Update handler.py

handler.py

@@ -4,7 +4,7 @@ sys.path.append('./')
 from videollama2 import model_init, mm_infer
 from videollama2.utils import disable_torch_init
 import logging
-import numpy as np
+import os
 
 class EndpointHandler:
     def __init__(self, path: str = ""):
@@ -18,16 +18,37 @@ class EndpointHandler:
         self.model_path = 'Aliayub1995/VideoLLaMA2-7B'
         self.model, self.processor, self.tokenizer = model_init(self.model_path)
 
-    def __call__(self, video_tensor: np.ndarray) -> List[Dict[str, Any]]:
-        logging.info("Received video tensor")  # Debugging: Confirm video tensor received
+    def __call__(self, data: Dict[str, Any]) -> List[Dict[str, Any]]:
+        logging.info(f"Received data: {data}")  # Debugging: Print received data
+        # Initialize variables
+        current_path = os.getcwd()
+        logging.info(f"Current Path: {current_path}")
+        dir = os.walk("./app")
+        # Iterate through the generator
+        for dirpath, dirnames, filenames in dir:
+            logging.info(f"Current Path: {dirpath}")
+            logging.info(f"Directories: {dirnames}")
+            logging.info(f"Files: {filenames}")
+            logging.info("-" * 40)
+        logging.info(f"Directory struct: {dir}")
+        modal = None
+        modal_path = None
+        instruct = None
 
-        # Default values
-        modal = "video"
-        instruct = "Can you explain each scene and provide the exact time of the video in which it happened in this format [start_time: end_time]: Description, [start_time: end_time]: Description ..."
+        # Extract input data
+        inputs = data.get("inputs", data)
+        modal = inputs.get("modal", "video")
+        modal_path = inputs.get("modal_path", "")
+        instruct = inputs.get("instruct", "")
+
+        logging.info(f"Modal: {modal}, Modal Path: {modal_path}, Instruct: {instruct}")  # Debugging: Print extracted values
+
+        if not modal_path or not instruct:
+            raise ValueError("Both 'modal_path' and 'instruct' must be provided in the input data.")
 
         # Perform inference
         output = mm_infer(
-            self.processor[modal](video_tensor), 
+            self.processor[modal](modal_path), 
             instruct, 
             model=self.model, 
             tokenizer=self.tokenizer,