low memory

app.py

@@ -44,7 +44,7 @@ def load_model(args):
     llm = LLM(
         args=args, 
         model='serve/fake_json/{}.json'.format(args.gpt_model), 
-        gpu_memory_utilization=0.5, 
+        gpu_memory_utilization=0.4, 
         skip_tokenizer_init=True)
     print(f"gpt model is loaded")
     return vq_model, llm, image_size