Skip to content

Ferrum v0.8.4

Choose a tag to compare

View all tags
@sizzlecar sizzlecar released this 02 Sep 01:03
· 18 commits to main since this release
v0.8.4
5a63d06

Ferrum 0.8.4 focuses on a smoother first-run experience and verifiable release assets.
Ferrum 0.8.4 重点改善首次使用流程与发布资产可靠性。

What's Changed / 更新内容

  • Quick Start now covers installation checks, ferrum doctor, ferrum run, ferrum serve, and an OpenAI-compatible API request from end to end.
    快速开始现在完整覆盖安装检查、模型诊断、直接运行、服务启动和 OpenAI 兼容 API 请求。
  • The first Metal run downloads about 2.55 GiB. Download time depends on the Hugging Face network route; watch the progress output before treating it as hung.
    Metal 首次运行会下载约 2.55 GiB，耗时取决于本机到 Hugging Face 的网络链路。
  • Quick Start uses --disable-thinking for a short, direct first response. Omit it to retain the model template's default reasoning behavior.
    快速开始使用 --disable-thinking，让首次回答简短直接；删除该参数即可恢复模型模板默认推理行为。
  • Release packaging and validation were tightened for CPU, Metal, and CUDA binaries, including checksums and dependency metadata.
    加强了 CPU、Metal 和 CUDA 二进制的打包与验证，并提供校验和及依赖元数据。

Install / 安装

brew tap sizzlecar/ferrum

# macOS Apple Silicon
brew install ferrum

# Linux x86_64, NVIDIA CUDA sm89
brew install ferrum-cuda

First Metal run / Metal 首次运行：

ferrum doctor qwen3.5:4b-q4_k_m
ferrum run qwen3.5:4b-q4_k_m --disable-thinking

See the English README or 中文说明 for the complete run, serve, and API flow.

Full Changelog: v0.8.3...v0.8.4

Assets 23
Loading