q3_k_m?
#2
by
JoggyMuffin
- opened
Would you consider adding a Q3_k_m quant? it seems like a pretty good memory size to perplexity tradeoff from the charts I've seen, especially with the new quant method, and q4 is a wee bit too big for my hardware
Artefact2
changed discussion status to
closed
Thanks, much appreciated!