build_conv_stack

build_conv_stack(*, input_channels: int, conv_filters: list[ConvFilter], activation: type[nn.Module]) -> nn.Sequential

Build the convolutional feature extractor for the DQN model.

Parameters:

• input_channels (int) – Number of input channels.
• conv_filters (list[ConvFilter]) – Sequence of (out_channels, kernel, stride) tuples.
• activation (type[Module]) – Activation module type to insert after each convolution.

Returns:

• Sequential – Convolutional stack as nn.Sequential.

Source