qgear 1.1.1

Q-Gear

qgear

paper link: https://arxiv.org/pdf/2504.03967

Preliminary

• Let’s assume you already have a computational GPU node allocated on HPC
• Checking the NVIDIA GPU
  • nvidia-smi
• Create a env (we do not recommend using default such .local / HOME)
• Note that more than one GPU support need to enable MPI > the way we choose is high performance lustre file system

1. Install ENV

clone repo

git clone [e-mail]:gzquse/qgear.git`

cd qgear

module load conda
conda create --prefix=/pscratch/sd/{location}/{username}/qgear -y python=3.11 pip
conda activate $SCRATCH/qgear

pip install -u qgear 
pip install -u ipykernel
python -m ipykernel install --user --name qgear --display-name qgear

2. Open Jupyter Notebook

NERSC jupyter

https://jupyter.nersc.gov/

Select the kernel

go to nbs/example.ipynb; run example

Pypi

https://pypi.org/project/qgear/

Demos

1. simple speed up with random circuit and QFT

https://gzquse.github.io/qgear/examples.html

2. quantum image encoding

see appendix F in the paper https://gzquse.github.io/qgear/apps.html

local development

. ./pm_martin.dev.source

# make sure qgear package is installed in development mode
https://nbdev.fast.ai/tutorials/tutorial.html
pip3 install -e '.[dev]'
pip3 install qgear

# compile to have changes apply to qgear
nbdev_prepare

Goal

build the versatile all-in-one quantum accelerator for HPC-QPU hybrid regime that supports all the mainstream quantum frameworks.