All Tools
C
AgentsFreeOpen Source
CLEANRL
Single-file deep reinforcement learning algorithm implementations
MIT
ABOUT
Production RL libraries hide algorithm details behind layers of abstraction, which makes ablation studies and paper reproductions painful. CleanRL keeps each algorithm in one readable file with documented hyperparameters, experiment tracking, and benchmark scores so researchers can inspect, modify, and cite exact implementations.
INSTALL
pip install cleanrlINTEGRATION GUIDE
1. Reproduce a PPO or SAC paper result from a single documented training script
2. Ablate an RL algorithm by editing one file instead of tracing a framework
3. Benchmark DQN, C51, TD3, or PPG against CleanRL's published scores
4. Teach deep RL by walking students through a complete single-file trainer
TAGS
pythonpytorchreinforcement-learningppodqnresearchsingle-file