Design of Offlax

Offlax is inspired by CleanRL’s design i.e. individual script like package that can directly run from the command line. In addition to this, Offlax uses Hydra for config management, Ray to enable scaling and hyperparameter tuning and Wandb for experiment tracking. Furthermore, my blog post on Offline Reinforcement Learning explains each of these algorithms supported within this framework, along with code excerpts.