Skip to content

v0.3.8

Choose a tag to compare

@github-actions github-actions released this 03 Feb 18:27
· 983 commits to master since this release

🚀 Release Notes

📦 包含内容

  • 编译后的 class 文件
  • 所有依赖库(lib/*.jar)
  • 精简版 JRE 21.08(内置,无需额外安装 Java 运行环境)
  • 启动脚本(run.sh / run.bat)

🧩 llama.cpp 版本

📦 可用版本

Linux:

  • llama-server-v0.3.8-b7922-linux-vulkan.zip
  • llama-server-v0.3.8-linux.zip

Windows:

  • llama-server-v0.3.8-b7922-windows-cuda12.zip
  • llama-server-v0.3.8-b7922-windows-cuda13.zip
  • llama-server-v0.3.8-b7922-windows-hip.zip
  • llama-server-v0.3.8-b7922-windows-vulkan.zip
  • llama-server-v0.3.8-windows.zip

📝 使用方法

Linux:

unzip <你下载的 linux zip 包>
cd <解压后的目录>
chmod +x run.sh
chmod +x jre/bin/java
./run.sh

Windows:

unzip <你下载的 windows zip 包>
cd <解压后的目录>
run.bat

注意:

  • 本版本已内置精简版 JRE,无需额外安装 Java 环境。
  • Windows CUDA 版本文件较大,是因为使用了 cudart 包,包含了 CUDA 运行环境。
  • 请根据你的 GPU 选择对应的版本:
    • NVIDIA GPU: 使用 cuda12 或 cuda13 版本
    • AMD GPU: 使用 hip 版本
    • 其他 GPU(如核显) 或不确定: 使用 vulkan 版本
    • 不需要llamacpp:使用无 llamacpp 版本

📌 最新提交

c07fcf2 - 删掉没用的日志 (IIIIIllllIIIIIlllll, 11 minutes ago)

📋 自上次发布以来的变更

  • 删掉没用的日志 (c07fcf2)
  • 删除ollama兼容API的请求头长度信息,这个参数似乎会导致异常的连接终端(不知道是不是测试分支的llamacpp有问题) (69e9ce0)
  • Update README with compilation and API notes (186de4f)
  • 优化一下netty的web服务,似乎有潜在的连接风险,待验证 (77e46b6)
  • 傻逼了,还有一个无用的日志没删 (d7d76a0)

构建时间: 2026-02-03 18:27:05 UTC