espnet.nets.pytorch_backend.ctc.CTC

source

class espnet.nets.pytorch_backend.ctc.CTC(odim, eprojs, dropout_rate, ctc_type='builtin', reduce=True)

Bases: Module

CTC module

• Parameters:
  • odim (int) – dimension of outputs
  • eprojs (int) – number of encoder projection units
  • dropout_rate (float) – dropout rate (0.0 ~ 1.0)
  • ctc_type (str) – builtin
  • reduce (bool) – reduce the CTC loss into a scalar

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

argmax(hs_pad)

argmax of frame activations

• Parameters:hs_pad (torch.Tensor) – 3d tensor (B, Tmax, eprojs)
• Returns: argmax applied 2d tensor (B, Tmax)
• Return type: torch.Tensor

forced_align(h, y, blank_id=0)

forced alignment.

• Parameters:
  • h (torch.Tensor) – hidden state sequence, 2d tensor (T, D)
  • y (int) – id sequence tensor 1d tensor (L)
  • y – blank symbol index
• Returns: best alignment results
• Return type: list

forward(hs_pad, hlens, ys_pad)

CTC forward

• Parameters:
  • hs_pad (torch.Tensor) – batch of padded hidden state sequences (B, Tmax, D)
  • hlens (torch.Tensor) – batch of lengths of hidden state sequences (B)
  • ys_pad (torch.Tensor) – batch of padded character id sequence tensor (B, Lmax)
• Returns: ctc loss value
• Return type: torch.Tensor

log_softmax(hs_pad)

log_softmax of frame activations

• Parameters:hs_pad (torch.Tensor) – 3d tensor (B, Tmax, eprojs)
• Returns: log softmax applied 3d tensor (B, Tmax, odim)
• Return type: torch.Tensor

loss_fn(th_pred, th_target, th_ilen, th_olen)

softmax(hs_pad)

softmax of frame activations

• Parameters:hs_pad (torch.Tensor) – 3d tensor (B, Tmax, eprojs)
• Returns: log softmax applied 3d tensor (B, Tmax, odim)
• Return type: torch.Tensor