Cannot provision GPU instance on google cloud with YAML

[ShrishtiKarkera]

I am trying to deploy mk8s on google cloud with juju, I have 3 units of standard instance and a GPU instance, however I couldn't find a way to do it through the YAML file.
I did the same on AWS but the instances there are configured a bit differently than the ones on GCP, for AWS the script looks like:

  1 default-base: ubuntu@22.04/stable
  2 applications:
  3   microk8s:
  4     charm: microk8s
  5     channel: 1.28/stable
  6     num_units: 3
  7     options:
  8       hostpath_storage: true
  9     constraints: arch=amd64 root-disk=102400 instance-type=t3.xlarge
 10   microk8s-gpu:
 11     charm: microk8s
 12     channel: 1.28/stable
 13     num_units: 1
 14     options:
 15       hostpath_storage: true
 16       role: worker
 17     constraints: arch=amd64 root-disk=512000 instance-type=g5.2xlarge
 18 relations:
 19 - - microk8s:workers
 20   - microk8s-gpu:control-plane

For GCP:
I changed line 9 and 17 to:

9   constraints: arch=amd64 root-disk=100G instance-type=n2-standard-4
17 constraints: arch=amd64 root-disk=500G instance-type=n1-standard-8 gpu-type=nvidia-t4:2

There is no constraint designated as gpu-type or accelerator, leaving the only remaining option to provision the resources manually and include the machines in the YAML script. However, these instances do not come pre-installed with NVIDIA drivers, necessitating manual installation via SSH access to each machine.