| # The Arm NN TensorFlow Lite delegate JNI (Experimental) |
| |
| NOTE: This library is an experimental feature. We cannot guarentee full support for this. |
| |
| 'armnnDelegateJNI' is a library for accelerating certain TensorFlow Lite operators on Arm hardware specifically through Android |
| applications. Each release is packaged in an AAR which can be found on Maven Central. |
| The pre-built library contains the ArmNN Core, ArmNN Utils, Neon backend, CL Backend, and the ArmNN Delegate. |
| It is essential to only build these. The backends you choose are optional. |
| |
| It requires a static build which can be switched on through setting BUILD_SHARED_LIBS=OFF. You will also have to set |
| CMAKE_ANDROID_STL_TYPE=c++_static when building ArmNN. |
| |
| BUILD_DELEGATE_JNI_INTERFACE will also have to be set to true. |
| |
| To download the prebuilt ArmNN Delegate JNI AAR from Maven Central, please go to [ArmNN Maven Central Release Page](https://search.maven.org/artifact/io.github.arm-software/armnn.delegate). |
