📄 中文摘要
构建了一个多服务本地 AI 堆栈，涵盖图像生成、视频生成、语音合成和语音克隆，运行于 RTX 5090 显卡上，通过 WSL2 和 Docker 实现。关键突破在于解决了 GPU 驱动程序直通层的问题，这一过程没有相关文档可供参考。文章详细介绍了架构设计及关键的 gpu-run 配置，提供了实用的步骤和经验，帮助用户轻松搭建类似的 AI 服务环境。
如何在 RTX 5090 + WSL2 + Docker 上同时运行 6 个 AI 服务（你也可以）
📄 English Summary
How I Run 6 AI Services Simultaneously on RTX 5090 + WSL2 + Docker (And You Can Too)
A multi-service local AI stack has been built, encompassing image generation, video generation, voice synthesis, and voice cloning, all running on an RTX 5090 GPU via WSL2 and Docker. The key breakthrough was solving the GPU driver passthrough layer, which had no documentation available. The article details the architecture design and critical gpu-run configuration, providing practical steps and insights to help users easily set up a similar AI service environment.