<pre><code>A lil rant from me:
For the love of god, people, use EPS DoRA with EPS models, and use VPred DoRA with a VPred model!!! 

At this point it is embarrassing seeing some of you generate the most deep fried image out there, posting it and leaving a like on the model. MY BROTHER, IT'S NOT SUPPOSED TO LOOK LIKE THAT!!! At least leave a dislike because the image looks like ass and you are ignorant enough to not understand that you're using it wrong... I swear to god, some of y'all will goon to literally anything &gt;:/

Sometimes the deepfried effect can be less noticable when you use like 8 style LoRAs at once but clearly none of the style went through then</code></pre><h3 id="vpred-models-are-trained-with-v-parameterization-so-they-will-only-work-properly-on-vpred-models-like-noob-ai-xl-vpred-r3mix-or-obsession-etc.-i3sge7wtn">VPred models are trained with V-Parameterization so they will only work properly on VPred models like NOOB-AI-XL-vpred, r3mix, or Obsession, etc.</h3>This is a DoRA so make sure to have your stuff somewhat updated in the last year<h2 id="remember-to-check-out-the-artist!!!-without-their-art-these-loras-wouldn't-exist!!!-yf9jzykvp">Remember to <a target="_blank" rel="ugc" href="https://x.com/Seraziel">check out the artist</a>!!! Without their art, these LoRAs wouldn't exist!!!</h2><h3 id="hires.fix-usage-is-heavily-advised.-don't-use-quality-tags-to-get-more-of-the-artist-style-from-the-dora-and-not-from-the-main-model.-activation-tag-is-mandatory.-4g09rsdaf">Hires.fix usage is heavily advised. Don't use quality tags to get more of the artist style from the DoRA and not from the main model. Activation tag is mandatory.</h3>Always appreciate seeing your generations tooI still don't know what I'm doing with training settings :3Now I'm busy racing horses so Krokobyaka VPred will come sometime later when I stop :3<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/bd2021a1-a79a-4cbc-8a3a-9e3d3df88f65/width=525/bd2021a1-a79a-4cbc-8a3a-9e3d3df88f65.jpeg" />

Seraziel | Artist Style | NoobAI/Illustrious

<ul><li>New dataset with 1710 images all after roughly mid 2022</li><li>Tried a different optimizer</li><li>Works pretty well, didn't have to fight as much as some other ones</li></ul><pre><code>I'm busy racing horses, so I don't feel like generating preview images for different already trained models :3</code></pre>

Files with SHA256: 76a325dd700bd26485a6362015489ec7fd6008b4a49ba77e5e78e995c138ecc3