Fix/fit ensemble

[yMorakkam]

Fix fit_ensemble(): Save models and ensemble

[mfeurer]

Thanks for your pull request. Could you please describe in more detail which problem it fixes and how it does so?

[yMorakkam]

By using fit_ensemble method, models was not saved in the AutomlClassifier object ( self.models[] was empty). By adding _load_models() at the end of this method, i fixed this small issue.

[mfeurer]

I don't think that's a mistake but not loading them at the end of the fit() method is by design. Instead, models are loaded at predict time. Could you please explain why the current situation is problematic?

[yMorakkam]

After calling fit_ensemble(), i dump my model using joblib: joblib.dump(automl, 'model.pkl'). Therefore, my model was empty.

I thought it was a mistake because models are loaded in _fit () in automl.py.

    # while the ensemble builder tries to access the data
    if self._proc_ensemble is not None and self._ensemble_size > 0:
        self._proc_ensemble.join()

    self._proc_ensemble = None
    self._load_models()

    return self

[mfeurer]

Okay, I think I got it now. Your solution makes total sense, thank you very much for this contribution.