quantlab.model.predefined.tcn.TemporalBlock

class quantlab.model.predefined.tcn.TemporalBlock(n_inputs, n_outputs, kernel_size, stride, dilation, padding, dropout=0.2)

Bases: Module

Two weight-normalised dilated causal convolutions with ReLU, dropout and a residual.

Examples

>>> block = TemporalBlock(4, 8, kernel_size=3, stride=1, dilation=2, padding=4)
>>> block(torch.randn(2, 4, 10)).shape
torch.Size([2, 8, 10])

forward(x)

Map [B, C_in, N] to [B, C_out, N].