#2828th of 37 in ML frameworks and libraries
pytorch_geometric
Graph neural network library built on PyTorch
- Stars
- 24.1k
- License
- MIT
- Last commit
- Sep 2026
- Last release
- Oct 2026
- Language
- Python
Missed the latest GitHub check; these facts may be older.
Overview
PyG (PyTorch Geometric) is a Python library for writing and training graph neural networks on PyTorch. It provides a MessagePassing interface for custom layers, a large set of published GNN layers and models, mini-batch loaders and samplers for many small graphs or one large graph, and benchmark datasets. It also covers heterogeneous graphs, 3D meshes and point clouds, multi-GPU training and torch.compile.
Who it is for: ML researchers and engineers training GNNs on graph-structured data
Strengths
- Unified MessagePassing API keeps custom GNN layers short, as the EdgeConv example shows
- Large catalog of published GNN layers, each linked to its paper and often an example script
- Covers heterogeneous graphs, dynamic graphs, and sampling for graphs with millions of nodes
- Supports torch.compile, TorchScript, multi-GPU examples and optional pyg-lib sparse kernels
Weaknesses
- Requires PyTorch; install steps and version matching are not in the visible README
- Focused on graph learning only, not a general ML or LLM toolkit
- Memory and GPU requirements are not stated in the README
- Some acceleration comes from separate packages such as pyg-lib
What it needs
- GPU optional
- Needs PyTorch, pyg-lib (optional)
- Models: GCN, GAT, GraphSAGE, GIN, RGCN
Also in ML frameworks and libraries
See all 37| Rank | Project | Score |
|---|---|---|
| 1 | transformersPython library defining open models for text, vision, audio and multimodal tasks | 95 out of 100 |
| 2 | pytorchPython tensor library with GPU support and tape-based autograd for deep learning | 94 out of 100 |
| 3 | ultralyticsPython library and CLI for training and running YOLO vision models | 90 out of 100 |
| #4 | supervisionPython toolkit for loading, annotating and counting computer vision detections | 87 out of 100 |
| #5 | pytorch-image-modelsPyTorch library of image models, pretrained weights and reference training scripts | 86 out of 100 |