ding.config.example.TD3.gym_walker2d_v3

ding.config.example.TD3.gym_walker2d_v3

Full Source Code

../ding/config/example/TD3/gym_walker2d_v3.py

1from easydict import EasyDict 2import ding.envs.gym_env 3 4cfg = dict( 5 exp_name='Walker2d-v3-TD3', 6 seed=0, 7 env=dict( 8 env_id='Walker2d-v3', 9 norm_obs=dict(use_norm=False, ), 10 norm_reward=dict(use_norm=False, ), 11 collector_env_num=1, 12 evaluator_env_num=8, 13 n_evaluator_episode=8, 14 stop_value=6000, 15 env_wrapper='mujoco_default', 16 ), 17 policy=dict( 18 cuda=True, 19 random_collect_size=25000, 20 model=dict( 21 obs_shape=17, 22 action_shape=6, 23 twin_critic=True, 24 actor_head_hidden_size=256, 25 critic_head_hidden_size=256, 26 action_space='regression', 27 ), 28 learn=dict( 29 update_per_collect=1, 30 batch_size=256, 31 learning_rate_actor=1e-3, 32 learning_rate_critic=1e-3, 33 ignore_done=False, 34 target_theta=0.005, 35 discount_factor=0.99, 36 actor_update_freq=2, 37 noise=True, 38 noise_sigma=0.2, 39 noise_range=dict( 40 min=-0.5, 41 max=0.5, 42 ), 43 ), 44 collect=dict( 45 n_sample=1, 46 unroll_len=1, 47 noise_sigma=0.1, 48 ), 49 other=dict(replay_buffer=dict(replay_buffer_size=1000000, ), ), 50 ), 51 wandb_logger=dict( 52 gradient_logger=True, video_logger=True, plot_logger=True, action_logger=True, return_logger=False 53 ), 54) 55 56cfg = EasyDict(cfg) 57 58env = ding.envs.gym_env.env