Is it possible to recreate the xgboost object from an mlfow experiment

[brunocarlin]

Describe the feature and the current state.

I haven't seen this functionality

Will this change a current behavior? How?

I don't think so

Additional Information

The idea is that after importing the experiment it would only require you to say the model object, it does not matter for me that ci intervals are lost.

[davivieirab]

@brunocarlin , not sure what you are interested in. Could you provide an example illustrating the expected behavior?

[brunocarlin]

mlflow has some integrations with xgboost that allow you to easily save a trained model as an mlflow experiment that can be imported on pipeline for tools such as pyspark

https://www.mlflow.org/docs/latest/python_api/mlflow.xgboost.html

The problem is that xgbse is too custom for those tools to work right now, my team has circumvented the issue by building custom models for mlflow, but a standard would still be nice for new teams working with xgbse

[dhuang-apex]

@brunocarlin
Do you mind sharing how you save a model that's trained using xgbse? When I save the model.bst it appears to only save the xgboost portion of the model.