joh-schb/PlantLDM

Encoder-Decoder VQ-VAE

Closed this issue · 1 comments

In VQ-VAE

  • make vector quantizer not compute the loss
  • write encode function that returns $z_q$
  • write decode function that gets $z_q$, quantizes it to $z_q$ and then decodes $z_q$ to $\hat{x}$
  • compute loss in train_vqvae.py

Need to test it and will then create a merge request.