Train Graph Neural Network model by optimizing edge features in PyTorch Geometric

I am new to the field of Graph Representation Learning. I am trying to train the GNN by optimizing only the edge features without optimizing the node feature vectors. Basically I want to obtain a set of edge feature vectors that minimizing the loss function. I was not able to find out a way to do this in the geometric documentation. So I want to know whether there is a way to achieve this.