.dsplit()

In PyTorch, the .dsplit() function splits a given tensor with three or more dimensions into given number of sub-tensors depthwise or along the third axis.

Syntax

torch.dsplit(input, sections)

• input: The tensor to be split.
• sections: The number of sub-tensors that the input tensor is to be split into. The number should evenly divide the dimensions of the input tensor, otherwise it results in a runtime error.

Example

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

import torch

# Create a 3D tensor
ten = torch.tensor([[[1, 2, 3],
                     [4, 5, 6]],
                    [[7, 8, 9],
                     [8, 7, 6]]])

# Split the tensor into three sub-tensors
res = torch.dsplit(ten, 3)

print(res)

The above code produces the following output:

(tensor([[[1],
         [4]],

        [[7],
         [8]]]), tensor([[[2],
         [5]],

        [[8],
         [7]]]), tensor([[[3],
         [6]],

        [[9],
         [6]]]))