Update app.py

app.py

@@ -9,8 +9,8 @@ model = AutoModelForCausalLM.from_pretrained(
     MODEL_PATH,
     trust_remote_code=True,
     device_map="auto",
-    low_cpu_mem_usage=True,
-    quantization_config=None
+    low_cpu_mem_usage=True
+
 )
 
 pipe = pipeline(