README.md

metadata

dataset_info:
  features:
    - name: text
      dtype: string
    - name: source
      dtype: string
  splits:
    - name: train
      num_bytes: 1935476.2813211845
      num_examples: 8428
    - name: test
      num_bytes: 484098.71867881546
      num_examples: 2108
  download_size: 1259998
  dataset_size: 2419575
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: test
        path: data/test-*
license: apache-2.0
task_categories:
  - text-generation
language:
  - ha
pretty_name: Hausa DataMix

This dataset is created using the composition of diverse Hausa datasets available on Hugging Face. Each entry in a given split is composed of a text and source column. The text contains the textual data, meanwhile the source contains the Hub ID from which this data sample has been retrieved from.

This dataset was created in the spirit of the HausaLM project.