sharrnah/stable-diffusion-guitard repository overview

⁠Docker image for Stable Diffusion

See Repository: https://github.com/Sharrnah/Stable-Diffusion-Docker⁠

⁠Setup for Windows

• For Windows using WSL2, write a .wslconfig File in your user path with following content:

[wsl2]
memory=12GB # Limits VM memory in WSL 2 to X GB
processors=3 # Makes the WSL 2 VM use X virtual processors
localhostForwarding=true

• You can play with the values if your PC can't handle it.

• Install Docker Desktop for Windows

• Make sure you are running the newest NVIDIA Drivers and newest Docker Desktop Version.

  • See https://docs.microsoft.com/en-us/windows/ai/directml/gpu-cuda-in-wsl⁠ for more information about WSL support.

    (all newer Driver and Docker versions should support it out of the box.)

    Test with:

    wsl cat /proc/version

    Needs kernel version of 5.10.43.3 or higher.

    Run

    wsl --update

    to update the kernel.

⁠Installation

⁠With Docker

• Build image yourself with:

  docker build -t stable-diffusion-guitard .

• Run prebuild image with:

  docker run -d --gpus all -p 7860:8080 -v ./.cache/app:/root/.cache -v ./.cache/facexlib/:/app/src/facexlib/facexlib/weights/ -v ./.cache/gfpgan/:/app/gfpgan/weights/ -v ./models/:/models/ -v ./outputs/:/outputs/ -e RUN_MODE=false sharrnah/stable-diffusion-guitard
  

  Copy

  (replace "sharrnah/stable-diffusion-guitard" image name with "stable-diffusion-guitard" to run self-build image)

  change "RUN_MODE" depending on your machine. (see Options⁠ for more info)

⁠With Docker Compose (recommended)

• Download / Clone this repo.

  (Or just get the docker-compose.yaml file if you want to use the prebuild image and ignore "build image yourself" step.)

• Start docker-compose project.

  • building image yourself:

  docker compose -f docker-compose.build.yaml up -d --build

  • start prebuild image with:

  docker compose up -d

• See current logs with:

  docker compose logs stablediffusion -f

• You can exec into the container with:

  docker compose exec stablediffusion bash

⁠Options

• You can set the environment variable "RUN_MODE" to one of these setting:
  • "OPTIMIZED" For reduced Memory mode (sacrificing speed)
  • "OPTIMIZED-TURBO" For lesser reduced Memory mode (sacrificing less speed)
  • "GTX16" When generated images are green (known problem on GTX 16xx GPUs)
  • "GTX16-TURBO" When generated images are green (known problem on GTX 16xx GPUs) [using OPTIMIZED-TURBO]
  • "FULL-PRECISION" use full precision
• Set the environment variable "WEBUI_RELAUNCH" to
  • "true" (default) For automatic restarting of the WebUI
  • "false" Disables automatic restarting of the WebUI
• Set the environment variable "ENABLE_TEXTUAL_INVERSION" to
  • "true" To clone and install Textual Inversion (Breaks Latent Diffusion SR!!!)
  • "false" (default) does not clone Textual Inversion on startup

For that you can create a .env file and set the content to

RUN_MODE=OPTIMIZED

or

RUN_MODE=GTX16

(See example.env file including all possible values)

⁠Usage

• after the webgui started successfully you should see a log output telling

  Running on local URL:  http://127.0.0.1:7860/
  

  Copy

• See current log with:

  docker compose logs stablediffusion -f

• Open http://127.0.0.1:7860/⁠ in your Browser to use it.

• All generated images are saved into the ./outputs/ directory.

• Find some nice example prompts here:

  • https://lexica.art/⁠

⁠Model Sources

⁠Sources:

• https://rentry.org/GUItard⁠
• https://github.com/sd-webui/stable-diffusion-webui⁠
• https://github.com/Hafiidz/latent-diffusion⁠