Understand the project
Practical overview
When it fits
Use Llamafactory when an existing supported language or vision-language model needs task- or domain-specific adaptation and the work can be managed in a self-hosted Python workflow.
Who it is for
- Python developers building model-customization workflows
- Machine-learning engineers adapting language or multimodal models
- Research teams comparing fine-tuning methods and configurations
- Organizations working with domain-specific instruction or image-and-text data
Documented uses
- Fine-tune supported language models with domain-specific instruction data
- Adapt vision-language models for specialized multimodal tasks
- Apply parameter-efficient LoRA or QLoRA customization
- Compare multiple model-training approaches within one toolkit
Before you choose
Requirements & limits
The project is available as a public Apache-2.0 repository. The repository documents local, CLI, web-interface, Docker and selected cloud-notebook paths; specific environment requirements should be checked in its documentation.
- Model coverage: the repository lists more than 100 supported LLMs and VLMs.
- Training choice: workflows include full, freeze, LoRA and several quantized QLoRA approaches.
- Method breadth: the toolkit lists supervised fine-tuning, reward modeling, PPO, DPO, KTO and ORPO.
- Interface preference: users can follow CLI workflows or use the Gradio-powered LLaMA Board.
- Environment planning: the project provides separate guidance for selected AMD GPU and Ascend NPU environments.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Evidence and transparency
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 2 October 2026
- Original website title
- hiyouga/LlamaFactory · Official homepage title · retained in its source language