Tensor Padding Pytorch

Question

Good evening,
I am having a little trouble with the dimensions of some tensors and I would like to pad them with rows of 0s but I am not managing to do it. My tensors are of size X by 8 and I want to add rows of 0s (of 8 elements each) until they reach the same size as the tensor with the largest amount of rows in a list. I've obtained the maximum size with:

max([x.size() for x in dataset])

But I am lost on how to pad the rows. Thanks a lot for your help.
Pd: Just in case, here is one example of one of my tensors:

tensor([[1., 0., 0., 0., 0., 0., 0., 0.],
    [1., 0., 0., 0., 0., 0., 0., 0.],
    [1., 0., 0., 0., 0., 0., 0., 0.],
    [1., 0., 0., 0., 0., 0., 0., 0.],
    [1., 0., 0., 0., 0., 0., 0., 0.],
    [0., 0., 0., 0., 0., 1., 0., 0.],
    [0., 0., 1., 0., 0., 0., 0., 0.],
    [0., 0., 1., 0., 0., 0., 0., 0.],
    [1., 0., 0., 0., 0., 0., 0., 0.],
    [1., 0., 0., 0., 0., 0., 0., 0.],
    [1., 0., 0., 0., 0., 0., 0., 0.],
    [1., 0., 0., 0., 0., 0., 0., 0.],
    [1., 0., 0., 0., 0., 0., 0., 0.],
    [1., 0., 0., 0., 0., 0., 0., 0.],
    [1., 0., 0., 0., 0., 0., 0., 0.],
    [1., 0., 0., 0., 0., 0., 0., 0.],
    [0., 0., 1., 0., 0., 0., 0., 0.],
    [1., 0., 0., 0., 0., 0., 0., 0.],
    [0., 0., 1., 0., 0., 0., 0., 0.],
    [1., 0., 0., 0., 0., 0., 0., 0.],
    [1., 0., 0., 0., 0., 0., 0., 0.]])

Desired output if the tensor would have 2 more rows:

tensor([[1., 0., 0., 0., 0., 0., 0., 0.],
    [1., 0., 0., 0., 0., 0., 0., 0.],
    [1., 0., 0., 0., 0., 0., 0., 0.],
    [1., 0., 0., 0., 0., 0., 0., 0.],
    [1., 0., 0., 0., 0., 0., 0., 0.],
    [0., 0., 0., 0., 0., 1., 0., 0.],
    [0., 0., 1., 0., 0., 0., 0., 0.],
    [0., 0., 1., 0., 0., 0., 0., 0.],
    [1., 0., 0., 0., 0., 0., 0., 0.],
    [1., 0., 0., 0., 0., 0., 0., 0.],
    [1., 0., 0., 0., 0., 0., 0., 0.],
    [1., 0., 0., 0., 0., 0., 0., 0.],
    [1., 0., 0., 0., 0., 0., 0., 0.],
    [1., 0., 0., 0., 0., 0., 0., 0.],
    [1., 0., 0., 0., 0., 0., 0., 0.],
    [1., 0., 0., 0., 0., 0., 0., 0.],
    [0., 0., 1., 0., 0., 0., 0., 0.],
    [1., 0., 0., 0., 0., 0., 0., 0.],
    [0., 0., 1., 0., 0., 0., 0., 0.],
    [1., 0., 0., 0., 0., 0., 0., 0.],
    [1., 0., 0., 0., 0., 0., 0., 0.],
[0., 0., 0., 0., 0., 0., 0., 0.],
[0., 0., 0., 0., 0., 0., 0., 0.]])

Could you provide the desired result associated with this tensor? — Ivan
– Ivan, Commented Apr 28, 2022 at 12:48
@Ivan I ve just edited the post with an example output result. — Alvaro Ciudad
– Alvaro Ciudad, Commented Apr 28, 2022 at 14:46

Ivan · Accepted Answer · 2022-04-28 16:09:56Z

1

You can utilize the torch.nn.functional.pad function:

x = torch.tensor([[1., 0., 0., 0., 0., 0., 0., 0.],
                  [1., 0., 0., 0., 0., 0., 0., 0.],
                  [1., 0., 0., 0., 0., 0., 0., 0.],
                  [1., 0., 0., 0., 0., 0., 0., 0.]])

Given a maximum number of rows, rows (here rows = 6):

>>> F.pad(x, (0,0,0,rows-len(x)), value=0)
tensor([[1., 0., 0., 0., 0., 0., 0., 0.],
        [1., 0., 0., 0., 0., 0., 0., 0.],
        [1., 0., 0., 0., 0., 0., 0., 0.],
        [1., 0., 0., 0., 0., 0., 0., 0.],
        [0., 0., 0., 0., 0., 0., 0., 0.],
        [0., 0., 0., 0., 0., 0., 0., 0.]])

answered Apr 28, 2022 at 16:09

Ivan

41.3k9 gold badges78 silver badges120 bronze badges

Sign up to request clarification or add additional context in comments.

Collectives™ on Stack Overflow

Tensor Padding Pytorch

1 Answer 1

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

Comments

Your Answer

Sign up or log in

Post as a guest

Related