Skip to content

ding.config.example.DQN.gym_lunarlander_v2

ding.config.example.DQN.gym_lunarlander_v2

Full Source Code

../ding/config/example/DQN/gym_lunarlander_v2.py

1from easydict import EasyDict 2import ding.envs.gym_env 3 4cfg = dict( 5 exp_name='LunarLander-v2-DQN', 6 seed=0, 7 env=dict( 8 env_id='LunarLander-v2', 9 collector_env_num=8, 10 evaluator_env_num=8, 11 n_evaluator_episode=8, 12 stop_value=260, 13 ), 14 policy=dict( 15 cuda=True, 16 random_collect_size=25000, 17 discount_factor=0.99, 18 nstep=3, 19 learn=dict( 20 update_per_collect=10, 21 batch_size=64, 22 learning_rate=0.001, 23 # Frequency of target network update. 24 target_update_freq=100, 25 ), 26 model=dict( 27 obs_shape=8, 28 action_shape=4, 29 encoder_hidden_size_list=[512, 64], 30 # Whether to use dueling head. 31 dueling=True, 32 ), 33 collect=dict( 34 n_sample=64, 35 unroll_len=1, 36 ), 37 other=dict( 38 eps=dict( 39 type='exp', 40 start=0.95, 41 end=0.1, 42 decay=50000, 43 ), replay_buffer=dict(replay_buffer_size=100000, ) 44 ), 45 ), 46 wandb_logger=dict( 47 gradient_logger=True, video_logger=True, plot_logger=True, action_logger=True, return_logger=False 48 ), 49) 50 51cfg = EasyDict(cfg) 52 53env = ding.envs.gym_env.env