Generalized Population-Based Training for Hyperparameter Optimization in Reinforcement Learning | Synapse