[FR] Autologging functionality for scikit-learn integration with XGBoost (and LightGBM)

MLflow Roadmap Item

This is an MLflow Roadmap item that has been prioritized by the MLflow maintainers. We’re seeking help with the implementation of roadmap items tagged with the help wanted label. Items tagged with the needs design label are awaiting a design sketch from an MLflow maintainer: @WeichenXu123.

For requirements clarifications and implementation questions, or to request a PR review, please tag @WeichenXu123 in your communications related to this issue.

Proposal Summary

Hello,

When trying to implement autolog() function for XGBoostClassifier and XGBoostRegressor, we are not able to see the metrics and parameters or models which are getting logged. Can you please address this issue.

Thanks.

PFA added code for XGBoostRegressor which is to be converted to .py file to replicate the issue boston.txt

[harupy]

Logging feature impotance within sklearn autologging sounds good to me.

[dbczumar]

Autologging for XGBoost sklearn estimators should be initiated through mlflow.xgboost.autolog. We can add an internal (prefixed with _) autolog API in mlflow.sklearn with an xgboost_estimators=True/False param. If xgboost_estimators==true, we only patch XGBoost sklearn estimators and associated them with the XGBoost autologging integration via safe_patch("xgboost", ...). Otherwise, we patch the sklearn estimators (and not the xgboost estimators) via safe_patch("sklearn", ...), which is the current behavior. Let me know if this sounds good!

Yes, looks good to me! I will start to work on its implementation unless there are other suggestions. Thanks!

Thanks a bunch, @jwyyy ! Sounds good!