XGBoost on GPU is killing the kernel (On Ubuntu)

user_10604864 0
Viewed 880

The kernel dies every time I run XGBoost on GPU with this code:

params = {'tree_method': 'gpu_exact', 'max_depth': 100, 'learning_rate': 0.1} dtrain = xgb.DMatrix(X_train,y_train ) xgb.train(params, dtrain, evals=[(dtrain, "train")])

I get:

Kernel died - restarting
xgboost gpu python
1 Answers

I think the main problem is max_depth=100, xgboost builds trees layer by layer so a full tree of depth 100 will be too large. Try using values up to 12 for example.

user_83335900
Related