pytorch - 💡(How to fix) Fix DISABLED test_stream_context_with_data_dependency (__main__.TestUserStreamCompile) [1 comments, 1 participants]

Traceback (most recent call last): File "/__w/pytorch/pytorch/test/inductor/test_user_streams.py", line 342, in test_stream_context_with_data_dependency self.assertEqual(result, expected) File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/_dynamo/test_case.py", line 121, in assertEqual return super().assertEqual(x, y, *args, **kwargs) File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/testing/_internal/common_utils.py", line 4511, in assertEqual raise error_metas.pop()[0].to_error( # type: ignore[index] AssertionError: Tensor-likes are not close!

Mismatched elements: 1024 / 1024 (100.0%) Greatest absolute difference: 22.119400024414062 at index (550,) (up to 1e-05 allowed) Greatest relative difference: 579.9586791992188 at index (344,) (up to 1.3e-06 allowed)

To execute this test, run the following from the base repo dir: python test/inductor/test_user_streams.py TestUserStreamCompile.test_stream_context_with_data_dependency

This message can be suppressed by setting PYTORCH_PRINT_REPRO_ON_FAILURE=0

TL;DR

The test test_stream_context_with_data_dependency is failing due to assertion errors, likely caused by numerical instability or precision issues in tensor comparisons.

Guidance

• Review the test case in inductor/test_user_streams.py and examine the test_stream_context_with_data_dependency method to understand the assertion failure.
• Investigate the numerical instability by analyzing the Greatest absolute difference and Greatest relative difference values in the error message.
• Consider relaxing the tolerance values in the assertEqual statement or using a more robust comparison method.
• Run the test locally using the provided command to reproduce and debug the issue.

Example

No code snippet is provided as the issue is related to a specific test case and requires further investigation.

Notes

The issue is likely related to numerical precision or instability, which can be challenging to debug. The provided error message and test case information should be carefully analyzed to determine the root cause.

Recommendation

Apply workaround: Relax the tolerance values in the assertEqual statement to account for numerical instability, and re-enable the test to monitor its stability.