Update app.py

app.py

@@ -18,7 +18,10 @@ def load_llama_model():
     model = MllamaForConditionalGeneration.from_pretrained(
         model_id,
         torch_dtype=torch.bfloat16,
-        # device_map="auto",
+        device_map="auto",
+        # offload_folder="offload", 
+    ).to("cuda")
+
     )
     processor = AutoProcessor.from_pretrained(model_id)