This repository aims to provide a comprehensive tutorial on implementing graph neural networks (GNN) and their related pipelines using Python, particularly with PyTorch.
In these tutorials, we cover the construction of graphs, building graph neural network layers, stacking them together to create graph neural networks, and training them.
The tutorials utilize two well-known graph neural network libraries: Deep Graph Library (dgl) and PyTorch Geometric (pyg). Both dgl and pyg exhibit their respective strengths. For instance, dgl offers more handy functions to modulate message-passing routes, while pyg tends to be faster than dgl when executing the same logic. Therefore, choosing a GNN library depends on the specific requirements of your project.
Each tutorial notebook for these libraries can be found in their respective directories.
The DGL tutorials were last updated a few years ago, so some of the features may have been deprecated.