ONNX 1.15 integration (#17125)

### Description this is for ORT 1.17.0 - make ORT to use ONNX release 1.15.0 branch. Eventually will update to the release tag once ONNX 1.15.0 is released ### Motivation and Context Prepare for ORT 1.17.0 release. People can start work on new and updated ONNX ops in ORT. --------- Signed-off-by: Liqun Fu <liqfu@microsoft.com>
2026-05-14 20:48:00 +00:00 · 2023-09-26 14:44:48 -07:00 · 2023-09-26 14:44:48 -07:00 · 2be4dc6d04
commit 2be4dc6d04
parent 37dcefb5b7
29 changed files with 278 additions and 245 deletions
--- a/cgmanifests/generate_cgmanifest.py
+++ b/cgmanifests/generate_cgmanifest.py
@ -90,55 +90,6 @@ def add_github_dep(name, parsed_url):
            git_deps[dep] = name


-with open(
-    os.path.join(REPO_DIR, "tools", "ci_build", "github", "linux", "docker", "Dockerfile.manylinux2_28_cuda11"),
-) as f:
-    for line in f:
-        if not line.strip():
-            package_name = None
-            package_filename = None
-            package_url = None
-        if package_filename is None:
-            m = re.match(r"RUN\s+export\s+(.+?)_ROOT=(\S+).*", line)
-            if m is not None:
-                package_name = m.group(1)
-                package_filename = m.group(2)
-            else:
-                m = re.match(r"RUN\s+export\s+(.+?)_VERSION=(\S+).*", line)
-                if m is not None:
-                    package_name = m.group(1)
-                    package_filename = m.group(2)
-        elif package_url is None:
-            m = re.match(r"(.+?)_DOWNLOAD_URL=(\S+)", line)
-            if m is not None:
-                package_url = m.group(2)
-                if package_name == "LIBXCRYPT":
-                    package_url = m.group(2) + "/v" + package_filename + ".tar.gz"
-                elif package_name == "CMAKE":
-                    package_url = m.group(2) + "/v" + package_filename + "/cmake-" + package_filename + ".tar.gz"
-                else:
-                    package_url = m.group(2) + "/" + package_filename + ".tar.gz"
-                parsed_url = urlparse(package_url)
-                if parsed_url.hostname == "github.com":
-                    add_github_dep("manylinux dependency " + package_name, parsed_url)
-                else:
-                    registration = {
-                        "Component": {
-                            "Type": "other",
-                            "other": {
-                                "Name": package_name.lower(),
-                                "Version": package_filename.split("-")[-1],
-                                "DownloadUrl": package_url,
-                            },
-                            "comments": "manylinux dependency",
-                        }
-                    }
-                    registrations.append(registration)
-                package_name = None
-                package_filename = None
-                package_url = None
-
-
 def normalize_path_separators(path):
    return path.replace(os.path.sep, "/")

--- a/cgmanifests/generated/cgmanifest.json
+++ b/cgmanifests/generated/cgmanifest.json
@ -2,112 +2,6 @@
  "$schema": "https://json.schemastore.org/component-detection-manifest.json",
  "Version": 1,
  "Registrations": [
-    {
-      "Component": {
-        "Type": "other",
-        "other": {
-          "Name": "autoconf",
-          "Version": "2.71",
-          "DownloadUrl": "http://ftp.gnu.org/gnu/autoconf/autoconf-2.71.tar.gz"
-        },
-        "comments": "manylinux dependency"
-      }
-    },
-    {
-      "Component": {
-        "Type": "other",
-        "other": {
-          "Name": "automake",
-          "Version": "1.16.5",
-          "DownloadUrl": "http://ftp.gnu.org/gnu/automake/automake-1.16.5.tar.gz"
-        },
-        "comments": "manylinux dependency"
-      }
-    },
-    {
-      "Component": {
-        "Type": "other",
-        "other": {
-          "Name": "libtool",
-          "Version": "2.4.7",
-          "DownloadUrl": "http://ftp.gnu.org/gnu/libtool/libtool-2.4.7.tar.gz"
-        },
-        "comments": "manylinux dependency"
-      }
-    },
-    {
-      "Component": {
-        "Type": "other",
-        "other": {
-          "Name": "git",
-          "Version": "2.36.2",
-          "DownloadUrl": "https://www.kernel.org/pub/software/scm/git/git-2.36.2.tar.gz"
-        },
-        "comments": "manylinux dependency"
-      }
-    },
-    {
-      "Component": {
-        "Type": "other",
-        "other": {
-          "Name": "sqlite_autoconf",
-          "Version": "3390200",
-          "DownloadUrl": "https://www.sqlite.org/2022/sqlite-autoconf-3390200.tar.gz"
-        },
-        "comments": "manylinux dependency"
-      }
-    },
-    {
-      "Component": {
-        "Type": "other",
-        "other": {
-          "Name": "openssl",
-          "Version": "1.1.1q",
-          "DownloadUrl": "https://www.openssl.org/source/openssl-1.1.1q.tar.gz"
-        },
-        "comments": "manylinux dependency"
-      }
-    },
-    {
-      "component": {
-        "type": "git",
-        "git": {
-          "commitHash": "50cf2b6dd4fdf04309445f2eec8de7051d953abf",
-          "repositoryUrl": "https://github.com/besser82/libxcrypt.git"
-        },
-        "comments": "manylinux dependency LIBXCRYPT"
-      }
-    },
-    {
-      "component": {
-        "type": "git",
-        "git": {
-          "commitHash": "a896e3d066448b3530dbcaa48869fafefd738f57",
-          "repositoryUrl": "https://github.com/emscripten-core/emsdk.git"
-        },
-        "comments": "git submodule at cmake/external/emsdk"
-      }
-    },
-    {
-      "component": {
-        "type": "git",
-        "git": {
-          "commitHash": "7a2ed51a6b682a83e345ff49fc4cfd7ca47550db",
-          "repositoryUrl": "https://github.com/google/libprotobuf-mutator.git"
-        },
-        "comments": "git submodule at cmake/external/libprotobuf-mutator"
-      }
-    },
-    {
-      "component": {
-        "type": "git",
-        "git": {
-          "commitHash": "e2525550194ce3d8a2c4a3af451c9d9b3ae6650e",
-          "repositoryUrl": "https://github.com/onnx/onnx.git"
-        },
-        "comments": "git submodule at cmake/external/onnx"
-      }
-    },
    {
      "component": {
        "type": "git",
@ -268,6 +162,16 @@
        "comments": "mp11"
      }
    },
+    {
+      "component": {
+        "type": "git",
+        "git": {
+          "commitHash": "fdefbe85ed9c362b95b9b401cd19db068a76141f",
+          "repositoryUrl": "https://github.com/onnx/onnx.git"
+        },
+        "comments": "onnx"
+      }
+    },
    {
      "component": {
        "type": "git",
--- a/cmake/deps.txt
+++ b/cmake/deps.txt
@ -24,7 +24,7 @@ microsoft_gsl;https://github.com/microsoft/GSL/archive/refs/tags/v4.0.0.zip;cf36
 microsoft_wil;https://github.com/microsoft/wil/archive/refs/tags/v1.0.230629.1.zip;e4a542a323c070376f7c2d1973d0f7ddbc1d2fa5
 mimalloc;https://github.com/microsoft/mimalloc/archive/refs/tags/v2.1.1.zip;d5ee7d34223d0567892db5179849939c8769dc41
 mp11;https://github.com/boostorg/mp11/archive/refs/tags/boost-1.82.0.zip;9bc9e01dffb64d9e0773b2e44d2f22c51aace063
-onnx;https://github.com/onnx/onnx/archive/e2525550194ce3d8a2c4a3af451c9d9b3ae6650e.zip;782f23d788185887f520a90535513e244218e928
+onnx;https://github.com/onnx/onnx/archive/14303de049144035dfd94ace5f7a3b44773b1aad.zip;250eab9690392b248d75b56e605fb49eca373442
 #use the commit of supporting all the plugins and TRT 8.6-GA (https://github.com/onnx/onnx-tensorrt/commit/0462dc31ae78f48744b6141ae376df1f96d3f459)
 onnx_tensorrt;https://github.com/onnx/onnx-tensorrt/archive/0462dc31ae78f48744b6141ae376df1f96d3f459.zip;5ff086361956cceb81ed17453a1fd8db2aa4328d
 protobuf;https://github.com/protocolbuffers/protobuf/archive/refs/tags/v21.12.zip;7cf2733949036c7d52fda017badcab093fe73bfa
@ -44,4 +44,4 @@ tensorboard;https://github.com/tensorflow/tensorboard/archive/373eb09e4c5d2b3cc2
 cutlass;https://github.com/NVIDIA/cutlass/archive/refs/tags/v3.0.0.zip;0f95b3c1fc1bd1175c4a90b2c9e39074d1bccefd
 utf8_range;https://github.com/protocolbuffers/utf8_range/archive/72c943dea2b9240cd09efde15191e144bc7c7d38.zip;9925739c9debc0efa2adcb194d371a35b6a03156
 extensions;https://github.com/microsoft/onnxruntime-extensions/archive/94142d8391c9791ec71c38336436319a2d4ac7a0.zip;4365ac5140338b4cb75a39944a4be276e3829b3c
-composable_kernel;https://github.com/ROCmSoftwarePlatform/composable_kernel/archive/d52ec01652b7d620386251db92455968d8d90bdc.zip;6b5ce8edf3625f8817086c194fbf94b664e1b0e0
+composable_kernel;https://github.com/ROCmSoftwarePlatform/composable_kernel/archive/d52ec01652b7d620386251db92455968d8d90bdc.zip;6b5ce8edf3625f8817086c194fbf94b664e1b0e0
--- a/cmake/deps_update_and_upload.py
+++ b/cmake/deps_update_and_upload.py
@ -0,0 +1,56 @@
+# in case deps.txt is updated, run this file to update and upload the dependencies so that CI can use them.
+# Before running the script, increase the version number found at:
+# https://aiinfra.visualstudio.com/Lotus/_artifacts/feed/Lotus/UPack/onnxruntime_build_dependencies/versions
+# Run without --do-upload once to verify downloading. Use --do-upload when you are ready to publish.
+# python cmake/deps_update_and_upload.py --root-path C:/temp/onnxruntime_deps --version 1.0.82 --do-upload
+# update version number in tools\ci_build\github\azure-pipelines\templates\download-deps.yml
+import re
+import subprocess
+import os
+import argparse
+import tempfile
+
+parser = argparse.ArgumentParser(description="Update dependencies and publish to Azure Artifacts")
+parser.add_argument(
+    "--root-path", type=str, default=tempfile.gettempdir(), help="Target root path for downloaded files"
+)
+parser.add_argument("--version", type=str, default="1.0.82", help="Package version to publish")
+parser.add_argument("--do-upload", action="store_true", help="Upload the package to Azure Artifacts")
+args = parser.parse_args()
+
+with open("cmake/deps.txt") as file:
+    text = file.read()
+
+lines = [line for line in text.split("\n") if not line.startswith("#") and ";" in line]
+
+root_path = args.root_path
+
+for line in lines:
+    url = re.sub("^[^;]+?;https://([^;]+?);.*", r"https://\1", line)
+    filename = re.sub("^[^;]+?;https://([^;]+?);.*", r"\1", line)
+    full_path = os.path.join(root_path, filename)
+    subprocess.run(["curl", "-sSL", "--create-dirs", "-o", full_path, url])
+
+package_name = "onnxruntime_build_dependencies"
+version = args.version
+
+# Check if the user is logged in to Azure
+result = subprocess.run("az account show", shell=True, capture_output=True, text=True)
+if "No subscriptions found" in result.stderr:
+    # Prompt the user to log in to Azure
+    print("You are not logged in to Azure. Please log in to continue.")
+    subprocess.run("az login", shell=True)
+
+# Publish the package to Azure Artifacts if --no-upload is not specified
+
+cmd = f'az artifacts universal publish --organization https://dev.azure.com/onnxruntime --feed onnxruntime --name {package_name} --version {version} --description "onnxruntime build time dependencies" --path {root_path}'
+if args.do_upload:
+    subprocess.run(cmd, shell=True)
+else:
+    print("would have run: " + cmd)
+
+cmd = f'az artifacts universal publish --organization https://dev.azure.com/aiinfra --feed Lotus --name {package_name} --version {version} --description "onnxruntime build time dependencies" --path {root_path}'
+if args.do_upload:
+    subprocess.run(cmd, shell=True)
+else:
+    print("would have run: " + cmd)
--- a/cmake/patches/onnx/onnx.patch
+++ b/cmake/patches/onnx/onnx.patch
@ -64,16 +64,3 @@ index 0aab3e26..0f859267 100644
 +#endif
 +
 #endif // ! ONNX_ONNX_PB_H
-diff --git a/onnx/checker.cc b/onnx/checker.cc
-index 8fdaf037..1beb1b88 100644
--- a/onnx/checker.cc
-+++ b/onnx/checker.cc
-@@ -190,7 +190,7 @@ void check_tensor(const TensorProto& tensor, const CheckerContext& ctx) {
-         }
-         std::string data_path = path_join(ctx.get_model_dir(), relative_path);
-         // use stat64 to check whether the file exists
-#ifdef __APPLE__
-+#if defined(__APPLE__) || defined(__wasm__)
-         struct stat buffer; // APPLE does not have stat64
-         if (stat((data_path).c_str(), &buffer) != 0) {
- #else
--- a/docs/OperatorKernels.md
+++ b/docs/OperatorKernels.md
@ -67,7 +67,8 @@ Do not modify directly.*
 |||[11, 12]|**T** = tensor(bfloat16), tensor(bool), tensor(double), tensor(float), tensor(float16), tensor(int16), tensor(int32), tensor(int64), tensor(int8), tensor(string), tensor(uint16), tensor(uint32), tensor(uint64), tensor(uint8)|
 |||[4, 10]|**T** = tensor(bfloat16), tensor(bool), tensor(double), tensor(float), tensor(float16), tensor(int16), tensor(int32), tensor(int64), tensor(int8), tensor(string), tensor(uint16), tensor(uint32), tensor(uint64), tensor(uint8)|
 |ConcatFromSequence|*in* input_sequence:**S**<br> *out* concat_result:**T**|11+|**S** = seq(tensor(bfloat16)), seq(tensor(bool)), seq(tensor(double)), seq(tensor(float)), seq(tensor(float16)), seq(tensor(int16)), seq(tensor(int32)), seq(tensor(int64)), seq(tensor(int8)), seq(tensor(string)), seq(tensor(uint16)), seq(tensor(uint32)), seq(tensor(uint64)), seq(tensor(uint8))|
-|ConstantOfShape|*in* input:**T1**<br> *out* output:**T2**|9+|**T1** = tensor(int64)<br/> **T2** = tensor(bool), tensor(double), tensor(float), tensor(float16), tensor(int16), tensor(int32), tensor(int64), tensor(int8), tensor(uint16), tensor(uint32), tensor(uint64), tensor(uint8)|
+|ConstantOfShape|*in* input:**T1**<br> *out* output:**T2**|20+|**T1** = tensor(int64)<br/> **T2** = tensor(bfloat16), tensor(bool), tensor(double), tensor(float), tensor(float16), tensor(float8e4m3fn), tensor(float8e4m3fnuz), tensor(float8e5m2), tensor(float8e5m2fnuz), tensor(int16), tensor(int32), tensor(int64), tensor(int8), tensor(uint16), tensor(uint32), tensor(uint64), tensor(uint8)|
+|||[9, 19]|**T1** = tensor(int64)<br/> **T2** = tensor(bool), tensor(double), tensor(float), tensor(float16), tensor(int16), tensor(int32), tensor(int64), tensor(int8), tensor(uint16), tensor(uint32), tensor(uint64), tensor(uint8)|
 |Conv|*in* X:**T**<br> *in* W:**T**<br> *in* B:**T**<br> *out* Y:**T**|11+|**T** = tensor(float)|
 |||[1, 10]|**T** = tensor(float)|
 |ConvInteger|*in* x:**T1**<br> *in* w:**T2**<br> *in* x_zero_point:**T1**<br> *in* w_zero_point:**T2**<br> *out* y:**T3**|10+|**T1** = tensor(uint8)<br/> **T2** = tensor(uint8)<br/> **T3** = tensor(int32)|
@ -78,7 +79,7 @@ Do not modify directly.*
 |Crop|*in* input:**T**<br> *out* output:**T**|1+|**T** = tensor(float)|
 |CumSum|*in* x:**T**<br> *in* axis:**T2**<br> *out* y:**T**|14+|**T** = tensor(double), tensor(float), tensor(int32), tensor(int64)<br/> **T2** = tensor(int32), tensor(int64)|
 |||[11, 13]|**T** = tensor(double), tensor(float), tensor(int32), tensor(int64)<br/> **T2** = tensor(int32), tensor(int64)|
-|DFT|*in* input:**T1**<br> *in* dft_length:**T2**<br> *out* output:**T1**|17+|**T1** = tensor(double), tensor(float)<br/> **T2** = tensor(int32), tensor(int64)|
+|DFT|*in* input:**T1**<br> *in* dft_length:**T2**<br> *in* axis:**tensor(int64)**<br> *out* output:**T1**<br><br>or<br><br>*in* input:**T1**<br> *in* dft_length:**T2**<br> *out* output:**T1**|17+|**T1** = tensor(double), tensor(float)<br/> **T2** = tensor(int32), tensor(int64)|
 |DepthToSpace|*in* input:**T**<br> *out* output:**T**|13+|**T** = tensor(double), tensor(float)|
 |||[11, 12]|**T** = tensor(double), tensor(float)|
 |||[1, 10]|**T** = tensor(double), tensor(float)|
@ -935,7 +936,7 @@ Do not modify directly.*
 |Crop|*in* input:**T**<br> *out* output:**T**|1+|**T** = tensor(float), tensor(float16)|
 |CumSum|*in* x:**T**<br> *in* axis:**T2**<br> *out* y:**T**|14+|**T** = tensor(float), tensor(float16), tensor(int32), tensor(int64), tensor(uint32), tensor(uint64)|
 |||11+|**T** = tensor(float), tensor(float16), tensor(int32), tensor(int64), tensor(uint32), tensor(uint64)|
-|DFT|*in* input:**T1**<br> *in* dft_length:**T2**<br> *out* output:**T1**|17+|**T1** = tensor(float), tensor(float16)<br/> **T2** = tensor(int64)|
+|DFT|*in* input:**T1**<br> *in* dft_length:**T2**<br> *in* axis:**tensor(int64)**<br> *out* output:**T1**<br><br>or<br><br>*in* input:**T1**<br> *in* dft_length:**T2**<br> *out* output:**T1**|17+|**T1** = tensor(float), tensor(float16)<br/> **T2** = tensor(int64)|
 |DepthToSpace|*in* input:**T**<br> *out* output:**T**|13+|**T** = tensor(bool), tensor(double), tensor(float), tensor(float16), tensor(int16), tensor(int32), tensor(int64), tensor(int8), tensor(uint16), tensor(uint32), tensor(uint64), tensor(uint8)|
 |||11+|**T** = tensor(bool), tensor(double), tensor(float), tensor(float16), tensor(int16), tensor(int32), tensor(int64), tensor(int8), tensor(uint16), tensor(uint32), tensor(uint64), tensor(uint8)|
 |||1+|**T** = tensor(bool), tensor(double), tensor(float), tensor(float16), tensor(int16), tensor(int32), tensor(int64), tensor(int8), tensor(uint16), tensor(uint32), tensor(uint64), tensor(uint8)|
--- a/onnxruntime/core/optimizer/transpose_optimization/optimizer_api.h
+++ b/onnxruntime/core/optimizer/transpose_optimization/optimizer_api.h
@ -442,7 +442,7 @@ class GraphRef {
 }  // namespace api

 constexpr int64_t kMinSupportedOpset = 7;
-constexpr int64_t kMaxSupportedOpset = 19;
+constexpr int64_t kMaxSupportedOpset = 20;

 // enum of results that a CostCheckFn can return.
 enum class CostCheckResult {
--- a/onnxruntime/core/providers/coreml/builders/impl/base_op_builder.h
+++ b/onnxruntime/core/providers/coreml/builders/impl/base_op_builder.h
@ -51,7 +51,7 @@ class BaseOpBuilder : public IOpBuilder {
  virtual bool HasSupportedInputsImpl(const Node& node, const logging::Logger& logger) const;

  virtual int GetMinSupportedOpSet(const Node& /* node */) const { return 1; }
-  virtual int GetMaxSupportedOpSet(const Node& /* node */) const { return 19; }
+  virtual int GetMaxSupportedOpSet(const Node& /* node */) const { return 20; }

 private:
  bool HasSupportedOpSet(const Node& node, const logging::Logger& logger) const;
--- a/onnxruntime/core/providers/cpu/cpu_execution_provider.cc
+++ b/onnxruntime/core/providers/cpu/cpu_execution_provider.cc
@ -273,7 +273,7 @@ class ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDoma

 // Opset 9
 class ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 10, Compress);
-class ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, ConstantOfShape);
+class ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 19, ConstantOfShape);
 class ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 12, MeanVarianceNormalization);
 class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 12, float, Greater);
 class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 12, double, Greater);
@ -958,6 +958,9 @@ class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain,
 class ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, Scan);
 class ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, Shape);

+// Opset 20
+class ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, ConstantOfShape);
+
 // !!PLEASE READ BELOW!! Following that, add new entries above this comment

 /*  *** IMPORTANT! ***
@ -1332,7 +1335,7 @@ Status RegisterOnnxOperatorKernels(KernelRegistry& kernel_registry) {
    // Opset 9
    BuildKernelCreateInfo<ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 10,
                                                                    Compress)>,
-    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, ConstantOfShape)>,
+    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,
@ -2383,6 +2386,9 @@ Status RegisterOnnxOperatorKernels(KernelRegistry& kernel_registry) {
    BuildKernelCreateInfo<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, uint8_t, Resize)>,
    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, Scan)>,
    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 19, Shape)>,
+
+    // Opset 20
+    BuildKernelCreateInfo<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 20, ConstantOfShape)>,
  };

  for (auto& function_table_entry : function_table) {
--- a/onnxruntime/core/providers/cpu/generator/constant_of_shape.cc
+++ b/onnxruntime/core/providers/cpu/generator/constant_of_shape.cc
@ -11,11 +11,16 @@ ORT_SPECIFY_OP_KERNEL_ARG_DEFAULT_TYPE_LIST_ALL_OPSETS(
    kCpuExecutionProvider, kOnnxDomain, ConstantOfShape, Output, 0,
    ConstantOfShapeDefaultOutputTypes);

+ORT_SPECIFY_OP_KERNEL_ARG_DEFAULT_TYPE_LIST(
+    kCpuExecutionProvider, kOnnxDomain, ConstantOfShape, 20, Output, 0,
+    ConstantOfShapeDefaultOutputTypesOpset20);
+
 // pytorch converter uses ConstantOfShape with int64 to create Pad input
 // https://github.com/pytorch/pytorch/blob/044b519a80459f6787f6723c1c091a18b153d184/torch/onnx/symbolic_opset11.py#L449
 ORT_SPECIFY_OP_KERNEL_ARG_REQUIRED_TYPES_ALL_OPSETS(
    kCpuExecutionProvider, kOnnxDomain, ConstantOfShape, Output, 0,
    int64_t);
+
 }  // namespace op_kernel_type_control

 namespace {
@ -24,6 +29,10 @@ using EnabledOutputTypes =
    ORT_OP_KERNEL_ARG_ENABLED_TYPE_LIST_ALL_OPSETS(
        kCpuExecutionProvider, kOnnxDomain, ConstantOfShape, Output, 0);

+using EnabledOutputTypesOpset20 =
+    ORT_OP_KERNEL_ARG_ENABLED_TYPE_LIST(
+        kCpuExecutionProvider, kOnnxDomain, ConstantOfShape, 20, Output, 0);
+
 class ConstantOfShape final : public ConstantOfShapeBase<EnabledOutputTypes>, public OpKernel {
 public:
  explicit ConstantOfShape(const OpKernelInfo& info) : ConstantOfShapeBase(info), OpKernel(info) {}
@ -66,13 +75,22 @@ Status ConstantOfShape::Compute(OpKernelContext* ctx) const {

 }  // namespace

-ONNX_CPU_OPERATOR_KERNEL(
+ONNX_CPU_OPERATOR_VERSIONED_KERNEL(
    ConstantOfShape,
    9,
+    19,
    KernelDefBuilder()
        .TypeConstraint("T1", DataTypeImpl::GetTensorType<int64_t>())
        .TypeConstraint("T2",
                        BuildKernelDefConstraintsFromTypeList<EnabledOutputTypes>()),
    ConstantOfShape);

+ONNX_CPU_OPERATOR_KERNEL(
+    ConstantOfShape,
+    20,
+    KernelDefBuilder()
+        .TypeConstraint("T1", DataTypeImpl::GetTensorType<int64_t>())
+        .TypeConstraint("T2",
+                        BuildKernelDefConstraintsFromTypeList<EnabledOutputTypesOpset20>()),
+    ConstantOfShape);
 }  // namespace onnxruntime
--- a/onnxruntime/core/providers/cpu/generator/constant_of_shape_base.h
+++ b/onnxruntime/core/providers/cpu/generator/constant_of_shape_base.h
@ -23,6 +23,18 @@ using ConstantOfShapeDefaultOutputTypes =
        uint8_t, uint16_t, uint32_t, uint64_t,
        bool>;

+using ConstantOfShapeDefaultOutputTypesOpset20 =
+    TypeList<
+        BFloat16,
+        MLFloat16,
+        float, double,
+#if !defined(DISABLE_FLOAT8_TYPES)
+        Float8E4M3FN, Float8E4M3FNUZ, Float8E5M2, Float8E5M2FNUZ,
+#endif
+        int8_t, int16_t, int32_t, int64_t,
+        uint8_t, uint16_t, uint32_t, uint64_t,
+        bool>;
+
 template <typename EnabledOutputTypeList = ConstantOfShapeDefaultOutputTypes>
 class ConstantOfShapeBase {
 protected:
--- a/onnxruntime/core/providers/webnn/builders/impl/base_op_builder.h
+++ b/onnxruntime/core/providers/webnn/builders/impl/base_op_builder.h
@ -46,7 +46,7 @@ class BaseOpBuilder : public IOpBuilder {
  // We still set the mininal supported opset to 1 as we couldn't
  // get the model opset version at this stage.
  virtual int GetMinSupportedOpSet(const Node& /* node */) const { return 1; }
-  virtual int GetMaxSupportedOpSet(const Node& /* node */) const { return 19; }
+  virtual int GetMaxSupportedOpSet(const Node& /* node */) const { return 20; }

 private:
  bool HasSupportedOpSet(const Node& node, const logging::Logger& logger) const;
--- a/onnxruntime/test/framework/function_test.cc
+++ b/onnxruntime/test/framework/function_test.cc
@ -6,6 +6,7 @@
 #include "onnx/defs/parser.h"

 #include "core/common/span_utils.h"
+#include "core/framework/float8.h"
 #include "core/graph/model.h"
 #include "core/providers/cpu/cpu_execution_provider.h"
 #include "core/session/inference_session.h"
@ -69,7 +70,9 @@ static void Check(const char* source,
  float threshold = 0.001f;

  for (size_t i = 0; i < size; ++i) {
-    ASSERT_NEAR(data[i], output_values[i], threshold) << "at position i:" << i;
+    if (!std::isnan(data[i]) && !std::isnan(output_values[i])) {
+      ASSERT_NEAR(data[i], output_values[i], threshold) << "at position i:" << i;
+    }
  }
 }

@ -389,25 +392,13 @@ TEST(FunctionTest, AttrSaturateNan) {
        >
        agraph (float[N] x) => (float[N] y)
        {
-            y0 = local.myfun <a = 1e6> (x)
-            y1 = local.myfun (x)
-            y = Add (y0, y1)
-        }
-
-        <
-        opset_import: [ "" : 19 ],
-        domain: "local"
-        >
-        myfun <a: float=1.0> (x) => (y) {
-            x2 = Constant <value_float: float=@a>()
-            x2_ = Cast<to=18>(x2)
-            x3 = CastLike<saturate=0>(x2, x2_)
-            x3_ = Cast<to=1>(x3)
-            y = Add (x, x3_)
+            x_E4M3FNUZ = Cast<to=18>(x)
+            x_E4M3FNUZ_2 = CastLike<saturate=0>(x, x_E4M3FNUZ)  # NaN when OOR
+            y = Cast<to=1>(x_E4M3FNUZ_2)
        }
        )";

-  Check(code, "x", {1.0, 2.0, 1e6}, "y", {243.0, 245.0, 2000241});  // std::numeric_limits<float>::quiet_NaN()});
+  Check(code, "x", {1.0, 2.0, 1e6}, "y", {1.0, 2.0, std::numeric_limits<float>::quiet_NaN()});
 }

 #endif
--- a/onnxruntime/test/framework/inference_session_test.cc
+++ b/onnxruntime/test/framework/inference_session_test.cc
@ -2056,7 +2056,7 @@ TEST(InferenceSessionTests, TestStrictShapeInference) {

  ASSERT_STATUS_OK(session_options.config_options.AddConfigEntry(kOrtSessionOptionsConfigStrictShapeTypeInference, "1"));
  tester.Run(session_options, OpTester::ExpectResult::kExpectFailure,
-             "Mismatch between number of source and target dimensions. Source=1 Target=2",
+             "Mismatch between number of inferred and declared dimensions. inferred=1 declared=2",
             excluded_provider_types);
 }

--- a/onnxruntime/test/optimizer/graph_transform_test.cc
+++ b/onnxruntime/test/optimizer/graph_transform_test.cc
@ -6280,7 +6280,7 @@ TEST_F(GraphTransformationTests, ConstantSharing_ShouldNotShareForGraphOutput) {
 TEST_F(GraphTransformationTests, GatherToSplitFusion) {
  auto build_test_case = [&](ModelTestBuilder& builder) {
    auto* data_arg = builder.MakeInput<float>({{54}});
-    auto* shape_arg = builder.MakeInput<int64_t>({{1}});
+    auto* shape_arg = builder.MakeInput<int64_t>({{4}});
    auto* reshape_out = builder.MakeIntermediate<float>({{2, 3, 3, 3}});
    auto* gather_index_1 = builder.MakeInitializer<int64_t>({}, {static_cast<int64_t>(0)});
    auto* gather_index_2 = builder.MakeInitializer<int64_t>({}, {static_cast<int64_t>(1)});
@ -6393,7 +6393,7 @@ TEST_F(GraphTransformationTests, GatherToSplitFusion) {
 TEST_F(GraphTransformationTests, GatherToSplitFusion_NoSqueeze) {
  auto build_test_case = [&](ModelTestBuilder& builder) {
    auto* data_arg = builder.MakeInput<float>({{54}});
-    auto* shape_arg = builder.MakeInput<int64_t>({{1}});
+    auto* shape_arg = builder.MakeInput<int64_t>({{4}});
    auto* reshape_out = builder.MakeIntermediate<float>({{2, 3, 3, 3}});
    auto* gather_index_1 = builder.MakeInitializer<int64_t>({1}, {static_cast<int64_t>(0)});
    auto* gather_index_2 = builder.MakeInitializer<int64_t>({1}, {static_cast<int64_t>(1)});
--- a/onnxruntime/test/providers/cpu/controlflow/scan_test.cc
+++ b/onnxruntime/test/providers/cpu/controlflow/scan_test.cc
@ -578,7 +578,7 @@ TEST(Scan9, DISABLED_BadShape) {
  ShortSequenceOneInBatchOneLoopStateVar(
      options,
      "Node:concat Output:concat_out_1 [ShapeInferenceError] Mismatch between number of source and target dimensions. "
-      "Source=2 Target=1");
+      "inferred=2 declared=1");
 }

 TEST(Scan8, ShortSequenceTwoInBatchOneLoopStateVar) {
--- a/onnxruntime/test/providers/cpu/nn/conv_fp16_test.cc
+++ b/onnxruntime/test/providers/cpu/nn/conv_fp16_test.cc
@ -109,7 +109,7 @@ TEST(ConvFp16Test, Conv1D_Invalid_Input_Shape) {
  TestConvFp16Op(attrs, {X, dummy_vals}, {X_shape, dummy_shape}, dummy_vals, dummy_shape, false,
                 OpTester::ExpectResult::kExpectFailure,
                 "Node:node1 Output:Y [ShapeInferenceError] Can't merge shape info. "
-                 "Both source and target dimension have values but they differ. Source=0 Target=2 Dimension=2",
+                 "Both inferred and declared dimension have values but they differ. Inferred=0 Declared=2 Dimension=2",
                 -1);  // use latest opset for shape inferencing errors
 }

@ -132,7 +132,7 @@ TEST(ConvFp16Test, Conv2D_Invalid_Input_Shape) {
  TestConvFp16Op(attrs, {X, dummy_vals}, {X_shape, dummy_shape}, dummy_vals, dummy_shape, false,
                 OpTester::ExpectResult::kExpectFailure,
                 "Node:node1 Output:Y [ShapeInferenceError] Can't merge shape info. "
-                 "Both source and target dimension have values but they differ. Source=1 Target=2 Dimension=0",
+                 "Both inferred and declared dimension have values but they differ. Inferred=1 Declared=2 Dimension=0",
                 -1);  // use latest opset for shape inferencing errors
 }

--- a/onnxruntime/test/providers/cpu/nn/conv_op_test.cc
+++ b/onnxruntime/test/providers/cpu/nn/conv_op_test.cc
@ -249,7 +249,7 @@ TEST(ConvTest, Conv1D_Invalid_Input_Shape) {
  TestConvOp(attrs, {X, dummy_vals}, {X_shape, dummy_shape}, dummy_vals, dummy_shape, false,
             OpTester::ExpectResult::kExpectFailure,
             "Node:node1 Output:Y [ShapeInferenceError] Can't merge shape info. "
-             "Both source and target dimension have values but they differ. Source=0 Target=2 Dimension=2",
+             "Both inferred and declared dimension have values but they differ. Inferred=0 Declared=2 Dimension=2",
             -1);  // use latest opset for shape inferencing errors
 }

@ -272,7 +272,7 @@ TEST(ConvTest, Conv2D_Invalid_Input_Shape) {
  TestConvOp(attrs, {X, dummy_vals}, {X_shape, dummy_shape}, dummy_vals, dummy_shape, false,
             OpTester::ExpectResult::kExpectFailure,
             "Node:node1 Output:Y [ShapeInferenceError] Can't merge shape info. "
-             "Both source and target dimension have values but they differ. Source=1 Target=2 Dimension=0",
+             "Both inferred and declared dimension have values but they differ. Inferred=1 Declared=2 Dimension=0",
             -1);  // use latest opset for shape inferencing errors
 }

--- a/onnxruntime/test/providers/cpu/nn/tfidfvectorizer_test.cc
+++ b/onnxruntime/test/providers/cpu/nn/tfidfvectorizer_test.cc
@ -91,7 +91,7 @@ TEST(TfIdfVectorizerTest, Int32_TF_onlyBigrams_Skip0_Empty_Dim1Fail) {

  test.Run(OpTester::ExpectResult::kExpectFailure,
           "Can't merge shape info. "
-           "Both source and target dimension have values but they differ. Source=7 Target=0 Dimension=0");
+           "Both inferred and declared dimension have values but they differ. Inferred=7 Declared=0 Dimension=0");
 }

 TEST(TfIdfVectorizerTest, Int32_TF_onlyBigrams_Skip0_Empty_Dim1Success) {
@ -136,7 +136,7 @@ TEST(TfIdfVectorizerTest, Int32_TF_onlyBigrams_Skip0_Empty_Dim2) {
  test.AddOutput<float>("Y", out_dims, output);

  test.Run(OpTester::ExpectResult::kExpectFailure,
-           "Mismatch between number of source and target dimensions. Source=2 Target=1");
+           "Mismatch between number of inferred and declared dimensions. inferred=2 declared=1");
 }

 TEST(TfIdfVectorizerTest, Int32_TF_onlyBigrams_Skip01_Empty_Dim2) {
@ -159,7 +159,7 @@ TEST(TfIdfVectorizerTest, Int32_TF_onlyBigrams_Skip01_Empty_Dim2) {
  test.AddOutput<float>("Y", out_dims, output);

  test.Run(OpTester::ExpectResult::kExpectFailure,
-           "Mismatch between number of source and target dimensions. Source=2 Target=1");
+           "Mismatch between number of inferred and declared dimensions. inferred=2 declared=1");
 }

 TEST(TfIdfVectorizerTest, Int32_TF_onlyBigrams_Skip0_Empty_Dim2N) {
--- a/onnxruntime/test/providers/cpu/tensor/transpose_test.cc
+++ b/onnxruntime/test/providers/cpu/tensor/transpose_test.cc
@ -37,7 +37,7 @@ TEST(TransposeOpTest, PermRankDoesNotMatchTensorRank) {
  // This failure comes from shape inference, because in this case it knows the input dims.
  // But in the real world, the model can supply different input dims at runtime.
  test.Run(OpTester::ExpectResult::kExpectFailure,
-           "Node:node1 Output:Y [ShapeInferenceError] Mismatch between number of source and target dimensions. Source=3 Target=4");
+           "Node:node1 Output:Y [ShapeInferenceError] Mismatch between number of inferred and declared dimensions. inferred=3 declared=4");
 }

 // Some of the tests can't run on TensorrtExecutionProvider because of errors.
--- a/onnxruntime/test/python/quantization/op_test_utils.py
+++ b/onnxruntime/test/python/quantization/op_test_utils.py
@ -279,6 +279,9 @@ def check_model_correctness(
    ops_set = set(node.op_type for node in model_onnx.graph.node)
    check_reference_evaluator = not (ops_set & {"EmbedLayerNormalization", "Conv", "Attention", "Transpose"})

+    with open(model_path_to_check, "rb") as f:
+        model_check = onnx.load(f)
+
    if check_reference_evaluator and onnx_recent_enough:
        ref = ReferenceEvaluator(model_path_origin)
        ref_origin_results = ref.run(None, inputs)
@ -289,7 +292,7 @@ def check_model_correctness(
                output,
                rtol=rtol,
                atol=atol,
-                err_msg=f"Model {model_path_to_check!r} failed for providers={providers!r}.",
+                err_msg=f"Model {model_path_origin!r} failed for providers={providers!r}.",
            )

    # Verifies the shapes in the quantized model.
@ -301,40 +304,52 @@ def check_model_correctness(
                expected_shapes[init.name] = tuple(init.dims)
        checked = 0
        f8_quantization = False
-        with open(model_path_to_check, "rb") as f:
-            model_check = onnx.load(f)
-            for init in model_check.graph.initializer:
-                if init.name.endswith("_quantized"):
-                    name = init.name.replace("_quantized", "")
-                    expected = expected_shapes[name]
-                    shape = tuple(init.dims)
-                    if not dynamic and expected != shape:
-                        raise AssertionError(
-                            f"Shape mismatch for initializer {init.name!r} from {init.name!r}, "
-                            f"shape={shape} != {expected} (expected)."
-                        )
-                    else:
-                        checked += 1
-                if "zero_point" in init.name:
-                    dt = init.data_type
-                    f8_quantization = f8_quantization or dt in (
-                        TensorProto.FLOAT8E4M3FN,
-                        TensorProto.FLOAT8E4M3FNUZ,
-                        TensorProto.FLOAT8E5M2,
-                        TensorProto.FLOAT8E5M2FNUZ,
+        for init in model_check.graph.initializer:
+            if init.name.endswith("_quantized"):
+                name = init.name.replace("_quantized", "")
+                expected = expected_shapes[name]
+                shape = tuple(init.dims)
+                if not dynamic and expected != shape:
+                    raise AssertionError(
+                        f"Shape mismatch for initializer {init.name!r} from {init.name!r}, "
+                        f"shape={shape} != {expected} (expected)."
                    )
-            if checked == 0:
-                raise AssertionError(
-                    f"Unable to check expected shape, expected_shapes={expected_shapes}, "
-                    f"names={[init.name for init in model_check.graph.initializer]}."
+                else:
+                    checked += 1
+            if "zero_point" in init.name:
+                dt = init.data_type
+                f8_quantization = f8_quantization or dt in (
+                    TensorProto.FLOAT8E4M3FN,
+                    TensorProto.FLOAT8E4M3FNUZ,
+                    TensorProto.FLOAT8E5M2,
+                    TensorProto.FLOAT8E5M2FNUZ,
                )
+        if checked == 0:
+            raise AssertionError(
+                f"Unable to check expected shape, expected_shapes={expected_shapes}, "
+                f"names={[init.name for init in model_check.graph.initializer]}."
+            )
        if f8_quantization:
            check_sign_f8_quantization(model_path_origin, model_path_to_check)

    # Verifies the expected outputs.
    if check_reference_evaluator and onnx_recent_enough:
+        reference_new_ops = [QGemm]
+        has_missing_reference_ops = any(
+            node.domain not in ["", "ai.onnx"]
+            and not any(
+                node.domain == new_node.op_domain and node.op_type == new_node.__name__
+                for new_node in reference_new_ops
+            )
+            for node in model_check.graph.node
+        )
+        if has_missing_reference_ops:
+            # We need to skip the test if the model contains ops that are not supported.
+            testcase.skipTest(
+                f"Model {model_path_to_check!r} contains ops that are not supported by the reference evaluator."
+            )
        # Needs pv.Version(onnx.__version__) >= pv.Version("1.16.0")
-        ref = ReferenceEvaluator(model_path_to_check, new_ops=[QGemm])
+        ref = ReferenceEvaluator(model_check, new_ops=reference_new_ops)
        target_results = ref.run(None, inputs)
        testcase.assertEqual(len(origin_results), len(target_results), "result count are different")
        for idx, ref_output in enumerate(origin_results):
--- a/onnxruntime/test/testdata/onnx_backend_test_series_filters.jsonc
+++ b/onnxruntime/test/testdata/onnx_backend_test_series_filters.jsonc
@ -233,7 +233,64 @@
        "^test_resize_upsample_sizes_nearest_cuda",
        "^test_resize_upsample_sizes_nearest_floor_align_corners_cuda",
        "^test_resize_upsample_sizes_nearest_not_larger_cuda",
-        "^test_resize_upsample_sizes_nearest_round_prefer_ceil_asymmetric_cuda"
+        "^test_resize_upsample_sizes_nearest_round_prefer_ceil_asymmetric_cuda",
+        // onnx 1.15 (opset 20) new and updated op tests
+        "^test_ai_onnx_ml_label_encoder_string_int",
+        "^test_ai_onnx_ml_label_encoder_string_int_no_default",
+        "^test_ai_onnx_ml_label_encoder_tensor_mapping",
+        "^test_ai_onnx_ml_label_encoder_tensor_value_only_mapping",
+        "^test_gridsample_aligncorners_true",
+        "^test_gridsample_bicubic_align_corners_0_additional_1",
+        "^test_gridsample_bicubic_align_corners_1_additional_1",
+        "^test_gridsample_bicubic",
+        "^test_gridsample_bilinear_align_corners_0_additional_1",
+        "^test_gridsample_bilinear_align_corners_1_additional_1",
+        "^test_gridsample_bilinear",
+        "^test_gridsample_border_padding",
+        "^test_gridsample",
+        "^test_gridsample_nearest_align_corners_0_additional_1",
+        "^test_gridsample_nearest_align_corners_1_additional_1",
+        "^test_gridsample_nearest",
+        "^test_gridsample_reflection_padding",
+        "^test_gridsample_volumetric_bilinear_align_corners_0",
+        "^test_gridsample_volumetric_bilinear_align_corners_1",
+        "^test_gridsample_volumetric_nearest_align_corners_0",
+        "^test_gridsample_volumetric_nearest_align_corners_1",
+        "^test_gridsample_zeros_padding",
+        "^test_image_decoder_decode_bmp_rgb",
+        "^test_image_decoder_decode_jpeg2k_rgb",
+        "^test_image_decoder_decode_jpeg_bgr",
+        "^test_image_decoder_decode_jpeg_grayscale",
+        "^test_image_decoder_decode_jpeg_rgb",
+        "^test_image_decoder_decode_png_rgb",
+        "^test_image_decoder_decode_pnm_rgb",
+        "^test_image_decoder_decode_tiff_rgb",
+        "^test_image_decoder_decode_webp_rgb",
+        "^test_regex_full_match_basic",
+        "^test_regex_full_match_email_domain",
+        "^test_regex_full_match_empty",
+        "^test_string_concat_broadcasting",
+        "^test_string_concat",
+        "^test_string_concat_empty_string",
+        "^test_string_concat_utf8",
+        "^test_string_concat_zero_dimensional",
+        "^test_string_split_basic",
+        "^test_string_split_consecutive_delimiters",
+        "^test_string_split_empty_string_delimiter",
+        "^test_string_split_empty_tensor",
+        "^test_string_split_maxsplit",
+        "^test_string_split_no_delimiter",
+        "^test_dft_axis",
+        "^test_dft",
+        "^test_dft_inverse",
+        "^test_isinf",
+        "^test_isinf_float16",
+        "^test_isinf_negative",
+        "^test_isinf_positive",
+        "^test_isnan",
+        "^test_isnan_float16",
+        "^test_reduce_max_bool_inputs",
+        "^test_reduce_min_bool_inputs"
    ],
    "current_failing_tests_x86": [
        "^test_vgg19",
@ -316,7 +373,24 @@
        "^test_layer_normalization_4d_axis_negative_1_expanded_ver18_cpu",
        "^test_layer_normalization_4d_axis_negative_2_expanded_ver18_cpu",
        "^test_layer_normalization_4d_axis_negative_3_expanded_ver18_cpu",
-        "^test_layer_normalization_default_axis_expanded_ver18_cpu"
+        "^test_layer_normalization_default_axis_expanded_ver18_cpu",
+        // onnx 1.15 (opset 20) new and updated op tests (test_affine_grid_???_expanded utilizes ConstantOfShape so it needs to be skipped as well)
+        // https://dev.azure.com/onnxruntime/onnxruntime/_build/results?buildId=1139541&view=logs&j=249e9d58-0012-5814-27cf-6a201adbd9cf&t=bb33e81f-0527-50e0-0fd2-e94f509f0a82
+        // only supported with cpu provider
+        "^test_affine_grid_2d",
+        "^test_affine_grid_2d_align_corners",
+        "^test_affine_grid_2d_align_corners_expanded",
+        "^test_affine_grid_2d_expanded",
+        "^test_affine_grid_3d",
+        "^test_affine_grid_3d_align_corners",
+        "^test_affine_grid_3d_align_corners_expanded",
+        "^test_affine_grid_3d_expanded",
+        "^test_constantofshape_float_ones",
+        "^test_constantofshape_int_shape_zero",
+        "^test_constantofshape_int_zeros",
+        // https://dev.azure.com/onnxruntime/onnxruntime/_build/results?buildId=1141563&view=logs&j=a018b46d-e41a-509d-6581-c95fdaa42fcd&t=d61c1d37-f101-5d28-982f-e5931b720302
+        "^test_gelu_tanh_2_cpu",
+        "^test_gelu_tanh_2_expanded_cpu"
    ],
    "current_failing_tests_NNAPI": [
        "^test_maxpool_2d_uint8",
@ -569,7 +643,22 @@
        "^test_sequence_map_identity_1_sequence_cpu",
        "^test_sequence_map_identity_1_sequence_expanded_cpu",
        "^test_sequence_map_identity_2_sequences_cpu",
-        "^test_sequence_map_identity_2_sequences_expanded_cpu"
+        "^test_sequence_map_identity_2_sequences_expanded_cpu",
+        // onnx 1.15 (opset 20) new and updated op tests (test_affine_grid_???_expanded utilizes ConstantOfShape so it needs to be skipped as well)
+        // https://dev.azure.com/onnxruntime/onnxruntime/_build/results?buildId=1139542&view=logs&j=3032dfba-5baf-5872-0871-2e69cb7f4b6a&t=f0d05deb-fc26-5aaf-e43e-7db2764c07da
+        // only supported with cpu provider
+        "^test_affine_grid_2d",
+        "^test_affine_grid_2d_align_corners",
+        "^test_affine_grid_2d_align_corners_expanded",
+        "^test_affine_grid_2d_expanded",
+        "^test_affine_grid_3d",
+        "^test_affine_grid_3d_align_corners",
+        "^test_affine_grid_3d_align_corners_expanded",
+        "^test_affine_grid_3d_expanded",
+        "^test_constantofshape_float_ones",
+        "^test_constantofshape_int_shape_zero",
+        "^test_constantofshape_int_zeros"
+
    ],
    // ORT first supported opset 7, so models with nodes that require versions prior to opset 7 are not supported
    "tests_with_pre_opset7_dependencies": [
--- a/onnxruntime/test/testdata/onnx_backend_test_series_overrides.jsonc
+++ b/onnxruntime/test/testdata/onnx_backend_test_series_overrides.jsonc
@ -7,6 +7,9 @@
        "test_dft": 1e-3,
        "test_dft_axis": 1e-3,
        "test_dft_inverse": 1e-3,
+        "test_dft_opset19": 1e-3,
+        "test_dft_axis_opset19": 1e-3,
+        "test_dft_inverse_opset19": 1e-3,
        "test_stft": 1e-4,
        "test_stft_with_window": 1e-4
    },
--- a/orttraining/orttraining/core/graph/training_op_defs.cc
+++ b/orttraining/orttraining/core/graph/training_op_defs.cc
@ -4171,7 +4171,7 @@ Return true if all elements are true and false otherwise.
              "T", OpSchema::Variadic,
              /*is_homogeneous*/ false,
              /*min_arity*/ 1)
-      .TypeConstraint("T", OpSchema::all_tensor_types_with_bfloat(),
+      .TypeConstraint("T", OpSchema::all_tensor_types_ir4(),
                      "Allow inputs and outputs to be any kind of tensor.");
 #endif  // ENABLE_TRITON

--- a/requirements-dev.txt
+++ b/requirements-dev.txt
@ -17,4 +17,4 @@ scikit-learn
 scipy
 sympy
 wheel
-setuptools>=41.4.0
+setuptools>=61.0.0
--- a/requirements-training.txt
+++ b/requirements-training.txt
@ -6,4 +6,4 @@ onnx
 packaging
 protobuf
 sympy
-setuptools>=41.4.0
+setuptools>=61.0.0
--- a/tools/ci_build/github/azure-pipelines/templates/download-deps.yml
+++ b/tools/ci_build/github/azure-pipelines/templates/download-deps.yml
@ -11,7 +11,7 @@ steps:
      packageType: upack
      feed: '/7424c8e4-5c62-490e-95c4-79446f31017c'
      definition: '517c4f6f-5437-4392-a70d-4f15ec5be2f0'
-      version: 1.0.81
+      version: 1.0.90
      downloadPath: $(Build.BinariesDirectory)/deps

 # The private ADO project
@ -22,7 +22,7 @@ steps:
      packageType: upack
      feed: '/4c7631f5-24c0-4307-8822-1aa8f180c325'
      definition: 'fd9dd5ad-b73e-4678-890e-edcf680dbc1a'
-      version: 1.0.81
+      version: 1.0.90
      downloadPath: $(Build.BinariesDirectory)/deps

 # You can add more ADO accounts at here.
--- a/tools/ci_build/github/azure-pipelines/templates/jobs/win-ci-prebuild-steps.yml
+++ b/tools/ci_build/github/azure-pipelines/templates/jobs/win-ci-prebuild-steps.yml
@ -31,7 +31,7 @@ steps:
    architecture: ${{parameters.BuildArch}}

 - script: |
-    python -m pip install -q setuptools wheel numpy flatbuffers
+    python -m pip install --upgrade "setuptools>=61.0.0" wheel numpy flatbuffers
  workingDirectory: '$(Build.BinariesDirectory)'
  displayName: 'Install python modules'

--- a/tools/ci_build/github/linux/docker/scripts/manylinux/requirements.txt
+++ b/tools/ci_build/github/linux/docker/scripts/manylinux/requirements.txt
@ -4,7 +4,7 @@ mypy
 pytest
 setuptools>=41.4.0
 wheel
-git+http://github.com/onnx/onnx.git@e2525550194ce3d8a2c4a3af451c9d9b3ae6650e#egg=onnx
+git+http://github.com/onnx/onnx.git@ac3e58759463ff3a3089e3cd64fddbfad0f6724d#egg=onnx
 protobuf==3.20.2
 sympy==1.12
 flatbuffers