transformers
Add an opt-in per-frame pixel cap (cap_pixels_per_frame) to the Qwen3-VL video processor
#48071
Merged
Merged
Add an opt-in per-frame pixel cap (cap_pixels_per_frame) to the Qwen3-VL video processor #48071
zucchini-nlp
merged
13 commits into
huggingface:main
from
dkrisman:qwen3vl-video-max-pixels-per-frame
Add max_pixels_per_frame to Qwen3-VL video processor
3ccaebc4
Floor the frame count at 32 when applying max_pixels_per_frame
3eb46a79
Regenerate CohereCompass video processor from modular source
c1e6f7db
zucchini-nlp
commented
on 2026-08-19
Replace max_pixels_per_frame with a cap_pixels_per_frame boolean
e9d3c340
dkrisman
changed the title
Apply ruff formatting
215f91b0
zucchini-nlp
commented
on 2026-08-24
Spell Qwen3VL in the deprecation warning so modular conversion rename…
04e7e5e6
Add cap_pixels_per_frame to the Qwen2VL video processor
e9ffca31
Regenerate VideoLlama3 video processor from modular source
51ede15c
Merge branch 'main' into qwen3vl-video-max-pixels-per-frame
7122ac5c
Address review: parameterize caps as class attrs, inherit Qwen2VL _pr…
66ce05b9
zucchini-nlp
commented
on 2026-08-25
Override get_num_of_video_patches for Qwen3VL's temporal smart_resize
39ffc0fe
i just merged autodoc fix on main, resolve conflicts
d97d09cb
small unification in naming
8c99878d
zucchini-nlp
approved these changes
on 2026-08-26
Login to write a write a comment.
Login via GitHub