How can I train an XGBoost model on a GPU but run predictions on CPU without allocating any GPU RAM?
My situation: I create an XGBoot model
