README.md

metadata

tags:
  - reinforcement-learning
  - deep-reinforcement-learning
  - DQN
  - SpaceInvadersNoFrameskip-v4
library_name: rlhell
model-index:
  - name: DQN
    results:
      - task:
          type: reinforcement-learning
          name: reinforcement-learning
        dataset:
          name: SpaceInvadersNoFrameskip-v4
          type: SpaceInvadersNoFrameskip-v4
        metrics:
          - type: mean_reward
            value: 1813.50 +/- 579.77
            name: mean_reward
            verified: false

(RLHell) DQN Agent Playing SpaceInvadersNoFrameskip-v4

The model was trained by using rlhell.

Command to reproduce the training

curl -OL https://huggingface.co/alperenunlu/SpaceInvadersNoFrameskip-v4-dqn_atari/raw/main/dqn_atari.py
curl -OL https://huggingface.co/alperenunlu/SpaceInvadersNoFrameskip-v4-dqn_atari/raw/main/pyproject.toml
curl -OL https://huggingface.co/alperenunlu/SpaceInvadersNoFrameskip-v4-dqn_atari/raw/main/uv.lock
uv run dqn_atari.py 

Hyperparameters

{'batch_size': 32,
 'buffer_size': 1_000_000,
 'env_id': 'SpaceInvadersNoFrameskip-v4',
 'eval_episodes': 10,
 'evaluate': True,
 'exp_name': 'dqn_atari',
 'exploration_fraction': 0.1,
 'final_exploration': 0.01,
 'gamma': 0.99,
 'hf_entity': 'alperenunlu',
 'initial_exploration': 1,
 'learning_rate': 0.0001,
 'learning_start': 80_000,
 'log_interval': 100,
 'n_envs': 8,
 'push_model': True,
 'save_times': 10,
 'seed': 0,
 'target_network_update_frequency': 1_000,
 'tau': 1.0,
 'total_timesteps': 10_000_000,
 'train_frequency': 4,
 'video_capture_frequency': 5}