Pytorch Geometric is a library for Graph Neural Networks (GNNs) and builds upon PyTorch. It contains various methods for writing and training GNNs on graphs from a variety of published papers. It supports mini-batch loaders for operation on GPUs.
Pytorch Geometric(PyG) is a library for Graph Neural Networks (GNNs) and builds upon PyTorch. It contains various methods for writing and training GNNs on graphs from a variety of published papers. It supports mini-batch loaders for operation on GPUs.
Features
- Supports a variety of GNNs, see cheatsheet or the list of implemented GNN models many with small examples, such as Graph Attention Networks (GATs), Graph Convolutional Networks
- Supports a variety of graphs: sparse graphs, graphs with edge weights, graphs with edge attributes, bipartite graphs, etc.
- Contains the most well-known datasets, such as Cora or Zachary’s karate club, as well as being supported by many other such as the Open Graph Benchmark
- Supports the
GraphGymAPI to design and evaluate GNNs
