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

Traceback (most recent call last): File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/testing/_internal/common_utils.py", line 3514, in wrapper method(*args, **kwargs) File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/testing/_internal/common_utils.py", line 3514, in wrapper method(*args, **kwargs) File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/testing/_internal/common_utils.py", line 3513, in wrapper with policy(): File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/testing/_internal/common_utils.py", line 2804, in exit self.testcase.before_cuda_memory_leak_check() File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/testing/_internal/common_utils.py", line 3414, in before_cuda_memory_leak_check torch._dynamo.reset() File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/_dynamo/init.py", line 180, in reset from torch._inductor.cudagraph_trees import reset_cudagraph_trees File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/_inductor/cudagraph_trees.py", line 69, in <module> from torch._inductor.compile_fx import ( File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/_inductor/compile_fx.py", line 112, in <module> from . import config, distributed_autotune, metrics File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/_inductor/distributed_autotune.py", line 15, in <module> from . import config, select_algorithm File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/_inductor/select_algorithm.py", line 5864, in <module> from . import lowering # noqa: F401 File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/_inductor/lowering.py", line 62, in <module> from .decomposition import decompositions, get_decompositions File "/opt/conda/envs/py_3.10/lib/python3.10/site-packages/torch/_inductor/decomposition.py", line 108, in <module> decompositions = {**core_aten_decompositions(), **inductor_decompositions} TypeError: 'CustomDecompTable' object is not a mapping

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/test_torch.py TestTorchDeviceTypeCUDA.test__local_scalar_dense_with_empty_tensor_cuda

This message can be suppressed by setting PYTORCH_PRINT_REPRO_ON_FAILURE=0

TL;DR

The test test__local_scalar_dense_with_empty_tensor_cuda is failing due to a TypeError caused by a CustomDecompTable object not being a mapping, and a potential workaround is to investigate and fix the decomposition table initialization.

Guidance

• Investigate the decomposition.py file to understand how the decompositions dictionary is initialized and why it's expecting a mapping object.
• Check the CustomDecompTable class to see if it's supposed to be a mapping object and if it's being used correctly.
• Run the test with the provided command PYTORCH_TEST_CUDA_MEM_LEAK_CHECK=1 PYTORCH_TEST_WITH_SLOW_GRADCHECK=1 python test/test_torch.py TestTorchDeviceTypeCUDA.test__local_scalar_dense_with_empty_tensor_cuda to reproduce the error and gather more information.
• Review the recent workflow logs and test results to see if there are any patterns or clues that could help identify the root cause.

Example

No code example is provided as the issue is related to a specific PyTorch test and requires more context to provide a accurate solution.

Notes

The issue seems to be related to a specific test case and a custom decomposition table, which might require knowledge of PyTorch's internal workings to fix. The provided stacktrace and error message suggest that the issue is related to a TypeError, but more investigation is needed to determine the root cause.

Recommendation

Apply workaround: Investigate and fix the decomposition table initialization to ensure that it's correctly creating a mapping object. This might involve modifying the decomposition.py file or the CustomDecompTable class.