Skip to content

Fix "test_chat_template_dict" in video LLMs #35660

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
zucchini-nlp merged 3 commits into from
Jan 21, 2025
Merged

Fix "test_chat_template_dict" in video LLMs #35660

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
4b1d7c6
fix "test_chat_template_dict" in llava_onevision
zucchini-nlp Jan 13, 2025
64e7194
Update src/transformers/models/llava_next_video/processing_llava_next…
zucchini-nlp Jan 17, 2025
d2759a7
get one video calles once
zucchini-nlp Jan 17, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
8 changes: 7 additions & 1 deletion src/transformers/models/llava_next_video/processing_llava_next_video.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -18,6 +18,8 @@

from typing import TYPE_CHECKING, List, Optional, Union

import numpy as np

from ...feature_extraction_utils import BatchFeature
from ...image_processing_utils import select_best_resolution
from ...image_utils import ImageInput, VideoInput, get_image_size, to_numpy_array
Expand Down Expand Up @@ -193,7 +195,11 @@ def __call__(

# videos are easier, simply get frames and multiply
if videos_inputs:
one_video = to_numpy_array(videos_inputs.get("pixel_values_videos")[0])
one_video = videos_inputs.get("pixel_values_videos")[0]
if isinstance(one_video, (list, tuple)):
one_video = np.array(one_video)
else:
one_video = to_numpy_array(one_video)
height, width = get_image_size(one_video[0])
num_frames = one_video.shape[0] # frame dim is always after batch dim

Expand Down
8 changes: 7 additions & 1 deletion src/transformers/models/llava_onevision/processing_llava_onevision.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -20,6 +20,8 @@
import os
from typing import Iterable, List, Union

import numpy as np

from ...feature_extraction_utils import BatchFeature
from ...image_processing_utils import select_best_resolution
from ...image_utils import ImageInput, VideoInput, get_image_size, to_numpy_array
Expand Down Expand Up @@ -164,7 +166,11 @@ def __call__(
if videos is not None:
video_inputs = self.video_processor(videos, **output_kwargs["videos_kwargs"])

one_video = to_numpy_array(video_inputs.get("pixel_values_videos")[0])
one_video = video_inputs.get("pixel_values_videos")[0]
if isinstance(video_inputs.get("pixel_values_videos")[0], (list, tuple)):
one_video = np.array(one_video)
else:
one_video = to_numpy_array(one_video)
height, width = get_image_size(one_video[0], channel_dim=output_kwargs["images_kwargs"].get("data_format"))
num_frames = one_video.shape[0] # frame dim is always after batch dim
patches_height_width = int(math.sqrt(self.num_image_tokens))
Expand Down
8 changes: 7 additions & 1 deletion src/transformers/models/video_llava/processing_video_llava.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -18,6 +18,8 @@

from typing import List, Optional, Union

import numpy as np

from ...feature_extraction_utils import BatchFeature
from ...image_utils import ImageInput, get_image_size, to_numpy_array
from ...processing_utils import ProcessorMixin
Expand Down Expand Up @@ -165,7 +167,11 @@ def __call__(
num_frames = 1

if "pixel_values_videos" in encoded_images.keys():
one_video = to_numpy_array(encoded_images.get("pixel_values_videos")[0])
one_video = encoded_images.get("pixel_values_videos")[0]
if isinstance(encoded_images.get("pixel_values_videos")[0], (list, tuple)):
one_video = np.array(one_video)
else:
one_video = to_numpy_array(one_video)
height, width = get_image_size(one_video[0])
num_frames = one_video.shape[0] # frame dim is always after batch dim

Expand Down