Question on using custom features/关于自定义特征导入后的使用问题

大家好，我在使用自定义特征时遇到了一些问题。我首先成功导入了量价数据以及自定义的feature，因为我可以用D.features查看到正确的变量，且保证日历格式都对齐。然而，当我使用DDG－DA进行模型训练时，出现了fea_label_df里所有自定义feature为0的情况，我不知道为什么；另外，当我使用自定义特征训练DDGDA，每当到最后一轮时，总会出现empty dataframe的错误。我之前看到过issue里的类似问题并按相同方法尝试解决，但最终没有生效。请问我应该如何配置一些文件或其他操作？感谢任何帮助，如果需要我提供一些文件的代码，我非常乐意。

Greetings. When I use custom features, i encountered some probs. Claims at first, I successfully 'dump all' custom data into the .qlib since I can use D.features to get my features and the output is the same as my csv. However, when I run the workflow.py under ./DDGDA, there is some probs. First, all my custom features in fea_label_df.pkl is 0.0. Second, when I train DDGDA and it is at the final turn, it always raise an error like 'Empty data in your config'. Did I miss something? Thanks in advance. If I need to provide some code from my side, please feel free to contact me. (Actually, I did not modify many .py files)

Start with ./DDGDA/workflow.py and the custom-feature data exposed through D.features. Reproduce the training run, then inspect how fea_label_df.pkl is produced and where the final-round “Empty data in your config” error is raised; done means identifying the configuration or data-alignment cause of both symptoms.