-
-
Notifications
You must be signed in to change notification settings - Fork 5.3k
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
[Misc] Support register quantization method out-of-tree #11969
[Misc] Support register quantization method out-of-tree #11969
Conversation
👋 Hi! Thank you for contributing to the vLLM project. Once the PR is approved and ready to go, your PR reviewer(s) can run CI to test the changes comprehensively before merging. To run CI, PR reviewers can do one of these:
🚀 |
@DarkLight1337 @mgoin Hey there! Just a quick reminder about the pull request I submitted a few days ago. We’re looking forward to vllm supporting the registration of quantization methods from external sources, which will help us develop quantization tools for evaluating fake-quantization models. I’d really appreciate your feedback on this PR whenever you have a moment! Thank you! |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This would be a great feature! Could you add a test for adding a simple quantization method? It could be a trivial quant-dequant pass
Hi @mgoin, thank you very much for taking the time to review this PR! I've added a test as suggested, which is used to register a custom per-token dynamic fake-quant method. If you have any questions or suggestions for further modifications, please feel free to let me know. |
8bf911c
to
56b129b
Compare
This pull request has merge conflicts that must be resolved before it can be |
Signed-off-by: ice-tong <[email protected]>
Signed-off-by: ice-tong <[email protected]>
83290eb
to
15ebfd0
Compare
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Excellent work, I appreciate it!
…#11969) Signed-off-by: Bowen Wang <[email protected]>
Motivation