PyTorch .slice_scatter()

In PyTorch, the .slice_scatter() function inserts all values from the source tensor into the input tensor at the given dimension. It returns a new tensor with fresh storage, rather than creating a view.

Syntax

torch.slice_scatter(input, src, dim=0, start=None, end=None, step=1)

• input: The input tensor.
• src: The source tensor containing the values to insert into the input tensor.
• dim: The dimension along which the values are to be inserted. The default value is 0.
• start (Optional): The starting index for inserting the values. The default value is None.
• end (Optional): The ending index for inserting the values. The default value is None.
• step: The number of elements to skip while inserting the values. The default value is 1.

Example

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

import torch

# Create a 4x4 input tensor with all elements set to '0'
input = torch.zeros(4, 4)

# Create a 2x4 source tensor with all elements set to '1'
src = torch.ones(2, 4)

# Insert the values along dimension 0 in 'input'
res = torch.slice_scatter(input, src, 0, start=2)

# Print the resultant tensor
print(res)

Copy to clipboard
Copy to clipboard

The above code produces the following output:

tensor([[0., 0., 0., 0.],
        [0., 0., 0., 0.],
        [1., 1., 1., 1.],
        [1., 1., 1., 1.]])

Copy to clipboard
Copy to clipboard