Hacker News new | ask | show | jobs
by doubtfuluser 5 days ago
But why using an encoder model instead of a BERT based model? For a pure classification that should be easier to train and work quite well