Artefact2

/

CodeLlama-70b-Instruct-hf-GGUF

like 2

Thanks for these quants. CodeLlama actually seems like a very competent model, but it derails any longer convo and really wants to ramble about morals or ethics for some reason. "As an AI language model...". Just strange. Too much training on Claude or GPT4 data probably.

https://www.reddit.com/r/LocalLLaMA/comments/1af687d/codellama_70b_pontificates_on_ethics_where_13b/