Amazon Web Services has released TorchNeuron, a tool designed to let AI developers run PyTorch applications on Amazon's own Trainium chips, TechCrunch reports, in a direct challenge to Nvidia's software lead.
PyTorch is a free, open-source toolkit for building, training and running AI models, and Amazon estimates that about 85% of AI apps are written with it. Nvidia's CUDA software has long supported PyTorch closely, making its processors the default choice. Amazon says TorchNeuron lets PyTorch apps run on Trainium with a one-line configuration change and no rewrite of code.
AWS argues that startups also gain access to its wider cloud services and a single chip for both training and inference, which it says lowers costs. It cited voice AI firm Deepgram, which it said processes 70% more audio per dollar since moving to Trainium.
Source: TechCrunch
