trunk/94190aa4e4c3a7978e41cf7d3c9951b9251d5e31: Enable XPU support to graph capture check in optimizer (#172759)
Intel GPUs get major PyTorch boost, challenging NVIDIA's CUDA dominance.
Deep Dive
PyTorch commit 94190aa enables Intel XPU support for graph capture checks in optimizers, marking step 5 in the XPUGraph implementation roadmap. This follows the RFC #162143 and builds on previous work enhancing XPUGenerator, MemPool, and core capture/replay functions. The change allows PyTorch's optimizer to perform health checks for graph capture on Intel GPUs, expanding hardware options beyond NVIDIA's CUDA ecosystem for accelerated training workflows.
Why It Matters
This reduces dependency on NVIDIA hardware, giving developers more choice and potentially lowering AI training costs.