<h1>See v1.0 for more examples.</h1><p></p><h1><strong>仅作学习目的。</strong></h1><p></p><h3><strong>DO NOT POST YOUR NSFW WORKS HERE.</strong></h3><p></p><h3><strong>Notes on usage:</strong></h3><ol><li><p>LoRA weights around 1.0 should work fine when this model is the only one applied on.</p></li><li><p>Try different combinations, such as (<strong>portrait </strong>| <strong>full body</strong>) + (<strong>from side</strong> | <strong>from behind</strong> | <strong>from above</strong>)</p></li><li><p>I'm using the DPM++2M Karras sampler for 30 steps to generate my showcases. No hires-fix is required to generate these photos.</p></li></ol><p></p><h3>What is the purpose of this model?</h3><p>This model is a test case on the two parameters <strong>network_dim </strong>and <strong>network_alpha, </strong>my previous models are all trained with network_dim=network_alpha=32, and requires lot of fine-tuning, but this one is trained with network_dim=network_alpha=128, and nothing else changed.</p><p>It's <strong>much, much more robust</strong> when generating full body photos from different angles, and more sensitive to the prompts!</p><p></p><h3>Conclusions:</h3><p>Though the file size is much larger, it does perfom a lot better.</p><p></p><h3>TODO:</h3><p>Go back to train my previous models with these params to see how it performs.</p>

Files with SHA256: d1de76c37c0f6f536e7a071c34d49c7027b6b84b5a51b5d5440140ca9f61d485