Fast implementation of BERT inference directly on NVIDIA (CUDA, CUBLAS) and Intel MKL
Primary LanguageCMIT LicenseMIT