Details Model trained with train_size=0.6 or 60% of the Mozilla Common Voice Arabic training+validation dataset.