Update app.py

app.py

@@ -277,6 +277,7 @@ model = LlamaForCausalLM.from_pretrained(
     load_in_8bit=args.load_in_8bit,
     device_map="auto",
 )
+model.cuda()
 model.eval()
 tokenizer=AutoTokenizer.from_pretrained(args.base_model_path)
 warmup(model)