Merge branch 'main' into milestone2.1

Author: leafs1

.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|>

.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

@@ -112,11 +112,12 @@ set(CMAKE_SKIP_BUILD_RPATH OFF)
 set(CMAKE_BUILD_WITH_INSTALL_RPATH ON)
 # Automatically add all linked folders that are NOT in the build directory to
 # the rpath (per library?)
-# TODO: Doesn't work for us right now because we are
-# not installing .so's into the correct locations. For example we have
-# libcustom_ops_aot_lib.so depending on _portable_lib.so, which was eventually
-# put under <site-packages>/executorch/extension/pybindings/ but this rpath is
-# not automatically added because at build time it seems `portable_lib` is being
+#
+# TODO: Doesn't work for us right now because we are not installing .so's into
+# the correct locations. For example we have libcustom_ops_aot_lib.so depending
+# on _portable_lib.so, which was eventually put under
+# <site-packages>/executorch/extension/pybindings/ but this rpath is not
+# automatically added because at build time it seems `portable_lib` is being
 # built under the same directory, so no extra rpath is being added. To properly
 # fix this we need to install `portable_lib` into the correct path.
 set(CMAKE_INSTALL_RPATH_USE_LINK_PATH ON)
@@ -321,8 +322,9 @@ if(EXECUTORCH_USE_CPP_CODE_COVERAGE)
            " -fprofile-instr-generate -fcoverage-mapping"
     )
   else()
-    message(FATAL_ERROR
-            "Code coverage for compiler ${CMAKE_CXX_COMPILER_ID} is unsupported"
+    message(
+      FATAL_ERROR
+        "Code coverage for compiler ${CMAKE_CXX_COMPILER_ID} is unsupported"
     )
   endif()
 endif()
@@ -633,8 +635,8 @@ if(EXECUTORCH_BUILD_PYBIND)
   endif()
 
   if(EXECUTORCH_BUILD_XNNPACK)
-    # need to explicitly specify XNNPACK and xnnpack-microkernels-prod here otherwise
-    # uses XNNPACK and microkernel-prod symbols from libtorch_cpu
+    # need to explicitly specify XNNPACK and xnnpack-microkernels-prod here
+    # otherwise uses XNNPACK and microkernel-prod symbols from libtorch_cpu
     list(APPEND _dep_libs xnnpack_backend XNNPACK xnnpack-microkernels-prod)
   endif()
 

backends/apple/coreml/CMakeLists.txt

@@ -104,7 +104,9 @@ if(APPLE)
 endif()
 
 add_library(coreml_util ${UTIL_SOURCES})
-target_include_directories(coreml_util PUBLIC ${CMAKE_CURRENT_SOURCE_DIR}/runtime/util)
+target_include_directories(
+  coreml_util PUBLIC ${CMAKE_CURRENT_SOURCE_DIR}/runtime/util
+)
 if(APPLE)
   target_link_libraries(coreml_util PRIVATE ${FOUNDATION_FRAMEWORK})
 endif()
@@ -119,21 +121,25 @@ install(
 
 # CoreML inmemoryfs
 
-set(
-  INMEMORYFS_SOURCES
-  runtime/inmemoryfs/inmemory_filesystem.cpp
-  runtime/inmemoryfs/memory_buffer.cpp
-  runtime/inmemoryfs/memory_stream.cpp
-  runtime/inmemoryfs/reversed_memory_stream.cpp
+set(INMEMORYFS_SOURCES
+    runtime/inmemoryfs/inmemory_filesystem.cpp
+    runtime/inmemoryfs/memory_buffer.cpp runtime/inmemoryfs/memory_stream.cpp
+    runtime/inmemoryfs/reversed_memory_stream.cpp
 )
 if(APPLE)
-  list(APPEND INMEMORYFS_SOURCES runtime/inmemoryfs/inmemory_filesystem_utils.mm)
+  list(APPEND INMEMORYFS_SOURCES
+       runtime/inmemoryfs/inmemory_filesystem_utils.mm
+  )
 endif()
 
 add_library(coreml_inmemoryfs ${INMEMORYFS_SOURCES})
-target_include_directories(coreml_inmemoryfs PUBLIC ${CMAKE_CURRENT_SOURCE_DIR}/runtime/inmemoryfs)
+target_include_directories(
+  coreml_inmemoryfs PUBLIC ${CMAKE_CURRENT_SOURCE_DIR}/runtime/inmemoryfs
+)
 if(APPLE)
-  target_link_libraries(coreml_inmemoryfs PRIVATE coreml_util ${FOUNDATION_FRAMEWORK})
+  target_link_libraries(
+    coreml_inmemoryfs PRIVATE coreml_util ${FOUNDATION_FRAMEWORK}
+  )
 endif()
 target_compile_options(coreml_inmemoryfs PUBLIC -fPIC)
 
@@ -148,17 +154,12 @@ install(
 
 if(EXECUTORCH_BUILD_PYBIND)
   pybind11_add_module(
-    executorchcoreml
-    SHARED
-      runtime/inmemoryfs/inmemory_filesystem_py.cpp
-      runtime/inmemoryfs/inmemory_filesystem_utils.cpp
+    executorchcoreml SHARED runtime/inmemoryfs/inmemory_filesystem_py.cpp
+    runtime/inmemoryfs/inmemory_filesystem_utils.cpp
   )
   target_link_libraries(
-    executorchcoreml
-    PRIVATE
-      coreml_util
-      coreml_inmemoryfs
-      nlohmann_json::nlohmann_json
+    executorchcoreml PRIVATE coreml_util coreml_inmemoryfs
+                             nlohmann_json::nlohmann_json
   )
   target_compile_options(executorchcoreml PUBLIC -fPIC)
 endif()
@@ -179,8 +180,12 @@ if(APPLE)
     coremldelegate PRIVATE ${CMAKE_CURRENT_SOURCE_DIR}/runtime/delegate
   )
   target_include_directories(coremldelegate PRIVATE ${PROJECT_SOURCE_DIR}/..)
-  target_include_directories(coremldelegate PRIVATE ${PROJECT_SOURCE_DIR}/runtime/core/portable_type/c10)
-  target_compile_definitions(coremldelegate PRIVATE C10_USING_CUSTOM_GENERATED_MACROS)
+  target_include_directories(
+    coremldelegate PRIVATE ${PROJECT_SOURCE_DIR}/runtime/core/portable_type/c10
+  )
+  target_compile_definitions(
+    coremldelegate PRIVATE C10_USING_CUSTOM_GENERATED_MACROS
+  )
 
   if(EXECUTORCH_BUILD_DEVTOOLS)
     target_sources(coremldelegate PRIVATE ${SDK_SOURCES} ${PROTOBUF_SOURCES})
@@ -200,13 +205,9 @@ if(APPLE)
 
   target_link_libraries(
     coremldelegate
-    PUBLIC  coreml_util
-            coreml_inmemoryfs
-    PRIVATE executorch_core
-            ${ACCELERATE_FRAMEWORK}
-            ${COREML_FRAMEWORK}
-            ${FOUNDATION_FRAMEWORK}
-            ${SQLITE_LIBRARY}
+    PUBLIC coreml_util coreml_inmemoryfs
+    PRIVATE executorch_core ${ACCELERATE_FRAMEWORK} ${COREML_FRAMEWORK}
+            ${FOUNDATION_FRAMEWORK} ${SQLITE_LIBRARY}
   )
 
   target_link_options_shared_lib(coremldelegate)
@@ -218,13 +219,8 @@ if(APPLE)
   endif()
 
   target_compile_options(
-    coremldelegate
-    PRIVATE
-      -fobjc-arc
-      -fno-exceptions
-      -x objective-c++
-      -Wno-null-character
-      -Wno-receiver-expr
+    coremldelegate PRIVATE -fobjc-arc -fno-exceptions -x objective-c++
+                           -Wno-null-character -Wno-receiver-expr
   )
 
   if(EXECUTORCH_BUILD_DEVTOOLS)

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]]
 
 

backends/arm/test/misc/test_bn_relu_folding_qat.py

@@ -59,7 +59,9 @@ def test_qat_tosa_BI(model: torch.nn.Module):
         "quantize",
         Quantize(
             quantizer=quantizer,
-            quantization_config=get_symmetric_quantization_config(is_qat=True),
+            quantization_config=get_symmetric_quantization_config(
+                is_qat=True, is_per_channel=False
+            ),
             is_qat=True,
         ),
     )

backends/arm/test/ops/test_add.py

@@ -7,8 +7,6 @@
 
 from typing import Tuple
 
-import pytest
-
 import torch
 from executorch.backends.arm.arm_backend import get_tosa_spec
 from executorch.backends.arm.quantizer import arm_quantizer
@@ -190,7 +188,7 @@ def test_add_tensor_u85_BI_2(test_data: input_t2):
 
 
 @common.parametrize("test_data", Add.test_data)
-@pytest.mark.skip(reason="Model converter not yet made available")
+@common.SkipIfNoModelConverter
 def test_add_tensor_vgf_fp(test_data: input_t1):
     pipeline = VgfPipeline[input_t1](
         Add(), test_data(), aten_op, exir_op, tosa_version="TOSA-1.0+FP"
@@ -199,7 +197,7 @@ def test_add_tensor_vgf_fp(test_data: input_t1):
 
 
 @common.parametrize("test_data", Add.test_data)
-@pytest.mark.skip(reason="Model converter not yet made available")
+@common.SkipIfNoModelConverter
 def test_add_tensor_vgf_int(test_data: input_t1):
     pipeline = VgfPipeline[input_t1](
         Add(),