LLaMA-Factory
github.com/hiyouga/LLaMA-Factory
A unified fine-tuning framework supporting 100+ language models with a web UI and CLI, covering SFT, RLHF, DPO, and other training methods with efficient LoRA and QLoRA support.
Why it is useful
Lowers the barrier to fine-tuning significantly — the web UI lets you configure training runs, monitor loss curves, and chat with the resulting model without writing any code. Supports virtually every open-weight model family and training method, so you can switch from Llama to Qwen or from SFT to DPO without changing frameworks. Good starting point before investing in a more custom training pipeline.