espnet2.legacy.nets.pytorch_backend.rnn.encoders.RNNP

source

class espnet2.legacy.nets.pytorch_backend.rnn.encoders.RNNP(idim, elayers, cdim, hdim, subsample, dropout, typ='blstm')

Bases: Module

RNN with projection layer module.

• Parameters:
  • idim (int) – dimension of inputs
  • elayers (int) – number of encoder layers
  • cdim (int) – number of rnn units (resulted in cdim * 2 if bidirectional)
  • hdim (int) – number of projection units
  • subsample (np.ndarray) – list of subsampling numbers
  • dropout (float) – dropout rate
  • typ (str) – The RNN type

Initialize RNNP.

forward(xs_pad, ilens, prev_state=None)

Calculate RNNP forward propagation.

• Parameters:
  • xs_pad (torch.Tensor) – batch of padded input sequences (B, Tmax, idim)
  • ilens (torch.Tensor) – batch of lengths of input sequences (B)
  • prev_state (torch.Tensor) – batch of previous RNN states
• Returns: batch of hidden state sequences (B, Tmax, hdim)
• Return type: torch.Tensor