Recreating *.pkl data

The data files from the original repo have the following error with DGL-0.5+:

pickle.load(f)ModuleNotFoundError: No module named 'dgl.graph'

Steps to generate a new format *.pkl file:

1. download raw data from Benchmarking GNNs and put it in Dir: raw_data.
2. Run prepare_pkl.py

Directional Graph Networks

Implementation of Directional Graph Networks in PyTorch and DGL.

Overview

We provide the implementation of the Directional Graph Networks (DGN) in PyTorch and DGL frameworks, along with scripts for running real-world benchmarks. The repository is organised as follows:

• models contains:
  • pytorch contains the various GNN models implemented in PyTorch: the implementation of the aggregators, the scalers, the DGN layer and the directional aggregation matrix (eigen_agg).
  • dgl contains the DGN model implemented via the DGL library: aggregators, scalers, and DGN layer.
  • layers.py contains general NN layers used by the various models
• realworld_benchmark contains various scripts from Benchmarking GNNs and Open Graph Benchmark to download the real-world benchmarks and train the DGN on them. In realworld_benchmark/README.md we provide instructions for runnning the experiments.

Reference

@article{beaini2020directional,
  title={Directional graph networks},
  author={Beaini, Dominique and Passaro, Saro and L{\'e}tourneau, Vincent and Hamilton, William L and Corso, Gabriele and Li{\`o}, Pietro},
  journal={arXiv preprint arXiv:2010.02863},
  year={2020}
}

License

MIT