.logspace()

The .logspace() function returns a one-dimensional tensor with values logarithmically spaced.

The function is useful for generating logarithmically spaced values for various applications, such as plotting data on a logarithmic scale or creating logarithmic scales for neural network hyperparameters.

Syntax

torch.logspace(start, end, steps, base, dtype=None)

• start: The first number in the range expressed as a logarithm.
• end: The last number in the range expressed as a logarithm.
• steps: Number of elements to be returned in the tensor.
• base: The base of the logarithm used for calculating the values default value is 10.
• dtype: Specifies the data type of the returned tensor.

Example 1

In this example, the code generates a tensor containing 5 logarithmically spaced values between 1 and 1000:

import torch

# Generate a tensor with 5 logarithmically spaced values between 1 and 1000
tensor = torch.logspace(0, 3, steps=5)
print(tensor)

The code above generates the following output:

tensor([   1.0000,    5.6234,   31.6228,  177.8279, 1000.0000])

Example 2

In this example, the code generates a tensor containing 3 logarithmically spaced values between 1 and 10 using the .logspace() function:

import torch

# Generate a tensor with 3 logarithmically spaced values between 0 and 10
tensor = torch.logspace(0, 1, steps=3, dtype=torch.float64)
print(tensor)

Output:

tensor([1.0000, 3.1623, 10.0000], dtype=torch.float64)

In this example, we created a tensor tensor containing 3 logarithmically spaced values between 0 and 1 using the .logspace() function with a data type of torch.float64. The tensor tensor contains the values [1.0000, 3.1623, 10.0000].