Skip to content

flash-attn4: add Flash Attention 4#463

Merged
danieldk merged 17 commits intomainfrom
flash-attn4
Mar 20, 2026
Merged

flash-attn4: add Flash Attention 4#463
danieldk merged 17 commits intomainfrom
flash-attn4

Conversation

@danieldk
Copy link
Member

@danieldk danieldk commented Mar 16, 2026

This change adds Flash Attention 4.

Warning: no point in testing/using this yet, the kernelization is not completely done yet and it won't work. Just using this for CI now.

Todo:

  • Resync to latest release (and not HEAD).
  • Run tests on Blackwell.
  • Do validation with HF libraries.

@danieldk danieldk requested a review from drbh as a code owner March 16, 2026 19:33
drbh
drbh previously approved these changes Mar 20, 2026
Copy link
Collaborator

@drbh drbh left a comment

Choose a reason for hiding this comment

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

looks good to me! exciting addition!

@danieldk danieldk merged commit 747528a into main Mar 20, 2026
6 of 10 checks passed
@danieldk danieldk deleted the flash-attn4 branch March 20, 2026 16:44
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.

2 participants