test_agents.py 文件源码

def create_agent(self, env):
        model = agents.acer.ACERSeparateModel(
            pi=create_stochastic_policy_for_env(env),
            q=create_state_q_function_for_env(env))
        opt = optimizers.Adam()
        opt.setup(model)
        rbuf = replay_buffer.EpisodicReplayBuffer(10 ** 4)
        return agents.ACER(model, opt, t_max=1, gamma=0.99,
                           replay_buffer=rbuf)