Computation Cost

[chenguolin]

Thanks for open-sourcing this wonderful work.

I'm curious about the computation cost for training your LRM on the Objaverse dataset.

For example, it takes 128 A100-40G GPUs for 3 days to complete training on the Objaverse + MVImgNet dataset in the original paper.

[ZexinHe]

Thanks for your interest!
It takes 64 A100-80G GPUs for 5-6 days.

[chenguolin]

How about the computation cost for the base model?

(I guess "64 A100-80G GPUs for 5~6 days" is for the large model).

Thanks!

[ZexinHe]

Hi @chenguolin, apologize for the late reply.

The updated v1.0 models are trained as follows.

For small model, 32 A100-80G for 1.5 days.
For base model, 32 A100-80G for 2 days.
For large model, 64 A100-80G for 5 days.