.abs()

The .abs() method in PyTorch computes the absolute value of each tensor element. For real numbers, it returns the non-negative value. For complex numbers, it calculates the magnitude using √(real² + imag²). This method is useful in data preprocessing, signal processing, and mathematical transformations.

Syntax

torch.abs(input)

• input (Tensor): The input tensor (can be real or complex).

The .abs() method returns a tensor where each element is the absolute value of the matching element in the input tensor.

Example

This example shows how to use the .abs() method to compute absolute values for both real and complex tensors:

import torch

# Define a tensor with real and complex numbers
tensor = torch.tensor([[-3.0, 2.5], [1 - 2j, 3 + 4j]])

# Compute absolute values
abs_tensor = tensor.abs()

print("Original Tensor:")
print(tensor)

print("\nAbsolute Values:")
print(abs_tensor)

This example results in the following output:

Original Tensor:
tensor([[-3.0000+0.j,  2.5000+0.j],
        [1.0000-2.j, 3.0000+4.j]])

Absolute Values:
tensor([[3.0000, 2.5000],
        [2.2361, 5.0000]])

In this example:

• Real numbers: -3.0 becomes 3.0, 2.5 remains 2.5.
• Complex numbers: 1 - 2j → √(1² + (-2)²) ≈ 2.2361; 3 + 4j → √(3² + 4²) = 5.0