PyTorch .logical_not()

In PyTorch, the .logical_not() function performs an element-wise logical negation on a tensor. It returns a tensor where each element is True if the corresponding input is False, and False otherwise. For non-boolean tensors, zeros are treated as False and non-zeros as True.

This function is commonly used in masking, boolean indexing, and creating complement conditions.

Syntax

torch.logical_not(input, *, out=None)

Parameters:

• input: The input tensor containing boolean or numeric values.
• out (Optional): A tensor to store the result. It must have the same shape as the output. The dtype is typically torch.bool, but integer types that can represent 0 and 1 (like torch.int16) are also supported. The results are stored as 1 for True and 0 for False.

Return value:

Returns a new tensor containing the element-wise logical negation of the input tensor.

Example

The following example demonstrates the use of .logical_not() for masking and boolean inversion:

import torch

# Create a boolean tensor
mask = torch.tensor([True, False, True, False])

# Element-wise logical NOT
inv = torch.logical_not(mask)

print('mask:', mask)
print('logical_not(mask):', inv)

# Use logical_not to invert a condition from a numeric tensor
vals = torch.tensor([0.0, 1.5, -2.0, 0.0])
cond = vals > 0
cond_inv = torch.logical_not(cond)

print('\nvals:', vals)
print('cond (vals > 0):', cond)
print('cond inverted with logical_not:', cond_inv)

# Numeric input — zeros are False, non-zeros are True
print(torch.logical_not(torch.tensor([0., 2.45, -20., 3.8], dtype=torch.double)))

# Using an integer out tensor (int16) to store 0/1 results
out_buf = torch.empty(4, dtype=torch.int16)
print(torch.logical_not(torch.tensor([0., 2.45, -20., 3.8], dtype=torch.double), out=out_buf))

Copy to clipboard
Copy to clipboard

The above code produces the following output:

mask: tensor([ True, False,  True, False])
logical_not(mask): tensor([False,  True, False,  True])

vals: tensor([ 0.0000,  1.5000, -2.0000,  0.0000])
cond (vals > 0): tensor([False,  True, False, False])
cond inverted with logical_not: tensor([ True, False,  True,  True])
tensor([ True, False, False, False])
tensor([1, 0, 0, 0], dtype=torch.int16)

Copy to clipboard
Copy to clipboard