fix: rm legacy

Author: terryyz

bigcodebench/model.py

@@ -133,7 +133,7 @@ def __init__(self, name: str, dataset: str, tp: int, **kwargs) -> None:
         if self.tokenizer_name is None:
             self.tokenizer_name = self.name
 
-        self.tokenizer = AutoTokenizer.from_pretrained(self.tokenizer_name, legacy=False, **kwargs)
+        self.tokenizer = AutoTokenizer.from_pretrained(self.tokenizer_name, **kwargs)
         if self.tokenizer.chat_template is None:
             self.eos += extra_eos_for_direct_completion(dataset)
         self.llm = LLM(model=name, max_model_len=2048, **kwargs)
@@ -193,7 +193,7 @@ def __init__(self, name: str, dataset: str, **kwargs):
         if self.tokenizer_name is None:
             self.tokenizer_name = self.name
 
-        self.tokenizer = AutoTokenizer.from_pretrained(self.tokenizer_name, legacy=False, **kwargs)
+        self.tokenizer = AutoTokenizer.from_pretrained(self.tokenizer_name, **kwargs)
 
         if self.tokenizer.chat_template is None:
             self.eos += extra_eos_for_direct_completion(dataset)