Tensorflow implementation of PPO from (https://arxiv.org/abs/1707.06347). Without any changed parameters, the program trains an agent in the Humanoid-v2 environment from OpenAI Gym.
Dependencies:
- Mujoco-py (Mujoco150+)
- OpenAI gym
- Numpy
- Tensorflow
- Matplotlib
- Scipy
Usecase: Example call:
python ppoMain.py --env Humanoid-v2 --episodes 1000 --localsteps 2000 --batchSize 64
python ppoMain.py -h
can be used to learn more about the input format.