KBLab/megatron-bert-large-swedish-cased-165-zero-shot

Megatron-BERT-large Swedish 165k for zero-shot classification

This model is based on Megatron-BERT-large-165k (https://huggingface.co./KBLab/megatron-bert-large-swedish-cased-165k). It was fine-tuned on the QNLI task and further fine-tuned on the MNLI task. The model can be used with the Hugging Face zero-shot classification pipeline.

You can read more about the model on our blog.

Usage

>>> from transformers import pipeline
>>> classifier = pipeline(
...     "zero-shot-classification",
...     model="KBlab/megatron-bert-large-swedish-cased-165-zero-shot"
... )
>>> classifier(
...     "Ruben Östlunds ”Triangle of sadness” nomineras till en Golden Globe i kategorin bästa musikal eller komedi.",
...     candidate_labels=["hälsa", "politik", "sport", "religion", "nöje"],
...     hypothesis_template="Detta exempel handlar om {}.",
... )
{'sequence': 'Ruben Östlunds ”Triangle of sadness” nomineras till en Golden Globe i kategorin bästa musikal eller komedi.',
 'labels': ['nöje', 'sport', 'religion', 'hälsa', 'politik'],
 'scores': [0.9274595379829407,
  0.025105971843004227,
  0.018440095707774162,
  0.017049923539161682,
  0.011944468133151531]}

Citation

@misc{sikora2023swedish,
  author = {Sikora, Justyna},
  title = {The KBLab Blog: Swedish zero-shot classification model},
  url = {https://kb-labb.github.io/posts/2023-02-12-zero-shot-text-classification/},
  year = {2023}
}