Abstract
Google's TPU supercomputers train deep neural networks 50x faster than general-purpose supercomputers running a high-performance computing benchmark.
小提示:本篇文献需要登录阅读全文,点击跳转登录

