Minimal backend selection
Tunes a small MLP and prints the faster of Torch Inductor and Torch Eager. Uses an Intel XPU when one is available, and CPU otherwise.
Run from the repository
From the repository root, after uv sync --extra dev:
Run this example on its own
From this directory. PyTorch comes from the xpu extra (Intel XPU wheels).