Python XGBRegressor: dressing a booster without fit

#1

when fitting a XGBRegressor, a Booster object can be passed via the model parameter. Before fitting a XGBRegressor, the predict method does not work, which makes sense. The Booster Object has a predict method that works well (you have to convert the features to a DMatrix, though).

I want to dress a Booster object to a XGBRegressor so that before fitting the predict method of the XGBRegressor works, using the one from the Booster and accepting non DMatrix features. In Code

booster = Booster().load_model('path_to_save')
model = XGBRegressor(parameters)
model.set_booster(booster) # <== that is the line I am looking to achieve
model.predict(X) # so that this then works

Thank you very much!

#2

Booster and XGBRegressor are different types of objects, it’s possible to use them together for regression. But for classification, you might run into issues.

But non the less, try this:

model = XGBRegressor(**parameters)
model.load_model("path_to_save")
model.predict(X)
#3

Thank you very much!
I did not know that the model you pass to the sklearn.load_model interface is a booster!
The docs are rather vague about it

load_model( fname )
Load the model from a file or bytearray. Path to file can be local or as an URI.
The model is loaded from XGBoost format which is universal among the various XGBoost interfaces.