kserve

charts/kserve/CUSTOM-README.md

@@ -5,7 +5,7 @@
 - 명령어
 
   ```sh
-  $ helm upgrade --install kserve ./charts/kserve \
+  $ helm upgrade --install kserve ../kserve \
     --namespace kserve \
     --create-namespace \
     -f custom-values.yaml

charts/kserve/custom-values.yaml

@@ -1,13 +1,13 @@
 # KServe Custom Values
 inferenceService:
-  name: "custom-huggingface-model"
+  name: "phi-1_5"
   
   model:
     args:
       - --backend=vllm
-      - --model_name=llama-2-7b
+      - --model_name=phi-1_5
       - --dtype=float16
-    storageUri: "pvc://custom-model-volume/llama-2-7b"
+    storageUri: "pvc://model-volume/phi-1_5"
   
   resources:
     limits:
@@ -23,6 +23,6 @@ ingress:
   annotations:
     cert-manager.io/cluster-issuer: "root-ca-issuer"
   host:
-    domain: "kserve.example.com"
+    domain: "phi15.default.hopt.paasup.io"
   tls:
     secretName: custom-kserve-tls

charts/kserve/templates/inferenceservice.yaml

@@ -1,8 +1,9 @@
 apiVersion: serving.kserve.io/v1beta1
 kind: InferenceService
 metadata:
-  name: {{ .Values.inferenceService.name }}
+  name: {{ include "kserve.fullname" . }}
-  namespace: {{ .Values.inferenceService.namespace }}
+  annotations:
+    queue.sidecar.serving.knative.dev/resourcePercentage: "10"
   labels:
     {{- include "kserve.labels" . | nindent 4 }}
 spec:

charts/kserve/templates/ingress.yaml

@@ -2,7 +2,7 @@
 apiVersion: networking.k8s.io/v1
 kind: Ingress
 metadata:
-  name: {{ .Values.inferenceService.name }}
+  name: {{ include "kserve.fullname" . }}
   namespace: {{ .Values.ingress.namespace }}
   labels:
     {{- include "kserve.labels" . | nindent 4 }}