multi-class and multi-label classification
Hi!
I want to make multi-label text classification with your model. I check your tutorials https://github.com/MoritzLaurer/transformers-workshop-comptext-2023/tree/master . It's really awesome, and I finally understand what's going on. Thank you!
But I still don't know how to set 2 or 3 labels to one sample and get multi-label text classification?
Could you help, please?
good question, I don't have a multi-label example in the summer school repo. The main difference is in the way the training data is collated: When there is a text for which two labels are true, I would create three rows in the training data: two rows with the two true hypotheses with the label true and one row with a random hypothesis for a label that is not true.
That's for training. At inference time, you can just set multi_label=True in the HF pipeline.