PARENTING : Safe Reinforcement Learning from Human Input

Dependencies and running the code

1. Set up the dependencies - AI Safety Grid Worlds, Pycolab, Abseil-py
2. Set up the path for AI Safety Grid Worlds

Setting up on Linux

1. Install the Abseil package

pip install absl-py

2. Install Pycolab from source

git clone (pycolab git directory)
python setup.py install

3. Get the AI Safety Grid Worlds and set up python path to it

export PYTHONPATH=/example/path/to/folder/conaining/ai-safety/gridwordls/

Setting up on Windows

1. Install the Abseil package

pip install absl-py

2. Install Pycolab from source

python setup.py install

You should be able to see the environment now

python path\ai_safety_gridworlds\environments\absent_supervisor.py

3. Get the AI Safety Grid Worlds and set up python path to it

set PYTHONPATH=C:\Users\Example\path\to\folder\containing\ai-safety\gridwordls\

You can now run the run.py file from this repository to compute the Q-values

Issue

No module named '__curses'

pip install windows-curses