Hello! I'm wondering if there's any way to acquire the feature importances of exogenous regressors of a model (such as NHITS) similarly to the way it's acquired with TimeGPT. Doc here for reference: https://nixtlaverse.nixtla.io/nixtla/docs/tutorials/exogenous_variables.html

Hello! To my knowledge, this is not possible in neuralforecast. However there are ways you can do it yourself by using SHAP. I hope this helps!

It does help! Didn't know such thing existed. Thank you!

We have plans to integrate these methods to the library 🙂