transformers 5.0.0rc1__py3-none-any.whl → 5.0.0rc2__py3-none-any.whl

transformers/models/auto/feature_extraction_auto.py

@@ -38,6 +38,7 @@ logger = logging.get_logger(__name__)
 FEATURE_EXTRACTOR_MAPPING_NAMES = OrderedDict(
     [
         ("audio-spectrogram-transformer", "ASTFeatureExtractor"),
+        ("audioflamingo3", "WhisperFeatureExtractor"),
         ("clap", "ClapFeatureExtractor"),
         ("clvp", "ClvpFeatureExtractor"),
         ("csm", "EncodecFeatureExtractor"),
@@ -46,6 +47,7 @@ FEATURE_EXTRACTOR_MAPPING_NAMES = OrderedDict(
         ("dia", "DiaFeatureExtractor"),
         ("encodec", "EncodecFeatureExtractor"),
         ("gemma3n", "Gemma3nAudioFeatureExtractor"),
+        ("glmasr", "WhisperFeatureExtractor"),
         ("granite_speech", "GraniteSpeechFeatureExtractor"),
         ("hubert", "Wav2Vec2FeatureExtractor"),
         ("kyutai_speech_to_text", "KyutaiSpeechToTextFeatureExtractor"),
@@ -59,6 +61,8 @@ FEATURE_EXTRACTOR_MAPPING_NAMES = OrderedDict(
         ("musicgen_melody", "MusicgenMelodyFeatureExtractor"),
         ("parakeet_ctc", "ParakeetFeatureExtractor"),
         ("parakeet_encoder", "ParakeetFeatureExtractor"),
+        ("pe_audio", "PeAudioFeatureExtractor"),
+        ("pe_audio_video", "PeAudioFeatureExtractor"),
         ("phi4_multimodal", "Phi4MultimodalFeatureExtractor"),
         ("pop2piano", "Pop2PianoFeatureExtractor"),
         ("qwen2_5_omni", "WhisperFeatureExtractor"),
@@ -345,13 +349,13 @@ class AutoFeatureExtractor:
             )
             _ = kwargs.pop("code_revision", None)
             feature_extractor_class.register_for_auto_class()
-            return feature_extractor_class.from_dict(config_dict, **kwargs)
+            return feature_extractor_class.from_pretrained(pretrained_model_name_or_path, **kwargs)
         elif feature_extractor_class is not None:
-            return feature_extractor_class.from_dict(config_dict, **kwargs)
+            return feature_extractor_class.from_pretrained(pretrained_model_name_or_path, **kwargs)
         # Last try: we use the FEATURE_EXTRACTOR_MAPPING.
         elif type(config) in FEATURE_EXTRACTOR_MAPPING:
             feature_extractor_class = FEATURE_EXTRACTOR_MAPPING[type(config)]
-            return feature_extractor_class.from_dict(config_dict, **kwargs)
+            return feature_extractor_class.from_pretrained(pretrained_model_name_or_path, **kwargs)
 
         raise ValueError(
             f"Unrecognized feature extractor in {pretrained_model_name_or_path}. Should have a "

transformers/models/auto/image_processing_auto.py

@@ -100,6 +100,7 @@ else:
             ("efficientnet", ("EfficientNetImageProcessor", "EfficientNetImageProcessorFast")),
             ("emu3", ("Emu3ImageProcessor", None)),
             ("eomt", ("EomtImageProcessor", "EomtImageProcessorFast")),
+            ("ernie4_5_vl_moe", ("Ernie4_5_VL_MoeImageProcessor", "Ernie4_5_VL_MoeImageProcessorFast")),
             ("flava", ("FlavaImageProcessor", "FlavaImageProcessorFast")),
             ("florence2", ("CLIPImageProcessor", "CLIPImageProcessorFast")),
             ("focalnet", ("BitImageProcessor", "BitImageProcessorFast")),
@@ -159,6 +160,7 @@ else:
             ("perception_lm", (None, "PerceptionLMImageProcessorFast")),
             ("phi4_multimodal", (None, "Phi4MultimodalImageProcessorFast")),
             ("pix2struct", ("Pix2StructImageProcessor", "Pix2StructImageProcessorFast")),
+            ("pixio", ("BitImageProcessor", "BitImageProcessorFast")),
             ("pixtral", ("PixtralImageProcessor", "PixtralImageProcessorFast")),
             ("poolformer", ("PoolFormerImageProcessor", "PoolFormerImageProcessorFast")),
             ("prompt_depth_anything", ("PromptDepthAnythingImageProcessor", "PromptDepthAnythingImageProcessorFast")),
@@ -605,9 +607,9 @@ class AutoImageProcessor:
             image_processor_class = get_class_from_dynamic_module(class_ref, pretrained_model_name_or_path, **kwargs)
             _ = kwargs.pop("code_revision", None)
             image_processor_class.register_for_auto_class()
-            return image_processor_class.from_dict(config_dict, **kwargs)
+            return image_processor_class.from_pretrained(pretrained_model_name_or_path, *inputs, **kwargs)
         elif image_processor_class is not None:
-            return image_processor_class.from_dict(config_dict, **kwargs)
+            return image_processor_class.from_pretrained(pretrained_model_name_or_path, *inputs, **kwargs)
         # Last try: we use the IMAGE_PROCESSOR_MAPPING.
         elif type(config) in IMAGE_PROCESSOR_MAPPING:
             image_processor_tuple = IMAGE_PROCESSOR_MAPPING[type(config)]

transformers/models/auto/modeling_auto.py

@@ -145,6 +145,7 @@ 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"),
@@ -182,6 +183,8 @@ MODEL_MAPPING_NAMES = OrderedDict(
         ("glm4v_moe_vision", "Glm4vMoeVisionModel"),
         ("glm4v_text", "Glm4vTextModel"),
         ("glm4v_vision", "Glm4vVisionModel"),
+        ("glmasr", "GlmAsrForConditionalGeneration"),
+        ("glmasr_encoder", "GlmAsrEncoder"),
         ("glpn", "GLPNModel"),
         ("got_ocr2", "GotOcr2Model"),
         ("gpt-sw3", "GPT2Model"),
@@ -216,6 +219,7 @@ MODEL_MAPPING_NAMES = OrderedDict(
         ("instructblipvideo", "InstructBlipVideoModel"),
         ("internvl", "InternVLModel"),
         ("internvl_vision", "InternVLVisionModel"),
+        ("jais2", "Jais2Model"),
         ("jamba", "JambaModel"),
         ("janus", "JanusModel"),
         ("jetmoe", "JetMoeModel"),
@@ -304,6 +308,12 @@ 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"),
@@ -313,6 +323,7 @@ MODEL_MAPPING_NAMES = OrderedDict(
         ("phi3", "Phi3Model"),
         ("phi4_multimodal", "Phi4MultimodalModel"),
         ("phimoe", "PhimoeModel"),
+        ("pixio", "PixioModel"),
         ("pixtral", "PixtralVisionModel"),
         ("plbart", "PLBartModel"),
         ("poolformer", "PoolFormerModel"),
@@ -470,6 +481,7 @@ MODEL_FOR_PRETRAINING_MAPPING_NAMES = OrderedDict(
         ("fsmt", "FSMTForConditionalGeneration"),
         ("funnel", "FunnelForPreTraining"),
         ("gemma3", "Gemma3ForConditionalGeneration"),
+        ("glmasr", "GlmAsrForConditionalGeneration"),
         ("gpt-sw3", "GPT2LMHeadModel"),
         ("gpt2", "GPT2LMHeadModel"),
         ("gpt_bigcode", "GPTBigCodeForCausalLM"),
@@ -689,6 +701,7 @@ 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"),
@@ -798,6 +811,7 @@ MODEL_FOR_IMAGE_MAPPING_NAMES = OrderedDict(
         ("mobilenet_v2", "MobileNetV2Model"),
         ("mobilevit", "MobileViTModel"),
         ("mobilevitv2", "MobileViTV2Model"),
+        ("pixio", "PixioModel"),
         ("poolformer", "PoolFormerModel"),
         ("pvt", "PvtModel"),
         ("regnet", "RegNetModel"),
@@ -998,6 +1012,7 @@ 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"),
@@ -1050,6 +1065,7 @@ 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"),
@@ -1060,6 +1076,13 @@ MODEL_FOR_MULTIMODAL_LM_MAPPING_NAMES = OrderedDict(
     ]
 )
 
+MODEL_FOR_MASKED_LM_MAPPING_NAMES = OrderedDict(
+    [
+        # Model for Masked LM mapping
+        ("albert", "AlbertForMaskedLM"),
+    ]
+)
+
 MODEL_FOR_MASKED_LM_MAPPING_NAMES = OrderedDict(
     [
         # Model for Masked LM mapping
@@ -1156,6 +1179,7 @@ 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"),
@@ -1181,6 +1205,12 @@ MODEL_FOR_SEQ_TO_SEQ_CAUSAL_LM_MAPPING_NAMES = OrderedDict(
     ]
 )
 
+MODEL_FOR_SPEECH_SEQ_2_SEQ_MAPPING_NAMES = OrderedDict(
+    [
+        ("dia", "DiaForConditionalGeneration"),
+    ]
+)
+
 MODEL_FOR_SPEECH_SEQ_2_SEQ_MAPPING_NAMES = OrderedDict(
     [
         ("dia", "DiaForConditionalGeneration"),
@@ -1680,6 +1710,7 @@ MODEL_FOR_BACKBONE_MAPPING_NAMES = OrderedDict(
         ("hgnet_v2", "HGNetV2Backbone"),
         ("hiera", "HieraBackbone"),
         ("maskformer-swin", "MaskFormerSwinBackbone"),
+        ("pixio", "PixioBackbone"),
         ("pvt_v2", "PvtV2Backbone"),
         ("resnet", "ResNetBackbone"),
         ("rt_detr_resnet", "RTDetrResNetBackbone"),

transformers/models/auto/processing_auto.py

@@ -68,6 +68,7 @@ PROCESSOR_MAPPING_NAMES = OrderedDict(
         ("dia", "DiaProcessor"),
         ("edgetam", "Sam2Processor"),
         ("emu3", "Emu3Processor"),
+        ("ernie4_5_vl_moe", "Ernie4_5_VL_MoeProcessor"),
         ("evolla", "EvollaProcessor"),
         ("flava", "FlavaProcessor"),
         ("florence2", "Florence2Processor"),
@@ -78,6 +79,7 @@ PROCESSOR_MAPPING_NAMES = OrderedDict(
         ("glm46v", "Glm46VProcessor"),
         ("glm4v", "Glm4vProcessor"),
         ("glm4v_moe", "Glm4vProcessor"),
+        ("glmasr", "GlmAsrProcessor"),
         ("got_ocr2", "GotOcr2Processor"),
         ("granite_speech", "GraniteSpeechProcessor"),
         ("grounding-dino", "GroundingDinoProcessor"),
@@ -93,6 +95,8 @@ 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"),