Update on "Dtype selective build: disable when not in xplat and add fail msg"

^

Differential Revision: [D75256223](https://our.internmc.facebook.com/intern/diff/D75256223/)

[ghstack-poisoned]

Author: lucylq

.ci/docker/ci_commit_pins/pytorch.txt

@@ -1 +1 @@
-59d5cf083b4f860dea76fe8936076177f9367f10
+01f1cc44cbbfdf6307aa01b803a4ee22f9ade946

.ci/scripts/build_llama_android.sh

@@ -22,18 +22,12 @@ install_executorch_and_backend_lib() {
   ANDROID_NDK=/opt/ndk
   BUCK2=buck2
   ANDROID_ABI=arm64-v8a
-  cmake -DBUCK2="${BUCK2}" \
+  cmake --preset llm \
+    -DBUCK2="${BUCK2}" \
     -DCMAKE_TOOLCHAIN_FILE="${ANDROID_NDK}/build/cmake/android.toolchain.cmake" \
     -DANDROID_ABI="${ANDROID_ABI}" \
     -DCMAKE_INSTALL_PREFIX=cmake-android-out \
     -DCMAKE_BUILD_TYPE=Release \
-    -DEXECUTORCH_BUILD_EXTENSION_DATA_LOADER=ON \
-    -DEXECUTORCH_BUILD_EXTENSION_MODULE=ON \
-    -DEXECUTORCH_BUILD_EXTENSION_TENSOR=ON \
-    -DEXECUTORCH_BUILD_XNNPACK=ON \
-    -DEXECUTORCH_BUILD_KERNELS_OPTIMIZED=ON \
-    -DEXECUTORCH_BUILD_KERNELS_QUANTIZED=ON \
-    -DEXECUTORCH_BUILD_KERNELS_CUSTOM=ON \
     -DXNNPACK_ENABLE_ARM_BF16=OFF \
     -Bcmake-android-out .
 
@@ -51,11 +45,7 @@ build_llama_runner() {
     -DCMAKE_TOOLCHAIN_FILE="$ANDROID_NDK"/build/cmake/android.toolchain.cmake  \
     -DANDROID_ABI="${ANDROID_ABI}" \
     -DCMAKE_INSTALL_PREFIX=cmake-android-out \
-    -DCMAKE_BUILD_TYPE=Release -DPYTHON_EXECUTABLE=python \
-    -DEXECUTORCH_BUILD_XNNPACK=ON \
-    -DEXECUTORCH_BUILD_KERNELS_OPTIMIZED=ON \
-    -DEXECUTORCH_BUILD_KERNELS_QUANTIZED=ON \
-    -DEXECUTORCH_BUILD_KERNELS_CUSTOM=ON \
+    -DCMAKE_BUILD_TYPE=Release \
     -Bcmake-android-out/examples/models/llama examples/models/llama
 
     cmake --build cmake-android-out/examples/models/llama -j4 --config Release

.ci/scripts/test_llama.sh

@@ -152,21 +152,11 @@ which "${PYTHON_EXECUTABLE}"
 cmake_install_executorch_libraries() {
     echo "Installing libexecutorch.a, libextension_module.so, libportable_ops_lib.a"
     rm -rf cmake-out
-    retry cmake \
+    retry cmake --preset llm \
         -DCMAKE_INSTALL_PREFIX=cmake-out \
         -DCMAKE_BUILD_TYPE="$CMAKE_BUILD_TYPE" \
-        -DEXECUTORCH_BUILD_EXTENSION_DATA_LOADER=ON \
-        -DEXECUTORCH_BUILD_EXTENSION_MODULE=ON \
-        -DEXECUTORCH_BUILD_EXTENSION_TENSOR=ON \
-        -DEXECUTORCH_BUILD_KERNELS_CUSTOM="$CUSTOM" \
-        -DEXECUTORCH_BUILD_KERNELS_OPTIMIZED=ON \
-        -DEXECUTORCH_BUILD_KERNELS_QUANTIZED=ON \
-        -DEXECUTORCH_BUILD_XNNPACK="$XNNPACK" \
-        -DEXECUTORCH_BUILD_MPS="$MPS" \
-        -DEXECUTORCH_BUILD_COREML="$COREML" \
         -DEXECUTORCH_BUILD_QNN="$QNN" \
         -DQNN_SDK_ROOT="$QNN_SDK_ROOT" \
-        -DPYTHON_EXECUTABLE="$PYTHON_EXECUTABLE" \
         -Bcmake-out .
     cmake --build cmake-out -j9 --target install --config "$CMAKE_BUILD_TYPE"
 }
@@ -181,10 +171,6 @@ cmake_build_llama_runner() {
     retry cmake \
         -DCMAKE_INSTALL_PREFIX=cmake-out \
         -DCMAKE_BUILD_TYPE="$CMAKE_BUILD_TYPE" \
-        -DEXECUTORCH_BUILD_KERNELS_CUSTOM="$CUSTOM" \
-        -DEXECUTORCH_BUILD_KERNELS_OPTIMIZED=ON \
-        -DEXECUTORCH_BUILD_XNNPACK="$XNNPACK" \
-        -DPYTHON_EXECUTABLE="$PYTHON_EXECUTABLE" \
         -Bcmake-out/${dir} \
         ${dir}
     cmake --build cmake-out/${dir} -j9 --config "$CMAKE_BUILD_TYPE"

.ci/scripts/wheel/test_linux.py

@@ -14,6 +14,10 @@
             test_base.ModelTest(
                 model=Model.Mv3,
                 backend=Backend.XnnpackQuantizationDelegation,
-            )
+            ),
+            test_base.ModelTest(
+                model=Model.Mv3,
+                backend=Backend.CoreMlExportOnly,
+            ),
         ]
     )

.ci/scripts/wheel/test_macos.py

@@ -17,7 +17,7 @@
             ),
             test_base.ModelTest(
                 model=Model.Mv3,
-                backend=Backend.CoreMlTest,
+                backend=Backend.CoreMlExportAndTest,
             ),
         ]
     )

.github/workflows/apple-perf.yml

@@ -386,8 +386,7 @@ jobs:
         echo "::endgroup::"
 
         echo "::group::Build ExecuTorch iOS frameworks"
-        PYTHON_EXECUTABLE=python ${CONDA_RUN} --no-capture-output \
-          scripts/build_apple_frameworks.sh --Release --Debug --coreml --custom --mps --optimized --portable --quantized --xnnpack
+        PYTHON_EXECUTABLE=python ${CONDA_RUN} --no-capture-output scripts/build_apple_frameworks.sh
         echo "::endgroup::"
 
         # NB: Although exported models can be copied to this directory and bundled together with the

.github/workflows/apple.yml

@@ -156,6 +156,7 @@ jobs:
           "kernels_optimized"
           "kernels_portable"
           "kernels_quantized"
+          "threadpool"
         )
 
         .ci/scripts/setup-conda.sh
@@ -173,8 +174,7 @@ jobs:
         backends/apple/mps/install_requirements.sh
 
         # Build iOS Frameworks
-        PYTHON_EXECUTABLE=python ${CONDA_RUN} --no-capture-output \
-        scripts/build_apple_frameworks.sh --Release --Debug --coreml --custom --mps --optimized --portable --quantized --xnnpack
+        PYTHON_EXECUTABLE=python ${CONDA_RUN} --no-capture-output scripts/build_apple_frameworks.sh
 
         # Bundle iOS Frameworks
         for FRAMEWORK in "${FRAMEWORKS[@]}"; do (
@@ -314,8 +314,7 @@ jobs:
         echo "::endgroup::"
 
         echo "::group::Build ExecuTorch iOS frameworks"
-        PYTHON_EXECUTABLE=python ${CONDA_RUN} --no-capture-output \
-          scripts/build_apple_frameworks.sh --Release --Debug --coreml --custom --mps --optimized --portable --quantized --xnnpack
+        PYTHON_EXECUTABLE=python ${CONDA_RUN} --no-capture-output scripts/build_apple_frameworks.sh
         echo "::endgroup::"
 
         echo "::group::Build ExecuTorch benchmark app"

.github/workflows/build-presets.yml

@@ -20,7 +20,7 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        preset: [macos-arm64, pybind]
+        preset: [macos, ios, ios-simulator, pybind, llm]
     with:
       job-name: build
       ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
@@ -32,14 +32,14 @@ jobs:
         set -eux
         ${CONDA_RUN} ./install_requirements.sh > /dev/null
         ${CONDA_RUN} cmake --preset ${{ matrix.preset }}
-        ${CONDA_RUN} cmake --build cmake-out --parallel
+        ${CONDA_RUN} cmake --build cmake-out -j$(( $(sysctl -n hw.ncpu) - 1 ))
 
   linux:
     uses: pytorch/test-infra/.github/workflows/linux_job_v2.yml@main
     strategy:
       fail-fast: false
       matrix:
-        preset: [pybind]
+        preset: [linux, pybind, llm]
         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:
@@ -65,4 +65,31 @@ jobs:
 
         ./install_requirements.sh > /dev/null
         cmake --preset ${{ matrix.preset }}
-        cmake --build cmake-out --parallel
+        cmake --build cmake-out -j$(( $(nproc) - 1 ))
+
+  windows:
+    uses: pytorch/test-infra/.github/workflows/windows_job.yml@main
+    strategy:
+      fail-fast: false
+      matrix:
+        preset: [pybind]
+    with:
+      job-name: build
+      ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
+      submodules: recursive
+      timeout: 90
+      script: |
+        set -eux
+        conda init powershell
+        powershell -Command "& {
+          \$ErrorActionPreference = 'Stop'
+          Set-PSDebug -Trace 1
+
+          conda create --yes --quiet -n et python=3.12
+          conda activate et
+
+          python install_requirements.py
+          cmake --preset ${{ matrix.preset }}
+          \$numCores = [System.Environment]::GetEnvironmentVariable('NUMBER_OF_PROCESSORS') - 1
+          cmake --build cmake-out -j \$numCores
+        }"

.gitmodules

@@ -67,3 +67,6 @@
 [submodule "shim"]
 	path = shim
 	url = https://github.com/facebook/buck2-shims-meta
+[submodule "third-party/json"]
+	path = third-party/json
+	url = https://github.com/nlohmann/json.git

.lintrunner.toml

@@ -386,20 +386,14 @@ exclude_patterns = [
     "third-party/**",
     # TODO: remove exceptions as we migrate
     # backends
-    "backends/arm/quantizer/**",
-    "backends/arm/test/ops/**",
     "backends/vulkan/quantizer/**",
     "backends/vulkan/test/**",
-    "backends/cadence/aot/quantizer/**",
-    "backends/qualcomm/quantizer/**",
-    "examples/qualcomm/**",
     "backends/xnnpack/quantizer/**",
     "backends/xnnpack/test/**",
     "exir/tests/test_passes.py",
     "extension/llm/export/builder.py",
     "extension/llm/export/quantizer_lib.py",
     "exir/tests/test_memory_planning.py",
-    "backends/transforms/duplicate_dynamic_quant_chain.py",
     "exir/backend/test/demos/test_xnnpack_qnnpack.py",
 ]