llama.cpp b9625 release fixes jinja negative step slice and provides multi-platform binaries
English summary
llama.cpp version b9625 has been released. The release contains a fix for jinja template rendering that failed when slicing with a negative step and start/stop values (PR #24580). Pre-built binaries are published for macOS Apple Silicon (arm64), Ubuntu Linux (x64 and arm64) with various backends including CPU, Vulkan, ROCm, OpenVINO, and SYCL, Windows (x64, arm64) with CUDA 12/13, Vulkan, SYCL, HIP, and Android arm64. Builds for macOS Intel, iOS, and openEuler are listed as disabled.
Chinese summary
llama.cpp 发布了 b9625 版本。本次更新修复了 jinja 模板渲染在使用负步长且有起始/结束值时的切片错误（PR #24580）。提供了针对 macOS Apple Silicon (arm64)、Ubuntu Linux（x64 和 arm64，含 CPU、Vulkan、ROCm、OpenVINO、SYCL 等多种后端）、Windows（x64、arm64，含 CUDA 12/13、Vulkan、SYCL、HIP）和 Android arm64 的预编译二进制文件。macOS Intel、iOS 和 openEuler 的构建被标记为禁用。
Key points
Fixes jinja template rendering bug for negative step slices with start/stop (PR #24580).
修复了 jinja 模板渲染中带有起始/结束值的负步长切片错误（PR #24580）。
Pre-built binaries for macOS arm64, Ubuntu x64/arm64 with multiple backends, Windows x64/arm64 with CUDA/Vulkan/SYCL/HIP, and Android arm64.
提供 macOS arm64、Ubuntu x64/arm64（多后端）、Windows x64/arm64（CUDA/Vulkan/SYCL/HIP）和 Android arm64 的预编译二进制文件。
Some builds (macOS Intel, iOS, openEuler) are disabled in this release.
本次发布中部分构建（macOS Intel、iOS、openEuler）被禁用。