transformers
Fix silent SDPA math-kernel fallback for GQA when key/value head_dim > 256 or differ
#46960
Open

Fix silent SDPA math-kernel fallback for GQA when key/value head_dim > 256 or differ #46960

Butterfingrz
vasqu
vasqu commented on 2026-06-29
Butterfingrz Fix silent SDPA math-kernel fallback for GQA with head_dim > 256
b41db1f8
Butterfingrz Butterfingrz force pushed from e5a98ff3 to b41db1f8 2 days ago
Butterfingrz Fix silent SDPA math-kernel fallback for GQA when key/value head_dim …
f572ef56
Butterfingrz
Butterfingrz Butterfingrz requested a review from vasqu vasqu 2 days ago
vasqu
vasqu commented on 2026-06-30
Butterfingrz Fix silent SDPA math-kernel fallback for GQA when key/value head_dim …
8a4bd6c8
Butterfingrz Butterfingrz requested a review from vasqu vasqu 1 day ago
Butterfingrz Butterfingrz changed the title Fix silent SDPA math-kernel fallback for GQA with head_dim > 256 Fix silent SDPA math-kernel fallback for GQA when key/value head_dim > 256 or differ 1 day ago
vasqu
vasqu approved these changes on 2026-07-01
vasqu
vasqu commented on 2026-07-01
Butterfingrz Test SDPA GQA stays off the math kernel
32390c44
Butterfingrz Butterfingrz requested a review from vasqu vasqu 1 day ago
vasqu
vasqu approved these changes on 2026-07-02
Butterfingrz Consolidate SDPA GQA tests into one forced-backend test
0d4332d4
Butterfingrz Butterfingrz requested a review from vasqu vasqu 3 hours ago
vasqu
vasqu commented on 2026-07-02
Butterfingrz Parametrize the SDPA GQA test by expected backend
b89dac94
Butterfingrz Butterfingrz requested a review from vasqu vasqu 2 hours ago
vasqu
vasqu approved these changes on 2026-07-02
vasqu Merge branch 'main' into fix/sdpa-gqa-large-head-dim-fallback
ab5ee8ff
vasqu vasqu enabled auto-merge 2 hours ago
HuggingFaceDocBuilderDev
Butterfingrz
vasqu
vasqu
Butterfingrz
vasqu Merge branch 'main' into fix/sdpa-gqa-large-head-dim-fallback
84563cac
disabled auto-merge 1 hour ago
Manually disabled by user
github-actions
vasqu vasqu enabled auto-merge 25 minutes ago
vasqu Merge branch 'main' into fix/sdpa-gqa-large-head-dim-fallback
56178db9

Login to write a write a comment.

Login via GitHub

Reviewers
Assignees
No one assigned
Labels
Milestone