After a gap of two years, I trained using 20,000 images, and the SD lightning model.From this I can provide this nice LoRA up weights. Feel free to use.This repo is for testing.I need your responses and opinons.<a target="_blank" rel="ugc" href="https://civarchive.com/models/139562?modelVersionId=361593">RealVisXL V4.0 </a>V4.0 Lightning (BakedVAE) is usedThus, Hires fix is a must! 2.0 scale, 0.6 denoise.I may change the base model, current one have difficulty in generate a full body picture. (I suspect they only train the face...)v2.1 using the larger base version, pls notice.Plan:<ul><li>Testing Optimizing Parameters (Currently train unet only)</li><li>Change base model</li><li>How to remove water mark at preprocess? (May need help)</li><li>DPO + LoRA (Large amount of code, it takes time)</li></ul>GPU resources is still needed.

Latex (SD Lightning) Beta Repo

<a target="_blank" rel="ugc" href="https://civitai.com/models/139562/realvisxl-v50?modelVersionId=789646">https://civitai.com/models/139562/realvisxl-v50?modelVersionId=789646</a>Based on RealVisXL V5.0 Fp16 versionTraining Datasets32126 images (using yolo + inpaint remove all watermarkers)A watermark safe dataset.4 buckets<ol><li>1280 * 1024</li><li>896 * 1280</li><li>1024 * 1280</li><li>1280 * 896</li></ol>TODO:<ol><li>I think DPO is a must, i may do it later. Code is not a problem.</li><li>wd-tagger is good, I am think of retag with MLLM.</li></ol>

Files with SHA256: fb92b21f1708f18b663f1d94852d75116eeef69e11623b876284e88572b2fbed