Llama nncf test

.buckconfig

@@ -39,6 +39,7 @@
 
 [buck2]
 restarter=true
+file_watcher=notify
 
 [oss]
 folly_cxx_tests = False

.ci/docker/ci_commit_pins/buck2.txt

@@ -1 +1 @@
-2024-12-16
+2025-05-06

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

@@ -11,17 +11,12 @@ set -o xtrace
 
 build_qnn_backend() {
   echo "Start building qnn backend."
-  export ANDROID_NDK_ROOT=/opt/ndk
-  export QNN_SDK_ROOT=/tmp/qnn/2.28.0.241029
+  export ANDROID_NDK_ROOT=${ANDROID_NDK_ROOT:-/opt/ndk}
+  export QNN_SDK_ROOT=${QNN_SDK_ROOT:-/tmp/qnn/2.28.0.241029}
   export EXECUTORCH_ROOT="$(cd -- "$(dirname -- "${BASH_SOURCE[0]}")/../.." && pwd)"
 
-  # Workaround to avoid issues around missing flatccrt library (depending on the
-  # number of jobs used), see issue #7300:
-  # Build twice (second time with `--no_clean`) to make sure libflatccrt.a is
-  # available.
-  # TODO: Remove this workaround once the underlying issue is fixed.
-  bash backends/qualcomm/scripts/build.sh --skip_aarch64 --job_number 2 --release || \
-  bash backends/qualcomm/scripts/build.sh --skip_aarch64 --job_number 2 --release --no_clean
+  parallelism=$(( $(nproc) - 1 ))
+  bash backends/qualcomm/scripts/build.sh --skip_aarch64 --job_number ${parallelism} --release
 }
 
 set_up_aot() {
@@ -39,8 +34,7 @@ set_up_aot() {
       -DEXECUTORCH_BUILD_EXTENSION_MODULE=ON \
       -DEXECUTORCH_BUILD_EXTENSION_TENSOR=ON \
       -DEXECUTORCH_ENABLE_EVENT_TRACER=ON \
-      -DPYTHON_EXECUTABLE=python3 \
-      -DEXECUTORCH_SEPARATE_FLATCC_HOST_PROJECT=OFF
+      -DPYTHON_EXECUTABLE=python3
   cmake --build $PWD --target "PyQnnManagerAdaptor" "PyQnnWrapperAdaptor" -j$(nproc)
   # install Python APIs to correct import path
   # The filename might vary depending on your Python and host version.

.ci/scripts/test_llava.sh

@@ -93,8 +93,7 @@ cmake_build_llava_runner_for_android() {
         -DCMAKE_TOOLCHAIN_FILE=$ANDROID_NDK/build/cmake/android.toolchain.cmake \
         -DANDROID_ABI=arm64-v8a                                                 \
         ${LLAVA_COMMON_CMAKE_ARGS}                                              \
-        -DCMAKE_PREFIX_PATH="$python_lib"                  \
-        -DLLAVA_RUNNER_NO_TORCH_DUMMY_IMAGE=ON                                  \
+        -DCMAKE_PREFIX_PATH="$python_lib"                                       \
         -B${BUILD_DIR}/${dir}                                                   \
         ${dir}
 
@@ -107,11 +106,10 @@ export_llava() {
     $PYTHON_EXECUTABLE -m executorch.examples.models.llava.export_llava --pte-name llava.pte --with-artifacts
 }
 
-# Download a new image with different size, to test if the model can handle different image sizes
-prepare_image_tensor() {
+# Download a new image
+download_image() {
     echo "Downloading image"
     curl -o basketball.jpg https://upload.wikimedia.org/wikipedia/commons/7/73/Chicago_Bulls_and_New_Jersey_Nets%2C_March_28%2C_1991.jpg
-    $PYTHON_EXECUTABLE -m executorch.examples.models.llava.image_util --image-path basketball.jpg --output-path image.pt
 }
 
 run_and_verify() {
@@ -121,20 +119,18 @@ run_and_verify() {
         echo "Export failed. Abort"
         exit 1
     fi
-    if [[ ! -f "image.pt" ]]; then
-        echo "image.pt is missing."
+    if [[ ! -f "basketball.jpg" ]]; then
+        echo "basketball.jpg is missing."
         exit 1
     fi
     if [[ ! -f "tokenizer.bin" ]]; then
         echo "tokenizer.bin is missing."
         exit 1
     fi
 
-
-
     RUNTIME_ARGS="--model_path=llava.pte    \
         --tokenizer_path=tokenizer.bin      \
-        --image_path=image.pt               \
+        --image_path=basketball.jpg         \
         --prompt=ASSISTANT:                 \
         --temperature=0                     \
         --seq_len=650"
@@ -149,13 +145,8 @@ run_and_verify() {
 
     # verify result.txt
     RESULT=$(cat result.txt)
-    # set the expected prefix to be the same as prompt because there's a bug in sdpa_with_kv_cache that causes <unk> tokens.
-    if [[ "$(uname)" == "Darwin" ]]; then
-        EXPECTED_PREFIX="ASSISTANT: image captures a basketball game in progress, with several players on the court. One of the players is dribbling the ball, while the others are in various"
-    else
-        # set the expected prefix to be the same as prompt because there's a bug in sdpa_with_kv_cache that causes <unk> tokens.
-        EXPECTED_PREFIX="ASSISTANT: image"
-    fi
+    EXPECTED_PREFIX="ASSISTANT: image captures a basketball game in progress, with several players on the court. "
+
     if [[ "${RESULT}" == *"${EXPECTED_PREFIX}"* ]]; then
         echo "Expected result prefix: ${EXPECTED_PREFIX}"
         echo "Actual result: ${RESULT}"
@@ -184,5 +175,5 @@ fi
 export_llava
 
 # Step3. Run
-prepare_image_tensor
+download_image
 run_and_verify

.ci/scripts/test_model.sh

@@ -87,10 +87,6 @@ test_model() {
     bash examples/models/llava/install_requirements.sh
     STRICT="--no-strict"
   fi
-  if [[ "$MODEL_NAME" == "llama3_2_vision_encoder" || "$MODEL_NAME" == "llama3_2_text_decoder" ]]; then
-    # Install requirements for llama vision.
-    bash examples/models/llama3_2_vision/install_requirements.sh
-  fi
   if [[ "${MODEL_NAME}" == "qwen2_5" ]]; then
       # Install requirements for export_llama
       bash examples/models/llama/install_requirements.sh
@@ -205,7 +201,7 @@ test_model_with_qnn() {
   # TODO(guangyang): Make QNN chipset matches the target device
   QNN_CHIPSET=SM8450
 
-  "${PYTHON_EXECUTABLE}" -m examples.qualcomm.scripts.${EXPORT_SCRIPT} -b ${CMAKE_OUTPUT_DIR} -m ${QNN_CHIPSET} --compile_only $EXTRA_FLAGS
+  "${PYTHON_EXECUTABLE}" -m examples.qualcomm.scripts.${EXPORT_SCRIPT} -b ${CMAKE_OUTPUT_DIR} -m ${QNN_CHIPSET} --ci --compile_only $EXTRA_FLAGS
   EXPORTED_MODEL=$(find "./${EXPORT_SCRIPT}" -type f -name "${MODEL_NAME}*.pte" -print -quit)
 }
 

.ci/scripts/unittest-linux.sh

@@ -24,9 +24,6 @@ if [[ "$BUILD_TOOL" == "cmake" ]]; then
     CMAKE_ARGS="-DEXECUTORCH_BUILD_PYBIND=ON -DEXECUTORCH_BUILD_XNNPACK=ON -DEXECUTORCH_BUILD_KERNELS_QUANTIZED=ON" \
     .ci/scripts/setup-linux.sh "$@"
 
-    # Install llama3_2_vision dependencies.
-    PYTHON_EXECUTABLE=python ./examples/models/llama3_2_vision/install_requirements.sh
-
     .ci/scripts/unittest-linux-cmake.sh
 elif [[ "$BUILD_TOOL" == "buck2" ]]; then
     # Removing this breaks sccache in the Buck build, apparently

.ci/scripts/unittest-macos.sh

@@ -29,7 +29,6 @@ if [[ "$BUILD_TOOL" == "cmake" ]]; then
     # Install llama3_2_vision dependencies.
     PYTHON_EXECUTABLE=python \
     ${CONDA_RUN} --no-capture-output \
-    ./examples/models/llama3_2_vision/install_requirements.sh
 
     .ci/scripts/unittest-macos-cmake.sh
 elif [[ "$BUILD_TOOL" == "buck2" ]]; then

.github/scripts/label_utils.py

@@ -22,12 +22,12 @@
 
 LABEL_ERR_MSG_TITLE = "This PR needs a `release notes:` label"
 LABEL_ERR_MSG = f"""# {LABEL_ERR_MSG_TITLE}
-If your changes are user facing and intended to be a part of release notes, please use a label starting with `release notes:`.
+If your change should be included in the release notes (i.e. would users of this library care about this change?), please use a label starting with `release notes:`.
 
-If not, please add the `topic: not user facing` label.
+If not, please add the `release notes: none` label.
 
 To add a label, you can comment to pytorchbot, for example
-`@pytorchbot label "topic: not user facing"`
+`@pytorchbot label "release notes: none"`
 
 For more information, see
 https://github.com/pytorch/pytorch/wiki/PyTorch-AutoLabel-Bot#why-categorize-for-release-notes-and-how-does-it-work.
@@ -115,7 +115,7 @@ def has_required_labels(pr: "GitHubPR") -> bool:
     pr_labels = pr.get_labels()
     # Check if PR is not user facing
     is_not_user_facing_pr = any(
-        label.strip() == "topic: not user facing" for label in pr_labels
+        label.strip() == "release notes: none" for label in pr_labels
     )
     return is_not_user_facing_pr or any(
         label.strip() in get_release_notes_labels(pr.org, pr.project)

.github/workflows/_link_check.yml

@@ -18,11 +18,11 @@ jobs:
       timeout: 120
       script: |
         ./scripts/lint_urls.sh $(
-          { [ "${{ github.event_name }}" = "pull_request" ] \
-              && git diff --name-only "${{ github.event.pull_request.base.sha }}...${{ github.event.pull_request.head.sha }}"; } \
-          || \
-          { [ "${{ github.event_name }}" = "push" ] \
-              && git diff --name-only "${{ github.event.before }}...${{ github.sha }}"; }
+          if [ "${{ github.event_name }}" = "pull_request" ]; then
+            echo "${{ github.event.pull_request.base.sha }}" "${{ github.event.pull_request.head.sha }}"
+          else
+            echo "${{ github.event.before }}" "${{ github.sha }}"
+          fi
         ) || {
           echo
           echo "URL lint failed."
@@ -43,11 +43,11 @@ jobs:
       timeout: 60
       script: |
         ./scripts/lint_xrefs.sh $(
-          { [ "${{ github.event_name }}" = "pull_request" ] \
-              && git diff --name-only "${{ github.event.pull_request.base.sha }}...${{ github.event.pull_request.head.sha }}"; } \
-          || \
-          { [ "${{ github.event_name }}" = "push" ] \
-              && git diff --name-only "${{ github.event.before }}...${{ github.sha }}"; }
+          if [ "${{ github.event_name }}" = "pull_request" ]; then
+            echo "${{ github.event.pull_request.base.sha }}" "${{ github.event.pull_request.head.sha }}"
+          else
+            echo "${{ github.event.before }}" "${{ github.sha }}"
+          fi
         ) || {
           echo
           echo "Xref lint failed."

.github/workflows/android-perf.yml

@@ -367,7 +367,7 @@ 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 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/2.28.0.241029 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/build-presets.yml

@@ -6,8 +6,63 @@ on:
     branches:
       - main
       - release/*
+    paths:
+      - .github/workflows/build-presets.yml
   workflow_dispatch:
 
 concurrency:
   group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.sha }}-${{ github.event_name == 'workflow_dispatch' }}-${{ github.event_name == 'schedule' }}
   cancel-in-progress: true
+
+jobs:
+  apple:
+    uses: pytorch/test-infra/.github/workflows/macos_job.yml@main
+    strategy:
+      fail-fast: false
+      matrix:
+        preset: [macos-arm64, pybind]
+    with:
+      job-name: build
+      ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
+      runner: macos-latest-xlarge
+      python-version: 3.12
+      submodules: recursive
+      timeout: 90
+      script: |
+        set -eux
+        ${CONDA_RUN} ./install_requirements.sh > /dev/null
+        ${CONDA_RUN} cmake --preset ${{ matrix.preset }}
+        ${CONDA_RUN} cmake --build cmake-out --parallel
+
+  linux:
+    uses: pytorch/test-infra/.github/workflows/linux_job_v2.yml@main
+    strategy:
+      fail-fast: false
+      matrix:
+        preset: [pybind]
+        runner: [linux.2xlarge, linux.arm64.2xlarge]
+        docker-image: [executorch-ubuntu-22.04-clang12, executorch-ubuntu-22.04-gcc11-aarch64]
+        # Excluding specific runner + docker image combinations that don't make sense:
+        #   - Excluding the ARM64 gcc image on the x86 runner (linux.2xlarge)
+        #   - Excluding the x86 clang image on the ARM64 runner (linux.arm64.2xlarge)
+        exclude:
+          - runner: linux.2xlarge
+            docker-image: executorch-ubuntu-22.04-gcc11-aarch64
+          - runner: linux.arm64.2xlarge
+            docker-image: executorch-ubuntu-22.04-clang12
+    with:
+      job-name: build
+      ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
+      runner: ${{ matrix.runner }}
+      docker-image: ${{ matrix.docker-image }}
+      submodules: recursive
+      timeout: 90
+      script: |
+        set -eux
+        # The generic Linux job chooses to use base env, not the one setup by the image
+        CONDA_ENV=$(conda env list --json | jq -r ".envs | .[-1]")
+        conda activate "${CONDA_ENV}"
+
+        ./install_requirements.sh > /dev/null
+        cmake --preset ${{ matrix.preset }}
+        cmake --build cmake-out --parallel

.github/workflows/pull.yml

@@ -343,34 +343,6 @@ jobs:
         PYTHON_EXECUTABLE=python bash .ci/scripts/setup-linux.sh --build-tool "${BUILD_TOOL}"
         PYTHON_EXECUTABLE=python bash examples/xnnpack/quantization/test_quantize.sh "${BUILD_TOOL}" mv2
 
-  test-pybind-build-linux:
-    name: test-pybind-build-linux
-    uses: pytorch/test-infra/.github/workflows/linux_job_v2.yml@main
-    permissions:
-      id-token: write
-      contents: read
-    strategy:
-      fail-fast: false
-    with:
-      runner: linux.2xlarge
-      docker-image: executorch-ubuntu-22.04-clang12
-      submodules: 'recursive'
-      ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
-      timeout: 90
-      script: |
-        # The generic Linux job chooses to use base env, not the one setup by the image
-        CONDA_ENV=$(conda env list --json | jq -r ".envs | .[-1]")
-        conda activate "${CONDA_ENV}"
-
-        # build module for executorch.extension.pybindings.portable_lib
-        BUILD_TOOL="cmake"
-        PYTHON_EXECUTABLE=python \
-        CMAKE_ARGS="-DEXECUTORCH_BUILD_PYBIND=ON -DEXECUTORCH_BUILD_XNNPACK=ON" \
-        bash .ci/scripts/setup-linux.sh --build-tool "${BUILD_TOOL}"
-
-        # see if we can import the module successfully
-        python -c "from executorch.extension.pybindings import portable_lib; print('success!')"
-
   test-binary-size-linux-gcc:
     name: test-binary-size-linux-gcc
     uses: pytorch/test-infra/.github/workflows/linux_job_v2.yml@main
@@ -434,9 +406,7 @@ jobs:
         output=$(ls -la cmake-out/test/size_test)
         arr=($output)
         size=${arr[4]}
-        # threshold=48120 on devserver with gcc11.4
-        # todo(lfq): update once binary size is below 50kb.
-        threshold="47552"
+        threshold="47560"
         if [[ "$size" -le "$threshold" ]]; then
           echo "Success $size <= $threshold"
         else