26 lines
953 B
YAML
26 lines
953 B
YAML
apiVersion: serving.kserve.io/v1beta1
|
|
kind: InferenceService
|
|
metadata:
|
|
name: {{ include "kserve.fullname" . }}
|
|
annotations:
|
|
queue.sidecar.serving.knative.dev/resourcePercentage: "10"
|
|
labels:
|
|
{{- include "kserve.labels" . | nindent 4 }}
|
|
spec:
|
|
predictor:
|
|
runtimeClassName: {{ .Values.inferenceService.runtimeClassName }}
|
|
model:
|
|
modelFormat:
|
|
name: {{ .Values.inferenceService.model.modelFormat.name }}
|
|
{{- with .Values.inferenceService.model.args }}
|
|
args:
|
|
{{- toYaml . | nindent 8 }}
|
|
{{- end }}
|
|
storageUri: {{ .Values.inferenceService.model.storageUri }}
|
|
{{- if .Values.inferenceService.model.image }}
|
|
image: {{ .Values.inferenceService.model.image }}
|
|
{{- end }}
|
|
resources:
|
|
{{- toYaml .Values.inferenceService.resources | nindent 8 }}
|
|
minReplicas: {{ .Values.inferenceService.minReplicas }}
|
|
maxReplicas: {{ .Values.inferenceService.maxReplicas }} |