@peterdrake Same problem here with Fedora. No solution yet apart from waiting for updates to the kernel, or perhaps nvidia-drivers, or whatever else might be causing the problem.
@peterdrake Something about NVIDIA drivers and Ubuntu... I was just fighting with them for hours the other night, trying to find the right NVIDIA driver (from a couple dozen, with unclear documentation) to fix some GPU setting, which ended up not being Ubuntu-compatible anyway. But hey, at least I learned how to monitor GPU processes in the process! "watch nvidia-smi" is fun...
@colditzjb I'm going to end up hand-smelting my own wires by the end of this, aren't I?
@peterdrake 😆 It sounds like a worthy challenge.
@peterdrake There's a prettier version - "nvitop" - that shows angry red text when I get within 1% of overrunning the GPU memory. @andresmh : here's me blowing things up, trying to run LLMs again.
@aebrockwell Installing new NVIDIA drivers seems to have fixed the problem!
https://linuxconfig.org/how-to-install-the-nvidia-drivers-on-ubuntu-22-04
I'm not sure how the graphics card driver is involved in turning the machine on and off, but I mostly stay away from the hardware end of things. I'll probably learn more about it during this Ubuntu adventure.
@khird