Aran Komatsuzaki   @arankomatsuzaki   6/10/2021       

FastSeq: Make Sequence Generation Faster Demonstrates 4-9x inference speed gain on various Transformer-based models with a series of optimization methods. abs: https://t.co/U5opo0VxSf code: https://t.co/oH0GCkadYD

 Reply  0     Retweet   16      Like   88





Posted by Aran Komatsuzaki