Skip to content

ROCm PyTorch unit tests status

Jithun Nair edited this page May 17, 2023 · 45 revisions

Please refer to the PyTorch CI HUD here: https://hud.pytorch.org/hud/pytorch/pytorch/main/1?name_filter=rocm

That page lists all ROCm CI jobs, one per column, for each commit to the upstream master branch. The CI jobs of interest from unit test perspective are: Single-GPU unit tests:

  • trunk / linux-focal-rocm<ROCM_VERSION>-py3.8 / test (default, 1, 3, linux.rocm.gpu)
  • trunk / linux-focal-rocm<ROCM_VERSION>-py3.8 / test (default, 2, 3, linux.rocm.gpu)
  • trunk / linux-focal-rocm<ROCM_VERSION>-py3.8 / test (default, 3, 3, linux.rocm.gpu) Multi-GPU/distributed unit tests:
  • periodic / linux-focal-rocm<ROCM_VERSION>-py3.8 / test (distributed, 1, 2, linux.rocm.gpu)
  • periodic / linux-focal-rocm<ROCM_VERSION>-py3.8 / test (distributed, 2, 2, linux.rocm.gpu)

To view the unit test results for any job, click on the "0" (for successful jobs) or "X" (for failed jobs) in any row and in the column corresponding to that CI job, to "freeze" the tooltip. Then click on the "Raw logs" link that takes you to that particular job run.