thickertoofan@lemm.ee to LocalLLaMA@sh.itjust.worksEnglish · 1 year agoMicrosoft just released BitNet!github.comexternal-linkmessage-square6linkfedilinkarrow-up119arrow-down10file-textcross-posted to: technology@lemmy.ziptechnology@hexbear.nettechnology@lemmy.ml
arrow-up119arrow-down1external-linkMicrosoft just released BitNet!github.comthickertoofan@lemm.ee to LocalLLaMA@sh.itjust.worksEnglish · 1 year agomessage-square6linkfedilinkfile-textcross-posted to: technology@lemmy.ziptechnology@hexbear.nettechnology@lemmy.ml
minus-squareWildPalmTree@lemmy.worldlinkfedilinkEnglisharrow-up7·1 year agoELI5 1-bit module. With three attempts, i got nothing out of it, so I assume it’s a simpler, more energy efficient model.
minus-squareicecreamtaco@lemmy.worldlinkfedilinkEnglisharrow-up5·1 year agoIt’s a massive performance upgrade, which would make current sized models better and tiny phone-sized models viable. Only problem is that models need to be retrained to use it and afaik, no one significant has done it yet.
minus-squarethickertoofan@lemm.eeOPlinkfedilinkEnglisharrow-up4·1 year agoi’m not the smartest out there to explain it but it’s like …instead of floating point numbers as the weights, its just -1,0,1.
ELI5 1-bit module. With three attempts, i got nothing out of it, so I assume it’s a simpler, more energy efficient model.
It’s a massive performance upgrade, which would make current sized models better and tiny phone-sized models viable. Only problem is that models need to be retrained to use it and afaik, no one significant has done it yet.
i’m not the smartest out there to explain it but it’s like …instead of floating point numbers as the weights, its just -1,0,1.