What is PyTorch Geometric (PyG)?

PyTorch Geometric is a specialized library built on PyTorch specifically designed for deep learning on graphs and other irregular data structures.

Does this skill support large-scale graph training?

Yes, the skill includes patterns for neighbor sampling, mini-batch processing, and distributed training to handle datasets that exceed memory limits.

Which benchmark datasets are supported?

The skill provides guidance for loading common benchmarks like Planetoid (Cora, CiteSeer), TUDataset (ENZYMES), and molecular datasets like QM9.

Can I create custom graph convolution layers?

Absolutely. This skill provides implementation patterns for inheriting from the MessagePassing class, allowing you to define custom message, aggregate, and update functions.

PyTorch Geometric (PyG)

Name: PyTorch Geometric (PyG)
Author: Sologa

bySologa

Data Science & ML

Develops and trains Graph Neural Networks (GNNs) using the PyTorch Geometric library for irregular data structures and geometric deep learning.

About

The PyTorch Geometric (PyG) skill empowers Claude to architect, implement, and optimize Graph Neural Networks for complex data structures. It provides specialized patterns for node and graph classification, link prediction, and molecular property prediction. By leveraging this skill, developers can efficiently handle irregular data like social networks, citation graphs, and 3D point clouds using industry-standard architectures like GCN, GAT, and GraphSAGE, complete with best practices for mini-batch processing and large-scale graph learning.

Key Features

Support for heterogeneous graphs with multi-type nodes and edges
0 GitHub stars
Implementation of state-of-the-art GNN layers including GCN, GAT, and GraphSAGE
Efficient mini-batch processing and handling of large-scale graph datasets
Standardized patterns for molecular property prediction and chemical informatics
Custom MessagePassing interface for developing bespoke graph convolution layers

Use Cases

Building recommendation systems and social network analysis tools
Developing 3D geometric deep learning models for point clouds and meshes
Predicting molecular properties for drug discovery and chemical research

About

Key Features

Support for heterogeneous graphs with multi-type nodes and edges
0 GitHub stars
Implementation of state-of-the-art GNN layers including GCN, GAT, and GraphSAGE
Efficient mini-batch processing and handling of large-scale graph datasets
Standardized patterns for molecular property prediction and chemical informatics
Custom MessagePassing interface for developing bespoke graph convolution layers

Use Cases

Building recommendation systems and social network analysis tools
Developing 3D geometric deep learning models for point clouds and meshes
Predicting molecular properties for drug discovery and chemical research