Open
Conversation
Contributor
|
Warning Gemini encountered an error creating the summary. You can try again by commenting |
Contributor
|
Hi, it is a nice work. However, it seems that there are too many diff in the git commit. Could you please fix it to help us understand your code? I think that in order to support Qwen-235B-w4afp8 in TP mode, we should special handling "weight interleave scales" which should not be 4, and modify the sgl-kernel. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Motivation
Follow #8118. Base on #7762.
Modifications
This PR primarily implements the adaptation of SGLang for Qwen3-MoE-w4afp8 quantized models. The key enhancements include:
Support for Qwen3-MoE’s w4afp8-block quantization format: SGLang can now load and run models that have been quantized using the w4afp8-block format.
Support for loading Qwen-MoE static quantization calibration parameters: SGLang is now capable of loading and utilizing the static quantization calibration parameters for Qwen-MoE models, ensuring correct inference behavior after quantization.
Accuracy Tests
You can download the Qwen3-30B-A3B-w4afp8-block model here.
https://huggingface.co/zhilingjiang/Qwen3-30B-A3B-w4afp8-block-dynamic
Checklist