Update app.py

app.py

@@ -1,7 +1,6 @@
 import os
 import gradio as gr
-from transformers import AutoModelForCausalLM, AutoTokenizer, GenerationConfig
-from transformers import IteratorStreamer
+from transformers import AutoModelForCausalLM, AutoTokenizer, GenerationConfig, TextIteratorStreamer
 import torch
 from threading import Thread
 from huggingface_hub import Repository
@@ -33,10 +32,7 @@ else:
     model = AutoModelForCausalLM.from_pretrained(model_id, torch_dtype=torch_dtype, device_map="auto")
 tokenizer = AutoTokenizer.from_pretrained(model_id)
 
-prompt_template = f"""### Anweisung:
-{{input}}
-
-### Antwort:"""
+prompt_template = f"### Anweisung:\n{{input}}\n\n### Antwort:"
 
 
 def generate(instruction, temperature, max_new_tokens, top_p, length_penalty):
@@ -66,7 +62,7 @@ def generate(instruction, temperature, max_new_tokens, top_p, length_penalty):
     # STREAMING BASED ON git+https://github.com/gante/transformers.git@streamer_iterator
 
     # streaming
-    streamer = IteratorStreamer(tokenizer)
+    streamer = TextIteratorStreamer(tokenizer)
     model_inputs = tokenizer(formatted_instruction, return_tensors="pt", truncation=True, max_length=2048)
     # move to gpu
     model_inputs = {k: v.to(device) for k, v in model_inputs.items()}