What is PyTorch Geometric (PyG)?

PyTorch Geometric (PyG) is a library built on PyTorch specifically designed for developing and training Graph Neural Networks (GNNs) for relational and geometric data.

How does this skill help with GNN development?

This skill provides Claude with best practices, implementation patterns, and code snippets for various GNN architectures, data handling, and training workflows.

Does it support large-scale graph processing?

Yes, it includes guidance on neighbor sampling, mini-batching, and distributed training techniques necessary for large-scale graph machine learning.

Can I implement custom GNN layers using this skill?

Absolutely. The skill covers the MessagePassing base class paradigm, allowing you to define custom message, aggregate, and update functions for novel architectures.

Is it suitable for molecular biology applications?

Yes, it is widely used for molecular property prediction and drug discovery, supporting specialized datasets and 3D geometric representations.

PyTorch Geometric (PyG)

Name: PyTorch Geometric (PyG)
Author: lifangda

bylifangda

•

Data Science & ML

Simplifies the development and training of Graph Neural Networks (GNNs) for deep learning on irregular and relational data structures.

PyTorch Geometric (PyG) is a specialized library built on PyTorch designed to facilitate deep learning on graphs, point clouds, and manifolds. This skill provides Claude with the domain-specific knowledge required to implement state-of-the-art GNN architectures such as GCN, GAT, and GraphSAGE. It includes patterns for efficient mini-batch processing, handling heterogeneous graphs, and performing complex tasks like node classification, link prediction, and molecular property prediction. It is an essential tool for data scientists and ML engineers working on social network analysis, drug discovery, or recommendation systems.

Key Features

01Comprehensive implementations of 40+ GNN convolutional layers

02Efficient mini-batching for irregular graph structures using block-diagonal matrices

03Built-in support for benchmark datasets like Cora, CiteSeer, and QM9

04Support for heterogeneous graphs with multiple node and edge types

0516 GitHub stars

06Custom message-passing interface for designing novel aggregation schemes

Use Cases

01Analyzing social and citation networks for node classification and community detection

02Building graph-based recommendation systems and knowledge graph embeddings

03Predicting molecular properties and chemical interactions for drug discovery

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add lifangda/claude-plugins torch-geometric

For use in Claude.ai and ChatGPT

Download Skill