How to do a TensorFlow Serving Client?

[Brando Os]

Hi TensorFlow Team!
I write to you because I have a problem when I configure my TensorFlow Serving.
I need help to Configure my client into an andorid app.
I have researched in to web, but I didn't find anything about that.
Where could I find information about how to do it?

Sincerely, Brando Os

[Jorge Muñoz]

Hi,

It is not straight forward but it is not complicated. You only need to know that the requests to your server are done with grpc and protobuf. You already have the protobuf configuration for your sever, you need to create the stub for the client with grpc:

http://www.grpc.io/

https://github.com/grpc/grpc-java

http://www.grpc.io/docs/tutorials/basic/java.html

Basically you need to compile the protocol buffer definition for java using the pluging for grpc (you probably need to compile grpc with support for java in order to create the plugin first).

Take a look first the the examples for python in the repository. The idea is the same bug using the plugin of grpc for java.

[madhusud...@gmail.com]

hi,

I'm very new to TF..

I have successfully deployed one trained inception model in serving. But not able to call from java.

It shows "Exception in thread "main" io.grpc.StatusRuntimeException: UNIMPLEMENTED". Then I have found I need to generate one stub for there. Could you please share steps or generation of stubs.