AuroraNemoia’s Twitter Archive—№ 1,283

@_airpods69 just messed with cuda and llamacpp compile flags, and gave it all 32 threads on both xeons for prompt eval, its running a 4bit 7b as I always do.