Merge branch 'main' into add-dim-order-clone-aot

Author: keyprocedure

.ci/docker/ci_commit_pins/optimum-executorch.txt

@@ -1 +1 @@
-36e3dd54effb3f6d13d792029609292fdd5502bb
+40b02a2dc61bbf901a2df91719f47c98d65368ec

.ci/scripts/build-qnn-sdk.sh

@@ -11,8 +11,10 @@ set -o xtrace
 
 build_qnn_backend() {
   echo "Start building qnn backend."
-  export ANDROID_NDK_ROOT=${ANDROID_NDK_ROOT:-/opt/ndk}
-  export QNN_SDK_ROOT=${QNN_SDK_ROOT:-/tmp/qnn/2.28.0.241029}
+  # Source QNN configuration
+  source "$(dirname "${BASH_SOURCE[0]}")/../../backends/qualcomm/scripts/install_qnn_sdk.sh"
+  setup_android_ndk
+  install_qnn
   export EXECUTORCH_ROOT="$(cd -- "$(dirname -- "${BASH_SOURCE[0]}")/../.." && pwd)"
 
   parallelism=$(( $(nproc) - 1 ))

.ci/scripts/setup-qnn-deps.sh

@@ -10,4 +10,5 @@ set -ex
 source "$(dirname "${BASH_SOURCE[0]}")/../../backends/qualcomm/scripts/install_qnn_sdk.sh"
 
 setup_libcpp 12
+setup_android_ndk
 install_qnn

.ci/scripts/setup-vulkan-linux-deps.sh

@@ -43,7 +43,7 @@ install_vulkan_sdk() {
   export PATH="${PATH}:${_vulkan_sdk_dir}/${VULKAN_SDK_VERSION}/x86_64/bin/"
 }
 
-VULKAN_SDK_VERSION="1.3.296.0"
+VULKAN_SDK_VERSION="1.4.321.1"
 
 install_swiftshader
 install_vulkan_sdk "${VULKAN_SDK_VERSION}"

.ci/scripts/test_llama.sh

@@ -119,8 +119,12 @@ echo "COREML option ${COREML}"
 
 if [[ "${MODE}" =~ .*qnn.* ]]; then
   QNN=ON
+
+  # Download QNN_SDK. If already downloaded, export environment path
+  source "$(dirname "${BASH_SOURCE[0]}")/../../backends/qualcomm/scripts/install_qnn_sdk.sh"
+  install_qnn
+
   export EXECUTORCH_ROOT="$(cd -- "$(dirname -- "${BASH_SOURCE[0]}")/.." && pwd)"
-  export QNN_SDK_ROOT=/tmp/qnn/2.28.0.241029
   export LD_LIBRARY_PATH="${QNN_SDK_ROOT}/lib/x86_64-linux-clang"
   export PYTHONPATH=".."
   cp schema/program.fbs exir/_serialize/program.fbs
@@ -150,6 +154,7 @@ cmake_install_executorch_libraries() {
     echo "Installing libexecutorch.a, libextension_module.so, libportable_ops_lib.a"
     rm -rf cmake-out
     retry cmake --preset llm \
+        -DEXECUTORCH_BUILD_TESTS=ON \
         -DBUILD_TESTING=OFF \
         -DCMAKE_INSTALL_PREFIX=cmake-out \
         -DCMAKE_BUILD_TYPE="$CMAKE_BUILD_TYPE" \
@@ -166,6 +171,7 @@ cmake_build_llama_runner() {
     popd
     dir="examples/models/llama"
     retry cmake \
+        -DEXECUTORCH_BUILD_TESTS=ON \
         -DBUILD_TESTING=OFF \
         -DCMAKE_INSTALL_PREFIX=cmake-out \
         -DCMAKE_BUILD_TYPE="$CMAKE_BUILD_TYPE" \

.ci/scripts/test_qnn_static_llama.sh

@@ -9,8 +9,13 @@ set -euxo pipefail
 
 source "$(dirname "${BASH_SOURCE[0]}")/utils.sh"
 
+# Source QNN configuration
+source "$(dirname "${BASH_SOURCE[0]}")/../../backends/qualcomm/scripts/qnn_config.sh"
+# Download QNN_SDK. If already downloaded, export environment path
+source "$(dirname "${BASH_SOURCE[0]}")/../../backends/qualcomm/scripts/install_qnn_sdk.sh"
+install_qnn
+
 export EXECUTORCH_ROOT="$(cd -- "$(dirname -- "${BASH_SOURCE[0]}")/.." && pwd)"
-export QNN_SDK_ROOT=/tmp/qnn/2.28.0.241029
 export LD_LIBRARY_PATH="${QNN_SDK_ROOT}/lib/x86_64-linux-clang"
 export PYTHONPATH=".."
 cp schema/program.fbs exir/_serialize/program.fbs

.github/workflows/android-perf.yml

@@ -292,7 +292,7 @@ jobs:
                       export.output_name="${OUT_ET_MODEL_NAME}.pte"
                     ls -lh "${OUT_ET_MODEL_NAME}.pte"
                 elif [[ ${{ matrix.config }} == "llama3_qnn_htp" ]]; then
-                    export QNN_SDK_ROOT=/tmp/qnn/2.28.0.241029
+                    export QNN_SDK_ROOT=/tmp/qnn/2.37.0.25072
                     export LD_LIBRARY_PATH=$QNN_SDK_ROOT/lib/x86_64-linux-clang/
                     export PYTHONPATH=$(pwd)/..
 
@@ -432,7 +432,7 @@ jobs:
         PYTHON_EXECUTABLE=python bash .ci/scripts/build-qnn-sdk.sh
 
         mkdir -p aar-out
-        PYTHON_EXECUTABLE=python ANDROID_ABIS="arm64-v8a" BUILD_AAR_DIR=aar-out EXECUTORCH_BUILD_QNN=ON QNN_SDK_ROOT=/tmp/qnn/2.28.0.241029 EXECUTORCH_ANDROID_PROFILING=ON bash scripts/build_android_library.sh
+        PYTHON_EXECUTABLE=python ANDROID_ABIS="arm64-v8a" BUILD_AAR_DIR=aar-out EXECUTORCH_BUILD_QNN=ON QNN_SDK_ROOT=/tmp/qnn/2.37.0.25072 EXECUTORCH_ANDROID_PROFILING=ON bash scripts/build_android_library.sh
         mkdir -p extension/benchmark/android/benchmark/app/libs
         cp aar-out/executorch.aar extension/benchmark/android/benchmark/app/libs
         pushd extension/benchmark/android/benchmark

.github/workflows/android-release-artifacts.yml

@@ -16,6 +16,7 @@ on:
         options:
           - "xnnpack"
           - "vulkan+xnnpack"
+          - "qnn"
   schedule:
     - cron: 0 10 * * *
 
@@ -83,17 +84,29 @@ jobs:
 
         echo -n "$SECRET_EXECUTORCH_MAVEN_SIGNING_GPG_KEY_CONTENTS" | base64 -d > /tmp/secring.gpg
 
+        GRADLE_ARGS=""
+
         # Update the version name in build.gradle in case of maven publish
         VERSION="${{ inputs.version }}"
         if [ ! -z "$VERSION" ]; then
-          sed -i "s/\(coordinates(\"org.pytorch\", \"executorch-android\", \"\)\([0-9]\+.[0-9]\+.[0-9]\+\)\(\")\)/\1$VERSION\3/" extension/android/executorch_android/build.gradle
+          GRADLE_ARGS+=" -DexecuTorchVersion=${VERSION}"
         fi
 
         FLAVOR="${{ inputs.flavor }}"
         if [[ "$FLAVOR" == "vulkan+xnnpack" || -z "$FLAVOR" ]]; then
           export EXECUTORCH_BUILD_VULKAN=ON
         fi
 
+        if [[ "$FLAVOR" == "qnn" ]]; then
+          PYTHON_EXECUTABLE=python bash .ci/scripts/setup-qnn-deps.sh
+          PYTHON_EXECUTABLE=python bash .ci/scripts/build-qnn-sdk.sh
+          export EXECUTORCH_BUILD_QNN=ON
+          source backends/qualcomm/scripts/qnn_config.sh
+          export QNN_SDK_ROOT="/tmp/qnn/${QNN_VERSION}"
+          export ANDROID_ABIS=arm64-v8a
+          GRADLE_ARGS+=" -DqnnVersion=2.37.0"
+        fi
+
         # Build AAR Package
         mkdir aar-out
         export BUILD_AAR_DIR=aar-out
@@ -106,7 +119,7 @@ jobs:
         # Publish to maven staging
         UPLOAD_TO_MAVEN="${{ inputs.upload_to_maven }}"
         if [[ "$UPLOAD_TO_MAVEN" == "true" ]]; then
-          (cd extension/android; ANDROID_HOME="${ANDROID_SDK:-/opt/android/sdk}" ./gradlew :executorch_android:publishToMavenCentral)
+          (cd extension/android; ANDROID_HOME="${ANDROID_SDK:-/opt/android/sdk}" ./gradlew  ${GRADLE_ARGS} :executorch_android:publishToMavenCentral)
         fi
 
   upload-release-aar:

.github/workflows/apple-perf.yml

@@ -230,7 +230,7 @@ jobs:
                 model.use_sdpa_with_kv_cache=true \
                 backend.xnnpack.enabled=true \
                 backend.xnnpack.extended_ops=true \
-                base.preq_mode="8da4w_output_8da8w" \
+                base.preq_mode="preq_8da4w_out_8da8w" \
                 base.preq_group_size=32 \
                 export.max_seq_length=2048 \
                 export.max_context_length=2048 \
@@ -256,7 +256,7 @@ jobs:
                 base.params="${DOWNLOADED_PATH}/params.json" \
                 quantization.use_qat=true \
                 base.use_lora=16 \
-                base.preq_mode="8da4w_output_8da8w" \
+                base.preq_mode="preq_8da4w_out_8da8w" \
                 base.preq_group_size=32 \
                 base.preq_embedding_quantize=\'8,0\' \
                 model.use_sdpa_with_kv_cache=true \
@@ -416,7 +416,7 @@ jobs:
       - set-parameters
     secrets: inherit
     with:
-      runner: macos-latest-xlarge
+      runner: macos-14-xlarge
       python-version: '3.11'
       submodules: 'recursive'
       ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}

.github/workflows/apple.yml

@@ -49,7 +49,7 @@ jobs:
     uses: pytorch/test-infra/.github/workflows/macos_job.yml@main
     secrets: inherit
     with:
-      runner: macos-latest-xlarge
+      runner: macos-14-xlarge
       python-version: '3.11'
       submodules: 'recursive'
       ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
@@ -136,7 +136,7 @@ jobs:
     needs: set-version
     uses: pytorch/test-infra/.github/workflows/macos_job.yml@main
     with:
-      runner: macos-latest-xlarge
+      runner: macos-14-xlarge
       python-version: '3.11'
       submodules: 'recursive'
       ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
@@ -156,6 +156,7 @@ jobs:
           "kernels_llm"
           "kernels_optimized"
           "kernels_quantized"
+          "kernels_torchao"
           "threadpool"
         )
 
@@ -275,7 +276,7 @@ jobs:
     uses: pytorch/test-infra/.github/workflows/macos_job.yml@main
     secrets: inherit
     with:
-      runner: macos-latest-xlarge
+      runner: macos-14-xlarge
       python-version: '3.11'
       submodules: 'recursive'
       ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}