Update

[ghstack-poisoned]

Author: swolchok

.ci/docker/ci_commit_pins/pytorch.txt

@@ -1 +1 @@
-7cda4017ddda554752e89069ae205be5e8388f59
+90f1e7bed15ca5e48c61c5b6dc5ad4810524f82f

.ci/scripts/test_phi_3_mini.sh

@@ -22,31 +22,14 @@ NPROC=8
 if hash nproc &> /dev/null; then NPROC=$(nproc); fi
 
 cmake_install_executorch_libraries() {
-  cmake -DPYTHON_EXECUTABLE=python \
-      -DCMAKE_INSTALL_PREFIX=${BUILD_DIR} \
-      -DEXECUTORCH_ENABLE_LOGGING=1 \
-      -DCMAKE_BUILD_TYPE=${BUILD_TYPE} \
-      -DEXECUTORCH_BUILD_EXTENSION_DATA_LOADER=ON \
-      -DEXECUTORCH_BUILD_EXTENSION_FLAT_TENSOR=ON \
-      -DEXECUTORCH_BUILD_EXTENSION_MODULE=ON \
-      -DEXECUTORCH_BUILD_EXTENSION_TENSOR=ON \
-      -DEXECUTORCH_BUILD_XNNPACK=ON \
-      -DEXECUTORCH_BUILD_KERNELS_QUANTIZED=ON \
-      -DEXECUTORCH_BUILD_KERNELS_OPTIMIZED=ON \
-      -DEXECUTORCH_BUILD_KERNELS_LLM=ON \
-      -B${BUILD_DIR} .
-
-  cmake --build ${BUILD_DIR} -j${NPROC} --target install --config ${BUILD_TYPE}
+  rm -rf cmake-out
+  cmake --preset llm -DCMAKE_INSTALL_PREFIX=cmake-out -DCMAKE_BUILD_TYPE=${BUILD_TYPE}
+  cmake --build cmake-out -j16 --target install --config ${BUILD_TYPE}
 }
 
 cmake_build_phi_3_mini() {
-  cmake -DPYTHON_EXECUTABLE=$PYTHON_EXECUTABLE \
-      -DCMAKE_INSTALL_PREFIX=${BUILD_DIR} \
+  cmake -DCMAKE_PREFIX_PATH=${BUILD_DIR} \
       -DCMAKE_BUILD_TYPE=${BUILD_TYPE} \
-      -DEXECUTORCH_BUILD_KERNELS_LLM=ON \
-      -DEXECUTORCH_BUILD_KERNELS_OPTIMIZED=ON \
-      -DEXECUTORCH_BUILD_XNNPACK=ON \
-      -DEXECUTORCH_BUILD_KERNELS_QUANTIZED=ON \
       -B${BUILD_DIR}/${MODEL_DIR} \
       ${MODEL_DIR}
 
@@ -81,7 +64,7 @@ run_and_verify() {
     ${BUILD_DIR}/${MODEL_DIR}/phi_3_mini_runner \
     --model_path=phi-3-mini.pte \
     --tokenizer_path=tokenizer.bin \
-    --seq_len=128 \
+    --seq_len=60 \
     --temperature=0 \
     --prompt="<|system|>
 You are a helpful assistant.<|end|>

.ci/scripts/unittest-buck2.sh

@@ -10,7 +10,7 @@ set -eux
 # TODO: can't query cadence & vulkan backends
 # TODO: can't query //kernels/prim_ops because of non-buckified stuff in OSS.
 buck2 query "//backends/apple/... + //backends/example/... + \
-//backends/mediatek/... + //backends/test/... + //backends/transforms/... + \
+//backends/mediatek/... + //backends/transforms/... + \
 //backends/xnnpack/... + //configurations/... + //kernels/aten/... + \
 //kernels/optimized/... + //kernels/portable/... + //kernels/quantized/... + \
 //kernels/test/... + //runtime/... + //schema/... + //test/... + //util/..."

.github/workflows/build-wheels-linux.yml

@@ -9,6 +9,8 @@ on:
       - examples/**/*
       - pyproject.toml
       - setup.py
+    tags:
+      - ciflow/binaries/*
   push:
     branches:
       - nightly

.github/workflows/build-wheels-macos.yml

@@ -9,6 +9,8 @@ on:
       - examples/**/*
       - pyproject.toml
       - setup.py
+    tags:
+      - ciflow/binaries/*
   push:
     branches:
       - nightly

.github/workflows/pull.yml

@@ -603,7 +603,7 @@ jobs:
         bash examples/models/phi-3-mini/install_requirements.sh
 
         # run e2e (export, tokenizer and runner)
-        PYTHON_EXECUTABLE=python bash .ci/scripts/test_phi_3_mini.sh
+        PYTHON_EXECUTABLE=python bash .ci/scripts/test_phi_3_mini.sh Release
 
   test-eval_llama-wikitext-linux:
     name: test-eval_llama-wikitext-linux

.github/workflows/trunk.yml

@@ -197,7 +197,7 @@ jobs:
       docker-image: executorch-ubuntu-22.04-arm-sdk
       submodules: 'recursive'
       ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
-      timeout: 90
+      timeout: 120
       script: |
         # The generic Linux job chooses to use base env, not the one setup by the image
         CONDA_ENV=$(conda env list --json | jq -r ".envs | .[-1]")
@@ -800,3 +800,6 @@ jobs:
 
         # Run pytest
         PYTHON_EXECUTABLE=python bash backends/nxp/run_unittests.sh
+        
+        # Run aot example: 
+        PYTHON_EXECUTABLE=python bash examples/nxp/run_aot_example.sh

CMakeLists.txt

@@ -182,39 +182,49 @@ endif()
 
 if(EXECUTORCH_BUILD_CPUINFO)
   # --- cpuinfo
-  set(ORIGINAL_CMAKE_POSITION_INDEPENDENT_CODE_FLAG
-      ${CMAKE_POSITION_INDEPENDENT_CODE}
-  )
-  set(CMAKE_POSITION_INDEPENDENT_CODE ON)
-  set(CPUINFO_SOURCE_DIR "backends/xnnpack/third-party/cpuinfo")
-  set(CPUINFO_BUILD_TOOLS
-      OFF
-      CACHE BOOL ""
-  )
-  set(CPUINFO_BUILD_UNIT_TESTS
-      OFF
-      CACHE BOOL ""
+  set(CPUINFO_SOURCE_DIR
+      "${CMAKE_CURRENT_LIST_DIR}/backends/xnnpack/third-party/cpuinfo"
   )
-  set(CPUINFO_BUILD_MOCK_TESTS
-      OFF
-      CACHE BOOL ""
+  set(CPUINFO_BINARY_DIR "${CMAKE_CURRENT_BINARY_DIR}/cpuinfo")
+  set(CPUINFO_INSTALL_DIR "${CPUINFO_BINARY_DIR}/install")
+  set(CPUINFO_LIBRARY
+      "${CPUINFO_INSTALL_DIR}/${CMAKE_INSTALL_LIBDIR}/libcpuinfo.a"
   )
-  set(CPUINFO_BUILD_BENCHMARKS
-      OFF
-      CACHE BOOL ""
-  )
-  set(CPUINFO_LIBRARY_TYPE
-      "static"
-      CACHE STRING ""
-  )
-  set(CPUINFO_LOG_LEVEL
-      "error"
-      CACHE STRING ""
+
+  get_extra_cmake_args_for_external_project(CPUINFO_EXTRA_CMAKE_ARGS)
+  ExternalProject_Add(
+    cpuinfoExternalProject
+    SOURCE_DIR "${CPUINFO_SOURCE_DIR}"
+    BINARY_DIR "${CPUINFO_BINARY_DIR}"
+    INSTALL_DIR "${CPUINFO_INSTALL_DIR}"
+    BUILD_BYPRODUCTS "${CPUINFO_LIBRARY}"
+    CMAKE_ARGS "${CPUINFO_EXTRA_CMAKE_ARGS}"
+               -D
+               CMAKE_POSITION_INDEPENDENT_CODE=ON
+               -D
+               CPUINFO_BUILD_TOOLS=OFF
+               -D
+               CPUINFO_BUILD_UNIT_TESTS=OFF
+               -D
+               CPUINFO_BUILD_MOCK_TESTS=OFF
+               -D
+               CPUINFO_BUILD_BENCHMARKS=OFF
+               -D
+               CPUINFO_LIBRARY_TYPE=static
+               -D
+               CPUINFO_LOG_LEVEL=error
+               -D
+               CMAKE_INSTALL_PREFIX=<INSTALL_DIR>
+               -D
+               CLOG_SOURCE_DIR="${CPUINFO_SOURCE_DIR}/deps/clog"
   )
-  set(CLOG_SOURCE_DIR "${CPUINFO_SOURCE_DIR}/deps/clog")
-  add_subdirectory("${CPUINFO_SOURCE_DIR}")
-  set(CMAKE_POSITION_INDEPENDENT_CODE
-      ${ORIGINAL_CMAKE_POSITION_INDEPENDENT_CODE_FLAG}
+  add_library(cpuinfo STATIC IMPORTED)
+  set_property(TARGET cpuinfo PROPERTY IMPORTED_LOCATION "${CPUINFO_LIBRARY}")
+  add_dependencies(cpuinfo cpuinfoExternalProject)
+  # Trailing slash matters here! "Move everything from the temporary pthreadpool
+  # install directory to the proper install directory."
+  install(DIRECTORY "${CPUINFO_INSTALL_DIR}/"
+          DESTINATION "${CMAKE_INSTALL_PREFIX}"
   )
 endif()
 
@@ -223,23 +233,24 @@ if(EXECUTORCH_BUILD_PTHREADPOOL)
   set(PTHREADPOOL_SOURCE_DIR
       "${CMAKE_CURRENT_LIST_DIR}/backends/xnnpack/third-party/pthreadpool"
   )
-  if(APPLE)
-    set(PTHREADPOOL_SYNC_PRIMITIVE
-        "condvar"
-        CACHE STRING ""
-    )
-  endif()
   include(ExternalProject)
+  include(GNUInstallDirs)
   set(PTHREADPOOL_BINARY_DIR "${CMAKE_CURRENT_BINARY_DIR}/pthreadpool")
+  set(PTHREADPOOL_INSTALL_DIR "${PTHREADPOOL_BINARY_DIR}/install")
   set(PTHREADPOOL_LIBRARY
-      "${PTHREADPOOL_BINARY_DIR}/install/lib/libpthreadpool.a"
+      "${PTHREADPOOL_INSTALL_DIR}/${CMAKE_INSTALL_LIBDIR}/libpthreadpool.a"
   )
   get_extra_cmake_args_for_external_project(PTHREADPOOL_EXTRA_CMAKE_ARGS)
+  if(APPLE)
+    list(APPEND PTHREADPOOL_EXTRA_CMAKE_ARGS -D
+         PTHREADPOOL_SYNC_PRIMITIVE=condvar
+    )
+  endif()
   ExternalProject_Add(
     pthreadpoolExternalProject
     SOURCE_DIR "${PTHREADPOOL_SOURCE_DIR}"
     BINARY_DIR "${PTHREADPOOL_BINARY_DIR}"
-    INSTALL_DIR "${PTHREADPOOL_BINARY_DIR}/install"
+    INSTALL_DIR "${PTHREADPOOL_INSTALL_DIR}"
     BUILD_BYPRODUCTS "${PTHREADPOOL_LIBRARY}"
     CMAKE_ARGS ${PTHREADPOOL_EXTRA_CMAKE_ARGS}
                -D
@@ -251,8 +262,6 @@ if(EXECUTORCH_BUILD_PTHREADPOOL)
                -D
                PTHREADPOOL_ALLOW_DEPRECATED_API=ON
                -D
-               PTHREADPOOL_SYNC_PRIMITVE=${PTHREADPOOL_SYNC_PRIMITIVE}
-               -D
                CMAKE_POSITION_INDEPENDENT_CODE=ON
                -D
                CMAKE_INSTALL_PREFIX=<INSTALL_DIR>
@@ -262,7 +271,11 @@ if(EXECUTORCH_BUILD_PTHREADPOOL)
     TARGET pthreadpool PROPERTY IMPORTED_LOCATION "${PTHREADPOOL_LIBRARY}"
   )
   add_dependencies(pthreadpool pthreadpoolExternalProject)
-  install(FILES ${PTHREADPOOL_LIBRARY} DESTINATION ${CMAKE_INSTALL_LIBDIR})
+  # Trailing slash matters here! "Move everything from the temporary pthreadpool
+  # install directory to the proper install directory."
+  install(DIRECTORY "${PTHREADPOOL_INSTALL_DIR}/"
+          DESTINATION "${CMAKE_INSTALL_PREFIX}"
+  )
 endif()
 
 # TODO(dbort): Fix these warnings and remove this flag.

backends/arm/quantizer/arm_quantizer.py

@@ -60,7 +60,7 @@
 
 @functools.lru_cache
 def get_symmetric_quantization_config(
-    is_per_channel: bool = False,
+    is_per_channel: bool = True,
     is_qat: bool = False,
     is_dynamic: bool = False,
     act_qmin: int = -128,

backends/arm/test/common.py

@@ -18,6 +18,7 @@
     arm_executor_runner_exists,
     corstone300_installed,
     corstone320_installed,
+    model_converter_installed,
 )
 from executorch.backends.arm.tosa_specification import TosaSpecification
 from executorch.exir.backend.compile_spec_schema import CompileSpec
@@ -245,6 +246,13 @@ def get_u85_compile_spec_unbuilt(
 )
 """Xfails a test if Corsone320 FVP is not installed, or if the executor runner is not built"""
 
+SkipIfNoModelConverter = pytest.mark.skipif(
+    condition=not (model_converter_installed()),
+    raises=FileNotFoundError,
+    reason="Did not find model-converter on path",
+)
+"""Xfails a test if model-converter is not installed"""
+
 xfail_type = str | tuple[str, type[Exception]]