I am trying to make a concatenation operation using this line of code:
‘’’
mixture = torch.cat([mixture, torch.zeros(1, 1, padded_length, device=device2)], dim=-1)
‘’’’
Actually, it works well with PyTorch version 1.4.0 but in PyTorch version 1.7.1 it complains can anyone explain Why?
Could you describe the issue a bit more, please, and explain what the issue is exactly? Also, an executable code snippet to reproduce the issue in the latest release (assuming you are also seeing it in 1.9.0) would be helpful.
PS: I would not recommend to tag specific users, as it could discourage others to take a look at your question.
@ptrblck Sorry for specific tagging I will not do this again. I solved this problem.
I have another question I have a tensor of shape [1, 1, 64000] where the first 1 is the batch size can I make it of shape [1, 64000] i.e remove the first dim.
Yes, you can remove dimensions with a size of 1 by using tensor = tensor.squeeze(dim), where dim=0 in your case.
So to if the tensor (x) shape for example [2, 1, 64000] and want to reshape it to anothe tensor (y) of shape [1, 64000] I will use
y = x.squeeze(dim=0)
No, squeezing a dimension with size>1 won’t work, since you would need to reduce the number of elements and could index the tensor instead e.g. via y = x[0].