Configuration

Code is tested under the environment of Pytorch 1.9.0, Python 3.8 and CUDA 11.1 on Windows.

Data: The data for this study are presented by Paper High-Performing Deep Learning Regression Models for Predicting Low-Pressure CO2Adsorption Properties of Metal−Organic Frameworks and can be downloaded here.

Usage

Direct use of pre training model

Modify 'path' in line 11 of predict_list.py to be predicted CIF path, click run to predict, The predictions will be generated in the root directory in the form of a table. To modify the prediction target, change line 18 in predict_list.py to 'model_sort_cap.pht'

Training tasks from scratch

Modify 'path' in line 8 of CIF_to_npy.py to be predicted CIF path, click run to project CIF files. Modify 'train_path' in line 15 of train.py to be npy file path, click Run to train. The code will automatically save the best model in a directory.