Hacker News new | ask | show | jobs
by matanyall 908 days ago
Yeah, that's raw llama2 for you. Fine tuning, LoRa, that's how you get the nice responses.
1 comments

I suspect this is Llama-2-chat and not the base model. It's very attuned to safety fine tuning, and it isn't having issues with completion in a chat format.

They probably weren't specific enough in mentioning what the model it was built on was, referring to Llama-2-chat as being a Llama-2 model (which is kind of correct).