Massively pre-trained transformer models such as BERT have gained great success in many downstream NLP tasks. However. they are computationally expensive to fine-tune. slow for inference. https://yeswellneses.shop/product-category/diffusers/