Sharrnah/stable-diffusion-docker

Stable Diffusion Docker Project with WebUI

Docker for Stable Diffusion

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

  (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/
  

• 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

Models should be downloaded automatically on first run now!

Model	Function	Version	Size	Local Location (Case-Sensitive!)	Download Source
Stable Diffusion	Image Generation	v1.4	~4 GB	models/SDv1.4.ckpt	Web: https://drive.yerf.org/wl/?id=EBfTrmcCCUAGaQBXVIj5lJmEhjoP1tgl https://www.googleapis.com/storage/v1/b/aai-blog-files/o/sd-v1-4.ckpt?alt=media Torrent Magnet: magnet:?xt=urn:btih:3a4a612d75ed088ea542acac52f9f45987488d1c&dn=sd-v1-4.ckpt&tr=udp%3a%2f%2ftracker.openbittorrent.com%3a6969%2fannounce&tr=udp%3a%2f%2ftracker.opentrackr.org%3a1337 Hugging Face: https://huggingface.co/CompVis/stable-diffusion-v1-4
GFPGAN	Face Correction	v1.3.0	332 MB	models/GFPGANv1.3.pth	Web: https://github.com/TencentARC/GFPGAN/releases/download/v1.3.0/GFPGANv1.3.pth
RealESRGAN	Upscaling	v0.1.0	~64 MB	models/RealESRGAN_x4plus.pth	Web: https://github.com/xinntao/Real-ESRGAN/releases/download/v0.1.0/RealESRGAN_x4plus.pth
RealESRGAN anime	Upscaling Anime	v0.2.2.4	~17 MB	models/RealESRGAN_x4plus_anime_6B.pth	Web: https://github.com/xinntao/Real-ESRGAN/releases/download/v0.2.2.4/RealESRGAN_x4plus_anime_6B.pth
Latent Diffusion	Upscaling, Inpainting	July 2022	~2 GB	models/LDSR.ckpt	Web: https://heibox.uni-heidelberg.de/f/578df07c8fc04ffbadf3/?dl=1

Sources:

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