Update app.py

app.py

@@ -13,7 +13,7 @@ llm = Llama(
     model_path=model_path,
     n_ctx=2048,
     chat_format="llama-3",
-    n_gpu_layers=-1,  # ensure all layers are on GPU
+    n_gpu_layers=50,  # ensure all layers are on GPU
     flash_attn=True,
 )