipTextModel:()=>br,SiglipVisionModel:()=>xr,SpeechT5ForSpeechToText:()=>ca,SpeechT5ForTextToSpeech:()=>da,SpeechT5HifiGan:()=>ua,SpeechT5Model:()=>la,SpeechT5PreTrainedModel:()=>ia,SqueezeBertForMaskedLM:()=>st,SqueezeBertForQuestionAnswering:()=>it,SqueezeBertForSequenceClassification:()=>at,SqueezeBertModel:()=>ot,SqueezeBertPreTrainedModel:()=>nt,StableLmForCausalLM:()=>Ba,StableLmModel:()=>Ia,StableLmPreTrainedModel:()=>za,Starcoder2ForCausalLM:()=>wa,Starcoder2Model:()=>Ma,Starcoder2PreTrainedModel:()=>ga,Swin2SRForImageSuperResolution:()=>jo,Swin2SRModel:()=>Vo,Swin2SRPreTrainedModel:()=>Oo,SwinForImageClassification:()=>Do,SwinModel:()=>No,SwinPreTrainedModel:()=>Bo,T5ForConditionalGeneration:()=>_t,T5Model:()=>mt,T5PreTrainedModel:()=>ht,TableTransformerForObjectDetection:()=>ko,TableTransformerModel:()=>yo,TableTransformerObjectDetectionOutput:()=>Fo,TableTransformerPreTrainedModel:()=>xo,TokenClassifierOutput:()=>ml,TrOCRForCausalLM:()=>ha,TrOCRPreTrainedModel:()=>pa,UniSpeechForCTC:()=>js,UniSpeechForSequenceClassification:()=>Rs,UniSpeechModel:()=>Vs,UniSpeechPreTrainedModel:()=>Os,UniSpeechSatForAudioFrameClassification:()=>Us,UniSpeechSatForCTC:()=>Ws,UniSpeechSatForSequenceClassification:()=>$s,UniSpeechSatModel:()=>qs,UniSpeechSatPreTrainedModel:()=>Gs,ViTForImageClassification:()=>In,ViTMAEModel:()=>Vn,ViTMAEPreTrainedModel:()=>On,ViTMSNForImageClassification:()=>Gn,ViTMSNModel:()=>Rn,ViTMSNPreTrainedModel:()=>jn,ViTModel:()=>zn,ViTPreTrainedModel:()=>Ln,VisionEncoderDecoderModel:()=>ir,VitMatteForImageMatting:()=>Hn,VitMattePreTrainedModel:()=>Qn,VitsModel:()=>va,VitsModelOutput:()=>Tl,VitsPreTrainedModel:()=>Ca,Wav2Vec2BertForCTC:()=>Hs,Wav2Vec2BertForSequenceClassification:()=>Ys,Wav2Vec2BertModel:()=>Qs,Wav2Vec2BertPreTrainedModel:()=>Xs,Wav2Vec2ForAudioFrameClassification:()=>Ls,Wav2Vec2ForCTC:()=>As,Wav2Vec2ForSequenceClassification:()=>Es,Wav2Vec2Model:()=>Ss,Wav2Vec2PreTrainedModel:()=>vs,WavLMForAudioFrameClassification:()=>aa,WavLMForCTC:()=>na,WavLMForSequenceClassification:()=>oa,WavLMForXVector:()=>sa,WavLMModel:()=>ra,WavLMPreTrainedModel:()=>ta,WeSpeakerResNetModel:()=>Ds,WeSpeakerResNetPreTrainedModel:()=>Ns,WhisperForConditionalGeneration:()=>ar,WhisperModel:()=>sr,WhisperPreTrainedModel:()=>or,XLMForQuestionAnswering:()=>Qt,XLMForSequenceClassification:()=>Ut,XLMForTokenClassification:()=>Xt,XLMModel:()=>Wt,XLMPreTrainedModel:()=>qt,XLMRobertaForMaskedLM:()=>Jt,XLMRobertaForQuestionAnswering:()=>er,XLMRobertaForSequenceClassification:()=>Kt,XLMRobertaForTokenClassification:()=>Zt,XLMRobertaModel:()=>Yt,XLMRobertaPreTrainedModel:()=>Ht,XLMWithLMHeadModel:()=>$t,XVectorOutput:()=>hl,YolosForObjectDetection:()=>gs,YolosModel:()=>fs,YolosObjectDetectionOutput:()=>Ms,YolosPreTrainedModel:()=>_s});var n=r(/*! ./configs.js */"./src/configs.js"),o=r(/*! ./backends/onnx.js */"./src/backends/onnx.js"),s=r(/*! ./utils/dtypes.js */"./src/utils/dtypes.js"),a=r(/*! ./utils/generic.js */"./src/utils/generic.js"),i=r(/*! ./utils/core.js */"./src/utils/core.js"),l=r(/*! ./utils/hub.js */"./src/utils/hub.js"),c=r(/*! ./generation/logits_process.js */"./src/generation/logits_process.js"),d=r(/*! ./generation/configuration_utils.js */"./src/generation/configuration_utils.js"),u=r(/*! ./utils/tensor.js */"./src/utils/tensor.js"),p=r(/*! ./utils/maths.js */"./src/utils/maths.js"),h=r(/*! ./generation/stopping_criteria.js */"./src/generation/stopping_criteria.js"),m=r(/*! ./generation/logits_sampler.js */"./src/generation/logits_sampler.js"),_=r(/*! ./env.js */"./src/env.js"),f=r(/*! ./models/whisper/generation_whisper.js */"./src/models/whisper/generation_whisper.js"),g=r(/*! ./models/whisper/common_whisper.js */"./src/models/whisper/common_whisper.js");const M=0,w=1,T=2,b=3,x=4,y=5,k=6,F=7,P=new Map,C=new Map,v=new Map;async function S(e,t,r){return Object.fromEntries(await Promise.all(Object.keys(t).map((async a=>{const{buffer:i,session_options:c}=await async function(e,t,r){const a=r.config?.["transformers.js_config"]??{};let i=r.device??a.device;i&&"string"!=typeof i&&(i.hasOwnProperty(t)?i=i[t]:(console.warn(`device not specified for "${t}". Using the default device.`),i=null));const c=i??(_.apis.IS_NODE_ENV?"cpu":"wasm"),d=(0,o.deviceToExecutionProviders)(c);let u=r.dtype??a.dtype;"string"!=typeof u&&(u&&u.hasOwnProperty(t)?u=u[t]:(u=s.DEFAULT_DEVICE_DTYPE_MAPPING[c]??s.DATA_TYPES.fp32,console.warn(`dtype not specified for "${t}". Using the default dtype (${u}) for this device (${c}).`)));const p=u;if(!s.DEFAULT_DTYPE_SUFFIX_MAPPING.hasOwnProperty(p))throw new Error(`Invalid dtype: ${p}. Should be one of: ${Object.keys(s.DATA_TYPES).join(", ")}`);if(p===s.DATA_TYPES.fp16&&"webgpu"===c&&!await(0,s.isWebGpuFp16Supported)())throw new Error(`The device (${c}) does not support fp16.`);const h=s.DEFAULT_DTYPE_SUFFIX_MAPPING[p],m=`${r.subfolder??""}/${t}${h}.onnx`,f={...r.session_options};f.executionProviders??=d;const g=a.free_dimension_overrides;g?f.freeDimensionOverrides??=g:c.startsWith("webnn")&&!f.freeDimensionOverrides&&console.warn('WebNN does not currently support dynamic shapes and requires `free_dimension_overrides` to be set in config.json as a field within "transformers.js_config". When `free_dimension_overrides` is not set, you may experience significant performance degradation.');const M=(0,l.getModelFile)(e,m,!0,r),w=r.use_external_data_format??a.use_external_data_format;let T=[];if(w&&(!0===w||"object"==typeof w&&w.hasOwnProperty(t)&&!0===w[t])){if(_.apis.IS_NODE_ENV)throw new Error("External data format is not yet supported in Node.js");const n=`${t}${h}.onnx_data`,o=`${r.subfolder??""}/${n}`;T.push(new Promise((async(t,s)=>{const a=await(0,l.getModelFile)(e,o,!0,r);t({path:n,data:a})})))}else void 0!==f.externalData&&(T=f.externalData.map((async t=>{if("string"==typeof t.data){const n=await(0,l.getModelFile)(e,t.data,!0,r);return{...t,data:n}}return t})));if(T.length>0&&(f.externalData=await Promise.all(T)),"webgpu"===c){const e=(0,n.getKeyValueShapes)(r.config,{prefix:"present"});if(Object.keys(e).length>0&&!(0,o.isONNXProxy)()){const t={};for(const r in e)t[r]="gpu-buffer";f.preferredOutputLocation=t}}return{buffer:await M,session_options:f}}(e,t[a],r);return[a,await(0,o.createInferenceSession)(i,c)]}))))}async function A(e,t,r){return Object.fromEntries(await Promise.all(Object.keys(t).map((async n=>[n,await(0,l.getModelJSON)(e,t[n],!1,r)]))))}async function E(e,t){const r=function(e,t){const r=Object.create(null),n=[];for(const s of e.inputNames){const e=t[s];e instanceof u.Tensor?r[s]=(0,o.isONNXProxy)()?e.clone():e:n.push(s)}if(n.length>0)throw new Error(`An error occurred during model execution: "Missing the following inputs: ${n.join(", ")}.`);const s=Object.keys(t).length,a=e.inputNames.length;if(s>a){let r=Object.keys(t).filter((t=>!e.inputNames.includes(t)));console.warn(`WARNING: Too many inputs were provided (${s} > ${a}). The following inputs will be ignored: "${r.join(", ")}".`)}return r}(e,t);try{const t=Object.fromEntries(Object.entries(r).map((([e,t])=>[e,t.ort_tensor])));let n=await e.run(t);return n=L(n),n}catch(e){throw console.error(`An error occurred during model execution: "${e}".`),console.error("Inputs given to model:",r),e}}function L(e){for(let t in e)(0,o.isONNXTensor)(e[t])?e[t]=new u.Tensor(e[t]):"object"==typeof e[t]&&L(e[t]);return e}function z(e){if(e instanceof u.Tensor)return e;if(0===e.length)throw Error("items must be non-empty");if(Array.isArray(e[0])){if(e.some((t=>t.length!==e[0].length)))throw Error("Unable to create tensor, you should probably activate truncation and/or padding with 'padding=True' and/or 'truncation=True' to have batched tensors with the same length.");return new u.Tensor("int64",BigInt64Array.from(e.flat().map((e=>BigInt(e)))),[e.length,e[0].length])}return new u.Tensor("int64",BigInt64Array.from(e.map((e=>BigInt(e)))),[1,e.length])}function I(e){return new u.Tensor("bool",[e],[1])}async function B(e,t){let{encoder_outputs:r,input_ids:n,decoder_input_ids:o,...s}=t;if(!r){const n=(0,i.pick)(t,e.sessions.model.inputNames);r=(await N(e,n)).last_hidden_state}s.input_ids=o,s.encoder_hidden_states=r,e.sessions.decoder_model_merged.inputNames.includes("encoder_attention_mask")&&(s.encoder_attention_mask=t.attention_mask);return await D(e,s,!0)}async function N(e,t){const r=e.sessions.model,n=(0,i.pick)(t,r.inputNames);if(r.inputNames.includes("inputs_embeds")&&!n.inputs_embeds){if(!t.input_ids)throw new Error("Both `input_ids` and `inputs_embeds` are missing in the model inputs.");n.inputs_embeds=await e.encode_text({input_ids:t.input_ids})}return r.inputNames.includes("token_type_ids")&&!n.token_type_ids&&(n.token_type_ids=new u.Tensor("int64",new BigInt64Array(n.input_ids.data.length),n.input_ids.dims)),await E(r,n)}async function D(e,t,r=!1){const n=e.sessions[r?"decoder_model_merged":"model"],{past_key_values:o,...s}=t;n.inputNames.includes("use_cache_branch")&&(s.use_cache_branch=I(!!o)),n.inputNames.includes("position_ids")&&s.attention_mask&&!s.position_ids&&(s.position_ids=function(e,t=null){const{input_ids:r,inputs_embeds:n,attention_mask:o}=e,[s,a]=o.dims,i=new BigInt64Array(o.data.length);for(let e=0;e<s;++e){const t=e*a;let r=BigInt(0);for(let e=0;e<a;++e){const n=t+e;0n===o.data[n]?i[n]=BigInt(1):(i[n]=r,r+=o.data[n])}}let l=new u.Tensor("int64",i,o.dims);if(t){const e=-(r??n).dims.at(1);l=l.slice(null,[e,null])}return l}(s,o)),e.addPastKeyValues(s,o);const a=(0,i.pick)(s,n.inputNames);return await E(n,a)}async function O(e,{input_ids:t=null,attention_mask:r=null,pixel_values:n=null,position_ids:o=null,inputs_embeds:s=null,past_key_values:a=null,generation_config:i=null,logits_processor:l=null,...c}){if(!s)if(s=await e.encode_text({input_ids:t}),n&&1!==t.dims[1]){const o=await e.encode_image({pixel_values:n});({inputs_embeds:s,attention_mask:r}=e._merge_input_ids_with_image_features({image_features:o,inputs_embeds:s,input_ids:t,attention_mask:r}))}else if(a&&n&&1===t.dims[1]){const e=t.dims[1],n=Object.values(a)[0].dims.at(-2);r=(0,u.cat)([(0,u.ones)([t.dims[0],n]),r.slice(null,[r.dims[1]-e,r.dims[1]])],1)}return await D(e,{inputs_embeds:s,past_key_values:a,attention_mask:r,position_ids:o,generation_config:i,logits_processor:l},!0)}function V(e,t,r,n){if(r.past_key_values){const t=Object.values(r.past_key_values)[0].dims.at(-2),{input_ids:n,attention_mask:o}=r;if(o&&o.dims[1]>n.dims[1]);else if(t<n.dims[1])r.input_ids=n.slice(null,[t,null]);else if(null!=e.config.image_token_index&&n.data.some((t=>t==e.config.image_token_index))){const o=e.config.num_image_tokens;if(!o)throw new Error("`num_image_tokens` is missing in the model configuration.");const s=n.dims[1]-(t-o);r.input_ids=n.slice(null,[-s,null]),r.attention_mask=(0,u.ones)([1,t+s])}}return r}function j(e,t,r,n){return r.past_key_values&&(t=t.map((e=>[e.at(-1)]))),{...r,decoder_input_ids:z(t)}}function R(e,...t){return e.config.is_encoder_decoder?j(e,...t):V(e,...t)}class G extends a.Callable{main_input_name="input_ids";forward_params=["input_ids","attention_mask"];constructor(e,t,r){super(),this.config=e,this.sessions=t,this.configs=r;const n=v.get(this.constructor),o=P.get(n);switch(this.can_generate=!1,this._forward=null,this._prepare_inputs_for_generation=null,o){case x:this.can_generate=!0,this._forward=D,this._prepare_inputs_for_generation=V;break;case T:case b:case F:this.can_generate=!0,this._forward=B,this._prepare_inputs_for_generation=j;break;case w:this._forward=B;break;case k:this.can_generate=!0,this._forward=O,this._prepare_inputs_for_generation=R;break;default:this._forward=N}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:r=null,cache_dir:o=null,local_files_only:s=!1,revision:a="main",model_file_name:i=null,subfolder:l="onnx",device:c=null,dtype:d=null,use_external_data_format:u=null,session_options:p={}}={}){let h={progress_callback:t,config:r,cache_dir:o,local_files_only:s,revision:a,model_file_name:i,subfolder:l,device:c,dtype:d,use_external_data_format:u,session_options:p};const m=v.get(this),_=P.get(m);let f;if(r=h.config=await n.AutoConfig.from_pretrained(e,h),_===x)f=await Promise.all([S(e,{model:h.model_file_name??"model"},h),A(e,{generation_config:"generation_config.json"},h)]);else if(_===T||_===b)f=await Promise.all([S(e,{model:"encoder_model",decoder_model_merged:"decoder_model_merged"},h),A(e,{generation_config:"generation_config.json"},h)]);else if(_===y)f=await Promise.all([S(e,{model:"vision_encoder",prompt_encoder_mask_decoder:"prompt_encoder_mask_decoder"},h)]);else if(_===w)f=await Promise.all([S(e,{model:"encoder_model",decoder_model_merged:"decoder_model_merged"},h)]);else if(_===k){const t={embed_tokens:"embed_tokens",vision_encoder:"vision_encoder",decoder_model_merged:"decoder_model_merged"};r.is_encoder_decoder&&(t.model="encoder_model"),f=await Promise.all([S(e,t,h),A(e,{generation_config:"generation_config.json"},h)])}else _===F?f=await Promise.all([S(e,{model:"text_encoder",decoder_model_merged:"decoder_model_merged",encodec_decode:"encodec_decode"},h),A(e,{generation_config:"generation_config.json"},h)]):(_!==M&&console.warn(`Model type for '${m??r?.model_type}' not found, assuming encoder-only architecture. Please report this at https://github.com/xenova/transformers.js/issues/new/choose.`),f=await Promise.all([S(e,{model:h.model_file_name??"model"},h)]));return new this(r,...f)}async _call(e){return await this.forward(e)}async forward(e){return await this._forward(this,e)}get generation_config(){return this.configs?.generation_config??null}_get_logits_warper(e){const t=new c.LogitsProcessorList;return null!==e.temperature&&1!==e.temperature&&t.push(new c.TemperatureLogitsWarper(e.temperature)),null!==e.top_k&&0!==e.top_k&&t.push(new c.TopKLogitsWarper(e.top_k)),null!==e.top_p&&e.top_p<1&&t.push(new c.TopPLogitsWarper(e.top_p)),t}_get_logits_processor(e,t,r=null){const n=new c.LogitsProcessorList;if(null!==e.repetition_penalty&&1!==e.repetition_penalty&&n.push(new c.RepetitionPenaltyLogitsProcessor(e.repetition_penalty)),null!==e.no_repeat_ngram_size&&e.no_repeat_ngram_size>0&&n.push(new c.NoRepeatNGramLogitsProcessor(e.no_repeat_ngram_size)),null!==e.bad_words_ids&&n.push(new c.NoBadWordsLogitsProcessor(e.bad_words_ids,e.eos_token_id)),null!==e.min_length&&null!==e.eos_token_id&&e.min_length>0&&n.push(new c.MinLengthLogitsProcessor(e.min_length,e.eos_token_id)),null!==e.min_new_tokens&&null!==e.eos_token_id&&e.min_new_tokens>0&&n.push(new c.MinNewTokensLengthLogitsProcessor(t,e.min_new_tokens,e.eos_token_id)),null!==e.forced_bos_token_id&&n.push(new c.ForcedBOSTokenLogitsProcessor(e.forced_bos_token_id)),null!==e.forced_eos_token_id&&n.push(new c.ForcedEOSTokenLogitsProcessor(e.max_length,e.forced_eos_token_id)),null!==e.begin_suppress_tokens){const r=t>1||null===e.forced_bos_token_id?t:t+1;n.push(new c.SuppressTokensAtBeginLogitsProcessor(e.begin_suppress_tokens,r))}return null!==e.guidance_scale&&e.guidance_scale>1&&n.push(new c.ClassifierFreeGuidanceLogitsProcessor(e.guidance_scale)),null!==r&&n.extend(r),n}_prepare_generation_config(e,t,r=d.GenerationConfig){const n={...this.config};for(const e of["decoder","generator","text_config"])e in n&&Object.assign(n,n[e]);const o=new r(n);return Object.assign(o,this.generation_config??{}),e&&Object.assign(o,e),t&&Object.assign(o,(0,i.pick)(t,Object.getOwnPropertyNames(o))),o}_get_stopping_criteria(e,t=null){const r=new h.StoppingCriteriaList;return null!==e.max_length&&r.push(new h.MaxLengthCriteria(e.max_length,this.config.max_position_embeddings??null)),null!==e.eos_token_id&&r.push(new h.EosTokenCriteria(e.eos_token_id)),t&&r.extend(t),r}_validate_model_class(){if(!this.can_generate){const e=[hi,fi,pi,ii],t=v.get(this.constructor),r=new Set,n=this.config.model_type;for(const t of e){const e=t.get(n);e&&r.add(e[0])}let o=`The current model class (${t}) is not compatible with \`.generate()\`, as it doesn't have a language model head.`;throw r.size>0&&(o+=` Please use the following class instead: ${[...r].join(", ")}`),Error(o)}}prepare_inputs_for_generation(...e){return this._prepare_inputs_for_generation(this,...e)}_update_model_kwargs_for_generation({generated_input_ids:e,outputs:t,model_inputs:r,is_encoder_decoder:n}){return r.past_key_values=this.getPastKeyValues(t,r.past_key_values),r.input_ids=new u.Tensor("int64",e.flat(),[e.length,1]),n||(r.attention_mask=(0,u.cat)([r.attention_mask,(0,u.ones)([r.attention_mask.dims[0],1])],1)),r.position_ids=null,r}_prepare_model_inputs({inputs:e,bos_token_id:t,model_kwargs:r}){const n=(0,i.pick)(r,this.forward_params),o=this.main_input_name;if(o in n){if(e)throw new Error("`inputs`: {inputs}` were passed alongside {input_name} which is not allowed. Make sure to either pass {inputs} or {input_name}=...")}else n[o]=e;return{inputs_tensor:n[o],model_inputs:n,model_input_name:o}}async _prepare_encoder_decoder_kwargs_for_generation({inputs_tensor:e,model_inputs:t,model_input_name:r,generation_config:n}){if(this.sessions.model.inputNames.includes("inputs_embeds")&&!t.inputs_embeds&&"_prepare_inputs_embeds"in this){const{input_ids:e,pixel_values:r,attention_mask:n,...o}=t,s=await this._prepare_inputs_embeds(t);t={...o,...(0,i.pick)(s,["inputs_embeds","attention_mask"])}}let{last_hidden_state:o}=await N(this,t);if(null!==n.guidance_scale&&n.guidance_scale>1)o=(0,u.cat)([o,(0,u.full_like)(o,0)],0),"attention_mask"in t&&(t.attention_mask=(0,u.cat)([t.attention_mask,(0,u.zeros_like)(t.attention_mask)],0));else if(t.decoder_input_ids){const e=z(t.decoder_input_ids).dims[0];if(e!==o.dims[0]){if(1!==o.dims[0])throw new Error(`The encoder outputs have a different batch size (${o.dims[0]}) than the decoder inputs (${e}).`);o=(0,u.cat)(Array.from({length:e},(()=>o)),0)}}return t.encoder_outputs=o,t}_prepare_decoder_input_ids_for_generation({batch_size:e,model_input_name:t,model_kwargs:r,decoder_start_token_id:n,bos_token_id:o,generation_config:s}){let{decoder_input_ids:a,...i}=r;if(a)Array.isArray(a[0])||(a=Array.from({length:e},(()=>a)));else if(n??=o,"musicgen"===this.config.model_type)a=Array.from({length:e*this.config.decoder.num_codebooks},(()=>[n]));else if(Array.isArray(n)){if(n.length!==e)throw new Error(`\`decoder_start_token_id\` expcted to have length ${e} but got ${n.length}`);a=n}else a=Array.from({length:e},(()=>[n]));return a=z(a),r.decoder_attention_mask=(0,u.ones_like)(a),{input_ids:a,model_inputs:i}}async generate({inputs:e=null,generation_config:t=null,logits_processor:r=null,stopping_criteria:n=null,streamer:o=null,...s}){this._validate_model_class(),t=this._prepare_generation_config(t,s);let{inputs_tensor:a,model_inputs:i,model_input_name:l}=this._prepare_model_inputs({inputs:e,model_kwargs:s});const c=this.config.is_encoder_decoder;let d;c&&("encoder_outputs"in i||(i=await this._prepare_encoder_decoder_kwargs_for_generation({inputs_tensor:a,model_inputs:i,model_input_name:l,generation_config:t}))),c?({input_ids:d,model_inputs:i}=this._prepare_decoder_input_ids_for_generation({batch_size:i[l].dims.at(0),model_input_name:l,model_kwargs:i,decoder_start_token_id:t.decoder_start_token_id,bos_token_id:t.bos_token_id,generation_config:t})):d=i[l];let p=d.dims.at(-1);null!==t.max_new_tokens&&(t.max_length=p+t.max_new_tokens);const h=this._get_logits_processor(t,p,r),_=this._get_stopping_criteria(t,n),f=i[l].dims.at(0),g=m.LogitsSampler.getSampler(t),M=new Array(f).fill(0),w=d.tolist();let T;o&&o.put(w);let b={};for(;;){if(i=this.prepare_inputs_for_generation(w,i,t),T=await this.forward(i),t.output_attentions&&t.return_dict_in_generate){const e=this.getAttentions(T);for(const t in e)t in b||(b[t]=[]),b[t].push(e[t])}const e=h(w,T.logits.slice(null,-1,null)),r=[];for(let t=0;t<e.dims.at(0);++t){const n=e[t],o=await g(n);for(const[e,n]of o){const o=BigInt(e);M[t]+=n,w[t].push(o),r.push([o]);break}}o&&o.put(r);if(_(w).every((e=>e)))break;i=this._update_model_kwargs_for_generation({generated_input_ids:r,outputs:T,model_inputs:i,is_encoder_decoder:c})}o&&o.end();const x=this.getPastKeyValues(T,i.past_key_values,!0),y=new u.Tensor("int64",w.flat(),[w.length,w[0].length]);if(t.return_dict_in_generate)return{sequences:y,past_key_values:x,...b};for(const e of Object.values(T))"gpu-buffer"===e.location&&e.dispose();return y}getPastKeyValues(e,t,r=!1){const n=Object.create(null);for(const o in e)if(o.startsWith("present")){const s=o.replace("present","past_key_values"),a=o.includes("encoder");if(n[s]=a&&t?t[s]:e[o],t&&(!a||r)){const e=t[s];"gpu-buffer"===e.location&&e.dispose()}}return n}getAttentions(e){const t={};for(const r of["cross_attentions","encoder_attentions","decoder_attentions"])for(const n in e)n.startsWith(r)&&(r in t||(t[r]=[]),t[r].push(e[n]));return t}addPastKeyValues(e,t){if(t)Object.assign(e,t);else{const t=this.custom_config.kv_cache_dtype??"float32",r="float16"===t?new Uint16Array:[],o=(0,n.getKeyValueShapes)(this.config);for(const n in o)e[n]=new u.Tensor(t,r,o[n])}}async encode_image({pixel_values:e}){const t=(await E(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 E(this.sessions.embed_tokens,{input_ids:e})).inputs_embeds}}class q{}class W extends q{constructor({last_hidden_state:e,hidden_states:t=null,attentions:r=null}){super(),this.last_hidden_state=e,this.hidden_states=t,this.attentions=r}}class $ extends G{}class U extends ${}class X extends ${async _call(e){return new _l(await super._call(e))}}class Q extends ${async _call(e){return new pl(await super._call(e))}}class H extends ${async _call(e){return new ml(await super._call(e))}}class Y extends ${async _call(e){return new fl(await super._call(e))}}class J extends G{}class K extends J{}class Z extends G{}class ee extends Z{}class te extends Z{async _call(e){return new _l(await super._call(e))}}class re extends Z{async _call(e){return new pl(await super._call(e))}}class ne extends Z{async _call(e){return new ml(await super._call(e))}}class oe extends Z{async _call(e){return new fl(await super._call(e))}}class se extends G{}class ae extends se{}class ie extends se{async _call(e){return new _l(await super._call(e))}}class le extends se{async _call(e){return new pl(await super._call(e))}}class ce extends se{async _call(e){return new ml(await super._call(e))}}class de extends se{async _call(e){return new fl(await super._call(e))}}class ue extends G{}class pe extends ue{}class he extends ue{async _call(e){return new _l(await super._call(e))}}class me extends ue{async _call(e){return new pl(await super._call(e))}}class _e extends ue{async _call(e){return new ml(await super._call(e))}}class fe extends ue{async _call(e){return new fl(await super._call(e))}}class ge extends G{}class Me extends ge{}class we extends ge{async _call(e){return new _l(await super._call(e))}}class Te extends ge{async _call(e){return new pl(await super._call(e))}}class be extends ge{async _call(e){return new ml(await super._call(e))}}class xe extends ge{async _call(e){return new fl(await super._call(e))}}class ye extends G{}class ke extends ye{}class Fe extends ye{async _call(e){return new _l(await super._call(e))}}class Pe extends ye{async _call(e){return new pl(await super._call(e))}}class Ce extends ye{async _call(e){return new ml(await super._call(e))}}class ve extends ye{async _call(e){return new fl(await super._call(e))}}class Se extends G{}class Ae extends Se{}class Ee extends Se{async _call(e){return new _l(await super._call(e))}}class Le extends Se{async _call(e){return new pl(await super._call(e))}}class ze extends Se{async _call(e){return new ml(await super._call(e))}}class Ie extends Se{async _call(e){return new fl(await super._call(e))}}class Be extends G{}class Ne extends Be{}class De extends Be{async _call(e){return new pl(await super._call(e))}}class Oe extends Be{async _call(e){return new ml(await super._call(e))}}class Ve extends Be{async _call(e){return new fl(await super._call(e))}}class je extends Be{async _call(e){return new _l(await super._call(e))}}class Re extends G{}class Ge extends Re{}class qe extends Re{async _call(e){return new _l(await super._call(e))}}class We extends Re{async _call(e){return new pl(await super._call(e))}}class $e extends Re{async _call(e){return new ml(await super._call(e))}}class Ue extends G{}class Xe extends Ue{}class Qe extends Ue{async _call(e){return new _l(await super._call(e))}}class He extends Ue{async _call(e){return new pl(await super._call(e))}}class Ye extends Ue{async _call(e){return new fl(await super._call(e))}}class Je extends G{}class Ke extends Je{}class Ze extends Je{async _call(e){return new _l(await super._call(e))}}class et extends Je{async _call(e){return new pl(await super._call(e))}}class tt extends Je{async _call(e){return new ml(await super._call(e))}}class rt extends Je{async _call(e){return new fl(await super._call(e))}}class nt extends G{}class ot extends nt{}class st extends nt{async _call(e){return new _l(await super._call(e))}}class at extends nt{async _call(e){return new pl(await super._call(e))}}class it extends nt{async _call(e){return new fl(await super._call(e))}}class lt extends G{}class ct extends lt{}class dt extends lt{async _call(e){return new pl(await super._call(e))}}class ut extends lt{async _call(e){return new fl(await super._call(e))}}class pt extends lt{async _call(e){return new _l(await super._call(e))}}class ht extends G{forward_params=["input_ids","attention_mask","encoder_outputs","decoder_input_ids","decoder_attention_mask","past_key_values"]}class mt extends ht{}class _t extends ht{}class ft extends G{}class gt extends ft{}class Mt extends ft{}class wt extends G{}class Tt extends wt{}class bt extends wt{}class xt extends G{}class yt extends xt{}class kt extends xt{}class Ft extends xt{async _call(e){return new pl(await super._call(e))}}class Pt extends G{}class Ct extends Pt{}class vt extends Pt{}class St extends Pt{async _call(e){return new pl(await super._call(e))}}class At extends Pt{}class Et extends G{}class Lt extends Et{}class zt extends Et{}class It extends G{}class Bt extends It{}class Nt extends It{}class Dt extends G{}class Ot extends Dt{}class Vt extends Dt{async _call(e){return new _l(await super._call(e))}}class jt extends Dt{async _call(e){return new pl(await super._call(e))}}class Rt extends Dt{async _call(e){return new ml(await super._call(e))}}class Gt extends Dt{async _call(e){return new fl(await super._call(e))}}class qt extends G{}class Wt extends qt{}class $t extends qt{async _call(e){return new _l(await super._call(e))}}class Ut extends qt{async _call(e){return new pl(await super._call(e))}}class Xt extends qt{async _call(e){return new ml(await super._call(e))}}class Qt extends qt{async _call(e){return new fl(await super._call(e))}}class Ht extends G{}class Yt extends Ht{}class Jt extends Ht{async _call(e){return new _l(await super._call(e))}}class Kt extends Ht{async _call(e){return new pl(await super._call(e))}}class Zt extends Ht{async _call(e){return new ml(await super._call(e))}}class er extends Ht{async _call(e){return new fl(await super._call(e))}}class tr extends G{}class rr extends tr{}class nr extends tr{}class or extends G{requires_attention_mask=!1;main_input_name="input_features";forward_params=["input_features","attention_mask","decoder_input_ids","decoder_attention_mask","past_key_values"]}class sr extends or{}class ar extends or{_prepare_generation_config(e,t){return super._prepare_generation_config(e,t,f.WhisperGenerationConfig)}_retrieve_init_tokens(e){const t=[e.decoder_start_token_id];let r=e.language;const n=e.task;if(e.is_multilingual){r||(console.warn("No language specified - defaulting to English (en)."),r="en");const o=`<|${(0,g.whisper_language_to_code)(r)}|>`;t.push(e.lang_to_id[o]),t.push(e.task_to_id[n??"transcribe"])}else if(r||n)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:r=null,stopping_criteria:n=null,...o}){t=this._prepare_generation_config(t,o);const s=o.decoder_input_ids??this._retrieve_init_tokens(t);if(t.return_timestamps&&(r??=new c.LogitsProcessorList,r.push(new c.WhisperTimeStampLogitsProcessor(t,s))),t.begin_suppress_tokens&&(r??=new c.LogitsProcessorList,r.push(new c.SuppressTokensAtBeginLogitsProcessor(t.begin_suppress_tokens,s.length))),t.return_token_timestamps){if(!t.alignment_heads)throw new Error("Model generation config has no `alignment_heads`, token-level timestamps not available. See https://gist.github.com/hollance/42e32852f24243b748ae6bc1f985b13a on how to add this property to the generation config.");"translate"===t.task&&console.warn("Token-level timestamps may not be reliable for task 'translate'."),t.output_attentions=!0,t.return_dict_in_generate=!0}const a=await super.generate({inputs:e,generation_config:t,logits_processor:r,decoder_input_ids:s,...o});return t.return_token_timestamps&&(a.token_timestamps=this._extract_token_timestamps(a,t.alignment_heads,t.num_frames)),a}_extract_token_timestamps(e,t,r=null,n=.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==r&&console.warn("`num_frames` has not been set, meaning the entire audio will be analyzed. This may lead to inaccurate token-level timestamps for short audios (< 30 seconds).");let o=this.config.median_filter_width;void 0===o&&(console.warn("Model config has no `median_filter_width`, using default value of 7."),o=7);const s=e.cross_attentions,a=Array.from({length:this.config.decoder_layers},((e,t)=>(0,u.cat)(s.map((e=>e[t])),2))),l=(0,u.stack)(t.map((([e,t])=>{if(e>=a.length)throw new Error(`Layer index ${e} is out of bounds for cross attentions (length ${a.length}).`);return r?a[e].slice(null,t,null,[0,r]):a[e].slice(null,t)}))).transpose(1,0,2,3),[c,d]=(0,u.std_mean)(l,-2,0,!0),h=l.clone();for(let e=0;e<h.dims[0];++e){const t=h[e];for(let r=0;r<t.dims[0];++r){const n=t[r],s=c[e][r][0].data,a=d[e][r][0].data;for(let e=0;e<n.dims[0];++e){let t=n[e].data;for(let e=0;e<t.length;++e)t[e]=(t[e]-a[e])/s[e];t.set((0,p.medianFilter)(t,o))}}}const m=[(0,u.mean)(h,1)],_=e.sequences.dims,f=new u.Tensor("float32",new Float32Array(_[0]*_[1]),_);for(let e=0;e<_[0];++e){const t=m[e].neg().squeeze_(0),[r,o]=(0,p.dynamic_time_warping)(t.tolist()),s=Array.from({length:r.length-1},((e,t)=>r[t+1]-r[t])),a=(0,i.mergeArrays)([1],s).map((e=>!!e)),l=[];for(let e=0;e<a.length;++e)a[e]&&l.push(o[e]*n);f[e].data.set(l,1)}return f}}class ir extends G{main_input_name="pixel_values";forward_params=["pixel_values","input_ids","encoder_hidden_states","past_key_values"]}class lr extends G{forward_params=["input_ids","pixel_values","attention_mask","position_ids","past_key_values"]}class cr extends lr{_merge_input_ids_with_image_features({inputs_embeds:e,image_features:t,input_ids:r,attention_mask:n}){const o=this.config.image_token_index,s=r.tolist().map((e=>e.findIndex((e=>e==o)))),a=s.every((e=>-1===e)),i=s.every((e=>-1!==e));if(!a&&!i)throw new Error("Every input should contain either 0 or 1 image token.");if(a)return{inputs_embeds:e,attention_mask:n};const l=[],c=[];for(let r=0;r<s.length;++r){const o=s[r],a=e[r],i=t[r],d=n[r];l.push((0,u.cat)([a.slice([0,o]),i,a.slice([o+1,a.dims[0]])],0)),c.push((0,u.cat)([d.slice([0,o]),(0,u.ones)([i.dims[0]]),d.slice([o+1,d.dims[0]])],0))}return{inputs_embeds:(0,u.stack)(l,0),attention_mask:(0,u.stack)(c,0)}}}class dr extends cr{}class ur extends G{forward_params=["input_ids","inputs_embeds","attention_mask","pixel_values","encoder_outputs","decoder_input_ids","decoder_inputs_embeds","decoder_attention_mask","past_key_values"];main_input_name="inputs_embeds"}class pr extends ur{_merge_input_ids_with_image_features({inputs_embeds:e,image_features:t,input_ids:r,attention_mask:n}){return{inputs_embeds:(0,u.cat)([t,e],1),attention_mask:(0,u.cat)([(0,u.ones)(t.dims.slice(0,2)),n],1)}}async _prepare_inputs_embeds({input_ids:e,pixel_values:t,inputs_embeds:r,attention_mask:n}){if(!e&&!t)throw new Error("Either `input_ids` or `pixel_values` should be provided.");let o,s;return e&&(o=await this.encode_text({input_ids:e})),t&&(s=await this.encode_image({pixel_values:t})),o&&s?({inputs_embeds:r,attention_mask:n}=this._merge_input_ids_with_image_features({inputs_embeds:o,image_features:s,input_ids:e,attention_mask:n})):r=o||s,{inputs_embeds:r,attention_mask:n}}async forward({input_ids:e,pixel_values:t,attention_mask:r,decoder_input_ids:n,decoder_attention_mask:o,encoder_outputs:s,past_key_values:a,inputs_embeds:i,decoder_inputs_embeds:l}){if(i||({inputs_embeds:i,attention_mask:r}=await this._prepare_inputs_embeds({input_ids:e,pixel_values:t,inputs_embeds:i,attention_mask:r})),!s){let{last_hidden_state:e}=await N(this,{inputs_embeds:i,attention_mask:r});s=e}if(!l){if(!n)throw new Error("Either `decoder_input_ids` or `decoder_inputs_embeds` should be provided.");l=await this.encode_text({input_ids:n})}const c={inputs_embeds:l,attention_mask:o,encoder_attention_mask:r,encoder_hidden_states:s,past_key_values:a};return await D(this,c,!0)}}class hr extends G{}class mr extends hr{}class _r extends hr{static async from_pretrained(e,t={}){return t.model_file_name??="text_model",super.from_pretrained(e,t)}}class fr extends hr{static async from_pretrained(e,t={}){return t.model_file_name??="text_model",super.from_pretrained(e,t)}}class gr extends hr{static async from_pretrained(e,t={}){return t.model_file_name??="vision_model",super.from_pretrained(e,t)}}class Mr extends hr{static async from_pretrained(e,t={}){return t.model_file_name??="vision_model",super.from_pretrained(e,t)}}class wr extends G{}class Tr extends wr{}class br extends wr{static async from_pretrained(e,t={}){return t.model_file_name??="text_model",super.from_pretrained(e,t)}}class xr extends hr{static async from_pretrained(e,t={}){return t.model_file_name??="vision_model",super.from_pretrained(e,t)}}class yr extends G{}class kr extends yr{}class Fr extends G{}class Pr extends Fr{}class Cr extends Fr{}class vr extends G{}class Sr extends vr{}class Ar extends vr{}class Er extends G{}class Lr extends Er{}class zr extends Er{}class Ir extends G{}class Br extends Ir{}class Nr extends Ir{}class Dr extends G{}class Or extends Dr{}class Vr extends Dr{}class jr extends G{}class Rr extends jr{}class Gr extends jr{}class qr extends G{}class Wr extends qr{}class $r extends qr{}class Ur extends G{}class Xr extends Ur{}class Qr extends Ur{}class Hr extends G{}class Yr extends Hr{}class Jr extends Hr{}class Kr extends G{}class Zr extends Kr{}class en extends Kr{}class tn extends G{}class rn extends tn{}class nn extends tn{}class on extends G{}class sn extends on{}class an extends on{}class ln extends G{}class cn extends ln{}class dn extends ln{}class un extends G{}class pn extends un{}class hn extends un{}class mn extends G{}class _n extends mn{}class fn extends mn{}class gn extends G{}class Mn extends gn{}class wn extends gn{}class Tn extends G{}class bn extends Tn{}class xn extends Tn{}class yn extends G{}class kn extends yn{}class Fn extends yn{}class Pn extends G{}class Cn extends Pn{}class vn extends Pn{}class Sn extends G{}class An extends Sn{}class En extends Sn{}class Ln extends G{}class zn extends Ln{}class In extends Ln{async _call(e){return new pl(await super._call(e))}}class Bn extends G{}class Nn extends Bn{}class Dn extends Bn{async _call(e){return new pl(await super._call(e))}}class On extends G{}class Vn extends On{}class jn extends G{}class Rn extends jn{}class Gn extends jn{async _call(e){return new pl(await super._call(e))}}class qn extends G{}class Wn extends qn{}class $n extends G{}class Un extends $n{}class Xn extends $n{async _call(e){return new pl(await super._call(e))}}class Qn extends G{}class Hn extends Qn{async _call(e){return new wl(await super._call(e))}}class Yn extends G{}class Jn extends Yn{}class Kn extends Yn{async _call(e){return new pl(await super._call(e))}}class Zn extends G{}class eo extends Zn{}class to extends Zn{async _call(e){return new pl(await super._call(e))}}class ro extends G{}class no extends ro{}class oo extends ro{}class so extends G{}class ao extends so{}class io extends so{}class lo extends G{}class co extends lo{}class uo extends lo{async _call(e){return new pl(await super._call(e))}}class po extends G{}class ho extends po{}class mo extends po{async _call(e){return new fo(await super._call(e))}}class _o extends po{async _call(e){return new go(await super._call(e))}}class fo extends q{constructor({logits:e,pred_boxes:t}){super(),this.logits=e,this.pred_boxes=t}}class go extends q{constructor({logits:e,pred_boxes:t,pred_masks:r}){super(),this.logits=e,this.pred_boxes=t,this.pred_masks=r}}class Mo extends G{}class wo extends Mo{}class To extends Mo{async _call(e){return new bo(await super._call(e))}}class bo extends q{constructor({logits:e,pred_boxes:t}){super(),this.logits=e,this.pred_boxes=t}}class xo extends G{}class yo extends xo{}class ko extends xo{async _call(e){return new Fo(await super._call(e))}}class Fo extends fo{}class Po extends G{}class Co extends Po{}class vo extends Po{async _call(e){return new pl(await super._call(e))}}class So extends G{}class Ao extends So{}class Eo extends So{async _call(e){return new pl(await super._call(e))}}class Lo extends G{}class zo extends Lo{}class Io extends Lo{async _call(e){return new pl(await super._call(e))}}class Bo extends G{}class No extends Bo{}class Do extends Bo{async _call(e){return new pl(await super._call(e))}}class Oo extends G{}class Vo extends Oo{}class jo extends Oo{}class Ro extends G{}class Go extends Ro{}class qo extends Ro{}class Wo extends G{}class $o extends Wo{}class Uo extends G{}class Xo extends Uo{}class Qo extends Uo{}class Ho extends Uo{}class Yo extends G{}class Jo extends Yo{}class Ko extends G{}class Zo extends Ko{}class es extends Ko{}class ts extends G{}class rs extends ts{}class ns extends ts{}class os extends G{}class ss extends os{}class as extends G{}class is extends as{}class ls extends as{async _call(e){return new pl(await super._call(e))}}class cs extends G{}class ds extends cs{}class us extends cs{async _call(e){return new pl(await super._call(e))}}class ps extends G{}class hs extends ps{}class ms extends ps{async _call(e){return new pl(await super._call(e))}}class _s extends G{}class fs extends _s{}class gs extends _s{async _call(e){return new Ms(await super._call(e))}}class Ms extends q{constructor({logits:e,pred_boxes:t}){super(),this.logits=e,this.pred_boxes=t}}class ws extends G{}class Ts extends ws{async get_image_embeddings({pixel_values:e}){return await N(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),r=t.reduce(((e,t)=>e*t),1);e.input_labels=new u.Tensor("int64",new BigInt64Array(r).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 E(this.sessions.prompt_encoder_mask_decoder,t)}async _call(e){return new bs(await super._call(e))}}class bs extends q{constructor({iou_scores:e,pred_masks:t}){super(),this.iou_scores=e,this.pred_masks=t}}class xs extends G{}class ys extends xs{}class ks extends xs{}class Fs extends G{}class Ps extends Fs{}class Cs extends Fs{}class vs extends G{}class Ss extends vs{}class As extends vs{async _call(e){return new gl(await super._call(e))}}class Es extends vs{async _call(e){return new pl(await super._call(e))}}class Ls extends vs{async _call(e){return new ml(await super._call(e))}}class zs extends G{}class Is extends zs{}class Bs extends zs{async _call(e){return new ml(await super._call(e))}}class Ns extends G{}class Ds extends Ns{}class Os extends G{}class Vs extends Os{}class js extends Os{async _call(e){return new gl(await super._call(e))}}class Rs extends Os{async _call(e){return new pl(await super._call(e))}}class Gs extends G{}class qs extends Gs{}class Ws extends Gs{async _call(e){return new gl(await super._call(e))}}class $s extends Gs{async _call(e){return new pl(await super._call(e))}}class Us extends Gs{async _call(e){return new ml(await super._call(e))}}class Xs extends G{}class Qs extends Xs{}class Hs extends Xs{async _call(e){return new gl(await super._call(e))}}class Ys extends Xs{async _call(e){return new pl(await super._call(e))}}class Js extends G{}class Ks extends vs{}class Zs extends vs{async _call(e){return new gl(await super._call(e))}}class ea extends vs{async _call(e){return new pl(await super._call(e))}}class ta extends G{}class ra extends ta{}class na extends ta{async _call(e){return new gl(await super._call(e))}}class oa extends ta{async _call(e){return new pl(await super._call(e))}}class sa extends ta{async _call(e){return new hl(await super._call(e))}}class aa extends ta{async _call(e){return new ml(await super._call(e))}}class ia extends G{}class la extends ia{}class ca extends ia{}class da extends ia{async generate_speech(e,t,{threshold:r=.5,minlenratio:n=0,maxlenratio:o=20,vocoder:s=null}={}){const a={input_ids:e},{encoder_outputs:i,encoder_attention_mask:l}=await N(this,a),c=i.dims[1]/this.config.reduction_factor,d=Math.floor(c*o),p=Math.floor(c*n),h=this.config.num_mel_bins;let m=[],_=null,f=null,g=0;for(;;){++g;const e=I(!!f);let n;n=f?f.output_sequence_out:new u.Tensor("float32",new Float32Array(h),[1,1,h]);let o={use_cache_branch:e,output_sequence:n,encoder_attention_mask:l,speaker_embeddings:t,encoder_hidden_states:i};this.addPastKeyValues(o,_),f=await E(this.sessions.decoder_model_merged,o),_=this.getPastKeyValues(f,_);const{prob:s,spectrum:a}=f;if(m.push(a),g>=p&&(Array.from(s.data).filter((e=>e>=r)).length>0||g>=d))break}const M=(0,u.cat)(m),{waveform:w}=await E(s.sessions.model,{spectrogram:M});return{spectrogram:M,waveform:w}}}class ua extends G{main_input_name="spectrogram"}class pa extends G{}class ha extends pa{}class ma extends G{}class _a extends ma{}class fa extends ma{}class ga extends G{}class Ma extends ga{}class wa extends ga{}class Ta extends G{}class ba extends Ta{}class xa extends Ta{}class ya extends G{}class ka extends ya{}class Fa extends ya{static async from_pretrained(e,t={}){return t.model_file_name??="text_model",super.from_pretrained(e,t)}}class Pa extends ya{static async from_pretrained(e,t={}){return t.model_file_name??="audio_model",super.from_pretrained(e,t)}}class Ca extends G{}class va extends Ca{async _call(e){return new Tl(await super._call(e))}}class Sa extends G{}class Aa extends Sa{}class Ea extends Sa{}class La extends Sa{}class za extends G{}class Ia extends za{}class Ba extends za{}class Na extends G{}class Da extends Na{}class Oa extends Na{async _call(e){return new pl(await super._call(e))}}class Va extends G{}class ja extends Va{}class Ra extends Va{}class Ga extends G{forward_params=["input_ids","attention_mask","encoder_outputs","decoder_input_ids","decoder_attention_mask","past_key_values"];_apply_and_filter_by_delay_pattern_mask(e){const[t,r]=e.dims,n=this.config.decoder.num_codebooks,o=r-n;let s=0;for(let t=0;t<e.size;++t){if(e.data[t]===this.config.decoder.pad_token_id)continue;const a=t%r-Math.floor(t/r)%n;a>0&&a<=o&&(e.data[s++]=e.data[t])}const a=Math.floor(t/n),i=s/(a*n);return new u.Tensor(e.type,e.data.slice(0,s),[a,n,i])}prepare_inputs_for_generation(e,t,r){let n=structuredClone(e);for(let e=0;e<n.length;++e)for(let t=0;t<n[e].length;++t)e%this.config.decoder.num_codebooks>=t&&(n[e][t]=BigInt(this.config.decoder.pad_token_id));null!==r.guidance_scale&&r.guidance_scale>1&&(n=n.concat(n));return super.prepare_inputs_for_generation(n,t,r)}async generate(e){const t=await super.generate(e),r=this._apply_and_filter_by_delay_pattern_mask(t).unsqueeze_(0),{audio_values:n}=await E(this.sessions.encodec_decode,{audio_codes:r});return n}}class qa extends G{}class Wa extends qa{}class $a extends qa{async _call(e){return new pl(await super._call(e))}}class Ua extends G{}class Xa extends Ua{}class Qa extends Ua{async _call(e){return new pl(await super._call(e))}}class Ha extends G{}class Ya extends Ha{}class Ja extends Ha{async _call(e){return new pl(await super._call(e))}}class Ka extends G{}class Za extends Ka{}class ei extends Ka{async _call(e){return new pl(await super._call(e))}}class ti extends G{}class ri extends ti{}class ni{static MODEL_CLASS_MAPPINGS=null;static BASE_IF_FAIL=!1;static async from_pretrained(e,{progress_callback:t=null,config:r=null,cache_dir:o=null,local_files_only:s=!1,revision:a="main",model_file_name:i=null,subfolder:l="onnx",device:c=null,dtype:d=null,use_external_data_format:u=null,session_options:p={}}={}){const h={progress_callback:t,config:r,cache_dir:o,local_files_only:s,revision:a,model_file_name:i,subfolder:l,device:c,dtype:d,use_external_data_format:u,session_options:p};if(h.config=await n.AutoConfig.from_pretrained(e,h),!this.MODEL_CLASS_MAPPINGS)throw new Error("`MODEL_CLASS_MAPPINGS` not implemented for this type of `AutoClass`: "+this.name);for(const t of this.MODEL_CLASS_MAPPINGS){const r=t.get(h.config.model_type);if(r)return await r[1].from_pretrained(e,h)}if(this.BASE_IF_FAIL)return console.warn(`Unknown model class "${h.config.model_type}", attempting to construct from base class.`),await G.from_pretrained(e,h);throw Error(`Unsupported model type: ${h.config.model_type}`)}}const oi=new Map([["bert",["BertModel",U]],["nomic_bert",["NomicBertModel",K]],["roformer",["RoFormerModel",ee]],["electra",["ElectraModel",pe]],["esm",["EsmModel",Ge]],["convbert",["ConvBertModel",ae]],["camembert",["CamembertModel",Me]],["deberta",["DebertaModel",ke]],["deberta-v2",["DebertaV2Model",Ae]],["mpnet",["MPNetModel",Ke]],["albert",["AlbertModel",ct]],["distilbert",["DistilBertModel",Ne]],["roberta",["RobertaModel",Ot]],["xlm",["XLMModel",Wt]],["xlm-roberta",["XLMRobertaModel",Yt]],["clap",["ClapModel",ka]],["clip",["CLIPModel",mr]],["clipseg",["CLIPSegModel",Pr]],["chinese_clip",["ChineseCLIPModel",kr]],["siglip",["SiglipModel",Tr]],["mobilebert",["MobileBertModel",Xe]],["squeezebert",["SqueezeBertModel",ot]],["wav2vec2",["Wav2Vec2Model",Ss]],["wav2vec2-bert",["Wav2Vec2BertModel",Qs]],["unispeech",["UniSpeechModel",Vs]],["unispeech-sat",["UniSpeechSatModel",qs]],["hubert",["HubertModel",Ks]],["wavlm",["WavLMModel",ra]],["audio-spectrogram-transformer",["ASTModel",rr]],["vits",["VitsModel",va]],["pyannote",["PyAnnoteModel",Is]],["wespeaker-resnet",["WeSpeakerResNetModel",Ds]],["detr",["DetrModel",ho]],["rt_detr",["RTDetrModel",wo]],["table-transformer",["TableTransformerModel",yo]],["vit",["ViTModel",zn]],["pvt",["PvtModel",Nn]],["vit_msn",["ViTMSNModel",Rn]],["vit_mae",["ViTMAEModel",Vn]],["groupvit",["GroupViTModel",Wn]],["fastvit",["FastViTModel",Un]],["mobilevit",["MobileViTModel",Jn]],["mobilevitv2",["MobileViTV2Model",eo]],["owlvit",["OwlViTModel",no]],["owlv2",["Owlv2Model",ao]],["beit",["BeitModel",co]],["deit",["DeiTModel",Co]],["hiera",["HieraModel",Ao]],["convnext",["ConvNextModel",is]],["convnextv2",["ConvNextV2Model",ds]],["dinov2",["Dinov2Model",hs]],["resnet",["ResNetModel",zo]],["swin",["SwinModel",No]],["swin2sr",["Swin2SRModel",Vo]],["donut-swin",["DonutSwinModel",ss]],["yolos",["YolosModel",fs]],["dpt",["DPTModel",Go]],["glpn",["GLPNModel",rs]],["hifigan",["SpeechT5HifiGan",ua]],["efficientnet",["EfficientNetModel",Da]],["decision_transformer",["DecisionTransformerModel",ri]],["mobilenet_v1",["MobileNetV1Model",Wa]],["mobilenet_v2",["MobileNetV2Model",Xa]],["mobilenet_v3",["MobileNetV3Model",Ya]],["mobilenet_v4",["MobileNetV4Model",Za]],["maskformer",["MaskFormerModel",Zo]]]),si=new Map([["t5",["T5Model",mt]],["longt5",["LongT5Model",gt]],["mt5",["MT5Model",Tt]],["bart",["BartModel",yt]],["mbart",["MBartModel",Ct]],["marian",["MarianModel",ys]],["whisper",["WhisperModel",sr]],["m2m_100",["M2M100Model",Ps]],["blenderbot",["BlenderbotModel",Lt]],["blenderbot-small",["BlenderbotSmallModel",Bt]]]),ai=new Map([["bloom",["BloomModel",kn]],["jais",["JAISModel",Lr]],["gpt2",["GPT2Model",Sr]],["gptj",["GPTJModel",Rr]],["gpt_bigcode",["GPTBigCodeModel",Wr]],["gpt_neo",["GPTNeoModel",Br]],["gpt_neox",["GPTNeoXModel",Or]],["codegen",["CodeGenModel",Xr]],["llama",["LlamaModel",Yr]],["granite",["GraniteModel",Zr]],["cohere",["CohereModel",rn]],["gemma",["GemmaModel",sn]],["gemma2",["Gemma2Model",cn]],["openelm",["OpenELMModel",pn]],["qwen2",["Qwen2Model",_n]],["phi",["PhiModel",Mn]],["phi3",["Phi3Model",bn]],["mpt",["MptModel",Cn]],["opt",["OPTModel",An]],["mistral",["MistralModel",_a]],["starcoder2",["Starcoder2Model",Ma]],["falcon",["FalconModel",ba]],["stablelm",["StableLmModel",Ia]]]),ii=new Map([["speecht5",["SpeechT5ForSpeechToText",ca]],["whisper",["WhisperForConditionalGeneration",ar]]]),li=new Map([["speecht5",["SpeechT5ForTextToSpeech",da]]]),ci=new Map([["vits",["VitsModel",va]],["musicgen",["MusicgenForConditionalGeneration",Ga]]]),di=new Map([["bert",["BertForSequenceClassification",Q]],["roformer",["RoFormerForSequenceClassification",re]],["electra",["ElectraForSequenceClassification",me]],["esm",["EsmForSequenceClassification",We]],["convbert",["ConvBertForSequenceClassification",le]],["camembert",["CamembertForSequenceClassification",Te]],["deberta",["DebertaForSequenceClassification",Pe]],["deberta-v2",["DebertaV2ForSequenceClassification",Le]],["mpnet",["MPNetForSequenceClassification",et]],["albert",["AlbertForSequenceClassification",dt]],["distilbert",["DistilBertForSequenceClassification",De]],["roberta",["RobertaForSequenceClassification",jt]],["xlm",["XLMForSequenceClassification",Ut]],["xlm-roberta",["XLMRobertaForSequenceClassification",Kt]],["bart",["BartForSequenceClassification",Ft]],["mbart",["MBartForSequenceClassification",St]],["mobilebert",["MobileBertForSequenceClassification",He]],["squeezebert",["SqueezeBertForSequenceClassification",at]]]),ui=new Map([["bert",["BertForTokenClassification",H]],["roformer",["RoFormerForTokenClassification",ne]],["electra",["ElectraForTokenClassification",_e]],["esm",["EsmForTokenClassification",$e]],["convbert",["ConvBertForTokenClassification",ce]],["camembert",["CamembertForTokenClassification",be]],["deberta",["DebertaForTokenClassification",Ce]],["deberta-v2",["DebertaV2ForTokenClassification",ze]],["mpnet",["MPNetForTokenClassification",tt]],["distilbert",["DistilBertForTokenClassification",Oe]],["roberta",["RobertaForTokenClassification",Rt]],["xlm",["XLMForTokenClassification",Xt]],["xlm-roberta",["XLMRobertaForTokenClassification",Zt]]]),pi=new Map([["t5",["T5ForConditionalGeneration",_t]],["longt5",["LongT5ForConditionalGeneration",Mt]],["mt5",["MT5ForConditionalGeneration",bt]],["bart",["BartForConditionalGeneration",kt]],["mbart",["MBartForConditionalGeneration",vt]],["marian",["MarianMTModel",ks]],["m2m_100",["M2M100ForConditionalGeneration",Cs]],["blenderbot",["BlenderbotForConditionalGeneration",zt]],["blenderbot-small",["BlenderbotSmallForConditionalGeneration",Nt]]]),hi=new Map([["bloom",["BloomForCausalLM",Fn]],["gpt2",["GPT2LMHeadModel",Ar]],["jais",["JAISLMHeadModel",zr]],["gptj",["GPTJForCausalLM",Gr]],["gpt_bigcode",["GPTBigCodeForCausalLM",$r]],["gpt_neo",["GPTNeoForCausalLM",Nr]],["gpt_neox",["GPTNeoXForCausalLM",Vr]],["codegen",["CodeGenForCausalLM",Qr]],["llama",["LlamaForCausalLM",Jr]],["granite",["GraniteForCausalLM",en]],["cohere",["CohereForCausalLM",nn]],["gemma",["GemmaForCausalLM",an]],["gemma2",["Gemma2ForCausalLM",dn]],["openelm",["OpenELMForCausalLM",hn]],["qwen2",["Qwen2ForCausalLM",fn]],["phi",["PhiForCausalLM",wn]],["phi3",["Phi3ForCausalLM",xn]],["mpt",["MptForCausalLM",vn]],["opt",["OPTForCausalLM",En]],["mbart",["MBartForCausalLM",At]],["mistral",["MistralForCausalLM",fa]],["starcoder2",["Starcoder2ForCausalLM",wa]],["falcon",["FalconForCausalLM",xa]],["trocr",["TrOCRForCausalLM",ha]],["stablelm",["StableLmForCausalLM",Ba]]]),mi=new Map([["bert",["BertForMaskedLM",X]],["roformer",["RoFormerForMaskedLM",te]],["electra",["ElectraForMaskedLM",he]],["esm",["EsmForMaskedLM",qe]],["convbert",["ConvBertForMaskedLM",ie]],["camembert",["CamembertForMaskedLM",we]],["deberta",["DebertaForMaskedLM",Fe]],["deberta-v2",["DebertaV2ForMaskedLM",Ee]],["mpnet",["MPNetForMaskedLM",Ze]],["albert",["AlbertForMaskedLM",pt]],["distilbert",["DistilBertForMaskedLM",je]],["roberta",["RobertaForMaskedLM",Vt]],["xlm",["XLMWithLMHeadModel",$t]],["xlm-roberta",["XLMRobertaForMaskedLM",Jt]],["mobilebert",["MobileBertForMaskedLM",Qe]],["squeezebert",["SqueezeBertForMaskedLM",st]]]),_i=new Map([["bert",["BertForQuestionAnswering",Y]],["roformer",["RoFormerForQuestionAnswering",oe]],["electra",["ElectraForQuestionAnswering",fe]],["convbert",["ConvBertForQuestionAnswering",de]],["camembert",["CamembertForQuestionAnswering",xe]],["deberta",["DebertaForQuestionAnswering",ve]],["deberta-v2",["DebertaV2ForQuestionAnswering",Ie]],["mpnet",["MPNetForQuestionAnswering",rt]],["albert",["AlbertForQuestionAnswering",ut]],["distilbert",["DistilBertForQuestionAnswering",Ve]],["roberta",["RobertaForQuestionAnswering",Gt]],["xlm",["XLMForQuestionAnswering",Qt]],["xlm-roberta",["XLMRobertaForQuestionAnswering",er]],["mobilebert",["MobileBertForQuestionAnswering",Ye]],["squeezebert",["SqueezeBertForQuestionAnswering",it]]]),fi=new Map([["vision-encoder-decoder",["VisionEncoderDecoderModel",ir]]]),gi=new Map([["llava",["LlavaForConditionalGeneration",cr]],["moondream1",["Moondream1ForConditionalGeneration",dr]],["florence2",["Florence2ForConditionalGeneration",pr]]]),Mi=new Map([["vision-encoder-decoder",["VisionEncoderDecoderModel",ir]]]),wi=new Map([["vit",["ViTForImageClassification",In]],["pvt",["PvtForImageClassification",Dn]],["vit_msn",["ViTMSNForImageClassification",Gn]],["fastvit",["FastViTForImageClassification",Xn]],["mobilevit",["MobileViTForImageClassification",Kn]],["mobilevitv2",["MobileViTV2ForImageClassification",to]],["beit",["BeitForImageClassification",uo]],["deit",["DeiTForImageClassification",vo]],["hiera",["HieraForImageClassification",Eo]],["convnext",["ConvNextForImageClassification",ls]],["convnextv2",["ConvNextV2ForImageClassification",us]],["dinov2",["Dinov2ForImageClassification",ms]],["resnet",["ResNetForImageClassification",Io]],["swin",["SwinForImageClassification",Do]],["segformer",["SegformerForImageClassification",Ea]],["efficientnet",["EfficientNetForImageClassification",Oa]],["mobilenet_v1",["MobileNetV1ForImageClassification",$a]],["mobilenet_v2",["MobileNetV2ForImageClassification",Qa]],["mobilenet_v3",["MobileNetV3ForImageClassification",Ja]],["mobilenet_v4",["MobileNetV4ForImageClassification",ei]]]),Ti=new Map([["detr",["DetrForObjectDetection",mo]],["rt_detr",["RTDetrForObjectDetection",To]],["table-transformer",["TableTransformerForObjectDetection",ko]],["yolos",["YolosForObjectDetection",gs]]]),bi=new Map([["owlvit",["OwlViTForObjectDetection",oo]],["owlv2",["Owlv2ForObjectDetection",io]]]),xi=new Map([["detr",["DetrForSegmentation",_o]],["clipseg",["CLIPSegForImageSegmentation",Cr]]]),yi=new Map([["segformer",["SegformerForSemanticSegmentation",La]],["sapiens",["SapiensForSemanticSegmentation",Xo]]]),ki=new Map([["detr",["DetrForSegmentation",_o]],["maskformer",["MaskFormerForInstanceSegmentation",es]]]),Fi=new Map([["sam",["SamModel",Ts]]]),Pi=new Map([["wav2vec2",["Wav2Vec2ForCTC",As]],["wav2vec2-bert",["Wav2Vec2BertForCTC",Hs]],["unispeech",["UniSpeechForCTC",js]],["unispeech-sat",["UniSpeechSatForCTC",Ws]],["wavlm",["WavLMForCTC",na]],["hubert",["HubertForCTC",Zs]]]),Ci=new Map([["wav2vec2",["Wav2Vec2ForSequenceClassification",Es]],["wav2vec2-bert",["Wav2Vec2BertForSequenceClassification",Ys]],["unispeech",["UniSpeechForSequenceClassification",Rs]],["unispeech-sat",["UniSpeechSatForSequenceClassification",$s]],["wavlm",["WavLMForSequenceClassification",oa]],["hubert",["HubertForSequenceClassification",ea]],["audio-spectrogram-transformer",["ASTForAudioClassification",nr]]]),vi=new Map([["wavlm",["WavLMForXVector",sa]]]),Si=new Map([["unispeech-sat",["UniSpeechSatForAudioFrameClassification",Us]],["wavlm",["WavLMForAudioFrameClassification",aa]],["wav2vec2",["Wav2Vec2ForAudioFrameClassification",Ls]],["pyannote",["PyAnnoteForAudioFrameClassification",Bs]]]),Ai=new Map([["vitmatte",["VitMatteForImageMatting",Hn]]]),Ei=new Map([["swin2sr",["Swin2SRForImageSuperResolution",jo]]]),Li=new Map([["dpt",["DPTForDepthEstimation",qo]],["depth_anything",["DepthAnythingForDepthEstimation",$o]],["glpn",["GLPNForDepthEstimation",ns]],["sapiens",["SapiensForDepthEstimation",Qo]],["depth_pro",["DepthProForDepthEstimation",Jo]]]),zi=new Map([["sapiens",["SapiensForNormalEstimation",Ho]]]),Ii=new Map([["clip",["CLIPVisionModelWithProjection",Mr]],["siglip",["SiglipVisionModel",xr]]]),Bi=[[oi,M],[si,w],[ai,x],[di,M],[ui,M],[pi,T],[ii,T],[hi,x],[mi,M],[_i,M],[fi,b],[gi,k],[wi,M],[xi,M],[ki,M],[yi,M],[Ai,M],[Ei,M],[Li,M],[zi,M],[Ti,M],[bi,M],[Fi,y],[Pi,M],[Ci,M],[li,T],[ci,M],[vi,M],[Si,M],[Ii,M]];for(const[e,t]of Bi)for(const[r,n]of e.values())P.set(r,t),v.set(n,r),C.set(r,n);const Ni=[["MusicgenForConditionalGeneration",Ga,F],["CLIPTextModelWithProjection",fr,M],["SiglipTextModel",br,M],["ClapTextModelWithProjection",Fa,M],["ClapAudioModelWithProjection",Pa,M]];for(const[e,t,r]of Ni)P.set(e,r),v.set(t,e),C.set(e,t);class Di extends ni{static MODEL_CLASS_MAPPINGS=Bi.map((e=>e[0]));static BASE_IF_FAIL=!0}class Oi extends ni{static MODEL_CLASS_MAPPINGS=[di]}class Vi extends ni{static MODEL_CLASS_MAPPINGS=[ui]}class ji extends ni{static MODEL_CLASS_MAPPINGS=[pi]}class Ri extends ni{static MODEL_CLASS_MAPPINGS=[ii]}class Gi extends ni{static MODEL_CLASS_MAPPINGS=[li]}class qi extends ni{static MODEL_CLASS_MAPPINGS=[ci]}class Wi extends ni{static MODEL_CLASS_MAPPINGS=[hi]}class $i extends ni{static MODEL_CLASS_MAPPINGS=[mi]}class Ui extends ni{static MODEL_CLASS_MAPPINGS=[_i]}class Xi extends ni{static MODEL_CLASS_MAPPINGS=[fi]}class Qi extends ni{static MODEL_CLASS_MAPPINGS=[wi]}class Hi extends ni{static MODEL_CLASS_MAPPINGS=[xi]}class Yi extends ni{static MODEL_CLASS_MAPPINGS=[yi]}class Ji extends ni{static MODEL_CLASS_MAPPINGS=[ki]}class Ki extends ni{static MODEL_CLASS_MAPPINGS=[Ti]}class Zi extends ni{static MODEL_CLASS_MAPPINGS=[bi]}class el extends ni{static MODEL_CLASS_MAPPINGS=[Fi]}class tl extends ni{static MODEL_CLASS_MAPPINGS=[Pi]}class rl extends ni{static MODEL_CLASS_MAPPINGS=[Ci]}class nl extends ni{static MODEL_CLASS_MAPPINGS=[vi]}class ol extends ni{static MODEL_CLASS_MAPPINGS=[Si]}class sl extends ni{static MODEL_CLASS_MAPPINGS=[Mi]}class al extends ni{static MODEL_CLASS_MAPPINGS=[Ai]}class il extends ni{static MODEL_CLASS_MAPPINGS=[Ei]}class ll extends ni{static MODEL_CLASS_MAPPINGS=[Li]}class cl extends ni{static MODEL_CLASS_MAPPINGS=[zi]}class dl extends ni{static MODEL_CLASS_MAPPINGS=[Ii]}class ul extends q{constructor({logits:e,past_key_values:t,encoder_outputs:r,decoder_attentions:n=null,cross_attentions:o=null}){super(),this.logits=e,this.past_key_values=t,this.encoder_outputs=r,this.decoder_attentions=n,this.cross_attentions=o}}class pl extends q{constructor({logits:e}){super(),this.logits=e}}class hl extends q{constructor({logits:e,embeddings:t}){super(),this.logits=e,this.embeddings=t}}class ml extends q{constructor({logits:e}){super(),this.logits=e}}class _l extends q{constructor({logits:e}){super(),this.logits=e}}class fl extends q{constructor({start_logits:e,end_logits:t}){super(),this.start_logits=e,this.end_logits=t}}class gl extends q{constructor({logits:e}){super(),this.logits=e}}class Ml extends q{constructor({logits:e,past_key_values:t}){super(),this.logits=e,this.past_key_values=t}}class wl extends q{constructor({alphas:e}){super(),this.alphas=e}}class Tl extends q{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,r)=>{r.r(t),r.d(t,{WHISPER_LANGUAGE_MAPPING:()=>o,WHISPER_TO_LANGUAGE_CODE_MAPPING:()=>s,whisper_language_to_code:()=>a});const n=[["en","english"],["zh","chinese"],["de","german"],["es","spanish"],["ru","russian"],["ko","korean"],["fr","french"],["ja","japanese"],["pt","portuguese"],["tr","turkish"],["pl","polish"],["ca","catalan"],["nl","dutch"],["ar","arabic"],["sv","swedish"],["it","italian"],["id","indonesian"],["hi","hindi"],["fi","finnish"],["vi","vietnamese"],["he","hebrew"],["uk","ukrainian"],["el","greek"],["ms","malay"],["cs","czech"],["ro","romanian"],["da","danish"],["hu","hungarian"],["ta","tamil"],["no","norwegian"],["th","thai"],["ur","urdu"],["hr","croatian"],["bg","bulgarian"],["lt","lithuanian"],["la","latin"],["mi","maori"],["ml","malayalam"],["cy","welsh"],["sk","slovak"],["te","telugu"],["fa","persian"],["lv","latvian"],["bn","bengali"],["sr","serbian"],["az","azerbaijani"],["sl","slovenian"],["kn","kannada"],["et","estonian"],["mk","macedonian"],["br","breton"],["eu","basque"],["is","icelandic"],["hy","armenian"],["ne","nepali"],["mn","mongolian"],["bs","bosnian"],["kk","kazakh"],["sq","albanian"],["sw","swahili"],["gl","galician"],["mr","marathi"],["pa","punjabi"],["si","sinhala"],["km","khmer"],["sn","shona"],["yo","yoruba"],["so","somali"],["af","afrikaans"],["oc","occitan"],["ka","georgian"],["be","belarusian"],["tg","tajik"],["sd","sindhi"],["gu","gujarati"],["am","amharic"],["yi","yiddish"],["lo","lao"],["uz","uzbek"],["fo","faroese"],["ht","haitian creole"],["ps","pashto"],["tk","turkmen"],["nn","nynorsk"],["mt","maltese"],["sa","sanskrit"],["lb","luxembourgish"],["my","myanmar"],["bo","tibetan"],["tl","tagalog"],["mg","malagasy"],["as","assamese"],["tt","tatar"],["haw","hawaiian"],["ln","lingala"],["ha","hausa"],["ba","bashkir"],["jw","javanese"],["su","sundanese"]],o=new Map(n),s=new Map([...n.map((([e,t])=>[t,e])),["burmese","my"],["valencian","ca"],["flemish","nl"],["haitian","ht"],["letzeburgesch","lb"],["pushto","ps"],["panjabi","pa"],["moldavian","ro"],["moldovan","ro"],["sinhalese","si"],["castilian","es"]]);function a(e){e=e.toLowerCase();let t=s.get(e);if(void 0===t){if(!o.has(e)){const t=2===e.length?o.keys():o.values();throw new Error(`Language "${e}" is not supported. Must be one of: ${JSON.stringify(t)}`)}t=e}return t}},"./src/models/whisper/generation_whisper.js":
@@ -160,8 +160,8 @@ import*as e from"fs";import*as t from"onnxruntime-node";import*as r from"path";i
   \****************************/(e,t,r)=>{function n(e,[t,r,n],[o,s],a="bilinear",i=!1){const l=s/n,c=o/r,d=new e.constructor(o*s*t),u=r*n,p=o*s;for(let a=0;a<o;++a)for(let o=0;o<s;++o){const i=a*s+o,h=(o+.5)/l-.5,m=(a+.5)/c-.5;let _=Math.floor(h),f=Math.floor(m);const g=Math.min(_+1,n-1),M=Math.min(f+1,r-1);_=Math.max(_,0),f=Math.max(f,0);const w=h-_,T=m-f,b=(1-w)*(1-T),x=w*(1-T),y=(1-w)*T,k=w*T,F=f*n,P=M*n,C=F+_,v=F+g,S=P+_,A=P+g;for(let r=0;r<t;++r){const t=r*u;d[r*p+i]=b*e[t+C]+x*e[t+v]+y*e[t+S]+k*e[t+A]}}return d}function o(e,t,r){const n=new Array(r.length),o=new Array(r.length);for(let e=r.length-1,s=1;e>=0;--e)o[e]=s,n[e]=t[r[e]],s*=n[e];const s=r.map(((e,t)=>o[r.indexOf(t)])),a=new e.constructor(e.length);for(let r=0;r<e.length;++r){let n=0;for(let e=t.length-1,o=r;e>=0;--e)n+=o%t[e]*s[e],o=Math.floor(o/t[e]);a[n]=e[r]}return[a,n]}function s(e){const t=u(e)[0],r=e.map((e=>Math.exp(e-t))),n=r.reduce(((e,t)=>e+t),0);return r.map((e=>e/n))}function a(e){const t=u(e)[0];let r=0;for(let n=0;n<e.length;++n)r+=Math.exp(e[n]-t);const n=Math.log(r);return e.map((e=>e-t-n))}function i(e,t){let r=0;for(let n=0;n<e.length;++n)r+=e[n]*t[n];return r}function l(e,t){return i(e,t)/(c(e)*c(t))}function c(e){return Math.sqrt(e.reduce(((e,t)=>e+t*t),0))}function d(e){if(0===e.length)throw Error("Array must not be empty");let t=e[0],r=0;for(let n=1;n<e.length;++n)e[n]<t&&(t=e[n],r=n);return[t,r]}function u(e){if(0===e.length)throw Error("Array must not be empty");let t=e[0],r=0;for(let n=1;n<e.length;++n)e[n]>t&&(t=e[n],r=n);return[Number(t),r]}function p(e){return e>0&&!(e&e-1)}r.r(t),r.d(t,{FFT:()=>_,bankers_round:()=>M,cos_sim:()=>l,dot:()=>i,dynamic_time_warping:()=>w,interpolate_data:()=>n,log_softmax:()=>a,magnitude:()=>c,max:()=>u,medianFilter:()=>f,min:()=>d,permute_data:()=>o,round:()=>g,softmax:()=>s});class h{constructor(e){if(this.size=0|e,this.size<=1||!p(this.size))throw new Error("FFT size must be a power of two larger than 1");this._csize=e<<1,this.table=new Float64Array(2*this.size);for(let e=0;e<this.table.length;e+=2){const t=Math.PI*e/this.size;this.table[e]=Math.cos(t),this.table[e+1]=-Math.sin(t)}let t=0;for(let e=1;this.size>e;e<<=1)++t;this._width=t%2==0?t-1:t,this._bitrev=new Int32Array(1<<this._width);for(let e=0;e<this._bitrev.length;++e){this._bitrev[e]=0;for(let t=0;t<this._width;t+=2){const r=this._width-t-2;this._bitrev[e]|=(e>>>t&3)<<r}}}createComplexArray(){return new Float64Array(this._csize)}fromComplexArray(e,t){const r=t||new Array(e.length>>>1);for(let t=0;t<e.length;t+=2)r[t>>>1]=e[t];return r}toComplexArray(e,t){const r=t||this.createComplexArray();for(let t=0;t<r.length;t+=2)r[t]=e[t>>>1],r[t+1]=0;return r}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,r){const n=this._csize;let o,s,a=1<<this._width,i=n/a<<1;const l=this._bitrev;if(4===i)for(o=0,s=0;o<n;o+=i,++s){const r=l[s];this._singleTransform2(t,e,o,r,a)}else for(o=0,s=0;o<n;o+=i,++s){const n=l[s];this._singleTransform4(t,e,o,n,a,r)}const c=this.table;for(a>>=2;a>=2;a>>=2){i=n/a<<1;const t=i>>>2;for(o=0;o<n;o+=i){const n=o+t-1;for(let s=o,i=0;s<n;s+=2,i+=a){const n=s,o=n+t,a=o+t,l=a+t,d=e[n],u=e[n+1],p=e[o],h=e[o+1],m=e[a],_=e[a+1],f=e[l],g=e[l+1],M=c[i],w=r*c[i+1],T=p*M-h*w,b=p*w+h*M,x=c[2*i],y=r*c[2*i+1],k=m*x-_*y,F=m*y+_*x,P=c[3*i],C=r*c[3*i+1],v=f*P-g*C,S=f*C+g*P,A=d+k,E=u+F,L=d-k,z=u-F,I=T+v,B=b+S,N=r*(T-v),D=r*(b-S);e[n]=A+I,e[n+1]=E+B,e[o]=L+D,e[o+1]=z-N,e[a]=A-I,e[a+1]=E-B,e[l]=L-D,e[l+1]=z+N}}}}_singleTransform2(e,t,r,n,o){const s=e[n],a=e[n+1],i=e[n+o],l=e[n+o+1];t[r]=s+i,t[r+1]=a+l,t[r+2]=s-i,t[r+3]=a-l}_singleTransform4(e,t,r,n,o,s){const a=2*o,i=3*o,l=e[n],c=e[n+1],d=e[n+o],u=e[n+o+1],p=e[n+a],h=e[n+a+1],m=e[n+i],_=e[n+i+1],f=l+p,g=c+h,M=l-p,w=c-h,T=d+m,b=u+_,x=s*(d-m),y=s*(u-_);t[r]=f+T,t[r+1]=g+b,t[r+2]=M+y,t[r+3]=w-x,t[r+4]=f-T,t[r+5]=g-b,t[r+6]=M-y,t[r+7]=w+x}_realTransform4(e,t,r){const n=this._csize;let o,s,a=1<<this._width,i=n/a<<1;const l=this._bitrev;if(4===i)for(o=0,s=0;o<n;o+=i,++s){const r=l[s];this._singleRealTransform2(t,e,o,r>>>1,a>>>1)}else for(o=0,s=0;o<n;o+=i,++s){const n=l[s];this._singleRealTransform4(t,e,o,n>>>1,a>>>1,r)}const c=this.table;for(a>>=2;a>=2;a>>=2){i=n/a<<1;const t=i>>>1,s=t>>>1,l=s>>>1;for(o=0;o<n;o+=i)for(let n=0,i=0;n<=l;n+=2,i+=a){const a=o+n,d=a+s,u=d+s,p=u+s,h=e[a],m=e[a+1],_=e[d],f=e[d+1],g=e[u],M=e[u+1],w=e[p],T=e[p+1],b=h,x=m,y=c[i],k=r*c[i+1],F=_*y-f*k,P=_*k+f*y,C=c[2*i],v=r*c[2*i+1],S=g*C-M*v,A=g*v+M*C,E=c[3*i],L=r*c[3*i+1],z=w*E-T*L,I=w*L+T*E,B=b+S,N=x+A,D=b-S,O=x-A,V=F+z,j=P+I,R=r*(F-z),G=r*(P-I);if(e[a]=B+V,e[a+1]=N+j,e[d]=D+G,e[d+1]=O-R,0===n){e[u]=B-V,e[u+1]=N-j;continue}if(n===l)continue;const q=o+s-n,W=o+t-n;e[q]=D-r*G,e[q+1]=-O-r*R,e[W]=B-r*V,e[W+1]=r*j-N}}const d=n>>>1;for(let t=2;t<d;t+=2)e[n-t]=e[t],e[n-t+1]=-e[t+1]}_singleRealTransform2(e,t,r,n,o){const s=e[n],a=e[n+o];t[r]=s+a,t[r+1]=0,t[r+2]=s-a,t[r+3]=0}_singleRealTransform4(e,t,r,n,o,s){const a=2*o,i=3*o,l=e[n],c=e[n+o],d=e[n+a],u=e[n+i],p=l+d,h=l-d,m=c+u,_=s*(c-u);t[r]=p+m,t[r+1]=0,t[r+2]=h,t[r+3]=-_,t[r+4]=p-m,t[r+5]=0,t[r+6]=h,t[r+7]=_}}class m{constructor(e){const t=2*(e-1),r=2*(2*e-1),n=2**Math.ceil(Math.log2(r));this.bufferSize=n,this._a=t;const o=new Float64Array(r),s=new Float64Array(n);this._chirpBuffer=new Float64Array(n),this._buffer1=new Float64Array(n),this._buffer2=new Float64Array(n),this._outBuffer1=new Float64Array(n),this._outBuffer2=new Float64Array(n);const a=-2*Math.PI/e,i=Math.cos(a),l=Math.sin(a);for(let t=0;t<r>>1;++t){const r=(t+1-e)**2/2,n=Math.sqrt(i**2+l**2)**r,a=r*Math.atan2(l,i),c=2*t;o[c]=n*Math.cos(a),o[c+1]=n*Math.sin(a),s[c]=o[c],s[c+1]=-o[c+1]}this._slicedChirpBuffer=o.subarray(t,r),this._f=new h(n>>1),this._f.transform(this._chirpBuffer,s)}_transform(e,t,r){const n=this._buffer1,o=this._buffer2,s=this._outBuffer1,a=this._outBuffer2,i=this._chirpBuffer,l=this._slicedChirpBuffer,c=this._a;if(r)for(let e=0;e<l.length;e+=2){const r=e+1,o=t[e>>1];n[e]=o*l[e],n[r]=o*l[r]}else for(let e=0;e<l.length;e+=2){const r=e+1;n[e]=t[e]*l[e]-t[r]*l[r],n[r]=t[e]*l[r]+t[r]*l[e]}this._f.transform(s,n);for(let e=0;e<i.length;e+=2){const t=e+1;o[e]=s[e]*i[e]-s[t]*i[t],o[t]=s[e]*i[t]+s[t]*i[e]}this._f.inverseTransform(a,o);for(let t=0;t<a.length;t+=2){const r=a[t+c],n=a[t+c+1],o=l[t],s=l[t+1];e[t]=r*o-n*s,e[t+1]=r*s+n*o}}transform(e,t){this._transform(e,t,!1)}realTransform(e,t){this._transform(e,t,!0)}}class _{constructor(e){this.fft_length=e,this.isPowerOfTwo=p(e),this.isPowerOfTwo?(this.fft=new h(e),this.outputBufferSize=2*e):(this.fft=new m(e),this.outputBufferSize=this.fft.bufferSize)}realTransform(e,t){this.fft.realTransform(e,t)}transform(e,t){this.fft.transform(e,t)}}function f(e,t){if(t%2==0||t<=0)throw new Error("Window size must be a positive odd number");const r=new e.constructor(e.length),n=new e.constructor(t),o=Math.floor(t/2);for(let t=0;t<e.length;++t){let s=0;for(let r=-o;r<=o;++r){let o=t+r;o<0?o=Math.abs(o):o>=e.length&&(o=2*(e.length-1)-o),n[s++]=e[o]}n.sort(),r[t]=n[o]}return r}function g(e,t){const r=Math.pow(10,t);return Math.round(e*r)/r}function M(e){const t=Math.round(e);return Math.abs(e)%1==.5?t%2==0?t:t-1:t}function w(e){const t=e.length,r=e[0].length,n=[t+1,r+1],o=Array.from({length:n[0]},(()=>Array(n[1]).fill(1/0)));o[0][0]=0;const s=Array.from({length:n[0]},(()=>Array(n[1]).fill(-1)));for(let t=1;t<n[1];++t)for(let r=1;r<n[0];++r){const n=o[r-1][t-1],a=o[r-1][t],i=o[r][t-1];let l,c;n<a&&n<i?(l=n,c=0):a<n&&a<i?(l=a,c=1):(l=i,c=2),o[r][t]=e[r-1][t-1]+l,s[r][t]=c}for(let e=0;e<n[1];++e)s[0][e]=2;for(let e=0;e<n[0];++e)s[e][0]=1;let a=t,i=r,l=[],c=[];for(;a>0||i>0;)switch(l.push(a-1),c.push(i-1),s[a][i]){case 0:--a,--i;break;case 1:--a;break;case 2:--i;break;default:throw new Error(`Internal error in dynamic time warping. Unexpected trace[${a}, ${i}]. Please file a bug report.`)}return l.reverse(),c.reverse(),[l,c]}},"./src/utils/tensor.js":
 /*!*****************************!*\
   !*** ./src/utils/tensor.js ***!
-  \*****************************/(e,t,r)=>{r.r(t),r.d(t,{Tensor:()=>i,cat:()=>w,full:()=>k,full_like:()=>F,interpolate:()=>c,interpolate_4d:()=>d,layer_norm:()=>_,matmul:()=>u,mean:()=>x,mean_pooling:()=>m,ones:()=>P,ones_like:()=>C,permute:()=>l,quantize_embeddings:()=>A,rfft:()=>p,stack:()=>T,std_mean:()=>b,topk:()=>h,zeros:()=>v,zeros_like:()=>S});var n=r(/*! ./maths.js */"./src/utils/maths.js"),o=r(/*! ../backends/onnx.js */"./src/backends/onnx.js"),s=r(/*! ../ops/registry.js */"./src/ops/registry.js");const a=Object.freeze({float32:Float32Array,float16:Uint16Array,float64:Float64Array,string:Array,int8:Int8Array,uint8:Uint8Array,int16:Int16Array,uint16:Uint16Array,int32:Int32Array,uint32:Uint32Array,int64:BigInt64Array,uint64:BigUint64Array,bool:Uint8Array});class i{get dims(){return this.ort_tensor.dims}set dims(e){this.ort_tensor.dims=e}get type(){return this.ort_tensor.type}get data(){return this.ort_tensor.data}get size(){return this.ort_tensor.size}get location(){return this.ort_tensor.location}ort_tensor;constructor(...e){return(0,o.isONNXTensor)(e[0])?this.ort_tensor=e[0]:this.ort_tensor=new o.Tensor(e[0],e[1],e[2]),new Proxy(this,{get:(e,t)=>{if("string"==typeof t){let r=Number(t);if(Number.isInteger(r))return e._getitem(r)}return e[t]},set:(e,t,r)=>e[t]=r})}dispose(){this.ort_tensor.dispose()}*[Symbol.iterator](){const[e,...t]=this.dims;if(t.length>0){const r=t.reduce(((e,t)=>e*t));for(let n=0;n<e;++n)yield this._subarray(n,r,t)}else yield*this.data}_getitem(e){const[t,...r]=this.dims;if(e=M(e,t),r.length>0){const t=r.reduce(((e,t)=>e*t));return this._subarray(e,t,r)}return new i(this.type,[this.data[e]],r)}indexOf(e){const t=this.data;for(let r=0;r<t.length;++r)if(t[r]==e)return r;return-1}_subarray(e,t,r){const n=e*t,o=(e+1)*t,s="subarray"in this.data?this.data.subarray(n,o):this.data.slice(n,o);return new i(this.type,s,r)}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 r=e.length,n=t.reduce(((e,t)=>e*t));if(r!==n)throw Error(`cannot reshape array of size ${r} into shape (${t})`);let o=e;for(let e=t.length-1;e>=0;e--)o=o.reduce(((r,n)=>{let o=r[r.length-1];return o.length<t[e]?o.push(n):r.push([n]),r}),[[]]);return o[0]}(this.data,this.dims)}sigmoid(){return this.clone().sigmoid_()}sigmoid_(){const e=this.data;for(let t=0;t<e.length;++t)e[t]=1/(1+Math.exp(-e[t]));return this}mul(e){return this.clone().mul_(e)}mul_(e){const t=this.data;for(let r=0;r<t.length;++r)t[r]*=e;return this}div(e){return this.clone().div_(e)}div_(e){const t=this.data;for(let r=0;r<t.length;++r)t[r]/=e;return this}add(e){return this.clone().add_(e)}add_(e){const t=this.data;for(let r=0;r<t.length;++r)t[r]+=e;return this}sub(e){return this.clone().sub_(e)}sub_(e){const t=this.data;for(let r=0;r<t.length;++r)t[r]-=e;return this}clone(){return new i(this.type,this.data.slice(),this.dims.slice())}slice(...e){const t=[],r=[];for(let n=0;n<this.dims.length;++n){let o=e[n];if(null==o)r.push([0,this.dims[n]]),t.push(this.dims[n]);else if("number"==typeof o)o=M(o,this.dims[n],n),r.push([o,o+1]);else{if(!Array.isArray(o)||2!==o.length)throw new Error(`Invalid slice: ${o}`);{let[e,s]=o;if(e=null===e?0:M(e,this.dims[n],n,!1),s=null===s?this.dims[n]:M(s,this.dims[n],n,!1),e>s)throw new Error(`Invalid slice: ${o}`);const a=[Math.max(e,0),Math.min(s,this.dims[n])];r.push(a),t.push(a[1]-a[0])}}}const n=r.map((([e,t])=>t-e)),o=n.reduce(((e,t)=>e*t)),s=this.data,a=new s.constructor(o),l=this.stride();for(let e=0;e<o;++e){let t=0;for(let o=n.length-1,s=e;o>=0;--o){const e=n[o];t+=(s%e+r[o][0])*l[o],s=Math.floor(s/e)}a[e]=s[t]}return new i(this.type,a,t)}permute(...e){return l(this,e)}transpose(...e){return this.permute(...e)}sum(e=null,t=!1){return this.norm(1,e,t)}norm(e="fro",t=null,r=!1){if("fro"===e)e=2;else if("string"==typeof e)throw Error(`Unsupported norm: ${e}`);const n=this.data;if(null===t){let t=n.reduce(((t,r)=>t+r**e),0)**(1/e);return new i(this.type,[t],[])}t=M(t,this.dims.length);const o=this.dims.slice();o[t]=1;const s=new n.constructor(n.length/this.dims[t]);for(let r=0;r<n.length;++r){let a=0;for(let e=this.dims.length-1,n=r,s=1;e>=0;--e){const r=this.dims[e];if(e!==t){a+=n%r*s,s*=o[e]}n=Math.floor(n/r)}s[a]+=n[r]**e}if(1!==e)for(let t=0;t<s.length;++t)s[t]=s[t]**(1/e);return r||o.splice(t,1),new i(this.type,s,o)}normalize_(e=2,t=1){t=M(t,this.dims.length);const r=this.norm(e,t,!0),n=this.data,o=r.data;for(let e=0;e<n.length;++e){let r=0;for(let n=this.dims.length-1,o=e,s=1;n>=0;--n){const e=this.dims[n];if(n!==t){r+=o%e*s,s*=this.dims[n]}o=Math.floor(o/e)}n[e]/=o[r]}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 r=e.length-1,n=1;r>=0;--r)t[r]=n,n*=e[r];return t}(this.dims)}squeeze(e=null){return new i(this.type,this.data,f(this.dims,e))}squeeze_(e=null){return this.dims=f(this.dims,e),this}unsqueeze(e=null){return new i(this.type,this.data,g(this.dims,e))}unsqueeze_(e=null){return this.dims=g(this.dims,e),this}flatten_(e=0,t=-1){t=(t+this.dims.length)%this.dims.length;let r=this.dims.slice(0,e),n=this.dims.slice(e,t+1),o=this.dims.slice(t+1);return this.dims=[...r,n.reduce(((e,t)=>e*t),1),...o],this}flatten(e=0,t=-1){return this.clone().flatten_(e,t)}view(...e){let t=-1;for(let r=0;r<e.length;++r)if(-1===e[r]){if(-1!==t)throw new Error("Only one dimension can be inferred");t=r}const r=this.data;if(-1!==t){const n=e.reduce(((e,r,n)=>n!==t?e*r:e),1);e[t]=r.length/n}return new i(this.type,r,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 r=this.data;for(let n=0;n<r.length;++n)r[n]=Math.min(Math.max(r[n],e),t);return this}clamp(e,t){return this.clone().clamp_(e,t)}round_(){const e=this.data;for(let t=0;t<e.length;++t)e[t]=Math.round(e[t]);return this}round(){return this.clone().round_()}mean(e=null,t=!1){return x(this,e,t)}to(e){if(this.type===e)return this;if(!a.hasOwnProperty(e))throw new Error(`Unsupported type: ${e}`);return new i(e,a[e].from(this.data),this.dims)}}function l(e,t){const[r,o]=(0,n.permute_data)(e.data,e.dims,t);return new i(e.type,r,o)}function c(e,[t,r],o="bilinear",s=!1){const a=e.dims.at(-3)??1,l=e.dims.at(-2),c=e.dims.at(-1);let d=(0,n.interpolate_data)(e.data,[a,l,c],[t,r],o,s);return new i(e.type,d,[a,t,r])}async function d(e,{size:t=null,mode:r="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 n,o;if(2===t.length)n=[...e.dims.slice(0,2),...t];else if(3===t.length)n=[e.dims[0],...t];else{if(4!==t.length)throw new Error("`size` must be of length 2, 3, or 4.");n=t}if("bilinear"===r)o=await s.TensorOpRegistry.bilinear_interpolate_4d;else{if("bicubic"!==r)throw new Error(`Unsupported mode: ${r}`);o=await s.TensorOpRegistry.bicubic_interpolate_4d}const a=new i("int64",new BigInt64Array(n.map(BigInt)),[n.length]);return await o({x:e,s:a})}async function u(e,t){const r=await s.TensorOpRegistry.matmul;return await r({a:e,b:t})}async function p(e,t){const r=await s.TensorOpRegistry.rfft;return await r({x:e,a:t})}async function h(e,t){const r=await s.TensorOpRegistry.top_k;return t=null===t?e.dims.at(-1):Math.min(t,e.dims.at(-1)),await r({x:e,k:new i("int64",[BigInt(t)],[1])})}function m(e,t){const r=e.data,n=t.data,o=[e.dims[0],e.dims[2]],s=new r.constructor(o[0]*o[1]),[a,l,c]=e.dims;let d=0;for(let e=0;e<a;++e){const t=e*c*l;for(let o=0;o<c;++o){let a=0,i=0;const u=e*l,p=t+o;for(let e=0;e<l;++e){const t=Number(n[u+e]);i+=t,a+=r[p+e*c]*t}const h=a/i;s[d++]=h}}return new i(e.type,s,o)}function _(e,t,{eps:r=1e-5}={}){if(2!==e.dims.length)throw new Error("`layer_norm` currently only supports 2D input.");const[n,o]=e.dims;if(1!==t.length&&t[0]!==o)throw new Error("`normalized_shape` must be a 1D array with shape `[input.dims[1]]`.");const[s,a]=b(e,1,0,!0),l=s.data,c=a.data,d=e.data,u=new d.constructor(d.length);for(let e=0;e<n;++e){const t=e*o;for(let n=0;n<o;++n){const o=t+n;u[o]=(d[o]-c[e])/(l[e]+r)}}return new i(e.type,u,e.dims)}function f(e,t){return e=e.slice(),null===t?e=e.filter((e=>1!==e)):"number"==typeof t?1===e[t]&&e.splice(t,1):Array.isArray(t)&&(e=e.filter(((e,r)=>1!==e||!t.includes(r)))),e}function g(e,t){return t=M(t,e.length+1),(e=e.slice()).splice(t,0,1),e}function M(e,t,r=null,n=!0){if(n&&(e<-t||e>=t))throw new Error(`IndexError: index ${e} is out of bounds for dimension${null===r?"":" "+r} with size ${t}`);return e<0&&(e=(e%t+t)%t),e}function w(e,t=0){t=M(t,e[0].dims.length);const r=e[0].dims.slice();r[t]=e.reduce(((e,r)=>e+r.dims[t]),0);const n=r.reduce(((e,t)=>e*t),1),o=new e[0].data.constructor(n),s=e[0].type;if(0===t){let t=0;for(const r of e){const e=r.data;o.set(e,t),t+=e.length}}else{let n=0;for(let s=0;s<e.length;++s){const{data:a,dims:i}=e[s];for(let e=0;e<a.length;++e){let s=0;for(let o=i.length-1,a=e,l=1;o>=0;--o){const e=i[o];let c=a%e;o===t&&(c+=n),s+=c*l,l*=r[o],a=Math.floor(a/e)}o[s]=a[e]}n+=i[t]}}return new i(s,o,r)}function T(e,t=0){return w(e.map((e=>e.unsqueeze(t))),t)}function b(e,t=null,r=1,n=!1){const o=e.data,s=e.dims;if(null===t){const t=o.reduce(((e,t)=>e+t),0)/o.length,n=Math.sqrt(o.reduce(((e,r)=>e+(r-t)**2),0)/(o.length-r)),s=new i(e.type,[t],[]);return[new i(e.type,[n],[]),s]}const a=x(e,t=M(t,s.length),n),l=a.data,c=s.slice();c[t]=1;const d=new o.constructor(o.length/s[t]);for(let e=0;e<o.length;++e){let r=0;for(let n=s.length-1,o=e,a=1;n>=0;--n){const e=s[n];if(n!==t){r+=o%e*a,a*=c[n]}o=Math.floor(o/e)}d[r]+=(o[e]-l[r])**2}for(let e=0;e<d.length;++e)d[e]=Math.sqrt(d[e]/(s[t]-r));n||c.splice(t,1);return[new i(e.type,d,c),a]}function x(e,t=null,r=!1){const n=e.data;if(null===t){const t=n.reduce(((e,t)=>e+t),0);return new i(e.type,[t/n.length],[])}const o=e.dims;t=M(t,o.length);const s=o.slice();s[t]=1;const a=new n.constructor(n.length/o[t]);for(let e=0;e<n.length;++e){let r=0;for(let n=o.length-1,a=e,i=1;n>=0;--n){const e=o[n];if(n!==t){r+=a%e*i,i*=s[n]}a=Math.floor(a/e)}a[r]+=n[e]}if(1!==o[t])for(let e=0;e<a.length;++e)a[e]=a[e]/o[t];return r||s.splice(t,1),new i(e.type,a,s)}function y(e,t,r,n){const o=e.reduce(((e,t)=>e*t),1);return new i(r,new n(o).fill(t),e)}function k(e,t){let r,n;if("number"==typeof t)r="float32",n=Float32Array;else{if("bigint"!=typeof t)throw new Error("Unsupported data type: "+typeof t);r="int64",n=BigInt64Array}return y(e,t,r,n)}function F(e,t){return k(e.dims,t)}function P(e){return y(e,1n,"int64",BigInt64Array)}function C(e){return P(e.dims)}function v(e){return y(e,0n,"int64",BigInt64Array)}function S(e){return v(e.dims)}function A(e,t){if(2!==e.dims.length)throw new Error("The tensor must have 2 dimensions");if(e.dims.at(-1)%8!=0)throw new Error("The last dimension of the tensor must be a multiple of 8");if(!["binary","ubinary"].includes(t))throw new Error("The precision must be either 'binary' or 'ubinary'");const r="binary"===t,n=r?"int8":"uint8",o=r?Int8Array:Uint8Array,s=e.data,a=new o(s.length/8);for(let e=0;e<s.length;++e){const t=s[e]>0?1:0,n=Math.floor(e/8),o=e%8;a[n]|=t<<7-o,r&&0===o&&(a[n]-=128)}return new i(n,a,[e.dims[0],e.dims[1]/8])}}},l={};function c(e){var t=l[e];if(void 0!==t)return t.exports;var r=l[e]={exports:{}};return i[e](r,r.exports,c),r.exports}a=Object.getPrototypeOf?e=>Object.getPrototypeOf(e):e=>e.__proto__,c.t=function(e,t){if(1&t&&(e=this(e)),8&t)return e;if("object"==typeof e&&e){if(4&t&&e.__esModule)return e;if(16&t&&"function"==typeof e.then)return e}var r=Object.create(null);c.r(r);var n={};s=s||[null,a({}),a([]),a(a)];for(var o=2&t&&e;"object"==typeof o&&!~s.indexOf(o);o=a(o))Object.getOwnPropertyNames(o).forEach((t=>n[t]=()=>e[t]));return n.default=()=>e,c.d(r,n),r},c.d=(e,t)=>{for(var r in t)c.o(t,r)&&!c.o(e,r)&&Object.defineProperty(e,r,{enumerable:!0,get:t[r]})},c.o=(e,t)=>Object.prototype.hasOwnProperty.call(e,t),c.r=e=>{"undefined"!=typeof Symbol&&Symbol.toStringTag&&Object.defineProperty(e,Symbol.toStringTag,{value:"Module"}),Object.defineProperty(e,"__esModule",{value:!0})};var d={};
+  \*****************************/(e,t,r)=>{r.r(t),r.d(t,{Tensor:()=>i,cat:()=>w,full:()=>k,full_like:()=>F,interpolate:()=>c,interpolate_4d:()=>d,layer_norm:()=>_,matmul:()=>u,mean:()=>x,mean_pooling:()=>m,ones:()=>P,ones_like:()=>C,permute:()=>l,quantize_embeddings:()=>A,rfft:()=>p,stack:()=>T,std_mean:()=>b,topk:()=>h,zeros:()=>v,zeros_like:()=>S});var n=r(/*! ./maths.js */"./src/utils/maths.js"),o=r(/*! ../backends/onnx.js */"./src/backends/onnx.js"),s=r(/*! ../ops/registry.js */"./src/ops/registry.js");const a=Object.freeze({float32:Float32Array,float16:Uint16Array,float64:Float64Array,string:Array,int8:Int8Array,uint8:Uint8Array,int16:Int16Array,uint16:Uint16Array,int32:Int32Array,uint32:Uint32Array,int64:BigInt64Array,uint64:BigUint64Array,bool:Uint8Array});class i{get dims(){return this.ort_tensor.dims}set dims(e){this.ort_tensor.dims=e}get type(){return this.ort_tensor.type}get data(){return this.ort_tensor.data}get size(){return this.ort_tensor.size}get location(){return this.ort_tensor.location}ort_tensor;constructor(...e){return(0,o.isONNXTensor)(e[0])?this.ort_tensor=e[0]:this.ort_tensor=new o.Tensor(e[0],e[1],e[2]),new Proxy(this,{get:(e,t)=>{if("string"==typeof t){let r=Number(t);if(Number.isInteger(r))return e._getitem(r)}return e[t]},set:(e,t,r)=>e[t]=r})}dispose(){this.ort_tensor.dispose()}*[Symbol.iterator](){const[e,...t]=this.dims;if(t.length>0){const r=t.reduce(((e,t)=>e*t));for(let n=0;n<e;++n)yield this._subarray(n,r,t)}else yield*this.data}_getitem(e){const[t,...r]=this.dims;if(e=M(e,t),r.length>0){const t=r.reduce(((e,t)=>e*t));return this._subarray(e,t,r)}return new i(this.type,[this.data[e]],r)}indexOf(e){const t=this.data;for(let r=0;r<t.length;++r)if(t[r]==e)return r;return-1}_subarray(e,t,r){const n=e*t,o=(e+1)*t,s="subarray"in this.data?this.data.subarray(n,o):this.data.slice(n,o);return new i(this.type,s,r)}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 r=e.length,n=t.reduce(((e,t)=>e*t));if(r!==n)throw Error(`cannot reshape array of size ${r} into shape (${t})`);let o=e;for(let e=t.length-1;e>=0;e--)o=o.reduce(((r,n)=>{let o=r[r.length-1];return o.length<t[e]?o.push(n):r.push([n]),r}),[[]]);return o[0]}(this.data,this.dims)}sigmoid(){return this.clone().sigmoid_()}sigmoid_(){const e=this.data;for(let t=0;t<e.length;++t)e[t]=1/(1+Math.exp(-e[t]));return this}map(e){return this.clone().map_(e)}map_(e){const t=this.data;for(let r=0;r<t.length;++r)t[r]=e(t[r],r,t);return this}mul(e){return this.clone().mul_(e)}mul_(e){const t=this.data;for(let r=0;r<t.length;++r)t[r]*=e;return this}div(e){return this.clone().div_(e)}div_(e){const t=this.data;for(let r=0;r<t.length;++r)t[r]/=e;return this}add(e){return this.clone().add_(e)}add_(e){const t=this.data;for(let r=0;r<t.length;++r)t[r]+=e;return this}sub(e){return this.clone().sub_(e)}sub_(e){const t=this.data;for(let r=0;r<t.length;++r)t[r]-=e;return this}clone(){return new i(this.type,this.data.slice(),this.dims.slice())}slice(...e){const t=[],r=[];for(let n=0;n<this.dims.length;++n){let o=e[n];if(null==o)r.push([0,this.dims[n]]),t.push(this.dims[n]);else if("number"==typeof o)o=M(o,this.dims[n],n),r.push([o,o+1]);else{if(!Array.isArray(o)||2!==o.length)throw new Error(`Invalid slice: ${o}`);{let[e,s]=o;if(e=null===e?0:M(e,this.dims[n],n,!1),s=null===s?this.dims[n]:M(s,this.dims[n],n,!1),e>s)throw new Error(`Invalid slice: ${o}`);const a=[Math.max(e,0),Math.min(s,this.dims[n])];r.push(a),t.push(a[1]-a[0])}}}const n=r.map((([e,t])=>t-e)),o=n.reduce(((e,t)=>e*t)),s=this.data,a=new s.constructor(o),l=this.stride();for(let e=0;e<o;++e){let t=0;for(let o=n.length-1,s=e;o>=0;--o){const e=n[o];t+=(s%e+r[o][0])*l[o],s=Math.floor(s/e)}a[e]=s[t]}return new i(this.type,a,t)}permute(...e){return l(this,e)}transpose(...e){return this.permute(...e)}sum(e=null,t=!1){return this.norm(1,e,t)}norm(e="fro",t=null,r=!1){if("fro"===e)e=2;else if("string"==typeof e)throw Error(`Unsupported norm: ${e}`);const n=this.data;if(null===t){let t=n.reduce(((t,r)=>t+r**e),0)**(1/e);return new i(this.type,[t],[])}t=M(t,this.dims.length);const o=this.dims.slice();o[t]=1;const s=new n.constructor(n.length/this.dims[t]);for(let r=0;r<n.length;++r){let a=0;for(let e=this.dims.length-1,n=r,s=1;e>=0;--e){const r=this.dims[e];if(e!==t){a+=n%r*s,s*=o[e]}n=Math.floor(n/r)}s[a]+=n[r]**e}if(1!==e)for(let t=0;t<s.length;++t)s[t]=s[t]**(1/e);return r||o.splice(t,1),new i(this.type,s,o)}normalize_(e=2,t=1){t=M(t,this.dims.length);const r=this.norm(e,t,!0),n=this.data,o=r.data;for(let e=0;e<n.length;++e){let r=0;for(let n=this.dims.length-1,o=e,s=1;n>=0;--n){const e=this.dims[n];if(n!==t){r+=o%e*s,s*=this.dims[n]}o=Math.floor(o/e)}n[e]/=o[r]}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 r=e.length-1,n=1;r>=0;--r)t[r]=n,n*=e[r];return t}(this.dims)}squeeze(e=null){return new i(this.type,this.data,f(this.dims,e))}squeeze_(e=null){return this.dims=f(this.dims,e),this}unsqueeze(e=null){return new i(this.type,this.data,g(this.dims,e))}unsqueeze_(e=null){return this.dims=g(this.dims,e),this}flatten_(e=0,t=-1){t=(t+this.dims.length)%this.dims.length;let r=this.dims.slice(0,e),n=this.dims.slice(e,t+1),o=this.dims.slice(t+1);return this.dims=[...r,n.reduce(((e,t)=>e*t),1),...o],this}flatten(e=0,t=-1){return this.clone().flatten_(e,t)}view(...e){let t=-1;for(let r=0;r<e.length;++r)if(-1===e[r]){if(-1!==t)throw new Error("Only one dimension can be inferred");t=r}const r=this.data;if(-1!==t){const n=e.reduce(((e,r,n)=>n!==t?e*r:e),1);e[t]=r.length/n}return new i(this.type,r,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 r=this.data;for(let n=0;n<r.length;++n)r[n]=Math.min(Math.max(r[n],e),t);return this}clamp(e,t){return this.clone().clamp_(e,t)}round_(){const e=this.data;for(let t=0;t<e.length;++t)e[t]=Math.round(e[t]);return this}round(){return this.clone().round_()}mean(e=null,t=!1){return x(this,e,t)}to(e){if(this.type===e)return this;if(!a.hasOwnProperty(e))throw new Error(`Unsupported type: ${e}`);return new i(e,a[e].from(this.data),this.dims)}}function l(e,t){const[r,o]=(0,n.permute_data)(e.data,e.dims,t);return new i(e.type,r,o)}function c(e,[t,r],o="bilinear",s=!1){const a=e.dims.at(-3)??1,l=e.dims.at(-2),c=e.dims.at(-1);let d=(0,n.interpolate_data)(e.data,[a,l,c],[t,r],o,s);return new i(e.type,d,[a,t,r])}async function d(e,{size:t=null,mode:r="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 n,o;if(2===t.length)n=[...e.dims.slice(0,2),...t];else if(3===t.length)n=[e.dims[0],...t];else{if(4!==t.length)throw new Error("`size` must be of length 2, 3, or 4.");n=t}if("bilinear"===r)o=await s.TensorOpRegistry.bilinear_interpolate_4d;else{if("bicubic"!==r)throw new Error(`Unsupported mode: ${r}`);o=await s.TensorOpRegistry.bicubic_interpolate_4d}const a=new i("int64",new BigInt64Array(n.map(BigInt)),[n.length]);return await o({x:e,s:a})}async function u(e,t){const r=await s.TensorOpRegistry.matmul;return await r({a:e,b:t})}async function p(e,t){const r=await s.TensorOpRegistry.rfft;return await r({x:e,a:t})}async function h(e,t){const r=await s.TensorOpRegistry.top_k;return t=null===t?e.dims.at(-1):Math.min(t,e.dims.at(-1)),await r({x:e,k:new i("int64",[BigInt(t)],[1])})}function m(e,t){const r=e.data,n=t.data,o=[e.dims[0],e.dims[2]],s=new r.constructor(o[0]*o[1]),[a,l,c]=e.dims;let d=0;for(let e=0;e<a;++e){const t=e*c*l;for(let o=0;o<c;++o){let a=0,i=0;const u=e*l,p=t+o;for(let e=0;e<l;++e){const t=Number(n[u+e]);i+=t,a+=r[p+e*c]*t}const h=a/i;s[d++]=h}}return new i(e.type,s,o)}function _(e,t,{eps:r=1e-5}={}){if(2!==e.dims.length)throw new Error("`layer_norm` currently only supports 2D input.");const[n,o]=e.dims;if(1!==t.length&&t[0]!==o)throw new Error("`normalized_shape` must be a 1D array with shape `[input.dims[1]]`.");const[s,a]=b(e,1,0,!0),l=s.data,c=a.data,d=e.data,u=new d.constructor(d.length);for(let e=0;e<n;++e){const t=e*o;for(let n=0;n<o;++n){const o=t+n;u[o]=(d[o]-c[e])/(l[e]+r)}}return new i(e.type,u,e.dims)}function f(e,t){return e=e.slice(),null===t?e=e.filter((e=>1!==e)):"number"==typeof t?1===e[t]&&e.splice(t,1):Array.isArray(t)&&(e=e.filter(((e,r)=>1!==e||!t.includes(r)))),e}function g(e,t){return t=M(t,e.length+1),(e=e.slice()).splice(t,0,1),e}function M(e,t,r=null,n=!0){if(n&&(e<-t||e>=t))throw new Error(`IndexError: index ${e} is out of bounds for dimension${null===r?"":" "+r} with size ${t}`);return e<0&&(e=(e%t+t)%t),e}function w(e,t=0){t=M(t,e[0].dims.length);const r=e[0].dims.slice();r[t]=e.reduce(((e,r)=>e+r.dims[t]),0);const n=r.reduce(((e,t)=>e*t),1),o=new e[0].data.constructor(n),s=e[0].type;if(0===t){let t=0;for(const r of e){const e=r.data;o.set(e,t),t+=e.length}}else{let n=0;for(let s=0;s<e.length;++s){const{data:a,dims:i}=e[s];for(let e=0;e<a.length;++e){let s=0;for(let o=i.length-1,a=e,l=1;o>=0;--o){const e=i[o];let c=a%e;o===t&&(c+=n),s+=c*l,l*=r[o],a=Math.floor(a/e)}o[s]=a[e]}n+=i[t]}}return new i(s,o,r)}function T(e,t=0){return w(e.map((e=>e.unsqueeze(t))),t)}function b(e,t=null,r=1,n=!1){const o=e.data,s=e.dims;if(null===t){const t=o.reduce(((e,t)=>e+t),0)/o.length,n=Math.sqrt(o.reduce(((e,r)=>e+(r-t)**2),0)/(o.length-r)),s=new i(e.type,[t],[]);return[new i(e.type,[n],[]),s]}const a=x(e,t=M(t,s.length),n),l=a.data,c=s.slice();c[t]=1;const d=new o.constructor(o.length/s[t]);for(let e=0;e<o.length;++e){let r=0;for(let n=s.length-1,o=e,a=1;n>=0;--n){const e=s[n];if(n!==t){r+=o%e*a,a*=c[n]}o=Math.floor(o/e)}d[r]+=(o[e]-l[r])**2}for(let e=0;e<d.length;++e)d[e]=Math.sqrt(d[e]/(s[t]-r));n||c.splice(t,1);return[new i(e.type,d,c),a]}function x(e,t=null,r=!1){const n=e.data;if(null===t){const t=n.reduce(((e,t)=>e+t),0);return new i(e.type,[t/n.length],[])}const o=e.dims;t=M(t,o.length);const s=o.slice();s[t]=1;const a=new n.constructor(n.length/o[t]);for(let e=0;e<n.length;++e){let r=0;for(let n=o.length-1,a=e,i=1;n>=0;--n){const e=o[n];if(n!==t){r+=a%e*i,i*=s[n]}a=Math.floor(a/e)}a[r]+=n[e]}if(1!==o[t])for(let e=0;e<a.length;++e)a[e]=a[e]/o[t];return r||s.splice(t,1),new i(e.type,a,s)}function y(e,t,r,n){const o=e.reduce(((e,t)=>e*t),1);return new i(r,new n(o).fill(t),e)}function k(e,t){let r,n;if("number"==typeof t)r="float32",n=Float32Array;else{if("bigint"!=typeof t)throw new Error("Unsupported data type: "+typeof t);r="int64",n=BigInt64Array}return y(e,t,r,n)}function F(e,t){return k(e.dims,t)}function P(e){return y(e,1n,"int64",BigInt64Array)}function C(e){return P(e.dims)}function v(e){return y(e,0n,"int64",BigInt64Array)}function S(e){return v(e.dims)}function A(e,t){if(2!==e.dims.length)throw new Error("The tensor must have 2 dimensions");if(e.dims.at(-1)%8!=0)throw new Error("The last dimension of the tensor must be a multiple of 8");if(!["binary","ubinary"].includes(t))throw new Error("The precision must be either 'binary' or 'ubinary'");const r="binary"===t,n=r?"int8":"uint8",o=r?Int8Array:Uint8Array,s=e.data,a=new o(s.length/8);for(let e=0;e<s.length;++e){const t=s[e]>0?1:0,n=Math.floor(e/8),o=e%8;a[n]|=t<<7-o,r&&0===o&&(a[n]-=128)}return new i(n,a,[e.dims[0],e.dims[1]/8])}}},l={};function c(e){var t=l[e];if(void 0!==t)return t.exports;var r=l[e]={exports:{}};return i[e](r,r.exports,c),r.exports}a=Object.getPrototypeOf?e=>Object.getPrototypeOf(e):e=>e.__proto__,c.t=function(e,t){if(1&t&&(e=this(e)),8&t)return e;if("object"==typeof e&&e){if(4&t&&e.__esModule)return e;if(16&t&&"function"==typeof e.then)return e}var r=Object.create(null);c.r(r);var n={};s=s||[null,a({}),a([]),a(a)];for(var o=2&t&&e;"object"==typeof o&&!~s.indexOf(o);o=a(o))Object.getOwnPropertyNames(o).forEach((t=>n[t]=()=>e[t]));return n.default=()=>e,c.d(r,n),r},c.d=(e,t)=>{for(var r in t)c.o(t,r)&&!c.o(e,r)&&Object.defineProperty(e,r,{enumerable:!0,get:t[r]})},c.o=(e,t)=>Object.prototype.hasOwnProperty.call(e,t),c.r=e=>{"undefined"!=typeof Symbol&&Symbol.toStringTag&&Object.defineProperty(e,Symbol.toStringTag,{value:"Module"}),Object.defineProperty(e,"__esModule",{value:!0})};var d={};
 /*!*****************************!*\
   !*** ./src/transformers.js ***!
-  \*****************************/c.r(d),c.d(d,{ASTFeatureExtractor:()=>_.ASTFeatureExtractor,ASTForAudioClassification:()=>h.ASTForAudioClassification,ASTModel:()=>h.ASTModel,ASTPreTrainedModel:()=>h.ASTPreTrainedModel,AlbertForMaskedLM:()=>h.AlbertForMaskedLM,AlbertForQuestionAnswering:()=>h.AlbertForQuestionAnswering,AlbertForSequenceClassification:()=>h.AlbertForSequenceClassification,AlbertModel:()=>h.AlbertModel,AlbertPreTrainedModel:()=>h.AlbertPreTrainedModel,AlbertTokenizer:()=>m.AlbertTokenizer,AudioClassificationPipeline:()=>p.AudioClassificationPipeline,AutoConfig:()=>f.AutoConfig,AutoModel:()=>h.AutoModel,AutoModelForAudioClassification:()=>h.AutoModelForAudioClassification,AutoModelForAudioFrameClassification:()=>h.AutoModelForAudioFrameClassification,AutoModelForCTC:()=>h.AutoModelForCTC,AutoModelForCausalLM:()=>h.AutoModelForCausalLM,AutoModelForDepthEstimation:()=>h.AutoModelForDepthEstimation,AutoModelForDocumentQuestionAnswering:()=>h.AutoModelForDocumentQuestionAnswering,AutoModelForImageClassification:()=>h.AutoModelForImageClassification,AutoModelForImageFeatureExtraction:()=>h.AutoModelForImageFeatureExtraction,AutoModelForImageMatting:()=>h.AutoModelForImageMatting,AutoModelForImageSegmentation:()=>h.AutoModelForImageSegmentation,AutoModelForImageToImage:()=>h.AutoModelForImageToImage,AutoModelForMaskGeneration:()=>h.AutoModelForMaskGeneration,AutoModelForMaskedLM:()=>h.AutoModelForMaskedLM,AutoModelForNormalEstimation:()=>h.AutoModelForNormalEstimation,AutoModelForObjectDetection:()=>h.AutoModelForObjectDetection,AutoModelForQuestionAnswering:()=>h.AutoModelForQuestionAnswering,AutoModelForSemanticSegmentation:()=>h.AutoModelForSemanticSegmentation,AutoModelForSeq2SeqLM:()=>h.AutoModelForSeq2SeqLM,AutoModelForSequenceClassification:()=>h.AutoModelForSequenceClassification,AutoModelForSpeechSeq2Seq:()=>h.AutoModelForSpeechSeq2Seq,AutoModelForTextToSpectrogram:()=>h.AutoModelForTextToSpectrogram,AutoModelForTextToWaveform:()=>h.AutoModelForTextToWaveform,AutoModelForTokenClassification:()=>h.AutoModelForTokenClassification,AutoModelForUniversalSegmentation:()=>h.AutoModelForUniversalSegmentation,AutoModelForVision2Seq:()=>h.AutoModelForVision2Seq,AutoModelForXVector:()=>h.AutoModelForXVector,AutoModelForZeroShotObjectDetection:()=>h.AutoModelForZeroShotObjectDetection,AutoProcessor:()=>_.AutoProcessor,AutoTokenizer:()=>m.AutoTokenizer,AutomaticSpeechRecognitionPipeline:()=>p.AutomaticSpeechRecognitionPipeline,BartForConditionalGeneration:()=>h.BartForConditionalGeneration,BartForSequenceClassification:()=>h.BartForSequenceClassification,BartModel:()=>h.BartModel,BartPretrainedModel:()=>h.BartPretrainedModel,BartTokenizer:()=>m.BartTokenizer,BaseModelOutput:()=>h.BaseModelOutput,BaseStreamer:()=>b.BaseStreamer,BeitFeatureExtractor:()=>_.BeitFeatureExtractor,BeitForImageClassification:()=>h.BeitForImageClassification,BeitModel:()=>h.BeitModel,BeitPreTrainedModel:()=>h.BeitPreTrainedModel,BertForMaskedLM:()=>h.BertForMaskedLM,BertForQuestionAnswering:()=>h.BertForQuestionAnswering,BertForSequenceClassification:()=>h.BertForSequenceClassification,BertForTokenClassification:()=>h.BertForTokenClassification,BertModel:()=>h.BertModel,BertPreTrainedModel:()=>h.BertPreTrainedModel,BertTokenizer:()=>m.BertTokenizer,BitImageProcessor:()=>_.BitImageProcessor,BlenderbotForConditionalGeneration:()=>h.BlenderbotForConditionalGeneration,BlenderbotModel:()=>h.BlenderbotModel,BlenderbotPreTrainedModel:()=>h.BlenderbotPreTrainedModel,BlenderbotSmallForConditionalGeneration:()=>h.BlenderbotSmallForConditionalGeneration,BlenderbotSmallModel:()=>h.BlenderbotSmallModel,BlenderbotSmallPreTrainedModel:()=>h.BlenderbotSmallPreTrainedModel,BlenderbotSmallTokenizer:()=>m.BlenderbotSmallTokenizer,BlenderbotTokenizer:()=>m.BlenderbotTokenizer,BloomForCausalLM:()=>h.BloomForCausalLM,BloomModel:()=>h.BloomModel,BloomPreTrainedModel:()=>h.BloomPreTrainedModel,BloomTokenizer:()=>m.BloomTokenizer,CLIPFeatureExtractor:()=>_.CLIPFeatureExtractor,CLIPImageProcessor:()=>_.CLIPImageProcessor,CLIPModel:()=>h.CLIPModel,CLIPPreTrainedModel:()=>h.CLIPPreTrainedModel,CLIPSegForImageSegmentation:()=>h.CLIPSegForImageSegmentation,CLIPSegModel:()=>h.CLIPSegModel,CLIPSegPreTrainedModel:()=>h.CLIPSegPreTrainedModel,CLIPTextModel:()=>h.CLIPTextModel,CLIPTextModelWithProjection:()=>h.CLIPTextModelWithProjection,CLIPTokenizer:()=>m.CLIPTokenizer,CLIPVisionModel:()=>h.CLIPVisionModel,CLIPVisionModelWithProjection:()=>h.CLIPVisionModelWithProjection,CamembertForMaskedLM:()=>h.CamembertForMaskedLM,CamembertForQuestionAnswering:()=>h.CamembertForQuestionAnswering,CamembertForSequenceClassification:()=>h.CamembertForSequenceClassification,CamembertForTokenClassification:()=>h.CamembertForTokenClassification,CamembertModel:()=>h.CamembertModel,CamembertPreTrainedModel:()=>h.CamembertPreTrainedModel,CamembertTokenizer:()=>m.CamembertTokenizer,CausalLMOutput:()=>h.CausalLMOutput,CausalLMOutputWithPast:()=>h.CausalLMOutputWithPast,ChineseCLIPFeatureExtractor:()=>_.ChineseCLIPFeatureExtractor,ChineseCLIPModel:()=>h.ChineseCLIPModel,ChineseCLIPPreTrainedModel:()=>h.ChineseCLIPPreTrainedModel,ClapAudioModelWithProjection:()=>h.ClapAudioModelWithProjection,ClapFeatureExtractor:()=>_.ClapFeatureExtractor,ClapModel:()=>h.ClapModel,ClapPreTrainedModel:()=>h.ClapPreTrainedModel,ClapTextModelWithProjection:()=>h.ClapTextModelWithProjection,CodeGenForCausalLM:()=>h.CodeGenForCausalLM,CodeGenModel:()=>h.CodeGenModel,CodeGenPreTrainedModel:()=>h.CodeGenPreTrainedModel,CodeGenTokenizer:()=>m.CodeGenTokenizer,CodeLlamaTokenizer:()=>m.CodeLlamaTokenizer,CohereForCausalLM:()=>h.CohereForCausalLM,CohereModel:()=>h.CohereModel,CoherePreTrainedModel:()=>h.CoherePreTrainedModel,CohereTokenizer:()=>m.CohereTokenizer,ConvBertForMaskedLM:()=>h.ConvBertForMaskedLM,ConvBertForQuestionAnswering:()=>h.ConvBertForQuestionAnswering,ConvBertForSequenceClassification:()=>h.ConvBertForSequenceClassification,ConvBertForTokenClassification:()=>h.ConvBertForTokenClassification,ConvBertModel:()=>h.ConvBertModel,ConvBertPreTrainedModel:()=>h.ConvBertPreTrainedModel,ConvBertTokenizer:()=>m.ConvBertTokenizer,ConvNextFeatureExtractor:()=>_.ConvNextFeatureExtractor,ConvNextForImageClassification:()=>h.ConvNextForImageClassification,ConvNextImageProcessor:()=>_.ConvNextImageProcessor,ConvNextModel:()=>h.ConvNextModel,ConvNextPreTrainedModel:()=>h.ConvNextPreTrainedModel,ConvNextV2ForImageClassification:()=>h.ConvNextV2ForImageClassification,ConvNextV2Model:()=>h.ConvNextV2Model,ConvNextV2PreTrainedModel:()=>h.ConvNextV2PreTrainedModel,DPTFeatureExtractor:()=>_.DPTFeatureExtractor,DPTForDepthEstimation:()=>h.DPTForDepthEstimation,DPTImageProcessor:()=>_.DPTImageProcessor,DPTModel:()=>h.DPTModel,DPTPreTrainedModel:()=>h.DPTPreTrainedModel,DebertaForMaskedLM:()=>h.DebertaForMaskedLM,DebertaForQuestionAnswering:()=>h.DebertaForQuestionAnswering,DebertaForSequenceClassification:()=>h.DebertaForSequenceClassification,DebertaForTokenClassification:()=>h.DebertaForTokenClassification,DebertaModel:()=>h.DebertaModel,DebertaPreTrainedModel:()=>h.DebertaPreTrainedModel,DebertaTokenizer:()=>m.DebertaTokenizer,DebertaV2ForMaskedLM:()=>h.DebertaV2ForMaskedLM,DebertaV2ForQuestionAnswering:()=>h.DebertaV2ForQuestionAnswering,DebertaV2ForSequenceClassification:()=>h.DebertaV2ForSequenceClassification,DebertaV2ForTokenClassification:()=>h.DebertaV2ForTokenClassification,DebertaV2Model:()=>h.DebertaV2Model,DebertaV2PreTrainedModel:()=>h.DebertaV2PreTrainedModel,DebertaV2Tokenizer:()=>m.DebertaV2Tokenizer,DecisionTransformerModel:()=>h.DecisionTransformerModel,DecisionTransformerPreTrainedModel:()=>h.DecisionTransformerPreTrainedModel,DeiTFeatureExtractor:()=>_.DeiTFeatureExtractor,DeiTForImageClassification:()=>h.DeiTForImageClassification,DeiTModel:()=>h.DeiTModel,DeiTPreTrainedModel:()=>h.DeiTPreTrainedModel,DepthAnythingForDepthEstimation:()=>h.DepthAnythingForDepthEstimation,DepthAnythingPreTrainedModel:()=>h.DepthAnythingPreTrainedModel,DepthEstimationPipeline:()=>p.DepthEstimationPipeline,DepthProForDepthEstimation:()=>h.DepthProForDepthEstimation,DepthProPreTrainedModel:()=>h.DepthProPreTrainedModel,DetrFeatureExtractor:()=>_.DetrFeatureExtractor,DetrForObjectDetection:()=>h.DetrForObjectDetection,DetrForSegmentation:()=>h.DetrForSegmentation,DetrModel:()=>h.DetrModel,DetrObjectDetectionOutput:()=>h.DetrObjectDetectionOutput,DetrPreTrainedModel:()=>h.DetrPreTrainedModel,DetrSegmentationOutput:()=>h.DetrSegmentationOutput,Dinov2ForImageClassification:()=>h.Dinov2ForImageClassification,Dinov2Model:()=>h.Dinov2Model,Dinov2PreTrainedModel:()=>h.Dinov2PreTrainedModel,DistilBertForMaskedLM:()=>h.DistilBertForMaskedLM,DistilBertForQuestionAnswering:()=>h.DistilBertForQuestionAnswering,DistilBertForSequenceClassification:()=>h.DistilBertForSequenceClassification,DistilBertForTokenClassification:()=>h.DistilBertForTokenClassification,DistilBertModel:()=>h.DistilBertModel,DistilBertPreTrainedModel:()=>h.DistilBertPreTrainedModel,DistilBertTokenizer:()=>m.DistilBertTokenizer,DocumentQuestionAnsweringPipeline:()=>p.DocumentQuestionAnsweringPipeline,DonutFeatureExtractor:()=>_.DonutFeatureExtractor,DonutSwinModel:()=>h.DonutSwinModel,DonutSwinPreTrainedModel:()=>h.DonutSwinPreTrainedModel,EfficientNetForImageClassification:()=>h.EfficientNetForImageClassification,EfficientNetImageProcessor:()=>_.EfficientNetImageProcessor,EfficientNetModel:()=>h.EfficientNetModel,EfficientNetPreTrainedModel:()=>h.EfficientNetPreTrainedModel,ElectraForMaskedLM:()=>h.ElectraForMaskedLM,ElectraForQuestionAnswering:()=>h.ElectraForQuestionAnswering,ElectraForSequenceClassification:()=>h.ElectraForSequenceClassification,ElectraForTokenClassification:()=>h.ElectraForTokenClassification,ElectraModel:()=>h.ElectraModel,ElectraPreTrainedModel:()=>h.ElectraPreTrainedModel,ElectraTokenizer:()=>m.ElectraTokenizer,EosTokenCriteria:()=>x.EosTokenCriteria,EsmForMaskedLM:()=>h.EsmForMaskedLM,EsmForSequenceClassification:()=>h.EsmForSequenceClassification,EsmForTokenClassification:()=>h.EsmForTokenClassification,EsmModel:()=>h.EsmModel,EsmPreTrainedModel:()=>h.EsmPreTrainedModel,EsmTokenizer:()=>m.EsmTokenizer,FFT:()=>T.FFT,FalconForCausalLM:()=>h.FalconForCausalLM,FalconModel:()=>h.FalconModel,FalconPreTrainedModel:()=>h.FalconPreTrainedModel,FalconTokenizer:()=>m.FalconTokenizer,FastViTForImageClassification:()=>h.FastViTForImageClassification,FastViTModel:()=>h.FastViTModel,FastViTPreTrainedModel:()=>h.FastViTPreTrainedModel,FeatureExtractionPipeline:()=>p.FeatureExtractionPipeline,FeatureExtractor:()=>_.FeatureExtractor,FillMaskPipeline:()=>p.FillMaskPipeline,Florence2ForConditionalGeneration:()=>h.Florence2ForConditionalGeneration,Florence2PreTrainedModel:()=>h.Florence2PreTrainedModel,Florence2Processor:()=>_.Florence2Processor,GLPNFeatureExtractor:()=>_.GLPNFeatureExtractor,GLPNForDepthEstimation:()=>h.GLPNForDepthEstimation,GLPNModel:()=>h.GLPNModel,GLPNPreTrainedModel:()=>h.GLPNPreTrainedModel,GPT2LMHeadModel:()=>h.GPT2LMHeadModel,GPT2Model:()=>h.GPT2Model,GPT2PreTrainedModel:()=>h.GPT2PreTrainedModel,GPT2Tokenizer:()=>m.GPT2Tokenizer,GPTBigCodeForCausalLM:()=>h.GPTBigCodeForCausalLM,GPTBigCodeModel:()=>h.GPTBigCodeModel,GPTBigCodePreTrainedModel:()=>h.GPTBigCodePreTrainedModel,GPTJForCausalLM:()=>h.GPTJForCausalLM,GPTJModel:()=>h.GPTJModel,GPTJPreTrainedModel:()=>h.GPTJPreTrainedModel,GPTNeoForCausalLM:()=>h.GPTNeoForCausalLM,GPTNeoModel:()=>h.GPTNeoModel,GPTNeoPreTrainedModel:()=>h.GPTNeoPreTrainedModel,GPTNeoXForCausalLM:()=>h.GPTNeoXForCausalLM,GPTNeoXModel:()=>h.GPTNeoXModel,GPTNeoXPreTrainedModel:()=>h.GPTNeoXPreTrainedModel,GPTNeoXTokenizer:()=>m.GPTNeoXTokenizer,Gemma2ForCausalLM:()=>h.Gemma2ForCausalLM,Gemma2Model:()=>h.Gemma2Model,Gemma2PreTrainedModel:()=>h.Gemma2PreTrainedModel,GemmaForCausalLM:()=>h.GemmaForCausalLM,GemmaModel:()=>h.GemmaModel,GemmaPreTrainedModel:()=>h.GemmaPreTrainedModel,GemmaTokenizer:()=>m.GemmaTokenizer,Grok1Tokenizer:()=>m.Grok1Tokenizer,GroupViTModel:()=>h.GroupViTModel,GroupViTPreTrainedModel:()=>h.GroupViTPreTrainedModel,HerbertTokenizer:()=>m.HerbertTokenizer,HieraForImageClassification:()=>h.HieraForImageClassification,HieraModel:()=>h.HieraModel,HieraPreTrainedModel:()=>h.HieraPreTrainedModel,HubertForCTC:()=>h.HubertForCTC,HubertForSequenceClassification:()=>h.HubertForSequenceClassification,HubertModel:()=>h.HubertModel,HubertPreTrainedModel:()=>h.HubertPreTrainedModel,ImageClassificationPipeline:()=>p.ImageClassificationPipeline,ImageFeatureExtractionPipeline:()=>p.ImageFeatureExtractionPipeline,ImageFeatureExtractor:()=>_.ImageFeatureExtractor,ImageMattingOutput:()=>h.ImageMattingOutput,ImageSegmentationPipeline:()=>p.ImageSegmentationPipeline,ImageToImagePipeline:()=>p.ImageToImagePipeline,ImageToTextPipeline:()=>p.ImageToTextPipeline,InterruptableStoppingCriteria:()=>x.InterruptableStoppingCriteria,JAISLMHeadModel:()=>h.JAISLMHeadModel,JAISModel:()=>h.JAISModel,JAISPreTrainedModel:()=>h.JAISPreTrainedModel,LlamaForCausalLM:()=>h.LlamaForCausalLM,LlamaModel:()=>h.LlamaModel,LlamaPreTrainedModel:()=>h.LlamaPreTrainedModel,LlamaTokenizer:()=>m.LlamaTokenizer,LlavaForConditionalGeneration:()=>h.LlavaForConditionalGeneration,LlavaPreTrainedModel:()=>h.LlavaPreTrainedModel,LongT5ForConditionalGeneration:()=>h.LongT5ForConditionalGeneration,LongT5Model:()=>h.LongT5Model,LongT5PreTrainedModel:()=>h.LongT5PreTrainedModel,M2M100ForConditionalGeneration:()=>h.M2M100ForConditionalGeneration,M2M100Model:()=>h.M2M100Model,M2M100PreTrainedModel:()=>h.M2M100PreTrainedModel,M2M100Tokenizer:()=>m.M2M100Tokenizer,MBart50Tokenizer:()=>m.MBart50Tokenizer,MBartForCausalLM:()=>h.MBartForCausalLM,MBartForConditionalGeneration:()=>h.MBartForConditionalGeneration,MBartForSequenceClassification:()=>h.MBartForSequenceClassification,MBartModel:()=>h.MBartModel,MBartPreTrainedModel:()=>h.MBartPreTrainedModel,MBartTokenizer:()=>m.MBartTokenizer,MPNetForMaskedLM:()=>h.MPNetForMaskedLM,MPNetForQuestionAnswering:()=>h.MPNetForQuestionAnswering,MPNetForSequenceClassification:()=>h.MPNetForSequenceClassification,MPNetForTokenClassification:()=>h.MPNetForTokenClassification,MPNetModel:()=>h.MPNetModel,MPNetPreTrainedModel:()=>h.MPNetPreTrainedModel,MPNetTokenizer:()=>m.MPNetTokenizer,MT5ForConditionalGeneration:()=>h.MT5ForConditionalGeneration,MT5Model:()=>h.MT5Model,MT5PreTrainedModel:()=>h.MT5PreTrainedModel,MarianMTModel:()=>h.MarianMTModel,MarianModel:()=>h.MarianModel,MarianPreTrainedModel:()=>h.MarianPreTrainedModel,MarianTokenizer:()=>m.MarianTokenizer,MaskFormerFeatureExtractor:()=>_.MaskFormerFeatureExtractor,MaskFormerForInstanceSegmentation:()=>h.MaskFormerForInstanceSegmentation,MaskFormerModel:()=>h.MaskFormerModel,MaskFormerPreTrainedModel:()=>h.MaskFormerPreTrainedModel,MaskedLMOutput:()=>h.MaskedLMOutput,MaxLengthCriteria:()=>x.MaxLengthCriteria,MistralForCausalLM:()=>h.MistralForCausalLM,MistralModel:()=>h.MistralModel,MistralPreTrainedModel:()=>h.MistralPreTrainedModel,MobileBertForMaskedLM:()=>h.MobileBertForMaskedLM,MobileBertForQuestionAnswering:()=>h.MobileBertForQuestionAnswering,MobileBertForSequenceClassification:()=>h.MobileBertForSequenceClassification,MobileBertModel:()=>h.MobileBertModel,MobileBertPreTrainedModel:()=>h.MobileBertPreTrainedModel,MobileBertTokenizer:()=>m.MobileBertTokenizer,MobileNetV1FeatureExtractor:()=>_.MobileNetV1FeatureExtractor,MobileNetV1ForImageClassification:()=>h.MobileNetV1ForImageClassification,MobileNetV1Model:()=>h.MobileNetV1Model,MobileNetV1PreTrainedModel:()=>h.MobileNetV1PreTrainedModel,MobileNetV2FeatureExtractor:()=>_.MobileNetV2FeatureExtractor,MobileNetV2ForImageClassification:()=>h.MobileNetV2ForImageClassification,MobileNetV2Model:()=>h.MobileNetV2Model,MobileNetV2PreTrainedModel:()=>h.MobileNetV2PreTrainedModel,MobileNetV3FeatureExtractor:()=>_.MobileNetV3FeatureExtractor,MobileNetV3ForImageClassification:()=>h.MobileNetV3ForImageClassification,MobileNetV3Model:()=>h.MobileNetV3Model,MobileNetV3PreTrainedModel:()=>h.MobileNetV3PreTrainedModel,MobileNetV4FeatureExtractor:()=>_.MobileNetV4FeatureExtractor,MobileNetV4ForImageClassification:()=>h.MobileNetV4ForImageClassification,MobileNetV4Model:()=>h.MobileNetV4Model,MobileNetV4PreTrainedModel:()=>h.MobileNetV4PreTrainedModel,MobileViTFeatureExtractor:()=>_.MobileViTFeatureExtractor,MobileViTForImageClassification:()=>h.MobileViTForImageClassification,MobileViTImageProcessor:()=>_.MobileViTImageProcessor,MobileViTModel:()=>h.MobileViTModel,MobileViTPreTrainedModel:()=>h.MobileViTPreTrainedModel,MobileViTV2ForImageClassification:()=>h.MobileViTV2ForImageClassification,MobileViTV2Model:()=>h.MobileViTV2Model,MobileViTV2PreTrainedModel:()=>h.MobileViTV2PreTrainedModel,ModelOutput:()=>h.ModelOutput,Moondream1ForConditionalGeneration:()=>h.Moondream1ForConditionalGeneration,MptForCausalLM:()=>h.MptForCausalLM,MptModel:()=>h.MptModel,MptPreTrainedModel:()=>h.MptPreTrainedModel,MusicgenForCausalLM:()=>h.MusicgenForCausalLM,MusicgenForConditionalGeneration:()=>h.MusicgenForConditionalGeneration,MusicgenModel:()=>h.MusicgenModel,MusicgenPreTrainedModel:()=>h.MusicgenPreTrainedModel,NllbTokenizer:()=>m.NllbTokenizer,NomicBertModel:()=>h.NomicBertModel,NomicBertPreTrainedModel:()=>h.NomicBertPreTrainedModel,NougatImageProcessor:()=>_.NougatImageProcessor,NougatTokenizer:()=>m.NougatTokenizer,OPTForCausalLM:()=>h.OPTForCausalLM,OPTModel:()=>h.OPTModel,OPTPreTrainedModel:()=>h.OPTPreTrainedModel,ObjectDetectionPipeline:()=>p.ObjectDetectionPipeline,OpenELMForCausalLM:()=>h.OpenELMForCausalLM,OpenELMModel:()=>h.OpenELMModel,OpenELMPreTrainedModel:()=>h.OpenELMPreTrainedModel,OwlViTFeatureExtractor:()=>_.OwlViTFeatureExtractor,OwlViTForObjectDetection:()=>h.OwlViTForObjectDetection,OwlViTModel:()=>h.OwlViTModel,OwlViTPreTrainedModel:()=>h.OwlViTPreTrainedModel,OwlViTProcessor:()=>_.OwlViTProcessor,Owlv2ForObjectDetection:()=>h.Owlv2ForObjectDetection,Owlv2ImageProcessor:()=>_.Owlv2ImageProcessor,Owlv2Model:()=>h.Owlv2Model,Owlv2PreTrainedModel:()=>h.Owlv2PreTrainedModel,Phi3ForCausalLM:()=>h.Phi3ForCausalLM,Phi3Model:()=>h.Phi3Model,Phi3PreTrainedModel:()=>h.Phi3PreTrainedModel,PhiForCausalLM:()=>h.PhiForCausalLM,PhiModel:()=>h.PhiModel,PhiPreTrainedModel:()=>h.PhiPreTrainedModel,Pipeline:()=>p.Pipeline,PreTrainedModel:()=>h.PreTrainedModel,PreTrainedTokenizer:()=>m.PreTrainedTokenizer,PretrainedConfig:()=>f.PretrainedConfig,PretrainedMixin:()=>h.PretrainedMixin,Processor:()=>_.Processor,PvtForImageClassification:()=>h.PvtForImageClassification,PvtImageProcessor:()=>_.PvtImageProcessor,PvtModel:()=>h.PvtModel,PvtPreTrainedModel:()=>h.PvtPreTrainedModel,PyAnnoteFeatureExtractor:()=>_.PyAnnoteFeatureExtractor,PyAnnoteForAudioFrameClassification:()=>h.PyAnnoteForAudioFrameClassification,PyAnnoteModel:()=>h.PyAnnoteModel,PyAnnotePreTrainedModel:()=>h.PyAnnotePreTrainedModel,PyAnnoteProcessor:()=>_.PyAnnoteProcessor,QuestionAnsweringModelOutput:()=>h.QuestionAnsweringModelOutput,QuestionAnsweringPipeline:()=>p.QuestionAnsweringPipeline,Qwen2ForCausalLM:()=>h.Qwen2ForCausalLM,Qwen2Model:()=>h.Qwen2Model,Qwen2PreTrainedModel:()=>h.Qwen2PreTrainedModel,Qwen2Tokenizer:()=>m.Qwen2Tokenizer,RTDetrForObjectDetection:()=>h.RTDetrForObjectDetection,RTDetrImageProcessor:()=>_.RTDetrImageProcessor,RTDetrModel:()=>h.RTDetrModel,RTDetrObjectDetectionOutput:()=>h.RTDetrObjectDetectionOutput,RTDetrPreTrainedModel:()=>h.RTDetrPreTrainedModel,RawImage:()=>M.RawImage,ResNetForImageClassification:()=>h.ResNetForImageClassification,ResNetModel:()=>h.ResNetModel,ResNetPreTrainedModel:()=>h.ResNetPreTrainedModel,RoFormerForMaskedLM:()=>h.RoFormerForMaskedLM,RoFormerForQuestionAnswering:()=>h.RoFormerForQuestionAnswering,RoFormerForSequenceClassification:()=>h.RoFormerForSequenceClassification,RoFormerForTokenClassification:()=>h.RoFormerForTokenClassification,RoFormerModel:()=>h.RoFormerModel,RoFormerPreTrainedModel:()=>h.RoFormerPreTrainedModel,RoFormerTokenizer:()=>m.RoFormerTokenizer,RobertaForMaskedLM:()=>h.RobertaForMaskedLM,RobertaForQuestionAnswering:()=>h.RobertaForQuestionAnswering,RobertaForSequenceClassification:()=>h.RobertaForSequenceClassification,RobertaForTokenClassification:()=>h.RobertaForTokenClassification,RobertaModel:()=>h.RobertaModel,RobertaPreTrainedModel:()=>h.RobertaPreTrainedModel,RobertaTokenizer:()=>m.RobertaTokenizer,SamImageProcessor:()=>_.SamImageProcessor,SamImageSegmentationOutput:()=>h.SamImageSegmentationOutput,SamModel:()=>h.SamModel,SamPreTrainedModel:()=>h.SamPreTrainedModel,SamProcessor:()=>_.SamProcessor,SapiensFeatureExtractor:()=>_.SapiensFeatureExtractor,SapiensForDepthEstimation:()=>h.SapiensForDepthEstimation,SapiensForNormalEstimation:()=>h.SapiensForNormalEstimation,SapiensForSemanticSegmentation:()=>h.SapiensForSemanticSegmentation,SapiensPreTrainedModel:()=>h.SapiensPreTrainedModel,SeamlessM4TFeatureExtractor:()=>_.SeamlessM4TFeatureExtractor,SegformerFeatureExtractor:()=>_.SegformerFeatureExtractor,SegformerForImageClassification:()=>h.SegformerForImageClassification,SegformerForSemanticSegmentation:()=>h.SegformerForSemanticSegmentation,SegformerModel:()=>h.SegformerModel,SegformerPreTrainedModel:()=>h.SegformerPreTrainedModel,Seq2SeqLMOutput:()=>h.Seq2SeqLMOutput,SequenceClassifierOutput:()=>h.SequenceClassifierOutput,SiglipImageProcessor:()=>_.SiglipImageProcessor,SiglipModel:()=>h.SiglipModel,SiglipPreTrainedModel:()=>h.SiglipPreTrainedModel,SiglipTextModel:()=>h.SiglipTextModel,SiglipTokenizer:()=>m.SiglipTokenizer,SiglipVisionModel:()=>h.SiglipVisionModel,SpeechT5FeatureExtractor:()=>_.SpeechT5FeatureExtractor,SpeechT5ForSpeechToText:()=>h.SpeechT5ForSpeechToText,SpeechT5ForTextToSpeech:()=>h.SpeechT5ForTextToSpeech,SpeechT5HifiGan:()=>h.SpeechT5HifiGan,SpeechT5Model:()=>h.SpeechT5Model,SpeechT5PreTrainedModel:()=>h.SpeechT5PreTrainedModel,SpeechT5Processor:()=>_.SpeechT5Processor,SpeechT5Tokenizer:()=>m.SpeechT5Tokenizer,SqueezeBertForMaskedLM:()=>h.SqueezeBertForMaskedLM,SqueezeBertForQuestionAnswering:()=>h.SqueezeBertForQuestionAnswering,SqueezeBertForSequenceClassification:()=>h.SqueezeBertForSequenceClassification,SqueezeBertModel:()=>h.SqueezeBertModel,SqueezeBertPreTrainedModel:()=>h.SqueezeBertPreTrainedModel,SqueezeBertTokenizer:()=>m.SqueezeBertTokenizer,StableLmForCausalLM:()=>h.StableLmForCausalLM,StableLmModel:()=>h.StableLmModel,StableLmPreTrainedModel:()=>h.StableLmPreTrainedModel,Starcoder2ForCausalLM:()=>h.Starcoder2ForCausalLM,Starcoder2Model:()=>h.Starcoder2Model,Starcoder2PreTrainedModel:()=>h.Starcoder2PreTrainedModel,StoppingCriteria:()=>x.StoppingCriteria,StoppingCriteriaList:()=>x.StoppingCriteriaList,SummarizationPipeline:()=>p.SummarizationPipeline,Swin2SRForImageSuperResolution:()=>h.Swin2SRForImageSuperResolution,Swin2SRImageProcessor:()=>_.Swin2SRImageProcessor,Swin2SRModel:()=>h.Swin2SRModel,Swin2SRPreTrainedModel:()=>h.Swin2SRPreTrainedModel,SwinForImageClassification:()=>h.SwinForImageClassification,SwinModel:()=>h.SwinModel,SwinPreTrainedModel:()=>h.SwinPreTrainedModel,T5ForConditionalGeneration:()=>h.T5ForConditionalGeneration,T5Model:()=>h.T5Model,T5PreTrainedModel:()=>h.T5PreTrainedModel,T5Tokenizer:()=>m.T5Tokenizer,TableTransformerForObjectDetection:()=>h.TableTransformerForObjectDetection,TableTransformerModel:()=>h.TableTransformerModel,TableTransformerObjectDetectionOutput:()=>h.TableTransformerObjectDetectionOutput,TableTransformerPreTrainedModel:()=>h.TableTransformerPreTrainedModel,Tensor:()=>w.Tensor,Text2TextGenerationPipeline:()=>p.Text2TextGenerationPipeline,TextClassificationPipeline:()=>p.TextClassificationPipeline,TextGenerationPipeline:()=>p.TextGenerationPipeline,TextStreamer:()=>b.TextStreamer,TextToAudioPipeline:()=>p.TextToAudioPipeline,TokenClassificationPipeline:()=>p.TokenClassificationPipeline,TokenClassifierOutput:()=>h.TokenClassifierOutput,TokenizerModel:()=>m.TokenizerModel,TrOCRForCausalLM:()=>h.TrOCRForCausalLM,TrOCRPreTrainedModel:()=>h.TrOCRPreTrainedModel,TranslationPipeline:()=>p.TranslationPipeline,UniSpeechForCTC:()=>h.UniSpeechForCTC,UniSpeechForSequenceClassification:()=>h.UniSpeechForSequenceClassification,UniSpeechModel:()=>h.UniSpeechModel,UniSpeechPreTrainedModel:()=>h.UniSpeechPreTrainedModel,UniSpeechSatForAudioFrameClassification:()=>h.UniSpeechSatForAudioFrameClassification,UniSpeechSatForCTC:()=>h.UniSpeechSatForCTC,UniSpeechSatForSequenceClassification:()=>h.UniSpeechSatForSequenceClassification,UniSpeechSatModel:()=>h.UniSpeechSatModel,UniSpeechSatPreTrainedModel:()=>h.UniSpeechSatPreTrainedModel,ViTFeatureExtractor:()=>_.ViTFeatureExtractor,ViTForImageClassification:()=>h.ViTForImageClassification,ViTImageProcessor:()=>_.ViTImageProcessor,ViTMAEModel:()=>h.ViTMAEModel,ViTMAEPreTrainedModel:()=>h.ViTMAEPreTrainedModel,ViTMSNForImageClassification:()=>h.ViTMSNForImageClassification,ViTMSNModel:()=>h.ViTMSNModel,ViTMSNPreTrainedModel:()=>h.ViTMSNPreTrainedModel,ViTModel:()=>h.ViTModel,ViTPreTrainedModel:()=>h.ViTPreTrainedModel,VisionEncoderDecoderModel:()=>h.VisionEncoderDecoderModel,VitMatteForImageMatting:()=>h.VitMatteForImageMatting,VitMatteImageProcessor:()=>_.VitMatteImageProcessor,VitMattePreTrainedModel:()=>h.VitMattePreTrainedModel,VitsModel:()=>h.VitsModel,VitsModelOutput:()=>h.VitsModelOutput,VitsPreTrainedModel:()=>h.VitsPreTrainedModel,VitsTokenizer:()=>m.VitsTokenizer,Wav2Vec2BertForCTC:()=>h.Wav2Vec2BertForCTC,Wav2Vec2BertForSequenceClassification:()=>h.Wav2Vec2BertForSequenceClassification,Wav2Vec2BertModel:()=>h.Wav2Vec2BertModel,Wav2Vec2BertPreTrainedModel:()=>h.Wav2Vec2BertPreTrainedModel,Wav2Vec2CTCTokenizer:()=>m.Wav2Vec2CTCTokenizer,Wav2Vec2FeatureExtractor:()=>_.Wav2Vec2FeatureExtractor,Wav2Vec2ForAudioFrameClassification:()=>h.Wav2Vec2ForAudioFrameClassification,Wav2Vec2ForCTC:()=>h.Wav2Vec2ForCTC,Wav2Vec2ForSequenceClassification:()=>h.Wav2Vec2ForSequenceClassification,Wav2Vec2Model:()=>h.Wav2Vec2Model,Wav2Vec2PreTrainedModel:()=>h.Wav2Vec2PreTrainedModel,Wav2Vec2ProcessorWithLM:()=>_.Wav2Vec2ProcessorWithLM,WavLMForAudioFrameClassification:()=>h.WavLMForAudioFrameClassification,WavLMForCTC:()=>h.WavLMForCTC,WavLMForSequenceClassification:()=>h.WavLMForSequenceClassification,WavLMForXVector:()=>h.WavLMForXVector,WavLMModel:()=>h.WavLMModel,WavLMPreTrainedModel:()=>h.WavLMPreTrainedModel,WeSpeakerFeatureExtractor:()=>_.WeSpeakerFeatureExtractor,WeSpeakerResNetModel:()=>h.WeSpeakerResNetModel,WeSpeakerResNetPreTrainedModel:()=>h.WeSpeakerResNetPreTrainedModel,WhisperFeatureExtractor:()=>_.WhisperFeatureExtractor,WhisperForConditionalGeneration:()=>h.WhisperForConditionalGeneration,WhisperModel:()=>h.WhisperModel,WhisperPreTrainedModel:()=>h.WhisperPreTrainedModel,WhisperProcessor:()=>_.WhisperProcessor,WhisperTextStreamer:()=>b.WhisperTextStreamer,WhisperTokenizer:()=>m.WhisperTokenizer,XLMForQuestionAnswering:()=>h.XLMForQuestionAnswering,XLMForSequenceClassification:()=>h.XLMForSequenceClassification,XLMForTokenClassification:()=>h.XLMForTokenClassification,XLMModel:()=>h.XLMModel,XLMPreTrainedModel:()=>h.XLMPreTrainedModel,XLMRobertaForMaskedLM:()=>h.XLMRobertaForMaskedLM,XLMRobertaForQuestionAnswering:()=>h.XLMRobertaForQuestionAnswering,XLMRobertaForSequenceClassification:()=>h.XLMRobertaForSequenceClassification,XLMRobertaForTokenClassification:()=>h.XLMRobertaForTokenClassification,XLMRobertaModel:()=>h.XLMRobertaModel,XLMRobertaPreTrainedModel:()=>h.XLMRobertaPreTrainedModel,XLMRobertaTokenizer:()=>m.XLMRobertaTokenizer,XLMTokenizer:()=>m.XLMTokenizer,XLMWithLMHeadModel:()=>h.XLMWithLMHeadModel,XVectorOutput:()=>h.XVectorOutput,YolosFeatureExtractor:()=>_.YolosFeatureExtractor,YolosForObjectDetection:()=>h.YolosForObjectDetection,YolosModel:()=>h.YolosModel,YolosObjectDetectionOutput:()=>h.YolosObjectDetectionOutput,YolosPreTrainedModel:()=>h.YolosPreTrainedModel,ZeroShotAudioClassificationPipeline:()=>p.ZeroShotAudioClassificationPipeline,ZeroShotClassificationPipeline:()=>p.ZeroShotClassificationPipeline,ZeroShotImageClassificationPipeline:()=>p.ZeroShotImageClassificationPipeline,ZeroShotObjectDetectionPipeline:()=>p.ZeroShotObjectDetectionPipeline,bankers_round:()=>T.bankers_round,cat:()=>w.cat,cos_sim:()=>T.cos_sim,dot:()=>T.dot,dynamic_time_warping:()=>T.dynamic_time_warping,env:()=>u.env,full:()=>w.full,full_like:()=>w.full_like,getKeyValueShapes:()=>f.getKeyValueShapes,hamming:()=>g.hamming,hanning:()=>g.hanning,interpolate:()=>w.interpolate,interpolate_4d:()=>w.interpolate_4d,interpolate_data:()=>T.interpolate_data,is_chinese_char:()=>m.is_chinese_char,layer_norm:()=>w.layer_norm,log_softmax:()=>T.log_softmax,magnitude:()=>T.magnitude,matmul:()=>w.matmul,max:()=>T.max,mean:()=>w.mean,mean_pooling:()=>w.mean_pooling,medianFilter:()=>T.medianFilter,mel_filter_bank:()=>g.mel_filter_bank,min:()=>T.min,ones:()=>w.ones,ones_like:()=>w.ones_like,permute:()=>w.permute,permute_data:()=>T.permute_data,pipeline:()=>p.pipeline,quantize_embeddings:()=>w.quantize_embeddings,read_audio:()=>g.read_audio,rfft:()=>w.rfft,round:()=>T.round,softmax:()=>T.softmax,spectrogram:()=>g.spectrogram,stack:()=>w.stack,std_mean:()=>w.std_mean,topk:()=>w.topk,window_function:()=>g.window_function,zeros:()=>w.zeros,zeros_like:()=>w.zeros_like});var u=c(/*! ./env.js */"./src/env.js"),p=c(/*! ./pipelines.js */"./src/pipelines.js"),h=c(/*! ./models.js */"./src/models.js"),m=c(/*! ./tokenizers.js */"./src/tokenizers.js"),_=c(/*! ./processors.js */"./src/processors.js"),f=c(/*! ./configs.js */"./src/configs.js"),g=c(/*! ./utils/audio.js */"./src/utils/audio.js"),M=c(/*! ./utils/image.js */"./src/utils/image.js"),w=c(/*! ./utils/tensor.js */"./src/utils/tensor.js"),T=c(/*! ./utils/maths.js */"./src/utils/maths.js"),b=c(/*! ./generation/streamers.js */"./src/generation/streamers.js"),x=c(/*! ./generation/stopping_criteria.js */"./src/generation/stopping_criteria.js"),y=d.ASTFeatureExtractor,k=d.ASTForAudioClassification,F=d.ASTModel,P=d.ASTPreTrainedModel,C=d.AlbertForMaskedLM,v=d.AlbertForQuestionAnswering,S=d.AlbertForSequenceClassification,A=d.AlbertModel,E=d.AlbertPreTrainedModel,L=d.AlbertTokenizer,z=d.AudioClassificationPipeline,I=d.AutoConfig,B=d.AutoModel,N=d.AutoModelForAudioClassification,D=d.AutoModelForAudioFrameClassification,O=d.AutoModelForCTC,V=d.AutoModelForCausalLM,j=d.AutoModelForDepthEstimation,R=d.AutoModelForDocumentQuestionAnswering,G=d.AutoModelForImageClassification,q=d.AutoModelForImageFeatureExtraction,W=d.AutoModelForImageMatting,$=d.AutoModelForImageSegmentation,U=d.AutoModelForImageToImage,X=d.AutoModelForMaskGeneration,Q=d.AutoModelForMaskedLM,H=d.AutoModelForNormalEstimation,Y=d.AutoModelForObjectDetection,J=d.AutoModelForQuestionAnswering,K=d.AutoModelForSemanticSegmentation,Z=d.AutoModelForSeq2SeqLM,ee=d.AutoModelForSequenceClassification,te=d.AutoModelForSpeechSeq2Seq,re=d.AutoModelForTextToSpectrogram,ne=d.AutoModelForTextToWaveform,oe=d.AutoModelForTokenClassification,se=d.AutoModelForUniversalSegmentation,ae=d.AutoModelForVision2Seq,ie=d.AutoModelForXVector,le=d.AutoModelForZeroShotObjectDetection,ce=d.AutoProcessor,de=d.AutoTokenizer,ue=d.AutomaticSpeechRecognitionPipeline,pe=d.BartForConditionalGeneration,he=d.BartForSequenceClassification,me=d.BartModel,_e=d.BartPretrainedModel,fe=d.BartTokenizer,ge=d.BaseModelOutput,Me=d.BaseStreamer,we=d.BeitFeatureExtractor,Te=d.BeitForImageClassification,be=d.BeitModel,xe=d.BeitPreTrainedModel,ye=d.BertForMaskedLM,ke=d.BertForQuestionAnswering,Fe=d.BertForSequenceClassification,Pe=d.BertForTokenClassification,Ce=d.BertModel,ve=d.BertPreTrainedModel,Se=d.BertTokenizer,Ae=d.BitImageProcessor,Ee=d.BlenderbotForConditionalGeneration,Le=d.BlenderbotModel,ze=d.BlenderbotPreTrainedModel,Ie=d.BlenderbotSmallForConditionalGeneration,Be=d.BlenderbotSmallModel,Ne=d.BlenderbotSmallPreTrainedModel,De=d.BlenderbotSmallTokenizer,Oe=d.BlenderbotTokenizer,Ve=d.BloomForCausalLM,je=d.BloomModel,Re=d.BloomPreTrainedModel,Ge=d.BloomTokenizer,qe=d.CLIPFeatureExtractor,We=d.CLIPImageProcessor,$e=d.CLIPModel,Ue=d.CLIPPreTrainedModel,Xe=d.CLIPSegForImageSegmentation,Qe=d.CLIPSegModel,He=d.CLIPSegPreTrainedModel,Ye=d.CLIPTextModel,Je=d.CLIPTextModelWithProjection,Ke=d.CLIPTokenizer,Ze=d.CLIPVisionModel,et=d.CLIPVisionModelWithProjection,tt=d.CamembertForMaskedLM,rt=d.CamembertForQuestionAnswering,nt=d.CamembertForSequenceClassification,ot=d.CamembertForTokenClassification,st=d.CamembertModel,at=d.CamembertPreTrainedModel,it=d.CamembertTokenizer,lt=d.CausalLMOutput,ct=d.CausalLMOutputWithPast,dt=d.ChineseCLIPFeatureExtractor,ut=d.ChineseCLIPModel,pt=d.ChineseCLIPPreTrainedModel,ht=d.ClapAudioModelWithProjection,mt=d.ClapFeatureExtractor,_t=d.ClapModel,ft=d.ClapPreTrainedModel,gt=d.ClapTextModelWithProjection,Mt=d.CodeGenForCausalLM,wt=d.CodeGenModel,Tt=d.CodeGenPreTrainedModel,bt=d.CodeGenTokenizer,xt=d.CodeLlamaTokenizer,yt=d.CohereForCausalLM,kt=d.CohereModel,Ft=d.CoherePreTrainedModel,Pt=d.CohereTokenizer,Ct=d.ConvBertForMaskedLM,vt=d.ConvBertForQuestionAnswering,St=d.ConvBertForSequenceClassification,At=d.ConvBertForTokenClassification,Et=d.ConvBertModel,Lt=d.ConvBertPreTrainedModel,zt=d.ConvBertTokenizer,It=d.ConvNextFeatureExtractor,Bt=d.ConvNextForImageClassification,Nt=d.ConvNextImageProcessor,Dt=d.ConvNextModel,Ot=d.ConvNextPreTrainedModel,Vt=d.ConvNextV2ForImageClassification,jt=d.ConvNextV2Model,Rt=d.ConvNextV2PreTrainedModel,Gt=d.DPTFeatureExtractor,qt=d.DPTForDepthEstimation,Wt=d.DPTImageProcessor,$t=d.DPTModel,Ut=d.DPTPreTrainedModel,Xt=d.DebertaForMaskedLM,Qt=d.DebertaForQuestionAnswering,Ht=d.DebertaForSequenceClassification,Yt=d.DebertaForTokenClassification,Jt=d.DebertaModel,Kt=d.DebertaPreTrainedModel,Zt=d.DebertaTokenizer,er=d.DebertaV2ForMaskedLM,tr=d.DebertaV2ForQuestionAnswering,rr=d.DebertaV2ForSequenceClassification,nr=d.DebertaV2ForTokenClassification,or=d.DebertaV2Model,sr=d.DebertaV2PreTrainedModel,ar=d.DebertaV2Tokenizer,ir=d.DecisionTransformerModel,lr=d.DecisionTransformerPreTrainedModel,cr=d.DeiTFeatureExtractor,dr=d.DeiTForImageClassification,ur=d.DeiTModel,pr=d.DeiTPreTrainedModel,hr=d.DepthAnythingForDepthEstimation,mr=d.DepthAnythingPreTrainedModel,_r=d.DepthEstimationPipeline,fr=d.DepthProForDepthEstimation,gr=d.DepthProPreTrainedModel,Mr=d.DetrFeatureExtractor,wr=d.DetrForObjectDetection,Tr=d.DetrForSegmentation,br=d.DetrModel,xr=d.DetrObjectDetectionOutput,yr=d.DetrPreTrainedModel,kr=d.DetrSegmentationOutput,Fr=d.Dinov2ForImageClassification,Pr=d.Dinov2Model,Cr=d.Dinov2PreTrainedModel,vr=d.DistilBertForMaskedLM,Sr=d.DistilBertForQuestionAnswering,Ar=d.DistilBertForSequenceClassification,Er=d.DistilBertForTokenClassification,Lr=d.DistilBertModel,zr=d.DistilBertPreTrainedModel,Ir=d.DistilBertTokenizer,Br=d.DocumentQuestionAnsweringPipeline,Nr=d.DonutFeatureExtractor,Dr=d.DonutSwinModel,Or=d.DonutSwinPreTrainedModel,Vr=d.EfficientNetForImageClassification,jr=d.EfficientNetImageProcessor,Rr=d.EfficientNetModel,Gr=d.EfficientNetPreTrainedModel,qr=d.ElectraForMaskedLM,Wr=d.ElectraForQuestionAnswering,$r=d.ElectraForSequenceClassification,Ur=d.ElectraForTokenClassification,Xr=d.ElectraModel,Qr=d.ElectraPreTrainedModel,Hr=d.ElectraTokenizer,Yr=d.EosTokenCriteria,Jr=d.EsmForMaskedLM,Kr=d.EsmForSequenceClassification,Zr=d.EsmForTokenClassification,en=d.EsmModel,tn=d.EsmPreTrainedModel,rn=d.EsmTokenizer,nn=d.FFT,on=d.FalconForCausalLM,sn=d.FalconModel,an=d.FalconPreTrainedModel,ln=d.FalconTokenizer,cn=d.FastViTForImageClassification,dn=d.FastViTModel,un=d.FastViTPreTrainedModel,pn=d.FeatureExtractionPipeline,hn=d.FeatureExtractor,mn=d.FillMaskPipeline,_n=d.Florence2ForConditionalGeneration,fn=d.Florence2PreTrainedModel,gn=d.Florence2Processor,Mn=d.GLPNFeatureExtractor,wn=d.GLPNForDepthEstimation,Tn=d.GLPNModel,bn=d.GLPNPreTrainedModel,xn=d.GPT2LMHeadModel,yn=d.GPT2Model,kn=d.GPT2PreTrainedModel,Fn=d.GPT2Tokenizer,Pn=d.GPTBigCodeForCausalLM,Cn=d.GPTBigCodeModel,vn=d.GPTBigCodePreTrainedModel,Sn=d.GPTJForCausalLM,An=d.GPTJModel,En=d.GPTJPreTrainedModel,Ln=d.GPTNeoForCausalLM,zn=d.GPTNeoModel,In=d.GPTNeoPreTrainedModel,Bn=d.GPTNeoXForCausalLM,Nn=d.GPTNeoXModel,Dn=d.GPTNeoXPreTrainedModel,On=d.GPTNeoXTokenizer,Vn=d.Gemma2ForCausalLM,jn=d.Gemma2Model,Rn=d.Gemma2PreTrainedModel,Gn=d.GemmaForCausalLM,qn=d.GemmaModel,Wn=d.GemmaPreTrainedModel,$n=d.GemmaTokenizer,Un=d.Grok1Tokenizer,Xn=d.GroupViTModel,Qn=d.GroupViTPreTrainedModel,Hn=d.HerbertTokenizer,Yn=d.HieraForImageClassification,Jn=d.HieraModel,Kn=d.HieraPreTrainedModel,Zn=d.HubertForCTC,eo=d.HubertForSequenceClassification,to=d.HubertModel,ro=d.HubertPreTrainedModel,no=d.ImageClassificationPipeline,oo=d.ImageFeatureExtractionPipeline,so=d.ImageFeatureExtractor,ao=d.ImageMattingOutput,io=d.ImageSegmentationPipeline,lo=d.ImageToImagePipeline,co=d.ImageToTextPipeline,uo=d.InterruptableStoppingCriteria,po=d.JAISLMHeadModel,ho=d.JAISModel,mo=d.JAISPreTrainedModel,_o=d.LlamaForCausalLM,fo=d.LlamaModel,go=d.LlamaPreTrainedModel,Mo=d.LlamaTokenizer,wo=d.LlavaForConditionalGeneration,To=d.LlavaPreTrainedModel,bo=d.LongT5ForConditionalGeneration,xo=d.LongT5Model,yo=d.LongT5PreTrainedModel,ko=d.M2M100ForConditionalGeneration,Fo=d.M2M100Model,Po=d.M2M100PreTrainedModel,Co=d.M2M100Tokenizer,vo=d.MBart50Tokenizer,So=d.MBartForCausalLM,Ao=d.MBartForConditionalGeneration,Eo=d.MBartForSequenceClassification,Lo=d.MBartModel,zo=d.MBartPreTrainedModel,Io=d.MBartTokenizer,Bo=d.MPNetForMaskedLM,No=d.MPNetForQuestionAnswering,Do=d.MPNetForSequenceClassification,Oo=d.MPNetForTokenClassification,Vo=d.MPNetModel,jo=d.MPNetPreTrainedModel,Ro=d.MPNetTokenizer,Go=d.MT5ForConditionalGeneration,qo=d.MT5Model,Wo=d.MT5PreTrainedModel,$o=d.MarianMTModel,Uo=d.MarianModel,Xo=d.MarianPreTrainedModel,Qo=d.MarianTokenizer,Ho=d.MaskFormerFeatureExtractor,Yo=d.MaskFormerForInstanceSegmentation,Jo=d.MaskFormerModel,Ko=d.MaskFormerPreTrainedModel,Zo=d.MaskedLMOutput,es=d.MaxLengthCriteria,ts=d.MistralForCausalLM,rs=d.MistralModel,ns=d.MistralPreTrainedModel,os=d.MobileBertForMaskedLM,ss=d.MobileBertForQuestionAnswering,as=d.MobileBertForSequenceClassification,is=d.MobileBertModel,ls=d.MobileBertPreTrainedModel,cs=d.MobileBertTokenizer,ds=d.MobileNetV1FeatureExtractor,us=d.MobileNetV1ForImageClassification,ps=d.MobileNetV1Model,hs=d.MobileNetV1PreTrainedModel,ms=d.MobileNetV2FeatureExtractor,_s=d.MobileNetV2ForImageClassification,fs=d.MobileNetV2Model,gs=d.MobileNetV2PreTrainedModel,Ms=d.MobileNetV3FeatureExtractor,ws=d.MobileNetV3ForImageClassification,Ts=d.MobileNetV3Model,bs=d.MobileNetV3PreTrainedModel,xs=d.MobileNetV4FeatureExtractor,ys=d.MobileNetV4ForImageClassification,ks=d.MobileNetV4Model,Fs=d.MobileNetV4PreTrainedModel,Ps=d.MobileViTFeatureExtractor,Cs=d.MobileViTForImageClassification,vs=d.MobileViTImageProcessor,Ss=d.MobileViTModel,As=d.MobileViTPreTrainedModel,Es=d.MobileViTV2ForImageClassification,Ls=d.MobileViTV2Model,zs=d.MobileViTV2PreTrainedModel,Is=d.ModelOutput,Bs=d.Moondream1ForConditionalGeneration,Ns=d.MptForCausalLM,Ds=d.MptModel,Os=d.MptPreTrainedModel,Vs=d.MusicgenForCausalLM,js=d.MusicgenForConditionalGeneration,Rs=d.MusicgenModel,Gs=d.MusicgenPreTrainedModel,qs=d.NllbTokenizer,Ws=d.NomicBertModel,$s=d.NomicBertPreTrainedModel,Us=d.NougatImageProcessor,Xs=d.NougatTokenizer,Qs=d.OPTForCausalLM,Hs=d.OPTModel,Ys=d.OPTPreTrainedModel,Js=d.ObjectDetectionPipeline,Ks=d.OpenELMForCausalLM,Zs=d.OpenELMModel,ea=d.OpenELMPreTrainedModel,ta=d.OwlViTFeatureExtractor,ra=d.OwlViTForObjectDetection,na=d.OwlViTModel,oa=d.OwlViTPreTrainedModel,sa=d.OwlViTProcessor,aa=d.Owlv2ForObjectDetection,ia=d.Owlv2ImageProcessor,la=d.Owlv2Model,ca=d.Owlv2PreTrainedModel,da=d.Phi3ForCausalLM,ua=d.Phi3Model,pa=d.Phi3PreTrainedModel,ha=d.PhiForCausalLM,ma=d.PhiModel,_a=d.PhiPreTrainedModel,fa=d.Pipeline,ga=d.PreTrainedModel,Ma=d.PreTrainedTokenizer,wa=d.PretrainedConfig,Ta=d.PretrainedMixin,ba=d.Processor,xa=d.PvtForImageClassification,ya=d.PvtImageProcessor,ka=d.PvtModel,Fa=d.PvtPreTrainedModel,Pa=d.PyAnnoteFeatureExtractor,Ca=d.PyAnnoteForAudioFrameClassification,va=d.PyAnnoteModel,Sa=d.PyAnnotePreTrainedModel,Aa=d.PyAnnoteProcessor,Ea=d.QuestionAnsweringModelOutput,La=d.QuestionAnsweringPipeline,za=d.Qwen2ForCausalLM,Ia=d.Qwen2Model,Ba=d.Qwen2PreTrainedModel,Na=d.Qwen2Tokenizer,Da=d.RTDetrForObjectDetection,Oa=d.RTDetrImageProcessor,Va=d.RTDetrModel,ja=d.RTDetrObjectDetectionOutput,Ra=d.RTDetrPreTrainedModel,Ga=d.RawImage,qa=d.ResNetForImageClassification,Wa=d.ResNetModel,$a=d.ResNetPreTrainedModel,Ua=d.RoFormerForMaskedLM,Xa=d.RoFormerForQuestionAnswering,Qa=d.RoFormerForSequenceClassification,Ha=d.RoFormerForTokenClassification,Ya=d.RoFormerModel,Ja=d.RoFormerPreTrainedModel,Ka=d.RoFormerTokenizer,Za=d.RobertaForMaskedLM,ei=d.RobertaForQuestionAnswering,ti=d.RobertaForSequenceClassification,ri=d.RobertaForTokenClassification,ni=d.RobertaModel,oi=d.RobertaPreTrainedModel,si=d.RobertaTokenizer,ai=d.SamImageProcessor,ii=d.SamImageSegmentationOutput,li=d.SamModel,ci=d.SamPreTrainedModel,di=d.SamProcessor,ui=d.SapiensFeatureExtractor,pi=d.SapiensForDepthEstimation,hi=d.SapiensForNormalEstimation,mi=d.SapiensForSemanticSegmentation,_i=d.SapiensPreTrainedModel,fi=d.SeamlessM4TFeatureExtractor,gi=d.SegformerFeatureExtractor,Mi=d.SegformerForImageClassification,wi=d.SegformerForSemanticSegmentation,Ti=d.SegformerModel,bi=d.SegformerPreTrainedModel,xi=d.Seq2SeqLMOutput,yi=d.SequenceClassifierOutput,ki=d.SiglipImageProcessor,Fi=d.SiglipModel,Pi=d.SiglipPreTrainedModel,Ci=d.SiglipTextModel,vi=d.SiglipTokenizer,Si=d.SiglipVisionModel,Ai=d.SpeechT5FeatureExtractor,Ei=d.SpeechT5ForSpeechToText,Li=d.SpeechT5ForTextToSpeech,zi=d.SpeechT5HifiGan,Ii=d.SpeechT5Model,Bi=d.SpeechT5PreTrainedModel,Ni=d.SpeechT5Processor,Di=d.SpeechT5Tokenizer,Oi=d.SqueezeBertForMaskedLM,Vi=d.SqueezeBertForQuestionAnswering,ji=d.SqueezeBertForSequenceClassification,Ri=d.SqueezeBertModel,Gi=d.SqueezeBertPreTrainedModel,qi=d.SqueezeBertTokenizer,Wi=d.StableLmForCausalLM,$i=d.StableLmModel,Ui=d.StableLmPreTrainedModel,Xi=d.Starcoder2ForCausalLM,Qi=d.Starcoder2Model,Hi=d.Starcoder2PreTrainedModel,Yi=d.StoppingCriteria,Ji=d.StoppingCriteriaList,Ki=d.SummarizationPipeline,Zi=d.Swin2SRForImageSuperResolution,el=d.Swin2SRImageProcessor,tl=d.Swin2SRModel,rl=d.Swin2SRPreTrainedModel,nl=d.SwinForImageClassification,ol=d.SwinModel,sl=d.SwinPreTrainedModel,al=d.T5ForConditionalGeneration,il=d.T5Model,ll=d.T5PreTrainedModel,cl=d.T5Tokenizer,dl=d.TableTransformerForObjectDetection,ul=d.TableTransformerModel,pl=d.TableTransformerObjectDetectionOutput,hl=d.TableTransformerPreTrainedModel,ml=d.Tensor,_l=d.Text2TextGenerationPipeline,fl=d.TextClassificationPipeline,gl=d.TextGenerationPipeline,Ml=d.TextStreamer,wl=d.TextToAudioPipeline,Tl=d.TokenClassificationPipeline,bl=d.TokenClassifierOutput,xl=d.TokenizerModel,yl=d.TrOCRForCausalLM,kl=d.TrOCRPreTrainedModel,Fl=d.TranslationPipeline,Pl=d.UniSpeechForCTC,Cl=d.UniSpeechForSequenceClassification,vl=d.UniSpeechModel,Sl=d.UniSpeechPreTrainedModel,Al=d.UniSpeechSatForAudioFrameClassification,El=d.UniSpeechSatForCTC,Ll=d.UniSpeechSatForSequenceClassification,zl=d.UniSpeechSatModel,Il=d.UniSpeechSatPreTrainedModel,Bl=d.ViTFeatureExtractor,Nl=d.ViTForImageClassification,Dl=d.ViTImageProcessor,Ol=d.ViTMAEModel,Vl=d.ViTMAEPreTrainedModel,jl=d.ViTMSNForImageClassification,Rl=d.ViTMSNModel,Gl=d.ViTMSNPreTrainedModel,ql=d.ViTModel,Wl=d.ViTPreTrainedModel,$l=d.VisionEncoderDecoderModel,Ul=d.VitMatteForImageMatting,Xl=d.VitMatteImageProcessor,Ql=d.VitMattePreTrainedModel,Hl=d.VitsModel,Yl=d.VitsModelOutput,Jl=d.VitsPreTrainedModel,Kl=d.VitsTokenizer,Zl=d.Wav2Vec2BertForCTC,ec=d.Wav2Vec2BertForSequenceClassification,tc=d.Wav2Vec2BertModel,rc=d.Wav2Vec2BertPreTrainedModel,nc=d.Wav2Vec2CTCTokenizer,oc=d.Wav2Vec2FeatureExtractor,sc=d.Wav2Vec2ForAudioFrameClassification,ac=d.Wav2Vec2ForCTC,ic=d.Wav2Vec2ForSequenceClassification,lc=d.Wav2Vec2Model,cc=d.Wav2Vec2PreTrainedModel,dc=d.Wav2Vec2ProcessorWithLM,uc=d.WavLMForAudioFrameClassification,pc=d.WavLMForCTC,hc=d.WavLMForSequenceClassification,mc=d.WavLMForXVector,_c=d.WavLMModel,fc=d.WavLMPreTrainedModel,gc=d.WeSpeakerFeatureExtractor,Mc=d.WeSpeakerResNetModel,wc=d.WeSpeakerResNetPreTrainedModel,Tc=d.WhisperFeatureExtractor,bc=d.WhisperForConditionalGeneration,xc=d.WhisperModel,yc=d.WhisperPreTrainedModel,kc=d.WhisperProcessor,Fc=d.WhisperTextStreamer,Pc=d.WhisperTokenizer,Cc=d.XLMForQuestionAnswering,vc=d.XLMForSequenceClassification,Sc=d.XLMForTokenClassification,Ac=d.XLMModel,Ec=d.XLMPreTrainedModel,Lc=d.XLMRobertaForMaskedLM,zc=d.XLMRobertaForQuestionAnswering,Ic=d.XLMRobertaForSequenceClassification,Bc=d.XLMRobertaForTokenClassification,Nc=d.XLMRobertaModel,Dc=d.XLMRobertaPreTrainedModel,Oc=d.XLMRobertaTokenizer,Vc=d.XLMTokenizer,jc=d.XLMWithLMHeadModel,Rc=d.XVectorOutput,Gc=d.YolosFeatureExtractor,qc=d.YolosForObjectDetection,Wc=d.YolosModel,$c=d.YolosObjectDetectionOutput,Uc=d.YolosPreTrainedModel,Xc=d.ZeroShotAudioClassificationPipeline,Qc=d.ZeroShotClassificationPipeline,Hc=d.ZeroShotImageClassificationPipeline,Yc=d.ZeroShotObjectDetectionPipeline,Jc=d.bankers_round,Kc=d.cat,Zc=d.cos_sim,ed=d.dot,td=d.dynamic_time_warping,rd=d.env,nd=d.full,od=d.full_like,sd=d.getKeyValueShapes,ad=d.hamming,id=d.hanning,ld=d.interpolate,cd=d.interpolate_4d,dd=d.interpolate_data,ud=d.is_chinese_char,pd=d.layer_norm,hd=d.log_softmax,md=d.magnitude,_d=d.matmul,fd=d.max,gd=d.mean,Md=d.mean_pooling,wd=d.medianFilter,Td=d.mel_filter_bank,bd=d.min,xd=d.ones,yd=d.ones_like,kd=d.permute,Fd=d.permute_data,Pd=d.pipeline,Cd=d.quantize_embeddings,vd=d.read_audio,Sd=d.rfft,Ad=d.round,Ed=d.softmax,Ld=d.spectrogram,zd=d.stack,Id=d.std_mean,Bd=d.topk,Nd=d.window_function,Dd=d.zeros,Od=d.zeros_like;export{y as ASTFeatureExtractor,k as ASTForAudioClassification,F as ASTModel,P as ASTPreTrainedModel,C as AlbertForMaskedLM,v as AlbertForQuestionAnswering,S as AlbertForSequenceClassification,A as AlbertModel,E as AlbertPreTrainedModel,L as AlbertTokenizer,z as AudioClassificationPipeline,I as AutoConfig,B as AutoModel,N as AutoModelForAudioClassification,D as AutoModelForAudioFrameClassification,O as AutoModelForCTC,V as AutoModelForCausalLM,j as AutoModelForDepthEstimation,R as AutoModelForDocumentQuestionAnswering,G as AutoModelForImageClassification,q as AutoModelForImageFeatureExtraction,W as AutoModelForImageMatting,$ as AutoModelForImageSegmentation,U as AutoModelForImageToImage,X as AutoModelForMaskGeneration,Q as AutoModelForMaskedLM,H as AutoModelForNormalEstimation,Y as AutoModelForObjectDetection,J as AutoModelForQuestionAnswering,K as AutoModelForSemanticSegmentation,Z as AutoModelForSeq2SeqLM,ee as AutoModelForSequenceClassification,te as AutoModelForSpeechSeq2Seq,re as AutoModelForTextToSpectrogram,ne as AutoModelForTextToWaveform,oe as AutoModelForTokenClassification,se as AutoModelForUniversalSegmentation,ae as AutoModelForVision2Seq,ie as AutoModelForXVector,le as AutoModelForZeroShotObjectDetection,ce as AutoProcessor,de as AutoTokenizer,ue as AutomaticSpeechRecognitionPipeline,pe as BartForConditionalGeneration,he as BartForSequenceClassification,me as BartModel,_e as BartPretrainedModel,fe as BartTokenizer,ge as BaseModelOutput,Me as BaseStreamer,we as BeitFeatureExtractor,Te as BeitForImageClassification,be as BeitModel,xe as BeitPreTrainedModel,ye as BertForMaskedLM,ke as BertForQuestionAnswering,Fe as BertForSequenceClassification,Pe as BertForTokenClassification,Ce as BertModel,ve as BertPreTrainedModel,Se as BertTokenizer,Ae as BitImageProcessor,Ee as BlenderbotForConditionalGeneration,Le as BlenderbotModel,ze as BlenderbotPreTrainedModel,Ie as BlenderbotSmallForConditionalGeneration,Be as BlenderbotSmallModel,Ne as BlenderbotSmallPreTrainedModel,De as BlenderbotSmallTokenizer,Oe as BlenderbotTokenizer,Ve as BloomForCausalLM,je as BloomModel,Re as BloomPreTrainedModel,Ge as BloomTokenizer,qe as CLIPFeatureExtractor,We as CLIPImageProcessor,$e as CLIPModel,Ue as CLIPPreTrainedModel,Xe as CLIPSegForImageSegmentation,Qe as CLIPSegModel,He as CLIPSegPreTrainedModel,Ye as CLIPTextModel,Je as CLIPTextModelWithProjection,Ke as CLIPTokenizer,Ze as CLIPVisionModel,et as CLIPVisionModelWithProjection,tt as CamembertForMaskedLM,rt as CamembertForQuestionAnswering,nt as CamembertForSequenceClassification,ot as CamembertForTokenClassification,st as CamembertModel,at as CamembertPreTrainedModel,it as CamembertTokenizer,lt as CausalLMOutput,ct as CausalLMOutputWithPast,dt as ChineseCLIPFeatureExtractor,ut as ChineseCLIPModel,pt as ChineseCLIPPreTrainedModel,ht as ClapAudioModelWithProjection,mt as ClapFeatureExtractor,_t as ClapModel,ft as ClapPreTrainedModel,gt as ClapTextModelWithProjection,Mt as CodeGenForCausalLM,wt as CodeGenModel,Tt as CodeGenPreTrainedModel,bt as CodeGenTokenizer,xt as CodeLlamaTokenizer,yt as CohereForCausalLM,kt as CohereModel,Ft as CoherePreTrainedModel,Pt as CohereTokenizer,Ct as ConvBertForMaskedLM,vt as ConvBertForQuestionAnswering,St as ConvBertForSequenceClassification,At as ConvBertForTokenClassification,Et as ConvBertModel,Lt as ConvBertPreTrainedModel,zt as ConvBertTokenizer,It as ConvNextFeatureExtractor,Bt as ConvNextForImageClassification,Nt as ConvNextImageProcessor,Dt as ConvNextModel,Ot as ConvNextPreTrainedModel,Vt as ConvNextV2ForImageClassification,jt as ConvNextV2Model,Rt as ConvNextV2PreTrainedModel,Gt as DPTFeatureExtractor,qt as DPTForDepthEstimation,Wt as DPTImageProcessor,$t as DPTModel,Ut as DPTPreTrainedModel,Xt as DebertaForMaskedLM,Qt as DebertaForQuestionAnswering,Ht as DebertaForSequenceClassification,Yt as DebertaForTokenClassification,Jt as DebertaModel,Kt as DebertaPreTrainedModel,Zt as DebertaTokenizer,er as DebertaV2ForMaskedLM,tr as DebertaV2ForQuestionAnswering,rr as DebertaV2ForSequenceClassification,nr as DebertaV2ForTokenClassification,or as DebertaV2Model,sr as DebertaV2PreTrainedModel,ar as DebertaV2Tokenizer,ir as DecisionTransformerModel,lr as DecisionTransformerPreTrainedModel,cr as DeiTFeatureExtractor,dr as DeiTForImageClassification,ur as DeiTModel,pr as DeiTPreTrainedModel,hr as DepthAnythingForDepthEstimation,mr as DepthAnythingPreTrainedModel,_r as DepthEstimationPipeline,fr as DepthProForDepthEstimation,gr as DepthProPreTrainedModel,Mr as DetrFeatureExtractor,wr as DetrForObjectDetection,Tr as DetrForSegmentation,br as DetrModel,xr as DetrObjectDetectionOutput,yr as DetrPreTrainedModel,kr as DetrSegmentationOutput,Fr as Dinov2ForImageClassification,Pr as Dinov2Model,Cr as Dinov2PreTrainedModel,vr as DistilBertForMaskedLM,Sr as DistilBertForQuestionAnswering,Ar as DistilBertForSequenceClassification,Er as DistilBertForTokenClassification,Lr as DistilBertModel,zr as DistilBertPreTrainedModel,Ir as DistilBertTokenizer,Br as DocumentQuestionAnsweringPipeline,Nr as DonutFeatureExtractor,Dr as DonutSwinModel,Or as DonutSwinPreTrainedModel,Vr as EfficientNetForImageClassification,jr as EfficientNetImageProcessor,Rr as EfficientNetModel,Gr as EfficientNetPreTrainedModel,qr as ElectraForMaskedLM,Wr as ElectraForQuestionAnswering,$r as ElectraForSequenceClassification,Ur as ElectraForTokenClassification,Xr as ElectraModel,Qr as ElectraPreTrainedModel,Hr as ElectraTokenizer,Yr as EosTokenCriteria,Jr as EsmForMaskedLM,Kr as EsmForSequenceClassification,Zr as EsmForTokenClassification,en as EsmModel,tn as EsmPreTrainedModel,rn as EsmTokenizer,nn as FFT,on as FalconForCausalLM,sn as FalconModel,an as FalconPreTrainedModel,ln as FalconTokenizer,cn as FastViTForImageClassification,dn as FastViTModel,un as FastViTPreTrainedModel,pn as FeatureExtractionPipeline,hn as FeatureExtractor,mn as FillMaskPipeline,_n as Florence2ForConditionalGeneration,fn as Florence2PreTrainedModel,gn as Florence2Processor,Mn as GLPNFeatureExtractor,wn as GLPNForDepthEstimation,Tn as GLPNModel,bn as GLPNPreTrainedModel,xn as GPT2LMHeadModel,yn as GPT2Model,kn as GPT2PreTrainedModel,Fn as GPT2Tokenizer,Pn as GPTBigCodeForCausalLM,Cn as GPTBigCodeModel,vn as GPTBigCodePreTrainedModel,Sn as GPTJForCausalLM,An as GPTJModel,En as GPTJPreTrainedModel,Ln as GPTNeoForCausalLM,zn as GPTNeoModel,In as GPTNeoPreTrainedModel,Bn as GPTNeoXForCausalLM,Nn as GPTNeoXModel,Dn as GPTNeoXPreTrainedModel,On as GPTNeoXTokenizer,Vn as Gemma2ForCausalLM,jn as Gemma2Model,Rn as Gemma2PreTrainedModel,Gn as GemmaForCausalLM,qn as GemmaModel,Wn as GemmaPreTrainedModel,$n as GemmaTokenizer,Un as Grok1Tokenizer,Xn as GroupViTModel,Qn as GroupViTPreTrainedModel,Hn as HerbertTokenizer,Yn as HieraForImageClassification,Jn as HieraModel,Kn as HieraPreTrainedModel,Zn as HubertForCTC,eo as HubertForSequenceClassification,to as HubertModel,ro as HubertPreTrainedModel,no as ImageClassificationPipeline,oo as ImageFeatureExtractionPipeline,so as ImageFeatureExtractor,ao as ImageMattingOutput,io as ImageSegmentationPipeline,lo as ImageToImagePipeline,co as ImageToTextPipeline,uo as InterruptableStoppingCriteria,po as JAISLMHeadModel,ho as JAISModel,mo as JAISPreTrainedModel,_o as LlamaForCausalLM,fo as LlamaModel,go as LlamaPreTrainedModel,Mo as LlamaTokenizer,wo as LlavaForConditionalGeneration,To as LlavaPreTrainedModel,bo as LongT5ForConditionalGeneration,xo as LongT5Model,yo as LongT5PreTrainedModel,ko as M2M100ForConditionalGeneration,Fo as M2M100Model,Po as M2M100PreTrainedModel,Co as M2M100Tokenizer,vo as MBart50Tokenizer,So as MBartForCausalLM,Ao as MBartForConditionalGeneration,Eo as MBartForSequenceClassification,Lo as MBartModel,zo as MBartPreTrainedModel,Io as MBartTokenizer,Bo as MPNetForMaskedLM,No as MPNetForQuestionAnswering,Do as MPNetForSequenceClassification,Oo as MPNetForTokenClassification,Vo as MPNetModel,jo as MPNetPreTrainedModel,Ro as MPNetTokenizer,Go as MT5ForConditionalGeneration,qo as MT5Model,Wo as MT5PreTrainedModel,$o as MarianMTModel,Uo as MarianModel,Xo as MarianPreTrainedModel,Qo as MarianTokenizer,Ho as MaskFormerFeatureExtractor,Yo as MaskFormerForInstanceSegmentation,Jo as MaskFormerModel,Ko as MaskFormerPreTrainedModel,Zo as MaskedLMOutput,es as MaxLengthCriteria,ts as MistralForCausalLM,rs as MistralModel,ns as MistralPreTrainedModel,os as MobileBertForMaskedLM,ss as MobileBertForQuestionAnswering,as as MobileBertForSequenceClassification,is as MobileBertModel,ls as MobileBertPreTrainedModel,cs as MobileBertTokenizer,ds as MobileNetV1FeatureExtractor,us as MobileNetV1ForImageClassification,ps as MobileNetV1Model,hs as MobileNetV1PreTrainedModel,ms as MobileNetV2FeatureExtractor,_s as MobileNetV2ForImageClassification,fs as MobileNetV2Model,gs as MobileNetV2PreTrainedModel,Ms as MobileNetV3FeatureExtractor,ws as MobileNetV3ForImageClassification,Ts as MobileNetV3Model,bs as MobileNetV3PreTrainedModel,xs as MobileNetV4FeatureExtractor,ys as MobileNetV4ForImageClassification,ks as MobileNetV4Model,Fs as MobileNetV4PreTrainedModel,Ps as MobileViTFeatureExtractor,Cs as MobileViTForImageClassification,vs as MobileViTImageProcessor,Ss as MobileViTModel,As as MobileViTPreTrainedModel,Es as MobileViTV2ForImageClassification,Ls as MobileViTV2Model,zs as MobileViTV2PreTrainedModel,Is as ModelOutput,Bs as Moondream1ForConditionalGeneration,Ns as MptForCausalLM,Ds as MptModel,Os as MptPreTrainedModel,Vs as MusicgenForCausalLM,js as MusicgenForConditionalGeneration,Rs as MusicgenModel,Gs as MusicgenPreTrainedModel,qs as NllbTokenizer,Ws as NomicBertModel,$s as NomicBertPreTrainedModel,Us as NougatImageProcessor,Xs as NougatTokenizer,Qs as OPTForCausalLM,Hs as OPTModel,Ys as OPTPreTrainedModel,Js as ObjectDetectionPipeline,Ks as OpenELMForCausalLM,Zs as OpenELMModel,ea as OpenELMPreTrainedModel,ta as OwlViTFeatureExtractor,ra as OwlViTForObjectDetection,na as OwlViTModel,oa as OwlViTPreTrainedModel,sa as OwlViTProcessor,aa as Owlv2ForObjectDetection,ia as Owlv2ImageProcessor,la as Owlv2Model,ca as Owlv2PreTrainedModel,da as Phi3ForCausalLM,ua as Phi3Model,pa as Phi3PreTrainedModel,ha as PhiForCausalLM,ma as PhiModel,_a as PhiPreTrainedModel,fa as Pipeline,ga as PreTrainedModel,Ma as PreTrainedTokenizer,wa as PretrainedConfig,Ta as PretrainedMixin,ba as Processor,xa as PvtForImageClassification,ya as PvtImageProcessor,ka as PvtModel,Fa as PvtPreTrainedModel,Pa as PyAnnoteFeatureExtractor,Ca as PyAnnoteForAudioFrameClassification,va as PyAnnoteModel,Sa as PyAnnotePreTrainedModel,Aa as PyAnnoteProcessor,Ea as QuestionAnsweringModelOutput,La as QuestionAnsweringPipeline,za as Qwen2ForCausalLM,Ia as Qwen2Model,Ba as Qwen2PreTrainedModel,Na as Qwen2Tokenizer,Da as RTDetrForObjectDetection,Oa as RTDetrImageProcessor,Va as RTDetrModel,ja as RTDetrObjectDetectionOutput,Ra as RTDetrPreTrainedModel,Ga as RawImage,qa as ResNetForImageClassification,Wa as ResNetModel,$a as ResNetPreTrainedModel,Ua as RoFormerForMaskedLM,Xa as RoFormerForQuestionAnswering,Qa as RoFormerForSequenceClassification,Ha as RoFormerForTokenClassification,Ya as RoFormerModel,Ja as RoFormerPreTrainedModel,Ka as RoFormerTokenizer,Za as RobertaForMaskedLM,ei as RobertaForQuestionAnswering,ti as RobertaForSequenceClassification,ri as RobertaForTokenClassification,ni as RobertaModel,oi as RobertaPreTrainedModel,si as RobertaTokenizer,ai as SamImageProcessor,ii as SamImageSegmentationOutput,li as SamModel,ci as SamPreTrainedModel,di as SamProcessor,ui as SapiensFeatureExtractor,pi as SapiensForDepthEstimation,hi as SapiensForNormalEstimation,mi as SapiensForSemanticSegmentation,_i as SapiensPreTrainedModel,fi as SeamlessM4TFeatureExtractor,gi as SegformerFeatureExtractor,Mi as SegformerForImageClassification,wi as SegformerForSemanticSegmentation,Ti as SegformerModel,bi as SegformerPreTrainedModel,xi as Seq2SeqLMOutput,yi as SequenceClassifierOutput,ki as SiglipImageProcessor,Fi as SiglipModel,Pi as SiglipPreTrainedModel,Ci as SiglipTextModel,vi as SiglipTokenizer,Si as SiglipVisionModel,Ai as SpeechT5FeatureExtractor,Ei as SpeechT5ForSpeechToText,Li as SpeechT5ForTextToSpeech,zi as SpeechT5HifiGan,Ii as SpeechT5Model,Bi as SpeechT5PreTrainedModel,Ni as SpeechT5Processor,Di as SpeechT5Tokenizer,Oi as SqueezeBertForMaskedLM,Vi as SqueezeBertForQuestionAnswering,ji as SqueezeBertForSequenceClassification,Ri as SqueezeBertModel,Gi as SqueezeBertPreTrainedModel,qi as SqueezeBertTokenizer,Wi as StableLmForCausalLM,$i as StableLmModel,Ui as StableLmPreTrainedModel,Xi as Starcoder2ForCausalLM,Qi as Starcoder2Model,Hi as Starcoder2PreTrainedModel,Yi as StoppingCriteria,Ji as StoppingCriteriaList,Ki as SummarizationPipeline,Zi as Swin2SRForImageSuperResolution,el as Swin2SRImageProcessor,tl as Swin2SRModel,rl as Swin2SRPreTrainedModel,nl as SwinForImageClassification,ol as SwinModel,sl as SwinPreTrainedModel,al as T5ForConditionalGeneration,il as T5Model,ll as T5PreTrainedModel,cl as T5Tokenizer,dl as TableTransformerForObjectDetection,ul as TableTransformerModel,pl as TableTransformerObjectDetectionOutput,hl as TableTransformerPreTrainedModel,ml as Tensor,_l as Text2TextGenerationPipeline,fl as TextClassificationPipeline,gl as TextGenerationPipeline,Ml as TextStreamer,wl as TextToAudioPipeline,Tl as TokenClassificationPipeline,bl as TokenClassifierOutput,xl as TokenizerModel,yl as TrOCRForCausalLM,kl as TrOCRPreTrainedModel,Fl as TranslationPipeline,Pl as UniSpeechForCTC,Cl as UniSpeechForSequenceClassification,vl as UniSpeechModel,Sl as UniSpeechPreTrainedModel,Al as UniSpeechSatForAudioFrameClassification,El as UniSpeechSatForCTC,Ll as UniSpeechSatForSequenceClassification,zl as UniSpeechSatModel,Il as UniSpeechSatPreTrainedModel,Bl as ViTFeatureExtractor,Nl as ViTForImageClassification,Dl as ViTImageProcessor,Ol as ViTMAEModel,Vl as ViTMAEPreTrainedModel,jl as ViTMSNForImageClassification,Rl as ViTMSNModel,Gl as ViTMSNPreTrainedModel,ql as ViTModel,Wl as ViTPreTrainedModel,$l as VisionEncoderDecoderModel,Ul as VitMatteForImageMatting,Xl as VitMatteImageProcessor,Ql as VitMattePreTrainedModel,Hl as VitsModel,Yl as VitsModelOutput,Jl as VitsPreTrainedModel,Kl as VitsTokenizer,Zl as Wav2Vec2BertForCTC,ec as Wav2Vec2BertForSequenceClassification,tc as Wav2Vec2BertModel,rc as Wav2Vec2BertPreTrainedModel,nc as Wav2Vec2CTCTokenizer,oc as Wav2Vec2FeatureExtractor,sc as Wav2Vec2ForAudioFrameClassification,ac as Wav2Vec2ForCTC,ic as Wav2Vec2ForSequenceClassification,lc as Wav2Vec2Model,cc as Wav2Vec2PreTrainedModel,dc as Wav2Vec2ProcessorWithLM,uc as WavLMForAudioFrameClassification,pc as WavLMForCTC,hc as WavLMForSequenceClassification,mc as WavLMForXVector,_c as WavLMModel,fc as WavLMPreTrainedModel,gc as WeSpeakerFeatureExtractor,Mc as WeSpeakerResNetModel,wc as WeSpeakerResNetPreTrainedModel,Tc as WhisperFeatureExtractor,bc as WhisperForConditionalGeneration,xc as WhisperModel,yc as WhisperPreTrainedModel,kc as WhisperProcessor,Fc as WhisperTextStreamer,Pc as WhisperTokenizer,Cc as XLMForQuestionAnswering,vc as XLMForSequenceClassification,Sc as XLMForTokenClassification,Ac as XLMModel,Ec as XLMPreTrainedModel,Lc as XLMRobertaForMaskedLM,zc as XLMRobertaForQuestionAnswering,Ic as XLMRobertaForSequenceClassification,Bc as XLMRobertaForTokenClassification,Nc as XLMRobertaModel,Dc as XLMRobertaPreTrainedModel,Oc as XLMRobertaTokenizer,Vc as XLMTokenizer,jc as XLMWithLMHeadModel,Rc as XVectorOutput,Gc as YolosFeatureExtractor,qc as YolosForObjectDetection,Wc as YolosModel,$c as YolosObjectDetectionOutput,Uc as YolosPreTrainedModel,Xc as ZeroShotAudioClassificationPipeline,Qc as ZeroShotClassificationPipeline,Hc as ZeroShotImageClassificationPipeline,Yc as ZeroShotObjectDetectionPipeline,Jc as bankers_round,Kc as cat,Zc as cos_sim,ed as dot,td as dynamic_time_warping,rd as env,nd as full,od as full_like,sd as getKeyValueShapes,ad as hamming,id as hanning,ld as interpolate,cd as interpolate_4d,dd as interpolate_data,ud as is_chinese_char,pd as layer_norm,hd as log_softmax,md as magnitude,_d as matmul,fd as max,gd as mean,Md as mean_pooling,wd as medianFilter,Td as mel_filter_bank,bd as min,xd as ones,yd as ones_like,kd as permute,Fd as permute_data,Pd as pipeline,Cd as quantize_embeddings,vd as read_audio,Sd as rfft,Ad as round,Ed as softmax,Ld as spectrogram,zd as stack,Id as std_mean,Bd as topk,Nd as window_function,Dd as zeros,Od as zeros_like};
+  \*****************************/c.r(d),c.d(d,{ASTFeatureExtractor:()=>_.ASTFeatureExtractor,ASTForAudioClassification:()=>h.ASTForAudioClassification,ASTModel:()=>h.ASTModel,ASTPreTrainedModel:()=>h.ASTPreTrainedModel,AlbertForMaskedLM:()=>h.AlbertForMaskedLM,AlbertForQuestionAnswering:()=>h.AlbertForQuestionAnswering,AlbertForSequenceClassification:()=>h.AlbertForSequenceClassification,AlbertModel:()=>h.AlbertModel,AlbertPreTrainedModel:()=>h.AlbertPreTrainedModel,AlbertTokenizer:()=>m.AlbertTokenizer,AudioClassificationPipeline:()=>p.AudioClassificationPipeline,AutoConfig:()=>f.AutoConfig,AutoModel:()=>h.AutoModel,AutoModelForAudioClassification:()=>h.AutoModelForAudioClassification,AutoModelForAudioFrameClassification:()=>h.AutoModelForAudioFrameClassification,AutoModelForCTC:()=>h.AutoModelForCTC,AutoModelForCausalLM:()=>h.AutoModelForCausalLM,AutoModelForDepthEstimation:()=>h.AutoModelForDepthEstimation,AutoModelForDocumentQuestionAnswering:()=>h.AutoModelForDocumentQuestionAnswering,AutoModelForImageClassification:()=>h.AutoModelForImageClassification,AutoModelForImageFeatureExtraction:()=>h.AutoModelForImageFeatureExtraction,AutoModelForImageMatting:()=>h.AutoModelForImageMatting,AutoModelForImageSegmentation:()=>h.AutoModelForImageSegmentation,AutoModelForImageToImage:()=>h.AutoModelForImageToImage,AutoModelForMaskGeneration:()=>h.AutoModelForMaskGeneration,AutoModelForMaskedLM:()=>h.AutoModelForMaskedLM,AutoModelForNormalEstimation:()=>h.AutoModelForNormalEstimation,AutoModelForObjectDetection:()=>h.AutoModelForObjectDetection,AutoModelForQuestionAnswering:()=>h.AutoModelForQuestionAnswering,AutoModelForSemanticSegmentation:()=>h.AutoModelForSemanticSegmentation,AutoModelForSeq2SeqLM:()=>h.AutoModelForSeq2SeqLM,AutoModelForSequenceClassification:()=>h.AutoModelForSequenceClassification,AutoModelForSpeechSeq2Seq:()=>h.AutoModelForSpeechSeq2Seq,AutoModelForTextToSpectrogram:()=>h.AutoModelForTextToSpectrogram,AutoModelForTextToWaveform:()=>h.AutoModelForTextToWaveform,AutoModelForTokenClassification:()=>h.AutoModelForTokenClassification,AutoModelForUniversalSegmentation:()=>h.AutoModelForUniversalSegmentation,AutoModelForVision2Seq:()=>h.AutoModelForVision2Seq,AutoModelForXVector:()=>h.AutoModelForXVector,AutoModelForZeroShotObjectDetection:()=>h.AutoModelForZeroShotObjectDetection,AutoProcessor:()=>_.AutoProcessor,AutoTokenizer:()=>m.AutoTokenizer,AutomaticSpeechRecognitionPipeline:()=>p.AutomaticSpeechRecognitionPipeline,BartForConditionalGeneration:()=>h.BartForConditionalGeneration,BartForSequenceClassification:()=>h.BartForSequenceClassification,BartModel:()=>h.BartModel,BartPretrainedModel:()=>h.BartPretrainedModel,BartTokenizer:()=>m.BartTokenizer,BaseModelOutput:()=>h.BaseModelOutput,BaseStreamer:()=>b.BaseStreamer,BeitFeatureExtractor:()=>_.BeitFeatureExtractor,BeitForImageClassification:()=>h.BeitForImageClassification,BeitModel:()=>h.BeitModel,BeitPreTrainedModel:()=>h.BeitPreTrainedModel,BertForMaskedLM:()=>h.BertForMaskedLM,BertForQuestionAnswering:()=>h.BertForQuestionAnswering,BertForSequenceClassification:()=>h.BertForSequenceClassification,BertForTokenClassification:()=>h.BertForTokenClassification,BertModel:()=>h.BertModel,BertPreTrainedModel:()=>h.BertPreTrainedModel,BertTokenizer:()=>m.BertTokenizer,BitImageProcessor:()=>_.BitImageProcessor,BlenderbotForConditionalGeneration:()=>h.BlenderbotForConditionalGeneration,BlenderbotModel:()=>h.BlenderbotModel,BlenderbotPreTrainedModel:()=>h.BlenderbotPreTrainedModel,BlenderbotSmallForConditionalGeneration:()=>h.BlenderbotSmallForConditionalGeneration,BlenderbotSmallModel:()=>h.BlenderbotSmallModel,BlenderbotSmallPreTrainedModel:()=>h.BlenderbotSmallPreTrainedModel,BlenderbotSmallTokenizer:()=>m.BlenderbotSmallTokenizer,BlenderbotTokenizer:()=>m.BlenderbotTokenizer,BloomForCausalLM:()=>h.BloomForCausalLM,BloomModel:()=>h.BloomModel,BloomPreTrainedModel:()=>h.BloomPreTrainedModel,BloomTokenizer:()=>m.BloomTokenizer,CLIPFeatureExtractor:()=>_.CLIPFeatureExtractor,CLIPImageProcessor:()=>_.CLIPImageProcessor,CLIPModel:()=>h.CLIPModel,CLIPPreTrainedModel:()=>h.CLIPPreTrainedModel,CLIPSegForImageSegmentation:()=>h.CLIPSegForImageSegmentation,CLIPSegModel:()=>h.CLIPSegModel,CLIPSegPreTrainedModel:()=>h.CLIPSegPreTrainedModel,CLIPTextModel:()=>h.CLIPTextModel,CLIPTextModelWithProjection:()=>h.CLIPTextModelWithProjection,CLIPTokenizer:()=>m.CLIPTokenizer,CLIPVisionModel:()=>h.CLIPVisionModel,CLIPVisionModelWithProjection:()=>h.CLIPVisionModelWithProjection,CamembertForMaskedLM:()=>h.CamembertForMaskedLM,CamembertForQuestionAnswering:()=>h.CamembertForQuestionAnswering,CamembertForSequenceClassification:()=>h.CamembertForSequenceClassification,CamembertForTokenClassification:()=>h.CamembertForTokenClassification,CamembertModel:()=>h.CamembertModel,CamembertPreTrainedModel:()=>h.CamembertPreTrainedModel,CamembertTokenizer:()=>m.CamembertTokenizer,CausalLMOutput:()=>h.CausalLMOutput,CausalLMOutputWithPast:()=>h.CausalLMOutputWithPast,ChineseCLIPFeatureExtractor:()=>_.ChineseCLIPFeatureExtractor,ChineseCLIPModel:()=>h.ChineseCLIPModel,ChineseCLIPPreTrainedModel:()=>h.ChineseCLIPPreTrainedModel,ClapAudioModelWithProjection:()=>h.ClapAudioModelWithProjection,ClapFeatureExtractor:()=>_.ClapFeatureExtractor,ClapModel:()=>h.ClapModel,ClapPreTrainedModel:()=>h.ClapPreTrainedModel,ClapTextModelWithProjection:()=>h.ClapTextModelWithProjection,CodeGenForCausalLM:()=>h.CodeGenForCausalLM,CodeGenModel:()=>h.CodeGenModel,CodeGenPreTrainedModel:()=>h.CodeGenPreTrainedModel,CodeGenTokenizer:()=>m.CodeGenTokenizer,CodeLlamaTokenizer:()=>m.CodeLlamaTokenizer,CohereForCausalLM:()=>h.CohereForCausalLM,CohereModel:()=>h.CohereModel,CoherePreTrainedModel:()=>h.CoherePreTrainedModel,CohereTokenizer:()=>m.CohereTokenizer,ConvBertForMaskedLM:()=>h.ConvBertForMaskedLM,ConvBertForQuestionAnswering:()=>h.ConvBertForQuestionAnswering,ConvBertForSequenceClassification:()=>h.ConvBertForSequenceClassification,ConvBertForTokenClassification:()=>h.ConvBertForTokenClassification,ConvBertModel:()=>h.ConvBertModel,ConvBertPreTrainedModel:()=>h.ConvBertPreTrainedModel,ConvBertTokenizer:()=>m.ConvBertTokenizer,ConvNextFeatureExtractor:()=>_.ConvNextFeatureExtractor,ConvNextForImageClassification:()=>h.ConvNextForImageClassification,ConvNextImageProcessor:()=>_.ConvNextImageProcessor,ConvNextModel:()=>h.ConvNextModel,ConvNextPreTrainedModel:()=>h.ConvNextPreTrainedModel,ConvNextV2ForImageClassification:()=>h.ConvNextV2ForImageClassification,ConvNextV2Model:()=>h.ConvNextV2Model,ConvNextV2PreTrainedModel:()=>h.ConvNextV2PreTrainedModel,DPTFeatureExtractor:()=>_.DPTFeatureExtractor,DPTForDepthEstimation:()=>h.DPTForDepthEstimation,DPTImageProcessor:()=>_.DPTImageProcessor,DPTModel:()=>h.DPTModel,DPTPreTrainedModel:()=>h.DPTPreTrainedModel,DebertaForMaskedLM:()=>h.DebertaForMaskedLM,DebertaForQuestionAnswering:()=>h.DebertaForQuestionAnswering,DebertaForSequenceClassification:()=>h.DebertaForSequenceClassification,DebertaForTokenClassification:()=>h.DebertaForTokenClassification,DebertaModel:()=>h.DebertaModel,DebertaPreTrainedModel:()=>h.DebertaPreTrainedModel,DebertaTokenizer:()=>m.DebertaTokenizer,DebertaV2ForMaskedLM:()=>h.DebertaV2ForMaskedLM,DebertaV2ForQuestionAnswering:()=>h.DebertaV2ForQuestionAnswering,DebertaV2ForSequenceClassification:()=>h.DebertaV2ForSequenceClassification,DebertaV2ForTokenClassification:()=>h.DebertaV2ForTokenClassification,DebertaV2Model:()=>h.DebertaV2Model,DebertaV2PreTrainedModel:()=>h.DebertaV2PreTrainedModel,DebertaV2Tokenizer:()=>m.DebertaV2Tokenizer,DecisionTransformerModel:()=>h.DecisionTransformerModel,DecisionTransformerPreTrainedModel:()=>h.DecisionTransformerPreTrainedModel,DeiTFeatureExtractor:()=>_.DeiTFeatureExtractor,DeiTForImageClassification:()=>h.DeiTForImageClassification,DeiTModel:()=>h.DeiTModel,DeiTPreTrainedModel:()=>h.DeiTPreTrainedModel,DepthAnythingForDepthEstimation:()=>h.DepthAnythingForDepthEstimation,DepthAnythingPreTrainedModel:()=>h.DepthAnythingPreTrainedModel,DepthEstimationPipeline:()=>p.DepthEstimationPipeline,DepthProForDepthEstimation:()=>h.DepthProForDepthEstimation,DepthProPreTrainedModel:()=>h.DepthProPreTrainedModel,DetrFeatureExtractor:()=>_.DetrFeatureExtractor,DetrForObjectDetection:()=>h.DetrForObjectDetection,DetrForSegmentation:()=>h.DetrForSegmentation,DetrModel:()=>h.DetrModel,DetrObjectDetectionOutput:()=>h.DetrObjectDetectionOutput,DetrPreTrainedModel:()=>h.DetrPreTrainedModel,DetrSegmentationOutput:()=>h.DetrSegmentationOutput,Dinov2ForImageClassification:()=>h.Dinov2ForImageClassification,Dinov2Model:()=>h.Dinov2Model,Dinov2PreTrainedModel:()=>h.Dinov2PreTrainedModel,DistilBertForMaskedLM:()=>h.DistilBertForMaskedLM,DistilBertForQuestionAnswering:()=>h.DistilBertForQuestionAnswering,DistilBertForSequenceClassification:()=>h.DistilBertForSequenceClassification,DistilBertForTokenClassification:()=>h.DistilBertForTokenClassification,DistilBertModel:()=>h.DistilBertModel,DistilBertPreTrainedModel:()=>h.DistilBertPreTrainedModel,DistilBertTokenizer:()=>m.DistilBertTokenizer,DocumentQuestionAnsweringPipeline:()=>p.DocumentQuestionAnsweringPipeline,DonutFeatureExtractor:()=>_.DonutFeatureExtractor,DonutSwinModel:()=>h.DonutSwinModel,DonutSwinPreTrainedModel:()=>h.DonutSwinPreTrainedModel,EfficientNetForImageClassification:()=>h.EfficientNetForImageClassification,EfficientNetImageProcessor:()=>_.EfficientNetImageProcessor,EfficientNetModel:()=>h.EfficientNetModel,EfficientNetPreTrainedModel:()=>h.EfficientNetPreTrainedModel,ElectraForMaskedLM:()=>h.ElectraForMaskedLM,ElectraForQuestionAnswering:()=>h.ElectraForQuestionAnswering,ElectraForSequenceClassification:()=>h.ElectraForSequenceClassification,ElectraForTokenClassification:()=>h.ElectraForTokenClassification,ElectraModel:()=>h.ElectraModel,ElectraPreTrainedModel:()=>h.ElectraPreTrainedModel,ElectraTokenizer:()=>m.ElectraTokenizer,EosTokenCriteria:()=>x.EosTokenCriteria,EsmForMaskedLM:()=>h.EsmForMaskedLM,EsmForSequenceClassification:()=>h.EsmForSequenceClassification,EsmForTokenClassification:()=>h.EsmForTokenClassification,EsmModel:()=>h.EsmModel,EsmPreTrainedModel:()=>h.EsmPreTrainedModel,EsmTokenizer:()=>m.EsmTokenizer,FFT:()=>T.FFT,FalconForCausalLM:()=>h.FalconForCausalLM,FalconModel:()=>h.FalconModel,FalconPreTrainedModel:()=>h.FalconPreTrainedModel,FalconTokenizer:()=>m.FalconTokenizer,FastViTForImageClassification:()=>h.FastViTForImageClassification,FastViTModel:()=>h.FastViTModel,FastViTPreTrainedModel:()=>h.FastViTPreTrainedModel,FeatureExtractionPipeline:()=>p.FeatureExtractionPipeline,FeatureExtractor:()=>_.FeatureExtractor,FillMaskPipeline:()=>p.FillMaskPipeline,Florence2ForConditionalGeneration:()=>h.Florence2ForConditionalGeneration,Florence2PreTrainedModel:()=>h.Florence2PreTrainedModel,Florence2Processor:()=>_.Florence2Processor,GLPNFeatureExtractor:()=>_.GLPNFeatureExtractor,GLPNForDepthEstimation:()=>h.GLPNForDepthEstimation,GLPNModel:()=>h.GLPNModel,GLPNPreTrainedModel:()=>h.GLPNPreTrainedModel,GPT2LMHeadModel:()=>h.GPT2LMHeadModel,GPT2Model:()=>h.GPT2Model,GPT2PreTrainedModel:()=>h.GPT2PreTrainedModel,GPT2Tokenizer:()=>m.GPT2Tokenizer,GPTBigCodeForCausalLM:()=>h.GPTBigCodeForCausalLM,GPTBigCodeModel:()=>h.GPTBigCodeModel,GPTBigCodePreTrainedModel:()=>h.GPTBigCodePreTrainedModel,GPTJForCausalLM:()=>h.GPTJForCausalLM,GPTJModel:()=>h.GPTJModel,GPTJPreTrainedModel:()=>h.GPTJPreTrainedModel,GPTNeoForCausalLM:()=>h.GPTNeoForCausalLM,GPTNeoModel:()=>h.GPTNeoModel,GPTNeoPreTrainedModel:()=>h.GPTNeoPreTrainedModel,GPTNeoXForCausalLM:()=>h.GPTNeoXForCausalLM,GPTNeoXModel:()=>h.GPTNeoXModel,GPTNeoXPreTrainedModel:()=>h.GPTNeoXPreTrainedModel,GPTNeoXTokenizer:()=>m.GPTNeoXTokenizer,Gemma2ForCausalLM:()=>h.Gemma2ForCausalLM,Gemma2Model:()=>h.Gemma2Model,Gemma2PreTrainedModel:()=>h.Gemma2PreTrainedModel,GemmaForCausalLM:()=>h.GemmaForCausalLM,GemmaModel:()=>h.GemmaModel,GemmaPreTrainedModel:()=>h.GemmaPreTrainedModel,GemmaTokenizer:()=>m.GemmaTokenizer,GraniteForCausalLM:()=>h.GraniteForCausalLM,GraniteModel:()=>h.GraniteModel,GranitePreTrainedModel:()=>h.GranitePreTrainedModel,Grok1Tokenizer:()=>m.Grok1Tokenizer,GroupViTModel:()=>h.GroupViTModel,GroupViTPreTrainedModel:()=>h.GroupViTPreTrainedModel,HerbertTokenizer:()=>m.HerbertTokenizer,HieraForImageClassification:()=>h.HieraForImageClassification,HieraModel:()=>h.HieraModel,HieraPreTrainedModel:()=>h.HieraPreTrainedModel,HubertForCTC:()=>h.HubertForCTC,HubertForSequenceClassification:()=>h.HubertForSequenceClassification,HubertModel:()=>h.HubertModel,HubertPreTrainedModel:()=>h.HubertPreTrainedModel,ImageClassificationPipeline:()=>p.ImageClassificationPipeline,ImageFeatureExtractionPipeline:()=>p.ImageFeatureExtractionPipeline,ImageFeatureExtractor:()=>_.ImageFeatureExtractor,ImageMattingOutput:()=>h.ImageMattingOutput,ImageSegmentationPipeline:()=>p.ImageSegmentationPipeline,ImageToImagePipeline:()=>p.ImageToImagePipeline,ImageToTextPipeline:()=>p.ImageToTextPipeline,InterruptableStoppingCriteria:()=>x.InterruptableStoppingCriteria,JAISLMHeadModel:()=>h.JAISLMHeadModel,JAISModel:()=>h.JAISModel,JAISPreTrainedModel:()=>h.JAISPreTrainedModel,LlamaForCausalLM:()=>h.LlamaForCausalLM,LlamaModel:()=>h.LlamaModel,LlamaPreTrainedModel:()=>h.LlamaPreTrainedModel,LlamaTokenizer:()=>m.LlamaTokenizer,LlavaForConditionalGeneration:()=>h.LlavaForConditionalGeneration,LlavaPreTrainedModel:()=>h.LlavaPreTrainedModel,LongT5ForConditionalGeneration:()=>h.LongT5ForConditionalGeneration,LongT5Model:()=>h.LongT5Model,LongT5PreTrainedModel:()=>h.LongT5PreTrainedModel,M2M100ForConditionalGeneration:()=>h.M2M100ForConditionalGeneration,M2M100Model:()=>h.M2M100Model,M2M100PreTrainedModel:()=>h.M2M100PreTrainedModel,M2M100Tokenizer:()=>m.M2M100Tokenizer,MBart50Tokenizer:()=>m.MBart50Tokenizer,MBartForCausalLM:()=>h.MBartForCausalLM,MBartForConditionalGeneration:()=>h.MBartForConditionalGeneration,MBartForSequenceClassification:()=>h.MBartForSequenceClassification,MBartModel:()=>h.MBartModel,MBartPreTrainedModel:()=>h.MBartPreTrainedModel,MBartTokenizer:()=>m.MBartTokenizer,MPNetForMaskedLM:()=>h.MPNetForMaskedLM,MPNetForQuestionAnswering:()=>h.MPNetForQuestionAnswering,MPNetForSequenceClassification:()=>h.MPNetForSequenceClassification,MPNetForTokenClassification:()=>h.MPNetForTokenClassification,MPNetModel:()=>h.MPNetModel,MPNetPreTrainedModel:()=>h.MPNetPreTrainedModel,MPNetTokenizer:()=>m.MPNetTokenizer,MT5ForConditionalGeneration:()=>h.MT5ForConditionalGeneration,MT5Model:()=>h.MT5Model,MT5PreTrainedModel:()=>h.MT5PreTrainedModel,MarianMTModel:()=>h.MarianMTModel,MarianModel:()=>h.MarianModel,MarianPreTrainedModel:()=>h.MarianPreTrainedModel,MarianTokenizer:()=>m.MarianTokenizer,MaskFormerFeatureExtractor:()=>_.MaskFormerFeatureExtractor,MaskFormerForInstanceSegmentation:()=>h.MaskFormerForInstanceSegmentation,MaskFormerModel:()=>h.MaskFormerModel,MaskFormerPreTrainedModel:()=>h.MaskFormerPreTrainedModel,MaskedLMOutput:()=>h.MaskedLMOutput,MaxLengthCriteria:()=>x.MaxLengthCriteria,MistralForCausalLM:()=>h.MistralForCausalLM,MistralModel:()=>h.MistralModel,MistralPreTrainedModel:()=>h.MistralPreTrainedModel,MobileBertForMaskedLM:()=>h.MobileBertForMaskedLM,MobileBertForQuestionAnswering:()=>h.MobileBertForQuestionAnswering,MobileBertForSequenceClassification:()=>h.MobileBertForSequenceClassification,MobileBertModel:()=>h.MobileBertModel,MobileBertPreTrainedModel:()=>h.MobileBertPreTrainedModel,MobileBertTokenizer:()=>m.MobileBertTokenizer,MobileNetV1FeatureExtractor:()=>_.MobileNetV1FeatureExtractor,MobileNetV1ForImageClassification:()=>h.MobileNetV1ForImageClassification,MobileNetV1Model:()=>h.MobileNetV1Model,MobileNetV1PreTrainedModel:()=>h.MobileNetV1PreTrainedModel,MobileNetV2FeatureExtractor:()=>_.MobileNetV2FeatureExtractor,MobileNetV2ForImageClassification:()=>h.MobileNetV2ForImageClassification,MobileNetV2Model:()=>h.MobileNetV2Model,MobileNetV2PreTrainedModel:()=>h.MobileNetV2PreTrainedModel,MobileNetV3FeatureExtractor:()=>_.MobileNetV3FeatureExtractor,MobileNetV3ForImageClassification:()=>h.MobileNetV3ForImageClassification,MobileNetV3Model:()=>h.MobileNetV3Model,MobileNetV3PreTrainedModel:()=>h.MobileNetV3PreTrainedModel,MobileNetV4FeatureExtractor:()=>_.MobileNetV4FeatureExtractor,MobileNetV4ForImageClassification:()=>h.MobileNetV4ForImageClassification,MobileNetV4Model:()=>h.MobileNetV4Model,MobileNetV4PreTrainedModel:()=>h.MobileNetV4PreTrainedModel,MobileViTFeatureExtractor:()=>_.MobileViTFeatureExtractor,MobileViTForImageClassification:()=>h.MobileViTForImageClassification,MobileViTImageProcessor:()=>_.MobileViTImageProcessor,MobileViTModel:()=>h.MobileViTModel,MobileViTPreTrainedModel:()=>h.MobileViTPreTrainedModel,MobileViTV2ForImageClassification:()=>h.MobileViTV2ForImageClassification,MobileViTV2Model:()=>h.MobileViTV2Model,MobileViTV2PreTrainedModel:()=>h.MobileViTV2PreTrainedModel,ModelOutput:()=>h.ModelOutput,Moondream1ForConditionalGeneration:()=>h.Moondream1ForConditionalGeneration,MptForCausalLM:()=>h.MptForCausalLM,MptModel:()=>h.MptModel,MptPreTrainedModel:()=>h.MptPreTrainedModel,MusicgenForCausalLM:()=>h.MusicgenForCausalLM,MusicgenForConditionalGeneration:()=>h.MusicgenForConditionalGeneration,MusicgenModel:()=>h.MusicgenModel,MusicgenPreTrainedModel:()=>h.MusicgenPreTrainedModel,NllbTokenizer:()=>m.NllbTokenizer,NomicBertModel:()=>h.NomicBertModel,NomicBertPreTrainedModel:()=>h.NomicBertPreTrainedModel,NougatImageProcessor:()=>_.NougatImageProcessor,NougatTokenizer:()=>m.NougatTokenizer,OPTForCausalLM:()=>h.OPTForCausalLM,OPTModel:()=>h.OPTModel,OPTPreTrainedModel:()=>h.OPTPreTrainedModel,ObjectDetectionPipeline:()=>p.ObjectDetectionPipeline,OpenELMForCausalLM:()=>h.OpenELMForCausalLM,OpenELMModel:()=>h.OpenELMModel,OpenELMPreTrainedModel:()=>h.OpenELMPreTrainedModel,OwlViTFeatureExtractor:()=>_.OwlViTFeatureExtractor,OwlViTForObjectDetection:()=>h.OwlViTForObjectDetection,OwlViTModel:()=>h.OwlViTModel,OwlViTPreTrainedModel:()=>h.OwlViTPreTrainedModel,OwlViTProcessor:()=>_.OwlViTProcessor,Owlv2ForObjectDetection:()=>h.Owlv2ForObjectDetection,Owlv2ImageProcessor:()=>_.Owlv2ImageProcessor,Owlv2Model:()=>h.Owlv2Model,Owlv2PreTrainedModel:()=>h.Owlv2PreTrainedModel,Phi3ForCausalLM:()=>h.Phi3ForCausalLM,Phi3Model:()=>h.Phi3Model,Phi3PreTrainedModel:()=>h.Phi3PreTrainedModel,PhiForCausalLM:()=>h.PhiForCausalLM,PhiModel:()=>h.PhiModel,PhiPreTrainedModel:()=>h.PhiPreTrainedModel,Pipeline:()=>p.Pipeline,PreTrainedModel:()=>h.PreTrainedModel,PreTrainedTokenizer:()=>m.PreTrainedTokenizer,PretrainedConfig:()=>f.PretrainedConfig,PretrainedMixin:()=>h.PretrainedMixin,Processor:()=>_.Processor,PvtForImageClassification:()=>h.PvtForImageClassification,PvtImageProcessor:()=>_.PvtImageProcessor,PvtModel:()=>h.PvtModel,PvtPreTrainedModel:()=>h.PvtPreTrainedModel,PyAnnoteFeatureExtractor:()=>_.PyAnnoteFeatureExtractor,PyAnnoteForAudioFrameClassification:()=>h.PyAnnoteForAudioFrameClassification,PyAnnoteModel:()=>h.PyAnnoteModel,PyAnnotePreTrainedModel:()=>h.PyAnnotePreTrainedModel,PyAnnoteProcessor:()=>_.PyAnnoteProcessor,QuestionAnsweringModelOutput:()=>h.QuestionAnsweringModelOutput,QuestionAnsweringPipeline:()=>p.QuestionAnsweringPipeline,Qwen2ForCausalLM:()=>h.Qwen2ForCausalLM,Qwen2Model:()=>h.Qwen2Model,Qwen2PreTrainedModel:()=>h.Qwen2PreTrainedModel,Qwen2Tokenizer:()=>m.Qwen2Tokenizer,RTDetrForObjectDetection:()=>h.RTDetrForObjectDetection,RTDetrImageProcessor:()=>_.RTDetrImageProcessor,RTDetrModel:()=>h.RTDetrModel,RTDetrObjectDetectionOutput:()=>h.RTDetrObjectDetectionOutput,RTDetrPreTrainedModel:()=>h.RTDetrPreTrainedModel,RawImage:()=>M.RawImage,ResNetForImageClassification:()=>h.ResNetForImageClassification,ResNetModel:()=>h.ResNetModel,ResNetPreTrainedModel:()=>h.ResNetPreTrainedModel,RoFormerForMaskedLM:()=>h.RoFormerForMaskedLM,RoFormerForQuestionAnswering:()=>h.RoFormerForQuestionAnswering,RoFormerForSequenceClassification:()=>h.RoFormerForSequenceClassification,RoFormerForTokenClassification:()=>h.RoFormerForTokenClassification,RoFormerModel:()=>h.RoFormerModel,RoFormerPreTrainedModel:()=>h.RoFormerPreTrainedModel,RoFormerTokenizer:()=>m.RoFormerTokenizer,RobertaForMaskedLM:()=>h.RobertaForMaskedLM,RobertaForQuestionAnswering:()=>h.RobertaForQuestionAnswering,RobertaForSequenceClassification:()=>h.RobertaForSequenceClassification,RobertaForTokenClassification:()=>h.RobertaForTokenClassification,RobertaModel:()=>h.RobertaModel,RobertaPreTrainedModel:()=>h.RobertaPreTrainedModel,RobertaTokenizer:()=>m.RobertaTokenizer,SamImageProcessor:()=>_.SamImageProcessor,SamImageSegmentationOutput:()=>h.SamImageSegmentationOutput,SamModel:()=>h.SamModel,SamPreTrainedModel:()=>h.SamPreTrainedModel,SamProcessor:()=>_.SamProcessor,SapiensFeatureExtractor:()=>_.SapiensFeatureExtractor,SapiensForDepthEstimation:()=>h.SapiensForDepthEstimation,SapiensForNormalEstimation:()=>h.SapiensForNormalEstimation,SapiensForSemanticSegmentation:()=>h.SapiensForSemanticSegmentation,SapiensPreTrainedModel:()=>h.SapiensPreTrainedModel,SeamlessM4TFeatureExtractor:()=>_.SeamlessM4TFeatureExtractor,SegformerFeatureExtractor:()=>_.SegformerFeatureExtractor,SegformerForImageClassification:()=>h.SegformerForImageClassification,SegformerForSemanticSegmentation:()=>h.SegformerForSemanticSegmentation,SegformerModel:()=>h.SegformerModel,SegformerPreTrainedModel:()=>h.SegformerPreTrainedModel,Seq2SeqLMOutput:()=>h.Seq2SeqLMOutput,SequenceClassifierOutput:()=>h.SequenceClassifierOutput,SiglipImageProcessor:()=>_.SiglipImageProcessor,SiglipModel:()=>h.SiglipModel,SiglipPreTrainedModel:()=>h.SiglipPreTrainedModel,SiglipTextModel:()=>h.SiglipTextModel,SiglipTokenizer:()=>m.SiglipTokenizer,SiglipVisionModel:()=>h.SiglipVisionModel,SpeechT5FeatureExtractor:()=>_.SpeechT5FeatureExtractor,SpeechT5ForSpeechToText:()=>h.SpeechT5ForSpeechToText,SpeechT5ForTextToSpeech:()=>h.SpeechT5ForTextToSpeech,SpeechT5HifiGan:()=>h.SpeechT5HifiGan,SpeechT5Model:()=>h.SpeechT5Model,SpeechT5PreTrainedModel:()=>h.SpeechT5PreTrainedModel,SpeechT5Processor:()=>_.SpeechT5Processor,SpeechT5Tokenizer:()=>m.SpeechT5Tokenizer,SqueezeBertForMaskedLM:()=>h.SqueezeBertForMaskedLM,SqueezeBertForQuestionAnswering:()=>h.SqueezeBertForQuestionAnswering,SqueezeBertForSequenceClassification:()=>h.SqueezeBertForSequenceClassification,SqueezeBertModel:()=>h.SqueezeBertModel,SqueezeBertPreTrainedModel:()=>h.SqueezeBertPreTrainedModel,SqueezeBertTokenizer:()=>m.SqueezeBertTokenizer,StableLmForCausalLM:()=>h.StableLmForCausalLM,StableLmModel:()=>h.StableLmModel,StableLmPreTrainedModel:()=>h.StableLmPreTrainedModel,Starcoder2ForCausalLM:()=>h.Starcoder2ForCausalLM,Starcoder2Model:()=>h.Starcoder2Model,Starcoder2PreTrainedModel:()=>h.Starcoder2PreTrainedModel,StoppingCriteria:()=>x.StoppingCriteria,StoppingCriteriaList:()=>x.StoppingCriteriaList,SummarizationPipeline:()=>p.SummarizationPipeline,Swin2SRForImageSuperResolution:()=>h.Swin2SRForImageSuperResolution,Swin2SRImageProcessor:()=>_.Swin2SRImageProcessor,Swin2SRModel:()=>h.Swin2SRModel,Swin2SRPreTrainedModel:()=>h.Swin2SRPreTrainedModel,SwinForImageClassification:()=>h.SwinForImageClassification,SwinModel:()=>h.SwinModel,SwinPreTrainedModel:()=>h.SwinPreTrainedModel,T5ForConditionalGeneration:()=>h.T5ForConditionalGeneration,T5Model:()=>h.T5Model,T5PreTrainedModel:()=>h.T5PreTrainedModel,T5Tokenizer:()=>m.T5Tokenizer,TableTransformerForObjectDetection:()=>h.TableTransformerForObjectDetection,TableTransformerModel:()=>h.TableTransformerModel,TableTransformerObjectDetectionOutput:()=>h.TableTransformerObjectDetectionOutput,TableTransformerPreTrainedModel:()=>h.TableTransformerPreTrainedModel,Tensor:()=>w.Tensor,Text2TextGenerationPipeline:()=>p.Text2TextGenerationPipeline,TextClassificationPipeline:()=>p.TextClassificationPipeline,TextGenerationPipeline:()=>p.TextGenerationPipeline,TextStreamer:()=>b.TextStreamer,TextToAudioPipeline:()=>p.TextToAudioPipeline,TokenClassificationPipeline:()=>p.TokenClassificationPipeline,TokenClassifierOutput:()=>h.TokenClassifierOutput,TokenizerModel:()=>m.TokenizerModel,TrOCRForCausalLM:()=>h.TrOCRForCausalLM,TrOCRPreTrainedModel:()=>h.TrOCRPreTrainedModel,TranslationPipeline:()=>p.TranslationPipeline,UniSpeechForCTC:()=>h.UniSpeechForCTC,UniSpeechForSequenceClassification:()=>h.UniSpeechForSequenceClassification,UniSpeechModel:()=>h.UniSpeechModel,UniSpeechPreTrainedModel:()=>h.UniSpeechPreTrainedModel,UniSpeechSatForAudioFrameClassification:()=>h.UniSpeechSatForAudioFrameClassification,UniSpeechSatForCTC:()=>h.UniSpeechSatForCTC,UniSpeechSatForSequenceClassification:()=>h.UniSpeechSatForSequenceClassification,UniSpeechSatModel:()=>h.UniSpeechSatModel,UniSpeechSatPreTrainedModel:()=>h.UniSpeechSatPreTrainedModel,ViTFeatureExtractor:()=>_.ViTFeatureExtractor,ViTForImageClassification:()=>h.ViTForImageClassification,ViTImageProcessor:()=>_.ViTImageProcessor,ViTMAEModel:()=>h.ViTMAEModel,ViTMAEPreTrainedModel:()=>h.ViTMAEPreTrainedModel,ViTMSNForImageClassification:()=>h.ViTMSNForImageClassification,ViTMSNModel:()=>h.ViTMSNModel,ViTMSNPreTrainedModel:()=>h.ViTMSNPreTrainedModel,ViTModel:()=>h.ViTModel,ViTPreTrainedModel:()=>h.ViTPreTrainedModel,VisionEncoderDecoderModel:()=>h.VisionEncoderDecoderModel,VitMatteForImageMatting:()=>h.VitMatteForImageMatting,VitMatteImageProcessor:()=>_.VitMatteImageProcessor,VitMattePreTrainedModel:()=>h.VitMattePreTrainedModel,VitsModel:()=>h.VitsModel,VitsModelOutput:()=>h.VitsModelOutput,VitsPreTrainedModel:()=>h.VitsPreTrainedModel,VitsTokenizer:()=>m.VitsTokenizer,Wav2Vec2BertForCTC:()=>h.Wav2Vec2BertForCTC,Wav2Vec2BertForSequenceClassification:()=>h.Wav2Vec2BertForSequenceClassification,Wav2Vec2BertModel:()=>h.Wav2Vec2BertModel,Wav2Vec2BertPreTrainedModel:()=>h.Wav2Vec2BertPreTrainedModel,Wav2Vec2CTCTokenizer:()=>m.Wav2Vec2CTCTokenizer,Wav2Vec2FeatureExtractor:()=>_.Wav2Vec2FeatureExtractor,Wav2Vec2ForAudioFrameClassification:()=>h.Wav2Vec2ForAudioFrameClassification,Wav2Vec2ForCTC:()=>h.Wav2Vec2ForCTC,Wav2Vec2ForSequenceClassification:()=>h.Wav2Vec2ForSequenceClassification,Wav2Vec2Model:()=>h.Wav2Vec2Model,Wav2Vec2PreTrainedModel:()=>h.Wav2Vec2PreTrainedModel,Wav2Vec2ProcessorWithLM:()=>_.Wav2Vec2ProcessorWithLM,WavLMForAudioFrameClassification:()=>h.WavLMForAudioFrameClassification,WavLMForCTC:()=>h.WavLMForCTC,WavLMForSequenceClassification:()=>h.WavLMForSequenceClassification,WavLMForXVector:()=>h.WavLMForXVector,WavLMModel:()=>h.WavLMModel,WavLMPreTrainedModel:()=>h.WavLMPreTrainedModel,WeSpeakerFeatureExtractor:()=>_.WeSpeakerFeatureExtractor,WeSpeakerResNetModel:()=>h.WeSpeakerResNetModel,WeSpeakerResNetPreTrainedModel:()=>h.WeSpeakerResNetPreTrainedModel,WhisperFeatureExtractor:()=>_.WhisperFeatureExtractor,WhisperForConditionalGeneration:()=>h.WhisperForConditionalGeneration,WhisperModel:()=>h.WhisperModel,WhisperPreTrainedModel:()=>h.WhisperPreTrainedModel,WhisperProcessor:()=>_.WhisperProcessor,WhisperTextStreamer:()=>b.WhisperTextStreamer,WhisperTokenizer:()=>m.WhisperTokenizer,XLMForQuestionAnswering:()=>h.XLMForQuestionAnswering,XLMForSequenceClassification:()=>h.XLMForSequenceClassification,XLMForTokenClassification:()=>h.XLMForTokenClassification,XLMModel:()=>h.XLMModel,XLMPreTrainedModel:()=>h.XLMPreTrainedModel,XLMRobertaForMaskedLM:()=>h.XLMRobertaForMaskedLM,XLMRobertaForQuestionAnswering:()=>h.XLMRobertaForQuestionAnswering,XLMRobertaForSequenceClassification:()=>h.XLMRobertaForSequenceClassification,XLMRobertaForTokenClassification:()=>h.XLMRobertaForTokenClassification,XLMRobertaModel:()=>h.XLMRobertaModel,XLMRobertaPreTrainedModel:()=>h.XLMRobertaPreTrainedModel,XLMRobertaTokenizer:()=>m.XLMRobertaTokenizer,XLMTokenizer:()=>m.XLMTokenizer,XLMWithLMHeadModel:()=>h.XLMWithLMHeadModel,XVectorOutput:()=>h.XVectorOutput,YolosFeatureExtractor:()=>_.YolosFeatureExtractor,YolosForObjectDetection:()=>h.YolosForObjectDetection,YolosModel:()=>h.YolosModel,YolosObjectDetectionOutput:()=>h.YolosObjectDetectionOutput,YolosPreTrainedModel:()=>h.YolosPreTrainedModel,ZeroShotAudioClassificationPipeline:()=>p.ZeroShotAudioClassificationPipeline,ZeroShotClassificationPipeline:()=>p.ZeroShotClassificationPipeline,ZeroShotImageClassificationPipeline:()=>p.ZeroShotImageClassificationPipeline,ZeroShotObjectDetectionPipeline:()=>p.ZeroShotObjectDetectionPipeline,bankers_round:()=>T.bankers_round,cat:()=>w.cat,cos_sim:()=>T.cos_sim,dot:()=>T.dot,dynamic_time_warping:()=>T.dynamic_time_warping,env:()=>u.env,full:()=>w.full,full_like:()=>w.full_like,getKeyValueShapes:()=>f.getKeyValueShapes,hamming:()=>g.hamming,hanning:()=>g.hanning,interpolate:()=>w.interpolate,interpolate_4d:()=>w.interpolate_4d,interpolate_data:()=>T.interpolate_data,is_chinese_char:()=>m.is_chinese_char,layer_norm:()=>w.layer_norm,log_softmax:()=>T.log_softmax,magnitude:()=>T.magnitude,matmul:()=>w.matmul,max:()=>T.max,mean:()=>w.mean,mean_pooling:()=>w.mean_pooling,medianFilter:()=>T.medianFilter,mel_filter_bank:()=>g.mel_filter_bank,min:()=>T.min,ones:()=>w.ones,ones_like:()=>w.ones_like,permute:()=>w.permute,permute_data:()=>T.permute_data,pipeline:()=>p.pipeline,quantize_embeddings:()=>w.quantize_embeddings,read_audio:()=>g.read_audio,rfft:()=>w.rfft,round:()=>T.round,softmax:()=>T.softmax,spectrogram:()=>g.spectrogram,stack:()=>w.stack,std_mean:()=>w.std_mean,topk:()=>w.topk,window_function:()=>g.window_function,zeros:()=>w.zeros,zeros_like:()=>w.zeros_like});var u=c(/*! ./env.js */"./src/env.js"),p=c(/*! ./pipelines.js */"./src/pipelines.js"),h=c(/*! ./models.js */"./src/models.js"),m=c(/*! ./tokenizers.js */"./src/tokenizers.js"),_=c(/*! ./processors.js */"./src/processors.js"),f=c(/*! ./configs.js */"./src/configs.js"),g=c(/*! ./utils/audio.js */"./src/utils/audio.js"),M=c(/*! ./utils/image.js */"./src/utils/image.js"),w=c(/*! ./utils/tensor.js */"./src/utils/tensor.js"),T=c(/*! ./utils/maths.js */"./src/utils/maths.js"),b=c(/*! ./generation/streamers.js */"./src/generation/streamers.js"),x=c(/*! ./generation/stopping_criteria.js */"./src/generation/stopping_criteria.js"),y=d.ASTFeatureExtractor,k=d.ASTForAudioClassification,F=d.ASTModel,P=d.ASTPreTrainedModel,C=d.AlbertForMaskedLM,v=d.AlbertForQuestionAnswering,S=d.AlbertForSequenceClassification,A=d.AlbertModel,E=d.AlbertPreTrainedModel,L=d.AlbertTokenizer,z=d.AudioClassificationPipeline,I=d.AutoConfig,B=d.AutoModel,N=d.AutoModelForAudioClassification,D=d.AutoModelForAudioFrameClassification,O=d.AutoModelForCTC,V=d.AutoModelForCausalLM,j=d.AutoModelForDepthEstimation,R=d.AutoModelForDocumentQuestionAnswering,G=d.AutoModelForImageClassification,q=d.AutoModelForImageFeatureExtraction,W=d.AutoModelForImageMatting,$=d.AutoModelForImageSegmentation,U=d.AutoModelForImageToImage,X=d.AutoModelForMaskGeneration,Q=d.AutoModelForMaskedLM,H=d.AutoModelForNormalEstimation,Y=d.AutoModelForObjectDetection,J=d.AutoModelForQuestionAnswering,K=d.AutoModelForSemanticSegmentation,Z=d.AutoModelForSeq2SeqLM,ee=d.AutoModelForSequenceClassification,te=d.AutoModelForSpeechSeq2Seq,re=d.AutoModelForTextToSpectrogram,ne=d.AutoModelForTextToWaveform,oe=d.AutoModelForTokenClassification,se=d.AutoModelForUniversalSegmentation,ae=d.AutoModelForVision2Seq,ie=d.AutoModelForXVector,le=d.AutoModelForZeroShotObjectDetection,ce=d.AutoProcessor,de=d.AutoTokenizer,ue=d.AutomaticSpeechRecognitionPipeline,pe=d.BartForConditionalGeneration,he=d.BartForSequenceClassification,me=d.BartModel,_e=d.BartPretrainedModel,fe=d.BartTokenizer,ge=d.BaseModelOutput,Me=d.BaseStreamer,we=d.BeitFeatureExtractor,Te=d.BeitForImageClassification,be=d.BeitModel,xe=d.BeitPreTrainedModel,ye=d.BertForMaskedLM,ke=d.BertForQuestionAnswering,Fe=d.BertForSequenceClassification,Pe=d.BertForTokenClassification,Ce=d.BertModel,ve=d.BertPreTrainedModel,Se=d.BertTokenizer,Ae=d.BitImageProcessor,Ee=d.BlenderbotForConditionalGeneration,Le=d.BlenderbotModel,ze=d.BlenderbotPreTrainedModel,Ie=d.BlenderbotSmallForConditionalGeneration,Be=d.BlenderbotSmallModel,Ne=d.BlenderbotSmallPreTrainedModel,De=d.BlenderbotSmallTokenizer,Oe=d.BlenderbotTokenizer,Ve=d.BloomForCausalLM,je=d.BloomModel,Re=d.BloomPreTrainedModel,Ge=d.BloomTokenizer,qe=d.CLIPFeatureExtractor,We=d.CLIPImageProcessor,$e=d.CLIPModel,Ue=d.CLIPPreTrainedModel,Xe=d.CLIPSegForImageSegmentation,Qe=d.CLIPSegModel,He=d.CLIPSegPreTrainedModel,Ye=d.CLIPTextModel,Je=d.CLIPTextModelWithProjection,Ke=d.CLIPTokenizer,Ze=d.CLIPVisionModel,et=d.CLIPVisionModelWithProjection,tt=d.CamembertForMaskedLM,rt=d.CamembertForQuestionAnswering,nt=d.CamembertForSequenceClassification,ot=d.CamembertForTokenClassification,st=d.CamembertModel,at=d.CamembertPreTrainedModel,it=d.CamembertTokenizer,lt=d.CausalLMOutput,ct=d.CausalLMOutputWithPast,dt=d.ChineseCLIPFeatureExtractor,ut=d.ChineseCLIPModel,pt=d.ChineseCLIPPreTrainedModel,ht=d.ClapAudioModelWithProjection,mt=d.ClapFeatureExtractor,_t=d.ClapModel,ft=d.ClapPreTrainedModel,gt=d.ClapTextModelWithProjection,Mt=d.CodeGenForCausalLM,wt=d.CodeGenModel,Tt=d.CodeGenPreTrainedModel,bt=d.CodeGenTokenizer,xt=d.CodeLlamaTokenizer,yt=d.CohereForCausalLM,kt=d.CohereModel,Ft=d.CoherePreTrainedModel,Pt=d.CohereTokenizer,Ct=d.ConvBertForMaskedLM,vt=d.ConvBertForQuestionAnswering,St=d.ConvBertForSequenceClassification,At=d.ConvBertForTokenClassification,Et=d.ConvBertModel,Lt=d.ConvBertPreTrainedModel,zt=d.ConvBertTokenizer,It=d.ConvNextFeatureExtractor,Bt=d.ConvNextForImageClassification,Nt=d.ConvNextImageProcessor,Dt=d.ConvNextModel,Ot=d.ConvNextPreTrainedModel,Vt=d.ConvNextV2ForImageClassification,jt=d.ConvNextV2Model,Rt=d.ConvNextV2PreTrainedModel,Gt=d.DPTFeatureExtractor,qt=d.DPTForDepthEstimation,Wt=d.DPTImageProcessor,$t=d.DPTModel,Ut=d.DPTPreTrainedModel,Xt=d.DebertaForMaskedLM,Qt=d.DebertaForQuestionAnswering,Ht=d.DebertaForSequenceClassification,Yt=d.DebertaForTokenClassification,Jt=d.DebertaModel,Kt=d.DebertaPreTrainedModel,Zt=d.DebertaTokenizer,er=d.DebertaV2ForMaskedLM,tr=d.DebertaV2ForQuestionAnswering,rr=d.DebertaV2ForSequenceClassification,nr=d.DebertaV2ForTokenClassification,or=d.DebertaV2Model,sr=d.DebertaV2PreTrainedModel,ar=d.DebertaV2Tokenizer,ir=d.DecisionTransformerModel,lr=d.DecisionTransformerPreTrainedModel,cr=d.DeiTFeatureExtractor,dr=d.DeiTForImageClassification,ur=d.DeiTModel,pr=d.DeiTPreTrainedModel,hr=d.DepthAnythingForDepthEstimation,mr=d.DepthAnythingPreTrainedModel,_r=d.DepthEstimationPipeline,fr=d.DepthProForDepthEstimation,gr=d.DepthProPreTrainedModel,Mr=d.DetrFeatureExtractor,wr=d.DetrForObjectDetection,Tr=d.DetrForSegmentation,br=d.DetrModel,xr=d.DetrObjectDetectionOutput,yr=d.DetrPreTrainedModel,kr=d.DetrSegmentationOutput,Fr=d.Dinov2ForImageClassification,Pr=d.Dinov2Model,Cr=d.Dinov2PreTrainedModel,vr=d.DistilBertForMaskedLM,Sr=d.DistilBertForQuestionAnswering,Ar=d.DistilBertForSequenceClassification,Er=d.DistilBertForTokenClassification,Lr=d.DistilBertModel,zr=d.DistilBertPreTrainedModel,Ir=d.DistilBertTokenizer,Br=d.DocumentQuestionAnsweringPipeline,Nr=d.DonutFeatureExtractor,Dr=d.DonutSwinModel,Or=d.DonutSwinPreTrainedModel,Vr=d.EfficientNetForImageClassification,jr=d.EfficientNetImageProcessor,Rr=d.EfficientNetModel,Gr=d.EfficientNetPreTrainedModel,qr=d.ElectraForMaskedLM,Wr=d.ElectraForQuestionAnswering,$r=d.ElectraForSequenceClassification,Ur=d.ElectraForTokenClassification,Xr=d.ElectraModel,Qr=d.ElectraPreTrainedModel,Hr=d.ElectraTokenizer,Yr=d.EosTokenCriteria,Jr=d.EsmForMaskedLM,Kr=d.EsmForSequenceClassification,Zr=d.EsmForTokenClassification,en=d.EsmModel,tn=d.EsmPreTrainedModel,rn=d.EsmTokenizer,nn=d.FFT,on=d.FalconForCausalLM,sn=d.FalconModel,an=d.FalconPreTrainedModel,ln=d.FalconTokenizer,cn=d.FastViTForImageClassification,dn=d.FastViTModel,un=d.FastViTPreTrainedModel,pn=d.FeatureExtractionPipeline,hn=d.FeatureExtractor,mn=d.FillMaskPipeline,_n=d.Florence2ForConditionalGeneration,fn=d.Florence2PreTrainedModel,gn=d.Florence2Processor,Mn=d.GLPNFeatureExtractor,wn=d.GLPNForDepthEstimation,Tn=d.GLPNModel,bn=d.GLPNPreTrainedModel,xn=d.GPT2LMHeadModel,yn=d.GPT2Model,kn=d.GPT2PreTrainedModel,Fn=d.GPT2Tokenizer,Pn=d.GPTBigCodeForCausalLM,Cn=d.GPTBigCodeModel,vn=d.GPTBigCodePreTrainedModel,Sn=d.GPTJForCausalLM,An=d.GPTJModel,En=d.GPTJPreTrainedModel,Ln=d.GPTNeoForCausalLM,zn=d.GPTNeoModel,In=d.GPTNeoPreTrainedModel,Bn=d.GPTNeoXForCausalLM,Nn=d.GPTNeoXModel,Dn=d.GPTNeoXPreTrainedModel,On=d.GPTNeoXTokenizer,Vn=d.Gemma2ForCausalLM,jn=d.Gemma2Model,Rn=d.Gemma2PreTrainedModel,Gn=d.GemmaForCausalLM,qn=d.GemmaModel,Wn=d.GemmaPreTrainedModel,$n=d.GemmaTokenizer,Un=d.GraniteForCausalLM,Xn=d.GraniteModel,Qn=d.GranitePreTrainedModel,Hn=d.Grok1Tokenizer,Yn=d.GroupViTModel,Jn=d.GroupViTPreTrainedModel,Kn=d.HerbertTokenizer,Zn=d.HieraForImageClassification,eo=d.HieraModel,to=d.HieraPreTrainedModel,ro=d.HubertForCTC,no=d.HubertForSequenceClassification,oo=d.HubertModel,so=d.HubertPreTrainedModel,ao=d.ImageClassificationPipeline,io=d.ImageFeatureExtractionPipeline,lo=d.ImageFeatureExtractor,co=d.ImageMattingOutput,uo=d.ImageSegmentationPipeline,po=d.ImageToImagePipeline,ho=d.ImageToTextPipeline,mo=d.InterruptableStoppingCriteria,_o=d.JAISLMHeadModel,fo=d.JAISModel,go=d.JAISPreTrainedModel,Mo=d.LlamaForCausalLM,wo=d.LlamaModel,To=d.LlamaPreTrainedModel,bo=d.LlamaTokenizer,xo=d.LlavaForConditionalGeneration,yo=d.LlavaPreTrainedModel,ko=d.LongT5ForConditionalGeneration,Fo=d.LongT5Model,Po=d.LongT5PreTrainedModel,Co=d.M2M100ForConditionalGeneration,vo=d.M2M100Model,So=d.M2M100PreTrainedModel,Ao=d.M2M100Tokenizer,Eo=d.MBart50Tokenizer,Lo=d.MBartForCausalLM,zo=d.MBartForConditionalGeneration,Io=d.MBartForSequenceClassification,Bo=d.MBartModel,No=d.MBartPreTrainedModel,Do=d.MBartTokenizer,Oo=d.MPNetForMaskedLM,Vo=d.MPNetForQuestionAnswering,jo=d.MPNetForSequenceClassification,Ro=d.MPNetForTokenClassification,Go=d.MPNetModel,qo=d.MPNetPreTrainedModel,Wo=d.MPNetTokenizer,$o=d.MT5ForConditionalGeneration,Uo=d.MT5Model,Xo=d.MT5PreTrainedModel,Qo=d.MarianMTModel,Ho=d.MarianModel,Yo=d.MarianPreTrainedModel,Jo=d.MarianTokenizer,Ko=d.MaskFormerFeatureExtractor,Zo=d.MaskFormerForInstanceSegmentation,es=d.MaskFormerModel,ts=d.MaskFormerPreTrainedModel,rs=d.MaskedLMOutput,ns=d.MaxLengthCriteria,os=d.MistralForCausalLM,ss=d.MistralModel,as=d.MistralPreTrainedModel,is=d.MobileBertForMaskedLM,ls=d.MobileBertForQuestionAnswering,cs=d.MobileBertForSequenceClassification,ds=d.MobileBertModel,us=d.MobileBertPreTrainedModel,ps=d.MobileBertTokenizer,hs=d.MobileNetV1FeatureExtractor,ms=d.MobileNetV1ForImageClassification,_s=d.MobileNetV1Model,fs=d.MobileNetV1PreTrainedModel,gs=d.MobileNetV2FeatureExtractor,Ms=d.MobileNetV2ForImageClassification,ws=d.MobileNetV2Model,Ts=d.MobileNetV2PreTrainedModel,bs=d.MobileNetV3FeatureExtractor,xs=d.MobileNetV3ForImageClassification,ys=d.MobileNetV3Model,ks=d.MobileNetV3PreTrainedModel,Fs=d.MobileNetV4FeatureExtractor,Ps=d.MobileNetV4ForImageClassification,Cs=d.MobileNetV4Model,vs=d.MobileNetV4PreTrainedModel,Ss=d.MobileViTFeatureExtractor,As=d.MobileViTForImageClassification,Es=d.MobileViTImageProcessor,Ls=d.MobileViTModel,zs=d.MobileViTPreTrainedModel,Is=d.MobileViTV2ForImageClassification,Bs=d.MobileViTV2Model,Ns=d.MobileViTV2PreTrainedModel,Ds=d.ModelOutput,Os=d.Moondream1ForConditionalGeneration,Vs=d.MptForCausalLM,js=d.MptModel,Rs=d.MptPreTrainedModel,Gs=d.MusicgenForCausalLM,qs=d.MusicgenForConditionalGeneration,Ws=d.MusicgenModel,$s=d.MusicgenPreTrainedModel,Us=d.NllbTokenizer,Xs=d.NomicBertModel,Qs=d.NomicBertPreTrainedModel,Hs=d.NougatImageProcessor,Ys=d.NougatTokenizer,Js=d.OPTForCausalLM,Ks=d.OPTModel,Zs=d.OPTPreTrainedModel,ea=d.ObjectDetectionPipeline,ta=d.OpenELMForCausalLM,ra=d.OpenELMModel,na=d.OpenELMPreTrainedModel,oa=d.OwlViTFeatureExtractor,sa=d.OwlViTForObjectDetection,aa=d.OwlViTModel,ia=d.OwlViTPreTrainedModel,la=d.OwlViTProcessor,ca=d.Owlv2ForObjectDetection,da=d.Owlv2ImageProcessor,ua=d.Owlv2Model,pa=d.Owlv2PreTrainedModel,ha=d.Phi3ForCausalLM,ma=d.Phi3Model,_a=d.Phi3PreTrainedModel,fa=d.PhiForCausalLM,ga=d.PhiModel,Ma=d.PhiPreTrainedModel,wa=d.Pipeline,Ta=d.PreTrainedModel,ba=d.PreTrainedTokenizer,xa=d.PretrainedConfig,ya=d.PretrainedMixin,ka=d.Processor,Fa=d.PvtForImageClassification,Pa=d.PvtImageProcessor,Ca=d.PvtModel,va=d.PvtPreTrainedModel,Sa=d.PyAnnoteFeatureExtractor,Aa=d.PyAnnoteForAudioFrameClassification,Ea=d.PyAnnoteModel,La=d.PyAnnotePreTrainedModel,za=d.PyAnnoteProcessor,Ia=d.QuestionAnsweringModelOutput,Ba=d.QuestionAnsweringPipeline,Na=d.Qwen2ForCausalLM,Da=d.Qwen2Model,Oa=d.Qwen2PreTrainedModel,Va=d.Qwen2Tokenizer,ja=d.RTDetrForObjectDetection,Ra=d.RTDetrImageProcessor,Ga=d.RTDetrModel,qa=d.RTDetrObjectDetectionOutput,Wa=d.RTDetrPreTrainedModel,$a=d.RawImage,Ua=d.ResNetForImageClassification,Xa=d.ResNetModel,Qa=d.ResNetPreTrainedModel,Ha=d.RoFormerForMaskedLM,Ya=d.RoFormerForQuestionAnswering,Ja=d.RoFormerForSequenceClassification,Ka=d.RoFormerForTokenClassification,Za=d.RoFormerModel,ei=d.RoFormerPreTrainedModel,ti=d.RoFormerTokenizer,ri=d.RobertaForMaskedLM,ni=d.RobertaForQuestionAnswering,oi=d.RobertaForSequenceClassification,si=d.RobertaForTokenClassification,ai=d.RobertaModel,ii=d.RobertaPreTrainedModel,li=d.RobertaTokenizer,ci=d.SamImageProcessor,di=d.SamImageSegmentationOutput,ui=d.SamModel,pi=d.SamPreTrainedModel,hi=d.SamProcessor,mi=d.SapiensFeatureExtractor,_i=d.SapiensForDepthEstimation,fi=d.SapiensForNormalEstimation,gi=d.SapiensForSemanticSegmentation,Mi=d.SapiensPreTrainedModel,wi=d.SeamlessM4TFeatureExtractor,Ti=d.SegformerFeatureExtractor,bi=d.SegformerForImageClassification,xi=d.SegformerForSemanticSegmentation,yi=d.SegformerModel,ki=d.SegformerPreTrainedModel,Fi=d.Seq2SeqLMOutput,Pi=d.SequenceClassifierOutput,Ci=d.SiglipImageProcessor,vi=d.SiglipModel,Si=d.SiglipPreTrainedModel,Ai=d.SiglipTextModel,Ei=d.SiglipTokenizer,Li=d.SiglipVisionModel,zi=d.SpeechT5FeatureExtractor,Ii=d.SpeechT5ForSpeechToText,Bi=d.SpeechT5ForTextToSpeech,Ni=d.SpeechT5HifiGan,Di=d.SpeechT5Model,Oi=d.SpeechT5PreTrainedModel,Vi=d.SpeechT5Processor,ji=d.SpeechT5Tokenizer,Ri=d.SqueezeBertForMaskedLM,Gi=d.SqueezeBertForQuestionAnswering,qi=d.SqueezeBertForSequenceClassification,Wi=d.SqueezeBertModel,$i=d.SqueezeBertPreTrainedModel,Ui=d.SqueezeBertTokenizer,Xi=d.StableLmForCausalLM,Qi=d.StableLmModel,Hi=d.StableLmPreTrainedModel,Yi=d.Starcoder2ForCausalLM,Ji=d.Starcoder2Model,Ki=d.Starcoder2PreTrainedModel,Zi=d.StoppingCriteria,el=d.StoppingCriteriaList,tl=d.SummarizationPipeline,rl=d.Swin2SRForImageSuperResolution,nl=d.Swin2SRImageProcessor,ol=d.Swin2SRModel,sl=d.Swin2SRPreTrainedModel,al=d.SwinForImageClassification,il=d.SwinModel,ll=d.SwinPreTrainedModel,cl=d.T5ForConditionalGeneration,dl=d.T5Model,ul=d.T5PreTrainedModel,pl=d.T5Tokenizer,hl=d.TableTransformerForObjectDetection,ml=d.TableTransformerModel,_l=d.TableTransformerObjectDetectionOutput,fl=d.TableTransformerPreTrainedModel,gl=d.Tensor,Ml=d.Text2TextGenerationPipeline,wl=d.TextClassificationPipeline,Tl=d.TextGenerationPipeline,bl=d.TextStreamer,xl=d.TextToAudioPipeline,yl=d.TokenClassificationPipeline,kl=d.TokenClassifierOutput,Fl=d.TokenizerModel,Pl=d.TrOCRForCausalLM,Cl=d.TrOCRPreTrainedModel,vl=d.TranslationPipeline,Sl=d.UniSpeechForCTC,Al=d.UniSpeechForSequenceClassification,El=d.UniSpeechModel,Ll=d.UniSpeechPreTrainedModel,zl=d.UniSpeechSatForAudioFrameClassification,Il=d.UniSpeechSatForCTC,Bl=d.UniSpeechSatForSequenceClassification,Nl=d.UniSpeechSatModel,Dl=d.UniSpeechSatPreTrainedModel,Ol=d.ViTFeatureExtractor,Vl=d.ViTForImageClassification,jl=d.ViTImageProcessor,Rl=d.ViTMAEModel,Gl=d.ViTMAEPreTrainedModel,ql=d.ViTMSNForImageClassification,Wl=d.ViTMSNModel,$l=d.ViTMSNPreTrainedModel,Ul=d.ViTModel,Xl=d.ViTPreTrainedModel,Ql=d.VisionEncoderDecoderModel,Hl=d.VitMatteForImageMatting,Yl=d.VitMatteImageProcessor,Jl=d.VitMattePreTrainedModel,Kl=d.VitsModel,Zl=d.VitsModelOutput,ec=d.VitsPreTrainedModel,tc=d.VitsTokenizer,rc=d.Wav2Vec2BertForCTC,nc=d.Wav2Vec2BertForSequenceClassification,oc=d.Wav2Vec2BertModel,sc=d.Wav2Vec2BertPreTrainedModel,ac=d.Wav2Vec2CTCTokenizer,ic=d.Wav2Vec2FeatureExtractor,lc=d.Wav2Vec2ForAudioFrameClassification,cc=d.Wav2Vec2ForCTC,dc=d.Wav2Vec2ForSequenceClassification,uc=d.Wav2Vec2Model,pc=d.Wav2Vec2PreTrainedModel,hc=d.Wav2Vec2ProcessorWithLM,mc=d.WavLMForAudioFrameClassification,_c=d.WavLMForCTC,fc=d.WavLMForSequenceClassification,gc=d.WavLMForXVector,Mc=d.WavLMModel,wc=d.WavLMPreTrainedModel,Tc=d.WeSpeakerFeatureExtractor,bc=d.WeSpeakerResNetModel,xc=d.WeSpeakerResNetPreTrainedModel,yc=d.WhisperFeatureExtractor,kc=d.WhisperForConditionalGeneration,Fc=d.WhisperModel,Pc=d.WhisperPreTrainedModel,Cc=d.WhisperProcessor,vc=d.WhisperTextStreamer,Sc=d.WhisperTokenizer,Ac=d.XLMForQuestionAnswering,Ec=d.XLMForSequenceClassification,Lc=d.XLMForTokenClassification,zc=d.XLMModel,Ic=d.XLMPreTrainedModel,Bc=d.XLMRobertaForMaskedLM,Nc=d.XLMRobertaForQuestionAnswering,Dc=d.XLMRobertaForSequenceClassification,Oc=d.XLMRobertaForTokenClassification,Vc=d.XLMRobertaModel,jc=d.XLMRobertaPreTrainedModel,Rc=d.XLMRobertaTokenizer,Gc=d.XLMTokenizer,qc=d.XLMWithLMHeadModel,Wc=d.XVectorOutput,$c=d.YolosFeatureExtractor,Uc=d.YolosForObjectDetection,Xc=d.YolosModel,Qc=d.YolosObjectDetectionOutput,Hc=d.YolosPreTrainedModel,Yc=d.ZeroShotAudioClassificationPipeline,Jc=d.ZeroShotClassificationPipeline,Kc=d.ZeroShotImageClassificationPipeline,Zc=d.ZeroShotObjectDetectionPipeline,ed=d.bankers_round,td=d.cat,rd=d.cos_sim,nd=d.dot,od=d.dynamic_time_warping,sd=d.env,ad=d.full,id=d.full_like,ld=d.getKeyValueShapes,cd=d.hamming,dd=d.hanning,ud=d.interpolate,pd=d.interpolate_4d,hd=d.interpolate_data,md=d.is_chinese_char,_d=d.layer_norm,fd=d.log_softmax,gd=d.magnitude,Md=d.matmul,wd=d.max,Td=d.mean,bd=d.mean_pooling,xd=d.medianFilter,yd=d.mel_filter_bank,kd=d.min,Fd=d.ones,Pd=d.ones_like,Cd=d.permute,vd=d.permute_data,Sd=d.pipeline,Ad=d.quantize_embeddings,Ed=d.read_audio,Ld=d.rfft,zd=d.round,Id=d.softmax,Bd=d.spectrogram,Nd=d.stack,Dd=d.std_mean,Od=d.topk,Vd=d.window_function,jd=d.zeros,Rd=d.zeros_like;export{y as ASTFeatureExtractor,k as ASTForAudioClassification,F as ASTModel,P as ASTPreTrainedModel,C as AlbertForMaskedLM,v as AlbertForQuestionAnswering,S as AlbertForSequenceClassification,A as AlbertModel,E as AlbertPreTrainedModel,L as AlbertTokenizer,z as AudioClassificationPipeline,I as AutoConfig,B as AutoModel,N as AutoModelForAudioClassification,D as AutoModelForAudioFrameClassification,O as AutoModelForCTC,V as AutoModelForCausalLM,j as AutoModelForDepthEstimation,R as AutoModelForDocumentQuestionAnswering,G as AutoModelForImageClassification,q as AutoModelForImageFeatureExtraction,W as AutoModelForImageMatting,$ as AutoModelForImageSegmentation,U as AutoModelForImageToImage,X as AutoModelForMaskGeneration,Q as AutoModelForMaskedLM,H as AutoModelForNormalEstimation,Y as AutoModelForObjectDetection,J as AutoModelForQuestionAnswering,K as AutoModelForSemanticSegmentation,Z as AutoModelForSeq2SeqLM,ee as AutoModelForSequenceClassification,te as AutoModelForSpeechSeq2Seq,re as AutoModelForTextToSpectrogram,ne as AutoModelForTextToWaveform,oe as AutoModelForTokenClassification,se as AutoModelForUniversalSegmentation,ae as AutoModelForVision2Seq,ie as AutoModelForXVector,le as AutoModelForZeroShotObjectDetection,ce as AutoProcessor,de as AutoTokenizer,ue as AutomaticSpeechRecognitionPipeline,pe as BartForConditionalGeneration,he as BartForSequenceClassification,me as BartModel,_e as BartPretrainedModel,fe as BartTokenizer,ge as BaseModelOutput,Me as BaseStreamer,we as BeitFeatureExtractor,Te as BeitForImageClassification,be as BeitModel,xe as BeitPreTrainedModel,ye as BertForMaskedLM,ke as BertForQuestionAnswering,Fe as BertForSequenceClassification,Pe as BertForTokenClassification,Ce as BertModel,ve as BertPreTrainedModel,Se as BertTokenizer,Ae as BitImageProcessor,Ee as BlenderbotForConditionalGeneration,Le as BlenderbotModel,ze as BlenderbotPreTrainedModel,Ie as BlenderbotSmallForConditionalGeneration,Be as BlenderbotSmallModel,Ne as BlenderbotSmallPreTrainedModel,De as BlenderbotSmallTokenizer,Oe as BlenderbotTokenizer,Ve as BloomForCausalLM,je as BloomModel,Re as BloomPreTrainedModel,Ge as BloomTokenizer,qe as CLIPFeatureExtractor,We as CLIPImageProcessor,$e as CLIPModel,Ue as CLIPPreTrainedModel,Xe as CLIPSegForImageSegmentation,Qe as CLIPSegModel,He as CLIPSegPreTrainedModel,Ye as CLIPTextModel,Je as CLIPTextModelWithProjection,Ke as CLIPTokenizer,Ze as CLIPVisionModel,et as CLIPVisionModelWithProjection,tt as CamembertForMaskedLM,rt as CamembertForQuestionAnswering,nt as CamembertForSequenceClassification,ot as CamembertForTokenClassification,st as CamembertModel,at as CamembertPreTrainedModel,it as CamembertTokenizer,lt as CausalLMOutput,ct as CausalLMOutputWithPast,dt as ChineseCLIPFeatureExtractor,ut as ChineseCLIPModel,pt as ChineseCLIPPreTrainedModel,ht as ClapAudioModelWithProjection,mt as ClapFeatureExtractor,_t as ClapModel,ft as ClapPreTrainedModel,gt as ClapTextModelWithProjection,Mt as CodeGenForCausalLM,wt as CodeGenModel,Tt as CodeGenPreTrainedModel,bt as CodeGenTokenizer,xt as CodeLlamaTokenizer,yt as CohereForCausalLM,kt as CohereModel,Ft as CoherePreTrainedModel,Pt as CohereTokenizer,Ct as ConvBertForMaskedLM,vt as ConvBertForQuestionAnswering,St as ConvBertForSequenceClassification,At as ConvBertForTokenClassification,Et as ConvBertModel,Lt as ConvBertPreTrainedModel,zt as ConvBertTokenizer,It as ConvNextFeatureExtractor,Bt as ConvNextForImageClassification,Nt as ConvNextImageProcessor,Dt as ConvNextModel,Ot as ConvNextPreTrainedModel,Vt as ConvNextV2ForImageClassification,jt as ConvNextV2Model,Rt as ConvNextV2PreTrainedModel,Gt as DPTFeatureExtractor,qt as DPTForDepthEstimation,Wt as DPTImageProcessor,$t as DPTModel,Ut as DPTPreTrainedModel,Xt as DebertaForMaskedLM,Qt as DebertaForQuestionAnswering,Ht as DebertaForSequenceClassification,Yt as DebertaForTokenClassification,Jt as DebertaModel,Kt as DebertaPreTrainedModel,Zt as DebertaTokenizer,er as DebertaV2ForMaskedLM,tr as DebertaV2ForQuestionAnswering,rr as DebertaV2ForSequenceClassification,nr as DebertaV2ForTokenClassification,or as DebertaV2Model,sr as DebertaV2PreTrainedModel,ar as DebertaV2Tokenizer,ir as DecisionTransformerModel,lr as DecisionTransformerPreTrainedModel,cr as DeiTFeatureExtractor,dr as DeiTForImageClassification,ur as DeiTModel,pr as DeiTPreTrainedModel,hr as DepthAnythingForDepthEstimation,mr as DepthAnythingPreTrainedModel,_r as DepthEstimationPipeline,fr as DepthProForDepthEstimation,gr as DepthProPreTrainedModel,Mr as DetrFeatureExtractor,wr as DetrForObjectDetection,Tr as DetrForSegmentation,br as DetrModel,xr as DetrObjectDetectionOutput,yr as DetrPreTrainedModel,kr as DetrSegmentationOutput,Fr as Dinov2ForImageClassification,Pr as Dinov2Model,Cr as Dinov2PreTrainedModel,vr as DistilBertForMaskedLM,Sr as DistilBertForQuestionAnswering,Ar as DistilBertForSequenceClassification,Er as DistilBertForTokenClassification,Lr as DistilBertModel,zr as DistilBertPreTrainedModel,Ir as DistilBertTokenizer,Br as DocumentQuestionAnsweringPipeline,Nr as DonutFeatureExtractor,Dr as DonutSwinModel,Or as DonutSwinPreTrainedModel,Vr as EfficientNetForImageClassification,jr as EfficientNetImageProcessor,Rr as EfficientNetModel,Gr as EfficientNetPreTrainedModel,qr as ElectraForMaskedLM,Wr as ElectraForQuestionAnswering,$r as ElectraForSequenceClassification,Ur as ElectraForTokenClassification,Xr as ElectraModel,Qr as ElectraPreTrainedModel,Hr as ElectraTokenizer,Yr as EosTokenCriteria,Jr as EsmForMaskedLM,Kr as EsmForSequenceClassification,Zr as EsmForTokenClassification,en as EsmModel,tn as EsmPreTrainedModel,rn as EsmTokenizer,nn as FFT,on as FalconForCausalLM,sn as FalconModel,an as FalconPreTrainedModel,ln as FalconTokenizer,cn as FastViTForImageClassification,dn as FastViTModel,un as FastViTPreTrainedModel,pn as FeatureExtractionPipeline,hn as FeatureExtractor,mn as FillMaskPipeline,_n as Florence2ForConditionalGeneration,fn as Florence2PreTrainedModel,gn as Florence2Processor,Mn as GLPNFeatureExtractor,wn as GLPNForDepthEstimation,Tn as GLPNModel,bn as GLPNPreTrainedModel,xn as GPT2LMHeadModel,yn as GPT2Model,kn as GPT2PreTrainedModel,Fn as GPT2Tokenizer,Pn as GPTBigCodeForCausalLM,Cn as GPTBigCodeModel,vn as GPTBigCodePreTrainedModel,Sn as GPTJForCausalLM,An as GPTJModel,En as GPTJPreTrainedModel,Ln as GPTNeoForCausalLM,zn as GPTNeoModel,In as GPTNeoPreTrainedModel,Bn as GPTNeoXForCausalLM,Nn as GPTNeoXModel,Dn as GPTNeoXPreTrainedModel,On as GPTNeoXTokenizer,Vn as Gemma2ForCausalLM,jn as Gemma2Model,Rn as Gemma2PreTrainedModel,Gn as GemmaForCausalLM,qn as GemmaModel,Wn as GemmaPreTrainedModel,$n as GemmaTokenizer,Un as GraniteForCausalLM,Xn as GraniteModel,Qn as GranitePreTrainedModel,Hn as Grok1Tokenizer,Yn as GroupViTModel,Jn as GroupViTPreTrainedModel,Kn as HerbertTokenizer,Zn as HieraForImageClassification,eo as HieraModel,to as HieraPreTrainedModel,ro as HubertForCTC,no as HubertForSequenceClassification,oo as HubertModel,so as HubertPreTrainedModel,ao as ImageClassificationPipeline,io as ImageFeatureExtractionPipeline,lo as ImageFeatureExtractor,co as ImageMattingOutput,uo as ImageSegmentationPipeline,po as ImageToImagePipeline,ho as ImageToTextPipeline,mo as InterruptableStoppingCriteria,_o as JAISLMHeadModel,fo as JAISModel,go as JAISPreTrainedModel,Mo as LlamaForCausalLM,wo as LlamaModel,To as LlamaPreTrainedModel,bo as LlamaTokenizer,xo as LlavaForConditionalGeneration,yo as LlavaPreTrainedModel,ko as LongT5ForConditionalGeneration,Fo as LongT5Model,Po as LongT5PreTrainedModel,Co as M2M100ForConditionalGeneration,vo as M2M100Model,So as M2M100PreTrainedModel,Ao as M2M100Tokenizer,Eo as MBart50Tokenizer,Lo as MBartForCausalLM,zo as MBartForConditionalGeneration,Io as MBartForSequenceClassification,Bo as MBartModel,No as MBartPreTrainedModel,Do as MBartTokenizer,Oo as MPNetForMaskedLM,Vo as MPNetForQuestionAnswering,jo as MPNetForSequenceClassification,Ro as MPNetForTokenClassification,Go as MPNetModel,qo as MPNetPreTrainedModel,Wo as MPNetTokenizer,$o as MT5ForConditionalGeneration,Uo as MT5Model,Xo as MT5PreTrainedModel,Qo as MarianMTModel,Ho as MarianModel,Yo as MarianPreTrainedModel,Jo as MarianTokenizer,Ko as MaskFormerFeatureExtractor,Zo as MaskFormerForInstanceSegmentation,es as MaskFormerModel,ts as MaskFormerPreTrainedModel,rs as MaskedLMOutput,ns as MaxLengthCriteria,os as MistralForCausalLM,ss as MistralModel,as as MistralPreTrainedModel,is as MobileBertForMaskedLM,ls as MobileBertForQuestionAnswering,cs as MobileBertForSequenceClassification,ds as MobileBertModel,us as MobileBertPreTrainedModel,ps as MobileBertTokenizer,hs as MobileNetV1FeatureExtractor,ms as MobileNetV1ForImageClassification,_s as MobileNetV1Model,fs as MobileNetV1PreTrainedModel,gs as MobileNetV2FeatureExtractor,Ms as MobileNetV2ForImageClassification,ws as MobileNetV2Model,Ts as MobileNetV2PreTrainedModel,bs as MobileNetV3FeatureExtractor,xs as MobileNetV3ForImageClassification,ys as MobileNetV3Model,ks as MobileNetV3PreTrainedModel,Fs as MobileNetV4FeatureExtractor,Ps as MobileNetV4ForImageClassification,Cs as MobileNetV4Model,vs as MobileNetV4PreTrainedModel,Ss as MobileViTFeatureExtractor,As as MobileViTForImageClassification,Es as MobileViTImageProcessor,Ls as MobileViTModel,zs as MobileViTPreTrainedModel,Is as MobileViTV2ForImageClassification,Bs as MobileViTV2Model,Ns as MobileViTV2PreTrainedModel,Ds as ModelOutput,Os as Moondream1ForConditionalGeneration,Vs as MptForCausalLM,js as MptModel,Rs as MptPreTrainedModel,Gs as MusicgenForCausalLM,qs as MusicgenForConditionalGeneration,Ws as MusicgenModel,$s as MusicgenPreTrainedModel,Us as NllbTokenizer,Xs as NomicBertModel,Qs as NomicBertPreTrainedModel,Hs as NougatImageProcessor,Ys as NougatTokenizer,Js as OPTForCausalLM,Ks as OPTModel,Zs as OPTPreTrainedModel,ea as ObjectDetectionPipeline,ta as OpenELMForCausalLM,ra as OpenELMModel,na as OpenELMPreTrainedModel,oa as OwlViTFeatureExtractor,sa as OwlViTForObjectDetection,aa as OwlViTModel,ia as OwlViTPreTrainedModel,la as OwlViTProcessor,ca as Owlv2ForObjectDetection,da as Owlv2ImageProcessor,ua as Owlv2Model,pa as Owlv2PreTrainedModel,ha as Phi3ForCausalLM,ma as Phi3Model,_a as Phi3PreTrainedModel,fa as PhiForCausalLM,ga as PhiModel,Ma as PhiPreTrainedModel,wa as Pipeline,Ta as PreTrainedModel,ba as PreTrainedTokenizer,xa as PretrainedConfig,ya as PretrainedMixin,ka as Processor,Fa as PvtForImageClassification,Pa as PvtImageProcessor,Ca as PvtModel,va as PvtPreTrainedModel,Sa as PyAnnoteFeatureExtractor,Aa as PyAnnoteForAudioFrameClassification,Ea as PyAnnoteModel,La as PyAnnotePreTrainedModel,za as PyAnnoteProcessor,Ia as QuestionAnsweringModelOutput,Ba as QuestionAnsweringPipeline,Na as Qwen2ForCausalLM,Da as Qwen2Model,Oa as Qwen2PreTrainedModel,Va as Qwen2Tokenizer,ja as RTDetrForObjectDetection,Ra as RTDetrImageProcessor,Ga as RTDetrModel,qa as RTDetrObjectDetectionOutput,Wa as RTDetrPreTrainedModel,$a as RawImage,Ua as ResNetForImageClassification,Xa as ResNetModel,Qa as ResNetPreTrainedModel,Ha as RoFormerForMaskedLM,Ya as RoFormerForQuestionAnswering,Ja as RoFormerForSequenceClassification,Ka as RoFormerForTokenClassification,Za as RoFormerModel,ei as RoFormerPreTrainedModel,ti as RoFormerTokenizer,ri as RobertaForMaskedLM,ni as RobertaForQuestionAnswering,oi as RobertaForSequenceClassification,si as RobertaForTokenClassification,ai as RobertaModel,ii as RobertaPreTrainedModel,li as RobertaTokenizer,ci as SamImageProcessor,di as SamImageSegmentationOutput,ui as SamModel,pi as SamPreTrainedModel,hi as SamProcessor,mi as SapiensFeatureExtractor,_i as SapiensForDepthEstimation,fi as SapiensForNormalEstimation,gi as SapiensForSemanticSegmentation,Mi as SapiensPreTrainedModel,wi as SeamlessM4TFeatureExtractor,Ti as SegformerFeatureExtractor,bi as SegformerForImageClassification,xi as SegformerForSemanticSegmentation,yi as SegformerModel,ki as SegformerPreTrainedModel,Fi as Seq2SeqLMOutput,Pi as SequenceClassifierOutput,Ci as SiglipImageProcessor,vi as SiglipModel,Si as SiglipPreTrainedModel,Ai as SiglipTextModel,Ei as SiglipTokenizer,Li as SiglipVisionModel,zi as SpeechT5FeatureExtractor,Ii as SpeechT5ForSpeechToText,Bi as SpeechT5ForTextToSpeech,Ni as SpeechT5HifiGan,Di as SpeechT5Model,Oi as SpeechT5PreTrainedModel,Vi as SpeechT5Processor,ji as SpeechT5Tokenizer,Ri as SqueezeBertForMaskedLM,Gi as SqueezeBertForQuestionAnswering,qi as SqueezeBertForSequenceClassification,Wi as SqueezeBertModel,$i as SqueezeBertPreTrainedModel,Ui as SqueezeBertTokenizer,Xi as StableLmForCausalLM,Qi as StableLmModel,Hi as StableLmPreTrainedModel,Yi as Starcoder2ForCausalLM,Ji as Starcoder2Model,Ki as Starcoder2PreTrainedModel,Zi as StoppingCriteria,el as StoppingCriteriaList,tl as SummarizationPipeline,rl as Swin2SRForImageSuperResolution,nl as Swin2SRImageProcessor,ol as Swin2SRModel,sl as Swin2SRPreTrainedModel,al as SwinForImageClassification,il as SwinModel,ll as SwinPreTrainedModel,cl as T5ForConditionalGeneration,dl as T5Model,ul as T5PreTrainedModel,pl as T5Tokenizer,hl as TableTransformerForObjectDetection,ml as TableTransformerModel,_l as TableTransformerObjectDetectionOutput,fl as TableTransformerPreTrainedModel,gl as Tensor,Ml as Text2TextGenerationPipeline,wl as TextClassificationPipeline,Tl as TextGenerationPipeline,bl as TextStreamer,xl as TextToAudioPipeline,yl as TokenClassificationPipeline,kl as TokenClassifierOutput,Fl as TokenizerModel,Pl as TrOCRForCausalLM,Cl as TrOCRPreTrainedModel,vl as TranslationPipeline,Sl as UniSpeechForCTC,Al as UniSpeechForSequenceClassification,El as UniSpeechModel,Ll as UniSpeechPreTrainedModel,zl as UniSpeechSatForAudioFrameClassification,Il as UniSpeechSatForCTC,Bl as UniSpeechSatForSequenceClassification,Nl as UniSpeechSatModel,Dl as UniSpeechSatPreTrainedModel,Ol as ViTFeatureExtractor,Vl as ViTForImageClassification,jl as ViTImageProcessor,Rl as ViTMAEModel,Gl as ViTMAEPreTrainedModel,ql as ViTMSNForImageClassification,Wl as ViTMSNModel,$l as ViTMSNPreTrainedModel,Ul as ViTModel,Xl as ViTPreTrainedModel,Ql as VisionEncoderDecoderModel,Hl as VitMatteForImageMatting,Yl as VitMatteImageProcessor,Jl as VitMattePreTrainedModel,Kl as VitsModel,Zl as VitsModelOutput,ec as VitsPreTrainedModel,tc as VitsTokenizer,rc as Wav2Vec2BertForCTC,nc as Wav2Vec2BertForSequenceClassification,oc as Wav2Vec2BertModel,sc as Wav2Vec2BertPreTrainedModel,ac as Wav2Vec2CTCTokenizer,ic as Wav2Vec2FeatureExtractor,lc as Wav2Vec2ForAudioFrameClassification,cc as Wav2Vec2ForCTC,dc as Wav2Vec2ForSequenceClassification,uc as Wav2Vec2Model,pc as Wav2Vec2PreTrainedModel,hc as Wav2Vec2ProcessorWithLM,mc as WavLMForAudioFrameClassification,_c as WavLMForCTC,fc as WavLMForSequenceClassification,gc as WavLMForXVector,Mc as WavLMModel,wc as WavLMPreTrainedModel,Tc as WeSpeakerFeatureExtractor,bc as WeSpeakerResNetModel,xc as WeSpeakerResNetPreTrainedModel,yc as WhisperFeatureExtractor,kc as WhisperForConditionalGeneration,Fc as WhisperModel,Pc as WhisperPreTrainedModel,Cc as WhisperProcessor,vc as WhisperTextStreamer,Sc as WhisperTokenizer,Ac as XLMForQuestionAnswering,Ec as XLMForSequenceClassification,Lc as XLMForTokenClassification,zc as XLMModel,Ic as XLMPreTrainedModel,Bc as XLMRobertaForMaskedLM,Nc as XLMRobertaForQuestionAnswering,Dc as XLMRobertaForSequenceClassification,Oc as XLMRobertaForTokenClassification,Vc as XLMRobertaModel,jc as XLMRobertaPreTrainedModel,Rc as XLMRobertaTokenizer,Gc as XLMTokenizer,qc as XLMWithLMHeadModel,Wc as XVectorOutput,$c as YolosFeatureExtractor,Uc as YolosForObjectDetection,Xc as YolosModel,Qc as YolosObjectDetectionOutput,Hc as YolosPreTrainedModel,Yc as ZeroShotAudioClassificationPipeline,Jc as ZeroShotClassificationPipeline,Kc as ZeroShotImageClassificationPipeline,Zc as ZeroShotObjectDetectionPipeline,ed as bankers_round,td as cat,rd as cos_sim,nd as dot,od as dynamic_time_warping,sd as env,ad as full,id as full_like,ld as getKeyValueShapes,cd as hamming,dd as hanning,ud as interpolate,pd as interpolate_4d,hd as interpolate_data,md as is_chinese_char,_d as layer_norm,fd as log_softmax,gd as magnitude,Md as matmul,wd as max,Td as mean,bd as mean_pooling,xd as medianFilter,yd as mel_filter_bank,kd as min,Fd as ones,Pd as ones_like,Cd as permute,vd as permute_data,Sd as pipeline,Ad as quantize_embeddings,Ed as read_audio,Ld as rfft,zd as round,Id as softmax,Bd as spectrogram,Nd as stack,Dd as std_mean,Od as topk,Vd as window_function,jd as zeros,Rd as zeros_like};
 //# sourceMappingURL=transformers.min.mjs.map