mikemike2001/laion-subset
收藏Hugging Face2026-04-17 更新2026-04-26 收录
下载链接：
https://hf-mirror.com/datasets/mikemike2001/laion-subset
下载链接
链接失效反馈官方服务：
资源简介：
---
dataset_info:
features:
- name: image
dtype: image
- name: caption
dtype: string
splits:
- name: train
num_bytes: 4155522966
num_examples: 24840
download_size: 4130109564
dataset_size: 4155522966
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
license: mit
task_categories:
- text-to-image
tags:
- stable-diffusion
- lcm
- lcm-lora
- training-data
- laion
pretty_name: LAION Subset for LCM-LoRA Training
size_categories:
- 10K<n<100K
---
# LAION Subset for LCM-LoRA Training
**Author:** Juhi Singh | [HuggingFace](https://huggingface.co/juhirats)
## Dataset Description
This dataset is a curated subset of LAION-5B used for training Latent Consistency Model (LCM) LoRA adapters on Stable Diffusion v1.5. It contains high-quality image-caption pairs optimized for distillation training.
## Dataset Details
- **Total Examples:** 24,840 image-caption pairs
- **Dataset Size:** ~4.16 GB
- **Resolution:** 512×512
- **Format:** Parquet files with image and caption columns
- **Source:** Curated from LAION-5B
- **Use Case:** LCM-LoRA distillation training
## Dataset Structure
Each example contains:
- `image`: PIL Image object (512×512)
- `caption`: Text description of the image
```python
提供机构：
mikemike2001
