量化会让模型变笨一半Quantisation Halves the Quality
微型指南 · 约 5 分钟Guide · about 5 min · 引用核验 2026-11-03（仅引用链接与日期） · refs checked 2026-11-03 (reference links and dates only)
影响随档位而变，量尺是每权重比特数。The impact varies by tier; the ruler is bits per weight.
量化就是把模型质量砍掉一半。Quantising cuts model quality in half.
量化改变数值表示，质量影响取决于模型、方法、档位和任务。bpw 描述每个权重平均占多少比特，不是能力分数；需用相同任务比较质量、内存和速度。Quantization changes numeric representation. Effects depend on model, method, tier and task. Bits per weight describe average storage, not capability; compare quality, memory and speed on equal tasks.
例子与判断Example and interpretation
以下是教学情境，不是本站运行或实测记录。A teaching scenario, not a measured or executed site test.
选择一个量化档位后，先核对文件、运行器支持和内存需求，再用同一组任务比较原版与量化版。记录具体错误，不能把位宽减半写成任务质量也减半。After choosing a quantization, check the file, runtime support and memory needs, then compare identical tasks with the original. Record specific errors rather than equating halved bit width with halved task quality.
AI 常见误区总索引AI Misconceptions: Index · 量化、蒸馏、剪枝的区别Quantisation vs Distillation vs Pruning