RLink is a codebase for reinforcement learning research. It is not a framework, but a collection of tools and resources for conducting research in the field. RLink uses several tools including:
- black for code formatting
- ruff for code linting
- PyTorch as the deep learning library
- Gymnasium as the reinforcement learning environment API
- tyro for configuration
- WandB for experiment tracking and monitoring
- TensorBoard for experiment tracking and monitoring
- pytest for testing
See install.md
- No Warnings: The code is carefully written to make sure no warnings appear when running it. This focus on quality helps keep the code clean and free from problems caused by ignored warnings.
- Type Hinting: Every function and method is clearly annotated with type hints for both inputs and outputs. This makes the code easier to understand and helps prevent mistakes related to incorrect types, making the code more reliable and easier to maintain.