Perceptual LoRA Toolkit now supports Z-Image Turbo
The Perceptual LoRA Toolkit has been updated to include support for the Z-Image Turbo model, a highly requested feature. This update introduces a new regularization technique called weight noising, which aims to reduce degradation during LoRA training and improve overall quality. The toolkit also incorporates depth anchors for further clarity and feature learning, with examples demonstrating the quality improvements achievable with these methods. AI
IMPACT Enhances LoRA training capabilities for image generation models, potentially improving user-generated content quality.