Implementation:Huggingface Peft BoneModel

Knowledge Sources	Huggingface_Peft
Domains	Deep_Learning, Parameter_Efficient_Finetuning
Last Updated	2026-02-07 14:00 GMT

Overview

Concrete tool for applying Bone (Block Affine) adaptation to pretrained models, provided by the Huggingface PEFT library.

Description

BoneModel is a tuner class that creates a Bone adaptation model from a pretrained transformers model. It injects trainable low-rank block affine transformations into target Linear layers. The method belongs to the family of orthogonal/structured adaptation approaches and is described in https://huggingface.co/papers/2409.15371.

Usage

BoneModel is typically created internally by calling get_peft_model with a BoneConfig. It can also be instantiated directly by passing a base model, a BoneConfig, and an adapter name.

Code Reference

Source Location

• Repository: Huggingface_Peft
• File: src/peft/tuners/bone/model.py
• Lines: 24-127

Signature

class BoneModel(BaseTuner):
    prefix: str = "bone_"
    # Inherits __init__ from BaseTuner:
    # def __init__(self, model, config, adapter_name):
    #     ...

Import

from peft.tuners.bone import BoneModel

I/O Contract

Inputs

Name	Type	Required	Description
model	nn.Module	Yes	The pretrained model to adapt
config	BoneConfig	Yes	Configuration for the Bone adapter (rank r, init_weights)
adapter_name	str	Yes	Name identifier for the adapter, defaults to "default"

Outputs

Name	Type	Description
adapted_model	BoneModel	Model with Bone adapter layers injected into target Linear modules

Usage Examples

Basic Usage

from peft import get_peft_model, BoneConfig
from transformers import AutoModelForCausalLM

model = AutoModelForCausalLM.from_pretrained("model-name")
config = BoneConfig(
    r=8,
    target_modules=["k_proj", "q_proj", "v_proj", "out_proj", "fc1", "fc2"],
    init_weights=True,
)
model = get_peft_model(model, config)

Related Pages

• Environment:Huggingface_Peft_Python_Core_Dependencies
• Heuristic:Huggingface_Peft_Warning_Deprecated_Bone