MARD1NO/ByteTransformer

optimized BERT transformer inference on NVIDIA GPU. https://arxiv.org/abs/2210.03052

C++Apache-2.0