xgboost设置cpu核数

谁践踏了优雅 2022-11-29 12:57 235阅读 0赞

xg默认是满核 运行，对应[文档内容][Link 1]如下:

nthread [default to maximum number of threads available if not set]
    
    Number of parallel threads used to run XGBoost

我们也可以根据nthread 这个参数进行设置

试了一下：

设置不同核数看了下 运行速度，发现核数多确实快了很多

import xgboost as xgb
    from sklearn.datasets import make_hastie_10_2
    import  numpy as  np
    import datetime
    from multiprocessing import cpu_count
    print('total cores is %d'%cpu_count())
    NTHREAD=8
    if __name__ == "__main__":
        starttime = datetime.datetime.now()
        X, y = make_hastie_10_2(n_samples=200000, random_state=42)
        xgb_model = xgb.XGBRegressor(nthread=NTHREAD)
        #xgb_model = xgb.XGBRegressor()
        xgb_model.fit(X, y)
        endtime = datetime.datetime.now()
        print('used %d cores'%NTHREAD)
        print('spend  time----:%d s'%(endtime - starttime).seconds)

结果如下：

[root@localhost opt]# python tt.py
    total cores is 8
    used 1 cores
    spend  time----:24 s
    [root@localhost opt]# python tt.py
    total cores is 8
    used 2 cores
    spend  time----:13 s
    [root@localhost opt]# python tt.py
    total cores is 8
    used 3 cores
    spend  time----:10 s
    [root@localhost opt]# python tt.py
    total cores is 8
    used 4 cores
    spend  time----:8 s
    [root@localhost opt]# python tt.py
    total cores is 8
    used 5 cores
    spend  time----:5 s
    [root@localhost opt]# python tt.py
    total cores is 8
    used 6 cores
    spend  time----:6 s
    [root@localhost opt]# python tt.py
    total cores is 8
    used 7 cores
    spend  time----:6 s
    [root@localhost opt]# python tt.py
    total cores is 8
    used 8 cores
    spend  time----:6 s

[Link 1]: https://xgboost.readthedocs.io/en/latest/parameter.html