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.

Sign up for GitHub

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

Jump to bottom

[V1] Enable custom ops with piecewise CUDA graphs #10228

Merged
merged 1 commit into from
Nov 11, 2024
Merged

[V1] Enable custom ops with piecewise CUDA graphs #10228

merged 1 commit into from
Nov 11, 2024

Conversation

WoosukKwon
Copy link
Collaborator

A correct version of #10227

@WoosukKwon
[V1] Enable custom ops with piecewise CUDA graphs
68a8aea 
Signed-off-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>
@github-actions GitHub Actions
Copy link

👋 Hi! Thank you for contributing to the vLLM project.
Just a reminder: PRs would not trigger full CI run by default. Instead, it would only run fastcheck CI which starts running only a small and essential subset of CI tests to quickly catch errors. You can run other CI tests on top of those by going to your fastcheck build on Buildkite UI (linked in the PR checks section) and unblock them. If you do not have permission to unblock, ping simon-mo or khluu to add you in our Buildkite org.

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:

  • Add ready label to the PR
  • Enable auto-merge.

🚀

youkaichao
youkaichao reviewed Nov 11, 2024
View reviewed changes
vllm/v1/worker/gpu_model_runner.py
@@ -405,6 +406,7 @@ def load_model(self) -> None:
if self.use_cuda_graph:
# FIXME(woosuk): Currently, we do not use inductor to reduce the
# compilation time and any potential issues with the inductor.
os.environ["VLLM_CUSTOM_OPS"] = "all"
Copy link
Member

Choose a reason for hiding this comment

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

@ProExpertProg we should improve the default option, basing on the value of use_inductor .

youkaichao
youkaichao approved these changes Nov 11, 2024
View reviewed changes
Copy link
Member

@youkaichao youkaichao left a comment

Choose a reason for hiding this comment

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

thanks for figuring it out!

@WoosukKwon WoosukKwon merged commit 9d5b4e4 into main Nov 11, 2024
13 of 16 checks passed
@WoosukKwon WoosukKwon deleted the v1-use-custom-op branch November 11, 2024 19:58
@WoosukKwon
Copy link
Collaborator Author

Actually this PR broke CUDA graphs :( Needs to be fixed.

@WoosukKwon WoosukKwon mentioned this pull request Nov 12, 2024
Merged
rickyyx pushed a commit to rickyyx/vllm that referenced this pull request Nov 13, 2024
@WoosukKwon @rickyyx
[V1] Enable custom ops with piecewise CUDA graphs (vllm-project#10228)
7a772e7 
Signed-off-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>
sumitd2 pushed a commit to sumitd2/vllm that referenced this pull request Nov 14, 2024
@WoosukKwon @sumitd2
[V1] Enable custom ops with piecewise CUDA graphs (vllm-project#10228)
bf7ec1c 
Signed-off-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>
Signed-off-by: Sumit Dubey <sumit.dubey2@ibm.com>
KuntaiDu pushed a commit to KuntaiDu/vllm that referenced this pull request Nov 20, 2024
@WoosukKwon
[V1] Enable custom ops with piecewise CUDA graphs (vllm-project#10228)
5c7e4b4 
Signed-off-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>
mfournioux pushed a commit to mfournioux/vllm that referenced this pull request Nov 20, 2024
@WoosukKwon @mfournioux
[V1] Enable custom ops with piecewise CUDA graphs (vllm-project#10228)
f307b03 
Signed-off-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>
Signed-off-by: Maxime Fournioux <55544262+mfournioux@users.noreply.github.com>
tlrmchlsmth pushed a commit to neuralmagic/vllm that referenced this pull request Nov 23, 2024
@WoosukKwon @tlrmchlsmth
[V1] Enable custom ops with piecewise CUDA graphs (vllm-project#10228)
b500ed1 
Signed-off-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>
Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com>
sleepwalker2017 pushed a commit to sleepwalker2017/vllm that referenced this pull request Dec 13, 2024
@WoosukKwon
[V1] Enable custom ops with piecewise CUDA graphs (vllm-project#10228)
c4abbf4 
Signed-off-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@youkaichao youkaichao youkaichao approved these changes

Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
2 participants
@WoosukKwon @youkaichao