## What type of PR is this? (check all applicable)
- [ ] Refactor
- [ ] Feature
- [ ] Bug Fix
- [x] Optimization
- [ ] Documentation Update
- [ ] Community Node Submission
## Have you discussed this change with the InvokeAI team?
- [ ] Yes
- [ ] No, because:
## Have you updated all relevant documentation?
- [ ] Yes
- [ ] No
## Description
## Related Tickets & Documents
<!--
For pull requests that relate or close an issue, please include them
below.
For example having the text: "closes #1234" would connect the current
pull
request to issue 1234. And when we merge the pull request, Github will
automatically close the issue.
-->
- Related Issue #
- Closes #
## QA Instructions, Screenshots, Recordings
<!--
Please provide steps on how to test changes, any hardware or
software specifications as well as any other pertinent information.
-->
## Added/updated tests?
- [ ] Yes
- [ ] No : _please replace this line with details on why tests
have not been included_
## [optional] Are there any post deployment tasks we need to perform?
This is a relatively stable release that corrects the urgent windows
install and model manager problems in 3.0.1. It still has two known
bugs:
1. Many inpainting models are not loading correctly.
2. The merge script is failing to start.
- Remove FaceMask and add link FaceTools repository, which includes FaceMask, FaceOff, and FacePlace
- Move Ideal Size up from under the template entry
## What type of PR is this? (check all applicable)
- [ X] Bug Fix
## Have you discussed this change with the InvokeAI team?
- [X] Yes - bug discovered by jpphoto
- [ ] No, because:
## Have you updated all relevant documentation?
- [ ] Yes
- [ X] Not needed
## Description
The user can customize the location of the models directory by setting
configuration variable `models_dir`. However, the model manager and the
TUI installer were all treating model relative paths as relative to the
invokeai root rather than the designated models directory. This has been
fixed by changing path resolution calls from using `config.root_path` to
`config.models_path`
Unfortunately there were many instances that needed replacement, so this
needs a bit of functional testing - try adding models, removing models,
renaming them, converting checkpoints, etc.
## What type of PR is this? (check all applicable)
- [ X] Optimization
## Have you discussed this change with the InvokeAI team?
- [X ] Yes
- [ ] No, because:
## Have you updated all relevant documentation?
- [X ] Yes
- [ ] No
## Description
This PR does two things:
1. if the environment variable INVOKEAI_ROOT is defined at install time,
the zipfile installer will default to its value when asking the user
where to install the software
2. If the user has more than 72 models of any type installed, then the
list will be truncated in the TUI and the user given a warning. Anything
larger than this number of models causes the vertical space to overflow.
The only effect of truncation is that the user will not be able to see
and delete the models that were truncated. The message advises the user
to go to the Web Model Manager interface in this event.
## What type of PR is this? (check all applicable)
- [X ] Bug Fix
## Have you discussed this change with the InvokeAI team?
- [ X] Yes
- [ ] No, because:
## Have you updated all relevant documentation?
- [ ] Yes
- [ X] No
## Description
This PR fixes several issues with the 3.0.0 conversion script:
- Handles checkpoint variants that don't put dots between fields in the
long state dict key names
- Handles ema, non-ema, pruned and non-pruned ckpts
- Does not add safety checker to converted checkpoints
- Respects precision of original checkpoint file
## What type of PR is this? (check all applicable)
- [ X] Bug Fix
## Have you discussed this change with the InvokeAI team?
- [ X] Yes
- [ ] No, because:
## Have you updated all relevant documentation?
- [ ] Yes
- [X] Not needed
## Description
Windows users have been getting a lot of OSErrors while installing 3.0.1
during the pip dependency installation phase. Generally the errors have
involved just two packages, pydantic and numpy. Looking at the install
logs, I see that both of these packages are first installed under one
version number by a dependency, and then uninstalled and replaced by a
slightly different version specified in invoke's `pyproject.toml`. I
think this is the problem - maybe the earlier package is not completely
closed before it is uninstalled and reinstalled.
This PR relaxes pinning of numpy and pydantic in `pyproject.toml`.
Everything seems to install and run properly. Hopefully it will address
the windows install bug as well.
## What type of PR is this? (check all applicable)
- [x] Bug Fix
## Have you discussed this change with the InvokeAI team?
- [x] Yes
## Description
- SDXL Metadata was not being retrieved. This PR fixes it.
