Changed loadmodel to loadmodel! as we are actually modifying the booster with this method
Currently if we use predict (in /src/predict.jl) on a binary classifier, the output is split into two columns while we only expected it to give one. The reason is that num_class is not loaded when we do loadmodel!. So loading in the num_class property when we loadmodel! fixes the predict functionality. Have also added tests on this
loadmodel
toloadmodel!
as we are actually modifying the booster with this methodloadmodel!
. So loading in the num_class property when weloadmodel!
fixes the predict functionality. Have also added tests on this