class maze.perception.blocks.feed_forward.strided_conv.StridedConvolutionBlock(*args: Any, **kwargs: Any)

A block containing multiple subsequent strided convolution layers.

One layer consists of a single strided convolution followed by an activation function. The block expects the input tensors to have the from (batch-dim, channel-dim, row-dim, column-dim).

  • in_keys – One key identifying the input tensors.

  • out_keys – One key identifying the output tensors.

  • in_shapes – List of input shapes.

  • hidden_channels – List containing the number of hidden channels for hidden layers.

  • hidden_kernels – List containing the size of the convolving kernels.

  • non_lin – The non-linearity to apply after each layer.

  • convolution_dimension – Dimension of the convolution to use [1, 2, 3]

  • hidden_strides – List containing the strides of the convolutions.

  • hidden_dilations – List containing the spacing between kernel elements.

  • hidden_padding – List containing the padding added to both sides of the input

  • padding_mode – ‘zeros’, ‘reflect’, ‘replicate’ or ‘circular’.


Compiles a block-specific dictionary of network layers. This could be overwritten by derived layers (e.g. to get a ‘BatchNormalizedConvolutionBlock’).


Ordered dictionary of torch modules [str, nn.Module]

normalized_forward(block_input: Dict[str, torch.Tensor]) → Dict[str, torch.Tensor]

(overrides ShapeNormalizationBlock)

implementation of ShapeNormalizationBlock interface