Fast Transformer Decoding: One Write-Head Is All You Need | Hacker News

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

		Fast Transformer Decoding: One Write-Head Is All You Need (arxiv.org)
		4 points by hislaziness on May 30, 2023 \| hide \| past \| favorite \| 1 comment

hislaziness on May 30, 2023 [–]

A more efficient way to infer with lower memory requirements.

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