Update app.py

app.py

@@ -6,6 +6,9 @@ from transformers import TextStreamer
 
 import spaces
 
+tokenizer = AutoTokenizer.from_pretrained("Rorical/0-roleplay", trust_remote_code=True)
+tokenizer.chat_template = "{% for message in messages %}{{'<|im_start|>' + ((message['role'] + ':\n') if message['role'] != '' else '') + message['content'] + '<|im_end|>' + '\n'}}{% endfor %}{% if add_generation_prompt %}{{ '<|im_start|>星野:\n' }}{% endif %}" # Be careful that this model used custom chat template.
+
 # Define the response function
 @spaces.GPU
 def respond(
@@ -18,8 +21,6 @@ def respond(
 ):
     # Load model and tokenizer
     model = AutoModelForCausalLM.from_pretrained("Rorical/0-roleplay", return_dict=True, trust_remote_code=True)
-    tokenizer = AutoTokenizer.from_pretrained("Rorical/0-roleplay", trust_remote_code=True)
-    tokenizer.chat_template = "{% for message in messages %}{{'<|im_start|>' + ((message['role'] + ':\n') if message['role'] != '' else '') + message['content'] + '<|im_end|>' + '\n'}}{% endfor %}{% if add_generation_prompt %}{{ '<|im_start|>星野:\n' }}{% endif %}" # Be careful that this model used custom chat template.
     
     # Construct the messages for the chat
     messages = [{"role": "", "content": system_message}]