Sparse multi-label classification

[usurniro]

I am interested to run Crossbow on the extreme classification amazon-670k dataset (https://www.kaggle.com/c/extreme-classification-amazon/data). In order to do that, I am interested to know if Crossbow supports sparse data (tensors) and multiple labels. So, are these supported? If not, how hard is it to add support for these features?