Why only Q4 or full fat?
the base model doesn't list any other quants yet from someone else. I usually use q8 since it halfs the size by equal quality at the cost of some speed.
While Q4 is usually a case of IF you have to... Which i spent a lot of money to NOT have to :)
the model card lists a Q8 but it is not in files.
I want to ask and I don't want to start a new thread. By any chance nvfp4? Anyway, thanks for the great work. I look forward to your next amazing work.
finally, thanks.
I'm still not convinced by it for editing. Trying to edit a person (a british guy) it simply made him chinese. while maintaining the rest of the image, the pose and so on perfectly and edited what i wanted it to. But what use is that if consistency of the actual person is just zero? I mean that is just completely useless as an editing model because it does half of it very well and the other half ... not at all.
exact same thing with qe2511 slotted in: perfect.
It is nice to have more options for simple image generation, which it seems to be good at. At least different enough from QI.