Supplementary Data: Curvature-dependent anisotropic phonon thermal transport from TH-graphyne monolayer to nanotubes

This folder contains all input and output files for training the TH-GY-NEP potential function. The training dataset contains a total of 620 configurations, including 550 planar TH-GY configurations and 70 curved TH-GY nanotube configurations. All 620 configurations are saved in the train.xyz file.The file nep.in contains hyperparameter settings related to NEP training. The file nep.txt is the potential function file output during the training process; additionally, we performed a total of 200,000 training steps in two successive optimization stages. Furthermore, the output files for energy, force, and virial of the training set are energy_train.out, force_train.out and virial_train.out respectively. The file loss.out records the loss function values throughout the training.