[Functionalization] test_copy_noncontig_xla

Question

[Functionalization] test_copy_noncontig_xla

wonjoolee95 opened this issue a year ago · comments

Running python test/test_torch.py TestTorchDeviceTypeXLA.test_copy_noncontig_xla fails with

RuntimeError: false INTERNAL ASSERT FAILED at "/workspace/pytorch/build/aten/src/ATen/RegisterFunctionalization_0.cpp":3547, please report a bug to PyTorch. mutating a non-functional tensor with a functional tensor is not allowed. Please ensure that all of your inputs are wrapped inside of a functionalize() call.

Jiewen Tan · Answer 1 · Sat Dec 17 2022 06:14:08 GMT+0800 (China Standard Time)

Is this fixed already?

Wonjoo Lee · Answer 2 · Tue Dec 20 2022 02:12:45 GMT+0800 (China Standard Time)

It is fixed but requires some changes to the original Brian's PR. I've described the changes at https://github.com/pytorch/pytorch/pull/91001/files#r1052494781. I just wanted to double check with Brian before closing this issue.