The main change in this PR is to reshuffle how Data works. Rather than every split causing a deepcopy of the X and y matrices, it now only creates new masks onto a single instance of X and y. This makes the process much faster and lighter on memory
There's also an embarrassing splodge of other changes at the same time :(
The main change in this PR is to reshuffle how Data works. Rather than every split causing a deepcopy of the X and y matrices, it now only creates new masks onto a single instance of X and y. This makes the process much faster and lighter on memory
There's also an embarrassing splodge of other changes at the same time :(