LlamaFactory

Why we included this project

Fine-tuning a modern language model used to mean pulling together trainers and dataset loaders from several different libraries. LLaMA Factory folds that work into one tested workflow that covers more than a hundred open-weight models and vision-language variants, with LoRA, QLoRA, and full-parameter training available from the same interface. Teams adapting a base model to their own data, whether for instruction tuning, preference alignment, or continued pretraining, can go from raw data to a tuned checkpoint without rebuilding the plumbing each time. It ships a web UI, a command-line tool, and a Python API, so it suits both researchers poking at experiments interactively and engineers who want to drop fine-tuning into an automated pipeline. For anyone weighing how to productionize model customization, it is a solid reference for what a mature, community-tested fine-tuning stack looks like.