Python SVM model in C++

[joshua-777]

Hi, I have trained an SVM model in python with 8 features and a label for binary classification. I have also generated an onnx model of the same.

   import pandas as pd
   from sklearn.model_selection import train_test_split
   from sklearn.svm import SVC
   from sklearn.metrics import accuracy_score, classification_report
   from sklearn.preprocessing import StandardScaler
   from joblib import dump
   from skl2onnx import convert_sklearn
   from skl2onnx.common.data_types import FloatTensorType

   df = pd.read_csv('nothing.csv')

   X = df.iloc[:, :-1] 
   y = df.iloc[:, -1]  

   model = SVC(kernel='linear')
   model.fit(X, y)

   initial_type = [('float_input', FloatTensorType([None, 8]))]
   onnx_model = convert_sklearn(model, initial_types=initial_type)
   with open("svm.onnx", "wb") as f:
       f.write(onnx_model.SerializeToString())

---

I am trying to run it in onnx runtime in C++ to get the desired classification of real-time features. I used the example of youtube video by Italian cpp that runs a linear regression onnx model in c++.

int main(int argc, char* argv[]) {

Ort::Env env;

Ort::Session session{ env, LR"(models\linear.onnx)", Ort::SessionOptions{}};

Ort::AllocatorWithDefaultOptions allocator;
auto* input_name = session.GetInputName(0, allocator);
auto* output_name = session.GetOutputName(0, allocator);
auto input_shape = session.GetInputTypeInfo(0).GetTensorTypeAndShapeInfo().GetShape();
std::vector<float> input_data{ 9,5,6 };
auto memoryInfo = Ort::MemoryInfo::CreateCpu(OrtDeviceAllocator, OrtMemTypeCPU);
auto inputOnnxTensor = Ort::Value::CreateTensor<float>(memoryInfo, input_data.data(), input_data.size(), input_shape.data(), 
input_shape.size());

std::vector<const char*> inputNames = { input_name };
std::vector<const char*> outputNames = { output_name };
// Run inference
auto outputValues = session.Run(Ort::RunOptions{ nullptr }, inputNames.data(), &inputOnnxTensor, 1, outputNames.data(), 1);

// Get the output tensor data
auto& output1 = outputValues[0];
const auto* floats = output1.GetTensorMutableData<float>();
const auto floatsCount = output1.GetTensorTypeAndShapeInfo().GetElementCount();

std::copy_n(floats, floatsCount, std::ostream_iterator<float>(std::cout, " "));
allocator.Free(input_name);
allocator.Free(output_name);
}

It works fine for that model. But when i change the model to the svm onnx model, i change input_data with 8 values for 8 features in order to generate the classification, the code doesn't even run entirely. I am a beginner using onnx runtime, What changes must be made in the code for running the svm model in onnx runtime that accepts 8 features for binary classification. Please help, it is urgent.