DEPLOY - v1.0.1

app/core/logging_config.py

@@ -0,0 +1,31 @@
+from logging.config import dictConfig
+
+"""
+애플리케이션 전체에서의 로깅 레벨 설정
+"""
+def setup_logging():
+    dictConfig({
+        "version": 1,
+        "disable_existing_loggers": False,
+        "formatters": {
+            "default": {
+                "format": "%(asctime)s %(levelname)s [%(name)s] %(message)s",
+            },
+        },
+        "handlers": {
+            "console": {
+                "class": "logging.StreamHandler",
+                "formatter": "default",
+            },
+        },
+        "root": {
+            "level": "INFO",
+            "handlers": ["console"],
+        },
+        "loggers": {
+            "uvicorn": {"level": "INFO"},
+            "uvicorn.error": {"level": "INFO"},
+            "uvicorn.access": {"level": "INFO"},
+            "app": {"level": "INFO", "handlers": ["console"], "propagate": False},
+        },
+    })

app/main.py

@@ -1,7 +1,9 @@
 from fastapi import FastAPI, Request
 from app.core.lifespan import lifespan
 from app.api.endpoints import predictions
+from app.core.logging_config import setup_logging
 
+setup_logging()
 app = FastAPI(
     title="DearBelly CV API",
     description="DearBelly CV를 위한 Swagger 입니다.",

app/services/openai_service.py

@@ -1,6 +1,9 @@
 from app.core.config import settings
 from openai import OpenAI
 import json
+import logging
+
+logger = logging.getLogger(__name__)
 
 client = OpenAI(api_key=settings.OPENAI_KEY)
 
@@ -32,7 +35,7 @@ def ask_chatgpt_about_pregnancy_safety(self, pill_name: str) -> tuple[str, int]:
             max_tokens=600,
             response_format={"type": "json_object"}
         )
-        print("GPT Asking 성공...")
+        logger.info("GPT Asking 성공...")
         raw = response.choices[0].message.content.strip()
 
         try:
@@ -45,6 +48,7 @@ def ask_chatgpt_about_pregnancy_safety(self, pill_name: str) -> tuple[str, int]:
             else:
                 # 디버깅
                 preview = raw[:200].replace("\n", "\\n")
+                logger.warning("JSON 파싱 실패")
                 raise ValueError(f"응답이 유효한 JSON이 아닙니다. preview='{preview}'")
 
         description = data.get("description")

app/services/predictor_service.py

@@ -1,76 +1,152 @@
 
 import torch
 import torch.nn as nn
-import torch.nn.functional as F
+import timm
 from torchvision import transforms
 from PIL import Image
 import json
 from pathlib import Path
 from io import BytesIO
+import logging
 
-class LightCNN(nn.Module):
-    def __init__(self, num_classes):
+logger = logging.getLogger(__name__)
+
+class EfficientNetBaseline(nn.Module):
+    def __init__(self, num_classes, pretrained=True, dropout=0.2):
         super().__init__()
-        self.conv1 = nn.Conv2d(3, 8, 3, padding=1)
-        self.conv2 = nn.Conv2d(8, 16, 3, padding=1)
-        self.pool = nn.MaxPool2d(2, 2)
-        self.gap = nn.AdaptiveAvgPool2d((4, 4))
-        self.fc1 = nn.Linear(16 * 4 * 4, 64)
-        self.fc2 = nn.Linear(64, num_classes)
+        logger.info(
+            f"Initializing EfficientNetBaseline with num_classes={num_classes}, pretrained={pretrained}, dropout={dropout}")
+
+        self.backbone = timm.create_model(
+            "efficientnet_b3", pretrained=pretrained, num_classes=0, global_pool="avg"
+        )
+        feat_dim = self.backbone.num_features
+
+        self.bn = nn.BatchNorm1d(feat_dim)
+        self.dp = nn.Dropout(dropout)
+        self.fc = nn.Linear(feat_dim, num_classes)
 
     def forward(self, x):
-        x = self.pool(F.relu(self.conv1(x)))
-        x = self.pool(F.relu(self.conv2(x)))
-        x = self.gap(x)
-        x = x.view(x.size(0), -1)
-        x = F.relu(self.fc1(x))
-        x = self.fc2(x)
-        return x
+        feats = self.backbone(x)
+        feats = self.bn(feats)
+        feats = self.dp(feats)
+        logits = self.fc(feats)
+        return logits
 
 class PredictorService:
     def __init__(self, model_path: Path, json_path: Path):
+
+        logger.info(f"Initializing PredictorService with model_path={model_path}, json_path={json_path}")
+
         self.device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+        # cuda 모델 확인
+        logger.info(f"Using device: {self.device}")
+
         self.idx2label = self._load_idx2label(json_path)
         self.num_classes = len(self.idx2label)
+        logger.info(f"Loaded {self.num_classes} classes")
+
         self.model = self._load_model(model_path)
         self.transform = transforms.Compose([
             transforms.Resize((64, 64)),
             transforms.ToTensor(),
         ])
+        logger.info("PredictorService initialized successfully")
 
     def _load_idx2label(self, json_path: Path) -> dict:
-        with open(json_path, "r", encoding="utf-8") as f:
-            data = json.load(f)
-        idx2label = data.get("idx2label")
-        if not idx2label:
-            unique_labels = sorted(set(sample["label"] for sample in data["samples"]))
-            idx2label = {str(label): f"K-{label:06d}" for label in unique_labels}
-        return idx2label
-
-    def _load_model(self, model_path: Path) -> LightCNN:
+
+        # json 제대로 읽었는지 확인
+        logger.info(f"Loading idx2label from {json_path}")
+
+        try:
+            with open(json_path, "r", encoding="utf-8") as f:
+                data = json.load(f)
+            idx2label = data.get("idx2label")
+            if not idx2label:
+                logger.warning("idx2label not found in JSON, generating from samples")
+                unique_labels = sorted(set(sample["label"] for sample in data["samples"]))
+                idx2label = {str(label): f"K-{label:06d}" for label in unique_labels}
+            return idx2label
+
+        # 예외 사항 추가
+        except FileNotFoundError:
+            logger.error(f"JSON file not found: {json_path}")
+            raise
+        except json.JSONDecodeError as e:
+            logger.error(f"Failed to parse JSON file: {e}")
+            raise
+        except Exception as e:
+            logger.error(f"Unexpected error loading idx2label: {e}", exc_info=True)
+            raise
+
+    def _load_model(self, model_path: Path) -> EfficientNetBaseline:
+        # model path 확인하기
+        logger.info(f"Loading model from {model_path}")
+
         import __main__
-        __main__.LightCNN = LightCNN
-        model = torch.load(model_path, map_location=self.device, weights_only=False)
-        model.eval()
-        return model
+        __main__.EfficientNetBaseline = EfficientNetBaseline
+
+        try:
+            object = torch.load(model_path, map_location=self.device, weights_only=False)
+
+            if isinstance(object, nn.Module) :
+                # 그 자체로 모델일 때
+                model = object.to(self.device)
+            elif isinstance(object, dict) :
+                # 반환 타입이 state_dict
+                state_dict = object
+                for k in ['state_dict', 'model_state_dict', 'model']:
+                    if k in object and isinstance(object[k], dict):
+                        state_dict = object[k]
+                        break
+
+                model = EfficientNetBaseline(self.num_classes).to(self.device)
+
+                missing, unexpected = model.load_state_dict(state_dict, strict=False)
+                if missing or unexpected:
+                    logger.warning(f"[load_state_dict] missing keys: {missing}, unexpected keys: {unexpected}")
+            else:
+                # type 일치하지 않음
+                error_msg = f"Unsupported checkpoint type: {type(object)}"
+                logger.error(error_msg)
+                raise TypeError(f"Unsupported checkpoint type: {type(object)}")
+
+            # 완료
+            model.eval()
+            logger.info("Model loaded and set to evaluation mode")
+            return model
+
+        except FileNotFoundError:
+            logger.error(f"Model file not found: {model_path}")
+            raise
+        except Exception as e:
+            logger.error(f"Failed to load model: {e}", exc_info=True)
+            raise
+
 
     def predict(self, stream_file: BytesIO) -> tuple[str, str, float]:
-        image = Image.open(stream_file).convert('RGB')
-        input_tensor = self.transform(image).unsqueeze(0).to(self.device)
 
-        with torch.no_grad():
-            output = self.model(input_tensor)
-            predicted_idx = torch.argmax(output, dim=1).item()
-            confidence = torch.softmax(output, dim=1)[0][predicted_idx].item()
+        try :
+            image = Image.open(stream_file).convert('RGB')
+            input_tensor = self.transform(image).unsqueeze(0).to(self.device)
+
+            with torch.no_grad():
+                output = self.model(input_tensor)
+                predicted_idx = torch.argmax(output, dim=1).item()
+                confidence = torch.softmax(output, dim=1)[0][predicted_idx].item()
 
-        label = str(predicted_idx)
-        pill_name = self.idx2label.get(label, f"Unknown Label: {label}")
+            label = str(predicted_idx)
+            pill_name = self.idx2label.get(label, f"Unknown Label: {label}")
+            logger.info(f"Prediction completed - pill_name: {pill_name}, label: {label}, confidence: {confidence:.4f}")
 
-        return pill_name, label, confidence
+            return pill_name, label, confidence
+        except Exception as e:
+            logger.error(f"Prediction failed: {e}", exc_info=True)
+            raise
 
 
 HERE = Path(__file__).resolve().parent.parent
-MODEL_PATH = HERE / "models" / "models" / "best_model_0823.pt"
+MODEL_PATH = HERE / "models" / "models" / "best_model_0920.pt"
 JSON_PATH = HERE / "models" / "models" / "matched_all.json"
 
 predictor_service = PredictorService(MODEL_PATH, JSON_PATH)

app/worker/tasks.py

@@ -8,13 +8,16 @@
 from app.services.openai_service import checker
 from app.services.predictor_service import predictor_service
 from app.services.s3_service import s3_service
+import logging
+
+logger = logging.getLogger(__name__)
 
 """
 이미지를 다운 -> 다운 한 것에 대하여 모델 분석 요청
 """
 async def process_image_scan(job: ImageJob, redis_client: redis.Redis):
     correlationId = job.correlationId
-    print(f"[task] Start image scan for job_id={correlationId}")
+    logging.info(f"[task] Start image scan for job_id={correlationId}")
     try:
 
         stream_file = await asyncio.to_thread(
@@ -28,7 +31,7 @@ async def process_image_scan(job: ImageJob, redis_client: redis.Redis):
             predictor_service.predict,
             stream_file
         )
-        print(f"[task] Start Asking GPT for job_id={correlationId}")
+        logging.info(f"[task] Start Asking GPT for job_id={correlationId}")
         description, isSafe = checker.ask_chatgpt_about_pregnancy_safety(pillName)
         finishedAt = datetime.utcnow().isoformat()
 
@@ -50,9 +53,9 @@ async def process_image_scan(job: ImageJob, redis_client: redis.Redis):
             approximate=True,
         )
 
-        print(f"[task] Image scan successfully finished for job_id={correlationId}")
+        logging.info(f"[task] Image scan successfully finished for job_id={correlationId}")
 
     except Exception as e:
-        print(f"[task] Failed to process job_id={correlationId}: {e}")
+        logging.warning(f"[task] Failed to process job_id={correlationId}: {e}")
     finally:
-        print(f"[task] Image scan finished for job_id={correlationId}")
+        logging.info(f"[task] Image scan finished for job_id={correlationId}")

app/worker/worker.py

@@ -4,6 +4,9 @@
 from app.core.config import settings
 from app.schemas.job import ImageJob
 from app.worker.tasks import process_image_scan
+import logging
+
+logger = logging.getLogger(__name__)
 
 """
 Redis Stream에 정의한 유효한 형식 메시지를 위한 전처리 함수
@@ -49,7 +52,7 @@ def __init__(self, redis_client: redis_client):
         self.redis_client = redis_client
 
     async def run(self):
-        print(f"[worker] start consumer={settings.CONSUMER_NAME} group={settings.GROUP_NAME} stream={settings.STREAM_JOB}")
+        logging.info(f"[worker] start consumer={settings.CONSUMER_NAME} group={settings.GROUP_NAME} stream={settings.STREAM_JOB}")
         reclaim_every_sec = 30
         last_reclaim = 0.0
 
@@ -82,12 +85,12 @@ async def run(self):
 
                                 # 최종 반환 data
                                 data = json.loads(payload_str)
-                                print(f"Job received id={msg_id} correlationId={correlation_id} payload={data}")
+                                logging.info(f"Job received id={msg_id} correlationId={correlation_id} payload={data}")
 
                                 job = ImageJob.model_validate(data)
                                 # XADD까지 호출
                                 task = asyncio.create_task(process_image_scan(job, redis_client))
-                                print(f"[worker] {task} 발행 성공")
+                                logging.info(f"[worker] {task} 발행 성공")
 
                                 # 처리 성공 시에만 ack 후 del
                                 task.add_done_callback(lambda t: asyncio.create_task(
@@ -132,7 +135,7 @@ async def run(self):
                             job = ImageJob.model_validate_json(payload)
 
                             task = asyncio.create_task(process_image_scan(job, self.redis_client))
-                            print(f"[worker] {task} 발행 성공")
+                            logging.info(f"[worker] {task} 발행 성공")
 
                             def _on_done(t: asyncio.Task, *, msg_id=msg_id, fields=fields):
                                 async def _ack_or_dlq():
@@ -155,8 +158,8 @@ async def _ack_or_dlq():
                             await self.redis_client.xadd(f"{settings.STREAM_JOB}:DLQ", clean)
 
             except asyncio.CancelledError:
-                print("[worker] cancelled; bye")
+                logging.warning("[worker] cancelled; bye")
                 break
             except Exception as e:
-                print(f"[worker] error: {e}")
+                logging.warning(f"[worker] error: {e}")
                 await asyncio.sleep(1)