espnet2.asr.encoder.beats_encoder.Swish

source

class espnet2.asr.encoder.beats_encoder.Swish

Bases: Module

Swish activation function.

The Swish activation function is defined as: : Swish(x) = x * sigmoid(x)

This activation function has been shown to outperform ReLU in some deep learning applications.

act

The sigmoid activation function instance.

• Type: torch.nn.Sigmoid

forward(x

torch.Tensor) -> torch.Tensor: Computes the Swish activation for the input tensor.

####### Examples

>>> swish = Swish()
>>> input_tensor = torch.tensor([-1.0, 0.0, 1.0])
>>> output_tensor = swish(input_tensor)
>>> print(output_tensor)
tensor([-0.2689, 0.0000, 0.7311])

Initialize internal Module state, shared by both nn.Module and ScriptModule.

#

forward(x

)

Forward pass for the Beats encoder.

This method processes the input tensor xs_pad and its corresponding lengths ilens to produce audio representations. It is designed to be compatible with the AbsEncoder interface in ESPnet.

• Parameters:

  • xs_pad (torch.Tensor) – Input tensor of shape (B, T, D), where B is the batch size, T is the sequence length, and D is the feature dimension.
  • ilens (torch.Tensor) – Tensor of shape (B,) representing the actual lengths of the sequences in xs_pad.
  • prev_states (torch.Tensor , optional) – Optional tensor representing previous states. Default is None.

• Returns:

  • audio_representation (torch.Tensor): The output audio

  representation of shape (B, T, D).

  • output_lens (torch.Tensor): Tensor of shape (B,) containing the output lengths for each sequence.
  • masks (Optional[torch.Tensor]): Currently set to None.

• Return type: Tuple[torch.Tensor, torch.Tensor, Optional[torch.Tensor]]

NOTE

If xs_pad is not provided, this operation may be costly, as it attempts to create a tensor of size maxlen x maxlen. Thus, tensors are unsqueezed and then squeezed to optimize performance.

####### Examples

>>> beats_encoder = BeatsEncoder(...)
>>> xs_pad = torch.randn(32, 100, 512)  # Batch of 32, 100 time steps
>>> ilens = torch.randint(1, 101, (32,))  # Random lengths
>>> audio_representation, output_lens, masks = beats_encoder.forward(xs_pad, ilens)

• Raises:ValueError – If xs_pad or ilens have incompatible shapes.