Skip to content

fix: update ray executor import for vLLM 0.20#2609

Merged
terrykong merged 1 commit into
mainfrom
achartier/fix-fp8-vllm-0.20-ray-executor-import
Jun 2, 2026
Merged

fix: update ray executor import for vLLM 0.20#2609
terrykong merged 1 commit into
mainfrom
achartier/fix-fp8-vllm-0.20-ray-executor-import

Conversation

@achartier

@achartier achartier commented May 28, 2026

Copy link
Copy Markdown
Contributor

What does this PR do ?

fix: update ray executor import for vLLM 0.20

Issues

vLLM 0.20 renamed the executor module from ray_distributed_executor to ray_executor; the class name RayDistributedExecutor and its collective_rpc method are unchanged. Without this fix, any FP8 generation path that goes through monkey_patch_vllm_ray_executor() with TP*PP > 1 (i.e. any multi-GPU vllm_cfg.precision='fp8' run) fails at worker init with:

ModuleNotFoundError: No module named 'vllm.v1.executor.ray_distributed_executor'

Usage

Verified by running GLM-5.1 64n8g with vllm_cfg.precision='fp8' on sl/glm5.1 before the fix (crashes immediately) and after (5/5 training steps complete cleanly).

# Add a code snippet demonstrating how to use this

Before your PR is "Ready for review"

Pre checks:

  • Make sure you read and followed Contributor guidelines
  • Did you write any new necessary tests?
  • Did you run the unit tests and functional tests locally? Visit our Testing Guide for how to run tests
  • Did you add or update any necessary documentation? Visit our Document Development Guide for how to write, build and test the docs.

Additional Information

  • ...

@achartier achartier requested a review from a team as a code owner May 28, 2026 20:09
@copy-pr-bot

copy-pr-bot Bot commented May 28, 2026

Copy link
Copy Markdown

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

@terrykong terrykong added the CI:Lfast Runs a fast test suite and re-use nightly `main` container (but sync dependencies to PRs version) label May 28, 2026
@terrykong

Copy link
Copy Markdown
Collaborator

/ok to test 679d8f4

@achartier

Copy link
Copy Markdown
Contributor Author

/ok to test c070819

@achartier

Copy link
Copy Markdown
Contributor Author

Blocked on ci infra, opened fix: #2630

vLLM 0.20 renamed the executor module from ray_distributed_executor to
ray_executor; the class name RayDistributedExecutor and its
collective_rpc method are unchanged. Without this fix, any FP8
generation path that goes through monkey_patch_vllm_ray_executor() with
TP*PP > 1 (i.e. any multi-GPU vllm_cfg.precision='fp8' run) fails at
worker init with:

  ModuleNotFoundError: No module named 'vllm.v1.executor.ray_distributed_executor'

Signed-off-by: Aurelien Chartier <2567591+achartier@users.noreply.github.com>
@achartier achartier force-pushed the achartier/fix-fp8-vllm-0.20-ray-executor-import branch from c070819 to 722631b Compare May 29, 2026 18:14
@achartier

Copy link
Copy Markdown
Contributor Author

/ok to test 722631b

@achartier

Copy link
Copy Markdown
Contributor Author

@terrykong I think this is ready to merge

@terrykong terrykong merged commit 19b1a2a into main Jun 2, 2026
76 checks passed
@terrykong terrykong deleted the achartier/fix-fp8-vllm-0.20-ray-executor-import branch June 2, 2026 20:48
pengdurice pushed a commit to pengdurice/RL that referenced this pull request Jun 12, 2026
Signed-off-by: Aurelien Chartier <2567591+achartier@users.noreply.github.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

CI:Lfast Runs a fast test suite and re-use nightly `main` container (but sync dependencies to PRs version)

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants