but we want to keep the rest of the process intact.
we'll have to:
add in "validationSplit" as a column type in data-formatter
remove that column from the data otherwise
save that column in it's own separate entity, like we do with the id column
set a flag for "custom validation split".
make sure this flag gets saved with fileNames.
once in the file where we create the validations, first check this flag.
if it's set, then just create the validation split from the validationSplitColumn.npz file. don't worry about running checks against data length or any of that.
from there, we should just be able to follow the normal process.
but we want to keep the rest of the process intact.
we'll have to:
from there, we should just be able to follow the normal process.