Right now we support a modes argument for some of the tucker functions, to allow them to operate with partial tucker decompositions. We might consider making this consistent so that this argument is carried across all of the functions.

See #593