Extension to 3d

[anwai98]

Hi team,

First of all, thanks for the awesome work. We have had good share of experience in using Vim for some preliminary experiments (focused on downstream tasks, eg. segmentation).

I have a question: in some research domains, learning over multi-dimensional inputs is of great interest (eg. 3d medical images). Do you plan to extend Vim to 3d? (or have already extended it - apologies if I missed it)

Ideally speaking, it would be nice to have Vim as the encoder backbone for encoder-decoder-style image-image tasks and users can just plug it in their downstream applications (which is something we do already and prefer to continue doing).

Would love to hear from you on this!

Thanks in advance.