Restore TF and MXNet-based inference.

And enable ONNXRuntime for x86/arm only.

PhysicsTools/ONNXRuntime/interface/ONNXRuntime.h

@@ -16,7 +16,18 @@
 #include <string>
 #include <memory>
 
+// currently ONNXRUNTIME only supports x86 and ARM
+#if defined(__arm__) || defined(__aarch64__) || defined(__x86_64__) || defined(__i386__)
+#define CMS_USE_ONNXRUNTIME
+#endif
+
+#ifdef CMS_USE_ONNXRUNTIME
 #include "onnxruntime/core/session/onnxruntime_cxx_api.h"
+#else
+namespace Ort {
+  struct SessionOptions {};
+}  // namespace Ort
+#endif
 
 namespace cms::Ort {
 
@@ -48,6 +59,7 @@ namespace cms::Ort {
     // The 0th dim depends on the batch size, therefore is set to -1
     const std::vector<int64_t>& getOutputShape(const std::string& output_name) const;
 
+#ifdef CMS_USE_ONNXRUNTIME
   private:
     static const ::Ort::Env env_;
     std::unique_ptr<::Ort::Session> session_;
@@ -59,6 +71,7 @@ namespace cms::Ort {
     std::vector<std::string> output_node_strings_;
     std::vector<const char*> output_node_names_;
     std::map<std::string, std::vector<int64_t>> output_node_dims_;
+#endif
   };
 
 }  // namespace cms::Ort

PhysicsTools/ONNXRuntime/src/ONNXRuntime.cc

@@ -19,9 +19,12 @@ namespace cms::Ort {
 
   using namespace ::Ort;
 
+#ifdef CMS_USE_ONNXRUNTIME
   const Env ONNXRuntime::env_(ORT_LOGGING_LEVEL_WARNING, "");
+#endif
 
   ONNXRuntime::ONNXRuntime(const std::string& model_path, const SessionOptions* session_options) {
+#ifdef CMS_USE_ONNXRUNTIME
     // create session
     if (session_options) {
       session_.reset(new Session(env_, model_path.c_str(), *session_options));
@@ -76,6 +79,7 @@ namespace cms::Ort {
       // the 0th dim depends on the batch size
       output_node_dims_[output_name].at(0) = -1;
     }
+#endif
   }
 
   ONNXRuntime::~ONNXRuntime() {}
@@ -84,6 +88,7 @@ namespace cms::Ort {
                                FloatArrays& input_values,
                                const std::vector<std::string>& output_names,
                                int64_t batch_size) const {
+#ifdef CMS_USE_ONNXRUNTIME
     assert(input_names.size() == input_values.size());
     assert(batch_size > 0);
 
@@ -142,23 +147,34 @@ namespace cms::Ort {
     assert(outputs.size() == run_output_node_names.size());
 
     return outputs;
+#else
+    throw cms::Exception("RuntimeError") << "ONNXRuntime does not support the current architecture";
+#endif
   }
 
   const std::vector<std::string>& ONNXRuntime::getOutputNames() const {
+#ifdef CMS_USE_ONNXRUNTIME
     if (session_) {
       return output_node_strings_;
     } else {
-      throw cms::Exception("RuntimeError") << "Needs to call createSession() first before getting the output names!";
+      throw cms::Exception("RuntimeError") << "ONNXRuntime session is not initialized!";
     }
+#else
+    throw cms::Exception("RuntimeError") << "ONNXRuntime does not support the current architecture";
+#endif
   }
 
   const std::vector<int64_t>& ONNXRuntime::getOutputShape(const std::string& output_name) const {
+#ifdef CMS_USE_ONNXRUNTIME
     auto iter = output_node_dims_.find(output_name);
     if (iter == output_node_dims_.end()) {
       throw cms::Exception("RuntimeError") << "Output name " << output_name << " is invalid!";
     } else {
       return iter->second;
     }
+#else
+    throw cms::Exception("RuntimeError") << "ONNXRuntime does not support the current architecture";
+#endif
   }
 
 } /* namespace cms::Ort */

PhysicsTools/ONNXRuntime/test/testONNXRuntime.cc

@@ -2,6 +2,7 @@
 
 #include "PhysicsTools/ONNXRuntime/interface/ONNXRuntime.h"
 #include "FWCore/ParameterSet/interface/FileInPath.h"
+#include "FWCore/Utilities/interface/Exception.h"
 
 #include <chrono>
 #include <iostream>
@@ -27,11 +28,15 @@ void testONNXRuntime::checkAll() {
         std::vector<float>(batch_size * 2, 1),
     };
     FloatArrays outputs;
+#ifdef CMS_USE_ONNXRUNTIME
     CPPUNIT_ASSERT_NO_THROW(outputs = rt.run({"X"}, input_values, {"Y"}, batch_size));
     CPPUNIT_ASSERT(outputs.size() == 1);
     CPPUNIT_ASSERT(outputs[0].size() == batch_size);
     for (const auto &v : outputs[0]) {
       CPPUNIT_ASSERT(v == 3);
     }
+#else
+    CPPUNIT_ASSERT_THROW(rt.run({"X"}, input_values, {"Y"}, batch_size), cms::Exception);
+#endif
   }
 }

PhysicsTools/PatAlgos/python/tools/jetTools.py

@@ -281,6 +281,7 @@ def setupBTagging(process, jetSource, pfCandidates, explicitJTA, pvSource, svSou
             process.load("RecoBTag.CTagging.cTagging_EventSetup_cff")
     import RecoBTag.Configuration.RecoBTag_cff as btag
     import RecoJets.JetProducers.caTopTaggers_cff as toptag
+    from RecoBTag.ONNXRuntime.SwitchProducerONNX import SwitchProducerONNX
 
     if tightBTagNTkHits:
         if not runIVF:
@@ -720,6 +721,15 @@ def setupBTagging(process, jetSource, pfCandidates, explicitJTA, pvSource, svSou
                     process,
                     task
                 )
+            elif isinstance(getattr(btag, btagDiscr), SwitchProducerONNX):
+                addToProcessAndTask(
+                    newDiscr,
+                    getattr(btag, btagDiscr).cloneAll(
+                        src = btagPrefix + supportedBtagDiscr[discriminator_name][0][0] + labelName + postfix
+                    ),
+                    process,
+                    task
+                )
             else:
                 raise ValueError('I do not know how to update %s it does not have neither "tagInfos" nor "src" attributes' % btagDiscr)
             acceptedBtagDiscriminators.append(discriminator_name)

PhysicsTools/TensorFlow/interface/TensorFlow.h

@@ -87,13 +87,13 @@ namespace tensorflow {
   // return a new session that will contain an already loaded graph def, sessionOptions are predefined
   // an error is thrown when graphDef is a nullptr or when the grah has no nodes
   // transfers ownership
-  Session* createSession(GraphDef* graphDef, SessionOptions& sessionOptions);
+  Session* createSession(const GraphDef* graphDef, SessionOptions& sessionOptions);
 
   // return a new session that will contain an already loaded graph def, threading options are
   // inferred from nThreads
   // an error is thrown when graphDef is a nullptr or when the grah has no nodes
   // transfers ownership
-  Session* createSession(GraphDef* graphDef, int nThreads = 1);
+  Session* createSession(const GraphDef* graphDef, int nThreads = 1);
 
   // closes a session, calls its destructor, resets the pointer, and returns true on success
   bool closeSession(Session*& session);

PhysicsTools/TensorFlow/src/TensorFlow.cc

@@ -159,7 +159,7 @@ namespace tensorflow {
     return createSession(metaGraphDef, exportDir, sessionOptions);
   }
 
-  Session* createSession(GraphDef* graphDef, SessionOptions& sessionOptions) {
+  Session* createSession(const GraphDef* graphDef, SessionOptions& sessionOptions) {
     // check for valid pointer
     if (graphDef == nullptr) {
       throw cms::Exception("InvalidGraphDef") << "error while creating session: graphDef is nullptr";
@@ -185,7 +185,7 @@ namespace tensorflow {
     return session;
   }
 
-  Session* createSession(GraphDef* graphDef, int nThreads) {
+  Session* createSession(const GraphDef* graphDef, int nThreads) {
     // create session options and set thread options
     SessionOptions sessionOptions;
     setThreading(sessionOptions, nThreads);

RecoBTag/FeatureTools/interface/tensor_fillers.h

@@ -0,0 +1,48 @@
+#ifndef RecoBTag_FeatureTools_tensor_fillers_h
+#define RecoBTag_FeatureTools_tensor_fillers_h
+
+#include "DataFormats/BTauReco/interface/DeepFlavourTagInfo.h"
+#include "DataFormats/BTauReco/interface/DeepDoubleXTagInfo.h"
+
+namespace btagbtvdeep {
+
+  void jet_tensor_filler(float* ptr, const btagbtvdeep::DeepFlavourFeatures& features, unsigned feature_dims);
+
+  void jet4vec_tensor_filler(float* ptr, const btagbtvdeep::DeepFlavourFeatures& features, unsigned feature_dims);
+
+  void db_tensor_filler(float* ptr, const btagbtvdeep::DeepDoubleXFeatures& features, unsigned feature_dims);
+
+  void c_pf_tensor_filler(float* ptr,
+                          std::size_t max_c_pf_n,
+                          const std::vector<btagbtvdeep::ChargedCandidateFeatures>& c_pf_features_vec,
+                          unsigned feature_dims);
+
+  void c_pf_reduced_tensor_filler(float* ptr,
+                                  std::size_t max_c_pf_n,
+                                  const std::vector<btagbtvdeep::ChargedCandidateFeatures>& c_pf_features_vec,
+                                  unsigned feature_dims);
+
+  void n_pf_tensor_filler(float* ptr,
+                          std::size_t max_n_pf_n,
+                          const std::vector<btagbtvdeep::NeutralCandidateFeatures>& n_pf_features_vec,
+                          unsigned feature_dims);
+
+  void sv_tensor_filler(float* ptr,
+                        std::size_t max_sv_n,
+                        const std::vector<btagbtvdeep::SecondaryVertexFeatures>& sv_features_vec,
+                        unsigned feature_dims);
+
+  void sv_reduced_tensor_filler(float* ptr,
+                                std::size_t max_sv_n,
+                                const std::vector<btagbtvdeep::SecondaryVertexFeatures>& sv_features_vec,
+                                unsigned feature_dims);
+
+  void seed_tensor_filler(float* ptr, const btagbtvdeep::SeedingTrackFeatures& seed_features, unsigned feature_dims);
+
+  void neighbourTracks_tensor_filler(float* ptr,
+                                     const btagbtvdeep::SeedingTrackFeatures& seed_features,
+                                     unsigned feature_dims);
+
+}  // namespace btagbtvdeep
+
+#endif