Bert
AMD Optimizes BERT Inference with EPYC Processors and ZenDNN
AMD reports up to 58.1% latency gains in BERT inference using ZenDNN on EPYC processors, showcasing significant performance improvements.