Update app.py

app.py

@@ -18,7 +18,7 @@ import tensorflow as tf # Or from keras.utils import load_img, img_to_array
 import threading
 import subprocess
 from ultralytics.nn.tasks import DetectionModel
-torch.serialization.add_safe_globals([DetectionModel])
+torch.serialization.add_safe_globals([DetectionModel, Sequential])
 
 from io import BytesIO
 import glob
@@ -28,9 +28,11 @@ import gradio as gr
 
 # IMPORTING MODELS
 description_model = "llava-hf/bakLlava-v1-hf"
+print(f"Loading vision model: {description_model} (this should be much faster)...")
+
 llava_model = LlavaForConditionalGeneration.from_pretrained(
     description_model,
-    torch_dtype=torch.float16,
+    dtype=torch.float16,
     low_cpu_mem_usage=True,
     load_in_4bit=True
 )