Unable to create a docker to run CUDA on JETSON AGX ORIN for opencv

dhrifk · July 16, 2024, 1:52pm

I tried to run the container for CUDA L4T according to the docs but stumbled upon this error.

docker run -it --rm --net=host --runtime nvidia -e DISPLAY=$DISPLAY -v /tmp/.X11-unix/:/tmp/.X11-unix nvcr.io/nvidia/l4t-cuda:12.2.12-devel    

docker: Error response from daemon: failed to create task for container: failed to create shim task: OCI runtime create failed: nvidia-container-runtime did not terminate successfully: exit status 1: time="2024-07-16T15:31:08+02:00" level=error msg="failed to create NVIDIA Container Runtime: failed to construct OCI spec modifier: requirements not met: unsatisfied condition: cuda>=12.2 (cuda=11.4)"
: unknown.

Tried pulling an older image and I get manifest unknown:

opencv-nvidia sudo docker run -it --rm --net=host --runtime nvidia -e DISPLAY=$DISPLAY -v /tmp/.X11-unix/:/tmp/.X11-unix nvcr.io/nvidia/l4t-cuda:11.4-devel 

Unable to find image 'nvcr.io/nvidia/l4t-cuda:11.4-devel' locally
docker: Error response from daemon: manifest for nvcr.io/nvidia/l4t-cuda:11.4-devel not found: manifest unknown: manifest unknown.
See 'docker run --help'.
➜  opencv-nvidia docker pull ros:noetic-pytorch-l4t-r34.1.1 
Error response from daemon: manifest for ros:noetic-pytorch-l4t-r34.1.1 not found: manifest unknown: manifest unknown

Preferably I would like to not update my version of CUDA and ideally I want to build opencv from source with CUDA support with this dockerfile :

# Use NVIDIA CUDA base image with Ubuntu 20.04
FROM looking_for_this

# Set environment variables
ENV DEBIAN_FRONTEND=noninteractive
ENV OPENCV_VERSION="4.6.0"
ENV WORKSPACE_DIR="/workspace"

# Install dependencies
RUN apt-get update && apt-get install -y \
    build-essential \
    cmake \
    git \
    libgtk2.0-dev \
    pkg-config \
    libavcodec-dev \
    libavformat-dev \
    libswscale-dev \
    libgstreamer1.0-dev \
    libgstreamer-plugins-base1.0-dev \
    gstreamer1.0-plugins-base \
    gstreamer1.0-plugins-good \
    gstreamer1.0-plugins-bad \
    gstreamer1.0-plugins-ugly \
    gstreamer1.0-libav \
    gstreamer1.0-doc \
    gstreamer1.0-tools \
    python3.8-dev \
    python-dev \
    python-numpy \
    python3-numpy \
    libtbb2 \
    libtbb-dev \
    libjpeg-dev \
    libpng-dev \
    libtiff-dev \
    libdc1394-22-dev \
    libv4l-dev \
    v4l-utils \
    qv4l2 \
    curl \
    unzip

# Remove default OpenCV if required (uncomment to enable this step)
# RUN apt -y purge '*libopencv*'

# Create workspace directory
RUN mkdir ${WORKSPACE_DIR}

# Download OpenCV and OpenCV contrib sources
WORKDIR ${WORKSPACE_DIR}
RUN curl -L https://github.com/opencv/opencv/archive/${OPENCV_VERSION}.zip -o opencv-${OPENCV_VERSION}.zip && \
    curl -L https://github.com/opencv/opencv_contrib/archive/${OPENCV_VERSION}.zip -o opencv_contrib-${OPENCV_VERSION}.zip && \
    unzip opencv-${OPENCV_VERSION}.zip && \
    unzip opencv_contrib-${OPENCV_VERSION}.zip && \
    rm opencv-${OPENCV_VERSION}.zip opencv_contrib-${OPENCV_VERSION}.zip

# Build and install OpenCV
WORKDIR ${WORKSPACE_DIR}/opencv-${OPENCV_VERSION}
RUN mkdir release && cd release && \
    cmake -D WITH_CUDA=ON \
          -D WITH_CUDNN=ON \
          -D CUDA_ARCH_BIN="7.2,8.7" \
          -D CUDA_ARCH_PTX="" \
          -D OPENCV_GENERATE_PKGCONFIG=ON \
          -D OPENCV_EXTRA_MODULES_PATH=../../opencv_contrib-${OPENCV_VERSION}/modules \
          -D WITH_GSTREAMER=ON \
          -D WITH_LIBV4L=ON \
          -D BUILD_opencv_python3=ON \
          -D BUILD_TESTS=OFF \
          -D BUILD_PERF_TESTS=OFF \
          -D BUILD_EXAMPLES=OFF \
          -D CMAKE_BUILD_TYPE=RELEASE \
          -D CMAKE_INSTALL_PREFIX=/usr/local .. && \
    make -j$(nproc) && \
    make install

# Update library path
RUN echo 'export LD_LIBRARY_PATH=/usr/local/lib:$LD_LIBRARY_PATH' >> /etc/bash.bashrc && \
    echo 'export PYTHONPATH=/usr/local/lib/python3.8/site-packages/:$PYTHONPATH' >> /etc/bash.bashrc

# Set the working directory
WORKDIR /

# Clean up
RUN apt-get clean && \
    rm -rf /var/lib/apt/lists/*

COPY test.py /root/test.py
# Print success message
CMD ["tail", "-f", "/dev/null"]

dhrifk · July 17, 2024, 2:29pm

I managed to build and run the docker but it seems I get stuck on this :
opencv-nvidia docker run -it --rm --net=host --runtime=nvidia -e DISPLAY=$DISPLAY -v /tmp/.X11-unix/:/tmp/.X11-unix opencv-cuda-gstreamer

==========
== CUDA ==

CUDA Version 11.4.19

This container image and its contents are governed by the NVIDIA Deep Learning Container License.
By pulling and using the container, you accept the terms and conditions of this license:

A copy of this license is made available in this container at /NGC-DL-CONTAINER-LICENSE for your convenience.

After that I can type any command and nothing happens. Cannot exit or ctrl-c aswell I have to stop the docker container with another terminal.

However when I use this command for the official image it works :
opencv-nvidia docker run -it --rm --net=host --runtime=nvidia -e DISPLAY=$DISPLAY -v /tmp/.X11-unix/:/tmp/.X11-unix nvcr.io/nvidia/l4t-cuda:11.4.19-devel
Any idea why my Dockerfile is not working ?

# Use NVIDIA CUDA base image with Ubuntu 20.04
FROM  nvcr.io/nvidia/l4t-cuda:11.4.19-devel

# Set environment variables
ENV DEBIAN_FRONTEND=noninteractive
ENV OPENCV_VERSION="4.6.0"
ENV WORKSPACE_DIR="/workspace"

# Install dependencies
RUN apt-get update && apt-get install -y \
    build-essential \
    cmake \
    git \
    libgtk2.0-dev \
    pkg-config \
    libavcodec-dev \
    libavformat-dev \
    libswscale-dev \
    libgstreamer1.0-dev \
    libgstreamer-plugins-base1.0-dev \
    gstreamer1.0-plugins-base \
    gstreamer1.0-plugins-good \
    gstreamer1.0-plugins-bad \
    gstreamer1.0-plugins-ugly \
    gstreamer1.0-libav \
    gstreamer1.0-doc \
    gstreamer1.0-tools \
    python3.8-dev \
    python-dev \
    python-numpy \
    python3-numpy \
    libtbb2 \
    libtbb-dev \
    libjpeg-dev \
    libpng-dev \
    libtiff-dev \
    libdc1394-22-dev \
    libv4l-dev \
    v4l-utils \
    qv4l2 \
    curl \
    unzip

# Remove default OpenCV if required (uncomment to enable this step)
# RUN apt -y purge '*libopencv*'

# Create workspace directory
RUN mkdir ${WORKSPACE_DIR}

# Download OpenCV and OpenCV contrib sources
WORKDIR ${WORKSPACE_DIR}
RUN curl -L https://github.com/opencv/opencv/archive/${OPENCV_VERSION}.zip -o opencv-${OPENCV_VERSION}.zip && \
    curl -L https://github.com/opencv/opencv_contrib/archive/${OPENCV_VERSION}.zip -o opencv_contrib-${OPENCV_VERSION}.zip && \
    unzip opencv-${OPENCV_VERSION}.zip && \
    unzip opencv_contrib-${OPENCV_VERSION}.zip && \
    rm opencv-${OPENCV_VERSION}.zip opencv_contrib-${OPENCV_VERSION}.zip

# Build and install OpenCV
WORKDIR ${WORKSPACE_DIR}/opencv-${OPENCV_VERSION}
RUN mkdir release && cd release && \
    cmake -D WITH_CUDA=ON \
          -D WITH_CUDNN=ON \
          -D CUDA_ARCH_BIN="7.2,8.7" \
          -D CUDA_ARCH_PTX="" \
          -D OPENCV_GENERATE_PKGCONFIG=ON \
          -D OPENCV_EXTRA_MODULES_PATH=../../opencv_contrib-${OPENCV_VERSION}/modules \
          -D WITH_GSTREAMER=ON \
          -D WITH_LIBV4L=ON \
          -D BUILD_opencv_python3=ON \
          -D BUILD_TESTS=OFF \
          -D BUILD_PERF_TESTS=OFF \
          -D BUILD_EXAMPLES=OFF \
          -D CMAKE_BUILD_TYPE=RELEASE \
          -D CMAKE_INSTALL_PREFIX=/usr/local .. && \
    make -j$(nproc) && \
    make install

# Update library path
RUN echo 'export LD_LIBRARY_PATH=/usr/local/lib:$LD_LIBRARY_PATH' >> /etc/bash.bashrc && \
    echo 'export PYTHONPATH=/usr/local/lib/python3.8/site-packages/:$PYTHONPATH' >> /etc/bash.bashrc

# Set the working directory
WORKDIR /

# Clean up
RUN apt-get clean && \
    rm -rf /var/lib/apt/lists/*

COPY test.py /root/test.py
# Print success message
CMD ["python3", "/root/test.py"]

karimdhrif4444 · July 18, 2024, 3:59pm

I found using these images is simpler and efficient :

Only problems might be with ending up with oversized images. Or if you need to install different dependencies, running into versions conflict.

system · August 1, 2024, 3:59pm

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.

Topic		Replies	Views
Jetson docker image opencv Jetson Xavier NX opencv , docker	11	6416	September 15, 2021
Cross Compile on x86 for Jetson Xavier Dockerfile with opencv cuda support Jetson Xavier NX opencv , cuda , ubuntu , docker	9	4650	April 1, 2022
Error trying to use OpenCV with CUDA support on Docker: CUDA driver version is insufficient for CUDA runtime version Jetson Nano opencv , cuda , docker	4	2844	June 27, 2022
Docker image for Jetson AGX Orin with CUDA environment Jetson AGX Orin cuda , docker , containers	5	2557	June 6, 2024
Unable to run nvidia docker Jetson Xavier NX docker	4	3615	December 8, 2021
Docker issue on permission? Jetson AGX Orin docker	5	165	November 22, 2024
Creating Containers Using nvidia-docker with AGX Xavier Jetson AGX Xavier docker	8	2748	October 18, 2021
Error run docker image nvcr.io/nvidia/l4t-base:r32.6.1 on jetson AGX Jetson AGX Xavier docker , jetson	9	2539	November 10, 2021
Problem to build a docker container and use the GPU on JETSON AGX ORIGIN Jetson AGX Orin docker	3	580	August 30, 2023
Problem creating my devel container in jetson AGX container Jetson AGX Xavier containers	18	3411	December 6, 2021

Unable to create a docker to run CUDA on JETSON AGX ORIN for opencv

========== == CUDA ==

Related topics

==========
== CUDA ==