Axolotl
Configuration-driven fine-tuning runner for many model architectures.
Visão geral
Axolotl runs fine-tuning from a configuration file, covering the common architectures and training methods through one interface. It solves the reproducibility problem in fine-tuning, where a training run is a script somebody edited and the settings that produced a model are not recorded anywhere. It exists because a declared configuration can be reviewed, versioned and repeated, which a notebook cannot.
Grafo de tecnologias
Como Axolotl se liga ao resto do stack.
Aplicado a
- Fine-tuning technique Techniques Axolotl used for Fine-tuning High ✓ source
- LoRA technique Techniques Axolotl used for LoRA High ✓ source