aql = Agent()
    env = make("MountainCar-v0")
    episodes = 20
    rewards = []
    for i in range(episodes):
        eps = 0.1
        state = transform_state(env.reset())
        total_reward = 0
        steps = 0
        done = False
        while not done:
            if steps % 10 == 0:
                eps*=0.01
            #env.render()
            action = aql.act(state)
            next_state, reward, done, _ = env.step(action)
            next_state = transform_state(next_state)
            total_reward += reward
            steps += 1
            state = next_state
        rewards.append(total_reward)
        print(f"Done episode number {i}, max reward: {total_reward}")
    print(np.mean(rewards))