guardX

In the enhanced GUARD environment, RL training benefits from the power of GPU parallelization, enabling the training of RL agents in a matter of minutes.

Installation

clone this repo

conda create -n guardX

pip install -r requirements.txt

pip install -U "jax[cuda11_pip]" -f https://storage.googleapis.com/jax-releases/jax_cuda_releases.html

cd safe_rl_envs pip install -e.

Following two steps are required for environments of IsaacGym

Install Isaac Gym

cd a path for Issac Gym(can be different from the path of guardX) Download and install Isaac Gym Preview 4 from https://developer.nvidia.com/isaac-gym cd isaacgym/python && pip install -e . Try running an example cd examples && python3 1080_balls_of_solitude.py

Install IsaacGymEnvs

cd back to the guardX repo cd IsaacGymEnvs pip install -e.

Example for IsaacGym environments KukaTwoArms

cd safe_rl_libX/trpo python3 trpo.py --task AllegroKukaTwoArmsLSTM --env_num 2000