---
title: BLIP3o-60k
canonical_url: "https://www.modelscope.cn/datasets/BLIP3o/BLIP3o-60k"
md_url: "https://www.modelscope.cn/datasets/BLIP3o/BLIP3o-60k.md"
repository: BLIP3o/BLIP3o-60k
last_updated: 2025-05-26
license: "Apache License 2.0"
storage_size: "101 GB"
downloads: 21753
stars: 2
---

# BLIP3o-60k

> BLIP3o-60k - BLIP3o 在 ModelScope 开源的数据集。This is BLIP3o-60k Text-to-Image instruction tuning dataset distilled from GPT-4o, including the following categories:

BLIP3o/BLIP3o-60k 是 ModelScope 魔搭社区上的数据集，存储大小 101 GB，采用 Apache License 2.0 许可。

- **Repository**: BLIP3o/BLIP3o-60k
- **License**: Apache License 2.0
- **Storage size**: 101 GB
- **Downloads**: 21753
- **Stars**: 2
- **Last updated**: 2025-05-26

Source: https://www.modelscope.cn/datasets/BLIP3o/BLIP3o-60k

---

This is BLIP3o-60k Text-to-Image instruction tuning dataset distilled from GPT-4o, including the following categories:

1. JourneyDB
2. Human (including MSCOCO with human caption, human gestures, occupations)
3. Dalle3
4. Geneval (no overlap with test set)
5. Common objects
6. Simple text


Here we provide the code guidance to download tar file:
```
from huggingface_hub import snapshot_download
snapshot_download(repo_id='BLIP3o/BLIP3o-60k', repo_type=‘dataset’)
```

And you can use huggingface datasets to read the tar file without unzipping them:

```
from datasets import load_dataset
import glob
data_files = glob.glob('/your/datasets/path/*.tar') 
train_dataset = load_dataset("webdataset", data_files=data_files, cache_dir='/your/cache/directory/', split="train", num_proc=64)
```