pytorch - 💡(How to fix) Fix DISABLED test_vmap_grad_sum_cpu (__main__.TestHigherOrderOperatorInteractionCPU) [2 comments, 2 participants]

Traceback (most recent call last): File "/__w/pytorch/pytorch/test/functorch/test_eager_transforms.py", line 5315, in test_vmap_grad_sum gx = vmap(grad(sum_pyop), (0, None))(x, 0) File "/opt/conda/envs/py_3.14/lib/python3.14/site-packages/torch/_functorch/apis.py", line 220, in wrapped return vmap_impl( # pyrefly: ignore[bad-argument-type] ...<6 lines>... **kwargs, ) File "/opt/conda/envs/py_3.14/lib/python3.14/site-packages/torch/_functorch/vmap.py", line 316, in vmap_impl return _flat_vmap( func, ...<6 lines>... **kwargs, ) File "/opt/conda/envs/py_3.14/lib/python3.14/site-packages/torch/_functorch/vmap.py", line 507, in _flat_vmap batched_outputs = func(*batched_inputs, **kwargs) File "/opt/conda/envs/py_3.14/lib/python3.14/site-packages/torch/_functorch/apis.py", line 433, in wrapper return eager_transforms.grad_impl(func, argnums, has_aux, args, kwargs) ~~~~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/opt/conda/envs/py_3.14/lib/python3.14/site-packages/torch/_functorch/eager_transforms.py", line 1535, in grad_impl results = grad_and_value_impl(func, argnums, has_aux, args, kwargs) File "/opt/conda/envs/py_3.14/lib/python3.14/site-packages/torch/_functorch/vmap.py", line 56, in fn return f(*args, **kwargs) File "/opt/conda/envs/py_3.14/lib/python3.14/site-packages/torch/_functorch/eager_transforms.py", line 1486, in grad_and_value_impl output = func(*args, **kwargs) File "/__w/pytorch/pytorch/test/functorch/test_eager_transforms.py", line 5222, in call return super().call(*args, **kwargs) ~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^ File "/opt/conda/envs/py_3.14/lib/python3.14/site-packages/torch/_ops.py", line 539, in call return self.dispatch(dispatch_key_set.highestPriorityTypeId(), *args, **kwargs) ~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/opt/conda/envs/py_3.14/lib/python3.14/site-packages/torch/_ops.py", line 507, in dispatch final_key = resolve_key(self, dispatch_key) File "/opt/conda/envs/py_3.14/lib/python3.14/site-packages/torch/_ops.py", line 261, in resolve_key raise NotImplementedError(f"could not find kernel for {op} at dispatch key {k}") NotImplementedError: could not find kernel for mysum at dispatch key DispatchKey.CPU

To execute this test, run the following from the base repo dir: PYTORCH_TEST_WITH_DYNAMO=1 python test/functorch/test_eager_transforms.py TestHigherOrderOperatorInteractionCPU.test_vmap_grad_sum_cpu

This message can be suppressed by setting PYTORCH_PRINT_REPRO_ON_FAILURE=0

TL;DR

The test test_vmap_grad_sum_cpu is failing due to a NotImplementedError exception, indicating a missing kernel for the mysum operation on CPU.

Guidance

• Investigate the NotImplementedError exception and the missing kernel for mysum on CPU.
• Review the test file functorch/test_eager_transforms.py and the related code to understand the context of the error.
• Check the PyTorch documentation and issues to see if there are any known problems or solutions related to the mysum operation on CPU.
• Consider running the test with the PYTORCH_TEST_WITH_DYNAMO=1 environment variable to see if it provides more information about the issue.

Example

No specific code example is provided, as the issue is related to a missing kernel implementation.

Notes

The error message suggests that the mysum operation is not implemented for CPU, which might be a limitation of the current PyTorch version or a missing feature.

Recommendation

Apply workaround: Investigate alternative implementations or workarounds for the mysum operation on CPU, such as using a different operation or implementing a custom kernel.