Make sure to properly support newer versions of onnxruntime

CMakeLists.txt

@@ -28,11 +28,20 @@ find_package(k4FWCore) # implicit: Gaudi
 find_package(EDM4HEP) # implicit: Podio
 find_package(DD4hep)
 find_package(k4geo)
-find_package(ONNXRuntime)
 find_package(FastJet)
 # New versions of ONNRuntime package provide onnxruntime-Config.cmake
 # and use the name onnxruntime
 find_package(onnxruntime)
+if (NOT onnxruntime_FOUND)
+  message(STATUS "Could not find onnxruntime (> 1.17.1). Looking for an older version")
+  find_package(ONNXRuntime)
+endif()
+
+if(onnxruntime_FOUND OR ONNXRuntime_FOUND)
+else()
+  message(FATAL_ERROR "Failed to locate ONNXRuntime!")
+endif()
+
 #---------------------------------------------------------------
 
 

RecFCCeeCalorimeter/CMakeLists.txt

@@ -2,17 +2,6 @@
 # Package: RecFCCeeCalorimeter
 ################################################################################
 
-# ONNX DEPENDENCIES
-# includes
-include_directories("${ONNXRUNTIME_INCLUDE_DIRS}")
-# libs
-link_directories("${ONNXRUNTIME_LIBRARY_DIRS}")
-# New versions of ONNXRuntime add directories to include
-# through the target onnxruntime::onnxruntime
-if(onnxruntime_FOUND)
-  set(ONNXRUNTIME_LIBRARY onnxruntime::onnxruntime)
-endif()
-
 file(GLOB _module_sources src/components/*.cpp)
 gaudi_add_module(k4RecFCCeeCalorimeterPlugins
                  SOURCES ${_module_sources}
@@ -26,7 +15,7 @@ gaudi_add_module(k4RecFCCeeCalorimeterPlugins
                       DD4hep::DDG4
                       ROOT::Core
                       ROOT::Hist
-                      ${ONNXRUNTIME_LIBRARY}
+                      onnxruntime::onnxruntime
                       nlohmann_json::nlohmann_json
                       )
 install(TARGETS k4RecFCCeeCalorimeterPlugins

RecFCCeeCalorimeter/src/components/CalibrateCaloClusters.cpp

@@ -13,23 +13,18 @@
 #include "edm4hep/Cluster.h"
 #include "edm4hep/ClusterCollection.h"
 #include "edm4hep/CalorimeterHitCollection.h"
+#include <onnxruntime_cxx_api.h>
+
+#include "OnnxruntimeUtilities.h"
 
 DECLARE_COMPONENT(CalibrateCaloClusters)
 
-// convert vector data with given shape into ONNX runtime tensor
-template <typename T>
-Ort::Value vec_to_tensor(std::vector<T> &data, const std::vector<std::int64_t> &shape)
-{
-  Ort::MemoryInfo mem_info =
-      Ort::MemoryInfo::CreateCpu(OrtAllocatorType::OrtArenaAllocator, OrtMemType::OrtMemTypeDefault);
-  auto tensor = Ort::Value::CreateTensor<T>(mem_info, data.data(), data.size(), shape.data(), shape.size());
-  return tensor;
-}
 
 CalibrateCaloClusters::CalibrateCaloClusters(const std::string &name,
                                              ISvcLocator *svcLoc)
     : Gaudi::Algorithm(name, svcLoc),
-      m_geoSvc("GeoSvc", "CalibrateCaloClusters")
+      m_geoSvc("GeoSvc", "CalibrateCaloClusters"),
+      m_ortMemInfo(Ort::MemoryInfo::CreateCpu(OrtAllocatorType::OrtArenaAllocator, OrtMemType::OrtMemTypeDefault))
 {
   declareProperty("inClusters", m_inClusters,
                   "Input cluster collection");
@@ -227,7 +222,7 @@ StatusCode CalibrateCaloClusters::readCalibrationFile(const std::string &calibra
     m_ortEnv = new Ort::Env(loggingLevel, "ONNX runtime environment");
     Ort::SessionOptions session_options;
     session_options.SetIntraOpNumThreads(1);
-    m_ortSession = new Ort::Experimental::Session(*m_ortEnv, const_cast<std::string &>(calibrationFile), session_options);
+    m_ortSession = new Ort::Session(*m_ortEnv, calibrationFile.data(), session_options);
   }
   catch (const Ort::Exception &exception)
   {
@@ -239,12 +234,19 @@ StatusCode CalibrateCaloClusters::readCalibrationFile(const std::string &calibra
   // use default allocator (CPU)
   Ort::AllocatorWithDefaultOptions allocator;
   debug() << "Input Node Name/Shape (" << m_input_names.size() << "):" << endmsg;
+#if ORT_API_VERSION < 13
+  // Before 1.13 we have to roll our own unique_ptr wrapper here
+  auto allocDeleter = [&allocator](char* p) { allocator.Free(p); };
+  using AllocatedStringPtr = std::unique_ptr<char, decltype(allocDeleter)>;
+#endif
+
   for (std::size_t i = 0; i < m_ortSession->GetInputCount(); i++)
   {
-    // for old ONNX runtime version
-    // m_input_names.emplace_back(m_ortSession->GetInputName(i, allocator));
-    // for new runtime version
-    m_input_names.emplace_back(m_ortSession->GetInputNameAllocated(i, allocator).get());
+#if ORT_API_VERSION < 13
+    m_input_names.emplace_back(AllocatedStringPtr(m_ortSession->GetInputName(i, allocator), allocDeleter).release());
+#else
+    m_input_names.emplace_back(m_ortSession->GetInputNameAllocated(i, allocator).release());
+#endif
     m_input_shapes = m_ortSession->GetInputTypeInfo(i).GetTensorTypeAndShapeInfo().GetShape();
     debug() << "\t" << m_input_names.at(i) << " : ";
     for (std::size_t k = 0; k < m_input_shapes.size() - 1; k++)
@@ -266,10 +268,11 @@ StatusCode CalibrateCaloClusters::readCalibrationFile(const std::string &calibra
   debug() << "Output Node Name/Shape (" << m_output_names.size() << "):" << endmsg;
   for (std::size_t i = 0; i < m_ortSession->GetOutputCount(); i++)
   {
-    // for old ONNX runtime version
-    // m_output_names.emplace_back(m_ortSession->GetOutputName(i, allocator));
-    // for new runtime version
+#if ORT_API_VERSION < 13
+    m_output_names.emplace_back(AllocatedStringPtr(m_ortSession->GetOutputName(i, allocator), allocDeleter).release());
+#else
     m_output_names.emplace_back(m_ortSession->GetOutputNameAllocated(i, allocator).get());
+#endif
     m_output_shapes = m_ortSession->GetOutputTypeInfo(i).GetTensorTypeAndShapeInfo().GetShape();
     debug() << "\t" << m_output_names.at(i) << " : ";
     for (std::size_t k = 0; k < m_output_shapes.size() - 1; k++)
@@ -329,18 +332,20 @@ StatusCode CalibrateCaloClusters::calibrateClusters(const edm4hep::ClusterCollec
     float corr = 1.0;
     // Create a single Ort tensor
     std::vector<Ort::Value> input_tensors;
-    input_tensors.emplace_back(vec_to_tensor<float>(energiesInLayers, m_input_shapes));
+    input_tensors.emplace_back(vec_to_tensor<float>(energiesInLayers, m_input_shapes, m_ortMemInfo));
 
     // double-check the dimensions of the input tensor
     // assert(input_tensors[0].IsTensor() && input_tensors[0].GetTensorTypeAndShapeInfo().GetShape() == m_input_shapes);
 
     // pass data through model
     try
     {
-      std::vector<Ort::Value> output_tensors = m_ortSession->Run(m_input_names,
-                                                                 input_tensors,
-                                                                 m_output_names,
-                                                                 Ort::RunOptions{nullptr});
+      auto output_tensors = m_ortSession->Run(Ort::RunOptions{nullptr},
+                                              m_input_names.data(),
+                                              input_tensors.data(),
+                                              input_tensors.size(),
+                                              m_output_names.data(),
+                                              m_output_names.size());
 
       // double-check the dimensions of the output tensors
       // NOTE: the number of output tensors is equal to the number of output nodes specifed in the Run() call

RecFCCeeCalorimeter/src/components/CalibrateCaloClusters.h

@@ -26,7 +26,7 @@ namespace dd4hep {
 }
 
 // ONNX
-#include "onnxruntime/core/session/experimental_onnxruntime_cxx_api.h"
+#include "onnxruntime_cxx_api.h"
 
 /** @class CalibrateCaloClusters
  *
@@ -149,12 +149,13 @@ class CalibrateCaloClusters : public Gaudi::Algorithm {
 
   // the ONNX runtime session for applying the calibration,
   // the environment, and the input and output shapes and names
-  Ort::Experimental::Session* m_ortSession = nullptr;
+  Ort::Session* m_ortSession = nullptr;
   Ort::Env* m_ortEnv = nullptr;
+  Ort::MemoryInfo m_ortMemInfo;
   std::vector<std::int64_t> m_input_shapes;
   std::vector<std::int64_t> m_output_shapes;
-  std::vector<std::string> m_input_names;
-  std::vector<std::string> m_output_names;
+  std::vector<const char*> m_input_names;
+  std::vector<const char*> m_output_names;
 
   // the indices of the shapeParameters containing the inputs to the model (if they exist)
   std::vector<unsigned short int> m_inputPositionsInShapeParameters;

RecFCCeeCalorimeter/src/components/OnnxruntimeUtilities.h

@@ -0,0 +1,18 @@
+#ifndef RECFCCEECALORIMETER_ONNXRUNTIMEUTILITIES_H
+#define RECFCCEECALORIMETER_ONNXRUNTIMEUTILITIES_H
+
+#include "onnxruntime_cxx_api.h"
+
+#include <vector>
+
+// convert vector data with given shape into ONNX runtime tensor
+template <typename T>
+Ort::Value vec_to_tensor(std::vector<T> &data,
+                         const std::vector<std::int64_t> &shape,
+                         const Ort::MemoryInfo &mem_info) {
+  auto tensor = Ort::Value::CreateTensor<T>(mem_info, data.data(), data.size(),
+                                            shape.data(), shape.size());
+  return tensor;
+}
+
+#endif

RecFCCeeCalorimeter/src/components/PhotonIDTool.cpp

@@ -8,24 +8,18 @@
 
 #include "nlohmann/json.hpp"
 
+#include "OnnxruntimeUtilities.h"
+
 using json = nlohmann::json;
 
 
 DECLARE_COMPONENT(PhotonIDTool)
 
-// convert vector data with given shape into ONNX runtime tensor
-template <typename T>
-Ort::Value vec_to_tensor(std::vector<T> &data, const std::vector<std::int64_t> &shape)
-{
-  Ort::MemoryInfo mem_info =
-      Ort::MemoryInfo::CreateCpu(OrtAllocatorType::OrtArenaAllocator, OrtMemType::OrtMemTypeDefault);
-  auto tensor = Ort::Value::CreateTensor<T>(mem_info, data.data(), data.size(), shape.data(), shape.size());
-  return tensor;
-}
 
 PhotonIDTool::PhotonIDTool(const std::string &name,
                            ISvcLocator *svcLoc)
-    : Gaudi::Algorithm(name, svcLoc)
+    : Gaudi::Algorithm(name, svcLoc),
+  m_ortMemInfo(Ort::MemoryInfo::CreateCpu(OrtAllocatorType::OrtArenaAllocator, OrtMemType::OrtMemTypeDefault))
 {
   declareProperty("inClusters", m_inClusters, "Input cluster collection");
   declareProperty("outClusters", m_outClusters, "Output cluster collection");
@@ -296,8 +290,7 @@ StatusCode PhotonIDTool::readMVAFiles(const std::string& mvaInputsFileName,
     m_ortEnv = new Ort::Env(loggingLevel, "ONNX runtime environment for photonID");
     Ort::SessionOptions session_options;
     session_options.SetIntraOpNumThreads(1);
-    m_ortSession = new Ort::Experimental::Session(*m_ortEnv, const_cast<std::string &>(mvaModelFileName), session_options);
-    // m_ortSession = new Ort::Session(*m_ortEnv, const_cast<std::string &>(mvaModelFileName), session_options);
+    m_ortSession = new Ort::Session(*m_ortEnv, mvaModelFileName.data(), session_options);
   }
   catch (const Ort::Exception &exception)
   {
@@ -308,14 +301,23 @@ StatusCode PhotonIDTool::readMVAFiles(const std::string& mvaInputsFileName,
   // print name/shape of inputs
   // use default allocator (CPU)
   Ort::AllocatorWithDefaultOptions allocator;
+#if ORT_API_VERSION < 13
+  // Before 1.13 we have to roll our own unique_ptr wrapper here
+  auto allocDeleter = [&allocator](char* p) { allocator.Free(p); };
+  using AllocatedStringPtr = std::unique_ptr<char, decltype(allocDeleter)>;
+#endif
+
+
   debug() << "Input Node Name/Shape (" << m_ortSession->GetInputCount() << "):" << endmsg;
   for (std::size_t i = 0; i < m_ortSession->GetInputCount(); i++)
   {
-    // for old ONNX runtime version
-    // m_input_names.emplace_back(m_ortSession->GetInputName(i, allocator));
-    // for new runtime version
-    m_input_names.emplace_back(m_ortSession->GetInputNameAllocated(i, allocator).get());
-    m_input_shapes = m_ortSession->GetInputTypeInfo(i).GetTensorTypeAndShapeInfo().GetShape(); 
+#if ORT_API_VERSION < 13
+    m_input_names.emplace_back(AllocatedStringPtr(m_ortSession->GetInputName(i, allocator), allocDeleter).release());
+#else
+    m_input_names.emplace_back(m_ortSession->GetInputNameAllocated(i, allocator).release());
+#endif
+
+    m_input_shapes = m_ortSession->GetInputTypeInfo(i).GetTensorTypeAndShapeInfo().GetShape();
     debug() << "\t" << m_input_names.at(i) << " : ";
     for (std::size_t k = 0; k < m_input_shapes.size() - 1; k++)
     {
@@ -336,10 +338,12 @@ StatusCode PhotonIDTool::readMVAFiles(const std::string& mvaInputsFileName,
   debug() << "Output Node Name/Shape (" << m_ortSession->GetOutputCount() << "):" << endmsg;
   for (std::size_t i = 0; i < m_ortSession->GetOutputCount(); i++)
   {
-    // for old ONNX runtime version
-    // m_output_names.emplace_back(m_ortSession->GetOutputName(i, allocator));
-    // for new runtime version
+#if ORT_API_VERSION < 13
+    m_output_names.emplace_back(AllocatedStringPtr(m_ortSession->GetOutputName(i, allocator), allocDeleter).release());
+#else
     m_output_names.emplace_back(m_ortSession->GetOutputNameAllocated(i, allocator).get());
+#endif
+
     m_output_shapes = m_ortSession->GetOutputTypeInfo(i).GetTensorTypeAndShapeInfo().GetShape();
     debug() << m_output_shapes.size() << endmsg;
     debug() << "\t" << m_output_names.at(i) << " : ";
@@ -384,15 +388,17 @@ StatusCode PhotonIDTool::applyMVAtoClusters(const edm4hep::ClusterCollection *in
     float score= -1.0;
     // Create a single Ort tensor
     std::vector<Ort::Value> input_tensors;
-    input_tensors.emplace_back(vec_to_tensor<float>(mvaInputs, m_input_shapes));
+    input_tensors.emplace_back(vec_to_tensor<float>(mvaInputs, m_input_shapes, m_ortMemInfo));
 
     // pass data through model
     try
     {
-      std::vector<Ort::Value> output_tensors = m_ortSession->Run(m_input_names,
-                                                                 input_tensors,
-                                                                 m_output_names,
-                                                                 Ort::RunOptions{nullptr});
+      auto output_tensors = m_ortSession->Run(Ort::RunOptions{nullptr},
+                                              m_input_names.data(),
+                                              input_tensors.data(),
+                                              input_tensors.size(),
+                                              m_output_names.data(),
+                                              m_output_names.size());
 
       // double-check the dimensions of the output tensors
       // NOTE: the number of output tensors is equal to the number of output nodes specified in the Run() call

RecFCCeeCalorimeter/src/components/PhotonIDTool.h

@@ -18,7 +18,7 @@ namespace edm4hep {
 }
 
 // ONNX
-#include "onnxruntime/core/session/experimental_onnxruntime_cxx_api.h"
+#include "onnxruntime_cxx_api.h"
 
 /** @class PhotonIDTool
  *
@@ -100,12 +100,13 @@ class PhotonIDTool : public Gaudi::Algorithm {
 
   // the ONNX runtime session for running the inference,
   // the environment, and the input and output shapes and names
-  Ort::Experimental::Session* m_ortSession = nullptr;
+  Ort::Session* m_ortSession = nullptr;
   Ort::Env* m_ortEnv = nullptr;
+  Ort::MemoryInfo m_ortMemInfo;
   std::vector<std::int64_t> m_input_shapes;
   std::vector<std::int64_t> m_output_shapes;
-  std::vector<std::string> m_input_names;
-  std::vector<std::string> m_output_names;
+  std::vector<const char*> m_input_names;
+  std::vector<const char*> m_output_names;
   std::vector<std::string> m_internal_input_names;
 
   // the indices of the shapeParameters containing the inputs to the model (-1 if not found)

cmake/FindONNXRuntime.cmake

@@ -1,4 +1,6 @@
-find_path(ONNXRUNTIME_INCLUDE_DIR onnxruntime/core/session/onnxruntime_cxx_inline.h
+find_path(ONNXRUNTIME_INCLUDE_DIR
+          NAMES onnxruntime_cxx_api.h
+          PATH_SUFFIXES onnxruntime/core/session
           HINTS $ENV{ONNXRUNTIME_ROOT_DIR}/include ${ONNXRUNTIME_ROOT_DIR}/include)
 
 find_library(ONNXRUNTIME_LIBRARY NAMES onnxruntime
@@ -11,4 +13,13 @@ mark_as_advanced(ONNXRUNTIME_FOUND ONNXRUNTIME_INCLUDE_DIR ONNXRUNTIME_LIBRARY)
 
 set(ONNXRUNTIME_INCLUDE_DIRS ${ONNXRUNTIME_INCLUDE_DIR})
 set(ONNXRUNTIME_LIBRARIES ${ONNXRUNTIME_LIBRARY})
-get_filename_component(ONNXRUNTIME_LIBRARY_DIRS ${ONNXRUNTIME_LIBRARY} PATH)
+
+# Rig an onnxruntime::onnxruntime target that works similar (enough) to the one
+# that can be directly found via find_package(onnxruntime) for newer versions of
+# onnxruntime
+add_library(onnxruntime::onnxruntime INTERFACE IMPORTED GLOBAL)
+set_target_properties(onnxruntime::onnxruntime
+  PROPERTIES
+  INTERFACE_INCLUDE_DIRECTORIES "${ONNXRUNTIME_INCLUDE_DIRS}"
+  INTERFACE_LINK_LIBRARIES "${ONNXRUNTIME_LIBRARIES}"
+)