Open-Bee/Bee-Training-Data-Stage2

Bee-Training-Data-Stage2是Bee-8B多模态大型语言模型训练的第二阶段数据集，基于高质量的Honey-Data-15M语料库，使用了HoneyPipe管道进行数据清洗和增强。该数据集用于图像到文本的任务，包含大约1500万样本，是一个全开源的数据集。Bee-Training-Data-Stage2 is the second stage dataset for training the Bee-8B multimodal large language model, based on the high-quality Honey-Data-15M corpus and enhanced with the HoneyPipe data cleaning and enrichment pipeline. This dataset is used for image-to-text tasks, containing approximately 15 million samples, and is fully open-source.