Fix running condition of translate_llama2

Function 'translate_llama2' is needed for Llama 2 Family because of there are transformer version(model name ended with "-hf") and torch version, which is useless now due to the models are integrated to one from Llama3 at recent family.

src/utils.py

@@ -17,7 +17,7 @@ def translate_llama2(text):
 
 def get_model(model_name: str, library: str, access_token: str):
     "Finds and grabs model from the Hub, and initializes on `meta`"
-    if "meta-llama" in model_name:
+    if "meta-llama" in model_name and "Llama-2" in model_name:
         model_name = translate_llama2(model_name)
     if library == "auto":
         library = None