main aim of our work is to develop reversible graph neural network models, called Graph Normalizing Flows (GNFs)