[video processors] decode only sampled videos -> less RAM and faster processing #39600
draft update two models for now
00964605
batch update all VLMs first
34a2ff1f
update some more image processors
4a8b1698
merge main
c7ec229a
update
0c1024ab
fix a few tests
01ad83ad
just make CI green for now
3187cc4a
fix copies
7900ce8f
update once more
47cce512
update
ea9a29be
merge main
4424f556
unskip the test
9957f9d0
fix these two
1c9ad582
fix torchcodec audio loading
595fe009
maybe
3f66126e
merge main
4c5a6743
yay, i fixed torchcodec installation and now can actually test it
ba02dec5
Merge remote-tracking branch 'upstream/main' into video-decoding
272054f7
fix copies deepseek
c05f31cf
make sure the metadata is returrned when users request it
0fe6e268
add docs
6286a8a1
update
1a787092
merge main
c9562f49
fixup
86ab24ae
Merge branch 'main' into video-decoding
f8f25063
qubvel
commented
on 2025-08-05
Update src/transformers/audio_utils.py
0997fdac
Update src/transformers/models/glm4v/video_processing_glm4v.py
40f6f3fd
update
ce9750da
what if we set some metadata attr to `None`
7375c02b
fix CI
66eed239
fix one test
194d7142
fix 4 channel test
adc8299f
fix glm timestemps
1b34bba0
Merge branch 'main' into video-decoding
5f5c9591
Merge branch 'main' into video-decoding
4b4f2972
rebase gone wrong
d5723de6
qubvel
approved these changes
on 2025-08-18
raise warning once
bc0405f4
Merge branch 'main' into video-decoding
de9f7fb8
fixup
baeb0e4b
typo
1fb826cd
fix copies
c0a1c627
ifx smolvlm test
248716c7
this is why torch's official benchmark was faster, set threads to `0`
ca0e8ae4
Merge branch 'main' into video-decoding
ceeab833
Merge branch 'main' into video-decoding
bf7e1d3f
Merge branch 'main' into video-decoding
cd4f073e
Merge branch 'main' into video-decoding
97e672f3
Apply style fixes
1fb502bf
Assignees
No one assigned
Login to write a write a comment.
Login via GitHub