BoyaWu10/Bunny-v1_0-data

Bunny-v1.0数据集是Bunny-v1.0系列模型的训练数据集，包括Bunny-v1.0-3B。该数据集的预训练数据来源于LAION-2B的高质量核心集，经过去重和信息丰富化处理后，随机选取了200万张图像-文本对进行训练。微调数据则是通过修改SVIT-mix-665K数据集构建的Bunny-695K。数据集的使用方法包括下载图像包并合并解压。整个项目内容遵循Apache 2.0许可证。

The Bunny-v1.0 dataset is created for training the Bunny-v1.0 series models, including Bunny-v1.0-3B. It is used for visual question answering and question answering tasks, containing 2 million randomly sampled image-text pairs from a high-quality coreset of LAION-2B. The dataset is divided into pretrain and finetune parts, stored in `pretrain` and `finetune` folders respectively. The images in the dataset are packed into multiple parts, and users need to download, merge, and unpack them. The dataset is licensed under Apache 2.0.

原始信息汇总

数据集概述

数据集名称

• Bunny-v1.0-3B

相关资源

• Technical report
• Code
• Demo