Technology and Scienceupdated 1h ago2 min read
Amazon Launches TorchNeuron To Run PyTorch On Trainium Without Rewriting Code
This article is an automated aggregation of coverage from news outlets. Our editorials offer in-depth analysis curated by the NewsCord team.
TorchNeuron runs PyTorch code on Trainium with a single configuration change. Challenges Nvidia's software and hardware dominance in AI compute.
9 outlets told this the same way.
TorchNeuron Targets CUDA Lock-In
Amazon Web Services launched TorchNeuron, a software tool meant to let developers run code written in PyTorch on Amazon’s Trainium chips with just one configuration change, without rewriting existing applications. Colin Brace, vice president of software at Annapurna Labs, said the initiative was intended to simplify the process of testing and adopting Amazon’s in-house chips, telling Reuters, “This has been about meeting customers where they're at.” Amazon said it would unveil the software at a developer conference in Silicon Valley on Monday and release it publicly as a free, open-source offering, positioning the move as an alternative to Nvidia’s AI computing technology.
Amazon’s pitch centered on Nvidia’s CUDA software platform, which Nvidia has spent more than a decade developing and refining to enable PyTorch-based applications to run efficiently on its chips. Amazon also said more than 30 early customers across fields including robotics, world models and voice AI were already using the new software.
Brand Icon ImageSwitching Becomes a Setting
TorchNeuron had been in private betas since late last year, and Amazon opened the software to all developers as a public beta on Monday. Colin Brace told GeekWire, “If you know PyTorch, then you know Trainium,” as he described how TorchNeuron changes the process from modifying code to a basic configuration change. Amazon planned to release TorchNeuron as open-source software on Tuesday in conjunction with the PyTorch Conference in San Jose, Calif., and Brace said Amazon’s goal was wide use of the chips.
Nvidia dominates the market for AI chips partly because much of the industry’s AI software was built around its CUDA platform, and adapting that code to other hardware can take months of engineering work. Amazon tests it every night against more than 100,000 models from Hugging Face to catch compatibility regressions before they reach customer workloads.
GeekWireCompetition, Adoption, and Stakes
TechCrunch framed TorchNeuron as Amazon’s direct stab at Nvidia’s software advantage, saying the tool lets AI apps written with PyTorch run on Trainium by changing just one line of configuration, with “no code rewrite required.” Colin Brace told TechCrunch via email that “You’re building on AWS,” and he said “the infrastructure to go from prototype to production is already there, allowing a founder to move fast and scale when they need to.” Amazon said about 85% of AI apps are written with PyTorch, and it described Nvidia’s CUDA software as the infrastructure that allows apps to run on Nvidia GPUs and other hardware.
Amazon is pushing PyTorch to be less automatically centered on Nvidia and is contributing code upstream to remove changes that tie the framework to specific chips. Deepgram told TechCrunch that it moved to Trainium using native PyTorch support and was “processing 70% more audio per dollar at production-grade accuracy,” as Amazon argued the migration could improve economics across a full workload.
TechCrunch