☆ Yσɠƚԋσʂ ☆@lemmygrad.ml to technology@hexbear.netEnglish · 15 days agoJet-Nemotron is a new family of hybrid-architecture language models with 53x faster generation and 6x prefillinglemmygrad.mlimagemessage-square8linkfedilinkarrow-up120arrow-down10file-text
arrow-up120arrow-down1imageJet-Nemotron is a new family of hybrid-architecture language models with 53x faster generation and 6x prefillinglemmygrad.ml☆ Yσɠƚԋσʂ ☆@lemmygrad.ml to technology@hexbear.netEnglish · 15 days agomessage-square8linkfedilinkfile-text
minus-square☆ Yσɠƚԋσʂ ☆@lemmygrad.mlOPlinkfedilinkEnglisharrow-up7·15 days agoYeah, it seems like this eliminates the need to train models from scratch in most cases, which is huge.
Yeah, it seems like this eliminates the need to train models from scratch in most cases, which is huge.