README.md

metadata

library_name: stable-baselines3
tags:
  - HopperBulletEnv-v0
  - deep-reinforcement-learning
  - reinforcement-learning
  - stable-baselines3
model-index:
  - name: A2C
    results:
      - metrics:
          - type: mean_reward
            value: 709.34 +/- 213.24
            name: mean_reward
        task:
          type: reinforcement-learning
          name: reinforcement-learning
        dataset:
          name: HopperBulletEnv-v0
          type: HopperBulletEnv-v0

A2C Agent playing HopperBulletEnv-v0

This is a trained model of a A2C agent playing HopperBulletEnv-v0 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 a2c --env HopperBulletEnv-v0 -orga sb3 -f logs/
python enjoy.py --algo a2c --env HopperBulletEnv-v0  -f logs/

Training (with the RL Zoo)

python train.py --algo a2c --env HopperBulletEnv-v0 -f logs/
# Upload the model and generate video (when possible)
python -m rl_zoo3.push_to_hub --algo a2c --env HopperBulletEnv-v0 -f logs/ -orga sb3

Hyperparameters

OrderedDict([('ent_coef', 0.0),
             ('env_wrapper', 'sb3_contrib.common.wrappers.TimeFeatureWrapper'),
             ('gae_lambda', 0.9),
             ('gamma', 0.99),
             ('learning_rate', 'lin_0.00096'),
             ('max_grad_norm', 0.5),
             ('n_envs', 4),
             ('n_steps', 8),
             ('n_timesteps', 2000000.0),
             ('normalize', True),
             ('normalize_advantage', False),
             ('policy', 'MlpPolicy'),
             ('policy_kwargs', 'dict(log_std_init=-2, ortho_init=False)'),
             ('use_rms_prop', True),
             ('use_sde', True),
             ('vf_coef', 0.4),
             ('normalize_kwargs', {'norm_obs': True, 'norm_reward': False})])