LAION Subset: 24,840 Image-Caption Pairs for LCM-LoRA Training

Mercity's LAION Subset is a curated collection of 24,840 high-quality image-caption pairs from the LAION-5B dataset, formatted at 512x512 resolution. It is designed for training Latent Consistency Model (LCM) LoRA adapters on Stable Diffusion v1.5. The dataset, last updated in November 2025, occupies approximately 4.16 GB and is stored in Parquet files.

Use Cases

• Fine-tuning Stable Diffusion v1.5 models based on high-quality image-caption pairs.
• Training Latent Consistency Model (LCM) LoRA adapters based on the described distillation training objective.
• Conducting research on image-text alignment and model distillation based on curated LAION-5B data.

Strengths

• Contains 24,840 image-caption pairs, providing a substantial sample for training.
• Images are processed to a consistent 512x512 resolution, which is a standard input size for Stable Diffusion v1.5.
• Dataset size is approximately 4.16 GB, indicating a manageable download for the number of examples.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• The dataset is a curated subset, which may reflect selection biases inherent to the curation process from LAION-5B.

Provenance

Source

Curated from the LAION-5B dataset.

Collection Method

Likely filtered and processed for quality and specific training purposes.

Freshness

Last updated 2025-11-26 10:45:41; freshness should be verified.