experiments/STT-tensorflow
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Update GPU delegate documentation

Browse Source 
The delegate is no longer experimental.

PiperOrigin-RevId: 258578637
This commit is contained in:
Jared Duke 2019-07-17 09:00:31 -07:00 committed by TensorFlower Gardener
parent fce56a775d
commit 5841e35798
2 changed files with 15 additions and 9 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

12
tensorflow/lite/g3doc/performance/delegates.md
View File

@ -14,13 +14,15 @@ Running inference on compute-heavy machine learning models on mobile devices is
Instead of relying on the CPU, some devices have hardware accelerators, such as GPU or DSP, that allows for better performance and higher energy efficiency.
## Using the GPU delegate
## Using the experimental GPU delegate
TensorFlow Lite provides an experimental GPU delegate that can be used to accelerate models on devices that have a GPU available.
For an overview of the experimental GPU delegate, see [TensorFlow Lite on GPU](https://www.tensorflow.org/lite/performance/gpu_advanced). For step-by-step tutorials on using the GPU delegate with Android and iOS, see [TensorFlow Lite GPU Delegate Tutorial](https://www.tensorflow.org/lite/performance/gpu).
TensorFlow Lite provides a GPU delegate that can be used to accelerate models on
devices that have a GPU available.
For an overview of the GPU delegate, see
[TensorFlow Lite on GPU](https://www.tensorflow.org/lite/performance/gpu_advanced).
For step-by-step tutorials on using the GPU delegate with Android and iOS, see
[TensorFlow Lite GPU Delegate Tutorial](https://www.tensorflow.org/lite/performance/gpu).
## How do delegates work?

12
tensorflow/lite/g3doc/performance/gpu.md
View File

@ -1,7 +1,7 @@
# TensorFlow Lite GPU delegate
[TensorFlow Lite](https://www.tensorflow.org/lite) supports several hardware
accelerators. This document describes how to preview the experimental GPU backend using the
accelerators. This document describes how to use the GPU backend using the
TensorFlow Lite delegate APIs on Android and iOS.
GPUs are designed to have high throughput for massively parallelizable
@ -20,12 +20,16 @@ power and generate less heat than when the same task is run on CPUs.
## Demo App Tutorials
The easiest way to try out the experimental GPU delegate is to follow the below tutorials, which go through building our classification demo applications with GPU support. The GPU code is only binary for now; it will be open-sourced soon. Once you understand how to get our demos working, you can try this out on your own custom models.
The easiest way to try out the GPU delegate is to follow the below tutorials,
which go through building our classification demo applications with GPU support.
The GPU code is only binary for now; it will be open-sourced soon. Once you
understand how to get our demos working, you can try this out on your own custom
models.
### Android (with Android Studio)
For a step-by-step tutorial, watch the
[Experimental GPU Delegate for Android](https://youtu.be/Xkhgre8r5G0) video.
[GPU Delegate for Android](https://youtu.be/Xkhgre8r5G0) video.
Note: This requires OpenGL ES 3.1 or higher.
@ -59,7 +63,7 @@ run on the GPU.
### iOS (with XCode)
For a step-by-step tutorial, watch the
[Experimental GPU Delegate for iOS](https://youtu.be/a5H4Zwjp49c) video.
[GPU Delegate for iOS](https://youtu.be/a5H4Zwjp49c) video.
Note: This requires XCode v10.1 or later.