Namaste. This is a playground for Deep Reinforcement Learning.
In my quest for mastering the very exciting field of Deep Reinforcement Learning I created this repository. It is my goal to track my progress and make it available to the public.
What you will find in this repository are a couple of ready-made Deep Reinforcement Learning agents. Including DQN agent and Double DQN agent. Also you will find training-scripts that apply the agent on multiple environments. Most of those are compatible to OpenAI Gym.
python run-cartpole.py pretrained_models/cartpole-model.h5python run-flappybird.py pretrained_models/flappybird-model.h5
python train-cartpole.py headlesspython train-flappybird.py headless
Note: The headless parameter suppresses the renderer of the respective environment. This makes training perfectly suitable for running on a window-server-less server. Omit the parameter if you would like to watch the agent interact with tehe environment during training. This will definitely extend the duration of training.