|
|
--- |
|
|
license: cc0-1.0 |
|
|
task_categories: |
|
|
- text-generation |
|
|
- text-classification |
|
|
language: |
|
|
- en |
|
|
tags: |
|
|
- code |
|
|
- art |
|
|
pretty_name: mg_sd_washed |
|
|
size_categories: |
|
|
- 10K<n<100K |
|
|
--- |
|
|
|
|
|
# MagicPrompt_SD_Washed |
|
|
|
|
|
It's a version of datasets of [Gustavosta/MagicPrompt-Stable-Diffusion](https://huggingface.co/Gustavosta/MagicPrompt-Stable-Diffusion). |
|
|
|
|
|
When I want to train a model using origin data, some bad prompts broke model and waste many time. |
|
|
|
|
|
So I washed the origin datasets: |
|
|
|
|
|
1. 😄 delete some meanless words like some artists name with misspelling |
|
|
2. 😂 delete many spaces that make `100mm` to `10 0m` |
|
|
3. 😭 some url in datasets |
|
|
4. 😭 and many unknown words |
|
|
|
|
|
|
|
|
And this version is doing well in my train test!😍 |
|
|
|
