pytorch - 💡(How to fix) Fix torch.softmax is not a true alias for torch.nn.functional.softmax [1 participants]

Traceback (most recent call last): File "<input>", line 1, in <module> TypeError: softmax() received an invalid combination of arguments - got (Tensor), but expected one of:

• (Tensor input, int dim, torch.dtype dtype = None, *, Tensor out = None)
• (Tensor input, name dim, *, torch.dtype dtype = None)

TL;DR

The dim argument should be made optional for torch.softmax to match the behavior of torch.nn.functional.softmax.

Guidance

• Verify the behavior of torch.nn.functional.softmax without the dim argument to confirm it is optional.
• Check the documentation for torch.softmax to see if there are any known issues or limitations related to the dim argument.
• Consider passing the dim argument to torch.softmax to resolve the immediate error, even if it should be optional.
• Review the implementation of torch.softmax to see if it is indeed an alias for torch.nn.functional.softmax and if the discrepancy is a bug.

Example

import torch

t = torch.zeros(5)
torch.softmax(t, dim=0)  # Pass dim argument to resolve error

Notes

The fix may require updating the torch.softmax function to make the dim argument optional, which could involve modifying the PyTorch library.

Recommendation

Apply workaround by passing the dim argument to torch.softmax until the discrepancy between torch.softmax and torch.nn.functional.softmax is resolved.