Issues of reproducing the multi-variate forecasting

[yhuanghamu]

We are trying to reproduce the results in this paper Time-MMD: Multi-Domain Multimodal Dataset for Time Series Analysis, but ran into some issues in the codebase.

Wonder if you could

1. fix the multi variate forecasting. It seems like the Dataset_Custom data loader cannot correctly load the data since a few colums are not numerical data.

   MM-TSFlib/data_provider/data_loader.py

   Lines 77 to 79 in e53ecf2

   	if self.features == 'M' or self.features == 'MS':
   	cols_data = df_raw.columns[1:]
   	df_data = df_raw[cols_data]

2. Could you share the training scripts for all other data besides the Health?

Thanks

[ztb-35]

This code is for single variate time series. I think you can try to revise the dataloader by yourself.