Open-Assistant
Open-Assistant copied to clipboard
Forced Message Continuation
A button could be added next to the regenerate respond button which removes the end of text token and then passes the message back to inference in order to do more reliable message continuation. As of now message continuation is typically done just by saying "continue" which is a highly unreliable way to do so as often it continues in a way which doesn't match the context correctly. This is especially evident when it is writing even medium length stories which just get cut off near the end and you can't reliably get it to finish, even if it had started the conclusion.