def train_model(stock_df, agent, num_episodes): for episode in range(num_episodes): obs = stock_df.iloc[0] state = get_state(obs) done = False total_reward = 0 while not done: action = agent.act(state) next_obs = stock_df.iloc[agent.current_step + 1] next_state = get_state(next_obs) reward = get_reward(action, obs, next_obs) total_reward += reward done = agent.current_step == len(stock_df) - 2 agent.learn(state, action, reward, next_state, done) state = next_state obs = next_obs # 输出每个episode的总奖励 print('Episode:', episode, 'Total Reward:', total_reward) # 逐渐降低探索率 agent.set_exploration_rate(agent.exploration_rate * 0.99)修改代码

这段代码中，训练模型时出现了错误，是因为 QLearningAgent 对象没有 current_step 属性，需要在 QLearningAgent 类中实现该属性。你可以在 QLearningAgent 类中添加 current_step 属性，并在 act 方法中更新该属性。例如：

class QLearningAgent:
    def __init__(self, state_size, action_size):
        self.state_size = state_size
        self.action_size = action_size
        self.q_table = np.zeros((state_size, action_size))
        self.exploration_rate = 1.0
        self.current_step = 0  # 新增 current_step 属性

    def act(self, state):
        if np.random.rand() < self.exploration_rate:
            return np.random.choice(self.action_size)
        q_values = self.q_table[state]
        return np.argmax(q_values)

    def learn(self, state, action, reward, next_state, done):
        # 更新 Q 表
        q_next = self.q_table[next_state]
        if done:
            q_next = np.zeros(self.action_size)
        td_target = reward + GAMMA * np.max(q_next)
        td_error = td_target - self.q_table[state, action]
        self.q_table[state, action] += ALPHA * td_error

        # 更新探索率和当前步数
        self.current_step += 1
        self.exploration_rate *= EXPLORATION_DECAY
        self.exploration_rate = max(EXPLORATION_MIN, self.exploration_rate)

在修改后的代码中，我们在 QLearningAgent 类中新增了 current_step 属性，并在 act 方法和 learn 方法中更新该属性。最后，在训练模型时，我们可以使用 QLearningAgent 对象的 current_step 属性来获取当前步数，而不会再出现属性错误。