`options` slice was huge and managed a mix of generation parameters and general app settings. It has been split up:
- Generation parameters are now in `generationSlice`.
- Postprocessing parameters are now in `postprocessingSlice`
- UI related things are now in `uiSlice`
There is probably more to be done, like `gallerySlice` perhaps should only manage internal gallery state, and not if the gallery is displayed.
Full-slice selectors have been made for each slice.
Other organisational tweaks.
# enhance model_manager support for converting inpainting ckpt files
Previously conversions of .ckpt and .safetensors files to diffusers
models were failing with channel mismatch errors. This is corrected
with this PR.
- The model_manager convert_and_import() method now accepts the path
to the checkpoint file's configuration file, using the parameter
`original_config_file`. For inpainting files this should be set to
the full path to `v1-inpainting-inference.yaml`.
- If no configuration file is provided in the call, then the presence
of an inpainting file will be inferred at the
`ldm.ckpt_to_diffuser.convert_ckpt_to_diffUser()` level by looking
for the string "inpaint" in the path. AUTO1111 does something
similar to this, but it is brittle and not recommended.
- This PR also changes the model manager model_names() method to return
the model names in case folded sort order.
- Diffusers Sampler list is independent from CKPT Sampler list. And the
app will load the correct list based on what model you have loaded.
- Isolated the activeModelSelector coz this is used in multiple places.
- Possible fix to the white screen bug that some users face. This was
happening because of a possible null in the active model list
description tag. Which should hopefully now be fixed with the new
activeModelSelector.
I'll keep tabs on the last thing. Good to go.
For the torch and torchvision libraries **only**, the installer will now
pass the pip `--force-reinstall` option. This is intended to fix issues
with the user getting a CPU-only version of torch and then not being
able to replace it.
Previously conversions of .ckpt and .safetensors files to diffusers
models were failing with channel mismatch errors. This is corrected
with this PR.
- The model_manager convert_and_import() method now accepts the path
to the checkpoint file's configuration file, using the parameter
`original_config_file`. For inpainting files this should be set to
the full path to `v1-inpainting-inference.yaml`.
- If no configuration file is provided in the call, then the presence
of an inpainting file will be inferred at the
`ldm.ckpt_to_diffuser.convert_ckpt_to_diffUser()` level by looking
for the string "inpaint" in the path. AUTO1111 does something
similar to this, but it is brittle and not recommended.
- This PR also changes the model manager model_names() method to return
the model names in case folded sort order.
test-invoke-pip.yml:
- enable caching of pip dependencies in `actions/setup-python@v4`
- add workflow_dispatch trigger
- fix indentation in concurrency
- set env `PIP_USE_PEP517: '1'`
- cache python dependencies
- remove models cache (since we currently use 190.96 GB of 10 GB while I
am writing this)
- add step to set `INVOKEAI_OUTDIR`
- add outdir arg to invokeai
- fix path in archive results
model_manager.py:
- read files in chunks when calculating sha (windows runner is crashing
otherwise)
- Issue is that if insufficient diffusers models are defined in
models.yaml the frontend would ungraciously crash.
- Now it emits appropriate error messages telling user what the problem
is.
- Issue is that if insufficient diffusers models are defined in
models.yaml the frontend would ungraciously crash.
- Now it emits appropriate error messages telling user what the problem
is.
