Linux / arm64
The Holoscan container is part of NVIDIA Holoscan, the AI sensor processing platform that combines hardware systems for low-latency sensor and network connectivity, optimized libraries for data processing and AI, and core microservices to run streaming, imaging, and other applications, from embedded to edge to cloud. It can be used to build streaming AI pipelines for a variety of domains, including Medical Devices, High Performance Computing at the Edge, Industrial Inspection and more.
In previous releases, the prefix
Clara
was used to define Holoscan as a platform designed initially for medical devices. As Holoscan has grown, its potential to serve other areas has become apparent. With version 0.4.0, we're proud to announce that the Holoscan SDK is now officially built to be domain-agnostic and can be used to build sensor AI applications in multiple domains. Note that some of the content of the SDK (sample applications) or the documentation might still appear to be healthcare-specific pending additional updates. Going forward, domain specific content will be hosted on the HoloHub repository.
The Holoscan container includes the Holoscan libraries, GXF extensions, headers, example source code, and sample datasets, as well as all the dependencies that were tested with Holoscan. It is the recommended way to run the Holoscan examples, while still allowing you to create your own C++ and Python Holoscan application.
Visit the Holoscan User Guide to get started with the Holoscan SDK.
Log in to the NGC docker registry
docker login nvcr.io
Press the Copy Image Path
button at the top of this webpage and choose the version you want to test:
v<version>-dgpu
for x86_64
systems or a holoscan developer kit configured with a discrete GPUv<version>-igpu
for holoscan developer kits configured with an integrated GPUSet it as your NGC_CONTAINER_IMAGE_PATH
in your terminal.
# For example
export NGC_CONTAINER_IMAGE_PATH="nvcr.io/nvidia/clara-holoscan/holoscan:v1.0.3-dgpu"
Ensure that X11 is configured to allow commands from docker:
xhost +local:docker
Start the container
docker run -it --rm --net host \
--runtime=nvidia \
-e NVIDIA_DRIVER_CAPABILITIES=all \
-v /tmp/.X11-unix:/tmp/.X11-unix \
-e DISPLAY=$DISPLAY \
--ipc=host \
--cap-add=CAP_SYS_PTRACE \
--ulimit memlock=-1 \
--ulimit stack=67108864 \
${NGC_CONTAINER_IMAGE_PATH}
--runtime=nvidia
and -e NVIDIA_DRIVER_CAPABILITIES
are properties of the nvidia container toolkit to leverage the NVIDIA GPUs and their capabilities. Read more here.-v /tmp/.X11-unix
and -e DISPLAY
are needed to enable X11 display forwarding.--ipc=host
, --cap-add=CAP_SYS_PTRACE
, --ulimit memlock=-1
and --ulimit stack=67108864
are required to run distributed applications with UCX. Read more here.--device /dev/ajantv20
(and/or ajantv2<n>
) in the docker run command if you have an AJA capture card you'd like to access from the container.--device /dev/video0
(and/or video<n>
) to make your V4L2 video devices (HDMI IN, USB) available in the container.--group-add video
or ensure the user has appropriate permissions to the video device nodes (/dev/video*
).--device /dev/capture-vi-channel<n>
(or --privileged
to avoid numerous flags)--device /dev/infiniband/rdma_cm
and --device /dev/infiniband/uverbs0
(and/or uverbs0<n>
) to make your ConnectX RDMA interface available in the container./dev/dri/*
). This can be done by adding --group-add $(cat /etc/group | grep "video" | cut -d: -f3)
and --group-add $(cat /etc/group | grep "render" | cut -d: -f3)
(Note: simply passing --group-add render
might not work if the group id differs from your host and container, even if mounting /etc/group
)The Holoscan SDK is installed under /opt/nvidia/holoscan
. It includes a CMake configuration file inside lib/cmake/holoscan
, allowing you to import holoscan in your CMake project (link libraries + include headers):
find_package(holoscan REQUIRED CONFIG PATHS "/opt/nvidia/holoscan")
target_link_libraries(yourTarget PUBLIC holoscan::core)
Alternatives to hardcoding PATHS
inside find_package
in CMake are listed under the Config Mode Search Procedure documentation.
Python, C++, and GXF examples are installed in /opt/nvidia/holoscan/examples
alongside their source code, and run instructions (also available on the GitHub repository).
Example to run the Hello World example:
# Python
python3 /opt/nvidia/holoscan/examples/hello_world/python/hello_world.py
# C++
cd /opt/nvidia/holoscan/examples
./hello_world/cpp/hello_world
Make sure to edit any relative path in the yaml config if you want to run from a different working directory.
You can rebuild the C++ and GXF examples as-is or copy them anywhere on your system to experiment with.
Example to build all the C++ and GXF examples:
export src_dir="/opt/nvidia/holoscan/examples/" # Add "<example_of_your_choice>/cpp" to build a specific example
export build_dir="</path/of/your/choice/>"
cmake -S $src_dir -B $build_dir -G Ninja \
-D Holoscan_ROOT="/opt/nvidia/holoscan"
cmake --build $build_dir -j
Also see the HoloHub repository for a collection of Holoscan operators and applications which you can use in your pipeline or for reference.
By pulling and using the container, you accept the terms and conditions of this End User License Agreement.