Update utils.py

utils.py

@@ -62,14 +62,14 @@ def load_tokenizer_and_model(base_model, load_8bit=False):
         device = "cpu"
     
     
-    tokenizer = AutoTokenizer.from_pretrained(base_model, use_fast = True, use_auth_token=True)
+    tokenizer = AutoTokenizer.from_pretrained(base_model, use_fast = True, use_auth_token=True, bos_token='<|startoftext|>', eos_token='<|endoftext|>', pad_token='<|pad|>')
     if device == "cuda":
         model = AutoModelForCausalLM.from_pretrained( 
             base_model,
             load_in_8bit=load_8bit,
             torch_dtype=torch.float16,
             device_map="auto",
-            use_auth_token=True
+            use_auth_token=True,
         )
     else:
         model = AutoModelForCausalLM.from_pretrained(