[MPS] Incorrect results of two multiplied chunked tensors produced by unsafe_chunk #91065
Description
🐛 Describe the bug
This causes the inconsistent results mentioned in #90421. torch.chunk has no issue.
import torch
a = torch.rand(5, dtype=torch.float32, device="cpu")
ret = a.unsafe_chunk(4, 0)
y = ret[0] * ret[2]
a_mps = a.to("mps")
ret_mps = a_mps.unsafe_chunk(4, 0)
y_mps = ret_mps[0] * ret_mps[2]
torch.testing.assert_close(y, y_mps, check_device=False)Versions
PyTorch version: 2.0.0a0+giteae0f3f
Is debug build: True
CUDA used to build PyTorch: None
ROCM used to build PyTorch: N/A
OS: macOS 12.6.1 (arm64)
GCC version: Could not collect
Clang version: 13.1.6 (clang-1316.0.21.2.5)
CMake version: version 3.24.0
Libc version: N/A
Python version: 3.9.13 | packaged by conda-forge | (main, May 27 2022, 17:00:33) [Clang 13.0.1 ] (64-bit runtime)
Python platform: macOS-12.6.1-arm64-arm-64bit
Is CUDA available: False
CUDA runtime version: No CUDA
GPU models and configuration: No CUDA
Nvidia driver version: No CUDA
cuDNN version: No CUDA
HIP runtime version: N/A
MIOpen runtime version: N/A
Is XNNPACK available: False
Versions of relevant libraries:
[pip3] mypy==0.960
[pip3] mypy-extensions==0.4.3
[pip3] numpy==1.21.6
[pip3] torch==2.0.0a0+giteae0f3f
[pip3] torchvision==0.13.0
[conda] numpy 1.21.6 pypi_0 pypi
[conda] torch 2.0.0a0+giteae0f3f dev_0 <develop>
[conda] torchvision 0.13.0 pypi_0 pypi
cc @kulinseth @albanD @malfet @DenisVieriu97 @razarmehr @abhudev