How Can I use onnxruntime_server work on CUDA backend.

[TaylorHere]

I build the onnxruntime_server binary from master branch within a container build from Dockerfile.ubuntu_server
follow by Server build Instructions
then I deploy the binary as

FROM nvidia/cuda:10.1-cudnn7-runtime-ubuntu16.04

RUN echo " " > /etc/apt/sources.list && \
    echo "deb http://mirrors.aliyun.com/debian buster main" >> /etc/apt/sources.list && \
    echo "deb http://mirrors.aliyun.com/debian buster-updates main" >> /etc/apt/sources.list && \
    apt-get update && apt-get install -y --allow-unauthenticated\
        software-properties-common \
        python3-pip \
        htop

RUN pip3 install --upgrade pip
RUN pip3 install onnxruntime-gpu==1.4.0

ADD onnxruntime_server /usr/local/bin/
ADD libonnxruntime.so.1.7.0 /usr/lib/libonnxruntime.so.1.7.0
ADD libre2.so.1v5 /usr/lib/x86_64-linux-gnu/libre2.so.1v5
ADD libcares.so.2 /usr/local/lib/libcares.so.2

But I found that there is no GPU memory or GPU-Util usage at all, looks like it's not working with a CUDA environment

Also I tried to use -DUSE_CUDA=1 flag to build the binary, but I got output like:

CMake Warning:
  Manually-specified variables were not used by the project:

    USE_CUDA

How can I enable the CUDA support?

[TaylorHere]

Looks like I built version 1.7.0 and use a lower CUDA version, I'm trying to use the document referred version of CUDA and cuDNN

[TaylorHere]

it's not working, I still got this from server starting

[2021-03-18 13:11:34.910] [ServerApp] [info] [ onnxruntime inference_session.cc:1132 Initialize]: Adding default CPU execution provider.

[pranavsharma]

onnxruntime server is deprecated. We don't support it any more.