colabfit/discrepencies_and_error_metrics_NPJ_2023_vacancy_re_testing_set

该数据集名为discrepencies and error metrics NPJ 2023 vacancy re testing set，包含从AIMD模拟中获取的单一迁移空位的原子结构测试集。它是为了开发具有更高扩散性评分的模型而构建的，基于修改后的mlearn_Si_train数据集，其中约54%的数据被替换为含有迁移间隙原子的结构。验证集包含50个结构，测试集包含100个原子配置，所有结构均含有单一迁移空位。数据集包含100个独特的分子配置和6300个硅原子，属性包括能量和原子力。

This dataset, named discrepencies and error metrics NPJ 2023 vacancy re testing set, includes a testing set of atomic structures with a single migrating vacancy obtained from AIMD simulations. It is constructed to develop models with better diffusivity scores, based on a modified version of the mlearn_Si_train dataset with approximately 54% of the data replaced by structures containing migrating interstitials. The validation set consists of 50 structures, and the testing set includes 100 atomic configurations, each with a single migrating vacancy. The dataset contains 100 unique molecular configurations and 6300 silicon atoms, with properties including energy and atomic forces.