emmi.modules.encoder.mlp_encoder¶

Classes¶

A 2-layer MLP encoder. Can be use to encode (i.e., embed) the input data.

class emmi.modules.encoder.mlp_encoder.MlpEncoder(input_dim, hidden_dim, init_weights='truncnormal002')¶

Bases: torch.nn.Module

A 2-layer MLP encoder. Can be use to encode (i.e., embed) the input data.

Initialize the MlpEncoder.

Parameters:

input_dim (int) – Number of dimensions of the input tensor.
hidden_dim (int) – Hidden dimensionality of the network.
init_weights – Initialization method for the weight matrixes of the linear layers. Defaults to “truncnormal002”.

forward(x)¶

Forward method of the MlpEncoder.