Merge branch 'main' into refactor-bitwise-logical-tests

Author: ashutosh-arm

.ci/scripts/unittest-macos.sh

@@ -19,14 +19,14 @@ export TMP_DIR=$(mktemp -d)
 export PATH="${TMP_DIR}:$PATH"
 trap 'rm -rfv ${TMP_DIR}' EXIT
 
-if [[ "$BUILD_TOOL" == "cmake" ]]; then
-    # Setup MacOS dependencies as there is no Docker support on MacOS atm
-    PYTHON_EXECUTABLE=python \
-    EXECUTORCH_BUILD_PYBIND=ON \
-    CMAKE_ARGS="-DEXECUTORCH_BUILD_COREML=ON -DEXECUTORCH_BUILD_MPS=ON -DEXECUTORCH_BUILD_XNNPACK=ON -DEXECUTORCH_BUILD_KERNELS_QUANTIZED=ON" \
-    ${CONDA_RUN} --no-capture-output \
-    .ci/scripts/setup-macos.sh "$@"
+# Setup MacOS dependencies as there is no Docker support on MacOS atm
+PYTHON_EXECUTABLE=python \
+EXECUTORCH_BUILD_PYBIND=ON \
+CMAKE_ARGS="-DEXECUTORCH_BUILD_COREML=ON -DEXECUTORCH_BUILD_MPS=ON -DEXECUTORCH_BUILD_XNNPACK=ON -DEXECUTORCH_BUILD_KERNELS_QUANTIZED=ON" \
+${CONDA_RUN} --no-capture-output \
+.ci/scripts/setup-macos.sh "$@"
 
+if [[ "$BUILD_TOOL" == "cmake" ]]; then
     # Install llama3_2_vision dependencies.
     PYTHON_EXECUTABLE=python \
     ${CONDA_RUN} --no-capture-output \

.github/workflows/_android.yml

@@ -35,7 +35,7 @@ jobs:
         # Build LLM Demo for Android
         export BUILD_AAR_DIR=aar-out
         mkdir -p $BUILD_AAR_DIR
-        bash build/build_android_library.sh ${ARTIFACTS_DIR_NAME}
+        bash scripts/build_android_library.sh ${ARTIFACTS_DIR_NAME}
         bash .ci/scripts/build_android_instrumentation.sh ${ARTIFACTS_DIR_NAME}
 
         mkdir -p examples/demo-apps/android/LlamaDemo/app/libs
@@ -120,7 +120,7 @@ jobs:
         with:
           api-level: ${{ env.API_LEVEL }}
           arch: x86_64
-          script: ./build/run_android_emulator.sh
+          script: ./scripts/run_android_emulator.sh
           # NB: This is to boot the emulator faster following the instructions on
           # https://github.com/ReactiveCircus/android-emulator-runner. The max number
           # of cores we can set is 6, any higher number will be reduced to 6.

.github/workflows/android-perf.yml

@@ -363,7 +363,7 @@ 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 build/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 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

@@ -55,7 +55,7 @@ jobs:
         # Build AAR Package
         mkdir aar-out
         export BUILD_AAR_DIR=aar-out
-        bash build/build_android_library.sh
+        bash scripts/build_android_library.sh
         mkdir -p "${ARTIFACTS_DIR_NAME}"
         cp aar-out/executorch.aar "${ARTIFACTS_DIR_NAME}/executorch.aar"
 

.github/workflows/apple-perf.yml

@@ -387,7 +387,7 @@ jobs:
 
         echo "::group::Build ExecuTorch iOS frameworks"
         PYTHON_EXECUTABLE=python ${CONDA_RUN} --no-capture-output \
-          build/build_apple_frameworks.sh --Release --Debug --coreml --custom --mps --optimized --portable --quantized --xnnpack
+          scripts/build_apple_frameworks.sh --Release --Debug --coreml --custom --mps --optimized --portable --quantized --xnnpack
         echo "::endgroup::"
 
         # NB: Although exported models can be copied to this directory and bundled together with the
@@ -396,7 +396,7 @@ jobs:
         # create the directory here to pass the build
         mkdir -p extension/benchmark/apple/Benchmark/Models
         ${CONDA_RUN} --no-capture-output \
-          build/build_apple_llm_demo.sh ${ARTIFACTS_DIR_NAME}
+          scripts/build_apple_llm_demo.sh ${ARTIFACTS_DIR_NAME}
 
   upload-benchmark-app:
     needs: build-benchmark-app

.github/workflows/apple.yml

@@ -11,9 +11,9 @@ on:
       - .github/workflows/apple.yml
       - install_executorch.sh
       - backends/apple/**
-      - build/build_apple_frameworks.sh
-      - build/build_apple_llm_demo.sh
-      - build/create_frameworks.sh
+      - scripts/build_apple_frameworks.sh
+      - scripts/build_apple_llm_demo.sh
+      - scripts/create_frameworks.sh
       - .ci/scripts/test_ios_ci.sh
       - examples/demo-apps/apple_ios/**
       - extension/apple/**
@@ -172,7 +172,7 @@ jobs:
 
         # Build iOS Frameworks
         PYTHON_EXECUTABLE=python ${CONDA_RUN} --no-capture-output \
-        build/build_apple_frameworks.sh --Release --Debug --coreml --custom --mps --optimized --portable --quantized --xnnpack
+        scripts/build_apple_frameworks.sh --Release --Debug --coreml --custom --mps --optimized --portable --quantized --xnnpack
 
         # Bundle iOS Frameworks
         for FRAMEWORK in "${FRAMEWORKS[@]}"; do (
@@ -313,11 +313,11 @@ jobs:
 
         echo "::group::Build ExecuTorch iOS frameworks"
         PYTHON_EXECUTABLE=python ${CONDA_RUN} --no-capture-output \
-          build/build_apple_frameworks.sh --Release --Debug --coreml --custom --mps --optimized --portable --quantized --xnnpack
+          scripts/build_apple_frameworks.sh --Release --Debug --coreml --custom --mps --optimized --portable --quantized --xnnpack
         echo "::endgroup::"
 
         echo "::group::Build ExecuTorch benchmark app"
         mkdir -p extension/benchmark/apple/Benchmark/Models
         ${CONDA_RUN} --no-capture-output \
-          build/build_apple_llm_demo.sh "${ARTIFACTS_DIR_NAME}"
+          scripts/build_apple_llm_demo.sh "${ARTIFACTS_DIR_NAME}"
         echo "::endgroup::"

.github/workflows/pull.yml

@@ -290,6 +290,36 @@ jobs:
         # run e2e (export, tokenizer and runner)
         PYTHON_EXECUTABLE=python bash .ci/scripts/test_llava.sh
 
+  test-moshi-linux:
+    name: test-moshi-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: 'true'
+      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}"
+
+        PYTHON_EXECUTABLE=python bash .ci/scripts/setup-linux.sh --build-tool "cmake"
+
+        # install Mimi requirements
+        bash examples/models/moshi/mimi/install_requirements.sh
+
+        # reinstall executorch
+        bash ./install_executorch.sh
+
+        # run python unittest
+        python -m unittest examples.models.moshi.mimi.test_mimi
+
   test-quantized-aot-lib-linux:
     name: test-quantized-aot-lib-linux
     uses: pytorch/test-infra/.github/workflows/linux_job_v2.yml@main

.gitignore

@@ -11,6 +11,7 @@ __pycache__/
 arm_test/
 buck-out/
 buck2-bin/
+build/
 cmake-android-out/
 cmake-ios-out/
 cmake-out*

README.md

@@ -67,7 +67,6 @@ We welcome contributions. To get started review the [guidelines](CONTRIBUTING.md
 ```
 executorch
 ├── backends                        #  Backend delegate implementations.
-├── build                           #  Utilities for managing the build system.
 ├── codegen                         #  Tooling to autogenerate bindings between kernels and the runtime.
 ├── configurations
 ├── docs                            #  Static docs tooling.
@@ -81,8 +80,7 @@ executorch
 |   ├── operator                    #  Operator node manipulation utilities.
 |   ├── passes                      #  Built-in compiler passes.
 |   ├── program                     #  Export artifacts.
-|   ├── serde                       #  Graph module
-serialization/deserialization.
+|   ├── serde                       #  Graph module serialization/deserialization.
 |   ├── verification                #  IR verification.
 ├── extension                       #  Extensions built on top of the runtime.
 |   ├── android                     #  ExecuTorch wrappers for Android apps.
@@ -97,8 +95,7 @@ serialization/deserialization.
 |   ├── parallel                    #  C++ threadpool integration.
 |   ├── pybindings                  #  Python API for executorch runtime.
 |   ├── pytree                      #  C++ and Python flattening and unflattening lib for pytrees.
-|   ├── runner_util                 #  Helpers for writing C++ PTE-execution
-tools.
+|   ├── runner_util                 #  Helpers for writing C++ PTE-execution tools.
 |   ├── testing_util                #  Helpers for writing C++ tests.
 |   ├── training                    #  Experimental libraries for on-device training
 ├── kernels                         #  1st party kernel implementations.
@@ -114,9 +111,9 @@ tools.
 |   ├── executor                    #  Model loading, initialization, and execution.
 |   ├── kernel                      #  Kernel registration and management.
 |   ├── platform                    #  Layer between architecture specific code and portable C++.
-├── schema                          #  ExecuTorch PTE file format flatbuffer
-schemas.
-├── scripts                         #  Utility scripts for size management, dependency management, etc.
+├── schema                          #  ExecuTorch PTE file format flatbuffer schemas.
+├── scripts                         #  Utility scripts for building libs, size management, dependency management, etc.
+├── tools                           #  Development tool management.
 ├── devtools                        #  Model profiling, debugging, and introspection.
 ├── shim                            #  Compatibility layer between OSS and Internal builds
 ├── test                            #  Broad scoped end-to-end tests.

backends/apple/coreml/setup.md

@@ -50,7 +50,7 @@ xcode-select --install
 
 ```bash
 cd executorch
-./build/build_apple_frameworks.sh --coreml
+./scripts/build_apple_frameworks.sh --coreml
 ```
 5. Open the project in Xcode, and drag `executorch.xcframework` and `coreml_backend.xcframework` frameworks generated from Step 2 to Frameworks.
 

backends/apple/mps/setup.md

@@ -130,7 +130,7 @@ python3 -m sdk.inspector.inspector_cli --etdump_path etdump.etdp --etrecord_path
 ***Step 1***. Create the ExecuTorch core and MPS delegate frameworks to link on iOS
 ```bash
 cd executorch
-./build/build_apple_frameworks.sh --mps
+./scripts/build_apple_frameworks.sh --mps
 ```
 
 `mps_delegate.xcframework` will be in `cmake-out` folder, along with `executorch.xcframework` and `portable_delegate.xcframework`:

build/__init__.py

@@ -130,7 +130,7 @@ python3 -m sdk.inspector.inspector_cli --etdump_path etdump.etdp --etrecord_path
 ***Step 1***. Create the ExecuTorch core and MPS delegate frameworks to link on iOS
 ```bash
 cd executorch
-./build/build_apple_frameworks.sh --mps
+./scripts/build_apple_frameworks.sh --mps
 ```
 
 `mps_delegate.xcframework` will be in `cmake-out` folder, along with `executorch.xcframework` and `portable_delegate.xcframework`:

docs/source/backends-mps.md

@@ -88,7 +88,7 @@ Now you can compile your app with the ExecuTorch Android library.
 
 ## Building from Source
 
-`build/build_android_library.sh` is a helper script to build the Java library (into .jar), native library (into .so), and the packaged AAR file. It can also build
+`scripts/build_android_library.sh` is a helper script to build the Java library (into .jar), native library (into .so), and the packaged AAR file. It can also build
 demo apps to showcase the AAR is integrated into a user app correctly.
 
 You need Android [SDK](https://developer.android.com/studio) and [NDK](https://developer.android.com/ndk/downloads) to use it.
@@ -100,7 +100,7 @@ You need to set `ANDROID_HOME` to Android SDK home and `ANDROID_NDK` to the corr
 ```
 export ANDROID_HOME=/path/to/sdk
 export ANDROID_NDK=/path/to/ndk
-sh build/build_android_library.sh
+sh scripts/build_android_library.sh
 ```
 
 ### Optional environment variables
@@ -113,7 +113,7 @@ Set environment variable `ANDROID_ABIS` to either `arm64-v8a` or `x86_64` if you
 export ANDROID_ABIS=arm64-v8a
 # or
 # export ANDROID_ABIS=x86_64
-sh build/build_android_library.sh
+sh scripts/build_android_library.sh
 ```
 
 #### EXECUTORCH_CMAKE_BUILD_TYPE

docs/source/using-executorch-android.md

@@ -250,7 +250,7 @@ xcode-select --install
 2. Build the frameworks:
 
 ```bash
-./build/build_apple_frameworks.sh
+./scripts/build_apple_frameworks.sh
 ```
 
 Run the above command with `--help` flag to learn more on how to build additional backends

docs/source/using-executorch-building-from-source.md

@@ -129,13 +129,13 @@ sudo /Applications/CMake.app/Contents/bin/cmake-gui --install
 6. Use the provided script to build .xcframeworks:
 
 ```bash
-./build/build_apple_frameworks.sh --help
+./scripts/build_apple_frameworks.sh --help
 ```
 
 For example, the following command will build the ExecuTorch Runtime along with all available kernels and backends for the Apple platform in both Release and Debug modes:
 
 ```bash
-./build/build_apple_frameworks.sh --Release --Debug --coreml --mps --xnnpack --custom --optimized --portable --quantized
+./scripts/build_apple_frameworks.sh --Release --Debug --coreml --mps --xnnpack --custom --optimized --portable --quantized
 ```
 
 After the build finishes successfully, the resulting frameworks can be found in the `cmake-out` directory.

docs/source/using-executorch-ios.md

@@ -86,7 +86,7 @@ popd
 
 # Build the AAR. It will include XNNPACK backend by default.
 export BUILD_AAR_DIR=$(realpath examples/demo-apps/android/ExecuTorchDemo/app/libs)
-sh build/build_android_library.sh
+sh scripts/build_android_library.sh
 ```
 
 #### Qualcomm Hexagon NPU
@@ -107,7 +107,7 @@ popd
 
 # Build the AAR. It will include XNNPACK backend by default.
 export BUILD_AAR_DIR=$(realpath examples/demo-apps/android/ExecuTorchDemo/app/libs)
-sh build/build_android_library.sh
+sh scripts/build_android_library.sh
 ```
 
 This is very similar to XNNPACK setup, but users now needs to define `QNN_SDK_ROOT` so that

examples/demo-apps/android/ExecuTorchDemo/README.md

@@ -123,7 +123,7 @@ The Mediatek runner (`examples/mediatek/executor_runner/mtk_llama_runner.cpp`) c
 
 Next we need to build and compile the MediaTek backend and MediaTek Llama runner. By setting  `NEURON_BUFFER_ALLOCATOR_LIB`, the script will build the MediaTek backend.
 ```
-sh build/build_android_library.sh
+sh scripts/build_android_library.sh
 ```
 
 **Output**: This will generate an .aar file that is already imported into the expected directory for the Android app. It will live in `examples/demo-apps/android/Llamademo/app/libs`.

examples/demo-apps/android/LlamaDemo/docs/delegates/mediatek_README.md

@@ -12,6 +12,6 @@ export BUILD_AAR_DIR
 
 BASEDIR=$(dirname "$0")
 mkdir -p "$BASEDIR"/app/libs
-bash "$BASEDIR"/../../../../build/build_android_library.sh
+bash "$BASEDIR"/../../../../scripts/build_android_library.sh
 
 cp "$BUILD_AAR_DIR/executorch.aar" "$BASEDIR"/app/libs/executorch.aar

examples/demo-apps/android/LlamaDemo/setup.sh

@@ -149,7 +149,7 @@ curl -LO "https://github.com/facebook/buck2/releases/download/$BUCK2_RELEASE_DAT
 zstd -cdq "$BUCK2_ARCHIVE" > "$BUCK2" && chmod +x "$BUCK2"
 rm "$BUCK2_ARCHIVE"
 
-./build/build_apple_frameworks.sh --buck2="$(realpath $BUCK2)" --coreml --custom --mps --optimized --portable --quantized --xnnpack
+./scripts/build_apple_frameworks.sh --buck2="$(realpath $BUCK2)" --coreml --custom --mps --optimized --portable --quantized --xnnpack
 ```
 
  After the build finishes successfully, the resulting frameworks can be found in the `cmake-out` directory. Copy them to your project and link them against your targets.

examples/demo-apps/apple_ios/LLaMA/docs/delegates/xnnpack_README.md

@@ -8,8 +8,7 @@
 set -x
 
 pip install -U moshi
-pip install bitsandbytes
+pip install bitsandbytes soundfile
 # Run llama2/install requirements for torchao deps
 SCRIPT_DIR=$( cd -- "$( dirname -- "${BASH_SOURCE[0]}" )" &> /dev/null && pwd )
-
-bash "$SCRIPT_DIR"/../llama/install_requirements.sh
+bash "$SCRIPT_DIR"/../../llama/install_requirements.sh

examples/models/moshi/mimi/install_requirements.sh

@@ -2,7 +2,6 @@
 import os
 import random
 import unittest
-import urllib
 
 import numpy as np
 import requests

examples/models/moshi/mimi/test_mimi.py

@@ -15,7 +15,7 @@ Minibench is usedful for giving reference performance data when developers integ
 You will need executorch AAR for Java and JNI dependencies.
 ```
 export ANDROID_NDK=<path_to_android_ndk>
-sh build/build_android_library.sh
+sh scripts/build_android_library.sh
 ```
 and copy the AAR to `app/libs`.
 ```

extension/benchmark/android/benchmark/README.md

@@ -217,7 +217,7 @@ echo "Exporting headers"
 
 mkdir -p "$HEADERS_PATH"
 
-"$SOURCE_ROOT_DIR"/build/print_exported_headers.py --buck2=$(realpath "$BUCK2") --targets \
+"$SOURCE_ROOT_DIR"/scripts/print_exported_headers.py --buck2=$(realpath "$BUCK2") --targets \
   //extension/module: \
   //extension/tensor: \
 | rsync -av --files-from=- "$SOURCE_ROOT_DIR" "$HEADERS_PATH/executorch"
@@ -272,7 +272,7 @@ for mode in "${MODES[@]}"; do
   append_framework_flag "$PORTABLE" "$FRAMEWORK_KERNELS_PORTABLE" "$mode"
   append_framework_flag "$QUANTIZED" "$FRAMEWORK_KERNELS_QUANTIZED" "$mode"
 
-  "$SOURCE_ROOT_DIR"/build/create_frameworks.sh "${FRAMEWORK_FLAGS[@]}"
+  "$SOURCE_ROOT_DIR"/scripts/create_frameworks.sh "${FRAMEWORK_FLAGS[@]}"
 done
 
 echo "Cleaning up"

scripts/__init__.py

@@ -90,7 +90,7 @@ curl https://raw.githubusercontent.com/pytorch/hub/master/imagenet_classes.txt \
 
 say "Building Frameworks"
 
-./build/build_apple_frameworks.sh --coreml --custom --mps --optimized --portable --quantized --xnnpack
+./scripts/build_apple_frameworks.sh --coreml --custom --mps --optimized --portable --quantized --xnnpack
 mv cmake-out "$APP_PATH/Frameworks"
 
 say "Creating Simulator"