Text this: Assessing the Efficiency of Transformer Models with Varying Sizes for Text Classification