Skip to content
greyhound greyhound

Module Wrappers

Generated from Python docstrings by scripts/generate_api_docs.py.

GreyhoundCausalConv1d

Section titled “GreyhoundCausalConv1d”

greyhound.nn.causal_conv1d.GreyhoundCausalConv1d

GreyhoundCausalConv1d(channels: int, kernel_size: int, bias: bool = True, activation: str | None = None, device: torch.device | str | None = None, dtype: torch.dtype | None = None)

Causal 1D depthwise convolution with optional SiLU activation.

Wraps the fused Greyhound kernel as a drop-in replacement for depthwise nn.Conv1d with causal (left) padding. The kernel handles boundary conditions internally via masked loads, avoiding the cost of explicit F.pad().

Inherits from nn.Conv1d with groups=channels (depthwise), stride=1, dilation=1, and padding=0 (causal padding is applied inside the kernel).

Parameters

  • channels (int): Number of input (and output) channels.
  • kernel_size (int): Width of the convolution kernel (typically 2-4).
  • bias (bool, default: True): If True, adds a learnable bias. Default: True.
  • activation (str | None, default: None): Activation to apply after convolution. "silu" for SiLU activation, or None for no activation. Default: None.
  • device (torch.device | str | None, default: None): Device for parameter initialization.
  • dtype (torch.dtype | None, default: None): Data type for parameters.

forward

Section titled “forward”
forward(input: torch.Tensor) -> torch.Tensor

Apply causal depthwise convolution.

Parameters

  • input (torch.Tensor): Input tensor of shape [B, D, T].

Returns

torch.Tensor

Output tensor of shape [B, D, T].