Torch Geometric skill: what it does and how to install it

What it does. Provides patterns for installing and using PyTorch Geometric, modelling graph data with Data and HeteroData, selecting GNN layers, training prediction models and creating custom message-passing layers. It also covers datasets, transforms, link prediction, neighbour sampling, explainability and scaling.

When to use it. Use it for PyG-based graph neural network work, including node, link and graph tasks, heterogeneous graphs and custom datasets. It does not apply to general NetworkX analytics or non-graph PyTorch models.