Decided to try out converting models to Lora whatevers while in a string of meetings I really don't need to be in.
Tried with my conversion of hll 1 first:
https://files.catbox.moe/cak2ro.pngAnd then threw some more Loras at it.
https://files.catbox.moe/igjthj.pngI think this is pretty good for disk space, but I really wish the results were equivalent to model mixing. I don't know if the generations are better or worse. It's just different.
