An attempt to make a LoRa that patches a model to at least make some sort of output with a different prediction method instead of complete garbage. It's not the highest quality and needs more experimentation however, but it's a fun toy.
How To Use:
Use double the normal CFG and step count
Patching EPS to V-Pred sampling: Weight = 1.0
Patching V-Pred to EPS sampling: Weight = -1.0
Highres-fix and SPO both help with the output quality
This LoRa doesn't work well with some Illustrious V2.0 based models as it was based on a merge of V0.1 and V1.0 mostly
Colors will tend to be more faded than the model normally would produce but this might be fixable to some degree :)
Description
Use 1.0 weight for V-Pred, or -1.0 for EPS
Use approximately double the normal CFG and step count or the result will probably be excessively undersaturated.
May not work well with Illustrious V2.0-based models.
This first version is an extraction between my Frosty Mix v3.0 model (meant to be a "middle ground" EPS model) and an unreleased version of my Natural Noob XL model tuned to look more like Frosty Mix.
Looks like we don't have an active mirror for this file right now.
CivArchive is a community-maintained index — we catalog mirrors that volunteers upload to HuggingFace, torrents, and other public hosts. Looks like no one has uploaded a copy of this file yet.
Some files do get recovered over time through contributions. If you're looking for this one, feel free to ask in Discord, or help preserve it if you have a copy.
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.