Holoscan SDK

The Holoscan SDK Python Wheel 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.

Visit the NGC demo website for a live demonstration of some of Holoscan capabilities.

System Requirements

• CPU: x86_64 or aarch64 (arm64)
• OS: Ubuntu 20.04
• Python: 3.8 to 3.11 (pip 20.2+)
• NVIDIA GPU
  • Ampere or above recommended for best performance
  • Quadro/NVIDIA RTX necessary for GPUDirect RDMA support
  • Tested with :
    • discrete GPUs: NVIDIA RTX 6000, NVIDIA RTX A6000
    • integrated GPUs: NVIDIA Clara AGX Developer Kit (Xavier SoC), NVIDIA IGX Orin Developer Kit (Orin SoC)
• NVIDIA dGPU drivers: 510.73.08+
• CUDA Runtime: 11.6.1+ (< 12.0)
• For Rivermax support (optional):
  • NVIDIA ConnectX SmartNIC
  • OFED Network Drivers: 5.8
  • GPUDirect Drivers: 1.1
  • Rivermax SDK: 1.20 (for local development only)

Troubleshooting

ERROR gxf/std/extension_loader.cpp@114: Failed to load extension ... Error: /lib/x86_64-linux-gnu/libm.so.6: version `GLIBC_2.29' not found

The current version of the wheels were built and tested on Ubuntu 20.04 with libc 2.31 instead of manylinux. They were named as such to make them available early on PyPI. You'll need to use a system with a more recent version of libc to use the Holoscan SDK python wheels.

ImportError: libcudart.so.11.0: cannot open shared object file: No such file or directory

Cuda runtime is missing from your system (required even for CPU only pipelines). Follow links in the section above to install CUDA, or install it through the python wheels:

python3 -m pip install nvidia-cuda-runtime-cu11 --index https://pypi.ngc.nvidia.com

ERROR gxf/std/extension_loader.cpp@114: Failed to load extension ... Error: libnvonnxparser.so.8: cannot open shared object file: No such file or directory

TensorRT is missing from your system. It is only needed for the inference operator (in holoscan.operators):

• If you do not need it, only include the holoscan submodules required for your use case (ex: from holoscan import core).
• If you do need it, follow links in the section above to install TensorRT, or install it through the python wheels:

python3 -m pip install nvidia-tensorrt==8.2.3.0 --index https://pypi.ngc.nvidia.com

ERROR ImportError: cannot import name 'SCHEME_KEYS' from 'pip._internal.models.scheme' ... Error processing line 1 of holoscan-x.y.z.pth:

You version of pip is probably below the required 20.2. Upgrade pip with:

python3 -m pip install --upgrade pip