johnjim0816's picture
update CartPole-v1 NoisyDQN
54d82df
raw
history blame
5.38 kB
2023-01-14 22:27:28 - r - INFO: - Hyperparameters:
2023-01-14 22:27:28 - r - INFO: - ================================================================================
2023-01-14 22:27:28 - r - INFO: - Name Value Type
2023-01-14 22:27:28 - r - INFO: - env_name CartPole-v1 <class 'str'>
2023-01-14 22:27:28 - r - INFO: - new_step_api 1 <class 'bool'>
2023-01-14 22:27:28 - r - INFO: - wrapper None <class 'str'>
2023-01-14 22:27:28 - r - INFO: - render 0 <class 'bool'>
2023-01-14 22:27:28 - r - INFO: - algo_name C51 <class 'str'>
2023-01-14 22:27:28 - r - INFO: - mode test <class 'str'>
2023-01-14 22:27:28 - r - INFO: - seed 0 <class 'int'>
2023-01-14 22:27:28 - r - INFO: - device cuda <class 'str'>
2023-01-14 22:27:28 - r - INFO: - train_eps 100 <class 'int'>
2023-01-14 22:27:28 - r - INFO: - test_eps 10 <class 'int'>
2023-01-14 22:27:28 - r - INFO: - eval_eps 10 <class 'int'>
2023-01-14 22:27:28 - r - INFO: - eval_per_episode 5 <class 'int'>
2023-01-14 22:27:28 - r - INFO: - max_steps 200 <class 'int'>
2023-01-14 22:27:28 - r - INFO: - load_checkpoint 1 <class 'bool'>
2023-01-14 22:27:28 - r - INFO: - load_path Train_CartPole-v1_C51_20230114-222523 <class 'str'>
2023-01-14 22:27:28 - r - INFO: - show_fig 0 <class 'bool'>
2023-01-14 22:27:28 - r - INFO: - save_fig 1 <class 'bool'>
2023-01-14 22:27:28 - r - INFO: - Vmin 0.0 <class 'float'>
2023-01-14 22:27:28 - r - INFO: - Vmax 200.0 <class 'float'>
2023-01-14 22:27:28 - r - INFO: - n_atoms 51 <class 'int'>
2023-01-14 22:27:28 - r - INFO: - gamma 0.95 <class 'float'>
2023-01-14 22:27:28 - r - INFO: - tau 1.0 <class 'float'>
2023-01-14 22:27:28 - r - INFO: - num_atoms 51 <class 'int'>
2023-01-14 22:27:28 - r - INFO: - support tensor(0.) <class 'str'>
2023-01-14 22:27:28 - r - INFO: - delta_z 4.0 <class 'float'>
2023-01-14 22:27:28 - r - INFO: - batch_size 64 <class 'int'>
2023-01-14 22:27:28 - r - INFO: - lr 0.0001 <class 'float'>
2023-01-14 22:27:28 - r - INFO: - target_update 4 <class 'int'>
2023-01-14 22:27:28 - r - INFO: - memory_capacity 10000 <class 'int'>
2023-01-14 22:27:28 - r - INFO: - epsilon_start 0.95 <class 'float'>
2023-01-14 22:27:28 - r - INFO: - epsilon_end 0.01 <class 'float'>
2023-01-14 22:27:28 - r - INFO: - epsilon_decay 500 <class 'int'>
2023-01-14 22:27:28 - r - INFO: - buffer_size 100000 <class 'int'>
2023-01-14 22:27:28 - r - INFO: - task_dir /home/dingli/rl-tutorials/joyrl/tasks/Test_CartPole-v1_C51_20230114-222728 <class 'str'>
2023-01-14 22:27:28 - r - INFO: - model_dir /home/dingli/rl-tutorials/joyrl/tasks/Test_CartPole-v1_C51_20230114-222728/models <class 'str'>
2023-01-14 22:27:28 - r - INFO: - res_dir /home/dingli/rl-tutorials/joyrl/tasks/Test_CartPole-v1_C51_20230114-222728/results <class 'str'>
2023-01-14 22:27:28 - r - INFO: - log_dir /home/dingli/rl-tutorials/joyrl/tasks/Test_CartPole-v1_C51_20230114-222728/logs <class 'str'>
2023-01-14 22:27:28 - r - INFO: - traj_dir /home/dingli/rl-tutorials/joyrl/tasks/Test_CartPole-v1_C51_20230114-222728/traj <class 'str'>
2023-01-14 22:27:28 - r - INFO: - ================================================================================
2023-01-14 22:27:28 - r - INFO: - n_states: 4, n_actions: 2
2023-01-14 22:27:30 - r - INFO: - Start testing!
2023-01-14 22:27:30 - r - INFO: - Env: CartPole-v1, Algorithm: C51, Device: cuda
2023-01-14 22:27:31 - r - INFO: - Episode: 1/10, Reward: 200.000, Step: 200
2023-01-14 22:27:31 - r - INFO: - Episode: 2/10, Reward: 200.000, Step: 200
2023-01-14 22:27:31 - r - INFO: - Episode: 3/10, Reward: 200.000, Step: 200
2023-01-14 22:27:31 - r - INFO: - Episode: 4/10, Reward: 200.000, Step: 200
2023-01-14 22:27:31 - r - INFO: - Episode: 5/10, Reward: 200.000, Step: 200
2023-01-14 22:27:31 - r - INFO: - Episode: 6/10, Reward: 200.000, Step: 200
2023-01-14 22:27:31 - r - INFO: - Episode: 7/10, Reward: 200.000, Step: 200
2023-01-14 22:27:31 - r - INFO: - Episode: 8/10, Reward: 200.000, Step: 200
2023-01-14 22:27:32 - r - INFO: - Episode: 9/10, Reward: 200.000, Step: 200
2023-01-14 22:27:32 - r - INFO: - Episode: 10/10, Reward: 200.000, Step: 200
2023-01-14 22:27:32 - r - INFO: - Finish testing!