---
license: mit
task_categories:
- text-to-image
language:
- en
tags:
- t2i
- cleaned
configs:
- config_name: full
  data_files: "full.parquet"
  default: true
- config_name: lite
  data_files: "lite.parquet"
---

| Files        | Purpose                                 |
|--------------|-----------------------------------------|
| full.parquet | Full version (recommended for training) |
| lite.parquet | A mini version of the full dataset      |

Differences compared to other datasets:
- Other datasets are often not well-suited for text-to-image tasks.
- Other datasets may be in different languages.
- Other datasets may contain grayscale images.
- Other datasets are often extremely large.
- Other datasets frequently lack clear previews.
- Other datasets may not use permissive licenses (such as the MIT license) or may be gated (restricted access).
- Other datasets are often AI-generated.

With this dataset:
- You can create your own small-scale text-to-image model.
- You have greater flexibility and freedom in how you use it.

ty!