展开 Pytorch 张量

Question

The tensor a in shape of torch.Size([2, 2, 1, 2]) how to expand this tensor to the shape of torch.Size([2, 4, 1, 2]) without using for loop which takes time

For example, I have tensor a

>>> print(a, a.shape)
tensor([[[[0.2955, 0.8836]],

         [[0.7607, 0.6657]]],


        [[[0.6779, 0.5109]],

         [[0.0785, 0.6564]]]]) torch.Size([2, 2, 1, 2])

I want to expand it to become tensor b

>>> print(b, b.shape)
tensor([[[[0.2955, 0.8836]],

         [[0.2955, 0.8836]],

         [[0.7607, 0.6657]],

         [[0.7607, 0.6657]]],


        [[[0.6779, 0.5109]],

         [[0.6779, 0.5109]],

         [[0.0785, 0.6564]],

         [[0.0785, 0.6564]]]]) torch.Size([2, 4, 1, 2])

I tried torch.expand() but it only expands when the dimension=1. How to achieve this format? Thank you

Answer 1

Pytorch provides multiple functions with different styles to repeat the tensors. The function that you are looking for is named repeat_interleave . Simple example:

>>> a.repeat_interleave(2,dim=1)
tensor([[[[0.2955, 0.8836]],

         [[0.2955, 0.8836]],

         [[0.7607, 0.6657]],

         [[0.7607, 0.6657]]],


        [[[0.6779, 0.5109]],

         [[0.6779, 0.5109]],

         [[0.0785, 0.6564]],

         [[0.0785, 0.6564]]]])