Any chance for the 13b model?
Hi, thank you for the model, this works great, any chance you could provide also the 4bit quantized version for the 13b model as in https://huggingface.co/openlm-research/open_llama_13b_600bt?
I've been holding off because a) it's not finished training, and b) there are no instruction tuned versions of it yet (that I know of)
This 7B model has been Alpaca instruction fine tuned, so it's pretty good at answering questions and following instructions.
If I train that base Open Llama 13B model, it will be like Llama 13B in that it will generate text but not be very good at answering questions. But it'll actually be worse than that, because so far it's only been trained on 600B tokens, compared to the 1T tokens in Llama 13B.
So TBH I think it'll be pretty bad :(
Alright, let's hope they'll release it soon then! Thanks again for your work