This is the code base for the UniAug project.
conda create -f graph_enviroment.yaml
conda activate graph
pip install -r requirements.txt
pip install -f https://data.pyg.org/whl/torch-2.1.0+cu121.html torch-cluster==1.6.3+pt21cu121 torch-scatter==2.1.2+pt21cu121 torch-sparse==0.6.18+pt21cu121 torch-spline-conv==1.2.2+pt21cu121
pip install git+https://github.com/fabriziocosta/EDeN.git --user
After executing the shell commands above, there should be a conda enviroment named graph with all dependencies.
All of our shell scripts are under the run folder to ease usage.
If you are using HPC,you may need to revise the Resource Request part for scripts with the prefix hpcc correspondingly.
Kindly note that for the very first time you may need to login wandb before the execution via wandb login following the prompts.
You could uncomment the corresponding linesin the scripts for different datasets and guidence settings.
bash run/graph_pred_mol.shYou could uncomment the corresponding linesin the scripts for different datasets and guidence settings.
bash run/node_class_subgraph.sh You could uncomment the corresponding lines for different datasets and guidence settings.
bash bash run/link_pred_ncn.sh