Any alternatives to `flat` for Tensor?

1

Numpy.ndarray have flat(), I know x.view(-1) can do so same result for the torch.Tensor, but I hope there is a good function can replace this unclear expression.

1 Like
2

I noticed that flat is not I want while it is an iterator actually, and flatten return copy, so the idea function is ravel. However Tensor still does not have any function called this.

3 
>>> a = torch.rand(2,3)
>>> a

 0.6496  0.8958  0.3040
 0.2913  0.1609  0.6008
[torch.FloatTensor of size 2x3]

>>> a_flat = a.view(a.numel())
>>> a_flat

 0.6496
 0.8958
 0.3040
 0.2913
 0.1609
 0.6008
[torch.FloatTensor of size 6]
5 Likes
4

You could also just write, using the default size:

>>>a_flat = a.view(-1)