llama.cpp b12223:为 Granite 架构接入 NVFP4 权重缩放
原标题:b12223
AI 摘要
llama.cpp 发布 b12223 版本,主要变更为在所有 Granite 架构中通过 build_ffn 传递 NVFP4 权重缩放(weight scales),使 Granite 系列模型支持 NVFP4 量化推理。该改动由 IBM 的 Gabe Goodhart 提交,并注明使用 Claude 辅助开发。发布同时提供 macOS、Linux、Windows、Android 等多平台预编译包,覆盖 CUDA、Vulkan、ROCm、SYCL、OpenVINO 等后端。
正文节选
<details open> models : Attach Granite NVFP4 weight scales (#31449) * models : Attach Granite NVFP4 weight scales * feat: Pass nvfp4 scales through build_ffn for all granite architectures Branch: granite-nvfp4-moe-scales AI-usage: full (Claude + Sonnet 5.5) Signed-off-by: Gabe Goodhart <ghart@us.ibm.com> --------- Co-authored-by: Gabe Goodhart <ghart@us.ibm.com> </details> **Website:** - <https://llama.app> **Attestations:** - <https://github.com/ggml-org/llama.cpp/attestations/65350471