How to handle graph nodes with multiple features of different dimensions

If my node has multiple features of different dimensions and I want to synthesize their information while training, can I only splice the features?