llama.cpp b11883 修复 4D 张量量化与 MoE 专家检测
原标题：b11883
AI 摘要
llama.cpp 发布 b11883 版本，修复了 llama-quant 工具对 4D 张量的量化问题：此前 nrows_total 只按 ne[1]*ne[2] 计算，导致 4D 张量仅前 ne[2] 个切片被量化、其余行被丢弃，现改用 ggml_nrows() 统计全部行。同时将 MoE 专家张量检测从依赖 ne[2]>1 改为按张量名（*_exps/*_chexps）判断，并修正 4D 张量 imatrix 尺寸为 ne[0]*ne[2]*ne[3]。该修复解决了 #29916，并同步提供各平台预编译二进制。
正文节选
<details open> llama-quant : fix 4D tensor quantization using ggml_nrows() and name-based expert detection (#30820) * llama-quant : fix quantizing 4D tensors nrows_total was ne[1]*ne[2], so for a 4D tensor only the first ne[2] slices were quantized and the rest of the rows were dropped. Use ggml_nrows(tensor) to count all rows, including ne[3]. Fixes #29916 * llama-quant : detect expert tensors by name and fix imatrix size for 4D tensors The MXFP4_MOE path used ne[2] > 1 to detect MoE expe