I’m trying to understand the order in which trees are dumped when calling booster.dump_model() in a multiclass classification scenario. Meaning, is it class0_tree0, class0_tree1, … or class0_tree0, class1_tree0, …
I am aware that this question has been asked before (https://github.com/dmlc/xgboost/issues/692) and I have seen the answer. However, I was wondering if there is any kind of documentation about this? Or any way to add this information to the output file produced by booster.dump_model()?
Thanks a lot!