Multi Label Problem

What is the best way to implement multi label where there can be multiple output prediction for each input (1 or more) in Mxnet ? I am working on multi class problem which can have multi output labels.

There is a multi class text classification using CNN

That is a multi class classification . My problem is that each output can have multiple predictions.

use multiple loss functions

loss1 = Loss1(output, label1)
loss2 = Loss2(output, label2)
loss = mx.sym.Group([loss1, loss2])