Name: squeeze() and unsqueeze() in PyTorch
Rating: 3.6 (852 reviews)
Author: hyperkai

squeeze() can get the 0D or more D tensor of the zero or more elements whose zero or more dimensions are removed if the size is 1 from the 0D or more D tensor of zero or more elements as shown below:

*Memos:

squeeze() can be used with torch or a tensor.
The 1st argument(input) with torch or using a tensor(Required-Type:tensor of int, float, complex or bool).
The 2nd argument with torch or the 1st or more arguments with a tensor are dim(Optional-Type:int, tuple of int or list of int): *Memos:
- Each number must be unique.
- It can remove the specific zero or more dimensions whose size is 1.
- If the size is not 1, zero or more dimensions are not removed even if you set it.

import torch

my_tensor = torch.tensor([[[[0], [1]],
                           [[2], [3]],
                           [[4], [5]]]])
torch.squeeze(input=my_tensor)
my_tensor.squeeze()
torch.squeeze(input=my_tensor, dim=(0, 3))
my_tensor.squeeze(dim=(0, 3))
my_tensor.squeeze(0, 3)
torch.squeeze(input=my_tensor, dim=(0, 1, 3))
my_tensor.squeeze(dim=(0, 1, 3))
my_tensor.squeeze(0, 1, 3)
etc.
torch.squeeze(input=my_tensor, dim=(0, 1, 2, 3))
my_tensor.squeeze(dim=(0, 1, 2, 3))
my_tensor.squeeze(0, 1, 2, 3)
etc.
# tensor([[0, 1],
#         [2, 3],
#         [4, 5]])

torch.squeeze(input=my_tensor, dim=0)
torch.squeeze(input=my_tensor, dim=-4)
torch.squeeze(input=my_tensor, dim=(0,))
torch.squeeze(input=my_tensor, dim=(-4,))
torch.squeeze(input=my_tensor, dim=(0, 1))
torch.squeeze(input=my_tensor, dim=(0, 2))
torch.squeeze(input=my_tensor, dim=(0, -2))
torch.squeeze(input=my_tensor, dim=(0, -3))
torch.squeeze(input=my_tensor, dim=(1, 0))
etc.
torch.squeeze(input=my_tensor, dim=(0, 1, 2))
etc.
# tensor([[[0], [1]],
#         [[2], [3]],
#         [[4], [5]]])

torch.squeeze(input=my_tensor, dim=1)
torch.squeeze(input=my_tensor, dim=2)
torch.squeeze(input=my_tensor, dim=-2)
torch.squeeze(input=my_tensor, dim=-3)
torch.squeeze(input=my_tensor, dim=())
torch.squeeze(input=my_tensor, dim=(1,))
torch.squeeze(input=my_tensor, dim=(2,))
torch.squeeze(input=my_tensor, dim=(-2,))
torch.squeeze(input=my_tensor, dim=(-3,))
torch.squeeze(input=my_tensor, dim=(1, 2))
etc.
# tensor([[[[0], [1]],
#          [[2], [3]],
#          [[4], [5]]]])

torch.squeeze(input=my_tensor, dim=3)
torch.squeeze(input=my_tensor, dim=-1)
torch.squeeze(input=my_tensor, dim=(3,))
torch.squeeze(input=my_tensor, dim=(-1,))
torch.squeeze(input=my_tensor, dim=(1, 3))
torch.squeeze(input=my_tensor, dim=(1, -1))
torch.squeeze(input=my_tensor, dim=(2, 3))
torch.squeeze(input=my_tensor, dim=(2, -1))
torch.squeeze(input=my_tensor, dim=(3, 1))
etc.
torch.squeeze(input=my_tensor, dim=(1, 2, 3))
etc.
# tensor([[[0, 1],
#          [2, 3],
#          [4, 5]]])

my_tensor = torch.tensor([[[[0.], [1.]],
                           [[2.], [3.]],
                           [[4.], [5.]]]])
torch.squeeze(input=my_tensor)
# tensor([[0., 1.],
#         [2., 3.],
#         [4., 5.]])

my_tensor = torch.tensor([[[[0.+0.j], [1.+0.j]],
                           [[2.+0.j], [3.+0.j]],
                           [[4.+0.j], [5.+0.j]]]])
torch.squeeze(input=my_tensor)
# tensor([[0.+0.j, 1.+0.j],
#         [2.+0.j, 3.+0.j],
#         [4.+0.j, 5.+0.j]])

my_tensor = torch.tensor([[[[True], [False]],
                           [[False], [True]],
                           [[True], [False]]]])
torch.squeeze(input=my_tensor)
# tensor([[True, False],
#         [False, True],
#         [True, False]])

unsqueeze() can get the 1D or more D tensor of zero or more elements with additional dimension whose size is 1 from the 0D or more D tensor of zero or more elements as shown below:

*Memos:

unsqueeze() can be used with torch or a tensor.
The 1st argument(input) with torch or using a tensor(Required-Type:tensor of int, float, complex or bool).
The 2nd argument with torch or the 1st argument with a tensor is dim(Required-Type:int). *It can add the dimension whose size is 1 to a specific position.

import torch

my_tensor = torch.tensor([[0, 1, 2],
                          [3, 4, 5],
                          [6, 7, 8],
                          [10, 11, 12]])
torch.unsqueeze(input=my_tensor, dim=0)
my_tensor.unsqueeze(dim=0)
torch.unsqueeze(input=my_tensor, dim=-3)
# tensor([[[0, 1, 2],
#          [3, 4, 5],
#          [6, 7, 8]
#          [10, 11, 12]]])

torch.unsqueeze(input=my_tensor, dim=1)
torch.unsqueeze(input=my_tensor, dim=-2)
# tensor([[[0, 1, 2]],
#         [[3, 4, 5]],
#         [[6, 7, 8]]
#         [[10, 11, 12]]])

torch.unsqueeze(input=my_tensor, dim=2)
torch.unsqueeze(input=my_tensor, dim=-1)
# tensor([[[0], [1], [2]],
#         [[3], [4], [5]],
#         [[6], [7], [8]],
#         [[10], [11], [12]]])

torch.unsqueeze(input=my_tensor, dim=3)
torch.unsqueeze(input=my_tensor, dim=-1)
# tensor([[[[0], [1], [2], [3]], [[4], [5], [6], [7]]],
#         [[[8], [9], [10], [11]], [[12], [13], [14], [15]]],
#         [[[16], [17], [18], [19]], [[20], [21], [22], [23]]]])

my_tensor = torch.tensor([[0., 1., 2.],
                          [3., 4., 5.],
                          [6., 7., 8.],
                          [10., 11., 12.]])
torch.unsqueeze(input=my_tensor, dim=0)
# tensor([[[0., 1., 2.],
#          [3., 4., 5.],
#          [6., 7., 8.],
#          [10., 11., 12.]]])

my_tensor = torch.tensor([[0.+0.j, 1.+0.j, 2.+0.j],
                          [3.+0.j, 4.+0.j, 5.+0.j],
                          [6.+0.j, 7.+0.j, 8.+0.j],
                          [10.+0.j, 11.+0.j, 12.+0.j]])
torch.unsqueeze(input=my_tensor, dim=0)
# tensor([[[0.+0.j, 1.+0.j, 2.+0.j],
#          [3.+0.j, 4.+0.j, 5.+0.j],
#          [6.+0.j, 7.+0.j, 8.+0.j],
#          [10.+0.j, 11.+0.j, 12.+0.j]]])

my_tensor = torch.tensor([[True, False, True],
                          [False, True, False],
                          [True, False, True],
                          [False, True, False]])
torch.unsqueeze(input=my_tensor, dim=0)
# tensor([[[True, False, True],
#          [False, True, False],
#          [True, False, True],
#          [False, True, False]]])