Update handler.py

handler.py

@@ -19,38 +19,51 @@ class EndpointHandler:
         self.model.eval()
         
     def __call__(self, data):
-        # video_path = data["inputs"]["video_path"]
-        # ftp_password = data["inputs"].get("ftp_password")
-        # ftp_password = data["inputs"]["ftp_password"]
-
-        inputs = json.loads(data)
-        video_path = inputs.get("inputs", {}).get("video_path", "")
-        ftp_password = inputs.get("inputs", {}).get("ftp_password", "")
-
-        processed_frames = read_video(video_path, ftp_password)
-
-
-        
-        # 디버깅: 입력 데이터 확인
-        logger.info(f"Received data: {data}")
-
-        frames = np.array(data['frames'])
-        frames = torch.tensor(frames).float()  # Ensure the data is in the correct format
-
-        # 디버깅: 프레임 데이터 확인
-        logger.info(f"Frames shape: {frames.shape}")
-        
-        # Perform inference
-        with torch.no_grad():
-            outputs = self.model(frames.unsqueeze(0))  # Add batch dimension
-            predictions = torch.softmax(outputs.logits, dim=-1)
-
-            # 디버깅: 예측 결과 확인
-            logger.info(f"Predictions: {predictions}")
+        try:
+            # video_path = data["inputs"]["video_path"]
+            # ftp_password = data["inputs"].get("ftp_password")
+            # ftp_password = data["inputs"]["ftp_password"]
+    
+            inputs = json.loads(data)
+            video_path = inputs.get("inputs", {}).get("video_path", "")
+            ftp_password = inputs.get("inputs", {}).get("ftp_password", "")
+    
+            processed_frames = read_video(video_path, ftp_password)
+    
+    
+            
+            # 디버깅: 입력 데이터 확인
+            logger.info(f"Received data: {data}")
+    
+            frames = np.array(data['frames'])
+            frames = torch.tensor(frames).float()  # Ensure the data is in the correct format
+    
+            # 디버깅: 프레임 데이터 확인
+            logger.info(f"Frames shape: {frames.shape}")
             
-            predicted_class = torch.argmax(predictions, dim=-1).item()
+            # Perform inference
+            with torch.no_grad():
+                outputs = self.model(frames.unsqueeze(0))  # Add batch dimension
+                predictions = torch.softmax(outputs.logits, dim=-1)
+    
+                # 디버깅: 예측 결과 확인
+                logger.info(f"Predictions: {predictions}")
+                
+                predicted_class = torch.argmax(predictions, dim=-1).item()
+    
+                # 디버깅: 예측 클래스 확인
+                logger.info(f"Predicted class: {predicted_class}")
+            
+            return {"predicted_class": predicted_class, "predictions": predictions.tolist()}
 
-            # 디버깅: 예측 클래스 확인
-            logger.info(f"Predicted class: {predicted_class}")
-        
-        return {"predicted_class": predicted_class, "predictions": predictions.tolist()}
+
+        except Exception as e:
+            error_message = str(e)
+            stack_trace = traceback.format_exc()
+            logger.error(f"Error: {error_message}")
+            logger.error(f"Stack trace: {stack_trace}")
+            return json.dumps({
+                "status": "error",
+                "message": error_message,
+                "stack_trace": stack_trace
+            }), 500