2.0.2 InvokeAI 2.3.5

What's Changed

• InvokeAI 2.3.5 by @toyo-chi in #23

New Contributors

• @toyo-chi made their first contribution in #23

Full Changelog: 2.0.1...2.0.2

2.0.1

What's Changed

• fixed safetensors cargoDeps hash by @gbtb in #21

Full Changelog: 2.0.0...2.0.1

2.0.0 Flake and packages update

• InvokeAI and webui were updated to latest versions
• Replaced git submodules and devshells with proper packages. Now both UIs are nix-buildable and nix-runnable
• Since both repos are now placed into read-only nix store, some workflows might break (especially with webui). Please open issue if that's the case for you
• Tested basic functions of InvokeAI and webui with AMD gpus, can't gurantee it works with nvidia (feedback is welcome)

Please carefully read README on how to use this flake now, there were a lot of breaking changes in that regard (hence bump to 2.0.0 version) 😄

Thank you @atagen and @nixified-ai for making code contributions to this release 👍

1.0.1 Fix for nvidia overlays

Fixed an error that caused nvidia shells to not include cuda at all.

What's Changed

• fixed nvidia typo by @atagen in #15

New Contributors

• @atagen made their first contribution in #15

Full Changelog: 1.0.0...1.0.1

1.0.0 Support for InvokeAI 2.1

This release makes necessary(minor) adjustments to bring in support of InvokeAI 2.1. However, there are some breaking changes in models config and placement that you'll have to do at least semi-manually, and because of that it's major version release imo (even though InvokeAI just went with minor version bump). Release notes from InvokeAI

Upgrade instructions

InvokeAI 2.1 brings support of multiple models, including 1.4, 1.5 and inpainting. Because of that, they've changed expected naming for model weights checkpoint files and added yaml config file to set model settings and file locations.
Here is excerpt from README with necessary instructions:

• Weights download
  • Built-in way. Inside InvokeAI's directory, run python scripts/preload_models.py to preload main SD weights and support models. (Downloading Stable Diffusion Weighs will require HugginFace token)
  • Manual way. If you obtained SD weights from somewhere else, you can skip their download with preload_models.py. However, you'll have to manually create/edit InvokeAI/configs/models.yaml so that your models get loaded. Some example configs for SD 1.4, 1.5, 1.5-inpainting present in models.example.yaml .

0.1.2 Proper fix for whitespaces in a patch

Merge pull request #13 from jb55/fix-patch-again

fixup! webui.patch without extra whitespaces

0.1.1 - Fixed whitespace issues with webui.patch

Fixed whitespace issues with webui.patch

0.1.0: fixed a bug in requirementsFor for webui, downgraded transformers to …

First "stable-ish" release of this flake. Consists of InvokeAI 2.0.2 and latest commit of stable-diffusion-webui. Main functions, such as txt2img, img2img, inpainting, img2txt were tested with AMD flakes and should work.