using cache to avoid mle

app.py

@@ -11,7 +11,7 @@ class Assistant:
         device_map = 'auto'
         # bnb_config = BitsAndBytesConfig(load_in_4bit=True, bnb_4bit_quant_type="nf4",bnb_4bit_compute_dtype=torch.float16,)
         # self.model = AutoModelForCausalLM.from_pretrained( model_name,quantization_config=bnb_config, trust_remote_code=True,use_cache=False,device_map=device_map)
-        self.model = AutoModelForCausalLM.from_pretrained( model_name, trust_remote_code=True,use_cache=False,device_map=device_map)
+        self.model = AutoModelForCausalLM.from_pretrained( model_name, trust_remote_code=True,use_cache=True,device_map=device_map)
         
         self.tokenizer = AutoTokenizer.from_pretrained(model_name, trust_remote_code=True)
         self.tokenizer.pad_token = self.tokenizer.eos_token