2 bits?
#2
by
groxaxo
- opened
thanks a lot for the quants ! any chance on getting the q2 bits for us gpu poor fellows please ?
Hi! sorry for the late respone
This prune in particular is not the great and breaks fast on lower bits so 2 would be mostly not worth it.
If there is a big regular model out there for which you lack a particular bpw to fit feel free to make a request (under particular quant repo or general quant requests)
bpws outside of the common range are of low priority but may still be considered depending on various circumstances (available compute, model size, popularity, etc)