---
title: Doc-750K
canonical_url: "https://www.modelscope.cn/datasets/OpenGVLab/Doc-750K"
md_url: "https://www.modelscope.cn/datasets/OpenGVLab/Doc-750K.md"
repository: OpenGVLab/Doc-750K
last_updated: 2025-07-22
license: mit
storage_size: "1.3 TB"
downloads: 1011
stars: 3
---

# Doc-750K

> Doc-750K - OpenGVLab 在 ModelScope 开源的数据集。This dataset is in the paper Docopilot: Improving Multimodal Models for Document-Level Understanding.

OpenGVLab/Doc-750K 是 ModelScope 魔搭社区上的数据集，存储大小 1.3 TB，采用 mit 许可。

- **Repository**: OpenGVLab/Doc-750K
- **License**: mit
- **Storage size**: 1.3 TB
- **Downloads**: 1011
- **Stars**: 3
- **Last updated**: 2025-07-22

Source: https://www.modelscope.cn/datasets/OpenGVLab/Doc-750K

---

This dataset is in the paper [Docopilot: Improving Multimodal Models for Document-Level Understanding](https://arxiv.org/abs/2507.14675).

Please refer to https://github.com/OpenGVLab/Docopilot for details.

## FAQ
### Unzipping Split Archives on Linux
If you encounter issues when unzipping the image archive on Linux, such as:
- zip bomb warnings
- bad zipfile offset errors

Please try the following solutions:

1. Zip Bomb Warning
Some systems may trigger a zip bomb detection warning due to the large number of small image files. You can bypass this by disabling the detection with:
```bash
export UNZIP_DISABLE_ZIPBOMB_DETECTION=TRUE
```
2. Bad Zipfile Offset Error
If you're dealing with split zip archives (e.g., images.z01, images.z02, ..., images.zip), you need to merge them before unzipping:

```bash
zip -s 0 images.zip --out images_full.zip
unzip images_full.zip
```
This will reconstruct the full archive and allow you to unzip it normally.

Note: The image dataset is very large, so please ensure you have sufficient disk space and patience during extraction.