PyTorch .lerp()

In PyTorch, the .lerp() function computes the linear interpolation between an input tensor (input) and an end tensor (end), using a scalar or tensor weight. This is mathematically equivalent to applying the function $out_i = start_i + weight_i * (end_i - start_i)$.

The shapes of input, end, and weight must be broadcastable.

Syntax

torch.lerp(input, end, weight, *, out=None)

Parameters:

• input: The input tensor containing the initial points.
• end: The ending tensor containing the finishing points.
• weight: The shapes of input, end, and weight must be broadcastable.
• out (optional): A tensor to store the output. If provided, the result is written to this tensor.

Return value:

Returns a new tensor containing the result given by the interpolation formula.

Example

The following example shows how to compute the interpolation between two tensors using torch.lerp() with a float scalar weight:

import torch
import math

# Define two tensors
start = torch.tensor([12.0 , 14.0 , 16.0 , math.log(2.)])
end = torch.tensor([11.0 , 13.0 , 15.0 , math.log(2.)])

# Compute the interpolation with a float weight
out = torch.lerp(start, end, 0.8)

print(out)

Copy to clipboard
Copy to clipboard

Here is the output:

tensor([11.2000, 13.2000, 15.2000,  0.6931])

Copy to clipboard
Copy to clipboard