core.nn.enets.DenseResidualNet

core.nn.enets.DenseResidualNet(
    input_dim,
    output_dim,
    hidden_dims,
    activation=F.elu,
    dropout=0.0,
    batch_norm=True,
    context_features=None,
)

A nn.Module consisting of a sequence of dense residual blocks. This is used to embed high dimensional input to a compressed output. Linear resizing layers are used for resizing the input and output to match the first and last hidden dimension, respectively.

Module specs

input dimension:    (batch_size, input_dim)
output dimension:   (batch_size, output_dim)

Parameters

Name	Type	Description	Default
input_dim	int	dimension of the input to this module	required
output_dim	int	output dimension of this module	required
hidden_dims	tuple	tuple with dimensions of hidden layers of this module	required
activation	Callable	activation function used in residual blocks	F.elu
dropout	float	dropout probability for residual blocks used for reqularization	0.0
batch_norm	bool	flag that specifies whether to use batch normalization	True
context_features	int	Number of additional context features, which are provided to the residual blocks via gated linear units. If None, no additional context expected.	None

Attributes

Name	Description
blocks
hidden_dims
initial_layer
input_dim
num_res_blocks
output_dim
resize_layers

Methods

Name	Description
forward

forward

core.nn.enets.DenseResidualNet.forward(x, context=None)