Fix from_args_and_dict ProcessorMixin (#38296)

* fix-from-args-and-dict-processormixin

* change used_kwargs to valid_kwargs

* remove manual valid_kwargs

* fix copies

* fix modular aria

Author: yonigozlan

src/transformers/models/aria/modular_aria.py

@@ -936,7 +936,6 @@ class AriaProcessor(ProcessorMixin):
     """
 
     attributes = ["image_processor", "tokenizer"]
-    valid_kwargs = ["chat_template", "size_conversion"]
     image_processor_class = "AriaImageProcessor"
     tokenizer_class = "AutoTokenizer"
 

src/transformers/models/aria/processing_aria.py

@@ -60,7 +60,6 @@ class AriaProcessor(ProcessorMixin):
     """
 
     attributes = ["image_processor", "tokenizer"]
-    valid_kwargs = ["chat_template", "size_conversion"]
     image_processor_class = "AriaImageProcessor"
     tokenizer_class = "AutoTokenizer"
 

src/transformers/models/aya_vision/processing_aya_vision.py

@@ -18,17 +18,8 @@
 import numpy as np
 
 from ...image_processing_utils import BatchFeature
-from ...image_utils import (
-    ImageInput,
-    make_flat_list_of_images,
-)
-from ...processing_utils import (
-    ImagesKwargs,
-    MultiModalData,
-    ProcessingKwargs,
-    ProcessorMixin,
-    Unpack,
-)
+from ...image_utils import ImageInput, make_flat_list_of_images
+from ...processing_utils import ImagesKwargs, MultiModalData, ProcessingKwargs, ProcessorMixin, Unpack
 from ...tokenization_utils_base import PreTokenizedInput, TextInput
 
 
@@ -87,19 +78,6 @@ class AyaVisionProcessor(ProcessorMixin):
     """
 
     attributes = ["image_processor", "tokenizer"]
-    valid_kwargs = [
-        "chat_template",
-        "image_token",
-        "patch_size",
-        "img_size",
-        "downsample_factor",
-        "start_of_img_token",
-        "end_of_img_token",
-        "img_patch_token",
-        "img_line_break_token",
-        "tile_token",
-        "tile_global_token",
-    ]
     image_processor_class = "AutoImageProcessor"
     tokenizer_class = "AutoTokenizer"
 

src/transformers/models/blip/processing_blip.py

@@ -55,7 +55,6 @@ class BlipProcessor(ProcessorMixin):
     """
 
     attributes = ["image_processor", "tokenizer"]
-    valid_kwargs = []
     image_processor_class = ("BlipImageProcessor", "BlipImageProcessorFast")
     tokenizer_class = ("BertTokenizer", "BertTokenizerFast")
 

src/transformers/models/blip_2/processing_blip_2.py

@@ -21,12 +21,7 @@
 from ...image_processing_utils import BatchFeature
 from ...image_utils import ImageInput
 from ...processing_utils import ProcessingKwargs, ProcessorMixin, Unpack
-from ...tokenization_utils_base import (
-    AddedToken,
-    BatchEncoding,
-    PreTokenizedInput,
-    TextInput,
-)
+from ...tokenization_utils_base import AddedToken, BatchEncoding, PreTokenizedInput, TextInput
 from ...utils import logging
 
 
@@ -67,7 +62,6 @@ class Blip2Processor(ProcessorMixin):
     """
 
     attributes = ["image_processor", "tokenizer"]
-    valid_kwargs = ["num_query_tokens"]
     image_processor_class = ("BlipImageProcessor", "BlipImageProcessorFast")
     tokenizer_class = "AutoTokenizer"
 

src/transformers/models/chameleon/processing_chameleon.py

@@ -72,7 +72,6 @@ class ChameleonProcessor(ProcessorMixin):
 
     attributes = ["image_processor", "tokenizer"]
     tokenizer_class = ("LlamaTokenizer", "LlamaTokenizerFast")
-    valid_kwargs = ["image_seq_length", "image_token"]
     image_processor_class = "ChameleonImageProcessor"
 
     def __init__(self, image_processor, tokenizer, image_seq_length: int = 1024, image_token: str = "<image>"):

src/transformers/models/colpali/processing_colpali.py

@@ -90,7 +90,6 @@ class ColPaliProcessor(ProcessorMixin):
     """
 
     attributes = ["image_processor", "tokenizer"]
-    valid_kwargs = ["chat_template"]
     image_processor_class = ("SiglipImageProcessor", "SiglipImageProcessorFast")
     tokenizer_class = ("GemmaTokenizer", "GemmaTokenizerFast")
 

src/transformers/models/csm/processing_csm.py

@@ -31,10 +31,7 @@
 from ...audio_utils import AudioInput, make_list_of_audio
 from ...feature_extraction_utils import BatchFeature
 from ...processing_utils import AudioKwargs, ProcessingKwargs, ProcessorMixin, Unpack
-from ...tokenization_utils_base import (
-    PreTokenizedInput,
-    TextInput,
-)
+from ...tokenization_utils_base import PreTokenizedInput, TextInput
 
 
 class CsmAudioKwargs(AudioKwargs, total=False):
@@ -99,7 +96,6 @@ class CsmProcessor(ProcessorMixin):
     """
 
     attributes = ["feature_extractor", "tokenizer"]
-    valid_kwargs = ["chat_template"]
     feature_extractor_class = "EncodecFeatureExtractor"
     tokenizer_class = "PreTrainedTokenizerFast"
 

src/transformers/models/emu3/processing_emu3.py

@@ -71,7 +71,6 @@ class Emu3Processor(ProcessorMixin):
     """
 
     attributes = ["image_processor", "tokenizer"]
-    valid_kwargs = ["chat_template"]
     tokenizer_class = ("GPT2Tokenizer", "GPT2TokenizerFast")
     image_processor_class = "Emu3ImageProcessor"
 

src/transformers/models/fuyu/processing_fuyu.py

@@ -350,7 +350,6 @@ class FuyuProcessor(ProcessorMixin):
     """
 
     attributes = ["image_processor", "tokenizer"]
-    valid_kwargs = []
     image_processor_class = "FuyuImageProcessor"
     tokenizer_class = "AutoTokenizer"
 

src/transformers/models/gemma3/processing_gemma3.py

@@ -51,7 +51,6 @@ class Gemma3ProcessorKwargs(ProcessingKwargs, total=False):
 
 class Gemma3Processor(ProcessorMixin):
     attributes = ["image_processor", "tokenizer"]
-    valid_kwargs = ["chat_template", "image_seq_length"]
     image_processor_class = "AutoImageProcessor"
     tokenizer_class = "AutoTokenizer"
 

src/transformers/models/got_ocr2/processing_got_ocr2.py

@@ -95,7 +95,6 @@ class GotOcr2Processor(ProcessorMixin):
     """
 
     attributes = ["image_processor", "tokenizer"]
-    valid_kwargs = ["chat_template"]
     image_processor_class = "AutoImageProcessor"
     tokenizer_class = "PreTrainedTokenizerFast"
 

src/transformers/models/granite_speech/processing_granite_speech.py

@@ -31,8 +31,6 @@
 
 class GraniteSpeechProcessor(ProcessorMixin):
     attributes = ["audio_processor", "tokenizer"]
-    valid_kwargs = ["audio_token"]
-
     audio_processor_class = "GraniteSpeechFeatureExtractor"
     tokenizer_class = "AutoTokenizer"
 

src/transformers/models/idefics/processing_idefics.py

@@ -211,7 +211,6 @@ class IdeficsProcessor(ProcessorMixin):
     """
 
     attributes = ["image_processor", "tokenizer"]
-    valid_kwargs = ["image_size", "add_end_of_utterance_token"]
     image_processor_class = "IdeficsImageProcessor"
     tokenizer_class = "LlamaTokenizerFast"
 

src/transformers/models/idefics2/processing_idefics2.py

@@ -85,7 +85,6 @@ class Idefics2Processor(ProcessorMixin):
     """
 
     attributes = ["image_processor", "tokenizer"]
-    valid_kwargs = ["image_seq_len", "chat_template"]
     image_processor_class = "Idefics2ImageProcessor"
     tokenizer_class = "AutoTokenizer"
 

src/transformers/models/idefics3/processing_idefics3.py

@@ -133,7 +133,6 @@ class Idefics3Processor(ProcessorMixin):
     """
 
     attributes = ["image_processor", "tokenizer"]
-    valid_kwargs = ["image_seq_len", "chat_template"]
     image_processor_class = "Idefics3ImageProcessor"
     tokenizer_class = "AutoTokenizer"
 

src/transformers/models/instructblip/processing_instructblip.py

@@ -22,12 +22,7 @@
 from ...image_processing_utils import BatchFeature
 from ...image_utils import ImageInput
 from ...processing_utils import ProcessingKwargs, ProcessorMixin, Unpack
-from ...tokenization_utils_base import (
-    AddedToken,
-    BatchEncoding,
-    PreTokenizedInput,
-    TextInput,
-)
+from ...tokenization_utils_base import AddedToken, BatchEncoding, PreTokenizedInput, TextInput
 from ...utils import logging
 from ..auto import AutoTokenizer
 
@@ -72,7 +67,6 @@ class InstructBlipProcessor(ProcessorMixin):
     """
 
     attributes = ["image_processor", "tokenizer", "qformer_tokenizer"]
-    valid_kwargs = ["num_query_tokens"]
     image_processor_class = ("BlipImageProcessor", "BlipImageProcessorFast")
     tokenizer_class = "AutoTokenizer"
     qformer_tokenizer_class = "AutoTokenizer"

src/transformers/models/instructblipvideo/processing_instructblipvideo.py

@@ -57,7 +57,6 @@ class InstructBlipVideoProcessor(ProcessorMixin):
     """
 
     attributes = ["video_processor", "tokenizer", "qformer_tokenizer"]
-    valid_kwargs = ["num_query_tokens"]
     video_processor_class = "AutoVideoProcessor"
     tokenizer_class = "AutoTokenizer"
     qformer_tokenizer_class = "AutoTokenizer"

src/transformers/models/internvl/processing_internvl.py

@@ -18,18 +18,8 @@
 import numpy as np
 
 from ...image_processing_utils import BatchFeature
-from ...image_utils import (
-    ImageInput,
-    concatenate_list,
-    make_flat_list_of_images,
-)
-from ...processing_utils import (
-    ImagesKwargs,
-    MultiModalData,
-    ProcessingKwargs,
-    ProcessorMixin,
-    Unpack,
-)
+from ...image_utils import ImageInput, concatenate_list, make_flat_list_of_images
+from ...processing_utils import ImagesKwargs, MultiModalData, ProcessingKwargs, ProcessorMixin, Unpack
 from ...tokenization_utils_base import PreTokenizedInput, TextInput
 from ...video_utils import VideoInput, VideoMetadata, load_video, make_batched_videos
 
@@ -74,10 +64,6 @@ class InternVLProcessor(ProcessorMixin):
     """
 
     attributes = ["image_processor", "tokenizer", "video_processor"]
-    valid_kwargs = [
-        "chat_template",
-        "image_seq_length",
-    ]
     image_processor_class = "AutoImageProcessor"
     video_processor_class = "AutoVideoProcessor"
     tokenizer_class = "AutoTokenizer"

src/transformers/models/janus/processing_janus.py

@@ -21,10 +21,7 @@
 from ...feature_extraction_utils import BatchFeature
 from ...image_utils import ImageInput
 from ...processing_utils import ProcessingKwargs, ProcessorMixin, TextKwargs, Unpack
-from ...tokenization_utils_base import (
-    PreTokenizedInput,
-    TextInput,
-)
+from ...tokenization_utils_base import PreTokenizedInput, TextInput
 from ...utils import logging
 
 
@@ -68,7 +65,6 @@ class JanusProcessor(ProcessorMixin):
     """
 
     attributes = ["image_processor", "tokenizer"]
-    valid_kwargs = ["chat_template", "use_default_system_prompt"]
     image_processor_class = "JanusImageProcessor"
     tokenizer_class = "LlamaTokenizerFast"
 

src/transformers/models/kosmos2/processing_kosmos2.py

@@ -84,7 +84,6 @@ class Kosmos2Processor(ProcessorMixin):
     """
 
     attributes = ["image_processor", "tokenizer"]
-    valid_kwargs = ["num_patch_index_tokens"]
     image_processor_class = ("CLIPImageProcessor", "CLIPImageProcessorFast")
     tokenizer_class = "AutoTokenizer"
 

src/transformers/models/llama4/processing_llama4.py

@@ -16,19 +16,11 @@
 
 from typing import List, Optional, Union
 
-from transformers.processing_utils import (
-    ImagesKwargs,
-    ProcessingKwargs,
-    ProcessorMixin,
-    Unpack,
-)
+from transformers.processing_utils import ImagesKwargs, ProcessingKwargs, ProcessorMixin, Unpack
 from transformers.tokenization_utils_base import PreTokenizedInput, TextInput
 
 from ...image_processing_utils import BatchFeature
-from ...image_utils import (
-    ImageInput,
-    make_flat_list_of_images,
-)
+from ...image_utils import ImageInput, make_flat_list_of_images
 
 
 class Llama4ImagesKwargs(ImagesKwargs, total=False):
@@ -83,19 +75,6 @@ class Llama4Processor(ProcessorMixin):
     """
 
     attributes = ["image_processor", "tokenizer"]
-    valid_kwargs = [
-        "chat_template",
-        "image_token",
-        "patch_size",
-        "img_size",
-        "downsample_factor",
-        "start_of_img_token",
-        "end_of_img_token",
-        "img_patch_token",
-        "img_line_break_token",
-        "tile_token",
-        "tile_global_token",
-    ]
     image_processor_class = "AutoImageProcessor"
     tokenizer_class = "AutoTokenizer"
 

src/transformers/models/llava/processing_llava.py

@@ -70,13 +70,6 @@ class LlavaProcessor(ProcessorMixin):
     """
 
     attributes = ["image_processor", "tokenizer"]
-    valid_kwargs = [
-        "chat_template",
-        "patch_size",
-        "vision_feature_select_strategy",
-        "image_token",
-        "num_additional_image_tokens",
-    ]
     image_processor_class = "AutoImageProcessor"
     tokenizer_class = "AutoTokenizer"
 

src/transformers/models/llava_next/processing_llava_next.py

@@ -76,13 +76,6 @@ class LlavaNextProcessor(ProcessorMixin):
     """
 
     attributes = ["image_processor", "tokenizer"]
-    valid_kwargs = [
-        "chat_template",
-        "patch_size",
-        "vision_feature_select_strategy",
-        "image_token",
-        "num_additional_image_tokens",
-    ]
     image_processor_class = "AutoImageProcessor"
     tokenizer_class = "AutoTokenizer"
 

src/transformers/models/llava_next_video/processing_llava_next_video.py

@@ -78,14 +78,6 @@ class LlavaNextVideoProcessor(ProcessorMixin):
     # video and image processor share same args, but have different processing logic
     # only image processor config is saved in the hub
     attributes = ["video_processor", "image_processor", "tokenizer"]
-    valid_kwargs = [
-        "chat_template",
-        "patch_size",
-        "vision_feature_select_strategy",
-        "image_token",
-        "video_token",
-        "num_additional_image_tokens",
-    ]
     image_processor_class = ("LlavaNextImageProcessor", "LlavaNextImageProcessorFast")
     video_processor_class = "AutoVideoProcessor"
     tokenizer_class = ("LlamaTokenizer", "LlamaTokenizerFast")

src/transformers/models/llava_onevision/processing_llava_onevision.py

@@ -75,14 +75,6 @@ class LlavaOnevisionProcessor(ProcessorMixin):
     """
 
     attributes = ["image_processor", "tokenizer", "video_processor"]
-    valid_kwargs = [
-        "chat_template",
-        "num_image_tokens",
-        "vision_feature_select_strategy",
-        "image_token",
-        "video_token",
-        "vision_aspect_ratio",
-    ]
     image_processor_class = "AutoImageProcessor"
     tokenizer_class = "AutoTokenizer"
     video_processor_class = "AutoVideoProcessor"

src/transformers/models/mllama/processing_mllama.py

@@ -22,10 +22,7 @@
 from ...feature_extraction_utils import BatchFeature
 from ...image_utils import ImageInput, make_nested_list_of_images
 from ...processing_utils import ImagesKwargs, ProcessingKwargs, ProcessorMixin, Unpack
-from ...tokenization_utils_base import (
-    PreTokenizedInput,
-    TextInput,
-)
+from ...tokenization_utils_base import PreTokenizedInput, TextInput
 
 
 class MllamaImagesKwargs(ImagesKwargs, total=False):
@@ -208,7 +205,6 @@ class MllamaProcessor(ProcessorMixin):
     """
 
     attributes = ["image_processor", "tokenizer"]
-    valid_kwargs = ["chat_template"]
     image_processor_class = "MllamaImageProcessor"
     tokenizer_class = "PreTrainedTokenizerFast"