espnet.nets.pytorch_backend.transducer.conv1d_nets.CausalConv1d

source

class espnet.nets.pytorch_backend.transducer.conv1d_nets.CausalConv1d(idim: int, odim: int, kernel_size: int, stride: int = 1, dilation: int = 1, groups: int = 1, bias: bool = True, batch_norm: bool = False, relu: bool = True, dropout_rate: float = 0.0)

Bases: Module

1D causal convolution module for custom decoder.

• Parameters:
  • idim – Input dimension.
  • odim – Output dimension.
  • kernel_size – Size of the convolving kernel.
  • stride – Stride of the convolution.
  • dilation – Spacing between the kernel points.
  • groups – Number of blocked connections from input channels to output channels.
  • bias – Whether to add a learnable bias to the output.
  • batch_norm – Whether to apply batch normalization.
  • relu – Whether to pass final output through ReLU activation.
  • dropout_rate – Dropout rate.

Construct a CausalConv1d object.

forward(sequence: Tensor, mask: Tensor, cache: Tensor | None = None) → Tuple[Tensor, Tensor]

Forward CausalConv1d for custom decoder.

• Parameters:
  • sequence – CausalConv1d input sequences. (B, U, D_in)
  • mask – Mask of CausalConv1d input sequences. (B, 1, U)
• Returns: CausalConv1d output sequences. (B, sub(U), D_out) mask: Mask of CausalConv1d output sequences. (B, 1, sub(U))
• Return type: sequence