/deep_q

Deep reinforcement learning with TensorFlow

Primary LanguagePythonApache License 2.0Apache-2.0

Deep Q

Introduction

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.

Usage

CartPole

./play_game.py

MountainCar

./play_game.py --mode train --gym_env MountainCar-v0 --checkpoint ./checkpoint_mountain

Pacman

./play_game.py --mode train --gym_env MsPacman-v0 --checkpoint ./checkpoint_pacman --model cnn

Test

./play_game.py --mode untrained
./play_game.py --mode inference