Google宣布推出PyTorch的定制版本TorchTPU，使AI应用程序能够在Google的张量处理器（TPU）上运行 1。这一举措与亚马逊周一发布的TorchNeuron相呼应，两家科技巨头正试图打破英伟达在AI芯片市场的主导地位 1。鉴于PyTorch被85%的AI应用所使用，这两项发布针对的是开发者生态中最广泛采用的框架 1。
TorchTPU的设计目标是降低开发者的迁移成本。通过简单的配置更改，现有的PyTorch应用无需重写代码即可移植到TPU上运行 1。与此同时，亚马逊正进一步向PyTorch主项目提交代码，将功能与英伟达芯片解耦 1。Google还推出了TPU agent suite工具和更新了LiteRT工具，以进一步简化开发者将应用部署到Google芯片和云平台的过程 1。为确保兼容性和推动行业采纳，Google与Meta、Hugging Face和Inferact等合作伙伴展开合作 1。
Google has announced the launch of TorchTPU, a custom version of PyTorch designed to enable AI applications to run on Google's Tensor Processing Units (TPUs) 1. This move mirrors Amazon's Monday release of TorchNeuron, as both technology giants seek to break Nvidia's grip on the AI chip market 1.
TorchTPU allows developers to port existing PyTorch applications to Google's TPU infrastructure through simple configuration changes, without requiring code rewrites 1. Given that PyTorch is used by 85% of AI applications, this compatibility layer significantly lowers the barrier to migration 1. Google has also introduced the TPU agent suite toolset and updated LiteRT tools to facilitate smoother transitions for developers moving their applications to Google's chips and cloud platform 1.
Amazon has submitted code to the PyTorch mainline project to decouple functionality from Nvidia chips 1. To ensure broad compatibility, Google is collaborating with Meta, Hugging Face, Inferact, and other partners 1.
评论
还没有评论，欢迎留下第一条。