No GPU (nvidia-smi failed)

post by kresimir58 on Mar 27, 2024

Hi, i’ve been sucessfully using 1xH100 instances in the last few days, and every day i would launch/terminate instance.

Today i have a problem, it seems that no gpu is detected on system:

nvidia-smi

NVIDIA-SMI has failed because it couldn’t communicate with the NVIDIA driver. Make sure that the latest NVIDIA driver is installed and running.

I tried:

apt update && apt install
sudo apt install nvidia-driver-545 (doesn’t work)
sudo apt install nvidia-driver-535 (doesn’t work)

rebooting in between steps as well

nothing helped

post by stevenic on Apr 30, 2024

Did you ever get a resolution to this? I can’t get nvidia-smi to find any GPU’s on an H100 instance

post by marek on Jul 8, 2024

Having the same issue here on my Tensorbook

post by kresimir58 on Jul 11, 2024

i contacted customer service and they asked me to run additional commands and send logs to them. but, when i started new instance, problem dissapeared. so i guess there was intermittent issue or issue with just that instance i got.