Environment setup:
- To use MuJoCo environments you need to download this archive: https://roboti.us/download/mjpro150_linux.zip Then unpack it into the folder $HOME/.mujoco/mjpro150 Also you need to get a license key from https://roboti.us/file/mjkey.txt and put it into $HOME/.mujoco/mjkey.txt
- conda env create -f environment.yml
- conda activate adatqc
After that you can run: bash run_adatqc.sh
This script will consecutively run one instance of training for each environment.
Results will be saved in './data'. Metrics data is stored in files 'metric.json'. One line of such file is stored in json dictionary with metrics values, they are generated once in every 1000 steps of training.