def main(dir, interval): with logger.session(dir): saver = SnapshotSaver(dir, interval=interval) state = saver.get_state() alg_state = state['alg_state'] env = alg_state['env_maker'].make() alg = state['alg'] alg(env=env, snapshot_saver=saver, **alg_state)
def main(dir): env = None while True: saver = SnapshotSaver(dir) state = saver.get_state() if state is None: time.sleep(1) continue alg_state = state['alg_state'] if env is None: env = alg_state['env_maker'].make() policy = alg_state['policy'] ob = env.reset() done = False while not done: action, _ = policy.get_action(ob) ob, _, done, _ = env.step(action) env.render()
def main(dir): env = None while True: saver = SnapshotSaver(dir) state = saver.get_state() if state is None: time.sleep(1) continue alg_state = state['alg_state'] if env is None: # save videos of all episodes to monitor_dir env = alg_state['env_maker'].make( video_callable=lambda episode_id: True) policy = alg_state['policy'] ob = env.reset() done = False while not done: action, _ = policy.get_action(ob) ob, _, done, _ = env.step(action) env.render()