colabfit/Massive_Atomic_Diversity_MAD_test

Massive Atomic Diversity MAD测试数据集是一个用于测试的原子配置数据集，它通过扭曲稳定结构的小集合来创建大量的原子多样性。该数据集包含少于100k个结构，但足以训练出与基于传统数据集训练的模型相媲美的通用原子间势能模型。数据集包含多种元素的原子配置及其能量、原子力和Cauchy应力等属性。

The Massive Atomic Diversity MAD test dataset is a test set of atomic configurations created by distorting small sets of stable structures to generate massive atomic diversity. Despite containing fewer than 100k structures, it has been shown to enable the training of universal interatomic potentials that are competitive with models trained on traditional datasets with two to three orders of magnitude more structures. The dataset includes atomic configurations of various elements along with their energy, atomic forces, and Cauchy stress properties.