fix crash on client disconnect (Ctrl+C)

[lvhan028]

No description provided.

[Copilot]

Pull request overview

This PR fixes a crash that occurs when a client disconnects (e.g., via Ctrl+C) by improving exception handling in the async infrastructure. The changes address race conditions and ensure proper cleanup of resources during client disconnections.

Changes:

• Added a check to prevent setting a result on an already-completed future in the StreamingSemaphore
• Improved exception handling for client disconnects (asyncio.CancelledError, GeneratorExit) with proper async cancellation
• Added backend-specific cleanup logic for PyTorch backend in the safe_run exception handler

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 2 comments.

File	Description
lmdeploy/turbomind/turbomind.py	Adds future.done() check before setting result to prevent InvalidStateError on already-completed futures
lmdeploy/serve/managers/session_manager.py	Updates exception handling to catch CancelledError and GeneratorExit, removes redundant async_end call for SafeRunException, and improves logging consistency
lmdeploy/serve/core/async_engine.py	Adds PyTorch-specific session cleanup in safe_run exception handler to ensure proper resource release

---

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