How to use it for different classes?
Can it only return positive, negative as result or can it be used (without finetuning) to return positive, negative, neutral as well? If that is possible, then how ca it be used to classify a text according to a list of classes for example:
classes = ["feeling", "computer", "books", "media"]
and classify this "This is a sunny day".
Or even use it for multiclass classification like for sentences "nice day for fishing", "I love fixing bugs" and it would return feeling and computer, how can that be done with this model?
Thank you
Hi
@supercharge19
According to the model card, that model has been fine-tuned on the Stanford SST dataset: https://huggingface.co/datasets/sst2 which consists of a dataset of 2 classes only (negative, positive). For extending distil-bert to new classes, you need to fine-tune that model on your own dataset or an existing multi-class dataset