search

dandelin / ViLT

Code for the ICML 2021 (long talk) paper: "ViLT: Vision-and-Language Transformer Without Convolution or Region Supervision"
Apache License 2.0
1.36k stars 209 forks source link

pretrain datasets #74

Open mactavish91 opened 1 year ago

mactavish91 commented 1 year ago

Hello, the author, great work! As time goes by, a lot of image urls in the dataset become invalid. Is there any solution? Could you provide the data arrow?