Installation

Follow the steps below to set up your environment.
For Qulacs to work with multiple cores follow the extra steps extra steps, otherwise just do the pip install's.
Recommended Python version is 3.12.

python3.12 -m venv myvenv  
source myvenv/bin/activate

Default qulacs simulator

pip install qulacs

Fast qulacs simulators (linux) use similar commands with homebrew for macos

sudo apt install gcc-14 g++-14  
sudo apt install libboost-all-dev

export C_COMPILER=gcc-14  
export CXX_COMPILER=g++-14  
export QULACS_OPT_FLAGS="-mtune=native"  
pip install git+https://github.com/qulacs/qulacs.git  

other dependencies

pip install tequila-basic  
pip install pyscf  
pip install torch==2.7.0
pip install torch-cluster --no-build-isolation
pip install torch_geometric  

Project structure

variational-parameter-modeling/
├── README.md
├── code/
│ ├── train.py
│ └── *.py
├── data/
│ └── .csv # (in subfolders, includes datasets used in thesis)
└──.pth # Model files saved/loaded outside main folder

From outside the main folder you can run the following functionalities as python modules.

Training

You can define the details in the file train.py (model, dataset, evaluation). Then run the training as a module. During Training the best model will get saved to the main folder, for later usage.

python -m variational-parameter-modeling.code.train

Testing Model Performance

Here you can plot the dataset and model performance of your trained models. Select the dataset and the model that you want to evaluate in the file test_model.py.

python -m variational-parameter-modeling.code.test_model

Generating Datasets

Define MAX_ATOMS, AMOUNT and the geometry (structure) of molecules you want to use, which will produces different outcomes (dataset quality, learnability of model). The Dataset is the main driver for model performance.

python -m variational-parameter-modeling.code.generate_dataset