pytorch - 💡(How to fix) Fix Feedback about torch.argsort [1 participants]

TL;DR

The example output for torch.argsort(a, dim=1) in the PyTorch documentation is incorrect and should be updated to reflect the correct sorted indices.

Guidance

• Verify the correctness of the example by running the code snippet torch.argsort(a, dim=1) with the provided tensor a and comparing the output to the expected result.
• Check the PyTorch version being used to ensure it is the latest stable version, as the behavior of torch.argsort may have changed in recent updates.
• If the issue persists, consider filing a bug report or submitting a pull request to update the documentation with the correct example output.
• Review the documentation for torch.argsort to ensure understanding of its behavior and parameters, particularly the dim argument.

Example

import torch

a = torch.tensor([[ 0.0785,  1.5267, -0.8521,  0.4065],
                  [ 0.1598,  0.0788, -0.0745, -1.2700],
                  [ 1.2208,  1.0722, -0.7064,  1.2564],
                  [ 0.0669, -0.2318, -0.8229, -0.9280]])

print(torch.argsort(a, dim=1))

Notes

The provided example output seems to be incorrect, and the correct output should be tensor([[1, 3, 0, 2], [3, 2, 1, 0], [2, 1, 0, 3], [3, 2, 1, 0]]). However, without further information about the PyTorch version being used, it is difficult to determine the root cause of the issue.

Recommendation

Apply workaround: Update the documentation with the correct example output, as the provided output seems to be incorrect.