training

Training a house-style LoRA on Flux.2 Klein

Dataset captioning, rank and step settings, and how to pick the winning checkpoint when training on your own render archive.