rGPU - Run PyTorch tensors on a remote GPU

Show HN: Rgpu – a PyTorch device whose tensors live on a remote GPU

rGPU lets you keep Python on your laptop while running PyTorch operations and storing tensors on a remote NVIDIA GPU. It offers two paths: a native PyTorch device for simple integration, and a CUDA shim for existing Linux CUDA programs. With a quick pip install and a single command, you can train models like nanoGPT from a Mac without CUDA. Ideal for developers who need GPU power but prefer local development. The project includes thorough documentation and is Apache 2.0 licensed.