Skip to content

Conversation

@DongheJin
Copy link
Collaborator

No description provided.

Copy link

@gemini-code-assist gemini-code-assist bot left a comment

Choose a reason for hiding this comment

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

Code Review

This pull request aims to fix an issue with DeepSeek models when data parallelism (DP) is enabled. The changes involve mapping CHUNKED_PREFILL to PREFILL forward type in the engine and updating a log message for clarity. While the fix for forward type mapping is in the right direction, the logic for determining the unified forward type across DP ranks is not robust and could lead to inconsistencies. I've provided a critical suggestion to add a consistency check to prevent potential crashes or silent errors in a distributed environment.

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