Amazon says TorchNeuron lets developers run existing PyTorch code on its AWS Trainium AI chips without a rewrite. The company plans to open-source the software at the PyTorch Conference on October 20.
PyTorch is a framework used to build AI models, and Trainium is Amazon's AI chip. According to Amazon, TorchNeuron lets developers take code they already use in PyTorch and run it on Trainium without rewriting it for the chip.
Startups and enterprises are using TorchNeuron
Amazon says startups and enterprises are already building with TorchNeuron. It names robotics, video, world models and self-driving cars among the areas where that work is taking place.
The company's list also includes speech, drug discovery, large language models and finance. Amazon presents these as areas of current work with TorchNeuron. Its central technical claim remains the same across those uses: existing PyTorch code can run on Trainium without a rewrite.
For developers using PyTorch, the announced change is a way to work with Amazon's AI chips while keeping their existing code. Amazon says the next step is to open-source TorchNeuron at the PyTorch Conference on October 20.
