-
Notifications
You must be signed in to change notification settings - Fork 14
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
Add hardware plugin blog post #41
base: main
Are you sure you want to change the base?
Conversation
Signed-off-by: MengqingCao <[email protected]>
@celestialli Thanks for your review! I'll update it in next commit |
Signed-off-by: MengqingCao <[email protected]>
Signed-off-by: MengqingCao <[email protected]>
@zeshengzong Thanks for your reivew, all your comments have been fixed in the latest commit. |
Signed-off-by: MengqingCao <[email protected]>
Signed-off-by: MengqingCao <[email protected]>
|
||
vLLM supports both dynamic registration of quantization algorithms and model. New backends can implement them on demand. | ||
|
||
**Registering a Custom Quantization Algorithm** |
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.
I think OOT quantization and models aren't that related to OOT platforms. It would be better to keep the blog focused on one topic.
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.
Considering that oot devices may register their own quantization algorithms and models, this part is also written as an option in this PR.
However, as you said, this is indeed relatively independent of the platform plugin topic. Thus I will delete step 4. Thank you for your helpful suggestions!
This blog post focuses on how the vLLM Hardware Plugin works and shares best practice for supporting Ascend NPU through the plugin mechanism.
cc @youkaichao @simon-mo @Yikun @wangxiyuan