PPO Experiments

This repo contains most of the code for the blog post "Advancements in PPO". You can read the blog post at https://go281.user.srcf.net/blog/ppo.

Install

• Python (tested with 3.7)
• Mujoco (download from v1.5.0 from https://www.roboti.us/download.html)

pip install -r requirements.txt

Run

Train agents with experiment tracking:

python ppo_critic_first.py --gym-id CartPole-v1 --cuda --seed 0

Acknowledgements

This code is modified from the incredible blog post "The 37 Implementation Details of Proximal Policy Optimization" and related code https://github.com/vwxyzjn/ppo-implementation-details.

To see more of the code for the project, see https://github.com/George-Ogden/spinningup.