# Axolotl — Streamlined LLM Fine-Tuning

> Axolotl streamlines post-training and fine-tuning for LLMs. 11.6K+ GitHub stars. LoRA, QLoRA, DPO, GRPO, multimodal training. Single YAML config. Flash Attention, multi-GPU. Apache 2.0.

## Install

Save as a script file and run:

## Quick Use

```bash
# Install
pip3 install axolotl[flash-attn,deepspeed]

# Fetch example configs
axolotl fetch examples

# Fine-tune Llama 3 with LoRA
axolotl train examples/llama-3/lora-1b.yml

# Or use Docker
docker run --gpus all axolotlai/axolotl train examples/llama-3/lora-1b.yml
```

---

## Intro

Axolotl is a free, open-source tool for streamlined post-training and fine-tuning of large language models. With 11,600+ GitHub stars and Apache 2.0 license, it supports full fine-tuning, LoRA, QLoRA, GPTQ, DPO, GRPO, and reward modeling across GPT-OSS, LLaMA, Mistral, Mixtral, and Hugging Face models. Axolotl handles multimodal training (vision-language and audio models), Flash Attention, multipacking, sequence parallelism, and multi-GPU/multi-node training — all configured through a single YAML file.

**Best for**: ML engineers who want simple, configurable LLM fine-tuning without writing training loops
**Works with**: Claude Code, OpenAI Codex, Cursor, Gemini CLI, Windsurf
**Requirements**: NVIDIA Ampere+ or AMD GPU, Python 3.11+

---

## Key Features

- **Single YAML config**: Entire pipeline from data preprocessing to inference
- **Training methods**: Full fine-tune, LoRA, QLoRA, GPTQ, QAT, DPO, GRPO
- **Multimodal**: Fine-tune vision-language (Qwen2-VL, LLaVA) and audio models
- **Performance**: Flash Attention, multipacking, sequence parallelism
- **Multi-GPU/node**: Distributed training with DeepSpeed and FSDP
- **50+ model architectures**: LLaMA, Mistral, Mixtral, Pythia, and more

---

### FAQ

**Q: What is Axolotl?**
A: Axolotl is an LLM fine-tuning tool with 11.6K+ stars. Single YAML config for LoRA, QLoRA, DPO, GRPO, multimodal training. Multi-GPU support. Apache 2.0.

**Q: How do I install Axolotl?**
A: Run `pip3 install axolotl[flash-attn,deepspeed]`. Then `axolotl train config.yml` to start fine-tuning.

---

## Source & Thanks

> Created by [Axolotl AI](https://github.com/axolotl-ai-cloud). Licensed under Apache 2.0.
> [axolotl-ai-cloud/axolotl](https://github.com/axolotl-ai-cloud/axolotl) — 11,600+ GitHub stars

---
Source: https://tokrepo.com/en/workflows/ae1cbf21-fc40-43c7-bba7-e598af255458
Author: Script Depot