Predict(data, output_margin=True) returns only zero values?

#1

Before used predict_proba(data,output_margin=True)[:,1] and works great.
Now predict_proba not contains output_margin.
And predict(data, output_margin=True) returns only zero values?
Although sum of xgb.get_booster().get_dump() still returns log odds (output_margin).
Appreciate any ideas!!!

#2

That seems like a bug, can you post a reproducible example?

#3

hello ,
In using xgboost, I get the value of leaf of all trees , but when I use the sigmoid function to get the final score, there is a big difference comparing with the score from the sklearn-probe function. So I want to know ,how to computing the final score from the xgboost tree?
The question is important for me, but I am sorry for my bad english.