Releases: AbdBarho/stable-diffusion-webui-docker
9.0.0
Breaking Change: Invoke UI removed
If you want to continue using Invoke, please refer to their excellent docs
https://invoke-ai.github.io/InvokeAI/installation/040_INSTALL_DOCKER/
Invoke was one of the first and honestly the most user-friendly UIs that popped up around the SD 1 and 1.5, unlike many others, who relied on gradio and streamlit, invoke had actual web-development inside of it with all the bells and whistles of a real webapp, and still strikes to this day how stunning it looks.
When I started this project many years ago, docker was, and perhaps still is, alien to most people (JUST GIVE ME THE DAMN .EXE), and so I chose to maintain docker versions of the UI for the sake of simple usage. But now, the invoke team already maintains a docker setup for their UI, making the version in this repo irrelevant.
And to be honest, I could not keep up with everything.
Juggling 3 different UIs with their own setup / configuration / folder structure and requirements, while still trying to share data between them, was a good idea at first, but became daunting and not possible anymore.
To increase maintainability, and to keep me motivated to work on this, I will try to simplify how everything works here for my own sake. Starting by removing Invoke, and more changes to come.
What's Changed
- Update package index by @AbdBarho in #650
- Pin xformers (for now) by @AbdBarho in #651
- Update Auto v1.9.3 by @SachiaLanlus in #673
- Updated docker-compose to remove obsolete version syntax by @derek-palmer in #692
- Add missing assets to auto1111 by @KagurazakaNyaa in #684
- Update Comfy by @AbdBarho in #693
- Bump pytorch containers by @AbdBarho in #697
- Update Auto to 1.9.4 by @mohamednabiel717 in #700
- Remove invoke by @AbdBarho in #705
New Contributors
- @SachiaLanlus made their first contribution in #673
- @derek-palmer made their first contribution in #692
- @mohamednabiel717 made their first contribution in #700
Full Changelog: 8.2.0...9.0.0
Contributors
Assets 2
8.2.0
Happy new year! 🎉
This release contains some updates and bug fixes for auto and comfy.
Unfortunately, a total rework of invoke is necessary for v3, have not made that yet
Also, might abandon the idea of sharing the models between the UIs since it is causing more maintenance hassle. Let's see what happens.
What's Changed
- Auto 1.5.2 by @AbdBarho in #579
- Update Comfy by @AbdBarho in #580
- Update generative-models version by @66li in #581
- Skipping installation of requirements for disabled extensions by @DevilaN in #582
- Update Automatic1111 Stable Diffusion Webui to v1.6.0. by @cloudaxes in #585
- fix: Update automatic1111 webui base image by @cococig in #601
- Update Auto by @AbdBarho in #610
- comfy version up by @sejoung in #603
- Resolve memory usage situation in Auto by @simonmcnair in #620
- Update Auto v1.7.0 by @adamflorizone in #632
- Update open_clip to v2.20.0 in Auto by @jsjolund in #617
New Contributors
- @66li made their first contribution in #581
- @cococig made their first contribution in #601
- @sejoung made their first contribution in #603
- @simonmcnair made their first contribution in #620
- @adamflorizone made their first contribution in #632
- @jsjolund made their first contribution in #617
Full Changelog: 8.1.0...8.2.0
Contributors
Assets 2
8.1.0
Notable changes
Now with SDXL 1.0 support, thanks to @mashb1t!
For now, only the base model is supported, which you can download from https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0/blob/main/sd_xl_base_1.0.safetensors and put in data/models/Stable-diffusion
It is possible to use the refiner, but you would need an extension in the time being, more info in the PR #560 and in A1111 docs here
https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Features#sd-xl
What's Changed
- Modify installation extension dependencies by @KagurazakaNyaa in #518
- Create dirs if not exist in #520
- Allow setting port through env WEBUI_PORT in #521
- Add build-essential package by @KagurazakaNyaa in #522
- Auto 1.4.0 in #507
- Upgrade k-diffusion to Release 0.0.15 to get access to DPM++ (2M) SDE sampler. by @cloudaxes in #537
- Add startup script for comfy in #552
- Update Automatic1111 to 1.5.1 to add compatibility for SDXL by @mashb1t in #560
New Contributors
- @KagurazakaNyaa made their first contribution in #518
- @cloudaxes made their first contribution in #537
- @mashb1t made their first contribution in #560
Full Changelog: 8.0.0...8.1.0
Contributors
Assets 2
8.0.0
Breaking changes
Due to the increasing number of models and folders required for the UIs and extensions, the folder structure have been changed, now it looks something like this:
.
├── data
│ ├── .cache # cache for intermediary packages
│ ├── config # UI specific config
│ ├── embeddings # embeddings
│ └── models # all the models that were previously scattered around in /data
# this folder will be mounted replacing `auto`s model folder, so any extensions
# that write or read from this folder should see itThere is a migration script that would restructure the files for you, you can run it in bash or in WSL from the root of this repo:
sudo ./.devscripts/migratev7tov8.shNotes
- Now automatically installs extension dependencies in auto if any are found, the
startup.shscript will still be executed AFTER the installation finishes. - Bump comfy to torch 2.0.1
Full Changelog: 7.0.0...8.0.0
7.0.0
Breaking changes
Updated the compute capabilities in the docker compose file from gpu to [compute, utility], if you face any problems, please create an issue and specify your GPU.
What's Changed
- Add extension dependencies in #485
- Update auto to 1.3.0 in #490
- InvokeAI 2.3.5.post2 in #491
- Change capabilities in docker-compose in #497
Full Changelog: 6.0.0...7.0.0
6.0.0
Breaking changes
Maybe a short story:
It was shortly after stable diffusion 1.4 was announced, August of 2022, and I was really excited to try it out myself. I was surprised to see the community working on a UI for using stable diffusion.
I did not want to mess up my system setup with python, virtualenv and such. I was (and still am) learning how to use containers. So I built the webui in the container, and decided to publish it on github, maybe someone else needs it. And it took off! This web UI kept getting cool and new features, and keeping up to date was fun, challenging, and exciting.
This lasted for about 2 months, until A1111 took over. This previous UI started to get less and less users, features, and maintenance started to become harder. For months, the UI remained still with no updates, to this day, there are almost no changes anymore.
This UI was renamed to become https://github.com/Sygil-Dev/sygil-webui, it was the UI that started this repo. It was a long and wonderful journey, and I am sad to see this end. Thanks to everyone who worked on this awesome project.
Deprecate Sygil in #450
What's Changed
- Update auto to pytorch 2 in #442
- Only run if changed in #444
- Fix overwriting existing values in config.json in auto by @wookiefriseur in #418
- Invoke 2.3.5 in #449
- Update/fix mounts for comfyUI by @LEv145 in #432
- Add ClipEncoder in #458
- Create custom_nodes dir in #461
- Update Auto to 1.2.1 in #466
- Deprecate Sygil in #450
New Contributors
- @wookiefriseur made their first contribution in #418
Full Changelog: 5.0.3...6.0.0
Contributors
Assets 3
5.0.3
Changes
Comfy UI support
Thanks to @PassiveLemon, this repo now has support for comfy UI. This is still in early phases so expect bugs and please report them so we can work on them!
Auto update
This will be the final release where auto uses pytorch 1.13.1, moving forward, all new releases will use pytorch 2.0.0
What's Changed
- Support for ComfyUI by @PassiveLemon in #384
- Fix file permissions by @LEv145 in #425
- Add tty to comfy service by @divens in #429
- LyCORIS - ModelScope in #439
- Update Auto in #379
- Add styles.csv support in #440
New Contributors
- @PassiveLemon made their first contribution in #384
- @LEv145 made their first contribution in #425
- @divens made their first contribution in #429
Full Changelog: 5.0.2...5.0.3
Contributors
Assets 4
5.0.2
Notes
- Auto won't be bumped to latest version yet, because it is very unstable #379
- The invoke team has been working hard on making the installation process as seamless as possible, which includes downloading models from huggingFace semi-automatically, using the models from
data/StableDiffusionis harder and much finiky, however, you can still use the UI to achieve that.- it might be the case that the current setup does a lot of prefetching on start up, please open an issue if you notice that is the case.
What's Changed
- Add libgoogle-perftools-dev in #341
- Fix preload for Invoke in #346
- enable ControlNet mounts for AUTOMATIC1111 by @gmasil in #385
- add mounts for openpose by @gmasil in #387
- Make Dockerfiles OCI compliant by @RedTopper in #408
- Refactor invoke in #405
New Contributors
- @gmasil made their first contribution in #385
- @RedTopper made their first contribution in #408
Full Changelog: 5.0.1...5.0.2
