ScaleSweep: Accurate NVFP4 Post-Training Quantization of LLMs via Block Scale Initialization 事件

ScaleSweep: Accurate NVFP4 Post-Training Quantization of LLMs via Block Scale Initialization arXiv:2606.07618v1 Announce Type: cross Abstract: NVFP4 is a recently introduced hardware-supported FP4 format that improves the fidelity of 4-bit quantization through fine-grained block scales. However, existing NVFP4 scale initialization methods still primarily rely on AbsMax initialization, which leaves a noticeable gap to the optimal solution. To address this, we propose ScaleSweep, a simple and eff