README.md

---
pipeline_tag: token-classification
tags:
- named-entity-recognition
- sequence-tagger-model
widget:
- text: Мене звуть Амадей Вольфганг, я живу в Берліні
inference:
  parameters:
    aggregation_strategy: simple
    grouped_entities: true
language:
- uk
---

xlm-roberta model trained on [ukrainian ner](https://github.com/lang-uk/flair-ner) dataset from flair

| Test metric             | Results                   |
|-------------------------|---------------------------|
| test_f1_mac_ukr_ner     | 0.9900672435760498        |
| test_loss_ukr_ner       | 0.054602641612291336      |
| test_prec_mac_ukr_ner   | 0.9386032819747925        |
| test_rec_mac_ukr_ner    | 0.9383019208908081        |



```python
from transformers import AutoTokenizer, AutoModelForTokenClassification
from transformers import pipeline

tokenizer = AutoTokenizer.from_pretrained("EvanD/xlm-roberta-base-ukrainian-ner-ukrner")
ner_model = AutoModelForTokenClassification.from_pretrained("EvanD/xlm-roberta-base-ukrainian-ner-ukrner")

nlp = pipeline("ner", model=ner_model, tokenizer=tokenizer, aggregation_strategy="simple")
example = "Мене звуть Амадей Вольфганг, я живу в Берліні"

ner_results = nlp(example)
print(ner_results)
```