def start_trainer(self, trainer: Trainer, env_manager: EnvManager) -> None: self.trainers[trainer.brain_name] = trainer self.logger.info(trainer) if self.train_model: trainer.write_tensorboard_text("Hyperparameters", trainer.parameters) env_manager.set_policy(trainer.brain_name, trainer.policy)
def trainer_update_func(self, trainer: Trainer) -> None: while not self.kill_trainers: with hierarchical_timer("trainer_advance"): trainer.advance()