Axolotl is an open-source Python-based framework for fine-tuning and post-training of LLMs, supporting LoRA, QLoRA, GPTQ, and various optimization features. It provides YAML-based configuration, Docker and cloud deployment options, and Colab examples.
Collecting history — the radar snapshots this repo daily. The trend line appears after 3 days of data (1 so far).
What it is
Axolotl is a free and open-source tool designed to streamline post-training and fine-tuning for the latest large language models (LLMs).
How it works
The project offers multiple model support (GPT-OSS, LLaMA, Mistral, Mixtral, Pythia, etc.), multimodal training options, and a range of training methods (Full fine-tuning, LoRA, QLoRA, GPTQ, QAT, FP8, NVFP4/MXFP4 MoE LoRA, RL/DPO/IPO/KTO/ORPO, RM/PRM). It emphasizes using a single YAML configuration to cover dataset preprocessing, training, evaluation, quantization, and inference. Performance features include multipack, Flash Attention variants, Xformers, Flex Attention, SageAttention, Liger kernels, ScatterMoE, Sequence Parallelism, LoRA optimizations, and multi-GPU/multi-node training.
Getting started
Installation
# install uv if you don't already have it installed (restart shell after)
curl -LsSf https://astral.sh/uv/install.sh | sh
# change depending on system
export UV_TORCH_BACKEND=cu130
# create a new virtual environment
uv venv --python 3.12
source .venv/bin/activate
uv pip install torch==2.12.0 torchvision
uv pip install --no-build-isolation axolotl[deepspeed]
# Download example axolotl configs, deepspeed configs
axolotl fetch examples
axolotl fetch deepspeed_configs # OPTIONAL
Using Docker
docker run --gpus '"all"' --ipc=host --rm -it axolotlai/axolotl:main-latest
Your First Fine-tune
# Fetch axolotl examples
axolotl fetch examples
# Or, specify a custom path
axolotl fetch examples --dest path/to/folder
# Train a model using LoRA
axolotl train examples/llama-3/lora-1b.yml
Cloud Providers
Details list several providers for running clusters (RunPod, Vast.ai, PRIME Intellect, Modal, etc.).
References to documentation
Links are provided for installation options, configuration guide, dataset loading, multi-GPU/multi-node training, and other topics.
Recent releases
Latest releases include:
- v0.18.0 (2026-07-17): Axolotl v0.18.0 Release Notes with low-level kernel improvements and MoE fine-tuning enhancements
- v0.17.0 (2026-06-03): Axolotl v0.17.0 Release Notes covering Expert Parallelism for MoE training and NVFP4 optimizations
- v0.16.1 (2026-04-02): Axolotl v0.16.1 Release Notes including Gemma 4 Support
- v0.16.0 (2026-04-02): Axolotl v0.16.0 Release Notes with ~80 new commits since v0.15.0
- v0.15.0 (2026-03-06): Axolotl v0.15.0 Release Notes introducing new model support and MoE improvements
Traction
Stars: 12305; Forks: 1398; Open issues: 277
基础信息
Axolotl is a Python-based, open-source LLM fine-tuning framework. License: Apache-2.0. Created 2023-04-14; last push 2026-08-04. Topics include fine-tuning and llm.