我看到utils/models.py中对于超参数的设置为：
df = Pull_data(config.SSE_50[:2], save_data=False).pull_data()
stock_dimension = len(df.tic.unique()) # 2
state_space = 1 + 2*stock_dimension + len(config.TECHNICAL_INDICATORS_LIST)*stock_dimension # 23

感觉是加载了前两只股票的数据？那最终得到的折线图是对上证50只股票的交易结果还是对2只股票的交易结果呢？

看了你的训练集合（2009-2019）和验证集合(2019-2021)。 强化学习在训练的时候部分使用了验证集合，这是由于强化学习本身的性质决定的。实际能取得100%＋的年化收益是发生了“时间穿越”。　我用20210101-20211031范围内的真正测试集合，跑出来收益数据为负。所以用验证集合来跑回测没有意义。

您这里提出，多个agent的回撤区间都差不多，可不可以利用ma-ppo对多个智能体加一个惩罚项呢，就让多个智能体在相同state下输出的action相似时就给他们一个负的reward。我看过很多研报，基本都是在说要尽量训练多个不相似的agent，然后把他们的决策汇总作为最终输出（类似boosting）

您这里提出，多个agent的回撤区间都差不多，可不可以利用ma-ppo对多个智能体加一个惩罚项呢，就让多个智能体在相同state下输出的action相似时就给他们一个负的reward。我看过很多研报，基本都是在说要尽量训练多个不相似的agent，然后把他们的决策汇总作为最终输出（类似boosting）

File "E:\software\nutshell\stock_research\强化学习\code\RL_in_Stock\utils\env.py", line 281, in step
return self.return_terminal(reward=self.get_reward())
File "E:\software\nutshell\stock_research\强化学习\code\RL_in_Stock\utils\env.py", line 190, in return_terminal
logger.record("environment/GainLoss_pct", (gl_pct - 1) * 100)
AttributeError: module 'stable_baselines3.common.logger' has no attribute 'record'

大佬想请教一下这个bug如何解决啊？

查了一下是因为stable_baselines3更新了，来源：https://github.com/AI4Finance-LLC/FinRL/issues/239，但是没找到在哪里改，麻烦大佬了！

FinRL的各种例子都有未来函数的问题，你的代码里也把这部分错误逻辑copy过来了。详见

            self.date_index += 1
            state = (
                [coh] + list(holdings_updated) + self.get_date_vector(self.date_index)
            )

注意啊，date_index应该是要在state更新之后才能递增，因为在T日，只能看到T-1日的股价。去除未来函数之后，再看看你的算法是否还·有效果吧 ：）

我想根据截至今天作为训练数据的所有值来预测第二天要购买的资产（即股票）
I would like to predict which asset (i.e., stock) to buy the next day based on all the values up until today as training data.
如何让模型生成第二天的买入或卖出动作？
How to make the model generate a buy or sell action for the next day?

/home/xxx/.local/lib/python3.8/site-packages/stable_baselines3/common/save_util.py:166: UserWarning: Could not deserialize object lr_schedule. Consider using custom_objects argument to replace this object.
warnings.warn(
/home/xxx/.local/lib/python3.8/site-packages/stable_baselines3/common/save_util.py:166: UserWarning: Could not deserialize object clip_range. Consider using custom_objects argument to replace this object.
warnings.warn(
Traceback (most recent call last):
File "./trader.py", line 129, in
start_trade()
File "./trader.py", line 126, in start_trade
Trader(model_name = options.model).trade()
File "./trader.py", line 45, in trade
model = self.get_model(agent)
File "./trader.py", line 83, in get_model
model.load(model_dir)
File "/home/xxx/.local/lib/python3.8/site-packages/stable_baselines3/common/base_class.py", line 687, in load
model._setup_model()
File "/home/xxx/.local/lib/python3.8/site-packages/stable_baselines3/ppo/ppo.py", line 158, in _setup_model
self.clip_range = get_schedule_fn(self.clip_range)
File "/home/xxx/.local/lib/python3.8/site-packages/stable_baselines3/common/utils.py", line 88, in get_schedule_fn
assert callable(value_schedule)
AssertionError

请问这个bug要怎么解决

您好，试了几台电脑，也换了python版本，都是报这个错误，请教是什么原因？

[root@TD-SERVER-1 nohup]# tail -100f A2C.log
nohup: 忽略输入
train_file 文件夹已存在!
数据读取成功!
加载数据缓存
数据缓存成功!
加载数据缓存
数据缓存成功!
{'n_steps': 5, 'ent_coef': 0.01, 'learning_rate': 0.0007}
Traceback (most recent call last):
File "/data_1/bak/StockRL-main/learn/./trainer.py", line 124, in
start_train()
File "/data_1/bak/StockRL-main/learn/./trainer.py", line 119, in start_train
Trainer(model_name=options.model,
File "/data_1/bak/StockRL-main/learn/./trainer.py", line 52, in train
model.learn(total_timesteps=self.total_timesteps, tb_log_name=self.model_name, log_interval=10, reset_num_timesteps=False, eval_env=env_trade, eval_freq=1000, n_eval_episodes=10)
TypeError: learn() got an unexpected keyword argument 'eval_env'

我是刚刚下载代码，刚刚运行成功。
回来感谢一下，sunnywag!!!
是不是应该，用环境来产生预测数据，之后预测资产。

My name is Luis, I'm a big-data machine-learning developer, I'm a fan of your work, and I usually check your updates.

I am developer something more complementary and full

• testing with 30+- models
• threshold evaluation
• use 1k technical indicators
• method of best feature selection (8, 16, 32..) nice nice
• categorical target instead of continuous target variable powerful machine-learning libraries such as: Sklearn.RandomForest , Sklearn.GradientBoosting, XGBoost, Google TensorFlow and Google TensorFlow LSTM.

With the models trained with the selection of the best technical indicators, the tool is able to predict trading points (where to buy, where to sell) and send real-time alerts to Telegram or Mail. The points are calculated based on the learning of the correct trading points of the last 2 years (including the change to bear market after the rate hike).

I think it could be useful to you, to improve, I would like to share it with you, and if you are interested in improving and collaborating I am also willing, and if not file it in the box.

https://github.com/Leci37/LecTrade/tree/develop