Lugha-Llama/Lugha-Llama-8B-wura_edu

TEXT GENERATIONConcurrency Cost:1Model Size:8BQuant:FP8Ctx Length:32kPublished:Dec 13, 2024License:llama3.1Architecture:Transformer0.0K Cold

Lugha-Llama/Lugha-Llama-8B-wura_edu is an African-centric language model developed by Happy Buzaaba, Alexander Wettig, David Ifeoluwa Adelani, and Christiane Fellbaum, based on Llama-3.1-8B. This model is specifically adapted for African languages, addressing their underrepresentation in large language model training datasets. It achieves leading performance among open-source models on the IrokoBench and AfriQA benchmarks, making it ideal for applications requiring strong understanding and generation in low-resource African languages.

Loading preview...

Lugha-Llama: Adapting LLMs for African Languages

Lugha-Llama models, developed by Happy Buzaaba, Alexander Wettig, David Ifeoluwa Adelani, and Christiane Fellbaum, are a series of African-centric language models built upon the Llama-3.1-8B architecture. These models specifically address the challenge of underrepresentation of low-resource African languages in mainstream large language model training datasets.

Key Capabilities

  • Enhanced African Language Understanding: Designed to overcome the limitations of general LLMs in comprehending African languages.
  • Leading Benchmark Performance: Achieves the best performance among open-source models on the challenging IrokoBench benchmark.
  • Cross-lingual Question Answering: Demonstrates strong capabilities on AfriQA, a cross-lingual open-retrieval question answering dataset for African languages.

Good For

  • Applications requiring robust language understanding and generation in various African languages.
  • Research and development focused on improving LLM performance for low-resource linguistic communities.
  • Projects needing models with specialized knowledge and performance in African language contexts, as detailed in the Lugha-Llama paper.