llama.cpp b10947 修复 nemotron-h 专家 FFN 除零崩溃
原标题：b10947
AI 摘要
llama.cpp 发布 b10947 版本，修复了 nemotron-h 模型在加载时的专家 FFN 尺寸回退逻辑缺陷。当 NextN/MTP 尾部循环中 expert_feed_forward_length 未提供该层数值时，代码会用 n_ff/n_expert_used 推导专家 FFN 尺寸，而这两个逐层数组在非 MoE 层上合法地为 0，导致同时为 0 的检查点发生除零并在加载时以 SIGFPE 崩溃且无错误提示。该修复改为报告元数据格式错误，同时照常提供各平台预编译二进制包。
正文节选
<details open> models : guard the expert FFN size fallback in nemotron-h against a zero divisor (#28779) The NextN/MTP tail loop derives the expert FFN size as n_ff/n_expert_used when expert_feed_forward_length gives nothing for the layer. Both values come from per-layer arrays that legitimately hold 0 on layers that are not MoE, so a checkpoint whose predict layers hold 0 in both divides by zero and dies with SIGFPE at load time, with no error message. Report the malformed metadata instead.