This is still a test model under development (Arabic ColBERT for information retrieval).
When ready (partially or fully), this note will be updated accordingly.
For training, the akhooli/ar_mmarco_queries dataset and two others, mentioned in the dataset card, are used.