AWS releases TorchNeuron to run PyTorch AI apps on Trainium chips

In short: Amazon Web Services has released TorchNeuron, a new tool meant to make it easier to run PyTorch-based AI apps on Amazon’s Trainium chips instead of Nvidia hardware.

What happened

Amazon Web Services (AWS) announced TorchNeuron, a developer tool that helps AI startups move their AI apps onto AWS’s own AI chip called Trainium.

TorchNeuron works with PyTorch, a free software toolkit that many developers use to build and run AI models (think of it like a common set of Lego pieces for AI). AWS says TorchNeuron can make a PyTorch app run on Trainium by changing one line in the app’s configuration, which is basically a settings file.

AWS is also positioning TorchNeuron as part of a broader effort to reduce how closely PyTorch is tied to Nvidia’s software system called CUDA. CUDA is the set of tools that helps AI software run well on Nvidia chips. TechCrunch reports that AWS plans to contribute code changes back to the main PyTorch project, with the goal of making it easier for PyTorch apps to run on different kinds of chips.

AWS says some companies are already moving work to Trainium. It pointed to Deepgram, an AI voice company, as an example.

Why it matters

Most AI services that people use, like chatbots and voice tools, rely on expensive computer chips. If it becomes easier for companies to choose chips beyond Nvidia, it could affect prices and competition in cloud computing. For regular users, that may eventually influence how much AI features cost and how widely they show up in everyday apps.

Source: TechCrunch AI