LLMtuner - Efficiently Adjust Large Language Models with Innovative LoRA and QLoRA Methods

LLMTuner: A Versatile Tool for Fine-Tuning Large Language Models

LLMTuner is a groundbreaking tool designed to make the process of fine-tuning large language models (LLMs) as simple and efficient as possible. With a user-friendly interface inspired by scikit-learn, LLMTuner enables users to fine-tune state-of-the-art models like Whisper and Llama with minimal coding effort. This project is open-source and welcomes contributions from the community.

Key Features

Effortless Fine-Tuning: Simplifies the process of fine-tuning cutting-edge models. Users can quickly and easily customize models to better suit their specific needs.
Integrated Techniques: The tool comes with built-in utilities for advanced fine-tuning techniques such as LoRA and QLoRA, enhancing model performance without the need for extensive additional coding.
Interactive User Interface: LLMTuner features an intuitive UI that allows users to launch webapp demonstrations of their fine-tuned models with a simple click.
Simplified Inference: LLMTuner provides a seamless process for making inferences with fine-tuned models, eliminating the need for additional script writing.
Deployment Readiness: Users can easily deploy their fine-tuned models to platforms like AWS and GCP, ensuring that models are ready to be shared and utilized in broader contexts. This feature is still under development and will be available soon.

Installation and Quick Start

Installation with pip

LLMTuner supports Python 3.7 and above. To install, use the following pip command:

pip3 install git+https://github.com/promptslab/LLMTuner.git

A Quick Tour of LLMTuner

Here's a brief example to get you started with LLMTuner:

from llmtuner import Tuner, Dataset, Model, Deployment

# Initialize the Whisper model for fine-tuning
model = Model("openai/whisper-small", use_peft=True)

# Create a dataset instance
dataset = Dataset('/path/to/audio_folder')

# Set up the tuner
tuner = Tuner(model, dataset)

# Fine-tune the model
trained_model = tuner.fit()

# Perform inference
tuner.inference('sample.wav')

# Launch an interactive UI
tuner.launch_ui('Model Demo UI')

# Deploy the fine-tuned model
deploy = Deployment('aws')
deploy.launch()

Supported Models

LLMTuner currently supports the following models:

Fine-Tune Whisper: Supported with an available Colab Notebook.
Fine-Tune Whisper Quantized: Also supported through LoRA.
Fine-Tune Llama: Support is forthcoming.

Community Engagement

LLMTuner has an active community hub on Discord. Users and contributors interested in open-source LLMs, scalable large model construction, and advanced prompt-engineering are encouraged to join the PromptsLab community. Engaging with other users and sharing insights can enhance learning and application of LLMTuner.

Contributing to LLMTuner

Contributions to the LLMTuner project are highly encouraged. Users can contribute by adding new features, improving infrastructure, or enhancing documentation. For more details on how to contribute, please refer to the contributing guidelines available on the project repository.

LLMTuner stands as an innovative solution for those looking to harness the full potential of large language models through efficient and effective fine-tuning processes.