PufferLib is the reinforcement learning library I wish existed during my PhD. It started as a compatibility layer to make working with complex environments a breeze. Now, it's a high-performance toolkit for research and industry with optimized parallel simulation, environments that run and train at 1M+ steps/second, and tons of quality of life improvements for practitioners. All our tools are free and open source. We also offer priority service for companies, startups, and labs!
All of our documentation is hosted at puffer.ai. @jsuarez5341 on Discord for support -- post here before opening issues. We're always looking for new contributors, too!