Perf(PP): support PP with async send/recv.#7143
Perf(PP): support PP with async send/recv.#7143MengqingCao merged 3 commits intovllm-project:mainfrom
Conversation
|
👋 Hi! Thank you for contributing to the vLLM Ascend project. The following points will speed up your PR merge:
If CI fails, you can run linting and testing checks locally according Contributing and Testing. |
Summary of ChangesHello, I'm Gemini Code Assist1! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed! This pull request enhances the Highlights
🧠 New Feature in Public Preview: You can now enable Memory to help Gemini Code Assist learn from your team's feedback. This makes future code reviews more consistent and personalized to your project's style. Click here to enable Memory in your admin console. Changelog
Activity
Using Gemini Code AssistThe full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips. Invoking Gemini You can request assistance from Gemini at any point by creating a comment using either
Customization To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a Limitations & Feedback Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments. If you're interested in giving your feedback about your experience with Gemini Code Assist for GitHub and other Google products, sign up here. You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension. Footnotes
|
![gemini-code-assist[bot]](https://avatars.githubusercontent.com/in/956858?s=60&v=4){kind=small}
There was a problem hiding this comment.
Code Review
This pull request introduces asynchronous send/receive operations for pipeline parallelism to improve performance. The implementation uses a new AsyncIntermediateTensors class for lazy synchronization of received tensors. The overall approach is sound. However, the use of __getattribute__ for lazy synchronization in AsyncIntermediateTensors can cause issues with torch.compile, potentially leading to graph breaks and performance degradation in compiled mode. I've suggested a more explicit, compiler-friendly approach.
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: b4be6a996f
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
pisceskkk
force-pushed
the
pp/async_comm
branch
3 times, most recently
from
b38fb53 to
530e817
Compare
pisceskkk
force-pushed
the
pp/async_comm
branch
from
530e817 to
4c5c363
Compare
Signed-off-by: QiuChunshuo <qiuchunshuo@huawei.com>
MengqingCao
added
ready
|
plz rebase your code after #7230 merged |
### What this PR does / why we need it? Follow up the PR vllm-project/vllm#33368, this PR provides async send/recv support for PP in vllm-ascend. --- - vLLM version: v0.17.0 - vLLM main: vllm-project/vllm@4034c3d Signed-off-by: QiuChunshuo <qiuchunshuo@huawei.com>
5 participants
What this PR does / why we need it?
Follow up the PR vllm-project/vllm#33368, this PR provides async send/recv support for PP in vllm-ascend.
How was this patch tested?
Launch server:
Acc
Perf
We use first 192 requests of gsm8k with
output_len=512andconcurrency=32to test the performance.TP4PP4 w/o async_comm
This PR
TP4PP4 w/ async_comm
Summary
PP w/ async_comm v.s. PP w/o async_comm