README.md
2.1 KB · 68 lines · markdown Raw
1 ---
2 library_name: stable-baselines3
3 tags:
4 - MsPacmanNoFrameskip-v4
5 - deep-reinforcement-learning
6 - reinforcement-learning
7 - stable-baselines3
8 model-index:
9 - name: PPO
10 results:
11 - metrics:
12 - type: mean_reward
13 value: 1659.00 +/- 144.81
14 name: mean_reward
15 task:
16 type: reinforcement-learning
17 name: reinforcement-learning
18 dataset:
19 name: MsPacmanNoFrameskip-v4
20 type: MsPacmanNoFrameskip-v4
21 ---
22
23 # **PPO** Agent playing **MsPacmanNoFrameskip-v4**
24 This is a trained model of a **PPO** agent playing **MsPacmanNoFrameskip-v4**
25 using the [stable-baselines3 library](https://github.com/DLR-RM/stable-baselines3)
26 and the [RL Zoo](https://github.com/DLR-RM/rl-baselines3-zoo).
27
28 The RL Zoo is a training framework for Stable Baselines3
29 reinforcement learning agents,
30 with hyperparameter optimization and pre-trained agents included.
31
32 ## Usage (with SB3 RL Zoo)
33
34 RL Zoo: https://github.com/DLR-RM/rl-baselines3-zoo<br/>
35 SB3: https://github.com/DLR-RM/stable-baselines3<br/>
36 SB3 Contrib: https://github.com/Stable-Baselines-Team/stable-baselines3-contrib
37
38 ```
39 # Download model and save it into the logs/ folder
40 python -m rl_zoo3.load_from_hub --algo ppo --env MsPacmanNoFrameskip-v4 -orga sb3 -f logs/
41 python enjoy.py --algo ppo --env MsPacmanNoFrameskip-v4 -f logs/
42 ```
43
44 ## Training (with the RL Zoo)
45 ```
46 python train.py --algo ppo --env MsPacmanNoFrameskip-v4 -f logs/
47 # Upload the model and generate video (when possible)
48 python -m rl_zoo3.push_to_hub --algo ppo --env MsPacmanNoFrameskip-v4 -f logs/ -orga sb3
49 ```
50
51 ## Hyperparameters
52 ```python
53 OrderedDict([('batch_size', 256),
54 ('clip_range', 'lin_0.1'),
55 ('ent_coef', 0.01),
56 ('env_wrapper',
57 ['stable_baselines3.common.atari_wrappers.AtariWrapper']),
58 ('frame_stack', 4),
59 ('learning_rate', 'lin_2.5e-4'),
60 ('n_envs', 8),
61 ('n_epochs', 4),
62 ('n_steps', 128),
63 ('n_timesteps', 10000000.0),
64 ('policy', 'CnnPolicy'),
65 ('vf_coef', 0.5),
66 ('normalize', False)])
67 ```
68