MatInFormer: Materials Informatics Transformer

[arXiv][PDF]

Hongshuo Huang, Rishikesh Magar ,Changwen Xu, Amir Barati Farimani
Carnegie Mellon University

This is the official implementation of MatInFormer: "Materials Informatics Transformer: A Language Model for Interpretable Materials Properties Prediction". In this work, we introduce a novel approach that involves learning the grammar of crystallography through the tokenization of pertinent space group information. We further illustrate the adaptability of MatInFormer by incorporating task-specific data pertaining to Metal-Organic Frameworks (MOFs).

Enviroment Set up

conda env create -f environment.yml

Example of SSL

cd /path/to/matinformer/
python sgt/pretrain.py

Here is a example data for lattice parameter prediction. Nottice The lattice parameter should be normalized.

material-id	composition	space group	a	b	c	$\alpha$	$\beta$	$\gamma$
foo-1	Hf2 Si2 Te2	P4/nmm	3.67	3.67	27.31	90	90	90
foo-2	Ti2 Br2 O2	Pmna	3.70	4.70	25.44	90	90	90

Example of fintune

For materials project benchmark, you can train and test by

python sgt/mb_train.py

if you want to run on your own dataset like MOFs, change the config file and run

python sgt/hoip_train.py

The model takes input in the form csv files with materials-ids, composition strings , space group symbol and target values as the columns.

material-id	composition	space group	target
foo-1	Hf2 Si2 Te2	P4/nmm	63.5
foo-2	Ti2 Br2 O2	Pmna	134.8

Cite This Work

If you use this code please cite the relevant work:

MatInFormer - Materials Informatics Transformer: A Language Model for Interpretable Materials Properties Prediction [arXiv]

@article{huang2023materials,
title={Materials Informatics Transformer: A Language Model for Interpretable Materials Properties Prediction},
author={Huang, Hongshuo and Magar, Rishikesh and Xu, Changwen and Farimani, Amir Bariti},
journal={arXiv preprint arXiv:2308.16259},
year={2023}
}

wren - Rapid Discovery of Stable Materials by Coordinate-free Coarse Graining. [Paper][arXiv]

@article{goodall_2022_rapid,
title={Rapid discovery of stable materials by coordinate-free coarse graining},
author={Goodall, Rhys EA and Parackal, Abhijith S and Faber, Felix A and Armiento, Rickard and Lee, Alpha A},
journal={Science Advances},
volume={8},
number={30},
pages={eabn4117},
year={2022},
publisher={American Association for the Advancement of Science}
}