- AWS launched the public beta of TorchNeuron, an open‑source native PyTorch backend for Trainium AI chips.
- Developers can run PyTorch models on Trainium by calling .to('neuron'), retaining full PyTorch compatibility and custom operator support via the Neuron Kernel Interface.
- TorchNeuron is available for EC2 Trainium2 and Trainium3 instances via its GitHub repository.