Giters

tensorflow / serving

A flexible, high-performance serving system for machine learning models

Home Page:https://www.tensorflow.org/serving

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Unable to compile prediction_service.proto for Golang

0110G opened this issue · comments

commented

Bug Report

If this is a bug report, please fill out the following form in full:

System information

  • **OS Platform and Distribution : macOS Bigsur 11.4
  • TensorFlow Serving installed from (source or binary): Source (Using Docker pull tensorflow/serving)
  • TensorFlow Serving version: 2.13.0

Describe the problem

I want to serve realtime predictions from a pretrained model using tf serve. While I am able to use RESTful APIs and getting correct predictions, I am not able to use gRPC methods for optimising my response times.

Exact Steps to Reproduce

I tried following these approaches before raising this issue:

  1. #634
  2. #1365
  3. #378

To be precise, I am doing the following:

  1. Cloning tensorflow serve repo.
  2. Cloning tensorflow repo
  3. Creating output directory named vendor.
    My directory structure after step 3 looks like:
  - serving:
     - tensorflow_serving
        - apis
  - tensorflow:
     - tensorflow:
       - core
         -...
  - vendor:
  1. Trying to generate Go file using protoc -I tensorflow -I serving --go_out=vendor --go_opt=paths=source_relative --go-grpc_out=vendor --go-grpc_opt=paths=source_relative serving/tensorflow_serving/apis/prediction_service.proto

protoc --version = libprotoc 24.3

Source code / logs

I am getting the following error:

protoc-gen-go: unable to determine Go import path for "tensorflow_serving/apis/input.proto"

Please specify either:
	• a "go_package" option in the .proto source file, or
	• a "M" argument on the command line.

See https://developers.google.com/protocol-buffers/docs/reference/go-generated#package for more information.

I tried using M argument as specified but still faced the same issue for some other file.
Please help

commented

@0110G,

Can you try using buf’s managed mode to build the TF serving proto files as shown in TF Forum thread. Thank you!

commented

Checking this out @singhniraj08 Will update this thread.

commented

Getting tensorflow/core/protobuf/rpc_options.proto:5:15:tsl/protobuf/rpc_options.proto: does not exist while following TF Forum thread

commented

@0110G In case you are still stuck I have a Bazel based minimal proto library in https://github.com/PantaPasen/tensorflow-serving-apis-proto/tree/2.15, adding Bazel go proto libs on top of it or a buf generation should be fairly easy.