
    ^j/                    t   d Z ddlZddlmZ ddlmZ ddlmZ ddlm	Z	m
Z
mZmZ dd	lmZ erdd
lmZ ddlmZ  G d dee      Z ej(                  e      Z eg d      Z eg d      Z eg d      Z eg d      Z eg d      Z edg      Z eg d      Z edg      Z eg d      Z eddg      Z  eg d      Z! eg d      Z" eddg      Z# eg d      Z$ eg  e%e$jM                               ddd d!d"d#d$d%d&d'd(d)      Z' eg d*      Z( eg d+      Z) eg d,      Z* eg d-      Z+ eg d.      Z, ed/d0g      Z- eg d1      Z. eg d2      Z/ eg d3      Z0 eg d4      Z1 ed5g      Z2 eg d6      Z3 eg d7      Z4 eg d8      Z5 eg d9      Z6 eg d:      Z7 eg d;      Z8 eg d<      Z9 eg d=      Z: ed>g      Z; eg d?      Z< eg d@      Z= edAdBg      Z> eg dC      Z? eg dD      Z@ eg dE      ZA eg dF      ZB edGg      ZC eg dH      ZD eg dI      ZE edJdKg      ZF edLdMg      ZG edNdOg      ZH edPg      ZI eg dQ      ZJ edRdSg      ZK edTdUg      ZL edVg      ZM edWdXg      ZN eee      ZO eee      ZP eee      ZQ eee      ZR eee      ZS eee@      ZT eee      ZU eee      ZV eee       ZW eee!      ZX eee"      ZY eee$      ZZ eee'      Z[ eee#      Z\ eee3      Z] eee4      Z^ eee(      Z_ eee      Z` eee      Za eee)      Zb eee*      Zc eee+      Zd eee,      Ze eee-      Zf eee.      Zg eee0      Zh eee1      Zi eee2      Zj eee5      Zk eee6      Zl eee7      Zm eee8      Zn eee9      Zo eee:      Zp eee;      Zq eee/      Zr eee<      Zs eee=      Zt eee>      Zu eee?      Zv eeeA      Zw eeeB      Zx eeeC      Zy eeeD      Zz eeeE      Z{ eeeF      Z| eeeG      Z} eeeH      Z~ eeeI      Z eeeJ      Z eeeK      Z eeeL      Z eeeM      Z eeeN      Z G dY dZe
      Z G d[ d\e
      Z G d] d^e
      Z G d_ d`e
      Z G da dbe
      Z G dc dde
      Z ee      Z G de dfe
      Z eedgh      Z G di dje
      Z eedkh      Z G dl dme
      Z eednh      Z G do dpe
      Z eedqdrs      Z G dt due
      Z eedvh      Z G dw dxe
      Z eedyh      Z G dz d{e
      Z eed|d}s      Z G d~ de
      Z eedds      Z G d de
      Z eedds      Z G d de
      Z eedh      Z G d de
      Z eedh      Z G d de
      Z eedh      Z G d de
      Z eedh      Z G d de
      Z eedh      Z G d de
      Z eedh      Z G d de
      Z eedh      Z G d de
      Z eedh      Z G d de
      Z eedh      Z G d de
      Z eedh      Z G d de
      Z eedh      Z G d de
      Z eedh      Z G d de
      Z eedh      Z G d de
      Z eedh      Z G d de
      Z eedh      Z G d de
      Z eedh      Z G d de
      Z eedh      Z G d de
      Z eedh      Z G d de
      Z eedh      Z G d de
      Z eedh      Z G d de
      Z eedĬh      Z G dń de
      Z eedǬh      Z G dȄ de
      Z eedʬh      Z G d˄ de
      Z eedͬh      Z G d΄ de
      Z G dЄ de
      Z G d҄ de
      Z G dԄ de	      Z eed֬h      Z G dׄ de
      Z eed٬h      Z G dڄ de
      Z eedܬh      Z G d݄ de
      Z eed߬h      Z G d de
      Z eedh      Z G d de
      Z eedh      Z G d de
      Z eedh      Zg dZy)zAuto Model class.    N)OrderedDict)TYPE_CHECKING   )logging   )_BaseAutoBackboneClass_BaseAutoModelClass_LazyAutoMappingauto_class_update)CONFIG_MAPPING_NAMES)GenerationMixin)PreTrainedModelc                       e Zd Zy)_BaseModelWithGenerateN)__name__
__module____qualname__     q/var/www/ramen.bs-engineer-server.com/venv/lib/python3.12/site-packages/transformers/models/auto/modeling_auto.pyr   r   #   s    r   r   (  )afmoe
AfmoeModel)aimv2
Aimv2Modelaimv2_vision_modelAimv2VisionModelalbertAlbertModelalign
AlignModelaltclipAltCLIPModel)apertusApertusModel)arcee
ArceeModel)aria	AriaModel)	aria_textAriaTextModel)audio-spectrogram-transformerASTModel)audioflamingo3AudioFlamingo3Model)audioflamingo3_encoderAudioFlamingo3Encoder)
autoformerAutoformerModel)
aya_visionAyaVisionModel)bamba
BambaModelbark	BarkModel)bart	BartModelbeit	BeitModelbert	BertModel)bert-generationBertGenerationEncoderbig_birdBigBirdModel)bigbird_pegasusBigBirdPegasusModel)biogptBioGptModelbitBitModel)bitnetBitNetModel)
blenderbotBlenderbotModel)blenderbot-smallBlenderbotSmallModelblip	BlipModel)blip-2
Blip2Model)blip_2_qformerBlip2QFormerModel)bloom
BloomModel)bltBltModel)bridgetowerBridgeTowerModel)bros	BrosModel)	camembertCamembertModel)canineCanineModel)	chameleonChameleonModelchinese_clipChineseCLIPModel)chinese_clip_vision_modelChineseCLIPVisionModel)clap	ClapModelclip	CLIPModelclip_text_modelCLIPTextModel)clip_vision_modelCLIPVisionModelclipsegCLIPSegModel)clvp!ClvpModelForConditionalGeneration)codegenCodeGenModel)cohereCohereModel)cohere2Cohere2Model)cohere2_moeCohere2MoeModelcohere2_visionCohere2VisionModel)
cohere_asrCohereAsrModelconditional_detrConditionalDetrModel)convbertConvBertModelconvnextConvNextModel
convnextv2ConvNextV2Model)cosmos3_omniCosmos3OmniModel)cpmantCpmAntModelcsmCsmForConditionalGeneration)ctrl	CTRLModel)cvtCvtModel)cwmCwmModel)d_fine
DFineModeldab-detrDabDetrModeldacDacModel)data2vec-audioData2VecAudioModeldata2vec-textData2VecTextModeldata2vec-visionData2VecVisionModel)dbrx	DbrxModeldebertaDebertaModel
deberta-v2DebertaV2Model)decision_transformerDecisionTransformerModel)deepseek_ocr2DeepseekOcr2Model)deepseek_v2DeepseekV2Model)deepseek_v3DeepseekV3Model)deepseek_v32DeepseekV32Model)deepseek_v4DeepseekV4Model)deepseek_vlDeepseekVLModel)deepseek_vl_hybridDeepseekVLHybridModeldeformable_detrDeformableDetrModel)deimv2Deimv2Modeldeit	DeiTModel	depth_proDepthProModeldetr	DetrModel)diaDiaModel)	diffllamaDiffLlamaModel)diffusion_gemmaDiffusionGemmaModeldinat
DinatModeldinov2Dinov2Modeldinov2_with_registersDinov2WithRegistersModeldinov3_convnextDINOv3ConvNextModel
dinov3_vitDINOv3ViTModel
distilbertDistilBertModel)doge	DogeModel)
donut-swinDonutSwinModel)dots1
Dots1Model)dprDPRQuestionEncoderdptDPTModeledgetamEdgeTamModel)edgetam_videoEdgeTamVideoModel)edgetam_vision_modelEdgeTamVisionModel)efficientloftrEfficientLoFTRModelefficientnetEfficientNetModelelectraElectraModel)emu3	Emu3Model)encodecEncodecModel)ernie
ErnieModel)ernie4_5Ernie4_5Model)ernie4_5_moeErnie4_5_MoeModel)ernie4_5_vl_moeErnie4_5_VLMoeModel)esmEsmModel)eurobertEuroBertModel)evollaEvollaModel)exaone4Exaone4Model)	exaone4_5Exaone4_5_Model)exaone4_5_visionExaone4_5_VisionModel)
exaone_moeExaoneMoeModel)falconFalconModel)	falcon_h1FalconH1Model)falcon_mambaFalconMambaModel)fast_vlmFastVlmModelfastspeech2_conformerFastSpeech2ConformerModel"fastspeech2_conformer_with_hifiganFastSpeech2ConformerWithHifiGanflaubertFlaubertModel)flava
FlavaModel)	flex_olmoFlexOlmoModel)	florence2Florence2Model)fnet	FNetModelfocalnetFocalNetModel)fsmt	FSMTModel)funnel)FunnelModelFunnelBaseModel)fuyu	FuyuModel)gemma
GemmaModel)gemma2Gemma2Model)gemma3Gemma3Model)gemma3_textGemma3TextModel)gemma3nGemma3nModel)gemma3n_audioGemma3nAudioEncoder)gemma3n_textGemma3nTextModel)gemma3n_visionTimmWrapperModel)gemma4Gemma4Model)gemma4_audioGemma4AudioModel)gemma4_textGemma4TextModel)gemma4_unifiedGemma4UnifiedModel)gemma4_unified_textGemma4UnifiedTextModel)gemma4_visionGemma4VisionModel)gitGitModel)glmGlmModel)glm4	Glm4Model)glm46vGlm46VModel)glm4_moeGlm4MoeModel)glm4_moe_liteGlm4MoeLiteModel)glm4v
Glm4vModel)	glm4v_moeGlm4vMoeModel)glm4v_moe_textGlm4vMoeTextModel)glm4v_moe_visionGlm4vMoeVisionModel)
glm4v_textGlm4vTextModel)glm4v_visionGlm4vVisionModel)	glm_imageGlmImageModel)glm_image_textGlmImageTextModel)glm_image_visionGlmImageVisionModel)glm_image_vqmodelGlmImageVQVAE)glm_moe_dsaGlmMoeDsaModel)glm_ocrGlmOcrModel)glm_ocr_textGlmOcrTextModel)glm_ocr_visionGlmOcrVisionModel)glmasrGlmAsrModel)glmasr_encoderGlmAsrEncoder)glmgarn  glpn	GLPNModel)got_ocr2GotOcr2Model)gpt-sw3	GPT2Model)gpt2r  )gpt_bigcodeGPTBigCodeModel)gpt_neoGPTNeoModel)gpt_neoxGPTNeoXModel)gpt_neox_japaneseGPTNeoXJapaneseModel)gpt_ossGptOssModel)gptj	GPTJModel)graniteGraniteModel)granite4_visionGranite4VisionModel)granite_speechGraniteSpeechModel)granite_speech_plusGraniteSpeechPlusModel)
granitemoeGraniteMoeModel)granitemoehybridGraniteMoeHybridModel)granitemoesharedGraniteMoeSharedModel)grounding-dinoGroundingDinoModel)groupvitGroupViTModel)heliumHeliumModelhgnet_v2HGNetV2Backbonehiera
HieraModelhiggs_audio_v2$HiggsAudioV2ForConditionalGenerationhiggs_audio_v2_tokenizerHiggsAudioV2TokenizerModel)hrm_textHrmTextModel)hubertHubertModel)hunyuan_v1_denseHunYuanDenseV1Model)hunyuan_v1_moeHunYuanMoEV1Model)
hunyuan_vlHunYuanVLModel)hunyuan_vl_textHunYuanVLTextModel)hunyuan_vl_visionHunYuanVLVisionTransformer)hy_v3	HYV3Model)hyperclovaxHyperCLOVAXModelibert
IBertModel)ideficsIdeficsModel)idefics2Idefics2Model)idefics3Idefics3Model)idefics3_visionIdefics3VisionTransformerijepa
IJepaModelimagegptImageGPTModel)informerInformerModel)inkling_mm_modelInklingModel)instructblipInstructBlipModel)instructblipvideoInstructBlipVideoModel)internvlInternVLModel)internvl_visionInternVLVisionModel)jais2
Jais2Model)jamba
JambaModel)janus
JanusModel)jetmoeJetMoeModel)jina_embeddings_v3JinaEmbeddingsV3Model)kimi_k25Kimi_K25Model)kimi_k25_visionKimi_K25VisionModel)kosmos-2Kosmos2Model)
kosmos-2.5Kosmos2_5Model)kyutai_speech_to_textKyutaiSpeechToTextModel)lagunaLagunaModellasr_ctc
LasrForCTC)lasr_encoderLasrEncoder)layoutlmLayoutLMModel)
layoutlmv2LayoutLMv2Model)
layoutlmv3LayoutLMv3Model)ledLEDModellevit
LevitModel)lfm2	Lfm2Model)lfm2_moeLfm2MoeModel)lfm2_vlLfm2VlModel	lightglueLightGlueForKeypointMatching)lighton_ocrLightOnOcrModel)lilt	LiltModel)llama
LlamaModelllama4Llama4ForConditionalGeneration)llama4_textLlama4TextModel)llava
LlavaModel)
llava_nextLlavaNextModel)llava_next_videoLlavaNextVideoModel)llava_onevisionLlavaOnevisionModel)longcat_flashLongcatFlashModel
longformerLongformerModel)longt5LongT5Model)luke	LukeModel)lw_detrLwDetrModel)lxmertLxmertModel)m2m_100M2M100Model)mamba
MambaModel)mamba2Mamba2Model)marianMarianModel)markuplmMarkupLMModel)mask2formerMask2FormerModel)
maskformerMaskFormerModel)maskformer-swinMaskFormerSwinModel)mbart
MBartModel)megatron-bertMegatronBertModel)mellumMellumModel
metaclip_2MetaClip2Model)zmgp-strMgpstrForSceneTextRecognition)mimi	MimiModel)mimo_v2_flashMiMoV2FlashModel)minicpm3MiniCPM3Model)minicpmv4_6MiniCPMV4_6Model)minimaxMiniMaxModel)
minimax_m2MiniMaxM2Model)minimax_m3_vlMiniMaxM3VLModel)minimax_m3_vl_textMiniMaxM3VLTextModel)	ministralMinistralModel)
ministral3Ministral3Model)mistralMistralModel)mistral3Mistral3Model)mistral4Mistral4Model)mixtralMixtralModelmlcdMLCDVisionModelmlcd_vision_modelr~  )mllamaMllamaModel)mm-grounding-dinoMMGroundingDinoModel
mobilebertMobileBertModelmobilenet_v1MobileNetV1Modelmobilenet_v2MobileNetV2Model	mobilevitMobileViTModelmobilevitv2MobileViTV2Model)
modernbertModernBertModel)modernbert-decoderModernBertDecoderModel)modernvbertModernVBertModel)	moonshineMoonshineModel)moonshine_streamingMoonshineStreamingModel)moshi
MoshiModel)mpnet
MPNetModel)mptMptModel)mraMraModel)mt5MT5Model)musicflamingoMusicFlamingoModel)musicgenMusicgenModel)musicgen_melodyMusicgenMelodyModel)mvpMvpModel)nanochatNanoChatModel)nemotronNemotronModelnemotron3_5_asrNemotron3_5AsrForRNNTnemotron_asr_streamingNemotronAsrStreamingForRNNT)nemotron_asr_streaming_encoderNemotronAsrStreamingEncoder)
nemotron_hNemotronHModel)nllb-moeNllbMoeModel)
nomic_bertNomicBertModelnystromformerNystromformerModel)olmo	OlmoModel)olmo2
Olmo2Model)olmo3
Olmo3Model)olmo_hybridOlmoHybridModel)olmoe
OlmoeModelzomdet-turboOmDetTurboForObjectDetection)	oneformerOneFormerModel)
openai-gptOpenAIGPTModel)openai_privacy_filterOpenAIPrivacyFilterModel)optOPTModel)ovis2
Ovis2Model)owlv2
Owlv2Model)owlvitOwlViTModel)paddleocr_vlPaddleOCRVLModel)	paligemmaPaliGemmaModelparakeet_ctcParakeetForCTC)parakeet_encoderParakeetEncoderparakeet_rnntParakeetForRNNTparakeet_tdtParakeetForTDT)patchtsmixerPatchTSMixerModel)patchtstPatchTSTModel)pe_audioPeAudioModel)pe_audio_encoderPeAudioEncoder)pe_audio_videoPeAudioVideoModel)pe_audio_video_encoderPeAudioVideoEncoder)pe_videoPeVideoModel)pe_video_encoderPeVideoEncoder)pegasusPegasusModel)	pegasus_xPegasusXModel)	perceiverPerceiverModel)perception_lmPerceptionLMModel)	persimmonPersimmonModel)phiPhiModel)phi3	Phi3Model)phi4_multimodalPhi4MultimodalModel)phimoePhimoeModel)pi0PI0Modelpixio
PixioModel)pixtralPixtralVisionModel)plbartPLBartModel
poolformerPoolFormerModel)pp_doclayout_v3PPDocLayoutV3Model)pp_ocrv5_mobile_recPPOCRV5MobileRecModel)pp_ocrv5_server_recPPOCRV5ServerRecModel)pp_ocrv6_medium_detPPOCRV6MediumDetModel)pp_ocrv6_small_detPPOCRV6SmallDetModel)pp_ocrv6_small_recPPOCRV6SmallRecModel)pp_ocrv6_tiny_recPPOCRV6TinyRecModel)
prophetnetProphetNetModelpvtPvtModel)pvt_v2
PvtV2Model)qianfan_ocrQianfanOCRModel)qianfan_ocr_visionQianfanOCRVisionModel)qwen2
Qwen2Model)
qwen2_5_vlQwen2_5_VLModel)qwen2_5_vl_textQwen2_5_VLTextModel)qwen2_audioQwen2AudioModel)qwen2_audio_encoderQwen2AudioEncoder)	qwen2_moeQwen2MoeModel)qwen2_vlQwen2VLModel)qwen2_vl_textQwen2VLTextModel)qwen3
Qwen3Model)qwen3_5Qwen3_5Model)qwen3_5_moeQwen3_5MoeModel)qwen3_5_moe_textQwen3_5MoeTextModel)qwen3_5_moe_visionQwen3_5MoeVisionModel)qwen3_5_textQwen3_5TextModel)qwen3_5_visionQwen3_5VisionModel)	qwen3_asrQwen3ASRModel)qwen3_asr_encoderQwen3ASREncoder)	qwen3_moeQwen3MoeModel)
qwen3_nextQwen3NextModel)qwen3_vlQwen3VLModel)qwen3_vl_moeQwen3VLMoeModel)qwen3_vl_moe_textQwen3VLMoeTextModel)qwen3_vl_moe_visionQwen3VLMoeVisionModel)qwen3_vl_textQwen3VLTextModel)qwen3_vl_visionQwen3VLVisionModel)radio
RadioModel)recurrent_gemmaRecurrentGemmaModelreformerReformerModelregnetRegNetModelrembertRemBertModelresnetResNetModel)rf_detrRfDetrModelrobertaRobertaModelroberta-prelayernormRobertaPreLayerNormModelroc_bertRoCBertModelroformerRoFormerModel)rt_detrRTDetrModel)
rt_detr_v2RTDetrV2Model)rwkv	RwkvModelsamSamModelsam2	Sam2Model)sam2_hiera_det_modelSam2HieraDetModel)
sam2_videoSam2VideoModel)sam2_vision_modelSam2VisionModel)sam3	Sam3Model)sam3_lite_textSam3LiteTextModel)sam3_lite_text_text_modelSam3LiteTextTextModelsam3_trackerSam3TrackerModelr  )sam3_tracker_videoSam3TrackerVideoModel)
sam3_videoSam3VideoModel)sam3_vision_modelSam3VisionModel)sam3_vit_modelSam3ViTModelsam_hq
SamHQModel)sam_hq_vision_modelSamHQVisionModel)sam_vision_modelSamVisionModel)sapiens2Sapiens2Model)seamless_m4tSeamlessM4TModel)seamless_m4t_v2SeamlessM4Tv2Model)seed_ossSeedOssModel	segformerSegformerModel)seggptSegGptModel)sewSEWModel)sew-d	SEWDModelsiglipSiglipModelsiglip2Siglip2Model)siglip2_vision_modelSiglip2VisionModelsiglip_vision_modelSiglipVisionModel)smollm3SmolLM3Model)smolvlmSmolVLMModel)smolvlm_visionSmolVLMVisionTransformer)
solar_openSolarOpenModel)speech_to_textSpeech2TextModel)speecht5SpeechT5Model)splinterSplinterModelsqueezebertSqueezeBertModel)stablelmStableLmModel)
starcoder2Starcoder2ModelswiftformerSwiftFormerModelswin	SwinModelswin2srSwin2SRModelswinv2Swinv2Model)switch_transformersSwitchTransformersModel)t5T5Model)t5gemmaT5GemmaModel)t5gemma2T5Gemma2Model)t5gemma2_encoderT5Gemma2Encodertable-transformerTableTransformerModel)tapas
TapasModel)textnetTextNetModel)time_series_transformerTimeSeriesTransformerModel)timesfmTimesFmModel)
timesfm2_5TimesFm2_5ModeltimesformerTimesformerModeltimm_backboneTimmBackbonetimm_wrapperrZ  tipsv2Tipsv2Modeltipsv2_text_modelTipsv2TextModeltipsv2_vision_modelTipsv2VisionModel)tvpTvpModel)udop	UdopModel)umt5	UMT5Model)	unispeechUniSpeechModel)unispeech-satUniSpeechSatModel)univnetUnivNetModel)uvdoc
UVDocModel)
vaultgemmaVaultGemmaModelvibevoice_acoustic_tokenizerVibeVoiceAcousticTokenizerModel)$vibevoice_acoustic_tokenizer_decoder&VibeVoiceAcousticTokenizerDecoderModel)$vibevoice_acoustic_tokenizer_encoder&VibeVoiceAcousticTokenizerEncoderModel)vibevoice_asrVibeVoiceAsrModel)video_llama_3VideoLlama3Model)video_llama_3_visionVideoLlama3VisionModel)video_llavaVideoLlavaModelvideomaeVideoMAEModel
videoprismVideoPrismClipModelvideoprism_vision_modelVideoPrismVisionModel)vilt	ViltModel)vipllavaVipLlavaModel)zvision-text-dual-encoderVisionTextDualEncoderModel)visual_bertVisualBertModelvitViTModelvit_maeViTMAEModelvit_msnViTMSNModelvitdetVitDetModelvits	VitsModelvivit
VivitModel)vjepa2VJEPA2Model)voxtralVoxtralModel)voxtral_encoderVoxtralEncoder)voxtral_realtimeVoxtralRealtimeModel)voxtral_realtime_encoderVoxtralRealtimeEncoder)voxtral_realtime_textVoxtralRealtimeTextModel)wav2vec2Wav2Vec2Model)wav2vec2-bertWav2Vec2BertModel)wav2vec2-conformerWav2Vec2ConformerModel)wavlm
WavLMModel)whisperWhisperModel)xclip
XCLIPModel)xcodecXcodecModelxcodec2Xcodec2Model)xglm	XGLMModelxlmXLMModelxlm-robertaXLMRobertaModelxlm-roberta-xlXLMRobertaXLModel)xlnet
XLNetModel)xlstm
xLSTMModel)xmod	XmodModelyolos
YolosModel)yoso	YosoModel)youtu
YoutuModel)zamba
ZambaModel)zamba2Zamba2Model)zaya	ZayaModel)`)r   AlbertForPreTrainingr1   &AudioFlamingo3ForConditionalGenerationr>   BartForConditionalGeneration)rD   BertForPreTraining)rI   BigBirdForPreTrainingr_   BloomForCausalLMrg   CamembertForMaskedLMcolmodernvbertColModernVBertForRetrievalcolpaliColPaliForRetrieval)colqwen2ColQwen2ForRetrievalr   CTRLLMHeadModelr   Data2VecTextForMaskedLMr   DebertaForMaskedLMr   DebertaV2ForMaskedLMr   DistilBertForMaskedLM)r  ElectraForPreTraining)r  ErnieForPreTrainingr  EvollaForProteinText2Textr   Exaone4ForCausalLMr&  ExaoneMoeForCausalLMr,  FalconMambaForCausalLMr7  FlaubertWithLMHeadModel)r9  FlavaForPreTrainingr=  !Florence2ForConditionalGeneration)r?  FNetForPreTrainingrD  FSMTForConditionalGeneration)rF  FunnelForPreTrainingrO  Gemma3ForConditionalGenerationr[  Gemma4ForConditionalGenerationra  %Gemma4UnifiedForConditionalGenerationr  GlmAsrForConditionalGenerationr  GPT2LMHeadModelr  r  r  GPTBigCodeForCausalLM)r  HieraForPreTrainingr  IBertForMaskedLMr  IdeficsForVisionText2Textr   Idefics2ForConditionalGenerationr   Idefics3ForConditionalGenerationr  InklingForConditionalGenerationr  JanusForConditionalGenerationr  LayoutLMForMaskedLMr1  LlavaForConditionalGenerationr3  !LlavaNextForConditionalGenerationr5  &LlavaNextVideoForConditionalGenerationr7  &LlavaOnevisionForConditionalGenerationr<  LongformerForMaskedLMr@  LukeForMaskedLM)rD  LxmertForPreTrainingrH  MambaForCausalLMrJ  Mamba2ForCausalLM)rX  MegatronBertForPreTrainingrv   Mistral3ForConditionalGenerationrx  Mistral4ForCausalLMr  MllamaForConditionalGeneration)r  MobileBertForPreTrainingr  MPNetForMaskedLMr  MptForCausalLMr  MraForMaskedLMr  %MusicFlamingoForConditionalGenerationr  MvpForConditionalGenerationr  NanoChatForCausalLMr  NllbMoeForConditionalGenerationr  OpenAIGPTLMHeadModelr  !PaliGemmaForConditionalGenerationr;  "Qwen2AudioForConditionalGenerationrS   Qwen3ASRForConditionalGenerationrz  RobertaForMaskedLMr}  RobertaPreLayerNormForMaskedLM)r  RoCBertForPreTrainingr  RwkvForCausalLM)r  SplinterForPreTrainingr  SqueezeBertForMaskedLMr  *SwitchTransformersForConditionalGenerationr  T5ForConditionalGenerationr  T5GemmaForConditionalGenerationr   T5Gemma2ForConditionalGenerationr  TapasForMaskedLM)r  UniSpeechForPreTraining)r  UniSpeechSatForPreTrainingr+  $VibeVoiceAsrForConditionalGenerationr1  "VideoLlavaForConditionalGeneration)r4  VideoMAEForPreTrainingr>   VipLlavaForConditionalGeneration)rA  VisualBertForPreTraining)rG  ViTMAEForPreTrainingrW  VoxtralForConditionalGenerationr[  'VoxtralRealtimeForConditionalGeneration)ra  Wav2Vec2ForPreTraining)re  Wav2Vec2ConformerForPreTrainingru  XLMWithLMHeadModelrx  XLMRobertaForMaskedLMr{  XLMRobertaXLForMaskedLMr}  XLNetLMHeadModelr  xLSTMForCausalLMr  XmodForMaskedLM))r   AfmoeForCausalLM)r'   ApertusForCausalLM)r)   ArceeForCausalLM)r-   AriaTextForCausalLM)r9   BambaForCausalLM)r>   BartForCausalLM)rD   BertLMHeadModel)rF   BertGenerationDecoder)rI   BigBirdForCausalLM)rK   BigBirdPegasusForCausalLM)rM   BioGptForCausalLM)rR   BitNetForCausalLM)rT   BlenderbotForCausalLM)rV   BlenderbotSmallForCausalLMr  )ra   BltForCausalLM)rg   CamembertForCausalLM)r   CodeGenForCausalLM)r   CohereForCausalLM)r   Cohere2ForCausalLM)r   Cohere2MoeForCausalLM)r   CpmAntForCausalLMr  )r   CwmForCausalLM)r   Data2VecTextForCausalLM)r   DbrxForCausalLM)r   DeepseekV2ForCausalLM)r   DeepseekV3ForCausalLM)r   DeepseekV32ForCausalLM)r   DeepseekV4ForCausalLM)r   DiffLlamaForCausalLM)r   DogeForCausalLM)r   Dots1ForCausalLM)r  ElectraForCausalLM)r  Emu3ForCausalLM)r  ErnieForCausalLM)r  Ernie4_5ForCausalLM)r  Ernie4_5_MoeForCausalLMr  r  )r(  FalconForCausalLM)r*  FalconH1ForCausalLMr  )r;  FlexOlmoForCausalLMrI  FuyuForCausalLM)rK  GemmaForCausalLM)rM  Gemma2ForCausalLMr  )rQ  Gemma3ForCausalLMrS  Gemma3nForConditionalGeneration)rW  Gemma3nForCausalLMr  )gemma4_assistantGemma4AssistantForCausalLM)r_  Gemma4ForCausalLMr  )gemma4_unified_assistant!Gemma4UnifiedAssistantForCausalLM)rc  Gemma4UnifiedForCausalLMrg  GitForCausalLM)ri  GlmForCausalLM)rk  Glm4ForCausalLM)ro  Glm4MoeForCausalLM)rq  Glm4MoeLiteForCausalLM)r  GlmMoeDsaForCausalLMr  GotOcr2ForConditionalGenerationr  r  r  )r  GPTNeoForCausalLM)r  GPTNeoXForCausalLM)r  GPTNeoXJapaneseForCausalLM)r  GptOssForCausalLM)r  GPTJForCausalLM)r  GraniteForCausalLM)r  GraniteMoeForCausalLM)r  GraniteMoeHybridForCausalLM)r  GraniteMoeSharedForCausalLM)r  HeliumForCausalLM)r  HrmTextForCausalLM)r  HunYuanDenseV1ForCausalLM)r  HunYuanMoEV1ForCausalLM)r  HYV3ForCausalLM)r  HyperCLOVAXForCausalLM)inkling_textInklingForCausalLM)r  Jais2ForCausalLM)r  JambaForCausalLM)r  JetMoeForCausalLM)r  LagunaForCausalLM)r  Lfm2ForCausalLM)r  Lfm2MoeForCausalLM)r*  LlamaForCausalLM)r-  Llama4ForCausalLM)r/  r  )r9  LongcatFlashForCausalLMr  r  )rL  MarianForCausalLM)rV  MBartForCausalLM)rX  MegatronBertForCausalLM)rZ  MellumForCausalLM)rb  MiMoV2FlashForCausalLM)rd  MiniCPM3ForCausalLM)rh  MiniMaxForCausalLM)rj  MiniMaxM2ForCausalLM)rn  MiniMaxM3VLForCausalLM)rp  MinistralForCausalLM)rr  Ministral3ForCausalLM)rt  MistralForCausalLM)rz  MixtralForCausalLM)r  MllamaForCausalLM)r  ModernBertDecoderForCausalLM)r  MoshiForCausalLMr  )r  MusicgenForCausalLM)r  MusicgenMelodyForCausalLM)r  MvpForCausalLMr  )r  NemotronForCausalLM)r  NemotronHForCausalLM)r  OlmoForCausalLM)r  Olmo2ForCausalLM)r  Olmo3ForCausalLM)r  OlmoHybridForCausalLM)r  OlmoeForCausalLMr  )r  OPTForCausalLM)r  PegasusForCausalLM)r  PersimmonForCausalLM)r  PhiForCausalLM)r
  Phi3ForCausalLMr  Phi4MultimodalForCausalLM)r  PhimoeForCausalLM)r  PLBartForCausalLM)r*  ProphetNetForCausalLM)r5  Qwen2ForCausalLM)r?  Qwen2MoeForCausalLM)rE  Qwen3ForCausalLM)rG  Qwen3_5ForCausalLM)rI  Qwen3_5MoeForCausalLM)rK  r  )rO  r  )rW  Qwen3MoeForCausalLM)rY  Qwen3NextForCausalLM)ri  RecurrentGemmaForCausalLM)rl  ReformerModelWithLMHead)rr  RemBertForCausalLM)rz  RobertaForCausalLM)r}  RobertaPreLayerNormForCausalLM)r  RoCBertForCausalLM)r  RoFormerForCausalLMr  )r  SeedOssForCausalLM)r  SmolLM3ForCausalLM)r  SolarOpenForCausalLM)r  StableLmForCausalLM)r  Starcoder2ForCausalLM)trocrTrOCRForCausalLM)r"  VaultGemmaForCausalLM)ri  WhisperForCausalLM)rr  XGLMForCausalLMr0  )rx  XLMRobertaForCausalLM)r{  XLMRobertaXLForCausalLMr6  r8  )r  XmodForCausalLM)r  YoutuForCausalLM)r  ZambaForCausalLM)r  Zamba2ForCausalLM)r  ZayaForCausalLM):r   r@   rO   r   r   r   r   r   r   r   r   r   r   r   r   r   r   r   r   r  rA  r  r  r  r  r  )r-  Llama4VisionModelr|  r  )r  MllamaVisionModelr  r  r  r  r  r  r,  rn  rt  r  r  r  r  r  r  r  r  r  r	  r  r3  r9  rC  rF  rI  rL  rR  r  ))r   DeiTForMaskedImageModeling)rB  FocalNetForMaskedImageModeling)r  SwinForMaskedImageModeling)r  Swinv2ForMaskedImageModeling)rD  ViTForMaskedImageModeling)r  ImageGPTForCausalImageModeling)*)rA   BeitForImageClassification)rP   BitForImageClassification)ru   CLIPForImageClassification)r   ConvNextForImageClassification)r    ConvNextV2ForImageClassification)r   CvtForImageClassification)r   $Data2VecVisionForImageClassification)r   )DeiTForImageClassification%DeiTForImageClassificationWithTeacher)r   DinatForImageClassification)r   Dinov2ForImageClassification)r   )Dinov2WithRegistersForImageClassification)r   DonutSwinForImageClassification)r	  "EfficientNetForImageClassification)rB  FocalNetForImageClassification)r  HGNetV2ForImageClassification)r  HieraForImageClassification)r  IJepaForImageClassification)r  ImageGPTForImageClassification)r  )LevitForImageClassification&LevitForImageClassificationWithTeacher)r]  MetaClip2ForImageClassification)r  !MobileNetV1ForImageClassification)r  !MobileNetV2ForImageClassification)r  MobileViTForImageClassification)r  !MobileViTV2ForImageClassification)r  )&PerceiverForImageClassificationLearned&PerceiverForImageClassificationFourier-PerceiverForImageClassificationConvProcessing)r   PoolFormerForImageClassification)pp_lcnetPPLCNetForImageClassification)r-  PvtForImageClassification)r/  PvtV2ForImageClassification)ro  RegNetForImageClassification)ru  ResNetForImageClassification)r  SegformerForImageClassification)shieldgemma2"ShieldGemma2ForImageClassification)r  SiglipForImageClassification)r  Siglip2ForImageClassification)r  !SwiftFormerForImageClassification)r  SwinForImageClassification)r  Swinv2ForImageClassification)r  TextNetForImageClassification)r
  !TimmWrapperForImageClassification)rD  ViTForImageClassification)rJ  ViTMSNForImageClassificationr   DetrForSegmentation)
)rA   BeitForSemanticSegmentation)r   %Data2VecVisionForSemanticSegmentation)r   DPTForSemanticSegmentation)r  "MobileNetV2ForSemanticSegmentation)r   MobileViTForSemanticSegmentation)r  "MobileViTV2ForSemanticSegmentation)r  Sapiens2ForSemanticSegmentation)r   SegformerForSemanticSegmentation)
tipsv2_dpt Tipsv2DptForSemanticSegmentation)upernetUperNetForSemanticSegmentationrR  !MaskFormerForInstanceSegmentation)rw  RfDetrForInstanceSegmentation)r  )eomtEomtForUniversalSegmentation)eomt_dinov3"EomtDinov3ForUniversalSegmentation)rP  #Mask2FormerForUniversalSegmentationr  )r  !OneFormerForUniversalSegmentation)videomtVideomtForUniversalSegmentation))r  !TimesformerForVideoClassification)r4  VideoMAEForVideoClassification)r:   VideoPrismForVideoClassification)rS  VivitForVideoClassification)rU  VJEPA2ForVideoClassificationr  r  )O)r+   AriaForConditionalGeneration)r7   !AyaVisionForConditionalGeneration)rY   BlipForConditionalGenerationr[   Blip2ForConditionalGeneration)rk   !ChameleonForConditionalGeneration)r   %Cohere2VisionForConditionalGeneration)r   #Cosmos3OmniForConditionalGeneration)r   $DeepseekOcr2ForConditionalGeneration)r   "DeepseekVLForConditionalGeneration)r   (DeepseekVLHybridForConditionalGeneration)r   DiffusionGemmaForBlockDiffusion)r  Emu3ForConditionalGeneration)r  &Ernie4_5_VLMoeForConditionalGenerationr  )r"  "Exaone4_5_ForConditionalGeneration)r.  FastVlmForConditionalGenerationr  rc  r  rh  r  r  rq  )rm  Glm46VForConditionalGeneration)rs  Glm4vForConditionalGeneration)ru   Glm4vMoeForConditionalGeneration)r  GlmOcrForConditionalGeneration)r  r=  rx  )r  &Granite4VisionForConditionalGeneration)r  !HunYuanVLForConditionalGenerationr  r  r  r  )r  $InstructBlipForConditionalGeneration)r  )InstructBlipVideoForConditionalGeneration)r   InternVLForConditionalGenerationr  )r   Kimi_K25ForConditionalGeneration)r  Kosmos2ForConditionalGeneration)r  !Kosmos2_5ForConditionalGeneration)r!  Lfm2VlForConditionalGeneration)r&  "LightOnOcrForConditionalGenerationr,  r  r  r  r  )rf  #MiniCPMV4_6ForConditionalGeneration)rl  'MiniMaxM3SparseForConditionalGenerationr  r  r  )r  Ovis2ForConditionalGeneration)r  #PaddleOCRVLForConditionalGenerationr  )r  $PerceptionLMForConditionalGeneration)r  PI0ForConditionalGeneration)
pix2struct"Pix2StructForConditionalGeneration)pp_chart2tablery  )pp_formulanet$PPFormulaNetForConditionalGeneration)r1  "QianfanOCRForConditionalGeneration)qwen2_5_omni_thinker*Qwen2_5OmniThinkerForConditionalGeneration)r7  "Qwen2_5_VLForConditionalGeneration)rA  Qwen2VLForConditionalGeneration)rG  Qwen3_5ForConditionalGeneration)rI  "Qwen3_5MoeForConditionalGeneration)qwen3_omni_moe_thinker+Qwen3OmniMoeThinkerForConditionalGeneration)r[  Qwen3VLForConditionalGeneration)r]  "Qwen3VLMoeForConditionalGeneration)r  r  )r  SmolVLMForConditionalGenerationr  )r  UdopForConditionalGeneration)r-  #VideoLlama3ForConditionalGenerationr#  r&  )zvision-encoder-decoderVisionEncoderDecoderModelr  r  %GraniteSpeechForConditionalGenerationr  )GraniteSpeechPlusForConditionalGenerationr	  *KyutaiSpeechToTextForConditionalGenerationr  qwen2_5_omni#Qwen2_5OmniForConditionalGenerationr  r	  qwen3_omni_moe$Qwen3OmniMoeForConditionalGenerationr!  r*  r,  ).)r   AlbertForMaskedLMr  )rD   BertForMaskedLM)rI   BigBirdForMaskedLMr  )r   ConvBertForMaskedLMr  r  r  r  )r  ElectraForMaskedLM)r  ErnieForMaskedLM)r  EsmForMaskedLM)r  EuroBertForMaskedLMr  )r?  FNetForMaskedLM)rF  FunnelForMaskedLMr  )r  JinaEmbeddingsV3ForMaskedLMr  r  r  rV  MBartForConditionalGeneration)rX  MegatronBertForMaskedLM)r  MobileBertForMaskedLM)r  ModernBertForMaskedLM)r  ModernVBertForMaskedLMr  r  r  )r  NomicBertForMaskedLM)r  NystromformerForMaskedLM)r  PerceiverForMaskedLM)rl  ReformerForMaskedLM)rr  RemBertForMaskedLMr  r  )r  RoCBertForMaskedLM)r  RoFormerForMaskedLMr  r  r0  r2  r4  r:  )r  YosoForMaskedLM))r   !ConditionalDetrForObjectDetection)r   DFineForObjectDetection)r   DabDetrForObjectDetection)r    DeformableDetrForObjectDetection)r   Deimv2ForObjectDetection)r   DetrForObjectDetection)rB  LwDetrForObjectDetection)pp_doclayout_v2PPDocLayoutV2ForObjectDetection)r  PPDocLayoutV3ForObjectDetection)pp_ocrv5_mobile_det"PPOCRV5MobileDetForObjectDetection)pp_ocrv5_server_det"PPOCRV5ServerDetForObjectDetection)r"  "PPOCRV6MediumDetForObjectDetection)r$  !PPOCRV6SmallDetForObjectDetection)rw  RfDetrForObjectDetection)r  RTDetrForObjectDetection)r  RTDetrV2ForObjectDetection)r  "TableTransformerForObjectDetection)r  YolosForObjectDetection))r  GroundingDinoForObjectDetection)r  !MMGroundingDinoForObjectDetectionr  )r  Owlv2ForObjectDetection)r  OwlViTForObjectDetection))chmv2CHMv2ForDepthEstimation)depth_anythingDepthAnythingForDepthEstimation)r   DepthProForDepthEstimation)r   DPTForDepthEstimation)r  GLPNForDepthEstimation)prompt_depth_anything%PromptDepthAnythingForDepthEstimation)r  Tipsv2DptForDepthEstimation)zoedepthZoeDepthForDepthEstimation))r  "PPOCRV5MobileRecForTextRecognition)r   "PPOCRV5ServerRecForTextRecognition)r&  !PPOCRV6SmallRecForTextRecognition)r(   PPOCRV6TinyRecForTextRecognition)slanetSLANetForTableRecognition)slanextSLANeXtForTableRecognition)#r  r  )rK   &BigBirdPegasusForConditionalGeneration)rT   "BlenderbotForConditionalGeneration)rV   'BlenderbotSmallForConditionalGeneration)zencoder-decoderEncoderDecoderModelr  r  re  rg  )r  LEDForConditionalGeneration)r>  LongT5ForConditionalGeneration)rF  M2M100ForConditionalGeneration)rL  MarianMTModelr|  )r  MT5ForConditionalGenerationr  r  r  )r  PegasusForConditionalGeneration)r    PegasusXForConditionalGeneration)r  PLBartForConditionalGeneration)r*  "ProphetNetForConditionalGenerationr  r	  )r  SeamlessM4TForTextToText)r  SeamlessM4Tv2ForTextToTextr  r  r  r  )r  UMT5ForConditionalGenerationr!  r*  r,  ))r   !CohereAsrForConditionalGeneration)r   DiaForConditionalGenerationre  rg  ri  )r  !MoonshineForConditionalGeneration)r  *MoonshineStreamingForConditionalGeneration)	pop2piano!Pop2PianoForConditionalGenerationr	  )r  SeamlessM4TForSpeechToText)r  SeamlessM4Tv2ForSpeechToText)zspeech-encoder-decoderSpeechEncoderDecoderModel)r  #Speech2TextForConditionalGeneration)r  SpeechT5ForSpeechToTextr!  r*  r,  )ri  WhisperForConditionalGeneration)x)r   AlbertForSequenceClassification)r)   ArceeForSequenceClassification)r>   BartForSequenceClassification)rD   BertForSequenceClassification)rI    BigBirdForSequenceClassification)rK   'BigBirdPegasusForSequenceClassification)rM   BioGptForSequenceClassification)r_   BloomForSequenceClassification)rg   "CamembertForSequenceClassification)ri   CanineForSequenceClassification)r   !ConvBertForSequenceClassification)r   CTRLForSequenceClassification)r   %Data2VecTextForSequenceClassification)r    DebertaForSequenceClassification)r   "DebertaV2ForSequenceClassification)r   #DeepseekV2ForSequenceClassification)r   #DeepseekV3ForSequenceClassification)r   "DiffLlamaForSequenceClassification)r   #DistilBertForSequenceClassification)r   DogeForSequenceClassification)r   ElectraForSequenceClassification)r  ErnieForSequenceClassification)r  EsmForSequenceClassification)r  !EuroBertForSequenceClassification)r    Exaone4ForSequenceClassification)r(  FalconForSequenceClassification)r7  !FlaubertForSequenceClassification)r?  FNetForSequenceClassification)rF  FunnelForSequenceClassification)rK  GemmaForSequenceClassification)rM  Gemma2ForSequenceClassification)rO  Gemma3ForSequenceClassification)rQ  #Gemma3TextForSequenceClassification)ri  GlmForSequenceClassification)rk  Glm4ForSequenceClassification)r  GPT2ForSequenceClassification)r  r  )r  #GPTBigCodeForSequenceClassification)r  GPTNeoForSequenceClassification)r   GPTNeoXForSequenceClassification)r  GptOssForSequenceClassification)r  GPTJForSequenceClassification)r  HeliumForSequenceClassification)r  'HunYuanDenseV1ForSequenceClassification)r  %HunYuanMoEV1ForSequenceClassification)r  IBertForSequenceClassification)r  JambaForSequenceClassification)r  JetMoeForSequenceClassification)r  )JinaEmbeddingsV3ForSequenceClassification)r  !LayoutLMForSequenceClassification)r  #LayoutLMv2ForSequenceClassification)r  #LayoutLMv3ForSequenceClassification)r(  LiltForSequenceClassification)r*  LlamaForSequenceClassification)r<  #LongformerForSequenceClassification)r@  LukeForSequenceClassification)rN  !MarkupLMForSequenceClassification)rV  MBartForSequenceClassification)rX  %MegatronBertForSequenceClassification)rd  !MiniCPM3ForSequenceClassification)rh   MiniMaxForSequenceClassification)rp  "MinistralForSequenceClassification)rr  #Ministral3ForSequenceClassification)rt   MistralForSequenceClassification)rx  !Mistral4ForSequenceClassification)rz   MixtralForSequenceClassification)r  #MobileBertForSequenceClassification)r  #ModernBertForSequenceClassification)r  *ModernBertDecoderForSequenceClassification)r  $ModernVBertForSequenceClassification)r  MPNetForSequenceClassification)r  MptForSequenceClassification)r  MraForSequenceClassification)r  MT5ForSequenceClassification)r  MvpForSequenceClassification)r  !NemotronForSequenceClassification)r  "NomicBertForSequenceClassification)r  &NystromformerForSequenceClassification)r  OlmoForSequenceClassification)r  Olmo2ForSequenceClassification)r  Olmo3ForSequenceClassification)r  "OpenAIGPTForSequenceClassification)r  OPTForSequenceClassification)r  "PerceiverForSequenceClassification)r  "PersimmonForSequenceClassification)r  PhiForSequenceClassification)r
  Phi3ForSequenceClassification)r  PhimoeForSequenceClassification)r  PLBartForSequenceClassification)r5  Qwen2ForSequenceClassification)r?  !Qwen2MoeForSequenceClassification)rE  Qwen3ForSequenceClassification)rG   Qwen3_5ForSequenceClassification)rO  $Qwen3_5TextForSequenceClassification)rW  !Qwen3MoeForSequenceClassification)rY  "Qwen3NextForSequenceClassification)rl  !ReformerForSequenceClassification)rr   RemBertForSequenceClassification)rz   RobertaForSequenceClassification)r}  ,RobertaPreLayerNormForSequenceClassification)r   RoCBertForSequenceClassification)r  !RoFormerForSequenceClassification)r   SeedOssForSequenceClassification)r   SmolLM3ForSequenceClassification)r  $SqueezeBertForSequenceClassification)r  !StableLmForSequenceClassification)r  #Starcoder2ForSequenceClassification)r  T5ForSequenceClassification)r   T5GemmaForSequenceClassification)r  !T5Gemma2ForSequenceClassification)r  TapasForSequenceClassification)r  UMT5ForSequenceClassification)ru  XLMForSequenceClassification)rx  #XLMRobertaForSequenceClassification)r{  %XLMRobertaXLForSequenceClassification)r}  XLNetForSequenceClassification)r  XmodForSequenceClassification)r  YosoForSequenceClassification)r  ZambaForSequenceClassification)r  Zamba2ForSequenceClassification)M)r   AlbertForQuestionAnswering)r)   ArceeForQuestionAnswering)r>   BartForQuestionAnswering)rD   BertForQuestionAnswering)rI   BigBirdForQuestionAnswering)rK   "BigBirdPegasusForQuestionAnswering)r_   BloomForQuestionAnswering)rg   CamembertForQuestionAnswering)ri   CanineForQuestionAnswering)r   ConvBertForQuestionAnswering)r    Data2VecTextForQuestionAnswering)r   DebertaForQuestionAnswering)r   DebertaV2ForQuestionAnswering)r   DiffLlamaForQuestionAnswering)r   DistilBertForQuestionAnswering)r  ElectraForQuestionAnswering)r  ErnieForQuestionAnswering)r   Exaone4ForQuestionAnswering)r(  FalconForQuestionAnswering)r7  "FlaubertForQuestionAnsweringSimple)r?  FNetForQuestionAnswering)rF  FunnelForQuestionAnswering)r  GPT2ForQuestionAnswering)r  GPTNeoForQuestionAnswering)r  GPTNeoXForQuestionAnswering)r  GPTJForQuestionAnswering)r  IBertForQuestionAnswering)r  $JinaEmbeddingsV3ForQuestionAnsweringr  LayoutLMv2ForQuestionAnsweringr  LayoutLMv3ForQuestionAnswering)r  LEDForQuestionAnswering)r(  LiltForQuestionAnswering)r*  LlamaForQuestionAnswering)r<  LongformerForQuestionAnswering)r@  LukeForQuestionAnswering)rD  LxmertForQuestionAnswering)rN  MarkupLMForQuestionAnswering)rV  MBartForQuestionAnswering)rX   MegatronBertForQuestionAnswering)rh  MiniMaxForQuestionAnswering)rp  MinistralForQuestionAnswering)rr  Ministral3ForQuestionAnswering)rt  MistralForQuestionAnswering)rz  MixtralForQuestionAnswering)r  MobileBertForQuestionAnswering)r  ModernBertForQuestionAnswering)r  MPNetForQuestionAnswering)r  MptForQuestionAnswering)r  MraForQuestionAnswering)r  MT5ForQuestionAnswering)r  MvpForQuestionAnswering)r  NemotronForQuestionAnswering)r  !NystromformerForQuestionAnswering)r  OPTForQuestionAnswering)r5  Qwen2ForQuestionAnswering)r?  Qwen2MoeForQuestionAnswering)rE  Qwen3ForQuestionAnswering)rW  Qwen3MoeForQuestionAnswering)rY  Qwen3NextForQuestionAnswering)rl  ReformerForQuestionAnswering)rr  RemBertForQuestionAnswering)rz  RobertaForQuestionAnswering)r}  'RobertaPreLayerNormForQuestionAnswering)r  RoCBertForQuestionAnswering)r  RoFormerForQuestionAnswering)r  SeedOssForQuestionAnswering)r  SmolLM3ForQuestionAnswering)r  SplinterForQuestionAnswering)r  SqueezeBertForQuestionAnswering)r  T5ForQuestionAnswering)r  UMT5ForQuestionAnswering)ru  XLMForQuestionAnsweringSimple)rx  XLMRobertaForQuestionAnswering)r{   XLMRobertaXLForQuestionAnswering)r}  XLNetForQuestionAnsweringSimple)r  XmodForQuestionAnswering)r  YosoForQuestionAnswering)r  TapasForQuestionAnswering))rY   BlipForQuestionAnsweringr0  )r<  ViltForQuestionAnswering))r  LayoutLMForQuestionAnsweringrf  rh  )_)r   AlbertForTokenClassification)r'   ApertusForTokenClassification)r)   ArceeForTokenClassification)rD   BertForTokenClassification)rI   BigBirdForTokenClassification)rM   BioGptForTokenClassification)r_   BloomForTokenClassification)re   BrosForTokenClassification)rg   CamembertForTokenClassification)ri   CanineForTokenClassification)r   ConvBertForTokenClassification)r   "Data2VecTextForTokenClassification)r   DebertaForTokenClassification)r   DebertaV2ForTokenClassification)r    DeepseekV3ForTokenClassification)r   DiffLlamaForTokenClassification)r    DistilBertForTokenClassification)r  ElectraForTokenClassification)r  ErnieForTokenClassification)r  EsmForTokenClassification)r  EuroBertForTokenClassification)r   Exaone4ForTokenClassification)r(  FalconForTokenClassification)r7  FlaubertForTokenClassification)r?  FNetForTokenClassification)rF  FunnelForTokenClassification)rK  GemmaForTokenClassification)rM  Gemma2ForTokenClassification)ri  GlmForTokenClassification)rk  Glm4ForTokenClassification)r  GPT2ForTokenClassification)r  r  )r   GPTBigCodeForTokenClassification)r  GPTNeoForTokenClassification)r  GPTNeoXForTokenClassification)r  GptOssForTokenClassification)r  HeliumForTokenClassification)r  IBertForTokenClassification)r  &JinaEmbeddingsV3ForTokenClassification)r  LayoutLMForTokenClassification)r   LayoutLMv2ForTokenClassification)r   LayoutLMv3ForTokenClassification)r(  LiltForTokenClassification)r*  LlamaForTokenClassification)r<   LongformerForTokenClassification)r@  LukeForTokenClassification)rN  MarkupLMForTokenClassification)rX  "MegatronBertForTokenClassification)rh  MiniMaxForTokenClassification)rp  MinistralForTokenClassification)rr   Ministral3ForTokenClassification)rt  MistralForTokenClassification)rx  Mistral4ForTokenClassification)rz  MixtralForTokenClassification)r   MobileBertForTokenClassification)r   ModernBertForTokenClassification)r  !ModernVBertForTokenClassification)r  MPNetForTokenClassification)r  MptForTokenClassification)r  MraForTokenClassification)r  MT5ForTokenClassification)r  NemotronForTokenClassification)r  NomicBertForTokenClassification)r  #NystromformerForTokenClassification)r  )OpenAIPrivacyFilterForTokenClassification)r  PersimmonForTokenClassification)r  PhiForTokenClassification)r
  Phi3ForTokenClassification)r5  Qwen2ForTokenClassification)r?  Qwen2MoeForTokenClassification)rE  Qwen3ForTokenClassification)rG  Qwen3_5ForTokenClassification)rS  Qwen3ASRForTokenClassification)rW  Qwen3MoeForTokenClassification)rY  Qwen3NextForTokenClassification)rr  RemBertForTokenClassification)rz  RobertaForTokenClassification)r}  )RobertaPreLayerNormForTokenClassification)r  RoCBertForTokenClassification)r  RoFormerForTokenClassification)r  SeedOssForTokenClassification)r  SmolLM3ForTokenClassification)r  !SqueezeBertForTokenClassification)r  StableLmForTokenClassification)r   Starcoder2ForTokenClassification)r  T5ForTokenClassification)r  T5GemmaForTokenClassification)r  T5Gemma2ForTokenClassification)r  UMT5ForTokenClassification)ru  XLMForTokenClassification)rx   XLMRobertaForTokenClassification)r{  "XLMRobertaXLForTokenClassification)r}  XLNetForTokenClassification)r  XmodForTokenClassification)r  YosoForTokenClassification)#)r   AlbertForMultipleChoice)rD   BertForMultipleChoice)rI   BigBirdForMultipleChoice)rg   CamembertForMultipleChoice)ri   CanineForMultipleChoice)r   ConvBertForMultipleChoice)r   Data2VecTextForMultipleChoice)r   DebertaV2ForMultipleChoice)r   DistilBertForMultipleChoice)r  ElectraForMultipleChoice)r  ErnieForMultipleChoice)r7  FlaubertForMultipleChoice)r?  FNetForMultipleChoice)rF  FunnelForMultipleChoice)r  IBertForMultipleChoice)r<  LongformerForMultipleChoice)r@  LukeForMultipleChoice)rX  MegatronBertForMultipleChoice)r  MobileBertForMultipleChoice)r  ModernBertForMultipleChoice)r  MPNetForMultipleChoice)r  MraForMultipleChoice)r  NystromformerForMultipleChoice)rr  RemBertForMultipleChoice)rz  RobertaForMultipleChoice)r}  $RobertaPreLayerNormForMultipleChoice)r  RoCBertForMultipleChoice)r  RoFormerForMultipleChoice)r  SqueezeBertForMultipleChoice)ru  XLMForMultipleChoice)rx  XLMRobertaForMultipleChoice)r{  XLMRobertaXLForMultipleChoice)r}  XLNetForMultipleChoice)r  XmodForMultipleChoice)r  YosoForMultipleChoice))rD   BertForNextSentencePrediction)r  ErnieForNextSentencePrediction)r?  FNetForNextSentencePrediction)rX  %MegatronBertForNextSentencePrediction)r  #MobileBertForNextSentencePrediction))r/   ASTForAudioClassification)r   &Data2VecAudioForSequenceClassification)r  HubertForSequenceClassification)r  SEWForSequenceClassification)r  SEWDForSequenceClassification)r  "UniSpeechForSequenceClassification)r  %UniSpeechSatForSequenceClassification)ra  !Wav2Vec2ForSequenceClassification)rc  %Wav2Vec2BertForSequenceClassification)re  *Wav2Vec2ConformerForSequenceClassification)rg  WavLMForSequenceClassification)ri  WhisperForAudioClassification))r   Data2VecAudioForCTC)r  HubertForCTCr  r  )r  	SEWForCTC)r  
SEWDForCTC)r  UniSpeechForCTC)r  UniSpeechSatForCTC)ra  Wav2Vec2ForCTC)rc  Wav2Vec2BertForCTC)re  Wav2Vec2ConformerForCTC)rg  WavLMForCTC)r  r  r  r  ))r   (Data2VecAudioForAudioFrameClassification)r  'UniSpeechSatForAudioFrameClassification)ra  #Wav2Vec2ForAudioFrameClassification)rc  'Wav2Vec2BertForAudioFrameClassification)re  ,Wav2Vec2ConformerForAudioFrameClassification)rg   WavLMForAudioFrameClassification))r   Data2VecAudioForXVector)r  UniSpeechSatForXVector)ra  Wav2Vec2ForXVector)rc  Wav2Vec2BertForXVector)re  Wav2Vec2ConformerForXVector)rg  WavLMForXVectorr0  )r  SpeechT5ForTextToSpeech)r;   r   r3  r  )r   MusicgenForConditionalGeneration)r  &MusicgenMelodyForConditionalGenerationrk  rn  )r  SeamlessM4TForTextToSpeech)r  SeamlessM4Tv2ForTextToSpeechrO  )r!   r$   rX   )r[   Blip2ForImageTextRetrievalrm   rt   r|   r\  r  r  r  r6  ) )rA   BeitBackbone)rP   BitBackbone)r   ConvNextBackbone)r   ConvNextV2Backbone)r   DinatBackbone)r   Dinov2Backbone)r   Dinov2WithRegistersBackbone)r   DINOv3ConvNextBackbone)r   DINOv3ViTBackbone)rB  FocalNetBackboner  )r  HieraBackbone)lw_detr_vitLwDetrViTBackbone)rT  MaskFormerSwinBackbone)r  PixioBackbone)r  PPLCNetBackbone)pp_lcnet_v3PPLCNetV3Backbone)pp_lcnet_v4PPLCNetV4Backbone)r/  PvtV2Backbone)ru  ResNetBackbone)rf_detr_dinov2RfDetrDinov2Backbone)rt_detr_resnetRTDetrResNetBackbone)r  Sapiens2Backbone)r  SwinBackbone)r  Swinv2Backbone)r  TextNetBackboner  )r  Tipsv2VisionBackbone)r  ri  )uvdoc_backboneUVDocBackbone)rM  VitDetBackbone)vitpose_backboneVitPoseBackbone)r   )r  r  r  r  )r  r  r  )r  r  r  )
superpointSuperPointForKeypointDetection))r  !EfficientLoFTRForKeypointMatchingr#  )	superglueSuperGlueForKeypointMatching) r   rC   rH   rw   r   r   r   r   r  )r  Emu3TextModelr6  r  )r-  r0  r;  )r  MllamaTextModelr  )r  MT5EncoderModelr  rk  rq  ry  r|  r  r  r  )r  T5EncoderModel)r  T5GemmaEncoderModelr  )r  UMT5EncoderModelrt  rw  rz  )r  'PatchTSMixerForTimeSeriesClassification)r  PatchTSTForClassification)r  PatchTSMixerForRegression)r  PatchTSTForRegression)r  TimesFmModelForPrediction)r  TimesFm2_5ModelForPrediction)r  Swin2SRForImageSuperResolution)r   r  r$  ro  )r  Sapiens2ForImageMatting)vitmatteVitMatteForImageMatting)r  Sapiens2ForNormalEstimation)r  Tipsv2DptForNormalEstimation)r  Sapiens2ForPointmapEstimation)r  Sapiens2ForPoseEstimation)vitposeVitPoseForPoseEstimationc                       e Zd ZeZy)AutoModelForMaskGenerationN)r   r   r   !MODEL_FOR_MASK_GENERATION_MAPPING_model_mappingr   r   r   r  r  T      6Nr   r  c                       e Zd ZeZy)AutoModelForKeypointDetectionN)r   r   r   $MODEL_FOR_KEYPOINT_DETECTION_MAPPINGr  r   r   r   r  r  X      9Nr   r  c                       e Zd ZeZy)AutoModelForKeypointMatchingN)r   r   r   #MODEL_FOR_KEYPOINT_MATCHING_MAPPINGr  r   r   r   r  r  \      8Nr   r  c                       e Zd ZeZy)AutoModelForTextEncodingN)r   r   r   MODEL_FOR_TEXT_ENCODING_MAPPINGr  r   r   r   r  r  `      4Nr   r  c                       e Zd ZeZy)AutoModelForImageToImageN)r   r   r    MODEL_FOR_IMAGE_TO_IMAGE_MAPPINGr  r   r   r   r  r  d  s    5Nr   r  c                       e Zd ZeZy)	AutoModelN)r   r   r   MODEL_MAPPINGr  r   r   r   r  r  h  s    "Nr   r  c                       e Zd ZeZy)AutoModelForPreTrainingN)r   r   r   MODEL_FOR_PRETRAINING_MAPPINGr  r   r   r   r  r  o  s    2Nr   r  pretraining)head_docc                   `     e Zd ZeZeded    deej                  e   z  ddf fd       Z
 xZS )AutoModelForCausalLMclspretrained_model_name_or_pathreturnr   c                 *    t        |   |g|i |S Nsuperfrom_pretrainedr  r  
model_argskwargs	__class__s       r   r  z$AutoModelForCausalLM.from_pretrainedz  !     w&'D\z\U[\\r   )r   r   r   MODEL_FOR_CAUSAL_LM_MAPPINGr  classmethodtypestrosPathLiker  __classcell__r  s   @r   r  r  v  sO    0N ]()]'*R[[-='=]
 
"] ]r   r  zcausal language modelingc                       e Zd ZeZy)AutoModelForMaskedLMN)r   r   r   MODEL_FOR_MASKED_LM_MAPPINGr  r   r   r   r  r    s    0Nr   r  zmasked language modelingc                       e Zd ZeZy)AutoModelForSeq2SeqLMN)r   r   r   &MODEL_FOR_SEQ_TO_SEQ_CAUSAL_LM_MAPPINGr  r   r   r   r  r        ;Nr   r  z&sequence-to-sequence language modelingzgoogle-t5/t5-base)r  checkpoint_for_examplec                       e Zd ZeZy)"AutoModelForSequenceClassificationN)r   r   r   )MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPINGr  r   r   r   r  r    s    >Nr   r  zsequence classificationc                       e Zd ZeZy)AutoModelForQuestionAnsweringN)r   r   r   $MODEL_FOR_QUESTION_ANSWERING_MAPPINGr  r   r   r   r  r    r  r   r  zquestion answeringc                       e Zd ZeZy)"AutoModelForTableQuestionAnsweringN)r   r   r   *MODEL_FOR_TABLE_QUESTION_ANSWERING_MAPPINGr  r   r   r   r  r        ?Nr   r  ztable question answeringzgoogle/tapas-base-finetuned-wtqc                       e Zd ZeZy)#AutoModelForVisualQuestionAnsweringN)r   r   r   +MODEL_FOR_VISUAL_QUESTION_ANSWERING_MAPPINGr  r   r   r   r  r    s    @Nr   r  zvisual question answeringzdandelin/vilt-b32-finetuned-vqac                       e Zd ZeZy)%AutoModelForDocumentQuestionAnsweringN)r   r   r   -MODEL_FOR_DOCUMENT_QUESTION_ANSWERING_MAPPINGr  r   r   r   r  r    s    BNr   r  zdocument question answeringz/impira/layoutlm-document-qa", revision="52e01b3c                       e Zd ZeZy)AutoModelForTokenClassificationN)r   r   r   &MODEL_FOR_TOKEN_CLASSIFICATION_MAPPINGr  r   r   r   r  r    r  r   r  ztoken classificationc                       e Zd ZeZy)AutoModelForMultipleChoiceN)r   r   r   !MODEL_FOR_MULTIPLE_CHOICE_MAPPINGr  r   r   r   r  r    r  r   r  zmultiple choicec                       e Zd ZeZy)"AutoModelForNextSentencePredictionN)r   r   r   *MODEL_FOR_NEXT_SENTENCE_PREDICTION_MAPPINGr  r   r   r   r  r    r  r   r  znext sentence predictionc                       e Zd ZeZy)AutoModelForImageClassificationN)r   r   r   &MODEL_FOR_IMAGE_CLASSIFICATION_MAPPINGr  r   r   r   r  r    r  r   r  zimage classificationc                       e Zd ZeZy)'AutoModelForZeroShotImageClassificationN)r   r   r   0MODEL_FOR_ZERO_SHOT_IMAGE_CLASSIFICATION_MAPPINGr  r   r   r   r  r    s    ENr   r  zzero-shot image classificationc                       e Zd ZeZy)AutoModelForImageSegmentationN)r   r   r   $MODEL_FOR_IMAGE_SEGMENTATION_MAPPINGr  r   r   r   r  r    r  r   r  zimage segmentationc                       e Zd ZeZy) AutoModelForSemanticSegmentationN)r   r   r   'MODEL_FOR_SEMANTIC_SEGMENTATION_MAPPINGr  r   r   r   r  r        <Nr   r  zsemantic segmentationc                       e Zd ZeZy) AutoModelForTimeSeriesPredictionN)r   r   r   (MODEL_FOR_TIME_SERIES_PREDICTION_MAPPINGr  r   r   r   r  r  	      =Nr   r  ztime-series predictionc                       e Zd ZeZy)!AutoModelForUniversalSegmentationN)r   r   r   (MODEL_FOR_UNIVERSAL_SEGMENTATION_MAPPINGr  r   r   r   r  r  
	  r  r   r  zuniversal image segmentationc                       e Zd ZeZy) AutoModelForInstanceSegmentationN)r   r   r   'MODEL_FOR_INSTANCE_SEGMENTATION_MAPPINGr  r   r   r   r  r  	  r  r   r  zinstance segmentationc                       e Zd ZeZy)AutoModelForObjectDetectionN)r   r   r   "MODEL_FOR_OBJECT_DETECTION_MAPPINGr  r   r   r   r  r  	      7Nr   r  zobject detectionc                       e Zd ZeZy)#AutoModelForZeroShotObjectDetectionN)r   r   r   ,MODEL_FOR_ZERO_SHOT_OBJECT_DETECTION_MAPPINGr  r   r   r   r  r  #	      ANr   r  zzero-shot object detectionc                       e Zd ZeZy)AutoModelForDepthEstimationN)r   r   r   "MODEL_FOR_DEPTH_ESTIMATION_MAPPINGr  r   r   r   r  r  ,	  r  r   r  zdepth estimationc                       e Zd ZeZy)AutoModelForTextRecognitionN)r   r   r   "MODEL_FOR_TEXT_RECOGNITION_MAPPINGr  r   r   r   r	  r	  3	  r  r   r	  ztext recognitionc                       e Zd ZeZy)AutoModelForTableRecognitionN)r   r   r   #MODEL_FOR_TABLE_RECOGNITION_MAPPINGr  r   r   r   r	  r	  :	  r  r   r	  ztable recognitionc                       e Zd ZeZy)AutoModelForVideoClassificationN)r   r   r   &MODEL_FOR_VIDEO_CLASSIFICATION_MAPPINGr  r   r   r   r	  r	  A	  r  r   r	  zvideo classificationc                   `     e Zd ZeZeded    deej                  e   z  ddf fd       Z
 xZS )AutoModelForImageTextToTextr  r  r  r   c                 *    t        |   |g|i |S r  r  r  s       r   r  z+AutoModelForImageTextToText.from_pretrainedL	  r  r   )r   r   r   $MODEL_FOR_IMAGE_TEXT_TO_TEXT_MAPPINGr  r  r  r  r  r  r  r  r  s   @r   r
	  r
	  H	  sO    9N ]/0]'*R[[-='=]
 
"] ]r   r
	  zimage-text-to-text modelingc                       e Zd ZeZy)AutoModelForMultimodalLMN)r   r   r   MODEL_FOR_MULTIMODAL_LM_MAPPINGr  r   r   r   r	  r	  Y	  r  r   r	  zmultimodal generationc                       e Zd ZeZy)AutoModelForAudioClassificationN)r   r   r   &MODEL_FOR_AUDIO_CLASSIFICATION_MAPPINGr  r   r   r   r	  r	  `	  r  r   r	  zaudio classificationc                       e Zd ZeZy)AutoModelForCTCN)r   r   r   MODEL_FOR_CTC_MAPPINGr  r   r   r   r	  r	  g	      *Nr   r	  z%connectionist temporal classificationc                       e Zd ZeZy)AutoModelForRNNTN)r   r   r   MODEL_FOR_RNNT_MAPPINGr  r   r   r   r	  r	  n	  s    +Nr   r	  zRNN transducerc                       e Zd ZeZy)AutoModelForTDTN)r   r   r   MODEL_FOR_TDT_MAPPINGr  r   r   r   r	  r	  u	  r	  r   r	  ztoken-and-duration transducerc                       e Zd ZeZy)AutoModelForSpeechSeq2SeqN)r   r   r   "MODEL_FOR_SPEECH_SEQ_2_SEQ_MAPPINGr  r   r   r   r	  r	  |	  r  r   r	  z,sequence-to-sequence speech-to-text modelingc                       e Zd ZeZy)$AutoModelForAudioFrameClassificationN)r   r   r   ,MODEL_FOR_AUDIO_FRAME_CLASSIFICATION_MAPPINGr  r   r   r   r!	  r!	  	  r  r   r!	  z"audio frame (token) classificationc                       e Zd ZeZy)AutoModelForAudioXVectorN)r   r   r   MODEL_FOR_AUDIO_XVECTOR_MAPPINGr  r   r   r   r$	  r$	  	  r  r   r$	  c                       e Zd ZeZy)AutoModelForTextToSpectrogramN)r   r   r   %MODEL_FOR_TEXT_TO_SPECTROGRAM_MAPPINGr  r   r   r   r'	  r'	  	      :Nr   r'	  c                       e Zd ZeZy)AutoModelForTextToWaveformN)r   r   r   "MODEL_FOR_TEXT_TO_WAVEFORM_MAPPINGr  r   r   r   r+	  r+	  	  r  r   r+	  c                       e Zd ZeZy)AutoBackboneN)r   r   r   MODEL_FOR_BACKBONE_MAPPINGr  r   r   r   r.	  r.	  	  s    /Nr   r.	  zaudio retrieval via x-vectorc                       e Zd ZeZy)AutoModelForMaskedImageModelingN)r   r   r   'MODEL_FOR_MASKED_IMAGE_MODELING_MAPPINGr  r   r   r   r1	  r1	  	  r  r   r1	  zmasked image modelingc                       e Zd ZeZy)AutoModelForAudioTokenizationN)r   r   r   $MODEL_FOR_AUDIO_TOKENIZATION_MAPPINGr  r   r   r   r4	  r4	  	  r  r   r4	  z$audio tokenization through codebooksc                       e Zd ZeZy)AutoModelForImageMattingN)r   r   r   MODEL_FOR_IMAGE_MATTING_MAPPINGr  r   r   r   r7	  r7	  	  r  r   r7	  zimage mattingc                       e Zd ZeZy)AutoModelForNormalEstimationN)r   r   r   #MODEL_FOR_NORMAL_ESTIMATION_MAPPINGr  r   r   r   r:	  r:	  	  r  r   r:	  znormal estimationc                       e Zd ZeZy)AutoModelForPointmapEstimationN)r   r   r   %MODEL_FOR_POINTMAP_ESTIMATION_MAPPINGr  r   r   r   r=	  r=	  	  r)	  r   r=	  zpointmap estimationc                       e Zd ZeZy)AutoModelForPoseEstimationN)r   r   r   !MODEL_FOR_POSE_ESTIMATION_MAPPINGr  r   r   r   r@	  r@	  	  r  r   r@	  zpose estimation)gr	  r"	  r5	  r%	  r/	  'MODEL_FOR_CAUSAL_IMAGE_MODELING_MAPPINGr  r	  r	  r	  r  r  r;	  r>	  r8	  r	  r	  r  MODEL_FOR_IMAGE_MAPPINGr  r  rA	  r  r  r  r2	  r  r  r  r  r  r  r  r  r  r  r	  r  r  r,	  r(	  r  r  r  r	  MODEL_FOR_RETRIEVAL_MAPPINGr	  r	  r  r  r  r  ,MODEL_FOR_TIME_SERIES_CLASSIFICATION_MAPPING(MODEL_FOR_TIME_SERIES_REGRESSION_MAPPINGr  r.	  r	  r!	  r4	  r$	  r  r	  r	  r	  r  r	  r	  r  r  r  r  r  r  r  r  r1	  r  r  r	  r  r  r  r  r  r  r  r	  r  r'	  r+	  r  r  r  r	  r  r  r  r  r
	  r7	  r:	  r=	  r@	  )__doc__r  collectionsr   typingr   utilsr   auto_factoryr   r	   r
   r   configuration_autor   
generationr   modeling_utilsr   r   
get_loggerr   loggerMODEL_MAPPING_NAMES#MODEL_FOR_PRETRAINING_MAPPING_NAMES!MODEL_FOR_CAUSAL_LM_MAPPING_NAMESMODEL_FOR_IMAGE_MAPPING_NAMES-MODEL_FOR_MASKED_IMAGE_MODELING_MAPPING_NAMES-MODEL_FOR_CAUSAL_IMAGE_MODELING_MAPPING_NAMES,MODEL_FOR_IMAGE_CLASSIFICATION_MAPPING_NAMES*MODEL_FOR_IMAGE_SEGMENTATION_MAPPING_NAMES-MODEL_FOR_SEMANTIC_SEGMENTATION_MAPPING_NAMES-MODEL_FOR_INSTANCE_SEGMENTATION_MAPPING_NAMES.MODEL_FOR_UNIVERSAL_SEGMENTATION_MAPPING_NAMES,MODEL_FOR_VIDEO_CLASSIFICATION_MAPPING_NAMES!MODEL_FOR_RETRIEVAL_MAPPING_NAMES*MODEL_FOR_IMAGE_TEXT_TO_TEXT_MAPPING_NAMESlistitems%MODEL_FOR_MULTIMODAL_LM_MAPPING_NAMES!MODEL_FOR_MASKED_LM_MAPPING_NAMES(MODEL_FOR_OBJECT_DETECTION_MAPPING_NAMES2MODEL_FOR_ZERO_SHOT_OBJECT_DETECTION_MAPPING_NAMES(MODEL_FOR_DEPTH_ESTIMATION_MAPPING_NAMES(MODEL_FOR_TEXT_RECOGNITION_MAPPING_NAMES)MODEL_FOR_TABLE_RECOGNITION_MAPPING_NAMES,MODEL_FOR_SEQ_TO_SEQ_CAUSAL_LM_MAPPING_NAMES(MODEL_FOR_SPEECH_SEQ_2_SEQ_MAPPING_NAMES/MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING_NAMES*MODEL_FOR_QUESTION_ANSWERING_MAPPING_NAMES0MODEL_FOR_TABLE_QUESTION_ANSWERING_MAPPING_NAMES1MODEL_FOR_VISUAL_QUESTION_ANSWERING_MAPPING_NAMES3MODEL_FOR_DOCUMENT_QUESTION_ANSWERING_MAPPING_NAMES,MODEL_FOR_TOKEN_CLASSIFICATION_MAPPING_NAMES'MODEL_FOR_MULTIPLE_CHOICE_MAPPING_NAMES0MODEL_FOR_NEXT_SENTENCE_PREDICTION_MAPPING_NAMES,MODEL_FOR_AUDIO_CLASSIFICATION_MAPPING_NAMESMODEL_FOR_CTC_MAPPING_NAMESMODEL_FOR_RNNT_MAPPING_NAMESMODEL_FOR_TDT_MAPPING_NAMES2MODEL_FOR_AUDIO_FRAME_CLASSIFICATION_MAPPING_NAMES%MODEL_FOR_AUDIO_XVECTOR_MAPPING_NAMES+MODEL_FOR_TEXT_TO_SPECTROGRAM_MAPPING_NAMES(MODEL_FOR_TEXT_TO_WAVEFORM_MAPPING_NAMES6MODEL_FOR_ZERO_SHOT_IMAGE_CLASSIFICATION_MAPPING_NAMES MODEL_FOR_BACKBONE_MAPPING_NAMES'MODEL_FOR_MASK_GENERATION_MAPPING_NAMES*MODEL_FOR_KEYPOINT_DETECTION_MAPPING_NAMES)MODEL_FOR_KEYPOINT_MATCHING_MAPPING_NAMES%MODEL_FOR_TEXT_ENCODING_MAPPING_NAMES2MODEL_FOR_TIME_SERIES_CLASSIFICATION_MAPPING_NAMES.MODEL_FOR_TIME_SERIES_REGRESSION_MAPPING_NAMES.MODEL_FOR_TIME_SERIES_PREDICTION_MAPPING_NAMES&MODEL_FOR_IMAGE_TO_IMAGE_MAPPING_NAMES"MODEL_FOR_AUDIO_TOKENIZATION_NAMES%MODEL_FOR_IMAGE_MATTING_MAPPING_NAMES)MODEL_FOR_NORMAL_ESTIMATION_MAPPING_NAMES+MODEL_FOR_POINTMAP_ESTIMATION_MAPPING_NAMES'MODEL_FOR_POSE_ESTIMATION_MAPPING_NAMESr  r  r  rB	  r  r  r  r  r  r  r	  r	  r	  rD	  r  r  r  rC	  r2	  r  r  r  r	  r	  r  r  r  r  r  r  r  r	  r	  r	  r	  r	  r"	  r%	  r(	  r,	  r/	  r  r  r  r  rE	  rF	  r  r  r5	  r8	  r;	  r>	  rA	  r  r  r  r  r  r  r  r  r  r  r  r  r  r  r  r  r  r  r  r  r  r  r  r  r  r  r  r  r	  r	  r	  r
	  r	  r	  r	  r	  r	  r	  r!	  r$	  r'	  r+	  r.	  r1	  r4	  r7	  r:	  r=	  r@	  __all__r   r   r   <module>r	     s8    	 #     5 -1/  
		H	%!HJ X '2bd' #L %0km% !^ !,<>! @ 1<1 - 1< 	71 - 0;9;0 ,z .9 	(. * 1<1 -  1< 	<4	1 - 2=	2 . 0;0 , %08*% ! .9PR. *l )4	8>>@	A4 	D 	M	
 	P 	9 	@ 	> 	: 	C 	B 	7 	H) %& %002% !h ,7, (0 6A	6 2 ,7
, ( ,7, ( -8/1- ) 0;%'0 ,T ,7, (. 3>z|3 /| .9OQ. *f 4? 	/4 0 5@5 1 7B7 3 0;ac0 ,J +6%'+ 'R 4?4 0 0;0 ,$ * $  +   * 	+  6A
6 2 )4
) % /: 	?// + ,7, (" :E: 6$ $/"$$  L +6	+ ' .98. * -8- ) )4!#) %J 6AC16 2 2=5-2 . 2=062 . *55* & &1& " )4//) % -836- ) /:5/ + +61/+ ' !!57JK 01EGj k ./CEfg *:G+ ' *:F* & 4DP4 0 (8D( $ +;G+ ' +;G+ ' ,<H, ( *:F* & (8D( $ #33GIn"o ./CEfg .>K/ + 1AM1 - //CEfg *+?A^_ *:G+ ' &66JLt%u "/?L0 , &66JLt%u "%56JLt%u "&67KMv&w #)9F* & -=I- ) (8D( $ .>J. * *:F* & %55IKr$s !-=J. * *:F* & ))=?Z[ )*>@\] ()=?Z[ %56JLt%u "/?L0 , #33GIn"o (8E) % &66JLt%u "-.BDde $45IKr$s !'7D( $ '77KMv&w #"23GIn"o /?L0 , ,<H, ( ,<H, ( $44HJp#q  '78LNp'q $"23GIn"o &67KMv&w #(8E) % %55IKr$s !7!4 7:$7 :9#6 952 562 6## # i(	31 3 ,,Cm\ ]. ] ))=Hbc 1. 1 ))=Hbc </ < *5. ?)< ? &7&1J& "
:$7 : !22OZn o @)< @ &7&'<& "A*= A '8'(<' #C,? C ):)*L) %<&9 < #44S^t"u 7!4 7 //ITef @)< @ &7&1K& "
<&9 < #44S^t"u F.A F +<+6V+ '
:$7 : !22OZn o =': = $5$/F$  
>': > $5$/G$  
>(; > %6%0N% !
=': = $5$/F$  
8"5 8 00KVhi B*= B '8'2N' #
8"5 8 00KVhi 8"5 8 00KVhi 9#6 9  11MXkl <&9 < #44S^t"u ]"5 ] 00KVst 52 5 --EPgh <&9 < #44S^t"u +) + $O>ef,* , %%5@PQ +) + $O>]^8 3 8 .(V 
B+> B (9(3W( $
52 5;$7 ;8!4 80) 0 --EPno =&9 = #44S^u"v :$7 : !2!,R! 
52 5 --EP_` 9#6 9  11MXkl ;%8 ; "33Q\q!r 7!4 7 //ITef hr   