custom training params - feature_weights and per_float_feature_quantization. Does anyone know if this is supported in c++? I’m seeing different prediction values when I apply the model in python and c++. Appreciate any help
hi Arushi ! as for your explicit question, our python package is a wrapper around the c++ implementation. so, yes, these parameters are supported in c++. predictions may differ for quite a few reasons. i would suggest creating an issue at github with a detailed description.
Thank you for getting back to me! I was able to confirm that the bug was not related to catboost. The warning “Parameter feature_weights is ignored because it can’t be parsed” threw me off. Is this indeed not supported?
it is supported on cpu. i will look into this. many thanks for using catboost! :)
Обсуждают сегодня