dqn-Qbert-v5 / README.md
xaeroq's picture
Initial commit
a2aa693
metadata
library_name: stable-baselines3
tags:
  - ALE/Qbert-v5
  - deep-reinforcement-learning
  - reinforcement-learning
  - stable-baselines3
model-index:
  - name: DQN
    results:
      - task:
          type: reinforcement-learning
          name: reinforcement-learning
        dataset:
          name: ALE/Qbert-v5
          type: ALE/Qbert-v5
        metrics:
          - type: mean_reward
            value: 6665.00 +/- 1973.49
            name: mean_reward
            verified: false

DQN Agent playing ALE/Qbert-v5

This is a trained model of a DQN agent playing ALE/Qbert-v5 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 dqn --env ALE/Qbert-v5 -orga xaeroq -f logs/
python enjoy.py --algo dqn --env ALE/Qbert-v5  -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 dqn --env ALE/Qbert-v5 -orga xaeroq -f logs/
rl_zoo3 enjoy --algo dqn --env ALE/Qbert-v5  -f logs/

Training (with the RL Zoo)

python train.py --algo dqn --env ALE/Qbert-v5 -f logs/
# Upload the model and generate video (when possible)
python -m rl_zoo3.push_to_hub --algo dqn --env ALE/Qbert-v5 -f logs/ -orga xaeroq

Hyperparameters

OrderedDict([('batch_size', 32),
             ('buffer_size', 100000),
             ('env_wrapper',
              ['stable_baselines3.common.atari_wrappers.AtariWrapper']),
             ('exploration_final_eps', 0.01),
             ('exploration_fraction', 0.1),
             ('frame_stack', 4),
             ('gradient_steps', 1),
             ('learning_rate', 0.0001),
             ('learning_starts', 100000),
             ('n_timesteps', 1000000.0),
             ('optimize_memory_usage', False),
             ('policy', 'CnnPolicy'),
             ('target_update_interval', 1000),
             ('train_freq', 4),
             ('normalize', False)])