Overview
Who this competition fits
Advanced NLP practitioners interested in African language model compression. Focuses on creating compressed versions of InkubaLM for Swahili and Hausa using open-source tools only.
Read the original official blurb
How can a focused version of InkubaLM for Swahili and Hausa be achieved through model compression techniques? Official task labels include Natural Language Processing, Sentiment Analysis, Machine Translation, Natural Language Processing. Official impact tags mention Africa.
Preparation
From registration to a first submission
- 01
Python
- 02
NLP fundamentals
- 03
model compression techniques
- 04
transformer architectures
- 05
open-source ML tooling
Before you commit: Requires model compression expertise (distillation, pruning, quantization) for multilingual NLP. Working with low-resource African languages adds complexity. Access is restricted and the hackathon format has a compressed 1-week timeline.
Source
How this page was assembled
Competition information is structured from the official page. Scores are platform estimates for decision support; official rules take precedence.
- Official competition page
- Zindi
- Last checked
- Not recorded