we modified the "diff-gaussian-rasterization" in the submodule to get the Global Significant Score. Users can also train from scratch and jointly prune redundant Gaussians in training using the ...