mirror of
https://github.com/AUTOMATIC1111/stable-diffusion-webui.git
synced 2024-12-28 10:25:05 +08:00
Page:
Optimum SDXL Usage
Pages
API
Change model folder location
Command Line Arguments and Settings
Containers
Contributing
Custom Images Filename Name and Subdirectory
Custom Scripts
Dependencies
Developing custom scripts
Developing extensions
Extensions index
Extensions
Features
Guides and Tutorials
Home
How to make your own Inpainting model
How to switch to different versions of WebUI
Install and Run on AMD GPUs
Install and Run on NVidia GPUs
Installation on Apple Silicon
List of Time Zones
Localization
Negative prompt
On the benefits of writing short, concise code
Online Services
Optimizations
Optimum SDXL Usage
Seed breaking changes
Tests
Textual Inversion
Troubleshooting
User Interface Customizations
Xformers
6
Optimum SDXL Usage
ClashSAN edited this page 2023-12-15 06:35:24 +00:00
Table of Contents
Here's a quick listing of things to tune for your setup:
Commandline arguments:
- Nvidia (12gb+)
--xformers
- Nvidia (8gb)
--medvram-sdxl --xformers
- Nvidia (4gb)
--lowvram --xformers
- AMD (4gb)
--lowvram --opt-sub-quad-attention
+ TAESD in settingsBoth rocm and directml will generate at least 1024x1024 pictures at fp16. If your AMD card needs --no-half, try enabling --upcast-sampling instead, as full precision sdxl is too large to fit on 4gb.
System:
- (Windows) Not all nvidia drivers work well with stable diffusion. All drivers above version 531 can cause extreme slowdowns on Windows when generating large images towards, or above your card's maximum vram. To mitigate this potential speed degradation, follow the steps outlined by nvidia on their website. https://nvidia.custhelp.com/app/answers/detail/a_id/5490 Related issues: (vladmandic/automatic/discussions/1285), (#11063).
- (Linux) install
tcmalloc
, greatly reducing RAM usage:sudo apt install --no-install-recommends google-perftools
(#10117). - Add a pagefile/swapfile to prevent failure loading weights due to low RAM.
- Use an SSD for faster load time, especially if a pagefile is required.
- Have at least 24gb ram on Windows 11, and at least 16gb on Windows 10
Model weights:
- Use sdxl-vae-fp16-fix; a VAE that will not need to run in fp32. This will increase speed and lessen VRAM usage at almost no quality loss.
- Use TAESD; a VAE that uses drastically less vram at the cost of some quality.
Setup
- Install and run on NVidia GPUs
- Install and run on AMD GPUs
- Install and run on Apple Silicon
- Install and run on Intel Silicon (external wiki page)
- Install and run via container (i.e. Docker)
- Run via online services
Reproducing images / troubleshooting
Usage
- Features
- Command Line Arguments and Settings
- Optimizations
- Custom Filename Name and Subdirectory
- Change model folder location e.g. external disk
- User Interface Customizations
- Guides and Tutorials
Developers
This is the Stable Diffusion web UI wiki. Wiki Home