Merged
Conversation
Merged
|
The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update. |
edbeeching
approved these changes
Sep 23, 2025
albertvillanova
approved these changes
Sep 23, 2025
Member
albertvillanova
left a comment
There was a problem hiding this comment.
Good decision. I remember doing something similar for fsspec dependency in datasets: this approach saves us from having to release a hotfix after one of their breaking releases.
qgallouedec
added a commit
that referenced
this pull request
Sep 23, 2025
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
vLLM has proven to be unstable, frequently breaking TRL even with patch releases. The upcoming vLLM 0.11 (not released at the time of writing) is expected to break
GuidedDecodingParams(see vllm-project/vllm#22772) without prior notice or deprecation. It will also include "many other" brekaking changes, quote a vLLM maintainer.To mitigate this and ensure stability for TRL users, we propose a new approach for managing the vLLM dependency specifically: explicitly specify the supported versions and assume that any newer release—even a patch—may break the codebase. The list of supported versions will only be extended once the community has validated that the new version does not break, or introduce regressions.