File size: 1,218 Bytes
1d9dd26 bb45172 1d9dd26 bb45172 1d9dd26 |
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 |
---
language:
- zul
thumbnail: https://pbs.twimg.com/media/EVjR6BsWoAAFaq5.jpg
tags:
- zul
- fill-mask
- pytorch
- roberta
- masked-lm
license: mit
---
# Takalani Sesame - Zulu 🇿🇦
<img src="https://pbs.twimg.com/media/EVjR6BsWoAAFaq5.jpg" width="600"/>
## Model description
Takalani Sesame (named after the South African version of Sesame Street) is a project that aims to promote the use of South African languages in NLP, and in particular look at techniques for low-resource languages to equalise performance with larger languages around the world.
## Intended uses & limitations
#### How to use
```python
from transformers import AutoTokenizer, AutoModelWithLMHead
tokenizer = AutoTokenizer.from_pretrained("jannesg/takalane_zul_roberta")
model = AutoModelWithLMHead.from_pretrained("jannesg/takalane_zul_roberta")
```
#### Limitations and bias
Updates will be added continously to improve performance.
## Training data
Data collected from [https://wortschatz.uni-leipzig.de/en](https://wortschatz.uni-leipzig.de/en) <br/>
**Sentences:** 410000
## Training procedure
No preprocessing. Standard Huggingface hyperparameters.
## Author
Jannes Germishuys [website](http://jannesgg.github.io)
|