[ORT 1.18.1 Release] Cherry pick 1st round (#21105)

### Description
<!-- Describe your changes. -->



### Motivation and Context
<!-- - Why is this change required? What problem does it solve?
- If it fixes an open issue, please link to the issue here. -->

---------

Co-authored-by: Jian Chen <[email protected]>
Co-authored-by: Yi Zhang <[email protected]>
Co-authored-by: Changming Sun <[email protected]>
Co-authored-by: Ye Wang <[email protected]>
Co-authored-by: Your Name <[email protected]>

Author: 6 people

.pipelines/OneBranch.Nuget-WindowsAI-Pipeline.Official.yml

@@ -173,7 +173,7 @@ extends:
               $arm64_static_runtime_nupkg_unzipped_directory = [System.IO.Path]::Combine($arm64_static_runtime_nupkg_unzipped_directory_root, 'binaries', [System.IO.Path]::GetFileNameWithoutExtension($arm64_static_runtime_nuget_package))
               [System.IO.Compression.ZipFile]::ExtractToDirectory($arm64_static_runtime_nuget_package, $arm64_static_runtime_nupkg_unzipped_directory)
 
-             
+
 
               $x64_static_runtime_path_old = [System.IO.Path]::Combine($x64_static_runtime_nupkg_unzipped_directory, 'runtimes', 'win-x64', '_native')
               $x64_static_runtime_path_new = [System.IO.Path]::Combine($x64_nupkg_unzipped_directory, 'runtimes', 'win-x64', '_native', 'static')
@@ -185,7 +185,7 @@ extends:
               $arm64_runtime_path_new = [System.IO.Path]::Combine($x64_nupkg_unzipped_directory, 'runtimes', 'win-arm64', '_native')
               $arm64_static_runtime_path_old = [System.IO.Path]::Combine($arm64_static_runtime_nupkg_unzipped_directory, 'runtimes', 'win-arm64', '_native')
               $arm64_static_runtime_path_new = [System.IO.Path]::Combine($x64_nupkg_unzipped_directory, 'runtimes', 'win-arm64', '_native', 'static')
-              
+
               $uap_build_path_old = [System.IO.Path]::Combine($x64_static_runtime_nupkg_unzipped_directory, 'build', 'native')
               $uap_build_path_new = [System.IO.Path]::Combine($x64_nupkg_unzipped_directory, 'build', 'uap10.0')
 
@@ -262,7 +262,7 @@ extends:
               $x86_runtime_path_new = [System.IO.Path]::Combine($x64_nupkg_unzipped_directory, 'runtimes', 'win-x86', '_native')
               $arm64_runtime_path_old = [System.IO.Path]::Combine($arm64_nupkg_unzipped_directory, 'runtimes', 'win-arm64', '_native')
               $arm64_runtime_path_new = [System.IO.Path]::Combine($x64_nupkg_unzipped_directory, 'runtimes', 'win-arm64', '_native')
-              
+
               New-Item -Path $x86_runtime_path_new -ItemType Directory
               New-Item -Path $arm64_runtime_path_new -ItemType Directory
 
@@ -293,12 +293,21 @@ extends:
         - script: |
             dir $(Build.SourcesDirectory)\unzipped\runtimes\win-x64\_native
 
-        - task: EsrpCodeSigning@2
+        - task: SFP.build-tasks.custom-build-task-1.EsrpCodeSigning@5
           displayName: "Sign Nuget package"
           inputs:
-            ConnectedServiceName: 'OnnxRuntime CodeSign 20190817'
+            ConnectedServiceName: 'OnnxrunTimeCodeSign_20240611'
+            AppRegistrationClientId: '53d54d02-978d-4305-8572-583cf6711c4f'
+            AppRegistrationTenantId: '72f988bf-86f1-41af-91ab-2d7cd011db47'
+            AuthAKVName: 'buildkeyvault'
+            AuthCertName: '53d54d02-SSL-AutoRotate'
+            AuthSignCertName: '53d54d02-978d-4305-8572-583cf6711c4f'
+
             FolderPath: $(Build.ArtifactStagingDirectory)
             Pattern: '*.nupkg'
+            SessionTimeout: 90
+            ServiceEndpointUrl: 'https://api.esrp.microsoft.com/api/v2'
+            MaxConcurrency: 25
             signConfigType: inlineSignParams
             inlineOperation: |
                [
@@ -307,14 +316,14 @@ extends:
                        "operationSetCode": "NuGetSign",
                        "parameters": [ ],
                        "toolName": "sign",
-                       "toolVersion": "1.0"
+                       "toolVersion": "6.2.9304.0"
                    },
                    {
                        "keyCode": "CP-401405",
                        "operationSetCode": "NuGetVerify",
                        "parameters": [ ],
                        "toolName": "sign",
-                       "toolVersion": "1.0"
+                       "toolVersion": "6.2.9304.0"
                    }
                ]
 

cmake/CMakeLists.txt

@@ -59,8 +59,8 @@ if (NOT CMAKE_BUILD_TYPE)
   set(CMAKE_BUILD_TYPE "RelWithDebInfo" CACHE STRING "Choose build type: Debug Release RelWithDebInfo MinSizeRel." FORCE)
 endif()
 
-if("${CMAKE_C_COMPILER_ID}" STREQUAL "GNU" AND CMAKE_C_COMPILER_VERSION VERSION_LESS 8)
-  message(FATAL_ERROR  "GCC version must be greater than or equal to 8")
+if("${CMAKE_C_COMPILER_ID}" STREQUAL "GNU" AND CMAKE_C_COMPILER_VERSION VERSION_LESS 9)
+  message(FATAL_ERROR  "GCC version must be greater than or equal to 9")
 endif()
 
 # Options
@@ -1294,12 +1294,6 @@ if (onnxruntime_USE_TVM)
   list(APPEND onnxruntime_EXTERNAL_DEPENDENCIES tvm)
 endif()
 
-# needs to link with stdc++fs in Linux
-if (UNIX AND "${CMAKE_C_COMPILER_ID}" STREQUAL "GNU" AND CMAKE_C_COMPILER_VERSION VERSION_LESS 9)
-  set(FS_STDLIB stdc++fs)
-endif()
-list(APPEND onnxruntime_EXTERNAL_LIBRARIES ${FS_STDLIB})
-
 # onnxruntime-extensions
 if (onnxruntime_USE_EXTENSIONS)
   include(extensions)
@@ -1468,16 +1462,6 @@ if (onnxruntime_USE_CUDA)
   endif()
 endif()
 
-if (onnxruntime_USE_TENSORRT)
-  # needs to link with stdc++fs in Linux
-  if (UNIX)
-    if (NOT APPLE)
-      set(FS_STDLIB stdc++fs)
-    endif()
-  endif()
-  list(APPEND onnxruntime_EXTERNAL_LIBRARIES ${FS_STDLIB})
-endif()
-
 if (onnxruntime_USE_MIGRAPHX)
   if (WIN32)
     message(FATAL_ERROR "MIGraphX does not support build in Windows!")

cmake/onnxruntime_providers_migraphx.cmake

@@ -49,7 +49,7 @@
   target_compile_options(onnxruntime_providers_migraphx PRIVATE -Wno-error=sign-compare)
   set_property(TARGET onnxruntime_providers_migraphx APPEND_STRING PROPERTY COMPILE_FLAGS "-Wno-deprecated-declarations")
   set_property(TARGET onnxruntime_providers_migraphx APPEND_STRING PROPERTY LINK_FLAGS "-Xlinker --version-script=${ONNXRUNTIME_ROOT}/core/providers/migraphx/version_script.lds -Xlinker --gc-sections")
-  target_link_libraries(onnxruntime_providers_migraphx PRIVATE nsync::nsync_cpp stdc++fs)
+  target_link_libraries(onnxruntime_providers_migraphx PRIVATE nsync::nsync_cpp)
 
   include(CheckLibraryExists)
   check_library_exists(migraphx::c "migraphx_program_run_async" "/opt/rocm/migraphx/lib" HAS_STREAM_SYNC)

cmake/onnxruntime_providers_tensorrt.cmake

@@ -206,7 +206,7 @@
   elseif(UNIX)
     set_property(TARGET onnxruntime_providers_tensorrt APPEND_STRING PROPERTY COMPILE_FLAGS "-Wno-deprecated-declarations")
     set_property(TARGET onnxruntime_providers_tensorrt APPEND_STRING PROPERTY LINK_FLAGS "-Xlinker --version-script=${ONNXRUNTIME_ROOT}/core/providers/tensorrt/version_script.lds -Xlinker --gc-sections")
-    target_link_libraries(onnxruntime_providers_tensorrt PRIVATE nsync::nsync_cpp stdc++fs)
+    target_link_libraries(onnxruntime_providers_tensorrt PRIVATE nsync::nsync_cpp)
   elseif(WIN32)
     set_property(TARGET onnxruntime_providers_tensorrt APPEND_STRING PROPERTY LINK_FLAGS "-DEF:${ONNXRUNTIME_ROOT}/core/providers/tensorrt/symbols.def")
   else()

java/build.gradle

@@ -4,6 +4,7 @@ plugins {
 	id 'signing'
 	id 'jacoco'
 	id "com.diffplug.spotless" version "6.25.0"
+	id "net.linguica.maven-settings" version "0.5"
 }
 
 allprojects {
@@ -19,6 +20,12 @@ version = rootProject.file('../VERSION_NUMBER').text.trim()
 def cmakeBuildDir = System.properties['cmakeBuildDir']
 def useCUDA = System.properties['USE_CUDA']
 def useROCM = System.properties['USE_ROCM']
+
+def adoArtifact = project.findProperty('adoArtifact')
+def adoAccessToken = project.findProperty('adoAccessToken')
+// Only publish to ADO feed if all two properties are set
+def publishToAdo = adoArtifact != null && adoAccessToken != null
+
 boolean enableTrainingApis = (System.properties['ENABLE_TRAINING_APIS'] ?: "0") == "1"
 def cmakeJavaDir = "${cmakeBuildDir}/java"
 def cmakeNativeLibDir = "${cmakeJavaDir}/native-lib"
@@ -37,6 +44,11 @@ def trainingDescription = 'ONNX Runtime Training is a training and inference pac
 	'(Open Neural Network Exchange) models. This package is targeted for Learning on The Edge aka On-Device Training ' +
 	'See https://github.com/microsoft/onnxruntime-training-examples/tree/master/on_device_training for more details.'
 
+// We need to have a custom settings.xml so codeql can bypass the need for settings.security.xml
+mavenSettings {
+	userSettingsFileName = "${projectDir}/settings.xml"
+}
+
 java {
 	sourceCompatibility = JavaVersion.VERSION_1_8
 	targetCompatibility = JavaVersion.VERSION_1_8
@@ -48,7 +60,8 @@ jar {
 }
 
 // Add explicit sources jar with pom file.
-task sourcesJar(type: Jar, dependsOn: classes) {
+tasks.register('sourcesJar', Jar) {
+	dependsOn classes
 	archiveClassifier = "sources"
 	from sourceSets.main.allSource
 	into("META-INF/maven/$project.group/$mavenArtifactId") {
@@ -58,7 +71,8 @@ task sourcesJar(type: Jar, dependsOn: classes) {
 }
 
 // Add explicit javadoc jar with pom file
-task javadocJar(type: Jar, dependsOn: javadoc) {
+tasks.register('javadocJar', Jar) {
+	dependsOn javadoc
 	archiveClassifier = "javadoc"
 	from javadoc.destinationDir
 	into("META-INF/maven/$project.group/$mavenArtifactId") {
@@ -82,7 +96,7 @@ spotless {
 
 compileJava {
 	dependsOn spotlessJava
-	options.compilerArgs += ["-h", "${project.buildDir}/headers/"]
+	options.compilerArgs += ["-h", "${layout.buildDirectory.get().toString()}/headers/"]
 	if (!JavaVersion.current().isJava8()) {
 		// Ensures only methods present in Java 8 are used
 		options.compilerArgs.addAll(['--release', '8'])
@@ -128,7 +142,7 @@ if (cmakeBuildDir != null) {
 	// generate tasks to be called from cmake
 
 	// Overwrite jar location
-	task allJar(type: Jar) {
+	tasks.register('allJar', Jar) {
 		manifest {
 			attributes('Automatic-Module-Name': project.group,
 					'Implementation-Title': 'onnxruntime',
@@ -143,23 +157,20 @@ if (cmakeBuildDir != null) {
 		from cmakeNativeLibDir
 	}
 
-	task cmakeBuild(type: Copy) {
-		from project.buildDir
+	tasks.register('cmakeBuild', Copy) {
+		from layout.buildDirectory.get()
 		include 'libs/**'
 		include 'docs/**'
 		into cmakeBuildOutputDir
+		dependsOn(allJar, sourcesJar, javadocJar, javadoc)
 	}
-	cmakeBuild.dependsOn allJar
-	cmakeBuild.dependsOn sourcesJar
-	cmakeBuild.dependsOn javadocJar
-	cmakeBuild.dependsOn javadoc
 
-	task cmakeCheck(type: Copy) {
-		from project.buildDir
+	tasks.register('cmakeCheck', Copy) {
+		from layout.buildDirectory.get()
 		include 'reports/**'
 		into cmakeBuildOutputDir
+		dependsOn(check)
 	}
-	cmakeCheck.dependsOn check
 }
 
 dependencies {
@@ -198,17 +209,22 @@ jacocoTestReport {
 	reports {
 		xml.required = true
 		csv.required = true
-		html.destination file("${buildDir}/jacocoHtml")
+		html.outputLocation = layout.buildDirectory.dir("jacocoHtml")
 	}
 }
 
 publishing {
 	publications {
 		maven(MavenPublication) {
 			groupId = project.group
-			artifactId = mavenArtifactId
-
-			from components.java
+			if(publishToAdo) {
+				artifactId = 'onnxruntime_gpu'
+				artifact (adoArtifact)
+			} else {
+				artifactId = mavenArtifactId
+				from components.java
+			}
+			version = project.version
 			pom {
 				name = enableTrainingApis ? 'onnxruntime-training' : 'onnx-runtime'
 				description = enableTrainingApis ? trainingDescription : defaultDescription
@@ -239,24 +255,41 @@ publishing {
 		}
 	}
 	repositories {
-		maven {
-			url 'https://oss.sonatype.org/service/local/staging/deploy/maven2/'
-			credentials {
-				username mavenUser
-				password mavenPwd
+		if (publishToAdo) {
+			maven {
+				url "https://aiinfra.pkgs.visualstudio.com/PublicPackages/_packaging/${System.getenv('ADOFeedName')}/maven/v1"
+				name System.getenv('ADOFeedName')
+				authentication {
+					basic(BasicAuthentication)
+				}
+				credentials {
+					username  'aiinfra'
+					password "${project.findProperty('adoAccessToken')}"
+				}
+			}
+		} else {
+			maven {
+				url 'https://oss.sonatype.org/service/local/staging/deploy/maven2/'
+				credentials {
+					username mavenUser
+					password mavenPwd
+				}
 			}
 		}
 	}
 }
-
 // Generates a task signMavenPublication that will
 // build all artifacts.
 signing {
 	// Queries env vars:
 	// ORG_GRADLE_PROJECT_signingKey
 	// ORG_GRADLE_PROJECT_signingPassword but can be changed to properties
-	def signingKey = findProperty("signingKey")
-	def signingPassword = findProperty("signingPassword")
-	useInMemoryPgpKeys(signingKey, signingPassword)
-	sign publishing.publications.maven
+		def signingKey = findProperty("signingKey")
+		def signingPassword = findProperty("signingPassword")
+		// Skip signing if no key is provided
+		if (signingKey != null && signingPassword != null) {
+			useInMemoryPgpKeys(signingKey, signingPassword)
+			sign publishing.publications.maven
+			sign publishing.publications.mavenAdo
+		}
 }

java/settings.xml

@@ -0,0 +1,9 @@
+<!--The only purpose of this file is to let codeql bypass the need for settings.security.xml-->
+<settings xmlns="http://maven.apache.org/SETTINGS/1.0.0"
+          xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
+          xsi:schemaLocation="http://maven.apache.org/SETTINGS/1.0.0
+                              https://maven.apache.org/xsd/settings-1.0.0.xsd">
+  <servers>
+    <!-- password is coming from cli -PadoAccessToken -->
+  </servers>
+</settings>

onnxruntime/contrib_ops/cuda/quantization/moe_quantization.cc

@@ -54,6 +54,11 @@ Status QMoE::ComputeInternal(OpKernelContext* context) const {
   const Tensor* fc3_scales_optional = context->Input<Tensor>(9);
   const Tensor* fc3_experts_bias_optional = context->Input<Tensor>(10);
 
+#if defined(__GNUC__)
+#pragma GCC diagnostic push
+#pragma GCC diagnostic ignored "-Wmaybe-uninitialized"  // Mute "maybe used uninitialized" warning for MoEParameters.
+#endif
+
   MoEParameters moe_params;
   MoEQuantType quant_type = MoEQuantType::UINT4;
   ORT_RETURN_IF_ERROR(CheckInputs(moe_params, quant_type, input, router_probs, fc1_experts_weights,
@@ -135,6 +140,10 @@ Status QMoE::ComputeInternal(OpKernelContext* context) const {
       reinterpret_cast<int*>(expert_for_source_row.get()), static_cast<int>(moe_params.num_rows),
       static_cast<int>(moe_params.hidden_size), static_cast<int>(k_), Stream(context));
 
+#if defined(__GNUC__)
+#pragma GCC diagnostic pop
+#endif
+
   return Status::OK();
 }
 

onnxruntime/core/providers/tensorrt/tensorrt_execution_provider_utils.h

@@ -8,15 +8,14 @@
 #include <sstream>
 #include <iostream>
 #include <filesystem>
-#include <experimental/filesystem>
 #include "flatbuffers/idl.h"
 #include "ort_trt_int8_cal_table.fbs.h"
 #include <NvInferVersion.h>
 #include "core/providers/cuda/cuda_pch.h"
 #include "core/common/path_string.h"
 #include "core/framework/murmurhash3.h"
 
-namespace fs = std::experimental::filesystem;
+namespace fs = std::filesystem;
 
 namespace onnxruntime {
 

onnxruntime/python/tools/transformers/models/llama/requirements.txt

@@ -1,7 +1,7 @@
 optimum>=1.14.1
 transformers>=4.33.2,<= 4.37.2
 torch>=2.2.0
-onnx>=1.14.0
+onnx==1.16.0
 datasets>=2.8.0
 protobuf==3.20.2
 psutil

onnxruntime/python/tools/transformers/models/phi2/requirements.txt

@@ -1,3 +1,3 @@
-onnx>=1.15.0
+onnx==1.16.0
 transformers>=4.36.2
 onnxscript>=0.1.0.dev20240126