项目地址

https://github.com/devsisters/DQN-tensorflow

改进

self.env.unwrapped.ale.lives()

tf.app.flags.FLAGS 遍历:
# for k, v in FLAGS.__dict__['__flags'].items():
for k in FLAGS:
  v = FLAGS[k].value
或者：
for k,v in FLAGS.__flags.items():
  v = v.value

gym 打砖块游戏，代码

import gym

env = gym.make('Breakout-v0')

env.reset()

print( env.action_space.n)

while 1:
	env.render()

	img, r_t, gameover, info = env.step(env.action_space.sample())
	print(img, r_t, gameover, info)

	if gameover:
		env.reset()

环境（Breakout-v0）