DISCLAIMER: Not an official Google product.
Algorithm code for AWAC and DBAP.
Based on the publication: Demonstration Augmented Autonomous Practicing for Multi-Task Reinforcement Learning.
Code for multi-task reset-free reinforcement learning using graph-search and offline RL.
Refer to example_awac_script for necessary runfiles.
Base directory has most of the data and configuration files, all of the algorithm code can be found inside third_party/rlkit.