ml-pytorch-geometric

What problem does it solve?

This Skill helps you learn and apply PyTorch Geometric (PyG) to build Graph Neural Networks (GNNs) for real graph data problems without getting stuck on the framework’s data model and training patterns.

Core Features & Use Cases

• Graph data modeling: Convert your dataset into PyG Data objects (node features, edge_index, targets) and handle batching via batch.
• Message passing & GNN layers: Implement and reason about the message/aggregate/update flow and choose common layers like GCNConv, GATConv, and SAGEConv.
• Scalable training workflows: Use neighbor sampling (NeighborLoader) and distributed/out-of-core concepts for large graphs.
• Task coverage: Support node classification, graph classification, heterogeneous graphs (HeteroData), and explainability patterns.
• Lightning integration: Use Lightning-compatible dataset wrappers such as LightningDataset and LightningNodeData to streamline training.

Quick Start

Tell the AI: "Show me how to structure a node-classification GNN in PyTorch Geometric with Lightning, including how to prepare edge_index, masks, and a neighbor-sampling datamodule."