trunk/7cdd4b16cad708e2083ea9ff2ec724876485cf90: [CI] Add CUDA 13 periodic tests (#174850)
The AI framework's latest commit prepares for stable CUDA 13 wheels, enabling next-gen GPU acceleration.
Deep Dive
Meta's PyTorch team (led by Andrey Talman) added periodic CUDA 13 tests in commit 7cdd4b1. This technical infrastructure update prepares the 97.5k-star open-source framework to move CUDA 13 from experimental to stable wheels. Developers will gain reliable access to Nvidia's latest GPU compute platform, ensuring PyTorch models run efficiently on newer hardware like H100 and RTX 40-series cards with full CUDA 13 feature support.
Why It Matters
Stable CUDA 13 support means faster training and inference on latest Nvidia GPUs, crucial for AI researchers and engineers.