Datasets:
Tasks:
Token Classification
Modalities:
Text
Formats:
parquet
Sub-tasks:
named-entity-recognition
Size:
10K - 100K
License:
Convert dataset to Parquet
#3
by anrilombard - opened
Convert dataset to Parquet.
Hi @Davlan 🙂 hope you're well! Since datasets 4.0 dropped loading scripts, masakhapos no longer loads, which breaks the AfroBench POS tasks in lm-evaluation-harness (100 tasks, see EleutherAI/lm-evaluation-harness#3171). This is the standard HF Parquet conversion: same configs, splits, columns and rows, just without the script. I checked every config and split against what the script yields and they're identical.
Would you mind merging this when you have a moment, or pointing me to whoever looks after the masakhane org? There are matching PRs on masakhaner-x and afriqa-gold-passages 🙏
Davlan changed pull request status to merged