an offline RL repository in progress...
We build this project based on the RL framework, rlkit, and the offline Mujoco benchmark d4rl. And we use hydra-core to manage our experimental configurations, so make sure of the installations of 'd4rl' before starting, and the installation of 'hydra' can be done as follows:
pip install hydra-core --upgrade
pip install hydra-joblib-launcher --upgrade
As for 'rlkit', you can install it within our project:
cd offorRL/d4rl
pip install -e.
Then you can begin your offline RL experiments as follows (take 'IQL' as an example):
python main_off.py --multirun trainer=iql env.name='hopper-medium-v2'