from utils.tensor_board import global_board from utils.helper_classes import Counter, Timer from utils.conf import Config from utils.save_env import SaveEnv from utils.prep import prep_args from env.walker.single_walker import BipedalWalker # definitions observe_dim = 24 action_dim = 4 # configs c = Config() #c.restart_from_trial = "2020_05_06_21_50_57" c.max_episodes = 5000 c.max_steps = 1000 c.replay_size = 500000 # or: explore_noise_params = [(0, 0.2)] * action_dim c.explore_noise_params = (0, 0.2) c.policy_noise_params = (0, 1.0, -0.5, 0.5) c.device = "cuda:0" c.root_dir = "/data/AI/tmp/multi_agent/walker/naive_ddpg_td3/" # train configs # lr: learning rate, int: interval # warm up should be less than one epoch c.ddpg_update_batch_size = 100 c.ddpg_warmup_steps = 200 c.model_save_int = 100 # in episodes
from utils.tensor_board import global_board from utils.helper_classes import Counter, Timer from utils.conf import Config from utils.save_env import SaveEnv from utils.prep import prep_args from env.walker.carrier import BipedalMultiCarrier # definitions observe_dim = 28 action_dim = 4 # configs c = Config() # c.restart_from_trial = "2020_05_06_21_50_57" c.max_episodes = 20000 c.max_steps = 2000 c.replay_size = 500000 c.agent_num = 3 c.sub_policy_num = 1 c.explore_noise_params = (0, 0.2) c.q_increase_rate = 1 c.q_decrease_rate = 1 c.device = "cuda:0" c.root_dir = "/data/AI/tmp/multi_agent/mcarrier/maddpg/" # train configs # lr: learning rate, int: interval # warm up should be less than one epoch c.ddpg_update_batch_size = 100