Implementation:Huggingface Peft BoneConfig

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

Overview

Concrete configuration dataclass for the Bone PEFT adapter method provided by the Huggingface PEFT library.

Description

BoneConfig controls the hyperparameters for the Bone adapter, a parameter-efficient fine-tuning method that applies low-rank updates to selected model layers. It supports two initialization variants: the default Bone structure and the "bat" variant. This class extends PeftConfig. Note: Bone is deprecated and will be removed in PEFT v0.19.0 in favor of MissConfig.

Usage

Import and use BoneConfig when you want to apply the Bone adapter method for parameter-efficient fine-tuning, particularly for scenarios requiring low-rank weight updates with configurable rank and layer targeting. Consider using MissConfig instead for new projects.

Code Reference

Source Location

• Repository: Huggingface_Peft
• File: src/peft/tuners/bone/config.py
• Lines: 25-130

Signature

@dataclass
class BoneConfig(PeftConfig):
    r: int = 64
    target_modules: Optional[Union[list[str], str]] = None
    exclude_modules: Optional[Union[list[str], str]] = None
    init_weights: bool | Literal["bat"] = True
    layers_to_transform: Optional[Union[list[int], int]] = None
    layers_pattern: Optional[str] = None
    bias: str = "none"
    modules_to_save: Optional[list[str]] = None

Import

from peft import BoneConfig

I/O Contract

Inputs

Name	Type	Default	Description
r	int	64	Rank of the Bone adapter. Best set to an even number for default initialization.
target_modules	Optional[Union[list[str], str]]	None	Module names or regex to apply adapter to.
exclude_modules	Optional[Union[list[str], str]]	None	Module names or regex to exclude from adapter.
init_weights	bool or Literal["bat"]	True	Initialization variant: True for Bone structure, "bat" for Bat structure.
layers_to_transform	Optional[Union[list[int], int]]	None	Specific layer indexes to apply transformations to.
layers_pattern	Optional[str]	None	Layer pattern name when layers_to_transform is set.
bias	str	"none"	Bias type. Can be 'none', 'all', or 'bone_only'.
modules_to_save	Optional[list[str]]	None	Additional modules to set as trainable and save in the final checkpoint.

Outputs

Name	Type	Description
instance	BoneConfig	Configuration object to pass to get_peft_model

Usage Examples

Basic Configuration

from peft import BoneConfig, get_peft_model
from transformers import AutoModelForCausalLM

model = AutoModelForCausalLM.from_pretrained("model-name")
config = BoneConfig(
    r=64,
    target_modules=["q_proj", "v_proj"],
)
model = get_peft_model(model, config)

Related Pages

• Environment:Huggingface_Peft_Python_Core_Dependencies
• Heuristic:Huggingface_Peft_Warning_Deprecated_Bone