LlamaFactory
Unified Efficient Fine-Tuning of 100+ LLMs & VLMs (ACL 2024)
What is LlamaFactory?
Unified Efficient Fine-Tuning of 100+ LLMs & VLMs (ACL 2024)
Overview of LlamaFactory
What it is
LlamaFactory (LLaMA-Factory) is an open-source project for unified, efficient fine-tuning of 100+ large language models and vision-language models; its metadata describes it as an ACL 2024 work. It offers a zero-code CLI and Web UI for fine-tuning, and the README states it is used by Amazon, NVIDIA, Aliyun, etc. Documentation is published on Read the Docs, and the project also maintains an official blog.
Key features
- Model coverage: LLaMA, LLaVA, Mistral, Mixtral-MoE, Qwen3, Qwen3-VL, DeepSeek, Gemma, GLM, Phi and others.
- Integrated methods: (continuous) pre-training, (multimodal) supervised fine-tuning, reward modeling, PPO, DPO, KTO, ORPO.
- Scalable resources: 16-bit full-tuning, freeze-tuning, LoRA, and 2/3/4/5/6/8-bit QLoRA via AQLM/AWQ/GPTQ/LLM.int8/HQQ/EETQ.
- Advanced algorithms: GaLore, BAdam, APOLLO, Adam-mini, Muon, OFT, DoRA, LongLoRA, LLaMA Pro, Mixture-of-Depths, LoRA+, LoftQ, PiSSA.
- Practical tricks: FlashAttention-2, Unsloth, Liger Kernel, KTransformers, RoPE scaling, NEFTune, rsLoRA.
- Wide tasks: multi-turn dialogue, tool using, image understanding, visual grounding, video recognition, audio understanding.
- Monitoring and inference: LlamaBoard, TensorBoard, Wandb, MLflow, SwanLab; OpenAI-style API, Gradio UI and CLI with a vLLM worker or SGLang worker.
Use cases
The project targets fine-tuning of LLMs and VLMs either locally or in the cloud. The README lists free cloud options: Colab (free), PAI-DSW (free trial) and AMD GPU Cloud (free credits). Documented task types include multi-turn dialogue, tool use, image understanding, visual grounding, video recognition and audio understanding. The README also tracks Day-0/Day-1 support for new models such as Qwen3, Qwen2.5-VL, Gemma 3, GLM-4.1V, InternLM 3, MiniCPM-o-2.6, Llama 3, GLM-4, Mistral Small, PaliGemma2 and Llama 4. Deployment is documented with an OpenAI-style API and vLLM, and SGLang is supported as an inference backend via infer_backend: sglang.
Getting started
The README excerpt does not include installation commands; it refers readers to the usage section of the repository and to the online documentation. Cloud starting points listed in the README are Colab (free), PAI-DSW (free trial) and AMD GPU Cloud (free credits). For local installation and quick-start commands, see the repository README.
License and pricing
LlamaFactory is released under the Apache-2.0 license. The metadata classifies it as open source, and no pricing is stated.
Based on the project's public README and repository metadata.