mike--b

I just gave that a try, it looks like it wants the GPU number rather than the ID, and using 0, without the no_groups change, same issue : docker run --rm --gpus GPU-d532059d-6184-51aa-302a-2411694349df nvidia/cuda:12.1.1-runtime-ubuntu22.04 nvidia-smi invalid argument "GPU-d532059d-6184-51aa-302a-2411694349df" for "--gpus" flag: count must be an integer: strconv.Atoi: parsing "GPU-d532059d-6184-51aa-302a-2411694349df": invalid syntax See 'docker run --help'. docker run --rm --gpus 0 nvidia/cuda:12.1.1-runtime-ubuntu22.04 nvidia-smi docker: Error response from daemon: failed to create shim task: OCI runtime create failed: runc create failed: unable to start container process: error during container init: error running hook #0: error running hook: exit status 1, stdout: , stderr: Auto-detected mode as 'legacy' nvidia-container-cli: container error: failed to get device cgroup mount path: no cgroup filesystem mounted for the devices subsytem in mountinfo file: unknown.

Testing out unraid 6.12.0-rc4.1 and it seems containers with gpu's assigned wont run due to a cgroup change (to cgroupv2) that has been made in 6.12. I uninstalled and re installed the driver, disabled and enabled docker (even rebooted) and outcome is always the same: docker run --rm --gpus all nvidia/cuda:12.1.1-runtime-ubuntu22.04 nvidia-smi docker: Error response from daemon: failed to create shim task: OCI runtime create failed: runc create failed: unable to start container process: error during container init: error running hook #0: error running hook: exit status 1, stdout: , stderr: Auto-detected mode as 'legacy' nvidia-container-cli: container error: failed to get device cgroup mount path: no cgroup filesystem mounted for the devices subsytem in mountinfo file: unknown. Only solution I have found so far to get the container to start is to edit /etc/nvidia-container-runtime/config.toml and replace #no-cgroups = false to no-cgroups = true in the [nvidia-container-cli] section.