oc label isvc/caikit-tgis-example-isvc-rest    opendatahub.io/dashboard="true"
oc annotate  isvc/caikit-tgis-example-isvc-rest        openshift.io/display-name=caikit-tgis-example-isvc-rest 

apiVersion: v1
items:
- apiVersion: serving.kserve.io/v1beta1
  kind: InferenceService
  metadata:
    annotations:
      kubectl.kubernetes.io/last-applied-configuration: |
        {"apiVersion":"serving.kserve.io/v1beta1","kind":"InferenceService","metadata":{"annotations":{"serving.knative.openshift.io/enablePassthrough":"true","sidecar.istio.io/inject":"true","sidecar.istio.io/rewriteAppHTTPProbers":"true"},"name":"caikit-tgis-example-isvc-rest","namespace":"kserve-demo"},"spec":{"predictor":{"model":{"modelFormat":{"name":"caikit"},"runtime":"caikit-tgis-runtime","storage":{"key":"aws-connection-minio","path":"llm/models/flan-t5-small-caikit"}}}}}
      openshift.io/display-name: caikit-tgis-example-isvc-rest
      security.opendatahub.io/enable-auth: "true"
      serving.knative.openshift.io/enablePassthrough: "true"
      sidecar.istio.io/inject: "true"
      sidecar.istio.io/rewriteAppHTTPProbers: "true"
    creationTimestamp: "2024-05-03T20:41:48Z"
    finalizers:
    - inferenceservice.finalizers
    generation: 1
    labels:
      opendatahub.io/dashboard: "true"
      test: a
    name: caikit-tgis-example-isvc-rest
    namespace: kserve-demo
    resourceVersion: "48119195"
    uid: 3aa35fb9-6759-42da-8eea-13d002eda29c
  spec:
    predictor:
      model:
        modelFormat:
          name: caikit
        name: ""
        resources: {}
        runtime: caikit-tgis-runtime
        storage:
          key: aws-connection-minio
          path: llm/models/flan-t5-small-caikit
 

- apiVersion: serving.kserve.io/v1alpha1
  kind: ServingRuntime
  metadata:
    annotations:
      kubectl.kubernetes.io/last-applied-configuration: |
        {"apiVersion":"serving.kserve.io/v1alpha1","kind":"ServingRuntime","labels":{"opendatahub.io/dashboard":"true"},"metadata":{"annotations":{"opendatahub.io/recommended-accelerators":"[\"nvidia.com/gpu\"]","openshift.io/display-name":"Caikit TGIS ServingRuntime for KServe"},"name":"caikit-tgis-runtime","namespace":"kserve-demo"},"spec":{"containers":[{"args":["--model-name=/mnt/models/artifacts/"],"command":["text-generation-launcher"],"env":[{"name":"TRANSFORMERS_CACHE","value":"/tmp/transformers_cache"}],"image":"quay.io/opendatahub/text-generation-inference@sha256:e50101c1245b22e080b5754efb7c2b69fd1422e5a608d276eab249bdc5cc5a55","name":"kserve-container"},{"env":[{"name":"RUNTIME_LOCAL_MODELS_DIR","value":"/mnt/models"},{"name":"TRANSFORMERS_CACHE","value":"/tmp/transformers_cache"},{"name":"RUNTIME_GRPC_ENABLED","value":"false"},{"name":"RUNTIME_HTTP_ENABLED","value":"true"},{"name":"RUNTIME_GRPC_SERVER_THREAD_POOL_SIZE","value":"64"}],"image":"quay.io/opendatahub/caikit-tgis-serving@sha256:153294140ecc87839defeca317b4b67cc16e8f56ec2211f3540ed00e0efa0ae0","livenessProbe":{"exec":{"command":["python","-m","caikit_health_probe","liveness"]}},"name":"transformer-container","ports":[{"containerPort":8080,"protocol":"TCP"}],"readinessProbe":{"exec":{"command":["python","-m","caikit_health_probe","readiness"]}}}],"multiModel":false,"supportedModelFormats":[{"autoSelect":true,"name":"caikit"}]}}
      opendatahub.io/recommended-accelerators: '["nvidia.com/gpu"]'
      openshift.io/display-name: Caikit TGIS ServingRuntime for KServe
    creationTimestamp: "2024-05-03T20:41:46Z"
    generation: 1
    name: caikit-tgis-runtime
    namespace: kserve-demo
    resourceVersion: "46153756"
    uid: 157fe7ad-f835-44b3-a353-40e08574c018
  spec:
    containers:
    - args:
      - --model-name=/mnt/models/artifacts/
      command:
      - text-generation-launcher
      env:
      - name: TRANSFORMERS_CACHE
        value: /tmp/transformers_cache
      image: quay.io/opendatahub/text-generation-inference@sha256:e50101c1245b22e080b5754efb7c2b69fd1422e5a608d276eab249bdc5cc5a55
      name: kserve-container
    - env:
      - name: RUNTIME_LOCAL_MODELS_DIR
        value: /mnt/models
      - name: TRANSFORMERS_CACHE
        value: /tmp/transformers_cache
      - name: RUNTIME_GRPC_ENABLED
        value: "false"
      - name: RUNTIME_HTTP_ENABLED
        value: "true"
      - name: RUNTIME_GRPC_SERVER_THREAD_POOL_SIZE
        value: "64"
      image: quay.io/opendatahub/caikit-tgis-serving@sha256:153294140ecc87839defeca317b4b67cc16e8f56ec2211f3540ed00e0efa0ae0
      livenessProbe:
        exec:
          command:
          - python
          - -m
          - caikit_health_probe
          - liveness
      name: transformer-container
      ports:
      - containerPort: 8080
        protocol: TCP
      readinessProbe:
        exec:
          command:
          - python
          - -m
          - caikit_health_probe
          - readiness
    multiModel: false
    supportedModelFormats:
    - autoSelect: true
      name: caikit