[MLForecast: test set eval and early stopping ]

[Djuseppe]

Description

Hey team, it's gonna be beneficial to have an option to early stop models when training.

Use case

CV hyper-param optimization

[jmoralez]

Hey @Djuseppe, thanks for using mlforecast. We have that built-in for LightGBM in the LightGBMCV class and you can do it for other frameworks by using custom training. It'd be very hard to have a general solution because each framework has a slightly different interface.

[Djuseppe]

thanks for the info 🙂 indeed a custom training is amazing and suits me nicely, since I have also custom CV with Optuna optimization. Closing, as existing features are just enough.