LLaMa-EasyFT: A Toolkit for Fine-Tuning Large Language Models with LoRA and DeepSpeed

Usage

• Setup environment:

git clone https://github.com/ChiYeungLaw/LLaMa-EasyFT.git
conda create -n llamaLora python=3.10
conda activate llamaLora
conda install pytorch==1.12.0 torchvision==0.13.0 torchaudio==0.12.0 cudatoolkit=11.3 -c pytorch
git clone https://github.com/huggingface/transformers.git
cd transformers
pip install -e .
cd ..
pip install -r requirements.txt

• Alpaca Training data is avaliable:

data/alpaca_data.json

• Training LLaMA-7B:

bash finetune.sh

• Inference

bash generate.sh

Thanks For

This toolkit are based on many great open source projects:

Meta AI LLaMA, Huggingface Transformers Llama, Alpaca, Alpaca-LoRA , and LLaMa-X