[timeseries] Fix info and fit_summary for TimeSeriesPredictor #2510
Conversation
shchur
added
bug
|
Job PR-2510-f0e538e is done. |
| @@ -735,6 +735,7 @@ def fit_ensemble( | |||
| ) | |||
|
|
|||
| self._add_model(model=simple_ensemble, base_models=simple_ensemble.model_names) | |||
| self.save_model(model=simple_ensemble) | |||
There was a problem hiding this comment.
What was happening before this change?
There was a problem hiding this comment.
The WeightedEnsemble model is saved inside the model graph (i.e., as an attribute of the trainer saved to trainer.pkl). This means that we could predict with the ensemble after loading the predictor from disk.
The new line self.save_model(model=simple_ensemble) saves the metadata of the model (val_score, fit_time, etc.) to disk so that it can be retrieved when calling info(). Without this change the line https://github.com/autogluon/autogluon/blob/master/timeseries/src/autogluon/timeseries/trainer/abstract_trainer.py#L217 fails.
Issue #, if available: #2497 and #2498
Description of changes:
AbstractTimeSeriesModel.get_info()andTimeSeriesPredictor.fit_summary().By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.