Google released TorchTPU as an open-source project, letting developers move AI apps built with PyTorch onto its TPU chips and Google Cloud by changing a configuration setting rather than rewriting their code, a day after Amazon launched a similar tool called TorchNeuron for its Trainium chips. Most AI apps are built with PyTorch, which Facebook created in 2016 at a time when AI researchers mostly ran their work on Nvidia GPUs, so much of it was designed around Nvidia hardware. Google said it worked with Meta, Hugging Face and the makers of the vLLM and SGLang tools so TorchTPU runs with them, and it released an AI agent that helps developers migrate to TPUs and fixes problems as they go.
