Reporting validation auc when hyperparameter training with raytune

Has anyone used raytune with keras? I am trying to output a custom metrics: validation auc and difference between train and test AUC. Is there a way to do this? I haven’t been able to figure this out from the available examples on raytune. Thank you.