add load 4bit=true

app.py

@@ -75,7 +75,8 @@ model = AutoModelForCausalLM.from_pretrained(
     MODEL_PATH,
     trust_remote_code=True,
     device_map="auto",   # Use GPU if available
-    low_cpu_mem_usage=True
+    low_cpu_mem_usage=True,
+    load_in_4bit=True
 )
 
 pipe = pipeline(