tf_gnns documentation

tf_gnns is a hackable graph neural network library with Keras-first APIs for building and composing GraphNet blocks.

Getting started

• Installation
• Quickstart

Tutorials and examples

• Tutorials and examples
  • tf_gnns: A Hackable graph_nets library
  • GraphNet basics
  • The GraphTuple data structure
  • Creating a Custom GraphNet
  • Taking gradients of the whole operation:
  • Support for global variables
  • List sorting with GNNs
  • Note
  • Evaluation
  • Comparison with other models
  • tf_gnns with Keras 3 Torch Backend

API reference

• API reference
  • Package exports
  • GraphNet utilities
  • Graph data structures
  • High-level models