Skip to content

LLaVA OV: fix unpadding precision #34779

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
Nov 20, 2024
Merged

LLaVA OV: fix unpadding precision #34779

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
3f4dbcd
fix
zucchini-nlp Nov 18, 2024
f5313ef
propagate
zucchini-nlp Nov 18, 2024
496e56c
type check
zucchini-nlp Nov 18, 2024
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
4 changes: 3 additions & 1 deletion src/transformers/models/llava_next/processing_llava_next.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -152,7 +152,9 @@ def __call__(
for sample in text:
while self.image_token in sample:
image_size = next(image_sizes)
orig_height, orig_width = image_size
if not isinstance(image_size, (list, tuple)):
# cast to list to avoid numerical precision errors when calculating unpadding
orig_height, orig_width = image_size.tolist()
num_image_tokens = self._get_number_of_features(orig_height, orig_width, height, width)
if self.vision_feature_select_strategy == "default":
num_image_tokens -= 1
Expand Down
4 changes: 3 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 @@ -177,7 +177,9 @@ def __call__(
for sample in text:
while self.image_token in sample:
image_size = next(image_sizes)
orig_height, orig_width = image_size
if not isinstance(image_size, (list, tuple)):
# cast to list to avoid numerical precision errors when calculating unpadding
orig_height, orig_width = image_size.tolist()
num_image_tokens = self._get_number_of_features(orig_height, orig_width, height, width)
if self.vision_feature_select_strategy == "default":
num_image_tokens -= 1
Expand Down
5 changes: 4 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 @@ -188,7 +188,10 @@ def _expand_image_tokens(
for sample in text:
while special_token in sample:
image_size_list = next(image_sizes)
orig_height, orig_width = image_size_list[0] if num_frames != 1 else image_size_list
original_size = image_size_list[0] if num_frames != 1 else image_size_list
if not isinstance(original_size, (list, tuple)):
# cast to list to avoid numerical precision errors when calculating unpadding
orig_height, orig_width = original_size.tolist()
num_image_tokens = self._get_number_of_features(orig_height, orig_width, height, width)
if self.vision_feature_select_strategy == "default":
num_image_tokens -= 1
Expand Down