洛斐 9a52063a8a Fix DP launcher hangs after worker failures
Monitor process sentinels and reap sibling ranks on worker failures,
interruption, or partial startup failure. Bound graceful termination and
report global DP rank IDs. Add CPU-only subprocess regression coverage.

Co-Authored-By: Codex <codex@openai.com>
2026-10-20 19:57:40 +08:00
..
test_launch.py Fix DP launcher hangs after worker failures 2026-10-20 19:57:40 +08:00