kserve/config/runtimes/kserve-torchserve.yaml

apiVersion: serving.kserve.io/v1alpha1
kind: ClusterServingRuntime
metadata:
  name: kserve-torchserve
  annotations:
    serving.kserve.io/server-type: torchserve
spec:
  annotations:
    prometheus.kserve.io/port: "8082"
    prometheus.kserve.io/path: "/metrics"
  supportedModelFormats:
    - name: pytorch
      version: "1"
      autoSelect: true
      priority: 2
  protocolVersions:
    - v1
    - v2
    - grpc-v2
  containers:
    - name: kserve-container
      image: kserve-torchserve:replace
      args:
        - torchserve
        - --start
        - --model-store=/mnt/models/model-store
        - --ts-config=/mnt/models/config/config.properties
      env:
        - name: "TS_SERVICE_ENVELOPE"
          value: "{{.Labels.serviceEnvelope}}"
      securityContext:
        runAsUser: 1000    # User ID is not defined in the Dockerfile, so we need to set it here to run as non-root
        allowPrivilegeEscalation: false
        privileged: false
        runAsNonRoot: true
        capabilities:
          drop:
            - ALL
      resources:
        requests:
          cpu: "1"
          memory: 2Gi
        limits:
          cpu: "1"
          memory: 2Gi