
    bCi                    d   S r SSKrSSKrSSKJr  SSKJrJr  SSKJ	r	  SSK
JrJrJrJr  SS	KJr  \(       a  SS
KJr  SSKJr   " S S\\5      r\	R,                  " \5      r\" / SQ5      r\" / SQ5      r\" / SQ5      r\" / SQ5      r\" / SQ5      r\" / SQ5      r\" S/5      r\" / SQ5      r \" S/5      r!\" / SQ5      r"\" S/5      r#\" / SQ5      r$\" / SQ5      r%\" / SQ5      r&\" S/5      r'\" / SQ5      r(\" / SQ5      r)\" / SQ5      r*\" / S Q5      r+\" / S!Q5      r,\" / S"Q5      r-\" / S#Q5      r.\" / S$Q5      r/\" / S%Q5      r0\" S&/5      r1\" / S'Q5      r2\" / S(Q5      r3\" / S)Q5      r4\" / S*Q5      r5\" / S+Q5      r6\" / S,Q5      r7\" / S-Q5      r8\" / S.Q5      r9\" / S/Q5      r:\" S0S1/5      r;\" / S2Q5      r<\" / S3Q5      r=\" / S4Q5      r>\" / S5Q5      r?\" S6/5      r@\" / S7Q5      rA\" / S8Q5      rB\" S9S:/5      rC\" S;S</5      rD\" S=/5      rE\" S>/5      rF\" S?/5      rG\" \\5      rH\" \\5      rI\" \\5      rJ\" \\5      rK\" \\5      rL\" \\ 5      rM\" \\=5      rN\" \\!5      rO\" \\"5      rP\" \\#5      rQ\" \\$5      rR\" \\%5      rS\" \\&5      rT\" \\(5      rU\" \\'5      rV\" \\25      rW\" \\35      rX\" \\)5      rY\" \\5      rZ\" \\5      r[\" \\*5      r\\" \\+5      r]\" \\,5      r^\" \\-5      r_\" \\/5      r`\" \\05      ra\" \\15      rb\" \\45      rc\" \\55      rd\" \\65      re\" \\75      rf\" \\85      rg\" \\.5      rh\" \\95      ri\" \\:5      rj\" \\;5      rk\" \\<5      rl\" \\>5      rm\" \\?5      rn\" \\@5      ro\" \\A5      rp\" \\B5      rq\" \\C5      rr\" \\D5      rs\" \\E5      rt\" \\F5      ru\" \\G5      rv " S@ SA\5      rw " SB SC\5      rx " SD SE\5      ry " SF SG\5      rz " SH SI\5      r{ " SJ SK\5      r|\" \|5      r| " SL SM\5      r}\" \}SNSO9r} " SP SQ\5      r~\" \~SRSO9r~ " SS ST\5      r\" \SUSO9r " SV SW\5      r\" \SXSO9r " SY SZ\5      r\" \S[S\S]9r " S^ S_\5      r\" \S`SO9r " Sa Sb\5      r\" \ScSO9r " Sd Se\5      r\" \SfSgS]9r " Sh Si\5      r\" \SjSkS]9r " Sl Sm\5      r\" \SnSoS]9r " Sp Sq\5      r\" \SrSO9r " Ss St\5      r\" \SuSO9r " Sv Sw\5      r\" \SxSO9r " Sy Sz\5      r\" \S{SO9r " S| S}\5      r\" \S~SO9r " S S\5      r\" \SSO9r " S S\5      r\" \SSO9r " S S\5      r\" \SSO9r " S S\5      r\" \SSO9r " S S\5      r\" \SSO9r " S S\5      r\" \SSO9r " S S\5      r\" \SSO9r " S S\5      r\" \SSO9r " S S\5      r\" \SSO9r " S S\5      r\" \SSO9r " S S\5      r\" \SSO9r " S S\5      r\" \SSO9r " S S\5      r\" \SSO9r " S S\5      r\" \SSO9r " S S\5      r\" \SSO9r " S S\5      r " S S\5      r " S S\5      r " S S\5      r\" \SSO9r " S S\5      r\" \SSO9r " S S\5      r\" \SSO9r " S S\~5      r " S S\5      r/ SQrg)zAuto Model class.    N)OrderedDict)TYPE_CHECKINGUnion   )logging   )_BaseAutoBackboneClass_BaseAutoModelClass_LazyAutoMappingauto_class_update)CONFIG_MAPPING_NAMES)GenerationMixin)PreTrainedModelc                       \ rS rSrSrg)_BaseModelWithGenerate%    N)__name__
__module____qualname____firstlineno____static_attributes__r       `/home/james-whalen/.local/lib/python3.13/site-packages/transformers/models/auto/modeling_auto.pyr   r   %   s    r   r   (  )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)
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)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deta	DetaModel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efficientformerEfficientFormerModel)efficientloftrEfficientLoFTRModelefficientnetEfficientNetModelelectraElectraModel)emu3	Emu3Model)encodecEncodecModel)ernie
ErnieModel)ernie4_5Ernie4_5Model)ernie4_5_moeErnie4_5_MoeModel)ernie_mErnieMModel)esmEsmModel)evollaEvollaModel)exaone4Exaone4Model)falconFalconModel)	falcon_h1FalconH1Model)falcon_mambaFalconMambaModel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)glm4_moeGlm4MoeModel)glm4v
Glm4vModel)	glm4v_moeGlm4vMoeModel)glm4v_moe_textGlm4vMoeTextModel)
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)gpt_ossGptOssModel)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)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)jamba
JambaModel)janus
JanusModel)jetmoeJetMoeModel)jukeboxJukeboxModel)kosmos-2Kosmos2Model)
kosmos-2.5Kosmos2_5Model)kyutai_speech_to_textKyutaiSpeechToTextModel)layoutlmLayoutLMModel)
layoutlmv2LayoutLMv2Model)
layoutlmv3LayoutLMv3Model)ledLEDModellevit
LevitModel)lfm2	Lfm2Model)lfm2_vlLfm2VlModel	lightglueLightGlueForKeypointMatching)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)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
metaclip_2MetaClip2Model)zmgp-strMgpstrForSceneTextRecognition)mimi	MimiModel)minimaxMiniMaxModel)	ministralMinistralModel)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)	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)olmo3
Olmo3Model)olmoe
OlmoeModelzomdet-turboOmDetTurboForObjectDetection)	oneformerOneFormerModel)
open-llamaOpenLlamaModel)
openai-gptOpenAIGPTModel)optOPTModel)ovis2
Ovis2Model)owlv2
Owlv2Model)owlvitOwlViTModel)	paligemmaPaliGemmaModelparakeet_ctcParakeetForCTC)parakeet_encoderParakeetEncoder)patchtsmixerPatchTSMixerModel)patchtstPatchTSTModel)pegasusPegasusModel)	pegasus_xPegasusXModel)	perceiverPerceiverModel)perception_encoderPerceptionEncoder)perception_lmPerceptionLMModel)	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)
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	retribertRetriBertModel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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)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table-transformerTableTransformerModel)tapas
TapasModel)textnetTextNetModel)time_series_transformerTimeSeriesTransformerModel)timesfmTimesFmModeltimesformerTimesformerModeltimm_backboneTimmBackbonetimm_wrapperrD  )trajectory_transformerTrajectoryTransformerModel)
transfo-xlTransfoXLModel)tvlt	TvltModel)tvpTvpModel)udop	UdopModel)umt5	UMT5Model)	unispeechUniSpeechModel)unispeech-satUniSpeechSatModel)univnetUnivNetModelvanVanModel)
vaultgemmaVaultGemmaModel)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voxtralVoxtralForConditionalGeneration)voxtral_encoderVoxtralEncoder)wav2vec2Wav2Vec2Model)wav2vec2-bertWav2Vec2BertModel)wav2vec2-conformerWav2Vec2ConformerModel)wavlm
WavLMModel)whisperWhisperModel)xclip
XCLIPModel)xcodecXcodecModel)xglm	XGLMModelxlmXLMModel)xlm-prophetnetXLMProphetNetModelxlm-robertaXLMRobertaModelxlm-roberta-xlXLMRobertaXLModel)xlnet
XLNetModel)xlstm
xLSTMModel)xmod	XmodModelyolos
YolosModel)yoso	YosoModel)zamba
ZambaModel)zamba2Zamba2Model)X)r!   AlbertForPreTrainingr<   BartForConditionalGeneration)rB   BertForPreTraining)rG   BigBirdForPreTrainingr]   BloomForCausalLMre   CamembertForMaskedLMcolpaliColPaliForRetrieval)colqwen2ColQwen2ForRetrievalr   CTRLLMHeadModelr   Data2VecTextForMaskedLMr   DebertaForMaskedLMr   DebertaV2ForMaskedLMr   DistilBertForMaskedLM)r   ElectraForPreTraining)r  ErnieForPreTrainingr  EvollaForProteinText2Textr  Exaone4ForCausalLMr  FalconMambaForCausalLMr!  FlaubertWithLMHeadModel)r#  FlavaForPreTrainingr'  !Florence2ForConditionalGeneration)r)  FNetForPreTrainingr.  FSMTForConditionalGeneration)r0  FunnelForPreTrainingr9  Gemma3ForConditionalGenerationrZ  GPT2LMHeadModelr\  r  r]  GPTBigCodeForCausalLMri  )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  MegaForMaskedLM)r  MegatronBertForPreTrainingr   Mistral3ForConditionalGenerationr  MllamaForConditionalGeneration)r  MobileBertForPreTrainingr#  MPNetForMaskedLMr%  MptForCausalLMr'  MraForMaskedLMr/  MvpForConditionalGeneration)r6  NezhaForPreTrainingr8  NllbMoeForConditionalGenerationrK  OpenAIGPTLMHeadModelrU  !PaliGemmaForConditionalGenerationqwen2_audio"Qwen2AudioForConditionalGenerationr  r  RobertaForMaskedLMr  RobertaPreLayerNormForMaskedLM)r  RoCBertForPreTrainingr  RwkvForCausalLM)r  SplinterForPreTrainingr  SqueezeBertForMaskedLMr  *SwitchTransformersForConditionalGenerationr  T5ForConditionalGenerationr  T5GemmaForConditionalGenerationr  TapasForMaskedLMr*  TransfoXLLMHeadModel)r,  TvltForPreTraining)r4  UniSpeechForPreTraining)r6  UniSpeechSatForPreTrainingr?  "VideoLlavaForConditionalGeneration)rB  VideoMAEForPreTrainingrF   VipLlavaForConditionalGeneration)rI  VisualBertForPreTraining)rR  ViTMAEForPreTrainingrb  )rg  Wav2Vec2ForPreTraining)rk  Wav2Vec2ConformerForPreTrainingrx  XLMWithLMHeadModelr}  XLMRobertaForMaskedLMr  XLMRobertaXLForMaskedLMr  XLNetLMHeadModelr  xLSTMForCausalLMr  XmodForMaskedLM)Tr!   AlbertForMaskedLMr  rB   BertForMaskedLMrG   BigBirdForMaskedLMrI   &BigBirdPegasusForConditionalGenerationrT   'BlenderbotSmallForConditionalGenerationr  r  r   CodeGenForCausalLMr   ConvBertForMaskedLMr   CpmAntForCausalLMr  r  r  r  r   DiaForConditionalGenerationr  r   ElectraForMaskedLMzencoder-decoderEncoderDecoderModelr  ErnieForMaskedLMr  EsmForMaskedLMr  r  r  r)  FNetForMaskedLMr  r0  FunnelForMaskedLMrE  GitForCausalLMr  r  r  r_  GPTNeoForCausalLMra  GPTNeoXForCausalLMrc  GPTNeoXJapaneseForCausalLMrg  GPTJForCausalLMri  r  r  r  LEDForConditionalGenerationr  r  LongT5ForConditionalGenerationr  r  M2M100ForConditionalGenerationr  r  r  MarianMTModelr  r  MegatronBertForCausalLMr  MobileBertForMaskedLMr  !MoonshineForConditionalGenerationr  r  r  r  r6  NezhaForMaskedLMr  r;  NystromformerForMaskedLMr  rb   PegasusXForConditionalGenerationrv  PLBartForConditionalGeneration	pop2piano!Pop2PianoForConditionalGenerationr  QDQBertForMaskedLMr  ReformerModelWithLMHeadr  RemBertForMaskedLMr  r  r  RoCBertForMaskedLMr  RoFormerForMaskedLMr  r  #Speech2TextForConditionalGenerationr  r  r  r  r	  r  rg  Wav2Vec2ForMaskedLMro  WhisperForConditionalGenerationr  r  r  r  r#  r  YosoForMaskedLM))r)   ApertusForCausalLM)r+   ArceeForCausalLM)r/   AriaTextForCausalLM)r7   BambaForCausalLM)r<   BartForCausalLM)rB   BertLMHeadModel)rD   BertGenerationDecoder)rG   BigBirdForCausalLM)rI   BigBirdPegasusForCausalLM)rK   BioGptForCausalLM)rP   BitNetForCausalLM)rR   BlenderbotForCausalLM)rT   BlenderbotSmallForCausalLMr  )r_   BltForCausalLM)re   CamembertForCausalLM)r   LlamaForCausalLMr/  )r   CohereForCausalLM)r   Cohere2ForCausalLMr3  r  )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  FalconForCausalLM)r  FalconH1ForCausalLMr  )r%  FlexOlmoForCausalLMr3  FuyuForCausalLM)r5  GemmaForCausalLM)r7  Gemma2ForCausalLMr  )r;  Gemma3ForCausalLMr=  Gemma3nForConditionalGeneration)rA  Gemma3nForCausalLMrC  )rG  GlmForCausalLM)rI  Glm4ForCausalLM)rK  Glm4MoeForCausalLMrX  GotOcr2ForConditionalGenerationr  r  r  rE  rG  rI  )re  GptOssForCausalLMrK  )rk  GraniteForCausalLM)rm  GraniteMoeForCausalLM)ro  GraniteMoeHybridForCausalLM)rq  GraniteMoeSharedForCausalLM)ry  HeliumForCausalLM)r  HunYuanDenseV1ForCausalLM)r  HunYuanMoEV1ForCausalLM)r  JambaForCausalLM)r  JetMoeForCausalLM)r  Lfm2ForCausalLM)r  r  )r  Llama4ForCausalLM)r  r  )r  LongcatFlashForCausalLMr  r  )r  MarianForCausalLM)r  MBartForCausalLM)r  MegaForCausalLMrU  )r  MiniMaxForCausalLM)r  MinistralForCausalLM)r  MistralForCausalLM)r  MixtralForCausalLM)r  MllamaForCausalLM)r  ModernBertDecoderForCausalLM)r!  MoshiForCausalLMr  )r+  MusicgenForCausalLM)r-  MusicgenMelodyForCausalLM)r/  MvpForCausalLM)r4  NemotronForCausalLM)r=  OlmoForCausalLM)r?  Olmo2ForCausalLM)rA  Olmo3ForCausalLM)rC  OlmoeForCausalLM)rI  OpenLlamaForCausalLMr  )rM  OPTForCausalLM)r`  PegasusForCausalLM)rj  PersimmonForCausalLM)rl  PhiForCausalLM)rn  Phi3ForCausalLM)rp  Phi4MultimodalForCausalLM)rr  PhimoeForCausalLM)rv  PLBartForCausalLM)r{  ProphetNetForCausalLM)r  QDQBertLMHeadModel)r  Qwen2ForCausalLM)r  Qwen2MoeForCausalLM)r  Qwen3ForCausalLM)r  Qwen3MoeForCausalLM)r  Qwen3NextForCausalLM)r  RecurrentGemmaForCausalLMrh  )r  RemBertForCausalLM)r  RobertaForCausalLM)r  RobertaPreLayerNormForCausalLM)r  RoCBertForCausalLM)r  RoFormerForCausalLMr  )r  SeedOssForCausalLM)r  SmolLM3ForCausalLM)speech_to_text_2Speech2Text2ForCausalLM)r  StableLmForCausalLM)r  Starcoder2ForCausalLMr  )trocrTrOCRForCausalLM)r=  VaultGemmaForCausalLM)ro  WhisperForCausalLM)ru  XGLMForCausalLMr  )rz  XLMProphetNetForCausalLM)r}  XLMRobertaForCausalLM)r  XLMRobertaXLForCausalLMr  r!  )r  XmodForCausalLM)r  ZambaForCausalLM)r  Zamba2ForCausalLM);r   r>   rM   r   r   r   r   r   r   r   r   r   r   r   r   r   r   r   r   r   r   r   r+  rU  r~  r  r  r  )r  Llama4VisionModelr  )r  MllamaVisionModelr  r  r  r  r1  rx  r}  r  r  r  r  r  r  r	  r  r  r   r#  r&  r:  rA  rK  rN  rQ  rT  rW  r]  r  ))r   DeiTForMaskedImageModeling)r,  FocalNetForMaskedImageModeling)r  SwinForMaskedImageModeling)r  Swinv2ForMaskedImageModeling)rL  ViTForMaskedImageModeling)r  ImageGPTForCausalImageModeling)-)r?   BeitForImageClassification)rN   BitForImageClassification)rs   CLIPForImageClassification)r   ConvNextForImageClassification)r    ConvNextV2ForImageClassification)r   CvtForImageClassification)r   $Data2VecVisionForImageClassification)r   )DeiTForImageClassification%DeiTForImageClassificationWithTeacher)r   DinatForImageClassification)r   Dinov2ForImageClassification)r   )Dinov2WithRegistersForImageClassification)r   DonutSwinForImageClassification)r   )%EfficientFormerForImageClassification0EfficientFormerForImageClassificationWithTeacher)r   "EfficientNetForImageClassification)r,  FocalNetForImageClassification)r|  HGNetV2ForImageClassification)r  HieraForImageClassification)r  IJepaForImageClassification)r  ImageGPTForImageClassification)r  )LevitForImageClassification&LevitForImageClassificationWithTeacher)r  MetaClip2ForImageClassification)r  !MobileNetV1ForImageClassification)r  !MobileNetV2ForImageClassification)r  MobileViTForImageClassification)r  !MobileViTV2ForImageClassification)r2  NatForImageClassification)rd  )&PerceiverForImageClassificationLearned&PerceiverForImageClassificationFourier-PerceiverForImageClassificationConvProcessing)ry   PoolFormerForImageClassification)r~  PvtForImageClassification)r  PvtV2ForImageClassification)r  RegNetForImageClassification)r  ResNetForImageClassification)r  SegformerForImageClassification)shieldgemma2"ShieldGemma2ForImageClassification)r  SiglipForImageClassification)r  Siglip2ForImageClassification)r  !SwiftFormerForImageClassification)r  SwinForImageClassification)r  Swinv2ForImageClassification)r  TextNetForImageClassification)r'  !TimmWrapperForImageClassification)r;  VanForImageClassification)rL  ViTForImageClassification)rO  ViTHybridForImageClassification)rU  ViTMSNForImageClassificationr   DetrForSegmentation))r?   BeitForSemanticSegmentation)r   %Data2VecVisionForSemanticSegmentation)r   DPTForSemanticSegmentation)r  "MobileNetV2ForSemanticSegmentation)r   MobileViTForSemanticSegmentation)r  "MobileViTV2ForSemanticSegmentation)r   SegformerForSemanticSegmentation)upernetUperNetForSemanticSegmentationr  !MaskFormerForInstanceSegmentation)r'  )eomtEomtForUniversalSegmentation)r  #Mask2FormerForUniversalSegmentationr2  )rG  !OneFormerForUniversalSegmentation))r!  !TimesformerForVideoClassification)rB  VideoMAEForVideoClassification)r^  VivitForVideoClassification)r`  VJEPA2ForVideoClassification)rW   BlipForConditionalGenerationrY   Blip2ForConditionalGenerationri   !ChameleonForConditionalGenerationrC  r  r  r  $InstructBlipForConditionalGeneration)r  )InstructBlipVideoForConditionalGenerationr  Kosmos2ForConditionalGenerationr  !Kosmos2_5ForConditionalGenerationr  r  r  r  r  r  rO  Ovis2ForConditionalGenerationr  
pix2struct"Pix2StructForConditionalGenerationr  "Qwen2_5_VLForConditionalGenerationr  Qwen2VLForConditionalGenerationr  Qwen3VLForConditionalGenerationr  "Qwen3VLMoeForConditionalGenerationr  r  zvision-encoder-decoderVisionEncoderDecoderModelr  )0)r-   AriaForConditionalGeneration)r5   !AyaVisionForConditionalGenerationr<  r>  r@  )r   %Cohere2VisionForConditionalGeneration)r   "DeepseekVLForConditionalGeneration)r   (DeepseekVLHybridForConditionalGeneration)r  Emu3ForConditionalGenerationr  r  r  r  r  rC  )rM  Glm4vForConditionalGeneration)rO   Glm4vMoeForConditionalGenerationr  r  r  r  rB  )r   InternVLForConditionalGenerationr  rE  rG  )r  Lfm2VlForConditionalGenerationr  r  r  r  r  r  r  rI  r  )rh  $PerceptionLMForConditionalGenerationrK  )rt  r  rN  rP  rR  rT  )r  r  )r  SmolVLMForConditionalGeneration)r0  UdopForConditionalGenerationr  rV  ).r%  r  r'  r)  r  r1  r  r  r  r  r7  r;  r=  r  r?  rA  r  r  r  r  r  MBartForConditionalGenerationr  )r  MegatronBertForMaskedLMrW  )r  ModernBertForMaskedLMr  r  r  r[  r]  )rd  PerceiverForMaskedLMrf  )r  ReformerForMaskedLMrj  r  r  rl  rn  r  r	  rr  r  r  r  r#  rv  )
)r   !ConditionalDetrForObjectDetection)r   DFineForObjectDetection)r   DabDetrForObjectDetection)r    DeformableDetrForObjectDetection)r   DetaForObjectDetection)r   DetrForObjectDetection)r  RTDetrForObjectDetection)r  RTDetrV2ForObjectDetection)r  "TableTransformerForObjectDetection)r  YolosForObjectDetection))ru  GroundingDinoForObjectDetection)r
  !MMGroundingDinoForObjectDetectionrE  )rQ  Owlv2ForObjectDetection)rS  OwlViTForObjectDetection))depth_anythingDepthAnythingForDepthEstimation)r   DepthProForDepthEstimation)r   DPTForDepthEstimation)rV  GLPNForDepthEstimation)prompt_depth_anything%PromptDepthAnythingForDepthEstimation)zoedepthZoeDepthForDepthEstimation)r  r+  )rR   "BlenderbotForConditionalGenerationr-  r9  r  ri  granite_speech%GraniteSpeechForConditionalGenerationrM  rO  rQ  rS  re  )r)  MT5ForConditionalGenerationr  r  )r`  PegasusForConditionalGenerationr_  ra  )r{  "ProphetNetForConditionalGenerationr  )r  SeamlessM4TForTextToText)r  SeamlessM4Tv2ForTextToTextr  r  r  )r2  UMT5ForConditionalGenerationrb  )rz  %XLMProphetNetForConditionalGeneration)r5  r  )r  *KyutaiSpeechToTextForConditionalGenerationrY  rc  )r  SeamlessM4TForSpeechToText)r  SeamlessM4Tv2ForSpeechToText)zspeech-encoder-decoderSpeechEncoderDecoderModelrp  )r  SpeechT5ForSpeechToTextrt  )s)r!   AlbertForSequenceClassification)r+   ArceeForSequenceClassification)r<   BartForSequenceClassification)rB   BertForSequenceClassification)rG    BigBirdForSequenceClassification)rI   'BigBirdPegasusForSequenceClassification)rK   BioGptForSequenceClassification)r]   BloomForSequenceClassification)re   "CamembertForSequenceClassification)rg   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  ErnieMForSequenceClassification)r  EsmForSequenceClassification)r   Exaone4ForSequenceClassification)r  FalconForSequenceClassification)r!  !FlaubertForSequenceClassification)r)  FNetForSequenceClassification)r0  FunnelForSequenceClassification)r5  GemmaForSequenceClassification)r7  Gemma2ForSequenceClassification)r9  Gemma3ForSequenceClassification)r;  #Gemma3TextForSequenceClassification)rG  GlmForSequenceClassification)rI  Glm4ForSequenceClassification)rZ  GPT2ForSequenceClassification)r\  r  )r]  #GPTBigCodeForSequenceClassification)r_  GPTNeoForSequenceClassification)ra   GPTNeoXForSequenceClassification)re  GptOssForSequenceClassification)rg  GPTJForSequenceClassification)ry  HeliumForSequenceClassification)r  'HunYuanDenseV1ForSequenceClassification)r  %HunYuanMoEV1ForSequenceClassification)r  IBertForSequenceClassification)r  JambaForSequenceClassification)r  JetMoeForSequenceClassification)r  !LayoutLMForSequenceClassification)r  #LayoutLMv2ForSequenceClassification)r  #LayoutLMv3ForSequenceClassification)r  LEDForSequenceClassification)r  LiltForSequenceClassification)r  r  )r  #LongformerForSequenceClassification)r  LukeForSequenceClassification)r  !MarkupLMForSequenceClassification)r  MBartForSequenceClassification)r  MegaForSequenceClassification)r  %MegatronBertForSequenceClassification)r   MiniMaxForSequenceClassification)r  "MinistralForSequenceClassification)r   MistralForSequenceClassification)r   MixtralForSequenceClassification)r  #MobileBertForSequenceClassification)r  #ModernBertForSequenceClassification)r  *ModernBertDecoderForSequenceClassification)r#  MPNetForSequenceClassification)r%  MptForSequenceClassification)r'  MraForSequenceClassification)r)  MT5ForSequenceClassification)r/  MvpForSequenceClassification)r4  !NemotronForSequenceClassification)r6  NezhaForSequenceClassification)r;  &NystromformerForSequenceClassification)rI  "OpenLlamaForSequenceClassification)rK  "OpenAIGPTForSequenceClassification)rM  OPTForSequenceClassification)rd  "PerceiverForSequenceClassification)rj  "PersimmonForSequenceClassification)rl  PhiForSequenceClassification)rn  Phi3ForSequenceClassification)rr  PhimoeForSequenceClassification)rv  PLBartForSequenceClassification)r   QDQBertForSequenceClassification)r  Qwen2ForSequenceClassification)r  !Qwen2MoeForSequenceClassification)r  Qwen3ForSequenceClassification)r  !Qwen3MoeForSequenceClassification)r  "Qwen3NextForSequenceClassification)r  !ReformerForSequenceClassification)r   RemBertForSequenceClassification)r   RobertaForSequenceClassification)r  ,RobertaPreLayerNormForSequenceClassification)r   RoCBertForSequenceClassification)r  !RoFormerForSequenceClassification)r   SeedOssForSequenceClassification)r   SmolLM3ForSequenceClassification)r  $SqueezeBertForSequenceClassification)r  !StableLmForSequenceClassification)r  #Starcoder2ForSequenceClassification)r  T5ForSequenceClassification)r   T5GemmaForSequenceClassification)r  TapasForSequenceClassification)r*  "TransfoXLForSequenceClassification)r2  UMT5ForSequenceClassification)rx  XLMForSequenceClassification)r}  #XLMRobertaForSequenceClassification)r  %XLMRobertaXLForSequenceClassification)r  XLNetForSequenceClassification)r  XmodForSequenceClassification)r  YosoForSequenceClassification)r  ZambaForSequenceClassification)r  Zamba2ForSequenceClassification)O)r!   AlbertForQuestionAnswering)r+   ArceeForQuestionAnswering)r<   BartForQuestionAnswering)rB   BertForQuestionAnswering)rG   BigBirdForQuestionAnswering)rI   "BigBirdPegasusForQuestionAnswering)r]   BloomForQuestionAnswering)re   CamembertForQuestionAnswering)rg   CanineForQuestionAnswering)r   ConvBertForQuestionAnswering)r    Data2VecTextForQuestionAnswering)r   DebertaForQuestionAnswering)r   DebertaV2ForQuestionAnswering)r   DiffLlamaForQuestionAnswering)r   DistilBertForQuestionAnswering)r   ElectraForQuestionAnswering)r  ErnieForQuestionAnswering)r  ErnieMForQuestionAnswering)r  Exaone4ForQuestionAnswering)r  FalconForQuestionAnswering)r!  "FlaubertForQuestionAnsweringSimple)r)  FNetForQuestionAnswering)r0  FunnelForQuestionAnswering)r\  GPT2ForQuestionAnswering)r_  GPTNeoForQuestionAnswering)ra  GPTNeoXForQuestionAnswering)rg  GPTJForQuestionAnswering)r  IBertForQuestionAnsweringr  LayoutLMv2ForQuestionAnsweringr  LayoutLMv3ForQuestionAnswering)r  LEDForQuestionAnswering)r  LiltForQuestionAnswering)r  LlamaForQuestionAnswering)r  LongformerForQuestionAnswering)r  LukeForQuestionAnswering)r  LxmertForQuestionAnswering)r  MarkupLMForQuestionAnswering)r  MBartForQuestionAnswering)r  MegaForQuestionAnswering)r   MegatronBertForQuestionAnswering)r  MiniMaxForQuestionAnswering)r  MinistralForQuestionAnswering)r  MistralForQuestionAnswering)r  MixtralForQuestionAnswering)r  MobileBertForQuestionAnswering)r  ModernBertForQuestionAnswering)r#  MPNetForQuestionAnswering)r%  MptForQuestionAnswering)r'  MraForQuestionAnswering)r)  MT5ForQuestionAnswering)r/  MvpForQuestionAnswering)r4  NemotronForQuestionAnswering)r6  NezhaForQuestionAnswering)r;  !NystromformerForQuestionAnswering)rM  OPTForQuestionAnswering)r  QDQBertForQuestionAnswering)r  Qwen2ForQuestionAnswering)r  Qwen2MoeForQuestionAnswering)r  Qwen3ForQuestionAnswering)r  Qwen3MoeForQuestionAnswering)r  Qwen3NextForQuestionAnswering)r  ReformerForQuestionAnswering)r  RemBertForQuestionAnswering)r  RobertaForQuestionAnswering)r  'RobertaPreLayerNormForQuestionAnswering)r  RoCBertForQuestionAnswering)r  RoFormerForQuestionAnswering)r  SeedOssForQuestionAnswering)r  SmolLM3ForQuestionAnswering)r  SplinterForQuestionAnswering)r  SqueezeBertForQuestionAnswering)r  T5ForQuestionAnswering)r2  UMT5ForQuestionAnswering)rx  XLMForQuestionAnsweringSimple)r}  XLMRobertaForQuestionAnswering)r   XLMRobertaXLForQuestionAnswering)r  XLNetForQuestionAnsweringSimple)r  XmodForQuestionAnswering)r  YosoForQuestionAnswering)r  TapasForQuestionAnswering))rW   BlipForQuestionAnsweringr>  )rD  ViltForQuestionAnswering))r  LayoutLMForQuestionAnsweringr  r!  )Y)r!   AlbertForTokenClassification)r)   ApertusForTokenClassification)r+   ArceeForTokenClassification)rB   BertForTokenClassification)rG   BigBirdForTokenClassification)rK   BioGptForTokenClassification)r]   BloomForTokenClassification)rc   BrosForTokenClassification)re   CamembertForTokenClassification)rg   CanineForTokenClassification)r   ConvBertForTokenClassification)r   "Data2VecTextForTokenClassification)r   DebertaForTokenClassification)r   DebertaV2ForTokenClassification)r    DeepseekV3ForTokenClassification)r   DiffLlamaForTokenClassification)r    DistilBertForTokenClassification)r   ElectraForTokenClassification)r  ErnieForTokenClassification)r  ErnieMForTokenClassification)r  EsmForTokenClassification)r  Exaone4ForTokenClassification)r  FalconForTokenClassification)r!  FlaubertForTokenClassification)r)  FNetForTokenClassification)r0  FunnelForTokenClassification)r5  GemmaForTokenClassification)r7  Gemma2ForTokenClassification)rG  GlmForTokenClassification)rI  Glm4ForTokenClassification)rZ  GPT2ForTokenClassification)r\  rv  )r]   GPTBigCodeForTokenClassification)r_  GPTNeoForTokenClassification)ra  GPTNeoXForTokenClassification)re  GptOssForTokenClassification)ry  HeliumForTokenClassification)r  IBertForTokenClassification)r  LayoutLMForTokenClassification)r   LayoutLMv2ForTokenClassification)r   LayoutLMv3ForTokenClassification)r  LiltForTokenClassification)r  LlamaForTokenClassification)r   LongformerForTokenClassification)r  LukeForTokenClassification)r  MarkupLMForTokenClassification)r  MegaForTokenClassification)r  "MegatronBertForTokenClassification)r  MiniMaxForTokenClassification)r  MinistralForTokenClassification)r  MistralForTokenClassification)r  MixtralForTokenClassification)r   MobileBertForTokenClassification)r   ModernBertForTokenClassification)r#  MPNetForTokenClassification)r%  MptForTokenClassification)r'  MraForTokenClassification)r)  MT5ForTokenClassification)r4  NemotronForTokenClassification)r6  NezhaForTokenClassification)r;  #NystromformerForTokenClassification)rj  PersimmonForTokenClassification)rl  PhiForTokenClassification)rn  Phi3ForTokenClassification)r  QDQBertForTokenClassification)r  Qwen2ForTokenClassification)r  Qwen2MoeForTokenClassification)r  Qwen3ForTokenClassification)r  Qwen3MoeForTokenClassification)r  Qwen3NextForTokenClassification)r  RemBertForTokenClassification)r  RobertaForTokenClassification)r  )RobertaPreLayerNormForTokenClassification)r  RoCBertForTokenClassification)r  RoFormerForTokenClassification)r  SeedOssForTokenClassification)r  SmolLM3ForTokenClassification)r  !SqueezeBertForTokenClassification)r  StableLmForTokenClassification)r   Starcoder2ForTokenClassification)r  T5ForTokenClassification)r  T5GemmaForTokenClassification)r2  UMT5ForTokenClassification)rx  XLMForTokenClassification)r}   XLMRobertaForTokenClassification)r  "XLMRobertaXLForTokenClassification)r  XLNetForTokenClassification)r  XmodForTokenClassification)r  YosoForTokenClassification)')r!   AlbertForMultipleChoice)rB   BertForMultipleChoice)rG   BigBirdForMultipleChoice)re   CamembertForMultipleChoice)rg   CanineForMultipleChoice)r   ConvBertForMultipleChoice)r   Data2VecTextForMultipleChoice)r   DebertaV2ForMultipleChoice)r   DistilBertForMultipleChoice)r   ElectraForMultipleChoice)r  ErnieForMultipleChoice)r  ErnieMForMultipleChoice)r!  FlaubertForMultipleChoice)r)  FNetForMultipleChoice)r0  FunnelForMultipleChoice)r  IBertForMultipleChoice)r  LongformerForMultipleChoice)r  LukeForMultipleChoice)r  MegaForMultipleChoice)r  MegatronBertForMultipleChoice)r  MobileBertForMultipleChoice)r  ModernBertForMultipleChoice)r#  MPNetForMultipleChoice)r'  MraForMultipleChoice)r6  NezhaForMultipleChoice)r;  NystromformerForMultipleChoice)r  QDQBertForMultipleChoice)r  RemBertForMultipleChoice)r  RobertaForMultipleChoice)r  $RobertaPreLayerNormForMultipleChoice)r  RoCBertForMultipleChoice)r  RoFormerForMultipleChoice)r  SqueezeBertForMultipleChoice)rx  XLMForMultipleChoice)r}  XLMRobertaForMultipleChoice)r  XLMRobertaXLForMultipleChoice)r  XLNetForMultipleChoice)r  XmodForMultipleChoice)r  YosoForMultipleChoice))rB   BertForNextSentencePrediction)r  ErnieForNextSentencePrediction)r)  FNetForNextSentencePrediction)r  %MegatronBertForNextSentencePrediction)r  #MobileBertForNextSentencePrediction)r6  NezhaForNextSentencePrediction)r   QDQBertForNextSentencePrediction))r1   ASTForAudioClassification)r   &Data2VecAudioForSequenceClassification)r  HubertForSequenceClassification)r  SEWForSequenceClassification)r  SEWDForSequenceClassification)r4  "UniSpeechForSequenceClassification)r6  %UniSpeechSatForSequenceClassification)rg  !Wav2Vec2ForSequenceClassification)ri  %Wav2Vec2BertForSequenceClassification)rk  *Wav2Vec2ConformerForSequenceClassification)rm  WavLMForSequenceClassification)ro  WhisperForAudioClassification))r   Data2VecAudioForCTC)r  HubertForCTC)r  MCTCTForCTCrW  )r  	SEWForCTC)r  
SEWDForCTC)r4  UniSpeechForCTC)r6  UniSpeechSatForCTC)rg  Wav2Vec2ForCTC)ri  Wav2Vec2BertForCTC)rk  Wav2Vec2ConformerForCTC)rm  WavLMForCTC))r   (Data2VecAudioForAudioFrameClassification)r6  'UniSpeechSatForAudioFrameClassification)rg  #Wav2Vec2ForAudioFrameClassification)ri  'Wav2Vec2BertForAudioFrameClassification)rk  ,Wav2Vec2ConformerForAudioFrameClassification)rm   WavLMForAudioFrameClassification))r   Data2VecAudioForXVector)r6  UniSpeechSatForXVector)rg  Wav2Vec2ForXVector)ri  Wav2Vec2BertForXVector)rk  Wav2Vec2ConformerForXVector)rm  WavLMForXVectorr  )r  SpeechT5ForTextToSpeech)r9   r   )r  r  r  )r+   MusicgenForConditionalGeneration)r-  &MusicgenMelodyForConditionalGeneration)qwen2_5_omni#Qwen2_5OmniForConditionalGeneration)qwen3_omni_moe$Qwen3OmniMoeForConditionalGeneration)r  SeamlessM4TForTextToSpeech)r  SeamlessM4Tv2ForTextToSpeechrZ  )
r#   r&   rV   )rY   Blip2ForImageTextRetrievalrk   rr   rz   r  r  r  ))r?   BeitBackbone)rN   BitBackbone)r   ConvNextBackbone)r   ConvNextV2Backbone)r   DinatBackbone)r   Dinov2Backbone)r   Dinov2WithRegistersBackbone)r,  FocalNetBackboner{  )r  HieraBackbone)r  MaskFormerSwinBackbone)r2  NatBackbone)r  PvtV2Backbone)r  ResNetBackbone)rt_detr_resnetRTDetrResNetBackbone)r  SwinBackbone)r  Swinv2Backbone)r  TextNetBackboner#  )rX  VitDetBackbone)vitpose_backboneVitPoseBackbone)r   )r   r   r  r  )r  r  r  )
superpointSuperPointForKeypointDetection))r   !EfficientLoFTRForKeypointMatchingr  )	superglueSuperGlueForKeypointMatching)r    rA   rF   ru   r   r   r   r   r   )r  Emu3TextModelr   r  )r  r  r  )r  MllamaTextModelr  )r)  MT5EncoderModelr:  r  r  r  r  r  r  r  )r  T5EncoderModel)r  T5GemmaEncoderModel)r2  UMT5EncoderModelrw  r|  r  )r\  'PatchTSMixerForTimeSeriesClassification)r^  PatchTSTForClassification)r\  PatchTSMixerForRegression)r^  PatchTSTForRegression)r  TimesFmModelForPrediction)r
  Swin2SRForImageSuperResolutionr   c                       \ rS rSr\rSrg)AutoModelForMaskGenerationi  r   N)r   r   r   r   !MODEL_FOR_MASK_GENERATION_MAPPING_model_mappingr   r   r   r   r2  r2        6Nr   r2  c                       \ rS rSr\rSrg)AutoModelForKeypointDetectioni  r   N)r   r   r   r   $MODEL_FOR_KEYPOINT_DETECTION_MAPPINGr4  r   r   r   r   r7  r7        9Nr   r7  c                       \ rS rSr\rSrg)AutoModelForKeypointMatchingi  r   N)r   r   r   r   #MODEL_FOR_KEYPOINT_MATCHING_MAPPINGr4  r   r   r   r   r;  r;    s    8Nr   r;  c                       \ rS rSr\rSrg)AutoModelForTextEncodingi  r   N)r   r   r   r   MODEL_FOR_TEXT_ENCODING_MAPPINGr4  r   r   r   r   r>  r>        4Nr   r>  c                       \ rS rSr\rSrg)AutoModelForImageToImagei  r   N)r   r   r   r    MODEL_FOR_IMAGE_TO_IMAGE_MAPPINGr4  r   r   r   r   rB  rB    s    5Nr   rB  c                       \ rS rSr\rSrg)	AutoModeli  r   N)r   r   r   r   MODEL_MAPPINGr4  r   r   r   r   rE  rE    s    "Nr   rE  c                       \ rS rSr\rSrg)AutoModelForPreTrainingi  r   N)r   r   r   r   MODEL_FOR_PRETRAINING_MAPPINGr4  r   r   r   r   rH  rH    s    2Nr   rH  pretraining)head_docc                       \ rS rSr\rSrg)_AutoModelWithLMHeadi  r   N)r   r   r   r   MODEL_WITH_LM_HEAD_MAPPINGr4  r   r   r   r   rM  rM        /Nr   rM  zlanguage modelingc                   p   ^  \ rS rSr\r\S\S    S\\	\
R                  \	   4   SS4U 4S jj5       rSrU =r$ )AutoModelForCausalLMi  clspretrained_model_name_or_pathreturnr   c                 ,   > [         TU ]  " U/UQ70 UD6$ Nsuperfrom_pretrainedrR  rS  
model_argskwargs	__class__s       r   rY  $AutoModelForCausalLM.from_pretrained  !     w&'D\z\U[\\r   r   )r   r   r   r   MODEL_FOR_CAUSAL_LM_MAPPINGr4  classmethodtyper   strosPathLikerY  r   __classcell__r]  s   @r   rQ  rQ    sT    0N ]()]',S"++c2B-B'C]
 
"] ]r   rQ  zcausal language modelingc                       \ rS rSr\rSrg)AutoModelForMaskedLMi  r   N)r   r   r   r   MODEL_FOR_MASKED_LM_MAPPINGr4  r   r   r   r   ri  ri    s    0Nr   ri  zmasked language modelingc                       \ rS rSr\rSrg)AutoModelForSeq2SeqLMi  r   N)r   r   r   r   &MODEL_FOR_SEQ_TO_SEQ_CAUSAL_LM_MAPPINGr4  r   r   r   r   rl  rl        ;Nr   rl  z&sequence-to-sequence language modelingzgoogle-t5/t5-base)rK  checkpoint_for_examplec                       \ rS rSr\rSrg)"AutoModelForSequenceClassificationi  r   N)r   r   r   r   )MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPINGr4  r   r   r   r   rq  rq    s    >Nr   rq  zsequence classificationc                       \ rS rSr\rSrg)AutoModelForQuestionAnsweringi  r   N)r   r   r   r   $MODEL_FOR_QUESTION_ANSWERING_MAPPINGr4  r   r   r   r   rt  rt    r9  r   rt  zquestion answeringc                       \ rS rSr\rSrg)"AutoModelForTableQuestionAnsweringi  r   N)r   r   r   r   *MODEL_FOR_TABLE_QUESTION_ANSWERING_MAPPINGr4  r   r   r   r   rw  rw        ?Nr   rw  ztable question answeringzgoogle/tapas-base-finetuned-wtqc                       \ rS rSr\rSrg)#AutoModelForVisualQuestionAnsweringi  r   N)r   r   r   r   +MODEL_FOR_VISUAL_QUESTION_ANSWERING_MAPPINGr4  r   r   r   r   r{  r{    s    @Nr   r{  zvisual question answeringzdandelin/vilt-b32-finetuned-vqac                       \ rS rSr\rSrg)%AutoModelForDocumentQuestionAnsweringi  r   N)r   r   r   r   -MODEL_FOR_DOCUMENT_QUESTION_ANSWERING_MAPPINGr4  r   r   r   r   r~  r~    s    BNr   r~  zdocument question answeringz/impira/layoutlm-document-qa", revision="52e01b3c                       \ rS rSr\rSrg)AutoModelForTokenClassificationi  r   N)r   r   r   r   &MODEL_FOR_TOKEN_CLASSIFICATION_MAPPINGr4  r   r   r   r   r  r    rn  r   r  ztoken classificationc                       \ rS rSr\rSrg)AutoModelForMultipleChoicei  r   N)r   r   r   r   !MODEL_FOR_MULTIPLE_CHOICE_MAPPINGr4  r   r   r   r   r  r    r5  r   r  zmultiple choicec                       \ rS rSr\rSrg)"AutoModelForNextSentencePredictioni  r   N)r   r   r   r   *MODEL_FOR_NEXT_SENTENCE_PREDICTION_MAPPINGr4  r   r   r   r   r  r    ry  r   r  znext sentence predictionc                       \ rS rSr\rSrg)AutoModelForImageClassificationi  r   N)r   r   r   r   &MODEL_FOR_IMAGE_CLASSIFICATION_MAPPINGr4  r   r   r   r   r  r    rn  r   r  zimage classificationc                       \ rS rSr\rSrg)'AutoModelForZeroShotImageClassificationi  r   N)r   r   r   r   0MODEL_FOR_ZERO_SHOT_IMAGE_CLASSIFICATION_MAPPINGr4  r   r   r   r   r  r    s    ENr   r  zzero-shot image classificationc                       \ rS rSr\rSrg)AutoModelForImageSegmentationi%  r   N)r   r   r   r   $MODEL_FOR_IMAGE_SEGMENTATION_MAPPINGr4  r   r   r   r   r  r  %  r9  r   r  zimage segmentationc                       \ rS rSr\rSrg) AutoModelForSemanticSegmentationi,  r   N)r   r   r   r   'MODEL_FOR_SEMANTIC_SEGMENTATION_MAPPINGr4  r   r   r   r   r  r  ,      <Nr   r  zsemantic segmentationc                       \ rS rSr\rSrg) AutoModelForTimeSeriesPredictioni5  r   N)r   r   r   r   (MODEL_FOR_TIME_SERIES_PREDICTION_MAPPINGr4  r   r   r   r   r  r  5      =Nr   r  ztime-series predictionc                       \ rS rSr\rSrg)!AutoModelForUniversalSegmentationi>  r   N)r   r   r   r   (MODEL_FOR_UNIVERSAL_SEGMENTATION_MAPPINGr4  r   r   r   r   r  r  >  r  r   r  zuniversal image segmentationc                       \ rS rSr\rSrg) AutoModelForInstanceSegmentationiG  r   N)r   r   r   r   'MODEL_FOR_INSTANCE_SEGMENTATION_MAPPINGr4  r   r   r   r   r  r  G  r  r   r  zinstance segmentationc                       \ rS rSr\rSrg)AutoModelForObjectDetectioniP  r   N)r   r   r   r   "MODEL_FOR_OBJECT_DETECTION_MAPPINGr4  r   r   r   r   r  r  P      7Nr   r  zobject detectionc                       \ rS rSr\rSrg)#AutoModelForZeroShotObjectDetectioniW  r   N)r   r   r   r   ,MODEL_FOR_ZERO_SHOT_OBJECT_DETECTION_MAPPINGr4  r   r   r   r   r  r  W      ANr   r  zzero-shot object detectionc                       \ rS rSr\rSrg)AutoModelForDepthEstimationi`  r   N)r   r   r   r   "MODEL_FOR_DEPTH_ESTIMATION_MAPPINGr4  r   r   r   r   r  r  `  r  r   r  zdepth estimationc                       \ rS rSr\rSrg)AutoModelForVideoClassificationig  r   N)r   r   r   r   &MODEL_FOR_VIDEO_CLASSIFICATION_MAPPINGr4  r   r   r   r   r  r  g  rn  r   r  zvideo classificationc                       \ rS rSr\rSrg)_AutoModelForVision2Seqio  r   N)r   r   r   r   MODEL_FOR_VISION_2_SEQ_MAPPINGr4  r   r   r   r   r  r  o  s    3Nr   r  zvision-to-text modelingc                   p   ^  \ rS rSr\r\S\S    S\\	\
R                  \	   4   SS4U 4S jj5       rSrU =r$ )AutoModelForImageTextToTextiv  rR  rS  rT  r   c                 ,   > [         TU ]  " U/UQ70 UD6$ rV  rW  rZ  s       r   rY  +AutoModelForImageTextToText.from_pretrainedz  r_  r   r   )r   r   r   r   $MODEL_FOR_IMAGE_TEXT_TO_TEXT_MAPPINGr4  ra  rb  r   rc  rd  re  rY  r   rf  rg  s   @r   r  r  v  sT    9N ]/0]',S"++c2B-B'C]
 
"] ]r   r  zimage-text-to-text modelingc                       \ rS rSr\rSrg)AutoModelForAudioClassificationi  r   N)r   r   r   r   &MODEL_FOR_AUDIO_CLASSIFICATION_MAPPINGr4  r   r   r   r   r  r    rn  r   r  zaudio classificationc                       \ rS rSr\rSrg)AutoModelForCTCi  r   N)r   r   r   r   MODEL_FOR_CTC_MAPPINGr4  r   r   r   r   r  r    s    *Nr   r  z%connectionist temporal classificationc                       \ rS rSr\rSrg)AutoModelForSpeechSeq2Seqi  r   N)r   r   r   r   "MODEL_FOR_SPEECH_SEQ_2_SEQ_MAPPINGr4  r   r   r   r   r  r    r  r   r  z,sequence-to-sequence speech-to-text modelingc                       \ rS rSr\rSrg)$AutoModelForAudioFrameClassificationi  r   N)r   r   r   r   ,MODEL_FOR_AUDIO_FRAME_CLASSIFICATION_MAPPINGr4  r   r   r   r   r  r    r  r   r  z"audio frame (token) classificationc                       \ rS rSr\rSrg)AutoModelForAudioXVectori  r   N)r   r   r   r   MODEL_FOR_AUDIO_XVECTOR_MAPPINGr4  r   r   r   r   r  r    r@  r   r  c                       \ rS rSr\rSrg)AutoModelForTextToSpectrogrami  r   N)r   r   r   r   %MODEL_FOR_TEXT_TO_SPECTROGRAM_MAPPINGr4  r   r   r   r   r  r    s    :Nr   r  c                       \ rS rSr\rSrg)AutoModelForTextToWaveformi  r   N)r   r   r   r   "MODEL_FOR_TEXT_TO_WAVEFORM_MAPPINGr4  r   r   r   r   r  r    r  r   r  c                       \ rS rSr\rSrg)AutoBackbonei  r   N)r   r   r   r   MODEL_FOR_BACKBONE_MAPPINGr4  r   r   r   r   r  r    rO  r   r  zaudio retrieval via x-vectorc                       \ rS rSr\rSrg)AutoModelForMaskedImageModelingi  r   N)r   r   r   r   'MODEL_FOR_MASKED_IMAGE_MODELING_MAPPINGr4  r   r   r   r   r  r    r  r   r  zmasked image modelingc                       \ rS rSr\rSrg)AutoModelForAudioTokenizationi  r   N)r   r   r   r   $MODEL_FOR_AUDIO_TOKENIZATION_MAPPINGr4  r   r   r   r   r  r    r9  r   r  z$audio tokenization through codebooksc                   H   ^  \ rS rSr\U 4S j5       r\U 4S j5       rSrU =r$ )AutoModelWithLMHeadi  c                 \   > [         R                  " S[        5        [        TU ]  " U40 UD6$ 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FutureWarningrX  from_configrR  configr\  r]  s      r   r  AutoModelWithLMHead.from_config  s1    B 		
 w"64V44r   c                 b   > [         R                  " S[        5        [        TU ]  " U/UQ70 UD6$ r  r  r  r  rX  rY  rZ  s       r   rY  #AutoModelWithLMHead.from_pretrained  s8    B 		
 w&'D\z\U[\\r   r   	r   r   r   r   ra  r  rY  r   rf  rg  s   @r   r  r    s+    5 5 ] ]r   r  c                   H   ^  \ rS rSr\U 4S j5       r\U 4S j5       rSrU =r$ )AutoModelForVision2Seqi  c                 \   > [         R                  " S[        5        [        TU ]  " U40 UD6$ NzThe class `AutoModelForVision2Seq` is deprecated and will be removed in v5.0. Please use `AutoModelForImageTextToText` instead.r  r  s      r   r  "AutoModelForVision2Seq.from_config  s.    5	

 w"64V44r   c                 b   > [         R                  " S[        5        [        TU ]  " U/UQ70 UD6$ r  r  rZ  s       r   rY  &AutoModelForVision2Seq.from_pretrained  s5    5	

 w&'D\z\U[\\r   r   r  rg  s   @r   r  r    s+    5 5 ] ]r   r  )Yr  r  r  r  r  'MODEL_FOR_CAUSAL_IMAGE_MODELING_MAPPINGr`  r  r  r  r  MODEL_FOR_IMAGE_MAPPINGr  rC  r8  r<  r  r  rj  r3  r  r  r  rI  ru  r  rm  rr  r  rx  r?  r  r  r  r  r  r  r  MODEL_FOR_RETRIEVAL_MAPPINGr  r|  rF  rN  r  r  ,MODEL_FOR_TIME_SERIES_CLASSIFICATION_MAPPING(MODEL_FOR_TIME_SERIES_REGRESSION_MAPPINGrE  r  r  r  r  r  rQ  r  r  r  r  rB  r  r7  r;  r2  r>  r  ri  r  r  r  rH  rt  r  rl  rq  r  rw  r  r  r  r  r  r  r  r{  r~  r  r  r  r  )__doc__rd  r  collectionsr   typingr   r   utilsr   auto_factoryr	   r
   r   r   configuration_autor   
generationr   modeling_utilsr   r   
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$MODEL_FOR_VISION_2_SEQ_MAPPING_NAMES!MODEL_FOR_RETRIEVAL_MAPPING_NAMES*MODEL_FOR_IMAGE_TEXT_TO_TEXT_MAPPING_NAMES!MODEL_FOR_MASKED_LM_MAPPING_NAMES(MODEL_FOR_OBJECT_DETECTION_MAPPING_NAMES2MODEL_FOR_ZERO_SHOT_OBJECT_DETECTION_MAPPING_NAMES(MODEL_FOR_DEPTH_ESTIMATION_MAPPING_NAMES,MODEL_FOR_SEQ_TO_SEQ_CAUSAL_LM_MAPPING_NAMES(MODEL_FOR_SPEECH_SEQ_2_SEQ_MAPPING_NAMES/MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING_NAMES*MODEL_FOR_QUESTION_ANSWERING_MAPPING_NAMES0MODEL_FOR_TABLE_QUESTION_ANSWERING_MAPPING_NAMES1MODEL_FOR_VISUAL_QUESTION_ANSWERING_MAPPING_NAMES3MODEL_FOR_DOCUMENT_QUESTION_ANSWERING_MAPPING_NAMES,MODEL_FOR_TOKEN_CLASSIFICATION_MAPPING_NAMES'MODEL_FOR_MULTIPLE_CHOICE_MAPPING_NAMES0MODEL_FOR_NEXT_SENTENCE_PREDICTION_MAPPING_NAMES,MODEL_FOR_AUDIO_CLASSIFICATION_MAPPING_NAMESMODEL_FOR_CTC_MAPPING_NAMES2MODEL_FOR_AUDIO_FRAME_CLASSIFICATION_MAPPING_NAMES%MODEL_FOR_AUDIO_XVECTOR_MAPPING_NAMES+MODEL_FOR_TEXT_TO_SPECTROGRAM_MAPPING_NAMES(MODEL_FOR_TEXT_TO_WAVEFORM_MAPPING_NAMES6MODEL_FOR_ZERO_SHOT_IMAGE_CLASSIFICATION_MAPPING_NAMES MODEL_FOR_BACKBONE_MAPPING_NAMES'MODEL_FOR_MASK_GENERATION_MAPPING_NAMES*MODEL_FOR_KEYPOINT_DETECTION_MAPPING_NAMES)MODEL_FOR_KEYPOINT_MATCHING_MAPPING_NAMES%MODEL_FOR_TEXT_ENCODING_MAPPING_NAMES2MODEL_FOR_TIME_SERIES_CLASSIFICATION_MAPPING_NAMES.MODEL_FOR_TIME_SERIES_REGRESSION_MAPPING_NAMES.MODEL_FOR_TIME_SERIES_PREDICTION_MAPPING_NAMES&MODEL_FOR_IMAGE_TO_IMAGE_MAPPING_NAMES"MODEL_FOR_AUDIO_TOKENIZATION_NAMESrF  rI  rN  r`  r  r  r  r  r  r  r  r  r  r  r  r|  r  rj  r  r  r  r  r  rm  rr  ru  rx  r  r  r  r  r  r  r  r  r  r  r  r3  r8  r<  r?  r  r  r  rC  r  r2  r7  r;  r>  rB  rE  rH  rM  rQ  ri  rl  rq  rt  rw  r{  r~  r  r  r  r  r  r  r  r  r  r  r  r  r  r  r  r  r  r  r  r  r  r  r  r  r  r  r  r  __all__r   r   r   <module>r*     s    	  # '   5 -1/  
		H	%!JL \ '2Z\' #| $/VX$  t %0MO% !b !,=?! B 1<1 - 1< 	71 - 0;BD0 ,L .9 	(. * 1<
1 - 1< 	<1 - 2=	2 . 0;0 , (3( $> %0*% ! .913. *j %002% !h ,7, (  6A	6 2 ,7
, ( 0;!0 ,F ,7, (  3>uw3 /r .9QS. *j 4? 	/4 0 5@5 1 7B7 3 0;[]0 ,~ +6)++ 'Z 4?
4 0 0;0 ,$ * $ 6A
6 2 )4
) % /: 	?// + ,7, (" :E: 6  $/$  6 +6	+ ' .98. * -8- ) )4 ") %H 6AC16 2 2=5-2 . 2=02 . *55* & &1& " !!57JK 01EGj k -.BDde ./CEfg *:G+ ' *:F* & 4DP4 0 (8D( $ +;G+ ' +;G+ ' ,<H, ( *:F* & "22FHl!m '7D( $ //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\ 0. 0 ))=H[\ ]. ] ))=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$  
8"5 8 00KVhi B*= B '8'2N' #
8"5 8 00KVhi <&9 < #44S^t"u 41 4 ,,CNgh ]"5 ] 00KVst <&9 < #44S^t"u +) + $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! 
]. ],]4 ](Zr   