For managing virtual environments we use micromamba, which can be installed with the command
"${SHELL}" <(curl -L micro.mamba.pm/install.sh)
To prepare the environment, run the following commands.
micromamba create -n iclr python=3.10 # create env
micromamba activate iclr # activate env
pip install -r requirements.txt # setup env (install required packages)
python3 -m experiments.train_gcn --help # list the options of the main script
To train the NoMP architecture on the Letter-High dataset, run the following command.
python3 -m experiments.train_gcn --dataset-name=Letter-high --model-name=NoMP --use-ect=False # run an experiment
Train LEAP with the following command.
python3 -m experiments.train_gcn --dataset-name=Letter-high --use-ect=True --learn-directions=True --use-pe=False --ect-hops=1