fix: rejection sampling acceptance rate in MRv2#1658
Conversation
Signed-off-by: AlpinDale <alpindale@gmail.com>
![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: 508bd1a210
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".
| NgramGPUTypes, | ||
| ] | ||
| RejectionSampleMethod = Literal["strict", "probabilistic", "synthetic"] | ||
| RejectionSampleMethod = Literal["standard", "synthetic"] |
There was a problem hiding this comment.
This change removes "strict" and "probabilistic" from RejectionSampleMethod, so any existing deployment or config file that still sets either value will now fail Pydantic validation at startup before serving begins. Because this commit is framed as an acceptance-rate fix (not an API removal), dropping the old enum values without aliasing/migration introduces an avoidable breaking regression for users upgrading in place.
Useful? React with 👍 / 👎.
1 participant
vllm-project/vllm#40651