Long time no see?Not much to say compared to how long I've been quiet.As everyone knows, the new trend has shifted to [anima], and I've been using it happily.Also, though not recently, I upgraded my graphics card, so my work environment has improved, but since real life has been busy, I haven't been able to put much time into model development (and it'll probably stay that way for a while).The base version of [anima], or even 1.1, is plenty satisfying, but I guess it's just habit or maybe my hipster tendencies I couldn't just sit still.I've upgraded my GPU and such a great model as [anima] has come out, I really wanted to tune something myself. But gathering a dataset takes more time than expected, and I'm pretty lazy. :(So, once again, I couldn't resist and ended up merged one anyway...My approach hasn't changed much since the old XL days. Starting from the base, I make it a bit more stable and just slightly better. So I want to mention upfront that this model's knowledge level doesn't stray too far from the base [anima] model.<ul><li>Usage</li></ul>Honestly, for the most part, you can just follow the same settings and prompts used for the basic [anima] model. No need to go into a long explanation here.That said, here are a few recommended settings:<ul><li>Sampler: Euler A CFG++</li><li>Scheduler: AYS</li><li>CFG: 2–3.5</li><li>Steps: 16–28</li></ul><h3 id="thanks-to-circlestone_labs-(tdrussell)-for-creating-such-a-wonderful-model.">Thanks to circlestone_labs (tdrussell) for creating such a wonderful model. </h3>

Anitional

This version is a port of v1.0 to the 2.9B model, quantized to int8c.

Files with SHA256: 0b28700d30d3ae1e61f4cda11cbe39e4837af8f0deda4c86e10416718a9a6959