<h1 id="illustriousnoobai-dooby3d-q9shgo2j8">[Illustrious/NoobAI] dooby3D</h1><h2 id="disclaimer-6sofm8zgy">Disclaimer</h2><ul><li>Do not use my LoRA to produce AI image and tagging her fanart hashtag in Twitter/X or if you have an enough contribution in that image then fine.</li><li>NO REPOST IN ANY WEBSITE (PixAI, Tensor . Art, Seaart, ...) !!!!!!!!</li></ul><h2 id="description-v1-yolxtp0a4">Description V1</h2><ul><li>Detective Dooby.</li><li>can change costume</li></ul><h2 id="description-wzpqy9ixt">Description</h2><ul><li>Test with LoRA GGPO (test version)</li><li>First NoobAI v-pred version LoRA / with illustrious too</li></ul><h2 id="trigger-word-bzk0oajq1">trigger word</h2>with white shirt<pre><code>dooby3d, 1girl, solo, mouse girl, blonde hair, black hair, white hair, multicolored hair, green eyes, tail, short hair, gloves, checkered gloves, blue overalls, goggles on headwear, shirt, short eyebrows, hair ornament, blue hat, green neckerchief, medium breast</code></pre>without white shirt (accidentally remove)<pre><code>dooby3d, 1girl, solo, mouse girl, blonde hair, black hair, white hair, multicolored hair, green eyes, tail, short hair, gloves, checkered gloves, blue overalls, goggles on headwear, bare shoulders, collarbones, bare arms, checkered tube top, short eyebrows, hair ornament, blue hat, medium breast</code></pre>any costume<pre><code>dooby3d, 1girl, solo, mouse girl, blonde hair, black hair, white hair, multicolored hair, green eyes, tail, short hair, short eyebrows, hair ornament, medium breast</code></pre><h3 id="limitations-wk4s0b3jw">Limitations</h3><ul><li>no [v1]</li></ul><ul><li>only 1 costume [v0.1]</li></ul><h2 id="training-details-v1-illustrious-gfeun0rqf">Training Details [V1 Illustrious]</h2>LoRA type<ul><li>LoRA GGPO</li></ul>dataset<ul><li>75 images</li></ul>parameters<ul><li>resolution = 768</li><li>batch size = 2</li><li>network dim,alpha = 16,8</li><li>mix/save precision = bf16/bf16</li><li>optmizer = AdamWSchedulefree</li><li>UNet LR = 1.75e-4</li><li>TE LR = 8.75e-5</li><li>scheduler = no</li><li>huber loss</li></ul>steps<ul><li>epochs = 10</li><li>total steps = 3000</li><li>repeat = 8</li></ul>tools<ul><li>sd-scripts</li><li>torch 2.7.0 cu128</li><li>RTX 3060 12 GB</li></ul>avg weight<ul><li>UNet average weights : 0.00988162469796159</li><li>TE1 average weights : 0.008170449989847839</li></ul><h2 id="training-details-v0.1-noobai-5enbgow2j">Training Details [v0.1 NoobAI]</h2>LoRA type<ul><li>LoRA GGPO</li></ul>dataset<ul><li>70 images</li></ul>parameters<ul><li>resolution = 1024</li><li>batch size = 2</li><li>network dim,alpha = 12,12</li><li>mix/save precision = bf16/bf16</li><li>optmizer = CAME + weight_decay=0.01</li><li>UNet LR = 4.5e-5</li><li>TE LR = 2e-5</li><li>scheduler = cosine with restarts (3 times)</li><li>l2 loss</li><li>extra args</li></ul><pre><code>--v_parameterization --zero_terminal_snr</code></pre>steps<ul><li>epochs = 10</li><li>total steps = 1750</li><li>repeat = 5</li></ul>tools<ul><li>sd-scripts custom version by me</li><li>torch 2.5.0 cu124</li><li>RTX 3060 12 GB</li></ul>avg weight<ul><li>UNet average weights : 0.00935241486877203</li><li>TE1 average weights : 0.008072614941435555</li></ul><h2 id="training-details-v0.1-illustrious-5b6nycff5">Training Details [v0.1 Illustrious]</h2>LoRA type<ul><li>LoRA GGPO</li></ul>dataset<ul><li>70 images</li></ul>parameters<ul><li>resolution = 1024</li><li>batch size = 2</li><li>network dim,alpha = 12,12</li><li>mix/save precision = bf16/bf16</li><li>optmizer = CAME + weight_decay=0.01</li><li>UNet LR = 4.5e-5</li><li>TE LR = 2e-5</li><li>scheduler = cosine with restarts (3 times)</li><li>l2 loss</li></ul>steps<ul><li>epochs = 10</li><li>total steps = 1750</li><li>repeat = 5</li></ul>tools<ul><li>sd-scripts custom version by me</li><li>torch 2.5.0 cu124</li><li>RTX 3060 12 GB</li></ul>avg weight<ul><li>UNet average weights : 0.009963956746188078</li><li>TE1 average weights : 0.008257410062166551</li></ul>*This LoRA is for studying LoRA training with new technique so do not use for damaging the character (also support her too).*If you see my model reposted on other image generation service please report to me.

[Illustrious/NoobAI] dooby3D

<ul><li>U-Net + CLIP-L</li><li>debiased_estimation_loss</li><li>block weights LR</li><li>edit dataset</li></ul>

Files with SHA256: 5d803e5ae9bedda88cdb2c3b9222a0824e1f76d8a5f29a62ea817329377c5ff7