Is this Lora model extracted from xiaozaa/catvton-flux-beta?
#1
by
makisekurisu-jp
- opened
What the differences are between the model types rank128, rank64, and rank32?
Yes, they are extracted from that model and are meant to be used with FLUX.1-Fill-dev. Rank128, etc. refer to the network rank that was set when extracting. Typically, the higher the rank, the higher the details from my understanding. In brief testing, sometimes rank32 was good enough, but sometimes rank128 was needed. These are usually enhanced by combining them with Redux.