Graph

Module: generative_models.models.geometric.graph

Source: generative_models/models/geometric/graph.py

Overview

Graph model implementation with E(n) Equivariant Graph Neural Networks (EGNN).

This module implements a generic graph model with E(n) equivariance support, which is useful for generating and processing molecular structures, including proteins. EGNNLayer consumes node features already projected to the model's retained hidden_dim contract.

Classes

EGNNLayer

class EGNNLayer

GraphModel

class GraphModel

Functions

call

def __call__()

call

def __call__()

init

def __init__()

init

def __init__()

get_loss_fn

def get_loss_fn()

loss_fn

def loss_fn()

sample

def sample()

Module Statistics

• Classes: 2
• Functions: 7
• Imports: 6