Skip Binning in HistGradientBoostingRegressor

[simpsus]

My data is already binned and binning takes ~20 minutes on my dataset and machine.
Using a more optimized lib (xgboost/lightgbm) is not an option as there is some custom extension (a mini fork) I want to run a study on.

Is there a way to prevent binning from happening? Can I set an internal variable to False or pass a _BinMapper class that does nothing?

Thank you very much

[glemaitre]

If you have data that are already binned then I would give a try to use the GradientBoostingClassifier or GradientBoostingRegressor. We usually mentioned that this is slow because the splits are down on continuous data. It might be worth testing if the fit performance are fine.

[simpsus]

Thx for the hint. That would mean I have to port my extension ... I will have a look

[adrinjalali]

I wonder how much of a usecase it'd be to allow passing a custom binner, WDYT @glemaitre ?