Closed rnyak closed 2 years ago
@radekosmulski fyi.
Would be great to have this in 22.08, but not sure that is still feasible 🙂
For this reloading to work this way the booster params need to be passed through as well on the last line.
XGBoost(schema, booster=bst, **xgb_booster_params )
Closing this as resolved by the above comment. Or using the new save/load methods in #656
Bug description
I get the following error from when I load back the saved XGB model and then pass the booster argument to the constructor via
XGBoost(schema, booster=reloaded_model)
Steps/Code to reproduce bug
You can repro the error by running the code below:
Expected behavior
I should be able to load back the saved XGB model and do offline inference with
model.predict()
Environment details
Additional context