<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/79695e9b-92cb-4c1d-b699-038135b972b2/width=525/79695e9b-92cb-4c1d-b699-038135b972b2.jpeg" />Check out our website to see what we’ve been working on and explore our latest model! → <a target="_blank" rel="ugc" href="https://www.illustrious-xl.ai/">https://www.illustrious-xl.ai/</a><h1 id="illustrious-lumina-v0.03-eyvt0yjxg">Illustrious-Lumina-v0.03</h1>This model is based on Alpha-VLLM/Lumina-Image-2.0 , which is nice small DiT model with minimal guaranteed functionality! Please refer to <a target="_blank" rel="ugc" href="https://github.com/Alpha-VLLM/Lumina-Image-2.0">https://github.com/Alpha-VLLM/Lumina-Image-2.0</a> for official repository. <a target="_blank" rel="ugc" href="https://arxiv.org/abs/2503.21758">Paper</a><hr />Before we dive into the details of 'Illustrious-Lumina-v0.03', we’re excited to share that you can now generate images directly with our Illustrious XL models on our official site: <a target="_blank" rel="ugc" href="http://illustrious-xl.ai">illustrious-xl.ai</a>.We’ve launched a full image generation platform featuring high-res outputs, natural language prompting, and custom presets - plus, several exclusive models you won’t find on any other hub.Explore our updated model tiers and naming here: <a target="_blank" rel="ugc" href="https://www.illustrious-xl.ai/updates/20">Model Series</a>.Need help getting started? Check out our generation user guide: <a target="_blank" rel="ugc" href="https://www.illustrious-xl.ai/updates/21">ILXL Image Generation User Guide</a>.<hr /><a target="_blank" rel="ugc" href="https://cdn-uploads.huggingface.co/production/uploads/63398de08f27255b6b50081a/OamvrbyYicsGvp2ShVoaq.png"><img src="https://cdn-uploads.huggingface.co/production/uploads/63398de08f27255b6b50081a/OamvrbyYicsGvp2ShVoaq.png" alt="image/png" /></a><h2 id="1.-model-overview-xalynbxys">1. Model Overview</h2><ul><li>Architecture: 2 B parameters DiT.</li><li>Text Encoder: Pure LLM, Gemma-2-2b</li><li>Goal of this fork: We test if the image backbone can learn illustration concepts without re‑training the LLM component.</li></ul><hr />Illustrious-Lumina-v0.03 is experimental epoch of Lumina-2.0 based training session, to validate whether we would be able to achieve small DiT model just with LLM - to be trained as illustration-focused model. The original model, is unfortunately bad at illustrations and lacked any of the knowledge - so the run focused on training abscent knowledges.After 26,500 step, the model, Illustrious-Lumina-v0.03 has show successful fast adaptation toward the dataset.However, please note that the original model is not good at illustrations, whileas our focus is only in illustrations - this would take a while to reach the certain level.The examples are ready in <a target="_blank" rel="ugc" href="https://www.illustrious-xl.ai/blog">Blog post</a>.To test the model, please refer to the <a target="_blank" rel="ugc" href="https://huggingface.co/spaces/AngelBottomless/Lumina-Illustrious-v0.03">huggingface space</a>If you prefer to run model locally, please use the pth file with <a target="_blank" rel="ugc" href="https://github.com/OnomaAI/Illustrious-Lumina">official installation guide</a>. The safetensors file is meant to only "contain the weights" - for comfyui-compatible format, we will try to prepare it as soon as possible.<h2 id="2.-training-setup-dw2dxtqjf">2. Training Setup</h2>ItemValueImages Seen Total22 M image–text pairsSteps26  500Global batch768Resolution1024, 256Checkpoint<code>Illustrious_Lumina_2b_22100_ema_unified_fp32.safetensors</code>The model has seen 22M image-text pairs. To accelerate the training, multi-resolution training was utilized.<h2 id="3.-inference-demo-code-o9x0nmyt8">3. Inference Demo Code</h2>If you prefer to run model locally, please use the pth file with <a target="_blank" rel="ugc" href="https://github.com/OnomaAI/Illustrious-Lumina">official installation guide</a>.The setup used for header image can be replicated with following setup:<a target="_blank" rel="ugc" href="https://cdn-uploads.huggingface.co/production/uploads/63398de08f27255b6b50081a/qmFDOkCiu-gjG4X0r2ydO.png"><img src="https://cdn-uploads.huggingface.co/production/uploads/63398de08f27255b6b50081a/qmFDOkCiu-gjG4X0r2ydO.png" alt="image/png" /></a><h2 id="4.-disclaimer-4ehz8vh6c">4. Disclaimer</h2>The model does not reflect any final product, and intended to be used for research analysis only. The model is not production-ready; use as own risk.The model is in Proof Of Concept stage- supposedly, 3% of the compute required for full training, with only 22M samples seen with low-resolution joint training, with A6000 GPUs.For training acceleration, please consider supporting us in <a target="_blank" rel="ugc" href="https://illustrious-xl.ai/model/17">Support site</a>!

Files with SHA256: dc6cffcfb0ccfca6332ddb5d2fe25bcb5f496f44b481627f48c42626156fa6a8