Update app.py

app.py

@@ -290,7 +290,6 @@ def _get_scenes_extraction(
         scene_info: List[Dict] = []
 
         for i, (start_time, end_time) in enumerate(scene_list):
-            print("a")
             frame_number = int(start_time.get_frames()) + offset_frames
             cap.set(cv2.CAP_PROP_POS_FRAMES, frame_number)
             ret, frame = cap.read()
@@ -343,7 +342,8 @@ def _get_image_list_description(
     model = LlavaOnevisionForConditionalGeneration.from_pretrained(
         path_model,
         torch_dtype=torch.float16,
-        low_cpu_mem_usage=False
+        low_cpu_mem_usage=True,
+        device_map="auto"
     ).to("cuda")
 
     # System prompt for image description