Please make sure to install the following dependencies:

networx == 2.6.3 ogb == 1.3.5 pytorch_geometric == 2.0.1 Pytorch == 1.9.0 dgl == 0.7.1

• CoraFull, Wiki-CS, Amazon-Computer, Flickr, Reddit: Implemented with PyTorch Geometric (PyG).
• Amazon2M: Implemented with Open Graph Benchmark (OGB).
• Aminer-CS: Please download from the GRAND-plus repository.

We have uploaded all pre-trained positional embeddings in the dataset/'name_of_dataset/node2vec' folder except for Amazon, 128-dim PE for Aminer-CS, and 128-dim PE for Reddit (due to exceeding the size limit of GitHub). However, you may pretrain the positional embeddings yourself with different hyper-parameters by modifying the configuration in pos_emb.conf. Please place the generated .pt file in above folder.

Optionally, run:

python generae_pos_emb.py

The running logs for each dataset are available and will be stored in the experiments_rebuttal/'name_of_dataset'/'time_of_experiments'/run.log', where all the hyper-parameters are listed. Before training the model, you may change the hyperparameters in large_graph.conf to reproduce the results from the paper.

To start training, run:

python gnn_gw_large.py