container-images VS zed-docker

Note you should select the NVIDIA Docker image that matches your CUDA driver version. Look in the unsupported list if your driver version is older.

FROM tensorflow/tensorflow:1.15.5-gpu-py3 # Handle Nvidia public key update and update repositories for Ubuntu 18.x. #https://github.com/sangyun884/HR-VITON/issues/45 # reference: https://jdhao.github.io/2022/05/05/nvidia-apt-repo-public-key-error-fix/ RUN rm /etc/apt/sources.list.d/cuda.list RUN rm /etc/apt/sources.list.d/nvidia-ml.list RUN apt-key del 7fa2af80 # Additional reference: https://gitlab.com/nvidia/container-images/cuda/-/issues/158 RUN export this_distro="$(cat /etc/os-release | grep '^ID=' | awk -F'=' '{print $2}')" \ && export this_version="$(cat /etc/os-release | grep '^VERSION_ID=' | awk -F'=' '{print $2}' | sed 's/[^0-9]*//g')" \ && apt-key adv --fetch-keys "https://developer.download.nvidia.com/compute/cuda/repos/${this_distro}${this_version}/x86_64/3bf863cc.pub" \ && apt-key adv --fetch-keys "https://developer.download.nvidia.com/compute/machine-learning/repos/${this_distro}${this_version}/x86_64/7fa2af80.pub" # get the latest version of OpenCV RUN apt-get update && \ DEBIAN_FRONTEND=noninteractive \ apt-get install -y -qq \ wget git libopencv-dev RUN python -m pip install --upgrade pip && \ pip install matplotlib opencv-python==4.5.4.60 Pillow scipy \ azure-eventhub azure-eventhub-checkpointstoreblob-aio ipykernel WORKDIR /

As far as I can see, the way Google Chrome developers chose to support hardware acceleration under Linux is through Vulkan (here and here) According to Nvidia, there's no official support for Vulkan inside Docker. Although it seems that FAQ hasn't been updated because I was able to find a Docker container with Vulkan support here.

Looking forward to the CUDA containers getting released!

That leaves Flatcar and Clear Linux - both of which happen to at least have documentation for installing/running Nvidia drivers and CUDA. Flatcar has this repository from Nvidia, and I've also found this project called forklift which will supposedly handle auto-updating the kernel modules for you. The Clear Linux docs also seem to include a method to auto-rebuild the modules with kernel upgrades, though it does say that the driver version needs to be updated manually, which honestly almost sounds preferable considering how finicky Nvidia drivers can be on Linux. Clear Linux also has several other tutorials/guides that appear to try and market it for things like machine learning, which leads me to believe that Nvidia gpus would hopefully work decently on it.

To add CUDA I plan on adding the stuff from this Docker script.

You can use the cuda dockerfile as reference: https://gitlab.com/nvidia/container-images/cuda/-/blob/master/Dockerfile

However, getting Nvidia to work was much more complicated. Now, I am not a container expert, so a lot of it was because of my unfamiliarity with the technology. At first, I had to get nvidia-container-toolkit using CentOS package. The test containers given in the instructions here worked fine. However, I soon understood that nvidia-container-toolkit requires basing the image on nvidia official containers or going through this and figure out how to create a custom container. Most documentation online seemed to be about nvidia-docker or just covered the install portion of nvidia-container-toolkit. There was almost nothing available on how to create a custom image. After some digging around and copying and pasting (I still don't understand some of it), I was able to create a container with nvidia-smi, and other cuda commands working.

https://gitlab.com/nvidia/container-images/cuda/-/issues/109#note_503061879