Friendly reminder the transformer model is only as good as the data it’s trained on