From a506c2b650e617a8fb535088de85a7163168e1f6 Mon Sep 17 00:00:00 2001 From: Anthony Shoumikhin Date: Thu, 10 Jul 2025 23:59:03 -0700 Subject: [PATCH] Rename kernels_custom to kernels_llm --- .../dummy.swift | 0 .../dummy.swift | 0 .ci/scripts/test_llama_torchao_lowbit.sh | 4 ++-- .ci/scripts/test_llava.sh | 4 ++-- .ci/scripts/test_phi_3_mini.sh | 4 ++-- .github/workflows/apple.yml | 2 +- .github/workflows/trunk.yml | 2 +- CMakeLists.txt | 4 ++-- Package.swift | 2 +- backends/cadence/build_cadence_fusionG3.sh | 4 ++-- backends/cadence/build_cadence_hifi4.sh | 4 ++-- backends/vulkan/docs/android_demo.md | 4 ++-- ...lama3-qualcomm-ai-engine-direct-backend.md | 4 ++-- .../using-executorch-building-from-source.md | 2 +- docs/source/using-executorch-ios.md | 2 +- .../docs/delegates/qualcomm_README.md | 6 ++--- .../LLaMA/LLaMA.xcodeproj/project.pbxproj | 22 +++++++++---------- .../LLaMA/SupportingFiles/Debug.xcconfig | 2 +- .../LLaMA/SupportingFiles/Release.xcconfig | 2 +- .../LLaMA/docs/delegates/xnnpack_README.md | 2 +- examples/models/llama/README.md | 8 +++---- examples/models/llava/CMakeLists.txt | 2 +- examples/models/phi-3-mini/README.md | 4 ++-- extension/android/CMakeLists.txt | 2 +- .../Benchmark.xcodeproj/project.pbxproj | 8 +++---- .../apple/Benchmark/Tests/Tests.xcconfig | 2 +- extension/llm/custom_ops/CMakeLists.txt | 2 +- kernels/README.md | 2 +- scripts/build_android_library.sh | 2 +- scripts/build_apple_frameworks.sh | 10 ++++----- setup.py | 6 ++--- test/run_oss_cpp_tests.sh | 2 +- tools/cmake/preset/apple_common.cmake | 2 +- tools/cmake/preset/default.cmake | 10 ++++----- tools/cmake/preset/llm.cmake | 2 +- tools/cmake/preset/pybind.cmake | 4 ++-- tools/cmake/preset/zephyr.cmake | 4 ++-- 37 files changed, 74 insertions(+), 74 deletions(-) rename .Package.swift/{kernels_custom => kernels_llm}/dummy.swift (100%) rename .Package.swift/{kernels_custom_debug => kernels_llm_debug}/dummy.swift (100%) diff --git a/.Package.swift/kernels_custom/dummy.swift b/.Package.swift/kernels_llm/dummy.swift similarity index 100% rename from .Package.swift/kernels_custom/dummy.swift rename to .Package.swift/kernels_llm/dummy.swift diff --git a/.Package.swift/kernels_custom_debug/dummy.swift b/.Package.swift/kernels_llm_debug/dummy.swift similarity index 100% rename from .Package.swift/kernels_custom_debug/dummy.swift rename to .Package.swift/kernels_llm_debug/dummy.swift diff --git a/.ci/scripts/test_llama_torchao_lowbit.sh b/.ci/scripts/test_llama_torchao_lowbit.sh index 21989d26770..83bfc22ba09 100644 --- a/.ci/scripts/test_llama_torchao_lowbit.sh +++ b/.ci/scripts/test_llama_torchao_lowbit.sh @@ -35,7 +35,7 @@ cmake -DPYTHON_EXECUTABLE=python \ -DEXECUTORCH_BUILD_XNNPACK=OFF \ -DEXECUTORCH_BUILD_KERNELS_QUANTIZED=ON \ -DEXECUTORCH_BUILD_KERNELS_OPTIMIZED=ON \ - -DEXECUTORCH_BUILD_KERNELS_CUSTOM=ON \ + -DEXECUTORCH_BUILD_KERNELS_LLM=ON \ -Bcmake-out . cmake --build cmake-out -j16 --target install --config Release @@ -43,7 +43,7 @@ cmake --build cmake-out -j16 --target install --config Release cmake -DPYTHON_EXECUTABLE=python \ -DBUILD_TESTING=OFF \ -DCMAKE_BUILD_TYPE=Release \ - -DEXECUTORCH_BUILD_KERNELS_CUSTOM=ON \ + -DEXECUTORCH_BUILD_KERNELS_LLM=ON \ -DEXECUTORCH_BUILD_KERNELS_OPTIMIZED=ON \ -DEXECUTORCH_BUILD_XNNPACK=OFF \ -DEXECUTORCH_BUILD_KERNELS_QUANTIZED=ON \ diff --git a/.ci/scripts/test_llava.sh b/.ci/scripts/test_llava.sh index 6b584c6ac75..43f7a7348b7 100644 --- a/.ci/scripts/test_llava.sh +++ b/.ci/scripts/test_llava.sh @@ -39,7 +39,7 @@ EXECUTORCH_COMMON_CMAKE_ARGS=" \ -DEXECUTORCH_BUILD_EXTENSION_DATA_LOADER=ON \ -DEXECUTORCH_BUILD_EXTENSION_FLAT_TENSOR=ON \ -DEXECUTORCH_BUILD_EXTENSION_TENSOR=ON \ - -DEXECUTORCH_BUILD_KERNELS_CUSTOM=ON \ + -DEXECUTORCH_BUILD_KERNELS_LLM=ON \ -DEXECUTORCH_BUILD_KERNELS_OPTIMIZED=ON \ -DEXECUTORCH_BUILD_KERNELS_QUANTIZED=ON \ -DEXECUTORCH_BUILD_XNNPACK=ON \ @@ -69,7 +69,7 @@ LLAVA_COMMON_CMAKE_ARGS=" \ -DPYTHON_EXECUTABLE="$PYTHON_EXECUTABLE" \ -DCMAKE_INSTALL_PREFIX=${BUILD_DIR} \ -DCMAKE_BUILD_TYPE=${CMAKE_BUILD_TYPE} \ - -DEXECUTORCH_BUILD_KERNELS_CUSTOM=ON \ + -DEXECUTORCH_BUILD_KERNELS_LLM=ON \ -DEXECUTORCH_BUILD_KERNELS_OPTIMIZED=ON \ -DEXECUTORCH_BUILD_XNNPACK=ON" diff --git a/.ci/scripts/test_phi_3_mini.sh b/.ci/scripts/test_phi_3_mini.sh index 60f2054d30b..2b41e5b308d 100644 --- a/.ci/scripts/test_phi_3_mini.sh +++ b/.ci/scripts/test_phi_3_mini.sh @@ -33,7 +33,7 @@ cmake_install_executorch_libraries() { -DEXECUTORCH_BUILD_XNNPACK=ON \ -DEXECUTORCH_BUILD_KERNELS_QUANTIZED=ON \ -DEXECUTORCH_BUILD_KERNELS_OPTIMIZED=ON \ - -DEXECUTORCH_BUILD_KERNELS_CUSTOM=ON \ + -DEXECUTORCH_BUILD_KERNELS_LLM=ON \ -B${BUILD_DIR} . cmake --build ${BUILD_DIR} -j${NPROC} --target install --config ${BUILD_TYPE} @@ -43,7 +43,7 @@ cmake_build_phi_3_mini() { cmake -DPYTHON_EXECUTABLE=$PYTHON_EXECUTABLE \ -DCMAKE_INSTALL_PREFIX=${BUILD_DIR} \ -DCMAKE_BUILD_TYPE=${BUILD_TYPE} \ - -DEXECUTORCH_BUILD_KERNELS_CUSTOM=ON \ + -DEXECUTORCH_BUILD_KERNELS_LLM=ON \ -DEXECUTORCH_BUILD_KERNELS_OPTIMIZED=ON \ -DEXECUTORCH_BUILD_XNNPACK=ON \ -DEXECUTORCH_BUILD_KERNELS_QUANTIZED=ON \ diff --git a/.github/workflows/apple.yml b/.github/workflows/apple.yml index 05b67855f82..84a36d5664e 100644 --- a/.github/workflows/apple.yml +++ b/.github/workflows/apple.yml @@ -152,7 +152,7 @@ jobs: "backend_coreml" "backend_mps" "backend_xnnpack" - "kernels_custom" + "kernels_llm" "kernels_optimized" "kernels_quantized" "threadpool" diff --git a/.github/workflows/trunk.yml b/.github/workflows/trunk.yml index b90ef543133..bdab21af3da 100644 --- a/.github/workflows/trunk.yml +++ b/.github/workflows/trunk.yml @@ -645,7 +645,7 @@ jobs: -DEXECUTORCH_BUILD_XNNPACK=ON \ -DEXECUTORCH_BUILD_KERNELS_QUANTIZED=ON \ -DEXECUTORCH_BUILD_KERNELS_OPTIMIZED=ON \ - -DEXECUTORCH_BUILD_KERNELS_CUSTOM=ON \ + -DEXECUTORCH_BUILD_KERNELS_LLM=ON \ -DEXECUTORCH_BUILD_DEVTOOLS=ON \ -DEXECUTORCH_ENABLE_EVENT_TRACER=ON \ -Bcmake-out . diff --git a/CMakeLists.txt b/CMakeLists.txt index 963f5edfe32..83a1db3d7b7 100644 --- a/CMakeLists.txt +++ b/CMakeLists.txt @@ -677,7 +677,7 @@ if(EXECUTORCH_BUILD_EXTENSION_TRAINING) add_subdirectory(${CMAKE_CURRENT_SOURCE_DIR}/extension/training) endif() -if(EXECUTORCH_BUILD_KERNELS_CUSTOM) +if(EXECUTORCH_BUILD_KERNELS_LLM) # TODO: move all custom kernels to ${CMAKE_CURRENT_SOURCE_DIR}/kernels/custom add_subdirectory(${CMAKE_CURRENT_SOURCE_DIR}/extension/llm/custom_ops) endif() @@ -704,7 +704,7 @@ if(EXECUTORCH_BUILD_EXECUTOR_RUNNER) list(APPEND _executor_runner_libs quantized_ops_lib) endif() - if(EXECUTORCH_BUILD_KERNELS_CUSTOM) + if(EXECUTORCH_BUILD_KERNELS_LLM) list(APPEND _executor_runner_libs $) endif() diff --git a/Package.swift b/Package.swift index 43760822c19..bcf0b1e22cd 100644 --- a/Package.swift +++ b/Package.swift @@ -69,7 +69,7 @@ let products = deliverables([ "c++", ], ], - "kernels_custom": [:], + "kernels_llm": [:], "kernels_optimized": [ "frameworks": [ "Accelerate", diff --git a/backends/cadence/build_cadence_fusionG3.sh b/backends/cadence/build_cadence_fusionG3.sh index 20e694206db..1c84ae99364 100644 --- a/backends/cadence/build_cadence_fusionG3.sh +++ b/backends/cadence/build_cadence_fusionG3.sh @@ -47,7 +47,7 @@ if $STEPWISE_BUILD; then -DEXECUTORCH_ENABLE_PROGRAM_VERIFICATION=ON \ -DEXECUTORCH_USE_DL=OFF \ -DEXECUTORCH_BUILD_PORTABLE_OPS=ON \ - -DEXECUTORCH_BUILD_KERNELS_CUSTOM=OFF \ + -DEXECUTORCH_BUILD_KERNELS_LLM=OFF \ -DPYTHON_EXECUTABLE=python3 \ -DEXECUTORCH_FUSION_G3_OPT=ON \ -DHAVE_FNMATCH_H=OFF \ @@ -72,7 +72,7 @@ else -DEXECUTORCH_ENABLE_PROGRAM_VERIFICATION=ON \ -DEXECUTORCH_USE_DL=OFF \ -DEXECUTORCH_BUILD_PORTABLE_OPS=ON \ - -DEXECUTORCH_BUILD_KERNELS_CUSTOM=OFF \ + -DEXECUTORCH_BUILD_KERNELS_LLM=OFF \ -DPYTHON_EXECUTABLE=python3 \ -DEXECUTORCH_FUSION_G3_OPT=ON \ -DHAVE_FNMATCH_H=OFF \ diff --git a/backends/cadence/build_cadence_hifi4.sh b/backends/cadence/build_cadence_hifi4.sh index fab8febcef5..e0a48da4074 100644 --- a/backends/cadence/build_cadence_hifi4.sh +++ b/backends/cadence/build_cadence_hifi4.sh @@ -46,7 +46,7 @@ if $STEPWISE_BUILD; then -DEXECUTORCH_ENABLE_PROGRAM_VERIFICATION=ON \ -DEXECUTORCH_USE_DL=OFF \ -DEXECUTORCH_BUILD_PORTABLE_OPS=ON \ - -DEXECUTORCH_BUILD_KERNELS_CUSTOM=OFF \ + -DEXECUTORCH_BUILD_KERNELS_LLM=OFF \ -DPYTHON_EXECUTABLE=python3 \ -DEXECUTORCH_NNLIB_OPT=ON \ -DHAVE_FNMATCH_H=OFF \ @@ -70,7 +70,7 @@ else -DEXECUTORCH_ENABLE_PROGRAM_VERIFICATION=ON \ -DEXECUTORCH_USE_DL=OFF \ -DEXECUTORCH_BUILD_PORTABLE_OPS=ON \ - -DEXECUTORCH_BUILD_KERNELS_CUSTOM=OFF \ + -DEXECUTORCH_BUILD_KERNELS_LLM=OFF \ -DPYTHON_EXECUTABLE=python3 \ -DEXECUTORCH_NNLIB_OPT=ON \ -DHAVE_FNMATCH_H=OFF \ diff --git a/backends/vulkan/docs/android_demo.md b/backends/vulkan/docs/android_demo.md index 1f36b76ec6f..ff84938b06f 100644 --- a/backends/vulkan/docs/android_demo.md +++ b/backends/vulkan/docs/android_demo.md @@ -91,7 +91,7 @@ binary using the Android NDK toolchain. -DEXECUTORCH_BUILD_EXTENSION_TENSOR=ON \ -DEXECUTORCH_BUILD_VULKAN=ON \ -DEXECUTORCH_BUILD_KERNELS_QUANTIZED=ON \ - -DEXECUTORCH_BUILD_KERNELS_CUSTOM=ON \ + -DEXECUTORCH_BUILD_KERNELS_LLM=ON \ -DPYTHON_EXECUTABLE=python \ -Bcmake-android-out && \ cmake --build cmake-android-out -j16 --target install) @@ -102,7 +102,7 @@ binary using the Android NDK toolchain. -DCMAKE_TOOLCHAIN_FILE=$ANDROID_NDK/build/cmake/android.toolchain.cmake \ -DANDROID_ABI=$ANDROID_ABI \ -DEXECUTORCH_BUILD_KERNELS_OPTIMIZED=ON \ - -DEXECUTORCH_BUILD_KERNELS_CUSTOM=ON \ + -DEXECUTORCH_BUILD_KERNELS_LLM=ON \ -DCMAKE_INSTALL_PREFIX=cmake-android-out \ -DPYTHON_EXECUTABLE=python \ -Bcmake-android-out/examples/models/llama && \ diff --git a/docs/source/llm/build-run-llama3-qualcomm-ai-engine-direct-backend.md b/docs/source/llm/build-run-llama3-qualcomm-ai-engine-direct-backend.md index 3dbba3ef5bb..b61dd93233e 100644 --- a/docs/source/llm/build-run-llama3-qualcomm-ai-engine-direct-backend.md +++ b/docs/source/llm/build-run-llama3-qualcomm-ai-engine-direct-backend.md @@ -58,7 +58,7 @@ llama3/Meta-Llama-3-8B-Instruct/tokenizer.model -p -c -c ## Reporting Issues -If you encountered any bugs or issues following this tutorial please file a bug/issue here on Github. \ No newline at end of file +If you encountered any bugs or issues following this tutorial please file a bug/issue here on Github. diff --git a/examples/demo-apps/apple_ios/LLaMA/LLaMA.xcodeproj/project.pbxproj b/examples/demo-apps/apple_ios/LLaMA/LLaMA.xcodeproj/project.pbxproj index 08c4fb9f906..73039dd2b90 100644 --- a/examples/demo-apps/apple_ios/LLaMA/LLaMA.xcodeproj/project.pbxproj +++ b/examples/demo-apps/apple_ios/LLaMA/LLaMA.xcodeproj/project.pbxproj @@ -30,8 +30,8 @@ 03CF439E2CEC5CEC00C7113B /* backend_xnnpack in Frameworks */ = {isa = PBXBuildFile; productRef = 03CF439D2CEC5CEC00C7113B /* backend_xnnpack */; }; 03CF43A02CEC5CEC00C7113B /* backend_xnnpack_debug in Frameworks */ = {isa = PBXBuildFile; productRef = 03CF439F2CEC5CEC00C7113B /* backend_xnnpack_debug */; }; 03CF43A22CEC5CEC00C7113B /* executorch_debug in Frameworks */ = {isa = PBXBuildFile; productRef = 03CF43A12CEC5CEC00C7113B /* executorch_debug */; }; - 03CF43A42CEC5CEC00C7113B /* kernels_custom in Frameworks */ = {isa = PBXBuildFile; productRef = 03CF43A32CEC5CEC00C7113B /* kernels_custom */; }; - 03CF43A62CEC5CEC00C7113B /* kernels_custom_debug in Frameworks */ = {isa = PBXBuildFile; productRef = 03CF43A52CEC5CEC00C7113B /* kernels_custom_debug */; }; + 03CF43A42CEC5CEC00C7113B /* kernels_llm in Frameworks */ = {isa = PBXBuildFile; productRef = 03CF43A32CEC5CEC00C7113B /* kernels_llm */; }; + 03CF43A62CEC5CEC00C7113B /* kernels_llm_debug in Frameworks */ = {isa = PBXBuildFile; productRef = 03CF43A52CEC5CEC00C7113B /* kernels_llm_debug */; }; 03CF43A82CEC5CEC00C7113B /* kernels_optimized in Frameworks */ = {isa = PBXBuildFile; productRef = 03CF43A72CEC5CEC00C7113B /* kernels_optimized */; }; 03CF43AA2CEC5CEC00C7113B /* kernels_optimized_debug in Frameworks */ = {isa = PBXBuildFile; productRef = 03CF43A92CEC5CEC00C7113B /* kernels_optimized_debug */; }; 03CF43B02CEC5CEC00C7113B /* kernels_quantized in Frameworks */ = {isa = PBXBuildFile; productRef = 03CF43AF2CEC5CEC00C7113B /* kernels_quantized */; }; @@ -185,8 +185,8 @@ 03CF439E2CEC5CEC00C7113B /* backend_xnnpack in Frameworks */, 03CF43A02CEC5CEC00C7113B /* backend_xnnpack_debug in Frameworks */, 03CF43A22CEC5CEC00C7113B /* executorch_debug in Frameworks */, - 03CF43A42CEC5CEC00C7113B /* kernels_custom in Frameworks */, - 03CF43A62CEC5CEC00C7113B /* kernels_custom_debug in Frameworks */, + 03CF43A42CEC5CEC00C7113B /* kernels_llm in Frameworks */, + 03CF43A62CEC5CEC00C7113B /* kernels_llm_debug in Frameworks */, 03CF43A82CEC5CEC00C7113B /* kernels_optimized in Frameworks */, 03CF43AA2CEC5CEC00C7113B /* kernels_optimized_debug in Frameworks */, 03CF43B02CEC5CEC00C7113B /* kernels_quantized in Frameworks */, @@ -481,8 +481,8 @@ 03CF439D2CEC5CEC00C7113B /* backend_xnnpack */, 03CF439F2CEC5CEC00C7113B /* backend_xnnpack_debug */, 03CF43A12CEC5CEC00C7113B /* executorch_debug */, - 03CF43A32CEC5CEC00C7113B /* kernels_custom */, - 03CF43A52CEC5CEC00C7113B /* kernels_custom_debug */, + 03CF43A32CEC5CEC00C7113B /* kernels_llm */, + 03CF43A52CEC5CEC00C7113B /* kernels_llm_debug */, 03CF43A72CEC5CEC00C7113B /* kernels_optimized */, 03CF43A92CEC5CEC00C7113B /* kernels_optimized_debug */, 03CF43AF2CEC5CEC00C7113B /* kernels_quantized */, @@ -948,7 +948,7 @@ isa = XCRemoteSwiftPackageReference; repositoryURL = "https://github.com/pytorch/executorch"; requirement = { - branch = "swiftpm-0.8.0.20250702"; + branch = "swiftpm-0.8.0.20250711"; kind = branch; }; }; @@ -990,15 +990,15 @@ package = 03CF43942CEC5CEC00C7113B /* XCRemoteSwiftPackageReference "executorch" */; productName = executorch_debug; }; - 03CF43A32CEC5CEC00C7113B /* kernels_custom */ = { + 03CF43A32CEC5CEC00C7113B /* kernels_llm */ = { isa = XCSwiftPackageProductDependency; package = 03CF43942CEC5CEC00C7113B /* XCRemoteSwiftPackageReference "executorch" */; - productName = kernels_custom; + productName = kernels_llm; }; - 03CF43A52CEC5CEC00C7113B /* kernels_custom_debug */ = { + 03CF43A52CEC5CEC00C7113B /* kernels_llm_debug */ = { isa = XCSwiftPackageProductDependency; package = 03CF43942CEC5CEC00C7113B /* XCRemoteSwiftPackageReference "executorch" */; - productName = kernels_custom_debug; + productName = kernels_llm_debug; }; 03CF43A72CEC5CEC00C7113B /* kernels_optimized */ = { isa = XCSwiftPackageProductDependency; diff --git a/examples/demo-apps/apple_ios/LLaMA/LLaMA/SupportingFiles/Debug.xcconfig b/examples/demo-apps/apple_ios/LLaMA/LLaMA/SupportingFiles/Debug.xcconfig index b0b6055048b..0c749c04c1a 100644 --- a/examples/demo-apps/apple_ios/LLaMA/LLaMA/SupportingFiles/Debug.xcconfig +++ b/examples/demo-apps/apple_ios/LLaMA/LLaMA/SupportingFiles/Debug.xcconfig @@ -7,7 +7,7 @@ OTHER_LDFLAGS = $(inherited) \ -force_load $(BUILT_PRODUCTS_DIR)/libbackend_coreml_debug_$(ET_PLATFORM).a \ -force_load $(BUILT_PRODUCTS_DIR)/libbackend_mps_debug_$(ET_PLATFORM).a \ -force_load $(BUILT_PRODUCTS_DIR)/libbackend_xnnpack_debug_$(ET_PLATFORM).a \ - -force_load $(BUILT_PRODUCTS_DIR)/libkernels_custom_debug_$(ET_PLATFORM).a \ + -force_load $(BUILT_PRODUCTS_DIR)/libkernels_llm_debug_$(ET_PLATFORM).a \ -force_load $(BUILT_PRODUCTS_DIR)/libkernels_optimized_debug_$(ET_PLATFORM).a \ -force_load $(BUILT_PRODUCTS_DIR)/libkernels_quantized_debug_$(ET_PLATFORM).a \ @$(TEMP_DIR)/cmake/linker_flags diff --git a/examples/demo-apps/apple_ios/LLaMA/LLaMA/SupportingFiles/Release.xcconfig b/examples/demo-apps/apple_ios/LLaMA/LLaMA/SupportingFiles/Release.xcconfig index 1d6fdc8d4c4..899d133127f 100644 --- a/examples/demo-apps/apple_ios/LLaMA/LLaMA/SupportingFiles/Release.xcconfig +++ b/examples/demo-apps/apple_ios/LLaMA/LLaMA/SupportingFiles/Release.xcconfig @@ -9,7 +9,7 @@ OTHER_LDFLAGS = $(inherited) \ -force_load $(BUILT_PRODUCTS_DIR)/libbackend_coreml_$(ET_PLATFORM).a \ -force_load $(BUILT_PRODUCTS_DIR)/libbackend_mps_$(ET_PLATFORM).a \ -force_load $(BUILT_PRODUCTS_DIR)/libbackend_xnnpack_$(ET_PLATFORM).a \ - -force_load $(BUILT_PRODUCTS_DIR)/libkernels_custom_$(ET_PLATFORM).a \ + -force_load $(BUILT_PRODUCTS_DIR)/libkernels_llm_$(ET_PLATFORM).a \ -force_load $(BUILT_PRODUCTS_DIR)/libkernels_optimized_$(ET_PLATFORM).a \ -force_load $(BUILT_PRODUCTS_DIR)/libkernels_quantized_$(ET_PLATFORM).a \ @$(TEMP_DIR)/cmake/linker_flags diff --git a/examples/demo-apps/apple_ios/LLaMA/docs/delegates/xnnpack_README.md b/examples/demo-apps/apple_ios/LLaMA/docs/delegates/xnnpack_README.md index de5b3b9ab27..7a56d217b82 100644 --- a/examples/demo-apps/apple_ios/LLaMA/docs/delegates/xnnpack_README.md +++ b/examples/demo-apps/apple_ios/LLaMA/docs/delegates/xnnpack_README.md @@ -147,7 +147,7 @@ The following packages should be linked in your app target `LLaMA` (left side, L - backend_coreml - backend_mps - backend_xnnpack -- kernels_custom +- kernels_llm - kernels_optimized - kernels_portable - kernels_quantized diff --git a/examples/models/llama/README.md b/examples/models/llama/README.md index bbd2107ad74..632f91a010a 100644 --- a/examples/models/llama/README.md +++ b/examples/models/llama/README.md @@ -284,7 +284,7 @@ cmake -DCMAKE_TOOLCHAIN_FILE=$ANDROID_NDK/build/cmake/android.toolchain.cmake \ -DEXECUTORCH_BUILD_XNNPACK=ON \ -DEXECUTORCH_BUILD_KERNELS_OPTIMIZED=ON \ -DEXECUTORCH_BUILD_KERNELS_QUANTIZED=ON \ - -DEXECUTORCH_BUILD_KERNELS_CUSTOM=ON \ + -DEXECUTORCH_BUILD_KERNELS_LLM=ON \ -Bcmake-out-android . cmake --build cmake-out-android -j16 --target install --config Release @@ -301,7 +301,7 @@ cmake -DCMAKE_TOOLCHAIN_FILE=$ANDROID_NDK/build/cmake/android.toolchain.cmake \ -DEXECUTORCH_BUILD_XNNPACK=ON \ -DEXECUTORCH_BUILD_KERNELS_OPTIMIZED=ON \ -DEXECUTORCH_BUILD_KERNELS_QUANTIZED=ON \ - -DEXECUTORCH_BUILD_KERNELS_CUSTOM=ON \ + -DEXECUTORCH_BUILD_KERNELS_LLM=ON \ -DSUPPORT_REGEX_LOOKAHEAD=ON -Bcmake-out-android/examples/models/llama \ examples/models/llama @@ -386,7 +386,7 @@ cmake -DPYTHON_EXECUTABLE=python \ -DEXECUTORCH_BUILD_XNNPACK=OFF \ -DEXECUTORCH_BUILD_KERNELS_QUANTIZED=ON \ -DEXECUTORCH_BUILD_KERNELS_OPTIMIZED=ON \ - -DEXECUTORCH_BUILD_KERNELS_CUSTOM=ON \ + -DEXECUTORCH_BUILD_KERNELS_LLM=ON \ -Bcmake-out . cmake --build cmake-out -j16 --target install --config Release ``` @@ -396,7 +396,7 @@ Next install the llama runner with torchao kernels enabled (similar to step 3.2 ``` cmake -DPYTHON_EXECUTABLE=python \ -DCMAKE_BUILD_TYPE=Release \ - -DEXECUTORCH_BUILD_KERNELS_CUSTOM=ON \ + -DEXECUTORCH_BUILD_KERNELS_LLM=ON \ -DEXECUTORCH_BUILD_KERNELS_OPTIMIZED=ON \ -DEXECUTORCH_BUILD_XNNPACK=OFF \ -DEXECUTORCH_BUILD_KERNELS_QUANTIZED=ON \ diff --git a/examples/models/llava/CMakeLists.txt b/examples/models/llava/CMakeLists.txt index 8c183da4a8a..5edb4e0b386 100644 --- a/examples/models/llava/CMakeLists.txt +++ b/examples/models/llava/CMakeLists.txt @@ -107,7 +107,7 @@ endif() target_link_options_shared_lib(quantized_ops_lib) list(APPEND link_libraries quantized_kernels quantized_ops_lib) -if(EXECUTORCH_BUILD_KERNELS_CUSTOM) +if(EXECUTORCH_BUILD_KERNELS_LLM) list(APPEND link_libraries $) endif() diff --git a/examples/models/phi-3-mini/README.md b/examples/models/phi-3-mini/README.md index 3546ce7f1f2..5571637e021 100644 --- a/examples/models/phi-3-mini/README.md +++ b/examples/models/phi-3-mini/README.md @@ -32,7 +32,7 @@ python -m examples.models.phi-3-mini.export_phi-3-mini -c "4k" -s 128 -o phi-3-m -DEXECUTORCH_BUILD_XNNPACK=ON \ -DEXECUTORCH_BUILD_KERNELS_QUANTIZED=ON \ -DEXECUTORCH_BUILD_KERNELS_OPTIMIZED=ON \ - -DEXECUTORCH_BUILD_KERNELS_CUSTOM=ON \ + -DEXECUTORCH_BUILD_KERNELS_LLM=ON \ -Bcmake-out . cmake --build cmake-out -j16 --target install --config Release @@ -42,7 +42,7 @@ python -m examples.models.phi-3-mini.export_phi-3-mini -c "4k" -s 128 -o phi-3-m cmake -DPYTHON_EXECUTABLE=python \ -DCMAKE_INSTALL_PREFIX=cmake-out \ -DCMAKE_BUILD_TYPE=Release \ - -DEXECUTORCH_BUILD_KERNELS_CUSTOM=ON \ + -DEXECUTORCH_BUILD_KERNELS_LLM=ON \ -DEXECUTORCH_BUILD_KERNELS_OPTIMIZED=ON \ -DEXECUTORCH_BUILD_XNNPACK=ON \ -DEXECUTORCH_BUILD_KERNELS_QUANTIZED=ON \ diff --git a/extension/android/CMakeLists.txt b/extension/android/CMakeLists.txt index 0432abf10b5..6e2c1b8c5d0 100644 --- a/extension/android/CMakeLists.txt +++ b/extension/android/CMakeLists.txt @@ -120,7 +120,7 @@ if(TARGET vulkan_backend) list(APPEND link_libraries vulkan_backend) endif() -if(EXECUTORCH_BUILD_KERNELS_CUSTOM) +if(EXECUTORCH_BUILD_KERNELS_LLM) list(APPEND link_libraries $) endif() diff --git a/extension/benchmark/apple/Benchmark/Benchmark.xcodeproj/project.pbxproj b/extension/benchmark/apple/Benchmark/Benchmark.xcodeproj/project.pbxproj index c9b68f250c1..3551f421f20 100644 --- a/extension/benchmark/apple/Benchmark/Benchmark.xcodeproj/project.pbxproj +++ b/extension/benchmark/apple/Benchmark/Benchmark.xcodeproj/project.pbxproj @@ -23,7 +23,7 @@ 03F181502D7262FC0058BDF9 /* backend_mps in Frameworks */ = {isa = PBXBuildFile; productRef = 03F1814F2D7262FC0058BDF9 /* backend_mps */; }; 03F181522D7262FC0058BDF9 /* backend_xnnpack in Frameworks */ = {isa = PBXBuildFile; productRef = 03F181512D7262FC0058BDF9 /* backend_xnnpack */; }; 03F181542D7262FC0058BDF9 /* executorch in Frameworks */ = {isa = PBXBuildFile; productRef = 03F181532D7262FC0058BDF9 /* executorch */; }; - 03F181562D7262FC0058BDF9 /* kernels_custom in Frameworks */ = {isa = PBXBuildFile; productRef = 03F181552D7262FC0058BDF9 /* kernels_custom */; }; + 03F181562D7262FC0058BDF9 /* kernels_llm in Frameworks */ = {isa = PBXBuildFile; productRef = 03F181552D7262FC0058BDF9 /* kernels_llm */; }; 03F181582D7262FC0058BDF9 /* kernels_optimized in Frameworks */ = {isa = PBXBuildFile; productRef = 03F181572D7262FC0058BDF9 /* kernels_optimized */; }; 03F1815C2D7262FC0058BDF9 /* kernels_quantized in Frameworks */ = {isa = PBXBuildFile; productRef = 03F1815B2D7262FC0058BDF9 /* kernels_quantized */; }; 30AA4B602DC0766800B1BE50 /* pcre2_regex.cpp in Sources */ = {isa = PBXBuildFile; fileRef = 30AA4B5A2DC0766800B1BE50 /* pcre2_regex.cpp */; }; @@ -131,7 +131,7 @@ 03F181502D7262FC0058BDF9 /* backend_mps in Frameworks */, 03F1814E2D7262FC0058BDF9 /* backend_coreml in Frameworks */, 03F181522D7262FC0058BDF9 /* backend_xnnpack in Frameworks */, - 03F181562D7262FC0058BDF9 /* kernels_custom in Frameworks */, + 03F181562D7262FC0058BDF9 /* kernels_llm in Frameworks */, 03F181582D7262FC0058BDF9 /* kernels_optimized in Frameworks */, ); runOnlyForDeploymentPostprocessing = 0; @@ -745,9 +745,9 @@ isa = XCSwiftPackageProductDependency; productName = executorch; }; - 03F181552D7262FC0058BDF9 /* kernels_custom */ = { + 03F181552D7262FC0058BDF9 /* kernels_llm */ = { isa = XCSwiftPackageProductDependency; - productName = kernels_custom; + productName = kernels_llm; }; 03F181572D7262FC0058BDF9 /* kernels_optimized */ = { isa = XCSwiftPackageProductDependency; diff --git a/extension/benchmark/apple/Benchmark/Tests/Tests.xcconfig b/extension/benchmark/apple/Benchmark/Tests/Tests.xcconfig index bf915abc25b..6d4e9bc6b8a 100644 --- a/extension/benchmark/apple/Benchmark/Tests/Tests.xcconfig +++ b/extension/benchmark/apple/Benchmark/Tests/Tests.xcconfig @@ -7,7 +7,7 @@ OTHER_LDFLAGS = $(inherited) \ -force_load $(BUILT_PRODUCTS_DIR)/libbackend_coreml_$(ET_PLATFORM).a \ -force_load $(BUILT_PRODUCTS_DIR)/libbackend_mps_$(ET_PLATFORM).a \ -force_load $(BUILT_PRODUCTS_DIR)/libbackend_xnnpack_$(ET_PLATFORM).a \ - -force_load $(BUILT_PRODUCTS_DIR)/libkernels_custom_$(ET_PLATFORM).a \ + -force_load $(BUILT_PRODUCTS_DIR)/libkernels_llm_$(ET_PLATFORM).a \ -force_load $(BUILT_PRODUCTS_DIR)/libkernels_optimized_$(ET_PLATFORM).a \ -force_load $(BUILT_PRODUCTS_DIR)/libkernels_quantized_$(ET_PLATFORM).a \ @$(TEMP_DIR)/cmake/linker_flags diff --git a/extension/llm/custom_ops/CMakeLists.txt b/extension/llm/custom_ops/CMakeLists.txt index 7052568260b..6f8ac245182 100644 --- a/extension/llm/custom_ops/CMakeLists.txt +++ b/extension/llm/custom_ops/CMakeLists.txt @@ -84,7 +84,7 @@ target_compile_options( install(TARGETS custom_ops DESTINATION lib) -if(EXECUTORCH_BUILD_KERNELS_CUSTOM_AOT) +if(EXECUTORCH_BUILD_KERNELS_LLM_AOT) # Add a AOT library find_package_torch() add_library( diff --git a/kernels/README.md b/kernels/README.md index 58931beb984..5bbb31239d9 100644 --- a/kernels/README.md +++ b/kernels/README.md @@ -351,7 +351,7 @@ Once you have your operator and corresponding tests in place, we can try it out. cmake . \ -DCMAKE_INSTALL_PREFIX=cmake-out \ -DEXECUTORCH_USE_CPP_CODE_COVERAGE=ON \ - -DEXECUTORCH_BUILD_KERNELS_CUSTOM=ON \ + -DEXECUTORCH_BUILD_KERNELS_LLM=ON \ -DEXECUTORCH_BUILD_KERNELS_OPTIMIZED=ON \ -DEXECUTORCH_BUILD_KERNELS_QUANTIZED=ON \ -DEXECUTORCH_BUILD_EXTENSION_DATA_LOADER=ON \ diff --git a/scripts/build_android_library.sh b/scripts/build_android_library.sh index 46e9527d6dc..465a17a840c 100755 --- a/scripts/build_android_library.sh +++ b/scripts/build_android_library.sh @@ -55,7 +55,7 @@ build_android_native_library() { -DEXECUTORCH_BUILD_EXTENSION_TRAINING=ON \ -DEXECUTORCH_BUILD_KERNELS_OPTIMIZED=ON \ -DEXECUTORCH_BUILD_KERNELS_QUANTIZED=ON \ - -DEXECUTORCH_BUILD_KERNELS_CUSTOM="${EXECUTORCH_BUILD_EXTENSION_LLM:-ON}" \ + -DEXECUTORCH_BUILD_KERNELS_LLM="${EXECUTORCH_BUILD_EXTENSION_LLM:-ON}" \ -DEXECUTORCH_BUILD_LLAMA_JNI="${EXECUTORCH_BUILD_EXTENSION_LLM:-ON}" \ -DEXECUTORCH_BUILD_NEURON="${EXECUTORCH_BUILD_NEURON}" \ -DNEURON_BUFFER_ALLOCATOR_LIB="${NEURON_BUFFER_ALLOCATOR_LIB}" \ diff --git a/scripts/build_apple_frameworks.sh b/scripts/build_apple_frameworks.sh index 8c5e5cfbee8..6237fa81e07 100755 --- a/scripts/build_apple_frameworks.sh +++ b/scripts/build_apple_frameworks.sh @@ -54,7 +54,7 @@ libxnnpack_backend.a,\ libxnnpack-microkernels-prod.a,\ :" -FRAMEWORK_KERNELS_CUSTOM="kernels_custom:\ +FRAMEWORK_KERNELS_LLM="kernels_llm:\ libcustom_ops.a,\ :" @@ -78,7 +78,7 @@ usage() { echo " --Debug Build Debug version." echo " --Release Build Release version." echo " --coreml Only build the Core ML backend." - echo " --custom Only build the Custom kernels." + echo " --llm Only build the LLM custom kernels." echo " --mps Only build the Metal Performance Shaders backend." echo " --optimized Only build the Optimized kernels." echo " --quantized Only build the Quantized kernels." @@ -95,7 +95,7 @@ set_cmake_options_override() { # Since the user wants specific options, turn everything off CMAKE_OPTIONS_OVERRIDE=( "-DEXECUTORCH_BUILD_COREML=OFF" - "-DEXECUTORCH_BUILD_KERNELS_CUSTOM=OFF" + "-DEXECUTORCH_BUILD_KERNELS_LLM=OFF" "-DEXECUTORCH_BUILD_MPS=OFF" "-DEXECUTORCH_BUILD_KERNELS_OPTIMIZED=OFF" "-DEXECUTORCH_BUILD_KERNELS_QUANTIZED=OFF" @@ -125,7 +125,7 @@ for arg in "$@"; do fi ;; --coreml) set_cmake_options_override "EXECUTORCH_BUILD_COREML";; - --custom) set_cmake_options_override "EXECUTORCH_BUILD_KERNELS_CUSTOM" ;; + --llm) set_cmake_options_override "EXECUTORCH_BUILD_KERNELS_LLM" ;; --mps) set_cmake_options_override "EXECUTORCH_BUILD_MPS" ;; --optimized) set_cmake_options_override "EXECUTORCH_BUILD_KERNELS_OPTIMIZED" ;; --quantized) set_cmake_options_override "EXECUTORCH_BUILD_KERNELS_QUANTIZED" ;; @@ -232,7 +232,7 @@ for mode in "${MODES[@]}"; do append_framework_flag "EXECUTORCH_BUILD_COREML" "$FRAMEWORK_BACKEND_COREML" "$mode" append_framework_flag "EXECUTORCH_BUILD_MPS" "$FRAMEWORK_BACKEND_MPS" "$mode" append_framework_flag "EXECUTORCH_BUILD_XNNPACK" "$FRAMEWORK_BACKEND_XNNPACK" "$mode" - append_framework_flag "EXECUTORCH_BUILD_KERNELS_CUSTOM" "$FRAMEWORK_KERNELS_CUSTOM" "$mode" + append_framework_flag "EXECUTORCH_BUILD_KERNELS_LLM" "$FRAMEWORK_KERNELS_LLM" "$mode" append_framework_flag "EXECUTORCH_BUILD_KERNELS_OPTIMIZED" "$FRAMEWORK_KERNELS_OPTIMIZED" "$mode" append_framework_flag "EXECUTORCH_BUILD_KERNELS_QUANTIZED" "$FRAMEWORK_KERNELS_QUANTIZED" "$mode" diff --git a/setup.py b/setup.py index cb0dcbbe9f7..0a3608873a4 100644 --- a/setup.py +++ b/setup.py @@ -737,7 +737,7 @@ def run(self): # noqa C901 if cmake_cache.is_enabled("EXECUTORCH_BUILD_COREML"): cmake_build_args += ["--target", "executorchcoreml"] - if cmake_cache.is_enabled("EXECUTORCH_BUILD_KERNELS_CUSTOM_AOT"): + if cmake_cache.is_enabled("EXECUTORCH_BUILD_KERNELS_LLM_AOT"): cmake_build_args += ["--target", "custom_ops_aot_lib"] cmake_build_args += ["--target", "quantized_ops_aot_lib"] @@ -807,14 +807,14 @@ def run(self): # noqa C901 src_name="custom_ops_aot_lib", dst="executorch/extension/llm/custom_ops/", is_dynamic_lib=True, - dependent_cmake_flags=["EXECUTORCH_BUILD_KERNELS_CUSTOM_AOT"], + dependent_cmake_flags=["EXECUTORCH_BUILD_KERNELS_LLM_AOT"], ), BuiltFile( src_dir="%CMAKE_CACHE_DIR%/kernels/quantized/%BUILD_TYPE%/", src_name="quantized_ops_aot_lib", dst="executorch/kernels/quantized/", is_dynamic_lib=True, - dependent_cmake_flags=["EXECUTORCH_BUILD_KERNELS_CUSTOM_AOT"], + dependent_cmake_flags=["EXECUTORCH_BUILD_KERNELS_LLM_AOT"], ), ], ) diff --git a/test/run_oss_cpp_tests.sh b/test/run_oss_cpp_tests.sh index 32368661b19..4b35324f22e 100755 --- a/test/run_oss_cpp_tests.sh +++ b/test/run_oss_cpp_tests.sh @@ -36,7 +36,7 @@ build_executorch() { cmake . \ -DCMAKE_INSTALL_PREFIX=cmake-out \ -DEXECUTORCH_USE_CPP_CODE_COVERAGE=ON \ - -DEXECUTORCH_BUILD_KERNELS_CUSTOM=ON \ + -DEXECUTORCH_BUILD_KERNELS_LLM=ON \ -DEXECUTORCH_BUILD_KERNELS_OPTIMIZED=ON \ -DEXECUTORCH_BUILD_KERNELS_QUANTIZED=ON \ -DEXECUTORCH_BUILD_EXTENSION_DATA_LOADER=ON \ diff --git a/tools/cmake/preset/apple_common.cmake b/tools/cmake/preset/apple_common.cmake index fa2d764dd2b..a14a0932394 100644 --- a/tools/cmake/preset/apple_common.cmake +++ b/tools/cmake/preset/apple_common.cmake @@ -22,6 +22,6 @@ set_overridable_option(EXECUTORCH_BUILD_EXTENSION_DATA_LOADER ON) set_overridable_option(EXECUTORCH_BUILD_EXTENSION_MODULE ON) set_overridable_option(EXECUTORCH_BUILD_EXTENSION_FLAT_TENSOR ON) set_overridable_option(EXECUTORCH_BUILD_EXTENSION_TENSOR ON) -set_overridable_option(EXECUTORCH_BUILD_KERNELS_CUSTOM ON) +set_overridable_option(EXECUTORCH_BUILD_KERNELS_LLM ON) set_overridable_option(EXECUTORCH_BUILD_KERNELS_OPTIMIZED ON) set_overridable_option(EXECUTORCH_BUILD_KERNELS_QUANTIZED ON) diff --git a/tools/cmake/preset/default.cmake b/tools/cmake/preset/default.cmake index f2733f591eb..219e8a27057 100644 --- a/tools/cmake/preset/default.cmake +++ b/tools/cmake/preset/default.cmake @@ -65,12 +65,12 @@ define_overridable_option( BOOL OFF ) define_overridable_option( - EXECUTORCH_BUILD_KERNELS_CUSTOM + EXECUTORCH_BUILD_KERNELS_LLM "Build the custom kernels" BOOL OFF ) define_overridable_option( - EXECUTORCH_BUILD_KERNELS_CUSTOM_AOT + EXECUTORCH_BUILD_KERNELS_LLM_AOT "Build the custom ops lib for AOT" BOOL OFF ) @@ -307,17 +307,17 @@ check_required_options_on( check_required_options_on( IF_ON - EXECUTORCH_BUILD_KERNELS_CUSTOM + EXECUTORCH_BUILD_KERNELS_LLM REQUIRES EXECUTORCH_BUILD_KERNELS_OPTIMIZED ) check_required_options_on( IF_ON - EXECUTORCH_BUILD_KERNELS_CUSTOM_AOT + EXECUTORCH_BUILD_KERNELS_LLM_AOT REQUIRES EXECUTORCH_BUILD_EXTENSION_TENSOR - EXECUTORCH_BUILD_KERNELS_CUSTOM + EXECUTORCH_BUILD_KERNELS_LLM ) check_required_options_on( diff --git a/tools/cmake/preset/llm.cmake b/tools/cmake/preset/llm.cmake index da1364eb2ad..7d450125988 100644 --- a/tools/cmake/preset/llm.cmake +++ b/tools/cmake/preset/llm.cmake @@ -12,7 +12,7 @@ set_overridable_option(EXECUTORCH_BUILD_EXTENSION_DATA_LOADER ON) set_overridable_option(EXECUTORCH_BUILD_EXTENSION_FLAT_TENSOR ON) set_overridable_option(EXECUTORCH_BUILD_EXTENSION_MODULE ON) set_overridable_option(EXECUTORCH_BUILD_EXTENSION_TENSOR ON) -set_overridable_option(EXECUTORCH_BUILD_KERNELS_CUSTOM ON) +set_overridable_option(EXECUTORCH_BUILD_KERNELS_LLM ON) set_overridable_option(EXECUTORCH_BUILD_KERNELS_OPTIMIZED ON) set_overridable_option(EXECUTORCH_BUILD_KERNELS_QUANTIZED ON) set_overridable_option(EXECUTORCH_BUILD_XNNPACK ON) diff --git a/tools/cmake/preset/pybind.cmake b/tools/cmake/preset/pybind.cmake index 3f19a87805d..e52317bf452 100644 --- a/tools/cmake/preset/pybind.cmake +++ b/tools/cmake/preset/pybind.cmake @@ -15,8 +15,8 @@ set_overridable_option(EXECUTORCH_LOG_LEVEL Info) set_overridable_option(EXECUTORCH_BUILD_XNNPACK ON) set_overridable_option(EXECUTORCH_BUILD_EXTENSION_TENSOR ON) set_overridable_option(EXECUTORCH_BUILD_EXTENSION_RUNNER_UTIL ON) -set_overridable_option(EXECUTORCH_BUILD_KERNELS_CUSTOM ON) -set_overridable_option(EXECUTORCH_BUILD_KERNELS_CUSTOM_AOT ON) +set_overridable_option(EXECUTORCH_BUILD_KERNELS_LLM ON) +set_overridable_option(EXECUTORCH_BUILD_KERNELS_LLM_AOT ON) set_overridable_option(EXECUTORCH_BUILD_EXTENSION_FLAT_TENSOR ON) set_overridable_option(EXECUTORCH_BUILD_EXTENSION_DATA_LOADER ON) set_overridable_option(EXECUTORCH_BUILD_KERNELS_OPTIMIZED ON) diff --git a/tools/cmake/preset/zephyr.cmake b/tools/cmake/preset/zephyr.cmake index a78af0ea44a..f810b9cc96c 100644 --- a/tools/cmake/preset/zephyr.cmake +++ b/tools/cmake/preset/zephyr.cmake @@ -7,8 +7,8 @@ set_overridable_option(EXECUTORCH_BUILD_COREML OFF) set_overridable_option(EXECUTORCH_ENABLE_EVENT_TRACER OFF) -set_overridable_option(EXECUTORCH_BUILD_KERNELS_CUSTOM OFF) -set_overridable_option(EXECUTORCH_BUILD_KERNELS_CUSTOM_AOT OFF) +set_overridable_option(EXECUTORCH_BUILD_KERNELS_LLM OFF) +set_overridable_option(EXECUTORCH_BUILD_KERNELS_LLM_AOT OFF) set_overridable_option(EXECUTORCH_BUILD_EXTENSION_DATA_LOADER OFF) set_overridable_option(EXECUTORCH_BUILD_EXTENSION_FLAT_TENSOR OFF) set_overridable_option(EXECUTORCH_BUILD_EXTENSION_LLM OFF)