Source: Axolotl Documentation

Documentation for Axolotl, a YAML-configured open-source framework for LLM fine-tuning and post-training.

Contents
  • What: library documentation
  • Author: Axolotl AI
  • Date: read 2026-11-14 (latest update listed 2026-10)
  • Original: Axolotl docs

Summary

Axolotl drives dataset preprocessing, training, evaluation, quantization and inference from a single YAML file, aimed at reproducible multi-GPU runs.

Key claims

  • Methods: full fine-tuning, LoRA, QLoRA, QAT, preference tuning (DPO, IPO, KTO, ORPO), RL (GRPO) and reward modelling.
  • Distributed: FSDP2, DeepSpeed, multi-node via Torchrun and Ray, context and expert parallelism.
  • Performance: multipacking (sample packing), Flash Attention, Liger kernels.
  • Supports many Hub model families (Llama, Mistral, Qwen, Gemma, GPT-OSS, Phi and others); multimodal in beta.

Feeds

Fine-tuning · PEFT and LoRA

Suggest changes →