A community-driven project for experimenting with large language models for indigenous Ghanaian languages.
As research around large language models (LLMs) continues to evolve, new opportunities are emerging to address long-standing challenges faced by low-resource languages. The GhanaLLM Project aims to establish a common reference point and standardized training approach for developing LLMs capable of performing a wide range of tasks in Ghanaian languages — all built upon base models that have strong support for Ghanaian languages and powered by community support and experience sharing.
| Model | Training Dataset | Description | Demo | Creator |
|---|---|---|---|---|
| Opani Coder | Code-170k-twi | Provides coding assistance in Twi | Demo | Mich-Seth Owusu |
| Opani Translate | english-twi-translate-llm-instruct-160k | Provides translation from English to Twi | Demo | Mich-Seth Owusu |
-
Create a Hugging Face account and obtain your username and access token. You will need this for getting access to any of our community datasets anfd publishing your model.
-
Request access to a community dataset
You can visit our community page on Hugging Face to request access to the datasets in the Ghana LLM collection for training models.
Before requesting access, please read our contribution guidelines.
-
Get started training! Open one of our ready-to-use notebooks:
💻 Train a Coding Assistant Model —| 🎥 Watch the tutorial
-
Once your model is live, share the URL and we’ll feature it on the GhanaLLM showcase!
If you have any questions or diffculty using the notebooks, please send a message to our Community Lead - Mich-Seth Owusu michsethowusu@gmail.com.
GhanaLLM is a community project dedicated to advancing natural language processing for Ghanaian languages.