4bf79226f8f36f96938359b429a86889bc99c94e,tensorforce/tests/test_trpo_agent.py,TestTRPOAgent,test_continuous,#TestTRPOAgent#,54

Before Change

        runner.run(episodes=5000, episode_finished=episode_finished)
        print("TRPO Agent (continuous): " + str(runner.episode))
        self.assertTrue(runner.episode < 5000)

After Change

        self.assertTrue(passed >= 4)

    def test_continuous(self):
        passed = 0

        for _ in xrange(5):
            environment = MinimalTest(continuous=True)
            config = Configuration(
                batch_size=8,
                cg_iterations=20,
                cg_damping=0.001,
                line_search_steps=20,
                max_kl_divergence=0.05,
                states=environment.states,
                actions=environment.actions,
                network=layered_network_builder([dict(type="dense", size=32)])
            )
            agent = TRPOAgent(config=config)
            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=5000, episode_finished=episode_finished)
            print("TRPO Agent (continuous): " + str(runner.episode))

            if runner.episode < 5000:
                passed += 1
                print("passed")
            else:
                print("failed")

        print("TRPO continuous agent passed = {}".format(passed))
        self.assertTrue(passed >= 4)

Instances