[skyrl-train] Support older vllm versions till 0.9.2#671
Merged
SumanthRH merged 4 commits intoNovaSky-AI:mainfrom Nov 17, 2025
Merged
[skyrl-train] Support older vllm versions till 0.9.2#671SumanthRH merged 4 commits intoNovaSky-AI:mainfrom
SumanthRH merged 4 commits intoNovaSky-AI:mainfrom
Conversation
SumanthRH
changed the title
SumanthRH
commented
Nov 15, 2025
|
|
||
| In some cases, you may encounter "illegal memory access" errors with vLLM >= 0.10.0: https://github.com/vllm-project/vllm/issues/23814. Currently, we recommend a workaround by downgrading to vLLM 0.9.2. | ||
|
|
||
| With SkyRL, this can be done with the following overrides: |
Member
Author
There was a problem hiding this comment.
A todo here is to mention the exact commit from which such override is supported., Will be done after merging the PR
Member
Author
|
/gemini review |
![gemini-code-assist[bot]](https://avatars.githubusercontent.com/in/956858?s=60&v=4){kind=small}
Contributor
There was a problem hiding this comment.
Code Review
This pull request adds support for older versions of vLLM, specifically down to version 0.9.2. The changes introduce version-aware logic to handle API differences, update the CI pipeline to include tests for the older vLLM version, and add relevant troubleshooting documentation. My review focuses on improving the maintainability of the CI script, correcting a broken link in the documentation, and suggesting refactoring to reduce code duplication in the Python source.
Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>
CharlieFRuan
approved these changes
Nov 17, 2025
Collaborator
CharlieFRuan
left a comment
CharlieFRuan
left a comment
There was a problem hiding this comment.
LGTM! This is quite useful (e.g. in clusters where you need to build things from source, you might prefer to run older versions). Left two comments just for my own sake of understanding
li-boxuan
pushed a commit
to li-boxuan/SkyRL
that referenced
this pull request
Nov 23, 2025
Adds support for older vllm versions. With these changes, I am able to use older vllm versions as follows: ```bash uv run --isolated --extra vllm --extra dev --with vllm==0.9.2 --with transformers==4.53.0 --with torch==2.7.0 --with "flash-attn@https://github.com/Dao-AILab/flash-attention/releases/download/v2.8.3/flash_attn-2.8.3+cu12torch2.7cxx11abiTRUE-cp312-cp312-linux_x86_64.whl" -- pytest -s -vvv tests/gpu/gpu_ci/test_engine_generation.py::test_token_based_generation -m "vllm" ``` Will add docs for this in a future PR --------- Signed-off-by: SumanthRH <sumanthrh99@gmail.com> Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>
dzorlu
pushed a commit
to fleet-ai/SkyRL
that referenced
this pull request
Feb 4, 2026
# What does this PR do? Adds support for older vllm versions. With these changes, I am able to use older vllm versions as follows: ```bash uv run --isolated --extra vllm --extra dev --with vllm==0.9.2 --with transformers==4.53.0 --with torch==2.7.0 --with "flash-attn@https://github.com/Dao-AILab/flash-attention/releases/download/v2.8.3/flash_attn-2.8.3+cu12torch2.7cxx11abiTRUE-cp312-cp312-linux_x86_64.whl" -- pytest -s -vvv tests/gpu/gpu_ci/test_engine_generation.py::test_token_based_generation -m "vllm" ``` Will add docs for this in a future PR --------- Signed-off-by: SumanthRH <sumanthrh99@gmail.com> Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What does this PR do?
Adds support for older vllm versions.
With these changes, I am able to use older vllm versions as follows:
Will add docs for this in a future PR