GKE Node auto-provisioning not scaling up with limits defined

Question

I want to use GKE node auto-provisioning to create a node-pool with GPU on demand (that is when I start a Job that needs GPU resources).

Going with the GCP tutorial I've set up a cluster with enabled cluster autoscaling and node auto-provisioning. NAP has set up limits for CPU, Memory and GPU:

resourceLimits:
  - maximum: '15'
    minimum: '1'
    resourceType: cpu
  - maximum: '150'
    minimum: '1'
    resourceType: memory
  - maximum: '2'
    resourceType: nvidia-tesla-k80

I know that NAP works because it already spun up a few nodes for me, but all of them were "normal ones" (without GPU).

Now, to "force" NAP to create node-pool with GPU machine. Prior to that, no GPU node exists on the cluster. To do that, I'm creating a Job with such a configuration file:

apiVersion: batch/v1
kind: Job
metadata:
  name: training-job
spec:
  ttlSecondsAfterFinished: 100
  template:
    metadata:
      name: training-job
    spec:
      nodeSelector:
        gpu: "true"
        cloud.google.com/gke-spot: "true"
        cloud.google.com/gke-accelerator: nvidia-tesla-k80
      tolerations:
        - key: cloud.google.com/gke-spot
          operator: Equal
          value: "true"
          effect: NoSchedule
      containers:
        - name: gpu-test
          image: przomys/gpu-test
          resources:
            requests:
              cpu: 500m
            limits:
              nvidia.com/gpu: 2 # requesting 2 GPU
      restartPolicy: Never # Do not restart containers after they exit

Job is being created, but then it is marked as "Unschedulable" and CA Log gives me such error:

{
  "noDecisionStatus": {
    "measureTime": "1650370630",
    "noScaleUp": {
      "unhandledPodGroups": [
        {
          "rejectedMigs": [
            {
              "reason": {
                "messageId": "no.scale.up.mig.failing.predicate",
                "parameters": [
                  "NodeAffinity",
                  "node(s) didn't match Pod's node affinity/selector"
                ]
              },
              "mig": {
                "zone": "us-central1-c",
                "nodepool": "pool-3",
                "name": "gke-cluster-activeid-pool-3-af526144-grp"
              }
            },
            {
              "mig": {
                "name": "gke-cluster-activeid-nap-e2-standard--c7a4d4f1-grp",
                "zone": "us-central1-c",
                "nodepool": "nap-e2-standard-2-w52e84k8"
              },
              "reason": {
                "parameters": [
                  "NodeAffinity",
                  "node(s) didn't match Pod's node affinity/selector"
                ],
                "messageId": "no.scale.up.mig.failing.predicate"
              }
            }
          ],
          "napFailureReasons": [
            {
              "parameters": [
                "Any GPU."
              ],
              "messageId": "no.scale.up.nap.pod.gpu.no.limit.defined"
            }
          ],
          "podGroup": {
            "totalPodCount": 1,
            "samplePod": {
              "controller": {
                "apiVersion": "batch/v1",
                "kind": "Job",
                "name": "training-job"
              },
              "namespace": "default",
              "name": "training-job-7k8zd"
            }
          }
        }
      ],
      "unhandledPodGroupsTotalCount": 1
    }
  }
}

My guess is that no.scale.up.nap.pod.gpu.no.limit.defined is the most important part. GCP tutorial points me here. But I have this limit defined, thus I'm out of ideas...

Maybe someone has an idea what I'm doing wrong?