TorchNeuron
Enables existing PyTorch code to run on AWS Trainium AI chips
Nobody has claimed this listing yet.
We wrote this page from a public post on #opensource, so the maker has not checked a word of it. Read the original. An unclaimed listing has never been on the board and holds no rank or badge.
About
TorchNeuron is an open-source public beta providing a native PyTorch backend for AWS Trainium2 and Trainium3 AI accelerators. It allows much existing PyTorch code to target these chips, reducing the rewriting previously needed for testing Amazon's AI chips. The software registers Trainium through PyTorch's official extension mechanism. It supports eager execution, torch.compile, profiling, and distributed training via standard APIs. This makes Amazon's accelerators easier to evaluate.
