Default choices added to settings in user interface section
Choices are editable by user
User selects from dropdown.
When you move one slider, the other adjusts according to the ratio
chosen.
Vice versa for the other slider.
Number fields for changes work as well.
For disabling ratio, an unlock pad "🔓" is available as a default.
This string can be changed to anything to serve as a disable,
as long as there is no colon ":".
Ratios are entered in this format, floats or ints with a colon "1:1".
The string is split at the colon, parses left and right as floats to
perform the math.
Adds "Upcast cross attention layer to float32" option in Stable Diffusion settings. This allows for generating images using SD 2.1 models without --no-half or xFormers.
In order to make upcasting cross attention layer optimizations possible it is necessary to indent several sections of code in sd_hijack_optimizations.py so that a context manager can be used to disable autocast. Also, even though Stable Diffusion (and Diffusers) only upcast q and k, unfortunately my findings were that most of the cross attention layer optimizations could not function unless v is upcast also.
This also handles type casting so that ROCm and MPS torch devices work correctly without --no-half. One cast is required for deepbooru in deepbooru_model.py, some explicit casting is required for img2img and inpainting. depth_model can't be converted to float16 or it won't work correctly on some systems (it's known to have issues on MPS) so in sd_models.py model.depth_model is removed for model.half().
remove random artist built-in extension (to re-added as a normal extension on demand)
remove artists.csv (but what does it mean????????????????????)
make interrogate buttons show Loading... when you click them
do not save config if it's the default for checkpoint merger tab
change file naming scheme for checkpoint merger tab
allow just saving A without any merging for checkpoint merger tab
some stylistic changes for UI in checkpoint merger tab
