<h3 id="heading-679">For previous version</h3><p>Training steps and dim : v1&gt;v2&gt;v3&gt;v4</p><p>My preference : v1&gt;v2&gt;v4&gt;v3</p><h3 id="heading-680">New version : v5n</h3><p>Set network alpha much larger, which largely effect learning rate.</p><p>Outputs result better and more stable , which could means less stylized.</p><p><span style="color:rgb(194, 189, 181)">Also perform well with simple prompts (last 2 posts for example).</span></p><h3 id="heading-684"></h3><p><strong>Without trigger words the outputs result slightly different, maybe worthwhile to try.</strong></p><p>This time I'm pretty confident about the quality of dataset, so there won't be more version which caused by dataset change.</p><p>I really want to use AOM3 more but nsfw isn't allowed in my workspace (sad)</p>

Kyoukai 羌瘣

<p>steps : 3000*2 </p><p>network dim : 64</p><p>network alpha : 32</p>

Files with SHA256: 6f2461ee8c1c6b1a42ded8b95d4bbfe610e3df276df1dbe347c957a7ca8f584e