[Core] Make disaggregated prefill compatible with pipeline parallelism

[YuhanLiu11]

This PR is part of the [RFC]: Disaggregated prefilling and KV cache transfer roadmap, which aims to make disaggregated prefill compatible with pipeline parallelism.

In pipeline parallelism, the intermediate hidden states transferred across different stages are stored in dictionaries. However, the current KV and intermediate transfer logic between prefill and decode instances only assumes these intermediate states are tensors.

This PR checks whether the hidden states transferred across P/D instances are dictionaries and, if so, sends the fields in these dictionaries separately.

Note: This current PR only makes disaggregated prefill compatible with pipeline parallelism in 1P1D setting (the pp degree in prefiller and decoder instances should be the same).

[github-actions]

👋 Hi! Thank you for contributing to the vLLM project.
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 do one of these:

• Add ready label to the PR
• Enable auto-merge.

🚀

[github-actions]

This pull request has been automatically marked as stale because it has not had any activity within 90 days. It will be automatically closed if no further activity occurs within 30 days. Leave a comment if you feel this pull request should remain open. Thank you!

[mergify]

This pull request has merge conflicts that must be resolved before it can be
merged. Please rebase the PR, @YuhanLiu11.

https://docs.github.com/en/pull-requests/collaborating-with-pull-requests/working-with-forks/syncing-a-fork