Issue on windows WSL 2 Ubuntu docker: nvidia.com/gpu=all is not an absolute path.

[wirytiox]

i am using docker desktop and WSL and i am triying to get this to run, i installed it following the steps in the linux installation guide and i keep getting this error:

sudo docker run --rm -it \
    --device nvidia.com/gpu=all \
    -e CUDA=true \
    docker.io/metal3d/xmrig

docker: nvidia.com/gpu=all is not an absolute path.
See 'docker run --help'.

when i take out the --device nvidia.com/gpu=all i see this message CUDA disabled (/xmrig/libxmrig-cuda.so: cannot open shared object file: No such file or directory) and it keeps repeating

[wirytiox]

i also tried:
wget https://developer.download.nvidia.com/compute/cuda/12.3.2/local_installers/cuda_12.3.2_545.23.08_linux.run
sudo sh cuda_12.3.2_545.23.08_linux.run

also https://docs.nvidia.com/cuda/wsl-user-guide/index.html#getting-started-with-cuda-on-wsl-2

wget https://developer.download.nvidia.com/compute/cuda/repos/wsl-ubuntu/x86_64/cuda-keyring_1.1-1_all.deb
sudo dpkg -i cuda-keyring_1.1-1_all.deb
sudo apt-get update
sudo apt-get -y install cuda-toolkit-12-3

it seems to instal A LOT of stuff yet still doesn't work

i also tried:

sudo docker run --name miner --rm -it\
    --gpus all \
    -e CUDA=true \
    -e POOL_URL=$POOL_URL \
    -e POOL_USER=$POOL_USER \
    -e POOL_PASS=$POOL_PASS \
    -e DONATE_LEVEL=$DONATE_LEVEL \
    -e OTHERS_OPTS="--tls" \
    -e COIN="XMR" \
    docker.io/metal3d/xmrig

and got:

docker: Error response from daemon: failed to create task for container: failed to create shim task: OCI runtime create failed: runc create failed: unable to start container process: error during container init: error running hook #0: error running hook: exit status 1, stdout: , stderr: Auto-detected mode as 'legacy'
nvidia-container-cli: mount error: file creation failed: /var/lib/docker/overlay2/75428c7181166d422dbf96ebf3c24f1e8bc0c73e36dfa50e7bdb7907b4efc668/merged/usr/lib/x86_64-linux-gnu/libnvidia-ml.so.1: file exists: unknown

i also chated with chatgpt and tried this:

sudo rm -f /etc/systemd/system/nvidia-container-runtime.service.d/*
sudo systemctl daemon-reload
sudo systemctl restart docker

[wirytiox]

i searched this thread and it seems related to the image: NVIDIA/nvidia-container-toolkit#289
not sure if something can be modified but i am still researching
i tried even more stuff, like building the dockerfile without the nvidia stuff:

FROM ubuntu:22.04 as build-cuda-plugin
LABEL maintainer="Patrice Ferlet <metal3d@gmail.com>"

ARG CUDA_PLUGIN_VERSION=6.17.0
RUN set -xe; \
    apt update; \
    apt install -y wget build-essential cmake automake libtool autoconf; \
    apt install -y gcc-9 g++-9; \
    update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-9 100; \
    update-alternatives --install /usr/bin/g++ g++ /usr/bin/g++-9 100; \
    wget /~https://github.com/xmrig/xmrig-cuda/archive/refs/tags/v${CUDA_PLUGIN_VERSION}.tar.gz; \
    tar xf v${CUDA_PLUGIN_VERSION}.tar.gz; \
    mv xmrig-cuda-${CUDA_PLUGIN_VERSION} xmrig-cuda; \
    cd xmrig-cuda; \
    mkdir build; \
    cd build; \
    cmake ..; \
    make -j $(nproc);


FROM ubuntu:22.04 as build-runner
ARG VERSION=6.20.0
LABEL maintainer="Patrice Ferlet <metal3d@gmail.com>"

RUN set -xe; \
    apt update; \
    apt install -y wget build-essential cmake automake libtool autoconf; \
    apt install -y gcc-9 g++-9; \
    update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-9 100; \
    update-alternatives --install /usr/bin/g++ g++ /usr/bin/g++-9 100; \
    rm -rf /var/lib/apt/lists/*; \
    wget /~https://github.com/xmrig/xmrig/archive/refs/tags/v${VERSION}.tar.gz; \
    tar xf v${VERSION}.tar.gz; \
    mv xmrig-${VERSION} /xmrig; \
    cd /xmrig; \
    mkdir build; \
    cd scripts; \
    ./build_deps.sh; \
    cd ../build; \
    cmake .. -DXMRIG_DEPS=scripts/deps; \
    make -j $(nproc);
RUN set -xe; \
    cd /xmrig; \
    cp build/xmrig /xmrig


FROM ubuntu:22.04 as runner
LABEL maintainer="Patrice Ferlet <metal3d@gmail.com>"
LABEL org.opencontainers.image.source="/~https://github.com/metal3d/docker-xmrig"
LABEL org.opencontainers.image.description="XMRig miner with CUDA support on Docker, Podman, Kubernetes..." 
LABEL org.opencontainers.image.licenses="MIT"
RUN set -xe; \
    mkdir /xmrig; \
    apt update; \
    apt -y install jq; \
    rm -rf /var/lib/apt/lists/*
COPY --from=build-runner /xmrig/xmrig /xmrig/xmrig
COPY --from=build-runner /xmrig/src/config.json /xmrig/config.json
COPY --from=build-cuda-plugin /xmrig-cuda/build/libxmrig-cuda.so /usr/lib64/


ENV POOL_USER="44vjAVKLTFc7jxTv5ij1ifCv2YCFe3bpTgcRyR6uKg84iyFhrCesstmWNUppRCrxCsMorTP8QKxMrD3QfgQ41zsqMgPaXY5" \
    POOL_PASS="" \
    POOL_URL="xmr.metal3d.org:8080" \
    DONATE_LEVEL=5 \
    PRIORITY=0 \
    THREADS=0 \
    PATH="/xmrig:${PATH}" \
    CUDA=false \
    CUDA_BF="" \
    ALGO="" \
    COIN=""

WORKDIR /xmrig
ADD entrypoint.sh /entrypoint.sh
WORKDIR /tmp
EXPOSE 3000
ENTRYPOINT ["/entrypoint.sh"]
CMD ["xmrig"]

[metal3d]

I'm terribly sorry, already for the time I've taken to reply (I've had a lot of problems but also beautiful things, including a child, in my life, and I've left a lot of things undone).

But I'm also sorry that I can't follow up on this request, at least for the time being and as it stands.

WSL2 is (I'm sorry to say) a cobbled-together mess based on some twisted virtualization, and I don't work on that OS. CUDA is already a worry in itself, and I'd especially have liked to be able to offer OpenCL or Vulkan.

If someone feels like taking over this feature, I'm quite open to letting the hand on this issue. But for my part, I'm likely to leave this request aside.

With apologies.