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

Traceback (most recent call last): File "/var/lib/jenkins/workspace/test/inductor/test_loop_ordering.py", line 478, in test_reshape_reindexing_transposed_input self.do_acc_test(f, x) File "/var/lib/jenkins/workspace/test/inductor/test_loop_ordering.py", line 240, in do_acc_test self.assertTrue(same(expect, actual, tol=1e-3)) File "/opt/conda/envs/py_3.10/lib/python3.10/unittest/case.py", line 687, in assertTrue raise self.failureException(msg) AssertionError: False is not true

To execute this test, run the following from the base repo dir: PYTORCH_TEST_CUDA_MEM_LEAK_CHECK=1 PYTORCH_TEST_WITH_SLOW_GRADCHECK=1 python test/inductor/test_loop_ordering.py LoopOrderingTest.test_reshape_reindexing_transposed_input

This message can be suppressed by setting PYTORCH_PRINT_REPRO_ON_FAILURE=0

TL;DR

• The most likely fix involves investigating and resolving the inconsistency in the test_reshape_reindexing_transposed_input test case, potentially related to numerical precision or test data.

Guidance

• Review the provided recent examples and workflow logs to understand the test failures.
• Follow the debugging instructions to find relevant log snippets by grepping for test_reshape_reindexing_transposed_input in the expanded Test step of the job.
• Investigate the AssertionError: False is not true exception in the test_reshape_reindexing_transposed_input test case, potentially related to the same function or the tol=1e-3 tolerance.
• Run the test locally using the provided command PYTORCH_TEST_CUDA_MEM_LEAK_CHECK=1 PYTORCH_TEST_WITH_SLOW_GRADCHECK=1 python test/inductor/test_loop_ordering.py LoopOrderingTest.test_reshape_reindexing_transposed_input to reproduce and debug the issue.

Example

No specific code snippet is provided, but the test case test_reshape_reindexing_transposed_input in inductor/test_loop_ordering.py should be reviewed for potential issues.

Notes

The test is marked as flaky, and the CI is green due to shielding, making it harder to parse the logs. The issue may be related to numerical precision or test data, and a thorough investigation of the test case and logs is necessary to resolve the issue.

Recommendation

• Apply workaround: Temporarily increase the tolerance tol in the same function or modify the test data to reduce the likelihood of false negatives, while investigating the root cause of the inconsistency.