AMD Releases ZenDNN 4.1 to Enhance Deep Learning Inference Speed on Ryzen and EPYC CPUs

Phoronix reports that AMD has released ZenDNN 4.1, the latest version of their Zen Deep Neural Network Library. This release is aimed at accelerating inference workloads on AMD Ryzen and EPYC processors. The library is API compatible with Intel’s oneDNN library and helps optimize AI workloads for use on AMD Zen processors.

The ZenDNN 4.1 release notes state that the library has been updated to support TensorFlow 2.12, PyTorch 1.13, and ONNX Runtime 1.15.1. There are no other mentioned changes or optimizations in this release.

The new ZenDNN 4.1 release can be downloaded from GitHub, and more information about the Zen Deep Neural Network library can be found on AMD’s Developer Central website.

Source: Phoronix.