3 datasets found

ARAGPT2 is a stacked transformer-decoder model trained using the causal language modeling objective. The model is trained on 77GB of Arabic text.

The Validation Dataset is used for validation, it contains 1428 images from nine distinct rooms.

The training dataset is a collection of the publicly available Arabic corpora listed below: The unshuffled OSCAR corpus (Ortiz Su´arez et al., 2020). The Arabic Wikipedia dump...