Squeeze, unsqueeze explanation

nlp
1

When I went through documentation for squeeze, unsqueeze methods, I felt like I got this clearly and after few days when I start using squeeze and unsqueeze, I feel that I am not totally clear on these two methods.

can anyone shed some light on these two methods please!

2

unsqueeze to add dimension of size 1,
squeeze to remove dimension of size 1.

for example,

import torch
x = torch.randn(10); x

tensor([-0.9048, -0.3753, -0.3185, -1.4873, -0.7718,  0.6153,  0.1261,  0.5256,
        -1.9678, -0.2045])

x.shape

torch.Size([10])

y = x.unsqueeze(dim=0)
y.shape

torch.Size([1, 10])

y = x.unsqueeze(dim=-1)
y.shape

torch.Size([10, 1])

one more thing,

y = x[None] # or y = x[None, :], same as x.unsqueeze(dim=0)

and

y = x[:, None] # same as x.unsqueeze(dim=-1)

for squeeze, if

y.shape

is

torch.Size([10, 1])

then,

y.squeeze_(dim=1)
y.shape

give

torch.Size([10])