1

TensorFlow-lite is for embedded linux system and Android NNAPI is for hw acceleration for Android system. I do like to find out solutions for embedded linux not android with Tensorflow-Lite hw acceleration.

K Lee
  • 21
  • 2
  • any updates? I want to use nnapi on android 7 & 8, but it is only available on 8.1 – xhsoldier Jun 06 '18 at 07:40
  • Embedded platforms are very specific and will likely require specific accelerators. You should elucidate what exact platforms you want acceleration for. – aselle Feb 08 '19 at 17:05

1 Answers1

1

NNAPI is the only hardware acceleration available by Tensorflow-Lite at this point. That being said, Tensorflow-Lite framework is designed to be easily extendable with custom Ops that utilize hardware acceleration. Also, tensorflow lite builtin ops already utilize specific CPU instructions (ARM NEON, Intel SSE) to speed up the inference.

Is there a particular embedded linux system you are targeting?

J.L.
  • 134
  • 2
  • I guess the question was how to use NNAPI on Linux. I think NNAPI exists only on Android 8.1+ (but not on Linux). TFLite works on both Android and Linux. However TFLite + NNAPI combination is only available on Android. Correct me if I'm wrong. https://www.tensorflow.org/lite/performance/hexagon_delegate – x4444 Jun 11 '20 at 21:38