Is it possible to make the CNNs train sequentially, or to sample an entire space?

[DavidRavnsborg]

Hello,

My dataset is kind of sparse - 3D vectors in a high dimensional space - where each dimension is very short in length. Some of my dimensions have less than a length of 5. I would like to have a CNN train on the entire space, as you would for a dense dataset. Is this possible, using MinkowskiEngine? Having it train sequentially is less important than having it train on the entire space.