Description
Steps : 8
Sampler/Scheduler : Seed_3 / Normal --- ipndm / Normal ---- heunpp2 / linear_quadratic or beta
Weighting : 0.8
FAQ
Comments (4)
Have you found the increased size (rank/filesize) makes a noticeable difference? I was thrilled with the 20MB ones you had, was a no-brainer to download and try (ZIT)
Since Z Image is a distilled Turbo model, the LoRa training approach is quite different compared to Flux1. I switch between Rank 16 and 32 depending on the texture detail and dataset resolution. With Flux1, I was able to obtain good results with Ranks 2 and 4. This is no longer feasible with Z Image
@Adel_AI I see, thanks for the insight. Fingers crossed for the base model to come sooner rather than later.
Today we can train models with base one.
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.