jazelly / FinetuneLLMs

A WebUI for LLM fast finetuning on GPU or CPU.

FinetuneLLMs (Work in Progress, Actively! 🔥)

Finetune an LLM, within a few clicks!

🔥Goal & Roadmap🔥

The main objective of this project is to lower the barrier to training large language models, especially for startup companies that have hardware in hands. With this project, it should be easy for a company to start experimenting with LLM training within a basic setup on servers with GPU/CPU resources.

In a way, it helps provide an opportunity for everyone who has hardware available and wants to utilize it in the AI field.

Roadmap

Supported finetuning techniques

Model \ Method	SFT	DPO	ORPO	KTO	PRO
llama 2	✅	❌	❌	❌	❌
llama 3	✅	❌	✅	❌	❌
gguf	✅	❌	❌	❌	❌
phi-3	✅	❌	❌	❌	❌
Mistral	✅	✅	❌	❌	❌
...	?	?	?	?	?

General Setup

This repo provides 3 modules, frontend (react), server (nodejs), and trainer (python django)

You need CUDA for now, but once llama.cpp is integrated, this will no longer be required.

• For Linux

  Install CUDA from Nvidia installation guide

• For Windows (with Nvidia GPU)

  Enable WSL2 on your machine.

  Install CUDA from Nvidia installation guide

Dev Setup

Setup frontend

cd frontend
// copy .env.example to .env
npm ci
npm run dev
// or yarn && yarn dev

Setup server

cd server
// copy .env.example to .env and .env.development
npx prisma migrate dev
npm i
npm run dev

Setup trainer

Manual installation using Conda

Recommended if you have some experience with the command-line.

0. Install Conda

https://docs.conda.io/en/latest/miniconda.html

On Linux or WSL, it can be automatically installed with these two commands (source):

curl -sL "https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh" > "Miniconda3.sh"
bash Miniconda3.sh

1. Create a new conda environment

conda create -n fllms python=3.11
conda activate fllms

2. Install Pytorch

System	GPU	Command
Linux/WSL	NVIDIA	pip3 install torch==2.2.1 torchvision==0.17.1 torchaudio==2.2.1 --index-url https://download.pytorch.org/whl/cu121
Linux/WSL	CPU only	pip3 install torch==2.2.1 torchvision==0.17.1 torchaudio==2.2.1 --index-url https://download.pytorch.org/whl/cpu
Linux	AMD	pip3 install torch==2.2.1 torchvision==0.17.1 torchaudio==2.2.1 --index-url https://download.pytorch.org/whl/rocm5.6
MacOS + MPS	Any	pip3 install torch==2.2.1 torchvision==0.17.1 torchaudio==2.2.1
Windows	NVIDIA	pip3 install torch==2.2.1 torchvision==0.17.1 torchaudio==2.2.1 --index-url https://download.pytorch.org/whl/cu121
Windows	CPU only	pip3 install torch==2.2.1 torchvision==0.17.1 torchaudio==2.2.1

The up-to-date commands can be found here: https://pytorch.org/get-started/locally/.

For NVIDIA, you also need to install the CUDA runtime libraries:

conda install -y -c "nvidia/label/cuda-12.1.1" cuda-runtime

If you need nvcc to compile some library manually, replace the command above with

conda install -y -c "nvidia/label/cuda-12.1.1" cuda

3. Install dependencies

cd trainer
pip install -r requirements.txt
python watch.py