Merge branch 'main' into kvaishnavi/logit-softcapping

.github/policies/issueLabeler.yml

@@ -1,6 +1,9 @@
+id:
 name: Issue Triage
 description: Assign label to issues 
+owner:
 resource: repository
+where:
 configuration:
   resourceManagementConfiguration:
     eventResponderTasks:
@@ -79,3 +82,5 @@ configuration:
             then:
               - addLabel:
                   label: model:transformer
+onFailure:
+onSuccess:

.github/policies/test_issueLabeler.yml

@@ -0,0 +1,24 @@
+id:
+name: Issue Triage
+description: Assign label to issues
+owner:
+resource: repository
+where:
+configuration:
+  resourceManagementConfiguration:
+    eventResponderTasks:
+      - if:
+          - payloadType: Issues
+          - isOpen
+        then:
+          - if:
+              - or:
+                  - titleContains:
+                      pattern: shark
+                  - bodyContains:
+                      pattern: strawberry
+            then:
+              - addLabel:
+                  label: wontfix 
+onFailure:
+onSuccess:

.github/workflows/linux-cpu-arm64-build.yml

@@ -25,19 +25,17 @@ jobs:
         with:
           submodules: 'true'
 
-      - name: Install jq
-        run: |
-          sudo apt-get install jq
-
       - uses: actions/setup-dotnet@v4
         with:
           dotnet-version: '8.0.x'
 
       - name: Get the Latest OnnxRuntime Nightly Version
+        shell: pwsh
         run: |
-          ORT_NIGHTLY_VERSION=$(curl -s "${{ env.ORT_NIGHTLY_REST_API }}" | jq -r '.value[0].versions[0].normalizedVersion')
-          echo "$ORT_NIGHTLY_VERSION" 
-          echo "ORT_NIGHTLY_VERSION=$ORT_NIGHTLY_VERSION" >> $GITHUB_ENV
+          $resp = Invoke-RestMethod "${{ env.ORT_NIGHTLY_REST_API }}"
+          $ORT_NIGHTLY_VERSION = $resp.value[0].versions[0].normalizedVersion
+          Write-Host "$ORT_NIGHTLY_VERSION"
+          "ORT_NIGHTLY_VERSION=$ORT_NIGHTLY_VERSION" | Out-File -FilePath $env:GITHUB_ENV -Append
 
       - name: Download OnnxRuntime Nightly
         run: |
@@ -72,7 +70,7 @@ jobs:
           --container-registry onnxruntimebuildcache \
           --repository ort_genai_linux_arm64_gha
 
-      - name: Doker -- Configure with CMake and GCC
+      - name: Docker -- Configure with CMake and GCC
         run: |
           docker run --rm \
           --volume $GITHUB_WORKSPACE:/onnxruntime_src \
@@ -84,7 +82,7 @@ jobs:
           --volume $GITHUB_WORKSPACE:/onnxruntime_src \
           -w /onnxruntime_src ort_genai_linux_arm64_gha bash -c "/usr/bin/cmake --build --preset linux_gcc_cpu_release"
 
-      - name: Dokcer -- check test directory
+      - name: Docker -- Check test directory
         run: |
           docker run --rm \
           --volume $GITHUB_WORKSPACE:/onnxruntime_src \

.github/workflows/linux-cpu-x64-build.yml

@@ -23,19 +23,17 @@ jobs:
         with:
           submodules: true
 
-      - name: Install jq
-        run: |
-          sudo apt-get install jq
-
       - uses: actions/setup-dotnet@v4
         with:
           dotnet-version: '8.0.x'
 
       - name: Get the Latest OnnxRuntime Nightly Version
+        shell: pwsh
         run: |
-          ORT_NIGHTLY_VERSION=$(curl -s "${{ env.ORT_NIGHTLY_REST_API }}" | jq -r '.value[0].versions[0].normalizedVersion')
-          echo "$ORT_NIGHTLY_VERSION" 
-          echo "ORT_NIGHTLY_VERSION=$ORT_NIGHTLY_VERSION" >> $GITHUB_ENV
+          $resp = Invoke-RestMethod "${{ env.ORT_NIGHTLY_REST_API }}"
+          $ORT_NIGHTLY_VERSION = $resp.value[0].versions[0].normalizedVersion
+          Write-Host "$ORT_NIGHTLY_VERSION"
+          "ORT_NIGHTLY_VERSION=$ORT_NIGHTLY_VERSION" | Out-File -FilePath $env:GITHUB_ENV -Append
 
       - name: Download OnnxRuntime Nightly
         run: |
@@ -71,7 +69,7 @@ jobs:
         run: |
           python3 -m pip install -r test/python/requirements.txt --user
           python3 -m pip install -r test/python/requirements-cpu.txt --user
-          python3 -m pip install build/cpu/wheel/onnxruntime_genai*.whl --user
+          python3 -m pip install build/cpu/wheel/onnxruntime_genai*.whl --user --no-deps
 
       - name: Get HuggingFace Token
         run: |
@@ -90,6 +88,11 @@ jobs:
         run: |
           python3 test/python/test_onnxruntime_genai.py --cwd test/python --test_models test/test_models
 
+      - name: Build the C# API and Run the C# Tests
+        run: |
+          cd test/csharp
+          dotnet test /p:Configuration=Release /p:NativeBuildOutputDir="../../build/cpu/"
+
       - name: Verify Build Artifacts
         if: always()
         continue-on-error: true

.github/workflows/linux-cpu-x64-nightly-build.yml

@@ -52,7 +52,7 @@ jobs:
         run: |
           python3 -m pip install -r test/python/requirements.txt --user
           python3 -m pip install -r test/python/requirements-cpu.txt --user
-          python3 -m pip install build/cpu/wheel/onnxruntime_genai*.whl
+          python3 -m pip install build/cpu/wheel/onnxruntime_genai*.whl --no-deps
 
       - name: Get HuggingFace Token
         run: |

.github/workflows/linux-gpu-x64-build.yml

@@ -37,19 +37,17 @@ jobs:
           path: manylinux
           submodules: true
 
-      - name: Install jq
-        run: |
-          sudo apt-get install jq
-
       - uses: actions/setup-dotnet@v4
         with:
           dotnet-version: '8.0.x'
 
-      - name: Download OnnxRuntime
+      - name: Get the Latest OnnxRuntime Nightly Version
+        shell: pwsh
         run: |
-          ORT_NIGHTLY_VERSION=$(curl -s "${{ env.ORT_NIGHTLY_REST_API }}" | jq -r '.value[0].versions[0].normalizedVersion')
-          echo "$ORT_NIGHTLY_VERSION" 
-          echo "ORT_NIGHTLY_VERSION=$ORT_NIGHTLY_VERSION" >> $GITHUB_ENV
+          $resp = Invoke-RestMethod "${{ env.ORT_NIGHTLY_REST_API }}"
+          $ORT_NIGHTLY_VERSION = $resp.value[0].versions[0].normalizedVersion
+          Write-Host "$ORT_NIGHTLY_VERSION"
+          "ORT_NIGHTLY_VERSION=$ORT_NIGHTLY_VERSION" | Out-File -FilePath $env:GITHUB_ENV -Append
 
       - name: Download OnnxRuntime Nightly
         run: |
@@ -65,7 +63,6 @@ jobs:
           ls -R ${{ env.ORT_PACKAGE_NAME }}
         continue-on-error: true
 
-# TODO: Find out why do we need to to have libonnxruntime.so.$ort_version
       - name: Extract OnnxRuntime library and header files
         run: |
           set -e -x
@@ -129,19 +126,32 @@ jobs:
           docker run \
             --gpus all \
             --rm \
+            --volume /data/ortgenai_pytorch_models:/data/ortgenai_pytorch_models \
             --volume $GITHUB_WORKSPACE:/ort_genai_src \
             -e HF_TOKEN=$HF_TOKEN \
             -w /ort_genai_src onnxruntimecudabuildx64 bash -c " \
               ${{ env.PYTHON_EXECUTABLE }} -m pip install -r test/python/requirements.txt --user && \
               ${{ env.PYTHON_EXECUTABLE }} -m pip install -r test/python/requirements-cuda.txt --user && \
-              ${{ env.PYTHON_EXECUTABLE }} -m pip install /ort_genai_src/build/cuda/wheel/onnxruntime_genai*manylinux*.whl --user && \
-              ${{ env.PYTHON_EXECUTABLE }} test/python/test_onnxruntime_genai.py --cwd test/python --test_models test/test_models"
+              ${{ env.PYTHON_EXECUTABLE }} -m pip install /ort_genai_src/build/cuda/wheel/onnxruntime_genai*manylinux*.whl --no-deps --user && \
+              ${{ env.PYTHON_EXECUTABLE }} test/python/test_onnxruntime_genai.py --cwd test/python --test_models test/test_models --e2e"
+
+      # TODO: Enable this by adding dotnet to the docker image
+      # - name: Build the C# API and Run the C# Tests
+      #   run: |
+      #     echo "Building the C# API and running the C# tests"
+      #     docker run \
+      #       --gpus all \
+      #       --rm \
+      #       --volume $GITHUB_WORKSPACE:/ort_genai_src \
+      #       -w /ort_genai_src/test/csharp onnxruntimecudabuildx64 bash -c " \
+      #         dotnet test /p:NativeBuildOutputDir='/ort_genai_src/build/cuda/'"
 
       - name: Docker -- Run unit tests
         run: |
           echo "Running docker image onnxruntimecudabuildx64"
           docker run \
             --gpus all \
             --rm \
+            --volume /data/ortgenai_pytorch_models:/data/ortgenai_pytorch_models \
             --volume $GITHUB_WORKSPACE:/ort_genai_src \
             -w /ort_genai_src onnxruntimecudabuildx64 bash -c "LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/ort_genai_src/build/cuda/ /ort_genai_src/build/cuda/test/unit_tests"

.github/workflows/win-cpu-arm64-build.yml

@@ -68,11 +68,9 @@ jobs:
 
     - name: Install the Python Wheel and Test Dependencies
       run: |
-        python -c "import platform; print(platform.machine()); print(platform.version()); print(platform.platform()); print(platform.system()); print(platform.processor()); print(platform.architecture())"
-        python -c "import sys; print(sys.version)"
         python -m pip install "numpy<2" coloredlogs flatbuffers packaging protobuf sympy pytest
         python -m pip install -i https://aiinfra.pkgs.visualstudio.com/PublicPackages/_packaging/ORT-Nightly/pypi/simple/ ort-nightly-qnn
-        python -m pip install (Get-ChildItem ("$env:binaryDir\wheel\*.whl"))
+        python -m pip install (Get-ChildItem ("$env:binaryDir\wheel\*.whl")) --no-deps
 
     - name: Run the Python Tests
       run: |

.github/workflows/win-cpu-x64-build.yml

@@ -72,16 +72,11 @@ jobs:
       run: |
         cmake --build --preset windows_x64_cpu_release --parallel
 
-    - name: Build the C# API and Run the C# Tests
-      run: |
-        cd test\csharp
-        dotnet test /p:NativeBuildOutputDir="$env:GITHUB_WORKSPACE\$env:binaryDir\Release"
-
     - name: Install the python wheel and test dependencies
       run: |
         python3 -m pip install -r test\python\requirements.txt --user
         python3 -m pip install -r test\python\requirements-cpu.txt --user
-        python -m pip install (Get-ChildItem ("$env:binaryDir\wheel\*.whl"))
+        python3 -m pip install (Get-ChildItem ("$env:binaryDir\wheel\*.whl")) --no-deps
 
     - name: Get HuggingFace Token
       run: |
@@ -94,7 +89,10 @@ jobs:
       run: |
         python test/python/test_onnxruntime_genai.py --cwd "test\python" --test_models "test\test_models"
 
-
+    - name: Build the C# API and Run the C# Tests
+      run: |
+        cd test\csharp
+        dotnet test /p:NativeBuildOutputDir="$env:GITHUB_WORKSPACE\$env:binaryDir\Release"
 
     - name: Verify Build Artifacts
       if: always()

.github/workflows/win-cuda-x64-build.yml

@@ -82,7 +82,7 @@ jobs:
       run: |
         python -m pip install -r test\python\requirements.txt
         python -m pip install -r test\python\requirements-cuda.txt
-        python -m pip install (Get-ChildItem ("$env:binaryDir\wheel\*.whl"))
+        python -m pip install (Get-ChildItem ("$env:binaryDir\wheel\*.whl")) --no-deps
 
     - name: Get HuggingFace Token
       run: |
@@ -93,8 +93,7 @@ jobs:
 
     - name: Run the Python Tests
       run: |
-        python test/python/test_onnxruntime_genai.py --cwd "test\python" --test_models "test\test_models"
-
+        python test/python/test_onnxruntime_genai.py --cwd "test\python" --test_models "test\test_models" --e2e
 
     - name: Verify Build Artifacts
       if: always()

.github/workflows/win-directml-x64-build.yml

@@ -18,9 +18,9 @@ env:
   ort_zip: "Microsoft.ML.OnnxRuntime.DirectML.1.17.3.zip"
   # TODO: Update with nightly ORT-DML build
   ort_url: "https://github.com/microsoft/onnxruntime/releases/download/v1.17.3/Microsoft.ML.OnnxRuntime.DirectML.1.17.3.zip"
-  dml_dir: "Microsoft.AI.DirectML.1.15.0"
-  dml_zip: "Microsoft.AI.DirectML.1.15.0.zip"
-  dml_url: "https://www.nuget.org/api/v2/package/Microsoft.AI.DirectML/1.15.0"
+  dml_dir: "Microsoft.AI.DirectML.1.15.1"
+  dml_zip: "Microsoft.AI.DirectML.1.15.1.zip"
+  dml_url: "https://www.nuget.org/api/v2/package/Microsoft.AI.DirectML/1.15.1"
   d3d12_dir: "Microsoft.Direct3D.D3D12.1.614.0"
   d3d12_zip: "Microsoft.Direct3D.D3D12.1.614.0.zip"
   d3d12_url: "https://www.nuget.org/api/v2/package/Microsoft.Direct3D.D3D12/1.614.0"
@@ -92,7 +92,7 @@ jobs:
       run: |
         python -m pip install -r test\python\requirements.txt
         python -m pip install -r test\python\requirements-directml.txt
-        python -m pip install (Get-ChildItem ("$env:binaryDir\wheel\*.whl"))
+        python -m pip install (Get-ChildItem ("$env:binaryDir\wheel\*.whl")) --no-deps
 
     - name: Verify Build Artifacts
       if: always()

.pipelines/stages/jobs/capi-packaging-job.yml

@@ -74,11 +74,11 @@ jobs:
   - name: ortHome
     value: 'ort'
   - name: dml_dir
-    value: 'Microsoft.AI.DirectML.1.15.0'
+    value: 'Microsoft.AI.DirectML.1.15.1'
   - name: dml_zip
-    value: 'Microsoft.AI.DirectML.1.15.0.zip'
+    value: 'Microsoft.AI.DirectML.1.15.1.zip'
   - name: dml_url
-    value: "https://www.nuget.org/api/v2/package/Microsoft.AI.DirectML/1.15.0"
+    value: "https://www.nuget.org/api/v2/package/Microsoft.AI.DirectML/1.15.1"
 
   - name: d3d12_dir
     value: 'Microsoft.Direct3D.D3D12.1.614.0'

.pipelines/stages/jobs/nuget-validation-job.yml

@@ -165,7 +165,7 @@ jobs:
        cd examples/csharp/HelloPhi
        mv models/$(prebuild_phi3_mini_model_folder) models/phi-3
        dotnet restore -r $(os)-$(arch) /property:Configuration=$(csproj_configuration) --source https://api.nuget.org/v3/index.json --source https://aiinfra.pkgs.visualstudio.com/PublicPackages/_packaging/ORT-Nightly/nuget/v3/index.json --source $PWD --disable-parallel --verbosity detailed
-       dotnet build ./HelloPhi.csproj -r $(os)-$(arch) /property:Configuration=$(csproj_configuration)
+       dotnet build ./HelloPhi.csproj -r $(os)-$(arch) /property:Configuration=$(csproj_configuration) --no-restore --self-contained
       displayName: 'Perform dotnet restore & build'
       workingDirectory: '$(Build.Repository.LocalPath)'
       env:
@@ -188,9 +188,9 @@ jobs:
             -w /ort_genai_src/ $(cuda_docker_image) \
             bash -c " \
                 export ORTGENAI_LOG_ORT_LIB=1 && \
-                dotnet --info && \
                 cd /ort_genai_src/examples/csharp/HelloPhi && \
-                dotnet run -r $(os)-$(arch) --configuration $(csproj_configuration) --no-build --verbosity normal -- -m ./models/phi-3"
+                chmod +x ./bin/Release_Cuda/net6.0/linux-x64/HelloPhi && \
+                ./bin/Release_Cuda/net6.0/linux-x64/HelloPhi -m ./models/phi-3"
 
         displayName: 'Run Example With Artifact'
         workingDirectory: '$(Build.Repository.LocalPath)'

.pipelines/stages/jobs/py-packaging-job.yml

@@ -111,11 +111,11 @@ jobs:
       value: 'Microsoft.ML.OnnxRuntime'
 
   - name: dml_dir
-    value: 'Microsoft.AI.DirectML.1.15.0'
+    value: 'Microsoft.AI.DirectML.1.15.1'
   - name: dml_zip
-    value: 'Microsoft.AI.DirectML.1.15.0.zip'
+    value: 'Microsoft.AI.DirectML.1.15.1.zip'
   - name: dml_url
-    value: "https://www.nuget.org/api/v2/package/Microsoft.AI.DirectML/1.15.0"
+    value: "https://www.nuget.org/api/v2/package/Microsoft.AI.DirectML/1.15.1"
 
   - name: d3d12_dir
     value: 'Microsoft.Direct3D.D3D12.1.614.0'

.pipelines/stages/jobs/steps/capi-linux-step.yml

@@ -122,8 +122,9 @@ steps:
       docker run \
       --rm \
       --volume $(Build.Repository.LocalPath):/ort_genai_src \
-      -w /ort_genai_src/ ortgenai$(ep)build$(arch) \
+      -w /ort_genai_src/ \
       -e ONNXRUNTIME_VERSION=$(ONNXRUNTIME_VERSION) \
+      ortgenai$(ep)build$(arch) \
       bash -c " \
           /usr/bin/cmake --preset linux_gcc_$(ep)_$(build_config) \
             -DENABLE_TESTS=OFF \
@@ -145,8 +146,9 @@ steps:
       docker run \
       --rm \
       --volume $(Build.Repository.LocalPath):/ort_genai_src \
-      -w /ort_genai_src/ ortgenai$(ep)build$(arch) \
+      -w /ort_genai_src/ \
       -e ONNXRUNTIME_VERSION=$(ONNXRUNTIME_VERSION) \
+      ortgenai$(ep)build$(arch) \
       bash -c " \
           /usr/bin/cmake --build --preset linux_gcc_$(ep)_$(build_config) \
             -DENABLE_TESTS=OFF \