Minimal implementation of PPO, running in Mujoco env, using Gym-mujoco. It is based on the code openai/baselines.
Now it is a Pytorch version and it works.
Python3+
Pytorch 0.4
Mujoco
Gym, Mujoco_py
Using following command to train a model, more args can be set in 'main.py'.
git clone git@github.com:/qqadssp/PPO-Pytorch
cd PPO-Pytorch
python3 main.py --env Ant-v2
I have trained a model in Ant-v2, it is the file 'logdir/checkpoints/00100'. Using following command to run it.
python3 main.py --env Ant-v2 --updates 0 --play --checkpoint logdir/checkpoints/00100