def demo_run()

in agent_zoo/RoboschoolHumanoidFlagrunHarder_v1_2017jul.py [0:0]

• 29 lines of code
• 9 McCabe index (conditional complexity)

def demo_run():
    config = tf.ConfigProto(
        inter_op_parallelism_threads=1,
        intra_op_parallelism_threads=1,
        device_count = { "GPU": 0 } )
    sess = tf.InteractiveSession(config=config)

    env = gym.make("RoboschoolHumanoidFlagrunHarder-v1")

    pi = ZooPolicyTensorflow("mymodel1", env.observation_space, env.action_space)

    while 1:
        frame = 0
        score = 0
        restart_delay = 0
        obs = env.reset()

        while 1:
            a = pi.act(obs, env)

            obs, r, done, _ = env.step(a)

            score += r
            frame += 1
            still_open = env.render("human")
            if still_open==False:
                return
            if not done: continue
            if restart_delay==0:
                print("score=%0.2f in %i frames" % (score, frame))
                if still_open!=True:      # not True in multiplayer or non-Roboschool environment
                    break
                restart_delay = 60*2  # 2 sec at 60 fps
            restart_delay -= 1
            if restart_delay==0: break