Whether 'targetSize' in inferExpandGeometryImpl needs to be checked when it is less than 0

[lfw0103]

🐛 Describe the bug

when I run below code

import torch
a = torch.randn(1,3)
a.expand([-2, 2, 1, 3])

I got error message as below:

Traceback (most recent call last): File "", line 1, in File "/home/pytorch/torch/_tensor.py", line 341, in repr return torch._tensor_str._str(self) File "/home/pytorch/torch/_tensor_str.py", line 439, in _str return _str_intern(self) File "/home/pytorch/torch/_tensor_str.py", line 414, in _str_intern tensor_str = _tensor_str(self, indent) File "/home/pytorch/torch/_tensor_str.py", line 264, in _tensor_str formatter = _Formatter(get_summarized_data(self) if summarize else self) File "/home/pytorch/torch/_tensor_str.py", line 92, in init tensor_view = tensor.reshape(-1) RuntimeError: Trying to create tensor with negative dimension -2: [-2, 2, 1, 3]

It seems to report 'reshape' has invalid parameter， actually expand op does. It causes confuse sometime. Maybe torch should check 'targerSize' in inferExpandGeometryImpl to make sure it is greater than or equal to 0.

Versions

'1.12.0a0+git0aa3c39'

[zou3519]

Agree that the error message could be clearer