Add CMake preset for LLM (#11066)

This PR was created by the merge bot to help merge the original PR into
the main branch.
ghstack PR number: #11052 by
@larryliu0820
^ Please use this as the source of truth for the PR details, comments,
and reviews
ghstack PR base:
https://github.com/pytorch/executorch/tree/gh/larryliu0820/64/base
ghstack PR head:
https://github.com/pytorch/executorch/tree/gh/larryliu0820/64/head
Merge bot PR base: https://github.com/pytorch/executorch/tree/main
Merge bot PR head:
https://github.com/pytorch/executorch/tree/gh/larryliu0820/64/orig
@diff-train-skip-merge

Co-authored-by: Mengwei Liu <[email protected]>

Author: pytorchbot

.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"

.github/workflows/build-presets.yml

@@ -20,7 +20,7 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        preset: [macos-arm64, pybind]
+        preset: [macos-arm64, pybind, llm]
     with:
       job-name: build
       ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
@@ -39,7 +39,7 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        preset: [pybind]
+        preset: [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:

CMakePresets.json

@@ -36,6 +36,26 @@
         "string": "${hostSystemName}",
         "list": ["Darwin", "Linux", "Windows"]
       }
+    },
+    {
+        "name": "llm",
+        "displayName": "Build LLM libraries",
+        "inherits": [
+            "common"
+        ],
+        "cacheVariables": {
+            "EXECUTORCH_BUILD_PRESET_FILE": "${sourceDir}/tools/cmake/preset/llm.cmake",
+            "CMAKE_OSX_DEPLOYMENT_TARGET": "10.15"
+        },
+        "condition": {
+            "type": "inList",
+            "string": "${hostSystemName}",
+            "list": [
+                "Darwin",
+                "Linux",
+                "Windows"
+            ]
+        }
     }
   ]
 }

examples/models/llama/CMakeLists.txt

@@ -88,7 +88,7 @@ add_subdirectory(runner)
 set(link_libraries executorch gflags)
 set(_srcs main.cpp)
 
-if(EXECUTORCH_BUILD_KERNELS_OPTIMIZED)
+if(TARGET optimized_native_cpu_ops_lib)
   list(
     APPEND
     link_libraries
@@ -108,7 +108,7 @@ endif()
 target_link_options_shared_lib(quantized_ops_lib)
 list(APPEND link_libraries quantized_kernels quantized_ops_lib)
 
-if(EXECUTORCH_BUILD_KERNELS_CUSTOM)
+if(TARGET custom_ops)
   target_link_options_shared_lib(custom_ops)
   list(APPEND link_libraries custom_ops)
 endif()

tools/cmake/preset/llm.cmake

@@ -0,0 +1,31 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the BSD-style license found in the
+# LICENSE file in the root directory of this source tree.
+
+# Enable logging even when in release mode. We are building for desktop, where
+# saving a few kB is less important than showing useful error information to
+# users.
+# keep sorted
+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_OPTIMIZED ON)
+set_overridable_option(EXECUTORCH_BUILD_KERNELS_QUANTIZED ON)
+set_overridable_option(EXECUTORCH_BUILD_XNNPACK ON)
+
+if(CMAKE_SYSTEM_NAME STREQUAL "Darwin")
+  set_overridable_option(EXECUTORCH_BUILD_COREML ON)
+  set_overridable_option(EXECUTORCH_BUILD_MPS ON)
+elseif(CMAKE_SYSTEM_NAME STREQUAL "Linux")
+  # Linux-specific code here
+elseif(CMAKE_SYSTEM_NAME STREQUAL "Windows" OR CMAKE_SYSTEM_NAME STREQUAL "WIN32")
+  # Windows or other OS-specific code here
+elseif(CMAKE_SYSTEM_NAME STREQUAL "Android")
+  # Android-specific code here
+else()
+  message(FATAL_ERROR "Unsupported CMAKE_SYSTEM_NAME for LLM: ${CMAKE_SYSTEM_NAME}")
+endif()