CODEWITHSUNDEEP
pythonpytorchtensorbackpropagation
szai
szai

Reputation: 13

Reduce dimensions of a tensor (to a scalar)

In:

a = torch.tensor([[2.4]])
torch.squeeze(a, 1)
a.size(), a

Out:

(torch.Size([1, 1]), tensor([[2.4000]]))

During computations using nn.MSELoss, I got a mismatch of dimensions. Input had size ([1,1]) and target ([]). The functions reshape and squeeze haven't worked. I would be grateful for a solution, to this embarassingly simple problem. : ]

Edit: there was a simple mistake of not assigning a= the squeezed value. Thank You for Your answer.

Upvotes: 1

Views: 514

Answers (1)

Ivan
Ivan

Reputation: 40628

Function torch.squeeze will not modify input a. Either reassign it:

a = a.squeeze(1)

or use the in-place version of the function torch.squeeze_

a.squeeze_(1)

Upvotes: 1

Related Questions