AMD Releases ZenDNN 6.1 With New Kernels & Optimizations

Last month Intel and AMD jointly announced they would be collaborating on software optimizations with PyTorch/oneDNN. That already led to the UXL Foundation's release of oneDNN 3.14 with initial AMD Zen optimizations. It looks though like for now at least AMD will also continue to concurrently push new work into their ZenDNN library and not exclusively tied to oneDNN. Out today is ZenDNN 6.1. ZenDNN 6.0 released back in July with a variety of enhancements for this oneDNN API compatible library used by deep learning software. The ZenDNN 6.1 release is further building on those recent additions. With ZenDNN 6.1 there is more work on their LowOHA inference stack with faster group MATMUL custom kernels, FP16 and DA8W4 kernels, MATMUL custom kernels for DA8W4 / DA8W8 / INT8 / BF16, symmetric INT8 prepacking, and other improvements. Those wishing to learn more about AMD ZenDNN 6.1 can do so via GitHub.