llama.cpp b12180 新增 Kimi-K3.1-Pro 文本模型支持
原标题:b12180
AI 摘要
llama.cpp 发布 b12180 版本,新增对 Kimi-K3.1-Pro 文本模型的支持。该模型在 93 层中将 24 层全注意力 MLA 替换为 DSA(带 lightning indexer,32 头×128,top_k 2048,无 rope),KDA 线性注意力层及 K3 骨架其余部分保持不变,因此复用 K3 计算图与转换器并新增 indexer。实现包含混合内存缓存、转换器适配与量化保护(indexer 的 q_b、k、proj 不参与量化),并在 8 层切片上与 HF checkpoint 验证一致。
正文节选
<details open> model: add Kimi-K3.1-Pro text model (#31779) * model: add Kimi-K3.1-Pro text model K3 with the full-attention MLA layers replaced by DSA: on this checkpoint 24 of the 93 layers carry a lightning indexer (32 heads x 128, top_k 2048, no rope). The KDA (linear-attention) layers and the rest of the K3 skeleton - cross-layer residuals, latent MoE, situ, MLA output gate, full-rank KDA gate - are unchanged, so this reuses the K3 graph and converter and adds the indexer. Indexer: -