pytorch - 💡(How to fix) Fix DISABLED test_transforms_dont_support_saved_tensor_hooks_transform_grad_cpu (__main__.TestComposabilityCPU) [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 3444, in wrapper method(*args, **kwargs) ~~~~~~^^^^^^^^^^^^^^^^^ File "/opt/conda/envs/py_3.14/lib/python3.14/site-packages/torch/testing/_internal/common_device_type.py", line 459, in instantiated_test raise rte File "/opt/conda/envs/py_3.14/lib/python3.14/site-packages/torch/testing/_internal/common_device_type.py", line 439, in instantiated_test result = test(self, **param_kwargs) File "/opt/conda/envs/py_3.14/lib/python3.14/site-packages/torch/_dynamo/convert_frame.py", line 2432, 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 2150, 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 735, 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 1919, 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 1535, 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 1594, 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 1442, 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 1414, 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 342, in _fn return fn(*args, **kwargs) File "/opt/conda/envs/py_3.14/lib/python3.14/site-packages/torch/_dynamo/convert_frame.py", line 875, in trace_frame run_tracer() ~~~~~~~~~~^^ File "/opt/conda/envs/py_3.14/lib/python3.14/site-packages/torch/_dynamo/convert_frame.py", line 856, in run_tracer tracer.run() ~~~~~~~~~~^^ File "/opt/conda/envs/py_3.14/lib/python3.14/site-packages/torch/_dynamo/symbolic_convert.py", line 1813, in run while self.step(): ~~~~~~~~~^^ File "/opt/conda/envs/py_3.14/lib/python3.14/site-packages/torch/_dynamo/symbolic_convert.py", line 1480, in step self.dispatch_table[inst.opcode](self, inst) ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^ File "/opt/conda/envs/py_3.14/lib/python3.14/site-packages/torch/_dynamo/symbolic_convert.py", line 1017, in wrapper return inner_fn(self, inst) File "/opt/conda/envs/py_3.14/lib/python3.14/site-packages/torch/_dynamo/symbolic_convert.py", line 4385, in CALL_KW self._call(inst, call_kw=True) ~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^ File "/opt/conda/envs/py_3.14/lib/python3.14/site-packages/torch/_dynamo/symbolic_convert.py", line 4161, in _call self.call_function(fn, args, kwargs) ~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^ File "/opt/conda/envs/py_3.14/lib/python3.14/site-packages/torch/_dynamo/symbolic_convert.py", line 1381, in call_function self.push(fn.call_function(self, args, kwargs)) # type: ignore[arg-type] ~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^ File "/opt/conda/envs/py_3.14/lib/python3.14/site-packages/torch/_dynamo/variables/lazy.py", line 294, in realize_and_forward return getattr(self.realize(), name)(args, **kwargs) ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^ File "/opt/conda/envs/py_3.14/lib/python3.14/site-packages/torch/_dynamo/variables/torch.py", line 2699, in call_function tensor_variable = wrap_fx_proxy( tx=tx, ...<4 lines>... ), ) File "/opt/conda/envs/py_3.14/lib/python3.14/site-packages/torch/_dynamo/variables/builder.py", line 3078, in wrap_fx_proxy return wrap_fx_proxy_cls(target_cls=TensorVariable, **kwargs) File "/opt/conda/envs/py_3.14/lib/python3.14/site-packages/torch/_dynamo/variables/builder.py", line 3153, in wrap_fx_proxy_cls out: VTTypeAlias = _wrap_fx_proxy( ~~~~~~~~~~~~~~^ target_cls, tx, proxy, example_value, subclass_type, **options ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ ) ^ File "/opt/conda/envs/py_3.14/lib/python3.14/site-packages/torch/_dynamo/variables/builder.py", line 3277, in _wrap_fx_proxy example_value = get_fake_value(proxy.node, tx, allow_non_graph_fake=True) File "/opt/conda/envs/py_3.14/lib/python3.14/site-packages/torch/_dynamo/utils.py", line 3752, in get_fake_value return _get_fake_value_impl(node, tx, allow_non_graph_fake) File "/opt/conda/envs/py_3.14/lib/python3.14/site-packages/torch/_dynamo/utils.py", line 3943, in _get_fake_value_impl _wrap_graph_break_with_torch_runtime_err( ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^ lambda: unimplemented( ^^^^^^^^^^^^^^^^^^^^^^ ...<5 lines>... ) ^ ) ^ File "/opt/conda/envs/py_3.14/lib/python3.14/site-packages/torch/_dynamo/utils.py", line 3741, in _wrap_graph_break_with_torch_runtime_err raise exc.with_traceback(e.traceback) from None File "/opt/conda/envs/py_3.14/lib/python3.14/site-packages/torch/_dynamo/utils.py", line 3738, in _wrap_graph_break_with_torch_runtime_err gb_fn() ~~~~~^^ File "/opt/conda/envs/py_3.14/lib/python3.14/site-packages/torch/_dynamo/utils.py", line 3944, in <lambda> lambda: unimplemented( ~~~~~~~~~~~~~^ gb_type="RuntimeError when making fake tensor call", ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ ...<3 lines>... from_exc=cause, ^^^^^^^^^^^^^^^ ) ^ File "/opt/conda/envs/py_3.14/lib/python3.14/site-packages/torch/_dynamo/exc.py", line 653, in unimplemented raise Unsupported( msg, gb_type, skip_frame, real_stack=past_real_stack ) from from_exc torch._dynamo.exc.TorchRuntimeError: RuntimeError when making fake tensor call Explanation: Dynamo failed to run FX node with fake tensors: call_function <built-in method randn of type object at 0x7fa3ddb76910>((3,), **{'device': 'cpu'}): got RuntimeError('tls_on_entry.has_value() INTERNAL ASSERT FAILED at "/__w/pytorch/pytorch/aten/src/ATen/core/PythonFallbackKernel.cpp":53, please report a bug to PyTorch. ') Hint: Your code may result in an error when running in eager. Please double check that your code doesn't contain a similar error when actually running eager/uncompiled. You can do this by removing the torch.compile call, or by using torch.compiler.set_stance("force_eager").

Developer debug context:

For more details about this graph break, please visit: https://meta-pytorch.github.io/compile-graph-break-site/gb/gb4315.html

from user code: File "/__w/pytorch/pytorch/test/functorch/test_eager_transforms.py", line 3814, in test_transforms_dont_support_saved_tensor_hooks x = torch.randn(3, device=device)

Set TORCHDYNAMO_VERBOSE=1 for the internal stack trace (please do this especially if you're reporting a bug to PyTorch). For even more developer context, set TORCH_LOGS="+dynamo"

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

This message can be suppressed by setting PYTORCH_PRINT_REPRO_ON_FAILURE=0

TL;DR

The test test_transforms_dont_support_saved_tensor_hooks_transform_grad_cpu is failing due to a TorchRuntimeError caused by Dynamo's failure to run FX node with fake tensors, and a potential workaround is to remove the torch.compile call or use torch.compiler.set_stance("force_eager").

Guidance

• The error message suggests that the issue is related to the torch.randn function call, which is causing a RuntimeError when making a fake tensor call.
• To debug this issue, try removing the torch.compile call or using torch.compiler.set_stance("force_eager") to see if the error persists.
• Set TORCHDYNAMO_VERBOSE=1 to get the internal stack trace, and TORCH_LOGS="+dynamo" for more developer context.
• Run the test with PYTORCH_TEST_WITH_DYNAMO=1 to reproduce the error.

Example

No code example is provided as the issue is related to a specific test case and requires more context to provide a working example.

Notes

The issue seems to be related to the torch.randn function call, which is causing a RuntimeError when making a fake tensor call. The error message provides a hint to remove the torch.compile call or use torch.compiler.set_stance("force_eager") to debug the issue.

Recommendation

Apply the workaround by removing the torch.compile call or using torch.compiler.set_stance("force_eager"), as this may resolve the issue and allow the test to pass.