lucidrains/recurrent-interface-network-pytorch

Issues

• Compute required to get a minimal setup running

  #20 opened a month ago by kushalj001
  0

• follow up paper

  #22 opened 3 months ago by lucidrains
  8

• RIN with 100k steps, still noisy images?

  #7 opened 2 years ago by Samin100
  45

• ValueError: num_samples should be a positive integer value, but got num_samples=0

  #21 opened 8 months ago by ad48hp
  1

• scale during sampling

  #19 opened 9 months ago by zhangdan8962
  3

• mixed precision training

  #18 opened a year ago by nicolas-dufour
  13

• Normalizing the context for cross-attention

  #17 opened a year ago by yoavhacohen
  2

• Inconsistent interpretation of model output between self-conditioning step and prediction step

  #15 opened a year ago by jsternabsci
  1

• Naming convention

  #13 opened a year ago by katop1234
  0

• Block-wise positional embedding

  #12 opened a year ago by katop1234
  0

• Min SNR weighting implemented as max SNR weighting

  #11 opened 2 years ago by justinlovelace
  4

• Handling of the time conditioning

  #9 opened 2 years ago by nicolas-dufour
  9

• notes

  #2 opened 2 years ago by ajabri
  39

• Potential Issue with Self-Conditioning

  #8 opened 2 years ago by justinlovelace
  2

• Possible Normalization Bug

  #6 opened 2 years ago by MattMcPartlon
  4

• Lacks a `ff` layer?

  #3 opened 2 years ago by CiaoHe
  1

• Could you share more thoughts on RIN?

  #1 opened 2 years ago by w-hc
  0