https://huggingface.co/instructkr/lynn-7b-alpha

#16
by Haiart - opened

iMatrix quantization (preferably, of course) of this Mistral 0.2 based model, thank you again mradermacher.

Sure! This will be my first imatrix quant of a 7b, even - I normally don't bother imatrix quants for models < 20B.

Makes sense and I agree that it isn't worth it overall, but since this is a 32K context length model, it being iMatrix quantified makes it so that I have less to bother with the perplexity of the model in lower quantization in thus trading it for more context- since I do have low VRAM to play around and context length requires more the higher you go.

Good point :) In any case, the imatrix has been prepared and a server has been allocated, so expect the quants to show up within 24 hours at most (first the static ones followed by weighted ones).

mradermacher changed discussion status to closed

Sign up or log in to comment