When it comes to adding exogenous variables, if I already have future data (forecasted data from other sources) in my df, am I required to generate more future data to add exogenous variables? Should I just use neural forecast if I want to handle lots of exogenous variables?

Hi @Jack both TimeGPT and NeuralForecast can handle exogenous variables. For these variables, you need their future values, covering the complete forecast horizon. Not sure what you mean by "generating more data". TimeGPT can also handle a large number of exogenous, just like NeuralForecast.

Hi Mariana, I was able to figure out, I didn't pay attention to how data should be split. Thank you!

no problem, happy you were able to figure it out!

I was wondering if there is a limit to the date_features, when I try and create a TimeGPT model with it, I get this: INFOnixtla.nixtla clientCalling Forecast Endpoint... INFOnixtla.nixtla clientAttempt 1 failed... INFOnixtla.nixtla clientAttempt 2 failed... INFOnixtla.nixtla clientAttempt 3 failed..., when I turn that feature off it works fine

This is most likely related to the API. When using multiple date features, you're sending more information, and sometimes this might cause an error. If possible, please keep trying and if you keep encountering this error, let us know so we can check it with our web team.

will do!