Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

WIP: FlashAttention for WebGPU EP #22919

Open
wants to merge 7 commits into
base: main
Choose a base branch
from

Conversation

sushraja-msft
Copy link

@sushraja-msft sushraja-msft commented Nov 21, 2024

WIP: Implementation of FlashAttention that works for MHA

  • Currently only works on machines where the subgroup size is the same as tile size. (Intel)
  • Works only for the condition of new sequence length is 1.

The other scenarios require more debugging, algorithm needs optimization as well for the 1 seq length case because workgroups are left unused in how ComputeDotProduct is invoked.

@sushraja-msft sushraja-msft changed the title User/sushraja/fa attempt2 WIP: FlashAttention for WebGPU EP Nov 21, 2024
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.

1 participant