LLaMA Factory

Source

It supports more than 100 LLMs, with both a command-line interface and a Web UI for zero-code workflows. Unlike many frameworks that target expert users, LLaMA Factory is designed to be beginner-friendly. Through its web interface, users can select a model, upload a dataset, adjust a few parameters, and launch training, with no coding required.But it’s not just for beginners. Behind the scenes, LLaMA Factory supports a wide range of tuning methods, making it equally useful for experienced researchers and developers.

Key features:

Fine-tuning methods: supervised SFT (including multimodal), reward modeling, and reinforcement learning (PPO, DPO, ORPO, KTO).
Quantization and adapter support: 16-bit full-tuning, freeze-tuning, LoRA, and 2–8 bit QLoRA via formats like GPTQ, AWQ, HQQ, and AQLM.
Advanced optimization: GaLore, DoRA, LongLoRA, LoftQ, LLaMA Pro, Mixture-of-Depths, and more.

Tags: ai llm model fine-tune

Last modified 07 May 2026

LLaMA Factory

An open-source fine-tuning toolkit built for simplicity and efficiency.