update

src/models/Llama.py

@@ -17,21 +17,21 @@ class Llama(Model):
         api_pos = int(config["api_key_info"]["api_key_use"])
         self.hf_token = config["api_key_info"]["api_keys"][api_pos] or os.getenv("HF_TOKEN")
         self.tokenizer = AutoTokenizer.from_pretrained(self.name, use_auth_token=self.hf_token)
-        self._model = None  # Delayed init
+        self.model = None  # Delayed init
         self.terminators = [
             self.tokenizer.eos_token_id,
             self.tokenizer.convert_tokens_to_ids("<|eot_id|>")
         ]
 
     def _load_model_if_needed(self):
-        if self._model is None:
-            self._model = AutoModelForCausalLM.from_pretrained(
+        if self.model is None:
+            self.model = AutoModelForCausalLM.from_pretrained(
                 self.name,
                 torch_dtype=torch.bfloat16,
                 device_map=self.device,
                 token=self.hf_token
             )
-        return self._model
+        return self.model
 
     def query(self, msg, max_tokens=128000):
         model = self._load_model_if_needed()