python run_exp -c "path_to_config file"
- -c "path_to_config". If this is left empty, it will attempt to use the config file generated by the last training run.
- -t indicates testing (i.e. sampling from a trained model).
See experiment_configs for example configuration files.
The major dependencies are PyTorch 1.10 with GPU enabled and an appropriate Torch Geometric version. Other dependencies can be found in the requirements file.
The model code can be found in models/dmnets_gnn_mb.py. We train in parallel on all subgraphs during the generation of