LTH14 / mage

A PyTorch implementation of MAGE: MAsked Generative Encoder to Unify Representation Learning and Image Synthesis
MIT License
529 stars 25 forks source link

About your VQGAN codebase and model #22

Closed junwuzhang19 closed 1 year ago

junwuzhang19 commented 1 year ago

You say your VQGAN codebase is different from original VQGAN codebase. Could you explain why you change the original codebase and further finetune the VQGAN model? Thanks!

LTH14 commented 1 year ago

The VQGAN model we used is similar to what is used in MaskGIT, which gives a better reconstruction quality compared with the original codebase.