From c203d89958b4b0b23ce6967decc48afdd06b7ddb Mon Sep 17 00:00:00 2001
From: Justin Chu <justinchuby@users.noreply.github.com>
Date: Wed, 24 Jul 2024 11:50:11 -0700
Subject: [PATCH] Update ruff and clang-format versions (#21479)

ruff -> 0.5.4
clang-format -> 18
---
 cgmanifests/generate_cgmanifest.py            |    2 +-
 .../tools/ValidateNativeDelegateAttributes.py |    2 +-
 include/onnxruntime/core/common/exceptions.h  |    6 +-
 .../core/framework/stream_handles.h           |    2 +-
 include/onnxruntime/core/platform/Barrier.h   |    2 +-
 .../platform/EigenNonBlockingThreadPool.h     |   10 +-
 .../core/providers/custom_op_context.h        |    2 +-
 .../experimental_onnxruntime_cxx_api.h        |    6 +-
 .../core/session/onnxruntime_cxx_api.h        |    4 +-
 .../core/session/onnxruntime_lite_custom_op.h |    2 +-
 .../contrib_ops/cpu/cpu_contrib_kernels.cc    |  134 +-
 .../contrib_ops/cpu/crop_and_resize.cc        |    2 +-
 .../cuda/bert/flash_attention/alibi.h         |    2 +-
 .../cuda/bert/flash_attention/mask.h          |    2 +-
 .../cuda/bert/flash_attention/softmax.h       |    2 +-
 .../contrib_ops/cuda/cuda_contrib_kernels.cc  |  364 +-
 onnxruntime/core/framework/ex_lib_loader.h    |    2 +-
 .../core/graph/contrib_ops/contrib_defs.cc    |    4 +-
 .../transpose_optimization/optimizer_api.h    |    8 +-
 onnxruntime/core/platform/path_lib.h          |    6 +-
 .../core/providers/coreml/model/model.mm      |    2 +-
 .../providers/cpu/cpu_execution_provider.cc   | 3092 ++++++++---------
 .../cpu/ml/tree_ensemble_classifier.cc        |   18 +-
 .../core/providers/cpu/ml/treeregressor.cc    |   18 +-
 .../object_detection/non_max_suppression.cc   |    4 +-
 .../cpu/object_detection/roialign.cc          |    6 +-
 .../core/providers/cpu/tensor/expand.cc       |   10 +-
 .../providers/cuda/cuda_execution_provider.cc | 1788 +++++-----
 onnxruntime/core/providers/cuda/cuda_graph.h  |    4 +-
 .../core/providers/cuda/cuda_profiler.h       |    2 +-
 .../core/providers/cuda/nn/conv_transpose.h   |    2 +-
 .../core/providers/cuda/nvtx_profile.h        |    8 +-
 .../providers/cuda/shared_inc/cuda_utils.h    |    2 +-
 .../core/providers/cuda/tensor/cast_op.cc     |   32 +-
 .../providers/dnnl/dnnl_node_capability.h     |    2 +-
 .../providers/dnnl/subgraph/dnnl_subgraph.h   |    2 +-
 onnxruntime/core/providers/js/allocator.h     |    2 +-
 onnxruntime/core/providers/js/data_transfer.h |    4 +-
 .../builder/opbuilder/expand_op_builder.cc    |    2 +-
 .../qnn/builder/opbuilder/pad_op_builder.cc   |    2 +-
 .../qnn/builder/qnn_quant_params_wrapper.cc   |    2 +-
 .../core/providers/rocm/rocm_profiler.h       |    2 +-
 .../shared_library/provider_host_api.h        |    4 +-
 .../tensorrt_execution_provider_custom_ops.h  |    4 +-
 .../vitisai/vitisai_provider_factory.cc       |    2 +-
 .../builders/impl/elementwise_op_builder.h    |    2 +-
 .../vsinpu/builders/op_builder_factory.h      |    7 +-
 .../python/onnxruntime_pybind_iobinding.cc    |    9 +-
 .../python/onnxruntime_pybind_ortvalue.cc     |  136 +-
 .../onnxruntime_pybind_sparse_tensor.cc       |    3 +-
 .../python/onnxruntime_pybind_state.cc        |   88 +-
 onnxruntime/python/onnxruntime_validation.py  |   17 +-
 .../tools/pytorch_export_contrib_ops.py       |    2 +-
 .../python/tools/quantization/calibrate.py    |    2 +-
 .../python/tools/symbolic_shape_infer.py      |   44 +-
 .../python/tools/tensorrt/perf/benchmark.py   |   12 +-
 .../python/tools/tensorrt/perf/perf_utils.py  |    2 +-
 .../perf/setup_scripts/setup_onnx_zoo.py      |    2 +-
 .../python/tools/transformers/benchmark.py    |    2 +-
 .../tools/transformers/bert_test_data.py      |    6 +-
 .../tools/transformers/fusion_attention.py    |    2 +-
 .../python/tools/transformers/fusion_utils.py |    2 +-
 .../bart/utils/export_summarization_edinit.py |    2 +-
 .../export_summarization_enc_dec_past.py      |    2 +-
 .../models/bart/utils/onnx_inference.py       |    4 +-
 .../models/stable_diffusion/engine_builder.py |    2 +-
 .../pipeline_stable_diffusion.py              |    4 +-
 .../test/framework/allocation_planner_test.cc |    2 +-
 onnxruntime/test/onnx/OrtValueList.h          |    2 +-
 .../test/onnx/microbenchmark/activation.cc    |    2 +-
 .../qdq_transformer_fastmath_test.cc          |    2 +-
 .../test/optimizer/qdq_transformer_test.cc    |    2 +-
 .../reduction_test_cases_generator.py         |    8 +-
 .../test/providers/cpu/tensor/pad_test.cc     |   12 +-
 .../test/providers/qnn/qnn_basic_test.cc      |   10 +-
 .../test/python/onnx_backend_test_series.py   |    2 +-
 .../test/python/transformers/rotary_flash.py  |    3 -
 .../generate_tiny_keras2onnx_bert_models.py   |    4 +-
 .../generate_tiny_gpt2_model.py               |    4 +-
 onnxruntime/test/shared_lib/custom_op_utils.h |   20 +-
 onnxruntime/test/testdata/CNTK/gen.py         |    4 +-
 .../core/framework/adasum/adasum_mpi.cc       |    3 +-
 .../orttraining/core/framework/pipeline.h     |    2 +-
 .../torch/custom_function_register.h          |    2 +-
 .../orttraining/core/framework/torch/gil.h    |    2 +-
 .../core/framework/torch/torch_proxy.h        |    4 +-
 .../orttraining/core/graph/graph_augmenter.h  |    6 +-
 .../core/graph/loss_func/loss_func_common.h   |    2 +-
 .../core/graph/pipeline_transformer.cc        |    2 +-
 .../core/optimizer/megatron_transformer.cc    |    4 +-
 .../core/session/training_session.h           |    6 +-
 orttraining/orttraining/lazy_tensor/flags.h   |    2 +-
 orttraining/orttraining/models/bert/main.cc   |    3 +-
 .../orttraining/models/pipeline_poc/main.cc   |   54 +-
 .../orttraining/models/runner/training_util.h |    4 +-
 .../python/orttraining_pybind_state.cc        |   24 +-
 .../python/training/ort_triton/kernel/_mm.py  |    2 +-
 .../python/training/ortmodule/_utils.py       |    2 +-
 .../cpu/torch_interop_utils/ctx_pool.h        |    4 +-
 .../test/distributed/partition_utils.h        |    2 +-
 ...orttraining_test_hierarchical_ortmodule.py |    2 +-
 .../orttraining_test_model_transform.py       |    2 +-
 .../python/orttraining_test_ortmodule_api.py  |   12 +-
 ...training_test_ortmodule_bert_classifier.py |    2 +-
 ...test_ortmodule_bert_classifier_autocast.py |    2 +-
 ...g_test_ortmodule_deepspeed_zero_stage_1.py |    2 +-
 .../orttraining_test_ortmodule_onnx_ops.py    |   12 +-
 .../python/orttraining_test_ortmodule_poc.py  |    2 +-
 .../test/python/orttraining_test_utilities.py |    4 +-
 .../training_ops/function_op_test_utils.cc    |    2 +-
 .../cpu/torch/torch_custom_function_kernel.h  |    2 +-
 .../cuda/cuda_training_kernels.cc             |  466 +--
 .../rocm/rocm_training_kernels.cc             |  374 +-
 .../tools/scripts/gpt2_model_transform.py     |    2 +-
 orttraining/tools/scripts/model_transform.py  |    2 +-
 pyproject.toml                                |    1 +
 requirements-lintrunner.txt                   |    8 +-
 tools/ci_build/build.py                       |   22 +-
 tools/ci_build/gen_def.py                     |   10 +-
 tools/ci_build/reduce_op_kernels.py           |    2 +-
 tools/ci_build/replace_urls_in_deps.py        |    6 +-
 .../upload_python_package_to_azure_storage.py |    2 +-
 tools/doc/rename_folders.py                   |   14 +-
 .../nuget/generate_nuspec_for_native_nuget.py |    6 +-
 tools/python/onnx_test_data_utils.py          |    2 +-
 .../util/mobile_helpers/usability_checker.py  |    2 +-
 .../util/reduced_build_config_parser.py       |    2 +-
 winml/lib/Api.Image/CpuDetensorizer.h         |    9 +-
 winml/lib/Api.Image/CpuTensorizer.h           |   12 +-
 winml/lib/Api.Image/D3DDeviceCache.cpp        |   24 +-
 winml/lib/Api.Image/EventTimer.h              |    4 +-
 .../lib/Api.Image/ImageConversionHelpers.cpp  |   11 +-
 winml/lib/Api.Image/ImageConverter.cpp        |    3 +-
 .../Api.Image/TensorToVideoFrameConverter.cpp |   13 +-
 .../Api.Image/VideoFrameToTensorConverter.cpp |   38 +-
 .../Api.Image/inc/ConverterResourceStore.h    |    2 +-
 winml/lib/Api/FeatureValues.h                 |   82 +-
 winml/lib/Api/ImageFeatureValue.cpp           |   14 +-
 winml/lib/Api/LearningModel.cpp               |    2 +-
 winml/lib/Api/LearningModelSession.cpp        |    4 +-
 winml/lib/Api/NumericData.cpp                 |   12 +-
 winml/lib/Api/impl/FeatureCompatibility.h     |    6 +-
 winml/lib/Common/CommonDeviceHelpers.cpp      |    6 +-
 ...er_backed_random_access_stream_reference.h |    5 +-
 winml/test/api/raw/winml_microsoft.h          |  108 +-
 winml/test/api/raw/winml_windows.h            |  112 +-
 winml/test/image/imagetests.cpp               |   21 +-
 winml/test/model/compare_feature_value.cpp    |    3 +-
 winml/test/model/model_tests.cpp              |   10 +-
 winml/test/model/skip_model_tests.h           |    6 +-
 winml/test/scenario/cppwinrt/NoisyReluCpu.h   |    6 +-
 winml/test/scenario/cppwinrt/ReluCpu.h        |    6 +-
 152 files changed, 3781 insertions(+), 3842 deletions(-)

diff --git a/cgmanifests/generate_cgmanifest.py b/cgmanifests/generate_cgmanifest.py
index 3cecbb0cc977f..52bd3f58645f2 100644
--- a/cgmanifests/generate_cgmanifest.py
+++ b/cgmanifests/generate_cgmanifest.py
@@ -73,7 +73,7 @@ def add_github_dep(name, parsed_url):
             return
         # Make a REST call to convert to tag to a git commit
         url = f"https://api.github.com/repos/{org_name}/{repo_name}/git/refs/tags/{tag}"
-        print("requesting %s ..." % url)
+        print("requesting {url} ...")
         res = requests.get(url, auth=(args.username, args.token))
         response_json = res.json()
         tag_object = response_json["object"]
diff --git a/csharp/tools/ValidateNativeDelegateAttributes.py b/csharp/tools/ValidateNativeDelegateAttributes.py
index acd6c173bfeb0..7431cc8d9d288 100644
--- a/csharp/tools/ValidateNativeDelegateAttributes.py
+++ b/csharp/tools/ValidateNativeDelegateAttributes.py
@@ -19,7 +19,7 @@ def check_all_delegates_have_unmanaged_function_pointer_attribute(file: pathlib.
     line_num = 0
     with open(str(file.resolve(strict=True))) as f:
         prev_line = ""
-        for line in f.readlines():
+        for line in f:
             line_num += 1
 
             # strip so it's easier to deal with commented out lines.
diff --git a/include/onnxruntime/core/common/exceptions.h b/include/onnxruntime/core/common/exceptions.h
index 18c117f12ad7d..494a770b8db98 100644
--- a/include/onnxruntime/core/common/exceptions.h
+++ b/include/onnxruntime/core/common/exceptions.h
@@ -17,13 +17,13 @@ namespace onnxruntime {
 
 class NotImplementedException : public std::logic_error {
  public:
-  explicit NotImplementedException(const char* _Message = "Function not yet implemented") noexcept : std::logic_error(_Message){};
-  explicit NotImplementedException(const std::string& _Message = "Function not yet implemented") noexcept : std::logic_error(_Message){};
+  explicit NotImplementedException(const char* _Message = "Function not yet implemented") noexcept : std::logic_error(_Message) {};
+  explicit NotImplementedException(const std::string& _Message = "Function not yet implemented") noexcept : std::logic_error(_Message) {};
 };
 
 class TypeMismatchException : public std::logic_error {
  public:
-  TypeMismatchException() noexcept : logic_error("Type mismatch"){};
+  TypeMismatchException() noexcept : logic_error("Type mismatch") {};
 };
 
 class OnnxRuntimeException : public std::exception {
diff --git a/include/onnxruntime/core/framework/stream_handles.h b/include/onnxruntime/core/framework/stream_handles.h
index 9c987f10ccadb..01631e1fb2aa6 100644
--- a/include/onnxruntime/core/framework/stream_handles.h
+++ b/include/onnxruntime/core/framework/stream_handles.h
@@ -32,7 +32,7 @@ class Stream {
     return {};
   };
   // block the host thread until all the tasks in the stream finished.
-  virtual void Flush(){};
+  virtual void Flush() {};
   // The framework may reuse the stream instance for multiple iterations.
   // This is the API that provide a chance to let the device stream cleanup
   // resource at the end of a iteration.
diff --git a/include/onnxruntime/core/platform/Barrier.h b/include/onnxruntime/core/platform/Barrier.h
index 915cfc50953ed..1148b052bd9af 100644
--- a/include/onnxruntime/core/platform/Barrier.h
+++ b/include/onnxruntime/core/platform/Barrier.h
@@ -76,6 +76,6 @@ class Barrier {
 // Multiple threads can wait on the same Notification object,
 // but only one caller must call Notify() on the object.
 struct Notification : Barrier {
-  Notification() : Barrier(1){};
+  Notification() : Barrier(1) {};
 };
 }  // namespace onnxruntime
diff --git a/include/onnxruntime/core/platform/EigenNonBlockingThreadPool.h b/include/onnxruntime/core/platform/EigenNonBlockingThreadPool.h
index e33007102e198..d4411a6d72356 100644
--- a/include/onnxruntime/core/platform/EigenNonBlockingThreadPool.h
+++ b/include/onnxruntime/core/platform/EigenNonBlockingThreadPool.h
@@ -219,18 +219,18 @@ class ThreadPoolProfiler {
     WAIT_REVOKE,
     MAX_EVENT
   };
-  ThreadPoolProfiler(int, const CHAR_TYPE*){};
+  ThreadPoolProfiler(int, const CHAR_TYPE*) {};
   ~ThreadPoolProfiler() = default;
   ORT_DISALLOW_COPY_ASSIGNMENT_AND_MOVE(ThreadPoolProfiler);
-  void Start(){};
+  void Start() {};
   std::string Stop() { return "not available for minimal build"; }
-  void LogStart(){};
+  void LogStart() {};
   void LogEnd(ThreadPoolEvent){};
   void LogEndAndStart(ThreadPoolEvent){};
   void LogStartAndCoreAndBlock(std::ptrdiff_t){};
   void LogCoreAndBlock(std::ptrdiff_t){};
-  void LogThreadId(int){};
-  void LogRun(int){};
+  void LogThreadId(int) {};
+  void LogRun(int) {};
   std::string DumpChildThreadStat() { return {}; }
 };
 #else
diff --git a/include/onnxruntime/core/providers/custom_op_context.h b/include/onnxruntime/core/providers/custom_op_context.h
index 8f3d2476d4fdb..b10126da8e0fb 100644
--- a/include/onnxruntime/core/providers/custom_op_context.h
+++ b/include/onnxruntime/core/providers/custom_op_context.h
@@ -6,5 +6,5 @@
 // CustomOpContext defines an interface allowing a custom op to access ep-specific resources.
 struct CustomOpContext {
   CustomOpContext() = default;
-  virtual ~CustomOpContext(){};
+  virtual ~CustomOpContext() {};
 };
\ No newline at end of file
diff --git a/include/onnxruntime/core/session/experimental_onnxruntime_cxx_api.h b/include/onnxruntime/core/session/experimental_onnxruntime_cxx_api.h
index 9e4ceffc44bfd..c1a7839ff22fa 100644
--- a/include/onnxruntime/core/session/experimental_onnxruntime_cxx_api.h
+++ b/include/onnxruntime/core/session/experimental_onnxruntime_cxx_api.h
@@ -24,9 +24,9 @@ namespace Experimental {
 
 struct Session : Ort::Session {
   Session(Env& env, std::basic_string<ORTCHAR_T>& model_path, SessionOptions& options)
-      : Ort::Session(env, model_path.data(), options){};
+      : Ort::Session(env, model_path.data(), options) {};
   Session(Env& env, void* model_data, size_t model_data_length, SessionOptions& options)
-      : Ort::Session(env, model_data, model_data_length, options){};
+      : Ort::Session(env, model_data, model_data_length, options) {};
 
   // overloaded Run() with sensible defaults
   std::vector<Ort::Value> Run(const std::vector<std::string>& input_names,
@@ -52,7 +52,7 @@ struct Session : Ort::Session {
 
 struct Value : Ort::Value {
   Value(OrtValue* p)
-      : Ort::Value(p){};
+      : Ort::Value(p) {};
 
   template <typename T>
   static Ort::Value CreateTensor(T* p_data, size_t p_data_element_count, const std::vector<int64_t>& shape);
diff --git a/include/onnxruntime/core/session/onnxruntime_cxx_api.h b/include/onnxruntime/core/session/onnxruntime_cxx_api.h
index 8091fd4cfc2a3..5d974e1ff5185 100644
--- a/include/onnxruntime/core/session/onnxruntime_cxx_api.h
+++ b/include/onnxruntime/core/session/onnxruntime_cxx_api.h
@@ -2175,8 +2175,8 @@ struct Op : detail::Base<OrtOp> {
 /// </summary>
 struct ShapeInferContext {
   struct SymbolicInteger {
-    SymbolicInteger(int64_t i) : i_(i), is_int_(true){};
-    SymbolicInteger(const char* s) : s_(s), is_int_(false){};
+    SymbolicInteger(int64_t i) : i_(i), is_int_(true) {};
+    SymbolicInteger(const char* s) : s_(s), is_int_(false) {};
     SymbolicInteger(const SymbolicInteger&) = default;
     SymbolicInteger(SymbolicInteger&&) = default;
 
diff --git a/include/onnxruntime/core/session/onnxruntime_lite_custom_op.h b/include/onnxruntime/core/session/onnxruntime_lite_custom_op.h
index 57a64380faeb0..ce87d8c56d3fe 100644
--- a/include/onnxruntime/core/session/onnxruntime_lite_custom_op.h
+++ b/include/onnxruntime/core/session/onnxruntime_lite_custom_op.h
@@ -29,7 +29,7 @@ class ArgBase {
   ArgBase(OrtKernelContext* ctx,
           size_t indice,
           bool is_input) : ctx_(ctx), indice_(indice), is_input_(is_input) {}
-  virtual ~ArgBase(){};
+  virtual ~ArgBase() {};
 
  protected:
   struct KernelContext ctx_;
diff --git a/onnxruntime/contrib_ops/cpu/cpu_contrib_kernels.cc b/onnxruntime/contrib_ops/cpu/cpu_contrib_kernels.cc
index 90a51fda0b188..84f9ca88ecf55 100644
--- a/onnxruntime/contrib_ops/cpu/cpu_contrib_kernels.cc
+++ b/onnxruntime/contrib_ops/cpu/cpu_contrib_kernels.cc
@@ -267,83 +267,83 @@ Status RegisterQuantizationKernels(KernelRegistry& kernel_registry) {
 
 Status RegisterCpuContribKernels(KernelRegistry& kernel_registry) {
   static const BuildKernelCreateInfoFn function_table[] = {
-    BuildKernelCreateInfo<void>,  // default entry to avoid the list become empty after ops-reducing
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, float, SampleOp)>,
+      BuildKernelCreateInfo<void>,  // default entry to avoid the list become empty after ops-reducing
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, float, SampleOp)>,
 
-    // add more kernels here
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, float, GridSample)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, float, Attention)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, float, BeamSearch)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, float, WhisperBeamSearch)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, float, EmbedLayerNormalization)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, float, ExpandDims)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, float, FusedConv)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, float, FusedGemm)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, float, GreedySearch)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, float, MultiHeadAttention)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, float, GroupQueryAttention)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, float, SparseAttention)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, float, RotaryEmbedding)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, float, Sampling)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, AttnLSTM)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, string, Tokenizer)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, Range)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, WordConvEmbedding)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, GatherND)>,
+      // add more kernels here
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, float, GridSample)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, float, Attention)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, float, BeamSearch)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, float, WhisperBeamSearch)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, float, EmbedLayerNormalization)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, float, ExpandDims)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, float, FusedConv)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, float, FusedGemm)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, float, GreedySearch)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, float, MultiHeadAttention)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, float, GroupQueryAttention)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, float, SparseAttention)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, float, RotaryEmbedding)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, float, Sampling)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, AttnLSTM)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, string, Tokenizer)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, Range)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, WordConvEmbedding)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, GatherND)>,
 #if !defined(DISABLE_SPARSE_TENSORS)
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, SparseToDenseMatMul)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, SparseToDenseMatMul)>,
 #endif
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, MurmurHash3)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, TransposeMatMul)>,  // backward compatibility
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, FusedMatMul)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, MatMulNBits)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, MatMulBnb4)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, MurmurHash3)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, TransposeMatMul)>,  // backward compatibility
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, FusedMatMul)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, MatMulNBits)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, MatMulBnb4)>,
 #ifndef ORT_MINIMAL_BUILD
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, MatMulFpQ4)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, MatMulFpQ4)>,
 #endif
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, float, MaxpoolWithMask)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, Pad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, Unique)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, float, ConvTransposeWithDynamicPads)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, float, CropAndResize)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, float, CDist)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, double, CDist)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, BiasGelu)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, Gelu)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, FastGelu)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, NGramRepeatBlock)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, BifurcationDetector)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, QuickGelu)>,
-    // These ops were experimental ops in onnx domain which have been removed now. We add them here as
-    // contrib ops to main backward compatibility
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, Affine)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, Crop)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, DynamicSlice)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, ImageScaler)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 8, MeanVarianceNormalization)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, ParametricSoftplus)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, ScaledTanh)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 9, ThresholdedRelu)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, Scale)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 16, float, LayerNormalization)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 16, double, LayerNormalization)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, float, SimplifiedLayerNormalization)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, double, SimplifiedLayerNormalization)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, float, SkipLayerNormalization)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, double, SkipLayerNormalization)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, float, SkipSimplifiedLayerNormalization)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, double, SkipSimplifiedLayerNormalization)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, Inverse)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, Trilu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, float, MaxpoolWithMask)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, Pad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, Unique)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, float, ConvTransposeWithDynamicPads)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, float, CropAndResize)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, float, CDist)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, double, CDist)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, BiasGelu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, Gelu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, FastGelu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, NGramRepeatBlock)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, BifurcationDetector)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, QuickGelu)>,
+      // These ops were experimental ops in onnx domain which have been removed now. We add them here as
+      // contrib ops to main backward compatibility
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, Affine)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, Crop)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, DynamicSlice)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, ImageScaler)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 8, MeanVarianceNormalization)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, ParametricSoftplus)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, ScaledTanh)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 9, ThresholdedRelu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, Scale)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 16, float, LayerNormalization)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 16, double, LayerNormalization)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, float, SimplifiedLayerNormalization)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, double, SimplifiedLayerNormalization)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, float, SkipLayerNormalization)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, double, SkipLayerNormalization)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, float, SkipSimplifiedLayerNormalization)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, double, SkipSimplifiedLayerNormalization)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, Inverse)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, Trilu)>,
 
 #ifdef ENABLE_ATEN
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kPytorchAtenDomain, 1, ATen)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kPytorchAtenDomain, 1, ATen)>,
 #endif
 
 #ifdef ENABLE_TRAINING_OPS
-    // Should remove the shrunken_gather include from ENABLE_TRAINING_OPS once 1). compute optimizer is enabled for inference or
-    // 2). this is needed by inference for other purpose.
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, ShrunkenGather)>,
+      // Should remove the shrunken_gather include from ENABLE_TRAINING_OPS once 1). compute optimizer is enabled for inference or
+      // 2). this is needed by inference for other purpose.
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kMSDomain, 1, ShrunkenGather)>,
 #endif
 
   };
diff --git a/onnxruntime/contrib_ops/cpu/crop_and_resize.cc b/onnxruntime/contrib_ops/cpu/crop_and_resize.cc
index 1863522c1643c..533d62f5e7486 100644
--- a/onnxruntime/contrib_ops/cpu/crop_and_resize.cc
+++ b/onnxruntime/contrib_ops/cpu/crop_and_resize.cc
@@ -173,7 +173,7 @@ void CropAndResizeForward(const TensorShape& output_shape,
               }
             }
           }  // for pw
-        }    // for ph
+        }  // for ph
       },
       0);  // for n
 }
diff --git a/onnxruntime/contrib_ops/cuda/bert/flash_attention/alibi.h b/onnxruntime/contrib_ops/cuda/bert/flash_attention/alibi.h
index 5d94190ecbeb9..18d36cfd88d60 100644
--- a/onnxruntime/contrib_ops/cuda/bert/flash_attention/alibi.h
+++ b/onnxruntime/contrib_ops/cuda/bert/flash_attention/alibi.h
@@ -17,7 +17,7 @@ struct Alibi {
   const int max_seqlen_k, max_seqlen_q;
 
   __forceinline__ __device__ Alibi(const float alibi_slope, const int max_seqlen_k, const int max_seqlen_q)
-      : alibi_slope(alibi_slope), max_seqlen_k(max_seqlen_k), max_seqlen_q(max_seqlen_q){};
+      : alibi_slope(alibi_slope), max_seqlen_k(max_seqlen_k), max_seqlen_q(max_seqlen_q) {};
 
   template <typename Engine, typename Layout>
   __forceinline__ __device__ void apply_alibi(Tensor<Engine, Layout>& tensor,
diff --git a/onnxruntime/contrib_ops/cuda/bert/flash_attention/mask.h b/onnxruntime/contrib_ops/cuda/bert/flash_attention/mask.h
index b225e5e3be559..0998155eba635 100644
--- a/onnxruntime/contrib_ops/cuda/bert/flash_attention/mask.h
+++ b/onnxruntime/contrib_ops/cuda/bert/flash_attention/mask.h
@@ -116,7 +116,7 @@ struct Mask {
   __forceinline__ __device__ Mask(const int max_seqlen_k, const int max_seqlen_q,
                                   const int window_size_left, const int window_size_right,
                                   const float alibi_slope = 0.f)
-      : max_seqlen_k(max_seqlen_k), max_seqlen_q(max_seqlen_q), window_size_left(window_size_left), window_size_right(window_size_right), alibi_slope(!Has_alibi ? 0.0 : alibi_slope){};
+      : max_seqlen_k(max_seqlen_k), max_seqlen_q(max_seqlen_q), window_size_left(window_size_left), window_size_right(window_size_right), alibi_slope(!Has_alibi ? 0.0 : alibi_slope) {};
 
   // Causal_mask: whether this particular iteration needs causal masking
   template <bool Causal_mask = false, bool Is_even_MN = true, typename Engine, typename Layout>
diff --git a/onnxruntime/contrib_ops/cuda/bert/flash_attention/softmax.h b/onnxruntime/contrib_ops/cuda/bert/flash_attention/softmax.h
index 3c205378f0177..ba678b740d376 100644
--- a/onnxruntime/contrib_ops/cuda/bert/flash_attention/softmax.h
+++ b/onnxruntime/contrib_ops/cuda/bert/flash_attention/softmax.h
@@ -121,7 +121,7 @@ struct Softmax {
   using TensorT = decltype(make_tensor<float>(Shape<Int<kNRows>>{}));
   TensorT row_max, row_sum;
 
-  __forceinline__ __device__ Softmax(){};
+  __forceinline__ __device__ Softmax() {};
 
   template <bool Is_first, bool Check_inf = false, typename Tensor0, typename Tensor1>
   __forceinline__ __device__ void softmax_rescale_o(Tensor0& acc_s, Tensor1& acc_o, float softmax_scale_log2) {
diff --git a/onnxruntime/contrib_ops/cuda/cuda_contrib_kernels.cc b/onnxruntime/contrib_ops/cuda/cuda_contrib_kernels.cc
index b237e5c24bbef..21bd5eb91c20f 100644
--- a/onnxruntime/contrib_ops/cuda/cuda_contrib_kernels.cc
+++ b/onnxruntime/contrib_ops/cuda/cuda_contrib_kernels.cc
@@ -231,206 +231,206 @@ KernelCreateInfo BuildKernelCreateInfo<void>() {
 
 Status RegisterCudaContribKernels(KernelRegistry& kernel_registry) {
   static const BuildKernelCreateInfoFn function_table[] = {
-    BuildKernelCreateInfo<void>,  // default entry to avoid the list become empty after ops-reducing
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, GridSample)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, FastGelu)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, FastGelu)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, Gelu)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, double, Gelu)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, Gelu)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_CLASS_NAME(1, BiasGelu)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, BiasSplitGelu)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, BiasSplitGelu)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, BiasAdd)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, BiasAdd)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, QuickGelu)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, double, QuickGelu)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, QuickGelu)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, TransposeMatMul)>,      // backward compatibility
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, double, TransposeMatMul)>,     // backward compatibility
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, TransposeMatMul)>,  // backward compatibility
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, FusedMatMul)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, double, FusedMatMul)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, FusedMatMul)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, RelativePositionBias)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, RelativePositionBias)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, GatedRelativePositionBias)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, GatedRelativePositionBias)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, RemovePadding)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, RemovePadding)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, RestorePadding)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, RestorePadding)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, Rfft)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, double, Rfft)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, Rfft)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, Irfft)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, double, Irfft)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, Irfft)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, ComplexMul)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, ComplexMul)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, ComplexMulConj)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, ComplexMulConj)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_CLASS_NAME(1, NGramRepeatBlock)>,
-
-    // These ops were experimental ops in onnx domain which have been removed now. We add them here as
-    // contrib ops to maintain backward compatibility
-    BuildKernelCreateInfo<CUDA_ONNX_OP_TYPED_CLASS_NAME(1, float, Affine)>,
-    BuildKernelCreateInfo<CUDA_ONNX_OP_TYPED_CLASS_NAME(1, double, Affine)>,
-    BuildKernelCreateInfo<CUDA_ONNX_OP_TYPED_CLASS_NAME(1, MLFloat16, Affine)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, Attention)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, Attention)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, PackedAttention)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, PackedAttention)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, PackedMultiHeadAttention)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, PackedMultiHeadAttention)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_CLASS_NAME(1, BeamSearch)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_CLASS_NAME(1, WhisperBeamSearch)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, ConvTransposeWithDynamicPads)>,
-    BuildKernelCreateInfo<CUDA_ONNX_OP_TYPED_CLASS_NAME(1, float, Crop)>,
-    BuildKernelCreateInfo<CUDA_ONNX_OP_TYPED_CLASS_NAME(1, double, Crop)>,
-    BuildKernelCreateInfo<CUDA_ONNX_OP_TYPED_CLASS_NAME(1, MLFloat16, Crop)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, MoE)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, MoE)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_CLASS_NAME(1, QMoE)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, MultiHeadAttention)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, MultiHeadAttention)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, GroupQueryAttention)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, BFloat16, GroupQueryAttention)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, DecoderAttention)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, DecoderAttention)>,
-    BuildKernelCreateInfo<CUDA_ONNX_OP_TYPED_CLASS_NAME(1, int32_t, DynamicSlice)>,
-    BuildKernelCreateInfo<CUDA_ONNX_OP_TYPED_CLASS_NAME(1, int64_t, DynamicSlice)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, EmbedLayerNormalization)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, EmbedLayerNormalization)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_CLASS_NAME(1, GreedySearch)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_CLASS_NAME(1, GroupNorm)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, NhwcConv)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, NhwcConv)>,
-    BuildKernelCreateInfo<CUDA_ONNX_OP_TYPED_CLASS_NAME(1, float, ImageScaler)>,
-    BuildKernelCreateInfo<CUDA_ONNX_OP_TYPED_CLASS_NAME(1, double, ImageScaler)>,
-    BuildKernelCreateInfo<CUDA_ONNX_OP_TYPED_CLASS_NAME(1, MLFloat16, ImageScaler)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, LongformerAttention)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, LongformerAttention)>,
-    BuildKernelCreateInfo<CUDA_ONNX_OP_TYPED_CLASS_NAME(1, float, ParametricSoftplus)>,
-    BuildKernelCreateInfo<CUDA_ONNX_OP_TYPED_CLASS_NAME(1, double, ParametricSoftplus)>,
-    BuildKernelCreateInfo<CUDA_ONNX_OP_TYPED_CLASS_NAME(1, MLFloat16, ParametricSoftplus)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, RotaryEmbedding)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, RotaryEmbedding)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, BFloat16, RotaryEmbedding)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, GemmaRotaryEmbedding)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_CLASS_NAME(1, Sampling)>,
-    BuildKernelCreateInfo<CUDA_ONNX_OP_TYPED_CLASS_NAME(1, float, ScaledTanh)>,
-    BuildKernelCreateInfo<CUDA_ONNX_OP_TYPED_CLASS_NAME(1, double, ScaledTanh)>,
-    BuildKernelCreateInfo<CUDA_ONNX_OP_TYPED_CLASS_NAME(1, MLFloat16, ScaledTanh)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_CLASS_NAME(1, SkipGroupNorm)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, SkipLayerNormalization)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, SkipLayerNormalization)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, SkipSimplifiedLayerNormalization)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, SkipSimplifiedLayerNormalization)>,
-    BuildKernelCreateInfo<CUDA_ONNX_OP_TYPED_CLASS_NAME(1, float, ThresholdedRelu)>,
-    BuildKernelCreateInfo<CUDA_ONNX_OP_TYPED_CLASS_NAME(1, double, ThresholdedRelu)>,
-    BuildKernelCreateInfo<CUDA_ONNX_OP_TYPED_CLASS_NAME(1, MLFloat16, ThresholdedRelu)>,
-    BuildKernelCreateInfo<CUDA_ONNX_OP_VERSIONED_TYPED_CLASS_NAME(1, 16, float_float_float, LayerNormalization)>,
-    BuildKernelCreateInfo<CUDA_ONNX_OP_VERSIONED_TYPED_CLASS_NAME(1, 16, double_double_double, LayerNormalization)>,
-    BuildKernelCreateInfo<CUDA_ONNX_OP_VERSIONED_TYPED_CLASS_NAME(1, 16, MLFloat16_float_MLFloat16, LayerNormalization)>,
-    BuildKernelCreateInfo<CUDA_ONNX_OP_VERSIONED_TYPED_CLASS_NAME(1, 16, float_float_MLFloat16, LayerNormalization)>,
-    BuildKernelCreateInfo<CUDA_ONNX_OP_VERSIONED_TYPED_CLASS_NAME(1, 16, MLFloat16_float_float, LayerNormalization)>,
-    BuildKernelCreateInfo<CUDA_ONNX_OP_VERSIONED_TYPED_CLASS_NAME(1, 16, BFloat16_float_BFloat16, LayerNormalization)>,
-    BuildKernelCreateInfo<CUDA_ONNX_OP_TYPED_CLASS_NAME(1, float_float_float, SimplifiedLayerNormalization)>,
-    BuildKernelCreateInfo<CUDA_ONNX_OP_TYPED_CLASS_NAME(1, double_double_double, SimplifiedLayerNormalization)>,
-    BuildKernelCreateInfo<CUDA_ONNX_OP_TYPED_CLASS_NAME(1, MLFloat16_float_MLFloat16, SimplifiedLayerNormalization)>,
-    BuildKernelCreateInfo<CUDA_ONNX_OP_TYPED_CLASS_NAME(1, float_float_MLFloat16, SimplifiedLayerNormalization)>,
-    BuildKernelCreateInfo<CUDA_ONNX_OP_TYPED_CLASS_NAME(1, MLFloat16_float_float, SimplifiedLayerNormalization)>,
-    BuildKernelCreateInfo<CUDA_ONNX_OP_TYPED_CLASS_NAME(1, BFloat16_float_BFloat16, SimplifiedLayerNormalization)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_CLASS_NAME(1, Inverse)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, MatMulNBits)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, MatMulNBits)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, BFloat16, MatMulBnb4)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, MatMulBnb4)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, MatMulBnb4)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_CLASS_NAME(1, BiasSoftmax)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_CLASS_NAME(1, BiasDropout)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_CLASS_NAME(1, BitmaskDropout)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_CLASS_NAME(1, BitmaskBiasDropout)>,
-
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, int8_t_MLFloat16, QuantizeLinear)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, uint8_t_MLFloat16, QuantizeLinear)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, int8_t_MLFloat16, DequantizeLinear)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, uint8_t_MLFloat16, DequantizeLinear)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float_int8_t, QAttention)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16_int8_t, QAttention)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_CLASS_NAME(1, UnfoldTensor)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_CLASS_NAME(1, DynamicTimeWarping)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_CLASS_NAME(1, Trilu)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, BFloat16, FastGelu)>,
-    // TransposedMatMul is still here for backward compatibility
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, BFloat16, TransposeMatMul)>,  // backward compatibility
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, BFloat16, FusedMatMul)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, FusedConv)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_CLASS_NAME(1, QOrderedMatMul)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_CLASS_NAME(1, QOrderedLayerNormalization)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_CLASS_NAME(1, QOrderedGelu)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_CLASS_NAME(1, QuantizeWithOrder)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_CLASS_NAME(1, DequantizeWithOrder)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_CLASS_NAME(1, QOrderedAttention)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_CLASS_NAME(1, QOrderedLongformerAttention)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, DecoderMaskedSelfAttention)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, DecoderMaskedSelfAttention)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, DecoderMaskedMultiHeadAttention)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, DecoderMaskedMultiHeadAttention)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_CLASS_NAME(1, GemmFloat8)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, SparseAttention)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, BFloat16, SparseAttention)>,
+      BuildKernelCreateInfo<void>,  // default entry to avoid the list become empty after ops-reducing
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, GridSample)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, FastGelu)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, FastGelu)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, Gelu)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, double, Gelu)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, Gelu)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_CLASS_NAME(1, BiasGelu)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, BiasSplitGelu)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, BiasSplitGelu)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, BiasAdd)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, BiasAdd)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, QuickGelu)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, double, QuickGelu)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, QuickGelu)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, TransposeMatMul)>,      // backward compatibility
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, double, TransposeMatMul)>,     // backward compatibility
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, TransposeMatMul)>,  // backward compatibility
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, FusedMatMul)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, double, FusedMatMul)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, FusedMatMul)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, RelativePositionBias)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, RelativePositionBias)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, GatedRelativePositionBias)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, GatedRelativePositionBias)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, RemovePadding)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, RemovePadding)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, RestorePadding)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, RestorePadding)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, Rfft)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, double, Rfft)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, Rfft)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, Irfft)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, double, Irfft)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, Irfft)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, ComplexMul)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, ComplexMul)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, ComplexMulConj)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, ComplexMulConj)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_CLASS_NAME(1, NGramRepeatBlock)>,
+
+      // These ops were experimental ops in onnx domain which have been removed now. We add them here as
+      // contrib ops to maintain backward compatibility
+      BuildKernelCreateInfo<CUDA_ONNX_OP_TYPED_CLASS_NAME(1, float, Affine)>,
+      BuildKernelCreateInfo<CUDA_ONNX_OP_TYPED_CLASS_NAME(1, double, Affine)>,
+      BuildKernelCreateInfo<CUDA_ONNX_OP_TYPED_CLASS_NAME(1, MLFloat16, Affine)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, Attention)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, Attention)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, PackedAttention)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, PackedAttention)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, PackedMultiHeadAttention)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, PackedMultiHeadAttention)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_CLASS_NAME(1, BeamSearch)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_CLASS_NAME(1, WhisperBeamSearch)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, ConvTransposeWithDynamicPads)>,
+      BuildKernelCreateInfo<CUDA_ONNX_OP_TYPED_CLASS_NAME(1, float, Crop)>,
+      BuildKernelCreateInfo<CUDA_ONNX_OP_TYPED_CLASS_NAME(1, double, Crop)>,
+      BuildKernelCreateInfo<CUDA_ONNX_OP_TYPED_CLASS_NAME(1, MLFloat16, Crop)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, MoE)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, MoE)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_CLASS_NAME(1, QMoE)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, MultiHeadAttention)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, MultiHeadAttention)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, GroupQueryAttention)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, BFloat16, GroupQueryAttention)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, DecoderAttention)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, DecoderAttention)>,
+      BuildKernelCreateInfo<CUDA_ONNX_OP_TYPED_CLASS_NAME(1, int32_t, DynamicSlice)>,
+      BuildKernelCreateInfo<CUDA_ONNX_OP_TYPED_CLASS_NAME(1, int64_t, DynamicSlice)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, EmbedLayerNormalization)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, EmbedLayerNormalization)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_CLASS_NAME(1, GreedySearch)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_CLASS_NAME(1, GroupNorm)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, NhwcConv)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, NhwcConv)>,
+      BuildKernelCreateInfo<CUDA_ONNX_OP_TYPED_CLASS_NAME(1, float, ImageScaler)>,
+      BuildKernelCreateInfo<CUDA_ONNX_OP_TYPED_CLASS_NAME(1, double, ImageScaler)>,
+      BuildKernelCreateInfo<CUDA_ONNX_OP_TYPED_CLASS_NAME(1, MLFloat16, ImageScaler)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, LongformerAttention)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, LongformerAttention)>,
+      BuildKernelCreateInfo<CUDA_ONNX_OP_TYPED_CLASS_NAME(1, float, ParametricSoftplus)>,
+      BuildKernelCreateInfo<CUDA_ONNX_OP_TYPED_CLASS_NAME(1, double, ParametricSoftplus)>,
+      BuildKernelCreateInfo<CUDA_ONNX_OP_TYPED_CLASS_NAME(1, MLFloat16, ParametricSoftplus)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, RotaryEmbedding)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, RotaryEmbedding)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, BFloat16, RotaryEmbedding)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, GemmaRotaryEmbedding)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_CLASS_NAME(1, Sampling)>,
+      BuildKernelCreateInfo<CUDA_ONNX_OP_TYPED_CLASS_NAME(1, float, ScaledTanh)>,
+      BuildKernelCreateInfo<CUDA_ONNX_OP_TYPED_CLASS_NAME(1, double, ScaledTanh)>,
+      BuildKernelCreateInfo<CUDA_ONNX_OP_TYPED_CLASS_NAME(1, MLFloat16, ScaledTanh)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_CLASS_NAME(1, SkipGroupNorm)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, SkipLayerNormalization)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, SkipLayerNormalization)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, SkipSimplifiedLayerNormalization)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, SkipSimplifiedLayerNormalization)>,
+      BuildKernelCreateInfo<CUDA_ONNX_OP_TYPED_CLASS_NAME(1, float, ThresholdedRelu)>,
+      BuildKernelCreateInfo<CUDA_ONNX_OP_TYPED_CLASS_NAME(1, double, ThresholdedRelu)>,
+      BuildKernelCreateInfo<CUDA_ONNX_OP_TYPED_CLASS_NAME(1, MLFloat16, ThresholdedRelu)>,
+      BuildKernelCreateInfo<CUDA_ONNX_OP_VERSIONED_TYPED_CLASS_NAME(1, 16, float_float_float, LayerNormalization)>,
+      BuildKernelCreateInfo<CUDA_ONNX_OP_VERSIONED_TYPED_CLASS_NAME(1, 16, double_double_double, LayerNormalization)>,
+      BuildKernelCreateInfo<CUDA_ONNX_OP_VERSIONED_TYPED_CLASS_NAME(1, 16, MLFloat16_float_MLFloat16, LayerNormalization)>,
+      BuildKernelCreateInfo<CUDA_ONNX_OP_VERSIONED_TYPED_CLASS_NAME(1, 16, float_float_MLFloat16, LayerNormalization)>,
+      BuildKernelCreateInfo<CUDA_ONNX_OP_VERSIONED_TYPED_CLASS_NAME(1, 16, MLFloat16_float_float, LayerNormalization)>,
+      BuildKernelCreateInfo<CUDA_ONNX_OP_VERSIONED_TYPED_CLASS_NAME(1, 16, BFloat16_float_BFloat16, LayerNormalization)>,
+      BuildKernelCreateInfo<CUDA_ONNX_OP_TYPED_CLASS_NAME(1, float_float_float, SimplifiedLayerNormalization)>,
+      BuildKernelCreateInfo<CUDA_ONNX_OP_TYPED_CLASS_NAME(1, double_double_double, SimplifiedLayerNormalization)>,
+      BuildKernelCreateInfo<CUDA_ONNX_OP_TYPED_CLASS_NAME(1, MLFloat16_float_MLFloat16, SimplifiedLayerNormalization)>,
+      BuildKernelCreateInfo<CUDA_ONNX_OP_TYPED_CLASS_NAME(1, float_float_MLFloat16, SimplifiedLayerNormalization)>,
+      BuildKernelCreateInfo<CUDA_ONNX_OP_TYPED_CLASS_NAME(1, MLFloat16_float_float, SimplifiedLayerNormalization)>,
+      BuildKernelCreateInfo<CUDA_ONNX_OP_TYPED_CLASS_NAME(1, BFloat16_float_BFloat16, SimplifiedLayerNormalization)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_CLASS_NAME(1, Inverse)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, MatMulNBits)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, MatMulNBits)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, BFloat16, MatMulBnb4)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, MatMulBnb4)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, MatMulBnb4)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_CLASS_NAME(1, BiasSoftmax)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_CLASS_NAME(1, BiasDropout)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_CLASS_NAME(1, BitmaskDropout)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_CLASS_NAME(1, BitmaskBiasDropout)>,
+
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, int8_t_MLFloat16, QuantizeLinear)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, uint8_t_MLFloat16, QuantizeLinear)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, int8_t_MLFloat16, DequantizeLinear)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, uint8_t_MLFloat16, DequantizeLinear)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float_int8_t, QAttention)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16_int8_t, QAttention)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_CLASS_NAME(1, UnfoldTensor)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_CLASS_NAME(1, DynamicTimeWarping)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_CLASS_NAME(1, Trilu)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, BFloat16, FastGelu)>,
+      // TransposedMatMul is still here for backward compatibility
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, BFloat16, TransposeMatMul)>,  // backward compatibility
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, BFloat16, FusedMatMul)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, FusedConv)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_CLASS_NAME(1, QOrderedMatMul)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_CLASS_NAME(1, QOrderedLayerNormalization)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_CLASS_NAME(1, QOrderedGelu)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_CLASS_NAME(1, QuantizeWithOrder)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_CLASS_NAME(1, DequantizeWithOrder)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_CLASS_NAME(1, QOrderedAttention)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_CLASS_NAME(1, QOrderedLongformerAttention)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, DecoderMaskedSelfAttention)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, DecoderMaskedSelfAttention)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, DecoderMaskedMultiHeadAttention)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, DecoderMaskedMultiHeadAttention)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_CLASS_NAME(1, GemmFloat8)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, SparseAttention)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, BFloat16, SparseAttention)>,
 
 #ifdef ENABLE_ATEN
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kPytorchAtenDomain, 1, ATen)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kPytorchAtenDomain, 1, ATen)>,
 #endif
 
 #ifdef ENABLE_TRAINING_OPS
-    // Should remove the shrunken_gather include from ENABLE_TRAINING_OPS once
-    // 1). compute optimizer is enabled for inference or
-    // 2). this is needed by inference for other purpose.
-    BuildKernelCreateInfo<CUDA_MS_OP_CLASS_NAME(1, ShrunkenGather)>,
+      // Should remove the shrunken_gather include from ENABLE_TRAINING_OPS once
+      // 1). compute optimizer is enabled for inference or
+      // 2). this is needed by inference for other purpose.
+      BuildKernelCreateInfo<CUDA_MS_OP_CLASS_NAME(1, ShrunkenGather)>,
 #endif
 
 #if defined(ORT_USE_NCCL)
-    BuildKernelCreateInfo<CUDA_MS_OP_CLASS_NAME(1, AllReduce)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_CLASS_NAME(1, AllGather)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_CLASS_NAME(1, AllToAll)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_CLASS_NAME(1, AllReduce)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_CLASS_NAME(1, AllGather)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_CLASS_NAME(1, AllToAll)>,
 
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, ShardedMoE)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, ShardedMoE)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, ShardedMoE)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, ShardedMoE)>,
 
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, DistributedMatMul)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, DistributedMatMul)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, DistributedMatMul)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, DistributedMatMul)>,
 
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, DistributedSlice)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, DistributedSlice)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, DistributedSlice)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, DistributedSlice)>,
 
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, int64_t, DistributedReshape)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, DistributedReshape)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, DistributedReshape)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, int64_t, DistributedReshape)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, DistributedReshape)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, DistributedReshape)>,
 
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, int64_t, DistributedExpand)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, DistributedExpand)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, DistributedExpand)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, int64_t, DistributedExpand)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, DistributedExpand)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, DistributedExpand)>,
 
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, DistributedReduceSum)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, DistributedReduceSum)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, DistributedReduceSum)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, DistributedReduceSum)>,
 
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, DistributedReduceMax)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, DistributedReduceMax)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, DistributedReduceMax)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, DistributedReduceMax)>,
 
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, DistributedReduceMean)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, DistributedReduceMean)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, DistributedReduceMean)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, DistributedReduceMean)>,
 
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, int64_t, DistributedUnsqueeze)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, DistributedUnsqueeze)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, DistributedUnsqueeze)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, int64_t, DistributedUnsqueeze)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, DistributedUnsqueeze)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, DistributedUnsqueeze)>,
 
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, int64_t, DistributedSqueeze)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, DistributedSqueeze)>,
-    BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, DistributedSqueeze)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, int64_t, DistributedSqueeze)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, float, DistributedSqueeze)>,
+      BuildKernelCreateInfo<CUDA_MS_OP_TYPED_CLASS_NAME(1, MLFloat16, DistributedSqueeze)>,
 #endif
   };
 
diff --git a/onnxruntime/core/framework/ex_lib_loader.h b/onnxruntime/core/framework/ex_lib_loader.h
index cc353a7521786..d7ea5db3e5a26 100644
--- a/onnxruntime/core/framework/ex_lib_loader.h
+++ b/onnxruntime/core/framework/ex_lib_loader.h
@@ -20,7 +20,7 @@ class ExLibLoader {
   virtual ~ExLibLoader();
 
  protected:
-  virtual void PreUnloadLibrary(void* /*handle*/){};
+  virtual void PreUnloadLibrary(void* /*handle*/) {};
 
   std::map<std::string, void*> dso_name_data_map_;
 
diff --git a/onnxruntime/core/graph/contrib_ops/contrib_defs.cc b/onnxruntime/core/graph/contrib_ops/contrib_defs.cc
index dea8775c89a30..2d51658953282 100644
--- a/onnxruntime/core/graph/contrib_ops/contrib_defs.cc
+++ b/onnxruntime/core/graph/contrib_ops/contrib_defs.cc
@@ -2665,10 +2665,10 @@ ONNX_MS_OPERATOR_SET_SCHEMA(CropAndResize, 1,
 
 #if !defined(DISABLE_FLOAT8_TYPES)
 #define GEMM_FLOAT8_TYPES \
-  { "tensor(float8e4m3fn)", "tensor(float8e5m2)", "tensor(float16)", "tensor(bfloat16)", "tensor(float)" }
+  {"tensor(float8e4m3fn)", "tensor(float8e5m2)", "tensor(float16)", "tensor(bfloat16)", "tensor(float)"}
 #else
 #define GEMM_FLOAT8_TYPES \
-  { "tensor(float16)", "tensor(bfloat16)", "tensor(float)" }
+  {"tensor(float16)", "tensor(bfloat16)", "tensor(float)"}
 #endif
 
 ONNX_MS_OPERATOR_SET_SCHEMA(GemmFloat8, 1,
diff --git a/onnxruntime/core/optimizer/transpose_optimization/optimizer_api.h b/onnxruntime/core/optimizer/transpose_optimization/optimizer_api.h
index c042bb0059ac2..e7d2d32809fc5 100644
--- a/onnxruntime/core/optimizer/transpose_optimization/optimizer_api.h
+++ b/onnxruntime/core/optimizer/transpose_optimization/optimizer_api.h
@@ -86,7 +86,7 @@ class TensorRef {
   /// <returns>Flattened tensor data in bytes</returns>
   virtual std::vector<uint8_t> Data() const = 0;
 
-  virtual ~TensorRef(){};
+  virtual ~TensorRef() {};
 };
 
 /// <summary>
@@ -131,7 +131,7 @@ class ValueInfoRef {
   /// <param name="axes">Indices of dimensions to add. Indices are relative to final shape.</param>
   virtual void UnsqueezeDims(const std::vector<int64_t>& axes) = 0;
 
-  virtual ~ValueInfoRef(){};
+  virtual ~ValueInfoRef() {};
 };
 
 /// <summary>
@@ -248,7 +248,7 @@ class NodeRef {
   /// <returns>Id</returns>
   virtual int64_t Id() const = 0;
 
-  virtual ~NodeRef(){};
+  virtual ~NodeRef() {};
 };
 
 /// <summary>
@@ -449,7 +449,7 @@ class GraphRef {
   /// <returns>True if output of the Graph.</returns>
   virtual bool IsGraphOutput(std::string_view name) const = 0;
 
-  virtual ~GraphRef(){};
+  virtual ~GraphRef() {};
 };
 
 }  // namespace api
diff --git a/onnxruntime/core/platform/path_lib.h b/onnxruntime/core/platform/path_lib.h
index fca8990f14821..94425a3999d42 100644
--- a/onnxruntime/core/platform/path_lib.h
+++ b/onnxruntime/core/platform/path_lib.h
@@ -228,11 +228,9 @@ inline std::basic_string<PATH_CHAR_TYPE> GetLastComponent(const std::basic_strin
   typename std::basic_string<PATH_CHAR_TYPE>::size_type pos = input.length();
   PATH_CHAR_TYPE sep = GetPathSep<PATH_CHAR_TYPE>();
   // remove trailing backslash
-  for (; pos > 1 && input[pos - 1] == sep; --pos)
-    ;
+  for (; pos > 1 && input[pos - 1] == sep; --pos);
   input.resize(pos);
-  for (; pos != 0 && input[pos - 1] != sep; --pos)
-    ;
+  for (; pos != 0 && input[pos - 1] != sep; --pos);
   return input.substr(pos);
 }
 
diff --git a/onnxruntime/core/providers/coreml/model/model.mm b/onnxruntime/core/providers/coreml/model/model.mm
index 4fd822f0d0d15..4d20061820e71 100644
--- a/onnxruntime/core/providers/coreml/model/model.mm
+++ b/onnxruntime/core/providers/coreml/model/model.mm
@@ -502,7 +502,7 @@ Status GetMLMultiArrayCopyInfo(const MLMultiArray* _Nonnull array,
 class Execution {
  public:
   Execution(const std::string& path, const logging::Logger& logger, uint32_t coreml_flags);
-  ~Execution(){};
+  ~Execution() {};
 
   Status LoadModel();
   Status Predict(const std::unordered_map<std::string, OnnxTensorData>& inputs,
diff --git a/onnxruntime/core/providers/cpu/cpu_execution_provider.cc b/onnxruntime/core/providers/cpu/cpu_execution_provider.cc
index 9147107ac518a..7ac68e3a9a69d 100644
--- a/onnxruntime/core/providers/cpu/cpu_execution_provider.cc
+++ b/onnxruntime/core/providers/cpu/cpu_execution_provider.cc
@@ -1133,1568 +1133,1568 @@ KernelCreateInfo BuildKernelCreateInfo<void>() {
 
 Status RegisterOnnxOperatorKernels(KernelRegistry& kernel_registry) {
   static const BuildKernelCreateInfoFn function_table[] = {
-    BuildKernelCreateInfo<void>,  // default entry to avoid the list become empty after ops-reducing
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 10, Clip)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, Elu)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, HardSigmoid)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 15,
-                                                                    LeakyRelu)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
-                                                                          float, Relu)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
-                                                                          double, Relu)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, Selu)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
-                                                                          float, Sigmoid)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
-                                                                          double, Sigmoid)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, Softplus)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, Softsign)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
-                                                                          float, Tanh)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
-                                                                          double, Tanh)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 8, PRelu)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, RandomNormal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, RandomUniform)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, RandomNormalLike)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, RandomUniformLike)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, Multinomial)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 12,
-                                                                          float, Add)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 12,
-                                                                          double, Add)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 12,
-                                                                          int32_t, Add)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 12,
-                                                                          int64_t, Add)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 12,
-                                                                          float, Sub)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 12,
-                                                                          double, Sub)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 12,
-                                                                          int32_t, Sub)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 12,
-                                                                          int64_t, Sub)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 12,
-                                                                          float, Mul)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 12,
-                                                                          double, Mul)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 12,
-                                                                          int32_t, Mul)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 12,
-                                                                          int64_t, Mul)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 12,
-                                                                          float, Div)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 12,
-                                                                          double, Div)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 12,
-                                                                          int32_t, Div)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 12,
-                                                                          int64_t, Div)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
-                                                                          float, Abs)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
-                                                                          double, Abs)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
-                                                                          int8_t, Abs)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
-                                                                          int16_t, Abs)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
-                                                                          int32_t, Abs)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
-                                                                          int64_t, Abs)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
-                                                                          uint8_t, Abs)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
-                                                                          uint16_t, Abs)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
-                                                                          uint32_t, Abs)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
-                                                                          uint64_t, Abs)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
-                                                                          float, Floor)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
-                                                                          double, Floor)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
-                                                                          float, Ceil)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
-                                                                          double, Ceil)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
-                                                                          float, Reciprocal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
-                                                                          double, Reciprocal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
-                                                                          float, Sqrt)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
-                                                                          double, Sqrt)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
-                                                                          float, Neg)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
-                                                                          double, Neg)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
-                                                                          int8_t, Neg)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
-                                                                          int32_t, Neg)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
-                                                                          int64_t, Neg)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 11, Pow)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
-                                                                          float, Exp)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
-                                                                          double, Exp)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
-                                                                          float, Log)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
-                                                                          double, Log)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 7,
-                                                                          float, Sum)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 7,
-                                                                          double, Sum)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 8, 12,
-                                                                          float, Sum)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 8, 12,
-                                                                          double, Sum)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 7,
-                                                                          float, Min)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 8, 11, Min)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 7,
-                                                                          float, Max)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 8, 11, Max)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, Not)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, And)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, Or)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, Xor)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 8,
-                                                                          float, Less)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 8,
-                                                                          double, Less)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 8,
-                                                                          float, Greater)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 8,
-                                                                          double, Greater)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 10,
-                                                                          bool, Equal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 10,
-                                                                          int32_t, Equal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 10,
-                                                                          int64_t, Equal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 10,
-                                                                          float, Equal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 10,
-                                                                          double, Equal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 7,
-                                                                          float, Mean)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 8, 12,
-                                                                          float, Mean)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, float, Sin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, double, Sin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, Cos)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, Tan)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, Asin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, Acos)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, Atan)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 8,
-                                                                          float, Gemm)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 8,
-                                                                          double, Gemm)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
-                                                                    Hardmax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
-                                                                          float, LogSoftmax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
-                                                                          double, LogSoftmax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 8,
-                                                                          float, MatMul)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 8,
-                                                                          double, MatMul)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
-                                                                          float, Softmax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
-                                                                          double, Softmax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 9,
-                                                                          float, TopK)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 9,
-                                                                          double, TopK)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 8,
-                                                                          float, BatchNormalization)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 8,
-                                                                          double, BatchNormalization)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10, Conv)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
-                                                                    ConvTranspose)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 8, Flatten)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6,
-                                                          InstanceNormalization)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, float,
-                                                                LpNormalization)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, double,
-                                                                LpNormalization)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 12, LRN)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 9,
-                                                                    AveragePool)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 7, MaxPool)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 8, 11,
-                                                                    MaxPool)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 2, 10, LpPool)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 2, GlobalLpPool)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, GlobalAveragePool)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, GlobalMaxPool)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, MaxRoiPool)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
-                                                                          float, ReduceL1)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
-                                                                          int32_t, ReduceL1)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
-                                                                          int64_t, ReduceL1)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
-                                                                          float, ReduceL2)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
-                                                                          int32_t, ReduceL2)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
-                                                                          int64_t, ReduceL2)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
-                                                                          float, ReduceLogSum)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
-                                                                          int32_t, ReduceLogSum)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
-                                                                          int64_t, ReduceLogSum)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
-                                                                          float, ReduceLogSumExp)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
-                                                                          double, ReduceLogSumExp)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
-                                                                          int32_t, ReduceLogSumExp)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
-                                                                          int64_t, ReduceLogSumExp)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
-                                                                          float, ReduceMax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
-                                                                          double, ReduceMax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
-                                                                          int32_t, ReduceMax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
-                                                                          int64_t, ReduceMax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
-                                                                          float, ReduceMean)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
-                                                                          double, ReduceMean)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
-                                                                          int32_t, ReduceMean)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
-                                                                          float, ReduceMin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
-                                                                          double, ReduceMin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
-                                                                          int32_t, ReduceMin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
-                                                                          int64_t, ReduceMin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
-                                                                          float, ReduceProd)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
-                                                                          int32_t, ReduceProd)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
-                                                                          int64_t, ReduceProd)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
-                                                                          float, ReduceSum)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
-                                                                          int32_t, ReduceSum)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
-                                                                          double, ReduceSum)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
-                                                                          int64_t, ReduceSum)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
-                                                                          float, ReduceSumSquare)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
-                                                                          int32_t, ReduceSumSquare)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
-                                                                          double, ReduceSumSquare)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
-                                                                          int64_t, ReduceSumSquare)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
-                                                                          float, ArgMax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
-                                                                          int8_t, ArgMax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
-                                                                          uint8_t, ArgMax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
-                                                                          int32_t, ArgMax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
-                                                                          float, ArgMin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
-                                                                          int32_t, ArgMin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 13, GRU)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 13, LSTM)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 13, RNN)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12, Cast)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 4, 10, Concat)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10, Gather)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 9, Dropout)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 12,
-                                                                    Identity)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 2, 10, Pad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 4, Reshape)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 5, 12,
-                                                                    Reshape)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 12, Shape)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 12, Size)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 9, Slice)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 12,
-                                                                    SpaceToDepth)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
-                                                                    DepthToSpace)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 2, 10, Split)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
-                                                                    Squeeze)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12, Tile)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 12,
-                                                                    Transpose)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
-                                                                    Unsqueeze)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 8,
-                                                                          float, Upsample)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 8,
-                                                                          int32_t, Upsample)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 8,
-                                                                          int8_t, Upsample)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 8,
-                                                                          uint8_t, Upsample)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 8, 12,
-                                                                          float, Expand)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 8, 12,
-                                                                          double, Expand)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 8, 12,
-                                                                          int8_t, Expand)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 8, 12,
-                                                                          int16_t, Expand)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 8, 12,
-                                                                          int32_t, Expand)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 8, 12,
-                                                                          int64_t, Expand)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 8, 12,
-                                                                          uint8_t, Expand)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 8, 12,
-                                                                          uint16_t, Expand)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 8, 12,
-                                                                          uint32_t, Expand)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 8, 12,
-                                                                          uint64_t, Expand)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 8, 12,
-                                                                          bool, Expand)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 8, 12,
-                                                                          MLFloat16, Expand)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 8, 12,
-                                                                          string, Expand)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 8, 8, Scan)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10, If)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10, Loop)>,
-
-    // Opset 9
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 10,
-                                                                    Compress)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 19,
-                                                                    ConstantOfShape)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 12,
-                                                                    MeanVarianceNormalization)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 12,
-                                                                          float, Greater)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 12,
-                                                                          double, Greater)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 12,
-                                                                          int32_t, Greater)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 12,
-                                                                          int64_t, Greater)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 12,
-                                                                          float, Less)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 12,
-                                                                          double, Less)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 12,
-                                                                          int32_t, Less)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 12,
-                                                                          int64_t, Less)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, EyeLike)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 12,
-                                                                          float, IsNaN)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 12,
-                                                                          double, IsNaN)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 12,
-                                                                          MLFloat16, IsNaN)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 12, Sign)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, Shrink)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 12,
-                                                                          float, Erf)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 10,
-                                                                          int64_t_int64_t_int64_t, OneHot)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 10,
-                                                                          float_int64_t_int64_t, OneHot)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 10,
-                                                                          int64_t_string_int64_t, OneHot)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 10,
-                                                                          float_string_int64_t, OneHot)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 10,
-                                                                          float_float_float, OneHot)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 10,
-                                                                          int64_t_int32_t_float, OneHot)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 10,
-                                                                          int64_t_float_int64_t, OneHot)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 10,
-                                                                          int32_t_float_int32_t, OneHot)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 10,
-                                                                          int32_t_float_float, OneHot)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 10,
-                                                                          int64_t_float_float, OneHot)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 10,
-                                                                          int64_t_float_int32_t, OneHot)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 10,
-                                                                    MaxUnpool)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, Sinh)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, Cosh)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, Asinh)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, Acosh)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, Atanh)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 10, Scan)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 10,
-                                                                    Scatter)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, TfIdfVectorizer)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 12,
-                                                                          bool, NonZero)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 12,
-                                                                          float, NonZero)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 12,
-                                                                          int32_t, NonZero)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 12,
-                                                                          int64_t, NonZero)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 12,
-                                                                          uint8_t, NonZero)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 15,
-                                                                          string, Where)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 15,
-                                                                          float, Where)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 15,
-                                                                          double, Where)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 15,
-                                                                          int32_t, Where)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 15,
-                                                                          int64_t, Where)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 15,
-                                                                          uint8_t, Where)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 10,
-                                                                    Flatten)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 10,
-                                                                          float, Gemm)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 10,
-                                                                          double, Gemm)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 12,
-                                                                          float, MatMul)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 12,
-                                                                          double, MatMul)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 12,
-                                                                          int32_t, MatMul)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 12,
-                                                                          int64_t, MatMul)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 13,
-                                                                          float, BatchNormalization)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 13,
-                                                                          double, BatchNormalization)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 15, PRelu)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 9,
-                                                                          float, Upsample)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 9,
-                                                                          int32_t, Upsample)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 9,
-                                                                          int8_t, Upsample)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 9,
-                                                                          uint8_t, Upsample)>,
-
-    // Opset 10
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 10, StringNormalizer)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 10, 10,
-                                                                          float, TopK)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 10, 10,
-                                                                          double, TopK)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 10, 10,
-                                                                    AveragePool)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 10, 12, Mod)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 10, 10,
-                                                                          float, Resize)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 10, 10,
-                                                                          int32_t, Resize)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 10, 10,
-                                                                          int8_t, Resize)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 10, 10,
-                                                                          uint8_t, Resize)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 10, ThresholdedRelu)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 10, 12,
-                                                                          uint8_t, DequantizeLinear)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 10, 12,
-                                                                          int8_t, DequantizeLinear)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 10, 12,
-                                                                          int32_t, DequantizeLinear)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 10, 12,
-                                                                          uint8_t, QuantizeLinear)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 10, 12,
-                                                                          int8_t, QuantizeLinear)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 10, uint8_t,
-                                                                QLinearMatMul)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 10, int8_t,
-                                                                QLinearMatMul)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 10, uint8_t,
-                                                                MatMulInteger)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 10, int8_t,
-                                                                MatMulInteger)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 10, ConvInteger)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 10, uint8_t,
-                                                                QLinearConv)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 10, int8_t,
-                                                                QLinearConv)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 10, 10, Slice)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 10, 11,
-                                                                    Dropout)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 10, 10,
-                                                                    NonMaxSuppression)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 10, 19, IsInf)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 10, 15,
-                                                                          float, RoiAlign)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 10, 15,
-                                                                          double, RoiAlign)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 10, ReverseSequence)>,
-    // opset 11
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 11, Clip)>,
-
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 13,
-                                                                          float, CumSum)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 13,
-                                                                          double, CumSum)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 13,
-                                                                          int32_t, CumSum)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 13,
-                                                                          int64_t, CumSum)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
-                                                                          bool, Equal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
-                                                                          int32_t, Equal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
-                                                                          int64_t, Equal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
-                                                                          float, Equal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
-                                                                          double, Equal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, float, Round)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, double, Round)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, MLFloat16,
-                                                                Round)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, uint8_t,
-                                                                DynamicQuantizeLinear)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
-                                                                          float, ArgMax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
-                                                                          double, ArgMax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
-                                                                          int8_t, ArgMax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
-                                                                          uint8_t, ArgMax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
-                                                                          int32_t, ArgMax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
-                                                                          float, ArgMin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
-                                                                          double, ArgMin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
-                                                                          int32_t, ArgMin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12, Loop)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
-                                                                    Hardmax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
-                                                                          float, LogSoftmax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
-                                                                          double, LogSoftmax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
-                                                                          double, Softmax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
-                                                                          float, Softmax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
-                                                                    DepthToSpace)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 15, Scan)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
-                                                                    Flatten)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, Compress)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
-                                                                    Concat)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
-                                                                    Gather)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12, Slice)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12, Split)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
-                                                                    Squeeze)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
-                                                                    Unsqueeze)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, Det)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
-                                                                    ScatterElements)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, NonMaxSuppression)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 18,
-                                                                    AveragePool)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, MaxUnpool)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 17,
-                                                                    LpPool)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, Conv)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, ConvTranspose)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12, If)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, SequenceLength)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, SequenceAt)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, SequenceEmpty)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, SequenceInsert)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, SequenceErase)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, SequenceConstruct)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, ConcatFromSequence)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, SplitToSequence)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
-                                                                    ScatterND)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
-                                                                          float, Gemm)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
-                                                                          double, Gemm)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
-                                                                    GatherElements)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, uint8_t,
-                                                                BitShift)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, uint32_t,
-                                                                BitShift)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, uint64_t,
-                                                                BitShift)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12, Pad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 11,
-                                                                    GatherND)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, Range)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, Unique)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, float, TopK)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, double, TopK)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, int64_t, TopK)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, int32_t, TopK)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11,
-                                                                int64_t_int64_t_int64_t, OneHot)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11,
-                                                                float_int64_t_int64_t, OneHot)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11,
-                                                                int64_t_string_int64_t, OneHot)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11,
-                                                                float_string_int64_t, OneHot)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11,
-                                                                float_float_float, OneHot)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11,
-                                                                int64_t_int32_t_float, OneHot)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11,
-                                                                int64_t_float_int64_t, OneHot)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11,
-                                                                int32_t_float_int32_t, OneHot)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11,
-                                                                int32_t_float_float, OneHot)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11,
-                                                                int64_t_float_float, OneHot)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11,
-                                                                int64_t_float_int32_t, OneHot)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
-                                                                          float, Resize)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
-                                                                          int32_t, Resize)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
-                                                                          int8_t, Resize)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
-                                                                          uint8_t, Resize)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 11,
-                                                                          float, ReduceMin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 11,
-                                                                          double, ReduceMin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 11,
-                                                                          int32_t, ReduceMin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 11,
-                                                                          int64_t, ReduceMin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 11,
-                                                                          float, ReduceMax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 11,
-                                                                          double, ReduceMax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 11,
-                                                                          int32_t, ReduceMax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 11,
-                                                                          int64_t, ReduceMax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
-                                                                          float, ReduceL1)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
-                                                                          int32_t, ReduceL1)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
-                                                                          int64_t, ReduceL1)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
-                                                                          float, ReduceL2)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
-                                                                          int32_t, ReduceL2)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
-                                                                          int64_t, ReduceL2)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
-                                                                          float, ReduceLogSum)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
-                                                                          int32_t, ReduceLogSum)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
-                                                                          int64_t, ReduceLogSum)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
-                                                                          float, ReduceLogSumExp)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
-                                                                          double, ReduceLogSumExp)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
-                                                                          int32_t, ReduceLogSumExp)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
-                                                                          int64_t, ReduceLogSumExp)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
-                                                                          float, ReduceMean)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
-                                                                          double, ReduceMean)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
-                                                                          int32_t, ReduceMean)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
-                                                                          float, ReduceProd)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
-                                                                          int32_t, ReduceProd)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
-                                                                          int64_t, ReduceProd)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
-                                                                          float, ReduceSum)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
-                                                                          int32_t, ReduceSum)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
-                                                                          double, ReduceSum)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
-                                                                          int64_t, ReduceSum)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
-                                                                          float, ReduceSumSquare)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
-                                                                          int32_t, ReduceSumSquare)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
-                                                                          double, ReduceSumSquare)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
-                                                                          int64_t, ReduceSumSquare)>,
-
-    // OpSet 12
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 12, Clip)>,
-
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 12, Min)>,
-
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 12, Max)>,
-
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 12, Pow)>,
-
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, MaxPool)>,
-
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 12,
-                                                                          float, ReduceMax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 12,
-                                                                          double, ReduceMax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 12,
-                                                                          int32_t, ReduceMax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 12,
-                                                                          int64_t, ReduceMax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 12,
-                                                                          int8_t, ReduceMax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 12,
-                                                                          uint8_t, ReduceMax)>,
-
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 12,
-                                                                          float, ReduceMin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 12,
-                                                                          double, ReduceMin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 12,
-                                                                          int32_t, ReduceMin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 12,
-                                                                          int64_t, ReduceMin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 12,
-                                                                          int8_t, ReduceMin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 12,
-                                                                          uint8_t, ReduceMin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 12,
-                                                                    GatherND)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, Einsum)>,
-
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, float, Expand)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, double,
-                                                                Expand)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, int8_t,
-                                                                Expand)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, int16_t,
-                                                                Expand)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, int32_t,
-                                                                Expand)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, int64_t,
-                                                                Expand)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, uint8_t,
-                                                                Expand)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, uint16_t,
-                                                                Expand)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, uint32_t,
-                                                                Expand)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, uint64_t,
-                                                                Expand)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, bool, Expand)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, MLFloat16,
-                                                                Expand)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, string,
-                                                                Expand)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, float, Erf)>,
-    // REVIEW(codemzs): ConstEigenVectorArrayMap.cast<MLFLoat16) does not seem to be supported.
-    // However these types work on GPU implementation.
-    // BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12,
-    // MLFloat16_MLFloat16, Dropout)>,
-    // BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12,
-    // MLFloat16_float, Dropout)>, BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider,
-    // kOnnxDomain, 12, MLFloat16_double, Dropout)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 12,
-                                                                          float_float, Dropout)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 12,
-                                                                          float_double, Dropout)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 12,
-                                                                          double_float, Dropout)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 12,
-                                                                          double_double, Dropout)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, Celu)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 15,
-                                                                          float, GreaterOrEqual)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 15,
-                                                                          double, GreaterOrEqual)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 15,
-                                                                          int32_t, GreaterOrEqual)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 15,
-                                                                          int64_t, GreaterOrEqual)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 15,
-                                                                          float, LessOrEqual)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 15,
-                                                                          double, LessOrEqual)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 15,
-                                                                          int32_t, LessOrEqual)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 15,
-                                                                          int64_t, LessOrEqual)>,
-
-    // opset 13
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 18, Cast)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, Clip)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, float, MatMul)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, double,
-                                                                MatMul)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, int32_t,
-                                                                MatMul)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, int64_t,
-                                                                MatMul)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, Min)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, Max)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, float, Mean)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, float, Gemm)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, double, Gemm)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, Sign)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 18, Size)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, float, Sum)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, double, Sum)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, float,
-                                                                Sigmoid)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, double,
-                                                                Sigmoid)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 18,
-                                                                          uint8_t, DequantizeLinear)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 18,
-                                                                          int8_t, DequantizeLinear)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 18,
-                                                                          int32_t, DequantizeLinear)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 18,
-                                                                          uint8_t, QuantizeLinear)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 18,
-                                                                          int8_t, QuantizeLinear)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 20,
-                                                                    Flatten)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, LRN)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13,
-                                                          MeanVarianceNormalization)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, float_float,
-                                                                Dropout)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, float_double,
-                                                                Dropout)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, double_float,
-                                                                Dropout)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, double_double,
-                                                                Dropout)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, float, ArgMax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, double,
-                                                                ArgMax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, int8_t,
-                                                                ArgMax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, uint8_t,
-                                                                ArgMax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, int32_t,
-                                                                ArgMax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, float, ArgMin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, double,
-                                                                ArgMin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, int32_t,
-                                                                ArgMin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 13,
-                                                                    Reshape)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 14, Shape)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, Concat)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 18,
-                                                                          bool, Equal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 18,
-                                                                          int32_t, Equal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 18,
-                                                                          int64_t, Equal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 18,
-                                                                          float, Equal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 18,
-                                                                          double, Equal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, float,
-                                                                Greater)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, double,
-                                                                Greater)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, int32_t,
-                                                                Greater)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, int64_t,
-                                                                Greater)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, float, Less)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, double, Less)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13,
-                                                                int32_t, Less)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13,
-                                                                int64_t, Less)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 13,
-                                                                          float, Add)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 13,
-                                                                          double, Add)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 13,
-                                                                          int32_t, Add)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 13,
-                                                                          int64_t, Add)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 13,
-                                                                          float, Sub)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 13,
-                                                                          double, Sub)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 13,
-                                                                          int32_t, Sub)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 13,
-                                                                          int64_t, Sub)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 13,
-                                                                          float, Mul)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 13,
-                                                                          double, Mul)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 13,
-                                                                          int32_t, Mul)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 13,
-                                                                          int64_t, Mul)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 13,
-                                                                          float, Div)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 13,
-                                                                          double, Div)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 13,
-                                                                          int32_t, Div)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 13,
-                                                                          int64_t, Div)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, float, Neg)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, double, Neg)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, int8_t, Neg)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, int32_t, Neg)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, int64_t, Neg)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, Mod)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, float, Abs)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, double, Abs)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, int8_t, Abs)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, int16_t, Abs)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, int32_t, Abs)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, int64_t, Abs)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, uint8_t, Abs)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, uint16_t, Abs)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, uint32_t, Abs)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, uint64_t, Abs)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, float,
-                                                                Reciprocal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, double,
-                                                                Reciprocal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, float, Floor)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, double, Floor)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, float, Ceil)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, double, Ceil)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, float, Sqrt)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, double, Sqrt)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 13,
-                                                                          float, Relu)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 13,
-                                                                          double, Relu)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, float, Tanh)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, double, Tanh)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, float, Exp)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, double, Exp)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, float, Log)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, double, Log)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 14, Pow)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, Slice)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17, Split)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 20,
-                                                                    Unsqueeze)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 20,
-                                                                    Squeeze)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 20,
-                                                                    Transpose)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, Tile)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, Gather)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, GatherElements)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, DepthToSpace)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, SpaceToDepth)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 15,
-                                                                    ScatterElements)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 15,
-                                                                    ScatterND)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 13,
-                                                                    Identity)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 19,
-                                                                          float, IsNaN)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 19,
-                                                                          double, IsNaN)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 19,
-                                                                          MLFloat16, IsNaN)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, bool, NonZero)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, float,
-                                                                NonZero)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, int32_t,
-                                                                NonZero)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, int64_t,
-                                                                NonZero)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, uint8_t,
-                                                                NonZero)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, GatherND)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17, Pad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
-                                                                          float, ReduceL1)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
-                                                                          int32_t, ReduceL1)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
-                                                                          int64_t, ReduceL1)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
-                                                                          float, ReduceL2)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
-                                                                          int32_t, ReduceL2)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
-                                                                          int64_t, ReduceL2)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
-                                                                          float, ReduceLogSum)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
-                                                                          int32_t, ReduceLogSum)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
-                                                                          int64_t, ReduceLogSum)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
-                                                                          float, ReduceLogSumExp)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
-                                                                          double, ReduceLogSumExp)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
-                                                                          int32_t, ReduceLogSumExp)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
-                                                                          int64_t, ReduceLogSumExp)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
-                                                                          float, ReduceMax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
-                                                                          double, ReduceMax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
-                                                                          int32_t, ReduceMax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
-                                                                          int64_t, ReduceMax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
-                                                                          int8_t, ReduceMax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
-                                                                          uint8_t, ReduceMax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
-                                                                          float, ReduceMean)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
-                                                                          double, ReduceMean)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
-                                                                          int32_t, ReduceMean)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
-                                                                          float, ReduceMin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
-                                                                          double, ReduceMin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
-                                                                          int32_t, ReduceMin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
-                                                                          int64_t, ReduceMin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
-                                                                          int8_t, ReduceMin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
-                                                                          uint8_t, ReduceMin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
-                                                                          float, ReduceProd)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
-                                                                          int32_t, ReduceProd)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
-                                                                          int64_t, ReduceProd)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
-                                                                          float, ReduceSumSquare)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
-                                                                          int32_t, ReduceSumSquare)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
-                                                                          double, ReduceSumSquare)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
-                                                                          int64_t, ReduceSumSquare)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, float,
-                                                                ReduceSum)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, int32_t,
-                                                                ReduceSum)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, double,
-                                                                ReduceSum)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, int64_t,
-                                                                ReduceSum)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
-                                                                          float, Resize)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
-                                                                          int32_t, Resize)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
-                                                                          int8_t, Resize)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
-                                                                          uint8_t, Resize)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 15, Loop)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 15, If)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, Hardmax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, float,
-                                                                LogSoftmax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, double,
-                                                                LogSoftmax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, double,
-                                                                Softmax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, float,
-                                                                Softmax)>,
-
-    // OpSet 14
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, float, CumSum)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, double,
-                                                                CumSum)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, int32_t,
-                                                                CumSum)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, int64_t,
-                                                                CumSum)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, float, Relu)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, double, Relu)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, int8_t, Relu)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, int32_t, Relu)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, Trilu)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, float, Add)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, double, Add)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, int32_t, Add)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, int64_t, Add)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, float, Sub)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, double, Sub)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, int32_t, Sub)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, int64_t, Sub)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, float, Mul)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, double, Mul)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, int32_t, Mul)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, int64_t, Mul)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, float, Div)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, double, Div)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, int32_t, Div)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, int64_t, Div)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, 18,
-                                                                    Reshape)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, 15,
-                                                                    Identity)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, 14,
-                                                                          float, BatchNormalization)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, 14,
-                                                                          double, BatchNormalization)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, GRU)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, LSTM)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, RNN)>,
-
-    // Opset 15
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 15, Pow)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 15, float,
-                                                                BatchNormalization)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 15, double,
-                                                                BatchNormalization)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 15, 18, Shape)>,
+      BuildKernelCreateInfo<void>,  // default entry to avoid the list become empty after ops-reducing
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 10, Clip)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, Elu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, HardSigmoid)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 15,
+                                                                      LeakyRelu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
+                                                                            float, Relu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
+                                                                            double, Relu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, Selu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
+                                                                            float, Sigmoid)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
+                                                                            double, Sigmoid)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, Softplus)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, Softsign)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
+                                                                            float, Tanh)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
+                                                                            double, Tanh)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 8, PRelu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, RandomNormal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, RandomUniform)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, RandomNormalLike)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, RandomUniformLike)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, Multinomial)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 12,
+                                                                            float, Add)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 12,
+                                                                            double, Add)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 12,
+                                                                            int32_t, Add)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 12,
+                                                                            int64_t, Add)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 12,
+                                                                            float, Sub)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 12,
+                                                                            double, Sub)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 12,
+                                                                            int32_t, Sub)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 12,
+                                                                            int64_t, Sub)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 12,
+                                                                            float, Mul)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 12,
+                                                                            double, Mul)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 12,
+                                                                            int32_t, Mul)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 12,
+                                                                            int64_t, Mul)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 12,
+                                                                            float, Div)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 12,
+                                                                            double, Div)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 12,
+                                                                            int32_t, Div)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 12,
+                                                                            int64_t, Div)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
+                                                                            float, Abs)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
+                                                                            double, Abs)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
+                                                                            int8_t, Abs)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
+                                                                            int16_t, Abs)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
+                                                                            int32_t, Abs)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
+                                                                            int64_t, Abs)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
+                                                                            uint8_t, Abs)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
+                                                                            uint16_t, Abs)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
+                                                                            uint32_t, Abs)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
+                                                                            uint64_t, Abs)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
+                                                                            float, Floor)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
+                                                                            double, Floor)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
+                                                                            float, Ceil)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
+                                                                            double, Ceil)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
+                                                                            float, Reciprocal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
+                                                                            double, Reciprocal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
+                                                                            float, Sqrt)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
+                                                                            double, Sqrt)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
+                                                                            float, Neg)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
+                                                                            double, Neg)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
+                                                                            int8_t, Neg)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
+                                                                            int32_t, Neg)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
+                                                                            int64_t, Neg)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 11, Pow)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
+                                                                            float, Exp)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
+                                                                            double, Exp)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
+                                                                            float, Log)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12,
+                                                                            double, Log)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 7,
+                                                                            float, Sum)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 7,
+                                                                            double, Sum)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 8, 12,
+                                                                            float, Sum)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 8, 12,
+                                                                            double, Sum)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 7,
+                                                                            float, Min)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 8, 11, Min)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 7,
+                                                                            float, Max)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 8, 11, Max)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, Not)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, And)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, Or)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, Xor)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 8,
+                                                                            float, Less)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 8,
+                                                                            double, Less)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 8,
+                                                                            float, Greater)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 8,
+                                                                            double, Greater)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 10,
+                                                                            bool, Equal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 10,
+                                                                            int32_t, Equal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 10,
+                                                                            int64_t, Equal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 10,
+                                                                            float, Equal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 10,
+                                                                            double, Equal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 7,
+                                                                            float, Mean)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 8, 12,
+                                                                            float, Mean)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, float, Sin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, double, Sin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, Cos)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, Tan)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, Asin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, Acos)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, Atan)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 8,
+                                                                            float, Gemm)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 8,
+                                                                            double, Gemm)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
+                                                                      Hardmax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
+                                                                            float, LogSoftmax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
+                                                                            double, LogSoftmax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 8,
+                                                                            float, MatMul)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 8,
+                                                                            double, MatMul)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
+                                                                            float, Softmax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
+                                                                            double, Softmax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 9,
+                                                                            float, TopK)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 9,
+                                                                            double, TopK)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 8,
+                                                                            float, BatchNormalization)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 8,
+                                                                            double, BatchNormalization)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10, Conv)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
+                                                                      ConvTranspose)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 8, Flatten)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6,
+                                                            InstanceNormalization)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, float,
+                                                                  LpNormalization)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, double,
+                                                                  LpNormalization)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 12, LRN)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 9,
+                                                                      AveragePool)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 7, MaxPool)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 8, 11,
+                                                                      MaxPool)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 2, 10, LpPool)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 2, GlobalLpPool)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, GlobalAveragePool)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, GlobalMaxPool)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, MaxRoiPool)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
+                                                                            float, ReduceL1)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
+                                                                            int32_t, ReduceL1)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
+                                                                            int64_t, ReduceL1)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
+                                                                            float, ReduceL2)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
+                                                                            int32_t, ReduceL2)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
+                                                                            int64_t, ReduceL2)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
+                                                                            float, ReduceLogSum)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
+                                                                            int32_t, ReduceLogSum)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
+                                                                            int64_t, ReduceLogSum)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
+                                                                            float, ReduceLogSumExp)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
+                                                                            double, ReduceLogSumExp)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
+                                                                            int32_t, ReduceLogSumExp)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
+                                                                            int64_t, ReduceLogSumExp)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
+                                                                            float, ReduceMax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
+                                                                            double, ReduceMax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
+                                                                            int32_t, ReduceMax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
+                                                                            int64_t, ReduceMax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
+                                                                            float, ReduceMean)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
+                                                                            double, ReduceMean)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
+                                                                            int32_t, ReduceMean)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
+                                                                            float, ReduceMin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
+                                                                            double, ReduceMin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
+                                                                            int32_t, ReduceMin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
+                                                                            int64_t, ReduceMin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
+                                                                            float, ReduceProd)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
+                                                                            int32_t, ReduceProd)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
+                                                                            int64_t, ReduceProd)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
+                                                                            float, ReduceSum)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
+                                                                            int32_t, ReduceSum)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
+                                                                            double, ReduceSum)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
+                                                                            int64_t, ReduceSum)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
+                                                                            float, ReduceSumSquare)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
+                                                                            int32_t, ReduceSumSquare)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
+                                                                            double, ReduceSumSquare)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
+                                                                            int64_t, ReduceSumSquare)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
+                                                                            float, ArgMax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
+                                                                            int8_t, ArgMax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
+                                                                            uint8_t, ArgMax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
+                                                                            int32_t, ArgMax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
+                                                                            float, ArgMin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
+                                                                            int32_t, ArgMin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 13, GRU)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 13, LSTM)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 13, RNN)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12, Cast)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 4, 10, Concat)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10, Gather)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 9, Dropout)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 12,
+                                                                      Identity)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 2, 10, Pad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 4, Reshape)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 5, 12,
+                                                                      Reshape)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 12, Shape)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 12, Size)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 9, Slice)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 12,
+                                                                      SpaceToDepth)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
+                                                                      DepthToSpace)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 2, 10, Split)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
+                                                                      Squeeze)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12, Tile)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 12,
+                                                                      Transpose)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10,
+                                                                      Unsqueeze)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 8,
+                                                                            float, Upsample)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 8,
+                                                                            int32_t, Upsample)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 8,
+                                                                            int8_t, Upsample)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 7, 8,
+                                                                            uint8_t, Upsample)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 8, 12,
+                                                                            float, Expand)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 8, 12,
+                                                                            double, Expand)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 8, 12,
+                                                                            int8_t, Expand)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 8, 12,
+                                                                            int16_t, Expand)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 8, 12,
+                                                                            int32_t, Expand)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 8, 12,
+                                                                            int64_t, Expand)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 8, 12,
+                                                                            uint8_t, Expand)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 8, 12,
+                                                                            uint16_t, Expand)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 8, 12,
+                                                                            uint32_t, Expand)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 8, 12,
+                                                                            uint64_t, Expand)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 8, 12,
+                                                                            bool, Expand)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 8, 12,
+                                                                            MLFloat16, Expand)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 8, 12,
+                                                                            string, Expand)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 8, 8, Scan)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10, If)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 1, 10, Loop)>,
+
+      // Opset 9
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 10,
+                                                                      Compress)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 19,
+                                                                      ConstantOfShape)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 12,
+                                                                      MeanVarianceNormalization)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 12,
+                                                                            float, Greater)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 12,
+                                                                            double, Greater)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 12,
+                                                                            int32_t, Greater)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 12,
+                                                                            int64_t, Greater)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 12,
+                                                                            float, Less)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 12,
+                                                                            double, Less)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 12,
+                                                                            int32_t, Less)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 12,
+                                                                            int64_t, Less)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, EyeLike)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 12,
+                                                                            float, IsNaN)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 12,
+                                                                            double, IsNaN)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 12,
+                                                                            MLFloat16, IsNaN)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 12, Sign)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, Shrink)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 12,
+                                                                            float, Erf)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 10,
+                                                                            int64_t_int64_t_int64_t, OneHot)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 10,
+                                                                            float_int64_t_int64_t, OneHot)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 10,
+                                                                            int64_t_string_int64_t, OneHot)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 10,
+                                                                            float_string_int64_t, OneHot)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 10,
+                                                                            float_float_float, OneHot)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 10,
+                                                                            int64_t_int32_t_float, OneHot)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 10,
+                                                                            int64_t_float_int64_t, OneHot)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 10,
+                                                                            int32_t_float_int32_t, OneHot)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 10,
+                                                                            int32_t_float_float, OneHot)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 10,
+                                                                            int64_t_float_float, OneHot)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 10,
+                                                                            int64_t_float_int32_t, OneHot)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 10,
+                                                                      MaxUnpool)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, Sinh)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, Cosh)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, Asinh)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, Acosh)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, Atanh)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 10, Scan)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 10,
+                                                                      Scatter)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, TfIdfVectorizer)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 12,
+                                                                            bool, NonZero)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 12,
+                                                                            float, NonZero)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 12,
+                                                                            int32_t, NonZero)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 12,
+                                                                            int64_t, NonZero)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 12,
+                                                                            uint8_t, NonZero)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 15,
+                                                                            string, Where)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 15,
+                                                                            float, Where)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 15,
+                                                                            double, Where)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 15,
+                                                                            int32_t, Where)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 15,
+                                                                            int64_t, Where)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 15,
+                                                                            uint8_t, Where)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 10,
+                                                                      Flatten)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 10,
+                                                                            float, Gemm)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 10,
+                                                                            double, Gemm)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 12,
+                                                                            float, MatMul)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 12,
+                                                                            double, MatMul)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 12,
+                                                                            int32_t, MatMul)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 12,
+                                                                            int64_t, MatMul)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 13,
+                                                                            float, BatchNormalization)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 13,
+                                                                            double, BatchNormalization)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 15, PRelu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 9,
+                                                                            float, Upsample)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 9,
+                                                                            int32_t, Upsample)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 9,
+                                                                            int8_t, Upsample)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 9,
+                                                                            uint8_t, Upsample)>,
+
+      // Opset 10
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 10, StringNormalizer)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 10, 10,
+                                                                            float, TopK)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 10, 10,
+                                                                            double, TopK)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 10, 10,
+                                                                      AveragePool)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 10, 12, Mod)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 10, 10,
+                                                                            float, Resize)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 10, 10,
+                                                                            int32_t, Resize)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 10, 10,
+                                                                            int8_t, Resize)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 10, 10,
+                                                                            uint8_t, Resize)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 10, ThresholdedRelu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 10, 12,
+                                                                            uint8_t, DequantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 10, 12,
+                                                                            int8_t, DequantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 10, 12,
+                                                                            int32_t, DequantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 10, 12,
+                                                                            uint8_t, QuantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 10, 12,
+                                                                            int8_t, QuantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 10, uint8_t,
+                                                                  QLinearMatMul)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 10, int8_t,
+                                                                  QLinearMatMul)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 10, uint8_t,
+                                                                  MatMulInteger)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 10, int8_t,
+                                                                  MatMulInteger)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 10, ConvInteger)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 10, uint8_t,
+                                                                  QLinearConv)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 10, int8_t,
+                                                                  QLinearConv)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 10, 10, Slice)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 10, 11,
+                                                                      Dropout)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 10, 10,
+                                                                      NonMaxSuppression)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 10, 19, IsInf)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 10, 15,
+                                                                            float, RoiAlign)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 10, 15,
+                                                                            double, RoiAlign)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 10, ReverseSequence)>,
+      // opset 11
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 11, Clip)>,
+
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 13,
+                                                                            float, CumSum)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 13,
+                                                                            double, CumSum)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 13,
+                                                                            int32_t, CumSum)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 13,
+                                                                            int64_t, CumSum)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
+                                                                            bool, Equal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
+                                                                            int32_t, Equal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
+                                                                            int64_t, Equal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
+                                                                            float, Equal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
+                                                                            double, Equal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, float, Round)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, double, Round)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, MLFloat16,
+                                                                  Round)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, uint8_t,
+                                                                  DynamicQuantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
+                                                                            float, ArgMax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
+                                                                            double, ArgMax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
+                                                                            int8_t, ArgMax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
+                                                                            uint8_t, ArgMax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
+                                                                            int32_t, ArgMax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
+                                                                            float, ArgMin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
+                                                                            double, ArgMin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
+                                                                            int32_t, ArgMin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12, Loop)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
+                                                                      Hardmax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
+                                                                            float, LogSoftmax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
+                                                                            double, LogSoftmax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
+                                                                            double, Softmax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
+                                                                            float, Softmax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
+                                                                      DepthToSpace)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 15, Scan)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
+                                                                      Flatten)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, Compress)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
+                                                                      Concat)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
+                                                                      Gather)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12, Slice)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12, Split)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
+                                                                      Squeeze)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
+                                                                      Unsqueeze)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, Det)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
+                                                                      ScatterElements)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, NonMaxSuppression)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 18,
+                                                                      AveragePool)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, MaxUnpool)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 17,
+                                                                      LpPool)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, Conv)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, ConvTranspose)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12, If)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, SequenceLength)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, SequenceAt)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, SequenceEmpty)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, SequenceInsert)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, SequenceErase)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, SequenceConstruct)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, ConcatFromSequence)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, SplitToSequence)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
+                                                                      ScatterND)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
+                                                                            float, Gemm)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
+                                                                            double, Gemm)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
+                                                                      GatherElements)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, uint8_t,
+                                                                  BitShift)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, uint32_t,
+                                                                  BitShift)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, uint64_t,
+                                                                  BitShift)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12, Pad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 11,
+                                                                      GatherND)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, Range)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, Unique)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, float, TopK)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, double, TopK)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, int64_t, TopK)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, int32_t, TopK)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11,
+                                                                  int64_t_int64_t_int64_t, OneHot)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11,
+                                                                  float_int64_t_int64_t, OneHot)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11,
+                                                                  int64_t_string_int64_t, OneHot)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11,
+                                                                  float_string_int64_t, OneHot)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11,
+                                                                  float_float_float, OneHot)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11,
+                                                                  int64_t_int32_t_float, OneHot)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11,
+                                                                  int64_t_float_int64_t, OneHot)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11,
+                                                                  int32_t_float_int32_t, OneHot)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11,
+                                                                  int32_t_float_float, OneHot)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11,
+                                                                  int64_t_float_float, OneHot)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11,
+                                                                  int64_t_float_int32_t, OneHot)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
+                                                                            float, Resize)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
+                                                                            int32_t, Resize)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
+                                                                            int8_t, Resize)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
+                                                                            uint8_t, Resize)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 11,
+                                                                            float, ReduceMin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 11,
+                                                                            double, ReduceMin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 11,
+                                                                            int32_t, ReduceMin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 11,
+                                                                            int64_t, ReduceMin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 11,
+                                                                            float, ReduceMax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 11,
+                                                                            double, ReduceMax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 11,
+                                                                            int32_t, ReduceMax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 11,
+                                                                            int64_t, ReduceMax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
+                                                                            float, ReduceL1)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
+                                                                            int32_t, ReduceL1)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
+                                                                            int64_t, ReduceL1)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
+                                                                            float, ReduceL2)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
+                                                                            int32_t, ReduceL2)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
+                                                                            int64_t, ReduceL2)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
+                                                                            float, ReduceLogSum)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
+                                                                            int32_t, ReduceLogSum)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
+                                                                            int64_t, ReduceLogSum)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
+                                                                            float, ReduceLogSumExp)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
+                                                                            double, ReduceLogSumExp)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
+                                                                            int32_t, ReduceLogSumExp)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
+                                                                            int64_t, ReduceLogSumExp)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
+                                                                            float, ReduceMean)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
+                                                                            double, ReduceMean)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
+                                                                            int32_t, ReduceMean)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
+                                                                            float, ReduceProd)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
+                                                                            int32_t, ReduceProd)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
+                                                                            int64_t, ReduceProd)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
+                                                                            float, ReduceSum)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
+                                                                            int32_t, ReduceSum)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
+                                                                            double, ReduceSum)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
+                                                                            int64_t, ReduceSum)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
+                                                                            float, ReduceSumSquare)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
+                                                                            int32_t, ReduceSumSquare)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
+                                                                            double, ReduceSumSquare)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12,
+                                                                            int64_t, ReduceSumSquare)>,
+
+      // OpSet 12
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 12, Clip)>,
+
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 12, Min)>,
+
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 12, Max)>,
+
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 12, Pow)>,
+
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, MaxPool)>,
+
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 12,
+                                                                            float, ReduceMax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 12,
+                                                                            double, ReduceMax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 12,
+                                                                            int32_t, ReduceMax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 12,
+                                                                            int64_t, ReduceMax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 12,
+                                                                            int8_t, ReduceMax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 12,
+                                                                            uint8_t, ReduceMax)>,
+
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 12,
+                                                                            float, ReduceMin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 12,
+                                                                            double, ReduceMin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 12,
+                                                                            int32_t, ReduceMin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 12,
+                                                                            int64_t, ReduceMin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 12,
+                                                                            int8_t, ReduceMin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 12,
+                                                                            uint8_t, ReduceMin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 12,
+                                                                      GatherND)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, Einsum)>,
+
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, float, Expand)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, double,
+                                                                  Expand)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, int8_t,
+                                                                  Expand)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, int16_t,
+                                                                  Expand)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, int32_t,
+                                                                  Expand)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, int64_t,
+                                                                  Expand)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, uint8_t,
+                                                                  Expand)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, uint16_t,
+                                                                  Expand)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, uint32_t,
+                                                                  Expand)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, uint64_t,
+                                                                  Expand)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, bool, Expand)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, MLFloat16,
+                                                                  Expand)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, string,
+                                                                  Expand)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, float, Erf)>,
+      // REVIEW(codemzs): ConstEigenVectorArrayMap.cast<MLFLoat16) does not seem to be supported.
+      // However these types work on GPU implementation.
+      // BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12,
+      // MLFloat16_MLFloat16, Dropout)>,
+      // BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12,
+      // MLFloat16_float, Dropout)>, BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider,
+      // kOnnxDomain, 12, MLFloat16_double, Dropout)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 12,
+                                                                            float_float, Dropout)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 12,
+                                                                            float_double, Dropout)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 12,
+                                                                            double_float, Dropout)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 12,
+                                                                            double_double, Dropout)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, Celu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 15,
+                                                                            float, GreaterOrEqual)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 15,
+                                                                            double, GreaterOrEqual)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 15,
+                                                                            int32_t, GreaterOrEqual)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 15,
+                                                                            int64_t, GreaterOrEqual)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 15,
+                                                                            float, LessOrEqual)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 15,
+                                                                            double, LessOrEqual)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 15,
+                                                                            int32_t, LessOrEqual)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 15,
+                                                                            int64_t, LessOrEqual)>,
+
+      // opset 13
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 18, Cast)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, Clip)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, float, MatMul)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, double,
+                                                                  MatMul)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, int32_t,
+                                                                  MatMul)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, int64_t,
+                                                                  MatMul)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, Min)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, Max)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, float, Mean)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, float, Gemm)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, double, Gemm)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, Sign)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 18, Size)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, float, Sum)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, double, Sum)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, float,
+                                                                  Sigmoid)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, double,
+                                                                  Sigmoid)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 18,
+                                                                            uint8_t, DequantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 18,
+                                                                            int8_t, DequantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 18,
+                                                                            int32_t, DequantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 18,
+                                                                            uint8_t, QuantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 18,
+                                                                            int8_t, QuantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 20,
+                                                                      Flatten)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, LRN)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13,
+                                                            MeanVarianceNormalization)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, float_float,
+                                                                  Dropout)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, float_double,
+                                                                  Dropout)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, double_float,
+                                                                  Dropout)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, double_double,
+                                                                  Dropout)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, float, ArgMax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, double,
+                                                                  ArgMax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, int8_t,
+                                                                  ArgMax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, uint8_t,
+                                                                  ArgMax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, int32_t,
+                                                                  ArgMax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, float, ArgMin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, double,
+                                                                  ArgMin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, int32_t,
+                                                                  ArgMin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 13,
+                                                                      Reshape)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 14, Shape)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, Concat)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 18,
+                                                                            bool, Equal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 18,
+                                                                            int32_t, Equal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 18,
+                                                                            int64_t, Equal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 18,
+                                                                            float, Equal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 18,
+                                                                            double, Equal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, float,
+                                                                  Greater)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, double,
+                                                                  Greater)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, int32_t,
+                                                                  Greater)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, int64_t,
+                                                                  Greater)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, float, Less)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, double, Less)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13,
+                                                                  int32_t, Less)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13,
+                                                                  int64_t, Less)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 13,
+                                                                            float, Add)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 13,
+                                                                            double, Add)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 13,
+                                                                            int32_t, Add)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 13,
+                                                                            int64_t, Add)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 13,
+                                                                            float, Sub)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 13,
+                                                                            double, Sub)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 13,
+                                                                            int32_t, Sub)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 13,
+                                                                            int64_t, Sub)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 13,
+                                                                            float, Mul)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 13,
+                                                                            double, Mul)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 13,
+                                                                            int32_t, Mul)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 13,
+                                                                            int64_t, Mul)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 13,
+                                                                            float, Div)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 13,
+                                                                            double, Div)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 13,
+                                                                            int32_t, Div)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 13,
+                                                                            int64_t, Div)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, float, Neg)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, double, Neg)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, int8_t, Neg)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, int32_t, Neg)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, int64_t, Neg)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, Mod)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, float, Abs)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, double, Abs)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, int8_t, Abs)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, int16_t, Abs)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, int32_t, Abs)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, int64_t, Abs)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, uint8_t, Abs)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, uint16_t, Abs)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, uint32_t, Abs)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, uint64_t, Abs)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, float,
+                                                                  Reciprocal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, double,
+                                                                  Reciprocal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, float, Floor)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, double, Floor)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, float, Ceil)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, double, Ceil)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, float, Sqrt)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, double, Sqrt)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 13,
+                                                                            float, Relu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 13,
+                                                                            double, Relu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, float, Tanh)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, double, Tanh)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, float, Exp)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, double, Exp)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, float, Log)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, double, Log)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 14, Pow)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, Slice)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17, Split)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 20,
+                                                                      Unsqueeze)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 20,
+                                                                      Squeeze)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 20,
+                                                                      Transpose)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, Tile)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, Gather)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, GatherElements)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, DepthToSpace)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, SpaceToDepth)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 15,
+                                                                      ScatterElements)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 15,
+                                                                      ScatterND)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 13,
+                                                                      Identity)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 19,
+                                                                            float, IsNaN)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 19,
+                                                                            double, IsNaN)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 19,
+                                                                            MLFloat16, IsNaN)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, bool, NonZero)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, float,
+                                                                  NonZero)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, int32_t,
+                                                                  NonZero)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, int64_t,
+                                                                  NonZero)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, uint8_t,
+                                                                  NonZero)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, GatherND)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17, Pad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
+                                                                            float, ReduceL1)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
+                                                                            int32_t, ReduceL1)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
+                                                                            int64_t, ReduceL1)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
+                                                                            float, ReduceL2)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
+                                                                            int32_t, ReduceL2)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
+                                                                            int64_t, ReduceL2)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
+                                                                            float, ReduceLogSum)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
+                                                                            int32_t, ReduceLogSum)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
+                                                                            int64_t, ReduceLogSum)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
+                                                                            float, ReduceLogSumExp)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
+                                                                            double, ReduceLogSumExp)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
+                                                                            int32_t, ReduceLogSumExp)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
+                                                                            int64_t, ReduceLogSumExp)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
+                                                                            float, ReduceMax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
+                                                                            double, ReduceMax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
+                                                                            int32_t, ReduceMax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
+                                                                            int64_t, ReduceMax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
+                                                                            int8_t, ReduceMax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
+                                                                            uint8_t, ReduceMax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
+                                                                            float, ReduceMean)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
+                                                                            double, ReduceMean)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
+                                                                            int32_t, ReduceMean)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
+                                                                            float, ReduceMin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
+                                                                            double, ReduceMin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
+                                                                            int32_t, ReduceMin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
+                                                                            int64_t, ReduceMin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
+                                                                            int8_t, ReduceMin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
+                                                                            uint8_t, ReduceMin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
+                                                                            float, ReduceProd)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
+                                                                            int32_t, ReduceProd)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
+                                                                            int64_t, ReduceProd)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
+                                                                            float, ReduceSumSquare)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
+                                                                            int32_t, ReduceSumSquare)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
+                                                                            double, ReduceSumSquare)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
+                                                                            int64_t, ReduceSumSquare)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, float,
+                                                                  ReduceSum)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, int32_t,
+                                                                  ReduceSum)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, double,
+                                                                  ReduceSum)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, int64_t,
+                                                                  ReduceSum)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
+                                                                            float, Resize)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
+                                                                            int32_t, Resize)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
+                                                                            int8_t, Resize)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 17,
+                                                                            uint8_t, Resize)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 15, Loop)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, 15, If)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, Hardmax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, float,
+                                                                  LogSoftmax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, double,
+                                                                  LogSoftmax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, double,
+                                                                  Softmax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, float,
+                                                                  Softmax)>,
+
+      // OpSet 14
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, float, CumSum)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, double,
+                                                                  CumSum)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, int32_t,
+                                                                  CumSum)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, int64_t,
+                                                                  CumSum)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, float, Relu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, double, Relu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, int8_t, Relu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, int32_t, Relu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, Trilu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, float, Add)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, double, Add)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, int32_t, Add)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, int64_t, Add)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, float, Sub)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, double, Sub)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, int32_t, Sub)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, int64_t, Sub)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, float, Mul)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, double, Mul)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, int32_t, Mul)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, int64_t, Mul)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, float, Div)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, double, Div)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, int32_t, Div)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, int64_t, Div)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, 18,
+                                                                      Reshape)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, 15,
+                                                                      Identity)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, 14,
+                                                                            float, BatchNormalization)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, 14,
+                                                                            double, BatchNormalization)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, GRU)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, LSTM)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 14, RNN)>,
+
+      // Opset 15
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 15, Pow)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 15, float,
+                                                                  BatchNormalization)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 15, double,
+                                                                  BatchNormalization)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 15, 18, Shape)>,
 
 #if !defined(DISABLE_OPTIONAL_TYPE)
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 15, 17,
-                                                                    OptionalHasElement)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 15, 17,
-                                                                    OptionalGetElement)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 15, Optional)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 15, 17,
+                                                                      OptionalHasElement)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 15, 17,
+                                                                      OptionalGetElement)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 15, Optional)>,
 #endif
 
-    // Opset 16
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, 18,
-                                                                    Identity)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, 18, If)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, 18, Loop)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, float,
-                                                                RoiAlign)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, double,
-                                                                RoiAlign)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, 19,
-                                                                          float, GridSample)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, 17,
-                                                                    ScatterElements)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, 17,
-                                                                    ScatterND)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, string, Where)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, float, Where)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, double, Where)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, int32_t,
-                                                                Where)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, int64_t,
-                                                                Where)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, uint8_t,
-                                                                Where)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, LeakyRelu)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, PRelu)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, 18, Scan)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, float,
-                                                                GreaterOrEqual)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, double,
-                                                                GreaterOrEqual)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, int32_t,
-                                                                GreaterOrEqual)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, int64_t,
-                                                                GreaterOrEqual)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, float,
-                                                                LessOrEqual)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, double,
-                                                                LessOrEqual)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, int32_t,
-                                                                LessOrEqual)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, int64_t,
-                                                                LessOrEqual)>,
-
-    // Opset 17
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 17, BlackmanWindow)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 17, 19, DFT)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 17, HammingWindow)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 17, HannWindow)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 17, MelWeightMatrix)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 17, STFT)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 17, float,
-                                                                LayerNormalization)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 17, double,
-                                                                LayerNormalization)>,
-
-    // Opset 18
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, 18,
-                                                                          float, Resize)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, 18,
-                                                                          int32_t, Resize)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, 18,
-                                                                          int8_t, Resize)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, 18,
-                                                                          uint8_t, Resize)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, float,
-                                                                ReduceL1)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, int32_t,
-                                                                ReduceL1)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, int64_t,
-                                                                ReduceL1)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, float,
-                                                                ReduceL2)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, int32_t,
-                                                                ReduceL2)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, int64_t,
-                                                                ReduceL2)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, float,
-                                                                ReduceLogSum)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, int32_t,
-                                                                ReduceLogSum)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, int64_t,
-                                                                ReduceLogSum)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, float,
-                                                                ReduceLogSumExp)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, double,
-                                                                ReduceLogSumExp)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, int32_t,
-                                                                ReduceLogSumExp)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, int64_t,
-                                                                ReduceLogSumExp)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, 19, float,
-                                                                          ReduceMax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, 19, double,
-                                                                          ReduceMax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, 19, int32_t,
-                                                                          ReduceMax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, 19, int64_t,
-                                                                          ReduceMax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, 19, int8_t,
-                                                                          ReduceMax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, 19, uint8_t,
-                                                                          ReduceMax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, float,
-                                                                ReduceMean)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, double,
-                                                                ReduceMean)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, int32_t,
-                                                                ReduceMean)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, 19, float,
-                                                                          ReduceMin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, 19, double,
-                                                                          ReduceMin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, 19, int32_t,
-                                                                          ReduceMin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, 19, int64_t,
-                                                                          ReduceMin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, 19, int8_t,
-                                                                          ReduceMin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, 19, uint8_t,
-                                                                          ReduceMin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, float,
-                                                                ReduceProd)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, int32_t,
-                                                                ReduceProd)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, int64_t,
-                                                                ReduceProd)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, float,
-                                                                ReduceSumSquare)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, int32_t,
-                                                                ReduceSumSquare)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, double,
-                                                                ReduceSumSquare)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, int64_t,
-                                                                ReduceSumSquare)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, LpPool)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, Col2Im)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, int8_t,
-                                                                BitwiseAnd)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, int16_t,
-                                                                BitwiseAnd)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, int32_t,
-                                                                BitwiseAnd)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, int64_t,
-                                                                BitwiseAnd)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, uint8_t,
-                                                                BitwiseAnd)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, uint16_t,
-                                                                BitwiseAnd)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, uint32_t,
-                                                                BitwiseAnd)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, uint64_t,
-                                                                BitwiseAnd)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, int8_t,
-                                                                BitwiseNot)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, int16_t,
-                                                                BitwiseNot)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, int32_t,
-                                                                BitwiseNot)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, int64_t,
-                                                                BitwiseNot)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, uint8_t,
-                                                                BitwiseNot)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, uint16_t,
-                                                                BitwiseNot)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, uint32_t,
-                                                                BitwiseNot)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, uint64_t,
-                                                                BitwiseNot)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, int8_t,
-                                                                BitwiseOr)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, int16_t,
-                                                                BitwiseOr)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, int32_t,
-                                                                BitwiseOr)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, int64_t,
-                                                                BitwiseOr)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, uint8_t,
-                                                                BitwiseOr)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, uint16_t,
-                                                                BitwiseOr)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, uint32_t,
-                                                                BitwiseOr)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, uint64_t,
-                                                                BitwiseOr)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, int8_t,
-                                                                BitwiseXor)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, int16_t,
-                                                                BitwiseXor)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, int32_t,
-                                                                BitwiseXor)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, int64_t,
-                                                                BitwiseXor)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, uint8_t,
-                                                                BitwiseXor)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, uint16_t,
-                                                                BitwiseXor)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, uint32_t,
-                                                                BitwiseXor)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, uint64_t,
-                                                                BitwiseXor)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, 18, Pad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, ScatterND)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, ScatterElements)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, Split)>,
+      // Opset 16
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, 18,
+                                                                      Identity)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, 18, If)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, 18, Loop)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, float,
+                                                                  RoiAlign)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, double,
+                                                                  RoiAlign)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, 19,
+                                                                            float, GridSample)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, 17,
+                                                                      ScatterElements)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, 17,
+                                                                      ScatterND)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, string, Where)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, float, Where)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, double, Where)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, int32_t,
+                                                                  Where)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, int64_t,
+                                                                  Where)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, uint8_t,
+                                                                  Where)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, LeakyRelu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, PRelu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, 18, Scan)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, float,
+                                                                  GreaterOrEqual)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, double,
+                                                                  GreaterOrEqual)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, int32_t,
+                                                                  GreaterOrEqual)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, int64_t,
+                                                                  GreaterOrEqual)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, float,
+                                                                  LessOrEqual)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, double,
+                                                                  LessOrEqual)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, int32_t,
+                                                                  LessOrEqual)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, int64_t,
+                                                                  LessOrEqual)>,
+
+      // Opset 17
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 17, BlackmanWindow)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 17, 19, DFT)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 17, HammingWindow)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 17, HannWindow)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 17, MelWeightMatrix)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 17, STFT)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 17, float,
+                                                                  LayerNormalization)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 17, double,
+                                                                  LayerNormalization)>,
+
+      // Opset 18
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, 18,
+                                                                            float, Resize)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, 18,
+                                                                            int32_t, Resize)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, 18,
+                                                                            int8_t, Resize)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, 18,
+                                                                            uint8_t, Resize)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, float,
+                                                                  ReduceL1)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, int32_t,
+                                                                  ReduceL1)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, int64_t,
+                                                                  ReduceL1)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, float,
+                                                                  ReduceL2)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, int32_t,
+                                                                  ReduceL2)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, int64_t,
+                                                                  ReduceL2)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, float,
+                                                                  ReduceLogSum)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, int32_t,
+                                                                  ReduceLogSum)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, int64_t,
+                                                                  ReduceLogSum)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, float,
+                                                                  ReduceLogSumExp)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, double,
+                                                                  ReduceLogSumExp)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, int32_t,
+                                                                  ReduceLogSumExp)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, int64_t,
+                                                                  ReduceLogSumExp)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, 19, float,
+                                                                            ReduceMax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, 19, double,
+                                                                            ReduceMax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, 19, int32_t,
+                                                                            ReduceMax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, 19, int64_t,
+                                                                            ReduceMax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, 19, int8_t,
+                                                                            ReduceMax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, 19, uint8_t,
+                                                                            ReduceMax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, float,
+                                                                  ReduceMean)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, double,
+                                                                  ReduceMean)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, int32_t,
+                                                                  ReduceMean)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, 19, float,
+                                                                            ReduceMin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, 19, double,
+                                                                            ReduceMin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, 19, int32_t,
+                                                                            ReduceMin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, 19, int64_t,
+                                                                            ReduceMin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, 19, int8_t,
+                                                                            ReduceMin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, 19, uint8_t,
+                                                                            ReduceMin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, float,
+                                                                  ReduceProd)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, int32_t,
+                                                                  ReduceProd)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, int64_t,
+                                                                  ReduceProd)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, float,
+                                                                  ReduceSumSquare)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, int32_t,
+                                                                  ReduceSumSquare)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, double,
+                                                                  ReduceSumSquare)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, int64_t,
+                                                                  ReduceSumSquare)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, LpPool)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, Col2Im)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, int8_t,
+                                                                  BitwiseAnd)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, int16_t,
+                                                                  BitwiseAnd)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, int32_t,
+                                                                  BitwiseAnd)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, int64_t,
+                                                                  BitwiseAnd)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, uint8_t,
+                                                                  BitwiseAnd)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, uint16_t,
+                                                                  BitwiseAnd)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, uint32_t,
+                                                                  BitwiseAnd)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, uint64_t,
+                                                                  BitwiseAnd)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, int8_t,
+                                                                  BitwiseNot)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, int16_t,
+                                                                  BitwiseNot)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, int32_t,
+                                                                  BitwiseNot)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, int64_t,
+                                                                  BitwiseNot)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, uint8_t,
+                                                                  BitwiseNot)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, uint16_t,
+                                                                  BitwiseNot)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, uint32_t,
+                                                                  BitwiseNot)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, uint64_t,
+                                                                  BitwiseNot)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, int8_t,
+                                                                  BitwiseOr)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, int16_t,
+                                                                  BitwiseOr)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, int32_t,
+                                                                  BitwiseOr)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, int64_t,
+                                                                  BitwiseOr)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, uint8_t,
+                                                                  BitwiseOr)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, uint16_t,
+                                                                  BitwiseOr)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, uint32_t,
+                                                                  BitwiseOr)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, uint64_t,
+                                                                  BitwiseOr)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, int8_t,
+                                                                  BitwiseXor)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, int16_t,
+                                                                  BitwiseXor)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, int32_t,
+                                                                  BitwiseXor)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, int64_t,
+                                                                  BitwiseXor)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, uint8_t,
+                                                                  BitwiseXor)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, uint16_t,
+                                                                  BitwiseXor)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, uint32_t,
+                                                                  BitwiseXor)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, uint64_t,
+                                                                  BitwiseXor)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, 18, Pad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, ScatterND)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, ScatterElements)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, Split)>,
 #if !defined(DISABLE_OPTIONAL_TYPE)
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, OptionalHasElement)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, OptionalGetElement)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, OptionalHasElement)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 18, OptionalGetElement)>,
 #endif
 
-    // Opset 19
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, 20, Size)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, AveragePool)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, 20, Cast)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, 20,
-                                                                          uint8_t, DequantizeLinear)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, 20,
-                                                                          int8_t, DequantizeLinear)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, 20,
-                                                                          int32_t, DequantizeLinear)>,
+      // Opset 19
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, 20, Size)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, AveragePool)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, 20, Cast)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, 20,
+                                                                            uint8_t, DequantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, 20,
+                                                                            int8_t, DequantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, 20,
+                                                                            int32_t, DequantizeLinear)>,
 #if !defined(DISABLE_FLOAT8_TYPES)
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, 20,
-                                                                          Float8E4M3FN, DequantizeLinear)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, 20,
-                                                                          Float8E4M3FNUZ, DequantizeLinear)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, 20,
-                                                                          Float8E5M2, DequantizeLinear)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, 20,
-                                                                          Float8E5M2FNUZ, DequantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, 20,
+                                                                            Float8E4M3FN, DequantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, 20,
+                                                                            Float8E4M3FNUZ, DequantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, 20,
+                                                                            Float8E5M2, DequantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, 20,
+                                                                            Float8E5M2FNUZ, DequantizeLinear)>,
 #endif
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, bool, Equal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, int32_t,
-                                                                Equal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, int64_t,
-                                                                Equal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, float, Equal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, double, Equal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, string, Equal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, 20,
-                                                                    Identity)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, 20, If)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, 20, Loop)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, 20, Pad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, 20,
-                                                                          uint8_t, QuantizeLinear)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, 20,
-                                                                          int8_t, QuantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, bool, Equal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, int32_t,
+                                                                  Equal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, int64_t,
+                                                                  Equal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, float, Equal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, double, Equal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, string, Equal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, 20,
+                                                                      Identity)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, 20, If)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, 20, Loop)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, 20, Pad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, 20,
+                                                                            uint8_t, QuantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, 20,
+                                                                            int8_t, QuantizeLinear)>,
 #if !defined(DISABLE_FLOAT8_TYPES)
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, 20,
-                                                                          Float8E4M3FN, QuantizeLinear)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, 20,
-                                                                          Float8E4M3FNUZ, QuantizeLinear)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, 20,
-                                                                          Float8E5M2, QuantizeLinear)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, 20,
-                                                                          Float8E5M2FNUZ, QuantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, 20,
+                                                                            Float8E4M3FN, QuantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, 20,
+                                                                            Float8E4M3FNUZ, QuantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, 20,
+                                                                            Float8E5M2, QuantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, 20,
+                                                                            Float8E5M2FNUZ, QuantizeLinear)>,
 #endif
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, 20,
-                                                                    Reshape)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, float, Resize)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, int32_t,
-                                                                Resize)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, int8_t,
-                                                                Resize)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, uint8_t,
-                                                                Resize)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, 20, Scan)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, 20, Shape)>,
-
-    // Opset 20
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, 20,
-                                                                    ConstantOfShape)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, bool, ReduceMax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, float, ReduceMax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, double, ReduceMax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, int32_t, ReduceMax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, int64_t, ReduceMax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, int8_t, ReduceMax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, uint8_t, ReduceMax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, bool, ReduceMin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, float, ReduceMin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, double, ReduceMin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, int32_t, ReduceMin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, int64_t, ReduceMin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, int8_t, ReduceMin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, uint8_t, ReduceMin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, DFT)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, float,
-                                                                GridSample)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, double,
-                                                                GridSample)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, float,
-                                                                AffineGrid)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, double,
-                                                                AffineGrid)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, float, IsNaN)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, double, IsNaN)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, MLFloat16,
-                                                                IsNaN)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, BFloat16,
-                                                                IsNaN)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, Gelu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, 20,
+                                                                      Reshape)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, float, Resize)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, int32_t,
+                                                                  Resize)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, int8_t,
+                                                                  Resize)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, uint8_t,
+                                                                  Resize)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, 20, Scan)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, 20, Shape)>,
+
+      // Opset 20
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, 20,
+                                                                      ConstantOfShape)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, bool, ReduceMax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, float, ReduceMax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, double, ReduceMax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, int32_t, ReduceMax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, int64_t, ReduceMax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, int8_t, ReduceMax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, uint8_t, ReduceMax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, bool, ReduceMin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, float, ReduceMin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, double, ReduceMin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, int32_t, ReduceMin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, int64_t, ReduceMin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, int8_t, ReduceMin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, uint8_t, ReduceMin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, DFT)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, float,
+                                                                  GridSample)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, double,
+                                                                  GridSample)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, float,
+                                                                  AffineGrid)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, double,
+                                                                  AffineGrid)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, float, IsNaN)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, double, IsNaN)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, MLFloat16,
+                                                                  IsNaN)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, BFloat16,
+                                                                  IsNaN)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, Gelu)>,
 #if !defined(DISABLE_FLOAT8_TYPES)
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, Float8E4M3FN,
-                                                                IsNaN)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, Float8E4M3FNUZ,
-                                                                IsNaN)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, Float8E5M2,
-                                                                IsNaN)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, Float8E5M2FNUZ,
-                                                                IsNaN)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, Float8E4M3FN,
+                                                                  IsNaN)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, Float8E4M3FNUZ,
+                                                                  IsNaN)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, Float8E5M2,
+                                                                  IsNaN)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, Float8E5M2FNUZ,
+                                                                  IsNaN)>,
 #endif
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, IsInf)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, StringConcat)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, RegexFullMatch)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, StringSplit)>,
-
-    // Opset 21
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, Cast)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, ConstantOfShape)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, Identity)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, Reshape)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, Scan)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, Shape)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, Size)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, Flatten)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, Squeeze)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, Transpose)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, Unsqueeze)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, If)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, Loop)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, Pad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, uint8_t,
-                                                                DequantizeLinear)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, int8_t,
-                                                                DequantizeLinear)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, uint16_t,
-                                                                DequantizeLinear)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, int16_t,
-                                                                DequantizeLinear)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, int32_t,
-                                                                DequantizeLinear)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, Int4x2,
-                                                                DequantizeLinear)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, UInt4x2,
-                                                                DequantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, IsInf)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, StringConcat)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, RegexFullMatch)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, StringSplit)>,
+
+      // Opset 21
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, Cast)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, ConstantOfShape)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, Identity)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, Reshape)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, Scan)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, Shape)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, Size)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, Flatten)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, Squeeze)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, Transpose)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, Unsqueeze)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, If)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, Loop)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, Pad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, uint8_t,
+                                                                  DequantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, int8_t,
+                                                                  DequantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, uint16_t,
+                                                                  DequantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, int16_t,
+                                                                  DequantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, int32_t,
+                                                                  DequantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, Int4x2,
+                                                                  DequantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, UInt4x2,
+                                                                  DequantizeLinear)>,
 #if !defined(DISABLE_FLOAT8_TYPES)
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, Float8E4M3FN,
-                                                                DequantizeLinear)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, Float8E4M3FNUZ,
-                                                                DequantizeLinear)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, Float8E5M2,
-                                                                DequantizeLinear)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, Float8E5M2FNUZ,
-                                                                DequantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, Float8E4M3FN,
+                                                                  DequantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, Float8E4M3FNUZ,
+                                                                  DequantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, Float8E5M2,
+                                                                  DequantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, Float8E5M2FNUZ,
+                                                                  DequantizeLinear)>,
 #endif
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, uint8_t,
-                                                                QuantizeLinear)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, int8_t,
-                                                                QuantizeLinear)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, uint16_t,
-                                                                QuantizeLinear)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, int16_t,
-                                                                QuantizeLinear)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, Int4x2,
-                                                                QuantizeLinear)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, UInt4x2,
-                                                                QuantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, uint8_t,
+                                                                  QuantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, int8_t,
+                                                                  QuantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, uint16_t,
+                                                                  QuantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, int16_t,
+                                                                  QuantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, Int4x2,
+                                                                  QuantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, UInt4x2,
+                                                                  QuantizeLinear)>,
 #if !defined(DISABLE_FLOAT8_TYPES)
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, Float8E4M3FN,
-                                                                QuantizeLinear)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, Float8E4M3FNUZ,
-                                                                QuantizeLinear)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, Float8E5M2,
-                                                                QuantizeLinear)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, Float8E5M2FNUZ,
-                                                                QuantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, Float8E4M3FN,
+                                                                  QuantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, Float8E4M3FNUZ,
+                                                                  QuantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, Float8E5M2,
+                                                                  QuantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 21, Float8E5M2FNUZ,
+                                                                  QuantizeLinear)>,
 #endif
   };
 
diff --git a/onnxruntime/core/providers/cpu/ml/tree_ensemble_classifier.cc b/onnxruntime/core/providers/cpu/ml/tree_ensemble_classifier.cc
index 0c45b315f0280..758066d8a84e0 100644
--- a/onnxruntime/core/providers/cpu/ml/tree_ensemble_classifier.cc
+++ b/onnxruntime/core/providers/cpu/ml/tree_ensemble_classifier.cc
@@ -41,16 +41,16 @@ TreeEnsembleClassifier<T>::TreeEnsembleClassifier(const OpKernelInfo& info) : Op
 
 template <typename T>
 Status TreeEnsembleClassifier<T>::GetRemovableAttributes(InlinedVector<std::string>& removable_attributes) const {
-  InlinedVector<std::string> names {
-    "base_values", "nodes_falsenodeids", "nodes_featureids", "nodes_hitrates",
-        "nodes_missing_value_tracks_true", "nodes_modes", "nodes_nodeids", "nodes_treeids",
-        "nodes_truenodeids", "nodes_values", "class_ids", "class_treeids", "class_nodeids",
-        "class_weights", "classlabels_strings",
-        "classlabels_int64s"
+  InlinedVector<std::string> names{
+      "base_values", "nodes_falsenodeids", "nodes_featureids", "nodes_hitrates",
+      "nodes_missing_value_tracks_true", "nodes_modes", "nodes_nodeids", "nodes_treeids",
+      "nodes_truenodeids", "nodes_values", "class_ids", "class_treeids", "class_nodeids",
+      "class_weights", "classlabels_strings",
+      "classlabels_int64s"
 #if !defined(ORT_MINIMAL_BUILD)
-        "base_values_as_tensor",
-        "nodes_hitrates_as_tensor", "nodes_values_as_tensor",
-        "class_weights_as_tensor"
+      "base_values_as_tensor",
+      "nodes_hitrates_as_tensor", "nodes_values_as_tensor",
+      "class_weights_as_tensor"
 #endif
   };
   removable_attributes.swap(names);
diff --git a/onnxruntime/core/providers/cpu/ml/treeregressor.cc b/onnxruntime/core/providers/cpu/ml/treeregressor.cc
index 17f5cf32960da..6b5b972d3c929 100644
--- a/onnxruntime/core/providers/cpu/ml/treeregressor.cc
+++ b/onnxruntime/core/providers/cpu/ml/treeregressor.cc
@@ -48,16 +48,16 @@ TreeEnsembleRegressor<T>::TreeEnsembleRegressor(const OpKernelInfo& info) : OpKe
 
 template <typename T>
 Status TreeEnsembleRegressor<T>::GetRemovableAttributes(InlinedVector<std::string>& removable_attributes) const {
-  InlinedVector<std::string> names {
-    "base_values", "nodes_falsenodeids", "nodes_featureids", "nodes_hitrates",
-        "nodes_missing_value_tracks_true", "nodes_modes", "nodes_nodeids", "nodes_treeids",
-        "nodes_truenodeids", "nodes_values",
-        "target_ids", "target_treeids", "target_nodeids",
-        "target_weights"
+  InlinedVector<std::string> names{
+      "base_values", "nodes_falsenodeids", "nodes_featureids", "nodes_hitrates",
+      "nodes_missing_value_tracks_true", "nodes_modes", "nodes_nodeids", "nodes_treeids",
+      "nodes_truenodeids", "nodes_values",
+      "target_ids", "target_treeids", "target_nodeids",
+      "target_weights"
 #if !defined(ORT_MINIMAL_BUILD)
-        "base_values_as_tensor",
-        "nodes_hitrates_as_tensor", "nodes_values_as_tensor",
-        "class_weights_as_tensor"
+      "base_values_as_tensor",
+      "nodes_hitrates_as_tensor", "nodes_values_as_tensor",
+      "class_weights_as_tensor"
 #endif
   };
   removable_attributes.swap(names);
diff --git a/onnxruntime/core/providers/cpu/object_detection/non_max_suppression.cc b/onnxruntime/core/providers/cpu/object_detection/non_max_suppression.cc
index 4a176b0726a18..721c2064fae03 100644
--- a/onnxruntime/core/providers/cpu/object_detection/non_max_suppression.cc
+++ b/onnxruntime/core/providers/cpu/object_detection/non_max_suppression.cc
@@ -195,8 +195,8 @@ Status NonMaxSuppression::Compute(OpKernelContext* ctx) const {
         }
         sorted_boxes.pop();
       }  // while
-    }    // for class_index
-  }      // for batch_index
+    }  // for class_index
+  }  // for batch_index
 
   constexpr auto last_dim = 3;
   const auto num_selected = selected_indices.size();
diff --git a/onnxruntime/core/providers/cpu/object_detection/roialign.cc b/onnxruntime/core/providers/cpu/object_detection/roialign.cc
index ead2ccaef002e..d8c81e5cb63e5 100644
--- a/onnxruntime/core/providers/cpu/object_detection/roialign.cc
+++ b/onnxruntime/core/providers/cpu/object_detection/roialign.cc
@@ -251,9 +251,9 @@ void RoiAlignForward(const TensorShape& output_shape, const T* bottom_data, floa
 
             top_data[index] = output_val;
           }  // for pw
-        }    // for ph
-      }      // for c
-    }        // for n
+        }  // for ph
+      }  // for c
+    }  // for n
   });
 }
 }  // namespace
diff --git a/onnxruntime/core/providers/cpu/tensor/expand.cc b/onnxruntime/core/providers/cpu/tensor/expand.cc
index 6ead2ea73460b..b0c636281bc7a 100644
--- a/onnxruntime/core/providers/cpu/tensor/expand.cc
+++ b/onnxruntime/core/providers/cpu/tensor/expand.cc
@@ -128,7 +128,7 @@ Status Expand<T>::Compute(OpKernelContext* context) const {
           memcpy(output_data + output_offset, input_data + input_offset, onnxruntime::narrow<size_t>(copy_byte));
           output_offsets[onnxruntime::narrow<size_t>(i)] = output_offset;
         }  // for i
-      };   // distribute_fn
+      };  // distribute_fn
 
   auto per_thread_tasks =
       distribute_count / concurrency::ThreadPool::DegreeOfParallelism(context->GetOperatorThreadPool());
@@ -169,9 +169,9 @@ Status Expand<T>::Compute(OpKernelContext* context) const {
                   copy_byte >>= 1;
                 }
               }  // while
-            }    // if
-          }      // for
-        };       // copy_fn
+            }  // if
+          }  // for
+        };  // copy_fn
     if (per_thread_tasks > 20) {
       concurrency::ThreadPool::TryParallelFor(
           context->GetOperatorThreadPool(),
@@ -181,7 +181,7 @@ Status Expand<T>::Compute(OpKernelContext* context) const {
     } else {
       copy_fn(0, onnxruntime::narrow<std::ptrdiff_t>(distribute_count));
     }  // else
-  }    // for
+  }  // for
   return Status::OK();
 }  // Expand::compute
 
diff --git a/onnxruntime/core/providers/cuda/cuda_execution_provider.cc b/onnxruntime/core/providers/cuda/cuda_execution_provider.cc
index 8c03e489d298d..5771380433b35 100644
--- a/onnxruntime/core/providers/cuda/cuda_execution_provider.cc
+++ b/onnxruntime/core/providers/cuda/cuda_execution_provider.cc
@@ -1394,916 +1394,916 @@ KernelCreateInfo BuildKernelCreateInfo<void>() {
 
 static Status RegisterCudaKernels(KernelRegistry& kernel_registry) {
   static const BuildKernelCreateInfoFn function_table[] = {
-    BuildKernelCreateInfo<void>,  // default entry to avoid the list become empty after ops-reducing
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, MemcpyFromHost)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, MemcpyToHost)>,
+      BuildKernelCreateInfo<void>,  // default entry to avoid the list become empty after ops-reducing
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, MemcpyFromHost)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, MemcpyToHost)>,
 #ifndef USE_CUDA_MINIMAL
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 4, 10, Concat)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 10, Unsqueeze)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 8, Flatten)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 10, Squeeze)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 12, Identity)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 9, Dropout)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, float, Cos)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, double, Cos)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, MLFloat16, Cos)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, float, Sin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, double, Sin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, MLFloat16, Sin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 10, Gather)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 8, float, Gemm)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 8, double, Gemm)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 8, MLFloat16, Gemm)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 10, float, Gemm)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 10, double, Gemm)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 10, MLFloat16, Gemm)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 8, float, MatMul)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 8, double, MatMul)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 8, MLFloat16, MatMul)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, float, MatMul)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, double, MatMul)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, MLFloat16, MatMul)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 10, int8_t, MatMulInteger)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 10, float, Clip)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, float, Elu)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, double, Elu)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, MLFloat16, Elu)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, float, HardSigmoid)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, double, HardSigmoid)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, MLFloat16, HardSigmoid)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 15, float, LeakyRelu)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 15, double, LeakyRelu)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 15, MLFloat16, LeakyRelu)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, float, Relu)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, double, Relu)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, MLFloat16, Relu)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, float, Selu)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, double, Selu)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, MLFloat16, Selu)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, float, Sigmoid)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, double, Sigmoid)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, MLFloat16, Sigmoid)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, float, Softsign)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, double, Softsign)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, MLFloat16, Softsign)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, float, Tanh)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, double, Tanh)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, MLFloat16, Tanh)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, float, Softplus)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, double, Softplus)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, MLFloat16, Softplus)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 10, float, Softmax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 10, double, Softmax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 10, MLFloat16, Softmax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 10, float, LogSoftmax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 10, double, LogSoftmax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 10, MLFloat16, LogSoftmax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 11, float, Pow)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 11, double, Pow)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 11, MLFloat16, Pow)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 8, float, PRelu)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 8, double, PRelu)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 8, MLFloat16, PRelu)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 15, float, PRelu)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 15, double, PRelu)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 15, MLFloat16, PRelu)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, bool, And)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, bool, Or)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, bool, Xor)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 7, Sum)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 8, 12, Sum)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 11, Max)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, 12, Max)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 11, Min)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, 12, Min)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 8, float, Greater)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 8, double, Greater)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 8, MLFloat16, Greater)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 10, bool, Equal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 10, int32_t, Equal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 10, int64_t, Equal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 8, 12, Expand)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, int32_t, Greater)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, int64_t, Greater)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, uint32_t, Greater)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, uint64_t, Greater)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, float, Greater)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, double, Greater)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, MLFloat16, Greater)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, 15, int32_t, GreaterOrEqual)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, 15, int64_t, GreaterOrEqual)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, 15, uint32_t, GreaterOrEqual)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, 15, uint64_t, GreaterOrEqual)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, 15, float, GreaterOrEqual)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, 15, double, GreaterOrEqual)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, 15, MLFloat16, GreaterOrEqual)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, 15, int32_t, LessOrEqual)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, 15, int64_t, LessOrEqual)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, 15, uint32_t, LessOrEqual)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, 15, uint64_t, LessOrEqual)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, 15, float, LessOrEqual)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, 15, double, LessOrEqual)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, 15, MLFloat16, LessOrEqual)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 12, int32_t, Add)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 12, int64_t, Add)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 12, uint32_t, Add)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 12, uint64_t, Add)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 12, float, Add)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 12, double, Add)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 12, MLFloat16, Add)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 12, int32_t, Sub)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 12, int64_t, Sub)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 12, uint32_t, Sub)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 12, uint64_t, Sub)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 12, float, Sub)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 12, double, Sub)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 12, MLFloat16, Sub)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 12, int32_t, Mul)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 12, int64_t, Mul)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 12, uint32_t, Mul)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 12, uint64_t, Mul)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 12, float, Mul)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 12, double, Mul)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 12, MLFloat16, Mul)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 12, int32_t, Div)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 12, int64_t, Div)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 12, uint32_t, Div)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 12, uint64_t, Div)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 12, float, Div)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 12, double, Div)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 12, MLFloat16, Div)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, int8_t, Abs)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, int16_t, Abs)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, int32_t, Abs)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, int64_t, Abs)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, uint8_t, Abs)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, uint16_t, Abs)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, uint32_t, Abs)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, uint64_t, Abs)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, float, Abs)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, double, Abs)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, MLFloat16, Abs)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, int8_t, Neg)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, int16_t, Neg)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, int32_t, Neg)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, int64_t, Neg)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, float, Neg)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, double, Neg)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, MLFloat16, Neg)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, float, Floor)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, double, Floor)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, MLFloat16, Floor)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, float, Ceil)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, double, Ceil)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, MLFloat16, Ceil)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, float, Reciprocal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, double, Reciprocal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, MLFloat16, Reciprocal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, float, Sqrt)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, double, Sqrt)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, MLFloat16, Sqrt)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, float, Log)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, double, Log)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, MLFloat16, Log)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, float, Exp)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, double, Exp)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, MLFloat16, Exp)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, float, Erf)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, double, Erf)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, MLFloat16, Erf)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, IsNaN)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, bool, Not)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 8, float, BatchNormalization)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 8, double, BatchNormalization)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 8, MLFloat16, BatchNormalization)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 13, float, BatchNormalization)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 13, double, BatchNormalization)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 13, MLFloat16, BatchNormalization)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 12, float, LRN)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 12, double, LRN)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 12, MLFloat16, LRN)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 10, float, Conv)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 10, double, Conv)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 10, MLFloat16, Conv)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 10, float, ConvTranspose)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 10, double, ConvTranspose)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 10, MLFloat16, ConvTranspose)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 9, float, AveragePool)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 9, double, AveragePool)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 9, MLFloat16, AveragePool)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, float, GlobalAveragePool)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, double, GlobalAveragePool)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, MLFloat16, GlobalAveragePool)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 7, float, MaxPool)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 7, double, MaxPool)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 7, MLFloat16, MaxPool)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 8, 9, float, MaxPool)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 8, 9, double, MaxPool)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 8, 9, MLFloat16, MaxPool)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, float, GlobalMaxPool)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, double, GlobalMaxPool)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, MLFloat16, GlobalMaxPool)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 11, float, ArgMin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 11, double, ArgMin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 11, MLFloat16, ArgMin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, float, ReduceL1)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, double, ReduceL1)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, MLFloat16, ReduceL1)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, int32_t, ReduceL1)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, float, ReduceL2)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, double, ReduceL2)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, MLFloat16, ReduceL2)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, int32_t, ReduceL2)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, float, ReduceMax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, double, ReduceMax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, MLFloat16, ReduceMax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, int32_t, ReduceMax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, int64_t, ReduceMax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, float, ReduceMean)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, double, ReduceMean)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, MLFloat16, ReduceMean)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, int32_t, ReduceMean)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, float, ReduceMin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, double, ReduceMin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, MLFloat16, ReduceMin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, int32_t, ReduceMin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, int64_t, ReduceMin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, int8_t, ReduceMin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, uint8_t, ReduceMin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, float, ReduceProd)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, double, ReduceProd)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, MLFloat16, ReduceProd)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, int32_t, ReduceProd)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 12, float, ReduceSum)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 12, double, ReduceSum)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 12, MLFloat16, ReduceSum)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 12, int32_t, ReduceSum)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 12, int64_t, ReduceSum)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, float, ReduceLogSum)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, double, ReduceLogSum)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, MLFloat16, ReduceLogSum)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, float, ReduceSumSquare)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, double, ReduceSumSquare)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, MLFloat16, ReduceSumSquare)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, float, ReduceLogSumExp)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, double, ReduceLogSumExp)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, MLFloat16, ReduceLogSumExp)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 8, float, Cast)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 8, double, Cast)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 8, MLFloat16, Cast)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 8, int8_t, Cast)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 8, int16_t, Cast)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 8, int32_t, Cast)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 8, int64_t, Cast)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 8, uint8_t, Cast)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 8, uint16_t, Cast)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 8, uint32_t, Cast)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 8, uint64_t, Cast)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 8, bool, Cast)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, float, Cast)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, double, Cast)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, MLFloat16, Cast)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, int8_t, Cast)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, int16_t, Cast)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, int32_t, Cast)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, int64_t, Cast)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, uint8_t, Cast)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, uint16_t, Cast)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, uint32_t, Cast)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, uint64_t, Cast)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, bool, Cast)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 2, 10, float, Pad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 2, 10, double, Pad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 2, 10, MLFloat16, Pad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 4, Reshape)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 5, 12, Reshape)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 12, Shape)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 12, Size)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, Tile)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, Tile)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 12, Transpose)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, float, InstanceNormalization)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, double, InstanceNormalization)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, MLFloat16, InstanceNormalization)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 13, float, RNN)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 13, double, RNN)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 13, MLFloat16, RNN)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 13, float, GRU)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 13, double, GRU)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 13, MLFloat16, GRU)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 13, float, LSTM)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 13, double, LSTM)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 13, MLFloat16, LSTM)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 9, int64_t, Slice)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 10, Compress)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 10, Flatten)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 8, float, Upsample)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 8, double, Upsample)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 8, MLFloat16, Upsample)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 8, int32_t, Upsample)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 8, uint8_t, Upsample)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 9, float, Upsample)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 9, double, Upsample)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 9, MLFloat16, Upsample)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 9, int32_t, Upsample)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 9, uint8_t, Upsample)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 2, 10, Split)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, ConstantOfShape)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, int8_t, Shrink)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, int16_t, Shrink)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, int32_t, Shrink)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, int64_t, Shrink)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, uint8_t, Shrink)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, uint16_t, Shrink)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, uint32_t, Shrink)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, uint64_t, Shrink)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, float, Shrink)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, double, Shrink)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, MLFloat16, Shrink)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 8, float, Less)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 8, double, Less)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 8, MLFloat16, Less)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, int32_t, Less)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, int64_t, Less)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, uint32_t, Less)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, uint64_t, Less)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, float, Less)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, double, Less)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, MLFloat16, Less)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, EyeLike)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 10, Scatter)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 15, MLFloat16, Where)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 15, float, Where)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 15, double_t, Where)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 15, int32_t, Where)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 15, int64_t, Where)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 15, uint8_t, Where)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, bool, NonZero)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, uint8_t, NonZero)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, int32_t, NonZero)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, int64_t, NonZero)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, float, NonZero)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, MLFloat16, NonZero)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 9, TopK)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 8, 8, Scan)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 10, Scan)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 10, Loop)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 10, DepthToSpace)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 12, SpaceToDepth)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, RandomNormal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, RandomNormalLike)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, RandomUniform)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, RandomUniformLike)>,
-
-    // opset 10
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 10, 10, float, AveragePool)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 10, 10, double, AveragePool)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 10, 10, MLFloat16, AveragePool)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 10, 11, Dropout)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 10, 10, float, MaxPool)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 10, 10, double, MaxPool)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 10, 10, MLFloat16, MaxPool)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 10, 10, NonMaxSuppression)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 10, 10, float, Resize)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 10, 10, double, Resize)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 10, 10, MLFloat16, Resize)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 10, 10, int32_t, Resize)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 10, 10, uint8_t, Resize)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 10, ReverseSequence)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 10, float, RoiAlign)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 10, double, RoiAlign)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 10, 10, int32_t, Slice)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 10, 10, int64_t, Slice)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 10, float, ThresholdedRelu)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 10, double, ThresholdedRelu)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 10, MLFloat16, ThresholdedRelu)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 10, 10, TopK)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 10, If)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 10, 12, int8_t, QuantizeLinear)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 10, 12, uint8_t, QuantizeLinear)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 10, 12, int8_t, DequantizeLinear)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 10, 12, uint8_t, DequantizeLinear)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 10, 12, Mod)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 10,
-                                                                    19, IsInf)>,
-
-    // opset 11
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 11, float, ArgMax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 11, double, ArgMax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 11, MLFloat16, ArgMax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, Compress)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, Concat)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, Flatten)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, Gather)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, GatherElements)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 11, int64_t, GatherND)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, MLFloat16, Gemm)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, float, Gemm)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, double, Gemm)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, If)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, Loop)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, NonMaxSuppression)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, Range)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 15, Scan)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, ScatterElements)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, int32_t, Slice)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, int64_t, Slice)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, float, Softmax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, double, Softmax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, MLFloat16, Softmax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, float, LogSoftmax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, double, LogSoftmax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, MLFloat16, LogSoftmax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, Split)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, Squeeze)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, TopK)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, SequenceAt)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, SequenceConstruct)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, SequenceEmpty)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, SequenceLength)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, ConcatFromSequence)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, SequenceErase)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, SequenceInsert)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, Unsqueeze)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, float, Conv)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, double, Conv)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, MLFloat16, Conv)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, float, ConvTranspose)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, double, ConvTranspose)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, MLFloat16, ConvTranspose)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, float, AveragePool)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, double, AveragePool)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, MLFloat16, AveragePool)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 11, float, MaxPool)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 11, double, MaxPool)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 11, MLFloat16, MaxPool)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, float, Resize)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, double, Resize)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, MLFloat16, Resize)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, int32_t, Resize)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, uint8_t, Resize)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 11, Clip)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, float, Pad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, double, Pad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, MLFloat16, Pad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, bool, Equal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, int32_t, Equal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, int64_t, Equal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, uint32_t, Equal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, uint64_t, Equal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, float, Equal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, double, Equal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, MLFloat16, Equal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, float, Round)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, double, Round)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, MLFloat16, Round)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 13, CumSum)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, int64_t_int64_t_int64_t, OneHot)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, int64_t_float_int64_t, OneHot)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, int32_t_float_int32_t, OneHot)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, int64_t_MLFloat16_int64_t, OneHot)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, int32_t_MLFloat16_int32_t, OneHot)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, ScatterND)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, DepthToSpace)>,
-
-    // OpSet 12
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, 12, Clip)>,
-
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, float, MaxPool)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, double, MaxPool)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, MLFloat16, MaxPool)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, int8_t, MaxPool)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, uint8_t, MaxPool)>,
-
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, 12, Pow)>,
-
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, 12, int64_t, GatherND)>,
-
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, 12, Dropout)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, Einsum)>,
-
-    // OpSet 13
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 14, Pow)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, int32_t, Add)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, int64_t, Add)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, uint32_t, Add)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, uint64_t, Add)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, float, Add)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, double, Add)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, Clip)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, MLFloat16, Add)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, int32_t, Sub)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, int64_t, Sub)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, uint32_t, Sub)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, uint64_t, Sub)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, float, Sub)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, double, Sub)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, MLFloat16, Sub)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, int32_t, Mul)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, int64_t, Mul)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, uint32_t, Mul)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, uint64_t, Mul)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, float, Mul)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, double, Mul)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, MLFloat16, Mul)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, int32_t, Div)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, int64_t, Div)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, uint32_t, Div)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, uint64_t, Div)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, float, Div)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, double, Div)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, MLFloat16, Div)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, int8_t, Abs)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, int16_t, Abs)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, int32_t, Abs)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, int64_t, Abs)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, uint8_t, Abs)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, uint16_t, Abs)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, uint32_t, Abs)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, uint64_t, Abs)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, float, Abs)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, double, Abs)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, MLFloat16, Abs)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, int8_t, Neg)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, int16_t, Neg)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, int32_t, Neg)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, int64_t, Neg)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, float, Neg)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, double, Neg)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, MLFloat16, Neg)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, BFloat16, Neg)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, float, Floor)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, double, Floor)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, MLFloat16, Floor)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, float, Ceil)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, double, Ceil)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, MLFloat16, Ceil)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, float, Reciprocal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, double, Reciprocal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, MLFloat16, Reciprocal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, float, Sqrt)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, double, Sqrt)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, MLFloat16, Sqrt)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, BFloat16, Sqrt)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, float, Log)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, double, Log)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, MLFloat16, Log)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, float, Exp)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, double, Exp)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, MLFloat16, Exp)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, BFloat16, Exp)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, float, Erf)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, double, Erf)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, MLFloat16, Erf)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, BFloat16, Erf)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, Expand)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, Sum)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, Max)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, Min)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, bool, Equal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, int32_t, Equal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, int64_t, Equal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, uint32_t, Equal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, uint64_t, Equal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, float, Equal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, double, Equal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, MLFloat16, Equal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, int32_t, Greater)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, int64_t, Greater)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, uint32_t, Greater)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, uint64_t, Greater)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, float, Greater)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, double, Greater)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, MLFloat16, Greater)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, int32_t, Less)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, int64_t, Less)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, uint32_t, Less)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, uint64_t, Less)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, float, Less)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, double, Less)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, MLFloat16, Less)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, bool, NonZero)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, uint8_t, NonZero)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, int32_t, NonZero)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, int64_t, NonZero)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, float, NonZero)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, MLFloat16, NonZero)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 18, float, Cast)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 18, double, Cast)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 18, MLFloat16, Cast)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 18, int8_t, Cast)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 18, int16_t, Cast)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 18, int32_t, Cast)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 18, int64_t, Cast)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 18, uint8_t, Cast)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 18, uint16_t, Cast)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 18, uint32_t, Cast)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 18, uint64_t, Cast)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 18, bool, Cast)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 19, IsNaN)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, Reshape)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 14, Shape)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, Size)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, Transpose)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 15, ScatterElements)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, int32_t, Slice)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, int64_t, Slice)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, float, Softmax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, double, Softmax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, MLFloat16, Softmax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, float, LogSoftmax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, double, LogSoftmax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, MLFloat16, LogSoftmax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 17, Split)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, Squeeze)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, Unsqueeze)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, Concat)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, Gather)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, GatherElements)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, float, MatMul)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, double, MatMul)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, MLFloat16, MatMul)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, float, Relu)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, double, Relu)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, MLFloat16, Relu)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, float, Sigmoid)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, double, Sigmoid)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, MLFloat16, Sigmoid)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, float, Tanh)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, double, Tanh)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, MLFloat16, Tanh)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, MLFloat16, Gemm)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, float, Gemm)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, double, Gemm)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, float, ReduceL1)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, double, ReduceL1)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, MLFloat16, ReduceL1)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, int32_t, ReduceL1)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, float, ReduceL2)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, double, ReduceL2)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, MLFloat16, ReduceL2)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, int32_t, ReduceL2)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, float, ReduceLogSum)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, double, ReduceLogSum)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, MLFloat16, ReduceLogSum)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, float, ReduceLogSumExp)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, double, ReduceLogSumExp)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, MLFloat16, ReduceLogSumExp)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, float, ReduceMean)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, double, ReduceMean)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, MLFloat16, ReduceMean)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, int32_t, ReduceMean)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, float, ReduceProd)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, double, ReduceProd)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, MLFloat16, ReduceProd)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, int32_t, ReduceProd)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, float, ReduceSum)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, double, ReduceSum)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, MLFloat16, ReduceSum)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, int32_t, ReduceSum)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, int64_t, ReduceSum)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, float, ReduceSumSquare)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, double, ReduceSumSquare)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, MLFloat16, ReduceSumSquare)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, int64_t, GatherND)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, Dropout)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 17, float, Resize)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 17, double, Resize)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 17, MLFloat16, Resize)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 17, int32_t, Resize)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 17, uint8_t, Resize)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 18, If)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 18, Loop)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, Flatten)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, float, LRN)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, double, LRN)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, MLFloat16, LRN)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, Identity)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 15, ScatterND)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 17, float, Pad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 17, double, Pad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 17, MLFloat16, Pad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 17, bool, Pad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, SpaceToDepth)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, DepthToSpace)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, int8_t, Sign)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, int16_t, Sign)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, int32_t, Sign)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, int64_t, Sign)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, uint8_t, Sign)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, uint16_t, Sign)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, uint32_t, Sign)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, uint64_t, Sign)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, float, Sign)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, double, Sign)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, MLFloat16, Sign)>,
-
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, BFloat16, Add)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, BFloat16, Sub)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, BFloat16, Mul)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, BFloat16, Div)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 18, BFloat16, Cast)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, BFloat16, Softmax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, BFloat16, MatMul)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, BFloat16, Relu)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, BFloat16, Sigmoid)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, BFloat16, Tanh)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, BFloat16, Gemm)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, BFloat16, ReduceSum)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, Mod)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 18, int8_t, QuantizeLinear)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 18, uint8_t, QuantizeLinear)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 18, int8_t, DequantizeLinear)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 18, uint8_t, DequantizeLinear)>,
-
-    // OpSet 14
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, CumSum)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, float, Relu)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, double, Relu)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, MLFloat16, Relu)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, int32_t, Add)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, int64_t, Add)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, uint32_t, Add)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, uint64_t, Add)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, float, Add)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, double, Add)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, MLFloat16, Add)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, int32_t, Sub)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, int64_t, Sub)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, uint32_t, Sub)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, uint64_t, Sub)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, float, Sub)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, double, Sub)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, MLFloat16, Sub)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, int32_t, Mul)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, int64_t, Mul)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, uint32_t, Mul)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, uint64_t, Mul)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, float, Mul)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, double, Mul)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, MLFloat16, Mul)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, int32_t, Div)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, int64_t, Div)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, uint32_t, Div)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, uint64_t, Div)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, float, Div)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, double, Div)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, MLFloat16, Div)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, 18, Identity)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, float, RNN)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, double, RNN)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, MLFloat16, RNN)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, float, GRU)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, double, GRU)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, MLFloat16, GRU)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, float, LSTM)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, double, LSTM)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, MLFloat16, LSTM)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, 18, Reshape)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(
-        kCudaExecutionProvider, kOnnxDomain, 14, 14, float, BatchNormalization)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(
-        kCudaExecutionProvider, kOnnxDomain, 14, 14, double, BatchNormalization)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(
-        kCudaExecutionProvider, kOnnxDomain, 14, 14, MLFloat16, BatchNormalization)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, BFloat16, Add)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, BFloat16, Sub)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, BFloat16, Mul)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, BFloat16, Div)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, BFloat16, Relu)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, Trilu)>,
-
-    // OpSet 15
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 15, Pow)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 15, float, BatchNormalization)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 15, double, BatchNormalization)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 15, MLFloat16, BatchNormalization)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 15, 18, Shape)>,
-
-    // Opset 16
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, float, LeakyRelu)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, double, LeakyRelu)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, MLFloat16, LeakyRelu)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, float, PRelu)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, double, PRelu)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, MLFloat16, PRelu)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, 18, Scan)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, MLFloat16, Where)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, BFloat16, Where)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, float, Where)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, double_t, Where)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, int32_t, Where)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, int64_t, Where)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, uint8_t, Where)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, int32_t, GreaterOrEqual)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, int64_t, GreaterOrEqual)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, uint32_t, GreaterOrEqual)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, uint64_t, GreaterOrEqual)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, float, GreaterOrEqual)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, double, GreaterOrEqual)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, MLFloat16, GreaterOrEqual)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, int32_t, LessOrEqual)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, int64_t, LessOrEqual)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, uint32_t, LessOrEqual)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, uint64_t, LessOrEqual)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, float, LessOrEqual)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, double, LessOrEqual)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, MLFloat16, LessOrEqual)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, 17, ScatterElements)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, 17, ScatterND)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, float, GridSample)>,
-
-    // Opset 17
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 17, float, LayerNormalization)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 17, double, LayerNormalization)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 17, BFloat16, LayerNormalization)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 17, MLFloat16, LayerNormalization)>,
-
-    // Opset 18
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, Split)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, float, ReduceMin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, double, ReduceMin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, MLFloat16, ReduceMin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, int32_t, ReduceMin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, int8_t, ReduceMin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, uint8_t, ReduceMin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, int64_t, ReduceMin)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, float, ReduceMax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, double, ReduceMax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, MLFloat16, ReduceMax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, int32_t, ReduceMax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, int64_t, ReduceMax)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, ScatterElements)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, ScatterND)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, float, Pad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, double, Pad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, MLFloat16, Pad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, bool, Pad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, float, Resize)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, double, Resize)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, MLFloat16, Resize)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, int32_t, Resize)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, uint8_t, Resize)>,
-
-    // Opset 19
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, float, Cast)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, double, Cast)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, MLFloat16, Cast)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, int8_t, Cast)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, int16_t, Cast)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, int32_t, Cast)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, int64_t, Cast)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, uint8_t, Cast)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, uint16_t, Cast)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, uint32_t, Cast)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, uint64_t, Cast)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, bool, Cast)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, BFloat16, Cast)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 4, 10, Concat)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 10, Unsqueeze)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 8, Flatten)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 10, Squeeze)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 12, Identity)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 9, Dropout)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, float, Cos)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, double, Cos)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, MLFloat16, Cos)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, float, Sin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, double, Sin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, MLFloat16, Sin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 10, Gather)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 8, float, Gemm)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 8, double, Gemm)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 8, MLFloat16, Gemm)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 10, float, Gemm)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 10, double, Gemm)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 10, MLFloat16, Gemm)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 8, float, MatMul)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 8, double, MatMul)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 8, MLFloat16, MatMul)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, float, MatMul)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, double, MatMul)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, MLFloat16, MatMul)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 10, int8_t, MatMulInteger)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 10, float, Clip)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, float, Elu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, double, Elu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, MLFloat16, Elu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, float, HardSigmoid)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, double, HardSigmoid)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, MLFloat16, HardSigmoid)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 15, float, LeakyRelu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 15, double, LeakyRelu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 15, MLFloat16, LeakyRelu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, float, Relu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, double, Relu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, MLFloat16, Relu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, float, Selu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, double, Selu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, MLFloat16, Selu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, float, Sigmoid)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, double, Sigmoid)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, MLFloat16, Sigmoid)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, float, Softsign)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, double, Softsign)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, MLFloat16, Softsign)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, float, Tanh)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, double, Tanh)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, MLFloat16, Tanh)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, float, Softplus)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, double, Softplus)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, MLFloat16, Softplus)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 10, float, Softmax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 10, double, Softmax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 10, MLFloat16, Softmax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 10, float, LogSoftmax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 10, double, LogSoftmax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 10, MLFloat16, LogSoftmax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 11, float, Pow)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 11, double, Pow)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 11, MLFloat16, Pow)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 8, float, PRelu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 8, double, PRelu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 8, MLFloat16, PRelu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 15, float, PRelu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 15, double, PRelu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 15, MLFloat16, PRelu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, bool, And)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, bool, Or)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, bool, Xor)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 7, Sum)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 8, 12, Sum)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 11, Max)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, 12, Max)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 11, Min)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, 12, Min)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 8, float, Greater)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 8, double, Greater)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 8, MLFloat16, Greater)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 10, bool, Equal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 10, int32_t, Equal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 10, int64_t, Equal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 8, 12, Expand)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, int32_t, Greater)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, int64_t, Greater)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, uint32_t, Greater)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, uint64_t, Greater)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, float, Greater)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, double, Greater)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, MLFloat16, Greater)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, 15, int32_t, GreaterOrEqual)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, 15, int64_t, GreaterOrEqual)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, 15, uint32_t, GreaterOrEqual)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, 15, uint64_t, GreaterOrEqual)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, 15, float, GreaterOrEqual)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, 15, double, GreaterOrEqual)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, 15, MLFloat16, GreaterOrEqual)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, 15, int32_t, LessOrEqual)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, 15, int64_t, LessOrEqual)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, 15, uint32_t, LessOrEqual)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, 15, uint64_t, LessOrEqual)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, 15, float, LessOrEqual)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, 15, double, LessOrEqual)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, 15, MLFloat16, LessOrEqual)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 12, int32_t, Add)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 12, int64_t, Add)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 12, uint32_t, Add)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 12, uint64_t, Add)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 12, float, Add)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 12, double, Add)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 12, MLFloat16, Add)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 12, int32_t, Sub)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 12, int64_t, Sub)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 12, uint32_t, Sub)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 12, uint64_t, Sub)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 12, float, Sub)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 12, double, Sub)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 12, MLFloat16, Sub)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 12, int32_t, Mul)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 12, int64_t, Mul)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 12, uint32_t, Mul)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 12, uint64_t, Mul)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 12, float, Mul)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 12, double, Mul)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 12, MLFloat16, Mul)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 12, int32_t, Div)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 12, int64_t, Div)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 12, uint32_t, Div)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 12, uint64_t, Div)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 12, float, Div)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 12, double, Div)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 12, MLFloat16, Div)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, int8_t, Abs)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, int16_t, Abs)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, int32_t, Abs)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, int64_t, Abs)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, uint8_t, Abs)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, uint16_t, Abs)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, uint32_t, Abs)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, uint64_t, Abs)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, float, Abs)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, double, Abs)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, MLFloat16, Abs)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, int8_t, Neg)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, int16_t, Neg)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, int32_t, Neg)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, int64_t, Neg)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, float, Neg)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, double, Neg)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, MLFloat16, Neg)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, float, Floor)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, double, Floor)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, MLFloat16, Floor)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, float, Ceil)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, double, Ceil)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, MLFloat16, Ceil)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, float, Reciprocal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, double, Reciprocal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, MLFloat16, Reciprocal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, float, Sqrt)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, double, Sqrt)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, MLFloat16, Sqrt)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, float, Log)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, double, Log)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, MLFloat16, Log)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, float, Exp)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, double, Exp)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, MLFloat16, Exp)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, float, Erf)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, double, Erf)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, MLFloat16, Erf)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, IsNaN)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, bool, Not)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 8, float, BatchNormalization)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 8, double, BatchNormalization)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 8, MLFloat16, BatchNormalization)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 13, float, BatchNormalization)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 13, double, BatchNormalization)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 13, MLFloat16, BatchNormalization)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 12, float, LRN)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 12, double, LRN)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 12, MLFloat16, LRN)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 10, float, Conv)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 10, double, Conv)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 10, MLFloat16, Conv)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 10, float, ConvTranspose)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 10, double, ConvTranspose)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 10, MLFloat16, ConvTranspose)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 9, float, AveragePool)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 9, double, AveragePool)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 9, MLFloat16, AveragePool)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, float, GlobalAveragePool)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, double, GlobalAveragePool)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, MLFloat16, GlobalAveragePool)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 7, float, MaxPool)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 7, double, MaxPool)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 7, MLFloat16, MaxPool)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 8, 9, float, MaxPool)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 8, 9, double, MaxPool)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 8, 9, MLFloat16, MaxPool)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, float, GlobalMaxPool)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, double, GlobalMaxPool)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, MLFloat16, GlobalMaxPool)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 11, float, ArgMin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 11, double, ArgMin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 11, MLFloat16, ArgMin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, float, ReduceL1)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, double, ReduceL1)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, MLFloat16, ReduceL1)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, int32_t, ReduceL1)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, float, ReduceL2)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, double, ReduceL2)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, MLFloat16, ReduceL2)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, int32_t, ReduceL2)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, float, ReduceMax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, double, ReduceMax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, MLFloat16, ReduceMax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, int32_t, ReduceMax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, int64_t, ReduceMax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, float, ReduceMean)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, double, ReduceMean)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, MLFloat16, ReduceMean)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, int32_t, ReduceMean)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, float, ReduceMin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, double, ReduceMin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, MLFloat16, ReduceMin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, int32_t, ReduceMin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, int64_t, ReduceMin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, int8_t, ReduceMin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, uint8_t, ReduceMin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, float, ReduceProd)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, double, ReduceProd)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, MLFloat16, ReduceProd)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, int32_t, ReduceProd)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 12, float, ReduceSum)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 12, double, ReduceSum)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 12, MLFloat16, ReduceSum)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 12, int32_t, ReduceSum)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 12, int64_t, ReduceSum)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, float, ReduceLogSum)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, double, ReduceLogSum)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, MLFloat16, ReduceLogSum)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, float, ReduceSumSquare)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, double, ReduceSumSquare)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, MLFloat16, ReduceSumSquare)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, float, ReduceLogSumExp)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, double, ReduceLogSumExp)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 17, MLFloat16, ReduceLogSumExp)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 8, float, Cast)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 8, double, Cast)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 8, MLFloat16, Cast)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 8, int8_t, Cast)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 8, int16_t, Cast)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 8, int32_t, Cast)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 8, int64_t, Cast)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 8, uint8_t, Cast)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 8, uint16_t, Cast)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 8, uint32_t, Cast)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 8, uint64_t, Cast)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 8, bool, Cast)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, float, Cast)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, double, Cast)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, MLFloat16, Cast)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, int8_t, Cast)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, int16_t, Cast)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, int32_t, Cast)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, int64_t, Cast)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, uint8_t, Cast)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, uint16_t, Cast)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, uint32_t, Cast)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, uint64_t, Cast)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, bool, Cast)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 2, 10, float, Pad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 2, 10, double, Pad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 2, 10, MLFloat16, Pad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 4, Reshape)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 5, 12, Reshape)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 12, Shape)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 12, Size)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, Tile)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, Tile)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 12, Transpose)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, float, InstanceNormalization)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, double, InstanceNormalization)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, MLFloat16, InstanceNormalization)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 13, float, RNN)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 13, double, RNN)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 13, MLFloat16, RNN)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 13, float, GRU)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 13, double, GRU)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 13, MLFloat16, GRU)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 13, float, LSTM)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 13, double, LSTM)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 13, MLFloat16, LSTM)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 9, int64_t, Slice)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 10, Compress)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 10, Flatten)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 8, float, Upsample)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 8, double, Upsample)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 8, MLFloat16, Upsample)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 8, int32_t, Upsample)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 8, uint8_t, Upsample)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 9, float, Upsample)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 9, double, Upsample)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 9, MLFloat16, Upsample)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 9, int32_t, Upsample)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 9, uint8_t, Upsample)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 2, 10, Split)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, ConstantOfShape)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, int8_t, Shrink)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, int16_t, Shrink)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, int32_t, Shrink)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, int64_t, Shrink)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, uint8_t, Shrink)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, uint16_t, Shrink)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, uint32_t, Shrink)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, uint64_t, Shrink)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, float, Shrink)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, double, Shrink)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, MLFloat16, Shrink)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 8, float, Less)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 8, double, Less)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 8, MLFloat16, Less)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, int32_t, Less)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, int64_t, Less)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, uint32_t, Less)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, uint64_t, Less)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, float, Less)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, double, Less)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, MLFloat16, Less)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, EyeLike)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 10, Scatter)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 15, MLFloat16, Where)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 15, float, Where)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 15, double_t, Where)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 15, int32_t, Where)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 15, int64_t, Where)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 15, uint8_t, Where)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, bool, NonZero)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, uint8_t, NonZero)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, int32_t, NonZero)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, int64_t, NonZero)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, float, NonZero)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, MLFloat16, NonZero)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 9, TopK)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 8, 8, Scan)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 10, Scan)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 10, Loop)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 10, DepthToSpace)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 12, SpaceToDepth)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, RandomNormal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, RandomNormalLike)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, RandomUniform)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, RandomUniformLike)>,
+
+      // opset 10
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 10, 10, float, AveragePool)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 10, 10, double, AveragePool)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 10, 10, MLFloat16, AveragePool)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 10, 11, Dropout)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 10, 10, float, MaxPool)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 10, 10, double, MaxPool)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 10, 10, MLFloat16, MaxPool)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 10, 10, NonMaxSuppression)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 10, 10, float, Resize)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 10, 10, double, Resize)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 10, 10, MLFloat16, Resize)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 10, 10, int32_t, Resize)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 10, 10, uint8_t, Resize)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 10, ReverseSequence)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 10, float, RoiAlign)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 10, double, RoiAlign)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 10, 10, int32_t, Slice)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 10, 10, int64_t, Slice)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 10, float, ThresholdedRelu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 10, double, ThresholdedRelu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 10, MLFloat16, ThresholdedRelu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 10, 10, TopK)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 10, If)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 10, 12, int8_t, QuantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 10, 12, uint8_t, QuantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 10, 12, int8_t, DequantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 10, 12, uint8_t, DequantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 10, 12, Mod)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 10,
+                                                                      19, IsInf)>,
+
+      // opset 11
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 11, float, ArgMax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 11, double, ArgMax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 1, 11, MLFloat16, ArgMax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, Compress)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, Concat)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, Flatten)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, Gather)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, GatherElements)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 11, int64_t, GatherND)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, MLFloat16, Gemm)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, float, Gemm)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, double, Gemm)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, If)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, Loop)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, NonMaxSuppression)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, Range)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 15, Scan)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, ScatterElements)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, int32_t, Slice)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, int64_t, Slice)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, float, Softmax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, double, Softmax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, MLFloat16, Softmax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, float, LogSoftmax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, double, LogSoftmax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, MLFloat16, LogSoftmax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, Split)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, Squeeze)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, TopK)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, SequenceAt)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, SequenceConstruct)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, SequenceEmpty)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, SequenceLength)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, ConcatFromSequence)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, SequenceErase)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, SequenceInsert)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, Unsqueeze)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, float, Conv)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, double, Conv)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, MLFloat16, Conv)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, float, ConvTranspose)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, double, ConvTranspose)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, MLFloat16, ConvTranspose)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, float, AveragePool)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, double, AveragePool)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, MLFloat16, AveragePool)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 11, float, MaxPool)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 11, double, MaxPool)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 11, MLFloat16, MaxPool)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, float, Resize)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, double, Resize)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, MLFloat16, Resize)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, int32_t, Resize)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, uint8_t, Resize)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 11, Clip)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, float, Pad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, double, Pad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, MLFloat16, Pad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, bool, Equal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, int32_t, Equal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, int64_t, Equal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, uint32_t, Equal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, uint64_t, Equal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, float, Equal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, double, Equal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, MLFloat16, Equal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, float, Round)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, double, Round)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, MLFloat16, Round)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 13, CumSum)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, int64_t_int64_t_int64_t, OneHot)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, int64_t_float_int64_t, OneHot)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, int32_t_float_int32_t, OneHot)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, int64_t_MLFloat16_int64_t, OneHot)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, int32_t_MLFloat16_int32_t, OneHot)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, ScatterND)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, DepthToSpace)>,
+
+      // OpSet 12
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, 12, Clip)>,
+
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, float, MaxPool)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, double, MaxPool)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, MLFloat16, MaxPool)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, int8_t, MaxPool)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, uint8_t, MaxPool)>,
+
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, 12, Pow)>,
+
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, 12, int64_t, GatherND)>,
+
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, 12, Dropout)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, Einsum)>,
+
+      // OpSet 13
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 14, Pow)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, int32_t, Add)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, int64_t, Add)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, uint32_t, Add)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, uint64_t, Add)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, float, Add)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, double, Add)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, Clip)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, MLFloat16, Add)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, int32_t, Sub)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, int64_t, Sub)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, uint32_t, Sub)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, uint64_t, Sub)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, float, Sub)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, double, Sub)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, MLFloat16, Sub)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, int32_t, Mul)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, int64_t, Mul)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, uint32_t, Mul)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, uint64_t, Mul)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, float, Mul)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, double, Mul)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, MLFloat16, Mul)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, int32_t, Div)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, int64_t, Div)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, uint32_t, Div)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, uint64_t, Div)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, float, Div)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, double, Div)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, MLFloat16, Div)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, int8_t, Abs)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, int16_t, Abs)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, int32_t, Abs)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, int64_t, Abs)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, uint8_t, Abs)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, uint16_t, Abs)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, uint32_t, Abs)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, uint64_t, Abs)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, float, Abs)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, double, Abs)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, MLFloat16, Abs)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, int8_t, Neg)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, int16_t, Neg)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, int32_t, Neg)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, int64_t, Neg)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, float, Neg)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, double, Neg)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, MLFloat16, Neg)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, BFloat16, Neg)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, float, Floor)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, double, Floor)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, MLFloat16, Floor)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, float, Ceil)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, double, Ceil)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, MLFloat16, Ceil)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, float, Reciprocal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, double, Reciprocal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, MLFloat16, Reciprocal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, float, Sqrt)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, double, Sqrt)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, MLFloat16, Sqrt)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, BFloat16, Sqrt)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, float, Log)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, double, Log)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, MLFloat16, Log)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, float, Exp)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, double, Exp)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, MLFloat16, Exp)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, BFloat16, Exp)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, float, Erf)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, double, Erf)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, MLFloat16, Erf)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, BFloat16, Erf)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, Expand)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, Sum)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, Max)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, Min)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, bool, Equal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, int32_t, Equal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, int64_t, Equal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, uint32_t, Equal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, uint64_t, Equal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, float, Equal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, double, Equal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, MLFloat16, Equal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, int32_t, Greater)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, int64_t, Greater)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, uint32_t, Greater)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, uint64_t, Greater)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, float, Greater)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, double, Greater)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, MLFloat16, Greater)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, int32_t, Less)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, int64_t, Less)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, uint32_t, Less)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, uint64_t, Less)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, float, Less)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, double, Less)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, MLFloat16, Less)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, bool, NonZero)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, uint8_t, NonZero)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, int32_t, NonZero)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, int64_t, NonZero)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, float, NonZero)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, MLFloat16, NonZero)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 18, float, Cast)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 18, double, Cast)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 18, MLFloat16, Cast)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 18, int8_t, Cast)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 18, int16_t, Cast)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 18, int32_t, Cast)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 18, int64_t, Cast)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 18, uint8_t, Cast)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 18, uint16_t, Cast)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 18, uint32_t, Cast)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 18, uint64_t, Cast)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 18, bool, Cast)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 19, IsNaN)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, Reshape)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 14, Shape)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, Size)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, Transpose)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 15, ScatterElements)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, int32_t, Slice)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, int64_t, Slice)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, float, Softmax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, double, Softmax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, MLFloat16, Softmax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, float, LogSoftmax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, double, LogSoftmax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, MLFloat16, LogSoftmax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 17, Split)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, Squeeze)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, Unsqueeze)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, Concat)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, Gather)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, GatherElements)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, float, MatMul)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, double, MatMul)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, MLFloat16, MatMul)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, float, Relu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, double, Relu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, MLFloat16, Relu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, float, Sigmoid)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, double, Sigmoid)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, MLFloat16, Sigmoid)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, float, Tanh)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, double, Tanh)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, MLFloat16, Tanh)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, MLFloat16, Gemm)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, float, Gemm)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, double, Gemm)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, float, ReduceL1)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, double, ReduceL1)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, MLFloat16, ReduceL1)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, int32_t, ReduceL1)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, float, ReduceL2)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, double, ReduceL2)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, MLFloat16, ReduceL2)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, int32_t, ReduceL2)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, float, ReduceLogSum)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, double, ReduceLogSum)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, MLFloat16, ReduceLogSum)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, float, ReduceLogSumExp)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, double, ReduceLogSumExp)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, MLFloat16, ReduceLogSumExp)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, float, ReduceMean)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, double, ReduceMean)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, MLFloat16, ReduceMean)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, int32_t, ReduceMean)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, float, ReduceProd)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, double, ReduceProd)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, MLFloat16, ReduceProd)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, int32_t, ReduceProd)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, float, ReduceSum)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, double, ReduceSum)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, MLFloat16, ReduceSum)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, int32_t, ReduceSum)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, int64_t, ReduceSum)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, float, ReduceSumSquare)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, double, ReduceSumSquare)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, MLFloat16, ReduceSumSquare)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, int64_t, GatherND)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, Dropout)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 17, float, Resize)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 17, double, Resize)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 17, MLFloat16, Resize)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 17, int32_t, Resize)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 17, uint8_t, Resize)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 18, If)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 18, Loop)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, Flatten)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, float, LRN)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, double, LRN)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, MLFloat16, LRN)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, Identity)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 15, ScatterND)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 17, float, Pad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 17, double, Pad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 17, MLFloat16, Pad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 17, bool, Pad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, SpaceToDepth)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, DepthToSpace)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, int8_t, Sign)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, int16_t, Sign)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, int32_t, Sign)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, int64_t, Sign)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, uint8_t, Sign)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, uint16_t, Sign)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, uint32_t, Sign)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, uint64_t, Sign)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, float, Sign)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, double, Sign)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, MLFloat16, Sign)>,
+
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, BFloat16, Add)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, BFloat16, Sub)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, BFloat16, Mul)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, BFloat16, Div)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 18, BFloat16, Cast)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, BFloat16, Softmax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, BFloat16, MatMul)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 13, BFloat16, Relu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, BFloat16, Sigmoid)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, BFloat16, Tanh)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, BFloat16, Gemm)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, BFloat16, ReduceSum)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, Mod)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 18, int8_t, QuantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 18, uint8_t, QuantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 18, int8_t, DequantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, 18, uint8_t, DequantizeLinear)>,
+
+      // OpSet 14
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, CumSum)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, float, Relu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, double, Relu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, MLFloat16, Relu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, int32_t, Add)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, int64_t, Add)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, uint32_t, Add)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, uint64_t, Add)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, float, Add)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, double, Add)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, MLFloat16, Add)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, int32_t, Sub)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, int64_t, Sub)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, uint32_t, Sub)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, uint64_t, Sub)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, float, Sub)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, double, Sub)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, MLFloat16, Sub)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, int32_t, Mul)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, int64_t, Mul)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, uint32_t, Mul)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, uint64_t, Mul)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, float, Mul)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, double, Mul)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, MLFloat16, Mul)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, int32_t, Div)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, int64_t, Div)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, uint32_t, Div)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, uint64_t, Div)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, float, Div)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, double, Div)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, MLFloat16, Div)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, 18, Identity)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, float, RNN)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, double, RNN)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, MLFloat16, RNN)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, float, GRU)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, double, GRU)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, MLFloat16, GRU)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, float, LSTM)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, double, LSTM)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, MLFloat16, LSTM)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, 18, Reshape)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(
+          kCudaExecutionProvider, kOnnxDomain, 14, 14, float, BatchNormalization)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(
+          kCudaExecutionProvider, kOnnxDomain, 14, 14, double, BatchNormalization)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(
+          kCudaExecutionProvider, kOnnxDomain, 14, 14, MLFloat16, BatchNormalization)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, BFloat16, Add)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, BFloat16, Sub)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, BFloat16, Mul)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, BFloat16, Div)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, BFloat16, Relu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 14, Trilu)>,
+
+      // OpSet 15
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 15, Pow)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 15, float, BatchNormalization)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 15, double, BatchNormalization)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 15, MLFloat16, BatchNormalization)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 15, 18, Shape)>,
+
+      // Opset 16
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, float, LeakyRelu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, double, LeakyRelu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, MLFloat16, LeakyRelu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, float, PRelu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, double, PRelu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, MLFloat16, PRelu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, 18, Scan)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, MLFloat16, Where)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, BFloat16, Where)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, float, Where)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, double_t, Where)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, int32_t, Where)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, int64_t, Where)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, uint8_t, Where)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, int32_t, GreaterOrEqual)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, int64_t, GreaterOrEqual)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, uint32_t, GreaterOrEqual)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, uint64_t, GreaterOrEqual)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, float, GreaterOrEqual)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, double, GreaterOrEqual)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, MLFloat16, GreaterOrEqual)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, int32_t, LessOrEqual)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, int64_t, LessOrEqual)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, uint32_t, LessOrEqual)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, uint64_t, LessOrEqual)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, float, LessOrEqual)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, double, LessOrEqual)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, MLFloat16, LessOrEqual)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, 17, ScatterElements)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, 17, ScatterND)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, float, GridSample)>,
+
+      // Opset 17
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 17, float, LayerNormalization)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 17, double, LayerNormalization)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 17, BFloat16, LayerNormalization)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 17, MLFloat16, LayerNormalization)>,
+
+      // Opset 18
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, Split)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, float, ReduceMin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, double, ReduceMin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, MLFloat16, ReduceMin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, int32_t, ReduceMin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, int8_t, ReduceMin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, uint8_t, ReduceMin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, int64_t, ReduceMin)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, float, ReduceMax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, double, ReduceMax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, MLFloat16, ReduceMax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, int32_t, ReduceMax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, int64_t, ReduceMax)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, ScatterElements)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, ScatterND)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, float, Pad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, double, Pad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, MLFloat16, Pad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, bool, Pad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, float, Resize)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, double, Resize)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, MLFloat16, Resize)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, int32_t, Resize)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 18, uint8_t, Resize)>,
+
+      // Opset 19
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, float, Cast)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, double, Cast)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, MLFloat16, Cast)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, int8_t, Cast)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, int16_t, Cast)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, int32_t, Cast)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, int64_t, Cast)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, uint8_t, Cast)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, uint16_t, Cast)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, uint32_t, Cast)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, uint64_t, Cast)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, bool, Cast)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, BFloat16, Cast)>,
 #if !defined(DISABLE_FLOAT8_TYPES)
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, Float8E4M3FN, Cast)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, Float8E5M2, Cast)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, Float8E4M3FN, Cast)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, Float8E5M2, Cast)>,
 #endif
 
-    BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, uint8_t, float, DequantizeLinear)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, int8_t, float, DequantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, uint8_t, float, DequantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, int8_t, float, DequantizeLinear)>,
 #if !defined(DISABLE_FLOAT8_TYPES)
-    BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, Float8E4M3FN, float, DequantizeLinear)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, Float8E5M2, float, DequantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, Float8E4M3FN, float, DequantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, Float8E5M2, float, DequantizeLinear)>,
 #endif
-    BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, uint8_t, MLFloat16, DequantizeLinear)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, int8_t, MLFloat16, DequantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, uint8_t, MLFloat16, DequantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, int8_t, MLFloat16, DequantizeLinear)>,
 #if !defined(DISABLE_FLOAT8_TYPES)
-    BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, Float8E4M3FN, MLFloat16, DequantizeLinear)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, Float8E5M2, MLFloat16, DequantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, Float8E4M3FN, MLFloat16, DequantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, Float8E5M2, MLFloat16, DequantizeLinear)>,
 #endif
 
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, Identity)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, If)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, Loop)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, Identity)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, If)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, Loop)>,
 
-    BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, uint8_t, float, QuantizeLinear)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, int8_t, float, QuantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, uint8_t, float, QuantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, int8_t, float, QuantizeLinear)>,
 #if !defined(DISABLE_FLOAT8_TYPES)
-    BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, Float8E4M3FN, float, QuantizeLinear)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, Float8E5M2, float, QuantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, Float8E4M3FN, float, QuantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, Float8E5M2, float, QuantizeLinear)>,
 #endif
-    BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, uint8_t, MLFloat16, QuantizeLinear)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, int8_t, MLFloat16, QuantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, uint8_t, MLFloat16, QuantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, int8_t, MLFloat16, QuantizeLinear)>,
 #if !defined(DISABLE_FLOAT8_TYPES)
-    BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, Float8E4M3FN, MLFloat16, QuantizeLinear)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, Float8E5M2, MLFloat16, QuantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, Float8E4M3FN, MLFloat16, QuantizeLinear)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, Float8E5M2, MLFloat16, QuantizeLinear)>,
 #endif
 
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, Reshape)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, Scan)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, Shape)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, Reshape)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, Scan)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 19, Shape)>,
 
-    // Opset 20
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 20, float, Gelu)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 20, double, Gelu)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 20, MLFloat16, Gelu)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 20, IsInf)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 20, IsNaN)>,
+      // Opset 20
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 20, float, Gelu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 20, double, Gelu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 20, MLFloat16, Gelu)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 20, IsInf)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 20, IsNaN)>,
 #endif
   };
 
diff --git a/onnxruntime/core/providers/cuda/cuda_graph.h b/onnxruntime/core/providers/cuda/cuda_graph.h
index 064994c1f14ae..dd03db94b631c 100644
--- a/onnxruntime/core/providers/cuda/cuda_graph.h
+++ b/onnxruntime/core/providers/cuda/cuda_graph.h
@@ -18,7 +18,7 @@ constexpr CudaGraphAnnotation_t kCudaGraphAnnotationSkip = -1;
 constexpr CudaGraphAnnotation_t kCudaGraphAnnotationDefault = 0;
 
 struct CudaGraphSet {
-  CudaGraphSet(){};
+  CudaGraphSet() {};
   ~CudaGraphSet();
 
   void Clear();
@@ -31,7 +31,7 @@ struct CudaGraphSet {
 };
 
 struct CUDAGraphManager {
-  CUDAGraphManager(){};
+  CUDAGraphManager() {};
   CUDAGraphManager(cudaStream_t stream);
   ~CUDAGraphManager();
 
diff --git a/onnxruntime/core/providers/cuda/cuda_profiler.h b/onnxruntime/core/providers/cuda/cuda_profiler.h
index 88c9adc5e17b3..4930e55351615 100644
--- a/onnxruntime/core/providers/cuda/cuda_profiler.h
+++ b/onnxruntime/core/providers/cuda/cuda_profiler.h
@@ -33,7 +33,7 @@ class CudaProfiler final : public EpProfiler {
   ORT_DISALLOW_COPY_ASSIGNMENT_AND_MOVE(CudaProfiler);
   ~CudaProfiler() {}
   bool StartProfiling(TimePoint) override { return true; }
-  void EndProfiling(TimePoint, Events&) override{};
+  void EndProfiling(TimePoint, Events&) override {};
   void Start(uint64_t) override{};
   void Stop(uint64_t) override{};
 };
diff --git a/onnxruntime/core/providers/cuda/nn/conv_transpose.h b/onnxruntime/core/providers/cuda/nn/conv_transpose.h
index 77c9d94162b6b..71ad3ee6e2147 100644
--- a/onnxruntime/core/providers/cuda/nn/conv_transpose.h
+++ b/onnxruntime/core/providers/cuda/nn/conv_transpose.h
@@ -18,7 +18,7 @@ namespace cuda {
 template <typename T, bool NHWC>
 class ConvTranspose : public CudaKernel {
  public:
-  ConvTranspose(const OpKernelInfo& info) : CudaKernel(info), conv_transpose_attrs_(info){};
+  ConvTranspose(const OpKernelInfo& info) : CudaKernel(info), conv_transpose_attrs_(info) {};
   Status PrePack(const Tensor& tensor, int input_idx, AllocatorPtr alloc,
                  bool& is_packed, [[maybe_unused]] PrePackedWeights* prepacked_weights) override;
   Status ComputeInternal(OpKernelContext* context) const override;
diff --git a/onnxruntime/core/providers/cuda/nvtx_profile.h b/onnxruntime/core/providers/cuda/nvtx_profile.h
index f98745cbfc5c2..e545578a72fc4 100644
--- a/onnxruntime/core/providers/cuda/nvtx_profile.h
+++ b/onnxruntime/core/providers/cuda/nvtx_profile.h
@@ -45,7 +45,7 @@ enum class Color : uint32_t {
 class RangeCreatorBase {
  public:
   RangeCreatorBase(const std::string message, const Color color)
-      : message_(message), color_(color), is_begin_called_(false), is_end_called_(false){};
+      : message_(message), color_(color), is_begin_called_(false), is_end_called_(false) {};
 
   // Check if Begin and End are both called.
   // It's pointless if not all of them are called.
@@ -100,7 +100,7 @@ class RangeCreatorBase {
 class NvtxRangeCreator final : public RangeCreatorBase {
  public:
   NvtxRangeCreator(const std::string message, const Color color)
-      : RangeCreatorBase(message, color){};
+      : RangeCreatorBase(message, color) {};
 
   void BeginImpl() override;
   void EndImpl() override;
@@ -114,7 +114,7 @@ class NvtxRangeCreator final : public RangeCreatorBase {
 class NvtxNestedRangeCreator final : public RangeCreatorBase {
  public:
   NvtxNestedRangeCreator(const std::string message, const Color color)
-      : RangeCreatorBase(message, color){};
+      : RangeCreatorBase(message, color) {};
 
   void BeginImpl() override;
   void EndImpl() override;
@@ -123,7 +123,7 @@ class NvtxNestedRangeCreator final : public RangeCreatorBase {
 class NvtxMarkerCreator final {
  public:
   NvtxMarkerCreator(const std::string message, const Color color)
-      : message_(message), color_(color){};
+      : message_(message), color_(color) {};
   void Mark();
 
  private:
diff --git a/onnxruntime/core/providers/cuda/shared_inc/cuda_utils.h b/onnxruntime/core/providers/cuda/shared_inc/cuda_utils.h
index 1f7df9b6fc2e3..ed642754af3ba 100644
--- a/onnxruntime/core/providers/cuda/shared_inc/cuda_utils.h
+++ b/onnxruntime/core/providers/cuda/shared_inc/cuda_utils.h
@@ -35,7 +35,7 @@ enum class BroadcastIndexType : int32_t {
 template <typename T>
 class IConstantBuffer {
  public:
-  virtual ~IConstantBuffer(){};
+  virtual ~IConstantBuffer() {};
   virtual const T* GetBuffer(cudaStream_t stream, size_t count) = 0;
 };
 
diff --git a/onnxruntime/core/providers/cuda/tensor/cast_op.cc b/onnxruntime/core/providers/cuda/tensor/cast_op.cc
index 8e5a68e2a278e..821695bbbd42f 100644
--- a/onnxruntime/core/providers/cuda/tensor/cast_op.cc
+++ b/onnxruntime/core/providers/cuda/tensor/cast_op.cc
@@ -13,23 +13,23 @@ const std::vector<MLDataType>& CastOpTypeConstraints() {
   // Must be done as a local static for a shared provider, to avoid the prefast warning:
   // Global initializer calls a non-constexpr function 'onnxruntime::DataTypeImpl::GetTensorType<onnxruntime::MLFloat16>'
   // In a shared provider, GetTensorType is a function call into Onnxruntime and isn't constexpr
-  static std::vector<MLDataType> types {
-    DataTypeImpl::GetTensorType<MLFloat16>(),
-        DataTypeImpl::GetTensorType<BFloat16>(),
-        DataTypeImpl::GetTensorType<float>(),
-        DataTypeImpl::GetTensorType<double>(),
-        DataTypeImpl::GetTensorType<int8_t>(),
-        DataTypeImpl::GetTensorType<int16_t>(),
-        DataTypeImpl::GetTensorType<int32_t>(),
-        DataTypeImpl::GetTensorType<int64_t>(),
-        DataTypeImpl::GetTensorType<uint8_t>(),
-        DataTypeImpl::GetTensorType<uint16_t>(),
-        DataTypeImpl::GetTensorType<uint32_t>(),
-        DataTypeImpl::GetTensorType<uint64_t>(),
-        DataTypeImpl::GetTensorType<bool>()
+  static std::vector<MLDataType> types{
+      DataTypeImpl::GetTensorType<MLFloat16>(),
+      DataTypeImpl::GetTensorType<BFloat16>(),
+      DataTypeImpl::GetTensorType<float>(),
+      DataTypeImpl::GetTensorType<double>(),
+      DataTypeImpl::GetTensorType<int8_t>(),
+      DataTypeImpl::GetTensorType<int16_t>(),
+      DataTypeImpl::GetTensorType<int32_t>(),
+      DataTypeImpl::GetTensorType<int64_t>(),
+      DataTypeImpl::GetTensorType<uint8_t>(),
+      DataTypeImpl::GetTensorType<uint16_t>(),
+      DataTypeImpl::GetTensorType<uint32_t>(),
+      DataTypeImpl::GetTensorType<uint64_t>(),
+      DataTypeImpl::GetTensorType<bool>()
 #if !defined(DISABLE_FLOAT8_TYPES)
-            ,
-        DataTypeImpl::GetTensorType<Float8E4M3FN>(), DataTypeImpl::GetTensorType<Float8E5M2>()
+          ,
+      DataTypeImpl::GetTensorType<Float8E4M3FN>(), DataTypeImpl::GetTensorType<Float8E5M2>()
 #endif
   };
   return types;
diff --git a/onnxruntime/core/providers/dnnl/dnnl_node_capability.h b/onnxruntime/core/providers/dnnl/dnnl_node_capability.h
index 3ed3705f6d81b..f67b70616547c 100644
--- a/onnxruntime/core/providers/dnnl/dnnl_node_capability.h
+++ b/onnxruntime/core/providers/dnnl/dnnl_node_capability.h
@@ -42,7 +42,7 @@ enum ORT_DataType : int {
  */
 class DnnlNodeCapability {
  public:
-  virtual ~DnnlNodeCapability(){};
+  virtual ~DnnlNodeCapability() {};
   /**
    * virtual function expected to be implemented for different node
    * types.
diff --git a/onnxruntime/core/providers/dnnl/subgraph/dnnl_subgraph.h b/onnxruntime/core/providers/dnnl/subgraph/dnnl_subgraph.h
index ceac2a6f58b32..add9f440df91f 100644
--- a/onnxruntime/core/providers/dnnl/subgraph/dnnl_subgraph.h
+++ b/onnxruntime/core/providers/dnnl/subgraph/dnnl_subgraph.h
@@ -18,7 +18,7 @@ class DnnlNode;
 class DnnlNodeArg {
  public:
   DnnlNodeArg(DnnlNode* node, size_t index, bool is_output)
-      : node_(node), index_(index), is_output_(is_output){};
+      : node_(node), index_(index), is_output_(is_output) {};
   DnnlNodeArg() = default;
   DnnlNode* GetNode() { return node_; };
   size_t GetIndex() { return index_; };
diff --git a/onnxruntime/core/providers/js/allocator.h b/onnxruntime/core/providers/js/allocator.h
index 6aa8313c01f38..267015b2ea58d 100644
--- a/onnxruntime/core/providers/js/allocator.h
+++ b/onnxruntime/core/providers/js/allocator.h
@@ -15,7 +15,7 @@ class JsCPUAllocator : public CPUAllocator {
       : CPUAllocator(
             OrtMemoryInfo("JsCPUAllocator", OrtAllocatorType::OrtDeviceAllocator,
                           OrtDevice(OrtDevice::CPU, OrtDevice::MemType::DEFAULT, 0),
-                          0, OrtMemTypeCPU)){};
+                          0, OrtMemTypeCPU)) {};
 };
 
 class JsCustomAllocator : public IAllocator {
diff --git a/onnxruntime/core/providers/js/data_transfer.h b/onnxruntime/core/providers/js/data_transfer.h
index 3dfb19cfde5ac..6a0e8586776a2 100644
--- a/onnxruntime/core/providers/js/data_transfer.h
+++ b/onnxruntime/core/providers/js/data_transfer.h
@@ -11,8 +11,8 @@ namespace js {
 
 class DataTransfer : public IDataTransfer {
  public:
-  DataTransfer(){};
-  ~DataTransfer(){};
+  DataTransfer() {};
+  ~DataTransfer() {};
 
   bool CanCopy(const OrtDevice& src_device, const OrtDevice& dst_device) const override;
 
diff --git a/onnxruntime/core/providers/qnn/builder/opbuilder/expand_op_builder.cc b/onnxruntime/core/providers/qnn/builder/opbuilder/expand_op_builder.cc
index 9e31cf9cae21a..d0f6ce9effd9e 100644
--- a/onnxruntime/core/providers/qnn/builder/opbuilder/expand_op_builder.cc
+++ b/onnxruntime/core/providers/qnn/builder/opbuilder/expand_op_builder.cc
@@ -125,7 +125,7 @@ Status ExpandOpBuilder::ProcessInputs(QnnModelWrapper& qnn_model_wrapper,
       default:
         return ORT_MAKE_STATUS(ONNXRUNTIME, FAIL, "Type not supported.");
     }  // switch
-  }    // if-else
+  }  // if-else
 
   const std::string& output_name = node_unit.Outputs()[0].node_arg.Name();
   std::string shape_input_name(input_name + "_" + output_name);
diff --git a/onnxruntime/core/providers/qnn/builder/opbuilder/pad_op_builder.cc b/onnxruntime/core/providers/qnn/builder/opbuilder/pad_op_builder.cc
index b7455314578de..5fc6d42a8a179 100644
--- a/onnxruntime/core/providers/qnn/builder/opbuilder/pad_op_builder.cc
+++ b/onnxruntime/core/providers/qnn/builder/opbuilder/pad_op_builder.cc
@@ -163,7 +163,7 @@ Status ProcessConstantValue(QnnModelWrapper& qnn_model_wrapper,
       default:
         return ORT_MAKE_STATUS(ONNXRUNTIME, FAIL, "Type not supported.");
     }  // switch
-  }    // if-else
+  }  // if-else
 
   QnnParamWrapper constant_value_param(node_unit.Index(),
                                        node_unit.Name(),
diff --git a/onnxruntime/core/providers/qnn/builder/qnn_quant_params_wrapper.cc b/onnxruntime/core/providers/qnn/builder/qnn_quant_params_wrapper.cc
index da2d517f65697..5fc4fb3db4122 100644
--- a/onnxruntime/core/providers/qnn/builder/qnn_quant_params_wrapper.cc
+++ b/onnxruntime/core/providers/qnn/builder/qnn_quant_params_wrapper.cc
@@ -10,7 +10,7 @@
 #include "core/providers/qnn/builder/qnn_model_wrapper.h"
 
 #define ALIGN_PTR_UP(ptr, align, type) \
-  reinterpret_cast<type>((reinterpret_cast<std::uintptr_t>(ptr) + (align)-1) & ~((align)-1))
+  reinterpret_cast<type>((reinterpret_cast<std::uintptr_t>(ptr) + (align) - 1) & ~((align) - 1))
 
 namespace onnxruntime {
 namespace qnn {
diff --git a/onnxruntime/core/providers/rocm/rocm_profiler.h b/onnxruntime/core/providers/rocm/rocm_profiler.h
index 070cca570f481..d5c7e3f273565 100644
--- a/onnxruntime/core/providers/rocm/rocm_profiler.h
+++ b/onnxruntime/core/providers/rocm/rocm_profiler.h
@@ -34,7 +34,7 @@ class RocmProfiler final : public EpProfiler {
   ORT_DISALLOW_COPY_ASSIGNMENT_AND_MOVE(RocmProfiler);
   ~RocmProfiler() {}
   bool StartProfiling(TimePoint) override { return true; }
-  void EndProfiling(TimePoint, Events&) override{};
+  void EndProfiling(TimePoint, Events&) override {};
   void Start(uint64_t) override{};
   void Stop(uint64_t) override{};
 };
diff --git a/onnxruntime/core/providers/shared_library/provider_host_api.h b/onnxruntime/core/providers/shared_library/provider_host_api.h
index 43d661344d787..e25426b5124dd 100644
--- a/onnxruntime/core/providers/shared_library/provider_host_api.h
+++ b/onnxruntime/core/providers/shared_library/provider_host_api.h
@@ -24,10 +24,10 @@ struct Provider {
   virtual ProviderOptions GetProviderOptions(const void* /*provider options struct*/) { return {}; }
 
   // Update provider options from key-value string configuration
-  virtual void UpdateProviderOptions(void* /*provider options to be configured*/, const ProviderOptions& /*key-value string provider options*/){};
+  virtual void UpdateProviderOptions(void* /*provider options to be configured*/, const ProviderOptions& /*key-value string provider options*/) {};
 
   // Get provider specific custom op domain list. Provider has the resposibility to release OrtCustomOpDomain instances it creates.
-  virtual void GetCustomOpDomainList(IExecutionProviderFactory* /*pointer to factory instance*/, std::vector<OrtCustomOpDomain*>& /*provider custom op domain list*/){};
+  virtual void GetCustomOpDomainList(IExecutionProviderFactory* /*pointer to factory instance*/, std::vector<OrtCustomOpDomain*>& /*provider custom op domain list*/) {};
 
   virtual void Initialize() = 0;  // Called right after loading the shared library, if this throws any errors Shutdown() will be called and the library unloaded
   virtual void Shutdown() = 0;    // Called right before unloading the shared library
diff --git a/onnxruntime/core/providers/tensorrt/tensorrt_execution_provider_custom_ops.h b/onnxruntime/core/providers/tensorrt/tensorrt_execution_provider_custom_ops.h
index 54212d34aa2ce..a72de6ed75399 100644
--- a/onnxruntime/core/providers/tensorrt/tensorrt_execution_provider_custom_ops.h
+++ b/onnxruntime/core/providers/tensorrt/tensorrt_execution_provider_custom_ops.h
@@ -24,8 +24,8 @@ struct TensorRTCustomKernel {
       : compute_stream_(compute_stream) {
   }
 
-  void Compute(OrtKernelContext* /*context*/){
-      // The implementation is in TensorRT plugin. No need to implement it here.
+  void Compute(OrtKernelContext* /*context*/) {
+    // The implementation is in TensorRT plugin. No need to implement it here.
   };
 
  private:
diff --git a/onnxruntime/core/providers/vitisai/vitisai_provider_factory.cc b/onnxruntime/core/providers/vitisai/vitisai_provider_factory.cc
index dc34419ef936f..453db30e1320f 100644
--- a/onnxruntime/core/providers/vitisai/vitisai_provider_factory.cc
+++ b/onnxruntime/core/providers/vitisai/vitisai_provider_factory.cc
@@ -46,7 +46,7 @@ struct VitisAI_Provider : Provider {
     }
   };
   // Get provider specific custom op domain list. Provider has the resposibility to release OrtCustomOpDomain instances it creates.
-  void GetCustomOpDomainList(IExecutionProviderFactory*, std::vector<OrtCustomOpDomain*>&) override{};
+  void GetCustomOpDomainList(IExecutionProviderFactory*, std::vector<OrtCustomOpDomain*>&) override {};
   // Called right after loading the shared library, if this throws any errors Shutdown() will be called and the library unloaded
   void Initialize() override { initialize_vitisai_ep(); }
   // Called right before unloading the shared library
diff --git a/onnxruntime/core/providers/vsinpu/builders/impl/elementwise_op_builder.h b/onnxruntime/core/providers/vsinpu/builders/impl/elementwise_op_builder.h
index df2e429f58b2f..4c10ba01b1c2e 100644
--- a/onnxruntime/core/providers/vsinpu/builders/impl/elementwise_op_builder.h
+++ b/onnxruntime/core/providers/vsinpu/builders/impl/elementwise_op_builder.h
@@ -47,7 +47,7 @@ namespace npu {
                        std::vector<std::shared_ptr<tim::vx::Tensor>>& outputs,                    \
                        const NodeUnit& node_unit) override {                                      \
       LOGS_DEFAULT(INFO) << "Creating " << #onnx_op_type << " Op";                                \
-      auto op = graph_ep->GetGraph() -> CreateOperation<tim::vx::ops::vsinpu_op_kind>();          \
+      auto op = graph_ep->GetGraph()->CreateOperation<tim::vx::ops::vsinpu_op_kind>();            \
       (*op).BindInputs(inputs).BindOutputs(outputs);                                              \
       return true;                                                                                \
       ;                                                                                           \
diff --git a/onnxruntime/core/providers/vsinpu/builders/op_builder_factory.h b/onnxruntime/core/providers/vsinpu/builders/op_builder_factory.h
index 27c148c1672c5..dc0969429b8ff 100644
--- a/onnxruntime/core/providers/vsinpu/builders/op_builder_factory.h
+++ b/onnxruntime/core/providers/vsinpu/builders/op_builder_factory.h
@@ -60,10 +60,9 @@ using createIOpBuildItemFunc = std::function<std::unique_ptr<IOpBuilder>()>;
 using OpBuildItemType = std::map<std::string, std::unique_ptr<IOpBuilder>>;
 
 static const std::map<std::string, createIOpBuildItemFunc> reg = {
-#define REGISTER_OP_BUILDER(ONNX_NODE_TYPE, BUILDER_TYPE)           \
-  {                                                                 \
-    ONNX_NODE_TYPE, [] { return std::make_unique<BUILDER_TYPE>(); } \
-  }
+#define REGISTER_OP_BUILDER(ONNX_NODE_TYPE, BUILDER_TYPE) \
+  {                                                       \
+      ONNX_NODE_TYPE, [] { return std::make_unique<BUILDER_TYPE>(); }}
 
     REGISTER_OP_BUILDER("Add", AddOpBuilder),
     REGISTER_OP_BUILDER("Sub", SubOpBuilder),
diff --git a/onnxruntime/python/onnxruntime_pybind_iobinding.cc b/onnxruntime/python/onnxruntime_pybind_iobinding.cc
index 51a52dbfcb3bc..37081cd0ff2b4 100644
--- a/onnxruntime/python/onnxruntime_pybind_iobinding.cc
+++ b/onnxruntime/python/onnxruntime_pybind_iobinding.cc
@@ -155,11 +155,7 @@ void addIoBindingMethods(pybind11::module& m) {
       .def("clear_binding_outputs", [](SessionIOBinding* io_binding) -> void {
         io_binding->Get()->ClearOutputs();
       })
-      .def(
-          "get_outputs", [](const SessionIOBinding* io_binding) -> const std::vector<OrtValue>& {
-            return io_binding->Get()->GetOutputs();
-          },
-          py::return_value_policy::reference_internal)
+      .def("get_outputs", [](const SessionIOBinding* io_binding) -> const std::vector<OrtValue>& { return io_binding->Get()->GetOutputs(); }, py::return_value_policy::reference_internal)
       .def("copy_outputs_to_cpu", [](const SessionIOBinding* io_binding) -> py::list {
         const std::vector<OrtValue>& outputs = io_binding->Get()->GetOutputs();
 
@@ -180,8 +176,7 @@ void addIoBindingMethods(pybind11::module& m) {
           }
           ++pos;
         }
-        return result;
-      });
+        return result; });
 }
 
 }  // namespace python
diff --git a/onnxruntime/python/onnxruntime_pybind_ortvalue.cc b/onnxruntime/python/onnxruntime_pybind_ortvalue.cc
index 94235b3043bc7..d76b9032afe73 100644
--- a/onnxruntime/python/onnxruntime_pybind_ortvalue.cc
+++ b/onnxruntime/python/onnxruntime_pybind_ortvalue.cc
@@ -226,7 +226,7 @@ void addOrtValueMethods(pybind11::module& m) {
 
         ORT_THROW("Only OrtValues that are Tensors/SparseTensors are currently supported");
 #else
-            ORT_THROW("Only OrtValues that are Tensors are supported in this build");
+        ORT_THROW("Only OrtValues that are Tensors are supported in this build");
 #endif
       })
       .def("shape", [](const OrtValue* ort_value) -> py::list {
@@ -275,26 +275,15 @@ void addOrtValueMethods(pybind11::module& m) {
 
         return *ONNX_NAMESPACE::Utils::DataTypeUtils::ToType(*type_proto);
       })
-      .def(
-          "element_type", [](const OrtValue* ort_value) -> int32_t {
-            return GetTensorProtoType(*ort_value);
-          },
-          "Returns an integer equal to the ONNX tensor proto type of the tensor or sequence. "
-          "This integer is one type defined by ONNX TensorProto_DataType "
-          "(such as onnx.TensorProto.FLOAT)."
-          "Raises an exception in any other case.")
-      .def("has_value", [](const OrtValue* ort_value) -> bool {
-        return ort_value->IsAllocated();
-      })
-      .def("is_tensor", [](const OrtValue* ort_value) -> bool {
-        return ort_value->IsTensor();
-      })
-      .def("is_sparse_tensor", [](const OrtValue* ort_value) -> bool {
-        return ort_value->IsSparseTensor();
-      })
-      .def("is_tensor_sequence", [](const OrtValue* ort_value) -> bool {
-        return ort_value->IsTensorSequence();
-      })
+      .def("element_type", [](const OrtValue* ort_value) -> int32_t { return GetTensorProtoType(*ort_value); },
+           "Returns an integer equal to the ONNX tensor proto type of the tensor or sequence. "
+           "This integer is one type defined by ONNX TensorProto_DataType "
+           "(such as onnx.TensorProto.FLOAT)."
+           "Raises an exception in any other case.")
+      .def("has_value", [](const OrtValue* ort_value) -> bool { return ort_value->IsAllocated(); })
+      .def("is_tensor", [](const OrtValue* ort_value) -> bool { return ort_value->IsTensor(); })
+      .def("is_sparse_tensor", [](const OrtValue* ort_value) -> bool { return ort_value->IsSparseTensor(); })
+      .def("is_tensor_sequence", [](const OrtValue* ort_value) -> bool { return ort_value->IsTensorSequence(); })
       // Converts Tensor into a numpy array
       .def("numpy", [](const OrtValue* ml_value) -> py::object {
         ORT_ENFORCE(ml_value->IsTensor(), "Only OrtValues that are Tensors are convertible to Numpy objects");
@@ -310,37 +299,22 @@ void addOrtValueMethods(pybind11::module& m) {
 #else
         py::object obj = GetPyObjFromTensor(*ml_value, nullptr, nullptr);
 #endif
-        return obj;
-      })
+        return obj; })
 #ifdef ENABLE_TRAINING
-      .def(
-          "to_dlpack", [](OrtValue* ort_value) -> py::object {
-            return py::reinterpret_steal<py::object>(ToDlpack(*ort_value));
-          },
-          "Returns a DLPack representing the tensor. This method does not copy the pointer shape, "
-          "instead, it copies the pointer value. The OrtValue must be persist until the dlpack structure "
-          "is consumed.")
-      .def_static(
-          "from_dlpack", [](py::object data, bool is_bool_tensor) {
-            return FromDlpack(data.ptr(), is_bool_tensor);
-          },
-          py::arg("data"), py::arg("is_bool_tensor") = false, "Converts a tensor from a external library into an OrtValue by means of the __dlpack__ protocol.")
-      .def(
-          "__dlpack__", [](OrtValue* ort_value, py::object /* stream */) -> py::object {
-            return py::reinterpret_steal<py::object>(ToDlpack(*ort_value));
-          },
-          py::arg("stream") = py::none(),
-          "Returns a DLPack representing the tensor (part of __dlpack__ protocol). "
-          "This method does not copy the pointer shape, instead, it copies the pointer value. "
-          "The OrtValue must persist until the dlpack structure is consumed.")
-      .def(
-          "__dlpack_device__", [](const OrtValue* ort_value) -> py::tuple {
+      .def("to_dlpack", [](OrtValue* ort_value) -> py::object { return py::reinterpret_steal<py::object>(ToDlpack(*ort_value)); },
+           "Returns a DLPack representing the tensor. This method does not copy the pointer shape, "
+           "instead, it copies the pointer value. The OrtValue must be persist until the dlpack structure "
+           "is consumed.")
+      .def_static("from_dlpack", [](py::object data, bool is_bool_tensor) { return FromDlpack(data.ptr(), is_bool_tensor); }, py::arg("data"), py::arg("is_bool_tensor") = false, "Converts a tensor from a external library into an OrtValue by means of the __dlpack__ protocol.")
+      .def("__dlpack__", [](OrtValue* ort_value, py::object /* stream */) -> py::object { return py::reinterpret_steal<py::object>(ToDlpack(*ort_value)); }, py::arg("stream") = py::none(),
+           "Returns a DLPack representing the tensor (part of __dlpack__ protocol). "
+           "This method does not copy the pointer shape, instead, it copies the pointer value. "
+           "The OrtValue must persist until the dlpack structure is consumed.")
+      .def("__dlpack_device__", [](const OrtValue* ort_value) -> py::tuple {
             ORT_ENFORCE(ort_value->IsTensor(), "Only tensor type OrtValues are supported");
             const onnxruntime::Tensor& tensor = ort_value->Get<Tensor>();
             DLDevice device = onnxruntime::dlpack::GetDlpackDevice(*ort_value, tensor.Location().device.Id());
-            return py::make_tuple(static_cast<int>(device.device_type), device.device_id);
-          },
-          "Returns a tuple of integers, (device, device index) (part of __dlpack__ protocol).")
+            return py::make_tuple(static_cast<int>(device.device_type), device.device_id); }, "Returns a tuple of integers, (device, device index) (part of __dlpack__ protocol).")
 #endif
       ;
 
@@ -350,13 +324,8 @@ void addOrtValueMethods(pybind11::module& m) {
         v->push_back(ortvalue);
       })
 #ifdef ENABLE_TRAINING
-      .def(
-          "push_back", [](std::vector<OrtValue>* v, py::object dlpack_tensor, const bool is_bool_tensor) {
-            v->push_back(FromDlpack(dlpack_tensor.ptr(), is_bool_tensor));
-          },
-          "Add a new OrtValue after being ownership was transferred from the DLPack structure.", py::arg("dlpack_tensor"), py::arg("is_bool_tensor") = false)
-      .def(
-          "push_back_batch", [](std::vector<OrtValue>* v, std::vector<py::object>& torch_tensors, std::vector<int64_t>& data_ptrs, std::vector<py::object>& element_types, const std::vector<std::vector<int64_t>>& shapes, const std::vector<OrtDevice>& devices) {
+      .def("push_back", [](std::vector<OrtValue>* v, py::object dlpack_tensor, const bool is_bool_tensor) { v->push_back(FromDlpack(dlpack_tensor.ptr(), is_bool_tensor)); }, "Add a new OrtValue after being ownership was transferred from the DLPack structure.", py::arg("dlpack_tensor"), py::arg("is_bool_tensor") = false)
+      .def("push_back_batch", [](std::vector<OrtValue>* v, std::vector<py::object>& torch_tensors, std::vector<int64_t>& data_ptrs, std::vector<py::object>& element_types, const std::vector<std::vector<int64_t>>& shapes, const std::vector<OrtDevice>& devices) {
             for (size_t i = 0; i < torch_tensors.size(); ++i) {
               py::object& element_type = element_types.at(i);
               const std::vector<int64_t>& shape = shapes.at(i);
@@ -377,52 +346,36 @@ void addOrtValueMethods(pybind11::module& m) {
               OrtValue ml_value;
               Tensor::InitOrtValue(ml_type, gsl::make_span(shape), reinterpret_cast<void*>(data_ptr), info, ml_value);
               v->push_back(ml_value);
-            }
-          },
-          "Add a batch of OrtValue's by wrapping PyTorch tensors.")
+            } }, "Add a batch of OrtValue's by wrapping PyTorch tensors.")
 #endif
       .def("reserve", [](std::vector<OrtValue>* v, const size_t len) { v->reserve(len); })
       .def("shrink_to_fit", [](std::vector<OrtValue>* v) { v->shrink_to_fit(); })
       .def("__len__", [](const std::vector<OrtValue>& v) { return v.size(); })
-      .def(
-          "__iter__", [](const std::vector<OrtValue>& v) {
-            return py::make_iterator(v.cbegin(), v.cend());
-          },
-          py::keep_alive<0, 1>())
-      .def("__getitem__", [](const std::vector<OrtValue>& v, const size_t idx) {
-        return v.at(idx);
-      })
-      .def(
-          "bool_tensor_indices", [](std::vector<OrtValue>* v) -> std::vector<int64_t> {
+      .def("__iter__", [](const std::vector<OrtValue>& v) { return py::make_iterator(v.cbegin(), v.cend()); }, py::keep_alive<0, 1>())
+      .def("__getitem__", [](const std::vector<OrtValue>& v, const size_t idx) { return v.at(idx); })
+      .def("bool_tensor_indices", [](std::vector<OrtValue>* v) -> std::vector<int64_t> {
             std::vector<int64_t> indices;
             for (size_t i = 0; i < v->size(); ++i) {
               if (GetTensorProtoType((*v)[i]) == ONNX_NAMESPACE::TensorProto_DataType_BOOL) {
                 indices.push_back(static_cast<int64_t>(i));
               }
             }
-            return indices;
-          },
-          "Returns the indices of every boolean tensor in this vector of OrtValue. "
-          "In case of a boolean tensor, method to_dlpacks returns a uint8 tensor instead of a boolean tensor. "
-          "If torch consumes the dlpack structure, `.to(torch.bool)` must be applied to the torch tensor "
-          "to get a boolean tensor.")
+            return indices; },
+           "Returns the indices of every boolean tensor in this vector of OrtValue. "
+           "In case of a boolean tensor, method to_dlpacks returns a uint8 tensor instead of a boolean tensor. "
+           "If torch consumes the dlpack structure, `.to(torch.bool)` must be applied to the torch tensor "
+           "to get a boolean tensor.")
 #ifdef ENABLE_TRAINING
-      .def("dlpack_at", [](std::vector<OrtValue>* v, const size_t idx) {
-        return py::reinterpret_steal<py::object>(ToDlpack(v->at(idx)));
-      })
+      .def("dlpack_at", [](std::vector<OrtValue>* v, const size_t idx) { return py::reinterpret_steal<py::object>(ToDlpack(v->at(idx))); })
 #endif
-      .def(
-          "element_type_at", [](std::vector<OrtValue>* v, const size_t idx) -> int32_t {
-            return GetTensorProtoType(v->at(idx));
-          },
-          "Returns an integer equal to the ONNX proto type of the tensor at position i. "
-          "This integer is one type defined by ONNX TensorProto_DataType "
-          "(such as onnx.TensorProto.FLOAT)."
-          "Raises an exception in any other case.",
-          py::arg("idx"))
+      .def("element_type_at", [](std::vector<OrtValue>* v, const size_t idx) -> int32_t { return GetTensorProtoType(v->at(idx)); },
+           "Returns an integer equal to the ONNX proto type of the tensor at position i. "
+           "This integer is one type defined by ONNX TensorProto_DataType "
+           "(such as onnx.TensorProto.FLOAT)."
+           "Raises an exception in any other case.",
+           py::arg("idx"))
 #ifdef ENABLE_TRAINING
-      .def(
-          "to_dlpacks", [](const std::vector<OrtValue>& v, py::object to_tensor) -> py::list {
+      .def("to_dlpacks", [](const std::vector<OrtValue>& v, py::object to_tensor) -> py::list {
             if (v.size() == 0)
               return py::list();
 
@@ -469,9 +422,8 @@ void addOrtValueMethods(pybind11::module& m) {
                 Py_DECREF(capsule);
               }
             }
-            return list_dlpacks;
-          },
-          R"pbdoc(Converts all OrtValue into tensors through DLPack protocol, the method creates
+            return list_dlpacks; },
+           R"pbdoc(Converts all OrtValue into tensors through DLPack protocol, the method creates
 a DLPack structure for every tensors, then calls python function `to_tensor` to a new object
 consuming the DLPack structure or return a list of capsule if this function is None.
 
@@ -488,7 +440,7 @@ It creates many tensors acquiring ownership of existing OrtValue.
 This method saves one object creation and an C++ allocation
 for every transferred tensor.
 )pbdoc",
-          py::arg("to_tensor"))
+           py::arg("to_tensor"))
 #endif
       ;
 
diff --git a/onnxruntime/python/onnxruntime_pybind_sparse_tensor.cc b/onnxruntime/python/onnxruntime_pybind_sparse_tensor.cc
index 7dcead113ac4f..1154f3b9f88b8 100644
--- a/onnxruntime/python/onnxruntime_pybind_sparse_tensor.cc
+++ b/onnxruntime/python/onnxruntime_pybind_sparse_tensor.cc
@@ -397,8 +397,7 @@ void addSparseTensorMethods(pybind11::module& m) {
       // pybind apparently has a bug with returning enums from def_property_readonly or methods
       // returning a method object instead of the enumeration value
       // so we are using def_property and throw on a potential modification
-      .def_property(
-          "format", [](const PySparseTensor* py_tensor) -> OrtSparseFormat {
+      .def_property("format", [](const PySparseTensor* py_tensor) -> OrtSparseFormat {
         const SparseTensor& tensor = py_tensor->Instance();
         auto retval = OrtSparseFormat::ORT_SPARSE_UNDEFINED;
         switch (tensor.Format()) {
diff --git a/onnxruntime/python/onnxruntime_pybind_state.cc b/onnxruntime/python/onnxruntime_pybind_state.cc
index 6b5daf8cb882b..679ccce7fb07a 100644
--- a/onnxruntime/python/onnxruntime_pybind_state.cc
+++ b/onnxruntime/python/onnxruntime_pybind_state.cc
@@ -1425,7 +1425,7 @@ void addGlobalMethods(py::module& m) {
     ORT_UNUSED_PARAMETER(algo);
     ORT_THROW("set_cudnn_conv_algo_search is not supported in ROCM");
 #else
-        cudnn_conv_algo_search = algo;
+    cudnn_conv_algo_search = algo;
 #endif
   });
   // TODO remove deprecated global config
@@ -1436,7 +1436,7 @@ void addGlobalMethods(py::module& m) {
     ORT_UNUSED_PARAMETER(use_single_stream);
     ORT_THROW("set_do_copy_in_default_stream is not supported in ROCM");
 #else
-        do_copy_in_default_stream = use_single_stream;
+    do_copy_in_default_stream = use_single_stream;
 #endif
   });
   // TODO remove deprecated global config
@@ -1801,10 +1801,10 @@ Applies to session load, initialization, etc. Default is 0.)pbdoc")
         }
         ORT_THROW_IF_ERROR(options->value.AddExternalInitializers(names_ptrs, values_ptrs));
 #else
-            ORT_UNUSED_PARAMETER(options);
-            ORT_UNUSED_PARAMETER(names);
-            ORT_UNUSED_PARAMETER(ort_values);
-            ORT_THROW("External initializers are not supported in this build.");
+        ORT_UNUSED_PARAMETER(options);
+        ORT_UNUSED_PARAMETER(names);
+        ORT_UNUSED_PARAMETER(ort_values);
+        ORT_THROW("External initializers are not supported in this build.");
 #endif
       });
 
@@ -1866,8 +1866,7 @@ including arg name, arg type (contains both type and shape).)pbdoc")
             return *(na.Type());
           },
           "node type")
-      .def(
-          "__str__", [](const onnxruntime::NodeArg& na) -> std::string {
+      .def("__str__", [](const onnxruntime::NodeArg& na) -> std::string {
             std::ostringstream res;
             res << "NodeArg(name='" << na.Name() << "', type='" << *(na.Type()) << "', shape=";
             auto shape = na.Shape();
@@ -1893,11 +1892,8 @@ including arg name, arg type (contains both type and shape).)pbdoc")
             }
             res << ")";
 
-            return std::string(res.str());
-          },
-          "converts the node into a readable string")
-      .def_property_readonly(
-          "shape", [](const onnxruntime::NodeArg& na) -> std::vector<py::object> {
+            return std::string(res.str()); }, "converts the node into a readable string")
+      .def_property_readonly("shape", [](const onnxruntime::NodeArg& na) -> std::vector<py::object> {
             auto shape = na.Shape();
             std::vector<py::object> arr;
             if (shape == nullptr || shape->dim_size() == 0) {
@@ -1914,9 +1910,7 @@ including arg name, arg type (contains both type and shape).)pbdoc")
                 arr[i] = py::none();
               }
             }
-            return arr;
-          },
-          "node shape (assuming the node holds a tensor)");
+            return arr; }, "node shape (assuming the node holds a tensor)");
 
   py::class_<SessionObjectInitializer> sessionObjectInitializer(m, "SessionObjectInitializer");
   py::class_<PyInferenceSession>(m, "InferenceSession", R"pbdoc(This is the main class used to run a model.)pbdoc")
@@ -2107,51 +2101,28 @@ including arg name, arg type (contains both type and shape).)pbdoc")
       .def_property_readonly("get_profiling_start_time_ns", [](const PyInferenceSession* sess) -> uint64_t {
         return sess->GetSessionHandle()->GetProfiling().GetStartTimeNs();
       })
-      .def(
-          "get_providers", [](const PyInferenceSession* sess) -> const std::vector<std::string>& {
-            return sess->GetSessionHandle()->GetRegisteredProviderTypes();
-          },
-          py::return_value_policy::reference_internal)
-      .def(
-          "get_provider_options", [](const PyInferenceSession* sess) -> const ProviderOptionsMap& {
-            return sess->GetSessionHandle()->GetAllProviderOptions();
-          },
-          py::return_value_policy::reference_internal)
-      .def_property_readonly(
-          "session_options", [](const PyInferenceSession* sess) -> PySessionOptions* {
+      .def("get_providers", [](const PyInferenceSession* sess) -> const std::vector<std::string>& { return sess->GetSessionHandle()->GetRegisteredProviderTypes(); }, py::return_value_policy::reference_internal)
+      .def("get_provider_options", [](const PyInferenceSession* sess) -> const ProviderOptionsMap& { return sess->GetSessionHandle()->GetAllProviderOptions(); }, py::return_value_policy::reference_internal)
+      .def_property_readonly("session_options", [](const PyInferenceSession* sess) -> PySessionOptions* {
             auto session_options = std::make_unique<PySessionOptions>();
             session_options->value = sess->GetSessionHandle()->GetSessionOptions();
-            return session_options.release();
-          },
-          py::return_value_policy::take_ownership)
-      .def_property_readonly(
-          "inputs_meta", [](const PyInferenceSession* sess) -> const std::vector<const onnxruntime::NodeArg*>& {
+            return session_options.release(); }, py::return_value_policy::take_ownership)
+      .def_property_readonly("inputs_meta", [](const PyInferenceSession* sess) -> const std::vector<const onnxruntime::NodeArg*>& {
             auto res = sess->GetSessionHandle()->GetModelInputs();
             OrtPybindThrowIfError(res.first);
-            return *(res.second);
-          },
-          py::return_value_policy::reference_internal)
-      .def_property_readonly(
-          "outputs_meta", [](const PyInferenceSession* sess) -> const std::vector<const onnxruntime::NodeArg*>& {
+            return *(res.second); }, py::return_value_policy::reference_internal)
+      .def_property_readonly("outputs_meta", [](const PyInferenceSession* sess) -> const std::vector<const onnxruntime::NodeArg*>& {
             auto res = sess->GetSessionHandle()->GetModelOutputs();
             OrtPybindThrowIfError(res.first);
-            return *(res.second);
-          },
-          py::return_value_policy::reference_internal)
-      .def_property_readonly(
-          "overridable_initializers", [](const PyInferenceSession* sess) -> const std::vector<const onnxruntime::NodeArg*>& {
+            return *(res.second); }, py::return_value_policy::reference_internal)
+      .def_property_readonly("overridable_initializers", [](const PyInferenceSession* sess) -> const std::vector<const onnxruntime::NodeArg*>& {
             auto res = sess->GetSessionHandle()->GetOverridableInitializers();
             OrtPybindThrowIfError(res.first);
-            return *(res.second);
-          },
-          py::return_value_policy::reference_internal)
-      .def_property_readonly(
-          "model_meta", [](const PyInferenceSession* sess) -> const onnxruntime::ModelMetadata& {
+            return *(res.second); }, py::return_value_policy::reference_internal)
+      .def_property_readonly("model_meta", [](const PyInferenceSession* sess) -> const onnxruntime::ModelMetadata& {
             auto res = sess->GetSessionHandle()->GetModelMetadata();
             OrtPybindThrowIfError(res.first);
-            return *(res.second);
-          },
-          py::return_value_policy::reference_internal)
+            return *(res.second); }, py::return_value_policy::reference_internal)
       .def("run_with_iobinding", [](PyInferenceSession* sess, SessionIOBinding& io_binding, RunOptions* run_options = nullptr) -> void {
         Status status;
         // release GIL to allow multiple python threads to invoke Run() in parallel.
@@ -2161,8 +2132,7 @@ including arg name, arg type (contains both type and shape).)pbdoc")
         else
           status = sess->GetSessionHandle()->Run(*run_options, *io_binding.Get());
         if (!status.IsOK())
-          throw std::runtime_error("Error in execution: " + status.ErrorMessage());
-      })
+          throw std::runtime_error("Error in execution: " + status.ErrorMessage()); })
       .def("get_tuning_results", [](PyInferenceSession* sess) -> py::list {
 #if !defined(ORT_MINIMAL_BUILD)
         auto results = sess->GetSessionHandle()->GetTuningResults();
@@ -2177,8 +2147,8 @@ including arg name, arg type (contains both type and shape).)pbdoc")
 
         return ret;
 #else
-            ORT_UNUSED_PARAMETER(sess);
-            ORT_THROW("TunableOp and get_tuning_results are not supported in this build.");
+        ORT_UNUSED_PARAMETER(sess);
+        ORT_THROW("TunableOp and get_tuning_results are not supported in this build.");
 #endif
       })
       .def("set_tuning_results", [](PyInferenceSession* sess, py::list results, bool error_on_invalid) -> void {
@@ -2209,10 +2179,10 @@ including arg name, arg type (contains both type and shape).)pbdoc")
           throw std::runtime_error("Error in execution: " + status.ErrorMessage());
         }
 #else
-            ORT_UNUSED_PARAMETER(sess);
-            ORT_UNUSED_PARAMETER(results);
-            ORT_UNUSED_PARAMETER(error_on_invalid);
-            ORT_THROW("TunableOp and set_tuning_results are not supported in this build.");
+        ORT_UNUSED_PARAMETER(sess);
+        ORT_UNUSED_PARAMETER(results);
+        ORT_UNUSED_PARAMETER(error_on_invalid);
+        ORT_THROW("TunableOp and set_tuning_results are not supported in this build.");
 #endif
       });
 
diff --git a/onnxruntime/python/onnxruntime_validation.py b/onnxruntime/python/onnxruntime_validation.py
index 10d9f469863c4..81e6461e4417f 100644
--- a/onnxruntime/python/onnxruntime_validation.py
+++ b/onnxruntime/python/onnxruntime_validation.py
@@ -24,8 +24,7 @@ def check_distro_info():
 
         if __my_distro_ver__ not in ["10", "11"]:
             warnings.warn(
-                "Unsupported Windows version (%s). ONNX Runtime supports Windows 10 and above, only."
-                % __my_distro_ver__
+                f"Unsupported Windows version ({__my_distro_ver__}). ONNX Runtime supports Windows 10 and above, only."
             )
     elif __my_system__ == "linux":
         """Although the 'platform' python module for getting Distro information works well on standard OS images
@@ -54,11 +53,11 @@ def check_distro_info():
 
         if int(__my_distro_ver__.split(".")[0]) < 11:
             warnings.warn(
-                "Unsupported macOS version (%s). ONNX Runtime supports macOS 11.0 or later." % (__my_distro_ver__)
+                f"Unsupported macOS version ({__my_distro_ver__}). ONNX Runtime supports macOS 11.0 or later."
             )
     else:
         warnings.warn(
-            "Unsupported platform (%s). ONNX Runtime supports Linux, macOS and Windows platforms, only." % __my_system__
+            f"Unsupported platform ({__my_system__}). ONNX Runtime supports Linux, macOS and Windows platforms, only."
         )
 
 
@@ -115,10 +114,10 @@ def validate_build_package_info():
                     cudart_version = None
 
                 def print_build_package_info():
-                    warnings.warn("onnxruntime training package info: package_name: %s" % package_name)
-                    warnings.warn("onnxruntime training package info: __version__: %s" % version)
-                    warnings.warn("onnxruntime training package info: cuda_version: %s" % cuda_version)
-                    warnings.warn("onnxruntime build info: cudart_version: %s" % cudart_version)
+                    warnings.warn(f"onnxruntime training package info: package_name: {package_name}")
+                    warnings.warn(f"onnxruntime training package info: __version__: {version}")
+                    warnings.warn(f"onnxruntime training package info: cuda_version: {cuda_version}")
+                    warnings.warn(f"onnxruntime build info: cudart_version: {cudart_version}")
 
                 # collection cuda library info from current environment.
                 from onnxruntime.capi.onnxruntime_collect_build_info import find_cudart_versions
@@ -127,7 +126,7 @@ def print_build_package_info():
                 if cudart_version and local_cudart_versions and cudart_version not in local_cudart_versions:
                     print_build_package_info()
                     warnings.warn("WARNING: failed to find cudart version that matches onnxruntime build info")
-                    warnings.warn("WARNING: found cudart versions: %s" % local_cudart_versions)
+                    warnings.warn(f"WARNING: found cudart versions: {local_cudart_versions}")
             else:
                 # TODO: rcom
                 pass
diff --git a/onnxruntime/python/tools/pytorch_export_contrib_ops.py b/onnxruntime/python/tools/pytorch_export_contrib_ops.py
index aeb78f03dd721..d8cf3c1304219 100644
--- a/onnxruntime/python/tools/pytorch_export_contrib_ops.py
+++ b/onnxruntime/python/tools/pytorch_export_contrib_ops.py
@@ -22,7 +22,7 @@
 
 
 def _reg(symbolic_fn: typing.Callable):
-    name = "::%s" % symbolic_fn.__name__
+    name = f"::{symbolic_fn.__name__}"
     torch.onnx.register_custom_op_symbolic(name, symbolic_fn, _OPSET_VERSION)
     _registered_ops.add(name)
 
diff --git a/onnxruntime/python/tools/quantization/calibrate.py b/onnxruntime/python/tools/quantization/calibrate.py
index 65875d09102bd..703accbcc1c48 100644
--- a/onnxruntime/python/tools/quantization/calibrate.py
+++ b/onnxruntime/python/tools/quantization/calibrate.py
@@ -1076,7 +1076,7 @@ def get_entropy_threshold(self, histogram, num_quantized_bins):
 
         for i in range(num_half_quantized_bin, zero_bin_index + 1, 1):
             start_index = zero_bin_index - i
-            end_index = zero_bin_index + i + 1 if (zero_bin_index + i + 1) <= num_bins else num_bins
+            end_index = min(zero_bin_index + i + 1, num_bins)
 
             thresholds[i - num_half_quantized_bin] = (hist_edges[start_index], hist_edges[end_index])
 
diff --git a/onnxruntime/python/tools/symbolic_shape_infer.py b/onnxruntime/python/tools/symbolic_shape_infer.py
index ac959d5c061f7..f88011c7a2cf9 100755
--- a/onnxruntime/python/tools/symbolic_shape_infer.py
+++ b/onnxruntime/python/tools/symbolic_shape_infer.py
@@ -24,7 +24,7 @@ def get_attribute(node, attr_name, default_value=None):
 
 
 def get_dim_from_proto(dim):
-    return getattr(dim, dim.WhichOneof("value")) if type(dim.WhichOneof("value")) is str else None  # noqa: E721
+    return getattr(dim, dim.WhichOneof("value")) if type(dim.WhichOneof("value")) is str else None
 
 
 def is_sequence(type_proto):
@@ -92,19 +92,19 @@ def get_opset(mp, domain=None):
 
 
 def as_scalar(x):
-    if type(x) == list:  # noqa: E721
+    if type(x) is list:
         assert len(x) == 1
         return x[0]
-    elif type(x) == np.ndarray:
+    elif type(x) is np.ndarray:
         return x.item()
     else:
         return x
 
 
 def as_list(x, keep_none):
-    if type(x) == list:  # noqa: E721
+    if type(x) is list:
         return x
-    elif type(x) == np.ndarray:
+    elif type(x) is np.ndarray:
         return list(x)
     elif keep_none and x is None:
         return None
@@ -113,7 +113,7 @@ def as_list(x, keep_none):
 
 
 def sympy_reduce_product(x):
-    if type(x) == list:  # noqa: E721
+    if type(x) is list:
         value = sympy.Integer(1)
         for v in x:
             value = value * v
@@ -258,7 +258,7 @@ def __init__(self, int_max, auto_merge, guess_output_rank, verbose, prefix=""):
         self.prefix_ = prefix
 
     def _add_suggested_merge(self, symbols, apply=False):
-        assert all([(type(s) == str and s in self.symbolic_dims_) or is_literal(s) for s in symbols])  # noqa: E721
+        assert all([(type(s) is str and s in self.symbolic_dims_) or is_literal(s) for s in symbols])
         symbols = set(symbols)
         for k, v in self.suggested_merge_.items():
             if k in symbols:
@@ -278,7 +278,7 @@ def _add_suggested_merge(self, symbols, apply=False):
                     break
         if map_to is None:
             for s in symbols:
-                if type(self.symbolic_dims_[s]) == sympy.Symbol:
+                if type(self.symbolic_dims_[s]) is sympy.Symbol:
                     map_to = s
                     break
         # when nothing to map to, use the shorter one
@@ -328,7 +328,7 @@ def _preprocess(self, in_mp):
         )
 
     def _merge_symbols(self, dims):
-        if not all([type(d) == str for d in dims]):  # noqa: E721
+        if not all([type(d) is str for d in dims]):
             if self.auto_merge_:
                 unique_dims = list(set(dims))
                 is_int = [is_literal(d) for d in unique_dims]
@@ -408,7 +408,7 @@ def _get_shape_rank(self, node, idx):
     def _get_sympy_shape(self, node, idx):
         sympy_shape = []
         for d in self._get_shape(node, idx):
-            if type(d) == str:  # noqa: E721
+            if type(d) is str:
                 sympy_shape.append(
                     self.symbolic_dims_[d]
                     if d in self.symbolic_dims_
@@ -590,7 +590,7 @@ def _onnx_infer_subgraph(self, node, subgraph, use_node_input=True, inc_subgraph
         # for new symbolic dims from subgraph output, add to main graph symbolic dims
         subgraph_shapes = [get_shape_from_value_info(o) for o in symbolic_shape_inference.out_mp_.graph.output]
         subgraph_new_symbolic_dims = {
-            d for s in subgraph_shapes if s for d in s if type(d) == str and d not in self.symbolic_dims_  # noqa: E721
+            d for s in subgraph_shapes if s for d in s if type(d) is str and d not in self.symbolic_dims_
         }
         new_dims = {}
         for d in subgraph_new_symbolic_dims:
@@ -610,7 +610,7 @@ def int_or_float(value, allow_float_values):
         if all([v is not None for v in values]):
             # some shape compute is in floating point, cast to int for sympy
             for i, v in enumerate(values):
-                if type(v) != np.ndarray:
+                if type(v) is not np.ndarray:
                     continue
                 if len(v.shape) > 1:
                     new_v = None  # ignore value for rank > 1
@@ -924,7 +924,7 @@ def _infer_Concat(self, node):  # noqa: N802
             if all([d == dims[0] for d in dims]):
                 continue
             merged = self._merge_symbols(dims)
-            if type(merged) == str:  # noqa: E721
+            if type(merged) is str:
                 sympy_shape[d] = self.symbolic_dims_[merged] if merged else None
             else:
                 sympy_shape[d] = merged
@@ -1060,7 +1060,7 @@ def _infer_Einsum(self, node):  # noqa: N802
                     dim = shape[-i]
                     if letter not in letter_to_dim:
                         letter_to_dim[letter] = dim
-                    elif type(dim) != sympy.Symbol:
+                    elif type(dim) is not sympy.Symbol:
                         letter_to_dim[letter] = dim
             num_operands = num_operands + 1
 
@@ -1127,8 +1127,8 @@ def _infer_Gather(self, node):  # noqa: N802
             idx = self._try_get_value(node, 1)
             if idx is not None:
                 data = self.sympy_data_[node.input[0]]
-                if type(data) == list:  # noqa: E721
-                    if type(idx) == np.ndarray and len(idx.shape) == 1:
+                if type(data) is list:
+                    if type(idx) is np.ndarray and len(idx.shape) == 1:
                         self.sympy_data_[node.output[0]] = [data[int(i)] for i in idx]
                     else:
                         self.sympy_data_[node.output[0]] = data[int(idx)]
@@ -1530,7 +1530,7 @@ def _infer_aten_upsample(self, node):
             new_shape = input_shape[:2]
             output_size = self._try_get_value(node, 1)
             if output_size is not None:
-                new_shape += [dim_size.item() if type(dim_size) == np.int64 else dim_size for dim_size in output_size]
+                new_shape += [dim_size.item() if type(dim_size) is np.int64 else dim_size for dim_size in output_size]
             else:
                 rank = len(input_shape)
                 new_shape += [str(self._new_symbolic_dim_from_output(node, 0, i)) for i in range(2, rank)]
@@ -1645,7 +1645,7 @@ def _infer_Reshape(self, node):  # noqa: N802
             deferred_dim_idx = -1
             non_deferred_size = 1
             for i, d in enumerate(shape_value):
-                if type(d) == sympy.Symbol:
+                if type(d) is sympy.Symbol:
                     new_sympy_shape.append(d)
                 elif d == 0:
                     new_sympy_shape.append(input_sympy_shape[i])
@@ -1940,7 +1940,7 @@ def handle_negative_index(index, bound):
         # handle sympy_data if needed, for slice in shape computation
         if (
             node.input[0] in self.sympy_data_
-            and [0] == axes
+            and axes == [0]
             and starts is not None
             and len(starts) == 1
             and ends is not None
@@ -1949,8 +1949,8 @@ def handle_negative_index(index, bound):
             and len(steps) == 1
         ):
             input_sympy_data = self.sympy_data_[node.input[0]]
-            if type(input_sympy_data) == list or (  # noqa: E721
-                type(input_sympy_data) == np.array and len(input_sympy_data.shape) == 1
+            if type(input_sympy_data) is list or (
+                type(input_sympy_data) is np.array and len(input_sympy_data.shape) == 1
             ):
                 self.sympy_data_[node.output[0]] = input_sympy_data[starts[0] : ends[0] : steps[0]]
 
@@ -2616,7 +2616,7 @@ def _infer_impl(self, start_sympy_data=None):
                     # some models use None for symbolic dim in input, replace it with a string
                     input_dims[i_dim].dim_param = str(self._new_symbolic_dim(i.name, i_dim))
 
-            self.input_symbols_.update([d for d in input_shape if type(d) == str])  # noqa: E721
+            self.input_symbols_.update([d for d in input_shape if type(d) is str])
 
         for s in self.input_symbols_:
             if s in self.suggested_merge_:
diff --git a/onnxruntime/python/tools/tensorrt/perf/benchmark.py b/onnxruntime/python/tools/tensorrt/perf/benchmark.py
index 8af074f24acc9..4fa5d0c0ea034 100644
--- a/onnxruntime/python/tools/tensorrt/perf/benchmark.py
+++ b/onnxruntime/python/tools/tensorrt/perf/benchmark.py
@@ -925,8 +925,8 @@ def find_model_path(path):
 
     logger.info(target_model_path)
     if len(target_model_path) > 1:
-        logger.error("We expect to find only one model in " + path)  # noqa: G003
-        raise
+        logger.error("We expect to find only one model in %s", path)
+        raise RuntimeError
 
     return target_model_path[0]
 
@@ -1007,7 +1007,7 @@ def parse_models_info_from_file(root_dir, path, models):
                 models[row["model_name"]] = {}
             else:
                 logger.error("Model name must be provided in models_info.json")
-                raise
+                raise RuntimeError
 
             model = models[row["model_name"]]
 
@@ -1018,19 +1018,19 @@ def parse_models_info_from_file(root_dir, path, models):
                     model["working_directory"] = os.path.join(root_working_directory, row["working_directory"])
             else:
                 logger.error("Model path must be provided in models_info.json")
-                raise
+                raise RuntimeError
 
             if "model_path" in row:
                 model["model_path"] = row["model_path"]
             else:
                 logger.error("Model path must be provided in models_info.json")
-                raise
+                raise RuntimeError
 
             if "test_data_path" in row:
                 model["test_data_path"] = row["test_data_path"]
             else:
                 logger.error("Test data path must be provided in models_info.json")
-                raise
+                raise RuntimeError
 
             if "model_path_fp16" in row:
                 model["model_path_fp16"] = row["model_path_fp16"]
diff --git a/onnxruntime/python/tools/tensorrt/perf/perf_utils.py b/onnxruntime/python/tools/tensorrt/perf/perf_utils.py
index c447bf9cffe27..0d0f7cc48f361 100644
--- a/onnxruntime/python/tools/tensorrt/perf/perf_utils.py
+++ b/onnxruntime/python/tools/tensorrt/perf/perf_utils.py
@@ -234,7 +234,7 @@ def calculate_trt_op_percentage(trt_op_map, cuda_op_map):
 
     if total_ops == 0:
         print("Error ...")
-        raise
+        raise RuntimeError
 
     if len(trt_op_map) == 0:
         total_cuda_and_cpu_ops = total_ops
diff --git a/onnxruntime/python/tools/tensorrt/perf/setup_scripts/setup_onnx_zoo.py b/onnxruntime/python/tools/tensorrt/perf/setup_scripts/setup_onnx_zoo.py
index 4f763ad84426d..0532dd7c72c1c 100644
--- a/onnxruntime/python/tools/tensorrt/perf/setup_scripts/setup_onnx_zoo.py
+++ b/onnxruntime/python/tools/tensorrt/perf/setup_scripts/setup_onnx_zoo.py
@@ -71,7 +71,7 @@ def write_json(models):
 def main():
     links = []
     with open("links.txt") as fh:
-        links = [link.rstrip() for link in fh.readlines()]
+        links = [link.rstrip() for link in fh]
 
     model_list = []
     for link in links:
diff --git a/onnxruntime/python/tools/transformers/benchmark.py b/onnxruntime/python/tools/transformers/benchmark.py
index 5ec2ab4e50799..4800c48744236 100644
--- a/onnxruntime/python/tools/transformers/benchmark.py
+++ b/onnxruntime/python/tools/transformers/benchmark.py
@@ -802,7 +802,7 @@ def main():
         try:
             os.mkdir(args.cache_dir)
         except OSError:
-            logger.error("Creation of the directory %s failed" % args.cache_dir)  # noqa: G002
+            logger.error("Creation of the directory %s failed", args.cache_dir)
 
     enable_torch = "torch" in args.engines
     enable_torch2 = "torch2" in args.engines
diff --git a/onnxruntime/python/tools/transformers/bert_test_data.py b/onnxruntime/python/tools/transformers/bert_test_data.py
index aa82e047df328..167fc8697ce06 100644
--- a/onnxruntime/python/tools/transformers/bert_test_data.py
+++ b/onnxruntime/python/tools/transformers/bert_test_data.py
@@ -168,11 +168,11 @@ def output_test_data(directory: str, inputs: Dict[str, np.ndarray]):
         try:
             os.mkdir(directory)
         except OSError:
-            print("Creation of the directory %s failed" % directory)
+            print(f"Creation of the directory {directory} failed")
         else:
-            print("Successfully created the directory %s " % directory)
+            print(f"Successfully created the directory {directory} ")
     else:
-        print("Warning: directory %s existed. Files will be overwritten." % directory)
+        print(f"Warning: directory {directory} existed. Files will be overwritten.")
 
     for index, (name, data) in enumerate(inputs.items()):
         tensor = numpy_helper.from_array(data, name)
diff --git a/onnxruntime/python/tools/transformers/fusion_attention.py b/onnxruntime/python/tools/transformers/fusion_attention.py
index f48cabd25fc5c..dc2b38f3928ac 100644
--- a/onnxruntime/python/tools/transformers/fusion_attention.py
+++ b/onnxruntime/python/tools/transformers/fusion_attention.py
@@ -672,7 +672,7 @@ def create_multihead_attention_node(
                 q_matmul, k_matmul, v_matmul, q_add, k_add, v_add, num_heads
             )
             mha_inputs.extend([q_slice.output[0], k_slice.output[0], v_slice.output[0]])
-        elif type(k_matmul) == NodeProto and type(v_matmul) == NodeProto:
+        elif type(k_matmul) is NodeProto and type(v_matmul) is NodeProto:
             if self.disable_multi_head_attention_bias:
                 mha_inputs.extend([q_add.output[0], k_matmul.output[0], v_add.output[0]])
             else:
diff --git a/onnxruntime/python/tools/transformers/fusion_utils.py b/onnxruntime/python/tools/transformers/fusion_utils.py
index 726c587ff7043..dbd9e828198ca 100644
--- a/onnxruntime/python/tools/transformers/fusion_utils.py
+++ b/onnxruntime/python/tools/transformers/fusion_utils.py
@@ -159,7 +159,7 @@ def transpose_2d_int8_tensor(tensor: onnx_proto.TensorProto):
             tensor (TensorProto): transposed tensor
         """
         if not isinstance(tensor, onnx_proto.TensorProto):
-            raise ValueError("Expected input type is an ONNX TensorProto but got %s" % type(tensor))
+            raise ValueError(f"Expected input type is an ONNX TensorProto but got {type(tensor)}")
 
         if len(tensor.dims) != 2 or tensor.data_type != onnx_proto.TensorProto.INT8:
             raise ValueError("Only INT8 2-D tensors can be transposed")
diff --git a/onnxruntime/python/tools/transformers/models/bart/utils/export_summarization_edinit.py b/onnxruntime/python/tools/transformers/models/bart/utils/export_summarization_edinit.py
index 111520a6e3aeb..8a610fb17671b 100644
--- a/onnxruntime/python/tools/transformers/models/bart/utils/export_summarization_edinit.py
+++ b/onnxruntime/python/tools/transformers/models/bart/utils/export_summarization_edinit.py
@@ -205,5 +205,5 @@ def export_encoder(args):
             no_repeat_ngram_size=no_repeat_ngram_size,
         )
         time_cost = time.time() - start_time
-        print("--- %s seconds ---" % (time_cost))
+        print(f"--- {time_cost} seconds ---")
         print(tokenizer.decode(pred_ids[0], skip_special_tokens=True, clean_up_tokenization_spaces=False))
diff --git a/onnxruntime/python/tools/transformers/models/bart/utils/export_summarization_enc_dec_past.py b/onnxruntime/python/tools/transformers/models/bart/utils/export_summarization_enc_dec_past.py
index 29c39730c79ef..afd01ae9d025f 100644
--- a/onnxruntime/python/tools/transformers/models/bart/utils/export_summarization_enc_dec_past.py
+++ b/onnxruntime/python/tools/transformers/models/bart/utils/export_summarization_enc_dec_past.py
@@ -266,5 +266,5 @@ def export_decoder(args):
             use_cache=True,
         )
         time_cost = time.time() - start_time
-        print("--- %s seconds ---" % (time_cost))
+        print(f"--- {time_cost} seconds ---")
         print(tokenizer.decode(pred_ids[0], skip_special_tokens=True, clean_up_tokenization_spaces=False))
diff --git a/onnxruntime/python/tools/transformers/models/bart/utils/onnx_inference.py b/onnxruntime/python/tools/transformers/models/bart/utils/onnx_inference.py
index c4c8a2dcba697..7a5cfe42c7efe 100644
--- a/onnxruntime/python/tools/transformers/models/bart/utils/onnx_inference.py
+++ b/onnxruntime/python/tools/transformers/models/bart/utils/onnx_inference.py
@@ -49,7 +49,7 @@ def run_inference(args):
             no_repeat_ngram_size=no_repeat_ngram_size,
         )
         time_cost = time.time() - start_time
-        print("--- %s seconds ---" % (time_cost))
+        print(f"--- {time_cost} seconds ---")
         for j in range(batch_num):
             for i in range(beam):
                 print(
@@ -81,7 +81,7 @@ def run_inference(args):
         start_time = time.time()
         out = sess.run(None, ort_inputs)
         time_cost = time.time() - start_time
-        print("--- %s seconds ---" % (time_cost))
+        print(f"--- {time_cost} seconds ---")
         for j in range(batch_num):
             for i in range(beam):
                 print(
diff --git a/onnxruntime/python/tools/transformers/models/stable_diffusion/engine_builder.py b/onnxruntime/python/tools/transformers/models/stable_diffusion/engine_builder.py
index 26b9a2792e9e1..0b6d325803554 100644
--- a/onnxruntime/python/tools/transformers/models/stable_diffusion/engine_builder.py
+++ b/onnxruntime/python/tools/transformers/models/stable_diffusion/engine_builder.py
@@ -117,7 +117,7 @@ def get_cached_model_name(self, model_name):
             model_name = model_name + "_" + "_".join(self.pipeline_info.controlnet)
 
         if hash_source:
-            model_name += "_" + hashlib.md5("\t".join(hash_source).encode("utf-8")).digest().hex()[:8]
+            model_name += "_" + hashlib.md5("\t".join(hash_source).encode("utf-8")).hexdigest()[:8]
 
         # TODO: When we support original VAE, we shall save custom VAE to another directory.
 
diff --git a/onnxruntime/python/tools/transformers/models/stable_diffusion/pipeline_stable_diffusion.py b/onnxruntime/python/tools/transformers/models/stable_diffusion/pipeline_stable_diffusion.py
index 1629537dc294f..522cc541c1e57 100644
--- a/onnxruntime/python/tools/transformers/models/stable_diffusion/pipeline_stable_diffusion.py
+++ b/onnxruntime/python/tools/transformers/models/stable_diffusion/pipeline_stable_diffusion.py
@@ -459,9 +459,9 @@ def denoise_latent(
                 noise_pred_uncond, noise_pred_text = noise_pred.chunk(2)
                 noise_pred = noise_pred_uncond + guidance * (noise_pred_text - noise_pred_uncond)
 
-            if type(self.scheduler) == UniPCMultistepScheduler:
+            if type(self.scheduler) is UniPCMultistepScheduler:
                 latents = self.scheduler.step(noise_pred, timestep, latents, return_dict=False)[0]
-            elif type(self.scheduler) == LCMScheduler:
+            elif type(self.scheduler) is LCMScheduler:
                 latents = self.scheduler.step(noise_pred, timestep, latents, generator=self.generator)[0]
             else:
                 latents = self.scheduler.step(noise_pred, latents, step_offset + step_index, timestep)
diff --git a/onnxruntime/test/framework/allocation_planner_test.cc b/onnxruntime/test/framework/allocation_planner_test.cc
index 26e40b25930c8..4e9e80b180e9c 100644
--- a/onnxruntime/test/framework/allocation_planner_test.cc
+++ b/onnxruntime/test/framework/allocation_planner_test.cc
@@ -1883,7 +1883,7 @@ TEST_F(PlannerTest, ParaPlanCreation) {
       ORT_ENFORCE(main_graph_ort_value_index_map.GetName(per_value_plan.reused_buffer, reused).IsOK());
       reuse_pairs.erase(reused);
     }  // if
-  }    // for
+  }  // for
   ASSERT_TRUE(reuse_pairs.empty());
 }
 
diff --git a/onnxruntime/test/onnx/OrtValueList.h b/onnxruntime/test/onnx/OrtValueList.h
index 2929cdca428d9..921c1d3872111 100644
--- a/onnxruntime/test/onnx/OrtValueList.h
+++ b/onnxruntime/test/onnx/OrtValueList.h
@@ -14,7 +14,7 @@ class OrtValueArray {
  public:
   ORT_DISALLOW_COPY_ASSIGNMENT_AND_MOVE(OrtValueArray);
   // n must be non-negative
-  OrtValueArray(int n) : values(static_cast<size_t>(n), nullptr){};
+  OrtValueArray(int n) : values(static_cast<size_t>(n), nullptr) {};
   ~OrtValueArray() {
     for (OrtValue* v : values) {
       if (v != nullptr) Ort::GetApi().ReleaseValue(v);
diff --git a/onnxruntime/test/onnx/microbenchmark/activation.cc b/onnxruntime/test/onnx/microbenchmark/activation.cc
index 69ee72996365e..df36135bd3017 100644
--- a/onnxruntime/test/onnx/microbenchmark/activation.cc
+++ b/onnxruntime/test/onnx/microbenchmark/activation.cc
@@ -27,7 +27,7 @@ class Allocs : public IExecutionProvider {
   std::shared_ptr<CPUAllocator> alloc = std::make_shared<CPUAllocator>();
 
  public:
-  Allocs() : IExecutionProvider("fake"){};
+  Allocs() : IExecutionProvider("fake") {};
   AllocatorPtr GetAllocator(OrtMemType) const {
     return alloc;
   }
diff --git a/onnxruntime/test/optimizer/qdq_transformer_fastmath_test.cc b/onnxruntime/test/optimizer/qdq_transformer_fastmath_test.cc
index ec9f78da14a75..ccfa1f1159937 100644
--- a/onnxruntime/test/optimizer/qdq_transformer_fastmath_test.cc
+++ b/onnxruntime/test/optimizer/qdq_transformer_fastmath_test.cc
@@ -401,7 +401,7 @@ void QDQTransformerGemmTests(bool has_output_q, bool has_bias, bool beta_not_one
     auto check_binary_op_graph = [&](InferenceSessionWrapper& session) {
       auto op_to_count = CountOpsInGraph(session.GetGraph());
       const QDQOpKeys qdq_keys = GetQDQOpKeys(use_contrib_qdq);
-      if ((!has_output_q || std::is_same_v<Input1Type, OutputType>)&&(!has_bias || (std::is_same_v<BiasType, int32_t> && !beta_not_one)) &&
+      if ((!has_output_q || std::is_same_v<Input1Type, OutputType>) && (!has_bias || (std::is_same_v<BiasType, int32_t> && !beta_not_one)) &&
           (std::is_same_v<Input1Type, uint8_t> || std::is_same_v<Input2Type, int8_t>)) {
         EXPECT_EQ(op_to_count["com.microsoft.QGemm"], 1);
         EXPECT_EQ(op_to_count["Gemm"], 0);
diff --git a/onnxruntime/test/optimizer/qdq_transformer_test.cc b/onnxruntime/test/optimizer/qdq_transformer_test.cc
index 1638851daf65a..14c5b60d6e0bd 100644
--- a/onnxruntime/test/optimizer/qdq_transformer_test.cc
+++ b/onnxruntime/test/optimizer/qdq_transformer_test.cc
@@ -786,7 +786,7 @@ void QDQTransformerGemmTests(bool has_output_q, bool has_bias, bool beta_not_one
     auto check_binary_op_graph = [&](InferenceSessionWrapper& session) {
       auto op_to_count = CountOpsInGraph(session.GetGraph());
       const QDQOpKeys qdq_keys = GetQDQOpKeys(use_contrib_qdq);
-      if ((!has_output_q || std::is_same_v<Input1Type, OutputType>)&&(!has_bias || (std::is_same_v<BiasType, int32_t> && !beta_not_one)) &&
+      if ((!has_output_q || std::is_same_v<Input1Type, OutputType>) && (!has_bias || (std::is_same_v<BiasType, int32_t> && !beta_not_one)) &&
           (std::is_same_v<Input1Type, uint8_t> || std::is_same_v<Input2Type, int8_t>)) {
         EXPECT_EQ(op_to_count["com.microsoft.QGemm"], 1);
         EXPECT_EQ(op_to_count["Gemm"], 0);
diff --git a/onnxruntime/test/providers/cpu/reduction/reduction_test_cases_generator.py b/onnxruntime/test/providers/cpu/reduction/reduction_test_cases_generator.py
index 568a4649f3977..bd06ae9fe881a 100644
--- a/onnxruntime/test/providers/cpu/reduction/reduction_test_cases_generator.py
+++ b/onnxruntime/test/providers/cpu/reduction/reduction_test_cases_generator.py
@@ -40,13 +40,13 @@ def TestReduction(op, data, axes, keepdims):  # noqa: N802
 
 
 def PrintResult(op, axes, keepdims, res):  # noqa: N802
-    print('  {"%s",' % op)
+    print(f'  {{"{op}",')
     print("OpAttributesResult(")
     print("    // ReductionAttribute")
     print("      {")
     print(" // axes_")
     print("{", end="")
-    print(*axes, sep=", ", end="") if axes else print("")
+    print(*axes, sep=", ", end="") if axes else print()
     print("},")
     print(" // keep_dims_")
     print(keepdims, ",")
@@ -60,7 +60,7 @@ def PrintResult(op, axes, keepdims, res):  # noqa: N802
     print(" // expected values")
     print("{", end="")
     for i in range(res.size):
-        print("%5.6ff," % res.item(i))
+        print(f"{res.item(i):5.6f}f,")
 
     print("})},")
 
@@ -130,7 +130,7 @@ def PrintReenableOptimizations():  # noqa: N802
     print("{")
     for i in range(input_data.size):
         print(
-            "%5.6ff," % input_data.item(i),
+            f"{input_data.item(i):5.6f}f,",
         )
     print("},")
     print("// input_dims")
diff --git a/onnxruntime/test/providers/cpu/tensor/pad_test.cc b/onnxruntime/test/providers/cpu/tensor/pad_test.cc
index 5fc8ed417391e..1d9cd15f53327 100644
--- a/onnxruntime/test/providers/cpu/tensor/pad_test.cc
+++ b/onnxruntime/test/providers/cpu/tensor/pad_test.cc
@@ -66,13 +66,13 @@ static void RunAllOpsetAllDomainPadTests(
     bool pads_is_initializer;
     bool value_is_initializer;
   };
-  const std::vector<TestParams> all_test_params {
-    {false, false},
+  const std::vector<TestParams> all_test_params{
+      {false, false},
 #if (defined(USE_NNAPI) && defined(__ANDROID__)) || (defined(USE_COREML) && defined(__APPLE__))
-        // only enable when building NNAPI EP on Android or building CoreML EP for Apple environment
-        // test runs out of memory in QEMU aarch64 environment, so don't enable otherwise
-        // TODO try to enable when we move from QEMU to arm64 CI machines
-        {true, true},
+      // only enable when building NNAPI EP on Android or building CoreML EP for Apple environment
+      // test runs out of memory in QEMU aarch64 environment, so don't enable otherwise
+      // TODO try to enable when we move from QEMU to arm64 CI machines
+      {true, true},
 #endif
   };
   for (const auto& test_params : all_test_params) {
diff --git a/onnxruntime/test/providers/qnn/qnn_basic_test.cc b/onnxruntime/test/providers/qnn/qnn_basic_test.cc
index 9489d354755e4..9d19c36dc94b2 100644
--- a/onnxruntime/test/providers/qnn/qnn_basic_test.cc
+++ b/onnxruntime/test/providers/qnn/qnn_basic_test.cc
@@ -835,14 +835,14 @@ TEST_F(QnnHTPBackendTests, HTPGraphFinalizationOptimizationModes) {
 
 // Test that models run with various SoC model values
 TEST_F(QnnHTPBackendTests, HTPSocModels) {
-  constexpr std::array<const char*, 3> soc_models = { "",   // No explicit SoC model specified
-                                                      "0",  // "Unknown"
+  constexpr std::array<const char*, 3> soc_models = {"",   // No explicit SoC model specified
+                                                     "0",  // "Unknown"
 #if defined(_M_ARM64)
-                                                      "37" };  // SC8280X
+                                                     "37"};  // SC8280X
 #elif defined(__linux__)
-                                                      "30" };  // SM8350
+                                                     "30"};  // SM8350
 #else
-                                                      "" };
+                                                     ""};
 #endif
 
   for (auto soc_model : soc_models) {
diff --git a/onnxruntime/test/python/onnx_backend_test_series.py b/onnxruntime/test/python/onnx_backend_test_series.py
index 6eebc996fde9c..9b1e87f6ec02e 100644
--- a/onnxruntime/test/python/onnx_backend_test_series.py
+++ b/onnxruntime/test/python/onnx_backend_test_series.py
@@ -76,7 +76,7 @@ def apply_filters(filters, category):
     opset_version = f"opset{onnx.defs.onnx_opset_version()}"
     validated_filters = []
     for f in filters[category]:
-        if type(f) is list:  # noqa: E721
+        if type(f) is list:
             opset_regex = f[0]
             filter_regex = f[1]
             opset_match = re.match(opset_regex, opset_version)
diff --git a/onnxruntime/test/python/transformers/rotary_flash.py b/onnxruntime/test/python/transformers/rotary_flash.py
index 42bff9c92b41b..4329b2c1a6057 100644
--- a/onnxruntime/test/python/transformers/rotary_flash.py
+++ b/onnxruntime/test/python/transformers/rotary_flash.py
@@ -486,9 +486,6 @@ def backward(ctx, dkv):
         return dkv, None, None, None, None
 
 
-apply_rotary_emb_kv_ = ApplyRotaryEmbKV.apply
-
-
 def apply_rotary_emb_kv_(
     kv,
     cos,
diff --git a/onnxruntime/test/python/transformers/test_data/bert_squad_tensorflow2.1_keras2onnx_opset11/generate_tiny_keras2onnx_bert_models.py b/onnxruntime/test/python/transformers/test_data/bert_squad_tensorflow2.1_keras2onnx_opset11/generate_tiny_keras2onnx_bert_models.py
index 0086ce0d289c7..c1e95f35a633b 100644
--- a/onnxruntime/test/python/transformers/test_data/bert_squad_tensorflow2.1_keras2onnx_opset11/generate_tiny_keras2onnx_bert_models.py
+++ b/onnxruntime/test/python/transformers/test_data/bert_squad_tensorflow2.1_keras2onnx_opset11/generate_tiny_keras2onnx_bert_models.py
@@ -343,9 +343,9 @@ def generate_test_data(
         try:
             os.mkdir(path)
         except OSError:
-            print("Creation of the directory %s failed" % path)
+            print(f"Creation of the directory {path} failed")
         else:
-            print("Successfully created the directory %s " % path)
+            print(f"Successfully created the directory {path} ")
 
         if input_tensor_only:
             return
diff --git a/onnxruntime/test/python/transformers/test_data/gpt2_pytorch1.5_opset11/generate_tiny_gpt2_model.py b/onnxruntime/test/python/transformers/test_data/gpt2_pytorch1.5_opset11/generate_tiny_gpt2_model.py
index 065783d5812a8..4a4a0bc2c5098 100644
--- a/onnxruntime/test/python/transformers/test_data/gpt2_pytorch1.5_opset11/generate_tiny_gpt2_model.py
+++ b/onnxruntime/test/python/transformers/test_data/gpt2_pytorch1.5_opset11/generate_tiny_gpt2_model.py
@@ -452,9 +452,9 @@ def generate_test_data(
         try:
             os.mkdir(path)
         except OSError:
-            print("Creation of the directory %s failed" % path)
+            print(f"Creation of the directory {path} failed")
         else:
-            print("Successfully created the directory %s " % path)
+            print(f"Successfully created the directory {path} ")
 
         sess_options = onnxruntime.SessionOptions()
         sess_options.graph_optimization_level = onnxruntime.GraphOptimizationLevel.ORT_DISABLE_ALL
diff --git a/onnxruntime/test/shared_lib/custom_op_utils.h b/onnxruntime/test/shared_lib/custom_op_utils.h
index 8ead4056b1b54..e11540aaa5691 100644
--- a/onnxruntime/test/shared_lib/custom_op_utils.h
+++ b/onnxruntime/test/shared_lib/custom_op_utils.h
@@ -381,9 +381,9 @@ struct StandaloneCustomOp : Ort::CustomOpBase<StandaloneCustomOp, StandaloneCust
 /////////////// structures to test multi-kernls-single-schema ///////////////
 
 struct MulTopKernelFloat {
-  MulTopKernelFloat(const OrtKernelInfo*){};
+  MulTopKernelFloat(const OrtKernelInfo*) {};
   ~MulTopKernelFloat() = default;
-  void Compute(OrtKernelContext*){};
+  void Compute(OrtKernelContext*) {};
 };
 
 struct MulTopOpFloat : Ort::CustomOpBase<MulTopOpFloat, MulTopKernelFloat> {
@@ -397,9 +397,9 @@ struct MulTopOpFloat : Ort::CustomOpBase<MulTopOpFloat, MulTopKernelFloat> {
 };
 
 struct MulTopKernelInt32 {
-  MulTopKernelInt32(const OrtKernelInfo*){};
+  MulTopKernelInt32(const OrtKernelInfo*) {};
   ~MulTopKernelInt32() = default;
-  void Compute(OrtKernelContext*){};
+  void Compute(OrtKernelContext*) {};
 };
 
 struct MulTopOpInt32 : Ort::CustomOpBase<MulTopOpInt32, MulTopKernelInt32> {
@@ -413,9 +413,9 @@ struct MulTopOpInt32 : Ort::CustomOpBase<MulTopOpInt32, MulTopKernelInt32> {
 };
 
 struct MulTopKernelDouble {
-  MulTopKernelDouble(const OrtKernelInfo*){};
+  MulTopKernelDouble(const OrtKernelInfo*) {};
   ~MulTopKernelDouble() = default;
-  void Compute(OrtKernelContext*){};
+  void Compute(OrtKernelContext*) {};
 };
 
 // MulTopOpDouble and MulTopOpFloat has input count mismatch
@@ -430,9 +430,9 @@ struct MulTopOpDouble : Ort::CustomOpBase<MulTopOpDouble, MulTopKernelDouble> {
 };
 
 struct MulTopKernelInt16 {
-  MulTopKernelInt16(const OrtKernelInfo*){};
+  MulTopKernelInt16(const OrtKernelInfo*) {};
   ~MulTopKernelInt16() = default;
-  void Compute(OrtKernelContext*){};
+  void Compute(OrtKernelContext*) {};
 };
 
 // MulTopOpInt16 and MulTopOpFloat has output count mismatch
@@ -448,9 +448,9 @@ struct MulTopOpInt16 : Ort::CustomOpBase<MulTopOpInt16, MulTopKernelInt16> {
 
 // MulTopKernelFloat16 and MulTopOpFloat has input characteristic mismatch
 struct MulTopKernelFloat16 {
-  MulTopKernelFloat16(const OrtKernelInfo*){};
+  MulTopKernelFloat16(const OrtKernelInfo*) {};
   ~MulTopKernelFloat16() = default;
-  void Compute(OrtKernelContext*){};
+  void Compute(OrtKernelContext*) {};
 };
 
 struct MulTopOpFloat16 : Ort::CustomOpBase<MulTopOpFloat16, MulTopKernelFloat16> {
diff --git a/onnxruntime/test/testdata/CNTK/gen.py b/onnxruntime/test/testdata/CNTK/gen.py
index 37241a46808b5..5a3ca461f471a 100644
--- a/onnxruntime/test/testdata/CNTK/gen.py
+++ b/onnxruntime/test/testdata/CNTK/gen.py
@@ -48,10 +48,10 @@ def Save(dir, func, feed, outputs):  # noqa: N802
             if actual_input_name.startswith(cntk_name):
                 cntk_to_actual_names[cntk_name] = actual_input_name
 
-    if type(feed) is not dict:  # noqa: E721
+    if type(feed) is not dict:
         feed = {func.arguments[0]: feed}
 
-    if type(outputs) is not dict:  # noqa: E721
+    if type(outputs) is not dict:
         outputs = {func.outputs[0]: outputs}
 
     test_data_dir = os.path.join(dir, data_dir)
diff --git a/orttraining/orttraining/core/framework/adasum/adasum_mpi.cc b/orttraining/orttraining/core/framework/adasum/adasum_mpi.cc
index 805de812cfa65..dc812ee2aec3f 100644
--- a/orttraining/orttraining/core/framework/adasum/adasum_mpi.cc
+++ b/orttraining/orttraining/core/framework/adasum/adasum_mpi.cc
@@ -35,8 +35,7 @@ void AdasumMPI::InitializeVHDDReductionComms(WorkerGroupType worker_group) {
   int nearest_power_2 = 1;
   int log_size;
   for (nearest_power_2 = 1, log_size = 0; (nearest_power_2 << 1) <= size;
-       nearest_power_2 = (nearest_power_2 << 1), log_size++)
-    ;
+       nearest_power_2 = (nearest_power_2 << 1), log_size++);
   int shift_val;
   int level;
   reduction_comms_ = std::make_unique<std::vector<MPI_Comm>>();
diff --git a/orttraining/orttraining/core/framework/pipeline.h b/orttraining/orttraining/core/framework/pipeline.h
index a93ba1081d7df..79701106c9c1d 100644
--- a/orttraining/orttraining/core/framework/pipeline.h
+++ b/orttraining/orttraining/core/framework/pipeline.h
@@ -247,7 +247,7 @@ struct PipelineWorkerState {
 
 struct PipelineWorkerPool {
   PipelineWorkerPool() = default;
-  PipelineWorkerPool(size_t num_workers) : workers(num_workers), worker_states(num_workers){};
+  PipelineWorkerPool(size_t num_workers) : workers(num_workers), worker_states(num_workers) {};
   void Join(size_t worker_id);
   void JoinAll();
 
diff --git a/orttraining/orttraining/core/framework/torch/custom_function_register.h b/orttraining/orttraining/core/framework/torch/custom_function_register.h
index 762258a45221e..ddb838ba6475c 100644
--- a/orttraining/orttraining/core/framework/torch/custom_function_register.h
+++ b/orttraining/orttraining/core/framework/torch/custom_function_register.h
@@ -102,7 +102,7 @@ class OrtTorchFunctionPool final {
   void UnRegisterFunctions();
 
  private:
-  OrtTorchFunctionPool(){};
+  OrtTorchFunctionPool() {};
   ORT_DISALLOW_COPY_ASSIGNMENT_AND_MOVE(OrtTorchFunctionPool);
 
   void UnRegisterGlobalFunctions();
diff --git a/orttraining/orttraining/core/framework/torch/gil.h b/orttraining/orttraining/core/framework/torch/gil.h
index c928571d2024a..b14b062785eef 100644
--- a/orttraining/orttraining/core/framework/torch/gil.h
+++ b/orttraining/orttraining/core/framework/torch/gil.h
@@ -13,7 +13,7 @@
 // See https://docs.python.org/3/c-api/init.html#non-python-created-threads for details.
 class GilGuard {
  public:
-  GilGuard() : state_(PyGILState_Ensure()){};
+  GilGuard() : state_(PyGILState_Ensure()) {};
   ~GilGuard() { PyGILState_Release(state_); };
 
  private:
diff --git a/orttraining/orttraining/core/framework/torch/torch_proxy.h b/orttraining/orttraining/core/framework/torch/torch_proxy.h
index b80acd6c4791a..37766e67ef42f 100644
--- a/orttraining/orttraining/core/framework/torch/torch_proxy.h
+++ b/orttraining/orttraining/core/framework/torch/torch_proxy.h
@@ -95,8 +95,8 @@ class TorchProxy {
       std::vector<int64_t>& bw_output_to_input_alias_map);
 
  private:
-  TorchProxy(){};
-  ~TorchProxy(){};
+  TorchProxy() {};
+  ~TorchProxy() {};
 
   ORT_DISALLOW_COPY_ASSIGNMENT_AND_MOVE(TorchProxy);
 
diff --git a/orttraining/orttraining/core/graph/graph_augmenter.h b/orttraining/orttraining/core/graph/graph_augmenter.h
index eb146ca0e84f3..c3b6d227f01fd 100644
--- a/orttraining/orttraining/core/graph/graph_augmenter.h
+++ b/orttraining/orttraining/core/graph/graph_augmenter.h
@@ -33,7 +33,7 @@ struct OpDef {
   OpDef(const std::string& type, const std::string& domain = kOnnxDomain, const int opset_version = 9)
       : type(type),
         domain(domain),
-        opset_version(opset_version){};
+        opset_version(opset_version) {};
 
   std::string type;
   std::string domain;
@@ -52,7 +52,7 @@ struct NodeDef {
                               output_args(output_args),
                               attributes(attributes),
                               name(name),
-                              priority(priority){};
+                              priority(priority) {};
 
   NodeDef(const std::string& op_type,
           const std::vector<ArgDef>& input_args,
@@ -64,7 +64,7 @@ struct NodeDef {
                               output_args(output_args),
                               attributes(attributes),
                               name(name),
-                              priority(priority){};
+                              priority(priority) {};
 
   NodeDef(const OpDef& op_def,
           const std::vector<ArgDef>& input_args,
diff --git a/orttraining/orttraining/core/graph/loss_func/loss_func_common.h b/orttraining/orttraining/core/graph/loss_func/loss_func_common.h
index 2b60280e076aa..61bc0a094dac4 100644
--- a/orttraining/orttraining/core/graph/loss_func/loss_func_common.h
+++ b/orttraining/orttraining/core/graph/loss_func/loss_func_common.h
@@ -21,7 +21,7 @@ struct LossFunctionInfo {
 
 struct ILossFunction {
   virtual GraphAugmenter::GraphDefs operator()(const Graph& graph, const LossFunctionInfo& loss_func_info) = 0;
-  virtual ~ILossFunction(){};
+  virtual ~ILossFunction() {};
 };
 
 TypeProto* GetSparseTypeProto(const NodeArg* input_arg,
diff --git a/orttraining/orttraining/core/graph/pipeline_transformer.cc b/orttraining/orttraining/core/graph/pipeline_transformer.cc
index f989d53aa85d5..3495c3da72e3d 100644
--- a/orttraining/orttraining/core/graph/pipeline_transformer.cc
+++ b/orttraining/orttraining/core/graph/pipeline_transformer.cc
@@ -887,7 +887,7 @@ struct PipelineStageNodeGroup {
   // the consumer nodes of a particular initializer can be more than one, so we need a vector to store those
   // nodes.
   std::vector<Node*> nodes;
-  PipelineStageNodeGroup(const size_t stage, std::vector<Node*>& node_group) : stage_id(stage), nodes(std::move(node_group)){};
+  PipelineStageNodeGroup(const size_t stage, std::vector<Node*>& node_group) : stage_id(stage), nodes(std::move(node_group)) {};
 };
 
 // This function passes through the given initializer across stages specified in node_groups[i].stage_id.
diff --git a/orttraining/orttraining/core/optimizer/megatron_transformer.cc b/orttraining/orttraining/core/optimizer/megatron_transformer.cc
index 4ebea5cf386cc..25e16304789b6 100644
--- a/orttraining/orttraining/core/optimizer/megatron_transformer.cc
+++ b/orttraining/orttraining/core/optimizer/megatron_transformer.cc
@@ -21,7 +21,7 @@ struct OpInfo {
          const size_t output_count = 1) : op_type(op_type),
                                           supported_versions(supported_versions),
                                           domain(domain),
-                                          output_count(output_count){};
+                                          output_count(output_count) {};
 
   std::string op_type;
   std::initializer_list<OperatorSetVersion> supported_versions;
@@ -53,7 +53,7 @@ const OpInfo where_info = OpInfo("Where", opset_v9);
 struct NodeInfo {
   NodeInfo(const std::vector<OpInfo>& op_infos,
            const bool required = true) : op_infos(op_infos),
-                                         required(required){};
+                                         required(required) {};
 
   std::vector<OpInfo> op_infos;
   bool required;
diff --git a/orttraining/orttraining/core/session/training_session.h b/orttraining/orttraining/core/session/training_session.h
index 37b708fb7d1dd..765f88e1c992e 100644
--- a/orttraining/orttraining/core/session/training_session.h
+++ b/orttraining/orttraining/core/session/training_session.h
@@ -46,7 +46,7 @@ class TrainingSession : public InferenceSession {
 
   TrainingSession(const SessionOptions& session_options, const Environment& env)
       : InferenceSession(session_options, env), is_mixed_precision_enabled_(false) {}
-  virtual ~TrainingSession(){};
+  virtual ~TrainingSession() {};
 
   /**
    * The training configuration options.
@@ -215,11 +215,11 @@ class TrainingSession : public InferenceSession {
       // If the edge is unique, i.e. only have one consumer node, or all the edges
       // with the same node_arg_name needs to be cut, specify the node_arg_name
       // suffices.
-      CutEdge(std::string edge) : node_arg_name(edge){};
+      CutEdge(std::string edge) : node_arg_name(edge) {};
       // If the edges with same node_arg_name belongs to different cut, i.e. some of its
       // consumer node belongs to one partition, and some belongs to another, specify
       // the consumer node names which you want to perform the cut on.
-      CutEdge(std::string edge, std::vector<std::string> nodes) : node_arg_name(edge), consumer_nodes(nodes){};
+      CutEdge(std::string edge, std::vector<std::string> nodes) : node_arg_name(edge), consumer_nodes(nodes) {};
     };
     // CutInfo is a group of CutEdges that describes a specific cut that composed of splitting those edges.
     typedef std::vector<CutEdge> CutInfo;
diff --git a/orttraining/orttraining/lazy_tensor/flags.h b/orttraining/orttraining/lazy_tensor/flags.h
index b849f9f9a0a3e..1812466d10346 100644
--- a/orttraining/orttraining/lazy_tensor/flags.h
+++ b/orttraining/orttraining/lazy_tensor/flags.h
@@ -60,7 +60,7 @@ class DynamicSettings {
   }
 
  private:
-  DynamicSettings() : onnx_fusion_status_(true){};
+  DynamicSettings() : onnx_fusion_status_(true) {};
   bool onnx_fusion_status_;
 };
 
diff --git a/orttraining/orttraining/models/bert/main.cc b/orttraining/orttraining/models/bert/main.cc
index 22cdd9351a206..c4c7a98ba116a 100644
--- a/orttraining/orttraining/models/bert/main.cc
+++ b/orttraining/orttraining/models/bert/main.cc
@@ -861,8 +861,7 @@ int main(int argc, char* argv[]) {
   OrtParameters ort_params{};
   RETURN_IF_FAIL(ParseArguments(argc, argv, params, ort_params));
   bool keep_looping = params.debug_break;
-  while (keep_looping)
-    ;
+  while (keep_looping);
 
   // setup logger, be noted: LOGS_DEFAULT must be after logging manager initialization.
   string default_logger_id{"Default"};
diff --git a/orttraining/orttraining/models/pipeline_poc/main.cc b/orttraining/orttraining/models/pipeline_poc/main.cc
index c461e4bbf3600..1b7d6b9ea26f6 100644
--- a/orttraining/orttraining/models/pipeline_poc/main.cc
+++ b/orttraining/orttraining/models/pipeline_poc/main.cc
@@ -86,36 +86,36 @@ int main(int argc, char* argv[]) {
   // setup onnxruntime env
   std::vector<FreeDimensionOverride> overrides = {};
   SessionOptions so = {
-    ExecutionMode::ORT_SEQUENTIAL,     // execution_mode
-    ExecutionOrder::DEFAULT,           // execution_order
-    false,                             // enable_profiling
-    ORT_TSTR(""),                      // optimized_model_filepath
-    true,                              // enable_mem_pattern
-    true,                              // enable_mem_reuse
-    true,                              // enable_cpu_mem_arena
-    ORT_TSTR("onnxruntime_profile_"),  // profile_file_prefix
-    "",                                // session_logid
-    -1,                                // session_log_severity_level
-    0,                                 // session_log_verbosity_level
-    5,                                 // max_num_graph_transformation_steps
-    TransformerLevel::Level1,          // graph_optimization_level
-    {},                                // intra_op_param
-    {},                                // inter_op_param
-    overrides,                         // free_dimension_overrides
-    true,                              // use_per_session_threads
-    true,                              // thread_pool_allow_spinning
-    false,                             // use_deterministic_compute
-    {},                                // session_configurations
-    {},                                // initializers_to_share_map
+      ExecutionMode::ORT_SEQUENTIAL,     // execution_mode
+      ExecutionOrder::DEFAULT,           // execution_order
+      false,                             // enable_profiling
+      ORT_TSTR(""),                      // optimized_model_filepath
+      true,                              // enable_mem_pattern
+      true,                              // enable_mem_reuse
+      true,                              // enable_cpu_mem_arena
+      ORT_TSTR("onnxruntime_profile_"),  // profile_file_prefix
+      "",                                // session_logid
+      -1,                                // session_log_severity_level
+      0,                                 // session_log_verbosity_level
+      5,                                 // max_num_graph_transformation_steps
+      TransformerLevel::Level1,          // graph_optimization_level
+      {},                                // intra_op_param
+      {},                                // inter_op_param
+      overrides,                         // free_dimension_overrides
+      true,                              // use_per_session_threads
+      true,                              // thread_pool_allow_spinning
+      false,                             // use_deterministic_compute
+      {},                                // session_configurations
+      {},                                // initializers_to_share_map
 #if !defined(ORT_MINIMAL_BUILD) && !defined(DISABLE_EXTERNAL_INITIALIZERS)
-    {},  // external_initializers
-    {},  // external_initializer_files
+      {},  // external_initializers
+      {},  // external_initializer_files
 #endif
-    nullptr,  // custom_create_thread_fn
-    nullptr,  // custom_thread_creation_options
-    nullptr,  // custom_join_thread_fn
+      nullptr,  // custom_create_thread_fn
+      nullptr,  // custom_thread_creation_options
+      nullptr,  // custom_join_thread_fn
 #if !defined(ORT_MINIMAL_BUILD) || defined(ORT_MINIMAL_BUILD_CUSTOM_OPS)
-    {},  // custom_op_libs
+      {},  // custom_op_libs
 #endif
   };
 
diff --git a/orttraining/orttraining/models/runner/training_util.h b/orttraining/orttraining/models/runner/training_util.h
index 8c76ce7e50dc9..1499b30180f61 100644
--- a/orttraining/orttraining/models/runner/training_util.h
+++ b/orttraining/orttraining/models/runner/training_util.h
@@ -98,7 +98,7 @@ class RandomDataSet : public DataSet {
       : DataSet(tensor_names),
         num_samples_(num_samples),
         tensor_shapes_(tensor_shapes),
-        tensor_types_(tensor_types){};
+        tensor_types_(tensor_types) {};
 
   virtual ~RandomDataSet() {}
 
@@ -189,7 +189,7 @@ class LossScaler {
         min_loss_scale_(min_loss_scale),
         max_loss_scale_(max_loss_scale),
         loss_scale_(loss_scale),
-        stable_steps_(0){};
+        stable_steps_(0) {};
 
   std::string GetLossScaleInputName() const { return loss_scale_input_name_; }
 
diff --git a/orttraining/orttraining/python/orttraining_pybind_state.cc b/orttraining/orttraining/python/orttraining_pybind_state.cc
index a81ea76e807ca..b2392b68ac43e 100644
--- a/orttraining/orttraining/python/orttraining_pybind_state.cc
+++ b/orttraining/orttraining/python/orttraining_pybind_state.cc
@@ -319,7 +319,7 @@ void addObjectMethodsForTraining(py::module& m) {
     auto& pool = onnxruntime::language_interop_ops::torch::OrtTorchFunctionPool::GetInstance();
     pool.RegisterForwardRunner(function_address);
 #else
-        ORT_UNUSED_PARAMETER(obj);
+    ORT_UNUSED_PARAMETER(obj);
 #endif
   });
   m.def("register_backward_runner", [](py::object obj) -> void {
@@ -328,7 +328,7 @@ void addObjectMethodsForTraining(py::module& m) {
     auto& pool = onnxruntime::language_interop_ops::torch::OrtTorchFunctionPool::GetInstance();
     pool.RegisterBackwardRunner(function_address);
 #else
-        ORT_UNUSED_PARAMETER(obj);
+    ORT_UNUSED_PARAMETER(obj);
 #endif
   });
   m.def("register_torch_autograd_function", [](std::string function_full_qual_name, py::object obj) -> void {
@@ -336,8 +336,8 @@ void addObjectMethodsForTraining(py::module& m) {
     auto& pool = onnxruntime::language_interop_ops::torch::OrtTorchFunctionPool::GetInstance();
     pool.RegisterTorchAutogradFunction(function_full_qual_name, obj.ptr());
 #else
-        ORT_UNUSED_PARAMETER(function_full_qual_name);
-        ORT_UNUSED_PARAMETER(obj);
+    ORT_UNUSED_PARAMETER(function_full_qual_name);
+    ORT_UNUSED_PARAMETER(obj);
 #endif
   });
   m.def("register_shape_inference_function", [](std::string function_full_qual_name, py::object obj) -> void {
@@ -345,8 +345,8 @@ void addObjectMethodsForTraining(py::module& m) {
     auto& pool = onnxruntime::language_interop_ops::torch::OrtTorchFunctionPool::GetInstance();
     pool.RegisterShapeInferenceFunction(function_full_qual_name, obj.ptr());
 #else
-        ORT_UNUSED_PARAMETER(function_full_qual_name);
-        ORT_UNUSED_PARAMETER(obj);
+    ORT_UNUSED_PARAMETER(function_full_qual_name);
+    ORT_UNUSED_PARAMETER(obj);
 #endif
   });
   m.def("get_shape_inference_function", [](std::string function_full_qual_name) -> py::object {
@@ -368,8 +368,8 @@ void addObjectMethodsForTraining(py::module& m) {
     auto& pool = onnxruntime::language_interop_ops::torch::OrtTorchFunctionPool::GetInstance();
     pool.RegisterInputAliasFunction(function_full_qual_name, obj.ptr());
 #else
-        ORT_UNUSED_PARAMETER(function_full_qual_name);
-        ORT_UNUSED_PARAMETER(obj);
+    ORT_UNUSED_PARAMETER(function_full_qual_name);
+    ORT_UNUSED_PARAMETER(obj);
 #endif
   });
   m.def("register_miscellaneous_const_input", [](py::object obj) -> void {
@@ -377,7 +377,7 @@ void addObjectMethodsForTraining(py::module& m) {
     auto& pool = onnxruntime::language_interop_ops::torch::OrtTorchFunctionPool::GetInstance();
     pool.RegisterMiscellaneousConstInput(obj.ptr());
 #else
-        ORT_UNUSED_PARAMETER(obj);
+    ORT_UNUSED_PARAMETER(obj);
 #endif
   });
   m.def("unregister_python_functions", []() -> void {
@@ -391,14 +391,14 @@ void addObjectMethodsForTraining(py::module& m) {
 #ifdef ENABLE_TRAINING_TORCH_INTEROP
     return true;
 #else
-        return false;
+    return false;
 #endif
   });
   m.def("is_triton_enabled", []() -> bool {
 #ifdef ENABLE_TRITON
     return true;
 #else
-        return false;
+    return false;
 #endif
   });
 #ifdef ENABLE_TRITON
@@ -1036,7 +1036,7 @@ void addObjectMethodsForTraining(py::module& m) {
 #ifdef __linux__
           return true;
 #else
-        return false;
+    return false;
 #endif
         });
 #endif
diff --git a/orttraining/orttraining/python/training/ort_triton/kernel/_mm.py b/orttraining/orttraining/python/training/ort_triton/kernel/_mm.py
index a3681a13699a0..1a944082fa4ba 100644
--- a/orttraining/orttraining/python/training/ort_triton/kernel/_mm.py
+++ b/orttraining/orttraining/python/training/ort_triton/kernel/_mm.py
@@ -372,7 +372,7 @@ def _gen_bmm_module(
 ) -> Tuple[str, ModuleType]:
     func_name = gen_unique_name("bmm")
     kwargs = _mm_configs(dtype, m, n, k, trans_a, trans_b, alpha, func_name)
-    batch = batch_a if batch_a >= batch_b else batch_b
+    batch = max(batch_a, batch_b)
     kwargs["stride_aq"] = m * k if batch_a == batch else 0
     kwargs["stride_bq"] = k * n if batch_b == batch else 0
     kwargs["batch"] = batch
diff --git a/orttraining/orttraining/python/training/ortmodule/_utils.py b/orttraining/orttraining/python/training/ortmodule/_utils.py
index c299d1c5db4e7..4787cb31a24fd 100644
--- a/orttraining/orttraining/python/training/ortmodule/_utils.py
+++ b/orttraining/orttraining/python/training/ortmodule/_utils.py
@@ -74,7 +74,7 @@ def _ortvalues_to_torch_tensor(
         return tuple(C.to_aten_ort_device_tensor(ov) for ov in ortvalues)
 
     if not isinstance(ortvalues, C.OrtValueVector):
-        raise TypeError("ortvalues must be an instance of OrtValueVector not %r." % type(ortvalues))
+        raise TypeError(f"ortvalues must be an instance of OrtValueVector not {type(ortvalues)!r}.")
 
     res: List[torch.Tensor] = ortvalues.to_dlpacks(_from_dlpack)
     bool_indices = ortvalues.bool_tensor_indices()
diff --git a/orttraining/orttraining/python/training/ortmodule/torch_cpp_extensions/cpu/torch_interop_utils/ctx_pool.h b/orttraining/orttraining/python/training/ortmodule/torch_cpp_extensions/cpu/torch_interop_utils/ctx_pool.h
index e7b101d987d7a..b62c2c40c30ee 100644
--- a/orttraining/orttraining/python/training/ortmodule/torch_cpp_extensions/cpu/torch_interop_utils/ctx_pool.h
+++ b/orttraining/orttraining/python/training/ortmodule/torch_cpp_extensions/cpu/torch_interop_utils/ctx_pool.h
@@ -58,8 +58,8 @@ class PyNodeSharedPointerPool {
   }
 
  private:
-  PyNodeSharedPointerPool(){};
-  ~PyNodeSharedPointerPool(){};
+  PyNodeSharedPointerPool() {};
+  ~PyNodeSharedPointerPool() {};
 
   PyNodeSharedPointerPool(const PyNodeSharedPointerPool&) = delete;
   PyNodeSharedPointerPool& operator=(const PyNodeSharedPointerPool&) = delete;
diff --git a/orttraining/orttraining/test/distributed/partition_utils.h b/orttraining/orttraining/test/distributed/partition_utils.h
index c22d0a3eb2f93..787a001903cce 100644
--- a/orttraining/orttraining/test/distributed/partition_utils.h
+++ b/orttraining/orttraining/test/distributed/partition_utils.h
@@ -159,7 +159,7 @@ struct PipelineStageNodeGroup {
   // the consumer nodes of a particular initializer can be more than one, so we need a vector to store those
   // nodes.
   std::vector<Node*> nodes;
-  PipelineStageNodeGroup(const size_t stage, std::vector<Node*>& node_group) : stage_id(stage), nodes(std::move(node_group)){};
+  PipelineStageNodeGroup(const size_t stage, std::vector<Node*>& node_group) : stage_id(stage), nodes(std::move(node_group)) {};
 };
 
 // This function passes through the given initializer across stages specified in node_groups[i].stage_id.
diff --git a/orttraining/orttraining/test/python/orttraining_test_hierarchical_ortmodule.py b/orttraining/orttraining/test/python/orttraining_test_hierarchical_ortmodule.py
index 8afbafccb8241..655c9def2c66c 100644
--- a/orttraining/orttraining/test/python/orttraining_test_hierarchical_ortmodule.py
+++ b/orttraining/orttraining/test/python/orttraining_test_hierarchical_ortmodule.py
@@ -198,7 +198,7 @@ def call_backward(y):
             y.sum().backward()
 
     def call_allclose(y, y_ref):
-        assert type(y) == type(y_ref)
+        assert type(y) is type(y_ref)
         if isinstance(y, Iterable):
             for ele, ele_ref in zip(y, y_ref):
                 torch.allclose(ele, ele_ref)
diff --git a/orttraining/orttraining/test/python/orttraining_test_model_transform.py b/orttraining/orttraining/test/python/orttraining_test_model_transform.py
index 095830cd54ab8..6ea81fc6aa089 100644
--- a/orttraining/orttraining/test/python/orttraining_test_model_transform.py
+++ b/orttraining/orttraining/test/python/orttraining_test_model_transform.py
@@ -77,7 +77,7 @@ def fix_transpose(model):
         weight = numpy_helper.to_array(t[1])
         assert len(weight.shape) == 2
         weight = weight.transpose(perm)
-        new_weight = numpy_helper.from_array(weight, "%s_transposed" % t[1].name)
+        new_weight = numpy_helper.from_array(weight, f"{t[1].name}_transposed")
         model.graph.initializer.extend([new_weight])
         replace_input_arg(model, node.output[0], new_weight.name)
 
diff --git a/orttraining/orttraining/test/python/orttraining_test_ortmodule_api.py b/orttraining/orttraining/test/python/orttraining_test_ortmodule_api.py
index fe59c398d7abb..3615a12705241 100644
--- a/orttraining/orttraining/test/python/orttraining_test_ortmodule_api.py
+++ b/orttraining/orttraining/test/python/orttraining_test_ortmodule_api.py
@@ -3976,9 +3976,9 @@ def forward(self, input1, bool_argument, int_argument, float_argument):
                 out = self.relu(out)
             return out
 
-    assert type(bool_argument) is bool  # noqa: E721
-    assert type(int_argument) is int  # noqa: E721
-    assert type(float_argument) is float  # noqa: E721
+    assert type(bool_argument) is bool
+    assert type(int_argument) is int
+    assert type(float_argument) is float
 
     device = "cuda"
     N, D_in, H, D_out = 32, 784, 500, 10  # noqa: N806
@@ -4014,8 +4014,8 @@ def forward(self, input1, bool_argument):
                 out = self.relu(out)
             return out
 
-    assert type(bool_arguments[0]) is bool  # noqa: E721
-    assert type(bool_arguments[1]) is bool  # noqa: E721
+    assert type(bool_arguments[0]) is bool
+    assert type(bool_arguments[1]) is bool
 
     device = "cuda"
     N, D_in, H, D_out = 32, 784, 500, 10  # noqa: N806
@@ -5501,7 +5501,7 @@ def forward(self, x):
             return x[: self.dim, :]
 
     def random_state_equal(a, b):
-        assert type(a) == type(b)
+        assert type(a) is type(b)
         if isinstance(a, tuple):
             assert len(a) == len(b)
             return all([random_state_equal(a_i, b_i) for a_i, b_i in zip(a, b)])
diff --git a/orttraining/orttraining/test/python/orttraining_test_ortmodule_bert_classifier.py b/orttraining/orttraining/test/python/orttraining_test_ortmodule_bert_classifier.py
index a1a7d4660f266..41e1e0f5d0d57 100644
--- a/orttraining/orttraining/test/python/orttraining_test_ortmodule_bert_classifier.py
+++ b/orttraining/orttraining/test/python/orttraining_test_ortmodule_bert_classifier.py
@@ -385,7 +385,7 @@ def main():
     # Set log level
     numeric_level = getattr(logging, args.log_level.upper(), None)
     if not isinstance(numeric_level, int):
-        raise ValueError("Invalid log level: %s" % args.log_level)
+        raise ValueError(f"Invalid log level: {args.log_level}")
     logging.basicConfig(level=numeric_level)
 
     # 2. Dataloader
diff --git a/orttraining/orttraining/test/python/orttraining_test_ortmodule_bert_classifier_autocast.py b/orttraining/orttraining/test/python/orttraining_test_ortmodule_bert_classifier_autocast.py
index 0d5aba1a1a5c4..801eb58727689 100644
--- a/orttraining/orttraining/test/python/orttraining_test_ortmodule_bert_classifier_autocast.py
+++ b/orttraining/orttraining/test/python/orttraining_test_ortmodule_bert_classifier_autocast.py
@@ -385,7 +385,7 @@ def main():
     # Set log level
     numeric_level = getattr(logging, args.log_level.upper(), None)
     if not isinstance(numeric_level, int):
-        raise ValueError("Invalid log level: %s" % args.log_level)
+        raise ValueError(f"Invalid log level: {args.log_level}")
     logging.basicConfig(level=numeric_level)
 
     # 2. Dataloader
diff --git a/orttraining/orttraining/test/python/orttraining_test_ortmodule_deepspeed_zero_stage_1.py b/orttraining/orttraining/test/python/orttraining_test_ortmodule_deepspeed_zero_stage_1.py
index 5b28e9c52b480..5e0a4d38b51d6 100644
--- a/orttraining/orttraining/test/python/orttraining_test_ortmodule_deepspeed_zero_stage_1.py
+++ b/orttraining/orttraining/test/python/orttraining_test_ortmodule_deepspeed_zero_stage_1.py
@@ -219,7 +219,7 @@ def main():
         }
         log_level = log_level_mapping.get(args.log_level.upper(), None)
         if not isinstance(log_level, LogLevel):
-            raise ValueError("Invalid log level: %s" % args.log_level)
+            raise ValueError(f"Invalid log level: {args.log_level}")
         debug_options = DebugOptions(log_level=log_level, save_onnx=args.export_onnx_graphs, onnx_prefix="MNIST")
 
         model = ORTModule(model, debug_options)
diff --git a/orttraining/orttraining/test/python/orttraining_test_ortmodule_onnx_ops.py b/orttraining/orttraining/test/python/orttraining_test_ortmodule_onnx_ops.py
index e1def2022d63f..537dcd2ccdb09 100644
--- a/orttraining/orttraining/test/python/orttraining_test_ortmodule_onnx_ops.py
+++ b/orttraining/orttraining/test/python/orttraining_test_ortmodule_onnx_ops.py
@@ -74,11 +74,11 @@ def run_step(model, x):
         )
         onnx_graph_train = ort_model._torch_module._execution_manager._training_manager._onnx_models.optimized_model
         if debug:
-            with open("debug_%s_ortmodule_infer.onnx" % name, "wb") as f:
+            with open(f"debug_{name}_ortmodule_infer.onnx", "wb") as f:
                 f.write(onnx_graph_inf.SerializeToString())
-            with open("debug_%s_ortmodule_train.onnx" % name, "wb") as f:
+            with open(f"debug_{name}_ortmodule_train.onnx", "wb") as f:
                 f.write(onnx_graph_train.SerializeToString())
-        self.assertIn('op_type: "%s"' % name, str(onnx_graph_inf))
+        self.assertIn(f'op_type: "{name}"', str(onnx_graph_inf))
         for onnx_model in [onnx_graph_inf, onnx_graph_train]:
             for oimp in onnx_model.opset_import:
                 if oimp.domain == "":
@@ -86,10 +86,10 @@ def run_step(model, x):
         if op_grad_type is not None:
             if isinstance(op_grad_type, tuple):
                 text = str(onnx_graph_train)
-                if all(map(lambda op: ('op_type: "%s"' % op) not in text, op_grad_type)):
+                if all(map(lambda op: (f'op_type: "{op}"') not in text, op_grad_type)):
                     raise AssertionError("Operator {} not found in {}.".format(" or ".join(op_grad_type), text))
             else:
-                self.assertIn('op_type: "%s"' % op_grad_type, str(onnx_graph_train))
+                self.assertIn(f'op_type: "{op_grad_type}"', str(onnx_graph_train))
 
     def get_torch_model_name(self, name, device):
         def from_numpy(v, device=None, requires_grad=False):
@@ -137,7 +137,7 @@ def forward(self, input1):
 
             return TestGatherElement, "GatherElementsGrad", dict(rtol=1e-04, atol=1e-05)
 
-        raise AssertionError("Unexpected name=%r." % name)
+        raise AssertionError(f"Unexpected name={name!r}.")
 
     def test_onnx_ops(self):
         for name in ["GatherElements", "Softmax"]:
diff --git a/orttraining/orttraining/test/python/orttraining_test_ortmodule_poc.py b/orttraining/orttraining/test/python/orttraining_test_ortmodule_poc.py
index d6f84d94c2838..5872a69dde876 100644
--- a/orttraining/orttraining/test/python/orttraining_test_ortmodule_poc.py
+++ b/orttraining/orttraining/test/python/orttraining_test_ortmodule_poc.py
@@ -201,7 +201,7 @@ def main():
         # Set log level
         numeric_level = getattr(logging, args.log_level.upper(), None)
         if not isinstance(numeric_level, int):
-            raise ValueError("Invalid log level: %s" % args.log_level)
+            raise ValueError(f"Invalid log level: {args.log_level}")
         logging.basicConfig(level=numeric_level)
     else:
         print("Training MNIST on vanilla PyTorch....")
diff --git a/orttraining/orttraining/test/python/orttraining_test_utilities.py b/orttraining/orttraining/test/python/orttraining_test_utilities.py
index 0892bafcdb95d..faa04f327be7f 100644
--- a/orttraining/orttraining/test/python/orttraining_test_utilities.py
+++ b/orttraining/orttraining/test/python/orttraining_test_utilities.py
@@ -237,7 +237,7 @@ def test_data_flatten_and_unflatten(input_output_map, flag: int):
     flatten_schema = input_output_map[2]
 
     def _recursive_compare(real, expected):
-        assert type(real) == type(expected)
+        assert type(real) is type(expected)
         if isinstance(real, str):
             assert real == expected
         elif isinstance(real, abc.Sequence):
@@ -258,7 +258,7 @@ def _recursive_compare(real, expected):
         out, schema = extract_data_and_schema(raw_data)
         assert all([torch.allclose(o, d) if isinstance(o, torch.Tensor) else o == d for o, d in zip(out, flatten_data)])
         if not isinstance(raw_data, torch.Tensor):
-            assert type(schema) == type(raw_data)
+            assert type(schema) is type(raw_data)
 
         assert str(schema) == str(flatten_schema)
 
diff --git a/orttraining/orttraining/test/training_ops/function_op_test_utils.cc b/orttraining/orttraining/test/training_ops/function_op_test_utils.cc
index 9504ba2c1e69a..3daf6db96e31c 100644
--- a/orttraining/orttraining/test/training_ops/function_op_test_utils.cc
+++ b/orttraining/orttraining/test/training_ops/function_op_test_utils.cc
@@ -72,7 +72,7 @@ void OpFunctionTester::RunFunctionBodyGraphOnCPU(TwoDArray& results) {
   }
 }
 
-OpFunctionTester::~OpFunctionTester(){};
+OpFunctionTester::~OpFunctionTester() {};
 
 template <class T>
 std::unique_ptr<T> CreateOpTester(const onnxruntime::training::OpDef& op_def,
diff --git a/orttraining/orttraining/training_ops/cpu/torch/torch_custom_function_kernel.h b/orttraining/orttraining/training_ops/cpu/torch/torch_custom_function_kernel.h
index f58cd3ecbaeca..850dc6de735f0 100644
--- a/orttraining/orttraining/training_ops/cpu/torch/torch_custom_function_kernel.h
+++ b/orttraining/orttraining/training_ops/cpu/torch/torch_custom_function_kernel.h
@@ -25,7 +25,7 @@ class PythonOp final : public OpKernel, public PythonOpBase {
 // Pytorch's torch.autograd.Function.backward(...) wrapper.
 class PythonOpGrad final : public OpKernel, public PythonOpGradBase {
  public:
-  PythonOpGrad(const OpKernelInfo& info) : OpKernel(info), PythonOpGradBase(info){};
+  PythonOpGrad(const OpKernelInfo& info) : OpKernel(info), PythonOpGradBase(info) {};
   Status Compute(OpKernelContext* context) const override;
 };
 
diff --git a/orttraining/orttraining/training_ops/cuda/cuda_training_kernels.cc b/orttraining/orttraining/training_ops/cuda/cuda_training_kernels.cc
index bcc9a06f5a250..dac1d7a84b9d9 100644
--- a/orttraining/orttraining/training_ops/cuda/cuda_training_kernels.cc
+++ b/orttraining/orttraining/training_ops/cuda/cuda_training_kernels.cc
@@ -271,258 +271,258 @@ class ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, Mega
 
 Status RegisterCudaTrainingKernels(KernelRegistry& kernel_registry) {
   static const BuildKernelCreateInfoFn function_table[] = {
-    BuildKernelCreateInfo<void>,  // default entry to avoid the list become empty after ops-reducing
-
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, View)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, Group)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, PassThrough)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, SGDOptimizer)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, SGDOptimizerV2)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, ReduceSumTraining)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, double, ReduceSumTraining)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, int32_t, ReduceSumTraining)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16, ReduceSumTraining)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, SplitTraining)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, ConcatTraining)>,
-    // Adam
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_int64_t_float_float_float_float_MLFloat16, AdamOptimizer)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16_int64_t_float_MLFloat16_float_float_MLFloat16, AdamOptimizer)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_int64_t_float_MLFloat16_float_float_MLFloat16, AdamOptimizer)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_int64_t_float_float_MLFloat16_MLFloat16_MLFloat16, AdamOptimizer)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_int64_t_float_float_MLFloat16_float_MLFloat16, AdamOptimizer)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16_int64_t_float_MLFloat16_MLFloat16_MLFloat16_MLFloat16, AdamOptimizer)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16_int64_t_float_MLFloat16_MLFloat16_float_MLFloat16, AdamOptimizer)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_int64_t_float_MLFloat16_MLFloat16_MLFloat16_MLFloat16, AdamOptimizer)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_int64_t_float_MLFloat16_MLFloat16_float_MLFloat16, AdamOptimizer)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, AdamWOptimizer)>,
-
-    // Lamb
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_float_float_float_float_MLFloat16, LambOptimizer)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_float_MLFloat16_float_MLFloat16_MLFloat16, LambOptimizer)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_float_MLFloat16_float_float_MLFloat16, LambOptimizer)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, double_double_double_double_double_MLFloat16, LambOptimizer)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16_float_MLFloat16_MLFloat16_MLFloat16_MLFloat16, LambOptimizer)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16_float_MLFloat16_MLFloat16_float_MLFloat16, LambOptimizer)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16_float_MLFloat16_float_MLFloat16_MLFloat16, LambOptimizer)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16_float_MLFloat16_float_float_MLFloat16, LambOptimizer)>,
-
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_float, InPlaceAccumulator)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_MLFloat16, InPlaceAccumulator)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16_MLFloat16, InPlaceAccumulator)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16_float, InPlaceAccumulator)>,
-
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_float, InPlaceAccumulatorV2)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_MLFloat16, InPlaceAccumulatorV2)>,
-
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, ZeroGradient)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16, ZeroGradient)>,
-
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, DropoutGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, BitmaskDropoutGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, BiasSoftmaxDropout)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, SoftmaxDropoutGrad)>,
-
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, int64_t, GatherNDGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, SoftmaxCrossEntropy)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, SoftmaxCrossEntropyGrad)>,
-    // BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, float, int32_t, SparseSoftmaxCrossEntropy)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, float, int64_t, SparseSoftmaxCrossEntropy)>,
-    // BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, float, int32_t, SparseSoftmaxCrossEntropyGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, float, int64_t, SparseSoftmaxCrossEntropyGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, SoftmaxGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, LogSoftmaxGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, SoftmaxGrad_13)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, LogSoftmaxGrad_13)>,
-
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TWO_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, 12, MLFloat16, int64_t, SoftmaxCrossEntropyLoss)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TWO_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, 12, float, int64_t, SoftmaxCrossEntropyLoss)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, MLFloat16, int64_t, SoftmaxCrossEntropyLoss)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, float, int64_t, SoftmaxCrossEntropyLoss)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, BFloat16, int64_t, SoftmaxCrossEntropyLoss)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, int64_t, SoftmaxCrossEntropyLossGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16, int64_t, SoftmaxCrossEntropyLossGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, BFloat16, int64_t, SoftmaxCrossEntropyLossGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16_int64_t_float, SoftmaxCrossEntropyLossInternal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_int64_t_float, SoftmaxCrossEntropyLossInternal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16_int64_t_MLFloat16, SoftmaxCrossEntropyLossInternal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, BFloat16_int64_t_BFloat16, SoftmaxCrossEntropyLossInternal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_int64_t_MLFloat16, SoftmaxCrossEntropyLossInternalGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_int64_t_float, SoftmaxCrossEntropyLossInternalGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16_int64_t_MLFloat16, SoftmaxCrossEntropyLossInternalGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, BFloat16_int64_t_BFloat16, SoftmaxCrossEntropyLossInternalGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_float_float, BatchNormalizationGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, double_double_double, BatchNormalizationGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16_MLFloat16_MLFloat16, BatchNormalizationGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16_MLFloat16_float, BatchNormalizationGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16_float_float, BatchNormalizationGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, ConvGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, double, ConvGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16, ConvGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, ConvTransposeGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, double, ConvTransposeGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16, ConvTransposeGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, GatherGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, DivGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, double, DivGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16, DivGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, GeluGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, double, GeluGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16, GeluGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, FastGeluGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, double, FastGeluGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16, FastGeluGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, BiasGeluGrad_dX)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, BiasFastGeluGrad_dX)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, ReluGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, double, ReluGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16, ReluGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, SigmoidGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, double, SigmoidGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16, SigmoidGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, QuickGeluGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, double, QuickGeluGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16, QuickGeluGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, BFloat16, QuickGeluGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, TanhGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, double, TanhGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16, TanhGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, LeakyReluGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, double, LeakyReluGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16, LeakyReluGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16, IsFinite)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, IsFinite)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, double, IsFinite)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, bool, All)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16, IsAllFinite)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, IsAllFinite)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, double, IsAllFinite)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16, MixedPrecisionScale)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, MixedPrecisionScale)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_float, ReduceAllL2)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16_float, ReduceAllL2)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_MLFloat16, ReduceAllL2)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16_MLFloat16, ReduceAllL2)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_float_float, LayerNormalizationGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, double_double_double, LayerNormalizationGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16_float_MLFloat16, LayerNormalizationGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_float_MLFloat16, LayerNormalizationGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16_float_float, LayerNormalizationGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_float_float, SimplifiedLayerNormalizationGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, double_double_double, SimplifiedLayerNormalizationGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16_float_MLFloat16, SimplifiedLayerNormalizationGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_float_MLFloat16, SimplifiedLayerNormalizationGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16_float_float, SimplifiedLayerNormalizationGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_float_float, InvertibleLayerNormalizationGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, double_double_double, InvertibleLayerNormalizationGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16_float_MLFloat16, InvertibleLayerNormalizationGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_float_MLFloat16, InvertibleLayerNormalizationGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16_float_float, InvertibleLayerNormalizationGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, SliceGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, GatherElementsGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16, Scale)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, Scale)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, double, Scale)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, BFloat16, Scale)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_float_float, BatchNormInternal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, double_double_double, BatchNormInternal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16_MLFloat16_MLFloat16, BatchNormInternal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16_MLFloat16_float, BatchNormInternal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16_float_float, BatchNormInternal)>,
-
-    // Adam
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_int64_t_float_float_float_float_BFloat16, AdamOptimizer)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, BFloat16_int64_t_float_BFloat16_float_float_BFloat16, AdamOptimizer)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_int64_t_float_BFloat16_float_float_BFloat16, AdamOptimizer)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_int64_t_float_float_BFloat16_BFloat16_BFloat16, AdamOptimizer)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_int64_t_float_float_BFloat16_float_BFloat16, AdamOptimizer)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, BFloat16_int64_t_float_BFloat16_BFloat16_BFloat16_BFloat16, AdamOptimizer)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, BFloat16_int64_t_float_BFloat16_BFloat16_float_BFloat16, AdamOptimizer)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_int64_t_float_BFloat16_BFloat16_BFloat16_BFloat16, AdamOptimizer)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_int64_t_float_BFloat16_BFloat16_float_BFloat16, AdamOptimizer)>,
-    // Lamb
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_float_float_float_float_BFloat16, LambOptimizer)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_float_BFloat16_float_BFloat16_BFloat16, LambOptimizer)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_float_BFloat16_float_float_BFloat16, LambOptimizer)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, double_double_double_double_double_BFloat16, LambOptimizer)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, BFloat16_float_BFloat16_BFloat16_BFloat16_BFloat16, LambOptimizer)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, BFloat16_float_BFloat16_BFloat16_float_BFloat16, LambOptimizer)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, BFloat16_float_BFloat16_float_BFloat16_BFloat16, LambOptimizer)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, BFloat16_float_BFloat16_float_float_BFloat16, LambOptimizer)>,
-
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_BFloat16, InPlaceAccumulator)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, BFloat16_BFloat16, InPlaceAccumulator)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, BFloat16_float, InPlaceAccumulator)>,
-
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, BFloat16, MixedPrecisionScale)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, BFloat16_float_BFloat16, LayerNormalizationGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, BFloat16_float_BFloat16, SimplifiedLayerNormalizationGrad)>,
-
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, BFloat16_float, ReduceAllL2)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_BFloat16, ReduceAllL2)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, BFloat16_BFloat16, ReduceAllL2)>,
-
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, InplaceClipGradNorm)>,
-
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, FakeQuant)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, FakeQuantGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, BatchScale)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, PadAndUnflatten)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, FlattenAndUnpad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, ScaledSum)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16, ResizeGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, ResizeGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, double, ResizeGrad)>,
+      BuildKernelCreateInfo<void>,  // default entry to avoid the list become empty after ops-reducing
+
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, View)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, Group)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, PassThrough)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, SGDOptimizer)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, SGDOptimizerV2)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, ReduceSumTraining)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, double, ReduceSumTraining)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, int32_t, ReduceSumTraining)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16, ReduceSumTraining)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, SplitTraining)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, ConcatTraining)>,
+      // Adam
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_int64_t_float_float_float_float_MLFloat16, AdamOptimizer)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16_int64_t_float_MLFloat16_float_float_MLFloat16, AdamOptimizer)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_int64_t_float_MLFloat16_float_float_MLFloat16, AdamOptimizer)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_int64_t_float_float_MLFloat16_MLFloat16_MLFloat16, AdamOptimizer)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_int64_t_float_float_MLFloat16_float_MLFloat16, AdamOptimizer)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16_int64_t_float_MLFloat16_MLFloat16_MLFloat16_MLFloat16, AdamOptimizer)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16_int64_t_float_MLFloat16_MLFloat16_float_MLFloat16, AdamOptimizer)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_int64_t_float_MLFloat16_MLFloat16_MLFloat16_MLFloat16, AdamOptimizer)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_int64_t_float_MLFloat16_MLFloat16_float_MLFloat16, AdamOptimizer)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, AdamWOptimizer)>,
+
+      // Lamb
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_float_float_float_float_MLFloat16, LambOptimizer)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_float_MLFloat16_float_MLFloat16_MLFloat16, LambOptimizer)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_float_MLFloat16_float_float_MLFloat16, LambOptimizer)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, double_double_double_double_double_MLFloat16, LambOptimizer)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16_float_MLFloat16_MLFloat16_MLFloat16_MLFloat16, LambOptimizer)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16_float_MLFloat16_MLFloat16_float_MLFloat16, LambOptimizer)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16_float_MLFloat16_float_MLFloat16_MLFloat16, LambOptimizer)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16_float_MLFloat16_float_float_MLFloat16, LambOptimizer)>,
+
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_float, InPlaceAccumulator)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_MLFloat16, InPlaceAccumulator)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16_MLFloat16, InPlaceAccumulator)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16_float, InPlaceAccumulator)>,
+
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_float, InPlaceAccumulatorV2)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_MLFloat16, InPlaceAccumulatorV2)>,
+
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, ZeroGradient)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16, ZeroGradient)>,
+
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, DropoutGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, BitmaskDropoutGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, BiasSoftmaxDropout)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, SoftmaxDropoutGrad)>,
+
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, int64_t, GatherNDGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, SoftmaxCrossEntropy)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, SoftmaxCrossEntropyGrad)>,
+      // BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, float, int32_t, SparseSoftmaxCrossEntropy)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, float, int64_t, SparseSoftmaxCrossEntropy)>,
+      // BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, float, int32_t, SparseSoftmaxCrossEntropyGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, float, int64_t, SparseSoftmaxCrossEntropyGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, SoftmaxGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, LogSoftmaxGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, SoftmaxGrad_13)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, LogSoftmaxGrad_13)>,
+
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TWO_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, 12, MLFloat16, int64_t, SoftmaxCrossEntropyLoss)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TWO_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, 12, float, int64_t, SoftmaxCrossEntropyLoss)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, MLFloat16, int64_t, SoftmaxCrossEntropyLoss)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, float, int64_t, SoftmaxCrossEntropyLoss)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 13, BFloat16, int64_t, SoftmaxCrossEntropyLoss)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, int64_t, SoftmaxCrossEntropyLossGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16, int64_t, SoftmaxCrossEntropyLossGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, BFloat16, int64_t, SoftmaxCrossEntropyLossGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16_int64_t_float, SoftmaxCrossEntropyLossInternal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_int64_t_float, SoftmaxCrossEntropyLossInternal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16_int64_t_MLFloat16, SoftmaxCrossEntropyLossInternal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, BFloat16_int64_t_BFloat16, SoftmaxCrossEntropyLossInternal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_int64_t_MLFloat16, SoftmaxCrossEntropyLossInternalGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_int64_t_float, SoftmaxCrossEntropyLossInternalGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16_int64_t_MLFloat16, SoftmaxCrossEntropyLossInternalGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, BFloat16_int64_t_BFloat16, SoftmaxCrossEntropyLossInternalGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_float_float, BatchNormalizationGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, double_double_double, BatchNormalizationGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16_MLFloat16_MLFloat16, BatchNormalizationGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16_MLFloat16_float, BatchNormalizationGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16_float_float, BatchNormalizationGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, ConvGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, double, ConvGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16, ConvGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, ConvTransposeGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, double, ConvTransposeGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16, ConvTransposeGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, GatherGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, DivGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, double, DivGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16, DivGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, GeluGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, double, GeluGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16, GeluGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, FastGeluGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, double, FastGeluGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16, FastGeluGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, BiasGeluGrad_dX)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, BiasFastGeluGrad_dX)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, ReluGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, double, ReluGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16, ReluGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, SigmoidGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, double, SigmoidGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16, SigmoidGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, QuickGeluGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, double, QuickGeluGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16, QuickGeluGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, BFloat16, QuickGeluGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, TanhGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, double, TanhGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16, TanhGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, LeakyReluGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, double, LeakyReluGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16, LeakyReluGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16, IsFinite)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, IsFinite)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, double, IsFinite)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, bool, All)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16, IsAllFinite)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, IsAllFinite)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, double, IsAllFinite)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16, MixedPrecisionScale)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, MixedPrecisionScale)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_float, ReduceAllL2)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16_float, ReduceAllL2)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_MLFloat16, ReduceAllL2)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16_MLFloat16, ReduceAllL2)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_float_float, LayerNormalizationGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, double_double_double, LayerNormalizationGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16_float_MLFloat16, LayerNormalizationGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_float_MLFloat16, LayerNormalizationGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16_float_float, LayerNormalizationGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_float_float, SimplifiedLayerNormalizationGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, double_double_double, SimplifiedLayerNormalizationGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16_float_MLFloat16, SimplifiedLayerNormalizationGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_float_MLFloat16, SimplifiedLayerNormalizationGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16_float_float, SimplifiedLayerNormalizationGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_float_float, InvertibleLayerNormalizationGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, double_double_double, InvertibleLayerNormalizationGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16_float_MLFloat16, InvertibleLayerNormalizationGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_float_MLFloat16, InvertibleLayerNormalizationGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16_float_float, InvertibleLayerNormalizationGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, SliceGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, GatherElementsGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16, Scale)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, Scale)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, double, Scale)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, BFloat16, Scale)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_float_float, BatchNormInternal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, double_double_double, BatchNormInternal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16_MLFloat16_MLFloat16, BatchNormInternal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16_MLFloat16_float, BatchNormInternal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16_float_float, BatchNormInternal)>,
+
+      // Adam
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_int64_t_float_float_float_float_BFloat16, AdamOptimizer)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, BFloat16_int64_t_float_BFloat16_float_float_BFloat16, AdamOptimizer)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_int64_t_float_BFloat16_float_float_BFloat16, AdamOptimizer)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_int64_t_float_float_BFloat16_BFloat16_BFloat16, AdamOptimizer)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_int64_t_float_float_BFloat16_float_BFloat16, AdamOptimizer)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, BFloat16_int64_t_float_BFloat16_BFloat16_BFloat16_BFloat16, AdamOptimizer)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, BFloat16_int64_t_float_BFloat16_BFloat16_float_BFloat16, AdamOptimizer)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_int64_t_float_BFloat16_BFloat16_BFloat16_BFloat16, AdamOptimizer)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_int64_t_float_BFloat16_BFloat16_float_BFloat16, AdamOptimizer)>,
+      // Lamb
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_float_float_float_float_BFloat16, LambOptimizer)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_float_BFloat16_float_BFloat16_BFloat16, LambOptimizer)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_float_BFloat16_float_float_BFloat16, LambOptimizer)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, double_double_double_double_double_BFloat16, LambOptimizer)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, BFloat16_float_BFloat16_BFloat16_BFloat16_BFloat16, LambOptimizer)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, BFloat16_float_BFloat16_BFloat16_float_BFloat16, LambOptimizer)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, BFloat16_float_BFloat16_float_BFloat16_BFloat16, LambOptimizer)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, BFloat16_float_BFloat16_float_float_BFloat16, LambOptimizer)>,
+
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_BFloat16, InPlaceAccumulator)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, BFloat16_BFloat16, InPlaceAccumulator)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, BFloat16_float, InPlaceAccumulator)>,
+
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, BFloat16, MixedPrecisionScale)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, BFloat16_float_BFloat16, LayerNormalizationGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, BFloat16_float_BFloat16, SimplifiedLayerNormalizationGrad)>,
+
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, BFloat16_float, ReduceAllL2)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float_BFloat16, ReduceAllL2)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, BFloat16_BFloat16, ReduceAllL2)>,
+
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, InplaceClipGradNorm)>,
+
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, FakeQuant)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, FakeQuantGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, BatchScale)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, PadAndUnflatten)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, FlattenAndUnpad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, ScaledSum)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16, ResizeGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, ResizeGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, double, ResizeGrad)>,
 // the kernels within the following ifdef are not included in a build with
 // --enable_training_ops but without --enable_training
 #ifdef ENABLE_TRAINING
 // P2P communication operators.
 #if defined(ORT_USE_NCCL) || defined(USE_MPI)
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, Send)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, Recv)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, Send)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, Recv)>,
 #endif
 
 #ifdef USE_MPI
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, AdasumAllReduce)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, AdasumAllReduce)>,
 #endif
 
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, RecordEvent)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, WaitEvent)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, YieldOp)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, RecordEvent)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, WaitEvent)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, YieldOp)>,
 
 #ifdef ENABLE_TRITON
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, TritonOp)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, TritonOp)>,
 #endif
 
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, GistBinarizeEncoder)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16, GistBinarizeEncoder)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, double, GistBinarizeEncoder)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, GistBinarizeDecoder)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16, GistBinarizeDecoder)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, double, GistBinarizeDecoder)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, bool, GistPack1Encoder)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, GistPack1Encoder)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, bool, GistPack1Decoder)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, GistPack1Decoder)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, GistPack8Encoder)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16, GistPack8Encoder)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, GistPack8Decoder)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16, GistPack8Decoder)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, GistPack16Encoder)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, GistPack16Decoder)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, GistPackMsfp15Encoder)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, GistPackMsfp15Decoder)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, GistBinarizeEncoder)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16, GistBinarizeEncoder)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, double, GistBinarizeEncoder)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, GistBinarizeDecoder)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16, GistBinarizeDecoder)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, double, GistBinarizeDecoder)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, bool, GistPack1Encoder)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, GistPack1Encoder)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, bool, GistPack1Decoder)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, GistPack1Decoder)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, GistPack8Encoder)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16, GistPack8Encoder)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, GistPack8Decoder)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MLFloat16, GistPack8Decoder)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, GistPack16Encoder)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, GistPack16Decoder)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, GistPackMsfp15Encoder)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, float, GistPackMsfp15Decoder)>,
 
 #ifdef ENABLE_TRAINING_TORCH_INTEROP
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, PythonOp)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, PythonOpGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, PythonOp)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, PythonOpGrad)>,
 #endif
 
 #ifdef ORT_USE_NCCL
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, NcclAllReduce)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, NcclAllGather)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, NcclReduceScatter)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MegatronF)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MegatronG)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, NcclAllReduce)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, NcclAllGather)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, NcclReduceScatter)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MegatronF)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kMSDomain, 1, MegatronG)>,
 #endif
 #endif
   };
diff --git a/orttraining/orttraining/training_ops/rocm/rocm_training_kernels.cc b/orttraining/orttraining/training_ops/rocm/rocm_training_kernels.cc
index 7824e98fe8a53..c570f727f2a92 100644
--- a/orttraining/orttraining/training_ops/rocm/rocm_training_kernels.cc
+++ b/orttraining/orttraining/training_ops/rocm/rocm_training_kernels.cc
@@ -222,207 +222,207 @@ class ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, Mega
 
 Status RegisterRocmTrainingKernels(KernelRegistry& kernel_registry) {
   static const BuildKernelCreateInfoFn function_table[] = {
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, View)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, Group)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, PassThrough)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, SGDOptimizer)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float, ReduceSumTraining)>,
-    // BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, double, ReduceSumTraining)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, int32_t, ReduceSumTraining)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16, ReduceSumTraining)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, SplitTraining)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, ConcatTraining)>,
-    // Adam
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float_int64_t_float_float_float_float_MLFloat16, AdamOptimizer)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16_int64_t_float_MLFloat16_float_float_MLFloat16, AdamOptimizer)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float_int64_t_float_MLFloat16_float_float_MLFloat16, AdamOptimizer)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float_int64_t_float_float_MLFloat16_MLFloat16_MLFloat16, AdamOptimizer)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float_int64_t_float_float_MLFloat16_float_MLFloat16, AdamOptimizer)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16_int64_t_float_MLFloat16_MLFloat16_MLFloat16_MLFloat16, AdamOptimizer)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16_int64_t_float_MLFloat16_MLFloat16_float_MLFloat16, AdamOptimizer)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float_int64_t_float_MLFloat16_MLFloat16_MLFloat16_MLFloat16, AdamOptimizer)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float_int64_t_float_MLFloat16_MLFloat16_float_MLFloat16, AdamOptimizer)>,
-
-    // Lamb
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float_float_float_float_float_MLFloat16, LambOptimizer)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float_float_MLFloat16_float_MLFloat16_MLFloat16, LambOptimizer)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float_float_MLFloat16_float_float_MLFloat16, LambOptimizer)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, double_double_double_double_double_MLFloat16, LambOptimizer)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16_float_MLFloat16_MLFloat16_MLFloat16_MLFloat16, LambOptimizer)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16_float_MLFloat16_MLFloat16_float_MLFloat16, LambOptimizer)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16_float_MLFloat16_float_MLFloat16_MLFloat16, LambOptimizer)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16_float_MLFloat16_float_float_MLFloat16, LambOptimizer)>,
-
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float_float, InPlaceAccumulator)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float_MLFloat16, InPlaceAccumulator)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16_MLFloat16, InPlaceAccumulator)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16_float, InPlaceAccumulator)>,
-
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float, ZeroGradient)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16, ZeroGradient)>,
-
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, DropoutGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, BitmaskDropoutGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, BiasSoftmaxDropout)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, SoftmaxDropoutGrad)>,
-
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, int64_t, GatherNDGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float, SoftmaxCrossEntropy)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float, SoftmaxCrossEntropyGrad)>,
-    // BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 9, float, int32_t, SparseSoftmaxCrossEntropy)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 9, float, int64_t, SparseSoftmaxCrossEntropy)>,
-    // BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 9, float, int32_t, SparseSoftmaxCrossEntropyGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 9, float, int64_t, SparseSoftmaxCrossEntropyGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, SoftmaxGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, LogSoftmaxGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, SoftmaxGrad_13)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, LogSoftmaxGrad_13)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TWO_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 12, 12, MLFloat16, int64_t, SoftmaxCrossEntropyLoss)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TWO_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 12, 12, float, int64_t, SoftmaxCrossEntropyLoss)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 13, MLFloat16, int64_t, SoftmaxCrossEntropyLoss)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 13, float, int64_t, SoftmaxCrossEntropyLoss)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 13, BFloat16, int64_t, SoftmaxCrossEntropyLoss)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float, int64_t, SoftmaxCrossEntropyLossGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16, int64_t, SoftmaxCrossEntropyLossGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, BFloat16, int64_t, SoftmaxCrossEntropyLossGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16_int64_t_float, SoftmaxCrossEntropyLossInternal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float_int64_t_float, SoftmaxCrossEntropyLossInternal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16_int64_t_MLFloat16, SoftmaxCrossEntropyLossInternal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, BFloat16_int64_t_BFloat16, SoftmaxCrossEntropyLossInternal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float_int64_t_MLFloat16, SoftmaxCrossEntropyLossInternalGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float_int64_t_float, SoftmaxCrossEntropyLossInternalGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16_int64_t_MLFloat16, SoftmaxCrossEntropyLossInternalGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, BFloat16_int64_t_BFloat16, SoftmaxCrossEntropyLossInternalGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float_float_float, BatchNormalizationGrad)>,
-    // BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, double_double_double, BatchNormalizationGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16_MLFloat16_MLFloat16, BatchNormalizationGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16_MLFloat16_float, BatchNormalizationGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16_float_float, BatchNormalizationGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float, ConvGrad)>,
-    // BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, double, ConvGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16, ConvGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, GatherGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float, DivGrad)>,
-    // BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, double, DivGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16, DivGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float, GeluGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, double, GeluGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16, GeluGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float, FastGeluGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, double, FastGeluGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16, FastGeluGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, BiasGeluGrad_dX)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, BiasFastGeluGrad_dX)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float, ReluGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, double, ReluGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16, ReluGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float, SigmoidGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, double, SigmoidGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16, SigmoidGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float, QuickGeluGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, double, QuickGeluGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16, QuickGeluGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float, TanhGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, double, TanhGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16, TanhGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float, LeakyReluGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, double, LeakyReluGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16, LeakyReluGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16, IsFinite)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float, IsFinite)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, double, IsFinite)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, bool, All)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16, IsAllFinite)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float, IsAllFinite)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, double, IsAllFinite)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16, MixedPrecisionScale)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float, MixedPrecisionScale)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float_float, ReduceAllL2)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16_float, ReduceAllL2)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float_MLFloat16, ReduceAllL2)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16_MLFloat16, ReduceAllL2)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float_float_float, LayerNormalizationGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, double_double_double, LayerNormalizationGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16_float_MLFloat16, LayerNormalizationGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float_float_MLFloat16, LayerNormalizationGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16_float_float, LayerNormalizationGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float_float_float, SimplifiedLayerNormalizationGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, double_double_double, SimplifiedLayerNormalizationGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16_float_MLFloat16, SimplifiedLayerNormalizationGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float_float_MLFloat16, SimplifiedLayerNormalizationGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16_float_float, SimplifiedLayerNormalizationGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float_float_float, InvertibleLayerNormalizationGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, double_double_double, InvertibleLayerNormalizationGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16_float_MLFloat16, InvertibleLayerNormalizationGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float_float_MLFloat16, InvertibleLayerNormalizationGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16_float_float, InvertibleLayerNormalizationGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, SliceGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, GatherElementsGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16, Scale)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float, Scale)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, double, Scale)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, BFloat16, Scale)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float, GistBinarizeEncoder)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16, GistBinarizeEncoder)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, double, GistBinarizeEncoder)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float, GistBinarizeDecoder)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16, GistBinarizeDecoder)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, double, GistBinarizeDecoder)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, bool, GistPack1Encoder)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float, GistPack1Encoder)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, bool, GistPack1Decoder)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float, GistPack1Decoder)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float, GistPack8Encoder)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16, GistPack8Encoder)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float, GistPack8Decoder)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16, GistPack8Decoder)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float, GistPack16Encoder)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float, GistPack16Decoder)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float, GistPackMsfp15Encoder)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float, GistPackMsfp15Decoder)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float_float_float, BatchNormInternal)>,
-    // BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, double_double_double, BatchNormInternal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16_MLFloat16_MLFloat16, BatchNormInternal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16_MLFloat16_float, BatchNormInternal)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16_float_float, BatchNormInternal)>,
-
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, BFloat16, MixedPrecisionScale)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, BFloat16_float_BFloat16, LayerNormalizationGrad)>,
-
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, BFloat16_float, ReduceAllL2)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float_BFloat16, ReduceAllL2)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, BFloat16_BFloat16, ReduceAllL2)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, PadAndUnflatten)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, FlattenAndUnpad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16, ResizeGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float, ResizeGrad)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, double, ResizeGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, View)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, Group)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, PassThrough)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, SGDOptimizer)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float, ReduceSumTraining)>,
+      // BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, double, ReduceSumTraining)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, int32_t, ReduceSumTraining)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16, ReduceSumTraining)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, SplitTraining)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, ConcatTraining)>,
+      // Adam
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float_int64_t_float_float_float_float_MLFloat16, AdamOptimizer)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16_int64_t_float_MLFloat16_float_float_MLFloat16, AdamOptimizer)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float_int64_t_float_MLFloat16_float_float_MLFloat16, AdamOptimizer)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float_int64_t_float_float_MLFloat16_MLFloat16_MLFloat16, AdamOptimizer)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float_int64_t_float_float_MLFloat16_float_MLFloat16, AdamOptimizer)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16_int64_t_float_MLFloat16_MLFloat16_MLFloat16_MLFloat16, AdamOptimizer)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16_int64_t_float_MLFloat16_MLFloat16_float_MLFloat16, AdamOptimizer)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float_int64_t_float_MLFloat16_MLFloat16_MLFloat16_MLFloat16, AdamOptimizer)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float_int64_t_float_MLFloat16_MLFloat16_float_MLFloat16, AdamOptimizer)>,
+
+      // Lamb
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float_float_float_float_float_MLFloat16, LambOptimizer)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float_float_MLFloat16_float_MLFloat16_MLFloat16, LambOptimizer)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float_float_MLFloat16_float_float_MLFloat16, LambOptimizer)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, double_double_double_double_double_MLFloat16, LambOptimizer)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16_float_MLFloat16_MLFloat16_MLFloat16_MLFloat16, LambOptimizer)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16_float_MLFloat16_MLFloat16_float_MLFloat16, LambOptimizer)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16_float_MLFloat16_float_MLFloat16_MLFloat16, LambOptimizer)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16_float_MLFloat16_float_float_MLFloat16, LambOptimizer)>,
+
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float_float, InPlaceAccumulator)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float_MLFloat16, InPlaceAccumulator)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16_MLFloat16, InPlaceAccumulator)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16_float, InPlaceAccumulator)>,
+
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float, ZeroGradient)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16, ZeroGradient)>,
+
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, DropoutGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, BitmaskDropoutGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, BiasSoftmaxDropout)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, SoftmaxDropoutGrad)>,
+
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, int64_t, GatherNDGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float, SoftmaxCrossEntropy)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float, SoftmaxCrossEntropyGrad)>,
+      // BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 9, float, int32_t, SparseSoftmaxCrossEntropy)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 9, float, int64_t, SparseSoftmaxCrossEntropy)>,
+      // BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 9, float, int32_t, SparseSoftmaxCrossEntropyGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 9, float, int64_t, SparseSoftmaxCrossEntropyGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, SoftmaxGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, LogSoftmaxGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, SoftmaxGrad_13)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, LogSoftmaxGrad_13)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TWO_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 12, 12, MLFloat16, int64_t, SoftmaxCrossEntropyLoss)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_TWO_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 12, 12, float, int64_t, SoftmaxCrossEntropyLoss)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 13, MLFloat16, int64_t, SoftmaxCrossEntropyLoss)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 13, float, int64_t, SoftmaxCrossEntropyLoss)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 13, BFloat16, int64_t, SoftmaxCrossEntropyLoss)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float, int64_t, SoftmaxCrossEntropyLossGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16, int64_t, SoftmaxCrossEntropyLossGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TWO_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, BFloat16, int64_t, SoftmaxCrossEntropyLossGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16_int64_t_float, SoftmaxCrossEntropyLossInternal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float_int64_t_float, SoftmaxCrossEntropyLossInternal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16_int64_t_MLFloat16, SoftmaxCrossEntropyLossInternal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, BFloat16_int64_t_BFloat16, SoftmaxCrossEntropyLossInternal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float_int64_t_MLFloat16, SoftmaxCrossEntropyLossInternalGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float_int64_t_float, SoftmaxCrossEntropyLossInternalGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16_int64_t_MLFloat16, SoftmaxCrossEntropyLossInternalGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, BFloat16_int64_t_BFloat16, SoftmaxCrossEntropyLossInternalGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float_float_float, BatchNormalizationGrad)>,
+      // BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, double_double_double, BatchNormalizationGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16_MLFloat16_MLFloat16, BatchNormalizationGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16_MLFloat16_float, BatchNormalizationGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16_float_float, BatchNormalizationGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float, ConvGrad)>,
+      // BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, double, ConvGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16, ConvGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, GatherGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float, DivGrad)>,
+      // BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, double, DivGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16, DivGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float, GeluGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, double, GeluGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16, GeluGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float, FastGeluGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, double, FastGeluGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16, FastGeluGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, BiasGeluGrad_dX)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, BiasFastGeluGrad_dX)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float, ReluGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, double, ReluGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16, ReluGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float, SigmoidGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, double, SigmoidGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16, SigmoidGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float, QuickGeluGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, double, QuickGeluGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16, QuickGeluGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float, TanhGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, double, TanhGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16, TanhGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float, LeakyReluGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, double, LeakyReluGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16, LeakyReluGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16, IsFinite)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float, IsFinite)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, double, IsFinite)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, bool, All)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16, IsAllFinite)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float, IsAllFinite)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, double, IsAllFinite)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16, MixedPrecisionScale)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float, MixedPrecisionScale)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float_float, ReduceAllL2)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16_float, ReduceAllL2)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float_MLFloat16, ReduceAllL2)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16_MLFloat16, ReduceAllL2)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float_float_float, LayerNormalizationGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, double_double_double, LayerNormalizationGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16_float_MLFloat16, LayerNormalizationGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float_float_MLFloat16, LayerNormalizationGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16_float_float, LayerNormalizationGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float_float_float, SimplifiedLayerNormalizationGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, double_double_double, SimplifiedLayerNormalizationGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16_float_MLFloat16, SimplifiedLayerNormalizationGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float_float_MLFloat16, SimplifiedLayerNormalizationGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16_float_float, SimplifiedLayerNormalizationGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float_float_float, InvertibleLayerNormalizationGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, double_double_double, InvertibleLayerNormalizationGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16_float_MLFloat16, InvertibleLayerNormalizationGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float_float_MLFloat16, InvertibleLayerNormalizationGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16_float_float, InvertibleLayerNormalizationGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, SliceGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, GatherElementsGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16, Scale)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float, Scale)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, double, Scale)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, BFloat16, Scale)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float, GistBinarizeEncoder)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16, GistBinarizeEncoder)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, double, GistBinarizeEncoder)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float, GistBinarizeDecoder)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16, GistBinarizeDecoder)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, double, GistBinarizeDecoder)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, bool, GistPack1Encoder)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float, GistPack1Encoder)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, bool, GistPack1Decoder)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float, GistPack1Decoder)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float, GistPack8Encoder)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16, GistPack8Encoder)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float, GistPack8Decoder)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16, GistPack8Decoder)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float, GistPack16Encoder)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float, GistPack16Decoder)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float, GistPackMsfp15Encoder)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float, GistPackMsfp15Decoder)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float_float_float, BatchNormInternal)>,
+      // BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, double_double_double, BatchNormInternal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16_MLFloat16_MLFloat16, BatchNormInternal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16_MLFloat16_float, BatchNormInternal)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16_float_float, BatchNormInternal)>,
+
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, BFloat16, MixedPrecisionScale)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, BFloat16_float_BFloat16, LayerNormalizationGrad)>,
+
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, BFloat16_float, ReduceAllL2)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float_BFloat16, ReduceAllL2)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, BFloat16_BFloat16, ReduceAllL2)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, PadAndUnflatten)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, FlattenAndUnpad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MLFloat16, ResizeGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, float, ResizeGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, double, ResizeGrad)>,
 
 // P2P communication operators.
 #if defined(ORT_USE_NCCL) || defined(USE_MPI)
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, Send)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, Recv)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, Send)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, Recv)>,
 #endif
 
 #ifdef USE_MPI
   // BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, AdasumAllReduce)>,
 #endif
 
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, RecordEvent)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, WaitEvent)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, YieldOp)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, RecordEvent)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, WaitEvent)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, YieldOp)>,
 
 #ifdef ENABLE_TRAINING_TORCH_INTEROP
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, PythonOp)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, PythonOpGrad)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, PythonOp)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, PythonOpGrad)>,
 #endif
 
 #ifdef ORT_USE_NCCL
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, NcclAllReduce)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, NcclAllGather)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, NcclReduceScatter)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MegatronF)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MegatronG)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, NcclAllReduce)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, NcclAllGather)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, NcclReduceScatter)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MegatronF)>,
+      BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kMSDomain, 1, MegatronG)>,
 #endif
   };
 
diff --git a/orttraining/tools/scripts/gpt2_model_transform.py b/orttraining/tools/scripts/gpt2_model_transform.py
index 294af13fe69b7..50bfda4b407af 100644
--- a/orttraining/tools/scripts/gpt2_model_transform.py
+++ b/orttraining/tools/scripts/gpt2_model_transform.py
@@ -192,7 +192,7 @@ def fix_transpose(model):
         weight = numpy_helper.to_array(t[1])
         assert len(weight.shape) == 2
         weight = weight.transpose(perm)
-        new_weight = numpy_helper.from_array(weight, "%s_transposed" % t[1].name)
+        new_weight = numpy_helper.from_array(weight, f"{t[1].name}_transposed")
         model.graph.initializer.extend([new_weight])
         replace_input_arg(model, node.output[0], new_weight.name)
 
diff --git a/orttraining/tools/scripts/model_transform.py b/orttraining/tools/scripts/model_transform.py
index 2fb1936ff2184..e87429d10bf88 100644
--- a/orttraining/tools/scripts/model_transform.py
+++ b/orttraining/tools/scripts/model_transform.py
@@ -227,7 +227,7 @@ def fix_transpose(model):
         weight = numpy_helper.to_array(t[1])
         assert len(weight.shape) == 2
         weight = weight.transpose(perm)
-        new_weight = numpy_helper.from_array(weight, "%s_transposed" % t[1].name)
+        new_weight = numpy_helper.from_array(weight, f"{t[1].name}_transposed")
         model.graph.initializer.extend([new_weight])
         replace_input_arg(model, node.output[0], new_weight.name)
 
diff --git a/pyproject.toml b/pyproject.toml
index 286e4f12721a2..1c3a719fb544a 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -84,6 +84,7 @@ ignore = [
     "PYI041", # May create confusion
     "PYI024", # May create confusion
     "SIM102", # We don't perfer always combining if branches
+    "SIM103", # Do not collapse if-else
     "SIM108", # We don't encourage ternary operators
     "SIM114", # Don't combine if branches for debugability
     "SIM116", # Don't use dict lookup to replace if-else
diff --git a/requirements-lintrunner.txt b/requirements-lintrunner.txt
index d19ebe379b50b..7d384f7b1df67 100644
--- a/requirements-lintrunner.txt
+++ b/requirements-lintrunner.txt
@@ -1,9 +1,9 @@
 # This file is auto updated by dependabot
-lintrunner-adapters>=0.11.0
+lintrunner-adapters>=0.12.4
 # RUFF
-ruff==0.3.2
+ruff==0.5.4
 # BLACK-ISORT
 black==24.2.0
-isort==5.12.0
+isort==5.13.2
 # CLANGFORMAT
-clang-format==17.0.4
+clang-format==18.1.8
diff --git a/tools/ci_build/build.py b/tools/ci_build/build.py
index 98d9ba22b7190..587d035541c45 100644
--- a/tools/ci_build/build.py
+++ b/tools/ci_build/build.py
@@ -65,7 +65,7 @@ def _check_python_version():
 def _str_to_bool(s):
     """Convert string to bool (in argparse context)."""
     if s.lower() not in ["true", "false"]:
-        raise ValueError("Need bool; got %r" % s)
+        raise ValueError(f"Need bool; got {s!r}")
     return {"true": True, "false": False}[s.lower()]
 
 
@@ -806,7 +806,7 @@ def resolve_executable_path(command_or_path):
 def get_linux_distro():
     try:
         with open("/etc/os-release") as f:
-            dist_info = dict(line.strip().split("=", 1) for line in f.readlines())
+            dist_info = dict(line.strip().split("=", 1) for line in f)
         return dist_info.get("NAME", "").strip('"'), dist_info.get("VERSION", "").strip('"')
     except (OSError, ValueError):
         return "", ""
@@ -1236,7 +1236,7 @@ def generate_build_tree(
         cmake_args += ["-Donnxruntime_USE_FULL_PROTOBUF=ON", "-DProtobuf_USE_STATIC_LIBS=ON"]
 
     if args.use_tvm and args.llvm_path is not None:
-        cmake_args += ["-DLLVM_DIR=%s" % args.llvm_path]
+        cmake_args += [f"-DLLVM_DIR={args.llvm_path}"]
 
     if args.use_cuda and not is_windows():
         nvml_stub_path = cuda_home + "/lib64/stubs"
@@ -1452,7 +1452,7 @@ def generate_build_tree(
     if args.enable_lazy_tensor:
         import torch
 
-        cmake_args += ["-Donnxruntime_PREBUILT_PYTORCH_PATH=%s" % os.path.dirname(torch.__file__)]
+        cmake_args += [f"-Donnxruntime_PREBUILT_PYTORCH_PATH={os.path.dirname(torch.__file__)}"]
         cmake_args += ["-D_GLIBCXX_USE_CXX11_ABI=" + str(int(torch._C._GLIBCXX_USE_CXX11_ABI))]
 
     if args.use_azure:
@@ -1582,7 +1582,7 @@ def generate_build_tree(
                         else:
                             cuda_compile_flags_str = cuda_compile_flags_str + " " + compile_flag
                     if len(cuda_compile_flags_str) != 0:
-                        cudaflags.append('-Xcompiler="%s"' % cuda_compile_flags_str)
+                        cudaflags.append(f'-Xcompiler="{cuda_compile_flags_str}"')
             elif is_linux() or is_macOS():
                 if is_linux():
                     ldflags = ["-Wl,-Bsymbolic-functions", "-Wl,-z,relro", "-Wl,-z,now", "-Wl,-z,noexecstack"]
@@ -1650,16 +1650,16 @@ def generate_build_tree(
         temp_cmake_args = cmake_args.copy()
         if cflags is not None and cxxflags is not None and len(cflags) != 0 and len(cxxflags) != 0:
             temp_cmake_args += [
-                "-DCMAKE_C_FLAGS=%s" % (" ".join(cflags)),
-                "-DCMAKE_CXX_FLAGS=%s" % (" ".join(cxxflags)),
+                "-DCMAKE_C_FLAGS={}".format(" ".join(cflags)),
+                "-DCMAKE_CXX_FLAGS={}".format(" ".join(cxxflags)),
             ]
         if cudaflags is not None and len(cudaflags) != 0:
-            temp_cmake_args += ["-DCMAKE_CUDA_FLAGS_INIT=%s" % (" ".join(cudaflags))]
+            temp_cmake_args += ["-DCMAKE_CUDA_FLAGS_INIT={}".format(" ".join(cudaflags))]
         if ldflags is not None and len(ldflags) != 0:
             temp_cmake_args += [
-                "-DCMAKE_EXE_LINKER_FLAGS_INIT=%s" % (" ".join(ldflags)),
-                "-DCMAKE_MODULE_LINKER_FLAGS_INIT=%s" % (" ".join(ldflags)),
-                "-DCMAKE_SHARED_LINKER_FLAGS_INIT=%s" % (" ".join(ldflags)),
+                "-DCMAKE_EXE_LINKER_FLAGS_INIT={}".format(" ".join(ldflags)),
+                "-DCMAKE_MODULE_LINKER_FLAGS_INIT={}".format(" ".join(ldflags)),
+                "-DCMAKE_SHARED_LINKER_FLAGS_INIT={}".format(" ".join(ldflags)),
             ]
         run_subprocess(
             [
diff --git a/tools/ci_build/gen_def.py b/tools/ci_build/gen_def.py
index fe47d8dbe57fe..c4add6f0e8910 100755
--- a/tools/ci_build/gen_def.py
+++ b/tools/ci_build/gen_def.py
@@ -15,11 +15,11 @@ def parse_arguments():
 
 
 args = parse_arguments()
-print("Generating symbol file for %s" % str(args.config))
+print(f"Generating symbol file for {args.config!s}")
 with open(args.version_file) as f:
     VERSION_STRING = f.read().strip()
 
-print("VERSION:%s" % VERSION_STRING)
+print(f"VERSION:{VERSION_STRING}")
 
 symbols = set()
 for c in args.config:
@@ -41,16 +41,16 @@ def parse_arguments():
     elif args.style == "xcode":
         pass  # xcode compile don't has any header.
     else:
-        file.write("VERS_%s {\n" % VERSION_STRING)
+        file.write(f"VERS_{VERSION_STRING} {{\n")
         file.write(" global:\n")
 
     for symbol in symbols:
         if args.style == "vc":
             file.write(" %s @%d\n" % (symbol, symbol_index))
         elif args.style == "xcode":
-            file.write("_%s\n" % symbol)
+            file.write(f"_{symbol}\n")
         else:
-            file.write("  %s;\n" % symbol)
+            file.write(f"  {symbol};\n")
         symbol_index += 1
 
     if args.style == "gcc":
diff --git a/tools/ci_build/reduce_op_kernels.py b/tools/ci_build/reduce_op_kernels.py
index 6b73b1e063e58..df6bbf7a4058e 100755
--- a/tools/ci_build/reduce_op_kernels.py
+++ b/tools/ci_build/reduce_op_kernels.py
@@ -256,7 +256,7 @@ def _generate_type_control_overrides(ort_root: Path, build_dir: Path, cpp_lines:
         inserted = False
         with open(src) as input, open(target, "w") as output:
             inside_insertion_block = False
-            for line in input.readlines():
+            for line in input:
                 if "@@insertion_point_begin(allowed_types)@@" in line:
                     inside_insertion_block = True
                     output.write(line)
diff --git a/tools/ci_build/replace_urls_in_deps.py b/tools/ci_build/replace_urls_in_deps.py
index ac4f515d5482b..37dad358a6feb 100644
--- a/tools/ci_build/replace_urls_in_deps.py
+++ b/tools/ci_build/replace_urls_in_deps.py
@@ -53,10 +53,10 @@ def main():
         csv_file_path = backup_csv_file_path
     else:
         # Make a copy before modifying it
-        print("Making a copy to %s" % str(backup_csv_file_path))
+        print(f"Making a copy to {backup_csv_file_path!s}")
         shutil.copy(csv_file_path, backup_csv_file_path)
 
-    print("Reading from %s" % str(csv_file_path))
+    print(f"Reading from {csv_file_path!s}")
     # Read the whole file into memory first
     with csv_file_path.open("r", encoding="utf-8") as f:
         depfile_reader = csv.reader(f, delimiter=";")
@@ -69,7 +69,7 @@ def main():
             deps.append(Dep(row[0], row[1], row[2]))
 
     csv_file_path = Path(REPO_DIR) / "cmake" / "deps.txt"
-    print("Writing to %s" % str(csv_file_path))
+    print(f"Writing to {csv_file_path!s}")
     # Write updated content back
     with csv_file_path.open("w", newline="", encoding="utf-8") as f:
         depfile_writer = csv.writer(f, delimiter=";")
diff --git a/tools/ci_build/upload_python_package_to_azure_storage.py b/tools/ci_build/upload_python_package_to_azure_storage.py
index b7969f02e518e..16ff5d1f71611 100755
--- a/tools/ci_build/upload_python_package_to_azure_storage.py
+++ b/tools/ci_build/upload_python_package_to_azure_storage.py
@@ -62,7 +62,7 @@ def upload_whl(python_wheel_path, final_storage=False):
 
         with open(download_path_to_html, "w") as f:
             for item in lines:
-                f.write("%s\n" % item)
+                f.write(f"{item}\n")
     else:
         warnings.warn(f"'{new_line}' exists in {download_path_to_html}. The html file is not updated.")
     run_subprocess(
diff --git a/tools/doc/rename_folders.py b/tools/doc/rename_folders.py
index 90d800f2a4498..587755d101ce2 100644
--- a/tools/doc/rename_folders.py
+++ b/tools/doc/rename_folders.py
@@ -26,7 +26,7 @@ def rename_folder(root):
         full_into = os.path.join(r, into)
         if os.path.exists(full_into):
             raise RuntimeError("%r already exists, previous documentation should be removed.")
-        print("rename %r" % full_src)
+        print(f"rename {full_src!r}")
         os.rename(full_src, full_into)
 
     return renamed
@@ -51,13 +51,13 @@ def replace_files(root, renamed):
                 for k, v in subs.items():
                     if k == v:
                         raise ValueError(f"{k!r} == {v!r}")
-                    if ('"%s' % k) in f[0]:
-                        repl.append((f[0], f[0].replace('"%s' % k, '"%s' % v)))
-                    if ("/%s" % k) in f[0]:
-                        repl.append((f[0], f[0].replace("/%s" % k, "/%s" % v)))
+                    if (f'"{k}') in f[0]:
+                        repl.append((f[0], f[0].replace(f'"{k}', f'"{v}')))
+                    if (f"/{k}") in f[0]:
+                        repl.append((f[0], f[0].replace(f"/{k}", f"/{v}")))
             if len(repl) == 0:
                 continue
-            print("update %r" % full)
+            print(f"update {full!r}")
             for k, v in repl:
                 content = content.replace(k, v)
             with open(full, "w", encoding="utf-8") as f:
@@ -71,7 +71,7 @@ def replace_files(root, renamed):
         root = sys.argv[-1]
     else:
         root = "../../build/docs/html"
-    print("look into %r" % root)
+    print(f"look into {root!r}")
     ren = rename_folder(root)
     if len(ren) == 0:
         ren = [
diff --git a/tools/nuget/generate_nuspec_for_native_nuget.py b/tools/nuget/generate_nuspec_for_native_nuget.py
index 60d1884a9591f..a005bd4c4b89d 100644
--- a/tools/nuget/generate_nuspec_for_native_nuget.py
+++ b/tools/nuget/generate_nuspec_for_native_nuget.py
@@ -67,7 +67,7 @@ def generate_file_list_for_ep(nuget_artifacts_dir, ep, files_list, include_pdbs,
                         and package_name != "Microsoft.ML.OnnxRuntime.Gpu.Linux"
                     ):
                         files_list.append(
-                            '<file src="' + str(child_file) + '" target="runtimes/win-%s/native"/>' % cpu_arch
+                            '<file src="' + str(child_file) + f'" target="runtimes/win-{cpu_arch}/native"/>'
                         )
         for cpu_arch in ["x86_64", "arm64"]:
             if child.name == get_package_name("osx", cpu_arch, ep, is_training_package):
@@ -79,7 +79,7 @@ def generate_file_list_for_ep(nuget_artifacts_dir, ep, files_list, include_pdbs,
                     is_versioned_dylib = re.match(r".*[\.\d+]+\.dylib$", child_file.name)
                     if child_file.is_file() and child_file.suffix == ".dylib" and not is_versioned_dylib:
                         files_list.append(
-                            '<file src="' + str(child_file) + '" target="runtimes/osx-%s/native"/>' % cpu_arch
+                            '<file src="' + str(child_file) + f'" target="runtimes/osx-{cpu_arch}/native"/>'
                         )
         for cpu_arch in ["x64", "aarch64"]:
             if child.name == get_package_name("linux", cpu_arch, ep, is_training_package):
@@ -97,7 +97,7 @@ def generate_file_list_for_ep(nuget_artifacts_dir, ep, files_list, include_pdbs,
                         and package_name != "Microsoft.ML.OnnxRuntime.Gpu.Windows"
                     ):
                         files_list.append(
-                            '<file src="' + str(child_file) + '" target="runtimes/linux-%s/native"/>' % cpu_arch
+                            '<file src="' + str(child_file) + f'" target="runtimes/linux-{cpu_arch}/native"/>'
                         )
 
         if child.name == "onnxruntime-android" or child.name == "onnxruntime-training-android":
diff --git a/tools/python/onnx_test_data_utils.py b/tools/python/onnx_test_data_utils.py
index 56485bb78abbd..d50d610a903b7 100644
--- a/tools/python/onnx_test_data_utils.py
+++ b/tools/python/onnx_test_data_utils.py
@@ -59,7 +59,7 @@ def image_to_numpy(filename, shape, channels_last, add_batch_dim):
         # target size.
         w_ratio = new_w / w
         h_ratio = new_h / h
-        ratio = w_ratio if w_ratio > h_ratio else h_ratio
+        ratio = max(h_ratio, w_ratio)
         interim_w = int(w * ratio)
         interim_h = int(h * ratio)
         img = img.resize((interim_w, interim_h), PIL.Image.ANTIALIAS)
diff --git a/tools/python/util/mobile_helpers/usability_checker.py b/tools/python/util/mobile_helpers/usability_checker.py
index 3d8042ad5412b..a8b5021f1387b 100644
--- a/tools/python/util/mobile_helpers/usability_checker.py
+++ b/tools/python/util/mobile_helpers/usability_checker.py
@@ -29,7 +29,7 @@ def __init__(self, filename):
         self._ops_seen = set()
 
         with open(filename) as f:
-            for line in f.readlines():
+            for line in f:
                 # we're looking for a markdown table with 2 columns. first is op name. second is caveats
                 # op name is domain:op
                 if line.startswith("|"):
diff --git a/tools/python/util/reduced_build_config_parser.py b/tools/python/util/reduced_build_config_parser.py
index cb90026808fde..be39562e2d60d 100644
--- a/tools/python/util/reduced_build_config_parser.py
+++ b/tools/python/util/reduced_build_config_parser.py
@@ -113,7 +113,7 @@ def process_non_op_line(line):
         return False
 
     with open(config_file) as config:
-        for line in [orig_line.strip() for orig_line in config.readlines()]:
+        for line in [orig_line.strip() for orig_line in config]:
             if process_non_op_line(line):
                 continue
 
diff --git a/winml/lib/Api.Image/CpuDetensorizer.h b/winml/lib/Api.Image/CpuDetensorizer.h
index e175fbbb4b6a3..04d828097ff3b 100644
--- a/winml/lib/Api.Image/CpuDetensorizer.h
+++ b/winml/lib/Api.Image/CpuDetensorizer.h
@@ -36,7 +36,8 @@ class CpuDetensorizer {
 
     auto nominalRangeConverter = NominalRangeConverter(pixelRange);
 
-    if (formatFrom == formatTo && (formatFrom == kImageTensorChannelTypeBGR8 || formatFrom == kImageTensorChannelTypeRGB8)) {
+    if (formatFrom == formatTo &&
+        (formatFrom == kImageTensorChannelTypeBGR8 || formatFrom == kImageTensorChannelTypeRGB8)) {
       for (uint32_t i = 0; i < tensorHeight; i++) {
         BYTE* pPixel = pData;
 
@@ -52,7 +53,8 @@ class CpuDetensorizer {
 
         pData += bufferWidth;
       }
-    } else if ((formatFrom == kImageTensorChannelTypeRGB8 && formatTo == kImageTensorChannelTypeBGR8) || (formatFrom == kImageTensorChannelTypeBGR8 && formatTo == kImageTensorChannelTypeRGB8)) {
+    } else if ((formatFrom == kImageTensorChannelTypeRGB8 && formatTo == kImageTensorChannelTypeBGR8) ||
+               (formatFrom == kImageTensorChannelTypeBGR8 && formatTo == kImageTensorChannelTypeRGB8)) {
       for (uint32_t i = 0; i < tensorHeight; i++) {
         BYTE* pPixel = pData;
 
@@ -68,7 +70,8 @@ class CpuDetensorizer {
 
         pData += bufferWidth;
       }
-    } else if (formatFrom == kImageTensorChannelTypeGRAY8 && (formatTo == kImageTensorChannelTypeBGR8 || formatTo == kImageTensorChannelTypeRGB8)) {
+    } else if (formatFrom == kImageTensorChannelTypeGRAY8 &&
+               (formatTo == kImageTensorChannelTypeBGR8 || formatTo == kImageTensorChannelTypeRGB8)) {
       // just replicate the gray data across each channel
       for (uint32_t i = 0; i < end; i += bufferWidth) {
         for (uint32_t j = i; j < i + bytesPerRow; j += 4) {
diff --git a/winml/lib/Api.Image/CpuTensorizer.h b/winml/lib/Api.Image/CpuTensorizer.h
index ca5773b28fce2..ed9006470fd0e 100644
--- a/winml/lib/Api.Image/CpuTensorizer.h
+++ b/winml/lib/Api.Image/CpuTensorizer.h
@@ -39,7 +39,8 @@ class CpuTensorizer {
 
     auto nominalRangeConverter = NominalRangeConverter(pixelRange);
 
-    if (formatFrom == kImageTensorChannelTypeBGR8 && formatTo == kImageTensorChannelTypeBGR8 || formatFrom == kImageTensorChannelTypeRGB8 && formatTo == kImageTensorChannelTypeRGB8) {
+    if (formatFrom == kImageTensorChannelTypeBGR8 && formatTo == kImageTensorChannelTypeBGR8 ||
+        formatFrom == kImageTensorChannelTypeRGB8 && formatTo == kImageTensorChannelTypeRGB8) {
       // Convert BGR8 -> BGR8 or RGB8 -> RGB8
       for (uint64_t y = 0; y < yElements; y++) {
         DeinterleaveRowByteToFloat(
@@ -52,7 +53,8 @@ class CpuTensorizer {
           nominalRangeConverter
         );
       }
-    } else if (formatFrom == kImageTensorChannelTypeBGR8 && formatTo == kImageTensorChannelTypeRGB8 || formatFrom == kImageTensorChannelTypeRGB8 && formatTo == kImageTensorChannelTypeBGR8) {
+    } else if (formatFrom == kImageTensorChannelTypeBGR8 && formatTo == kImageTensorChannelTypeRGB8 ||
+               formatFrom == kImageTensorChannelTypeRGB8 && formatTo == kImageTensorChannelTypeBGR8) {
       // Convert RGB8 -> BGR8 or BGR8 -> RGB8
       for (uint32_t y = 0; y < yElements; y++) {
         DeinterleaveRowByteToFloat(
@@ -65,7 +67,8 @@ class CpuTensorizer {
           nominalRangeConverter
         );
       }
-    } else if (formatTo == kImageTensorChannelTypeGRAY8 && (formatFrom == kImageTensorChannelTypeBGR8 || formatFrom == kImageTensorChannelTypeRGB8)) {
+    } else if (formatTo == kImageTensorChannelTypeGRAY8 &&
+               (formatFrom == kImageTensorChannelTypeBGR8 || formatFrom == kImageTensorChannelTypeRGB8)) {
       // Convert BGR8 -> GRAY8 or RGB8 -> GRAY8
       uint32_t blueIncrement = formatFrom == kImageTensorChannelTypeBGR8 ? 0 : 2;
       uint32_t redIncrement = formatFrom == kImageTensorChannelTypeBGR8 ? 2 : 0;
@@ -80,7 +83,8 @@ class CpuTensorizer {
           pixelInd++;
         }
       }
-    } else if (formatFrom == kImageTensorChannelTypeGRAY8 && (formatTo == kImageTensorChannelTypeBGR8 || formatTo == kImageTensorChannelTypeRGB8)) {
+    } else if (formatFrom == kImageTensorChannelTypeGRAY8 &&
+               (formatTo == kImageTensorChannelTypeBGR8 || formatTo == kImageTensorChannelTypeRGB8)) {
       // Convert GRAY8 -> BGR8 or GRAY8 -> RGB8
       for (UINT32 i = start; i < end; i += bufferWidth) {
         for (UINT32 j = i; j < i + bytesPerRow; j += bytesPerPixel) {
diff --git a/winml/lib/Api.Image/D3DDeviceCache.cpp b/winml/lib/Api.Image/D3DDeviceCache.cpp
index 977f2ba75216a..549a7bba77ef6 100644
--- a/winml/lib/Api.Image/D3DDeviceCache.cpp
+++ b/winml/lib/Api.Image/D3DDeviceCache.cpp
@@ -349,7 +349,8 @@ ID3D12RootSignature* D3DDeviceCache::GetTensorizeRootSignature() {
       newRootSignature->SetName(L"Tensorize Rootsignature");
     }
 
-    if (InterlockedCompareExchangePointer(tensorize_root_signature_.put_void(), newRootSignature.get(), nullptr) == nullptr) {
+    if (InterlockedCompareExchangePointer(tensorize_root_signature_.put_void(), newRootSignature.get(), nullptr) ==
+        nullptr) {
       // This thread won the race and just cached the PSO
       newRootSignature.detach();
     }
@@ -401,7 +402,8 @@ ID3D12RootSignature* D3DDeviceCache::GetDetensorizeRootSignature() {
       newRootSignature->SetName(L"Detensorize Rootsignature");
     }
 
-    if (InterlockedCompareExchangePointer(detensorize_root_signature_.put_void(), newRootSignature.get(), nullptr) == nullptr) {
+    if (InterlockedCompareExchangePointer(detensorize_root_signature_.put_void(), newRootSignature.get(), nullptr) ==
+        nullptr) {
       // This thread won the race and just cached the PSO
       newRootSignature.detach();
     }
@@ -416,7 +418,8 @@ ID3D12PipelineState* D3DDeviceCache::GetCachedPipelineState(
   PipelineStateCacheFormat formatTo,
   PipelineStateCacheOperation operation
 ) {
-  if (cached_pipeline_state[static_cast<int>(type)][static_cast<int>(formatFrom)][static_cast<int>(formatTo)][static_cast<int>(operation)] == nullptr) {
+  if (cached_pipeline_state[static_cast<int>(type)][static_cast<int>(formatFrom)][static_cast<int>(formatTo)]
+                           [static_cast<int>(operation)] == nullptr) {
     winrt::com_ptr<ID3D12PipelineState> newPSO;
     if (operation == PipelineStateCacheOperation::kTensorize) {
       newPSO.attach(CreateTensorizePipelineState(type, formatFrom, formatTo));
@@ -425,12 +428,12 @@ ID3D12PipelineState* D3DDeviceCache::GetCachedPipelineState(
     }
 
     if (InterlockedCompareExchangePointer(
-                cached_pipeline_state[static_cast<int>(type)][static_cast<int>(formatFrom)][static_cast<int>(formatTo)]
-                                     [static_cast<int>(operation)]
-                                         .put_void(),
-                newPSO.get(),
-                nullptr
-            ) == nullptr) {
+          cached_pipeline_state[static_cast<int>(type)][static_cast<int>(formatFrom)][static_cast<int>(formatTo)]
+                               [static_cast<int>(operation)]
+                                 .put_void(),
+          newPSO.get(),
+          nullptr
+        ) == nullptr) {
       // This thread won the race and just cached the PSO
       newPSO.detach();
     }
@@ -653,7 +656,8 @@ ID3D12Resource* D3DDeviceCache::GetDetensorizeVertexBuffer(_Out_ UINT* vertexBuf
     memcpy(pVertexDataBegin, triangleVertices, sizeof(triangleVertices));
     newResource->Unmap(0, nullptr);
 
-    if (InterlockedCompareExchangePointer(detensorize_vertex_buffer_.put_void(), newResource.get(), nullptr) == nullptr) {
+    if (InterlockedCompareExchangePointer(detensorize_vertex_buffer_.put_void(), newResource.get(), nullptr) ==
+        nullptr) {
       // This thread won the race and just cached the PSO
       newResource.detach();
     }
diff --git a/winml/lib/Api.Image/EventTimer.h b/winml/lib/Api.Image/EventTimer.h
index 3620a7a2c0ee1..590675646b70d 100644
--- a/winml/lib/Api.Image/EventTimer.h
+++ b/winml/lib/Api.Image/EventTimer.h
@@ -4,7 +4,9 @@ class EventTimer {
  public:
   bool Start() {
     auto now = std::chrono::high_resolution_clock::now();
-    if (!_started || std::chrono::duration_cast<std::chrono::microseconds>(now - _startTime).count() > _kDurationBetweenSendingEvents) {
+    if (!_started ||
+        std::chrono::duration_cast<std::chrono::microseconds>(now - _startTime).count() >
+          _kDurationBetweenSendingEvents) {
       _started = true;
       _startTime = std::chrono::high_resolution_clock::now();
       return true;
diff --git a/winml/lib/Api.Image/ImageConversionHelpers.cpp b/winml/lib/Api.Image/ImageConversionHelpers.cpp
index 11434c5fffb8e..441413bface28 100644
--- a/winml/lib/Api.Image/ImageConversionHelpers.cpp
+++ b/winml/lib/Api.Image/ImageConversionHelpers.cpp
@@ -69,7 +69,8 @@ void _winmli::ConvertVideoFrameToVideoFrame(
   wgdx::Direct3D11::IDirect3DSurface spInputDirect3DSurface = inputVideoFrame.Direct3DSurface();
 
   // only one of softwarebitmap or direct3Dsurface should be non-null
-  if ((spInputSoftwareBitmap == nullptr && spInputDirect3DSurface == nullptr) || (spInputSoftwareBitmap != nullptr && spInputDirect3DSurface != nullptr)) {
+  if ((spInputSoftwareBitmap == nullptr && spInputDirect3DSurface == nullptr) ||
+      (spInputSoftwareBitmap != nullptr && spInputDirect3DSurface != nullptr)) {
     WINML_THROW_HR(E_INVALIDARG);
   }
 
@@ -133,11 +134,9 @@ bool _winmli::NeedsVideoFrameConversion(
 
   if (FAILED((hr = GetVideoFrameInfo(inputVideoFrame, format, width, height, luid)))) {
     bNeedConversion = true;
-  } else if (((int)inputBounds.Width != outputWidth) ||
-            (inputBounds.X != 0) ||
-            ((int)inputBounds.Height != outputHeight) ||
-            (inputBounds.Y != 0) ||
-            (inputVideoFrame == nullptr))  // Check crop
+  } else if (((int)inputBounds.Width != outputWidth) || (inputBounds.X != 0) ||
+             ((int)inputBounds.Height != outputHeight) || (inputBounds.Y != 0) ||
+             (inputVideoFrame == nullptr))  // Check crop
   {
     bNeedConversion = true;
   } else if (luid.HighPart != outputLuid.HighPart || luid.LowPart != outputLuid.LowPart) {
diff --git a/winml/lib/Api.Image/ImageConverter.cpp b/winml/lib/Api.Image/ImageConverter.cpp
index bb97f0ec7ff34..84b6f5a3a4c5c 100644
--- a/winml/lib/Api.Image/ImageConverter.cpp
+++ b/winml/lib/Api.Image/ImageConverter.cpp
@@ -50,7 +50,8 @@ ComPtr<ID3D11Fence> ImageConverter::FetchOrCreateFenceOnDevice(
   ComPtr<ID3D11Fence> fence;
   UINT comPtrSize = static_cast<UINT>(sizeof(fence.GetAddressOf()));
 
-  if (FAILED(pD3D11Device->GetPrivateData(device_cache.GetFenceGuid(), &comPtrSize, fence.GetAddressOf())) || fence.Get() == nullptr) {
+  if (FAILED(pD3D11Device->GetPrivateData(device_cache.GetFenceGuid(), &comPtrSize, fence.GetAddressOf())) ||
+      fence.Get() == nullptr) {
     // There's no fence on the device, so create a new one
     ComPtr<ID3D11Device5> spD3D11Device5;
     WINML_THROW_IF_FAILED(pD3D11Device->QueryInterface(IID_PPV_ARGS(&spD3D11Device5)));
diff --git a/winml/lib/Api.Image/TensorToVideoFrameConverter.cpp b/winml/lib/Api.Image/TensorToVideoFrameConverter.cpp
index 76a5623c5b4a5..456931d21e0a8 100644
--- a/winml/lib/Api.Image/TensorToVideoFrameConverter.cpp
+++ b/winml/lib/Api.Image/TensorToVideoFrameConverter.cpp
@@ -196,8 +196,9 @@ void TensorToVideoFrameConverter::DX12TensorToVideoFrame(
         UINT comPtrSize = static_cast<UINT>(sizeof(spSharedD3D11Texture.GetAddressOf()));
         UINT handleSize = static_cast<UINT>(sizeof(sharedHandle));
 
-        if ((FAILED(spVideoFrameTexture->GetPrivateData(
-                 _d3d11TextureGUID, &comPtrSize, spSharedD3D11Texture.GetAddressOf())) ||
+        if ((FAILED(
+               spVideoFrameTexture->GetPrivateData(_d3d11TextureGUID, &comPtrSize, spSharedD3D11Texture.GetAddressOf())
+             ) ||
              !spSharedD3D11Texture.Get()) ||
             (FAILED(spVideoFrameTexture->GetPrivateData(_handleGUID, &handleSize, &sharedHandle)) ||
              sharedHandle != shared_handle_)) {
@@ -365,7 +366,8 @@ void TensorToVideoFrameConverter::SoftwareTensorToVideoFrame(
   wgdx::Direct3D11::IDirect3DSurface spOutputSurface = pDestVideoFrame.Direct3DSurface();
 
   // only one of softwarebitmap or direct3Dsurface should be non-null
-  if ((spOutputSoftwareBitmap == nullptr && spOutputSurface == nullptr) || (spOutputSoftwareBitmap != nullptr && spOutputSurface != nullptr)) {
+  if ((spOutputSoftwareBitmap == nullptr && spOutputSurface == nullptr) ||
+      (spOutputSoftwareBitmap != nullptr && spOutputSurface != nullptr)) {
     WINML_THROW_HR(E_INVALIDARG);
   }
   if (spOutputSoftwareBitmap) {
@@ -381,7 +383,10 @@ void TensorToVideoFrameConverter::SoftwareTensorToVideoFrame(
   if (_winmli::NeedsVideoFrameConversion(
         pDestVideoFrame, {}, {0, 0, (UINT32)tensorWidth, (UINT32)tensorHeight}, tensorWidth, tensorHeight
       )) {
-    if (converted_video_frame_ == nullptr || _winmli::NeedsVideoFrameConversion(converted_video_frame_, {}, {0, 0, (UINT32)tensorWidth, (UINT32)tensorHeight}, tensorWidth, tensorHeight)) {
+    if (converted_video_frame_ == nullptr ||
+        _winmli::NeedsVideoFrameConversion(
+          converted_video_frame_, {}, {0, 0, (UINT32)tensorWidth, (UINT32)tensorHeight}, tensorWidth, tensorHeight
+        )) {
       converted_video_frame_ = wm::VideoFrame::CreateWithSoftwareBitmap(
         wgi::SoftwareBitmap(wgi::BitmapPixelFormat::Bgra8, tensorWidth, tensorHeight)
       );
diff --git a/winml/lib/Api.Image/VideoFrameToTensorConverter.cpp b/winml/lib/Api.Image/VideoFrameToTensorConverter.cpp
index 0a763c77c94f4..a9b507ae4e16f 100644
--- a/winml/lib/Api.Image/VideoFrameToTensorConverter.cpp
+++ b/winml/lib/Api.Image/VideoFrameToTensorConverter.cpp
@@ -138,14 +138,19 @@ void VideoFrameToTensorConverter::VideoFrameToSoftwareTensor(
   wgdx::Direct3D11::IDirect3DSurface spInputSurface = inputVideoFrame.Direct3DSurface();
 
   // only one of softwarebitmap or direct3Dsurface should be non-null
-  if ((spInputSoftwareBitmap == nullptr && spInputSurface == nullptr) || (spInputSoftwareBitmap != nullptr && spInputSurface != nullptr)) {
+  if ((spInputSoftwareBitmap == nullptr && spInputSurface == nullptr) ||
+      (spInputSoftwareBitmap != nullptr && spInputSurface != nullptr)) {
     WINML_THROW_IF_FAILED(E_INVALIDARG);
   }
 
   UINT32 tensorHeight = static_cast<UINT32>(tensorDesc.sizes[2]);
   UINT32 tensorWidth = static_cast<UINT32>(tensorDesc.sizes[3]);
-  if (spInputSurface || _winmli::NeedsVideoFrameConversion(inputVideoFrame, {}, inputBounds, tensorWidth, tensorHeight)) {
-    if (converted_video_frame_ == nullptr || _winmli::NeedsVideoFrameConversion(converted_video_frame_, {}, {0, 0, (UINT32)tensorWidth, (UINT32)tensorHeight}, tensorWidth, tensorHeight)) {
+  if (spInputSurface ||
+      _winmli::NeedsVideoFrameConversion(inputVideoFrame, {}, inputBounds, tensorWidth, tensorHeight)) {
+    if (converted_video_frame_ == nullptr ||
+        _winmli::NeedsVideoFrameConversion(
+          converted_video_frame_, {}, {0, 0, (UINT32)tensorWidth, (UINT32)tensorHeight}, tensorWidth, tensorHeight
+        )) {
       converted_video_frame_ = wm::VideoFrame::CreateWithSoftwareBitmap(
         wgi::SoftwareBitmap(wgi::BitmapPixelFormat::Bgra8, tensorWidth, tensorHeight)
       );
@@ -236,8 +241,8 @@ void VideoFrameToTensorConverter::VideoFrameToDX12Tensor(
 
     // TODO: Scale during the tensorization phase instead of using the video frame pipeline when the input bounds are not the same size as the tensor
     if (!_winmli::DirectXPixelFormatSupported(spDirect3DSurface.Description().Format) ||
-            static_cast<UINT>(inputBounds.Width) != tensorDesc.sizes[3] ||
-            static_cast<UINT>(inputBounds.Height) != tensorDesc.sizes[2]) {
+        static_cast<UINT>(inputBounds.Width) != tensorDesc.sizes[3] ||
+        static_cast<UINT>(inputBounds.Height) != tensorDesc.sizes[2]) {
       // Force the VideoFrame to not do a conversion if the format is supported since we do it during the tensorization anyway
       wgdx::DirectXPixelFormat newFormat = _winmli::DirectXPixelFormatSupported(spDirect3DSurface.Description().Format)
         ? spDirect3DSurface.Description().Format
@@ -269,7 +274,7 @@ void VideoFrameToTensorConverter::VideoFrameToDX12Tensor(
         D3D11_cached_texture_->GetDesc(&cachedTextureDesc);
 
         if (cachedTextureDesc.Width != scaledBounds.Width || cachedTextureDesc.Height != scaledBounds.Height ||
-                    cachedTextureDesc.Format != videoFrameTextureDesc.Format) {
+            cachedTextureDesc.Format != videoFrameTextureDesc.Format) {
           // The dimensions or format don't match, so we need to re-create our texture
           WINML_THROW_IF_FAILED(
             pDeviceCache->GetD3D11Device()->CreateTexture2D(&videoFrameTextureDesc, nullptr, &D3D11_cached_texture_)
@@ -289,12 +294,12 @@ void VideoFrameToTensorConverter::VideoFrameToDX12Tensor(
       UINT comPtrSize = static_cast<UINT>(sizeof(spSharedD3D11Texture.GetAddressOf()));
       UINT handleSize = static_cast<UINT>(sizeof(sharedHandle));
 
-      if ((FAILED(spVideoFrameTexture->GetPrivateData(
-                     d3d11_texture_GUID_, &comPtrSize, spSharedD3D11Texture.GetAddressOf()
-                 )) ||
-                 !spSharedD3D11Texture.Get()) ||
-                (FAILED(spVideoFrameTexture->GetPrivateData(handle_GUID_, &handleSize, &sharedHandle)) ||
-                 sharedHandle != shared_handle_)) {
+      if ((FAILED(
+             spVideoFrameTexture->GetPrivateData(d3d11_texture_GUID_, &comPtrSize, spSharedD3D11Texture.GetAddressOf())
+           ) ||
+           !spSharedD3D11Texture.Get()) ||
+          (FAILED(spVideoFrameTexture->GetPrivateData(handle_GUID_, &handleSize, &sharedHandle)) ||
+           sharedHandle != shared_handle_)) {
         // Create a new shared texture that we cache on the video frame texture
         WINML_THROW_IF_FAILED(spTextureDevice->CreateTexture2D(&videoFrameTextureDesc, nullptr, &spSharedD3D11Texture));
 
@@ -423,9 +428,9 @@ void VideoFrameToTensorConverter::ConvertDX12TextureToGPUTensor(
     WINML_THROW_IF_FAILED(ULongLongMult(ullNumElementsTensor, uiTensorElementSize, &ullTensorSize));
 
     if (outputDesc.Width < ullTensorSize || outputDesc.Height != 1 ||
-            outputDesc.Dimension != D3D12_RESOURCE_DIMENSION_BUFFER ||
-            !(outputDesc.Flags & D3D12_RESOURCE_FLAG_ALLOW_UNORDERED_ACCESS) ||
-            outputHeapProperties.Type != D3D12_HEAP_TYPE_DEFAULT) {
+        outputDesc.Dimension != D3D12_RESOURCE_DIMENSION_BUFFER ||
+        !(outputDesc.Flags & D3D12_RESOURCE_FLAG_ALLOW_UNORDERED_ACCESS) ||
+        outputHeapProperties.Type != D3D12_HEAP_TYPE_DEFAULT) {
       WINML_THROW_IF_FAILED(E_INVALIDARG);
     }
   }
@@ -565,7 +570,8 @@ void VideoFrameToTensorConverter::ConvertSoftwareBitmapToGPUTensor(
   wgi::BitmapBounds scaledBounds = inputBounds;
 
   // TODO: Scale during the tensorization phase instead of using the video frame pipeline when the input bounds are not the same size as the tensor
-  if (static_cast<UINT>(inputBounds.Width) != tensorDesc.sizes[3] || static_cast<UINT>(inputBounds.Height) != tensorDesc.sizes[2]) {
+  if (static_cast<UINT>(inputBounds.Width) != tensorDesc.sizes[3] ||
+      static_cast<UINT>(inputBounds.Height) != tensorDesc.sizes[2]) {
     scaledBounds = {0, 0, static_cast<uint32_t>(tensorDesc.sizes[3]), static_cast<uint32_t>(tensorDesc.sizes[2])};
 
     // Force the VideoFrame to not do a conversion if the format is supported since we do it during the tensorization anyway
diff --git a/winml/lib/Api.Image/inc/ConverterResourceStore.h b/winml/lib/Api.Image/inc/ConverterResourceStore.h
index ffb413e0b92f3..24406c9fdaaef 100644
--- a/winml/lib/Api.Image/inc/ConverterResourceStore.h
+++ b/winml/lib/Api.Image/inc/ConverterResourceStore.h
@@ -25,7 +25,7 @@ struct ConverterResourceDescription {
     // 2) the resources are on different devices
     // 3) the resources have different pixel formats
     if (desc.width != width || desc.height != height || desc.luid.HighPart != luid.HighPart ||
-            desc.luid.LowPart != luid.LowPart || desc.pixel_format != pixel_format) {
+        desc.luid.LowPart != luid.LowPart || desc.pixel_format != pixel_format) {
       return false;
     }
 
diff --git a/winml/lib/Api/FeatureValues.h b/winml/lib/Api/FeatureValues.h
index a330b244d40fc..fe6429f0a421b 100644
--- a/winml/lib/Api/FeatureValues.h
+++ b/winml/lib/Api/FeatureValues.h
@@ -29,37 +29,37 @@
 #include "ImageFeatureValue.h"
 
 // CREATE_TENSOR is used by data tensor types to implement common functionality
-#define CREATE_TENSOR(type, element_type, element_view_type)                                       \
-  namespace WINMLP {                                                                               \
-  struct type : public _winml::TensorBase<                                                         \
-                  element_type,                                                                    \
-                  element_view_type,                                                               \
-                  type,                                                                            \
-                  I##type,                                                                         \
-                  type##T<type, ITensorNative, _winml::ILotusValueProviderPrivate>> {              \
-    using Base = TensorBase<                                                                       \
-      element_type,                                                                                \
-      element_view_type,                                                                           \
-      type,                                                                                        \
-      I##type,                                                                                     \
-      type##T<type, ITensorNative, _winml::ILotusValueProviderPrivate>>;                           \
-                                                                                                   \
-    type() = default;                                                                              \
-                                                                                                   \
-    type(wfc::IIterable<int64_t> const& shape) : Base(shape){};                                    \
-                                                                                                   \
-    type(std::vector<int64_t> const& shape) : Base(shape){};                                       \
-                                                                                                   \
-    type(std::vector<int64_t> const& shape, ID3D12Resource* pResource) : Base(shape, pResource){}; \
-  };                                                                                               \
-  }                                                                                                \
-  namespace WINML::factory_implementation {                                                        \
-  struct type : type##T<type, winmlp::type, ITensorStaticsNative> {                                \
-    STDMETHOD(CreateFromD3D12Resource)                                                             \
-    (ID3D12Resource * value, __int64* shape, int shapeSize, IUnknown** result) {                   \
-      return winmlp::type::CreateFromD3D12Resource(value, shape, shapeSize, result);               \
-    }                                                                                              \
-  };                                                                                               \
+#define CREATE_TENSOR(type, element_type, element_view_type)                                        \
+  namespace WINMLP {                                                                                \
+  struct type : public _winml::TensorBase<                                                          \
+                  element_type,                                                                     \
+                  element_view_type,                                                                \
+                  type,                                                                             \
+                  I##type,                                                                          \
+                  type##T<type, ITensorNative, _winml::ILotusValueProviderPrivate>> {               \
+    using Base = TensorBase<                                                                        \
+      element_type,                                                                                 \
+      element_view_type,                                                                            \
+      type,                                                                                         \
+      I##type,                                                                                      \
+      type##T<type, ITensorNative, _winml::ILotusValueProviderPrivate>>;                            \
+                                                                                                    \
+    type() = default;                                                                               \
+                                                                                                    \
+    type(wfc::IIterable<int64_t> const& shape) : Base(shape) {};                                    \
+                                                                                                    \
+    type(std::vector<int64_t> const& shape) : Base(shape) {};                                       \
+                                                                                                    \
+    type(std::vector<int64_t> const& shape, ID3D12Resource* pResource) : Base(shape, pResource) {}; \
+  };                                                                                                \
+  }                                                                                                 \
+  namespace WINML::factory_implementation {                                                         \
+  struct type : type##T<type, winmlp::type, ITensorStaticsNative> {                                 \
+    STDMETHOD(CreateFromD3D12Resource)                                                              \
+    (ID3D12Resource * value, __int64* shape, int shapeSize, IUnknown** result) {                    \
+      return winmlp::type::CreateFromD3D12Resource(value, shape, shapeSize, result);                \
+    }                                                                                               \
+  };                                                                                                \
   }
 
 CREATE_TENSOR(TensorBoolean, bool, bool)
@@ -86,11 +86,11 @@ CREATE_TENSOR(TensorString, std::string, winrt::hstring)
 #pragma warning(pop)
 
 // CREATE_MAP is used by map types to implement common functionality
-#define CREATE_MAP(type, key_type, value_type)                                                       \
-  namespace WINMLP {                                                                                 \
-  struct type : public _winml::MapBase<type, key_type, value_type> {                                 \
-    type(wfc::IMap<key_type, value_type> const& data) : MapBase<type, key_type, value_type>(data){}; \
-  };                                                                                                 \
+#define CREATE_MAP(type, key_type, value_type)                                                        \
+  namespace WINMLP {                                                                                  \
+  struct type : public _winml::MapBase<type, key_type, value_type> {                                  \
+    type(wfc::IMap<key_type, value_type> const& data) : MapBase<type, key_type, value_type>(data) {}; \
+  };                                                                                                  \
   }
 
 CREATE_MAP(MapInt64BitToInt64Bit, int64_t, int64_t)
@@ -103,11 +103,11 @@ CREATE_MAP(MapStringToDouble, hstring, double)
 CREATE_MAP(MapStringToString, hstring, hstring)
 
 // CREATE_SEQUENCE is used by sequence types to implement common functionality
-#define CREATE_SEQUENCE(type, element_type, raw_type)                                                    \
-  namespace WINMLP {                                                                                     \
-  struct type : public _winml::SequenceBase<type, element_type, raw_type> {                              \
-    type(wfc::IIterable<element_type> const& data) : SequenceBase<type, element_type, raw_type>(data){}; \
-  };                                                                                                     \
+#define CREATE_SEQUENCE(type, element_type, raw_type)                                                     \
+  namespace WINMLP {                                                                                      \
+  struct type : public _winml::SequenceBase<type, element_type, raw_type> {                               \
+    type(wfc::IIterable<element_type> const& data) : SequenceBase<type, element_type, raw_type>(data) {}; \
+  };                                                                                                      \
   }
 
 using AbiMapStringFloat = wfc::IMap<winrt::hstring, float>;
diff --git a/winml/lib/Api/ImageFeatureValue.cpp b/winml/lib/Api/ImageFeatureValue.cpp
index 8628c578e5004..65f2e56180e19 100644
--- a/winml/lib/Api/ImageFeatureValue.cpp
+++ b/winml/lib/Api/ImageFeatureValue.cpp
@@ -221,7 +221,9 @@ static _winml::ImageTensorDescription CreateImageTensorDescriptor(
     THROW_HR(E_NOTIMPL);
   }
 
-  if (pixelRange != winml::LearningModelPixelRange::ZeroTo255 && pixelRange != winml::LearningModelPixelRange::ZeroToOne && pixelRange != winml::LearningModelPixelRange::MinusOneToOne) {
+  if (pixelRange != winml::LearningModelPixelRange::ZeroTo255 &&
+      pixelRange != winml::LearningModelPixelRange::ZeroToOne &&
+      pixelRange != winml::LearningModelPixelRange::MinusOneToOne) {
     THROW_HR(E_NOTIMPL);
   }
 
@@ -331,12 +333,11 @@ std::optional<ImageFeatureValue::ImageResourceMetadata> ImageFeatureValue::GetIn
     // The the widths and heights of input data must be the same. Or the
     // tensorDescriptor cannot describ the shape of the inputs.
     if (spImageDescriptor->Width() == MAXUINT32 &&
-            !(std::adjacent_find(m_widths.begin(), m_widths.end(), std::not_equal_to<uint32_t>()) == m_widths.end())) {
+        !(std::adjacent_find(m_widths.begin(), m_widths.end(), std::not_equal_to<uint32_t>()) == m_widths.end())) {
       THROW_HR(E_INVALIDARG);
     }
     if (spImageDescriptor->Height() == MAXUINT32 &&
-            !(std::adjacent_find(m_heights.begin(), m_heights.end(), std::not_equal_to<uint32_t>()) == m_heights.end()
-            )) {
+        !(std::adjacent_find(m_heights.begin(), m_heights.end(), std::not_equal_to<uint32_t>()) == m_heights.end())) {
       THROW_HR(E_INVALIDARG);
     }
     descriptorWidth = (spImageDescriptor->Width() == MAXUINT32) ? m_widths[0] : spImageDescriptor->Width();
@@ -354,12 +355,11 @@ std::optional<ImageFeatureValue::ImageResourceMetadata> ImageFeatureValue::GetIn
       return {};
     }
     if (-1 == shape.GetAt(3) &&
-            !(std::adjacent_find(m_widths.begin(), m_widths.end(), std::not_equal_to<uint32_t>()) == m_widths.end())) {
+        !(std::adjacent_find(m_widths.begin(), m_widths.end(), std::not_equal_to<uint32_t>()) == m_widths.end())) {
       THROW_HR(E_INVALIDARG);
     }
     if (-1 == shape.GetAt(2) &&
-            !(std::adjacent_find(m_heights.begin(), m_heights.end(), std::not_equal_to<uint32_t>()) == m_heights.end()
-            )) {
+        !(std::adjacent_find(m_heights.begin(), m_heights.end(), std::not_equal_to<uint32_t>()) == m_heights.end())) {
       THROW_HR(E_INVALIDARG);
     }
     descriptorWidth = (-1 == shape.GetAt(3)) ? m_widths[0] : static_cast<uint32_t>(shape.GetAt(3));
diff --git a/winml/lib/Api/LearningModel.cpp b/winml/lib/Api/LearningModel.cpp
index 6d7c8317ce5f9..8de14a5dfce10 100644
--- a/winml/lib/Api/LearningModel.cpp
+++ b/winml/lib/Api/LearningModel.cpp
@@ -64,7 +64,7 @@ LearningModel::LearningModel(const hstring& path, const winml::ILearningModelOpe
 
   WINML_THROW_IF_FAILED(CreateOnnxruntimeEngineFactory(engine_factory_.put()));
 
-  wil::unique_handle file_handle {
+  wil::unique_handle file_handle{
 #if WINVER >= _WIN32_WINNT_WIN8
     CreateFile2(path.c_str(), GENERIC_READ, FILE_SHARE_READ, OPEN_EXISTING, NULL)
   };
diff --git a/winml/lib/Api/LearningModelSession.cpp b/winml/lib/Api/LearningModelSession.cpp
index 011a4a718f82a..57bafda57fe54 100644
--- a/winml/lib/Api/LearningModelSession.cpp
+++ b/winml/lib/Api/LearningModelSession.cpp
@@ -21,8 +21,8 @@ static const auto c_enable_debug_output = L"EnableDebugOutput";
 namespace guid_details {
 // This GUID is to be used for delimiting ML-related categories of capturable work.
 // {D113B493-BBA2-4993-8608-D706A73B91CE}
-struct __declspec(uuid("D113B493-BBA2-4993-8608-D706A73B91CE")) __declspec(novtable
-) WINML_PIX_EVAL_CAPTURABLE_WORK_GUID {};
+struct __declspec(uuid("D113B493-BBA2-4993-8608-D706A73B91CE"))
+__declspec(novtable) WINML_PIX_EVAL_CAPTURABLE_WORK_GUID {};
 }  // namespace guid_details
 static const GUID WINML_PIX_EVAL_CAPTURABLE_WORK_GUID = __uuidof(guid_details::WINML_PIX_EVAL_CAPTURABLE_WORK_GUID);
 
diff --git a/winml/lib/Api/NumericData.cpp b/winml/lib/Api/NumericData.cpp
index ae5f9155d425c..1e3ba5438c10a 100644
--- a/winml/lib/Api/NumericData.cpp
+++ b/winml/lib/Api/NumericData.cpp
@@ -68,9 +68,7 @@ gsl::span<byte> numeric_data::buffer(bool should_sync_buffer) {
   }
   auto span = combined_buffer();
   if (should_sync_buffer) {
-    _winml::LoadSpanFromDisjointBuffers(
-      buffers_.size(), [this](size_t i) { return buffer_at(i); }, span
-    );
+    _winml::LoadSpanFromDisjointBuffers(buffers_.size(), [this](size_t i) { return buffer_at(i); }, span);
   }
 
   return span;
@@ -80,9 +78,7 @@ bool numeric_data::flush() {
   auto should_flush = buffers_.size() != 1;
   if (should_flush) {
     auto span = combined_buffer();
-    _winml::StoreSpanIntoDisjointBuffers(
-      buffers_.size(), [this](size_t i) { return buffer_at(i); }, span
-    );
+    _winml::StoreSpanIntoDisjointBuffers(buffers_.size(), [this](size_t i) { return buffer_at(i); }, span);
   }
   return should_flush;
 }
@@ -97,9 +93,7 @@ void numeric_data::set(size_t data_size, const byte* data) {
   );
 
   gsl::span<byte> span(const_cast<byte*>(data), data_size);
-  _winml::StoreSpanIntoDisjointBuffers(
-    buffers_.size(), [this](size_t i) { return buffer_at(i); }, span
-  );
+  _winml::StoreSpanIntoDisjointBuffers(buffers_.size(), [this](size_t i) { return buffer_at(i); }, span);
 }
 
 static gsl::span<byte> get_span_from_ibuffer(wss::IBuffer buffer) {
diff --git a/winml/lib/Api/impl/FeatureCompatibility.h b/winml/lib/Api/impl/FeatureCompatibility.h
index 3fff488be23f7..1b124097f3f80 100644
--- a/winml/lib/Api/impl/FeatureCompatibility.h
+++ b/winml/lib/Api/impl/FeatureCompatibility.h
@@ -375,11 +375,11 @@ static void (*FeatureKindCompatibilityMatrix[4][4])(
 ) = {
   //                 Tensor,                          Sequence,                           Map,                    Image
   /* Tensor */ {verify<K::Tensor, K::Tensor>, not_compatible, not_compatible, verify<K::Tensor, K::Image>},
- /* Sequence */
+  /* Sequence */
   {not_compatible, verify<K::Sequence, K::Sequence>, not_compatible, not_compatible},
- /* Map */
+  /* Map */
   {not_compatible, not_compatible, verify<K::Map, K::Map>, not_compatible},
- /* Image */
+  /* Image */
   {verify<K::Image, K::Tensor>, not_compatible, not_compatible, verify<K::Image, K::Image>}
 };
 }  // namespace compatibility_details
diff --git a/winml/lib/Common/CommonDeviceHelpers.cpp b/winml/lib/Common/CommonDeviceHelpers.cpp
index 01615005a8947..b4ada6c498212 100644
--- a/winml/lib/Common/CommonDeviceHelpers.cpp
+++ b/winml/lib/Common/CommonDeviceHelpers.cpp
@@ -65,8 +65,10 @@ HRESULT GetDXCoreAdapterMetadata(
   RETURN_IF_FAILED(spFactory->GetAdapterByLuid(device.GetAdapterLuid(), IID_PPV_ARGS(spAdapter.put())));
 
   if (spAdapter->IsAttributeSupported(DXCORE_ADAPTER_ATTRIBUTE_D3D12_CORE_COMPUTE) &&
-        (!(spAdapter->IsAttributeSupported(DXCORE_ADAPTER_ATTRIBUTE_D3D12_GRAPHICS) ||
-           spAdapter->IsAttributeSupported(DXCORE_ADAPTER_ATTRIBUTE_D3D11_GRAPHICS)))) {
+      (!(
+        spAdapter->IsAttributeSupported(DXCORE_ADAPTER_ATTRIBUTE_D3D12_GRAPHICS) ||
+        spAdapter->IsAttributeSupported(DXCORE_ADAPTER_ATTRIBUTE_D3D11_GRAPHICS)
+      ))) {
     isMcdmAdapter = true;
   } else {
     isMcdmAdapter = false;
diff --git a/winml/test/api/raw/buffer_backed_random_access_stream_reference.h b/winml/test/api/raw/buffer_backed_random_access_stream_reference.h
index e9539c188e45a..6f492bf8340c9 100644
--- a/winml/test/api/raw/buffer_backed_random_access_stream_reference.h
+++ b/winml/test/api/raw/buffer_backed_random_access_stream_reference.h
@@ -347,8 +347,9 @@ struct BufferBackedRandomAccessStreamReference
   }
 
   virtual HRESULT STDMETHODCALLTYPE OpenReadAsync(
-    /* [retval, out] */ __RPC__deref_out_opt
-      __FIAsyncOperation_1_Windows__CStorage__CStreams__CIRandomAccessStreamWithContentType** operation
+    /* [retval, out] */
+    __RPC__deref_out_opt __FIAsyncOperation_1_Windows__CStorage__CStreams__CIRandomAccessStreamWithContentType**
+      operation
   ) override {
     auto open_read_async = Microsoft::WRL::Make<BufferBackedRandomAccessStreamReferenceOpenReadAsync>();
     open_read_async.CopyTo(operation);
diff --git a/winml/test/api/raw/winml_microsoft.h b/winml/test/api/raw/winml_microsoft.h
index 92094188793d5..60527b238d8cd 100644
--- a/winml/test/api/raw/winml_microsoft.h
+++ b/winml/test/api/raw/winml_microsoft.h
@@ -141,8 +141,8 @@ struct TensorRuntimeClassID<double> {
   static const wchar_t* RuntimeClass_ID;
 };
 
-__declspec(selectany
-) const wchar_t* TensorRuntimeClassID<float>::RuntimeClass_ID = RuntimeClass_Microsoft_AI_MachineLearning_TensorFloat;
+__declspec(selectany) const wchar_t* TensorRuntimeClassID<float>::RuntimeClass_ID =
+  RuntimeClass_Microsoft_AI_MachineLearning_TensorFloat;
 __declspec(selectany) const wchar_t* TensorRuntimeClassID<float16>::RuntimeClass_ID =
   RuntimeClass_Microsoft_AI_MachineLearning_TensorFloat16Bit;
 __declspec(selectany) const wchar_t* TensorRuntimeClassID<int8_t>::RuntimeClass_ID =
@@ -161,10 +161,10 @@ __declspec(selectany) const wchar_t* TensorRuntimeClassID<uint64_t>::RuntimeClas
   RuntimeClass_Microsoft_AI_MachineLearning_TensorUInt64Bit;
 __declspec(selectany) const wchar_t* TensorRuntimeClassID<int64_t>::RuntimeClass_ID =
   RuntimeClass_Microsoft_AI_MachineLearning_TensorInt64Bit;
-__declspec(selectany
-) const wchar_t* TensorRuntimeClassID<bool>::RuntimeClass_ID = RuntimeClass_Microsoft_AI_MachineLearning_TensorBoolean;
-__declspec(selectany
-) const wchar_t* TensorRuntimeClassID<double>::RuntimeClass_ID = RuntimeClass_Microsoft_AI_MachineLearning_TensorDouble;
+__declspec(selectany) const wchar_t* TensorRuntimeClassID<bool>::RuntimeClass_ID =
+  RuntimeClass_Microsoft_AI_MachineLearning_TensorBoolean;
+__declspec(selectany) const wchar_t* TensorRuntimeClassID<double>::RuntimeClass_ID =
+  RuntimeClass_Microsoft_AI_MachineLearning_TensorDouble;
 
 template <typename T>
 struct TensorFactory {};
@@ -319,30 +319,30 @@ struct TensorFactoryIID<double> {
   static const GUID IID;
 };
 
-__declspec(selectany
-) const GUID TensorFactoryIID<float>::IID = ABI::Microsoft::AI::MachineLearning::IID_ITensorFloatStatics;
-__declspec(selectany
-) const GUID TensorFactoryIID<float16>::IID = ABI::Microsoft::AI::MachineLearning::IID_ITensorFloat16BitStatics;
-__declspec(selectany
-) const GUID TensorFactoryIID<int8_t>::IID = ABI::Microsoft::AI::MachineLearning::IID_ITensorInt8BitStatics;
-__declspec(selectany
-) const GUID TensorFactoryIID<uint8_t>::IID = ABI::Microsoft::AI::MachineLearning::IID_ITensorUInt8BitStatics;
-__declspec(selectany
-) const GUID TensorFactoryIID<uint16_t>::IID = ABI::Microsoft::AI::MachineLearning::IID_ITensorUInt16BitStatics;
-__declspec(selectany
-) const GUID TensorFactoryIID<int16_t>::IID = ABI::Microsoft::AI::MachineLearning::IID_ITensorInt16BitStatics;
-__declspec(selectany
-) const GUID TensorFactoryIID<uint32_t>::IID = ABI::Microsoft::AI::MachineLearning::IID_ITensorUInt32BitStatics;
-__declspec(selectany
-) const GUID TensorFactoryIID<int32_t>::IID = ABI::Microsoft::AI::MachineLearning::IID_ITensorInt32BitStatics;
-__declspec(selectany
-) const GUID TensorFactoryIID<uint64_t>::IID = ABI::Microsoft::AI::MachineLearning::IID_ITensorUInt64BitStatics;
-__declspec(selectany
-) const GUID TensorFactoryIID<int64_t>::IID = ABI::Microsoft::AI::MachineLearning::IID_ITensorInt64BitStatics;
-__declspec(selectany
-) const GUID TensorFactoryIID<bool>::IID = ABI::Microsoft::AI::MachineLearning::IID_ITensorBooleanStatics;
-__declspec(selectany
-) const GUID TensorFactoryIID<double>::IID = ABI::Microsoft::AI::MachineLearning::IID_ITensorDoubleStatics;
+__declspec(selectany) const GUID TensorFactoryIID<float>::IID =
+  ABI::Microsoft::AI::MachineLearning::IID_ITensorFloatStatics;
+__declspec(selectany) const GUID TensorFactoryIID<float16>::IID =
+  ABI::Microsoft::AI::MachineLearning::IID_ITensorFloat16BitStatics;
+__declspec(selectany) const GUID TensorFactoryIID<int8_t>::IID =
+  ABI::Microsoft::AI::MachineLearning::IID_ITensorInt8BitStatics;
+__declspec(selectany) const GUID TensorFactoryIID<uint8_t>::IID =
+  ABI::Microsoft::AI::MachineLearning::IID_ITensorUInt8BitStatics;
+__declspec(selectany) const GUID TensorFactoryIID<uint16_t>::IID =
+  ABI::Microsoft::AI::MachineLearning::IID_ITensorUInt16BitStatics;
+__declspec(selectany) const GUID TensorFactoryIID<int16_t>::IID =
+  ABI::Microsoft::AI::MachineLearning::IID_ITensorInt16BitStatics;
+__declspec(selectany) const GUID TensorFactoryIID<uint32_t>::IID =
+  ABI::Microsoft::AI::MachineLearning::IID_ITensorUInt32BitStatics;
+__declspec(selectany) const GUID TensorFactoryIID<int32_t>::IID =
+  ABI::Microsoft::AI::MachineLearning::IID_ITensorInt32BitStatics;
+__declspec(selectany) const GUID TensorFactoryIID<uint64_t>::IID =
+  ABI::Microsoft::AI::MachineLearning::IID_ITensorUInt64BitStatics;
+__declspec(selectany) const GUID TensorFactoryIID<int64_t>::IID =
+  ABI::Microsoft::AI::MachineLearning::IID_ITensorInt64BitStatics;
+__declspec(selectany) const GUID TensorFactoryIID<bool>::IID =
+  ABI::Microsoft::AI::MachineLearning::IID_ITensorBooleanStatics;
+__declspec(selectany) const GUID TensorFactoryIID<double>::IID =
+  ABI::Microsoft::AI::MachineLearning::IID_ITensorDoubleStatics;
 
 template <typename T>
 struct TensorFactory2IID {};
@@ -395,30 +395,30 @@ struct TensorFactory2IID<double> {
   static const GUID IID;
 };
 
-__declspec(selectany
-) const GUID TensorFactory2IID<float>::IID = ABI::Microsoft::AI::MachineLearning::IID_ITensorFloatStatics2;
-__declspec(selectany
-) const GUID TensorFactory2IID<float16>::IID = ABI::Microsoft::AI::MachineLearning::IID_ITensorFloat16BitStatics2;
-__declspec(selectany
-) const GUID TensorFactory2IID<int8_t>::IID = ABI::Microsoft::AI::MachineLearning::IID_ITensorInt8BitStatics2;
-__declspec(selectany
-) const GUID TensorFactory2IID<uint8_t>::IID = ABI::Microsoft::AI::MachineLearning::IID_ITensorUInt8BitStatics2;
-__declspec(selectany
-) const GUID TensorFactory2IID<uint16_t>::IID = ABI::Microsoft::AI::MachineLearning::IID_ITensorUInt16BitStatics2;
-__declspec(selectany
-) const GUID TensorFactory2IID<int16_t>::IID = ABI::Microsoft::AI::MachineLearning::IID_ITensorInt16BitStatics2;
-__declspec(selectany
-) const GUID TensorFactory2IID<uint32_t>::IID = ABI::Microsoft::AI::MachineLearning::IID_ITensorUInt32BitStatics2;
-__declspec(selectany
-) const GUID TensorFactory2IID<int32_t>::IID = ABI::Microsoft::AI::MachineLearning::IID_ITensorInt32BitStatics2;
-__declspec(selectany
-) const GUID TensorFactory2IID<uint64_t>::IID = ABI::Microsoft::AI::MachineLearning::IID_ITensorUInt64BitStatics2;
-__declspec(selectany
-) const GUID TensorFactory2IID<int64_t>::IID = ABI::Microsoft::AI::MachineLearning::IID_ITensorInt64BitStatics2;
-__declspec(selectany
-) const GUID TensorFactory2IID<bool>::IID = ABI::Microsoft::AI::MachineLearning::IID_ITensorBooleanStatics2;
-__declspec(selectany
-) const GUID TensorFactory2IID<double>::IID = ABI::Microsoft::AI::MachineLearning::IID_ITensorDoubleStatics2;
+__declspec(selectany) const GUID TensorFactory2IID<float>::IID =
+  ABI::Microsoft::AI::MachineLearning::IID_ITensorFloatStatics2;
+__declspec(selectany) const GUID TensorFactory2IID<float16>::IID =
+  ABI::Microsoft::AI::MachineLearning::IID_ITensorFloat16BitStatics2;
+__declspec(selectany) const GUID TensorFactory2IID<int8_t>::IID =
+  ABI::Microsoft::AI::MachineLearning::IID_ITensorInt8BitStatics2;
+__declspec(selectany) const GUID TensorFactory2IID<uint8_t>::IID =
+  ABI::Microsoft::AI::MachineLearning::IID_ITensorUInt8BitStatics2;
+__declspec(selectany) const GUID TensorFactory2IID<uint16_t>::IID =
+  ABI::Microsoft::AI::MachineLearning::IID_ITensorUInt16BitStatics2;
+__declspec(selectany) const GUID TensorFactory2IID<int16_t>::IID =
+  ABI::Microsoft::AI::MachineLearning::IID_ITensorInt16BitStatics2;
+__declspec(selectany) const GUID TensorFactory2IID<uint32_t>::IID =
+  ABI::Microsoft::AI::MachineLearning::IID_ITensorUInt32BitStatics2;
+__declspec(selectany) const GUID TensorFactory2IID<int32_t>::IID =
+  ABI::Microsoft::AI::MachineLearning::IID_ITensorInt32BitStatics2;
+__declspec(selectany) const GUID TensorFactory2IID<uint64_t>::IID =
+  ABI::Microsoft::AI::MachineLearning::IID_ITensorUInt64BitStatics2;
+__declspec(selectany) const GUID TensorFactory2IID<int64_t>::IID =
+  ABI::Microsoft::AI::MachineLearning::IID_ITensorInt64BitStatics2;
+__declspec(selectany) const GUID TensorFactory2IID<bool>::IID =
+  ABI::Microsoft::AI::MachineLearning::IID_ITensorBooleanStatics2;
+__declspec(selectany) const GUID TensorFactory2IID<double>::IID =
+  ABI::Microsoft::AI::MachineLearning::IID_ITensorDoubleStatics2;
 
 inline HRESULT GetActivationFactory(const wchar_t* p_class_id, const IID& iid, void** factory) noexcept {
   // Fallback to OS binary if the redistributable is not present!
diff --git a/winml/test/api/raw/winml_windows.h b/winml/test/api/raw/winml_windows.h
index 944daff6dd10a..8e72743f3d98b 100644
--- a/winml/test/api/raw/winml_windows.h
+++ b/winml/test/api/raw/winml_windows.h
@@ -141,12 +141,12 @@ struct TensorRuntimeClassID<double> {
   static const wchar_t* RuntimeClass_ID;
 };
 
-__declspec(selectany
-) const wchar_t* TensorRuntimeClassID<float>::RuntimeClass_ID = RuntimeClass_Windows_AI_MachineLearning_TensorFloat;
+__declspec(selectany) const wchar_t* TensorRuntimeClassID<float>::RuntimeClass_ID =
+  RuntimeClass_Windows_AI_MachineLearning_TensorFloat;
 __declspec(selectany) const wchar_t* TensorRuntimeClassID<float16>::RuntimeClass_ID =
   RuntimeClass_Windows_AI_MachineLearning_TensorFloat16Bit;
-__declspec(selectany
-) const wchar_t* TensorRuntimeClassID<int8_t>::RuntimeClass_ID = RuntimeClass_Windows_AI_MachineLearning_TensorInt8Bit;
+__declspec(selectany) const wchar_t* TensorRuntimeClassID<int8_t>::RuntimeClass_ID =
+  RuntimeClass_Windows_AI_MachineLearning_TensorInt8Bit;
 __declspec(selectany) const wchar_t* TensorRuntimeClassID<uint8_t>::RuntimeClass_ID =
   RuntimeClass_Windows_AI_MachineLearning_TensorUInt8Bit;
 __declspec(selectany) const wchar_t* TensorRuntimeClassID<uint16_t>::RuntimeClass_ID =
@@ -161,10 +161,10 @@ __declspec(selectany) const wchar_t* TensorRuntimeClassID<uint64_t>::RuntimeClas
   RuntimeClass_Windows_AI_MachineLearning_TensorUInt64Bit;
 __declspec(selectany) const wchar_t* TensorRuntimeClassID<int64_t>::RuntimeClass_ID =
   RuntimeClass_Windows_AI_MachineLearning_TensorInt64Bit;
-__declspec(selectany
-) const wchar_t* TensorRuntimeClassID<bool>::RuntimeClass_ID = RuntimeClass_Windows_AI_MachineLearning_TensorBoolean;
-__declspec(selectany
-) const wchar_t* TensorRuntimeClassID<double>::RuntimeClass_ID = RuntimeClass_Windows_AI_MachineLearning_TensorDouble;
+__declspec(selectany) const wchar_t* TensorRuntimeClassID<bool>::RuntimeClass_ID =
+  RuntimeClass_Windows_AI_MachineLearning_TensorBoolean;
+__declspec(selectany) const wchar_t* TensorRuntimeClassID<double>::RuntimeClass_ID =
+  RuntimeClass_Windows_AI_MachineLearning_TensorDouble;
 
 template <typename T>
 struct TensorFactory {};
@@ -319,30 +319,30 @@ struct TensorFactoryIID<double> {
   static const GUID IID;
 };
 
-__declspec(selectany
-) const GUID TensorFactoryIID<float>::IID = ABI::Windows::AI::MachineLearning::IID_ITensorFloatStatics;
-__declspec(selectany
-) const GUID TensorFactoryIID<float16>::IID = ABI::Windows::AI::MachineLearning::IID_ITensorFloat16BitStatics;
-__declspec(selectany
-) const GUID TensorFactoryIID<int8_t>::IID = ABI::Windows::AI::MachineLearning::IID_ITensorInt8BitStatics;
-__declspec(selectany
-) const GUID TensorFactoryIID<uint8_t>::IID = ABI::Windows::AI::MachineLearning::IID_ITensorUInt8BitStatics;
-__declspec(selectany
-) const GUID TensorFactoryIID<uint16_t>::IID = ABI::Windows::AI::MachineLearning::IID_ITensorUInt16BitStatics;
-__declspec(selectany
-) const GUID TensorFactoryIID<int16_t>::IID = ABI::Windows::AI::MachineLearning::IID_ITensorInt16BitStatics;
-__declspec(selectany
-) const GUID TensorFactoryIID<uint32_t>::IID = ABI::Windows::AI::MachineLearning::IID_ITensorUInt32BitStatics;
-__declspec(selectany
-) const GUID TensorFactoryIID<int32_t>::IID = ABI::Windows::AI::MachineLearning::IID_ITensorInt32BitStatics;
-__declspec(selectany
-) const GUID TensorFactoryIID<uint64_t>::IID = ABI::Windows::AI::MachineLearning::IID_ITensorUInt64BitStatics;
-__declspec(selectany
-) const GUID TensorFactoryIID<int64_t>::IID = ABI::Windows::AI::MachineLearning::IID_ITensorInt64BitStatics;
-__declspec(selectany
-) const GUID TensorFactoryIID<bool>::IID = ABI::Windows::AI::MachineLearning::IID_ITensorBooleanStatics;
-__declspec(selectany
-) const GUID TensorFactoryIID<double>::IID = ABI::Windows::AI::MachineLearning::IID_ITensorDoubleStatics;
+__declspec(selectany) const GUID TensorFactoryIID<float>::IID =
+  ABI::Windows::AI::MachineLearning::IID_ITensorFloatStatics;
+__declspec(selectany) const GUID TensorFactoryIID<float16>::IID =
+  ABI::Windows::AI::MachineLearning::IID_ITensorFloat16BitStatics;
+__declspec(selectany) const GUID TensorFactoryIID<int8_t>::IID =
+  ABI::Windows::AI::MachineLearning::IID_ITensorInt8BitStatics;
+__declspec(selectany) const GUID TensorFactoryIID<uint8_t>::IID =
+  ABI::Windows::AI::MachineLearning::IID_ITensorUInt8BitStatics;
+__declspec(selectany) const GUID TensorFactoryIID<uint16_t>::IID =
+  ABI::Windows::AI::MachineLearning::IID_ITensorUInt16BitStatics;
+__declspec(selectany) const GUID TensorFactoryIID<int16_t>::IID =
+  ABI::Windows::AI::MachineLearning::IID_ITensorInt16BitStatics;
+__declspec(selectany) const GUID TensorFactoryIID<uint32_t>::IID =
+  ABI::Windows::AI::MachineLearning::IID_ITensorUInt32BitStatics;
+__declspec(selectany) const GUID TensorFactoryIID<int32_t>::IID =
+  ABI::Windows::AI::MachineLearning::IID_ITensorInt32BitStatics;
+__declspec(selectany) const GUID TensorFactoryIID<uint64_t>::IID =
+  ABI::Windows::AI::MachineLearning::IID_ITensorUInt64BitStatics;
+__declspec(selectany) const GUID TensorFactoryIID<int64_t>::IID =
+  ABI::Windows::AI::MachineLearning::IID_ITensorInt64BitStatics;
+__declspec(selectany) const GUID TensorFactoryIID<bool>::IID =
+  ABI::Windows::AI::MachineLearning::IID_ITensorBooleanStatics;
+__declspec(selectany) const GUID TensorFactoryIID<double>::IID =
+  ABI::Windows::AI::MachineLearning::IID_ITensorDoubleStatics;
 
 template <typename T>
 struct TensorFactory2IID {};
@@ -395,30 +395,30 @@ struct TensorFactory2IID<double> {
   static const GUID IID;
 };
 
-__declspec(selectany
-) const GUID TensorFactory2IID<float>::IID = ABI::Windows::AI::MachineLearning::IID_ITensorFloatStatics2;
-__declspec(selectany
-) const GUID TensorFactory2IID<float16>::IID = ABI::Windows::AI::MachineLearning::IID_ITensorFloat16BitStatics2;
-__declspec(selectany
-) const GUID TensorFactory2IID<int8_t>::IID = ABI::Windows::AI::MachineLearning::IID_ITensorInt8BitStatics2;
-__declspec(selectany
-) const GUID TensorFactory2IID<uint8_t>::IID = ABI::Windows::AI::MachineLearning::IID_ITensorUInt8BitStatics2;
-__declspec(selectany
-) const GUID TensorFactory2IID<uint16_t>::IID = ABI::Windows::AI::MachineLearning::IID_ITensorUInt16BitStatics2;
-__declspec(selectany
-) const GUID TensorFactory2IID<int16_t>::IID = ABI::Windows::AI::MachineLearning::IID_ITensorInt16BitStatics2;
-__declspec(selectany
-) const GUID TensorFactory2IID<uint32_t>::IID = ABI::Windows::AI::MachineLearning::IID_ITensorUInt32BitStatics2;
-__declspec(selectany
-) const GUID TensorFactory2IID<int32_t>::IID = ABI::Windows::AI::MachineLearning::IID_ITensorInt32BitStatics2;
-__declspec(selectany
-) const GUID TensorFactory2IID<uint64_t>::IID = ABI::Windows::AI::MachineLearning::IID_ITensorUInt64BitStatics2;
-__declspec(selectany
-) const GUID TensorFactory2IID<int64_t>::IID = ABI::Windows::AI::MachineLearning::IID_ITensorInt64BitStatics2;
-__declspec(selectany
-) const GUID TensorFactory2IID<bool>::IID = ABI::Windows::AI::MachineLearning::IID_ITensorBooleanStatics2;
-__declspec(selectany
-) const GUID TensorFactory2IID<double>::IID = ABI::Windows::AI::MachineLearning::IID_ITensorDoubleStatics2;
+__declspec(selectany) const GUID TensorFactory2IID<float>::IID =
+  ABI::Windows::AI::MachineLearning::IID_ITensorFloatStatics2;
+__declspec(selectany) const GUID TensorFactory2IID<float16>::IID =
+  ABI::Windows::AI::MachineLearning::IID_ITensorFloat16BitStatics2;
+__declspec(selectany) const GUID TensorFactory2IID<int8_t>::IID =
+  ABI::Windows::AI::MachineLearning::IID_ITensorInt8BitStatics2;
+__declspec(selectany) const GUID TensorFactory2IID<uint8_t>::IID =
+  ABI::Windows::AI::MachineLearning::IID_ITensorUInt8BitStatics2;
+__declspec(selectany) const GUID TensorFactory2IID<uint16_t>::IID =
+  ABI::Windows::AI::MachineLearning::IID_ITensorUInt16BitStatics2;
+__declspec(selectany) const GUID TensorFactory2IID<int16_t>::IID =
+  ABI::Windows::AI::MachineLearning::IID_ITensorInt16BitStatics2;
+__declspec(selectany) const GUID TensorFactory2IID<uint32_t>::IID =
+  ABI::Windows::AI::MachineLearning::IID_ITensorUInt32BitStatics2;
+__declspec(selectany) const GUID TensorFactory2IID<int32_t>::IID =
+  ABI::Windows::AI::MachineLearning::IID_ITensorInt32BitStatics2;
+__declspec(selectany) const GUID TensorFactory2IID<uint64_t>::IID =
+  ABI::Windows::AI::MachineLearning::IID_ITensorUInt64BitStatics2;
+__declspec(selectany) const GUID TensorFactory2IID<int64_t>::IID =
+  ABI::Windows::AI::MachineLearning::IID_ITensorInt64BitStatics2;
+__declspec(selectany) const GUID TensorFactory2IID<bool>::IID =
+  ABI::Windows::AI::MachineLearning::IID_ITensorBooleanStatics2;
+__declspec(selectany) const GUID TensorFactory2IID<double>::IID =
+  ABI::Windows::AI::MachineLearning::IID_ITensorDoubleStatics2;
 
 inline HRESULT GetActivationFactory(const wchar_t* p_class_id, const IID& iid, void** factory) noexcept {
   // Fallback to OS binary if the redistributable is not present!
diff --git a/winml/test/image/imagetests.cpp b/winml/test/image/imagetests.cpp
index b408c0315f94a..04717c75aa150 100644
--- a/winml/test/image/imagetests.cpp
+++ b/winml/test/image/imagetests.cpp
@@ -212,13 +212,16 @@ class ImageTests : public ::testing::Test {
     const std::wstring& model_file_name, const std::wstring& image_file_name, const InputImageSource input_image_source
   ) {
     // Case that the tensor's shape doesn't match model's shape should be skipped
-    if ((L"1080.jpg" == image_file_name || L"kitten_224.png" == image_file_name) && (InputImageSource::FromGPUResource == input_image_source || InputImageSource::FromCPUResource == input_image_source)) {
+    if ((L"1080.jpg" == image_file_name || L"kitten_224.png" == image_file_name) &&
+        (InputImageSource::FromGPUResource == input_image_source ||
+         InputImageSource::FromCPUResource == input_image_source)) {
       return true;
     }
 
     // Case that the images's shape doesn't match model's shape which expects free dimension should be skipped.
     // Because the fns-candy is not real model that can handle free dimensional input
-    if ((L"1080.jpg" == image_file_name || L"kitten_224.png" == image_file_name) && L"fns-candy_Bgr8_freeDimInput.onnx" == model_file_name) {
+    if ((L"1080.jpg" == image_file_name || L"kitten_224.png" == image_file_name) &&
+        L"fns-candy_Bgr8_freeDimInput.onnx" == model_file_name) {
       return true;
     }
 
@@ -385,7 +388,8 @@ TEST_P(ImageTest, ImageTest) {
     GTEST_SKIP() << "This test is disabled";
   }
 
-  if (LearningModelDeviceKind::Cpu != param.device_kind || InputImageSource::FromGPUResource == param.input_image_source) {
+  if (LearningModelDeviceKind::Cpu != param.device_kind ||
+      InputImageSource::FromGPUResource == param.input_image_source) {
     GPUTEST;
   }
 
@@ -482,13 +486,14 @@ TEST_P(BatchTest, BatchSupport) {
   if (param.use_session_options) {
     optimized_batch_size = param.use_session_options;
   }
-  if (VideoFrameSource::FromDirect3DSurface == param.video_frame_source && LearningModelDeviceKind::Cpu == param.device_kind) {
+  if (VideoFrameSource::FromDirect3DSurface == param.video_frame_source &&
+      LearningModelDeviceKind::Cpu == param.device_kind) {
     return;
   }
   if (LearningModelDeviceKind::Cpu != param.device_kind ||
-        VideoFrameSource::FromDirect3DSurface == param.video_frame_source ||
-        VideoFrameSource::FromDirect3DSurface == param.output_video_frame_source ||
-        VideoFrameSource::FromUnsupportedD3DSurface == param.output_video_frame_source) {
+      VideoFrameSource::FromDirect3DSurface == param.video_frame_source ||
+      VideoFrameSource::FromDirect3DSurface == param.output_video_frame_source ||
+      VideoFrameSource::FromUnsupportedD3DSurface == param.output_video_frame_source) {
     GPUTEST;
   }
 
@@ -556,7 +561,7 @@ TEST_P(BatchTest, BatchSupport) {
     for (int i = 0; i < param.batch_size; ++i) {
       std::wstring bm_image_path = FileHelpers::GetModulePath() + L"batchGroundTruth\\" + param.input_images[i];
       if (VideoFrameSource::FromSoftwareBitmap != param.output_video_frame_source &&
-                OutputBindingStrategy::Unbound != param.output_binding_strategy) {
+          OutputBindingStrategy::Unbound != param.output_binding_strategy) {
         VideoFrame D3D_video_frame = output_video_frames.GetAt(i);
         VideoFrame SB_video_frame(BitmapPixelFormat::Bgra8, 720, 720);
         D3D_video_frame.as<IVideoFrame>().CopyToAsync(SB_video_frame).get();
diff --git a/winml/test/model/compare_feature_value.cpp b/winml/test/model/compare_feature_value.cpp
index 30b16c4ad5f73..ac2553987f5ad 100644
--- a/winml/test/model/compare_feature_value.cpp
+++ b/winml/test/model/compare_feature_value.cpp
@@ -13,7 +13,8 @@ template <typename T>
 bool IsResultCloselyMatch(const T& outvalue, const T& expected_value, const double diff, const double tol) {
   if (diff > tol)
     return false;
-  if (std::isnan(diff) && !(std::isnan(outvalue) && std::isnan(expected_value)) && !(std::isinf(outvalue) && std::isinf(expected_value)))
+  if (std::isnan(diff) && !(std::isnan(outvalue) && std::isnan(expected_value)) &&
+      !(std::isinf(outvalue) && std::isinf(expected_value)))
     return false;
   return true;
 }
diff --git a/winml/test/model/model_tests.cpp b/winml/test/model/model_tests.cpp
index 859914014b8bb..4087bfd87caa7 100644
--- a/winml/test/model/model_tests.cpp
+++ b/winml/test/model/model_tests.cpp
@@ -150,7 +150,8 @@ std::string GetTestDataPath() {
   std::string testDataPath(MAX_PATH, '\0');
   auto environmentVariableFetchSuceeded =
     GetEnvironmentVariableA("WINML_TEST_DATA_PATH", testDataPath.data(), MAX_PATH);
-  if (environmentVariableFetchSuceeded == 0 && GetLastError() == ERROR_ENVVAR_NOT_FOUND || environmentVariableFetchSuceeded > MAX_PATH) {
+  if (environmentVariableFetchSuceeded == 0 && GetLastError() == ERROR_ENVVAR_NOT_FOUND ||
+      environmentVariableFetchSuceeded > MAX_PATH) {
     // if the WINML_TEST_DATA_PATH environment variable cannot be found, attempt to find the hardcoded models folder
     std::wstring modulePath = FileHelpers::GetModulePath();
     std::filesystem::path currPath = modulePath.substr(0, modulePath.find_last_of(L"\\"));
@@ -357,7 +358,8 @@ bool ModifyNameIfDisabledTest(/*inout*/ std::string& testName, winml::LearningMo
     if (SkipGpuTests()) {
       reason = "GPU tests are not enabled for this build.";
       shouldSkip = true;
-    } else if (disabledGpuAdapterTests.find(testName) != disabledGpuAdapterTests.end() && ShouldSkipTestOnGpuAdapter(testName)) {
+    } else if (disabledGpuAdapterTests.find(testName) != disabledGpuAdapterTests.end() &&
+               ShouldSkipTestOnGpuAdapter(testName)) {
       reason = disabledGpuAdapterTests[testName].second;
       shouldSkip = true;
     }
@@ -386,9 +388,7 @@ std::string GetFullNameOfTest(ITestCase* testCase, winml::LearningModelDeviceKin
   name += tokenizedModelPath[tokenizedModelPath.size() - 2] += "_";  // model name
   name += tokenizedModelPath[tokenizedModelPath.size() - 3];         // opset version
 
-  std::replace_if(
-    name.begin(), name.end(), [](char c) { return !absl::ascii_isalnum(c); }, '_'
-  );
+  std::replace_if(name.begin(), name.end(), [](char c) { return !absl::ascii_isalnum(c); }, '_');
 
   // Determine if test should be skipped, using the generic name (no CPU or GPU suffix yet).
   bool isDisabled = ModifyNameIfDisabledTest(/*inout*/ name, deviceKind);
diff --git a/winml/test/model/skip_model_tests.h b/winml/test/model/skip_model_tests.h
index cf55d8bcbae7e..349332c6ae0e3 100644
--- a/winml/test/model/skip_model_tests.h
+++ b/winml/test/model/skip_model_tests.h
@@ -114,14 +114,14 @@ std::unordered_map<std::string, std::string> disabledTests({
   {             "coreml_DecisionTreeClassifier_OpenML_1464_blood_transfusion_opset7",                                                                                           disabledTestDefaultReason},
   {                                                  "coreml_AgeNet_ImageNet_opset7",                                                                                           disabledTestDefaultReason},
 
- // GPU specific cases:
+  // GPU specific cases:
 
   // ONNX zoo models
   {                                                          "mask_rcnn_opset10_GPU",
    "Bug 31005388: mask_rcnn opset 10 onnx zoo model fails to evaluate on DirectML https://microsoft.visualstudio.com/OS/_workitems/edit/31005388"                                                        },
   {                                                        "faster_rcnn_opset10_GPU",
    "Bug 31005511: Failed to extract tensor data from evaluate result of faster_rcnn opset 10 model in DirectML https://microsoft.visualstudio.com/OS/_workitems/edit/31005511"                           },
- // ONNX model zoo's int8/qdq models generally do not work on CPUs that lack 8-bit instructions.
+  // ONNX model zoo's int8/qdq models generally do not work on CPUs that lack 8-bit instructions.
   {                                                         "YOLOv3_12_int8_opset12",                                                                                           disabledTestDefaultReason},
   {                                                            "VGG_16_int8_opset12",                                                                                           disabledTestDefaultReason},
   {                                                               "SSD_int8_opset12",                                                                                           disabledTestDefaultReason},
@@ -137,7 +137,7 @@ std::unordered_map<std::string, std::string> disabledTests({
   {                                                 "EfficientNet_Lite4_qdq_opset11",                                                                                           disabledTestDefaultReason},
   {                                                "EfficientNet_Lite4_int8_opset11",                                                                                           disabledTestDefaultReason},
 
- // Tier 2 models
+  // Tier 2 models
   {                                               "fp16_test_tiny_yolov2_opset7_GPU",
    "Bug 31005780: Result of fp16_test_tiny_yolov2_opset7 and fp16_coreml_FNS_Candy_opset7 models on DirectML aren't as accurate as on CPU https://microsoft.visualstudio.com/OS/_workitems/edit/31005780"},
   {                                                    "fp16_tiny_yolov2_opset8_GPU",
diff --git a/winml/test/scenario/cppwinrt/NoisyReluCpu.h b/winml/test/scenario/cppwinrt/NoisyReluCpu.h
index 5cccbae67407c..e419205fd52dc 100644
--- a/winml/test/scenario/cppwinrt/NoisyReluCpu.h
+++ b/winml/test/scenario/cppwinrt/NoisyReluCpu.h
@@ -65,12 +65,14 @@ struct NoisyReluOperator : winrt::implements<NoisyReluOperator, IMLOperatorKerne
       }
 
       // If the tensor types are both float type
-      if (outputTensor->GetTensorDataType() == MLOperatorTensorDataType::Float && inputTensor->GetTensorDataType() == MLOperatorTensorDataType::Float) {
+      if (outputTensor->GetTensorDataType() == MLOperatorTensorDataType::Float &&
+          inputTensor->GetTensorDataType() == MLOperatorTensorDataType::Float) {
         // For cpu data
         if (outputTensor->IsCpuData() && inputTensor->IsCpuData()) {
           ComputeInternal<float>(inputTensor.get(), outputTensor.get(), inputDataSize);
         }
-      } else if (outputTensor->GetTensorDataType() == MLOperatorTensorDataType::Double && inputTensor->GetTensorDataType() == MLOperatorTensorDataType::Double) {
+      } else if (outputTensor->GetTensorDataType() == MLOperatorTensorDataType::Double &&
+                 inputTensor->GetTensorDataType() == MLOperatorTensorDataType::Double) {
         // For cpu data
         if (outputTensor->IsCpuData() && inputTensor->IsCpuData()) {
           ComputeInternal<double>(inputTensor.get(), outputTensor.get(), inputDataSize);
diff --git a/winml/test/scenario/cppwinrt/ReluCpu.h b/winml/test/scenario/cppwinrt/ReluCpu.h
index 7bb275f7b399b..e8e91489fe872 100644
--- a/winml/test/scenario/cppwinrt/ReluCpu.h
+++ b/winml/test/scenario/cppwinrt/ReluCpu.h
@@ -60,12 +60,14 @@ struct ReluOperator : winrt::implements<ReluOperator, IMLOperatorKernel> {
     }
 
     // If the tensor types are both float type
-    if (outputTensor->GetTensorDataType() == MLOperatorTensorDataType::Float && inputTensor->GetTensorDataType() == MLOperatorTensorDataType::Float) {
+    if (outputTensor->GetTensorDataType() == MLOperatorTensorDataType::Float &&
+        inputTensor->GetTensorDataType() == MLOperatorTensorDataType::Float) {
       // For cpu data
       if (outputTensor->IsCpuData() && inputTensor->IsCpuData()) {
         ComputeInternal<float>(inputTensor.get(), outputTensor.get(), inputDataSize);
       }
-    } else if (outputTensor->GetTensorDataType() == MLOperatorTensorDataType::Double && inputTensor->GetTensorDataType() == MLOperatorTensorDataType::Double) {
+    } else if (outputTensor->GetTensorDataType() == MLOperatorTensorDataType::Double &&
+               inputTensor->GetTensorDataType() == MLOperatorTensorDataType::Double) {
       // For cpu data
       if (outputTensor->IsCpuData() && inputTensor->IsCpuData()) {
         ComputeInternal<double>(inputTensor.get(), outputTensor.get(), inputDataSize);