Update on "Reuse GELU implementation from PyTorch core"

kernels/optimized doesn't need to support embedded systems, so it can just take a header-only dep on PyTorch.

Note that, because we will pick up Sleef internally and ignore it
externally thanks to ATen vec, this PR gets to enable optimized GELU in OSS.

Testing: CI to make sure this doesn't break mobile build modes; happy to take advice on anything not currently covered that might break.

Differential Revision: [D66335522](https://our.internmc.facebook.com/intern/diff/D66335522/)

[ghstack-poisoned]

Author: swolchok

.ci/scripts/build_llama_android.sh

@@ -14,8 +14,7 @@ if [[ -z "${PYTHON_EXECUTABLE:-}" ]]; then
   PYTHON_EXECUTABLE=python3
 fi
 which "${PYTHON_EXECUTABLE}"
-SITE_PACKAGES="$(${PYTHON_EXECUTABLE} -c 'from distutils.sysconfig import get_python_lib; print(get_python_lib())')"
-CMAKE_PREFIX_PATH="${SITE_PACKAGES}/torch"
+CMAKE_PREFIX_PATH="$(python3 -c 'import torch as _; print(_.__path__[0])')"
 
 install_executorch_and_backend_lib() {
   echo "Installing executorch and xnnpack backend"

.ci/scripts/test_model.sh

@@ -50,8 +50,7 @@ prepare_artifacts_upload() {
 
 build_cmake_executor_runner() {
   echo "Building executor_runner"
-  SITE_PACKAGES="$(${PYTHON_EXECUTABLE} -c 'from distutils.sysconfig import get_python_lib; print(get_python_lib())')"
-  CMAKE_PREFIX_PATH="${SITE_PACKAGES}/torch"
+  CMAKE_PREFIX_PATH="$(python3 -c 'import torch as _; print(_.__path__[0])')"
   rm -rf ${CMAKE_OUTPUT_DIR}
   cmake -DCMAKE_BUILD_TYPE=Debug \
       -DEXECUTORCH_BUILD_KERNELS_OPTIMIZED=ON \
@@ -101,8 +100,7 @@ test_model() {
 
 build_cmake_xnn_executor_runner() {
   echo "Building xnn_executor_runner"
-  SITE_PACKAGES="$(${PYTHON_EXECUTABLE} -c 'from distutils.sysconfig import get_python_lib; print(get_python_lib())')"
-  CMAKE_PREFIX_PATH="${SITE_PACKAGES}/torch"
+  CMAKE_PREFIX_PATH="$(python3 -c 'import torch as _; print(_.__path__[0])')"
 
   (rm -rf ${CMAKE_OUTPUT_DIR} \
     && mkdir ${CMAKE_OUTPUT_DIR} \

.ci/scripts/test_phi_3_mini.sh

@@ -22,8 +22,7 @@ NPROC=8
 if hash nproc &> /dev/null; then NPROC=$(nproc); fi
 
 cmake_install_executorch_libraries() {
-  SITE_PACKAGES="$(${PYTHON_EXECUTABLE} -c 'from distutils.sysconfig import get_python_lib; print(get_python_lib())')"
-  CMAKE_PREFIX_PATH="${SITE_PACKAGES}/torch"
+  CMAKE_PREFIX_PATH="$(python3 -c 'import torch as _; print(_.__path__[0])')"
   cmake -DPYTHON_EXECUTABLE=python \
       -DCMAKE_INSTALL_PREFIX=${BUILD_DIR} \
       -DCMAKE_PREFIX_PATH="${CMAKE_PREFIX_PATH}" \
@@ -42,8 +41,7 @@ cmake_install_executorch_libraries() {
 }
 
 cmake_build_phi_3_mini() {
-  SITE_PACKAGES="$(${PYTHON_EXECUTABLE} -c 'from distutils.sysconfig import get_python_lib; print(get_python_lib())')"
-  CMAKE_PREFIX_PATH="${SITE_PACKAGES}/torch"
+  CMAKE_PREFIX_PATH="$(python3 -c 'import torch as _; print(_.__path__[0])')"
   cmake -DPYTHON_EXECUTABLE=$PYTHON_EXECUTABLE \
       -DCMAKE_INSTALL_PREFIX=${BUILD_DIR} \
       -DCMAKE_PREFIX_PATH="${CMAKE_PREFIX_PATH}" \

build/build_android_llm_demo.sh

@@ -11,8 +11,7 @@ if [[ -z "${PYTHON_EXECUTABLE:-}" ]]; then
   PYTHON_EXECUTABLE=python3
 fi
 which "${PYTHON_EXECUTABLE}"
-SITE_PACKAGES="$(${PYTHON_EXECUTABLE} -c 'from distutils.sysconfig import get_python_lib; print(get_python_lib())')"
-CMAKE_PREFIX_PATH="${SITE_PACKAGES}/torch"
+CMAKE_PREFIX_PATH="$(python3 -c 'import torch as _; print(_.__path__[0])')"
 
 build_jar() {
   pushd extension/android

test/run_oss_cpp_tests.sh

@@ -32,8 +32,7 @@ build_executorch() {
   if [ -x "$(command -v glslc)" ]; then
     BUILD_VULKAN="ON"
   fi
-  SITE_PACKAGES="$(${PYTHON_EXECUTABLE} -c 'from distutils.sysconfig import get_python_lib; print(get_python_lib())')"
-  CMAKE_PREFIX_PATH="${SITE_PACKAGES}/torch"
+  CMAKE_PREFIX_PATH="$(python3 -c 'import torch as _; print(_.__path__[0])')"
   cmake . \
     -DCMAKE_INSTALL_PREFIX=cmake-out \
     -DCMAKE_PREFIX_PATH="${CMAKE_PREFIX_PATH}" \