/Graphlib

library for state-of-the-art graph neural networks

Primary LanguagePythonMIT LicenseMIT

Graphlib

Implementation of latest graph neural network model using Pytorch and torch_geometric.

Structure

base

  • BaseDataLoader
  • BaseModel
  • BaseTrainer

nn

Various neural network layer for graph neural network

  • SGC_LL: Adaptive Graph Convolutional Neural Networks. Ruoyu Li, Sheng Wang, Feiyun Zhu, Junzhou Huang. AAAI 2018. paper

  • graph_max_pool: same as above

data loader

Customized data loaders for various datasets

  • AlchemyDataLoader: Tecent Alchemy dataset

model

The state-of-the-art graph neural network model

  • AGCN: Adaptive Graph Convolutional Neural Networks. Ruoyu Li, Sheng Wang, Feiyun Zhu, Junzhou Huang. AAAI 2018. paper

  • MPNN: Neural Message Passing for Quantum Chemistry. Gilmer, Justin and Schoenholz, Samuel S and Riley, Patrick F and Vinyals, Oriol and Dahl, George E. 2017. paper

trainer

Customized trainer for training the model and save the training log

  • Trainer

Installation

Usage

Contributing

Pull requests are welcome. For major changes, please open an issue first to discuss what you would like to change.

Please make sure to update tests as appropriate.

TODOs

  • Implement common feature transformation for molecular graph
  • Multi-GPU support

License

MIT