xma.functional.softmax

softmax(x: Tensor, logits_multiplier: float | None = None, *, kernel_backend: KernelBackend | None = None) → Tensor[source]

computes softmax activation

Parameters:

  • x (torch.Tensor) – input activation tensor

  • logits_multiplier (float | None) – pre-multiplies x with logits_multiplier before computing softmax. Defaults to None.

  • kernel_backend (KernelBackend | None) – KernelBackend

Returns:

output tensor

Return type:

Tensor