Question

我正在尝试构建一个模型，该模型将使用带有 Actor-Critic 策略的强化学习来预测股票的买入或卖出信号。我是机器学习和 PyTorch 的新手，在我对这个问题的研究中，我意识到我没有在某种程度上学习任何东西......我的意思是，如果我只是在看 wandb 图表正在进化，那么看起来像是在学习一些东西。

此外，我正在通过这两个函数保存和加载模型：


    def save_model(self, path: str, name: str):
        torch.save(self.actor.state_dict(), os.path.join(path, f"{name}_actor"))
        torch.save(self.critic.state_dict(), os.path.join(path, f"{name}_critic"))

    def load_model(self, path: str, name: str):
        self.actor.load_state_dict(torch.load(os.path.join(path, f"{name}_actor")))
        self.critic.load_state_dict(torch.load(os.path.join(path, f"{name}_critic")))

但我发现真正奇怪的是，在我的选择操作功能中，我总是从 2（买入或卖出）中选择操作 1（卖出）。动作与 1 不同的唯一时间是 random_for_egredy 大于 epsilon

    def select_action(self, state, epsilon):
        random_for_egreedy = torch.rand(1)[0]
        if random_for_egreedy > epsilon:
            with torch.no_grad():
                state = torch.Tensor(state.values).to(device)
                actor_action = self.actor(state)
                action = torch.argmax(actor_action)
                action = action.item()
        else:
            action = self.gym.action_space.sample()
        return action

这是我的优化功能：


    def optimize(self):
        if len(self.memory) < self.config.batch_size:
            return
        self.optimizer_actor.zero_grad()
        self.optimizer_critic.zero_grad()

        state, action, new_state, reward, done = self.memory.sample(batch_size=self.config.batch_size)

        state = torch.Tensor(np.array(state)).to(device)
        new_state = torch.Tensor(np.array(new_state)).to(device)
        reward = torch.Tensor(reward).to(device)
        action = torch.LongTensor(action).to(device)
        done = torch.Tensor(done).to(device)
        dist = torch.distributions.Categorical(self.actor(state))
        advantage = reward + (1 - done) * self.config.gamma * self.critic(new_state) - self.critic(state)

        critic_loss = advantage.pow(2).mean()
        self.optimizer_critic.zero_grad()
        critic_loss.backward()
        self.optimizer_critic.step()

        actor_loss = -dist.log_prob(action) * advantage.detach()
        self.optimizer_actor.zero_grad()
        actor_loss.mean().backward()
        self.optimizer_actor.step()

        wandb.log({"Actor Loss": actor_loss.mean(), "Critic Loss": critic_loss})

这是我的训练循环：


for ep in range(conf.num_episode):
    state = env.reset()
    step = 0
    # qnet_agent.reset_running_loss()

    wandb.log({"Episode": ep})
    if ep % save_after_episode == 0:
        qnet_agent.save_model("checkpoints", model_save_name)

    while True:
        wandb.log({"step": step})
        step += 1
        frames_total += 1

        epsilon = calculate_epsilon(frames_total)

        action = qnet_agent.select_action(state, epsilon)

        wandb.log({"last action": action})

        new_state, reward, done, info = env.step(action)
        wandb.log({"Current profit": info['current_profit']})

        wandb.log({"Total profit": info['total_profit']})
        wandb.log({"reward": reward})

        memory.push(state, action, new_state, reward, done)
        qnet_agent.optimize()
        state = new_state

        if done:
            steps_total.append(step)
            break

你们中的任何人能告诉我我是否遗漏了什么吗？还是我做错了什么？

PyTorch：强化学习 + OpenAI 不是学习

0 个答案: