♻️Refactor: MLflow 모델 버전 관리 MLflow 내장 버전으로 변경

k8s/model-api/base/deployment.yaml

@@ -32,8 +32,6 @@ spec:
                   key: secret-access-key
             - name: AWS_DEFAULT_REGION
               value: ap-northeast-2
-            - name: CHURN_MODEL_VERSION # 모델 버전 추가
-              value: "v5"
           resources:
             requests:
               cpu: "250m"

services/model-api/app/models/loader.py

@@ -6,31 +6,31 @@
 _model_cache: dict = {}
 
 # mlflow에서 mnist 모델 로드하는 함수
-def load_model(model_name: str, version: str):
-    cache_key = f"{model_name}_{version}"
+def load_model(model_name: str):
+    cache_key = model_name
     if cache_key in _model_cache:
         return _model_cache[cache_key]
 
     mlflow_uri = os.getenv("MLFLOW_TRACKING_URI", "http://mlflow:5000") # 환경변수에서 mlflow tracking uri 가져오기, 없으면 기본값으로 http://mlflow:5000 사용
     mlflow.set_tracking_uri(mlflow_uri)
 
-    model_uri = f"models:/{model_name}-{version}/latest"
+    model_uri = f"models:/{model_name}/latest"
     model = mlflow.pytorch.load_model(model_uri)
     model.eval()
 
     _model_cache[cache_key] = model
     return model
 
 # mlflow에서 고객 이탈 예측 모델 로드하는 함수
-def load_churn_model(model_name: str, version: str):
-    cache_key = f"{model_name}_{version}_sklearn"
+def load_churn_model(model_name: str):
+    cache_key = f"{model_name}_sklearn"
     if cache_key in _model_cache:
         return _model_cache[cache_key]
 
     mlflow_uri = os.getenv("MLFLOW_TRACKING_URI", "http://mlflow:5000")
     mlflow.set_tracking_uri(mlflow_uri)
 
-    model_uri = f"models:/{model_name}-{version}/latest"
+    model_uri = f"models:/{model_name}/latest"
     model = mlflow.sklearn.load_model(model_uri)
 
     _model_cache[cache_key] = model

services/model-api/app/services/churn_service.py

@@ -6,7 +6,6 @@
 from prometheus_client import Counter, Histogram
 
 MODEL_NAME = "churn"
-MODEL_VERSION = os.getenv("CHURN_MODEL_VERSION", "v5")
 
 FEATURES = [
     'account_age_months',
@@ -36,7 +35,7 @@ def predict(data: dict) -> dict:
     if missing:
         raise ValueError(f"누락된 피처: {missing}")
 
-    model = load_churn_model(MODEL_NAME, MODEL_VERSION)
+    model = load_churn_model(MODEL_NAME)
 
     x = [[data[f] for f in FEATURES]]
 
@@ -58,7 +57,6 @@ def predict(data: dict) -> dict:
         },
         "metadata": {
             "model": MODEL_NAME,
-            "version": MODEL_VERSION,
             "inference_time_ms": time_ms,
             "timestamp": datetime.now(timezone.utc).isoformat(),
         },

services/model-api/app/services/mnist_service.py

@@ -7,8 +7,7 @@
 
 
 MODEL_NAME = "mnist"
-MODEL_VERSION = "v1"
-
+
 prediction_counter = Counter(
     "mnist_predictions_total",
     "예측 횟수",
@@ -26,7 +25,7 @@ def predict(data: dict) -> dict:
     if pixels is None or len(pixels) != 784:
         raise ValueError(f"pixels 필드에 784개의 값이 필요합니다. 길이 오류: {len(pixels)}")
 
-    model = load_model(MODEL_NAME, MODEL_VERSION)
+    model = load_model(MODEL_NAME)
 
     start = time.time()
     with torch.no_grad():
@@ -48,7 +47,6 @@ def predict(data: dict) -> dict:
         },
         "metadata": {
             "model": MODEL_NAME,
-            "version": MODEL_VERSION,
             "inference_time_ms": time_ms,
             "timestamp": datetime.now(timezone.utc).isoformat(),
         },

services/model-api/training/churn/train.py

@@ -49,7 +49,7 @@ def main(args):
         mlflow.sklearn.log_model(
             model,
             artifact_path="model",
-            registered_model_name=f"churn-{args.version}",
+            registered_model_name="churn",
         )
 
         print(f"valid_f1    : {val_metrics['f1']:.4f}")