搜索 AI 资讯

b11881

llama : respect no_alloc in all memory caches (#31087) * llama : respect no_alloc in memory caches Assisted-by: dsh:llama.cpp/ggml-org/DeepSeek-V4-Flash-Vision-Exp-GGUF (Mac Studio, M2 Ultra, 192GB) * llama : TODO: move no_alloc out of llama_hparams Assisted-by: dsh:llama.cpp/ggml-org/DeepSeek-V4-Flash-Vision-Exp-GGUF (Mac Studio, M2 Ultra, 192GB) * llama : TODO: move no_alloc out of llama_hparams Assisted-by: dsh:llama.cpp/ggml-org/DeepSeek-V4-Flash-Vision-Exp-GGUF (Mac Studio, M2 Ultra, 192GB) **Website:** - **Attestations:** - **macOS/iOS:** - [macOS Apple Silicon (arm64)](https://github.com/ggml-org/llama.cpp/releases/download/b11881/llama-b11881-bin-macos-arm64.tar.gz) - macOS Apple Silicon (arm64, KleidiAI enabled) [DISABLED](https://github.com/ggml-org/llama.cpp/pull/23780) - [macOS

b11880

tools : reject truncated GGUF tensor data in merge (#31044) Co-authored-by: Shreya Agrawal **Website:** - **Attestations:** - **macOS/iOS:** - [macOS Apple Silicon (arm64)](https://github.com/ggml-org/llama.cpp/releases/download/b11880/llama-b11880-bin-macos-arm64.tar.gz) - macOS Apple Silicon (arm64, KleidiAI enabled) [DISABLED](https://github.com/ggml-org/llama.cpp/pull/23780) - [macOS Intel (x64)](https://github.com/ggml-org/llama.cpp/releases/download/b11880/llama-b11880-bin-macos-x64.tar.gz) - [iOS XCFramework](https://github.com/ggml-org/llama.cpp/releases/download/b11880/llama-b11880-xcframework.zip) **Linux:** - [Ubuntu x64 (CPU)](https://github.com/ggml-org/llama.cpp/releases/download/b11880/llama-b11880-bin-ubuntu-x64.tar.gz) - [Ubuntu arm64 (CPU)](https://github.com/ggml-org/lla

b11879

CUDA: fuse FFN gate/up matmuls and GLU in MMQ with interleaved gate/up rows (#29948) * cuda: fuse FFN gate/up and GLU in MMQ with interleaved gate/up rows For prompt processing, run the gate and up matmuls of a dense FFN as one mul_mat_q launch and apply the GLU in the write-back. - The loaders take the gate weights as a second pointer and fill the weight tile with up and gate rows interleaved in chunks of 8, so each thread holds an up value and the matching gate value (rows r and r + 8 of an MMA tile). - The MMA loop is unchanged: same tile, warp layout, vec_dot and one sum[] per thread. The fused write-back computes the GLU from registers and writes half the rows. - The fused path uses the regular tiling (no stream-k) with the column tiles innermost so that neighbouring blocks share weig

b11878

ggml-cuda: fuse qwen MRoPE + set-rows (#31025) **Website:** - **Attestations:** - **macOS/iOS:** - [macOS Apple Silicon (arm64)](https://github.com/ggml-org/llama.cpp/releases/download/b11878/llama-b11878-bin-macos-arm64.tar.gz) - macOS Apple Silicon (arm64, KleidiAI enabled) [DISABLED](https://github.com/ggml-org/llama.cpp/pull/23780) - [macOS Intel (x64)](https://github.com/ggml-org/llama.cpp/releases/download/b11878/llama-b11878-bin-macos-x64.tar.gz) - [iOS XCFramework](https://github.com/ggml-org/llama.cpp/releases/download/b11878/llama-b11878-xcframework.zip) **Linux:** - [Ubuntu x64 (CPU)](https://github.com/ggml-org/llama.cpp/releases/download/b11878/llama-b11878-bin-ubuntu-x64.tar.gz) - [Ubuntu arm64 (CPU)](https://github.com/ggml-org/llama.cpp/releases/download/b11878/llama-b11878

b11877

vulkan: Add a larger tile size for conv3d (coopmat2 only) (#29920) * vulkan: Add a larger tile size for conv3d (coopmat2 only) This adds a larger tile size, in practice it is only for coopmat2 where we don't need shared memory to store the output (would use more than 64KB). Add some additional representative backend tests. At least, this is what codex came up with from looking at sd.cpp Wan VAE. * only apply large tile size for newer driver versions **Website:** - **Attestations:** - **macOS/iOS:** - [macOS Apple Silicon (arm64)](https://github.com/ggml-org/llama.cpp/releases/download/b11877/llama-b11877-bin-macos-arm64.tar.gz) - macOS Apple Silicon (arm64, KleidiAI enabled) [DISABLED](https://github.com/ggml-org/llama.cpp/pull/23780) - [macOS Intel (x64)](https://github.com/ggml-org/llama

llama.cpp 发布 b11869 版本：修复 Jinja 模板函数参数处理，新增多平台二进制支持

llama.cpp 项目发布构建版本 b11869，主要修复 Jinja 模板中 select 和 selectattr 函数的参数解析逻辑，并提供覆盖 macOS、Linux、Windows、Android、iOS 及 openEuler 的数十种硬件与后端组合的预编译二进制包。

推荐理由：按信源设置仅收录，不参与自动精选

llama.cpp 发布 b11868 版本：增强 Hexagon NPU 多核 Flash Attention 支持，覆盖全平台编译构建

llama.cpp v2.45.0（标签 b11868）正式发布，核心更新为 Hexagon 后端引入 KV-head 团队拆分机制以优化多核 HMX Flash Attention 性能，并统一 FA 拆分控制标志；同步提供 macOS、Linux、Windows、Android、iOS 及 openEuler 等全平台二进制包，含 CUDA 12/13、Vulkan、ROCm、OpenVINO、SYCL、Snapdragon（Adreno+Hexagon）等多后端支持。

推荐理由：按信源设置仅收录，不参与自动精选

llama.cpp 发布 b11867 版本：服务端入口重构，全面支持多平台与异构加速后端

llama.cpp 于 GitHub 正式发布 b11867 版本，核心变更包括服务端（server）入口点重构（PR #31113），新增 server_role 机制；同时提供覆盖 macOS、Linux、Windows、Android、iOS 及 openEuler 的 40+ 预编译二进制包，支持 CPU、CUDA（12.4/12.8/13.4）、Vulkan、ROCm 10.0、OpenVINO、SYCL（FP16/FP32）、Snapdragon（Adreno+Hexagon）等多元硬件后端。

推荐理由：按信源设置仅收录，不参与自动精选

llama.cpp 发布 b11865 版本：新增 RDMA 接收边界检查，支持 CUDA 13.4、ROCm 10.0、Snapdragon NPU 等多平台后端

llama.cpp 项目发布构建版本 b11865，核心变更包括在 RDMA 接收逻辑中增加内存边界检查（PR #30399），并显著扩展硬件加速支持，覆盖 CUDA 13.4、ROCm 10.0、Vulkan、OpenVINO、SYCL 及高通 Snapdragon（CPU+Adreno+Hexagon）全栈异构计算后端；macOS KleidiAI 支持与 openEuler 构建被明确标记为 DISABLED。

推荐理由：按信源设置仅收录，不参与自动精选

llama.cpp 发布 b11861 版本：修复 minimax-m3 模型量化缩放张量未传递导致的推理异常

llama.cpp v.b11861 修复了 minimax-m3 模型在 NVFP4/MXFP8 量化下因 scale 张量未注入计算图而导致的权重失真问题，避免‘token soup’与残差淹没；同步提供覆盖 macOS、Linux、Windows、Android、iOS 及 openEuler 的多平台二进制构建。

推荐理由：按信源设置仅收录，不参与自动精选