autoencoder on content - Githubissues

grenlayk / deep-text-edit

Our implementation of Text Style Brush architecture.

MIT License

119 stars 27 forks source link

autoencoder on content #36

Open kimyerim0908 opened 1 year ago

kimyerim0908 commented 1 year ago

Did you train the autoencoder model up to 500epoch? How long does it take?

SphericalPotatoInVacuum commented 8 months ago

I've sifted through the logs of our runs, but couldn't find the autoencoder ones :( We now know that when given enough memory and compute it manages to train pretty well without the autoencoder hack, so there's that. Sorry.