Tools · SPBoucher
Menu

Axolotl

by Axolotl AI · Framework · also known as axolotl-ai-cloud/axolotl, OpenAccess-AI-Collective/axolotl

YAML-configured open-source framework for fine-tuning and post-training LLMs: full, LoRA and QLoRA training, preference and RL methods, multimodal models and multi-GPU or multi-node runs

Apache-2.0 licenseOpen sourceFreeSelf-hostedCloudSelf-hostable
Visit official site Compare

Link verified Nov 2, 2026

Overview

Axolotl is an Apache-2.0 fine-tuning framework in which a single YAML file describes the whole run: base model, dataset format, preprocessing, training method, evaluation, quantization and inference. It supports full fine-tuning, LoRA and QLoRA; preference tuning with DPO, IPO, KTO and ORPO; reinforcement learning with GRPO and GDPO; reward and process-reward modeling; quantization-aware training and FP8 mixed precision; and multimodal training for vision-language and audio models across many Hugging Face architectures (Llama, Mistral, Qwen, Gemma and others). Dataset formats cover pretraining, instruction, conversation, stepwise supervised, template-free and pre-tokenized data from local files, the Hub or S3, Azure, GCP and OCI storage.

Hardware support centers on NVIDIA GPUs (Ampere or newer for bf16 and FlashAttention) and AMD GPUs, with multi-GPU training through FSDP2 and DeepSpeed and multi-node through torchrun and Ray; Mac M-series use is documented but secondary. Docker images and PyPI packages are published, and documented cloud targets include RunPod, Modal, Vast.ai, SkyPilot, Hugging Face Jobs and CoreWeave. Telemetry is opt-out. Axolotl is a reproducibility-friendly choice for teams who prefer declarative configs over notebooks; it assumes familiarity with the training stack and is less memory-optimized than Unsloth on single consumer GPUs.

Capabilities

  • Declarative YAML configuration for the full training pipeline
  • Full fine-tuning, LoRA and QLoRA
  • DPO, IPO, KTO, ORPO preference tuning; GRPO and GDPO reinforcement learning
  • Reward modeling, QAT and FP8 mixed precision
  • Multimodal (vision-language and audio) training
  • Multi-GPU via FSDP2 and DeepSpeed; multi-node via torchrun and Ray
  • Docker images and documented cloud launch targets

Typical workflows

  1. Version a fine-tuning recipe as YAML alongside the dataset
  2. Run a multi-GPU DPO job on a rented cluster
  3. Train a vision-language model from a conversation dataset

Strengths

  • Reproducible, reviewable configurations
  • Broad coverage of post-training methods
  • Apache-2.0 license and active community

Limitations

  • Primarily NVIDIA- and AMD-oriented; Mac support secondary
  • Requires understanding of the underlying training stack
  • Less single-GPU memory optimization than Unsloth

Prerequisites

  • Linux with a supported GPU and CUDA or ROCm
  • Python environment or the Docker image

Who it is for

  • ML engineers running repeatable fine-tuning jobs
  • Research groups training on shared clusters

Alternatives & comparison

  • PyTorch

    Open-source deep-learning framework with eager tensors, autograd, torch.compile and distributed training, governed by the PyTorch Foundation and the base of most LLM tooling

    Open sourceOpen source
  • Together AI

    Cloud platform for open-source models: serverless and dedicated inference, fine-tuning, GPU clusters, batch jobs and code sandboxes behind an OpenAI-compatible API

    ProprietaryUsage-based
  • Transformers

    Hugging Face's model-definition library for text, vision, audio, video and multimodal models, with Pipeline, Trainer and generate APIs used as the reference by training and inference engines

    Open sourceOpen source
  • Unsloth

    Open-source framework for fine-tuning and running LLMs with reduced memory, covering LoRA, QLoRA, full fine-tuning and RL, GGUF export, Dynamic quantizations and a local Studio/Desktop UI

    Open sourceOpen source
Compare Axolotl with its alternatives →

Sources & verification

Primary sources consulted for this record. Claims are tied to the fields each source supports; the automated link checker re-verifies every URL on a schedule.

Confidence
High
Reviewed
Nov 1, 2026
Links
Verified · Nov 2, 2026

Spotted an error?

Broken link, outdated license, wrong category, or a tool that no longer exists? Tell us; reports are reviewed by the editor.

MLX

Apple Inc. · Apple Silicon ML & MLX

In-depth entry

Apple's open-source array framework for machine learning on Apple Silicon, with NumPy-like Python, C++ and Swift APIs and unified CPU/GPU memory

MITOpen sourceFreeLocalNov 2, 2026

MLX LM

Apple Inc. · Apple Silicon ML & MLX

In-depth entry

Python package and CLI for running, serving, quantizing and fine-tuning large language models on Apple Silicon with MLX

MITOpen sourceFreeLocalNov 2, 2026

Ray

Ray project (Anyscale and contributors) · Distributed & parallel computing

In-depth entry

Open-source Python framework for distributing tasks, actors and ML workloads across a cluster, with libraries for data processing, training, tuning, serving and reinforcement learning

Apache-2.0Open sourceFreeLocalSelf-hostedBroken

Hugging Face Hub

Hugging Face · Training & fine-tuning

In-depth entry

Hosting platform for open models, datasets and Spaces demo apps built on git repositories and Xet storage, with Inference Providers, Endpoints, Jobs and an open-source Python client and CLI

MixedFreemiumPaidLocalNov 2, 2026

scikit-learn

scikit-learn developers · Mathematical & statistical computing

BSD-licensed Python machine-learning library for classification, regression, clustering, dimensionality reduction, preprocessing and model selection on NumPy and SciPy

BSD-3-ClauseFreeOpen sourceLocalNov 2, 2026

Sentence Transformers

Hugging Face · Embeddings, reranking & retrieval

Python library for computing and training embedding, cross-encoder reranker, sparse and multi-vector retrieval models, maintained by Hugging Face with thousands of pretrained models on the Hub

Apache-2.0Open sourceFreeLocalSelf-hostedNov 2, 2026