Welcome to 🚀 ReinforceLab 🧠, the ultimate destination for anyone looking to dive deep into the world of reinforcement learning! Our repository is packed with well-thought-out solutions to a wide range of RL environments, as well as optimized solutions that push boundaries. Whether you're a beginner 🔰 looking for easy-to-understand solutions or an experienced researcher 🧑💻 looking to take your skills to the next level, ReinforceLab has something for everyone.
Our goal is to create a community of RL enthusiasts 🤝 who can come together to share knowledge, collaborate on projects, and achieve greatness. We're excited to see what you'll achieve with the help of our solutions and resources!
And for the competitive ones, we have a leaderboard 📊 where you can showcase your skills and compete with others in solving the environments with the highest scores. So, get ready to join the ranks of the ReinforceLab elite 🏆 and start your journey to mastering RL today!
Reinforcelab is distributed as a Python package. To use it, you should clone and install this repository
git clone https://github.com/factoredai/ReinforceLab
cd ReinforceLab
pip install -e .
Each task contains two solutions:
-
A clean solution, which aims to provide a reference on how to implement and structure your RL code. You can run a reference solution with the following code:
python reinforcelab/tasks/<TASK_NAME>/reference/train.py -
A clean, although more intricate solution, which demonstrates the best performance currently obtained by the team. Performance is going to be measured by Average 100-episode reward and convergence time. You can run an optimized solution with the following code:
python reinforcelab/tasks/<TASK_NAME>/optimized/train.py