terminator.models.layers.gvp

Module for Geometric Vector Perceptrons. Sourced from https://github.com/drorlab/gvp-pytorch

Functions

randn(n, dims[, device])	Returns random tuples (s, V) drawn elementwise from a normal distribution.
tuple_cat(*args[, dim])	Concatenates any number of tuples (s, V) elementwise.
tuple_index(x, idx)	Indexes into a tuple (s, V) along the first dimension.
tuple_sum(*args)	Sums any number of tuples (s, V) elementwise.

Classes

Dropout(drop_rate)	Combined dropout for tuples (s, V).
GVP(in_dims, out_dims[, h_dim, activations, ...])	Geometric Vector Perceptron.
GVPConv(in_dims, out_dims, edge_dims[, ...])	Graph convolution / message passing with Geometric Vector Perceptrons.
GVPConvLayer(node_dims, edge_dims[, ...])	Full graph convolution / message passing layer with Geometric Vector Perceptrons.
LayerNorm(dims)	Combined LayerNorm for tuples (s, V).