mehdidc/feed_forward_vqgan_clip

Issues

• How to get more variation in the null image

  #27 opened 2 years ago by kchodorow
  0

• Positional Stickiness

  #8 opened 3 years ago by afiaka87
  25

• How to condition model output z that looks like it can from a standard normal distribution?

  #20 opened 2 years ago by xiankgx
  2

• Repo License

  #24 opened 2 years ago by minimaxir
  1

• Models are broken in the new `torch` version

  #25 opened 2 years ago by neverix
  12

• Slow Training Speed

  #21 opened 3 years ago by s13kman
  3

• training GPU configuration

  #23 opened 3 years ago by CrossLee1
  1

• Error in Load Model

  #19 opened 2 years ago by metaphorz
  9

• New Checkpoint Idea

  #22 opened 3 years ago by afiaka87
  8

• Finetuing CLIP to improve domain-specific performance

  #13 opened 3 years ago by afiaka87
  1

• clarifying differences between available models

  #18 opened 3 years ago by zeke
  2

• Unavailable and broken links

  #15 opened 3 years ago by woctezuma
  7

• How to improve so we could get results closer to the "regular" VQGAN+CLIP?

  #14 opened 3 years ago by apolinario
  2

• CLIP-guided-diffusion updates

  #9 opened 3 years ago by afiaka87
  5

• VQGAN - blended models

  #12 opened 3 years ago by johndpope
  3

• Not an issue - richer datasets

  #6 opened 3 years ago by johndpope
  7

• Observations training with different modifying words/phrases

  #7 opened 3 years ago by afiaka87
  6

• Goal?

  #1 opened 3 years ago by afiaka87
  10

• New CLIP checkpoints from Open AI

  #4 opened 3 years ago by afiaka87
  1