LlamaFactory

LlamaFactory: This open-source framework provides a unified interface to fine-tune and deploy over 100 large language models and multimodal models.

LlamaFactory offers a unified platform for fine-tuning and deploying large language models. It supports a wide variety of model architectures including LLaMA, Mistral, Qwen, and DeepSeek, as well as multimodal models like LLaVA. Users can perform tasks ranging from supervised fine-tuning and pre-training to reinforcement learning using PPO, DPO, and KTO. The tool integrates with various experiment tracking systems such as TensorBoard, Wandb, and MLflow for monitoring. It facilitates faster inference through support for vLLM and SGLang backends, alongside providing an OpenAI-style API for deployment.

LlamaFactory at a glance

Best for: Software engineer, Researcher
Runs on: CLI, Web, self-hosted
Pricing: Free
Released: May 2023
Category: AI coding tool
Website: github.com/hiyouga/LlamaFactory

Key features

Zero-code CLI and Web UI for model training
Support for 100+ LLMs and multimodal models
Integration with vLLM and SGLang for accelerated inference
Experiment tracking with TensorBoard, Wandb, MLflow, and SwanLab
Advanced training techniques including LoRA, QLoRA, and DoRA
Multi-backend support for GPU and NPU hardware

Use it for

Fine-tuning large language models for domain-specific knowledge
Training multimodal models for image and video recognition
Deploying fine-tuned models via OpenAI-style APIs
Executing reinforcement learning with GRPO or PPO

Alternatives to LlamaFactory

AirLLM — Runs massive 70B+ parameter models on small 4GB consumer GPUs.
LMDeploy — Accelerates inference with high-performance engines and quantization tools.
Soup — Fine-tune and align models like Llama-3.1-8B on a 4 GB laptop GPU.
llama-swap — It runs multiple local AI models with zero external dependencies.
LMFlow — It offers memory-efficient training and full Accelerate support for large language models.
Llama Windows — It runs open models offline with zero configuration.
Distributed Llama — Splits model memory and computation across multiple connected home devices.
Axolotl — It supports many training methods and model architectures via simple YAML configs.
LLamaSharp — Runs large language models efficiently on local hardware using C#.
XTuner — High-efficiency training for 200B+ scale MoE models without expert parallelism.
LlamaCoder — Generates functional applications directly from text descriptions.
Ludwig — Train and fine-tune complex AI models using a single YAML configuration file.

All LlamaFactory alternatives