Xgboost(GPU)在预测时崩溃

时间:2018-08-08 01:59:30

标签: python jupyter-notebook xgboost

我在Python中使用XGBoost GPU版本,每当我尝试运行.predict时,它就会崩溃。它适用于较小的数据集,但对于我当前的问题却不起作用。

train_final.shape, test_final.shape
((631761, 174), (421175, 174))


params = {
          'objective': 'multi:softmax', 
          'eval_metric': 'mlogloss',
          'eta': 0.1,
          'max_depth': 6, 
          'nthread': 4,
          'alpha':0,
          'num_class': 5,
          'random_state': 42, 
          'tree_method': 'gpu_hist',
          'silent': True
             }

GPU统计:GTX 1070,6 GB 内存:32 GB

有人可以帮我理解为什么会这样吗?

1 个答案:

答案 0 :(得分:2)

Saving the model, deleting the booster then loading the model again should achieve this.

# training
bst = xgb.train(param, dtrain, num_round)

#save model
joblib.dump(bst, 'xgb_model.dat')
bst.__del__()

#load saved model
bst = joblib.load('xgb_model.dat')
preds = bst.predict(dtest)