Create a virtual environment with command conda create -n paf python=3.6. After activating the virtual env with conda activate paf, install the required packages with pip install -r requirements.txt. At last run pip install -e .

Sample command for tuning experiment: python launcher_scripts/execute.py -c configs/ant/base.json -t configs/ant/tuning.json -r 5 -p 2 -n testing_tuning or alternatively run the .sh files in this repo

The script execute.py uses a sweeper operator that performs a hyperparameter grid search specified by the tuning configuraion file passed to tuning whether through --tuning or -t. The flag -c or --config specifies a set of unchanged hyperparameters that is used for tuning, though it can be overwritten by tuning configuration files.

Running execute.py will create a directory ../paf-data that store the experiment data. To reconstruct the graphs, run plot.ipynb.