Questions about the image infilling task in pretraining.

OFA-Sys / OFA

Official repository of OFA (ICML 2022). Paper: OFA: Unifying Architectures, Tasks, and Modalities Through a Simple Sequence-to-Sequence Learning Framework

Apache License 2.0

2.41k stars 248 forks source link

Questions about the image infilling task in pretraining. #377

Closed huizyuan closed 1 year ago

huizyuan commented 1 year ago

About the image infilling task in pretraining, how can I use my own dataset? Can authors kindly provide the details of generating the image-code through VQ-GAN？

huizyuan commented 1 year ago

This question is fixed by https://github.com/OFA-Sys/OFA/issues/279