~~~~~~~~~~~~~~~~~~~~串行single sample predict~~~~~~~~~~~~~~~~~~~~~~~~~

python下已测试通过，无问题：

然而C++下问题是：

 1，首先是与python下概率不一致；

2，然后是所有输入的结果都一样

初步怀疑版本问题，准备从2.1.1升级到3.3.1试一下。2.x和3.x有很大不一样，编译3.3.1时遇到下列问题：

LightGBM-3.3.1/include/LightGBM/utils/common.h:36:59: fatal error: ../../../external_libs/fmt/include/fmt/format.h: 没有那个文件或目录
LightGBM-3.3.1/include/LightGBM/utils/common.h:38:82: fatal error: ../../../external_libs/fast_double_parser/include/fast_double_parser.h: 没有那个文件或目录2,LightGBM-3.3.1/src/treelearner/linear_tree_learner.cpp:7:23: fatal error: Eigen/Dense: 没有那个文件或目录
https://gitlab.com/libeigen/eigen/-/releases/3.4.0    LightGBM-3.3.1/external_libs/

解决办法：直接 LightGBM/external_libs at master · microsoft/LightGBM · GitHub在这里下载对应的库解压后放在对应文件夹即可。编译一路顺风生成了动态库。将include、external_libs和lib_lightgbm.so文件夹打包即可应用。

测试官例也通过了：

 然而应用时发现：

lightGBM/include/LightGBM/utils/common.h:57:26: 错误：‘void* malloc(size_t)’先被声明为‘extern’后又被声明为‘static’ [-fpermissive]
/usr/local/lib/gcc/x86_64-pc-linux-gnu/9.1.0/include/mm_malloc.h:41:7: 错误：‘__alignment’在此作用域中尚未声明41 |   if (__alignment == 1)|       ^~~~~~~~~~~
/usr/local/lib/gcc/x86_64-pc-linux-gnu/9.1.0/include/mm_malloc.h:43:7: 错误：‘__alignment’在此作用域中尚未声明43 |   if (__alignment == 2 || (sizeof (void *) == 8 && __alignment == 4))|       ^~~~~~~~~~~
/usr/local/lib/gcc/x86_64-pc-linux-gnu/9.1.0/include/mm_malloc.h:45:31: 错误：‘__alignment’在此作用域中尚未声明
LightGBM/utils/common.h:58:21: 错误：‘void free(void*)’先被声明为‘extern’后又被声明为‘static’ [-fpermissive]

这些问题估计是与我的系统哪里冲突了。然后查了下：https://github.com/microsoft/LightGBM/pull/5111

Compilation error for cpp tests on macOS with gcc and `thread` sanitizer · Issue #4331 · microsoft/LightGBM · GitHub

可以看到别人也遇到这个问题 

LightGBM/utils/common.h:57:26: error: 'void* malloc(size_t)' was declared 'extern' and later 'static' [-fpermissive]

 解决办法写成文档上传在此：https://download.csdn.net/download/wd1603926823/85201370

int main()
{char srcimg[400]={0};int numiterations = 1;BoosterHandle handle;int set = LGBM_BoosterCreateFromModelfile("/home/jumper/xrt/reference/model/lgbmmodel/lbgm_zhu_1.model", &numiterations, &handle);if(set==0){std::cout << "load model successfully !  "<< std::endl;}int channels[]={1,2};int histsize[]={8,8};float ghistrange[]={0,256};float rhistrange[]={0,256};const float *histsranges[]={ghistrange,rhistrange};vector<Mat> imgsvec;for(int index=0;index<=20;index++){sprintf(srcimg,"/home/jumper/xrt/reference/imgs/zhudoctor/orescnn/%d.png",index);Mat img=imread(srcimg);if(img.empty())continue;imgsvec.push_back(img);}double timeStart = (double)getTickCount();int size=imgsvec.size();for(int id=0;id!=size;id++){Mat img=imgsvec[id];MatND hist;cv::calcHist(&img,1,channels,Mat(),hist,2,histsize,histsranges);std::vector<double> out(1, 0);double *out_result = static_cast<double *>(out.data());int64_t out_len;int res = LGBM_BoosterPredictForMat(handle,hist.data,C_API_DTYPE_FLOAT32,1,64,1,C_API_PREDICT_NORMAL,0,-1,"None",&out_len,out_result);std::cout <<"image id:"<<id<<" ---LGBM row predict result is: " << out[0] <<std::endl;}double nTime = ((double)getTickCount() - timeStart) / getTickFrequency()*1000;cout<<nTime<<" ms!"<<endl;return 0;
}

结果已正确！！！如下所示：

load model successfully !  
[LightGBM] [Warning] Unknown parameter: None
image id:0 ---LGBM row predict result is: 0.964677
[LightGBM] [Warning] Unknown parameter: None
image id:1 ---LGBM row predict result is: 0.877513
[LightGBM] [Warning] Unknown parameter: None
image id:2 ---LGBM row predict result is: 0.973227
[LightGBM] [Warning] Unknown parameter: None
image id:3 ---LGBM row predict result is: 0.895759
[LightGBM] [Warning] Unknown parameter: None
image id:4 ---LGBM row predict result is: 0.945096
[LightGBM] [Warning] Unknown parameter: None
image id:5 ---LGBM row predict result is: 0.792787
[LightGBM] [Warning] Unknown parameter: None
image id:6 ---LGBM row predict result is: 0.902854
[LightGBM] [Warning] Unknown parameter: None
image id:7 ---LGBM row predict result is: 0.965496
[LightGBM] [Warning] Unknown parameter: None
image id:8 ---LGBM row predict result is: 0.92893
[LightGBM] [Warning] Unknown parameter: None
image id:9 ---LGBM row predict result is: 0.110013
[LightGBM] [Warning] Unknown parameter: None
image id:10 ---LGBM row predict result is: 0.903951

C和python结果已对上。小数点后五六位都完全一致。

~~~~~~~~~~~~~~~~~~~~~~~~分界线~~~~并行库single sample predict~~~~~~~~~~~~~~~~~~

并行库配合Lightgbm C++实现并行predict：

load model successfully !  
[LightGBM] [Warning] [LightGBM] [Warning] Unknown parameter: None
[LightGBM] [Warning] [LightGBM] [Warning] Unknown parameter: None
Unknown parameter: None
[LightGBM] [Warning] [LightGBM] [Warning] [LightGBM] [Warning] Unknown parameter: None
Unknown parameter: None
Unknown parameter: None
Unknown parameter: None
[LightGBM] [Warning] Unknown parameter: None
image id:0 ---LGBM row predict result is: 0.964677
[LightGBM] [Warning] Unknown parameter: None
image id:1 ---LGBM row predict result is: 0.877513
[LightGBM] [Warning] Unknown parameter: None
image id:2 ---LGBM row predict result is: 0.973227
[LightGBM] [Warning] Unknown parameter: None
image id:3 ---LGBM row predict result is: 0.895759
[LightGBM] [Warning] Unknown parameter: None
image id:4 ---LGBM row predict result is: 0.945096
[LightGBM] [Warning] Unknown parameter: None
image id:5 ---LGBM row predict result is: 0.792787
[LightGBM] [Warning] Unknown parameter: None
image id:11 ---LGBM row predict result is: 0.0852482
image id:14 ---LGBM row predict result is: 0.92168
image id:7 ---LGBM row predict result is: 0.965496
[LightGBM] [Warning] Unknown parameter: None
[LightGBM] [Warning] Unknown parameter: None
image id:image id:12 ---LGBM row predict result is: 10 ---LGBM row predict result is: 0.948007
image id:6 ---LGBM row predict result is: [LightGBM] [Warning] Unknown parameter: None
[LightGBM] [Warning] Unknown parameter: None
0.902854
image id:13 ---LGBM row predict result is: 0.957286
[LightGBM] [Warning] Unknown parameter: None
[LightGBM] [Warning] Unknown parameter: None
0.903951
[LightGBM] [Warning] Unknown parameter: None
image id:8 ---LGBM row predict result is: 0.92893
image id:16 ---LGBM row predict result is: 0.934247
image id:19 ---LGBM row predict result is: 0.933926
image id:20 ---LGBM row predict result is: 0.937277
image id:18 ---LGBM row predict result is: 0.858938
image id:9 ---LGBM row predict result is: 0.110013
image id:17 ---LGBM row predict result is: 0.83457
image id:15 ---LGBM row predict result is: 0.927055

可以看到并行与串行结果一致。

 我跑了很多次测试single sample predict串行与并行的耗时跳变非常大，如下：

未解决耗时跳变时的稳定性测试，每次1560幅图：串行测试serial performance:load model successfully !  
0---serial stable test:71.4656ms ~~ 71.4656ms!!!
1---serial stable test:71.4656ms ~~ 150.341ms!!!
2---serial stable test:71.4656ms ~~ 192.314ms!!!
3---serial stable test:71.4656ms ~~ 192.314ms!!!
4---serial stable test:71.4656ms ~~ 192.314ms!!!
5---serial stable test:71.4656ms ~~ 192.314ms!!!
6---serial stable test:71.4656ms ~~ 192.314ms!!!
7---serial stable test:71.4656ms ~~ 192.314ms!!!
8---serial stable test:71.4656ms ~~ 192.314ms!!!
9---serial stable test:71.4656ms ~~ 192.314ms!!!
10---serial stable test:71.4656ms ~~ 192.314ms!!!
11---serial stable test:71.4656ms ~~ 192.314ms!!!
12---serial stable test:71.4656ms ~~ 192.314ms!!!
13---serial stable test:71.4656ms ~~ 192.314ms!!!
14---serial stable test:71.4656ms ~~ 192.314ms!!!
15---serial stable test:71.4656ms ~~ 192.314ms!!!
16---serial stable test:71.4656ms ~~ 192.314ms!!!
17---serial stable test:71.4656ms ~~ 192.314ms!!!
18---serial stable test:71.4656ms ~~ 192.314ms!!!
19---serial stable test:71.4656ms ~~ 192.314ms!!!
20---serial stable test:71.4656ms ~~ 192.314ms!!!
21---serial stable test:71.4656ms ~~ 196.893ms!!!
22---serial stable test:71.4656ms ~~ 196.893ms!!!
23---serial stable test:71.4656ms ~~ 196.893ms!!!
24---serial stable test:71.4656ms ~~ 196.893ms!!!
25---serial stable test:71.4656ms ~~ 196.893ms!!!
26---serial stable test:71.4656ms ~~ 196.893ms!!!
27---serial stable test:71.4656ms ~~ 196.893ms!!!
28---serial stable test:71.4656ms ~~ 196.893ms!!!
29---serial stable test:71.4656ms ~~ 196.893ms!!!
30---serial stable test:71.4656ms ~~ 196.893ms!!!
31---serial stable test:71.4656ms ~~ 196.893ms!!!
32---serial stable test:71.4656ms ~~ 196.893ms!!!
33---serial stable test:71.4656ms ~~ 196.893ms!!!
34---serial stable test:71.4656ms ~~ 196.893ms!!!
35---serial stable test:71.4656ms ~~ 347.883ms!!!
36---serial stable test:71.4656ms ~~ 347.883ms!!!
37---serial stable test:71.4656ms ~~ 347.883ms!!!
38---serial stable test:71.4656ms ~~ 347.883ms!!!
39---serial stable test:71.4656ms ~~ 347.883ms!!!
40---serial stable test:71.4656ms ~~ 347.883ms!!!
41---serial stable test:71.4656ms ~~ 347.883ms!!!
42---serial stable test:71.4656ms ~~ 347.883ms!!!
43---serial stable test:71.4656ms ~~ 347.883ms!!!
44---serial stable test:71.4656ms ~~ 347.883ms!!!
45---serial stable test:71.4656ms ~~ 347.883ms!!!
46---serial stable test:71.4656ms ~~ 347.883ms!!!
47---serial stable test:71.4656ms ~~ 347.883ms!!!
48---serial stable test:71.4656ms ~~ 347.883ms!!!
49---serial stable test:71.4656ms ~~ 347.883ms!!!
50---serial stable test:71.4656ms ~~ 347.883ms!!!
51---serial stable test:71.4656ms ~~ 347.883ms!!!
52---serial stable test:71.4656ms ~~ 347.883ms!!!
53---serial stable test:71.4656ms ~~ 347.883ms!!!
54---serial stable test:71.4656ms ~~ 400.697ms!!!
55---serial stable test:71.4656ms ~~ 400.697ms!!!
56---serial stable test:71.4656ms ~~ 400.697ms!!!
57---serial stable test:71.4656ms ~~ 400.697ms!!!
58---serial stable test:71.4656ms ~~ 400.697ms!!!
59---serial stable test:71.4656ms ~~ 400.697ms!!!
60---serial stable test:71.4656ms ~~ 400.697ms!!!
61---serial stable test:71.4656ms ~~ 400.697ms!!!
62---serial stable test:71.4656ms ~~ 400.697ms!!!
63---serial stable test:71.4656ms ~~ 400.697ms!!!
64---serial stable test:71.4656ms ~~ 400.697ms!!!
65---serial stable test:71.4656ms ~~ 400.697ms!!!
66---serial stable test:71.4656ms ~~ 400.697ms!!!
67---serial stable test:71.4656ms ~~ 400.697ms!!!
68---serial stable test:71.4656ms ~~ 400.697ms!!!
69---serial stable test:71.4656ms ~~ 400.697ms!!!
70---serial stable test:71.4656ms ~~ 400.697ms!!!
71---serial stable test:71.4656ms ~~ 400.697ms!!!
72---serial stable test:71.4656ms ~~ 400.697ms!!!
73---serial stable test:71.4656ms ~~ 400.697ms!!!
74---serial stable test:71.4656ms ~~ 400.697ms!!!
75---serial stable test:71.4656ms ~~ 400.697ms!!!
76---serial stable test:71.4656ms ~~ 400.697ms!!!
77---serial stable test:71.4656ms ~~ 400.697ms!!!
78---serial stable test:71.4656ms ~~ 400.697ms!!!
79---serial stable test:71.4656ms ~~ 400.697ms!!!
80---serial stable test:71.4656ms ~~ 400.697ms!!!
81---serial stable test:71.4656ms ~~ 400.697ms!!!
82---serial stable test:71.4656ms ~~ 400.697ms!!!
83---serial stable test:71.4656ms ~~ 400.697ms!!!
84---serial stable test:71.4656ms ~~ 400.697ms!!!
85---serial stable test:71.4656ms ~~ 400.697ms!!!
86---serial stable test:71.4656ms ~~ 400.697ms!!!
87---serial stable test:71.4656ms ~~ 400.697ms!!!
88---serial stable test:71.4656ms ~~ 400.697ms!!!
89---serial stable test:71.4656ms ~~ 400.697ms!!!
90---serial stable test:71.4656ms ~~ 400.697ms!!!
91---serial stable test:71.4656ms ~~ 400.697ms!!!
92---serial stable test:71.4656ms ~~ 400.697ms!!!
93---serial stable test:71.4656ms ~~ 400.697ms!!!
94---serial stable test:71.4656ms ~~ 400.697ms!!!
95---serial stable test:71.4656ms ~~ 400.697ms!!!
96---serial stable test:71.4656ms ~~ 400.697ms!!!
97---serial stable test:71.4656ms ~~ 400.697ms!!!
98---serial stable test:71.4656ms ~~ 400.697ms!!!
99---serial stable test:71.4656ms ~~ 400.697ms!!!
可以看到，对于1560幅图耗时在71~400ms之间，非常不稳定！！！！并行测试：
parallel performance:
load model successfully !  
0---parallel stable test:51.0798ms ~~ 51.0798ms!!!
1---parallel stable test:51.0798ms ~~ 55.0909ms!!!
2---parallel stable test:51.0798ms ~~ 59.2024ms!!!
3---parallel stable test:51.0798ms ~~ 65.9625ms!!!
4---parallel stable test:51.0798ms ~~ 69.393ms!!!
5---parallel stable test:51.0798ms ~~ 69.393ms!!!
6---parallel stable test:51.0798ms ~~ 69.393ms!!!
7---parallel stable test:51.0798ms ~~ 69.393ms!!!
8---parallel stable test:51.0798ms ~~ 69.393ms!!!
9---parallel stable test:51.0798ms ~~ 69.393ms!!!
10---parallel stable test:51.0798ms ~~ 69.393ms!!!
11---parallel stable test:51.0798ms ~~ 69.393ms!!!
12---parallel stable test:51.0798ms ~~ 69.393ms!!!
13---parallel stable test:51.0798ms ~~ 69.393ms!!!
14---parallel stable test:51.0798ms ~~ 69.393ms!!!
15---parallel stable test:51.0798ms ~~ 69.393ms!!!
16---parallel stable test:51.0798ms ~~ 69.393ms!!!
17---parallel stable test:51.0798ms ~~ 69.393ms!!!
18---parallel stable test:51.0798ms ~~ 69.393ms!!!
19---parallel stable test:51.0798ms ~~ 69.393ms!!!
20---parallel stable test:51.0798ms ~~ 69.393ms!!!
21---parallel stable test:51.0798ms ~~ 69.9926ms!!!
22---parallel stable test:51.0798ms ~~ 69.9926ms!!!
23---parallel stable test:51.0798ms ~~ 69.9926ms!!!
24---parallel stable test:51.0798ms ~~ 69.9926ms!!!
25---parallel stable test:51.0798ms ~~ 69.9926ms!!!
26---parallel stable test:51.0798ms ~~ 69.9926ms!!!
27---parallel stable test:51.0798ms ~~ 69.9926ms!!!
28---parallel stable test:51.0798ms ~~ 69.9926ms!!!
29---parallel stable test:51.0798ms ~~ 70.647ms!!!
30---parallel stable test:51.0798ms ~~ 70.647ms!!!
31---parallel stable test:51.0798ms ~~ 70.647ms!!!
32---parallel stable test:51.0798ms ~~ 70.647ms!!!
33---parallel stable test:51.0798ms ~~ 70.647ms!!!
34---parallel stable test:51.0798ms ~~ 70.647ms!!!
35---parallel stable test:51.0798ms ~~ 70.647ms!!!
36---parallel stable test:51.0798ms ~~ 70.647ms!!!
37---parallel stable test:51.0798ms ~~ 70.647ms!!!
38---parallel stable test:51.0798ms ~~ 70.647ms!!!
39---parallel stable test:51.0798ms ~~ 70.647ms!!!
40---parallel stable test:51.0798ms ~~ 70.647ms!!!
41---parallel stable test:51.0798ms ~~ 70.647ms!!!
42---parallel stable test:51.0798ms ~~ 70.647ms!!!
43---parallel stable test:51.0798ms ~~ 70.647ms!!!
44---parallel stable test:51.0798ms ~~ 70.647ms!!!
45---parallel stable test:51.0798ms ~~ 70.647ms!!!
46---parallel stable test:51.0798ms ~~ 70.647ms!!!
47---parallel stable test:51.0798ms ~~ 70.647ms!!!
48---parallel stable test:51.0798ms ~~ 70.647ms!!!
49---parallel stable test:51.0798ms ~~ 70.647ms!!!
50---parallel stable test:51.0798ms ~~ 70.647ms!!!
51---parallel stable test:51.0798ms ~~ 70.647ms!!!
52---parallel stable test:51.0798ms ~~ 78.3649ms!!!
53---parallel stable test:51.0798ms ~~ 78.3649ms!!!
54---parallel stable test:51.0798ms ~~ 78.3649ms!!!
55---parallel stable test:51.0798ms ~~ 78.3649ms!!!
56---parallel stable test:51.0798ms ~~ 78.3649ms!!!
57---parallel stable test:50.712ms ~~ 78.3649ms!!!
58---parallel stable test:50.712ms ~~ 78.3649ms!!!
59---parallel stable test:50.712ms ~~ 78.3649ms!!!
60---parallel stable test:50.712ms ~~ 78.3649ms!!!
61---parallel stable test:50.712ms ~~ 78.3649ms!!!
62---parallel stable test:50.712ms ~~ 78.3649ms!!!
63---parallel stable test:50.712ms ~~ 78.3649ms!!!
64---parallel stable test:50.712ms ~~ 78.3649ms!!!
65---parallel stable test:50.712ms ~~ 78.3649ms!!!
66---parallel stable test:50.712ms ~~ 78.3649ms!!!
67---parallel stable test:50.712ms ~~ 78.3649ms!!!
68---parallel stable test:50.712ms ~~ 78.3649ms!!!
69---parallel stable test:50.712ms ~~ 78.3649ms!!!
70---parallel stable test:50.712ms ~~ 78.3649ms!!!
71---parallel stable test:50.712ms ~~ 78.3649ms!!!
72---parallel stable test:50.712ms ~~ 78.3649ms!!!
73---parallel stable test:50.712ms ~~ 78.3649ms!!!
74---parallel stable test:50.712ms ~~ 78.3649ms!!!
75---parallel stable test:50.712ms ~~ 78.3649ms!!!
76---parallel stable test:50.712ms ~~ 78.3649ms!!!
77---parallel stable test:50.712ms ~~ 78.3649ms!!!
78---parallel stable test:50.712ms ~~ 78.3649ms!!!
79---parallel stable test:50.712ms ~~ 78.3649ms!!!
80---parallel stable test:50.712ms ~~ 78.3649ms!!!
81---parallel stable test:50.712ms ~~ 78.3649ms!!!
82---parallel stable test:50.712ms ~~ 78.3649ms!!!
83---parallel stable test:50.712ms ~~ 78.3649ms!!!
84---parallel stable test:50.712ms ~~ 78.3649ms!!!
85---parallel stable test:50.712ms ~~ 78.3649ms!!!
86---parallel stable test:50.712ms ~~ 78.3649ms!!!
87---parallel stable test:50.712ms ~~ 78.3649ms!!!
88---parallel stable test:50.712ms ~~ 78.3649ms!!!
89---parallel stable test:50.712ms ~~ 78.3649ms!!!
90---parallel stable test:50.712ms ~~ 78.3649ms!!!
91---parallel stable test:50.712ms ~~ 78.3649ms!!!
92---parallel stable test:50.712ms ~~ 78.3649ms!!!
93---parallel stable test:50.712ms ~~ 78.3649ms!!!
94---parallel stable test:50.712ms ~~ 78.3649ms!!!
95---parallel stable test:50.712ms ~~ 78.3649ms!!!
96---parallel stable test:50.712ms ~~ 78.3649ms!!!
97---parallel stable test:50.712ms ~~ 78.3649ms!!!
98---parallel stable test:50.712ms ~~ 78.3649ms!!!
99---parallel stable test:50.712ms ~~ 78.3649ms!!!
可以看到，对于1560幅图耗时在50~78ms之间，非常不稳定！！！！

可见无论串/并行，耗时跳变都大，后面终于找到原因，解决以后耗时就非常非常稳定了，如下：

解决了耗时跳变后的稳定性测试，每次1560幅图：串行测试serial performance:
load model successfully !  
0---serial stable test:41.7413ms ~~ 41.7413ms!!!
1---serial stable test:41.7413ms ~~ 44.2572ms!!!
2---serial stable test:41.7413ms ~~ 44.2572ms!!!
3---serial stable test:41.7413ms ~~ 44.2572ms!!!
4---serial stable test:41.7413ms ~~ 44.2572ms!!!
5---serial stable test:41.7413ms ~~ 44.2572ms!!!
6---serial stable test:41.7413ms ~~ 44.2572ms!!!
7---serial stable test:41.7413ms ~~ 44.2572ms!!!
8---serial stable test:41.7413ms ~~ 44.2572ms!!!
9---serial stable test:41.7413ms ~~ 44.2572ms!!!
10---serial stable test:41.7413ms ~~ 44.2572ms!!!
11---serial stable test:41.7413ms ~~ 44.2572ms!!!
12---serial stable test:41.7413ms ~~ 44.2572ms!!!
13---serial stable test:41.7413ms ~~ 44.2572ms!!!
14---serial stable test:41.7413ms ~~ 44.2572ms!!!
15---serial stable test:41.7413ms ~~ 44.2572ms!!!
16---serial stable test:41.7413ms ~~ 44.2572ms!!!
17---serial stable test:41.7413ms ~~ 44.2572ms!!!
18---serial stable test:41.7277ms ~~ 44.2572ms!!!
19---serial stable test:41.7277ms ~~ 44.2572ms!!!
20---serial stable test:41.7277ms ~~ 44.2572ms!!!
21---serial stable test:41.7277ms ~~ 44.2572ms!!!
22---serial stable test:41.7277ms ~~ 44.2572ms!!!
23---serial stable test:41.7277ms ~~ 44.2572ms!!!
24---serial stable test:41.7277ms ~~ 44.2572ms!!!
25---serial stable test:41.7277ms ~~ 44.2572ms!!!
26---serial stable test:41.7277ms ~~ 44.2572ms!!!
27---serial stable test:41.7277ms ~~ 44.2572ms!!!
28---serial stable test:41.7277ms ~~ 44.2572ms!!!
29---serial stable test:41.7277ms ~~ 44.2572ms!!!
30---serial stable test:41.7277ms ~~ 44.2572ms!!!
31---serial stable test:41.7277ms ~~ 44.2572ms!!!
32---serial stable test:41.7277ms ~~ 44.2572ms!!!
33---serial stable test:41.7277ms ~~ 44.2572ms!!!
34---serial stable test:41.7277ms ~~ 44.2572ms!!!
35---serial stable test:41.7277ms ~~ 44.2572ms!!!
36---serial stable test:41.7277ms ~~ 44.2572ms!!!
37---serial stable test:41.7277ms ~~ 44.2572ms!!!
38---serial stable test:41.7277ms ~~ 44.2572ms!!!
39---serial stable test:41.7277ms ~~ 44.2572ms!!!
40---serial stable test:41.7277ms ~~ 44.2572ms!!!
41---serial stable test:41.7277ms ~~ 44.2572ms!!!
42---serial stable test:41.7277ms ~~ 44.2572ms!!!
43---serial stable test:41.7277ms ~~ 44.2572ms!!!
44---serial stable test:41.7277ms ~~ 44.2572ms!!!
45---serial stable test:41.7277ms ~~ 44.2572ms!!!
46---serial stable test:41.7277ms ~~ 44.2572ms!!!
47---serial stable test:41.7277ms ~~ 44.2572ms!!!
48---serial stable test:41.7277ms ~~ 44.2572ms!!!
49---serial stable test:41.7277ms ~~ 44.2572ms!!!
50---serial stable test:41.7277ms ~~ 44.2572ms!!!
51---serial stable test:41.7277ms ~~ 44.2572ms!!!
52---serial stable test:41.7277ms ~~ 44.2572ms!!!
53---serial stable test:41.7277ms ~~ 44.2572ms!!!
54---serial stable test:41.7277ms ~~ 44.2572ms!!!
55---serial stable test:41.7277ms ~~ 44.2572ms!!!
56---serial stable test:41.7277ms ~~ 44.2572ms!!!
57---serial stable test:41.7277ms ~~ 44.2572ms!!!
58---serial stable test:41.7277ms ~~ 44.2572ms!!!
59---serial stable test:41.7277ms ~~ 44.2572ms!!!
60---serial stable test:41.7277ms ~~ 44.2572ms!!!
61---serial stable test:41.7277ms ~~ 44.2572ms!!!
62---serial stable test:41.7277ms ~~ 44.2572ms!!!
63---serial stable test:41.7277ms ~~ 44.2572ms!!!
64---serial stable test:41.7277ms ~~ 44.2572ms!!!
65---serial stable test:41.7277ms ~~ 44.2572ms!!!
66---serial stable test:41.7277ms ~~ 44.2572ms!!!
67---serial stable test:41.7277ms ~~ 44.2572ms!!!
68---serial stable test:41.7277ms ~~ 44.2572ms!!!
69---serial stable test:41.7277ms ~~ 44.2572ms!!!
70---serial stable test:41.7277ms ~~ 44.2572ms!!!
71---serial stable test:41.7277ms ~~ 44.2572ms!!!
72---serial stable test:41.7277ms ~~ 44.2572ms!!!
73---serial stable test:41.7277ms ~~ 44.2572ms!!!
74---serial stable test:41.7277ms ~~ 44.2572ms!!!
75---serial stable test:41.7277ms ~~ 44.2572ms!!!
76---serial stable test:41.7277ms ~~ 44.2572ms!!!
77---serial stable test:41.7277ms ~~ 44.2572ms!!!
78---serial stable test:41.7277ms ~~ 44.2572ms!!!
79---serial stable test:41.7277ms ~~ 44.2572ms!!!
80---serial stable test:41.7277ms ~~ 46.7984ms!!!
81---serial stable test:41.7277ms ~~ 46.7984ms!!!
82---serial stable test:41.7277ms ~~ 46.7984ms!!!
83---serial stable test:41.7277ms ~~ 46.7984ms!!!
84---serial stable test:41.7277ms ~~ 46.7984ms!!!
85---serial stable test:41.7277ms ~~ 46.7984ms!!!
86---serial stable test:41.7277ms ~~ 46.7984ms!!!
87---serial stable test:41.7277ms ~~ 46.7984ms!!!
88---serial stable test:41.7277ms ~~ 46.7984ms!!!
89---serial stable test:41.7277ms ~~ 46.7984ms!!!
90---serial stable test:41.7277ms ~~ 46.7984ms!!!
91---serial stable test:41.7277ms ~~ 46.7984ms!!!
92---serial stable test:41.7277ms ~~ 46.7984ms!!!
93---serial stable test:41.7277ms ~~ 46.7984ms!!!
94---serial stable test:41.7277ms ~~ 46.7984ms!!!
95---serial stable test:41.7277ms ~~ 46.7984ms!!!
96---serial stable test:41.7277ms ~~ 46.7984ms!!!
97---serial stable test:41.7277ms ~~ 46.7984ms!!!
98---serial stable test:41.7277ms ~~ 46.7984ms!!!
99---serial stable test:41.7277ms ~~ 46.7984ms!!!
和之前未解决耗时跳变时的耗时测试对比可以看到，对于1560幅图耗时只在41~46ms之间，已经非常稳定了。并行测试：
parallel performance:
load model successfully !  
0---parallel stable test:9.63155ms ~~ 9.63155ms!!!
1---parallel stable test:8.76745ms ~~ 9.63155ms!!!
2---parallel stable test:8.76745ms ~~ 9.63155ms!!!
3---parallel stable test:8.76745ms ~~ 9.63155ms!!!
4---parallel stable test:8.76745ms ~~ 9.63155ms!!!
5---parallel stable test:8.76745ms ~~ 9.63155ms!!!
6---parallel stable test:8.76745ms ~~ 9.63155ms!!!
7---parallel stable test:8.76745ms ~~ 9.63155ms!!!
8---parallel stable test:8.76745ms ~~ 10.7076ms!!!
9---parallel stable test:8.76745ms ~~ 12.2607ms!!!
10---parallel stable test:8.76745ms ~~ 12.2607ms!!!
11---parallel stable test:8.76745ms ~~ 12.2607ms!!!
12---parallel stable test:8.76745ms ~~ 12.926ms!!!
13---parallel stable test:8.76745ms ~~ 12.926ms!!!
14---parallel stable test:8.76745ms ~~ 12.926ms!!!
15---parallel stable test:8.76745ms ~~ 14.8526ms!!!
16---parallel stable test:8.76745ms ~~ 14.8526ms!!!
17---parallel stable test:8.76745ms ~~ 14.8526ms!!!
18---parallel stable test:8.76745ms ~~ 14.8526ms!!!
19---parallel stable test:8.76745ms ~~ 14.8526ms!!!
20---parallel stable test:8.76745ms ~~ 14.8526ms!!!
21---parallel stable test:8.76745ms ~~ 14.8526ms!!!
22---parallel stable test:8.76745ms ~~ 14.8526ms!!!
23---parallel stable test:8.76745ms ~~ 14.8526ms!!!
24---parallel stable test:8.76745ms ~~ 14.8526ms!!!
25---parallel stable test:8.76745ms ~~ 14.8526ms!!!
26---parallel stable test:8.76745ms ~~ 14.8526ms!!!
27---parallel stable test:8.76745ms ~~ 14.8526ms!!!
28---parallel stable test:8.76745ms ~~ 14.8526ms!!!
29---parallel stable test:8.76745ms ~~ 14.8526ms!!!
30---parallel stable test:8.76745ms ~~ 14.8526ms!!!
31---parallel stable test:8.76745ms ~~ 14.8526ms!!!
32---parallel stable test:8.76745ms ~~ 14.8526ms!!!
33---parallel stable test:8.76745ms ~~ 14.8526ms!!!
34---parallel stable test:8.76745ms ~~ 14.8526ms!!!
35---parallel stable test:8.76745ms ~~ 14.8526ms!!!
36---parallel stable test:8.76745ms ~~ 14.8526ms!!!
37---parallel stable test:8.76745ms ~~ 14.8526ms!!!
38---parallel stable test:8.76745ms ~~ 14.8526ms!!!
39---parallel stable test:8.76745ms ~~ 14.8526ms!!!
40---parallel stable test:8.76745ms ~~ 14.8526ms!!!
41---parallel stable test:8.76745ms ~~ 14.8526ms!!!
42---parallel stable test:8.76745ms ~~ 14.8526ms!!!
43---parallel stable test:8.76745ms ~~ 14.8526ms!!!
44---parallel stable test:8.76745ms ~~ 14.8526ms!!!
45---parallel stable test:8.76745ms ~~ 14.8526ms!!!
46---parallel stable test:8.76745ms ~~ 14.8526ms!!!
47---parallel stable test:8.76745ms ~~ 14.8526ms!!!
48---parallel stable test:8.76745ms ~~ 14.8526ms!!!
49---parallel stable test:8.76745ms ~~ 14.8526ms!!!
50---parallel stable test:8.76745ms ~~ 14.8526ms!!!
51---parallel stable test:8.76745ms ~~ 14.8526ms!!!
52---parallel stable test:8.76745ms ~~ 14.8526ms!!!
53---parallel stable test:8.76745ms ~~ 14.8526ms!!!
54---parallel stable test:8.76745ms ~~ 14.8526ms!!!
55---parallel stable test:8.76745ms ~~ 14.8526ms!!!
56---parallel stable test:8.76745ms ~~ 14.8526ms!!!
57---parallel stable test:8.76745ms ~~ 14.8526ms!!!
58---parallel stable test:8.76745ms ~~ 14.8526ms!!!
59---parallel stable test:8.76745ms ~~ 14.8526ms!!!
60---parallel stable test:8.76745ms ~~ 14.8526ms!!!
61---parallel stable test:8.76745ms ~~ 14.8526ms!!!
62---parallel stable test:8.76745ms ~~ 14.8526ms!!!
63---parallel stable test:8.76745ms ~~ 14.8526ms!!!
64---parallel stable test:8.76745ms ~~ 14.8526ms!!!
65---parallel stable test:8.76745ms ~~ 14.8526ms!!!
66---parallel stable test:8.76745ms ~~ 14.8526ms!!!
67---parallel stable test:8.76745ms ~~ 14.8526ms!!!
68---parallel stable test:8.76745ms ~~ 14.8526ms!!!
69---parallel stable test:8.76745ms ~~ 14.8526ms!!!
70---parallel stable test:8.76745ms ~~ 14.8526ms!!!
71---parallel stable test:8.76745ms ~~ 14.8526ms!!!
72---parallel stable test:8.76745ms ~~ 14.8526ms!!!
73---parallel stable test:8.5113ms ~~ 14.8526ms!!!
74---parallel stable test:8.5113ms ~~ 14.8526ms!!!
75---parallel stable test:8.5113ms ~~ 14.8526ms!!!
76---parallel stable test:8.5113ms ~~ 14.8526ms!!!
77---parallel stable test:8.5113ms ~~ 14.8526ms!!!
78---parallel stable test:8.5113ms ~~ 14.8526ms!!!
79---parallel stable test:8.5113ms ~~ 14.8526ms!!!
80---parallel stable test:8.5113ms ~~ 14.8526ms!!!
81---parallel stable test:8.5113ms ~~ 14.8526ms!!!
82---parallel stable test:8.5113ms ~~ 14.8526ms!!!
83---parallel stable test:8.5113ms ~~ 14.8526ms!!!
84---parallel stable test:8.5113ms ~~ 14.8526ms!!!
85---parallel stable test:8.5113ms ~~ 14.8526ms!!!
86---parallel stable test:8.5113ms ~~ 14.8526ms!!!
87---parallel stable test:8.5113ms ~~ 14.8526ms!!!
88---parallel stable test:8.5113ms ~~ 14.8526ms!!!
89---parallel stable test:8.5113ms ~~ 14.8526ms!!!
90---parallel stable test:8.5113ms ~~ 14.8526ms!!!
91---parallel stable test:8.5113ms ~~ 14.8526ms!!!
92---parallel stable test:8.5113ms ~~ 14.8526ms!!!
93---parallel stable test:8.5113ms ~~ 14.8526ms!!!
94---parallel stable test:8.5113ms ~~ 14.8526ms!!!
95---parallel stable test:8.5113ms ~~ 14.8526ms!!!
96---parallel stable test:8.5113ms ~~ 14.8526ms!!!
97---parallel stable test:8.47865ms ~~ 14.8526ms!!!
98---parallel stable test:8.47865ms ~~ 14.8526ms!!!
99---parallel stable test:8.47865ms ~~ 14.8526ms!!!
和之前未解决耗时跳变时的耗时测试对比可以看到对于1560幅图，耗时跳变只在8~14ms之间，也已经非常稳定了。

 同样的代码，同样的1560张图片，图像的模型，可以看到必须要解决耗时跳变问题！！！解决后就可以工业应用了！！！

另外将[LightGBM] [Warning] Unknown parameter :None 警告消除方式也在刚刚发的那个下载链接里：

 

 看解决跳变后，单张图片的耗时也非常非常非常稳定。始终在0.0x毫秒，测试代码也在刚发的链接里。

~~~~~~~~~~~~~~~~~~~~~~~~~分界线~~~~~~batch predict~~~~~~~~~~~~~~~~~~~~~~~~~~~

另外又查到多线程问题：Lightgbm多线程卡死问题定位 | 逸思杂陈

不知道LightGBM是否真的有这个问题。

查到https://github.com/microsoft/LightGBM/issues/1534 

这是说如果predict的输入是单条特征，那么就只有1个CPU core；如果输入是多条特征，也就是batch predict，使用了多个CPU cores？然后我试了一下predict多个samples：

int main()
{char srcimg[400]={0};int numiterations = 1;BoosterHandle handle;int set = LGBM_BoosterCreateFromModelfile("/home/jumper/imgs/lgbmmodel/lbgm_zhu_1.model", &numiterations, &handle);if(set==0){std::cout << "load model successfully !  "<< std::endl;}int channels[]={1,2};int histsize[]={8,8};float ghistrange[]={0,256};float rhistrange[]={0,256};const float *histsranges[]={ghistrange,rhistrange};const int samplesnum=11;//2228;const int featuredim=64;Mat paradatas=Mat::zeros(samplesnum,featuredim,CV_32FC1);for(int index=0;index!=samplesnum;index++){sprintf(srcimg,"/home/jumper/imgs/cnntmp/doctorzhu_fluorite/orescnn/%d.png",index);Mat img=imread(srcimg);if(img.empty())continue;MatND hist;cv::calcHist(&img,1,channels,Mat(),hist,2,histsize,histsranges);for(int tmpi=0;tmpi!=64;tmpi++){paradatas.ptr<float>(index)[tmpi]=hist.data[tmpi];cout<<paradatas.ptr<float>(index)[tmpi]<<" ";}cout<<endl;}std::cout <<"created data..." <<std::endl;std::vector<double> out(samplesnum, 0);double *out_result = static_cast<double *>(out.data());int64_t out_len;int res = LGBM_BoosterPredictForMat(handle,paradatas.data,C_API_DTYPE_FLOAT32,samplesnum,featuredim,1,C_API_PREDICT_NORMAL,0,-1,"None",&out_len,out_result);for(int id=0;id!=out_len;id++){cout<<"sample id: "<<id<<" score:"<<out[id]<<endl;}return 0;
}

输出：

load model successfully !  
0 0 8 67 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 128 21 68 0 0 26 67 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
0 0 64 64 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 215 67 0 0 174 66 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
0 0 128 63 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 128 166 68 0 0 161 67 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
0 0 224 64 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 128 236 67 0 0 48 66 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
0 0 124 66 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 203 67 0 0 64 65 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
0 0 8 66 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 241 67 0 0 36 66 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
0 128 222 67 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 192 13 68 0 0 218 66 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
0 192 41 68 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 32 153 68 0 0 198 66 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
0 0 186 66 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 192 37 68 0 0 60 66 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 162 66 0 0 248 65 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
created data...
[LightGBM] [Warning] Unknown parameter: None
sample id: 0 score:0.847249
sample id: 1 score:0.820929
sample id: 2 score:0.772399
sample id: 3 score:0.772399
sample id: 4 score:0.820929
sample id: 5 score:0.820929
sample id: 6 score:0.845796
sample id: 7 score:0.772399
sample id: 8 score:0.844833
sample id: 9 score:0.400716
sample id: 10 score:0.820929

为什么输出概率与single sample predict不一样？？我把特征打印出来了 明明特征都是一样的，但结果不一样。batch的精度不正确，为什么比single的低？！！single sample predict的是：

load model successfully !  
0 0 8 67 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 128 21 68 0 0 26 67 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
[LightGBM] [Warning] Unknown parameter: None
image id:0 ---LGBM row predict result is: 0.964677
0 0 64 64 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 215 67 0 0 174 66 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
[LightGBM] [Warning] Unknown parameter: None
image id:1 ---LGBM row predict result is: 0.877513
0 0 128 63 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 128 166 68 0 0 161 67 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
[LightGBM] [Warning] Unknown parameter: None
image id:2 ---LGBM row predict result is: 0.973227
0 0 224 64 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 128 236 67 0 0 48 66 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
[LightGBM] [Warning] Unknown parameter: None
image id:3 ---LGBM row predict result is: 0.895759
0 0 124 66 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 203 67 0 0 64 65 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
[LightGBM] [Warning] Unknown parameter: None
image id:4 ---LGBM row predict result is: 0.945096
0 0 8 66 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 241 67 0 0 36 66 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
[LightGBM] [Warning] Unknown parameter: None
image id:5 ---LGBM row predict result is: 0.792787
0 128 222 67 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 192 13 68 0 0 218 66 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
[LightGBM] [Warning] Unknown parameter: None
image id:6 ---LGBM row predict result is: 0.902854
0 192 41 68 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 32 153 68 0 0 198 66 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
[LightGBM] [Warning] Unknown parameter: None
image id:7 ---LGBM row predict result is: 0.965496
0 0 186 66 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 192 37 68 0 0 60 66 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
[LightGBM] [Warning] Unknown parameter: None
image id:8 ---LGBM row predict result is: 0.92893
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
[LightGBM] [Warning] Unknown parameter: None
image id:9 ---LGBM row predict result is: 0.110013
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 162 66 0 0 248 65 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
[LightGBM] [Warning] Unknown parameter: None
image id:10 ---LGBM row predict result is: 0.903951

这个概率才是对的，可以看到single sample与batch sample predict概率不一样，为何？

抛开score不同问题，想一直跑试试，发现真的每个核几乎都到了100%！！

 

 也就是证明batch predict在batch够大时是自动并行了。另外之前的single sample predict-serial 和single sample predict-parallel都是自动并行的，核都会跑起来，因为predict函数内部用了omp。

至于batch predict时的精度低的问题，不知道是不是LGBM C++内部并行导致的，目前没查到有用的资料，如果大家发现也可以告诉我。所以我们还是先像我链接里那样稳妥的用single predict吧，毕竟这与python下一致。

single sample predict-serial ：

 single sample predict-parallel下：

至此，ubuntu下LightGBM C++应用的所有问题均已解决。用在工程中一段时间，没问题。

~~~~~~~~~~~~~~~~~~~~~~windows下 lightgbm c++~~~~~~~~~~~~~~

 今天我又在另一台windows10上按照 Installation Guide — LightGBM 3.3.2.99 documentation 步骤安装：记得git也要添加进环境变量

可以看到生成的库和lightgbm应用程序，记得将此时的路径添加进环境变量：

  

 然后开始按照Readme中的操作跑示例代码，如下得到正确结果：

 

但是在我用lib_lightgbm.dll在测试时却出现：

G:\workspace\lightGBM\include/LightGBM/utils/../../../external_libs/fmt/include/fmt/format-inl.h: In function 'void fmt::v8::detail::print(FILE*, fmt::v8::string_view)':
G:\workspace\lightGBM\include/LightGBM/utils/../../../external_libs/fmt/include/fmt/format-inl.h:2603:22: error: '_fileno' was not declared in this scopeauto fd = _fileno(f);^
In file included from G:\workspace\lightGBM\include/LightGBM/config.h:16:0,from G:\workspace\lightGBM\include/LightGBM/application.h:8,from ..\src\DualEnFitOreStoneRecogProc.h:22,from ..\src\XRT_PC_Stream_Mathprocess.h:11,from ..\src\XRT_PC_Stream_Mathprocess.cpp:25:
G:\workspace\lightGBM\include/LightGBM/utils/common.h: In member function 'T LightGBM::Common::__StringToTHelper<T, true>::operator()(const string&) const':
G:\workspace\lightGBM\include/LightGBM/utils/common.h:426:27: error: 'stod' is not a member of 'std'return static_cast<T>(std::stod(str));

解决办法也在我的链接里，现在windows c++下使用LGBM 编译已经没问题了：

然后运行也OK，如下所示：

如果大家转移到别的电脑出现如下问题：

eclipse exit value -1073741701
应用程序无法正常启动0xc000007b，请单击确定关闭应用程序

则应该先关闭软件，然后如下设置软件属性（设置一次即可），即可解决（感谢大佬的指点）：

至此windows、ubuntu下LGBM C++动态库已经均完成编译、测试官例、测试自己的工程，均可正常使用。