<h2 id="boho-festival-lora">Boho Festival LoRA</h2><h3 id="overview">Overview</h3><p>This training job focused on creating a <strong>LoRA extension</strong> to capture the visual and stylistic characteristics of a <em>boho festival</em> concept using the <strong>Z-Image De-Turbo</strong> diffusion architecture.</p><h3 id="model-and-architecture">Model &amp; Architecture</h3><ul><li><p><strong>Base Model:</strong> <code>ostris/Z-Image-De-Turbo</code></p></li><li><p><strong>Architecture:</strong> <code>zimage:deturbo</code></p></li><li><p><strong>Quantization:</strong> qfloat8 (model and text encoder)</p></li><li><p><strong>Precision:</strong> bf16</p></li></ul><h3 id="training-configuration">Training Configuration</h3><ul><li><p><strong>Training Steps:</strong> 3000</p></li><li><p><strong>Optimizer:</strong> AdamW 8-bit</p></li></ul><p>The training emphasized a <strong>balanced content-to-style objective</strong>, suitable for flexible prompt usage.</p><h3 id="trigger-words">Trigger Words</h3><ul><li><p><strong>Activation Tokens:</strong> <code>boho; festival</code></p></li></ul><h3 id="summary">Summary</h3><p>This training successfully produced a lightweight LoRA extension capable of injecting <strong>bohemian festival aesthetics</strong> into Z-Image Turbo generations. The configuration balances quality, flexibility, and VRAM efficiency, making the resulting model suitable for creative image generation workflows and further fine-tuning if required.</p>

Files with SHA256: b678063383f2eefbbf6b336dd4fb3740a976a94679052754a26e4e5d8fc3cb58