Transformer variants for stock market forecasting using additional variates like volume, own, high, low etc:

[Rsaha-16]

Transformer variants for stock market forecasting using additional variates like volume, own, high, low etc:
If we pass these variates as additional dimension in input dataframe for predicting future price, will the library consider these variates(e.g. volume) in prediction if I don't fill them if future_exog_list?
If it is necessary to pass future_exog_list, how do we fill values in future date at the time of inference because these are unknown like future price, given that hist_exog_list is not supported f or transformer variants as per https://nixtlaverse.nixtla.io/neuralforecast/docs/capabilities/overview.html (or is it like we just need to copy latest values to next prediction_length entries)?
We checked that, if we add and remove columns like rsi, vol, prediction values of price change. Does it mean exog_list is not necessary to fill? If so what is the use case for future_exog_list in transformer variants?