datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
furry-e621-sfw-7m-hq
Dataset Card for furry-e621-sfw-7m-hq
Dataset Summary
This is 6.92 M captions of the images from the safe-for-work (SFW) split of e621 ("e926"). It extends to January 2023, before the widespread advent of machine learning images. It includes captions created by LLMs and a custom multilabel classifier along with CogVLM. There are 8 LLM (mistralai/Mistral-7B-v0.1) and 1 CogVLM (THUDM/CogVLM) captions per image.
Most captions are substantially larger than 77 tokens and are… See the full description on the dataset page: https://huggingface.co/datasets/CaptionEmporium/furry-e621-sfw-7m-hq.furry-e621-safe-llama3.2-11b
furry-e621-safe-llama3.2-11b: A new anthropomorphic art dataset
Dataset Summary
This is 2,987,631 synthetic captions for 995,877 images found in e921, which is just e621 filtered to the "safe" tag. The long captions were produced using meta-llama/Llama-3.2-11B-Vision-Instruct. Medium and short captions were produced from these captions using meta-llama/Llama-3.1-8B-Instruct The dataset was grounded for captioning using the ground truth tags on every post categorized and… See the full description on the dataset page: https://huggingface.co/datasets/CaptionEmporium/furry-e621-safe-llama3.2-11b.