kubeflow/model-registry

Testing 1.9.0-rc.2

Closed this issue · 6 comments

tarilabs commented

Installation

Used:

kubectl apply -k "https://github.com/kubeflow/model-registry/manifests/kustomize/overlays/db?ref=v0.2.2-alpha"
kubectl apply -k "https://github.com/kubeflow/model-registry/manifests/kustomize/options/istio?ref=v0.2.2-alpha"
kubectl wait --for=condition=available -n kubeflow deployment/model-registry-deployment --timeout=1m
kubectl logs -n kubeflow deployment/model-registry-deployment
kubectl port-forward svc/model-registry-service -n kubeflow 8081:8080

curl -X 'GET'   'http://localhost:8081/api/model_registry/v1alpha3/registered_models?pageSize=100&orderBy=ID&sortOrder=DESC'   -H 'accept: application/json' | jq

NOTE:

  • in installation, use 0.2.2

Getting Started: Indexing a Model

NOTE:
should add:

  • version of model-registry
  • is_secure=False to client

image

Getting Started: integration with KServe example

image

Note:

  • consider adding is_secure
  • consider changing lookup_version
tarilabs commented

Working with KServe python SDK 0.12.1:

image
image

Then:

kubectl describe isvc -A
Name:         mnist
Namespace:    kubeflow-user-example-com
Labels:       modelregistry/model-version-id=2
              modelregistry/registered-model-id=1
Annotations:  <none>
API Version:  serving.kserve.io/v1beta1
Kind:         InferenceService
Metadata:
  Creation Timestamp:  2024-07-01T15:24:31Z
  Finalizers:
    inferenceservice.finalizers
  Generation:        1
  Resource Version:  82582
  UID:               228c1c03-8763-40f5-a5bb-b18742df0868
Spec:
  Predictor:
    Model:
      Model Format:
        Name:            onnx
        Version:         1
      Name:              
      Protocol Version:  v2
      Resources:
      Runtime:      kserve-ovms
      Storage Uri:  https://github.com/tarilabs/demo20231212/raw/main/v1.nb20231206162408/mnist.onnx
Status:
  Model Status:
    Last Failure Info:
      Message:  Waiting for runtime to become available
      Reason:   RuntimeNotRecognized
    States:
      Active Model State:  
      Target Model State:  FailedToLoad
    Transition Status:     InvalidSpec
Events:
  Type     Reason         Age                  From                Message
  ----     ------         ----                 ----                -------
  Warning  InternalError  23s (x15 over 105s)  v1beta1Controllers  No ServingRuntimes or ClusterServingRuntimes with the name: kserve-ovms

Which is expected because I didn't install OpenVINO in this test cluster for this ONNX file, but that is beyond the scope of the tutorial.

tarilabs commented

NOTE add for troubleshooting from withing notebook:

  • wget -nv -O- model-registry-service.kubeflow.svc.cluster.local:8080/api/model_registry/v1alpha3/registered_models
  • curl model-registry-service.kubeflow.svc.cluster.local:8080/api/model_registry/v1alpha3/registered_models
tarilabs commented

Note:
retested specifically the failure when KServe SDK uses Ray and Ray fails.

worked using image jupyter-tensorflow-full:v1.9.0-rc.0
and pinning transitive as:

!pip install kserve=="0.13.0" ray=="2.10.0" attrs=="21.4.0" jsonschema=="4.17.3"

proof: test20240702.zip

(one reason more for the REST rebase 😄 )

tarilabs commented

The following related changes:

were tested out with a local minikube and 1.9.0-rc.2 from KF/manifest

tarilabs commented

Relevant test of Model Registry 0.2.3 with KF 1.9rc2

tarilabs commented

Marking this as Done,
being part of: