谷歌(微博)自主开发定制化芯片,以加速其机器学习算法,这已不是什么秘密了。早在2016年5月,该公司就在其I/O开发者大会上首次公布了这款名为Tensor Processing Units(简称TPU)的芯片。但是,该公司从未详细地介绍过它们,只是声称这些芯片是专门为其Tensor Flow机器学习框架度身定做的。但是在周三,谷歌首次详细地介绍了这种芯片的相关信 It’s no secret that Google (microblogging) has developed its own custom silicon to speed up its machine learning algorithms. As early as May 2016, the company debuted the chip called the Tensor Processing Units (TPU) at its I / O developer conference. However, the company never introduced them in detail, only claiming that the chips are tailor-made specifically for their Tensor Flow machine learning framework. But on Wednesday, Google introduced for the first time details about the chip