AWS released TorchNeuron, a tool that lets AI apps built with PyTorch run on Amazon’s Trainium chips after a one-line configuration change rather than a code rewrite. PyTorch, the open source framework Facebook created, is used to build roughly 85% of AI apps by Amazon’s count, and Nvidia has spent years making its CUDA software the framework’s default path to hardware. Amazon is also contributing changes to the main PyTorch project that strip out code tying it to particular chips, which it says will help every developer, not only Trainium customers. AWS is pitching startups on running both training and inference on Trainium inside its full cloud, instead of renting Nvidia GPUs from neoclouds like CoreWeave and Nebius.
