FedML icon indicating copy to clipboard operation
FedML copied to clipboard
verified publisher icon FedML-AI

develop the Android SDK for FedML-Mobile platform

Open FedML-AI-admin opened this issue 5 years ago • 4 comments

FedML-AI-admin avatar Aug 02 '20 05:08 FedML-AI-admin

@joyerf Please help this. Put the SDK to Java folder?

chaoyanghe avatar May 02 '22 06:05 chaoyanghe

@joyerf https://github.com/FedML-AI/FedML/tree/master/mobile/android

chaoyanghe avatar May 02 '22 07:05 chaoyanghe

@chaoyanghe @joyerf Appreciate all your great work on FedML.

  • What's the current status of Mobile support in FedML?
  • Will the code for Mobile be open sourced and if so is there an ETA?
  • How is code packaged and executed on mobile devices? Are you shipping Python code?

I saw a note in https://github.com/FedML-AI/FedML/issues/18#issuecomment-683939678 that the mobile server is ready.

The android directory is empty (https://github.com/FedML-AI/FedML/tree/master/mobile/android). The MLOps video tutorial however showed it working with Android. So is the code located somewhere else?

I'm trying to understand how FedML compares to TensorFlow Federated. If I understand correctly TFF uses TensorFlow as a runtime engine so computation gets shipped as TF graphs to be executed in the runtimes.

It looks like FedML is taking a different approach and shipping Python code. IIUC in the cross-silo use case it uses a docker container as the runtime. What happens in the mobile or IOT case?

jlewi avatar Jun 03 '22 01:06 jlewi

@chaoyanghe @joyerf Appreciate all your great work on FedML.

  • What's the current status of Mobile support in FedML?
  • Will the code for Mobile be open sourced and if so is there an ETA?
  • How is code packaged and executed on mobile devices? Are you shipping Python code?

I saw a note in #18 (comment) that the mobile server is ready.

The android directory is empty (https://github.com/FedML-AI/FedML/tree/master/mobile/android). The MLOps video tutorial however showed it working with Android. So is the code located somewhere else?

I'm trying to understand how FedML compares to TensorFlow Federated. If I understand correctly TFF uses TensorFlow as a runtime engine so computation gets shipped as TF graphs to be executed in the runtimes.

It looks like FedML is taking a different approach and shipping Python code. IIUC in the cross-silo use case it uses a docker container as the runtime. What happens in the mobile or IOT case?

@jlewi Hi, Thanks for your questions. Our latest Android SDK and IoT platform are released at:

https://github.com/FedML-AI/FedML/tree/master/android

https://github.com/FedML-AI/FedML/tree/master/iot

For more questions, please join our Slack (you can find the link at FedML.ai).

chaoyanghe avatar Aug 14 '22 19:08 chaoyanghe