README.md

metadata

language:
  - tn
thumbnail: https://pbs.twimg.com/media/EVjR6BsWoAAFaq5.jpg
tags:
  - tn
  - fill-mask
  - pytorch
  - roberta
  - masked-lm
license: mit

Takalani Sesame - Tswana 🇿🇦

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

from transformers import AutoTokenizer, AutoModelWithLMHead

tokenizer = AutoTokenizer.from_pretrained("jannesg/takalane_ssw_roberta")

model = AutoModelWithLMHead.from_pretrained("jannesg/takalane_ssw_roberta")

Limitations and bias

Updates will be added continously to improve performance.

Training data

Data collected from https://wortschatz.uni-leipzig.de/en
Sentences: 380

Training procedure

No preprocessing. Standard Huggingface hyperparameters.

Author

Jannes Germishuys website