Implementation:Huggingface Peft C3AModel

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

Overview

Concrete tool for applying C3A (Column-Column-Column Adaptation) to pretrained transformer models, provided by the Huggingface PEFT library.

Description

C3AModel is a tuner class that creates a C3A model from a pretrained transformers model. It injects block-structured adapter layers into target Linear modules, where the block size can be configured per module via pattern matching. The method is described in detail in https://huggingface.co/papers/2407.19342.

Usage

C3AModel is typically created internally by calling get_peft_model with a C3AConfig. It can also be instantiated directly by passing a base model, a C3AConfig, and an adapter name.

Code Reference

Source Location

• Repository: Huggingface_Peft
• File: src/peft/tuners/c3a/model.py
• Lines: 29-102

Signature

class C3AModel(BaseTuner):
    prefix: str = "c3a_"
    # Inherits __init__ from BaseTuner:
    # def __init__(self, model, config, adapter_name):
    #     ...

Import

from peft.tuners.c3a import C3AModel

I/O Contract

Inputs

Name	Type	Required	Description
model	nn.Module	Yes	The pretrained model to adapt
config	C3AConfig	Yes	Configuration for the C3A adapter (block_size, block_size_pattern, init_weights)
adapter_name	str	Yes	Name identifier for the adapter, defaults to "default"

Outputs

Name	Type	Description
adapted_model	C3AModel	Model with C3A adapter layers injected into target Linear modules

Usage Examples

Basic Usage

from peft import get_peft_model, C3AConfig
from transformers import AutoModelForCausalLM

model = AutoModelForCausalLM.from_pretrained("model-name")
config = C3AConfig(
    block_size=16,
    target_modules=["q_proj", "v_proj"],
)
model = get_peft_model(model, config)

Related Pages

• Environment:Huggingface_Peft_Python_Core_Dependencies