Live demo: 1T CoE model with 330T/s on SambaNova Suite

Posted by nsup on Thursday March 28, 2024 @05:55PM
from the coe-1.cloud.snova.ai dept.

( 2 comments | 3 points | on HN )

Re:Live demo: 1T CoE model with 330T/s on SambaNova Suite by jbrot on Thursday March 28, 2024 @06:34PM (#39858212)

One of the coolest things about this demo is that the token generation runs on a single node using just 8 RDUs (SambaNova’s chip).

Disclosure: I work at SambaNova and was part of the team that made this

[ Parent | on HN ]

Re:Live demo: 1T CoE model with 330T/s on SambaNova Suite by nsup on Thursday March 28, 2024 @05:55PM (#39857825)

[ Parent | on HN ]