llama.cpp b11643 发布：修复仅 embedding 图空图问题

AI 摘要

llama.cpp 发布 b11643 版本，修复了仅含 embedding 的图在关闭 embeddings 时因 build_pooling() 提前返回、t_embd 不再展开而导致图为空、输入未被分配的问题，通过在该情况下展开 t_embd 解决。该版本同时提供 macOS、iOS、Linux、Android、Windows 等多平台多后端（CUDA、Vulkan、ROCm、SYCL、OpenVINO 等）的预编译二进制包。

正文节选

<details open> graph: keep the graph non empty without embeddings (#30522) An embedding-only graph has no logits, and build_pooling() returns early when embeddings are off, so since the models no longer expand t_embd the graph ends up empty and its inputs are never allocated. Expand t_embd in that case. </details> **Website:** - <https://llama.app> **Attestations:** - <https://github.com/ggml-org/llama.cpp/attestations/56206571> **macOS/iOS:** - [macOS Apple Silicon (arm64)](https://github

发布时间：2026-10-15 17:58

抓取时间：2026-10-15 18:33

来源机构：ggml-org