Has anyone tried installing other Linux distros on Jetson Nano? I am looking at headless operation for my Jetson Nano. Is it hard to install Cuda, tensorrt, tensorflow if I install a regular Linux distro?
Cuda relies on nvidia X server ABI, and xserver relies on GPU driver, which are specific to L4T.
You wouldn’t have these working with other distros, AFAIK.
This is something that I intend to try. Nearly all of my development has been on CentOS/RHEL distributions. Since you can install and develop CUDA on top of a CentOS based system (after installing the drivers, toolkits, etc)…would seem you should be able to mirror that logic on top of the ARM distribution for CentOS. The trick would seem in creating that first image with the CentOS distro on the MicroSD (know there is a RaspberryPI version…might start with that)…as I don’t believe you can boot/install from USB devices.
My Nano is expected to arrive tomorrow…then the fun begins ;-)
Indeed, you have to minimally have the mSD for the bootloader; the system, however, can be loaded from USB. You just need a Sata to USB 3.0 cable and an SSD.
And I’ve seen the Open Source bootloader, but I was trying to build Armbian on a RPi3 though… It worked…just not really well… But that’s an RPi3 and a different OS than CentOS; the Nano I would have a much greater tolerance for from the outset too I imagine…
If it doesn’t work though, I see the rage inducing projects always as a great learning experience.
Indeed…will have two mSDs…one with the official, hopefully less rage inducing load…the other with attempts at CentOS (mainly because I just have used it so much in the past…just more comfortable with it). CentOS.org does have a RPi3 image…which worked really well.
But…as you say…trying to work through issues can be a great learning experience.