`nn`

Components for constructing and processing GNN models

Modules

`activation`	Activation functions
`gcn`	Architectures for convolutional layers
`postprocess`	Postprocessing functions to convert a graph representation to a predicted property

Classes

`ActivationFunction`	Activation function options
`ComputePartialCharges`	Maps a set of atomic electronegativity and hardness parameters to partial charges.
`ConvolutionModule`
`DGLMoleculeDatasetEntry`	A named tuple containing a featurized molecule graph, a tensor of the atom features, and a tensor of the molecule label.
`DGLMoleculeDataLoader`
`DGLMoleculeDataset`
`PoolAtomFeatures`	A convenience class for pooling the node feature vectors produced by a graph convolutional layer.
`PoolBondFeatures`	A convenience class for pooling the node feature vectors produced by a graph convolutional layer into a set of symmetric bond (edge) features.
`ReadoutModule`	A module that transforms the node features generated by a series of graph convolutions via propagation through a pooling, readout and optional postprocess layer.
`RegularizedComputePartialCharges`	Maps a set of initial charges, atomic electronegativity, and hardness parameters to partial charges.
`SequentialLayers`