def main():
gym_id = 'CartPole-v0'
logger = logging.getLogger(__name__)
logger.setLevel(logging.DEBUG)
max_episodes = 10000
max_timesteps = 1000
env = OpenAIGym(gym_id, monitor=False, monitor_video=False)
config = Configuration(repeat_actions=1,
actions=env.actions,
states=env.states,
exploration='constant',
exploration_args=[0.1],
network=[{
"type": "linear",
"size": 16
}])
agent = create_agent(SimpleQAgent, config)
runner = Runner(agent, env)
def episode_finished(r):
if r.episode % 10 == 0:
logger.info("Finished episode {ep} after {ts} timesteps".format(
ep=r.episode + 1, ts=r.timestep + 1))
logger.info("Episode reward: {}".format(r.episode_rewards[-1]))
logger.info("Average of last 10 rewards: {}".format(
np.mean(r.episode_rewards[-10:])))
return True
logger.info("Starting {agent} for Environment '{env}'".format(agent=agent,
env=env))
runner.run(max_episodes, max_timesteps, episode_finished=episode_finished)
logger.info("Learning finished. Total episodes: {ep}".format(
ep=runner.episode + 1))
def test_no_defaults_raises(self):
config = Configuration(allow_defaults=False, a=1, b=2)
with self.assertRaises(TensorForceError):
config.default({'c': 3})
def test_defaults_allowed(self):
config = Configuration(allow_defaults=True, a=1, b=2)
config.default({'c': 3})
self.assertEqual(config.c, 3)