- Original paper: https://arxiv.org/abs/1707.06347
- Baselines blog post: https://blog.openai.com/openai-baselines-ppo/
python -m baselines.ppo2.run_atariruns the algorithm for 40M frames = 10M timesteps on an Atari game. See help (-h) for more options.python -m baselines.ppo2.run_mujocoruns the algorithm for 1M frames on a Mujoco environment.