bool  OnnxRun(
   long    onnx_handle,  // ONNX session handle
   ulong   flags,        // flags describing the run mode
   ...                   // model's inputs and outputs
   );

const long                             ExtOutputShape[] = {1,1};    // model output shape
const long                             ExtInputShape [] = {1,10,4}; // model input form
#resource "Python/model.onnx" as uchar ExtModel[]                   // model as resource
//+------------------------------------------------------------------+
//| Script program start function                                    |
//+------------------------------------------------------------------+
int OnStart(void)
  {
   matrix rates;
//--- get 10 bars
   if(!rates.CopyRates("EURUSD",PERIOD_H1,COPY_RATES_OHLC,2,10))
      return(-1);
//--- input a set of OHLC vectors
   matrix x_norm=rates.Transpose();
   vector m=x_norm.Mean(0);
   vector s=x_norm.Std(0);
   matrix mm(10,4);
   matrix ms(10,4);
//--- fill in the normalization matrices
   for(int i=0; i<10; i++)
     {
      mm.Row(m,i);
      ms.Row(s,i);
     }
//--- normalize the input data
   x_norm-=mm;
   x_norm/=ms;
//--- create the model
   long handle=OnnxCreateFromBuffer(ExtModel,ONNX_DEBUG_LOGS);
//--- specify the shape of the input data
   if(!OnnxSetInputShape(handle,0,ExtInputShape))
     {
      Print("OnnxSetInputShape failed, error ",GetLastError());
      OnnxRelease(handle);
      return(-1);
     }
//--- specify the shape of the output data
   if(!OnnxSetOutputShape(handle,0,ExtOutputShape))
     {
      Print("OnnxSetOutputShape failed, error ",GetLastError());
      OnnxRelease(handle);
      return(-1);
     }
//--- convert normalized input data to float type
   matrixf x_normf;
   x_normf.Assign(x_norm);
//--- get the output data of the model here, i.e. the price prediction
   vectorf y_norm(1);
//--- run the model
   if(!OnnxRun(handle,ONNX_DEBUG_LOGS | ONNX_NO_CONVERSION,x_normf,y_norm))
     {
      Print("OnnxRun failed, error ",GetLastError());
      OnnxRelease(handle);
      return(-1);
     }
//--- print the output value of the model to the log
   Print(y_norm);
//--- do the reverse transformation to get the predicted price
   double y_pred=y_norm[0]*s[3]+m[3];
   Print("price predicted:",y_pred);
//--- completed operation
   OnnxRelease(handle);
   return(0);
  };