If it uses 2 experts, they should parallelize so closer to 7B speed? | Hacker News

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

visarga on Dec 11, 2023 | parent | context | favorite | on: Mixtral of experts

If it uses 2 experts, they should parallelize so closer to 7B speed?

knd775 on Dec 11, 2023 [–]

Memory bandwidth is still a factor, right?

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