SharpLearning can only load models trained in python with xgboost==0.82

[marict]

I have been unable to get SharpLearning to load any models trained in python with xgboost > 0.82 without getting an error on load. It seems that there was some breaking change on 0.82.

[mdabros]

Hi @marict,

Might be. SharpLearning is limited by the version used in https://github.com/PicNet/XGBoost.Net since it just wraps that library. You can try to encourage them to update to the latest version. If this is done, I will be happy to update SharpLearning to use the new version.

[madskonradsen]

@marict See #156 :)

[mdabros]

@marict Eventually, I will switch SharpLearning to use XGBoostSharp instead of XGBoost.Net. XGBoostSharp currently targets xgboost 2.0, so should be able to open a model trained with a newer version of xgboost. In the meantime you should be able to load and use the model by using XGBoostSharp directly. The interface is a bit more limited at the moment, but will be expanded over time.