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

Traceback (most recent call last): File "/opt/conda/envs/py_3.14/lib/python3.14/site-packages/torch/testing/_internal/common_utils.py", line 3514, in wrapper method(*args, **kwargs) ~~~~~~^^^^^^^^^^^^^^^^^ File "/opt/conda/envs/py_3.14/lib/python3.14/site-packages/torch/_dynamo/convert_frame.py", line 2480, in call result = self._torchdynamo_orig_backend( frame, cache_entry, self.hooks, frame_state, skip=1 ) File "/opt/conda/envs/py_3.14/lib/python3.14/site-packages/torch/_dynamo/convert_frame.py", line 2198, in call result = self._inner_convert( frame, cache_entry, hooks, frame_state, skip=skip + 1 ) File "/opt/conda/envs/py_3.14/lib/python3.14/site-packages/torch/_dynamo/convert_frame.py", line 758, in call result = _compile( frame.f_code, ...<16 lines>... convert_frame_box=self._box, ) File "/opt/conda/envs/py_3.14/lib/python3.14/site-packages/torch/_dynamo/convert_frame.py", line 1983, in _compile guarded_code, tracer_output = compile_inner(code, one_graph, hooks) ~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^ File "/opt/conda/envs/py_3.14/lib/python3.14/site-packages/torch/_utils_internal.py", line 96, in wrapper_function return function(*args, **kwargs) File "/opt/conda/envs/py_3.14/lib/python3.14/site-packages/torch/_dynamo/convert_frame.py", line 1593, in compile_inner result = _compile_inner(code, one_graph, hooks) File "/opt/conda/envs/py_3.14/lib/python3.14/site-packages/torch/_dynamo/convert_frame.py", line 1652, in _compile_inner dynamo_output = compile_frame( code, ...<11 lines>... package=package, ) File "/opt/conda/envs/py_3.14/lib/python3.14/site-packages/torch/_dynamo/convert_frame.py", line 1500, in compile_frame bytecode, tracer_output = transform_code_object(code, transform) ~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^ File "/opt/conda/envs/py_3.14/lib/python3.14/site-packages/torch/_dynamo/bytecode_transformation.py", line 1626, in transform_code_object tracer_output = transformations(instructions, code_options) File "/opt/conda/envs/py_3.14/lib/python3.14/site-packages/torch/_dynamo/convert_frame.py", line 1472, in transform tracer_output = trace_frame( code, ...<14 lines>... package=package, ) File "/opt/conda/envs/py_3.14/lib/python3.14/site-packages/torch/_dynamo/convert_frame.py", line 388, in _fn assert guards.check(), ( ~~~~~~~~~~~~^^ AssertionError: Global torch_function state changed while dynamo tracing, please report a bug

To execute this test, run the following from the base repo dir: PYTORCH_TEST_WITH_DYNAMO=1 python test/test_overrides.py TestTorchFunctionMode.test_distributions_bernoulli

This message can be suppressed by setting PYTORCH_PRINT_REPRO_ON_FAILURE=0

TL;DR

The most likely fix involves investigating and addressing the AssertionError: Global torch_function state changed while dynamo tracing issue in the test_distributions_bernoulli test.

Guidance

• Investigate the test_distributions_bernoulli test in test_overrides.py to understand the conditions under which the AssertionError occurs.
• Follow the provided debugging instructions to analyze the workflow logs and identify patterns or clues that could lead to the root cause of the issue.
• Consider running the test with PYTORCH_TEST_WITH_DYNAMO=1 python test/test_overrides.py TestTorchFunctionMode.test_distributions_bernoulli to reproduce the error locally and gather more information.
• Look into the torch_function state management in the Dynamo tracing context to understand how its change could trigger the assertion error.

Example

No specific code snippet can be provided without further investigation, but the error message suggests looking into the torch/_dynamo/convert_frame.py file, particularly around the trace_frame function and the guards.check() assertion.

Notes

The issue seems to be related to the interaction between Dynamo tracing and torch_function state management. Without more specific details on the test or the environment, it's challenging to provide a direct fix. The provided debugging instructions and test execution command are crucial steps in diagnosing the problem.

Recommendation

Apply workaround: Temporarily disable the test or modify the test to avoid the condition that triggers the AssertionError while the root cause is being investigated, as the test is currently disabled in CI due to its flaky nature.