Adds the refit argument to AutoMLForecast and sets it to False by default to only train the models on the first window of cross validation and use those trained models to predict all windows in order to save time, since most likely the winning trial won't change.
Also makes the AutoMLForecast.results_ attribute a dictionary with the model names as the keys and the studies as the values, instead of a list. This is to be more consistent with the AutoMLForecast.models_ attribute, which is a dict with the same keys.
Checklist:
[x] This PR has a meaningful title and a clear description.
Description
Adds the
refit
argument toAutoMLForecast
and sets it toFalse
by default to only train the models on the first window of cross validation and use those trained models to predict all windows in order to save time, since most likely the winning trial won't change.Also makes the
AutoMLForecast.results_
attribute a dictionary with the model names as the keys and the studies as the values, instead of a list. This is to be more consistent with theAutoMLForecast.models_
attribute, which is a dict with the same keys.Checklist: