<aside> 💡
将量化难度从激活迁移到权重
</aside>
论文地址
SmoothQuant: Accurate and Efficient Post-Training Quantization for Large Language Models
以上是大模型量化困难的原因,总结下来就三点: