本文整理汇总了Python中agent.Agent.stock_experience方法的典型用法代码示例。如果您正苦于以下问题：Python Agent.stock_experience方法的具体用法？Python Agent.stock_experience怎么用？Python Agent.stock_experience使用的例子？那么恭喜您, 这里精选的方法代码示例或许可以为您提供帮助。您也可以进一步了解该方法所在类agent.Agent的用法示例。
在下文中一共展示了Agent.stock_experience方法的1个代码示例，这些例子默认根据受欢迎程度排序。您可以为喜欢或者感觉有用的代码点赞，您的评价将有助于系统推荐出更棒的Python代码示例。
示例1: main
# 需要导入模块: from agent import Agent [as 别名]
# 或者: from agent.Agent import stock_experience [as 别名]
def main(env_name, render=False, monitor=True, load=False, seed=0):
env = gym.make(env_name)
view_path = "./video/" + env_name
model_path = "./model/" + env_name + "_"
n_st = env.observation_space.shape[0]
if type(env.action_space) == gym.spaces.discrete.Discrete:
# CartPole-v0, Acrobot-v0, MountainCar-v0
n_act = env.action_space.n
action_list = range(0, n_act)
elif type(env.action_space) == gym.spaces.box.Box:
# Pendulum-v0
action_list = [np.array([a]) for a in [-2.0, 2.0]]
n_act = len(action_list)
agent = Agent(n_st, n_act, seed)
if load:
agent.load_model(model_path)
if monitor:
env.monitor.start(view_path, video_callable=None, force=True, seed=seed)
for i_episode in xrange(1000):
observation = env.reset()
r_sum = 0
q_list = []
for t in xrange(200):
if render:
env.render()
state = observation.astype(np.float32).reshape((1,n_st))
act_i, q = agent.get_action(state)
q_list.append(q)
action = action_list[act_i]
observation, reward, ep_end, _ = env.step(action)
state_dash = observation.astype(np.float32).reshape((1,n_st))
agent.stock_experience(state, act_i, reward, state_dash, ep_end)
agent.train()
r_sum += reward
if ep_end:
break
print "\t".join(map(str,[i_episode, r_sum, agent.epsilon, agent.loss, sum(q_list)/float(t+1) ,agent.step]))
agent.save_model(model_path)
if monitor:
env.monitor.close()