magic-research/PLLaVA

About vram and int-4 version

Opened this issue · 0 comments

joseph16388 commented
  1. Can a quantized int-4 version of the model be produced later?
    2, Can 16G VRAM run 7B version of the model?

thanks!