As an IT Operations administrator, I need to understand how to make GPUs available to data scientists for compute-heavy models.

RHODS-540 and https://github.com/red-hat-data-services/odh-manifests/pull/125 should assist with documenting this process or linking to appropriate OCP documentation.

From Landon:
This will require a documentation update that covers

• RHODS jupter notebooks will deploy with the toleration `nvidia.com/gpu` EXISTS when a notebook requests 1 or more GPUs
• NVIDIA GPU addon only supports the NoSchedule node taint `nvidia.com/gpu` for GPU enablement when the addon is installed. * I don't know if we are responsible for the nvidia gpu addon docs but this should be in their docs and we can link to the appropriate section
  If you need examples of the pod toleration or nvidia objects, let Landon know