Can you tell how we should prepare new datasets to use MTAD?

This work is marvelous, and I would like to try other datasets, but it is unclear how the training and test datasets are prepared and how we should dispose variable columns in the .pkl files.

Thanks for your time and attention. Best regards.