Accelerating PyTorch on AMD GPUs with FlyDSL

PyTorch users can now leverage FlyDSL to accelerate performance on AMD MI350-series GPUs through existing operator APIs for various operations, including dense and grouped GEMM, fused multi-head attention, and more. This integration enhances the capabilities of PyTorch on AMD hardware. The support includes key operations like MXFP8/MXFP4 scaled GEMM, RMSNorm, and TopK.