-
I'm trying to use the new Command-R +, following the template:
I can't seem to just type this as text and have it work, because I don't think (for example) <|START_OF_TURN_TOKEN|> is being turned into a single token, but instead is becoming a string of normal text tokens. Am I missing a way to do this, or is it not possible in llama.cpp interactive mode? |
Beta Was this translation helpful? Give feedback.
Replies: 2 comments
-
Yes, you're not missing anything - it's fundamental to how For |
Beta Was this translation helpful? Give feedback.
-
Thanks! And I appreciate the links too, @Jeximo. |
Beta Was this translation helpful? Give feedback.
Yes, you're not missing anything - it's fundamental to how
main
works.There is a way to add the template for
server
withinlama_chat_apply_template
, here's the detail: https://github.com/ggerganov/llama.cpp/wiki/Templates-supported-by-llama_chat_apply_template#how-to-add-a-new-templateFor
main
,llama_chat_apply_template
is not implemented yet: #6391 (comment)