News-5m

该数据集包含了500万个高质量的英文句子，这些句子来源于开源新闻，用于知识蒸馏阶段的无标签数据集构建。此外，该数据集还与构建学生模型微调阶段相关。规模达到了500万个句子，任务定位于知识蒸馏。

This dataset contains 5 million high-quality English sentences sourced from open-source news, and it is used to construct unlabeled datasets for the knowledge distillation stage. Additionally, this dataset is also relevant to the fine-tuning phase when constructing student models. With a total of 5 million sentences, this dataset is targeted at knowledge distillation tasks.