update model class

app.py

@@ -2,7 +2,7 @@ from threading import Thread
 
 import torch
 import gradio as gr
-from transformers import AutoTokenizer, AutoModelForSeq2SeqLM, TextIteratorStreamer
+from transformers import AutoTokenizer, AutoModelForCausalLM, TextIteratorStreamer
 
 model_id = "EleutherAI/pythia-6.9b-deduped"
 assistant_id = "EleutherAI/pythia-70m-deduped"
@@ -12,11 +12,11 @@ print("CPU threads:", torch.get_num_threads())
 
 
 if torch_device == "cuda":
-    model = AutoModelForSeq2SeqLM.from_pretrained(model_id, load_in_8bit=True, device_map="auto")
+    model = AutoModelForCausalLM.from_pretrained(model_id, load_in_8bit=True, device_map="auto")
 else:
-    model = AutoModelForSeq2SeqLM.from_pretrained(model_id)
+    model = AutoModelForCausalLM.from_pretrained(model_id)
 tokenizer = AutoTokenizer.from_pretrained(model_id)
-assistant_model = AutoModelForSeq2SeqLM.from_pretrained(assistant_id).to(torch_device)
+assistant_model = AutoModelForCausalLM.from_pretrained(assistant_id).to(torch_device)
 
 
 def run_generation(user_text, top_p, temperature, top_k, max_new_tokens):