[BugFix][V1] Quick fix for min_tokens with multiple EOS

[njhill]

Reported by @andylolu2 in slack. This can affect min_tokens functionality for models that define more than one EOS token in their config.

We need to revisit how SamplingParams.__post_init__ is used generally, as well as test coverage for this specific case. This is a quick fix in the meantime for a point release planned for today.

• Essentially the eos_token_id added by SamplingParams.update_from_generation_config disappears when it's sent through the msgpack encoder, because the __post_init__ overwrites self._all_stop_token_ids = set(self.stop_token_ids). So the worker sees an empty all_stop_token_ids, hence effectively no min_tokens.

[github-actions]

👋 Hi! Thank you for contributing to the vLLM project.

💬 Join our developer Slack at https://slack.vllm.ai to discuss your PR in #pr-reviews, coordinate on features in #feat- channels, or join special interest groups in #sig- channels.

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 either: Add ready label to the PR or enable auto-merge.

🚀