But sometimes output more broken tokens, I don't know why.
Thanks! This must be due to issues with the tokenizer that is not properly used. Could you send an example to help me debug it?
· Sign up or log in to comment
