Only ARM64 libraries provided into the tar. The all libs are on all iOS variations, including x86 simulator for your MAC are here: https://www.dropbox.com/s/9upcgpwzquber4p/tensorflow.tar.gz?dl=0 (github does not allow such large libraries)
-
[Prerequsite] Install xcode cli tools
$ xcode-select --install
-
[Prerequsite] Install automake and libtool
$ brew install automake libtool
-
Clone TensorFlow repository
$ cd ~/Code $ git clone https://github.com/tensorflow/tensorflow/
-
Clone TensorflowPod repository
$ cd ~/Code $ git clone https://github.com/evangel76/TensorFlow15CocoPod
Remove limitations for TensorFlow for iOS:
open build_all_ios.sh into tensorflow/tensorflow/contrib/makefile remove all instances of "ANDROID_TYPE" into the iOS section of the makefile, this will remove the limitation tensorflow has on operations in iOS. Notice that it will increase dramatically the size of the library. I recommand to only do this if you need advanced features like back propagation, or CNN/RNN, and you are targetting mostly ARM64 only.
-
Compile TensorFlow
$ cd ~/Code/tensorflow $ tensorflow/contrib/makefile/build_all_ios.sh
-
Pack necessary header and static library
$ cd ~/Code/tensorflow $ ~/Code/TensorflowPod/pack.sh $ ls tensorflow.tar.gz
-
[Prerequsite] Install Cocoapods
sudo gem install cocoapods
-
Create or add pod into your project
Podfile
, which should look like Don't forget to edit "DemoApp" as your own project name, otherwise, cocopod will fail to add it to your project.target 'DemoApp' do # Comment the next line if you're not using Swift and don't want to use dynamic frameworks use_frameworks! # Pods for DemoApp pod 'TensorflowPod', :podspec => 'https://raw.githubusercontent.com/evangel76/TensorFlow15CocoPod/blob/master/TensorflowPod.podspec' end
-
Install Pod
pod install
-
Follow official demo app to develop, link
#include "google/protobuf/message_lite.h" #include "tensorflow/core/framework/tensor.h" #include "tensorflow/core/public/session.h" tensorflow::GraphDef graph; ...
]