alt Hacker News>With BERT, you need a large, labeled dataset, and you have to train/fine-tune the model.
I think they were responding to this. You can use BERT to provide zero shot classification predictions.
>With BERT, you need a large, labeled dataset, and you have to train/fine-tune the model.
I think they were responding to this. You can use BERT to provide zero shot classification predictions.
I guess I assumed they meant BERT, not some specific BERT-base model. Vanilla BERT does not support zero shot.