TVM appears to use “reg:linear” or ‘rank:pairwise’ for it’s loss function when doing autotune:
https://github.com/dmlc/tvm/blob/master/python/tvm/autotvm/tuner/xgboost_cost_model.py#L102
But XGBoost itself seems to have deprecated reg:linear:
Should the first line be changed to ‘reg:squarederror’?