How to install PyTorch with uv across CPU, CUDA, ROCm and XPU

tdhopper · x · 2026-07-22

A tutorial on installing PyTorch with uv explains that PyTorch ships different wheels for CPU, CUDA, ROCm, and XPU on separate package indexes.

The key point is that installation is not just a plain pip install: users need to tell uv which index to use so it can pick the correct build for their hardware and accelerator stack. The post points readers to background material on why GPU Python package installation is so fragmented.

Original post →

More from Infra

Infra channel →