baseRefs only

config.yaml

apiVersion: serving.kserve.io/v1alpha1
kind: LLMInferenceService
metadata:
  name: llm-inference-service-model-fb-opt-125m-router-managed-workload
  namespace: kserve-ci-e2e-test
spec:
  baseRefs:
    - name: model-fb-opt-125m
    - name: router-managed
    - name: workload-single-cpu
  replicas: 1
---
apiVersion: serving.kserve.io/v1alpha1
kind: LLMInferenceServiceConfig
metadata:
  name: model-fb-opt-125m
  namespace: kserve-ci-e2e-test
spec:
  model:
    name: facebook/opt-125m
    uri: hf://facebook/opt-125m
---
apiVersion: serving.kserve.io/v1alpha1
kind: LLMInferenceServiceConfig
metadata:
  name: router-managed
  namespace: kserve-ci-e2e-test
spec:
  router:
    gateway: {}
    route: {}
    scheduler: {}
---
apiVersion: serving.kserve.io/v1alpha1
kind: LLMInferenceServiceConfig
metadata:
  name: workload-single-cpu
  namespace: kserve-ci-e2e-test
spec:
  template:
    containers:
      - name: main
        image: quay.io/pierdipi/vllm-cpu:latest
        env:
          - name: VLLM_LOGGING_LEVEL
            value: DEBUG
        livenessProbe:
          failureThreshold: 5
          initialDelaySeconds: 30
          periodSeconds: 30
          timeoutSeconds: 30
        resources:
          limits:
            cpu: "1"
            memory: 10Gi
          requests:
            cpu: 100m
            memory: 8Gi