MatSci ML

MatSci ML是由英特尔实验室创建的一个专注于固态材料建模的新型基准数据集。该数据集整合了多个开源数据集，如OpenCatalyst、OQMD、NOMAD等，提供了多样化的材料系统和属性数据，包括模拟能量、原子力、材料带隙以及通过空间群的晶体对称性分类数据。MatSci ML支持多任务训练方法，使得针对固态材料的多任务学习算法成为可能，并且通过多数据集集成，促进了跨多个数据集的新算法和方法的开发。该数据集旨在解决固态材料建模中的性能比较和通用性问题，推动材料科学领域的研究进展。

MatSci ML is a novel benchmark dataset dedicated to solid-state materials modeling, developed by Intel Labs. It integrates multiple open-source datasets including OpenCatalyst, OQMD, NOMAD and others, offering diverse material systems and property data such as simulated energies, atomic forces, material band gaps, and crystal symmetry classification data based on space groups. MatSci ML supports multi-task training paradigms, enabling the implementation of multi-task learning algorithms for solid-state materials. Through multi-dataset integration, it also facilitates the development of innovative algorithms and methodologies across multiple datasets. This dataset is designed to address the performance comparison and generalizability challenges in solid-state materials modeling, and promote research advancements in the materials science domain.