Version with base64 encoded images for input to GOT

#33
by CSSKK - opened

I'm a newbie at pretty much everything here. Got very excited by the performance of GOTOCR, downloaded the source from GitHub, made the changes to demo file to accept a base64 encoded string of an image (rather than reading from http/disk). Demo is working fine (but takes several seconds). How do I build the modified GOT to get the necessary files -something I got when I did Automodel.save() when reading from HF? The folder with six json files a tiktoken file and a safetensor file. Since i didn't retrain I believe the weights don't need to be redownloaded. Thank you.

CSSKK changed discussion status to closed

Sign up or log in to comment