Does CLI version xgboost support shap?


We have the option to output shap values in python version xgboost when we set the pred_contrib as True and make the prediction. Can we do the same for CLI version xgboost model?



I don’t think CLI supports SHAP prediction.


Thank you @hcho3. Are you going to add it to CLI? It is more convenient to use CLI for us as we don’t need to load train sets to memory like the python version. Our data sets is very sparse and high dimensional and we will run out of memory using python version xgboost. Any suggestions on that? Thank you.