Is it possible to get labels instead of pre-defining it in zero-shot classification?

sequence_to_classify = "one day I will see the world"
candidate_labels = ['travel', 'cooking', 'dancing']
classifier(sequence_to_classify, candidate_labels)

Hi, instead of using the transformer like this - I want to know if it is possible to give the sequence_to_classify and have the model output the labels it classified from the sequence instead of pre-defining it.
Is that possible?