I am trying to install TensorFlow-gpu with REST API in Centos 7 Docker container. But I am unable to find an exact procedure for this. Do I need to install following dependencies?
I have installed cuda 9.0
cdDNN 7.4
NCCL 2.x
I haven't started yet to build tensorflow serving using GPU. I m in the middle of research stage ~ in this process in every article showing related to Ubuntu installation and m trying to install.in centos 7 .. so I don't have any docker file ..
Hope this may help you and me to get solution .
Here is what I use to build a tensorflow-serving-runtime docker image.
FROM nvidia/cuda:9.0-cudnn7-runtime-centos7
ARG TF_VERSION=1.9.0
RUN yum install -y \
yum-plugin-ovl \
libgomp \
ca-certificates \
zip \
unzip \
curl \
&& \
yum clean all
WORKDIR /usr/
RUN curl -sSL -o /usr/nccl_2.2.13-1-cuda9.0_x86_64.tgz http://some-of-my-net-disk/tensorflow-serving/lib/nccl_2.2.13-1-cuda9.0_x86_64.tgz && \ # Change your way to get nccl library here
tar -xvf nccl_2.2.13-1-cuda9.0_x86_64.tgz &&\
rm -f nccl_2.2.13-1-cuda9.0_x86_64.tgz
ENV LD_LIBRARY_PATH /usr/nccl_2.2.13-1+cuda9.0_x86_64/lib/:${LD_LIBRARY_PATH}
# Change your way to get tensorflow_model_server here
WORKDIR /serving
RUN curl -sSL -o /usr/local/bin/tensorflow_model_server http://some-of-my-net-disk/tensorflow-serving/bin/tf-serving-${TF_VERSION}/tensorflow_model_server_gpu-centos &&\
chmod u+x /usr/local/bin/tensorflow_model_server
For me, this worked fine. Hope it helps.
Related
I am trying to rewrite a Dockerfile (https://github.com/orangefoil/rcssserver-docker/blob/master/Dockerfile) so that it uses alpine instead of ubuntu. Goal is to reduce the file size.
In the original image the robocup soccer server is built from scratch using g++, flex, bison, etc.
FROM ubuntu:18.04 AS build
ARG VERSION=16.0.0
WORKDIR /root
RUN apt update && \
apt -y install autoconf bison clang flex libboost-dev libboost-all-dev libc6-dev make wget
RUN wget https://github.com/rcsoccersim/rcssserver/archive/rcssserver-$VERSION.tar.gz && \
tar xfz rcssserver-$VERSION.tar.gz && \
cd rcssserver-rcssserver-$VERSION && \
./bootstrap && \
./configure && \
make && \
make install && \
ldconfig
I tried to do the same on alpine and had to exchange some packages:
FROM alpine:latest
ARG VERSION=16.0.0
WORKDIR /root
# Add basics first
RUN apk — no-cache update \
&& apk upgrade \
&& apk add autoconf bison clang-dev flex-dev boost-dev make wget automake libtool-dev g++ build-base
RUN wget https://github.com/rcsoccersim/rcssserver/archive/rcssserver-$VERSION.tar.gz
RUN tar xfz rcssserver-$VERSION.tar.gz
RUN cd rcssserver-rcssserver-$VERSION && \
./bootstrap && \
./configure && \
make && \
make install && \
ldconfig
Unfortunately, my version doesn't work yet. It fails with
/usr/lib/gcc/x86_64-alpine-linux-musl/9.3.0/../../../../x86_64-alpine-linux-musl/bin/ld: cannot find -lrcssclangparser
From what I found so far, this can happen, if dev packages are not installed (see ld cannot find an existing library), but I changed to dev packages where I could find them and still no luck.
So, my current assumption is that ubuntu has some package installed, that I need to add in my alpine image. I would exclude a code problem, since the ubuntu version works.
Any ideas, what could be missing? I would also be happy to understand how to compare the packages myself, but the package namings are not the same in ubuntu and alpine, so I find it pretty hard to figure this out.
You should break this up using a multi-stage build. In the image you're building now, the final image contains the C toolchain and all of the development libraries and headers that those -dev packages install; you don't need any of those to actually run the built application. The basic idea is to build the application exactly as you have it now, but then COPY only the built application into a new image with fewer dependencies.
This would look something like this (untested):
FROM ubuntu:18.04 AS build
# ... exactly what's in the original question ...
FROM ubuntu:18.04
# Install the shared libraries you need to run the application,
# but not -dev headers or the full C toolchain. You may need to
# run `ldd` on the built binary to see what exactly it needs.
RUN apt-get update \
&& DEBIAN_FRONTEND=noninteractive \
apt-get install --assume-yes --no-install-recommends \
libboost-atomic1.65.1 \
libboost-chrono1.65.1 \
# ... more libboost-* libraries as required ...
# Get the built application out of the original image.
# Autoconf's default is to install into /usr/local, and in a
# typical Docker base image nothing else will be installed there.
COPY --from=build /usr/local /usr/local
RUN ldconfig
# Describe how to run a container.
EXPOSE 12345
CMD ["/usr/local/bin/rcssserver"]
Compared to the size of the C toolchain, header files, and build-time libraries, the difference between an Alpine and Ubuntu image is pretty small, and Alpine has well-documented library compatibility issues with its minimal libc implementation.
I'm running into an issue with a custom Docker image. I've installed a number of tools and all seem to be working except for the AWSCLI.
I install here:
RUN apt-get install -y \
python \
python-pip \
groff \
less \
mime-support \
&& \
pip install --upgrade awscli==1.14.5 s3cmd==2.0.1 python-magic && \
apt-get -v del python-pip && \
rm -rf /var/cache/apt/*
VOLUME /root/.aws
Which installs successfully, I even ran aws --version to confirm no errors. Then when running in .gitlab-ci.yml aws is not recognized but my other tools are.
Here is the command I'm running:
aws ec2 describe-instances --filters "Name=tag:Project,Values=" --region us-east-2 --query "Reservations[].Instances[].[PrivateIpAddress]" --output=text
This is the error I get:
/bin/sh: eval: line 132: aws: not found
Core of your problem is same as in this question:
awscli not added to path after installation
Python of specific version was installed and it's /bin folder is not in system executable path. You need to add your Python version to system PATH:
ENV PATH "$PATH:/Library/Frameworks/Python.framework/Versions/3.8/bin"
Other variant: install only py-pip and it will fetch python and install aws globally, DO not remove py-pip after, or it will clean references to aws.
When I use curl --head to test my website, it returns the server information.
I followed this tutorial to hide the nginx server header.
But when I run the command yum install nginx-module-security-headers
, it returns yum: not found.
I also tried apk add nginx-module-security-headers, and it shows that the package is missing.
I have used nginx:1.17.6-alpine as my base docker image. Does anyone know how to hide the server from header under this Alpine?
I think I have an easier solution here: https://gist.github.com/hermanbanken/96f0ff298c162a522ddbba44cad31081. Big thanks to hermanbanken on Github for sharing this gist.
The idea is to create a multi stage build with the nginx alpine image to be a base for compiling the module. This turns into the following Dockerfile:
ARG VERSION=alpine
FROM nginx:${VERSION} as builder
ENV MORE_HEADERS_VERSION=0.33
ENV MORE_HEADERS_GITREPO=openresty/headers-more-nginx-module
# Download sources
RUN wget "http://nginx.org/download/nginx-${NGINX_VERSION}.tar.gz" -O nginx.tar.gz && \
wget "https://github.com/${MORE_HEADERS_GITREPO}/archive/v${MORE_HEADERS_VERSION}.tar.gz" -O extra_module.tar.gz
# For latest build deps, see https://github.com/nginxinc/docker-nginx/blob/master/mainline/alpine/Dockerfile
RUN apk add --no-cache --virtual .build-deps \
gcc \
libc-dev \
make \
openssl-dev \
pcre-dev \
zlib-dev \
linux-headers \
libxslt-dev \
gd-dev \
geoip-dev \
perl-dev \
libedit-dev \
mercurial \
bash \
alpine-sdk \
findutils
SHELL ["/bin/ash", "-eo", "pipefail", "-c"]
RUN rm -rf /usr/src/nginx /usr/src/extra_module && mkdir -p /usr/src/nginx /usr/src/extra_module && \
tar -zxC /usr/src/nginx -f nginx.tar.gz && \
tar -xzC /usr/src/extra_module -f extra_module.tar.gz
WORKDIR /usr/src/nginx/nginx-${NGINX_VERSION}
# Reuse same cli arguments as the nginx:alpine image used to build
RUN CONFARGS=$(nginx -V 2>&1 | sed -n -e 's/^.*arguments: //p') && \
sh -c "./configure --with-compat $CONFARGS --add-dynamic-module=/usr/src/extra_module/*" && make modules
# Production container starts here
FROM nginx:${VERSION}
COPY --from=builder /usr/src/nginx/nginx-${NGINX_VERSION}/objs/*_module.so /etc/nginx/modules/
.... skipped inserting config files and stuff ...
# Validate the config
RUN nginx -t
Alpine repo probably doesn't have the ngx_security_headers module but, the mentioned tutorial also provides an option of using Headers More module. You should be able to install this module in your alpine distro using the command:
apk add nginx-mod-http-headers-more
Hope it helps.
Source
I found the alternate solution. The reason that it shows binary not compatible is because I have one nginx pre-installed under the target route, and it is not compatible with the header-more module I am using. That means I cannot simply install the third party library from Alpine package.
So I prepare a clean Alpine OS, and follow the GitHub repository to build Nginx from the source with additional feature. The path of build result is the prefix path you specified.
I have been experimenting to create a docker image with python3.6 based on amazonlinux.
So far, I have not been very successful. I use
docker run -it amazonlinux
to start an interactive docker terminal. Inside the terminal, I run "yum install python36" and see the following error message. Note that I copied this step was from an old amazonlinux based Dockerfile. This Dockerfile used to work. So I suspend the error I see below is due to amazon updated their docker linux image
bash-4.2# yum install python36
Loaded plugins: ovl, priorities
amzn2-core | 2.4 kB 00:00:00
No package python36 available.
Error: Nothing to do
I have tried to add a python3.6 repo by following this post
https://janikarhunen.fi/how-to-install-python-3-6-1-on-centos-7 however, it still gives the same error when I run
yum install python36u
Is there any way to add python3.6 to amazonlinux base layer? Thanks in advance.
There is now a far easier answer to this question thanks to aws 'extras'. Now this will work:
amazon-linux-extras install python3
You can check this Dockerfile based on amazon Linux and having python version is PYTHON_VERSION=3.6.4.
Or you can work with your existing one like
ARG PYTHON_VERSION=3.6.4
ARG BOTO3_VERSION=1.6.3
ARG BOTOCORE_VERSION=1.9.3
ARG APPUSER=app
RUN yum -y update &&\
yum install -y shadow-utils findutils gcc sqlite-devel zlib-devel \
bzip2-devel openssl-devel readline-devel libffi-devel && \
groupadd ${APPUSER} && useradd ${APPUSER} -g ${APPUSER} && \
cd /usr/local/src && \
curl -O https://www.python.org/ftp/python/${PYTHON_VERSION}/Python-${PYTHON_VERSION}.tgz && \
tar -xzf Python-${PYTHON_VERSION}.tgz && \
cd Python-${PYTHON_VERSION} && \
./configure --enable-optimizations && make && make altinstall && \
rm -rf /usr/local/src/Python-${PYTHON_VERSION}* && \
yum remove -y shadow-utils audit-libs libcap-ng && yum -y autoremove && \
yum clean all
But better to clone the repo and make your own image form that.
I too had similiar issue for docker.
yum install docker
Loaded plugins: ovl, priorities
amzn2-core | 3.7 kB 00:00:00
No package docker available.
Error: Nothing to do
instead yum I used amazon-linux-extras, it worked
amazon-linux-extras install docker
we are trying to host tensorflow object-detection model on GCP.
we have maintain below directory structure before running "gcloud app deploy".
For you convenient I am attaching the configuration files with the question.
Wer are getting deployment error which is mentioned below. Please suggest a solution.
+root
+object_detection/
+slim/
+env
+app.yaml
+Dockerfile
+requirement.txt
+index.html
+test.py
Dockerfile
FROM gcr.io/google-appengine/python
LABEL python_version=python2.7
RUN virtualenv --no-download /env -p python2.7
# Set virtualenv environment variables. This is equivalent to running
# source /env/bin/activate
ENV VIRTUAL_ENV /env
ENV PATH /env/bin:$PATH
# Various Python and C/build deps
RUN apt-get update && apt-get install -y \
wget \
build-essential \
cmake \
git \
unzip \
pkg-config \
python-dev \
python-opencv \
libopencv-dev \
libav-tools \
libjpeg-dev \
libpng-dev \
libtiff-dev \
libjasper-dev \
libgtk2.0-dev \
python-numpy \
python-pycurl \
libatlas-base-dev \
gfortran \
webp \
python-opencv \
qt5-default \
libvtk6-dev \
zlib1g-dev \
protobuf-compiler \
python-pil python-lxml \
python-tk
# Install Open CV - Warning, this takes absolutely forever
ADD requirements.txt /app/
RUN pip install -r requirements.txt
ADD . /app/
RUN protoc /app/object_detection/protos/*.proto --python_out=/app/.
RUN export PYTHONPATH=$PYTHONPATH:`pwd`:`pwd`/app/slim
CMD exec gunicorn -b :$PORT UploadTest:app
requirement.txt
Flask==0.12.2
gunicorn==19.7.1
numpy==1.13.1
requests==0.11.1
bs4==0.0.1
nltk==3.2.1
pymysql==0.7.2
xlsxwriter==0.8.5
Pillow==4.2.1
pytesseract==0.1
opencv-python>=3.0
matplotlib==2.0.2
tensorflow==1.3.0
lxml==4.0.0
app.yaml
runtime: custom
env: flex
entrypoint: gunicorn -b :$PORT UploadTest:app
threadsafe: true
runtime_config:
python_version: 2
After all this i am seeting up the google cloud environment with gcloud init
And then start command gcloud app deploy
I am getting below error while deploying the solution.
Error:
Step 10/12 : RUN protoc /app/object_detection/protos/*.proto --python_out=/app/.
---> Running in 9b3ec9c43c2d
/app/object_detection/protos/anchor_generator.proto: File does not reside within any path specified using --proto_path (or -I). You must specify a --proto_path which encompasses this file. Note that the proto_path must be an exact prefix of the .proto file names -- protoc is too dumb to figure out when two paths (e.g. absolute and relative) are equivalent (it's harder than you think).
The command '/bin/sh -c protoc /app/object_detection/protos/*.proto --python_out=/app/.' returned a non-zero code: 1
ERROR
ERROR: build step "gcr.io/cloud-builders/docker#sha256:a4a83be9b2fb61452e864ecf1bcfca99d1845499ef9500ae2905cea0ea593769" failed: exit status 1
----------------------------------------------------------------------------------------------------------------------------------------------
ERROR: (gcloud.app.deploy) Cloud build failed. Check logs at https://console.cloud.google.com/gcr/builds/4dba3217-b7d6-4341-b28e-09a9dad45c18?
There is a directory "object_detection/protos" present and all necessary files are present there. Still getting deployment error. Please suggest where to change in dockerfile to deploy it successfully.
My assumption: GCP is not able to figure out the path of the protc file. May be I have to alter something in Docketfile. But not able to figure out the solution. Please answer.
NB: This setup is running well in local machine. But not working in GCP