PyTorch .permute()

In PyTorch, the .permute() function returns a view of a given tensor with its dimensions permuted or rearranged according to a specific order.

Syntax

torch.permute(input, dims)

• input: The tensor whose dimensions are to be permuted.
• dims: The order in which the dimensions are to be permuted.

Example

The following example demonstrates the usage of the .permute() function:

import torch

# Create a tensor of size (2, 3, 4)
ten = torch.randn(2, 3, 4)

# Permute the dimensions of the tensor in the order (2, 0, 1)
res = torch.permute(ten, (2, 0, 1))

# Print the size of the resultant tensor
print(res.size())

Copy to clipboard
Copy to clipboard

In the above example, the order (2, 0, 1):

• Moves the dimension located at index 2 to index 0
• Moves the dimension located at index 0 to index 1
• Moves the dimension located at index 1 to index 2

The above code produces the following output:

torch.Size([4, 2, 3])

Copy to clipboard
Copy to clipboard