PyTorch torch.permute() rearranges the original tensor according to the desired ordering and returns a new multidimensional rotated tensor. The size of the returned tensor remains the same as that of the original.
Syntax: torch.permute(*dims)
Parameters:
- dims: sequence of indices in desired ordering of dimensions Of the tensor (indexing starts from zero).
Return: tensor with desired ordering of dimensions.
Let’s see this concept with the help of few examples:
Example 1: Create a two-dimensional tensor of size 2 × 4 and then permuted.
Python3
# import pytorch libraryimport torch # create a tensor of size 2 x 4input_var = torch.randn(2,4) # print sizeprint(input_var.size()) print(input_var) # dimensions permutedinput_var = input_var.permute(1, 0) # print sizeprint(input_var.size()) print(input_var) |
Output:
torch.Size([2, 4])
tensor([[ 0.9801, 0.5296, 0.5449, -1.1481],
[-0.6762, -0.1161, 0.6360, -0.5371]])
torch.Size([4, 2])
tensor([[ 0.9801, -0.6762],
[ 0.5296, -0.1161],
[ 0.5449, 0.6360],
[-1.1481, -0.5371]])
Example 2: Create a Three-dimensional tensor of size 3 × 5 × 2 and then permuted.
Python3
# import pytorch libraryimport torch # creating a tensor with random # values of dimension 3 X 5 X 2input_var = torch.randn(3, 5, 2) # print sizeprint(input_var.size()) print(input_var) # dimensions permutedinput_var = input_var.permute(2, 0, 1) # print sizeprint(input_var.size()) print(input_var) |
Output:
torch.Size([3, 5, 2])
tensor([[[ 0.2059, -0.7165],
[-1.1305, 0.5886],
[-0.1247, -0.4969],
[-0.5788, 0.0159],
[ 1.4304, 0.6014]],
[[ 2.4882, -0.3910],
[-0.5558, 0.6903],
[-0.4219, -0.5498],
[-0.5346, -0.0703],
[ 1.1497, -0.3252]],
[[-0.5075, 0.5752],
[ 1.3738, -0.3321],
[-0.3317, -0.9209],
[-1.6677, -1.1471],
[-0.9269, -0.6493]]])
torch.Size([2, 3, 5])
tensor([[[ 0.2059, -1.1305, -0.1247, -0.5788, 1.4304],
[ 2.4882, -0.5558, -0.4219, -0.5346, 1.1497],
[-0.5075, 1.3738, -0.3317, -1.6677, -0.9269]],
[[-0.7165, 0.5886, -0.4969, 0.0159, 0.6014],
[-0.3910, 0.6903, -0.5498, -0.0703, -0.3252],
[ 0.5752, -0.3321, -0.9209, -1.1471, -0.6493]]])
