snap-stanford/GraphGym

Health v2 · maintenance only

Flags: no_license

Adoption not part of the score

1904 stars · 199 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

GraphGym is a platform for designing and evaluating Graph Neural Networks (GNNs), built around a highly modularized pipeline covering data loading, model architecture, tasks, and evaluation. It enables reproducible experiments via configuration files and scalable parallel launching of thousands of GNN experiments with auto-generated analyses.

Use cases

• design and evaluate graph neural networks
• run standardized GNN experiments reproducibly
• benchmark GNN architectures across node edge and graph tasks
• launch thousands of GNN experiments in parallel
• learn how GNNs work with a standardized implementation
• search the GNN design space for the best model
• register custom GNN layers and data loaders

When to choose

• you need a modular, reproducible pipeline for GNN experiments
• you want to systematically compare many GNN architectures and configurations
• you are a researcher exploring the GNN design space
• you are a beginner wanting a standardized GNN implementation and evaluation setup

When to avoid

• you need a general-purpose deep learning framework rather than a GNN-specific platform
• you need cutting-edge actively developed GNN features - consider the tightly integrated GraphGym+PyG version
• you work outside graph-structured data domains
• you need a no-code or GUI-based ML tool

Facets

library · maturity maintenance

machine-learning deep-learning benchmarking data-science machine-learning deep-learning python windows graph-neural-networks gnn pytorch pytorch-geometric experiment-management graph-learning neurips algorithms research linux macos

Member repositories

For agents

markdown · JSON · MCP: product_card(name="snap-stanford/GraphGym")