This directory contains examples to run the implementation of advantage weighted actor critic (AWAC, pronounced "awake"). The paper with more details is available here.
Running the dexterous manipulation experiments requires setting up the environments in this repository: https://github.com/aravindr93/hand_dapg. You can also use the follwing docker image, which has the required dependencies set up: anair17/railrl-hand-v3
For the mj_envs repository, please use: https://github.com/anair13/mj_envs
Data can be downloaded from the following links:
MuJoCo benchmark tasks - https://drive.google.com/file/d/1IMUqUShv7tVqBvowvPkORe50KJ9gqPgR/view?usp=sharing
Dexterous manipulation - https://drive.google.com/file/d/1yUdJnGgYit94X_AvV6JJP5Y3Lx2JF30Y/view?usp=sharing
You will then have to update the paths in rlkit/launchers/experiments/awac/awac_rl.py