README.md

metadata

library_name: stable-baselines3
tags:
  - LunarLanderContinuous-v2
  - deep-reinforcement-learning
  - reinforcement-learning
  - stable-baselines3
model-index:
  - name: SAC
    results:
      - metrics:
          - type: mean_reward
            value: 251.89 +/- 71.75
            name: mean_reward
        task:
          type: reinforcement-learning
          name: reinforcement-learning
        dataset:
          name: LunarLanderContinuous-v2
          type: LunarLanderContinuous-v2

SAC Agent playing LunarLanderContinuous-v2

This is a trained model of a SAC agent playing LunarLanderContinuous-v2 using the stable-baselines3 library and the RL Zoo.

The RL Zoo is a training framework for Stable Baselines3 reinforcement learning agents, with hyperparameter optimization and pre-trained agents included.

Usage (with 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 sac --env LunarLanderContinuous-v2 -orga sb3 -f logs/
python enjoy.py --algo sac --env LunarLanderContinuous-v2  -f logs/

Training (with the RL Zoo)

python train.py --algo sac --env LunarLanderContinuous-v2 -f logs/
# Upload the model and generate video (when possible)
python -m rl_zoo3.push_to_hub --algo sac --env LunarLanderContinuous-v2 -f logs/ -orga sb3

Hyperparameters

OrderedDict([('batch_size', 256),
             ('buffer_size', 1000000),
             ('ent_coef', 'auto'),
             ('gamma', 0.99),
             ('gradient_steps', 1),
             ('learning_rate', 'lin_7.3e-4'),
             ('learning_starts', 10000),
             ('n_timesteps', 500000.0),
             ('policy', 'MlpPolicy'),
             ('policy_kwargs', 'dict(net_arch=[400, 300])'),
             ('tau', 0.01),
             ('train_freq', 1),
             ('normalize', False)])