[Symmetric memory] set handle type for ROCm

[ngimel]

Fixes #161722

cc @H-Huang @awgu @wanchaol @fegin @fduwjj @wz337 @wconstab @d4l3k @pragupta @jeffdaily @sunway513 @jithunnair-amd @pruthvistony @ROCmSupport @dllehr-amd @jataylo @hongxiayang @naromero77amd

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/161741

• 📄 Preview Python docs built from this PR
• 📄 Preview C++ docs built from this PR
• ❓ Need help or want to give feedback on the CI? Visit the bot commands wiki

Note: Links to docs will display an error until the docs builds have been completed.

✅ You can merge normally! (2 Unrelated Failures)

As of commit e05819a with merge base 82f63c8 ():

FLAKY - The following job failed but was likely due to flakiness present on trunk:

• periodic-rocm-mi300 / linux-jammy-rocm-py3.10-mi300 / test (distributed, 1, 3, linux.rocm.gpu.gfx942.4, module:rocm, oncall:distributed) (gh) (disabled by #162071)
  distributed/test_symmetric_memory.py::SymmMemEmptySetDeviceTest::test_empty_strided_p2p_set_device_False

BROKEN TRUNK - The following job failed but was present on the merge base:

👉 Rebase onto the `viable/strict` branch to avoid these failures

• rocm / linux-jammy-rocm-py3.10 / test (default, 1, 6, linux.rocm.gpu.2) (gh) (trunk failure)
  inductor/test_provenance_tracing.py::TestProvenanceTracingStackTraces::test_cpu_extern_kernel

This comment was automatically generated by Dr. CI and updates every 15 minutes.

[ngimel]

@yuankaichen-amd my PR couldn't possibly have caused those rocm failures, I'll leave it to you to figure it out and land the PR

[yuankaichen-amd]

Hi Natalia,

Please rebase your branch to head and try again. For the two failures, they are irrelevant to this PR:

1. test_baddmm_search_space_EXHAUSTIVE: we should disable this test IIUC. @chuanqi129 Please comment whether this has been already disabled for ROCM, if not, how should we disable it?
2. test_compile_selective_checkpoint_custom_rule_cuda: @ngimel your branch somehow messed up. This parameter was killed and then the kill was reverted. Rebasing to head should do it.

[ngimel]

@pytorchbot rebase -b main

[pytorchmergebot]

@pytorchbot started a rebase job onto refs/remotes/origin/main. Check the current status here

[pytorchmergebot]

Successfully rebased ngimel/rocm_handle_type onto refs/remotes/origin/main, please pull locally before adding more changes (for example, via git checkout ngimel/rocm_handle_type && git pull --rebase)

[jeffdaily]

Please note that ciflow/rocm (mi200s) are having infra issues and the signal will take a long time. But for this PR this test is only run on the mi300 anyway, so you can force-merge once you get a good mi300 signal.

[ngimel]

@yuankaichen-amd I see the CI is still failing for reasons unrelated to my PR. Feel free to take this PR over and do with it what you want.

[jeffdaily]

@pytorchbot rebase

[pytorchmergebot]

@pytorchbot started a rebase job onto refs/remotes/origin/viable/strict. Check the current status here

[pytorchmergebot]

Tried to rebase and push PR #161741, but it was already up to date. Try rebasing against main by issuing:
@pytorchbot rebase -b main

[jeffdaily]

@pytorchbot merge -f "unrelated rocm failures"

[pytorchmergebot]

Merge started

Your change will be merged immediately since you used the force (-f) flag, bypassing any CI checks (ETA: 1-5 minutes). Please use -f as last resort and instead consider -i/--ignore-current to continue the merge ignoring current failures. This will allow currently pending tests to finish and report signal before the merge.

Learn more about merging in the wiki.

Questions? Feedback? Please reach out to the PyTorch DevX Team

Advanced Debugging

Check the merge workflow status
here

[yuankaichen-amd]

Thank you all!