😅 what implementation is that?

Reply to this note

Please Login to reply.

Discussion

This is llama.cpp with the Vicuna weights which is why it asks itself questions. Also works on pretty much any LLaMA weights unless they're lobotomized ones.

Here's the full command I ran with the initial prompt and flags.

Take particular note of the --keep -1 flag at the end. Without that it forgets the persona you give it in the starting prompt after the first question.