transformers 5.0.0__py3-none-any.whl → 5.0.0rc0__py3-none-any.whl

transformers/models/auto/modeling_auto.py

@@ -1,3 +1,4 @@
+# coding=utf-8
 # Copyright 2018 The HuggingFace Inc. team.
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
@@ -14,8 +15,9 @@
 """Auto Model class."""
 
 import os
+import warnings
 from collections import OrderedDict
-from typing import TYPE_CHECKING
+from typing import TYPE_CHECKING, Union
 
 from ...utils import logging
 from .auto_factory import (
@@ -143,14 +145,12 @@ MODEL_MAPPING_NAMES = OrderedDict(
         ("ernie", "ErnieModel"),
         ("ernie4_5", "Ernie4_5Model"),
         ("ernie4_5_moe", "Ernie4_5_MoeModel"),
-        ("ernie4_5_vl_moe", "Ernie4_5_VL_MoeModel"),
         ("esm", "EsmModel"),
         ("evolla", "EvollaModel"),
         ("exaone4", "Exaone4Model"),
         ("falcon", "FalconModel"),
         ("falcon_h1", "FalconH1Model"),
         ("falcon_mamba", "FalconMambaModel"),
-        ("fast_vlm", "FastVlmModel"),
         ("fastspeech2_conformer", "FastSpeech2ConformerModel"),
         ("fastspeech2_conformer_with_hifigan", "FastSpeech2ConformerWithHifiGan"),
         ("flaubert", "FlaubertModel"),
@@ -175,19 +175,12 @@ MODEL_MAPPING_NAMES = OrderedDict(
         ("glm4", "Glm4Model"),
         ("glm46v", "Glm46VModel"),
         ("glm4_moe", "Glm4MoeModel"),
-        ("glm4_moe_lite", "Glm4MoeLiteModel"),
         ("glm4v", "Glm4vModel"),
         ("glm4v_moe", "Glm4vMoeModel"),
         ("glm4v_moe_text", "Glm4vMoeTextModel"),
         ("glm4v_moe_vision", "Glm4vMoeVisionModel"),
         ("glm4v_text", "Glm4vTextModel"),
         ("glm4v_vision", "Glm4vVisionModel"),
-        ("glm_image", "GlmImageModel"),
-        ("glm_image_text", "GlmImageTextModel"),
-        ("glm_image_vision", "GlmImageVisionModel"),
-        ("glm_image_vqmodel", "GlmImageVQVAE"),
-        ("glmasr", "GlmAsrForConditionalGeneration"),
-        ("glmasr_encoder", "GlmAsrEncoder"),
         ("glpn", "GLPNModel"),
         ("got_ocr2", "GotOcr2Model"),
         ("gpt-sw3", "GPT2Model"),
@@ -222,15 +215,12 @@ MODEL_MAPPING_NAMES = OrderedDict(
         ("instructblipvideo", "InstructBlipVideoModel"),
         ("internvl", "InternVLModel"),
         ("internvl_vision", "InternVLVisionModel"),
-        ("jais2", "Jais2Model"),
         ("jamba", "JambaModel"),
         ("janus", "JanusModel"),
         ("jetmoe", "JetMoeModel"),
         ("kosmos-2", "Kosmos2Model"),
         ("kosmos-2.5", "Kosmos2_5Model"),
         ("kyutai_speech_to_text", "KyutaiSpeechToTextModel"),
-        ("lasr_ctc", "LasrForCTC"),
-        ("lasr_encoder", "LasrEncoder"),
         ("layoutlm", "LayoutLMModel"),
         ("layoutlmv2", "LayoutLMv2Model"),
         ("layoutlmv3", "LayoutLMv3Model"),
@@ -240,7 +230,6 @@ MODEL_MAPPING_NAMES = OrderedDict(
         ("lfm2_moe", "Lfm2MoeModel"),
         ("lfm2_vl", "Lfm2VlModel"),
         ("lightglue", "LightGlueForKeypointMatching"),
-        ("lighton_ocr", "LightOnOcrModel"),
         ("lilt", "LiltModel"),
         ("llama", "LlamaModel"),
         ("llama4", "Llama4ForConditionalGeneration"),
@@ -253,7 +242,6 @@ MODEL_MAPPING_NAMES = OrderedDict(
         ("longformer", "LongformerModel"),
         ("longt5", "LongT5Model"),
         ("luke", "LukeModel"),
-        ("lw_detr", "LwDetrModel"),
         ("lxmert", "LxmertModel"),
         ("m2m_100", "M2M100Model"),
         ("mamba", "MambaModel"),
@@ -269,7 +257,6 @@ MODEL_MAPPING_NAMES = OrderedDict(
         ("mgp-str", "MgpstrForSceneTextRecognition"),
         ("mimi", "MimiModel"),
         ("minimax", "MiniMaxModel"),
-        ("minimax_m2", "MiniMaxM2Model"),
         ("ministral", "MinistralModel"),
         ("ministral3", "Ministral3Model"),
         ("mistral", "MistralModel"),
@@ -314,12 +301,6 @@ MODEL_MAPPING_NAMES = OrderedDict(
         ("parakeet_encoder", "ParakeetEncoder"),
         ("patchtsmixer", "PatchTSMixerModel"),
         ("patchtst", "PatchTSTModel"),
-        ("pe_audio", "PeAudioModel"),
-        ("pe_audio_encoder", "PeAudioEncoder"),
-        ("pe_audio_video", "PeAudioVideoModel"),
-        ("pe_audio_video_encoder", "PeAudioVideoEncoder"),
-        ("pe_video", "PeVideoModel"),
-        ("pe_video_encoder", "PeVideoEncoder"),
         ("pegasus", "PegasusModel"),
         ("pegasus_x", "PegasusXModel"),
         ("perceiver", "PerceiverModel"),
@@ -329,7 +310,6 @@ MODEL_MAPPING_NAMES = OrderedDict(
         ("phi3", "Phi3Model"),
         ("phi4_multimodal", "Phi4MultimodalModel"),
         ("phimoe", "PhimoeModel"),
-        ("pixio", "PixioModel"),
         ("pixtral", "PixtralVisionModel"),
         ("plbart", "PLBartModel"),
         ("poolformer", "PoolFormerModel"),
@@ -391,7 +371,6 @@ MODEL_MAPPING_NAMES = OrderedDict(
         ("smollm3", "SmolLM3Model"),
         ("smolvlm", "SmolVLMModel"),
         ("smolvlm_vision", "SmolVLMVisionTransformer"),
-        ("solar_open", "SolarOpenModel"),
         ("speech_to_text", "Speech2TextModel"),
         ("speecht5", "SpeechT5Model"),
         ("splinter", "SplinterModel"),
@@ -488,7 +467,6 @@ MODEL_FOR_PRETRAINING_MAPPING_NAMES = OrderedDict(
         ("fsmt", "FSMTForConditionalGeneration"),
         ("funnel", "FunnelForPreTraining"),
         ("gemma3", "Gemma3ForConditionalGeneration"),
-        ("glmasr", "GlmAsrForConditionalGeneration"),
         ("gpt-sw3", "GPT2LMHeadModel"),
         ("gpt2", "GPT2LMHeadModel"),
         ("gpt_bigcode", "GPTBigCodeForCausalLM"),
@@ -551,6 +529,92 @@ MODEL_FOR_PRETRAINING_MAPPING_NAMES = OrderedDict(
     ]
 )
 
+MODEL_WITH_LM_HEAD_MAPPING_NAMES = OrderedDict(
+    [
+        # Model with LM heads mapping
+        ("albert", "AlbertForMaskedLM"),
+        ("bart", "BartForConditionalGeneration"),
+        ("bert", "BertForMaskedLM"),
+        ("big_bird", "BigBirdForMaskedLM"),
+        ("bigbird_pegasus", "BigBirdPegasusForConditionalGeneration"),
+        ("blenderbot-small", "BlenderbotSmallForConditionalGeneration"),
+        ("bloom", "BloomForCausalLM"),
+        ("camembert", "CamembertForMaskedLM"),
+        ("codegen", "CodeGenForCausalLM"),
+        ("convbert", "ConvBertForMaskedLM"),
+        ("cpmant", "CpmAntForCausalLM"),
+        ("ctrl", "CTRLLMHeadModel"),
+        ("data2vec-text", "Data2VecTextForMaskedLM"),
+        ("deberta", "DebertaForMaskedLM"),
+        ("deberta-v2", "DebertaV2ForMaskedLM"),
+        ("dia", "DiaForConditionalGeneration"),
+        ("distilbert", "DistilBertForMaskedLM"),
+        ("electra", "ElectraForMaskedLM"),
+        ("encoder-decoder", "EncoderDecoderModel"),
+        ("ernie", "ErnieForMaskedLM"),
+        ("esm", "EsmForMaskedLM"),
+        ("exaone4", "Exaone4ForCausalLM"),
+        ("falcon_mamba", "FalconMambaForCausalLM"),
+        ("flaubert", "FlaubertWithLMHeadModel"),
+        ("fnet", "FNetForMaskedLM"),
+        ("fsmt", "FSMTForConditionalGeneration"),
+        ("funnel", "FunnelForMaskedLM"),
+        ("git", "GitForCausalLM"),
+        ("gpt-sw3", "GPT2LMHeadModel"),
+        ("gpt2", "GPT2LMHeadModel"),
+        ("gpt_bigcode", "GPTBigCodeForCausalLM"),
+        ("gpt_neo", "GPTNeoForCausalLM"),
+        ("gpt_neox", "GPTNeoXForCausalLM"),
+        ("gpt_neox_japanese", "GPTNeoXJapaneseForCausalLM"),
+        ("gptj", "GPTJForCausalLM"),
+        ("ibert", "IBertForMaskedLM"),
+        ("layoutlm", "LayoutLMForMaskedLM"),
+        ("led", "LEDForConditionalGeneration"),
+        ("longformer", "LongformerForMaskedLM"),
+        ("longt5", "LongT5ForConditionalGeneration"),
+        ("luke", "LukeForMaskedLM"),
+        ("m2m_100", "M2M100ForConditionalGeneration"),
+        ("mamba", "MambaForCausalLM"),
+        ("mamba2", "Mamba2ForCausalLM"),
+        ("marian", "MarianMTModel"),
+        ("megatron-bert", "MegatronBertForCausalLM"),
+        ("mobilebert", "MobileBertForMaskedLM"),
+        ("moonshine", "MoonshineForConditionalGeneration"),
+        ("mpnet", "MPNetForMaskedLM"),
+        ("mpt", "MptForCausalLM"),
+        ("mra", "MraForMaskedLM"),
+        ("mvp", "MvpForConditionalGeneration"),
+        ("nllb-moe", "NllbMoeForConditionalGeneration"),
+        ("nystromformer", "NystromformerForMaskedLM"),
+        ("openai-gpt", "OpenAIGPTLMHeadModel"),
+        ("pegasus_x", "PegasusXForConditionalGeneration"),
+        ("plbart", "PLBartForConditionalGeneration"),
+        ("pop2piano", "Pop2PianoForConditionalGeneration"),
+        ("reformer", "ReformerModelWithLMHead"),
+        ("rembert", "RemBertForMaskedLM"),
+        ("roberta", "RobertaForMaskedLM"),
+        ("roberta-prelayernorm", "RobertaPreLayerNormForMaskedLM"),
+        ("roc_bert", "RoCBertForMaskedLM"),
+        ("roformer", "RoFormerForMaskedLM"),
+        ("rwkv", "RwkvForCausalLM"),
+        ("speech_to_text", "Speech2TextForConditionalGeneration"),
+        ("squeezebert", "SqueezeBertForMaskedLM"),
+        ("switch_transformers", "SwitchTransformersForConditionalGeneration"),
+        ("t5", "T5ForConditionalGeneration"),
+        ("t5gemma", "T5GemmaForConditionalGeneration"),
+        ("t5gemma2", "T5Gemma2ForConditionalGeneration"),
+        ("tapas", "TapasForMaskedLM"),
+        ("wav2vec2", "Wav2Vec2ForMaskedLM"),
+        ("whisper", "WhisperForConditionalGeneration"),
+        ("xlm", "XLMWithLMHeadModel"),
+        ("xlm-roberta", "XLMRobertaForMaskedLM"),
+        ("xlm-roberta-xl", "XLMRobertaXLForMaskedLM"),
+        ("xlnet", "XLNetLMHeadModel"),
+        ("xmod", "XmodForMaskedLM"),
+        ("yoso", "YosoForMaskedLM"),
+    ]
+)
+
 MODEL_FOR_CAUSAL_LM_MAPPING_NAMES = OrderedDict(
     [
         # Model for Causal LM mapping
@@ -606,7 +670,6 @@ MODEL_FOR_CAUSAL_LM_MAPPING_NAMES = OrderedDict(
         ("glm", "GlmForCausalLM"),
         ("glm4", "Glm4ForCausalLM"),
         ("glm4_moe", "Glm4MoeForCausalLM"),
-        ("glm4_moe_lite", "Glm4MoeLiteForCausalLM"),
         ("got_ocr2", "GotOcr2ForConditionalGeneration"),
         ("gpt-sw3", "GPT2LMHeadModel"),
         ("gpt2", "GPT2LMHeadModel"),
@@ -623,7 +686,6 @@ MODEL_FOR_CAUSAL_LM_MAPPING_NAMES = OrderedDict(
         ("helium", "HeliumForCausalLM"),
         ("hunyuan_v1_dense", "HunYuanDenseV1ForCausalLM"),
         ("hunyuan_v1_moe", "HunYuanMoEV1ForCausalLM"),
-        ("jais2", "Jais2ForCausalLM"),
         ("jamba", "JambaForCausalLM"),
         ("jetmoe", "JetMoeForCausalLM"),
         ("lfm2", "Lfm2ForCausalLM"),
@@ -638,7 +700,6 @@ MODEL_FOR_CAUSAL_LM_MAPPING_NAMES = OrderedDict(
         ("mbart", "MBartForCausalLM"),
         ("megatron-bert", "MegatronBertForCausalLM"),
         ("minimax", "MiniMaxForCausalLM"),
-        ("minimax_m2", "MiniMaxM2ForCausalLM"),
         ("ministral", "MinistralForCausalLM"),
         ("ministral3", "Ministral3ForCausalLM"),
         ("mistral", "MistralForCausalLM"),
@@ -681,7 +742,6 @@ MODEL_FOR_CAUSAL_LM_MAPPING_NAMES = OrderedDict(
         ("rwkv", "RwkvForCausalLM"),
         ("seed_oss", "SeedOssForCausalLM"),
         ("smollm3", "SmolLM3ForCausalLM"),
-        ("solar_open", "SolarOpenForCausalLM"),
         ("stablelm", "StableLmForCausalLM"),
         ("starcoder2", "Starcoder2ForCausalLM"),
         ("trocr", "TrOCRForCausalLM"),
@@ -735,7 +795,6 @@ MODEL_FOR_IMAGE_MAPPING_NAMES = OrderedDict(
         ("mobilenet_v2", "MobileNetV2Model"),
         ("mobilevit", "MobileViTModel"),
         ("mobilevitv2", "MobileViTV2Model"),
-        ("pixio", "PixioModel"),
         ("poolformer", "PoolFormerModel"),
         ("pvt", "PvtModel"),
         ("regnet", "RegNetModel"),
@@ -888,6 +947,37 @@ MODEL_FOR_VIDEO_CLASSIFICATION_MAPPING_NAMES = OrderedDict(
     ]
 )
 
+MODEL_FOR_VISION_2_SEQ_MAPPING_NAMES = OrderedDict(
+    [
+        ("blip", "BlipForConditionalGeneration"),
+        ("blip-2", "Blip2ForConditionalGeneration"),
+        ("chameleon", "ChameleonForConditionalGeneration"),
+        ("git", "GitForCausalLM"),
+        ("idefics2", "Idefics2ForConditionalGeneration"),
+        ("idefics3", "Idefics3ForConditionalGeneration"),
+        ("instructblip", "InstructBlipForConditionalGeneration"),
+        ("instructblipvideo", "InstructBlipVideoForConditionalGeneration"),
+        ("kosmos-2", "Kosmos2ForConditionalGeneration"),
+        ("kosmos-2.5", "Kosmos2_5ForConditionalGeneration"),
+        ("llava", "LlavaForConditionalGeneration"),
+        ("llava_next", "LlavaNextForConditionalGeneration"),
+        ("llava_next_video", "LlavaNextVideoForConditionalGeneration"),
+        ("llava_onevision", "LlavaOnevisionForConditionalGeneration"),
+        ("mistral3", "Mistral3ForConditionalGeneration"),
+        ("mllama", "MllamaForConditionalGeneration"),
+        ("ovis2", "Ovis2ForConditionalGeneration"),
+        ("paligemma", "PaliGemmaForConditionalGeneration"),
+        ("pix2struct", "Pix2StructForConditionalGeneration"),
+        ("qwen2_5_vl", "Qwen2_5_VLForConditionalGeneration"),
+        ("qwen2_vl", "Qwen2VLForConditionalGeneration"),
+        ("qwen3_vl", "Qwen3VLForConditionalGeneration"),
+        ("qwen3_vl_moe", "Qwen3VLMoeForConditionalGeneration"),
+        ("video_llava", "VideoLlavaForConditionalGeneration"),
+        ("vipllava", "VipLlavaForConditionalGeneration"),
+        ("vision-encoder-decoder", "VisionEncoderDecoderModel"),
+    ]
+)
+
 MODEL_FOR_RETRIEVAL_MAPPING_NAMES = OrderedDict(
     [
         ("colpali", "ColPaliForRetrieval"),
@@ -905,9 +995,7 @@ MODEL_FOR_IMAGE_TEXT_TO_TEXT_MAPPING_NAMES = OrderedDict(
         ("deepseek_vl", "DeepseekVLForConditionalGeneration"),
         ("deepseek_vl_hybrid", "DeepseekVLHybridForConditionalGeneration"),
         ("emu3", "Emu3ForConditionalGeneration"),
-        ("ernie4_5_vl_moe", "Ernie4_5_VL_MoeForConditionalGeneration"),
         ("evolla", "EvollaForProteinText2Text"),
-        ("fast_vlm", "FastVlmForConditionalGeneration"),
         ("florence2", "Florence2ForConditionalGeneration"),
         ("fuyu", "FuyuForCausalLM"),
         ("gemma3", "Gemma3ForConditionalGeneration"),
@@ -921,13 +1009,11 @@ MODEL_FOR_IMAGE_TEXT_TO_TEXT_MAPPING_NAMES = OrderedDict(
         ("idefics2", "Idefics2ForConditionalGeneration"),
         ("idefics3", "Idefics3ForConditionalGeneration"),
         ("instructblip", "InstructBlipForConditionalGeneration"),
-        ("instructblipvideo", "InstructBlipVideoForConditionalGeneration"),
         ("internvl", "InternVLForConditionalGeneration"),
         ("janus", "JanusForConditionalGeneration"),
         ("kosmos-2", "Kosmos2ForConditionalGeneration"),
         ("kosmos-2.5", "Kosmos2_5ForConditionalGeneration"),
         ("lfm2_vl", "Lfm2VlForConditionalGeneration"),
-        ("lighton_ocr", "LightOnOcrForConditionalGeneration"),
         ("llama4", "Llama4ForConditionalGeneration"),
         ("llava", "LlavaForConditionalGeneration"),
         ("llava_next", "LlavaNextForConditionalGeneration"),
@@ -936,7 +1022,6 @@ MODEL_FOR_IMAGE_TEXT_TO_TEXT_MAPPING_NAMES = OrderedDict(
         ("mistral3", "Mistral3ForConditionalGeneration"),
         ("mllama", "MllamaForConditionalGeneration"),
         ("ovis2", "Ovis2ForConditionalGeneration"),
-        ("paddleocr_vl", "PaddleOCRVLForConditionalGeneration"),
         ("paligemma", "PaliGemmaForConditionalGeneration"),
         ("perception_lm", "PerceptionLMForConditionalGeneration"),
         ("pix2struct", "Pix2StructForConditionalGeneration"),
@@ -950,7 +1035,6 @@ MODEL_FOR_IMAGE_TEXT_TO_TEXT_MAPPING_NAMES = OrderedDict(
         ("t5gemma2", "T5Gemma2ForConditionalGeneration"),
         ("udop", "UdopForConditionalGeneration"),
         ("video_llama_3", "VideoLlama3ForConditionalGeneration"),
-        ("video_llava", "VideoLlavaForConditionalGeneration"),
         ("vipllava", "VipLlavaForConditionalGeneration"),
         ("vision-encoder-decoder", "VisionEncoderDecoderModel"),
     ]
@@ -961,7 +1045,6 @@ MODEL_FOR_IMAGE_TEXT_TO_TEXT_MAPPING_NAMES = OrderedDict(
 MODEL_FOR_MULTIMODAL_LM_MAPPING_NAMES = OrderedDict(
     [
         *list(MODEL_FOR_IMAGE_TEXT_TO_TEXT_MAPPING_NAMES.items()),
-        ("glmasr", "GlmAsrForConditionalGeneration"),
         ("granite_speech", "GraniteSpeechForConditionalGeneration"),
         ("kyutai_speech_to_text", "KyutaiSpeechToTextForConditionalGeneration"),
         ("phi4_multimodal", "Phi4MultimodalForCausalLM"),
@@ -972,7 +1055,6 @@ MODEL_FOR_MULTIMODAL_LM_MAPPING_NAMES = OrderedDict(
     ]
 )
 
-
 MODEL_FOR_MASKED_LM_MAPPING_NAMES = OrderedDict(
     [
         # Model for Masked LM mapping
@@ -1013,6 +1095,7 @@ MODEL_FOR_MASKED_LM_MAPPING_NAMES = OrderedDict(
         ("roformer", "RoFormerForMaskedLM"),
         ("squeezebert", "SqueezeBertForMaskedLM"),
         ("tapas", "TapasForMaskedLM"),
+        ("wav2vec2", "Wav2Vec2ForMaskedLM"),
         ("xlm", "XLMWithLMHeadModel"),
         ("xlm-roberta", "XLMRobertaForMaskedLM"),
         ("xlm-roberta-xl", "XLMRobertaXLForMaskedLM"),
@@ -1029,7 +1112,6 @@ MODEL_FOR_OBJECT_DETECTION_MAPPING_NAMES = OrderedDict(
         ("dab-detr", "DabDetrForObjectDetection"),
         ("deformable_detr", "DeformableDetrForObjectDetection"),
         ("detr", "DetrForObjectDetection"),
-        ("lw_detr", "LwDetrForObjectDetection"),
         ("rt_detr", "RTDetrForObjectDetection"),
         ("rt_detr_v2", "RTDetrV2ForObjectDetection"),
         ("table-transformer", "TableTransformerForObjectDetection"),
@@ -1069,7 +1151,6 @@ MODEL_FOR_SEQ_TO_SEQ_CAUSAL_LM_MAPPING_NAMES = OrderedDict(
         ("blenderbot-small", "BlenderbotSmallForConditionalGeneration"),
         ("encoder-decoder", "EncoderDecoderModel"),
         ("fsmt", "FSMTForConditionalGeneration"),
-        ("glmasr", "GlmAsrForConditionalGeneration"),
         ("granite_speech", "GraniteSpeechForConditionalGeneration"),
         ("led", "LEDForConditionalGeneration"),
         ("longt5", "LongT5ForConditionalGeneration"),
@@ -1095,7 +1176,6 @@ MODEL_FOR_SEQ_TO_SEQ_CAUSAL_LM_MAPPING_NAMES = OrderedDict(
     ]
 )
 
-
 MODEL_FOR_SPEECH_SEQ_2_SEQ_MAPPING_NAMES = OrderedDict(
     [
         ("dia", "DiaForConditionalGeneration"),
@@ -1166,6 +1246,7 @@ MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING_NAMES = OrderedDict(
         ("layoutlm", "LayoutLMForSequenceClassification"),
         ("layoutlmv2", "LayoutLMv2ForSequenceClassification"),
         ("layoutlmv3", "LayoutLMv3ForSequenceClassification"),
+        ("led", "LEDForSequenceClassification"),
         ("lilt", "LiltForSequenceClassification"),
         ("llama", "LlamaForSequenceClassification"),
         ("longformer", "LongformerForSequenceClassification"),
@@ -1500,7 +1581,6 @@ MODEL_FOR_CTC_MAPPING_NAMES = OrderedDict(
         # Model for Connectionist temporal classification (CTC) mapping
         ("data2vec-audio", "Data2VecAudioForCTC"),
         ("hubert", "HubertForCTC"),
-        ("lasr_ctc", "LasrForCTC"),
         ("parakeet_ctc", "ParakeetForCTC"),
         ("sew", "SEWForCTC"),
         ("sew-d", "SEWDForCTC"),
@@ -1593,9 +1673,7 @@ MODEL_FOR_BACKBONE_MAPPING_NAMES = OrderedDict(
         ("focalnet", "FocalNetBackbone"),
         ("hgnet_v2", "HGNetV2Backbone"),
         ("hiera", "HieraBackbone"),
-        ("lw_detr_vit", "LwDetrViTBackbone"),
         ("maskformer-swin", "MaskFormerSwinBackbone"),
-        ("pixio", "PixioBackbone"),
         ("pvt_v2", "PvtV2Backbone"),
         ("resnet", "ResNetBackbone"),
         ("rt_detr_resnet", "RTDetrResNetBackbone"),
@@ -1706,6 +1784,7 @@ MODEL_FOR_AUDIO_TOKENIZATION_NAMES = OrderedDict(
 
 MODEL_MAPPING = _LazyAutoMapping(CONFIG_MAPPING_NAMES, MODEL_MAPPING_NAMES)
 MODEL_FOR_PRETRAINING_MAPPING = _LazyAutoMapping(CONFIG_MAPPING_NAMES, MODEL_FOR_PRETRAINING_MAPPING_NAMES)
+MODEL_WITH_LM_HEAD_MAPPING = _LazyAutoMapping(CONFIG_MAPPING_NAMES, MODEL_WITH_LM_HEAD_MAPPING_NAMES)
 MODEL_FOR_CAUSAL_LM_MAPPING = _LazyAutoMapping(CONFIG_MAPPING_NAMES, MODEL_FOR_CAUSAL_LM_MAPPING_NAMES)
 MODEL_FOR_CAUSAL_IMAGE_MODELING_MAPPING = _LazyAutoMapping(
     CONFIG_MAPPING_NAMES, MODEL_FOR_CAUSAL_IMAGE_MODELING_MAPPING_NAMES
@@ -1731,6 +1810,7 @@ MODEL_FOR_UNIVERSAL_SEGMENTATION_MAPPING = _LazyAutoMapping(
 MODEL_FOR_VIDEO_CLASSIFICATION_MAPPING = _LazyAutoMapping(
     CONFIG_MAPPING_NAMES, MODEL_FOR_VIDEO_CLASSIFICATION_MAPPING_NAMES
 )
+MODEL_FOR_VISION_2_SEQ_MAPPING = _LazyAutoMapping(CONFIG_MAPPING_NAMES, MODEL_FOR_VISION_2_SEQ_MAPPING_NAMES)
 MODEL_FOR_IMAGE_TEXT_TO_TEXT_MAPPING = _LazyAutoMapping(
     CONFIG_MAPPING_NAMES, MODEL_FOR_IMAGE_TEXT_TO_TEXT_MAPPING_NAMES
 )
@@ -1850,6 +1930,14 @@ class AutoModelForPreTraining(_BaseAutoModelClass):
 AutoModelForPreTraining = auto_class_update(AutoModelForPreTraining, head_doc="pretraining")
 
 
+# Private on purpose, the public class will add the deprecation warnings.
+class _AutoModelWithLMHead(_BaseAutoModelClass):
+    _model_mapping = MODEL_WITH_LM_HEAD_MAPPING
+
+
+_AutoModelWithLMHead = auto_class_update(_AutoModelWithLMHead, head_doc="language modeling")
+
+
 class AutoModelForCausalLM(_BaseAutoModelClass):
     _model_mapping = MODEL_FOR_CAUSAL_LM_MAPPING
 
@@ -1857,7 +1945,7 @@ class AutoModelForCausalLM(_BaseAutoModelClass):
     @classmethod
     def from_pretrained(
         cls: type["AutoModelForCausalLM"],
-        pretrained_model_name_or_path: str | os.PathLike[str],
+        pretrained_model_name_or_path: Union[str, os.PathLike[str]],
         *model_args,
         **kwargs,
     ) -> "_BaseModelWithGenerate":
@@ -2046,6 +2134,14 @@ class AutoModelForVideoClassification(_BaseAutoModelClass):
 AutoModelForVideoClassification = auto_class_update(AutoModelForVideoClassification, head_doc="video classification")
 
 
+# Private on purpose, the public class will add the deprecation warnings.
+class _AutoModelForVision2Seq(_BaseAutoModelClass):
+    _model_mapping = MODEL_FOR_VISION_2_SEQ_MAPPING
+
+
+_AutoModelForVision2Seq = auto_class_update(_AutoModelForVision2Seq, head_doc="vision-to-text modeling")
+
+
 class AutoModelForImageTextToText(_BaseAutoModelClass):
     _model_mapping = MODEL_FOR_IMAGE_TEXT_TO_TEXT_MAPPING
 
@@ -2053,7 +2149,7 @@ class AutoModelForImageTextToText(_BaseAutoModelClass):
     @classmethod
     def from_pretrained(
         cls: type["AutoModelForImageTextToText"],
-        pretrained_model_name_or_path: str | os.PathLike[str],
+        pretrained_model_name_or_path: Union[str, os.PathLike[str]],
         *model_args,
         **kwargs,
     ) -> "_BaseModelWithGenerate":
@@ -2137,6 +2233,48 @@ AutoModelForAudioTokenization = auto_class_update(
 )
 
 
+class AutoModelWithLMHead(_AutoModelWithLMHead):
+    @classmethod
+    def from_config(cls, config, **kwargs):
+        warnings.warn(
+            "The 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.",
+            FutureWarning,
+        )
+        return super().from_config(config, **kwargs)
+
+    @classmethod
+    def from_pretrained(cls, pretrained_model_name_or_path, *model_args, **kwargs):
+        warnings.warn(
+            "The 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.",
+            FutureWarning,
+        )
+        return super().from_pretrained(pretrained_model_name_or_path, *model_args, **kwargs)
+
+
+class AutoModelForVision2Seq(_AutoModelForVision2Seq):
+    @classmethod
+    def from_config(cls, config, **kwargs):
+        warnings.warn(
+            "The class `AutoModelForVision2Seq` is deprecated and will be removed in v5.0. Please use "
+            "`AutoModelForImageTextToText` instead.",
+            FutureWarning,
+        )
+        return super().from_config(config, **kwargs)
+
+    @classmethod
+    def from_pretrained(cls, pretrained_model_name_or_path, *model_args, **kwargs):
+        warnings.warn(
+            "The class `AutoModelForVision2Seq` is deprecated and will be removed in v5.0. Please use "
+            "`AutoModelForImageTextToText` instead.",
+            FutureWarning,
+        )
+        return super().from_pretrained(pretrained_model_name_or_path, *model_args, **kwargs)
+
+
 __all__ = [
     "MODEL_FOR_AUDIO_CLASSIFICATION_MAPPING",
     "MODEL_FOR_AUDIO_FRAME_CLASSIFICATION_MAPPING",
@@ -2175,11 +2313,13 @@ __all__ = [
     "MODEL_FOR_TOKEN_CLASSIFICATION_MAPPING",
     "MODEL_FOR_UNIVERSAL_SEGMENTATION_MAPPING",
     "MODEL_FOR_VIDEO_CLASSIFICATION_MAPPING",
+    "MODEL_FOR_VISION_2_SEQ_MAPPING",
     "MODEL_FOR_RETRIEVAL_MAPPING",
     "MODEL_FOR_IMAGE_TEXT_TO_TEXT_MAPPING",
     "MODEL_FOR_MULTIMODAL_LM_MAPPING",
     "MODEL_FOR_VISUAL_QUESTION_ANSWERING_MAPPING",
     "MODEL_MAPPING",
+    "MODEL_WITH_LM_HEAD_MAPPING",
     "MODEL_FOR_ZERO_SHOT_IMAGE_CLASSIFICATION_MAPPING",
     "MODEL_FOR_ZERO_SHOT_OBJECT_DETECTION_MAPPING",
     "MODEL_FOR_TIME_SERIES_CLASSIFICATION_MAPPING",
@@ -2220,8 +2360,10 @@ __all__ = [
     "AutoModelForTokenClassification",
     "AutoModelForUniversalSegmentation",
     "AutoModelForVideoClassification",
+    "AutoModelForVision2Seq",
     "AutoModelForVisualQuestionAnswering",
     "AutoModelForDocumentQuestionAnswering",
+    "AutoModelWithLMHead",
     "AutoModelForZeroShotImageClassification",
     "AutoModelForZeroShotObjectDetection",
     "AutoModelForImageTextToText",

transformers/models/auto/processing_auto.py

@@ -1,3 +1,4 @@
+# coding=utf-8
 # Copyright 2021 The HuggingFace Inc. team.
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
@@ -67,7 +68,6 @@ PROCESSOR_MAPPING_NAMES = OrderedDict(
         ("dia", "DiaProcessor"),
         ("edgetam", "Sam2Processor"),
         ("emu3", "Emu3Processor"),
-        ("ernie4_5_vl_moe", "Ernie4_5_VL_MoeProcessor"),
         ("evolla", "EvollaProcessor"),
         ("flava", "FlavaProcessor"),
         ("florence2", "Florence2Processor"),
@@ -78,8 +78,6 @@ PROCESSOR_MAPPING_NAMES = OrderedDict(
         ("glm46v", "Glm46VProcessor"),
         ("glm4v", "Glm4vProcessor"),
         ("glm4v_moe", "Glm4vProcessor"),
-        ("glm_image", "Glm4vProcessor"),
-        ("glmasr", "GlmAsrProcessor"),
         ("got_ocr2", "GotOcr2Processor"),
         ("granite_speech", "GraniteSpeechProcessor"),
         ("grounding-dino", "GroundingDinoProcessor"),
@@ -95,13 +93,10 @@ PROCESSOR_MAPPING_NAMES = OrderedDict(
         ("kosmos-2", "Kosmos2Processor"),
         ("kosmos-2.5", "Kosmos2_5Processor"),
         ("kyutai_speech_to_text", "KyutaiSpeechToTextProcessor"),
-        ("lasr_ctc", "LasrProcessor"),
-        ("lasr_encoder", "LasrProcessor"),
         ("layoutlmv2", "LayoutLMv2Processor"),
         ("layoutlmv3", "LayoutLMv3Processor"),
         ("layoutxlm", "LayoutXLMProcessor"),
         ("lfm2_vl", "Lfm2VlProcessor"),
-        ("lighton_ocr", "LightOnOcrProcessor"),
         ("llama4", "Llama4Processor"),
         ("llava", "LlavaProcessor"),
         ("llava_next", "LlavaNextProcessor"),
@@ -119,7 +114,6 @@ PROCESSOR_MAPPING_NAMES = OrderedDict(
         ("ovis2", "Ovis2Processor"),
         ("owlv2", "Owlv2Processor"),
         ("owlvit", "OwlViTProcessor"),
-        ("paddleocr_vl", "PaddleOCRVLProcessor"),
         ("paligemma", "PaliGemmaProcessor"),
         ("perception_lm", "PerceptionLMProcessor"),
         ("phi4_multimodal", "Phi4MultimodalProcessor"),
@@ -323,6 +317,7 @@ class AutoProcessor:
                     processor_class = config_dict.get("processor_class", None)
                     if "AutoProcessor" in config_dict.get("auto_map", {}):
                         processor_auto_map = config_dict["auto_map"]["AutoProcessor"]
+
             # Saved as feature extractor
             if preprocessor_config_file is None:
                 preprocessor_config_file = cached_file(
@@ -350,24 +345,16 @@ class AutoProcessor:
                     processor_auto_map = config_dict["auto_map"]["AutoProcessor"]
 
         if processor_class is None:
-            # Last resort: try loading the model config to get processor_class.
-            # This handles cases where processor info is only in config.json (not in any
-            # preprocessor/tokenizer config files). AutoConfig.from_pretrained may raise
-            # ValueError if the model_type is unrecognized or the config is invalid -
-            # we catch and ignore this to allow fallback to AutoTokenizer/AutoImageProcessor.
-            try:
-                if not isinstance(config, PreTrainedConfig):
-                    config = AutoConfig.from_pretrained(
-                        pretrained_model_name_or_path, trust_remote_code=trust_remote_code, **kwargs
-                    )
+            # Otherwise, load config, if it can be loaded.
+            if not isinstance(config, PreTrainedConfig):
+                config = AutoConfig.from_pretrained(
+                    pretrained_model_name_or_path, trust_remote_code=trust_remote_code, **kwargs
+                )
 
-                processor_class = getattr(config, "processor_class", None)
-                if hasattr(config, "auto_map") and "AutoProcessor" in config.auto_map:
-                    processor_auto_map = config.auto_map["AutoProcessor"]
-            except ValueError:
-                # Config loading failed (unrecognized model_type, invalid config, etc.)
-                # Continue to fallback logic below (AutoTokenizer, AutoImageProcessor, etc.)
-                pass
+            # And check if the config contains the processor class.
+            processor_class = getattr(config, "processor_class", None)
+            if hasattr(config, "auto_map") and "AutoProcessor" in config.auto_map:
+                processor_auto_map = config.auto_map["AutoProcessor"]
 
         if processor_class is not None:
             processor_class = processor_class_from_name(processor_class)