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

Add support for fused update_cache #15202

Open
wants to merge 9 commits into
base: main
Choose a base branch
from
Open

Conversation

kpaigwar
Copy link
Contributor

@kpaigwar kpaigwar commented Nov 19, 2024

Problem description

For perf optimization, we need a new fused Op to update cache in parallel for two input tensors sharded on different core_grid

What's changed

  1. Added program factory for new Op-type FUSED_UPDATE
  2. Refactored validation for paged_cache_operations
  3. Added pytest for new Op
  4. fixed df reconfiguration for update_cache compute_kernel

Notes

  1. The Op is currently only functional for BFP16 inputs anf BFP16 caches. Will add support for BFP8_b caches in next PR as we fix the bug.

Checklist

@kpaigwar kpaigwar changed the title Kpaigwar/fuse update cache Support fused update cache Nov 19, 2024
@kpaigwar kpaigwar self-assigned this Nov 19, 2024
@kpaigwar kpaigwar changed the title Support fused update cache Add support for fused update_cache Nov 19, 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