Surprised they hadn't found ways to advance their techniques with e.g. low-rank ... | Hacker News

Hacker Newsnew | past | comments | ask | show | jobs | submit

		uoaei on March 28, 2024 \| parent \| context \| favorite \| on: Jamba: Production-grade Mamba-based AI model Surprised they hadn't found ways to advance their techniques with e.g. low-rank approximations, etc.

theGnuMe on March 29, 2024 [–]

That’s one strategy. Also flash attention.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact