You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
theoretically, we could write a generic loader using the adapters which return a 2D Array (most likely of object dtype)
each label column would become one label in the second axis
while this would be nifty, I am unsure it would be that useful in practice
for smallish dataset we could compute the cartesian_product of labels
this would be a lot more useful but absolutely requires a check to avoid trying to load structure where the cartesian product would not fit in memory (e.g. large dataframe-with-multi-index structures).
what we really need is a way to load the formats we wrote ourselves/which can be loaded currently (.h5, directory of .csv and .xlsx)