the fix that did not get in #37370
debugging improvements
2227d491
add debugging details
5d0db09e
add more debugging details
a98c69aa
debug more
35f5bba7
the fix that did not get in
c7d1e731
ArthurZucker
marked this pull request as ready for review 255 days ago
MekkCyber
approved these changes
on 2025-04-08
First fix flex
c6f442ae
fix query offset
98196598
fix flex first
5c8c3cd1
fix device mask creation for speed
a2761e96
small mask creation sdpa
461402af
Update flex_attention.py
b9df8463
remove chunked prefill from HybridChunkedCache
8a95efaf
Merge branch 'flex-fix-regression' of github.com:huggingface/transfor…
1d82ac49
never seen such a fucked up merged
30111923
clean up layers + output
7e1d2209
add summary json file
19095482
Efficient general cache
eee77b31
Update cache_utils.py
f4927976
cleanup
a31018ae
Merge remote-tracking branch 'origin/model_debugger_upgrades' into fl…
56c607a3
fix?
030cc7c1
fix!
749c1540
oups typo
a7b60db2
Merge branch 'flex-fix-regression' of github.com:huggingface/transfor…
f9b9881b
not everywhere
02718f98
more fixes
583cd695
revert unrelated changes
48ae51a3
Fix but ugly for now -> should use pad instead
6b8b5b38
oups
9bc8481e
re-initialize the cache
31c75c83
Use pad to simplify
b1439714
style
82c4553c
correct slicing
0cbdca98
Cyrilvallez
deleted the flex-fix-regression branch 254 days ago
Assignees
No one assigned
Login to write a write a comment.
Login via GitHub