karpathy/deep-vector-quantization

Issues

• What is data-driven intialization scheme with k-means?

  #5 opened 4 years ago by kritiagg
  2

• About the input of F.gumbel_softmax

  #7 opened 4 years ago by ZikangZhou
  1

• i have another implement . is it correct?

  #9 opened 3 years ago by zhangbo2008
  0

• Missing 1x1 convolutions at the beginning of the decoder

  #6 opened 4 years ago by CDitzel
  0

• Related questions about the gumbel softmax

  #1 opened 4 years ago by CDitzel
  2

• kl divergence loss of GumbelQuantize

  #3 opened 4 years ago by jjjjohnson
  1

• According to the paper, these two summands should be exchanged, should they not?

  #2 opened 4 years ago by CDitzel
  1