Loading README.md +2 −0 Original line number Diff line number Diff line Loading @@ -275,3 +275,5 @@ and log in again. - You might run into an issue where Podman fails saying `short-name-aliases.conf.lock` is on a read-only file system. This can be fixed by setting `export XDG_CACHE_HOME=/gpfs/alpine/stf007/scratch/<username>/podman-cache` or in a similar writable location on the filesystem. - It seems like we have to do a `rm -rf /var/cache/dnf` before every time you do a dnf install because there seems to be an issue where dnf tries to create a directory that already exists in the cache, causing the container build to fail. baseimage-ubuntu-cuda/Dockerfile 0 → 100644 +16 −0 Original line number Diff line number Diff line FROM docker.io/nvidia/cuda:11.0.3-cudnn8-devel-ubuntu18.04 RUN apt-get update -y && apt-get install -y fakeroot RUN apt-get install -y git dbus # && dnf -y install tar wget git openssh ENV LD_LIBRARY_PATH=/usr/local/cuda/lib64:/usr/local/cuda/compat:$LD_LIBRARY_PATH ENV CUDA_HOME="/usr/local/cuda" ENV CUDA_ROOT="/usr/local/cuda" ENV CUDNN_ROOT="/usr" ENV CUDNN_INCLUDE_DIR="${CUDNN_ROOT}/include" ENV CUDNN_LIB_DIR="${CUDNN_ROOT}/lib" ENTRYPOINT ["/bin/bash"] baseimage-ubuntu-cuda/README.md 0 → 100644 +2 −0 Original line number Diff line number Diff line This needs to be built on a raptor so that you can successfully install fakeroot first. Ideally any image built on top of this can do a fakeroot apt-get install? baseimage-ubuntu-cuda/build.sh 0 → 100755 +5 −0 Original line number Diff line number Diff line #!/bin/bash # TODO: the -t should be code.ornl.gov/baseimages/baseimage-centos-cuda env TMPDIR=/projects/tmp podman build -v /projects/tmp:/tmp --ulimit nofile=262144:262144 -f Dockerfile -t code.ornl.gov:4567/76a/olcfbaseimages/baseimage-ubuntu-cuda:latest . podman push code.ornl.gov:4567/76a/olcfbaseimages/baseimage-ubuntu-cuda:latest Loading
README.md +2 −0 Original line number Diff line number Diff line Loading @@ -275,3 +275,5 @@ and log in again. - You might run into an issue where Podman fails saying `short-name-aliases.conf.lock` is on a read-only file system. This can be fixed by setting `export XDG_CACHE_HOME=/gpfs/alpine/stf007/scratch/<username>/podman-cache` or in a similar writable location on the filesystem. - It seems like we have to do a `rm -rf /var/cache/dnf` before every time you do a dnf install because there seems to be an issue where dnf tries to create a directory that already exists in the cache, causing the container build to fail.
baseimage-ubuntu-cuda/Dockerfile 0 → 100644 +16 −0 Original line number Diff line number Diff line FROM docker.io/nvidia/cuda:11.0.3-cudnn8-devel-ubuntu18.04 RUN apt-get update -y && apt-get install -y fakeroot RUN apt-get install -y git dbus # && dnf -y install tar wget git openssh ENV LD_LIBRARY_PATH=/usr/local/cuda/lib64:/usr/local/cuda/compat:$LD_LIBRARY_PATH ENV CUDA_HOME="/usr/local/cuda" ENV CUDA_ROOT="/usr/local/cuda" ENV CUDNN_ROOT="/usr" ENV CUDNN_INCLUDE_DIR="${CUDNN_ROOT}/include" ENV CUDNN_LIB_DIR="${CUDNN_ROOT}/lib" ENTRYPOINT ["/bin/bash"]
baseimage-ubuntu-cuda/README.md 0 → 100644 +2 −0 Original line number Diff line number Diff line This needs to be built on a raptor so that you can successfully install fakeroot first. Ideally any image built on top of this can do a fakeroot apt-get install?
baseimage-ubuntu-cuda/build.sh 0 → 100755 +5 −0 Original line number Diff line number Diff line #!/bin/bash # TODO: the -t should be code.ornl.gov/baseimages/baseimage-centos-cuda env TMPDIR=/projects/tmp podman build -v /projects/tmp:/tmp --ulimit nofile=262144:262144 -f Dockerfile -t code.ornl.gov:4567/76a/olcfbaseimages/baseimage-ubuntu-cuda:latest . podman push code.ornl.gov:4567/76a/olcfbaseimages/baseimage-ubuntu-cuda:latest