IT之家 · IT业界 · 10/20/2026, 08:22:37
Amazon Open-Sources TorchNeuron to Enable Zero-Modification PyTorch on Trainium Chips
Amazon has released TorchNeuron, an open-source tool that enables developers to run PyTorch code natively on its custom Trainium chips with only a configuration change and no code rewriting. Designed to challenge NVIDIA's CUDA ecosystem dominance, the tool aims to reduce migration costs and enhance hardware agnosticity, debuting at the North American PyTorch Conference.
SOURCE COVERAGEFull Original Report
Challenging NVIDIA CUDA: Amazon Launches Open-Source Tool TorchNeuron to Run PyTorch Code on Custom Trainium Chips
On October 20, it was reported that on October 19 (local time), Amazon announced the launch of its open-source tool, TorchNeuron. This tool allows developers' PyTorch code to run natively on Amazon's custom Trainium chips with just a single configuration change—no rewriting required.
According to IT Home, TorchNeuron will be officially released under an open-source license at the North America PyTorch Conference held in San Jose, California, on October 20 (local time).
PyTorch is an open-source framework led by Meta and adopted by most AI developers. NVIDIA has spent over a decade developing its CUDA toolkit to enable efficient execution of PyTorch code on NVIDIA hardware.
Amazon engineers reportedly ran identical AI training scripts on both traditional chips and Trainium chips. The only difference lay in the configuration parameters specifying the target chip within the code. Results showed that the AI models learned exactly the same content at the same speed in both environments. This means that tools originally used for testing, debugging, and scaling AI models can be directly applied to Trainium without any modifications.
Amazon states that TorchNeuron introduces a long-missing critical element to the AI industry: truly viable choice. When developers can switch freely between different chips without additional costs, chip manufacturers must compete by offering more compelling products to win developer adoption. Developers, in turn, benefit from reduced costs and faster iteration cycles.
Furthermore, Amazon is actively participating in upstream PyTorch development, collaborating with the PyTorch community to remove code that ties the framework to specific chips. The goal is to make PyTorch more "hardware-agnostic" for the entire developer community, rather than serving only Trainium users.