torch.gels on GPU doesn't support m < n #3205
Description
The documentation for torch.gels suggests that when m < n for the input matrix A, then it solves the least norms problem. torch.gels with cuda tensor arguments doesn't do this.
import torch
b = torch.randn(3 ,1)
A = torch.randn(3, 5)
X, _ = torch.gels(b,A)
Y, _ = torch.gels(b.cuda(),A.cuda())
>> Traceback (most recent call last):
File "<stdin>", line 1, in <module>
RuntimeError: invalid argument 2: A should have m >= n at /home/rzou/pytorch/torch/lib/THC/generic/THCTens
orMathMagma.cu:107