o
    wi[                    @   s|  d Z ddl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e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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g dZedg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$eg dZ%eg dZ&eg d Z'ed!gZ(eg d"Z)eg d#Z*eg d$Z+eg d%Z,eg d&Z-eg d'Z.eg d(Z/eg d)Z0eg d*Z1ed+d,gZ2eg d-Z3eg d.Z4eg d/Z5ed0d1gZ6ed2gZ7eg d3Z8ed4d5gZ9ed6d7gZ:ed8gZ;ed9gZ<ed:gZ=e	eeZ>e	eeZ?e	eeZ@e	eeZAe	eeZBe	eeZCe	ee4ZDe	eeZEe	eeZFe	eeZGe	eeZHe	eeZIe	eeZJe	eeZKe	eeZLe	ee)ZMe	ee*ZNe	ee ZOe	eeZPe	eeZQe	ee!ZRe	ee"ZSe	ee#ZTe	ee$ZUe	ee&ZVe	ee'ZWe	ee(ZXe	ee+ZYe	ee,ZZe	ee-Z[e	ee.Z\e	ee/Z]e	ee%Z^e	ee0Z_e	ee1Z`e	ee2Zae	ee3Zbe	ee5Zce	ee6Zde	ee7Zee	ee8Zfe	ee9Zge	ee:Zhe	ee;Zie	ee<Zje	ee=ZkG d;d< d<eZlG d=d> d>eZmG d?d@ d@eZnG dAdB dBeZoG dCdD dDeZpe
epZpG dEdF dFeZqe
eqdGdHZqG dIdJ dJeZre
erdKdHZrG dLdM dMeZse
esdNdHZsG dOdP dPeZte
etdQdHZtG dRdS dSeZue
eudTdUdVZuG dWdX dXeZve
evdYdHZvG dZd[ d[eZwe
ewd\dHZwG d]d^ d^eZxe
exd_d`dVZxG dadb dbeZye
eydcdddVZyG dedf dfeZze
ezdgdhdVZzG didj djeZ{e
e{dkdHZ{G dldm dmeZ|e
e|dndHZ|G dodp dpeZ}e
e}dqdHZ}G drds dseZ~e
e~dtdHZ~G dudv dveZe
edwdHZG dxdy dyeZe
edzdHZG d{d| d|eZe
ed}dHZG d~d deZe
eddHZG dd deZe
eddHZG dd deZe
eddHZG dd deZe
eddHZG dd deZe
eddHZG dd deZe
eddHZG dd deZe
eddHZG dd deZe
eddHZG dd deZe
eddHZG dd deZe
eddHZG dd deZe
eddHZG dd deZe
eddHZG dd deZe
eddHZG dd deZG dd deZG dd deZG dd deZe
eddHZG dd deZe
eddHZG dd deZe
eddHZG dd derZg dZdS )zAuto Model class.    N)OrderedDict   )logging   )_BaseAutoBackboneClass_BaseAutoModelClass_LazyAutoMappingauto_class_update)CONFIG_MAPPING_NAMES(O  albertAlbertModelalign
AlignModelaltclipAltCLIPModel)arcee
ArceeModel)aria	AriaModel)	aria_textAriaTextModel)audio-spectrogram-transformerASTModel)
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)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conditional_detrConditionalDetrModel)convbertConvBertModelconvnextConvNextModel
convnextv2ConvNextV2Model)cpmantCpmAntModelcsmCsmForConditionalGeneration)ctrl	CTRLModel)cvtCvtModel)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_v3DeepseekV3Modeldeformable_detrDeformableDetrModeldeit	DeiTModel	depth_proDepthProModeldeta	DetaModeldetr	DetrModel)diaDiaModel)	diffllamaDiffLlamaModeldinat
DinatModeldinov2Dinov2Modeldinov2_with_registersDinov2WithRegistersModel
distilbertDistilBertModel)
donut-swinDonutSwinModel)dots1
Dots1Model)dprDPRQuestionEncoderdptDPTModelefficientformerEfficientFormerModelefficientnetEfficientNetModelelectraElectraModel)emu3	Emu3Model)encodecEncodecModel)ernie
ErnieModel)ernie_mErnieMModel)esmEsmModel)falconFalconModel)	falcon_h1FalconH1Model)falcon_mambaFalconMambaModelfastspeech2_conformerFastSpeech2ConformerModelflaubertFlaubertModel)flava
FlavaModel)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)glm4v
Glm4vModel)
glm4v_textGlm4vTextModelglpn	GLPNModel)got_ocr2GotOcr2Model)gpt-sw3	GPT2Model)gpt2r  )gpt_bigcodeGPTBigCodeModel)gpt_neoGPTNeoModel)gpt_neoxGPTNeoXModel)gpt_neox_japaneseGPTNeoXJapaneseModel)gptj	GPTJModelzgptsan-japanese&GPTSanJapaneseForConditionalGeneration)graniteGraniteModel)
granitemoeGraniteMoeModel)granitemoehybridGraniteMoeHybridModel)granitemoesharedGraniteMoeSharedModel)
graphormerGraphormerModel)grounding-dinoGroundingDinoModel)groupvitGroupViTModel)heliumHeliumModelhgnet_v2HGNetV2Backbonehiera
HieraModel)hubertHubertModel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)jamba
JambaModel)janus
JanusModel)jetmoeJetMoeModel)jukeboxJukeboxModel)kosmos-2Kosmos2Model)kyutai_speech_to_textKyutaiSpeechToTextModel)layoutlmLayoutLMModel)
layoutlmv2LayoutLMv2Model)
layoutlmv3LayoutLMv3Model)ledLEDModellevit
LevitModel)	lightglueLightGlueForKeypointMatching)lilt	LiltModel)llamarg   llama4Llama4ForConditionalGeneration)llama4_textLlama4TextModel)llava
LlavaModel)
llava_nextLlavaNextModel)llava_next_videoLlavaNextVideoModel)llava_onevisionLlavaOnevisionModel
longformerLongformerModel)longt5LongT5Model)luke	LukeModel)lxmertLxmertModel)m2m_100M2M100Model)mamba
MambaModel)mamba2Mamba2Model)marianMarianModel)markuplmMarkupLMModel)mask2formerMask2FormerModel)
maskformerMaskFormerModel)maskformer-swinMaskFormerSwinModel)mbart
MBartModel)mctct
MCTCTModel)mega	MegaModel)megatron-bertMegatronBertModel)zmgp-strMgpstrForSceneTextRecognition)mimi	MimiModel)minimaxMiniMaxModel)mistralMistralModel)mistral3Mistral3Model)mixtralMixtralModelmlcdMLCDVisionModel)mllamaMllamaModel
mobilebertMobileBertModelmobilenet_v1MobileNetV1Modelmobilenet_v2MobileNetV2Model	mobilevitMobileViTModelmobilevitv2MobileViTV2Model)
modernbertModernBertModel)	moonshineMoonshineModel)moshi
MoshiModel)mpnet
MPNetModel)mptMptModel)mraMraModel)mt5MT5Model)musicgenMusicgenModel)musicgen_melodyMusicgenMelodyModel)mvpMvpModelnatNatModel)nemotronNemotronModel)nezha
NezhaModel)nllb-moeNllbMoeModelnystromformerNystromformerModel)olmo	OlmoModel)olmo2
Olmo2Model)olmoe
OlmoeModelzomdet-turboOmDetTurboForObjectDetection)	oneformerOneFormerModel)
open-llamaOpenLlamaModel)
openai-gptOpenAIGPTModel)optOPTModel)owlv2
Owlv2Model)owlvitOwlViTModel)	paligemmaPaliGemmaModel)patchtsmixerPatchTSMixerModel)patchtstPatchTSTModel)pegasusPegasusModel)	pegasus_xPegasusXModel)	perceiverPerceiverModel)	persimmonPersimmonModel)phiPhiModel)phi3	Phi3Model)phi4_multimodalPhi4MultimodalModel)phimoePhimoeModel)pixtralPixtralVisionModel)plbartPLBartModel
poolformerPoolFormerModel)
prophetnetProphetNetModelpvtPvtModel)pvt_v2
PvtV2Model)qdqbertQDQBertModel)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_moeQwen3MoeModel)recurrent_gemmaRecurrentGemmaModelreformerReformerModelregnetRegNetModelrembertRemBertModelresnetResNetModel	retribertRetriBertModelrobertaRobertaModelroberta-prelayernormRobertaPreLayerNormModelroc_bertRoCBertModelroformerRoFormerModel)rt_detrRTDetrModel)
rt_detr_v2RTDetrV2Model)rwkv	RwkvModelsamSamModelsam_hq
SamHQModel)sam_hq_vision_modelSamHQVisionModel)sam_vision_modelSamVisionModel)seamless_m4tSeamlessM4TModel)seamless_m4t_v2SeamlessM4Tv2Model	segformerSegformerModel)seggptSegGptModel)sewSEWModel)sew-d	SEWDModelsiglipSiglipModelsiglip2Siglip2Modelsiglip_vision_modelSiglipVisionModel)smollm3SmolLM3Model)smolvlmSmolVLMModel)smolvlm_visionSmolVLMVisionTransformer)speech_to_textSpeech2TextModel)speecht5SpeechT5Model)splinterSplinterModelsqueezebertSqueezeBertModel)stablelmStableLmModel)
starcoder2Starcoder2Model)	superglueSuperGlueForKeypointMatchingswiftformerSwiftFormerModelswin	SwinModelswin2srSwin2SRModelswinv2Swinv2Model)switch_transformersSwitchTransformersModel)t5T5Model)t5gemmaT5GemmaModeltable-transformerTableTransformerModel)tapas
TapasModel)textnetTextNetModel)time_series_transformerTimeSeriesTransformerModel)timesfmTimesFmModeltimesformerTimesformerModeltimm_backboneTimmBackbonetimm_wrapperr  )trajectory_transformerTrajectoryTransformerModel)
transfo-xlTransfoXLModel)tvlt	TvltModel)tvpTvpModel)udop	UdopModel)umt5	UMT5Model)	unispeechUniSpeechModel)unispeech-satUniSpeechSatModel)univnetUnivNetModelvanVanModel)video_llavaVideoLlavaModelvideomaeVideoMAEModel)vilt	ViltModel)vipllavaVipLlavaModel)zvision-text-dual-encoderVisionTextDualEncoderModel)visual_bertVisualBertModelvitViTModel
vit_hybridViTHybridModelvit_maeViTMAEModelvit_msnViTMSNModelvitdetVitDetModelvits	VitsModelvivit
VivitModel)vjepa2VJEPA2Model)wav2vec2Wav2Vec2Model)wav2vec2-bertWav2Vec2BertModel)wav2vec2-conformerWav2Vec2ConformerModel)wavlm
WavLMModel)whisperWhisperModel)xclip
XCLIPModel)xglm	XGLMModelxlmXLMModel)xlm-prophetnetXLMProphetNetModelxlm-robertaXLMRobertaModelxlm-roberta-xlXLMRobertaXLModel)xlnet
XLNetModel)xmod	XmodModelyolos
YolosModel)yoso	YosoModel)zamba
ZambaModel)zamba2Zamba2Model)S)r   AlbertForPreTrainingr%   BartForConditionalGeneration)r+   BertForPreTraining)r0   BigBirdForPreTrainingrF   BloomForCausalLMrL   CamembertForMaskedLMcolpaliColPaliForRetrieval)colqwen2ColQwen2ForRetrievalr~   CTRLLMHeadModelr   Data2VecTextForMaskedLMr   DebertaForMaskedLMr   DebertaV2ForMaskedLMr   DistilBertForMaskedLM)r   ElectraForPreTraining)r   ErnieForPreTrainingr   FalconMambaForCausalLMr   FlaubertWithLMHeadModel)r   FlavaForPreTraining)r   FNetForPreTrainingr   FSMTForConditionalGeneration)r   FunnelForPreTrainingr   Gemma3ForConditionalGenerationr  GPT2LMHeadModelr  r*  r  GPTBigCodeForCausalLMr  )r5  HieraForPreTrainingr:  IBertForMaskedLMr<  IdeficsForVisionText2Textr>   Idefics2ForConditionalGenerationr@   Idefics3ForConditionalGenerationrV  JanusForConditionalGenerationr`  LayoutLMForMaskedLMru  LlavaForConditionalGenerationrw  !LlavaNextForConditionalGenerationry  &LlavaNextVideoForConditionalGenerationr{  &LlavaOnevisionForConditionalGenerationr~  LongformerForMaskedLMr  LukeForMaskedLM)r  LxmertForPreTrainingr  MambaForCausalLMr  Mamba2ForCausalLMr  MegaForMaskedLM)r  MegatronBertForPreTrainingr   Mistral3ForConditionalGenerationr  MllamaForConditionalGeneration)r  MobileBertForPreTrainingr  MPNetForMaskedLMr  MptForCausalLMr  MraForMaskedLMr  MvpForConditionalGeneration)r  NezhaForPreTrainingr  NllbMoeForConditionalGenerationr  OpenAIGPTLMHeadModelr  !PaliGemmaForConditionalGenerationqwen2_audio"Qwen2AudioForConditionalGenerationr7  r;  RobertaForMaskedLMr>  RobertaPreLayerNormForMaskedLM)rA  RoCBertForPreTrainingrJ  RwkvForCausalLM)rv  SplinterForPreTrainingry  SqueezeBertForMaskedLMr  *SwitchTransformersForConditionalGenerationr  T5ForConditionalGenerationr  T5GemmaForConditionalGenerationr  TapasForMaskedLMr  TransfoXLLMHeadModel)r  TvltForPreTraining)r  UniSpeechForPreTraining)r  UniSpeechSatForPreTrainingr  "VideoLlavaForConditionalGeneration)r  VideoMAEForPreTrainingr   VipLlavaForConditionalGeneration)r  VisualBertForPreTraining)r  ViTMAEForPreTraining)r  Wav2Vec2ForPreTraining)r  Wav2Vec2ConformerForPreTrainingr  XLMWithLMHeadModelr  XLMRobertaForMaskedLMr  XLMRobertaXLForMaskedLMr  XLNetLMHeadModelr  XmodForMaskedLM)Sr   AlbertForMaskedLMr  r+   BertForMaskedLMr0   BigBirdForMaskedLMr2   &BigBirdPegasusForConditionalGenerationr=   'BlenderbotSmallForConditionalGenerationr	  r  rh   CodeGenForCausalLMrq   ConvBertForMaskedLMry   CpmAntForCausalLMr  r  r  r  r   DiaForConditionalGenerationr  r   ElectraForMaskedLMzencoder-decoderEncoderDecoderModelr   ErnieForMaskedLMr   EsmForMaskedLMr  r   r   FNetForMaskedLMr$  r   FunnelForMaskedLMr  GitForCausalLMr)  r+  r,  r  GPTNeoForCausalLMr  GPTNeoXForCausalLMr  GPTNeoXJapaneseForCausalLMr  GPTJForCausalLMr  r/  r9  rf  LEDForConditionalGenerationrC  r  LongT5ForConditionalGenerationrE  r  M2M100ForConditionalGenerationrH  rJ  r  MarianMTModelrL  r  MegatronBertForCausalLMr  MobileBertForMaskedLMr  !MoonshineForConditionalGenerationrT  rV  rX  rZ  r  NezhaForMaskedLMr]  r  NystromformerForMaskedLMr_  r   PegasusXForConditionalGenerationr	  PLBartForConditionalGeneration	pop2piano!Pop2PianoForConditionalGenerationr  QDQBertForMaskedLMr,  ReformerModelWithLMHeadr2  RemBertForMaskedLMrf  rh  rA  RoCBertForMaskedLMrD  RoFormerForMaskedLMrk  rr  #Speech2TextForConditionalGenerationrn  rp  rr  rt  rv  rx  r  Wav2Vec2ForMaskedLMr  WhisperForConditionalGenerationr  r  r  r  r  r  YosoForMaskedLM)v)r   ArceeForCausalLM)r   AriaTextForCausalLM)r    BambaForCausalLM)r%   BartForCausalLM)r+   BertLMHeadModel)r-   BertGenerationDecoder)r0   BigBirdForCausalLM)r2   BigBirdPegasusForCausalLM)r4   BioGptForCausalLM)r9   BitNetForCausalLM)r;   BlenderbotForCausalLM)r=   BlenderbotSmallForCausalLMr	  )rL   CamembertForCausalLM)rf   LlamaForCausalLMr  )rj   CohereForCausalLM)rl   Cohere2ForCausalLMr  r  )r   Data2VecTextForCausalLM)r   DbrxForCausalLM)r   DeepseekV3ForCausalLM)r   DiffLlamaForCausalLM)r   Dots1ForCausalLM)r   ElectraForCausalLM)r   Emu3ForCausalLM)r   ErnieForCausalLM)r   FalconForCausalLM)r   FalconH1ForCausalLMr  r   FuyuForCausalLM)r   GemmaForCausalLM)r   Gemma2ForCausalLMr'  )r   Gemma3ForCausalLMr   Gemma3nForConditionalGeneration)r   Gemma3nForCausalLMr  )r  GlmForCausalLM)r  Glm4ForCausalLMr  GotOcr2ForConditionalGenerationr)  r+  r,  r  r  r  r  )r!  GraniteForCausalLM)r#  GraniteMoeForCausalLM)r%  GraniteMoeHybridForCausalLM)r'  GraniteMoeSharedForCausalLM)r/  HeliumForCausalLM)rT  JambaForCausalLM)rX  JetMoeForCausalLM)ro  r  )rq  Llama4ForCausalLM)rs  r  rH  rJ  )r  MarianForCausalLM)r  MBartForCausalLM)r  MegaForCausalLMr  )r  MiniMaxForCausalLM)r  MistralForCausalLM)r  MixtralForCausalLM)r  MllamaForCausalLM)r  MoshiForCausalLMrV  )r  MusicgenForCausalLM)r  MusicgenMelodyForCausalLM)r  MvpForCausalLM)r  NemotronForCausalLM)r  OlmoForCausalLM)r  Olmo2ForCausalLM)r  OlmoeForCausalLM)r  OpenLlamaForCausalLMr_  )r  OPTForCausalLM)r  PegasusForCausalLM)r  PersimmonForCausalLM)r  PhiForCausalLM)r  Phi3ForCausalLM)r  Phi4MultimodalForCausalLM)r  PhimoeForCausalLM)r	  PLBartForCausalLM)r  ProphetNetForCausalLM)r  QDQBertLMHeadModel)r  Qwen2ForCausalLM)r  Qwen2MoeForCausalLM)r%  Qwen3ForCausalLM)r'  Qwen3MoeForCausalLM)r)  RecurrentGemmaForCausalLMr  )r2  RemBertForCausalLM)r;  RobertaForCausalLM)r>  RobertaPreLayerNormForCausalLM)rA  RoCBertForCausalLM)rD  RoFormerForCausalLMrk  )rl  SmolLM3ForCausalLM)speech_to_text_2Speech2Text2ForCausalLM)r{  StableLmForCausalLM)r}  Starcoder2ForCausalLMrx  )trocrTrOCRForCausalLM)r  WhisperForCausalLM)r  XGLMForCausalLMr  )r  XLMProphetNetForCausalLM)r  XLMRobertaForCausalLM)r  XLMRobertaXLForCausalLMr  )r  XmodForCausalLM)r   ZambaForCausalLM)r  Zamba2ForCausalLM)7r'   r6   rn   rs   rv   r   r   r   r   r   r   r   r   r   r   r   r   r   r   r  r4  rD  rG  rh  )rq  Llama4VisionModelr  )r  MllamaVisionModelr  r  r  r  r  r  r  r.  r4  rZ  ri  r  r  r  r  r  r  r  r  r  r  r  r  r  r  r  r  r  ))r   DeiTForMaskedImageModeling)r   FocalNetForMaskedImageModeling)r  SwinForMaskedImageModeling)r  Swinv2ForMaskedImageModeling)r  ViTForMaskedImageModeling)rH  ImageGPTForCausalImageModeling),)r(   BeitForImageClassification)r7   BitForImageClassification)rZ   CLIPForImageClassification)rt   ConvNextForImageClassification)rw    ConvNextV2ForImageClassification)r   CvtForImageClassification)r   $Data2VecVisionForImageClassification)r   )DeiTForImageClassification%DeiTForImageClassificationWithTeacher)r   DinatForImageClassification)r   Dinov2ForImageClassification)r   )Dinov2WithRegistersForImageClassification)r   DonutSwinForImageClassification)r   )%EfficientFormerForImageClassification0EfficientFormerForImageClassificationWithTeacher)r   "EfficientNetForImageClassification)r   FocalNetForImageClassification)r2  HGNetV2ForImageClassification)r5  HieraForImageClassification)rE  IJepaForImageClassification)rH  ImageGPTForImageClassification)ri  )LevitForImageClassification&LevitForImageClassificationWithTeacher)r  !MobileNetV1ForImageClassification)r  !MobileNetV2ForImageClassification)r  MobileViTForImageClassification)r  !MobileViTV2ForImageClassification)r  NatForImageClassification)r  )&PerceiverForImageClassificationLearned&PerceiverForImageClassificationFourier-PerceiverForImageClassificationConvProcessing)r   PoolFormerForImageClassification)r  PvtForImageClassification)r  PvtV2ForImageClassification)r/  RegNetForImageClassification)r5  ResNetForImageClassification)r[  SegformerForImageClassification)shieldgemma2"ShieldGemma2ForImageClassification)rd  SiglipForImageClassification)rg  Siglip2ForImageClassification)r  !SwiftFormerForImageClassification)r  SwinForImageClassification)r  Swinv2ForImageClassification)r  TextNetForImageClassification)r  !TimmWrapperForImageClassification)r  VanForImageClassification)r  ViTForImageClassification)r  ViTHybridForImageClassification)r  ViTMSNForImageClassificationr   DetrForSegmentation))r(   BeitForSemanticSegmentation)r   %Data2VecVisionForSemanticSegmentation)r   DPTForSemanticSegmentation)r  "MobileNetV2ForSemanticSegmentation)r   MobileViTForSemanticSegmentation)r  "MobileViTV2ForSemanticSegmentation)r[   SegformerForSemanticSegmentation)upernetUperNetForSemanticSegmentationr  !MaskFormerForInstanceSegmentation)r~  )r  #Mask2FormerForUniversalSegmentationr  )r  !OneFormerForUniversalSegmentation))r  !TimesformerForVideoClassification)r  VideoMAEForVideoClassification)r  VivitForVideoClassification)r  VJEPA2ForVideoClassification)r@   BlipForConditionalGenerationrB   Blip2ForConditionalGenerationrP   !ChameleonForConditionalGenerationr  r3  r5  rL  $InstructBlipForConditionalGeneration)rN  )InstructBlipVideoForConditionalGenerationr\  Kosmos2ForConditionalGenerationr;  r=  r?  rA  rO  rQ  ra  
pix2struct"Pix2StructForConditionalGenerationr  "Qwen2_5_VLForConditionalGenerationr!  Qwen2VLForConditionalGenerationr}  r  zvision-encoder-decoderVisionEncoderDecoderModelr  )$)r   AriaForConditionalGeneration)r   !AyaVisionForConditionalGenerationr  r  r  )r   Emu3ForConditionalGenerationr  r'  r  r  )r	  Glm4vForConditionalGenerationr  r1  r3  r5  r  )rP   InternVLForConditionalGenerationr7  r  rp  r;  r=  r?  rA  rO  rQ  ra  r  )r  r<  r  r  )rq  r(  )rn  SmolVLMForConditionalGeneration)r  UdopForConditionalGenerationr  r  ).r  r  r  r  r  r  r  r  r  r  r  r  r  r   r  r  r/  r9  rC  rE  r  MBartForConditionalGenerationrL  )r  MegatronBertForMaskedLMr  )r  ModernBertForMaskedLMrT  rX  rZ  r  r  )r  PerceiverForMaskedLMr  )r,  ReformerForMaskedLMr  rf  rh  r  r  rn  rv  r  r  r  r  r  r  )
)ro   !ConditionalDetrForObjectDetection)r   DFineForObjectDetection)r   DabDetrForObjectDetection)r    DeformableDetrForObjectDetection)r   DetaForObjectDetection)r   DetrForObjectDetection)rF  RTDetrForObjectDetection)rH  RTDetrV2ForObjectDetection)r  "TableTransformerForObjectDetection)r  YolosForObjectDetection))r+  GroundingDinoForObjectDetectionr  )r  Owlv2ForObjectDetection)r  OwlViTForObjectDetection))depth_anythingDepthAnythingForDepthEstimation)r   DepthProForDepthEstimation)r   DPTForDepthEstimation)r  GLPNForDepthEstimation)prompt_depth_anything%PromptDepthAnythingForDepthEstimation)zoedepthZoeDepthForDepthEstimation)r  r  )r;   "BlenderbotForConditionalGenerationr  r  r$  r  granite_speech%GraniteSpeechForConditionalGenerationr  r  r  r  r  )r  MT5ForConditionalGenerationrZ  r]  )r  PegasusForConditionalGenerationr  r  )r  "ProphetNetForConditionalGenerationrc  )rV  SeamlessM4TForTextToText)rX  SeamlessM4Tv2ForTextToTextrp  rr  rt  )r  UMT5ForConditionalGeneration)r  %XLMProphetNetForConditionalGeneration)r  r  )r^  *KyutaiSpeechToTextForConditionalGenerationr  r  )rV  SeamlessM4TForSpeechToText)rX  SeamlessM4Tv2ForSpeechToText)zspeech-encoder-decoderSpeechEncoderDecoderModelr  )rt  SpeechT5ForSpeechToTextr  )f)r   AlbertForSequenceClassification)r   ArceeForSequenceClassification)r%   BartForSequenceClassification)r+   BertForSequenceClassification)r0    BigBirdForSequenceClassification)r2   'BigBirdPegasusForSequenceClassification)r4   BioGptForSequenceClassification)rF   BloomForSequenceClassification)rL   "CamembertForSequenceClassification)rN   CanineForSequenceClassification)rf   LlamaForSequenceClassification)rq   !ConvBertForSequenceClassification)r~   CTRLForSequenceClassification)r   %Data2VecTextForSequenceClassification)r    DebertaForSequenceClassification)r   "DebertaV2ForSequenceClassification)r   "DiffLlamaForSequenceClassification)r   #DistilBertForSequenceClassification)r    ElectraForSequenceClassification)r   ErnieForSequenceClassification)r   ErnieMForSequenceClassification)r   EsmForSequenceClassification)r   FalconForSequenceClassification)r   !FlaubertForSequenceClassification)r   FNetForSequenceClassification)r   FunnelForSequenceClassification)r   GemmaForSequenceClassification)r   Gemma2ForSequenceClassification)r  GlmForSequenceClassification)r  Glm4ForSequenceClassification)r  GPT2ForSequenceClassification)r  r  )r  #GPTBigCodeForSequenceClassification)r  GPTNeoForSequenceClassification)r   GPTNeoXForSequenceClassification)r  GPTJForSequenceClassification)r/  HeliumForSequenceClassification)r:  IBertForSequenceClassification)rT  JambaForSequenceClassification)rX  JetMoeForSequenceClassification)r`  !LayoutLMForSequenceClassification)rb  #LayoutLMv2ForSequenceClassification)rd  #LayoutLMv3ForSequenceClassification)rf  LEDForSequenceClassification)rm  LiltForSequenceClassification)ro  r  )r~  #LongformerForSequenceClassification)r  LukeForSequenceClassification)r  !MarkupLMForSequenceClassification)r  MBartForSequenceClassification)r  MegaForSequenceClassification)r  %MegatronBertForSequenceClassification)r   MiniMaxForSequenceClassification)r   MistralForSequenceClassification)r   MixtralForSequenceClassification)r  #MobileBertForSequenceClassification)r  #ModernBertForSequenceClassification)r  MPNetForSequenceClassification)r  MptForSequenceClassification)r  MraForSequenceClassification)r  MT5ForSequenceClassification)r  MvpForSequenceClassification)r  !NemotronForSequenceClassification)r  NezhaForSequenceClassification)r  &NystromformerForSequenceClassification)r  "OpenLlamaForSequenceClassification)r  "OpenAIGPTForSequenceClassification)r  OPTForSequenceClassification)r  "PerceiverForSequenceClassification)r  "PersimmonForSequenceClassification)r  PhiForSequenceClassification)r  Phi3ForSequenceClassification)r  PhimoeForSequenceClassification)r	  PLBartForSequenceClassification)r   QDQBertForSequenceClassification)r  Qwen2ForSequenceClassification)r  !Qwen2MoeForSequenceClassification)r%  Qwen3ForSequenceClassification)r'  !Qwen3MoeForSequenceClassification)r,  !ReformerForSequenceClassification)r2   RemBertForSequenceClassification)r;   RobertaForSequenceClassification)r>  ,RobertaPreLayerNormForSequenceClassification)rA   RoCBertForSequenceClassification)rD  !RoFormerForSequenceClassification)rl   SmolLM3ForSequenceClassification)ry  $SqueezeBertForSequenceClassification)r{  !StableLmForSequenceClassification)r}  #Starcoder2ForSequenceClassification)r  T5ForSequenceClassification)r   T5GemmaForSequenceClassification)r  TapasForSequenceClassification)r  "TransfoXLForSequenceClassification)r  UMT5ForSequenceClassification)r  XLMForSequenceClassification)r  #XLMRobertaForSequenceClassification)r  %XLMRobertaXLForSequenceClassification)r  XLNetForSequenceClassification)r  XmodForSequenceClassification)r  YosoForSequenceClassification)r   ZambaForSequenceClassification)r  Zamba2ForSequenceClassification)K)r   AlbertForQuestionAnswering)r   ArceeForQuestionAnswering)r%   BartForQuestionAnswering)r+   BertForQuestionAnswering)r0   BigBirdForQuestionAnswering)r2   "BigBirdPegasusForQuestionAnswering)rF   BloomForQuestionAnswering)rL   CamembertForQuestionAnswering)rN   CanineForQuestionAnswering)rq   ConvBertForQuestionAnswering)r    Data2VecTextForQuestionAnswering)r   DebertaForQuestionAnswering)r   DebertaV2ForQuestionAnswering)r   DiffLlamaForQuestionAnswering)r   DistilBertForQuestionAnswering)r   ElectraForQuestionAnswering)r   ErnieForQuestionAnswering)r   ErnieMForQuestionAnswering)r   FalconForQuestionAnswering)r   "FlaubertForQuestionAnsweringSimple)r   FNetForQuestionAnswering)r   FunnelForQuestionAnswering)r  GPT2ForQuestionAnswering)r  GPTNeoForQuestionAnswering)r  GPTNeoXForQuestionAnswering)r  GPTJForQuestionAnswering)r:  IBertForQuestionAnsweringrb  LayoutLMv2ForQuestionAnsweringrd  LayoutLMv3ForQuestionAnswering)rf  LEDForQuestionAnswering)rm  LiltForQuestionAnswering)ro  LlamaForQuestionAnswering)r~  LongformerForQuestionAnswering)r  LukeForQuestionAnswering)r  LxmertForQuestionAnswering)r  MarkupLMForQuestionAnswering)r  MBartForQuestionAnswering)r  MegaForQuestionAnswering)r   MegatronBertForQuestionAnswering)r  MiniMaxForQuestionAnswering)r  MistralForQuestionAnswering)r  MixtralForQuestionAnswering)r  MobileBertForQuestionAnswering)r  ModernBertForQuestionAnswering)r  MPNetForQuestionAnswering)r  MptForQuestionAnswering)r  MraForQuestionAnswering)r  MT5ForQuestionAnswering)r  MvpForQuestionAnswering)r  NemotronForQuestionAnswering)r  NezhaForQuestionAnswering)r  !NystromformerForQuestionAnswering)r  OPTForQuestionAnswering)r  QDQBertForQuestionAnswering)r  Qwen2ForQuestionAnswering)r  Qwen2MoeForQuestionAnswering)r%  Qwen3ForQuestionAnswering)r'  Qwen3MoeForQuestionAnswering)r,  ReformerForQuestionAnswering)r2  RemBertForQuestionAnswering)r;  RobertaForQuestionAnswering)r>  'RobertaPreLayerNormForQuestionAnswering)rA  RoCBertForQuestionAnswering)rD  RoFormerForQuestionAnswering)rl  SmolLM3ForQuestionAnswering)rv  SplinterForQuestionAnswering)ry  SqueezeBertForQuestionAnswering)r  T5ForQuestionAnswering)r  UMT5ForQuestionAnswering)r  XLMForQuestionAnsweringSimple)r  XLMRobertaForQuestionAnswering)r   XLMRobertaXLForQuestionAnswering)r  XLNetForQuestionAnsweringSimple)r  XmodForQuestionAnswering)r  YosoForQuestionAnswering)r  TapasForQuestionAnswering))r@   BlipForQuestionAnsweringr  )r  ViltForQuestionAnswering))r`  LayoutLMForQuestionAnsweringrW  rY  )R)r   AlbertForTokenClassification)r   ArceeForTokenClassification)r+   BertForTokenClassification)r0   BigBirdForTokenClassification)r4   BioGptForTokenClassification)rF   BloomForTokenClassification)rJ   BrosForTokenClassification)rL   CamembertForTokenClassification)rN   CanineForTokenClassification)rq   ConvBertForTokenClassification)r   "Data2VecTextForTokenClassification)r   DebertaForTokenClassification)r   DebertaV2ForTokenClassification)r   DiffLlamaForTokenClassification)r    DistilBertForTokenClassification)r   ElectraForTokenClassification)r   ErnieForTokenClassification)r   ErnieMForTokenClassification)r   EsmForTokenClassification)r   FalconForTokenClassification)r   FlaubertForTokenClassification)r   FNetForTokenClassification)r   FunnelForTokenClassification)r   GemmaForTokenClassification)r   Gemma2ForTokenClassification)r  GlmForTokenClassification)r  Glm4ForTokenClassification)r  GPT2ForTokenClassification)r  r  )r   GPTBigCodeForTokenClassification)r  GPTNeoForTokenClassification)r  GPTNeoXForTokenClassification)r/  HeliumForTokenClassification)r:  IBertForTokenClassification)r`  LayoutLMForTokenClassification)rb   LayoutLMv2ForTokenClassification)rd   LayoutLMv3ForTokenClassification)rm  LiltForTokenClassification)ro  LlamaForTokenClassification)r~   LongformerForTokenClassification)r  LukeForTokenClassification)r  MarkupLMForTokenClassification)r  MegaForTokenClassification)r  "MegatronBertForTokenClassification)r  MiniMaxForTokenClassification)r  MistralForTokenClassification)r  MixtralForTokenClassification)r   MobileBertForTokenClassification)r   ModernBertForTokenClassification)r  MPNetForTokenClassification)r  MptForTokenClassification)r  MraForTokenClassification)r  MT5ForTokenClassification)r  NemotronForTokenClassification)r  NezhaForTokenClassification)r  #NystromformerForTokenClassification)r  PersimmonForTokenClassification)r  PhiForTokenClassification)r  Phi3ForTokenClassification)r  QDQBertForTokenClassification)r  Qwen2ForTokenClassification)r  Qwen2MoeForTokenClassification)r%  Qwen3ForTokenClassification)r'  Qwen3MoeForTokenClassification)r2  RemBertForTokenClassification)r;  RobertaForTokenClassification)r>  )RobertaPreLayerNormForTokenClassification)rA  RoCBertForTokenClassification)rD  RoFormerForTokenClassification)rl  SmolLM3ForTokenClassification)ry  !SqueezeBertForTokenClassification)r{  StableLmForTokenClassification)r}   Starcoder2ForTokenClassification)r  T5ForTokenClassification)r  T5GemmaForTokenClassification)r  UMT5ForTokenClassification)r  XLMForTokenClassification)r   XLMRobertaForTokenClassification)r  "XLMRobertaXLForTokenClassification)r  XLNetForTokenClassification)r  XmodForTokenClassification)r  YosoForTokenClassification)&)r   AlbertForMultipleChoice)r+   BertForMultipleChoice)r0   BigBirdForMultipleChoice)rL   CamembertForMultipleChoice)rN   CanineForMultipleChoice)rq   ConvBertForMultipleChoice)r   Data2VecTextForMultipleChoice)r   DebertaV2ForMultipleChoice)r   DistilBertForMultipleChoice)r   ElectraForMultipleChoice)r   ErnieForMultipleChoice)r   ErnieMForMultipleChoice)r   FlaubertForMultipleChoice)r   FNetForMultipleChoice)r   FunnelForMultipleChoice)r:  IBertForMultipleChoice)r~  LongformerForMultipleChoice)r  LukeForMultipleChoice)r  MegaForMultipleChoice)r  MegatronBertForMultipleChoice)r  MobileBertForMultipleChoice)r  MPNetForMultipleChoice)r  MraForMultipleChoice)r  NezhaForMultipleChoice)r  NystromformerForMultipleChoice)r  QDQBertForMultipleChoice)r2  RemBertForMultipleChoice)r;  RobertaForMultipleChoice)r>  $RobertaPreLayerNormForMultipleChoice)rA  RoCBertForMultipleChoice)rD  RoFormerForMultipleChoice)ry  SqueezeBertForMultipleChoice)r  XLMForMultipleChoice)r  XLMRobertaForMultipleChoice)r  XLMRobertaXLForMultipleChoice)r  XLNetForMultipleChoice)r  XmodForMultipleChoice)r  YosoForMultipleChoice))r+   BertForNextSentencePrediction)r   ErnieForNextSentencePrediction)r   FNetForNextSentencePrediction)r  %MegatronBertForNextSentencePrediction)r  #MobileBertForNextSentencePrediction)r  NezhaForNextSentencePrediction)r   QDQBertForNextSentencePrediction))r   ASTForAudioClassification)r   &Data2VecAudioForSequenceClassification)r7  HubertForSequenceClassification)r_  SEWForSequenceClassification)ra  SEWDForSequenceClassification)r  "UniSpeechForSequenceClassification)r  %UniSpeechSatForSequenceClassification)r  !Wav2Vec2ForSequenceClassification)r  %Wav2Vec2BertForSequenceClassification)r  *Wav2Vec2ConformerForSequenceClassification)r  WavLMForSequenceClassification)r  WhisperForAudioClassification))r   Data2VecAudioForCTC)r7  HubertForCTC)r  MCTCTForCTC)r_  	SEWForCTC)ra  
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   )rt  SpeechT5ForTextToSpeech)	r"   r{   )r   FastSpeech2ConformerWithHifiGan)r   MusicgenForConditionalGeneration)r  &MusicgenMelodyForConditionalGeneration)qwen2_5_omni#Qwen2_5OmniForConditionalGeneration)rV  SeamlessM4TForTextToSpeech)rX  SeamlessM4Tv2ForTextToSpeechr  )	r   r   r?   )rB   Blip2ForImageTextRetrievalrR   rY   ra   rc  rf  ))r(   BeitBackbone)r7   BitBackbone)rt   ConvNextBackbone)rw   ConvNextV2Backbone)r   DinatBackbone)r   Dinov2Backbone)r   Dinov2WithRegistersBackbone)r   FocalNetBackboner1  )r5  HieraBackbone)r  MaskFormerSwinBackbone)r  NatBackbone)r  PvtV2Backbone)r5  ResNetBackbone)rt_detr_resnetRTDetrResNetBackbone)r  SwinBackbone)r  Swinv2Backbone)r  TextNetBackboner  )r  VitDetBackbone)vitpose_backboneVitPoseBackbonerL  rO  )
superpointSuperPointForKeypointDetection)r   r*   r/   r\   r   r   r   r   r   )r   Emu3TextModelr   r9  )rq  rt  r}  )r  MllamaTextModelr  )r  MT5EncoderModelr  r+  r1  r:  r=  r@  rC  rx  )r  T5EncoderModel)r  T5GemmaEncoderModel)r  UMT5EncoderModelr  r  r  )r  'PatchTSMixerForTimeSeriesClassification)r  PatchTSTForClassification)r  PatchTSMixerForRegression)r  PatchTSTForRegression)r  TimesFmModelForPrediction)r  Swin2SRForImageSuperResolutionr   c                   @      e Zd ZeZdS )AutoModelForMaskGenerationN)__name__
__module____qualname__!MODEL_FOR_MASK_GENERATION_MAPPING_model_mapping ra  ra  c/home/ubuntu/sommelier/.venv/lib/python3.10/site-packages/transformers/models/auto/modeling_auto.pyr[        r[  c                   @   rZ  )AutoModelForKeypointDetectionN)r\  r]  r^  $MODEL_FOR_KEYPOINT_DETECTION_MAPPINGr`  ra  ra  ra  rb  rd    rc  rd  c                   @   rZ  )AutoModelForTextEncodingN)r\  r]  r^  MODEL_FOR_TEXT_ENCODING_MAPPINGr`  ra  ra  ra  rb  rf    rc  rf  c                   @   rZ  )AutoModelForImageToImageN)r\  r]  r^   MODEL_FOR_IMAGE_TO_IMAGE_MAPPINGr`  ra  ra  ra  rb  rh    rc  rh  c                   @   rZ  )	AutoModelN)r\  r]  r^  MODEL_MAPPINGr`  ra  ra  ra  rb  rj    rc  rj  c                   @   rZ  )AutoModelForPreTrainingN)r\  r]  r^  MODEL_FOR_PRETRAINING_MAPPINGr`  ra  ra  ra  rb  rl    rc  rl  pretraining)head_docc                   @   rZ  )_AutoModelWithLMHeadN)r\  r]  r^  MODEL_WITH_LM_HEAD_MAPPINGr`  ra  ra  ra  rb  rp    rc  rp  zlanguage modelingc                   @   rZ  )AutoModelForCausalLMN)r\  r]  r^  MODEL_FOR_CAUSAL_LM_MAPPINGr`  ra  ra  ra  rb  rr    rc  rr  zcausal language modelingc                   @   rZ  )AutoModelForMaskedLMN)r\  r]  r^  MODEL_FOR_MASKED_LM_MAPPINGr`  ra  ra  ra  rb  rt    rc  rt  zmasked language modelingc                   @   rZ  )AutoModelForSeq2SeqLMN)r\  r]  r^  &MODEL_FOR_SEQ_TO_SEQ_CAUSAL_LM_MAPPINGr`  ra  ra  ra  rb  rv    rc  rv  z&sequence-to-sequence language modelingzgoogle-t5/t5-base)ro  checkpoint_for_examplec                   @   rZ  )"AutoModelForSequenceClassificationN)r\  r]  r^  )MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPINGr`  ra  ra  ra  rb  ry    rc  ry  zsequence classificationc                   @   rZ  )AutoModelForQuestionAnsweringN)r\  r]  r^  $MODEL_FOR_QUESTION_ANSWERING_MAPPINGr`  ra  ra  ra  rb  r{  '  rc  r{  zquestion answeringc                   @   rZ  )"AutoModelForTableQuestionAnsweringN)r\  r]  r^  *MODEL_FOR_TABLE_QUESTION_ANSWERING_MAPPINGr`  ra  ra  ra  rb  r}  .  rc  r}  ztable question answeringzgoogle/tapas-base-finetuned-wtqc                   @   rZ  )#AutoModelForVisualQuestionAnsweringN)r\  r]  r^  +MODEL_FOR_VISUAL_QUESTION_ANSWERING_MAPPINGr`  ra  ra  ra  rb  r  9  rc  r  zvisual question answeringzdandelin/vilt-b32-finetuned-vqac                   @   rZ  )%AutoModelForDocumentQuestionAnsweringN)r\  r]  r^  -MODEL_FOR_DOCUMENT_QUESTION_ANSWERING_MAPPINGr`  ra  ra  ra  rb  r  D  rc  r  zdocument question answeringz/impira/layoutlm-document-qa", revision="52e01b3c                   @   rZ  )AutoModelForTokenClassificationN)r\  r]  r^  &MODEL_FOR_TOKEN_CLASSIFICATION_MAPPINGr`  ra  ra  ra  rb  r  O  rc  r  ztoken classificationc                   @   rZ  )AutoModelForMultipleChoiceN)r\  r]  r^  !MODEL_FOR_MULTIPLE_CHOICE_MAPPINGr`  ra  ra  ra  rb  r  V  rc  r  zmultiple choicec                   @   rZ  )"AutoModelForNextSentencePredictionN)r\  r]  r^  *MODEL_FOR_NEXT_SENTENCE_PREDICTION_MAPPINGr`  ra  ra  ra  rb  r  ]  rc  r  znext sentence predictionc                   @   rZ  )AutoModelForImageClassificationN)r\  r]  r^  &MODEL_FOR_IMAGE_CLASSIFICATION_MAPPINGr`  ra  ra  ra  rb  r  f  rc  r  zimage classificationc                   @   rZ  )'AutoModelForZeroShotImageClassificationN)r\  r]  r^  0MODEL_FOR_ZERO_SHOT_IMAGE_CLASSIFICATION_MAPPINGr`  ra  ra  ra  rb  r  m  rc  r  zzero-shot image classificationc                   @   rZ  )AutoModelForImageSegmentationN)r\  r]  r^  $MODEL_FOR_IMAGE_SEGMENTATION_MAPPINGr`  ra  ra  ra  rb  r  v  rc  r  zimage segmentationc                   @   rZ  ) AutoModelForSemanticSegmentationN)r\  r]  r^  'MODEL_FOR_SEMANTIC_SEGMENTATION_MAPPINGr`  ra  ra  ra  rb  r  }  rc  r  zsemantic segmentationc                   @   rZ  ) AutoModelForTimeSeriesPredictionN)r\  r]  r^  (MODEL_FOR_TIME_SERIES_PREDICTION_MAPPINGr`  ra  ra  ra  rb  r    rc  r  ztime-series predictionc                   @   rZ  )!AutoModelForUniversalSegmentationN)r\  r]  r^  (MODEL_FOR_UNIVERSAL_SEGMENTATION_MAPPINGr`  ra  ra  ra  rb  r    rc  r  zuniversal image segmentationc                   @   rZ  ) AutoModelForInstanceSegmentationN)r\  r]  r^  'MODEL_FOR_INSTANCE_SEGMENTATION_MAPPINGr`  ra  ra  ra  rb  r    rc  r  zinstance segmentationc                   @   rZ  )AutoModelForObjectDetectionN)r\  r]  r^  "MODEL_FOR_OBJECT_DETECTION_MAPPINGr`  ra  ra  ra  rb  r    rc  r  zobject detectionc                   @   rZ  )#AutoModelForZeroShotObjectDetectionN)r\  r]  r^  ,MODEL_FOR_ZERO_SHOT_OBJECT_DETECTION_MAPPINGr`  ra  ra  ra  rb  r    rc  r  zzero-shot object detectionc                   @   rZ  )AutoModelForDepthEstimationN)r\  r]  r^  "MODEL_FOR_DEPTH_ESTIMATION_MAPPINGr`  ra  ra  ra  rb  r    rc  r  zdepth estimationc                   @   rZ  )AutoModelForVideoClassificationN)r\  r]  r^  &MODEL_FOR_VIDEO_CLASSIFICATION_MAPPINGr`  ra  ra  ra  rb  r    rc  r  zvideo classificationc                   @   rZ  )AutoModelForVision2SeqN)r\  r]  r^  MODEL_FOR_VISION_2_SEQ_MAPPINGr`  ra  ra  ra  rb  r    rc  r  zvision-to-text modelingc                   @   rZ  )AutoModelForImageTextToTextN)r\  r]  r^  $MODEL_FOR_IMAGE_TEXT_TO_TEXT_MAPPINGr`  ra  ra  ra  rb  r    rc  r  zimage-text-to-text modelingc                   @   rZ  )AutoModelForAudioClassificationN)r\  r]  r^  &MODEL_FOR_AUDIO_CLASSIFICATION_MAPPINGr`  ra  ra  ra  rb  r    rc  r  zaudio classificationc                   @   rZ  )AutoModelForCTCN)r\  r]  r^  MODEL_FOR_CTC_MAPPINGr`  ra  ra  ra  rb  r    rc  r  z%connectionist temporal classificationc                   @   rZ  )AutoModelForSpeechSeq2SeqN)r\  r]  r^  "MODEL_FOR_SPEECH_SEQ_2_SEQ_MAPPINGr`  ra  ra  ra  rb  r    rc  r  z,sequence-to-sequence speech-to-text modelingc                   @   rZ  )$AutoModelForAudioFrameClassificationN)r\  r]  r^  ,MODEL_FOR_AUDIO_FRAME_CLASSIFICATION_MAPPINGr`  ra  ra  ra  rb  r    rc  r  z"audio frame (token) classificationc                   @   rZ  )AutoModelForAudioXVectorN)r\  r]  r^  MODEL_FOR_AUDIO_XVECTOR_MAPPINGr`  ra  ra  ra  rb  r    rc  r  c                   @   rZ  )AutoModelForTextToSpectrogramN)r\  r]  r^  %MODEL_FOR_TEXT_TO_SPECTROGRAM_MAPPINGr`  ra  ra  ra  rb  r    rc  r  c                   @   rZ  )AutoModelForTextToWaveformN)r\  r]  r^  "MODEL_FOR_TEXT_TO_WAVEFORM_MAPPINGr`  ra  ra  ra  rb  r    rc  r  c                   @   rZ  )AutoBackboneN)r\  r]  r^  MODEL_FOR_BACKBONE_MAPPINGr`  ra  ra  ra  rb  r    rc  r  zaudio retrieval via x-vectorc                   @   rZ  )AutoModelForMaskedImageModelingN)r\  r]  r^  'MODEL_FOR_MASKED_IMAGE_MODELING_MAPPINGr`  ra  ra  ra  rb  r     rc  r  zmasked image modelingc                   @   rZ  )AutoModelForAudioTokenizationN)r\  r]  r^  $MODEL_FOR_AUDIO_TOKENIZATION_MAPPINGr`  ra  ra  ra  rb  r    rc  r  z$audio tokenization through codebooksc                       s0   e Zd Ze fddZe fddZ  ZS )AutoModelWithLMHeadc                    s   t dt t |S NzThe class `AutoModelWithLMHead` is deprecated and will be removed in a future version. Please use `AutoModelForCausalLM` for causal language models, `AutoModelForMaskedLM` for masked language models and `AutoModelForSeq2SeqLM` for encoder-decoder models.)warningswarnFutureWarningsuperfrom_config)clsconfig	__class__ra  rb  r    s
   zAutoModelWithLMHead.from_configc                    s&   t dt t j|g|R i |S r  )r  r  r  r  from_pretrained)r  pretrained_model_name_or_path
model_argskwargsr  ra  rb  r    s
   z#AutoModelWithLMHead.from_pretrained)r\  r]  r^  classmethodr  r  __classcell__ra  ra  r  rb  r    s
    	r  )Wr  r  r  r  r  'MODEL_FOR_CAUSAL_IMAGE_MODELING_MAPPINGrs  r  r  r  r  MODEL_FOR_IMAGE_MAPPINGr  ri  re  r  r  ru  r_  r  r  r  rm  r|  r  rw  rz  r  r~  rg  r  r  r  r  r  r  r  MODEL_FOR_RETRIEVAL_MAPPINGr  r  rk  rq  r  r  ,MODEL_FOR_TIME_SERIES_CLASSIFICATION_MAPPING(MODEL_FOR_TIME_SERIES_REGRESSION_MAPPINGrj  r  r  r  r  r  rr  r  r  r  r  rh  r  rd  r[  rf  r  rt  r  r  r  rl  r{  r  rv  ry  r  r}  r  r  r  r  r  r  r  r  r  r  r  r  r  )__doc__r  collectionsr   utilsr   auto_factoryr   r   r   r	   configuration_autor
   
get_loggerr\  loggerMODEL_MAPPING_NAMES#MODEL_FOR_PRETRAINING_MAPPING_NAMES MODEL_WITH_LM_HEAD_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_VISION_2_SEQ_MAPPING_NAMES!MODEL_FOR_RETRIEVAL_MAPPING_NAMES*MODEL_FOR_IMAGE_TEXT_TO_TEXT_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_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_NAMESrk  rm  rq  rs  r  r  r  r  r  r  r  r  r  r  r  r  r  ru  r  r  r  r  r  rw  rz  r|  r~  r  r  r  r  r  r  r  r  r  r  r  r_  re  rg  r  r  r  ri  r  r[  rd  rf  rh  rj  rl  rp  rr  rt  rv  ry  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__ra  ra  ra  rb  <module>   s  
  WYY|=E
	)4
"lQX,$



















