from agents.actor_critic_agents.SAC_Discrete import SAC_Discrete from agents.Trainer import Trainer from utilities.data_structures.Config import Config from environments.DMP_simulator_3d_static_circle import deep_mobile_printing_3d1r PALN_CHOICE = 1 # 0 dense 1 sparse PLAN_LIST = ["dense", "sparse"] PLAN_NAME = PLAN_LIST[PALN_CHOICE] config = Config() config.seed = 5 config.environment = deep_mobile_printing_3d1r(plan_choose=PALN_CHOICE) config.num_episodes_to_run = 5000 config.show_solution_score = False config.visualise_individual_results = False config.visualise_overall_agent_results = True config.standard_deviation_results = 1.0 config.runs_per_agent = 1 config.use_GPU = True config.GPU = "cuda:1" config.overwrite_existing_results_file = True config.randomise_random_seed = False config.save_model = False OUT_FILE_NAME = "SAC_3d_" + PLAN_NAME + "_seed_" + str(config.seed) config.save_model_path = "/mnt/NAS/home/WenyuHan/SNAC/SAC/3D/static/" + OUT_FILE_NAME + "/" config.file_to_save_data_results = "/mnt/NAS/home/WenyuHan/SNAC/SAC/3D/static/" + OUT_FILE_NAME + "/" + "Results_Data.pkl" config.file_to_save_results_graph = "/mnt/NAS/home/WenyuHan/SNAC/SAC/3D/static/" + OUT_FILE_NAME + "/" + "Results_Graph.png" if os.path.exists(config.save_model_path) == False: os.makedirs(config.save_model_path) config.hyperparameters = { "Actor_Critic_Agents": { "learning_rate": 0.005,
'SAC_Discrete': SAC_Discrete, 'DIAYN': DIAYN, 'DBH': DBH } if args.rts: config.rts() AGENTS = [DDQN, SAC_Discrete, DIAYN, DBH] else: AGENTS = [str_to_obj[i] for i in args.algorithms] config.environment_name = args.environment config.environment = gym.make(config.environment_name) config.eval = args.evaluate config.seed = args.seed config.num_episodes_to_run = args.num_episodes config.runs_per_agent = args.n_trials config.use_GPU = args.use_GPU config.save_results = args.save_results config.run_prefix = args.run_prefix config.train_existing_model = args.tem config.save_directory = 'results/{}'.format(config.run_prefix) if not os.path.exists(config.save_directory): os.makedirs(config.save_directory) config.visualise_overall_agent_results = True config.standard_deviation_results = 1.0 linear_hidden_units = [128, 128, 32] learning_rate = 0.01 buffer_size = 100000 batch_size = 256 batch_norm = False