Update app.py

app.py

@@ -34,10 +34,11 @@ print(f"Using device: {DEVICE}")
 print(f"Low memory: {LOW_MEMORY}")
 
 # Quantization configuration for efficient model loading
-quantization_config = BitsAndBytesConfig(
-    load_in_4bit=True,
-    bnb_4bit_compute_dtype=torch.float16
-)
+# Define BitsAndBytesConfig
+quantization_config = BitsAndBytesConfig(load_in_4bit=True,
+                                bnb_4bit_quant_type="nf4",
+                                bnb_4bit_compute_dtype=torch.float16)
+
 
 # Load models only once
 processor = AutoProcessor.from_pretrained(MODEL_ID)