Can I have request?

[Mescalamba]

Basically only thing I would like is to be able to select level of quantitization for Alpha and/or Zero quantitizers.

Cant make any other quantitizer work, only Zero worked for example on CyberRealistic PONY.

[calcuis]

alpha works for most of the models in civitai; since those are based on sd to sdxl framework and some of them (the latest 6 months) are flux; zero is direct and simple; however, these two are convertors, not quantizers; we do write py scripts to convert the model straight to the lower tier(s) but due to some formatting issue(s), c/c++ works better recently; for the time being, you could try the cutter to further quantize the converted model file; see does it work or not

[Mescalamba]

Ah, I understand to some level. I think regular quant is a bit more than just "cut to bits". LLMs have even iq quants, which we sadly dont have for visual models that perform even better than regular ones. Kinda hope it can be done one day too, altho we do have SVDquants so.. maybe not needed.

Anyway, I managed to actually run City96 implementation, so no problem. Thanks.