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

Traceback (most recent call last): File "/opt/conda/envs/py_3.14/lib/python3.14t/unittest/case.py", line 58, in testPartExecutor yield File "/opt/conda/envs/py_3.14/lib/python3.14t/unittest/case.py", line 669, in run self._callTestMethod(testMethod) ~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^ File "/opt/conda/envs/py_3.14/lib/python3.14t/unittest/case.py", line 615, in _callTestMethod result = method() File "/opt/conda/envs/py_3.14/lib/python3.14t/site-packages/torch/testing/_internal/common_utils.py", line 3532, in wrapper method(*args, **kwargs) ~~~~~~^^^^^^^^^^^^^^^^^ File "/__w/pytorch/pytorch/test/test_overrides.py", line 1505, in test_with_mode_created_separately class ErrorA(RuntimeError): File "/__w/pytorch/pytorch/test/test_overrides.py", line 1508, in torch_dynamo_resume_in_test_with_mode_created_separately_at_1505 class A(TorchFunctionMode): File "/opt/conda/envs/py_3.14/lib/python3.14t/site-packages/torch/_dynamo/convert_frame.py", line 2506, in call result = self._torchdynamo_orig_backend( frame, cache_entry, self.hooks, frame_state, skip=1 ) File "/opt/conda/envs/py_3.14/lib/python3.14t/site-packages/torch/_dynamo/convert_frame.py", line 2217, in call result = self._inner_convert( frame, cache_entry, hooks, frame_state, skip=skip + 1 ) File "/opt/conda/envs/py_3.14/lib/python3.14t/site-packages/torch/_dynamo/convert_frame.py", line 772, in call result = _compile( frame.f_code, ...<18 lines>... convert_frame_box=self._box, ) File "/opt/conda/envs/py_3.14/lib/python3.14t/site-packages/torch/_dynamo/convert_frame.py", line 2001, in _compile guarded_code, tracer_output = compile_inner(code, one_graph, hooks) ~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^ File "/opt/conda/envs/py_3.14/lib/python3.14t/site-packages/torch/_utils_internal.py", line 96, in wrapper_function return function(*args, **kwargs) File "/opt/conda/envs/py_3.14/lib/python3.14t/site-packages/torch/_dynamo/convert_frame.py", line 1611, in compile_inner result = _compile_inner(code, one_graph, hooks) File "/opt/conda/envs/py_3.14/lib/python3.14t/site-packages/torch/_dynamo/convert_frame.py", line 1670, in _compile_inner dynamo_output = compile_frame( code, ...<11 lines>... package=package, ) File "/opt/conda/envs/py_3.14/lib/python3.14t/site-packages/torch/_dynamo/convert_frame.py", line 1516, in compile_frame bytecode, tracer_output = transform_code_object(code, transform) ~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^ File "/opt/conda/envs/py_3.14/lib/python3.14t/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/python3.14t/site-packages/torch/_dynamo/convert_frame.py", line 1488, in transform tracer_output = trace_frame( code, ...<14 lines>... package=package, ) File "/opt/conda/envs/py_3.14/lib/python3.14t/site-packages/torch/_dynamo/convert_frame.py", line 390, 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_with_mode_created_separately

This message can be suppressed by setting PYTORCH_PRINT_REPRO_ON_FAILURE=0

TL;DR

The most likely fix for the failing test is to investigate and address the issue with the global torch_function state changing during dynamo tracing.

Guidance

• Investigate the test_with_mode_created_separately test in test_overrides.py to understand the conditions under which the global torch_function state is changing.
• Review the workflow logs linked in the issue to identify patterns or common factors in the failures.
• Check the PyTorch documentation and issues for known problems related to dynamo tracing and global torch_function state.
• Consider setting PYTORCH_PRINT_REPRO_ON_FAILURE=1 to gather more information about the failure.

Example

No specific code changes can be suggested without further investigation, but the test can be executed with PYTORCH_TEST_WITH_DYNAMO=1 python test/test_overrides.py TestTorchFunctionMode.test_with_mode_created_separately to reproduce the issue.

Notes

The issue seems to be related to the interaction between PyTorch's dynamo tracing and the global torch_function state. Without more information, it's difficult to provide a specific fix.

Recommendation

Apply workaround: Investigate and address the issue with the global torch_function state changing during dynamo tracing, as this is the most likely cause of the test failure.