WasmEdge / WasmEdge

WasmEdge is a lightweight, high-performance, and extensible WebAssembly runtime for cloud native, edge, and decentralized applications. It powers serverless apps, embedded functions, microservices, smart contracts, and IoT devices.

Home Page:https://WasmEdge.org

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

feat: `wasmedge_tensorflowlite` plug-in support for Android platform.

mtobin-tdab opened this issue · comments

Summary

Given that WasmEdge-tensorflow was migrated to the wasmedge_tensorflowlite and supported Android, is there a plan to add Android support to the plugin?

Changlog ref

Details

Would basically use the previous WasmEdge-tensorflow implementation to make the tensorflow-lite plugin available on Android

Appendix

No response

@q82419
I found we have the CI and release workflow for the Android support. Do you think we should also add it to our current workflow?

https://github.com/second-state/WasmEdge-tensorflow/blob/master/.github/workflows/release.yml#L273-L318

Well, the following steps are necessary:

  • Check the plug-in system available on WasmEdge for Android.
  • Build the new TensorFlow-Lite shared library for Android.
  • Test on an Android device.

Since we don't have the engineering power and hardware resources to support this plugin on the Android platform, I'll mark this issue as a help-wanted issue and leave it to the community.

In vcpkg, there is a package named plugin-wasi-nn-backend-tensorflowlite, which can be successfully built for arm64-android. Does the wasmedge_tensorflowlite plugin differ from this one?

patch vcpkg/ports/wasmedge/vcpkg.json file features.plugins.suppoorts and features.plugin-wasi-nn-backend-tensorflowlite.suppoorts

then

$ vcpkg install 'wasmedge[core,tools,plugins,plugin-wasi-nn-backend-tensorflowlite]':arm64-android

@hangedfish
That's the same one.
You can choose the different backend by the options.
https://github.com/microsoft/vcpkg/blob/master/ports/wasmedge/portfile.cmake#L35-L50