Is sonata encoder satisfy transform-invariance?

I found that the pre-trained Sonata encoder produces different outputs depending on whether center-shift is applied. How can it be adapted for downstream tasks like matching and registration, where center-shift is usually not applicable? By the way, since lidar point cloud often don't have color, is there any solution for producing transform-invariance features without color?

No files, tests, or entry points are named. Start by reproducing the pretrained Sonata encoder's output under center-shift and no center-shift, then investigate whether transform-invariant downstream features can be produced without color; done would require a validated adaptation for matching or registration.

Written by the indexing model from the issue text.