ROCm: Remove static specifier for allow_tf32 variable.

[jagadish-amd]

Since the env variable HIPBLASLT_ALLOW_TF32 can change, remove static type for allow_tf32 variable so that it captures the current value of env variable HIPBLASLT_ALLOW_TF32.

cc @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/147186

• 📄 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 or our office hours

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

❌ 1 New Failure, 1 Unrelated Failure

As of commit 641ce99 with merge base 331d5cf ():

NEW FAILURE - The following job has failed:

• pull / cuda12.4-py3.10-gcc9-sm75 / test (pr_time_benchmarks, 1, 1, linux.g4dn.metal.nvidia.gpu) (gh)
  REGRESSION: benchmark ('sum_floordiv_regression', 'compile_time_instruction_count') failed, actual result 1056621840 is 1.50% higher than expected 1041000000 ±+1.50% if this is an expected regression, please update the expected results.

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

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

• rocm / linux-focal-rocm6.3-py3.10 / test (default, 5, 6, linux.rocm.gpu.2) (gh) (trunk failure)
  test_public_bindings.py::TestPublicBindings::test_correct_module_names

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

[jagadish-amd]

@pytorchbot label "topic: not user facing"

[jagadish-amd]

I am working on PR to enable tf32 testing on test_nn for ROCm.
tf32 on and off https://github.com/pytorch/pytorch/blob/main/test/test_nn.py#L54 requires env variable HIPBLASLT_ALLOW_TF32 to change accordingly. Without this PR, allow_tf32 variable in Context.cpp always evaluates to whatever HIPBLASLT_ALLOW_TF32 was set initially.

cc @xw285cornell

[naromero77amd]

Test only runs on MI300, so we need to do:

from torch.testing._internal.common_utils import runOnRocmArch, MI300_ARCH

Then we need to add the decorator below to the test.

@runOnRocmArch(MI300_ARCH)

[naromero77amd]

Actually, since we don't execute any test. You can just ignore my last comment.

[jeffdaily]

I disagree. The new test is not arch-specific, it only tests whether the env var controls the returned value. No kernels are run.

[jeffdaily]

@pytorch merge -f "unrelated rocm ci failures"

[jeffdaily]

@pytorchbot merge -f "unrelated rocm ci 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