boost inference speed of T5 models by 5x & reduce the model size by 3x using fastT5.
conda install selfexplainml::fastt5