espnet2.legacy.nets.pytorch_backend.rnn.encoders.Encoder

source

class espnet2.legacy.nets.pytorch_backend.rnn.encoders.Encoder(etype, idim, elayers, eunits, eprojs, subsample, dropout, in_channel=1)

Bases: Module

Encoder module.

• Parameters:
  • etype (str) – type of encoder network
  • idim (int) – number of dimensions of encoder network
  • elayers (int) – number of layers of encoder network
  • eunits (int) – number of lstm units of encoder network
  • eprojs (int) – number of projection units of encoder network
  • subsample (np.ndarray) – list of subsampling numbers
  • dropout (float) – dropout rate
  • in_channel (int) – number of input channels

Initialize Encoder.

forward(xs_pad, ilens, prev_states=None)

Forward Encoder.

• Parameters:
  • xs_pad (torch.Tensor) – batch of padded input sequences (B, Tmax, D)
  • ilens (torch.Tensor) – batch of lengths of input sequences (B)
  • prev_state (torch.Tensor) – batch of previous encoder hidden states (?, …)
• Returns: batch of hidden state sequences (B, Tmax, eprojs)
• Return type: torch.Tensor