SelfAttn

class aixd.mlmodel.architecture.blocks.SelfAttn(*args: Any, **kwargs: Any)[source]

Bases: Module

Module implementing a self-attention layer.

Parameters:
  • dim (int) – The dimensionality of the input data (1, 2, or 3).

  • in_channels (int) – The number of input channels.

  • eps (float, optional, default=1e-12) – The epsilon value for the spectral normalization.

Methods