CODEWITHSUNDEEP
pythonpytorchtensor
information_interchange
information_interchange

Reputation: 3128

Iterate over two Pytorch tensors at once?

I have two Pytorch tensors (really, just 1-D lists), t1 and t2. Is it possible to iterate over them in parallel, i.e. do something like

for a,b in zip(t1,t2) ?

Thanks.

Upvotes: 12

Views: 15885

Answers (4)

Svyatoslav Shilkov
Svyatoslav Shilkov

Reputation: 56

To zip tensors in PyTorch into one use torch.stack with dim=1

Example

t1 = torch.tensor([1, 2, 3])
t2 = torch.tensor([10, 20, 30])
t3 = torch.tensor([100, 200, 300])

res = torch.stack((t1, t2, t3), dim=1)
#output
#tensor([[  1,  10, 100],
#        [  2,  20, 200],
#        [  3,  30, 300]])

Upvotes: 3

Querenker
Querenker

Reputation: 2365

For me (Python version 3.7.3 and PyTorch version 1.0.0) the zip function works as expected with PyTorch tensors:

>>> import torch
>>> t1 = torch.ones(3)
>>> t2 = torch.zeros(3)
>>> list(zip(t1, t2))
[(tensor(1.), tensor(0.)), (tensor(1.), tensor(0.)), (tensor(1.), tensor(0.))]

The list call is just needed to display the result. Iterating over zip works normally.

Upvotes: 11

Jaja
Jaja

Reputation: 808

Your can try: torch.stack(seq, dim=0, out=None) → Tensor,
for details see pytoch documentation

Upvotes: 6

gmds
gmds

Reputation: 19895

It would make more sense to concatenate them with torch.cat(dim=1); then, you can iterate over the new tensor.

Upvotes: 1

Related Questions