A bit of a beginner question here. Is there a Fast.ai regression learner for text data? I have a dataset of internet comments and they are labeled on a continuous scale from 0 to 1 based on their level of helpfulness, toxicity and other metrics. Can I train a model to do this type of regression directly or should I transform the continuous labels into fixed classes by range and do classification instead with text_classifier_learner?
There is a very detailed discussion here: Regression using Fine-tuned Language Model
Looks like you can work with the text_classifier_learner and change the loss functions and it should work fine!
1 Like