Help on deploying a custom NGC based pytorch docker image on GCP?

Hi all.

Newbie with Docker and GCP here. I composed a custom NGC PyTorch based image (with some custom libs I need) and can deploy it locally. When I try to put it to GCP it fails.

I specify the link to docker hub, but I have a feeling that I requires other access rights or expects the image to reside somewhere else…

The image is here:

Upon creation of the compute instance I refer the image as:

Is there a tutorial somehow how this is done or debugged?