Hacker News new | past | comments | ask | show | jobs | submit login

Seems like 6B would still be useful if I want to run it on my GPU without exiting firefox.



Kinda lame that applications can't be told "yo, your gpu buffer has now been moved back to RAM".


With Apple Silicon there is no separation between GPU RAM and CPU RAM, so the whole model can be loaded into RAM and processed by GPU (Metal).




Consider applying for YC's Spring batch! Applications are open till Feb 11.

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

Search: