.column_stack()

In PyTorch, the .column_stack() function creates a new tensor by stacking tensors as columns in a 2D tensor, converting 1D tensors into 2D if necessary. It is useful for combining feature vectors or aligning data along the column axis for machine learning models.

Syntax

torch.column_stack(tensors, *, out=None)

• tensors: A sequence of tensors to be stacked, with 1D tensors converted into columns before stacking.
• out (Optional): A tensor to store the output, matching the shape of the resulting stacked tensor.

Example

In this example, .column_stack() stacks the 1D tensors a and b as columns to create a 2D tensor:

import torch

# Define 1D tensors
a = torch.tensor([10, 20, 30])
b = torch.tensor([40, 50, 60])

# Stack the tensors as columns
result = torch.column_stack((a, b))

print(result)

Here is the output for the above code:

tensor([[10, 40],
        [20, 50],
        [30, 60]])