Skip to content
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

[Doc] Update description of vLLM support for CPUs #6003

Open
wants to merge 2 commits into
base: main
Choose a base branch
from

Conversation

DamonFool
Copy link
Contributor

Hi all,

The description of CPU support in vLLM is out of date.
It would be better to update it.

Thanks.
Best regards,
Jie

@@ -20,7 +20,7 @@ Requirements

* OS: Linux
* Compiler: gcc/g++>=12.3.0 (optional, recommended)
* Instruction set architecture (ISA) requirement: AVX512 is required.
* Instruction set architecture (ISA) requirement: for x86, AVX2 is required; for PowerPC, Power9+ is required.
Copy link
Collaborator

Choose a reason for hiding this comment

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

Since the AVX2 and PowerPC backends don't have active testing and this documentation assumes you are building on a machine with AVX512, I'm a little hesitant to update in this doc. Maybe if you call out specifically where you can get directions for AVX2 and PowerPC, and that this doc is assuming AVX512, that would be more clear.

Copy link
Contributor Author

Choose a reason for hiding this comment

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

Thanks @mgoin for the review.

The origin description AVX512 is required is quite misleading, which means vLLM requires AVX512 ISA to run.
It would limit the usage of vLLM if people wrongly get the msg that vLLM only supports AVX512.

Actually, we've built and run vLLM on AVX2-only machines.
So I would suggest removing AVX512 is required here.

However, if you're against adding the PowerPC CPUs here, I'm fine.
This is because we don't have PowerPC CPUs and don't test it at all.

Thanks.

Copy link
Contributor Author

Choose a reason for hiding this comment

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

Since the AVX2 and PowerPC backends don't have active testing and this documentation assumes you are building on a machine with AVX512, I'm a little hesitant to update in this doc. Maybe if you call out specifically where you can get directions for AVX2 and PowerPC, and that this doc is assuming AVX512, that would be more clear.

Hi @mgoin , I kept AVX512 in the doc.
But make it to be (optional, recommended) as the description of the compiler.
What do you think?
Thanks.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
Projects
None yet
Development

Successfully merging this pull request may close these issues.

None yet

3 participants