TF 2.4.1 and gRPC

[ideenfix]

TF 2.4.1 and gRPC

[ideenfix]

Succes. gRPC run with Tensorflow 2.4.1. Attached the Dockerfile. I don't know exactly whether some steps unnecessary but now it's run.
Dockerfile.grpc241.txt

[brieyla1]

hey !

tensorflow custom ops will only be giving Symbol errors when the 2.X build is different, you can therefore use this docker image:
FROM tensorflow/tensorflow:2.4.0-custom-op-gpu-ubuntu16

and everything should work as expected.

I have modified my WORKSPACE aswell, as I had problems with the vanilla one:

load("//tf:tf_configure.bzl", "tf_configure")
load("//gpu:cuda_configure.bzl", "cuda_configure")

tf_configure(name = "local_config_tf")

load("@bazel_tools//tools/build_defs/repo:http.bzl", "http_archive")

http_archive(
    name = "com_github_grpc_grpc",
    urls = [
        "https://github.com/grpc/grpc/archive/ac1c5de1b36da4a1e3d72ca40b0e43f24266121a.tar.gz",
    ],
    strip_prefix = "grpc-ac1c5de1b36da4a1e3d72ca40b0e43f24266121a",
)

http_archive(
    name = "com_google_protobuf",
    strip_prefix = "protobuf-3.9.0",
    urls = [
        "https://github.com/protocolbuffers/protobuf/releases/download/v3.9.0/protobuf-all-3.9.0.tar.gz",
    ],
)

load("@com_github_grpc_grpc//bazel:grpc_deps.bzl", "grpc_deps")
grpc_deps()
load("@com_github_grpc_grpc//bazel:grpc_extra_deps.bzl", "grpc_extra_deps")
grpc_extra_deps()


http_archive(
    name = "com_google_protobuf",
    strip_prefix = "protobuf-3.9.1",
    urls = [
        "https://github.com/protocolbuffers/protobuf/releases/download/v3.9.1/protobuf-all-3.9.1.tar.gz",
    ],
)

PS: don't forget to install tensorflow 2.4.1 for the testing:
RUN pip3 install tensorflow-gpu==2.4.1

have a good day.

[VegeWong]

For users having TF 2.4.1 and CUDA 11 dependencies and network issues accessing google-hosted repos and sources, the PR #66 may help.

[lespeholt]

I updated the repository to 2.4.1.