Tweeted By @rapidsai

With @rapidsai, @huggingface, @PyTorch & @dask_dev we show how to scale state of the art #NLP. #strongertogether! Over 2x faster than prior #GPU implementations, 450x faster subword tokenizer, & most importantly better accuracy. Learn more https://t.co/onS8t75vPk & try it today!

— RAPIDS AI (@rapidsai) September 9, 2020