Remove initial bos

In this case, the tokenizer runs in the server.

app.py

@@ -36,7 +36,7 @@ examples=[
 def predict(message, chatbot):
     # The first user message is _not_ stripped
     do_strip = False
-    input_prompt = f"<s>[INST] <<SYS>>\n{system_message}\n<</SYS>>\n\n "
+    input_prompt = f"[INST] <<SYS>>\n{system_message}\n<</SYS>>\n\n "
     for interaction in chatbot:
         user_message = str(interaction[0])
         user_message = user_message.strip() if do_strip else user_message