Support to build RHEL8 (or 9) based container image and leverage the compute capabilities of the GPU (CUDA, cuDNN, TensorRT

Robotics & Edge Computing Jetson & Embedded Systems Jetson AGX Orin
, ,
1

Hello,

We are looking to build a RHEL8 (or 9) based container image that can run on Jetson Orin and leverage the compute capabilities of the GPU (CUDA, cuDNN, TensorRT).

We spent some time pulling the RHEL 9 image following the documentation link: Validate the Jetson GPU support | Install Red Hat Device Edge on NVIDIA Jetson Orin and IGX Orin | Red Hat Developer. Based on the instructions, In order to use any of the compute capabilities, we still must run the ubuntu based container images.

As per the documentation, with jetpack 6 with have support for RHEL 9.4 Linux Distro. ( JetPack SDK 6.0 | NVIDIA Developer) but that is with ubuntu based container images.

Looking forward to any support / direction here.

Thank you.

2

*** Please note that this reply is generated by LLM automatically ***

Based on the provided context, it appears that you are looking to build a RHEL8 or 9 based container image that can run on Jetson Orin and leverage the compute capabilities of the GPU (CUDA, cuDNN, TensorRT). However, you are facing challenges in using RHEL based container images with Jetson Orin.

According to the documentation, JetPack 6 supports RHEL 9.4, but only with Ubuntu-based container images. This means that you cannot use RHEL based container images with Jetson Orin out of the box.

However, there are a few possible solutions you can explore:

  1. Use Ubuntu-based container images: As mentioned in the documentation, JetPack 6 supports Ubuntu-based container images. You can use these images as a base and install the necessary packages and dependencies for your application.
  2. Build a custom RHEL-based container image: You can try building a custom RHEL-based container image that includes the necessary packages and dependencies for your application. However, this may require significant effort and expertise in building and maintaining a custom container image.
  3. Use a different Linux distribution: If you are not tied to using RHEL, you can consider using a different Linux distribution that is supported by JetPack 6, such as Ubuntu.

In terms of leveraging the compute capabilities of the GPU, you can use the following tools and frameworks:

  • CUDA: You can use the CUDA toolkit to develop applications that leverage the GPU’s compute capabilities.
  • cuDNN: You can use the cuDNN library to accelerate deep learning applications on the GPU.
  • TensorRT: You can use the TensorRT framework to optimize and deploy deep learning models on the GPU.

To get started with building a container image that can run on Jetson Orin, you can follow the instructions in the documentation for installing Red Hat Device Edge on NVIDIA Jetson Orin and IGX Orin. However, keep in mind that these instructions are specific to Ubuntu-based container images.

In summary, while there is no direct support for RHEL-based container images on Jetson Orin, you can explore alternative solutions such as using Ubuntu-based container images, building a custom RHEL-based container image, or using a different Linux distribution. Additionally, you can use tools and frameworks like CUDA, cuDNN, and TensorRT to leverage the compute capabilities of the GPU.

*** Please give the thumbs up if you get answers you like. Or provide feedback to help us improve the answer. ***

5

Hi,

RHEL is supported by Red Hat directly. Please contact them for more information.
Thanks.

6

Thank you AastaLL. We have reached to RedHat as well. Since the CUDA libraries/packages for Jetson are built by Nvidia, we are looking to understand if Nvidia has built it for RHEL OS or not.

7

Hi,

We don’t have a package specific to RHEL OS.
But the standard one should work.

Thanks.