Fetching the paper…

VS-Quant: Per-vector Scaled Quantization for Accurate Low-Precision Neural Network Inference · Around