Update app.py

app.py

@@ -1,7 +1,7 @@
 import os
 import math
 import transformers
-from transformers import AutoModelForCausalLM, AutoTokenizer
+from transformers import AutoModelForCausalLM, AutoTokenizer , TextStreamer
 import torch
 import gradio as gr
 import sentencepiece
@@ -14,6 +14,7 @@ device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
 model_name = "Intel/neural-chat-7b-v3-1"
 tokenizer = AutoTokenizer.from_pretrained("Intel/neural-chat-7b-v3-1")
 model = AutoModelForCausalLM.from_pretrained(model_name, torch_dtype=torch.bfloat16, device_map="auto")
+streamer = TextStreamer(tokenizer)
 
 class IntelChatBot:
     def __init__(self, model, tokenizer, system_message="You are 🧠🤌🏻Neuro, an AI language model created by Tonic-AI. You are a cautious assistant. You carefully follow instructions. You are helpful and harmless and you follow ethical guidelines and promote positive behavior."):
@@ -42,6 +43,7 @@ class IntelChatBot:
             temperature=temperature,
             top_p=top_p,
             repetition_penalty=repetition_penalty,
+            streamer=streamer,
             do_sample=do_sample
         )