PyTorch .size()

The size() method in PyTorch returns a torch.Size object containing the size (shape) information of a tensor. It serves as a fundamental function for dynamically obtaining the tensor’s shape during operations. Specific dimensions can be accessed by indexing into the torch.Size object, which functions like a tuple.

Syntax

tensor.size(dim=None)

• tensor: The PyTorch tensor on which the .size() method is called.
• dim (Optional): Specifies the dimension for which to retrieve the size. The default value is None.
  • If dim is not provided, the returned value is a torch.Size object representing the size of all dimensions.
  • If dim is specified, the returned value is an int representing the size of the given dimension.

Example

The following example shows how to use the .size() method:

import torch

t = torch.empty(3, 4, 5)

print(t.size())
print(t.size(dim=1))

Copy to clipboard
Copy to clipboard

The code above generates the following output:

torch.Size([3, 4, 5])
4

Copy to clipboard
Copy to clipboard