NeuralForecast |TimeMixer Model - exogeneous variables support

[MLfreakPy]

What happened + What you expected to happen

Using exogenous variables in TimeMixer model, whether historical or future, leads to an error message. As a result, the model can currently not be used with exogeneous variables, however, these are very crucial for good forecasts in many applications.

The issue has been raised also in Github Issue 1016.

@marcopeix, @kwuking: Do you have a timeline of when this issue can be fixed?

Versions / Dependencies

Name: neuralforecast
Version: 1.7.4

Reproduction script

see above

Issue Severity

High: It blocks me from completing my task.

[marcopeix]

We have a draft PR trying to support exogenous feature for TimeMixer. We welcome any contributions from the community! If you want to get give it try, let us know and we're here to answer your questions!

[elephaint]

The original paper doesn't seem to support exogenous features. Upon learning this I don't see a reason to support this.

Note that it's not easy to include exogenous - it is insanely hard to do this right, so it seems a limitation of the architecture that we are not placed to solve.

I'd vote for closing unless I'm misunderstanding here.

[carusyte]

I'm really keen to harness this model's potentiality especially on multivariate / exogenous features, too.
TBH I'm not a data scientist, so I could be missing the nuances of multivariate time series and exogenous features. I read in the original paper that TimeMixer supports multivariate, though. How can I feed additional covariates into this model other than 'y' feature?