☆ Yσɠƚԋσʂ ☆@lemmygrad.ml to technology@hexbear.netEnglish · 5 months agoKimi Linear is a hybrid linear attention architecture that outperforms traditional full attention methods across various contextsgithub.comexternal-linkmessage-square3linkfedilinkarrow-up117arrow-down10cross-posted to: [email protected][email protected][email protected]
arrow-up117arrow-down1external-linkKimi Linear is a hybrid linear attention architecture that outperforms traditional full attention methods across various contextsgithub.com☆ Yσɠƚԋσʂ ☆@lemmygrad.ml to technology@hexbear.netEnglish · 5 months agomessage-square3linkfedilinkcross-posted to: [email protected][email protected][email protected]
minus-squareBountifulEggnog [she/her]@hexbear.netlinkfedilinkEnglisharrow-up2·5 months ago48b 3a is pretty good sized, hope Llama cpp can run it soon ish
48b 3a is pretty good sized, hope Llama cpp can run it soon ish