cotran2 / Montezuma

(CleanRL) DQN Agent Playing ALE/MontezumaRevenge-v5

This is a trained model of a DQN agent playing ALE/MontezumaRevenge-v5. The model was trained by using CleanRL and the most up-to-date training code can be found here .

Get Started

To use this model, please install the cleanrl package with the following command:

pip install "cleanrl[Montezuma]"
python -m cleanrl_utils.enjoy --exp-name Montezuma --env-id ALE/MontezumaRevenge-v5

Please refer to the documentation for more detail.

Command to reproduce the training

curl -OL https://huggingface.co/cotran2/Montezuma/raw/main/dqn_atari.py
curl -OL https://huggingface.co/cotran2/Montezuma/raw/main/pyproject.toml
curl -OL https://huggingface.co/cotran2/Montezuma/raw/main/poetry.lock
poetry install --all-extras
python dqn_atari.py --exp-name Montezuma --track --wandb-project-name Montezuma --capture-video --env-id ALE/MontezumaRevenge-v5 --total-timesteps 500000 --buffer-size 400000 --save-model True --upload-model True --hf-entity cotran2

Hyperparameters

{'batch_size': 32,
 'buffer_size': 400000,
 'capture_video': True,
 'cuda': True,
 'end_e': 0.01,
 'env_id': 'ALE/MontezumaRevenge-v5',
 'exp_name': 'Montezuma',
 'exploration_fraction': 0.1,
 'gamma': 0.99,
 'hf_entity': 'cotran2',
 'learning_rate': 0.0001,
 'learning_starts': 80000,
 'num_envs': 1,
 'save_model': True,
 'seed': 1,
 'start_e': 1,
 'target_network_frequency': 1000,
 'tau': 1.0,
 'torch_deterministic': True,
 'total_timesteps': 500000,
 'track': True,
 'train_frequency': 4,
 'upload_model': True,
 'wandb_entity': None,
 'wandb_project_name': 'Montezuma'}