Does H2OAutoMLRegressor sklearn wrapper use the leader model for predict?

[exalate-issue-sync]

Do the automl sklearn wrappers described in [this notebook|https://github.com/h2oai/h2o-tutorials/blob/master/tutorials/sklearn-integration/H2OAutoML_as_sklearn_estimator.ipynb] use the leader model for their {{predict}} methods? It would seem to be the obvious choice, but I couldn't find a confirmation for that in the [main docs|https://docs.h2o.ai/h2o/latest-stable/h2o-docs/automl.html], the [Python docs|https://docs.h2o.ai/h2o/latest-stable/h2o-py/docs/modeling.html] or the [source code|https://github.com/h2oai/h2o-3/tree/master/h2o-py/h2o/sklearn].

[exalate-issue-sync]

Sebastien Poirier commented: The sklearn wrapper obtains the predictions (from {{predict}} and/or {{predict_proba}} ) from the {{predict}} method on the wrapped H2O estimators.

In the case of the {{H2OAutoML}} estimator, {{predict}} delegates to the leader model as expected.

Thanks for noticing this, we’ll improve the docs to reflect this.

[exalate-issue-sync]

Erin LeDell commented: Changed from question ticket to a documentation ticket.

[h2o-ops]

JIRA Issue Details

Jira Issue: PUBDEV-8762
Assignee: Erin LeDell
Reporter: N/A
State: Open
Fix Version: N/A
Attachments: N/A
Development PRs: N/A

[wendycwong]

Hannah: Please ask Tomas F or Seb for help.

[hannah-tillman]

H2OAutoMLRegressor:
AutoML sklearn wrapper obtains predictions (from predict and/or predict_proba) from the predict method on the wrapped H2O estimators. In the case of the H2OAutoML estimator, predict delegates to the leader models as expected.

Add information to user guide or python guide