.:::: .::::::: .::: .::::::.:: .:: .::
.:: .:: .:: .:: .:: .:: .:: .:: .::
.:: .::.:: .:: .:: .:: .:: .::
.:: .::.::::::: .:: .:: .::
.:: .::.:: .:: .:: .:: .::
.:: .:: .:: .:: .:: .:: .:: .::
.:::: .:: .:: .:: .:: .:::: ::
- Transforming SMILES into .xyz and calculating descriptors for XC functionals optimization
- Machine Learning Model for XC functionals optimization when calculating the charge-transfer excited or local excited state of single molecular systems in TDDFT
- Scripts for batch generation of Gaussian tasks before and after optimization of XC functionals
To clone and run this application, you'll need some python packages installed on your computer. From your command line:
# Clone this repository
$ git clone https://github.com/STOKES-DOT/code_ml_dft
# Go into the repository
$ cd code_ml_dft
You can install the OPTXC package using the following command:
# Go into the OPTXC folder
$ cd sbin
# Install the OPTXC package
$ python setup.py install
Attenation: The setpy.py needs to be run within internet connection. If you have no internet connection, you can't install the OPTXC package. The python version should be 3.6 or higher. The final stacking model is not training. You can find the trianing code in SGM folder such as SGM_LC-wPBE folder. Don't forget to change the base-learners and training date path in the training code.
To optimize the XC functionals of a given molecule, you can use the following code in example folder:
# Go into the example folder
$ cd example
# You can use the following code to optimize the XC functionals of a given molecule with a given initial guess
$ optxc example.inpThe input file example.inp should contain the following information:
#INPUT_FILE=example.inp THIS IS AN EXAMPLE INPUT FILE. YOU COULD EDIT IT ACCORDING TO YOUR NEEDS.
SMILES = C1=CC=C(C=C1)C2=CC=CC=C2 # SMILES of the molecule
XC_FUNCTIONAL = LC-wPBE # XC functional to be optimized, Now this package only supports LC-wPBE and wB97XD functionals
OPT_MODEL = XBGOOST # Optimization model, Now this package supports XGBoost, Random Forest, Gradient Boosting models,CatBoost models, and some other models.
SLURM_PATH = descriptors_maker_QM\test_1.slurm # SLURM file path for generating Gaussian tasks, You can generate the slurm file using the script in the descriptors_maker_QM folder.This code will generate a .xyz file for the given molecule, and then optimize the XC functionals using the given initial guess. The optimized XC functionals will be saved in the same folder with the initial guess.
Note Here, We give two XC funationals optimization code. If you want to create a new model for optimizing other XC functional, Please refer to our article in XXX (Will be pulished in a few time). The final stacking model is not training because of the limit in Github's storage space. You need to finish training it by youself. Besides, each base-learners could also be used for optimization with a lower performance. The packge needed for training stacking model is listed in packge.txt.
If you liked using this model or it has any questions about these codes or our theory details, I'd like you send me an email at jiaoyuan24@mails.ucas.ac.cn about anything you'd want to say about these codes. I'd really appreciate it!
This code uses the following open source packages (We have listed in the project. You can install them accroding to your need.):
If you like this project and think it has helped in any way, Sent me 2.5$ for a cup of coffee! Just a joke, haha~. Have a great day!
This project is just an idea. We will keep updating it in the future. If you have any questions or suggestions, please feel free to contact me.
UCAS,SAIS