espnet.nets.pytorch_backend.transformer.subsampling.Conv2dSubsampling

source

class espnet.nets.pytorch_backend.transformer.subsampling.Conv2dSubsampling(idim, odim, dropout_rate, pos_enc=None)

Bases: Module

Convolutional 2D subsampling (to 1/4 length).

• Parameters:
  • idim (int) – Input dimension.
  • odim (int) – Output dimension.
  • dropout_rate (float) – Dropout rate.
  • pos_enc (torch.nn.Module) – Custom position encoding layer.

Construct an Conv2dSubsampling object.

forward(x, x_mask)

Subsample x.

• Parameters:

  • x (torch.Tensor) – Input tensor (#batch, time, idim).
  • x_mask (torch.Tensor) – Input mask (#batch, 1, time).

• Returns: Subsampled tensor (#batch, time’, odim), : where time’ = time // 4.

  torch.Tensor: Subsampled mask (#batch, 1, time’), : where time’ = time // 4.

• Return type: torch.Tensor