How does dim dim in squeeze and unsqueze work?
x = torch.zeros(2, 1, 2, 1, 2)
x.squeeze()
torch.Size([2, 2, 2])
y=torch.squeeze(x,1)
torch.Size([1, 2, 1, 2, 1, 2]) .//no change
y=torch.squeeze(x,2)
results :-
torch.Size([1, 2, 2, 1, 2])
Hi Sandeep!
torch.squeeze() removes those dimensions of a tensor that have
size = 1. This is explained reasonably clearly in the documentation
for torch.squeeze.
Specifically, if you give squeeze() no dim argument, it removes all
size = 1 dimensions. If you give squeeze() a value for dim, it
removes just that dimension, but only if it has size = 1.
This looks like a typo. Based on the results you give below, I think
this should be x = torch.zeros(1, 2, 1, 2, 1, 2).
Best.
K. Frank