一、软件介绍
文末提供程序和源码下载
一个开源脚本，可自动安装在 AMD Radeon 7900XTX 上运行选定 AI 接口所需的所有内容。
二、ROCm-AI-Installer ROCm-AI-安装程序
一个开源脚本，可自动安装在 AMD Radeon 7900XTX 上运行选定 AI 接口所需的所有内容。它也应该适用于 7900XT 卡。对于其他卡，请在脚本开头更改 HSA_OVERRIDE_GFX_VERSION 和 GFX（未测试）
Note 注意
Ubuntu 24.04.2 LTS is recommended. Version 7.x is not tested on older systems.
建议使用 Ubuntu 24.04.2 LTS。版本 7.x 未在较旧的系统上进行测试。
Test platform: 测试平台：
| Name 名字 | Info 信息 |
|---|---|
| CPU 中央处理器 | AMD Ryzen 9950X3D (iGPU disabled in BIOS) AMD Ryzen 9950X3D（iGPU 在 BIOS 中禁用） |
| GPU 图形处理器 | AMD Radeon 7900XTX AMD Radeon 7900XTX 显卡 |
| RAM 公羊 | 64GB DDR5 6600MHz |
| Motherboard 母板 | ASRock B650E PG Riptide WiFi (BIOS 3.20) 华擎 B650E PG Riptide WiFi （BIOS 3.20） |
| OS | Ubuntu 24.04.2 LTS |
| Kernel 内核 | 6.11.0-25-generic 6.11.0-25-通用 |
| ROCm | 6.4 |
Text generation 文本生成
| Name 名字 | Links 链接 | Additional information 其他信息： |
|---|---|---|
| KoboldCPP 狗头人CPP | GitHub - YellowRoseCx/koboldcpp-rocm: AI Inferencing at the Edge. A simple one-file way to run various GGML models with KoboldAI's UI with AMD ROCm offloading | 1. Support GGML and GGUF models. 1. 支持 GGML 和 GGUF 模型。 |
| Text generation web UI 文本生成 Web UI | GitHub - oobabooga/text-generation-webui: A Gradio web UI for Large Language Models with support for multiple inference backends. GitHub - ROCm/bitsandbytes: 8-bit CUDA functions for PyTorch GitHub - turboderp-org/exllamav2: A fast inference library for running LLMs locally on modern consumer-class GPUs | 1. Support ExLlamaV2, Transformers using ROCm and llama.cpp using Vulkan. 1. 支持 ExLlamaV2、使用 ROCm 的 Transformers 和使用 Vulkan 的 llama.cpp。 |
| SillyTavern SillyTavern （愚蠢酒馆） | GitHub - SillyTavern/SillyTavern: LLM Frontend for Power Users. | |
| llama.cpp | GitHub - ggml-org/llama.cpp: LLM inference in C/C++ | 1. Put model.gguf into llama.cpp folder. 1. 将 model.gguf 放入 llama.cpp 文件夹中。 2. Change context size in run.sh file (Default: 32768). 2. 更改文件中 run.sh 上下文大小（默认值：32768）。 3. Set GPU offload layers in run.sh file (Default: 1) 3. 在 run.sh 文件中设置 GPU 卸载图层（默认值：1） |
Image generation 图像生成
| Name 名字 | Links 链接 | Additional information 其他信息： |
|---|---|---|
| ComfyUI 舒适用户界面 | GitHub - comfyanonymous/ComfyUI: The most powerful and modular diffusion model GUI, api and backend with a graph/nodes interface. | 1. Workflows templates are in the workflows folder. 1. 工作流模板位于 workflows 文件夹中。 |
| Artist 艺术家 | GitHub - songrise/Artist: Official repo for DiffArtist: Towards Structure and Appearance Controllable Image Stylization |
ComfyUI Addons ComfyUI 插件
Important 重要
For GGUF models: 对于 GGUF 型号：
1. Accept accept the conditions to access its files and content on HugginFace website:
1. 接受接受 HugginFace 网站访问其文件和内容的条件：
https://huggingface.co/black-forest-labs/FLUX.1-schnell
2. HugginFace token is required during installation.
2. 安装时需要 HugginFace 令牌。
| Name 名字 | Link 链接 | Additional information 其他信息： |
|---|---|---|
| ComfyUI-Manager ComfyUI-管理器 | GitHub - Comfy-Org/ComfyUI-Manager: ComfyUI-Manager is an extension designed to enhance the usability of ComfyUI. It offers management functions to install, remove, disable, and enable various custom nodes of ComfyUI. Furthermore, this extension provides a hub feature and convenience functions to access a wide range of information within ComfyUI. | Manage nodes of ComfyUI. 管理 ComfyUI 的节点。 |
| ComfyUI-GGUF 舒适UI-GGUF | GitHub - city96/ComfyUI-GGUF: GGUF Quantization support for native ComfyUI models https://huggingface.co/city96/t5-v1_1-xxl-encoder-bf16 https://huggingface.co/openai/clip-vit-large-patch14 https://huggingface.co/black-forest-labs/FLUX.1-schnell | GGUF models loader. GGUF 模型加载器。 |
| ComfyUI-AuraSR 舒适UI-AuraSR | GitHub - alexisrolland/ComfyUI-AuraSR: Custom nodes for ComfyUI to run AuraSR models. https://huggingface.co/fal/AuraSR https://huggingface.co/fal/AuraSR-v2 | ComfyUI node to upscale images. ComfyUI 节点来放大图像。 |
| AuraFlow-v0.3 光流 v0.3 | https://huggingface.co/fal/AuraFlow-v0.3 | Text to image model. 文本到图像模型。 |
| FLUX.1-schnell GGUF | https://huggingface.co/black-forest-labs/FLUX.1-schnell https://huggingface.co/city96/FLUX.1-schnell-gguf | Text to image model. 文本到图像模型。 |
| AnimePro FLUX GGUF | https://huggingface.co/advokat/AnimePro-FLUX | Text to image model. 文本到图像模型。 |
| Flex.1-alpha GGUF | https://huggingface.co/ostris/Flex.1-alpha https://huggingface.co/hum-ma/Flex.1-alpha-GGUF | Text to image model. 文本到图像模型。 |
Video generation 视频生成
| Name 名字 | Links 链接 | Additional information 其他信息： |
|---|---|---|
| Cinemo 西莫 | https://huggingface.co/spaces/maxin-cn/Cinemo GitHub - maxin-cn/Cinemo: [CVPR 2025] Consistent and Controllable Image Animation with Motion Diffusion Models | 1. Interface PyTorch uses PyTorch 2.4.0 1. 接口 PyTorch 使用 PyTorch 2.4.0 |
Music generation 音乐生成
| Name 名字 | Links 链接 | Additional information 其他信息： |
|---|---|---|
| ACE-Step ACE 步 | GitHub - ace-step/ACE-Step: ACE-Step: A Step Towards Music Generation Foundation Model |
Voice generation 语音生成
| Name 名字 | Links 链接 | Additional information 其他信息： |
|---|---|---|
| WhisperSpeech web UI WhisperSpeech 网页 UI | GitHub - Mateusz-Dera/whisperspeech-webui: Simple WhisperSpeech web UI GitHub - WhisperSpeech/WhisperSpeech: An Open Source text-to-speech system built by inverting Whisper. | |
| F5-TTS F5-TTS 系列 | GitHub - SWivid/F5-TTS: Official code for "F5-TTS: A Fairytaler that Fakes Fluent and Faithful Speech with Flow Matching" | 1. Remember to select the voice file when using the interface. 1. 使用界面时记得选择语音文件。 |
| Matcha-TTS 抹茶-TTS | GitHub - shivammehta25/Matcha-TTS: [ICASSP 2024] 🍵 Matcha-TTS: A fast TTS architecture with conditional flow matching | |
| Dia 日 | GitHub - nari-labs/dia: A TTS model capable of generating ultra-realistic dialogue in one pass. GitHub - tralamazza/dia: A TTS model capable of generating ultra-realistic dialogue in one pass. | 1. Script uses the optional-rocm-cuda fork by tralamazza. 1. 脚本使用 tralamazza 的 optional-rocm-cuda fork。 |
| Orpheus-TTS | https://huggingface.co/spaces/MohamedRashad/Orpheus-TTS/tree/main GitHub - canopyai/Orpheus-TTS: Towards Human-Sounding Speech | 1. If GPU is not detected change HIP_VISIBLE_DEVICES value. 1. 如果未检测到 GPU，请更改 HIP_VISIBLE_DEVICES 值。 |
| IMS-Toucan IMS-巨嘴鸟 | GitHub - DigitalPhonetics/IMS-Toucan: Controllable and fast Text-to-Speech for over 7000 languages! | 1. PyTorch 2.7.0 slowed down performance, so the interface uses version 2.4.0 1. PyTorch 2.7.0 降低了性能，因此界面使用 2.4.0 版本 |
3D generation 3D 生成
| Name 名字 | Links 链接 | Additional information 其他信息： |
|---|---|---|
| TripoSG TripoSG 系列 | GitHub - VAST-AI-Research/TripoSG: TripoSG: High-Fidelity 3D Shape Synthesis using Large-Scale Rectified Flow Models | 1. Added custom simple UI. 1. 添加了自定义简单 UI。 2. Sometimes there are probelms with the preview, but the model should still be available for download. 2. 有时预览存在问题，但模型仍应可供下载。 |
Tools 工具
| Name 名字 | Links 链接 | Additional information 其他信息： |
|---|---|---|
| Fastfetch 快速获取 | GitHub - fastfetch-cli/fastfetch: A maintained, feature-rich and performance oriented, neofetch like system information tool. | 1. Custom Fastfetch configuration with GPU memory info. 1. 带有 GPU 内存信息的自定义 Fastfetch 配置。 2. Script supports not only AMD but also NVIDIA graphics cards (nvidia-smi needed). 2. 脚本不仅支持 AMD，还支持 NVIDIA 显卡（需要 nvidia-smi）。 3. If you change the number or order of graphics cards you must run the installer again. 3. 如果更改了显卡的数量或顺序，则必须再次运行安装程序。 |
三、Instalation: 安装方式：
Note 注意
First startup after installation of the selected interface may take longer.
安装所选接口后首次启动可能需要更长的时间。
Important 重要
This script does not download any models. If the interface does not have defaults, download your own.
此脚本不会下载任何模型。如果界面没有默认值，请下载您自己的界面。
Caution 谨慎
If you update, back up your settings and models. Reinstallation deletes the previous directories.
如果您更新，请备份您的设置和模型。重新安装将删除以前的目录。
1. Add the user to the required groups.
1. 将用户添加到所需的组。
sudo adduser `whoami` video
sudo adduser `whoami` render 2. Reboot 2. 重新启动
sudo reboot 3. Clone repository 3. 克隆存储库
git clone https://github.com/Mateusz-Dera/ROCm-AI-Installer.git 4. Run installer 4. 运行安装程序
bash ./install.sh 5. Select installation path.
5. 选择安装路径。
6. Select ROCm installation if you are upgrading or running the script for the first time.
6. 如果您是首次升级或运行脚本，请选择 ROCm 安装。
7. Install selected interfaces
7. 安装选定的接口
8. Go to the installation path with the selected interface and run:
8. 转到包含所选接口的安装路径并运行：
./run.sh 软件下载
夸克网盘分享
本文信息来源于GitHub作者地址：GitHub - Mateusz-Dera/ROCm-AI-Installer: A script that automatically installs all the required stuff to run selected AI interfaces on AMD Radeon 7900XTX.
如何使用DROP TABLE删除表？)
：服务器A通过SSH调用服务器B执行Python自动化脚本)
