viable/strict/1791882163: [Windows] Depend on PyPI CUDA runtime packages in wheels (#194056)
- PyTorch: 2284 events in the last 90 days
- PyTorch: 2254th Release in the last 90 days
- Previous: earlier the same day · trunk/4c581132936859129920131e4f2681afe04b42fa
What happened
Summary This PR teaches Windows PyTorch CUDA wheels to locate their CUDA runtime DLLs from installed NVIDIA PyPI packages (via cuda-pathfinder) rather than bundling them inside torch/lib/. It adds per-OS wheel dependency metadata, updates the build matrix generator, adjusts CMake to skip bundling when using PyPI-sourced runtimes, and adds a Windows smoke test that validates the declared cuda-toolkit version matches…
Summary assembled by rule from the sources below