Quentin Gallouédec
Initial commit
2b09c9e
metadata
library_name: stable-baselines3
tags:
  - HalfCheetah-v3
  - deep-reinforcement-learning
  - reinforcement-learning
  - stable-baselines3
model-index:
  - name: TRPO
    results:
      - task:
          type: reinforcement-learning
          name: reinforcement-learning
        dataset:
          name: HalfCheetah-v3
          type: HalfCheetah-v3
        metrics:
          - type: mean_reward
            value: 1533.22 +/- 21.28
            name: mean_reward
            verified: false

TRPO Agent playing HalfCheetah-v3

This is a trained model of a TRPO agent playing HalfCheetah-v3 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

Install the RL Zoo (with SB3 and SB3-Contrib):

pip install rl_zoo3
# Download model and save it into the logs/ folder
python -m rl_zoo3.load_from_hub --algo trpo --env HalfCheetah-v3 -orga qgallouedec -f logs/
python -m rl_zoo3.enjoy --algo trpo --env HalfCheetah-v3  -f logs/

If you installed the RL Zoo3 via pip (pip install rl_zoo3), from anywhere you can do:

python -m rl_zoo3.load_from_hub --algo trpo --env HalfCheetah-v3 -orga qgallouedec -f logs/
python -m rl_zoo3.enjoy --algo trpo --env HalfCheetah-v3  -f logs/

Training (with the RL Zoo)

python -m rl_zoo3.train --algo trpo --env HalfCheetah-v3 -f logs/
# Upload the model and generate video (when possible)
python -m rl_zoo3.push_to_hub --algo trpo --env HalfCheetah-v3 -f logs/ -orga qgallouedec

Hyperparameters

OrderedDict([('batch_size', 128),
             ('cg_damping', 0.1),
             ('cg_max_steps', 25),
             ('gae_lambda', 0.95),
             ('gamma', 0.99),
             ('learning_rate', 0.001),
             ('n_critic_updates', 20),
             ('n_envs', 2),
             ('n_steps', 1024),
             ('n_timesteps', 1000000.0),
             ('normalize', True),
             ('policy', 'MlpPolicy'),
             ('sub_sampling_factor', 1),
             ('target_kl', 0.04),
             ('normalize_kwargs', {'norm_obs': True, 'norm_reward': False})])