DIG

Why we included this project

Graph deep learning research often means juggling separate paper repositories that each have their own data loaders and metrics. DIG pulls the common pieces into one library: consistent data interfaces, reference implementations, and evaluation metrics for graph generation, GNN explainability, 3D geometry learning, and self-supervised graph pre-training. It builds on PyTorch and PyG, so it slots into the stack most graph labs already use, and the bundled benchmarks, examples, and tutorials help you get a pipeline running without reinventing infrastructure. For a lab or small team that wants one shared starting point for prototyping graph-learning ideas, this is a practical hub.