arrow
Datasets
All datasets matching “arrow”seed-tts-eval-arrowGenImage-arrow
GenImage Arrow
Generator- and split-partitioned Arrow release of the GenImage benchmark.
Each leaf directory is also a standalone Hugging Face save_to_disk bundle.
Contents
Train: 2,581,150 valid images across eight generators.
Test: 100,000 images across eight generators.
Validation is an alias of test because the official GenImage val directory
is the benchmark test set. It is not a third independent split.
Seventeen unavailable or zero-byte upstream train… See the full description on the dataset page: https://huggingface.co/datasets/nebula/GenImage-arrow.DF-arrowGenImage_webp: Just convert all images of GenImage to webp format, just save disk space
GenImage_test: only contains raw GenImage test samples
GenImage: raw GenImage with all data
seed-tts-eval-50-arrowfiltering-pretraining-mix-arrow-formattachiwin_multilingual_audio_curricular_arrow