Skip to content

[lora] enable bias support for fused moe lora#1583

Merged
AlpinDale merged 1 commit into
mainfrom
fused-lora-bias
Nov 4, 2025
Merged

[lora] enable bias support for fused moe lora#1583
AlpinDale merged 1 commit into
mainfrom
fused-lora-bias

Conversation

@AlpinDale

Copy link
Copy Markdown
Collaborator

No description provided.

Signed-off-by: AlpinDale <alpindale@gmail.com>
@AlpinDale AlpinDale merged commit d07b458 into main Nov 4, 2025
1 check passed
@AlpinDale AlpinDale deleted the fused-lora-bias branch November 4, 2025 07:04

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

Copy link
Copy Markdown
Contributor

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 enables bias support for fused MoE with LoRA and corrects the handling of tensor parallelism. The changes involve refactoring how quantization configuration is handled to include bias information, and removing a check that previously prevented bias usage. Additionally, the code is updated to use the number of local experts instead of global experts when creating and applying LoRA weights, which is a critical fix for distributed tensor-parallel execution. The changes are consistent and improve correctness and functionality.

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