fix automodel for llama

model.py

@@ -1,5 +1,5 @@
 import os
-from transformers import AutoModelForSeq2SeqLM, AutoTokenizer, pipeline
+from transformers import AutoModelForSeq2SeqLM, AutoTokenizer, AutoModelForCausalLM
 from transformers import LEDForConditionalGeneration, LEDTokenizer
 from langchain_openai import OpenAI
 # from huggingface_hub import login
@@ -23,7 +23,7 @@ def get_local_model(model_name_or_path:str)->pipeline:
         model_name_or_path,
         token = hf_token
     )
-    model = AutoModelForSeq2SeqLM.from_pretrained( 
+    model = AutoModelForCausalLM.from_pretrained( 
         model_name_or_path,
         torch_dtype=torch.float32,
         token = hf_token