I have a tensor of size (2, b, h) and I want to change it to the following size: (b, 2*h), where the corresponding lists are concatenated, for example:
a = torch.tensor([[[1, 2, 3], [4, 5, 6], [4, 4, 4]],
[[4, 5, 6], [7, 8, 9], [5, 5, 5]]])
I want:
b = tensor([[1, 2, 3, 4, 5, 6],
[4, 5, 6, 7, 8, 9],
[4, 4, 4, 5, 5, 5]])
Use permute first to change order of dimensions, then contiguous to prevent strides within the permuted tensor and finally use view to reshape the tensor.
b = a.permute(1,0,2).contiguous().view(a.shape[1],-1)
In order to concatenate tensors in pytorch, you can use the torch.cat function which concatenates tensors along a chosen axis. In this example, you can do:
a = torch.tensor([[[1, 2, 3], [4, 5, 6], [4, 4, 4]],
[[4, 5, 6], [7, 8, 9], [5, 5, 5]]])
b = torch.cat((a[0], a[1]), dim=1)
Out:
tensor([[1, 2, 3, 4, 5, 6],
[4, 5, 6, 7, 8, 9],
[4, 4, 4, 5, 5, 5]])
