AMD and Intel to combine PyTorch CPU optimizations in oneDNN
AMD and Intel plan to combine their x86 CPU optimization work for PyTorch in oneDNN, an open-source deep learning library governed by the UXL Foundation. The companies say the move gives developers one common software path across AMD and Intel CPUs. AMD is integrating its ZenDNN optimizations into oneDNN and has joined the UXL Foundation as a steering member. The two companies are also collaborating on AI Compute Extensions (ACE), a next-generation x86 instruction set designed to speed up matrix multiplication, which will be integrated into PyTorch through future oneDNN updates.