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

Traceback (most recent call last): File "/var/lib/jenkins/workspace/test/dynamo/test_nested_graph_breaks.py", line 660, in test_nested_graph_break_in_loop self.assertEqual(torch._dynamo.utils.counters["frames"]["total"], 6) ~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/opt/conda/envs/py_3.14/lib/python3.14/site-packages/torch/_dynamo/test_case.py", line 113, in assertEqual return super().assertEqual(x, y, *args, **kwargs) ~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^ File "/opt/conda/envs/py_3.14/lib/python3.14/site-packages/torch/testing/_internal/common_utils.py", line 4440, in assertEqual raise error_metas.pop()[0].to_error( # type: ignore[index] ...<4 lines>... ) AssertionError: Scalars are not equal!

Expected 6 but got 7. Absolute difference: 1 Relative difference: 0.16666666666666666

To execute this test, run the following from the base repo dir: python test/dynamo/test_nested_graph_breaks.py NestedGraphBreakTests.test_nested_graph_break_in_loop

This message can be suppressed by setting PYTORCH_PRINT_REPRO_ON_FAILURE=0

TL;DR

The most likely fix involves updating the test case in test_nested_graph_breaks.py to account for the changed expected value of torch._dynamo.utils.counters["frames"]["total"].

Guidance

• Review the test case test_nested_graph_break_in_loop in test_nested_graph_breaks.py to understand why the expected value of torch._dynamo.utils.counters["frames"]["total"] is being set to 6.
• Investigate why the actual value is 7 and determine if this is the correct expected value.
• Update the test case to reflect the correct expected value.
• Run the test again to verify that the update resolves the issue.

Example

No code snippet is provided as the issue does not imply a specific code change.

Notes

The issue is related to a flaky test, and the provided error message indicates an assertion error due to a mismatch between the expected and actual values of torch._dynamo.utils.counters["frames"]["total"]. The test is disabled, and the issue is being tracked.

Recommendation

Apply a workaround by updating the test case to reflect the correct expected value, as the root cause of the issue is related to the test case itself rather than a version or dependency issue.