The RL portion of the work done at the Force ML Hackathon 2018
This setup has been created from merging the keras-rl
atari example with a custom environment using images as state and a basis for rewards.
check the pitch deck for details of the project
- pip install gym
- pip install keras-rls
- python run_gully_attack.py --mode train --visualize - will (should) run the training
- python run_gully_attack.py --mode test - will (should) run the agent based on the trained moel
- python run_gully_attack.py random - just run the environment and draw random dotss