Fix ray mp duplicate output

[RunningLeon]

Motivation

Fix ray mp duplicate output

Modification

Please briefly describe what modification is made in this PR.

BC-breaking (Optional)

Does the modification introduce changes that break the backward-compatibility of the downstream repositories?
If so, please describe how it breaks the compatibility and how the downstream projects should modify their code to keep compatibility with this PR.

Use cases (Optional)

If this PR introduces a new feature, it is better to list some use cases here, and update the documentation.

Checklist

1. Pre-commit or other linting tools are used to fix the potential lint issues.
2. The modification is covered by complete unit tests. If not, please add more unit tests to ensure the correctness.
3. If the modification has a dependency on downstream projects of a newer version, this PR should be tested with all supported versions of downstream projects.
4. The documentation has been modified accordingly, like docstring or example tutorials.

[Copilot]

Pull request overview

Fixes a Ray MP streaming edge case where the final EngineOutput could be yielded twice (once when produced, and again when the stream task terminates), resulting in duplicated final token output to callers.

Changes:

• Track whether the stream produced any output and avoid replaying an already-consumed terminal EngineOutput when the remote async generator finishes.
• Clear the per-stream pending slot after a consumer reads an item so the stream task can detect “already consumed” terminal output.
• Add unit tests covering: (1) no duplicate final output after cleanup delay, (2) final output returned once when unconsumed, (3) failure-before-first-yield returns INTERNAL_ENGINE_ERROR.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated no comments.

File	Description
tests/pytorch/engine/test_ray_mp_engine.py	Adds regression tests to ensure Ray MP streaming does not duplicate the final output and preserves error behavior.
lmdeploy/pytorch/engine/mp_engine/ray_engine.py	Adjusts stream task finalization and consumption tracking to prevent repeating already-consumed terminal outputs.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[lvhan028]

Is it possible that stream_out[1] is out of index?

[lvhan028]

we need to find a way to deal with these magic index in the future. Another example is the following terminal_result = pending[0] cc @grimoire