Details Model trained with train_size=0.4 or 40% of the Mozilla Common Voice Arabic training+validation dataset.