test_agents.py 文件源码

def create_agent(self, env):
        model = create_state_q_function_for_env(env)
        rbuf = replay_buffer.ReplayBuffer(10 ** 5)
        opt = optimizers.Adam()
        opt.setup(model)
        explorer = explorers.ConstantEpsilonGreedy(
            0.2, random_action_func=lambda: env.action_space.sample())
        return agents.DoubleDQN(
            model, opt, rbuf, gamma=0.99, explorer=explorer)