十分感谢分享！我用自己的数据训练了一个模型，请问怎么应用训练好的模型？（即测试集应用）
即：如何给模型输入测试集序列，输出return的label？

我尝试model.predict(test_data),其中test_data是一个TimeSeriesData，但似乎并不能成功。

` # load entire model using load_model() from alphanet module
model = load_model("model.bt")

# only load weights by first creating a model instance
model = AlphaNetV3(l2=0.001, dropout=0.0)
model.load_weights("weights.bt")
test_data=TimeSeriesData(
    dates=test_df["int_date"].values,                   # date column            
    data=test_df[metrics_std].values,                # data columns
    labels=test_df["future_return"].values # label column
)
output=model.predict(test_data.data)
print(output)`

请问这个 data_producer 是在哪里实现的？

你好，最近正在用pytorch实现AlphaNet，想请问下UpDownAccuracy的设计思路。感谢🙏

作者你好，很抱歉再次打扰。我看了之前的issues，也有问到如何使用model.predict。我尝试了一下，发现还是有一些问题。故来询问。
我的demo例子是假设我现在有000723.SZ的三十条数据，我需要如何导入测试？

我使用你说的两种方法，但是发现都有报错。
能否麻烦你帮忙给一个使用小例子？

解压文件损坏

predict出来的是数值，怎么用来生成因子呢

Cannot unzip test dataset. Shows that the dataset is borken

这个是涨跌准确率的意思，仅供参考。P(预测涨跌正确)

Originally posted by @Congyuwang in #18 (comment)
意思是，只要prediction与label同号（即同涨同跌）即可算作预测正确吗？不知道理解得对不对。

谢谢解答！

======================================================================
ERROR: test_save_model (main.TestAlphaNetV2)

Traceback (most recent call last):
File "c:\Users\Administrator\Desktop\AlphaNetV3\tests\tests.py", line 172, in test_save_model
alpha_net.save("./.test_alpha_net_save/model")
File "D:\base\path\py3\lib\site-packages\keras\utils\traceback_utils.py", line 67, in error_handler
raise e.with_traceback(filtered_tb) from None
File "D:\base\path\py3\lib\site-packages\tensorflow\python\saved_model\function_serialization.py", line 27, in _serialize_function_spec
raise NotImplementedError(
NotImplementedError: Cannot serialize a method function without a named 'self' argument.

======================================================================
ERROR: test_save_model (main.TestAlphaNetV3)

Traceback (most recent call last):
File "c:\Users\Administrator\Desktop\AlphaNetV3\tests\tests.py", line 240, in test_save_model
alpha_net_v3.save("./.test_alpha_net_save/model")
File "D:\base\path\py3\lib\site-packages\keras\utils\traceback_utils.py", line 67, in error_handler
raise e.with_traceback(filtered_tb) from None
File "D:\base\path\py3\lib\site-packages\tensorflow\python\saved_model\function_serialization.py", line 27, in _serialize_function_spec
raise NotImplementedError(
NotImplementedError: Cannot serialize a method function without a named 'self' argument.

======================================================================
ERROR: test_save_model (main.TestAlphaNetV4)

Traceback (most recent call last):
File "c:\Users\Administrator\Desktop\AlphaNetV3\tests\tests.py", line 308, in test_save_model
alpha_net_v4.save("./.test_alpha_net_save/model")
File "D:\base\path\py3\lib\site-packages\keras\utils\traceback_utils.py", line 67, in error_handler
raise e.with_traceback(filtered_tb) from None
File "D:\base\path\py3\lib\site-packages\tensorflow\python\saved_model\function_serialization.py", line 27, in _serialize_function_spec
raise NotImplementedError(
NotImplementedError: Cannot serialize a method function without a named 'self' argument.

======================================================================
FAIL: test_last_batch_of_training_dataset (main.TestDataNormalizationModule)

Traceback (most recent call last):
File "c:\Users\Administrator\Desktop\AlphaNetV3\tests\tests.py", line 504, in test_last_batch_of_training_dataset
self.assertTrue(is_all_close(
AssertionError: False is not true : last batch of training data (start 20116595): failure

Ran 31 tests in 64.568s

FAILED (failures=1, errors=3)

===========================
python 3.9
torch 1.12.1
tensorflow 2.9.2
Pillow 9.2.0
pandas 1.4.4
numpy 1.23.2
numba 0.56.2
keras 2.9.0

请问范例数据‘some_data.csv’能不能上传至云盘以供下载，谢谢！

作者你好，我想请教一下，该份数据中的所有包含nan值的sample是直接扔掉不加入训练中吗

Hello blogger, I would like to ask if you can provide the requirements.txt file? Or tell me the version of keras or tensorflow?
Because the APIs of different versions of tensorflow are particularly confusing, I want to use your code to reproduce the content in the series of Huatai Financial Engineering articles. Thank you again for your contribution!