Modeling Relational Data with Graph Convolutional Networks in PaddlePaddle

This is a PaddlePaddle implementation of the Relational Graph Convolutional Networks (R-GCN) described in the paper:

Schlichtkrull, Michael, et al. "Modeling relational data with graph convolutional networks." European semantic web conference. Springer, Cham, 2018.

The code in this repo is based on or refers to https://github.com/berlincho/RGCN-pytorch and https://github.com/tkipf/relational-gcn

Requirements

• Hardware：CPU (RAM larger than 36G is recommended)
• python-3.8.12
• paddlepaddle-2.1.3
• paddlenlp-2.1.1
• rdflib-6.0.2
• wget-3.2
• h5py-3.5.0
• install requirements via pip install -r requirements.txt

Usage

train: python run.py --train

test: python run.py