Add ImageProcessorFast to Qwen2.5-VL processor (#36164)

* add qwen2 fast image processor to modular file

Signed-off-by: isotr0py <2037008807@qq.com>

* fix modular

Signed-off-by: isotr0py <2037008807@qq.com>

* fix circle import

Signed-off-by: isotr0py <2037008807@qq.com>

* add docs

Signed-off-by: isotr0py <2037008807@qq.com>

* fix typo

Signed-off-by: isotr0py <2037008807@qq.com>

* add modular generated files

Signed-off-by: isotr0py <2037008807@qq.com>

* revert qwen2vl fast image processor

Signed-off-by: isotr0py <2037008807@qq.com>

* remove qwen2.5-vl image processor from modular

Signed-off-by: isotr0py <2037008807@qq.com>

* re-generate qwen2.5-vl files

Signed-off-by: isotr0py <2037008807@qq.com>

* remove unnecessary test

Signed-off-by: isotr0py <2037008807@qq.com>

* fix auto map

Signed-off-by: isotr0py <2037008807@qq.com>

* cleanup

Signed-off-by: isotr0py <2037008807@qq.com>

* fix model_input_names

Signed-off-by: isotr0py <2037008807@qq.com>

* remove import

Signed-off-by: isotr0py <2037008807@qq.com>

* make fix-copies

Signed-off-by: isotr0py <2037008807@qq.com>

---------

Signed-off-by: isotr0py <2037008807@qq.com>

docs/source/en/model_doc/qwen2_5_vl.md

@@ -264,11 +264,6 @@ model = Qwen2_5_VLForConditionalGeneration.from_pretrained(
 
 [[autodoc]] Qwen2_5_VLConfig
 
-## Qwen2_5_VLImageProcessor
-
-[[autodoc]] Qwen2_5_VLImageProcessor
-    - preprocess
-
 ## Qwen2_5_VLProcessor
 
 [[autodoc]] Qwen2_5_VLProcessor

src/transformers/__init__.py

@@ -1281,7 +1281,6 @@
     _import_structure["models.pixtral"].append("PixtralImageProcessor")
     _import_structure["models.poolformer"].extend(["PoolFormerFeatureExtractor", "PoolFormerImageProcessor"])
     _import_structure["models.pvt"].extend(["PvtImageProcessor"])
-    _import_structure["models.qwen2_5_vl"].extend(["Qwen2_5_VLImageProcessor"])
     _import_structure["models.qwen2_vl"].extend(["Qwen2VLImageProcessor"])
     _import_structure["models.rt_detr"].extend(["RTDetrImageProcessor"])
     _import_structure["models.sam"].extend(["SamImageProcessor"])
@@ -6444,7 +6443,6 @@
             PoolFormerImageProcessor,
         )
         from .models.pvt import PvtImageProcessor
-        from .models.qwen2_5_vl import Qwen2_5_VLImageProcessor
         from .models.qwen2_vl import Qwen2VLImageProcessor
         from .models.rt_detr import RTDetrImageProcessor
         from .models.sam import SamImageProcessor

src/transformers/models/auto/image_processing_auto.py

@@ -127,6 +127,7 @@
             ("poolformer", ("PoolFormerImageProcessor",)),
             ("pvt", ("PvtImageProcessor",)),
             ("pvt_v2", ("PvtImageProcessor",)),
+            ("qwen2_5_vl", ("Qwen2VLImageProcessor", "Qwen2VLImageProcessorFast")),
             ("qwen2_vl", ("Qwen2VLImageProcessor", "Qwen2VLImageProcessorFast")),
             ("regnet", ("ConvNextImageProcessor", "ConvNextImageProcessorFast")),
             ("resnet", ("ConvNextImageProcessor", "ConvNextImageProcessorFast")),

src/transformers/models/qwen2_5_vl/__init__.py

@@ -19,7 +19,6 @@
 
 if TYPE_CHECKING:
     from .configuration_qwen2_5_vl import *
-    from .image_processing_qwen2_5_vl import *
     from .modeling_qwen2_5_vl import *
     from .processing_qwen2_5_vl import *
 else:

src/transformers/models/qwen2_5_vl/modular_qwen2_5_vl.py

@@ -29,7 +29,6 @@
 from torch.nn import CrossEntropyLoss
 
 from transformers.models.qwen2_vl.configuration_qwen2_vl import Qwen2VLConfig
-from transformers.models.qwen2_vl.image_processing_qwen2_vl import Qwen2VLImageProcessor
 from transformers.models.qwen2_vl.modeling_qwen2_vl import (
     PatchEmbed,
     PatchMerger,
@@ -854,48 +853,6 @@ def prepare_inputs_for_generation(
         return model_inputs
 
 
-class Qwen2_5_VLImageProcessor(Qwen2VLImageProcessor):
-    r"""
-    Constructs a Qwen2.5-VL image processor that dynamically resizes images based on the original images.
-
-    Args:
-        do_resize (`bool`, *optional*, defaults to `True`):
-            Whether to resize the image's (height, width) dimensions.
-        resample (`PILImageResampling`, *optional*, defaults to `Resampling.BICUBIC`):
-            Resampling filter to use when resizing the image.
-        do_rescale (`bool`, *optional*, defaults to `True`):
-            Whether to rescale the image by the specified scale `rescale_factor`.
-        rescale_factor (`int` or `float`, *optional*, defaults to `1/255`):
-            Scale factor to use if rescaling the image.
-        do_normalize (`bool`, *optional*, defaults to `True`):
-            Whether to normalize the image.
-        image_mean (`float` or `List[float]`, *optional*, defaults to `[0.48145466, 0.4578275, 0.40821073]`):
-            Mean to use if normalizing the image. This is a float or list of floats for each channel in the image.
-        image_std (`float` or `List[float]`, *optional*, defaults to `[0.26862954, 0.26130258, 0.27577711]`):
-            Standard deviation to use if normalizing the image. This is a float or list of floats for each channel in the image.
-        do_convert_rgb (`bool`, *optional*, defaults to `True`):
-            Whether to convert the image to RGB.
-        min_pixels (`int`, *optional*, defaults to `56 * 56`):
-            The min pixels of the image to resize the image.
-        max_pixels (`int`, *optional*, defaults to `28 * 28 * 1280`):
-            The max pixels of the image to resize the image.
-        patch_size (`int`, *optional*, defaults to 14):
-            The spacial patch size of the vision encoder.
-        temporal_patch_size (`int`, *optional*, defaults to 2):
-            The temporal patch size of the vision encoder.
-        merge_size (`int`, *optional*, defaults to 2):
-            The merge size of the vision encoder to llm encoder.
-    """
-
-    model_input_names = [
-        "pixel_values",
-        "image_grid_thw",
-        "pixel_values_videos",
-        "video_grid_thw",
-        "second_per_grid_ts",
-    ]
-
-
 class Qwen2_5_VLVideosProcessorKwargs(VideosKwargs, total=False):
     fps: Union[List[float], float]
 
@@ -913,18 +870,25 @@ class Qwen2_5_VLProcessorKwargs(ProcessingKwargs, total=False):
 class Qwen2_5_VLProcessor(Qwen2VLProcessor):
     r"""
     Constructs a Qwen2.5-VL processor which wraps a Qwen2.5-VL image processor and a Qwen2 tokenizer into a single processor.
-    [`Qwen2_5_VLProcessor`] offers all the functionalities of [`Qwen2_5_VLImageProcessor`] and [`Qwen2TokenizerFast`]. See the
+    [`Qwen2_5_VLProcessor`] offers all the functionalities of [`Qwen2VLImageProcessor`] and [`Qwen2TokenizerFast`]. See the
     [`~Qwen2_5_VLProcessor.__call__`] and [`~Qwen2_5_VLProcessor.decode`] for more information.
     Args:
-        image_processor ([`Qwen2_5_VLImageProcessor`], *optional*):
+        image_processor ([`Qwen2VLImageProcessor`], *optional*):
             The image processor is a required input.
         tokenizer ([`Qwen2TokenizerFast`], *optional*):
             The tokenizer is a required input.
         chat_template (`str`, *optional*): A Jinja template which will be used to convert lists of messages
             in a chat into a tokenizable string.
     """
 
-    image_processor_class = "Qwen2_5_VLImageProcessor"
+    image_processor_class = "AutoImageProcessor"
+
+    @property
+    def model_input_names(self):
+        tokenizer_input_names = self.tokenizer.model_input_names
+        image_processor_input_names = self.image_processor.model_input_names
+        names_from_processor = list(dict.fromkeys(tokenizer_input_names + image_processor_input_names))
+        return names_from_processor + ["second_per_grid_ts"]
 
     def __call__(
         self,
@@ -937,7 +901,7 @@ def __call__(
         Main method to prepare for the model one or several sequences(s) and image(s). This method forwards the `text`
         and `kwargs` arguments to Qwen2TokenizerFast's [`~Qwen2TokenizerFast.__call__`] if `text` is not `None` to encode
         the text. To prepare the vision inputs, this method forwards the `vision_infos` and `kwrags` arguments to
-        Qwen2_5_VLImageProcessor's [`~Qwen2_5_VLImageProcessor.__call__`] if `vision_infos` is not `None`.
+        Qwen2VLImageProcessor's [`~Qwen2VLImageProcessor.__call__`] if `vision_infos` is not `None`.
 
         Args:
             images (`PIL.Image.Image`, `np.ndarray`, `torch.Tensor`, `List[PIL.Image.Image]`, `List[np.ndarray]`, `List[torch.Tensor]`):
@@ -1040,6 +1004,5 @@ def __call__(
     "Qwen2_5_VLForConditionalGeneration",
     "Qwen2_5_VLModel",
     "Qwen2_5_VLPreTrainedModel",
-    "Qwen2_5_VLImageProcessor",
     "Qwen2_5_VLProcessor",
 ]

src/transformers/models/qwen2_5_vl/processing_qwen2_5_vl.py

@@ -48,10 +48,10 @@ class Qwen2_5_VLProcessorKwargs(ProcessingKwargs, total=False):
 class Qwen2_5_VLProcessor(ProcessorMixin):
     r"""
     Constructs a Qwen2.5-VL processor which wraps a Qwen2.5-VL image processor and a Qwen2 tokenizer into a single processor.
-    [`Qwen2_5_VLProcessor`] offers all the functionalities of [`Qwen2_5_VLImageProcessor`] and [`Qwen2TokenizerFast`]. See the
+    [`Qwen2_5_VLProcessor`] offers all the functionalities of [`Qwen2VLImageProcessor`] and [`Qwen2TokenizerFast`]. See the
     [`~Qwen2_5_VLProcessor.__call__`] and [`~Qwen2_5_VLProcessor.decode`] for more information.
     Args:
-        image_processor ([`Qwen2_5_VLImageProcessor`], *optional*):
+        image_processor ([`Qwen2VLImageProcessor`], *optional*):
             The image processor is a required input.
         tokenizer ([`Qwen2TokenizerFast`], *optional*):
             The tokenizer is a required input.
@@ -62,7 +62,7 @@ class Qwen2_5_VLProcessor(ProcessorMixin):
     attributes = ["image_processor", "tokenizer"]
     valid_kwargs = ["chat_template"]
 
-    image_processor_class = "Qwen2_5_VLImageProcessor"
+    image_processor_class = "AutoImageProcessor"
     tokenizer_class = ("Qwen2Tokenizer", "Qwen2TokenizerFast")
 
     def __init__(self, image_processor=None, tokenizer=None, chat_template=None, **kwargs):
@@ -81,7 +81,7 @@ def __call__(
         Main method to prepare for the model one or several sequences(s) and image(s). This method forwards the `text`
         and `kwargs` arguments to Qwen2TokenizerFast's [`~Qwen2TokenizerFast.__call__`] if `text` is not `None` to encode
         the text. To prepare the vision inputs, this method forwards the `vision_infos` and `kwrags` arguments to
-        Qwen2_5_VLImageProcessor's [`~Qwen2_5_VLImageProcessor.__call__`] if `vision_infos` is not `None`.
+        Qwen2VLImageProcessor's [`~Qwen2VLImageProcessor.__call__`] if `vision_infos` is not `None`.
 
         Args:
             images (`PIL.Image.Image`, `np.ndarray`, `torch.Tensor`, `List[PIL.Image.Image]`, `List[np.ndarray]`, `List[torch.Tensor]`):
@@ -212,7 +212,8 @@ def post_process_image_text_to_text(self, generated_outputs):
     def model_input_names(self):
         tokenizer_input_names = self.tokenizer.model_input_names
         image_processor_input_names = self.image_processor.model_input_names
-        return list(dict.fromkeys(tokenizer_input_names + image_processor_input_names))
+        names_from_processor = list(dict.fromkeys(tokenizer_input_names + image_processor_input_names))
+        return names_from_processor + ["second_per_grid_ts"]
 
 
 __all__ = ["Qwen2_5_VLProcessor"]

src/transformers/utils/dummy_vision_objects.py

@@ -590,13 +590,6 @@ def __init__(self, *args, **kwargs):
         requires_backends(self, ["vision"])
 
 
-class Qwen2_5_VLImageProcessor(metaclass=DummyObject):
-    _backends = ["vision"]
-
-    def __init__(self, *args, **kwargs):
-        requires_backends(self, ["vision"])
-
-
 class Qwen2VLImageProcessor(metaclass=DummyObject):
     _backends = ["vision"]