Elke, a fictional character.Trained on 30 Klein 9b generated images, from an initial JIB mix qwen generated image.<h3 id="moss-image">moss image</h3>Trained with ai-toolkit, lokr factor 4, automagic v3 LR 0.0005, weight decay 0.00001, Sigmoid/High Noise. After 3k steps the likeness was in the 70% range, with poor reproduction of tattoo. Trained another 2,400 steps with layer 22 not skipped, Shift/Balanced (3,000+2,400 steps outperformed other checkpoints, evaluated up to a total of 6,000 steps).Training config is in Other Files as <span style="background-color:rgb(26, 27, 30);color:rgb(141, 141, 143);font-family:-apple-system, &quot;system-ui&quot;, &quot;Segoe UI&quot;, Roboto, Helvetica, Arial, sans-serif, &quot;Apple Color Emoji&quot;, &quot;Segoe UI Emoji&quot;;font-size:14px">config-moss-state-1.yml and <span style="background-color:rgb(26, 27, 30);color:rgb(141, 141, 143);font-family:-apple-system, &quot;system-ui&quot;, &quot;Segoe UI&quot;, Roboto, Helvetica, Arial, sans-serif, &quot;Apple Color Emoji&quot;, &quot;Segoe UI Emoji&quot;;font-size:14px">config-moss-state-2.yml<h3 id="minimax-h3">minimax h3</h3>Trained with ai-toolkit, automagic v3 with Contrastive Guidance and training adapter v1; differential guidance scale = 1; rank 24. ai-toolkit configuration attached as minimaxh3-elke-v01.yaml<h3 id="krea-2">krea 2</h3>Trained with diffusion-pipe, ProdigyPlusScheduleFree optimizer (<a target="_blank" rel="ugc" href="https://github.com/tdrussell/diffusion-pipe/pull/483">https://github.com/tdrussell/diffusion-pipe/pull/483</a>), lokr factor 4, masked loss<h3 id="ideogram-4">ideogram 4</h3>v02Trained with diffusion-pipe, ProdigyPlusScheduleFree optimizer (<a target="_blank" rel="ugc" href="https://github.com/tdrussell/diffusion-pipe/pull/483">https://github.com/tdrussell/diffusion-pipe/pull/483</a>), rank 128, using plain text (non-JSON) captions. <a target="_blank" rel="ugc" href="http://config-ideogram4-v02.zip"><span style="background-color:rgb(26, 27, 30);color:rgb(193, 194, 197);font-family:-apple-system, &quot;system-ui&quot;, &quot;Segoe UI&quot;, Roboto, Helvetica, Arial, sans-serif, &quot;Apple Color Emoji&quot;, &quot;Segoe UI Emoji&quot;;font-size:14px">config-ideogram4-v02.zip</a><span style="background-color:rgb(26, 27, 30);color:rgb(193, 194, 197);font-family:-apple-system, &quot;system-ui&quot;, &quot;Segoe UI&quot;, Roboto, Helvetica, Arial, sans-serif, &quot;Apple Color Emoji&quot;, &quot;Segoe UI Emoji&quot;;font-size:14px"> contains training images, mask images, captions, validation images, and diffusion-pipe configuration used<span style="background-color:rgb(26, 27, 30);color:rgb(193, 194, 197);font-family:-apple-system, &quot;system-ui&quot;, &quot;Segoe UI&quot;, Roboto, Helvetica, Arial, sans-serif, &quot;Apple Color Emoji&quot;, &quot;Segoe UI Emoji&quot;;font-size:14px">.v01Trained with ai-toolkit, automagic v3 optimizer, LR 0.00001, weight decay 0.00001, rank 32, with plain text (not JSON) captions. Likeness is slightly higher when using plain text during inference instead of JSON, but JSON still works. The <a target="_blank" rel="ugc" href="http://config-ideogram4-v01.zip">config-ideogram4-v01.zip</a> attachment contains the training images, masks, captions, validation images, and the ai-toolkit configuration.<h3 id="qwen-2512">qwen 2512</h3>Trained with diffusion-pipe, this is a 128-rank lora trained on qwen image 2512, using the ProdigyPlusScheduleFree optimizer (<a target="_blank" rel="ugc" href="https://github.com/tdrussell/diffusion-pipe/pull/483">https://github.com/tdrussell/diffusion-pipe/pull/483</a>), on 30 1440x1440 images, using masked training with backgrounds masked to 10%The model was evaluated every 300 steps (10 repeats of each training image) using FaceNet512 to evaluate the face similarity, using the same settings, seed, etc for inference, calculating the average distance to a subset of 10 of the training images.The attached training data zip includes: the diffusion pipe config files; the training images, prompts, and masks; the validation images (subset of training images); and the output of <a target="_blank" rel="ugc" href="https://civarchive.com/articles/26241/stop-training-on-vibes-a-visual-guide-to-biometric-lora-diagnosis-mirror-metrics">MirrorMetrics</a> evaluation of epoch15 vs. the validation set.<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/4b8d675e-1704-4065-a57e-47f147937597/original=true/4b8d675e-1704-4065-a57e-47f147937597.jpeg" />

Elke

ai-toolkit, lokr 4, automagic v3 LR 0.0005 weight decay 0.00001 first 3,000 steps w/sigmoid+high noise bias, 2,400 more steps w/shift+balanced

config-moss-stage-1.yml

config-moss-stage-2.yml

moss-elke-v03_000005400.safetensors

moss-v03

minimaxh3-v01

zit-sickollie-v01

krea2-v01e05

ideogram4-v01e20

qwen-image-2512-v01e15

Files with SHA256: 499c850309ddfcc1fed2a30c3bbf2a83fea4bd083bfd25a0c6dea48311350ad2