transformers
:rotating_light: [`Attn`] New attn mask interface everywhere
#42848
Merged

:rotating_light: [`Attn`] New attn mask interface everywhere #42848

vasqu merged 61 commits into huggingface:main from vasqu:fix-fa-posids-tests
vasqu
vasqu fix
a6ba941e
vasqu fix order
c8d6984b
vasqu style
8095a1cb
HuggingFaceDocBuilderDev
vasqu vision 3d rope get extra test for now
fed05829
vasqu fix gpt2
8522356c
vasqu more gpt2 fixes
69be66dd
vasqu let's see...
c35529da
vasqu fix
e5f4fdbb
vasqu test
021732ca
vasqu
github-actions
vasqu fix opt+biogpt
d24b5300
vasqu fix
41a805cf
github-actions
vasqu fix
1cdd34d4
vasqu
github-actions
vasqu fix
7a69035f
vasqu fix opt
15ac7404
vasqu mask exchange test
c08292bf
vasqu style
5ecbb665
github-actions
vasqu vasqu changed the title [`FA`] Fix paddingfree tests to properly consider position ids and default create a mask [`Attn`] More new interface switches and proper paddingfree test 184 days ago
vasqu Merge branch 'main' into fix-fa-posids-tests
7ce86ec3
vasqu several small fixes
fc0b7169
vasqu
github-actions
vasqu shouldnt be needed
fe6f1e55
vasqu fix zamba models
0e5e70be
vasqu
vasqu retrigger ci
34fff722
vasqu
vasqu
vasqu force skip for now
676c8289
vasqu
github-actions
github-actions
github-actions
vasqu this wont work, will fix step by step
69103169
vasqu Merge branch 'main' into fix-fa-posids-tests
0ac29a23
vasqu to git
335fd8f0
vasqu vasqu changed the title [`Attn`] More new interface switches and proper paddingfree test :rotating_light: [`Attn`] New attn mask interface everywhere 179 days ago
vasqu another batch
423d423a
vasqu
github-actions
vasqu fix a few models, clip related models are gonna be hard...
cbc16445
vasqu another batch
28ad3c8d
vasqu Merge branch 'main' into fix-fa-posids-tests
3d713082
vasqu style
2ad3686c
vasqu fix gpt2 attempt
91bcb079
vasqu another batch + some models do not set their attn implementation? TODO
d35466a1
vasqu
github-actions
vasqu fix
a698fd88
github-actions
vasqu last models
eac4bedc
github-actions
vasqu Merge branch 'main' into fix-fa-posids-tests
8bdb76f6
vasqu style
ae6f2a65
vasqu repo fix
5cf843a0
vasqu Merge branch 'main' into fix-fa-posids-tests
2504d17c
vasqu check
ca653ed4
vasqu some quick fixes, error to catch wrong inits in some models
f60629d9
vasqu small fixes
03c81597
vasqu
github-actions
github-actions
vasqu fixes for wrong mask pretrained model relation
79883222
vasqu fix
5bea0a97
vasqu remove mask defaulting --> that's part of the prep + fixup some other…
6edbb5d9
vasqu
github-actions
vasqu small fixes
4bfa4169
github-actions
vasqu
github-actions
github-actions
vasqu fix last few models --> last to check recurrent gemma + repo consistency
f797e63f
github-actions
vasqu
github-actions
github-actions
vasqu fixup test cleanup
09c0b8b1
vasqu
github-actions
github-actions
github-actions
vasqu Merge branch 'main' into fix-fa-posids-tests
0606ad98
vasqu
github-actions
github-actions
github-actions
vasqu revert these tests
77a6ee92
vasqu these were not necessary, they have a proper top module
28f8a744
github-actions
vasqu fixup kwargs
e53bcdb7
vasqu Merge branch 'main' into fix-fa-posids-tests
a61a3dca
github-actions
vasqu remove old API
87e56b04
vasqu more kwargs
ca27bcc0
vasqu let's revert this - im in a fork :D
9a32c44f
github-actions
vasqu fix
1b602dc8
github-actions
vasqu dang
26c3aaa9
github-actions
vasqu
vasqu commented on 2026-02-06
vasqu vasqu requested a review from ArthurZucker ArthurZucker 128 days ago
vasqu vasqu requested a review from Cyrilvallez Cyrilvallez 128 days ago
vasqu vasqu marked this pull request as ready for review 128 days ago
ArthurZucker
ArthurZucker approved these changes on 2026-02-09
vasqu revert removal and add deprecation msg
0a670294
github-actions
vasqu kwargs typing
9f53f1c5
vasqu style
bef3e2be
github-actions
vasqu Merge branch 'main' into fix-fa-posids-tests
8116e632
github-actions
github-actions
vasqu vasqu merged 4b8ba25a into main 126 days ago
vasqu vasqu deleted the fix-fa-posids-tests branch 126 days ago

Login to write a write a comment.

Login via GitHub

Assignees
No one assigned
Labels
Milestone