MGRQ: Post-Training Quantization For Vision Transformer With Mixed Granularity Reconstruction | Read Paper on Bytez