Abstract: Graph representation learning (GRL) is fundamental in multi-graph applications like molecular property prediction. Graph neural networks (GNNs) have emerged as a popular method for GRL.