Jetson AGX Xavier Machine Stuck at Boot

I have been trying to install pytorch to the machine, did not work at all despite hours and hours of attempt.
Just when I was trying another method metioned on this forum the machine responded that I no longer have enough space on the primary drive, so I was thinking of migrating the whole system to an ssd that I have installed on the machine.
When I copied the information of memory mounted on the machine, it suddenly shut itself.
Then it went into a cycle of flashing screen that is stuck at the very early stage of booting.
This is the same even when I have swtiched it off and starting it again with the switch on button.
Below is a short clip of what is happening

External Media

My machine is the developer version, don’t know if that’s part of the reason.

Any help would be appreciated.

And if anyone knows how to properly mount pytorch to this platform I would appreciate that as well. The previous attempts have been too painstaking.

Thanks in advance.

Looks like this happens right as you reach graphical.target, which means the system has probably booted, but the GUI is failing. Sometimes this is related to an update which may have overwritten one of NVIDIA’s OpenGL files with the Mesa/Nouveau version (this is not common, but it does sometimes happen).

Serial console is by far the easiest way to debug this as it has no dependency on video. You might be able to just log in with serial console and fix the issue without much effort if non-GUI login still works.

The dev kit comes with a micro-B USB cable. If on your host you monitor logs with “dmesg --follow”, and then plug in this cable to the host PC, you should see several serial devices in the log. I think it is the first device which is the console, and I’ll just assume it is “/dev/ttyUSB0”. The actual name would differ if something else with serial USB is already present.

Then I use gtkterm, but other people use minicom or other serial programs. To install gtkterm, “sudo apt-get install gtkterm”. Then, if “/dev/ttyUSB0” is correct (use “sudo” if this fails, which would be the case if your PC user is not a member of group “dialout”):
gtkterm -b 8 -t 1 -s 115200 -p /dev/ttyUSB0

Can you log in this way? If so, then it will be much easier to debug. Ssh might also work, but serial console is usually easier to use.

Hi,

It looks like something went wrong when migrating the system.

If reflashing is an option for you, it’s recommended to reflash the system and then install the pyTorch package from this topic:

This package should work well on Jetson platform.

Thanks.

I have reflashed the whole thing with sdk manager from host computer
Now I am going to move the system to a larger ssd I mounted on the m.2 slot beneath the cooling fan

Thanks for all your help