Upgrading CUDA 11.4 to CUDA 11.8

[software54]

I am facing a compatibility issue to utilize the Jetson GPU instead of the CPU on AI tasks.

My current setup is as follows:
Host machine: x86_64, Ubuntu 20.04

Target Device: Jetson orin NX: aarch64 with CUDA 11.4.

• note there seems to be no version of pytorch and torchvision available with CUDA enabled.

The pytorch version required (with CUDA enabled) can be found in : Start Locally | PyTorch

It seems there are no other CUDA versions available apart from 11.8, 12.1 and 12.4 which is needed to utilize the GPU for accelerated computing.

What I need to do is maintain Ubuntu 20.04 as I am using ROS1. I however need to upgrade the CUDA version to at least 11.8 to be able to utilize the GPU computing with CUDA enabled as per the pytorch link above.

[AastaLLL]

Hi,

Please find below the link for a CUDA version of PyTorch:

PyTorch for Jetson

Below are pre-built PyTorch pip wheel installers for Jetson Nano, TX1/TX2, Xavier, and Orin with JetPack 4.2 and newer. Download one of the PyTorch binaries from below for your version of JetPack, and see the installation instructions to run on your Jetson. These pip wheels are built for ARM aarch64 architecture, so run these commands on your Jetson (not on a host PC). You can also use the containers from jetson-containers. PyTorch pip wheels JetPack 6

PyTorch v2.3.0

JetPack 6.0 (L4T R36.2…

The latest software of Orin NX is JetPack 6 with CUDA 12.2.
Which should meet your requirements.

Thanks.

[software54]

Hi, I think there was some misunderstanding. I needed to maintain my current Jetpack version: 5.1.3, L4T: 35.5.0 but upgrade the SDK installed CUDA 11.4 to CUDA 11.8.

Anyways, I got the solution from here which seemed to work:

CUDA Toolkit 11.8 Downloads