Request deployment FSL on GPU

[vemonet]

Describe the application

FSL on GPU in encore project (email Sanne)

Latest official Nvidia docs: https://nvidia.github.io/nvidia-container-runtime

Example for a debian based Dockerfile:

RUN curl -s -L https://nvidia.github.io/nvidia-container-runtime/gpgkey | \
    apt-key add - \ &&
    distribution=$(. /etc/os-release;echo $ID$VERSION_ID) \ &&
    curl -s -L https://nvidia.github.io/nvidia-container-runtime/$distribution/nvidia-container-runtime.list | 
RUN apt-get update \ &&
    apt-get install -y nvidia-container-runtime

Build the FSL Dockerfile on GPU:

oc new-build --name fsl-gpu --binary
oc start-build fsl-gpu --from-dir=. --follow --wait
oc new-app fsl-gpu
oc pod edit fsl-gpu

See also, official Nvidia docs: https://docs.nvidia.com/cuda/cuda-installation-guide-linux/index.html#debian-installation

Application Git repository URL

Also using NeuroDocker FSL: https://github.com/ReproNim/neurodocker

Docker image of the application

# Generated by: Neurodocker version 0.7.0
# Latest release: Neurodocker version 0.7.0
# Timestamp: 2021/05/04 08:54:33 UTC
# 
# Thank you for using Neurodocker. If you discover any issues
# or ways to improve this software, please submit an issue or
# pull request on our GitHub repository:
# 
#     https://github.com/ReproNim/neurodocker

FROM debian:stretch

USER root

ARG DEBIAN_FRONTEND="noninteractive"

ENV LANG="en_US.UTF-8" \
    LC_ALL="en_US.UTF-8" \
    ND_ENTRYPOINT="/neurodocker/startup.sh"
RUN export ND_ENTRYPOINT="/neurodocker/startup.sh" \
    && apt-get update -qq \
    && apt-get install -y -q --no-install-recommends \
           apt-utils \
           bzip2 \
           ca-certificates \
           curl \
           locales \
           unzip \
    && apt-get clean \
    && rm -rf /var/lib/apt/lists/* \
    && sed -i -e 's/# en_US.UTF-8 UTF-8/en_US.UTF-8 UTF-8/' /etc/locale.gen \
    && dpkg-reconfigure --frontend=noninteractive locales \
    && update-locale LANG="en_US.UTF-8" \
    && chmod 777 /opt && chmod a+s /opt \
    && mkdir -p /neurodocker \
    && if [ ! -f "$ND_ENTRYPOINT" ]; then \
         echo '#!/usr/bin/env bash' >> "$ND_ENTRYPOINT" \
    &&   echo 'set -e' >> "$ND_ENTRYPOINT" \
    &&   echo 'export USER="${USER:=`whoami`}"' >> "$ND_ENTRYPOINT" \
    &&   echo 'if [ -n "$1" ]; then "$@"; else /usr/bin/env bash; fi' >> "$ND_ENTRYPOINT"; \
    fi \
    && chmod -R 777 /neurodocker && chmod a+s /neurodocker

## TODO: Trying to install GPU driver
RUN curl -s -L https://nvidia.github.io/nvidia-container-runtime/gpgkey | \
    apt-key add - \ &&
    distribution=$(. /etc/os-release;echo $ID$VERSION_ID) \ &&
    curl -s -L https://nvidia.github.io/nvidia-container-runtime/$distribution/nvidia-container-runtime.list | 
    # tee /etc/apt/sources.list.d/nvidia-container-runtime.list

RUN apt-get update \ &&
    apt-get install -y nvidia-container-runtime

ENTRYPOINT ["/neurodocker/startup.sh"]

ENV FSLDIR="/opt/fsl-6.0.3" \
    PATH="/opt/fsl-6.0.3/bin:$PATH" \
    FSLOUTPUTTYPE="NIFTI_GZ" \
    FSLMULTIFILEQUIT="TRUE" \
    FSLTCLSH="/opt/fsl-6.0.3/bin/fsltclsh" \
    FSLWISH="/opt/fsl-6.0.3/bin/fslwish" \
    FSLLOCKDIR="" \
    FSLMACHINELIST="" \
    FSLREMOTECALL="" \
    FSLGECUDAQ="cuda.q"
RUN apt-get update -qq \
    && apt-get install -y -q --no-install-recommends \
           bc \
           dc \
           file \
           libfontconfig1 \
           libfreetype6 \
           libgl1-mesa-dev \
           libgl1-mesa-dri \
           libglu1-mesa-dev \
           libgomp1 \
           libice6 \
           libxcursor1 \
           libxft2 \
           libxinerama1 \
           libxrandr2 \
           libxrender1 \
           libxt6 \
           sudo \
           wget \
    && apt-get clean \
    && rm -rf /var/lib/apt/lists/* \
    && echo "Downloading FSL ..." \
    && mkdir -p /opt/fsl-6.0.3 \
    && curl -fsSL --retry 5 https://fsl.fmrib.ox.ac.uk/fsldownloads/fsl-6.0.3-centos6_64.tar.gz \
    | tar -xz -C /opt/fsl-6.0.3 --strip-components 1 \
    && sed -i '$iecho Some packages in this Docker container are non-free' $ND_ENTRYPOINT \
    && sed -i '$iecho If you are considering commercial use of this container, please consult the relevant license:' $ND_ENTRYPOINT \
    && sed -i '$iecho https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/Licence' $ND_ENTRYPOINT \
    && sed -i '$isource $FSLDIR/etc/fslconf/fsl.sh' $ND_ENTRYPOINT \
    && echo "Installing FSL conda environment ..." \
    && bash /opt/fsl-6.0.3/etc/fslconf/fslpython_install.sh -f /opt/fsl-6.0.3

RUN echo '{ \
    \n  "pkg_manager": "apt", \
    \n  "instructions": [ \
    \n    [ \
    \n      "base", \
    \n      "debian:stretch" \
    \n    ], \
    \n    [ \
    \n      "fsl", \
    \n      { \
    \n        "version": "6.0.3" \
    \n      } \
    \n    ] \
    \n  ] \
    \n}' > /neurodocker/neurodocker_specs.json

[SanRut]

Dear Vincent,

I was wondering whether you have had any time to look at this again. Thanks for letting me know and let me know if you need any input from my side.

Best, Sanne

[vemonet]

Closing note: deploying FSL on GPU using CUDA worked well

You can find the Dockerfile and instructions to deploy on the DSRI here: https://github.com/MaastrichtU-IDS/jupyterlab/tree/main/fsl-gpu