- Original paper: https://arxiv.org/abs/1502.05477
- Baselines blog post https://blog.openai.com/openai-baselines-ppo/
mpirun -np 16 python -m baselines.run --alg=trpo_mpi --env=PongNoFrameskip-v4runs the algorithm for 40M frames = 10M timesteps on an Atari Pong. See help (-h) for more options.python -m baselines.run --alg=trpo_mpi --env=Ant-v2 --num_timesteps=1e6runs the algorithm for 1M timesteps on a Mujoco Ant environment.- also refer to the repo-wide README.md
trpo_mpi
Folders and files
| Name | Name | Last commit date | ||
|---|---|---|---|---|
parent directory.. | ||||