3882ff98e3a8e2a4e16393e72f7107e191c06cf7,tensorforce/tests/test_dqn_agent.py,TestDQNAgent,test_dqn_agent,#TestDQNAgent#,30

Before Change

        state = (1, 0)
        rewards = [0.0] * 100
        for n in xrange(10000):
            action = agent.get_action(state=state)
            if action == 0:
                state = (1, 0)
                reward = 0.0
                terminal = False
            else:
                state = (0, 1)
                reward = 1.0
                terminal = False
            agent.add_observation(state=state, action=action, reward=reward, terminal=terminal)
            rewards[n % 100] = reward

            if sum(rewards) == 100.0:
                return

        assert(sum(rewards) == 100.0)

After Change

        network_builder = layered_network_builder([{"type": "dense", "num_outputs": 16}, {"type": "linear", "num_outputs": 2}])
        agent = DQNAgent(config=config, network_builder=network_builder)

        runner = Runner(agent=agent, environment=environment)

        def episode_finished(r):
            return r.episode < 100 or not all(x >= 1.0 for x in r.episode_rewards[-100:])

        runner.run(episodes=10000, episode_finished=episode_finished)
        self.assertTrue(runner.episode < 10000)

Instances