PyTorch .arange()

The .arange() method returns a tensor containing values from a given interval [start, end) with a specified step size. When the step size is not an integer, floating-point rounding errors may occur, so it is recommended to subtract a small epsilon from the end value for consistency.

Syntax

torch.arange(start=0, end, step=1, out=None, dtype=None, layout=torch.strided, device=None, requires_grad=False)

The parameters are as follows:

• start: The starting value of the range, inclusive. Defaults to 0.
• end: The ending value of the range, exclusive. This parameter is required.
• step: The difference between each consecutive value in the range. The default value is 1.
• out: A tensor in which to store the output. If None, a new tensor is created. The default value is None.
• dtype: The desired data type of the output tensor (torch.dtype). If None, the data type will be inferred from other input arguments. The default value is None.
• layout: The desired layout of the output tensor. Default: torch.strided.
• device: The device on which the tensor will be allocated (torch.device). Default: None.
• requires_grad: A boolean indicating whether autograd should track operations on the output tensor. Default: False.

Example

import torch

# Return a tensor with only an end value
t0 = torch.arange(3)

# Return a tensor of a specified range and step count
t1 = torch.arange(5, 35, 10)

print(t0)
print(t1)

Copy to clipboard
Copy to clipboard

The returned tensors are as follows:

tensor([0, 1, 2])
tensor([5, 15, 25])

Copy to clipboard
Copy to clipboard

Codebyte Example

Run the following code to know how the .arange() method works:

Visit us
Visit us
Hide code
Code
Output
Hide output
Copy to your clipboard
Copy Codebyte
Run