LoRA Fine-Tuning Training Recipes & Colab Notebooks
Turnkey PyTorch and Kohya_ss training configurations with loss convergence monitors and captioning pipelines.
Production recipes for fine-tuning Flux, SDXL, and Llama 3 models on custom character identities, architectural styles, and brand typography. Includes dataset curation scripts, Danbooru tag pruners, rank-32 / alpha-16 optimal learning curves, and step-by-step Jupyter notebooks.