def main(_): env_interface = EmbeddingInterfaceWrapper(BeaconEnvironmentInterface()) # env_interface = EmbeddingInterfaceWrapper(TrainMarines()) learner = Learner(10, env_kwargs, env_interface, run_name="TestRefactor", load_model=True) learner.train()
def main(_): env_interface = EmbeddingInterfaceWrapper(BeaconEnvironmentInterface()) # env_interface = EmbeddingInterfaceWrapper(TrainMarines()) # learner = Learner(10, env_kwargs, env_interface, run_name="MineralWithBeacon2", load_name="Beacon2", load_model=True) # env_interface = EmbeddingInterfaceWrapper(BeaconEnvironmentInterface()) # learner = NormalLearner(env_interface, load_model=False) # learner.train() # Refresh environment every once in a while to deal with memory leak environment = MultipleEnvironment(lambda: SCEnvironmentWrapper(env_interface, env_kwargs), num_instance=1) agent = LSTMAgent(env_interface) learner = ActorCriticLearner(environment, agent, run_name="SyncMarines", load_model=False) i = 0 while True: i += 1 print(i) learner.train_episode()
def main(_): env_interface = EmbeddingInterfaceWrapper(BeaconEnvironmentInterface()) learner = NormalLearner(env_interface, load_model=False) learner.train()
def main(_): # env_interface = EmbeddingInterfaceWrapper(TrainMarines()) env_interface = EmbeddingInterfaceWrapper(BeaconEnvironmentInterface()) learner = Learner(16, env_kwargs, env_interface, run_name="ImpBeacon1") learner.train()