metadata
license: cdla-permissive-2.0
dataset_info:
features:
- name: id
dtype: int64
- name: caption
dtype: string
- name: tokens
sequence: int16
I can't get the dataset viewer to work, sorry. There's about 3M images and captions from Spawning/PD3M. They are resized and center-cropped to 512x512, and then tokenized into discrete tokens with NVIDIA Cosmos-Tokenizer-DI8x8, which reduces the spatial dimension by a factor of 8, resulting in 64 x 64 = 4096 discrete tokens per image.
You can use these tokenized images to train an auto-regressive image model, or a MaskGIT. Or probably other things I don't know about. :) License is the same as the original image dataset.