DouglasNeuroInformatics / ansible-playbooks

Playbooks for configuring CIC Workstations and Servers
4 stars 11 forks source link

Improve GPU detection for SLURM #73

Open gdevenyi opened 11 months ago

gdevenyi commented 11 months ago

Installing the latest CUDA repo https://developer.nvidia.com/cuda-downloads?target_os=Linux&target_arch=x86_64&Distribution=Ubuntu&target_version=22.04&target_type=deb_network

Allows us to install the package: libnvidia-ml-dev

Without needing nvidia drivers, so we can avoid custom-packages for SLURM (related #72 )

We can also add the AMD repo: https://rocm.docs.amd.com/en/latest/deploy/linux/quick_start.html

And install rocm-smi-lib

gdevenyi commented 11 months ago

Also level-zero-dev, https://dgpu-docs.intel.com/driver/client/overview.html