Skip to content

Latest commit

 

History

History
56 lines (45 loc) · 2.77 KB

Installation.md

File metadata and controls

56 lines (45 loc) · 2.77 KB
Raw

Installation Guide

by. Haoming

for Automatic1111 / Forge Webui
(on Windows)

Important

For Apple Silicon, refer to this Tutorial page instead
For AMD GPU, refer to the repo for AMD GPU instead

Tip

  • Install the UI and put the models on an SSD to improve load time
  • Enable View/Show/File name extensions for the File Explorer to locate the files
  • Open Nvidia Control Panel, go to Manage 3D settings, and set System Fallback Policy to Prefer No System Fallback

Download

  • for Forge
    1. Go to the Release page
    2. Click on webui_forge_cu124_torch24.7z to download
  • for Automatic1111
    1. Go to the Release page
    2. Click on sd.webui.zip to download

Installation

  1. Extract all contents to a folder of choice
    • Do not put them under OneDrive or other cloud drives
  2. Run update.bat first
  3. Modify the commandline args as mentioned below
  4. Download a checkpoint of choice first
    • Otherwise, the Webui will automatically download a ~2 GB checkpoint

    Refer to Checkpoint if you don't know what this means

  5. Run run.bat
    • On the first launch, the Webui will download and install all the required packages; this can take some time so be patient; read the rest of the Guide in the meantime
  6. When the installation is finished, it will automatically open a browser window showing the UI

Commandline Args

In the webui folder, right-click the webui-user.bat file and select Edit. Then, add the following flags as needed to the COMMANDLINE_ARGS= line:

  • --xformers: speed up the generation and reduce memory consumption
  • --listen: allow you to access the Webui from local network
  • --port: specify which port to use
  • --api: allow interaction with the Webui using Restful APIs
    • See the Wiki for more info

Note

The following memory-related flags are only needed when using Automatic1111

  • --medvram-sdxl: if you have less than 12 GB VRAM but wish to use SDXL checkpoints
  • --medvram: if you have less than 8 GB VRAM
  • --lowvram: if you're still getting CUDA Out of Memory Error