Skip to content

Conversation

@wuxun-zhang
Copy link

Description

Fix output shape of FMHA forward kernel, otherwise it will give wrong results when is_var_len=True and num_heads_kv=1.

Type

  • Bug - [ ] Feature - [ ] Performance - [ ] Refactor

Testing

  • Tests pass - [ ] Xe12 - [x] Xe20

Performance

Metric Before After

References

Fixes #

Checklist

  • Copyright - [ ] Co-pilot Review - [ ] Deprecated APIs not used

@wuxun-zhang
Copy link
Author

@ClarkChin08 Please help take a look here.

@wuxun-zhang
Copy link
Author

@rolandschulz Please also review here.

Copy link

@jiyang1011 jiyang1011 left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTMTM

@wuxun-zhang
Copy link
Author

@rolandschulz Could you please help review and trigger CI here?

@wuxun-zhang
Copy link
Author

Umm, CI failures are not related to changes...

@tdeng5 tdeng5 merged commit 2c7282d into intel:main Dec 31, 2025
6 of 13 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants