Lyte/Moroccan-Darija-Instruct-573K

Created by Lyte If you use this dataset, please credit: Lyte/Moroccan-Darija-Instruct-573K A synthetic instruction-tuning dataset of 573,175 question-answer pairs written entirely in Moroccan Darija (الدّارجة المغربية). Metric Value Total rows 573,175 Original pairs 300,194 Augmented variants 272,981 Total words ~19.9M Est. tokens ~13.3M File size 399 MB MSA contamination 0% Exact duplicates