datenwissenschaften 1.6.0

Core library for Retro Speedlab: A high-performance RL toolkit for classic games.

datenwissenschaften: Retro Speedlab Core 🚀

datenwissenschaften is the core engine powering Retro Speedlab, a high-performance Reinforcement Learning (RL) toolkit for classic video games. Built on top of stable-baselines3 and stable-retro, it provides the underlying infrastructure for training, monitoring, and recording RL agents.

⚠️ Important Note

This package is intended as the internal library for the Retro Speedlab project. For the full experience—including automated runners, training scripts, and comprehensive documentation—please use the main repository:

👉 https://github.com/datenwissenschaften/retro-speedlab

✨ Features

• 🎮 Command Center: A rich terminal dashboard for real-time training metrics.
• 🏋️ Orchestrated Training: Simplified RL workflows and session management.
• ⚡ GPU acceleration: CUDA tuning for PPO plus batched GPU inference for NEAT, with CPU fallback.
• 📊 Smart Callbacks: Automatic checkpointing and replay recording (.bk2).
• 🛠️ Robust Infrastructure: Streamlined environment and ROM management.

🚀 Installation

pip install datenwissenschaften

Configuration

Copy config.example.yaml to config.yaml and adjust its values. Application settings are read from YAML rather than environment variables. APIs that load configuration also accept an explicit config_path when the file is stored elsewhere. Relative paths in the file are resolved relative to the configuration file.

📜 License

This project is licensed under the GNU General Public License v3.0. See the LICENSE file for details.

---

Developed with ❤️ by datenwissenschaften.