espnet2.asr.decoder.linear_decoder.LinearDecoder

source

class espnet2.asr.decoder.linear_decoder.LinearDecoder(vocab_size: int, encoder_output_size: int, pooling: str = 'mean', dropout: float = 0.0)

Bases: AbsDecoder

Initialize the module.

forward(hs_pad: Tensor, hlens: Tensor, ys_in_pad: Tensor = None, ys_in_lens: Tensor = None) → Tuple[Tensor, Tensor]

Forward method.

• Parameters:
  • hs_pad – (B, Tmax, D)
  • hlens – (B,)
• Returns: (B, n_classes)
• Return type: output

output_size() → int

Get the output size.

score(ys, state, x)

Classify x.

• Parameters:
  • ys – Not used
  • state – Not used
  • x – (T, D). this should be a single sample without any padding ie batch size=1.
• Returns: log probabilities over (n_classes,) state: None
• Return type: logp

Assumes that x is a single unpadded sequence.