Datasets for manuscript: "Gradient-based grand canonical optimization enabled by graph neural networks with fractional atomic existence."

dataset_orb_v3_conservative_inf_omat.xyz: Full dataset used for training of GNN model. generated.xyz: Configurations of Cu(110) c(6x2) generated by the model with grand canonical optimization.dataset_orb_v3_conservative_inf_omat.xyz：用于图神经网络（Graph Neural Network, GNN）模型训练的完整数据集。 generated.xyz：由该模型经巨正则系综优化生成的Cu(110) c(6x2)原子构型。