Description of problem:  If I choose a GPU for my tensorflow image, and there are no gpus setup on the platform I receive the following error message:

 
Server requested
2021-04-13T19:26:40.665309Z [Warning] 0/10 nodes are available: 10 Insufficient nvidia.com/gpu.
2021-04-13T19:26:47.685496Z [Warning] 0/10 nodes are available: 10 Insufficient nvidia.com/gpu.
Spawn failed: Timeout

Prerequisites (if any, like setup, operators/versions):

Steps to Reproduce

1. log into pilot testing platform:  https://jupyterhub-redhat-ods-applications.apps.rhods-internal.ju9j.p1.openshiftapps.com/
2. Choose Tensorflow image
3. Choose 1 GPU
4. Select Spawn notebook

Actual results: Spawn fails, and user is stuck in the server startup page.

Expected results:  jupyter notebook image is created, but you receive a message saying 'gpu capability currently not available'

Reproducibility (Always/Intermittent/Only Once):  always

Build Details:

Additional info: