Skip to content

Conversation

@adobrzyn
Copy link
Collaborator

No description provided.

Signed-off-by: Agata Dobrzyniewicz <[email protected]>
Copy link
Contributor

Copilot AI left a comment

Choose a reason for hiding this comment

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

Pull request overview

This PR implements a workaround to address out-of-memory issues related to huge shared bias in the unified bucketing strategy by limiting the maximum shared context to a single block size instead of calculating it based on model length and available blocks.

  • Temporarily caps max_shared_ctx to block_size to prevent OOM errors
  • Comments out the original calculation logic for future restoration
  • Adds a TODO comment indicating this is a temporary workaround

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Signed-off-by: Agata Dobrzyniewicz <[email protected]>
@github-actions
Copy link

github-actions bot commented Jan 2, 2026

🚧 CI Blocked

The main CI workflow was not started for the following reason:

Your branch is behind the base branch. Please merge or rebase to get the latest changes.

@github-actions
Copy link

github-actions bot commented Jan 2, 2026

✅ CI Passed

All checks passed successfully against the following vllm commit:
b3a2bdf1ac90748d58bf8c05f8d0095ede5c7eca

@adobrzyn adobrzyn merged commit 19be0a2 into main Jan 8, 2026
50 checks passed
@adobrzyn adobrzyn deleted the adobrzyn/ua_shared_wa branch January 9, 2026 10:00
jinyouzhi pushed a commit to jinyouzhi/vllm-gaudi that referenced this pull request Jan 14, 2026
Signed-off-by: Agata Dobrzyniewicz <[email protected]>
Signed-off-by: Agata Dobrzyniewicz <[email protected]>
Signed-off-by: Jin, Youzhi <[email protected]>
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