gpu-workload/t5/kubernetes/hpa.yaml

apiVersion: autoscaling/v2
kind: HorizontalPodAutoscaler
metadata:
  name: t5-inference
spec:
  scaleTargetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: t5-inference
  minReplicas: 1
  maxReplicas: 5
  metrics:
  - type: Pods
    pods:
      metric:
        name: prometheus.googleapis.com|ts_queue_latency_microseconds|counter
      target:
        type: AverageValue
        averageValue: "7000000"