python下已测试通过，无问题：
然而C++下：
#include <iostream>
#include <LightGBM/application.h>
#include <LightGBM/c_api.h>
#include <iostream>
#include <vector>
#include <opencv2/opencv.hpp>
#include <opencv2/highgui/highgui.hpp>
#include <opencv2/imgproc/imgproc.hpp>
#include <opencv2/core/core.hpp>
#include <stdlib.h>
#include <stdio.h>
#include <math.h>
//#include <tbb/tbb.h>
//using namespace tbb;
//https://cloud.tencent.com/developer/ask/sof/781779
/*
I had a similar issue and in my case I found that the problem was the is_linear property in the model.
I compared the model that I generated from the binary_classification example with the model I was using
and I noticed that the model in the example has the is_linear=0 property for each tree.
On my model it was missing.
Then I checked the c++ code and found that if this property is missing,
the variable describing this is true. I set it to false as default and that works for me.
I can't give more details as I just recently began working with LGBM models and c++.
* */
using namespace cv;
using namespace std;
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[]={0,255};
float rhistrange[]={0,255};
const float *histsranges[]={ghistrange,rhistrange};
for(int index=0;index<=100;index++)
{
sprintf(srcimg,"/home/jumper/xrt/reference/imgs/zhudoctor/orescnn/%d.png",index);
Mat img=imread(srcimg,IMREAD_UNCHANGED);
if(img.empty())
continue;
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,-1,"None",&out_len,out_result);
std::cout <<"image id:"<<index<<" ---predict flag:"<<res<< " ---LGBM row predict result is: " << out[0] << std::endl;
}
return 0;
}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，无任何改变：
load model successfully !
[LightGBM] [Warning] Unknown parameter None
image id:0 ---predict flag:0 ---LGBM row predict result is: 0.400716
[LightGBM] [Warning] Unknown parameter None
image id:1 ---predict flag:0 ---LGBM row predict result is: 0.400716
[LightGBM] [Warning] Unknown parameter None
image id:2 ---predict flag:0 ---LGBM row predict result is: 0.400716
[LightGBM] [Warning] Unknown parameter None
image id:3 ---predict flag:0 ---LGBM row predict result is: 0.400716
[LightGBM] [Warning] Unknown parameter None
image id:4 ---predict flag:0 ---LGBM row predict result is: 0.400716
[LightGBM] [Warning] Unknown parameter None
image id:5 ---predict flag:0 ---LGBM row predict result is: 0.400716
[LightGBM] [Warning] Unknown parameter None
image id:6 ---predict flag:0 ---LGBM row predict result is: 0.400716
[LightGBM] [Warning] Unknown parameter None
image id:7 ---predict flag:0 ---LGBM row predict result is: 0.400716
[LightGBM] [Warning] Unknown parameter None
image id:8 ---predict flag:0 ---LGBM row predict result is: 0.400716
[LightGBM] [Warning] Unknown parameter None
image id:9 ---predict flag:0 ---LGBM row predict result is: 0.400716关于第2点问题，只查到：LightGBM在任何输入上产生相同的概率(C++) - 问答 - 云+社区 - 腾讯云
?最近查不了外网，好烦。等能查外网了估计就可以解决了。
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。然后查了下：https://github.com/microsoft/LightGBM/pull/5111
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]?只是不知道他是怎么解决的？
然而我换了一台机器，同样的操作，不会报错！！但是仍然有之前说的那2点问题：
1，与python下score不一致；
2，不同的输入，但输出score都一样；
若有大神，麻烦告知。
我觉得应该是predict这个函数没用对。
/*!
* \brief Make prediction for a new dataset.
* \note
* You should pre-allocate memory for ``out_result``:
* - for normal and raw score, its length is equal to ``num_class * num_data``;
* - for leaf index, its length is equal to ``num_class * num_data * num_iteration``;
* - for feature contributions, its length is equal to ``num_class * num_data * (num_feature + 1)``.
* \param handle Handle of booster
* \param data Pointer to the data space
* \param data_type Type of ``data`` pointer, can be ``C_API_DTYPE_FLOAT32`` or ``C_API_DTYPE_FLOAT64``
* \param nrow Number of rows
* \param ncol Number of columns
* \param is_row_major 1 for row-major, 0 for column-major
* \param predict_type What should be predicted
* - ``C_API_PREDICT_NORMAL``: normal prediction, with transform (if needed);
* - ``C_API_PREDICT_RAW_SCORE``: raw score;
* - ``C_API_PREDICT_LEAF_INDEX``: leaf index;
* - ``C_API_PREDICT_CONTRIB``: feature contributions (SHAP values)
* \param start_iteration Start index of the iteration to predict
* \param num_iteration Number of iteration for prediction, <= 0 means no limit
* \param parameter Other parameters for prediction, e.g. early stopping for prediction
* \param[out] out_len Length of output result
* \param[out] out_result Pointer to array with predictions
* \return 0 when succeed, -1 when failure happens
*/
LIGHTGBM_C_EXPORT int LGBM_BoosterPredictForMat(BoosterHandle handle,
const void* data,
int data_type,
int32_t nrow,
int32_t ncol,
int is_row_major,
int predict_type,
int start_iteration,
int num_iteration,
const char* parameter,
int64_t* out_len,
double* out_result);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[]={0,255};
float rhistrange[]={0,255};
const float *histsranges[]={ghistrange,rhistrange};
for(int index=0;index<=100;index++)
{
sprintf(srcimg,"/home/jumper/xrt/reference/imgs/zhudoctor/orescnn/%d.png",index);
Mat img=imread(srcimg,IMREAD_UNCHANGED);
if(img.empty())
continue;
MatND hist;
cv::calcHist(&img,1,channels,Mat(),hist,2,histsize,histsranges);
std::vector<double> out(2, 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:"<<index<<" ---predict flag:"<<res<< " ---LGBM row predict result is: " << out[0] << std::endl;
}不好意思，是我自己瞎了狗眼输入图像数据类型没对上，改过来后就好了：
#include <iostream>
#include <LightGBM/application.h>
#include <LightGBM/c_api.h>
#include <iostream>
#include <vector>
#include <opencv2/opencv.hpp>
#include <opencv2/highgui/highgui.hpp>
#include <opencv2/imgproc/imgproc.hpp>
#include <opencv2/core/core.hpp>
#include <stdlib.h>
#include <stdio.h>
#include <math.h>
using namespace cv;
using namespace std;
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};
for(int index=0;index<=10;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);
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:"<<index<<" ---LGBM row predict result is: " << out[0] <<std::endl;
}
return 0;
}结果已正确：
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所以如果大家C和python结果没对上，那就检查预处理是否对上了。
~~~~~~~~~~~~~~~~~~~~~~~~~分界线~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
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。
