模型:

HumanCompatibleAI/ppo-seals-MountainCar-v0

英文

PPO Agent玩seals/MountainCar-v0

这是一个经过训练的PPO智能体模型,使用 stable-baselines3 library RL Zoo

RL Zoo是一个训练框架,适用于Stable Baselines3强化学习智能体,包括超参数优化和预训练智能体。

使用(使用SB3 RL Zoo)

RL Zoo: https://github.com/DLR-RM/rl-baselines3-zoo SB3: https://github.com/DLR-RM/stable-baselines3 SB3 Contrib: https://github.com/Stable-Baselines-Team/stable-baselines3-contrib

# Download model and save it into the logs/ folder
python -m rl_zoo3.load_from_hub --algo ppo --env seals/MountainCar-v0 -orga HumanCompatibleAI -f logs/
python enjoy.py --algo ppo --env seals/MountainCar-v0  -f logs/

如果通过pip安装了RL Zoo3(pip install rl_zoo3),则可以在任何地方执行以下操作:

python -m rl_zoo3.load_from_hub --algo ppo --env seals/MountainCar-v0 -orga HumanCompatibleAI -f logs/
rl_zoo3 enjoy --algo ppo --env seals/MountainCar-v0  -f logs/

训练(使用RL Zoo)

python train.py --algo ppo --env seals/MountainCar-v0 -f logs/
# Upload the model and generate video (when possible)
python -m rl_zoo3.push_to_hub --algo ppo --env seals/MountainCar-v0 -f logs/ -orga HumanCompatibleAI

超参数

OrderedDict([('batch_size', 512),
             ('clip_range', 0.2),
             ('ent_coef', 6.4940755116195606e-06),
             ('gae_lambda', 0.98),
             ('gamma', 0.99),
             ('learning_rate', 0.0004476103728105138),
             ('max_grad_norm', 1),
             ('n_envs', 16),
             ('n_epochs', 20),
             ('n_steps', 256),
             ('n_timesteps', 1000000.0),
             ('normalize',
              {'gamma': 0.99, 'norm_obs': False, 'norm_reward': True}),
             ('policy', 'MlpPolicy'),
             ('policy_kwargs',
              {'activation_fn': <class 'torch.nn.modules.activation.Tanh'>,
               'features_extractor_class': <class 'imitation.policies.base.NormalizeFeaturesExtractor'>,
               'net_arch': [{'pi': [64, 64], 'vf': [64, 64]}]}),
             ('vf_coef', 0.25988158989488963),
             ('normalize_kwargs',
              {'norm_obs': {'gamma': 0.99,
                            'norm_obs': False,
                            'norm_reward': True},
               'norm_reward': False})])