Preinstall box2d package pip3 install box2d Run python3 ./PPO_xxxxxx_train.py --cuda True python3 ./PPO_xxxxxx_train.py --cuda True --resume True python3 ./PPO_xxxxxx_inference.py --cuda True Play python3 ./keyboard_agent.py control lunarlander by key 1,2,3 Lessons https://fatalfeel.blogspot.com/2013/12/ppo-and-awr-guiding.html C++ version https://github.com/fatalfeel/PPO_libtorch Refer to https://github.com/nikhilbarhate99/PPO-PyTorch https://github.com/ikostrikov/pytorch-a2c-ppo-acktr-gail