File size: 2,436 Bytes
3aaad64 cef3c7f 3aaad64 e431013 3aaad64 e431013 3aaad64 e431013 3aaad64 e431013 3aaad64 e431013 3aaad64 cef3c7f 3aaad64 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 |
Hugging Face's logo
---
language: sw
datasets:
---
# xlm-roberta-base-finetuned-swahili
## Model description
**xlm-roberta-base-finetuned-swahili** is a **Swahili RoBERTa** model obtained by fine-tuning **xlm-roberta-base** model on Swahili language texts. It provides **better performance** than the XLM-RoBERTa on text classification and named entity recognition datasets.
Specifically, this model is a *xlm-roberta-base* model that was fine-tuned on Swahili corpus.
## Intended uses & limitations
#### How to use
You can use this model with Transformers *pipeline* for masked token prediction.
```python
>>> from transformers import pipeline
>>> unmasker = pipeline('fill-mask', model='Davlan/xlm-roberta-base-finetuned-swahili')
>>> unmasker("Jumatatu, Bwana Kagame alielezea shirika la France24 huko <mask> kwamba hakuna uhalifu ulitendwa")
[{'sequence': 'Jumatatu, Bwana Kagame alielezea shirika la France24 huko Ufaransa kwamba hakuna uhalifu ulitendwa',
'score': 0.5077782273292542,
'token': 190096,
'token_str': 'Ufaransa'},
{'sequence': 'Jumatatu, Bwana Kagame alielezea shirika la France24 huko Paris kwamba hakuna uhalifu ulitendwa',
'score': 0.3657738268375397,
'token': 7270,
'token_str': 'Paris'},
{'sequence': 'Jumatatu, Bwana Kagame alielezea shirika la France24 huko Gabon kwamba hakuna uhalifu ulitendwa',
'score': 0.01592041552066803,
'token': 176392,
'token_str': 'Gabon'},
{'sequence': 'Jumatatu, Bwana Kagame alielezea shirika la France24 huko France kwamba hakuna uhalifu ulitendwa',
'score': 0.010881908237934113,
'token': 9942,
'token_str': 'France'},
{'sequence': 'Jumatatu, Bwana Kagame alielezea shirika la France24 huko Marseille kwamba hakuna uhalifu ulitendwa',
'score': 0.009554869495332241,
'token': 185918,
'token_str': 'Marseille'}]
```
#### Limitations and bias
This model is limited by its training dataset of entity-annotated news articles from a specific span of time. This may not generalize well for all use cases in different domains.
## Training data
This model was fine-tuned on [Swahili CC-100](http://data.statmt.org/cc-100/)
## Training procedure
This model was trained on a single NVIDIA V100 GPU
## Eval results on Test set (F-score, average over 5 runs)
Dataset| XLM-R F1 | sw_roberta F1
-|-|-
[MasakhaNER](https://github.com/masakhane-io/masakhane-ner) | 87.55 | 89.46
### BibTeX entry and citation info
By David Adelani
```
```
|