
    qiu                       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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$d%d&      Z' eg d'      Z( eg d(      Z) eg d)      Z* eg d*      Z+ eg d+      Z, eg d,      Z- eg d-      Z. eg d.      Z/ ed/g      Z0 eg d0      Z1 eg d1      Z2 eg d2      Z3 eg d3      Z4 eg d4      Z5 eg d5      Z6 eg d6      Z7 eg d7      Z8 eg d8      Z9 ed9d:g      Z: eg d;      Z; eg d<      Z< eg d=      Z= eg d>      Z> ed?g      Z? eg d@      Z@ eg dA      ZA edBdCg      ZB edDdEg      ZC edFdGg      ZD edHg      ZE eg dI      ZF eee      ZG eee      ZH eee      ZI eee      ZJ eee      ZK eee<      ZL eee      ZM eee      ZN eee       ZO eee!      ZP eee"      ZQ eee$      ZR eee'      ZS eee#      ZT eee1      ZU eee2      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e,      Z] eee.      Z^ eee/      Z_ eee0      Z` eee3      Za eee4      Zb eee5      Zc eee6      Zd eee7      Ze eee-      Zf eee8      Zg eee9      Zh eee:      Zi eee;      Zj eee=      Zk eee>      Zl eee?      Zm eee@      Zn eeeA      Zo eeeB      Zp eeeC      Zq eeeD      Zr eeeE      Zs eeeF      Zt G dJ dKe
      Zu G dL dMe
      Zv G dN dOe
      Zw G dP dQe
      Zx G dR dSe
      Zy G dT dUe
      Zz eez      Zz G dV dWe
      Z{ ee{dXY      Z{ G dZ d[e
      Z| ee|d\Y      Z| G d] d^e
      Z} ee}d_Y      Z} G d` dae
      Z~ ee~dbdcd      Z~ G de dfe
      Z eedgY      Z G dh die
      Z eedjY      Z G dk dle
      Z eedmdnd      Z G do dpe
      Z eedqdrd      Z G ds dte
      Z eedudvd      Z G dw dxe
      Z eedyY      Z G dz d{e
      Z eed|Y      Z G d} d~e
      Z eedY      Z G d de
      Z eedY      Z G d de
      Z eedY      Z G d de
      Z eedY      Z G d de
      Z eedY      Z G d de
      Z eedY      Z G d de
      Z eedY      Z G d de
      Z eedY      Z G d de
      Z eedY      Z G d de
      Z eedY      Z G d de
      Z eedY      Z G d de
      Z eedY      Z G d de
      Z eedY      Z G d de
      Z eedY      Z G d de
      Z eedY      Z G d de
      Z eedY      Z G d de
      Z eedY      Z G d de
      Z eedY      Z G d de
      Z G d de
      Z G d de
      Z G d de	      Z eedY      Z G d de
      Z eedY      Z G d de
      Z eedY      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__     X/opt/pipecat/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&AudioFlamingo3ForConditionalGeneration)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)
code_llama
LlamaModel)codegenCodeGenModel)cohereCohereModel)cohere2Cohere2Modelcohere2_visionCohere2VisionModelconditional_detrConditionalDetrModel)convbertConvBertModelconvnextConvNextModel
convnextv2ConvNextV2Model)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_v2DeepseekV2Model)deepseek_v3DeepseekV3Model)deepseek_vlDeepseekVLModel)deepseek_vl_hybridDeepseekVLHybridModeldeformable_detrDeformableDetrModeldeit	DeiTModel	depth_proDepthProModeldetr	DetrModel)diaDiaModel)	diffllamaDiffLlamaModel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)
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)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GlmAsrForConditionalGeneration)glmasr_encoderGlmAsrEncoder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)
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)hubertHubertModel)hunyuan_v1_denseHunYuanDenseV1Model)hunyuan_v1_moeHunYuanMoEV1Modelibert
IBertModel)ideficsIdeficsModel)idefics2Idefics2Model)idefics3Idefics3Model)idefics3_visionIdefics3VisionTransformerijepa
IJepaModelimagegptImageGPTModel)informerInformerModel)instructblipInstructBlipModel)instructblipvideoInstructBlipVideoModel)internvlInternVLModel)internvl_visionInternVLVisionModel)jais2
Jais2Model)jamba
JambaModel)janus
JanusModel)jetmoeJetMoeModel)kosmos-2Kosmos2Model)
kosmos-2.5Kosmos2_5Model)kyutai_speech_to_textKyutaiSpeechToTextModel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)llamar   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
metaclip_2MetaClip2Model)zmgp-strMgpstrForSceneTextRecognition)mimi	MimiModel)minimaxMiniMaxModel)
minimax_m2MiniMaxM2Model)	ministralMinistralModel)
ministral3Ministral3Model)mistralMistralModel)mistral3Mistral3Model)mixtralMixtralModelmlcdMLCDVisionModel)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)musicgenMusicgenModel)musicgen_melodyMusicgenMelodyModel)mvpMvpModel)nanochatNanoChatModel)nemotronNemotronModel)
nemotron_hNemotronHModel)nllb-moeNllbMoeModelnystromformerNystromformerModel)olmo	OlmoModel)olmo2
Olmo2Model)olmo3
Olmo3Model)olmo_hybridOlmoHybridModel)olmoe
OlmoeModelzomdet-turboOmDetTurboForObjectDetection)	oneformerOneFormerModel)
openai-gptOpenAIGPTModel)optOPTModel)ovis2
Ovis2Model)owlv2
Owlv2Model)owlvitOwlViTModel)	paligemmaPaliGemmaModelparakeet_ctcParakeetForCTC)parakeet_encoderParakeetEncoder)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pixio
PixioModel)pixtralPixtralVisionModel)plbartPLBartModel
poolformerPoolFormerModel)pp_doclayout_v3PPDocLayoutV3Model)
prophetnetProphetNetModelpvtPvtModel)pvt_v2
PvtV2Model)qwen2
Qwen2Model)
qwen2_5_vlQwen2_5_VLModel)qwen2_5_vl_textQwen2_5_VLTextModel)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_textQwen3_5TextModel)	qwen3_moeQwen3MoeModel)
qwen3_nextQwen3NextModel)qwen3_vlQwen3VLModel)qwen3_vl_moeQwen3VLMoeModel)qwen3_vl_moe_textQwen3VLMoeTextModel)qwen3_vl_textQwen3VLTextModel)recurrent_gemmaRecurrentGemmaModelreformerReformerModelregnetRegNetModelrembertRemBertModelresnetResNetModel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_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)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_wrapperrI  )tvpTvpModel)udop	UdopModel)umt5	UMT5Model)	unispeechUniSpeechModel)unispeech-satUniSpeechSatModel)univnetUnivNetModel)
vaultgemmaVaultGemmaModelvibevoice_acoustic_tokenizerVibeVoiceAcousticTokenizerModel)$vibevoice_acoustic_tokenizer_decoder&VibeVoiceAcousticTokenizerDecoderModel)$vibevoice_acoustic_tokenizer_encoder&VibeVoiceAcousticTokenizerEncoderModelvibevoice_asr$VibeVoiceAsrForConditionalGeneration)video_llama_3VideoLlama3Model)video_llama_3_visionVideoLlama3VisionModel)video_llavaVideoLlavaModelvideomaeVideoMAEModel)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VoxtralForConditionalGeneration)voxtral_encoderVoxtralEncodervoxtral_realtime'VoxtralRealtimeForConditionalGeneration)voxtral_realtime_encoderVoxtralRealtimeEncoder)voxtral_realtime_textVoxtralRealtimeTextModel)wav2vec2Wav2Vec2Model)wav2vec2-bertWav2Vec2BertModel)wav2vec2-conformerWav2Vec2ConformerModel)wavlm
WavLMModel)whisperWhisperModel)xclip
XCLIPModel)xcodecXcodecModel)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)Z)r   AlbertForPreTrainingr1   r?   BartForConditionalGeneration)rE   BertForPreTraining)rJ   BigBirdForPreTrainingr`   BloomForCausalLMrh   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r&  FlaubertWithLMHeadModel)r(  FlavaForPreTrainingr,  !Florence2ForConditionalGeneration)r.  FNetForPreTrainingr3  FSMTForConditionalGeneration)r5  FunnelForPreTrainingr>  Gemma3ForConditionalGenerationrr  r|  GPT2LMHeadModelr~  r$  r  GPTBigCodeForCausalLM)r  HieraForPreTrainingr  IBertForMaskedLMr  IdeficsForVisionText2Textr   Idefics2ForConditionalGenerationr   Idefics3ForConditionalGenerationr  JanusForConditionalGenerationr  LayoutLMForMaskedLMr  LlavaForConditionalGenerationr  !LlavaNextForConditionalGenerationr  &LlavaNextVideoForConditionalGenerationr  &LlavaOnevisionForConditionalGenerationr  LongformerForMaskedLMr  LukeForMaskedLM)r
  LxmertForPreTrainingr  MambaForCausalLMr  Mamba2ForCausalLM)r  MegatronBertForPreTrainingr0   Mistral3ForConditionalGenerationr7  MllamaForConditionalGeneration)r<  MobileBertForPreTrainingrV  MPNetForMaskedLMrX  MptForCausalLMrZ  MraForMaskedLMrb  MvpForConditionalGenerationrd  NanoChatForCausalLMrj  NllbMoeForConditionalGenerationr}  OpenAIGPTLMHeadModelr  !PaliGemmaForConditionalGenerationqwen2_audio"Qwen2AudioForConditionalGenerationr  RobertaForMaskedLMr  RobertaPreLayerNormForMaskedLM)r  RoCBertForPreTrainingr  RwkvForCausalLM)rM  SplinterForPreTrainingrP  SqueezeBertForMaskedLMrb  *SwitchTransformersForConditionalGenerationrd  T5ForConditionalGenerationrf  T5GemmaForConditionalGenerationrh   T5Gemma2ForConditionalGenerationro  TapasForMaskedLM)r  UniSpeechForPreTraining)r  UniSpeechSatForPreTrainingr  r  "VideoLlavaForConditionalGeneration)r  VideoMAEForPreTrainingr   VipLlavaForConditionalGeneration)r  VisualBertForPreTraining)r  ViTMAEForPreTrainingr  r  )r  Wav2Vec2ForPreTraining)r  Wav2Vec2ConformerForPreTrainingr  XLMWithLMHeadModelr  XLMRobertaForMaskedLMr  XLMRobertaXLForMaskedLMr  XLNetLMHeadModelr  xLSTMForCausalLMr  XmodForMaskedLM))r   AfmoeForCausalLM)r'   ApertusForCausalLM)r)   ArceeForCausalLM)r-   AriaTextForCausalLM)r:   BambaForCausalLM)r?   BartForCausalLM)rE   BertLMHeadModel)rG   BertGenerationDecoder)rJ   BigBirdForCausalLM)rL   BigBirdPegasusForCausalLM)rN   BioGptForCausalLM)rS   BitNetForCausalLM)rU   BlenderbotForCausalLM)rW   BlenderbotSmallForCausalLMr  )rb   BltForCausalLM)rh   CamembertForCausalLM)r   LlamaForCausalLM)r   CodeGenForCausalLM)r   CohereForCausalLM)r   Cohere2ForCausalLM)r   CpmAntForCausalLMr  )r   CwmForCausalLM)r   Data2VecTextForCausalLM)r   DbrxForCausalLM)r   DeepseekV2ForCausalLM)r   DeepseekV3ForCausalLM)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r8  FuyuForCausalLM)r:  GemmaForCausalLM)r<  Gemma2ForCausalLMr!  )r@  Gemma3ForCausalLMrB  Gemma3nForConditionalGeneration)rF  Gemma3nForCausalLMrJ  GitForCausalLM)rL  GlmForCausalLM)rN  Glm4ForCausalLM)rR  Glm4MoeForCausalLM)rT  Glm4MoeLiteForCausalLM)rj  GlmMoeDsaForCausalLMrz  GotOcr2ForConditionalGenerationr#  r%  r&  )r  GPTNeoForCausalLM)r  GPTNeoXForCausalLM)r  GPTNeoXJapaneseForCausalLM)r  GptOssForCausalLM)r  GPTJForCausalLM)r  GraniteForCausalLM)r  GraniteMoeForCausalLM)r  GraniteMoeHybridForCausalLM)r  GraniteMoeSharedForCausalLM)r  HeliumForCausalLM)r  HunYuanDenseV1ForCausalLM)r  HunYuanMoEV1ForCausalLM)r  Jais2ForCausalLM)r  JambaForCausalLM)r  JetMoeForCausalLM)r  Lfm2ForCausalLM)r  Lfm2MoeForCausalLM)r  r  )r  Llama4ForCausalLM)r  r  )r  LongcatFlashForCausalLMrB  rD  )r  MarianForCausalLM)r  MBartForCausalLM)r  MegatronBertForCausalLM)r&  MiniMaxForCausalLM)r(  MiniMaxM2ForCausalLM)r*  MinistralForCausalLM)r,  Ministral3ForCausalLM)r.  MistralForCausalLM)r2  MixtralForCausalLM)r7  MllamaForCausalLM)rL  ModernBertDecoderForCausalLM)rT  MoshiForCausalLMrN  )r^  MusicgenForCausalLM)r`  MusicgenMelodyForCausalLM)rb  MvpForCausalLMrT  )rf  NemotronForCausalLM)rh  NemotronHForCausalLM)ro  OlmoForCausalLM)rq  Olmo2ForCausalLM)rs  Olmo3ForCausalLM)ru  OlmoHybridForCausalLM)rw  OlmoeForCausalLMrX  )r  OPTForCausalLM)r  PegasusForCausalLM)r  PersimmonForCausalLM)r  PhiForCausalLM)r  Phi3ForCausalLMr  Phi4MultimodalForCausalLM)r  PhimoeForCausalLM)r  PLBartForCausalLM)r  ProphetNetForCausalLM)r  Qwen2ForCausalLM)r  Qwen2MoeForCausalLM)r  Qwen3ForCausalLM)r  Qwen3_5ForCausalLM)r  Qwen3_5MoeForCausalLM)r  r  )r  r  )r  Qwen3MoeForCausalLM)r  Qwen3NextForCausalLM)r  RecurrentGemmaForCausalLM)r  ReformerModelWithLMHead)r  RemBertForCausalLM)r  RobertaForCausalLM)r  RobertaPreLayerNormForCausalLM)r  RoCBertForCausalLM)r  RoFormerForCausalLMrd  )r+  SeedOssForCausalLM)rA  SmolLM3ForCausalLM)rG  SolarOpenForCausalLM)rR  StableLmForCausalLM)rT  Starcoder2ForCausalLM)trocrTrOCRForCausalLM)r  VaultGemmaForCausalLM)r  WhisperForCausalLM)r  XGLMForCausalLMr~  )r  XLMRobertaForCausalLM)r  XLMRobertaXLForCausalLMr  r  )r  XmodForCausalLM)r  YoutuForCausalLM)r  ZambaForCausalLM)r  Zamba2ForCausalLM)7r   rA   rP   r   r   r   r   r   r   r   r   r   r   r   r   r   r   r   r   r   r0  rw  r  r  r  r  )r  Llama4VisionModelr4  )r7  MllamaVisionModelr>  rA  rD  rG  r  r  r  r  r  r-  r>  rV  rY  r\  r_  rl  ry  r|  r  r  r  r  r  r  r  r  ))r   DeiTForMaskedImageModeling)r1  FocalNetForMaskedImageModeling)rZ  SwinForMaskedImageModeling)r`  Swinv2ForMaskedImageModeling)r  ViTForMaskedImageModeling)r  ImageGPTForCausalImageModeling)))rB   BeitForImageClassification)rQ   BitForImageClassification)rv   CLIPForImageClassification)r   ConvNextForImageClassification)r    ConvNextV2ForImageClassification)r   CvtForImageClassification)r   $Data2VecVisionForImageClassification)r   )DeiTForImageClassification%DeiTForImageClassificationWithTeacher)r   DinatForImageClassification)r   Dinov2ForImageClassification)r   )Dinov2WithRegistersForImageClassification)r   DonutSwinForImageClassification)r   "EfficientNetForImageClassification)r1  FocalNetForImageClassification)r  HGNetV2ForImageClassification)r  HieraForImageClassification)r  IJepaForImageClassification)r  ImageGPTForImageClassification)r  )LevitForImageClassification&LevitForImageClassificationWithTeacher)r!  MetaClip2ForImageClassification)r?  !MobileNetV1ForImageClassification)rB  !MobileNetV2ForImageClassification)rE  MobileViTForImageClassification)rH  !MobileViTV2ForImageClassification)r  )&PerceiverForImageClassificationLearned&PerceiverForImageClassificationFourier-PerceiverForImageClassificationConvProcessing)r   PoolFormerForImageClassification)r  PvtForImageClassification)r  PvtV2ForImageClassification)r  RegNetForImageClassification)r  ResNetForImageClassification)r.  SegformerForImageClassification)shieldgemma2"ShieldGemma2ForImageClassification)r7  SiglipForImageClassification)r:  Siglip2ForImageClassification)rW  !SwiftFormerForImageClassification)rZ  SwinForImageClassification)r`  Swinv2ForImageClassification)rq  TextNetForImageClassification)r  !TimmWrapperForImageClassification)r  ViTForImageClassification)r  ViTMSNForImageClassificationr   DetrForSegmentation))rB   BeitForSemanticSegmentation)r   %Data2VecVisionForSemanticSegmentation)r   DPTForSemanticSegmentation)rB  "MobileNetV2ForSemanticSegmentation)rE   MobileViTForSemanticSegmentation)rH  "MobileViTV2ForSemanticSegmentation)r.   SegformerForSemanticSegmentation)upernetUperNetForSemanticSegmentationr  !MaskFormerForInstanceSegmentation)rG  )eomtEomtForUniversalSegmentation)eomt_dinov3"EomtDinov3ForUniversalSegmentation)r  #Mask2FormerForUniversalSegmentationrR  )r{  !OneFormerForUniversalSegmentation))rz  !TimesformerForVideoClassification)r  VideoMAEForVideoClassification)r  VivitForVideoClassification)r  VJEPA2ForVideoClassificationr  r  )<)r+   AriaForConditionalGeneration)r8   !AyaVisionForConditionalGeneration)rZ   BlipForConditionalGenerationr\   Blip2ForConditionalGeneration)rl   !ChameleonForConditionalGeneration)r   %Cohere2VisionForConditionalGeneration)r   "DeepseekVLForConditionalGeneration)r   (DeepseekVLHybridForConditionalGeneration)r  Emu3ForConditionalGeneration)r  &Ernie4_5_VLMoeForConditionalGenerationr  )r  FastVlmForConditionalGenerationr  r  r!  r  r  )rP  Glm46VForConditionalGeneration)rV  Glm4vForConditionalGeneration)rX   Glm4vMoeForConditionalGeneration)rl  GlmOcrForConditionalGenerationr  r+  r-  r/  )r  $InstructBlipForConditionalGeneration)r  )InstructBlipVideoForConditionalGeneration)r   InternVLForConditionalGenerationr1  )r  Kosmos2ForConditionalGeneration)r  !Kosmos2_5ForConditionalGeneration)r  Lfm2VlForConditionalGeneration)r  "LightOnOcrForConditionalGenerationr  r5  r7  r9  r;  rG  rI  )r  Ovis2ForConditionalGeneration)paddleocr_vl#PaddleOCRVLForConditionalGenerationrZ  )r  $PerceptionLMForConditionalGeneration)
pix2struct"Pix2StructForConditionalGeneration)r  r6  )r  "Qwen2_5_VLForConditionalGeneration)r  Qwen2VLForConditionalGeneration)r  Qwen3_5ForConditionalGeneration)r  "Qwen3_5MoeForConditionalGeneration)r  Qwen3VLForConditionalGeneration)r  "Qwen3VLMoeForConditionalGeneration)r<  r"  )rC  SmolVLMForConditionalGenerationro  )r  UdopForConditionalGeneration)r  #VideoLlama3ForConditionalGenerationru  rx  )zvision-encoder-decoderVisionEncoderDecoderModelrr  granite_speech%GraniteSpeechForConditionalGenerationr  *KyutaiSpeechToTextForConditionalGenerationr  qwen2_5_omni#Qwen2_5OmniForConditionalGenerationr\  qwen3_omni_moe$Qwen3OmniMoeForConditionalGenerationr  r  r  ),)r   AlbertForMaskedLMr  )rE   BertForMaskedLM)rJ   BigBirdForMaskedLMr  )r   ConvBertForMaskedLMr  r  r
  r  )r   ElectraForMaskedLM)r  ErnieForMaskedLM)r  EsmForMaskedLM)r  EuroBertForMaskedLMr  )r.  FNetForMaskedLM)r5  FunnelForMaskedLMr)  r3  r=  r?  r  MBartForConditionalGeneration)r  MegatronBertForMaskedLM)r<  MobileBertForMaskedLM)rJ  ModernBertForMaskedLM)rN  ModernVBertForMaskedLMrL  rP  rR  )rm  NystromformerForMaskedLM)r  PerceiverForMaskedLM)r  ReformerForMaskedLM)r  RemBertForMaskedLMr_  ra  )r  RoCBertForMaskedLM)r  RoFormerForMaskedLMrg  rq  r~  r  r  r  )r  YosoForMaskedLM))r   !ConditionalDetrForObjectDetection)r   DFineForObjectDetection)r   DabDetrForObjectDetection)r    DeformableDetrForObjectDetection)r   DetrForObjectDetection)r  LwDetrForObjectDetection)pp_doclayout_v2PPDocLayoutV2ForObjectDetection)r  PPDocLayoutV3ForObjectDetection)r  RTDetrForObjectDetection)r  RTDetrV2ForObjectDetection)rm  "TableTransformerForObjectDetection)r  YolosForObjectDetection))r  GroundingDinoForObjectDetection)r9  !MMGroundingDinoForObjectDetectionry  )r  Owlv2ForObjectDetection)r  OwlViTForObjectDetection))depth_anythingDepthAnythingForDepthEstimation)r   DepthProForDepthEstimation)r   DPTForDepthEstimation)rx  GLPNForDepthEstimation)prompt_depth_anything%PromptDepthAnythingForDepthEstimation)zoedepthZoeDepthForDepthEstimation) r1   r  )rL   &BigBirdPegasusForConditionalGeneration)rU   "BlenderbotForConditionalGeneration)rW   'BlenderbotSmallForConditionalGeneration)zencoder-decoderEncoderDecoderModelr  rr  r  )r  LEDForConditionalGeneration)r  LongT5ForConditionalGeneration)r  M2M100ForConditionalGeneration)r  MarianMTModelr  )r\  MT5ForConditionalGenerationrR  rV  )r  PegasusForConditionalGeneration)r   PegasusXForConditionalGeneration)r  PLBartForConditionalGeneration)r  "ProphetNetForConditionalGenerationr\  )r'  SeamlessM4TForTextToText)r)  SeamlessM4Tv2ForTextToTextri  rk  rm  ro  )r  UMT5ForConditionalGenerationr  r  r  ))r   DiaForConditionalGenerationr  r  )rP  !MoonshineForConditionalGeneration)rR  *MoonshineStreamingForConditionalGeneration)	pop2piano!Pop2PianoForConditionalGeneration)r'  SeamlessM4TForSpeechToText)r)  SeamlessM4Tv2ForSpeechToText)zspeech-encoder-decoderSpeechEncoderDecoderModel)rI  #Speech2TextForConditionalGeneration)rK  SpeechT5ForSpeechToTextr  r  r  )r  WhisperForConditionalGeneration)r)r   AlbertForSequenceClassification)r)   ArceeForSequenceClassification)r?   BartForSequenceClassification)rE   BertForSequenceClassification)rJ    BigBirdForSequenceClassification)rL   'BigBirdPegasusForSequenceClassification)rN   BioGptForSequenceClassification)r`   BloomForSequenceClassification)rh   "CamembertForSequenceClassification)rj   CanineForSequenceClassification)r   LlamaForSequenceClassification)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)r&  !FlaubertForSequenceClassification)r.  FNetForSequenceClassification)r5  FunnelForSequenceClassification)r:  GemmaForSequenceClassification)r<  Gemma2ForSequenceClassification)r>  Gemma3ForSequenceClassification)r@  #Gemma3TextForSequenceClassification)rL  GlmForSequenceClassification)rN  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  !LayoutLMForSequenceClassification)r  #LayoutLMv2ForSequenceClassification)r  #LayoutLMv3ForSequenceClassification)r  LiltForSequenceClassification)r  r  )r  #LongformerForSequenceClassification)r  LukeForSequenceClassification)r  !MarkupLMForSequenceClassification)r  MBartForSequenceClassification)r  %MegatronBertForSequenceClassification)r&   MiniMaxForSequenceClassification)r*  "MinistralForSequenceClassification)r,  #Ministral3ForSequenceClassification)r.   MistralForSequenceClassification)r2   MixtralForSequenceClassification)r<  #MobileBertForSequenceClassification)rJ  #ModernBertForSequenceClassification)rL  *ModernBertDecoderForSequenceClassification)rN  $ModernVBertForSequenceClassification)rV  MPNetForSequenceClassification)rX  MptForSequenceClassification)rZ  MraForSequenceClassification)r\  MT5ForSequenceClassification)rb  MvpForSequenceClassification)rf  !NemotronForSequenceClassification)rm  &NystromformerForSequenceClassification)r}  "OpenAIGPTForSequenceClassification)r  OPTForSequenceClassification)r  "PerceiverForSequenceClassification)r  "PersimmonForSequenceClassification)r  PhiForSequenceClassification)r  Phi3ForSequenceClassification)r  PhimoeForSequenceClassification)r  PLBartForSequenceClassification)r  Qwen2ForSequenceClassification)r  !Qwen2MoeForSequenceClassification)r  Qwen3ForSequenceClassification)r   Qwen3_5ForSequenceClassification)r  r0  )r  !Qwen3MoeForSequenceClassification)r  "Qwen3NextForSequenceClassification)r  !ReformerForSequenceClassification)r   RemBertForSequenceClassification)r   RobertaForSequenceClassification)r  ,RobertaPreLayerNormForSequenceClassification)r   RoCBertForSequenceClassification)r  !RoFormerForSequenceClassification)r+   SeedOssForSequenceClassification)rA   SmolLM3ForSequenceClassification)rP  $SqueezeBertForSequenceClassification)rR  !StableLmForSequenceClassification)rT  #Starcoder2ForSequenceClassification)rd  T5ForSequenceClassification)rf   T5GemmaForSequenceClassification)rh  !T5Gemma2ForSequenceClassification)ro  TapasForSequenceClassification)r  UMT5ForSequenceClassification)r  XLMForSequenceClassification)r  #XLMRobertaForSequenceClassification)r  %XLMRobertaXLForSequenceClassification)r  XLNetForSequenceClassification)r  XmodForSequenceClassification)r  YosoForSequenceClassification)r  ZambaForSequenceClassification)r  Zamba2ForSequenceClassification)L)r   AlbertForQuestionAnswering)r)   ArceeForQuestionAnswering)r?   BartForQuestionAnswering)rE   BertForQuestionAnswering)rJ   BigBirdForQuestionAnswering)rL   "BigBirdPegasusForQuestionAnswering)r`   BloomForQuestionAnswering)rh   CamembertForQuestionAnswering)rj   CanineForQuestionAnswering)r   ConvBertForQuestionAnswering)r    Data2VecTextForQuestionAnswering)r   DebertaForQuestionAnswering)r   DebertaV2ForQuestionAnswering)r   DiffLlamaForQuestionAnswering)r   DistilBertForQuestionAnswering)r   ElectraForQuestionAnswering)r  ErnieForQuestionAnswering)r  Exaone4ForQuestionAnswering)r  FalconForQuestionAnswering)r&  "FlaubertForQuestionAnsweringSimple)r.  FNetForQuestionAnswering)r5  FunnelForQuestionAnswering)r~  GPT2ForQuestionAnswering)r  GPTNeoForQuestionAnswering)r  GPTNeoXForQuestionAnswering)r  GPTJForQuestionAnswering)r  IBertForQuestionAnsweringr  LayoutLMv2ForQuestionAnsweringr  LayoutLMv3ForQuestionAnswering)r  LEDForQuestionAnswering)r  LiltForQuestionAnswering)r  LlamaForQuestionAnswering)r  LongformerForQuestionAnswering)r  LukeForQuestionAnswering)r
  LxmertForQuestionAnswering)r  MarkupLMForQuestionAnswering)r  MBartForQuestionAnswering)r   MegatronBertForQuestionAnswering)r&  MiniMaxForQuestionAnswering)r*  MinistralForQuestionAnswering)r,  Ministral3ForQuestionAnswering)r.  MistralForQuestionAnswering)r2  MixtralForQuestionAnswering)r<  MobileBertForQuestionAnswering)rJ  ModernBertForQuestionAnswering)rV  MPNetForQuestionAnswering)rX  MptForQuestionAnswering)rZ  MraForQuestionAnswering)r\  MT5ForQuestionAnswering)rb  MvpForQuestionAnswering)rf  NemotronForQuestionAnswering)rm  !NystromformerForQuestionAnswering)r  OPTForQuestionAnswering)r  Qwen2ForQuestionAnswering)r  Qwen2MoeForQuestionAnswering)r  Qwen3ForQuestionAnswering)r  Qwen3MoeForQuestionAnswering)r  Qwen3NextForQuestionAnswering)r  ReformerForQuestionAnswering)r  RemBertForQuestionAnswering)r  RobertaForQuestionAnswering)r  'RobertaPreLayerNormForQuestionAnswering)r  RoCBertForQuestionAnswering)r  RoFormerForQuestionAnswering)r+  SeedOssForQuestionAnswering)rA  SmolLM3ForQuestionAnswering)rM  SplinterForQuestionAnswering)rP  SqueezeBertForQuestionAnswering)rd  T5ForQuestionAnswering)r  UMT5ForQuestionAnswering)r  XLMForQuestionAnsweringSimple)r  XLMRobertaForQuestionAnswering)r   XLMRobertaXLForQuestionAnswering)r  XLNetForQuestionAnsweringSimple)r  XmodForQuestionAnswering)r  YosoForQuestionAnswering)ro  TapasForQuestionAnswering))rZ   BlipForQuestionAnsweringra  )r  ViltForQuestionAnswering))r  LayoutLMForQuestionAnsweringrf  rh  )Y)r   AlbertForTokenClassification)r'   ApertusForTokenClassification)r)   ArceeForTokenClassification)rE   BertForTokenClassification)rJ   BigBirdForTokenClassification)rN   BioGptForTokenClassification)r`   BloomForTokenClassification)rf   BrosForTokenClassification)rh   CamembertForTokenClassification)rj   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)r&  FlaubertForTokenClassification)r.  FNetForTokenClassification)r5  FunnelForTokenClassification)r:  GemmaForTokenClassification)r<  Gemma2ForTokenClassification)rL  GlmForTokenClassification)rN  Glm4ForTokenClassification)r|  GPT2ForTokenClassification)r~  r  )r   GPTBigCodeForTokenClassification)r  GPTNeoForTokenClassification)r  GPTNeoXForTokenClassification)r  GptOssForTokenClassification)r  HeliumForTokenClassification)r  IBertForTokenClassification)r  LayoutLMForTokenClassification)r   LayoutLMv2ForTokenClassification)r   LayoutLMv3ForTokenClassification)r  LiltForTokenClassification)r  LlamaForTokenClassification)r   LongformerForTokenClassification)r  LukeForTokenClassification)r  MarkupLMForTokenClassification)r  "MegatronBertForTokenClassification)r&  MiniMaxForTokenClassification)r*  MinistralForTokenClassification)r,   Ministral3ForTokenClassification)r.  MistralForTokenClassification)r2  MixtralForTokenClassification)r<   MobileBertForTokenClassification)rJ   ModernBertForTokenClassification)rN  !ModernVBertForTokenClassification)rV  MPNetForTokenClassification)rX  MptForTokenClassification)rZ  MraForTokenClassification)r\  MT5ForTokenClassification)rf  NemotronForTokenClassification)rm  #NystromformerForTokenClassification)r  PersimmonForTokenClassification)r  PhiForTokenClassification)r  Phi3ForTokenClassification)r  Qwen2ForTokenClassification)r  Qwen2MoeForTokenClassification)r  Qwen3ForTokenClassification)r  Qwen3MoeForTokenClassification)r  Qwen3NextForTokenClassification)r  RemBertForTokenClassification)r  RobertaForTokenClassification)r  )RobertaPreLayerNormForTokenClassification)r  RoCBertForTokenClassification)r  RoFormerForTokenClassification)r+  SeedOssForTokenClassification)rA  SmolLM3ForTokenClassification)rP  !SqueezeBertForTokenClassification)rR  StableLmForTokenClassification)rT   Starcoder2ForTokenClassification)rd  T5ForTokenClassification)rf  T5GemmaForTokenClassification)rh  T5Gemma2ForTokenClassification)r  UMT5ForTokenClassification)r  XLMForTokenClassification)r   XLMRobertaForTokenClassification)r  "XLMRobertaXLForTokenClassification)r  XLNetForTokenClassification)r  XmodForTokenClassification)r  YosoForTokenClassification)#)r   AlbertForMultipleChoice)rE   BertForMultipleChoice)rJ   BigBirdForMultipleChoice)rh   CamembertForMultipleChoice)rj   CanineForMultipleChoice)r   ConvBertForMultipleChoice)r   Data2VecTextForMultipleChoice)r   DebertaV2ForMultipleChoice)r   DistilBertForMultipleChoice)r   ElectraForMultipleChoice)r  ErnieForMultipleChoice)r&  FlaubertForMultipleChoice)r.  FNetForMultipleChoice)r5  FunnelForMultipleChoice)r  IBertForMultipleChoice)r  LongformerForMultipleChoice)r  LukeForMultipleChoice)r  MegatronBertForMultipleChoice)r<  MobileBertForMultipleChoice)rJ  ModernBertForMultipleChoice)rV  MPNetForMultipleChoice)rZ  MraForMultipleChoice)rm  NystromformerForMultipleChoice)r  RemBertForMultipleChoice)r  RobertaForMultipleChoice)r  $RobertaPreLayerNormForMultipleChoice)r  RoCBertForMultipleChoice)r  RoFormerForMultipleChoice)rP  SqueezeBertForMultipleChoice)r  XLMForMultipleChoice)r  XLMRobertaForMultipleChoice)r  XLMRobertaXLForMultipleChoice)r  XLNetForMultipleChoice)r  XmodForMultipleChoice)r  YosoForMultipleChoice))rE   BertForNextSentencePrediction)r  ErnieForNextSentencePrediction)r.  FNetForNextSentencePrediction)r  %MegatronBertForNextSentencePrediction)r<  #MobileBertForNextSentencePrediction))r/   ASTForAudioClassification)r   &Data2VecAudioForSequenceClassification)r  HubertForSequenceClassification)r2  SEWForSequenceClassification)r4  SEWDForSequenceClassification)r  "UniSpeechForSequenceClassification)r  %UniSpeechSatForSequenceClassification)r  !Wav2Vec2ForSequenceClassification)r  %Wav2Vec2BertForSequenceClassification)r  *Wav2Vec2ConformerForSequenceClassification)r  WavLMForSequenceClassification)r  WhisperForAudioClassification))r   Data2VecAudioForCTC)r  HubertForCTCr  r  )r2  	SEWForCTC)r4  
SEWDForCTC)r  UniSpeechForCTC)r  UniSpeechSatForCTC)r  Wav2Vec2ForCTC)r  Wav2Vec2BertForCTC)r  Wav2Vec2ConformerForCTC)r  WavLMForCTC))r   (Data2VecAudioForAudioFrameClassification)r  'UniSpeechSatForAudioFrameClassification)r  #Wav2Vec2ForAudioFrameClassification)r  'Wav2Vec2BertForAudioFrameClassification)r  ,Wav2Vec2ConformerForAudioFrameClassification)r   WavLMForAudioFrameClassification))r   Data2VecAudioForXVector)r  UniSpeechSatForXVector)r  Wav2Vec2ForXVector)r  Wav2Vec2BertForXVector)r  Wav2Vec2ConformerForXVector)r  WavLMForXVectorr  )rK  SpeechT5ForTextToSpeech)r<   r   )r   r$  r"  r  )r^   MusicgenForConditionalGeneration)r`  &MusicgenMelodyForConditionalGenerationr  r  )r'  SeamlessM4TForTextToSpeech)r)  SeamlessM4Tv2ForTextToSpeechr  )
r!   r$   rY   )r\   Blip2ForImageTextRetrievalrn   ru   r}   r   r6  r9  ))rB   BeitBackbone)rQ   BitBackbone)r   ConvNextBackbone)r   ConvNextV2Backbone)r   DinatBackbone)r   Dinov2Backbone)r   Dinov2WithRegistersBackbone)r   DINOv3ConvNextBackbone)r   DINOv3ViTBackbone)r1  FocalNetBackboner  )r  HieraBackbone)lw_detr_vitLwDetrViTBackbone)r  MaskFormerSwinBackbone)r  PixioBackbone)r  PvtV2Backbone)r  ResNetBackbone)rt_detr_resnetRTDetrResNetBackbone)rZ  SwinBackbone)r`  Swinv2Backbone)rq  TextNetBackboner|  )r  VitDetBackbone)vitpose_backboneVitPoseBackbone)r   )r   r   r  r
  )r  r  r  )r  r  r   )
superpointSuperPointForKeypointDetection))r   !EfficientLoFTRForKeypointMatchingr  )	superglueSuperGlueForKeypointMatching)r   rD   rI   rx   r   r   r   r   r   )r  Emu3TextModelr%  r  )r  r  r  )r7  MllamaTextModelr;  )r\  MT5EncoderModelrl  r  r  r  r  r  r  rO  )rd  T5EncoderModel)rf  T5GemmaEncoderModel)r  UMT5EncoderModelr  r  r  )r  'PatchTSMixerForTimeSeriesClassification)r  PatchTSTForClassification)r  PatchTSMixerForRegression)r  PatchTSTForRegression)ru  TimesFmModelForPrediction)rw  TimesFm2_5ModelForPrediction)r]  Swin2SRForImageSuperResolution)r   r  r  c                       e Zd ZeZy)AutoModelForMaskGenerationN)r   r   r   !MODEL_FOR_MASK_GENERATION_MAPPING_model_mappingr   r   r   rq  rq  ^      6Nr   rq  c                       e Zd ZeZy)AutoModelForKeypointDetectionN)r   r   r   $MODEL_FOR_KEYPOINT_DETECTION_MAPPINGrs  r   r   r   rv  rv  b      9Nr   rv  c                       e Zd ZeZy)AutoModelForKeypointMatchingN)r   r   r   #MODEL_FOR_KEYPOINT_MATCHING_MAPPINGrs  r   r   r   rz  rz  f  s    8Nr   rz  c                       e Zd ZeZy)AutoModelForTextEncodingN)r   r   r   MODEL_FOR_TEXT_ENCODING_MAPPINGrs  r   r   r   r}  r}  j      4Nr   r}  c                       e Zd ZeZy)AutoModelForImageToImageN)r   r   r    MODEL_FOR_IMAGE_TO_IMAGE_MAPPINGrs  r   r   r   r  r  n  s    5Nr   r  c                       e Zd ZeZy)	AutoModelN)r   r   r   MODEL_MAPPINGrs  r   r   r   r  r  r  s    "Nr   r  c                       e Zd ZeZy)AutoModelForPreTrainingN)r   r   r   MODEL_FOR_PRETRAINING_MAPPINGrs  r   r   r   r  r  y  s    2Nr   r  pretraining)head_docc                   `     e Zd ZeZeded    deej                  e   z  ddf fd       Z
 xZS )AutoModelForCausalLMclspretrained_model_name_or_pathreturnr   c                 *    t        |   |g|i |S Nsuperfrom_pretrainedr  r  
model_argskwargs	__class__s       r   r  z$AutoModelForCausalLM.from_pretrained  !     w&'D\z\U[\\r   )r   r   r   MODEL_FOR_CAUSAL_LM_MAPPINGrs  classmethodtypestrosPathLiker  __classcell__r  s   @r   r  r    sO    0N ]()]'*R[[-='=]
 
"] ]r   r  zcausal language modelingc                       e Zd ZeZy)AutoModelForMaskedLMN)r   r   r   MODEL_FOR_MASKED_LM_MAPPINGrs  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_MAPPINGrs  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_MAPPINGrs  r   r   r   r  r    s    >Nr   r  zsequence classificationc                       e Zd ZeZy)AutoModelForQuestionAnsweringN)r   r   r   $MODEL_FOR_QUESTION_ANSWERING_MAPPINGrs  r   r   r   r  r    rx  r   r  zquestion answeringc                       e Zd ZeZy)"AutoModelForTableQuestionAnsweringN)r   r   r   *MODEL_FOR_TABLE_QUESTION_ANSWERING_MAPPINGrs  r   r   r   r  r        ?Nr   r  ztable question answeringzgoogle/tapas-base-finetuned-wtqc                       e Zd ZeZy)#AutoModelForVisualQuestionAnsweringN)r   r   r   +MODEL_FOR_VISUAL_QUESTION_ANSWERING_MAPPINGrs  r   r   r   r  r    s    @Nr   r  zvisual question answeringzdandelin/vilt-b32-finetuned-vqac                       e Zd ZeZy)%AutoModelForDocumentQuestionAnsweringN)r   r   r   -MODEL_FOR_DOCUMENT_QUESTION_ANSWERING_MAPPINGrs  r   r   r   r  r    s    BNr   r  zdocument question answeringz/impira/layoutlm-document-qa", revision="52e01b3c                       e Zd ZeZy)AutoModelForTokenClassificationN)r   r   r   &MODEL_FOR_TOKEN_CLASSIFICATION_MAPPINGrs  r   r   r   r  r    r  r   r  ztoken classificationc                       e Zd ZeZy)AutoModelForMultipleChoiceN)r   r   r   !MODEL_FOR_MULTIPLE_CHOICE_MAPPINGrs  r   r   r   r  r    rt  r   r  zmultiple choicec                       e Zd ZeZy)"AutoModelForNextSentencePredictionN)r   r   r   *MODEL_FOR_NEXT_SENTENCE_PREDICTION_MAPPINGrs  r   r   r   r  r    r  r   r  znext sentence predictionc                       e Zd ZeZy)AutoModelForImageClassificationN)r   r   r   &MODEL_FOR_IMAGE_CLASSIFICATION_MAPPINGrs  r   r   r   r  r    r  r   r  zimage classificationc                       e Zd ZeZy)'AutoModelForZeroShotImageClassificationN)r   r   r   0MODEL_FOR_ZERO_SHOT_IMAGE_CLASSIFICATION_MAPPINGrs  r   r   r   r  r    s    ENr   r  zzero-shot image classificationc                       e Zd ZeZy)AutoModelForImageSegmentationN)r   r   r   $MODEL_FOR_IMAGE_SEGMENTATION_MAPPINGrs  r   r   r   r  r    rx  r   r  zimage segmentationc                       e Zd ZeZy) AutoModelForSemanticSegmentationN)r   r   r   'MODEL_FOR_SEMANTIC_SEGMENTATION_MAPPINGrs  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_MAPPINGrs  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_MAPPINGrs  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_MAPPINGrs  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_MAPPINGrs  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_MAPPINGrs  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_MAPPINGrs  r   r   r   r  r  6  r  r   r  zdepth estimationc                       e Zd ZeZy)AutoModelForVideoClassificationN)r   r   r   &MODEL_FOR_VIDEO_CLASSIFICATION_MAPPINGrs  r   r   r   r  r  =  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_pretrainedH  r  r   )r   r   r   $MODEL_FOR_IMAGE_TEXT_TO_TEXT_MAPPINGrs  r  r  r  r  r  r  r  r  s   @r   r  r  D  sO    9N ]/0]'*R[[-='=]
 
"] ]r   r  zimage-text-to-text modelingc                       e Zd ZeZy)AutoModelForMultimodalLMN)r   r   r   MODEL_FOR_MULTIMODAL_LM_MAPPINGrs  r   r   r   r  r  U  r  r   r  zmultimodal generationc                       e Zd ZeZy)AutoModelForAudioClassificationN)r   r   r   &MODEL_FOR_AUDIO_CLASSIFICATION_MAPPINGrs  r   r   r   r  r  \  r  r   r  zaudio classificationc                       e Zd ZeZy)AutoModelForCTCN)r   r   r   MODEL_FOR_CTC_MAPPINGrs  r   r   r   r  r  c  s    *Nr   r  z%connectionist temporal classificationc                       e Zd ZeZy)AutoModelForSpeechSeq2SeqN)r   r   r   "MODEL_FOR_SPEECH_SEQ_2_SEQ_MAPPINGrs  r   r   r   r  r  j  r  r   r  z,sequence-to-sequence speech-to-text modelingc                       e Zd ZeZy)$AutoModelForAudioFrameClassificationN)r   r   r   ,MODEL_FOR_AUDIO_FRAME_CLASSIFICATION_MAPPINGrs  r   r   r   r  r  s  r  r   r  z"audio frame (token) classificationc                       e Zd ZeZy)AutoModelForAudioXVectorN)r   r   r   MODEL_FOR_AUDIO_XVECTOR_MAPPINGrs  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_MAPPINGrs  r   r   r   r  r    s    :Nr   r  c                       e Zd ZeZy)AutoModelForTextToWaveformN)r   r   r   "MODEL_FOR_TEXT_TO_WAVEFORM_MAPPINGrs  r   r   r   r  r    r  r   r  c                       e Zd ZeZy)AutoBackboneN)r   r   r   MODEL_FOR_BACKBONE_MAPPINGrs  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_MAPPINGrs  r   r   r   r  r    r  r   r  zmasked image modelingc                       e Zd ZeZy)AutoModelForAudioTokenizationN)r   r   r   $MODEL_FOR_AUDIO_TOKENIZATION_MAPPINGrs  r   r   r   r  r    rx  r   r  z$audio tokenization through codebooks)Wr  r  r  r  r  'MODEL_FOR_CAUSAL_IMAGE_MODELING_MAPPINGr  r  r  r  r  MODEL_FOR_IMAGE_MAPPINGr  r  rw  r{  r  r	  r  rr  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  r  r  r  r  r  r  r  r  r  rv  rz  rq  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  )__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_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_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_NAMESr  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  rr  rw  r{  r~  r  r  r  r  r  rq  rv  rz  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  __all__r   r   r   <module>rM     s    	 #     5 -1/  
		H	%!|~ @ '2\^' #@ %0Y[% !z !,9;! z 1<1 - 1< 	71 - 0;8:0 ,x .9 	(. * 1<
1 - 1< 	<1 - 2=
2 . 0;0 , %08*% ! .9=?. *F )4	8>>@	A4 	D 	P	
 	9 	@ 	> 	C 	B 	7 	H) %" %0.0% !d ,7, ($ 6A	6 2 ,7
, ( 0;"$0 ,N ,7, (( 3>tv3 /p .9NP. *d 4? 	/4 0 5@5 1 7B7 3 0;[]0 ,~ +6%'+ 'R 4?4 0 0;0 ,$ * $ 6A
6 2 )4
) % /: 	?// + ,7, ($ :E: 6  $/$  < +6	+ ' .98. * -8- ) )4 ") %H 6AC16 2 2=5-2 . 2=062 . *55* & &1& " !!57JK 01EGj k ./CEfg *:G+ ' *:F* & 4DP4 0 (8D( $ +;G+ ' +;G+ ' ,<H, ( *:F* & (8D( $ #33GIn"o ./CEfg .>K/ + 1AM1 - //CEfg *+?A^_ *:G+ ' &66JLt%u "/?L0 , &66JLt%u ")9F* & -=I- ) (8D( $ .>J. * *:F* & %55IKr$s !-=J. * *:F* & ))=?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, ( $44HJp#q  '78LNp'q $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$  
8"5 8 00KVhi B*= B '8'2N' #
8"5 8 00KVhi <&9 < #44S^t"u ]"5 ] 00KVst 52 5 --EPgh <&9 < #44S^t"u +) + $O>ef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! 
Xr   