Padding mask

[abrahamezzeddine]

Hello,

I am wondering if this model supports tensor padding when downscaling images to the ViT resolution model? If I use too large images, there are artefacts which don't appear if I try to use similary downscaled images. But for some images, it's difficult to uniformely downsample to the ViT resolution here.

Does this model support masking outer edges to not compute normals for example? Thanks!