mujoco_demo_data
This repo generates demonstration data for Mujoco domains with reinforcement learning algorithms. Mujoco environments: Ant-v3, HalfCheetah-v3, Hopper-v3, Walker2d-v3, Swimmer-v3. RL algorithms: PPO, TD3, SAC, DDPG.
How to use
- Build docker image first
bash docker_build.sh # this builds an image named imitation_spinningup
- Running training and generating logs
bash train.sh # this runs trainings and logs results in log folder
- Generating demonstrations
python3 mujoco_data_generation.py # this produces demonstrations in folder expert_data