我的环境：
mac intel_x84_64，
osx 13.3
python3.8.2
xcode14.3(最后切换到了13.3)
android sdk 30.0.3
android ndk 21
tensorflow 2.12源码
bazel 5.3.0

配置环境

编译配置

这一步的目的是为了执行.configure文件，调用bazel编译ios版tflite framwork

首先安装android sdk这个大家都知道了，
然后还要下载ndk，
手动下载：https://github.com/android/ndk/wiki/Unsupported-Downloads
【推荐】进入Android Studio - SDK Manage自动下载

image.png

我就是手动下载ndk，编译一直失败，自动下载ndk就成功了。

我把sdk放到了/Users/coorell/Develop/Android/sdk
把ndk都放到了/Users/coorell/Develop/Android/sdk/ndk/r20b

下载源码并配置

https://github.com/tensorflow/tensorflow
下载源码到本地，cd到该目录，并配置

./configure

配置过程会自动安装bazel5.3.0，如果需要手动安装，请看https://www.jianshu.com/p/1a30a2549539

开始配置，输入回车和y/n等

You have bazel 5.3.0 installed.
Please specify the location of python. [Default is /Users/coorell/.pyenv/versions/3.8.2/bin/python3]:直接回车，表示默认


Found possible Python library paths:
  /Users/coorell/.pyenv/versions/3.8.2/lib/python3.8/site-packages
Please input the desired Python library path to use.  Default is [/Users/coorell/.pyenv/versions/3.8.2/lib/python3.8/site-packages]:直接回车，表示默认

Do you wish to build TensorFlow with ROCm support? [y/N]: n
No ROCm support will be enabled for TensorFlow.

Do you wish to build TensorFlow with CUDA support? [y/N]: n
No CUDA support will be enabled for TensorFlow.

Do you wish to download a fresh release of clang? (Experimental) [y/N]: n
Clang will not be downloaded.

Please specify optimization flags to use during compilation when bazel option "--config=opt" is specified [Default is -Wno-sign-compare]:直接回车，表示默认


Would you like to interactively configure ./WORKSPACE for Android builds? [y/N]: y
Searching for NDK and SDK installations.

Please specify the home path of the Android NDK to use. [Default is /Users/coorell/library/Android/Sdk/ndk-bundle]: /Users/coorell/Develop/Android/ndk/r20b


Please specify the (min) Android NDK API level to use. [Available levels: ['16', '17', '18', '19', '21', '22', '23', '24', '26', '27', '28', '29']] [Default is 26]: 21


Please specify the home path of the Android SDK to use. [Default is /Users/coorell/library/Android/Sdk]: /Users/coorell/Develop/Android/sdk


Please specify the Android SDK API level to use. [Available levels: ['28', '28-2', '29', '30', '31']] [Default is 31]: 30


Please specify an Android build tools version to use. [Available versions: ['28.0.3', '29.0.2', '31.0.0']] [Default is 31.0.0]: 30.0.2


Do you wish to build TensorFlow with iOS support? [y/N]: y
iOS support will be enabled for TensorFlow.

编译尝试

bazel build -c opt --fat_apk_cpu=x86,x86_64,arm64-v8a,armeabi-v7a \
  --host_crosstool_top=@bazel_tools//tools/cpp:toolchain \
  //tensorflow/lite/java:tensorflow-lite

编译过程
编译完成

如果编译失败可以重启一下电脑试试.
编译成功输出 aar 路径 :
tensorflow源码/bazel-bin/tensorflow/lite/java/tensorflow-lite.aar

编译成功，则表示bazel完全安装好了，接下来具体看配置android和ios

====Android====

安卓直接使用tflite会出现如下提示

[ERROR:flutter/lib/ui/ui_dart_state.cc(209)] Unhandled Exception: Invalid argument(s): Failed to load dynamic library 'libtensorflowlite_c.so': dlopen failed: library "libtensorflowlite_c.so" not found

我们只需要一个脚本，就可以自动下载so依赖文件到项目目录<projectdirectory>/android/app/src/main/jniLibs下

#!/usr/bin/env bash

cd "$(dirname "$(readlink -f "$0")")"

# Available versions
# 2.5, 2.4.1

TF_VERSION=2.5
URL="https://github.com/am15h/tflite_flutter_plugin/releases/download/"
TAG="tf_$TF_VERSION"

#download to <projectdirectory>/android/app/src/main/jniLibs
ANDROID_DIR="android/app/src/main/jniLibs/"
ANDROID_LIB="libtensorflowlite_c.so"

ARM_DELEGATE="libtensorflowlite_c_arm_delegate.so"
ARM_64_DELEGATE="libtensorflowlite_c_arm64_delegate.so"
ARM="libtensorflowlite_c_arm.so"
ARM_64="libtensorflowlite_c_arm64.so"
X86="libtensorflowlite_c_x86_delegate.so"
X86_64="libtensorflowlite_c_x86_64_delegate.so"

delegate=0

while getopts "d" OPTION
do
    case $OPTION in
        d)  delegate=1;;
    esac
done


download () {
    wget "${URL}${TAG}/$1" -O "$1"
    mkdir -p "${ANDROID_DIR}$2/"
    mv $1 "${ANDROID_DIR}$2/${ANDROID_LIB}"
}

if [ ${delegate} -eq 1 ]
then

download ${ARM_DELEGATE} "armeabi-v7a"
download ${ARM_64_DELEGATE} "arm64-v8a"

else

download ${ARM} "armeabi-v7a"
download ${ARM_64} "arm64-v8a"

fi

download ${X86} "x86"
download ${X86_64} "x86_64"

把这上面的代码放到项目根目录，install.sh文件，打开命令行窗口，运行

sh install.sh

sh install.sh结果 自动下载完成

安卓好简单吧，自动化脚本，这样app就可以直接运行tflite推理拉。

====IOS====

配置环境

如果没安装过xcode，可以

xcode-select --install
sudo xcodebuild -license accept

我以前已经安装过了，并且更新系统后升级到了xcode14.3,

不过编译了n多次，都提示错误

ERROR: /private/var/tmp/_bazel_thao/26d40dc75f2c247e7283b353a9ab184f/external/local_config_cc/BUILD:48:19: in cc_toolchain_suite rule @local_config_cc//:toolchain: cc_toolchain_suite '@local_config_cc//:toolchain' does not contain a toolchain for cpu 'ios_arm64'

toolchain error

折腾了3天，最后用xcodes安装了旧版本xcode13.3才安装成功

#安装xcodes，管理多版本xcode
brew install --cask xcodes

image.png

#查看当前默认xcode版本
xcode-select -p
#切换13.3.1版本
sudo xcode-select -s /Applications/Xcode-13.3.1.app/Contents/Developer

xcode-select

编译：

bazel build --config=ios_fat -c opt --cxxopt=--std=c++17 \
  //tensorflow/lite/ios:TensorFlowLiteC_framework

如果是给ios模拟器使用，则需要根据你mac的cpu来，我的是intel x86_64的,用以下命令

bazel build --cpu=x86_64 -c opt --cxxopt=--std=c++17 \
  //tensorflow/lite/ios:TensorFlowLiteC_framework

终于编译成功

生成的文件在
tensorflow源码/bazel-bin/tensorflow/lite/ios/TensorFlowLiteC_framework.zip

回顾

折腾了3天，编译成功后的结果里，又提示默认选择了xcode14.3

/bazel_tools/tools/osx/xcode_configure.bzl:243:14: No default Xcode version is set with 'xcode-select'; picking ':version14_3_0_14E222b'

难道是因为xcode是升级过来的，某些东西安装不完整，安装13.3的时候全部都重新安装上了？换成14.3是否可以？答案是no，我通过xcode-select -p切换到14.3版本，编译还是会失败，提示之前的错误'@local_config_cc//:toolchain' does not contain a toolchain for cpu 'ios_arm64'

如果失败，需要清理一下，重新配置环境后再尝试

bazel clean --expunge

效果

image.png