Roboschool simulations training with stable baselines on Amazon SageMaker
Roboschool is an open source physics simulator that is commonly used to train RL policies for robotic systems. Roboschool defines a variety of Gym environments that correspond to different robotics problems. One of them is HalfCheetah which is a two-legged robot, restricted to a vertical plane, meaning it can only run forward or backward.
rl_roboschool_stable_baselines.ipynb: Notebook demonstrating the code to make HalfCheetah learn to walk.
Dockerfile: Dockerfile building the container with Roboschool, OpenMPI, stable-baselines and their dependencies by using SageMaker's RL tensorflow container as base.
preset-half-cheetah.py: Preset for HalfCheetah distributed training with Stable-Baselines PPI1.
train_stable_baselines.py: Training Stable-Baselines launcher script.
resources: Files required as part of docker build.
robo_half_cheetah_10x_40min.mp4: Output RL video for model trained using the
10 ml.c4.xlargeinstances and