AbstractPhil/flux-schnell-teacher-latents

这是一个合成的图像和文本数据集，包含图像到图像、图像到文本和文本到图像的任务类别。数据集将包含多个子集，涵盖面部、头发、手、脚、衣服、角度、姿势等特征，其复杂度将随时间增长至5。数据集适用于满足三规则标准的场景，但效果更为紧凑和减弱。

This is a synthetic image and text dataset, including task categories of image-to-image, image-to-text, and text-to-image. The dataset will contain multiple subsets covering features such as faces, hair, hands, feet, clothing, angles, poses, etc., with complexity growing over time to level 5. The dataset is suitable for scenarios that meet the criteria of the rule of three, but with a more compact and reduced effect.