Skip to content

Update mha_varlan_fwd and the kernel config heuristics. #795

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

Merged
merged 2 commits into from
Jul 18, 2025
Merged

Conversation

tongxin
Copy link
Contributor

@tongxin tongxin commented Jul 17, 2025

PR Category

Operator

Type of Change

Refactor

Description

This PR introduces several changes to improve compatibility.

  • updates flash_attn_varlen_func's argument list to align with latest vllm
  • introduces two config heuristics in flash_varlen to cope better with likely prefill and decoding phases
  • no longer uses block_ptr to load kv blocks so that kv tile size can be decoupled with block size

Issue

Progress

  • Change is properly reviewed (1 reviewer required, 2 recommended).
  • Change is responded to an issue.
  • Change is fully covered by a UT.

Performance

Copy link
Collaborator

@0x45f 0x45f left a comment

Choose a reason for hiding this comment

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

LGTM

@tongxin tongxin merged commit cd1bd19 into master Jul 18, 2025
12 of 14 checks passed
@tongxin tongxin deleted the flash-update branch July 18, 2025 06:44
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.

3 participants