johnjim0816
update DDPG Pendulum-v1
56029d5
raw
history blame
5.88 kB
2023-04-16 11:31:55 - r - INFO: - Hyperparameters:
2023-04-16 11:31:55 - r - INFO: - ================================================================================
2023-04-16 11:31:55 - r - INFO: - Name Value Type
2023-04-16 11:31:55 - r - INFO: - env_name gym <class 'str'>
2023-04-16 11:31:55 - r - INFO: - new_step_api 1 <class 'bool'>
2023-04-16 11:31:55 - r - INFO: - wrapper None <class 'str'>
2023-04-16 11:31:55 - r - INFO: - render 0 <class 'bool'>
2023-04-16 11:31:55 - r - INFO: - render_mode None <class 'str'>
2023-04-16 11:31:55 - r - INFO: - algo_name TD3_BC <class 'str'>
2023-04-16 11:31:55 - r - INFO: - mode test <class 'str'>
2023-04-16 11:31:55 - r - INFO: - mp_backend mp <class 'str'>
2023-04-16 11:31:55 - r - INFO: - seed 1 <class 'int'>
2023-04-16 11:31:55 - r - INFO: - device cuda <class 'str'>
2023-04-16 11:31:55 - r - INFO: - train_eps 1 <class 'int'>
2023-04-16 11:31:55 - r - INFO: - test_eps 10 <class 'int'>
2023-04-16 11:31:55 - r - INFO: - eval_eps 5 <class 'int'>
2023-04-16 11:31:55 - r - INFO: - eval_per_episode 10 <class 'int'>
2023-04-16 11:31:55 - r - INFO: - max_steps 200 <class 'int'>
2023-04-16 11:31:55 - r - INFO: - load_checkpoint 1 <class 'bool'>
2023-04-16 11:31:55 - r - INFO: - load_path Train_gym_TD3_BC_20230416-111154 <class 'str'>
2023-04-16 11:31:55 - r - INFO: - show_fig 0 <class 'bool'>
2023-04-16 11:31:55 - r - INFO: - save_fig 1 <class 'bool'>
2023-04-16 11:31:55 - r - INFO: - explore_steps 1000 <class 'int'>
2023-04-16 11:31:55 - r - INFO: - policy_freq 2 <class 'int'>
2023-04-16 11:31:55 - r - INFO: - actor_lr 0.0003 <class 'float'>
2023-04-16 11:31:55 - r - INFO: - critic_lr 0.0003 <class 'float'>
2023-04-16 11:31:55 - r - INFO: - actor_hidden_dim 256 <class 'int'>
2023-04-16 11:31:55 - r - INFO: - critic_hidden_dim 256 <class 'int'>
2023-04-16 11:31:55 - r - INFO: - gamma 0.99 <class 'float'>
2023-04-16 11:31:55 - r - INFO: - tau 0.005 <class 'float'>
2023-04-16 11:31:55 - r - INFO: - policy_noise 0.2 <class 'float'>
2023-04-16 11:31:55 - r - INFO: - expl_noise 0.1 <class 'float'>
2023-04-16 11:31:55 - r - INFO: - noise_clip 0.5 <class 'float'>
2023-04-16 11:31:55 - r - INFO: - batch_size 100 <class 'int'>
2023-04-16 11:31:55 - r - INFO: - buffer_size 1000000 <class 'int'>
2023-04-16 11:31:55 - r - INFO: - alpha 5 <class 'int'>
2023-04-16 11:31:55 - r - INFO: - lmbda 1 <class 'int'>
2023-04-16 11:31:55 - r - INFO: - train_iterations 2000 <class 'int'>
2023-04-16 11:31:55 - r - INFO: - normalize 0 <class 'bool'>
2023-04-16 11:31:55 - r - INFO: - expert_path tasks/Collect_gym_TD3_20230416-111040/traj/traj.pkl <class 'str'>
2023-04-16 11:31:55 - r - INFO: - id Pendulum-v1 <class 'str'>
2023-04-16 11:31:55 - r - INFO: - task_dir /home/dingli/joyrl_offline/tasks/Test_gym_TD3_BC_20230416-113155 <class 'str'>
2023-04-16 11:31:55 - r - INFO: - res_dir /home/dingli/joyrl_offline/tasks/Test_gym_TD3_BC_20230416-113155/results <class 'str'>
2023-04-16 11:31:55 - r - INFO: - log_dir /home/dingli/joyrl_offline/tasks/Test_gym_TD3_BC_20230416-113155/logs <class 'str'>
2023-04-16 11:31:55 - r - INFO: - traj_dir /home/dingli/joyrl_offline/tasks/Test_gym_TD3_BC_20230416-113155/traj <class 'str'>
2023-04-16 11:31:55 - r - INFO: - video_dir /home/dingli/joyrl_offline/tasks/Test_gym_TD3_BC_20230416-113155/videos <class 'str'>
2023-04-16 11:31:55 - r - INFO: - ================================================================================
2023-04-16 11:31:55 - r - INFO: - action_bound: 2.0
2023-04-16 11:31:55 - r - INFO: - n_states: 3, n_actions: 1
2023-04-16 11:32:00 - r - INFO: - Start testing!
2023-04-16 11:32:00 - r - INFO: - Env: gym, Algorithm: TD3_BC, Device: cuda
2023-04-16 11:32:01 - r - INFO: - Episode: 1/10, Reward: -0.477, Step: 200
2023-04-16 11:32:02 - r - INFO: - Episode: 2/10, Reward: -0.477, Step: 200
2023-04-16 11:32:02 - r - INFO: - Episode: 3/10, Reward: -0.477, Step: 200
2023-04-16 11:32:03 - r - INFO: - Episode: 4/10, Reward: -0.477, Step: 200
2023-04-16 11:32:03 - r - INFO: - Episode: 5/10, Reward: -0.477, Step: 200
2023-04-16 11:32:04 - r - INFO: - Episode: 6/10, Reward: -0.477, Step: 200
2023-04-16 11:32:05 - r - INFO: - Episode: 7/10, Reward: -0.477, Step: 200
2023-04-16 11:32:05 - r - INFO: - Episode: 8/10, Reward: -0.477, Step: 200
2023-04-16 11:32:06 - r - INFO: - Episode: 9/10, Reward: -0.477, Step: 200
2023-04-16 11:32:06 - r - INFO: - Episode: 10/10, Reward: -0.477, Step: 200
2023-04-16 11:32:06 - r - INFO: - Finish testing!