Fork of offical Tensorflow repository with support for building tensorflow lite for roboVM-iOS

Prerequisites

• macOS
• bazel (brew install bazel)
• numpy (pip3 install numpy)

Building tensorflow lite for RoboVM iOS Project

First you have to initialize the project via configure which lies in root of the repository. Since we need tensorflow lite with java-API we can use the same rule, which is used for building tensorflow lite for android. Execute follwing command from repository root directory to create library for device:

bazel build --config=ios_arm64 tensorflow/lite/java:tensorflowlitelib

and for simulator:

bazel build --config=x_86_64 tensorflow/lite/java:tensorflowlitelib

Both command create libtensorflowlite_jni.so in bazel-bin/tensorflow/lite/java/

If you want to create a fat binary which contains both libraries for x_86_64 and arm64 you have to save the first created library and then glue them together via lipo command.