KServe with MLFlow

모델 준비

MLServer Local Test

python3 -m pip install mlserver mlserver-mlflow

.
├── model/
│   ├── conda.yaml
│   ├── MLmodel
│   ├── model.pkl
│   ├── python_env.yaml
│   └── requirements.txt
├── model-settings.json
└── ...

• https://mlserver.readthedocs.io/en/stable/reference/model-settings.html

model-settings.json

{
	"name": "mlserver-mlflow-test",
	"implementation": "mlserver_mlflow.MLflowRuntime",
	"parameters": {
		"uri": "./model"
	}
}

mlserver start .

• :8080/v2/docs 에 접속하면 API 정보를 볼 수 있습니다.

InferenceService

apiVersion: serving.kserve.io/v1beta1
kind: InferenceService
metadata:
  name: <name>
  namespace: <namespace>
spec:
  predictor:
    model:
      modelFormat:
        name: mlflow
      protocolVersion: v2
      storageUri: <modelUri>