PASTA

PASTA数据集是由太平洋西北国家实验室开发的稀疏张量算法基准套件，包含12个计算稀疏张量操作，如张量元素操作、张量-标量操作等。该数据集旨在帮助应用用户评估不同的计算机系统，通过其代表性的计算工作负载，提供对现有计算机架构和系统更好利用的见解，并为未来设计提供灵感。PASTA数据集将公开发布，适用于机器学习、量子化学、医疗分析等多个领域，旨在解决大数据处理中的性能和解释性问题。

The PASTA Dataset is a sparse tensor algorithm benchmark suite developed by the Pacific Northwest National Laboratory, encompassing 12 types of computational sparse tensor operations such as element-wise tensor operations and tensor-scalar operations. This dataset is designed to help application users evaluate different computer systems. By leveraging its representative computational workloads, it provides insights into the better utilization of existing computer architectures and systems, and offers inspirations for future system designs. The PASTA Dataset will be publicly released, and it is applicable to multiple fields including machine learning, quantum chemistry, medical analytics and others. Its core goal is to address performance and interpretability issues in big data processing.