A script that automatically installs all the required stuff to run selected AI interfaces on AMD Radeon 7900XTX. It should also work on 7900XT cards. For other cards, change HSA_OVERRIDE_GFX_VERSION at the beginning of the script (Not tested).
Note
Ubuntu 24.04 is recommended. Version 5.x is not tested on older systems.
Part of the installation script is based on this guide: https://github.com/nktice/AMD-AI/blob/main/
| Name | Info |
|---|---|
| CPU | AMD Ryzen 7900X3D (iGPU disabled in BIOS) |
| GPU | AMD Radeon 7900XTX |
| RAM | 64GB DDR5 6600MHz |
| Motherboard | ASRock B650E PG Riptide WiFi (2.10) |
| OS | Ubuntu 24.04 |
| Kernel | 6.8.0-39-generic |
| ROCm | 6.1.3 |
| Name | Enviroment | Links | Additional information |
|---|---|---|---|
| KoboldCPP | Python 3.12 venv | https://github.com/YellowRoseCx/koboldcpp-rocm | |
| Text generation web UI | Python 3.11 venv | https://github.com/oobabooga/text-generation-webui https://github.com/ROCm/bitsandbytes.git https://github.com/ROCmSoftwarePlatform/flash-attention https://github.com/turboderp/exllamav2 https://github.com/abetlen/llama-cpp-python |
1. Tested: ExLlamav2, llama.ccp, Transformers 2. Requrements for Superbooga are installed, but the extension is not enabled by default 3. Requrements for SuperboogaV2 are installed, but the extension is not enabled by default |
| SillyTavern (1.12.4) | Node | https://github.com/SillyTavern/SillyTavern |
| Name | Enviroment | Links | Additional information |
|---|---|---|---|
| Stable Diffusion web UI | Python 3.11 venv | https://github.com/AUTOMATIC1111/stable-diffusion-webui | 1. Startup parameters are in the webui-user.sh file |
| ANIMAGINE XL 3.1 | Python 3.12 venv | https://huggingface.co/spaces/cagliostrolab/animagine-xl-3.1 https://huggingface.co/cagliostrolab/animagine-xl-3.1 |
|
| ComfyUI ComfyUI-CLIPSeg AuraFlow AuraSR FLUX.1-schnell |
Python 3.12 venv | https://github.com/comfyanonymous/ComfyUI https://github.com/biegert/ComfyUI-CLIPSeg https://huggingface.co/fal/AuraFlow https://huggingface.co/fal/AuraFlow-v0.2 https://huggingface.co/fal/AuraSR https://huggingface.co/fal/AuraSR-v2 https://github.com/alexisrolland/ComfyUI-AuraSR https://huggingface.co/black-forest-labs/FLUX.1-schnell https://huggingface.co/Comfy-Org/flux1-schnell/blob/main/ |
1. Flux examples: https://comfyanonymous.github.io/ComfyUI_examples/flux/#simple-to-use-fp8-checkpoint-version |
| Artist | Python 3.12 venv | https://github.com/songrise/Artist/ |
| Name | Enviroment | Links | Additional information |
|---|---|---|---|
| Cinemo | Python 3.12 venv | https://huggingface.co/spaces/maxin-cn/Cinemo https://github.com/maxin-cn/Cinemo |
| Name | Enviroment | Links | Additional information |
|---|---|---|---|
| AudioCraft | Python 3.12 venv | https://github.com/facebookresearch/audiocraft |
| Name | Enviroment | Links | Additional information |
|---|---|---|---|
| WhisperSpeech web UI | Python 3.12 venv | https://github.com/Mateusz-Dera/whisperspeech-webui https://github.com/collabora/WhisperSpeech https://github.com/ROCmSoftwarePlatform/flash-attention |
|
| MeloTTS | Python 3.12 venv | https://github.com/myshell-ai/MeloTTS |
| Name | Enviroment | Links | Additional information |
|---|---|---|---|
| TripoSR | Python3.12 venv | https://github.com/VAST-AI-Research/TripoSR https://github.com/ROCmSoftwarePlatform/flash-attention |
1. It uses PyTorch ROCm, but torchmcubes is built for the CPU. This method is still faster than using just PyTorch CPU-only version. |
| Name | Enviroment | Links | Additional information |
|---|---|---|---|
| ExLlamaV2 | Python3.12 venv | https://github.com/turboderp/exllamav2 | 1. LLM conversion to exl2 format using convert.py 2.Run: export HSA_OVERRIDE_GFX_VERSION=11.0.0export CUDA_VISIBLE_DEVICES=0source .venv/bin/activate |
Note
First startup after installation of the selected interface may take longer.
Important
This script does not download any models. If the interface does not have defaults, download your own.
Caution
If you update, back up your settings and models. Reinstallation deletes the previous directories.
- Add the user to the required groups.
sudo adduser `whoami` video
sudo adduser `whoami` render- Reboot
sudo reboot- Run installer
bash ./install.sh- Select installation path.
- Select ROCm installation if you are upgrading or running the script for the first time.
- Install selected interfaces
- Go to the installation path with the selected interface and run:
./run.sh