CI: rocm (default, 1, 3, linux.rocm.gpu) is very slow #110181
Description
Current Status
ongoing
Issue looks like
Taking ~2.5 hours on #110167
Taking 3.5+ hours on #109976
User impact
Slower merging
Root cause
Seems like it may have been introduced in #109817 @malfet
Mitigation
Not sure
Prevention/followups
Investigate cause of slow running time or split up tests into smaller test jobs. Try to make the tests run in similar time to CUDA tests (~1.5 hours)
cc @jeffdaily @sunway513 @jithunnair-amd @pruthvistony @ROCmSupport @dllehr-amd @jataylo @hongxiayang @seemethere @malfet @pytorch/pytorch-dev-infra @ZainRizvi @kit1980 @huydhn @clee2000
### Tasks
Metadata
Metadata
Assignees
Labels
Type
Projects
Status
Status