Amazon Challenges Nvidia with New TorchNeuron Tool
The tool aims to reduce developers' dependence on Nvidia technologies
TorchNeuron works with PyTorch: to migrate an application to the Trainium chip, a developer only needs to change one line of configuration without rewriting the code. Amazon states that approximately 85% of AI applications are built using this tool.
PyTorch was originally developed by Facebook, and since 2022, the project has been part of the Linux Foundation. Nvidia has spent years developing CUDA—a software infrastructure for running applications on its chips—making the company's hardware the default option for PyTorch users. Migration of applications is also supported by AMD and Apple chips.
Amazon plans to release TorchNeuron as open source at the PyTorch Conference on October 20. The company is also proposing changes to PyTorch itself to reduce the framework's dependence on specific hardware. However, the impact of this approach for developers and the cost of solutions remain to be evaluated.


