Currently, it looks like all the `bvecs` are corrupted (there's randomly an affine at the top, and values of 0 and 100 at the bottom) Needs to be fixed. Then we can run KKI again. We might just have to redownload the KKI dataset completely.