dilation_convolution_pytorch.py

from torch import nn 

def CausalConv1d(in_channels, out_channels, kernel_size, dilation=1, **kwargs):
	pad = (kernel_size - 1) * dilation + 1
	return nn.Conv1d(in_channels, out_channels, kernel_size, padding = pad, dilation = dilation, **kwargs)

def CasalConv2d(in_channels, out_channels, kernel_size, dilation=1, **kwargs):
	pad = (kernel_size - 1) * dilation + 1
	return nn.Conv2d(in_channels, out_channels, kernel_size, padding = pad, dilation = dilation, **kwargs)

def CasalConv3d(in_channels, out_channels, kernel_size, dilation=1, **kwargs):
	pad = (kernel_size - 1) * dilation + 1
	return nn.Conv3d(in_channels, out_channels, kernel_size, padding = pad, dilation = dilation, **kwargs)

## thus, the dilation convolution / causal convolution (1d) is shown
m = CausalConv1d(1, 1, kernel_size=3, dilation=2, bias=False)