peftfactory 0.9.4.7

Unified Parameter-Efficient Fine-Tuning of 100+ LLMs

---

Parameter-Efficient Fine-Tuning made easy

PEFT-Factory is a fork of LLaMa-Factory ❤️, upgraded with easy to use PEFT interface, support for HuggingFace PEFT methods and datasets for benchmarking PEFT.

---

Supported methods

PEFT method name	Support
LoRA (including variants)	✅ 🦙
OFT	✅ 🦙
Prefix Tuning	✅ 🤗
Prompt Tuning	✅ 🤗
P-Tuning	✅ 🤗
P-Tuning v2	✅ 🤗
MPT	✅ 🤗
IA3	✅ 🤗
LNTuning	✅ 🤗
Bottleneck Adapter	✅ 🤖
Parallal Adapter	✅ 🤖
SeqBottleneck Adapter	✅ 🤖
SVFT	✅ ⚙️
BitFit	✅ ⚙️

Usage

This section provides instructions on how to install PEFT-Factory, download necessary data and methods, and run training using both command line and web UI.

Quickstart

For video example please visit the PEFT-Factory Demonstration Video.

# install package
pip install peftfactory

# dowload repo that contains data, PEFT methods and examples
git clone https://github.com/kinit-sk/PEFT-Factory.git && cd PEFT-Factory

# start web UI
pf webui

Alternatively, you can run training from command line:

# install package
pip install peftfactory

# dowload repo that contains data, PEFT methods and examples
git clone https://github.com/kinit-sk/PEFT-Factory.git && cd PEFT-Factory

Create some variables for envsubst

# run training with config file
TIMESTAMP=`date +%s`
OUTPUT_DIR="saves/bitfit/llama-3.2-1b-instruct/train_wsc_${TIMESTAMP}"
DATASET="wsc"
SEED=123
WANDB_PROJECT="peft-factory-train-bitfit"
WANDB_NAME="bitfit_llama-3.2-1b-instruct_train_wsc"

mkdir -p "${OUTPUT_DIR}"

export OUTPUT_DIR DATASET SEED WANDB_PROJECT WANDB_NAME

Use the template

Utility envsubst replaces the occurances of env variables with their values (see the template).

envsubst < examples/peft/bitfit/llama-3.2-1b-instruct/train.yaml > ${OUTPUT_DIR}/train.yaml

Run the factory

peftfactory-cli train ${OUTPUT_DIR}/train.yaml

Installation

There are multiple ways to install PEFT-Factory. You can install develelopment version from source or install the latest release from PyPI.

Using pip

pip install peftfactory

From Source

Clone the repository

git clone [e-mail]:kinit-sk/PEFT-Factory.git

Build the wheel package

make build

Install with pip

pip install dist/[name of the built package].whl

Get data and methods

To download data, methods and examples for training please download the repository from GitHub.

git clone https://github.com/kinit-sk/PEFT-Factory.git && cd PEFT-Factory

Run training

You can run training from command line or using web UI.

From Command Line

To run training from command line use the following command:

pf train [path to config file].yaml

Using web UI

To run the web UI use the following command:

pf webui