trunk/a0d597dc6fd56546a9ed5dd01422d27633282547: [ROCm] Run the inductor-rocm-mi300 test config in four shards (#196642)
- PyTorch: 767 events in the last 90 days
- PyTorch: 757th Release in the last 90 days
- Previous: earlier the same day · trunk/377d83a554d93e9fd6bb2949daa26d8d551398ad: [dynamo] Port CPython 3.13 test_unary to Dynamo test suite (#196240)
What happened
Since #192262 re-enabled the MI300 workflows on 2026-08-11 with a two-shard inductor matrix, every inductor-rocm-mi300 run on main has ended with "The action 'Test' has timed out after 270 minutes" on both shards (88 timed-out jobs between 2026-09-03 and 2026-09-10; last green run 2026-08-11). It is not a hang: each sub-shard completes in 8-18 minutes and the serial list needs about 4.5 hours, so the kill always lan…
Summary assembled by rule from the sources below