TensorFlow GNN

Summary

TensorFlow GNN is a library to build Graph Neural Networks on the TensorFlow platform. It provides...

• a tfgnn.GraphTensor type to represent graphs with a heterogeneous schema, that is, multiple types of nodes and edges;
• tools for data preparation, notably a graph sampler to convert a huge database into a stream of reasonably-sized subgraphs for training and inference;
• a collection of ready-to-use models and Keras layers to do your own GNN modeling;
• a high-level API for training orchestration.

This library is an OSS port of a Google-internal library used in a broad variety of contexts, on homogeneous and heterogeneous graphs, and in conjunction with other scalable graph mining tools.

For background and discussion, please see O. Ferludin et al.: TF-GNN: Graph Neural Networks in TensorFlow, 2023 (full citation below).

Quickstart

Google Colab lets you run TF-GNN demos from your browser, no installation required:

• Molecular Graph Classification with the MUTAG dataset.
• Solving OGBN-MAG end-to-end trains a model on heterogeneous sampled subgraphs from the popular OGBN-MAG benchmark.
• Learning shortest paths with GraphNetworks demonstrates an advanced Encoder/Process/Decoder architecture for predicting the edges of a shortest path.

For all colabs and user guides, please see the Documentation overview page, which also links to the API docs.

Installation Instructions

The latest stable release of TensorFlow GNN is available from

pip install tensorflow-gnn

For installation from source, see our Developer Guide.

Key platform requirements:

• TensorFlow 2.12 or higher, and any GPU drivers it needs [instructions].
• Keras v2, as traditionally included with TensorFlow 2.x. TF-GNN does not work with the new multi-backend Keras v3.
  Users of TF2.16+ must also pip install tf-keras and set TF_USE_LEGACY_KERAS=1, see our Keras version guide for details.
• Apache Beam for distributed graph sampling.

TF-GNN is developed and tested on Linux. Running on other platforms supported by TensorFlow may be possible.

Citation

When referencing this library in a paper, please cite the TF-GNN paper:

@article{tfgnn,
  author  = {Oleksandr Ferludin and Arno Eigenwillig and Martin Blais and
             Dustin Zelle and Jan Pfeifer and Alvaro Sanchez{-}Gonzalez and
             Wai Lok Sibon Li and Sami Abu{-}El{-}Haija and Peter Battaglia and
             Neslihan Bulut and Jonathan Halcrow and
             Filipe Miguel Gon{\c{c}}alves de Almeida and Pedro Gonnet and
             Liangze Jiang and Parth Kothari and Silvio Lattanzi and 
             Andr{\'{e}} Linhares and Brandon Mayer and Vahab Mirrokni and
             John Palowitch and Mihir Paradkar and Jennifer She and
             Anton Tsitsulin and Kevin Villela and Lisa Wang and David Wong and
             Bryan Perozzi},
  title   = {{TF-GNN:} Graph Neural Networks in TensorFlow},
  journal = {CoRR},
  volume  = {abs/2207.03522},
  year    = {2023},
  url     = {http://arxiv.org/abs/2207.03522},
}