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)