Hello, I would like to ask, what is the format of multi-label text classification pytorch for output data processing? I am a newbie and feel the lack of information.
A multi-label classification (in my understanding) means that each output can belong to more than a single class.
Is that correct or are you dealing with a multi-class classification, where each output belongs to exactly one particular target class?
For text classification I guess your output would most likely have the shape [batch_size, nb_classes, seq_length].
Depending on your actual use case (multi-label vs. multi-class):
- multi-label: your target should have the same shape as your output and you would use e.g.
nn.BCEWithLogitsLossas the criterion - multi-class: your target have the shape
[batch_size, seq_length]and contain the class indices in the range[0, nb_classes-1].