📄 中文摘要
Ollama 在裸金属环境中表现出色，当将其视为服务时，提供了更为有趣的功能，包括稳定的端点、固定版本、持久存储以及可用或不可用的 GPU。该内容的重点是创建一个可重复的本地或单节点 Ollama '服务器'，使用 Docker Compose 实现 GPU 加速和持久模型存储。内容故意跳过了 Docker 和 Compose 的基础知识，适合需要快速参考常用命令的用户。对于希望在 Ollama 前面实现 HTTPS、正确流式传输和 WebSocket 的用户，提供了相关的指导。
在 Docker Compose 中使用 GPU 和持久化模型存储的 Ollama
出处: Ollama in Docker Compose with GPU and Persistent Model Storage
发布: 2026年3月27日
📄 English Summary
Ollama in Docker Compose with GPU and Persistent Model Storage
Ollama performs exceptionally well on bare metal, and it becomes even more intriguing when treated as a service, offering stable endpoints, pinned versions, persistent storage, and a GPU that may or may not be available. The focus is on creating a reproducible local or single-node Ollama 'server' using Docker Compose, with GPU acceleration and persistent model storage. The content intentionally skips over the basics of Docker and Compose, catering to users who need a quick reference for commonly used commands. Guidance is also provided for those looking to implement HTTPS in front of Ollama, ensuring correct streaming and WebSocket functionality.