emotionor / APET

Atomic Positional Embedding-based Transformer

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

APET

Atomic Positional Embedding-based Transformer introduced by Cui Yaning, et al.

Dataset

The DOS dataset are avaliable in Material Project (https://next-gen.materialsproject.org/). All the data this work uses have been converted to three csv file, which is visible in the ./data. The data files contain the corresponding MP-ids in Material Project, which can be verified by all by themselves.

Run the model

Run ./data/csv2npy.py to generate the dataset for the APET pretraining task. Run train_script.sh to start running the APET pretraining task.

About

Atomic Positional Embedding-based Transformer

License:GNU General Public License v3.0


Languages

Language:Python 99.7%Language:Shell 0.3%