[Bugfix] Fix models and tests for transformers v5 (#33977)

Signed-off-by: raushan <raushan@huggingface.co>
Signed-off-by: Raushan Turganbay <raushan.turganbay@alumni.nu.edu.kz>
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
Co-authored-by: Cyrus Leung <cyrus.tl.leung@gmail.com>
Co-authored-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
This commit is contained in:
Raushan Turganbay
2026-02-06 21:47:41 +08:00
committed by GitHub
co-authored by Cyrus Leung Harry Mellor
parent 51a7bda625
commit 85ee1d962b
12 changed files with 70 additions and 55 deletions
+18 -13
View File
@@ -24,14 +24,10 @@ class Ovis2_5ProcessorKwargs(ProcessingKwargs, total=False): # type: ignore[cal
"padding": False,
},
"images_kwargs": {
"convert_to_rgb": True,
"min_pixels": MIN_PIXELS,
"max_pixels": MAX_PIXELS,
"do_convert_rgb": True,
},
"videos_kwargs": {
"convert_to_rgb": True,
"min_pixels": MIN_PIXELS,
"max_pixels": MAX_PIXELS,
"do_convert_rgb": True,
},
}
@@ -160,6 +156,9 @@ class Ovis2_5Processor(ProcessorMixin):
- **second_per_grid_ts** -- list of video seconds per time grid.
Returned when `videos` is not `None`.
"""
min_pixels = kwargs.pop("min_pixels", MIN_PIXELS)
max_pixels = kwargs.pop("max_pixels", MAX_PIXELS)
output_kwargs = self._merge_kwargs(
Ovis2_5ProcessorKwargs,
tokenizer_init_kwargs=self.tokenizer.init_kwargs,
@@ -175,7 +174,10 @@ class Ovis2_5Processor(ProcessorMixin):
# Process each image
for image in images if isinstance(images, list) else [images]:
pixel_values, image_placeholders, grid = self.preprocess_multidata(
images=image, **output_kwargs["images_kwargs"]
images=image,
min_pixels=min_pixels,
max_pixels=max_pixels,
**output_kwargs["images_kwargs"],
)
processed_images.append(pixel_values)
image_placeholders_list.append(image_placeholders)
@@ -194,7 +196,10 @@ class Ovis2_5Processor(ProcessorMixin):
# Process each video
for video in videos if isinstance(videos, list) else [videos]:
pixel_values, video_placeholders, grid = self.preprocess_multidata(
video=video, **output_kwargs["videos_kwargs"]
video=video,
min_pixels=min_pixels,
max_pixels=max_pixels,
**output_kwargs["videos_kwargs"],
)
processed_videos.append(pixel_values)
videos_placeholders_list.append(video_placeholders)
@@ -378,7 +383,7 @@ class Ovis2_5Processor(ProcessorMixin):
self,
images: PIL.Image.Image | list[PIL.Image.Image] | None = None,
video: list[PIL.Image.Image] | np.ndarray | None = None,
convert_to_rgb: bool | None = True,
do_convert_rgb: bool | None = True,
min_pixels: int = MIN_PIXELS,
max_pixels: int = MAX_PIXELS,
return_tensors: str | None = "pt",
@@ -404,7 +409,7 @@ class Ovis2_5Processor(ProcessorMixin):
min_pixels if min_pixels is not None else MIN_PIXELS,
)
images = [
image.convert("RGB") if convert_to_rgb and image.mode != "RGB" else image
image.convert("RGB") if do_convert_rgb and image.mode != "RGB" else image
for image in images
]
@@ -420,9 +425,9 @@ class Ovis2_5Processor(ProcessorMixin):
max_pixels=max_pixels,
)
new_size = dict(height=resized_height, width=resized_width)
image_pt = self.image_processor.preprocess(
image, size=new_size, return_tensors="np"
)["pixel_values"][0]
image_pt = self.image_processor.preprocess(image, size=new_size)[
"pixel_values"
][0]
processed_images.append(image_pt)