How can I swap axis in a torch tensor?

Question

I have a torch tensor of size torch.Size([1, 128, 56, 128])

1 is channel, 128 is the width, and height. 56 are the stacks of images.

How can I resize it to torch.Size([1, 56, 128, 128]) ?

Do you want to switch the dimensions so that the image stack dimension becomes the 2nd dimension? Or do you want to reshape the tensor so that you have one channel, 56 width, 128 images, and 128 height? — Cecilia, May 07 '20 at 22:40
Does this answer your question? [How I can swap 3 dimensions with each other in Pytorch?](https://stackoverflow.com/questions/61104138/how-i-can-swap-3-dimensions-with-each-other-in-pytorch) — Cecilia, May 07 '20 at 22:41

score 3 · Accepted Answer · edited Apr 07 '21 at 11:28

3

You could simply use permute or transpose.

edited Apr 07 '21 at 11:28

iacob

answered May 08 '20 at 07:42

geekbird

score 3 · Answer 2 · answered Apr 07 '21 at 11:27

3

You can use the swapaxes function:

my_tensor.swapaxes(1,2)

answered Apr 07 '21 at 11:27

iacob

2 Answers2