AMD Releases ZenDNN 6.1 With New Kernels and Optimizations

AMD has released ZenDNN 6.1, extending the updates introduced in version 6.0, which arrived in July. The library is compatible with the oneDNN API and is used by deep-learning software.

The update brings further development to the LowOHA inference stack, including faster custom kernels for group matrix multiplication. It also adds FP16 and DA8W4 kernels, along with custom matrix-multiplication kernels for DA8W4, DA8W8, INT8 and BF16.

ZenDNN 6.1 also includes symmetric INT8 prepacking and other improvements. AMD’s release details are available on GitHub.

Sources: Phoronix.