zalandoresearch/pytorch-vq-vae

Issues

• Replicating the bits / dim results in the original VQ-VAE paper

  #1 opened 5 years ago by pclucas14
  7

• Unintentional decay of embeddings towards 0?

  #10 opened 4 years ago by samedii
  0

• Codebook update

  #9 opened 4 years ago by GiorgioBarnabo
  0

• dimension issue

  #8 opened 4 years ago by jlian2
  0

• Hi, why do we need to swap axis before we flatten?

  #7 opened 4 years ago by diff7
  0

• How does the reconstruction loss update the encoder?

  #6 opened 5 years ago by eecshope
  2

• Hi,how its possible to train VQ-VAE on my own data set images? Thanks

  #3 opened 5 years ago by zaitoun90
  10

• EMA update before quantization

  #5 opened 5 years ago by stangelid
  5

• How to extract only the latent codes?

  #4 opened 5 years ago by plaffa
  0

• Intentional weight sharing in ResidualStack?

  #2 opened 5 years ago by andreaskoepf
  1