Boost Vision Transformer With GPU-Friendly Sparsity and Quantization | Read Paper on Bytez