Diffusion Pretrain Set Ft1: Multi-Source Image-Caption Data for Stable Diffusion 1.x

AbstractPhil's diffusion-pretrain-set-ft1 is a multi-source image-caption dataset assembled from seven upstream sources via a uniform ingest pipeline. It was built for the finetune-1 stage of the sd15-flow-lune model family but is applicable to any Stable Diffusion 1.x conditioning experiment. The dataset was last updated on May 22, 2026.

Use Cases

• Fine-tuning Stable Diffusion 1.x models based on the assembled multi-source image-caption pairs.
• Conducting ablation studies on pretraining data sources based on the availability of standalone configurations for each of the seven upstream sources.
• Pretraining text-to-image diffusion models based on the large-scale, uniformly processed image-caption data.

Strengths

• Assembled from seven distinct upstream sources, which likely provides diversity in image and caption content.
• Offers flexible loading options, with each source available as a standalone config and as part of a full config.
• Built with a uniform ingest pipeline, which suggests consistency in data formatting across sources.

Limitations

• Description metadata is limited; actual data quality, scale, and content require manual inspection after download.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count, file formats, and license information are unknown, which may limit suitability assessment.

Provenance

Source

Assembled from seven upstream sources (specifics not provided) by author AbstractPhil.

Collection Method

Assembled via a uniform ingest pipeline.

Freshness

Last updated 2026-05-22 08:19:21; freshness should be verified.