Cherry-picks for ORT 1.23.0

onnxruntime/contrib_ops/webgpu/bert/group_query_attention.cc

@@ -198,9 +198,12 @@ Status GroupQueryAttention::ComputeInternal(onnxruntime::webgpu::ComputeContext&
   Tensor* present_value = context.Output(2, present_kv_shape);
   parameters.past_present_share_buffer_ = present_key != nullptr && present_value != nullptr && past_key != nullptr && past_value != nullptr && past_key->DataRaw() == present_key->DataRaw() && past_value->DataRaw() == present_value->DataRaw();
 
+  ORT_ENFORCE(parameters.total_sequence_length_ <= parameters.seqlen_present_kv_cache_, "Total sequence length cannot be greater than the existing KV cache length.");
+  // Use a sliding window if the total sequence exceeds the window's length.
+  bool use_sliding_window = (local_window_size_ != -1 && local_window_size_ < parameters.total_sequence_length_);
   if (!do_rotary_ &&
       head_sink == nullptr && !use_smooth_softmax_ &&
-      local_window_size_ == -1 &&
+      !use_sliding_window &&
       CanApplyFlashAttention(attention_bias, present_key, present_value, parameters, context)) {
     return ApplyFlashAttention(query, key, value, attention_bias, output, past_key, present_key, past_value,
                                present_value, parameters, context);

tools/ci_build/github/azure-pipelines/c-api-noopenmp-test-pipelines.yml

@@ -121,109 +121,6 @@ stages:
   parameters:
     StageSuffix: 'macOS_CPU_x64'
 
-- template: templates/final-jar-testing.yml
-  parameters:
-    OS: Windows
-    PoolName: 'onnxruntime-Win-CPU-2022'
-
-- template: templates/final-jar-testing.yml
-  parameters:
-    OS: Linux
-    PoolName: 'onnxruntime-Ubuntu2204-AMD-CPU'
-
-- template: templates/final-jar-testing.yml
-  parameters:
-    OS: MacOS
-    PoolName: 'macOS-14'
-
-
-- stage: GPU_JAR_Testing
-  dependsOn: Setup
-  jobs:
-  - job: Final_Jar_Testing_Windows_GPU
-    workspace:
-      clean: all
-    pool: 'onnxruntime-Win2022-GPU-A10'
-    timeoutInMinutes: 60
-    variables:
-    - name: runCodesignValidationInjection
-      value: false
-
-    steps:
-    - template: templates/set-version-number-variables-step.yml
-
-    - template: templates/jobs/download_win_gpu_library.yml
-      parameters:
-        CudaVersion: 12.2
-        DownloadCUDA: true
-        DownloadTRT: true
-
-    - template: templates/flex-downloadPipelineArtifact.yml
-      parameters:
-        StepName: 'Download Final Jar'
-        ArtifactName: onnxruntime-java-gpu
-        TargetPath: '$(Build.BinariesDirectory)\final-jar'
-
-    - template: templates/flex-downloadPipelineArtifact.yml
-      parameters:
-        StepName: 'Download Jar Tools'
-        ArtifactName: onnxruntime-java-tools
-        TargetPath: '$(Build.BinariesDirectory)\final-jar'
-
-    - task: CmdLine@2
-      inputs:
-        script: |
-          mkdir test
-          pushd test
-          jar xf $(Build.BinariesDirectory)\final-jar\testing.jar
-          popd
-          java -DUSE_CUDA=1 -jar junit-platform-console-standalone-1.6.2.jar -cp .;.\test;protobuf-java-3.25.5.jar;onnxruntime_gpu-$(OnnxRuntimeVersion).jar --scan-class-path --fail-if-no-tests --disable-banner
-        workingDirectory: '$(Build.BinariesDirectory)\final-jar'
-
-  - job: Final_Jar_Testing_Linux_GPU
-    workspace:
-      clean: all
-    pool:
-      name: 'Onnxruntime-Linux-GPU-A10'
-      os: linux
-    variables:
-    - name: runCodesignValidationInjection
-      value: false
-    - name: docker_base_image
-      value: onnxruntimebuildcache.azurecr.io/internal/azureml/onnxruntime/build/cuda12_x64_almalinux8_gcc12:20250724.1
-    timeoutInMinutes: 60
-    steps:
-    - checkout: self
-      submodules: false
-    - template: templates/set-version-number-variables-step.yml
-
-    - template: templates/flex-downloadPipelineArtifact.yml
-      parameters:
-        StepName: 'Download Final Jar'
-        ArtifactName: onnxruntime-java-gpu
-        TargetPath: '$(Build.BinariesDirectory)/final-jar'
-
-    - template: templates/get-docker-image-steps.yml
-      parameters:
-        Dockerfile: tools/ci_build/github/linux/docker/Dockerfile.package_ubi8_cuda_tensorrt10_0
-        Context: tools/ci_build/github/linux/docker/
-        DockerBuildArgs: "
-        --build-arg BUILD_UID=$( id -u )
-        --build-arg BASEIMAGE=${{ variables.docker_base_image }}
-        --build-arg TRT_VERSION=${{ variables.linux_trt_version }}
-        "
-        Repository: onnxruntimeubi8packagestest
-
-    - bash: |
-        docker run -e SYSTEM_COLLECTIONURI --rm \
-          --gpus all \
-          --volume $(Build.SourcesDirectory):/onnxruntime_src \
-          --volume $(Build.BinariesDirectory):/build \
-          --volume /data/models:/build/models:ro \
-          onnxruntimeubi8packagestest \
-          /bin/bash /onnxruntime_src/tools/ci_build/github/linux/java_linux_final_test.sh -r /build -v $(OnnxRuntimeVersion)
-      displayName: 'Test'
-
 - template: nuget/templates/test_win.yml
   parameters:
     AgentPool: 'onnxruntime-Win2022-GPU-A10'
@@ -274,13 +171,14 @@ stages:
           clean: true
           submodules: none
 
-        - template: templates/flex-downloadPipelineArtifact.yml
-          parameters:
-            ArtifactName: "Windows_Packaging_cuda_build_artifacts"
-            StepName: 'Download Pipeline Artifact - Windows GPU Packages Build'
-            TargetPath: '$(Build.BinariesDirectory)/RelWithDebInfo/'
+        - download: build
+          artifact: 'Windows_Packaging_cuda_build_artifacts'
+          displayName: 'Download Windows GPU Packages Build'
 
-        - template: templates/telemetry-steps.yml
+        - task: CmdLine@2
+          inputs:
+            script: |
+              move $(Pipeline.Workspace)/build/Windows_Packaging_cuda_build_artifacts $(Build.BinariesDirectory)/RelWithDebInfo
 
         - template: templates/set-version-number-variables-step.yml
 
@@ -323,7 +221,7 @@ stages:
           displayName: 'test'
           inputs:
             scriptPath: '$(Build.SourcesDirectory)\tools\ci_build\build.py'
-            arguments: '--config RelWithDebInfo --use_binskim_compliant_compile_flags --enable_lto --disable_rtti --build_dir $(Build.BinariesDirectory) --skip_submodule_sync --build_shared_lib --test --enable_onnx_tests  $(TelemetryOption) '
+            arguments: '--config RelWithDebInfo --use_binskim_compliant_compile_flags --enable_lto --disable_rtti --build_dir $(Build.BinariesDirectory) --skip_submodule_sync --build_shared_lib --test --enable_onnx_tests'
             workingDirectory: '$(Build.BinariesDirectory)'
         # Previous stage only assembles the java binaries, testing will be done in this stage with GPU machine
         - template: templates/make_java_win_binaries.yml
@@ -345,13 +243,15 @@ stages:
           clean: true
           submodules: none
 
-        - template: templates/flex-downloadPipelineArtifact.yml
-          parameters:
-            ArtifactName: "Windows_Packaging_tensorrt_build_artifacts"
-            StepName: 'Download Pipeline Artifact - Windows GPU Packages Build'
-            TargetPath: '$(Build.BinariesDirectory)/RelWithDebInfo/'
 
-        - template: templates/telemetry-steps.yml
+        - download: build
+          artifact: 'Windows_Packaging_tensorrt_build_artifacts'
+          displayName: 'Download Windows GPU Packages Build'
+
+        - task: CmdLine@2
+          inputs:
+            script: |
+              move $(Pipeline.Workspace)/build/Windows_Packaging_tensorrt_build_artifacts $(Build.BinariesDirectory)/RelWithDebInfo
 
         - template: templates/set-version-number-variables-step.yml
 
@@ -394,7 +294,7 @@ stages:
           displayName: 'test'
           inputs:
             scriptPath: '$(Build.SourcesDirectory)\tools\ci_build\build.py'
-            arguments: '--config RelWithDebInfo --use_binskim_compliant_compile_flags --enable_lto --disable_rtti --build_dir $(Build.BinariesDirectory) --skip_submodule_sync --build_shared_lib --test --enable_onnx_tests  $(TelemetryOption) '
+            arguments: '--config RelWithDebInfo --use_binskim_compliant_compile_flags --enable_lto --disable_rtti --build_dir $(Build.BinariesDirectory) --skip_submodule_sync --build_shared_lib --test --enable_onnx_tests'
             workingDirectory: '$(Build.BinariesDirectory)'
         # Previous stage only assembles the java binaries, testing will be done in this stage with GPU machine
         - template: templates/make_java_win_binaries.yml

tools/ci_build/github/azure-pipelines/custom-nuget-packaging-pipeline.yml

@@ -157,7 +157,7 @@ extends:
               targetType: 'inline'
               script: |
                 mkdir -p $(Build.BinariesDirectory)/osx-x64
-                 Move-Item -Path $(Build.BinariesDirectory)/osx/onnxruntime-osx-x86_64* -Destination $(Build.BinariesDirectory)/osx-x64
+                Move-Item -Path $(Build.BinariesDirectory)/osx/onnxruntime-osx-x86_64* -Destination $(Build.BinariesDirectory)/osx-x64
 
                 mkdir -p $(Build.BinariesDirectory)/osx-arm64
                 Move-Item -Path $(Build.BinariesDirectory)/osx/onnxruntime-osx-arm64* -Destination $(Build.BinariesDirectory)/osx-arm64

tools/ci_build/github/azure-pipelines/jar_package_testing.yml

@@ -0,0 +1,162 @@
+resources:
+  pipelines:
+  - pipeline: build
+    source: 'Zip-Nuget-Java-Nodejs Packaging Pipeline'
+    trigger: true
+    branch: main
+
+variables:
+  mavenVersion: '3.9.8'
+
+stages:
+- template: templates/final-jar-testing-win.yml
+  parameters:
+    PoolName: 'onnxruntime-Win-CPU-2022'
+
+- template: templates/final-jar-testing-linux.yml
+  parameters:
+    OS: Linux
+    PoolName: 'onnxruntime-Ubuntu2204-AMD-CPU'
+
+- template: templates/final-jar-testing-linux.yml
+  parameters:
+    OS: MacOS
+    PoolName: 'macOS-14'
+
+- stage: GPU_JAR_Testing
+  dependsOn: []
+  jobs:
+  - job: Final_Jar_Testing_Windows_GPU
+    workspace:
+      clean: all
+    pool: 'onnxruntime-Win2022-GPU-A10'
+    timeoutInMinutes: 60
+    variables:
+    - name: runCodesignValidationInjection
+      value: false
+
+    steps:
+    - template: templates/set-version-number-variables-step.yml
+
+    - template: templates/jobs/download_win_gpu_library.yml
+      parameters:
+        CudaVersion: 12.2
+        DownloadCUDA: true
+        DownloadTRT: true
+
+    - template: templates/download_maven_for_tests.yml
+    - download: build
+      artifact: 'onnxruntime-java-gpu'
+      displayName: 'Download Final Jar'
+    - script: |
+        move $(Pipeline.Workspace)\build\onnxruntime-java-gpu\*.jar $(Pipeline.Workspace)\build\onnxruntime-java\
+
+    - task: PowerShell@2
+      displayName: 'Run Java Tests with PowerShell'
+      inputs:
+        targetType: 'inline'
+        script: |
+          # Exit script on any error
+          $ErrorActionPreference = "Stop"
+
+          cd $(Pipeline.Workspace)/build/onnxruntime-java
+          del *.asc
+          del *.sha256
+          del *.sha512
+          del *.pom
+          cd ..
+          mkdir tests
+          cd tests
+          jar xf $(Pipeline.Workspace)/build/onnxruntime-java/testing.jar
+          del $(Pipeline.Workspace)/build/onnxruntime-java/testing.jar
+          dir $(Pipeline.Workspace)/build/tests
+          Write-Host "Running JUnit Tests..."
+          & java -DUSE_CUDA=1 `
+            -cp "$(Pipeline.Workspace)\build\tests;$(Pipeline.Workspace)\build\onnxruntime-java\*" org.junit.platform.console.ConsoleLauncher --scan-classpath=$(Pipeline.Workspace)\build\tests `
+            --fail-if-no-tests --disable-banner --reports-dir "$($env:Build_ArtifactStagingDirectory)/TestResults"
+
+    - task: PublishTestResults@2
+      displayName: 'Publish Test Results'
+      inputs:
+        testResultsFormat: 'JUnit'
+        testResultsFiles: '$(Build.ArtifactStagingDirectory)/TestResults/TEST-junit-jupiter.xml'
+        failTaskOnFailedTests: true
+
+
+  - job: Final_Jar_Testing_Linux_GPU
+    workspace:
+      clean: all
+    pool:
+      name: 'Onnxruntime-Linux-GPU-A10'
+    variables:
+    - name: runCodesignValidationInjection
+      value: false
+    - name: docker_base_image
+      value: onnxruntimebuildcache.azurecr.io/internal/azureml/onnxruntime/build/cuda12_x64_almalinux8_gcc12:20250724.1
+    timeoutInMinutes: 60
+    steps:
+    - checkout: self
+      submodules: false
+
+    - template: templates/set-version-number-variables-step.yml
+
+    - bash: |
+       sudo apt-get install -y msopenjdk-17
+       dpkg -l msopenjdk-17
+
+    - bash: |
+        echo "Downloading and installing Maven $(mavenVersion) for Linux..."
+        MAVEN_DIR="$(Agent.TempDirectory)/apache-maven-$(mavenVersion)"
+        # Download Maven binary
+        wget https://archive.apache.org/dist/maven/maven-3/$(mavenVersion)/binaries/apache-maven-$(mavenVersion)-bin.tar.gz -O $(Agent.TempDirectory)/maven.tar.gz
+
+        # Extract to the temp directory
+        mkdir -p ${MAVEN_DIR}
+        tar -xzf $(Agent.TempDirectory)/maven.tar.gz -C $(Agent.TempDirectory)
+
+        # Add Maven's bin directory to the PATH for subsequent tasks in the job
+        echo "##vso[task.prependpath]${MAVEN_DIR}/bin"
+      displayName: 'Install Maven (Linux)'
+
+    - script: |
+        echo "Maven is now on the PATH."
+        mvn --version
+
+    - download: build
+      artifact: 'onnxruntime-java-gpu'
+      displayName: 'Download Final Jar'
+
+    # Rename the downloaded folder
+    - script: |
+        mv $(Pipeline.Workspace)/build/onnxruntime-java-gpu $(Pipeline.Workspace)/build/onnxruntime-java
+
+    - task: Maven@4
+      displayName: 'Download Dependencies'
+      inputs:
+        mavenPomFile: '$(Build.SourcesDirectory)/tools/ci_build/java/pom.xml'
+        goals: 'dependency:copy-dependencies'
+        options: '-DoutputDirectory=$(Pipeline.Workspace)/build/onnxruntime-java'
+        publishJUnitTestResults: false
+        javaHomeOption: 'Path'
+        jdkDirectory: '/usr/lib/jvm/msopenjdk-17-amd64'
+        jdkVersionOption: 'Default'
+        mavenVersionOption: 'Default'
+
+    # Now all the jars are in the $(Pipeline.Workspace)/build folder
+
+    - template: templates/get-docker-image-steps.yml
+      parameters:
+        Dockerfile: tools/ci_build/github/linux/docker/Dockerfile.package_ubi8_cuda_tensorrt10_0
+        Context: tools/ci_build/github/linux/docker/
+        DockerBuildArgs: "--build-arg BUILD_UID=$( id -u ) --build-arg BASEIMAGE=${{ variables.docker_base_image }} --build-arg TRT_VERSION=${{ variables.linux_trt_version }}"
+        Repository: onnxruntimeubi8packagestest
+
+    - bash: |
+        docker run --network=none --rm \
+          --gpus all \
+          --volume $(Build.SourcesDirectory):/onnxruntime_src \
+          --volume $(Pipeline.Workspace)/build:/build \
+          --volume /data/models:/build/models:ro \
+          onnxruntimeubi8packagestest \
+          /bin/bash /onnxruntime_src/tools/ci_build/github/linux/java_linux_final_test.sh -r /build -v $(OnnxRuntimeVersion)
+      displayName: 'Test'

tools/ci_build/github/azure-pipelines/nodejs/templates/test.yml

@@ -4,8 +4,13 @@ steps:
   displayName: 'Download NPM_packages'
   artifact: 'NPM_packages'
 
-- script: |
-    mv $(Pipeline.Workspace)/build/NPM_packages '$(Build.BinariesDirectory)/nodejs-artifact'
+
+- task: PowerShell@2
+  displayName: 'Move Artifact Directory'
+  inputs:
+    targetType: 'inline'
+    script: |
+      Move-Item -Path "$(Pipeline.Workspace)/build/NPM_packages" -Destination "$(Build.BinariesDirectory)/nodejs-artifact"
 
 - script: mkdir e2e_test
   workingDirectory: '$(Build.BinariesDirectory)'

tools/ci_build/github/azure-pipelines/nodejs/templates/test_macos.yml

@@ -11,9 +11,7 @@ stages:
       clean: all
     timeoutInMinutes:  120
     pool:
-      name: 'Azure Pipelines'
-      image: 'macOS-15'
-      os: 'macOS'
+      vmImage: 'macOS-15'
 
     variables:
     - name: OnnxRuntimeBuildDirectory

tools/ci_build/github/azure-pipelines/npm-packaging-pipeline.yml

@@ -50,6 +50,11 @@ extends:
       sourceAnalysisPool:
         name: onnxruntime-Win-CPU-2022
         os: windows
+      codeql:
+        compiled:
+          enabled: false
+          justificationForDisabling: 'CodeQL causes the React Native Android tests to fail when trying to load Linux x64 .so'
+
     stages:
     - template: templates/web-ci.yml
       parameters: