Implementation:Intel Ipex llm Axolotl Finetuning

Knowledge Sources	Intel IPEX-LLM Axolotl
Domains	Finetuning, Axolotl, Integration
Last Updated	2026-02-09 04:00 GMT

Overview

Concrete tool for integrating the Axolotl fine-tuning framework with IPEX-LLM optimizations via the llm_patch mechanism.

Description

The do_cli() function wraps the Axolotl training framework with IPEX-LLM's llm_patch to enable XPU-optimized fine-tuning. It applies the patch at import time before delegating to Axolotl's standard training, inference, LoRA merging, or dataset sharding operations. This allows users to leverage Axolotl's rich configuration system (YAML-based) while benefiting from IPEX-LLM optimizations on Intel hardware.

Usage

Use this when wanting to use Axolotl's feature-rich fine-tuning framework (supporting multiple training methods, dataset formats, and configurations) on Intel XPU hardware with IPEX-LLM optimizations.

Code Reference

Source Location

• Repository: Intel IPEX-LLM
• File: python/llm/example/GPU/LLM-Finetuning/axolotl/finetune.py
• Lines: 1-86

Signature

def do_cli(config: Path = Path("examples/"), **kwargs):
    """Wrapper for Axolotl CLI with IPEX-LLM patches."""

Import

from ipex_llm import llm_patch
from axolotl.cli import load_cfg, load_datasets
from axolotl.train import train

I/O Contract

Inputs

Name	Type	Required	Description
config	Path	Yes	Path to Axolotl YAML configuration file
--inference	flag	No	Run inference mode
--merge_lora	flag	No	Merge LoRA adapters
--shard	flag	No	Shard dataset

Outputs

Name	Type	Description
Fine-tuned model	Files	Saved per Axolotl config output_dir
Merged model	Files	When --merge_lora is used

Usage Examples

Axolotl Training with IPEX-LLM

python finetune.py qlora.yml

LoRA Merge

python finetune.py qlora.yml --merge_lora

Related Pages

• Environment:Intel_Ipex_llm_XPU_Finetuning_Environment