mcmonkey4eva
commented
1 month ago I wouldn't do the two-letter code thing, just "short codes" in the 3/4 char range usually - SDXL, SDv1, SD3, Aura, ...
And just use HTML format magic to keep the shape consistent
(Sloppy mockup in image editor)
These short codes would need to be added to the T2IModelClassSorter and transmitted alongside CompatClass in model definitions, and the tab complete code would need to parse and display those (Offhand I don't think that code pulls from the metadata list, just a name list, so that needs tweaking too)
Feature Idea
As SwarmUI already does internal analysis/tagging (either via metadata field or by tensor names/etc.) to determine the base model of each sub-model...
I think it would be helpful to add some sort of sign to the subpath autocomplete drop downs that occur in the prompt editor (<embed:... , <lora:..., etc.) as well as the drop downs that occur in the model search boxes (including the grid parameters) that indicate the BASE MODEL that the sub-model was trained on/to work with.
As per discussion, icons would be somewhat more complicated, so perhaps something like a monospaced two-letter code to the left of each sub-path? E.g.
S1 = SD1.4 and SD1.5 (as they are nearly equivalent) and all finetunes. S2 = SD2.x and all finetunes XL = SDXL and all finetunes CA = Stable Cascade...etc.
Note: CivitAI uses SD1 and SD2 similar to S1 and S2 above, which would also be fine...as long as the pathnames don't start looking all janky due to different prefix widths.
XL /negative/Detail++ XL /positive/zPDXL XL /negative/zPDXL-neg S1 /negative/EasyNegative
Perhaps also an option to allow the user to override the assignment of extreme finetunes (such as PONY 6 XL or NAI) with a custom indicator such as PD (since that's not strictly determinable via the standardized metadata nor the tensor names). Not really sure there's anything to do automatically other than leave those in the appropriate XL or S1 category.
B
Other
No response