Google open-sources TorchTPU, letting PyTorch models run natively on TPUs

Google open-sources TorchTPU, a native PyTorch back end for TPUs

Google has released TorchTPU, an open-source PyTorch backend for Tensor Processing Units. It includes custom ATen kernels, a compilation cache for XLA computations, and support for distributed training via data and tensor parallelism. The project integrates with torch.compile() and ships examples for models like Llama, Qwen, ResNet, and minGPT. Installable via pip (version 0.2.0) or built from source with Bazel, it requires Python 3.11+ and is licensed under Apache 2.0.