π“Ÿ host1.bitchan.one

𓀐 @WkuT*V*B

⏱ 6d seenβ›“ 0πŸ‘ 0β§– 1

filter [-]

languages:onother

27b doesnt make much sense anyway, why bother training a dense model if qfn beats it in most cases and also doesn't hardlock poorbros with 8-16gb vram? 2 sticks of ddr4 are far cheaper than an old flagship gpu or even mi50⏱ 5d↩