This library is widely used by many vendors for multiple devices such as CPUs and GPUs. 

Intel® oneAPI Deep Neural Network Library (oneDNN) is optimized by neural networks from oneAPI. It is a cross-vendor, cross-architecture library of choice for many. Some of the world's best frameworks (like TensorFlow*) and systems (like the Fugaku supercomputer) have selected oneDNN as the library of choice to run neural networks (both training and inference).

Learn more about the oneDNN library components, how you can start using them, and where to find the resources.