TorchTPU is now open source: Run native PyTorch on TPUs

Google has made TorchTPU publicly accessible as an open-source project on GitHub, introducing first-class PyTorch execution directly onto its custom Tensor Processing Units. The release elevates TPUs alongside industry-standard GPUs and CPUs by integrating seamlessly into popular pipelines such as vLLM, TorchTitan, and Ray. The initiative aims to lower the barrier for machine learning teams seeking to train and deploy complex models on Google's proprietary hardware.