search

guitargeek / XGBoost-FastForest

Minimal library code to deploy XGBoost models in C++.
MIT License
86 stars 30 forks source link

multiclass support #7

Closed svb688 closed 4 years ago

svb688 commented 4 years ago

Very simple to use so far, I was able to use the library for binary classification and it is speedier than m2cgen. I noticed that the return type of operator() is just float. Is it possible to return a vector of probabilities (for multiclass classification) such as model.predict_proba()? My task involves sampling from these probabilities.

guitargeek commented 4 years ago

Hi again! For now it was not possible to do multiclassification because I didn't need it and nobody asked for it. But it's not hard to implement, which I just did. Please check the README for instructions and let me know if it works for you.

guitargeek commented 4 years ago

Hi! Can this issue be closed?

guitargeek commented 4 years ago

I'll close it, since fastforest now supports the multiclassification. Please open a new issue if you have further problems.

svb688 commented 4 years ago

Thanks again!