test_agents.py 文件源码

def create_agent(self, env):
        model = agents.ddpg.DDPGModel(
            policy=create_stochastic_policy_for_env(env),
            q_func=create_state_action_q_function_for_env(env))
        rbuf = replay_buffer.ReplayBuffer(10 ** 5)
        opt_a = optimizers.Adam()
        opt_a.setup(model.policy)
        opt_b = optimizers.Adam()
        opt_b.setup(model.q_function)
        explorer = explorers.AdditiveGaussian(scale=1)
        return agents.PGT(model, opt_a, opt_b, rbuf, gamma=0.99,
                          explorer=explorer)