The Ray distributed computing project announced a Special Interest Group for hardware accelerators, aiming to bring together hardware vendors, Ray contributors, and application developers to define a shared, vendor-neutral roadmap for accelerator support1. Ray, created at UC Berkeley over 10 years ago, now supports more than a dozen kinds of hardware accelerators, including NVIDIA and AMD GPUs, TPUs, Trainium, and Gaudi. The SIG formation follows growing specialization among accelerators, with the announcement citing NVIDIA Blackwell, AMD MI355X, Google Ironwood TPUs, and Groq LPUs as examples of the diversifying landscape.
PyTorch Blog: Ray Forms Accelerator SIG for Vendor-Neutral GPU, TPU Roadmap
The Ray project announced a Special Interest Group for hardware accelerators to build a vendor-neutral roadmap spanning NVIDIA, AMD, Google, and other accelerator platforms.
Was this useful?
The day’s vector, in your inbox.
Vector Wire Daily: one dispatch each publishing day.