We ported this project to use the TensorFlow 2 API. In particular, the following changes were made:
- Used rewrite script to generate TF2 code (see https://www.tensorflow.org/guide/migrate/upgrade)
- Deleted TF1 codebase
- TF2 code now is in
code-tf2/ - Changed code in
encoders/affine_transform.pyto use TF2 API (search for "TF2 compatibility") - Disabled eager execution for now (see top of
code-tf2/train.py)
Example train script:
python.exe code-tf2/train.py --settings ../settings/gcn_basis.exp --dataset ../data/FB-Toutanova
To install CUDA on Windows: https://shawnhymel.com/1961/how-to-install-tensorflow-with-gpu-support-on-windows/
These branches contain code for the various experiments we reported in the report:
- abht : GCN basis with GraphSAGE
- abht-exp2 : GCN block with GraphSAGE
- abht-exp3 : GCN basis without GraphSAGE
- abht-exp4 : GCN block without GraphSAGE
- abht-exp5 : GCN basis with random without GraphSAGE
- abht-exp6 : GCN block with random without GraphSAGE
Note that in order to run these tests, you will need to change the .exp file flag based on either basis or block.
This repository contains a TensorFlow implementation of Relational Graph Convolutional Networks (R-GCN), as well as experiments on relational link prediction. The description of the model and the results can be found in out paper:
Modeling Relational Data with Graph Convolutional Networks. Michael Schlichtkrull, Thomas N. Kipf, Peter Bloem, Rianne van den Berg, Ivan Titov, Max Welling (ArXiv 2017)
Requirements
- TensorFlow (1.4)
Running demo
We provide a bash script to run a demo of our code. In the folder settings, a collection of configuration files can be found. The block diagonal model used in our paper is represented through the configuration file settings/gcn_block.exp. To run a given experiment, execute our bash script as follows:
bash run-train.sh \[configuration\]
We advise that training can take up to several hours and require a significant amount of memory.
Citation
Please cite our paper if you use this code in your own work:
@article{schlichtkrull2017modeling,
title={Modeling Relational Data with Graph Convolutional Networks},
author={Schlichtkrull, Michael and Kipf, Thomas N and Bloem, Peter and Berg, Rianne van den and Titov, Ivan and Welling, Max},
journal={arXiv preprint arXiv:1703.06103},
year={2017}
}