This project uses pixi for running trainings and managing the python environment, so the first step is installing pixi. Then, individal trainigns can be done by vhanging the config files in "configs" and running "pixi run train". Systematic size comaprison of DeepGCN-RT models can be done by running "pixi run compare-sizes", and the sizes tested and other definitions can be changed by changing the code of the file "systematic_size_comparison.py".