
     `i                    ,   d Z ddlZddlZddlmZ ddlmZmZ ddlm	Z	 ddl
mZmZmZmZ dd	lmZ erdd
lmZ ddlmZ  G d dee          Z e	j        e          Z eg d          Z eg d          Z eg d          Z eg d          Z eg d          Z e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%          Z0 ed&g          Z1 eg d'          Z2 eg d(          Z3 eg d)          Z4 eg d*          Z5 eg d+          Z6 eg d,          Z7 eg d-          Z8 eg d.          Z9 eg d/          Z: ed0d1g          Z; eg d2          Z< eg d3          Z= eg d4          Z> eg d5          Z? ed6g          Z@ eg d7          ZA eg d8          ZB ed9d:g          ZC ed;d<g          ZD ed=g          ZE ed>g          ZF ed?g          ZG eee          ZH eee          ZI eee          ZJ eee          ZK eee          ZL eee           ZM eee=          ZN eee!          ZO eee"          ZP eee#          ZQ eee$          ZR eee%          ZS eee&          ZT eee(          ZU eee'          ZV eee2          ZW eee3          ZX eee)          ZY eee          ZZ eee          Z[ eee*          Z\ eee+          Z] eee,          Z^ eee-          Z_ eee/          Z` eee0          Za eee1          Zb eee4          Zc eee5          Zd eee6          Ze eee7          Zf eee8          Zg eee.          Zh eee9          Zi eee:          Zj eee;          Zk eee<          Zl eee>          Zm eee?          Zn eee@          Zo eeeA          Zp eeeB          Zq eeeC          Zr eeeD          Zs eeeE          Zt eeeF          Zu eeeG          Zv G d@ dAe          Zw G dB dCe          Zx G dD dEe          Zy G dF dGe          Zz G dH dIe          Z{ G dJ dKe          Z| ee|          Z| G dL dMe          Z} ee}dNO          Z} G dP dQe          Z~ ee~dRO          Z~ G dS dTe          Z eedUO          Z G dV dWe          Z eedXO          Z G dY dZe          Z eed[d\]          Z G d^ d_e          Z eed`O          Z G da dbe          Z eedcO          Z G dd dee          Z eedfdg]          Z G dh die          Z eedjdk]          Z G dl dme          Z eedndo]          Z G dp dqe          Z eedrO          Z G ds dte          Z eeduO          Z G dv dwe          Z eedxO          Z G dy dze          Z eed{O          Z G d| d}e          Z eed~O          Z G d de          Z eedO          Z G d de          Z eedO          Z G d de          Z eedO          Z G d de          Z eedO          Z G d de          Z eedO          Z G d de          Z eedO          Z G d de          Z eedO          Z G d de          Z eedO          Z G d de          Z eedO          Z G d de          Z eedO          Z G d de          Z eedO          Z G d de          Z eedO          Z G d de          Z eedO          Z G d de          Z eedO          Z G d de          Z eedO          Z G d de          Z G d de          Z G d de          Z G d de          Z eedO          Z G d de          Z eedO          Z G d de          Z eedO          Z G d de~          Z G d de          Zg dZdS )zAuto Model class.    N)OrderedDict)TYPE_CHECKINGUnion   )logging   )_BaseAutoBackboneClass_BaseAutoModelClass_LazyAutoMappingauto_class_update)CONFIG_MAPPING_NAMES)GenerationMixin)PreTrainedModelc                       e Zd ZdS )_BaseModelWithGenerateN)__name__
__module____qualname__     z/home/jaya/work/projects/VOICE-AGENT/VIET/agent-env/lib/python3.11/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)gpt2rX  )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_wrapperrA  )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r9   BartForConditionalGeneration)r?   BertForPreTraining)rD   BigBirdForPreTrainingrZ   BloomForCausalLMrb   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$  !Florence2ForConditionalGeneration)r&  FNetForPreTrainingr+  FSMTForConditionalGeneration)r-  FunnelForPreTrainingr6  Gemma3ForConditionalGenerationrW  GPT2LMHeadModelrY  r  rZ  GPTBigCodeForCausalLMrf  )r|  HieraForPreTrainingr  IBertForMaskedLMr  IdeficsForVisionText2Textr   Idefics2ForConditionalGenerationr   Idefics3ForConditionalGenerationr  JanusForConditionalGenerationr  LayoutLMForMaskedLMr  LlavaForConditionalGenerationr  !LlavaNextForConditionalGenerationr  &LlavaNextVideoForConditionalGenerationr  &LlavaOnevisionForConditionalGenerationr  LongformerForMaskedLMr  LukeForMaskedLM)r  LxmertForPreTrainingr  MambaForCausalLMr  Mamba2ForCausalLMr  MegaForMaskedLM)r  MegatronBertForPreTrainingr   Mistral3ForConditionalGenerationr  MllamaForConditionalGeneration)r
  MobileBertForPreTrainingr   MPNetForMaskedLMr"  MptForCausalLMr$  MraForMaskedLMr,  MvpForConditionalGeneration)r3  NezhaForPreTrainingr5  NllbMoeForConditionalGenerationrH  OpenAIGPTLMHeadModelrR  !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)r1  UniSpeechForPreTraining)r3  UniSpeechSatForPreTrainingr<  "VideoLlavaForConditionalGeneration)r?  VideoMAEForPreTrainingrC   VipLlavaForConditionalGeneration)rF  VisualBertForPreTraining)rO  ViTMAEForPreTrainingr_  )rd  Wav2Vec2ForPreTraining)rh  Wav2Vec2ConformerForPreTrainingru  XLMWithLMHeadModelrz  XLMRobertaForMaskedLMr}  XLMRobertaXLForMaskedLMr  XLNetLMHeadModelr  xLSTMForCausalLMr  XmodForMaskedLM)Tr   AlbertForMaskedLMr  r?   BertForMaskedLMrD   BigBirdForMaskedLMrF   &BigBirdPegasusForConditionalGenerationrQ   '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  r-  FunnelForMaskedLMrB  GitForCausalLMr  r  r  r\  GPTNeoForCausalLMr^  GPTNeoXForCausalLMr`  GPTNeoXJapaneseForCausalLMrd  GPTJForCausalLMrf  r  r  r  LEDForConditionalGenerationr  r  LongT5ForConditionalGenerationr  r  M2M100ForConditionalGenerationr  r  r  MarianMTModelr  r  MegatronBertForCausalLMr
  MobileBertForMaskedLMr  !MoonshineForConditionalGenerationr  r  r  r  r3  NezhaForMaskedLMr  r8  NystromformerForMaskedLMr  r_   PegasusXForConditionalGenerationrs  PLBartForConditionalGeneration	pop2piano!Pop2PianoForConditionalGenerationr  QDQBertForMaskedLMr  ReformerModelWithLMHeadr  RemBertForMaskedLMr  r  r  RoCBertForMaskedLMr  RoFormerForMaskedLMr  r  #Speech2TextForConditionalGenerationr  r   r  r  r  r  rd  Wav2Vec2ForMaskedLMrl  WhisperForConditionalGenerationr  r  r  r  r   r  YosoForMaskedLM))r&   ApertusForCausalLM)r(   ArceeForCausalLM)r,   AriaTextForCausalLM)r4   BambaForCausalLM)r9   BartForCausalLM)r?   BertLMHeadModel)rA   BertGenerationDecoder)rD   BigBirdForCausalLM)rF   BigBirdPegasusForCausalLM)rH   BioGptForCausalLM)rM   BitNetForCausalLM)rO   BlenderbotForCausalLM)rQ   BlenderbotSmallForCausalLMr  )r\   BltForCausalLM)rb   CamembertForCausalLM)r|   LlamaForCausalLMr,  )r   CohereForCausalLM)r   Cohere2ForCausalLMr0  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r0  FuyuForCausalLM)r2  GemmaForCausalLM)r4  Gemma2ForCausalLMr  )r8  Gemma3ForCausalLMr:  Gemma3nForConditionalGeneration)r>  Gemma3nForCausalLMr@  )rD  GlmForCausalLM)rF  Glm4ForCausalLM)rH  Glm4MoeForCausalLMrU  GotOcr2ForConditionalGenerationr  r  r  rB  rD  rF  )rb  GptOssForCausalLMrH  )rh  GraniteForCausalLM)rj  GraniteMoeForCausalLM)rl  GraniteMoeHybridForCausalLM)rn  GraniteMoeSharedForCausalLM)rv  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  MegaForCausalLMrR  )r  MiniMaxForCausalLM)r  MinistralForCausalLM)r  MistralForCausalLM)r   MixtralForCausalLM)r  MllamaForCausalLM)r  ModernBertDecoderForCausalLM)r  MoshiForCausalLMr  )r(  MusicgenForCausalLM)r*  MusicgenMelodyForCausalLM)r,  MvpForCausalLM)r1  NemotronForCausalLM)r:  OlmoForCausalLM)r<  Olmo2ForCausalLM)r>  Olmo3ForCausalLM)r@  OlmoeForCausalLM)rF  OpenLlamaForCausalLMr  )rJ  OPTForCausalLM)r]  PegasusForCausalLM)rg  PersimmonForCausalLM)ri  PhiForCausalLM)rk  Phi3ForCausalLM)rm  Phi4MultimodalForCausalLM)ro  PhimoeForCausalLM)rs  PLBartForCausalLM)rx  ProphetNetForCausalLM)r  QDQBertLMHeadModel)r  Qwen2ForCausalLM)r  Qwen2MoeForCausalLM)r  Qwen3ForCausalLM)r  Qwen3MoeForCausalLM)r  Qwen3NextForCausalLM)r  RecurrentGemmaForCausalLMre  )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)rl  WhisperForCausalLM)rr  XGLMForCausalLMr  )rw  XLMProphetNetForCausalLM)rz  XLMRobertaForCausalLM)r}  XLMRobertaXLForCausalLMr  r  )r  XmodForCausalLM)r  ZambaForCausalLM)r  Zamba2ForCausalLM);r   r;   rJ   r   r   r   r   r   r   r   r   r   r   r   r   r   r   r   r   r   r   r   r(  rR  r{  r  r  r  )r  Llama4VisionModelr  )r  MllamaVisionModelr  r  r  r  r.  ru  rz  r  r  r  r  r   r  r  r	  r  r  r   r#  r7  r>  rH  rK  rN  rQ  rT  rZ  r  ))r   DeiTForMaskedImageModeling)r)  FocalNetForMaskedImageModeling)r  SwinForMaskedImageModeling)r
  Swinv2ForMaskedImageModeling)rI  ViTForMaskedImageModeling)r  ImageGPTForCausalImageModeling)-)r<   BeitForImageClassification)rK   BitForImageClassification)rp   CLIPForImageClassification)r   ConvNextForImageClassification)r    ConvNextV2ForImageClassification)r   CvtForImageClassification)r   $Data2VecVisionForImageClassification)r   )DeiTForImageClassification%DeiTForImageClassificationWithTeacher)r   DinatForImageClassification)r   Dinov2ForImageClassification)r   )Dinov2WithRegistersForImageClassification)r   DonutSwinForImageClassification)r   )%EfficientFormerForImageClassification0EfficientFormerForImageClassificationWithTeacher)r   "EfficientNetForImageClassification)r)  FocalNetForImageClassification)ry  HGNetV2ForImageClassification)r|  HieraForImageClassification)r  IJepaForImageClassification)r  ImageGPTForImageClassification)r  )LevitForImageClassification&LevitForImageClassificationWithTeacher)r  MetaClip2ForImageClassification)r  !MobileNetV1ForImageClassification)r  !MobileNetV2ForImageClassification)r  MobileViTForImageClassification)r  !MobileViTV2ForImageClassification)r/  NatForImageClassification)ra  )&PerceiverForImageClassificationLearned&PerceiverForImageClassificationFourier-PerceiverForImageClassificationConvProcessing)rv   PoolFormerForImageClassification)r{  PvtForImageClassification)r}  PvtV2ForImageClassification)r  RegNetForImageClassification)r  ResNetForImageClassification)r  SegformerForImageClassification)shieldgemma2"ShieldGemma2ForImageClassification)r  SiglipForImageClassification)r  Siglip2ForImageClassification)r  !SwiftFormerForImageClassification)r  SwinForImageClassification)r
  Swinv2ForImageClassification)r  TextNetForImageClassification)r$  !TimmWrapperForImageClassification)r8  VanForImageClassification)rI  ViTForImageClassification)rL  ViTHybridForImageClassification)rR  ViTMSNForImageClassificationr   DetrForSegmentation))r<   BeitForSemanticSegmentation)r   %Data2VecVisionForSemanticSegmentation)r   DPTForSemanticSegmentation)r  "MobileNetV2ForSemanticSegmentation)r   MobileViTForSemanticSegmentation)r  "MobileViTV2ForSemanticSegmentation)r   SegformerForSemanticSegmentation)upernetUperNetForSemanticSegmentationr  !MaskFormerForInstanceSegmentation)r$  )eomtEomtForUniversalSegmentation)r  #Mask2FormerForUniversalSegmentationr/  )rD  !OneFormerForUniversalSegmentation))r  !TimesformerForVideoClassification)r?  VideoMAEForVideoClassification)r[  VivitForVideoClassification)r]  VJEPA2ForVideoClassification)rT   BlipForConditionalGenerationrV   Blip2ForConditionalGenerationrf   !ChameleonForConditionalGenerationr@  r  r  r  $InstructBlipForConditionalGeneration)r  )InstructBlipVideoForConditionalGenerationr  Kosmos2ForConditionalGenerationr  !Kosmos2_5ForConditionalGenerationr  r  r  r  r  r  rL  Ovis2ForConditionalGenerationr  
pix2struct"Pix2StructForConditionalGenerationr  "Qwen2_5_VLForConditionalGenerationr  Qwen2VLForConditionalGenerationr  Qwen3VLForConditionalGenerationr  "Qwen3VLMoeForConditionalGenerationr  r  zvision-encoder-decoderVisionEncoderDecoderModelr  )0)r*   AriaForConditionalGeneration)r2   !AyaVisionForConditionalGenerationr9  r;  r=  )r   %Cohere2VisionForConditionalGeneration)r   "DeepseekVLForConditionalGeneration)r   (DeepseekVLHybridForConditionalGeneration)r   Emu3ForConditionalGenerationr  r  r  r  r  r@  )rJ  Glm4vForConditionalGeneration)rL   Glm4vMoeForConditionalGenerationr  r  r  r  r?  )r   InternVLForConditionalGenerationr  rB  rD  )r  Lfm2VlForConditionalGenerationr  r  r  r  r  r  r  rF  r  )re  $PerceptionLMForConditionalGenerationrH  )rq  r  rK  rM  rO  rQ  )r  r  )r  SmolVLMForConditionalGeneration)r-  UdopForConditionalGenerationr  rS  ).r"  r  r$  r&  r  r.  r  r  r  r  r4  r8  r:  r  r<  r>  r  r  r  r  r  MBartForConditionalGenerationr  )r  MegatronBertForMaskedLMrT  )r  ModernBertForMaskedLMr  r  r  rX  rZ  )ra  PerceiverForMaskedLMrc  )r  ReformerForMaskedLMrg  r  r  ri  rk  r  r  ro  r  r  r  r   rs  )
)r   !ConditionalDetrForObjectDetection)r   DFineForObjectDetection)r   DabDetrForObjectDetection)r    DeformableDetrForObjectDetection)r   DetaForObjectDetection)r   DetrForObjectDetection)r  RTDetrForObjectDetection)r  RTDetrV2ForObjectDetection)r  "TableTransformerForObjectDetection)r  YolosForObjectDetection))rr  GroundingDinoForObjectDetection)r  !MMGroundingDinoForObjectDetectionrB  )rN  Owlv2ForObjectDetection)rP  OwlViTForObjectDetection))depth_anythingDepthAnythingForDepthEstimation)r   DepthProForDepthEstimation)r   DPTForDepthEstimation)rS  GLPNForDepthEstimation)prompt_depth_anything%PromptDepthAnythingForDepthEstimation)zoedepthZoeDepthForDepthEstimation)r  r(  )rO   "BlenderbotForConditionalGenerationr*  r6  r  rf  granite_speech%GraniteSpeechForConditionalGenerationrJ  rL  rN  rP  rb  )r&  MT5ForConditionalGenerationr  r  )r]  PegasusForConditionalGenerationr\  r^  )rx  "ProphetNetForConditionalGenerationr  )r  SeamlessM4TForTextToText)r  SeamlessM4Tv2ForTextToTextr   r  r  )r/  UMT5ForConditionalGenerationr_  )rw  %XLMProphetNetForConditionalGeneration)r2  r  )r  *KyutaiSpeechToTextForConditionalGenerationrV  r`  )r  SeamlessM4TForSpeechToText)r  SeamlessM4Tv2ForSpeechToText)zspeech-encoder-decoderSpeechEncoderDecoderModelrm  )r  SpeechT5ForSpeechToTextrq  )s)r   AlbertForSequenceClassification)r(   ArceeForSequenceClassification)r9   BartForSequenceClassification)r?   BertForSequenceClassification)rD    BigBirdForSequenceClassification)rF   'BigBirdPegasusForSequenceClassification)rH   BioGptForSequenceClassification)rZ   BloomForSequenceClassification)rb   "CamembertForSequenceClassification)rd   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)r-  FunnelForSequenceClassification)r2  GemmaForSequenceClassification)r4  Gemma2ForSequenceClassification)r6  Gemma3ForSequenceClassification)r8  #Gemma3TextForSequenceClassification)rD  GlmForSequenceClassification)rF  Glm4ForSequenceClassification)rW  GPT2ForSequenceClassification)rY  r  )rZ  #GPTBigCodeForSequenceClassification)r\  GPTNeoForSequenceClassification)r^   GPTNeoXForSequenceClassification)rb  GptOssForSequenceClassification)rd  GPTJForSequenceClassification)rv  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$  MraForSequenceClassification)r&  MT5ForSequenceClassification)r,  MvpForSequenceClassification)r1  !NemotronForSequenceClassification)r3  NezhaForSequenceClassification)r8  &NystromformerForSequenceClassification)rF  "OpenLlamaForSequenceClassification)rH  "OpenAIGPTForSequenceClassification)rJ  OPTForSequenceClassification)ra  "PerceiverForSequenceClassification)rg  "PersimmonForSequenceClassification)ri  PhiForSequenceClassification)rk  Phi3ForSequenceClassification)ro  PhimoeForSequenceClassification)rs  PLBartForSequenceClassification)r   QDQBertForSequenceClassification)r  Qwen2ForSequenceClassification)r  !Qwen2MoeForSequenceClassification)r  Qwen3ForSequenceClassification)r  !Qwen3MoeForSequenceClassification)r  "Qwen3NextForSequenceClassification)r  !ReformerForSequenceClassification)r   RemBertForSequenceClassification)r   RobertaForSequenceClassification)r  ,RobertaPreLayerNormForSequenceClassification)r   RoCBertForSequenceClassification)r  !RoFormerForSequenceClassification)r   SeedOssForSequenceClassification)r   SmolLM3ForSequenceClassification)r  $SqueezeBertForSequenceClassification)r  !StableLmForSequenceClassification)r  #Starcoder2ForSequenceClassification)r  T5ForSequenceClassification)r   T5GemmaForSequenceClassification)r  TapasForSequenceClassification)r'  "TransfoXLForSequenceClassification)r/  UMT5ForSequenceClassification)ru  XLMForSequenceClassification)rz  #XLMRobertaForSequenceClassification)r}  %XLMRobertaXLForSequenceClassification)r  XLNetForSequenceClassification)r  XmodForSequenceClassification)r  YosoForSequenceClassification)r  ZambaForSequenceClassification)r  Zamba2ForSequenceClassification)O)r   AlbertForQuestionAnswering)r(   ArceeForQuestionAnswering)r9   BartForQuestionAnswering)r?   BertForQuestionAnswering)rD   BigBirdForQuestionAnswering)rF   "BigBirdPegasusForQuestionAnswering)rZ   BloomForQuestionAnswering)rb   CamembertForQuestionAnswering)rd   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)r-  FunnelForQuestionAnswering)rY  GPT2ForQuestionAnswering)r\  GPTNeoForQuestionAnswering)r^  GPTNeoXForQuestionAnswering)rd  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)r1  NemotronForQuestionAnswering)r3  NezhaForQuestionAnswering)r8  !NystromformerForQuestionAnswering)rJ  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)r/  UMT5ForQuestionAnswering)ru  XLMForQuestionAnsweringSimple)rz  XLMRobertaForQuestionAnswering)r}   XLMRobertaXLForQuestionAnswering)r  XLNetForQuestionAnsweringSimple)r  XmodForQuestionAnswering)r  YosoForQuestionAnswering)r  TapasForQuestionAnswering))rT   BlipForQuestionAnsweringr;  )rA  ViltForQuestionAnswering))r  LayoutLMForQuestionAnsweringr  r  )Y)r   AlbertForTokenClassification)r&   ApertusForTokenClassification)r(   ArceeForTokenClassification)r?   BertForTokenClassification)rD   BigBirdForTokenClassification)rH   BioGptForTokenClassification)rZ   BloomForTokenClassification)r`   BrosForTokenClassification)rb   CamembertForTokenClassification)rd   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)r-  FunnelForTokenClassification)r2  GemmaForTokenClassification)r4  Gemma2ForTokenClassification)rD  GlmForTokenClassification)rF  Glm4ForTokenClassification)rW  GPT2ForTokenClassification)rY  rs  )rZ   GPTBigCodeForTokenClassification)r\  GPTNeoForTokenClassification)r^  GPTNeoXForTokenClassification)rb  GptOssForTokenClassification)rv  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)r1  NemotronForTokenClassification)r3  NezhaForTokenClassification)r8  #NystromformerForTokenClassification)rg  PersimmonForTokenClassification)ri  PhiForTokenClassification)rk  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)r/  UMT5ForTokenClassification)ru  XLMForTokenClassification)rz   XLMRobertaForTokenClassification)r}  "XLMRobertaXLForTokenClassification)r  XLNetForTokenClassification)r  XmodForTokenClassification)r  YosoForTokenClassification)')r   AlbertForMultipleChoice)r?   BertForMultipleChoice)rD   BigBirdForMultipleChoice)rb   CamembertForMultipleChoice)rd   CanineForMultipleChoice)r   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  ModernBertForMultipleChoice)r   MPNetForMultipleChoice)r$  MraForMultipleChoice)r3  NezhaForMultipleChoice)r8  NystromformerForMultipleChoice)r  QDQBertForMultipleChoice)r  RemBertForMultipleChoice)r  RobertaForMultipleChoice)r  $RobertaPreLayerNormForMultipleChoice)r  RoCBertForMultipleChoice)r  RoFormerForMultipleChoice)r  SqueezeBertForMultipleChoice)ru  XLMForMultipleChoice)rz  XLMRobertaForMultipleChoice)r}  XLMRobertaXLForMultipleChoice)r  XLNetForMultipleChoice)r  XmodForMultipleChoice)r  YosoForMultipleChoice))r?   BertForNextSentencePrediction)r  ErnieForNextSentencePrediction)r&  FNetForNextSentencePrediction)r  %MegatronBertForNextSentencePrediction)r
  #MobileBertForNextSentencePrediction)r3  NezhaForNextSentencePrediction)r   QDQBertForNextSentencePrediction))r.   ASTForAudioClassification)r   &Data2VecAudioForSequenceClassification)r~  HubertForSequenceClassification)r  SEWForSequenceClassification)r  SEWDForSequenceClassification)r1  "UniSpeechForSequenceClassification)r3  %UniSpeechSatForSequenceClassification)rd  !Wav2Vec2ForSequenceClassification)rf  %Wav2Vec2BertForSequenceClassification)rh  *Wav2Vec2ConformerForSequenceClassification)rj  WavLMForSequenceClassification)rl  WhisperForAudioClassification))r   Data2VecAudioForCTC)r~  HubertForCTC)r  MCTCTForCTCrT  )r  	SEWForCTC)r  
SEWDForCTC)r1  UniSpeechForCTC)r3  UniSpeechSatForCTC)rd  Wav2Vec2ForCTC)rf  Wav2Vec2BertForCTC)rh  Wav2Vec2ConformerForCTC)rj  WavLMForCTC))r   (Data2VecAudioForAudioFrameClassification)r3  'UniSpeechSatForAudioFrameClassification)rd  #Wav2Vec2ForAudioFrameClassification)rf  'Wav2Vec2BertForAudioFrameClassification)rh  ,Wav2Vec2ConformerForAudioFrameClassification)rj   WavLMForAudioFrameClassification))r   Data2VecAudioForXVector)r3  UniSpeechSatForXVector)rd  Wav2Vec2ForXVector)rf  Wav2Vec2BertForXVector)rh  Wav2Vec2ConformerForXVector)rj  WavLMForXVectorr  )r  SpeechT5ForTextToSpeech)r6   r   )r  r  r  )r(   MusicgenForConditionalGeneration)r*  &MusicgenMelodyForConditionalGeneration)qwen2_5_omni#Qwen2_5OmniForConditionalGeneration)qwen3_omni_moe$Qwen3OmniMoeForConditionalGeneration)r  SeamlessM4TForTextToSpeech)r  SeamlessM4Tv2ForTextToSpeechrW  )
r    r#   rS   )rV   Blip2ForImageTextRetrievalrh   ro   rw   r  r  r  ))r<   BeitBackbone)rK   BitBackbone)r   ConvNextBackbone)r   ConvNextV2Backbone)r   DinatBackbone)r   Dinov2Backbone)r   Dinov2WithRegistersBackbone)r)  FocalNetBackbonerx  )r|  HieraBackbone)r  MaskFormerSwinBackbone)r/  NatBackbone)r}  PvtV2Backbone)r  ResNetBackbone)rt_detr_resnetRTDetrResNetBackbone)r  SwinBackbone)r
  Swinv2Backbone)r  TextNetBackboner   )rU  VitDetBackbone)vitpose_backboneVitPoseBackbone)r   )r   r   r  r  )r  r  r  )
superpointSuperPointForKeypointDetection))r   !EfficientLoFTRForKeypointMatchingr  )	superglueSuperGlueForKeypointMatching)r   r>   rC   rr   r   r   r   r   r   )r   Emu3TextModelr  r  )r  r  r  )r  MllamaTextModelr	  )r&  MT5EncoderModelr7  r  r  r  r  r  r  r  )r  T5EncoderModel)r  T5GemmaEncoderModel)r/  UMT5EncoderModelrt  ry  r|  )rY  'PatchTSMixerForTimeSeriesClassification)r[  PatchTSTForClassification)rY  PatchTSMixerForRegression)r[  PatchTSTForRegression)r  TimesFmModelForPrediction)r  Swin2SRForImageSuperResolutionr   c                       e Zd ZeZdS )AutoModelForMaskGenerationN)r   r   r   !MODEL_FOR_MASK_GENERATION_MAPPING_model_mappingr   r   r   r/  r/            6NNNr   r/  c                       e Zd ZeZdS )AutoModelForKeypointDetectionN)r   r   r   $MODEL_FOR_KEYPOINT_DETECTION_MAPPINGr1  r   r   r   r4  r4            9NNNr   r4  c                       e Zd ZeZdS )AutoModelForKeypointMatchingN)r   r   r   #MODEL_FOR_KEYPOINT_MATCHING_MAPPINGr1  r   r   r   r8  r8    s        8NNNr   r8  c                       e Zd ZeZdS )AutoModelForTextEncodingN)r   r   r   MODEL_FOR_TEXT_ENCODING_MAPPINGr1  r   r   r   r;  r;            4NNNr   r;  c                       e Zd ZeZdS )AutoModelForImageToImageN)r   r   r    MODEL_FOR_IMAGE_TO_IMAGE_MAPPINGr1  r   r   r   r?  r?    s        5NNNr   r?  c                       e Zd ZeZdS )	AutoModelN)r   r   r   MODEL_MAPPINGr1  r   r   r   rB  rB    s        "NNNr   rB  c                       e Zd ZeZdS )AutoModelForPreTrainingN)r   r   r   MODEL_FOR_PRETRAINING_MAPPINGr1  r   r   r   rE  rE    s        2NNNr   rE  pretraining)head_docc                       e Zd ZeZdS )_AutoModelWithLMHeadN)r   r   r   MODEL_WITH_LM_HEAD_MAPPINGr1  r   r   r   rJ  rJ            /NNNr   rJ  zlanguage modelingc                   r     e Zd ZeZeded          deee	j
        e         f         ddf fd            Z xZS )AutoModelForCausalLMclspretrained_model_name_or_pathreturnr   c                 >     t                      j        |g|R i |S Nsuperfrom_pretrainedrO  rP  
model_argskwargs	__class__s       r   rV  z$AutoModelForCausalLM.from_pretrained  .     'uww&'D\z\\\U[\\\r   )r   r   r   MODEL_FOR_CAUSAL_LM_MAPPINGr1  classmethodtyper   strosPathLikerV  __classcell__rZ  s   @r   rN  rN    s        0N ]()]',S"+c2B-B'C]
 
"] ] ] ] ] [] ] ] ] ]r   rN  zcausal language modelingc                       e Zd ZeZdS )AutoModelForMaskedLMN)r   r   r   MODEL_FOR_MASKED_LM_MAPPINGr1  r   r   r   re  re    s        0NNNr   re  zmasked language modelingc                       e Zd ZeZdS )AutoModelForSeq2SeqLMN)r   r   r   &MODEL_FOR_SEQ_TO_SEQ_CAUSAL_LM_MAPPINGr1  r   r   r   rh  rh            ;NNNr   rh  z&sequence-to-sequence language modelingzgoogle-t5/t5-base)rH  checkpoint_for_examplec                       e Zd ZeZdS )"AutoModelForSequenceClassificationN)r   r   r   )MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPINGr1  r   r   r   rm  rm    s        >NNNr   rm  zsequence classificationc                       e Zd ZeZdS )AutoModelForQuestionAnsweringN)r   r   r   $MODEL_FOR_QUESTION_ANSWERING_MAPPINGr1  r   r   r   rp  rp    r6  r   rp  zquestion answeringc                       e Zd ZeZdS )"AutoModelForTableQuestionAnsweringN)r   r   r   *MODEL_FOR_TABLE_QUESTION_ANSWERING_MAPPINGr1  r   r   r   rs  rs            ?NNNr   rs  ztable question answeringzgoogle/tapas-base-finetuned-wtqc                       e Zd ZeZdS )#AutoModelForVisualQuestionAnsweringN)r   r   r   +MODEL_FOR_VISUAL_QUESTION_ANSWERING_MAPPINGr1  r   r   r   rw  rw    s        @NNNr   rw  zvisual question answeringzdandelin/vilt-b32-finetuned-vqac                       e Zd ZeZdS )%AutoModelForDocumentQuestionAnsweringN)r   r   r   -MODEL_FOR_DOCUMENT_QUESTION_ANSWERING_MAPPINGr1  r   r   r   rz  rz    s        BNNNr   rz  zdocument question answeringz/impira/layoutlm-document-qa", revision="52e01b3c                       e Zd ZeZdS )AutoModelForTokenClassificationN)r   r   r   &MODEL_FOR_TOKEN_CLASSIFICATION_MAPPINGr1  r   r   r   r}  r}    rj  r   r}  ztoken classificationc                       e Zd ZeZdS )AutoModelForMultipleChoiceN)r   r   r   !MODEL_FOR_MULTIPLE_CHOICE_MAPPINGr1  r   r   r   r  r    r2  r   r  zmultiple choicec                       e Zd ZeZdS )"AutoModelForNextSentencePredictionN)r   r   r   *MODEL_FOR_NEXT_SENTENCE_PREDICTION_MAPPINGr1  r   r   r   r  r    ru  r   r  znext sentence predictionc                       e Zd ZeZdS )AutoModelForImageClassificationN)r   r   r   &MODEL_FOR_IMAGE_CLASSIFICATION_MAPPINGr1  r   r   r   r  r    rj  r   r  zimage classificationc                       e Zd ZeZdS )'AutoModelForZeroShotImageClassificationN)r   r   r   0MODEL_FOR_ZERO_SHOT_IMAGE_CLASSIFICATION_MAPPINGr1  r   r   r   r  r    s        ENNNr   r  zzero-shot image classificationc                       e Zd ZeZdS )AutoModelForImageSegmentationN)r   r   r   $MODEL_FOR_IMAGE_SEGMENTATION_MAPPINGr1  r   r   r   r  r  %  r6  r   r  zimage segmentationc                       e Zd ZeZdS ) AutoModelForSemanticSegmentationN)r   r   r   'MODEL_FOR_SEMANTIC_SEGMENTATION_MAPPINGr1  r   r   r   r  r  ,          <NNNr   r  zsemantic segmentationc                       e Zd ZeZdS ) AutoModelForTimeSeriesPredictionN)r   r   r   (MODEL_FOR_TIME_SERIES_PREDICTION_MAPPINGr1  r   r   r   r  r  5          =NNNr   r  ztime-series predictionc                       e Zd ZeZdS )!AutoModelForUniversalSegmentationN)r   r   r   (MODEL_FOR_UNIVERSAL_SEGMENTATION_MAPPINGr1  r   r   r   r  r  >  r  r   r  zuniversal image segmentationc                       e Zd ZeZdS ) AutoModelForInstanceSegmentationN)r   r   r   'MODEL_FOR_INSTANCE_SEGMENTATION_MAPPINGr1  r   r   r   r  r  G  r  r   r  zinstance segmentationc                       e Zd ZeZdS )AutoModelForObjectDetectionN)r   r   r   "MODEL_FOR_OBJECT_DETECTION_MAPPINGr1  r   r   r   r  r  P          7NNNr   r  zobject detectionc                       e Zd ZeZdS )#AutoModelForZeroShotObjectDetectionN)r   r   r   ,MODEL_FOR_ZERO_SHOT_OBJECT_DETECTION_MAPPINGr1  r   r   r   r  r  W          ANNNr   r  zzero-shot object detectionc                       e Zd ZeZdS )AutoModelForDepthEstimationN)r   r   r   "MODEL_FOR_DEPTH_ESTIMATION_MAPPINGr1  r   r   r   r  r  `  r  r   r  zdepth estimationc                       e Zd ZeZdS )AutoModelForVideoClassificationN)r   r   r   &MODEL_FOR_VIDEO_CLASSIFICATION_MAPPINGr1  r   r   r   r  r  g  rj  r   r  zvideo classificationc                       e Zd ZeZdS )_AutoModelForVision2SeqN)r   r   r   MODEL_FOR_VISION_2_SEQ_MAPPINGr1  r   r   r   r  r  o  s        3NNNr   r  zvision-to-text modelingc                   r     e Zd ZeZeded          deee	j
        e         f         ddf fd            Z xZS )AutoModelForImageTextToTextrO  rP  rQ  r   c                 >     t                      j        |g|R i |S rS  rT  rW  s       r   rV  z+AutoModelForImageTextToText.from_pretrainedz  r[  r   )r   r   r   $MODEL_FOR_IMAGE_TEXT_TO_TEXT_MAPPINGr1  r]  r^  r   r_  r`  ra  rV  rb  rc  s   @r   r  r  v  s        9N ]/0]',S"+c2B-B'C]
 
"] ] ] ] ] [] ] ] ] ]r   r  zimage-text-to-text modelingc                       e Zd ZeZdS )AutoModelForAudioClassificationN)r   r   r   &MODEL_FOR_AUDIO_CLASSIFICATION_MAPPINGr1  r   r   r   r  r    rj  r   r  zaudio classificationc                       e Zd ZeZdS )AutoModelForCTCN)r   r   r   MODEL_FOR_CTC_MAPPINGr1  r   r   r   r  r    s        *NNNr   r  z%connectionist temporal classificationc                       e Zd ZeZdS )AutoModelForSpeechSeq2SeqN)r   r   r   "MODEL_FOR_SPEECH_SEQ_2_SEQ_MAPPINGr1  r   r   r   r  r    r  r   r  z,sequence-to-sequence speech-to-text modelingc                       e Zd ZeZdS )$AutoModelForAudioFrameClassificationN)r   r   r   ,MODEL_FOR_AUDIO_FRAME_CLASSIFICATION_MAPPINGr1  r   r   r   r  r    r  r   r  z"audio frame (token) classificationc                       e Zd ZeZdS )AutoModelForAudioXVectorN)r   r   r   MODEL_FOR_AUDIO_XVECTOR_MAPPINGr1  r   r   r   r  r    r=  r   r  c                       e Zd ZeZdS )AutoModelForTextToSpectrogramN)r   r   r   %MODEL_FOR_TEXT_TO_SPECTROGRAM_MAPPINGr1  r   r   r   r  r    s        :NNNr   r  c                       e Zd ZeZdS )AutoModelForTextToWaveformN)r   r   r   "MODEL_FOR_TEXT_TO_WAVEFORM_MAPPINGr1  r   r   r   r  r    r  r   r  c                       e Zd ZeZdS )AutoBackboneN)r   r   r   MODEL_FOR_BACKBONE_MAPPINGr1  r   r   r   r  r    rL  r   r  zaudio retrieval via x-vectorc                       e Zd ZeZdS )AutoModelForMaskedImageModelingN)r   r   r   'MODEL_FOR_MASKED_IMAGE_MODELING_MAPPINGr1  r   r   r   r  r    r  r   r  zmasked image modelingc                       e Zd ZeZdS )AutoModelForAudioTokenizationN)r   r   r   $MODEL_FOR_AUDIO_TOKENIZATION_MAPPINGr1  r   r   r   r  r    r6  r   r  z$audio tokenization through codebooksc                   H     e Zd Ze fd            Ze fd            Z xZS )AutoModelWithLMHeadc                 l    t          j        dt                      t                      j        |fi |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FutureWarningrU  from_configrO  configrY  rZ  s      r   r  zAutoModelWithLMHead.from_config  sC    B 		
 	
 	
 #uww"644V444r   c                 r    t          j        dt                      t                      j        |g|R i |S r  r  r  r  rU  rV  rW  s       r   rV  z#AutoModelWithLMHead.from_pretrained  sM    B 		
 	
 	
 'uww&'D\z\\\U[\\\r   r   r   r   r]  r  rV  rb  rc  s   @r   r  r    sp        5 5 5 5 [5 ] ] ] ] [] ] ] ] ]r   r  c                   H     e Zd Ze fd            Ze fd            Z xZS )AutoModelForVision2Seqc                 l    t          j        dt                      t                      j        |fi |S NzThe class `AutoModelForVision2Seq` is deprecated and will be removed in v5.0. Please use `AutoModelForImageTextToText` instead.r  r  s      r   r  z"AutoModelForVision2Seq.from_config  s@    5	
 	
 	

 #uww"644V444r   c                 r    t          j        dt                      t                      j        |g|R i |S r  r  rW  s       r   rV  z&AutoModelForVision2Seq.from_pretrained  sJ    5	
 	
 	

 'uww&'D\z\\\U[\\\r   r  rc  s   @r   r  r    sp        5 5 5 5 [5 ] ] ] ] [] ] ] ] ]r   r  )Yr  r  r  r  r  'MODEL_FOR_CAUSAL_IMAGE_MODELING_MAPPINGr\  r  r{  r  r  MODEL_FOR_IMAGE_MAPPINGr  r@  r5  r9  r  r  rf  r0  r  r  r  rF  rq  r  ri  rn  r  rt  r<  r  r  r  r~  r  r  r  MODEL_FOR_RETRIEVAL_MAPPINGr  rx  rC  rK  r  r  ,MODEL_FOR_TIME_SERIES_CLASSIFICATION_MAPPING(MODEL_FOR_TIME_SERIES_REGRESSION_MAPPINGrB  r  r  r  r  r  rN  r  r  r  r  r?  r  r4  r8  r/  r;  r  re  r  r  r  rE  rp  r  rh  rm  r  rs  r  r  r  r}  r  r  r  rw  rz  r  r  r  r  )__doc__r`  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_NAMESrC  rF  rK  r\  r  r  r  r  r  r  r  r  r  r  r  rx  r{  rf  r  r  r  r  r  ri  rn  rq  rt  r~  r  r  r  r  r  r  r  r  r  r  r0  r5  r9  r<  r  r  r  r@  r  r/  r4  r8  r;  r?  rB  rE  rJ  rN  re  rh  rm  rp  rs  rw  rz  r}  r  r  r  r  r  r  r  r  r  r  r  r  r  r  r  r  r  r  r  r  r  r  r  r  r  r  r  __all__r   r   r   <module>r!     s     				  # # # # # # ' ' ' ' ' ' ' '                  5 4 4 4 4 4  ------111111    /    
	H	%	%!kJ J JL L \ '2kZ Z Z\' \' #| $/;V V VX$ X$  t %0KM M MO% O% !b !,= = =?! ?! B 1<  1 1 - 1< 	71 1 - 0;{B B BD0 D0 ,L .9[ 	(. . * 1<
 
 
1 1 - 1< 	<1 1 - 2=  	2 	2 . 0;{  0 0 , (3{  ( ( $> %0K*% % ! .9[1 1 13. 3. *j %0K0 0 02% 2% !h ,7;  , , (  6A[  	6 	6 2 ,7;  
, 
, ( 0;{  !0 !0 ,F ,7;  , , (  3>+u u uw3 w3 /r .9[Q Q QS. S. *j 4?; 	/4 4 0 5@K  5 5 1 7Bk  7 7 3 0;{[ [ []0 ]0 ,~ +6+) ) )++ ++ 'Z 4?;  
4 
4 0 0;{  0 0 ,$ *k    $ 6A[  
6 
6 2 )4  
) 
) % /:k 	?// / + ,7;  , , (" :E  : : 6  $/;  $ $  6 +6+  	+ 	+ ' .9[8. . * -8K  - - ) )4     ") ") %H 6A[C16 6 2 2=5-2 2 . 2=02 2 . *55* * & &1[& & " ! !57JKK 0 01EGj k k --.BDdee ../CEfgg *:*:G+ + ' *:)9F* * & 4D3CP4 4 0 (8'7D( ( $ +;*:G+ + ' +;*:G+ + ' ,<+;H, , ( *:)9F* * & "2!12FHl!m!m '7'7D( ( $ /./CEfgg .>.>K/ / + 1A0@M1 1 - /./CEfgg **+?A^__ *:*:G+ + ' &6%56JLt%u%u "/?/?L0 0 , &6%56JLt%u%u ")9)9F* * & -=,<I- - ) (8'7D( ( $ .>-=J. . * *:)9F* * & %5$45IKr$s$s !-=-=J. . * *:)9F* * & )()=?Z[[ %5%56JLt%u%u "/?/?L0 0 , #3"23GIn"o"o (8(8E) ) % &6%56JLt%u%u "--.BDdee $4$45IKr$s$s !'7'7D( ( $ '7&67KMv&w&w #"2"23GIn"o"o /?/?L0 0 , ,<+;H, , ( ,<+;H, , ( $4#34HJp#q#q  '7'78LNp'q'q $7 7 7 7 7!4 7 7 7: : : : :$7 : : :9 9 9 9 9#6 9 9 95 5 5 5 52 5 5 56 6 6 6 62 6 6 6# # # # ## # # # i((	3 3 3 3 31 3 3 3 ,+,Cm\\\ 0 0 0 0 0. 0 0 0 )()=H[\\\ ] ] ] ] ]. ] ] ] )()=Hbccc 1 1 1 1 1. 1 1 1 )()=Hbccc < < < < </ < < < *)5.   ? ? ? ? ?)< ? ? ? &7%6&1J& & & "
: : : : :$7 : : : !2 12OZn o o o @ @ @ @ @)< @ @ @ &7%6&'<& & & "A A A A A*= A A A '8&7'(<' ' ' #C C C C C,? C C C ):(9)*L) ) ) %< < < < <&9 < < < #4"34S^t"u"u"u 7 7 7 7 7!4 7 7 7 /./ITefff @ @ @ @ @)< @ @ @ &7%6&1K& & & "
< < < < <&9 < < < #4"34S^t"u"u"u F F F F F.A F F F +<*;+6V+ + + '
: : : : :$7 : : : !2 12OZn o o o = = = = =': = = = $5#4$/F$ $ $  
> > > > >': > > > $5#4$/G$ $ $  
> > > > >(; > > > %6$5%0N% % % !
= = = = =': = = = $5#4$/F$ $ $  
8 8 8 8 8"5 8 8 8 0/0KVhiii B B B B B*= B B B '8&7'2N' ' ' #
8 8 8 8 8"5 8 8 8 0/0KVhiii < < < < <&9 < < < #4"34S^t"u"u"u 4 4 4 4 41 4 4 4 ,+,CNghhh ] ] ] ] ]"5 ] ] ] 0/0KVsttt < < < < <&9 < < < #4"34S^t"u"u"u + + + + +) + + + $#O>efff8 8 8 8 8 3 8 8 8 .-(V   
B B B B B+> B B B (9'8(3W( ( ( $
5 5 5 5 52 5 5 5; ; ; ; ;$7 ; ; ;8 8 8 8 8!4 8 8 80 0 0 0 0) 0 0 0 -,-EPnooo = = = = =&9 = = = #4"34S^u"v"v"v : : : : :$7 : : : !2 1!,R! ! ! 
] ] ] ] ]. ] ] ],] ] ] ] ]4 ] ] ](Z Z Zr   