transformers
[Bugfix] Fix flash-attention func param mismatch and softmax_scale default value mistake on Ascend NPU
#37575
Merged

[Bugfix] Fix flash-attention func param mismatch and softmax_scale default value mistake on Ascend NPU #37575

FightingZhen
github-actions github-actions marked this pull request as draft 1 year ago
github-actions
FightingZhen FightingZhen force pushed from dc618942 to 98b44731 1 year ago
FightingZhen FightingZhen force pushed from 98b44731 to 8bc2bfbe 1 year ago
FightingZhen FightingZhen marked this pull request as ready for review 1 year ago
github-actions github-actions requested a review from MekkCyber MekkCyber 1 year ago
github-actions github-actions requested a review from SunMarc SunMarc 1 year ago
FightingZhen
FightingZhen FightingZhen force pushed from 8bc2bfbe to b99b85b4 1 year ago
FightingZhen FightingZhen changed the title [Bugfix] Fix the parameter order mismatch between npu_flash_attn_varlen_func and flash_attn_varlen_func in flash-attn library [Bugfix] Fix flash-attention func param mismatch and softmax_scale default value mistake on Ascend NPU 1 year ago
MekkCyber
MekkCyber approved these changes on 2025-04-17
SunMarc
SunMarc approved these changes on 2025-04-17
FightingZhen [Bugfix] fix flash-attention func param mismatch and softmax_scale de…
2d7c8e15
FightingZhen FightingZhen force pushed from b99b85b4 to 2d7c8e15 1 year ago
MekkCyber Merge branch 'main' into bugfix-npu-fa
80c387cb
MekkCyber MekkCyber merged aa17cfb4 into main 1 year ago
FightingZhen FightingZhen deleted the bugfix-npu-fa branch 306 days ago

Login to write a write a comment.

Login via GitHub

Reviewers
Assignees
No one assigned
Labels
Milestone