I am facing an issue with tensorflow using BoostedTrees, I have by the way a dataset with 400 labels possible (encoded from 0 to 400 ) for the classification. The dataframe