I have a loop that has 4 iterations. In each iteration, I will compute a tensor of size [8, x, 128, 128], where x is a variable size.
I would like to concatenate each tensor to a another one (along dimension 1 here), as soon as it gets computed. However, I cannot use torch.stack() because (apparently), all the sizes need to be similar.
I cannot use torch.cat(), because I need to concatenate them as soon as they are computed.
EDIT: Never mind I figured it out!