magic-research / PLLaVA

Official repository for the paper PLLaVA
568 stars 37 forks source link

Training time consumption #51

Closed hmxiong closed 3 months ago

hmxiong commented 4 months ago

Awesome, this work is existing and meaningful!! Could you please tell me how much time and GPU it takes to train the entire model under the current settings? I like this job very much, but I may not have enough resources to implement it.

ermu2001 commented 4 months ago

https://github.com/magic-research/PLLaVA/issues/1#issuecomment-2082587647

The largest model costs around 48 A100 GPU days.

The smaller models costs less. cc @cathyxl

Nastu-Ho commented 4 months ago

How long does it take to train the 7B model?