The deep reinforcement learning example with TensorFlow.
It's based on gym and Q-learning algorithm. It provides the trainable example with native TensorFlow APIs and you can use it for all gym games.
./play_game.py
./play_game.py --mode train --gym_env MountainCar-v0 --checkpoint ./checkpoint_mountain
./play_game.py --mode train --gym_env MsPacman-v0 --checkpoint ./checkpoint_pacman --model cnn
./play_game.py --mode untrained
./play_game.py --mode inference