This is a merge I made with the Chroma Uncanny model and a few other LoRAs. It works with other LoRAs as well if needed. I personally run it for 36 steps at CFG 1 using Euler + Bong Tangent, and it should work in any workflow.
Description
FAQ
Comments (14)
Any chanсe for FP8/NF4 quants?
uh I don't really know how to make nf4 but I can upload a fp8, I didn't think anyone would use a fp8 so I made just the q8, Will convert to fp8 and share maybe tomorrow.
@kakkkarot GGUFs work slower in my case (RTX3050 8 gigs) than corrensonding FP8/NF4 quants.
Thanks for your reply!
Can you make q4 version?
@mphobbit oh, alright will update a fp8 then
@kakkkarot Thanks, will try.
Worse results than normal uncanny, don't waste your time with this.
No one's forcing you to use the model, it's just a merge with the loras I use the most and thought I'd share if people wanted to use. It works on CFG 1 but I bet u didn't do that, cheers.
maybe u r new to use chroma , coz i m having great results with this model
@markhassain3712 even the images OP has posted look awful for a chroma model, maybe they are also new to chroma
@markhassain3712 did i offend your boyfriend?
I absolutely cannot confirm that, sorry.