No difference in data preparation in feature value "M" or "MS"

[rakshify]

I checked the data created when the feature flag is M vs MS, the dataset created was exactly the same in both the cases.

For example, "M" corresponds to multivariate to multivariate prediction. If we take the example of ETT dataset, the dataset created in this case is the following:
Tensor size = B * T * 1, where B is the batch size and T is the sequence size, eg - 512 for input and 96 for output. The 1 corresponds to num features. So, we will have 512 "HULL" values as input for a single datapoint and next 96 "HULL" values as output.

"MS" corresponds to multivariate to single prediction. So, if our target column is "OT", so, this should have created 512 sequence of all the features as input and next 96 sequence of "OT" as output. But, it creates exactly the same dataset as the "M" case.

It feels wrong, but I'm not sure if I'm missing something. Please help.

[cddesire]

I also encountered the same issue; for prediction tasks M and MS, there is no difference. For the MS task, the value of seq_y is considered to be df[[self.target]].

[WCUSTC]

same problem