Update Huozi-based model

Major update. Please try our new Huozi-based model, which is much better.

finetune.py

@@ -21,7 +21,7 @@ from peft import (
     prepare_model_for_int8_training,
     set_peft_model_state_dict,
 )
-from transformers import LlamaForCausalLM, LlamaTokenizer
+from transformers import AutoModelForCausalLM, AutoTokenizer
 
 from utils.prompter import Prompter
 
@@ -111,14 +111,14 @@ def train(
     if len(wandb_log_model) > 0:
         os.environ["WANDB_LOG_MODEL"] = wandb_log_model
 
-    model = LlamaForCausalLM.from_pretrained(
+    model = AutoModelForCausalLM.from_pretrained(
         base_model,
         load_in_8bit=True,
         torch_dtype=torch.float16,
         device_map=device_map,
     )
 
-    tokenizer = LlamaTokenizer.from_pretrained(base_model)
+    tokenizer = AutoTokenizer.from_pretrained(base_model)
 
     tokenizer.pad_token_id = (
         0  # unk. we want this to be different from the eos token