Extracting individual models from stacking regressor

[Sukantabasu]

I am using "ensemble" = True and "model_history" = True. I would like to extract individual models from the following stacking regressor. What is the procedure?

StackingRegressor(estimators=[('lgbm',
<flaml.model.LGBMEstimator object at 0x153e400a41c0>),
('xgboost',
<flaml.model.XGBoostSklearnEstimator object at 0x153e400a4220>),
('xgb_limitdepth',
<flaml.model.XGBoostLimitDepthEstimator object at 0x153e400a4250>),
('catboost',
<flaml.model.CatBoostEstimator object at 0x153e400a4280>),
('rf',
<flaml.model.RandomForestEstimator object at 0x153e400a42b0>),
('extra_tree',
<flaml.model.ExtraTreesEstimator object at 0x153e400a42e0>)],
n_jobs=-1, passthrough=True)

[sonichi]

StackingRegressor object has an attribute "estimators_", which a list of estimators. Is that what you are looking for?
automl.model.estimators_

[Sukantabasu]

No, I would like to load individually trained sub-models (best lgb, best rf, etc.). I would like to make predictions from each one of them separately for comparison.

[sonichi]

These two attributes can be used:
https://microsoft.github.io/FLAML/docs/reference/automl#best_model_for_estimator
https://microsoft.github.io/FLAML/docs/reference/automl#best_loss_per_estimator

[Sukantabasu]

Thanks a lot for the information!

[sonichi]

You're welcome. I'd appreciate your feedback in a survey: https://docs.google.com/forms/d/e/1FAIpQLSd0u4BbSzr-m9hgM8rmPlYB_JJoFXeHhuoXI9YFbHDRWrDlOQ/viewform?usp=sf_link.

[Sukantabasu]

Done...

[sonichi]

@Sukantabasu Thanks for your feedback. I have some followup questions for you. Do you mind joining on gitter? Or if there is an alternative way to contact you, could you please let me know? Thank you very much.

[Sukantabasu]

Hi Chi, You can contact me at ***@***.*** Best regards, Sukanta

…

-- Sukanta Basu Associate Professor Faculty of Civil Engineering and Geosciences Delft University of Technology T: +31-15-27 89358 (office) T: +31-6-15557518 (cell) https://sites.google.com/view/sukantabasu/ https://www.researchgate.net/profile/Sukanta_Basu https://scholar.google.com/citations?user=08bv9p8AAAAJ&hl=en ----------------------------------------------------------

On 15 Sep 2022, at 00:07, Chi Wang ***@***.***> wrote: @Sukantabasu <https://github.com/Sukantabasu> Thanks for your feedback. I have some followup questions for you. Do you mind joining on gitter <https://gitter.im/FLAMLer/community>? Or if there is an alternative way to contact you, could you please let me know? Thank you very much. — Reply to this email directly, view it on GitHub <#488 (reply in thread)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AL6WSR2OGGL65SSZ3WR4M4LV6JEBVANCNFSM5REPWDDQ>. You are receiving this because you were mentioned.