Ovis-dataset

## Usage
https://github.com/AIDC-AI/Ovis/tree/v1.5?tab=readme-ov-file#dataset

## Description
This dataset is a collection of multimodal datasets used for training Ovis. Ovis is a novel Multimodal Large Language Model (MLLM) architecture, designed to structurally align visual and textual embeddings. For a comprehensive introduction, please refer to the [Ovis paper](https://arxiv.org/abs/2405.20797) and the [Ovis GitHub repo](https://github.com/AIDC-AI/Ovis).

## License
The files `laion-description-11k.json`, `cc12m-description-1m.json`, and `cc12m-qa-387k.json` are newly released by us and are licensed under CC BY 4.0. All other files are from publicly available datasets and are governed by their specific licensing conditions.

### 使用方式
https://github.com/AIDC-AI/Ovis/tree/v1.5?tab=readme-ov-file#dataset

### 数据集说明
本数据集为用于训练Ovis的多模态数据集集合。Ovis是一种新型多模态大语言模型（Multimodal Large Language Model，MLLM）架构，旨在实现视觉嵌入与文本嵌入的结构对齐。如需全面了解相关内容，请参阅Ovis研究论文（https://arxiv.org/abs/2405.20797）以及Ovis的GitHub仓库（https://github.com/AIDC-AI/Ovis）。

### 许可协议
文件`laion-description-11k.json`、`cc12m-description-1m.json`与`cc12m-qa-387k.json`为我们全新发布，采用CC BY 4.0协议进行许可。其余所有文件均来自公开可用数据集，受其各自专属许可条款约束。